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 struct lwkt_token pf_token = LWKT_TOKEN_MP_INITIALIZER(pf_token);
 114
 115 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) kprintf x
 116
 117 /*
 118  * Global variables
 119  */
 120
 121 /* state tables */
 122 struct pf_state_tree_lan_ext     pf_statetbl_lan_ext;
 123 struct pf_state_tree_ext_gwy     pf_statetbl_ext_gwy;
 124
 125 struct pf_altqqueue      pf_altqs[2];
 126 struct pf_palist         pf_pabuf;
 127 struct pf_altqqueue     *pf_altqs_active;
 128 struct pf_altqqueue     *pf_altqs_inactive;
 129 struct pf_status         pf_status;
 130
 131 u_int32_t                ticket_altqs_active;
 132 u_int32_t                ticket_altqs_inactive;
 133 int                      altqs_inactive_open;
 134 u_int32_t                ticket_pabuf;
 135
 136 struct pf_anchor_stackframe {
 137         struct pf_ruleset                       *rs;
 138         struct pf_rule                          *r;
 139         struct pf_anchor_node                   *parent;
 140         struct pf_anchor                        *child;
 141 } pf_anchor_stack[64];
 142
 143 vm_zone_t                pf_src_tree_pl, pf_rule_pl, pf_pooladdr_pl;
 144 vm_zone_t                pf_state_pl, pf_state_key_pl;
 145 vm_zone_t                pf_altq_pl;
 146
 147 void                     pf_print_host(struct pf_addr *, u_int16_t, u_int8_t);
 148
 149 void                     pf_init_threshold(struct pf_threshold *, u_int32_t,
 150                             u_int32_t);
 151 void                     pf_add_threshold(struct pf_threshold *);
 152 int                      pf_check_threshold(struct pf_threshold *);
 153
 154 void                     pf_change_ap(struct pf_addr *, u_int16_t *,
 155                             u_int16_t *, u_int16_t *, struct pf_addr *,
 156                             u_int16_t, u_int8_t, sa_family_t);
 157 int                      pf_modulate_sack(struct mbuf *, int, struct pf_pdesc *,
 158                             struct tcphdr *, struct pf_state_peer *);
 159 #ifdef INET6
 160 void                     pf_change_a6(struct pf_addr *, u_int16_t *,
 161                             struct pf_addr *, u_int8_t);
 162 #endif /* INET6 */
 163 void                     pf_change_icmp(struct pf_addr *, u_int16_t *,
 164                             struct pf_addr *, struct pf_addr *, u_int16_t,
 165                             u_int16_t *, u_int16_t *, u_int16_t *,
 166                             u_int16_t *, u_int8_t, sa_family_t);
 167 void                     pf_send_tcp(const struct pf_rule *, sa_family_t,
 168                             const struct pf_addr *, const struct pf_addr *,
 169                             u_int16_t, u_int16_t, u_int32_t, u_int32_t,
 170                             u_int8_t, u_int16_t, u_int16_t, u_int8_t, int,
 171                             u_int16_t, struct ether_header *, struct ifnet *);
 172 void                     pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t,
 173                             sa_family_t, struct pf_rule *);
 174 struct pf_rule          *pf_match_translation(struct pf_pdesc *, struct mbuf *,
 175                             int, int, struct pfi_kif *,
 176                             struct pf_addr *, u_int16_t, struct pf_addr *,
 177                             u_int16_t, int);
 178 struct pf_rule          *pf_get_translation(struct pf_pdesc *, struct mbuf *,
 179                             int, int, struct pfi_kif *, struct pf_src_node **,
 180                             struct pf_addr *, u_int16_t,
 181                             struct pf_addr *, u_int16_t,
 182                             struct pf_addr *, u_int16_t *);
 183 void                     pf_attach_state(struct pf_state_key *,
 184                             struct pf_state *, int);
 185 void                     pf_detach_state(struct pf_state *, int);
 186 int                      pf_test_rule(struct pf_rule **, struct pf_state **,
 187                             int, struct pfi_kif *, struct mbuf *, int,
 188                             void *, struct pf_pdesc *, struct pf_rule **,
 189                             struct pf_ruleset **, struct ifqueue *, struct inpcb *);
 190 int                      pf_test_fragment(struct pf_rule **, int,
 191                             struct pfi_kif *, struct mbuf *, void *,
 192                             struct pf_pdesc *, struct pf_rule **,
 193                             struct pf_ruleset **);
 194 int                      pf_test_state_tcp(struct pf_state **, int,
 195                             struct pfi_kif *, struct mbuf *, int,
 196                             void *, struct pf_pdesc *, u_short *);
 197 int                      pf_test_state_udp(struct pf_state **, int,
 198                             struct pfi_kif *, struct mbuf *, int,
 199                             void *, struct pf_pdesc *);
 200 int                      pf_test_state_icmp(struct pf_state **, int,
 201                             struct pfi_kif *, struct mbuf *, int,
 202                             void *, struct pf_pdesc *, u_short *);
 203 int                      pf_test_state_other(struct pf_state **, int,
 204                             struct pfi_kif *, struct pf_pdesc *);
 205 int                      pf_match_tag(struct mbuf *, struct pf_rule *, int *);
 206 void                     pf_step_into_anchor(int *, struct pf_ruleset **, int,
 207                             struct pf_rule **, struct pf_rule **,  int *);
 208 int                      pf_step_out_of_anchor(int *, struct pf_ruleset **,
 209                              int, struct pf_rule **, struct pf_rule **,
 210                              int *);
 211 void                     pf_hash(struct pf_addr *, struct pf_addr *,
 212                             struct pf_poolhashkey *, sa_family_t);
 213 int                      pf_map_addr(u_int8_t, struct pf_rule *,
 214                             struct pf_addr *, struct pf_addr *,
 215                             struct pf_addr *, struct pf_src_node **);
 216 int                      pf_get_sport(sa_family_t, u_int8_t, struct pf_rule *,
 217                             struct pf_addr *, struct pf_addr *, u_int16_t,
 218                             struct pf_addr *, u_int16_t*, u_int16_t, u_int16_t,
 219                             struct pf_src_node **);
 220 void                     pf_route(struct mbuf **, struct pf_rule *, int,
 221                             struct ifnet *, struct pf_state *,
 222                             struct pf_pdesc *);
 223 void                     pf_route6(struct mbuf **, struct pf_rule *, int,
 224                             struct ifnet *, struct pf_state *,
 225                             struct pf_pdesc *);
 226 u_int8_t                 pf_get_wscale(struct mbuf *, int, u_int16_t,
 227                             sa_family_t);
 228 u_int16_t                pf_get_mss(struct mbuf *, int, u_int16_t,
 229                             sa_family_t);
 230 u_int16_t                pf_calc_mss(struct pf_addr *, sa_family_t,
 231                                 u_int16_t);
 232 void                     pf_set_rt_ifp(struct pf_state *,
 233                             struct pf_addr *);
 234 int                      pf_check_proto_cksum(struct mbuf *, int, int,
 235                             u_int8_t, sa_family_t);
 236 int                      pf_addr_wrap_neq(struct pf_addr_wrap *,
 237                             struct pf_addr_wrap *);
 238 struct pf_state         *pf_find_state(struct pfi_kif *,
 239                             struct pf_state_key_cmp *, u_int8_t);
 240 int                      pf_src_connlimit(struct pf_state **);
 241 void                     pf_stateins_err(const char *, struct pf_state *,
 242                             struct pfi_kif *);
 243 int                      pf_check_congestion(struct ifqueue *);
 244
 245 extern int pf_end_threads;
 246
 247 struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] = {
 248         { &pf_state_pl, PFSTATE_HIWAT },
 249         { &pf_src_tree_pl, PFSNODE_HIWAT },
 250         { &pf_frent_pl, PFFRAG_FRENT_HIWAT },
 251         { &pfr_ktable_pl, PFR_KTABLE_HIWAT },
 252         { &pfr_kentry_pl, PFR_KENTRY_HIWAT }
 253 };
 254
 255 #define STATE_LOOKUP()                                                  \
 256         do {                                                            \
 257                 if (direction == PF_IN)                                 \
 258                         *state = pf_find_state(kif, &key, PF_EXT_GWY);  \
 259                 else                                                    \
 260                         *state = pf_find_state(kif, &key, PF_LAN_EXT);  \
 261                 if (*state == NULL || (*state)->timeout == PFTM_PURGE)  \
 262                         return (PF_DROP);                               \
 263                 if (direction == PF_OUT &&                              \
 264                     (((*state)->rule.ptr->rt == PF_ROUTETO &&           \
 265                     (*state)->rule.ptr->direction == PF_OUT) ||         \
 266                     ((*state)->rule.ptr->rt == PF_REPLYTO &&            \
 267                     (*state)->rule.ptr->direction == PF_IN)) &&         \
 268                     (*state)->rt_kif != NULL &&                         \
 269                     (*state)->rt_kif != kif)                            \
 270                         return (PF_PASS);                               \
 271         } while (0)
 272
 273 #define STATE_TRANSLATE(sk) \
 274         (sk)->lan.addr.addr32[0] != (sk)->gwy.addr.addr32[0] || \
 275         ((sk)->af == AF_INET6 && \
 276         ((sk)->lan.addr.addr32[1] != (sk)->gwy.addr.addr32[1] || \
 277         (sk)->lan.addr.addr32[2] != (sk)->gwy.addr.addr32[2] || \
 278         (sk)->lan.addr.addr32[3] != (sk)->gwy.addr.addr32[3])) || \
 279         (sk)->lan.port != (sk)->gwy.port
 280
 281 #define BOUND_IFACE(r, k) \
 282         ((r)->rule_flag & PFRULE_IFBOUND) ? (k) : pfi_all
 283
 284 #define STATE_INC_COUNTERS(s)                           \
 285         do {                                            \
 286                 s->rule.ptr->states++;                  \
 287                 if (s->anchor.ptr != NULL)              \
 288                         s->anchor.ptr->states++;        \
 289                 if (s->nat_rule.ptr != NULL)            \
 290                         s->nat_rule.ptr->states++;      \
 291         } while (0)
 292
 293 #define STATE_DEC_COUNTERS(s)                           \
 294         do {                                            \
 295                 if (s->nat_rule.ptr != NULL)            \
 296                         s->nat_rule.ptr->states--;      \
 297                 if (s->anchor.ptr != NULL)              \
 298                         s->anchor.ptr->states--;        \
 299                 s->rule.ptr->states--;                  \
 300         } while (0)
 301
 302 static __inline int pf_src_compare(struct pf_src_node *, struct pf_src_node *);
 303 static __inline int pf_state_compare_lan_ext(struct pf_state_key *,
 304         struct pf_state_key *);
 305 static __inline int pf_state_compare_ext_gwy(struct pf_state_key *,
 306         struct pf_state_key *);
 307 static __inline int pf_state_compare_id(struct pf_state *,
 308         struct pf_state *);
 309
 310 struct pf_src_tree tree_src_tracking;
 311
 312 struct pf_state_tree_id tree_id;
 313 struct pf_state_queue state_list;
 314
 315 RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare);
 316 RB_GENERATE(pf_state_tree_lan_ext, pf_state_key,
 317     entry_lan_ext, pf_state_compare_lan_ext);
 318 RB_GENERATE(pf_state_tree_ext_gwy, pf_state_key,
 319     entry_ext_gwy, pf_state_compare_ext_gwy);
 320 RB_GENERATE(pf_state_tree_id, pf_state,
 321     entry_id, pf_state_compare_id);
 322
 323 #define PF_DT_SKIP_LANEXT       0x01
 324 #define PF_DT_SKIP_EXTGWY       0x02
 325
 326 static __inline int
 327 pf_src_compare(struct pf_src_node *a, struct pf_src_node *b)
 328 {
 329         int     diff;
 330
 331         if (a->rule.ptr > b->rule.ptr)
 332                 return (1);
 333         if (a->rule.ptr < b->rule.ptr)
 334                 return (-1);
 335         if ((diff = a->af - b->af) != 0)
 336                 return (diff);
 337         switch (a->af) {
 338 #ifdef INET
 339         case AF_INET:
 340                 if (a->addr.addr32[0] > b->addr.addr32[0])
 341                         return (1);
 342                 if (a->addr.addr32[0] < b->addr.addr32[0])
 343                         return (-1);
 344                 break;
 345 #endif /* INET */
 346 #ifdef INET6
 347         case AF_INET6:
 348                 if (a->addr.addr32[3] > b->addr.addr32[3])
 349                         return (1);
 350                 if (a->addr.addr32[3] < b->addr.addr32[3])
 351                         return (-1);
 352                 if (a->addr.addr32[2] > b->addr.addr32[2])
 353                         return (1);
 354                 if (a->addr.addr32[2] < b->addr.addr32[2])
 355                         return (-1);
 356                 if (a->addr.addr32[1] > b->addr.addr32[1])
 357                         return (1);
 358                 if (a->addr.addr32[1] < b->addr.addr32[1])
 359                         return (-1);
 360                 if (a->addr.addr32[0] > b->addr.addr32[0])
 361                         return (1);
 362                 if (a->addr.addr32[0] < b->addr.addr32[0])
 363                         return (-1);
 364                 break;
 365 #endif /* INET6 */
 366         }
 367         return (0);
 368 }
 369
 370 u_int32_t
 371 pf_state_hash(struct pf_state_key *sk)
 372 {
 373         u_int32_t hv = (intptr_t)sk / sizeof(*sk);
 374
 375         hv ^= crc32(&sk->lan, sizeof(sk->lan));
 376         hv ^= crc32(&sk->gwy, sizeof(sk->gwy));
 377         hv ^= crc32(&sk->ext, sizeof(sk->ext));
 378         if (hv == 0)    /* disallow 0 */
 379                 hv = 1;
 380         return(hv);
 381 }
 382
 383 static __inline int
 384 pf_state_compare_lan_ext(struct pf_state_key *a, struct pf_state_key *b)
 385 {
 386         int     diff;
 387
 388         if ((diff = a->proto - b->proto) != 0)
 389                 return (diff);
 390         if ((diff = a->af - b->af) != 0)
 391                 return (diff);
 392         switch (a->af) {
 393 #ifdef INET
 394         case AF_INET:
 395                 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
 396                         return (1);
 397                 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
 398                         return (-1);
 399                 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
 400                         return (1);
 401                 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
 402                         return (-1);
 403                 break;
 404 #endif /* INET */
 405 #ifdef INET6
 406         case AF_INET6:
 407                 if (a->lan.addr.addr32[3] > b->lan.addr.addr32[3])
 408                         return (1);
 409                 if (a->lan.addr.addr32[3] < b->lan.addr.addr32[3])
 410                         return (-1);
 411                 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
 412                         return (1);
 413                 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
 414                         return (-1);
 415                 if (a->lan.addr.addr32[2] > b->lan.addr.addr32[2])
 416                         return (1);
 417                 if (a->lan.addr.addr32[2] < b->lan.addr.addr32[2])
 418                         return (-1);
 419                 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
 420                         return (1);
 421                 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
 422                         return (-1);
 423                 if (a->lan.addr.addr32[1] > b->lan.addr.addr32[1])
 424                         return (1);
 425                 if (a->lan.addr.addr32[1] < b->lan.addr.addr32[1])
 426                         return (-1);
 427                 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
 428                         return (1);
 429                 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
 430                         return (-1);
 431                 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
 432                         return (1);
 433                 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
 434                         return (-1);
 435                 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
 436                         return (1);
 437                 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
 438                         return (-1);
 439                 break;
 440 #endif /* INET6 */
 441         }
 442
 443         if ((diff = a->lan.port - b->lan.port) != 0)
 444                 return (diff);
 445         if ((diff = a->ext.port - b->ext.port) != 0)
 446                 return (diff);
 447
 448         return (0);
 449 }
 450
 451 static __inline int
 452 pf_state_compare_ext_gwy(struct pf_state_key *a, struct pf_state_key *b)
 453 {
 454         int     diff;
 455
 456         if ((diff = a->proto - b->proto) != 0)
 457                 return (diff);
 458         if ((diff = a->af - b->af) != 0)
 459                 return (diff);
 460         switch (a->af) {
 461 #ifdef INET
 462         case AF_INET:
 463                 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
 464                         return (1);
 465                 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
 466                         return (-1);
 467                 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
 468                         return (1);
 469                 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
 470                         return (-1);
 471                 break;
 472 #endif /* INET */
 473 #ifdef INET6
 474         case AF_INET6:
 475                 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
 476                         return (1);
 477                 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
 478                         return (-1);
 479                 if (a->gwy.addr.addr32[3] > b->gwy.addr.addr32[3])
 480                         return (1);
 481                 if (a->gwy.addr.addr32[3] < b->gwy.addr.addr32[3])
 482                         return (-1);
 483                 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
 484                         return (1);
 485                 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
 486                         return (-1);
 487                 if (a->gwy.addr.addr32[2] > b->gwy.addr.addr32[2])
 488                         return (1);
 489                 if (a->gwy.addr.addr32[2] < b->gwy.addr.addr32[2])
 490                         return (-1);
 491                 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
 492                         return (1);
 493                 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
 494                         return (-1);
 495                 if (a->gwy.addr.addr32[1] > b->gwy.addr.addr32[1])
 496                         return (1);
 497                 if (a->gwy.addr.addr32[1] < b->gwy.addr.addr32[1])
 498                         return (-1);
 499                 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
 500                         return (1);
 501                 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
 502                         return (-1);
 503                 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
 504                         return (1);
 505                 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
 506                         return (-1);
 507                 break;
 508 #endif /* INET6 */
 509         }
 510
 511         if ((diff = a->ext.port - b->ext.port) != 0)
 512                 return (diff);
 513         if ((diff = a->gwy.port - b->gwy.port) != 0)
 514                 return (diff);
 515
 516         return (0);
 517 }
 518
 519 static __inline int
 520 pf_state_compare_id(struct pf_state *a, struct pf_state *b)
 521 {
 522         if (a->id > b->id)
 523                 return (1);
 524         if (a->id < b->id)
 525                 return (-1);
 526         if (a->creatorid > b->creatorid)
 527                 return (1);
 528         if (a->creatorid < b->creatorid)
 529                 return (-1);
 530
 531         return (0);
 532 }
 533
 534 #ifdef INET6
 535 void
 536 pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af)
 537 {
 538         switch (af) {
 539 #ifdef INET
 540         case AF_INET:
 541                 dst->addr32[0] = src->addr32[0];
 542                 break;
 543 #endif /* INET */
 544         case AF_INET6:
 545                 dst->addr32[0] = src->addr32[0];
 546                 dst->addr32[1] = src->addr32[1];
 547                 dst->addr32[2] = src->addr32[2];
 548                 dst->addr32[3] = src->addr32[3];
 549                 break;
 550         }
 551 }
 552 #endif /* INET6 */
 553
 554 struct pf_state *
 555 pf_find_state_byid(struct pf_state_cmp *key)
 556 {
 557         pf_status.fcounters[FCNT_STATE_SEARCH]++;
 558
 559         return (RB_FIND(pf_state_tree_id, &tree_id, (struct pf_state *)key));
 560 }
 561
 562 struct pf_state *
 563 pf_find_state(struct pfi_kif *kif, struct pf_state_key_cmp *key, u_int8_t tree)
 564 {
 565         struct pf_state_key     *sk;
 566         struct pf_state         *s;
 567
 568         pf_status.fcounters[FCNT_STATE_SEARCH]++;
 569
 570         switch (tree) {
 571         case PF_LAN_EXT:
 572                 sk = RB_FIND(pf_state_tree_lan_ext, &pf_statetbl_lan_ext,
 573                     (struct pf_state_key *)key);
 574                 break;
 575         case PF_EXT_GWY:
 576                 sk = RB_FIND(pf_state_tree_ext_gwy, &pf_statetbl_ext_gwy,
 577                     (struct pf_state_key *)key);
 578                 break;
 579         default:
 580                 panic("pf_find_state");
 581         }
 582
 583         /* list is sorted, if-bound states before floating ones */
 584         if (sk != NULL)
 585                 TAILQ_FOREACH(s, &sk->states, next)
 586                         if (s->kif == pfi_all || s->kif == kif)
 587                                 return (s);
 588
 589         return (NULL);
 590 }
 591
 592 struct pf_state *
 593 pf_find_state_all(struct pf_state_key_cmp *key, u_int8_t tree, int *more)
 594 {
 595         struct pf_state_key     *sk;
 596         struct pf_state         *s, *ret = NULL;
 597
 598         pf_status.fcounters[FCNT_STATE_SEARCH]++;
 599
 600         switch (tree) {
 601         case PF_LAN_EXT:
 602                 sk = RB_FIND(pf_state_tree_lan_ext,
 603                     &pf_statetbl_lan_ext, (struct pf_state_key *)key);
 604                 break;
 605         case PF_EXT_GWY:
 606                 sk = RB_FIND(pf_state_tree_ext_gwy,
 607                     &pf_statetbl_ext_gwy, (struct pf_state_key *)key);
 608                 break;
 609         default:
 610                 panic("pf_find_state_all");
 611         }
 612
 613         if (sk != NULL) {
 614                 ret = TAILQ_FIRST(&sk->states);
 615                 if (more == NULL)
 616                         return (ret);
 617
 618                 TAILQ_FOREACH(s, &sk->states, next)
 619                         (*more)++;
 620         }
 621
 622         return (ret);
 623 }
 624
 625 void
 626 pf_init_threshold(struct pf_threshold *threshold,
 627     u_int32_t limit, u_int32_t seconds)
 628 {
 629         threshold->limit = limit * PF_THRESHOLD_MULT;
 630         threshold->seconds = seconds;
 631         threshold->count = 0;
 632         threshold->last = time_second;
 633 }
 634
 635 void
 636 pf_add_threshold(struct pf_threshold *threshold)
 637 {
 638         u_int32_t t = time_second, diff = t - threshold->last;
 639
 640         if (diff >= threshold->seconds)
 641                 threshold->count = 0;
 642         else
 643                 threshold->count -= threshold->count * diff /
 644                     threshold->seconds;
 645         threshold->count += PF_THRESHOLD_MULT;
 646         threshold->last = t;
 647 }
 648
 649 int
 650 pf_check_threshold(struct pf_threshold *threshold)
 651 {
 652         return (threshold->count > threshold->limit);
 653 }
 654
 655 int
 656 pf_src_connlimit(struct pf_state **state)
 657 {
 658         int bad = 0;
 659
 660         (*state)->src_node->conn++;
 661         (*state)->src.tcp_est = 1;
 662         pf_add_threshold(&(*state)->src_node->conn_rate);
 663
 664         if ((*state)->rule.ptr->max_src_conn &&
 665             (*state)->rule.ptr->max_src_conn <
 666             (*state)->src_node->conn) {
 667                 pf_status.lcounters[LCNT_SRCCONN]++;
 668                 bad++;
 669         }
 670
 671         if ((*state)->rule.ptr->max_src_conn_rate.limit &&
 672             pf_check_threshold(&(*state)->src_node->conn_rate)) {
 673                 pf_status.lcounters[LCNT_SRCCONNRATE]++;
 674                 bad++;
 675         }
 676
 677         if (!bad)
 678                 return (0);
 679
 680         if ((*state)->rule.ptr->overload_tbl) {
 681                 struct pfr_addr p;
 682                 u_int32_t       killed = 0;
 683
 684                 pf_status.lcounters[LCNT_OVERLOAD_TABLE]++;
 685                 if (pf_status.debug >= PF_DEBUG_MISC) {
 686                         kprintf("pf_src_connlimit: blocking address ");
 687                         pf_print_host(&(*state)->src_node->addr, 0,
 688                             (*state)->state_key->af);
 689                 }
 690
 691                 bzero(&p, sizeof(p));
 692                 p.pfra_af = (*state)->state_key->af;
 693                 switch ((*state)->state_key->af) {
 694 #ifdef INET
 695                 case AF_INET:
 696                         p.pfra_net = 32;
 697                         p.pfra_ip4addr = (*state)->src_node->addr.v4;
 698                         break;
 699 #endif /* INET */
 700 #ifdef INET6
 701                 case AF_INET6:
 702                         p.pfra_net = 128;
 703                         p.pfra_ip6addr = (*state)->src_node->addr.v6;
 704                         break;
 705 #endif /* INET6 */
 706                 }
 707
 708                 pfr_insert_kentry((*state)->rule.ptr->overload_tbl,
 709                     &p, time_second);
 710
 711                 /* kill existing states if that's required. */
 712                 if ((*state)->rule.ptr->flush) {
 713                         struct pf_state_key *sk;
 714                         struct pf_state *st;
 715
 716                         pf_status.lcounters[LCNT_OVERLOAD_FLUSH]++;
 717                         RB_FOREACH(st, pf_state_tree_id, &tree_id) {
 718                                 sk = st->state_key;
 719                                 /*
 720                                  * Kill states from this source.  (Only those
 721                                  * from the same rule if PF_FLUSH_GLOBAL is not
 722                                  * set)
 723                                  */
 724                                 if (sk->af ==
 725                                     (*state)->state_key->af &&
 726                                     (((*state)->state_key->direction ==
 727                                         PF_OUT &&
 728                                     PF_AEQ(&(*state)->src_node->addr,
 729                                         &sk->lan.addr, sk->af)) ||
 730                                     ((*state)->state_key->direction == PF_IN &&
 731                                     PF_AEQ(&(*state)->src_node->addr,
 732                                         &sk->ext.addr, sk->af))) &&
 733                                     ((*state)->rule.ptr->flush &
 734                                     PF_FLUSH_GLOBAL ||
 735                                     (*state)->rule.ptr == st->rule.ptr)) {
 736                                         st->timeout = PFTM_PURGE;
 737                                         st->src.state = st->dst.state =
 738                                             TCPS_CLOSED;
 739                                         killed++;
 740                                 }
 741                         }
 742                         if (pf_status.debug >= PF_DEBUG_MISC)
 743                                 kprintf(", %u states killed", killed);
 744                 }
 745                 if (pf_status.debug >= PF_DEBUG_MISC)
 746                         kprintf("\n");
 747         }
 748
 749         /* kill this state */
 750         (*state)->timeout = PFTM_PURGE;
 751         (*state)->src.state = (*state)->dst.state = TCPS_CLOSED;
 752         return (1);
 753 }
 754
 755 int
 756 pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule,
 757     struct pf_addr *src, sa_family_t af)
 758 {
 759         struct pf_src_node      k;
 760
 761         if (*sn == NULL) {
 762                 k.af = af;
 763                 PF_ACPY(&k.addr, src, af);
 764                 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
 765                     rule->rpool.opts & PF_POOL_STICKYADDR)
 766                         k.rule.ptr = rule;
 767                 else
 768                         k.rule.ptr = NULL;
 769                 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
 770                 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
 771         }
 772         if (*sn == NULL) {
 773                 if (!rule->max_src_nodes ||
 774                     rule->src_nodes < rule->max_src_nodes)
 775                         (*sn) = pool_get(&pf_src_tree_pl, PR_NOWAIT);
 776                 else
 777                         pf_status.lcounters[LCNT_SRCNODES]++;
 778                 if ((*sn) == NULL)
 779                         return (-1);
 780                 bzero(*sn, sizeof(struct pf_src_node));
 781
 782                 pf_init_threshold(&(*sn)->conn_rate,
 783                     rule->max_src_conn_rate.limit,
 784                     rule->max_src_conn_rate.seconds);
 785
 786                 (*sn)->af = af;
 787                 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
 788                     rule->rpool.opts & PF_POOL_STICKYADDR)
 789                         (*sn)->rule.ptr = rule;
 790                 else
 791                         (*sn)->rule.ptr = NULL;
 792                 PF_ACPY(&(*sn)->addr, src, af);
 793                 if (RB_INSERT(pf_src_tree,
 794                     &tree_src_tracking, *sn) != NULL) {
 795                         if (pf_status.debug >= PF_DEBUG_MISC) {
 796                                 kprintf("pf: src_tree insert failed: ");
 797                                 pf_print_host(&(*sn)->addr, 0, af);
 798                                 kprintf("\n");
 799                         }
 800                         pool_put(&pf_src_tree_pl, *sn);
 801                         return (-1);
 802                 }
 803                 (*sn)->creation = time_second;
 804                 (*sn)->ruletype = rule->action;
 805                 if ((*sn)->rule.ptr != NULL)
 806                         (*sn)->rule.ptr->src_nodes++;
 807                 pf_status.scounters[SCNT_SRC_NODE_INSERT]++;
 808                 pf_status.src_nodes++;
 809         } else {
 810                 if (rule->max_src_states &&
 811                     (*sn)->states >= rule->max_src_states) {
 812                         pf_status.lcounters[LCNT_SRCSTATES]++;
 813                         return (-1);
 814                 }
 815         }
 816         return (0);
 817 }
 818
 819 void
 820 pf_stateins_err(const char *tree, struct pf_state *s, struct pfi_kif *kif)
 821 {
 822         struct pf_state_key     *sk = s->state_key;
 823
 824         if (pf_status.debug >= PF_DEBUG_MISC) {
 825                 kprintf("pf: state insert failed: %s %s", tree, kif->pfik_name);
 826                 kprintf(" lan: ");
 827                 pf_print_host(&sk->lan.addr, sk->lan.port,
 828                     sk->af);
 829                 kprintf(" gwy: ");
 830                 pf_print_host(&sk->gwy.addr, sk->gwy.port,
 831                     sk->af);
 832                 kprintf(" ext: ");
 833                 pf_print_host(&sk->ext.addr, sk->ext.port,
 834                     sk->af);
 835                 if (s->sync_flags & PFSTATE_FROMSYNC)
 836                         kprintf(" (from sync)");
 837                 kprintf("\n");
 838         }
 839 }
 840
 841 int
 842 pf_insert_state(struct pfi_kif *kif, struct pf_state *s)
 843 {
 844         struct pf_state_key     *cur;
 845         struct pf_state         *sp;
 846
 847         KKASSERT(s->state_key != NULL);
 848         s->kif = kif;
 849
 850         if ((cur = RB_INSERT(pf_state_tree_lan_ext, &pf_statetbl_lan_ext,
 851             s->state_key)) != NULL) {
 852                 /* key exists. check for same kif, if none, add to key */
 853                 TAILQ_FOREACH(sp, &cur->states, next)
 854                         if (sp->kif == kif) {   /* collision! */
 855                                 pf_stateins_err("tree_lan_ext", s, kif);
 856                                 pf_detach_state(s,
 857                                     PF_DT_SKIP_LANEXT|PF_DT_SKIP_EXTGWY);
 858                                 return (-1);
 859                         }
 860                 pf_detach_state(s, PF_DT_SKIP_LANEXT|PF_DT_SKIP_EXTGWY);
 861                 pf_attach_state(cur, s, kif == pfi_all ? 1 : 0);
 862         }
 863
 864         /* if cur != NULL, we already found a state key and attached to it */
 865         if (cur == NULL && (cur = RB_INSERT(pf_state_tree_ext_gwy,
 866             &pf_statetbl_ext_gwy, s->state_key)) != NULL) {
 867                 /* must not happen. we must have found the sk above! */
 868                 pf_stateins_err("tree_ext_gwy", s, kif);
 869                 pf_detach_state(s, PF_DT_SKIP_EXTGWY);
 870                 return (-1);
 871         }
 872
 873         if (s->id == 0 && s->creatorid == 0) {
 874                 s->id = htobe64(pf_status.stateid++);
 875                 s->creatorid = pf_status.hostid;
 876         }
 877         if (RB_INSERT(pf_state_tree_id, &tree_id, s) != NULL) {
 878                 if (pf_status.debug >= PF_DEBUG_MISC) {
 879                         kprintf("pf: state insert failed: "
 880                             "id: %016jx creatorid: %08x",
 881                             (uintmax_t)be64toh(s->id), ntohl(s->creatorid));
 882                         if (s->sync_flags & PFSTATE_FROMSYNC)
 883                                 kprintf(" (from sync)");
 884                         kprintf("\n");
 885                 }
 886                 pf_detach_state(s, 0);
 887                 return (-1);
 888         }
 889         TAILQ_INSERT_TAIL(&state_list, s, entry_list);
 890         pf_status.fcounters[FCNT_STATE_INSERT]++;
 891         pf_status.states++;
 892         pfi_kif_ref(kif, PFI_KIF_REF_STATE);
 893 #if NPFSYNC
 894         pfsync_insert_state(s);
 895 #endif
 896         return (0);
 897 }
 898
 899 void
 900 pf_purge_thread(void *v)
 901 {
 902         int nloops = 0;
 903         int locked = 0;
 904
 905         get_mplock();
 906         for (;;) {
 907                 tsleep(pf_purge_thread, PWAIT, "pftm", 1 * hz);
 908
 909                 lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
 910
 911                 if (pf_end_threads) {
 912                         pf_purge_expired_states(pf_status.states, 1);
 913                         pf_purge_expired_fragments();
 914                         pf_purge_expired_src_nodes(1);
 915                         pf_end_threads++;
 916
 917                         lockmgr(&pf_consistency_lock, LK_RELEASE);
 918                         wakeup(pf_purge_thread);
 919                         kthread_exit();
