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 static void
2809 in_pcblookup_hash_handler(netmsg_t msg)
2810 {
2811         struct netmsg_hashlookup *rmsg = (struct netmsg_hashlookup *)msg;
2812
2813         if (rmsg->nm_af == AF_INET)
2814                 *rmsg->nm_pinp = in_pcblookup_hash(rmsg->nm_pcbinfo,
2815                     rmsg->nm_saddr->v4, rmsg->nm_sport, rmsg->nm_daddr->v4,
2816                     rmsg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2817 #ifdef INET6
2818         else
2819                 *rmsg->nm_pinp = in6_pcblookup_hash(rmsg->nm_pcbinfo,
2820                     &rmsg->nm_saddr->v6, rmsg->nm_sport, &rmsg->nm_daddr->v6,
2821                     rmsg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2822 #endif /* INET6 */
2823         lwkt_replymsg(&rmsg->base.lmsg, 0);
2824 }
2825 #endif /* SMP */
2826
2827 int
2828 pf_socket_lookup(int direction, struct pf_pdesc *pd)
2829 {
2830         struct pf_addr          *saddr, *daddr;
2831         u_int16_t                sport, dport;
2832         struct inpcbinfo        *pi;
2833         struct inpcb            *inp;
2834 #ifdef SMP
2835         struct netmsg_hashlookup *msg = NULL;
2836 #endif
2837         int                      pi_cpu = 0;
2838
2839         if (pd == NULL)
2840                 return (-1);
2841         pd->lookup.uid = UID_MAX;
2842         pd->lookup.gid = GID_MAX;
2843         pd->lookup.pid = NO_PID;
2844         if (direction == PF_IN) {
2845                 saddr = pd->src;
2846                 daddr = pd->dst;
2847         } else {
2848                 saddr = pd->dst;
2849                 daddr = pd->src;
2850         }
2851         switch (pd->proto) {
2852         case IPPROTO_TCP:
2853                 if (pd->hdr.tcp == NULL)
2854                         return (-1);
2855                 sport = pd->hdr.tcp->th_sport;
2856                 dport = pd->hdr.tcp->th_dport;
2857
2858                 pi_cpu = tcp_addrcpu(saddr->v4.s_addr, sport, daddr->v4.s_addr, dport);
2859                 pi = &tcbinfo[pi_cpu];
2860 #ifdef SMP
2861                 /*
2862                  * Our netstack runs lockless on MP systems
2863                  * (only for TCP connections at the moment).
2864                  *
2865                  * As we are not allowed to read another CPU's tcbinfo,
2866                  * we have to ask that CPU via remote call to search the
2867                  * table for us.
2868                  *
2869                  * Prepare a msg iff data belongs to another CPU.
2870                  */
2871                 if (pi_cpu != mycpu->gd_cpuid) {
2872                         msg = kmalloc(sizeof(*msg), M_LWKTMSG, M_INTWAIT);
2873                         netmsg_init(&msg->base, NULL, &netisr_afree_rport,
2874                                     0, in_pcblookup_hash_handler);
2875                         msg->nm_pinp = &inp;
2876                         msg->nm_pcbinfo = pi;
2877                         msg->nm_saddr = saddr;
2878                         msg->nm_sport = sport;
2879                         msg->nm_daddr = daddr;
2880                         msg->nm_dport = dport;
2881                         msg->nm_af = pd->af;
2882                 }
2883 #endif /* SMP */
2884                 break;
2885         case IPPROTO_UDP:
2886                 if (pd->hdr.udp == NULL)
2887                         return (-1);
2888                 sport = pd->hdr.udp->uh_sport;
2889                 dport = pd->hdr.udp->uh_dport;
2890                 pi = &udbinfo;
2891                 break;
2892         default:
2893                 return (-1);
2894         }
2895         if (direction != PF_IN) {
2896                 u_int16_t       p;
2897
2898                 p = sport;
2899                 sport = dport;
2900                 dport = p;
2901         }
2902         switch (pd->af) {
2903 #ifdef INET6
2904         case AF_INET6:
2905 #ifdef SMP
2906                 /*
2907                  * Query other CPU, second part
2908                  *
2909                  * msg only gets initialized when:
2910                  * 1) packet is TCP
2911                  * 2) the info belongs to another CPU
2912                  *
2913                  * Use some switch/case magic to avoid code duplication.
2914                  */
2915                 if (msg == NULL)
2916 #endif /* SMP */
2917                 {
2918                         inp = in6_pcblookup_hash(pi, &saddr->v6, sport,
2919                             &daddr->v6, dport, INPLOOKUP_WILDCARD, NULL);
2920
2921                         if (inp == NULL)
2922                                 return (-1);
2923                         break;
2924                 }
2925                 /* FALLTHROUGH if SMP and on other CPU */
2926 #endif /* INET6 */
2927         case AF_INET:
2928 #ifdef SMP
2929                 if (msg != NULL) {
2930                         lwkt_domsg(cpu_portfn(pi_cpu),
2931                                      &msg->base.lmsg, 0);
2932                 } else
2933 #endif /* SMP */
2934                 {
2935                         inp = in_pcblookup_hash(pi, saddr->v4, sport, daddr->v4,
2936                             dport, INPLOOKUP_WILDCARD, NULL);
2937                 }
2938                 if (inp == NULL)
2939                         return (-1);
2940                 break;
2941
2942         default:
2943                 return (-1);
2944         }
2945         pd->lookup.uid = inp->inp_socket->so_cred->cr_uid;
2946         pd->lookup.gid = inp->inp_socket->so_cred->cr_groups[0];
2947         return (1);
2948 }
2949
2950 u_int8_t
2951 pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2952 {
2953         int              hlen;
2954         u_int8_t         hdr[60];
2955         u_int8_t        *opt, optlen;
2956         u_int8_t         wscale = 0;
2957
2958         hlen = th_off << 2;             /* hlen <= sizeof(hdr) */
2959         if (hlen <= sizeof(struct tcphdr))
2960                 return (0);
2961         if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2962                 return (0);
2963         opt = hdr + sizeof(struct tcphdr);
2964         hlen -= sizeof(struct tcphdr);
2965         while (hlen >= 3) {
2966                 switch (*opt) {
2967                 case TCPOPT_EOL:
2968                 case TCPOPT_NOP:
2969                         ++opt;
2970                         --hlen;
2971                         break;
2972                 case TCPOPT_WINDOW:
2973                         wscale = opt[2];
2974                         if (wscale > TCP_MAX_WINSHIFT)
2975                                 wscale = TCP_MAX_WINSHIFT;
2976                         wscale |= PF_WSCALE_FLAG;
2977                         /* FALLTHROUGH */
2978                 default:
2979                         optlen = opt[1];
2980                         if (optlen < 2)
2981                                 optlen = 2;
2982                         hlen -= optlen;
2983                         opt += optlen;
2984                         break;
2985                 }
2986         }
2987         return (wscale);
2988 }
2989
2990 u_int16_t
2991 pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2992 {
2993         int              hlen;
2994         u_int8_t         hdr[60];
2995         u_int8_t        *opt, optlen;
2996         u_int16_t        mss = tcp_mssdflt;
2997
2998         hlen = th_off << 2;     /* hlen <= sizeof(hdr) */
2999         if (hlen <= sizeof(struct tcphdr))
3000                 return (0);
3001         if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
3002                 return (0);
3003         opt = hdr + sizeof(struct tcphdr);
3004         hlen -= sizeof(struct tcphdr);
3005         while (hlen >= TCPOLEN_MAXSEG) {
3006                 switch (*opt) {
3007                 case TCPOPT_EOL:
3008                 case TCPOPT_NOP:
3009                         ++opt;
3010                         --hlen;
3011                         break;
3012                 case TCPOPT_MAXSEG:
3013                         bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2);
3014                         /* FALLTHROUGH */
3015                 default:
3016                         optlen = opt[1];
3017                         if (optlen < 2)
3018                                 optlen = 2;
3019                         hlen -= optlen;
3020                         opt += optlen;
3021                         break;
3022                 }
3023         }
3024         return (mss);
3025 }
3026
3027 u_int16_t
3028 pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer)
3029 {
3030 #ifdef INET
3031         struct sockaddr_in      *dst;
3032         struct route             ro;
3033 #endif /* INET */
3034 #ifdef INET6
3035         struct sockaddr_in6     *dst6;
3036         struct route_in6         ro6;
3037 #endif /* INET6 */
3038         struct rtentry          *rt = NULL;
3039         int                      hlen = 0;
3040         u_int16_t                mss = tcp_mssdflt;
3041
3042         switch (af) {
3043 #ifdef INET
3044         case AF_INET:
3045                 hlen = sizeof(struct ip);
3046                 bzero(&ro, sizeof(ro));
3047                 dst = (struct sockaddr_in *)&ro.ro_dst;
3048                 dst->sin_family = AF_INET;
3049                 dst->sin_len = sizeof(*dst);
3050                 dst->sin_addr = addr->v4;
3051                 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
3052                 rt = ro.ro_rt;
3053                 break;
3054 #endif /* INET */
3055 #ifdef INET6
3056         case AF_INET6:
3057                 hlen = sizeof(struct ip6_hdr);
3058                 bzero(&ro6, sizeof(ro6));
3059                 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst;
3060                 dst6->sin6_family = AF_INET6;
3061                 dst6->sin6_len = sizeof(*dst6);
3062                 dst6->sin6_addr = addr->v6;
3063                 rtalloc_ign((struct route *)&ro6, (RTF_CLONING | RTF_PRCLONING));
3064                 rt = ro6.ro_rt;
3065                 break;
3066 #endif /* INET6 */
3067         }
3068
3069         if (rt && rt->rt_ifp) {
3070                 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr);
3071                 mss = max(tcp_mssdflt, mss);
3072                 RTFREE(rt);
3073         }
3074         mss = min(mss, offer);
3075         mss = max(mss, 64);             /* sanity - at least max opt space */
3076         return (mss);
3077 }
3078
3079 void
3080 pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr)
3081 {
3082         struct pf_rule *r = s->rule.ptr;
3083
3084         s->rt_kif = NULL;
3085         if (!r->rt || r->rt == PF_FASTROUTE)
3086                 return;
3087         switch (s->key[PF_SK_WIRE]->af) {
3088 #ifdef INET
3089         case AF_INET:
3090                 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL,
3091                     &s->nat_src_node);
3092                 s->rt_kif = r->rpool.cur->kif;
3093                 break;
3094 #endif /* INET */
3095 #ifdef INET6
3096         case AF_INET6:
3097                 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL,
3098                     &s->nat_src_node);
3099                 s->rt_kif = r->rpool.cur->kif;
3100                 break;
3101 #endif /* INET6 */
3102         }
3103 }
3104
3105 u_int32_t
3106 pf_tcp_iss(struct pf_pdesc *pd)
3107 {
3108         MD5_CTX ctx;
3109         u_int32_t digest[4];
3110
3111         if (pf_tcp_secret_init == 0) {
3112                 karc4rand(pf_tcp_secret, sizeof(pf_tcp_secret));
3113                 MD5Init(&pf_tcp_secret_ctx);
3114                 MD5Update(&pf_tcp_secret_ctx, pf_tcp_secret,
3115                     sizeof(pf_tcp_secret));
3116                 pf_tcp_secret_init = 1;
3117         }
3118         ctx = pf_tcp_secret_ctx;
3119
3120         MD5Update(&ctx, (char *)&pd->hdr.tcp->th_sport, sizeof(u_short));
3121         MD5Update(&ctx, (char *)&pd->hdr.tcp->th_dport, sizeof(u_short));
3122         if (pd->af == AF_INET6) {
3123                 MD5Update(&ctx, (char *)&pd->src->v6, sizeof(struct in6_addr));
3124                 MD5Update(&ctx, (char *)&pd->dst->v6, sizeof(struct in6_addr));
3125         } else {
3126                 MD5Update(&ctx, (char *)&pd->src->v4, sizeof(struct in_addr));
3127                 MD5Update(&ctx, (char *)&pd->dst->v4, sizeof(struct in_addr));
3128         }
3129         MD5Final((u_char *)digest, &ctx);
3130         pf_tcp_iss_off += 4096;
3131         return (digest[0] + pd->hdr.tcp->th_seq + pf_tcp_iss_off);
3132 }
3133
3134 int
3135 pf_test_rule(struct pf_rule **rm, struct pf_state **sm, int direction,
3136     struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3137     struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
3138     struct ifqueue *ifq, struct inpcb *inp)
3139 {
3140         struct pf_rule          *nr = NULL;
3141         struct pf_addr          *saddr = pd->src, *daddr = pd->dst;
3142         sa_family_t              af = pd->af;
3143         struct pf_rule          *r, *a = NULL;
3144         struct pf_ruleset       *ruleset = NULL;
3145         struct pf_src_node      *nsn = NULL;
3146         struct tcphdr           *th = pd->hdr.tcp;
3147         struct pf_state_key     *skw = NULL, *sks = NULL;
3148         struct pf_state_key     *sk = NULL, *nk = NULL;
3149         u_short                  reason;
3150         int                      rewrite = 0, hdrlen = 0;
3151         int                      tag = -1, rtableid = -1;
3152         int                      asd = 0;
3153         int                      match = 0;
3154         int                      state_icmp = 0;
3155         u_int16_t                sport = 0, dport = 0;
3156         u_int16_t                nport = 0, bport = 0;
3157         u_int16_t                bproto_sum = 0, bip_sum = 0;
3158         u_int8_t                 icmptype = 0, icmpcode = 0;
3159
3160
3161         if (direction == PF_IN && pf_check_congestion(ifq)) {
3162                 REASON_SET(&reason, PFRES_CONGEST);
3163                 return (PF_DROP);
3164         }
3165
3166         if (inp != NULL)
3167                 pd->lookup.done = pf_socket_lookup(direction, pd);
3168         else if (debug_pfugidhack) {
3169                 DPFPRINTF(PF_DEBUG_MISC, ("pf: unlocked lookup\n"));
3170                 pd->lookup.done = pf_socket_lookup(direction, pd);
3171         }
3172
3173         switch (pd->proto) {
3174         case IPPROTO_TCP:
3175                 sport = th->th_sport;
3176                 dport = th->th_dport;
3177                 hdrlen = sizeof(*th);
3178                 break;
3179         case IPPROTO_UDP:
3180                 sport = pd->hdr.udp->uh_sport;
3181                 dport = pd->hdr.udp->uh_dport;
3182                 hdrlen = sizeof(*pd->hdr.udp);
3183                 break;
3184 #ifdef INET
3185         case IPPROTO_ICMP:
3186                 if (pd->af != AF_INET)
3187                         break;
3188                 sport = dport = pd->hdr.icmp->icmp_id;
3189                 hdrlen = sizeof(*pd->hdr.icmp);
3190                 icmptype = pd->hdr.icmp->icmp_type;
3191                 icmpcode = pd->hdr.icmp->icmp_code;
3192
3193                 if (icmptype == ICMP_UNREACH ||
3194                     icmptype == ICMP_SOURCEQUENCH ||
3195                     icmptype == ICMP_REDIRECT ||
3196                     icmptype == ICMP_TIMXCEED ||
3197                     icmptype == ICMP_PARAMPROB)
3198                         state_icmp++;
3199                 break;
3200 #endif /* INET */
3201 #ifdef INET6
3202         case IPPROTO_ICMPV6:
3203                 if (af != AF_INET6)
3204                         break;
3205                 sport = dport = pd->hdr.icmp6->icmp6_id;
3206                 hdrlen = sizeof(*pd->hdr.icmp6);
3207                 icmptype = pd->hdr.icmp6->icmp6_type;
3208                 icmpcode = pd->hdr.icmp6->icmp6_code;
3209
3210                 if (icmptype == ICMP6_DST_UNREACH ||
3211                     icmptype == ICMP6_PACKET_TOO_BIG ||
3212                     icmptype == ICMP6_TIME_EXCEEDED ||
3213                     icmptype == ICMP6_PARAM_PROB)
3214                         state_icmp++;
3215                 break;
3216 #endif /* INET6 */
3217         default:
3218                 sport = dport = hdrlen = 0;
3219                 break;
3220         }
3221
3222         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3223
3224         bport = nport = sport;
3225         /* check packet for BINAT/NAT/RDR */
3226         if ((nr = pf_get_translation(pd, m, off, direction, kif, &nsn,
3227             &skw, &sks, &sk, &nk, saddr, daddr, sport, dport)) != NULL) {
3228                 if (nk == NULL || sk == NULL) {
3229                         REASON_SET(&reason, PFRES_MEMORY);
3230                         goto cleanup;
3231                 }
3232
3233                 if (pd->ip_sum)
3234                         bip_sum = *pd->ip_sum;
3235
3236                 switch (pd->proto) {
3237                 case IPPROTO_TCP:
3238                         bproto_sum = th->th_sum;
3239                         pd->proto_sum = &th->th_sum;
3240
3241                         if (PF_ANEQ(saddr, &nk->addr[pd->sidx], af) ||
3242                             nk->port[pd->sidx] != sport) {
3243                                 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
3244                                     &th->th_sum, &nk->addr[pd->sidx],
3245                                     nk->port[pd->sidx], 0, af);
3246                                 pd->sport = &th->th_sport;
3247                                 sport = th->th_sport;
3248                         }
3249
3250                         if (PF_ANEQ(daddr, &nk->addr[pd->didx], af) ||
3251                             nk->port[pd->didx] != dport) {
3252                                 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
3253                                     &th->th_sum, &nk->addr[pd->didx],
3254                                     nk->port[pd->didx], 0, af);
3255                                 dport = th->th_dport;
3256                                 pd->dport = &th->th_dport;
3257                         }
3258                         rewrite++;
3259                         break;
3260                 case IPPROTO_UDP:
3261                         bproto_sum = pd->hdr.udp->uh_sum;
3262                         pd->proto_sum = &pd->hdr.udp->uh_sum;
3263
3264                         if (PF_ANEQ(saddr, &nk->addr[pd->sidx], af) ||
3265                             nk->port[pd->sidx] != sport) {
3266                                 pf_change_ap(saddr, &pd->hdr.udp->uh_sport,
3267                                     pd->ip_sum, &pd->hdr.udp->uh_sum,
3268                                     &nk->addr[pd->sidx],
3269                                     nk->port[pd->sidx], 1, af);
3270                                 sport = pd->hdr.udp->uh_sport;
3271                                 pd->sport = &pd->hdr.udp->uh_sport;
3272                         }
3273
3274                         if (PF_ANEQ(daddr, &nk->addr[pd->didx], af) ||
3275                             nk->port[pd->didx] != dport) {
3276                                 pf_change_ap(daddr, &pd->hdr.udp->uh_dport,
3277                                     pd->ip_sum, &pd->hdr.udp->uh_sum,
3278                                     &nk->addr[pd->didx],
3279                                     nk->port[pd->didx], 1, af);
3280                                 dport = pd->hdr.udp->uh_dport;
3281                                 pd->dport = &pd->hdr.udp->uh_dport;
3282                         }
3283                         rewrite++;
3284                         break;
3285 #ifdef INET
3286                 case IPPROTO_ICMP:
3287                         nk->port[0] = nk->port[1];
3288                         if (PF_ANEQ(saddr, &nk->addr[pd->sidx], AF_INET))
3289                                 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3290                                     nk->addr[pd->sidx].v4.s_addr, 0);
3291
3292                         if (PF_ANEQ(daddr, &nk->addr[pd->didx], AF_INET))
3293                                 pf_change_a(&daddr->v4.s_addr, pd->ip_sum,
3294                                     nk->addr[pd->didx].v4.s_addr, 0);
3295
3296                         if (nk->port[1] != pd->hdr.icmp->icmp_id) {
3297                                 pd->hdr.icmp->icmp_cksum = pf_cksum_fixup(
3298                                     pd->hdr.icmp->icmp_cksum, sport,
3299                                     nk->port[1], 0);
3300                                 pd->hdr.icmp->icmp_id = nk->port[1];
3301                                 pd->sport = &pd->hdr.icmp->icmp_id;
