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 <sys/proc.h>
  55 #include <sys/kthread.h>
  56
  57 #include <machine/inttypes.h>
  58
  59 #include <sys/md5.h>
  60
  61 #include <net/if.h>
  62 #include <net/if_types.h>
  63 #include <net/bpf.h>
  64 #include <net/netisr2.h>
  65 #include <net/route.h>
  66
  67 #include <netinet/in.h>
  68 #include <netinet/in_var.h>
  69 #include <netinet/in_systm.h>
  70 #include <netinet/ip.h>
  71 #include <netinet/ip_var.h>
  72 #include <netinet/tcp.h>
  73 #include <netinet/tcp_seq.h>
  74 #include <netinet/udp.h>
  75 #include <netinet/ip_icmp.h>
  76 #include <netinet/in_pcb.h>
  77 #include <netinet/tcp_timer.h>
  78 #include <netinet/tcp_var.h>
  79 #include <netinet/udp_var.h>
  80 #include <netinet/icmp_var.h>
  81 #include <netinet/if_ether.h>
  82
  83 #include <net/pf/pfvar.h>
  84 #include <net/pf/if_pflog.h>
  85
  86 #include <net/pf/if_pfsync.h>
  87
  88 #ifdef INET6
  89 #include <netinet/ip6.h>
  90 #include <netinet/icmp6.h>
  91 #include <netinet6/nd6.h>
  92 #include <netinet6/ip6_var.h>
  93 #include <netinet6/in6_pcb.h>
  94 #endif /* INET6 */
  95
  96 #include <sys/in_cksum.h>
  97 #include <sys/ucred.h>
  98 #include <machine/limits.h>
  99 #include <sys/msgport2.h>
 100 #include <net/netmsg2.h>
 101
 102 extern int ip_optcopy(struct ip *, struct ip *);
 103 extern int debug_pfugidhack;
 104
 105 struct lwkt_token pf_token = LWKT_TOKEN_INITIALIZER(pf_token);
 106
 107 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) kprintf x
 108
 109 /*
 110  * Global variables
 111  */
 112
 113 /* mask radix tree */
 114 struct radix_node_head  *pf_maskhead;
 115
 116 /* state tables */
 117 struct pf_state_tree     pf_statetbl;
 118
 119 struct pf_altqqueue      pf_altqs[2];
 120 struct pf_palist         pf_pabuf;
 121 struct pf_altqqueue     *pf_altqs_active;
 122 struct pf_altqqueue     *pf_altqs_inactive;
 123 struct pf_status         pf_status;
 124
 125 u_int32_t                ticket_altqs_active;
 126 u_int32_t                ticket_altqs_inactive;
 127 int                      altqs_inactive_open;
 128 u_int32_t                ticket_pabuf;
 129
 130 MD5_CTX                  pf_tcp_secret_ctx;
 131 u_char                   pf_tcp_secret[16];
 132 int                      pf_tcp_secret_init;
 133 int                      pf_tcp_iss_off;
 134
 135 struct pf_anchor_stackframe {
 136         struct pf_ruleset                       *rs;
 137         struct pf_rule                          *r;
 138         struct pf_anchor_node                   *parent;
 139         struct pf_anchor                        *child;
 140 } pf_anchor_stack[64];
 141
 142 struct malloc_type       *pf_src_tree_pl, *pf_rule_pl, *pf_pooladdr_pl;
 143 struct malloc_type       *pf_state_pl, *pf_state_key_pl, *pf_state_item_pl;
 144 struct malloc_type       *pf_altq_pl;
 145
 146 void                     pf_print_host(struct pf_addr *, u_int16_t, u_int8_t);
 147
 148 void                     pf_init_threshold(struct pf_threshold *, u_int32_t,
 149                             u_int32_t);
 150 void                     pf_add_threshold(struct pf_threshold *);
 151 int                      pf_check_threshold(struct pf_threshold *);
 152
 153 void                     pf_change_ap(struct pf_addr *, u_int16_t *,
 154                             u_int16_t *, u_int16_t *, struct pf_addr *,
 155                             u_int16_t, u_int8_t, sa_family_t);
 156 int                      pf_modulate_sack(struct mbuf *, int, struct pf_pdesc *,
 157                             struct tcphdr *, struct pf_state_peer *);
 158 #ifdef INET6
 159 void                     pf_change_a6(struct pf_addr *, u_int16_t *,
 160                             struct pf_addr *, u_int8_t);
 161 #endif /* INET6 */
 162 void                     pf_change_icmp(struct pf_addr *, u_int16_t *,
 163                             struct pf_addr *, struct pf_addr *, u_int16_t,
 164                             u_int16_t *, u_int16_t *, u_int16_t *,
 165                             u_int16_t *, u_int8_t, sa_family_t);
 166 void                     pf_send_tcp(const struct pf_rule *, sa_family_t,
 167                             const struct pf_addr *, const struct pf_addr *,
 168                             u_int16_t, u_int16_t, u_int32_t, u_int32_t,
 169                             u_int8_t, u_int16_t, u_int16_t, u_int8_t, int,
 170                             u_int16_t, struct ether_header *, struct ifnet *);
 171 void                     pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t,
 172                             sa_family_t, struct pf_rule *);
 173 struct pf_rule          *pf_match_translation(struct pf_pdesc *, struct mbuf *,
 174                             int, int, struct pfi_kif *,
 175                             struct pf_addr *, u_int16_t, struct pf_addr *,
 176                             u_int16_t, int);
 177 struct pf_rule          *pf_get_translation(struct pf_pdesc *, struct mbuf *,
 178                             int, int, struct pfi_kif *, struct pf_src_node **,
 179                             struct pf_state_key **, struct pf_state_key **,
 180                             struct pf_state_key **, struct pf_state_key **,
 181                             struct pf_addr *, struct pf_addr *,
 182                             u_int16_t, u_int16_t);
 183 void                     pf_detach_state(struct pf_state *);
 184 int                      pf_state_key_setup(struct pf_pdesc *, struct pf_rule *,
 185                             struct pf_state_key **, struct pf_state_key **,
 186                             struct pf_state_key **, struct pf_state_key **,
 187                             struct pf_addr *, struct pf_addr *,
 188                             u_int16_t, u_int16_t);
 189 void                     pf_state_key_detach(struct pf_state *, int);
 190 u_int32_t                pf_tcp_iss(struct pf_pdesc *);
 191 int                      pf_test_rule(struct pf_rule **, struct pf_state **,
 192                             int, struct pfi_kif *, struct mbuf *, int,
 193                             void *, struct pf_pdesc *, struct pf_rule **,
 194                             struct pf_ruleset **, struct ifqueue *, struct inpcb *);
 195 static __inline int      pf_create_state(struct pf_rule *, struct pf_rule *,
 196                             struct pf_rule *, struct pf_pdesc *,
 197                             struct pf_src_node *, struct pf_state_key *,
 198                             struct pf_state_key *, struct pf_state_key *,
 199                             struct pf_state_key *, struct mbuf *, int,
 200                             u_int16_t, u_int16_t, int *, struct pfi_kif *,
 201                             struct pf_state **, int, u_int16_t, u_int16_t,
 202                             int);
 203 int                      pf_test_fragment(struct pf_rule **, int,
 204                             struct pfi_kif *, struct mbuf *, void *,
 205                             struct pf_pdesc *, struct pf_rule **,
 206                             struct pf_ruleset **);
 207 int                      pf_tcp_track_full(struct pf_state_peer *,
 208                             struct pf_state_peer *, struct pf_state **,
 209                             struct pfi_kif *, struct mbuf *, int,
 210                             struct pf_pdesc *, u_short *, int *);
 211 int                     pf_tcp_track_sloppy(struct pf_state_peer *,
 212                             struct pf_state_peer *, struct pf_state **,
 213                             struct pf_pdesc *, u_short *);
 214 int                      pf_test_state_tcp(struct pf_state **, int,
 215                             struct pfi_kif *, struct mbuf *, int,
 216                             void *, struct pf_pdesc *, u_short *);
 217 int                      pf_test_state_udp(struct pf_state **, int,
 218                             struct pfi_kif *, struct mbuf *, int,
 219                             void *, struct pf_pdesc *);
 220 int                      pf_test_state_icmp(struct pf_state **, int,
 221                             struct pfi_kif *, struct mbuf *, int,
 222                             void *, struct pf_pdesc *, u_short *);
 223 int                      pf_test_state_other(struct pf_state **, int,
 224                             struct pfi_kif *, struct mbuf *, struct pf_pdesc *);
 225 void                     pf_step_into_anchor(int *, struct pf_ruleset **, int,
 226                             struct pf_rule **, struct pf_rule **, int *);
 227 int                      pf_step_out_of_anchor(int *, struct pf_ruleset **,
 228                              int, struct pf_rule **, struct pf_rule **,
 229                              int *);
 230 void                     pf_hash(struct pf_addr *, struct pf_addr *,
 231                             struct pf_poolhashkey *, sa_family_t);
 232 int                      pf_map_addr(u_int8_t, struct pf_rule *,
 233                             struct pf_addr *, struct pf_addr *,
 234                             struct pf_addr *, struct pf_src_node **);
 235 int                      pf_get_sport(sa_family_t, u_int8_t, struct pf_rule *,
 236                             struct pf_addr *, struct pf_addr *, u_int16_t,
 237                             struct pf_addr *, u_int16_t*, u_int16_t, u_int16_t,
 238                             struct pf_src_node **);
 239 void                     pf_route(struct mbuf **, struct pf_rule *, int,
 240                             struct ifnet *, struct pf_state *,
 241                             struct pf_pdesc *);
 242 void                     pf_route6(struct mbuf **, struct pf_rule *, int,
 243                             struct ifnet *, struct pf_state *,
 244                             struct pf_pdesc *);
 245 u_int8_t                 pf_get_wscale(struct mbuf *, int, u_int16_t,
 246                             sa_family_t);
 247 u_int16_t                pf_get_mss(struct mbuf *, int, u_int16_t,
 248                             sa_family_t);
 249 u_int16_t                pf_calc_mss(struct pf_addr *, sa_family_t,
 250                                 u_int16_t);
 251 void                     pf_set_rt_ifp(struct pf_state *,
 252                             struct pf_addr *);
 253 int                      pf_check_proto_cksum(struct mbuf *, int, int,
 254                             u_int8_t, sa_family_t);
 255 struct pf_divert        *pf_get_divert(struct mbuf *);
 256 void                     pf_print_state_parts(struct pf_state *,
 257                             struct pf_state_key *, struct pf_state_key *);
 258 int                      pf_addr_wrap_neq(struct pf_addr_wrap *,
 259                             struct pf_addr_wrap *);
 260 struct pf_state         *pf_find_state(struct pfi_kif *,
 261                             struct pf_state_key_cmp *, u_int, struct mbuf *);
 262 int                      pf_src_connlimit(struct pf_state **);
 263 int                      pf_check_congestion(struct ifqueue *);
 264
 265 extern int pf_end_threads;
 266
 267 struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] = {
 268         { &pf_state_pl, PFSTATE_HIWAT },
 269         { &pf_src_tree_pl, PFSNODE_HIWAT },
 270         { &pf_frent_pl, PFFRAG_FRENT_HIWAT },
 271         { &pfr_ktable_pl, PFR_KTABLE_HIWAT },
 272         { &pfr_kentry_pl, PFR_KENTRY_HIWAT }
 273 };
 274
 275 #define STATE_LOOKUP(i, k, d, s, m)                                     \
 276         do {                                                            \
 277                 s = pf_find_state(i, k, d, m);                  \
 278                 if (s == NULL || (s)->timeout == PFTM_PURGE)            \
 279                         return (PF_DROP);                               \
 280                 if (d == PF_OUT &&                                      \
 281                     (((s)->rule.ptr->rt == PF_ROUTETO &&                \
 282                     (s)->rule.ptr->direction == PF_OUT) ||              \
 283                     ((s)->rule.ptr->rt == PF_REPLYTO &&                 \
 284                     (s)->rule.ptr->direction == PF_IN)) &&              \
 285                     (s)->rt_kif != NULL &&                              \
 286                     (s)->rt_kif != i)                                   \
 287                         return (PF_PASS);                               \
 288         } while (0)
 289
 290 #define BOUND_IFACE(r, k) \
 291         ((r)->rule_flag & PFRULE_IFBOUND) ? (k) : pfi_all
 292
 293 #define STATE_INC_COUNTERS(s)                           \
 294         do {                                            \
 295                 s->rule.ptr->states_cur++;              \
 296                 s->rule.ptr->states_tot++;              \
 297                 if (s->anchor.ptr != NULL) {            \
 298                         s->anchor.ptr->states_cur++;    \
 299                         s->anchor.ptr->states_tot++;    \
 300                 }                                       \
 301                 if (s->nat_rule.ptr != NULL) {          \
 302                         s->nat_rule.ptr->states_cur++;  \
 303                         s->nat_rule.ptr->states_tot++;  \
 304                 }                                       \
 305         } while (0)
 306
 307 #define STATE_DEC_COUNTERS(s)                           \
 308         do {                                            \
 309                 if (s->nat_rule.ptr != NULL)            \
 310                         s->nat_rule.ptr->states_cur--;  \
 311                 if (s->anchor.ptr != NULL)              \
 312                         s->anchor.ptr->states_cur--;    \
 313                 s->rule.ptr->states_cur--;              \
 314         } while (0)
 315
 316 static MALLOC_DEFINE(M_PFSTATEPL, "pfstatepl", "pf state pool list");
 317 static MALLOC_DEFINE(M_PFSRCTREEPL, "pfsrctpl", "pf source tree pool list");
 318 static MALLOC_DEFINE(M_PFSTATEKEYPL, "pfstatekeypl", "pf state key pool list");
 319 static MALLOC_DEFINE(M_PFSTATEITEMPL, "pfstateitempl", "pf state item pool list");
 320
 321 static __inline int pf_src_compare(struct pf_src_node *, struct pf_src_node *);
 322 static __inline int pf_state_compare_key(struct pf_state_key *,
 323         struct pf_state_key *);
 324 static __inline int pf_state_compare_id(struct pf_state *,
 325         struct pf_state *);
 326
 327 struct pf_src_tree tree_src_tracking;
 328
 329 struct pf_state_tree_id tree_id;
 330 struct pf_state_queue state_list;
 331
 332 RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare);
 333 RB_GENERATE(pf_state_tree, pf_state_key, entry, pf_state_compare_key);
 334 RB_GENERATE(pf_state_tree_id, pf_state,
 335     entry_id, pf_state_compare_id);
 336
 337 static __inline int
 338 pf_src_compare(struct pf_src_node *a, struct pf_src_node *b)
 339 {
 340         int     diff;
 341
 342         if (a->rule.ptr > b->rule.ptr)
 343                 return (1);
 344         if (a->rule.ptr < b->rule.ptr)
 345                 return (-1);
 346         if ((diff = a->af - b->af) != 0)
 347                 return (diff);
 348         switch (a->af) {
 349 #ifdef INET
 350         case AF_INET:
 351                 if (a->addr.addr32[0] > b->addr.addr32[0])
 352                         return (1);
 353                 if (a->addr.addr32[0] < b->addr.addr32[0])
 354                         return (-1);
 355                 break;
 356 #endif /* INET */
 357 #ifdef INET6
 358         case AF_INET6:
 359                 if (a->addr.addr32[3] > b->addr.addr32[3])
 360                         return (1);
 361                 if (a->addr.addr32[3] < b->addr.addr32[3])
 362                         return (-1);
 363                 if (a->addr.addr32[2] > b->addr.addr32[2])
 364                         return (1);
 365                 if (a->addr.addr32[2] < b->addr.addr32[2])
 366                         return (-1);
 367                 if (a->addr.addr32[1] > b->addr.addr32[1])
 368                         return (1);
 369                 if (a->addr.addr32[1] < b->addr.addr32[1])
 370                         return (-1);
 371                 if (a->addr.addr32[0] > b->addr.addr32[0])
 372                         return (1);
 373                 if (a->addr.addr32[0] < b->addr.addr32[0])
 374                         return (-1);
 375                 break;
 376 #endif /* INET6 */
 377         }
 378         return (0);
 379 }
 380
 381 u_int32_t
 382 pf_state_hash(struct pf_state_key *sk)
 383 {
 384         u_int32_t hv = (u_int32_t)(((intptr_t)sk >> 6) ^ ((intptr_t)sk >> 15));
 385         if (hv == 0)    /* disallow 0 */
 386                 hv = 1;
 387         return(hv);
 388 }
 389
 390 #ifdef INET6
 391 void
 392 pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af)
 393 {
 394         switch (af) {
 395 #ifdef INET
 396         case AF_INET:
 397                 dst->addr32[0] = src->addr32[0];
 398                 break;
 399 #endif /* INET */
 400         case AF_INET6:
 401                 dst->addr32[0] = src->addr32[0];
 402                 dst->addr32[1] = src->addr32[1];
 403                 dst->addr32[2] = src->addr32[2];
 404                 dst->addr32[3] = src->addr32[3];
 405                 break;
 406         }
 407 }
 408 #endif /* INET6 */
 409
 410 void
 411 pf_init_threshold(struct pf_threshold *threshold,
 412     u_int32_t limit, u_int32_t seconds)
 413 {
 414         threshold->limit = limit * PF_THRESHOLD_MULT;
 415         threshold->seconds = seconds;
 416         threshold->count = 0;
 417         threshold->last = time_second;
 418 }
 419
 420 void
 421 pf_add_threshold(struct pf_threshold *threshold)
 422 {
 423         u_int32_t t = time_second, diff = t - threshold->last;
 424
 425         if (diff >= threshold->seconds)
 426                 threshold->count = 0;
 427         else
 428                 threshold->count -= threshold->count * diff /
 429                     threshold->seconds;
 430         threshold->count += PF_THRESHOLD_MULT;
 431         threshold->last = t;
 432 }
 433
 434 int
 435 pf_check_threshold(struct pf_threshold *threshold)
 436 {
 437         return (threshold->count > threshold->limit);
 438 }
 439
 440 int
 441 pf_src_connlimit(struct pf_state **state)
 442 {
 443         int bad = 0;
 444
 445         (*state)->src_node->conn++;
 446         (*state)->src.tcp_est = 1;
 447         pf_add_threshold(&(*state)->src_node->conn_rate);
 448
 449         if ((*state)->rule.ptr->max_src_conn &&
 450             (*state)->rule.ptr->max_src_conn <
 451             (*state)->src_node->conn) {
 452                 pf_status.lcounters[LCNT_SRCCONN]++;
 453                 bad++;
 454         }
 455
 456         if ((*state)->rule.ptr->max_src_conn_rate.limit &&
 457             pf_check_threshold(&(*state)->src_node->conn_rate)) {
 458                 pf_status.lcounters[LCNT_SRCCONNRATE]++;
 459                 bad++;
 460         }
 461
 462         if (!bad)
 463                 return (0);
 464
 465         if ((*state)->rule.ptr->overload_tbl) {
 466                 struct pfr_addr p;
 467                 u_int32_t       killed = 0;
 468
 469                 pf_status.lcounters[LCNT_OVERLOAD_TABLE]++;
 470                 if (pf_status.debug >= PF_DEBUG_MISC) {
 471                         kprintf("pf_src_connlimit: blocking address ");
 472                         pf_print_host(&(*state)->src_node->addr, 0,
 473                             (*state)->key[PF_SK_WIRE]->af);
 474                 }
 475
 476                 bzero(&p, sizeof(p));
 477                 p.pfra_af = (*state)->key[PF_SK_WIRE]->af;
 478                 switch ((*state)->key[PF_SK_WIRE]->af) {
 479 #ifdef INET
 480                 case AF_INET:
 481                         p.pfra_net = 32;
 482                         p.pfra_ip4addr = (*state)->src_node->addr.v4;
 483                         break;
 484 #endif /* INET */
 485 #ifdef INET6
 486                 case AF_INET6:
 487                         p.pfra_net = 128;
 488                         p.pfra_ip6addr = (*state)->src_node->addr.v6;
 489                         break;
 490 #endif /* INET6 */
 491                 }
 492
 493                 pfr_insert_kentry((*state)->rule.ptr->overload_tbl,
 494                     &p, time_second);
 495
 496                 /* kill existing states if that's required. */
 497                 if ((*state)->rule.ptr->flush) {
 498                         struct pf_state_key *sk;
 499                         struct pf_state *st;
 500
 501                         pf_status.lcounters[LCNT_OVERLOAD_FLUSH]++;
 502                         RB_FOREACH(st, pf_state_tree_id, &tree_id) {
 503                                 sk = st->key[PF_SK_WIRE];
 504                                 /*
 505                                  * Kill states from this source.  (Only those
 506                                  * from the same rule if PF_FLUSH_GLOBAL is not
 507                                  * set)
 508                                  */
 509                                 if (sk->af ==
 510                                     (*state)->key[PF_SK_WIRE]->af &&
 511                                     (((*state)->direction == PF_OUT &&
 512                                     PF_AEQ(&(*state)->src_node->addr,
 513                                         &sk->addr[0], sk->af)) ||
 514                                     ((*state)->direction == PF_IN &&
 515                                     PF_AEQ(&(*state)->src_node->addr,
 516                                         &sk->addr[1], sk->af))) &&
 517                                     ((*state)->rule.ptr->flush &
 518                                     PF_FLUSH_GLOBAL ||
 519                                     (*state)->rule.ptr == st->rule.ptr)) {
 520                                         st->timeout = PFTM_PURGE;
 521                                         st->src.state = st->dst.state =
 522                                             TCPS_CLOSED;
 523                                         killed++;
 524                                 }
 525                         }
 526                         if (pf_status.debug >= PF_DEBUG_MISC)
 527                                 kprintf(", %u states killed", killed);
 528                 }
 529                 if (pf_status.debug >= PF_DEBUG_MISC)
 530                         kprintf("\n");
 531         }
 532
 533         /* kill this state */
 534         (*state)->timeout = PFTM_PURGE;
 535         (*state)->src.state = (*state)->dst.state = TCPS_CLOSED;
 536         return (1);
 537 }
 538
 539 int
 540 pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule,
 541     struct pf_addr *src, sa_family_t af)
 542 {
 543         struct pf_src_node      k;
 544
 545         if (*sn == NULL) {
 546                 k.af = af;
 547                 PF_ACPY(&k.addr, src, af);
 548                 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
 549                     rule->rpool.opts & PF_POOL_STICKYADDR)
 550                         k.rule.ptr = rule;
 551                 else
 552                         k.rule.ptr = NULL;
 553                 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
 554                 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
 555         }
 556         if (*sn == NULL) {
 557                 if (!rule->max_src_nodes ||
 558                     rule->src_nodes < rule->max_src_nodes)
 559                         (*sn) = kmalloc(sizeof(struct pf_src_node), M_PFSRCTREEPL, M_NOWAIT|M_ZERO);
 560                 else
 561                         pf_status.lcounters[LCNT_SRCNODES]++;
 562                 if ((*sn) == NULL)
 563                         return (-1);
 564
 565                 pf_init_threshold(&(*sn)->conn_rate,
 566                     rule->max_src_conn_rate.limit,
 567                     rule->max_src_conn_rate.seconds);
 568
 569                 (*sn)->af = af;
 570                 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
 571                     rule->rpool.opts & PF_POOL_STICKYADDR)
 572                         (*sn)->rule.ptr = rule;
 573                 else
 574                         (*sn)->rule.ptr = NULL;
 575                 PF_ACPY(&(*sn)->addr, src, af);
 576                 if (RB_INSERT(pf_src_tree,
 577                     &tree_src_tracking, *sn) != NULL) {
 578                         if (pf_status.debug >= PF_DEBUG_MISC) {
 579                                 kprintf("pf: src_tree insert failed: ");
 580                                 pf_print_host(&(*sn)->addr, 0, af);
 581                                 kprintf("\n");
 582                         }
 583                         kfree(*sn, M_PFSRCTREEPL);
 584                         return (-1);
 585                 }
 586                 (*sn)->creation = time_second;
 587                 (*sn)->ruletype = rule->action;
 588                 if ((*sn)->rule.ptr != NULL)
 589                         (*sn)->rule.ptr->src_nodes++;
 590                 pf_status.scounters[SCNT_SRC_NODE_INSERT]++;
 591                 pf_status.src_nodes++;
 592         } else {
 593                 if (rule->max_src_states &&
 594                     (*sn)->states >= rule->max_src_states) {
 595                         pf_status.lcounters[LCNT_SRCSTATES]++;
 596                         return (-1);
 597                 }
 598         }
 599         return (0);
 600 }
 601
 602 /* state table stuff */
 603
 604 static __inline int
 605 pf_state_compare_key(struct pf_state_key *a, struct pf_state_key *b)
 606 {
 607         int     diff;
 608
 609         if ((diff = a->proto - b->proto) != 0)
 610                 return (diff);
 611         if ((diff = a->af - b->af) != 0)
 612                 return (diff);
 613         switch (a->af) {
 614 #ifdef INET
 615         case AF_INET:
 616                 if (a->addr[0].addr32[0] > b->addr[0].addr32[0])
 617                         return (1);
 618                 if (a->addr[0].addr32[0] < b->addr[0].addr32[0])
 619                         return (-1);
 620                 if (a->addr[1].addr32[0] > b->addr[1].addr32[0])
 621                         return (1);
 622                 if (a->addr[1].addr32[0] < b->addr[1].addr32[0])
 623                         return (-1);
 624                 break;
 625 #endif /* INET */
 626 #ifdef INET6
 627         case AF_INET6:
 628                 if (a->addr[0].addr32[3] > b->addr[0].addr32[3])
 629                         return (1);
 630                 if (a->addr[0].addr32[3] < b->addr[0].addr32[3])
 631                         return (-1);
 632                 if (a->addr[1].addr32[3] > b->addr[1].addr32[3])
 633                         return (1);
 634                 if (a->addr[1].addr32[3] < b->addr[1].addr32[3])
 635                         return (-1);
 636                 if (a->addr[0].addr32[2] > b->addr[0].addr32[2])
 637                         return (1);
 638                 if (a->addr[0].addr32[2] < b->addr[0].addr32[2])
 639                         return (-1);
 640                 if (a->addr[1].addr32[2] > b->addr[1].addr32[2])
 641                         return (1);
 642                 if (a->addr[1].addr32[2] < b->addr[1].addr32[2])
 643                         return (-1);
 644                 if (a->addr[0].addr32[1] > b->addr[0].addr32[1])
 645                         return (1);
 646                 if (a->addr[0].addr32[1] < b->addr[0].addr32[1])
 647                         return (-1);
 648                 if (a->addr[1].addr32[1] > b->addr[1].addr32[1])
 649                         return (1);
 650                 if (a->addr[1].addr32[1] < b->addr[1].addr32[1])
 651                         return (-1);
 652                 if (a->addr[0].addr32[0] > b->addr[0].addr32[0])
 653                         return (1);
 654                 if (a->addr[0].addr32[0] < b->addr[0].addr32[0])
 655                         return (-1);
 656                 if (a->addr[1].addr32[0] > b->addr[1].addr32[0])
 657                         return (1);
 658                 if (a->addr[1].addr32[0] < b->addr[1].addr32[0])
 659                         return (-1);
 660                 break;
 661 #endif /* INET6 */
 662         }
 663
 664         if ((diff = a->port[0] - b->port[0]) != 0)
 665                 return (diff);
 666         if ((diff = a->port[1] - b->port[1]) != 0)
 667                 return (diff);
 668
 669         return (0);
 670 }
 671
 672 static __inline int
 673 pf_state_compare_id(struct pf_state *a, struct pf_state *b)
 674 {
 675         if (a->id > b->id)
 676                 return (1);
 677         if (a->id < b->id)
 678                 return (-1);
 679         if (a->creatorid > b->creatorid)
 680                 return (1);
 681         if (a->creatorid < b->creatorid)
 682                 return (-1);
 683
 684         return (0);
 685 }
 686
 687 int
 688 pf_state_key_attach(struct pf_state_key *sk, struct pf_state *s, int idx)
 689 {
 690         struct pf_state_item    *si;
 691         struct pf_state_key     *cur;
 692
 693         KKASSERT(s->key[idx] == NULL);  /* XXX handle this? */
 694
 695         if ((cur = RB_INSERT(pf_state_tree, &pf_statetbl, sk)) != NULL) {
 696                 /* key exists. check for same kif, if none, add to key */
 697                 TAILQ_FOREACH(si, &cur->states, entry)
 698                         if (si->s->kif == s->kif &&
 699                             si->s->direction == s->direction) {
 700                                 if (pf_status.debug >= PF_DEBUG_MISC) {
 701                                         kprintf(
 702                                             "pf: %s key attach failed on %s: ",
 703                                             (idx == PF_SK_WIRE) ?