 920                 }
 921                 crit_enter();
 922
 923                 /* process a fraction of the state table every second */
 924                 if(!pf_purge_expired_states(1 + (pf_status.states
 925                     / pf_default_rule.timeout[PFTM_INTERVAL]), 0)) {
 926
 927                         pf_purge_expired_states(1 + (pf_status.states
 928                             / pf_default_rule.timeout[PFTM_INTERVAL]), 1);
 929                 }
 930
 931                 /* purge other expired types every PFTM_INTERVAL seconds */
 932                 if (++nloops >= pf_default_rule.timeout[PFTM_INTERVAL]) {
 933                         pf_purge_expired_fragments();
 934                         if (!pf_purge_expired_src_nodes(locked)) {
 935                                 pf_purge_expired_src_nodes(1);
 936                         }
 937                         nloops = 0;
 938                 }
 939                 crit_exit();
 940                 lockmgr(&pf_consistency_lock, LK_RELEASE);
 941         }
 942         rel_mplock();
 943 }
 944
 945 u_int32_t
 946 pf_state_expires(const struct pf_state *state)
 947 {
 948         u_int32_t       timeout;
 949         u_int32_t       start;
 950         u_int32_t       end;
 951         u_int32_t       states;
 952
 953         /* handle all PFTM_* > PFTM_MAX here */
 954         if (state->timeout == PFTM_PURGE)
 955                 return (time_second);
 956         if (state->timeout == PFTM_UNTIL_PACKET)
 957                 return (0);
 958         KKASSERT(state->timeout != PFTM_UNLINKED);
 959         KASSERT((state->timeout < PFTM_MAX),
 960                 ("pf_state_expires: timeout > PFTM_MAX"));
 961         timeout = state->rule.ptr->timeout[state->timeout];
 962         if (!timeout)
 963                 timeout = pf_default_rule.timeout[state->timeout];
 964         start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START];
 965         if (start) {
 966                 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END];
 967                 states = state->rule.ptr->states;
 968         } else {
 969                 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START];
 970                 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END];
 971                 states = pf_status.states;
 972         }
 973         if (end && states > start && start < end) {
 974                 if (states < end)
 975                         return (state->expire + timeout * (end - states) /
 976                             (end - start));
 977                 else
 978                         return (time_second);
 979         }
 980         return (state->expire + timeout);
 981 }
 982
 983 int
 984 pf_purge_expired_src_nodes(int waslocked)
 985 {
 986          struct pf_src_node             *cur, *next;
 987          int                             locked = waslocked;
 988
 989          for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) {
 990                  next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur);
 991
 992                  if (cur->states <= 0 && cur->expire <= time_second) {
 993                          if (! locked) {
 994                                  lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
 995                                  next = RB_NEXT(pf_src_tree,
 996                                      &tree_src_tracking, cur);
 997                                  locked = 1;
 998                          }
 999                          if (cur->rule.ptr != NULL) {
1000                                  cur->rule.ptr->src_nodes--;
1001                                  if (cur->rule.ptr->states <= 0 &&
1002                                      cur->rule.ptr->max_src_nodes <= 0)
1003                                          pf_rm_rule(NULL, cur->rule.ptr);
1004                          }
1005                          RB_REMOVE(pf_src_tree, &tree_src_tracking, cur);
1006                          pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
1007                          pf_status.src_nodes--;
1008                          pool_put(&pf_src_tree_pl, cur);
1009                  }
1010          }
1011
1012          if (locked && !waslocked)
1013                 lockmgr(&pf_consistency_lock, LK_RELEASE);
1014         return(1);
1015 }
1016
1017 void
1018 pf_src_tree_remove_state(struct pf_state *s)
1019 {
1020         u_int32_t timeout;
1021
1022         if (s->src_node != NULL) {
1023                 if (s->src.tcp_est)
1024                         --s->src_node->conn;
1025                 if (--s->src_node->states <= 0) {
1026                         timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1027                         if (!timeout)
1028                                 timeout =
1029                                     pf_default_rule.timeout[PFTM_SRC_NODE];
1030                         s->src_node->expire = time_second + timeout;
1031                 }
1032         }
1033         if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) {
1034                 if (--s->nat_src_node->states <= 0) {
1035                         timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1036                         if (!timeout)
1037                                 timeout =
1038                                     pf_default_rule.timeout[PFTM_SRC_NODE];
1039                         s->nat_src_node->expire = time_second + timeout;
1040                 }
1041         }
1042         s->src_node = s->nat_src_node = NULL;
1043 }
1044
1045 /* callers should be at crit_enter() */
1046 void
1047 pf_unlink_state(struct pf_state *cur)
1048 {
1049         if (cur->src.state == PF_TCPS_PROXY_DST) {
1050                 pf_send_tcp(cur->rule.ptr, cur->state_key->af,
1051                     &cur->state_key->ext.addr, &cur->state_key->lan.addr,
1052                     cur->state_key->ext.port, cur->state_key->lan.port,
1053                     cur->src.seqhi, cur->src.seqlo + 1,
1054                     TH_RST|TH_ACK, 0, 0, 0, 1, cur->tag, NULL, NULL);
1055         }
1056         RB_REMOVE(pf_state_tree_id, &tree_id, cur);
1057 #if NPFSYNC
1058         if (cur->creatorid == pf_status.hostid)
1059                 pfsync_delete_state(cur);
1060 #endif
1061         cur->timeout = PFTM_UNLINKED;
1062         pf_src_tree_remove_state(cur);
1063         pf_detach_state(cur, 0);
1064 }
1065
1066 /* callers should be at crit_enter() and hold the
1067  * write_lock on pf_consistency_lock */
1068 void
1069 pf_free_state(struct pf_state *cur)
1070 {
1071 #if NPFSYNC
1072         if (pfsyncif != NULL &&
1073             (pfsyncif->sc_bulk_send_next == cur ||
1074             pfsyncif->sc_bulk_terminator == cur))
1075                 return;
1076 #endif
1077         KKASSERT(cur->timeout == PFTM_UNLINKED);
1078         if (--cur->rule.ptr->states <= 0 &&
1079             cur->rule.ptr->src_nodes <= 0)
1080                 pf_rm_rule(NULL, cur->rule.ptr);
1081         if (cur->nat_rule.ptr != NULL)
1082                 if (--cur->nat_rule.ptr->states <= 0 &&
1083                         cur->nat_rule.ptr->src_nodes <= 0)
1084                         pf_rm_rule(NULL, cur->nat_rule.ptr);
1085         if (cur->anchor.ptr != NULL)
1086                 if (--cur->anchor.ptr->states <= 0)
1087                         pf_rm_rule(NULL, cur->anchor.ptr);
1088         pf_normalize_tcp_cleanup(cur);
1089         pfi_kif_unref(cur->kif, PFI_KIF_REF_STATE);
1090         TAILQ_REMOVE(&state_list, cur, entry_list);
1091         if (cur->tag)
1092                 pf_tag_unref(cur->tag);
1093         pool_put(&pf_state_pl, cur);
1094         pf_status.fcounters[FCNT_STATE_REMOVALS]++;
1095         pf_status.states--;
1096 }
1097
1098 int
1099 pf_purge_expired_states(u_int32_t maxcheck, int waslocked)
1100 {
1101         static struct pf_state  *cur = NULL;
1102         struct pf_state         *next;
1103         int                      locked = waslocked;
1104
1105         while (maxcheck--) {
1106                 /* wrap to start of list when we hit the end */
1107                 if (cur == NULL) {
1108                         cur = TAILQ_FIRST(&state_list);
1109                         if (cur == NULL)
1110                                 break;  /* list empty */
1111                 }
1112
1113                 /* get next state, as cur may get deleted */
1114                 next = TAILQ_NEXT(cur, entry_list);
1115
1116                 if (cur->timeout == PFTM_UNLINKED) {
1117                         /* free unlinked state */
1118                         if (! locked) {
1119                                 lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
1120                                 locked = 1;
1121                         }
1122                         pf_free_state(cur);
1123                 } else if (pf_state_expires(cur) <= time_second) {
1124                         /* unlink and free expired state */
1125                         pf_unlink_state(cur);
1126                         if (! locked) {
1127                                 if (!lockmgr(&pf_consistency_lock, LK_EXCLUSIVE))
1128                                         return (0);
1129                                 locked = 1;
1130                         }
1131                         pf_free_state(cur);
1132                 }
1133                 cur = next;
1134         }
1135
1136         if (locked)
1137                 lockmgr(&pf_consistency_lock, LK_RELEASE);
1138         return (1);
1139 }
1140
1141 int
1142 pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw)
1143 {
1144         if (aw->type != PF_ADDR_TABLE)
1145                 return (0);
1146         if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL)
1147                 return (1);
1148         return (0);
1149 }
1150
1151 void
1152 pf_tbladdr_remove(struct pf_addr_wrap *aw)
1153 {
1154         if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL)
1155                 return;
1156         pfr_detach_table(aw->p.tbl);
1157         aw->p.tbl = NULL;
1158 }
1159
1160 void
1161 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
1162 {
1163         struct pfr_ktable *kt = aw->p.tbl;
1164
1165         if (aw->type != PF_ADDR_TABLE || kt == NULL)
1166                 return;
1167         if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
1168                 kt = kt->pfrkt_root;
1169         aw->p.tbl = NULL;
1170         aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
1171                 kt->pfrkt_cnt : -1;
1172 }
1173
1174 void
1175 pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af)
1176 {
1177         switch (af) {
1178 #ifdef INET
1179         case AF_INET: {
1180                 u_int32_t a = ntohl(addr->addr32[0]);
1181                 kprintf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255,
1182                     (a>>8)&255, a&255);
1183                 if (p) {
1184                         p = ntohs(p);
1185                         kprintf(":%u", p);
1186                 }
1187                 break;
1188         }
1189 #endif /* INET */
1190 #ifdef INET6
1191         case AF_INET6: {
1192                 u_int16_t b;
1193                 u_int8_t i, curstart = 255, curend = 0,
1194                     maxstart = 0, maxend = 0;
1195                 for (i = 0; i < 8; i++) {
1196                         if (!addr->addr16[i]) {
1197                                 if (curstart == 255)
1198                                         curstart = i;
1199                                 else
1200                                         curend = i;
1201                         } else {
1202                                 if (curstart) {
1203                                         if ((curend - curstart) >
1204                                             (maxend - maxstart)) {
1205                                                 maxstart = curstart;
1206                                                 maxend = curend;
1207                                                 curstart = 255;
1208                                         }
1209                                 }
1210                         }
1211                 }
1212                 for (i = 0; i < 8; i++) {
1213                         if (i >= maxstart && i <= maxend) {
1214                                 if (maxend != 7) {
1215                                         if (i == maxstart)
1216                                                 kprintf(":");
1217                                 } else {
1218                                         if (i == maxend)
1219                                                 kprintf(":");
1220                                 }
1221                         } else {
1222                                 b = ntohs(addr->addr16[i]);
1223                                 kprintf("%x", b);
1224                                 if (i < 7)
1225                                         kprintf(":");
1226                         }
1227                 }
1228                 if (p) {
1229                         p = ntohs(p);
1230                         kprintf("[%u]", p);
1231                 }
1232                 break;
1233         }
1234 #endif /* INET6 */
1235         }
1236 }
1237
1238 void
1239 pf_print_state(struct pf_state *s)
1240 {
1241         struct pf_state_key *sk = s->state_key;
1242         switch (sk->proto) {
1243         case IPPROTO_TCP:
1244                 kprintf("TCP ");
1245                 break;
1246         case IPPROTO_UDP:
1247                 kprintf("UDP ");
1248                 break;
1249         case IPPROTO_ICMP:
1250                 kprintf("ICMP ");
1251                 break;
1252         case IPPROTO_ICMPV6:
1253                 kprintf("ICMPV6 ");
1254                 break;
1255         default:
1256                 kprintf("%u ", sk->proto);
1257                 break;
1258         }
1259         pf_print_host(&sk->lan.addr, sk->lan.port, sk->af);
1260         kprintf(" ");
1261         pf_print_host(&sk->gwy.addr, sk->gwy.port, sk->af);
1262         kprintf(" ");
1263         pf_print_host(&sk->ext.addr, sk->ext.port, sk->af);
1264         kprintf(" [lo=%u high=%u win=%u modulator=%u", s->src.seqlo,
1265             s->src.seqhi, s->src.max_win, s->src.seqdiff);
1266         if (s->src.wscale && s->dst.wscale)
1267                 kprintf(" wscale=%u", s->src.wscale & PF_WSCALE_MASK);
1268         kprintf("]");
1269         kprintf(" [lo=%u high=%u win=%u modulator=%u", s->dst.seqlo,
1270             s->dst.seqhi, s->dst.max_win, s->dst.seqdiff);
1271         if (s->src.wscale && s->dst.wscale)
1272                 kprintf(" wscale=%u", s->dst.wscale & PF_WSCALE_MASK);
1273         kprintf("]");
1274         kprintf(" %u:%u", s->src.state, s->dst.state);
1275 }
1276
1277 void
1278 pf_print_flags(u_int8_t f)
1279 {
1280         if (f)
1281                 kprintf(" ");
1282         if (f & TH_FIN)
1283                 kprintf("F");
1284         if (f & TH_SYN)
1285                 kprintf("S");
1286         if (f & TH_RST)
1287                 kprintf("R");
1288         if (f & TH_PUSH)
1289                 kprintf("P");
1290         if (f & TH_ACK)
1291                 kprintf("A");
1292         if (f & TH_URG)
1293                 kprintf("U");
1294         if (f & TH_ECE)
1295                 kprintf("E");
1296         if (f & TH_CWR)
1297                 kprintf("W");
1298 }
1299
1300 #define PF_SET_SKIP_STEPS(i)                                    \
1301         do {                                                    \
1302                 while (head[i] != cur) {                        \
1303                         head[i]->skip[i].ptr = cur;             \
1304                         head[i] = TAILQ_NEXT(head[i], entries); \
1305                 }                                               \
1306         } while (0)
1307
1308 void
1309 pf_calc_skip_steps(struct pf_rulequeue *rules)
1310 {
1311         struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT];
1312         int i;
1313
1314         cur = TAILQ_FIRST(rules);
1315         prev = cur;
1316         for (i = 0; i < PF_SKIP_COUNT; ++i)
1317                 head[i] = cur;
1318         while (cur != NULL) {
1319
1320                 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
1321                         PF_SET_SKIP_STEPS(PF_SKIP_IFP);
1322                 if (cur->direction != prev->direction)
1323                         PF_SET_SKIP_STEPS(PF_SKIP_DIR);
1324                 if (cur->af != prev->af)
1325                         PF_SET_SKIP_STEPS(PF_SKIP_AF);
1326                 if (cur->proto != prev->proto)
1327                         PF_SET_SKIP_STEPS(PF_SKIP_PROTO);
1328                 if (cur->src.neg != prev->src.neg ||
1329                     pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr))
1330                         PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR);
1331                 if (cur->src.port[0] != prev->src.port[0] ||
1332                     cur->src.port[1] != prev->src.port[1] ||
1333                     cur->src.port_op != prev->src.port_op)
1334                         PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
1335                 if (cur->dst.neg != prev->dst.neg ||
1336                     pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr))
1337                         PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR);
1338                 if (cur->dst.port[0] != prev->dst.port[0] ||
1339                     cur->dst.port[1] != prev->dst.port[1] ||
1340                     cur->dst.port_op != prev->dst.port_op)
1341                         PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
1342
1343                 prev = cur;
1344                 cur = TAILQ_NEXT(cur, entries);
1345         }
1346         for (i = 0; i < PF_SKIP_COUNT; ++i)
1347                 PF_SET_SKIP_STEPS(i);
1348 }
1349
1350 int
1351 pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2)
1352 {
1353         if (aw1->type != aw2->type)
1354                 return (1);
1355         switch (aw1->type) {
1356         case PF_ADDR_ADDRMASK:
1357                 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0))
1358                         return (1);
1359                 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0))
1360                         return (1);
1361                 return (0);
1362         case PF_ADDR_DYNIFTL:
1363                 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt);
1364         case PF_ADDR_NOROUTE:
1365         case PF_ADDR_URPFFAILED:
1366                 return (0);
1367         case PF_ADDR_TABLE:
1368                 return (aw1->p.tbl != aw2->p.tbl);
1369         case PF_ADDR_RTLABEL:
1370                 return (aw1->v.rtlabel != aw2->v.rtlabel);
1371         default:
1372                 kprintf("invalid address type: %d\n", aw1->type);
1373                 return (1);
1374         }
1375 }
1376
1377 u_int16_t
1378 pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp)
1379 {
1380         u_int32_t       l;
1381
1382         if (udp && !cksum)
1383                 return (0x0000);
1384         l = cksum + old - new;
1385         l = (l >> 16) + (l & 65535);
1386         l = l & 65535;
1387         if (udp && !l)
1388                 return (0xFFFF);
1389         return (l);
1390 }
1391
1392 void
1393 pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc,
1394     struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af)
1395 {
1396         struct pf_addr  ao;
1397         u_int16_t       po = *p;
1398
1399         PF_ACPY(&ao, a, af);
1400         PF_ACPY(a, an, af);
1401
1402         *p = pn;
1403
1404         switch (af) {
1405 #ifdef INET
1406         case AF_INET:
1407                 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1408                     ao.addr16[0], an->addr16[0], 0),
1409                     ao.addr16[1], an->addr16[1], 0);
1410                 *p = pn;
1411                 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1412                     ao.addr16[0], an->addr16[0], u),
1413                     ao.addr16[1], an->addr16[1], u),
1414                     po, pn, u);
1415                 break;
1416 #endif /* INET */
1417 #ifdef INET6
1418         case AF_INET6:
1419                 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1420                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1421                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1422                     ao.addr16[0], an->addr16[0], u),
1423                     ao.addr16[1], an->addr16[1], u),
1424                     ao.addr16[2], an->addr16[2], u),
1425                     ao.addr16[3], an->addr16[3], u),
1426                     ao.addr16[4], an->addr16[4], u),
1427                     ao.addr16[5], an->addr16[5], u),
1428                     ao.addr16[6], an->addr16[6], u),
1429                     ao.addr16[7], an->addr16[7], u),
1430                     po, pn, u);
1431                 break;
1432 #endif /* INET6 */
1433         }
1434 }
1435
1436
1437 /* Changes a u_int32_t.  Uses a void * so there are no align restrictions */
1438 void
1439 pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u)
1440 {
1441         u_int32_t       ao;
1442
1443         memcpy(&ao, a, sizeof(ao));
1444         memcpy(a, &an, sizeof(u_int32_t));
1445         *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u),
1446             ao % 65536, an % 65536, u);
1447 }
1448
1449 #ifdef INET6
1450 void
1451 pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u)
1452 {
1453         struct pf_addr  ao;
1454
1455         PF_ACPY(&ao, a, AF_INET6);
1456         PF_ACPY(a, an, AF_INET6);
1457
1458         *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1459             pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1460             pf_cksum_fixup(pf_cksum_fixup(*c,
1461             ao.addr16[0], an->addr16[0], u),
1462             ao.addr16[1], an->addr16[1], u),
1463             ao.addr16[2], an->addr16[2], u),
1464             ao.addr16[3], an->addr16[3], u),
1465             ao.addr16[4], an->addr16[4], u),
1466             ao.addr16[5], an->addr16[5], u),
1467             ao.addr16[6], an->addr16[6], u),
1468             ao.addr16[7], an->addr16[7], u);
1469 }
1470 #endif /* INET6 */
1471
1472 void
1473 pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa,
1474     struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c,
1475     u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af)
1476 {
1477         struct pf_addr  oia, ooa;
1478
1479         PF_ACPY(&oia, ia, af);
1480         PF_ACPY(&ooa, oa, af);
1481
1482         /* Change inner protocol port, fix inner protocol checksum. */
1483         if (ip != NULL) {
1484                 u_int16_t       oip = *ip;
1485                 u_int32_t       opc = 0;
1486
1487                 if (pc != NULL)
1488                         opc = *pc;
1489                 *ip = np;
1490                 if (pc != NULL)
1491                         *pc = pf_cksum_fixup(*pc, oip, *ip, u);
1492                 *ic = pf_cksum_fixup(*ic, oip, *ip, 0);
1493                 if (pc != NULL)
1494                         *ic = pf_cksum_fixup(*ic, opc, *pc, 0);
1495         }
1496         /* Change inner ip address, fix inner ip and icmp checksums. */
1497         PF_ACPY(ia, na, af);
1498         switch (af) {
1499 #ifdef INET
1500         case AF_INET: {
1501                 u_int32_t        oh2c = *h2c;
1502
1503                 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c,
1504                     oia.addr16[0], ia->addr16[0], 0),
1505                     oia.addr16[1], ia->addr16[1], 0);
1506                 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1507                     oia.addr16[0], ia->addr16[0], 0),
1508                     oia.addr16[1], ia->addr16[1], 0);
1509                 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0);
1510                 break;
1511         }
1512 #endif /* INET */
1513 #ifdef INET6
1514         case AF_INET6:
1515                 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1516                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1517                     pf_cksum_fixup(pf_cksum_fixup(*ic,
1518                     oia.addr16[0], ia->addr16[0], u),
1519                     oia.addr16[1], ia->addr16[1], u),
1520                     oia.addr16[2], ia->addr16[2], u),
1521                     oia.addr16[3], ia->addr16[3], u),
1522                     oia.addr16[4], ia->addr16[4], u),
1523                     oia.addr16[5], ia->addr16[5], u),
1524                     oia.addr16[6], ia->addr16[6], u),
1525                     oia.addr16[7], ia->addr16[7], u);
1526                 break;
1527 #endif /* INET6 */
1528         }
1529         /* Change outer ip address, fix outer ip or icmpv6 checksum. */
1530         PF_ACPY(oa, na, af);
1531         switch (af) {
1532 #ifdef INET
1533         case AF_INET:
1534                 *hc = pf_cksum_fixup(pf_cksum_fixup(*hc,
1535                     ooa.addr16[0], oa->addr16[0], 0),
1536                     ooa.addr16[1], oa->addr16[1], 0);
1537                 break;
1538 #endif /* INET */
1539 #ifdef INET6
1540         case AF_INET6:
1541                 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1542                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1543                     pf_cksum_fixup(pf_cksum_fixup(*ic,
1544                     ooa.addr16[0], oa->addr16[0], u),
1545                     ooa.addr16[1], oa->addr16[1], u),
1546                     ooa.addr16[2], oa->addr16[2], u),
1547                     ooa.addr16[3], oa->addr16[3], u),
1548                     ooa.addr16[4], oa->addr16[4], u),
1549                     ooa.addr16[5], oa->addr16[5], u),
1550                     ooa.addr16[6], oa->addr16[6], u),
1551                     ooa.addr16[7], oa->addr16[7], u);
1552                 break;
1553 #endif /* INET6 */
1554         }
1555 }
1556
1557
1558 /*
1559  * Need to modulate the sequence numbers in the TCP SACK option
1560  * (credits to Krzysztof Pfaff for report and patch)
1561  */
1562 int
1563 pf_modulate_sack(struct mbuf *m, int off, struct pf_pdesc *pd,
1564     struct tcphdr *th, struct pf_state_peer *dst)
1565 {
1566         int hlen = (th->th_off << 2) - sizeof(*th), thoptlen = hlen;
1567         u_int8_t opts[TCP_MAXOLEN], *opt = opts;
1568         int copyback = 0, i, olen;
1569         struct raw_sackblock sack;
1570
1571 #define TCPOLEN_SACKLEN (TCPOLEN_SACK + 2)
1572         if (hlen < TCPOLEN_SACKLEN ||
1573             !pf_pull_hdr(m, off + sizeof(*th), opts, hlen, NULL, NULL, pd->af))
1574                 return 0;
1575
1576         while (hlen >= TCPOLEN_SACKLEN) {
1577                 olen = opt[1];
1578                 switch (*opt) {
1579                 case TCPOPT_EOL:        /* FALLTHROUGH */
1580                 case TCPOPT_NOP:
1581                         opt++;
1582                         hlen--;
1583                         break;
1584                 case TCPOPT_SACK:
1585                         if (olen > hlen)
1586                                 olen = hlen;
1587                         if (olen >= TCPOLEN_SACKLEN) {
1588                                 for (i = 2; i + TCPOLEN_SACK <= olen;
1589                                     i += TCPOLEN_SACK) {
1590                                         memcpy(&sack, &opt[i], sizeof(sack));
1591                                         pf_change_a(&sack.rblk_start, &th->th_sum,
1592                                             htonl(ntohl(sack.rblk_start) -
1593                                             dst->seqdiff), 0);
1594                                         pf_change_a(&sack.rblk_end, &th->th_sum,
1595                                             htonl(ntohl(sack.rblk_end) -
1596                                             dst->seqdiff), 0);
1597                                         memcpy(&opt[i], &sack, sizeof(sack));
1598                                 }
1599                                 copyback = 1;
1600                         }
1601                         /* FALLTHROUGH */
1602                 default:
1603                         if (olen < 2)
1604                                 olen = 2;
1605                         hlen -= olen;
1606                         opt += olen;
1607                 }
1608         }
1609
1610         if (copyback)
1611                 m_copyback(m, off + sizeof(*th), thoptlen, opts);
1612         return (copyback);
1613 }
1614
1615 void
1616 pf_send_tcp(const struct pf_rule *r, sa_family_t af,
1617     const struct pf_addr *saddr, const struct pf_addr *daddr,
1618     u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack,
1619     u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl, int tag,
1620     u_int16_t rtag, struct ether_header *eh, struct ifnet *ifp)
1621 {
1622         struct mbuf     *m;
1623         int              len = 0, tlen;
1624 #ifdef INET
1625         struct ip       *h = NULL;
1626 #endif /* INET */
1627 #ifdef INET6
1628         struct ip6_hdr  *h6 = NULL;
1629 #endif /* INET6 */
1630         struct tcphdr   *th = NULL;
1631         char            *opt;
1632
1633         ASSERT_LWKT_TOKEN_HELD(&pf_token);
1634
1635         /* maximum segment size tcp option */
1636         tlen = sizeof(struct tcphdr);
1637         if (mss)
1638                 tlen += 4;
1639
1640         switch (af) {
1641 #ifdef INET
1642         case AF_INET:
1643                 len = sizeof(struct ip) + tlen;
1644                 break;
1645 #endif /* INET */
1646 #ifdef INET6
1647         case AF_INET6:
1648                 len = sizeof(struct ip6_hdr) + tlen;
1649                 break;
1650 #endif /* INET6 */
1651         }
1652
1653         /* create outgoing mbuf */
1654         m = m_gethdr(MB_DONTWAIT, MT_HEADER);
1655         if (m == NULL) {
1656                 return;
1657         }
1658         if (tag)
1659                 m->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
1660         m->m_pkthdr.pf.tag = rtag;
1661
1662         if (r != NULL && r->rtableid >= 0)
1663                 m->m_pkthdr.pf.rtableid = r->rtableid;
1664
1665 #ifdef ALTQ
1666         if (r != NULL && r->qid) {
1667                 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
1668                 m->m_pkthdr.pf.qid = r->qid;
1669                 m->m_pkthdr.pf.ecn_af = af;
1670                 m->m_pkthdr.pf.hdr = mtod(m, struct ip *);
1671         }
1672 #endif /* ALTQ */
1673         m->m_data += max_linkhdr;
1674         m->m_pkthdr.len = m->m_len = len;
1675         m->m_pkthdr.rcvif = NULL;
1676         bzero(m->m_data, len);
1677         switch (af) {
1678 #ifdef INET
1679         case AF_INET:
1680                 h = mtod(m, struct ip *);
1681
1682                 /* IP header fields included in the TCP checksum */
1683                 h->ip_p = IPPROTO_TCP;
1684                 h->ip_len = tlen;
1685                 h->ip_src.s_addr = saddr->v4.s_addr;
1686                 h->ip_dst.s_addr = daddr->v4.s_addr;
1687
1688                 th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip));
1689                 break;
1690 #endif /* INET */
1691 #ifdef INET6
1692         case AF_INET6:
1693                 h6 = mtod(m, struct ip6_hdr *);
1694
1695                 /* IP header fields included in the TCP checksum */
1696                 h6->ip6_nxt = IPPROTO_TCP;
1697                 h6->ip6_plen = htons(tlen);
1698                 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr));
1699                 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr));
1700
1701                 th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr));
1702                 break;
1703 #endif /* INET6 */
1704         }
1705
1706         /* TCP header */
1707         th->th_sport = sport;
1708         th->th_dport = dport;
1709         th->th_seq = htonl(seq);
1710         th->th_ack = htonl(ack);
1711         th->th_off = tlen >> 2;
1712         th->th_flags = flags;
1713         th->th_win = htons(win);
1714
1715         if (mss) {
1716                 opt = (char *)(th + 1);
1717                 opt[0] = TCPOPT_MAXSEG;
1718                 opt[1] = 4;
1719                 mss = htons(mss);
1720                 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2);
1721         }
1722
1723         switch (af) {
1724 #ifdef INET
1725         case AF_INET:
1726                 /* TCP checksum */
1727                 th->th_sum = in_cksum(m, len);
1728
1729                 /* Finish the IP header */
1730                 h->ip_v = 4;
1731                 h->ip_hl = sizeof(*h) >> 2;
1732                 h->ip_tos = IPTOS_LOWDELAY;
1733                 h->ip_len = len;
1734                 h->ip_off = path_mtu_discovery ? IP_DF : 0;
1735                 h->ip_ttl = ttl ? ttl : ip_defttl;
1736                 h->ip_sum = 0;
1737                 if (eh == NULL) {
1738                         lwkt_reltoken(&pf_token);
1739                         ip_output(m, NULL, NULL, 0, NULL, NULL);
1740                         lwkt_gettoken(&pf_token);
1741                 } else {
1742                         struct route             ro;
1743                         struct rtentry           rt;
1744                         struct ether_header     *e = (void *)ro.ro_dst.sa_data;
1745
1746                         if (ifp == NULL) {
1747                                 m_freem(m);
1748                                 return;
1749                         }
1750                         rt.rt_ifp = ifp;
1751                         ro.ro_rt = &rt;
1752                         ro.ro_dst.sa_len = sizeof(ro.ro_dst);
1753                         ro.ro_dst.sa_family = pseudo_AF_HDRCMPLT;
1754                         bcopy(eh->ether_dhost, e->ether_shost, ETHER_ADDR_LEN);
1755                         bcopy(eh->ether_shost, e->ether_dhost, ETHER_ADDR_LEN);
1756                         e->ether_type = eh->ether_type;
1757                         /* XXX_IMPORT: later */
1758                         lwkt_reltoken(&pf_token);
1759                         ip_output(m, (void *)NULL, &ro, 0,
1760                             (void *)NULL, (void *)NULL);
1761                         lwkt_gettoken(&pf_token);
1762                 }
1763                 break;
1764 #endif /* INET */
1765 #ifdef INET6
1766         case AF_INET6:
1767                 /* TCP checksum */
1768                 th->th_sum = in6_cksum(m, IPPROTO_TCP,
1769                     sizeof(struct ip6_hdr), tlen);
1770
1771                 h6->ip6_vfc |= IPV6_VERSION;
1772                 h6->ip6_hlim = IPV6_DEFHLIM;
1773
1774                 lwkt_reltoken(&pf_token);
1775                 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
1776                 lwkt_gettoken(&pf_token);
1777                 break;
1778 #endif /* INET6 */
1779         }
1780 }
1781
1782 void
1783 pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af,
1784     struct pf_rule *r)
1785 {
1786         struct mbuf     *m0;
1787
1788         m0 = m_copy(m, 0, M_COPYALL);
1789         m0->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
1790
1791         if (r->rtableid >= 0)
1792                 m0->m_pkthdr.pf.rtableid = r->rtableid;
1793
1794 #ifdef ALTQ
1795         if (r->qid) {
1796                 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
1797                 m0->m_pkthdr.pf.qid = r->qid;
1798                 m0->m_pkthdr.pf.ecn_af = af;
1799                 m0->m_pkthdr.pf.hdr = mtod(m0, struct ip *);
1800         }
1801 #endif /* ALTQ */
1802
1803         switch (af) {
1804 #ifdef INET
1805         case AF_INET:
1806                 icmp_error(m0, type, code, 0, 0);
1807                 break;
1808 #endif /* INET */
1809 #ifdef INET6
1810         case AF_INET6:
1811                 icmp6_error(m0, type, code, 0);
1812                 break;
1813 #endif /* INET6 */
1814         }
1815 }
1816
1817 /*
1818  * Return 1 if the addresses a and b match (with mask m), otherwise return 0.
