sys/net/pf/pf.c

   1 /*      $OpenBSD: pf.c,v 1.614 2008/08/02 12:34:37 henning Exp $ */
   2
   3 /*
   4  * Copyright (c) 2004 The DragonFly Project.  All rights reserved.
   5  *
   6  * Copyright (c) 2001 Daniel Hartmeier
   7  * Copyright (c) 2002 - 2008 Henning Brauer
   8  * All rights reserved.
   9  *
  10  * Redistribution and use in source and binary forms, with or without
  11  * modification, are permitted provided that the following conditions
  12  * are met:
  13  *
  14  *    - Redistributions of source code must retain the above copyright
  15  *      notice, this list of conditions and the following disclaimer.
  16  *    - Redistributions in binary form must reproduce the above
  17  *      copyright notice, this list of conditions and the following
  18  *      disclaimer in the documentation and/or other materials provided
  19  *      with the distribution.
  20  *
  21  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  22  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  23  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
  24  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
  25  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
  26  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
  27  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  28  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
  29  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
  31  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  32  * POSSIBILITY OF SUCH DAMAGE.
  33  *
  34  * Effort sponsored in part by the Defense Advanced Research Projects
  35  * Agency (DARPA) and Air Force Research Laboratory, Air Force
  36  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
  37  *
  38  */
  39
  40 #include "opt_inet.h"
  41 #include "opt_inet6.h"
  42
  43 #include <sys/param.h>
  44 #include <sys/systm.h>
  45 #include <sys/malloc.h>
  46 #include <sys/mbuf.h>
  47 #include <sys/filio.h>
  48 #include <sys/socket.h>
  49 #include <sys/socketvar.h>
  50 #include <sys/kernel.h>
  51 #include <sys/time.h>
  52 #include <sys/sysctl.h>
  53 #include <sys/endian.h>
  54 #include <vm/vm_zone.h>
  55 #include <sys/proc.h>
  56 #include <sys/kthread.h>
  57
  58 #include <machine/inttypes.h>
  59
  60 #include <sys/md5.h>
  61
  62 #include <net/if.h>
  63 #include <net/if_types.h>
  64 #include <net/bpf.h>
  65 #include <net/netisr.h>
  66 #include <net/route.h>
  67
  68 #include <netinet/in.h>
  69 #include <netinet/in_var.h>
  70 #include <netinet/in_systm.h>
  71 #include <netinet/ip.h>
  72 #include <netinet/ip_var.h>
  73 #include <netinet/tcp.h>
  74 #include <netinet/tcp_seq.h>
  75 #include <netinet/udp.h>
  76 #include <netinet/ip_icmp.h>
  77 #include <netinet/in_pcb.h>
  78 #include <netinet/tcp_timer.h>
  79 #include <netinet/tcp_var.h>
  80 #include <netinet/udp_var.h>
  81 #include <netinet/icmp_var.h>
  82 #include <netinet/if_ether.h>
  83
  84 #include <net/pf/pfvar.h>
  85 #include <net/pf/if_pflog.h>
  86
  87 #include <net/pf/if_pfsync.h>
  88
  89 #ifdef INET6
  90 #include <netinet/ip6.h>
  91 #include <netinet/in_pcb.h>
  92 #include <netinet/icmp6.h>
  93 #include <netinet6/nd6.h>
  94 #include <netinet6/ip6_var.h>
  95 #include <netinet6/in6_pcb.h>
  96 #endif /* INET6 */
  97
  98 #include <sys/in_cksum.h>
  99 #include <sys/ucred.h>
 100 #include <machine/limits.h>
 101 #include <sys/msgport2.h>
 102 #include <net/netmsg2.h>
 103
 104 extern int ip_optcopy(struct ip *, struct ip *);
 105 extern int debug_pfugidhack;
 106
 107 struct lwkt_token pf_token = LWKT_TOKEN_INITIALIZER(pf_token);
 108
 109 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) kprintf x
 110
 111 /*
 112  * Global variables
 113  */
 114
 115 /* state tables */
 116 struct pf_state_tree     pf_statetbl;
 117
 118 struct pf_altqqueue      pf_altqs[2];
 119 struct pf_palist         pf_pabuf;
 120 struct pf_altqqueue     *pf_altqs_active;
 121 struct pf_altqqueue     *pf_altqs_inactive;
 122 struct pf_status         pf_status;
 123
 124 u_int32_t                ticket_altqs_active;
 125 u_int32_t                ticket_altqs_inactive;
 126 int                      altqs_inactive_open;
 127 u_int32_t                ticket_pabuf;
 128
 129 MD5_CTX                  pf_tcp_secret_ctx;
 130 u_char                   pf_tcp_secret[16];
 131 int                      pf_tcp_secret_init;
 132 int                      pf_tcp_iss_off;
 133
 134 struct pf_anchor_stackframe {
 135         struct pf_ruleset                       *rs;
 136         struct pf_rule                          *r;
 137         struct pf_anchor_node                   *parent;
 138         struct pf_anchor                        *child;
 139 } pf_anchor_stack[64];
 140
 141 vm_zone_t                pf_src_tree_pl, pf_rule_pl, pf_pooladdr_pl;
 142 vm_zone_t                pf_state_pl, pf_state_key_pl, pf_state_item_pl;
 143 vm_zone_t                pf_altq_pl;
 144
 145 void                     pf_print_host(struct pf_addr *, u_int16_t, u_int8_t);
 146
 147 void                     pf_init_threshold(struct pf_threshold *, u_int32_t,
 148                             u_int32_t);
 149 void                     pf_add_threshold(struct pf_threshold *);
 150 int                      pf_check_threshold(struct pf_threshold *);
 151
 152 void                     pf_change_ap(struct pf_addr *, u_int16_t *,
 153                             u_int16_t *, u_int16_t *, struct pf_addr *,
 154                             u_int16_t, u_int8_t, sa_family_t);
 155 int                      pf_modulate_sack(struct mbuf *, int, struct pf_pdesc *,
 156                             struct tcphdr *, struct pf_state_peer *);
 157 #ifdef INET6
 158 void                     pf_change_a6(struct pf_addr *, u_int16_t *,
 159                             struct pf_addr *, u_int8_t);
 160 #endif /* INET6 */
 161 void                     pf_change_icmp(struct pf_addr *, u_int16_t *,
 162                             struct pf_addr *, struct pf_addr *, u_int16_t,
 163                             u_int16_t *, u_int16_t *, u_int16_t *,
 164                             u_int16_t *, u_int8_t, sa_family_t);
 165 void                     pf_send_tcp(const struct pf_rule *, sa_family_t,
 166                             const struct pf_addr *, const struct pf_addr *,
 167                             u_int16_t, u_int16_t, u_int32_t, u_int32_t,
 168                             u_int8_t, u_int16_t, u_int16_t, u_int8_t, int,
 169                             u_int16_t, struct ether_header *, struct ifnet *);
 170 void                     pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t,
 171                             sa_family_t, struct pf_rule *);
 172 struct pf_rule          *pf_match_translation(struct pf_pdesc *, struct mbuf *,
 173                             int, int, struct pfi_kif *,
 174                             struct pf_addr *, u_int16_t, struct pf_addr *,
 175                             u_int16_t, int);
 176 struct pf_rule          *pf_get_translation(struct pf_pdesc *, struct mbuf *,
 177                             int, int, struct pfi_kif *, struct pf_src_node **,
 178                             struct pf_state_key **, struct pf_state_key **,
 179                             struct pf_state_key **, struct pf_state_key **,
 180                             struct pf_addr *, struct pf_addr *,
 181                             u_int16_t, u_int16_t);
 182 void                     pf_detach_state(struct pf_state *);
 183 int                      pf_state_key_setup(struct pf_pdesc *, struct pf_rule *,
 184                             struct pf_state_key **, struct pf_state_key **,
 185                             struct pf_state_key **, struct pf_state_key **,
 186                             struct pf_addr *, struct pf_addr *,
 187                             u_int16_t, u_int16_t);
 188 void                     pf_state_key_detach(struct pf_state *, int);
 189 u_int32_t                pf_tcp_iss(struct pf_pdesc *);
 190 int                      pf_test_rule(struct pf_rule **, struct pf_state **,
 191                             int, struct pfi_kif *, struct mbuf *, int,
 192                             void *, struct pf_pdesc *, struct pf_rule **,
 193                             struct pf_ruleset **, struct ifqueue *, struct inpcb *);
 194 static __inline int      pf_create_state(struct pf_rule *, struct pf_rule *,
 195                             struct pf_rule *, struct pf_pdesc *,
 196                             struct pf_src_node *, struct pf_state_key *,
 197                             struct pf_state_key *, struct pf_state_key *,
 198                             struct pf_state_key *, struct mbuf *, int,
 199                             u_int16_t, u_int16_t, int *, struct pfi_kif *,
 200                             struct pf_state **, int, u_int16_t, u_int16_t,
 201                             int);
 202 int                      pf_test_fragment(struct pf_rule **, int,
 203                             struct pfi_kif *, struct mbuf *, void *,
 204                             struct pf_pdesc *, struct pf_rule **,
 205                             struct pf_ruleset **);
 206 int                      pf_tcp_track_full(struct pf_state_peer *,
 207                             struct pf_state_peer *, struct pf_state **,
 208                             struct pfi_kif *, struct mbuf *, int,
 209                             struct pf_pdesc *, u_short *, int *);
 210 int                     pf_tcp_track_sloppy(struct pf_state_peer *,
 211                             struct pf_state_peer *, struct pf_state **,
 212                             struct pf_pdesc *, u_short *);
 213 int                      pf_test_state_tcp(struct pf_state **, int,
 214                             struct pfi_kif *, struct mbuf *, int,
 215                             void *, struct pf_pdesc *, u_short *);
 216 int                      pf_test_state_udp(struct pf_state **, int,
 217                             struct pfi_kif *, struct mbuf *, int,
 218                             void *, struct pf_pdesc *);
 219 int                      pf_test_state_icmp(struct pf_state **, int,
 220                             struct pfi_kif *, struct mbuf *, int,
 221                             void *, struct pf_pdesc *, u_short *);
 222 int                      pf_test_state_other(struct pf_state **, int,
 223                             struct pfi_kif *, struct mbuf *, struct pf_pdesc *);
 224 void                     pf_step_into_anchor(int *, struct pf_ruleset **, int,
 225                             struct pf_rule **, struct pf_rule **, int *);
 226 int                      pf_step_out_of_anchor(int *, struct pf_ruleset **,
 227                              int, struct pf_rule **, struct pf_rule **,
 228                              int *);
 229 void                     pf_hash(struct pf_addr *, struct pf_addr *,
 230                             struct pf_poolhashkey *, sa_family_t);
 231 int                      pf_map_addr(u_int8_t, struct pf_rule *,
 232                             struct pf_addr *, struct pf_addr *,
 233                             struct pf_addr *, struct pf_src_node **);
 234 int                      pf_get_sport(sa_family_t, u_int8_t, struct pf_rule *,
 235                             struct pf_addr *, struct pf_addr *, u_int16_t,
 236                             struct pf_addr *, u_int16_t*, u_int16_t, u_int16_t,
 237                             struct pf_src_node **);
 238 void                     pf_route(struct mbuf **, struct pf_rule *, int,
 239                             struct ifnet *, struct pf_state *,
 240                             struct pf_pdesc *);
 241 void                     pf_route6(struct mbuf **, struct pf_rule *, int,
 242                             struct ifnet *, struct pf_state *,
 243                             struct pf_pdesc *);
 244 u_int8_t                 pf_get_wscale(struct mbuf *, int, u_int16_t,
 245                             sa_family_t);
 246 u_int16_t                pf_get_mss(struct mbuf *, int, u_int16_t,
 247                             sa_family_t);
 248 u_int16_t                pf_calc_mss(struct pf_addr *, sa_family_t,
 249                                 u_int16_t);
 250 void                     pf_set_rt_ifp(struct pf_state *,
 251                             struct pf_addr *);
 252 int                      pf_check_proto_cksum(struct mbuf *, int, int,
 253                             u_int8_t, sa_family_t);
 254 struct pf_divert        *pf_get_divert(struct mbuf *);
 255 void                     pf_print_state_parts(struct pf_state *,
 256                             struct pf_state_key *, struct pf_state_key *);
 257 int                      pf_addr_wrap_neq(struct pf_addr_wrap *,
 258                             struct pf_addr_wrap *);
 259 struct pf_state         *pf_find_state(struct pfi_kif *,
 260                             struct pf_state_key_cmp *, u_int, struct mbuf *);
 261 int                      pf_src_connlimit(struct pf_state **);
 262 int                      pf_check_congestion(struct ifqueue *);
 263
 264 extern int pf_end_threads;
 265
 266 struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] = {
 267         { &pf_state_pl, PFSTATE_HIWAT },
 268         { &pf_src_tree_pl, PFSNODE_HIWAT },
 269         { &pf_frent_pl, PFFRAG_FRENT_HIWAT },
 270         { &pfr_ktable_pl, PFR_KTABLE_HIWAT },
 271         { &pfr_kentry_pl, PFR_KENTRY_HIWAT }
 272 };
 273
 274 #define STATE_LOOKUP(i, k, d, s, m)                                     \
 275         do {                                                            \
 276                 s = pf_find_state(i, k, d, m);                  \
 277                 if (s == NULL || (s)->timeout == PFTM_PURGE)            \
 278                         return (PF_DROP);                               \
 279                 if (d == PF_OUT &&                                      \
 280                     (((s)->rule.ptr->rt == PF_ROUTETO &&                \
 281                     (s)->rule.ptr->direction == PF_OUT) ||              \
 282                     ((s)->rule.ptr->rt == PF_REPLYTO &&                 \
 283                     (s)->rule.ptr->direction == PF_IN)) &&              \
 284                     (s)->rt_kif != NULL &&                              \
 285                     (s)->rt_kif != i)                                   \
 286                         return (PF_PASS);                               \
 287         } while (0)
 288
 289 #define BOUND_IFACE(r, k) \
 290         ((r)->rule_flag & PFRULE_IFBOUND) ? (k) : pfi_all
 291
 292 #define STATE_INC_COUNTERS(s)                           \
 293         do {                                            \
 294                 s->rule.ptr->states_cur++;              \
 295                 s->rule.ptr->states_tot++;              \
 296                 if (s->anchor.ptr != NULL) {            \
 297                         s->anchor.ptr->states_cur++;    \
 298                         s->anchor.ptr->states_tot++;    \
 299                 }                                       \
 300                 if (s->nat_rule.ptr != NULL) {          \
 301                         s->nat_rule.ptr->states_cur++;  \
 302                         s->nat_rule.ptr->states_tot++;  \
 303                 }                                       \
 304         } while (0)
 305
 306 #define STATE_DEC_COUNTERS(s)                           \
 307         do {                                            \
 308                 if (s->nat_rule.ptr != NULL)            \
 309                         s->nat_rule.ptr->states_cur--;  \
 310                 if (s->anchor.ptr != NULL)              \
 311                         s->anchor.ptr->states_cur--;    \
 312                 s->rule.ptr->states_cur--;              \
 313         } while (0)
 314
 315 static __inline int pf_src_compare(struct pf_src_node *, struct pf_src_node *);
 316 static __inline int pf_state_compare_key(struct pf_state_key *,
 317         struct pf_state_key *);
 318 static __inline int pf_state_compare_id(struct pf_state *,
 319         struct pf_state *);
 320
 321 struct pf_src_tree tree_src_tracking;
 322
 323 struct pf_state_tree_id tree_id;
 324 struct pf_state_queue state_list;
 325
 326 RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare);
 327 RB_GENERATE(pf_state_tree, pf_state_key, entry, pf_state_compare_key);
 328 RB_GENERATE(pf_state_tree_id, pf_state,
 329     entry_id, pf_state_compare_id);
 330
 331 static __inline int
 332 pf_src_compare(struct pf_src_node *a, struct pf_src_node *b)
 333 {
 334         int     diff;
 335
 336         if (a->rule.ptr > b->rule.ptr)
 337                 return (1);
 338         if (a->rule.ptr < b->rule.ptr)
 339                 return (-1);
 340         if ((diff = a->af - b->af) != 0)
 341                 return (diff);
 342         switch (a->af) {
 343 #ifdef INET
 344         case AF_INET:
 345                 if (a->addr.addr32[0] > b->addr.addr32[0])
 346                         return (1);
 347                 if (a->addr.addr32[0] < b->addr.addr32[0])
 348                         return (-1);
 349                 break;
 350 #endif /* INET */
 351 #ifdef INET6
 352         case AF_INET6:
 353                 if (a->addr.addr32[3] > b->addr.addr32[3])
 354                         return (1);
 355                 if (a->addr.addr32[3] < b->addr.addr32[3])
 356                         return (-1);
 357                 if (a->addr.addr32[2] > b->addr.addr32[2])
 358                         return (1);
 359                 if (a->addr.addr32[2] < b->addr.addr32[2])
 360                         return (-1);
 361                 if (a->addr.addr32[1] > b->addr.addr32[1])
 362                         return (1);
 363                 if (a->addr.addr32[1] < b->addr.addr32[1])
 364                         return (-1);
 365                 if (a->addr.addr32[0] > b->addr.addr32[0])
 366                         return (1);
 367                 if (a->addr.addr32[0] < b->addr.addr32[0])
 368                         return (-1);
 369                 break;
 370 #endif /* INET6 */
 371         }
 372         return (0);
 373 }
 374
 375 u_int32_t
 376 pf_state_hash(struct pf_state_key *sk)
 377 {
 378         u_int32_t hv = (u_int32_t)(((intptr_t)sk >> 6) ^ ((intptr_t)sk >> 15));
 379         if (hv == 0)    /* disallow 0 */
 380                 hv = 1;
 381         return(hv);
 382 }
 383
 384 #ifdef INET6
 385 void
 386 pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af)
 387 {
 388         switch (af) {
 389 #ifdef INET
 390         case AF_INET:
 391                 dst->addr32[0] = src->addr32[0];
 392                 break;
 393 #endif /* INET */
 394         case AF_INET6:
 395                 dst->addr32[0] = src->addr32[0];
 396                 dst->addr32[1] = src->addr32[1];
 397                 dst->addr32[2] = src->addr32[2];
 398                 dst->addr32[3] = src->addr32[3];
 399                 break;
 400         }
 401 }
 402 #endif /* INET6 */
 403
 404 void
 405 pf_init_threshold(struct pf_threshold *threshold,
 406     u_int32_t limit, u_int32_t seconds)
 407 {
 408         threshold->limit = limit * PF_THRESHOLD_MULT;
 409         threshold->seconds = seconds;
 410         threshold->count = 0;
 411         threshold->last = time_second;
 412 }
 413
 414 void
 415 pf_add_threshold(struct pf_threshold *threshold)
 416 {
 417         u_int32_t t = time_second, diff = t - threshold->last;
 418
 419         if (diff >= threshold->seconds)
 420                 threshold->count = 0;
 421         else
 422                 threshold->count -= threshold->count * diff /
 423                     threshold->seconds;
 424         threshold->count += PF_THRESHOLD_MULT;
 425         threshold->last = t;
 426 }
 427
 428 int
 429 pf_check_threshold(struct pf_threshold *threshold)
 430 {
 431         return (threshold->count > threshold->limit);
 432 }
 433
 434 int
 435 pf_src_connlimit(struct pf_state **state)
 436 {
 437         int bad = 0;
 438
 439         (*state)->src_node->conn++;
 440         (*state)->src.tcp_est = 1;
 441         pf_add_threshold(&(*state)->src_node->conn_rate);
 442
 443         if ((*state)->rule.ptr->max_src_conn &&
 444             (*state)->rule.ptr->max_src_conn <
 445             (*state)->src_node->conn) {
 446                 pf_status.lcounters[LCNT_SRCCONN]++;
 447                 bad++;
 448         }
 449
 450         if ((*state)->rule.ptr->max_src_conn_rate.limit &&
 451             pf_check_threshold(&(*state)->src_node->conn_rate)) {
 452                 pf_status.lcounters[LCNT_SRCCONNRATE]++;
 453                 bad++;
 454         }
 455
 456         if (!bad)
 457                 return (0);
 458
 459         if ((*state)->rule.ptr->overload_tbl) {
 460                 struct pfr_addr p;
 461                 u_int32_t       killed = 0;
 462
 463                 pf_status.lcounters[LCNT_OVERLOAD_TABLE]++;
 464                 if (pf_status.debug >= PF_DEBUG_MISC) {
 465                         kprintf("pf_src_connlimit: blocking address ");
 466                         pf_print_host(&(*state)->src_node->addr, 0,
 467                             (*state)->key[PF_SK_WIRE]->af);
 468                 }
 469
 470                 bzero(&p, sizeof(p));
 471                 p.pfra_af = (*state)->key[PF_SK_WIRE]->af;
 472                 switch ((*state)->key[PF_SK_WIRE]->af) {
 473 #ifdef INET
 474                 case AF_INET:
 475                         p.pfra_net = 32;
 476                         p.pfra_ip4addr = (*state)->src_node->addr.v4;
 477                         break;
 478 #endif /* INET */
 479 #ifdef INET6
 480                 case AF_INET6:
 481                         p.pfra_net = 128;
 482                         p.pfra_ip6addr = (*state)->src_node->addr.v6;
 483                         break;
 484 #endif /* INET6 */
 485                 }
 486
 487                 pfr_insert_kentry((*state)->rule.ptr->overload_tbl,
 488                     &p, time_second);
 489
 490                 /* kill existing states if that's required. */
 491                 if ((*state)->rule.ptr->flush) {
 492                         struct pf_state_key *sk;
 493                         struct pf_state *st;
 494
 495                         pf_status.lcounters[LCNT_OVERLOAD_FLUSH]++;
 496                         RB_FOREACH(st, pf_state_tree_id, &tree_id) {
 497                                 sk = st->key[PF_SK_WIRE];
 498                                 /*
 499                                  * Kill states from this source.  (Only those
 500                                  * from the same rule if PF_FLUSH_GLOBAL is not
 501                                  * set)
 502                                  */
 503                                 if (sk->af ==
 504                                     (*state)->key[PF_SK_WIRE]->af &&
 505                                     (((*state)->direction == PF_OUT &&
 506                                     PF_AEQ(&(*state)->src_node->addr,
 507                                         &sk->addr[0], sk->af)) ||
 508                                     ((*state)->direction == PF_IN &&
 509                                     PF_AEQ(&(*state)->src_node->addr,
 510                                         &sk->addr[1], sk->af))) &&
 511                                     ((*state)->rule.ptr->flush &
 512                                     PF_FLUSH_GLOBAL ||
 513                                     (*state)->rule.ptr == st->rule.ptr)) {
 514                                         st->timeout = PFTM_PURGE;
 515                                         st->src.state = st->dst.state =
 516                                             TCPS_CLOSED;
 517                                         killed++;
 518                                 }
 519                         }
 520                         if (pf_status.debug >= PF_DEBUG_MISC)
 521                                 kprintf(", %u states killed", killed);
 522                 }
 523                 if (pf_status.debug >= PF_DEBUG_MISC)
 524                         kprintf("\n");
 525         }
 526
 527         /* kill this state */
 528         (*state)->timeout = PFTM_PURGE;
 529         (*state)->src.state = (*state)->dst.state = TCPS_CLOSED;
 530         return (1);
 531 }
 532
 533 int
 534 pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule,
 535     struct pf_addr *src, sa_family_t af)
 536 {
 537         struct pf_src_node      k;
 538
 539         if (*sn == NULL) {
 540                 k.af = af;
 541                 PF_ACPY(&k.addr, src, af);
 542                 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
 543                     rule->rpool.opts & PF_POOL_STICKYADDR)
 544                         k.rule.ptr = rule;
 545                 else
 546                         k.rule.ptr = NULL;
 547                 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
 548                 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
 549         }
 550         if (*sn == NULL) {
 551                 if (!rule->max_src_nodes ||
 552                     rule->src_nodes < rule->max_src_nodes)
 553                         (*sn) = pool_get(&pf_src_tree_pl, PR_NOWAIT | PR_ZERO);
 554                 else
 555                         pf_status.lcounters[LCNT_SRCNODES]++;
 556                 if ((*sn) == NULL)
 557                         return (-1);
 558
 559                 pf_init_threshold(&(*sn)->conn_rate,
 560                     rule->max_src_conn_rate.limit,
 561                     rule->max_src_conn_rate.seconds);
 562
 563                 (*sn)->af = af;
 564                 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
 565                     rule->rpool.opts & PF_POOL_STICKYADDR)
 566                         (*sn)->rule.ptr = rule;
 567                 else
 568                         (*sn)->rule.ptr = NULL;
 569                 PF_ACPY(&(*sn)->addr, src, af);
 570                 if (RB_INSERT(pf_src_tree,
 571                     &tree_src_tracking, *sn) != NULL) {
 572                         if (pf_status.debug >= PF_DEBUG_MISC) {
 573                                 kprintf("pf: src_tree insert failed: ");
 574                                 pf_print_host(&(*sn)->addr, 0, af);
 575                                 kprintf("\n");
 576                         }
 577                         pool_put(&pf_src_tree_pl, *sn);
 578                         return (-1);
 579                 }
 580                 (*sn)->creation = time_second;
 581                 (*sn)->ruletype = rule->action;
 582                 if ((*sn)->rule.ptr != NULL)
 583                         (*sn)->rule.ptr->src_nodes++;
 584                 pf_status.scounters[SCNT_SRC_NODE_INSERT]++;
 585                 pf_status.src_nodes++;
 586         } else {
 587                 if (rule->max_src_states &&
 588                     (*sn)->states >= rule->max_src_states) {
 589                         pf_status.lcounters[LCNT_SRCSTATES]++;
 590                         return (-1);
 591                 }
 592         }
 593         return (0);
 594 }
 595
 596 /* state table stuff */
 597
 598 static __inline int
 599 pf_state_compare_key(struct pf_state_key *a, struct pf_state_key *b)
 600 {
 601         int     diff;
 602
 603         if ((diff = a->proto - b->proto) != 0)
 604                 return (diff);
 605         if ((diff = a->af - b->af) != 0)
 606                 return (diff);
 607         switch (a->af) {
 608 #ifdef INET
 609         case AF_INET:
 610                 if (a->addr[0].addr32[0] > b->addr[0].addr32[0])
 611                         return (1);
 612                 if (a->addr[0].addr32[0] < b->addr[0].addr32[0])
 613                         return (-1);
 614                 if (a->addr[1].addr32[0] > b->addr[1].addr32[0])
 615                         return (1);
 616                 if (a->addr[1].addr32[0] < b->addr[1].addr32[0])
 617                         return (-1);
 618                 break;
 619 #endif /* INET */
 620 #ifdef INET6
 621         case AF_INET6:
 622                 if (a->addr[0].addr32[3] > b->addr[0].addr32[3])
 623                         return (1);
 624                 if (a->addr[0].addr32[3] < b->addr[0].addr32[3])
 625                         return (-1);
 626                 if (a->addr[1].addr32[3] > b->addr[1].addr32[3])
 627                         return (1);
 628                 if (a->addr[1].addr32[3] < b->addr[1].addr32[3])
 629                         return (-1);
 630                 if (a->addr[0].addr32[2] > b->addr[0].addr32[2])
 631                         return (1);
 632                 if (a->addr[0].addr32[2] < b->addr[0].addr32[2])
 633                         return (-1);
 634                 if (a->addr[1].addr32[2] > b->addr[1].addr32[2])
 635                         return (1);
 636                 if (a->addr[1].addr32[2] < b->addr[1].addr32[2])
 637                         return (-1);
 638                 if (a->addr[0].addr32[1] > b->addr[0].addr32[1])
 639                         return (1);
 640                 if (a->addr[0].addr32[1] < b->addr[0].addr32[1])
 641                         return (-1);
 642                 if (a->addr[1].addr32[1] > b->addr[1].addr32[1])
 643                         return (1);
 644                 if (a->addr[1].addr32[1] < b->addr[1].addr32[1])
 645                         return (-1);
 646                 if (a->addr[0].addr32[0] > b->addr[0].addr32[0])
 647                         return (1);
 648                 if (a->addr[0].addr32[0] < b->addr[0].addr32[0])
 649                         return (-1);
 650                 if (a->addr[1].addr32[0] > b->addr[1].addr32[0])
 651                         return (1);
 652                 if (a->addr[1].addr32[0] < b->addr[1].addr32[0])
 653                         return (-1);
 654                 break;
 655 #endif /* INET6 */
 656         }
 657
 658         if ((diff = a->port[0] - b->port[0]) != 0)
 659                 return (diff);
 660         if ((diff = a->port[1] - b->port[1]) != 0)
 661                 return (diff);
 662
 663         return (0);
 664 }
 665
 666 static __inline int
 667 pf_state_compare_id(struct pf_state *a, struct pf_state *b)
 668 {
 669         if (a->id > b->id)
 670                 return (1);
 671         if (a->id < b->id)
 672                 return (-1);
 673         if (a->creatorid > b->creatorid)
 674                 return (1);
 675         if (a->creatorid < b->creatorid)
 676                 return (-1);
 677
 678         return (0);
 679 }
 680
 681 int
 682 pf_state_key_attach(struct pf_state_key *sk, struct pf_state *s, int idx)
 683 {
 684         struct pf_state_item    *si;
 685         struct pf_state_key     *cur;
 686
 687         KKASSERT(s->key[idx] == NULL);  /* XXX handle this? */
 688
 689         if ((cur = RB_INSERT(pf_state_tree, &pf_statetbl, sk)) != NULL) {
 690                 /* key exists. check for same kif, if none, add to key */
 691                 TAILQ_FOREACH(si, &cur->states, entry)
 692                         if (si->s->kif == s->kif &&
 693                             si->s->direction == s->direction) {
 694                                 if (pf_status.debug >= PF_DEBUG_MISC) {
 695                                         kprintf(
 696                                             "pf: %s key attach failed on %s: ",
 697                                             (idx == PF_SK_WIRE) ?