 704                                             "wire" : "stack",
 705                                             s->kif->pfik_name);
 706                                         pf_print_state_parts(s,
 707                                             (idx == PF_SK_WIRE) ? sk : NULL,
 708                                             (idx == PF_SK_STACK) ? sk : NULL);
 709                                         kprintf("\n");
 710                                 }
 711                                 kfree(sk, M_PFSTATEKEYPL);
 712                                 return (-1);    /* collision! */
 713                         }
 714                 kfree(sk, M_PFSTATEKEYPL);
 715
 716                 s->key[idx] = cur;
 717         } else
 718                 s->key[idx] = sk;
 719
 720         if ((si = kmalloc(sizeof(struct pf_state_item), M_PFSTATEITEMPL, M_NOWAIT)) == NULL) {
 721                 pf_state_key_detach(s, idx);
 722                 return (-1);
 723         }
 724         si->s = s;
 725
 726         /* list is sorted, if-bound states before floating */
 727         if (s->kif == pfi_all)
 728                 TAILQ_INSERT_TAIL(&s->key[idx]->states, si, entry);
 729         else
 730                 TAILQ_INSERT_HEAD(&s->key[idx]->states, si, entry);
 731         return (0);
 732 }
 733
 734 void
 735 pf_detach_state(struct pf_state *s)
 736 {
 737         if (s->key[PF_SK_WIRE] == s->key[PF_SK_STACK])
 738                 s->key[PF_SK_WIRE] = NULL;
 739
 740         if (s->key[PF_SK_STACK] != NULL)
 741                 pf_state_key_detach(s, PF_SK_STACK);
 742
 743         if (s->key[PF_SK_WIRE] != NULL)
 744                 pf_state_key_detach(s, PF_SK_WIRE);
 745 }
 746
 747 void
 748 pf_state_key_detach(struct pf_state *s, int idx)
 749 {
 750         struct pf_state_item    *si;
 751         si = TAILQ_FIRST(&s->key[idx]->states);
 752         while (si && si->s != s)
 753             si = TAILQ_NEXT(si, entry);
 754
 755         if (si) {
 756                 TAILQ_REMOVE(&s->key[idx]->states, si, entry);
 757                 kfree(si, M_PFSTATEITEMPL);
 758         }
 759
 760         if (TAILQ_EMPTY(&s->key[idx]->states)) {
 761                 RB_REMOVE(pf_state_tree, &pf_statetbl, s->key[idx]);
 762                 if (s->key[idx]->reverse)
 763                         s->key[idx]->reverse->reverse = NULL;
 764                 if (s->key[idx]->inp)
 765                         s->key[idx]->inp->inp_pf_sk = NULL;
 766                 kfree(s->key[idx], M_PFSTATEKEYPL);
 767         }
 768         s->key[idx] = NULL;
 769 }
 770
 771 struct pf_state_key *
 772 pf_alloc_state_key(int pool_flags)
 773 {
 774         struct pf_state_key     *sk;
 775
 776         if ((sk = kmalloc(sizeof(struct pf_state_key), M_PFSTATEKEYPL, pool_flags)) == NULL)
 777                         return (NULL);
 778         TAILQ_INIT(&sk->states);
 779
 780         return (sk);
 781 }
 782
 783 int
 784 pf_state_key_setup(struct pf_pdesc *pd, struct pf_rule *nr,
 785         struct pf_state_key **skw, struct pf_state_key **sks,
 786         struct pf_state_key **skp, struct pf_state_key **nkp,
 787         struct pf_addr *saddr, struct pf_addr *daddr,
 788         u_int16_t sport, u_int16_t dport)
 789 {
 790         KKASSERT((*skp == NULL && *nkp == NULL));
 791
 792         if ((*skp = pf_alloc_state_key(M_NOWAIT | M_ZERO)) == NULL)
 793                 return (ENOMEM);
 794
 795         PF_ACPY(&(*skp)->addr[pd->sidx], saddr, pd->af);
 796         PF_ACPY(&(*skp)->addr[pd->didx], daddr, pd->af);
 797         (*skp)->port[pd->sidx] = sport;
 798         (*skp)->port[pd->didx] = dport;
 799         (*skp)->proto = pd->proto;
 800         (*skp)->af = pd->af;
 801
 802         if (nr != NULL) {
 803                 if ((*nkp = pf_alloc_state_key(M_NOWAIT | M_ZERO)) == NULL)
 804                         return (ENOMEM); /* caller must handle cleanup */
 805
 806                 /* XXX maybe just bcopy and TAILQ_INIT(&(*nkp)->states) */
 807                 PF_ACPY(&(*nkp)->addr[0], &(*skp)->addr[0], pd->af);
 808                 PF_ACPY(&(*nkp)->addr[1], &(*skp)->addr[1], pd->af);
 809                 (*nkp)->port[0] = (*skp)->port[0];
 810                 (*nkp)->port[1] = (*skp)->port[1];
 811                 (*nkp)->proto = pd->proto;
 812                 (*nkp)->af = pd->af;
 813         } else
 814                 *nkp = *skp;
 815
 816         if (pd->dir == PF_IN) {
 817                 *skw = *skp;
 818                 *sks = *nkp;
 819         } else {
 820                 *sks = *skp;
 821                 *skw = *nkp;
 822         }
 823         return (0);
 824 }
 825
 826
 827 int
 828 pf_state_insert(struct pfi_kif *kif, struct pf_state_key *skw,
 829     struct pf_state_key *sks, struct pf_state *s)
 830 {
 831         s->kif = kif;
 832
 833         if (skw == sks) {
 834                 if (pf_state_key_attach(skw, s, PF_SK_WIRE))
 835                         return (-1);
 836                 s->key[PF_SK_STACK] = s->key[PF_SK_WIRE];
 837         } else {
 838                 if (pf_state_key_attach(skw, s, PF_SK_WIRE)) {
 839                         kfree(sks, M_PFSTATEKEYPL);
 840                         return (-1);
 841                 }
 842                 if (pf_state_key_attach(sks, s, PF_SK_STACK)) {
 843                         pf_state_key_detach(s, PF_SK_WIRE);
 844                         return (-1);
 845                 }
 846         }
 847
 848         if (s->id == 0 && s->creatorid == 0) {
 849                 s->id = htobe64(pf_status.stateid++);
 850                 s->creatorid = pf_status.hostid;
 851         }
 852
 853         /*
 854          * Calculate hash code for altq
 855          */
 856         s->hash = crc32(s->key[PF_SK_WIRE], sizeof(*sks));
 857
 858         if (RB_INSERT(pf_state_tree_id, &tree_id, s) != NULL) {
 859                 if (pf_status.debug >= PF_DEBUG_MISC) {
 860                         kprintf("pf: state insert failed: "
 861                             "id: %016jx creatorid: %08x",
 862                               (uintmax_t)be64toh(s->id), ntohl(s->creatorid));
 863                         if (s->sync_flags & PFSTATE_FROMSYNC)
 864                                 kprintf(" (from sync)");
 865                         kprintf("\n");
 866                 }
 867                 pf_detach_state(s);
 868                 return (-1);
 869         }
 870         TAILQ_INSERT_TAIL(&state_list, s, entry_list);
 871         pf_status.fcounters[FCNT_STATE_INSERT]++;
 872         pf_status.states++;
 873         pfi_kif_ref(kif, PFI_KIF_REF_STATE);
 874         pfsync_insert_state(s);
 875         return (0);
 876 }
 877
 878 struct pf_state *
 879 pf_find_state_byid(struct pf_state_cmp *key)
 880 {
 881         pf_status.fcounters[FCNT_STATE_SEARCH]++;
 882
 883         return (RB_FIND(pf_state_tree_id, &tree_id, (struct pf_state *)key));
 884 }
 885
 886 struct pf_state *
 887 pf_find_state(struct pfi_kif *kif, struct pf_state_key_cmp *key, u_int dir,
 888     struct mbuf *m)
 889 {
 890         struct pf_state_key     *sk;
 891         struct pf_state_item    *si;
 892
 893         pf_status.fcounters[FCNT_STATE_SEARCH]++;
 894
 895         if (dir == PF_OUT && m->m_pkthdr.pf.statekey &&
 896             ((struct pf_state_key *)m->m_pkthdr.pf.statekey)->reverse)
 897                 sk = ((struct pf_state_key *)m->m_pkthdr.pf.statekey)->reverse;
 898         else {
 899                 if ((sk = RB_FIND(pf_state_tree, &pf_statetbl,
 900                     (struct pf_state_key *)key)) == NULL)
 901                         return (NULL);
 902                 if (dir == PF_OUT && m->m_pkthdr.pf.statekey) {
 903                         ((struct pf_state_key *)
 904                             m->m_pkthdr.pf.statekey)->reverse = sk;
 905                         sk->reverse = m->m_pkthdr.pf.statekey;
 906                 }
 907         }
 908
 909         if (dir == PF_OUT)
 910                 m->m_pkthdr.pf.statekey = NULL;
 911
 912         /* list is sorted, if-bound states before floating ones */
 913         TAILQ_FOREACH(si, &sk->states, entry)
 914                 if ((si->s->kif == pfi_all || si->s->kif == kif) &&
 915                     sk == (dir == PF_IN ? si->s->key[PF_SK_WIRE] :
 916                     si->s->key[PF_SK_STACK]))
 917                         return (si->s);
 918
 919         return (NULL);
 920 }
 921
 922 struct pf_state *
 923 pf_find_state_all(struct pf_state_key_cmp *key, u_int dir, int *more)
 924 {
 925         struct pf_state_key     *sk;
 926         struct pf_state_item    *si, *ret = NULL;
 927
 928         pf_status.fcounters[FCNT_STATE_SEARCH]++;
 929
 930         sk = RB_FIND(pf_state_tree, &pf_statetbl, (struct pf_state_key *)key);
 931
 932         if (sk != NULL) {
 933                 TAILQ_FOREACH(si, &sk->states, entry)
 934                         if (dir == PF_INOUT ||
 935                             (sk == (dir == PF_IN ? si->s->key[PF_SK_WIRE] :
 936                             si->s->key[PF_SK_STACK]))) {
 937                                 if (more == NULL)
 938                                         return (si->s);
 939
 940                                 if (ret)
 941                                         (*more)++;
 942                                 else
 943                                         ret = si;
 944                         }
 945         }
 946         return (ret ? ret->s : NULL);
 947 }
 948
 949 /* END state table stuff */
 950
 951
 952 void
 953 pf_purge_thread(void *v)
 954 {
 955         int nloops = 0;
 956         int locked = 0;
 957
 958         lwkt_gettoken(&pf_token);
 959         for (;;) {
 960                 tsleep(pf_purge_thread, PWAIT, "pftm", 1 * hz);
 961
 962                 lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
 963
 964                 if (pf_end_threads) {
 965                         pf_purge_expired_states(pf_status.states, 0);
 966                         pf_purge_expired_fragments();
 967                         pf_purge_expired_src_nodes(1);
 968                         pf_end_threads++;
 969
 970                         lockmgr(&pf_consistency_lock, LK_RELEASE);
 971                         wakeup(pf_purge_thread);
 972                         kthread_exit();
 973                 }
 974                 crit_enter();
 975
 976                 /* process a fraction of the state table every second */
 977                 if(!pf_purge_expired_states(1 + (pf_status.states
 978                     / pf_default_rule.timeout[PFTM_INTERVAL]), 0)) {
 979
 980                         pf_purge_expired_states(1 + (pf_status.states
 981                             / pf_default_rule.timeout[PFTM_INTERVAL]), 1);
 982                 }
 983
 984                 /* purge other expired types every PFTM_INTERVAL seconds */
 985                 if (++nloops >= pf_default_rule.timeout[PFTM_INTERVAL]) {
 986                         pf_purge_expired_fragments();
 987                         if (!pf_purge_expired_src_nodes(locked)) {
 988                                 pf_purge_expired_src_nodes(1);
 989                         }
 990                         nloops = 0;
 991                 }
 992                 crit_exit();
 993                 lockmgr(&pf_consistency_lock, LK_RELEASE);
 994         }
 995         lwkt_reltoken(&pf_token);
 996 }
 997
 998 u_int32_t
 999 pf_state_expires(const struct pf_state *state)
1000 {
1001         u_int32_t       timeout;
1002         u_int32_t       start;
1003         u_int32_t       end;
1004         u_int32_t       states;
1005
1006         /* handle all PFTM_* > PFTM_MAX here */
1007         if (state->timeout == PFTM_PURGE)
1008                 return (time_second);
1009         if (state->timeout == PFTM_UNTIL_PACKET)
1010                 return (0);
1011         KKASSERT(state->timeout != PFTM_UNLINKED);
1012         KKASSERT(state->timeout < PFTM_MAX);
1013         timeout = state->rule.ptr->timeout[state->timeout];
1014         if (!timeout)
1015                 timeout = pf_default_rule.timeout[state->timeout];
1016         start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START];
1017         if (start) {
1018                 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END];
1019                 states = state->rule.ptr->states_cur;
1020         } else {
1021                 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START];
1022                 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END];
1023                 states = pf_status.states;
1024         }
1025         if (end && states > start && start < end) {
1026                 if (states < end)
1027                         return (state->expire + timeout * (end - states) /
1028                             (end - start));
1029                 else
1030                         return (time_second);
1031         }
1032         return (state->expire + timeout);
1033 }
1034
1035 int
1036 pf_purge_expired_src_nodes(int waslocked)
1037 {
1038          struct pf_src_node             *cur, *next;
1039          int                             locked = waslocked;
1040
1041          for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) {
1042                  next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur);
1043
1044                  if (cur->states <= 0 && cur->expire <= time_second) {
1045                          if (! locked) {
1046                                  lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
1047                                  next = RB_NEXT(pf_src_tree,
1048                                      &tree_src_tracking, cur);
1049                                  locked = 1;
1050                          }
1051                          if (cur->rule.ptr != NULL) {
1052                                  cur->rule.ptr->src_nodes--;
1053                                  if (cur->rule.ptr->states_cur <= 0 &&
1054                                      cur->rule.ptr->max_src_nodes <= 0)
1055                                          pf_rm_rule(NULL, cur->rule.ptr);
1056                          }
1057                          RB_REMOVE(pf_src_tree, &tree_src_tracking, cur);
1058                          pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
1059                          pf_status.src_nodes--;
1060                          kfree(cur, M_PFSRCTREEPL);
1061                  }
1062          }
1063
1064          if (locked && !waslocked)
1065                 lockmgr(&pf_consistency_lock, LK_RELEASE);
1066         return(1);
1067 }
1068
1069 void
1070 pf_src_tree_remove_state(struct pf_state *s)
1071 {
1072         u_int32_t timeout;
1073
1074         if (s->src_node != NULL) {
1075                 if (s->src.tcp_est)
1076                         --s->src_node->conn;
1077                 if (--s->src_node->states <= 0) {
1078                         timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1079                         if (!timeout)
1080                                 timeout =
1081                                     pf_default_rule.timeout[PFTM_SRC_NODE];
1082                         s->src_node->expire = time_second + timeout;
1083                 }
1084         }
1085         if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) {
1086                 if (--s->nat_src_node->states <= 0) {
1087                         timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1088                         if (!timeout)
1089                                 timeout =
1090                                     pf_default_rule.timeout[PFTM_SRC_NODE];
1091                         s->nat_src_node->expire = time_second + timeout;
1092                 }
1093         }
1094         s->src_node = s->nat_src_node = NULL;
1095 }
1096
1097 /* callers should be at crit_enter() */
1098 void
1099 pf_unlink_state(struct pf_state *cur)
1100 {
1101         if (cur->src.state == PF_TCPS_PROXY_DST) {
1102                 /* XXX wire key the right one? */
1103                 pf_send_tcp(cur->rule.ptr, cur->key[PF_SK_WIRE]->af,
1104                     &cur->key[PF_SK_WIRE]->addr[1],
1105                     &cur->key[PF_SK_WIRE]->addr[0],
1106                     cur->key[PF_SK_WIRE]->port[1],
1107                     cur->key[PF_SK_WIRE]->port[0],
1108                     cur->src.seqhi, cur->src.seqlo + 1,
1109                     TH_RST|TH_ACK, 0, 0, 0, 1, cur->tag, NULL, NULL);
1110         }
1111         RB_REMOVE(pf_state_tree_id, &tree_id, cur);
1112         if (cur->creatorid == pf_status.hostid)
1113                 pfsync_delete_state(cur);
1114         cur->timeout = PFTM_UNLINKED;
1115         pf_src_tree_remove_state(cur);
1116         pf_detach_state(cur);
1117 }
1118
1119 static struct pf_state  *purge_cur;
1120
1121 /* callers should be at crit_enter() and hold the
1122  * write_lock on pf_consistency_lock */
1123 void
1124 pf_free_state(struct pf_state *cur)
1125 {
1126         if (pfsyncif != NULL &&
1127             (pfsyncif->sc_bulk_send_next == cur ||
1128             pfsyncif->sc_bulk_terminator == cur))
1129                 return;
1130         KKASSERT(cur->timeout == PFTM_UNLINKED);
1131         if (--cur->rule.ptr->states_cur <= 0 &&
1132             cur->rule.ptr->src_nodes <= 0)
1133                 pf_rm_rule(NULL, cur->rule.ptr);
1134         if (cur->nat_rule.ptr != NULL)
1135                 if (--cur->nat_rule.ptr->states_cur <= 0 &&
1136                         cur->nat_rule.ptr->src_nodes <= 0)
1137                         pf_rm_rule(NULL, cur->nat_rule.ptr);
1138         if (cur->anchor.ptr != NULL)
1139                 if (--cur->anchor.ptr->states_cur <= 0)
1140                         pf_rm_rule(NULL, cur->anchor.ptr);
1141         pf_normalize_tcp_cleanup(cur);
1142         pfi_kif_unref(cur->kif, PFI_KIF_REF_STATE);
1143
1144         /*
1145          * We may be freeing pf_purge_expired_states()'s saved scan entry,
1146          * adjust it if necessary.
1147          */
1148         if (purge_cur == cur) {
1149                 kprintf("PURGE CONFLICT\n");
1150                 purge_cur = TAILQ_NEXT(purge_cur, entry_list);
1151         }
1152         TAILQ_REMOVE(&state_list, cur, entry_list);
1153         if (cur->tag)
1154                 pf_tag_unref(cur->tag);
1155         kfree(cur, M_PFSTATEPL);
1156         pf_status.fcounters[FCNT_STATE_REMOVALS]++;
1157         pf_status.states--;
1158 }
1159
1160 int
1161 pf_purge_expired_states(u_int32_t maxcheck, int waslocked)
1162 {
1163         struct pf_state         *cur;
1164         int                      locked = waslocked;
1165
1166         while (maxcheck--) {
1167                 /*
1168                  * Wrap to start of list when we hit the end
1169                  */
1170                 cur = purge_cur;
1171                 if (cur == NULL) {
1172                         cur = TAILQ_FIRST(&state_list);
1173                         if (cur == NULL)
1174                                 break;  /* list empty */
1175                 }
1176
1177                 /*
1178                  * Setup next (purge_cur) while we process this one.  If we block and
1179                  * something else deletes purge_cur, pf_free_state() will adjust it further
1180                  * ahead.
1181                  */
1182                 purge_cur = TAILQ_NEXT(cur, entry_list);
1183
1184                 if (cur->timeout == PFTM_UNLINKED) {
1185                         /* free unlinked state */
1186                         if (! locked) {
1187                                 lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
1188                                 locked = 1;
1189                         }
1190                         pf_free_state(cur);
1191                 } else if (pf_state_expires(cur) <= time_second) {
1192                         /* unlink and free expired state */
1193                         pf_unlink_state(cur);
1194                         if (! locked) {
1195                                 if (!lockmgr(&pf_consistency_lock, LK_EXCLUSIVE))
1196                                         return (0);
1197                                 locked = 1;
1198                         }
1199                         pf_free_state(cur);
1200                 }
1201         }
1202
1203         if (locked)
1204                 lockmgr(&pf_consistency_lock, LK_RELEASE);
1205         return (1);
1206 }
1207
1208 int
1209 pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw)
1210 {
1211         if (aw->type != PF_ADDR_TABLE)
1212                 return (0);
1213         if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL)
1214                 return (1);
1215         return (0);
1216 }
1217
1218 void
1219 pf_tbladdr_remove(struct pf_addr_wrap *aw)
1220 {
1221         if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL)
1222                 return;
1223         pfr_detach_table(aw->p.tbl);
1224         aw->p.tbl = NULL;
1225 }
1226
1227 void
1228 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
1229 {
1230         struct pfr_ktable *kt = aw->p.tbl;
1231
1232         if (aw->type != PF_ADDR_TABLE || kt == NULL)
1233                 return;
1234         if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
1235                 kt = kt->pfrkt_root;
1236         aw->p.tbl = NULL;
1237         aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
1238                 kt->pfrkt_cnt : -1;
1239 }
1240
1241 void
1242 pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af)
1243 {
1244         switch (af) {
1245 #ifdef INET
1246         case AF_INET: {
1247                 u_int32_t a = ntohl(addr->addr32[0]);
1248                 kprintf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255,
1249                     (a>>8)&255, a&255);
1250                 if (p) {
1251                         p = ntohs(p);
1252                         kprintf(":%u", p);
1253                 }
1254                 break;
1255         }
1256 #endif /* INET */
1257 #ifdef INET6
1258         case AF_INET6: {
1259                 u_int16_t b;
1260                 u_int8_t i, curstart = 255, curend = 0,
1261                     maxstart = 0, maxend = 0;
1262                 for (i = 0; i < 8; i++) {
1263                         if (!addr->addr16[i]) {
1264                                 if (curstart == 255)
1265                                         curstart = i;
1266                                 else
1267                                         curend = i;
1268                         } else {
1269                                 if (curstart) {
1270                                         if ((curend - curstart) >
1271                                             (maxend - maxstart)) {
1272                                                 maxstart = curstart;
1273                                                 maxend = curend;
1274                                                 curstart = 255;
1275                                         }
1276                                 }
1277                         }
1278                 }
1279                 for (i = 0; i < 8; i++) {
1280                         if (i >= maxstart && i <= maxend) {
1281                                 if (maxend != 7) {
1282                                         if (i == maxstart)
1283                                                 kprintf(":");
1284                                 } else {
1285                                         if (i == maxend)
1286                                                 kprintf(":");
1287                                 }
1288                         } else {
1289                                 b = ntohs(addr->addr16[i]);
1290                                 kprintf("%x", b);
1291                                 if (i < 7)
1292                                         kprintf(":");
1293                         }
1294                 }
1295                 if (p) {
1296                         p = ntohs(p);
1297                         kprintf("[%u]", p);
1298                 }
1299                 break;
1300         }
1301 #endif /* INET6 */
1302         }
1303 }
1304
1305 void
1306 pf_print_state(struct pf_state *s)
1307 {
1308         pf_print_state_parts(s, NULL, NULL);
1309 }
1310
1311 void
1312 pf_print_state_parts(struct pf_state *s,
1313     struct pf_state_key *skwp, struct pf_state_key *sksp)
1314 {
1315         struct pf_state_key *skw, *sks;
1316         u_int8_t proto, dir;
1317
1318         /* Do our best to fill these, but they're skipped if NULL */
1319         skw = skwp ? skwp : (s ? s->key[PF_SK_WIRE] : NULL);
1320         sks = sksp ? sksp : (s ? s->key[PF_SK_STACK] : NULL);
1321         proto = skw ? skw->proto : (sks ? sks->proto : 0);
1322         dir = s ? s->direction : 0;
1323
1324         switch (proto) {
1325         case IPPROTO_TCP:
1326                 kprintf("TCP ");
1327                 break;
1328         case IPPROTO_UDP:
1329                 kprintf("UDP ");
1330                 break;
1331         case IPPROTO_ICMP:
1332                 kprintf("ICMP ");
1333                 break;
1334         case IPPROTO_ICMPV6:
1335                 kprintf("ICMPV6 ");
1336                 break;
1337         default:
1338                 kprintf("%u ", skw->proto);
1339                 break;
1340         }
1341         switch (dir) {
1342         case PF_IN:
1343                 kprintf(" in");
1344                 break;
1345         case PF_OUT:
1346                 kprintf(" out");
1347                 break;
1348         }
1349         if (skw) {
1350                 kprintf(" wire: ");
1351                 pf_print_host(&skw->addr[0], skw->port[0], skw->af);
1352                 kprintf(" ");
1353                 pf_print_host(&skw->addr[1], skw->port[1], skw->af);
1354         }
1355         if (sks) {
1356                 kprintf(" stack: ");
1357                 if (sks != skw) {
1358                         pf_print_host(&sks->addr[0], sks->port[0], sks->af);
1359                         kprintf(" ");
1360                         pf_print_host(&sks->addr[1], sks->port[1], sks->af);
1361                 } else
1362                         kprintf("-");
1363         }
1364         if (s) {
1365                 if (proto == IPPROTO_TCP) {
1366                         kprintf(" [lo=%u high=%u win=%u modulator=%u",
1367                             s->src.seqlo, s->src.seqhi,
1368                             s->src.max_win, s->src.seqdiff);
1369                         if (s->src.wscale && s->dst.wscale)
1370                                 kprintf(" wscale=%u",
1371                                     s->src.wscale & PF_WSCALE_MASK);
1372                         kprintf("]");
1373                         kprintf(" [lo=%u high=%u win=%u modulator=%u",
1374                             s->dst.seqlo, s->dst.seqhi,
1375                             s->dst.max_win, s->dst.seqdiff);
1376                         if (s->src.wscale && s->dst.wscale)
1377                                 kprintf(" wscale=%u",
1378                                 s->dst.wscale & PF_WSCALE_MASK);
1379                         kprintf("]");
1380                 }
1381                 kprintf(" %u:%u", s->src.state, s->dst.state);
1382         }
1383 }
1384
1385 void
1386 pf_print_flags(u_int8_t f)
1387 {
1388         if (f)
1389                 kprintf(" ");
1390         if (f & TH_FIN)
1391                 kprintf("F");
1392         if (f & TH_SYN)
1393                 kprintf("S");
1394         if (f & TH_RST)
1395                 kprintf("R");
1396         if (f & TH_PUSH)
1397                 kprintf("P");
1398         if (f & TH_ACK)
1399                 kprintf("A");
1400         if (f & TH_URG)
1401                 kprintf("U");
1402         if (f & TH_ECE)
1403                 kprintf("E");
1404         if (f & TH_CWR)
1405                 kprintf("W");
1406 }
1407
1408 #define PF_SET_SKIP_STEPS(i)                                    \
1409         do {                                                    \
1410                 while (head[i] != cur) {                        \
1411                         head[i]->skip[i].ptr = cur;             \
1412                         head[i] = TAILQ_NEXT(head[i], entries); \
1413                 }                                               \
1414         } while (0)
1415
1416 void
1417 pf_calc_skip_steps(struct pf_rulequeue *rules)
1418 {
1419         struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT];
1420         int i;
1421
1422         cur = TAILQ_FIRST(rules);
1423         prev = cur;
1424         for (i = 0; i < PF_SKIP_COUNT; ++i)
1425                 head[i] = cur;
1426         while (cur != NULL) {
1427
1428                 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
1429                         PF_SET_SKIP_STEPS(PF_SKIP_IFP);
1430                 if (cur->direction != prev->direction)
1431                         PF_SET_SKIP_STEPS(PF_SKIP_DIR);
1432                 if (cur->af != prev->af)
1433                         PF_SET_SKIP_STEPS(PF_SKIP_AF);
1434                 if (cur->proto != prev->proto)
1435                         PF_SET_SKIP_STEPS(PF_SKIP_PROTO);
1436                 if (cur->src.neg != prev->src.neg ||
1437                     pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr))
1438                         PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR);
1439                 if (cur->src.port[0] != prev->src.port[0] ||
1440                     cur->src.port[1] != prev->src.port[1] ||
1441                     cur->src.port_op != prev->src.port_op)
1442                         PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
1443                 if (cur->dst.neg != prev->dst.neg ||
1444                     pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr))
1445                         PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR);
1446                 if (cur->dst.port[0] != prev->dst.port[0] ||
1447                     cur->dst.port[1] != prev->dst.port[1] ||
1448                     cur->dst.port_op != prev->dst.port_op)
1449                         PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
1450
1451                 prev = cur;
1452                 cur = TAILQ_NEXT(cur, entries);
1453         }
1454         for (i = 0; i < PF_SKIP_COUNT; ++i)
1455                 PF_SET_SKIP_STEPS(i);
1456 }
1457
1458 int
1459 pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2)
1460 {
1461         if (aw1->type != aw2->type)
1462                 return (1);
1463         switch (aw1->type) {
1464         case PF_ADDR_ADDRMASK:
1465         case PF_ADDR_RANGE:
1466                 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0))
1467                         return (1);
1468                 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0))
1469                         return (1);
1470                 return (0);
1471         case PF_ADDR_DYNIFTL:
1472                 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt);
1473         case PF_ADDR_NOROUTE:
1474         case PF_ADDR_URPFFAILED:
1475                 return (0);
1476         case PF_ADDR_TABLE:
1477                 return (aw1->p.tbl != aw2->p.tbl);
1478         case PF_ADDR_RTLABEL:
1479                 return (aw1->v.rtlabel != aw2->v.rtlabel);
1480         default:
1481                 kprintf("invalid address type: %d\n", aw1->type);
1482                 return (1);
1483         }
1484 }
1485
1486 u_int16_t
1487 pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp)
1488 {
1489         u_int32_t       l;
1490
1491         if (udp && !cksum)
1492                 return (0x0000);
1493         l = cksum + old - new;
1494         l = (l >> 16) + (l & 65535);
1495         l = l & 65535;
1496         if (udp && !l)
1497                 return (0xFFFF);
1498         return (l);
1499 }
1500
1501 void
1502 pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc,
1503     struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af)
1504 {
1505         struct pf_addr  ao;
1506         u_int16_t       po = *p;
1507
1508         PF_ACPY(&ao, a, af);
1509         PF_ACPY(a, an, af);
1510
1511         *p = pn;
1512
1513         switch (af) {
1514 #ifdef INET
1515         case AF_INET:
1516                 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1517                     ao.addr16[0], an->addr16[0], 0),
1518                     ao.addr16[1], an->addr16[1], 0);
1519                 *p = pn;
1520                 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1521                     ao.addr16[0], an->addr16[0], u),
1522                     ao.addr16[1], an->addr16[1], u),
1523                     po, pn, u);
1524                 break;
1525 #endif /* INET */
1526 #ifdef INET6
1527         case AF_INET6:
1528                 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1529                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1530                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1531                     ao.addr16[0], an->addr16[0], u),
1532                     ao.addr16[1], an->addr16[1], u),
1533                     ao.addr16[2], an->addr16[2], u),
1534                     ao.addr16[3], an->addr16[3], u),
1535                     ao.addr16[4], an->addr16[4], u),
1536                     ao.addr16[5], an->addr16[5], u),
1537                     ao.addr16[6], an->addr16[6], u),
1538                     ao.addr16[7], an->addr16[7], u),
1539                     po, pn, u);
1540                 break;
1541 #endif /* INET6 */
1542         }
1543 }
1544
1545
1546 /* Changes a u_int32_t.  Uses a void * so there are no align restrictions */
1547 void
1548 pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u)
1549 {
1550         u_int32_t       ao;
1551
1552         memcpy(&ao, a, sizeof(ao));
1553         memcpy(a, &an, sizeof(u_int32_t));
1554         *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u),
1555             ao % 65536, an % 65536, u);
1556 }
1557
1558 #ifdef INET6
1559 void
1560 pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u)
1561 {
1562         struct pf_addr  ao;
1563
1564         PF_ACPY(&ao, a, AF_INET6);
1565         PF_ACPY(a, an, AF_INET6);
1566
1567         *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1568             pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1569             pf_cksum_fixup(pf_cksum_fixup(*c,
1570             ao.addr16[0], an->addr16[0], u),
1571             ao.addr16[1], an->addr16[1], u),
1572             ao.addr16[2], an->addr16[2], u),
1573             ao.addr16[3], an->addr16[3], u),
1574             ao.addr16[4], an->addr16[4], u),
1575             ao.addr16[5], an->addr16[5], u),
1576             ao.addr16[6], an->addr16[6], u),
1577             ao.addr16[7], an->addr16[7], u);
1578 }
1579 #endif /* INET6 */
1580
1581 void
1582 pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa,
1583     struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c,
1584     u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af)
1585 {
1586         struct pf_addr  oia, ooa;
1587
1588         PF_ACPY(&oia, ia, af);
1589         if (oa)
1590                 PF_ACPY(&ooa, oa, af);
1591
1592         /* Change inner protocol port, fix inner protocol checksum. */
1593         if (ip != NULL) {
1594                 u_int16_t       oip = *ip;
1595                 u_int32_t       opc = 0;
1596
1597                 if (pc != NULL)
1598                         opc = *pc;
1599                 *ip = np;
1600                 if (pc != NULL)
1601                         *pc = pf_cksum_fixup(*pc, oip, *ip, u);
1602                 *ic = pf_cksum_fixup(*ic, oip, *ip, 0);
1603                 if (pc != NULL)
1604                         *ic = pf_cksum_fixup(*ic, opc, *pc, 0);
1605         }
1606         /* Change inner ip address, fix inner ip and icmp checksums. */
1607         PF_ACPY(ia, na, af);
1608         switch (af) {
1609 #ifdef INET
1610         case AF_INET: {
1611                 u_int32_t        oh2c = *h2c;
1612
1613                 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c,
1614                     oia.addr16[0], ia->addr16[0], 0),
1615                     oia.addr16[1], ia->addr16[1], 0);
1616                 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1617                     oia.addr16[0], ia->addr16[0], 0),
1618                     oia.addr16[1], ia->addr16[1], 0);
1619                 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0);
1620                 break;
1621         }
1622 #endif /* INET */
1623 #ifdef INET6
1624         case AF_INET6:
1625                 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1626                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1627                     pf_cksum_fixup(pf_cksum_fixup(*ic,
1628                     oia.addr16[0], ia->addr16[0], u),
1629                     oia.addr16[1], ia->addr16[1], u),
1630                     oia.addr16[2], ia->addr16[2], u),
1631                     oia.addr16[3], ia->addr16[3], u),
1632                     oia.addr16[4], ia->addr16[4], u),
1633                     oia.addr16[5], ia->addr16[5], u),
1634                     oia.addr16[6], ia->addr16[6], u),
1635                     oia.addr16[7], ia->addr16[7], u);
1636                 break;
1637 #endif /* INET6 */
1638         }
1639         /* Outer ip address, fix outer ip or icmpv6 checksum, if necessary. */
1640         if (oa) {
1641                 PF_ACPY(oa, na, af);
1642                 switch (af) {
1643 #ifdef INET
1644                 case AF_INET:
1645                         *hc = pf_cksum_fixup(pf_cksum_fixup(*hc,
1646                             ooa.addr16[0], oa->addr16[0], 0),
1647                             ooa.addr16[1], oa->addr16[1], 0);
1648                         break;
1649 #endif /* INET */
1650 #ifdef INET6
1651                 case AF_INET6:
1652                         *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1653                             pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1654                             pf_cksum_fixup(pf_cksum_fixup(*ic,
1655                             ooa.addr16[0], oa->addr16[0], u),
1656                             ooa.addr16[1], oa->addr16[1], u),
1657                             ooa.addr16[2], oa->addr16[2], u),
1658                             ooa.addr16[3], oa->addr16[3], u),
1659                             ooa.addr16[4], oa->addr16[4], u),
1660                             ooa.addr16[5], oa->addr16[5], u),
1661                             ooa.addr16[6], oa->addr16[6], u),
1662                             ooa.addr16[7], oa->addr16[7], u);
1663                         break;
1664 #endif /* INET6 */
1665                 }
1666         }
1667 }
1668
1669
1670 /*
1671  * Need to modulate the sequence numbers in the TCP SACK option
1672  * (credits to Krzysztof Pfaff for report and patch)
1673  */
1674 int
1675 pf_modulate_sack(struct mbuf *m, int off, struct pf_pdesc *pd,
1676     struct tcphdr *th, struct pf_state_peer *dst)
1677 {
1678         int hlen = (th->th_off << 2) - sizeof(*th), thoptlen = hlen;
1679         u_int8_t opts[TCP_MAXOLEN], *opt = opts;
1680         int copyback = 0, i, olen;
1681         struct raw_sackblock sack;
1682
1683 #define TCPOLEN_SACKLEN (TCPOLEN_SACK + 2)
1684         if (hlen < TCPOLEN_SACKLEN ||
1685             !pf_pull_hdr(m, off + sizeof(*th), opts, hlen, NULL, NULL, pd->af))
1686                 return 0;
1687
1688         while (hlen >= TCPOLEN_SACKLEN) {
1689                 olen = opt[1];
1690                 switch (*opt) {
1691                 case TCPOPT_EOL:        /* FALLTHROUGH */
1692                 case TCPOPT_NOP:
1693                         opt++;
1694                         hlen--;
1695                         break;
1696                 case TCPOPT_SACK:
1697                         if (olen > hlen)
1698                                 olen = hlen;
1699                         if (olen >= TCPOLEN_SACKLEN) {
1700                                 for (i = 2; i + TCPOLEN_SACK <= olen;
1701                                     i += TCPOLEN_SACK) {
1702                                         memcpy(&sack, &opt[i], sizeof(sack));
1703                                         pf_change_a(&sack.rblk_start, &th->th_sum,
1704                                             htonl(ntohl(sack.rblk_start) -
1705                                             dst->seqdiff), 0);
1706                                         pf_change_a(&sack.rblk_end, &th->th_sum,
1707                                             htonl(ntohl(sack.rblk_end) -
1708                                             dst->seqdiff), 0);
1709                                         memcpy(&opt[i], &sack, sizeof(sack));
1710                                 }
1711                                 copyback = 1;
1712                         }
1713                         /* FALLTHROUGH */
1714                 default:
1715                         if (olen < 2)
1716                                 olen = 2;
1717                         hlen -= olen;
1718                         opt += olen;
1719                 }
1720         }
1721
1722         if (copyback)
1723                 m_copyback(m, off + sizeof(*th), thoptlen, opts);
1724         return (copyback);
1725 }
1726
1727 void
1728 pf_send_tcp(const struct pf_rule *r, sa_family_t af,
1729     const struct pf_addr *saddr, const struct pf_addr *daddr,
1730     u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack,
1731     u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl, int tag,
1732     u_int16_t rtag, struct ether_header *eh, struct ifnet *ifp)
1733 {
1734         struct mbuf     *m;
1735         int              len = 0, tlen;
1736 #ifdef INET
1737         struct ip       *h = NULL;
1738 #endif /* INET */
1739 #ifdef INET6
1740         struct ip6_hdr  *h6 = NULL;
1741 #endif /* INET6 */
1742         struct tcphdr   *th = NULL;
1743         char            *opt;
1744
1745         ASSERT_LWKT_TOKEN_HELD(&pf_token);
1746
1747         /* maximum segment size tcp option */
1748         tlen = sizeof(struct tcphdr);
1749         if (mss)
1750                 tlen += 4;
1751
1752         switch (af) {
1753 #ifdef INET
1754         case AF_INET:
1755                 len = sizeof(struct ip) + tlen;
1756                 break;
1757 #endif /* INET */
1758 #ifdef INET6
1759         case AF_INET6:
1760                 len = sizeof(struct ip6_hdr) + tlen;
1761                 break;
1762 #endif /* INET6 */
1763         }
1764
1765         /*
1766          * Create outgoing mbuf.