1819  * If n is 0, they match if they are equal. If n is != 0, they match if they
1820  * are different.
1821  */
1822 int
1823 pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m,
1824     struct pf_addr *b, sa_family_t af)
1825 {
1826         int     match = 0;
1827
1828         switch (af) {
1829 #ifdef INET
1830         case AF_INET:
1831                 if ((a->addr32[0] & m->addr32[0]) ==
1832                     (b->addr32[0] & m->addr32[0]))
1833                         match++;
1834                 break;
1835 #endif /* INET */
1836 #ifdef INET6
1837         case AF_INET6:
1838                 if (((a->addr32[0] & m->addr32[0]) ==
1839                      (b->addr32[0] & m->addr32[0])) &&
1840                     ((a->addr32[1] & m->addr32[1]) ==
1841                      (b->addr32[1] & m->addr32[1])) &&
1842                     ((a->addr32[2] & m->addr32[2]) ==
1843                      (b->addr32[2] & m->addr32[2])) &&
1844                     ((a->addr32[3] & m->addr32[3]) ==
1845                      (b->addr32[3] & m->addr32[3])))
1846                         match++;
1847                 break;
1848 #endif /* INET6 */
1849         }
1850         if (match) {
1851                 if (n)
1852                         return (0);
1853                 else
1854                         return (1);
1855         } else {
1856                 if (n)
1857                         return (1);
1858                 else
1859                         return (0);
1860         }
1861 }
1862
1863 int
1864 pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p)
1865 {
1866         switch (op) {
1867         case PF_OP_IRG:
1868                 return ((p > a1) && (p < a2));
1869         case PF_OP_XRG:
1870                 return ((p < a1) || (p > a2));
1871         case PF_OP_RRG:
1872                 return ((p >= a1) && (p <= a2));
1873         case PF_OP_EQ:
1874                 return (p == a1);
1875         case PF_OP_NE:
1876                 return (p != a1);
1877         case PF_OP_LT:
1878                 return (p < a1);
1879         case PF_OP_LE:
1880                 return (p <= a1);
1881         case PF_OP_GT:
1882                 return (p > a1);
1883         case PF_OP_GE:
1884                 return (p >= a1);
1885         }
1886         return (0); /* never reached */
1887 }
1888
1889 int
1890 pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p)
1891 {
1892         a1 = ntohs(a1);
1893         a2 = ntohs(a2);
1894         p = ntohs(p);
1895         return (pf_match(op, a1, a2, p));
1896 }
1897
1898 int
1899 pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u)
1900 {
1901         if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
1902                 return (0);
1903         return (pf_match(op, a1, a2, u));
1904 }
1905
1906 int
1907 pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g)
1908 {
1909         if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
1910                 return (0);
1911         return (pf_match(op, a1, a2, g));
1912 }
1913
1914 int
1915 pf_match_tag(struct mbuf *m, struct pf_rule *r, int *tag)
1916 {
1917         if (*tag == -1)
1918                 *tag = m->m_pkthdr.pf.tag;
1919
1920         return ((!r->match_tag_not && r->match_tag == *tag) ||
1921             (r->match_tag_not && r->match_tag != *tag));
1922 }
1923
1924 int
1925 pf_tag_packet(struct mbuf *m, int tag, int rtableid)
1926 {
1927         if (tag <= 0 && rtableid < 0)
1928                 return (0);
1929
1930         if (tag > 0)
1931                 m->m_pkthdr.pf.tag = tag;
1932         if (rtableid >= 0)
1933                 m->m_pkthdr.pf.rtableid = rtableid;
1934
1935         return (0);
1936 }
1937
1938 void
1939 pf_step_into_anchor(int *depth, struct pf_ruleset **rs, int n,
1940     struct pf_rule **r, struct pf_rule **a,  int *match)
1941 {
1942         struct pf_anchor_stackframe     *f;
1943
1944         (*r)->anchor->match = 0;
1945         if (match)
1946                 *match = 0;
1947         if (*depth >= sizeof(pf_anchor_stack) /
1948             sizeof(pf_anchor_stack[0])) {
1949                 kprintf("pf_step_into_anchor: stack overflow\n");
1950                 *r = TAILQ_NEXT(*r, entries);
1951                 return;
1952         } else if (*depth == 0 && a != NULL)
1953                 *a = *r;
1954         f = pf_anchor_stack + (*depth)++;
1955         f->rs = *rs;
1956         f->r = *r;
1957         if ((*r)->anchor_wildcard) {
1958                 f->parent = &(*r)->anchor->children;
1959                 if ((f->child = RB_MIN(pf_anchor_node, f->parent)) ==
1960                     NULL) {
1961                         *r = NULL;
1962                         return;
1963                 }
1964                 *rs = &f->child->ruleset;
1965         } else {
1966                 f->parent = NULL;
1967                 f->child = NULL;
1968                 *rs = &(*r)->anchor->ruleset;
1969         }
1970         *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
1971 }
1972
1973 int
1974 pf_step_out_of_anchor(int *depth, struct pf_ruleset **rs, int n,
1975     struct pf_rule **r, struct pf_rule **a, int *match)
1976 {
1977         struct pf_anchor_stackframe     *f;
1978         int quick = 0;
1979
1980         do {
1981                 if (*depth <= 0)
1982                         break;
1983                 f = pf_anchor_stack + *depth - 1;
1984                 if (f->parent != NULL && f->child != NULL) {
1985                         if (f->child->match ||
1986                             (match != NULL && *match)) {
1987                                 f->r->anchor->match = 1;
1988                                 *match = 0;
1989                         }
1990                         f->child = RB_NEXT(pf_anchor_node, f->parent, f->child);
1991                         if (f->child != NULL) {
1992                                 *rs = &f->child->ruleset;
1993                                 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
1994                                 if (*r == NULL)
1995                                         continue;
1996                                 else
1997                                         break;
1998                         }
1999                 }
2000                 (*depth)--;
2001                 if (*depth == 0 && a != NULL)
2002                         *a = NULL;
2003                 *rs = f->rs;
2004                 if (f->r->anchor->match || (match  != NULL && *match))
2005                         quick = f->r->quick;
2006                 *r = TAILQ_NEXT(f->r, entries);
2007         } while (*r == NULL);
2008
2009         return (quick);
2010 }
2011
2012 #ifdef INET6
2013 void
2014 pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr,
2015     struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af)
2016 {
2017         switch (af) {
2018 #ifdef INET
2019         case AF_INET:
2020                 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2021                 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2022                 break;
2023 #endif /* INET */
2024         case AF_INET6:
2025                 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2026                 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2027                 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) |
2028                 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]);
2029                 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) |
2030                 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]);
2031                 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) |
2032                 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]);
2033                 break;
2034         }
2035 }
2036
2037 void
2038 pf_addr_inc(struct pf_addr *addr, sa_family_t af)
2039 {
2040         switch (af) {
2041 #ifdef INET
2042         case AF_INET:
2043                 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1);
2044                 break;
2045 #endif /* INET */
2046         case AF_INET6:
2047                 if (addr->addr32[3] == 0xffffffff) {
2048                         addr->addr32[3] = 0;
2049                         if (addr->addr32[2] == 0xffffffff) {
2050                                 addr->addr32[2] = 0;
2051                                 if (addr->addr32[1] == 0xffffffff) {
2052                                         addr->addr32[1] = 0;
2053                                         addr->addr32[0] =
2054                                             htonl(ntohl(addr->addr32[0]) + 1);
2055                                 } else
2056                                         addr->addr32[1] =
2057                                             htonl(ntohl(addr->addr32[1]) + 1);
2058                         } else
2059                                 addr->addr32[2] =
2060                                     htonl(ntohl(addr->addr32[2]) + 1);
2061                 } else
2062                         addr->addr32[3] =
2063                             htonl(ntohl(addr->addr32[3]) + 1);
2064                 break;
2065         }
2066 }
2067 #endif /* INET6 */
2068
2069 #define mix(a,b,c) \
2070         do {                                    \
2071                 a -= b; a -= c; a ^= (c >> 13); \
2072                 b -= c; b -= a; b ^= (a << 8);  \
2073                 c -= a; c -= b; c ^= (b >> 13); \
2074                 a -= b; a -= c; a ^= (c >> 12); \
2075                 b -= c; b -= a; b ^= (a << 16); \
2076                 c -= a; c -= b; c ^= (b >> 5);  \
2077                 a -= b; a -= c; a ^= (c >> 3);  \
2078                 b -= c; b -= a; b ^= (a << 10); \
2079                 c -= a; c -= b; c ^= (b >> 15); \
2080         } while (0)
2081
2082 /*
2083  * hash function based on bridge_hash in if_bridge.c
2084  */
2085 void
2086 pf_hash(struct pf_addr *inaddr, struct pf_addr *hash,
2087     struct pf_poolhashkey *key, sa_family_t af)
2088 {
2089         u_int32_t       a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0];
2090
2091         switch (af) {
2092 #ifdef INET
2093         case AF_INET:
2094                 a += inaddr->addr32[0];
2095                 b += key->key32[1];
2096                 mix(a, b, c);
2097                 hash->addr32[0] = c + key->key32[2];
2098                 break;
2099 #endif /* INET */
2100 #ifdef INET6
2101         case AF_INET6:
2102                 a += inaddr->addr32[0];
2103                 b += inaddr->addr32[2];
2104                 mix(a, b, c);
2105                 hash->addr32[0] = c;
2106                 a += inaddr->addr32[1];
2107                 b += inaddr->addr32[3];
2108                 c += key->key32[1];
2109                 mix(a, b, c);
2110                 hash->addr32[1] = c;
2111                 a += inaddr->addr32[2];
2112                 b += inaddr->addr32[1];
2113                 c += key->key32[2];
2114                 mix(a, b, c);
2115                 hash->addr32[2] = c;
2116                 a += inaddr->addr32[3];
2117                 b += inaddr->addr32[0];
2118                 c += key->key32[3];
2119                 mix(a, b, c);
2120                 hash->addr32[3] = c;
2121                 break;
2122 #endif /* INET6 */
2123         }
2124 }
2125
2126 int
2127 pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr,
2128     struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn)
2129 {
2130         unsigned char            hash[16];
2131         struct pf_pool          *rpool = &r->rpool;
2132         struct pf_addr          *raddr = &rpool->cur->addr.v.a.addr;
2133         struct pf_addr          *rmask = &rpool->cur->addr.v.a.mask;
2134         struct pf_pooladdr      *acur = rpool->cur;
2135         struct pf_src_node       k;
2136
2137         if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR &&
2138             (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2139                 k.af = af;
2140                 PF_ACPY(&k.addr, saddr, af);
2141                 if (r->rule_flag & PFRULE_RULESRCTRACK ||
2142                     r->rpool.opts & PF_POOL_STICKYADDR)
2143                         k.rule.ptr = r;
2144                 else
2145                         k.rule.ptr = NULL;
2146                 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
2147                 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
2148                 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) {
2149                         PF_ACPY(naddr, &(*sn)->raddr, af);
2150                         if (pf_status.debug >= PF_DEBUG_MISC) {
2151                                 kprintf("pf_map_addr: src tracking maps ");
2152                                 pf_print_host(&k.addr, 0, af);
2153                                 kprintf(" to ");
2154                                 pf_print_host(naddr, 0, af);
2155                                 kprintf("\n");
2156                         }
2157                         return (0);
2158                 }
2159         }
2160
2161         if (rpool->cur->addr.type == PF_ADDR_NOROUTE)
2162                 return (1);
2163         if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2164                 switch (af) {
2165 #ifdef INET
2166                 case AF_INET:
2167                         if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 &&
2168                             (rpool->opts & PF_POOL_TYPEMASK) !=
2169                             PF_POOL_ROUNDROBIN)
2170                                 return (1);
2171                          raddr = &rpool->cur->addr.p.dyn->pfid_addr4;
2172                          rmask = &rpool->cur->addr.p.dyn->pfid_mask4;
2173                         break;
2174 #endif /* INET */
2175 #ifdef INET6
2176                 case AF_INET6:
2177                         if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 &&
2178                             (rpool->opts & PF_POOL_TYPEMASK) !=
2179                             PF_POOL_ROUNDROBIN)
2180                                 return (1);
2181                         raddr = &rpool->cur->addr.p.dyn->pfid_addr6;
2182                         rmask = &rpool->cur->addr.p.dyn->pfid_mask6;
2183                         break;
2184 #endif /* INET6 */
2185                 }
2186         } else if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2187                 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN)
2188                         return (1); /* unsupported */
2189         } else {
2190                 raddr = &rpool->cur->addr.v.a.addr;
2191                 rmask = &rpool->cur->addr.v.a.mask;
2192         }
2193
2194         switch (rpool->opts & PF_POOL_TYPEMASK) {
2195         case PF_POOL_NONE:
2196                 PF_ACPY(naddr, raddr, af);
2197                 break;
2198         case PF_POOL_BITMASK:
2199                 PF_POOLMASK(naddr, raddr, rmask, saddr, af);
2200                 break;
2201         case PF_POOL_RANDOM:
2202                 if (init_addr != NULL && PF_AZERO(init_addr, af)) {
2203                         switch (af) {
2204 #ifdef INET
2205                         case AF_INET:
2206                                 rpool->counter.addr32[0] = htonl(karc4random());
2207                                 break;
2208 #endif /* INET */
2209 #ifdef INET6
2210                         case AF_INET6:
2211                                 if (rmask->addr32[3] != 0xffffffff)
2212                                         rpool->counter.addr32[3] =
2213                                             htonl(karc4random());
2214                                 else
2215                                         break;
2216                                 if (rmask->addr32[2] != 0xffffffff)
2217                                         rpool->counter.addr32[2] =
2218                                             htonl(karc4random());
2219                                 else
2220                                         break;
2221                                 if (rmask->addr32[1] != 0xffffffff)
2222                                         rpool->counter.addr32[1] =
2223                                             htonl(karc4random());
2224                                 else
2225                                         break;
2226                                 if (rmask->addr32[0] != 0xffffffff)
2227                                         rpool->counter.addr32[0] =
2228                                             htonl(karc4random());
2229                                 break;
2230 #endif /* INET6 */
2231                         }
2232                         PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2233                         PF_ACPY(init_addr, naddr, af);
2234
2235                 } else {
2236                         PF_AINC(&rpool->counter, af);
2237                         PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2238                 }
2239                 break;
2240         case PF_POOL_SRCHASH:
2241                 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af);
2242                 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af);
2243                 break;
2244         case PF_POOL_ROUNDROBIN:
2245                 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2246                         if (!pfr_pool_get(rpool->cur->addr.p.tbl,
2247                             &rpool->tblidx, &rpool->counter,
2248                             &raddr, &rmask, af))
2249                                 goto get_addr;
2250                 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2251                         if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2252                             &rpool->tblidx, &rpool->counter,
2253                             &raddr, &rmask, af))
2254                                 goto get_addr;
2255                 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af))
2256                         goto get_addr;
2257
2258         try_next:
2259                 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL)
2260                         rpool->cur = TAILQ_FIRST(&rpool->list);
2261                 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2262                         rpool->tblidx = -1;
2263                         if (pfr_pool_get(rpool->cur->addr.p.tbl,
2264                             &rpool->tblidx, &rpool->counter,
2265                             &raddr, &rmask, af)) {
2266                                 /* table contains no address of type 'af' */
2267                                 if (rpool->cur != acur)
2268                                         goto try_next;
2269                                 return (1);
2270                         }
2271                 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2272                         rpool->tblidx = -1;
2273                         if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2274                             &rpool->tblidx, &rpool->counter,
2275                             &raddr, &rmask, af)) {
2276                                 /* table contains no address of type 'af' */
2277                                 if (rpool->cur != acur)
2278                                         goto try_next;
2279                                 return (1);
2280                         }
2281                 } else {
2282                         raddr = &rpool->cur->addr.v.a.addr;
2283                         rmask = &rpool->cur->addr.v.a.mask;
2284                         PF_ACPY(&rpool->counter, raddr, af);
2285                 }
2286
2287         get_addr:
2288                 PF_ACPY(naddr, &rpool->counter, af);
2289                 if (init_addr != NULL && PF_AZERO(init_addr, af))
2290                         PF_ACPY(init_addr, naddr, af);
2291                 PF_AINC(&rpool->counter, af);
2292                 break;
2293         }
2294         if (*sn != NULL)
2295                 PF_ACPY(&(*sn)->raddr, naddr, af);
2296
2297         if (pf_status.debug >= PF_DEBUG_MISC &&
2298             (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2299                 kprintf("pf_map_addr: selected address ");
2300                 pf_print_host(naddr, 0, af);
2301                 kprintf("\n");
2302         }
2303
2304         return (0);
2305 }
2306
2307 int
2308 pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r,
2309     struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport,
2310     struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high,
2311     struct pf_src_node **sn)
2312 {
2313         struct pf_state_key_cmp key;
2314         struct pf_addr          init_addr;
2315         u_int16_t               cut;
2316
2317         bzero(&init_addr, sizeof(init_addr));
2318         if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2319                 return (1);
2320
2321         if (proto == IPPROTO_ICMP) {
2322                 low = 1;
2323                 high = 65535;
2324         }
2325
2326         do {
2327                 key.af = af;
2328                 key.proto = proto;
2329                 PF_ACPY(&key.ext.addr, daddr, key.af);
2330                 PF_ACPY(&key.gwy.addr, naddr, key.af);
2331                 key.ext.port = dport;
2332
2333                 /*
2334                  * port search; start random, step;
2335                  * similar 2 portloop in in_pcbbind
2336                  */
2337                 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP ||
2338                     proto == IPPROTO_ICMP)) {
2339                         key.gwy.port = dport;
2340                         if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
2341                                 return (0);
2342                 } else if (low == 0 && high == 0) {
2343                         key.gwy.port = *nport;
2344                         if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
2345                                 return (0);
2346                 } else if (low == high) {
2347                         key.gwy.port = htons(low);
2348                         if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) {
2349                                 *nport = htons(low);
2350                                 return (0);
2351                         }
2352                 } else {
2353                         u_int16_t tmp;
2354
2355                         if (low > high) {
2356                                 tmp = low;
2357                                 low = high;
2358                                 high = tmp;
2359                         }
2360                         /* low < high */
2361                         cut = htonl(karc4random()) % (1 + high - low) + low;
2362                         /* low <= cut <= high */
2363                         for (tmp = cut; tmp <= high; ++(tmp)) {
2364                                 key.gwy.port = htons(tmp);
2365                                 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
2366                                     NULL) {
2367                                         *nport = htons(tmp);
2368                                         return (0);
2369                                 }
2370                         }
2371                         for (tmp = cut - 1; tmp >= low; --(tmp)) {
2372                                 key.gwy.port = htons(tmp);
2373                                 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
2374                                     NULL) {
2375                                         *nport = htons(tmp);
2376                                         return (0);
2377                                 }
2378                         }
2379                 }
2380
2381                 switch (r->rpool.opts & PF_POOL_TYPEMASK) {
2382                 case PF_POOL_RANDOM:
2383                 case PF_POOL_ROUNDROBIN:
2384                         if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2385                                 return (1);
2386                         break;
2387                 case PF_POOL_NONE:
2388                 case PF_POOL_SRCHASH:
2389                 case PF_POOL_BITMASK:
2390                 default:
2391                         return (1);
2392                 }
2393         } while (! PF_AEQ(&init_addr, naddr, af) );
2394
2395         return (1);                                     /* none available */
2396 }
2397
2398 struct pf_rule *
2399 pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off,
2400     int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport,
2401     struct pf_addr *daddr, u_int16_t dport, int rs_num)
2402 {
2403         struct pf_rule          *r, *rm = NULL;
2404         struct pf_ruleset       *ruleset = NULL;
2405         int                      tag = -1;
2406         int                      rtableid = -1;
2407         int                      asd = 0;
2408
2409         r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr);
2410         while (r && rm == NULL) {
2411                 struct pf_rule_addr     *src = NULL, *dst = NULL;
2412                 struct pf_addr_wrap     *xdst = NULL;
2413
2414                 if (r->action == PF_BINAT && direction == PF_IN) {
2415                         src = &r->dst;
2416                         if (r->rpool.cur != NULL)
2417                                 xdst = &r->rpool.cur->addr;
2418                 } else {
2419                         src = &r->src;
2420                         dst = &r->dst;
2421                 }
2422
2423                 r->evaluations++;
2424                 if (pfi_kif_match(r->kif, kif) == r->ifnot)
2425                         r = r->skip[PF_SKIP_IFP].ptr;
2426                 else if (r->direction && r->direction != direction)
2427                         r = r->skip[PF_SKIP_DIR].ptr;
2428                 else if (r->af && r->af != pd->af)
2429                         r = r->skip[PF_SKIP_AF].ptr;
2430                 else if (r->proto && r->proto != pd->proto)
2431                         r = r->skip[PF_SKIP_PROTO].ptr;
2432                 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af,
2433                     src->neg, kif))
2434                         r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR :
2435                             PF_SKIP_DST_ADDR].ptr;
2436                 else if (src->port_op && !pf_match_port(src->port_op,
2437                     src->port[0], src->port[1], sport))
2438                         r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT :
2439                             PF_SKIP_DST_PORT].ptr;
2440                 else if (dst != NULL &&
2441                     PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->neg, NULL))
2442                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
2443                 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af,
2444                     0, NULL))
2445                         r = TAILQ_NEXT(r, entries);
2446                 else if (dst != NULL && dst->port_op &&
2447                     !pf_match_port(dst->port_op, dst->port[0],
2448                     dst->port[1], dport))
2449                         r = r->skip[PF_SKIP_DST_PORT].ptr;
2450                 else if (r->match_tag && !pf_match_tag(m, r, &tag))
2451                         r = TAILQ_NEXT(r, entries);
2452                 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto !=
2453                     IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m,
2454                     off, pd->hdr.tcp), r->os_fingerprint)))
2455                         r = TAILQ_NEXT(r, entries);
2456                 else {
2457                         if (r->tag)
2458                                 tag = r->tag;
2459                         if (r->rtableid >= 0)
2460                                 rtableid = r->rtableid;
2461                         if (r->anchor == NULL) {
2462                                 rm = r;
2463                         } else
2464                                 pf_step_into_anchor(&asd, &ruleset, rs_num,
2465                                     &r, NULL, NULL);
2466                 }
2467                 if (r == NULL)
2468                         pf_step_out_of_anchor(&asd, &ruleset, rs_num, &r,
2469                             NULL, NULL);
2470         }
2471         if (pf_tag_packet(m, tag, rtableid))
2472                 return (NULL);
2473         if (rm != NULL && (rm->action == PF_NONAT ||
2474             rm->action == PF_NORDR || rm->action == PF_NOBINAT))
2475                 return (NULL);
2476         return (rm);
2477 }
2478
2479 struct pf_rule *
2480 pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction,
2481     struct pfi_kif *kif, struct pf_src_node **sn,
2482     struct pf_addr *saddr, u_int16_t sport,
2483     struct pf_addr *daddr, u_int16_t dport,
2484     struct pf_addr *naddr, u_int16_t *nport)
2485 {
2486         struct pf_rule  *r = NULL;
2487
2488         if (direction == PF_OUT) {
2489                 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2490                     sport, daddr, dport, PF_RULESET_BINAT);
2491                 if (r == NULL)
2492                         r = pf_match_translation(pd, m, off, direction, kif,
2493                             saddr, sport, daddr, dport, PF_RULESET_NAT);
2494         } else {
2495                 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2496                     sport, daddr, dport, PF_RULESET_RDR);
2497                 if (r == NULL)
2498                         r = pf_match_translation(pd, m, off, direction, kif,
2499                             saddr, sport, daddr, dport, PF_RULESET_BINAT);
2500         }
2501
2502         if (r != NULL) {
2503                 switch (r->action) {
2504                 case PF_NONAT:
2505                 case PF_NOBINAT:
2506                 case PF_NORDR:
2507                         return (NULL);
2508                 case PF_NAT:
2509                         if (pf_get_sport(pd->af, pd->proto, r, saddr,
2510                             daddr, dport, naddr, nport, r->rpool.proxy_port[0],
2511                             r->rpool.proxy_port[1], sn)) {
2512                                 DPFPRINTF(PF_DEBUG_MISC,
2513                                     ("pf: NAT proxy port allocation "
2514                                     "(%u-%u) failed\n",
2515                                     r->rpool.proxy_port[0],
2516                                     r->rpool.proxy_port[1]));
2517                                 return (NULL);
2518                         }
2519                         break;
2520                 case PF_BINAT:
2521                         switch (direction) {
2522                         case PF_OUT:
2523                                 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){
2524                                         switch (pd->af) {
2525 #ifdef INET
2526                                         case AF_INET:
2527                                                 if (r->rpool.cur->addr.p.dyn->
2528                                                     pfid_acnt4 < 1)
2529                                                         return (NULL);
2530                                                 PF_POOLMASK(naddr,
2531                                                     &r->rpool.cur->addr.p.dyn->
2532                                                     pfid_addr4,
2533                                                     &r->rpool.cur->addr.p.dyn->
2534                                                     pfid_mask4,
2535                                                     saddr, AF_INET);
2536                                                 break;
2537 #endif /* INET */
2538 #ifdef INET6
2539                                         case AF_INET6:
2540                                                 if (r->rpool.cur->addr.p.dyn->
2541                                                     pfid_acnt6 < 1)
2542                                                         return (NULL);
2543                                                 PF_POOLMASK(naddr,
2544                                                     &r->rpool.cur->addr.p.dyn->
2545                                                     pfid_addr6,
2546                                                     &r->rpool.cur->addr.p.dyn->
2547                                                     pfid_mask6,
2548                                                     saddr, AF_INET6);
2549                                                 break;
2550 #endif /* INET6 */
2551                                         }
2552                                 } else
2553                                         PF_POOLMASK(naddr,
2554                                             &r->rpool.cur->addr.v.a.addr,
2555                                             &r->rpool.cur->addr.v.a.mask,
2556                                             saddr, pd->af);
2557                                 break;
2558                         case PF_IN:
2559                                 if (r->src.addr.type == PF_ADDR_DYNIFTL) {
2560                                         switch (pd->af) {
2561 #ifdef INET
2562                                         case AF_INET:
2563                                                 if (r->src.addr.p.dyn->
2564                                                     pfid_acnt4 < 1)
2565                                                         return (NULL);
2566                                                 PF_POOLMASK(naddr,
2567                                                     &r->src.addr.p.dyn->
2568                                                     pfid_addr4,
2569                                                     &r->src.addr.p.dyn->
2570                                                     pfid_mask4,
2571                                                     daddr, AF_INET);
2572                                                 break;
2573 #endif /* INET */
2574 #ifdef INET6
2575                                         case AF_INET6:
2576                                                 if (r->src.addr.p.dyn->
2577                                                     pfid_acnt6 < 1)
2578                                                         return (NULL);
2579                                                 PF_POOLMASK(naddr,
2580                                                     &r->src.addr.p.dyn->
2581                                                     pfid_addr6,
2582                                                     &r->src.addr.p.dyn->
2583                                                     pfid_mask6,
2584                                                     daddr, AF_INET6);
2585                                                 break;
2586 #endif /* INET6 */
2587                                         }
2588                                 } else
2589                                         PF_POOLMASK(naddr,
2590                                             &r->src.addr.v.a.addr,
2591                                             &r->src.addr.v.a.mask, daddr,
2592                                             pd->af);
2593                                 break;
2594                         }
2595                         break;
2596                 case PF_RDR: {
2597                         if (pf_map_addr(pd->af, r, saddr, naddr, NULL, sn))
2598                                 return (NULL);
2599                         if ((r->rpool.opts & PF_POOL_TYPEMASK) ==
2600                             PF_POOL_BITMASK)
2601                                 PF_POOLMASK(naddr, naddr,
2602                                     &r->rpool.cur->addr.v.a.mask, daddr,
2603                                     pd->af);
2604
2605                         if (r->rpool.proxy_port[1]) {
2606                                 u_int32_t       tmp_nport;
2607
2608                                 tmp_nport = ((ntohs(dport) -
2609                                     ntohs(r->dst.port[0])) %
2610                                     (r->rpool.proxy_port[1] -
2611                                     r->rpool.proxy_port[0] + 1)) +
2612                                     r->rpool.proxy_port[0];
2613
2614                                 /* wrap around if necessary */
2615                                 if (tmp_nport > 65535)
2616                                         tmp_nport -= 65535;
2617                                 *nport = htons((u_int16_t)tmp_nport);
2618                         } else if (r->rpool.proxy_port[0])
2619                                 *nport = htons(r->rpool.proxy_port[0]);
2620                         break;
2621                 }
2622                 default:
2623                         return (NULL);
2624                 }
2625         }
2626
2627         return (r);
2628 }
2629
2630 #ifdef SMP
2631 struct netmsg_hashlookup {
2632         struct netmsg_base      base;
2633         struct inpcb            **nm_pinp;
2634         struct inpcbinfo        *nm_pcbinfo;
2635         struct pf_addr          *nm_saddr;
2636         struct pf_addr          *nm_daddr;
2637         uint16_t                nm_sport;
2638         uint16_t                nm_dport;
2639         sa_family_t             nm_af;
2640 };
2641
2642 static void
2643 in_pcblookup_hash_handler(netmsg_t msg)
2644 {
2645         struct netmsg_hashlookup *rmsg = (struct netmsg_hashlookup *)msg;
2646
2647         if (rmsg->nm_af == AF_INET)
2648                 *rmsg->nm_pinp = in_pcblookup_hash(rmsg->nm_pcbinfo,
2649                     rmsg->nm_saddr->v4, rmsg->nm_sport, rmsg->nm_daddr->v4,
2650                     rmsg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2651 #ifdef INET6
2652         else
2653                 *rmsg->nm_pinp = in6_pcblookup_hash(rmsg->nm_pcbinfo,
2654                     &rmsg->nm_saddr->v6, rmsg->nm_sport, &rmsg->nm_daddr->v6,
2655                     rmsg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2656 #endif /* INET6 */
2657         lwkt_replymsg(&rmsg->base.lmsg, 0);
2658 }
2659 #endif /* SMP */
2660
2661 int
2662 pf_socket_lookup(int direction, struct pf_pdesc *pd)
2663 {
2664         struct pf_addr          *saddr, *daddr;
2665         u_int16_t                sport, dport;
2666         struct inpcbinfo        *pi;
2667         struct inpcb            *inp;
2668 #ifdef SMP
2669         struct netmsg_hashlookup *msg = NULL;
2670 #endif
2671         int                      pi_cpu = 0;
2672
2673         if (pd == NULL)
2674                 return (-1);
2675         pd->lookup.uid = UID_MAX;
2676         pd->lookup.gid = GID_MAX;
2677         pd->lookup.pid = NO_PID;
2678         if (direction == PF_IN) {
2679                 saddr = pd->src;
2680                 daddr = pd->dst;
2681         } else {
2682                 saddr = pd->dst;
2683                 daddr = pd->src;
2684         }
2685         switch (pd->proto) {
2686         case IPPROTO_TCP:
2687                 if (pd->hdr.tcp == NULL)
2688                         return (-1);
2689                 sport = pd->hdr.tcp->th_sport;
2690                 dport = pd->hdr.tcp->th_dport;
2691
2692                 pi_cpu = tcp_addrcpu(saddr->v4.s_addr, sport, daddr->v4.s_addr, dport);
2693                 pi = &tcbinfo[pi_cpu];
2694 #ifdef SMP
2695                 /*
2696                  * Our netstack runs lockless on MP systems
2697                  * (only for TCP connections at the moment).
