sys/net/pf/pf.c

   1 /*      $OpenBSD: pf.c,v 1.552.2.1 2007/11/27 16:37:57 henning Exp $ */
   2 /* add $OpenBSD: pf.c,v 1.553 2007/08/23 11:15:49 dhartmei Exp $ */
   3 /* add $OpenBSD: pf.c,v 1.554 2007/08/28 16:09:12 henning Exp $ */
   4
   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 #include <sys/proc.h>
  60 #include <sys/kthread.h>
  61
  62 #include <sys/mplock2.h>
  63
  64 #include <machine/inttypes.h>
  65
  66 #include <net/if.h>
  67 #include <net/if_types.h>
  68 #include <net/bpf.h>
  69 #include <net/netisr.h>
  70 #include <net/route.h>
  71
  72 #include <netinet/in.h>
  73 #include <netinet/in_var.h>
  74 #include <netinet/in_systm.h>
  75 #include <netinet/ip.h>
  76 #include <netinet/ip_var.h>
  77 #include <netinet/tcp.h>
  78 #include <netinet/tcp_seq.h>
  79 #include <netinet/udp.h>
  80 #include <netinet/ip_icmp.h>
  81 #include <netinet/in_pcb.h>
  82 #include <netinet/tcp_timer.h>
  83 #include <netinet/tcp_var.h>
  84 #include <netinet/udp_var.h>
  85 #include <netinet/icmp_var.h>
  86 #include <netinet/if_ether.h>
  87
  88 #include <net/pf/pfvar.h>
  89 #include <net/pf/if_pflog.h>
  90
  91 #if NPFSYNC > 0
  92 #include <net/pf/if_pfsync.h>
  93 #endif /* NPFSYNC > 0 */
  94
  95 #ifdef INET6
  96 #include <netinet/ip6.h>
  97 #include <netinet/in_pcb.h>
  98 #include <netinet/icmp6.h>
  99 #include <netinet6/nd6.h>
 100 #include <netinet6/ip6_var.h>
 101 #include <netinet6/in6_pcb.h>
 102 #endif /* INET6 */
 103
 104 #include <sys/in_cksum.h>
 105 #include <sys/ucred.h>
 106 #include <machine/limits.h>
 107 #include <sys/msgport2.h>
 108 #include <net/netmsg2.h>
 109
 110 extern int ip_optcopy(struct ip *, struct ip *);
 111 extern int debug_pfugidhack;
 112
 113 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) kprintf x
 114
 115 /*
 116  * Global variables
 117  */
 118
 119 /* state tables */
 120 struct pf_state_tree_lan_ext     pf_statetbl_lan_ext;
 121 struct pf_state_tree_ext_gwy     pf_statetbl_ext_gwy;
 122
 123 struct pf_altqqueue      pf_altqs[2];
 124 struct pf_palist         pf_pabuf;
 125 struct pf_altqqueue     *pf_altqs_active;
 126 struct pf_altqqueue     *pf_altqs_inactive;
 127 struct pf_status         pf_status;
 128
 129 u_int32_t                ticket_altqs_active;
 130 u_int32_t                ticket_altqs_inactive;
 131 int                      altqs_inactive_open;
 132 u_int32_t                ticket_pabuf;
 133
 134 struct pf_anchor_stackframe {
 135         struct pf_ruleset                       *rs;
 136         struct pf_rule                          *r;
 137         struct pf_anchor_node                   *parent;
 138         struct pf_anchor                        *child;
 139 } pf_anchor_stack[64];
 140
 141 vm_zone_t                pf_src_tree_pl, pf_rule_pl, pf_pooladdr_pl;
 142 vm_zone_t                pf_state_pl, pf_state_key_pl;
 143 vm_zone_t                pf_altq_pl;
 144
 145 void                     pf_print_host(struct pf_addr *, u_int16_t, u_int8_t);
 146
 147 void                     pf_init_threshold(struct pf_threshold *, u_int32_t,
 148                             u_int32_t);
 149 void                     pf_add_threshold(struct pf_threshold *);
 150 int                      pf_check_threshold(struct pf_threshold *);
 151
 152 void                     pf_change_ap(struct pf_addr *, u_int16_t *,
 153                             u_int16_t *, u_int16_t *, struct pf_addr *,
 154                             u_int16_t, u_int8_t, sa_family_t);
 155 int                      pf_modulate_sack(struct mbuf *, int, struct pf_pdesc *,
 156                             struct tcphdr *, struct pf_state_peer *);
 157 #ifdef INET6
 158 void                     pf_change_a6(struct pf_addr *, u_int16_t *,
 159                             struct pf_addr *, u_int8_t);
 160 #endif /* INET6 */
 161 void                     pf_change_icmp(struct pf_addr *, u_int16_t *,
 162                             struct pf_addr *, struct pf_addr *, u_int16_t,
 163                             u_int16_t *, u_int16_t *, u_int16_t *,
 164                             u_int16_t *, u_int8_t, sa_family_t);
 165 void                     pf_send_tcp(const struct pf_rule *, sa_family_t,
 166                             const struct pf_addr *, const struct pf_addr *,
 167                             u_int16_t, u_int16_t, u_int32_t, u_int32_t,
 168                             u_int8_t, u_int16_t, u_int16_t, u_int8_t, int,
 169                             u_int16_t, struct ether_header *, struct ifnet *);
 170 void                     pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t,
 171                             sa_family_t, struct pf_rule *);
 172 struct pf_rule          *pf_match_translation(struct pf_pdesc *, struct mbuf *,
 173                             int, int, struct pfi_kif *,
 174                             struct pf_addr *, u_int16_t, struct pf_addr *,
 175                             u_int16_t, int);
 176 struct pf_rule          *pf_get_translation(struct pf_pdesc *, struct mbuf *,
 177                             int, int, struct pfi_kif *, struct pf_src_node **,
 178                             struct pf_addr *, u_int16_t,
 179                             struct pf_addr *, u_int16_t,
 180                             struct pf_addr *, u_int16_t *);
 181 void                     pf_attach_state(struct pf_state_key *,
 182                             struct pf_state *, int);
 183 void                     pf_detach_state(struct pf_state *, int);
 184 int                      pf_test_rule(struct pf_rule **, struct pf_state **,
 185                             int, struct pfi_kif *, struct mbuf *, int,
 186                             void *, struct pf_pdesc *, struct pf_rule **,
 187                             struct pf_ruleset **, struct ifqueue *, struct inpcb *);
 188 int                      pf_test_fragment(struct pf_rule **, int,
 189                             struct pfi_kif *, struct mbuf *, void *,
 190                             struct pf_pdesc *, struct pf_rule **,
 191                             struct pf_ruleset **);
 192 int                      pf_test_state_tcp(struct pf_state **, int,
 193                             struct pfi_kif *, struct mbuf *, int,
 194                             void *, struct pf_pdesc *, u_short *);
 195 int                      pf_test_state_udp(struct pf_state **, int,
 196                             struct pfi_kif *, struct mbuf *, int,
 197                             void *, struct pf_pdesc *);
 198 int                      pf_test_state_icmp(struct pf_state **, int,
 199                             struct pfi_kif *, struct mbuf *, int,
 200                             void *, struct pf_pdesc *, u_short *);
 201 int                      pf_test_state_other(struct pf_state **, int,
 202                             struct pfi_kif *, struct pf_pdesc *);
 203 int                      pf_match_tag(struct mbuf *, struct pf_rule *, int *);
 204 void                     pf_step_into_anchor(int *, struct pf_ruleset **, int,
 205                             struct pf_rule **, struct pf_rule **,  int *);
 206 int                      pf_step_out_of_anchor(int *, struct pf_ruleset **,
 207                              int, struct pf_rule **, struct pf_rule **,
 208                              int *);
 209 void                     pf_hash(struct pf_addr *, struct pf_addr *,
 210                             struct pf_poolhashkey *, sa_family_t);
 211 int                      pf_map_addr(u_int8_t, struct pf_rule *,
 212                             struct pf_addr *, struct pf_addr *,
 213                             struct pf_addr *, struct pf_src_node **);
 214 int                      pf_get_sport(sa_family_t, u_int8_t, struct pf_rule *,
 215                             struct pf_addr *, struct pf_addr *, u_int16_t,
 216                             struct pf_addr *, u_int16_t*, u_int16_t, u_int16_t,
 217                             struct pf_src_node **);
 218 void                     pf_route(struct mbuf **, struct pf_rule *, int,
 219                             struct ifnet *, struct pf_state *,
 220                             struct pf_pdesc *);
 221 void                     pf_route6(struct mbuf **, struct pf_rule *, int,
 222                             struct ifnet *, struct pf_state *,
 223                             struct pf_pdesc *);
 224 u_int8_t                 pf_get_wscale(struct mbuf *, int, u_int16_t,
 225                             sa_family_t);
 226 u_int16_t                pf_get_mss(struct mbuf *, int, u_int16_t,
 227                             sa_family_t);
 228 u_int16_t                pf_calc_mss(struct pf_addr *, sa_family_t,
 229                                 u_int16_t);
 230 void                     pf_set_rt_ifp(struct pf_state *,
 231                             struct pf_addr *);
 232 int                      pf_check_proto_cksum(struct mbuf *, int, int,
 233                             u_int8_t, sa_family_t);
 234 int                      pf_addr_wrap_neq(struct pf_addr_wrap *,
 235                             struct pf_addr_wrap *);
 236 struct pf_state         *pf_find_state(struct pfi_kif *,
 237                             struct pf_state_key_cmp *, u_int8_t);
 238 int                      pf_src_connlimit(struct pf_state **);
 239 void                     pf_stateins_err(const char *, struct pf_state *,
 240                             struct pfi_kif *);
 241 int                      pf_check_congestion(struct ifqueue *);
 242
 243 extern int pf_end_threads;
 244
 245 struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] = {
 246         { &pf_state_pl, PFSTATE_HIWAT },
 247         { &pf_src_tree_pl, PFSNODE_HIWAT },
 248         { &pf_frent_pl, PFFRAG_FRENT_HIWAT },
 249         { &pfr_ktable_pl, PFR_KTABLE_HIWAT },
 250         { &pfr_kentry_pl, PFR_KENTRY_HIWAT }
 251 };
 252
 253 #define STATE_LOOKUP()                                                  \
 254         do {                                                            \
 255                 if (direction == PF_IN)                                 \
 256                         *state = pf_find_state(kif, &key, PF_EXT_GWY);  \
 257                 else                                                    \
 258                         *state = pf_find_state(kif, &key, PF_LAN_EXT);  \
 259                 if (*state == NULL || (*state)->timeout == PFTM_PURGE)  \
 260                         return (PF_DROP);                               \
 261                 if (direction == PF_OUT &&                              \
 262                     (((*state)->rule.ptr->rt == PF_ROUTETO &&           \
 263                     (*state)->rule.ptr->direction == PF_OUT) ||         \
 264                     ((*state)->rule.ptr->rt == PF_REPLYTO &&            \
 265                     (*state)->rule.ptr->direction == PF_IN)) &&         \
 266                     (*state)->rt_kif != NULL &&                         \
 267                     (*state)->rt_kif != kif)                            \
 268                         return (PF_PASS);                               \
 269         } while (0)
 270
 271 #define STATE_TRANSLATE(sk) \
 272         (sk)->lan.addr.addr32[0] != (sk)->gwy.addr.addr32[0] || \
 273         ((sk)->af == AF_INET6 && \
 274         ((sk)->lan.addr.addr32[1] != (sk)->gwy.addr.addr32[1] || \
 275         (sk)->lan.addr.addr32[2] != (sk)->gwy.addr.addr32[2] || \
 276         (sk)->lan.addr.addr32[3] != (sk)->gwy.addr.addr32[3])) || \
 277         (sk)->lan.port != (sk)->gwy.port
 278
 279 #define BOUND_IFACE(r, k) \
 280         ((r)->rule_flag & PFRULE_IFBOUND) ? (k) : pfi_all
 281
 282 #define STATE_INC_COUNTERS(s)                           \
 283         do {                                            \
 284                 s->rule.ptr->states++;                  \
 285                 if (s->anchor.ptr != NULL)              \
 286                         s->anchor.ptr->states++;        \
 287                 if (s->nat_rule.ptr != NULL)            \
 288                         s->nat_rule.ptr->states++;      \
 289         } while (0)
 290
 291 #define STATE_DEC_COUNTERS(s)                           \
 292         do {                                            \
 293                 if (s->nat_rule.ptr != NULL)            \
 294                         s->nat_rule.ptr->states--;      \
 295                 if (s->anchor.ptr != NULL)              \
 296                         s->anchor.ptr->states--;        \
 297                 s->rule.ptr->states--;                  \
 298         } while (0)
 299
 300 static __inline int pf_src_compare(struct pf_src_node *, struct pf_src_node *);
 301 static __inline int pf_state_compare_lan_ext(struct pf_state_key *,
 302         struct pf_state_key *);
 303 static __inline int pf_state_compare_ext_gwy(struct pf_state_key *,
 304         struct pf_state_key *);
 305 static __inline int pf_state_compare_id(struct pf_state *,
 306         struct pf_state *);
 307
 308 struct pf_src_tree tree_src_tracking;
 309
 310 struct pf_state_tree_id tree_id;
 311 struct pf_state_queue state_list;
 312
 313 RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare);
 314 RB_GENERATE(pf_state_tree_lan_ext, pf_state_key,
 315     entry_lan_ext, pf_state_compare_lan_ext);
 316 RB_GENERATE(pf_state_tree_ext_gwy, pf_state_key,
 317     entry_ext_gwy, pf_state_compare_ext_gwy);
 318 RB_GENERATE(pf_state_tree_id, pf_state,
 319     entry_id, pf_state_compare_id);
 320
 321 #define PF_DT_SKIP_LANEXT       0x01
 322 #define PF_DT_SKIP_EXTGWY       0x02
 323
 324 static __inline int
 325 pf_src_compare(struct pf_src_node *a, struct pf_src_node *b)
 326 {
 327         int     diff;
 328
 329         if (a->rule.ptr > b->rule.ptr)
 330                 return (1);
 331         if (a->rule.ptr < b->rule.ptr)
 332                 return (-1);
 333         if ((diff = a->af - b->af) != 0)
 334                 return (diff);
 335         switch (a->af) {
 336 #ifdef INET
 337         case AF_INET:
 338                 if (a->addr.addr32[0] > b->addr.addr32[0])
 339                         return (1);
 340                 if (a->addr.addr32[0] < b->addr.addr32[0])
 341                         return (-1);
 342                 break;
 343 #endif /* INET */
 344 #ifdef INET6
 345         case AF_INET6:
 346                 if (a->addr.addr32[3] > b->addr.addr32[3])
 347                         return (1);
 348                 if (a->addr.addr32[3] < b->addr.addr32[3])
 349                         return (-1);
 350                 if (a->addr.addr32[2] > b->addr.addr32[2])
 351                         return (1);
 352                 if (a->addr.addr32[2] < b->addr.addr32[2])
 353                         return (-1);
 354                 if (a->addr.addr32[1] > b->addr.addr32[1])
 355                         return (1);
 356                 if (a->addr.addr32[1] < b->addr.addr32[1])
 357                         return (-1);
 358                 if (a->addr.addr32[0] > b->addr.addr32[0])
 359                         return (1);
 360                 if (a->addr.addr32[0] < b->addr.addr32[0])
 361                         return (-1);
 362                 break;
 363 #endif /* INET6 */
 364         }
 365         return (0);
 366 }
 367
 368 u_int32_t
 369 pf_state_hash(struct pf_state_key *sk)
 370 {
 371         u_int32_t hv = (intptr_t)sk / sizeof(*sk);
 372
 373         hv ^= crc32(&sk->lan, sizeof(sk->lan));
 374         hv ^= crc32(&sk->gwy, sizeof(sk->gwy));
 375         hv ^= crc32(&sk->ext, sizeof(sk->ext));
 376         if (hv == 0)    /* disallow 0 */
 377                 hv = 1;
 378         return(hv);
 379 }
 380
 381 static __inline int
 382 pf_state_compare_lan_ext(struct pf_state_key *a, struct pf_state_key *b)
 383 {
 384         int     diff;
 385
 386         if ((diff = a->proto - b->proto) != 0)
 387                 return (diff);
 388         if ((diff = a->af - b->af) != 0)
 389                 return (diff);
 390         switch (a->af) {
 391 #ifdef INET
 392         case AF_INET:
 393                 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
 394                         return (1);
 395                 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
 396                         return (-1);
 397                 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
 398                         return (1);
 399                 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
 400                         return (-1);
 401                 break;
 402 #endif /* INET */
 403 #ifdef INET6
 404         case AF_INET6:
 405                 if (a->lan.addr.addr32[3] > b->lan.addr.addr32[3])
 406                         return (1);
 407                 if (a->lan.addr.addr32[3] < b->lan.addr.addr32[3])
 408                         return (-1);
 409                 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
 410                         return (1);
 411                 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
 412                         return (-1);
 413                 if (a->lan.addr.addr32[2] > b->lan.addr.addr32[2])
 414                         return (1);
 415                 if (a->lan.addr.addr32[2] < b->lan.addr.addr32[2])
 416                         return (-1);
 417                 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
 418                         return (1);
 419                 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
 420                         return (-1);
 421                 if (a->lan.addr.addr32[1] > b->lan.addr.addr32[1])
 422                         return (1);
 423                 if (a->lan.addr.addr32[1] < b->lan.addr.addr32[1])
 424                         return (-1);
 425                 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
 426                         return (1);
 427                 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
 428                         return (-1);
 429                 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
 430                         return (1);
 431                 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
 432                         return (-1);
 433                 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
 434                         return (1);
 435                 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
 436                         return (-1);
 437                 break;
 438 #endif /* INET6 */
 439         }
 440
 441         if ((diff = a->lan.port - b->lan.port) != 0)
 442                 return (diff);
 443         if ((diff = a->ext.port - b->ext.port) != 0)
 444                 return (diff);
 445
 446         return (0);
 447 }
 448
 449 static __inline int
 450 pf_state_compare_ext_gwy(struct pf_state_key *a, struct pf_state_key *b)
 451 {
 452         int     diff;
 453
 454         if ((diff = a->proto - b->proto) != 0)
 455                 return (diff);
 456         if ((diff = a->af - b->af) != 0)
 457                 return (diff);
 458         switch (a->af) {
 459 #ifdef INET
 460         case AF_INET:
 461                 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
 462                         return (1);
 463                 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
 464                         return (-1);
 465                 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
 466                         return (1);
 467                 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
 468                         return (-1);
 469                 break;
 470 #endif /* INET */
 471 #ifdef INET6
 472         case AF_INET6:
 473                 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
 474                         return (1);
 475                 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
 476                         return (-1);
 477                 if (a->gwy.addr.addr32[3] > b->gwy.addr.addr32[3])
 478                         return (1);
 479                 if (a->gwy.addr.addr32[3] < b->gwy.addr.addr32[3])
 480                         return (-1);
 481                 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
 482                         return (1);
 483                 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
 484                         return (-1);
 485                 if (a->gwy.addr.addr32[2] > b->gwy.addr.addr32[2])
 486                         return (1);
 487                 if (a->gwy.addr.addr32[2] < b->gwy.addr.addr32[2])
 488                         return (-1);
 489                 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
 490                         return (1);
 491                 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
 492                         return (-1);
 493                 if (a->gwy.addr.addr32[1] > b->gwy.addr.addr32[1])
 494                         return (1);
 495                 if (a->gwy.addr.addr32[1] < b->gwy.addr.addr32[1])
 496                         return (-1);
 497                 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
 498                         return (1);
 499                 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
 500                         return (-1);
 501                 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
 502                         return (1);
 503                 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
 504                         return (-1);
 505                 break;
 506 #endif /* INET6 */
 507         }
 508
 509         if ((diff = a->ext.port - b->ext.port) != 0)
 510                 return (diff);
 511         if ((diff = a->gwy.port - b->gwy.port) != 0)
 512                 return (diff);
 513
 514         return (0);
 515 }
 516
 517 static __inline int
 518 pf_state_compare_id(struct pf_state *a, struct pf_state *b)
 519 {
 520         if (a->id > b->id)
 521                 return (1);
 522         if (a->id < b->id)
 523                 return (-1);
 524         if (a->creatorid > b->creatorid)
 525                 return (1);
 526         if (a->creatorid < b->creatorid)
 527                 return (-1);
 528
 529         return (0);
 530 }
 531
 532 #ifdef INET6
 533 void
 534 pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af)
 535 {
 536         switch (af) {
 537 #ifdef INET
 538         case AF_INET:
 539                 dst->addr32[0] = src->addr32[0];
 540                 break;
 541 #endif /* INET */
 542         case AF_INET6:
 543                 dst->addr32[0] = src->addr32[0];
 544                 dst->addr32[1] = src->addr32[1];
 545                 dst->addr32[2] = src->addr32[2];
 546                 dst->addr32[3] = src->addr32[3];
 547                 break;
 548         }
 549 }
 550 #endif /* INET6 */
 551
 552 struct pf_state *
 553 pf_find_state_byid(struct pf_state_cmp *key)
 554 {
 555         pf_status.fcounters[FCNT_STATE_SEARCH]++;
 556
 557         return (RB_FIND(pf_state_tree_id, &tree_id, (struct pf_state *)key));
 558 }
 559
 560 struct pf_state *
 561 pf_find_state(struct pfi_kif *kif, struct pf_state_key_cmp *key, u_int8_t tree)
 562 {
 563         struct pf_state_key     *sk;
 564         struct pf_state         *s;
 565
 566         pf_status.fcounters[FCNT_STATE_SEARCH]++;
 567
 568         switch (tree) {
 569         case PF_LAN_EXT:
 570                 sk = RB_FIND(pf_state_tree_lan_ext, &pf_statetbl_lan_ext,
 571                     (struct pf_state_key *)key);
 572                 break;
 573         case PF_EXT_GWY:
 574                 sk = RB_FIND(pf_state_tree_ext_gwy, &pf_statetbl_ext_gwy,
 575                     (struct pf_state_key *)key);
 576                 break;
 577         default:
 578                 panic("pf_find_state");
 579         }
 580
 581         /* list is sorted, if-bound states before floating ones */
 582         if (sk != NULL)
 583                 TAILQ_FOREACH(s, &sk->states, next)
 584                         if (s->kif == pfi_all || s->kif == kif)
 585                                 return (s);
 586
 587         return (NULL);
 588 }
 589
 590 struct pf_state *
 591 pf_find_state_all(struct pf_state_key_cmp *key, u_int8_t tree, int *more)
 592 {
 593         struct pf_state_key     *sk;
 594         struct pf_state         *s, *ret = NULL;
 595
 596         pf_status.fcounters[FCNT_STATE_SEARCH]++;
 597
 598         switch (tree) {
 599         case PF_LAN_EXT:
 600                 sk = RB_FIND(pf_state_tree_lan_ext,
 601                     &pf_statetbl_lan_ext, (struct pf_state_key *)key);
 602                 break;
 603         case PF_EXT_GWY:
 604                 sk = RB_FIND(pf_state_tree_ext_gwy,
 605                     &pf_statetbl_ext_gwy, (struct pf_state_key *)key);
 606                 break;
 607         default:
 608                 panic("pf_find_state_all");
 609         }
 610
 611         if (sk != NULL) {
 612                 ret = TAILQ_FIRST(&sk->states);
 613                 if (more == NULL)
 614                         return (ret);
 615
 616                 TAILQ_FOREACH(s, &sk->states, next)
 617                         (*more)++;
 618         }
 619
 620         return (ret);
 621 }
 622
 623 void
 624 pf_init_threshold(struct pf_threshold *threshold,
 625     u_int32_t limit, u_int32_t seconds)
 626 {
 627         threshold->limit = limit * PF_THRESHOLD_MULT;
 628         threshold->seconds = seconds;
 629         threshold->count = 0;
 630         threshold->last = time_second;
 631 }
 632
 633 void
 634 pf_add_threshold(struct pf_threshold *threshold)
 635 {
 636         u_int32_t t = time_second, diff = t - threshold->last;
 637
 638         if (diff >= threshold->seconds)
 639                 threshold->count = 0;
 640         else
 641                 threshold->count -= threshold->count * diff /
 642                     threshold->seconds;
 643         threshold->count += PF_THRESHOLD_MULT;
 644         threshold->last = t;
 645 }
 646
 647 int
 648 pf_check_threshold(struct pf_threshold *threshold)
 649 {
 650         return (threshold->count > threshold->limit);
 651 }
 652
 653 int
 654 pf_src_connlimit(struct pf_state **state)
 655 {
 656         int bad = 0;
 657
 658         (*state)->src_node->conn++;
 659         (*state)->src.tcp_est = 1;
 660         pf_add_threshold(&(*state)->src_node->conn_rate);
 661
 662         if ((*state)->rule.ptr->max_src_conn &&
 663             (*state)->rule.ptr->max_src_conn <
 664             (*state)->src_node->conn) {
 665                 pf_status.lcounters[LCNT_SRCCONN]++;
 666                 bad++;
 667         }
 668
 669         if ((*state)->rule.ptr->max_src_conn_rate.limit &&
 670             pf_check_threshold(&(*state)->src_node->conn_rate)) {
 671                 pf_status.lcounters[LCNT_SRCCONNRATE]++;
 672                 bad++;
 673         }
 674
 675         if (!bad)
 676                 return (0);
 677
 678         if ((*state)->rule.ptr->overload_tbl) {
 679                 struct pfr_addr p;
 680                 u_int32_t       killed = 0;
 681
 682                 pf_status.lcounters[LCNT_OVERLOAD_TABLE]++;
 683                 if (pf_status.debug >= PF_DEBUG_MISC) {
 684                         kprintf("pf_src_connlimit: blocking address ");
 685                         pf_print_host(&(*state)->src_node->addr, 0,
 686                             (*state)->state_key->af);
 687                 }
 688
 689                 bzero(&p, sizeof(p));
 690                 p.pfra_af = (*state)->state_key->af;
 691                 switch ((*state)->state_key->af) {
 692 #ifdef INET
 693                 case AF_INET:
 694                         p.pfra_net = 32;
 695                         p.pfra_ip4addr = (*state)->src_node->addr.v4;
 696                         break;
 697 #endif /* INET */
 698 #ifdef INET6
 699                 case AF_INET6:
 700                         p.pfra_net = 128;
 701                         p.pfra_ip6addr = (*state)->src_node->addr.v6;
 702                         break;
 703 #endif /* INET6 */
 704                 }
 705
 706                 pfr_insert_kentry((*state)->rule.ptr->overload_tbl,
 707                     &p, time_second);
 708
 709                 /* kill existing states if that's required. */
 710                 if ((*state)->rule.ptr->flush) {
 711                         struct pf_state_key *sk;
 712                         struct pf_state *st;
 713
 714                         pf_status.lcounters[LCNT_OVERLOAD_FLUSH]++;
 715                         RB_FOREACH(st, pf_state_tree_id, &tree_id) {
 716                                 sk = st->state_key;
 717                                 /*
 718                                  * Kill states from this source.  (Only those
 719                                  * from the same rule if PF_FLUSH_GLOBAL is not
 720                                  * set)
 721                                  */
 722                                 if (sk->af ==
 723                                     (*state)->state_key->af &&
 724                                     (((*state)->state_key->direction ==
 725                                         PF_OUT &&
 726                                     PF_AEQ(&(*state)->src_node->addr,
 727                                         &sk->lan.addr, sk->af)) ||
 728                                     ((*state)->state_key->direction == PF_IN &&
 729                                     PF_AEQ(&(*state)->src_node->addr,
 730                                         &sk->ext.addr, sk->af))) &&
 731                                     ((*state)->rule.ptr->flush &
 732                                     PF_FLUSH_GLOBAL ||
 733                                     (*state)->rule.ptr == st->rule.ptr)) {
 734                                         st->timeout = PFTM_PURGE;
 735                                         st->src.state = st->dst.state =
 736                                             TCPS_CLOSED;
 737                                         killed++;
 738                                 }
 739                         }
 740                         if (pf_status.debug >= PF_DEBUG_MISC)
 741                                 kprintf(", %u states killed", killed);
 742                 }
 743                 if (pf_status.debug >= PF_DEBUG_MISC)
 744                         kprintf("\n");
 745         }
 746
 747         /* kill this state */
 748         (*state)->timeout = PFTM_PURGE;
 749         (*state)->src.state = (*state)->dst.state = TCPS_CLOSED;
 750         return (1);
 751 }
 752
 753 int
 754 pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule,
 755     struct pf_addr *src, sa_family_t af)
 756 {
 757         struct pf_src_node      k;
 758
 759         if (*sn == NULL) {
 760                 k.af = af;
 761                 PF_ACPY(&k.addr, src, af);
 762                 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
 763                     rule->rpool.opts & PF_POOL_STICKYADDR)
 764                         k.rule.ptr = rule;
 765                 else
 766                         k.rule.ptr = NULL;
 767                 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
 768                 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
 769         }
 770         if (*sn == NULL) {
 771                 if (!rule->max_src_nodes ||
 772                     rule->src_nodes < rule->max_src_nodes)
 773                         (*sn) = pool_get(&pf_src_tree_pl, PR_NOWAIT);
 774                 else
 775                         pf_status.lcounters[LCNT_SRCNODES]++;
 776                 if ((*sn) == NULL)
 777                         return (-1);
 778                 bzero(*sn, sizeof(struct pf_src_node));
 779
 780                 pf_init_threshold(&(*sn)->conn_rate,
 781                     rule->max_src_conn_rate.limit,
 782                     rule->max_src_conn_rate.seconds);
 783
 784                 (*sn)->af = af;
 785                 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
 786                     rule->rpool.opts & PF_POOL_STICKYADDR)
 787                         (*sn)->rule.ptr = rule;
 788                 else
 789                         (*sn)->rule.ptr = NULL;
 790                 PF_ACPY(&(*sn)->addr, src, af);
 791                 if (RB_INSERT(pf_src_tree,
 792                     &tree_src_tracking, *sn) != NULL) {
 793                         if (pf_status.debug >= PF_DEBUG_MISC) {
 794                                 kprintf("pf: src_tree insert failed: ");
 795                                 pf_print_host(&(*sn)->addr, 0, af);
 796                                 kprintf("\n");
 797                         }
 798                         pool_put(&pf_src_tree_pl, *sn);
 799                         return (-1);
 800                 }
 801                 (*sn)->creation = time_second;
 802                 (*sn)->ruletype = rule->action;
 803                 if ((*sn)->rule.ptr != NULL)
 804                         (*sn)->rule.ptr->src_nodes++;
 805                 pf_status.scounters[SCNT_SRC_NODE_INSERT]++;
 806                 pf_status.src_nodes++;
 807         } else {
 808                 if (rule->max_src_states &&
 809                     (*sn)->states >= rule->max_src_states) {
 810                         pf_status.lcounters[LCNT_SRCSTATES]++;
 811                         return (-1);
 812                 }
 813         }
 814         return (0);
 815 }
 816
 817 void
 818 pf_stateins_err(const char *tree, struct pf_state *s, struct pfi_kif *kif)
 819 {
 820         struct pf_state_key     *sk = s->state_key;
 821
 822         if (pf_status.debug >= PF_DEBUG_MISC) {
 823                 kprintf("pf: state insert failed: %s %s", tree, kif->pfik_name);
 824                 kprintf(" lan: ");
 825                 pf_print_host(&sk->lan.addr, sk->lan.port,
 826                     sk->af);
 827                 kprintf(" gwy: ");
 828                 pf_print_host(&sk->gwy.addr, sk->gwy.port,
 829                     sk->af);
 830                 kprintf(" ext: ");
 831                 pf_print_host(&sk->ext.addr, sk->ext.port,
 832                     sk->af);
 833                 if (s->sync_flags & PFSTATE_FROMSYNC)
 834                         kprintf(" (from sync)");
 835                 kprintf("\n");
 836         }
 837 }
 838
 839 int
 840 pf_insert_state(struct pfi_kif *kif, struct pf_state *s)
 841 {
 842         struct pf_state_key     *cur;
 843         struct pf_state         *sp;
 844
 845         KKASSERT(s->state_key != NULL);
 846         s->kif = kif;
 847
 848         if ((cur = RB_INSERT(pf_state_tree_lan_ext, &pf_statetbl_lan_ext,
 849             s->state_key)) != NULL) {
 850                 /* key exists. check for same kif, if none, add to key */
 851                 TAILQ_FOREACH(sp, &cur->states, next)
 852                         if (sp->kif == kif) {   /* collision! */
 853                                 pf_stateins_err("tree_lan_ext", s, kif);
 854                                 pf_detach_state(s,
 855                                     PF_DT_SKIP_LANEXT|PF_DT_SKIP_EXTGWY);
 856                                 return (-1);
 857                         }
 858                 pf_detach_state(s, PF_DT_SKIP_LANEXT|PF_DT_SKIP_EXTGWY);
 859                 pf_attach_state(cur, s, kif == pfi_all ? 1 : 0);
 860         }
 861
 862         /* if cur != NULL, we already found a state key and attached to it */
 863         if (cur == NULL && (cur = RB_INSERT(pf_state_tree_ext_gwy,
 864             &pf_statetbl_ext_gwy, s->state_key)) != NULL) {
 865                 /* must not happen. we must have found the sk above! */
 866                 pf_stateins_err("tree_ext_gwy", s, kif);
 867                 pf_detach_state(s, PF_DT_SKIP_EXTGWY);
 868                 return (-1);
 869         }
 870
 871         if (s->id == 0 && s->creatorid == 0) {
 872                 s->id = htobe64(pf_status.stateid++);
 873                 s->creatorid = pf_status.hostid;
 874         }
 875         if (RB_INSERT(pf_state_tree_id, &tree_id, s) != NULL) {
 876                 if (pf_status.debug >= PF_DEBUG_MISC) {
 877                         kprintf("pf: state insert failed: "
 878                             "id: %016llx creatorid: %08x",
 879                             be64toh(s->id), ntohl(s->creatorid));
 880                         if (s->sync_flags & PFSTATE_FROMSYNC)
 881                                 kprintf(" (from sync)");
 882                         kprintf("\n");
 883                 }
 884                 pf_detach_state(s, 0);
 885                 return (-1);
 886         }
 887         TAILQ_INSERT_TAIL(&state_list, s, entry_list);