 698                                             "wire" : "stack",
 699                                             s->kif->pfik_name);
 700                                         pf_print_state_parts(s,
 701                                             (idx == PF_SK_WIRE) ? sk : NULL,
 702                                             (idx == PF_SK_STACK) ? sk : NULL);
 703                                         kprintf("\n");
 704                                 }
 705                                 pool_put(&pf_state_key_pl, sk);
 706                                 return (-1);    /* collision! */
 707                         }
 708                 pool_put(&pf_state_key_pl, sk);
 709                 s->key[idx] = cur;
 710         } else
 711                 s->key[idx] = sk;
 712
 713         if ((si = pool_get(&pf_state_item_pl, PR_NOWAIT)) == NULL) {
 714                 pf_state_key_detach(s, idx);
 715                 return (-1);
 716         }
 717         si->s = s;
 718
 719         /* list is sorted, if-bound states before floating */
 720         if (s->kif == pfi_all)
 721                 TAILQ_INSERT_TAIL(&s->key[idx]->states, si, entry);
 722         else
 723                 TAILQ_INSERT_HEAD(&s->key[idx]->states, si, entry);
 724         return (0);
 725 }
 726
 727 void
 728 pf_detach_state(struct pf_state *s)
 729 {
 730         if (s->key[PF_SK_WIRE] == s->key[PF_SK_STACK])
 731                 s->key[PF_SK_WIRE] = NULL;
 732
 733         if (s->key[PF_SK_STACK] != NULL)
 734                 pf_state_key_detach(s, PF_SK_STACK);
 735
 736         if (s->key[PF_SK_WIRE] != NULL)
 737                 pf_state_key_detach(s, PF_SK_WIRE);
 738 }
 739
 740 void
 741 pf_state_key_detach(struct pf_state *s, int idx)
 742 {
 743         struct pf_state_item    *si;
 744
 745         si = TAILQ_FIRST(&s->key[idx]->states);
 746         while (si && si->s != s)
 747             si = TAILQ_NEXT(si, entry);
 748
 749         if (si) {
 750                 TAILQ_REMOVE(&s->key[idx]->states, si, entry);
 751                 pool_put(&pf_state_item_pl, si);
 752         }
 753
 754         if (TAILQ_EMPTY(&s->key[idx]->states)) {
 755                 RB_REMOVE(pf_state_tree, &pf_statetbl, s->key[idx]);
 756                 if (s->key[idx]->reverse)
 757                         s->key[idx]->reverse->reverse = NULL;
 758                 if (s->key[idx]->inp)
 759                         s->key[idx]->inp->inp_pf_sk = NULL;
 760                 pool_put(&pf_state_key_pl, s->key[idx]);
 761         }
 762         s->key[idx] = NULL;
 763 }
 764
 765 struct pf_state_key *
 766 pf_alloc_state_key(int pool_flags)
 767 {
 768         struct pf_state_key     *sk;
 769
 770         if ((sk = pool_get(&pf_state_key_pl, pool_flags)) == NULL)
 771                 return (NULL);
 772         TAILQ_INIT(&sk->states);
 773
 774         return (sk);
 775 }
 776
 777 int
 778 pf_state_key_setup(struct pf_pdesc *pd, struct pf_rule *nr,
 779         struct pf_state_key **skw, struct pf_state_key **sks,
 780         struct pf_state_key **skp, struct pf_state_key **nkp,
 781         struct pf_addr *saddr, struct pf_addr *daddr,
 782         u_int16_t sport, u_int16_t dport)
 783 {
 784         KKASSERT((*skp == NULL && *nkp == NULL));
 785
 786         if ((*skp = pf_alloc_state_key(PR_NOWAIT | PR_ZERO)) == NULL)
 787                 return (ENOMEM);
 788
 789         PF_ACPY(&(*skp)->addr[pd->sidx], saddr, pd->af);
 790         PF_ACPY(&(*skp)->addr[pd->didx], daddr, pd->af);
 791         (*skp)->port[pd->sidx] = sport;
 792         (*skp)->port[pd->didx] = dport;
 793         (*skp)->proto = pd->proto;
 794         (*skp)->af = pd->af;
 795
 796         if (nr != NULL) {
 797                 if ((*nkp = pf_alloc_state_key(PR_NOWAIT | PR_ZERO)) == NULL)
 798                         return (ENOMEM); /* caller must handle cleanup */
 799
 800                 /* XXX maybe just bcopy and TAILQ_INIT(&(*nkp)->states) */
 801                 PF_ACPY(&(*nkp)->addr[0], &(*skp)->addr[0], pd->af);
 802                 PF_ACPY(&(*nkp)->addr[1], &(*skp)->addr[1], pd->af);
 803                 (*nkp)->port[0] = (*skp)->port[0];
 804                 (*nkp)->port[1] = (*skp)->port[1];
 805                 (*nkp)->proto = pd->proto;
 806                 (*nkp)->af = pd->af;
 807         } else
 808                 *nkp = *skp;
 809
 810         if (pd->dir == PF_IN) {
 811                 *skw = *skp;
 812                 *sks = *nkp;
 813         } else {
 814                 *sks = *skp;
 815                 *skw = *nkp;
 816         }
 817         return (0);
 818 }
 819
 820
 821 int
 822 pf_state_insert(struct pfi_kif *kif, struct pf_state_key *skw,
 823     struct pf_state_key *sks, struct pf_state *s)
 824 {
 825         s->kif = kif;
 826
 827         if (skw == sks) {
 828                 if (pf_state_key_attach(skw, s, PF_SK_WIRE))
 829                         return (-1);
 830                 s->key[PF_SK_STACK] = s->key[PF_SK_WIRE];
 831         } else {
 832                 if (pf_state_key_attach(skw, s, PF_SK_WIRE)) {
 833                         pool_put(&pf_state_key_pl, sks);
 834                         return (-1);
 835                 }
 836                 if (pf_state_key_attach(sks, s, PF_SK_STACK)) {
 837                         pf_state_key_detach(s, PF_SK_WIRE);
 838                         return (-1);
 839                 }
 840         }
 841
 842         if (s->id == 0 && s->creatorid == 0) {
 843                 s->id = htobe64(pf_status.stateid++);
 844                 s->creatorid = pf_status.hostid;
 845         }
 846         if (RB_INSERT(pf_state_tree_id, &tree_id, s) != NULL) {
 847                 if (pf_status.debug >= PF_DEBUG_MISC) {
 848                         kprintf("pf: state insert failed: "
 849                             "id: %016jx creatorid: %08x",
 850                               (uintmax_t)be64toh(s->id), ntohl(s->creatorid));
 851                         if (s->sync_flags & PFSTATE_FROMSYNC)
 852                                 kprintf(" (from sync)");
 853                         kprintf("\n");
 854                 }
 855                 pf_detach_state(s);
 856                 return (-1);
 857         }
 858         TAILQ_INSERT_TAIL(&state_list, s, entry_list);
 859         pf_status.fcounters[FCNT_STATE_INSERT]++;
 860         pf_status.states++;
 861         pfi_kif_ref(kif, PFI_KIF_REF_STATE);
 862         pfsync_insert_state(s);
 863         return (0);
 864 }
 865
 866 struct pf_state *
 867 pf_find_state_byid(struct pf_state_cmp *key)
 868 {
 869         pf_status.fcounters[FCNT_STATE_SEARCH]++;
 870
 871         return (RB_FIND(pf_state_tree_id, &tree_id, (struct pf_state *)key));
 872 }
 873
 874 struct pf_state *
 875 pf_find_state(struct pfi_kif *kif, struct pf_state_key_cmp *key, u_int dir,
 876     struct mbuf *m)
 877 {
 878         struct pf_state_key     *sk;
 879         struct pf_state_item    *si;
 880
 881         pf_status.fcounters[FCNT_STATE_SEARCH]++;
 882
 883         if (dir == PF_OUT && m->m_pkthdr.pf.statekey &&
 884             ((struct pf_state_key *)m->m_pkthdr.pf.statekey)->reverse)
 885                 sk = ((struct pf_state_key *)m->m_pkthdr.pf.statekey)->reverse;
 886         else {
 887                 if ((sk = RB_FIND(pf_state_tree, &pf_statetbl,
 888                     (struct pf_state_key *)key)) == NULL)
 889                         return (NULL);
 890                 if (dir == PF_OUT && m->m_pkthdr.pf.statekey) {
 891                         ((struct pf_state_key *)
 892                             m->m_pkthdr.pf.statekey)->reverse = sk;
 893                         sk->reverse = m->m_pkthdr.pf.statekey;
 894                 }
 895         }
 896
 897         if (dir == PF_OUT)
 898                 m->m_pkthdr.pf.statekey = NULL;
 899
 900         /* list is sorted, if-bound states before floating ones */
 901         TAILQ_FOREACH(si, &sk->states, entry)
 902                 if ((si->s->kif == pfi_all || si->s->kif == kif) &&
 903                     sk == (dir == PF_IN ? si->s->key[PF_SK_WIRE] :
 904                     si->s->key[PF_SK_STACK]))
 905                         return (si->s);
 906
 907         return (NULL);
 908 }
 909
 910 struct pf_state *
 911 pf_find_state_all(struct pf_state_key_cmp *key, u_int dir, int *more)
 912 {
 913         struct pf_state_key     *sk;
 914         struct pf_state_item    *si, *ret = NULL;
 915
 916         pf_status.fcounters[FCNT_STATE_SEARCH]++;
 917
 918         sk = RB_FIND(pf_state_tree, &pf_statetbl, (struct pf_state_key *)key);
 919
 920         if (sk != NULL) {
 921                 TAILQ_FOREACH(si, &sk->states, entry)
 922                         if (dir == PF_INOUT ||
 923                             (sk == (dir == PF_IN ? si->s->key[PF_SK_WIRE] :
 924                             si->s->key[PF_SK_STACK]))) {
 925                                 if (more == NULL)
 926                                         return (si->s);
 927
 928                                 if (ret)
 929                                         (*more)++;
 930                                 else
 931                                         ret = si;
 932                         }
 933         }
 934         return (ret ? ret->s : NULL);
 935 }
 936
 937 /* END state table stuff */
 938
 939
 940 void
 941 pf_purge_thread(void *v)
 942 {
 943         int nloops = 0;
 944         int locked = 0;
 945
 946         lwkt_gettoken(&pf_token);
 947         for (;;) {
 948                 tsleep(pf_purge_thread, PWAIT, "pftm", 1 * hz);
 949
 950                 lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
 951
 952                 if (pf_end_threads) {
 953                         pf_purge_expired_states(pf_status.states, 1);
 954                         pf_purge_expired_fragments();
 955                         pf_purge_expired_src_nodes(1);
 956                         pf_end_threads++;
 957
 958                         lockmgr(&pf_consistency_lock, LK_RELEASE);
 959                         wakeup(pf_purge_thread);
 960                         kthread_exit();
 961                 }
 962                 crit_enter();
 963
 964                 /* process a fraction of the state table every second */
 965                 if(!pf_purge_expired_states(1 + (pf_status.states
 966                     / pf_default_rule.timeout[PFTM_INTERVAL]), 0)) {
 967
 968                         pf_purge_expired_states(1 + (pf_status.states
 969                             / pf_default_rule.timeout[PFTM_INTERVAL]), 1);
 970                 }
 971
 972                 /* purge other expired types every PFTM_INTERVAL seconds */
 973                 if (++nloops >= pf_default_rule.timeout[PFTM_INTERVAL]) {
 974                         pf_purge_expired_fragments();
 975                         if (!pf_purge_expired_src_nodes(locked)) {
 976                                 pf_purge_expired_src_nodes(1);
 977                         }
 978                         nloops = 0;
 979                 }
 980                 crit_exit();
 981                 lockmgr(&pf_consistency_lock, LK_RELEASE);
 982         }
 983         lwkt_reltoken(&pf_token);
 984 }
 985
 986 u_int32_t
 987 pf_state_expires(const struct pf_state *state)
 988 {
 989         u_int32_t       timeout;
 990         u_int32_t       start;
 991         u_int32_t       end;
 992         u_int32_t       states;
 993
 994         /* handle all PFTM_* > PFTM_MAX here */
 995         if (state->timeout == PFTM_PURGE)
 996                 return (time_second);
 997         if (state->timeout == PFTM_UNTIL_PACKET)
 998                 return (0);
 999         KKASSERT(state->timeout != PFTM_UNLINKED);
1000         KKASSERT(state->timeout < PFTM_MAX);
1001         timeout = state->rule.ptr->timeout[state->timeout];
1002         if (!timeout)
1003                 timeout = pf_default_rule.timeout[state->timeout];
1004         start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START];
1005         if (start) {
1006                 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END];
1007                 states = state->rule.ptr->states_cur;
1008         } else {
1009                 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START];
1010                 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END];
1011                 states = pf_status.states;
1012         }
1013         if (end && states > start && start < end) {
1014                 if (states < end)
1015                         return (state->expire + timeout * (end - states) /
1016                             (end - start));
1017                 else
1018                         return (time_second);
1019         }
1020         return (state->expire + timeout);
1021 }
1022
1023 int
1024 pf_purge_expired_src_nodes(int waslocked)
1025 {
1026          struct pf_src_node             *cur, *next;
1027          int                             locked = waslocked;
1028
1029          for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) {
1030                  next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur);
1031
1032                  if (cur->states <= 0 && cur->expire <= time_second) {
1033                          if (! locked) {
1034                                  lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
1035                                  next = RB_NEXT(pf_src_tree,
1036                                      &tree_src_tracking, cur);
1037                                  locked = 1;
1038                          }
1039                          if (cur->rule.ptr != NULL) {
1040                                  cur->rule.ptr->src_nodes--;
1041                                  if (cur->rule.ptr->states_cur <= 0 &&
1042                                      cur->rule.ptr->max_src_nodes <= 0)
1043                                          pf_rm_rule(NULL, cur->rule.ptr);
1044                          }
1045                          RB_REMOVE(pf_src_tree, &tree_src_tracking, cur);
1046                          pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
1047                          pf_status.src_nodes--;
1048                          pool_put(&pf_src_tree_pl, cur);
1049                  }
1050          }
1051
1052          if (locked && !waslocked)
1053                 lockmgr(&pf_consistency_lock, LK_RELEASE);
1054         return(1);
1055 }
1056
1057 void
1058 pf_src_tree_remove_state(struct pf_state *s)
1059 {
1060         u_int32_t timeout;
1061
1062         if (s->src_node != NULL) {
1063                 if (s->src.tcp_est)
1064                         --s->src_node->conn;
1065                 if (--s->src_node->states <= 0) {
1066                         timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1067                         if (!timeout)
1068                                 timeout =
1069                                     pf_default_rule.timeout[PFTM_SRC_NODE];
1070                         s->src_node->expire = time_second + timeout;
1071                 }
1072         }
1073         if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) {
1074                 if (--s->nat_src_node->states <= 0) {
1075                         timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1076                         if (!timeout)
1077                                 timeout =
1078                                     pf_default_rule.timeout[PFTM_SRC_NODE];
1079                         s->nat_src_node->expire = time_second + timeout;
1080                 }
1081         }
1082         s->src_node = s->nat_src_node = NULL;
1083 }
1084
1085 /* callers should be at crit_enter() */
1086 void
1087 pf_unlink_state(struct pf_state *cur)
1088 {
1089         if (cur->src.state == PF_TCPS_PROXY_DST) {
1090                 /* XXX wire key the right one? */
1091                 pf_send_tcp(cur->rule.ptr, cur->key[PF_SK_WIRE]->af,
1092                     &cur->key[PF_SK_WIRE]->addr[1],
1093                     &cur->key[PF_SK_WIRE]->addr[0],
1094                     cur->key[PF_SK_WIRE]->port[1],
1095                     cur->key[PF_SK_WIRE]->port[0],
1096                     cur->src.seqhi, cur->src.seqlo + 1,
1097                     TH_RST|TH_ACK, 0, 0, 0, 1, cur->tag, NULL, NULL);
1098         }
1099         RB_REMOVE(pf_state_tree_id, &tree_id, cur);
1100         if (cur->creatorid == pf_status.hostid)
1101                 pfsync_delete_state(cur);
1102         cur->timeout = PFTM_UNLINKED;
1103         pf_src_tree_remove_state(cur);
1104         pf_detach_state(cur);
1105 }
1106
1107 static struct pf_state  *purge_cur;
1108
1109 /* callers should be at crit_enter() and hold the
1110  * write_lock on pf_consistency_lock */
1111 void
1112 pf_free_state(struct pf_state *cur)
1113 {
1114         if (pfsyncif != NULL &&
1115             (pfsyncif->sc_bulk_send_next == cur ||
1116             pfsyncif->sc_bulk_terminator == cur))
1117                 return;
1118         KKASSERT(cur->timeout == PFTM_UNLINKED);
1119         if (--cur->rule.ptr->states_cur <= 0 &&
1120             cur->rule.ptr->src_nodes <= 0)
1121                 pf_rm_rule(NULL, cur->rule.ptr);
1122         if (cur->nat_rule.ptr != NULL)
1123                 if (--cur->nat_rule.ptr->states_cur <= 0 &&
1124                         cur->nat_rule.ptr->src_nodes <= 0)
1125                         pf_rm_rule(NULL, cur->nat_rule.ptr);
1126         if (cur->anchor.ptr != NULL)
1127                 if (--cur->anchor.ptr->states_cur <= 0)
1128                         pf_rm_rule(NULL, cur->anchor.ptr);
1129         pf_normalize_tcp_cleanup(cur);
1130         pfi_kif_unref(cur->kif, PFI_KIF_REF_STATE);
1131
1132         /*
1133          * We may be freeing pf_purge_expired_states()'s saved scan entry,
1134          * adjust it if necessary.
1135          */
1136         if (purge_cur == cur) {
1137                 kprintf("PURGE CONFLICT\n");
1138                 purge_cur = TAILQ_NEXT(purge_cur, entry_list);
1139         }
1140         TAILQ_REMOVE(&state_list, cur, entry_list);
1141         if (cur->tag)
1142                 pf_tag_unref(cur->tag);
1143         pool_put(&pf_state_pl, cur);
1144         pf_status.fcounters[FCNT_STATE_REMOVALS]++;
1145         pf_status.states--;
1146 }
1147
1148 int
1149 pf_purge_expired_states(u_int32_t maxcheck, int waslocked)
1150 {
1151         struct pf_state         *cur;
1152         int                      locked = waslocked;
1153
1154         while (maxcheck--) {
1155                 /*
1156                  * Wrap to start of list when we hit the end
1157                  */
1158                 cur = purge_cur;
1159                 if (cur == NULL) {
1160                         cur = TAILQ_FIRST(&state_list);
1161                         if (cur == NULL)
1162                                 break;  /* list empty */
1163                 }
1164
1165                 /*
1166                  * Setup next (purge_cur) while we process this one.  If we block and
1167                  * something else deletes purge_cur, pf_free_state() will adjust it further
1168                  * ahead.
1169                  */
1170                 purge_cur = TAILQ_NEXT(cur, entry_list);
1171
1172                 if (cur->timeout == PFTM_UNLINKED) {
1173                         /* free unlinked state */
1174                         if (! locked) {
1175                                 lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
1176                                 locked = 1;
1177                         }
1178                         pf_free_state(cur);
1179                 } else if (pf_state_expires(cur) <= time_second) {
1180                         /* unlink and free expired state */
1181                         pf_unlink_state(cur);
1182                         if (! locked) {
1183                                 if (!lockmgr(&pf_consistency_lock, LK_EXCLUSIVE))
1184                                         return (0);
1185                                 locked = 1;
1186                         }
1187                         pf_free_state(cur);
1188                 }
1189         }
1190
1191         if (locked)
1192                 lockmgr(&pf_consistency_lock, LK_RELEASE);
1193         return (1);
1194 }
1195
1196 int
1197 pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw)
1198 {
1199         if (aw->type != PF_ADDR_TABLE)
1200                 return (0);
1201         if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL)
1202                 return (1);
1203         return (0);
1204 }
1205
1206 void
1207 pf_tbladdr_remove(struct pf_addr_wrap *aw)
1208 {
1209         if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL)
1210                 return;
1211         pfr_detach_table(aw->p.tbl);
1212         aw->p.tbl = NULL;
1213 }
1214
1215 void
1216 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
1217 {
1218         struct pfr_ktable *kt = aw->p.tbl;
1219
1220         if (aw->type != PF_ADDR_TABLE || kt == NULL)
1221                 return;
1222         if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
1223                 kt = kt->pfrkt_root;
1224         aw->p.tbl = NULL;
1225         aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
1226                 kt->pfrkt_cnt : -1;
1227 }
1228
1229 void
1230 pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af)
1231 {
1232         switch (af) {
1233 #ifdef INET
1234         case AF_INET: {
1235                 u_int32_t a = ntohl(addr->addr32[0]);
1236                 kprintf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255,
1237                     (a>>8)&255, a&255);
1238                 if (p) {
1239                         p = ntohs(p);
1240                         kprintf(":%u", p);
1241                 }
1242                 break;
1243         }
1244 #endif /* INET */
1245 #ifdef INET6
1246         case AF_INET6: {
1247                 u_int16_t b;
1248                 u_int8_t i, curstart = 255, curend = 0,
1249                     maxstart = 0, maxend = 0;
1250                 for (i = 0; i < 8; i++) {
1251                         if (!addr->addr16[i]) {
1252                                 if (curstart == 255)
1253                                         curstart = i;
1254                                 else
1255                                         curend = i;
1256                         } else {
1257                                 if (curstart) {
1258                                         if ((curend - curstart) >
1259                                             (maxend - maxstart)) {
1260                                                 maxstart = curstart;
1261                                                 maxend = curend;
1262                                                 curstart = 255;
1263                                         }
1264                                 }
1265                         }
1266                 }
1267                 for (i = 0; i < 8; i++) {
1268                         if (i >= maxstart && i <= maxend) {
1269                                 if (maxend != 7) {
1270                                         if (i == maxstart)
1271                                                 kprintf(":");
1272                                 } else {
1273                                         if (i == maxend)
1274                                                 kprintf(":");
1275                                 }
1276                         } else {
1277                                 b = ntohs(addr->addr16[i]);
1278                                 kprintf("%x", b);
1279                                 if (i < 7)
1280                                         kprintf(":");
1281                         }
1282                 }
1283                 if (p) {
1284                         p = ntohs(p);
1285                         kprintf("[%u]", p);
1286                 }
1287                 break;
1288         }
1289 #endif /* INET6 */
1290         }
1291 }
1292
1293 void
1294 pf_print_state(struct pf_state *s)
1295 {
1296         pf_print_state_parts(s, NULL, NULL);
1297 }
1298
1299 void
1300 pf_print_state_parts(struct pf_state *s,
1301     struct pf_state_key *skwp, struct pf_state_key *sksp)
1302 {
1303         struct pf_state_key *skw, *sks;
1304         u_int8_t proto, dir;
1305
1306         /* Do our best to fill these, but they're skipped if NULL */
1307         skw = skwp ? skwp : (s ? s->key[PF_SK_WIRE] : NULL);
1308         sks = sksp ? sksp : (s ? s->key[PF_SK_STACK] : NULL);
1309         proto = skw ? skw->proto : (sks ? sks->proto : 0);
1310         dir = s ? s->direction : 0;
1311
1312         switch (proto) {
1313         case IPPROTO_TCP:
1314                 kprintf("TCP ");
1315                 break;
1316         case IPPROTO_UDP:
1317                 kprintf("UDP ");
1318                 break;
1319         case IPPROTO_ICMP:
1320                 kprintf("ICMP ");
1321                 break;
1322         case IPPROTO_ICMPV6:
1323                 kprintf("ICMPV6 ");
1324                 break;
1325         default:
1326                 kprintf("%u ", skw->proto);
1327                 break;
1328         }
1329         switch (dir) {
1330         case PF_IN:
1331                 kprintf(" in");
1332                 break;
1333         case PF_OUT:
1334                 kprintf(" out");
1335                 break;
1336         }
1337         if (skw) {
1338                 kprintf(" wire: ");
1339                 pf_print_host(&skw->addr[0], skw->port[0], skw->af);
1340                 kprintf(" ");
1341                 pf_print_host(&skw->addr[1], skw->port[1], skw->af);
1342         }
1343         if (sks) {
1344                 kprintf(" stack: ");
1345                 if (sks != skw) {
1346                         pf_print_host(&sks->addr[0], sks->port[0], sks->af);
1347                         kprintf(" ");
1348                         pf_print_host(&sks->addr[1], sks->port[1], sks->af);
1349                 } else
1350                         kprintf("-");
1351         }
1352         if (s) {
1353                 if (proto == IPPROTO_TCP) {
1354                         kprintf(" [lo=%u high=%u win=%u modulator=%u",
1355                             s->src.seqlo, s->src.seqhi,
1356                             s->src.max_win, s->src.seqdiff);
1357                         if (s->src.wscale && s->dst.wscale)
1358                                 kprintf(" wscale=%u",
1359                                     s->src.wscale & PF_WSCALE_MASK);
1360                         kprintf("]");
1361                         kprintf(" [lo=%u high=%u win=%u modulator=%u",
1362                             s->dst.seqlo, s->dst.seqhi,
1363                             s->dst.max_win, s->dst.seqdiff);
1364                         if (s->src.wscale && s->dst.wscale)
1365                                 kprintf(" wscale=%u",
1366                                 s->dst.wscale & PF_WSCALE_MASK);
1367                         kprintf("]");
1368                 }
1369                 kprintf(" %u:%u", s->src.state, s->dst.state);
1370         }
1371 }
1372
1373 void
1374 pf_print_flags(u_int8_t f)
1375 {
1376         if (f)
1377                 kprintf(" ");
1378         if (f & TH_FIN)
1379                 kprintf("F");
1380         if (f & TH_SYN)
1381                 kprintf("S");
1382         if (f & TH_RST)
1383                 kprintf("R");
1384         if (f & TH_PUSH)
1385                 kprintf("P");
1386         if (f & TH_ACK)
1387                 kprintf("A");
1388         if (f & TH_URG)
1389                 kprintf("U");
1390         if (f & TH_ECE)
1391                 kprintf("E");
1392         if (f & TH_CWR)
1393                 kprintf("W");
1394 }
1395
1396 #define PF_SET_SKIP_STEPS(i)                                    \
1397         do {                                                    \
1398                 while (head[i] != cur) {                        \
1399                         head[i]->skip[i].ptr = cur;             \
1400                         head[i] = TAILQ_NEXT(head[i], entries); \
1401                 }                                               \
1402         } while (0)
1403
1404 void
1405 pf_calc_skip_steps(struct pf_rulequeue *rules)
1406 {
1407         struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT];
1408         int i;
1409
1410         cur = TAILQ_FIRST(rules);
1411         prev = cur;
1412         for (i = 0; i < PF_SKIP_COUNT; ++i)
1413                 head[i] = cur;
1414         while (cur != NULL) {
1415
1416                 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
1417                         PF_SET_SKIP_STEPS(PF_SKIP_IFP);
1418                 if (cur->direction != prev->direction)
1419                         PF_SET_SKIP_STEPS(PF_SKIP_DIR);
1420                 if (cur->af != prev->af)
1421                         PF_SET_SKIP_STEPS(PF_SKIP_AF);
1422                 if (cur->proto != prev->proto)
1423                         PF_SET_SKIP_STEPS(PF_SKIP_PROTO);
1424                 if (cur->src.neg != prev->src.neg ||
1425                     pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr))
1426                         PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR);
1427                 if (cur->src.port[0] != prev->src.port[0] ||
1428                     cur->src.port[1] != prev->src.port[1] ||
1429                     cur->src.port_op != prev->src.port_op)
1430                         PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
1431                 if (cur->dst.neg != prev->dst.neg ||
1432                     pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr))
1433                         PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR);
1434                 if (cur->dst.port[0] != prev->dst.port[0] ||
1435                     cur->dst.port[1] != prev->dst.port[1] ||
1436                     cur->dst.port_op != prev->dst.port_op)
1437                         PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
1438
1439                 prev = cur;
1440                 cur = TAILQ_NEXT(cur, entries);
1441         }
1442         for (i = 0; i < PF_SKIP_COUNT; ++i)
1443                 PF_SET_SKIP_STEPS(i);
1444 }
1445
1446 int
1447 pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2)
1448 {
1449         if (aw1->type != aw2->type)
1450                 return (1);
1451         switch (aw1->type) {
1452         case PF_ADDR_ADDRMASK:
1453         case PF_ADDR_RANGE:
1454                 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0))
1455                         return (1);
1456                 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0))
1457                         return (1);
1458                 return (0);
1459         case PF_ADDR_DYNIFTL:
1460                 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt);
1461         case PF_ADDR_NOROUTE:
1462         case PF_ADDR_URPFFAILED:
1463                 return (0);
1464         case PF_ADDR_TABLE:
1465                 return (aw1->p.tbl != aw2->p.tbl);
1466         case PF_ADDR_RTLABEL:
1467                 return (aw1->v.rtlabel != aw2->v.rtlabel);
1468         default:
1469                 kprintf("invalid address type: %d\n", aw1->type);
1470                 return (1);
1471         }
1472 }
1473
1474 u_int16_t
1475 pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp)
1476 {
1477         u_int32_t       l;
1478
1479         if (udp && !cksum)
1480                 return (0x0000);
1481         l = cksum + old - new;
1482         l = (l >> 16) + (l & 65535);
1483         l = l & 65535;
1484         if (udp && !l)
1485                 return (0xFFFF);
1486         return (l);
1487 }
1488
1489 void
1490 pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc,
1491     struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af)
1492 {
1493         struct pf_addr  ao;
1494         u_int16_t       po = *p;
1495
1496         PF_ACPY(&ao, a, af);
1497         PF_ACPY(a, an, af);
1498
1499         *p = pn;
1500
1501         switch (af) {
1502 #ifdef INET
1503         case AF_INET:
1504                 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1505                     ao.addr16[0], an->addr16[0], 0),
1506                     ao.addr16[1], an->addr16[1], 0);
1507                 *p = pn;
1508                 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1509                     ao.addr16[0], an->addr16[0], u),
1510                     ao.addr16[1], an->addr16[1], u),
1511                     po, pn, u);
1512                 break;
1513 #endif /* INET */
1514 #ifdef INET6
1515         case AF_INET6:
1516                 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1517                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1518                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1519                     ao.addr16[0], an->addr16[0], u),
1520                     ao.addr16[1], an->addr16[1], u),
1521                     ao.addr16[2], an->addr16[2], u),
1522                     ao.addr16[3], an->addr16[3], u),
1523                     ao.addr16[4], an->addr16[4], u),
1524                     ao.addr16[5], an->addr16[5], u),
1525                     ao.addr16[6], an->addr16[6], u),
1526                     ao.addr16[7], an->addr16[7], u),
1527                     po, pn, u);
1528                 break;
1529 #endif /* INET6 */
1530         }
1531 }
1532
1533
1534 /* Changes a u_int32_t.  Uses a void * so there are no align restrictions */
1535 void
1536 pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u)
1537 {
1538         u_int32_t       ao;
1539
1540         memcpy(&ao, a, sizeof(ao));
1541         memcpy(a, &an, sizeof(u_int32_t));
1542         *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u),
1543             ao % 65536, an % 65536, u);
1544 }
1545
1546 #ifdef INET6
1547 void
1548 pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u)
1549 {
1550         struct pf_addr  ao;
1551
1552         PF_ACPY(&ao, a, AF_INET6);
1553         PF_ACPY(a, an, AF_INET6);
1554
1555         *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1556             pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1557             pf_cksum_fixup(pf_cksum_fixup(*c,
1558             ao.addr16[0], an->addr16[0], u),
1559             ao.addr16[1], an->addr16[1], u),
1560             ao.addr16[2], an->addr16[2], u),
1561             ao.addr16[3], an->addr16[3], u),
1562             ao.addr16[4], an->addr16[4], u),
1563             ao.addr16[5], an->addr16[5], u),
1564             ao.addr16[6], an->addr16[6], u),
1565             ao.addr16[7], an->addr16[7], u);
1566 }
1567 #endif /* INET6 */
1568
1569 void
1570 pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa,
1571     struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c,
1572     u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af)
1573 {
1574         struct pf_addr  oia, ooa;
1575
1576         PF_ACPY(&oia, ia, af);
1577         if (oa)
1578                 PF_ACPY(&ooa, oa, af);
1579
1580         /* Change inner protocol port, fix inner protocol checksum. */
1581         if (ip != NULL) {
1582                 u_int16_t       oip = *ip;
1583                 u_int32_t       opc = 0;
1584
1585                 if (pc != NULL)
1586                         opc = *pc;
1587                 *ip = np;
1588                 if (pc != NULL)
1589                         *pc = pf_cksum_fixup(*pc, oip, *ip, u);
1590                 *ic = pf_cksum_fixup(*ic, oip, *ip, 0);
1591                 if (pc != NULL)
1592                         *ic = pf_cksum_fixup(*ic, opc, *pc, 0);
1593         }
1594         /* Change inner ip address, fix inner ip and icmp checksums. */
1595         PF_ACPY(ia, na, af);
1596         switch (af) {
1597 #ifdef INET
1598         case AF_INET: {
1599                 u_int32_t        oh2c = *h2c;
1600
1601                 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c,
1602                     oia.addr16[0], ia->addr16[0], 0),
1603                     oia.addr16[1], ia->addr16[1], 0);
1604                 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1605                     oia.addr16[0], ia->addr16[0], 0),
1606                     oia.addr16[1], ia->addr16[1], 0);
1607                 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0);
1608                 break;
1609         }
1610 #endif /* INET */
1611 #ifdef INET6
1612         case AF_INET6:
1613                 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1614                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1615                     pf_cksum_fixup(pf_cksum_fixup(*ic,
1616                     oia.addr16[0], ia->addr16[0], u),
1617                     oia.addr16[1], ia->addr16[1], u),
1618                     oia.addr16[2], ia->addr16[2], u),
1619                     oia.addr16[3], ia->addr16[3], u),
1620                     oia.addr16[4], ia->addr16[4], u),
1621                     oia.addr16[5], ia->addr16[5], u),
1622                     oia.addr16[6], ia->addr16[6], u),
1623                     oia.addr16[7], ia->addr16[7], u);
1624                 break;
1625 #endif /* INET6 */
1626         }
1627         /* Outer ip address, fix outer ip or icmpv6 checksum, if necessary. */
1628         if (oa) {
1629                 PF_ACPY(oa, na, af);
1630                 switch (af) {
1631 #ifdef INET
1632                 case AF_INET:
1633                         *hc = pf_cksum_fixup(pf_cksum_fixup(*hc,
1634                             ooa.addr16[0], oa->addr16[0], 0),
1635                             ooa.addr16[1], oa->addr16[1], 0);
1636                         break;
1637 #endif /* INET */
1638 #ifdef INET6
1639                 case AF_INET6:
1640                         *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1641                             pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1642                             pf_cksum_fixup(pf_cksum_fixup(*ic,
1643                             ooa.addr16[0], oa->addr16[0], u),
1644                             ooa.addr16[1], oa->addr16[1], u),
1645                             ooa.addr16[2], oa->addr16[2], u),
1646                             ooa.addr16[3], oa->addr16[3], u),
1647                             ooa.addr16[4], oa->addr16[4], u),
1648                             ooa.addr16[5], oa->addr16[5], u),
1649                             ooa.addr16[6], oa->addr16[6], u),
1650                             ooa.addr16[7], oa->addr16[7], u);
1651                         break;
1652 #endif /* INET6 */
1653                 }
1654         }
1655 }
1656
1657
1658 /*
1659  * Need to modulate the sequence numbers in the TCP SACK option
1660  * (credits to Krzysztof Pfaff for report and patch)
1661  */
1662 int
1663 pf_modulate_sack(struct mbuf *m, int off, struct pf_pdesc *pd,
1664     struct tcphdr *th, struct pf_state_peer *dst)
1665 {
1666         int hlen = (th->th_off << 2) - sizeof(*th), thoptlen = hlen;
1667         u_int8_t opts[TCP_MAXOLEN], *opt = opts;
1668         int copyback = 0, i, olen;
1669         struct raw_sackblock sack;
1670
1671 #define TCPOLEN_SACKLEN (TCPOLEN_SACK + 2)
1672         if (hlen < TCPOLEN_SACKLEN ||
1673             !pf_pull_hdr(m, off + sizeof(*th), opts, hlen, NULL, NULL, pd->af))
1674                 return 0;
1675
1676         while (hlen >= TCPOLEN_SACKLEN) {
1677                 olen = opt[1];
1678                 switch (*opt) {
1679                 case TCPOPT_EOL:        /* FALLTHROUGH */
1680                 case TCPOPT_NOP:
1681                         opt++;
1682                         hlen--;
1683                         break;
1684                 case TCPOPT_SACK:
1685                         if (olen > hlen)
1686                                 olen = hlen;
1687                         if (olen >= TCPOLEN_SACKLEN) {
1688                                 for (i = 2; i + TCPOLEN_SACK <= olen;
1689                                     i += TCPOLEN_SACK) {
1690                                         memcpy(&sack, &opt[i], sizeof(sack));
1691                                         pf_change_a(&sack.rblk_start, &th->th_sum,
1692                                             htonl(ntohl(sack.rblk_start) -
1693                                             dst->seqdiff), 0);
1694                                         pf_change_a(&sack.rblk_end, &th->th_sum,
1695                                             htonl(ntohl(sack.rblk_end) -
1696                                             dst->seqdiff), 0);
1697                                         memcpy(&opt[i], &sack, sizeof(sack));
1698                                 }
1699                                 copyback = 1;
1700                         }
1701                         /* FALLTHROUGH */
1702                 default:
1703                         if (olen < 2)
1704                                 olen = 2;
1705                         hlen -= olen;
1706                         opt += olen;
1707                 }
1708         }
1709
1710         if (copyback)
1711                 m_copyback(m, off + sizeof(*th), thoptlen, opts);
1712         return (copyback);
1713 }
1714
1715 void
1716 pf_send_tcp(const struct pf_rule *r, sa_family_t af,
1717     const struct pf_addr *saddr, const struct pf_addr *daddr,
1718     u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack,
1719     u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl, int tag,
1720     u_int16_t rtag, struct ether_header *eh, struct ifnet *ifp)
1721 {
1722         struct mbuf     *m;
1723         int              len = 0, tlen;
1724 #ifdef INET
1725         struct ip       *h = NULL;
1726 #endif /* INET */
1727 #ifdef INET6
1728         struct ip6_hdr  *h6 = NULL;
1729 #endif /* INET6 */
1730         struct tcphdr   *th = NULL;
1731         char            *opt;
1732
1733         ASSERT_LWKT_TOKEN_HELD(&pf_token);
1734
1735         /* maximum segment size tcp option */
1736         tlen = sizeof(struct tcphdr);
1737         if (mss)
1738                 tlen += 4;
1739
1740         switch (af) {
1741 #ifdef INET
1742         case AF_INET:
1743                 len = sizeof(struct ip) + tlen;
1744                 break;
1745 #endif /* INET */
1746 #ifdef INET6
1747         case AF_INET6:
1748                 len = sizeof(struct ip6_hdr) + tlen;
1749                 break;
1750 #endif /* INET6 */
1751         }
1752
1753         /*
1754          * Create outgoing mbuf.