2698                  *
2699                  * As we are not allowed to read another CPU's tcbinfo,
2700                  * we have to ask that CPU via remote call to search the
2701                  * table for us.
2702                  *
2703                  * Prepare a msg iff data belongs to another CPU.
2704                  */
2705                 if (pi_cpu != mycpu->gd_cpuid) {
2706                         msg = kmalloc(sizeof(*msg), M_LWKTMSG, M_INTWAIT);
2707                         netmsg_init(&msg->base, NULL, &netisr_afree_rport,
2708                                     0, in_pcblookup_hash_handler);
2709                         msg->nm_pinp = &inp;
2710                         msg->nm_pcbinfo = pi;
2711                         msg->nm_saddr = saddr;
2712                         msg->nm_sport = sport;
2713                         msg->nm_daddr = daddr;
2714                         msg->nm_dport = dport;
2715                         msg->nm_af = pd->af;
2716                 }
2717 #endif /* SMP */
2718                 break;
2719         case IPPROTO_UDP:
2720                 if (pd->hdr.udp == NULL)
2721                         return (-1);
2722                 sport = pd->hdr.udp->uh_sport;
2723                 dport = pd->hdr.udp->uh_dport;
2724                 pi = &udbinfo;
2725                 break;
2726         default:
2727                 return (-1);
2728         }
2729         if (direction != PF_IN) {
2730                 u_int16_t       p;
2731
2732                 p = sport;
2733                 sport = dport;
2734                 dport = p;
2735         }
2736         switch (pd->af) {
2737 #ifdef INET6
2738         case AF_INET6:
2739 #ifdef SMP
2740                 /*
2741                  * Query other CPU, second part
2742                  *
2743                  * msg only gets initialized when:
2744                  * 1) packet is TCP
2745                  * 2) the info belongs to another CPU
2746                  *
2747                  * Use some switch/case magic to avoid code duplication.
2748                  */
2749                 if (msg == NULL)
2750 #endif /* SMP */
2751                 {
2752                         inp = in6_pcblookup_hash(pi, &saddr->v6, sport,
2753                             &daddr->v6, dport, INPLOOKUP_WILDCARD, NULL);
2754
2755                         if (inp == NULL)
2756                                 return (-1);
2757                         break;
2758                 }
2759                 /* FALLTHROUGH if SMP and on other CPU */
2760 #endif /* INET6 */
2761         case AF_INET:
2762 #ifdef SMP
2763                 if (msg != NULL) {
2764                         lwkt_domsg(cpu_portfn(pi_cpu),
2765                                      &msg->base.lmsg, 0);
2766                 } else
2767 #endif /* SMP */
2768                 {
2769                         inp = in_pcblookup_hash(pi, saddr->v4, sport, daddr->v4,
2770                             dport, INPLOOKUP_WILDCARD, NULL);
2771                 }
2772                 if (inp == NULL)
2773                         return (-1);
2774                 break;
2775
2776         default:
2777                 return (-1);
2778         }
2779         pd->lookup.uid = inp->inp_socket->so_cred->cr_uid;
2780         pd->lookup.gid = inp->inp_socket->so_cred->cr_groups[0];
2781         return (1);
2782 }
2783
2784 u_int8_t
2785 pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2786 {
2787         int              hlen;
2788         u_int8_t         hdr[60];
2789         u_int8_t        *opt, optlen;
2790         u_int8_t         wscale = 0;
2791
2792         hlen = th_off << 2;             /* hlen <= sizeof(hdr) */
2793         if (hlen <= sizeof(struct tcphdr))
2794                 return (0);
2795         if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2796                 return (0);
2797         opt = hdr + sizeof(struct tcphdr);
2798         hlen -= sizeof(struct tcphdr);
2799         while (hlen >= 3) {
2800                 switch (*opt) {
2801                 case TCPOPT_EOL:
2802                 case TCPOPT_NOP:
2803                         ++opt;
2804                         --hlen;
2805                         break;
2806                 case TCPOPT_WINDOW:
2807                         wscale = opt[2];
2808                         if (wscale > TCP_MAX_WINSHIFT)
2809                                 wscale = TCP_MAX_WINSHIFT;
2810                         wscale |= PF_WSCALE_FLAG;
2811                         /* FALLTHROUGH */
2812                 default:
2813                         optlen = opt[1];
2814                         if (optlen < 2)
2815                                 optlen = 2;
2816                         hlen -= optlen;
2817                         opt += optlen;
2818                         break;
2819                 }
2820         }
2821         return (wscale);
2822 }
2823
2824 u_int16_t
2825 pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2826 {
2827         int              hlen;
2828         u_int8_t         hdr[60];
2829         u_int8_t        *opt, optlen;
2830         u_int16_t        mss = tcp_mssdflt;
2831
2832         hlen = th_off << 2;     /* hlen <= sizeof(hdr) */
2833         if (hlen <= sizeof(struct tcphdr))
2834                 return (0);
2835         if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2836                 return (0);
2837         opt = hdr + sizeof(struct tcphdr);
2838         hlen -= sizeof(struct tcphdr);
2839         while (hlen >= TCPOLEN_MAXSEG) {
2840                 switch (*opt) {
2841                 case TCPOPT_EOL:
2842                 case TCPOPT_NOP:
2843                         ++opt;
2844                         --hlen;
2845                         break;
2846                 case TCPOPT_MAXSEG:
2847                         bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2);
2848                         /* FALLTHROUGH */
2849                 default:
2850                         optlen = opt[1];
2851                         if (optlen < 2)
2852                                 optlen = 2;
2853                         hlen -= optlen;
2854                         opt += optlen;
2855                         break;
2856                 }
2857         }
2858         return (mss);
2859 }
2860
2861 u_int16_t
2862 pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer)
2863 {
2864 #ifdef INET
2865         struct sockaddr_in      *dst;
2866         struct route             ro;
2867 #endif /* INET */
2868 #ifdef INET6
2869         struct sockaddr_in6     *dst6;
2870         struct route_in6         ro6;
2871 #endif /* INET6 */
2872         struct rtentry          *rt = NULL;
2873         int                      hlen = 0;
2874         u_int16_t                mss = tcp_mssdflt;
2875
2876         switch (af) {
2877 #ifdef INET
2878         case AF_INET:
2879                 hlen = sizeof(struct ip);
2880                 bzero(&ro, sizeof(ro));
2881                 dst = (struct sockaddr_in *)&ro.ro_dst;
2882                 dst->sin_family = AF_INET;
2883                 dst->sin_len = sizeof(*dst);
2884                 dst->sin_addr = addr->v4;
2885                 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
2886                 rt = ro.ro_rt;
2887                 break;
2888 #endif /* INET */
2889 #ifdef INET6
2890         case AF_INET6:
2891                 hlen = sizeof(struct ip6_hdr);
2892                 bzero(&ro6, sizeof(ro6));
2893                 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst;
2894                 dst6->sin6_family = AF_INET6;
2895                 dst6->sin6_len = sizeof(*dst6);
2896                 dst6->sin6_addr = addr->v6;
2897                 rtalloc_ign((struct route *)&ro6, (RTF_CLONING | RTF_PRCLONING));
2898                 rt = ro6.ro_rt;
2899                 break;
2900 #endif /* INET6 */
2901         }
2902
2903         if (rt && rt->rt_ifp) {
2904                 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr);
2905                 mss = max(tcp_mssdflt, mss);
2906                 RTFREE(rt);
2907         }
2908         mss = min(mss, offer);
2909         mss = max(mss, 64);             /* sanity - at least max opt space */
2910         return (mss);
2911 }
2912
2913 void
2914 pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr)
2915 {
2916         struct pf_rule *r = s->rule.ptr;
2917
2918         s->rt_kif = NULL;
2919         if (!r->rt || r->rt == PF_FASTROUTE)
2920                 return;
2921         switch (s->state_key->af) {
2922 #ifdef INET
2923         case AF_INET:
2924                 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL,
2925                     &s->nat_src_node);
2926                 s->rt_kif = r->rpool.cur->kif;
2927                 break;
2928 #endif /* INET */
2929 #ifdef INET6
2930         case AF_INET6:
2931                 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL,
2932                     &s->nat_src_node);
2933                 s->rt_kif = r->rpool.cur->kif;
2934                 break;
2935 #endif /* INET6 */
2936         }
2937 }
2938
2939 void
2940 pf_attach_state(struct pf_state_key *sk, struct pf_state *s, int tail)
2941 {
2942         s->state_key = sk;
2943         sk->refcnt++;
2944
2945         /* list is sorted, if-bound states before floating */
2946         if (tail)
2947                 TAILQ_INSERT_TAIL(&sk->states, s, next);
2948         else
2949                 TAILQ_INSERT_HEAD(&sk->states, s, next);
2950 }
2951
2952 void
2953 pf_detach_state(struct pf_state *s, int flags)
2954 {
2955         struct pf_state_key     *sk = s->state_key;
2956
2957         if (sk == NULL)
2958                 return;
2959
2960         s->state_key = NULL;
2961         TAILQ_REMOVE(&sk->states, s, next);
2962         if (--sk->refcnt == 0) {
2963                 if (!(flags & PF_DT_SKIP_EXTGWY))
2964                         RB_REMOVE(pf_state_tree_ext_gwy,
2965                             &pf_statetbl_ext_gwy, sk);
2966                 if (!(flags & PF_DT_SKIP_LANEXT))
2967                         RB_REMOVE(pf_state_tree_lan_ext,
2968                             &pf_statetbl_lan_ext, sk);
2969                 pool_put(&pf_state_key_pl, sk);
2970         }
2971 }
2972
2973 struct pf_state_key *
2974 pf_alloc_state_key(struct pf_state *s)
2975 {
2976         struct pf_state_key     *sk;
2977
2978         if ((sk = pool_get(&pf_state_key_pl, PR_NOWAIT)) == NULL)
2979                 return (NULL);
2980         bzero(sk, sizeof(*sk));
2981         TAILQ_INIT(&sk->states);
2982         pf_attach_state(sk, s, 0);
2983
2984         return (sk);
2985 }
2986
2987 int
2988 pf_test_rule(struct pf_rule **rm, struct pf_state **sm, int direction,
2989     struct pfi_kif *kif, struct mbuf *m, int off, void *h,
2990     struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
2991     struct ifqueue *ifq, struct inpcb *inp)
2992 {
2993         struct pf_rule          *nr = NULL;
2994         struct pf_addr          *saddr = pd->src, *daddr = pd->dst;
2995         u_int16_t                bport, nport = 0;
2996         sa_family_t              af = pd->af;
2997         struct pf_rule          *r, *a = NULL;
2998         struct pf_ruleset       *ruleset = NULL;
2999         struct pf_src_node      *nsn = NULL;
3000         struct tcphdr           *th = pd->hdr.tcp;
3001         u_short                  reason;
3002         int                      rewrite = 0, hdrlen = 0;
3003         int                      tag = -1, rtableid = -1;
3004         int                      asd = 0;
3005         int                      match = 0;
3006         int                      state_icmp = 0;
3007         u_int16_t                mss = tcp_mssdflt;
3008         u_int16_t                sport, dport;
3009         u_int8_t                 icmptype = 0, icmpcode = 0;
3010
3011         if (direction == PF_IN && pf_check_congestion(ifq)) {
3012                 REASON_SET(&reason, PFRES_CONGEST);
3013                 return (PF_DROP);
3014         }
3015
3016         if (inp != NULL)
3017                 pd->lookup.done = pf_socket_lookup(direction, pd);
3018         else if (debug_pfugidhack) {
3019                 DPFPRINTF(PF_DEBUG_MISC, ("pf: unlocked lookup\n"));
3020                 pd->lookup.done = pf_socket_lookup(direction, pd);
3021         }
3022
3023         sport = dport = hdrlen = 0;
3024
3025         switch (pd->proto) {
3026         case IPPROTO_TCP:
3027                 sport = th->th_sport;
3028                 dport = th->th_dport;
3029                 hdrlen = sizeof(*th);
3030                 break;
3031         case IPPROTO_UDP:
3032                 sport = pd->hdr.udp->uh_sport;
3033                 dport = pd->hdr.udp->uh_dport;
3034                 hdrlen = sizeof(*pd->hdr.udp);
3035                 break;
3036 #ifdef INET
3037         case IPPROTO_ICMP:
3038                 if (pd->af != AF_INET)
3039                         break;
3040                 sport = dport = pd->hdr.icmp->icmp_id;
3041                 icmptype = pd->hdr.icmp->icmp_type;
3042                 icmpcode = pd->hdr.icmp->icmp_code;
3043
3044                 if (icmptype == ICMP_UNREACH ||
3045                     icmptype == ICMP_SOURCEQUENCH ||
3046                     icmptype == ICMP_REDIRECT ||
3047                     icmptype == ICMP_TIMXCEED ||
3048                     icmptype == ICMP_PARAMPROB)
3049                         state_icmp++;
3050                 break;
3051 #endif /* INET */
3052 #ifdef INET6
3053         case IPPROTO_ICMPV6:
3054                 if (pd->af != AF_INET6)
3055                         break;
3056                 sport = dport = pd->hdr.icmp6->icmp6_id;
3057                 hdrlen = sizeof(*pd->hdr.icmp6);
3058                 icmptype = pd->hdr.icmp6->icmp6_type;
3059                 icmpcode = pd->hdr.icmp6->icmp6_code;
3060
3061                 if (icmptype == ICMP6_DST_UNREACH ||
3062                     icmptype == ICMP6_PACKET_TOO_BIG ||
3063                     icmptype == ICMP6_TIME_EXCEEDED ||
3064                     icmptype == ICMP6_PARAM_PROB)
3065                         state_icmp++;
3066                 break;
3067 #endif /* INET6 */
3068         }
3069
3070         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3071
3072         if (direction == PF_OUT) {
3073                 bport = nport = sport;
3074                 /* check outgoing packet for BINAT/NAT */
3075                 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3076                     saddr, sport, daddr, dport, &pd->naddr, &nport)) != NULL) {
3077                         PF_ACPY(&pd->baddr, saddr, af);
3078                         switch (pd->proto) {
3079                         case IPPROTO_TCP:
3080                                 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
3081                                     &th->th_sum, &pd->naddr, nport, 0, af);
3082                                 sport = th->th_sport;
3083                                 rewrite++;
3084                                 break;
3085                         case IPPROTO_UDP:
3086                                 pf_change_ap(saddr, &pd->hdr.udp->uh_sport,
3087                                     pd->ip_sum, &pd->hdr.udp->uh_sum,
3088                                     &pd->naddr, nport, 1, af);
3089                                 sport = pd->hdr.udp->uh_sport;
3090                                 rewrite++;
3091                                 break;
3092 #ifdef INET
3093                         case IPPROTO_ICMP:
3094                                 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3095                                     pd->naddr.v4.s_addr, 0);
3096                                 pd->hdr.icmp->icmp_cksum = pf_cksum_fixup(
3097                                     pd->hdr.icmp->icmp_cksum, sport, nport, 0);
3098                                 pd->hdr.icmp->icmp_id = nport;
3099                                 m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
3100                                 break;
3101 #endif /* INET */
3102 #ifdef INET6
3103                         case IPPROTO_ICMPV6:
3104                                 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum,
3105                                     &pd->naddr, 0);
3106                                 rewrite++;
3107                                 break;
3108 #endif /* INET */
3109                         default:
3110                                 switch (af) {
3111 #ifdef INET
3112                                 case AF_INET:
3113                                         pf_change_a(&saddr->v4.s_addr,
3114                                             pd->ip_sum, pd->naddr.v4.s_addr, 0);
3115                                         break;
3116 #endif /* INET */
3117 #ifdef INET6
3118                                 case AF_INET6:
3119                                         PF_ACPY(saddr, &pd->naddr, af);
3120                                         break;
3121 #endif /* INET */
3122                                 }
3123                                 break;
3124                         }
3125
3126                         if (nr->natpass)
3127                                 r = NULL;
3128                         pd->nat_rule = nr;
3129                 }
3130         } else {
3131                 bport = nport = dport;
3132                 /* check incoming packet for BINAT/RDR */
3133                 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3134                     saddr, sport, daddr, dport, &pd->naddr, &nport)) != NULL) {
3135                         PF_ACPY(&pd->baddr, daddr, af);
3136                         switch (pd->proto) {
3137                         case IPPROTO_TCP:
3138                                 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
3139                                     &th->th_sum, &pd->naddr, nport, 0, af);
3140                                 dport = th->th_dport;
3141                                 rewrite++;
3142                                 break;
3143                         case IPPROTO_UDP:
3144                                 pf_change_ap(daddr, &pd->hdr.udp->uh_dport,
3145                                     pd->ip_sum, &pd->hdr.udp->uh_sum,
3146                                     &pd->naddr, nport, 1, af);
3147                                 dport = pd->hdr.udp->uh_dport;
3148                                 rewrite++;
3149                                 break;
3150 #ifdef INET
3151                         case IPPROTO_ICMP:
3152                                 pf_change_a(&daddr->v4.s_addr, pd->ip_sum,
3153                                     pd->naddr.v4.s_addr, 0);
3154                                 break;
3155 #endif /* INET */
3156 #ifdef INET6
3157                         case IPPROTO_ICMPV6:
3158                                 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum,
3159                                     &pd->naddr, 0);
3160                                 rewrite++;
3161                                 break;
3162 #endif /* INET6 */
3163                         default:
3164                                 switch (af) {
3165 #ifdef INET
3166                                 case AF_INET:
3167                                         pf_change_a(&daddr->v4.s_addr,
3168                                             pd->ip_sum, pd->naddr.v4.s_addr, 0);
3169                                         break;
3170 #endif /* INET */
3171 #ifdef INET6
3172                                 case AF_INET6:
3173                                         PF_ACPY(daddr, &pd->naddr, af);
3174                                         break;
3175 #endif /* INET */
3176                                 }
3177                                 break;
3178                         }
3179
3180                         if (nr->natpass)
3181                                 r = NULL;
3182                         pd->nat_rule = nr;
3183                 }
3184         }
3185
3186         while (r != NULL) {
3187                 r->evaluations++;
3188                 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3189                         r = r->skip[PF_SKIP_IFP].ptr;
3190                 else if (r->direction && r->direction != direction)
3191                         r = r->skip[PF_SKIP_DIR].ptr;
3192                 else if (r->af && r->af != af)
3193                         r = r->skip[PF_SKIP_AF].ptr;
3194                 else if (r->proto && r->proto != pd->proto)
3195                         r = r->skip[PF_SKIP_PROTO].ptr;
3196                 else if (PF_MISMATCHAW(&r->src.addr, saddr, af,
3197                     r->src.neg, kif))
3198                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3199                 /* tcp/udp only. port_op always 0 in other cases */
3200                 else if (r->src.port_op && !pf_match_port(r->src.port_op,
3201                     r->src.port[0], r->src.port[1], sport))
3202                         r = r->skip[PF_SKIP_SRC_PORT].ptr;
3203                 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af,
3204                     r->dst.neg, NULL))
3205                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
3206                 /* tcp/udp only. port_op always 0 in other cases */
3207                 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
3208                     r->dst.port[0], r->dst.port[1], dport))
3209                         r = r->skip[PF_SKIP_DST_PORT].ptr;
3210                 /* icmp only. type always 0 in other cases */
3211                 else if (r->type && r->type != icmptype + 1)
3212                         r = TAILQ_NEXT(r, entries);
3213                 /* icmp only. type always 0 in other cases */
3214                 else if (r->code && r->code != icmpcode + 1)
3215                         r = TAILQ_NEXT(r, entries);
3216                 else if (r->tos && !(r->tos == pd->tos))
3217                         r = TAILQ_NEXT(r, entries);
3218                 else if (r->rule_flag & PFRULE_FRAGMENT)
3219                         r = TAILQ_NEXT(r, entries);
3220                 else if (pd->proto == IPPROTO_TCP &&
3221                     (r->flagset & th->th_flags) != r->flags)
3222                         r = TAILQ_NEXT(r, entries);
3223                 /* tcp/udp only. uid.op always 0 in other cases */
3224                 else if (r->uid.op && (pd->lookup.done || (pd->lookup.done =
3225                     pf_socket_lookup(direction, pd), 1)) &&
3226                     !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
3227                     pd->lookup.uid))
3228                         r = TAILQ_NEXT(r, entries);
3229                 /* tcp/udp only. gid.op always 0 in other cases */
3230                 else if (r->gid.op && (pd->lookup.done || (pd->lookup.done =
3231                     pf_socket_lookup(direction, pd), 1)) &&
3232                     !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
3233                     pd->lookup.gid))
3234                         r = TAILQ_NEXT(r, entries);
3235                 else if (r->prob && r->prob <= karc4random())
3236                         r = TAILQ_NEXT(r, entries);
3237                 else if (r->match_tag && !pf_match_tag(m, r, &tag))
3238                         r = TAILQ_NEXT(r, entries);
3239                 else if (r->os_fingerprint != PF_OSFP_ANY &&
3240                     (pd->proto != IPPROTO_TCP || !pf_osfp_match(
3241                     pf_osfp_fingerprint(pd, m, off, th),
3242                     r->os_fingerprint)))
3243                         r = TAILQ_NEXT(r, entries);
3244                 else {
3245                         if (r->tag)
3246                                 tag = r->tag;
3247                         if (r->rtableid >= 0)
3248                                 rtableid = r->rtableid;
3249                         if (r->anchor == NULL) {
3250                                 match = 1;
3251                                 *rm = r;
3252                                 *am = a;
3253                                 *rsm = ruleset;
3254                                 if ((*rm)->quick)
3255                                         break;
3256                                 r = TAILQ_NEXT(r, entries);
3257                         } else
3258                                 pf_step_into_anchor(&asd, &ruleset,
3259                                     PF_RULESET_FILTER, &r, &a, &match);
3260                 }
3261                 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3262                     PF_RULESET_FILTER, &r, &a, &match))
3263                         break;
3264         }
3265         r = *rm;
3266         a = *am;
3267         ruleset = *rsm;
3268
3269         REASON_SET(&reason, PFRES_MATCH);
3270
3271         if (r->log || (nr != NULL && nr->log)) {
3272                 if (rewrite)
3273                         m_copyback(m, off, hdrlen, pd->hdr.any);
3274                 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr,
3275                     a, ruleset, pd);
3276         }
3277
3278         if ((r->action == PF_DROP) &&
3279             ((r->rule_flag & PFRULE_RETURNRST) ||
3280             (r->rule_flag & PFRULE_RETURNICMP) ||
3281             (r->rule_flag & PFRULE_RETURN))) {
3282                 /* undo NAT changes, if they have taken place */
3283                 if (nr != NULL) {
3284                         if (direction == PF_OUT) {
3285                                 switch (pd->proto) {
3286                                 case IPPROTO_TCP:
3287                                         pf_change_ap(saddr, &th->th_sport,
3288                                             pd->ip_sum, &th->th_sum,
3289                                             &pd->baddr, bport, 0, af);
3290                                         sport = th->th_sport;
3291                                         rewrite++;
3292                                         break;
3293                                 case IPPROTO_UDP:
3294                                         pf_change_ap(saddr,
3295                                             &pd->hdr.udp->uh_sport, pd->ip_sum,
3296                                             &pd->hdr.udp->uh_sum, &pd->baddr,
3297                                             bport, 1, af);
3298                                         sport = pd->hdr.udp->uh_sport;
3299                                         rewrite++;
3300                                         break;
3301                                 case IPPROTO_ICMP:
3302 #ifdef INET6
3303                                 case IPPROTO_ICMPV6:
3304 #endif
3305                                         /* nothing! */
3306                                         break;
3307                                 default:
3308                                         switch (af) {
3309                                         case AF_INET:
3310                                                 pf_change_a(&saddr->v4.s_addr,
3311                                                     pd->ip_sum,
3312                                                     pd->baddr.v4.s_addr, 0);
3313                                                 break;
3314                                         case AF_INET6:
3315                                                 PF_ACPY(saddr, &pd->baddr, af);
3316                                                 break;
3317                                         }
3318                                 }
3319                         } else {
3320                                 switch (pd->proto) {
3321                                 case IPPROTO_TCP:
3322                                         pf_change_ap(daddr, &th->th_dport,
3323                                             pd->ip_sum, &th->th_sum,
3324                                             &pd->baddr, bport, 0, af);
3325                                         dport = th->th_dport;
3326                                         rewrite++;
3327                                         break;
3328                                 case IPPROTO_UDP:
3329                                         pf_change_ap(daddr,
3330                                             &pd->hdr.udp->uh_dport, pd->ip_sum,
3331                                             &pd->hdr.udp->uh_sum, &pd->baddr,
3332                                             bport, 1, af);
3333                                         dport = pd->hdr.udp->uh_dport;
3334                                         rewrite++;
3335                                         break;
3336                                 case IPPROTO_ICMP:
3337 #ifdef INET6
3338                                 case IPPROTO_ICMPV6:
3339 #endif
3340                                         /* nothing! */
3341                                         break;
3342                                 default:
3343                                         switch (af) {
3344                                         case AF_INET:
3345                                                 pf_change_a(&daddr->v4.s_addr,
3346                                                     pd->ip_sum,
3347                                                     pd->baddr.v4.s_addr, 0);
3348                                                 break;
3349                                         case AF_INET6:
3350                                                 PF_ACPY(daddr, &pd->baddr, af);
3351                                                 break;
3352                                         }
3353                                 }
3354                         }
3355                 }
3356                 if (pd->proto == IPPROTO_TCP &&
3357                     ((r->rule_flag & PFRULE_RETURNRST) ||
3358                     (r->rule_flag & PFRULE_RETURN)) &&
3359                     !(th->th_flags & TH_RST)) {
3360                         u_int32_t        ack = ntohl(th->th_seq) + pd->p_len;
3361                         struct ip       *h = mtod(m, struct ip *);
3362
3363                         if (pf_check_proto_cksum(m, off,
3364                             h->ip_len - off, IPPROTO_TCP, AF_INET))
3365                                 REASON_SET(&reason, PFRES_PROTCKSUM);
3366                         else {
3367                                 if (th->th_flags & TH_SYN)
3368                                         ack++;
3369                                 if (th->th_flags & TH_FIN)
3370                                         ack++;
3371                                 pf_send_tcp(r, af, pd->dst,