 888         pf_status.fcounters[FCNT_STATE_INSERT]++;
 889         pf_status.states++;
 890         pfi_kif_ref(kif, PFI_KIF_REF_STATE);
 891 #if NPFSYNC
 892         pfsync_insert_state(s);
 893 #endif
 894         return (0);
 895 }
 896
 897 void
 898 pf_purge_thread(void *v)
 899 {
 900         int nloops = 0;
 901         int locked = 0;
 902
 903         get_mplock();
 904         for (;;) {
 905                 tsleep(pf_purge_thread, PWAIT, "pftm", 1 * hz);
 906
 907                 lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
 908
 909                 if (pf_end_threads) {
 910                         pf_purge_expired_states(pf_status.states, 1);
 911                         pf_purge_expired_fragments();
 912                         pf_purge_expired_src_nodes(1);
 913                         pf_end_threads++;
 914
 915                         lockmgr(&pf_consistency_lock, LK_RELEASE);
 916                         wakeup(pf_purge_thread);
 917                         kthread_exit();
 918                 }
 919                 crit_enter();
 920
 921                 /* process a fraction of the state table every second */
 922                 if(!pf_purge_expired_states(1 + (pf_status.states
 923                     / pf_default_rule.timeout[PFTM_INTERVAL]), 0)) {
 924
 925                         pf_purge_expired_states(1 + (pf_status.states
 926                             / pf_default_rule.timeout[PFTM_INTERVAL]), 1);
 927                 }
 928
 929                 /* purge other expired types every PFTM_INTERVAL seconds */
 930                 if (++nloops >= pf_default_rule.timeout[PFTM_INTERVAL]) {
 931                         pf_purge_expired_fragments();
 932                         if (!pf_purge_expired_src_nodes(locked)) {
 933                                 pf_purge_expired_src_nodes(1);
 934                         }
 935                         nloops = 0;
 936                 }
 937                 crit_exit();
 938                 lockmgr(&pf_consistency_lock, LK_RELEASE);
 939         }
 940         rel_mplock();
 941 }
 942
 943 u_int32_t
 944 pf_state_expires(const struct pf_state *state)
 945 {
 946         u_int32_t       timeout;
 947         u_int32_t       start;
 948         u_int32_t       end;
 949         u_int32_t       states;
 950
 951         /* handle all PFTM_* > PFTM_MAX here */
 952         if (state->timeout == PFTM_PURGE)
 953                 return (time_second);
 954         if (state->timeout == PFTM_UNTIL_PACKET)
 955                 return (0);
 956         KKASSERT(state->timeout != PFTM_UNLINKED);
 957         KASSERT((state->timeout < PFTM_MAX),
 958                 ("pf_state_expires: timeout > PFTM_MAX"));
 959         timeout = state->rule.ptr->timeout[state->timeout];
 960         if (!timeout)
 961                 timeout = pf_default_rule.timeout[state->timeout];
 962         start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START];
 963         if (start) {
 964                 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END];
 965                 states = state->rule.ptr->states;
 966         } else {
 967                 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START];
 968                 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END];
 969                 states = pf_status.states;
 970         }
 971         if (end && states > start && start < end) {
 972                 if (states < end)
 973                         return (state->expire + timeout * (end - states) /
 974                             (end - start));
 975                 else
 976                         return (time_second);
 977         }
 978         return (state->expire + timeout);
 979 }
 980
 981 int
 982 pf_purge_expired_src_nodes(int waslocked)
 983 {
 984          struct pf_src_node             *cur, *next;
 985          int                             locked = waslocked;
 986
 987          for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) {
 988                  next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur);
 989
 990                  if (cur->states <= 0 && cur->expire <= time_second) {
 991                          if (! locked) {
 992                                  lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
 993                                  next = RB_NEXT(pf_src_tree,
 994                                      &tree_src_tracking, cur);
 995                                  locked = 1;
 996                          }
 997                          if (cur->rule.ptr != NULL) {
 998                                  cur->rule.ptr->src_nodes--;
 999                                  if (cur->rule.ptr->states <= 0 &&
1000                                      cur->rule.ptr->max_src_nodes <= 0)
1001                                          pf_rm_rule(NULL, cur->rule.ptr);
1002                          }
1003                          RB_REMOVE(pf_src_tree, &tree_src_tracking, cur);
1004                          pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
1005                          pf_status.src_nodes--;
1006                          pool_put(&pf_src_tree_pl, cur);
1007                  }
1008          }
1009
1010          if (locked && !waslocked)
1011                 lockmgr(&pf_consistency_lock, LK_RELEASE);
1012         return(1);
1013 }
1014
1015 void
1016 pf_src_tree_remove_state(struct pf_state *s)
1017 {
1018         u_int32_t timeout;
1019
1020         if (s->src_node != NULL) {
1021                 if (s->src.tcp_est)
1022                         --s->src_node->conn;
1023                 if (--s->src_node->states <= 0) {
1024                         timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1025                         if (!timeout)
1026                                 timeout =
1027                                     pf_default_rule.timeout[PFTM_SRC_NODE];
1028                         s->src_node->expire = time_second + timeout;
1029                 }
1030         }
1031         if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) {
1032                 if (--s->nat_src_node->states <= 0) {
1033                         timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1034                         if (!timeout)
1035                                 timeout =
1036                                     pf_default_rule.timeout[PFTM_SRC_NODE];
1037                         s->nat_src_node->expire = time_second + timeout;
1038                 }
1039         }
1040         s->src_node = s->nat_src_node = NULL;
1041 }
1042
1043 /* callers should be at crit_enter() */
1044 void
1045 pf_unlink_state(struct pf_state *cur)
1046 {
1047         if (cur->src.state == PF_TCPS_PROXY_DST) {
1048                 pf_send_tcp(cur->rule.ptr, cur->state_key->af,
1049                     &cur->state_key->ext.addr, &cur->state_key->lan.addr,
1050                     cur->state_key->ext.port, cur->state_key->lan.port,
1051                     cur->src.seqhi, cur->src.seqlo + 1,
1052                     TH_RST|TH_ACK, 0, 0, 0, 1, cur->tag, NULL, NULL);
1053         }
1054         RB_REMOVE(pf_state_tree_id, &tree_id, cur);
1055 #if NPFSYNC
1056         if (cur->creatorid == pf_status.hostid)
1057                 pfsync_delete_state(cur);
1058 #endif
1059         cur->timeout = PFTM_UNLINKED;
1060         pf_src_tree_remove_state(cur);
1061         pf_detach_state(cur, 0);
1062 }
1063
1064 /* callers should be at crit_enter() and hold the
1065  * write_lock on pf_consistency_lock */
1066 void
1067 pf_free_state(struct pf_state *cur)
1068 {
1069 #if NPFSYNC
1070         if (pfsyncif != NULL &&
1071             (pfsyncif->sc_bulk_send_next == cur ||
1072             pfsyncif->sc_bulk_terminator == cur))
1073                 return;
1074 #endif
1075         KKASSERT(cur->timeout == PFTM_UNLINKED);
1076         if (--cur->rule.ptr->states <= 0 &&
1077             cur->rule.ptr->src_nodes <= 0)
1078                 pf_rm_rule(NULL, cur->rule.ptr);
1079         if (cur->nat_rule.ptr != NULL)
1080                 if (--cur->nat_rule.ptr->states <= 0 &&
1081                         cur->nat_rule.ptr->src_nodes <= 0)
1082                         pf_rm_rule(NULL, cur->nat_rule.ptr);
1083         if (cur->anchor.ptr != NULL)
1084                 if (--cur->anchor.ptr->states <= 0)
1085                         pf_rm_rule(NULL, cur->anchor.ptr);
1086         pf_normalize_tcp_cleanup(cur);
1087         pfi_kif_unref(cur->kif, PFI_KIF_REF_STATE);
1088         TAILQ_REMOVE(&state_list, cur, entry_list);
1089         if (cur->tag)
1090                 pf_tag_unref(cur->tag);
1091         pool_put(&pf_state_pl, cur);
1092         pf_status.fcounters[FCNT_STATE_REMOVALS]++;
1093         pf_status.states--;
1094 }
1095
1096 int
1097 pf_purge_expired_states(u_int32_t maxcheck, int waslocked)
1098 {
1099         static struct pf_state  *cur = NULL;
1100         struct pf_state         *next;
1101         int                      locked = waslocked;
1102
1103         while (maxcheck--) {
1104                 /* wrap to start of list when we hit the end */
1105                 if (cur == NULL) {
1106                         cur = TAILQ_FIRST(&state_list);
1107                         if (cur == NULL)
1108                                 break;  /* list empty */
1109                 }
1110
1111                 /* get next state, as cur may get deleted */
1112                 next = TAILQ_NEXT(cur, entry_list);
1113
1114                 if (cur->timeout == PFTM_UNLINKED) {
1115                         /* free unlinked state */
1116                         if (! locked) {
1117                                 lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
1118                                 locked = 1;
1119                         }
1120                         pf_free_state(cur);
1121                 } else if (pf_state_expires(cur) <= time_second) {
1122                         /* unlink and free expired state */
1123                         pf_unlink_state(cur);
1124                         if (! locked) {
1125                                 if (!lockmgr(&pf_consistency_lock, LK_EXCLUSIVE))
1126                                         return (0);
1127                                 locked = 1;
1128                         }
1129                         pf_free_state(cur);
1130                 }
1131                 cur = next;
1132         }
1133
1134         if (locked)
1135                 lockmgr(&pf_consistency_lock, LK_RELEASE);
1136         return (1);
1137 }
1138
1139 int
1140 pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw)
1141 {
1142         if (aw->type != PF_ADDR_TABLE)
1143                 return (0);
1144         if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL)
1145                 return (1);
1146         return (0);
1147 }
1148
1149 void
1150 pf_tbladdr_remove(struct pf_addr_wrap *aw)
1151 {
1152         if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL)
1153                 return;
1154         pfr_detach_table(aw->p.tbl);
1155         aw->p.tbl = NULL;
1156 }
1157
1158 void
1159 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
1160 {
1161         struct pfr_ktable *kt = aw->p.tbl;
1162
1163         if (aw->type != PF_ADDR_TABLE || kt == NULL)
1164                 return;
1165         if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
1166                 kt = kt->pfrkt_root;
1167         aw->p.tbl = NULL;
1168         aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
1169                 kt->pfrkt_cnt : -1;
1170 }
1171
1172 void
1173 pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af)
1174 {
1175         switch (af) {
1176 #ifdef INET
1177         case AF_INET: {
1178                 u_int32_t a = ntohl(addr->addr32[0]);
1179                 kprintf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255,
1180                     (a>>8)&255, a&255);
1181                 if (p) {
1182                         p = ntohs(p);
1183                         kprintf(":%u", p);
1184                 }
1185                 break;
1186         }
1187 #endif /* INET */
1188 #ifdef INET6
1189         case AF_INET6: {
1190                 u_int16_t b;
1191                 u_int8_t i, curstart = 255, curend = 0,
1192                     maxstart = 0, maxend = 0;
1193                 for (i = 0; i < 8; i++) {
1194                         if (!addr->addr16[i]) {
1195                                 if (curstart == 255)
1196                                         curstart = i;
1197                                 else
1198                                         curend = i;
1199                         } else {
1200                                 if (curstart) {
1201                                         if ((curend - curstart) >
1202                                             (maxend - maxstart)) {
1203                                                 maxstart = curstart;
1204                                                 maxend = curend;
1205                                                 curstart = 255;
1206                                         }
1207                                 }
1208                         }
1209                 }
1210                 for (i = 0; i < 8; i++) {
1211                         if (i >= maxstart && i <= maxend) {
1212                                 if (maxend != 7) {
1213                                         if (i == maxstart)
1214                                                 kprintf(":");
1215                                 } else {
1216                                         if (i == maxend)
1217                                                 kprintf(":");
1218                                 }
1219                         } else {
1220                                 b = ntohs(addr->addr16[i]);
1221                                 kprintf("%x", b);
1222                                 if (i < 7)
1223                                         kprintf(":");
1224                         }
1225                 }
1226                 if (p) {
1227                         p = ntohs(p);
1228                         kprintf("[%u]", p);
1229                 }
1230                 break;
1231         }
1232 #endif /* INET6 */
1233         }
1234 }
1235
1236 void
1237 pf_print_state(struct pf_state *s)
1238 {
1239         struct pf_state_key *sk = s->state_key;
1240         switch (sk->proto) {
1241         case IPPROTO_TCP:
1242                 kprintf("TCP ");
1243                 break;
1244         case IPPROTO_UDP:
1245                 kprintf("UDP ");
1246                 break;
1247         case IPPROTO_ICMP:
1248                 kprintf("ICMP ");
1249                 break;
1250         case IPPROTO_ICMPV6:
1251                 kprintf("ICMPV6 ");
1252                 break;
1253         default:
1254                 kprintf("%u ", sk->proto);
1255                 break;
1256         }
1257         pf_print_host(&sk->lan.addr, sk->lan.port, sk->af);
1258         kprintf(" ");
1259         pf_print_host(&sk->gwy.addr, sk->gwy.port, sk->af);
1260         kprintf(" ");
1261         pf_print_host(&sk->ext.addr, sk->ext.port, sk->af);
1262         kprintf(" [lo=%u high=%u win=%u modulator=%u", s->src.seqlo,
1263             s->src.seqhi, s->src.max_win, s->src.seqdiff);
1264         if (s->src.wscale && s->dst.wscale)
1265                 kprintf(" wscale=%u", s->src.wscale & PF_WSCALE_MASK);
1266         kprintf("]");
1267         kprintf(" [lo=%u high=%u win=%u modulator=%u", s->dst.seqlo,
1268             s->dst.seqhi, s->dst.max_win, s->dst.seqdiff);
1269         if (s->src.wscale && s->dst.wscale)
1270                 kprintf(" wscale=%u", s->dst.wscale & PF_WSCALE_MASK);
1271         kprintf("]");
1272         kprintf(" %u:%u", s->src.state, s->dst.state);
1273 }
1274
1275 void
1276 pf_print_flags(u_int8_t f)
1277 {
1278         if (f)
1279                 kprintf(" ");
1280         if (f & TH_FIN)
1281                 kprintf("F");
1282         if (f & TH_SYN)
1283                 kprintf("S");
1284         if (f & TH_RST)
1285                 kprintf("R");
1286         if (f & TH_PUSH)
1287                 kprintf("P");
1288         if (f & TH_ACK)
1289                 kprintf("A");
1290         if (f & TH_URG)
1291                 kprintf("U");
1292         if (f & TH_ECE)
1293                 kprintf("E");
1294         if (f & TH_CWR)
1295                 kprintf("W");
1296 }
1297
1298 #define PF_SET_SKIP_STEPS(i)                                    \
1299         do {                                                    \
1300                 while (head[i] != cur) {                        \
1301                         head[i]->skip[i].ptr = cur;             \
1302                         head[i] = TAILQ_NEXT(head[i], entries); \
1303                 }                                               \
1304         } while (0)
1305
1306 void
1307 pf_calc_skip_steps(struct pf_rulequeue *rules)
1308 {
1309         struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT];
1310         int i;
1311
1312         cur = TAILQ_FIRST(rules);
1313         prev = cur;
1314         for (i = 0; i < PF_SKIP_COUNT; ++i)
1315                 head[i] = cur;
1316         while (cur != NULL) {
1317
1318                 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
1319                         PF_SET_SKIP_STEPS(PF_SKIP_IFP);
1320                 if (cur->direction != prev->direction)
1321                         PF_SET_SKIP_STEPS(PF_SKIP_DIR);
1322                 if (cur->af != prev->af)
1323                         PF_SET_SKIP_STEPS(PF_SKIP_AF);
1324                 if (cur->proto != prev->proto)
1325                         PF_SET_SKIP_STEPS(PF_SKIP_PROTO);
1326                 if (cur->src.neg != prev->src.neg ||
1327                     pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr))
1328                         PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR);
1329                 if (cur->src.port[0] != prev->src.port[0] ||
1330                     cur->src.port[1] != prev->src.port[1] ||
1331                     cur->src.port_op != prev->src.port_op)
1332                         PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
1333                 if (cur->dst.neg != prev->dst.neg ||
1334                     pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr))
1335                         PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR);
1336                 if (cur->dst.port[0] != prev->dst.port[0] ||
1337                     cur->dst.port[1] != prev->dst.port[1] ||
1338                     cur->dst.port_op != prev->dst.port_op)
1339                         PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
1340
1341                 prev = cur;
1342                 cur = TAILQ_NEXT(cur, entries);
1343         }
1344         for (i = 0; i < PF_SKIP_COUNT; ++i)
1345                 PF_SET_SKIP_STEPS(i);
1346 }
1347
1348 int
1349 pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2)
1350 {
1351         if (aw1->type != aw2->type)
1352                 return (1);
1353         switch (aw1->type) {
1354         case PF_ADDR_ADDRMASK:
1355                 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0))
1356                         return (1);
1357                 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0))
1358                         return (1);
1359                 return (0);
1360         case PF_ADDR_DYNIFTL:
1361                 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt);
1362         case PF_ADDR_NOROUTE:
1363         case PF_ADDR_URPFFAILED:
1364                 return (0);
1365         case PF_ADDR_TABLE:
1366                 return (aw1->p.tbl != aw2->p.tbl);
1367         case PF_ADDR_RTLABEL:
1368                 return (aw1->v.rtlabel != aw2->v.rtlabel);
1369         default:
1370                 kprintf("invalid address type: %d\n", aw1->type);
1371                 return (1);
1372         }
1373 }
1374
1375 u_int16_t
1376 pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp)
1377 {
1378         u_int32_t       l;
1379
1380         if (udp && !cksum)
1381                 return (0x0000);
1382         l = cksum + old - new;
1383         l = (l >> 16) + (l & 65535);
1384         l = l & 65535;
1385         if (udp && !l)
1386                 return (0xFFFF);
1387         return (l);
1388 }
1389
1390 void
1391 pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc,
1392     struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af)
1393 {
1394         struct pf_addr  ao;
1395         u_int16_t       po = *p;
1396
1397         PF_ACPY(&ao, a, af);
1398         PF_ACPY(a, an, af);
1399
1400         *p = pn;
1401
1402         switch (af) {
1403 #ifdef INET
1404         case AF_INET:
1405                 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1406                     ao.addr16[0], an->addr16[0], 0),
1407                     ao.addr16[1], an->addr16[1], 0);
1408                 *p = pn;
1409                 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1410                     ao.addr16[0], an->addr16[0], u),
1411                     ao.addr16[1], an->addr16[1], u),
1412                     po, pn, u);
1413                 break;
1414 #endif /* INET */
1415 #ifdef INET6
1416         case AF_INET6:
1417                 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1418                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1419                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1420                     ao.addr16[0], an->addr16[0], u),
1421                     ao.addr16[1], an->addr16[1], u),
1422                     ao.addr16[2], an->addr16[2], u),
1423                     ao.addr16[3], an->addr16[3], u),
1424                     ao.addr16[4], an->addr16[4], u),
1425                     ao.addr16[5], an->addr16[5], u),
1426                     ao.addr16[6], an->addr16[6], u),
1427                     ao.addr16[7], an->addr16[7], u),
1428                     po, pn, u);
1429                 break;
1430 #endif /* INET6 */
1431         }
1432 }
1433
1434
1435 /* Changes a u_int32_t.  Uses a void * so there are no align restrictions */
1436 void
1437 pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u)
1438 {
1439         u_int32_t       ao;
1440
1441         memcpy(&ao, a, sizeof(ao));
1442         memcpy(a, &an, sizeof(u_int32_t));
1443         *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u),
1444             ao % 65536, an % 65536, u);
1445 }
1446
1447 #ifdef INET6
1448 void
1449 pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u)
1450 {
1451         struct pf_addr  ao;
1452
1453         PF_ACPY(&ao, a, AF_INET6);
1454         PF_ACPY(a, an, AF_INET6);
1455
1456         *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1457             pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1458             pf_cksum_fixup(pf_cksum_fixup(*c,
1459             ao.addr16[0], an->addr16[0], u),
1460             ao.addr16[1], an->addr16[1], u),
1461             ao.addr16[2], an->addr16[2], u),
1462             ao.addr16[3], an->addr16[3], u),
1463             ao.addr16[4], an->addr16[4], u),
1464             ao.addr16[5], an->addr16[5], u),
1465             ao.addr16[6], an->addr16[6], u),
1466             ao.addr16[7], an->addr16[7], u);
1467 }
1468 #endif /* INET6 */
1469
1470 void
1471 pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa,
1472     struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c,
1473     u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af)
1474 {
1475         struct pf_addr  oia, ooa;
1476
1477         PF_ACPY(&oia, ia, af);
1478         PF_ACPY(&ooa, oa, af);
1479
1480         /* Change inner protocol port, fix inner protocol checksum. */
1481         if (ip != NULL) {
1482                 u_int16_t       oip = *ip;
1483                 u_int32_t       opc = 0;
1484
1485                 if (pc != NULL)
1486                         opc = *pc;
1487                 *ip = np;
1488                 if (pc != NULL)
1489                         *pc = pf_cksum_fixup(*pc, oip, *ip, u);
1490                 *ic = pf_cksum_fixup(*ic, oip, *ip, 0);
1491                 if (pc != NULL)
1492                         *ic = pf_cksum_fixup(*ic, opc, *pc, 0);
1493         }
1494         /* Change inner ip address, fix inner ip and icmp checksums. */
1495         PF_ACPY(ia, na, af);
1496         switch (af) {
1497 #ifdef INET
1498         case AF_INET: {
1499                 u_int32_t        oh2c = *h2c;
1500
1501                 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c,
1502                     oia.addr16[0], ia->addr16[0], 0),
1503                     oia.addr16[1], ia->addr16[1], 0);
1504                 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1505                     oia.addr16[0], ia->addr16[0], 0),
1506                     oia.addr16[1], ia->addr16[1], 0);
1507                 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0);
1508                 break;
1509         }
1510 #endif /* INET */
1511 #ifdef INET6
1512         case AF_INET6:
1513                 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1514                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1515                     pf_cksum_fixup(pf_cksum_fixup(*ic,
1516                     oia.addr16[0], ia->addr16[0], u),
1517                     oia.addr16[1], ia->addr16[1], u),
1518                     oia.addr16[2], ia->addr16[2], u),
1519                     oia.addr16[3], ia->addr16[3], u),
1520                     oia.addr16[4], ia->addr16[4], u),
1521                     oia.addr16[5], ia->addr16[5], u),
1522                     oia.addr16[6], ia->addr16[6], u),
1523                     oia.addr16[7], ia->addr16[7], u);
1524                 break;
1525 #endif /* INET6 */
1526         }
1527         /* Change outer ip address, fix outer ip or icmpv6 checksum. */
1528         PF_ACPY(oa, na, af);
1529         switch (af) {
1530 #ifdef INET
1531         case AF_INET:
1532                 *hc = pf_cksum_fixup(pf_cksum_fixup(*hc,
1533                     ooa.addr16[0], oa->addr16[0], 0),
1534                     ooa.addr16[1], oa->addr16[1], 0);
1535                 break;
1536 #endif /* INET */
1537 #ifdef INET6
1538         case AF_INET6:
1539                 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1540                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1541                     pf_cksum_fixup(pf_cksum_fixup(*ic,
1542                     ooa.addr16[0], oa->addr16[0], u),
1543                     ooa.addr16[1], oa->addr16[1], u),
1544                     ooa.addr16[2], oa->addr16[2], u),
1545                     ooa.addr16[3], oa->addr16[3], u),
1546                     ooa.addr16[4], oa->addr16[4], u),
1547                     ooa.addr16[5], oa->addr16[5], u),
1548                     ooa.addr16[6], oa->addr16[6], u),
1549                     ooa.addr16[7], oa->addr16[7], u);
1550                 break;
1551 #endif /* INET6 */
1552         }
1553 }
1554
1555
1556 /*
1557  * Need to modulate the sequence numbers in the TCP SACK option
1558  * (credits to Krzysztof Pfaff for report and patch)
1559  */
1560 int
1561 pf_modulate_sack(struct mbuf *m, int off, struct pf_pdesc *pd,
1562     struct tcphdr *th, struct pf_state_peer *dst)
1563 {
1564         int hlen = (th->th_off << 2) - sizeof(*th), thoptlen = hlen;
1565         u_int8_t opts[TCP_MAXOLEN], *opt = opts;
1566         int copyback = 0, i, olen;
1567         struct raw_sackblock sack;
1568
1569 #define TCPOLEN_SACKLEN (TCPOLEN_SACK + 2)
1570         if (hlen < TCPOLEN_SACKLEN ||
1571             !pf_pull_hdr(m, off + sizeof(*th), opts, hlen, NULL, NULL, pd->af))
1572                 return 0;
1573
1574         while (hlen >= TCPOLEN_SACKLEN) {
1575                 olen = opt[1];
1576                 switch (*opt) {
1577                 case TCPOPT_EOL:        /* FALLTHROUGH */
1578                 case TCPOPT_NOP:
1579                         opt++;
1580                         hlen--;
1581                         break;
1582                 case TCPOPT_SACK:
1583                         if (olen > hlen)
1584                                 olen = hlen;
1585                         if (olen >= TCPOLEN_SACKLEN) {
1586                                 for (i = 2; i + TCPOLEN_SACK <= olen;
1587                                     i += TCPOLEN_SACK) {
1588                                         memcpy(&sack, &opt[i], sizeof(sack));
1589                                         pf_change_a(&sack.rblk_start, &th->th_sum,
1590                                             htonl(ntohl(sack.rblk_start) -
1591                                             dst->seqdiff), 0);
1592                                         pf_change_a(&sack.rblk_end, &th->th_sum,
1593                                             htonl(ntohl(sack.rblk_end) -
1594                                             dst->seqdiff), 0);
1595                                         memcpy(&opt[i], &sack, sizeof(sack));
1596                                 }
1597                                 copyback = 1;
1598                         }
1599                         /* FALLTHROUGH */
1600                 default:
1601                         if (olen < 2)
1602                                 olen = 2;
1603                         hlen -= olen;
1604                         opt += olen;
1605                 }
1606         }
1607
1608         if (copyback)
1609                 m_copyback(m, off + sizeof(*th), thoptlen, opts);
1610         return (copyback);
1611 }
1612
1613 void
1614 pf_send_tcp(const struct pf_rule *r, sa_family_t af,
1615     const struct pf_addr *saddr, const struct pf_addr *daddr,
1616     u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack,
1617     u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl, int tag,
1618     u_int16_t rtag, struct ether_header *eh, struct ifnet *ifp)
1619 {
1620         struct mbuf     *m;
1621         int              len = 0, tlen;
1622 #ifdef INET
1623         struct ip       *h = NULL;
1624 #endif /* INET */
1625 #ifdef INET6
1626         struct ip6_hdr  *h6 = NULL;
1627 #endif /* INET6 */
1628         struct tcphdr   *th = NULL;
1629         char            *opt;
1630
1631         /* maximum segment size tcp option */
1632         tlen = sizeof(struct tcphdr);
1633         if (mss)
1634                 tlen += 4;
1635
1636         switch (af) {
1637 #ifdef INET
1638         case AF_INET:
1639                 len = sizeof(struct ip) + tlen;
1640                 break;
1641 #endif /* INET */
1642 #ifdef INET6
1643         case AF_INET6:
1644                 len = sizeof(struct ip6_hdr) + tlen;
1645                 break;
1646 #endif /* INET6 */
1647         }
1648
1649         /* create outgoing mbuf */
1650         m = m_gethdr(MB_DONTWAIT, MT_HEADER);
1651         if (m == NULL)
1652                 return;
1653         if (tag)
1654                 m->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
1655         m->m_pkthdr.pf.tag = rtag;
1656
1657         if (r != NULL && r->rtableid >= 0)
1658                 m->m_pkthdr.pf.rtableid = r->rtableid;
1659
1660 #ifdef ALTQ
1661         if (r != NULL && r->qid) {
1662                 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
1663                 m->m_pkthdr.pf.qid = r->qid;
1664                 m->m_pkthdr.pf.ecn_af = af;
1665                 m->m_pkthdr.pf.hdr = mtod(m, struct ip *);
1666         }
1667 #endif /* ALTQ */
1668         m->m_data += max_linkhdr;
1669         m->m_pkthdr.len = m->m_len = len;
1670         m->m_pkthdr.rcvif = NULL;
1671         bzero(m->m_data, len);
1672         switch (af) {
1673 #ifdef INET
1674         case AF_INET:
1675                 h = mtod(m, struct ip *);
1676
1677                 /* IP header fields included in the TCP checksum */
1678                 h->ip_p = IPPROTO_TCP;
1679                 h->ip_len = tlen;
1680                 h->ip_src.s_addr = saddr->v4.s_addr;
1681                 h->ip_dst.s_addr = daddr->v4.s_addr;
1682
1683                 th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip));
1684                 break;
1685 #endif /* INET */
1686 #ifdef INET6
1687         case AF_INET6:
1688                 h6 = mtod(m, struct ip6_hdr *);
1689
1690                 /* IP header fields included in the TCP checksum */
1691                 h6->ip6_nxt = IPPROTO_TCP;
1692                 h6->ip6_plen = htons(tlen);
1693                 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr));
1694                 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr));
1695
1696                 th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr));
1697                 break;
1698 #endif /* INET6 */
1699         }
1700
1701         /* TCP header */
1702         th->th_sport = sport;
1703         th->th_dport = dport;
1704         th->th_seq = htonl(seq);
1705         th->th_ack = htonl(ack);
1706         th->th_off = tlen >> 2;
1707         th->th_flags = flags;
1708         th->th_win = htons(win);
1709
1710         if (mss) {
1711                 opt = (char *)(th + 1);
1712                 opt[0] = TCPOPT_MAXSEG;
1713                 opt[1] = 4;
1714                 mss = htons(mss);
1715                 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2);
1716         }
1717
1718         switch (af) {
1719 #ifdef INET
1720         case AF_INET:
1721                 /* TCP checksum */
1722                 th->th_sum = in_cksum(m, len);
1723
1724                 /* Finish the IP header */
1725                 h->ip_v = 4;
1726                 h->ip_hl = sizeof(*h) >> 2;
1727                 h->ip_tos = IPTOS_LOWDELAY;
1728                 h->ip_len = len;
1729                 h->ip_off = path_mtu_discovery ? IP_DF : 0;
1730                 h->ip_ttl = ttl ? ttl : ip_defttl;
1731                 h->ip_sum = 0;
1732                 if (eh == NULL) {
1733                         ip_output(m, NULL, NULL, 0, NULL, NULL);
1734                 } else {
1735                         struct route             ro;
1736                         struct rtentry           rt;
1737                         struct ether_header     *e = (void *)ro.ro_dst.sa_data;
1738
1739                         if (ifp == NULL) {
1740                                 m_freem(m);
1741                                 return;
1742                         }
1743                         rt.rt_ifp = ifp;
1744                         ro.ro_rt = &rt;
1745                         ro.ro_dst.sa_len = sizeof(ro.ro_dst);
1746                         ro.ro_dst.sa_family = pseudo_AF_HDRCMPLT;
1747                         bcopy(eh->ether_dhost, e->ether_shost, ETHER_ADDR_LEN);
1748                         bcopy(eh->ether_shost, e->ether_dhost, ETHER_ADDR_LEN);
1749                         e->ether_type = eh->ether_type;
1750                         /* XXX_IMPORT: later */
1751                         ip_output(m, (void *)NULL, &ro, 0,
1752                             (void *)NULL, (void *)NULL);
1753                 }
1754                 break;
1755 #endif /* INET */
1756 #ifdef INET6
1757         case AF_INET6:
1758                 /* TCP checksum */
1759                 th->th_sum = in6_cksum(m, IPPROTO_TCP,
1760                     sizeof(struct ip6_hdr), tlen);
1761
1762                 h6->ip6_vfc |= IPV6_VERSION;
1763                 h6->ip6_hlim = IPV6_DEFHLIM;
1764
1765                 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
1766                 break;
1767 #endif /* INET6 */
1768         }
1769 }
1770
1771 void
1772 pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af,
1773     struct pf_rule *r)
1774 {
1775         struct mbuf     *m0;
1776
1777         m0 = m_copy(m, 0, M_COPYALL);
1778         m0->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
1779
1780         if (r->rtableid >= 0)
1781                 m0->m_pkthdr.pf.rtableid = r->rtableid;
1782
1783 #ifdef ALTQ
1784         if (r->qid) {
1785                 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
1786                 m0->m_pkthdr.pf.qid = r->qid;
1787                 m0->m_pkthdr.pf.ecn_af = af;
1788                 m0->m_pkthdr.pf.hdr = mtod(m0, struct ip *);
1789         }
1790 #endif /* ALTQ */
1791
1792         switch (af) {
1793 #ifdef INET
1794         case AF_INET:
1795                 icmp_error(m0, type, code, 0, 0);
1796                 break;
1797 #endif /* INET */
1798 #ifdef INET6
1799         case AF_INET6:
1800                 icmp6_error(m0, type, code, 0);
1801                 break;
1802 #endif /* INET6 */
1803         }
1804 }
1805
1806 /*
1807  * Return 1 if the addresses a and b match (with mask m), otherwise return 0.