1767          *
1768          * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags,
1769          * so make sure pf.flags is clear.
1770          */
1771         m = m_gethdr(MB_DONTWAIT, MT_HEADER);
1772         if (m == NULL) {
1773                 return;
1774         }
1775         if (tag)
1776                 m->m_pkthdr.fw_flags |= PF_MBUF_TAGGED;
1777         m->m_pkthdr.pf.flags = 0;
1778         m->m_pkthdr.pf.tag = rtag;
1779         /* XXX Recheck when upgrading to > 4.4 */
1780         m->m_pkthdr.pf.statekey = NULL;
1781         if (r != NULL && r->rtableid >= 0)
1782                 m->m_pkthdr.pf.rtableid = r->rtableid;
1783
1784 #ifdef ALTQ
1785         if (r != NULL && r->qid) {
1786                 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
1787                 m->m_pkthdr.pf.qid = r->qid;
1788                 m->m_pkthdr.pf.ecn_af = af;
1789                 m->m_pkthdr.pf.hdr = mtod(m, struct ip *);
1790         }
1791 #endif /* ALTQ */
1792         m->m_data += max_linkhdr;
1793         m->m_pkthdr.len = m->m_len = len;
1794         m->m_pkthdr.rcvif = NULL;
1795         bzero(m->m_data, len);
1796         switch (af) {
1797 #ifdef INET
1798         case AF_INET:
1799                 h = mtod(m, struct ip *);
1800
1801                 /* IP header fields included in the TCP checksum */
1802                 h->ip_p = IPPROTO_TCP;
1803                 h->ip_len = tlen;
1804                 h->ip_src.s_addr = saddr->v4.s_addr;
1805                 h->ip_dst.s_addr = daddr->v4.s_addr;
1806
1807                 th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip));
1808                 break;
1809 #endif /* INET */
1810 #ifdef INET6
1811         case AF_INET6:
1812                 h6 = mtod(m, struct ip6_hdr *);
1813
1814                 /* IP header fields included in the TCP checksum */
1815                 h6->ip6_nxt = IPPROTO_TCP;
1816                 h6->ip6_plen = htons(tlen);
1817                 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr));
1818                 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr));
1819
1820                 th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr));
1821                 break;
1822 #endif /* INET6 */
1823         }
1824
1825         /* TCP header */
1826         th->th_sport = sport;
1827         th->th_dport = dport;
1828         th->th_seq = htonl(seq);
1829         th->th_ack = htonl(ack);
1830         th->th_off = tlen >> 2;
1831         th->th_flags = flags;
1832         th->th_win = htons(win);
1833
1834         if (mss) {
1835                 opt = (char *)(th + 1);
1836                 opt[0] = TCPOPT_MAXSEG;
1837                 opt[1] = 4;
1838                 mss = htons(mss);
1839                 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2);
1840         }
1841
1842         switch (af) {
1843 #ifdef INET
1844         case AF_INET:
1845                 /* TCP checksum */
1846                 th->th_sum = in_cksum(m, len);
1847
1848                 /* Finish the IP header */
1849                 h->ip_v = 4;
1850                 h->ip_hl = sizeof(*h) >> 2;
1851                 h->ip_tos = IPTOS_LOWDELAY;
1852                 h->ip_len = len;
1853                 h->ip_off = path_mtu_discovery ? IP_DF : 0;
1854                 h->ip_ttl = ttl ? ttl : ip_defttl;
1855                 h->ip_sum = 0;
1856                 if (eh == NULL) {
1857                         lwkt_reltoken(&pf_token);
1858                         ip_output(m, NULL, NULL, 0, NULL, NULL);
1859                         lwkt_gettoken(&pf_token);
1860                 } else {
1861                         struct route             ro;
1862                         struct rtentry           rt;
1863                         struct ether_header     *e = (void *)ro.ro_dst.sa_data;
1864
1865                         if (ifp == NULL) {
1866                                 m_freem(m);
1867                                 return;
1868                         }
1869                         rt.rt_ifp = ifp;
1870                         ro.ro_rt = &rt;
1871                         ro.ro_dst.sa_len = sizeof(ro.ro_dst);
1872                         ro.ro_dst.sa_family = pseudo_AF_HDRCMPLT;
1873                         bcopy(eh->ether_dhost, e->ether_shost, ETHER_ADDR_LEN);
1874                         bcopy(eh->ether_shost, e->ether_dhost, ETHER_ADDR_LEN);
1875                         e->ether_type = eh->ether_type;
1876                         /* XXX_IMPORT: later */
1877                         lwkt_reltoken(&pf_token);
1878                         ip_output(m, NULL, &ro, 0, NULL, NULL);
1879                         lwkt_gettoken(&pf_token);
1880                 }
1881                 break;
1882 #endif /* INET */
1883 #ifdef INET6
1884         case AF_INET6:
1885                 /* TCP checksum */
1886                 th->th_sum = in6_cksum(m, IPPROTO_TCP,
1887                     sizeof(struct ip6_hdr), tlen);
1888
1889                 h6->ip6_vfc |= IPV6_VERSION;
1890                 h6->ip6_hlim = IPV6_DEFHLIM;
1891
1892                 lwkt_reltoken(&pf_token);
1893                 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
1894                 lwkt_gettoken(&pf_token);
1895                 break;
1896 #endif /* INET6 */
1897         }
1898 }
1899
1900 void
1901 pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af,
1902     struct pf_rule *r)
1903 {
1904         struct mbuf     *m0;
1905
1906         /*
1907          * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags,
1908          * so make sure pf.flags is clear.
1909          */
1910         if ((m0 = m_copy(m, 0, M_COPYALL)) == NULL)
1911                 return;
1912
1913         m0->m_pkthdr.fw_flags |= PF_MBUF_TAGGED;
1914         m0->m_pkthdr.pf.flags = 0;
1915         /* XXX Re-Check when Upgrading to > 4.4 */
1916         m0->m_pkthdr.pf.statekey = NULL;
1917
1918         if (r->rtableid >= 0)
1919                 m0->m_pkthdr.pf.rtableid = r->rtableid;
1920
1921 #ifdef ALTQ
1922         if (r->qid) {
1923                 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
1924                 m0->m_pkthdr.pf.qid = r->qid;
1925                 m0->m_pkthdr.pf.ecn_af = af;
1926                 m0->m_pkthdr.pf.hdr = mtod(m0, struct ip *);
1927         }
1928 #endif /* ALTQ */
1929
1930         switch (af) {
1931 #ifdef INET
1932         case AF_INET:
1933                 icmp_error(m0, type, code, 0, 0);
1934                 break;
1935 #endif /* INET */
1936 #ifdef INET6
1937         case AF_INET6:
1938                 icmp6_error(m0, type, code, 0);
1939                 break;
1940 #endif /* INET6 */
1941         }
1942 }
1943
1944 /*
1945  * Return 1 if the addresses a and b match (with mask m), otherwise return 0.
1946  * If n is 0, they match if they are equal. If n is != 0, they match if they
1947  * are different.
1948  */
1949 int
1950 pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m,
1951     struct pf_addr *b, sa_family_t af)
1952 {
1953         int     match = 0;
1954
1955         switch (af) {
1956 #ifdef INET
1957         case AF_INET:
1958                 if ((a->addr32[0] & m->addr32[0]) ==
1959                     (b->addr32[0] & m->addr32[0]))
1960                         match++;
1961                 break;
1962 #endif /* INET */
1963 #ifdef INET6
1964         case AF_INET6:
1965                 if (((a->addr32[0] & m->addr32[0]) ==
1966                      (b->addr32[0] & m->addr32[0])) &&
1967                     ((a->addr32[1] & m->addr32[1]) ==
1968                      (b->addr32[1] & m->addr32[1])) &&
1969                     ((a->addr32[2] & m->addr32[2]) ==
1970                      (b->addr32[2] & m->addr32[2])) &&
1971                     ((a->addr32[3] & m->addr32[3]) ==
1972                      (b->addr32[3] & m->addr32[3])))
1973                         match++;
1974                 break;
1975 #endif /* INET6 */
1976         }
1977         if (match) {
1978                 if (n)
1979                         return (0);
1980                 else
1981                         return (1);
1982         } else {
1983                 if (n)
1984                         return (1);
1985                 else
1986                         return (0);
1987         }
1988 }
1989
1990 /*
1991  * Return 1 if b <= a <= e, otherwise return 0.
1992  */
1993 int
1994 pf_match_addr_range(struct pf_addr *b, struct pf_addr *e,
1995     struct pf_addr *a, sa_family_t af)
1996 {
1997         switch (af) {
1998 #ifdef INET
1999         case AF_INET:
2000                 if ((a->addr32[0] < b->addr32[0]) ||
2001                     (a->addr32[0] > e->addr32[0]))
2002                         return (0);
2003                 break;
2004 #endif /* INET */
2005 #ifdef INET6
2006         case AF_INET6: {
2007                 int     i;
2008
2009                 /* check a >= b */
2010                 for (i = 0; i < 4; ++i)
2011                         if (a->addr32[i] > b->addr32[i])
2012                                 break;
2013                         else if (a->addr32[i] < b->addr32[i])
2014                                 return (0);
2015                 /* check a <= e */
2016                 for (i = 0; i < 4; ++i)
2017                         if (a->addr32[i] < e->addr32[i])
2018                                 break;
2019                         else if (a->addr32[i] > e->addr32[i])
2020                                 return (0);
2021                 break;
2022         }
2023 #endif /* INET6 */
2024         }
2025         return (1);
2026 }
2027
2028 int
2029 pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p)
2030 {
2031         switch (op) {
2032         case PF_OP_IRG:
2033                 return ((p > a1) && (p < a2));
2034         case PF_OP_XRG:
2035                 return ((p < a1) || (p > a2));
2036         case PF_OP_RRG:
2037                 return ((p >= a1) && (p <= a2));
2038         case PF_OP_EQ:
2039                 return (p == a1);
2040         case PF_OP_NE:
2041                 return (p != a1);
2042         case PF_OP_LT:
2043                 return (p < a1);
2044         case PF_OP_LE:
2045                 return (p <= a1);
2046         case PF_OP_GT:
2047                 return (p > a1);
2048         case PF_OP_GE:
2049                 return (p >= a1);
2050         }
2051         return (0); /* never reached */
2052 }
2053
2054 int
2055 pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p)
2056 {
2057         a1 = ntohs(a1);
2058         a2 = ntohs(a2);
2059         p = ntohs(p);
2060         return (pf_match(op, a1, a2, p));
2061 }
2062
2063 int
2064 pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u)
2065 {
2066         if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
2067                 return (0);
2068         return (pf_match(op, a1, a2, u));
2069 }
2070
2071 int
2072 pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g)
2073 {
2074         if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
2075                 return (0);
2076         return (pf_match(op, a1, a2, g));
2077 }
2078
2079 int
2080 pf_match_tag(struct mbuf *m, struct pf_rule *r, int *tag)
2081 {
2082         if (*tag == -1)
2083                 *tag = m->m_pkthdr.pf.tag;
2084
2085         return ((!r->match_tag_not && r->match_tag == *tag) ||
2086             (r->match_tag_not && r->match_tag != *tag));
2087 }
2088
2089 int
2090 pf_tag_packet(struct mbuf *m, int tag, int rtableid)
2091 {
2092         if (tag <= 0 && rtableid < 0)
2093                 return (0);
2094
2095         if (tag > 0)
2096                 m->m_pkthdr.pf.tag = tag;
2097         if (rtableid >= 0)
2098                 m->m_pkthdr.pf.rtableid = rtableid;
2099
2100         return (0);
2101 }
2102
2103 void
2104 pf_step_into_anchor(int *depth, struct pf_ruleset **rs, int n,
2105     struct pf_rule **r, struct pf_rule **a, int *match)
2106 {
2107         struct pf_anchor_stackframe     *f;
2108
2109         (*r)->anchor->match = 0;
2110         if (match)
2111                 *match = 0;
2112         if (*depth >= NELEM(pf_anchor_stack)) {
2113                 kprintf("pf_step_into_anchor: stack overflow\n");
2114                 *r = TAILQ_NEXT(*r, entries);
2115                 return;
2116         } else if (*depth == 0 && a != NULL)
2117                 *a = *r;
2118         f = pf_anchor_stack + (*depth)++;
2119         f->rs = *rs;
2120         f->r = *r;
2121         if ((*r)->anchor_wildcard) {
2122                 f->parent = &(*r)->anchor->children;
2123                 if ((f->child = RB_MIN(pf_anchor_node, f->parent)) ==
2124                     NULL) {
2125                         *r = NULL;
2126                         return;
2127                 }
2128                 *rs = &f->child->ruleset;
2129         } else {
2130                 f->parent = NULL;
2131                 f->child = NULL;
2132                 *rs = &(*r)->anchor->ruleset;
2133         }
2134         *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
2135 }
2136
2137 int
2138 pf_step_out_of_anchor(int *depth, struct pf_ruleset **rs, int n,
2139     struct pf_rule **r, struct pf_rule **a, int *match)
2140 {
2141         struct pf_anchor_stackframe     *f;
2142         int quick = 0;
2143
2144         do {
2145                 if (*depth <= 0)
2146                         break;
2147                 f = pf_anchor_stack + *depth - 1;
2148                 if (f->parent != NULL && f->child != NULL) {
2149                         if (f->child->match ||
2150                             (match != NULL && *match)) {
2151                                 f->r->anchor->match = 1;
2152                                 *match = 0;
2153                         }
2154                         f->child = RB_NEXT(pf_anchor_node, f->parent, f->child);
2155                         if (f->child != NULL) {
2156                                 *rs = &f->child->ruleset;
2157                                 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
2158                                 if (*r == NULL)
2159                                         continue;
2160                                 else
2161                                         break;
2162                         }
2163                 }
2164                 (*depth)--;
2165                 if (*depth == 0 && a != NULL)
2166                         *a = NULL;
2167                 *rs = f->rs;
2168                 if (f->r->anchor->match || (match != NULL && *match))
2169                         quick = f->r->quick;
2170                 *r = TAILQ_NEXT(f->r, entries);
2171         } while (*r == NULL);
2172
2173         return (quick);
2174 }
2175
2176 #ifdef INET6
2177 void
2178 pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr,
2179     struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af)
2180 {
2181         switch (af) {
2182 #ifdef INET
2183         case AF_INET:
2184                 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2185                 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2186                 break;
2187 #endif /* INET */
2188         case AF_INET6:
2189                 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2190                 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2191                 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) |
2192                 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]);
2193                 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) |
2194                 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]);
2195                 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) |
2196                 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]);
2197                 break;
2198         }
2199 }
2200
2201 void
2202 pf_addr_inc(struct pf_addr *addr, sa_family_t af)
2203 {
2204         switch (af) {
2205 #ifdef INET
2206         case AF_INET:
2207                 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1);
2208                 break;
2209 #endif /* INET */
2210         case AF_INET6:
2211                 if (addr->addr32[3] == 0xffffffff) {
2212                         addr->addr32[3] = 0;
2213                         if (addr->addr32[2] == 0xffffffff) {
2214                                 addr->addr32[2] = 0;
2215                                 if (addr->addr32[1] == 0xffffffff) {
2216                                         addr->addr32[1] = 0;
2217                                         addr->addr32[0] =
2218                                             htonl(ntohl(addr->addr32[0]) + 1);
2219                                 } else
2220                                         addr->addr32[1] =
2221                                             htonl(ntohl(addr->addr32[1]) + 1);
2222                         } else
2223                                 addr->addr32[2] =
2224                                     htonl(ntohl(addr->addr32[2]) + 1);
2225                 } else
2226                         addr->addr32[3] =
2227                             htonl(ntohl(addr->addr32[3]) + 1);
2228                 break;
2229         }
2230 }
2231 #endif /* INET6 */
2232
2233 #define mix(a,b,c) \
2234         do {                                    \
2235                 a -= b; a -= c; a ^= (c >> 13); \
2236                 b -= c; b -= a; b ^= (a << 8);  \
2237                 c -= a; c -= b; c ^= (b >> 13); \
2238                 a -= b; a -= c; a ^= (c >> 12); \
2239                 b -= c; b -= a; b ^= (a << 16); \
2240                 c -= a; c -= b; c ^= (b >> 5);  \
2241                 a -= b; a -= c; a ^= (c >> 3);  \
2242                 b -= c; b -= a; b ^= (a << 10); \
2243                 c -= a; c -= b; c ^= (b >> 15); \
2244         } while (0)
2245
2246 /*
2247  * hash function based on bridge_hash in if_bridge.c
2248  */
2249 void
2250 pf_hash(struct pf_addr *inaddr, struct pf_addr *hash,
2251     struct pf_poolhashkey *key, sa_family_t af)
2252 {
2253         u_int32_t       a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0];
2254
2255         switch (af) {
2256 #ifdef INET
2257         case AF_INET:
2258                 a += inaddr->addr32[0];
2259                 b += key->key32[1];
2260                 mix(a, b, c);
2261                 hash->addr32[0] = c + key->key32[2];
2262                 break;
2263 #endif /* INET */
2264 #ifdef INET6
2265         case AF_INET6:
2266                 a += inaddr->addr32[0];
2267                 b += inaddr->addr32[2];
2268                 mix(a, b, c);
2269                 hash->addr32[0] = c;
2270                 a += inaddr->addr32[1];
2271                 b += inaddr->addr32[3];
2272                 c += key->key32[1];
2273                 mix(a, b, c);
2274                 hash->addr32[1] = c;
2275                 a += inaddr->addr32[2];
2276                 b += inaddr->addr32[1];
2277                 c += key->key32[2];
2278                 mix(a, b, c);
2279                 hash->addr32[2] = c;
2280                 a += inaddr->addr32[3];
2281                 b += inaddr->addr32[0];
2282                 c += key->key32[3];
2283                 mix(a, b, c);
2284                 hash->addr32[3] = c;
2285                 break;
2286 #endif /* INET6 */
2287         }
2288 }
2289
2290 int
2291 pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr,
2292     struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn)
2293 {
2294         unsigned char            hash[16];
2295         struct pf_pool          *rpool = &r->rpool;
2296         struct pf_addr          *raddr = &rpool->cur->addr.v.a.addr;
2297         struct pf_addr          *rmask = &rpool->cur->addr.v.a.mask;
2298         struct pf_pooladdr      *acur = rpool->cur;
2299         struct pf_src_node       k;
2300
2301         if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR &&
2302             (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2303                 k.af = af;
2304                 PF_ACPY(&k.addr, saddr, af);
2305                 if (r->rule_flag & PFRULE_RULESRCTRACK ||
2306                     r->rpool.opts & PF_POOL_STICKYADDR)
2307                         k.rule.ptr = r;
2308                 else
2309                         k.rule.ptr = NULL;
2310                 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
2311                 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
2312                 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) {
2313                         PF_ACPY(naddr, &(*sn)->raddr, af);
2314                         if (pf_status.debug >= PF_DEBUG_MISC) {
2315                                 kprintf("pf_map_addr: src tracking maps ");
2316                                 pf_print_host(&k.addr, 0, af);
2317                                 kprintf(" to ");
2318                                 pf_print_host(naddr, 0, af);
2319                                 kprintf("\n");
2320                         }
2321                         return (0);
2322                 }
2323         }
2324
2325         if (rpool->cur->addr.type == PF_ADDR_NOROUTE)
2326                 return (1);
2327         if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2328                 switch (af) {
2329 #ifdef INET
2330                 case AF_INET:
2331                         if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 &&
2332                             (rpool->opts & PF_POOL_TYPEMASK) !=
2333                             PF_POOL_ROUNDROBIN)
2334                                 return (1);
2335                         raddr = &rpool->cur->addr.p.dyn->pfid_addr4;
2336                         rmask = &rpool->cur->addr.p.dyn->pfid_mask4;
2337                         break;
2338 #endif /* INET */
2339 #ifdef INET6
2340                 case AF_INET6:
2341                         if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 &&
2342                             (rpool->opts & PF_POOL_TYPEMASK) !=
2343                             PF_POOL_ROUNDROBIN)
2344                                 return (1);
2345                         raddr = &rpool->cur->addr.p.dyn->pfid_addr6;
2346                         rmask = &rpool->cur->addr.p.dyn->pfid_mask6;
2347                         break;
2348 #endif /* INET6 */
2349                 }
2350         } else if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2351                 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN)
2352                         return (1); /* unsupported */
2353         } else {
2354                 raddr = &rpool->cur->addr.v.a.addr;
2355                 rmask = &rpool->cur->addr.v.a.mask;
2356         }
2357
2358         switch (rpool->opts & PF_POOL_TYPEMASK) {
2359         case PF_POOL_NONE:
2360                 PF_ACPY(naddr, raddr, af);
2361                 break;
2362         case PF_POOL_BITMASK:
2363                 PF_POOLMASK(naddr, raddr, rmask, saddr, af);
2364                 break;
2365         case PF_POOL_RANDOM:
2366                 if (init_addr != NULL && PF_AZERO(init_addr, af)) {
2367                         switch (af) {
2368 #ifdef INET
2369                         case AF_INET:
2370                                 rpool->counter.addr32[0] = htonl(karc4random());
2371                                 break;
2372 #endif /* INET */
2373 #ifdef INET6
2374                         case AF_INET6:
2375                                 if (rmask->addr32[3] != 0xffffffff)
2376                                         rpool->counter.addr32[3] =
2377                                             htonl(karc4random());
2378                                 else
2379                                         break;
2380                                 if (rmask->addr32[2] != 0xffffffff)
2381                                         rpool->counter.addr32[2] =
2382                                             htonl(karc4random());
2383                                 else
2384                                         break;
2385                                 if (rmask->addr32[1] != 0xffffffff)
2386                                         rpool->counter.addr32[1] =
2387                                             htonl(karc4random());
2388                                 else
2389                                         break;
2390                                 if (rmask->addr32[0] != 0xffffffff)
2391                                         rpool->counter.addr32[0] =
2392                                             htonl(karc4random());
2393                                 break;
2394 #endif /* INET6 */
2395                         }
2396                         PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2397                         PF_ACPY(init_addr, naddr, af);
2398
2399                 } else {
2400                         PF_AINC(&rpool->counter, af);
2401                         PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2402                 }
2403                 break;
2404         case PF_POOL_SRCHASH:
2405                 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af);
2406                 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af);
2407                 break;
2408         case PF_POOL_ROUNDROBIN:
2409                 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2410                         if (!pfr_pool_get(rpool->cur->addr.p.tbl,
2411                             &rpool->tblidx, &rpool->counter,
2412                             &raddr, &rmask, af))
2413                                 goto get_addr;
2414                 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2415                         if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2416                             &rpool->tblidx, &rpool->counter,
2417                             &raddr, &rmask, af))
2418                                 goto get_addr;
2419                 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af))
2420                         goto get_addr;
2421
2422         try_next:
2423                 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL)
2424                         rpool->cur = TAILQ_FIRST(&rpool->list);
2425                 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2426                         rpool->tblidx = -1;
2427                         if (pfr_pool_get(rpool->cur->addr.p.tbl,
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 if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2436                         rpool->tblidx = -1;
2437                         if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2438                             &rpool->tblidx, &rpool->counter,
2439                             &raddr, &rmask, af)) {
2440                                 /* table contains no address of type 'af' */
2441                                 if (rpool->cur != acur)
2442                                         goto try_next;
2443                                 return (1);
2444                         }
2445                 } else {
2446                         raddr = &rpool->cur->addr.v.a.addr;
2447                         rmask = &rpool->cur->addr.v.a.mask;
2448                         PF_ACPY(&rpool->counter, raddr, af);
2449                 }
2450
2451         get_addr:
2452                 PF_ACPY(naddr, &rpool->counter, af);
2453                 if (init_addr != NULL && PF_AZERO(init_addr, af))
2454                         PF_ACPY(init_addr, naddr, af);
2455                 PF_AINC(&rpool->counter, af);
2456                 break;
2457         }
2458         if (*sn != NULL)
2459                 PF_ACPY(&(*sn)->raddr, naddr, af);
2460
2461         if (pf_status.debug >= PF_DEBUG_MISC &&
2462             (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2463                 kprintf("pf_map_addr: selected address ");
2464                 pf_print_host(naddr, 0, af);
2465                 kprintf("\n");
2466         }
2467
2468         return (0);
2469 }
2470
2471 int
2472 pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r,
2473     struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport,
2474     struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high,
2475     struct pf_src_node **sn)
2476 {
2477         struct pf_state_key_cmp key;
2478         struct pf_addr          init_addr;
2479         u_int16_t               cut;
2480
2481         bzero(&init_addr, sizeof(init_addr));
2482         if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2483                 return (1);
2484
2485         if (proto == IPPROTO_ICMP) {
2486                 low = 1;
2487                 high = 65535;
2488         }
2489
2490         do {
2491                 key.af = af;
2492                 key.proto = proto;
2493                 PF_ACPY(&key.addr[1], daddr, key.af);
2494                 PF_ACPY(&key.addr[0], naddr, key.af);
2495                 key.port[1] = dport;
2496
2497                 /*
2498                  * port search; start random, step;
2499                  * similar 2 portloop in in_pcbbind
2500                  */
2501                 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP ||
2502                     proto == IPPROTO_ICMP)) {
2503                         key.port[0] = dport;
2504                         if (pf_find_state_all(&key, PF_IN, NULL) == NULL)
2505                                 return (0);
2506                 } else if (low == 0 && high == 0) {
2507                         key.port[0] = *nport;
2508                         if (pf_find_state_all(&key, PF_IN, NULL) == NULL)
2509                                 return (0);
2510                 } else if (low == high) {
2511                         key.port[0] = htons(low);
2512                         if (pf_find_state_all(&key, PF_IN, NULL) == NULL) {
2513                                 *nport = htons(low);
2514                                 return (0);
2515                         }
2516                 } else {
2517                         u_int16_t tmp;
2518
2519                         if (low > high) {
2520                                 tmp = low;
2521                                 low = high;
2522                                 high = tmp;
2523                         }
2524                         /* low < high */
2525                         cut = htonl(karc4random()) % (1 + high - low) + low;
2526                         /* low <= cut <= high */
2527                         for (tmp = cut; tmp <= high; ++(tmp)) {
2528                                 key.port[0] = htons(tmp);
2529                                 if (pf_find_state_all(&key, PF_IN, NULL) ==
2530                                     NULL && !in_baddynamic(tmp, proto)) {
2531                                         *nport = htons(tmp);
2532                                         return (0);
2533                                 }
2534                         }
2535                         for (tmp = cut - 1; tmp >= low; --(tmp)) {
2536                                 key.port[0] = htons(tmp);
2537                                 if (pf_find_state_all(&key, PF_IN, NULL) ==
2538                                     NULL && !in_baddynamic(tmp, proto)) {
2539                                         *nport = htons(tmp);
2540                                         return (0);
2541                                 }
2542                         }
2543                 }
2544
2545                 switch (r->rpool.opts & PF_POOL_TYPEMASK) {
2546                 case PF_POOL_RANDOM:
2547                 case PF_POOL_ROUNDROBIN:
2548                         if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2549                                 return (1);
2550                         break;
2551                 case PF_POOL_NONE:
2552                 case PF_POOL_SRCHASH:
2553                 case PF_POOL_BITMASK:
2554                 default:
2555                         return (1);
2556                 }
2557         } while (! PF_AEQ(&init_addr, naddr, af) );
2558         return (1);                                     /* none available */
2559 }
2560
2561 struct pf_rule *
2562 pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off,
2563     int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport,
2564     struct pf_addr *daddr, u_int16_t dport, int rs_num)
2565 {
2566         struct pf_rule          *r, *rm = NULL;
2567         struct pf_ruleset       *ruleset = NULL;
2568         int                      tag = -1;
2569         int                      rtableid = -1;
2570         int                      asd = 0;
2571
2572         r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr);
2573         while (r && rm == NULL) {
2574                 struct pf_rule_addr     *src = NULL, *dst = NULL;
2575                 struct pf_addr_wrap     *xdst = NULL;
2576
2577                 if (r->action == PF_BINAT && direction == PF_IN) {
2578                         src = &r->dst;
2579                         if (r->rpool.cur != NULL)
2580                                 xdst = &r->rpool.cur->addr;
2581                 } else {
2582                         src = &r->src;
2583                         dst = &r->dst;
2584                 }
2585
2586                 r->evaluations++;
2587                 if (pfi_kif_match(r->kif, kif) == r->ifnot)
2588                         r = r->skip[PF_SKIP_IFP].ptr;
2589                 else if (r->direction && r->direction != direction)
2590                         r = r->skip[PF_SKIP_DIR].ptr;
2591                 else if (r->af && r->af != pd->af)
2592                         r = r->skip[PF_SKIP_AF].ptr;
2593                 else if (r->proto && r->proto != pd->proto)
2594                         r = r->skip[PF_SKIP_PROTO].ptr;
2595                 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af,
2596                     src->neg, kif))
2597                         r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR :
2598                             PF_SKIP_DST_ADDR].ptr;
2599                 else if (src->port_op && !pf_match_port(src->port_op,
2600                     src->port[0], src->port[1], sport))
2601                         r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT :
2602                             PF_SKIP_DST_PORT].ptr;
2603                 else if (dst != NULL &&
2604                     PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->neg, NULL))
2605                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
2606                 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af,
2607                     0, NULL))
2608                         r = TAILQ_NEXT(r, entries);
2609                 else if (dst != NULL && dst->port_op &&
2610                     !pf_match_port(dst->port_op, dst->port[0],
2611                     dst->port[1], dport))
2612                         r = r->skip[PF_SKIP_DST_PORT].ptr;
2613                 else if (r->match_tag && !pf_match_tag(m, r, &tag))
2614                         r = TAILQ_NEXT(r, entries);
2615                 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto !=
2616                     IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m,
2617                     off, pd->hdr.tcp), r->os_fingerprint)))
2618                         r = TAILQ_NEXT(r, entries);
2619                 else {
2620                         if (r->tag)
2621                                 tag = r->tag;
2622                         if (r->rtableid >= 0)
2623                                 rtableid = r->rtableid;
2624                         if (r->anchor == NULL) {
2625                                 rm = r;
2626                         } else
2627                                 pf_step_into_anchor(&asd, &ruleset, rs_num,
2628                                     &r, NULL, NULL);
2629                 }
2630                 if (r == NULL)
2631                         pf_step_out_of_anchor(&asd, &ruleset, rs_num, &r,
2632                             NULL, NULL);
2633         }
2634         if (pf_tag_packet(m, tag, rtableid))
2635                 return (NULL);
2636         if (rm != NULL && (rm->action == PF_NONAT ||
2637             rm->action == PF_NORDR || rm->action == PF_NOBINAT))
2638                 return (NULL);
2639         return (rm);
2640 }
2641
2642 struct pf_rule *
2643 pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction,
2644     struct pfi_kif *kif, struct pf_src_node **sn,
2645     struct pf_state_key **skw, struct pf_state_key **sks,
2646     struct pf_state_key **skp, struct pf_state_key **nkp,
2647     struct pf_addr *saddr, struct pf_addr *daddr,
2648     u_int16_t sport, u_int16_t dport)
2649 {
2650         struct pf_rule  *r = NULL;
2651
2652
2653         if (direction == PF_OUT) {
2654                 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2655                     sport, daddr, dport, PF_RULESET_BINAT);
2656                 if (r == NULL)
2657                         r = pf_match_translation(pd, m, off, direction, kif,
2658                             saddr, sport, daddr, dport, PF_RULESET_NAT);
2659         } else {
2660                 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2661                     sport, daddr, dport, PF_RULESET_RDR);
2662                 if (r == NULL)
2663                         r = pf_match_translation(pd, m, off, direction, kif,
2664                             saddr, sport, daddr, dport, PF_RULESET_BINAT);
2665         }
2666
2667         if (r != NULL) {
2668                 struct pf_addr  *naddr;
2669                 u_int16_t       *nport;
2670
2671                 if (pf_state_key_setup(pd, r, skw, sks, skp, nkp,
2672                     saddr, daddr, sport, dport))
2673                         return r;
2674
2675                 /* XXX We only modify one side for now. */
2676                 naddr = &(*nkp)->addr[1];
2677                 nport = &(*nkp)->port[1];
2678
2679                 /*
2680                  * NOTE: Currently all translations will clear
2681                  *       BRIDGE_MBUF_TAGGED, telling the bridge to
2682                  *       ignore the original input encapsulation.