3372                                     pd->src, th->th_dport, th->th_sport,
3373                                     ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0,
3374                                     r->return_ttl, 1, 0, pd->eh, kif->pfik_ifp);
3375                         }
3376                 } else if ((af == AF_INET) && r->return_icmp)
3377                         pf_send_icmp(m, r->return_icmp >> 8,
3378                             r->return_icmp & 255, af, r);
3379                 else if ((af == AF_INET6) && r->return_icmp6)
3380                         pf_send_icmp(m, r->return_icmp6 >> 8,
3381                             r->return_icmp6 & 255, af, r);
3382         }
3383
3384         if (r->action == PF_DROP)
3385                 return (PF_DROP);
3386
3387         if (pf_tag_packet(m, tag, rtableid)) {
3388                 REASON_SET(&reason, PFRES_MEMORY);
3389                 return (PF_DROP);
3390         }
3391
3392         if (!state_icmp && (r->keep_state || nr != NULL ||
3393             (pd->flags & PFDESC_TCP_NORM))) {
3394                 /* create new state */
3395                 u_int16_t        len = 0;
3396                 struct pf_state *s = NULL;
3397                 struct pf_state_key *sk = NULL;
3398                 struct pf_src_node *sn = NULL;
3399
3400                 /* check maximums */
3401                 if (r->max_states && (r->states >= r->max_states)) {
3402                         pf_status.lcounters[LCNT_STATES]++;
3403                         REASON_SET(&reason, PFRES_MAXSTATES);
3404                         goto cleanup;
3405                 }
3406                 /* src node for filter rule */
3407                 if ((r->rule_flag & PFRULE_SRCTRACK ||
3408                     r->rpool.opts & PF_POOL_STICKYADDR) &&
3409                     pf_insert_src_node(&sn, r, saddr, af) != 0) {
3410                         REASON_SET(&reason, PFRES_SRCLIMIT);
3411                         goto cleanup;
3412                 }
3413                 /* src node for translation rule */
3414                 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3415                     ((direction == PF_OUT &&
3416                     pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3417                     (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) {
3418                         REASON_SET(&reason, PFRES_SRCLIMIT);
3419                         goto cleanup;
3420                 }
3421                 s = pool_get(&pf_state_pl, PR_NOWAIT);
3422                 if (s == NULL) {
3423                         REASON_SET(&reason, PFRES_MEMORY);
3424 cleanup:
3425                         if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3426                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3427                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3428                                 pf_status.src_nodes--;
3429                                 pool_put(&pf_src_tree_pl, sn);
3430                         }
3431                         if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3432                             nsn->expire == 0) {
3433                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3434                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3435                                 pf_status.src_nodes--;
3436                                 pool_put(&pf_src_tree_pl, nsn);
3437                         }
3438                         if (sk != NULL) {
3439                                 pool_put(&pf_state_key_pl, sk);
3440                         }
3441                         return (PF_DROP);
3442                 }
3443                 bzero(s, sizeof(*s));
3444                 s->rule.ptr = r;
3445                 s->nat_rule.ptr = nr;
3446                 s->anchor.ptr = a;
3447                 STATE_INC_COUNTERS(s);
3448                 s->allow_opts = r->allow_opts;
3449                 s->log = r->log & PF_LOG_ALL;
3450                 if (nr != NULL)
3451                         s->log |= nr->log & PF_LOG_ALL;
3452                 switch (pd->proto) {
3453                 case IPPROTO_TCP:
3454                         len = pd->tot_len - off - (th->th_off << 2);
3455                         s->src.seqlo = ntohl(th->th_seq);
3456                         s->src.seqhi = s->src.seqlo + len + 1;
3457                         if ((th->th_flags & (TH_SYN|TH_ACK)) ==
3458                         TH_SYN && r->keep_state == PF_STATE_MODULATE) {
3459                                 /* Generate sequence number modulator */
3460                                 while ((s->src.seqdiff =
3461                                     pf_new_isn(sk) - s->src.seqlo) == 0)
3462                                         ;
3463                                 pf_change_a(&th->th_seq, &th->th_sum,
3464                                     htonl(s->src.seqlo + s->src.seqdiff), 0);
3465                                 rewrite = 1;
3466                         } else
3467                                 s->src.seqdiff = 0;
3468                         if (th->th_flags & TH_SYN) {
3469                                 s->src.seqhi++;
3470                                 s->src.wscale = pf_get_wscale(m, off,
3471                                     th->th_off, af);
3472                         }
3473                         s->src.max_win = MAX(ntohs(th->th_win), 1);
3474                         if (s->src.wscale & PF_WSCALE_MASK) {
3475                                 /* Remove scale factor from initial window */
3476                                 int win = s->src.max_win;
3477                                 win += 1 << (s->src.wscale & PF_WSCALE_MASK);
3478                                 s->src.max_win = (win - 1) >>
3479                                     (s->src.wscale & PF_WSCALE_MASK);
3480                         }
3481                         if (th->th_flags & TH_FIN)
3482                                 s->src.seqhi++;
3483                         s->dst.seqhi = 1;
3484                         s->dst.max_win = 1;
3485                         s->src.state = TCPS_SYN_SENT;
3486                         s->dst.state = TCPS_CLOSED;
3487                         s->timeout = PFTM_TCP_FIRST_PACKET;
3488                         break;
3489                 case IPPROTO_UDP:
3490                         s->src.state = PFUDPS_SINGLE;
3491                         s->dst.state = PFUDPS_NO_TRAFFIC;
3492                         s->timeout = PFTM_UDP_FIRST_PACKET;
3493                         break;
3494                 case IPPROTO_ICMP:
3495 #ifdef INET6
3496                 case IPPROTO_ICMPV6:
3497 #endif
3498                         s->timeout = PFTM_ICMP_FIRST_PACKET;
3499                         break;
3500                 default:
3501                         s->src.state = PFOTHERS_SINGLE;
3502                         s->dst.state = PFOTHERS_NO_TRAFFIC;
3503                         s->timeout = PFTM_OTHER_FIRST_PACKET;
3504                 }
3505
3506                 s->creation = time_second;
3507                 s->expire = time_second;
3508
3509                 if (sn != NULL) {
3510                         s->src_node = sn;
3511                         s->src_node->states++;
3512                 }
3513                 if (nsn != NULL) {
3514                         PF_ACPY(&nsn->raddr, &pd->naddr, af);
3515                         s->nat_src_node = nsn;
3516                         s->nat_src_node->states++;
3517                 }
3518                 if (pd->proto == IPPROTO_TCP) {
3519                         if ((pd->flags & PFDESC_TCP_NORM) &&
3520                             pf_normalize_tcp_init(m, off, pd, th, &s->src,
3521                             &s->dst)) {
3522                                 REASON_SET(&reason, PFRES_MEMORY);
3523                                 pf_src_tree_remove_state(s);
3524                                 STATE_DEC_COUNTERS(s);
3525                                 pool_put(&pf_state_pl, s);
3526                                 return (PF_DROP);
3527                         }
3528                         if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
3529                             pf_normalize_tcp_stateful(m, off, pd, &reason,
3530                             th, s, &s->src, &s->dst, &rewrite)) {
3531                                 /* This really shouldn't happen!!! */
3532                                 DPFPRINTF(PF_DEBUG_URGENT,
3533                                     ("pf_normalize_tcp_stateful failed on "
3534                                     "first pkt"));
3535                                 pf_normalize_tcp_cleanup(s);
3536                                 pf_src_tree_remove_state(s);
3537                                 STATE_DEC_COUNTERS(s);
3538                                 pool_put(&pf_state_pl, s);
3539                                 return (PF_DROP);
3540                         }
3541                 }
3542
3543                 if ((sk = pf_alloc_state_key(s)) == NULL) {
3544                         REASON_SET(&reason, PFRES_MEMORY);
3545                         goto cleanup;
3546                 }
3547
3548                 sk->proto = pd->proto;
3549                 sk->direction = direction;
3550                 sk->af = af;
3551                 if (direction == PF_OUT) {
3552                         PF_ACPY(&sk->gwy.addr, saddr, af);
3553                         PF_ACPY(&sk->ext.addr, daddr, af);
3554                         switch (pd->proto) {
3555                         case IPPROTO_ICMP:
3556 #ifdef INET6
3557                         case IPPROTO_ICMPV6:
3558 #endif
3559                                 sk->gwy.port = nport;
3560                                 sk->ext.port = 0;
3561                                 break;
3562                         default:
3563                                 sk->gwy.port = sport;
3564                                 sk->ext.port = dport;
3565                         }
3566                         if (nr != NULL) {
3567                                 PF_ACPY(&sk->lan.addr, &pd->baddr, af);
3568                                 sk->lan.port = bport;
3569                         } else {
3570                                 PF_ACPY(&sk->lan.addr, &sk->gwy.addr, af);
3571                                 sk->lan.port = sk->gwy.port;
3572                         }
3573                 } else {
3574                         PF_ACPY(&sk->lan.addr, daddr, af);
3575                         PF_ACPY(&sk->ext.addr, saddr, af);
3576                         switch (pd->proto) {
3577                         case IPPROTO_ICMP:
3578 #ifdef INET6
3579                         case IPPROTO_ICMPV6:
3580 #endif
3581                                 sk->lan.port = nport;
3582                                 sk->ext.port = 0;
3583                                 break;
3584                         default:
3585                                 sk->lan.port = dport;
3586                                 sk->ext.port = sport;
3587                         }
3588                         if (nr != NULL) {
3589                                 PF_ACPY(&sk->gwy.addr, &pd->baddr, af);
3590                                 sk->gwy.port = bport;
3591                         } else {
3592                                 PF_ACPY(&sk->gwy.addr, &sk->lan.addr, af);
3593                                 sk->gwy.port = sk->lan.port;
3594                         }
3595                 }
3596
3597                 s->hash = pf_state_hash(sk);
3598                 s->pickup_mode = r->pickup_mode;
3599
3600                 pf_set_rt_ifp(s, saddr);        /* needs s->state_key set */
3601
3602                 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3603                         if (pd->proto == IPPROTO_TCP)
3604                                 pf_normalize_tcp_cleanup(s);
3605                         REASON_SET(&reason, PFRES_STATEINS);
3606                         pf_src_tree_remove_state(s);
3607                         STATE_DEC_COUNTERS(s);
3608                         pool_put(&pf_state_pl, s);
3609                         return (PF_DROP);
3610                 } else
3611                         *sm = s;
3612                 if (tag > 0) {
3613                         pf_tag_ref(tag);
3614                         s->tag = tag;
3615                 }
3616                 if (pd->proto == IPPROTO_TCP &&
3617                     (th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
3618                     r->keep_state == PF_STATE_SYNPROXY) {
3619                         s->src.state = PF_TCPS_PROXY_SRC;
3620                         if (nr != NULL) {
3621                                 if (direction == PF_OUT) {
3622                                         pf_change_ap(saddr, &th->th_sport,
3623                                             pd->ip_sum, &th->th_sum, &pd->baddr,
3624                                             bport, 0, af);
3625                                         sport = th->th_sport;
3626                                 } else {
3627                                         pf_change_ap(daddr, &th->th_dport,
3628                                             pd->ip_sum, &th->th_sum, &pd->baddr,
3629                                             bport, 0, af);
3630                                         sport = th->th_dport;
3631                                 }
3632                         }
3633                         s->src.seqhi = htonl(karc4random());
3634                         /* Find mss option */
3635                         mss = pf_get_mss(m, off, th->th_off, af);
3636                         mss = pf_calc_mss(saddr, af, mss);
3637                         mss = pf_calc_mss(daddr, af, mss);
3638                         s->src.mss = mss;
3639                         pf_send_tcp(r, af, daddr, saddr, th->th_dport,
3640                             th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1,
3641                             TH_SYN|TH_ACK, 0, s->src.mss, 0, 1, 0, NULL, NULL);
3642                         REASON_SET(&reason, PFRES_SYNPROXY);
3643                         return (PF_SYNPROXY_DROP);
3644                 }
3645         }
3646
3647         /* copy back packet headers if we performed NAT operations */
3648         if (rewrite)
3649                 m_copyback(m, off, hdrlen, pd->hdr.any);
3650
3651         return (PF_PASS);
3652 }
3653
3654 int
3655 pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif,
3656     struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am,
3657     struct pf_ruleset **rsm)
3658 {
3659         struct pf_rule          *r, *a = NULL;
3660         struct pf_ruleset       *ruleset = NULL;
3661         sa_family_t              af = pd->af;
3662         u_short                  reason;
3663         int                      tag = -1;
3664         int                      asd = 0;
3665         int                      match = 0;
3666
3667         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3668         while (r != NULL) {
3669                 r->evaluations++;
3670                 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3671                         r = r->skip[PF_SKIP_IFP].ptr;
3672                 else if (r->direction && r->direction != direction)
3673                         r = r->skip[PF_SKIP_DIR].ptr;
3674                 else if (r->af && r->af != af)
3675                         r = r->skip[PF_SKIP_AF].ptr;
3676                 else if (r->proto && r->proto != pd->proto)
3677                         r = r->skip[PF_SKIP_PROTO].ptr;
3678                 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af,
3679                     r->src.neg, kif))
3680                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3681                 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af,
3682                     r->dst.neg, NULL))
3683                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
3684                 else if (r->tos && !(r->tos == pd->tos))
3685                         r = TAILQ_NEXT(r, entries);
3686                 else if (r->os_fingerprint != PF_OSFP_ANY)
3687                         r = TAILQ_NEXT(r, entries);
3688                 else if (pd->proto == IPPROTO_UDP &&
3689                     (r->src.port_op || r->dst.port_op))
3690                         r = TAILQ_NEXT(r, entries);
3691                 else if (pd->proto == IPPROTO_TCP &&
3692                     (r->src.port_op || r->dst.port_op || r->flagset))
3693                         r = TAILQ_NEXT(r, entries);
3694                 else if ((pd->proto == IPPROTO_ICMP ||
3695                     pd->proto == IPPROTO_ICMPV6) &&
3696                     (r->type || r->code))
3697                         r = TAILQ_NEXT(r, entries);
3698                 else if (r->prob && r->prob <= karc4random())
3699                         r = TAILQ_NEXT(r, entries);
3700                 else if (r->match_tag && !pf_match_tag(m, r, &tag))
3701                         r = TAILQ_NEXT(r, entries);
3702                 else {
3703                         if (r->anchor == NULL) {
3704                                 match = 1;
3705                                 *rm = r;
3706                                 *am = a;
3707                                 *rsm = ruleset;
3708                                 if ((*rm)->quick)
3709                                         break;
3710                                 r = TAILQ_NEXT(r, entries);
3711                         } else
3712                                 pf_step_into_anchor(&asd, &ruleset,
3713                                     PF_RULESET_FILTER, &r, &a, &match);
3714                 }
3715                 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3716                     PF_RULESET_FILTER, &r, &a, &match))
3717                         break;
3718         }
3719         r = *rm;
3720         a = *am;
3721         ruleset = *rsm;
3722
3723         REASON_SET(&reason, PFRES_MATCH);
3724
3725         if (r->log)
3726                 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset,
3727                     pd);
3728
3729         if (r->action != PF_PASS)
3730                 return (PF_DROP);
3731
3732         if (pf_tag_packet(m, tag, -1)) {
3733                 REASON_SET(&reason, PFRES_MEMORY);
3734                 return (PF_DROP);
3735         }
3736
3737         return (PF_PASS);
3738 }
3739
3740 int
3741 pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
3742     struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
3743     u_short *reason)
3744 {
3745         struct pf_state_key_cmp  key;
3746         struct tcphdr           *th = pd->hdr.tcp;
3747         u_int16_t                win = ntohs(th->th_win);
3748         u_int32_t                ack, end, seq, orig_seq;
3749         u_int8_t                 sws, dws;
3750         int                      ackskew;
3751         int                      copyback = 0;
3752         struct pf_state_peer    *src, *dst;
3753
3754         key.af = pd->af;
3755         key.proto = IPPROTO_TCP;
3756         if (direction == PF_IN) {
3757                 PF_ACPY(&key.ext.addr, pd->src, key.af);
3758                 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
3759                 key.ext.port = th->th_sport;
3760                 key.gwy.port = th->th_dport;
3761         } else {
3762                 PF_ACPY(&key.lan.addr, pd->src, key.af);
3763                 PF_ACPY(&key.ext.addr, pd->dst, key.af);
3764                 key.lan.port = th->th_sport;
3765                 key.ext.port = th->th_dport;
3766         }
3767
3768         STATE_LOOKUP();
3769
3770         if (direction == (*state)->state_key->direction) {
3771                 src = &(*state)->src;
3772                 dst = &(*state)->dst;
3773         } else {
3774                 src = &(*state)->dst;
3775                 dst = &(*state)->src;
3776         }
3777
3778         if ((*state)->src.state == PF_TCPS_PROXY_SRC) {
3779                 if (direction != (*state)->state_key->direction) {
3780                         REASON_SET(reason, PFRES_SYNPROXY);
3781                         return (PF_SYNPROXY_DROP);
3782                 }
3783                 if (th->th_flags & TH_SYN) {
3784                         if (ntohl(th->th_seq) != (*state)->src.seqlo) {
3785                                 REASON_SET(reason, PFRES_SYNPROXY);
3786                                 return (PF_DROP);
3787                         }
3788                         pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
3789                             pd->src, th->th_dport, th->th_sport,
3790                             (*state)->src.seqhi, ntohl(th->th_seq) + 1,
3791                             TH_SYN|TH_ACK, 0, (*state)->src.mss, 0, 1,
3792                             0, NULL, NULL);
3793                         REASON_SET(reason, PFRES_SYNPROXY);
3794                         return (PF_SYNPROXY_DROP);
3795                 } else if (!(th->th_flags & TH_ACK) ||
3796                     (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
3797                     (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
3798                         REASON_SET(reason, PFRES_SYNPROXY);
3799                         return (PF_DROP);
3800                 } else if ((*state)->src_node != NULL &&
3801                     pf_src_connlimit(state)) {
3802                         REASON_SET(reason, PFRES_SRCLIMIT);
3803                         return (PF_DROP);
3804                 } else
3805                         (*state)->src.state = PF_TCPS_PROXY_DST;
3806         }
3807         if ((*state)->src.state == PF_TCPS_PROXY_DST) {
3808                 struct pf_state_host *src, *dst;
3809
3810                 if (direction == PF_OUT) {
3811                         src = &(*state)->state_key->gwy;
3812                         dst = &(*state)->state_key->ext;
3813                 } else {
3814                         src = &(*state)->state_key->ext;
3815                         dst = &(*state)->state_key->lan;
3816                 }
3817                 if (direction == (*state)->state_key->direction) {
3818                         if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) ||
3819                             (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
3820                             (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
3821                                 REASON_SET(reason, PFRES_SYNPROXY);
3822                                 return (PF_DROP);
3823                         }
3824                         (*state)->src.max_win = MAX(ntohs(th->th_win), 1);
3825                         if ((*state)->dst.seqhi == 1)
3826                                 (*state)->dst.seqhi = htonl(karc4random());
3827                         pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
3828                             &dst->addr, src->port, dst->port,
3829                             (*state)->dst.seqhi, 0, TH_SYN, 0,
3830                             (*state)->src.mss, 0, 0, (*state)->tag, NULL, NULL);
3831                         REASON_SET(reason, PFRES_SYNPROXY);
3832                         return (PF_SYNPROXY_DROP);
3833                 } else if (((th->th_flags & (TH_SYN|TH_ACK)) !=
3834                     (TH_SYN|TH_ACK)) ||
3835                     (ntohl(th->th_ack) != (*state)->dst.seqhi + 1)) {
3836                         REASON_SET(reason, PFRES_SYNPROXY);
3837                         return (PF_DROP);
3838                 } else {
3839                         (*state)->dst.max_win = MAX(ntohs(th->th_win), 1);
3840                         (*state)->dst.seqlo = ntohl(th->th_seq);
3841                         pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
3842                             pd->src, th->th_dport, th->th_sport,
3843                             ntohl(th->th_ack), ntohl(th->th_seq) + 1,
3844                             TH_ACK, (*state)->src.max_win, 0, 0, 0,
3845                             (*state)->tag, NULL, NULL);
3846                         pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
3847                             &dst->addr, src->port, dst->port,
3848                             (*state)->src.seqhi + 1, (*state)->src.seqlo + 1,
3849                             TH_ACK, (*state)->dst.max_win, 0, 0, 1,
3850                             0, NULL, NULL);
3851                         (*state)->src.seqdiff = (*state)->dst.seqhi -
3852                             (*state)->src.seqlo;
3853                         (*state)->dst.seqdiff = (*state)->src.seqhi -
3854                             (*state)->dst.seqlo;
3855                         (*state)->src.seqhi = (*state)->src.seqlo +
3856                             (*state)->dst.max_win;
3857                         (*state)->dst.seqhi = (*state)->dst.seqlo +
3858                             (*state)->src.max_win;
3859                         (*state)->src.wscale = (*state)->dst.wscale = 0;
3860                         (*state)->src.state = (*state)->dst.state =
3861                             TCPS_ESTABLISHED;
3862                         REASON_SET(reason, PFRES_SYNPROXY);
3863                         return (PF_SYNPROXY_DROP);
3864                 }
3865         }
3866
3867         if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) {
3868                 sws = src->wscale & PF_WSCALE_MASK;
3869                 dws = dst->wscale & PF_WSCALE_MASK;
3870         } else
3871                 sws = dws = 0;
3872
3873         /*
3874          * Sequence tracking algorithm from Guido van Rooij's paper:
3875          *   http://www.madison-gurkha.com/publications/tcp_filtering/
3876          *      tcp_filtering.ps
3877          */
3878
3879         orig_seq = seq = ntohl(th->th_seq);
3880         if (src->seqlo == 0) {
3881                 /* First packet from this end. Set its state */
3882
3883                 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) &&
3884                     src->scrub == NULL) {
3885                         if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) {
3886                                 REASON_SET(reason, PFRES_MEMORY);
3887                                 return (PF_DROP);
3888                         }
3889                 }
3890
3891                 /* Deferred generation of sequence number modulator */
3892                 if (dst->seqdiff && !src->seqdiff) {
3893
3894                         while ((src->seqdiff = pf_new_isn((struct pf_state_key *)&key) - seq) == 0)
3895                                 ;
3896                         ack = ntohl(th->th_ack) - dst->seqdiff;
3897                         pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3898                             src->seqdiff), 0);
3899                         pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3900                         copyback = 1;
3901                 } else {
3902                         ack = ntohl(th->th_ack);
3903                 }
3904
3905                 end = seq + pd->p_len;
3906                 if (th->th_flags & TH_SYN) {
3907                         end++;
3908                         (*state)->sync_flags |= PFSTATE_GOT_SYN2;
3909                         if (dst->wscale & PF_WSCALE_FLAG) {
3910                                 src->wscale = pf_get_wscale(m, off, th->th_off,
3911                                     pd->af);
3912                                 if (src->wscale & PF_WSCALE_FLAG) {
3913                                         /* Remove scale factor from initial
3914                                          * window */
3915                                         sws = src->wscale & PF_WSCALE_MASK;
3916                                         win = ((u_int32_t)win + (1 << sws) - 1)
3917                                             >> sws;
3918                                         dws = dst->wscale & PF_WSCALE_MASK;
3919                                 } else {
3920                                         /* fixup other window */
3921                                         dst->max_win <<= dst->wscale &
3922                                             PF_WSCALE_MASK;
3923                                         /* in case of a retrans SYN|ACK */
3924                                         dst->wscale = 0;
3925                                 }
3926                         }
3927                 }
3928                 if (th->th_flags & TH_FIN)
3929                         end++;
3930
3931                 src->seqlo = seq;
3932                 if (src->state < TCPS_SYN_SENT)
3933                         src->state = TCPS_SYN_SENT;
3934
3935                 /*
3936                  * May need to slide the window (seqhi may have been set by
3937                  * the crappy stack check or if we picked up the connection
3938                  * after establishment)
3939                  */
3940                 if (src->seqhi == 1 ||
3941                     SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi))
3942                         src->seqhi = end + MAX(1, dst->max_win << dws);
3943                 if (win > src->max_win)
3944                         src->max_win = win;
3945
3946         } else {
3947                 ack = ntohl(th->th_ack) - dst->seqdiff;
3948                 if (src->seqdiff) {
3949                         /* Modulate sequence numbers */
3950                         pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3951                             src->seqdiff), 0);
3952                         pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3953                         copyback = 1;
3954                 }
3955                 end = seq + pd->p_len;
3956                 if (th->th_flags & TH_SYN)
3957                         end++;
3958                 if (th->th_flags & TH_FIN)
3959                         end++;
3960         }
3961
3962         if ((th->th_flags & TH_ACK) == 0) {
3963                 /* Let it pass through the ack skew check */
3964                 ack = dst->seqlo;
3965         } else if ((ack == 0 &&
3966             (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) ||
3967             /* broken tcp stacks do not set ack */
3968             (dst->state < TCPS_SYN_SENT)) {
3969                 /*
3970                  * Many stacks (ours included) will set the ACK number in an
3971                  * FIN|ACK if the SYN times out -- no sequence to ACK.