3302                         }
3303                         m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
3304                         break;
3305 #endif /* INET */
3306 #ifdef INET6
3307                 case IPPROTO_ICMPV6:
3308                         nk->port[0] = nk->port[1];
3309                         if (PF_ANEQ(saddr, &nk->addr[pd->sidx], AF_INET6))
3310                                 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum,
3311                                     &nk->addr[pd->sidx], 0);
3312
3313                         if (PF_ANEQ(daddr, &nk->addr[pd->didx], AF_INET6))
3314                                 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum,
3315                                     &nk->addr[pd->didx], 0);
3316                         rewrite++;
3317                         break;
3318 #endif /* INET */
3319                 default:
3320                         switch (af) {
3321 #ifdef INET
3322                         case AF_INET:
3323                                 if (PF_ANEQ(saddr,
3324                                     &nk->addr[pd->sidx], AF_INET))
3325                                         pf_change_a(&saddr->v4.s_addr,
3326                                             pd->ip_sum,
3327                                             nk->addr[pd->sidx].v4.s_addr, 0);
3328
3329                                 if (PF_ANEQ(daddr,
3330                                     &nk->addr[pd->didx], AF_INET))
3331                                         pf_change_a(&daddr->v4.s_addr,
3332                                             pd->ip_sum,
3333                                             nk->addr[pd->didx].v4.s_addr, 0);
3334                                 break;
3335 #endif /* INET */
3336 #ifdef INET6
3337                         case AF_INET6:
3338                                 if (PF_ANEQ(saddr,
3339                                     &nk->addr[pd->sidx], AF_INET6))
3340                                         PF_ACPY(saddr, &nk->addr[pd->sidx], af);
3341
3342                                 if (PF_ANEQ(daddr,
3343                                     &nk->addr[pd->didx], AF_INET6))
3344                                         PF_ACPY(saddr, &nk->addr[pd->didx], af);
3345                                 break;
3346 #endif /* INET */
3347                         }
3348                         break;
3349                 }
3350                 if (nr->natpass)
3351                         r = NULL;
3352                 pd->nat_rule = nr;
3353         }
3354
3355         while (r != NULL) {
3356                 r->evaluations++;
3357                 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3358                         r = r->skip[PF_SKIP_IFP].ptr;
3359                 else if (r->direction && r->direction != direction)
3360                         r = r->skip[PF_SKIP_DIR].ptr;
3361                 else if (r->af && r->af != af)
3362                         r = r->skip[PF_SKIP_AF].ptr;
3363                 else if (r->proto && r->proto != pd->proto)
3364                         r = r->skip[PF_SKIP_PROTO].ptr;
3365                 else if (PF_MISMATCHAW(&r->src.addr, saddr, af,
3366                     r->src.neg, kif))
3367                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3368                 /* tcp/udp only. port_op always 0 in other cases */
3369                 else if (r->src.port_op && !pf_match_port(r->src.port_op,
3370                     r->src.port[0], r->src.port[1], sport))
3371                         r = r->skip[PF_SKIP_SRC_PORT].ptr;
3372                 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af,
3373                     r->dst.neg, NULL))
3374                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
3375                 /* tcp/udp only. port_op always 0 in other cases */
3376                 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
3377                     r->dst.port[0], r->dst.port[1], dport))
3378                         r = r->skip[PF_SKIP_DST_PORT].ptr;
3379                 /* icmp only. type always 0 in other cases */
3380                 else if (r->type && r->type != icmptype + 1)
3381                         r = TAILQ_NEXT(r, entries);
3382                 /* icmp only. type always 0 in other cases */
3383                 else if (r->code && r->code != icmpcode + 1)
3384                         r = TAILQ_NEXT(r, entries);
3385                 else if (r->tos && !(r->tos == pd->tos))
3386                         r = TAILQ_NEXT(r, entries);
3387                 else if (r->rule_flag & PFRULE_FRAGMENT)
3388                         r = TAILQ_NEXT(r, entries);
3389                 else if (pd->proto == IPPROTO_TCP &&
3390                     (r->flagset & th->th_flags) != r->flags)
3391                         r = TAILQ_NEXT(r, entries);
3392                 /* tcp/udp only. uid.op always 0 in other cases */
3393                 else if (r->uid.op && (pd->lookup.done || (pd->lookup.done =
3394                     pf_socket_lookup(direction, pd), 1)) &&
3395                     !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
3396                     pd->lookup.uid))
3397                         r = TAILQ_NEXT(r, entries);
3398                 /* tcp/udp only. gid.op always 0 in other cases */
3399                 else if (r->gid.op && (pd->lookup.done || (pd->lookup.done =
3400                     pf_socket_lookup(direction, pd), 1)) &&
3401                     !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
3402                     pd->lookup.gid))
3403                         r = TAILQ_NEXT(r, entries);
3404                 else if (r->prob &&
3405                   r->prob <= karc4random())
3406                         r = TAILQ_NEXT(r, entries);
3407                 else if (r->match_tag && !pf_match_tag(m, r, &tag))
3408                         r = TAILQ_NEXT(r, entries);
3409                 else if (r->os_fingerprint != PF_OSFP_ANY &&
3410                     (pd->proto != IPPROTO_TCP || !pf_osfp_match(
3411                     pf_osfp_fingerprint(pd, m, off, th),
3412                     r->os_fingerprint)))
3413                         r = TAILQ_NEXT(r, entries);
3414                 else {
3415                         if (r->tag)
3416                                 tag = r->tag;
3417                         if (r->rtableid >= 0)
3418                                 rtableid = r->rtableid;
3419                         if (r->anchor == NULL) {
3420                                 match = 1;
3421                                 *rm = r;
3422                                 *am = a;
3423                                 *rsm = ruleset;
3424                                 if ((*rm)->quick)
3425                                         break;
3426                                 r = TAILQ_NEXT(r, entries);
3427                         } else
3428                                 pf_step_into_anchor(&asd, &ruleset,
3429                                     PF_RULESET_FILTER, &r, &a, &match);
3430                 }
3431                 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3432                     PF_RULESET_FILTER, &r, &a, &match))
3433                         break;
3434         }
3435         r = *rm;
3436         a = *am;
3437         ruleset = *rsm;
3438
3439         REASON_SET(&reason, PFRES_MATCH);
3440
3441         if (r->log || (nr != NULL && nr->log)) {
3442                 if (rewrite)
3443                         m_copyback(m, off, hdrlen, pd->hdr.any);
3444                 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr,
3445                     a, ruleset, pd);
3446         }
3447
3448         if ((r->action == PF_DROP) &&
3449             ((r->rule_flag & PFRULE_RETURNRST) ||
3450             (r->rule_flag & PFRULE_RETURNICMP) ||
3451             (r->rule_flag & PFRULE_RETURN))) {
3452                 /* undo NAT changes, if they have taken place */
3453                 if (nr != NULL) {
3454                         PF_ACPY(saddr, &sk->addr[pd->sidx], af);
3455                         PF_ACPY(daddr, &sk->addr[pd->didx], af);
3456                         if (pd->sport)
3457                                 *pd->sport = sk->port[pd->sidx];
3458                         if (pd->dport)
3459                                 *pd->dport = sk->port[pd->didx];
3460                         if (pd->proto_sum)
3461                                 *pd->proto_sum = bproto_sum;
3462                         if (pd->ip_sum)
3463                                 *pd->ip_sum = bip_sum;
3464                         m_copyback(m, off, hdrlen, pd->hdr.any);
3465                 }
3466                 if (pd->proto == IPPROTO_TCP &&
3467                     ((r->rule_flag & PFRULE_RETURNRST) ||
3468                     (r->rule_flag & PFRULE_RETURN)) &&
3469                     !(th->th_flags & TH_RST)) {
3470                         u_int32_t        ack = ntohl(th->th_seq) + pd->p_len;
3471                         int              len = 0;
3472                         struct ip       *h4;
3473                         struct ip6_hdr  *h6;
3474
3475                         switch (af) {
3476                         case AF_INET:
3477                                 h4 = mtod(m, struct ip *);
3478                                 len = h4->ip_len - off;
3479                                 break;
3480 #ifdef INET6
3481                         case AF_INET6:
3482                                 h6 = mtod(m, struct ip6_hdr *);
3483                                 len = h6->ip6_plen - (off - sizeof(*h6));
3484                                 break;
3485 #endif
3486                         }
3487
3488                         if (pf_check_proto_cksum(m, off, len, IPPROTO_TCP, af))
3489                                 REASON_SET(&reason, PFRES_PROTCKSUM);
3490                         else {
3491                                 if (th->th_flags & TH_SYN)
3492                                         ack++;
3493                                 if (th->th_flags & TH_FIN)
3494                                         ack++;
3495                                 pf_send_tcp(r, af, pd->dst,
3496                                     pd->src, th->th_dport, th->th_sport,
3497                                     ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0,
3498                                     r->return_ttl, 1, 0, pd->eh, kif->pfik_ifp);
3499                         }
3500                 } else if (pd->proto != IPPROTO_ICMP && af == AF_INET &&
3501                     r->return_icmp)
3502                         pf_send_icmp(m, r->return_icmp >> 8,
3503                             r->return_icmp & 255, af, r);
3504                 else if (pd->proto != IPPROTO_ICMPV6 && af == AF_INET6 &&
3505                     r->return_icmp6)
3506                         pf_send_icmp(m, r->return_icmp6 >> 8,
3507                             r->return_icmp6 & 255, af, r);
3508         }
3509
3510         if (r->action == PF_DROP)
3511                 goto cleanup;
3512
3513         if (pf_tag_packet(m, tag, rtableid)) {
3514                 REASON_SET(&reason, PFRES_MEMORY);
3515                 goto cleanup;
3516         }
3517
3518         if (!state_icmp && (r->keep_state || nr != NULL ||
3519             (pd->flags & PFDESC_TCP_NORM))) {
3520                 int action;
3521                 action = pf_create_state(r, nr, a, pd, nsn, skw, sks, nk, sk, m,
3522                     off, sport, dport, &rewrite, kif, sm, tag, bproto_sum,
3523                     bip_sum, hdrlen);
3524                 if (action != PF_PASS)
3525                         return (action);
3526         }
3527
3528         /* copy back packet headers if we performed NAT operations */
3529         if (rewrite)
3530                 m_copyback(m, off, hdrlen, pd->hdr.any);
3531
3532         return (PF_PASS);
3533
3534 cleanup:
3535         if (sk != NULL)
3536                 pool_put(&pf_state_key_pl, sk);
3537         if (nk != NULL)
3538                 pool_put(&pf_state_key_pl, nk);
3539         return (PF_DROP);
3540 }
3541
3542 static __inline int
3543 pf_create_state(struct pf_rule *r, struct pf_rule *nr, struct pf_rule *a,
3544     struct pf_pdesc *pd, struct pf_src_node *nsn, struct pf_state_key *skw,
3545     struct pf_state_key *sks, struct pf_state_key *nk, struct pf_state_key *sk,
3546     struct mbuf *m, int off, u_int16_t sport, u_int16_t dport, int *rewrite,
3547     struct pfi_kif *kif, struct pf_state **sm, int tag, u_int16_t bproto_sum,
3548     u_int16_t bip_sum, int hdrlen)
3549 {
3550         struct pf_state         *s = NULL;
3551         struct pf_src_node      *sn = NULL;
3552         struct tcphdr           *th = pd->hdr.tcp;
3553         u_int16_t                mss = tcp_mssdflt;
3554         u_short                  reason;
3555
3556         /* check maximums */
3557         if (r->max_states && (r->states_cur >= r->max_states)) {
3558                 pf_status.lcounters[LCNT_STATES]++;
3559                 REASON_SET(&reason, PFRES_MAXSTATES);
3560                 return (PF_DROP);
3561         }
3562         /* src node for filter rule */
3563         if ((r->rule_flag & PFRULE_SRCTRACK ||
3564             r->rpool.opts & PF_POOL_STICKYADDR) &&
3565             pf_insert_src_node(&sn, r, pd->src, pd->af) != 0) {
3566                 REASON_SET(&reason, PFRES_SRCLIMIT);
3567                 goto csfailed;
3568         }
3569         /* src node for translation rule */
3570         if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3571             pf_insert_src_node(&nsn, nr, &sk->addr[pd->sidx], pd->af)) {
3572                 REASON_SET(&reason, PFRES_SRCLIMIT);
3573                 goto csfailed;
3574         }
3575         s = pool_get(&pf_state_pl, PR_NOWAIT | PR_ZERO);
3576         if (s == NULL) {
3577                 REASON_SET(&reason, PFRES_MEMORY);
3578                 goto csfailed;
3579         }
3580         s->id = 0; /* XXX Do we really need that? not in OpenBSD */
3581         s->creatorid = 0;
3582         s->rule.ptr = r;
3583         s->nat_rule.ptr = nr;
3584         s->anchor.ptr = a;
3585         STATE_INC_COUNTERS(s);
3586         if (r->allow_opts)
3587                 s->state_flags |= PFSTATE_ALLOWOPTS;
3588         if (r->rule_flag & PFRULE_STATESLOPPY)
3589                 s->state_flags |= PFSTATE_SLOPPY;
3590         s->log = r->log & PF_LOG_ALL;
3591         if (nr != NULL)
3592                 s->log |= nr->log & PF_LOG_ALL;
3593         switch (pd->proto) {
3594         case IPPROTO_TCP:
3595                 s->src.seqlo = ntohl(th->th_seq);
3596                 s->src.seqhi = s->src.seqlo + pd->p_len + 1;
3597                 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
3598                     r->keep_state == PF_STATE_MODULATE) {
3599                         /* Generate sequence number modulator */
3600                         if ((s->src.seqdiff = pf_tcp_iss(pd) - s->src.seqlo) ==
3601                             0)
3602                                 s->src.seqdiff = 1;
3603                         pf_change_a(&th->th_seq, &th->th_sum,
3604                             htonl(s->src.seqlo + s->src.seqdiff), 0);
3605                         *rewrite = 1;
3606                 } else
3607                         s->src.seqdiff = 0;
3608                 if (th->th_flags & TH_SYN) {
3609                         s->src.seqhi++;
3610                         s->src.wscale = pf_get_wscale(m, off,
3611                             th->th_off, pd->af);
3612                 }
3613                 s->src.max_win = MAX(ntohs(th->th_win), 1);
3614                 if (s->src.wscale & PF_WSCALE_MASK) {
3615                         /* Remove scale factor from initial window */
3616                         int win = s->src.max_win;
3617                         win += 1 << (s->src.wscale & PF_WSCALE_MASK);
3618                         s->src.max_win = (win - 1) >>
3619                             (s->src.wscale & PF_WSCALE_MASK);
3620                 }
3621                 if (th->th_flags & TH_FIN)
3622                         s->src.seqhi++;
3623                 s->dst.seqhi = 1;
3624                 s->dst.max_win = 1;
3625                 s->src.state = TCPS_SYN_SENT;
3626                 s->dst.state = TCPS_CLOSED;
3627                 s->timeout = PFTM_TCP_FIRST_PACKET;
3628                 break;
3629         case IPPROTO_UDP:
3630                 s->src.state = PFUDPS_SINGLE;
3631                 s->dst.state = PFUDPS_NO_TRAFFIC;
3632                 s->timeout = PFTM_UDP_FIRST_PACKET;
3633                 break;
3634         case IPPROTO_ICMP:
3635 #ifdef INET6
3636         case IPPROTO_ICMPV6:
3637 #endif
3638                 s->timeout = PFTM_ICMP_FIRST_PACKET;
3639                 break;
3640         default:
3641                 s->src.state = PFOTHERS_SINGLE;
3642                 s->dst.state = PFOTHERS_NO_TRAFFIC;
3643                 s->timeout = PFTM_OTHER_FIRST_PACKET;
3644         }
3645
3646         s->creation = time_second;
3647         s->expire = time_second;
3648
3649         if (sn != NULL) {
3650                 s->src_node = sn;
3651                 s->src_node->states++;
3652         }
3653         if (nsn != NULL) {
3654                 /* XXX We only modify one side for now. */
3655                 PF_ACPY(&nsn->raddr, &nk->addr[1], pd->af);
3656                 s->nat_src_node = nsn;
3657                 s->nat_src_node->states++;
3658         }
3659         if (pd->proto == IPPROTO_TCP) {
3660                 if ((pd->flags & PFDESC_TCP_NORM) && pf_normalize_tcp_init(m,
3661                     off, pd, th, &s->src, &s->dst)) {
3662                         REASON_SET(&reason, PFRES_MEMORY);
3663                         pf_src_tree_remove_state(s);
3664                         STATE_DEC_COUNTERS(s);
3665                         pool_put(&pf_state_pl, s);
3666                         return (PF_DROP);
3667                 }
3668                 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
3669                     pf_normalize_tcp_stateful(m, off, pd, &reason, th, s,
3670                     &s->src, &s->dst, rewrite)) {
3671                         /* This really shouldn't happen!!! */
3672                         DPFPRINTF(PF_DEBUG_URGENT,
3673                             ("pf_normalize_tcp_stateful failed on first pkt"));
3674                         pf_normalize_tcp_cleanup(s);
3675                         pf_src_tree_remove_state(s);
3676                         STATE_DEC_COUNTERS(s);
3677                         pool_put(&pf_state_pl, s);
3678                         return (PF_DROP);
3679                 }
3680         }
3681         s->direction = pd->dir;
3682
3683         if (sk == NULL && pf_state_key_setup(pd, nr, &skw, &sks, &sk, &nk,
3684             pd->src, pd->dst, sport, dport))
3685                 goto csfailed;
3686
3687         if (pf_state_insert(BOUND_IFACE(r, kif), skw, sks, s)) {
3688                 if (pd->proto == IPPROTO_TCP)
3689                         pf_normalize_tcp_cleanup(s);
3690                 REASON_SET(&reason, PFRES_STATEINS);
3691                 pf_src_tree_remove_state(s);
3692                 STATE_DEC_COUNTERS(s);
3693                 pool_put(&pf_state_pl, s);
3694                 return (PF_DROP);
3695         } else
3696                 *sm = s;
3697
3698         pf_set_rt_ifp(s, pd->src);      /* needs s->state_key set */
3699         if (tag > 0) {
3700                 pf_tag_ref(tag);
3701                 s->tag = tag;
3702         }
3703         if (pd->proto == IPPROTO_TCP && (th->th_flags & (TH_SYN|TH_ACK)) ==
3704             TH_SYN && r->keep_state == PF_STATE_SYNPROXY) {
3705                 s->src.state = PF_TCPS_PROXY_SRC;
3706                 /* undo NAT changes, if they have taken place */
3707                 if (nr != NULL) {
3708                         struct pf_state_key *skt = s->key[PF_SK_WIRE];
3709                         if (pd->dir == PF_OUT)
3710                                 skt = s->key[PF_SK_STACK];
3711                         PF_ACPY(pd->src, &skt->addr[pd->sidx], pd->af);
3712                         PF_ACPY(pd->dst, &skt->addr[pd->didx], pd->af);
3713                         if (pd->sport)
3714                                 *pd->sport = skt->port[pd->sidx];
3715                         if (pd->dport)
3716                                 *pd->dport = skt->port[pd->didx];
3717                         if (pd->proto_sum)
3718                                 *pd->proto_sum = bproto_sum;
3719                         if (pd->ip_sum)
3720                                 *pd->ip_sum = bip_sum;
3721                         m_copyback(m, off, hdrlen, pd->hdr.any);
3722                 }
3723                 s->src.seqhi = htonl(karc4random());