1755          *
1756          * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags,
1757          * so make sure pf.flags is clear.
1758          */
1759         m = m_gethdr(MB_DONTWAIT, MT_HEADER);
1760         if (m == NULL) {
1761                 return;
1762         }
1763         if (tag)
1764                 m->m_pkthdr.fw_flags |= PF_MBUF_TAGGED;
1765         m->m_pkthdr.pf.flags = 0;
1766         m->m_pkthdr.pf.tag = rtag;
1767         /* XXX Recheck when upgrading to > 4.4 */
1768         m->m_pkthdr.pf.statekey = NULL;
1769         if (r != NULL && r->rtableid >= 0)
1770                 m->m_pkthdr.pf.rtableid = r->rtableid;
1771
1772 #ifdef ALTQ
1773         if (r != NULL && r->qid) {
1774                 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
1775                 m->m_pkthdr.pf.qid = r->qid;
1776                 m->m_pkthdr.pf.ecn_af = af;
1777                 m->m_pkthdr.pf.hdr = mtod(m, struct ip *);
1778         }
1779 #endif /* ALTQ */
1780         m->m_data += max_linkhdr;
1781         m->m_pkthdr.len = m->m_len = len;
1782         m->m_pkthdr.rcvif = NULL;
1783         bzero(m->m_data, len);
1784         switch (af) {
1785 #ifdef INET
1786         case AF_INET:
1787                 h = mtod(m, struct ip *);
1788
1789                 /* IP header fields included in the TCP checksum */
1790                 h->ip_p = IPPROTO_TCP;
1791                 h->ip_len = tlen;
1792                 h->ip_src.s_addr = saddr->v4.s_addr;
1793                 h->ip_dst.s_addr = daddr->v4.s_addr;
1794
1795                 th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip));
1796                 break;
1797 #endif /* INET */
1798 #ifdef INET6
1799         case AF_INET6:
1800                 h6 = mtod(m, struct ip6_hdr *);
1801
1802                 /* IP header fields included in the TCP checksum */
1803                 h6->ip6_nxt = IPPROTO_TCP;
1804                 h6->ip6_plen = htons(tlen);
1805                 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr));
1806                 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr));
1807
1808                 th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr));
1809                 break;
1810 #endif /* INET6 */
1811         }
1812
1813         /* TCP header */
1814         th->th_sport = sport;
1815         th->th_dport = dport;
1816         th->th_seq = htonl(seq);
1817         th->th_ack = htonl(ack);
1818         th->th_off = tlen >> 2;
1819         th->th_flags = flags;
1820         th->th_win = htons(win);
1821
1822         if (mss) {
1823                 opt = (char *)(th + 1);
1824                 opt[0] = TCPOPT_MAXSEG;
1825                 opt[1] = 4;
1826                 mss = htons(mss);
1827                 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2);
1828         }
1829
1830         switch (af) {
1831 #ifdef INET
1832         case AF_INET:
1833                 /* TCP checksum */
1834                 th->th_sum = in_cksum(m, len);
1835
1836                 /* Finish the IP header */
1837                 h->ip_v = 4;
1838                 h->ip_hl = sizeof(*h) >> 2;
1839                 h->ip_tos = IPTOS_LOWDELAY;
1840                 h->ip_len = len;
1841                 h->ip_off = path_mtu_discovery ? IP_DF : 0;
1842                 h->ip_ttl = ttl ? ttl : ip_defttl;
1843                 h->ip_sum = 0;
1844                 if (eh == NULL) {
1845                         lwkt_reltoken(&pf_token);
1846                         ip_output(m, NULL, NULL, 0, NULL, NULL);
1847                         lwkt_gettoken(&pf_token);
1848                 } else {
1849                         struct route             ro;
1850                         struct rtentry           rt;
1851                         struct ether_header     *e = (void *)ro.ro_dst.sa_data;
1852
1853                         if (ifp == NULL) {
1854                                 m_freem(m);
1855                                 return;
1856                         }
1857                         rt.rt_ifp = ifp;
1858                         ro.ro_rt = &rt;
1859                         ro.ro_dst.sa_len = sizeof(ro.ro_dst);
1860                         ro.ro_dst.sa_family = pseudo_AF_HDRCMPLT;
1861                         bcopy(eh->ether_dhost, e->ether_shost, ETHER_ADDR_LEN);
1862                         bcopy(eh->ether_shost, e->ether_dhost, ETHER_ADDR_LEN);
1863                         e->ether_type = eh->ether_type;
1864                         /* XXX_IMPORT: later */
1865                         lwkt_reltoken(&pf_token);
1866                         ip_output(m, (void *)NULL, &ro, 0,
1867                             (void *)NULL, (void *)NULL);
1868                         lwkt_gettoken(&pf_token);
1869                 }
1870                 break;
1871 #endif /* INET */
1872 #ifdef INET6
1873         case AF_INET6:
1874                 /* TCP checksum */
1875                 th->th_sum = in6_cksum(m, IPPROTO_TCP,
1876                     sizeof(struct ip6_hdr), tlen);
1877
1878                 h6->ip6_vfc |= IPV6_VERSION;
1879                 h6->ip6_hlim = IPV6_DEFHLIM;
1880
1881                 lwkt_reltoken(&pf_token);
1882                 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
1883                 lwkt_gettoken(&pf_token);
1884                 break;
1885 #endif /* INET6 */
1886         }
1887 }
1888
1889 void
1890 pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af,
1891     struct pf_rule *r)
1892 {
1893         struct mbuf     *m0;
1894
1895         /*
1896          * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags,
1897          * so make sure pf.flags is clear.
1898          */
1899         if ((m0 = m_copy(m, 0, M_COPYALL)) == NULL)
1900                 return;
1901
1902         m0->m_pkthdr.fw_flags |= PF_MBUF_TAGGED;
1903         m0->m_pkthdr.pf.flags = 0;
1904         /* XXX Re-Check when Upgrading to > 4.4 */
1905         m0->m_pkthdr.pf.statekey = NULL;
1906
1907         if (r->rtableid >= 0)
1908                 m0->m_pkthdr.pf.rtableid = r->rtableid;
1909
1910 #ifdef ALTQ
1911         if (r->qid) {
1912                 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
1913                 m0->m_pkthdr.pf.qid = r->qid;
1914                 m0->m_pkthdr.pf.ecn_af = af;
1915                 m0->m_pkthdr.pf.hdr = mtod(m0, struct ip *);
1916         }
1917 #endif /* ALTQ */
1918
1919         switch (af) {
1920 #ifdef INET
1921         case AF_INET:
1922                 icmp_error(m0, type, code, 0, 0);
1923                 break;
1924 #endif /* INET */
1925 #ifdef INET6
1926         case AF_INET6:
1927                 icmp6_error(m0, type, code, 0);
1928                 break;
1929 #endif /* INET6 */
1930         }
1931 }
1932
1933 /*
1934  * Return 1 if the addresses a and b match (with mask m), otherwise return 0.
1935  * If n is 0, they match if they are equal. If n is != 0, they match if they
1936  * are different.
1937  */
1938 int
1939 pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m,
1940     struct pf_addr *b, sa_family_t af)
1941 {
1942         int     match = 0;
1943
1944         switch (af) {
1945 #ifdef INET
1946         case AF_INET:
1947                 if ((a->addr32[0] & m->addr32[0]) ==
1948                     (b->addr32[0] & m->addr32[0]))
1949                         match++;
1950                 break;
1951 #endif /* INET */
1952 #ifdef INET6
1953         case AF_INET6:
1954                 if (((a->addr32[0] & m->addr32[0]) ==
1955                      (b->addr32[0] & m->addr32[0])) &&
1956                     ((a->addr32[1] & m->addr32[1]) ==
1957                      (b->addr32[1] & m->addr32[1])) &&
1958                     ((a->addr32[2] & m->addr32[2]) ==
1959                      (b->addr32[2] & m->addr32[2])) &&
1960                     ((a->addr32[3] & m->addr32[3]) ==
1961                      (b->addr32[3] & m->addr32[3])))
1962                         match++;
1963                 break;
1964 #endif /* INET6 */
1965         }
1966         if (match) {
1967                 if (n)
1968                         return (0);
1969                 else
1970                         return (1);
1971         } else {
1972                 if (n)
1973                         return (1);
1974                 else
1975                         return (0);
1976         }
1977 }
1978
1979 /*
1980  * Return 1 if b <= a <= e, otherwise return 0.
1981  */
1982 int
1983 pf_match_addr_range(struct pf_addr *b, struct pf_addr *e,
1984     struct pf_addr *a, sa_family_t af)
1985 {
1986         switch (af) {
1987 #ifdef INET
1988         case AF_INET:
1989                 if ((a->addr32[0] < b->addr32[0]) ||
1990                     (a->addr32[0] > e->addr32[0]))
1991                         return (0);
1992                 break;
1993 #endif /* INET */
1994 #ifdef INET6
1995         case AF_INET6: {
1996                 int     i;
1997
1998                 /* check a >= b */
1999                 for (i = 0; i < 4; ++i)
2000                         if (a->addr32[i] > b->addr32[i])
2001                                 break;
2002                         else if (a->addr32[i] < b->addr32[i])
2003                                 return (0);
2004                 /* check a <= e */
2005                 for (i = 0; i < 4; ++i)
2006                         if (a->addr32[i] < e->addr32[i])
2007                                 break;
2008                         else if (a->addr32[i] > e->addr32[i])
2009                                 return (0);
2010                 break;
2011         }
2012 #endif /* INET6 */
2013         }
2014         return (1);
2015 }
2016
2017 int
2018 pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p)
2019 {
2020         switch (op) {
2021         case PF_OP_IRG:
2022                 return ((p > a1) && (p < a2));
2023         case PF_OP_XRG:
2024                 return ((p < a1) || (p > a2));
2025         case PF_OP_RRG:
2026                 return ((p >= a1) && (p <= a2));
2027         case PF_OP_EQ:
2028                 return (p == a1);
2029         case PF_OP_NE:
2030                 return (p != a1);
2031         case PF_OP_LT:
2032                 return (p < a1);
2033         case PF_OP_LE:
2034                 return (p <= a1);
2035         case PF_OP_GT:
2036                 return (p > a1);
2037         case PF_OP_GE:
2038                 return (p >= a1);
2039         }
2040         return (0); /* never reached */
2041 }
2042
2043 int
2044 pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p)
2045 {
2046         a1 = ntohs(a1);
2047         a2 = ntohs(a2);
2048         p = ntohs(p);
2049         return (pf_match(op, a1, a2, p));
2050 }
2051
2052 int
2053 pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u)
2054 {
2055         if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
2056                 return (0);
2057         return (pf_match(op, a1, a2, u));
2058 }
2059
2060 int
2061 pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g)
2062 {
2063         if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
2064                 return (0);
2065         return (pf_match(op, a1, a2, g));
2066 }
2067
2068 int
2069 pf_match_tag(struct mbuf *m, struct pf_rule *r, int *tag)
2070 {
2071         if (*tag == -1)
2072                 *tag = m->m_pkthdr.pf.tag;
2073
2074         return ((!r->match_tag_not && r->match_tag == *tag) ||
2075             (r->match_tag_not && r->match_tag != *tag));
2076 }
2077
2078 int
2079 pf_tag_packet(struct mbuf *m, int tag, int rtableid)
2080 {
2081         if (tag <= 0 && rtableid < 0)
2082                 return (0);
2083
2084         if (tag > 0)
2085                 m->m_pkthdr.pf.tag = tag;
2086         if (rtableid >= 0)
2087                 m->m_pkthdr.pf.rtableid = rtableid;
2088
2089         return (0);
2090 }
2091
2092 void
2093 pf_step_into_anchor(int *depth, struct pf_ruleset **rs, int n,
2094     struct pf_rule **r, struct pf_rule **a, int *match)
2095 {
2096         struct pf_anchor_stackframe     *f;
2097
2098         (*r)->anchor->match = 0;
2099         if (match)
2100                 *match = 0;
2101         if (*depth >= sizeof(pf_anchor_stack) /
2102             sizeof(pf_anchor_stack[0])) {
2103                 kprintf("pf_step_into_anchor: stack overflow\n");
2104                 *r = TAILQ_NEXT(*r, entries);
2105                 return;
2106         } else if (*depth == 0 && a != NULL)
2107                 *a = *r;
2108         f = pf_anchor_stack + (*depth)++;
2109         f->rs = *rs;
2110         f->r = *r;
2111         if ((*r)->anchor_wildcard) {
2112                 f->parent = &(*r)->anchor->children;
2113                 if ((f->child = RB_MIN(pf_anchor_node, f->parent)) ==
2114                     NULL) {
2115                         *r = NULL;
2116                         return;
2117                 }
2118                 *rs = &f->child->ruleset;
2119         } else {
2120                 f->parent = NULL;
2121                 f->child = NULL;
2122                 *rs = &(*r)->anchor->ruleset;
2123         }
2124         *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
2125 }
2126
2127 int
2128 pf_step_out_of_anchor(int *depth, struct pf_ruleset **rs, int n,
2129     struct pf_rule **r, struct pf_rule **a, int *match)
2130 {
2131         struct pf_anchor_stackframe     *f;
2132         int quick = 0;
2133
2134         do {
2135                 if (*depth <= 0)
2136                         break;
2137                 f = pf_anchor_stack + *depth - 1;
2138                 if (f->parent != NULL && f->child != NULL) {
2139                         if (f->child->match ||
2140                             (match != NULL && *match)) {
2141                                 f->r->anchor->match = 1;
2142                                 *match = 0;
2143                         }
2144                         f->child = RB_NEXT(pf_anchor_node, f->parent, f->child);
2145                         if (f->child != NULL) {
2146                                 *rs = &f->child->ruleset;
2147                                 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
2148                                 if (*r == NULL)
2149                                         continue;
2150                                 else
2151                                         break;
2152                         }
2153                 }
2154                 (*depth)--;
2155                 if (*depth == 0 && a != NULL)
2156                         *a = NULL;
2157                 *rs = f->rs;
2158                 if (f->r->anchor->match || (match != NULL && *match))
2159                         quick = f->r->quick;
2160                 *r = TAILQ_NEXT(f->r, entries);
2161         } while (*r == NULL);
2162
2163         return (quick);
2164 }
2165
2166 #ifdef INET6
2167 void
2168 pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr,
2169     struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af)
2170 {
2171         switch (af) {
2172 #ifdef INET
2173         case AF_INET:
2174                 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2175                 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2176                 break;
2177 #endif /* INET */
2178         case AF_INET6:
2179                 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2180                 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2181                 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) |
2182                 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]);
2183                 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) |
2184                 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]);
2185                 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) |
2186                 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]);
2187                 break;
2188         }
2189 }
2190
2191 void
2192 pf_addr_inc(struct pf_addr *addr, sa_family_t af)
2193 {
2194         switch (af) {
2195 #ifdef INET
2196         case AF_INET:
2197                 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1);
2198                 break;
2199 #endif /* INET */
2200         case AF_INET6:
2201                 if (addr->addr32[3] == 0xffffffff) {
2202                         addr->addr32[3] = 0;
2203                         if (addr->addr32[2] == 0xffffffff) {
2204                                 addr->addr32[2] = 0;
2205                                 if (addr->addr32[1] == 0xffffffff) {
2206                                         addr->addr32[1] = 0;
2207                                         addr->addr32[0] =
2208                                             htonl(ntohl(addr->addr32[0]) + 1);
2209                                 } else
2210                                         addr->addr32[1] =
2211                                             htonl(ntohl(addr->addr32[1]) + 1);
2212                         } else
2213                                 addr->addr32[2] =
2214                                     htonl(ntohl(addr->addr32[2]) + 1);
2215                 } else
2216                         addr->addr32[3] =
2217                             htonl(ntohl(addr->addr32[3]) + 1);
2218                 break;
2219         }
2220 }
2221 #endif /* INET6 */
2222
2223 #define mix(a,b,c) \
2224         do {                                    \
2225                 a -= b; a -= c; a ^= (c >> 13); \
2226                 b -= c; b -= a; b ^= (a << 8);  \
2227                 c -= a; c -= b; c ^= (b >> 13); \
2228                 a -= b; a -= c; a ^= (c >> 12); \
2229                 b -= c; b -= a; b ^= (a << 16); \
2230                 c -= a; c -= b; c ^= (b >> 5);  \
2231                 a -= b; a -= c; a ^= (c >> 3);  \
2232                 b -= c; b -= a; b ^= (a << 10); \
2233                 c -= a; c -= b; c ^= (b >> 15); \
2234         } while (0)
2235
2236 /*
2237  * hash function based on bridge_hash in if_bridge.c
2238  */
2239 void
2240 pf_hash(struct pf_addr *inaddr, struct pf_addr *hash,
2241     struct pf_poolhashkey *key, sa_family_t af)
2242 {
2243         u_int32_t       a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0];
2244
2245         switch (af) {
2246 #ifdef INET
2247         case AF_INET:
2248                 a += inaddr->addr32[0];
2249                 b += key->key32[1];
2250                 mix(a, b, c);
2251                 hash->addr32[0] = c + key->key32[2];
2252                 break;
2253 #endif /* INET */
2254 #ifdef INET6
2255         case AF_INET6:
2256                 a += inaddr->addr32[0];
2257                 b += inaddr->addr32[2];
2258                 mix(a, b, c);
2259                 hash->addr32[0] = c;
2260                 a += inaddr->addr32[1];
2261                 b += inaddr->addr32[3];
2262                 c += key->key32[1];
2263                 mix(a, b, c);
2264                 hash->addr32[1] = c;
2265                 a += inaddr->addr32[2];
2266                 b += inaddr->addr32[1];
2267                 c += key->key32[2];
2268                 mix(a, b, c);
2269                 hash->addr32[2] = c;
2270                 a += inaddr->addr32[3];
2271                 b += inaddr->addr32[0];
2272                 c += key->key32[3];
2273                 mix(a, b, c);
2274                 hash->addr32[3] = c;
2275                 break;
2276 #endif /* INET6 */
2277         }
2278 }
2279
2280 int
2281 pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr,
2282     struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn)
2283 {
2284         unsigned char            hash[16];
2285         struct pf_pool          *rpool = &r->rpool;
2286         struct pf_addr          *raddr = &rpool->cur->addr.v.a.addr;
2287         struct pf_addr          *rmask = &rpool->cur->addr.v.a.mask;
2288         struct pf_pooladdr      *acur = rpool->cur;
2289         struct pf_src_node       k;
2290
2291         if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR &&
2292             (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2293                 k.af = af;
2294                 PF_ACPY(&k.addr, saddr, af);
2295                 if (r->rule_flag & PFRULE_RULESRCTRACK ||
2296                     r->rpool.opts & PF_POOL_STICKYADDR)
2297                         k.rule.ptr = r;
2298                 else
2299                         k.rule.ptr = NULL;
2300                 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
2301                 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
2302                 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) {
2303                         PF_ACPY(naddr, &(*sn)->raddr, af);
2304                         if (pf_status.debug >= PF_DEBUG_MISC) {
2305                                 kprintf("pf_map_addr: src tracking maps ");
2306                                 pf_print_host(&k.addr, 0, af);
2307                                 kprintf(" to ");
2308                                 pf_print_host(naddr, 0, af);
2309                                 kprintf("\n");
2310                         }
2311                         return (0);
2312                 }
2313         }
2314
2315         if (rpool->cur->addr.type == PF_ADDR_NOROUTE)
2316                 return (1);
2317         if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2318                 switch (af) {
2319 #ifdef INET
2320                 case AF_INET:
2321                         if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 &&
2322                             (rpool->opts & PF_POOL_TYPEMASK) !=
2323                             PF_POOL_ROUNDROBIN)
2324                                 return (1);
2325                          raddr = &rpool->cur->addr.p.dyn->pfid_addr4;
2326                          rmask = &rpool->cur->addr.p.dyn->pfid_mask4;
2327                         break;
2328 #endif /* INET */
2329 #ifdef INET6
2330                 case AF_INET6:
2331                         if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 &&
2332                             (rpool->opts & PF_POOL_TYPEMASK) !=
2333                             PF_POOL_ROUNDROBIN)
2334                                 return (1);
2335                         raddr = &rpool->cur->addr.p.dyn->pfid_addr6;
2336                         rmask = &rpool->cur->addr.p.dyn->pfid_mask6;
2337                         break;
2338 #endif /* INET6 */
2339                 }
2340         } else if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2341                 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN)
2342                         return (1); /* unsupported */
2343         } else {
2344                 raddr = &rpool->cur->addr.v.a.addr;
2345                 rmask = &rpool->cur->addr.v.a.mask;
2346         }
2347
2348         switch (rpool->opts & PF_POOL_TYPEMASK) {
2349         case PF_POOL_NONE:
2350                 PF_ACPY(naddr, raddr, af);
2351                 break;
2352         case PF_POOL_BITMASK:
2353                 PF_POOLMASK(naddr, raddr, rmask, saddr, af);
2354                 break;
2355         case PF_POOL_RANDOM:
2356                 if (init_addr != NULL && PF_AZERO(init_addr, af)) {
2357                         switch (af) {
2358 #ifdef INET
2359                         case AF_INET:
2360                                 rpool->counter.addr32[0] = htonl(karc4random());
2361                                 break;
2362 #endif /* INET */
2363 #ifdef INET6
2364                         case AF_INET6:
2365                                 if (rmask->addr32[3] != 0xffffffff)
2366                                         rpool->counter.addr32[3] =
2367                                             htonl(karc4random());
2368                                 else
2369                                         break;
2370                                 if (rmask->addr32[2] != 0xffffffff)
2371                                         rpool->counter.addr32[2] =
2372                                             htonl(karc4random());
2373                                 else
2374                                         break;
2375                                 if (rmask->addr32[1] != 0xffffffff)
2376                                         rpool->counter.addr32[1] =
2377                                             htonl(karc4random());
2378                                 else
2379                                         break;
2380                                 if (rmask->addr32[0] != 0xffffffff)
2381                                         rpool->counter.addr32[0] =
2382                                             htonl(karc4random());
2383                                 break;
2384 #endif /* INET6 */
2385                         }
2386                         PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2387                         PF_ACPY(init_addr, naddr, af);
2388
2389                 } else {
2390                         PF_AINC(&rpool->counter, af);
2391                         PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2392                 }
2393                 break;
2394         case PF_POOL_SRCHASH:
2395                 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af);
2396                 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af);
2397                 break;
2398         case PF_POOL_ROUNDROBIN:
2399                 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2400                         if (!pfr_pool_get(rpool->cur->addr.p.tbl,
2401                             &rpool->tblidx, &rpool->counter,
2402                             &raddr, &rmask, af))
2403                                 goto get_addr;
2404                 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2405                         if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2406                             &rpool->tblidx, &rpool->counter,
2407                             &raddr, &rmask, af))
2408                                 goto get_addr;
2409                 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af))
2410                         goto get_addr;
2411
2412         try_next:
2413                 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL)
2414                         rpool->cur = TAILQ_FIRST(&rpool->list);
2415                 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2416                         rpool->tblidx = -1;
2417                         if (pfr_pool_get(rpool->cur->addr.p.tbl,
2418                             &rpool->tblidx, &rpool->counter,
2419                             &raddr, &rmask, af)) {
2420                                 /* table contains no address of type 'af' */
2421                                 if (rpool->cur != acur)
2422                                         goto try_next;
2423                                 return (1);
2424                         }
2425                 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2426                         rpool->tblidx = -1;
2427                         if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2428                             &rpool->tblidx, &rpool->counter,
2429                             &raddr, &rmask, af)) {
2430                                 /* table contains no address of type 'af' */
2431                                 if (rpool->cur != acur)
2432                                         goto try_next;
2433                                 return (1);
2434                         }
2435                 } else {
2436                         raddr = &rpool->cur->addr.v.a.addr;
2437                         rmask = &rpool->cur->addr.v.a.mask;
2438                         PF_ACPY(&rpool->counter, raddr, af);
2439                 }
2440
2441         get_addr:
2442                 PF_ACPY(naddr, &rpool->counter, af);
2443                 if (init_addr != NULL && PF_AZERO(init_addr, af))
2444                         PF_ACPY(init_addr, naddr, af);
2445                 PF_AINC(&rpool->counter, af);
2446                 break;
2447         }
2448         if (*sn != NULL)
2449                 PF_ACPY(&(*sn)->raddr, naddr, af);
2450
2451         if (pf_status.debug >= PF_DEBUG_MISC &&
2452             (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2453                 kprintf("pf_map_addr: selected address ");
2454                 pf_print_host(naddr, 0, af);
2455                 kprintf("\n");
2456         }
2457
2458         return (0);
2459 }
2460
2461 int
2462 pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r,
2463     struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport,
2464     struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high,
2465     struct pf_src_node **sn)
2466 {
2467         struct pf_state_key_cmp key;
2468         struct pf_addr          init_addr;
2469         u_int16_t               cut;
2470
2471         bzero(&init_addr, sizeof(init_addr));
2472         if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2473                 return (1);
2474
2475         if (proto == IPPROTO_ICMP) {
2476                 low = 1;
2477                 high = 65535;
2478         }
2479
2480         do {
2481                 key.af = af;
2482                 key.proto = proto;
2483                 PF_ACPY(&key.addr[1], daddr, key.af);
2484                 PF_ACPY(&key.addr[0], naddr, key.af);
2485                 key.port[1] = dport;
2486
2487                 /*
2488                  * port search; start random, step;
2489                  * similar 2 portloop in in_pcbbind
2490                  */
2491                 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP ||
2492                     proto == IPPROTO_ICMP)) {
2493                         key.port[0] = dport;
2494                         if (pf_find_state_all(&key, PF_IN, NULL) == NULL)
2495                                 return (0);
2496                 } else if (low == 0 && high == 0) {
2497                         key.port[0] = *nport;
2498                         if (pf_find_state_all(&key, PF_IN, NULL) == NULL)
2499                                 return (0);
2500                 } else if (low == high) {
2501                         key.port[0] = htons(low);
2502                         if (pf_find_state_all(&key, PF_IN, NULL) == NULL) {
2503                                 *nport = htons(low);
2504                                 return (0);
2505                         }
2506                 } else {
2507                         u_int16_t tmp;
2508
2509                         if (low > high) {
2510                                 tmp = low;
2511                                 low = high;
2512                                 high = tmp;
2513                         }
2514                         /* low < high */
2515                         cut = htonl(karc4random()) % (1 + high - low) + low;
2516                         /* low <= cut <= high */
2517                         for (tmp = cut; tmp <= high; ++(tmp)) {
2518                                 key.port[0] = htons(tmp);
2519                                 if (pf_find_state_all(&key, PF_IN, NULL) ==
2520                                     NULL && !in_baddynamic(tmp, proto)) {
2521                                         *nport = htons(tmp);
2522                                         return (0);
2523                                 }
2524                         }
2525                         for (tmp = cut - 1; tmp >= low; --(tmp)) {
2526                                 key.port[0] = htons(tmp);
2527                                 if (pf_find_state_all(&key, PF_IN, NULL) ==
2528                                     NULL && !in_baddynamic(tmp, proto)) {
2529                                         *nport = htons(tmp);
2530                                         return (0);
2531                                 }
2532                         }
2533                 }
2534
2535                 switch (r->rpool.opts & PF_POOL_TYPEMASK) {
2536                 case PF_POOL_RANDOM:
2537                 case PF_POOL_ROUNDROBIN:
2538                         if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2539                                 return (1);
2540                         break;
2541                 case PF_POOL_NONE:
2542                 case PF_POOL_SRCHASH:
2543                 case PF_POOL_BITMASK:
2544                 default:
2545                         return (1);
2546                 }
2547         } while (! PF_AEQ(&init_addr, naddr, af) );
2548         return (1);                                     /* none available */
2549 }
2550
2551 struct pf_rule *
2552 pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off,
2553     int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport,
2554     struct pf_addr *daddr, u_int16_t dport, int rs_num)
2555 {
2556         struct pf_rule          *r, *rm = NULL;
2557         struct pf_ruleset       *ruleset = NULL;
2558         int                      tag = -1;
2559         int                      rtableid = -1;
2560         int                      asd = 0;
2561
2562         r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr);
2563         while (r && rm == NULL) {
2564                 struct pf_rule_addr     *src = NULL, *dst = NULL;
2565                 struct pf_addr_wrap     *xdst = NULL;
2566
2567                 if (r->action == PF_BINAT && direction == PF_IN) {
2568                         src = &r->dst;
2569                         if (r->rpool.cur != NULL)
2570                                 xdst = &r->rpool.cur->addr;
2571                 } else {
2572                         src = &r->src;
2573                         dst = &r->dst;
2574                 }
2575
2576                 r->evaluations++;
2577                 if (pfi_kif_match(r->kif, kif) == r->ifnot)
2578                         r = r->skip[PF_SKIP_IFP].ptr;
2579                 else if (r->direction && r->direction != direction)
2580                         r = r->skip[PF_SKIP_DIR].ptr;
2581                 else if (r->af && r->af != pd->af)
2582                         r = r->skip[PF_SKIP_AF].ptr;
2583                 else if (r->proto && r->proto != pd->proto)
2584                         r = r->skip[PF_SKIP_PROTO].ptr;
2585                 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af,
2586                     src->neg, kif))
2587                         r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR :
2588                             PF_SKIP_DST_ADDR].ptr;
2589                 else if (src->port_op && !pf_match_port(src->port_op,
2590                     src->port[0], src->port[1], sport))
2591                         r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT :
2592                             PF_SKIP_DST_PORT].ptr;
2593                 else if (dst != NULL &&
2594                     PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->neg, NULL))
2595                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
2596                 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af,
2597                     0, NULL))
2598                         r = TAILQ_NEXT(r, entries);
2599                 else if (dst != NULL && dst->port_op &&
2600                     !pf_match_port(dst->port_op, dst->port[0],
2601                     dst->port[1], dport))
2602                         r = r->skip[PF_SKIP_DST_PORT].ptr;
2603                 else if (r->match_tag && !pf_match_tag(m, r, &tag))