3972                  */
3973                 ack = dst->seqlo;
3974         }
3975
3976         if (seq == end) {
3977                 /* Ease sequencing restrictions on no data packets */
3978                 seq = src->seqlo;
3979                 end = seq;
3980         }
3981
3982         ackskew = dst->seqlo - ack;
3983
3984
3985         /*
3986          * Need to demodulate the sequence numbers in any TCP SACK options
3987          * (Selective ACK). We could optionally validate the SACK values
3988          * against the current ACK window, either forwards or backwards, but
3989          * I'm not confident that SACK has been implemented properly
3990          * everywhere. It wouldn't surprise me if several stacks accidently
3991          * SACK too far backwards of previously ACKed data. There really aren't
3992          * any security implications of bad SACKing unless the target stack
3993          * doesn't validate the option length correctly. Someone trying to
3994          * spoof into a TCP connection won't bother blindly sending SACK
3995          * options anyway.
3996          */
3997         if (dst->seqdiff && (th->th_off << 2) > sizeof(struct tcphdr)) {
3998                 if (pf_modulate_sack(m, off, pd, th, dst))
3999                         copyback = 1;
4000         }
4001
4002
4003 #define MAXACKWINDOW (0xffff + 1500)    /* 1500 is an arbitrary fudge factor */
4004         if (SEQ_GEQ(src->seqhi, end) &&
4005             /* Last octet inside other's window space */
4006             SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) &&
4007             /* Retrans: not more than one window back */
4008             (ackskew >= -MAXACKWINDOW) &&
4009             /* Acking not more than one reassembled fragment backwards */
4010             (ackskew <= (MAXACKWINDOW << sws)) &&
4011             /* Acking not more than one window forward */
4012             ((th->th_flags & TH_RST) == 0 || orig_seq == src->seqlo ||
4013             (orig_seq == src->seqlo + 1) || (orig_seq + 1 == src->seqlo) ||
4014             (pd->flags & PFDESC_IP_REAS) == 0)) {
4015             /* Require an exact/+1 sequence match on resets when possible */
4016
4017                 if (dst->scrub || src->scrub) {
4018                         if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4019                             *state, src, dst, &copyback))
4020                                 return (PF_DROP);
4021                 }
4022
4023                 /* update max window */
4024                 if (src->max_win < win)
4025                         src->max_win = win;
4026                 /* synchronize sequencing */
4027                 if (SEQ_GT(end, src->seqlo))
4028                         src->seqlo = end;
4029                 /* slide the window of what the other end can send */
4030                 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4031                         dst->seqhi = ack + MAX((win << sws), 1);
4032
4033
4034                 /* update states */
4035                 if (th->th_flags & TH_SYN)
4036                         if (src->state < TCPS_SYN_SENT)
4037                                 src->state = TCPS_SYN_SENT;
4038                 if (th->th_flags & TH_FIN)
4039                         if (src->state < TCPS_CLOSING)
4040                                 src->state = TCPS_CLOSING;
4041                 if (th->th_flags & TH_ACK) {
4042                         if (dst->state == TCPS_SYN_SENT) {
4043                                 dst->state = TCPS_ESTABLISHED;
4044                                 if (src->state == TCPS_ESTABLISHED &&
4045                                     (*state)->src_node != NULL &&
4046                                     pf_src_connlimit(state)) {
4047                                         REASON_SET(reason, PFRES_SRCLIMIT);
4048                                         return (PF_DROP);
4049                                 }
4050                         } else if (dst->state == TCPS_CLOSING)
4051                                 dst->state = TCPS_FIN_WAIT_2;
4052                 }
4053                 if (th->th_flags & TH_RST)
4054                         src->state = dst->state = TCPS_TIME_WAIT;
4055
4056                 /* update expire time */
4057                 (*state)->expire = time_second;
4058                 if (src->state >= TCPS_FIN_WAIT_2 &&
4059                     dst->state >= TCPS_FIN_WAIT_2)
4060                         (*state)->timeout = PFTM_TCP_CLOSED;
4061                 else if (src->state >= TCPS_CLOSING &&
4062                     dst->state >= TCPS_CLOSING)
4063                         (*state)->timeout = PFTM_TCP_FIN_WAIT;
4064                 else if (src->state < TCPS_ESTABLISHED ||
4065                     dst->state < TCPS_ESTABLISHED)
4066                         (*state)->timeout = PFTM_TCP_OPENING;
4067                 else if (src->state >= TCPS_CLOSING ||
4068                     dst->state >= TCPS_CLOSING)
4069                         (*state)->timeout = PFTM_TCP_CLOSING;
4070                 else
4071                         (*state)->timeout = PFTM_TCP_ESTABLISHED;
4072
4073                 /* Fall through to PASS packet */
4074
4075         } else if ((dst->state < TCPS_SYN_SENT ||
4076                 dst->state >= TCPS_FIN_WAIT_2 ||
4077                 src->state >= TCPS_FIN_WAIT_2) &&
4078             SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) &&
4079             /* Within a window forward of the originating packet */
4080             SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) {
4081             /* Within a window backward of the originating packet */
4082
4083                 /*
4084                  * This currently handles three situations:
4085                  *  1) Stupid stacks will shotgun SYNs before their peer
4086                  *     replies.
4087                  *  2) When PF catches an already established stream (the
4088                  *     firewall rebooted, the state table was flushed, routes
4089                  *     changed...)
4090                  *  3) Packets get funky immediately after the connection
4091                  *     closes (this should catch Solaris spurious ACK|FINs
4092                  *     that web servers like to spew after a close)
4093                  *
4094                  * This must be a little more careful than the above code
4095                  * since packet floods will also be caught here. We don't
4096                  * update the TTL here to mitigate the damage of a packet
4097                  * flood and so the same code can handle awkward establishment
4098                  * and a loosened connection close.
4099                  * In the establishment case, a correct peer response will
4100                  * validate the connection, go through the normal state code
4101                  * and keep updating the state TTL.
4102                  */
4103
4104                 if (pf_status.debug >= PF_DEBUG_MISC) {
4105                         kprintf("pf: loose state match: ");
4106                         pf_print_state(*state);
4107                         pf_print_flags(th->th_flags);
4108                         kprintf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
4109                             "pkts=%llu:%llu dir=%s,%s\n", seq, orig_seq, ack, pd->p_len,
4110                             ackskew, (unsigned long long)(*state)->packets[0],
4111                             (unsigned long long)(*state)->packets[1],
4112                             direction == PF_IN ? "in" : "out",
4113                             direction == (*state)->state_key->direction ?
4114                                 "fwd" : "rev");
4115                 }
4116
4117                 if (dst->scrub || src->scrub) {
4118                         if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4119                             *state, src, dst, &copyback))
4120                                 return (PF_DROP);
4121                 }
4122
4123                 /* update max window */
4124                 if (src->max_win < win)
4125                         src->max_win = win;
4126                 /* synchronize sequencing */
4127                 if (SEQ_GT(end, src->seqlo))
4128                         src->seqlo = end;
4129                 /* slide the window of what the other end can send */
4130                 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4131                         dst->seqhi = ack + MAX((win << sws), 1);
4132
4133                 /*
4134                  * Cannot set dst->seqhi here since this could be a shotgunned
4135                  * SYN and not an already established connection.
4136                  */
4137
4138                 if (th->th_flags & TH_FIN)
4139                         if (src->state < TCPS_CLOSING)
4140                                 src->state = TCPS_CLOSING;
4141                 if (th->th_flags & TH_RST)
4142                         src->state = dst->state = TCPS_TIME_WAIT;
4143
4144                 /* Fall through to PASS packet */
4145
4146         } else if ((*state)->pickup_mode == PF_PICKUPS_HASHONLY ||
4147                     ((*state)->pickup_mode == PF_PICKUPS_ENABLED &&
4148                      ((*state)->sync_flags & PFSTATE_GOT_SYN_MASK) !=
4149                       PFSTATE_GOT_SYN_MASK)) {
4150                 /*
4151                  * If pickup mode is hash only, do not fail on sequence checks.
4152                  *
4153                  * If pickup mode is enabled and we did not see the SYN in
4154                  * both direction, do not fail on sequence checks because
4155                  * we do not have complete information on window scale.
4156                  *
4157                  * Adjust expiration and fall through to PASS packet.
4158                  * XXX Add a FIN check to reduce timeout?
4159                  */
4160                 (*state)->expire = time_second;
4161         } else  {
4162                 /*
4163                  * Failure processing
4164                  */
4165                 if ((*state)->dst.state == TCPS_SYN_SENT &&
4166                     (*state)->src.state == TCPS_SYN_SENT) {
4167                         /* Send RST for state mismatches during handshake */
4168                         if (!(th->th_flags & TH_RST))
4169                                 pf_send_tcp((*state)->rule.ptr, pd->af,
4170                                     pd->dst, pd->src, th->th_dport,
4171                                     th->th_sport, ntohl(th->th_ack), 0,
4172                                     TH_RST, 0, 0,
4173                                     (*state)->rule.ptr->return_ttl, 1, 0,
4174                                     pd->eh, kif->pfik_ifp);
4175                         src->seqlo = 0;
4176                         src->seqhi = 1;
4177                         src->max_win = 1;
4178                 } else if (pf_status.debug >= PF_DEBUG_MISC) {
4179                         kprintf("pf: BAD state: ");
4180                         pf_print_state(*state);
4181                         pf_print_flags(th->th_flags);
4182                         kprintf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
4183                             "pkts=%llu:%llu dir=%s,%s\n",
4184                             seq, orig_seq, ack, pd->p_len, ackskew,
4185                             (unsigned long long)(*state)->packets[0],
4186                                 (unsigned long long)(*state)->packets[1],
4187                             direction == PF_IN ? "in" : "out",
4188                             direction == (*state)->state_key->direction ?
4189                                 "fwd" : "rev");
4190                         kprintf("pf: State failure on: %c %c %c %c | %c %c\n",
4191                             SEQ_GEQ(src->seqhi, end) ? ' ' : '1',
4192                             SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ?
4193                             ' ': '2',
4194                             (ackskew >= -MAXACKWINDOW) ? ' ' : '3',
4195                             (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4',
4196                             SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5',
4197                             SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6');
4198                 }
4199                 REASON_SET(reason, PFRES_BADSTATE);
4200                 return (PF_DROP);
4201         }
4202
4203         /* Any packets which have gotten here are to be passed */
4204
4205         /* translate source/destination address, if necessary */
4206         if (STATE_TRANSLATE((*state)->state_key)) {
4207                 if (direction == PF_OUT) {
4208                         pf_change_ap(pd->src, &th->th_sport, pd->ip_sum,
4209                             &th->th_sum, &(*state)->state_key->gwy.addr,
4210                             (*state)->state_key->gwy.port, 0, pd->af);
4211                 } else {
4212                         /*
4213                          * If we don't redispatch the packet will go into
4214                          * the protocol stack on the wrong cpu for the
4215                          * post-translated address.
4216                          */
4217                         /* m->m_pkthdr.fw_flags |= FW_MBUF_REDISPATCH; */
4218                         m->m_flags &= ~M_HASH;
4219                         pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum,
4220                             &th->th_sum, &(*state)->state_key->lan.addr,
4221                             (*state)->state_key->lan.port, 0, pd->af);
4222                 }
4223                 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4224         } else if (copyback) {
4225                 /* Copyback sequence modulation or stateful scrub changes */
4226                 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4227         }
4228
4229         return (PF_PASS);
4230 }
4231
4232 int
4233 pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif,
4234     struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4235 {
4236         struct pf_state_peer    *src, *dst;
4237         struct pf_state_key_cmp  key;
4238         struct udphdr           *uh = pd->hdr.udp;
4239
4240         key.af = pd->af;
4241         key.proto = IPPROTO_UDP;
4242         if (direction == PF_IN) {
4243                 PF_ACPY(&key.ext.addr, pd->src, key.af);
4244                 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4245                 key.ext.port = uh->uh_sport;
4246                 key.gwy.port = uh->uh_dport;
4247         } else {
4248                 PF_ACPY(&key.lan.addr, pd->src, key.af);
4249                 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4250                 key.lan.port = uh->uh_sport;
4251                 key.ext.port = uh->uh_dport;
4252         }
4253
4254         STATE_LOOKUP();
4255
4256         if (direction == (*state)->state_key->direction) {
4257                 src = &(*state)->src;
4258                 dst = &(*state)->dst;
4259         } else {
4260                 src = &(*state)->dst;
4261                 dst = &(*state)->src;
4262         }
4263
4264         /* update states */
4265         if (src->state < PFUDPS_SINGLE)
4266                 src->state = PFUDPS_SINGLE;
4267         if (dst->state == PFUDPS_SINGLE)
4268                 dst->state = PFUDPS_MULTIPLE;
4269
4270         /* update expire time */
4271         (*state)->expire = time_second;
4272         if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE)
4273                 (*state)->timeout = PFTM_UDP_MULTIPLE;
4274         else
4275                 (*state)->timeout = PFTM_UDP_SINGLE;
4276
4277         /* translate source/destination address, if necessary */
4278         if (STATE_TRANSLATE((*state)->state_key)) {
4279                 if (direction == PF_OUT) {
4280                         pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum,
4281                             &uh->uh_sum, &(*state)->state_key->gwy.addr,
4282                             (*state)->state_key->gwy.port, 1, pd->af);
4283                 } else {
4284                         /*
4285                          * If we don't redispatch the packet will go into
4286                          * the protocol stack on the wrong cpu for the
4287                          * post-translated address.
4288                          */
4289                         /* m->m_pkthdr.fw_flags |= FW_MBUF_REDISPATCH; */
4290                         m->m_flags &= ~M_HASH;
4291                         pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum,
4292                             &uh->uh_sum, &(*state)->state_key->lan.addr,
4293                             (*state)->state_key->lan.port, 1, pd->af);
4294                 }
4295                 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
4296         }
4297
4298         return (PF_PASS);
4299 }
4300
4301 int
4302 pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif,
4303     struct mbuf *m, int off, void *h, struct pf_pdesc *pd, u_short *reason)
4304 {
4305         struct pf_addr  *saddr = pd->src, *daddr = pd->dst;
4306         u_int16_t        icmpid = 0, *icmpsum;
4307         u_int8_t         icmptype;
4308         int              state_icmp = 0;
4309         struct pf_state_key_cmp key;
4310
4311         switch (pd->proto) {
4312 #ifdef INET
4313         case IPPROTO_ICMP:
4314                 icmptype = pd->hdr.icmp->icmp_type;
4315                 icmpid = pd->hdr.icmp->icmp_id;
4316                 icmpsum = &pd->hdr.icmp->icmp_cksum;
4317
4318                 if (icmptype == ICMP_UNREACH ||
4319                     icmptype == ICMP_SOURCEQUENCH ||
4320                     icmptype == ICMP_REDIRECT ||
4321                     icmptype == ICMP_TIMXCEED ||
4322                     icmptype == ICMP_PARAMPROB)
4323                         state_icmp++;
4324                 break;
4325 #endif /* INET */
4326 #ifdef INET6
4327         case IPPROTO_ICMPV6:
4328                 icmptype = pd->hdr.icmp6->icmp6_type;
4329                 icmpid = pd->hdr.icmp6->icmp6_id;
4330                 icmpsum = &pd->hdr.icmp6->icmp6_cksum;
4331
4332                 if (icmptype == ICMP6_DST_UNREACH ||
4333                     icmptype == ICMP6_PACKET_TOO_BIG ||
4334                     icmptype == ICMP6_TIME_EXCEEDED ||
4335                     icmptype == ICMP6_PARAM_PROB)
4336                         state_icmp++;
4337                 break;
4338 #endif /* INET6 */
4339         }
4340
4341         if (!state_icmp) {
4342
4343                 /*
4344                  * ICMP query/reply message not related to a TCP/UDP packet.
4345                  * Search for an ICMP state.
4346                  */
4347                 key.af = pd->af;
4348                 key.proto = pd->proto;
4349                 if (direction == PF_IN) {
4350                         PF_ACPY(&key.ext.addr, pd->src, key.af);
4351                         PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4352                         key.ext.port = 0;
4353                         key.gwy.port = icmpid;
4354                 } else {
4355                         PF_ACPY(&key.lan.addr, pd->src, key.af);
4356                         PF_ACPY(&key.ext.addr, pd->dst, key.af);
4357                         key.lan.port = icmpid;
4358                         key.ext.port = 0;
4359                 }
4360
4361                 STATE_LOOKUP();
4362
4363                 (*state)->expire = time_second;
4364                 (*state)->timeout = PFTM_ICMP_ERROR_REPLY;
4365
4366                 /* translate source/destination address, if necessary */
4367                 if (STATE_TRANSLATE((*state)->state_key)) {
4368                         if (direction == PF_OUT) {
4369                                 switch (pd->af) {
4370 #ifdef INET
4371                                 case AF_INET:
4372                                         pf_change_a(&saddr->v4.s_addr,
4373                                             pd->ip_sum,
4374                                             (*state)->state_key->gwy.addr.v4.s_addr, 0);
4375                                         pd->hdr.icmp->icmp_cksum =
4376                                             pf_cksum_fixup(
4377                                             pd->hdr.icmp->icmp_cksum, icmpid,
4378                                             (*state)->state_key->gwy.port, 0);
4379                                         pd->hdr.icmp->icmp_id =
4380                                             (*state)->state_key->gwy.port;
4381                                         m_copyback(m, off, ICMP_MINLEN,
4382                                             (caddr_t)pd->hdr.icmp);
4383                                         break;
4384 #endif /* INET */
4385 #ifdef INET6
4386                                 case AF_INET6:
4387                                         pf_change_a6(saddr,
4388                                             &pd->hdr.icmp6->icmp6_cksum,
4389                                             &(*state)->state_key->gwy.addr, 0);
4390                                         m_copyback(m, off,
4391                                             sizeof(struct icmp6_hdr),
4392                                             (caddr_t)pd->hdr.icmp6);
4393                                         break;
4394 #endif /* INET6 */
4395                                 }
4396                         } else {
4397                                 switch (pd->af) {
4398 #ifdef INET
4399                                 case AF_INET:
4400                                         pf_change_a(&daddr->v4.s_addr,
4401                                             pd->ip_sum,
4402                                             (*state)->state_key->lan.addr.v4.s_addr, 0);
4403                                         pd->hdr.icmp->icmp_cksum =
4404                                             pf_cksum_fixup(
4405                                             pd->hdr.icmp->icmp_cksum, icmpid,
4406                                             (*state)->state_key->lan.port, 0);
4407                                         pd->hdr.icmp->icmp_id =
4408                                             (*state)->state_key->lan.port;
4409                                         m_copyback(m, off, ICMP_MINLEN,
4410                                             (caddr_t)pd->hdr.icmp);
4411                                         break;
4412 #endif /* INET */
4413 #ifdef INET6
4414                                 case AF_INET6:
4415                                         pf_change_a6(daddr,
4416                                             &pd->hdr.icmp6->icmp6_cksum,
4417                                             &(*state)->state_key->lan.addr, 0);
4418                                         m_copyback(m, off,
4419                                             sizeof(struct icmp6_hdr),
4420                                             (caddr_t)pd->hdr.icmp6);
4421                                         break;
4422 #endif /* INET6 */
4423                                 }
4424                         }
4425                 }
4426
4427                 return (PF_PASS);
4428
4429         } else {
4430                 /*
4431                  * ICMP error message in response to a TCP/UDP packet.
4432                  * Extract the inner TCP/UDP header and search for that state.
4433                  */
4434
4435                 struct pf_pdesc pd2;
4436 #ifdef INET
4437                 struct ip       h2;
4438 #endif /* INET */
4439 #ifdef INET6
4440                 struct ip6_hdr  h2_6;
4441                 int             terminal = 0;
4442 #endif /* INET6 */
4443                 int             ipoff2;
4444                 int             off2;
4445
4446                 pd2.af = pd->af;
4447                 switch (pd->af) {
4448 #ifdef INET
4449                 case AF_INET:
4450                         /* offset of h2 in mbuf chain */
4451                         ipoff2 = off + ICMP_MINLEN;
4452
4453                         if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2),
4454                             NULL, reason, pd2.af)) {
4455                                 DPFPRINTF(PF_DEBUG_MISC,
4456                                     ("pf: ICMP error message too short "
4457                                     "(ip)\n"));
4458                                 return (PF_DROP);
4459                         }
4460                         /*
4461                          * ICMP error messages don't refer to non-first
4462                          * fragments
4463                          */
4464                         if (h2.ip_off & htons(IP_OFFMASK)) {
4465                                 REASON_SET(reason, PFRES_FRAG);
4466                                 return (PF_DROP);
4467                         }
4468
4469                         /* offset of protocol header that follows h2 */
4470                         off2 = ipoff2 + (h2.ip_hl << 2);
4471
4472                         pd2.proto = h2.ip_p;
4473                         pd2.src = (struct pf_addr *)&h2.ip_src;
4474                         pd2.dst = (struct pf_addr *)&h2.ip_dst;
4475                         pd2.ip_sum = &h2.ip_sum;
4476                         break;
4477 #endif /* INET */
4478 #ifdef INET6
4479                 case AF_INET6:
4480                         ipoff2 = off + sizeof(struct icmp6_hdr);
4481
4482                         if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6),
4483                             NULL, reason, pd2.af)) {
4484                                 DPFPRINTF(PF_DEBUG_MISC,
4485                                     ("pf: ICMP error message too short "
4486                                     "(ip6)\n"));
4487                                 return (PF_DROP);
4488                         }
4489                         pd2.proto = h2_6.ip6_nxt;
4490                         pd2.src = (struct pf_addr *)&h2_6.ip6_src;
4491                         pd2.dst = (struct pf_addr *)&h2_6.ip6_dst;
4492                         pd2.ip_sum = NULL;
4493                         off2 = ipoff2 + sizeof(h2_6);
4494                         do {
4495                                 switch (pd2.proto) {
4496                                 case IPPROTO_FRAGMENT:
4497                                         /*
4498                                          * ICMPv6 error messages for
4499                                          * non-first fragments
4500                                          */
4501                                         REASON_SET(reason, PFRES_FRAG);
4502                                         return (PF_DROP);
4503                                 case IPPROTO_AH:
4504                                 case IPPROTO_HOPOPTS:
4505                                 case IPPROTO_ROUTING:
4506                                 case IPPROTO_DSTOPTS: {
4507                                         /* get next header and header length */
4508                                         struct ip6_ext opt6;
4509
4510                                         if (!pf_pull_hdr(m, off2, &opt6,
4511                                             sizeof(opt6), NULL, reason,
4512                                             pd2.af)) {
4513                                                 DPFPRINTF(PF_DEBUG_MISC,
4514                                                     ("pf: ICMPv6 short opt\n"));
4515                                                 return (PF_DROP);
4516                                         }
4517                                         if (pd2.proto == IPPROTO_AH)
4518                                                 off2 += (opt6.ip6e_len + 2) * 4;
4519                                         else
4520                                                 off2 += (opt6.ip6e_len + 1) * 8;
4521                                         pd2.proto = opt6.ip6e_nxt;
4522                                         /* goto the next header */
4523                                         break;
4524                                 }
4525                                 default:
4526                                         terminal++;
4527                                         break;
4528                                 }
4529                         } while (!terminal);
4530                         break;
4531 #endif /* INET6 */
4532                 default:
4533                         DPFPRINTF(PF_DEBUG_MISC,
4534                             ("pf: ICMP AF %d unknown (ip6)\n", pd->af));
4535                         return (PF_DROP);
4536                         break;
4537                 }
4538
4539                 switch (pd2.proto) {
4540                 case IPPROTO_TCP: {
4541                         struct tcphdr            th;
4542                         u_int32_t                seq;
4543                         struct pf_state_peer    *src, *dst;
4544                         u_int8_t                 dws;
4545                         int                      copyback = 0;
4546
4547                         /*
4548                          * Only the first 8 bytes of the TCP header can be
4549                          * expected. Don't access any TCP header fields after
4550                          * th_seq, an ackskew test is not possible.