3724                 /* Find mss option */
3725                 mss = pf_get_mss(m, off, th->th_off, pd->af);
3726                 mss = pf_calc_mss(pd->src, pd->af, mss);
3727                 mss = pf_calc_mss(pd->dst, pd->af, mss);
3728                 s->src.mss = mss;
3729                 pf_send_tcp(r, pd->af, pd->dst, pd->src, th->th_dport,
3730                     th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1,
3731                     TH_SYN|TH_ACK, 0, s->src.mss, 0, 1, 0, NULL, NULL);
3732                 REASON_SET(&reason, PFRES_SYNPROXY);
3733                 return (PF_SYNPROXY_DROP);
3734         }
3735
3736         return (PF_PASS);
3737
3738 csfailed:
3739         if (sk != NULL)
3740                 pool_put(&pf_state_key_pl, sk);
3741         if (nk != NULL)
3742                 pool_put(&pf_state_key_pl, nk);
3743
3744         if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3745                 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3746                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3747                 pf_status.src_nodes--;
3748                 pool_put(&pf_src_tree_pl, sn);
3749         }
3750         if (nsn != sn && nsn != NULL && nsn->states == 0 && nsn->expire == 0) {
3751                 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3752                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3753                 pf_status.src_nodes--;
3754                 pool_put(&pf_src_tree_pl, nsn);
3755         }
3756         return (PF_DROP);
3757 }
3758
3759 int
3760 pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif,
3761     struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am,
3762     struct pf_ruleset **rsm)
3763 {
3764         struct pf_rule          *r, *a = NULL;
3765         struct pf_ruleset       *ruleset = NULL;
3766         sa_family_t              af = pd->af;
3767         u_short                  reason;
3768         int                      tag = -1;
3769         int                      asd = 0;
3770         int                      match = 0;
3771
3772         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3773         while (r != NULL) {
3774                 r->evaluations++;
3775                 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3776                         r = r->skip[PF_SKIP_IFP].ptr;
3777                 else if (r->direction && r->direction != direction)
3778                         r = r->skip[PF_SKIP_DIR].ptr;
3779                 else if (r->af && r->af != af)
3780                         r = r->skip[PF_SKIP_AF].ptr;
3781                 else if (r->proto && r->proto != pd->proto)
3782                         r = r->skip[PF_SKIP_PROTO].ptr;
3783                 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af,
3784                     r->src.neg, kif))
3785                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3786                 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af,
3787                     r->dst.neg, NULL))
3788                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
3789                 else if (r->tos && !(r->tos == pd->tos))
3790                         r = TAILQ_NEXT(r, entries);
3791                 else if (r->os_fingerprint != PF_OSFP_ANY)
3792                         r = TAILQ_NEXT(r, entries);
3793                 else if (pd->proto == IPPROTO_UDP &&
3794                     (r->src.port_op || r->dst.port_op))
3795                         r = TAILQ_NEXT(r, entries);
3796                 else if (pd->proto == IPPROTO_TCP &&
3797                     (r->src.port_op || r->dst.port_op || r->flagset))
3798                         r = TAILQ_NEXT(r, entries);
3799                 else if ((pd->proto == IPPROTO_ICMP ||
3800                     pd->proto == IPPROTO_ICMPV6) &&
3801                     (r->type || r->code))
3802                         r = TAILQ_NEXT(r, entries);
3803                 else if (r->prob && r->prob <= karc4random())
3804                         r = TAILQ_NEXT(r, entries);
3805                 else if (r->match_tag && !pf_match_tag(m, r, &tag))
3806                         r = TAILQ_NEXT(r, entries);
3807                 else {
3808                         if (r->anchor == NULL) {
3809                                 match = 1;
3810                                 *rm = r;
3811                                 *am = a;
3812                                 *rsm = ruleset;
3813                                 if ((*rm)->quick)
3814                                         break;
3815                                 r = TAILQ_NEXT(r, entries);
3816                         } else
3817                                 pf_step_into_anchor(&asd, &ruleset,
3818                                     PF_RULESET_FILTER, &r, &a, &match);
3819                 }
3820                 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3821                     PF_RULESET_FILTER, &r, &a, &match))
3822                         break;
3823         }
3824         r = *rm;
3825         a = *am;
3826         ruleset = *rsm;
3827
3828         REASON_SET(&reason, PFRES_MATCH);
3829
3830         if (r->log)
3831                 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset,
3832                     pd);
3833
3834         if (r->action != PF_PASS)
3835                 return (PF_DROP);
3836
3837         if (pf_tag_packet(m, tag, -1)) {
3838                 REASON_SET(&reason, PFRES_MEMORY);
3839                 return (PF_DROP);
3840         }
3841
3842         return (PF_PASS);
3843 }
3844
3845 int
3846 pf_tcp_track_full(struct pf_state_peer *src, struct pf_state_peer *dst,
3847         struct pf_state **state, struct pfi_kif *kif, struct mbuf *m, int off,
3848         struct pf_pdesc *pd, u_short *reason, int *copyback)
3849 {
3850         struct tcphdr           *th = pd->hdr.tcp;
3851         u_int16_t                win = ntohs(th->th_win);
3852         u_int32_t                ack, end, seq, orig_seq;
3853         u_int8_t                 sws, dws;
3854         int                      ackskew;
3855
3856         if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) {
3857                 sws = src->wscale & PF_WSCALE_MASK;
3858                 dws = dst->wscale & PF_WSCALE_MASK;
3859         } else
3860                 sws = dws = 0;
3861
3862         /*
3863          * Sequence tracking algorithm from Guido van Rooij's paper:
3864          *   http://www.madison-gurkha.com/publications/tcp_filtering/
3865          *      tcp_filtering.ps
3866          */
3867
3868         orig_seq = seq = ntohl(th->th_seq);
3869         if (src->seqlo == 0) {
3870                 /* First packet from this end. Set its state */
3871
3872                 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) &&
3873                     src->scrub == NULL) {
3874                         if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) {
3875                                 REASON_SET(reason, PFRES_MEMORY);
3876                                 return (PF_DROP);
3877                         }
3878                 }
3879
3880                 /* Deferred generation of sequence number modulator */
3881                 if (dst->seqdiff && !src->seqdiff) {
3882                         /* use random iss for the TCP server */
3883                         while ((src->seqdiff = karc4random() - seq) == 0)
3884                                 ;
3885                         ack = ntohl(th->th_ack) - dst->seqdiff;
3886                         pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3887                             src->seqdiff), 0);
3888                         pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3889                         *copyback = 1;
3890                 } else {
3891                         ack = ntohl(th->th_ack);
3892                 }
3893
3894                 end = seq + pd->p_len;
3895                 if (th->th_flags & TH_SYN) {
3896                         end++;
3897                         (*state)->sync_flags |= PFSTATE_GOT_SYN2;
3898                         if (dst->wscale & PF_WSCALE_FLAG) {
3899                                 src->wscale = pf_get_wscale(m, off, th->th_off,
3900                                     pd->af);
3901                                 if (src->wscale & PF_WSCALE_FLAG) {
3902                                         /* Remove scale factor from initial
3903                                          * window */
3904                                         sws = src->wscale & PF_WSCALE_MASK;
3905                                         win = ((u_int32_t)win + (1 << sws) - 1)
3906                                             >> sws;
3907                                         dws = dst->wscale & PF_WSCALE_MASK;
3908                                 } else {
3909                                         /* fixup other window */
3910                                         dst->max_win <<= dst->wscale &
3911                                             PF_WSCALE_MASK;
3912                                         /* in case of a retrans SYN|ACK */
3913                                         dst->wscale = 0;
3914                                 }
3915                         }
3916                 }
3917                 if (th->th_flags & TH_FIN)
3918                         end++;
3919
3920                 src->seqlo = seq;
3921                 if (src->state < TCPS_SYN_SENT)
3922                         src->state = TCPS_SYN_SENT;
3923
3924                 /*
3925                  * May need to slide the window (seqhi may have been set by
3926                  * the crappy stack check or if we picked up the connection
3927                  * after establishment)
3928                  */
3929                 if (src->seqhi == 1 ||
3930                     SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi))
3931                         src->seqhi = end + MAX(1, dst->max_win << dws);
3932                 if (win > src->max_win)
3933                         src->max_win = win;
3934
3935         } else {
3936                 ack = ntohl(th->th_ack) - dst->seqdiff;
3937                 if (src->seqdiff) {
3938                         /* Modulate sequence numbers */
3939                         pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3940                             src->seqdiff), 0);
3941                         pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3942                         *copyback = 1;
3943                 }
3944                 end = seq + pd->p_len;
3945                 if (th->th_flags & TH_SYN)
3946                         end++;
3947                 if (th->th_flags & TH_FIN)
3948                         end++;
3949         }
3950
3951         if ((th->th_flags & TH_ACK) == 0) {
3952                 /* Let it pass through the ack skew check */
3953                 ack = dst->seqlo;
3954         } else if ((ack == 0 &&
3955             (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) ||
3956             /* broken tcp stacks do not set ack */
3957             (dst->state < TCPS_SYN_SENT)) {
3958                 /*
3959                  * Many stacks (ours included) will set the ACK number in an
3960                  * FIN|ACK if the SYN times out -- no sequence to ACK.
3961                  */
3962                 ack = dst->seqlo;
3963         }
3964
3965         if (seq == end) {
3966                 /* Ease sequencing restrictions on no data packets */
3967                 seq = src->seqlo;
3968                 end = seq;
3969         }
3970
3971         ackskew = dst->seqlo - ack;
3972
3973
3974         /*
3975          * Need to demodulate the sequence numbers in any TCP SACK options
3976          * (Selective ACK). We could optionally validate the SACK values
3977          * against the current ACK window, either forwards or backwards, but
3978          * I'm not confident that SACK has been implemented properly
3979          * everywhere. It wouldn't surprise me if several stacks accidently
3980          * SACK too far backwards of previously ACKed data. There really aren't
3981          * any security implications of bad SACKing unless the target stack
3982          * doesn't validate the option length correctly. Someone trying to
3983          * spoof into a TCP connection won't bother blindly sending SACK
3984          * options anyway.
3985          */
3986         if (dst->seqdiff && (th->th_off << 2) > sizeof(struct tcphdr)) {
3987                 if (pf_modulate_sack(m, off, pd, th, dst))
3988                         *copyback = 1;
3989         }
3990
3991
3992 #define MAXACKWINDOW (0xffff + 1500)    /* 1500 is an arbitrary fudge factor */
3993         if (SEQ_GEQ(src->seqhi, end) &&
3994             /* Last octet inside other's window space */
3995             SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) &&
3996             /* Retrans: not more than one window back */
3997             (ackskew >= -MAXACKWINDOW) &&
3998             /* Acking not more than one reassembled fragment backwards */
3999             (ackskew <= (MAXACKWINDOW << sws)) &&
4000             /* Acking not more than one window forward */
4001             ((th->th_flags & TH_RST) == 0 || orig_seq == src->seqlo ||
4002             (orig_seq == src->seqlo + 1) || (orig_seq + 1 == src->seqlo) ||
4003             (pd->flags & PFDESC_IP_REAS) == 0)) {
4004             /* Require an exact/+1 sequence match on resets when possible */
4005
4006                 if (dst->scrub || src->scrub) {
4007                         if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4008                             *state, src, dst, copyback))
4009                                 return (PF_DROP);
4010                 }
4011
4012                 /* update max window */
4013                 if (src->max_win < win)
4014                         src->max_win = win;
4015                 /* synchronize sequencing */
4016                 if (SEQ_GT(end, src->seqlo))
4017                         src->seqlo = end;
4018                 /* slide the window of what the other end can send */
4019                 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4020                         dst->seqhi = ack + MAX((win << sws), 1);
4021
4022
4023                 /* update states */
4024                 if (th->th_flags & TH_SYN)
4025                         if (src->state < TCPS_SYN_SENT)
4026                                 src->state = TCPS_SYN_SENT;
4027                 if (th->th_flags & TH_FIN)
4028                         if (src->state < TCPS_CLOSING)
4029                                 src->state = TCPS_CLOSING;
4030                 if (th->th_flags & TH_ACK) {
4031                         if (dst->state == TCPS_SYN_SENT) {
4032                                 dst->state = TCPS_ESTABLISHED;
4033                                 if (src->state == TCPS_ESTABLISHED &&
4034                                     (*state)->src_node != NULL &&
4035                                     pf_src_connlimit(state)) {
4036                                         REASON_SET(reason, PFRES_SRCLIMIT);
4037                                         return (PF_DROP);
4038                                 }
4039                         } else if (dst->state == TCPS_CLOSING)
4040                                 dst->state = TCPS_FIN_WAIT_2;
4041                 }
4042                 if (th->th_flags & TH_RST)
4043                         src->state = dst->state = TCPS_TIME_WAIT;
4044
4045                 /* update expire time */
4046                 (*state)->expire = time_second;
4047                 if (src->state >= TCPS_FIN_WAIT_2 &&
4048                     dst->state >= TCPS_FIN_WAIT_2)
4049                         (*state)->timeout = PFTM_TCP_CLOSED;
4050                 else if (src->state >= TCPS_CLOSING &&
4051                     dst->state >= TCPS_CLOSING)
4052                         (*state)->timeout = PFTM_TCP_FIN_WAIT;
4053                 else if (src->state < TCPS_ESTABLISHED ||
4054                     dst->state < TCPS_ESTABLISHED)
4055                         (*state)->timeout = PFTM_TCP_OPENING;
4056                 else if (src->state >= TCPS_CLOSING ||
4057                     dst->state >= TCPS_CLOSING)
4058                         (*state)->timeout = PFTM_TCP_CLOSING;
4059                 else
4060                         (*state)->timeout = PFTM_TCP_ESTABLISHED;
4061
4062                 /* Fall through to PASS packet */
4063
4064         } else if ((dst->state < TCPS_SYN_SENT ||
4065                 dst->state >= TCPS_FIN_WAIT_2 ||
4066                 src->state >= TCPS_FIN_WAIT_2) &&
4067             SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) &&
4068             /* Within a window forward of the originating packet */
4069             SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) {
4070             /* Within a window backward of the originating packet */
4071
4072                 /*
4073                  * This currently handles three situations:
4074                  *  1) Stupid stacks will shotgun SYNs before their peer
4075                  *     replies.
4076                  *  2) When PF catches an already established stream (the
4077                  *     firewall rebooted, the state table was flushed, routes
4078                  *     changed...)
4079                  *  3) Packets get funky immediately after the connection
4080                  *     closes (this should catch Solaris spurious ACK|FINs
4081                  *     that web servers like to spew after a close)
4082                  *
4083                  * This must be a little more careful than the above code
4084                  * since packet floods will also be caught here. We don't
4085                  * update the TTL here to mitigate the damage of a packet
4086                  * flood and so the same code can handle awkward establishment
4087                  * and a loosened connection close.
4088                  * In the establishment case, a correct peer response will
4089                  * validate the connection, go through the normal state code
4090                  * and keep updating the state TTL.
4091                  */
4092
4093                 if (pf_status.debug >= PF_DEBUG_MISC) {
4094                         kprintf("pf: loose state match: ");
4095                         pf_print_state(*state);
4096                         pf_print_flags(th->th_flags);
4097                         kprintf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
4098                             "pkts=%llu:%llu dir=%s,%s\n", seq, orig_seq, ack, pd->p_len,
4099                             ackskew, (unsigned long long)(*state)->packets[0],
4100                             (unsigned long long)(*state)->packets[1],
4101                             pd->dir == PF_IN ? "in" : "out",
4102                             pd->dir == (*state)->direction ? "fwd" : "rev");
4103                 }
4104
4105                 if (dst->scrub || src->scrub) {
4106                         if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4107                             *state, src, dst, copyback))
4108                                 return (PF_DROP);
4109                 }
4110
4111                 /* update max window */
4112                 if (src->max_win < win)
4113                         src->max_win = win;
4114                 /* synchronize sequencing */
4115                 if (SEQ_GT(end, src->seqlo))
4116                         src->seqlo = end;
4117                 /* slide the window of what the other end can send */
4118                 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4119                         dst->seqhi = ack + MAX((win << sws), 1);
4120
4121                 /*
4122                  * Cannot set dst->seqhi here since this could be a shotgunned
4123                  * SYN and not an already established connection.
4124                  */
4125
4126                 if (th->th_flags & TH_FIN)
4127                         if (src->state < TCPS_CLOSING)
4128                                 src->state = TCPS_CLOSING;
4129                 if (th->th_flags & TH_RST)
4130                         src->state = dst->state = TCPS_TIME_WAIT;
4131
4132                 /* Fall through to PASS packet */
4133
4134         } else if ((*state)->pickup_mode == PF_PICKUPS_HASHONLY ||
4135                     ((*state)->pickup_mode == PF_PICKUPS_ENABLED &&
4136                      ((*state)->sync_flags & PFSTATE_GOT_SYN_MASK) !=
4137                       PFSTATE_GOT_SYN_MASK)) {
4138                 /*
4139                  * If pickup mode is hash only, do not fail on sequence checks.