2604                         r = TAILQ_NEXT(r, entries);
2605                 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto !=
2606                     IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m,
2607                     off, pd->hdr.tcp), r->os_fingerprint)))
2608                         r = TAILQ_NEXT(r, entries);
2609                 else {
2610                         if (r->tag)
2611                                 tag = r->tag;
2612                         if (r->rtableid >= 0)
2613                                 rtableid = r->rtableid;
2614                         if (r->anchor == NULL) {
2615                                 rm = r;
2616                         } else
2617                                 pf_step_into_anchor(&asd, &ruleset, rs_num,
2618                                     &r, NULL, NULL);
2619                 }
2620                 if (r == NULL)
2621                         pf_step_out_of_anchor(&asd, &ruleset, rs_num, &r,
2622                             NULL, NULL);
2623         }
2624         if (pf_tag_packet(m, tag, rtableid))
2625                 return (NULL);
2626         if (rm != NULL && (rm->action == PF_NONAT ||
2627             rm->action == PF_NORDR || rm->action == PF_NOBINAT))
2628                 return (NULL);
2629         return (rm);
2630 }
2631
2632 struct pf_rule *
2633 pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction,
2634     struct pfi_kif *kif, struct pf_src_node **sn,
2635     struct pf_state_key **skw, struct pf_state_key **sks,
2636     struct pf_state_key **skp, struct pf_state_key **nkp,
2637     struct pf_addr *saddr, struct pf_addr *daddr,
2638     u_int16_t sport, u_int16_t dport)
2639 {
2640         struct pf_rule  *r = NULL;
2641
2642
2643         if (direction == PF_OUT) {
2644                 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2645                     sport, daddr, dport, PF_RULESET_BINAT);
2646                 if (r == NULL)
2647                         r = pf_match_translation(pd, m, off, direction, kif,
2648                             saddr, sport, daddr, dport, PF_RULESET_NAT);
2649         } else {
2650                 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2651                     sport, daddr, dport, PF_RULESET_RDR);
2652                 if (r == NULL)
2653                         r = pf_match_translation(pd, m, off, direction, kif,
2654                             saddr, sport, daddr, dport, PF_RULESET_BINAT);
2655         }
2656
2657         if (r != NULL) {
2658                 struct pf_addr  *naddr;
2659                 u_int16_t       *nport;
2660
2661                 if (pf_state_key_setup(pd, r, skw, sks, skp, nkp,
2662                     saddr, daddr, sport, dport))
2663                         return r;
2664
2665                 /* XXX We only modify one side for now. */
2666                 naddr = &(*nkp)->addr[1];
2667                 nport = &(*nkp)->port[1];
2668
2669                 switch (r->action) {
2670                 case PF_NONAT:
2671                 case PF_NOBINAT:
2672                 case PF_NORDR:
2673                         return (NULL);
2674                 case PF_NAT:
2675                         if (pf_get_sport(pd->af, pd->proto, r, saddr,
2676                             daddr, dport, naddr, nport, r->rpool.proxy_port[0],
2677                             r->rpool.proxy_port[1], sn)) {
2678                                 DPFPRINTF(PF_DEBUG_MISC,
2679                                     ("pf: NAT proxy port allocation "
2680                                     "(%u-%u) failed\n",
2681                                     r->rpool.proxy_port[0],
2682                                     r->rpool.proxy_port[1]));
2683                                 return (NULL);
2684                         }
2685                         break;
2686                 case PF_BINAT:
2687                         switch (direction) {
2688                         case PF_OUT:
2689                                 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){
2690                                         switch (pd->af) {
2691 #ifdef INET
2692                                         case AF_INET:
2693                                                 if (r->rpool.cur->addr.p.dyn->
2694                                                     pfid_acnt4 < 1)
2695                                                         return (NULL);
2696                                                 PF_POOLMASK(naddr,
2697                                                     &r->rpool.cur->addr.p.dyn->
2698                                                     pfid_addr4,
2699                                                     &r->rpool.cur->addr.p.dyn->
2700                                                     pfid_mask4,
2701                                                     saddr, AF_INET);
2702                                                 break;
2703 #endif /* INET */
2704 #ifdef INET6
2705                                         case AF_INET6:
2706                                                 if (r->rpool.cur->addr.p.dyn->
2707                                                     pfid_acnt6 < 1)
2708                                                         return (NULL);
2709                                                 PF_POOLMASK(naddr,
2710                                                     &r->rpool.cur->addr.p.dyn->
2711                                                     pfid_addr6,
2712                                                     &r->rpool.cur->addr.p.dyn->
2713                                                     pfid_mask6,
2714                                                     saddr, AF_INET6);
2715                                                 break;
2716 #endif /* INET6 */
2717                                         }
2718                                 } else
2719                                         PF_POOLMASK(naddr,
2720                                             &r->rpool.cur->addr.v.a.addr,
2721                                             &r->rpool.cur->addr.v.a.mask,
2722                                             saddr, pd->af);
2723                                 break;
2724                         case PF_IN:
2725                                 if (r->src.addr.type == PF_ADDR_DYNIFTL) {
2726                                         switch (pd->af) {
2727 #ifdef INET
2728                                         case AF_INET:
2729                                                 if (r->src.addr.p.dyn->
2730                                                     pfid_acnt4 < 1)
2731                                                         return (NULL);
2732                                                 PF_POOLMASK(naddr,
2733                                                     &r->src.addr.p.dyn->
2734                                                     pfid_addr4,
2735                                                     &r->src.addr.p.dyn->
2736                                                     pfid_mask4,
2737                                                     daddr, AF_INET);
2738                                                 break;
2739 #endif /* INET */
2740 #ifdef INET6
2741                                         case AF_INET6:
2742                                                 if (r->src.addr.p.dyn->
2743                                                     pfid_acnt6 < 1)
2744                                                         return (NULL);
2745                                                 PF_POOLMASK(naddr,
2746                                                     &r->src.addr.p.dyn->
2747                                                     pfid_addr6,
2748                                                     &r->src.addr.p.dyn->
2749                                                     pfid_mask6,
2750                                                     daddr, AF_INET6);
2751                                                 break;
2752 #endif /* INET6 */
2753                                         }
2754                                 } else
2755                                         PF_POOLMASK(naddr,
2756                                             &r->src.addr.v.a.addr,
2757                                             &r->src.addr.v.a.mask, daddr,
2758                                             pd->af);
2759                                 break;
2760                         }
2761                         break;
2762                 case PF_RDR: {
2763                         if (pf_map_addr(pd->af, r, saddr, naddr, NULL, sn))
2764                                 return (NULL);
2765                         if ((r->rpool.opts & PF_POOL_TYPEMASK) ==
2766                             PF_POOL_BITMASK)
2767                                 PF_POOLMASK(naddr, naddr,
2768                                     &r->rpool.cur->addr.v.a.mask, daddr,
2769                                     pd->af);
2770
2771                         if (r->rpool.proxy_port[1]) {
2772                                 u_int32_t       tmp_nport;
2773
2774                                 tmp_nport = ((ntohs(dport) -
2775                                     ntohs(r->dst.port[0])) %
2776                                     (r->rpool.proxy_port[1] -
2777                                     r->rpool.proxy_port[0] + 1)) +
2778                                     r->rpool.proxy_port[0];
2779
2780                                 /* wrap around if necessary */
2781                                 if (tmp_nport > 65535)
2782                                         tmp_nport -= 65535;
2783                                 *nport = htons((u_int16_t)tmp_nport);
2784                         } else if (r->rpool.proxy_port[0])
2785                                 *nport = htons(r->rpool.proxy_port[0]);
2786                         break;
2787                 }
2788                 default:
2789                         return (NULL);
2790                 }
2791         }
2792
2793         return (r);
2794 }
2795
2796 #ifdef SMP
2797 struct netmsg_hashlookup {
2798         struct netmsg_base      base;
2799         struct inpcb            **nm_pinp;
2800         struct inpcbinfo        *nm_pcbinfo;
2801         struct pf_addr          *nm_saddr;
2802         struct pf_addr          *nm_daddr;
2803         uint16_t                nm_sport;
2804         uint16_t                nm_dport;
2805         sa_family_t             nm_af;
2806 };
2807
2808 #ifdef PF_SOCKET_LOOKUP_DOMSG
2809 static void
2810 in_pcblookup_hash_handler(netmsg_t msg)
2811 {
2812         struct netmsg_hashlookup *rmsg = (struct netmsg_hashlookup *)msg;
2813
2814         if (rmsg->nm_af == AF_INET)
2815                 *rmsg->nm_pinp = in_pcblookup_hash(rmsg->nm_pcbinfo,
2816                     rmsg->nm_saddr->v4, rmsg->nm_sport, rmsg->nm_daddr->v4,
2817                     rmsg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2818 #ifdef INET6
2819         else
2820                 *rmsg->nm_pinp = in6_pcblookup_hash(rmsg->nm_pcbinfo,
2821                     &rmsg->nm_saddr->v6, rmsg->nm_sport, &rmsg->nm_daddr->v6,
2822                     rmsg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2823 #endif /* INET6 */
2824         lwkt_replymsg(&rmsg->base.lmsg, 0);
2825 }
2826 #endif  /* PF_SOCKET_LOOKUP_DOMSG */
2827
2828 #endif /* SMP */
2829
2830 int
2831 pf_socket_lookup(int direction, struct pf_pdesc *pd)
2832 {
2833         struct pf_addr          *saddr, *daddr;
2834         u_int16_t                sport, dport;
2835         struct inpcbinfo        *pi;
2836         struct inpcb            *inp;
2837 #ifdef SMP
2838         struct netmsg_hashlookup *msg = NULL;
2839 #ifdef PF_SOCKET_LOOKUP_DOMSG
2840         struct netmsg_hashlookup msg0;
2841 #endif
2842 #endif
2843         int                      pi_cpu = 0;
2844
2845         if (pd == NULL)
2846                 return (-1);
2847         pd->lookup.uid = UID_MAX;
2848         pd->lookup.gid = GID_MAX;
2849         pd->lookup.pid = NO_PID;
2850         if (direction == PF_IN) {
2851                 saddr = pd->src;
2852                 daddr = pd->dst;
2853         } else {
2854                 saddr = pd->dst;
2855                 daddr = pd->src;
2856         }
2857         switch (pd->proto) {
2858         case IPPROTO_TCP:
2859                 if (pd->hdr.tcp == NULL)
2860                         return (-1);
2861                 sport = pd->hdr.tcp->th_sport;
2862                 dport = pd->hdr.tcp->th_dport;
2863
2864                 pi_cpu = tcp_addrcpu(saddr->v4.s_addr, sport, daddr->v4.s_addr, dport);
2865                 pi = &tcbinfo[pi_cpu];
2866 #ifdef SMP
2867                 /*
2868                  * Our netstack runs lockless on MP systems
2869                  * (only for TCP connections at the moment).
2870                  *
2871                  * As we are not allowed to read another CPU's tcbinfo,
2872                  * we have to ask that CPU via remote call to search the
2873                  * table for us.
2874                  *
2875                  * Prepare a msg iff data belongs to another CPU.
2876                  */
2877                 if (pi_cpu != mycpu->gd_cpuid) {
2878 #ifdef PF_SOCKET_LOOKUP_DOMSG
2879                         /*
2880                          * NOTE:
2881                          *
2882                          * Following lwkt_domsg() is dangerous and could
2883                          * lockup the network system, e.g.
2884                          *
2885                          * On 2 CPU system:
2886                          * netisr0 domsg to netisr1 (due to lookup)
2887                          * netisr1 domsg to netisr0 (due to lookup)
2888                          *
2889                          * We simply return -1 here, since we are probably
2890                          * called before NAT, so the TCP packet should
2891                          * already be on the correct CPU.
2892                          */
2893                         msg = &msg0;
2894                         netmsg_init(&msg->base, NULL, &curthread->td_msgport,
2895                                     0, in_pcblookup_hash_handler);
2896                         msg->nm_pinp = &inp;
2897                         msg->nm_pcbinfo = pi;
2898                         msg->nm_saddr = saddr;
2899                         msg->nm_sport = sport;
2900                         msg->nm_daddr = daddr;
2901                         msg->nm_dport = dport;
2902                         msg->nm_af = pd->af;
2903 #else   /* !PF_SOCKET_LOOKUP_DOMSG */
2904                         kprintf("pf_socket_lookup: tcp packet not on the "
2905                                 "correct cpu %d, cur cpu %d\n",
2906                                 pi_cpu, mycpuid);
2907                         print_backtrace(-1);
2908                         return -1;
2909 #endif  /* PF_SOCKET_LOOKUP_DOMSG */
2910                 }
2911 #endif /* SMP */
2912                 break;
2913         case IPPROTO_UDP:
2914                 if (pd->hdr.udp == NULL)
2915                         return (-1);
2916                 sport = pd->hdr.udp->uh_sport;
2917                 dport = pd->hdr.udp->uh_dport;
2918                 pi = &udbinfo;
2919                 break;
2920         default:
2921                 return (-1);
2922         }
2923         if (direction != PF_IN) {
2924                 u_int16_t       p;
2925
2926                 p = sport;
2927                 sport = dport;
2928                 dport = p;
2929         }
2930         switch (pd->af) {
2931 #ifdef INET6
2932         case AF_INET6:
2933 #ifdef SMP
2934                 /*
2935                  * Query other CPU, second part
2936                  *
2937                  * msg only gets initialized when:
2938                  * 1) packet is TCP
2939                  * 2) the info belongs to another CPU
2940                  *
2941                  * Use some switch/case magic to avoid code duplication.
2942                  */
2943                 if (msg == NULL)
2944 #endif /* SMP */
2945                 {
2946                         inp = in6_pcblookup_hash(pi, &saddr->v6, sport,
2947                             &daddr->v6, dport, INPLOOKUP_WILDCARD, NULL);
2948
2949                         if (inp == NULL)
2950                                 return (-1);
2951                         break;
2952                 }
2953                 /* FALLTHROUGH if SMP and on other CPU */
2954 #endif /* INET6 */
2955         case AF_INET:
2956 #ifdef SMP
2957                 if (msg != NULL) {
2958                         lwkt_domsg(cpu_portfn(pi_cpu),
2959                                      &msg->base.lmsg, 0);
2960                 } else
2961 #endif /* SMP */
2962                 {
2963                         inp = in_pcblookup_hash(pi, saddr->v4, sport, daddr->v4,
2964                             dport, INPLOOKUP_WILDCARD, NULL);
2965                 }
2966                 if (inp == NULL)
2967                         return (-1);
2968                 break;
2969
2970         default:
2971                 return (-1);
2972         }
2973         pd->lookup.uid = inp->inp_socket->so_cred->cr_uid;
2974         pd->lookup.gid = inp->inp_socket->so_cred->cr_groups[0];
2975         return (1);
2976 }
2977
2978 u_int8_t
2979 pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2980 {
2981         int              hlen;
2982         u_int8_t         hdr[60];
2983         u_int8_t        *opt, optlen;
2984         u_int8_t         wscale = 0;
2985
2986         hlen = th_off << 2;             /* hlen <= sizeof(hdr) */
2987         if (hlen <= sizeof(struct tcphdr))
2988                 return (0);
2989         if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2990                 return (0);
2991         opt = hdr + sizeof(struct tcphdr);
2992         hlen -= sizeof(struct tcphdr);
2993         while (hlen >= 3) {
2994                 switch (*opt) {
2995                 case TCPOPT_EOL:
2996                 case TCPOPT_NOP:
2997                         ++opt;
2998                         --hlen;
2999                         break;
3000                 case TCPOPT_WINDOW:
3001                         wscale = opt[2];
3002                         if (wscale > TCP_MAX_WINSHIFT)
3003                                 wscale = TCP_MAX_WINSHIFT;
3004                         wscale |= PF_WSCALE_FLAG;
3005                         /* FALLTHROUGH */
3006                 default:
3007                         optlen = opt[1];
3008                         if (optlen < 2)
3009                                 optlen = 2;
3010                         hlen -= optlen;
3011                         opt += optlen;
3012                         break;
3013                 }
3014         }
3015         return (wscale);
3016 }
3017
3018 u_int16_t
3019 pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
3020 {
3021         int              hlen;
3022         u_int8_t         hdr[60];
3023         u_int8_t        *opt, optlen;
3024         u_int16_t        mss = tcp_mssdflt;
3025
3026         hlen = th_off << 2;     /* hlen <= sizeof(hdr) */
3027         if (hlen <= sizeof(struct tcphdr))
3028                 return (0);
3029         if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
3030                 return (0);
3031         opt = hdr + sizeof(struct tcphdr);
3032         hlen -= sizeof(struct tcphdr);
3033         while (hlen >= TCPOLEN_MAXSEG) {
3034                 switch (*opt) {
3035                 case TCPOPT_EOL:
3036                 case TCPOPT_NOP:
3037                         ++opt;
3038                         --hlen;
3039                         break;
3040                 case TCPOPT_MAXSEG:
3041                         bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2);
3042                         /* FALLTHROUGH */
3043                 default:
3044                         optlen = opt[1];
3045                         if (optlen < 2)
3046                                 optlen = 2;
3047                         hlen -= optlen;
3048                         opt += optlen;
3049                         break;
3050                 }
3051         }
3052         return (mss);
3053 }
3054
3055 u_int16_t
3056 pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer)
3057 {
3058 #ifdef INET
3059         struct sockaddr_in      *dst;
3060         struct route             ro;
3061 #endif /* INET */
3062 #ifdef INET6
3063         struct sockaddr_in6     *dst6;
3064         struct route_in6         ro6;
3065 #endif /* INET6 */
3066         struct rtentry          *rt = NULL;
3067         int                      hlen = 0;
3068         u_int16_t                mss = tcp_mssdflt;
3069
3070         switch (af) {
3071 #ifdef INET
3072         case AF_INET:
3073                 hlen = sizeof(struct ip);
3074                 bzero(&ro, sizeof(ro));
3075                 dst = (struct sockaddr_in *)&ro.ro_dst;
3076                 dst->sin_family = AF_INET;
3077                 dst->sin_len = sizeof(*dst);
3078                 dst->sin_addr = addr->v4;
3079                 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
3080                 rt = ro.ro_rt;
3081                 break;
3082 #endif /* INET */
3083 #ifdef INET6
3084         case AF_INET6:
3085                 hlen = sizeof(struct ip6_hdr);
3086                 bzero(&ro6, sizeof(ro6));
3087                 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst;
3088                 dst6->sin6_family = AF_INET6;
3089                 dst6->sin6_len = sizeof(*dst6);
3090                 dst6->sin6_addr = addr->v6;
3091                 rtalloc_ign((struct route *)&ro6, (RTF_CLONING | RTF_PRCLONING));
3092                 rt = ro6.ro_rt;
3093                 break;
3094 #endif /* INET6 */
3095         }
3096
3097         if (rt && rt->rt_ifp) {
3098                 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr);
3099                 mss = max(tcp_mssdflt, mss);
3100                 RTFREE(rt);
3101         }
3102         mss = min(mss, offer);
3103         mss = max(mss, 64);             /* sanity - at least max opt space */
3104         return (mss);
3105 }
3106
3107 void
3108 pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr)
3109 {
3110         struct pf_rule *r = s->rule.ptr;
3111
3112         s->rt_kif = NULL;
3113         if (!r->rt || r->rt == PF_FASTROUTE)
3114                 return;
3115         switch (s->key[PF_SK_WIRE]->af) {
3116 #ifdef INET
3117         case AF_INET:
3118                 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL,
3119                     &s->nat_src_node);
3120                 s->rt_kif = r->rpool.cur->kif;
3121                 break;
3122 #endif /* INET */
3123 #ifdef INET6
3124         case AF_INET6:
3125                 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL,
3126                     &s->nat_src_node);
3127                 s->rt_kif = r->rpool.cur->kif;
3128                 break;
3129 #endif /* INET6 */
3130         }
3131 }
3132
3133 u_int32_t
3134 pf_tcp_iss(struct pf_pdesc *pd)
3135 {
3136         MD5_CTX ctx;
3137         u_int32_t digest[4];
3138
3139         if (pf_tcp_secret_init == 0) {
3140                 karc4rand(pf_tcp_secret, sizeof(pf_tcp_secret));
3141                 MD5Init(&pf_tcp_secret_ctx);
3142                 MD5Update(&pf_tcp_secret_ctx, pf_tcp_secret,
3143                     sizeof(pf_tcp_secret));
3144                 pf_tcp_secret_init = 1;
3145         }
3146         ctx = pf_tcp_secret_ctx;
3147
3148         MD5Update(&ctx, (char *)&pd->hdr.tcp->th_sport, sizeof(u_short));
3149         MD5Update(&ctx, (char *)&pd->hdr.tcp->th_dport, sizeof(u_short));
3150         if (pd->af == AF_INET6) {
3151                 MD5Update(&ctx, (char *)&pd->src->v6, sizeof(struct in6_addr));
3152                 MD5Update(&ctx, (char *)&pd->dst->v6, sizeof(struct in6_addr));
3153         } else {
3154                 MD5Update(&ctx, (char *)&pd->src->v4, sizeof(struct in_addr));
3155                 MD5Update(&ctx, (char *)&pd->dst->v4, sizeof(struct in_addr));
3156         }
3157         MD5Final((u_char *)digest, &ctx);
3158         pf_tcp_iss_off += 4096;
3159         return (digest[0] + pd->hdr.tcp->th_seq + pf_tcp_iss_off);
3160 }
3161
3162 int
3163 pf_test_rule(struct pf_rule **rm, struct pf_state **sm, int direction,
3164     struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3165     struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
3166     struct ifqueue *ifq, struct inpcb *inp)
3167 {
3168         struct pf_rule          *nr = NULL;
3169         struct pf_addr          *saddr = pd->src, *daddr = pd->dst;
3170         sa_family_t              af = pd->af;
3171         struct pf_rule          *r, *a = NULL;
3172         struct pf_ruleset       *ruleset = NULL;
3173         struct pf_src_node      *nsn = NULL;
3174         struct tcphdr           *th = pd->hdr.tcp;
3175         struct pf_state_key     *skw = NULL, *sks = NULL;
3176         struct pf_state_key     *sk = NULL, *nk = NULL;
3177         u_short                  reason;
3178         int                      rewrite = 0, hdrlen = 0;
3179         int                      tag = -1, rtableid = -1;
3180         int                      asd = 0;
3181         int                      match = 0;
3182         int                      state_icmp = 0;
3183         u_int16_t                sport = 0, dport = 0;
3184         u_int16_t                nport = 0, bport = 0;
3185         u_int16_t                bproto_sum = 0, bip_sum = 0;
3186         u_int8_t                 icmptype = 0, icmpcode = 0;
3187
3188
3189         if (direction == PF_IN && pf_check_congestion(ifq)) {
3190                 REASON_SET(&reason, PFRES_CONGEST);
3191                 return (PF_DROP);
3192         }
3193
3194         if (inp != NULL)
3195                 pd->lookup.done = pf_socket_lookup(direction, pd);
3196         else if (debug_pfugidhack) {
3197                 DPFPRINTF(PF_DEBUG_MISC, ("pf: unlocked lookup\n"));
3198                 pd->lookup.done = pf_socket_lookup(direction, pd);
3199         }
3200
3201         switch (pd->proto) {
3202         case IPPROTO_TCP:
3203                 sport = th->th_sport;
3204                 dport = th->th_dport;
3205                 hdrlen = sizeof(*th);
3206                 break;
3207         case IPPROTO_UDP:
3208                 sport = pd->hdr.udp->uh_sport;
3209                 dport = pd->hdr.udp->uh_dport;
3210                 hdrlen = sizeof(*pd->hdr.udp);
3211                 break;
3212 #ifdef INET
3213         case IPPROTO_ICMP:
3214                 if (pd->af != AF_INET)
3215                         break;
3216                 sport = dport = pd->hdr.icmp->icmp_id;
3217                 hdrlen = sizeof(*pd->hdr.icmp);
3218                 icmptype = pd->hdr.icmp->icmp_type;
3219                 icmpcode = pd->hdr.icmp->icmp_code;
3220
3221                 if (icmptype == ICMP_UNREACH ||
3222                     icmptype == ICMP_SOURCEQUENCH ||
3223                     icmptype == ICMP_REDIRECT ||
3224                     icmptype == ICMP_TIMXCEED ||
3225                     icmptype == ICMP_PARAMPROB)
3226                         state_icmp++;
3227                 break;
3228 #endif /* INET */
3229 #ifdef INET6
3230         case IPPROTO_ICMPV6:
3231                 if (af != AF_INET6)
3232                         break;
3233                 sport = dport = pd->hdr.icmp6->icmp6_id;
3234                 hdrlen = sizeof(*pd->hdr.icmp6);
3235                 icmptype = pd->hdr.icmp6->icmp6_type;
3236                 icmpcode = pd->hdr.icmp6->icmp6_code;
3237
3238                 if (icmptype == ICMP6_DST_UNREACH ||
3239                     icmptype == ICMP6_PACKET_TOO_BIG ||
3240                     icmptype == ICMP6_TIME_EXCEEDED ||
3241                     icmptype == ICMP6_PARAM_PROB)
3242                         state_icmp++;
3243                 break;
3244 #endif /* INET6 */
3245         default:
3246                 sport = dport = hdrlen = 0;
3247                 break;
3248         }
3249
3250         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3251
3252         bport = nport = sport;
3253         /* check packet for BINAT/NAT/RDR */
3254         if ((nr = pf_get_translation(pd, m, off, direction, kif, &nsn,
3255             &skw, &sks, &sk, &nk, saddr, daddr, sport, dport)) != NULL) {
3256                 if (nk == NULL || sk == NULL) {
3257                         REASON_SET(&reason, PFRES_MEMORY);
3258                         goto cleanup;
3259                 }
3260
3261                 if (pd->ip_sum)
3262                         bip_sum = *pd->ip_sum;
3263
3264                 switch (pd->proto) {
3265                 case IPPROTO_TCP:
3266                         bproto_sum = th->th_sum;
3267                         pd->proto_sum = &th->th_sum;
3268
3269                         if (PF_ANEQ(saddr, &nk->addr[pd->sidx], af) ||
3270                             nk->port[pd->sidx] != sport) {
3271                                 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
3272                                     &th->th_sum, &nk->addr[pd->sidx],
3273                                     nk->port[pd->sidx], 0, af);
3274                                 pd->sport = &th->th_sport;
3275                                 sport = th->th_sport;
3276                         }
3277
3278                         if (PF_ANEQ(daddr, &nk->addr[pd->didx], af) ||
3279                             nk->port[pd->didx] != dport) {
3280                                 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
3281                                     &th->th_sum, &nk->addr[pd->didx],
3282                                     nk->port[pd->didx], 0, af);
3283                                 dport = th->th_dport;
3284                                 pd->dport = &th->th_dport;
3285                         }
3286                         rewrite++;
3287                         break;
3288                 case IPPROTO_UDP:
3289                         bproto_sum = pd->hdr.udp->uh_sum;
3290                         pd->proto_sum = &pd->hdr.udp->uh_sum;
3291
3292                         if (PF_ANEQ(saddr, &nk->addr[pd->sidx], af) ||
3293                             nk->port[pd->sidx] != sport) {
3294                                 pf_change_ap(saddr, &pd->hdr.udp->uh_sport,
3295                                     pd->ip_sum, &pd->hdr.udp->uh_sum,
3296                                     &nk->addr[pd->sidx],
3297                                     nk->port[pd->sidx], 1, af);
3298                                 sport = pd->hdr.udp->uh_sport;
3299                                 pd->sport = &pd->hdr.udp->uh_sport;
3300                         }
3301
3302                         if (PF_ANEQ(daddr, &nk->addr[pd->didx], af) ||
3303                             nk->port[pd->didx] != dport) {
3304                                 pf_change_ap(daddr, &pd->hdr.udp->uh_dport,
3305                                     pd->ip_sum, &pd->hdr.udp->uh_sum,
3306                                     &nk->addr[pd->didx],
3307                                     nk->port[pd->didx], 1, af);
3308                                 dport = pd->hdr.udp->uh_dport;
3309                                 pd->dport = &pd->hdr.udp->uh_dport;
3310                         }
3311                         rewrite++;
3312                         break;
3313 #ifdef INET
3314                 case IPPROTO_ICMP:
3315                         nk->port[0] = nk->port[1];
3316                         if (PF_ANEQ(saddr, &nk->addr[pd->sidx], AF_INET))
3317                                 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3318                                     nk->addr[pd->sidx].v4.s_addr, 0);
3319
3320                         if (PF_ANEQ(daddr, &nk->addr[pd->didx], AF_INET))
3321                                 pf_change_a(&daddr->v4.s_addr, pd->ip_sum,
3322                                     nk->addr[pd->didx].v4.s_addr, 0);
3323
3324                         if (nk->port[1] != pd->hdr.icmp->icmp_id) {
3325                                 pd->hdr.icmp->icmp_cksum = pf_cksum_fixup(
3326                                     pd->hdr.icmp->icmp_cksum, sport,
3327                                     nk->port[1], 0);
3328                                 pd->hdr.icmp->icmp_id = nk->port[1];
3329                                 pd->sport = &pd->hdr.icmp->icmp_id;
3330                         }
3331                         m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
3332                         break;
3333 #endif /* INET */
3334 #ifdef INET6
3335                 case IPPROTO_ICMPV6:
3336                         nk->port[0] = nk->port[1];
3337                         if (PF_ANEQ(saddr, &nk->addr[pd->sidx], AF_INET6))
3338                                 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum,
3339                                     &nk->addr[pd->sidx], 0);
3340
3341                         if (PF_ANEQ(daddr, &nk->addr[pd->didx], AF_INET6))
3342                                 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum,
3343                                     &nk->addr[pd->didx], 0);
3344                         rewrite++;
3345                         break;
3346 #endif /* INET */
3347                 default:
3348                         switch (af) {
3349 #ifdef INET
3350                         case AF_INET:
3351                                 if (PF_ANEQ(saddr,
3352                                     &nk->addr[pd->sidx], AF_INET))
3353                                         pf_change_a(&saddr->v4.s_addr,
3354                                             pd->ip_sum,
3355                                             nk->addr[pd->sidx].v4.s_addr, 0);
3356
3357                                 if (PF_ANEQ(daddr,
3358                                     &nk->addr[pd->didx], AF_INET))
3359                                         pf_change_a(&daddr->v4.s_addr,
3360                                             pd->ip_sum,
3361                                             nk->addr[pd->didx].v4.s_addr, 0);
3362                                 break;
3363 #endif /* INET */
3364 #ifdef INET6
3365                         case AF_INET6:
3366                                 if (PF_ANEQ(saddr,
3367                                     &nk->addr[pd->sidx], AF_INET6))
3368                                         PF_ACPY(saddr, &nk->addr[pd->sidx], af);
3369
3370                                 if (PF_ANEQ(daddr,
3371                                     &nk->addr[pd->didx], AF_INET6))
3372                                         PF_ACPY(saddr, &nk->addr[pd->didx], af);
3373                                 break;
3374 #endif /* INET */
3375                         }
3376                         break;
3377                 }
3378                 if (nr->natpass)
3379                         r = NULL;
3380                 pd->nat_rule = nr;
3381         }
3382
3383         while (r != NULL) {
3384                 r->evaluations++;
3385                 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3386                         r = r->skip[PF_SKIP_IFP].ptr;
3387                 else if (r->direction && r->direction != direction)