1808  * If n is 0, they match if they are equal. If n is != 0, they match if they
1809  * are different.
1810  */
1811 int
1812 pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m,
1813     struct pf_addr *b, sa_family_t af)
1814 {
1815         int     match = 0;
1816
1817         switch (af) {
1818 #ifdef INET
1819         case AF_INET:
1820                 if ((a->addr32[0] & m->addr32[0]) ==
1821                     (b->addr32[0] & m->addr32[0]))
1822                         match++;
1823                 break;
1824 #endif /* INET */
1825 #ifdef INET6
1826         case AF_INET6:
1827                 if (((a->addr32[0] & m->addr32[0]) ==
1828                      (b->addr32[0] & m->addr32[0])) &&
1829                     ((a->addr32[1] & m->addr32[1]) ==
1830                      (b->addr32[1] & m->addr32[1])) &&
1831                     ((a->addr32[2] & m->addr32[2]) ==
1832                      (b->addr32[2] & m->addr32[2])) &&
1833                     ((a->addr32[3] & m->addr32[3]) ==
1834                      (b->addr32[3] & m->addr32[3])))
1835                         match++;
1836                 break;
1837 #endif /* INET6 */
1838         }
1839         if (match) {
1840                 if (n)
1841                         return (0);
1842                 else
1843                         return (1);
1844         } else {
1845                 if (n)
1846                         return (1);
1847                 else
1848                         return (0);
1849         }
1850 }
1851
1852 int
1853 pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p)
1854 {
1855         switch (op) {
1856         case PF_OP_IRG:
1857                 return ((p > a1) && (p < a2));
1858         case PF_OP_XRG:
1859                 return ((p < a1) || (p > a2));
1860         case PF_OP_RRG:
1861                 return ((p >= a1) && (p <= a2));
1862         case PF_OP_EQ:
1863                 return (p == a1);
1864         case PF_OP_NE:
1865                 return (p != a1);
1866         case PF_OP_LT:
1867                 return (p < a1);
1868         case PF_OP_LE:
1869                 return (p <= a1);
1870         case PF_OP_GT:
1871                 return (p > a1);
1872         case PF_OP_GE:
1873                 return (p >= a1);
1874         }
1875         return (0); /* never reached */
1876 }
1877
1878 int
1879 pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p)
1880 {
1881         a1 = ntohs(a1);
1882         a2 = ntohs(a2);
1883         p = ntohs(p);
1884         return (pf_match(op, a1, a2, p));
1885 }
1886
1887 int
1888 pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u)
1889 {
1890         if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
1891                 return (0);
1892         return (pf_match(op, a1, a2, u));
1893 }
1894
1895 int
1896 pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g)
1897 {
1898         if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
1899                 return (0);
1900         return (pf_match(op, a1, a2, g));
1901 }
1902
1903 int
1904 pf_match_tag(struct mbuf *m, struct pf_rule *r, int *tag)
1905 {
1906         if (*tag == -1)
1907                 *tag = m->m_pkthdr.pf.tag;
1908
1909         return ((!r->match_tag_not && r->match_tag == *tag) ||
1910             (r->match_tag_not && r->match_tag != *tag));
1911 }
1912
1913 int
1914 pf_tag_packet(struct mbuf *m, int tag, int rtableid)
1915 {
1916         if (tag <= 0 && rtableid < 0)
1917                 return (0);
1918
1919         if (tag > 0)
1920                 m->m_pkthdr.pf.tag = tag;
1921         if (rtableid >= 0)
1922                 m->m_pkthdr.pf.rtableid = rtableid;
1923
1924         return (0);
1925 }
1926
1927 void
1928 pf_step_into_anchor(int *depth, struct pf_ruleset **rs, int n,
1929     struct pf_rule **r, struct pf_rule **a,  int *match)
1930 {
1931         struct pf_anchor_stackframe     *f;
1932
1933         (*r)->anchor->match = 0;
1934         if (match)
1935                 *match = 0;
1936         if (*depth >= sizeof(pf_anchor_stack) /
1937             sizeof(pf_anchor_stack[0])) {
1938                 kprintf("pf_step_into_anchor: stack overflow\n");
1939                 *r = TAILQ_NEXT(*r, entries);
1940                 return;
1941         } else if (*depth == 0 && a != NULL)
1942                 *a = *r;
1943         f = pf_anchor_stack + (*depth)++;
1944         f->rs = *rs;
1945         f->r = *r;
1946         if ((*r)->anchor_wildcard) {
1947                 f->parent = &(*r)->anchor->children;
1948                 if ((f->child = RB_MIN(pf_anchor_node, f->parent)) ==
1949                     NULL) {
1950                         *r = NULL;
1951                         return;
1952                 }
1953                 *rs = &f->child->ruleset;
1954         } else {
1955                 f->parent = NULL;
1956                 f->child = NULL;
1957                 *rs = &(*r)->anchor->ruleset;
1958         }
1959         *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
1960 }
1961
1962 int
1963 pf_step_out_of_anchor(int *depth, struct pf_ruleset **rs, int n,
1964     struct pf_rule **r, struct pf_rule **a, int *match)
1965 {
1966         struct pf_anchor_stackframe     *f;
1967         int quick = 0;
1968
1969         do {
1970                 if (*depth <= 0)
1971                         break;
1972                 f = pf_anchor_stack + *depth - 1;
1973                 if (f->parent != NULL && f->child != NULL) {
1974                         if (f->child->match ||
1975                             (match != NULL && *match)) {
1976                                 f->r->anchor->match = 1;
1977                                 *match = 0;
1978                         }
1979                         f->child = RB_NEXT(pf_anchor_node, f->parent, f->child);
1980                         if (f->child != NULL) {
1981                                 *rs = &f->child->ruleset;
1982                                 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
1983                                 if (*r == NULL)
1984                                         continue;
1985                                 else
1986                                         break;
1987                         }
1988                 }
1989                 (*depth)--;
1990                 if (*depth == 0 && a != NULL)
1991                         *a = NULL;
1992                 *rs = f->rs;
1993                 if (f->r->anchor->match || (match  != NULL && *match))
1994                         quick = f->r->quick;
1995                 *r = TAILQ_NEXT(f->r, entries);
1996         } while (*r == NULL);
1997
1998         return (quick);
1999 }
2000
2001 #ifdef INET6
2002 void
2003 pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr,
2004     struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af)
2005 {
2006         switch (af) {
2007 #ifdef INET
2008         case AF_INET:
2009                 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2010                 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2011                 break;
2012 #endif /* INET */
2013         case AF_INET6:
2014                 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2015                 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2016                 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) |
2017                 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]);
2018                 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) |
2019                 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]);
2020                 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) |
2021                 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]);
2022                 break;
2023         }
2024 }
2025
2026 void
2027 pf_addr_inc(struct pf_addr *addr, sa_family_t af)
2028 {
2029         switch (af) {
2030 #ifdef INET
2031         case AF_INET:
2032                 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1);
2033                 break;
2034 #endif /* INET */
2035         case AF_INET6:
2036                 if (addr->addr32[3] == 0xffffffff) {
2037                         addr->addr32[3] = 0;
2038                         if (addr->addr32[2] == 0xffffffff) {
2039                                 addr->addr32[2] = 0;
2040                                 if (addr->addr32[1] == 0xffffffff) {
2041                                         addr->addr32[1] = 0;
2042                                         addr->addr32[0] =
2043                                             htonl(ntohl(addr->addr32[0]) + 1);
2044                                 } else
2045                                         addr->addr32[1] =
2046                                             htonl(ntohl(addr->addr32[1]) + 1);
2047                         } else
2048                                 addr->addr32[2] =
2049                                     htonl(ntohl(addr->addr32[2]) + 1);
2050                 } else
2051                         addr->addr32[3] =
2052                             htonl(ntohl(addr->addr32[3]) + 1);
2053                 break;
2054         }
2055 }
2056 #endif /* INET6 */
2057
2058 #define mix(a,b,c) \
2059         do {                                    \
2060                 a -= b; a -= c; a ^= (c >> 13); \
2061                 b -= c; b -= a; b ^= (a << 8);  \
2062                 c -= a; c -= b; c ^= (b >> 13); \
2063                 a -= b; a -= c; a ^= (c >> 12); \
2064                 b -= c; b -= a; b ^= (a << 16); \
2065                 c -= a; c -= b; c ^= (b >> 5);  \
2066                 a -= b; a -= c; a ^= (c >> 3);  \
2067                 b -= c; b -= a; b ^= (a << 10); \
2068                 c -= a; c -= b; c ^= (b >> 15); \
2069         } while (0)
2070
2071 /*
2072  * hash function based on bridge_hash in if_bridge.c
2073  */
2074 void
2075 pf_hash(struct pf_addr *inaddr, struct pf_addr *hash,
2076     struct pf_poolhashkey *key, sa_family_t af)
2077 {
2078         u_int32_t       a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0];
2079
2080         switch (af) {
2081 #ifdef INET
2082         case AF_INET:
2083                 a += inaddr->addr32[0];
2084                 b += key->key32[1];
2085                 mix(a, b, c);
2086                 hash->addr32[0] = c + key->key32[2];
2087                 break;
2088 #endif /* INET */
2089 #ifdef INET6
2090         case AF_INET6:
2091                 a += inaddr->addr32[0];
2092                 b += inaddr->addr32[2];
2093                 mix(a, b, c);
2094                 hash->addr32[0] = c;
2095                 a += inaddr->addr32[1];
2096                 b += inaddr->addr32[3];
2097                 c += key->key32[1];
2098                 mix(a, b, c);
2099                 hash->addr32[1] = c;
2100                 a += inaddr->addr32[2];
2101                 b += inaddr->addr32[1];
2102                 c += key->key32[2];
2103                 mix(a, b, c);
2104                 hash->addr32[2] = c;
2105                 a += inaddr->addr32[3];
2106                 b += inaddr->addr32[0];
2107                 c += key->key32[3];
2108                 mix(a, b, c);
2109                 hash->addr32[3] = c;
2110                 break;
2111 #endif /* INET6 */
2112         }
2113 }
2114
2115 int
2116 pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr,
2117     struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn)
2118 {
2119         unsigned char            hash[16];
2120         struct pf_pool          *rpool = &r->rpool;
2121         struct pf_addr          *raddr = &rpool->cur->addr.v.a.addr;
2122         struct pf_addr          *rmask = &rpool->cur->addr.v.a.mask;
2123         struct pf_pooladdr      *acur = rpool->cur;
2124         struct pf_src_node       k;
2125
2126         if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR &&
2127             (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2128                 k.af = af;
2129                 PF_ACPY(&k.addr, saddr, af);
2130                 if (r->rule_flag & PFRULE_RULESRCTRACK ||
2131                     r->rpool.opts & PF_POOL_STICKYADDR)
2132                         k.rule.ptr = r;
2133                 else
2134                         k.rule.ptr = NULL;
2135                 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
2136                 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
2137                 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) {
2138                         PF_ACPY(naddr, &(*sn)->raddr, af);
2139                         if (pf_status.debug >= PF_DEBUG_MISC) {
2140                                 kprintf("pf_map_addr: src tracking maps ");
2141                                 pf_print_host(&k.addr, 0, af);
2142                                 kprintf(" to ");
2143                                 pf_print_host(naddr, 0, af);
2144                                 kprintf("\n");
2145                         }
2146                         return (0);
2147                 }
2148         }
2149
2150         if (rpool->cur->addr.type == PF_ADDR_NOROUTE)
2151                 return (1);
2152         if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2153                 switch (af) {
2154 #ifdef INET
2155                 case AF_INET:
2156                         if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 &&
2157                             (rpool->opts & PF_POOL_TYPEMASK) !=
2158                             PF_POOL_ROUNDROBIN)
2159                                 return (1);
2160                          raddr = &rpool->cur->addr.p.dyn->pfid_addr4;
2161                          rmask = &rpool->cur->addr.p.dyn->pfid_mask4;
2162                         break;
2163 #endif /* INET */
2164 #ifdef INET6
2165                 case AF_INET6:
2166                         if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 &&
2167                             (rpool->opts & PF_POOL_TYPEMASK) !=
2168                             PF_POOL_ROUNDROBIN)
2169                                 return (1);
2170                         raddr = &rpool->cur->addr.p.dyn->pfid_addr6;
2171                         rmask = &rpool->cur->addr.p.dyn->pfid_mask6;
2172                         break;
2173 #endif /* INET6 */
2174                 }
2175         } else if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2176                 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN)
2177                         return (1); /* unsupported */
2178         } else {
2179                 raddr = &rpool->cur->addr.v.a.addr;
2180                 rmask = &rpool->cur->addr.v.a.mask;
2181         }
2182
2183         switch (rpool->opts & PF_POOL_TYPEMASK) {
2184         case PF_POOL_NONE:
2185                 PF_ACPY(naddr, raddr, af);
2186                 break;
2187         case PF_POOL_BITMASK:
2188                 PF_POOLMASK(naddr, raddr, rmask, saddr, af);
2189                 break;
2190         case PF_POOL_RANDOM:
2191                 if (init_addr != NULL && PF_AZERO(init_addr, af)) {
2192                         switch (af) {
2193 #ifdef INET
2194                         case AF_INET:
2195                                 rpool->counter.addr32[0] = htonl(karc4random());
2196                                 break;
2197 #endif /* INET */
2198 #ifdef INET6
2199                         case AF_INET6:
2200                                 if (rmask->addr32[3] != 0xffffffff)
2201                                         rpool->counter.addr32[3] =
2202                                             htonl(karc4random());
2203                                 else
2204                                         break;
2205                                 if (rmask->addr32[2] != 0xffffffff)
2206                                         rpool->counter.addr32[2] =
2207                                             htonl(karc4random());
2208                                 else
2209                                         break;
2210                                 if (rmask->addr32[1] != 0xffffffff)
2211                                         rpool->counter.addr32[1] =
2212                                             htonl(karc4random());
2213                                 else
2214                                         break;
2215                                 if (rmask->addr32[0] != 0xffffffff)
2216                                         rpool->counter.addr32[0] =
2217                                             htonl(karc4random());
2218                                 break;
2219 #endif /* INET6 */
2220                         }
2221                         PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2222                         PF_ACPY(init_addr, naddr, af);
2223
2224                 } else {
2225                         PF_AINC(&rpool->counter, af);
2226                         PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2227                 }
2228                 break;
2229         case PF_POOL_SRCHASH:
2230                 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af);
2231                 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af);
2232                 break;
2233         case PF_POOL_ROUNDROBIN:
2234                 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2235                         if (!pfr_pool_get(rpool->cur->addr.p.tbl,
2236                             &rpool->tblidx, &rpool->counter,
2237                             &raddr, &rmask, af))
2238                                 goto get_addr;
2239                 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2240                         if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2241                             &rpool->tblidx, &rpool->counter,
2242                             &raddr, &rmask, af))
2243                                 goto get_addr;
2244                 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af))
2245                         goto get_addr;
2246
2247         try_next:
2248                 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL)
2249                         rpool->cur = TAILQ_FIRST(&rpool->list);
2250                 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2251                         rpool->tblidx = -1;
2252                         if (pfr_pool_get(rpool->cur->addr.p.tbl,
2253                             &rpool->tblidx, &rpool->counter,
2254                             &raddr, &rmask, af)) {
2255                                 /* table contains no address of type 'af' */
2256                                 if (rpool->cur != acur)
2257                                         goto try_next;
2258                                 return (1);
2259                         }
2260                 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2261                         rpool->tblidx = -1;
2262                         if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2263                             &rpool->tblidx, &rpool->counter,
2264                             &raddr, &rmask, af)) {
2265                                 /* table contains no address of type 'af' */
2266                                 if (rpool->cur != acur)
2267                                         goto try_next;
2268                                 return (1);
2269                         }
2270                 } else {
2271                         raddr = &rpool->cur->addr.v.a.addr;
2272                         rmask = &rpool->cur->addr.v.a.mask;
2273                         PF_ACPY(&rpool->counter, raddr, af);
2274                 }
2275
2276         get_addr:
2277                 PF_ACPY(naddr, &rpool->counter, af);
2278                 if (init_addr != NULL && PF_AZERO(init_addr, af))
2279                         PF_ACPY(init_addr, naddr, af);
2280                 PF_AINC(&rpool->counter, af);
2281                 break;
2282         }
2283         if (*sn != NULL)
2284                 PF_ACPY(&(*sn)->raddr, naddr, af);
2285
2286         if (pf_status.debug >= PF_DEBUG_MISC &&
2287             (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2288                 kprintf("pf_map_addr: selected address ");
2289                 pf_print_host(naddr, 0, af);
2290                 kprintf("\n");
2291         }
2292
2293         return (0);
2294 }
2295
2296 int
2297 pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r,
2298     struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport,
2299     struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high,
2300     struct pf_src_node **sn)
2301 {
2302         struct pf_state_key_cmp key;
2303         struct pf_addr          init_addr;
2304         u_int16_t               cut;
2305
2306         bzero(&init_addr, sizeof(init_addr));
2307         if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2308                 return (1);
2309
2310         if (proto == IPPROTO_ICMP) {
2311                 low = 1;
2312                 high = 65535;
2313         }
2314
2315         do {
2316                 key.af = af;
2317                 key.proto = proto;
2318                 PF_ACPY(&key.ext.addr, daddr, key.af);
2319                 PF_ACPY(&key.gwy.addr, naddr, key.af);
2320                 key.ext.port = dport;
2321
2322                 /*
2323                  * port search; start random, step;
2324                  * similar 2 portloop in in_pcbbind
2325                  */
2326                 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP ||
2327                     proto == IPPROTO_ICMP)) {
2328                         key.gwy.port = dport;
2329                         if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
2330                                 return (0);
2331                 } else if (low == 0 && high == 0) {
2332                         key.gwy.port = *nport;
2333                         if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
2334                                 return (0);
2335                 } else if (low == high) {
2336                         key.gwy.port = htons(low);
2337                         if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) {
2338                                 *nport = htons(low);
2339                                 return (0);
2340                         }
2341                 } else {
2342                         u_int16_t tmp;
2343
2344                         if (low > high) {
2345                                 tmp = low;
2346                                 low = high;
2347                                 high = tmp;
2348                         }
2349                         /* low < high */
2350                         cut = htonl(karc4random()) % (1 + high - low) + low;
2351                         /* low <= cut <= high */
2352                         for (tmp = cut; tmp <= high; ++(tmp)) {
2353                                 key.gwy.port = htons(tmp);
2354                                 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
2355                                     NULL) {
2356                                         *nport = htons(tmp);
2357                                         return (0);
2358                                 }
2359                         }
2360                         for (tmp = cut - 1; tmp >= low; --(tmp)) {
2361                                 key.gwy.port = htons(tmp);
2362                                 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
2363                                     NULL) {
2364                                         *nport = htons(tmp);
2365                                         return (0);
2366                                 }
2367                         }
2368                 }
2369
2370                 switch (r->rpool.opts & PF_POOL_TYPEMASK) {
2371                 case PF_POOL_RANDOM:
2372                 case PF_POOL_ROUNDROBIN:
2373                         if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2374                                 return (1);
2375                         break;
2376                 case PF_POOL_NONE:
2377                 case PF_POOL_SRCHASH:
2378                 case PF_POOL_BITMASK:
2379                 default:
2380                         return (1);
2381                 }
2382         } while (! PF_AEQ(&init_addr, naddr, af) );
2383
2384         return (1);                                     /* none available */
2385 }
2386
2387 struct pf_rule *
2388 pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off,
2389     int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport,
2390     struct pf_addr *daddr, u_int16_t dport, int rs_num)
2391 {
2392         struct pf_rule          *r, *rm = NULL;
2393         struct pf_ruleset       *ruleset = NULL;
2394         int                      tag = -1;
2395         int                      rtableid = -1;
2396         int                      asd = 0;
2397
2398         r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr);
2399         while (r && rm == NULL) {
2400                 struct pf_rule_addr     *src = NULL, *dst = NULL;
2401                 struct pf_addr_wrap     *xdst = NULL;
2402
2403                 if (r->action == PF_BINAT && direction == PF_IN) {
2404                         src = &r->dst;
2405                         if (r->rpool.cur != NULL)
2406                                 xdst = &r->rpool.cur->addr;
2407                 } else {
2408                         src = &r->src;
2409                         dst = &r->dst;
2410                 }
2411
2412                 r->evaluations++;
2413                 if (pfi_kif_match(r->kif, kif) == r->ifnot)
2414                         r = r->skip[PF_SKIP_IFP].ptr;
2415                 else if (r->direction && r->direction != direction)
2416                         r = r->skip[PF_SKIP_DIR].ptr;
2417                 else if (r->af && r->af != pd->af)
2418                         r = r->skip[PF_SKIP_AF].ptr;
2419                 else if (r->proto && r->proto != pd->proto)
2420                         r = r->skip[PF_SKIP_PROTO].ptr;
2421                 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af,
2422                     src->neg, kif))
2423                         r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR :
2424                             PF_SKIP_DST_ADDR].ptr;
2425                 else if (src->port_op && !pf_match_port(src->port_op,
2426                     src->port[0], src->port[1], sport))
2427                         r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT :
2428                             PF_SKIP_DST_PORT].ptr;
2429                 else if (dst != NULL &&
2430                     PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->neg, NULL))
2431                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
2432                 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af,
2433                     0, NULL))
2434                         r = TAILQ_NEXT(r, entries);
2435                 else if (dst != NULL && dst->port_op &&
2436                     !pf_match_port(dst->port_op, dst->port[0],
2437                     dst->port[1], dport))
2438                         r = r->skip[PF_SKIP_DST_PORT].ptr;
2439                 else if (r->match_tag && !pf_match_tag(m, r, &tag))
2440                         r = TAILQ_NEXT(r, entries);
2441                 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto !=
2442                     IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m,
2443                     off, pd->hdr.tcp), r->os_fingerprint)))
2444                         r = TAILQ_NEXT(r, entries);
2445                 else {
2446                         if (r->tag)
2447                                 tag = r->tag;
2448                         if (r->rtableid >= 0)
2449                                 rtableid = r->rtableid;
2450                         if (r->anchor == NULL) {
2451                                 rm = r;
2452                         } else
2453                                 pf_step_into_anchor(&asd, &ruleset, rs_num,
2454                                     &r, NULL, NULL);
2455                 }
2456                 if (r == NULL)
2457                         pf_step_out_of_anchor(&asd, &ruleset, rs_num, &r,
2458                             NULL, NULL);
2459         }
2460         if (pf_tag_packet(m, tag, rtableid))
2461                 return (NULL);
2462         if (rm != NULL && (rm->action == PF_NONAT ||
2463             rm->action == PF_NORDR || rm->action == PF_NOBINAT))
2464                 return (NULL);
2465         return (rm);
2466 }
2467
2468 struct pf_rule *
2469 pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction,
2470     struct pfi_kif *kif, struct pf_src_node **sn,
2471     struct pf_addr *saddr, u_int16_t sport,
2472     struct pf_addr *daddr, u_int16_t dport,
2473     struct pf_addr *naddr, u_int16_t *nport)
2474 {
2475         struct pf_rule  *r = NULL;
2476
2477         if (direction == PF_OUT) {
2478                 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2479                     sport, daddr, dport, PF_RULESET_BINAT);
2480                 if (r == NULL)
2481                         r = pf_match_translation(pd, m, off, direction, kif,
2482                             saddr, sport, daddr, dport, PF_RULESET_NAT);
2483         } else {
2484                 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2485                     sport, daddr, dport, PF_RULESET_RDR);
2486                 if (r == NULL)
2487                         r = pf_match_translation(pd, m, off, direction, kif,
2488                             saddr, sport, daddr, dport, PF_RULESET_BINAT);
2489         }
2490
2491         if (r != NULL) {
2492                 switch (r->action) {
2493                 case PF_NONAT:
2494                 case PF_NOBINAT:
2495                 case PF_NORDR:
2496                         return (NULL);
2497                 case PF_NAT:
2498                         if (pf_get_sport(pd->af, pd->proto, r, saddr,
2499                             daddr, dport, naddr, nport, r->rpool.proxy_port[0],
2500                             r->rpool.proxy_port[1], sn)) {
2501                                 DPFPRINTF(PF_DEBUG_MISC,
2502                                     ("pf: NAT proxy port allocation "
2503                                     "(%u-%u) failed\n",
2504                                     r->rpool.proxy_port[0],
2505                                     r->rpool.proxy_port[1]));
2506                                 return (NULL);
2507                         }
2508                         break;
2509                 case PF_BINAT:
2510                         switch (direction) {
2511                         case PF_OUT:
2512                                 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){
2513                                         switch (pd->af) {
2514 #ifdef INET
2515                                         case AF_INET:
2516                                                 if (r->rpool.cur->addr.p.dyn->
2517                                                     pfid_acnt4 < 1)
2518                                                         return (NULL);
2519                                                 PF_POOLMASK(naddr,
2520                                                     &r->rpool.cur->addr.p.dyn->
2521                                                     pfid_addr4,
2522                                                     &r->rpool.cur->addr.p.dyn->
2523                                                     pfid_mask4,
2524                                                     saddr, AF_INET);
2525                                                 break;
2526 #endif /* INET */
2527 #ifdef INET6
2528                                         case AF_INET6:
2529                                                 if (r->rpool.cur->addr.p.dyn->
2530                                                     pfid_acnt6 < 1)
2531                                                         return (NULL);
2532                                                 PF_POOLMASK(naddr,
2533                                                     &r->rpool.cur->addr.p.dyn->
2534                                                     pfid_addr6,
2535                                                     &r->rpool.cur->addr.p.dyn->
2536                                                     pfid_mask6,
2537                                                     saddr, AF_INET6);
2538                                                 break;
2539 #endif /* INET6 */
2540                                         }
2541                                 } else
2542                                         PF_POOLMASK(naddr,
2543                                             &r->rpool.cur->addr.v.a.addr,
2544                                             &r->rpool.cur->addr.v.a.mask,
2545                                             saddr, pd->af);
2546                                 break;
2547                         case PF_IN:
2548                                 if (r->src.addr.type == PF_ADDR_DYNIFTL) {
2549                                         switch (pd->af) {
2550 #ifdef INET
2551                                         case AF_INET:
2552                                                 if (r->src.addr.p.dyn->
2553                                                     pfid_acnt4 < 1)
2554                                                         return (NULL);
2555                                                 PF_POOLMASK(naddr,