4140                  *
4141                  * If pickup mode is enabled and we did not see the SYN in
4142                  * both direction, do not fail on sequence checks because
4143                  * we do not have complete information on window scale.
4144                  *
4145                  * Adjust expiration and fall through to PASS packet.
4146                  * XXX Add a FIN check to reduce timeout?
4147                  */
4148                 (*state)->expire = time_second;
4149         } else  {
4150                 /*
4151                  * Failure processing
4152                  */
4153                 if ((*state)->dst.state == TCPS_SYN_SENT &&
4154                     (*state)->src.state == TCPS_SYN_SENT) {
4155                         /* Send RST for state mismatches during handshake */
4156                         if (!(th->th_flags & TH_RST))
4157                                 pf_send_tcp((*state)->rule.ptr, pd->af,
4158                                     pd->dst, pd->src, th->th_dport,
4159                                     th->th_sport, ntohl(th->th_ack), 0,
4160                                     TH_RST, 0, 0,
4161                                     (*state)->rule.ptr->return_ttl, 1, 0,
4162                                     pd->eh, kif->pfik_ifp);
4163                         src->seqlo = 0;
4164                         src->seqhi = 1;
4165                         src->max_win = 1;
4166                 } else if (pf_status.debug >= PF_DEBUG_MISC) {
4167                         kprintf("pf: BAD state: ");
4168                         pf_print_state(*state);
4169                         pf_print_flags(th->th_flags);
4170                         kprintf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
4171                             "pkts=%llu:%llu dir=%s,%s\n",
4172                             seq, orig_seq, ack, pd->p_len, ackskew,
4173                             (unsigned long long)(*state)->packets[0],
4174                                 (unsigned long long)(*state)->packets[1],
4175                             pd->dir == PF_IN ? "in" : "out",
4176                             pd->dir == (*state)->direction ? "fwd" : "rev");
4177                         kprintf("pf: State failure on: %c %c %c %c | %c %c\n",
4178                             SEQ_GEQ(src->seqhi, end) ? ' ' : '1',
4179                             SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ?
4180                             ' ': '2',
4181                             (ackskew >= -MAXACKWINDOW) ? ' ' : '3',
4182                             (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4',
4183                             SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5',
4184                             SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6');
4185                 }
4186                 REASON_SET(reason, PFRES_BADSTATE);
4187                 return (PF_DROP);
4188         }
4189
4190         return (PF_PASS);
4191 }
4192
4193 int
4194 pf_tcp_track_sloppy(struct pf_state_peer *src, struct pf_state_peer *dst,
4195         struct pf_state **state, struct pf_pdesc *pd, u_short *reason)
4196 {
4197         struct tcphdr           *th = pd->hdr.tcp;
4198
4199         if (th->th_flags & TH_SYN)
4200                 if (src->state < TCPS_SYN_SENT)
4201                         src->state = TCPS_SYN_SENT;
4202         if (th->th_flags & TH_FIN)
4203                 if (src->state < TCPS_CLOSING)
4204                         src->state = TCPS_CLOSING;
4205         if (th->th_flags & TH_ACK) {
4206                 if (dst->state == TCPS_SYN_SENT) {
4207                         dst->state = TCPS_ESTABLISHED;
4208                         if (src->state == TCPS_ESTABLISHED &&
4209                             (*state)->src_node != NULL &&
4210                             pf_src_connlimit(state)) {
4211                                 REASON_SET(reason, PFRES_SRCLIMIT);
4212                                 return (PF_DROP);
4213                         }
4214                 } else if (dst->state == TCPS_CLOSING) {
4215                         dst->state = TCPS_FIN_WAIT_2;
4216                 } else if (src->state == TCPS_SYN_SENT &&
4217                     dst->state < TCPS_SYN_SENT) {
4218                         /*
4219                          * Handle a special sloppy case where we only see one
4220                          * half of the connection. If there is a ACK after
4221                          * the initial SYN without ever seeing a packet from
4222                          * the destination, set the connection to established.
4223                          */
4224                         dst->state = src->state = TCPS_ESTABLISHED;
4225                         if ((*state)->src_node != NULL &&
4226                             pf_src_connlimit(state)) {
4227                                 REASON_SET(reason, PFRES_SRCLIMIT);
4228                                 return (PF_DROP);
4229                         }
4230                 } else if (src->state == TCPS_CLOSING &&
4231                     dst->state == TCPS_ESTABLISHED &&
4232                     dst->seqlo == 0) {
4233                         /*
4234                          * Handle the closing of half connections where we
4235                          * don't see the full bidirectional FIN/ACK+ACK
4236                          * handshake.
4237                          */
4238                         dst->state = TCPS_CLOSING;
4239                 }
4240         }
4241         if (th->th_flags & TH_RST)
4242                 src->state = dst->state = TCPS_TIME_WAIT;
4243
4244         /* update expire time */
4245         (*state)->expire = time_second;
4246         if (src->state >= TCPS_FIN_WAIT_2 &&
4247             dst->state >= TCPS_FIN_WAIT_2)
4248                 (*state)->timeout = PFTM_TCP_CLOSED;
4249         else if (src->state >= TCPS_CLOSING &&
4250             dst->state >= TCPS_CLOSING)
4251                 (*state)->timeout = PFTM_TCP_FIN_WAIT;
4252         else if (src->state < TCPS_ESTABLISHED ||
4253             dst->state < TCPS_ESTABLISHED)
4254                 (*state)->timeout = PFTM_TCP_OPENING;
4255         else if (src->state >= TCPS_CLOSING ||
4256             dst->state >= TCPS_CLOSING)
4257                 (*state)->timeout = PFTM_TCP_CLOSING;
4258         else
4259                 (*state)->timeout = PFTM_TCP_ESTABLISHED;
4260
4261         return (PF_PASS);
4262 }
4263
4264 int
4265 pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
4266     struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
4267     u_short *reason)
4268 {
4269         struct pf_state_key_cmp  key;
4270         struct tcphdr           *th = pd->hdr.tcp;
4271         int                      copyback = 0;
4272         struct pf_state_peer    *src, *dst;
4273         struct pf_state_key     *sk;
4274
4275         key.af = pd->af;
4276         key.proto = IPPROTO_TCP;
4277         if (direction == PF_IN) {       /* wire side, straight */
4278                 PF_ACPY(&key.addr[0], pd->src, key.af);
4279                 PF_ACPY(&key.addr[1], pd->dst, key.af);
4280                 key.port[0] = th->th_sport;
4281                 key.port[1] = th->th_dport;
4282         } else {                        /* stack side, reverse */
4283                 PF_ACPY(&key.addr[1], pd->src, key.af);
4284                 PF_ACPY(&key.addr[0], pd->dst, key.af);
4285                 key.port[1] = th->th_sport;
4286                 key.port[0] = th->th_dport;
4287         }
4288
4289         STATE_LOOKUP(kif, &key, direction, *state, m);
4290
4291         if (direction == (*state)->direction) {
4292                 src = &(*state)->src;
4293                 dst = &(*state)->dst;
4294         } else {
4295                 src = &(*state)->dst;
4296                 dst = &(*state)->src;
4297         }
4298
4299         sk = (*state)->key[pd->didx];
4300
4301         if ((*state)->src.state == PF_TCPS_PROXY_SRC) {
4302                 if (direction != (*state)->direction) {
4303                         REASON_SET(reason, PFRES_SYNPROXY);
4304                         return (PF_SYNPROXY_DROP);
4305                 }
4306                 if (th->th_flags & TH_SYN) {
4307                         if (ntohl(th->th_seq) != (*state)->src.seqlo) {
4308                                 REASON_SET(reason, PFRES_SYNPROXY);
4309                                 return (PF_DROP);
4310                         }
4311                         pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
4312                             pd->src, th->th_dport, th->th_sport,
4313                             (*state)->src.seqhi, ntohl(th->th_seq) + 1,
4314                             TH_SYN|TH_ACK, 0, (*state)->src.mss, 0, 1,
4315                             0, NULL, NULL);
4316                         REASON_SET(reason, PFRES_SYNPROXY);
4317                         return (PF_SYNPROXY_DROP);
4318                 } else if (!(th->th_flags & TH_ACK) ||
4319                     (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
4320                     (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
4321                         REASON_SET(reason, PFRES_SYNPROXY);
4322                         return (PF_DROP);
4323                 } else if ((*state)->src_node != NULL &&
4324                     pf_src_connlimit(state)) {
4325                         REASON_SET(reason, PFRES_SRCLIMIT);
4326                         return (PF_DROP);
4327                 } else
4328                         (*state)->src.state = PF_TCPS_PROXY_DST;
4329         }
4330         if ((*state)->src.state == PF_TCPS_PROXY_DST) {
4331                 if (direction == (*state)->direction) {
4332                         if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) ||
4333                             (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
4334                             (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
4335                                 REASON_SET(reason, PFRES_SYNPROXY);
4336                                 return (PF_DROP);
4337                         }
4338                         (*state)->src.max_win = MAX(ntohs(th->th_win), 1);
4339                         if ((*state)->dst.seqhi == 1)
4340                                 (*state)->dst.seqhi = htonl(karc4random());
4341                         pf_send_tcp((*state)->rule.ptr, pd->af,
4342                             &sk->addr[pd->sidx], &sk->addr[pd->didx],
4343                             sk->port[pd->sidx], sk->port[pd->didx],
4344                             (*state)->dst.seqhi, 0, TH_SYN, 0,
4345                             (*state)->src.mss, 0, 0, (*state)->tag, NULL, NULL);
4346                         REASON_SET(reason, PFRES_SYNPROXY);
4347                         return (PF_SYNPROXY_DROP);
4348                 } else if (((th->th_flags & (TH_SYN|TH_ACK)) !=
4349                     (TH_SYN|TH_ACK)) ||
4350                     (ntohl(th->th_ack) != (*state)->dst.seqhi + 1)) {
4351                         REASON_SET(reason, PFRES_SYNPROXY);
4352                         return (PF_DROP);
4353                 } else {
4354                         (*state)->dst.max_win = MAX(ntohs(th->th_win), 1);
4355                         (*state)->dst.seqlo = ntohl(th->th_seq);
4356                         pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
4357                             pd->src, th->th_dport, th->th_sport,
4358                             ntohl(th->th_ack), ntohl(th->th_seq) + 1,
4359                             TH_ACK, (*state)->src.max_win, 0, 0, 0,
4360                             (*state)->tag, NULL, NULL);
4361                         pf_send_tcp((*state)->rule.ptr, pd->af,
4362                             &sk->addr[pd->sidx], &sk->addr[pd->didx],
4363                             sk->port[pd->sidx], sk->port[pd->didx],
4364                             (*state)->src.seqhi + 1, (*state)->src.seqlo + 1,
4365                             TH_ACK, (*state)->dst.max_win, 0, 0, 1,
4366                             0, NULL, NULL);
4367                         (*state)->src.seqdiff = (*state)->dst.seqhi -
4368                             (*state)->src.seqlo;
4369                         (*state)->dst.seqdiff = (*state)->src.seqhi -
4370                             (*state)->dst.seqlo;
4371                         (*state)->src.seqhi = (*state)->src.seqlo +
4372                             (*state)->dst.max_win;
4373                         (*state)->dst.seqhi = (*state)->dst.seqlo +
4374                             (*state)->src.max_win;
4375                         (*state)->src.wscale = (*state)->dst.wscale = 0;
4376                         (*state)->src.state = (*state)->dst.state =
4377                             TCPS_ESTABLISHED;
4378                         REASON_SET(reason, PFRES_SYNPROXY);
4379                         return (PF_SYNPROXY_DROP);
4380                 }
4381         }
4382
4383         if (((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN) &&
4384             dst->state >= TCPS_FIN_WAIT_2 &&
4385             src->state >= TCPS_FIN_WAIT_2) {
4386                 if (pf_status.debug >= PF_DEBUG_MISC) {
4387                         kprintf("pf: state reuse ");
4388                         pf_print_state(*state);
4389                         pf_print_flags(th->th_flags);
4390                         kprintf("\n");
4391                 }
4392                 /* XXX make sure it's the same direction ?? */
4393                 (*state)->src.state = (*state)->dst.state = TCPS_CLOSED;
4394                 pf_unlink_state(*state);
4395                 *state = NULL;
4396                 return (PF_DROP);
4397         }
4398
4399         if ((*state)->state_flags & PFSTATE_SLOPPY) {
4400                 if (pf_tcp_track_sloppy(src, dst, state, pd, reason) == PF_DROP)
4401                         return (PF_DROP);
4402         } else {
4403                 if (pf_tcp_track_full(src, dst, state, kif, m, off, pd, reason,
4404                     &copyback) == PF_DROP)
4405                         return (PF_DROP);
4406         }
4407
4408         /* translate source/destination address, if necessary */
4409         if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) {
4410                 struct pf_state_key *nk = (*state)->key[pd->didx];
4411
4412                 if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], pd->af) ||
4413                     nk->port[pd->sidx] != th->th_sport)
4414                         pf_change_ap(pd->src, &th->th_sport, pd->ip_sum,
4415                             &th->th_sum, &nk->addr[pd->sidx],
4416                             nk->port[pd->sidx], 0, pd->af);
4417
4418                 if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], pd->af) ||
4419                     nk->port[pd->didx] != th->th_dport) {
4420                         /*
4421                          * If we don't redispatch the packet will go into
4422                          * the protocol stack on the wrong cpu for the
4423                          * post-translated address.
4424                          */
4425                         m->m_flags &= ~M_HASH;
4426                         pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum,
4427                             &th->th_sum, &nk->addr[pd->didx],
4428                             nk->port[pd->didx], 0, pd->af);
4429                 }
4430                 copyback = 1;
4431         }
4432
4433         /* Copyback sequence modulation or stateful scrub changes if needed */
4434         if (copyback)
4435                 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4436
4437         return (PF_PASS);
4438 }
4439
4440 int
4441 pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif,
4442     struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4443 {
4444         struct pf_state_peer    *src, *dst;
4445         struct pf_state_key_cmp  key;
4446         struct udphdr           *uh = pd->hdr.udp;
4447
4448         key.af = pd->af;
4449         key.proto = IPPROTO_UDP;
4450         if (direction == PF_IN) {       /* wire side, straight */
4451                 PF_ACPY(&key.addr[0], pd->src, key.af);
4452                 PF_ACPY(&key.addr[1], pd->dst, key.af);
4453                 key.port[0] = uh->uh_sport;
4454                 key.port[1] = uh->uh_dport;
4455         } else {                        /* stack side, reverse */
4456                 PF_ACPY(&key.addr[1], pd->src, key.af);
4457                 PF_ACPY(&key.addr[0], pd->dst, key.af);
4458                 key.port[1] = uh->uh_sport;
4459                 key.port[0] = uh->uh_dport;
4460         }
4461
4462         STATE_LOOKUP(kif, &key, direction, *state, m);
4463
4464         if (direction == (*state)->direction) {
4465                 src = &(*state)->src;
4466                 dst = &(*state)->dst;
4467         } else {
4468                 src = &(*state)->dst;
4469                 dst = &(*state)->src;
4470         }
4471
4472         /* update states */
4473         if (src->state < PFUDPS_SINGLE)
4474                 src->state = PFUDPS_SINGLE;
4475         if (dst->state == PFUDPS_SINGLE)
4476                 dst->state = PFUDPS_MULTIPLE;
4477
4478         /* update expire time */
4479         (*state)->expire = time_second;
4480         if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE)
4481                 (*state)->timeout = PFTM_UDP_MULTIPLE;
4482         else
4483                 (*state)->timeout = PFTM_UDP_SINGLE;
4484
4485         /* translate source/destination address, if necessary */
4486         if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) {
4487                 struct pf_state_key *nk = (*state)->key[pd->didx];
4488
4489                 if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], pd->af) ||
4490                     nk->port[pd->sidx] != uh->uh_sport)
4491                         pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum,
4492                             &uh->uh_sum, &nk->addr[pd->sidx],
4493                             nk->port[pd->sidx], 1, pd->af);
4494
4495                 if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], pd->af) ||
4496                     nk->port[pd->didx] != uh->uh_dport) {
4497                         /*
4498                          * If we don't redispatch the packet will go into
4499                          * the protocol stack on the wrong cpu for the
4500                          * post-translated address.
4501                          */
4502                         m->m_flags &= ~M_HASH;
4503                         pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum,
4504                             &uh->uh_sum, &nk->addr[pd->didx],
4505                             nk->port[pd->didx], 1, pd->af);
4506                 }
4507                 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
4508         }
4509
4510         return (PF_PASS);
4511 }
4512
4513 int
4514 pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif,
4515     struct mbuf *m, int off, void *h, struct pf_pdesc *pd, u_short *reason)
4516 {
4517         struct pf_addr  *saddr = pd->src, *daddr = pd->dst;
4518         u_int16_t        icmpid = 0, *icmpsum;
4519         u_int8_t         icmptype;
4520         int              state_icmp = 0;
4521         struct pf_state_key_cmp key;
4522
4523         switch (pd->proto) {
4524 #ifdef INET
4525         case IPPROTO_ICMP:
4526                 icmptype = pd->hdr.icmp->icmp_type;
4527                 icmpid = pd->hdr.icmp->icmp_id;
4528                 icmpsum = &pd->hdr.icmp->icmp_cksum;
4529
4530                 if (icmptype == ICMP_UNREACH ||
4531                     icmptype == ICMP_SOURCEQUENCH ||
4532                     icmptype == ICMP_REDIRECT ||
4533                     icmptype == ICMP_TIMXCEED ||
4534                     icmptype == ICMP_PARAMPROB)
4535                         state_icmp++;
4536                 break;
4537 #endif /* INET */
4538 #ifdef INET6
4539         case IPPROTO_ICMPV6:
4540                 icmptype = pd->hdr.icmp6->icmp6_type;
4541                 icmpid = pd->hdr.icmp6->icmp6_id;
4542                 icmpsum = &pd->hdr.icmp6->icmp6_cksum;
4543
4544                 if (icmptype == ICMP6_DST_UNREACH ||
4545                     icmptype == ICMP6_PACKET_TOO_BIG ||
4546                     icmptype == ICMP6_TIME_EXCEEDED ||
4547                     icmptype == ICMP6_PARAM_PROB)
4548                         state_icmp++;
4549                 break;
4550 #endif /* INET6 */
4551         }
4552
4553         if (!state_icmp) {
4554
4555                 /*
4556                  * ICMP query/reply message not related to a TCP/UDP packet.
4557                  * Search for an ICMP state.