3388                         r = r->skip[PF_SKIP_DIR].ptr;
3389                 else if (r->af && r->af != af)
3390                         r = r->skip[PF_SKIP_AF].ptr;
3391                 else if (r->proto && r->proto != pd->proto)
3392                         r = r->skip[PF_SKIP_PROTO].ptr;
3393                 else if (PF_MISMATCHAW(&r->src.addr, saddr, af,
3394                     r->src.neg, kif))
3395                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3396                 /* tcp/udp only. port_op always 0 in other cases */
3397                 else if (r->src.port_op && !pf_match_port(r->src.port_op,
3398                     r->src.port[0], r->src.port[1], sport))
3399                         r = r->skip[PF_SKIP_SRC_PORT].ptr;
3400                 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af,
3401                     r->dst.neg, NULL))
3402                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
3403                 /* tcp/udp only. port_op always 0 in other cases */
3404                 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
3405                     r->dst.port[0], r->dst.port[1], dport))
3406                         r = r->skip[PF_SKIP_DST_PORT].ptr;
3407                 /* icmp only. type always 0 in other cases */
3408                 else if (r->type && r->type != icmptype + 1)
3409                         r = TAILQ_NEXT(r, entries);
3410                 /* icmp only. type always 0 in other cases */
3411                 else if (r->code && r->code != icmpcode + 1)
3412                         r = TAILQ_NEXT(r, entries);
3413                 else if (r->tos && !(r->tos == pd->tos))
3414                         r = TAILQ_NEXT(r, entries);
3415                 else if (r->rule_flag & PFRULE_FRAGMENT)
3416                         r = TAILQ_NEXT(r, entries);
3417                 else if (pd->proto == IPPROTO_TCP &&
3418                     (r->flagset & th->th_flags) != r->flags)
3419                         r = TAILQ_NEXT(r, entries);
3420                 /* tcp/udp only. uid.op always 0 in other cases */
3421                 else if (r->uid.op && (pd->lookup.done || (pd->lookup.done =
3422                     pf_socket_lookup(direction, pd), 1)) &&
3423                     !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
3424                     pd->lookup.uid))
3425                         r = TAILQ_NEXT(r, entries);
3426                 /* tcp/udp only. gid.op always 0 in other cases */
3427                 else if (r->gid.op && (pd->lookup.done || (pd->lookup.done =
3428                     pf_socket_lookup(direction, pd), 1)) &&
3429                     !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
3430                     pd->lookup.gid))
3431                         r = TAILQ_NEXT(r, entries);
3432                 else if (r->prob &&
3433                   r->prob <= karc4random())
3434                         r = TAILQ_NEXT(r, entries);
3435                 else if (r->match_tag && !pf_match_tag(m, r, &tag))
3436                         r = TAILQ_NEXT(r, entries);
3437                 else if (r->os_fingerprint != PF_OSFP_ANY &&
3438                     (pd->proto != IPPROTO_TCP || !pf_osfp_match(
3439                     pf_osfp_fingerprint(pd, m, off, th),
3440                     r->os_fingerprint)))
3441                         r = TAILQ_NEXT(r, entries);
3442                 else {
3443                         if (r->tag)
3444                                 tag = r->tag;
3445                         if (r->rtableid >= 0)
3446                                 rtableid = r->rtableid;
3447                         if (r->anchor == NULL) {
3448                                 match = 1;
3449                                 *rm = r;
3450                                 *am = a;
3451                                 *rsm = ruleset;
3452                                 if ((*rm)->quick)
3453                                         break;
3454                                 r = TAILQ_NEXT(r, entries);
3455                         } else
3456                                 pf_step_into_anchor(&asd, &ruleset,
3457                                     PF_RULESET_FILTER, &r, &a, &match);
3458                 }
3459                 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3460                     PF_RULESET_FILTER, &r, &a, &match))
3461                         break;
3462         }
3463         r = *rm;
3464         a = *am;
3465         ruleset = *rsm;
3466
3467         REASON_SET(&reason, PFRES_MATCH);
3468
3469         if (r->log || (nr != NULL && nr->log)) {
3470                 if (rewrite)
3471                         m_copyback(m, off, hdrlen, pd->hdr.any);
3472                 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr,
3473                     a, ruleset, pd);
3474         }
3475
3476         if ((r->action == PF_DROP) &&
3477             ((r->rule_flag & PFRULE_RETURNRST) ||
3478             (r->rule_flag & PFRULE_RETURNICMP) ||
3479             (r->rule_flag & PFRULE_RETURN))) {
3480                 /* undo NAT changes, if they have taken place */
3481                 if (nr != NULL) {
3482                         PF_ACPY(saddr, &sk->addr[pd->sidx], af);
3483                         PF_ACPY(daddr, &sk->addr[pd->didx], af);
3484                         if (pd->sport)
3485                                 *pd->sport = sk->port[pd->sidx];
3486                         if (pd->dport)
3487                                 *pd->dport = sk->port[pd->didx];
3488                         if (pd->proto_sum)
3489                                 *pd->proto_sum = bproto_sum;
3490                         if (pd->ip_sum)
3491                                 *pd->ip_sum = bip_sum;
3492                         m_copyback(m, off, hdrlen, pd->hdr.any);
3493                 }
3494                 if (pd->proto == IPPROTO_TCP &&
3495                     ((r->rule_flag & PFRULE_RETURNRST) ||
3496                     (r->rule_flag & PFRULE_RETURN)) &&
3497                     !(th->th_flags & TH_RST)) {
3498                         u_int32_t        ack = ntohl(th->th_seq) + pd->p_len;
3499                         int              len = 0;
3500                         struct ip       *h4;
3501                         struct ip6_hdr  *h6;
3502
3503                         switch (af) {
3504                         case AF_INET:
3505                                 h4 = mtod(m, struct ip *);
3506                                 len = h4->ip_len - off;
3507                                 break;
3508 #ifdef INET6
3509                         case AF_INET6:
3510                                 h6 = mtod(m, struct ip6_hdr *);
3511                                 len = h6->ip6_plen - (off - sizeof(*h6));
3512                                 break;
3513 #endif
3514                         }
3515
3516                         if (pf_check_proto_cksum(m, off, len, IPPROTO_TCP, af))
3517                                 REASON_SET(&reason, PFRES_PROTCKSUM);
3518                         else {
3519                                 if (th->th_flags & TH_SYN)
3520                                         ack++;
3521                                 if (th->th_flags & TH_FIN)
3522                                         ack++;
3523                                 pf_send_tcp(r, af, pd->dst,
3524                                     pd->src, th->th_dport, th->th_sport,
3525                                     ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0,
3526                                     r->return_ttl, 1, 0, pd->eh, kif->pfik_ifp);
3527                         }
3528                 } else if (pd->proto != IPPROTO_ICMP && af == AF_INET &&
3529                     r->return_icmp)
3530                         pf_send_icmp(m, r->return_icmp >> 8,
3531                             r->return_icmp & 255, af, r);
3532                 else if (pd->proto != IPPROTO_ICMPV6 && af == AF_INET6 &&
3533                     r->return_icmp6)
3534                         pf_send_icmp(m, r->return_icmp6 >> 8,
3535                             r->return_icmp6 & 255, af, r);
3536         }
3537
3538         if (r->action == PF_DROP)
3539                 goto cleanup;
3540
3541         if (pf_tag_packet(m, tag, rtableid)) {
3542                 REASON_SET(&reason, PFRES_MEMORY);
3543                 goto cleanup;
3544         }
3545
3546         if (!state_icmp && (r->keep_state || nr != NULL ||
3547             (pd->flags & PFDESC_TCP_NORM))) {
3548                 int action;
3549                 action = pf_create_state(r, nr, a, pd, nsn, skw, sks, nk, sk, m,
3550                     off, sport, dport, &rewrite, kif, sm, tag, bproto_sum,
3551                     bip_sum, hdrlen);
3552                 if (action != PF_PASS)
3553                         return (action);
3554         }
3555
3556         /* copy back packet headers if we performed NAT operations */
3557         if (rewrite)
3558                 m_copyback(m, off, hdrlen, pd->hdr.any);
3559
3560         return (PF_PASS);
3561
3562 cleanup:
3563         if (sk != NULL)
3564                 pool_put(&pf_state_key_pl, sk);
3565         if (nk != NULL)
3566                 pool_put(&pf_state_key_pl, nk);
3567         return (PF_DROP);
3568 }
3569
3570 static __inline int
3571 pf_create_state(struct pf_rule *r, struct pf_rule *nr, struct pf_rule *a,
3572     struct pf_pdesc *pd, struct pf_src_node *nsn, struct pf_state_key *skw,
3573     struct pf_state_key *sks, struct pf_state_key *nk, struct pf_state_key *sk,
3574     struct mbuf *m, int off, u_int16_t sport, u_int16_t dport, int *rewrite,
3575     struct pfi_kif *kif, struct pf_state **sm, int tag, u_int16_t bproto_sum,
3576     u_int16_t bip_sum, int hdrlen)
3577 {
3578         struct pf_state         *s = NULL;
3579         struct pf_src_node      *sn = NULL;
3580         struct tcphdr           *th = pd->hdr.tcp;
3581         u_int16_t                mss = tcp_mssdflt;
3582         u_short                  reason;
3583
3584         /* check maximums */
3585         if (r->max_states && (r->states_cur >= r->max_states)) {
3586                 pf_status.lcounters[LCNT_STATES]++;
3587                 REASON_SET(&reason, PFRES_MAXSTATES);
3588                 return (PF_DROP);
3589         }
3590         /* src node for filter rule */
3591         if ((r->rule_flag & PFRULE_SRCTRACK ||
3592             r->rpool.opts & PF_POOL_STICKYADDR) &&
3593             pf_insert_src_node(&sn, r, pd->src, pd->af) != 0) {
3594                 REASON_SET(&reason, PFRES_SRCLIMIT);
3595                 goto csfailed;
3596         }
3597         /* src node for translation rule */
3598         if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3599             pf_insert_src_node(&nsn, nr, &sk->addr[pd->sidx], pd->af)) {
3600                 REASON_SET(&reason, PFRES_SRCLIMIT);
3601                 goto csfailed;
3602         }
3603         s = pool_get(&pf_state_pl, PR_NOWAIT | PR_ZERO);
3604         if (s == NULL) {
3605                 REASON_SET(&reason, PFRES_MEMORY);
3606                 goto csfailed;
3607         }
3608         s->id = 0; /* XXX Do we really need that? not in OpenBSD */
3609         s->creatorid = 0;
3610         s->rule.ptr = r;
3611         s->nat_rule.ptr = nr;
3612         s->anchor.ptr = a;
3613         STATE_INC_COUNTERS(s);
3614         if (r->allow_opts)
3615                 s->state_flags |= PFSTATE_ALLOWOPTS;
3616         if (r->rule_flag & PFRULE_STATESLOPPY)
3617                 s->state_flags |= PFSTATE_SLOPPY;
3618         s->log = r->log & PF_LOG_ALL;
3619         if (nr != NULL)
3620                 s->log |= nr->log & PF_LOG_ALL;
3621         switch (pd->proto) {
3622         case IPPROTO_TCP:
3623                 s->src.seqlo = ntohl(th->th_seq);
3624                 s->src.seqhi = s->src.seqlo + pd->p_len + 1;
3625                 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
3626                     r->keep_state == PF_STATE_MODULATE) {
3627                         /* Generate sequence number modulator */
3628                         if ((s->src.seqdiff = pf_tcp_iss(pd) - s->src.seqlo) ==
3629                             0)
3630                                 s->src.seqdiff = 1;
3631                         pf_change_a(&th->th_seq, &th->th_sum,
3632                             htonl(s->src.seqlo + s->src.seqdiff), 0);
3633                         *rewrite = 1;
3634                 } else
3635                         s->src.seqdiff = 0;
3636                 if (th->th_flags & TH_SYN) {
3637                         s->src.seqhi++;
3638                         s->src.wscale = pf_get_wscale(m, off,
3639                             th->th_off, pd->af);
3640                 }
3641                 s->src.max_win = MAX(ntohs(th->th_win), 1);
3642                 if (s->src.wscale & PF_WSCALE_MASK) {
3643                         /* Remove scale factor from initial window */
3644                         int win = s->src.max_win;
3645                         win += 1 << (s->src.wscale & PF_WSCALE_MASK);
3646                         s->src.max_win = (win - 1) >>
3647                             (s->src.wscale & PF_WSCALE_MASK);
3648                 }
3649                 if (th->th_flags & TH_FIN)
3650                         s->src.seqhi++;
3651                 s->dst.seqhi = 1;
3652                 s->dst.max_win = 1;
3653                 s->src.state = TCPS_SYN_SENT;
3654                 s->dst.state = TCPS_CLOSED;
3655                 s->timeout = PFTM_TCP_FIRST_PACKET;
3656                 break;
3657         case IPPROTO_UDP:
3658                 s->src.state = PFUDPS_SINGLE;
3659                 s->dst.state = PFUDPS_NO_TRAFFIC;
3660                 s->timeout = PFTM_UDP_FIRST_PACKET;
3661                 break;
3662         case IPPROTO_ICMP:
3663 #ifdef INET6
3664         case IPPROTO_ICMPV6:
3665 #endif
3666                 s->timeout = PFTM_ICMP_FIRST_PACKET;
3667                 break;
3668         default:
3669                 s->src.state = PFOTHERS_SINGLE;
3670                 s->dst.state = PFOTHERS_NO_TRAFFIC;
3671                 s->timeout = PFTM_OTHER_FIRST_PACKET;
3672         }
3673
3674         s->creation = time_second;
3675         s->expire = time_second;
3676
3677         if (sn != NULL) {
3678                 s->src_node = sn;
3679                 s->src_node->states++;
3680         }
3681         if (nsn != NULL) {
3682                 /* XXX We only modify one side for now. */
3683                 PF_ACPY(&nsn->raddr, &nk->addr[1], pd->af);
3684                 s->nat_src_node = nsn;
3685                 s->nat_src_node->states++;
3686         }
3687         if (pd->proto == IPPROTO_TCP) {
3688                 if ((pd->flags & PFDESC_TCP_NORM) && pf_normalize_tcp_init(m,
3689                     off, pd, th, &s->src, &s->dst)) {
3690                         REASON_SET(&reason, PFRES_MEMORY);
3691                         pf_src_tree_remove_state(s);
3692                         STATE_DEC_COUNTERS(s);
3693                         pool_put(&pf_state_pl, s);
3694                         return (PF_DROP);
3695                 }
3696                 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
3697                     pf_normalize_tcp_stateful(m, off, pd, &reason, th, s,
3698                     &s->src, &s->dst, rewrite)) {
3699                         /* This really shouldn't happen!!! */
3700                         DPFPRINTF(PF_DEBUG_URGENT,
3701                             ("pf_normalize_tcp_stateful failed on first pkt"));
3702                         pf_normalize_tcp_cleanup(s);
3703                         pf_src_tree_remove_state(s);
3704                         STATE_DEC_COUNTERS(s);
3705                         pool_put(&pf_state_pl, s);
3706                         return (PF_DROP);
3707                 }
3708         }
3709         s->direction = pd->dir;
3710
3711         if (sk == NULL && pf_state_key_setup(pd, nr, &skw, &sks, &sk, &nk,
3712             pd->src, pd->dst, sport, dport))
3713                 goto csfailed;
3714
3715         if (pf_state_insert(BOUND_IFACE(r, kif), skw, sks, s)) {
3716                 if (pd->proto == IPPROTO_TCP)
3717                         pf_normalize_tcp_cleanup(s);
3718                 REASON_SET(&reason, PFRES_STATEINS);
3719                 pf_src_tree_remove_state(s);
3720                 STATE_DEC_COUNTERS(s);
3721                 pool_put(&pf_state_pl, s);
3722                 return (PF_DROP);
3723         } else
3724                 *sm = s;
3725
3726         pf_set_rt_ifp(s, pd->src);      /* needs s->state_key set */
3727         if (tag > 0) {
3728                 pf_tag_ref(tag);
3729                 s->tag = tag;
3730         }
3731         if (pd->proto == IPPROTO_TCP && (th->th_flags & (TH_SYN|TH_ACK)) ==
3732             TH_SYN && r->keep_state == PF_STATE_SYNPROXY) {
3733                 s->src.state = PF_TCPS_PROXY_SRC;
3734                 /* undo NAT changes, if they have taken place */
3735                 if (nr != NULL) {
3736                         struct pf_state_key *skt = s->key[PF_SK_WIRE];
3737                         if (pd->dir == PF_OUT)
3738                                 skt = s->key[PF_SK_STACK];
3739                         PF_ACPY(pd->src, &skt->addr[pd->sidx], pd->af);
3740                         PF_ACPY(pd->dst, &skt->addr[pd->didx], pd->af);
3741                         if (pd->sport)
3742                                 *pd->sport = skt->port[pd->sidx];
3743                         if (pd->dport)
3744                                 *pd->dport = skt->port[pd->didx];
3745                         if (pd->proto_sum)
3746                                 *pd->proto_sum = bproto_sum;
3747                         if (pd->ip_sum)
3748                                 *pd->ip_sum = bip_sum;
3749                         m_copyback(m, off, hdrlen, pd->hdr.any);
3750                 }
3751                 s->src.seqhi = htonl(karc4random());
3752                 /* Find mss option */
3753                 mss = pf_get_mss(m, off, th->th_off, pd->af);
3754                 mss = pf_calc_mss(pd->src, pd->af, mss);
3755                 mss = pf_calc_mss(pd->dst, pd->af, mss);
3756                 s->src.mss = mss;
3757                 pf_send_tcp(r, pd->af, pd->dst, pd->src, th->th_dport,
3758                     th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1,
3759                     TH_SYN|TH_ACK, 0, s->src.mss, 0, 1, 0, NULL, NULL);
3760                 REASON_SET(&reason, PFRES_SYNPROXY);
3761                 return (PF_SYNPROXY_DROP);
3762         }
3763
3764         return (PF_PASS);
3765
3766 csfailed:
3767         if (sk != NULL)
3768                 pool_put(&pf_state_key_pl, sk);
3769         if (nk != NULL)
3770                 pool_put(&pf_state_key_pl, nk);
3771
3772         if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3773                 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3774                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3775                 pf_status.src_nodes--;
3776                 pool_put(&pf_src_tree_pl, sn);
3777         }
3778         if (nsn != sn && nsn != NULL && nsn->states == 0 && nsn->expire == 0) {
3779                 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3780                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3781                 pf_status.src_nodes--;
3782                 pool_put(&pf_src_tree_pl, nsn);
3783         }
3784         return (PF_DROP);
3785 }
3786
3787 int
3788 pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif,
3789     struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am,
3790     struct pf_ruleset **rsm)
3791 {
3792         struct pf_rule          *r, *a = NULL;
3793         struct pf_ruleset       *ruleset = NULL;
3794         sa_family_t              af = pd->af;
3795         u_short                  reason;
3796         int                      tag = -1;
3797         int                      asd = 0;
3798         int                      match = 0;
3799
3800         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3801         while (r != NULL) {
3802                 r->evaluations++;
3803                 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3804                         r = r->skip[PF_SKIP_IFP].ptr;
3805                 else if (r->direction && r->direction != direction)
3806                         r = r->skip[PF_SKIP_DIR].ptr;
3807                 else if (r->af && r->af != af)
3808                         r = r->skip[PF_SKIP_AF].ptr;
3809                 else if (r->proto && r->proto != pd->proto)
3810                         r = r->skip[PF_SKIP_PROTO].ptr;
3811                 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af,
3812                     r->src.neg, kif))
3813                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3814                 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af,
3815                     r->dst.neg, NULL))
3816                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
3817                 else if (r->tos && !(r->tos == pd->tos))
3818                         r = TAILQ_NEXT(r, entries);
3819                 else if (r->os_fingerprint != PF_OSFP_ANY)
3820                         r = TAILQ_NEXT(r, entries);
3821                 else if (pd->proto == IPPROTO_UDP &&
3822                     (r->src.port_op || r->dst.port_op))
3823                         r = TAILQ_NEXT(r, entries);
3824                 else if (pd->proto == IPPROTO_TCP &&
3825                     (r->src.port_op || r->dst.port_op || r->flagset))
3826                         r = TAILQ_NEXT(r, entries);
3827                 else if ((pd->proto == IPPROTO_ICMP ||
3828                     pd->proto == IPPROTO_ICMPV6) &&
3829                     (r->type || r->code))
3830                         r = TAILQ_NEXT(r, entries);
3831                 else if (r->prob && r->prob <= karc4random())
3832                         r = TAILQ_NEXT(r, entries);
3833                 else if (r->match_tag && !pf_match_tag(m, r, &tag))
3834                         r = TAILQ_NEXT(r, entries);
3835                 else {
3836                         if (r->anchor == NULL) {
3837                                 match = 1;
3838                                 *rm = r;
3839                                 *am = a;
3840                                 *rsm = ruleset;
3841                                 if ((*rm)->quick)
3842                                         break;
3843                                 r = TAILQ_NEXT(r, entries);
3844                         } else
3845                                 pf_step_into_anchor(&asd, &ruleset,
3846                                     PF_RULESET_FILTER, &r, &a, &match);
3847                 }
3848                 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3849                     PF_RULESET_FILTER, &r, &a, &match))
3850                         break;
3851         }
3852         r = *rm;
3853         a = *am;
3854         ruleset = *rsm;
3855
3856         REASON_SET(&reason, PFRES_MATCH);
3857
3858         if (r->log)
3859                 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset,
3860                     pd);
3861
3862         if (r->action != PF_PASS)
3863                 return (PF_DROP);
3864
3865         if (pf_tag_packet(m, tag, -1)) {
3866                 REASON_SET(&reason, PFRES_MEMORY);
3867                 return (PF_DROP);
3868         }
3869
3870         return (PF_PASS);
3871 }
3872
3873 int
3874 pf_tcp_track_full(struct pf_state_peer *src, struct pf_state_peer *dst,
3875         struct pf_state **state, struct pfi_kif *kif, struct mbuf *m, int off,
3876         struct pf_pdesc *pd, u_short *reason, int *copyback)
3877 {
3878         struct tcphdr           *th = pd->hdr.tcp;
3879         u_int16_t                win = ntohs(th->th_win);
3880         u_int32_t                ack, end, seq, orig_seq;
3881         u_int8_t                 sws, dws;
3882         int                      ackskew;
3883
3884         if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) {
3885                 sws = src->wscale & PF_WSCALE_MASK;
3886                 dws = dst->wscale & PF_WSCALE_MASK;
3887         } else
3888                 sws = dws = 0;
3889
3890         /*
3891          * Sequence tracking algorithm from Guido van Rooij's paper:
3892          *   http://www.madison-gurkha.com/publications/tcp_filtering/
3893          *      tcp_filtering.ps
3894          */
3895
3896         orig_seq = seq = ntohl(th->th_seq);
3897         if (src->seqlo == 0) {
3898                 /* First packet from this end. Set its state */
3899
3900                 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) &&
3901                     src->scrub == NULL) {
3902                         if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) {
3903                                 REASON_SET(reason, PFRES_MEMORY);
3904                                 return (PF_DROP);
3905                         }
3906                 }
3907
3908                 /* Deferred generation of sequence number modulator */
3909                 if (dst->seqdiff && !src->seqdiff) {
3910                         /* use random iss for the TCP server */
3911                         while ((src->seqdiff = karc4random() - seq) == 0)
3912                                 ;
3913                         ack = ntohl(th->th_ack) - dst->seqdiff;
3914                         pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3915                             src->seqdiff), 0);
3916                         pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3917                         *copyback = 1;
3918                 } else {
3919                         ack = ntohl(th->th_ack);
3920                 }
3921
3922                 end = seq + pd->p_len;
3923                 if (th->th_flags & TH_SYN) {
3924                         end++;
3925                         (*state)->sync_flags |= PFSTATE_GOT_SYN2;
3926                         if (dst->wscale & PF_WSCALE_FLAG) {
3927                                 src->wscale = pf_get_wscale(m, off, th->th_off,
3928                                     pd->af);
3929                                 if (src->wscale & PF_WSCALE_FLAG) {
3930                                         /* Remove scale factor from initial
3931                                          * window */
3932                                         sws = src->wscale & PF_WSCALE_MASK;
3933                                         win = ((u_int32_t)win + (1 << sws) - 1)
3934                                             >> sws;
3935                                         dws = dst->wscale & PF_WSCALE_MASK;
3936                                 } else {
3937                                         /* fixup other window */
3938                                         dst->max_win <<= dst->wscale &
3939                                             PF_WSCALE_MASK;
3940                                         /* in case of a retrans SYN|ACK */
3941                                         dst->wscale = 0;
3942                                 }
3943                         }
3944                 }
3945                 if (th->th_flags & TH_FIN)
3946                         end++;
3947
3948                 src->seqlo = seq;
3949                 if (src->state < TCPS_SYN_SENT)
3950                         src->state = TCPS_SYN_SENT;
3951
3952                 /*
3953                  * May need to slide the window (seqhi may have been set by
3954                  * the crappy stack check or if we picked up the connection
3955                  * after establishment)
3956                  */
3957                 if (src->seqhi == 1 ||
3958                     SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi))
3959                         src->seqhi = end + MAX(1, dst->max_win << dws);
3960                 if (win > src->max_win)
3961                         src->max_win = win;
3962
3963         } else {
3964                 ack = ntohl(th->th_ack) - dst->seqdiff;
3965                 if (src->seqdiff) {
3966                         /* Modulate sequence numbers */
3967                         pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3968                             src->seqdiff), 0);
3969                         pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3970                         *copyback = 1;
3971                 }
3972                 end = seq + pd->p_len;
3973                 if (th->th_flags & TH_SYN)
3974                         end++;
3975                 if (th->th_flags & TH_FIN)
3976                         end++;
3977         }
3978
3979         if ((th->th_flags & TH_ACK) == 0) {
3980                 /* Let it pass through the ack skew check */
3981                 ack = dst->seqlo;
3982         } else if ((ack == 0 &&
3983             (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) ||
3984             /* broken tcp stacks do not set ack */
3985             (dst->state < TCPS_SYN_SENT)) {
3986                 /*
3987                  * Many stacks (ours included) will set the ACK number in an
3988                  * FIN|ACK if the SYN times out -- no sequence to ACK.
3989                  */
3990                 ack = dst->seqlo;
3991         }
3992
3993         if (seq == end) {
3994                 /* Ease sequencing restrictions on no data packets */
3995                 seq = src->seqlo;
3996                 end = seq;
3997         }
3998
3999         ackskew = dst->seqlo - ack;
4000
4001
4002         /*
4003          * Need to demodulate the sequence numbers in any TCP SACK options
4004          * (Selective ACK). We could optionally validate the SACK values
4005          * against the current ACK window, either forwards or backwards, but
4006          * I'm not confident that SACK has been implemented properly
4007          * everywhere. It wouldn't surprise me if several stacks accidently
4008          * SACK too far backwards of previously ACKed data. There really aren't
4009          * any security implications of bad SACKing unless the target stack
4010          * doesn't validate the option length correctly. Someone trying to
4011          * spoof into a TCP connection won't bother blindly sending SACK
4012          * options anyway.
4013          */
4014         if (dst->seqdiff && (th->th_off << 2) > sizeof(struct tcphdr)) {
4015                 if (pf_modulate_sack(m, off, pd, th, dst))
4016                         *copyback = 1;
4017         }
4018
4019
4020 #define MAXACKWINDOW (0xffff + 1500)    /* 1500 is an arbitrary fudge factor */
4021         if (SEQ_GEQ(src->seqhi, end) &&
4022             /* Last octet inside other's window space */
4023             SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) &&
4024             /* Retrans: not more than one window back */
4025             (ackskew >= -MAXACKWINDOW) &&
4026             /* Acking not more than one reassembled fragment backwards */
4027             (ackskew <= (MAXACKWINDOW << sws)) &&
4028             /* Acking not more than one window forward */
4029             ((th->th_flags & TH_RST) == 0 || orig_seq == src->seqlo ||
4030             (orig_seq == src->seqlo + 1) || (orig_seq + 1 == src->seqlo) ||
4031             (pd->flags & PFDESC_IP_REAS) == 0)) {
4032             /* Require an exact/+1 sequence match on resets when possible */
4033
4034                 if (dst->scrub || src->scrub) {
4035                         if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4036                             *state, src, dst, copyback))
4037                                 return (PF_DROP);
4038                 }
4039
4040                 /* update max window */
4041                 if (src->max_win < win)
4042                         src->max_win = win;
4043                 /* synchronize sequencing */
4044                 if (SEQ_GT(end, src->seqlo))
4045                         src->seqlo = end;
4046                 /* slide the window of what the other end can send */
4047                 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4048                         dst->seqhi = ack + MAX((win << sws), 1);
4049
4050
4051                 /* update states */
4052                 if (th->th_flags & TH_SYN)
4053                         if (src->state < TCPS_SYN_SENT)
4054                                 src->state = TCPS_SYN_SENT;
4055                 if (th->th_flags & TH_FIN)
4056                         if (src->state < TCPS_CLOSING)
4057                                 src->state = TCPS_CLOSING;
4058                 if (th->th_flags & TH_ACK) {
4059                         if (dst->state == TCPS_SYN_SENT) {
4060                                 dst->state = TCPS_ESTABLISHED;
4061                                 if (src->state == TCPS_ESTABLISHED &&
4062                                     (*state)->src_node != NULL &&
4063                                     pf_src_connlimit(state)) {
4064                                         REASON_SET(reason, PFRES_SRCLIMIT);
4065                                         return (PF_DROP);
4066                                 }
4067                         } else if (dst->state == TCPS_CLOSING)
4068                                 dst->state = TCPS_FIN_WAIT_2;
4069                 }
4070                 if (th->th_flags & TH_RST)
4071                         src->state = dst->state = TCPS_TIME_WAIT;
4072
4073                 /* update expire time */
4074                 (*state)->expire = time_second;
4075                 if (src->state >= TCPS_FIN_WAIT_2 &&
4076                     dst->state >= TCPS_FIN_WAIT_2)
4077                         (*state)->timeout = PFTM_TCP_CLOSED;
4078                 else if (src->state >= TCPS_CLOSING &&
4079                     dst->state >= TCPS_CLOSING)
4080                         (*state)->timeout = PFTM_TCP_FIN_WAIT;
4081                 else if (src->state < TCPS_ESTABLISHED ||
4082                     dst->state < TCPS_ESTABLISHED)
4083                         (*state)->timeout = PFTM_TCP_OPENING;
4084                 else if (src->state >= TCPS_CLOSING ||
4085                     dst->state >= TCPS_CLOSING)
4086                         (*state)->timeout = PFTM_TCP_CLOSING;
4087                 else
4088                         (*state)->timeout = PFTM_TCP_ESTABLISHED;
4089
4090                 /* Fall through to PASS packet */
4091
4092         } else if ((dst->state < TCPS_SYN_SENT ||
4093                 dst->state >= TCPS_FIN_WAIT_2 ||
4094                 src->state >= TCPS_FIN_WAIT_2) &&
4095             SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) &&
4096             /* Within a window forward of the originating packet */
4097             SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) {
4098             /* Within a window backward of the originating packet */
4099
4100                 /*
4101                  * This currently handles three situations:
4102                  *  1) Stupid stacks will shotgun SYNs before their peer
4103                  *     replies.
4104                  *  2) When PF catches an already established stream (the
4105                  *     firewall rebooted, the state table was flushed, routes
4106                  *     changed...)