2556                                                     &r->src.addr.p.dyn->
2557                                                     pfid_addr4,
2558                                                     &r->src.addr.p.dyn->
2559                                                     pfid_mask4,
2560                                                     daddr, AF_INET);
2561                                                 break;
2562 #endif /* INET */
2563 #ifdef INET6
2564                                         case AF_INET6:
2565                                                 if (r->src.addr.p.dyn->
2566                                                     pfid_acnt6 < 1)
2567                                                         return (NULL);
2568                                                 PF_POOLMASK(naddr,
2569                                                     &r->src.addr.p.dyn->
2570                                                     pfid_addr6,
2571                                                     &r->src.addr.p.dyn->
2572                                                     pfid_mask6,
2573                                                     daddr, AF_INET6);
2574                                                 break;
2575 #endif /* INET6 */
2576                                         }
2577                                 } else
2578                                         PF_POOLMASK(naddr,
2579                                             &r->src.addr.v.a.addr,
2580                                             &r->src.addr.v.a.mask, daddr,
2581                                             pd->af);
2582                                 break;
2583                         }
2584                         break;
2585                 case PF_RDR: {
2586                         if (pf_map_addr(pd->af, r, saddr, naddr, NULL, sn))
2587                                 return (NULL);
2588                         if ((r->rpool.opts & PF_POOL_TYPEMASK) ==
2589                             PF_POOL_BITMASK)
2590                                 PF_POOLMASK(naddr, naddr,
2591                                     &r->rpool.cur->addr.v.a.mask, daddr,
2592                                     pd->af);
2593
2594                         if (r->rpool.proxy_port[1]) {
2595                                 u_int32_t       tmp_nport;
2596
2597                                 tmp_nport = ((ntohs(dport) -
2598                                     ntohs(r->dst.port[0])) %
2599                                     (r->rpool.proxy_port[1] -
2600                                     r->rpool.proxy_port[0] + 1)) +
2601                                     r->rpool.proxy_port[0];
2602
2603                                 /* wrap around if necessary */
2604                                 if (tmp_nport > 65535)
2605                                         tmp_nport -= 65535;
2606                                 *nport = htons((u_int16_t)tmp_nport);
2607                         } else if (r->rpool.proxy_port[0])
2608                                 *nport = htons(r->rpool.proxy_port[0]);
2609                         break;
2610                 }
2611                 default:
2612                         return (NULL);
2613                 }
2614         }
2615
2616         return (r);
2617 }
2618
2619 #ifdef SMP
2620 struct netmsg_hashlookup {
2621         struct netmsg           nm_netmsg;
2622         struct inpcb            **nm_pinp;
2623         struct inpcbinfo        *nm_pcbinfo;
2624         struct pf_addr          *nm_saddr;
2625         struct pf_addr          *nm_daddr;
2626         uint16_t                nm_sport;
2627         uint16_t                nm_dport;
2628         sa_family_t             nm_af;
2629 };
2630
2631 static void
2632 in_pcblookup_hash_handler(struct netmsg *msg0)
2633 {
2634         struct netmsg_hashlookup *msg = (struct netmsg_hashlookup *)msg0;
2635
2636         if (msg->nm_af == AF_INET)
2637                 *msg->nm_pinp = in_pcblookup_hash(msg->nm_pcbinfo,
2638                     msg->nm_saddr->v4, msg->nm_sport, msg->nm_daddr->v4,
2639                     msg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2640 #ifdef INET6
2641         else
2642                 *msg->nm_pinp = in6_pcblookup_hash(msg->nm_pcbinfo,
2643                     &msg->nm_saddr->v6, msg->nm_sport, &msg->nm_daddr->v6,
2644                     msg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2645 #endif /* INET6 */
2646         lwkt_replymsg(&msg->nm_netmsg.nm_lmsg, 0);
2647 }
2648 #endif /* SMP */
2649
2650 int
2651 pf_socket_lookup(int direction, struct pf_pdesc *pd)
2652 {
2653         struct pf_addr          *saddr, *daddr;
2654         u_int16_t                sport, dport;
2655         struct inpcbinfo        *pi;
2656         struct inpcb            *inp;
2657 #ifdef SMP
2658         struct netmsg_hashlookup *msg = NULL;
2659 #endif
2660         int                      pi_cpu = 0;
2661
2662         if (pd == NULL)
2663                 return (-1);
2664         pd->lookup.uid = UID_MAX;
2665         pd->lookup.gid = GID_MAX;
2666         pd->lookup.pid = NO_PID;
2667         if (direction == PF_IN) {
2668                 saddr = pd->src;
2669                 daddr = pd->dst;
2670         } else {
2671                 saddr = pd->dst;
2672                 daddr = pd->src;
2673         }
2674         switch (pd->proto) {
2675         case IPPROTO_TCP:
2676                 if (pd->hdr.tcp == NULL)
2677                         return (-1);
2678                 sport = pd->hdr.tcp->th_sport;
2679                 dport = pd->hdr.tcp->th_dport;
2680
2681                 pi_cpu = tcp_addrcpu(saddr->v4.s_addr, sport, daddr->v4.s_addr, dport);
2682                 pi = &tcbinfo[pi_cpu];
2683 #ifdef SMP
2684                 /*
2685                  * Our netstack runs lockless on MP systems
2686                  * (only for TCP connections at the moment).
2687                  *
2688                  * As we are not allowed to read another CPU's tcbinfo,
2689                  * we have to ask that CPU via remote call to search the
2690                  * table for us.
2691                  *
2692                  * Prepare a msg iff data belongs to another CPU.
2693                  */
2694                 if (pi_cpu != mycpu->gd_cpuid) {
2695                         msg = kmalloc(sizeof(*msg), M_LWKTMSG, M_INTWAIT);
2696                         netmsg_init(&msg->nm_netmsg, NULL, &netisr_afree_rport,
2697                                     0, in_pcblookup_hash_handler);
2698                         msg->nm_pinp = &inp;
2699                         msg->nm_pcbinfo = pi;
2700                         msg->nm_saddr = saddr;
2701                         msg->nm_sport = sport;
2702                         msg->nm_daddr = daddr;
2703                         msg->nm_dport = dport;
2704                         msg->nm_af = pd->af;
2705                 }
2706 #endif /* SMP */
2707                 break;
2708         case IPPROTO_UDP:
2709                 if (pd->hdr.udp == NULL)
2710                         return (-1);
2711                 sport = pd->hdr.udp->uh_sport;
2712                 dport = pd->hdr.udp->uh_dport;
2713                 pi = &udbinfo;
2714                 break;
2715         default:
2716                 return (-1);
2717         }
2718         if (direction != PF_IN) {
2719                 u_int16_t       p;
2720
2721                 p = sport;
2722                 sport = dport;
2723                 dport = p;
2724         }
2725         switch (pd->af) {
2726 #ifdef INET6
2727         case AF_INET6:
2728 #ifdef SMP
2729                 /*
2730                  * Query other CPU, second part
2731                  *
2732                  * msg only gets initialized when:
2733                  * 1) packet is TCP
2734                  * 2) the info belongs to another CPU
2735                  *
2736                  * Use some switch/case magic to avoid code duplication.
2737                  */
2738                 if (msg == NULL)
2739 #endif /* SMP */
2740                 {
2741                         inp = in6_pcblookup_hash(pi, &saddr->v6, sport,
2742                             &daddr->v6, dport, INPLOOKUP_WILDCARD, NULL);
2743
2744                         if (inp == NULL)
2745                                 return (-1);
2746                         break;
2747                 }
2748                 /* FALLTHROUGH if SMP and on other CPU */
2749 #endif /* INET6 */
2750         case AF_INET:
2751 #ifdef SMP
2752                 if (msg != NULL) {
2753                         lwkt_sendmsg(cpu_portfn(pi_cpu),
2754                                      &msg->nm_netmsg.nm_lmsg);
2755                 } else
2756 #endif /* SMP */
2757                 {
2758                         inp = in_pcblookup_hash(pi, saddr->v4, sport, daddr->v4,
2759                             dport, INPLOOKUP_WILDCARD, NULL);
2760                 }
2761                 if (inp == NULL)
2762                         return (-1);
2763                 break;
2764
2765         default:
2766                 return (-1);
2767         }
2768         pd->lookup.uid = inp->inp_socket->so_cred->cr_uid;
2769         pd->lookup.gid = inp->inp_socket->so_cred->cr_groups[0];
2770         return (1);
2771 }
2772
2773 u_int8_t
2774 pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2775 {
2776         int              hlen;
2777         u_int8_t         hdr[60];
2778         u_int8_t        *opt, optlen;
2779         u_int8_t         wscale = 0;
2780
2781         hlen = th_off << 2;             /* hlen <= sizeof(hdr) */
2782         if (hlen <= sizeof(struct tcphdr))
2783                 return (0);
2784         if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2785                 return (0);
2786         opt = hdr + sizeof(struct tcphdr);
2787         hlen -= sizeof(struct tcphdr);
2788         while (hlen >= 3) {
2789                 switch (*opt) {
2790                 case TCPOPT_EOL:
2791                 case TCPOPT_NOP:
2792                         ++opt;
2793                         --hlen;
2794                         break;
2795                 case TCPOPT_WINDOW:
2796                         wscale = opt[2];
2797                         if (wscale > TCP_MAX_WINSHIFT)
2798                                 wscale = TCP_MAX_WINSHIFT;
2799                         wscale |= PF_WSCALE_FLAG;
2800                         /* FALLTHROUGH */
2801                 default:
2802                         optlen = opt[1];
2803                         if (optlen < 2)
2804                                 optlen = 2;
2805                         hlen -= optlen;
2806                         opt += optlen;
2807                         break;
2808                 }
2809         }
2810         return (wscale);
2811 }
2812
2813 u_int16_t
2814 pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2815 {
2816         int              hlen;
2817         u_int8_t         hdr[60];
2818         u_int8_t        *opt, optlen;
2819         u_int16_t        mss = tcp_mssdflt;
2820
2821         hlen = th_off << 2;     /* hlen <= sizeof(hdr) */
2822         if (hlen <= sizeof(struct tcphdr))
2823                 return (0);
2824         if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2825                 return (0);
2826         opt = hdr + sizeof(struct tcphdr);
2827         hlen -= sizeof(struct tcphdr);
2828         while (hlen >= TCPOLEN_MAXSEG) {
2829                 switch (*opt) {
2830                 case TCPOPT_EOL:
2831                 case TCPOPT_NOP:
2832                         ++opt;
2833                         --hlen;
2834                         break;
2835                 case TCPOPT_MAXSEG:
2836                         bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2);
2837                         /* FALLTHROUGH */
2838                 default:
2839                         optlen = opt[1];
2840                         if (optlen < 2)
2841                                 optlen = 2;
2842                         hlen -= optlen;
2843                         opt += optlen;
2844                         break;
2845                 }
2846         }
2847         return (mss);
2848 }
2849
2850 u_int16_t
2851 pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer)
2852 {
2853 #ifdef INET
2854         struct sockaddr_in      *dst;
2855         struct route             ro;
2856 #endif /* INET */
2857 #ifdef INET6
2858         struct sockaddr_in6     *dst6;
2859         struct route_in6         ro6;
2860 #endif /* INET6 */
2861         struct rtentry          *rt = NULL;
2862         int                      hlen = 0;
2863         u_int16_t                mss = tcp_mssdflt;
2864
2865         switch (af) {
2866 #ifdef INET
2867         case AF_INET:
2868                 hlen = sizeof(struct ip);
2869                 bzero(&ro, sizeof(ro));
2870                 dst = (struct sockaddr_in *)&ro.ro_dst;
2871                 dst->sin_family = AF_INET;
2872                 dst->sin_len = sizeof(*dst);
2873                 dst->sin_addr = addr->v4;
2874                 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
2875                 rt = ro.ro_rt;
2876                 break;
2877 #endif /* INET */
2878 #ifdef INET6
2879         case AF_INET6:
2880                 hlen = sizeof(struct ip6_hdr);
2881                 bzero(&ro6, sizeof(ro6));
2882                 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst;
2883                 dst6->sin6_family = AF_INET6;
2884                 dst6->sin6_len = sizeof(*dst6);
2885                 dst6->sin6_addr = addr->v6;
2886                 rtalloc_ign((struct route *)&ro6, (RTF_CLONING | RTF_PRCLONING));
2887                 rt = ro6.ro_rt;
2888                 break;
2889 #endif /* INET6 */
2890         }
2891
2892         if (rt && rt->rt_ifp) {
2893                 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr);
2894                 mss = max(tcp_mssdflt, mss);
2895                 RTFREE(rt);
2896         }
2897         mss = min(mss, offer);
2898         mss = max(mss, 64);             /* sanity - at least max opt space */
2899         return (mss);
2900 }
2901
2902 void
2903 pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr)
2904 {
2905         struct pf_rule *r = s->rule.ptr;
2906
2907         s->rt_kif = NULL;
2908         if (!r->rt || r->rt == PF_FASTROUTE)
2909                 return;
2910         switch (s->state_key->af) {
2911 #ifdef INET
2912         case AF_INET:
2913                 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL,
2914                     &s->nat_src_node);
2915                 s->rt_kif = r->rpool.cur->kif;
2916                 break;
2917 #endif /* INET */
2918 #ifdef INET6
2919         case AF_INET6:
2920                 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL,
2921                     &s->nat_src_node);
2922                 s->rt_kif = r->rpool.cur->kif;
2923                 break;
2924 #endif /* INET6 */
2925         }
2926 }
2927
2928 void
2929 pf_attach_state(struct pf_state_key *sk, struct pf_state *s, int tail)
2930 {
2931         s->state_key = sk;
2932         sk->refcnt++;
2933
2934         /* list is sorted, if-bound states before floating */
2935         if (tail)
2936                 TAILQ_INSERT_TAIL(&sk->states, s, next);
2937         else
2938                 TAILQ_INSERT_HEAD(&sk->states, s, next);
2939 }
2940
2941 void
2942 pf_detach_state(struct pf_state *s, int flags)
2943 {
2944         struct pf_state_key     *sk = s->state_key;
2945
2946         if (sk == NULL)
2947                 return;
2948
2949         s->state_key = NULL;
2950         TAILQ_REMOVE(&sk->states, s, next);
2951         if (--sk->refcnt == 0) {
2952                 if (!(flags & PF_DT_SKIP_EXTGWY))
2953                         RB_REMOVE(pf_state_tree_ext_gwy,
2954                             &pf_statetbl_ext_gwy, sk);
2955                 if (!(flags & PF_DT_SKIP_LANEXT))
2956                         RB_REMOVE(pf_state_tree_lan_ext,
2957                             &pf_statetbl_lan_ext, sk);
2958                 pool_put(&pf_state_key_pl, sk);
2959         }
2960 }
2961
2962 struct pf_state_key *
2963 pf_alloc_state_key(struct pf_state *s)
2964 {
2965         struct pf_state_key     *sk;
2966
2967         if ((sk = pool_get(&pf_state_key_pl, PR_NOWAIT)) == NULL)
2968                 return (NULL);
2969         bzero(sk, sizeof(*sk));
2970         TAILQ_INIT(&sk->states);
2971         pf_attach_state(sk, s, 0);
2972
2973         return (sk);
2974 }
2975
2976 int
2977 pf_test_rule(struct pf_rule **rm, struct pf_state **sm, int direction,
2978     struct pfi_kif *kif, struct mbuf *m, int off, void *h,
2979     struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
2980     struct ifqueue *ifq, struct inpcb *inp)
2981 {
2982         struct pf_rule          *nr = NULL;
2983         struct pf_addr          *saddr = pd->src, *daddr = pd->dst;
2984         u_int16_t                bport, nport = 0;
2985         sa_family_t              af = pd->af;
2986         struct pf_rule          *r, *a = NULL;
2987         struct pf_ruleset       *ruleset = NULL;
2988         struct pf_src_node      *nsn = NULL;
2989         struct tcphdr           *th = pd->hdr.tcp;
2990         u_short                  reason;
2991         int                      rewrite = 0, hdrlen = 0;
2992         int                      tag = -1, rtableid = -1;
2993         int                      asd = 0;
2994         int                      match = 0;
2995         int                      state_icmp = 0;
2996         u_int16_t                mss = tcp_mssdflt;
2997         u_int16_t                sport, dport;
2998         u_int8_t                 icmptype = 0, icmpcode = 0;
2999
3000         if (direction == PF_IN && pf_check_congestion(ifq)) {
3001                 REASON_SET(&reason, PFRES_CONGEST);
3002                 return (PF_DROP);
3003         }
3004
3005         if (inp != NULL)
3006                 pd->lookup.done = pf_socket_lookup(direction, pd);
3007         else if (debug_pfugidhack) {
3008                 crit_exit();
3009                 DPFPRINTF(PF_DEBUG_MISC, ("pf: unlocked lookup\n"));
3010                 pd->lookup.done = pf_socket_lookup(direction, pd);
3011                 crit_enter();
3012         }
3013
3014         sport = dport = hdrlen = 0;
3015
3016         switch (pd->proto) {
3017         case IPPROTO_TCP:
3018                 sport = th->th_sport;
3019                 dport = th->th_dport;
3020                 hdrlen = sizeof(*th);
3021                 break;
3022         case IPPROTO_UDP:
3023                 sport = pd->hdr.udp->uh_sport;
3024                 dport = pd->hdr.udp->uh_dport;
3025                 hdrlen = sizeof(*pd->hdr.udp);
3026                 break;
3027 #ifdef INET
3028         case IPPROTO_ICMP:
3029                 if (pd->af != AF_INET)
3030                         break;
3031                 sport = dport = pd->hdr.icmp->icmp_id;
3032                 icmptype = pd->hdr.icmp->icmp_type;
3033                 icmpcode = pd->hdr.icmp->icmp_code;
3034
3035                 if (icmptype == ICMP_UNREACH ||
3036                     icmptype == ICMP_SOURCEQUENCH ||
3037                     icmptype == ICMP_REDIRECT ||
3038                     icmptype == ICMP_TIMXCEED ||
3039                     icmptype == ICMP_PARAMPROB)
3040                         state_icmp++;
3041                 break;
3042 #endif /* INET */
3043 #ifdef INET6
3044         case IPPROTO_ICMPV6:
3045                 if (pd->af != AF_INET6)
3046                         break;
3047                 sport = dport = pd->hdr.icmp6->icmp6_id;
3048                 hdrlen = sizeof(*pd->hdr.icmp6);
3049                 icmptype = pd->hdr.icmp6->icmp6_type;
3050                 icmpcode = pd->hdr.icmp6->icmp6_code;
3051
3052                 if (icmptype == ICMP6_DST_UNREACH ||
3053                     icmptype == ICMP6_PACKET_TOO_BIG ||
3054                     icmptype == ICMP6_TIME_EXCEEDED ||
3055                     icmptype == ICMP6_PARAM_PROB)
3056                         state_icmp++;
3057                 break;
3058 #endif /* INET6 */
3059         }
3060
3061         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3062
3063         if (direction == PF_OUT) {
3064                 bport = nport = sport;
3065                 /* check outgoing packet for BINAT/NAT */
3066                 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3067                     saddr, sport, daddr, dport, &pd->naddr, &nport)) != NULL) {
3068                         PF_ACPY(&pd->baddr, saddr, af);
3069                         switch (pd->proto) {
3070                         case IPPROTO_TCP:
3071                                 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
3072                                     &th->th_sum, &pd->naddr, nport, 0, af);
3073                                 sport = th->th_sport;
3074                                 rewrite++;
3075                                 break;
3076                         case IPPROTO_UDP:
3077                                 pf_change_ap(saddr, &pd->hdr.udp->uh_sport,
3078                                     pd->ip_sum, &pd->hdr.udp->uh_sum,
3079                                     &pd->naddr, nport, 1, af);
3080                                 sport = pd->hdr.udp->uh_sport;
3081                                 rewrite++;
3082                                 break;
3083 #ifdef INET
3084                         case IPPROTO_ICMP:
3085                                 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3086                                     pd->naddr.v4.s_addr, 0);
3087                                 pd->hdr.icmp->icmp_cksum = pf_cksum_fixup(
3088                                     pd->hdr.icmp->icmp_cksum, sport, nport, 0);
3089                                 pd->hdr.icmp->icmp_id = nport;
3090                                 m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
3091                                 break;
3092 #endif /* INET */
3093 #ifdef INET6
3094                         case IPPROTO_ICMPV6:
3095                                 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum,
3096                                     &pd->naddr, 0);
3097                                 rewrite++;
3098                                 break;
3099 #endif /* INET */
3100                         default:
3101                                 switch (af) {
3102 #ifdef INET
3103                                 case AF_INET:
3104                                         pf_change_a(&saddr->v4.s_addr,
3105                                             pd->ip_sum, pd->naddr.v4.s_addr, 0);
3106                                         break;
3107 #endif /* INET */
3108 #ifdef INET6
3109                                 case AF_INET6:
3110                                         PF_ACPY(saddr, &pd->naddr, af);
3111                                         break;
3112 #endif /* INET */
3113                                 }
3114                                 break;
3115                         }
3116
3117                         if (nr->natpass)
3118                                 r = NULL;
3119                         pd->nat_rule = nr;
3120                 }
3121         } else {
3122                 bport = nport = dport;
3123                 /* check incoming packet for BINAT/RDR */
3124                 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3125                     saddr, sport, daddr, dport, &pd->naddr, &nport)) != NULL) {
3126                         PF_ACPY(&pd->baddr, daddr, af);
3127                         switch (pd->proto) {
3128                         case IPPROTO_TCP:
3129                                 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
3130                                     &th->th_sum, &pd->naddr, nport, 0, af);
3131                                 dport = th->th_dport;
3132                                 rewrite++;
3133                                 break;
3134                         case IPPROTO_UDP:
3135                                 pf_change_ap(daddr, &pd->hdr.udp->uh_dport,
3136                                     pd->ip_sum, &pd->hdr.udp->uh_sum,
3137                                     &pd->naddr, nport, 1, af);
3138                                 dport = pd->hdr.udp->uh_dport;
3139                                 rewrite++;
3140                                 break;
3141 #ifdef INET
3142                         case IPPROTO_ICMP:
3143                                 pf_change_a(&daddr->v4.s_addr, pd->ip_sum,
3144                                     pd->naddr.v4.s_addr, 0);
3145                                 break;
3146 #endif /* INET */
3147 #ifdef INET6
3148                         case IPPROTO_ICMPV6:
3149                                 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum,
3150                                     &pd->naddr, 0);
3151                                 rewrite++;
3152                                 break;
3153 #endif /* INET6 */
3154                         default:
3155                                 switch (af) {
3156 #ifdef INET
3157                                 case AF_INET:
3158                                         pf_change_a(&daddr->v4.s_addr,
3159                                             pd->ip_sum, pd->naddr.v4.s_addr, 0);
3160                                         break;
3161 #endif /* INET */
3162 #ifdef INET6
3163                                 case AF_INET6:
3164                                         PF_ACPY(daddr, &pd->naddr, af);
3165                                         break;
3166 #endif /* INET */
3167                                 }
3168                                 break;
3169                         }
3170
3171                         if (nr->natpass)
3172                                 r = NULL;
3173                         pd->nat_rule = nr;
3174                 }
3175         }
3176
3177         while (r != NULL) {
3178                 r->evaluations++;
3179                 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3180                         r = r->skip[PF_SKIP_IFP].ptr;
3181                 else if (r->direction && r->direction != direction)
3182                         r = r->skip[PF_SKIP_DIR].ptr;
3183                 else if (r->af && r->af != af)
3184                         r = r->skip[PF_SKIP_AF].ptr;
3185                 else if (r->proto && r->proto != pd->proto)
3186                         r = r->skip[PF_SKIP_PROTO].ptr;
3187                 else if (PF_MISMATCHAW(&r->src.addr, saddr, af,
3188                     r->src.neg, kif))
3189                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3190                 /* tcp/udp only. port_op always 0 in other cases */
3191                 else if (r->src.port_op && !pf_match_port(r->src.port_op,
3192                     r->src.port[0], r->src.port[1], sport))
3193                         r = r->skip[PF_SKIP_SRC_PORT].ptr;
3194                 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af,
3195                     r->dst.neg, NULL))
3196                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
3197                 /* tcp/udp only. port_op always 0 in other cases */
3198                 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
3199                     r->dst.port[0], r->dst.port[1], dport))
3200                         r = r->skip[PF_SKIP_DST_PORT].ptr;
3201                 /* icmp only. type always 0 in other cases */
3202                 else if (r->type && r->type != icmptype + 1)
3203                         r = TAILQ_NEXT(r, entries);
3204                 /* icmp only. type always 0 in other cases */
3205                 else if (r->code && r->code != icmpcode + 1)
3206                         r = TAILQ_NEXT(r, entries);
3207                 else if (r->tos && !(r->tos == pd->tos))
3208                         r = TAILQ_NEXT(r, entries);
3209                 else if (r->rule_flag & PFRULE_FRAGMENT)
3210                         r = TAILQ_NEXT(r, entries);
3211                 else if (pd->proto == IPPROTO_TCP &&
3212                     (r->flagset & th->th_flags) != r->flags)
3213                         r = TAILQ_NEXT(r, entries);
3214                 /* tcp/udp only. uid.op always 0 in other cases */
3215                 else if (r->uid.op && (pd->lookup.done || (pd->lookup.done =
3216                     pf_socket_lookup(direction, pd), 1)) &&
3217                     !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
3218                     pd->lookup.uid))
3219                         r = TAILQ_NEXT(r, entries);
3220                 /* tcp/udp only. gid.op always 0 in other cases */
3221                 else if (r->gid.op && (pd->lookup.done || (pd->lookup.done =
3222                     pf_socket_lookup(direction, pd), 1)) &&
3223                     !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
3224                     pd->lookup.gid))
3225                         r = TAILQ_NEXT(r, entries);
3226                 else if (r->prob && r->prob <= karc4random())
3227                         r = TAILQ_NEXT(r, entries);
3228                 else if (r->match_tag && !pf_match_tag(m, r, &tag))
3229                         r = TAILQ_NEXT(r, entries);
3230                 else if (r->os_fingerprint != PF_OSFP_ANY &&
3231                     (pd->proto != IPPROTO_TCP || !pf_osfp_match(
3232                     pf_osfp_fingerprint(pd, m, off, th),
3233                     r->os_fingerprint)))
3234                         r = TAILQ_NEXT(r, entries);
3235                 else {
3236                         if (r->tag)
3237                                 tag = r->tag;
3238                         if (r->rtableid >= 0)
3239                                 rtableid = r->rtableid;
3240                         if (r->anchor == NULL) {
3241                                 match = 1;
3242                                 *rm = r;
3243                                 *am = a;
3244                                 *rsm = ruleset;
3245                                 if ((*rm)->quick)
3246                                         break;
3247                                 r = TAILQ_NEXT(r, entries);
3248                         } else
3249                                 pf_step_into_anchor(&asd, &ruleset,
3250                                     PF_RULESET_FILTER, &r, &a, &match);
3251                 }
3252                 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3253                     PF_RULESET_FILTER, &r, &a, &match))
3254                         break;
3255         }
3256         r = *rm;
3257         a = *am;
3258         ruleset = *rsm;
3259
3260         REASON_SET(&reason, PFRES_MATCH);
3261
3262         if (r->log || (nr != NULL && nr->log)) {
3263                 if (rewrite)
3264                         m_copyback(m, off, hdrlen, pd->hdr.any);
3265                 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr,
3266                     a, ruleset, pd);
3267         }
3268
3269         if ((r->action == PF_DROP) &&
3270             ((r->rule_flag & PFRULE_RETURNRST) ||
3271             (r->rule_flag & PFRULE_RETURNICMP) ||
3272             (r->rule_flag & PFRULE_RETURN))) {
3273                 /* undo NAT changes, if they have taken place */
3274                 if (nr != NULL) {
3275                         if (direction == PF_OUT) {
3276                                 switch (pd->proto) {
3277                                 case IPPROTO_TCP:
3278                                         pf_change_ap(saddr, &th->th_sport,
3279                                             pd->ip_sum, &th->th_sum,
3280                                             &pd->baddr, bport, 0, af);
3281                                         sport = th->th_sport;
3282                                         rewrite++;
3283                                         break;
3284                                 case IPPROTO_UDP:
3285                                         pf_change_ap(saddr,
3286                                             &pd->hdr.udp->uh_sport, pd->ip_sum,
3287                                             &pd->hdr.udp->uh_sum, &pd->baddr,
3288                                             bport, 1, af);
3289                                         sport = pd->hdr.udp->uh_sport;
3290                                         rewrite++;
3291                                         break;
3292                                 case IPPROTO_ICMP:
3293 #ifdef INET6
3294                                 case IPPROTO_ICMPV6:
3295 #endif
3296                                         /* nothing! */
3297                                         break;
3298                                 default:
3299                                         switch (af) {
3300                                         case AF_INET:
3301                                                 pf_change_a(&saddr->v4.s_addr,
3302                                                     pd->ip_sum,
3303                                                     pd->baddr.v4.s_addr, 0);
3304                                                 break;
3305                                         case AF_INET6:
3306                                                 PF_ACPY(saddr, &pd->baddr, af);
3307                                                 break;
3308                                         }
3309                                 }
3310                         } else {