4558                  */
4559                 key.af = pd->af;
4560                 key.proto = pd->proto;
4561                 key.port[0] = key.port[1] = icmpid;
4562                 if (direction == PF_IN) {       /* wire side, straight */
4563                         PF_ACPY(&key.addr[0], pd->src, key.af);
4564                         PF_ACPY(&key.addr[1], pd->dst, key.af);
4565                 } else {                        /* stack side, reverse */
4566                         PF_ACPY(&key.addr[1], pd->src, key.af);
4567                         PF_ACPY(&key.addr[0], pd->dst, key.af);
4568                 }
4569
4570                 STATE_LOOKUP(kif, &key, direction, *state, m);
4571
4572                 (*state)->expire = time_second;
4573                 (*state)->timeout = PFTM_ICMP_ERROR_REPLY;
4574
4575                 /* translate source/destination address, if necessary */
4576                 if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) {
4577                         struct pf_state_key *nk = (*state)->key[pd->didx];
4578
4579                         switch (pd->af) {
4580 #ifdef INET
4581                         case AF_INET:
4582                                 if (PF_ANEQ(pd->src,
4583                                     &nk->addr[pd->sidx], AF_INET))
4584                                         pf_change_a(&saddr->v4.s_addr,
4585                                             pd->ip_sum,
4586                                             nk->addr[pd->sidx].v4.s_addr, 0);
4587
4588                                 if (PF_ANEQ(pd->dst, &nk->addr[pd->didx],
4589                                     AF_INET))
4590                                         pf_change_a(&daddr->v4.s_addr,
4591                                             pd->ip_sum,
4592                                             nk->addr[pd->didx].v4.s_addr, 0);
4593
4594                                 if (nk->port[0] !=
4595                                     pd->hdr.icmp->icmp_id) {
4596                                         pd->hdr.icmp->icmp_cksum =
4597                                             pf_cksum_fixup(
4598                                             pd->hdr.icmp->icmp_cksum, icmpid,
4599                                             nk->port[pd->sidx], 0);
4600                                         pd->hdr.icmp->icmp_id =
4601                                             nk->port[pd->sidx];
4602                                 }
4603
4604                                 m_copyback(m, off, ICMP_MINLEN,
4605                                     (caddr_t)pd->hdr.icmp);
4606                                 break;
4607 #endif /* INET */
4608 #ifdef INET6
4609                         case AF_INET6:
4610                                 if (PF_ANEQ(pd->src,
4611                                     &nk->addr[pd->sidx], AF_INET6))
4612                                         pf_change_a6(saddr,
4613                                             &pd->hdr.icmp6->icmp6_cksum,
4614                                             &nk->addr[pd->sidx], 0);
4615
4616                                 if (PF_ANEQ(pd->dst,
4617                                     &nk->addr[pd->didx], AF_INET6))
4618                                         pf_change_a6(daddr,
4619                                             &pd->hdr.icmp6->icmp6_cksum,
4620                                             &nk->addr[pd->didx], 0);
4621
4622                                 m_copyback(m, off,
4623                                         sizeof(struct icmp6_hdr),
4624                                         (caddr_t)pd->hdr.icmp6);
4625                                 break;
4626 #endif /* INET6 */
4627                         }
4628                 }
4629                 return (PF_PASS);
4630
4631         } else {
4632                 /*
4633                  * ICMP error message in response to a TCP/UDP packet.
4634                  * Extract the inner TCP/UDP header and search for that state.
4635                  */
4636
4637                 struct pf_pdesc pd2;
4638 #ifdef INET
4639                 struct ip       h2;
4640 #endif /* INET */
4641 #ifdef INET6
4642                 struct ip6_hdr  h2_6;
4643                 int             terminal = 0;
4644 #endif /* INET6 */
4645                 int             ipoff2;
4646                 int             off2;
4647
4648                 pd2.af = pd->af;
4649                 /* Payload packet is from the opposite direction. */
4650                 pd2.sidx = (direction == PF_IN) ? 1 : 0;
4651                 pd2.didx = (direction == PF_IN) ? 0 : 1;
4652                 switch (pd->af) {
4653 #ifdef INET
4654                 case AF_INET:
4655                         /* offset of h2 in mbuf chain */
4656                         ipoff2 = off + ICMP_MINLEN;
4657
4658                         if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2),
4659                             NULL, reason, pd2.af)) {
4660                                 DPFPRINTF(PF_DEBUG_MISC,
4661                                     ("pf: ICMP error message too short "
4662                                     "(ip)\n"));
4663                                 return (PF_DROP);
4664                         }
4665                         /*
4666                          * ICMP error messages don't refer to non-first
4667                          * fragments
4668                          */
4669                         if (h2.ip_off & htons(IP_OFFMASK)) {
4670                                 REASON_SET(reason, PFRES_FRAG);
4671                                 return (PF_DROP);
4672                         }
4673
4674                         /* offset of protocol header that follows h2 */
4675                         off2 = ipoff2 + (h2.ip_hl << 2);
4676
4677                         pd2.proto = h2.ip_p;
4678                         pd2.src = (struct pf_addr *)&h2.ip_src;
4679                         pd2.dst = (struct pf_addr *)&h2.ip_dst;
4680                         pd2.ip_sum = &h2.ip_sum;
4681                         break;
4682 #endif /* INET */
4683 #ifdef INET6
4684                 case AF_INET6:
4685                         ipoff2 = off + sizeof(struct icmp6_hdr);
4686
4687                         if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6),
4688                             NULL, reason, pd2.af)) {
4689                                 DPFPRINTF(PF_DEBUG_MISC,
4690                                     ("pf: ICMP error message too short "
4691                                     "(ip6)\n"));
4692                                 return (PF_DROP);
4693                         }
4694                         pd2.proto = h2_6.ip6_nxt;
4695                         pd2.src = (struct pf_addr *)&h2_6.ip6_src;
4696                         pd2.dst = (struct pf_addr *)&h2_6.ip6_dst;
4697                         pd2.ip_sum = NULL;
4698                         off2 = ipoff2 + sizeof(h2_6);
4699                         do {
4700                                 switch (pd2.proto) {
4701                                 case IPPROTO_FRAGMENT:
4702                                         /*
4703                                          * ICMPv6 error messages for
4704                                          * non-first fragments
4705                                          */
4706                                         REASON_SET(reason, PFRES_FRAG);
4707                                         return (PF_DROP);
4708                                 case IPPROTO_AH:
4709                                 case IPPROTO_HOPOPTS:
4710                                 case IPPROTO_ROUTING:
4711                                 case IPPROTO_DSTOPTS: {
4712                                         /* get next header and header length */
4713                                         struct ip6_ext opt6;
4714
4715                                         if (!pf_pull_hdr(m, off2, &opt6,
4716                                             sizeof(opt6), NULL, reason,
4717                                             pd2.af)) {
4718                                                 DPFPRINTF(PF_DEBUG_MISC,
4719                                                     ("pf: ICMPv6 short opt\n"));
4720                                                 return (PF_DROP);
4721                                         }
4722                                         if (pd2.proto == IPPROTO_AH)
4723                                                 off2 += (opt6.ip6e_len + 2) * 4;
4724                                         else
4725                                                 off2 += (opt6.ip6e_len + 1) * 8;
4726                                         pd2.proto = opt6.ip6e_nxt;
4727                                         /* goto the next header */
4728                                         break;
4729                                 }
4730                                 default:
4731                                         terminal++;
4732                                         break;
4733                                 }
4734                         } while (!terminal);
4735                         break;
4736 #endif /* INET6 */
4737                 default:
4738                         DPFPRINTF(PF_DEBUG_MISC,
4739                             ("pf: ICMP AF %d unknown (ip6)\n", pd->af));
4740                         return (PF_DROP);
4741                         break;
4742                 }
4743
4744                 switch (pd2.proto) {
4745                 case IPPROTO_TCP: {
4746                         struct tcphdr            th;
4747                         u_int32_t                seq;
4748                         struct pf_state_peer    *src, *dst;
4749                         u_int8_t                 dws;
4750                         int                      copyback = 0;
4751
4752                         /*
4753                          * Only the first 8 bytes of the TCP header can be
4754                          * expected. Don't access any TCP header fields after
4755                          * th_seq, an ackskew test is not possible.
4756                          */
4757                         if (!pf_pull_hdr(m, off2, &th, 8, NULL, reason,
4758                             pd2.af)) {
4759                                 DPFPRINTF(PF_DEBUG_MISC,
4760                                     ("pf: ICMP error message too short "
4761                                     "(tcp)\n"));
4762                                 return (PF_DROP);
4763                         }
4764
4765                         key.af = pd2.af;
4766                         key.proto = IPPROTO_TCP;
4767                         PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
4768                         PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
4769                         key.port[pd2.sidx] = th.th_sport;
4770                         key.port[pd2.didx] = th.th_dport;
4771
4772                         STATE_LOOKUP(kif, &key, direction, *state, m);
4773
4774                         if (direction == (*state)->direction) {
4775                                 src = &(*state)->dst;
4776                                 dst = &(*state)->src;
4777                         } else {
4778                                 src = &(*state)->src;
4779                                 dst = &(*state)->dst;
4780                         }
4781
4782                         if (src->wscale && dst->wscale)
4783                                 dws = dst->wscale & PF_WSCALE_MASK;
4784                         else
4785                                 dws = 0;
4786
4787                         /* Demodulate sequence number */
4788                         seq = ntohl(th.th_seq) - src->seqdiff;
4789                         if (src->seqdiff) {
4790                                 pf_change_a(&th.th_seq, icmpsum,
4791                                     htonl(seq), 0);
4792                                 copyback = 1;
4793                         }
4794
4795                         if (!((*state)->state_flags & PFSTATE_SLOPPY) &&
4796                             (!SEQ_GEQ(src->seqhi, seq) ||
4797                             !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)))) {
4798                                 if (pf_status.debug >= PF_DEBUG_MISC) {
4799                                         kprintf("pf: BAD ICMP %d:%d ",
4800                                             icmptype, pd->hdr.icmp->icmp_code);
4801                                         pf_print_host(pd->src, 0, pd->af);
4802                                         kprintf(" -> ");
4803                                         pf_print_host(pd->dst, 0, pd->af);
4804                                         kprintf(" state: ");
4805                                         pf_print_state(*state);
4806                                         kprintf(" seq=%u\n", seq);
4807                                 }
4808                                 REASON_SET(reason, PFRES_BADSTATE);
4809                                 return (PF_DROP);
4810                         } else {
4811                                 if (pf_status.debug >= PF_DEBUG_MISC) {
4812                                         kprintf("pf: OK ICMP %d:%d ",
4813                                             icmptype, pd->hdr.icmp->icmp_code);
4814                                         pf_print_host(pd->src, 0, pd->af);
4815                                         kprintf(" -> ");
4816                                         pf_print_host(pd->dst, 0, pd->af);
4817                                         kprintf(" state: ");
4818                                         pf_print_state(*state);
4819                                         kprintf(" seq=%u\n", seq);
4820                                 }
4821                         }
4822
4823                         /* translate source/destination address, if necessary */
4824                         if ((*state)->key[PF_SK_WIRE] !=
4825                             (*state)->key[PF_SK_STACK]) {
4826                                 struct pf_state_key *nk =
4827                                     (*state)->key[pd->didx];
4828
4829                                 if (PF_ANEQ(pd2.src,
4830                                     &nk->addr[pd2.sidx], pd2.af) ||
4831                                     nk->port[pd2.sidx] != th.th_sport)
4832                                         pf_change_icmp(pd2.src, &th.th_sport,
4833                                             daddr, &nk->addr[pd2.sidx],
4834                                             nk->port[pd2.sidx], NULL,
4835                                             pd2.ip_sum, icmpsum,
4836                                             pd->ip_sum, 0, pd2.af);
4837
4838                                 if (PF_ANEQ(pd2.dst,
4839                                     &nk->addr[pd2.didx], pd2.af) ||
4840                                     nk->port[pd2.didx] != th.th_dport)
4841                                         pf_change_icmp(pd2.dst, &th.th_dport,
4842                                             NULL, /* XXX Inbound NAT? */
4843                                             &nk->addr[pd2.didx],
4844                                             nk->port[pd2.didx], NULL,
4845                                             pd2.ip_sum, icmpsum,
4846                                             pd->ip_sum, 0, pd2.af);
4847                                 copyback = 1;
4848                         }
4849
4850                         if (copyback) {
4851                                 switch (pd2.af) {
4852 #ifdef INET
4853                                 case AF_INET:
4854                                         m_copyback(m, off, ICMP_MINLEN,
4855                                             (caddr_t)pd->hdr.icmp);
4856                                         m_copyback(m, ipoff2, sizeof(h2),
4857                                             (caddr_t)&h2);
4858                                         break;
4859 #endif /* INET */
4860 #ifdef INET6
4861                                 case AF_INET6:
4862                                         m_copyback(m, off,
4863                                             sizeof(struct icmp6_hdr),
4864                                             (caddr_t)pd->hdr.icmp6);
4865                                         m_copyback(m, ipoff2, sizeof(h2_6),
4866                                             (caddr_t)&h2_6);
4867                                         break;
4868 #endif /* INET6 */
4869                                 }
4870                                 m_copyback(m, off2, 8, (caddr_t)&th);
4871                         }
4872
4873                         return (PF_PASS);
4874                         break;
4875                 }
4876                 case IPPROTO_UDP: {
4877                         struct udphdr           uh;
4878
4879                         if (!pf_pull_hdr(m, off2, &uh, sizeof(uh),
4880                             NULL, reason, pd2.af)) {
4881                                 DPFPRINTF(PF_DEBUG_MISC,
4882                                     ("pf: ICMP error message too short "
4883                                     "(udp)\n"));
4884                                 return (PF_DROP);
4885                         }
4886
4887                         key.af = pd2.af;
4888                         key.proto = IPPROTO_UDP;
4889                         PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
4890                         PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
4891                         key.port[pd2.sidx] = uh.uh_sport;
4892                         key.port[pd2.didx] = uh.uh_dport;
4893
4894                         STATE_LOOKUP(kif, &key, direction, *state, m);
4895
4896                         /* translate source/destination address, if necessary */
4897                         if ((*state)->key[PF_SK_WIRE] !=
4898                             (*state)->key[PF_SK_STACK]) {
4899                                 struct pf_state_key *nk =
4900                                     (*state)->key[pd->didx];
4901
4902                                 if (PF_ANEQ(pd2.src,
4903                                     &nk->addr[pd2.sidx], pd2.af) ||
4904                                     nk->port[pd2.sidx] != uh.uh_sport)
4905                                         pf_change_icmp(pd2.src, &uh.uh_sport,
4906                                             daddr, &nk->addr[pd2.sidx],
4907                                             nk->port[pd2.sidx], &uh.uh_sum,
4908                                             pd2.ip_sum, icmpsum,
4909                                             pd->ip_sum, 1, pd2.af);
4910
4911                                 if (PF_ANEQ(pd2.dst,
4912                                     &nk->addr[pd2.didx], pd2.af) ||
4913                                     nk->port[pd2.didx] != uh.uh_dport)
4914                                         pf_change_icmp(pd2.dst, &uh.uh_dport,
4915                                             NULL, /* XXX Inbound NAT? */
4916                                             &nk->addr[pd2.didx],
4917                                             nk->port[pd2.didx], &uh.uh_sum,
4918                                             pd2.ip_sum, icmpsum,
4919                                             pd->ip_sum, 1, pd2.af);
4920
4921                                 switch (pd2.af) {
4922 #ifdef INET
4923                                 case AF_INET:
4924                                         m_copyback(m, off, ICMP_MINLEN,
4925                                             (caddr_t)pd->hdr.icmp);
4926                                         m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4927                                         break;
4928 #endif /* INET */
4929 #ifdef INET6
4930                                 case AF_INET6:
4931                                         m_copyback(m, off,
4932                                             sizeof(struct icmp6_hdr),
4933                                             (caddr_t)pd->hdr.icmp6);
4934                                         m_copyback(m, ipoff2, sizeof(h2_6),
4935                                             (caddr_t)&h2_6);
4936                                         break;
4937 #endif /* INET6 */
4938                                 }
4939                                 m_copyback(m, off2, sizeof(uh), (caddr_t)&uh);
4940                         }
4941
4942                         return (PF_PASS);
4943                         break;
4944                 }
4945 #ifdef INET
4946                 case IPPROTO_ICMP: {
4947                         struct icmp             iih;
4948
4949                         if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN,
4950                             NULL, reason, pd2.af)) {
4951                                 DPFPRINTF(PF_DEBUG_MISC,
4952                                     ("pf: ICMP error message too short i"
4953                                     "(icmp)\n"));
4954                                 return (PF_DROP);
4955                         }
4956
4957                         key.af = pd2.af;
4958                         key.proto = IPPROTO_ICMP;
4959                         PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
4960                         PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
4961                         key.port[0] = key.port[1] = iih.icmp_id;
4962
4963                         STATE_LOOKUP(kif, &key, direction, *state, m);
4964
4965                         /* translate source/destination address, if necessary */
4966                         if ((*state)->key[PF_SK_WIRE] !=
4967                             (*state)->key[PF_SK_STACK]) {
4968                                 struct pf_state_key *nk =
4969                                     (*state)->key[pd->didx];
4970
4971                                 if (PF_ANEQ(pd2.src,
4972                                     &nk->addr[pd2.sidx], pd2.af) ||
4973                                     nk->port[pd2.sidx] != iih.icmp_id)
4974                                         pf_change_icmp(pd2.src, &iih.icmp_id,
4975                                             daddr, &nk->addr[pd2.sidx],