2683                  */
2684                 switch (r->action) {
2685                 case PF_NONAT:
2686                 case PF_NOBINAT:
2687                 case PF_NORDR:
2688                         return (NULL);
2689                 case PF_NAT:
2690                         m->m_pkthdr.fw_flags &= ~BRIDGE_MBUF_TAGGED;
2691                         if (pf_get_sport(pd->af, pd->proto, r, saddr,
2692                             daddr, dport, naddr, nport, r->rpool.proxy_port[0],
2693                             r->rpool.proxy_port[1], sn)) {
2694                                 DPFPRINTF(PF_DEBUG_MISC,
2695                                     ("pf: NAT proxy port allocation "
2696                                     "(%u-%u) failed\n",
2697                                     r->rpool.proxy_port[0],
2698                                     r->rpool.proxy_port[1]));
2699                                 return (NULL);
2700                         }
2701                         break;
2702                 case PF_BINAT:
2703                         m->m_pkthdr.fw_flags &= ~BRIDGE_MBUF_TAGGED;
2704                         switch (direction) {
2705                         case PF_OUT:
2706                                 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){
2707                                         switch (pd->af) {
2708 #ifdef INET
2709                                         case AF_INET:
2710                                                 if (r->rpool.cur->addr.p.dyn->
2711                                                     pfid_acnt4 < 1)
2712                                                         return (NULL);
2713                                                 PF_POOLMASK(naddr,
2714                                                     &r->rpool.cur->addr.p.dyn->
2715                                                     pfid_addr4,
2716                                                     &r->rpool.cur->addr.p.dyn->
2717                                                     pfid_mask4,
2718                                                     saddr, AF_INET);
2719                                                 break;
2720 #endif /* INET */
2721 #ifdef INET6
2722                                         case AF_INET6:
2723                                                 if (r->rpool.cur->addr.p.dyn->
2724                                                     pfid_acnt6 < 1)
2725                                                         return (NULL);
2726                                                 PF_POOLMASK(naddr,
2727                                                     &r->rpool.cur->addr.p.dyn->
2728                                                     pfid_addr6,
2729                                                     &r->rpool.cur->addr.p.dyn->
2730                                                     pfid_mask6,
2731                                                     saddr, AF_INET6);
2732                                                 break;
2733 #endif /* INET6 */
2734                                         }
2735                                 } else
2736                                         PF_POOLMASK(naddr,
2737                                             &r->rpool.cur->addr.v.a.addr,
2738                                             &r->rpool.cur->addr.v.a.mask,
2739                                             saddr, pd->af);
2740                                 break;
2741                         case PF_IN:
2742                                 if (r->src.addr.type == PF_ADDR_DYNIFTL) {
2743                                         switch (pd->af) {
2744 #ifdef INET
2745                                         case AF_INET:
2746                                                 if (r->src.addr.p.dyn->
2747                                                     pfid_acnt4 < 1)
2748                                                         return (NULL);
2749                                                 PF_POOLMASK(naddr,
2750                                                     &r->src.addr.p.dyn->
2751                                                     pfid_addr4,
2752                                                     &r->src.addr.p.dyn->
2753                                                     pfid_mask4,
2754                                                     daddr, AF_INET);
2755                                                 break;
2756 #endif /* INET */
2757 #ifdef INET6
2758                                         case AF_INET6:
2759                                                 if (r->src.addr.p.dyn->
2760                                                     pfid_acnt6 < 1)
2761                                                         return (NULL);
2762                                                 PF_POOLMASK(naddr,
2763                                                     &r->src.addr.p.dyn->
2764                                                     pfid_addr6,
2765                                                     &r->src.addr.p.dyn->
2766                                                     pfid_mask6,
2767                                                     daddr, AF_INET6);
2768                                                 break;
2769 #endif /* INET6 */
2770                                         }
2771                                 } else
2772                                         PF_POOLMASK(naddr,
2773                                             &r->src.addr.v.a.addr,
2774                                             &r->src.addr.v.a.mask, daddr,
2775                                             pd->af);
2776                                 break;
2777                         }
2778                         break;
2779                 case PF_RDR: {
2780                         m->m_pkthdr.fw_flags &= ~BRIDGE_MBUF_TAGGED;
2781                         if (pf_map_addr(pd->af, r, saddr, naddr, NULL, sn))
2782                                 return (NULL);
2783                         if ((r->rpool.opts & PF_POOL_TYPEMASK) ==
2784                             PF_POOL_BITMASK)
2785                                 PF_POOLMASK(naddr, naddr,
2786                                     &r->rpool.cur->addr.v.a.mask, daddr,
2787                                     pd->af);
2788
2789                         if (r->rpool.proxy_port[1]) {
2790                                 u_int32_t       tmp_nport;
2791
2792                                 tmp_nport = ((ntohs(dport) -
2793                                     ntohs(r->dst.port[0])) %
2794                                     (r->rpool.proxy_port[1] -
2795                                     r->rpool.proxy_port[0] + 1)) +
2796                                     r->rpool.proxy_port[0];
2797
2798                                 /* wrap around if necessary */
2799                                 if (tmp_nport > 65535)
2800                                         tmp_nport -= 65535;
2801                                 *nport = htons((u_int16_t)tmp_nport);
2802                         } else if (r->rpool.proxy_port[0])
2803                                 *nport = htons(r->rpool.proxy_port[0]);
2804                         break;
2805                 }
2806                 default:
2807                         return (NULL);
2808                 }
2809         }
2810
2811         return (r);
2812 }
2813
2814 struct netmsg_hashlookup {
2815         struct netmsg_base      base;
2816         struct inpcb            **nm_pinp;
2817         struct inpcbinfo        *nm_pcbinfo;
2818         struct pf_addr          *nm_saddr;
2819         struct pf_addr          *nm_daddr;
2820         uint16_t                nm_sport;
2821         uint16_t                nm_dport;
2822         sa_family_t             nm_af;
2823 };
2824
2825 #ifdef PF_SOCKET_LOOKUP_DOMSG
2826 static void
2827 in_pcblookup_hash_handler(netmsg_t msg)
2828 {
2829         struct netmsg_hashlookup *rmsg = (struct netmsg_hashlookup *)msg;
2830
2831         if (rmsg->nm_af == AF_INET)
2832                 *rmsg->nm_pinp = in_pcblookup_hash(rmsg->nm_pcbinfo,
2833                     rmsg->nm_saddr->v4, rmsg->nm_sport, rmsg->nm_daddr->v4,
2834                     rmsg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2835 #ifdef INET6
2836         else
2837                 *rmsg->nm_pinp = in6_pcblookup_hash(rmsg->nm_pcbinfo,
2838                     &rmsg->nm_saddr->v6, rmsg->nm_sport, &rmsg->nm_daddr->v6,
2839                     rmsg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2840 #endif /* INET6 */
2841         lwkt_replymsg(&rmsg->base.lmsg, 0);
2842 }
2843 #endif  /* PF_SOCKET_LOOKUP_DOMSG */
2844
2845 int
2846 pf_socket_lookup(int direction, struct pf_pdesc *pd)
2847 {
2848         struct pf_addr          *saddr, *daddr;
2849         u_int16_t                sport, dport;
2850         struct inpcbinfo        *pi;
2851         struct inpcb            *inp;
2852         struct netmsg_hashlookup *msg = NULL;
2853 #ifdef PF_SOCKET_LOOKUP_DOMSG
2854         struct netmsg_hashlookup msg0;
2855 #endif
2856         int                      pi_cpu = 0;
2857
2858         if (pd == NULL)
2859                 return (-1);
2860         pd->lookup.uid = UID_MAX;
2861         pd->lookup.gid = GID_MAX;
2862         pd->lookup.pid = NO_PID;
2863         if (direction == PF_IN) {
2864                 saddr = pd->src;
2865                 daddr = pd->dst;
2866         } else {
2867                 saddr = pd->dst;
2868                 daddr = pd->src;
2869         }
2870         switch (pd->proto) {
2871         case IPPROTO_TCP:
2872                 if (pd->hdr.tcp == NULL)
2873                         return (-1);
2874                 sport = pd->hdr.tcp->th_sport;
2875                 dport = pd->hdr.tcp->th_dport;
2876
2877                 pi_cpu = tcp_addrcpu(saddr->v4.s_addr, sport, daddr->v4.s_addr, dport);
2878                 pi = &tcbinfo[pi_cpu];
2879                 /*
2880                  * Our netstack runs lockless on MP systems
2881                  * (only for TCP connections at the moment).
2882                  *
2883                  * As we are not allowed to read another CPU's tcbinfo,
2884                  * we have to ask that CPU via remote call to search the
2885                  * table for us.
2886                  *
2887                  * Prepare a msg iff data belongs to another CPU.
2888                  */
2889                 if (pi_cpu != mycpu->gd_cpuid) {
2890 #ifdef PF_SOCKET_LOOKUP_DOMSG
2891                         /*
2892                          * NOTE:
2893                          *
2894                          * Following lwkt_domsg() is dangerous and could
2895                          * lockup the network system, e.g.
2896                          *
2897                          * On 2 CPU system:
2898                          * netisr0 domsg to netisr1 (due to lookup)
2899                          * netisr1 domsg to netisr0 (due to lookup)
2900                          *
2901                          * We simply return -1 here, since we are probably
2902                          * called before NAT, so the TCP packet should
2903                          * already be on the correct CPU.
2904                          */
2905                         msg = &msg0;
2906                         netmsg_init(&msg->base, NULL, &curthread->td_msgport,
2907                                     0, in_pcblookup_hash_handler);
2908                         msg->nm_pinp = &inp;
2909                         msg->nm_pcbinfo = pi;
2910                         msg->nm_saddr = saddr;
2911                         msg->nm_sport = sport;
2912                         msg->nm_daddr = daddr;
2913                         msg->nm_dport = dport;
2914                         msg->nm_af = pd->af;
2915 #else   /* !PF_SOCKET_LOOKUP_DOMSG */
2916                         kprintf("pf_socket_lookup: tcp packet not on the "
2917                                 "correct cpu %d, cur cpu %d\n",
2918                                 pi_cpu, mycpuid);
2919                         print_backtrace(-1);
2920                         return -1;
2921 #endif  /* PF_SOCKET_LOOKUP_DOMSG */
2922                 }
2923                 break;
2924         case IPPROTO_UDP:
2925                 if (pd->hdr.udp == NULL)
2926                         return (-1);
2927                 sport = pd->hdr.udp->uh_sport;
2928                 dport = pd->hdr.udp->uh_dport;
2929                 pi = &udbinfo;
2930                 break;
2931         default:
2932                 return (-1);
2933         }
2934         if (direction != PF_IN) {
2935                 u_int16_t       p;
2936
2937                 p = sport;
2938                 sport = dport;
2939                 dport = p;
2940         }
2941         switch (pd->af) {
2942 #ifdef INET6
2943         case AF_INET6:
2944                 /*
2945                  * Query other CPU, second part
2946                  *
2947                  * msg only gets initialized when:
2948                  * 1) packet is TCP
2949                  * 2) the info belongs to another CPU
2950                  *
2951                  * Use some switch/case magic to avoid code duplication.
2952                  */
2953                 if (msg == NULL) {
2954                         inp = in6_pcblookup_hash(pi, &saddr->v6, sport,
2955                             &daddr->v6, dport, INPLOOKUP_WILDCARD, NULL);
2956
2957                         if (inp == NULL)
2958                                 return (-1);
2959                         break;
2960                 }
2961                 /* FALLTHROUGH if SMP and on other CPU */
2962 #endif /* INET6 */
2963         case AF_INET:
2964                 if (msg != NULL) {
2965                         lwkt_domsg(netisr_cpuport(pi_cpu),
2966                                      &msg->base.lmsg, 0);
2967                 } else
2968                 {
2969                         inp = in_pcblookup_hash(pi, saddr->v4, sport, daddr->v4,
2970                             dport, INPLOOKUP_WILDCARD, NULL);
2971                 }
2972                 if (inp == NULL)
2973                         return (-1);
2974                 break;
2975
2976         default:
2977                 return (-1);
2978         }
2979         pd->lookup.uid = inp->inp_socket->so_cred->cr_uid;
2980         pd->lookup.gid = inp->inp_socket->so_cred->cr_groups[0];
2981         return (1);
2982 }
2983
2984 u_int8_t
2985 pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2986 {
2987         int              hlen;
2988         u_int8_t         hdr[60];
2989         u_int8_t        *opt, optlen;
2990         u_int8_t         wscale = 0;
2991
2992         hlen = th_off << 2;             /* hlen <= sizeof(hdr) */
2993         if (hlen <= sizeof(struct tcphdr))
2994                 return (0);
2995         if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2996                 return (0);
2997         opt = hdr + sizeof(struct tcphdr);
2998         hlen -= sizeof(struct tcphdr);
2999         while (hlen >= 3) {
3000                 switch (*opt) {
3001                 case TCPOPT_EOL:
3002                 case TCPOPT_NOP:
3003                         ++opt;
3004                         --hlen;
3005                         break;
3006                 case TCPOPT_WINDOW:
3007                         wscale = opt[2];
3008                         if (wscale > TCP_MAX_WINSHIFT)
3009                                 wscale = TCP_MAX_WINSHIFT;
3010                         wscale |= PF_WSCALE_FLAG;
3011                         /* FALLTHROUGH */
3012                 default:
3013                         optlen = opt[1];
3014                         if (optlen < 2)
3015                                 optlen = 2;
3016                         hlen -= optlen;
3017                         opt += optlen;
3018                         break;
3019                 }
3020         }
3021         return (wscale);
3022 }
3023
3024 u_int16_t
3025 pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
3026 {
3027         int              hlen;
3028         u_int8_t         hdr[60];
3029         u_int8_t        *opt, optlen;
3030         u_int16_t        mss = tcp_mssdflt;
3031
3032         hlen = th_off << 2;     /* hlen <= sizeof(hdr) */
3033         if (hlen <= sizeof(struct tcphdr))
3034                 return (0);
3035         if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
3036                 return (0);
3037         opt = hdr + sizeof(struct tcphdr);
3038         hlen -= sizeof(struct tcphdr);
3039         while (hlen >= TCPOLEN_MAXSEG) {
3040                 switch (*opt) {
3041                 case TCPOPT_EOL:
3042                 case TCPOPT_NOP:
3043                         ++opt;
3044                         --hlen;
3045                         break;
3046                 case TCPOPT_MAXSEG:
3047                         bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2);
3048                         /* FALLTHROUGH */
3049                 default:
3050                         optlen = opt[1];
3051                         if (optlen < 2)
3052                                 optlen = 2;
3053                         hlen -= optlen;
3054                         opt += optlen;
3055                         break;
3056                 }
3057         }
3058         return (mss);
3059 }
3060
3061 u_int16_t
3062 pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer)
3063 {
3064 #ifdef INET
3065         struct sockaddr_in      *dst;
3066         struct route             ro;
3067 #endif /* INET */
3068 #ifdef INET6
3069         struct sockaddr_in6     *dst6;
3070         struct route_in6         ro6;
3071 #endif /* INET6 */
3072         struct rtentry          *rt = NULL;
3073         int                      hlen = 0;
3074         u_int16_t                mss = tcp_mssdflt;
3075
3076         switch (af) {
3077 #ifdef INET
3078         case AF_INET:
3079                 hlen = sizeof(struct ip);
3080                 bzero(&ro, sizeof(ro));
3081                 dst = (struct sockaddr_in *)&ro.ro_dst;
3082                 dst->sin_family = AF_INET;
3083                 dst->sin_len = sizeof(*dst);
3084                 dst->sin_addr = addr->v4;
3085                 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
3086                 rt = ro.ro_rt;
3087                 break;
3088 #endif /* INET */
3089 #ifdef INET6
3090         case AF_INET6:
3091                 hlen = sizeof(struct ip6_hdr);
3092                 bzero(&ro6, sizeof(ro6));
3093                 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst;
3094                 dst6->sin6_family = AF_INET6;
3095                 dst6->sin6_len = sizeof(*dst6);
3096                 dst6->sin6_addr = addr->v6;
3097                 rtalloc_ign((struct route *)&ro6, (RTF_CLONING | RTF_PRCLONING));
3098                 rt = ro6.ro_rt;
3099                 break;
3100 #endif /* INET6 */
3101         }
3102
3103         if (rt && rt->rt_ifp) {
3104                 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr);
3105                 mss = max(tcp_mssdflt, mss);
3106                 RTFREE(rt);
3107         }
3108         mss = min(mss, offer);
3109         mss = max(mss, 64);             /* sanity - at least max opt space */
3110         return (mss);
3111 }
3112
3113 void
3114 pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr)
3115 {
3116         struct pf_rule *r = s->rule.ptr;
3117
3118         s->rt_kif = NULL;
3119         if (!r->rt || r->rt == PF_FASTROUTE)
3120                 return;
3121         switch (s->key[PF_SK_WIRE]->af) {
3122 #ifdef INET
3123         case AF_INET:
3124                 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL,
3125                     &s->nat_src_node);
3126                 s->rt_kif = r->rpool.cur->kif;
3127                 break;
3128 #endif /* INET */
3129 #ifdef INET6
3130         case AF_INET6:
3131                 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL,
3132                     &s->nat_src_node);
3133                 s->rt_kif = r->rpool.cur->kif;
3134                 break;
3135 #endif /* INET6 */
3136         }
3137 }
3138
3139 u_int32_t
3140 pf_tcp_iss(struct pf_pdesc *pd)
3141 {
3142         MD5_CTX ctx;
3143         u_int32_t digest[4];
3144
3145         if (pf_tcp_secret_init == 0) {
3146                 karc4rand(pf_tcp_secret, sizeof(pf_tcp_secret));
3147                 MD5Init(&pf_tcp_secret_ctx);
3148                 MD5Update(&pf_tcp_secret_ctx, pf_tcp_secret,
3149                     sizeof(pf_tcp_secret));
3150                 pf_tcp_secret_init = 1;
3151         }
3152         ctx = pf_tcp_secret_ctx;
3153
3154         MD5Update(&ctx, (char *)&pd->hdr.tcp->th_sport, sizeof(u_short));
3155         MD5Update(&ctx, (char *)&pd->hdr.tcp->th_dport, sizeof(u_short));
3156         if (pd->af == AF_INET6) {
3157                 MD5Update(&ctx, (char *)&pd->src->v6, sizeof(struct in6_addr));
3158                 MD5Update(&ctx, (char *)&pd->dst->v6, sizeof(struct in6_addr));
3159         } else {
3160                 MD5Update(&ctx, (char *)&pd->src->v4, sizeof(struct in_addr));
3161                 MD5Update(&ctx, (char *)&pd->dst->v4, sizeof(struct in_addr));
3162         }
3163         MD5Final((u_char *)digest, &ctx);
3164         pf_tcp_iss_off += 4096;
3165         return (digest[0] + pd->hdr.tcp->th_seq + pf_tcp_iss_off);
3166 }
3167
3168 int
3169 pf_test_rule(struct pf_rule **rm, struct pf_state **sm, int direction,
3170     struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3171     struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
3172     struct ifqueue *ifq, struct inpcb *inp)
3173 {
3174         struct pf_rule          *nr = NULL;
3175         struct pf_addr          *saddr = pd->src, *daddr = pd->dst;
3176         sa_family_t              af = pd->af;
3177         struct pf_rule          *r, *a = NULL;
3178         struct pf_ruleset       *ruleset = NULL;
3179         struct pf_src_node      *nsn = NULL;
3180         struct tcphdr           *th = pd->hdr.tcp;
3181         struct pf_state_key     *skw = NULL, *sks = NULL;
3182         struct pf_state_key     *sk = NULL, *nk = NULL;
3183         u_short                  reason;
3184         int                      rewrite = 0, hdrlen = 0;
3185         int                      tag = -1, rtableid = -1;
3186         int                      asd = 0;
3187         int                      match = 0;
3188         int                      state_icmp = 0;
3189         u_int16_t                sport = 0, dport = 0;
3190         u_int16_t                bproto_sum = 0, bip_sum = 0;
3191         u_int8_t                 icmptype = 0, icmpcode = 0;
3192
3193
3194         if (direction == PF_IN && pf_check_congestion(ifq)) {
3195                 REASON_SET(&reason, PFRES_CONGEST);
3196                 return (PF_DROP);
3197         }
3198
3199         if (inp != NULL)
3200                 pd->lookup.done = pf_socket_lookup(direction, pd);
3201         else if (debug_pfugidhack) {
3202                 DPFPRINTF(PF_DEBUG_MISC, ("pf: unlocked lookup\n"));
3203                 pd->lookup.done = pf_socket_lookup(direction, pd);
3204         }
3205
3206         switch (pd->proto) {
3207         case IPPROTO_TCP:
3208                 sport = th->th_sport;
3209                 dport = th->th_dport;
3210                 hdrlen = sizeof(*th);
3211                 break;
3212         case IPPROTO_UDP:
3213                 sport = pd->hdr.udp->uh_sport;
3214                 dport = pd->hdr.udp->uh_dport;
3215                 hdrlen = sizeof(*pd->hdr.udp);
3216                 break;
3217 #ifdef INET
3218         case IPPROTO_ICMP:
3219                 if (pd->af != AF_INET)
3220                         break;
3221                 sport = dport = pd->hdr.icmp->icmp_id;
3222                 hdrlen = sizeof(*pd->hdr.icmp);
3223                 icmptype = pd->hdr.icmp->icmp_type;
3224                 icmpcode = pd->hdr.icmp->icmp_code;
3225
3226                 if (icmptype == ICMP_UNREACH ||
3227                     icmptype == ICMP_SOURCEQUENCH ||
3228                     icmptype == ICMP_REDIRECT ||
3229                     icmptype == ICMP_TIMXCEED ||
3230                     icmptype == ICMP_PARAMPROB)
3231                         state_icmp++;
3232                 break;
3233 #endif /* INET */
3234 #ifdef INET6
3235         case IPPROTO_ICMPV6:
3236                 if (af != AF_INET6)
3237                         break;
3238                 sport = dport = pd->hdr.icmp6->icmp6_id;
3239                 hdrlen = sizeof(*pd->hdr.icmp6);
3240                 icmptype = pd->hdr.icmp6->icmp6_type;
3241                 icmpcode = pd->hdr.icmp6->icmp6_code;
3242
3243                 if (icmptype == ICMP6_DST_UNREACH ||
3244                     icmptype == ICMP6_PACKET_TOO_BIG ||
3245                     icmptype == ICMP6_TIME_EXCEEDED ||
3246                     icmptype == ICMP6_PARAM_PROB)
3247                         state_icmp++;
3248                 break;
3249 #endif /* INET6 */
3250         default:
3251                 sport = dport = hdrlen = 0;
3252                 break;
3253         }
3254
3255         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3256
3257         /* check packet for BINAT/NAT/RDR */
3258         if ((nr = pf_get_translation(pd, m, off, direction, kif, &nsn,
3259             &skw, &sks, &sk, &nk, saddr, daddr, sport, dport)) != NULL) {
3260                 if (nk == NULL || sk == NULL) {
3261                         REASON_SET(&reason, PFRES_MEMORY);
3262                         goto cleanup;
3263                 }
3264
3265                 if (pd->ip_sum)
3266                         bip_sum = *pd->ip_sum;
3267
3268                 m->m_flags &= ~M_HASH;
3269                 switch (pd->proto) {
3270                 case IPPROTO_TCP:
3271                         bproto_sum = th->th_sum;
3272                         pd->proto_sum = &th->th_sum;
3273
3274                         if (PF_ANEQ(saddr, &nk->addr[pd->sidx], af) ||
3275                             nk->port[pd->sidx] != sport) {
3276                                 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
3277                                     &th->th_sum, &nk->addr[pd->sidx],
3278                                     nk->port[pd->sidx], 0, af);
3279                                 pd->sport = &th->th_sport;
3280                                 sport = th->th_sport;
3281                         }
3282
3283                         if (PF_ANEQ(daddr, &nk->addr[pd->didx], af) ||
3284                             nk->port[pd->didx] != dport) {
3285                                 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
3286                                     &th->th_sum, &nk->addr[pd->didx],
3287                                     nk->port[pd->didx], 0, af);
3288                                 dport = th->th_dport;
3289                                 pd->dport = &th->th_dport;
3290                         }
3291                         rewrite++;
3292                         break;
3293                 case IPPROTO_UDP:
3294                         bproto_sum = pd->hdr.udp->uh_sum;
3295                         pd->proto_sum = &pd->hdr.udp->uh_sum;
3296
3297                         if (PF_ANEQ(saddr, &nk->addr[pd->sidx], af) ||
3298                             nk->port[pd->sidx] != sport) {
3299                                 pf_change_ap(saddr, &pd->hdr.udp->uh_sport,
3300                                     pd->ip_sum, &pd->hdr.udp->uh_sum,
3301                                     &nk->addr[pd->sidx],
3302                                     nk->port[pd->sidx], 1, af);
3303                                 sport = pd->hdr.udp->uh_sport;
3304                                 pd->sport = &pd->hdr.udp->uh_sport;
3305                         }
3306
3307                         if (PF_ANEQ(daddr, &nk->addr[pd->didx], af) ||
3308                             nk->port[pd->didx] != dport) {
3309                                 pf_change_ap(daddr, &pd->hdr.udp->uh_dport,
3310                                     pd->ip_sum, &pd->hdr.udp->uh_sum,
3311                                     &nk->addr[pd->didx],
3312                                     nk->port[pd->didx], 1, af);
3313                                 dport = pd->hdr.udp->uh_dport;
3314                                 pd->dport = &pd->hdr.udp->uh_dport;
3315                         }
3316                         rewrite++;
3317                         break;
3318 #ifdef INET
3319                 case IPPROTO_ICMP:
3320                         nk->port[0] = nk->port[1];
3321                         if (PF_ANEQ(saddr, &nk->addr[pd->sidx], AF_INET))
3322                                 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3323                                     nk->addr[pd->sidx].v4.s_addr, 0);
3324
3325                         if (PF_ANEQ(daddr, &nk->addr[pd->didx], AF_INET))
3326                                 pf_change_a(&daddr->v4.s_addr, pd->ip_sum,
3327                                     nk->addr[pd->didx].v4.s_addr, 0);
3328
3329                         if (nk->port[1] != pd->hdr.icmp->icmp_id) {
3330                                 pd->hdr.icmp->icmp_cksum = pf_cksum_fixup(
3331                                     pd->hdr.icmp->icmp_cksum, sport,
3332                                     nk->port[1], 0);
3333                                 pd->hdr.icmp->icmp_id = nk->port[1];
3334                                 pd->sport = &pd->hdr.icmp->icmp_id;
3335                         }
3336                         m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
3337                         break;
3338 #endif /* INET */
3339 #ifdef INET6
3340                 case IPPROTO_ICMPV6:
3341                         nk->port[0] = nk->port[1];
3342                         if (PF_ANEQ(saddr, &nk->addr[pd->sidx], AF_INET6))
3343                                 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum,
3344                                     &nk->addr[pd->sidx], 0);
3345
3346                         if (PF_ANEQ(daddr, &nk->addr[pd->didx], AF_INET6))
3347                                 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum,
3348                                     &nk->addr[pd->didx], 0);
3349                         rewrite++;
3350                         break;
3351 #endif /* INET */
3352                 default:
3353                         switch (af) {
3354 #ifdef INET
3355                         case AF_INET:
3356                                 if (PF_ANEQ(saddr,
3357                                     &nk->addr[pd->sidx], AF_INET))
3358                                         pf_change_a(&saddr->v4.s_addr,
3359                                             pd->ip_sum,
3360                                             nk->addr[pd->sidx].v4.s_addr, 0);
3361
3362                                 if (PF_ANEQ(daddr,
3363                                     &nk->addr[pd->didx], AF_INET))
3364                                         pf_change_a(&daddr->v4.s_addr,
3365                                             pd->ip_sum,
3366                                             nk->addr[pd->didx].v4.s_addr, 0);
3367                                 break;
3368 #endif /* INET */
3369 #ifdef INET6
3370                         case AF_INET6:
3371                                 if (PF_ANEQ(saddr,
3372                                     &nk->addr[pd->sidx], AF_INET6))
3373                                         PF_ACPY(saddr, &nk->addr[pd->sidx], af);
3374
3375                                 if (PF_ANEQ(daddr,
3376                                     &nk->addr[pd->didx], AF_INET6))
3377                                         PF_ACPY(saddr, &nk->addr[pd->didx], af);
3378                                 break;
3379 #endif /* INET */
3380                         }
3381                         break;
3382                 }
3383                 if (nr->natpass)
3384                         r = NULL;
3385                 pd->nat_rule = nr;
3386         }
3387
3388         while (r != NULL) {
3389                 r->evaluations++;
3390                 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3391                         r = r->skip[PF_SKIP_IFP].ptr;
3392                 else if (r->direction && r->direction != direction)
3393                         r = r->skip[PF_SKIP_DIR].ptr;
3394                 else if (r->af && r->af != af)
3395                         r = r->skip[PF_SKIP_AF].ptr;
3396                 else if (r->proto && r->proto != pd->proto)
3397                         r = r->skip[PF_SKIP_PROTO].ptr;
3398                 else if (PF_MISMATCHAW(&r->src.addr, saddr, af,
3399                     r->src.neg, kif))
3400                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3401                 /* tcp/udp only. port_op always 0 in other cases */
3402                 else if (r->src.port_op && !pf_match_port(r->src.port_op,
3403                     r->src.port[0], r->src.port[1], sport))
3404                         r = r->skip[PF_SKIP_SRC_PORT].ptr;
3405                 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af,
3406                     r->dst.neg, NULL))
3407                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
3408                 /* tcp/udp only. port_op always 0 in other cases */
3409                 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
3410                     r->dst.port[0], r->dst.port[1], dport))
3411                         r = r->skip[PF_SKIP_DST_PORT].ptr;
3412                 /* icmp only. type always 0 in other cases */
3413                 else if (r->type && r->type != icmptype + 1)
3414                         r = TAILQ_NEXT(r, entries);
3415                 /* icmp only. type always 0 in other cases */
3416                 else if (r->code && r->code != icmpcode + 1)
3417                         r = TAILQ_NEXT(r, entries);
3418                 else if (r->tos && !(r->tos == pd->tos))
3419                         r = TAILQ_NEXT(r, entries);
3420                 else if (r->rule_flag & PFRULE_FRAGMENT)
3421                         r = TAILQ_NEXT(r, entries);
3422                 else if (pd->proto == IPPROTO_TCP &&