3311                                 switch (pd->proto) {
3312                                 case IPPROTO_TCP:
3313                                         pf_change_ap(daddr, &th->th_dport,
3314                                             pd->ip_sum, &th->th_sum,
3315                                             &pd->baddr, bport, 0, af);
3316                                         dport = th->th_dport;
3317                                         rewrite++;
3318                                         break;
3319                                 case IPPROTO_UDP:
3320                                         pf_change_ap(daddr,
3321                                             &pd->hdr.udp->uh_dport, pd->ip_sum,
3322                                             &pd->hdr.udp->uh_sum, &pd->baddr,
3323                                             bport, 1, af);
3324                                         dport = pd->hdr.udp->uh_dport;
3325                                         rewrite++;
3326                                         break;
3327                                 case IPPROTO_ICMP:
3328 #ifdef INET6
3329                                 case IPPROTO_ICMPV6:
3330 #endif
3331                                         /* nothing! */
3332                                         break;
3333                                 default:
3334                                         switch (af) {
3335                                         case AF_INET:
3336                                                 pf_change_a(&daddr->v4.s_addr,
3337                                                     pd->ip_sum,
3338                                                     pd->baddr.v4.s_addr, 0);
3339                                                 break;
3340                                         case AF_INET6:
3341                                                 PF_ACPY(daddr, &pd->baddr, af);
3342                                                 break;
3343                                         }
3344                                 }
3345                         }
3346                 }
3347                 if (pd->proto == IPPROTO_TCP &&
3348                     ((r->rule_flag & PFRULE_RETURNRST) ||
3349                     (r->rule_flag & PFRULE_RETURN)) &&
3350                     !(th->th_flags & TH_RST)) {
3351                         u_int32_t        ack = ntohl(th->th_seq) + pd->p_len;
3352                         struct ip       *h = mtod(m, struct ip *);
3353
3354                         if (pf_check_proto_cksum(m, off,
3355                             h->ip_len - off, IPPROTO_TCP, AF_INET))
3356                                 REASON_SET(&reason, PFRES_PROTCKSUM);
3357                         else {
3358                                 if (th->th_flags & TH_SYN)
3359                                         ack++;
3360                                 if (th->th_flags & TH_FIN)
3361                                         ack++;
3362                                 pf_send_tcp(r, af, pd->dst,
3363                                     pd->src, th->th_dport, th->th_sport,
3364                                     ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0,
3365                                     r->return_ttl, 1, 0, pd->eh, kif->pfik_ifp);
3366                         }
3367                 } else if ((af == AF_INET) && r->return_icmp)
3368                         pf_send_icmp(m, r->return_icmp >> 8,
3369                             r->return_icmp & 255, af, r);
3370                 else if ((af == AF_INET6) && r->return_icmp6)
3371                         pf_send_icmp(m, r->return_icmp6 >> 8,
3372                             r->return_icmp6 & 255, af, r);
3373         }
3374
3375         if (r->action == PF_DROP)
3376                 return (PF_DROP);
3377
3378         if (pf_tag_packet(m, tag, rtableid)) {
3379                 REASON_SET(&reason, PFRES_MEMORY);
3380                 return (PF_DROP);
3381         }
3382
3383         if (!state_icmp && (r->keep_state || nr != NULL ||
3384             (pd->flags & PFDESC_TCP_NORM))) {
3385                 /* create new state */
3386                 u_int16_t        len = 0;
3387                 struct pf_state *s = NULL;
3388                 struct pf_state_key *sk = NULL;
3389                 struct pf_src_node *sn = NULL;
3390
3391                 /* check maximums */
3392                 if (r->max_states && (r->states >= r->max_states)) {
3393                         pf_status.lcounters[LCNT_STATES]++;
3394                         REASON_SET(&reason, PFRES_MAXSTATES);
3395                         goto cleanup;
3396                 }
3397                 /* src node for filter rule */
3398                 if ((r->rule_flag & PFRULE_SRCTRACK ||
3399                     r->rpool.opts & PF_POOL_STICKYADDR) &&
3400                     pf_insert_src_node(&sn, r, saddr, af) != 0) {
3401                         REASON_SET(&reason, PFRES_SRCLIMIT);
3402                         goto cleanup;
3403                 }
3404                 /* src node for translation rule */
3405                 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3406                     ((direction == PF_OUT &&
3407                     pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3408                     (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) {
3409                         REASON_SET(&reason, PFRES_SRCLIMIT);
3410                         goto cleanup;
3411                 }
3412                 s = pool_get(&pf_state_pl, PR_NOWAIT);
3413                 if (s == NULL) {
3414                         REASON_SET(&reason, PFRES_MEMORY);
3415 cleanup:
3416                         if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3417                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3418                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3419                                 pf_status.src_nodes--;
3420                                 pool_put(&pf_src_tree_pl, sn);
3421                         }
3422                         if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3423                             nsn->expire == 0) {
3424                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3425                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3426                                 pf_status.src_nodes--;
3427                                 pool_put(&pf_src_tree_pl, nsn);
3428                         }
3429                         if (sk != NULL) {
3430                                 pool_put(&pf_state_key_pl, sk);
3431                         }
3432                         return (PF_DROP);
3433                 }
3434                 bzero(s, sizeof(*s));
3435                 s->rule.ptr = r;
3436                 s->nat_rule.ptr = nr;
3437                 s->anchor.ptr = a;
3438                 STATE_INC_COUNTERS(s);
3439                 s->allow_opts = r->allow_opts;
3440                 s->log = r->log & PF_LOG_ALL;
3441                 if (nr != NULL)
3442                         s->log |= nr->log & PF_LOG_ALL;
3443                 switch (pd->proto) {
3444                 case IPPROTO_TCP:
3445                         len = pd->tot_len - off - (th->th_off << 2);
3446                         s->src.seqlo = ntohl(th->th_seq);
3447                         s->src.seqhi = s->src.seqlo + len + 1;
3448                         if ((th->th_flags & (TH_SYN|TH_ACK)) ==
3449                         TH_SYN && r->keep_state == PF_STATE_MODULATE) {
3450                                 /* Generate sequence number modulator */
3451                                 while ((s->src.seqdiff =
3452                                     pf_new_isn(sk) - s->src.seqlo) == 0)
3453                                         ;
3454                                 pf_change_a(&th->th_seq, &th->th_sum,
3455                                     htonl(s->src.seqlo + s->src.seqdiff), 0);
3456                                 rewrite = 1;
3457                         } else
3458                                 s->src.seqdiff = 0;
3459                         if (th->th_flags & TH_SYN) {
3460                                 s->src.seqhi++;
3461                                 s->src.wscale = pf_get_wscale(m, off,
3462                                     th->th_off, af);
3463                         }
3464                         s->src.max_win = MAX(ntohs(th->th_win), 1);
3465                         if (s->src.wscale & PF_WSCALE_MASK) {
3466                                 /* Remove scale factor from initial window */
3467                                 int win = s->src.max_win;
3468                                 win += 1 << (s->src.wscale & PF_WSCALE_MASK);
3469                                 s->src.max_win = (win - 1) >>
3470                                     (s->src.wscale & PF_WSCALE_MASK);
3471                         }
3472                         if (th->th_flags & TH_FIN)
3473                                 s->src.seqhi++;
3474                         s->dst.seqhi = 1;
3475                         s->dst.max_win = 1;
3476                         s->src.state = TCPS_SYN_SENT;
3477                         s->dst.state = TCPS_CLOSED;
3478                         s->timeout = PFTM_TCP_FIRST_PACKET;
3479                         break;
3480                 case IPPROTO_UDP:
3481                         s->src.state = PFUDPS_SINGLE;
3482                         s->dst.state = PFUDPS_NO_TRAFFIC;
3483                         s->timeout = PFTM_UDP_FIRST_PACKET;
3484                         break;
3485                 case IPPROTO_ICMP:
3486 #ifdef INET6
3487                 case IPPROTO_ICMPV6:
3488 #endif
3489                         s->timeout = PFTM_ICMP_FIRST_PACKET;
3490                         break;
3491                 default:
3492                         s->src.state = PFOTHERS_SINGLE;
3493                         s->dst.state = PFOTHERS_NO_TRAFFIC;
3494                         s->timeout = PFTM_OTHER_FIRST_PACKET;
3495                 }
3496
3497                 s->creation = time_second;
3498                 s->expire = time_second;
3499
3500                 if (sn != NULL) {
3501                         s->src_node = sn;
3502                         s->src_node->states++;
3503                 }
3504                 if (nsn != NULL) {
3505                         PF_ACPY(&nsn->raddr, &pd->naddr, af);
3506                         s->nat_src_node = nsn;
3507                         s->nat_src_node->states++;
3508                 }
3509                 if (pd->proto == IPPROTO_TCP) {
3510                         if ((pd->flags & PFDESC_TCP_NORM) &&
3511                             pf_normalize_tcp_init(m, off, pd, th, &s->src,
3512                             &s->dst)) {
3513                                 REASON_SET(&reason, PFRES_MEMORY);
3514                                 pf_src_tree_remove_state(s);
3515                                 STATE_DEC_COUNTERS(s);
3516                                 pool_put(&pf_state_pl, s);
3517                                 return (PF_DROP);
3518                         }
3519                         if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
3520                             pf_normalize_tcp_stateful(m, off, pd, &reason,
3521                             th, s, &s->src, &s->dst, &rewrite)) {
3522                                 /* This really shouldn't happen!!! */
3523                                 DPFPRINTF(PF_DEBUG_URGENT,
3524                                     ("pf_normalize_tcp_stateful failed on "
3525                                     "first pkt"));
3526                                 pf_normalize_tcp_cleanup(s);
3527                                 pf_src_tree_remove_state(s);
3528                                 STATE_DEC_COUNTERS(s);
3529                                 pool_put(&pf_state_pl, s);
3530                                 return (PF_DROP);
3531                         }
3532                 }
3533
3534                 if ((sk = pf_alloc_state_key(s)) == NULL) {
3535                         REASON_SET(&reason, PFRES_MEMORY);
3536                         goto cleanup;
3537                 }
3538
3539                 sk->proto = pd->proto;
3540                 sk->direction = direction;
3541                 sk->af = af;
3542                 if (direction == PF_OUT) {
3543                         PF_ACPY(&sk->gwy.addr, saddr, af);
3544                         PF_ACPY(&sk->ext.addr, daddr, af);
3545                         switch (pd->proto) {
3546                         case IPPROTO_ICMP:
3547 #ifdef INET6
3548                         case IPPROTO_ICMPV6:
3549 #endif
3550                                 sk->gwy.port = nport;
3551                                 sk->ext.port = 0;
3552                                 break;
3553                         default:
3554                                 sk->gwy.port = sport;
3555                                 sk->ext.port = dport;
3556                         }
3557                         if (nr != NULL) {
3558                                 PF_ACPY(&sk->lan.addr, &pd->baddr, af);
3559                                 sk->lan.port = bport;
3560                         } else {
3561                                 PF_ACPY(&sk->lan.addr, &sk->gwy.addr, af);
3562                                 sk->lan.port = sk->gwy.port;
3563                         }
3564                 } else {
3565                         PF_ACPY(&sk->lan.addr, daddr, af);
3566                         PF_ACPY(&sk->ext.addr, saddr, af);
3567                         switch (pd->proto) {
3568                         case IPPROTO_ICMP:
3569 #ifdef INET6
3570                         case IPPROTO_ICMPV6:
3571 #endif
3572                                 sk->lan.port = nport;
3573                                 sk->ext.port = 0;
3574                                 break;
3575                         default:
3576                                 sk->lan.port = dport;
3577                                 sk->ext.port = sport;
3578                         }
3579                         if (nr != NULL) {
3580                                 PF_ACPY(&sk->gwy.addr, &pd->baddr, af);
3581                                 sk->gwy.port = bport;
3582                         } else {
3583                                 PF_ACPY(&sk->gwy.addr, &sk->lan.addr, af);
3584                                 sk->gwy.port = sk->lan.port;
3585                         }
3586                 }
3587
3588                 s->hash = pf_state_hash(sk);
3589                 s->pickup_mode = r->pickup_mode;
3590
3591                 pf_set_rt_ifp(s, saddr);        /* needs s->state_key set */
3592
3593                 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3594                         if (pd->proto == IPPROTO_TCP)
3595                                 pf_normalize_tcp_cleanup(s);
3596                         REASON_SET(&reason, PFRES_STATEINS);
3597                         pf_src_tree_remove_state(s);
3598                         STATE_DEC_COUNTERS(s);
3599                         pool_put(&pf_state_pl, s);
3600                         return (PF_DROP);
3601                 } else
3602                         *sm = s;
3603                 if (tag > 0) {
3604                         pf_tag_ref(tag);
3605                         s->tag = tag;
3606                 }
3607                 if (pd->proto == IPPROTO_TCP &&
3608                     (th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
3609                     r->keep_state == PF_STATE_SYNPROXY) {
3610                         s->src.state = PF_TCPS_PROXY_SRC;
3611                         if (nr != NULL) {
3612                                 if (direction == PF_OUT) {
3613                                         pf_change_ap(saddr, &th->th_sport,
3614                                             pd->ip_sum, &th->th_sum, &pd->baddr,
3615                                             bport, 0, af);
3616                                         sport = th->th_sport;
3617                                 } else {
3618                                         pf_change_ap(daddr, &th->th_dport,
3619                                             pd->ip_sum, &th->th_sum, &pd->baddr,
3620                                             bport, 0, af);
3621                                         sport = th->th_dport;
3622                                 }
3623                         }
3624                         s->src.seqhi = htonl(karc4random());
3625                         /* Find mss option */
3626                         mss = pf_get_mss(m, off, th->th_off, af);
3627                         mss = pf_calc_mss(saddr, af, mss);
3628                         mss = pf_calc_mss(daddr, af, mss);
3629                         s->src.mss = mss;
3630                         pf_send_tcp(r, af, daddr, saddr, th->th_dport,
3631                             th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1,
3632                             TH_SYN|TH_ACK, 0, s->src.mss, 0, 1, 0, NULL, NULL);
3633                         REASON_SET(&reason, PFRES_SYNPROXY);
3634                         return (PF_SYNPROXY_DROP);
3635                 }
3636         }
3637
3638         /* copy back packet headers if we performed NAT operations */
3639         if (rewrite)
3640                 m_copyback(m, off, hdrlen, pd->hdr.any);
3641
3642         return (PF_PASS);
3643 }
3644
3645 int
3646 pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif,
3647     struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am,
3648     struct pf_ruleset **rsm)
3649 {
3650         struct pf_rule          *r, *a = NULL;
3651         struct pf_ruleset       *ruleset = NULL;
3652         sa_family_t              af = pd->af;
3653         u_short                  reason;
3654         int                      tag = -1;
3655         int                      asd = 0;
3656         int                      match = 0;
3657
3658         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3659         while (r != NULL) {
3660                 r->evaluations++;
3661                 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3662                         r = r->skip[PF_SKIP_IFP].ptr;
3663                 else if (r->direction && r->direction != direction)
3664                         r = r->skip[PF_SKIP_DIR].ptr;
3665                 else if (r->af && r->af != af)
3666                         r = r->skip[PF_SKIP_AF].ptr;
3667                 else if (r->proto && r->proto != pd->proto)
3668                         r = r->skip[PF_SKIP_PROTO].ptr;
3669                 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af,
3670                     r->src.neg, kif))
3671                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3672                 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af,
3673                     r->dst.neg, NULL))
3674                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
3675                 else if (r->tos && !(r->tos == pd->tos))
3676                         r = TAILQ_NEXT(r, entries);
3677                 else if (r->os_fingerprint != PF_OSFP_ANY)
3678                         r = TAILQ_NEXT(r, entries);
3679                 else if (pd->proto == IPPROTO_UDP &&
3680                     (r->src.port_op || r->dst.port_op))
3681                         r = TAILQ_NEXT(r, entries);
3682                 else if (pd->proto == IPPROTO_TCP &&
3683                     (r->src.port_op || r->dst.port_op || r->flagset))
3684                         r = TAILQ_NEXT(r, entries);
3685                 else if ((pd->proto == IPPROTO_ICMP ||
3686                     pd->proto == IPPROTO_ICMPV6) &&
3687                     (r->type || r->code))
3688                         r = TAILQ_NEXT(r, entries);
3689                 else if (r->prob && r->prob <= karc4random())
3690                         r = TAILQ_NEXT(r, entries);
3691                 else if (r->match_tag && !pf_match_tag(m, r, &tag))
3692                         r = TAILQ_NEXT(r, entries);
3693                 else {
3694                         if (r->anchor == NULL) {
3695                                 match = 1;
3696                                 *rm = r;
3697                                 *am = a;
3698                                 *rsm = ruleset;
3699                                 if ((*rm)->quick)
3700                                         break;
3701                                 r = TAILQ_NEXT(r, entries);
3702                         } else
3703                                 pf_step_into_anchor(&asd, &ruleset,
3704                                     PF_RULESET_FILTER, &r, &a, &match);
3705                 }
3706                 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3707                     PF_RULESET_FILTER, &r, &a, &match))
3708                         break;
3709         }
3710         r = *rm;
3711         a = *am;
3712         ruleset = *rsm;
3713
3714         REASON_SET(&reason, PFRES_MATCH);
3715
3716         if (r->log)
3717                 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset,
3718                     pd);
3719
3720         if (r->action != PF_PASS)
3721                 return (PF_DROP);
3722
3723         if (pf_tag_packet(m, tag, -1)) {
3724                 REASON_SET(&reason, PFRES_MEMORY);
3725                 return (PF_DROP);
3726         }
3727
3728         return (PF_PASS);
3729 }
3730
3731 int
3732 pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
3733     struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
3734     u_short *reason)
3735 {
3736         struct pf_state_key_cmp  key;
3737         struct tcphdr           *th = pd->hdr.tcp;
3738         u_int16_t                win = ntohs(th->th_win);
3739         u_int32_t                ack, end, seq, orig_seq;
3740         u_int8_t                 sws, dws;
3741         int                      ackskew;
3742         int                      copyback = 0;
3743         struct pf_state_peer    *src, *dst;
3744
3745         key.af = pd->af;
3746         key.proto = IPPROTO_TCP;
3747         if (direction == PF_IN) {
3748                 PF_ACPY(&key.ext.addr, pd->src, key.af);
3749                 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
3750                 key.ext.port = th->th_sport;
3751                 key.gwy.port = th->th_dport;
3752         } else {
3753                 PF_ACPY(&key.lan.addr, pd->src, key.af);
3754                 PF_ACPY(&key.ext.addr, pd->dst, key.af);
3755                 key.lan.port = th->th_sport;
3756                 key.ext.port = th->th_dport;
3757         }
3758
3759         STATE_LOOKUP();
3760
3761         if (direction == (*state)->state_key->direction) {
3762                 src = &(*state)->src;
3763                 dst = &(*state)->dst;
3764         } else {
3765                 src = &(*state)->dst;
3766                 dst = &(*state)->src;
3767         }
3768
3769         if ((*state)->src.state == PF_TCPS_PROXY_SRC) {
3770                 if (direction != (*state)->state_key->direction) {
3771                         REASON_SET(reason, PFRES_SYNPROXY);
3772                         return (PF_SYNPROXY_DROP);
3773                 }
3774                 if (th->th_flags & TH_SYN) {
3775                         if (ntohl(th->th_seq) != (*state)->src.seqlo) {
3776                                 REASON_SET(reason, PFRES_SYNPROXY);
3777                                 return (PF_DROP);
3778                         }
3779                         pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
3780                             pd->src, th->th_dport, th->th_sport,
3781                             (*state)->src.seqhi, ntohl(th->th_seq) + 1,
3782                             TH_SYN|TH_ACK, 0, (*state)->src.mss, 0, 1,
3783                             0, NULL, NULL);
3784                         REASON_SET(reason, PFRES_SYNPROXY);
3785                         return (PF_SYNPROXY_DROP);
3786                 } else if (!(th->th_flags & TH_ACK) ||
3787                     (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
3788                     (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
3789                         REASON_SET(reason, PFRES_SYNPROXY);
3790                         return (PF_DROP);
3791                 } else if ((*state)->src_node != NULL &&
3792                     pf_src_connlimit(state)) {
3793                         REASON_SET(reason, PFRES_SRCLIMIT);
3794                         return (PF_DROP);
3795                 } else
3796                         (*state)->src.state = PF_TCPS_PROXY_DST;
3797         }
3798         if ((*state)->src.state == PF_TCPS_PROXY_DST) {
3799                 struct pf_state_host *src, *dst;
3800
3801                 if (direction == PF_OUT) {
3802                         src = &(*state)->state_key->gwy;
3803                         dst = &(*state)->state_key->ext;
3804                 } else {
3805                         src = &(*state)->state_key->ext;
3806                         dst = &(*state)->state_key->lan;
3807                 }
3808                 if (direction == (*state)->state_key->direction) {
3809                         if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) ||
3810                             (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
3811                             (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
3812                                 REASON_SET(reason, PFRES_SYNPROXY);
3813                                 return (PF_DROP);
3814                         }
3815                         (*state)->src.max_win = MAX(ntohs(th->th_win), 1);
3816                         if ((*state)->dst.seqhi == 1)
3817                                 (*state)->dst.seqhi = htonl(karc4random());
3818                         pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
3819                             &dst->addr, src->port, dst->port,
3820                             (*state)->dst.seqhi, 0, TH_SYN, 0,
3821                             (*state)->src.mss, 0, 0, (*state)->tag, NULL, NULL);
3822                         REASON_SET(reason, PFRES_SYNPROXY);
3823                         return (PF_SYNPROXY_DROP);
3824                 } else if (((th->th_flags & (TH_SYN|TH_ACK)) !=
3825                     (TH_SYN|TH_ACK)) ||
3826                     (ntohl(th->th_ack) != (*state)->dst.seqhi + 1)) {
3827                         REASON_SET(reason, PFRES_SYNPROXY);
3828                         return (PF_DROP);
3829                 } else {
3830                         (*state)->dst.max_win = MAX(ntohs(th->th_win), 1);
3831                         (*state)->dst.seqlo = ntohl(th->th_seq);
3832                         pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
3833                             pd->src, th->th_dport, th->th_sport,
3834                             ntohl(th->th_ack), ntohl(th->th_seq) + 1,
3835                             TH_ACK, (*state)->src.max_win, 0, 0, 0,
3836                             (*state)->tag, NULL, NULL);
3837                         pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
3838                             &dst->addr, src->port, dst->port,
3839                             (*state)->src.seqhi + 1, (*state)->src.seqlo + 1,
3840                             TH_ACK, (*state)->dst.max_win, 0, 0, 1,
3841                             0, NULL, NULL);
3842                         (*state)->src.seqdiff = (*state)->dst.seqhi -
3843                             (*state)->src.seqlo;
3844                         (*state)->dst.seqdiff = (*state)->src.seqhi -
3845                             (*state)->dst.seqlo;
3846                         (*state)->src.seqhi = (*state)->src.seqlo +
3847                             (*state)->dst.max_win;
3848                         (*state)->dst.seqhi = (*state)->dst.seqlo +
3849                             (*state)->src.max_win;
3850                         (*state)->src.wscale = (*state)->dst.wscale = 0;
3851                         (*state)->src.state = (*state)->dst.state =
3852                             TCPS_ESTABLISHED;
3853                         REASON_SET(reason, PFRES_SYNPROXY);
3854                         return (PF_SYNPROXY_DROP);
3855                 }
3856         }
3857
3858         if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) {
3859                 sws = src->wscale & PF_WSCALE_MASK;
3860                 dws = dst->wscale & PF_WSCALE_MASK;
3861         } else
3862                 sws = dws = 0;
3863
3864         /*
3865          * Sequence tracking algorithm from Guido van Rooij's paper:
3866          *   http://www.madison-gurkha.com/publications/tcp_filtering/
3867          *      tcp_filtering.ps
3868          */
3869
3870         orig_seq = seq = ntohl(th->th_seq);
3871         if (src->seqlo == 0) {
3872                 /* First packet from this end. Set its state */
3873
3874                 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) &&
3875                     src->scrub == NULL) {
3876                         if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) {
3877                                 REASON_SET(reason, PFRES_MEMORY);
3878                                 return (PF_DROP);
3879                         }
3880                 }
3881
3882                 /* Deferred generation of sequence number modulator */
3883                 if (dst->seqdiff && !src->seqdiff) {
3884
3885                         while ((src->seqdiff = pf_new_isn((struct pf_state_key *)&key) - seq) == 0)
3886                                 ;
3887                         ack = ntohl(th->th_ack) - dst->seqdiff;
3888                         pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3889                             src->seqdiff), 0);
3890                         pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3891                         copyback = 1;
3892                 } else {
3893                         ack = ntohl(th->th_ack);
3894                 }
3895
3896                 end = seq + pd->p_len;
3897                 if (th->th_flags & TH_SYN) {
3898                         end++;
3899                         (*state)->sync_flags |= PFSTATE_GOT_SYN2;
3900                         if (dst->wscale & PF_WSCALE_FLAG) {
3901                                 src->wscale = pf_get_wscale(m, off, th->th_off,
3902                                     pd->af);
3903                                 if (src->wscale & PF_WSCALE_FLAG) {
3904                                         /* Remove scale factor from initial
3905                                          * window */
3906                                         sws = src->wscale & PF_WSCALE_MASK;
3907                                         win = ((u_int32_t)win + (1 << sws) - 1)
3908                                             >> sws;
3909                                         dws = dst->wscale & PF_WSCALE_MASK;
3910                                 } else {
3911                                         /* fixup other window */
3912                                         dst->max_win <<= dst->wscale &
3913                                             PF_WSCALE_MASK;
3914                                         /* in case of a retrans SYN|ACK */
3915                                         dst->wscale = 0;
3916                                 }
3917                         }
3918                 }
3919                 if (th->th_flags & TH_FIN)
3920                         end++;
3921
3922                 src->seqlo = seq;
3923                 if (src->state < TCPS_SYN_SENT)
3924                         src->state = TCPS_SYN_SENT;
3925
3926                 /*
3927                  * May need to slide the window (seqhi may have been set by
3928                  * the crappy stack check or if we picked up the connection
3929                  * after establishment)
3930                  */
3931                 if (src->seqhi == 1 ||
3932                     SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi))
3933                         src->seqhi = end + MAX(1, dst->max_win << dws);
3934                 if (win > src->max_win)
3935                         src->max_win = win;
3936
3937         } else {
3938                 ack = ntohl(th->th_ack) - dst->seqdiff;
3939                 if (src->seqdiff) {
3940                         /* Modulate sequence numbers */
3941                         pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3942                             src->seqdiff), 0);
3943                         pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3944                         copyback = 1;
3945                 }
3946                 end = seq + pd->p_len;
3947                 if (th->th_flags & TH_SYN)
3948                         end++;
3949                 if (th->th_flags & TH_FIN)
3950                         end++;
3951         }
3952
3953         if ((th->th_flags & TH_ACK) == 0) {
3954                 /* Let it pass through the ack skew check */
3955                 ack = dst->seqlo;
3956         } else if ((ack == 0 &&
3957             (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) ||
3958             /* broken tcp stacks do not set ack */
3959             (dst->state < TCPS_SYN_SENT)) {
3960                 /*
3961                  * Many stacks (ours included) will set the ACK number in an
3962                  * FIN|ACK if the SYN times out -- no sequence to ACK.