4976                                             nk->port[pd2.sidx], NULL,
4977                                             pd2.ip_sum, icmpsum,
4978                                             pd->ip_sum, 0, AF_INET);
4979
4980                                 if (PF_ANEQ(pd2.dst,
4981                                     &nk->addr[pd2.didx], pd2.af) ||
4982                                     nk->port[pd2.didx] != iih.icmp_id)
4983                                         pf_change_icmp(pd2.dst, &iih.icmp_id,
4984                                             NULL, /* XXX Inbound NAT? */
4985                                             &nk->addr[pd2.didx],
4986                                             nk->port[pd2.didx], NULL,
4987                                             pd2.ip_sum, icmpsum,
4988                                             pd->ip_sum, 0, AF_INET);
4989
4990                                 m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
4991                                 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4992                                 m_copyback(m, off2, ICMP_MINLEN, (caddr_t)&iih);
4993                         }
4994                         return (PF_PASS);
4995                         break;
4996                 }
4997 #endif /* INET */
4998 #ifdef INET6
4999                 case IPPROTO_ICMPV6: {
5000                         struct icmp6_hdr        iih;
5001
5002                         if (!pf_pull_hdr(m, off2, &iih,
5003                             sizeof(struct icmp6_hdr), NULL, reason, pd2.af)) {
5004                                 DPFPRINTF(PF_DEBUG_MISC,
5005                                     ("pf: ICMP error message too short "
5006                                     "(icmp6)\n"));
5007                                 return (PF_DROP);
5008                         }
5009
5010                         key.af = pd2.af;
5011                         key.proto = IPPROTO_ICMPV6;
5012                         PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
5013                         PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
5014                         key.port[0] = key.port[1] = iih.icmp6_id;
5015
5016                         STATE_LOOKUP(kif, &key, direction, *state, m);
5017
5018                         /* translate source/destination address, if necessary */
5019                         if ((*state)->key[PF_SK_WIRE] !=
5020                             (*state)->key[PF_SK_STACK]) {
5021                                 struct pf_state_key *nk =
5022                                     (*state)->key[pd->didx];
5023
5024                                 if (PF_ANEQ(pd2.src,
5025                                     &nk->addr[pd2.sidx], pd2.af) ||
5026                                     nk->port[pd2.sidx] != iih.icmp6_id)
5027                                         pf_change_icmp(pd2.src, &iih.icmp6_id,
5028                                             daddr, &nk->addr[pd2.sidx],
5029                                             nk->port[pd2.sidx], NULL,
5030                                             pd2.ip_sum, icmpsum,
5031                                             pd->ip_sum, 0, AF_INET6);
5032
5033                                 if (PF_ANEQ(pd2.dst,
5034                                     &nk->addr[pd2.didx], pd2.af) ||
5035                                     nk->port[pd2.didx] != iih.icmp6_id)
5036                                         pf_change_icmp(pd2.dst, &iih.icmp6_id,
5037                                             NULL, /* XXX Inbound NAT? */
5038                                             &nk->addr[pd2.didx],
5039                                             nk->port[pd2.didx], NULL,
5040                                             pd2.ip_sum, icmpsum,
5041                                             pd->ip_sum, 0, AF_INET6);
5042
5043                                 m_copyback(m, off, sizeof(struct icmp6_hdr),
5044                                     (caddr_t)pd->hdr.icmp6);
5045                                 m_copyback(m, ipoff2, sizeof(h2_6), (caddr_t)&h2_6);
5046                                 m_copyback(m, off2, sizeof(struct icmp6_hdr),
5047                                     (caddr_t)&iih);
5048                         }
5049
5050                         return (PF_PASS);
5051                         break;
5052                 }
5053 #endif /* INET6 */
5054                 default: {
5055                         key.af = pd2.af;
5056                         key.proto = pd2.proto;
5057                         PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
5058                         PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
5059                         key.port[0] = key.port[1] = 0;
5060
5061                         STATE_LOOKUP(kif, &key, direction, *state, m);
5062
5063                         /* translate source/destination address, if necessary */
5064                         if ((*state)->key[PF_SK_WIRE] !=
5065                             (*state)->key[PF_SK_STACK]) {
5066                                 struct pf_state_key *nk =
5067                                     (*state)->key[pd->didx];
5068
5069                                 if (PF_ANEQ(pd2.src,
5070                                     &nk->addr[pd2.sidx], pd2.af))
5071                                         pf_change_icmp(pd2.src, NULL, daddr,
5072                                             &nk->addr[pd2.sidx], 0, NULL,
5073                                             pd2.ip_sum, icmpsum,
5074                                             pd->ip_sum, 0, pd2.af);
5075
5076                                 if (PF_ANEQ(pd2.dst,
5077                                     &nk->addr[pd2.didx], pd2.af))
5078                                         pf_change_icmp(pd2.src, NULL,
5079                                             NULL, /* XXX Inbound NAT? */
5080                                             &nk->addr[pd2.didx], 0, NULL,
5081                                             pd2.ip_sum, icmpsum,
5082                                             pd->ip_sum, 0, pd2.af);
5083
5084                                 switch (pd2.af) {
5085 #ifdef INET
5086                                 case AF_INET:
5087                                         m_copyback(m, off, ICMP_MINLEN,
5088                                             (caddr_t)pd->hdr.icmp);
5089                                         m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
5090                                         break;
5091 #endif /* INET */
5092 #ifdef INET6
5093                                 case AF_INET6:
5094                                         m_copyback(m, off,
5095                                             sizeof(struct icmp6_hdr),
5096                                             (caddr_t)pd->hdr.icmp6);
5097                                         m_copyback(m, ipoff2, sizeof(h2_6),
5098                                             (caddr_t)&h2_6);
5099                                         break;
5100 #endif /* INET6 */
5101                                 }
5102                         }
5103                         return (PF_PASS);
5104                         break;
5105                 }
5106                 }
5107         }
5108 }
5109
5110 int
5111 pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif,
5112     struct mbuf *m, struct pf_pdesc *pd)
5113 {
5114         struct pf_state_peer    *src, *dst;
5115         struct pf_state_key_cmp  key;
5116
5117         key.af = pd->af;
5118         key.proto = pd->proto;
5119         if (direction == PF_IN) {
5120                 PF_ACPY(&key.addr[0], pd->src, key.af);
5121                 PF_ACPY(&key.addr[1], pd->dst, key.af);
5122                 key.port[0] = key.port[1] = 0;
5123         } else {
5124                 PF_ACPY(&key.addr[1], pd->src, key.af);
5125                 PF_ACPY(&key.addr[0], pd->dst, key.af);
5126                 key.port[1] = key.port[0] = 0;
5127         }
5128
5129         STATE_LOOKUP(kif, &key, direction, *state, m);
5130
5131         if (direction == (*state)->direction) {
5132                 src = &(*state)->src;
5133                 dst = &(*state)->dst;
5134         } else {
5135                 src = &(*state)->dst;
5136                 dst = &(*state)->src;
5137         }
5138
5139         /* update states */
5140         if (src->state < PFOTHERS_SINGLE)
5141                 src->state = PFOTHERS_SINGLE;
5142         if (dst->state == PFOTHERS_SINGLE)
5143                 dst->state = PFOTHERS_MULTIPLE;
5144
5145         /* update expire time */
5146         (*state)->expire = time_second;
5147         if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE)
5148                 (*state)->timeout = PFTM_OTHER_MULTIPLE;
5149         else
5150                 (*state)->timeout = PFTM_OTHER_SINGLE;
5151
5152         /* translate source/destination address, if necessary */
5153         if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) {
5154                 struct pf_state_key *nk = (*state)->key[pd->didx];
5155
5156                 KKASSERT(nk);
5157                 KKASSERT(pd);
5158                 KKASSERT(pd->src);
5159                 KKASSERT(pd->dst);
5160                 switch (pd->af) {
5161 #ifdef INET
5162                 case AF_INET:
5163                         if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], AF_INET))
5164                                 pf_change_a(&pd->src->v4.s_addr,
5165                                     pd->ip_sum,
5166                                     nk->addr[pd->sidx].v4.s_addr,
5167                                     0);
5168
5169
5170                         if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], AF_INET))
5171                                 pf_change_a(&pd->dst->v4.s_addr,
5172                                     pd->ip_sum,
5173                                     nk->addr[pd->didx].v4.s_addr,
5174                                     0);
5175
5176                                 break;
5177 #endif /* INET */
5178 #ifdef INET6
5179                 case AF_INET6:
5180                         if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], AF_INET))
5181                                 PF_ACPY(pd->src, &nk->addr[pd->sidx], pd->af);
5182
5183                         if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], AF_INET))
5184                                 PF_ACPY(pd->dst, &nk->addr[pd->didx], pd->af);
5185 #endif /* INET6 */
5186                 }
5187         }
5188         return (PF_PASS);
5189 }
5190
5191 /*
5192  * ipoff and off are measured from the start of the mbuf chain.
5193  * h must be at "ipoff" on the mbuf chain.
5194  */
5195 void *
5196 pf_pull_hdr(struct mbuf *m, int off, void *p, int len,
5197     u_short *actionp, u_short *reasonp, sa_family_t af)
5198 {
5199         switch (af) {
5200 #ifdef INET
5201         case AF_INET: {
5202                 struct ip       *h = mtod(m, struct ip *);
5203                 u_int16_t        fragoff = (h->ip_off & IP_OFFMASK) << 3;
5204
5205                 if (fragoff) {
5206                         if (fragoff >= len)
5207                                 ACTION_SET(actionp, PF_PASS);
5208                         else {
5209                                 ACTION_SET(actionp, PF_DROP);
5210                                 REASON_SET(reasonp, PFRES_FRAG);
5211                         }
5212                         return (NULL);
5213                 }
5214                 if (m->m_pkthdr.len < off + len ||
5215                     h->ip_len < off + len) {
5216                         ACTION_SET(actionp, PF_DROP);
5217                         REASON_SET(reasonp, PFRES_SHORT);
5218                         return (NULL);
5219                 }
5220                 break;
5221         }
5222 #endif /* INET */
5223 #ifdef INET6
5224         case AF_INET6: {
5225                 struct ip6_hdr  *h = mtod(m, struct ip6_hdr *);
5226
5227                 if (m->m_pkthdr.len < off + len ||
5228                     (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) <
5229                     (unsigned)(off + len)) {
5230                         ACTION_SET(actionp, PF_DROP);
5231                         REASON_SET(reasonp, PFRES_SHORT);
5232                         return (NULL);
5233                 }
5234                 break;
5235         }
5236 #endif /* INET6 */
5237         }
5238         m_copydata(m, off, len, p);
5239         return (p);
5240 }
5241
5242 int
5243 pf_routable(struct pf_addr *addr, sa_family_t af, struct pfi_kif *kif)
5244 {
5245         struct sockaddr_in      *dst;
5246         int                      ret = 1;
5247         int                      check_mpath;
5248 #ifdef INET6
5249         struct sockaddr_in6     *dst6;
5250         struct route_in6         ro;
5251 #else
5252         struct route             ro;
5253 #endif
5254         struct radix_node       *rn;
5255         struct rtentry          *rt;
5256         struct ifnet            *ifp;
5257
5258         check_mpath = 0;
5259         bzero(&ro, sizeof(ro));
5260         switch (af) {
5261         case AF_INET:
5262                 dst = satosin(&ro.ro_dst);
5263                 dst->sin_family = AF_INET;
5264                 dst->sin_len = sizeof(*dst);
5265                 dst->sin_addr = addr->v4;
5266                 break;
5267 #ifdef INET6
5268         case AF_INET6:
5269                 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
5270                 dst6->sin6_family = AF_INET6;
5271                 dst6->sin6_len = sizeof(*dst6);
5272                 dst6->sin6_addr = addr->v6;
5273                 break;
5274 #endif /* INET6 */
5275         default:
5276                 return (0);
5277         }
5278
5279         /* Skip checks for ipsec interfaces */
5280         if (kif != NULL && kif->pfik_ifp->if_type == IFT_ENC)
5281                 goto out;
5282
5283         rtalloc_ign((struct route *)&ro, 0);
5284
5285         if (ro.ro_rt != NULL) {
5286                 /* No interface given, this is a no-route check */
5287                 if (kif == NULL)
5288                         goto out;
5289
5290                 if (kif->pfik_ifp == NULL) {
5291                         ret = 0;
5292                         goto out;
5293                 }
5294
5295                 /* Perform uRPF check if passed input interface */
5296                 ret = 0;
5297                 rn = (struct radix_node *)ro.ro_rt;
5298                 do {
5299                         rt = (struct rtentry *)rn;
5300                         ifp = rt->rt_ifp;
5301
5302                         if (kif->pfik_ifp == ifp)
5303                                 ret = 1;
5304                         rn = NULL;
5305                 } while (check_mpath == 1 && rn != NULL && ret == 0);
5306         } else
5307                 ret = 0;
5308 out:
5309         if (ro.ro_rt != NULL)
5310                 RTFREE(ro.ro_rt);
5311         return (ret);
5312 }
5313
5314 int
5315 pf_rtlabel_match(struct pf_addr *addr, sa_family_t af, struct pf_addr_wrap *aw)
5316 {
5317         struct sockaddr_in      *dst;
5318 #ifdef INET6
5319         struct sockaddr_in6     *dst6;
5320         struct route_in6         ro;
5321 #else
5322         struct route             ro;
5323 #endif
5324         int                      ret = 0;
5325
5326         ASSERT_LWKT_TOKEN_HELD(&pf_token);
5327
5328         bzero(&ro, sizeof(ro));
5329         switch (af) {
5330         case AF_INET:
5331                 dst = satosin(&ro.ro_dst);
5332                 dst->sin_family = AF_INET;
5333                 dst->sin_len = sizeof(*dst);
5334                 dst->sin_addr = addr->v4;
5335                 break;
5336 #ifdef INET6
5337         case AF_INET6:
5338                 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
5339                 dst6->sin6_family = AF_INET6;
5340                 dst6->sin6_len = sizeof(*dst6);
5341                 dst6->sin6_addr = addr->v6;
5342                 break;
5343 #endif /* INET6 */
5344         default:
5345                 return (0);
5346         }
5347
5348 rtalloc_ign((struct route *)&ro, (RTF_CLONING | RTF_PRCLONING));
5349
5350         if (ro.ro_rt != NULL) {
5351                 RTFREE(ro.ro_rt);
5352         }
5353
5354         return (ret);
5355 }
5356
5357 #ifdef INET
5358 void
5359 pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5360     struct pf_state *s, struct pf_pdesc *pd)
5361 {
5362         struct mbuf             *m0, *m1;
5363         struct route             iproute;
5364         struct route            *ro = NULL;
5365         struct sockaddr_in      *dst;
5366         struct ip               *ip;
5367         struct ifnet            *ifp = NULL;
5368         struct pf_addr           naddr;
5369         struct pf_src_node      *sn = NULL;
5370         int                      error = 0;
5371         int sw_csum;
5372 #ifdef IPSEC
5373         struct m_tag            *mtag;
5374 #endif /* IPSEC */
5375
5376         ASSERT_LWKT_TOKEN_HELD(&pf_token);
5377
5378         if (m == NULL || *m == NULL || r == NULL ||
5379             (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5380                 panic("pf_route: invalid parameters");
5381
5382         if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
5383                 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
5384                 (*m)->m_pkthdr.pf.routed = 1;
5385         } else {
5386                 if ((*m)->m_pkthdr.pf.routed++ > 3) {
5387                         m0 = *m;
5388                         *m = NULL;
5389                         goto bad;
5390                 }
5391         }
5392
5393         if (r->rt == PF_DUPTO) {
5394                 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL) {
5395                         return;
5396                 }
5397         } else {
5398                 if ((r->rt == PF_REPLYTO) == (r->direction == dir)) {
5399                         return;
5400                 }
5401                 m0 = *m;
5402         }
5403
5404         if (m0->m_len < sizeof(struct ip)) {
5405                 DPFPRINTF(PF_DEBUG_URGENT,
5406                     ("pf_route: m0->m_len < sizeof(struct ip)\n"));
5407                 goto bad;
5408         }
5409
5410         ip = mtod(m0, struct ip *);
5411
5412         ro = &iproute;
5413         bzero((caddr_t)ro, sizeof(*ro));
5414         dst = satosin(&ro->ro_dst);
5415         dst->sin_family = AF_INET;
5416         dst->sin_len = sizeof(*dst);
5417         dst->sin_addr = ip->ip_dst;
5418
5419         if (r->rt == PF_FASTROUTE) {
5420                 rtalloc(ro);
5421                 if (ro->ro_rt == 0) {
5422                         ipstat.ips_noroute++;
5423                         goto bad;
5424                 }
5425
5426                 ifp = ro->ro_rt->rt_ifp;
5427                 ro->ro_rt->rt_use++;
5428
5429                 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
5430                         dst = satosin(ro->ro_rt->rt_gateway);
5431         } else {
5432                 if (TAILQ_EMPTY(&r->rpool.list)) {
5433                         DPFPRINTF(PF_DEBUG_URGENT,
5434                             ("pf_route: TAILQ_EMPTY(&r->rpool.list)\n"));
5435                         goto bad;
5436                 }
5437                 if (s == NULL) {
5438                         pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src,
5439                             &naddr, NULL, &sn);
5440                         if (!PF_AZERO(&naddr, AF_INET))
5441                                 dst->sin_addr.s_addr = naddr.v4.s_addr;
5442                         ifp = r->rpool.cur->kif ?
5443                             r->rpool.cur->kif->pfik_ifp : NULL;
5444                 } else {
5445                         if (!PF_AZERO(&s->rt_addr, AF_INET))
5446                                 dst->sin_addr.s_addr =
5447                                     s->rt_addr.v4.s_addr;
5448                         ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5449                 }
5450         }
5451         if (ifp == NULL)
5452                 goto bad;
5453
5454         if (oifp != ifp) {
5455                 if (pf_test(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) {
5456                         goto bad;
5457                 } else if (m0 == NULL) {
5458                         goto done;
5459                 }
5460                 if (m0->m_len < sizeof(struct ip)) {
5461                         DPFPRINTF(PF_DEBUG_URGENT,
5462                             ("pf_route: m0->m_len < sizeof(struct ip)\n"));
5463                         goto bad;
5464                 }
5465                 ip = mtod(m0, struct ip *);
5466         }
5467
5468         /* Copied from FreeBSD 5.1-CURRENT ip_output. */
5469         m0->m_pkthdr.csum_flags |= CSUM_IP;
5470         sw_csum = m0->m_pkthdr.csum_flags & ~ifp->if_hwassist;
5471         if (sw_csum & CSUM_DELAY_DATA) {
5472                 in_delayed_cksum(m0);
5473                 sw_csum &= ~CSUM_DELAY_DATA;
5474         }
5475         m0->m_pkthdr.csum_flags &= ifp->if_hwassist;
5476
5477         if (ip->ip_len <= ifp->if_mtu ||
5478             (ifp->if_hwassist & CSUM_FRAGMENT &&
5479                 (ip->ip_off & IP_DF) == 0)) {
5480                 ip->ip_len = htons(ip->ip_len);
5481                 ip->ip_off = htons(ip->ip_off);
5482                 ip->ip_sum = 0;
5483                 if (sw_csum & CSUM_DELAY_IP) {
5484                         /* From KAME */
5485                         if (ip->ip_v == IPVERSION &&
5486                             (ip->ip_hl << 2) == sizeof(*ip)) {
5487                                 ip->ip_sum = in_cksum_hdr(ip);
5488                         } else {
5489                                 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
5490                         }
5491                 }
5492                 lwkt_reltoken(&pf_token);
5493                 error = ifp->if_output(ifp, m0, sintosa(dst), ro->ro_rt);
5494                 lwkt_gettoken(&pf_token);
5495                 goto done;
5496         }
5497
5498         /*
5499          * Too large for interface; fragment if possible.
5500          * Must be able to put at least 8 bytes per fragment.