3423                     (r->flagset & th->th_flags) != r->flags)
3424                         r = TAILQ_NEXT(r, entries);
3425                 /* tcp/udp only. uid.op always 0 in other cases */
3426                 else if (r->uid.op && (pd->lookup.done || (pd->lookup.done =
3427                     pf_socket_lookup(direction, pd), 1)) &&
3428                     !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
3429                     pd->lookup.uid))
3430                         r = TAILQ_NEXT(r, entries);
3431                 /* tcp/udp only. gid.op always 0 in other cases */
3432                 else if (r->gid.op && (pd->lookup.done || (pd->lookup.done =
3433                     pf_socket_lookup(direction, pd), 1)) &&
3434                     !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
3435                     pd->lookup.gid))
3436                         r = TAILQ_NEXT(r, entries);
3437                 else if (r->prob &&
3438                   r->prob <= karc4random())
3439                         r = TAILQ_NEXT(r, entries);
3440                 else if (r->match_tag && !pf_match_tag(m, r, &tag))
3441                         r = TAILQ_NEXT(r, entries);
3442                 else if (r->os_fingerprint != PF_OSFP_ANY &&
3443                     (pd->proto != IPPROTO_TCP || !pf_osfp_match(
3444                     pf_osfp_fingerprint(pd, m, off, th),
3445                     r->os_fingerprint)))
3446                         r = TAILQ_NEXT(r, entries);
3447                 else {
3448                         if (r->tag)
3449                                 tag = r->tag;
3450                         if (r->rtableid >= 0)
3451                                 rtableid = r->rtableid;
3452                         if (r->anchor == NULL) {
3453                                 match = 1;
3454                                 *rm = r;
3455                                 *am = a;
3456                                 *rsm = ruleset;
3457                                 if ((*rm)->quick)
3458                                         break;
3459                                 r = TAILQ_NEXT(r, entries);
3460                         } else
3461                                 pf_step_into_anchor(&asd, &ruleset,
3462                                     PF_RULESET_FILTER, &r, &a, &match);
3463                 }
3464                 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3465                     PF_RULESET_FILTER, &r, &a, &match))
3466                         break;
3467         }
3468         r = *rm;
3469         a = *am;
3470         ruleset = *rsm;
3471
3472         REASON_SET(&reason, PFRES_MATCH);
3473
3474         if (r->log || (nr != NULL && nr->log)) {
3475                 if (rewrite)
3476                         m_copyback(m, off, hdrlen, pd->hdr.any);
3477                 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr,
3478                     a, ruleset, pd);
3479         }
3480
3481         if ((r->action == PF_DROP) &&
3482             ((r->rule_flag & PFRULE_RETURNRST) ||
3483             (r->rule_flag & PFRULE_RETURNICMP) ||
3484             (r->rule_flag & PFRULE_RETURN))) {
3485                 /* undo NAT changes, if they have taken place */
3486                 if (nr != NULL) {
3487                         PF_ACPY(saddr, &sk->addr[pd->sidx], af);
3488                         PF_ACPY(daddr, &sk->addr[pd->didx], af);
3489                         if (pd->sport)
3490                                 *pd->sport = sk->port[pd->sidx];
3491                         if (pd->dport)
3492                                 *pd->dport = sk->port[pd->didx];
3493                         if (pd->proto_sum)
3494                                 *pd->proto_sum = bproto_sum;
3495                         if (pd->ip_sum)
3496                                 *pd->ip_sum = bip_sum;
3497                         m_copyback(m, off, hdrlen, pd->hdr.any);
3498                 }
3499                 if (pd->proto == IPPROTO_TCP &&
3500                     ((r->rule_flag & PFRULE_RETURNRST) ||
3501                     (r->rule_flag & PFRULE_RETURN)) &&
3502                     !(th->th_flags & TH_RST)) {
3503                         u_int32_t        ack = ntohl(th->th_seq) + pd->p_len;
3504                         int              len = 0;
3505                         struct ip       *h4;
3506 #ifdef INET6
3507                         struct ip6_hdr  *h6;
3508 #endif
3509                         switch (af) {
3510                         case AF_INET:
3511                                 h4 = mtod(m, struct ip *);
3512                                 len = h4->ip_len - off;
3513                                 break;
3514 #ifdef INET6
3515                         case AF_INET6:
3516                                 h6 = mtod(m, struct ip6_hdr *);
3517                                 len = h6->ip6_plen - (off - sizeof(*h6));
3518                                 break;
3519 #endif
3520                         }
3521
3522                         if (pf_check_proto_cksum(m, off, len, IPPROTO_TCP, af))
3523                                 REASON_SET(&reason, PFRES_PROTCKSUM);
3524                         else {
3525                                 if (th->th_flags & TH_SYN)
3526                                         ack++;
3527                                 if (th->th_flags & TH_FIN)
3528                                         ack++;
3529                                 pf_send_tcp(r, af, pd->dst,
3530                                     pd->src, th->th_dport, th->th_sport,
3531                                     ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0,
3532                                     r->return_ttl, 1, 0, pd->eh, kif->pfik_ifp);
3533                         }
3534                 } else if (pd->proto != IPPROTO_ICMP && af == AF_INET &&
3535                     r->return_icmp)
3536                         pf_send_icmp(m, r->return_icmp >> 8,
3537                             r->return_icmp & 255, af, r);
3538                 else if (pd->proto != IPPROTO_ICMPV6 && af == AF_INET6 &&
3539                     r->return_icmp6)
3540                         pf_send_icmp(m, r->return_icmp6 >> 8,
3541                             r->return_icmp6 & 255, af, r);
3542         }
3543
3544         if (r->action == PF_DROP)
3545                 goto cleanup;
3546
3547         if (pf_tag_packet(m, tag, rtableid)) {
3548                 REASON_SET(&reason, PFRES_MEMORY);
3549                 goto cleanup;
3550         }
3551
3552         if (!state_icmp && (r->keep_state || nr != NULL ||
3553             (pd->flags & PFDESC_TCP_NORM))) {
3554                 int action;
3555                 action = pf_create_state(r, nr, a, pd, nsn, skw, sks, nk, sk, m,
3556                     off, sport, dport, &rewrite, kif, sm, tag, bproto_sum,
3557                     bip_sum, hdrlen);
3558                 if (action != PF_PASS)
3559                         return (action);
3560         }
3561
3562         /* copy back packet headers if we performed NAT operations */
3563         if (rewrite)
3564                 m_copyback(m, off, hdrlen, pd->hdr.any);
3565
3566         return (PF_PASS);
3567
3568 cleanup:
3569         if (sk != NULL)
3570                 kfree(sk, M_PFSTATEKEYPL);
3571         if (nk != NULL)
3572                 kfree(nk, M_PFSTATEKEYPL);
3573         return (PF_DROP);
3574 }
3575
3576 static __inline int
3577 pf_create_state(struct pf_rule *r, struct pf_rule *nr, struct pf_rule *a,
3578     struct pf_pdesc *pd, struct pf_src_node *nsn, struct pf_state_key *skw,
3579     struct pf_state_key *sks, struct pf_state_key *nk, struct pf_state_key *sk,
3580     struct mbuf *m, int off, u_int16_t sport, u_int16_t dport, int *rewrite,
3581     struct pfi_kif *kif, struct pf_state **sm, int tag, u_int16_t bproto_sum,
3582     u_int16_t bip_sum, int hdrlen)
3583 {
3584         struct pf_state         *s = NULL;
3585         struct pf_src_node      *sn = NULL;
3586         struct tcphdr           *th = pd->hdr.tcp;
3587         u_int16_t                mss = tcp_mssdflt;
3588         u_short                  reason;
3589
3590         /* check maximums */
3591         if (r->max_states && (r->states_cur >= r->max_states)) {
3592                 pf_status.lcounters[LCNT_STATES]++;
3593                 REASON_SET(&reason, PFRES_MAXSTATES);
3594                 return (PF_DROP);
3595         }
3596         /* src node for filter rule */
3597         if ((r->rule_flag & PFRULE_SRCTRACK ||
3598             r->rpool.opts & PF_POOL_STICKYADDR) &&
3599             pf_insert_src_node(&sn, r, pd->src, pd->af) != 0) {
3600                 REASON_SET(&reason, PFRES_SRCLIMIT);
3601                 goto csfailed;
3602         }
3603         /* src node for translation rule */
3604         if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3605             pf_insert_src_node(&nsn, nr, &sk->addr[pd->sidx], pd->af)) {
3606                 REASON_SET(&reason, PFRES_SRCLIMIT);
3607                 goto csfailed;
3608         }
3609         s = kmalloc(sizeof(struct pf_state), M_PFSTATEPL, M_NOWAIT|M_ZERO);
3610         if (s == NULL) {
3611                 REASON_SET(&reason, PFRES_MEMORY);
3612                 goto csfailed;
3613         }
3614         s->id = 0; /* XXX Do we really need that? not in OpenBSD */
3615         s->creatorid = 0;
3616         s->rule.ptr = r;
3617         s->nat_rule.ptr = nr;
3618         s->anchor.ptr = a;
3619         STATE_INC_COUNTERS(s);
3620         if (r->allow_opts)
3621                 s->state_flags |= PFSTATE_ALLOWOPTS;
3622         if (r->rule_flag & PFRULE_STATESLOPPY)
3623                 s->state_flags |= PFSTATE_SLOPPY;
3624         s->log = r->log & PF_LOG_ALL;
3625         if (nr != NULL)
3626                 s->log |= nr->log & PF_LOG_ALL;
3627         switch (pd->proto) {
3628         case IPPROTO_TCP:
3629                 s->src.seqlo = ntohl(th->th_seq);
3630                 s->src.seqhi = s->src.seqlo + pd->p_len + 1;
3631                 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
3632                     r->keep_state == PF_STATE_MODULATE) {
3633                         /* Generate sequence number modulator */
3634                         if ((s->src.seqdiff = pf_tcp_iss(pd) - s->src.seqlo) ==
3635                             0)
3636                                 s->src.seqdiff = 1;
3637                         pf_change_a(&th->th_seq, &th->th_sum,
3638                             htonl(s->src.seqlo + s->src.seqdiff), 0);
3639                         *rewrite = 1;
3640                 } else
3641                         s->src.seqdiff = 0;
3642                 if (th->th_flags & TH_SYN) {
3643                         s->src.seqhi++;
3644                         s->src.wscale = pf_get_wscale(m, off,
3645                             th->th_off, pd->af);
3646                 }
3647                 s->src.max_win = MAX(ntohs(th->th_win), 1);
3648                 if (s->src.wscale & PF_WSCALE_MASK) {
3649                         /* Remove scale factor from initial window */
3650                         int win = s->src.max_win;
3651                         win += 1 << (s->src.wscale & PF_WSCALE_MASK);
3652                         s->src.max_win = (win - 1) >>
3653                             (s->src.wscale & PF_WSCALE_MASK);
3654                 }
3655                 if (th->th_flags & TH_FIN)
3656                         s->src.seqhi++;
3657                 s->dst.seqhi = 1;
3658                 s->dst.max_win = 1;
3659                 s->src.state = TCPS_SYN_SENT;
3660                 s->dst.state = TCPS_CLOSED;
3661                 s->timeout = PFTM_TCP_FIRST_PACKET;
3662                 break;
3663         case IPPROTO_UDP:
3664                 s->src.state = PFUDPS_SINGLE;
3665                 s->dst.state = PFUDPS_NO_TRAFFIC;
3666                 s->timeout = PFTM_UDP_FIRST_PACKET;
3667                 break;
3668         case IPPROTO_ICMP:
3669 #ifdef INET6
3670         case IPPROTO_ICMPV6:
3671 #endif
3672                 s->timeout = PFTM_ICMP_FIRST_PACKET;
3673                 break;
3674         default:
3675                 s->src.state = PFOTHERS_SINGLE;
3676                 s->dst.state = PFOTHERS_NO_TRAFFIC;
3677                 s->timeout = PFTM_OTHER_FIRST_PACKET;
3678         }
3679
3680         s->creation = time_second;
3681         s->expire = time_second;
3682
3683         if (sn != NULL) {
3684                 s->src_node = sn;
3685                 s->src_node->states++;
3686         }
3687         if (nsn != NULL) {
3688                 /* XXX We only modify one side for now. */
3689                 PF_ACPY(&nsn->raddr, &nk->addr[1], pd->af);
3690                 s->nat_src_node = nsn;
3691                 s->nat_src_node->states++;
3692         }
3693         if (pd->proto == IPPROTO_TCP) {
3694                 if ((pd->flags & PFDESC_TCP_NORM) && pf_normalize_tcp_init(m,
3695                     off, pd, th, &s->src, &s->dst)) {
3696                         REASON_SET(&reason, PFRES_MEMORY);
3697                         pf_src_tree_remove_state(s);
3698                         STATE_DEC_COUNTERS(s);
3699                         kfree(s, M_PFSTATEPL);
3700                         return (PF_DROP);
3701                 }
3702                 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
3703                     pf_normalize_tcp_stateful(m, off, pd, &reason, th, s,
3704                     &s->src, &s->dst, rewrite)) {
3705                         /* This really shouldn't happen!!! */
3706                         DPFPRINTF(PF_DEBUG_URGENT,
3707                             ("pf_normalize_tcp_stateful failed on first pkt"));
3708                         pf_normalize_tcp_cleanup(s);
3709                         pf_src_tree_remove_state(s);
3710                         STATE_DEC_COUNTERS(s);
3711                         kfree(s, M_PFSTATEPL);
3712                         return (PF_DROP);
3713                 }
3714         }
3715         s->direction = pd->dir;
3716
3717         if (sk == NULL && pf_state_key_setup(pd, nr, &skw, &sks, &sk, &nk,
3718             pd->src, pd->dst, sport, dport))
3719                 goto csfailed;
3720
3721         if (pf_state_insert(BOUND_IFACE(r, kif), skw, sks, s)) {
3722                 if (pd->proto == IPPROTO_TCP)
3723                         pf_normalize_tcp_cleanup(s);
3724                 REASON_SET(&reason, PFRES_STATEINS);
3725                 pf_src_tree_remove_state(s);
3726                 STATE_DEC_COUNTERS(s);
3727                 kfree(s, M_PFSTATEPL);
3728                 return (PF_DROP);
3729         } else
3730                 *sm = s;
3731
3732         pf_set_rt_ifp(s, pd->src);      /* needs s->state_key set */
3733         if (tag > 0) {
3734                 pf_tag_ref(tag);
3735                 s->tag = tag;
3736         }
3737         if (pd->proto == IPPROTO_TCP && (th->th_flags & (TH_SYN|TH_ACK)) ==
3738             TH_SYN && r->keep_state == PF_STATE_SYNPROXY) {
3739                 s->src.state = PF_TCPS_PROXY_SRC;
3740                 /* undo NAT changes, if they have taken place */
3741                 if (nr != NULL) {
3742                         struct pf_state_key *skt = s->key[PF_SK_WIRE];
3743                         if (pd->dir == PF_OUT)
3744                                 skt = s->key[PF_SK_STACK];
3745                         PF_ACPY(pd->src, &skt->addr[pd->sidx], pd->af);
3746                         PF_ACPY(pd->dst, &skt->addr[pd->didx], pd->af);
3747                         if (pd->sport)
3748                                 *pd->sport = skt->port[pd->sidx];
3749                         if (pd->dport)
3750                                 *pd->dport = skt->port[pd->didx];
3751                         if (pd->proto_sum)
3752                                 *pd->proto_sum = bproto_sum;
3753                         if (pd->ip_sum)
3754                                 *pd->ip_sum = bip_sum;
3755                         m_copyback(m, off, hdrlen, pd->hdr.any);
3756                 }
3757                 s->src.seqhi = htonl(karc4random());
3758                 /* Find mss option */
3759                 mss = pf_get_mss(m, off, th->th_off, pd->af);
3760                 mss = pf_calc_mss(pd->src, pd->af, mss);
3761                 mss = pf_calc_mss(pd->dst, pd->af, mss);
3762                 s->src.mss = mss;
3763                 pf_send_tcp(r, pd->af, pd->dst, pd->src, th->th_dport,
3764                     th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1,
3765                     TH_SYN|TH_ACK, 0, s->src.mss, 0, 1, 0, NULL, NULL);
3766                 REASON_SET(&reason, PFRES_SYNPROXY);
3767                 return (PF_SYNPROXY_DROP);
3768         }
3769
3770         return (PF_PASS);
3771
3772 csfailed:
3773         if (sk != NULL)
3774                 kfree(sk, M_PFSTATEKEYPL);
3775         if (nk != NULL)
3776                 kfree(nk, M_PFSTATEKEYPL);
3777
3778         if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3779                 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3780                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3781                 pf_status.src_nodes--;
3782                 kfree(sn, M_PFSRCTREEPL);
3783         }
3784         if (nsn != sn && nsn != NULL && nsn->states == 0 && nsn->expire == 0) {
3785                 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3786                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3787                 pf_status.src_nodes--;
3788                 kfree(nsn, M_PFSRCTREEPL);
3789         }
3790         return (PF_DROP);
3791 }
3792
3793 int
3794 pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif,
3795     struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am,
3796     struct pf_ruleset **rsm)
3797 {
3798         struct pf_rule          *r, *a = NULL;
3799         struct pf_ruleset       *ruleset = NULL;
3800         sa_family_t              af = pd->af;
3801         u_short                  reason;
3802         int                      tag = -1;
3803         int                      asd = 0;
3804         int                      match = 0;
3805
3806         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3807         while (r != NULL) {
3808                 r->evaluations++;
3809                 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3810                         r = r->skip[PF_SKIP_IFP].ptr;
3811                 else if (r->direction && r->direction != direction)
3812                         r = r->skip[PF_SKIP_DIR].ptr;
3813                 else if (r->af && r->af != af)
3814                         r = r->skip[PF_SKIP_AF].ptr;
3815                 else if (r->proto && r->proto != pd->proto)
3816                         r = r->skip[PF_SKIP_PROTO].ptr;
3817                 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af,
3818                     r->src.neg, kif))
3819                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3820                 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af,
3821                     r->dst.neg, NULL))
3822                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
3823                 else if (r->tos && !(r->tos == pd->tos))
3824                         r = TAILQ_NEXT(r, entries);
3825                 else if (r->os_fingerprint != PF_OSFP_ANY)
3826                         r = TAILQ_NEXT(r, entries);
3827                 else if (pd->proto == IPPROTO_UDP &&
3828                     (r->src.port_op || r->dst.port_op))
3829                         r = TAILQ_NEXT(r, entries);
3830                 else if (pd->proto == IPPROTO_TCP &&
3831                     (r->src.port_op || r->dst.port_op || r->flagset))
3832                         r = TAILQ_NEXT(r, entries);
3833                 else if ((pd->proto == IPPROTO_ICMP ||
3834                     pd->proto == IPPROTO_ICMPV6) &&
3835                     (r->type || r->code))
3836                         r = TAILQ_NEXT(r, entries);
3837                 else if (r->prob && r->prob <= karc4random())
3838                         r = TAILQ_NEXT(r, entries);
3839                 else if (r->match_tag && !pf_match_tag(m, r, &tag))
3840                         r = TAILQ_NEXT(r, entries);
3841                 else {
3842                         if (r->anchor == NULL) {
3843                                 match = 1;
3844                                 *rm = r;
3845                                 *am = a;
3846                                 *rsm = ruleset;
3847                                 if ((*rm)->quick)
3848                                         break;
3849                                 r = TAILQ_NEXT(r, entries);
3850                         } else
3851                                 pf_step_into_anchor(&asd, &ruleset,
3852                                     PF_RULESET_FILTER, &r, &a, &match);
3853                 }
3854                 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3855                     PF_RULESET_FILTER, &r, &a, &match))
3856                         break;
3857         }
3858         r = *rm;
3859         a = *am;
3860         ruleset = *rsm;
3861
3862         REASON_SET(&reason, PFRES_MATCH);
3863
3864         if (r->log)
3865                 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset,
3866                     pd);
3867
3868         if (r->action != PF_PASS)
3869                 return (PF_DROP);
3870
3871         if (pf_tag_packet(m, tag, -1)) {
3872                 REASON_SET(&reason, PFRES_MEMORY);
3873                 return (PF_DROP);
3874         }
3875
3876         return (PF_PASS);
3877 }
3878
3879 int
3880 pf_tcp_track_full(struct pf_state_peer *src, struct pf_state_peer *dst,
3881         struct pf_state **state, struct pfi_kif *kif, struct mbuf *m, int off,
3882         struct pf_pdesc *pd, u_short *reason, int *copyback)
3883 {
3884         struct tcphdr           *th = pd->hdr.tcp;
3885         u_int16_t                win = ntohs(th->th_win);
3886         u_int32_t                ack, end, seq, orig_seq;
3887         u_int8_t                 sws, dws;
3888         int                      ackskew;
3889
3890         if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) {
3891                 sws = src->wscale & PF_WSCALE_MASK;
3892                 dws = dst->wscale & PF_WSCALE_MASK;
3893         } else
3894                 sws = dws = 0;
3895
3896         /*
3897          * Sequence tracking algorithm from Guido van Rooij's paper:
3898          *   http://www.madison-gurkha.com/publications/tcp_filtering/
3899          *      tcp_filtering.ps
3900          */
3901
3902         orig_seq = seq = ntohl(th->th_seq);
3903         if (src->seqlo == 0) {
3904                 /* First packet from this end. Set its state */
3905
3906                 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) &&
3907                     src->scrub == NULL) {
3908                         if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) {
3909                                 REASON_SET(reason, PFRES_MEMORY);
3910                                 return (PF_DROP);
3911                         }
3912                 }
3913
3914                 /* Deferred generation of sequence number modulator */
3915                 if (dst->seqdiff && !src->seqdiff) {
3916                         /* use random iss for the TCP server */
3917                         while ((src->seqdiff = karc4random() - seq) == 0)
3918                                 ;
3919                         ack = ntohl(th->th_ack) - dst->seqdiff;
3920                         pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3921                             src->seqdiff), 0);
3922                         pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3923                         *copyback = 1;
3924                 } else {
3925                         ack = ntohl(th->th_ack);
3926                 }
3927
3928                 end = seq + pd->p_len;
3929                 if (th->th_flags & TH_SYN) {
3930                         end++;
3931                         (*state)->sync_flags |= PFSTATE_GOT_SYN2;
3932                         if (dst->wscale & PF_WSCALE_FLAG) {
3933                                 src->wscale = pf_get_wscale(m, off, th->th_off,
3934                                     pd->af);
3935                                 if (src->wscale & PF_WSCALE_FLAG) {
3936                                         /* Remove scale factor from initial
3937                                          * window */
3938                                         sws = src->wscale & PF_WSCALE_MASK;
3939                                         win = ((u_int32_t)win + (1 << sws) - 1)
3940                                             >> sws;
3941                                         dws = dst->wscale & PF_WSCALE_MASK;
3942                                 } else {
3943                                         /* fixup other window */
3944                                         dst->max_win <<= dst->wscale &
3945                                             PF_WSCALE_MASK;
3946                                         /* in case of a retrans SYN|ACK */
3947                                         dst->wscale = 0;
3948                                 }
3949                         }
3950                 }
3951                 if (th->th_flags & TH_FIN)
3952                         end++;
3953
3954                 src->seqlo = seq;
3955                 if (src->state < TCPS_SYN_SENT)
3956                         src->state = TCPS_SYN_SENT;
3957
3958                 /*
3959                  * May need to slide the window (seqhi may have been set by
3960                  * the crappy stack check or if we picked up the connection
3961                  * after establishment)
3962                  */
3963                 if (src->seqhi == 1 ||
3964                     SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi))
3965                         src->seqhi = end + MAX(1, dst->max_win << dws);
3966                 if (win > src->max_win)
3967                         src->max_win = win;
3968
3969         } else {
3970                 ack = ntohl(th->th_ack) - dst->seqdiff;
3971                 if (src->seqdiff) {
3972                         /* Modulate sequence numbers */
3973                         pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3974                             src->seqdiff), 0);
3975                         pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3976                         *copyback = 1;
3977                 }
3978                 end = seq + pd->p_len;
3979                 if (th->th_flags & TH_SYN)
3980                         end++;
3981                 if (th->th_flags & TH_FIN)
3982                         end++;
3983         }
3984
3985         if ((th->th_flags & TH_ACK) == 0) {
3986                 /* Let it pass through the ack skew check */
3987                 ack = dst->seqlo;
3988         } else if ((ack == 0 &&
3989             (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) ||
3990             /* broken tcp stacks do not set ack */
3991             (dst->state < TCPS_SYN_SENT)) {
3992                 /*
3993                  * Many stacks (ours included) will set the ACK number in an
3994                  * FIN|ACK if the SYN times out -- no sequence to ACK.
3995                  */
3996                 ack = dst->seqlo;
3997         }
3998
3999         if (seq == end) {
4000                 /* Ease sequencing restrictions on no data packets */
4001                 seq = src->seqlo;
4002                 end = seq;
4003         }
4004
4005         ackskew = dst->seqlo - ack;
4006
4007
4008         /*
4009          * Need to demodulate the sequence numbers in any TCP SACK options
4010          * (Selective ACK). We could optionally validate the SACK values
4011          * against the current ACK window, either forwards or backwards, but
4012          * I'm not confident that SACK has been implemented properly
4013          * everywhere. It wouldn't surprise me if several stacks accidently
4014          * SACK too far backwards of previously ACKed data. There really aren't
4015          * any security implications of bad SACKing unless the target stack
4016          * doesn't validate the option length correctly. Someone trying to
4017          * spoof into a TCP connection won't bother blindly sending SACK
4018          * options anyway.
4019          */
4020         if (dst->seqdiff && (th->th_off << 2) > sizeof(struct tcphdr)) {
4021                 if (pf_modulate_sack(m, off, pd, th, dst))
4022                         *copyback = 1;
4023         }
4024
4025
4026 #define MAXACKWINDOW (0xffff + 1500)    /* 1500 is an arbitrary fudge factor */
4027         if (SEQ_GEQ(src->seqhi, end) &&
4028             /* Last octet inside other's window space */
4029             SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) &&
4030             /* Retrans: not more than one window back */
4031             (ackskew >= -MAXACKWINDOW) &&
4032             /* Acking not more than one reassembled fragment backwards */
4033             (ackskew <= (MAXACKWINDOW << sws)) &&
4034             /* Acking not more than one window forward */
4035             ((th->th_flags & TH_RST) == 0 || orig_seq == src->seqlo ||
4036             (orig_seq == src->seqlo + 1) || (orig_seq + 1 == src->seqlo) ||
4037             (pd->flags & PFDESC_IP_REAS) == 0)) {
4038             /* Require an exact/+1 sequence match on resets when possible */
4039
4040                 if (dst->scrub || src->scrub) {
4041                         if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4042                             *state, src, dst, copyback))
4043                                 return (PF_DROP);
4044                 }
4045
4046                 /* update max window */
4047                 if (src->max_win < win)
4048                         src->max_win = win;
4049                 /* synchronize sequencing */
4050                 if (SEQ_GT(end, src->seqlo))
4051                         src->seqlo = end;
4052                 /* slide the window of what the other end can send */
4053                 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4054                         dst->seqhi = ack + MAX((win << sws), 1);
4055
4056
4057                 /* update states */
4058                 if (th->th_flags & TH_SYN)
4059                         if (src->state < TCPS_SYN_SENT)
4060                                 src->state = TCPS_SYN_SENT;
4061                 if (th->th_flags & TH_FIN)
4062                         if (src->state < TCPS_CLOSING)
4063                                 src->state = TCPS_CLOSING;
4064                 if (th->th_flags & TH_ACK) {
4065                         if (dst->state == TCPS_SYN_SENT) {
4066                                 dst->state = TCPS_ESTABLISHED;
4067                                 if (src->state == TCPS_ESTABLISHED &&
4068                                     (*state)->src_node != NULL &&
4069                                     pf_src_connlimit(state)) {
4070                                         REASON_SET(reason, PFRES_SRCLIMIT);
4071                                         return (PF_DROP);
4072                                 }
4073                         } else if (dst->state == TCPS_CLOSING)
4074                                 dst->state = TCPS_FIN_WAIT_2;
4075                 }
4076                 if (th->th_flags & TH_RST)
4077                         src->state = dst->state = TCPS_TIME_WAIT;
4078
4079                 /* update expire time */
4080                 (*state)->expire = time_second;
4081                 if (src->state >= TCPS_FIN_WAIT_2 &&
4082                     dst->state >= TCPS_FIN_WAIT_2)
4083                         (*state)->timeout = PFTM_TCP_CLOSED;
4084                 else if (src->state >= TCPS_CLOSING &&
4085                     dst->state >= TCPS_CLOSING)
4086                         (*state)->timeout = PFTM_TCP_FIN_WAIT;
4087                 else if (src->state < TCPS_ESTABLISHED ||
4088                     dst->state < TCPS_ESTABLISHED)
4089                         (*state)->timeout = PFTM_TCP_OPENING;
4090                 else if (src->state >= TCPS_CLOSING ||
4091                     dst->state >= TCPS_CLOSING)
4092                         (*state)->timeout = PFTM_TCP_CLOSING;
4093                 else
4094                         (*state)->timeout = PFTM_TCP_ESTABLISHED;
4095
4096                 /* Fall through to PASS packet */
4097
4098         } else if ((dst->state < TCPS_SYN_SENT ||
4099                 dst->state >= TCPS_FIN_WAIT_2 ||
4100                 src->state >= TCPS_FIN_WAIT_2) &&
4101             SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) &&
4102             /* Within a window forward of the originating packet */
4103             SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) {
4104             /* Within a window backward of the originating packet */
4105
4106                 /*
4107                  * This currently handles three situations:
4108                  *  1) Stupid stacks will shotgun SYNs before their peer
4109                  *     replies.
4110                  *  2) When PF catches an already established stream (the
4111                  *     firewall rebooted, the state table was flushed, routes
4112                  *     changed...)
4113                  *  3) Packets get funky immediately after the connection
4114                  *     closes (this should catch Solaris spurious ACK|FINs
4115                  *     that web servers like to spew after a close)
4116                  *
4117                  * This must be a little more careful than the above code
4118                  * since packet floods will also be caught here. We don't
4119                  * update the TTL here to mitigate the damage of a packet
4120                  * flood and so the same code can handle awkward establishment
4121                  * and a loosened connection close.
4122                  * In the establishment case, a correct peer response will
4123                  * validate the connection, go through the normal state code
4124                  * and keep updating the state TTL.