3963                  */
3964                 ack = dst->seqlo;
3965         }
3966
3967         if (seq == end) {
3968                 /* Ease sequencing restrictions on no data packets */
3969                 seq = src->seqlo;
3970                 end = seq;
3971         }
3972
3973         ackskew = dst->seqlo - ack;
3974
3975
3976         /*
3977          * Need to demodulate the sequence numbers in any TCP SACK options
3978          * (Selective ACK). We could optionally validate the SACK values
3979          * against the current ACK window, either forwards or backwards, but
3980          * I'm not confident that SACK has been implemented properly
3981          * everywhere. It wouldn't surprise me if several stacks accidently
3982          * SACK too far backwards of previously ACKed data. There really aren't
3983          * any security implications of bad SACKing unless the target stack
3984          * doesn't validate the option length correctly. Someone trying to
3985          * spoof into a TCP connection won't bother blindly sending SACK
3986          * options anyway.
3987          */
3988         if (dst->seqdiff && (th->th_off << 2) > sizeof(struct tcphdr)) {
3989                 if (pf_modulate_sack(m, off, pd, th, dst))
3990                         copyback = 1;
3991         }
3992
3993
3994 #define MAXACKWINDOW (0xffff + 1500)    /* 1500 is an arbitrary fudge factor */
3995         if (SEQ_GEQ(src->seqhi, end) &&
3996             /* Last octet inside other's window space */
3997             SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) &&
3998             /* Retrans: not more than one window back */
3999             (ackskew >= -MAXACKWINDOW) &&
4000             /* Acking not more than one reassembled fragment backwards */
4001             (ackskew <= (MAXACKWINDOW << sws)) &&
4002             /* Acking not more than one window forward */
4003             ((th->th_flags & TH_RST) == 0 || orig_seq == src->seqlo ||
4004             (orig_seq == src->seqlo + 1) || (orig_seq + 1 == src->seqlo) ||
4005             (pd->flags & PFDESC_IP_REAS) == 0)) {
4006             /* Require an exact/+1 sequence match on resets when possible */
4007
4008                 if (dst->scrub || src->scrub) {
4009                         if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4010                             *state, src, dst, &copyback))
4011                                 return (PF_DROP);
4012                 }
4013
4014                 /* update max window */
4015                 if (src->max_win < win)
4016                         src->max_win = win;
4017                 /* synchronize sequencing */
4018                 if (SEQ_GT(end, src->seqlo))
4019                         src->seqlo = end;
4020                 /* slide the window of what the other end can send */
4021                 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4022                         dst->seqhi = ack + MAX((win << sws), 1);
4023
4024
4025                 /* update states */
4026                 if (th->th_flags & TH_SYN)
4027                         if (src->state < TCPS_SYN_SENT)
4028                                 src->state = TCPS_SYN_SENT;
4029                 if (th->th_flags & TH_FIN)
4030                         if (src->state < TCPS_CLOSING)
4031                                 src->state = TCPS_CLOSING;
4032                 if (th->th_flags & TH_ACK) {
4033                         if (dst->state == TCPS_SYN_SENT) {
4034                                 dst->state = TCPS_ESTABLISHED;
4035                                 if (src->state == TCPS_ESTABLISHED &&
4036                                     (*state)->src_node != NULL &&
4037                                     pf_src_connlimit(state)) {
4038                                         REASON_SET(reason, PFRES_SRCLIMIT);
4039                                         return (PF_DROP);
4040                                 }
4041                         } else if (dst->state == TCPS_CLOSING)
4042                                 dst->state = TCPS_FIN_WAIT_2;
4043                 }
4044                 if (th->th_flags & TH_RST)
4045                         src->state = dst->state = TCPS_TIME_WAIT;
4046
4047                 /* update expire time */
4048                 (*state)->expire = time_second;
4049                 if (src->state >= TCPS_FIN_WAIT_2 &&
4050                     dst->state >= TCPS_FIN_WAIT_2)
4051                         (*state)->timeout = PFTM_TCP_CLOSED;
4052                 else if (src->state >= TCPS_CLOSING &&
4053                     dst->state >= TCPS_CLOSING)
4054                         (*state)->timeout = PFTM_TCP_FIN_WAIT;
4055                 else if (src->state < TCPS_ESTABLISHED ||
4056                     dst->state < TCPS_ESTABLISHED)
4057                         (*state)->timeout = PFTM_TCP_OPENING;
4058                 else if (src->state >= TCPS_CLOSING ||
4059                     dst->state >= TCPS_CLOSING)
4060                         (*state)->timeout = PFTM_TCP_CLOSING;
4061                 else
4062                         (*state)->timeout = PFTM_TCP_ESTABLISHED;
4063
4064                 /* Fall through to PASS packet */
4065
4066         } else if ((dst->state < TCPS_SYN_SENT ||
4067                 dst->state >= TCPS_FIN_WAIT_2 ||
4068                 src->state >= TCPS_FIN_WAIT_2) &&
4069             SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) &&
4070             /* Within a window forward of the originating packet */
4071             SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) {
4072             /* Within a window backward of the originating packet */
4073
4074                 /*
4075                  * This currently handles three situations:
4076                  *  1) Stupid stacks will shotgun SYNs before their peer
4077                  *     replies.
4078                  *  2) When PF catches an already established stream (the
4079                  *     firewall rebooted, the state table was flushed, routes
4080                  *     changed...)
4081                  *  3) Packets get funky immediately after the connection
4082                  *     closes (this should catch Solaris spurious ACK|FINs
4083                  *     that web servers like to spew after a close)
4084                  *
4085                  * This must be a little more careful than the above code
4086                  * since packet floods will also be caught here. We don't
4087                  * update the TTL here to mitigate the damage of a packet
4088                  * flood and so the same code can handle awkward establishment
4089                  * and a loosened connection close.
4090                  * In the establishment case, a correct peer response will
4091                  * validate the connection, go through the normal state code
4092                  * and keep updating the state TTL.
4093                  */
4094
4095                 if (pf_status.debug >= PF_DEBUG_MISC) {
4096                         kprintf("pf: loose state match: ");
4097                         pf_print_state(*state);
4098                         pf_print_flags(th->th_flags);
4099                         kprintf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
4100                             "pkts=%llu:%llu dir=%s,%s\n", seq, orig_seq, ack, pd->p_len,
4101                             ackskew, (unsigned long long)(*state)->packets[0],
4102                             (unsigned long long)(*state)->packets[1],
4103                             direction == PF_IN ? "in" : "out",
4104                             direction == (*state)->state_key->direction ?
4105                                 "fwd" : "rev");
4106                 }
4107
4108                 if (dst->scrub || src->scrub) {
4109                         if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4110                             *state, src, dst, &copyback))
4111                                 return (PF_DROP);
4112                 }
4113
4114                 /* update max window */
4115                 if (src->max_win < win)
4116                         src->max_win = win;
4117                 /* synchronize sequencing */
4118                 if (SEQ_GT(end, src->seqlo))
4119                         src->seqlo = end;
4120                 /* slide the window of what the other end can send */
4121                 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4122                         dst->seqhi = ack + MAX((win << sws), 1);
4123
4124                 /*
4125                  * Cannot set dst->seqhi here since this could be a shotgunned
4126                  * SYN and not an already established connection.
4127                  */
4128
4129                 if (th->th_flags & TH_FIN)
4130                         if (src->state < TCPS_CLOSING)
4131                                 src->state = TCPS_CLOSING;
4132                 if (th->th_flags & TH_RST)
4133                         src->state = dst->state = TCPS_TIME_WAIT;
4134
4135                 /* Fall through to PASS packet */
4136
4137         } else if ((*state)->pickup_mode == PF_PICKUPS_HASHONLY ||
4138                     ((*state)->pickup_mode == PF_PICKUPS_ENABLED &&
4139                      ((*state)->sync_flags & PFSTATE_GOT_SYN_MASK) !=
4140                       PFSTATE_GOT_SYN_MASK)) {
4141                 /*
4142                  * If pickup mode is hash only, do not fail on sequence checks.
4143                  *
4144                  * If pickup mode is enabled and we did not see the SYN in
4145                  * both direction, do not fail on sequence checks because
4146                  * we do not have complete information on window scale.
4147                  *
4148                  * Adjust expiration and fall through to PASS packet.
4149                  * XXX Add a FIN check to reduce timeout?
4150                  */
4151                 (*state)->expire = time_second;
4152         } else  {
4153                 /*
4154                  * Failure processing
4155                  */
4156                 if ((*state)->dst.state == TCPS_SYN_SENT &&
4157                     (*state)->src.state == TCPS_SYN_SENT) {
4158                         /* Send RST for state mismatches during handshake */
4159                         if (!(th->th_flags & TH_RST))
4160                                 pf_send_tcp((*state)->rule.ptr, pd->af,
4161                                     pd->dst, pd->src, th->th_dport,
4162                                     th->th_sport, ntohl(th->th_ack), 0,
4163                                     TH_RST, 0, 0,
4164                                     (*state)->rule.ptr->return_ttl, 1, 0,
4165                                     pd->eh, kif->pfik_ifp);
4166                         src->seqlo = 0;
4167                         src->seqhi = 1;
4168                         src->max_win = 1;
4169                 } else if (pf_status.debug >= PF_DEBUG_MISC) {
4170                         kprintf("pf: BAD state: ");
4171                         pf_print_state(*state);
4172                         pf_print_flags(th->th_flags);
4173                         kprintf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
4174                             "pkts=%llu:%llu dir=%s,%s\n",
4175                             seq, orig_seq, ack, pd->p_len, ackskew,
4176                             (unsigned long long)(*state)->packets[0],
4177                                 (unsigned long long)(*state)->packets[1],
4178                             direction == PF_IN ? "in" : "out",
4179                             direction == (*state)->state_key->direction ?
4180                                 "fwd" : "rev");
4181                         kprintf("pf: State failure on: %c %c %c %c | %c %c\n",
4182                             SEQ_GEQ(src->seqhi, end) ? ' ' : '1',
4183                             SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ?
4184                             ' ': '2',
4185                             (ackskew >= -MAXACKWINDOW) ? ' ' : '3',
4186                             (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4',
4187                             SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5',
4188                             SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6');
4189                 }
4190                 REASON_SET(reason, PFRES_BADSTATE);
4191                 return (PF_DROP);
4192         }
4193
4194         /* Any packets which have gotten here are to be passed */
4195
4196         /* translate source/destination address, if necessary */
4197         if (STATE_TRANSLATE((*state)->state_key)) {
4198                 if (direction == PF_OUT) {
4199                         pf_change_ap(pd->src, &th->th_sport, pd->ip_sum,
4200                             &th->th_sum, &(*state)->state_key->gwy.addr,
4201                             (*state)->state_key->gwy.port, 0, pd->af);
4202                 } else {
4203                         /*
4204                          * If we don't redispatch the packet will go into
4205                          * the protocol stack on the wrong cpu for the
4206                          * post-translated address.
4207                          */
4208                         m->m_pkthdr.fw_flags |= FW_MBUF_REDISPATCH;
4209                         pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum,
4210                             &th->th_sum, &(*state)->state_key->lan.addr,
4211                             (*state)->state_key->lan.port, 0, pd->af);
4212                 }
4213                 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4214         } else if (copyback) {
4215                 /* Copyback sequence modulation or stateful scrub changes */
4216                 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4217         }
4218
4219         return (PF_PASS);
4220 }
4221
4222 int
4223 pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif,
4224     struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4225 {
4226         struct pf_state_peer    *src, *dst;
4227         struct pf_state_key_cmp  key;
4228         struct udphdr           *uh = pd->hdr.udp;
4229
4230         key.af = pd->af;
4231         key.proto = IPPROTO_UDP;
4232         if (direction == PF_IN) {
4233                 PF_ACPY(&key.ext.addr, pd->src, key.af);
4234                 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4235                 key.ext.port = uh->uh_sport;
4236                 key.gwy.port = uh->uh_dport;
4237         } else {
4238                 PF_ACPY(&key.lan.addr, pd->src, key.af);
4239                 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4240                 key.lan.port = uh->uh_sport;
4241                 key.ext.port = uh->uh_dport;
4242         }
4243
4244         STATE_LOOKUP();
4245
4246         if (direction == (*state)->state_key->direction) {
4247                 src = &(*state)->src;
4248                 dst = &(*state)->dst;
4249         } else {
4250                 src = &(*state)->dst;
4251                 dst = &(*state)->src;
4252         }
4253
4254         /* update states */
4255         if (src->state < PFUDPS_SINGLE)
4256                 src->state = PFUDPS_SINGLE;
4257         if (dst->state == PFUDPS_SINGLE)
4258                 dst->state = PFUDPS_MULTIPLE;
4259
4260         /* update expire time */
4261         (*state)->expire = time_second;
4262         if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE)
4263                 (*state)->timeout = PFTM_UDP_MULTIPLE;
4264         else
4265                 (*state)->timeout = PFTM_UDP_SINGLE;
4266
4267         /* translate source/destination address, if necessary */
4268         if (STATE_TRANSLATE((*state)->state_key)) {
4269                 if (direction == PF_OUT) {
4270                         pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum,
4271                             &uh->uh_sum, &(*state)->state_key->gwy.addr,
4272                             (*state)->state_key->gwy.port, 1, pd->af);
4273                 } else {
4274                         /*
4275                          * If we don't redispatch the packet will go into
4276                          * the protocol stack on the wrong cpu for the
4277                          * post-translated address.
4278                          */
4279                         m->m_pkthdr.fw_flags |= FW_MBUF_REDISPATCH;
4280                         pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum,
4281                             &uh->uh_sum, &(*state)->state_key->lan.addr,
4282                             (*state)->state_key->lan.port, 1, pd->af);
4283                 }
4284                 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
4285         }
4286
4287         return (PF_PASS);
4288 }
4289
4290 int
4291 pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif,
4292     struct mbuf *m, int off, void *h, struct pf_pdesc *pd, u_short *reason)
4293 {
4294         struct pf_addr  *saddr = pd->src, *daddr = pd->dst;
4295         u_int16_t        icmpid = 0, *icmpsum;
4296         u_int8_t         icmptype;
4297         int              state_icmp = 0;
4298         struct pf_state_key_cmp key;
4299
4300         switch (pd->proto) {
4301 #ifdef INET
4302         case IPPROTO_ICMP:
4303                 icmptype = pd->hdr.icmp->icmp_type;
4304                 icmpid = pd->hdr.icmp->icmp_id;
4305                 icmpsum = &pd->hdr.icmp->icmp_cksum;
4306
4307                 if (icmptype == ICMP_UNREACH ||
4308                     icmptype == ICMP_SOURCEQUENCH ||
4309                     icmptype == ICMP_REDIRECT ||
4310                     icmptype == ICMP_TIMXCEED ||
4311                     icmptype == ICMP_PARAMPROB)
4312                         state_icmp++;
4313                 break;
4314 #endif /* INET */
4315 #ifdef INET6
4316         case IPPROTO_ICMPV6:
4317                 icmptype = pd->hdr.icmp6->icmp6_type;
4318                 icmpid = pd->hdr.icmp6->icmp6_id;
4319                 icmpsum = &pd->hdr.icmp6->icmp6_cksum;
4320
4321                 if (icmptype == ICMP6_DST_UNREACH ||
4322                     icmptype == ICMP6_PACKET_TOO_BIG ||
4323                     icmptype == ICMP6_TIME_EXCEEDED ||
4324                     icmptype == ICMP6_PARAM_PROB)
4325                         state_icmp++;
4326                 break;
4327 #endif /* INET6 */
4328         }
4329
4330         if (!state_icmp) {
4331
4332                 /*
4333                  * ICMP query/reply message not related to a TCP/UDP packet.
4334                  * Search for an ICMP state.
4335                  */
4336                 key.af = pd->af;
4337                 key.proto = pd->proto;
4338                 if (direction == PF_IN) {
4339                         PF_ACPY(&key.ext.addr, pd->src, key.af);
4340                         PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4341                         key.ext.port = 0;
4342                         key.gwy.port = icmpid;
4343                 } else {
4344                         PF_ACPY(&key.lan.addr, pd->src, key.af);
4345                         PF_ACPY(&key.ext.addr, pd->dst, key.af);
4346                         key.lan.port = icmpid;
4347                         key.ext.port = 0;
4348                 }
4349
4350                 STATE_LOOKUP();
4351
4352                 (*state)->expire = time_second;
4353                 (*state)->timeout = PFTM_ICMP_ERROR_REPLY;
4354
4355                 /* translate source/destination address, if necessary */
4356                 if (STATE_TRANSLATE((*state)->state_key)) {
4357                         if (direction == PF_OUT) {
4358                                 switch (pd->af) {
4359 #ifdef INET
4360                                 case AF_INET:
4361                                         pf_change_a(&saddr->v4.s_addr,
4362                                             pd->ip_sum,
4363                                             (*state)->state_key->gwy.addr.v4.s_addr, 0);
4364                                         pd->hdr.icmp->icmp_cksum =
4365                                             pf_cksum_fixup(
4366                                             pd->hdr.icmp->icmp_cksum, icmpid,
4367                                             (*state)->state_key->gwy.port, 0);
4368                                         pd->hdr.icmp->icmp_id =
4369                                             (*state)->state_key->gwy.port;
4370                                         m_copyback(m, off, ICMP_MINLEN,
4371                                             (caddr_t)pd->hdr.icmp);
4372                                         break;
4373 #endif /* INET */
4374 #ifdef INET6
4375                                 case AF_INET6:
4376                                         pf_change_a6(saddr,
4377                                             &pd->hdr.icmp6->icmp6_cksum,
4378                                             &(*state)->state_key->gwy.addr, 0);
4379                                         m_copyback(m, off,
4380                                             sizeof(struct icmp6_hdr),
4381                                             (caddr_t)pd->hdr.icmp6);
4382                                         break;
4383 #endif /* INET6 */
4384                                 }
4385                         } else {
4386                                 switch (pd->af) {
4387 #ifdef INET
4388                                 case AF_INET:
4389                                         pf_change_a(&daddr->v4.s_addr,
4390                                             pd->ip_sum,
4391                                             (*state)->state_key->lan.addr.v4.s_addr, 0);
4392                                         pd->hdr.icmp->icmp_cksum =
4393                                             pf_cksum_fixup(
4394                                             pd->hdr.icmp->icmp_cksum, icmpid,
4395                                             (*state)->state_key->lan.port, 0);
4396                                         pd->hdr.icmp->icmp_id =
4397                                             (*state)->state_key->lan.port;
4398                                         m_copyback(m, off, ICMP_MINLEN,
4399                                             (caddr_t)pd->hdr.icmp);
4400                                         break;
4401 #endif /* INET */
4402 #ifdef INET6
4403                                 case AF_INET6:
4404                                         pf_change_a6(daddr,
4405                                             &pd->hdr.icmp6->icmp6_cksum,
4406                                             &(*state)->state_key->lan.addr, 0);
4407                                         m_copyback(m, off,
4408                                             sizeof(struct icmp6_hdr),
4409                                             (caddr_t)pd->hdr.icmp6);
4410                                         break;
4411 #endif /* INET6 */
4412                                 }
4413                         }
4414                 }
4415
4416                 return (PF_PASS);
4417
4418         } else {
4419                 /*
4420                  * ICMP error message in response to a TCP/UDP packet.
4421                  * Extract the inner TCP/UDP header and search for that state.
4422                  */
4423
4424                 struct pf_pdesc pd2;
4425 #ifdef INET
4426                 struct ip       h2;
4427 #endif /* INET */
4428 #ifdef INET6
4429                 struct ip6_hdr  h2_6;
4430                 int             terminal = 0;
4431 #endif /* INET6 */
4432                 int             ipoff2;
4433                 int             off2;
4434
4435                 pd2.af = pd->af;
4436                 switch (pd->af) {
4437 #ifdef INET
4438                 case AF_INET:
4439                         /* offset of h2 in mbuf chain */
4440                         ipoff2 = off + ICMP_MINLEN;
4441
4442                         if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2),
4443                             NULL, reason, pd2.af)) {
4444                                 DPFPRINTF(PF_DEBUG_MISC,
4445                                     ("pf: ICMP error message too short "
4446                                     "(ip)\n"));
4447                                 return (PF_DROP);
4448                         }
4449                         /*
4450                          * ICMP error messages don't refer to non-first
4451                          * fragments
4452                          */
4453                         if (h2.ip_off & htons(IP_OFFMASK)) {
4454                                 REASON_SET(reason, PFRES_FRAG);
4455                                 return (PF_DROP);
4456                         }
4457
4458                         /* offset of protocol header that follows h2 */
4459                         off2 = ipoff2 + (h2.ip_hl << 2);
4460
4461                         pd2.proto = h2.ip_p;
4462                         pd2.src = (struct pf_addr *)&h2.ip_src;
4463                         pd2.dst = (struct pf_addr *)&h2.ip_dst;
4464                         pd2.ip_sum = &h2.ip_sum;
4465                         break;
4466 #endif /* INET */
4467 #ifdef INET6
4468                 case AF_INET6:
4469                         ipoff2 = off + sizeof(struct icmp6_hdr);
4470
4471                         if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6),
4472                             NULL, reason, pd2.af)) {
4473                                 DPFPRINTF(PF_DEBUG_MISC,
4474                                     ("pf: ICMP error message too short "
4475                                     "(ip6)\n"));
4476                                 return (PF_DROP);
4477                         }
4478                         pd2.proto = h2_6.ip6_nxt;
4479                         pd2.src = (struct pf_addr *)&h2_6.ip6_src;
4480                         pd2.dst = (struct pf_addr *)&h2_6.ip6_dst;
4481                         pd2.ip_sum = NULL;
4482                         off2 = ipoff2 + sizeof(h2_6);
4483                         do {
4484                                 switch (pd2.proto) {
4485                                 case IPPROTO_FRAGMENT:
4486                                         /*
4487                                          * ICMPv6 error messages for
4488                                          * non-first fragments
4489                                          */
4490                                         REASON_SET(reason, PFRES_FRAG);
4491                                         return (PF_DROP);
4492                                 case IPPROTO_AH:
4493                                 case IPPROTO_HOPOPTS:
4494                                 case IPPROTO_ROUTING:
4495                                 case IPPROTO_DSTOPTS: {
4496                                         /* get next header and header length */
4497                                         struct ip6_ext opt6;
4498
4499                                         if (!pf_pull_hdr(m, off2, &opt6,
4500                                             sizeof(opt6), NULL, reason,
4501                                             pd2.af)) {
4502                                                 DPFPRINTF(PF_DEBUG_MISC,
4503                                                     ("pf: ICMPv6 short opt\n"));
4504                                                 return (PF_DROP);
4505                                         }
4506                                         if (pd2.proto == IPPROTO_AH)
4507                                                 off2 += (opt6.ip6e_len + 2) * 4;
4508                                         else
4509                                                 off2 += (opt6.ip6e_len + 1) * 8;
4510                                         pd2.proto = opt6.ip6e_nxt;
4511                                         /* goto the next header */
4512                                         break;
4513                                 }
4514                                 default:
4515                                         terminal++;
4516                                         break;
4517                                 }
4518                         } while (!terminal);
4519                         break;
4520 #endif /* INET6 */
4521                 default:
4522                         DPFPRINTF(PF_DEBUG_MISC,
4523                             ("pf: ICMP AF %d unknown (ip6)\n", pd->af));
4524                         return (PF_DROP);
4525                         break;
4526                 }
4527
4528                 switch (pd2.proto) {
4529                 case IPPROTO_TCP: {
4530                         struct tcphdr            th;
4531                         u_int32_t                seq;
4532                         struct pf_state_peer    *src, *dst;
4533                         u_int8_t                 dws;
4534                         int                      copyback = 0;
4535
4536                         /*
4537                          * Only the first 8 bytes of the TCP header can be
4538                          * expected. Don't access any TCP header fields after
4539                          * th_seq, an ackskew test is not possible.