5501          */
5502         if (ip->ip_off & IP_DF) {
5503                 ipstat.ips_cantfrag++;
5504                 if (r->rt != PF_DUPTO) {
5505                         icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0,
5506                             ifp->if_mtu);
5507                         goto done;
5508                 } else
5509                         goto bad;
5510         }
5511
5512         m1 = m0;
5513         error = ip_fragment(ip, &m0, ifp->if_mtu, ifp->if_hwassist, sw_csum);
5514         if (error) {
5515                 goto bad;
5516         }
5517
5518         for (m0 = m1; m0; m0 = m1) {
5519                 m1 = m0->m_nextpkt;
5520                 m0->m_nextpkt = 0;
5521                 if (error == 0) {
5522                         lwkt_reltoken(&pf_token);
5523                         error = (*ifp->if_output)(ifp, m0, sintosa(dst),
5524                                                   NULL);
5525                         lwkt_gettoken(&pf_token);
5526                 } else
5527                         m_freem(m0);
5528         }
5529
5530         if (error == 0)
5531                 ipstat.ips_fragmented++;
5532
5533 done:
5534         if (r->rt != PF_DUPTO)
5535                 *m = NULL;
5536         if (ro == &iproute && ro->ro_rt)
5537                 RTFREE(ro->ro_rt);
5538         return;
5539
5540 bad:
5541         m_freem(m0);
5542         goto done;
5543 }
5544 #endif /* INET */
5545
5546 #ifdef INET6
5547 void
5548 pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5549     struct pf_state *s, struct pf_pdesc *pd)
5550 {
5551         struct mbuf             *m0;
5552         struct route_in6         ip6route;
5553         struct route_in6        *ro;
5554         struct sockaddr_in6     *dst;
5555         struct ip6_hdr          *ip6;
5556         struct ifnet            *ifp = NULL;
5557         struct pf_addr           naddr;
5558         struct pf_src_node      *sn = NULL;
5559         int                      error = 0;
5560
5561         if (m == NULL || *m == NULL || r == NULL ||
5562             (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5563                 panic("pf_route6: invalid parameters");
5564
5565         if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
5566                 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
5567                 (*m)->m_pkthdr.pf.routed = 1;
5568         } else {
5569                 if ((*m)->m_pkthdr.pf.routed++ > 3) {
5570                         m0 = *m;
5571                         *m = NULL;
5572                         goto bad;
5573                 }
5574         }
5575
5576         if (r->rt == PF_DUPTO) {
5577                 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL)
5578                         return;
5579         } else {
5580                 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5581                         return;
5582                 m0 = *m;
5583         }
5584
5585         if (m0->m_len < sizeof(struct ip6_hdr)) {
5586                 DPFPRINTF(PF_DEBUG_URGENT,
5587                     ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n"));
5588                 goto bad;
5589         }
5590         ip6 = mtod(m0, struct ip6_hdr *);
5591
5592         ro = &ip6route;
5593         bzero((caddr_t)ro, sizeof(*ro));
5594         dst = (struct sockaddr_in6 *)&ro->ro_dst;
5595         dst->sin6_family = AF_INET6;
5596         dst->sin6_len = sizeof(*dst);
5597         dst->sin6_addr = ip6->ip6_dst;
5598
5599         /*
5600          * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags,
5601          * so make sure pf.flags is clear.
5602          *
5603          * Cheat. XXX why only in the v6 case???
5604          */
5605         if (r->rt == PF_FASTROUTE) {
5606                 m0->m_pkthdr.fw_flags |= PF_MBUF_TAGGED;
5607                 m0->m_pkthdr.pf.flags = 0;
5608                 /* XXX Re-Check when Upgrading to > 4.4 */
5609                 m0->m_pkthdr.pf.statekey = NULL;
5610                 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
5611                 return;
5612         }
5613
5614         if (TAILQ_EMPTY(&r->rpool.list)) {
5615                 DPFPRINTF(PF_DEBUG_URGENT,
5616                     ("pf_route6: TAILQ_EMPTY(&r->rpool.list)\n"));
5617                 goto bad;
5618         }
5619         if (s == NULL) {
5620                 pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src,
5621                     &naddr, NULL, &sn);
5622                 if (!PF_AZERO(&naddr, AF_INET6))
5623                         PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5624                             &naddr, AF_INET6);
5625                 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL;
5626         } else {
5627                 if (!PF_AZERO(&s->rt_addr, AF_INET6))
5628                         PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5629                             &s->rt_addr, AF_INET6);
5630                 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5631         }
5632         if (ifp == NULL)
5633                 goto bad;
5634
5635         if (oifp != ifp) {
5636                 if (pf_test6(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) {
5637                         goto bad;
5638                 } else if (m0 == NULL) {
5639                         goto done;
5640                 }
5641                 if (m0->m_len < sizeof(struct ip6_hdr)) {
5642                         DPFPRINTF(PF_DEBUG_URGENT,
5643                             ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n"));
5644                         goto bad;
5645                 }
5646                 ip6 = mtod(m0, struct ip6_hdr *);
5647         }
5648
5649         /*
5650          * If the packet is too large for the outgoing interface,
5651          * send back an icmp6 error.
5652          */
5653         if (IN6_IS_ADDR_LINKLOCAL(&dst->sin6_addr))
5654                 dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
5655         if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) {
5656                 error = nd6_output(ifp, ifp, m0, dst, NULL);
5657         } else {
5658                 in6_ifstat_inc(ifp, ifs6_in_toobig);
5659                 if (r->rt != PF_DUPTO)
5660                         icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
5661                 else
5662                         goto bad;
5663         }
5664
5665 done:
5666         if (r->rt != PF_DUPTO)
5667                 *m = NULL;
5668         return;
5669
5670 bad:
5671         m_freem(m0);
5672         goto done;
5673 }
5674 #endif /* INET6 */
5675
5676
5677 /*
5678  * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag
5679  *   off is the offset where the protocol header starts
5680  *   len is the total length of protocol header plus payload
5681  * returns 0 when the checksum is valid, otherwise returns 1.
5682  */
5683 /*
5684  * XXX
5685  * FreeBSD supports cksum offload for the following drivers.
5686  * em(4), gx(4), lge(4), nge(4), ti(4), xl(4)
5687  * If we can make full use of it we would outperform ipfw/ipfilter in
5688  * very heavy traffic.
5689  * I have not tested 'cause I don't have NICs that supports cksum offload.
5690  * (There might be problems. Typical phenomena would be
5691  *   1. No route message for UDP packet.
5692  *   2. No connection acceptance from external hosts regardless of rule set.)
5693  */
5694 int
5695 pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p,
5696     sa_family_t af)
5697 {
5698         u_int16_t sum = 0;
5699         int hw_assist = 0;
5700         struct ip *ip;
5701
5702         if (off < sizeof(struct ip) || len < sizeof(struct udphdr))
5703                 return (1);
5704         if (m->m_pkthdr.len < off + len)
5705                 return (1);
5706
5707         switch (p) {
5708         case IPPROTO_TCP:
5709         case IPPROTO_UDP:
5710                 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
5711                         if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) {
5712                                 sum = m->m_pkthdr.csum_data;
5713                         } else {
5714                                 ip = mtod(m, struct ip *);
5715                                 sum = in_pseudo(ip->ip_src.s_addr,
5716                                         ip->ip_dst.s_addr, htonl((u_short)len +
5717                                         m->m_pkthdr.csum_data + p));
5718                         }
5719                         sum ^= 0xffff;
5720                         ++hw_assist;
5721                 }
5722                 break;
5723         case IPPROTO_ICMP:
5724 #ifdef INET6
5725         case IPPROTO_ICMPV6:
5726 #endif /* INET6 */
5727                 break;
5728         default:
5729                 return (1);
5730         }
5731
5732         if (!hw_assist) {
5733                 switch (af) {
5734                 case AF_INET:
5735                         if (p == IPPROTO_ICMP) {
5736                                 if (m->m_len < off)
5737                                         return (1);
5738                                 m->m_data += off;
5739                                 m->m_len -= off;
5740                                 sum = in_cksum(m, len);
5741                                 m->m_data -= off;
5742                                 m->m_len += off;
5743                         } else {
5744                                 if (m->m_len < sizeof(struct ip))
5745                                         return (1);
5746                                 sum = in_cksum_range(m, p, off, len);
5747                                 if (sum == 0) {
5748                                         m->m_pkthdr.csum_flags |=
5749                                             (CSUM_DATA_VALID |
5750                                              CSUM_PSEUDO_HDR);
5751                                         m->m_pkthdr.csum_data = 0xffff;
5752                                 }
5753                         }
5754                         break;
5755 #ifdef INET6
5756                 case AF_INET6:
5757                         if (m->m_len < sizeof(struct ip6_hdr))
5758                                 return (1);
5759                         sum = in6_cksum(m, p, off, len);
5760                         /*
5761                          * XXX
5762                          * IPv6 H/W cksum off-load not supported yet!
5763                          *
5764                          * if (sum == 0) {
5765                          *      m->m_pkthdr.csum_flags |=
5766                          *          (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
5767                          *      m->m_pkthdr.csum_data = 0xffff;
5768                          *}
5769                          */
5770                         break;
5771 #endif /* INET6 */
5772                 default:
5773                         return (1);
5774                 }
5775         }
5776         if (sum) {
5777                 switch (p) {
5778                 case IPPROTO_TCP:
5779                         tcpstat.tcps_rcvbadsum++;
5780                         break;
5781                 case IPPROTO_UDP:
5782                         udpstat.udps_badsum++;
5783                         break;
5784                 case IPPROTO_ICMP:
5785                         icmpstat.icps_checksum++;
5786                         break;
5787 #ifdef INET6
5788                 case IPPROTO_ICMPV6:
5789                         icmp6stat.icp6s_checksum++;
5790                         break;
5791 #endif /* INET6 */
5792                 }
5793                 return (1);
5794         }
5795         return (0);
5796 }
5797
5798 struct pf_divert *
5799 pf_find_divert(struct mbuf *m)
5800 {
5801         struct m_tag    *mtag;
5802
5803         if ((mtag = m_tag_find(m, PACKET_TAG_PF_DIVERT, NULL)) == NULL)
5804                 return (NULL);
5805
5806         return ((struct pf_divert *)(mtag + 1));
5807 }
5808
5809 struct pf_divert *
5810 pf_get_divert(struct mbuf *m)
5811 {
5812         struct m_tag    *mtag;
5813
5814         if ((mtag = m_tag_find(m, PACKET_TAG_PF_DIVERT, NULL)) == NULL) {
5815                 mtag = m_tag_get(PACKET_TAG_PF_DIVERT, sizeof(struct pf_divert),
5816                     M_NOWAIT);
5817                 if (mtag == NULL)
5818                         return (NULL);
5819                 bzero(mtag + 1, sizeof(struct pf_divert));
5820                 m_tag_prepend(m, mtag);
5821         }
5822
5823         return ((struct pf_divert *)(mtag + 1));
5824 }
5825
5826 #ifdef INET
5827 int
5828 pf_test(int dir, struct ifnet *ifp, struct mbuf **m0,
5829     struct ether_header *eh, struct inpcb *inp)
5830 {
5831         struct pfi_kif          *kif;
5832         u_short                  action, reason = 0, log = 0;
5833         struct mbuf             *m = *m0;
5834         struct ip               *h = NULL;
5835         struct pf_rule          *a = NULL, *r = &pf_default_rule, *tr, *nr;
5836         struct pf_state         *s = NULL;
5837         struct pf_ruleset       *ruleset = NULL;
5838         struct pf_pdesc          pd;
5839         int                      off, dirndx, pqid = 0;
5840
5841         if (!pf_status.running)
5842                 return (PF_PASS);
5843
5844         memset(&pd, 0, sizeof(pd));
5845         if (ifp->if_type == IFT_CARP && ifp->if_carpdev)
5846                 kif = (struct pfi_kif *)ifp->if_carpdev->if_pf_kif;
5847         else
5848                 kif = (struct pfi_kif *)ifp->if_pf_kif;
5849
5850         if (kif == NULL) {
5851                 DPFPRINTF(PF_DEBUG_URGENT,
5852                     ("pf_test: kif == NULL, if_xname %s\n", ifp->if_xname));
5853                 return (PF_DROP);
5854         }
5855         if (kif->pfik_flags & PFI_IFLAG_SKIP)
5856                 return (PF_PASS);
5857
5858 #ifdef DIAGNOSTIC
5859         if ((m->m_flags & M_PKTHDR) == 0)
5860                 panic("non-M_PKTHDR is passed to pf_test");
5861 #endif /* DIAGNOSTIC */
5862
5863         if (m->m_pkthdr.len < (int)sizeof(*h)) {
5864                 action = PF_DROP;
5865                 REASON_SET(&reason, PFRES_SHORT);
5866                 log = 1;
5867                 goto done;
5868         }
5869
5870         /*
5871          * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags,
5872          * so make sure pf.flags is clear.
5873          */
5874         if (m->m_pkthdr.fw_flags & PF_MBUF_TAGGED)
5875                 return (PF_PASS);
5876         m->m_pkthdr.pf.flags = 0;
5877         /* Re-Check when updating to > 4.4 */
5878         m->m_pkthdr.pf.statekey = NULL;
5879
5880         /* We do IP header normalization and packet reassembly here */
5881         if (pf_normalize_ip(m0, dir, kif, &reason, &pd) != PF_PASS) {
5882                 action = PF_DROP;
5883                 goto done;
5884         }
5885         m = *m0;        /* pf_normalize messes with m0 */
5886         h = mtod(m, struct ip *);
5887
5888         off = h->ip_hl << 2;
5889         if (off < (int)sizeof(*h)) {
5890                 action = PF_DROP;
5891                 REASON_SET(&reason, PFRES_SHORT);
5892                 log = 1;
5893                 goto done;
5894         }
5895
5896         pd.src = (struct pf_addr *)&h->ip_src;
5897         pd.dst = (struct pf_addr *)&h->ip_dst;
5898         pd.sport = pd.dport = NULL;
5899         pd.ip_sum = &h->ip_sum;
5900         pd.proto_sum = NULL;
5901         pd.proto = h->ip_p;
5902         pd.dir = dir;
5903         pd.sidx = (dir == PF_IN) ? 0 : 1;
5904         pd.didx = (dir == PF_IN) ? 1 : 0;
5905         pd.af = AF_INET;
5906         pd.tos = h->ip_tos;
5907         pd.tot_len = h->ip_len;
5908         pd.eh = eh;
5909
5910         /* handle fragments that didn't get reassembled by normalization */
5911         if (h->ip_off & (IP_MF | IP_OFFMASK)) {
5912                 action = pf_test_fragment(&r, dir, kif, m, h,
5913                     &pd, &a, &ruleset);
5914                 goto done;
5915         }
5916
5917         switch (h->ip_p) {
5918
5919         case IPPROTO_TCP: {
5920                 struct tcphdr   th;
5921
5922                 pd.hdr.tcp = &th;
5923                 if (!pf_pull_hdr(m, off, &th, sizeof(th),
5924                     &action, &reason, AF_INET)) {
5925                         log = action != PF_PASS;
5926                         goto done;
5927                 }
5928                 pd.p_len = pd.tot_len - off - (th.th_off << 2);
5929                 if ((th.th_flags & TH_ACK) && pd.p_len == 0)
5930                         pqid = 1;
5931                 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
5932                 if (action == PF_DROP)
5933                         goto done;
5934                 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
5935                     &reason);
5936                 if (action == PF_PASS) {
5937                         pfsync_update_state(s);
5938                         r = s->rule.ptr;
5939                         a = s->anchor.ptr;
5940                         log = s->log;
5941                 } else if (s == NULL)
5942                         action = pf_test_rule(&r, &s, dir, kif,
5943                             m, off, h, &pd, &a, &ruleset, NULL, inp);
5944                 break;
5945         }
5946
5947         case IPPROTO_UDP: {
5948                 struct udphdr   uh;
5949
5950                 pd.hdr.udp = &uh;
5951                 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
5952                     &action, &reason, AF_INET)) {
5953                         log = action != PF_PASS;
5954                         goto done;
5955                 }
5956                 if (uh.uh_dport == 0 ||
5957                     ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
5958                     ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
5959                         action = PF_DROP;
5960                         REASON_SET(&reason, PFRES_SHORT);
5961                         goto done;
5962                 }
5963                 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
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_ICMP: {
5976                 struct icmp     ih;
5977
5978                 pd.hdr.icmp = &ih;
5979                 if (!pf_pull_hdr(m, off, &ih, ICMP_MINLEN,
5980                     &action, &reason, AF_INET)) {
5981                         log = action != PF_PASS;
5982                         goto done;
5983                 }
5984                 action = pf_test_state_icmp(&s, dir, kif, m, off, h, &pd,
5985                     &reason);
5986                 if (action == PF_PASS) {
5987                         pfsync_update_state(s);
5988                         r = s->rule.ptr;
5989                         a = s->anchor.ptr;
5990                         log = s->log;
5991                 } else if (s == NULL)
5992                         action = pf_test_rule(&r, &s, dir, kif,
5993                             m, off, h, &pd, &a, &ruleset, NULL, inp);
5994                 break;
5995         }
5996
5997         default:
5998                 action = pf_test_state_other(&s, dir, kif, m, &pd);
5999                 if (action == PF_PASS) {
6000                         pfsync_update_state(s);
6001                         r = s->rule.ptr;
6002                         a = s->anchor.ptr;
6003                         log = s->log;
6004                 } else if (s == NULL)
6005                         action = pf_test_rule(&r, &s, dir, kif, m, off, h,
6006                             &pd, &a, &ruleset, NULL, inp);
6007                 break;
6008         }
6009
6010 done:
6011         if (action == PF_PASS && h->ip_hl > 5 &&
6012             !((s && s->state_flags & PFSTATE_ALLOWOPTS) || r->allow_opts)) {
6013                 action = PF_DROP;
6014                 REASON_SET(&reason, PFRES_IPOPTIONS);
6015                 log = 1;
6016                 DPFPRINTF(PF_DEBUG_MISC,
6017                     ("pf: dropping packet with ip options\n"));
6018         }
6019
6020         if ((s && s->tag) || r->rtableid)
6021                 pf_tag_packet(m, s ? s->tag : 0, r->rtableid);
6022
6023 #if 0
6024         if (dir == PF_IN && s && s->key[PF_SK_STACK])
6025                 m->m_pkthdr.pf.statekey = s->key[PF_SK_STACK];
6026 #endif
6027
6028 #ifdef ALTQ
6029         if (action == PF_PASS && r->qid) {
6030                 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
6031                 if (pqid || (pd.tos & IPTOS_LOWDELAY))
6032                         m->m_pkthdr.pf.qid = r->pqid;
6033                 else
6034                         m->m_pkthdr.pf.qid = r->qid;
6035                 m->m_pkthdr.pf.ecn_af = AF_INET;
6036                 m->m_pkthdr.pf.hdr = h;
6037                 /* add connection hash for fairq */
6038                 if (s) {
6039                         /* for fairq */
6040                         m->m_pkthdr.pf.state_hash = s->hash;
6041                         m->m_pkthdr.pf.flags |= PF_TAG_STATE_HASHED;
6042                 }
6043         }
6044 #endif /* ALTQ */
6045
6046         /*
6047          * connections redirected to loopback should not match sockets
6048          * bound specifically to loopback due to security implications,
6049          * see tcp_input() and in_pcblookup_listen().