4125                  */
4126
4127                 if (pf_status.debug >= PF_DEBUG_MISC) {
4128                         kprintf("pf: loose state match: ");
4129                         pf_print_state(*state);
4130                         pf_print_flags(th->th_flags);
4131                         kprintf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
4132                             "pkts=%llu:%llu dir=%s,%s\n", seq, orig_seq, ack, pd->p_len,
4133                             ackskew, (unsigned long long)(*state)->packets[0],
4134                             (unsigned long long)(*state)->packets[1],
4135                             pd->dir == PF_IN ? "in" : "out",
4136                             pd->dir == (*state)->direction ? "fwd" : "rev");
4137                 }
4138
4139                 if (dst->scrub || src->scrub) {
4140                         if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4141                             *state, src, dst, copyback))
4142                                 return (PF_DROP);
4143                 }
4144
4145                 /* update max window */
4146                 if (src->max_win < win)
4147                         src->max_win = win;
4148                 /* synchronize sequencing */
4149                 if (SEQ_GT(end, src->seqlo))
4150                         src->seqlo = end;
4151                 /* slide the window of what the other end can send */
4152                 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4153                         dst->seqhi = ack + MAX((win << sws), 1);
4154
4155                 /*
4156                  * Cannot set dst->seqhi here since this could be a shotgunned
4157                  * SYN and not an already established connection.
4158                  */
4159
4160                 if (th->th_flags & TH_FIN)
4161                         if (src->state < TCPS_CLOSING)
4162                                 src->state = TCPS_CLOSING;
4163                 if (th->th_flags & TH_RST)
4164                         src->state = dst->state = TCPS_TIME_WAIT;
4165
4166                 /* Fall through to PASS packet */
4167
4168         } else if ((*state)->pickup_mode == PF_PICKUPS_HASHONLY ||
4169                     ((*state)->pickup_mode == PF_PICKUPS_ENABLED &&
4170                      ((*state)->sync_flags & PFSTATE_GOT_SYN_MASK) !=
4171                       PFSTATE_GOT_SYN_MASK)) {
4172                 /*
4173                  * If pickup mode is hash only, do not fail on sequence checks.
4174                  *
4175                  * If pickup mode is enabled and we did not see the SYN in
4176                  * both direction, do not fail on sequence checks because
4177                  * we do not have complete information on window scale.
4178                  *
4179                  * Adjust expiration and fall through to PASS packet.
4180                  * XXX Add a FIN check to reduce timeout?
4181                  */
4182                 (*state)->expire = time_second;
4183         } else  {
4184                 /*
4185                  * Failure processing
4186                  */
4187                 if ((*state)->dst.state == TCPS_SYN_SENT &&
4188                     (*state)->src.state == TCPS_SYN_SENT) {
4189                         /* Send RST for state mismatches during handshake */
4190                         if (!(th->th_flags & TH_RST))
4191                                 pf_send_tcp((*state)->rule.ptr, pd->af,
4192                                     pd->dst, pd->src, th->th_dport,
4193                                     th->th_sport, ntohl(th->th_ack), 0,
4194                                     TH_RST, 0, 0,
4195                                     (*state)->rule.ptr->return_ttl, 1, 0,
4196                                     pd->eh, kif->pfik_ifp);
4197                         src->seqlo = 0;
4198                         src->seqhi = 1;
4199                         src->max_win = 1;
4200                 } else if (pf_status.debug >= PF_DEBUG_MISC) {
4201                         kprintf("pf: BAD state: ");
4202                         pf_print_state(*state);
4203                         pf_print_flags(th->th_flags);
4204                         kprintf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
4205                             "pkts=%llu:%llu dir=%s,%s\n",
4206                             seq, orig_seq, ack, pd->p_len, ackskew,
4207                             (unsigned long long)(*state)->packets[0],
4208                                 (unsigned long long)(*state)->packets[1],
4209                             pd->dir == PF_IN ? "in" : "out",
4210                             pd->dir == (*state)->direction ? "fwd" : "rev");
4211                         kprintf("pf: State failure on: %c %c %c %c | %c %c\n",
4212                             SEQ_GEQ(src->seqhi, end) ? ' ' : '1',
4213                             SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ?
4214                             ' ': '2',
4215                             (ackskew >= -MAXACKWINDOW) ? ' ' : '3',
4216                             (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4',
4217                             SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5',
4218                             SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6');
4219                 }
4220                 REASON_SET(reason, PFRES_BADSTATE);
4221                 return (PF_DROP);
4222         }
4223
4224         return (PF_PASS);
4225 }
4226
4227 int
4228 pf_tcp_track_sloppy(struct pf_state_peer *src, struct pf_state_peer *dst,
4229         struct pf_state **state, struct pf_pdesc *pd, u_short *reason)
4230 {
4231         struct tcphdr           *th = pd->hdr.tcp;
4232
4233         if (th->th_flags & TH_SYN)
4234                 if (src->state < TCPS_SYN_SENT)
4235                         src->state = TCPS_SYN_SENT;
4236         if (th->th_flags & TH_FIN)
4237                 if (src->state < TCPS_CLOSING)
4238                         src->state = TCPS_CLOSING;
4239         if (th->th_flags & TH_ACK) {
4240                 if (dst->state == TCPS_SYN_SENT) {
4241                         dst->state = TCPS_ESTABLISHED;
4242                         if (src->state == TCPS_ESTABLISHED &&
4243                             (*state)->src_node != NULL &&
4244                             pf_src_connlimit(state)) {
4245                                 REASON_SET(reason, PFRES_SRCLIMIT);
4246                                 return (PF_DROP);
4247                         }
4248                 } else if (dst->state == TCPS_CLOSING) {
4249                         dst->state = TCPS_FIN_WAIT_2;
4250                 } else if (src->state == TCPS_SYN_SENT &&
4251                     dst->state < TCPS_SYN_SENT) {
4252                         /*
4253                          * Handle a special sloppy case where we only see one
4254                          * half of the connection. If there is a ACK after
4255                          * the initial SYN without ever seeing a packet from
4256                          * the destination, set the connection to established.
4257                          */
4258                         dst->state = src->state = TCPS_ESTABLISHED;
4259                         if ((*state)->src_node != NULL &&
4260                             pf_src_connlimit(state)) {
4261                                 REASON_SET(reason, PFRES_SRCLIMIT);
4262                                 return (PF_DROP);
4263                         }
4264                 } else if (src->state == TCPS_CLOSING &&
4265                     dst->state == TCPS_ESTABLISHED &&
4266                     dst->seqlo == 0) {
4267                         /*
4268                          * Handle the closing of half connections where we
4269                          * don't see the full bidirectional FIN/ACK+ACK
4270                          * handshake.
4271                          */
4272                         dst->state = TCPS_CLOSING;
4273                 }
4274         }
4275         if (th->th_flags & TH_RST)
4276                 src->state = dst->state = TCPS_TIME_WAIT;
4277
4278         /* update expire time */
4279         (*state)->expire = time_second;
4280         if (src->state >= TCPS_FIN_WAIT_2 &&
4281             dst->state >= TCPS_FIN_WAIT_2)
4282                 (*state)->timeout = PFTM_TCP_CLOSED;
4283         else if (src->state >= TCPS_CLOSING &&
4284             dst->state >= TCPS_CLOSING)
4285                 (*state)->timeout = PFTM_TCP_FIN_WAIT;
4286         else if (src->state < TCPS_ESTABLISHED ||
4287             dst->state < TCPS_ESTABLISHED)
4288                 (*state)->timeout = PFTM_TCP_OPENING;
4289         else if (src->state >= TCPS_CLOSING ||
4290             dst->state >= TCPS_CLOSING)
4291                 (*state)->timeout = PFTM_TCP_CLOSING;
4292         else
4293                 (*state)->timeout = PFTM_TCP_ESTABLISHED;
4294
4295         return (PF_PASS);
4296 }
4297
4298 int
4299 pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
4300     struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
4301     u_short *reason)
4302 {
4303         struct pf_state_key_cmp  key;
4304         struct tcphdr           *th = pd->hdr.tcp;
4305         int                      copyback = 0;
4306         struct pf_state_peer    *src, *dst;
4307         struct pf_state_key     *sk;
4308
4309         key.af = pd->af;
4310         key.proto = IPPROTO_TCP;
4311         if (direction == PF_IN) {       /* wire side, straight */
4312                 PF_ACPY(&key.addr[0], pd->src, key.af);
4313                 PF_ACPY(&key.addr[1], pd->dst, key.af);
4314                 key.port[0] = th->th_sport;
4315                 key.port[1] = th->th_dport;
4316         } else {                        /* stack side, reverse */
4317                 PF_ACPY(&key.addr[1], pd->src, key.af);
4318                 PF_ACPY(&key.addr[0], pd->dst, key.af);
4319                 key.port[1] = th->th_sport;
4320                 key.port[0] = th->th_dport;
4321         }
4322
4323         STATE_LOOKUP(kif, &key, direction, *state, m);
4324
4325         if (direction == (*state)->direction) {
4326                 src = &(*state)->src;
4327                 dst = &(*state)->dst;
4328         } else {
4329                 src = &(*state)->dst;
4330                 dst = &(*state)->src;
4331         }
4332
4333         sk = (*state)->key[pd->didx];
4334
4335         if ((*state)->src.state == PF_TCPS_PROXY_SRC) {
4336                 if (direction != (*state)->direction) {
4337                         REASON_SET(reason, PFRES_SYNPROXY);
4338                         return (PF_SYNPROXY_DROP);
4339                 }
4340                 if (th->th_flags & TH_SYN) {
4341                         if (ntohl(th->th_seq) != (*state)->src.seqlo) {
4342                                 REASON_SET(reason, PFRES_SYNPROXY);
4343                                 return (PF_DROP);
4344                         }
4345                         pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
4346                             pd->src, th->th_dport, th->th_sport,
4347                             (*state)->src.seqhi, ntohl(th->th_seq) + 1,
4348                             TH_SYN|TH_ACK, 0, (*state)->src.mss, 0, 1,
4349                             0, NULL, NULL);
4350                         REASON_SET(reason, PFRES_SYNPROXY);
4351                         return (PF_SYNPROXY_DROP);
4352                 } else if (!(th->th_flags & TH_ACK) ||
4353                     (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
4354                     (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
4355                         REASON_SET(reason, PFRES_SYNPROXY);
4356                         return (PF_DROP);
4357                 } else if ((*state)->src_node != NULL &&
4358                     pf_src_connlimit(state)) {
4359                         REASON_SET(reason, PFRES_SRCLIMIT);
4360                         return (PF_DROP);
4361                 } else
4362                         (*state)->src.state = PF_TCPS_PROXY_DST;
4363         }
4364         if ((*state)->src.state == PF_TCPS_PROXY_DST) {
4365                 if (direction == (*state)->direction) {
4366                         if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) ||
4367                             (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
4368                             (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
4369                                 REASON_SET(reason, PFRES_SYNPROXY);
4370                                 return (PF_DROP);
4371                         }
4372                         (*state)->src.max_win = MAX(ntohs(th->th_win), 1);
4373                         if ((*state)->dst.seqhi == 1)
4374                                 (*state)->dst.seqhi = htonl(karc4random());
4375                         pf_send_tcp((*state)->rule.ptr, pd->af,
4376                             &sk->addr[pd->sidx], &sk->addr[pd->didx],
4377                             sk->port[pd->sidx], sk->port[pd->didx],
4378                             (*state)->dst.seqhi, 0, TH_SYN, 0,
4379                             (*state)->src.mss, 0, 0, (*state)->tag, NULL, NULL);
4380                         REASON_SET(reason, PFRES_SYNPROXY);
4381                         return (PF_SYNPROXY_DROP);
4382                 } else if (((th->th_flags & (TH_SYN|TH_ACK)) !=
4383                     (TH_SYN|TH_ACK)) ||
4384                     (ntohl(th->th_ack) != (*state)->dst.seqhi + 1)) {
4385                         REASON_SET(reason, PFRES_SYNPROXY);
4386                         return (PF_DROP);
4387                 } else {
4388                         (*state)->dst.max_win = MAX(ntohs(th->th_win), 1);
4389                         (*state)->dst.seqlo = ntohl(th->th_seq);
4390                         pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
4391                             pd->src, th->th_dport, th->th_sport,
4392                             ntohl(th->th_ack), ntohl(th->th_seq) + 1,
4393                             TH_ACK, (*state)->src.max_win, 0, 0, 0,
4394                             (*state)->tag, NULL, NULL);
4395                         pf_send_tcp((*state)->rule.ptr, pd->af,
4396                             &sk->addr[pd->sidx], &sk->addr[pd->didx],
4397                             sk->port[pd->sidx], sk->port[pd->didx],
4398                             (*state)->src.seqhi + 1, (*state)->src.seqlo + 1,
4399                             TH_ACK, (*state)->dst.max_win, 0, 0, 1,
4400                             0, NULL, NULL);
4401                         (*state)->src.seqdiff = (*state)->dst.seqhi -
4402                             (*state)->src.seqlo;
4403                         (*state)->dst.seqdiff = (*state)->src.seqhi -
4404                             (*state)->dst.seqlo;
4405                         (*state)->src.seqhi = (*state)->src.seqlo +
4406                             (*state)->dst.max_win;
4407                         (*state)->dst.seqhi = (*state)->dst.seqlo +
4408                             (*state)->src.max_win;
4409                         (*state)->src.wscale = (*state)->dst.wscale = 0;
4410                         (*state)->src.state = (*state)->dst.state =
4411                             TCPS_ESTABLISHED;
4412                         REASON_SET(reason, PFRES_SYNPROXY);
4413                         return (PF_SYNPROXY_DROP);
4414                 }
4415         }
4416
4417         if (((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN) &&
4418             dst->state >= TCPS_FIN_WAIT_2 &&
4419             src->state >= TCPS_FIN_WAIT_2) {
4420                 if (pf_status.debug >= PF_DEBUG_MISC) {
4421                         kprintf("pf: state reuse ");
4422                         pf_print_state(*state);
4423                         pf_print_flags(th->th_flags);
4424                         kprintf("\n");
4425                 }
4426                 /* XXX make sure it's the same direction ?? */
4427                 (*state)->src.state = (*state)->dst.state = TCPS_CLOSED;
4428                 pf_unlink_state(*state);
4429                 *state = NULL;
4430                 return (PF_DROP);
4431         }
4432
4433         if ((*state)->state_flags & PFSTATE_SLOPPY) {
4434                 if (pf_tcp_track_sloppy(src, dst, state, pd, reason) == PF_DROP)
4435                         return (PF_DROP);
4436         } else {
4437                 if (pf_tcp_track_full(src, dst, state, kif, m, off, pd, reason,
4438                     &copyback) == PF_DROP)
4439                         return (PF_DROP);
4440         }
4441
4442         /* translate source/destination address, if necessary */
4443         if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) {
4444                 struct pf_state_key *nk = (*state)->key[pd->didx];
4445
4446                 if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], pd->af) ||
4447                     nk->port[pd->sidx] != th->th_sport)  {
4448                         /*
4449                          * The translated source address may be completely
4450                          * unrelated to the saved link header, make sure
4451                          * a bridge doesn't try to use it.
4452                          */
4453                         m->m_pkthdr.fw_flags &= ~BRIDGE_MBUF_TAGGED;
4454                         m->m_flags &= ~M_HASH;
4455                         pf_change_ap(pd->src, &th->th_sport, pd->ip_sum,
4456                             &th->th_sum, &nk->addr[pd->sidx],
4457                             nk->port[pd->sidx], 0, pd->af);
4458                 }
4459
4460                 if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], pd->af) ||
4461                     nk->port[pd->didx] != th->th_dport) {
4462                         /*
4463                          * If we don't redispatch the packet will go into
4464                          * the protocol stack on the wrong cpu for the
4465                          * post-translated address.
4466                          */
4467                         m->m_flags &= ~M_HASH;
4468                         pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum,
4469                             &th->th_sum, &nk->addr[pd->didx],
4470                             nk->port[pd->didx], 0, pd->af);
4471                 }
4472                 copyback = 1;
4473         }
4474
4475         /* Copyback sequence modulation or stateful scrub changes if needed */
4476         if (copyback)
4477                 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4478
4479         return (PF_PASS);
4480 }
4481
4482 int
4483 pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif,
4484     struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4485 {
4486         struct pf_state_peer    *src, *dst;
4487         struct pf_state_key_cmp  key;
4488         struct udphdr           *uh = pd->hdr.udp;
4489
4490         key.af = pd->af;
4491         key.proto = IPPROTO_UDP;
4492         if (direction == PF_IN) {       /* wire side, straight */
4493                 PF_ACPY(&key.addr[0], pd->src, key.af);
4494                 PF_ACPY(&key.addr[1], pd->dst, key.af);
4495                 key.port[0] = uh->uh_sport;
4496                 key.port[1] = uh->uh_dport;
4497         } else {                        /* stack side, reverse */
4498                 PF_ACPY(&key.addr[1], pd->src, key.af);
4499                 PF_ACPY(&key.addr[0], pd->dst, key.af);
4500                 key.port[1] = uh->uh_sport;
4501                 key.port[0] = uh->uh_dport;
4502         }
4503
4504         STATE_LOOKUP(kif, &key, direction, *state, m);
4505
4506         if (direction == (*state)->direction) {
4507                 src = &(*state)->src;
4508                 dst = &(*state)->dst;
4509         } else {
4510                 src = &(*state)->dst;
4511                 dst = &(*state)->src;
4512         }
4513
4514         /* update states */
4515         if (src->state < PFUDPS_SINGLE)
4516                 src->state = PFUDPS_SINGLE;
4517         if (dst->state == PFUDPS_SINGLE)
4518                 dst->state = PFUDPS_MULTIPLE;
4519
4520         /* update expire time */
4521         (*state)->expire = time_second;
4522         if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE)
4523                 (*state)->timeout = PFTM_UDP_MULTIPLE;
4524         else
4525                 (*state)->timeout = PFTM_UDP_SINGLE;
4526
4527         /* translate source/destination address, if necessary */
4528         if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) {
4529                 struct pf_state_key *nk = (*state)->key[pd->didx];
4530
4531                 if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], pd->af) ||
4532                     nk->port[pd->sidx] != uh->uh_sport) {
4533                         /*
4534                          * The translated source address may be completely
4535                          * unrelated to the saved link header, make sure
4536                          * a bridge doesn't try to use it.
4537                          */
4538                         m->m_pkthdr.fw_flags &= ~BRIDGE_MBUF_TAGGED;
4539                         m->m_flags &= ~M_HASH;
4540                         pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum,
4541                             &uh->uh_sum, &nk->addr[pd->sidx],
4542                             nk->port[pd->sidx], 1, pd->af);
4543                 }
4544
4545                 if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], pd->af) ||
4546                     nk->port[pd->didx] != uh->uh_dport) {
4547                         /*
4548                          * If we don't redispatch the packet will go into
4549                          * the protocol stack on the wrong cpu for the
4550                          * post-translated address.
4551                          */
4552                         m->m_flags &= ~M_HASH;
4553                         pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum,
4554                             &uh->uh_sum, &nk->addr[pd->didx],
4555                             nk->port[pd->didx], 1, pd->af);
4556                 }
4557                 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
4558         }
4559
4560         return (PF_PASS);
4561 }
4562
4563 int
4564 pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif,
4565     struct mbuf *m, int off, void *h, struct pf_pdesc *pd, u_short *reason)
4566 {
4567         struct pf_addr  *saddr = pd->src, *daddr = pd->dst;
4568         u_int16_t        icmpid = 0, *icmpsum;
4569         u_int8_t         icmptype;
4570         int              state_icmp = 0;
4571         struct pf_state_key_cmp key;
4572
4573         switch (pd->proto) {
4574 #ifdef INET
4575         case IPPROTO_ICMP:
4576                 icmptype = pd->hdr.icmp->icmp_type;
4577                 icmpid = pd->hdr.icmp->icmp_id;
4578                 icmpsum = &pd->hdr.icmp->icmp_cksum;
4579
4580                 if (icmptype == ICMP_UNREACH ||
4581                     icmptype == ICMP_SOURCEQUENCH ||
4582                     icmptype == ICMP_REDIRECT ||
4583                     icmptype == ICMP_TIMXCEED ||
4584                     icmptype == ICMP_PARAMPROB)
4585                         state_icmp++;
4586                 break;
4587 #endif /* INET */
4588 #ifdef INET6
4589         case IPPROTO_ICMPV6:
4590                 icmptype = pd->hdr.icmp6->icmp6_type;
4591                 icmpid = pd->hdr.icmp6->icmp6_id;
4592                 icmpsum = &pd->hdr.icmp6->icmp6_cksum;
4593
4594                 if (icmptype == ICMP6_DST_UNREACH ||
4595                     icmptype == ICMP6_PACKET_TOO_BIG ||
4596                     icmptype == ICMP6_TIME_EXCEEDED ||
4597                     icmptype == ICMP6_PARAM_PROB)
4598                         state_icmp++;
4599                 break;
4600 #endif /* INET6 */
4601         }
4602
4603         if (!state_icmp) {
4604
4605                 /*
4606                  * ICMP query/reply message not related to a TCP/UDP packet.
4607                  * Search for an ICMP state.
4608                  */
4609                 key.af = pd->af;
4610                 key.proto = pd->proto;
4611                 key.port[0] = key.port[1] = icmpid;
4612                 if (direction == PF_IN) {       /* wire side, straight */
4613                         PF_ACPY(&key.addr[0], pd->src, key.af);
4614                         PF_ACPY(&key.addr[1], pd->dst, key.af);
4615                 } else {                        /* stack side, reverse */
4616                         PF_ACPY(&key.addr[1], pd->src, key.af);
4617                         PF_ACPY(&key.addr[0], pd->dst, key.af);
4618                 }
4619
4620                 STATE_LOOKUP(kif, &key, direction, *state, m);
4621
4622                 (*state)->expire = time_second;
4623                 (*state)->timeout = PFTM_ICMP_ERROR_REPLY;
4624
4625                 /* translate source/destination address, if necessary */
4626                 if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) {
4627                         struct pf_state_key *nk = (*state)->key[pd->didx];
4628
4629                         switch (pd->af) {
4630 #ifdef INET
4631                         case AF_INET:
4632                                 if (PF_ANEQ(pd->src,
4633                                     &nk->addr[pd->sidx], AF_INET))
4634                                         pf_change_a(&saddr->v4.s_addr,
4635                                             pd->ip_sum,
4636                                             nk->addr[pd->sidx].v4.s_addr, 0);
4637
4638                                 if (PF_ANEQ(pd->dst, &nk->addr[pd->didx],
4639                                     AF_INET))
4640                                         pf_change_a(&daddr->v4.s_addr,
4641                                             pd->ip_sum,
4642                                             nk->addr[pd->didx].v4.s_addr, 0);
4643
4644                                 if (nk->port[0] !=
4645                                     pd->hdr.icmp->icmp_id) {
4646                                         pd->hdr.icmp->icmp_cksum =
4647                                             pf_cksum_fixup(
4648                                             pd->hdr.icmp->icmp_cksum, icmpid,
4649                                             nk->port[pd->sidx], 0);
4650                                         pd->hdr.icmp->icmp_id =
4651                                             nk->port[pd->sidx];
4652                                 }
4653
4654                                 m_copyback(m, off, ICMP_MINLEN,
4655                                     (caddr_t)pd->hdr.icmp);
4656                                 break;
4657 #endif /* INET */
4658 #ifdef INET6
4659                         case AF_INET6:
4660                                 if (PF_ANEQ(pd->src,
4661                                     &nk->addr[pd->sidx], AF_INET6))
4662                                         pf_change_a6(saddr,
4663                                             &pd->hdr.icmp6->icmp6_cksum,
4664                                             &nk->addr[pd->sidx], 0);
4665
4666                                 if (PF_ANEQ(pd->dst,
4667                                     &nk->addr[pd->didx], AF_INET6))
4668                                         pf_change_a6(daddr,
4669                                             &pd->hdr.icmp6->icmp6_cksum,
4670                                             &nk->addr[pd->didx], 0);
4671
4672                                 m_copyback(m, off,
4673                                         sizeof(struct icmp6_hdr),
4674                                         (caddr_t)pd->hdr.icmp6);
4675                                 break;
4676 #endif /* INET6 */
4677                         }
4678                 }
4679                 return (PF_PASS);
4680
4681         } else {
4682                 /*
4683                  * ICMP error message in response to a TCP/UDP packet.
4684                  * Extract the inner TCP/UDP header and search for that state.
4685                  */
4686
4687                 struct pf_pdesc pd2;
4688 #ifdef INET
4689                 struct ip       h2;
4690 #endif /* INET */
4691 #ifdef INET6
4692                 struct ip6_hdr  h2_6;
4693                 int             terminal = 0;
4694 #endif /* INET6 */
4695                 int             ipoff2;
4696                 int             off2;
4697
4698                 pd2.af = pd->af;
4699                 /* Payload packet is from the opposite direction. */
4700                 pd2.sidx = (direction == PF_IN) ? 1 : 0;
4701                 pd2.didx = (direction == PF_IN) ? 0 : 1;
4702                 switch (pd->af) {
4703 #ifdef INET
4704                 case AF_INET:
4705                         /* offset of h2 in mbuf chain */
4706                         ipoff2 = off + ICMP_MINLEN;
4707
4708                         if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2),
4709                             NULL, reason, pd2.af)) {
4710                                 DPFPRINTF(PF_DEBUG_MISC,
4711                                     ("pf: ICMP error message too short "
4712                                     "(ip)\n"));
4713                                 return (PF_DROP);
4714                         }
4715                         /*
4716                          * ICMP error messages don't refer to non-first
4717                          * fragments
4718                          */
4719                         if (h2.ip_off & htons(IP_OFFMASK)) {
4720                                 REASON_SET(reason, PFRES_FRAG);
4721                                 return (PF_DROP);
4722                         }
4723
4724                         /* offset of protocol header that follows h2 */
4725                         off2 = ipoff2 + (h2.ip_hl << 2);
4726
4727                         pd2.proto = h2.ip_p;
4728                         pd2.src = (struct pf_addr *)&h2.ip_src;
4729                         pd2.dst = (struct pf_addr *)&h2.ip_dst;
4730                         pd2.ip_sum = &h2.ip_sum;
4731                         break;
4732 #endif /* INET */
4733 #ifdef INET6
4734                 case AF_INET6:
4735                         ipoff2 = off + sizeof(struct icmp6_hdr);
4736
4737                         if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6),
4738                             NULL, reason, pd2.af)) {
4739                                 DPFPRINTF(PF_DEBUG_MISC,
4740                                     ("pf: ICMP error message too short "
4741                                     "(ip6)\n"));
4742                                 return (PF_DROP);
4743                         }
4744                         pd2.proto = h2_6.ip6_nxt;
4745                         pd2.src = (struct pf_addr *)&h2_6.ip6_src;
4746                         pd2.dst = (struct pf_addr *)&h2_6.ip6_dst;
4747                         pd2.ip_sum = NULL;
4748                         off2 = ipoff2 + sizeof(h2_6);
4749                         do {
4750                                 switch (pd2.proto) {
4751                                 case IPPROTO_FRAGMENT:
4752                                         /*
4753                                          * ICMPv6 error messages for
4754                                          * non-first fragments
4755                                          */
4756                                         REASON_SET(reason, PFRES_FRAG);
4757                                         return (PF_DROP);
4758                                 case IPPROTO_AH:
4759                                 case IPPROTO_HOPOPTS:
4760                                 case IPPROTO_ROUTING:
4761                                 case IPPROTO_DSTOPTS: {
4762                                         /* get next header and header length */
4763                                         struct ip6_ext opt6;
4764
4765                                         if (!pf_pull_hdr(m, off2, &opt6,
4766                                             sizeof(opt6), NULL, reason,
4767                                             pd2.af)) {
4768                                                 DPFPRINTF(PF_DEBUG_MISC,
4769                                                     ("pf: ICMPv6 short opt\n"));
4770                                                 return (PF_DROP);
4771                                         }
4772                                         if (pd2.proto == IPPROTO_AH)
4773                                                 off2 += (opt6.ip6e_len + 2) * 4;
4774                                         else
4775                                                 off2 += (opt6.ip6e_len + 1) * 8;
4776                                         pd2.proto = opt6.ip6e_nxt;
4777                                         /* goto the next header */
4778                                         break;
4779                                 }
4780                                 default:
4781                                         terminal++;
4782                                         break;
4783                                 }
4784                         } while (!terminal);
4785                         break;
4786 #endif /* INET6 */
4787                 default:
4788                         DPFPRINTF(PF_DEBUG_MISC,
4789                             ("pf: ICMP AF %d unknown (ip6)\n", pd->af));
4790                         return (PF_DROP);
4791                         break;
4792                 }
4793
4794                 switch (pd2.proto) {
4795                 case IPPROTO_TCP: {
4796                         struct tcphdr            th;
4797                         u_int32_t                seq;
4798                         struct pf_state_peer    *src, *dst;
4799                         u_int8_t                 dws;
4800                         int                      copyback = 0;
4801
4802                         /*
4803                          * Only the first 8 bytes of the TCP header can be
4804                          * expected. Don't access any TCP header fields after
4805                          * th_seq, an ackskew test is not possible.