4540                          */
4541                         if (!pf_pull_hdr(m, off2, &th, 8, NULL, reason,
4542                             pd2.af)) {
4543                                 DPFPRINTF(PF_DEBUG_MISC,
4544                                     ("pf: ICMP error message too short "
4545                                     "(tcp)\n"));
4546                                 return (PF_DROP);
4547                         }
4548
4549                         key.af = pd2.af;
4550                         key.proto = IPPROTO_TCP;
4551                         if (direction == PF_IN) {
4552                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4553                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4554                                 key.ext.port = th.th_dport;
4555                                 key.gwy.port = th.th_sport;
4556                         } else {
4557                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4558                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4559                                 key.lan.port = th.th_dport;
4560                                 key.ext.port = th.th_sport;
4561                         }
4562
4563                         STATE_LOOKUP();
4564
4565                         if (direction == (*state)->state_key->direction) {
4566                                 src = &(*state)->dst;
4567                                 dst = &(*state)->src;
4568                         } else {
4569                                 src = &(*state)->src;
4570                                 dst = &(*state)->dst;
4571                         }
4572
4573                         if (src->wscale && dst->wscale)
4574                                 dws = dst->wscale & PF_WSCALE_MASK;
4575                         else
4576                                 dws = 0;
4577
4578                         /* Demodulate sequence number */
4579                         seq = ntohl(th.th_seq) - src->seqdiff;
4580                         if (src->seqdiff) {
4581                                 pf_change_a(&th.th_seq, icmpsum,
4582                                     htonl(seq), 0);
4583                                 copyback = 1;
4584                         }
4585
4586                         if (!SEQ_GEQ(src->seqhi, seq) ||
4587                             !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws))) {
4588                                 if (pf_status.debug >= PF_DEBUG_MISC) {
4589                                         kprintf("pf: BAD ICMP %d:%d ",
4590                                             icmptype, pd->hdr.icmp->icmp_code);
4591                                         pf_print_host(pd->src, 0, pd->af);
4592                                         kprintf(" -> ");
4593                                         pf_print_host(pd->dst, 0, pd->af);
4594                                         kprintf(" state: ");
4595                                         pf_print_state(*state);
4596                                         kprintf(" seq=%u\n", seq);
4597                                 }
4598                                 REASON_SET(reason, PFRES_BADSTATE);
4599                                 return (PF_DROP);
4600                         }
4601
4602                         if (STATE_TRANSLATE((*state)->state_key)) {
4603                                 if (direction == PF_IN) {
4604                                         pf_change_icmp(pd2.src, &th.th_sport,
4605                                             daddr, &(*state)->state_key->lan.addr,
4606                                             (*state)->state_key->lan.port, NULL,
4607                                             pd2.ip_sum, icmpsum,
4608                                             pd->ip_sum, 0, pd2.af);
4609                                 } else {
4610                                         pf_change_icmp(pd2.dst, &th.th_dport,
4611                                             saddr, &(*state)->state_key->gwy.addr,
4612                                             (*state)->state_key->gwy.port, NULL,
4613                                             pd2.ip_sum, icmpsum,
4614                                             pd->ip_sum, 0, pd2.af);
4615                                 }
4616                                 copyback = 1;
4617                         }
4618
4619                         if (copyback) {
4620                                 switch (pd2.af) {
4621 #ifdef INET
4622                                 case AF_INET:
4623                                         m_copyback(m, off, ICMP_MINLEN,
4624                                             (caddr_t)pd->hdr.icmp);
4625                                         m_copyback(m, ipoff2, sizeof(h2),
4626                                             (caddr_t)&h2);
4627                                         break;
4628 #endif /* INET */
4629 #ifdef INET6
4630                                 case AF_INET6:
4631                                         m_copyback(m, off,
4632                                             sizeof(struct icmp6_hdr),
4633                                             (caddr_t)pd->hdr.icmp6);
4634                                         m_copyback(m, ipoff2, sizeof(h2_6),
4635                                             (caddr_t)&h2_6);
4636                                         break;
4637 #endif /* INET6 */
4638                                 }
4639                                 m_copyback(m, off2, 8, (caddr_t)&th);
4640                         }
4641
4642                         return (PF_PASS);
4643                         break;
4644                 }
4645                 case IPPROTO_UDP: {
4646                         struct udphdr           uh;
4647
4648                         if (!pf_pull_hdr(m, off2, &uh, sizeof(uh),
4649                             NULL, reason, pd2.af)) {
4650                                 DPFPRINTF(PF_DEBUG_MISC,
4651                                     ("pf: ICMP error message too short "
4652                                     "(udp)\n"));
4653                                 return (PF_DROP);
4654                         }
4655
4656                         key.af = pd2.af;
4657                         key.proto = IPPROTO_UDP;
4658                         if (direction == PF_IN) {
4659                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4660                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4661                                 key.ext.port = uh.uh_dport;
4662                                 key.gwy.port = uh.uh_sport;
4663                         } else {
4664                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4665                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4666                                 key.lan.port = uh.uh_dport;
4667                                 key.ext.port = uh.uh_sport;
4668                         }
4669
4670                         STATE_LOOKUP();
4671
4672                         if (STATE_TRANSLATE((*state)->state_key)) {
4673                                 if (direction == PF_IN) {
4674                                         pf_change_icmp(pd2.src, &uh.uh_sport,
4675                                             daddr,
4676                                             &(*state)->state_key->lan.addr,
4677                                             (*state)->state_key->lan.port,
4678                                             &uh.uh_sum,
4679                                             pd2.ip_sum, icmpsum,
4680                                             pd->ip_sum, 1, pd2.af);
4681                                 } else {
4682                                         pf_change_icmp(pd2.dst, &uh.uh_dport,
4683                                             saddr,
4684                                             &(*state)->state_key->gwy.addr,
4685                                             (*state)->state_key->gwy.port, &uh.uh_sum,
4686                                             pd2.ip_sum, icmpsum,
4687                                             pd->ip_sum, 1, pd2.af);
4688                                 }
4689                                 switch (pd2.af) {
4690 #ifdef INET
4691                                 case AF_INET:
4692                                         m_copyback(m, off, ICMP_MINLEN,
4693                                             (caddr_t)pd->hdr.icmp);
4694                                         m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4695                                         break;
4696 #endif /* INET */
4697 #ifdef INET6
4698                                 case AF_INET6:
4699                                         m_copyback(m, off,
4700                                             sizeof(struct icmp6_hdr),
4701                                             (caddr_t)pd->hdr.icmp6);
4702                                         m_copyback(m, ipoff2, sizeof(h2_6),
4703                                             (caddr_t)&h2_6);
4704                                         break;
4705 #endif /* INET6 */
4706                                 }
4707                                 m_copyback(m, off2, sizeof(uh), (caddr_t)&uh);
4708                         }
4709
4710                         return (PF_PASS);
4711                         break;
4712                 }
4713 #ifdef INET
4714                 case IPPROTO_ICMP: {
4715                         struct icmp             iih;
4716
4717                         if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN,
4718                             NULL, reason, pd2.af)) {
4719                                 DPFPRINTF(PF_DEBUG_MISC,
4720                                     ("pf: ICMP error message too short i"
4721                                     "(icmp)\n"));
4722                                 return (PF_DROP);
4723                         }
4724
4725                         key.af = pd2.af;
4726                         key.proto = IPPROTO_ICMP;
4727                         if (direction == PF_IN) {
4728                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4729                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4730                                 key.ext.port = 0;
4731                                 key.gwy.port = iih.icmp_id;
4732                         } else {
4733                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4734                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4735                                 key.lan.port = iih.icmp_id;
4736                                 key.ext.port = 0;
4737                         }
4738
4739                         STATE_LOOKUP();
4740
4741                         if (STATE_TRANSLATE((*state)->state_key)) {
4742                                 if (direction == PF_IN) {
4743                                         pf_change_icmp(pd2.src, &iih.icmp_id,
4744                                             daddr,
4745                                             &(*state)->state_key->lan.addr,
4746                                             (*state)->state_key->lan.port, NULL,
4747                                             pd2.ip_sum, icmpsum,
4748                                             pd->ip_sum, 0, AF_INET);
4749                                 } else {
4750                                         pf_change_icmp(pd2.dst, &iih.icmp_id,
4751                                             saddr,
4752                                             &(*state)->state_key->gwy.addr,
4753                                             (*state)->state_key->gwy.port, NULL,
4754                                             pd2.ip_sum, icmpsum,
4755                                             pd->ip_sum, 0, AF_INET);
4756                                 }
4757                                 m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
4758                                 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4759                                 m_copyback(m, off2, ICMP_MINLEN, (caddr_t)&iih);
4760                         }
4761
4762                         return (PF_PASS);
4763                         break;
4764                 }
4765 #endif /* INET */
4766 #ifdef INET6
4767                 case IPPROTO_ICMPV6: {
4768                         struct icmp6_hdr        iih;
4769
4770                         if (!pf_pull_hdr(m, off2, &iih,
4771                             sizeof(struct icmp6_hdr), NULL, reason, pd2.af)) {
4772                                 DPFPRINTF(PF_DEBUG_MISC,
4773                                     ("pf: ICMP error message too short "
4774                                     "(icmp6)\n"));
4775                                 return (PF_DROP);
4776                         }
4777
4778                         key.af = pd2.af;
4779                         key.proto = IPPROTO_ICMPV6;
4780                         if (direction == PF_IN) {
4781                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4782                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4783                                 key.ext.port = 0;
4784                                 key.gwy.port = iih.icmp6_id;
4785                         } else {
4786                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4787                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4788                                 key.lan.port = iih.icmp6_id;
4789                                 key.ext.port = 0;
4790                         }
4791
4792                         STATE_LOOKUP();
4793
4794                         if (STATE_TRANSLATE((*state)->state_key)) {
4795                                 if (direction == PF_IN) {
4796                                         pf_change_icmp(pd2.src, &iih.icmp6_id,
4797                                             daddr,
4798                                             &(*state)->state_key->lan.addr,
4799                                             (*state)->state_key->lan.port, NULL,
4800                                             pd2.ip_sum, icmpsum,
4801                                             pd->ip_sum, 0, AF_INET6);
4802                                 } else {
4803                                         pf_change_icmp(pd2.dst, &iih.icmp6_id,
4804                                             saddr, &(*state)->state_key->gwy.addr,
4805                                             (*state)->state_key->gwy.port, NULL,
4806                                             pd2.ip_sum, icmpsum,
4807                                             pd->ip_sum, 0, AF_INET6);
4808                                 }
4809                                 m_copyback(m, off, sizeof(struct icmp6_hdr),
4810                                     (caddr_t)pd->hdr.icmp6);
4811                                 m_copyback(m, ipoff2, sizeof(h2_6), (caddr_t)&h2_6);
4812                                 m_copyback(m, off2, sizeof(struct icmp6_hdr),
4813                                     (caddr_t)&iih);
4814                         }
4815
4816                         return (PF_PASS);
4817                         break;
4818                 }
4819 #endif /* INET6 */
4820                 default: {
4821                         key.af = pd2.af;
4822                         key.proto = pd2.proto;
4823                         if (direction == PF_IN) {
4824                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4825                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4826                                 key.ext.port = 0;
4827                                 key.gwy.port = 0;
4828                         } else {
4829                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4830                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4831                                 key.lan.port = 0;
4832                                 key.ext.port = 0;
4833                         }
4834
4835                         STATE_LOOKUP();
4836
4837                         if (STATE_TRANSLATE((*state)->state_key)) {
4838                                 if (direction == PF_IN) {
4839                                         pf_change_icmp(pd2.src, NULL,
4840                                             daddr,
4841                                             &(*state)->state_key->lan.addr,
4842                                             0, NULL,
4843                                             pd2.ip_sum, icmpsum,
4844                                             pd->ip_sum, 0, pd2.af);
4845                                 } else {
4846                                         pf_change_icmp(pd2.dst, NULL,
4847                                             saddr,
4848                                             &(*state)->state_key->gwy.addr,
4849                                             0, NULL,
4850                                             pd2.ip_sum, icmpsum,
4851                                             pd->ip_sum, 0, pd2.af);
4852                                 }
4853                                 switch (pd2.af) {
4854 #ifdef INET
4855                                 case AF_INET:
4856                                         m_copyback(m, off, ICMP_MINLEN,
4857                                             (caddr_t)pd->hdr.icmp);
4858                                         m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4859                                         break;
4860 #endif /* INET */
4861 #ifdef INET6
4862                                 case AF_INET6:
4863                                         m_copyback(m, off,
4864                                             sizeof(struct icmp6_hdr),
4865                                             (caddr_t)pd->hdr.icmp6);
4866                                         m_copyback(m, ipoff2, sizeof(h2_6),
4867                                             (caddr_t)&h2_6);
4868                                         break;
4869 #endif /* INET6 */
4870                                 }
4871                         }
4872
4873                         return (PF_PASS);
4874                         break;
4875                 }
4876                 }
4877         }
4878 }
4879
4880 int
4881 pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif,
4882     struct pf_pdesc *pd)
4883 {
4884         struct pf_state_peer    *src, *dst;
4885         struct pf_state_key_cmp  key;
4886
4887         key.af = pd->af;
4888         key.proto = pd->proto;
4889         if (direction == PF_IN) {
4890                 PF_ACPY(&key.ext.addr, pd->src, key.af);
4891                 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4892                 key.ext.port = 0;
4893                 key.gwy.port = 0;
4894         } else {
4895                 PF_ACPY(&key.lan.addr, pd->src, key.af);
4896                 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4897                 key.lan.port = 0;
4898                 key.ext.port = 0;
4899         }
4900
4901         STATE_LOOKUP();
4902
4903         if (direction == (*state)->state_key->direction) {
4904                 src = &(*state)->src;
4905                 dst = &(*state)->dst;
4906         } else {
4907                 src = &(*state)->dst;
4908                 dst = &(*state)->src;
4909         }
4910
4911         /* update states */
4912         if (src->state < PFOTHERS_SINGLE)
4913                 src->state = PFOTHERS_SINGLE;
4914         if (dst->state == PFOTHERS_SINGLE)
4915                 dst->state = PFOTHERS_MULTIPLE;
4916
4917         /* update expire time */
4918         (*state)->expire = time_second;
4919         if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE)
4920                 (*state)->timeout = PFTM_OTHER_MULTIPLE;
4921         else
4922                 (*state)->timeout = PFTM_OTHER_SINGLE;
4923
4924         /* translate source/destination address, if necessary */
4925         if (STATE_TRANSLATE((*state)->state_key)) {
4926                 if (direction == PF_OUT)
4927                         switch (pd->af) {
4928 #ifdef INET
4929                         case AF_INET:
4930                                 pf_change_a(&pd->src->v4.s_addr,
4931                                     pd->ip_sum,
4932                                     (*state)->state_key->gwy.addr.v4.s_addr,
4933                                     0);
4934                                 break;
4935 #endif /* INET */
4936 #ifdef INET6
4937                         case AF_INET6:
4938                                 PF_ACPY(pd->src,
4939                                     &(*state)->state_key->gwy.addr, pd->af);
4940                                 break;
4941 #endif /* INET6 */
4942                         }
4943                 else
4944                         switch (pd->af) {
4945 #ifdef INET
4946                         case AF_INET:
4947                                 pf_change_a(&pd->dst->v4.s_addr,
4948                                     pd->ip_sum,
4949                                     (*state)->state_key->lan.addr.v4.s_addr,
4950                                     0);
4951                                 break;
4952 #endif /* INET */
4953 #ifdef INET6
4954                         case AF_INET6:
4955                                 PF_ACPY(pd->dst,
4956                                     &(*state)->state_key->lan.addr, pd->af);
4957                                 break;
4958 #endif /* INET6 */
4959                         }
4960         }
4961
4962         return (PF_PASS);
4963 }
4964
4965 /*
4966  * ipoff and off are measured from the start of the mbuf chain.
4967  * h must be at "ipoff" on the mbuf chain.
4968  */
4969 void *
4970 pf_pull_hdr(struct mbuf *m, int off, void *p, int len,
4971     u_short *actionp, u_short *reasonp, sa_family_t af)
4972 {
4973         switch (af) {
4974 #ifdef INET
4975         case AF_INET: {
4976                 struct ip       *h = mtod(m, struct ip *);
4977                 u_int16_t        fragoff = (h->ip_off & IP_OFFMASK) << 3;
4978
4979                 if (fragoff) {
4980                         if (fragoff >= len)
4981                                 ACTION_SET(actionp, PF_PASS);
4982                         else {
4983                                 ACTION_SET(actionp, PF_DROP);
4984                                 REASON_SET(reasonp, PFRES_FRAG);
4985                         }
4986                         return (NULL);
4987                 }
4988                 if (m->m_pkthdr.len < off + len ||
4989                     h->ip_len < off + len) {
4990                         ACTION_SET(actionp, PF_DROP);
4991                         REASON_SET(reasonp, PFRES_SHORT);
4992                         return (NULL);
4993                 }
4994                 break;
4995         }
4996 #endif /* INET */
4997 #ifdef INET6
4998         case AF_INET6: {
4999                 struct ip6_hdr  *h = mtod(m, struct ip6_hdr *);
5000
5001                 if (m->m_pkthdr.len < off + len ||
5002                     (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) <
5003                     (unsigned)(off + len)) {
5004                         ACTION_SET(actionp, PF_DROP);
5005                         REASON_SET(reasonp, PFRES_SHORT);
5006                         return (NULL);
5007                 }
5008                 break;
5009         }
5010 #endif /* INET6 */
5011         }
5012         m_copydata(m, off, len, p);
5013         return (p);
5014 }
5015
5016 int
5017 pf_routable(struct pf_addr *addr, sa_family_t af, struct pfi_kif *kif)
5018 {
5019         struct sockaddr_in      *dst;
5020         int                      ret = 1;
5021         int                      check_mpath;
5022 #ifdef INET6
5023         struct sockaddr_in6     *dst6;
5024         struct route_in6         ro;
5025 #else
5026         struct route             ro;
5027 #endif
5028         struct radix_node       *rn;
5029         struct rtentry          *rt;
5030         struct ifnet            *ifp;
5031
5032         check_mpath = 0;
5033         bzero(&ro, sizeof(ro));
5034         switch (af) {
5035         case AF_INET:
5036                 dst = satosin(&ro.ro_dst);
5037                 dst->sin_family = AF_INET;
5038                 dst->sin_len = sizeof(*dst);
5039                 dst->sin_addr = addr->v4;
5040                 break;
5041 #ifdef INET6
5042         case AF_INET6:
5043                 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
5044                 dst6->sin6_family = AF_INET6;
5045                 dst6->sin6_len = sizeof(*dst6);
5046                 dst6->sin6_addr = addr->v6;
5047                 break;
5048 #endif /* INET6 */
5049         default:
5050                 return (0);
5051         }
5052
5053         /* Skip checks for ipsec interfaces */
5054         if (kif != NULL && kif->pfik_ifp->if_type == IFT_ENC)
5055                 goto out;
5056
5057         rtalloc_ign((struct route *)&ro, 0);
5058
5059         if (ro.ro_rt != NULL) {
5060                 /* No interface given, this is a no-route check */
5061                 if (kif == NULL)
5062                         goto out;
5063
5064                 if (kif->pfik_ifp == NULL) {
5065                         ret = 0;
5066                         goto out;
5067                 }
5068
5069                 /* Perform uRPF check if passed input interface */
5070                 ret = 0;
5071                 rn = (struct radix_node *)ro.ro_rt;
5072                 do {
5073                         rt = (struct rtentry *)rn;
5074                         ifp = rt->rt_ifp;
5075
5076                         if (kif->pfik_ifp == ifp)
5077                                 ret = 1;
5078                         rn = NULL;
5079                 } while (check_mpath == 1 && rn != NULL && ret == 0);
5080         } else
5081                 ret = 0;
5082 out:
5083         if (ro.ro_rt != NULL)
5084                 RTFREE(ro.ro_rt);
5085         return (ret);
5086 }
5087
5088 int
5089 pf_rtlabel_match(struct pf_addr *addr, sa_family_t af, struct pf_addr_wrap *aw)
5090 {
5091         struct sockaddr_in      *dst;
5092 #ifdef INET6
5093         struct sockaddr_in6     *dst6;
5094         struct route_in6         ro;
5095 #else
5096         struct route             ro;
5097 #endif
5098         int                      ret = 0;
5099
5100         bzero(&ro, sizeof(ro));
5101         switch (af) {
5102         case AF_INET:
5103                 dst = satosin(&ro.ro_dst);
5104                 dst->sin_family = AF_INET;
5105                 dst->sin_len = sizeof(*dst);
5106                 dst->sin_addr = addr->v4;
5107                 break;
5108 #ifdef INET6
5109         case AF_INET6:
5110                 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
5111                 dst6->sin6_family = AF_INET6;
5112                 dst6->sin6_len = sizeof(*dst6);
5113                 dst6->sin6_addr = addr->v6;
5114                 break;
5115 #endif /* INET6 */
5116         default:
5117                 return (0);
5118         }
5119
5120 rtalloc_ign((struct route *)&ro, (RTF_CLONING | RTF_PRCLONING));
5121
5122         if (ro.ro_rt != NULL) {
5123                 RTFREE(ro.ro_rt);
5124         }
5125
5126         return (ret);
5127 }
5128
5129 #ifdef INET
5130 void
5131 pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5132     struct pf_state *s, struct pf_pdesc *pd)
5133 {
5134         struct mbuf             *m0, *m1;
5135         struct route             iproute;
5136         struct route            *ro = NULL;
5137         struct sockaddr_in      *dst;
5138         struct ip               *ip;
5139         struct ifnet            *ifp = NULL;
5140         struct pf_addr           naddr;
5141         struct pf_src_node      *sn = NULL;
5142         int                      error = 0;
5143         int sw_csum;
5144 #ifdef IPSEC
5145         struct m_tag            *mtag;
5146 #endif /* IPSEC */
5147
5148         if (m == NULL || *m == NULL || r == NULL ||
5149             (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5150                 panic("pf_route: invalid parameters");
5151
5152         if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
5153                 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
5154                 (*m)->m_pkthdr.pf.routed = 1;
5155         } else {
5156                 if ((*m)->m_pkthdr.pf.routed++ > 3) {
5157                         m0 = *m;
5158                         *m = NULL;
5159                         goto bad;
5160                 }
5161         }
5162
5163         if (r->rt == PF_DUPTO) {
5164                 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL)
5165                         return;
5166         } else {
5167                 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5168                         return;
5169                 m0 = *m;
5170         }
5171
5172         if (m0->m_len < sizeof(struct ip)) {
5173                 DPFPRINTF(PF_DEBUG_URGENT,
5174                     ("pf_route: m0->m_len < sizeof(struct ip)\n"));
5175                 goto bad;
5176         }
5177
5178         ip = mtod(m0, struct ip *);
5179
5180         ro = &iproute;
5181         bzero((caddr_t)ro, sizeof(*ro));
5182         dst = satosin(&ro->ro_dst);
5183         dst->sin_family = AF_INET;
5184         dst->sin_len = sizeof(*dst);
5185         dst->sin_addr = ip->ip_dst;
5186
5187         if (r->rt == PF_FASTROUTE) {
5188                 rtalloc(ro);
5189                 if (ro->ro_rt == 0) {
5190                         ipstat.ips_noroute++;
5191                         goto bad;
5192                 }
5193
5194                 ifp = ro->ro_rt->rt_ifp;
5195                 ro->ro_rt->rt_use++;
5196
5197                 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
5198                         dst = satosin(ro->ro_rt->rt_gateway);
5199         } else {
5200                 if (TAILQ_EMPTY(&r->rpool.list)) {
5201                         DPFPRINTF(PF_DEBUG_URGENT,
5202                             ("pf_route: TAILQ_EMPTY(&r->rpool.list)\n"));
5203                         goto bad;
5204                 }
5205                 if (s == NULL) {
5206                         pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src,
5207                             &naddr, NULL, &sn);
5208                         if (!PF_AZERO(&naddr, AF_INET))
5209                                 dst->sin_addr.s_addr = naddr.v4.s_addr;
5210                         ifp = r->rpool.cur->kif ?
5211                             r->rpool.cur->kif->pfik_ifp : NULL;
5212                 } else {
5213                         if (!PF_AZERO(&s->rt_addr, AF_INET))
5214                                 dst->sin_addr.s_addr =
5215                                     s->rt_addr.v4.s_addr;
5216                         ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5217                 }
5218         }
5219         if (ifp == NULL)
5220                 goto bad;
5221
5222         if (oifp != ifp) {
5223                 crit_exit();
5224                 if (pf_test(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) {
5225                         crit_enter();
5226                         goto bad;
5227                 } else if (m0 == NULL) {
5228                         crit_enter();
5229                         goto done;
5230                 }
5231                 crit_enter();
5232                 if (m0->m_len < sizeof(struct ip)) {
5233                         DPFPRINTF(PF_DEBUG_URGENT,
5234                             ("pf_route: m0->m_len < sizeof(struct ip)\n"));
5235                         goto bad;
5236                 }
5237                 ip = mtod(m0, struct ip *);
5238         }
5239
5240         /* Copied from FreeBSD 5.1-CURRENT ip_output. */
5241         m0->m_pkthdr.csum_flags |= CSUM_IP;
5242         sw_csum = m0->m_pkthdr.csum_flags & ~ifp->if_hwassist;
5243         if (sw_csum & CSUM_DELAY_DATA) {
5244                 in_delayed_cksum(m0);
5245                 sw_csum &= ~CSUM_DELAY_DATA;
5246         }
5247         m0->m_pkthdr.csum_flags &= ifp->if_hwassist;
5248
5249         if (ip->ip_len <= ifp->if_mtu ||
5250             (ifp->if_hwassist & CSUM_FRAGMENT &&
5251                 (ip->ip_off & IP_DF) == 0)) {
5252                 ip->ip_len = htons(ip->ip_len);
5253                 ip->ip_off = htons(ip->ip_off);
5254                 ip->ip_sum = 0;
5255                 if (sw_csum & CSUM_DELAY_IP) {
5256                         /* From KAME */
5257                         if (ip->ip_v == IPVERSION &&
5258                             (ip->ip_hl << 2) == sizeof(*ip)) {
5259                                 ip->ip_sum = in_cksum_hdr(ip);
5260                         } else {
5261                                 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
5262                         }
5263                 }
5264                 crit_exit();
5265                 error = ifp->if_output(ifp, m0, sintosa(dst), ro->ro_rt);
5266                 crit_enter();
5267                 goto done;
5268         }
5269
5270         /*
5271          * Too large for interface; fragment if possible.
5272          * Must be able to put at least 8 bytes per fragment.
5273          */
5274         if (ip->ip_off & IP_DF) {
5275                 ipstat.ips_cantfrag++;
5276                 if (r->rt != PF_DUPTO) {
5277                         crit_exit();
5278                         icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0,
5279                             ifp->if_mtu);
5280                         crit_enter();
5281                         goto done;
5282                 } else
5283                         goto bad;
5284         }
5285
5286         m1 = m0;
5287         error = ip_fragment(ip, &m0, ifp->if_mtu, ifp->if_hwassist, sw_csum);
5288         if (error) {
5289                 goto bad;
5290         }
5291
5292         for (m0 = m1; m0; m0 = m1) {
5293                 m1 = m0->m_nextpkt;
5294                 m0->m_nextpkt = 0;
5295                 if (error == 0) {
5296                         crit_exit();
5297                         error = (*ifp->if_output)(ifp, m0, sintosa(dst),
5298                             NULL);
5299                         crit_enter();
5300                 } else
5301                         m_freem(m0);
5302         }
5303
5304         if (error == 0)
5305                 ipstat.ips_fragmented++;
5306
5307 done:
5308         if (r->rt != PF_DUPTO)
5309                 *m = NULL;
5310         if (ro == &iproute && ro->ro_rt)
5311                 RTFREE(ro->ro_rt);
5312         return;
5313
5314 bad:
5315         m_freem(m0);
5316         goto done;
5317 }
5318 #endif /* INET */
5319
5320 #ifdef INET6
5321 void
5322 pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5323     struct pf_state *s, struct pf_pdesc *pd)
5324 {
5325         struct mbuf             *m0;
5326         struct route_in6         ip6route;
5327         struct route_in6        *ro;
5328         struct sockaddr_in6     *dst;
5329         struct ip6_hdr          *ip6;
5330         struct ifnet            *ifp = NULL;
5331         struct pf_addr           naddr;
5332         struct pf_src_node      *sn = NULL;
5333         int                      error = 0;
5334
5335         if (m == NULL || *m == NULL || r == NULL ||
5336             (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5337                 panic("pf_route6: invalid parameters");
5338
5339         if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
5340                 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
5341                 (*m)->m_pkthdr.pf.routed = 1;
5342         } else {
5343                 if ((*m)->m_pkthdr.pf.routed++ > 3) {
5344                         m0 = *m;
5345                         *m = NULL;
5346                         goto bad;
5347                 }
5348         }
5349
5350         if (r->rt == PF_DUPTO) {
5351                 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL)
5352                         return;
5353         } else {
5354                 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5355                         return;
5356                 m0 = *m;
5357         }
5358
5359         if (m0->m_len < sizeof(struct ip6_hdr)) {
5360                 DPFPRINTF(PF_DEBUG_URGENT,
5361                     ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n"));
5362                 goto bad;
5363         }
5364         ip6 = mtod(m0, struct ip6_hdr *);
5365
5366         ro = &ip6route;
5367         bzero((caddr_t)ro, sizeof(*ro));
5368         dst = (struct sockaddr_in6 *)&ro->ro_dst;
5369         dst->sin6_family = AF_INET6;
5370         dst->sin6_len = sizeof(*dst);
5371         dst->sin6_addr = ip6->ip6_dst;
5372
5373         /* Cheat. XXX why only in the v6 case??? */
5374         if (r->rt == PF_FASTROUTE) {
5375                 m0->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
5376                 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
5377                 crit_enter();
5378                 return;
5379         }
5380
5381         if (TAILQ_EMPTY(&r->rpool.list)) {
5382                 DPFPRINTF(PF_DEBUG_URGENT,
5383                     ("pf_route6: TAILQ_EMPTY(&r->rpool.list)\n"));
5384                 goto bad;
5385         }
5386         if (s == NULL) {
5387                 pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src,
5388                     &naddr, NULL, &sn);
5389                 if (!PF_AZERO(&naddr, AF_INET6))
5390                         PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5391                             &naddr, AF_INET6);
5392                 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL;
5393         } else {
5394                 if (!PF_AZERO(&s->rt_addr, AF_INET6))
5395                         PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5396                             &s->rt_addr, AF_INET6);
5397                 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5398         }
5399         if (ifp == NULL)
5400                 goto bad;
5401
5402         if (oifp != ifp) {
5403                 crit_exit();
5404                 if (pf_test6(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) {
5405                         crit_enter();
5406                         goto bad;
5407                 } else if (m0 == NULL) {
5408                         crit_enter();
5409                         goto done;
5410                 }
5411                 crit_enter();
5412                 if (m0->m_len < sizeof(struct ip6_hdr)) {
5413                         DPFPRINTF(PF_DEBUG_URGENT,
5414                             ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n"));
5415                         goto bad;
5416                 }
5417                 ip6 = mtod(m0, struct ip6_hdr *);
5418         }
5419
5420         /*
5421          * If the packet is too large for the outgoing interface,
5422          * send back an icmp6 error.
5423          */
5424         if (IN6_IS_ADDR_LINKLOCAL(&dst->sin6_addr))
5425                 dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
5426         if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) {
5427                 crit_exit();
5428                 error = nd6_output(ifp, ifp, m0, dst, NULL);
5429                 crit_enter();
5430         } else {
5431                 in6_ifstat_inc(ifp, ifs6_in_toobig);
5432                 if (r->rt != PF_DUPTO) {
5433                         crit_exit();
5434                         icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
5435                         crit_enter();
5436                  } else
5437                         goto bad;
5438         }
5439
5440 done:
5441         if (r->rt != PF_DUPTO)
5442                 *m = NULL;
5443         return;
5444
5445 bad:
5446         m_freem(m0);
5447         goto done;
5448 }
5449
5450 #endif /* INET6 */
5451
5452
5453 /*
5454  * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag
5455  *   off is the offset where the protocol header starts
5456  *   len is the total length of protocol header plus payload
5457  * returns 0 when the checksum is valid, otherwise returns 1.
5458  */
5459 /*
5460  * XXX
5461  * FreeBSD supports cksum offload for the following drivers.
5462  * em(4), gx(4), lge(4), nge(4), ti(4), xl(4)
5463  * If we can make full use of it we would outperform ipfw/ipfilter in
5464  * very heavy traffic.
5465  * I have not tested 'cause I don't have NICs that supports cksum offload.
5466  * (There might be problems. Typical phenomena would be
5467  *   1. No route message for UDP packet.
5468  *   2. No connection acceptance from external hosts regardless of rule set.)