6050          */
6051         if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
6052             pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
6053             (s->nat_rule.ptr->action == PF_RDR ||
6054             s->nat_rule.ptr->action == PF_BINAT) &&
6055             (ntohl(pd.dst->v4.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET)
6056                 m->m_pkthdr.pf.flags |= PF_TAG_TRANSLATE_LOCALHOST;
6057
6058         if (dir == PF_IN && action == PF_PASS && r->divert.port) {
6059                 struct pf_divert *divert;
6060
6061                 if ((divert = pf_get_divert(m))) {
6062                         m->m_pkthdr.pf.flags |= PF_TAG_DIVERTED;
6063                         divert->port = r->divert.port;
6064                         divert->addr.ipv4 = r->divert.addr.v4;
6065                 }
6066         }
6067
6068         if (log) {
6069                 struct pf_rule *lr;
6070
6071                 if (s != NULL && s->nat_rule.ptr != NULL &&
6072                     s->nat_rule.ptr->log & PF_LOG_ALL)
6073                         lr = s->nat_rule.ptr;
6074                 else
6075                         lr = r;
6076                 PFLOG_PACKET(kif, h, m, AF_INET, dir, reason, lr, a, ruleset,
6077                     &pd);
6078         }
6079
6080         kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
6081         kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++;
6082
6083         if (action == PF_PASS || r->action == PF_DROP) {
6084                 dirndx = (dir == PF_OUT);
6085                 r->packets[dirndx]++;
6086                 r->bytes[dirndx] += pd.tot_len;
6087                 if (a != NULL) {
6088                         a->packets[dirndx]++;
6089                         a->bytes[dirndx] += pd.tot_len;
6090                 }
6091                 if (s != NULL) {
6092                         if (s->nat_rule.ptr != NULL) {
6093                                 s->nat_rule.ptr->packets[dirndx]++;
6094                                 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len;
6095                         }
6096                         if (s->src_node != NULL) {
6097                                 s->src_node->packets[dirndx]++;
6098                                 s->src_node->bytes[dirndx] += pd.tot_len;
6099                         }
6100                         if (s->nat_src_node != NULL) {
6101                                 s->nat_src_node->packets[dirndx]++;
6102                                 s->nat_src_node->bytes[dirndx] += pd.tot_len;
6103                         }
6104                         dirndx = (dir == s->direction) ? 0 : 1;
6105                         s->packets[dirndx]++;
6106                         s->bytes[dirndx] += pd.tot_len;
6107                 }
6108                 tr = r;
6109                 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
6110                 if (nr != NULL && r == &pf_default_rule)
6111                         tr = nr;
6112                 if (tr->src.addr.type == PF_ADDR_TABLE)
6113                         pfr_update_stats(tr->src.addr.p.tbl,
6114                             (s == NULL) ? pd.src :
6115                             &s->key[(s->direction == PF_IN)]->
6116                                 addr[(s->direction == PF_OUT)],
6117                             pd.af, pd.tot_len, dir == PF_OUT,
6118                             r->action == PF_PASS, tr->src.neg);
6119                 if (tr->dst.addr.type == PF_ADDR_TABLE)
6120                         pfr_update_stats(tr->dst.addr.p.tbl,
6121                             (s == NULL) ? pd.dst :
6122                             &s->key[(s->direction == PF_IN)]->
6123                                 addr[(s->direction == PF_IN)],
6124                             pd.af, pd.tot_len, dir == PF_OUT,
6125                             r->action == PF_PASS, tr->dst.neg);
6126         }
6127
6128
6129         if (action == PF_SYNPROXY_DROP) {
6130                 m_freem(*m0);
6131                 *m0 = NULL;
6132                 action = PF_PASS;
6133         } else if (r->rt)
6134                 /* pf_route can free the mbuf causing *m0 to become NULL */
6135                 pf_route(m0, r, dir, kif->pfik_ifp, s, &pd);
6136
6137         return (action);
6138 }
6139 #endif /* INET */
6140
6141 #ifdef INET6
6142 int
6143 pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0,
6144     struct ether_header *eh, struct inpcb *inp)
6145 {
6146         struct pfi_kif          *kif;
6147         u_short                  action, reason = 0, log = 0;
6148         struct mbuf             *m = *m0, *n = NULL;
6149         struct ip6_hdr          *h = NULL;
6150         struct pf_rule          *a = NULL, *r = &pf_default_rule, *tr, *nr;
6151         struct pf_state         *s = NULL;
6152         struct pf_ruleset       *ruleset = NULL;
6153         struct pf_pdesc          pd;
6154         int                      off, terminal = 0, dirndx, rh_cnt = 0;
6155
6156         if (!pf_status.running)
6157                 return (PF_PASS);
6158
6159         memset(&pd, 0, sizeof(pd));
6160         if (ifp->if_type == IFT_CARP && ifp->if_carpdev)
6161                 kif = (struct pfi_kif *)ifp->if_carpdev->if_pf_kif;
6162         else
6163                 kif = (struct pfi_kif *)ifp->if_pf_kif;
6164
6165         if (kif == NULL) {
6166                 DPFPRINTF(PF_DEBUG_URGENT,
6167                     ("pf_test6: kif == NULL, if_xname %s\n", ifp->if_xname));
6168                 return (PF_DROP);
6169         }
6170         if (kif->pfik_flags & PFI_IFLAG_SKIP)
6171                 return (PF_PASS);
6172
6173 #ifdef DIAGNOSTIC
6174         if ((m->m_flags & M_PKTHDR) == 0)
6175                 panic("non-M_PKTHDR is passed to pf_test6");
6176 #endif /* DIAGNOSTIC */
6177
6178         if (m->m_pkthdr.len < (int)sizeof(*h)) {
6179                 action = PF_DROP;
6180                 REASON_SET(&reason, PFRES_SHORT);
6181                 log = 1;
6182                 goto done;
6183         }
6184
6185         /*
6186          * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags,
6187          * so make sure pf.flags is clear.
6188          */
6189         if (m->m_pkthdr.fw_flags & PF_MBUF_TAGGED)
6190                 return (PF_PASS);
6191         m->m_pkthdr.pf.flags = 0;
6192         /* Re-Check when updating to > 4.4 */
6193         m->m_pkthdr.pf.statekey = NULL;
6194
6195         /* We do IP header normalization and packet reassembly here */
6196         if (pf_normalize_ip6(m0, dir, kif, &reason, &pd) != PF_PASS) {
6197                 action = PF_DROP;
6198                 goto done;
6199         }
6200         m = *m0;        /* pf_normalize messes with m0 */
6201         h = mtod(m, struct ip6_hdr *);
6202
6203 #if 1
6204         /*
6205          * we do not support jumbogram yet.  if we keep going, zero ip6_plen
6206          * will do something bad, so drop the packet for now.
6207          */
6208         if (htons(h->ip6_plen) == 0) {
6209                 action = PF_DROP;
6210                 REASON_SET(&reason, PFRES_NORM);        /*XXX*/
6211                 goto done;
6212         }
6213 #endif
6214
6215         pd.src = (struct pf_addr *)&h->ip6_src;
6216         pd.dst = (struct pf_addr *)&h->ip6_dst;
6217         pd.sport = pd.dport = NULL;
6218         pd.ip_sum = NULL;
6219         pd.proto_sum = NULL;
6220         pd.dir = dir;
6221         pd.sidx = (dir == PF_IN) ? 0 : 1;
6222         pd.didx = (dir == PF_IN) ? 1 : 0;
6223         pd.af = AF_INET6;
6224         pd.tos = 0;
6225         pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr);
6226         pd.eh = eh;
6227
6228         off = ((caddr_t)h - m->m_data) + sizeof(struct ip6_hdr);
6229         pd.proto = h->ip6_nxt;
6230         do {
6231                 switch (pd.proto) {
6232                 case IPPROTO_FRAGMENT:
6233                         action = pf_test_fragment(&r, dir, kif, m, h,
6234                             &pd, &a, &ruleset);
6235                         if (action == PF_DROP)
6236                                 REASON_SET(&reason, PFRES_FRAG);
6237                         goto done;
6238                 case IPPROTO_ROUTING: {
6239                         struct ip6_rthdr rthdr;
6240
6241                         if (rh_cnt++) {
6242                                 DPFPRINTF(PF_DEBUG_MISC,
6243                                     ("pf: IPv6 more than one rthdr\n"));
6244                                 action = PF_DROP;
6245                                 REASON_SET(&reason, PFRES_IPOPTIONS);
6246                                 log = 1;
6247                                 goto done;
6248                         }
6249                         if (!pf_pull_hdr(m, off, &rthdr, sizeof(rthdr), NULL,
6250                             &reason, pd.af)) {
6251                                 DPFPRINTF(PF_DEBUG_MISC,
6252                                     ("pf: IPv6 short rthdr\n"));
6253                                 action = PF_DROP;
6254                                 REASON_SET(&reason, PFRES_SHORT);
6255                                 log = 1;
6256                                 goto done;
6257                         }
6258                         if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) {
6259                                 DPFPRINTF(PF_DEBUG_MISC,
6260                                     ("pf: IPv6 rthdr0\n"));
6261                                 action = PF_DROP;
6262                                 REASON_SET(&reason, PFRES_IPOPTIONS);
6263                                 log = 1;
6264                                 goto done;
6265                         }
6266                         /* FALLTHROUGH */
6267                 }
6268                 case IPPROTO_AH:
6269                 case IPPROTO_HOPOPTS:
6270                 case IPPROTO_DSTOPTS: {
6271                         /* get next header and header length */
6272                         struct ip6_ext  opt6;
6273
6274                         if (!pf_pull_hdr(m, off, &opt6, sizeof(opt6),
6275                             NULL, &reason, pd.af)) {
6276                                 DPFPRINTF(PF_DEBUG_MISC,
6277                                     ("pf: IPv6 short opt\n"));
6278                                 action = PF_DROP;
6279                                 log = 1;
6280                                 goto done;
6281                         }
6282                         if (pd.proto == IPPROTO_AH)
6283                                 off += (opt6.ip6e_len + 2) * 4;
6284                         else
6285                                 off += (opt6.ip6e_len + 1) * 8;
6286                         pd.proto = opt6.ip6e_nxt;
6287                         /* goto the next header */
6288                         break;
6289                 }
6290                 default:
6291                         terminal++;
6292                         break;
6293                 }
6294         } while (!terminal);
6295
6296         /* if there's no routing header, use unmodified mbuf for checksumming */
6297         if (!n)
6298                 n = m;
6299
6300         switch (pd.proto) {
6301
6302         case IPPROTO_TCP: {
6303                 struct tcphdr   th;
6304
6305                 pd.hdr.tcp = &th;
6306                 if (!pf_pull_hdr(m, off, &th, sizeof(th),
6307                     &action, &reason, AF_INET6)) {
6308                         log = action != PF_PASS;
6309                         goto done;
6310                 }
6311                 pd.p_len = pd.tot_len - off - (th.th_off << 2);
6312                 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
6313                 if (action == PF_DROP)
6314                         goto done;
6315                 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
6316                     &reason);
6317                 if (action == PF_PASS) {
6318                         pfsync_update_state(s);
6319                         r = s->rule.ptr;
6320                         a = s->anchor.ptr;
6321                         log = s->log;
6322                 } else if (s == NULL)
6323                         action = pf_test_rule(&r, &s, dir, kif,
6324                             m, off, h, &pd, &a, &ruleset, NULL, inp);
6325                 break;
6326         }
6327
6328         case IPPROTO_UDP: {
6329                 struct udphdr   uh;
6330
6331                 pd.hdr.udp = &uh;
6332                 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
6333                     &action, &reason, AF_INET6)) {
6334                         log = action != PF_PASS;
6335                         goto done;
6336                 }
6337                 if (uh.uh_dport == 0 ||
6338                     ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
6339                     ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
6340                         action = PF_DROP;
6341                         REASON_SET(&reason, PFRES_SHORT);
6342                         goto done;
6343                 }
6344                 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
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_ICMPV6: {
6357                 struct icmp6_hdr        ih;
6358
6359                 pd.hdr.icmp6 = &ih;
6360                 if (!pf_pull_hdr(m, off, &ih, sizeof(ih),
6361                     &action, &reason, AF_INET6)) {
6362                         log = action != PF_PASS;
6363                         goto done;
6364                 }
6365                 action = pf_test_state_icmp(&s, dir, kif,
6366                     m, off, h, &pd, &reason);
6367                 if (action == PF_PASS) {
6368                         pfsync_update_state(s);
6369                         r = s->rule.ptr;
6370                         a = s->anchor.ptr;
6371                         log = s->log;
6372                 } else if (s == NULL)
6373                         action = pf_test_rule(&r, &s, dir, kif,
6374                             m, off, h, &pd, &a, &ruleset, NULL, inp);
6375                 break;
6376         }
6377
6378         default:
6379                 action = pf_test_state_other(&s, dir, kif, m, &pd);
6380                 if (action == PF_PASS) {
6381                         pfsync_update_state(s);
6382                         r = s->rule.ptr;
6383                         a = s->anchor.ptr;
6384                         log = s->log;
6385                 } else if (s == NULL)
6386                         action = pf_test_rule(&r, &s, dir, kif, m, off, h,
6387                             &pd, &a, &ruleset, NULL, inp);
6388                 break;
6389         }
6390
6391 done:
6392         if (n != m) {
6393                 m_freem(n);
6394                 n = NULL;
6395         }
6396
6397         /* handle dangerous IPv6 extension headers. */
6398         if (action == PF_PASS && rh_cnt &&
6399             !((s && s->state_flags & PFSTATE_ALLOWOPTS) || r->allow_opts)) {
6400                 action = PF_DROP;
6401                 REASON_SET(&reason, PFRES_IPOPTIONS);
6402                 log = 1;
6403                 DPFPRINTF(PF_DEBUG_MISC,
6404                     ("pf: dropping packet with dangerous v6 headers\n"));
6405         }
6406
6407         if ((s && s->tag) || r->rtableid)
6408                 pf_tag_packet(m, s ? s->tag : 0, r->rtableid);
6409
6410 #if 0
6411         if (dir == PF_IN && s && s->key[PF_SK_STACK])
6412                 m->m_pkthdr.pf.statekey = s->key[PF_SK_STACK];
6413 #endif
6414
6415 #ifdef ALTQ
6416         if (action == PF_PASS && r->qid) {
6417                 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
6418                 if (pd.tos & IPTOS_LOWDELAY)
6419                         m->m_pkthdr.pf.qid = r->pqid;
6420                 else
6421                         m->m_pkthdr.pf.qid = r->qid;
6422                 m->m_pkthdr.pf.ecn_af = AF_INET6;
6423                 m->m_pkthdr.pf.hdr = h;
6424                 if (s) {
6425                         /* for fairq */
6426                         m->m_pkthdr.pf.state_hash = s->hash;
6427                         m->m_pkthdr.pf.flags |= PF_TAG_STATE_HASHED;
6428                 }
6429         }
6430 #endif /* ALTQ */
6431
6432         if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
6433             pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
6434             (s->nat_rule.ptr->action == PF_RDR ||
6435             s->nat_rule.ptr->action == PF_BINAT) &&
6436             IN6_IS_ADDR_LOOPBACK(&pd.dst->v6))
6437                 m->m_pkthdr.pf.flags |= PF_TAG_TRANSLATE_LOCALHOST;
6438
6439         if (dir == PF_IN && action == PF_PASS && r->divert.port) {
6440                 struct pf_divert *divert;
6441
6442                 if ((divert = pf_get_divert(m))) {
6443                         m->m_pkthdr.pf.flags |= PF_TAG_DIVERTED;
6444                         divert->port = r->divert.port;
6445                         divert->addr.ipv6 = r->divert.addr.v6;
6446                 }
6447         }
6448
6449         if (log) {
6450                 struct pf_rule *lr;
6451
6452                 if (s != NULL && s->nat_rule.ptr != NULL &&
6453                     s->nat_rule.ptr->log & PF_LOG_ALL)
6454                         lr = s->nat_rule.ptr;
6455                 else
6456                         lr = r;
6457                 PFLOG_PACKET(kif, h, m, AF_INET6, dir, reason, lr, a, ruleset,
6458                     &pd);
6459         }
6460
6461         kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
6462         kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++;
6463
6464         if (action == PF_PASS || r->action == PF_DROP) {
6465                 dirndx = (dir == PF_OUT);
6466                 r->packets[dirndx]++;
6467                 r->bytes[dirndx] += pd.tot_len;
6468                 if (a != NULL) {
6469                         a->packets[dirndx]++;
6470                         a->bytes[dirndx] += pd.tot_len;
6471                 }
6472                 if (s != NULL) {
6473                         if (s->nat_rule.ptr != NULL) {
6474                                 s->nat_rule.ptr->packets[dirndx]++;
6475                                 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len;
6476                         }
6477                         if (s->src_node != NULL) {
6478                                 s->src_node->packets[dirndx]++;
6479                                 s->src_node->bytes[dirndx] += pd.tot_len;
6480                         }
6481                         if (s->nat_src_node != NULL) {
6482                                 s->nat_src_node->packets[dirndx]++;
6483                                 s->nat_src_node->bytes[dirndx] += pd.tot_len;
6484                         }
6485                         dirndx = (dir == s->direction) ? 0 : 1;
6486                         s->packets[dirndx]++;
6487                         s->bytes[dirndx] += pd.tot_len;
6488                 }
6489                 tr = r;
6490                 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
6491                 if (nr != NULL && r == &pf_default_rule)
6492                         tr = nr;
6493                 if (tr->src.addr.type == PF_ADDR_TABLE)
6494                         pfr_update_stats(tr->src.addr.p.tbl,
6495                             (s == NULL) ? pd.src :
6496                             &s->key[(s->direction == PF_IN)]->addr[0],
6497                             pd.af, pd.tot_len, dir == PF_OUT,
6498                             r->action == PF_PASS, tr->src.neg);
6499                 if (tr->dst.addr.type == PF_ADDR_TABLE)
6500                         pfr_update_stats(tr->dst.addr.p.tbl,
6501                             (s == NULL) ? pd.dst :
6502                             &s->key[(s->direction == PF_IN)]->addr[1],
6503                             pd.af, pd.tot_len, dir == PF_OUT,
6504                             r->action == PF_PASS, tr->dst.neg);
6505         }
6506
6507
6508         if (action == PF_SYNPROXY_DROP) {
6509                 m_freem(*m0);
6510                 *m0 = NULL;
6511                 action = PF_PASS;
6512         } else if (r->rt)
6513                 /* pf_route6 can free the mbuf causing *m0 to become NULL */
6514                 pf_route6(m0, r, dir, kif->pfik_ifp, s, &pd);
6515
6516         return (action);
6517 }
6518 #endif /* INET6 */
6519
6520 int
6521 pf_check_congestion(struct ifqueue *ifq)
6522 {
6523                 return (0);
6524 }