4806                          */
4807                         if (!pf_pull_hdr(m, off2, &th, 8, NULL, reason,
4808                             pd2.af)) {
4809                                 DPFPRINTF(PF_DEBUG_MISC,
4810                                     ("pf: ICMP error message too short "
4811                                     "(tcp)\n"));
4812                                 return (PF_DROP);
4813                         }
4814
4815                         key.af = pd2.af;
4816                         key.proto = IPPROTO_TCP;
4817                         PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
4818                         PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
4819                         key.port[pd2.sidx] = th.th_sport;
4820                         key.port[pd2.didx] = th.th_dport;
4821
4822                         STATE_LOOKUP(kif, &key, direction, *state, m);
4823
4824                         if (direction == (*state)->direction) {
4825                                 src = &(*state)->dst;
4826                                 dst = &(*state)->src;
4827                         } else {
4828                                 src = &(*state)->src;
4829                                 dst = &(*state)->dst;
4830                         }
4831
4832                         if (src->wscale && dst->wscale)
4833                                 dws = dst->wscale & PF_WSCALE_MASK;
4834                         else
4835                                 dws = 0;
4836
4837                         /* Demodulate sequence number */
4838                         seq = ntohl(th.th_seq) - src->seqdiff;
4839                         if (src->seqdiff) {
4840                                 pf_change_a(&th.th_seq, icmpsum,
4841                                     htonl(seq), 0);
4842                                 copyback = 1;
4843                         }
4844
4845                         if (!((*state)->state_flags & PFSTATE_SLOPPY) &&
4846                             (!SEQ_GEQ(src->seqhi, seq) ||
4847                             !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)))) {
4848                                 if (pf_status.debug >= PF_DEBUG_MISC) {
4849                                         kprintf("pf: BAD ICMP %d:%d ",
4850                                             icmptype, pd->hdr.icmp->icmp_code);
4851                                         pf_print_host(pd->src, 0, pd->af);
4852                                         kprintf(" -> ");
4853                                         pf_print_host(pd->dst, 0, pd->af);
4854                                         kprintf(" state: ");
4855                                         pf_print_state(*state);
4856                                         kprintf(" seq=%u\n", seq);
4857                                 }
4858                                 REASON_SET(reason, PFRES_BADSTATE);
4859                                 return (PF_DROP);
4860                         } else {
4861                                 if (pf_status.debug >= PF_DEBUG_MISC) {
4862                                         kprintf("pf: OK ICMP %d:%d ",
4863                                             icmptype, pd->hdr.icmp->icmp_code);
4864                                         pf_print_host(pd->src, 0, pd->af);
4865                                         kprintf(" -> ");
4866                                         pf_print_host(pd->dst, 0, pd->af);
4867                                         kprintf(" state: ");
4868                                         pf_print_state(*state);
4869                                         kprintf(" seq=%u\n", seq);
4870                                 }
4871                         }
4872
4873                         /* translate source/destination address, if necessary */
4874                         if ((*state)->key[PF_SK_WIRE] !=
4875                             (*state)->key[PF_SK_STACK]) {
4876                                 struct pf_state_key *nk =
4877                                     (*state)->key[pd->didx];
4878
4879                                 if (PF_ANEQ(pd2.src,
4880                                     &nk->addr[pd2.sidx], pd2.af) ||
4881                                     nk->port[pd2.sidx] != th.th_sport)
4882                                         pf_change_icmp(pd2.src, &th.th_sport,
4883                                             daddr, &nk->addr[pd2.sidx],
4884                                             nk->port[pd2.sidx], NULL,
4885                                             pd2.ip_sum, icmpsum,
4886                                             pd->ip_sum, 0, pd2.af);
4887
4888                                 if (PF_ANEQ(pd2.dst,
4889                                     &nk->addr[pd2.didx], pd2.af) ||
4890                                     nk->port[pd2.didx] != th.th_dport)
4891                                         pf_change_icmp(pd2.dst, &th.th_dport,
4892                                             NULL, /* XXX Inbound NAT? */
4893                                             &nk->addr[pd2.didx],
4894                                             nk->port[pd2.didx], NULL,
4895                                             pd2.ip_sum, icmpsum,
4896                                             pd->ip_sum, 0, pd2.af);
4897                                 copyback = 1;
4898                         }
4899
4900                         if (copyback) {
4901                                 switch (pd2.af) {
4902 #ifdef INET
4903                                 case AF_INET:
4904                                         m_copyback(m, off, ICMP_MINLEN,
4905                                             (caddr_t)pd->hdr.icmp);
4906                                         m_copyback(m, ipoff2, sizeof(h2),
4907                                             (caddr_t)&h2);
4908                                         break;
4909 #endif /* INET */
4910 #ifdef INET6
4911                                 case AF_INET6:
4912                                         m_copyback(m, off,
4913                                             sizeof(struct icmp6_hdr),
4914                                             (caddr_t)pd->hdr.icmp6);
4915                                         m_copyback(m, ipoff2, sizeof(h2_6),
4916                                             (caddr_t)&h2_6);
4917                                         break;
4918 #endif /* INET6 */
4919                                 }
4920                                 m_copyback(m, off2, 8, (caddr_t)&th);
4921                         }
4922
4923                         return (PF_PASS);
4924                         break;
4925                 }
4926                 case IPPROTO_UDP: {
4927                         struct udphdr           uh;
4928
4929                         if (!pf_pull_hdr(m, off2, &uh, sizeof(uh),
4930                             NULL, reason, pd2.af)) {
4931                                 DPFPRINTF(PF_DEBUG_MISC,
4932                                     ("pf: ICMP error message too short "
4933                                     "(udp)\n"));
4934                                 return (PF_DROP);
4935                         }
4936
4937                         key.af = pd2.af;
4938                         key.proto = IPPROTO_UDP;
4939                         PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
4940                         PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
4941                         key.port[pd2.sidx] = uh.uh_sport;
4942                         key.port[pd2.didx] = uh.uh_dport;
4943
4944                         STATE_LOOKUP(kif, &key, direction, *state, m);
4945
4946                         /* translate source/destination address, if necessary */
4947                         if ((*state)->key[PF_SK_WIRE] !=
4948                             (*state)->key[PF_SK_STACK]) {
4949                                 struct pf_state_key *nk =
4950                                     (*state)->key[pd->didx];
4951
4952                                 if (PF_ANEQ(pd2.src,
4953                                     &nk->addr[pd2.sidx], pd2.af) ||
4954                                     nk->port[pd2.sidx] != uh.uh_sport)
4955                                         pf_change_icmp(pd2.src, &uh.uh_sport,
4956                                             daddr, &nk->addr[pd2.sidx],
4957                                             nk->port[pd2.sidx], &uh.uh_sum,
4958                                             pd2.ip_sum, icmpsum,
4959                                             pd->ip_sum, 1, pd2.af);
4960
4961                                 if (PF_ANEQ(pd2.dst,
4962                                     &nk->addr[pd2.didx], pd2.af) ||
4963                                     nk->port[pd2.didx] != uh.uh_dport)
4964                                         pf_change_icmp(pd2.dst, &uh.uh_dport,
4965                                             NULL, /* XXX Inbound NAT? */
4966                                             &nk->addr[pd2.didx],
4967                                             nk->port[pd2.didx], &uh.uh_sum,
4968                                             pd2.ip_sum, icmpsum,
4969                                             pd->ip_sum, 1, pd2.af);
4970
4971                                 switch (pd2.af) {
4972 #ifdef INET
4973                                 case AF_INET:
4974                                         m_copyback(m, off, ICMP_MINLEN,
4975                                             (caddr_t)pd->hdr.icmp);
4976                                         m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4977                                         break;
4978 #endif /* INET */
4979 #ifdef INET6
4980                                 case AF_INET6:
4981                                         m_copyback(m, off,
4982                                             sizeof(struct icmp6_hdr),
4983                                             (caddr_t)pd->hdr.icmp6);
4984                                         m_copyback(m, ipoff2, sizeof(h2_6),
4985                                             (caddr_t)&h2_6);
4986                                         break;
4987 #endif /* INET6 */
4988                                 }
4989                                 m_copyback(m, off2, sizeof(uh), (caddr_t)&uh);
4990                         }
4991
4992                         return (PF_PASS);
4993                         break;
4994                 }
4995 #ifdef INET
4996                 case IPPROTO_ICMP: {
4997                         struct icmp             iih;
4998
4999                         if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN,
5000                             NULL, reason, pd2.af)) {
5001                                 DPFPRINTF(PF_DEBUG_MISC,
5002                                     ("pf: ICMP error message too short i"
5003                                     "(icmp)\n"));
5004                                 return (PF_DROP);
5005                         }
5006
5007                         key.af = pd2.af;
5008                         key.proto = IPPROTO_ICMP;
5009                         PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
5010                         PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
5011                         key.port[0] = key.port[1] = iih.icmp_id;
5012
5013                         STATE_LOOKUP(kif, &key, direction, *state, m);
5014
5015                         /* translate source/destination address, if necessary */
5016                         if ((*state)->key[PF_SK_WIRE] !=
5017                             (*state)->key[PF_SK_STACK]) {
5018                                 struct pf_state_key *nk =
5019                                     (*state)->key[pd->didx];
5020
5021                                 if (PF_ANEQ(pd2.src,
5022                                     &nk->addr[pd2.sidx], pd2.af) ||
5023                                     nk->port[pd2.sidx] != iih.icmp_id)
5024                                         pf_change_icmp(pd2.src, &iih.icmp_id,
5025                                             daddr, &nk->addr[pd2.sidx],
5026                                             nk->port[pd2.sidx], NULL,
5027                                             pd2.ip_sum, icmpsum,
5028                                             pd->ip_sum, 0, AF_INET);
5029
5030                                 if (PF_ANEQ(pd2.dst,
5031                                     &nk->addr[pd2.didx], pd2.af) ||
5032                                     nk->port[pd2.didx] != iih.icmp_id)
5033                                         pf_change_icmp(pd2.dst, &iih.icmp_id,
5034                                             NULL, /* XXX Inbound NAT? */
5035                                             &nk->addr[pd2.didx],
5036                                             nk->port[pd2.didx], NULL,
5037                                             pd2.ip_sum, icmpsum,
5038                                             pd->ip_sum, 0, AF_INET);
5039
5040                                 m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
5041                                 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
5042                                 m_copyback(m, off2, ICMP_MINLEN, (caddr_t)&iih);
5043                         }
5044                         return (PF_PASS);
5045                         break;
5046                 }
5047 #endif /* INET */
5048 #ifdef INET6
5049                 case IPPROTO_ICMPV6: {
5050                         struct icmp6_hdr        iih;
5051
5052                         if (!pf_pull_hdr(m, off2, &iih,
5053                             sizeof(struct icmp6_hdr), NULL, reason, pd2.af)) {
5054                                 DPFPRINTF(PF_DEBUG_MISC,
5055                                     ("pf: ICMP error message too short "
5056                                     "(icmp6)\n"));
5057                                 return (PF_DROP);
5058                         }
5059
5060                         key.af = pd2.af;
5061                         key.proto = IPPROTO_ICMPV6;
5062                         PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
5063                         PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
5064                         key.port[0] = key.port[1] = iih.icmp6_id;
5065
5066                         STATE_LOOKUP(kif, &key, direction, *state, m);
5067
5068                         /* translate source/destination address, if necessary */
5069                         if ((*state)->key[PF_SK_WIRE] !=
5070                             (*state)->key[PF_SK_STACK]) {
5071                                 struct pf_state_key *nk =
5072                                     (*state)->key[pd->didx];
5073
5074                                 if (PF_ANEQ(pd2.src,
5075                                     &nk->addr[pd2.sidx], pd2.af) ||
5076                                     nk->port[pd2.sidx] != iih.icmp6_id)
5077                                         pf_change_icmp(pd2.src, &iih.icmp6_id,
5078                                             daddr, &nk->addr[pd2.sidx],
5079                                             nk->port[pd2.sidx], NULL,
5080                                             pd2.ip_sum, icmpsum,
5081                                             pd->ip_sum, 0, AF_INET6);
5082
5083                                 if (PF_ANEQ(pd2.dst,
5084                                     &nk->addr[pd2.didx], pd2.af) ||
5085                                     nk->port[pd2.didx] != iih.icmp6_id)
5086                                         pf_change_icmp(pd2.dst, &iih.icmp6_id,
5087                                             NULL, /* XXX Inbound NAT? */
5088                                             &nk->addr[pd2.didx],
5089                                             nk->port[pd2.didx], NULL,
5090                                             pd2.ip_sum, icmpsum,
5091                                             pd->ip_sum, 0, AF_INET6);
5092
5093                                 m_copyback(m, off, sizeof(struct icmp6_hdr),
5094                                     (caddr_t)pd->hdr.icmp6);
5095                                 m_copyback(m, ipoff2, sizeof(h2_6), (caddr_t)&h2_6);
5096                                 m_copyback(m, off2, sizeof(struct icmp6_hdr),
5097                                     (caddr_t)&iih);
5098                         }
5099
5100                         return (PF_PASS);
5101                         break;
5102                 }
5103 #endif /* INET6 */
5104                 default: {
5105                         key.af = pd2.af;
5106                         key.proto = pd2.proto;
5107                         PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
5108                         PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
5109                         key.port[0] = key.port[1] = 0;
5110
5111                         STATE_LOOKUP(kif, &key, direction, *state, m);
5112
5113                         /* translate source/destination address, if necessary */
5114                         if ((*state)->key[PF_SK_WIRE] !=
5115                             (*state)->key[PF_SK_STACK]) {
5116                                 struct pf_state_key *nk =
5117                                     (*state)->key[pd->didx];
5118
5119                                 if (PF_ANEQ(pd2.src,
5120                                     &nk->addr[pd2.sidx], pd2.af))
5121                                         pf_change_icmp(pd2.src, NULL, daddr,
5122                                             &nk->addr[pd2.sidx], 0, NULL,
5123                                             pd2.ip_sum, icmpsum,
5124                                             pd->ip_sum, 0, pd2.af);
5125
5126                                 if (PF_ANEQ(pd2.dst,
5127                                     &nk->addr[pd2.didx], pd2.af))
5128                                         pf_change_icmp(pd2.src, NULL,
5129                                             NULL, /* XXX Inbound NAT? */
5130                                             &nk->addr[pd2.didx], 0, NULL,
5131                                             pd2.ip_sum, icmpsum,
5132                                             pd->ip_sum, 0, pd2.af);
5133
5134                                 switch (pd2.af) {
5135 #ifdef INET
5136                                 case AF_INET:
5137                                         m_copyback(m, off, ICMP_MINLEN,
5138                                             (caddr_t)pd->hdr.icmp);
5139                                         m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
5140                                         break;
5141 #endif /* INET */
5142 #ifdef INET6
5143                                 case AF_INET6:
5144                                         m_copyback(m, off,
5145                                             sizeof(struct icmp6_hdr),
5146                                             (caddr_t)pd->hdr.icmp6);
5147                                         m_copyback(m, ipoff2, sizeof(h2_6),
5148                                             (caddr_t)&h2_6);
5149                                         break;
5150 #endif /* INET6 */
5151                                 }
5152                         }
5153                         return (PF_PASS);
5154                         break;
5155                 }
5156                 }
5157         }
5158 }
5159
5160 int
5161 pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif,
5162     struct mbuf *m, struct pf_pdesc *pd)
5163 {
5164         struct pf_state_peer    *src, *dst;
5165         struct pf_state_key_cmp  key;
5166
5167         key.af = pd->af;
5168         key.proto = pd->proto;
5169         if (direction == PF_IN) {
5170                 PF_ACPY(&key.addr[0], pd->src, key.af);
5171                 PF_ACPY(&key.addr[1], pd->dst, key.af);
5172                 key.port[0] = key.port[1] = 0;
5173         } else {
5174                 PF_ACPY(&key.addr[1], pd->src, key.af);
5175                 PF_ACPY(&key.addr[0], pd->dst, key.af);
5176                 key.port[1] = key.port[0] = 0;
5177         }
5178
5179         STATE_LOOKUP(kif, &key, direction, *state, m);
5180
5181         if (direction == (*state)->direction) {
5182                 src = &(*state)->src;
5183                 dst = &(*state)->dst;
5184         } else {
5185                 src = &(*state)->dst;
5186                 dst = &(*state)->src;
5187         }
5188
5189         /* update states */
5190         if (src->state < PFOTHERS_SINGLE)
5191                 src->state = PFOTHERS_SINGLE;
5192         if (dst->state == PFOTHERS_SINGLE)
5193                 dst->state = PFOTHERS_MULTIPLE;
5194
5195         /* update expire time */
5196         (*state)->expire = time_second;
5197         if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE)
5198                 (*state)->timeout = PFTM_OTHER_MULTIPLE;
5199         else
5200                 (*state)->timeout = PFTM_OTHER_SINGLE;
5201
5202         /* translate source/destination address, if necessary */
5203         if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) {
5204                 struct pf_state_key *nk = (*state)->key[pd->didx];
5205
5206                 KKASSERT(nk);
5207                 KKASSERT(pd);
5208                 KKASSERT(pd->src);
5209                 KKASSERT(pd->dst);
5210                 switch (pd->af) {
5211 #ifdef INET
5212                 case AF_INET:
5213                         if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], AF_INET))
5214                                 pf_change_a(&pd->src->v4.s_addr,
5215                                     pd->ip_sum,
5216                                     nk->addr[pd->sidx].v4.s_addr,
5217                                     0);
5218
5219
5220                         if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], AF_INET))
5221                                 pf_change_a(&pd->dst->v4.s_addr,
5222                                     pd->ip_sum,
5223                                     nk->addr[pd->didx].v4.s_addr,
5224                                     0);
5225
5226                         break;
5227 #endif /* INET */
5228 #ifdef INET6
5229                 case AF_INET6:
5230                         if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], AF_INET))
5231                                 PF_ACPY(pd->src, &nk->addr[pd->sidx], pd->af);
5232
5233                         if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], AF_INET))
5234                                 PF_ACPY(pd->dst, &nk->addr[pd->didx], pd->af);
5235 #endif /* INET6 */
5236                 }
5237         }
5238         return (PF_PASS);
5239 }
5240
5241 /*
5242  * ipoff and off are measured from the start of the mbuf chain.
5243  * h must be at "ipoff" on the mbuf chain.
5244  */
5245 void *
5246 pf_pull_hdr(struct mbuf *m, int off, void *p, int len,
5247     u_short *actionp, u_short *reasonp, sa_family_t af)
5248 {
5249         switch (af) {
5250 #ifdef INET
5251         case AF_INET: {
5252                 struct ip       *h = mtod(m, struct ip *);
5253                 u_int16_t        fragoff = (h->ip_off & IP_OFFMASK) << 3;
5254
5255                 if (fragoff) {
5256                         if (fragoff >= len)
5257                                 ACTION_SET(actionp, PF_PASS);
5258                         else {
5259                                 ACTION_SET(actionp, PF_DROP);
5260                                 REASON_SET(reasonp, PFRES_FRAG);
5261                         }
5262                         return (NULL);
5263                 }
5264                 if (m->m_pkthdr.len < off + len ||
5265                     h->ip_len < off + len) {
5266                         ACTION_SET(actionp, PF_DROP);
5267                         REASON_SET(reasonp, PFRES_SHORT);
5268                         return (NULL);
5269                 }
5270                 break;
5271         }
5272 #endif /* INET */
5273 #ifdef INET6
5274         case AF_INET6: {
5275                 struct ip6_hdr  *h = mtod(m, struct ip6_hdr *);
5276
5277                 if (m->m_pkthdr.len < off + len ||
5278                     (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) <
5279                     (unsigned)(off + len)) {
5280                         ACTION_SET(actionp, PF_DROP);
5281                         REASON_SET(reasonp, PFRES_SHORT);
5282                         return (NULL);
5283                 }
5284                 break;
5285         }
5286 #endif /* INET6 */
5287         }
5288         m_copydata(m, off, len, p);
5289         return (p);
5290 }
5291
5292 int
5293 pf_routable(struct pf_addr *addr, sa_family_t af, struct pfi_kif *kif)
5294 {
5295         struct sockaddr_in      *dst;
5296         int                      ret = 1;
5297         int                      check_mpath;
5298 #ifdef INET6
5299         struct sockaddr_in6     *dst6;
5300         struct route_in6         ro;
5301 #else
5302         struct route             ro;
5303 #endif
5304         struct radix_node       *rn;
5305         struct rtentry          *rt;
5306         struct ifnet            *ifp;
5307
5308         check_mpath = 0;
5309         bzero(&ro, sizeof(ro));
5310         switch (af) {
5311         case AF_INET:
5312                 dst = satosin(&ro.ro_dst);
5313                 dst->sin_family = AF_INET;
5314                 dst->sin_len = sizeof(*dst);
5315                 dst->sin_addr = addr->v4;
5316                 break;
5317 #ifdef INET6
5318         case AF_INET6:
5319                 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
5320                 dst6->sin6_family = AF_INET6;
5321                 dst6->sin6_len = sizeof(*dst6);
5322                 dst6->sin6_addr = addr->v6;
5323                 break;
5324 #endif /* INET6 */
5325         default:
5326                 return (0);
5327         }
5328
5329         /* Skip checks for ipsec interfaces */
5330         if (kif != NULL && kif->pfik_ifp->if_type == IFT_ENC)
5331                 goto out;
5332
5333         rtalloc_ign((struct route *)&ro, 0);
5334
5335         if (ro.ro_rt != NULL) {
5336                 /* No interface given, this is a no-route check */
5337                 if (kif == NULL)
5338                         goto out;
5339
5340                 if (kif->pfik_ifp == NULL) {
5341                         ret = 0;
5342                         goto out;
5343                 }
5344
5345                 /* Perform uRPF check if passed input interface */
5346                 ret = 0;
5347                 rn = (struct radix_node *)ro.ro_rt;
5348                 do {
5349                         rt = (struct rtentry *)rn;
5350                         ifp = rt->rt_ifp;
5351
5352                         if (kif->pfik_ifp == ifp)
5353                                 ret = 1;
5354                         rn = NULL;
5355                 } while (check_mpath == 1 && rn != NULL && ret == 0);
5356         } else
5357                 ret = 0;
5358 out:
5359         if (ro.ro_rt != NULL)
5360                 RTFREE(ro.ro_rt);
5361         return (ret);
5362 }
5363
5364 int
5365 pf_rtlabel_match(struct pf_addr *addr, sa_family_t af, struct pf_addr_wrap *aw)
5366 {
5367         struct sockaddr_in      *dst;
5368 #ifdef INET6
5369         struct sockaddr_in6     *dst6;
5370         struct route_in6         ro;
5371 #else
5372         struct route             ro;
5373 #endif
5374         int                      ret = 0;
5375
5376         ASSERT_LWKT_TOKEN_HELD(&pf_token);
5377
5378         bzero(&ro, sizeof(ro));
5379         switch (af) {
5380         case AF_INET:
5381                 dst = satosin(&ro.ro_dst);
5382                 dst->sin_family = AF_INET;
5383                 dst->sin_len = sizeof(*dst);
5384                 dst->sin_addr = addr->v4;
5385                 break;
5386 #ifdef INET6
5387         case AF_INET6:
5388                 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
5389                 dst6->sin6_family = AF_INET6;
5390                 dst6->sin6_len = sizeof(*dst6);
5391                 dst6->sin6_addr = addr->v6;
5392                 break;
5393 #endif /* INET6 */
5394         default:
5395                 return (0);
5396         }
5397
5398 rtalloc_ign((struct route *)&ro, (RTF_CLONING | RTF_PRCLONING));
5399
5400         if (ro.ro_rt != NULL) {
5401                 RTFREE(ro.ro_rt);
5402         }
5403
5404         return (ret);
5405 }
5406
5407 #ifdef INET
5408 void
5409 pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5410     struct pf_state *s, struct pf_pdesc *pd)
5411 {
5412         struct mbuf             *m0, *m1;
5413         struct route             iproute;
5414         struct route            *ro = NULL;
5415         struct sockaddr_in      *dst;
5416         struct ip               *ip;
5417         struct ifnet            *ifp = NULL;
5418         struct pf_addr           naddr;
5419         struct pf_src_node      *sn = NULL;
5420         int                      error = 0;
5421         int sw_csum;
5422 #ifdef IPSEC
5423         struct m_tag            *mtag;
5424 #endif /* IPSEC */
5425
5426         ASSERT_LWKT_TOKEN_HELD(&pf_token);
5427
5428         if (m == NULL || *m == NULL || r == NULL ||
5429             (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5430                 panic("pf_route: invalid parameters");
5431
5432         if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
5433                 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
5434                 (*m)->m_pkthdr.pf.routed = 1;
5435         } else {
5436                 if ((*m)->m_pkthdr.pf.routed++ > 3) {
5437                         m0 = *m;
5438                         *m = NULL;
5439                         goto bad;
5440                 }
5441         }
5442
5443         if (r->rt == PF_DUPTO) {
5444                 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL) {
5445                         return;
5446                 }
5447         } else {
5448                 if ((r->rt == PF_REPLYTO) == (r->direction == dir)) {
5449                         return;
5450                 }
5451                 m0 = *m;
5452         }
5453
5454         if (m0->m_len < sizeof(struct ip)) {
5455                 DPFPRINTF(PF_DEBUG_URGENT,
5456                     ("pf_route: m0->m_len < sizeof(struct ip)\n"));
5457                 goto bad;
5458         }
5459
5460         ip = mtod(m0, struct ip *);
5461
5462         ro = &iproute;
5463         bzero((caddr_t)ro, sizeof(*ro));
5464         dst = satosin(&ro->ro_dst);
5465         dst->sin_family = AF_INET;
5466         dst->sin_len = sizeof(*dst);
5467         dst->sin_addr = ip->ip_dst;
5468
5469         if (r->rt == PF_FASTROUTE) {
5470                 rtalloc(ro);
5471                 if (ro->ro_rt == 0) {
5472                         ipstat.ips_noroute++;
5473                         goto bad;
5474                 }
5475
5476                 ifp = ro->ro_rt->rt_ifp;
5477                 ro->ro_rt->rt_use++;
5478
5479                 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
5480                         dst = satosin(ro->ro_rt->rt_gateway);
5481         } else {
5482                 if (TAILQ_EMPTY(&r->rpool.list)) {
5483                         DPFPRINTF(PF_DEBUG_URGENT,
5484                             ("pf_route: TAILQ_EMPTY(&r->rpool.list)\n"));
5485                         goto bad;
5486                 }
5487                 if (s == NULL) {
5488                         pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src,
5489                             &naddr, NULL, &sn);
5490                         if (!PF_AZERO(&naddr, AF_INET))
5491                                 dst->sin_addr.s_addr = naddr.v4.s_addr;
5492                         ifp = r->rpool.cur->kif ?
5493                             r->rpool.cur->kif->pfik_ifp : NULL;
5494                 } else {
5495                         if (!PF_AZERO(&s->rt_addr, AF_INET))
5496                                 dst->sin_addr.s_addr =
5497                                     s->rt_addr.v4.s_addr;
5498                         ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5499                 }
5500         }
5501         if (ifp == NULL)
5502                 goto bad;
5503
5504         if (oifp != ifp) {
5505                 if (pf_test(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) {
5506                         goto bad;
5507                 } else if (m0 == NULL) {
5508                         goto done;
5509                 }
5510                 if (m0->m_len < sizeof(struct ip)) {
5511                         DPFPRINTF(PF_DEBUG_URGENT,
5512                             ("pf_route: m0->m_len < sizeof(struct ip)\n"));
5513                         goto bad;
5514                 }
5515                 ip = mtod(m0, struct ip *);
5516         }
5517
5518         /* Copied from FreeBSD 5.1-CURRENT ip_output. */
5519         m0->m_pkthdr.csum_flags |= CSUM_IP;
5520         sw_csum = m0->m_pkthdr.csum_flags & ~ifp->if_hwassist;
5521         if (sw_csum & CSUM_DELAY_DATA) {
5522                 in_delayed_cksum(m0);
5523                 sw_csum &= ~CSUM_DELAY_DATA;
5524         }
5525         m0->m_pkthdr.csum_flags &= ifp->if_hwassist;
5526         m0->m_pkthdr.csum_iphlen = (ip->ip_hl << 2);
5527
5528         if (ip->ip_len <= ifp->if_mtu ||
5529             (ifp->if_hwassist & CSUM_FRAGMENT &&
5530                 (ip->ip_off & IP_DF) == 0)) {
5531                 ip->ip_len = htons(ip->ip_len);
5532                 ip->ip_off = htons(ip->ip_off);
5533                 ip->ip_sum = 0;
5534                 if (sw_csum & CSUM_DELAY_IP) {
5535                         /* From KAME */
5536                         if (ip->ip_v == IPVERSION &&
5537                             (ip->ip_hl << 2) == sizeof(*ip)) {
5538                                 ip->ip_sum = in_cksum_hdr(ip);
5539                         } else {
5540                                 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
5541                         }
5542                 }
5543                 lwkt_reltoken(&pf_token);
5544                 error = ifp->if_output(ifp, m0, sintosa(dst), ro->ro_rt);
5545                 lwkt_gettoken(&pf_token);
5546                 goto done;
5547         }
5548
5549         /*
5550          * Too large for interface; fragment if possible.
5551          * Must be able to put at least 8 bytes per fragment.
5552          */
5553         if (ip->ip_off & IP_DF) {
5554                 ipstat.ips_cantfrag++;
5555                 if (r->rt != PF_DUPTO) {
5556                         icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0,
5557                             ifp->if_mtu);
5558                         goto done;
5559                 } else
5560                         goto bad;
5561         }
5562
5563         m1 = m0;
5564         error = ip_fragment(ip, &m0, ifp->if_mtu, ifp->if_hwassist, sw_csum);
5565         if (error) {
5566                 goto bad;
5567         }
5568
5569         for (m0 = m1; m0; m0 = m1) {
5570                 m1 = m0->m_nextpkt;
5571                 m0->m_nextpkt = 0;
5572                 if (error == 0) {
5573                         lwkt_reltoken(&pf_token);
5574                         error = (*ifp->if_output)(ifp, m0, sintosa(dst),
5575                                                   NULL);
5576                         lwkt_gettoken(&pf_token);
5577                 } else
5578                         m_freem(m0);
5579         }
5580
5581         if (error == 0)
5582                 ipstat.ips_fragmented++;
5583
5584 done:
5585         if (r->rt != PF_DUPTO)
5586                 *m = NULL;
5587         if (ro == &iproute && ro->ro_rt)
5588                 RTFREE(ro->ro_rt);
5589         return;
5590
5591 bad:
5592         m_freem(m0);
5593         goto done;
5594 }
5595 #endif /* INET */
5596
5597 #ifdef INET6
5598 void
5599 pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5600     struct pf_state *s, struct pf_pdesc *pd)
5601 {
5602         struct mbuf             *m0;
5603         struct route_in6         ip6route;
5604         struct route_in6        *ro;
5605         struct sockaddr_in6     *dst;
5606         struct ip6_hdr          *ip6;
5607         struct ifnet            *ifp = NULL;
5608         struct pf_addr           naddr;
5609         struct pf_src_node      *sn = NULL;
5610
5611         if (m == NULL || *m == NULL || r == NULL ||
5612             (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5613                 panic("pf_route6: invalid parameters");
5614
5615         if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
5616                 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
5617                 (*m)->m_pkthdr.pf.routed = 1;
5618         } else {
5619                 if ((*m)->m_pkthdr.pf.routed++ > 3) {
5620                         m0 = *m;
5621                         *m = NULL;
5622                         goto bad;
5623                 }
5624         }
5625
5626         if (r->rt == PF_DUPTO) {
5627                 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL)
5628                         return;
5629         } else {
5630                 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5631                         return;
5632                 m0 = *m;
5633         }
5634
5635         if (m0->m_len < sizeof(struct ip6_hdr)) {
5636                 DPFPRINTF(PF_DEBUG_URGENT,
5637                     ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n"));
5638                 goto bad;
5639         }
5640         ip6 = mtod(m0, struct ip6_hdr *);
5641
5642         ro = &ip6route;
5643         bzero((caddr_t)ro, sizeof(*ro));
5644         dst = (struct sockaddr_in6 *)&ro->ro_dst;
5645         dst->sin6_family = AF_INET6;
5646         dst->sin6_len = sizeof(*dst);
5647         dst->sin6_addr = ip6->ip6_dst;
5648
5649         /*
5650          * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags,
5651          * so make sure pf.flags is clear.
5652          *
5653          * Cheat. XXX why only in the v6 case???
5654          */
5655         if (r->rt == PF_FASTROUTE) {
5656                 m0->m_pkthdr.fw_flags |= PF_MBUF_TAGGED;
5657                 m0->m_pkthdr.pf.flags = 0;
5658                 /* XXX Re-Check when Upgrading to > 4.4 */
5659                 m0->m_pkthdr.pf.statekey = NULL;
5660                 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
5661                 return;
5662         }
5663
5664         if (TAILQ_EMPTY(&r->rpool.list)) {
5665                 DPFPRINTF(PF_DEBUG_URGENT,
5666                     ("pf_route6: TAILQ_EMPTY(&r->rpool.list)\n"));
5667                 goto bad;
5668         }
5669         if (s == NULL) {
5670                 pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src,
5671                     &naddr, NULL, &sn);
5672                 if (!PF_AZERO(&naddr, AF_INET6))
5673                         PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5674                             &naddr, AF_INET6);
5675                 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL;
5676         } else {
5677                 if (!PF_AZERO(&s->rt_addr, AF_INET6))
5678                         PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5679                             &s->rt_addr, AF_INET6);
5680                 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5681         }
5682         if (ifp == NULL)
5683                 goto bad;
5684
5685         if (oifp != ifp) {
5686                 if (pf_test6(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) {
5687                         goto bad;
5688                 } else if (m0 == NULL) {
5689                         goto done;
5690                 }
5691                 if (m0->m_len < sizeof(struct ip6_hdr)) {
5692                         DPFPRINTF(PF_DEBUG_URGENT,
5693                             ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n"));
5694                         goto bad;
5695                 }
5696                 ip6 = mtod(m0, struct ip6_hdr *);
5697         }
5698
5699         /*
5700          * If the packet is too large for the outgoing interface,
5701          * send back an icmp6 error.
5702          */
5703         if (IN6_IS_ADDR_LINKLOCAL(&dst->sin6_addr))
5704                 dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
5705         if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) {
5706                 nd6_output(ifp, ifp, m0, dst, NULL);
5707         } else {
5708                 in6_ifstat_inc(ifp, ifs6_in_toobig);
5709                 if (r->rt != PF_DUPTO)
5710                         icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
5711                 else
5712                         goto bad;
5713         }
5714
5715 done:
5716         if (r->rt != PF_DUPTO)
5717                 *m = NULL;
5718         return;
5719
5720 bad:
5721         m_freem(m0);
5722         goto done;
5723 }
5724 #endif /* INET6 */
5725
5726
5727 /*
5728  * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag
5729  *   off is the offset where the protocol header starts
5730  *   len is the total length of protocol header plus payload
5731  * returns 0 when the checksum is valid, otherwise returns 1.
5732  */
5733 /*
5734  * XXX
5735  * FreeBSD supports cksum offload for the following drivers.
5736  * em(4), gx(4), lge(4), nge(4), ti(4), xl(4)
5737  * If we can make full use of it we would outperform ipfw/ipfilter in
5738  * very heavy traffic.
5739  * I have not tested 'cause I don't have NICs that supports cksum offload.
5740  * (There might be problems. Typical phenomena would be
5741  *   1. No route message for UDP packet.
5742  *   2. No connection acceptance from external hosts regardless of rule set.)