4107                  *  3) Packets get funky immediately after the connection
4108                  *     closes (this should catch Solaris spurious ACK|FINs
4109                  *     that web servers like to spew after a close)
4110                  *
4111                  * This must be a little more careful than the above code
4112                  * since packet floods will also be caught here. We don't
4113                  * update the TTL here to mitigate the damage of a packet
4114                  * flood and so the same code can handle awkward establishment
4115                  * and a loosened connection close.
4116                  * In the establishment case, a correct peer response will
4117                  * validate the connection, go through the normal state code
4118                  * and keep updating the state TTL.
4119                  */
4120
4121                 if (pf_status.debug >= PF_DEBUG_MISC) {
4122                         kprintf("pf: loose state match: ");
4123                         pf_print_state(*state);
4124                         pf_print_flags(th->th_flags);
4125                         kprintf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
4126                             "pkts=%llu:%llu dir=%s,%s\n", seq, orig_seq, ack, pd->p_len,
4127                             ackskew, (unsigned long long)(*state)->packets[0],
4128                             (unsigned long long)(*state)->packets[1],
4129                             pd->dir == PF_IN ? "in" : "out",
4130                             pd->dir == (*state)->direction ? "fwd" : "rev");
4131                 }
4132
4133                 if (dst->scrub || src->scrub) {
4134                         if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4135                             *state, src, dst, copyback))
4136                                 return (PF_DROP);
4137                 }
4138
4139                 /* update max window */
4140                 if (src->max_win < win)
4141                         src->max_win = win;
4142                 /* synchronize sequencing */
4143                 if (SEQ_GT(end, src->seqlo))
4144                         src->seqlo = end;
4145                 /* slide the window of what the other end can send */
4146                 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4147                         dst->seqhi = ack + MAX((win << sws), 1);
4148
4149                 /*
4150                  * Cannot set dst->seqhi here since this could be a shotgunned
4151                  * SYN and not an already established connection.
4152                  */
4153
4154                 if (th->th_flags & TH_FIN)
4155                         if (src->state < TCPS_CLOSING)
4156                                 src->state = TCPS_CLOSING;
4157                 if (th->th_flags & TH_RST)
4158                         src->state = dst->state = TCPS_TIME_WAIT;
4159
4160                 /* Fall through to PASS packet */
4161
4162         } else if ((*state)->pickup_mode == PF_PICKUPS_HASHONLY ||
4163                     ((*state)->pickup_mode == PF_PICKUPS_ENABLED &&
4164                      ((*state)->sync_flags & PFSTATE_GOT_SYN_MASK) !=
4165                       PFSTATE_GOT_SYN_MASK)) {
4166                 /*
4167                  * If pickup mode is hash only, do not fail on sequence checks.
4168                  *
4169                  * If pickup mode is enabled and we did not see the SYN in
4170                  * both direction, do not fail on sequence checks because
4171                  * we do not have complete information on window scale.
4172                  *
4173                  * Adjust expiration and fall through to PASS packet.
4174                  * XXX Add a FIN check to reduce timeout?
4175                  */
4176                 (*state)->expire = time_second;
4177         } else  {
4178                 /*
4179                  * Failure processing
4180                  */
4181                 if ((*state)->dst.state == TCPS_SYN_SENT &&
4182                     (*state)->src.state == TCPS_SYN_SENT) {
4183                         /* Send RST for state mismatches during handshake */
4184                         if (!(th->th_flags & TH_RST))
4185                                 pf_send_tcp((*state)->rule.ptr, pd->af,
4186                                     pd->dst, pd->src, th->th_dport,
4187                                     th->th_sport, ntohl(th->th_ack), 0,
4188                                     TH_RST, 0, 0,
4189                                     (*state)->rule.ptr->return_ttl, 1, 0,
4190                                     pd->eh, kif->pfik_ifp);
4191                         src->seqlo = 0;
4192                         src->seqhi = 1;
4193                         src->max_win = 1;
4194                 } else if (pf_status.debug >= PF_DEBUG_MISC) {
4195                         kprintf("pf: BAD state: ");
4196                         pf_print_state(*state);
4197                         pf_print_flags(th->th_flags);
4198                         kprintf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
4199                             "pkts=%llu:%llu dir=%s,%s\n",
4200                             seq, orig_seq, ack, pd->p_len, ackskew,
4201                             (unsigned long long)(*state)->packets[0],
4202                                 (unsigned long long)(*state)->packets[1],
4203                             pd->dir == PF_IN ? "in" : "out",
4204                             pd->dir == (*state)->direction ? "fwd" : "rev");
4205                         kprintf("pf: State failure on: %c %c %c %c | %c %c\n",
4206                             SEQ_GEQ(src->seqhi, end) ? ' ' : '1',
4207                             SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ?
4208                             ' ': '2',
4209                             (ackskew >= -MAXACKWINDOW) ? ' ' : '3',
4210                             (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4',
4211                             SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5',
4212                             SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6');
4213                 }
4214                 REASON_SET(reason, PFRES_BADSTATE);
4215                 return (PF_DROP);
4216         }
4217
4218         return (PF_PASS);
4219 }
4220
4221 int
4222 pf_tcp_track_sloppy(struct pf_state_peer *src, struct pf_state_peer *dst,
4223         struct pf_state **state, struct pf_pdesc *pd, u_short *reason)
4224 {
4225         struct tcphdr           *th = pd->hdr.tcp;
4226
4227         if (th->th_flags & TH_SYN)
4228                 if (src->state < TCPS_SYN_SENT)
4229                         src->state = TCPS_SYN_SENT;
4230         if (th->th_flags & TH_FIN)
4231                 if (src->state < TCPS_CLOSING)
4232                         src->state = TCPS_CLOSING;
4233         if (th->th_flags & TH_ACK) {
4234                 if (dst->state == TCPS_SYN_SENT) {
4235                         dst->state = TCPS_ESTABLISHED;
4236                         if (src->state == TCPS_ESTABLISHED &&
4237                             (*state)->src_node != NULL &&
4238                             pf_src_connlimit(state)) {
4239                                 REASON_SET(reason, PFRES_SRCLIMIT);
4240                                 return (PF_DROP);
4241                         }
4242                 } else if (dst->state == TCPS_CLOSING) {
4243                         dst->state = TCPS_FIN_WAIT_2;
4244                 } else if (src->state == TCPS_SYN_SENT &&
4245                     dst->state < TCPS_SYN_SENT) {
4246                         /*
4247                          * Handle a special sloppy case where we only see one
4248                          * half of the connection. If there is a ACK after
4249                          * the initial SYN without ever seeing a packet from
4250                          * the destination, set the connection to established.
4251                          */
4252                         dst->state = src->state = TCPS_ESTABLISHED;
4253                         if ((*state)->src_node != NULL &&
4254                             pf_src_connlimit(state)) {
4255                                 REASON_SET(reason, PFRES_SRCLIMIT);
4256                                 return (PF_DROP);
4257                         }
4258                 } else if (src->state == TCPS_CLOSING &&
4259                     dst->state == TCPS_ESTABLISHED &&
4260                     dst->seqlo == 0) {
4261                         /*
4262                          * Handle the closing of half connections where we
4263                          * don't see the full bidirectional FIN/ACK+ACK
4264                          * handshake.
4265                          */
4266                         dst->state = TCPS_CLOSING;
4267                 }
4268         }
4269         if (th->th_flags & TH_RST)
4270                 src->state = dst->state = TCPS_TIME_WAIT;
4271
4272         /* update expire time */
4273         (*state)->expire = time_second;
4274         if (src->state >= TCPS_FIN_WAIT_2 &&
4275             dst->state >= TCPS_FIN_WAIT_2)
4276                 (*state)->timeout = PFTM_TCP_CLOSED;
4277         else if (src->state >= TCPS_CLOSING &&
4278             dst->state >= TCPS_CLOSING)
4279                 (*state)->timeout = PFTM_TCP_FIN_WAIT;
4280         else if (src->state < TCPS_ESTABLISHED ||
4281             dst->state < TCPS_ESTABLISHED)
4282                 (*state)->timeout = PFTM_TCP_OPENING;
4283         else if (src->state >= TCPS_CLOSING ||
4284             dst->state >= TCPS_CLOSING)
4285                 (*state)->timeout = PFTM_TCP_CLOSING;
4286         else
4287                 (*state)->timeout = PFTM_TCP_ESTABLISHED;
4288
4289         return (PF_PASS);
4290 }
4291
4292 int
4293 pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
4294     struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
4295     u_short *reason)
4296 {
4297         struct pf_state_key_cmp  key;
4298         struct tcphdr           *th = pd->hdr.tcp;
4299         int                      copyback = 0;
4300         struct pf_state_peer    *src, *dst;
4301         struct pf_state_key     *sk;
4302
4303         key.af = pd->af;
4304         key.proto = IPPROTO_TCP;
4305         if (direction == PF_IN) {       /* wire side, straight */
4306                 PF_ACPY(&key.addr[0], pd->src, key.af);
4307                 PF_ACPY(&key.addr[1], pd->dst, key.af);
4308                 key.port[0] = th->th_sport;
4309                 key.port[1] = th->th_dport;
4310         } else {                        /* stack side, reverse */
4311                 PF_ACPY(&key.addr[1], pd->src, key.af);
4312                 PF_ACPY(&key.addr[0], pd->dst, key.af);
4313                 key.port[1] = th->th_sport;
4314                 key.port[0] = th->th_dport;
4315         }
4316
4317         STATE_LOOKUP(kif, &key, direction, *state, m);
4318
4319         if (direction == (*state)->direction) {
4320                 src = &(*state)->src;
4321                 dst = &(*state)->dst;
4322         } else {
4323                 src = &(*state)->dst;
4324                 dst = &(*state)->src;
4325         }
4326
4327         sk = (*state)->key[pd->didx];
4328
4329         if ((*state)->src.state == PF_TCPS_PROXY_SRC) {
4330                 if (direction != (*state)->direction) {
4331                         REASON_SET(reason, PFRES_SYNPROXY);
4332                         return (PF_SYNPROXY_DROP);
4333                 }
4334                 if (th->th_flags & TH_SYN) {
4335                         if (ntohl(th->th_seq) != (*state)->src.seqlo) {
4336                                 REASON_SET(reason, PFRES_SYNPROXY);
4337                                 return (PF_DROP);
4338                         }
4339                         pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
4340                             pd->src, th->th_dport, th->th_sport,
4341                             (*state)->src.seqhi, ntohl(th->th_seq) + 1,
4342                             TH_SYN|TH_ACK, 0, (*state)->src.mss, 0, 1,
4343                             0, NULL, NULL);
4344                         REASON_SET(reason, PFRES_SYNPROXY);
4345                         return (PF_SYNPROXY_DROP);
4346                 } else if (!(th->th_flags & TH_ACK) ||
4347                     (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
4348                     (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
4349                         REASON_SET(reason, PFRES_SYNPROXY);
4350                         return (PF_DROP);
4351                 } else if ((*state)->src_node != NULL &&
4352                     pf_src_connlimit(state)) {
4353                         REASON_SET(reason, PFRES_SRCLIMIT);
4354                         return (PF_DROP);
4355                 } else
4356                         (*state)->src.state = PF_TCPS_PROXY_DST;
4357         }
4358         if ((*state)->src.state == PF_TCPS_PROXY_DST) {
4359                 if (direction == (*state)->direction) {
4360                         if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) ||
4361                             (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
4362                             (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
4363                                 REASON_SET(reason, PFRES_SYNPROXY);
4364                                 return (PF_DROP);
4365                         }
4366                         (*state)->src.max_win = MAX(ntohs(th->th_win), 1);
4367                         if ((*state)->dst.seqhi == 1)
4368                                 (*state)->dst.seqhi = htonl(karc4random());
4369                         pf_send_tcp((*state)->rule.ptr, pd->af,
4370                             &sk->addr[pd->sidx], &sk->addr[pd->didx],
4371                             sk->port[pd->sidx], sk->port[pd->didx],
4372                             (*state)->dst.seqhi, 0, TH_SYN, 0,
4373                             (*state)->src.mss, 0, 0, (*state)->tag, NULL, NULL);
4374                         REASON_SET(reason, PFRES_SYNPROXY);
4375                         return (PF_SYNPROXY_DROP);
4376                 } else if (((th->th_flags & (TH_SYN|TH_ACK)) !=
4377                     (TH_SYN|TH_ACK)) ||
4378                     (ntohl(th->th_ack) != (*state)->dst.seqhi + 1)) {
4379                         REASON_SET(reason, PFRES_SYNPROXY);
4380                         return (PF_DROP);
4381                 } else {
4382                         (*state)->dst.max_win = MAX(ntohs(th->th_win), 1);
4383                         (*state)->dst.seqlo = ntohl(th->th_seq);
4384                         pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
4385                             pd->src, th->th_dport, th->th_sport,
4386                             ntohl(th->th_ack), ntohl(th->th_seq) + 1,
4387                             TH_ACK, (*state)->src.max_win, 0, 0, 0,
4388                             (*state)->tag, NULL, NULL);
4389                         pf_send_tcp((*state)->rule.ptr, pd->af,
4390                             &sk->addr[pd->sidx], &sk->addr[pd->didx],
4391                             sk->port[pd->sidx], sk->port[pd->didx],
4392                             (*state)->src.seqhi + 1, (*state)->src.seqlo + 1,
4393                             TH_ACK, (*state)->dst.max_win, 0, 0, 1,
4394                             0, NULL, NULL);
4395                         (*state)->src.seqdiff = (*state)->dst.seqhi -
4396                             (*state)->src.seqlo;
4397                         (*state)->dst.seqdiff = (*state)->src.seqhi -
4398                             (*state)->dst.seqlo;
4399                         (*state)->src.seqhi = (*state)->src.seqlo +
4400                             (*state)->dst.max_win;
4401                         (*state)->dst.seqhi = (*state)->dst.seqlo +
4402                             (*state)->src.max_win;
4403                         (*state)->src.wscale = (*state)->dst.wscale = 0;
4404                         (*state)->src.state = (*state)->dst.state =
4405                             TCPS_ESTABLISHED;
4406                         REASON_SET(reason, PFRES_SYNPROXY);
4407                         return (PF_SYNPROXY_DROP);
4408                 }
4409         }
4410
4411         if (((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN) &&
4412             dst->state >= TCPS_FIN_WAIT_2 &&
4413             src->state >= TCPS_FIN_WAIT_2) {
4414                 if (pf_status.debug >= PF_DEBUG_MISC) {
4415                         kprintf("pf: state reuse ");
4416                         pf_print_state(*state);
4417                         pf_print_flags(th->th_flags);
4418                         kprintf("\n");
4419                 }
4420                 /* XXX make sure it's the same direction ?? */
4421                 (*state)->src.state = (*state)->dst.state = TCPS_CLOSED;
4422                 pf_unlink_state(*state);
4423                 *state = NULL;
4424                 return (PF_DROP);
4425         }
4426
4427         if ((*state)->state_flags & PFSTATE_SLOPPY) {
4428                 if (pf_tcp_track_sloppy(src, dst, state, pd, reason) == PF_DROP)
4429                         return (PF_DROP);
4430         } else {
4431                 if (pf_tcp_track_full(src, dst, state, kif, m, off, pd, reason,
4432                     &copyback) == PF_DROP)
4433                         return (PF_DROP);
4434         }
4435
4436         /* translate source/destination address, if necessary */
4437         if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) {
4438                 struct pf_state_key *nk = (*state)->key[pd->didx];
4439
4440                 if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], pd->af) ||
4441                     nk->port[pd->sidx] != th->th_sport)
4442                         pf_change_ap(pd->src, &th->th_sport, pd->ip_sum,
4443                             &th->th_sum, &nk->addr[pd->sidx],
4444                             nk->port[pd->sidx], 0, pd->af);
4445
4446                 if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], pd->af) ||
4447                     nk->port[pd->didx] != th->th_dport) {
4448                         /*
4449                          * If we don't redispatch the packet will go into
4450                          * the protocol stack on the wrong cpu for the
4451                          * post-translated address.
4452                          */
4453                         m->m_flags &= ~M_HASH;
4454                         pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum,
4455                             &th->th_sum, &nk->addr[pd->didx],
4456                             nk->port[pd->didx], 0, pd->af);
4457                 }
4458                 copyback = 1;
4459         }
4460
4461         /* Copyback sequence modulation or stateful scrub changes if needed */
4462         if (copyback)
4463                 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4464
4465         return (PF_PASS);
4466 }
4467
4468 int
4469 pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif,
4470     struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4471 {
4472         struct pf_state_peer    *src, *dst;
4473         struct pf_state_key_cmp  key;
4474         struct udphdr           *uh = pd->hdr.udp;
4475
4476         key.af = pd->af;
4477         key.proto = IPPROTO_UDP;
4478         if (direction == PF_IN) {       /* wire side, straight */
4479                 PF_ACPY(&key.addr[0], pd->src, key.af);
4480                 PF_ACPY(&key.addr[1], pd->dst, key.af);
4481                 key.port[0] = uh->uh_sport;
4482                 key.port[1] = uh->uh_dport;
4483         } else {                        /* stack side, reverse */
4484                 PF_ACPY(&key.addr[1], pd->src, key.af);
4485                 PF_ACPY(&key.addr[0], pd->dst, key.af);
4486                 key.port[1] = uh->uh_sport;
4487                 key.port[0] = uh->uh_dport;
4488         }
4489
4490         STATE_LOOKUP(kif, &key, direction, *state, m);
4491
4492         if (direction == (*state)->direction) {
4493                 src = &(*state)->src;
4494                 dst = &(*state)->dst;
4495         } else {
4496                 src = &(*state)->dst;
4497                 dst = &(*state)->src;
4498         }
4499
4500         /* update states */
4501         if (src->state < PFUDPS_SINGLE)
4502                 src->state = PFUDPS_SINGLE;
4503         if (dst->state == PFUDPS_SINGLE)
4504                 dst->state = PFUDPS_MULTIPLE;
4505
4506         /* update expire time */
4507         (*state)->expire = time_second;
4508         if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE)
4509                 (*state)->timeout = PFTM_UDP_MULTIPLE;
4510         else
4511                 (*state)->timeout = PFTM_UDP_SINGLE;
4512
4513         /* translate source/destination address, if necessary */
4514         if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) {
4515                 struct pf_state_key *nk = (*state)->key[pd->didx];
4516
4517                 if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], pd->af) ||
4518                     nk->port[pd->sidx] != uh->uh_sport)
4519                         pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum,
4520                             &uh->uh_sum, &nk->addr[pd->sidx],
4521                             nk->port[pd->sidx], 1, pd->af);
4522
4523                 if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], pd->af) ||
4524                     nk->port[pd->didx] != uh->uh_dport) {
4525                         /*
4526                          * If we don't redispatch the packet will go into
4527                          * the protocol stack on the wrong cpu for the
4528                          * post-translated address.
4529                          */
4530                         m->m_flags &= ~M_HASH;
4531                         pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum,
4532                             &uh->uh_sum, &nk->addr[pd->didx],
4533                             nk->port[pd->didx], 1, pd->af);
4534                 }
4535                 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
4536         }
4537
4538         return (PF_PASS);
4539 }
4540
4541 int
4542 pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif,
4543     struct mbuf *m, int off, void *h, struct pf_pdesc *pd, u_short *reason)
4544 {
4545         struct pf_addr  *saddr = pd->src, *daddr = pd->dst;
4546         u_int16_t        icmpid = 0, *icmpsum;
4547         u_int8_t         icmptype;
4548         int              state_icmp = 0;
4549         struct pf_state_key_cmp key;
4550
4551         switch (pd->proto) {
4552 #ifdef INET
4553         case IPPROTO_ICMP:
4554                 icmptype = pd->hdr.icmp->icmp_type;
4555                 icmpid = pd->hdr.icmp->icmp_id;
4556                 icmpsum = &pd->hdr.icmp->icmp_cksum;
4557
4558                 if (icmptype == ICMP_UNREACH ||
4559                     icmptype == ICMP_SOURCEQUENCH ||
4560                     icmptype == ICMP_REDIRECT ||
4561                     icmptype == ICMP_TIMXCEED ||
4562                     icmptype == ICMP_PARAMPROB)
4563                         state_icmp++;
4564                 break;
4565 #endif /* INET */
4566 #ifdef INET6
4567         case IPPROTO_ICMPV6:
4568                 icmptype = pd->hdr.icmp6->icmp6_type;
4569                 icmpid = pd->hdr.icmp6->icmp6_id;
4570                 icmpsum = &pd->hdr.icmp6->icmp6_cksum;
4571
4572                 if (icmptype == ICMP6_DST_UNREACH ||
4573                     icmptype == ICMP6_PACKET_TOO_BIG ||
4574                     icmptype == ICMP6_TIME_EXCEEDED ||
4575                     icmptype == ICMP6_PARAM_PROB)
4576                         state_icmp++;
4577                 break;
4578 #endif /* INET6 */
4579         }
4580
4581         if (!state_icmp) {
4582
4583                 /*
4584                  * ICMP query/reply message not related to a TCP/UDP packet.
4585                  * Search for an ICMP state.
4586                  */
4587                 key.af = pd->af;
4588                 key.proto = pd->proto;
4589                 key.port[0] = key.port[1] = icmpid;
4590                 if (direction == PF_IN) {       /* wire side, straight */
4591                         PF_ACPY(&key.addr[0], pd->src, key.af);
4592                         PF_ACPY(&key.addr[1], pd->dst, key.af);
4593                 } else {                        /* stack side, reverse */
4594                         PF_ACPY(&key.addr[1], pd->src, key.af);
4595                         PF_ACPY(&key.addr[0], pd->dst, key.af);
4596                 }
4597
4598                 STATE_LOOKUP(kif, &key, direction, *state, m);
4599
4600                 (*state)->expire = time_second;
4601                 (*state)->timeout = PFTM_ICMP_ERROR_REPLY;
4602
4603                 /* translate source/destination address, if necessary */
4604                 if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) {
4605                         struct pf_state_key *nk = (*state)->key[pd->didx];
4606
4607                         switch (pd->af) {
4608 #ifdef INET
4609                         case AF_INET:
4610                                 if (PF_ANEQ(pd->src,
4611                                     &nk->addr[pd->sidx], AF_INET))
4612                                         pf_change_a(&saddr->v4.s_addr,
4613                                             pd->ip_sum,
4614                                             nk->addr[pd->sidx].v4.s_addr, 0);
4615
4616                                 if (PF_ANEQ(pd->dst, &nk->addr[pd->didx],
4617                                     AF_INET))
4618                                         pf_change_a(&daddr->v4.s_addr,
4619                                             pd->ip_sum,
4620                                             nk->addr[pd->didx].v4.s_addr, 0);
4621
4622                                 if (nk->port[0] !=
4623                                     pd->hdr.icmp->icmp_id) {
4624                                         pd->hdr.icmp->icmp_cksum =
4625                                             pf_cksum_fixup(
4626                                             pd->hdr.icmp->icmp_cksum, icmpid,
4627                                             nk->port[pd->sidx], 0);
4628                                         pd->hdr.icmp->icmp_id =
4629                                             nk->port[pd->sidx];
4630                                 }
4631
4632                                 m_copyback(m, off, ICMP_MINLEN,
4633                                     (caddr_t)pd->hdr.icmp);
4634                                 break;
4635 #endif /* INET */
4636 #ifdef INET6
4637                         case AF_INET6:
4638                                 if (PF_ANEQ(pd->src,
4639                                     &nk->addr[pd->sidx], AF_INET6))
4640                                         pf_change_a6(saddr,
4641                                             &pd->hdr.icmp6->icmp6_cksum,
4642                                             &nk->addr[pd->sidx], 0);
4643
4644                                 if (PF_ANEQ(pd->dst,
4645                                     &nk->addr[pd->didx], AF_INET6))
4646                                         pf_change_a6(daddr,
4647                                             &pd->hdr.icmp6->icmp6_cksum,
4648                                             &nk->addr[pd->didx], 0);
4649
4650                                 m_copyback(m, off,
4651                                         sizeof(struct icmp6_hdr),
4652                                         (caddr_t)pd->hdr.icmp6);
4653                                 break;
4654 #endif /* INET6 */
4655                         }
4656                 }
4657                 return (PF_PASS);
4658
4659         } else {
4660                 /*
4661                  * ICMP error message in response to a TCP/UDP packet.
4662                  * Extract the inner TCP/UDP header and search for that state.
4663                  */
4664
4665                 struct pf_pdesc pd2;
4666 #ifdef INET
4667                 struct ip       h2;
4668 #endif /* INET */
4669 #ifdef INET6
4670                 struct ip6_hdr  h2_6;
4671                 int             terminal = 0;
4672 #endif /* INET6 */
4673                 int             ipoff2;
4674                 int             off2;
4675
4676                 pd2.af = pd->af;
4677                 /* Payload packet is from the opposite direction. */
4678                 pd2.sidx = (direction == PF_IN) ? 1 : 0;
4679                 pd2.didx = (direction == PF_IN) ? 0 : 1;
4680                 switch (pd->af) {
4681 #ifdef INET
4682                 case AF_INET:
4683                         /* offset of h2 in mbuf chain */
4684                         ipoff2 = off + ICMP_MINLEN;
4685
4686                         if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2),
4687                             NULL, reason, pd2.af)) {
4688                                 DPFPRINTF(PF_DEBUG_MISC,
4689                                     ("pf: ICMP error message too short "
4690                                     "(ip)\n"));
4691                                 return (PF_DROP);
4692                         }
4693                         /*
4694                          * ICMP error messages don't refer to non-first
4695                          * fragments
4696                          */
4697                         if (h2.ip_off & htons(IP_OFFMASK)) {
4698                                 REASON_SET(reason, PFRES_FRAG);
4699                                 return (PF_DROP);
4700                         }
4701
4702                         /* offset of protocol header that follows h2 */
4703                         off2 = ipoff2 + (h2.ip_hl << 2);
4704
4705                         pd2.proto = h2.ip_p;
4706                         pd2.src = (struct pf_addr *)&h2.ip_src;
4707                         pd2.dst = (struct pf_addr *)&h2.ip_dst;
4708                         pd2.ip_sum = &h2.ip_sum;
4709                         break;
4710 #endif /* INET */
4711 #ifdef INET6
4712                 case AF_INET6:
4713                         ipoff2 = off + sizeof(struct icmp6_hdr);
4714
4715                         if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6),
4716                             NULL, reason, pd2.af)) {
4717                                 DPFPRINTF(PF_DEBUG_MISC,
4718                                     ("pf: ICMP error message too short "
4719                                     "(ip6)\n"));
4720                                 return (PF_DROP);
4721                         }
4722                         pd2.proto = h2_6.ip6_nxt;
4723                         pd2.src = (struct pf_addr *)&h2_6.ip6_src;
4724                         pd2.dst = (struct pf_addr *)&h2_6.ip6_dst;
4725                         pd2.ip_sum = NULL;
4726                         off2 = ipoff2 + sizeof(h2_6);
4727                         do {
4728                                 switch (pd2.proto) {
4729                                 case IPPROTO_FRAGMENT:
4730                                         /*
4731                                          * ICMPv6 error messages for
4732                                          * non-first fragments
4733                                          */
4734                                         REASON_SET(reason, PFRES_FRAG);
4735                                         return (PF_DROP);
4736                                 case IPPROTO_AH:
4737                                 case IPPROTO_HOPOPTS:
4738                                 case IPPROTO_ROUTING:
4739                                 case IPPROTO_DSTOPTS: {
4740                                         /* get next header and header length */
4741                                         struct ip6_ext opt6;
4742
4743                                         if (!pf_pull_hdr(m, off2, &opt6,
4744                                             sizeof(opt6), NULL, reason,
4745                                             pd2.af)) {
4746                                                 DPFPRINTF(PF_DEBUG_MISC,
4747                                                     ("pf: ICMPv6 short opt\n"));
4748                                                 return (PF_DROP);
4749                                         }
4750                                         if (pd2.proto == IPPROTO_AH)
4751                                                 off2 += (opt6.ip6e_len + 2) * 4;
4752                                         else
4753                                                 off2 += (opt6.ip6e_len + 1) * 8;
4754                                         pd2.proto = opt6.ip6e_nxt;
4755                                         /* goto the next header */
4756                                         break;
4757                                 }
4758                                 default:
4759                                         terminal++;
4760                                         break;
4761                                 }
4762                         } while (!terminal);
4763                         break;
4764 #endif /* INET6 */
4765                 default:
4766                         DPFPRINTF(PF_DEBUG_MISC,
4767                             ("pf: ICMP AF %d unknown (ip6)\n", pd->af));
4768                         return (PF_DROP);
4769                         break;
4770                 }
4771
4772                 switch (pd2.proto) {
4773                 case IPPROTO_TCP: {
4774                         struct tcphdr            th;
4775                         u_int32_t                seq;
4776                         struct pf_state_peer    *src, *dst;
4777                         u_int8_t                 dws;
4778                         int                      copyback = 0;
4779
4780                         /*
4781                          * Only the first 8 bytes of the TCP header can be
4782                          * expected. Don't access any TCP header fields after
4783                          * th_seq, an ackskew test is not possible.