4551                          */
4552                         if (!pf_pull_hdr(m, off2, &th, 8, NULL, reason,
4553                             pd2.af)) {
4554                                 DPFPRINTF(PF_DEBUG_MISC,
4555                                     ("pf: ICMP error message too short "
4556                                     "(tcp)\n"));
4557                                 return (PF_DROP);
4558                         }
4559
4560                         key.af = pd2.af;
4561                         key.proto = IPPROTO_TCP;
4562                         if (direction == PF_IN) {
4563                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4564                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4565                                 key.ext.port = th.th_dport;
4566                                 key.gwy.port = th.th_sport;
4567                         } else {
4568                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4569                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4570                                 key.lan.port = th.th_dport;
4571                                 key.ext.port = th.th_sport;
4572                         }
4573
4574                         STATE_LOOKUP();
4575
4576                         if (direction == (*state)->state_key->direction) {
4577                                 src = &(*state)->dst;
4578                                 dst = &(*state)->src;
4579                         } else {
4580                                 src = &(*state)->src;
4581                                 dst = &(*state)->dst;
4582                         }
4583
4584                         if (src->wscale && dst->wscale)
4585                                 dws = dst->wscale & PF_WSCALE_MASK;
4586                         else
4587                                 dws = 0;
4588
4589                         /* Demodulate sequence number */
4590                         seq = ntohl(th.th_seq) - src->seqdiff;
4591                         if (src->seqdiff) {
4592                                 pf_change_a(&th.th_seq, icmpsum,
4593                                     htonl(seq), 0);
4594                                 copyback = 1;
4595                         }
4596
4597                         if (!SEQ_GEQ(src->seqhi, seq) ||
4598                             !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws))) {
4599                                 if (pf_status.debug >= PF_DEBUG_MISC) {
4600                                         kprintf("pf: BAD ICMP %d:%d ",
4601                                             icmptype, pd->hdr.icmp->icmp_code);
4602                                         pf_print_host(pd->src, 0, pd->af);
4603                                         kprintf(" -> ");
4604                                         pf_print_host(pd->dst, 0, pd->af);
4605                                         kprintf(" state: ");
4606                                         pf_print_state(*state);
4607                                         kprintf(" seq=%u\n", seq);
4608                                 }
4609                                 REASON_SET(reason, PFRES_BADSTATE);
4610                                 return (PF_DROP);
4611                         }
4612
4613                         if (STATE_TRANSLATE((*state)->state_key)) {
4614                                 if (direction == PF_IN) {
4615                                         pf_change_icmp(pd2.src, &th.th_sport,
4616                                             daddr, &(*state)->state_key->lan.addr,
4617                                             (*state)->state_key->lan.port, NULL,
4618                                             pd2.ip_sum, icmpsum,
4619                                             pd->ip_sum, 0, pd2.af);
4620                                 } else {
4621                                         pf_change_icmp(pd2.dst, &th.th_dport,
4622                                             saddr, &(*state)->state_key->gwy.addr,
4623                                             (*state)->state_key->gwy.port, NULL,
4624                                             pd2.ip_sum, icmpsum,
4625                                             pd->ip_sum, 0, pd2.af);
4626                                 }
4627                                 copyback = 1;
4628                         }
4629
4630                         if (copyback) {
4631                                 switch (pd2.af) {
4632 #ifdef INET
4633                                 case AF_INET:
4634                                         m_copyback(m, off, ICMP_MINLEN,
4635                                             (caddr_t)pd->hdr.icmp);
4636                                         m_copyback(m, ipoff2, sizeof(h2),
4637                                             (caddr_t)&h2);
4638                                         break;
4639 #endif /* INET */
4640 #ifdef INET6
4641                                 case AF_INET6:
4642                                         m_copyback(m, off,
4643                                             sizeof(struct icmp6_hdr),
4644                                             (caddr_t)pd->hdr.icmp6);
4645                                         m_copyback(m, ipoff2, sizeof(h2_6),
4646                                             (caddr_t)&h2_6);
4647                                         break;
4648 #endif /* INET6 */
4649                                 }
4650                                 m_copyback(m, off2, 8, (caddr_t)&th);
4651                         }
4652
4653                         return (PF_PASS);
4654                         break;
4655                 }
4656                 case IPPROTO_UDP: {
4657                         struct udphdr           uh;
4658
4659                         if (!pf_pull_hdr(m, off2, &uh, sizeof(uh),
4660                             NULL, reason, pd2.af)) {
4661                                 DPFPRINTF(PF_DEBUG_MISC,
4662                                     ("pf: ICMP error message too short "
4663                                     "(udp)\n"));
4664                                 return (PF_DROP);
4665                         }
4666
4667                         key.af = pd2.af;
4668                         key.proto = IPPROTO_UDP;
4669                         if (direction == PF_IN) {
4670                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4671                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4672                                 key.ext.port = uh.uh_dport;
4673                                 key.gwy.port = uh.uh_sport;
4674                         } else {
4675                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4676                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4677                                 key.lan.port = uh.uh_dport;
4678                                 key.ext.port = uh.uh_sport;
4679                         }
4680
4681                         STATE_LOOKUP();
4682
4683                         if (STATE_TRANSLATE((*state)->state_key)) {
4684                                 if (direction == PF_IN) {
4685                                         pf_change_icmp(pd2.src, &uh.uh_sport,
4686                                             daddr,
4687                                             &(*state)->state_key->lan.addr,
4688                                             (*state)->state_key->lan.port,
4689                                             &uh.uh_sum,
4690                                             pd2.ip_sum, icmpsum,
4691                                             pd->ip_sum, 1, pd2.af);
4692                                 } else {
4693                                         pf_change_icmp(pd2.dst, &uh.uh_dport,
4694                                             saddr,
4695                                             &(*state)->state_key->gwy.addr,
4696                                             (*state)->state_key->gwy.port, &uh.uh_sum,
4697                                             pd2.ip_sum, icmpsum,
4698                                             pd->ip_sum, 1, pd2.af);
4699                                 }
4700                                 switch (pd2.af) {
4701 #ifdef INET
4702                                 case AF_INET:
4703                                         m_copyback(m, off, ICMP_MINLEN,
4704                                             (caddr_t)pd->hdr.icmp);
4705                                         m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4706                                         break;
4707 #endif /* INET */
4708 #ifdef INET6
4709                                 case AF_INET6:
4710                                         m_copyback(m, off,
4711                                             sizeof(struct icmp6_hdr),
4712                                             (caddr_t)pd->hdr.icmp6);
4713                                         m_copyback(m, ipoff2, sizeof(h2_6),
4714                                             (caddr_t)&h2_6);
4715                                         break;
4716 #endif /* INET6 */
4717                                 }
4718                                 m_copyback(m, off2, sizeof(uh), (caddr_t)&uh);
4719                         }
4720
4721                         return (PF_PASS);
4722                         break;
4723                 }
4724 #ifdef INET
4725                 case IPPROTO_ICMP: {
4726                         struct icmp             iih;
4727
4728                         if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN,
4729                             NULL, reason, pd2.af)) {
4730                                 DPFPRINTF(PF_DEBUG_MISC,
4731                                     ("pf: ICMP error message too short i"
4732                                     "(icmp)\n"));
4733                                 return (PF_DROP);
4734                         }
4735
4736                         key.af = pd2.af;
4737                         key.proto = IPPROTO_ICMP;
4738                         if (direction == PF_IN) {
4739                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4740                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4741                                 key.ext.port = 0;
4742                                 key.gwy.port = iih.icmp_id;
4743                         } else {
4744                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4745                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4746                                 key.lan.port = iih.icmp_id;
4747                                 key.ext.port = 0;
4748                         }
4749
4750                         STATE_LOOKUP();
4751
4752                         if (STATE_TRANSLATE((*state)->state_key)) {
4753                                 if (direction == PF_IN) {
4754                                         pf_change_icmp(pd2.src, &iih.icmp_id,
4755                                             daddr,
4756                                             &(*state)->state_key->lan.addr,
4757                                             (*state)->state_key->lan.port, NULL,
4758                                             pd2.ip_sum, icmpsum,
4759                                             pd->ip_sum, 0, AF_INET);
4760                                 } else {
4761                                         pf_change_icmp(pd2.dst, &iih.icmp_id,
4762                                             saddr,
4763                                             &(*state)->state_key->gwy.addr,
4764                                             (*state)->state_key->gwy.port, NULL,
4765                                             pd2.ip_sum, icmpsum,
4766                                             pd->ip_sum, 0, AF_INET);
4767                                 }
4768                                 m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
4769                                 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4770                                 m_copyback(m, off2, ICMP_MINLEN, (caddr_t)&iih);
4771                         }
4772
4773                         return (PF_PASS);
4774                         break;
4775                 }
4776 #endif /* INET */
4777 #ifdef INET6
4778                 case IPPROTO_ICMPV6: {
4779                         struct icmp6_hdr        iih;
4780
4781                         if (!pf_pull_hdr(m, off2, &iih,
4782                             sizeof(struct icmp6_hdr), NULL, reason, pd2.af)) {
4783                                 DPFPRINTF(PF_DEBUG_MISC,
4784                                     ("pf: ICMP error message too short "
4785                                     "(icmp6)\n"));
4786                                 return (PF_DROP);
4787                         }
4788
4789                         key.af = pd2.af;
4790                         key.proto = IPPROTO_ICMPV6;
4791                         if (direction == PF_IN) {
4792                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4793                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4794                                 key.ext.port = 0;
4795                                 key.gwy.port = iih.icmp6_id;
4796                         } else {
4797                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4798                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4799                                 key.lan.port = iih.icmp6_id;
4800                                 key.ext.port = 0;
4801                         }
4802
4803                         STATE_LOOKUP();
4804
4805                         if (STATE_TRANSLATE((*state)->state_key)) {
4806                                 if (direction == PF_IN) {
4807                                         pf_change_icmp(pd2.src, &iih.icmp6_id,
4808                                             daddr,
4809                                             &(*state)->state_key->lan.addr,
4810                                             (*state)->state_key->lan.port, NULL,
4811                                             pd2.ip_sum, icmpsum,
4812                                             pd->ip_sum, 0, AF_INET6);
4813                                 } else {
4814                                         pf_change_icmp(pd2.dst, &iih.icmp6_id,
4815                                             saddr, &(*state)->state_key->gwy.addr,
4816                                             (*state)->state_key->gwy.port, NULL,
4817                                             pd2.ip_sum, icmpsum,
4818                                             pd->ip_sum, 0, AF_INET6);
4819                                 }
4820                                 m_copyback(m, off, sizeof(struct icmp6_hdr),
4821                                     (caddr_t)pd->hdr.icmp6);
4822                                 m_copyback(m, ipoff2, sizeof(h2_6), (caddr_t)&h2_6);
4823                                 m_copyback(m, off2, sizeof(struct icmp6_hdr),
4824                                     (caddr_t)&iih);
4825                         }
4826
4827                         return (PF_PASS);
4828                         break;
4829                 }
4830 #endif /* INET6 */
4831                 default: {
4832                         key.af = pd2.af;
4833                         key.proto = pd2.proto;
4834                         if (direction == PF_IN) {
4835                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4836                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4837                                 key.ext.port = 0;
4838                                 key.gwy.port = 0;
4839                         } else {
4840                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4841                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4842                                 key.lan.port = 0;
4843                                 key.ext.port = 0;
4844                         }
4845
4846                         STATE_LOOKUP();
4847
4848                         if (STATE_TRANSLATE((*state)->state_key)) {
4849                                 if (direction == PF_IN) {
4850                                         pf_change_icmp(pd2.src, NULL,
4851                                             daddr,
4852                                             &(*state)->state_key->lan.addr,
4853                                             0, NULL,
4854                                             pd2.ip_sum, icmpsum,
4855                                             pd->ip_sum, 0, pd2.af);
4856                                 } else {
4857                                         pf_change_icmp(pd2.dst, NULL,
4858                                             saddr,
4859                                             &(*state)->state_key->gwy.addr,
4860                                             0, NULL,
4861                                             pd2.ip_sum, icmpsum,
4862                                             pd->ip_sum, 0, pd2.af);
4863                                 }
4864                                 switch (pd2.af) {
4865 #ifdef INET
4866                                 case AF_INET:
4867                                         m_copyback(m, off, ICMP_MINLEN,
4868                                             (caddr_t)pd->hdr.icmp);
4869                                         m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4870                                         break;
4871 #endif /* INET */
4872 #ifdef INET6
4873                                 case AF_INET6:
4874                                         m_copyback(m, off,
4875                                             sizeof(struct icmp6_hdr),
4876                                             (caddr_t)pd->hdr.icmp6);
4877                                         m_copyback(m, ipoff2, sizeof(h2_6),
4878                                             (caddr_t)&h2_6);
4879                                         break;
4880 #endif /* INET6 */
4881                                 }
4882                         }
4883
4884                         return (PF_PASS);
4885                         break;
4886                 }
4887                 }
4888         }
4889 }
4890
4891 int
4892 pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif,
4893     struct pf_pdesc *pd)
4894 {
4895         struct pf_state_peer    *src, *dst;
4896         struct pf_state_key_cmp  key;
4897
4898         key.af = pd->af;
4899         key.proto = pd->proto;
4900         if (direction == PF_IN) {
4901                 PF_ACPY(&key.ext.addr, pd->src, key.af);
4902                 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4903                 key.ext.port = 0;
4904                 key.gwy.port = 0;
4905         } else {
4906                 PF_ACPY(&key.lan.addr, pd->src, key.af);
4907                 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4908                 key.lan.port = 0;
4909                 key.ext.port = 0;
4910         }
4911
4912         STATE_LOOKUP();
4913
4914         if (direction == (*state)->state_key->direction) {
4915                 src = &(*state)->src;
4916                 dst = &(*state)->dst;
4917         } else {
4918                 src = &(*state)->dst;
4919                 dst = &(*state)->src;
4920         }
4921
4922         /* update states */
4923         if (src->state < PFOTHERS_SINGLE)
4924                 src->state = PFOTHERS_SINGLE;
4925         if (dst->state == PFOTHERS_SINGLE)
4926                 dst->state = PFOTHERS_MULTIPLE;
4927
4928         /* update expire time */
4929         (*state)->expire = time_second;
4930         if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE)
4931                 (*state)->timeout = PFTM_OTHER_MULTIPLE;
4932         else
4933                 (*state)->timeout = PFTM_OTHER_SINGLE;
4934
4935         /* translate source/destination address, if necessary */
4936         if (STATE_TRANSLATE((*state)->state_key)) {
4937                 if (direction == PF_OUT)
4938                         switch (pd->af) {
4939 #ifdef INET
4940                         case AF_INET:
4941                                 pf_change_a(&pd->src->v4.s_addr,
4942                                     pd->ip_sum,
4943                                     (*state)->state_key->gwy.addr.v4.s_addr,
4944                                     0);
4945                                 break;
4946 #endif /* INET */
4947 #ifdef INET6
4948                         case AF_INET6:
4949                                 PF_ACPY(pd->src,
4950                                     &(*state)->state_key->gwy.addr, pd->af);
4951                                 break;
4952 #endif /* INET6 */
4953                         }
4954                 else
4955                         switch (pd->af) {
4956 #ifdef INET
4957                         case AF_INET:
4958                                 pf_change_a(&pd->dst->v4.s_addr,
4959                                     pd->ip_sum,
4960                                     (*state)->state_key->lan.addr.v4.s_addr,
4961                                     0);
4962                                 break;
4963 #endif /* INET */
4964 #ifdef INET6
4965                         case AF_INET6:
4966                                 PF_ACPY(pd->dst,
4967                                     &(*state)->state_key->lan.addr, pd->af);
4968                                 break;
4969 #endif /* INET6 */
4970                         }
4971         }
4972
4973         return (PF_PASS);
4974 }
4975
4976 /*
4977  * ipoff and off are measured from the start of the mbuf chain.
4978  * h must be at "ipoff" on the mbuf chain.
4979  */
4980 void *
4981 pf_pull_hdr(struct mbuf *m, int off, void *p, int len,
4982     u_short *actionp, u_short *reasonp, sa_family_t af)
4983 {
4984         switch (af) {
4985 #ifdef INET
4986         case AF_INET: {
4987                 struct ip       *h = mtod(m, struct ip *);
4988                 u_int16_t        fragoff = (h->ip_off & IP_OFFMASK) << 3;
4989
4990                 if (fragoff) {
4991                         if (fragoff >= len)
4992                                 ACTION_SET(actionp, PF_PASS);
4993                         else {
4994                                 ACTION_SET(actionp, PF_DROP);
4995                                 REASON_SET(reasonp, PFRES_FRAG);
4996                         }
4997                         return (NULL);
4998                 }
4999                 if (m->m_pkthdr.len < off + len ||
5000                     h->ip_len < off + len) {
5001                         ACTION_SET(actionp, PF_DROP);
5002                         REASON_SET(reasonp, PFRES_SHORT);
5003                         return (NULL);
5004                 }
5005                 break;
5006         }
5007 #endif /* INET */
5008 #ifdef INET6
5009         case AF_INET6: {
5010                 struct ip6_hdr  *h = mtod(m, struct ip6_hdr *);
5011
5012                 if (m->m_pkthdr.len < off + len ||
5013                     (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) <
5014                     (unsigned)(off + len)) {
5015                         ACTION_SET(actionp, PF_DROP);
5016                         REASON_SET(reasonp, PFRES_SHORT);
5017                         return (NULL);
5018                 }
5019                 break;
5020         }
5021 #endif /* INET6 */
5022         }
5023         m_copydata(m, off, len, p);
5024         return (p);
5025 }
5026
5027 int
5028 pf_routable(struct pf_addr *addr, sa_family_t af, struct pfi_kif *kif)
5029 {
5030         struct sockaddr_in      *dst;
5031         int                      ret = 1;
5032         int                      check_mpath;
5033 #ifdef INET6
5034         struct sockaddr_in6     *dst6;
5035         struct route_in6         ro;
5036 #else
5037         struct route             ro;
5038 #endif
5039         struct radix_node       *rn;
5040         struct rtentry          *rt;
5041         struct ifnet            *ifp;
5042
5043         check_mpath = 0;
5044         bzero(&ro, sizeof(ro));
5045         switch (af) {
5046         case AF_INET:
5047                 dst = satosin(&ro.ro_dst);
5048                 dst->sin_family = AF_INET;
5049                 dst->sin_len = sizeof(*dst);
5050                 dst->sin_addr = addr->v4;
5051                 break;
5052 #ifdef INET6
5053         case AF_INET6:
5054                 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
5055                 dst6->sin6_family = AF_INET6;
5056                 dst6->sin6_len = sizeof(*dst6);
5057                 dst6->sin6_addr = addr->v6;
5058                 break;
5059 #endif /* INET6 */
5060         default:
5061                 return (0);
5062         }
5063
5064         /* Skip checks for ipsec interfaces */
5065         if (kif != NULL && kif->pfik_ifp->if_type == IFT_ENC)
5066                 goto out;
5067
5068         rtalloc_ign((struct route *)&ro, 0);
5069
5070         if (ro.ro_rt != NULL) {
5071                 /* No interface given, this is a no-route check */
5072                 if (kif == NULL)
5073                         goto out;
5074
5075                 if (kif->pfik_ifp == NULL) {
5076                         ret = 0;
5077                         goto out;
5078                 }
5079
5080                 /* Perform uRPF check if passed input interface */
5081                 ret = 0;
5082                 rn = (struct radix_node *)ro.ro_rt;
5083                 do {
5084                         rt = (struct rtentry *)rn;
5085                         ifp = rt->rt_ifp;
5086
5087                         if (kif->pfik_ifp == ifp)
5088                                 ret = 1;
5089                         rn = NULL;
5090                 } while (check_mpath == 1 && rn != NULL && ret == 0);
5091         } else
5092                 ret = 0;
5093 out:
5094         if (ro.ro_rt != NULL)
5095                 RTFREE(ro.ro_rt);
5096         return (ret);
5097 }
5098
5099 int
5100 pf_rtlabel_match(struct pf_addr *addr, sa_family_t af, struct pf_addr_wrap *aw)
5101 {
5102         struct sockaddr_in      *dst;
5103 #ifdef INET6
5104         struct sockaddr_in6     *dst6;
5105         struct route_in6         ro;
5106 #else
5107         struct route             ro;
5108 #endif
5109         int                      ret = 0;
5110
5111         ASSERT_LWKT_TOKEN_HELD(&pf_token);
5112
5113         bzero(&ro, sizeof(ro));
5114         switch (af) {
5115         case AF_INET:
5116                 dst = satosin(&ro.ro_dst);
5117                 dst->sin_family = AF_INET;
5118                 dst->sin_len = sizeof(*dst);
5119                 dst->sin_addr = addr->v4;
5120                 break;
5121 #ifdef INET6
5122         case AF_INET6:
5123                 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
5124                 dst6->sin6_family = AF_INET6;
5125                 dst6->sin6_len = sizeof(*dst6);
5126                 dst6->sin6_addr = addr->v6;
5127                 break;
5128 #endif /* INET6 */
5129         default:
5130                 return (0);
5131         }
5132
5133 rtalloc_ign((struct route *)&ro, (RTF_CLONING | RTF_PRCLONING));
5134
5135         if (ro.ro_rt != NULL) {
5136                 RTFREE(ro.ro_rt);
5137         }
5138
5139         return (ret);
5140 }
5141
5142 #ifdef INET
5143 void
5144 pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5145     struct pf_state *s, struct pf_pdesc *pd)
5146 {
5147         struct mbuf             *m0, *m1;
5148         struct route             iproute;
5149         struct route            *ro = NULL;
5150         struct sockaddr_in      *dst;
5151         struct ip               *ip;
5152         struct ifnet            *ifp = NULL;
5153         struct pf_addr           naddr;
5154         struct pf_src_node      *sn = NULL;
5155         int                      error = 0;
5156         int sw_csum;
5157 #ifdef IPSEC
5158         struct m_tag            *mtag;
5159 #endif /* IPSEC */
5160
5161         ASSERT_LWKT_TOKEN_HELD(&pf_token);
5162
5163         if (m == NULL || *m == NULL || r == NULL ||
5164             (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5165                 panic("pf_route: invalid parameters");
5166
5167         if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
5168                 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
5169                 (*m)->m_pkthdr.pf.routed = 1;
5170         } else {
5171                 if ((*m)->m_pkthdr.pf.routed++ > 3) {
5172                         m0 = *m;
5173                         *m = NULL;
5174                         goto bad;
5175                 }
5176         }
5177
5178         if (r->rt == PF_DUPTO) {
5179                 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL) {
5180                         return;
5181                 }
5182         } else {
5183                 if ((r->rt == PF_REPLYTO) == (r->direction == dir)) {
5184                         return;
5185                 }
5186                 m0 = *m;
5187         }
5188
5189         if (m0->m_len < sizeof(struct ip)) {
5190                 DPFPRINTF(PF_DEBUG_URGENT,
5191                     ("pf_route: m0->m_len < sizeof(struct ip)\n"));
5192                 goto bad;
5193         }
5194
5195         ip = mtod(m0, struct ip *);
5196
5197         ro = &iproute;
5198         bzero((caddr_t)ro, sizeof(*ro));
5199         dst = satosin(&ro->ro_dst);
5200         dst->sin_family = AF_INET;
5201         dst->sin_len = sizeof(*dst);
5202         dst->sin_addr = ip->ip_dst;
5203
5204         if (r->rt == PF_FASTROUTE) {
5205                 rtalloc(ro);
5206                 if (ro->ro_rt == 0) {
5207                         ipstat.ips_noroute++;
5208                         goto bad;
5209                 }
5210
5211                 ifp = ro->ro_rt->rt_ifp;
5212                 ro->ro_rt->rt_use++;
5213
5214                 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
5215                         dst = satosin(ro->ro_rt->rt_gateway);
5216         } else {
5217                 if (TAILQ_EMPTY(&r->rpool.list)) {
5218                         DPFPRINTF(PF_DEBUG_URGENT,
5219                             ("pf_route: TAILQ_EMPTY(&r->rpool.list)\n"));
5220                         goto bad;
5221                 }
5222                 if (s == NULL) {
5223                         pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src,
5224                             &naddr, NULL, &sn);
5225                         if (!PF_AZERO(&naddr, AF_INET))
5226                                 dst->sin_addr.s_addr = naddr.v4.s_addr;
5227                         ifp = r->rpool.cur->kif ?
5228                             r->rpool.cur->kif->pfik_ifp : NULL;
5229                 } else {
5230                         if (!PF_AZERO(&s->rt_addr, AF_INET))
5231                                 dst->sin_addr.s_addr =
5232                                     s->rt_addr.v4.s_addr;
5233                         ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5234                 }
5235         }
5236         if (ifp == NULL)
5237                 goto bad;
5238
5239         if (oifp != ifp) {
5240                 if (pf_test(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) {
5241                         goto bad;
5242                 } else if (m0 == NULL) {
5243                         goto done;
5244                 }
5245                 if (m0->m_len < sizeof(struct ip)) {
5246                         DPFPRINTF(PF_DEBUG_URGENT,
5247                             ("pf_route: m0->m_len < sizeof(struct ip)\n"));
5248                         goto bad;
5249                 }
5250                 ip = mtod(m0, struct ip *);
5251         }
5252
5253         /* Copied from FreeBSD 5.1-CURRENT ip_output. */
5254         m0->m_pkthdr.csum_flags |= CSUM_IP;
5255         sw_csum = m0->m_pkthdr.csum_flags & ~ifp->if_hwassist;
5256         if (sw_csum & CSUM_DELAY_DATA) {
5257                 in_delayed_cksum(m0);
5258                 sw_csum &= ~CSUM_DELAY_DATA;
5259         }
5260         m0->m_pkthdr.csum_flags &= ifp->if_hwassist;
5261
5262         if (ip->ip_len <= ifp->if_mtu ||
5263             (ifp->if_hwassist & CSUM_FRAGMENT &&
5264                 (ip->ip_off & IP_DF) == 0)) {
5265                 ip->ip_len = htons(ip->ip_len);
5266                 ip->ip_off = htons(ip->ip_off);
5267                 ip->ip_sum = 0;
5268                 if (sw_csum & CSUM_DELAY_IP) {
5269                         /* From KAME */
5270                         if (ip->ip_v == IPVERSION &&
5271                             (ip->ip_hl << 2) == sizeof(*ip)) {
5272                                 ip->ip_sum = in_cksum_hdr(ip);
5273                         } else {
5274                                 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
5275                         }
5276                 }
5277                 lwkt_reltoken(&pf_token);
5278                 error = ifp->if_output(ifp, m0, sintosa(dst), ro->ro_rt);
5279                 lwkt_gettoken(&pf_token);
5280                 goto done;
5281         }
5282
5283         /*
5284          * Too large for interface; fragment if possible.
5285          * Must be able to put at least 8 bytes per fragment.
5286          */
5287         if (ip->ip_off & IP_DF) {
5288                 ipstat.ips_cantfrag++;
5289                 if (r->rt != PF_DUPTO) {
5290                         icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0,
5291                                    ifp->if_mtu);
5292                         goto done;
5293                 } else
5294                         goto bad;
5295         }
5296
5297         m1 = m0;
5298         error = ip_fragment(ip, &m0, ifp->if_mtu, ifp->if_hwassist, sw_csum);
5299         if (error) {
5300                 goto bad;
5301         }
5302
5303         for (m0 = m1; m0; m0 = m1) {
5304                 m1 = m0->m_nextpkt;
5305                 m0->m_nextpkt = 0;
5306                 if (error == 0) {
5307                         lwkt_reltoken(&pf_token);
5308                         error = (*ifp->if_output)(ifp, m0, sintosa(dst),
5309                                                   NULL);
5310                         lwkt_gettoken(&pf_token);
5311                 } else
5312                         m_freem(m0);
5313         }
5314
5315         if (error == 0)
5316                 ipstat.ips_fragmented++;
5317
5318 done:
5319         if (r->rt != PF_DUPTO)
5320                 *m = NULL;
5321         if (ro == &iproute && ro->ro_rt)
5322                 RTFREE(ro->ro_rt);
5323         return;
5324
5325 bad:
5326         m_freem(m0);
5327         goto done;
5328 }
5329 #endif /* INET */
5330
5331 #ifdef INET6
5332 void
5333 pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5334     struct pf_state *s, struct pf_pdesc *pd)
5335 {
5336         struct mbuf             *m0;
5337         struct route_in6         ip6route;
5338         struct route_in6        *ro;
5339         struct sockaddr_in6     *dst;
5340         struct ip6_hdr          *ip6;
5341         struct ifnet            *ifp = NULL;
5342         struct pf_addr           naddr;
5343         struct pf_src_node      *sn = NULL;
5344         int                      error = 0;
5345
5346         if (m == NULL || *m == NULL || r == NULL ||
5347             (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5348                 panic("pf_route6: invalid parameters");
5349
5350         if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
5351                 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
5352                 (*m)->m_pkthdr.pf.routed = 1;
5353         } else {
5354                 if ((*m)->m_pkthdr.pf.routed++ > 3) {
5355                         m0 = *m;
5356                         *m = NULL;
5357                         goto bad;
5358                 }
5359         }
5360
5361         if (r->rt == PF_DUPTO) {
5362                 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL)
5363                         return;
5364         } else {
5365                 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5366                         return;
5367                 m0 = *m;
5368         }
5369
5370         if (m0->m_len < sizeof(struct ip6_hdr)) {
5371                 DPFPRINTF(PF_DEBUG_URGENT,
5372                     ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n"));
5373                 goto bad;
5374         }
5375         ip6 = mtod(m0, struct ip6_hdr *);
5376
5377         ro = &ip6route;
5378         bzero((caddr_t)ro, sizeof(*ro));
5379         dst = (struct sockaddr_in6 *)&ro->ro_dst;
5380         dst->sin6_family = AF_INET6;
5381         dst->sin6_len = sizeof(*dst);
5382         dst->sin6_addr = ip6->ip6_dst;
5383
5384         /* Cheat. XXX why only in the v6 case??? */
5385         if (r->rt == PF_FASTROUTE) {
5386                 m0->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
5387                 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
5388                 crit_enter();
5389                 return;
5390         }
5391
5392         if (TAILQ_EMPTY(&r->rpool.list)) {
5393                 DPFPRINTF(PF_DEBUG_URGENT,
5394                     ("pf_route6: TAILQ_EMPTY(&r->rpool.list)\n"));
5395                 goto bad;
5396         }
5397         if (s == NULL) {
5398                 pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src,
5399                     &naddr, NULL, &sn);
5400                 if (!PF_AZERO(&naddr, AF_INET6))
5401                         PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5402                             &naddr, AF_INET6);
5403                 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL;
5404         } else {
5405                 if (!PF_AZERO(&s->rt_addr, AF_INET6))
5406                         PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5407                             &s->rt_addr, AF_INET6);
5408                 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5409         }
5410         if (ifp == NULL)
5411                 goto bad;
5412
5413         if (oifp != ifp) {
5414                 if (pf_test6(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) {
5415                         goto bad;
5416                 } else if (m0 == NULL) {
5417                         goto done;
5418                 }
5419                 if (m0->m_len < sizeof(struct ip6_hdr)) {
5420                         DPFPRINTF(PF_DEBUG_URGENT,
5421                             ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n"));
5422                         goto bad;
5423                 }
5424                 ip6 = mtod(m0, struct ip6_hdr *);
5425         }
5426
5427         /*
5428          * If the packet is too large for the outgoing interface,
5429          * send back an icmp6 error.
5430          */
5431         if (IN6_IS_ADDR_LINKLOCAL(&dst->sin6_addr))
5432                 dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
5433         if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) {
5434                 error = nd6_output(ifp, ifp, m0, dst, NULL);
5435         } else {
5436                 in6_ifstat_inc(ifp, ifs6_in_toobig);
5437                 if (r->rt != PF_DUPTO) {
5438                         icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
5439                  } else
5440                         goto bad;
5441         }
5442
5443 done:
5444         if (r->rt != PF_DUPTO)
5445                 *m = NULL;
5446         return;
5447
5448 bad:
5449         m_freem(m0);
5450         goto done;
5451 }
5452
5453 #endif /* INET6 */
5454
5455
5456 /*
5457  * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag
5458  *   off is the offset where the protocol header starts
5459  *   len is the total length of protocol header plus payload
5460  * returns 0 when the checksum is valid, otherwise returns 1.
5461  */
5462 /*
5463  * XXX
5464  * FreeBSD supports cksum offload for the following drivers.
5465  * em(4), gx(4), lge(4), nge(4), ti(4), xl(4)
5466  * If we can make full use of it we would outperform ipfw/ipfilter in
5467  * very heavy traffic.
5468  * I have not tested 'cause I don't have NICs that supports cksum offload.
5469  * (There might be problems. Typical phenomena would be
5470  *   1. No route message for UDP packet.
5471  *   2. No connection acceptance from external hosts regardless of rule set.)
5472  */
5473 int
5474 pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p,
5475     sa_family_t af)
5476 {
5477         u_int16_t sum = 0;
5478         int hw_assist = 0;
5479         struct ip *ip;
5480
5481         if (off < sizeof(struct ip) || len < sizeof(struct udphdr))
5482                 return (1);
5483         if (m->m_pkthdr.len < off + len)
5484                 return (1);
5485
5486         switch (p) {
5487         case IPPROTO_TCP:
5488         case IPPROTO_UDP:
5489                 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
5490                         if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) {
5491                                 sum = m->m_pkthdr.csum_data;
5492                         } else {
5493                                 ip = mtod(m, struct ip *);
5494                                 sum = in_pseudo(ip->ip_src.s_addr,
5495                                         ip->ip_dst.s_addr, htonl((u_short)len +
5496                                         m->m_pkthdr.csum_data + p));
5497                         }
5498                         sum ^= 0xffff;
5499                         ++hw_assist;
5500                 }
5501                 break;
5502         case IPPROTO_ICMP:
5503 #ifdef INET6
5504         case IPPROTO_ICMPV6:
5505 #endif /* INET6 */
5506                 break;
5507         default:
5508                 return (1);
5509         }
5510
5511         if (!hw_assist) {
5512                 switch (af) {
5513                 case AF_INET:
5514                         if (p == IPPROTO_ICMP) {
5515                                 if (m->m_len < off)
5516                                         return (1);
5517                                 m->m_data += off;
5518                                 m->m_len -= off;
5519                                 sum = in_cksum(m, len);
5520                                 m->m_data -= off;
5521                                 m->m_len += off;
5522                         } else {
5523                                 if (m->m_len < sizeof(struct ip))
5524                                         return (1);
5525                                 sum = in_cksum_range(m, p, off, len);
5526                                 if (sum == 0) {
5527                                         m->m_pkthdr.csum_flags |=
5528                                             (CSUM_DATA_VALID |
5529                                              CSUM_PSEUDO_HDR);
5530                                         m->m_pkthdr.csum_data = 0xffff;
5531                                 }
5532                         }
5533                         break;
5534 #ifdef INET6
5535                 case AF_INET6:
5536                         if (m->m_len < sizeof(struct ip6_hdr))
5537                                 return (1);
5538                         sum = in6_cksum(m, p, off, len);
5539                         /*
5540                          * XXX
5541                          * IPv6 H/W cksum off-load not supported yet!