5743  */
5744 int
5745 pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p,
5746     sa_family_t af)
5747 {
5748         u_int16_t sum = 0;
5749         int hw_assist = 0;
5750         struct ip *ip;
5751
5752         if (off < sizeof(struct ip) || len < sizeof(struct udphdr))
5753                 return (1);
5754         if (m->m_pkthdr.len < off + len)
5755                 return (1);
5756
5757         switch (p) {
5758         case IPPROTO_TCP:
5759         case IPPROTO_UDP:
5760                 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
5761                         if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) {
5762                                 sum = m->m_pkthdr.csum_data;
5763                         } else {
5764                                 ip = mtod(m, struct ip *);
5765                                 sum = in_pseudo(ip->ip_src.s_addr,
5766                                         ip->ip_dst.s_addr, htonl((u_short)len +
5767                                         m->m_pkthdr.csum_data + p));
5768                         }
5769                         sum ^= 0xffff;
5770                         ++hw_assist;
5771                 }
5772                 break;
5773         case IPPROTO_ICMP:
5774 #ifdef INET6
5775         case IPPROTO_ICMPV6:
5776 #endif /* INET6 */
5777                 break;
5778         default:
5779                 return (1);
5780         }
5781
5782         if (!hw_assist) {
5783                 switch (af) {
5784                 case AF_INET:
5785                         if (p == IPPROTO_ICMP) {
5786                                 if (m->m_len < off)
5787                                         return (1);
5788                                 m->m_data += off;
5789                                 m->m_len -= off;
5790                                 sum = in_cksum(m, len);
5791                                 m->m_data -= off;
5792                                 m->m_len += off;
5793                         } else {
5794                                 if (m->m_len < sizeof(struct ip))
5795                                         return (1);
5796                                 sum = in_cksum_range(m, p, off, len);
5797                                 if (sum == 0) {
5798                                         m->m_pkthdr.csum_flags |=
5799                                             (CSUM_DATA_VALID |
5800                                              CSUM_PSEUDO_HDR);
5801                                         m->m_pkthdr.csum_data = 0xffff;
5802                                 }
5803                         }
5804                         break;
5805 #ifdef INET6
5806                 case AF_INET6:
5807                         if (m->m_len < sizeof(struct ip6_hdr))
5808                                 return (1);
5809                         sum = in6_cksum(m, p, off, len);
5810                         /*
5811                          * XXX
5812                          * IPv6 H/W cksum off-load not supported yet!
5813                          *
5814                          * if (sum == 0) {
5815                          *      m->m_pkthdr.csum_flags |=
5816                          *          (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
5817                          *      m->m_pkthdr.csum_data = 0xffff;
5818                          *}
5819                          */
5820                         break;
5821 #endif /* INET6 */
5822                 default:
5823                         return (1);
5824                 }
5825         }
5826         if (sum) {
5827                 switch (p) {
5828                 case IPPROTO_TCP:
5829                         tcpstat.tcps_rcvbadsum++;
5830                         break;
5831                 case IPPROTO_UDP:
5832                         udp_stat.udps_badsum++;
5833                         break;
5834                 case IPPROTO_ICMP:
5835                         icmpstat.icps_checksum++;
5836                         break;
5837 #ifdef INET6
5838                 case IPPROTO_ICMPV6:
5839                         icmp6stat.icp6s_checksum++;
5840                         break;
5841 #endif /* INET6 */
5842                 }
5843                 return (1);
5844         }
5845         return (0);
5846 }
5847
5848 struct pf_divert *
5849 pf_find_divert(struct mbuf *m)
5850 {
5851         struct m_tag    *mtag;
5852
5853         if ((mtag = m_tag_find(m, PACKET_TAG_PF_DIVERT, NULL)) == NULL)
5854                 return (NULL);
5855
5856         return ((struct pf_divert *)(mtag + 1));
5857 }
5858
5859 struct pf_divert *
5860 pf_get_divert(struct mbuf *m)
5861 {
5862         struct m_tag    *mtag;
5863
5864         if ((mtag = m_tag_find(m, PACKET_TAG_PF_DIVERT, NULL)) == NULL) {
5865                 mtag = m_tag_get(PACKET_TAG_PF_DIVERT, sizeof(struct pf_divert),
5866                     M_NOWAIT);
5867                 if (mtag == NULL)
5868                         return (NULL);
5869                 bzero(mtag + 1, sizeof(struct pf_divert));
5870                 m_tag_prepend(m, mtag);
5871         }
5872
5873         return ((struct pf_divert *)(mtag + 1));
5874 }
5875
5876 #ifdef INET
5877 int
5878 pf_test(int dir, struct ifnet *ifp, struct mbuf **m0,
5879     struct ether_header *eh, struct inpcb *inp)
5880 {
5881         struct pfi_kif          *kif;
5882         u_short                  action, reason = 0, log = 0;
5883         struct mbuf             *m = *m0;
5884         struct ip               *h = NULL;
5885         struct pf_rule          *a = NULL, *r = &pf_default_rule, *tr, *nr;
5886         struct pf_state         *s = NULL;
5887         struct pf_ruleset       *ruleset = NULL;
5888         struct pf_pdesc          pd;
5889         int                      off, dirndx;
5890 #ifdef ALTQ
5891         int                      pqid = 0;
5892 #endif
5893
5894         if (!pf_status.running)
5895                 return (PF_PASS);
5896
5897         memset(&pd, 0, sizeof(pd));
5898 #ifdef foo
5899         if (ifp->if_type == IFT_CARP && ifp->if_carpdev)
5900                 kif = (struct pfi_kif *)ifp->if_carpdev->if_pf_kif;
5901         else
5902 #endif
5903                 kif = (struct pfi_kif *)ifp->if_pf_kif;
5904
5905         if (kif == NULL) {
5906                 DPFPRINTF(PF_DEBUG_URGENT,
5907                     ("pf_test: kif == NULL, if_xname %s\n", ifp->if_xname));
5908                 return (PF_DROP);
5909         }
5910         if (kif->pfik_flags & PFI_IFLAG_SKIP)
5911                 return (PF_PASS);
5912
5913 #ifdef DIAGNOSTIC
5914         if ((m->m_flags & M_PKTHDR) == 0)
5915                 panic("non-M_PKTHDR is passed to pf_test");
5916 #endif /* DIAGNOSTIC */
5917
5918         if (m->m_pkthdr.len < (int)sizeof(*h)) {
5919                 action = PF_DROP;
5920                 REASON_SET(&reason, PFRES_SHORT);
5921                 log = 1;
5922                 goto done;
5923         }
5924
5925         /*
5926          * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags,
5927          * so make sure pf.flags is clear.
5928          */
5929         if (m->m_pkthdr.fw_flags & PF_MBUF_TAGGED)
5930                 return (PF_PASS);
5931         m->m_pkthdr.pf.flags = 0;
5932         /* Re-Check when updating to > 4.4 */
5933         m->m_pkthdr.pf.statekey = NULL;
5934
5935         /* We do IP header normalization and packet reassembly here */
5936         if (pf_normalize_ip(m0, dir, kif, &reason, &pd) != PF_PASS) {
5937                 action = PF_DROP;
5938                 goto done;
5939         }
5940         m = *m0;        /* pf_normalize messes with m0 */
5941         h = mtod(m, struct ip *);
5942
5943         off = h->ip_hl << 2;
5944         if (off < (int)sizeof(*h)) {
5945                 action = PF_DROP;
5946                 REASON_SET(&reason, PFRES_SHORT);
5947                 log = 1;
5948                 goto done;
5949         }
5950
5951         pd.src = (struct pf_addr *)&h->ip_src;
5952         pd.dst = (struct pf_addr *)&h->ip_dst;
5953         pd.sport = pd.dport = NULL;
5954         pd.ip_sum = &h->ip_sum;
5955         pd.proto_sum = NULL;
5956         pd.proto = h->ip_p;
5957         pd.dir = dir;
5958         pd.sidx = (dir == PF_IN) ? 0 : 1;
5959         pd.didx = (dir == PF_IN) ? 1 : 0;
5960         pd.af = AF_INET;
5961         pd.tos = h->ip_tos;
5962         pd.tot_len = h->ip_len;
5963         pd.eh = eh;
5964
5965         /* handle fragments that didn't get reassembled by normalization */
5966         if (h->ip_off & (IP_MF | IP_OFFMASK)) {
5967                 action = pf_test_fragment(&r, dir, kif, m, h,
5968                     &pd, &a, &ruleset);
5969                 goto done;
5970         }
5971
5972         switch (h->ip_p) {
5973
5974         case IPPROTO_TCP: {
5975                 struct tcphdr   th;
5976
5977                 pd.hdr.tcp = &th;
5978                 if (!pf_pull_hdr(m, off, &th, sizeof(th),
5979                     &action, &reason, AF_INET)) {
5980                         log = action != PF_PASS;
5981                         goto done;
5982                 }
5983                 pd.p_len = pd.tot_len - off - (th.th_off << 2);
5984 #ifdef ALTQ
5985                 if ((th.th_flags & TH_ACK) && pd.p_len == 0)
5986                         pqid = 1;
5987 #endif
5988                 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
5989                 if (action == PF_DROP)
5990                         goto done;
5991                 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
5992                     &reason);
5993                 if (action == PF_PASS) {
5994                         pfsync_update_state(s);
5995                         r = s->rule.ptr;
5996                         a = s->anchor.ptr;
5997                         log = s->log;
5998                 } else if (s == NULL)
5999                         action = pf_test_rule(&r, &s, dir, kif,
6000                             m, off, h, &pd, &a, &ruleset, NULL, inp);
6001                 break;
6002         }
6003
6004         case IPPROTO_UDP: {
6005                 struct udphdr   uh;
6006
6007                 pd.hdr.udp = &uh;
6008                 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
6009                     &action, &reason, AF_INET)) {
6010                         log = action != PF_PASS;
6011                         goto done;
6012                 }
6013                 if (uh.uh_dport == 0 ||
6014                     ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
6015                     ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
6016                         action = PF_DROP;
6017                         REASON_SET(&reason, PFRES_SHORT);
6018                         goto done;
6019                 }
6020                 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
6021                 if (action == PF_PASS) {
6022                         pfsync_update_state(s);
6023                         r = s->rule.ptr;
6024                         a = s->anchor.ptr;
6025                         log = s->log;
6026                 } else if (s == NULL)
6027                         action = pf_test_rule(&r, &s, dir, kif,
6028                             m, off, h, &pd, &a, &ruleset, NULL, inp);
6029                 break;
6030         }
6031
6032         case IPPROTO_ICMP: {
6033                 struct icmp     ih;
6034
6035                 pd.hdr.icmp = &ih;
6036                 if (!pf_pull_hdr(m, off, &ih, ICMP_MINLEN,
6037                     &action, &reason, AF_INET)) {
6038                         log = action != PF_PASS;
6039                         goto done;
6040                 }
6041                 action = pf_test_state_icmp(&s, dir, kif, m, off, h, &pd,
6042                     &reason);
6043                 if (action == PF_PASS) {
6044                         pfsync_update_state(s);
6045                         r = s->rule.ptr;
6046                         a = s->anchor.ptr;
6047                         log = s->log;
6048                 } else if (s == NULL)
6049                         action = pf_test_rule(&r, &s, dir, kif,
6050                             m, off, h, &pd, &a, &ruleset, NULL, inp);
6051                 break;
6052         }
6053
6054         default:
6055                 action = pf_test_state_other(&s, dir, kif, m, &pd);
6056                 if (action == PF_PASS) {
6057                         pfsync_update_state(s);
6058                         r = s->rule.ptr;
6059                         a = s->anchor.ptr;
6060                         log = s->log;
6061                 } else if (s == NULL)
6062                         action = pf_test_rule(&r, &s, dir, kif, m, off, h,
6063                             &pd, &a, &ruleset, NULL, inp);
6064                 break;
6065         }
6066
6067 done:
6068         if (action == PF_PASS && h->ip_hl > 5 &&
6069             !((s && s->state_flags & PFSTATE_ALLOWOPTS) || r->allow_opts)) {
6070                 action = PF_DROP;
6071                 REASON_SET(&reason, PFRES_IPOPTIONS);
6072                 log = 1;
6073                 DPFPRINTF(PF_DEBUG_MISC,
6074                     ("pf: dropping packet with ip options\n"));
6075         }
6076
6077         if ((s && s->tag) || r->rtableid)
6078                 pf_tag_packet(m, s ? s->tag : 0, r->rtableid);
6079
6080 #if 0
6081         if (dir == PF_IN && s && s->key[PF_SK_STACK])
6082                 m->m_pkthdr.pf.statekey = s->key[PF_SK_STACK];
6083 #endif
6084
6085 #ifdef ALTQ
6086         if (action == PF_PASS && r->qid) {
6087                 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
6088                 if (pqid || (pd.tos & IPTOS_LOWDELAY))
6089                         m->m_pkthdr.pf.qid = r->pqid;
6090                 else
6091                         m->m_pkthdr.pf.qid = r->qid;
6092                 m->m_pkthdr.pf.ecn_af = AF_INET;
6093                 m->m_pkthdr.pf.hdr = h;
6094                 /* add connection hash for fairq */
6095                 if (s) {
6096                         /* for fairq */
6097                         m->m_pkthdr.pf.state_hash = s->hash;
6098                         m->m_pkthdr.pf.flags |= PF_TAG_STATE_HASHED;
6099                 }
6100         }
6101 #endif /* ALTQ */
6102
6103         /*
6104          * connections redirected to loopback should not match sockets
6105          * bound specifically to loopback due to security implications,
6106          * see tcp_input() and in_pcblookup_listen().
6107          */
6108         if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
6109             pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
6110             (s->nat_rule.ptr->action == PF_RDR ||
6111             s->nat_rule.ptr->action == PF_BINAT) &&
6112             (ntohl(pd.dst->v4.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET)
6113                 m->m_pkthdr.pf.flags |= PF_TAG_TRANSLATE_LOCALHOST;
6114
6115         if (dir == PF_IN && action == PF_PASS && r->divert.port) {
6116                 struct pf_divert *divert;
6117
6118                 if ((divert = pf_get_divert(m))) {
6119                         m->m_pkthdr.pf.flags |= PF_TAG_DIVERTED;
6120                         divert->port = r->divert.port;
6121                         divert->addr.ipv4 = r->divert.addr.v4;
6122                 }
6123         }
6124
6125         if (log) {
6126                 struct pf_rule *lr;
6127
6128                 if (s != NULL && s->nat_rule.ptr != NULL &&
6129                     s->nat_rule.ptr->log & PF_LOG_ALL)
6130                         lr = s->nat_rule.ptr;
6131                 else
6132                         lr = r;
6133                 PFLOG_PACKET(kif, h, m, AF_INET, dir, reason, lr, a, ruleset,
6134                     &pd);
6135         }
6136
6137         kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
6138         kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++;
6139
6140         if (action == PF_PASS || r->action == PF_DROP) {
6141                 dirndx = (dir == PF_OUT);
6142                 r->packets[dirndx]++;
6143                 r->bytes[dirndx] += pd.tot_len;
6144                 if (a != NULL) {
6145                         a->packets[dirndx]++;
6146                         a->bytes[dirndx] += pd.tot_len;
6147                 }
6148                 if (s != NULL) {
6149                         if (s->nat_rule.ptr != NULL) {
6150                                 s->nat_rule.ptr->packets[dirndx]++;
6151                                 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len;
6152                         }
6153                         if (s->src_node != NULL) {
6154                                 s->src_node->packets[dirndx]++;
6155                                 s->src_node->bytes[dirndx] += pd.tot_len;
6156                         }
6157                         if (s->nat_src_node != NULL) {
6158                                 s->nat_src_node->packets[dirndx]++;
6159                                 s->nat_src_node->bytes[dirndx] += pd.tot_len;
6160                         }
6161                         dirndx = (dir == s->direction) ? 0 : 1;
6162                         s->packets[dirndx]++;
6163                         s->bytes[dirndx] += pd.tot_len;
6164                 }
6165                 tr = r;
6166                 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
6167                 if (nr != NULL && r == &pf_default_rule)
6168                         tr = nr;
6169                 if (tr->src.addr.type == PF_ADDR_TABLE)
6170                         pfr_update_stats(tr->src.addr.p.tbl,
6171                             (s == NULL) ? pd.src :
6172                             &s->key[(s->direction == PF_IN)]->
6173                                 addr[(s->direction == PF_OUT)],
6174                             pd.af, pd.tot_len, dir == PF_OUT,
6175                             r->action == PF_PASS, tr->src.neg);
6176                 if (tr->dst.addr.type == PF_ADDR_TABLE)
6177                         pfr_update_stats(tr->dst.addr.p.tbl,
6178                             (s == NULL) ? pd.dst :
6179                             &s->key[(s->direction == PF_IN)]->
6180                                 addr[(s->direction == PF_IN)],
6181                             pd.af, pd.tot_len, dir == PF_OUT,
6182                             r->action == PF_PASS, tr->dst.neg);
6183         }
6184
6185
6186         if (action == PF_SYNPROXY_DROP) {
6187                 m_freem(*m0);
6188                 *m0 = NULL;
6189                 action = PF_PASS;
6190         } else if (r->rt)
6191                 /* pf_route can free the mbuf causing *m0 to become NULL */
6192                 pf_route(m0, r, dir, kif->pfik_ifp, s, &pd);
6193
6194         return (action);
6195 }
6196 #endif /* INET */
6197
6198 #ifdef INET6
6199 int
6200 pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0,
6201     struct ether_header *eh, struct inpcb *inp)
6202 {
6203         struct pfi_kif          *kif;
6204         u_short                  action, reason = 0, log = 0;
6205         struct mbuf             *m = *m0, *n = NULL;
6206         struct ip6_hdr          *h = NULL;
6207         struct pf_rule          *a = NULL, *r = &pf_default_rule, *tr, *nr;
6208         struct pf_state         *s = NULL;
6209         struct pf_ruleset       *ruleset = NULL;
6210         struct pf_pdesc          pd;
6211         int                      off, terminal = 0, dirndx, rh_cnt = 0;
6212
6213         if (!pf_status.running)
6214                 return (PF_PASS);
6215
6216         memset(&pd, 0, sizeof(pd));
6217 #ifdef foo
6218         if (ifp->if_type == IFT_CARP && ifp->if_carpdev)
6219                 kif = (struct pfi_kif *)ifp->if_carpdev->if_pf_kif;
6220         else
6221 #endif
6222                 kif = (struct pfi_kif *)ifp->if_pf_kif;
6223
6224         if (kif == NULL) {
6225                 DPFPRINTF(PF_DEBUG_URGENT,
6226                     ("pf_test6: kif == NULL, if_xname %s\n", ifp->if_xname));
6227                 return (PF_DROP);
6228         }
6229         if (kif->pfik_flags & PFI_IFLAG_SKIP)
6230                 return (PF_PASS);
6231
6232 #ifdef DIAGNOSTIC
6233         if ((m->m_flags & M_PKTHDR) == 0)
6234                 panic("non-M_PKTHDR is passed to pf_test6");
6235 #endif /* DIAGNOSTIC */
6236
6237         if (m->m_pkthdr.len < (int)sizeof(*h)) {
6238                 action = PF_DROP;
6239                 REASON_SET(&reason, PFRES_SHORT);
6240                 log = 1;
6241                 goto done;
6242         }
6243
6244         /*
6245          * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags,
6246          * so make sure pf.flags is clear.
6247          */
6248         if (m->m_pkthdr.fw_flags & PF_MBUF_TAGGED)
6249                 return (PF_PASS);
6250         m->m_pkthdr.pf.flags = 0;
6251         /* Re-Check when updating to > 4.4 */
6252         m->m_pkthdr.pf.statekey = NULL;
6253
6254         /* We do IP header normalization and packet reassembly here */
6255         if (pf_normalize_ip6(m0, dir, kif, &reason, &pd) != PF_PASS) {
6256                 action = PF_DROP;
6257                 goto done;
6258         }
6259         m = *m0;        /* pf_normalize messes with m0 */
6260         h = mtod(m, struct ip6_hdr *);
6261
6262 #if 1
6263         /*
6264          * we do not support jumbogram yet.  if we keep going, zero ip6_plen
6265          * will do something bad, so drop the packet for now.
6266          */
6267         if (htons(h->ip6_plen) == 0) {
6268                 action = PF_DROP;
6269                 REASON_SET(&reason, PFRES_NORM);        /*XXX*/
6270                 goto done;
6271         }
6272 #endif
6273
6274         pd.src = (struct pf_addr *)&h->ip6_src;
6275         pd.dst = (struct pf_addr *)&h->ip6_dst;
6276         pd.sport = pd.dport = NULL;
6277         pd.ip_sum = NULL;
6278         pd.proto_sum = NULL;
6279         pd.dir = dir;
6280         pd.sidx = (dir == PF_IN) ? 0 : 1;
6281         pd.didx = (dir == PF_IN) ? 1 : 0;
6282         pd.af = AF_INET6;
6283         pd.tos = 0;
6284         pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr);
6285         pd.eh = eh;
6286
6287         off = ((caddr_t)h - m->m_data) + sizeof(struct ip6_hdr);
6288         pd.proto = h->ip6_nxt;
6289         do {
6290                 switch (pd.proto) {
6291                 case IPPROTO_FRAGMENT:
6292                         action = pf_test_fragment(&r, dir, kif, m, h,
6293                             &pd, &a, &ruleset);
6294                         if (action == PF_DROP)
6295                                 REASON_SET(&reason, PFRES_FRAG);
6296                         goto done;
6297                 case IPPROTO_ROUTING: {
6298                         struct ip6_rthdr rthdr;
6299
6300                         if (rh_cnt++) {
6301                                 DPFPRINTF(PF_DEBUG_MISC,
6302                                     ("pf: IPv6 more than one rthdr\n"));
6303                                 action = PF_DROP;
6304                                 REASON_SET(&reason, PFRES_IPOPTIONS);
6305                                 log = 1;
6306                                 goto done;
6307                         }
6308                         if (!pf_pull_hdr(m, off, &rthdr, sizeof(rthdr), NULL,
6309                             &reason, pd.af)) {
6310                                 DPFPRINTF(PF_DEBUG_MISC,
6311                                     ("pf: IPv6 short rthdr\n"));
6312                                 action = PF_DROP;
6313                                 REASON_SET(&reason, PFRES_SHORT);
6314                                 log = 1;
6315                                 goto done;
6316                         }
6317                         if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) {
6318                                 DPFPRINTF(PF_DEBUG_MISC,
6319                                     ("pf: IPv6 rthdr0\n"));
6320                                 action = PF_DROP;
6321                                 REASON_SET(&reason, PFRES_IPOPTIONS);
6322                                 log = 1;
6323                                 goto done;
6324                         }
6325                         /* FALLTHROUGH */
6326                 }
6327                 case IPPROTO_AH:
6328                 case IPPROTO_HOPOPTS:
6329                 case IPPROTO_DSTOPTS: {
6330                         /* get next header and header length */
6331                         struct ip6_ext  opt6;
6332
6333                         if (!pf_pull_hdr(m, off, &opt6, sizeof(opt6),
6334                             NULL, &reason, pd.af)) {
6335                                 DPFPRINTF(PF_DEBUG_MISC,
6336                                     ("pf: IPv6 short opt\n"));
6337                                 action = PF_DROP;
6338                                 log = 1;
6339                                 goto done;
6340                         }
6341                         if (pd.proto == IPPROTO_AH)
6342                                 off += (opt6.ip6e_len + 2) * 4;
6343                         else
6344                                 off += (opt6.ip6e_len + 1) * 8;
6345                         pd.proto = opt6.ip6e_nxt;
6346                         /* goto the next header */
6347                         break;
6348                 }
6349                 default:
6350                         terminal++;
6351                         break;
6352                 }
6353         } while (!terminal);
6354
6355         /* if there's no routing header, use unmodified mbuf for checksumming */
6356         if (!n)
6357                 n = m;
6358
6359         switch (pd.proto) {
6360
6361         case IPPROTO_TCP: {
6362                 struct tcphdr   th;
6363
6364                 pd.hdr.tcp = &th;
6365                 if (!pf_pull_hdr(m, off, &th, sizeof(th),
6366                     &action, &reason, AF_INET6)) {
6367                         log = action != PF_PASS;
6368                         goto done;
6369                 }
6370                 pd.p_len = pd.tot_len - off - (th.th_off << 2);
6371                 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
6372                 if (action == PF_DROP)
6373                         goto done;
6374                 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
6375                     &reason);
6376                 if (action == PF_PASS) {
6377                         pfsync_update_state(s);
6378                         r = s->rule.ptr;
6379                         a = s->anchor.ptr;
6380                         log = s->log;
6381                 } else if (s == NULL)
6382                         action = pf_test_rule(&r, &s, dir, kif,
6383                             m, off, h, &pd, &a, &ruleset, NULL, inp);
6384                 break;
6385         }
6386
6387         case IPPROTO_UDP: {
6388                 struct udphdr   uh;
6389
6390                 pd.hdr.udp = &uh;
6391                 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
6392                     &action, &reason, AF_INET6)) {
6393                         log = action != PF_PASS;
6394                         goto done;
6395                 }
6396                 if (uh.uh_dport == 0 ||
6397                     ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
6398                     ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
6399                         action = PF_DROP;
6400                         REASON_SET(&reason, PFRES_SHORT);
6401                         goto done;
6402                 }
6403                 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
6404                 if (action == PF_PASS) {
6405                         pfsync_update_state(s);
6406                         r = s->rule.ptr;
6407                         a = s->anchor.ptr;
6408                         log = s->log;
6409                 } else if (s == NULL)
6410                         action = pf_test_rule(&r, &s, dir, kif,
6411                             m, off, h, &pd, &a, &ruleset, NULL, inp);
6412                 break;
6413         }
6414
6415         case IPPROTO_ICMPV6: {
6416                 struct icmp6_hdr        ih;
6417
6418                 pd.hdr.icmp6 = &ih;
6419                 if (!pf_pull_hdr(m, off, &ih, sizeof(ih),
6420                     &action, &reason, AF_INET6)) {
6421                         log = action != PF_PASS;
6422                         goto done;
6423                 }
6424                 action = pf_test_state_icmp(&s, dir, kif,
6425                     m, off, h, &pd, &reason);
6426                 if (action == PF_PASS) {
6427                         pfsync_update_state(s);
6428                         r = s->rule.ptr;
6429                         a = s->anchor.ptr;
6430                         log = s->log;
6431                 } else if (s == NULL)
6432                         action = pf_test_rule(&r, &s, dir, kif,
6433                             m, off, h, &pd, &a, &ruleset, NULL, inp);
6434                 break;
6435         }
6436
6437         default:
6438                 action = pf_test_state_other(&s, dir, kif, m, &pd);
6439                 if (action == PF_PASS) {
6440                         pfsync_update_state(s);
6441                         r = s->rule.ptr;
6442                         a = s->anchor.ptr;
6443                         log = s->log;
6444                 } else if (s == NULL)
6445                         action = pf_test_rule(&r, &s, dir, kif, m, off, h,
6446                             &pd, &a, &ruleset, NULL, inp);
6447                 break;
6448         }
6449
6450 done:
6451         if (n != m) {
6452                 m_freem(n);
6453                 n = NULL;
6454         }
6455
6456         /* handle dangerous IPv6 extension headers. */
6457         if (action == PF_PASS && rh_cnt &&
6458             !((s && s->state_flags & PFSTATE_ALLOWOPTS) || r->allow_opts)) {
6459                 action = PF_DROP;
6460                 REASON_SET(&reason, PFRES_IPOPTIONS);
6461                 log = 1;
6462                 DPFPRINTF(PF_DEBUG_MISC,
6463                     ("pf: dropping packet with dangerous v6 headers\n"));
6464         }
6465
6466         if ((s && s->tag) || r->rtableid)
6467                 pf_tag_packet(m, s ? s->tag : 0, r->rtableid);
6468
6469 #if 0
6470         if (dir == PF_IN && s && s->key[PF_SK_STACK])
6471                 m->m_pkthdr.pf.statekey = s->key[PF_SK_STACK];
6472 #endif
6473
6474 #ifdef ALTQ
6475         if (action == PF_PASS && r->qid) {
6476                 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
6477                 if (pd.tos & IPTOS_LOWDELAY)
6478                         m->m_pkthdr.pf.qid = r->pqid;
6479                 else
6480                         m->m_pkthdr.pf.qid = r->qid;
6481                 m->m_pkthdr.pf.ecn_af = AF_INET6;
6482                 m->m_pkthdr.pf.hdr = h;
6483                 if (s) {
6484                         /* for fairq */
6485                         m->m_pkthdr.pf.state_hash = s->hash;
6486                         m->m_pkthdr.pf.flags |= PF_TAG_STATE_HASHED;
6487                 }
6488         }
6489 #endif /* ALTQ */
6490
6491         if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
6492             pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
6493             (s->nat_rule.ptr->action == PF_RDR ||
6494             s->nat_rule.ptr->action == PF_BINAT) &&
6495             IN6_IS_ADDR_LOOPBACK(&pd.dst->v6))
6496                 m->m_pkthdr.pf.flags |= PF_TAG_TRANSLATE_LOCALHOST;
6497
6498         if (dir == PF_IN && action == PF_PASS && r->divert.port) {
6499                 struct pf_divert *divert;
6500
6501                 if ((divert = pf_get_divert(m))) {
6502                         m->m_pkthdr.pf.flags |= PF_TAG_DIVERTED;
6503                         divert->port = r->divert.port;
6504                         divert->addr.ipv6 = r->divert.addr.v6;
6505                 }
6506         }
6507
6508         if (log) {
6509                 struct pf_rule *lr;
6510
6511                 if (s != NULL && s->nat_rule.ptr != NULL &&
6512                     s->nat_rule.ptr->log & PF_LOG_ALL)
6513                         lr = s->nat_rule.ptr;
6514                 else
6515                         lr = r;
6516                 PFLOG_PACKET(kif, h, m, AF_INET6, dir, reason, lr, a, ruleset,
6517                     &pd);
6518         }
6519
6520         kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
6521         kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++;
6522
6523         if (action == PF_PASS || r->action == PF_DROP) {
6524                 dirndx = (dir == PF_OUT);
6525                 r->packets[dirndx]++;
6526                 r->bytes[dirndx] += pd.tot_len;
6527                 if (a != NULL) {
6528                         a->packets[dirndx]++;
6529                         a->bytes[dirndx] += pd.tot_len;
6530                 }
6531                 if (s != NULL) {
6532                         if (s->nat_rule.ptr != NULL) {
6533                                 s->nat_rule.ptr->packets[dirndx]++;
6534                                 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len;
6535                         }
6536                         if (s->src_node != NULL) {
6537                                 s->src_node->packets[dirndx]++;
6538                                 s->src_node->bytes[dirndx] += pd.tot_len;
6539                         }
6540                         if (s->nat_src_node != NULL) {
6541                                 s->nat_src_node->packets[dirndx]++;
6542                                 s->nat_src_node->bytes[dirndx] += pd.tot_len;
6543                         }
6544                         dirndx = (dir == s->direction) ? 0 : 1;
6545                         s->packets[dirndx]++;
6546                         s->bytes[dirndx] += pd.tot_len;
6547                 }
6548                 tr = r;
6549                 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
6550                 if (nr != NULL && r == &pf_default_rule)
6551                         tr = nr;
6552                 if (tr->src.addr.type == PF_ADDR_TABLE)
6553                         pfr_update_stats(tr->src.addr.p.tbl,
6554                             (s == NULL) ? pd.src :
6555                             &s->key[(s->direction == PF_IN)]->addr[0],
6556                             pd.af, pd.tot_len, dir == PF_OUT,
6557                             r->action == PF_PASS, tr->src.neg);
6558                 if (tr->dst.addr.type == PF_ADDR_TABLE)
6559                         pfr_update_stats(tr->dst.addr.p.tbl,
6560                             (s == NULL) ? pd.dst :
6561                             &s->key[(s->direction == PF_IN)]->addr[1],
6562                             pd.af, pd.tot_len, dir == PF_OUT,
6563                             r->action == PF_PASS, tr->dst.neg);
6564         }
6565
6566
6567         if (action == PF_SYNPROXY_DROP) {
6568                 m_freem(*m0);
6569                 *m0 = NULL;
6570                 action = PF_PASS;
6571         } else if (r->rt)
6572                 /* pf_route6 can free the mbuf causing *m0 to become NULL */
6573                 pf_route6(m0, r, dir, kif->pfik_ifp, s, &pd);
6574
6575         return (action);
6576 }
6577 #endif /* INET6 */
6578
6579 int
6580 pf_check_congestion(struct ifqueue *ifq)
6581 {
6582                 return (0);
6583 }