4784                          */
4785                         if (!pf_pull_hdr(m, off2, &th, 8, NULL, reason,
4786                             pd2.af)) {
4787                                 DPFPRINTF(PF_DEBUG_MISC,
4788                                     ("pf: ICMP error message too short "
4789                                     "(tcp)\n"));
4790                                 return (PF_DROP);
4791                         }
4792
4793                         key.af = pd2.af;
4794                         key.proto = IPPROTO_TCP;
4795                         PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
4796                         PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
4797                         key.port[pd2.sidx] = th.th_sport;
4798                         key.port[pd2.didx] = th.th_dport;
4799
4800                         STATE_LOOKUP(kif, &key, direction, *state, m);
4801
4802                         if (direction == (*state)->direction) {
4803                                 src = &(*state)->dst;
4804                                 dst = &(*state)->src;
4805                         } else {
4806                                 src = &(*state)->src;
4807                                 dst = &(*state)->dst;
4808                         }
4809
4810                         if (src->wscale && dst->wscale)
4811                                 dws = dst->wscale & PF_WSCALE_MASK;
4812                         else
4813                                 dws = 0;
4814
4815                         /* Demodulate sequence number */
4816                         seq = ntohl(th.th_seq) - src->seqdiff;
4817                         if (src->seqdiff) {
4818                                 pf_change_a(&th.th_seq, icmpsum,
4819                                     htonl(seq), 0);
4820                                 copyback = 1;
4821                         }
4822
4823                         if (!((*state)->state_flags & PFSTATE_SLOPPY) &&
4824                             (!SEQ_GEQ(src->seqhi, seq) ||
4825                             !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)))) {
4826                                 if (pf_status.debug >= PF_DEBUG_MISC) {
4827                                         kprintf("pf: BAD ICMP %d:%d ",
4828                                             icmptype, pd->hdr.icmp->icmp_code);
4829                                         pf_print_host(pd->src, 0, pd->af);
4830                                         kprintf(" -> ");
4831                                         pf_print_host(pd->dst, 0, pd->af);
4832                                         kprintf(" state: ");
4833                                         pf_print_state(*state);
4834                                         kprintf(" seq=%u\n", seq);
4835                                 }
4836                                 REASON_SET(reason, PFRES_BADSTATE);
4837                                 return (PF_DROP);
4838                         } else {
4839                                 if (pf_status.debug >= PF_DEBUG_MISC) {
4840                                         kprintf("pf: OK ICMP %d:%d ",
4841                                             icmptype, pd->hdr.icmp->icmp_code);
4842                                         pf_print_host(pd->src, 0, pd->af);
4843                                         kprintf(" -> ");
4844                                         pf_print_host(pd->dst, 0, pd->af);
4845                                         kprintf(" state: ");
4846                                         pf_print_state(*state);
4847                                         kprintf(" seq=%u\n", seq);
4848                                 }
4849                         }
4850
4851                         /* translate source/destination address, if necessary */
4852                         if ((*state)->key[PF_SK_WIRE] !=
4853                             (*state)->key[PF_SK_STACK]) {
4854                                 struct pf_state_key *nk =
4855                                     (*state)->key[pd->didx];
4856
4857                                 if (PF_ANEQ(pd2.src,
4858                                     &nk->addr[pd2.sidx], pd2.af) ||
4859                                     nk->port[pd2.sidx] != th.th_sport)
4860                                         pf_change_icmp(pd2.src, &th.th_sport,
4861                                             daddr, &nk->addr[pd2.sidx],
4862                                             nk->port[pd2.sidx], NULL,
4863                                             pd2.ip_sum, icmpsum,
4864                                             pd->ip_sum, 0, pd2.af);
4865
4866                                 if (PF_ANEQ(pd2.dst,
4867                                     &nk->addr[pd2.didx], pd2.af) ||
4868                                     nk->port[pd2.didx] != th.th_dport)
4869                                         pf_change_icmp(pd2.dst, &th.th_dport,
4870                                             NULL, /* XXX Inbound NAT? */
4871                                             &nk->addr[pd2.didx],
4872                                             nk->port[pd2.didx], NULL,
4873                                             pd2.ip_sum, icmpsum,
4874                                             pd->ip_sum, 0, pd2.af);
4875                                 copyback = 1;
4876                         }
4877
4878                         if (copyback) {
4879                                 switch (pd2.af) {
4880 #ifdef INET
4881                                 case AF_INET:
4882                                         m_copyback(m, off, ICMP_MINLEN,
4883                                             (caddr_t)pd->hdr.icmp);
4884                                         m_copyback(m, ipoff2, sizeof(h2),
4885                                             (caddr_t)&h2);
4886                                         break;
4887 #endif /* INET */
4888 #ifdef INET6
4889                                 case AF_INET6:
4890                                         m_copyback(m, off,
4891                                             sizeof(struct icmp6_hdr),
4892                                             (caddr_t)pd->hdr.icmp6);
4893                                         m_copyback(m, ipoff2, sizeof(h2_6),
4894                                             (caddr_t)&h2_6);
4895                                         break;
4896 #endif /* INET6 */
4897                                 }
4898                                 m_copyback(m, off2, 8, (caddr_t)&th);
4899                         }
4900
4901                         return (PF_PASS);
4902                         break;
4903                 }
4904                 case IPPROTO_UDP: {
4905                         struct udphdr           uh;
4906
4907                         if (!pf_pull_hdr(m, off2, &uh, sizeof(uh),
4908                             NULL, reason, pd2.af)) {
4909                                 DPFPRINTF(PF_DEBUG_MISC,
4910                                     ("pf: ICMP error message too short "
4911                                     "(udp)\n"));
4912                                 return (PF_DROP);
4913                         }
4914
4915                         key.af = pd2.af;
4916                         key.proto = IPPROTO_UDP;
4917                         PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
4918                         PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
4919                         key.port[pd2.sidx] = uh.uh_sport;
4920                         key.port[pd2.didx] = uh.uh_dport;
4921
4922                         STATE_LOOKUP(kif, &key, direction, *state, m);
4923
4924                         /* translate source/destination address, if necessary */
4925                         if ((*state)->key[PF_SK_WIRE] !=
4926                             (*state)->key[PF_SK_STACK]) {
4927                                 struct pf_state_key *nk =
4928                                     (*state)->key[pd->didx];
4929
4930                                 if (PF_ANEQ(pd2.src,
4931                                     &nk->addr[pd2.sidx], pd2.af) ||
4932                                     nk->port[pd2.sidx] != uh.uh_sport)
4933                                         pf_change_icmp(pd2.src, &uh.uh_sport,
4934                                             daddr, &nk->addr[pd2.sidx],
4935                                             nk->port[pd2.sidx], &uh.uh_sum,
4936                                             pd2.ip_sum, icmpsum,
4937                                             pd->ip_sum, 1, pd2.af);
4938
4939                                 if (PF_ANEQ(pd2.dst,
4940                                     &nk->addr[pd2.didx], pd2.af) ||
4941                                     nk->port[pd2.didx] != uh.uh_dport)
4942                                         pf_change_icmp(pd2.dst, &uh.uh_dport,
4943                                             NULL, /* XXX Inbound NAT? */
4944                                             &nk->addr[pd2.didx],
4945                                             nk->port[pd2.didx], &uh.uh_sum,
4946                                             pd2.ip_sum, icmpsum,
4947                                             pd->ip_sum, 1, pd2.af);
4948
4949                                 switch (pd2.af) {
4950 #ifdef INET
4951                                 case AF_INET:
4952                                         m_copyback(m, off, ICMP_MINLEN,
4953                                             (caddr_t)pd->hdr.icmp);
4954                                         m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4955                                         break;
4956 #endif /* INET */
4957 #ifdef INET6
4958                                 case AF_INET6:
4959                                         m_copyback(m, off,
4960                                             sizeof(struct icmp6_hdr),
4961                                             (caddr_t)pd->hdr.icmp6);
4962                                         m_copyback(m, ipoff2, sizeof(h2_6),
4963                                             (caddr_t)&h2_6);
4964                                         break;
4965 #endif /* INET6 */
4966                                 }
4967                                 m_copyback(m, off2, sizeof(uh), (caddr_t)&uh);
4968                         }
4969
4970                         return (PF_PASS);
4971                         break;
4972                 }
4973 #ifdef INET
4974                 case IPPROTO_ICMP: {
4975                         struct icmp             iih;
4976
4977                         if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN,
4978                             NULL, reason, pd2.af)) {
4979                                 DPFPRINTF(PF_DEBUG_MISC,
4980                                     ("pf: ICMP error message too short i"
4981                                     "(icmp)\n"));
4982                                 return (PF_DROP);
4983                         }
4984
4985                         key.af = pd2.af;
4986                         key.proto = IPPROTO_ICMP;
4987                         PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
4988                         PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
4989                         key.port[0] = key.port[1] = iih.icmp_id;
4990
4991                         STATE_LOOKUP(kif, &key, direction, *state, m);
4992
4993                         /* translate source/destination address, if necessary */
4994                         if ((*state)->key[PF_SK_WIRE] !=
4995                             (*state)->key[PF_SK_STACK]) {
4996                                 struct pf_state_key *nk =
4997                                     (*state)->key[pd->didx];
4998
4999                                 if (PF_ANEQ(pd2.src,
5000                                     &nk->addr[pd2.sidx], pd2.af) ||
5001                                     nk->port[pd2.sidx] != iih.icmp_id)
5002                                         pf_change_icmp(pd2.src, &iih.icmp_id,
5003                                             daddr, &nk->addr[pd2.sidx],
5004                                             nk->port[pd2.sidx], NULL,
5005                                             pd2.ip_sum, icmpsum,
5006                                             pd->ip_sum, 0, AF_INET);
5007
5008                                 if (PF_ANEQ(pd2.dst,
5009                                     &nk->addr[pd2.didx], pd2.af) ||
5010                                     nk->port[pd2.didx] != iih.icmp_id)
5011                                         pf_change_icmp(pd2.dst, &iih.icmp_id,
5012                                             NULL, /* XXX Inbound NAT? */
5013                                             &nk->addr[pd2.didx],
5014                                             nk->port[pd2.didx], NULL,
5015                                             pd2.ip_sum, icmpsum,
5016                                             pd->ip_sum, 0, AF_INET);
5017
5018                                 m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
5019                                 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
5020                                 m_copyback(m, off2, ICMP_MINLEN, (caddr_t)&iih);
5021                         }
5022                         return (PF_PASS);
5023                         break;
5024                 }
5025 #endif /* INET */
5026 #ifdef INET6
5027                 case IPPROTO_ICMPV6: {
5028                         struct icmp6_hdr        iih;
5029
5030                         if (!pf_pull_hdr(m, off2, &iih,
5031                             sizeof(struct icmp6_hdr), NULL, reason, pd2.af)) {
5032                                 DPFPRINTF(PF_DEBUG_MISC,
5033                                     ("pf: ICMP error message too short "
5034                                     "(icmp6)\n"));
5035                                 return (PF_DROP);
5036                         }
5037
5038                         key.af = pd2.af;
5039                         key.proto = IPPROTO_ICMPV6;
5040                         PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
5041                         PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
5042                         key.port[0] = key.port[1] = iih.icmp6_id;
5043
5044                         STATE_LOOKUP(kif, &key, direction, *state, m);
5045
5046                         /* translate source/destination address, if necessary */
5047                         if ((*state)->key[PF_SK_WIRE] !=
5048                             (*state)->key[PF_SK_STACK]) {
5049                                 struct pf_state_key *nk =
5050                                     (*state)->key[pd->didx];
5051
5052                                 if (PF_ANEQ(pd2.src,
5053                                     &nk->addr[pd2.sidx], pd2.af) ||
5054                                     nk->port[pd2.sidx] != iih.icmp6_id)
5055                                         pf_change_icmp(pd2.src, &iih.icmp6_id,
5056                                             daddr, &nk->addr[pd2.sidx],
5057                                             nk->port[pd2.sidx], NULL,
5058                                             pd2.ip_sum, icmpsum,
5059                                             pd->ip_sum, 0, AF_INET6);
5060
5061                                 if (PF_ANEQ(pd2.dst,
5062                                     &nk->addr[pd2.didx], pd2.af) ||
5063                                     nk->port[pd2.didx] != iih.icmp6_id)
5064                                         pf_change_icmp(pd2.dst, &iih.icmp6_id,
5065                                             NULL, /* XXX Inbound NAT? */
5066                                             &nk->addr[pd2.didx],
5067                                             nk->port[pd2.didx], NULL,
5068                                             pd2.ip_sum, icmpsum,
5069                                             pd->ip_sum, 0, AF_INET6);
5070
5071                                 m_copyback(m, off, sizeof(struct icmp6_hdr),
5072                                     (caddr_t)pd->hdr.icmp6);
5073                                 m_copyback(m, ipoff2, sizeof(h2_6), (caddr_t)&h2_6);
5074                                 m_copyback(m, off2, sizeof(struct icmp6_hdr),
5075                                     (caddr_t)&iih);
5076                         }
5077
5078                         return (PF_PASS);
5079                         break;
5080                 }
5081 #endif /* INET6 */
5082                 default: {
5083                         key.af = pd2.af;
5084                         key.proto = pd2.proto;
5085                         PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
5086                         PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
5087                         key.port[0] = key.port[1] = 0;
5088
5089                         STATE_LOOKUP(kif, &key, direction, *state, m);
5090
5091                         /* translate source/destination address, if necessary */
5092                         if ((*state)->key[PF_SK_WIRE] !=
5093                             (*state)->key[PF_SK_STACK]) {
5094                                 struct pf_state_key *nk =
5095                                     (*state)->key[pd->didx];
5096
5097                                 if (PF_ANEQ(pd2.src,
5098                                     &nk->addr[pd2.sidx], pd2.af))
5099                                         pf_change_icmp(pd2.src, NULL, daddr,
5100                                             &nk->addr[pd2.sidx], 0, NULL,
5101                                             pd2.ip_sum, icmpsum,
5102                                             pd->ip_sum, 0, pd2.af);
5103
5104                                 if (PF_ANEQ(pd2.dst,
5105                                     &nk->addr[pd2.didx], pd2.af))
5106                                         pf_change_icmp(pd2.src, NULL,
5107                                             NULL, /* XXX Inbound NAT? */
5108                                             &nk->addr[pd2.didx], 0, NULL,
5109                                             pd2.ip_sum, icmpsum,
5110                                             pd->ip_sum, 0, pd2.af);
5111
5112                                 switch (pd2.af) {
5113 #ifdef INET
5114                                 case AF_INET:
5115                                         m_copyback(m, off, ICMP_MINLEN,
5116                                             (caddr_t)pd->hdr.icmp);
5117                                         m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
5118                                         break;
5119 #endif /* INET */
5120 #ifdef INET6
5121                                 case AF_INET6:
5122                                         m_copyback(m, off,
5123                                             sizeof(struct icmp6_hdr),
5124                                             (caddr_t)pd->hdr.icmp6);
5125                                         m_copyback(m, ipoff2, sizeof(h2_6),
5126                                             (caddr_t)&h2_6);
5127                                         break;
5128 #endif /* INET6 */
5129                                 }
5130                         }
5131                         return (PF_PASS);
5132                         break;
5133                 }
5134                 }
5135         }
5136 }
5137
5138 int
5139 pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif,
5140     struct mbuf *m, struct pf_pdesc *pd)
5141 {
5142         struct pf_state_peer    *src, *dst;
5143         struct pf_state_key_cmp  key;
5144
5145         key.af = pd->af;
5146         key.proto = pd->proto;
5147         if (direction == PF_IN) {
5148                 PF_ACPY(&key.addr[0], pd->src, key.af);
5149                 PF_ACPY(&key.addr[1], pd->dst, key.af);
5150                 key.port[0] = key.port[1] = 0;
5151         } else {
5152                 PF_ACPY(&key.addr[1], pd->src, key.af);
5153                 PF_ACPY(&key.addr[0], pd->dst, key.af);
5154                 key.port[1] = key.port[0] = 0;
5155         }
5156
5157         STATE_LOOKUP(kif, &key, direction, *state, m);
5158
5159         if (direction == (*state)->direction) {
5160                 src = &(*state)->src;
5161                 dst = &(*state)->dst;
5162         } else {
5163                 src = &(*state)->dst;
5164                 dst = &(*state)->src;
5165         }
5166
5167         /* update states */
5168         if (src->state < PFOTHERS_SINGLE)
5169                 src->state = PFOTHERS_SINGLE;
5170         if (dst->state == PFOTHERS_SINGLE)
5171                 dst->state = PFOTHERS_MULTIPLE;
5172
5173         /* update expire time */
5174         (*state)->expire = time_second;
5175         if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE)
5176                 (*state)->timeout = PFTM_OTHER_MULTIPLE;
5177         else
5178                 (*state)->timeout = PFTM_OTHER_SINGLE;
5179
5180         /* translate source/destination address, if necessary */
5181         if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) {
5182                 struct pf_state_key *nk = (*state)->key[pd->didx];
5183
5184                 KKASSERT(nk);
5185                 KKASSERT(pd);
5186                 KKASSERT(pd->src);
5187                 KKASSERT(pd->dst);
5188                 switch (pd->af) {
5189 #ifdef INET
5190                 case AF_INET:
5191                         if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], AF_INET))
5192                                 pf_change_a(&pd->src->v4.s_addr,
5193                                     pd->ip_sum,
5194                                     nk->addr[pd->sidx].v4.s_addr,
5195                                     0);
5196
5197
5198                         if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], AF_INET))
5199                                 pf_change_a(&pd->dst->v4.s_addr,
5200                                     pd->ip_sum,
5201                                     nk->addr[pd->didx].v4.s_addr,
5202                                     0);
5203
5204                                 break;
5205 #endif /* INET */
5206 #ifdef INET6
5207                 case AF_INET6:
5208                         if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], AF_INET))
5209                                 PF_ACPY(pd->src, &nk->addr[pd->sidx], pd->af);
5210
5211                         if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], AF_INET))
5212                                 PF_ACPY(pd->dst, &nk->addr[pd->didx], pd->af);
5213 #endif /* INET6 */
5214                 }
5215         }
5216         return (PF_PASS);
5217 }
5218
5219 /*
5220  * ipoff and off are measured from the start of the mbuf chain.
5221  * h must be at "ipoff" on the mbuf chain.
5222  */
5223 void *
5224 pf_pull_hdr(struct mbuf *m, int off, void *p, int len,
5225     u_short *actionp, u_short *reasonp, sa_family_t af)
5226 {
5227         switch (af) {
5228 #ifdef INET
5229         case AF_INET: {
5230                 struct ip       *h = mtod(m, struct ip *);
5231                 u_int16_t        fragoff = (h->ip_off & IP_OFFMASK) << 3;
5232
5233                 if (fragoff) {
5234                         if (fragoff >= len)
5235                                 ACTION_SET(actionp, PF_PASS);
5236                         else {
5237                                 ACTION_SET(actionp, PF_DROP);
5238                                 REASON_SET(reasonp, PFRES_FRAG);
5239                         }
5240                         return (NULL);
5241                 }
5242                 if (m->m_pkthdr.len < off + len ||
5243                     h->ip_len < off + len) {
5244                         ACTION_SET(actionp, PF_DROP);
5245                         REASON_SET(reasonp, PFRES_SHORT);
5246                         return (NULL);
5247                 }
5248                 break;
5249         }
5250 #endif /* INET */
5251 #ifdef INET6
5252         case AF_INET6: {
5253                 struct ip6_hdr  *h = mtod(m, struct ip6_hdr *);
5254
5255                 if (m->m_pkthdr.len < off + len ||
5256                     (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) <
5257                     (unsigned)(off + len)) {
5258                         ACTION_SET(actionp, PF_DROP);
5259                         REASON_SET(reasonp, PFRES_SHORT);
5260                         return (NULL);
5261                 }
5262                 break;
5263         }
5264 #endif /* INET6 */
5265         }
5266         m_copydata(m, off, len, p);
5267         return (p);
5268 }
5269
5270 int
5271 pf_routable(struct pf_addr *addr, sa_family_t af, struct pfi_kif *kif)
5272 {
5273         struct sockaddr_in      *dst;
5274         int                      ret = 1;
5275         int                      check_mpath;
5276 #ifdef INET6
5277         struct sockaddr_in6     *dst6;
5278         struct route_in6         ro;
5279 #else
5280         struct route             ro;
5281 #endif
5282         struct radix_node       *rn;
5283         struct rtentry          *rt;
5284         struct ifnet            *ifp;
5285
5286         check_mpath = 0;
5287         bzero(&ro, sizeof(ro));
5288         switch (af) {
5289         case AF_INET:
5290                 dst = satosin(&ro.ro_dst);
5291                 dst->sin_family = AF_INET;
5292                 dst->sin_len = sizeof(*dst);
5293                 dst->sin_addr = addr->v4;
5294                 break;
5295 #ifdef INET6
5296         case AF_INET6:
5297                 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
5298                 dst6->sin6_family = AF_INET6;
5299                 dst6->sin6_len = sizeof(*dst6);
5300                 dst6->sin6_addr = addr->v6;
5301                 break;
5302 #endif /* INET6 */
5303         default:
5304                 return (0);
5305         }
5306
5307         /* Skip checks for ipsec interfaces */
5308         if (kif != NULL && kif->pfik_ifp->if_type == IFT_ENC)
5309                 goto out;
5310
5311         rtalloc_ign((struct route *)&ro, 0);
5312
5313         if (ro.ro_rt != NULL) {
5314                 /* No interface given, this is a no-route check */
5315                 if (kif == NULL)
5316                         goto out;
5317
5318                 if (kif->pfik_ifp == NULL) {
5319                         ret = 0;
5320                         goto out;
5321                 }
5322
5323                 /* Perform uRPF check if passed input interface */
5324                 ret = 0;
5325                 rn = (struct radix_node *)ro.ro_rt;
5326                 do {
5327                         rt = (struct rtentry *)rn;
5328                         ifp = rt->rt_ifp;
5329
5330                         if (kif->pfik_ifp == ifp)
5331                                 ret = 1;
5332                         rn = NULL;
5333                 } while (check_mpath == 1 && rn != NULL && ret == 0);
5334         } else
5335                 ret = 0;
5336 out:
5337         if (ro.ro_rt != NULL)
5338                 RTFREE(ro.ro_rt);
5339         return (ret);
5340 }
5341
5342 int
5343 pf_rtlabel_match(struct pf_addr *addr, sa_family_t af, struct pf_addr_wrap *aw)
5344 {
5345         struct sockaddr_in      *dst;
5346 #ifdef INET6
5347         struct sockaddr_in6     *dst6;
5348         struct route_in6         ro;
5349 #else
5350         struct route             ro;
5351 #endif
5352         int                      ret = 0;
5353
5354         ASSERT_LWKT_TOKEN_HELD(&pf_token);
5355
5356         bzero(&ro, sizeof(ro));
5357         switch (af) {
5358         case AF_INET:
5359                 dst = satosin(&ro.ro_dst);
5360                 dst->sin_family = AF_INET;
5361                 dst->sin_len = sizeof(*dst);
5362                 dst->sin_addr = addr->v4;
5363                 break;
5364 #ifdef INET6
5365         case AF_INET6:
5366                 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
5367                 dst6->sin6_family = AF_INET6;
5368                 dst6->sin6_len = sizeof(*dst6);
5369                 dst6->sin6_addr = addr->v6;
5370                 break;
5371 #endif /* INET6 */
5372         default:
5373                 return (0);
5374         }
5375
5376 rtalloc_ign((struct route *)&ro, (RTF_CLONING | RTF_PRCLONING));
5377
5378         if (ro.ro_rt != NULL) {
5379                 RTFREE(ro.ro_rt);
5380         }
5381
5382         return (ret);
5383 }
5384
5385 #ifdef INET
5386 void
5387 pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5388     struct pf_state *s, struct pf_pdesc *pd)
5389 {
5390         struct mbuf             *m0, *m1;
5391         struct route             iproute;
5392         struct route            *ro = NULL;
5393         struct sockaddr_in      *dst;
5394         struct ip               *ip;
5395         struct ifnet            *ifp = NULL;
5396         struct pf_addr           naddr;
5397         struct pf_src_node      *sn = NULL;
5398         int                      error = 0;
5399         int sw_csum;
5400 #ifdef IPSEC
5401         struct m_tag            *mtag;
5402 #endif /* IPSEC */
5403
5404         ASSERT_LWKT_TOKEN_HELD(&pf_token);
5405
5406         if (m == NULL || *m == NULL || r == NULL ||
5407             (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5408                 panic("pf_route: invalid parameters");
5409
5410         if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
5411                 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
5412                 (*m)->m_pkthdr.pf.routed = 1;
5413         } else {
5414                 if ((*m)->m_pkthdr.pf.routed++ > 3) {
5415                         m0 = *m;
5416                         *m = NULL;
5417                         goto bad;
5418                 }
5419         }
5420
5421         if (r->rt == PF_DUPTO) {
5422                 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL) {
5423                         return;
5424                 }
5425         } else {
5426                 if ((r->rt == PF_REPLYTO) == (r->direction == dir)) {
5427                         return;
5428                 }
5429                 m0 = *m;
5430         }
5431
5432         if (m0->m_len < sizeof(struct ip)) {
5433                 DPFPRINTF(PF_DEBUG_URGENT,
5434                     ("pf_route: m0->m_len < sizeof(struct ip)\n"));
5435                 goto bad;
5436         }
5437
5438         ip = mtod(m0, struct ip *);
5439
5440         ro = &iproute;
5441         bzero((caddr_t)ro, sizeof(*ro));
5442         dst = satosin(&ro->ro_dst);
5443         dst->sin_family = AF_INET;
5444         dst->sin_len = sizeof(*dst);
5445         dst->sin_addr = ip->ip_dst;
5446
5447         if (r->rt == PF_FASTROUTE) {
5448                 rtalloc(ro);
5449                 if (ro->ro_rt == 0) {
5450                         ipstat.ips_noroute++;
5451                         goto bad;
5452                 }
5453
5454                 ifp = ro->ro_rt->rt_ifp;
5455                 ro->ro_rt->rt_use++;
5456
5457                 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
5458                         dst = satosin(ro->ro_rt->rt_gateway);
5459         } else {
5460                 if (TAILQ_EMPTY(&r->rpool.list)) {
5461                         DPFPRINTF(PF_DEBUG_URGENT,
5462                             ("pf_route: TAILQ_EMPTY(&r->rpool.list)\n"));
5463                         goto bad;
5464                 }
5465                 if (s == NULL) {
5466                         pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src,
5467                             &naddr, NULL, &sn);
5468                         if (!PF_AZERO(&naddr, AF_INET))
5469                                 dst->sin_addr.s_addr = naddr.v4.s_addr;
5470                         ifp = r->rpool.cur->kif ?
5471                             r->rpool.cur->kif->pfik_ifp : NULL;
5472                 } else {
5473                         if (!PF_AZERO(&s->rt_addr, AF_INET))
5474                                 dst->sin_addr.s_addr =
5475                                     s->rt_addr.v4.s_addr;
5476                         ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5477                 }
5478         }
5479         if (ifp == NULL)
5480                 goto bad;
5481
5482         if (oifp != ifp) {
5483                 if (pf_test(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) {
5484                         goto bad;
5485                 } else if (m0 == NULL) {
5486                         goto done;
5487                 }
5488                 if (m0->m_len < sizeof(struct ip)) {
5489                         DPFPRINTF(PF_DEBUG_URGENT,
5490                             ("pf_route: m0->m_len < sizeof(struct ip)\n"));
5491                         goto bad;
5492                 }
5493                 ip = mtod(m0, struct ip *);
5494         }
5495
5496         /* Copied from FreeBSD 5.1-CURRENT ip_output. */
5497         m0->m_pkthdr.csum_flags |= CSUM_IP;
5498         sw_csum = m0->m_pkthdr.csum_flags & ~ifp->if_hwassist;
5499         if (sw_csum & CSUM_DELAY_DATA) {
5500                 in_delayed_cksum(m0);
5501                 sw_csum &= ~CSUM_DELAY_DATA;
5502         }
5503         m0->m_pkthdr.csum_flags &= ifp->if_hwassist;
5504
5505         if (ip->ip_len <= ifp->if_mtu ||
5506             (ifp->if_hwassist & CSUM_FRAGMENT &&
5507                 (ip->ip_off & IP_DF) == 0)) {
5508                 ip->ip_len = htons(ip->ip_len);
5509                 ip->ip_off = htons(ip->ip_off);
5510                 ip->ip_sum = 0;
5511                 if (sw_csum & CSUM_DELAY_IP) {
5512                         /* From KAME */
5513                         if (ip->ip_v == IPVERSION &&
5514                             (ip->ip_hl << 2) == sizeof(*ip)) {
5515                                 ip->ip_sum = in_cksum_hdr(ip);
5516                         } else {
5517                                 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
5518                         }
5519                 }
5520                 lwkt_reltoken(&pf_token);
5521                 error = ifp->if_output(ifp, m0, sintosa(dst), ro->ro_rt);
5522                 lwkt_gettoken(&pf_token);
5523                 goto done;
5524         }
5525
5526         /*
5527          * Too large for interface; fragment if possible.
5528          * Must be able to put at least 8 bytes per fragment.
5529          */
5530         if (ip->ip_off & IP_DF) {
5531                 ipstat.ips_cantfrag++;
5532                 if (r->rt != PF_DUPTO) {
5533                         icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0,
5534                             ifp->if_mtu);
5535                         goto done;
5536                 } else
5537                         goto bad;
5538         }
5539
5540         m1 = m0;
5541         error = ip_fragment(ip, &m0, ifp->if_mtu, ifp->if_hwassist, sw_csum);
5542         if (error) {
5543                 goto bad;
5544         }
5545
5546         for (m0 = m1; m0; m0 = m1) {
5547                 m1 = m0->m_nextpkt;
5548                 m0->m_nextpkt = 0;
5549                 if (error == 0) {
5550                         lwkt_reltoken(&pf_token);
5551                         error = (*ifp->if_output)(ifp, m0, sintosa(dst),
5552                                                   NULL);
5553                         lwkt_gettoken(&pf_token);
5554                 } else
5555                         m_freem(m0);
5556         }
5557
5558         if (error == 0)
5559                 ipstat.ips_fragmented++;
5560
5561 done:
5562         if (r->rt != PF_DUPTO)
5563                 *m = NULL;
5564         if (ro == &iproute && ro->ro_rt)
5565                 RTFREE(ro->ro_rt);
5566         return;
5567
5568 bad:
5569         m_freem(m0);
5570         goto done;
5571 }
5572 #endif /* INET */
5573
5574 #ifdef INET6
5575 void
5576 pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5577     struct pf_state *s, struct pf_pdesc *pd)
5578 {
5579         struct mbuf             *m0;
5580         struct route_in6         ip6route;
5581         struct route_in6        *ro;
5582         struct sockaddr_in6     *dst;
5583         struct ip6_hdr          *ip6;
5584         struct ifnet            *ifp = NULL;
5585         struct pf_addr           naddr;
5586         struct pf_src_node      *sn = NULL;
5587         int                      error = 0;
5588
5589         if (m == NULL || *m == NULL || r == NULL ||
5590             (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5591                 panic("pf_route6: invalid parameters");
5592
5593         if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
5594                 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
5595                 (*m)->m_pkthdr.pf.routed = 1;
5596         } else {
5597                 if ((*m)->m_pkthdr.pf.routed++ > 3) {
5598                         m0 = *m;
5599                         *m = NULL;
5600                         goto bad;
5601                 }
5602         }
5603
5604         if (r->rt == PF_DUPTO) {
5605                 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL)
5606                         return;
5607         } else {
5608                 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5609                         return;
5610                 m0 = *m;
5611         }
5612
5613         if (m0->m_len < sizeof(struct ip6_hdr)) {
5614                 DPFPRINTF(PF_DEBUG_URGENT,
5615                     ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n"));
5616                 goto bad;
5617         }
5618         ip6 = mtod(m0, struct ip6_hdr *);
5619
5620         ro = &ip6route;
5621         bzero((caddr_t)ro, sizeof(*ro));
5622         dst = (struct sockaddr_in6 *)&ro->ro_dst;
5623         dst->sin6_family = AF_INET6;
5624         dst->sin6_len = sizeof(*dst);
5625         dst->sin6_addr = ip6->ip6_dst;
5626
5627         /*
5628          * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags,
5629          * so make sure pf.flags is clear.
5630          *
5631          * Cheat. XXX why only in the v6 case???
5632          */
5633         if (r->rt == PF_FASTROUTE) {
5634                 m0->m_pkthdr.fw_flags |= PF_MBUF_TAGGED;
5635                 m0->m_pkthdr.pf.flags = 0;
5636                 /* XXX Re-Check when Upgrading to > 4.4 */
5637                 m0->m_pkthdr.pf.statekey = NULL;
5638                 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
5639                 return;
5640         }
5641
5642         if (TAILQ_EMPTY(&r->rpool.list)) {
5643                 DPFPRINTF(PF_DEBUG_URGENT,
5644                     ("pf_route6: TAILQ_EMPTY(&r->rpool.list)\n"));
5645                 goto bad;
5646         }
5647         if (s == NULL) {
5648                 pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src,
5649                     &naddr, NULL, &sn);
5650                 if (!PF_AZERO(&naddr, AF_INET6))
5651                         PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5652                             &naddr, AF_INET6);
5653                 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL;
5654         } else {
5655                 if (!PF_AZERO(&s->rt_addr, AF_INET6))
5656                         PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5657                             &s->rt_addr, AF_INET6);
5658                 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5659         }
5660         if (ifp == NULL)
5661                 goto bad;
5662
5663         if (oifp != ifp) {
5664                 if (pf_test6(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) {
5665                         goto bad;
5666                 } else if (m0 == NULL) {
5667                         goto done;
5668                 }
5669                 if (m0->m_len < sizeof(struct ip6_hdr)) {
5670                         DPFPRINTF(PF_DEBUG_URGENT,
5671                             ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n"));
5672                         goto bad;
5673                 }
5674                 ip6 = mtod(m0, struct ip6_hdr *);
5675         }
5676
5677         /*
5678          * If the packet is too large for the outgoing interface,
5679          * send back an icmp6 error.
5680          */
5681         if (IN6_IS_ADDR_LINKLOCAL(&dst->sin6_addr))
5682                 dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
5683         if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) {
5684                 error = nd6_output(ifp, ifp, m0, dst, NULL);
5685         } else {
5686                 in6_ifstat_inc(ifp, ifs6_in_toobig);
5687                 if (r->rt != PF_DUPTO)
5688                         icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
5689                 else
5690                         goto bad;
5691         }
5692
5693 done:
5694         if (r->rt != PF_DUPTO)
5695                 *m = NULL;
5696         return;
5697
5698 bad:
5699         m_freem(m0);
5700         goto done;
5701 }
5702 #endif /* INET6 */
5703
5704
5705 /*
5706  * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag
5707  *   off is the offset where the protocol header starts
5708  *   len is the total length of protocol header plus payload
5709  * returns 0 when the checksum is valid, otherwise returns 1.
5710  */
5711 /*
5712  * XXX
5713  * FreeBSD supports cksum offload for the following drivers.
5714  * em(4), gx(4), lge(4), nge(4), ti(4), xl(4)
5715  * If we can make full use of it we would outperform ipfw/ipfilter in
5716  * very heavy traffic.
5717  * I have not tested 'cause I don't have NICs that supports cksum offload.
5718  * (There might be problems. Typical phenomena would be
5719  *   1. No route message for UDP packet.
5720  *   2. No connection acceptance from external hosts regardless of rule set.)