5542                          *
5543                          * if (sum == 0) {
5544                          *      m->m_pkthdr.csum_flags |=
5545                          *          (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
5546                          *      m->m_pkthdr.csum_data = 0xffff;
5547                          *}
5548                          */
5549                         break;
5550 #endif /* INET6 */
5551                 default:
5552                         return (1);
5553                 }
5554         }
5555         if (sum) {
5556                 switch (p) {
5557                 case IPPROTO_TCP:
5558                         tcpstat.tcps_rcvbadsum++;
5559                         break;
5560                 case IPPROTO_UDP:
5561                         udpstat.udps_badsum++;
5562                         break;
5563                 case IPPROTO_ICMP:
5564                         icmpstat.icps_checksum++;
5565                         break;
5566 #ifdef INET6
5567                 case IPPROTO_ICMPV6:
5568                         icmp6stat.icp6s_checksum++;
5569                         break;
5570 #endif /* INET6 */
5571                 }
5572                 return (1);
5573         }
5574         return (0);
5575 }
5576
5577 #ifdef INET
5578 int
5579 pf_test(int dir, struct ifnet *ifp, struct mbuf **m0,
5580     struct ether_header *eh, struct inpcb *inp)
5581 {
5582         struct pfi_kif          *kif;
5583         u_short                  action, reason = 0, log = 0;
5584         struct mbuf             *m = *m0;
5585         struct ip               *h = NULL;
5586         struct pf_rule          *a = NULL, *r = &pf_default_rule, *tr, *nr;
5587         struct pf_state         *s = NULL;
5588         struct pf_state_key     *sk = NULL;
5589         struct pf_ruleset       *ruleset = NULL;
5590         struct pf_pdesc          pd;
5591         int                      off, dirndx, pqid = 0;
5592
5593         if (!pf_status.running)
5594                 return (PF_PASS);
5595
5596         memset(&pd, 0, sizeof(pd));
5597         if (ifp->if_type == IFT_CARP && ifp->if_carpdev)
5598                 kif = (struct pfi_kif *)ifp->if_carpdev->if_pf_kif;
5599         else
5600                 kif = (struct pfi_kif *)ifp->if_pf_kif;
5601
5602         if (kif == NULL) {
5603                 DPFPRINTF(PF_DEBUG_URGENT,
5604                     ("pf_test: kif == NULL, if_xname %s\n", ifp->if_xname));
5605                 return (PF_DROP);
5606         }
5607         if (kif->pfik_flags & PFI_IFLAG_SKIP)
5608                 return (PF_PASS);
5609
5610 #ifdef DIAGNOSTIC
5611         if ((m->m_flags & M_PKTHDR) == 0)
5612                 panic("non-M_PKTHDR is passed to pf_test");
5613 #endif /* DIAGNOSTIC */
5614
5615         if (m->m_pkthdr.len < (int)sizeof(*h)) {
5616                 action = PF_DROP;
5617                 REASON_SET(&reason, PFRES_SHORT);
5618                 log = 1;
5619                 goto done;
5620         }
5621
5622         if (m->m_pkthdr.pf.flags & PF_TAG_GENERATED)
5623                 return (PF_PASS);
5624
5625         /* We do IP header normalization and packet reassembly here */
5626         if (pf_normalize_ip(m0, dir, kif, &reason, &pd) != PF_PASS) {
5627                 action = PF_DROP;
5628                 goto done;
5629         }
5630         m = *m0;        /* pf_normalize messes with m0 */
5631         h = mtod(m, struct ip *);
5632
5633         off = h->ip_hl << 2;
5634         if (off < (int)sizeof(*h)) {
5635                 action = PF_DROP;
5636                 REASON_SET(&reason, PFRES_SHORT);
5637                 log = 1;
5638                 goto done;
5639         }
5640
5641         pd.src = (struct pf_addr *)&h->ip_src;
5642         pd.dst = (struct pf_addr *)&h->ip_dst;
5643         PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET);
5644         pd.ip_sum = &h->ip_sum;
5645         pd.proto = h->ip_p;
5646         pd.af = AF_INET;
5647         pd.tos = h->ip_tos;
5648         pd.tot_len = h->ip_len;
5649         pd.eh = eh;
5650
5651         /* handle fragments that didn't get reassembled by normalization */
5652         if (h->ip_off & (IP_MF | IP_OFFMASK)) {
5653                 action = pf_test_fragment(&r, dir, kif, m, h,
5654                     &pd, &a, &ruleset);
5655                 goto done;
5656         }
5657
5658         switch (h->ip_p) {
5659
5660         case IPPROTO_TCP: {
5661                 struct tcphdr   th;
5662
5663                 pd.hdr.tcp = &th;
5664                 if (!pf_pull_hdr(m, off, &th, sizeof(th),
5665                     &action, &reason, AF_INET)) {
5666                         log = action != PF_PASS;
5667                         goto done;
5668                 }
5669                 pd.p_len = pd.tot_len - off - (th.th_off << 2);
5670                 if ((th.th_flags & TH_ACK) && pd.p_len == 0)
5671                         pqid = 1;
5672                 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
5673                 if (action == PF_DROP)
5674                         goto done;
5675                 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
5676                     &reason);
5677                 if (action == PF_PASS) {
5678 #if NPFSYNC
5679                         pfsync_update_state(s);
5680 #endif /* NPFSYNC */
5681                         r = s->rule.ptr;
5682                         a = s->anchor.ptr;
5683                         log = s->log;
5684                 } else if (s == NULL)
5685                         action = pf_test_rule(&r, &s, dir, kif,
5686                             m, off, h, &pd, &a, &ruleset, NULL, inp);
5687                 break;
5688         }
5689
5690         case IPPROTO_UDP: {
5691                 struct udphdr   uh;
5692
5693                 pd.hdr.udp = &uh;
5694                 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
5695                     &action, &reason, AF_INET)) {
5696                         log = action != PF_PASS;
5697                         goto done;
5698                 }
5699                 if (uh.uh_dport == 0 ||
5700                     ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
5701                     ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
5702                         action = PF_DROP;
5703                         REASON_SET(&reason, PFRES_SHORT);
5704                         goto done;
5705                 }
5706                 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
5707                 if (action == PF_PASS) {
5708 #if NPFSYNC
5709                         pfsync_update_state(s);
5710 #endif /* NPFSYNC */
5711                         r = s->rule.ptr;
5712                         a = s->anchor.ptr;
5713                         log = s->log;
5714                 } else if (s == NULL)
5715                         action = pf_test_rule(&r, &s, dir, kif,
5716                             m, off, h, &pd, &a, &ruleset, NULL, inp);
5717                 break;
5718         }
5719
5720         case IPPROTO_ICMP: {
5721                 struct icmp     ih;
5722
5723                 pd.hdr.icmp = &ih;
5724                 if (!pf_pull_hdr(m, off, &ih, ICMP_MINLEN,
5725                     &action, &reason, AF_INET)) {
5726                         log = action != PF_PASS;
5727                         goto done;
5728                 }
5729                 action = pf_test_state_icmp(&s, dir, kif, m, off, h, &pd,
5730                     &reason);
5731                 if (action == PF_PASS) {
5732 #if NPFSYNC
5733                         pfsync_update_state(s);
5734 #endif /* NPFSYNC */
5735                         r = s->rule.ptr;
5736                         a = s->anchor.ptr;
5737                         log = s->log;
5738                 } else if (s == NULL)
5739                         action = pf_test_rule(&r, &s, dir, kif,
5740                             m, off, h, &pd, &a, &ruleset, NULL, inp);
5741                 break;
5742         }
5743
5744         default:
5745                 action = pf_test_state_other(&s, dir, kif, &pd);
5746                 if (action == PF_PASS) {
5747 #if NPFSYNC
5748                         pfsync_update_state(s);
5749 #endif /* NPFSYNC */
5750                         r = s->rule.ptr;
5751                         a = s->anchor.ptr;
5752                         log = s->log;
5753                 } else if (s == NULL)
5754                         action = pf_test_rule(&r, &s, dir, kif, m, off, h,
5755                             &pd, &a, &ruleset, NULL, inp);
5756                 break;
5757         }
5758
5759 done:
5760         if (action == PF_PASS && h->ip_hl > 5 &&
5761             !((s && s->allow_opts) || r->allow_opts)) {
5762                 action = PF_DROP;
5763                 REASON_SET(&reason, PFRES_IPOPTIONS);
5764                 log = 1;
5765                 DPFPRINTF(PF_DEBUG_MISC,
5766                     ("pf: dropping packet with ip options\n"));
5767         }
5768
5769         if ((s && s->tag) || r->rtableid)
5770                 pf_tag_packet(m, s ? s->tag : 0, r->rtableid);
5771
5772 #ifdef ALTQ
5773         if (action == PF_PASS && r->qid) {
5774                 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
5775                 if (pqid || (pd.tos & IPTOS_LOWDELAY))
5776                         m->m_pkthdr.pf.qid = r->pqid;
5777                 else
5778                         m->m_pkthdr.pf.qid = r->qid;
5779                 m->m_pkthdr.pf.ecn_af = AF_INET;
5780                 m->m_pkthdr.pf.hdr = h;
5781                 /* add connection hash for fairq */
5782                 if (s) {
5783                         /* for fairq */
5784                         m->m_pkthdr.pf.state_hash = s->hash;
5785                         m->m_pkthdr.pf.flags |= PF_TAG_STATE_HASHED;
5786                 }
5787         }
5788 #endif /* ALTQ */
5789
5790         /*
5791          * connections redirected to loopback should not match sockets
5792          * bound specifically to loopback due to security implications,
5793          * see tcp_input() and in_pcblookup_listen().
5794          */
5795         if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
5796             pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
5797             (s->nat_rule.ptr->action == PF_RDR ||
5798             s->nat_rule.ptr->action == PF_BINAT) &&
5799             (ntohl(pd.dst->v4.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET)
5800                 m->m_pkthdr.pf.flags |= PF_TAG_TRANSLATE_LOCALHOST;
5801
5802         if (log) {
5803                 struct pf_rule *lr;
5804
5805                 if (s != NULL && s->nat_rule.ptr != NULL &&
5806                     s->nat_rule.ptr->log & PF_LOG_ALL)
5807                         lr = s->nat_rule.ptr;
5808                 else
5809                         lr = r;
5810                 PFLOG_PACKET(kif, h, m, AF_INET, dir, reason, lr, a, ruleset,
5811                     &pd);
5812         }
5813
5814         kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
5815         kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++;
5816
5817         if (action == PF_PASS || r->action == PF_DROP) {
5818                 dirndx = (dir == PF_OUT);
5819                 r->packets[dirndx]++;
5820                 r->bytes[dirndx] += pd.tot_len;
5821                 if (a != NULL) {
5822                         a->packets[dirndx]++;
5823                         a->bytes[dirndx] += pd.tot_len;
5824                 }
5825                 if (s != NULL) {
5826                         sk = s->state_key;
5827                         if (s->nat_rule.ptr != NULL) {
5828                                 s->nat_rule.ptr->packets[dirndx]++;
5829                                 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len;
5830                         }
5831                         if (s->src_node != NULL) {
5832                                 s->src_node->packets[dirndx]++;
5833                                 s->src_node->bytes[dirndx] += pd.tot_len;
5834                         }
5835                         if (s->nat_src_node != NULL) {
5836                                 s->nat_src_node->packets[dirndx]++;
5837                                 s->nat_src_node->bytes[dirndx] += pd.tot_len;
5838                         }
5839                         dirndx = (dir == sk->direction) ? 0 : 1;
5840                         s->packets[dirndx]++;
5841                         s->bytes[dirndx] += pd.tot_len;
5842                 }
5843                 tr = r;
5844                 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
5845                 if (nr != NULL) {
5846                         struct pf_addr *x;
5847                         /*
5848                          * XXX: we need to make sure that the addresses
5849                          * passed to pfr_update_stats() are the same than
5850                          * the addresses used during matching (pfr_match)
5851                          */
5852                         if (r == &pf_default_rule) {
5853                                 tr = nr;
5854                                 x = (sk == NULL || sk->direction == dir) ?
5855                                     &pd.baddr : &pd.naddr;
5856                         } else
5857                                 x = (sk == NULL || sk->direction == dir) ?
5858                                     &pd.naddr : &pd.baddr;
5859                         if (x == &pd.baddr || s == NULL) {
5860                                 /* we need to change the address */
5861                                 if (dir == PF_OUT)
5862                                         pd.src = x;
5863                                 else
5864                                         pd.dst = x;
5865                         }
5866                 }
5867                 if (tr->src.addr.type == PF_ADDR_TABLE)
5868                         pfr_update_stats(tr->src.addr.p.tbl, (sk == NULL ||
5869                             sk->direction == dir) ?
5870                             pd.src : pd.dst, pd.af,
5871                             pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5872                             tr->src.neg);
5873                 if (tr->dst.addr.type == PF_ADDR_TABLE)
5874                         pfr_update_stats(tr->dst.addr.p.tbl, (sk == NULL ||
5875                             sk->direction == dir) ? pd.dst : pd.src, pd.af,
5876                             pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5877                             tr->dst.neg);
5878         }
5879
5880
5881         if (action == PF_SYNPROXY_DROP) {
5882                 m_freem(*m0);
5883                 *m0 = NULL;
5884                 action = PF_PASS;
5885         } else if (r->rt)
5886                 /* pf_route can free the mbuf causing *m0 to become NULL */
5887                 pf_route(m0, r, dir, kif->pfik_ifp, s, &pd);
5888
5889         return (action);
5890 }
5891 #endif /* INET */
5892
5893 #ifdef INET6
5894 int
5895 pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0,
5896     struct ether_header *eh, struct inpcb *inp)
5897 {
5898         struct pfi_kif          *kif;
5899         u_short                  action, reason = 0, log = 0;
5900         struct mbuf             *m = *m0, *n = NULL;
5901         struct ip6_hdr          *h = NULL;
5902         struct pf_rule          *a = NULL, *r = &pf_default_rule, *tr, *nr;
5903         struct pf_state         *s = NULL;
5904         struct pf_state_key     *sk = NULL;
5905         struct pf_ruleset       *ruleset = NULL;
5906         struct pf_pdesc          pd;
5907         int                      off, terminal = 0, dirndx, rh_cnt = 0;
5908
5909         if (!pf_status.running)
5910                 return (PF_PASS);
5911
5912         memset(&pd, 0, sizeof(pd));
5913         if (ifp->if_type == IFT_CARP && ifp->if_carpdev)
5914                 kif = (struct pfi_kif *)ifp->if_carpdev->if_pf_kif;
5915         else
5916                 kif = (struct pfi_kif *)ifp->if_pf_kif;
5917
5918         if (kif == NULL) {
5919                 DPFPRINTF(PF_DEBUG_URGENT,
5920                     ("pf_test6: kif == NULL, if_xname %s\n", ifp->if_xname));
5921                 return (PF_DROP);
5922         }
5923         if (kif->pfik_flags & PFI_IFLAG_SKIP)
5924                 return (PF_PASS);
5925
5926 #ifdef DIAGNOSTIC
5927         if ((m->m_flags & M_PKTHDR) == 0)
5928                 panic("non-M_PKTHDR is passed to pf_test6");
5929 #endif /* DIAGNOSTIC */
5930
5931         if (m->m_pkthdr.len < (int)sizeof(*h)) {
5932                 action = PF_DROP;
5933                 REASON_SET(&reason, PFRES_SHORT);
5934                 log = 1;
5935                 goto done;
5936         }
5937
5938         if (m->m_pkthdr.pf.flags & PF_TAG_GENERATED)
5939                 return (PF_PASS);
5940
5941         /* We do IP header normalization and packet reassembly here */
5942         if (pf_normalize_ip6(m0, dir, kif, &reason, &pd) != PF_PASS) {
5943                 action = PF_DROP;
5944                 goto done;
5945         }
5946         m = *m0;        /* pf_normalize messes with m0 */
5947         h = mtod(m, struct ip6_hdr *);
5948
5949 #if 1
5950         /*
5951          * we do not support jumbogram yet.  if we keep going, zero ip6_plen
5952          * will do something bad, so drop the packet for now.
5953          */
5954         if (htons(h->ip6_plen) == 0) {
5955                 action = PF_DROP;
5956                 REASON_SET(&reason, PFRES_NORM);        /*XXX*/
5957                 goto done;
5958         }
5959 #endif
5960
5961         pd.src = (struct pf_addr *)&h->ip6_src;
5962         pd.dst = (struct pf_addr *)&h->ip6_dst;
5963         PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET6);
5964         pd.ip_sum = NULL;
5965         pd.af = AF_INET6;
5966         pd.tos = 0;
5967         pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr);
5968         pd.eh = eh;
5969
5970         off = ((caddr_t)h - m->m_data) + sizeof(struct ip6_hdr);
5971         pd.proto = h->ip6_nxt;
5972         do {
5973                 switch (pd.proto) {
5974                 case IPPROTO_FRAGMENT:
5975                         action = pf_test_fragment(&r, dir, kif, m, h,
5976                             &pd, &a, &ruleset);
5977                         if (action == PF_DROP)
5978                                 REASON_SET(&reason, PFRES_FRAG);
5979                         goto done;
5980                 case IPPROTO_ROUTING: {
5981                         struct ip6_rthdr rthdr;
5982
5983                         if (rh_cnt++) {
5984                                 DPFPRINTF(PF_DEBUG_MISC,
5985                                     ("pf: IPv6 more than one rthdr\n"));
5986                                 action = PF_DROP;
5987                                 REASON_SET(&reason, PFRES_IPOPTIONS);
5988                                 log = 1;
5989                                 goto done;
5990                         }
5991                         if (!pf_pull_hdr(m, off, &rthdr, sizeof(rthdr), NULL,
5992                             &reason, pd.af)) {
5993                                 DPFPRINTF(PF_DEBUG_MISC,
5994                                     ("pf: IPv6 short rthdr\n"));
5995                                 action = PF_DROP;
5996                                 REASON_SET(&reason, PFRES_SHORT);
5997                                 log = 1;
5998                                 goto done;
5999                         }
6000                         if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) {
6001                                 DPFPRINTF(PF_DEBUG_MISC,
6002                                     ("pf: IPv6 rthdr0\n"));
6003                                 action = PF_DROP;
6004                                 REASON_SET(&reason, PFRES_IPOPTIONS);
6005                                 log = 1;
6006                                 goto done;
6007                         }
6008                         /* FALLTHROUGH */
6009                 }
6010                 case IPPROTO_AH:
6011                 case IPPROTO_HOPOPTS:
6012                 case IPPROTO_DSTOPTS: {
6013                         /* get next header and header length */
6014                         struct ip6_ext  opt6;
6015
6016                         if (!pf_pull_hdr(m, off, &opt6, sizeof(opt6),
6017                             NULL, &reason, pd.af)) {
6018                                 DPFPRINTF(PF_DEBUG_MISC,
6019                                     ("pf: IPv6 short opt\n"));
6020                                 action = PF_DROP;
6021                                 log = 1;
6022                                 goto done;
6023                         }
6024                         if (pd.proto == IPPROTO_AH)
6025                                 off += (opt6.ip6e_len + 2) * 4;
6026                         else
6027                                 off += (opt6.ip6e_len + 1) * 8;
6028                         pd.proto = opt6.ip6e_nxt;
6029                         /* goto the next header */
6030                         break;
6031                 }
6032                 default:
6033                         terminal++;
6034                         break;
6035                 }
6036         } while (!terminal);
6037
6038         /* if there's no routing header, use unmodified mbuf for checksumming */
6039         if (!n)
6040                 n = m;
6041
6042         switch (pd.proto) {
6043
6044         case IPPROTO_TCP: {
6045                 struct tcphdr   th;
6046
6047                 pd.hdr.tcp = &th;
6048                 if (!pf_pull_hdr(m, off, &th, sizeof(th),
6049                     &action, &reason, AF_INET6)) {
6050                         log = action != PF_PASS;
6051                         goto done;
6052                 }
6053                 pd.p_len = pd.tot_len - off - (th.th_off << 2);
6054                 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
6055                 if (action == PF_DROP)
6056                         goto done;
6057                 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
6058                     &reason);
6059                 if (action == PF_PASS) {
6060 #if NPFSYNC
6061                         pfsync_update_state(s);
6062 #endif /* NPFSYNC */
6063                         r = s->rule.ptr;
6064                         a = s->anchor.ptr;
6065                         log = s->log;
6066                 } else if (s == NULL)
6067                         action = pf_test_rule(&r, &s, dir, kif,
6068                             m, off, h, &pd, &a, &ruleset, NULL, inp);
6069                 break;
6070         }
6071
6072         case IPPROTO_UDP: {
6073                 struct udphdr   uh;
6074
6075                 pd.hdr.udp = &uh;
6076                 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
6077                     &action, &reason, AF_INET6)) {
6078                         log = action != PF_PASS;
6079                         goto done;
6080                 }
6081                 if (uh.uh_dport == 0 ||
6082                     ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
6083                     ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
6084                         action = PF_DROP;
6085                         REASON_SET(&reason, PFRES_SHORT);
6086                         goto done;
6087                 }
6088                 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
6089                 if (action == PF_PASS) {
6090 #if NPFSYNC
6091                         pfsync_update_state(s);
6092 #endif /* NPFSYNC */
6093                         r = s->rule.ptr;
6094                         a = s->anchor.ptr;
6095                         log = s->log;
6096                 } else if (s == NULL)
6097                         action = pf_test_rule(&r, &s, dir, kif,
6098                             m, off, h, &pd, &a, &ruleset, NULL, inp);
6099                 break;
6100         }
6101
6102         case IPPROTO_ICMPV6: {
6103                 struct icmp6_hdr        ih;
6104
6105                 pd.hdr.icmp6 = &ih;
6106                 if (!pf_pull_hdr(m, off, &ih, sizeof(ih),
6107                     &action, &reason, AF_INET6)) {
6108                         log = action != PF_PASS;
6109                         goto done;
6110                 }
6111                 action = pf_test_state_icmp(&s, dir, kif,
6112                     m, off, h, &pd, &reason);
6113                 if (action == PF_PASS) {
6114 #if NPFSYNC
6115                         pfsync_update_state(s);
6116 #endif /* NPFSYNC */
6117                         r = s->rule.ptr;
6118                         a = s->anchor.ptr;
6119                         log = s->log;
6120                 } else if (s == NULL)
6121                         action = pf_test_rule(&r, &s, dir, kif,
6122                             m, off, h, &pd, &a, &ruleset, NULL, inp);
6123                 break;
6124         }
6125
6126         default:
6127                 action = pf_test_state_other(&s, dir, kif, &pd);
6128                 if (action == PF_PASS) {
6129 #if NPFSYNC
6130                         pfsync_update_state(s);
6131 #endif /* NPFSYNC */
6132                         r = s->rule.ptr;
6133                         a = s->anchor.ptr;
6134                         log = s->log;
6135                 } else if (s == NULL)
6136                         action = pf_test_rule(&r, &s, dir, kif, m, off, h,
6137                             &pd, &a, &ruleset, NULL, inp);
6138                 break;
6139         }
6140
6141 done:
6142         if (n != m) {
6143                 m_freem(n);
6144                 n = NULL;
6145         }
6146
6147         /* handle dangerous IPv6 extension headers. */
6148         if (action == PF_PASS && rh_cnt &&
6149             !((s && s->allow_opts) || r->allow_opts)) {
6150                 action = PF_DROP;
6151                 REASON_SET(&reason, PFRES_IPOPTIONS);
6152                 log = 1;
6153                 DPFPRINTF(PF_DEBUG_MISC,
6154                     ("pf: dropping packet with dangerous v6 headers\n"));
6155         }
6156
6157         if ((s && s->tag) || r->rtableid)
6158                 pf_tag_packet(m, s ? s->tag : 0, r->rtableid);
6159
6160 #ifdef ALTQ
6161         if (action == PF_PASS && r->qid) {
6162                 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
6163                 if (pd.tos & IPTOS_LOWDELAY)
6164                         m->m_pkthdr.pf.qid = r->pqid;
6165                 else
6166                         m->m_pkthdr.pf.qid = r->qid;
6167                 m->m_pkthdr.pf.ecn_af = AF_INET6;
6168                 m->m_pkthdr.pf.hdr = h;
6169                 if (s) {
6170                         /* for fairq */
6171                         m->m_pkthdr.pf.state_hash = s->hash;
6172                         m->m_pkthdr.pf.flags |= PF_TAG_STATE_HASHED;
6173                 }
6174         }
6175 #endif /* ALTQ */
6176
6177         if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
6178             pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
6179             (s->nat_rule.ptr->action == PF_RDR ||
6180             s->nat_rule.ptr->action == PF_BINAT) &&
6181             IN6_IS_ADDR_LOOPBACK(&pd.dst->v6))
6182                 m->m_pkthdr.pf.flags |= PF_TAG_TRANSLATE_LOCALHOST;
6183
6184         if (log) {
6185                 struct pf_rule *lr;
6186
6187                 if (s != NULL && s->nat_rule.ptr != NULL &&
6188                     s->nat_rule.ptr->log & PF_LOG_ALL)
6189                         lr = s->nat_rule.ptr;
6190                 else
6191                         lr = r;
6192                 PFLOG_PACKET(kif, h, m, AF_INET6, dir, reason, lr, a, ruleset,
6193                     &pd);
6194         }
6195
6196         kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
6197         kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++;
6198
6199         if (action == PF_PASS || r->action == PF_DROP) {
6200                 dirndx = (dir == PF_OUT);
6201                 r->packets[dirndx]++;
6202                 r->bytes[dirndx] += pd.tot_len;
6203                 if (a != NULL) {
6204                         a->packets[dirndx]++;
6205                         a->bytes[dirndx] += pd.tot_len;
6206                 }
6207                 if (s != NULL) {
6208                         sk = s->state_key;
6209                         if (s->nat_rule.ptr != NULL) {
6210                                 s->nat_rule.ptr->packets[dirndx]++;
6211                                 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len;
6212                         }
6213                         if (s->src_node != NULL) {
6214                                 s->src_node->packets[dirndx]++;
6215                                 s->src_node->bytes[dirndx] += pd.tot_len;
6216                         }
6217                         if (s->nat_src_node != NULL) {
6218                                 s->nat_src_node->packets[dirndx]++;
6219                                 s->nat_src_node->bytes[dirndx] += pd.tot_len;
6220                         }
6221                         dirndx = (dir == sk->direction) ? 0 : 1;
6222                         s->packets[dirndx]++;
6223                         s->bytes[dirndx] += pd.tot_len;
6224                 }
6225                 tr = r;
6226                 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
6227                 if (nr != NULL) {
6228                         struct pf_addr *x;
6229                         /*
6230                          * XXX: we need to make sure that the addresses
6231                          * passed to pfr_update_stats() are the same than
6232                          * the addresses used during matching (pfr_match)
6233                          */
6234                         if (r == &pf_default_rule) {
6235                                 tr = nr;
6236                                 x = (s == NULL || sk->direction == dir) ?
6237                                     &pd.baddr : &pd.naddr;
6238                         } else {
6239                                 x = (s == NULL || sk->direction == dir) ?
6240                                     &pd.naddr : &pd.baddr;
6241                         }
6242                         if (x == &pd.baddr || s == NULL) {
6243                                 if (dir == PF_OUT)
6244                                         pd.src = x;
6245                                 else
6246                                         pd.dst = x;
6247                         }
6248                 }
6249                 if (tr->src.addr.type == PF_ADDR_TABLE)
6250                         pfr_update_stats(tr->src.addr.p.tbl, (sk == NULL ||
6251                             sk->direction == dir) ? pd.src : pd.dst, pd.af,
6252                             pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
6253                             tr->src.neg);
6254                 if (tr->dst.addr.type == PF_ADDR_TABLE)
6255                         pfr_update_stats(tr->dst.addr.p.tbl, (sk == NULL ||
6256                             sk->direction == dir) ? pd.dst : pd.src, pd.af,
6257                             pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
6258                             tr->dst.neg);
6259         }
6260
6261
6262         if (action == PF_SYNPROXY_DROP) {
6263                 m_freem(*m0);
6264                 *m0 = NULL;
6265                 action = PF_PASS;
6266         } else if (r->rt)
6267                 /* pf_route6 can free the mbuf causing *m0 to become NULL */
6268                 pf_route6(m0, r, dir, kif->pfik_ifp, s, &pd);
6269
6270         return (action);
6271 }
6272 #endif /* INET6 */
6273
6274 int
6275 pf_check_congestion(struct ifqueue *ifq)
6276 {
6277                 return (0);
6278 }