5469  */
5470 int
5471 pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p,
5472     sa_family_t af)
5473 {
5474         u_int16_t sum = 0;
5475         int hw_assist = 0;
5476         struct ip *ip;
5477
5478         if (off < sizeof(struct ip) || len < sizeof(struct udphdr))
5479                 return (1);
5480         if (m->m_pkthdr.len < off + len)
5481                 return (1);
5482
5483         switch (p) {
5484         case IPPROTO_TCP:
5485         case IPPROTO_UDP:
5486                 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
5487                         if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) {
5488                                 sum = m->m_pkthdr.csum_data;
5489                         } else {
5490                                 ip = mtod(m, struct ip *);
5491                                 sum = in_pseudo(ip->ip_src.s_addr,
5492                                         ip->ip_dst.s_addr, htonl((u_short)len +
5493                                         m->m_pkthdr.csum_data + p));
5494                         }
5495                         sum ^= 0xffff;
5496                         ++hw_assist;
5497                 }
5498                 break;
5499         case IPPROTO_ICMP:
5500 #ifdef INET6
5501         case IPPROTO_ICMPV6:
5502 #endif /* INET6 */
5503                 break;
5504         default:
5505                 return (1);
5506         }
5507
5508         if (!hw_assist) {
5509                 switch (af) {
5510                 case AF_INET:
5511                         if (p == IPPROTO_ICMP) {
5512                                 if (m->m_len < off)
5513                                         return (1);
5514                                 m->m_data += off;
5515                                 m->m_len -= off;
5516                                 sum = in_cksum(m, len);
5517                                 m->m_data -= off;
5518                                 m->m_len += off;
5519                         } else {
5520                                 if (m->m_len < sizeof(struct ip))
5521                                         return (1);
5522                                 sum = in_cksum_range(m, p, off, len);
5523                                 if (sum == 0) {
5524                                         m->m_pkthdr.csum_flags |=
5525                                             (CSUM_DATA_VALID |
5526                                              CSUM_PSEUDO_HDR);
5527                                         m->m_pkthdr.csum_data = 0xffff;
5528                                 }
5529                         }
5530                         break;
5531 #ifdef INET6
5532                 case AF_INET6:
5533                         if (m->m_len < sizeof(struct ip6_hdr))
5534                                 return (1);
5535                         sum = in6_cksum(m, p, off, len);
5536                         /*
5537                          * XXX
5538                          * IPv6 H/W cksum off-load not supported yet!
5539                          *
5540                          * if (sum == 0) {
5541                          *      m->m_pkthdr.csum_flags |=
5542                          *          (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
5543                          *      m->m_pkthdr.csum_data = 0xffff;
5544                          *}
5545                          */
5546                         break;
5547 #endif /* INET6 */
5548                 default:
5549                         return (1);
5550                 }
5551         }
5552         if (sum) {
5553                 switch (p) {
5554                 case IPPROTO_TCP:
5555                         tcpstat.tcps_rcvbadsum++;
5556                         break;
5557                 case IPPROTO_UDP:
5558                         udpstat.udps_badsum++;
5559                         break;
5560                 case IPPROTO_ICMP:
5561                         icmpstat.icps_checksum++;
5562                         break;
5563 #ifdef INET6
5564                 case IPPROTO_ICMPV6:
5565                         icmp6stat.icp6s_checksum++;
5566                         break;
5567 #endif /* INET6 */
5568                 }
5569                 return (1);
5570         }
5571         return (0);
5572 }
5573
5574 #ifdef INET
5575 int
5576 pf_test(int dir, struct ifnet *ifp, struct mbuf **m0,
5577     struct ether_header *eh, struct inpcb *inp)
5578 {
5579         struct pfi_kif          *kif;
5580         u_short                  action, reason = 0, log = 0;
5581         struct mbuf             *m = *m0;
5582         struct ip               *h = NULL;
5583         struct pf_rule          *a = NULL, *r = &pf_default_rule, *tr, *nr;
5584         struct pf_state         *s = NULL;
5585         struct pf_state_key     *sk = NULL;
5586         struct pf_ruleset       *ruleset = NULL;
5587         struct pf_pdesc          pd;
5588         int                      off, dirndx, pqid = 0;
5589
5590         if (!pf_status.running)
5591                 return (PF_PASS);
5592
5593         memset(&pd, 0, sizeof(pd));
5594         if (ifp->if_type == IFT_CARP && ifp->if_carpdev)
5595                 kif = (struct pfi_kif *)ifp->if_carpdev->if_pf_kif;
5596         else
5597                 kif = (struct pfi_kif *)ifp->if_pf_kif;
5598
5599         if (kif == NULL) {
5600                 DPFPRINTF(PF_DEBUG_URGENT,
5601                     ("pf_test: kif == NULL, if_xname %s\n", ifp->if_xname));
5602                 return (PF_DROP);
5603         }
5604         if (kif->pfik_flags & PFI_IFLAG_SKIP)
5605                 return (PF_PASS);
5606
5607 #ifdef DIAGNOSTIC
5608         if ((m->m_flags & M_PKTHDR) == 0)
5609                 panic("non-M_PKTHDR is passed to pf_test");
5610 #endif /* DIAGNOSTIC */
5611
5612         if (m->m_pkthdr.len < (int)sizeof(*h)) {
5613                 action = PF_DROP;
5614                 REASON_SET(&reason, PFRES_SHORT);
5615                 log = 1;
5616                 goto done;
5617         }
5618
5619         if (m->m_pkthdr.pf.flags & PF_TAG_GENERATED)
5620                 return (PF_PASS);
5621
5622         /* We do IP header normalization and packet reassembly here */
5623         if (pf_normalize_ip(m0, dir, kif, &reason, &pd) != PF_PASS) {
5624                 action = PF_DROP;
5625                 goto done;
5626         }
5627         m = *m0;        /* pf_normalize messes with m0 */
5628         h = mtod(m, struct ip *);
5629
5630         off = h->ip_hl << 2;
5631         if (off < (int)sizeof(*h)) {
5632                 action = PF_DROP;
5633                 REASON_SET(&reason, PFRES_SHORT);
5634                 log = 1;
5635                 goto done;
5636         }
5637
5638         pd.src = (struct pf_addr *)&h->ip_src;
5639         pd.dst = (struct pf_addr *)&h->ip_dst;
5640         PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET);
5641         pd.ip_sum = &h->ip_sum;
5642         pd.proto = h->ip_p;
5643         pd.af = AF_INET;
5644         pd.tos = h->ip_tos;
5645         pd.tot_len = h->ip_len;
5646         pd.eh = eh;
5647
5648         /* handle fragments that didn't get reassembled by normalization */
5649         if (h->ip_off & (IP_MF | IP_OFFMASK)) {
5650                 action = pf_test_fragment(&r, dir, kif, m, h,
5651                     &pd, &a, &ruleset);
5652                 goto done;
5653         }
5654
5655         switch (h->ip_p) {
5656
5657         case IPPROTO_TCP: {
5658                 struct tcphdr   th;
5659
5660                 pd.hdr.tcp = &th;
5661                 if (!pf_pull_hdr(m, off, &th, sizeof(th),
5662                     &action, &reason, AF_INET)) {
5663                         log = action != PF_PASS;
5664                         goto done;
5665                 }
5666                 pd.p_len = pd.tot_len - off - (th.th_off << 2);
5667                 if ((th.th_flags & TH_ACK) && pd.p_len == 0)
5668                         pqid = 1;
5669                 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
5670                 if (action == PF_DROP)
5671                         goto done;
5672                 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
5673                     &reason);
5674                 if (action == PF_PASS) {
5675 #if NPFSYNC
5676                         pfsync_update_state(s);
5677 #endif /* NPFSYNC */
5678                         r = s->rule.ptr;
5679                         a = s->anchor.ptr;
5680                         log = s->log;
5681                 } else if (s == NULL)
5682                         action = pf_test_rule(&r, &s, dir, kif,
5683                             m, off, h, &pd, &a, &ruleset, NULL, inp);
5684                 break;
5685         }
5686
5687         case IPPROTO_UDP: {
5688                 struct udphdr   uh;
5689
5690                 pd.hdr.udp = &uh;
5691                 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
5692                     &action, &reason, AF_INET)) {
5693                         log = action != PF_PASS;
5694                         goto done;
5695                 }
5696                 if (uh.uh_dport == 0 ||
5697                     ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
5698                     ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
5699                         action = PF_DROP;
5700                         REASON_SET(&reason, PFRES_SHORT);
5701                         goto done;
5702                 }
5703                 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
5704                 if (action == PF_PASS) {
5705 #if NPFSYNC
5706                         pfsync_update_state(s);
5707 #endif /* NPFSYNC */
5708                         r = s->rule.ptr;
5709                         a = s->anchor.ptr;
5710                         log = s->log;
5711                 } else if (s == NULL)
5712                         action = pf_test_rule(&r, &s, dir, kif,
5713                             m, off, h, &pd, &a, &ruleset, NULL, inp);
5714                 break;
5715         }
5716
5717         case IPPROTO_ICMP: {
5718                 struct icmp     ih;
5719
5720                 pd.hdr.icmp = &ih;
5721                 if (!pf_pull_hdr(m, off, &ih, ICMP_MINLEN,
5722                     &action, &reason, AF_INET)) {
5723                         log = action != PF_PASS;
5724                         goto done;
5725                 }
5726                 action = pf_test_state_icmp(&s, dir, kif, m, off, h, &pd,
5727                     &reason);
5728                 if (action == PF_PASS) {
5729 #if NPFSYNC
5730                         pfsync_update_state(s);
5731 #endif /* NPFSYNC */
5732                         r = s->rule.ptr;
5733                         a = s->anchor.ptr;
5734                         log = s->log;
5735                 } else if (s == NULL)
5736                         action = pf_test_rule(&r, &s, dir, kif,
5737                             m, off, h, &pd, &a, &ruleset, NULL, inp);
5738                 break;
5739         }
5740
5741         default:
5742                 action = pf_test_state_other(&s, dir, kif, &pd);
5743                 if (action == PF_PASS) {
5744 #if NPFSYNC
5745                         pfsync_update_state(s);
5746 #endif /* NPFSYNC */
5747                         r = s->rule.ptr;
5748                         a = s->anchor.ptr;
5749                         log = s->log;
5750                 } else if (s == NULL)
5751                         action = pf_test_rule(&r, &s, dir, kif, m, off, h,
5752                             &pd, &a, &ruleset, NULL, inp);
5753                 break;
5754         }
5755
5756 done:
5757         if (action == PF_PASS && h->ip_hl > 5 &&
5758             !((s && s->allow_opts) || r->allow_opts)) {
5759                 action = PF_DROP;
5760                 REASON_SET(&reason, PFRES_IPOPTIONS);
5761                 log = 1;
5762                 DPFPRINTF(PF_DEBUG_MISC,
5763                     ("pf: dropping packet with ip options\n"));
5764         }
5765
5766         if ((s && s->tag) || r->rtableid)
5767                 pf_tag_packet(m, s ? s->tag : 0, r->rtableid);
5768
5769 #ifdef ALTQ
5770         if (action == PF_PASS && r->qid) {
5771                 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
5772                 if (pqid || (pd.tos & IPTOS_LOWDELAY))
5773                         m->m_pkthdr.pf.qid = r->pqid;
5774                 else
5775                         m->m_pkthdr.pf.qid = r->qid;
5776                 m->m_pkthdr.pf.ecn_af = AF_INET;
5777                 m->m_pkthdr.pf.hdr = h;
5778                 /* add connection hash for fairq */
5779                 if (s) {
5780                         /* for fairq */
5781                         m->m_pkthdr.pf.state_hash = s->hash;
5782                         m->m_pkthdr.pf.flags |= PF_TAG_STATE_HASHED;
5783                 }
5784         }
5785 #endif /* ALTQ */
5786
5787         /*
5788          * connections redirected to loopback should not match sockets
5789          * bound specifically to loopback due to security implications,
5790          * see tcp_input() and in_pcblookup_listen().
5791          */
5792         if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
5793             pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
5794             (s->nat_rule.ptr->action == PF_RDR ||
5795             s->nat_rule.ptr->action == PF_BINAT) &&
5796             (ntohl(pd.dst->v4.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET)
5797                 m->m_pkthdr.pf.flags |= PF_TAG_TRANSLATE_LOCALHOST;
5798
5799         if (log) {
5800                 struct pf_rule *lr;
5801
5802                 if (s != NULL && s->nat_rule.ptr != NULL &&
5803                     s->nat_rule.ptr->log & PF_LOG_ALL)
5804                         lr = s->nat_rule.ptr;
5805                 else
5806                         lr = r;
5807                 PFLOG_PACKET(kif, h, m, AF_INET, dir, reason, lr, a, ruleset,
5808                     &pd);
5809         }
5810
5811         kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
5812         kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++;
5813
5814         if (action == PF_PASS || r->action == PF_DROP) {
5815                 dirndx = (dir == PF_OUT);
5816                 r->packets[dirndx]++;
5817                 r->bytes[dirndx] += pd.tot_len;
5818                 if (a != NULL) {
5819                         a->packets[dirndx]++;
5820                         a->bytes[dirndx] += pd.tot_len;
5821                 }
5822                 if (s != NULL) {
5823                         sk = s->state_key;
5824                         if (s->nat_rule.ptr != NULL) {
5825                                 s->nat_rule.ptr->packets[dirndx]++;
5826                                 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len;
5827                         }
5828                         if (s->src_node != NULL) {
5829                                 s->src_node->packets[dirndx]++;
5830                                 s->src_node->bytes[dirndx] += pd.tot_len;
5831                         }
5832                         if (s->nat_src_node != NULL) {
5833                                 s->nat_src_node->packets[dirndx]++;
5834                                 s->nat_src_node->bytes[dirndx] += pd.tot_len;
5835                         }
5836                         dirndx = (dir == sk->direction) ? 0 : 1;
5837                         s->packets[dirndx]++;
5838                         s->bytes[dirndx] += pd.tot_len;
5839                 }
5840                 tr = r;
5841                 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
5842                 if (nr != NULL) {
5843                         struct pf_addr *x;
5844                         /*
5845                          * XXX: we need to make sure that the addresses
5846                          * passed to pfr_update_stats() are the same than
5847                          * the addresses used during matching (pfr_match)
5848                          */
5849                         if (r == &pf_default_rule) {
5850                                 tr = nr;
5851                                 x = (sk == NULL || sk->direction == dir) ?
5852                                     &pd.baddr : &pd.naddr;
5853                         } else
5854                                 x = (sk == NULL || sk->direction == dir) ?
5855                                     &pd.naddr : &pd.baddr;
5856                         if (x == &pd.baddr || s == NULL) {
5857                                 /* we need to change the address */
5858                                 if (dir == PF_OUT)
5859                                         pd.src = x;
5860                                 else
5861                                         pd.dst = x;
5862                         }
5863                 }
5864                 if (tr->src.addr.type == PF_ADDR_TABLE)
5865                         pfr_update_stats(tr->src.addr.p.tbl, (sk == NULL ||
5866                             sk->direction == dir) ?
5867                             pd.src : pd.dst, pd.af,
5868                             pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5869                             tr->src.neg);
5870                 if (tr->dst.addr.type == PF_ADDR_TABLE)
5871                         pfr_update_stats(tr->dst.addr.p.tbl, (sk == NULL ||
5872                             sk->direction == dir) ? pd.dst : pd.src, pd.af,
5873                             pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5874                             tr->dst.neg);
5875         }
5876
5877
5878         if (action == PF_SYNPROXY_DROP) {
5879                 m_freem(*m0);
5880                 *m0 = NULL;
5881                 action = PF_PASS;
5882         } else if (r->rt)
5883                 /* pf_route can free the mbuf causing *m0 to become NULL */
5884                 pf_route(m0, r, dir, kif->pfik_ifp, s, &pd);
5885
5886         return (action);
5887 }
5888 #endif /* INET */
5889
5890 #ifdef INET6
5891 int
5892 pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0,
5893     struct ether_header *eh, struct inpcb *inp)
5894 {
5895         struct pfi_kif          *kif;
5896         u_short                  action, reason = 0, log = 0;
5897         struct mbuf             *m = *m0, *n = NULL;
5898         struct ip6_hdr          *h = NULL;
5899         struct pf_rule          *a = NULL, *r = &pf_default_rule, *tr, *nr;
5900         struct pf_state         *s = NULL;
5901         struct pf_state_key     *sk = NULL;
5902         struct pf_ruleset       *ruleset = NULL;
5903         struct pf_pdesc          pd;
5904         int                      off, terminal = 0, dirndx, rh_cnt = 0;
5905
5906         if (!pf_status.running)
5907                 return (PF_PASS);
5908
5909         memset(&pd, 0, sizeof(pd));
5910         if (ifp->if_type == IFT_CARP && ifp->if_carpdev)
5911                 kif = (struct pfi_kif *)ifp->if_carpdev->if_pf_kif;
5912         else
5913                 kif = (struct pfi_kif *)ifp->if_pf_kif;
5914
5915         if (kif == NULL) {
5916                 DPFPRINTF(PF_DEBUG_URGENT,
5917                     ("pf_test6: kif == NULL, if_xname %s\n", ifp->if_xname));
5918                 return (PF_DROP);
5919         }
5920         if (kif->pfik_flags & PFI_IFLAG_SKIP)
5921                 return (PF_PASS);
5922
5923 #ifdef DIAGNOSTIC
5924         if ((m->m_flags & M_PKTHDR) == 0)
5925                 panic("non-M_PKTHDR is passed to pf_test6");
5926 #endif /* DIAGNOSTIC */
5927
5928         if (m->m_pkthdr.len < (int)sizeof(*h)) {
5929                 action = PF_DROP;
5930                 REASON_SET(&reason, PFRES_SHORT);
5931                 log = 1;
5932                 goto done;
5933         }
5934
5935         if (m->m_pkthdr.pf.flags & PF_TAG_GENERATED)
5936                 return (PF_PASS);
5937
5938         /* We do IP header normalization and packet reassembly here */
5939         if (pf_normalize_ip6(m0, dir, kif, &reason, &pd) != PF_PASS) {
5940                 action = PF_DROP;
5941                 goto done;
5942         }
5943         m = *m0;        /* pf_normalize messes with m0 */
5944         h = mtod(m, struct ip6_hdr *);
5945
5946 #if 1
5947         /*
5948          * we do not support jumbogram yet.  if we keep going, zero ip6_plen
5949          * will do something bad, so drop the packet for now.
5950          */
5951         if (htons(h->ip6_plen) == 0) {
5952                 action = PF_DROP;
5953                 REASON_SET(&reason, PFRES_NORM);        /*XXX*/
5954                 goto done;
5955         }
5956 #endif
5957
5958         pd.src = (struct pf_addr *)&h->ip6_src;
5959         pd.dst = (struct pf_addr *)&h->ip6_dst;
5960         PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET6);
5961         pd.ip_sum = NULL;
5962         pd.af = AF_INET6;
5963         pd.tos = 0;
5964         pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr);
5965         pd.eh = eh;
5966
5967         off = ((caddr_t)h - m->m_data) + sizeof(struct ip6_hdr);
5968         pd.proto = h->ip6_nxt;
5969         do {
5970                 switch (pd.proto) {
5971                 case IPPROTO_FRAGMENT:
5972                         action = pf_test_fragment(&r, dir, kif, m, h,
5973                             &pd, &a, &ruleset);
5974                         if (action == PF_DROP)
5975                                 REASON_SET(&reason, PFRES_FRAG);
5976                         goto done;
5977                 case IPPROTO_ROUTING: {
5978                         struct ip6_rthdr rthdr;
5979
5980                         if (rh_cnt++) {
5981                                 DPFPRINTF(PF_DEBUG_MISC,
5982                                     ("pf: IPv6 more than one rthdr\n"));
5983                                 action = PF_DROP;
5984                                 REASON_SET(&reason, PFRES_IPOPTIONS);
5985                                 log = 1;
5986                                 goto done;
5987                         }
5988                         if (!pf_pull_hdr(m, off, &rthdr, sizeof(rthdr), NULL,
5989                             &reason, pd.af)) {
5990                                 DPFPRINTF(PF_DEBUG_MISC,
5991                                     ("pf: IPv6 short rthdr\n"));
5992                                 action = PF_DROP;
5993                                 REASON_SET(&reason, PFRES_SHORT);
5994                                 log = 1;
5995                                 goto done;
5996                         }
5997                         if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) {
5998                                 DPFPRINTF(PF_DEBUG_MISC,
5999                                     ("pf: IPv6 rthdr0\n"));
6000                                 action = PF_DROP;
6001                                 REASON_SET(&reason, PFRES_IPOPTIONS);
6002                                 log = 1;
6003                                 goto done;
6004                         }
6005                         /* FALLTHROUGH */
6006                 }
6007                 case IPPROTO_AH:
6008                 case IPPROTO_HOPOPTS:
6009                 case IPPROTO_DSTOPTS: {
6010                         /* get next header and header length */
6011                         struct ip6_ext  opt6;
6012
6013                         if (!pf_pull_hdr(m, off, &opt6, sizeof(opt6),
6014                             NULL, &reason, pd.af)) {
6015                                 DPFPRINTF(PF_DEBUG_MISC,
6016                                     ("pf: IPv6 short opt\n"));
6017                                 action = PF_DROP;
6018                                 log = 1;
6019                                 goto done;
6020                         }
6021                         if (pd.proto == IPPROTO_AH)
6022                                 off += (opt6.ip6e_len + 2) * 4;
6023                         else
6024                                 off += (opt6.ip6e_len + 1) * 8;
6025                         pd.proto = opt6.ip6e_nxt;
6026                         /* goto the next header */
6027                         break;
6028                 }
6029                 default:
6030                         terminal++;
6031                         break;
6032                 }
6033         } while (!terminal);
6034
6035         /* if there's no routing header, use unmodified mbuf for checksumming */
6036         if (!n)
6037                 n = m;
6038
6039         switch (pd.proto) {
6040
6041         case IPPROTO_TCP: {
6042                 struct tcphdr   th;
6043
6044                 pd.hdr.tcp = &th;
6045                 if (!pf_pull_hdr(m, off, &th, sizeof(th),
6046                     &action, &reason, AF_INET6)) {
6047                         log = action != PF_PASS;
6048                         goto done;
6049                 }
6050                 pd.p_len = pd.tot_len - off - (th.th_off << 2);
6051                 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
6052                 if (action == PF_DROP)
6053                         goto done;
6054                 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
6055                     &reason);
6056                 if (action == PF_PASS) {
6057 #if NPFSYNC
6058                         pfsync_update_state(s);
6059 #endif /* NPFSYNC */
6060                         r = s->rule.ptr;
6061                         a = s->anchor.ptr;
6062                         log = s->log;
6063                 } else if (s == NULL)
6064                         action = pf_test_rule(&r, &s, dir, kif,
6065                             m, off, h, &pd, &a, &ruleset, NULL, inp);
6066                 break;
6067         }
6068
6069         case IPPROTO_UDP: {
6070                 struct udphdr   uh;
6071
6072                 pd.hdr.udp = &uh;
6073                 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
6074                     &action, &reason, AF_INET6)) {
6075                         log = action != PF_PASS;
6076                         goto done;
6077                 }
6078                 if (uh.uh_dport == 0 ||
6079                     ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
6080                     ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
6081                         action = PF_DROP;
6082                         REASON_SET(&reason, PFRES_SHORT);
6083                         goto done;
6084                 }
6085                 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
6086                 if (action == PF_PASS) {
6087 #if NPFSYNC
6088                         pfsync_update_state(s);
6089 #endif /* NPFSYNC */
6090                         r = s->rule.ptr;
6091                         a = s->anchor.ptr;
6092                         log = s->log;
6093                 } else if (s == NULL)
6094                         action = pf_test_rule(&r, &s, dir, kif,
6095                             m, off, h, &pd, &a, &ruleset, NULL, inp);
6096                 break;
6097         }
6098
6099         case IPPROTO_ICMPV6: {
6100                 struct icmp6_hdr        ih;
6101
6102                 pd.hdr.icmp6 = &ih;
6103                 if (!pf_pull_hdr(m, off, &ih, sizeof(ih),
6104                     &action, &reason, AF_INET6)) {
6105                         log = action != PF_PASS;
6106                         goto done;
6107                 }
6108                 action = pf_test_state_icmp(&s, dir, kif,
6109                     m, off, h, &pd, &reason);
6110                 if (action == PF_PASS) {
6111 #if NPFSYNC
6112                         pfsync_update_state(s);
6113 #endif /* NPFSYNC */
6114                         r = s->rule.ptr;
6115                         a = s->anchor.ptr;
6116                         log = s->log;
6117                 } else if (s == NULL)
6118                         action = pf_test_rule(&r, &s, dir, kif,
6119                             m, off, h, &pd, &a, &ruleset, NULL, inp);
6120                 break;
6121         }
6122
6123         default:
6124                 action = pf_test_state_other(&s, dir, kif, &pd);
6125                 if (action == PF_PASS) {
6126 #if NPFSYNC
6127                         pfsync_update_state(s);
6128 #endif /* NPFSYNC */
6129                         r = s->rule.ptr;
6130                         a = s->anchor.ptr;
6131                         log = s->log;
6132                 } else if (s == NULL)
6133                         action = pf_test_rule(&r, &s, dir, kif, m, off, h,
6134                             &pd, &a, &ruleset, NULL, inp);
6135                 break;
6136         }
6137
6138 done:
6139         if (n != m) {
6140                 m_freem(n);
6141                 n = NULL;
6142         }
6143
6144         /* handle dangerous IPv6 extension headers. */
6145         if (action == PF_PASS && rh_cnt &&
6146             !((s && s->allow_opts) || r->allow_opts)) {
6147                 action = PF_DROP;
6148                 REASON_SET(&reason, PFRES_IPOPTIONS);
6149                 log = 1;
6150                 DPFPRINTF(PF_DEBUG_MISC,
6151                     ("pf: dropping packet with dangerous v6 headers\n"));
6152         }
6153
6154         if ((s && s->tag) || r->rtableid)
6155                 pf_tag_packet(m, s ? s->tag : 0, r->rtableid);
6156
6157 #ifdef ALTQ
6158         if (action == PF_PASS && r->qid) {
6159                 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
6160                 if (pd.tos & IPTOS_LOWDELAY)
6161                         m->m_pkthdr.pf.qid = r->pqid;
6162                 else
6163                         m->m_pkthdr.pf.qid = r->qid;
6164                 m->m_pkthdr.pf.ecn_af = AF_INET6;
6165                 m->m_pkthdr.pf.hdr = h;
6166                 if (s) {
6167                         /* for fairq */
6168                         m->m_pkthdr.pf.state_hash = s->hash;
6169                         m->m_pkthdr.pf.flags |= PF_TAG_STATE_HASHED;
6170                 }
6171         }
6172 #endif /* ALTQ */
6173
6174         if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
6175             pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
6176             (s->nat_rule.ptr->action == PF_RDR ||
6177             s->nat_rule.ptr->action == PF_BINAT) &&
6178             IN6_IS_ADDR_LOOPBACK(&pd.dst->v6))
6179                 m->m_pkthdr.pf.flags |= PF_TAG_TRANSLATE_LOCALHOST;
6180
6181         if (log) {
6182                 struct pf_rule *lr;
6183
6184                 if (s != NULL && s->nat_rule.ptr != NULL &&
6185                     s->nat_rule.ptr->log & PF_LOG_ALL)
6186                         lr = s->nat_rule.ptr;
6187                 else
6188                         lr = r;
6189                 PFLOG_PACKET(kif, h, m, AF_INET6, dir, reason, lr, a, ruleset,
6190                     &pd);
6191         }
6192
6193         kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
6194         kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++;
6195
6196         if (action == PF_PASS || r->action == PF_DROP) {
6197                 dirndx = (dir == PF_OUT);
6198                 r->packets[dirndx]++;
6199                 r->bytes[dirndx] += pd.tot_len;
6200                 if (a != NULL) {
6201                         a->packets[dirndx]++;
6202                         a->bytes[dirndx] += pd.tot_len;
6203                 }
6204                 if (s != NULL) {
6205                         sk = s->state_key;
6206                         if (s->nat_rule.ptr != NULL) {
6207                                 s->nat_rule.ptr->packets[dirndx]++;
6208                                 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len;
6209                         }
6210                         if (s->src_node != NULL) {
6211                                 s->src_node->packets[dirndx]++;
6212                                 s->src_node->bytes[dirndx] += pd.tot_len;
6213                         }
6214                         if (s->nat_src_node != NULL) {
6215                                 s->nat_src_node->packets[dirndx]++;
6216                                 s->nat_src_node->bytes[dirndx] += pd.tot_len;
6217                         }
6218                         dirndx = (dir == sk->direction) ? 0 : 1;
6219                         s->packets[dirndx]++;
6220                         s->bytes[dirndx] += pd.tot_len;
6221                 }
6222                 tr = r;
6223                 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
6224                 if (nr != NULL) {
6225                         struct pf_addr *x;
6226                         /*
6227                          * XXX: we need to make sure that the addresses
6228                          * passed to pfr_update_stats() are the same than
6229                          * the addresses used during matching (pfr_match)
6230                          */
6231                         if (r == &pf_default_rule) {
6232                                 tr = nr;
6233                                 x = (s == NULL || sk->direction == dir) ?
6234                                     &pd.baddr : &pd.naddr;
6235                         } else {
6236                                 x = (s == NULL || sk->direction == dir) ?
6237                                     &pd.naddr : &pd.baddr;
6238                         }
6239                         if (x == &pd.baddr || s == NULL) {
6240                                 if (dir == PF_OUT)
6241                                         pd.src = x;
6242                                 else
6243                                         pd.dst = x;
6244                         }
6245                 }
6246                 if (tr->src.addr.type == PF_ADDR_TABLE)
6247                         pfr_update_stats(tr->src.addr.p.tbl, (sk == NULL ||
6248                             sk->direction == dir) ? pd.src : pd.dst, pd.af,
6249                             pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
6250                             tr->src.neg);
6251                 if (tr->dst.addr.type == PF_ADDR_TABLE)
6252                         pfr_update_stats(tr->dst.addr.p.tbl, (sk == NULL ||
6253                             sk->direction == dir) ? pd.dst : pd.src, pd.af,
6254                             pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
6255                             tr->dst.neg);
6256         }
6257
6258
6259         if (action == PF_SYNPROXY_DROP) {
6260                 m_freem(*m0);
6261                 *m0 = NULL;
6262                 action = PF_PASS;
6263         } else if (r->rt)
6264                 /* pf_route6 can free the mbuf causing *m0 to become NULL */
6265                 pf_route6(m0, r, dir, kif->pfik_ifp, s, &pd);
6266
6267         return (action);
6268 }
6269 #endif /* INET6 */
6270
6271 int
6272 pf_check_congestion(struct ifqueue *ifq)
6273 {
6274                 return (0);
6275 }