5721  */
5722 int
5723 pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p,
5724     sa_family_t af)
5725 {
5726         u_int16_t sum = 0;
5727         int hw_assist = 0;
5728         struct ip *ip;
5729
5730         if (off < sizeof(struct ip) || len < sizeof(struct udphdr))
5731                 return (1);
5732         if (m->m_pkthdr.len < off + len)
5733                 return (1);
5734
5735         switch (p) {
5736         case IPPROTO_TCP:
5737         case IPPROTO_UDP:
5738                 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
5739                         if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) {
5740                                 sum = m->m_pkthdr.csum_data;
5741                         } else {
5742                                 ip = mtod(m, struct ip *);
5743                                 sum = in_pseudo(ip->ip_src.s_addr,
5744                                         ip->ip_dst.s_addr, htonl((u_short)len +
5745                                         m->m_pkthdr.csum_data + p));
5746                         }
5747                         sum ^= 0xffff;
5748                         ++hw_assist;
5749                 }
5750                 break;
5751         case IPPROTO_ICMP:
5752 #ifdef INET6
5753         case IPPROTO_ICMPV6:
5754 #endif /* INET6 */
5755                 break;
5756         default:
5757                 return (1);
5758         }
5759
5760         if (!hw_assist) {
5761                 switch (af) {
5762                 case AF_INET:
5763                         if (p == IPPROTO_ICMP) {
5764                                 if (m->m_len < off)
5765                                         return (1);
5766                                 m->m_data += off;
5767                                 m->m_len -= off;
5768                                 sum = in_cksum(m, len);
5769                                 m->m_data -= off;
5770                                 m->m_len += off;
5771                         } else {
5772                                 if (m->m_len < sizeof(struct ip))
5773                                         return (1);
5774                                 sum = in_cksum_range(m, p, off, len);
5775                                 if (sum == 0) {
5776                                         m->m_pkthdr.csum_flags |=
5777                                             (CSUM_DATA_VALID |
5778                                              CSUM_PSEUDO_HDR);
5779                                         m->m_pkthdr.csum_data = 0xffff;
5780                                 }
5781                         }
5782                         break;
5783 #ifdef INET6
5784                 case AF_INET6:
5785                         if (m->m_len < sizeof(struct ip6_hdr))
5786                                 return (1);
5787                         sum = in6_cksum(m, p, off, len);
5788                         /*
5789                          * XXX
5790                          * IPv6 H/W cksum off-load not supported yet!
5791                          *
5792                          * if (sum == 0) {
5793                          *      m->m_pkthdr.csum_flags |=
5794                          *          (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
5795                          *      m->m_pkthdr.csum_data = 0xffff;
5796                          *}
5797                          */
5798                         break;
5799 #endif /* INET6 */
5800                 default:
5801                         return (1);
5802                 }
5803         }
5804         if (sum) {
5805                 switch (p) {
5806                 case IPPROTO_TCP:
5807                         tcpstat.tcps_rcvbadsum++;
5808                         break;
5809                 case IPPROTO_UDP:
5810                         udpstat.udps_badsum++;
5811                         break;
5812                 case IPPROTO_ICMP:
5813                         icmpstat.icps_checksum++;
5814                         break;
5815 #ifdef INET6
5816                 case IPPROTO_ICMPV6:
5817                         icmp6stat.icp6s_checksum++;
5818                         break;
5819 #endif /* INET6 */
5820                 }
5821                 return (1);
5822         }
5823         return (0);
5824 }
5825
5826 struct pf_divert *
5827 pf_find_divert(struct mbuf *m)
5828 {
5829         struct m_tag    *mtag;
5830
5831         if ((mtag = m_tag_find(m, PACKET_TAG_PF_DIVERT, NULL)) == NULL)
5832                 return (NULL);
5833
5834         return ((struct pf_divert *)(mtag + 1));
5835 }
5836
5837 struct pf_divert *
5838 pf_get_divert(struct mbuf *m)
5839 {
5840         struct m_tag    *mtag;
5841
5842         if ((mtag = m_tag_find(m, PACKET_TAG_PF_DIVERT, NULL)) == NULL) {
5843                 mtag = m_tag_get(PACKET_TAG_PF_DIVERT, sizeof(struct pf_divert),
5844                     M_NOWAIT);
5845                 if (mtag == NULL)
5846                         return (NULL);
5847                 bzero(mtag + 1, sizeof(struct pf_divert));
5848                 m_tag_prepend(m, mtag);
5849         }
5850
5851         return ((struct pf_divert *)(mtag + 1));
5852 }
5853
5854 #ifdef INET
5855 int
5856 pf_test(int dir, struct ifnet *ifp, struct mbuf **m0,
5857     struct ether_header *eh, struct inpcb *inp)
5858 {
5859         struct pfi_kif          *kif;
5860         u_short                  action, reason = 0, log = 0;
5861         struct mbuf             *m = *m0;
5862         struct ip               *h = NULL;
5863         struct pf_rule          *a = NULL, *r = &pf_default_rule, *tr, *nr;
5864         struct pf_state         *s = NULL;
5865         struct pf_ruleset       *ruleset = NULL;
5866         struct pf_pdesc          pd;
5867         int                      off, dirndx, pqid = 0;
5868
5869         if (!pf_status.running)
5870                 return (PF_PASS);
5871
5872         memset(&pd, 0, sizeof(pd));
5873         if (ifp->if_type == IFT_CARP && ifp->if_carpdev)
5874                 kif = (struct pfi_kif *)ifp->if_carpdev->if_pf_kif;
5875         else
5876                 kif = (struct pfi_kif *)ifp->if_pf_kif;
5877
5878         if (kif == NULL) {
5879                 DPFPRINTF(PF_DEBUG_URGENT,
5880                     ("pf_test: kif == NULL, if_xname %s\n", ifp->if_xname));
5881                 return (PF_DROP);
5882         }
5883         if (kif->pfik_flags & PFI_IFLAG_SKIP)
5884                 return (PF_PASS);
5885
5886 #ifdef DIAGNOSTIC
5887         if ((m->m_flags & M_PKTHDR) == 0)
5888                 panic("non-M_PKTHDR is passed to pf_test");
5889 #endif /* DIAGNOSTIC */
5890
5891         if (m->m_pkthdr.len < (int)sizeof(*h)) {
5892                 action = PF_DROP;
5893                 REASON_SET(&reason, PFRES_SHORT);
5894                 log = 1;
5895                 goto done;
5896         }
5897
5898         /*
5899          * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags,
5900          * so make sure pf.flags is clear.
5901          */
5902         if (m->m_pkthdr.fw_flags & PF_MBUF_TAGGED)
5903                 return (PF_PASS);
5904         m->m_pkthdr.pf.flags = 0;
5905         /* Re-Check when updating to > 4.4 */
5906         m->m_pkthdr.pf.statekey = NULL;
5907
5908         /* We do IP header normalization and packet reassembly here */
5909         if (pf_normalize_ip(m0, dir, kif, &reason, &pd) != PF_PASS) {
5910                 action = PF_DROP;
5911                 goto done;
5912         }
5913         m = *m0;        /* pf_normalize messes with m0 */
5914         h = mtod(m, struct ip *);
5915
5916         off = h->ip_hl << 2;
5917         if (off < (int)sizeof(*h)) {
5918                 action = PF_DROP;
5919                 REASON_SET(&reason, PFRES_SHORT);
5920                 log = 1;
5921                 goto done;
5922         }
5923
5924         pd.src = (struct pf_addr *)&h->ip_src;
5925         pd.dst = (struct pf_addr *)&h->ip_dst;
5926         pd.sport = pd.dport = NULL;
5927         pd.ip_sum = &h->ip_sum;
5928         pd.proto_sum = NULL;
5929         pd.proto = h->ip_p;
5930         pd.dir = dir;
5931         pd.sidx = (dir == PF_IN) ? 0 : 1;
5932         pd.didx = (dir == PF_IN) ? 1 : 0;
5933         pd.af = AF_INET;
5934         pd.tos = h->ip_tos;
5935         pd.tot_len = h->ip_len;
5936         pd.eh = eh;
5937
5938         /* handle fragments that didn't get reassembled by normalization */
5939         if (h->ip_off & (IP_MF | IP_OFFMASK)) {
5940                 action = pf_test_fragment(&r, dir, kif, m, h,
5941                     &pd, &a, &ruleset);
5942                 goto done;
5943         }
5944
5945         switch (h->ip_p) {
5946
5947         case IPPROTO_TCP: {
5948                 struct tcphdr   th;
5949
5950                 pd.hdr.tcp = &th;
5951                 if (!pf_pull_hdr(m, off, &th, sizeof(th),
5952                     &action, &reason, AF_INET)) {
5953                         log = action != PF_PASS;
5954                         goto done;
5955                 }
5956                 pd.p_len = pd.tot_len - off - (th.th_off << 2);
5957                 if ((th.th_flags & TH_ACK) && pd.p_len == 0)
5958                         pqid = 1;
5959                 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
5960                 if (action == PF_DROP)
5961                         goto done;
5962                 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
5963                     &reason);
5964                 if (action == PF_PASS) {
5965                         pfsync_update_state(s);
5966                         r = s->rule.ptr;
5967                         a = s->anchor.ptr;
5968                         log = s->log;
5969                 } else if (s == NULL)
5970                         action = pf_test_rule(&r, &s, dir, kif,
5971                             m, off, h, &pd, &a, &ruleset, NULL, inp);
5972                 break;
5973         }
5974
5975         case IPPROTO_UDP: {
5976                 struct udphdr   uh;
5977
5978                 pd.hdr.udp = &uh;
5979                 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
5980                     &action, &reason, AF_INET)) {
5981                         log = action != PF_PASS;
5982                         goto done;
5983                 }
5984                 if (uh.uh_dport == 0 ||
5985                     ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
5986                     ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
5987                         action = PF_DROP;
5988                         REASON_SET(&reason, PFRES_SHORT);
5989                         goto done;
5990                 }
5991                 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
5992                 if (action == PF_PASS) {
5993                         pfsync_update_state(s);
5994                         r = s->rule.ptr;
5995                         a = s->anchor.ptr;
5996                         log = s->log;
5997                 } else if (s == NULL)
5998                         action = pf_test_rule(&r, &s, dir, kif,
5999                             m, off, h, &pd, &a, &ruleset, NULL, inp);
6000                 break;
6001         }
6002
6003         case IPPROTO_ICMP: {
6004                 struct icmp     ih;
6005
6006                 pd.hdr.icmp = &ih;
6007                 if (!pf_pull_hdr(m, off, &ih, ICMP_MINLEN,
6008                     &action, &reason, AF_INET)) {
6009                         log = action != PF_PASS;
6010                         goto done;
6011                 }
6012                 action = pf_test_state_icmp(&s, dir, kif, m, off, h, &pd,
6013                     &reason);
6014                 if (action == PF_PASS) {
6015                         pfsync_update_state(s);
6016                         r = s->rule.ptr;
6017                         a = s->anchor.ptr;
6018                         log = s->log;
6019                 } else if (s == NULL)
6020                         action = pf_test_rule(&r, &s, dir, kif,
6021                             m, off, h, &pd, &a, &ruleset, NULL, inp);
6022                 break;
6023         }
6024
6025         default:
6026                 action = pf_test_state_other(&s, dir, kif, m, &pd);
6027                 if (action == PF_PASS) {
6028                         pfsync_update_state(s);
6029                         r = s->rule.ptr;
6030                         a = s->anchor.ptr;
6031                         log = s->log;
6032                 } else if (s == NULL)
6033                         action = pf_test_rule(&r, &s, dir, kif, m, off, h,
6034                             &pd, &a, &ruleset, NULL, inp);
6035                 break;
6036         }
6037
6038 done:
6039         if (action == PF_PASS && h->ip_hl > 5 &&
6040             !((s && s->state_flags & PFSTATE_ALLOWOPTS) || r->allow_opts)) {
6041                 action = PF_DROP;
6042                 REASON_SET(&reason, PFRES_IPOPTIONS);
6043                 log = 1;
6044                 DPFPRINTF(PF_DEBUG_MISC,
6045                     ("pf: dropping packet with ip options\n"));
6046         }
6047
6048         if ((s && s->tag) || r->rtableid)
6049                 pf_tag_packet(m, s ? s->tag : 0, r->rtableid);
6050
6051 #if 0
6052         if (dir == PF_IN && s && s->key[PF_SK_STACK])
6053                 m->m_pkthdr.pf.statekey = s->key[PF_SK_STACK];
6054 #endif
6055
6056 #ifdef ALTQ
6057         if (action == PF_PASS && r->qid) {
6058                 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
6059                 if (pqid || (pd.tos & IPTOS_LOWDELAY))
6060                         m->m_pkthdr.pf.qid = r->pqid;
6061                 else
6062                         m->m_pkthdr.pf.qid = r->qid;
6063                 m->m_pkthdr.pf.ecn_af = AF_INET;
6064                 m->m_pkthdr.pf.hdr = h;
6065                 /* add connection hash for fairq */
6066                 if (s) {
6067                         /* for fairq */
6068                         m->m_pkthdr.pf.state_hash = s->hash;
6069                         m->m_pkthdr.pf.flags |= PF_TAG_STATE_HASHED;
6070                 }
6071         }
6072 #endif /* ALTQ */
6073
6074         /*
6075          * connections redirected to loopback should not match sockets
6076          * bound specifically to loopback due to security implications,
6077          * see tcp_input() and in_pcblookup_listen().
6078          */
6079         if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
6080             pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
6081             (s->nat_rule.ptr->action == PF_RDR ||
6082             s->nat_rule.ptr->action == PF_BINAT) &&
6083             (ntohl(pd.dst->v4.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET)
6084                 m->m_pkthdr.pf.flags |= PF_TAG_TRANSLATE_LOCALHOST;
6085
6086         if (dir == PF_IN && action == PF_PASS && r->divert.port) {
6087                 struct pf_divert *divert;
6088
6089                 if ((divert = pf_get_divert(m))) {
6090                         m->m_pkthdr.pf.flags |= PF_TAG_DIVERTED;
6091                         divert->port = r->divert.port;
6092                         divert->addr.ipv4 = r->divert.addr.v4;
6093                 }
6094         }
6095
6096         if (log) {
6097                 struct pf_rule *lr;
6098
6099                 if (s != NULL && s->nat_rule.ptr != NULL &&
6100                     s->nat_rule.ptr->log & PF_LOG_ALL)
6101                         lr = s->nat_rule.ptr;
6102                 else
6103                         lr = r;
6104                 PFLOG_PACKET(kif, h, m, AF_INET, dir, reason, lr, a, ruleset,
6105                     &pd);
6106         }
6107
6108         kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
6109         kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++;
6110
6111         if (action == PF_PASS || r->action == PF_DROP) {
6112                 dirndx = (dir == PF_OUT);
6113                 r->packets[dirndx]++;
6114                 r->bytes[dirndx] += pd.tot_len;
6115                 if (a != NULL) {
6116                         a->packets[dirndx]++;
6117                         a->bytes[dirndx] += pd.tot_len;
6118                 }
6119                 if (s != NULL) {
6120                         if (s->nat_rule.ptr != NULL) {
6121                                 s->nat_rule.ptr->packets[dirndx]++;
6122                                 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len;
6123                         }
6124                         if (s->src_node != NULL) {
6125                                 s->src_node->packets[dirndx]++;
6126                                 s->src_node->bytes[dirndx] += pd.tot_len;
6127                         }
6128                         if (s->nat_src_node != NULL) {
6129                                 s->nat_src_node->packets[dirndx]++;
6130                                 s->nat_src_node->bytes[dirndx] += pd.tot_len;
6131                         }
6132                         dirndx = (dir == s->direction) ? 0 : 1;
6133                         s->packets[dirndx]++;
6134                         s->bytes[dirndx] += pd.tot_len;
6135                 }
6136                 tr = r;
6137                 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
6138                 if (nr != NULL && r == &pf_default_rule)
6139                         tr = nr;
6140                 if (tr->src.addr.type == PF_ADDR_TABLE)
6141                         pfr_update_stats(tr->src.addr.p.tbl,
6142                             (s == NULL) ? pd.src :
6143                             &s->key[(s->direction == PF_IN)]->
6144                                 addr[(s->direction == PF_OUT)],
6145                             pd.af, pd.tot_len, dir == PF_OUT,
6146                             r->action == PF_PASS, tr->src.neg);
6147                 if (tr->dst.addr.type == PF_ADDR_TABLE)
6148                         pfr_update_stats(tr->dst.addr.p.tbl,
6149                             (s == NULL) ? pd.dst :
6150                             &s->key[(s->direction == PF_IN)]->
6151                                 addr[(s->direction == PF_IN)],
6152                             pd.af, pd.tot_len, dir == PF_OUT,
6153                             r->action == PF_PASS, tr->dst.neg);
6154         }
6155
6156
6157         if (action == PF_SYNPROXY_DROP) {
6158                 m_freem(*m0);
6159                 *m0 = NULL;
6160                 action = PF_PASS;
6161         } else if (r->rt)
6162                 /* pf_route can free the mbuf causing *m0 to become NULL */
6163                 pf_route(m0, r, dir, kif->pfik_ifp, s, &pd);
6164
6165         return (action);
6166 }
6167 #endif /* INET */
6168
6169 #ifdef INET6
6170 int
6171 pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0,
6172     struct ether_header *eh, struct inpcb *inp)
6173 {
6174         struct pfi_kif          *kif;
6175         u_short                  action, reason = 0, log = 0;
6176         struct mbuf             *m = *m0, *n = NULL;
6177         struct ip6_hdr          *h = NULL;
6178         struct pf_rule          *a = NULL, *r = &pf_default_rule, *tr, *nr;
6179         struct pf_state         *s = NULL;
6180         struct pf_ruleset       *ruleset = NULL;
6181         struct pf_pdesc          pd;
6182         int                      off, terminal = 0, dirndx, rh_cnt = 0;
6183
6184         if (!pf_status.running)
6185                 return (PF_PASS);
6186
6187         memset(&pd, 0, sizeof(pd));
6188         if (ifp->if_type == IFT_CARP && ifp->if_carpdev)
6189                 kif = (struct pfi_kif *)ifp->if_carpdev->if_pf_kif;
6190         else
6191                 kif = (struct pfi_kif *)ifp->if_pf_kif;
6192
6193         if (kif == NULL) {
6194                 DPFPRINTF(PF_DEBUG_URGENT,
6195                     ("pf_test6: kif == NULL, if_xname %s\n", ifp->if_xname));
6196                 return (PF_DROP);
6197         }
6198         if (kif->pfik_flags & PFI_IFLAG_SKIP)
6199                 return (PF_PASS);
6200
6201 #ifdef DIAGNOSTIC
6202         if ((m->m_flags & M_PKTHDR) == 0)
6203                 panic("non-M_PKTHDR is passed to pf_test6");
6204 #endif /* DIAGNOSTIC */
6205
6206         if (m->m_pkthdr.len < (int)sizeof(*h)) {
6207                 action = PF_DROP;
6208                 REASON_SET(&reason, PFRES_SHORT);
6209                 log = 1;
6210                 goto done;
6211         }
6212
6213         /*
6214          * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags,
6215          * so make sure pf.flags is clear.
6216          */
6217         if (m->m_pkthdr.fw_flags & PF_MBUF_TAGGED)
6218                 return (PF_PASS);
6219         m->m_pkthdr.pf.flags = 0;
6220         /* Re-Check when updating to > 4.4 */
6221         m->m_pkthdr.pf.statekey = NULL;
6222
6223         /* We do IP header normalization and packet reassembly here */
6224         if (pf_normalize_ip6(m0, dir, kif, &reason, &pd) != PF_PASS) {
6225                 action = PF_DROP;
6226                 goto done;
6227         }
6228         m = *m0;        /* pf_normalize messes with m0 */
6229         h = mtod(m, struct ip6_hdr *);
6230
6231 #if 1
6232         /*
6233          * we do not support jumbogram yet.  if we keep going, zero ip6_plen
6234          * will do something bad, so drop the packet for now.
6235          */
6236         if (htons(h->ip6_plen) == 0) {
6237                 action = PF_DROP;
6238                 REASON_SET(&reason, PFRES_NORM);        /*XXX*/
6239                 goto done;
6240         }
6241 #endif
6242
6243         pd.src = (struct pf_addr *)&h->ip6_src;
6244         pd.dst = (struct pf_addr *)&h->ip6_dst;
6245         pd.sport = pd.dport = NULL;
6246         pd.ip_sum = NULL;
6247         pd.proto_sum = NULL;
6248         pd.dir = dir;
6249         pd.sidx = (dir == PF_IN) ? 0 : 1;
6250         pd.didx = (dir == PF_IN) ? 1 : 0;
6251         pd.af = AF_INET6;
6252         pd.tos = 0;
6253         pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr);
6254         pd.eh = eh;
6255
6256         off = ((caddr_t)h - m->m_data) + sizeof(struct ip6_hdr);
6257         pd.proto = h->ip6_nxt;
6258         do {
6259                 switch (pd.proto) {
6260                 case IPPROTO_FRAGMENT:
6261                         action = pf_test_fragment(&r, dir, kif, m, h,
6262                             &pd, &a, &ruleset);
6263                         if (action == PF_DROP)
6264                                 REASON_SET(&reason, PFRES_FRAG);
6265                         goto done;
6266                 case IPPROTO_ROUTING: {
6267                         struct ip6_rthdr rthdr;
6268
6269                         if (rh_cnt++) {
6270                                 DPFPRINTF(PF_DEBUG_MISC,
6271                                     ("pf: IPv6 more than one rthdr\n"));
6272                                 action = PF_DROP;
6273                                 REASON_SET(&reason, PFRES_IPOPTIONS);
6274                                 log = 1;
6275                                 goto done;
6276                         }
6277                         if (!pf_pull_hdr(m, off, &rthdr, sizeof(rthdr), NULL,
6278                             &reason, pd.af)) {
6279                                 DPFPRINTF(PF_DEBUG_MISC,
6280                                     ("pf: IPv6 short rthdr\n"));
6281                                 action = PF_DROP;
6282                                 REASON_SET(&reason, PFRES_SHORT);
6283                                 log = 1;
6284                                 goto done;
6285                         }
6286                         if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) {
6287                                 DPFPRINTF(PF_DEBUG_MISC,
6288                                     ("pf: IPv6 rthdr0\n"));
6289                                 action = PF_DROP;
6290                                 REASON_SET(&reason, PFRES_IPOPTIONS);
6291                                 log = 1;
6292                                 goto done;
6293                         }
6294                         /* FALLTHROUGH */
6295                 }
6296                 case IPPROTO_AH:
6297                 case IPPROTO_HOPOPTS:
6298                 case IPPROTO_DSTOPTS: {
6299                         /* get next header and header length */
6300                         struct ip6_ext  opt6;
6301
6302                         if (!pf_pull_hdr(m, off, &opt6, sizeof(opt6),
6303                             NULL, &reason, pd.af)) {
6304                                 DPFPRINTF(PF_DEBUG_MISC,
6305                                     ("pf: IPv6 short opt\n"));
6306                                 action = PF_DROP;
6307                                 log = 1;
6308                                 goto done;
6309                         }
6310                         if (pd.proto == IPPROTO_AH)
6311                                 off += (opt6.ip6e_len + 2) * 4;
6312                         else
6313                                 off += (opt6.ip6e_len + 1) * 8;
6314                         pd.proto = opt6.ip6e_nxt;
6315                         /* goto the next header */
6316                         break;
6317                 }
6318                 default:
6319                         terminal++;
6320                         break;
6321                 }
6322         } while (!terminal);
6323
6324         /* if there's no routing header, use unmodified mbuf for checksumming */
6325         if (!n)
6326                 n = m;
6327
6328         switch (pd.proto) {
6329
6330         case IPPROTO_TCP: {
6331                 struct tcphdr   th;
6332
6333                 pd.hdr.tcp = &th;
6334                 if (!pf_pull_hdr(m, off, &th, sizeof(th),
6335                     &action, &reason, AF_INET6)) {
6336                         log = action != PF_PASS;
6337                         goto done;
6338                 }
6339                 pd.p_len = pd.tot_len - off - (th.th_off << 2);
6340                 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
6341                 if (action == PF_DROP)
6342                         goto done;
6343                 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
6344                     &reason);
6345                 if (action == PF_PASS) {
6346                         pfsync_update_state(s);
6347                         r = s->rule.ptr;
6348                         a = s->anchor.ptr;
6349                         log = s->log;
6350                 } else if (s == NULL)
6351                         action = pf_test_rule(&r, &s, dir, kif,
6352                             m, off, h, &pd, &a, &ruleset, NULL, inp);
6353                 break;
6354         }
6355
6356         case IPPROTO_UDP: {
6357                 struct udphdr   uh;
6358
6359                 pd.hdr.udp = &uh;
6360                 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
6361                     &action, &reason, AF_INET6)) {
6362                         log = action != PF_PASS;
6363                         goto done;
6364                 }
6365                 if (uh.uh_dport == 0 ||
6366                     ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
6367                     ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
6368                         action = PF_DROP;
6369                         REASON_SET(&reason, PFRES_SHORT);
6370                         goto done;
6371                 }
6372                 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
6373                 if (action == PF_PASS) {
6374                         pfsync_update_state(s);
6375                         r = s->rule.ptr;
6376                         a = s->anchor.ptr;
6377                         log = s->log;
6378                 } else if (s == NULL)
6379                         action = pf_test_rule(&r, &s, dir, kif,
6380                             m, off, h, &pd, &a, &ruleset, NULL, inp);
6381                 break;
6382         }
6383
6384         case IPPROTO_ICMPV6: {
6385                 struct icmp6_hdr        ih;
6386
6387                 pd.hdr.icmp6 = &ih;
6388                 if (!pf_pull_hdr(m, off, &ih, sizeof(ih),
6389                     &action, &reason, AF_INET6)) {
6390                         log = action != PF_PASS;
6391                         goto done;
6392                 }
6393                 action = pf_test_state_icmp(&s, dir, kif,
6394                     m, off, h, &pd, &reason);
6395                 if (action == PF_PASS) {
6396                         pfsync_update_state(s);
6397                         r = s->rule.ptr;
6398                         a = s->anchor.ptr;
6399                         log = s->log;
6400                 } else if (s == NULL)
6401                         action = pf_test_rule(&r, &s, dir, kif,
6402                             m, off, h, &pd, &a, &ruleset, NULL, inp);
6403                 break;
6404         }
6405
6406         default:
6407                 action = pf_test_state_other(&s, dir, kif, m, &pd);
6408                 if (action == PF_PASS) {
6409                         pfsync_update_state(s);
6410                         r = s->rule.ptr;
6411                         a = s->anchor.ptr;
6412                         log = s->log;
6413                 } else if (s == NULL)
6414                         action = pf_test_rule(&r, &s, dir, kif, m, off, h,
6415                             &pd, &a, &ruleset, NULL, inp);
6416                 break;
6417         }
6418
6419 done:
6420         if (n != m) {
6421                 m_freem(n);
6422                 n = NULL;
6423         }
6424
6425         /* handle dangerous IPv6 extension headers. */
6426         if (action == PF_PASS && rh_cnt &&
6427             !((s && s->state_flags & PFSTATE_ALLOWOPTS) || r->allow_opts)) {
6428                 action = PF_DROP;
6429                 REASON_SET(&reason, PFRES_IPOPTIONS);
6430                 log = 1;
6431                 DPFPRINTF(PF_DEBUG_MISC,
6432                     ("pf: dropping packet with dangerous v6 headers\n"));
6433         }
6434
6435         if ((s && s->tag) || r->rtableid)
6436                 pf_tag_packet(m, s ? s->tag : 0, r->rtableid);
6437
6438 #if 0
6439         if (dir == PF_IN && s && s->key[PF_SK_STACK])
6440                 m->m_pkthdr.pf.statekey = s->key[PF_SK_STACK];
6441 #endif
6442
6443 #ifdef ALTQ
6444         if (action == PF_PASS && r->qid) {
6445                 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
6446                 if (pd.tos & IPTOS_LOWDELAY)
6447                         m->m_pkthdr.pf.qid = r->pqid;
6448                 else
6449                         m->m_pkthdr.pf.qid = r->qid;
6450                 m->m_pkthdr.pf.ecn_af = AF_INET6;
6451                 m->m_pkthdr.pf.hdr = h;
6452                 if (s) {
6453                         /* for fairq */
6454                         m->m_pkthdr.pf.state_hash = s->hash;
6455                         m->m_pkthdr.pf.flags |= PF_TAG_STATE_HASHED;
6456                 }
6457         }
6458 #endif /* ALTQ */
6459
6460         if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
6461             pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
6462             (s->nat_rule.ptr->action == PF_RDR ||
6463             s->nat_rule.ptr->action == PF_BINAT) &&
6464             IN6_IS_ADDR_LOOPBACK(&pd.dst->v6))
6465                 m->m_pkthdr.pf.flags |= PF_TAG_TRANSLATE_LOCALHOST;
6466
6467         if (dir == PF_IN && action == PF_PASS && r->divert.port) {
6468                 struct pf_divert *divert;
6469
6470                 if ((divert = pf_get_divert(m))) {
6471                         m->m_pkthdr.pf.flags |= PF_TAG_DIVERTED;
6472                         divert->port = r->divert.port;
6473                         divert->addr.ipv6 = r->divert.addr.v6;
6474                 }
6475         }
6476
6477         if (log) {
6478                 struct pf_rule *lr;
6479
6480                 if (s != NULL && s->nat_rule.ptr != NULL &&
6481                     s->nat_rule.ptr->log & PF_LOG_ALL)
6482                         lr = s->nat_rule.ptr;
6483                 else
6484                         lr = r;
6485                 PFLOG_PACKET(kif, h, m, AF_INET6, dir, reason, lr, a, ruleset,
6486                     &pd);
6487         }
6488
6489         kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
6490         kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++;
6491
6492         if (action == PF_PASS || r->action == PF_DROP) {
6493                 dirndx = (dir == PF_OUT);
6494                 r->packets[dirndx]++;
6495                 r->bytes[dirndx] += pd.tot_len;
6496                 if (a != NULL) {
6497                         a->packets[dirndx]++;
6498                         a->bytes[dirndx] += pd.tot_len;
6499                 }
6500                 if (s != NULL) {
6501                         if (s->nat_rule.ptr != NULL) {
6502                                 s->nat_rule.ptr->packets[dirndx]++;
6503                                 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len;
6504                         }
6505                         if (s->src_node != NULL) {
6506                                 s->src_node->packets[dirndx]++;
6507                                 s->src_node->bytes[dirndx] += pd.tot_len;
6508                         }
6509                         if (s->nat_src_node != NULL) {
6510                                 s->nat_src_node->packets[dirndx]++;
6511                                 s->nat_src_node->bytes[dirndx] += pd.tot_len;
6512                         }
6513                         dirndx = (dir == s->direction) ? 0 : 1;
6514                         s->packets[dirndx]++;
6515                         s->bytes[dirndx] += pd.tot_len;
6516                 }
6517                 tr = r;
6518                 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
6519                 if (nr != NULL && r == &pf_default_rule)
6520                         tr = nr;
6521                 if (tr->src.addr.type == PF_ADDR_TABLE)
6522                         pfr_update_stats(tr->src.addr.p.tbl,
6523                             (s == NULL) ? pd.src :
6524                             &s->key[(s->direction == PF_IN)]->addr[0],
6525                             pd.af, pd.tot_len, dir == PF_OUT,
6526                             r->action == PF_PASS, tr->src.neg);
6527                 if (tr->dst.addr.type == PF_ADDR_TABLE)
6528                         pfr_update_stats(tr->dst.addr.p.tbl,
6529                             (s == NULL) ? pd.dst :
6530                             &s->key[(s->direction == PF_IN)]->addr[1],
6531                             pd.af, pd.tot_len, dir == PF_OUT,
6532                             r->action == PF_PASS, tr->dst.neg);
6533         }
6534
6535
6536         if (action == PF_SYNPROXY_DROP) {
6537                 m_freem(*m0);
6538                 *m0 = NULL;
6539                 action = PF_PASS;
6540         } else if (r->rt)
6541                 /* pf_route6 can free the mbuf causing *m0 to become NULL */
6542                 pf_route6(m0, r, dir, kif->pfik_ifp, s, &pd);
6543
6544         return (action);
6545 }
6546 #endif /* INET6 */
6547
6548 int
6549 pf_check_congestion(struct ifqueue *ifq)
6550 {
6551                 return (0);
6552 }