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/netisr.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/in_pcb.h>
  91 #include <netinet/icmp6.h>
  92 #include <netinet6/nd6.h>
  93 #include <netinet6/ip6_var.h>
  94 #include <netinet6/in6_pcb.h>
  95 #endif /* INET6 */
  96
  97 #include <sys/in_cksum.h>
  98 #include <sys/ucred.h>
  99 #include <machine/limits.h>
 100 #include <sys/msgport2.h>
 101 #include <net/netmsg2.h>
 102
 103 extern int ip_optcopy(struct ip *, struct ip *);
 104 extern int debug_pfugidhack;
 105
 106 struct lwkt_token pf_token = LWKT_TOKEN_INITIALIZER(pf_token);
 107
 108 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) kprintf x
 109
 110 /*
 111  * Global variables
 112  */
 113
 114 /* mask radix tree */
 115 struct radix_node_head  *pf_maskhead;
 116
 117 /* state tables */
 118 struct pf_state_tree     pf_statetbl;
 119
 120 struct pf_altqqueue      pf_altqs[2];
 121 struct pf_palist         pf_pabuf;
 122 struct pf_altqqueue     *pf_altqs_active;
 123 struct pf_altqqueue     *pf_altqs_inactive;
 124 struct pf_status         pf_status;
 125
 126 u_int32_t                ticket_altqs_active;
 127 u_int32_t                ticket_altqs_inactive;
 128 int                      altqs_inactive_open;
 129 u_int32_t                ticket_pabuf;
 130
 131 MD5_CTX                  pf_tcp_secret_ctx;
 132 u_char                   pf_tcp_secret[16];
 133 int                      pf_tcp_secret_init;
 134 int                      pf_tcp_iss_off;
 135
 136 struct pf_anchor_stackframe {
 137         struct pf_ruleset                       *rs;
 138         struct pf_rule                          *r;
 139         struct pf_anchor_node                   *parent;
 140         struct pf_anchor                        *child;
 141 } pf_anchor_stack[64];
 142
 143 struct malloc_type       *pf_src_tree_pl, *pf_rule_pl, *pf_pooladdr_pl;
 144 struct malloc_type       *pf_state_pl, *pf_state_key_pl, *pf_state_item_pl;
 145 struct malloc_type       *pf_altq_pl;
 146
 147 void                     pf_print_host(struct pf_addr *, u_int16_t, u_int8_t);
 148
 149 void                     pf_init_threshold(struct pf_threshold *, u_int32_t,
 150                             u_int32_t);
 151 void                     pf_add_threshold(struct pf_threshold *);
 152 int                      pf_check_threshold(struct pf_threshold *);
 153
 154 void                     pf_change_ap(struct pf_addr *, u_int16_t *,
 155                             u_int16_t *, u_int16_t *, struct pf_addr *,
 156                             u_int16_t, u_int8_t, sa_family_t);
 157 int                      pf_modulate_sack(struct mbuf *, int, struct pf_pdesc *,
 158                             struct tcphdr *, struct pf_state_peer *);
 159 #ifdef INET6
 160 void                     pf_change_a6(struct pf_addr *, u_int16_t *,
 161                             struct pf_addr *, u_int8_t);
 162 #endif /* INET6 */
 163 void                     pf_change_icmp(struct pf_addr *, u_int16_t *,
 164                             struct pf_addr *, struct pf_addr *, u_int16_t,
 165                             u_int16_t *, u_int16_t *, u_int16_t *,
 166                             u_int16_t *, u_int8_t, sa_family_t);
 167 void                     pf_send_tcp(const struct pf_rule *, sa_family_t,
 168                             const struct pf_addr *, const struct pf_addr *,
 169                             u_int16_t, u_int16_t, u_int32_t, u_int32_t,
 170                             u_int8_t, u_int16_t, u_int16_t, u_int8_t, int,
 171                             u_int16_t, struct ether_header *, struct ifnet *);
 172 void                     pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t,
 173                             sa_family_t, struct pf_rule *);
 174 struct pf_rule          *pf_match_translation(struct pf_pdesc *, struct mbuf *,
 175                             int, int, struct pfi_kif *,
 176                             struct pf_addr *, u_int16_t, struct pf_addr *,
 177                             u_int16_t, int);
 178 struct pf_rule          *pf_get_translation(struct pf_pdesc *, struct mbuf *,
 179                             int, int, struct pfi_kif *, struct pf_src_node **,
 180                             struct pf_state_key **, struct pf_state_key **,
 181                             struct pf_state_key **, struct pf_state_key **,
 182                             struct pf_addr *, struct pf_addr *,
 183                             u_int16_t, u_int16_t);
 184 void                     pf_detach_state(struct pf_state *);
 185 int                      pf_state_key_setup(struct pf_pdesc *, struct pf_rule *,
 186                             struct pf_state_key **, struct pf_state_key **,
 187                             struct pf_state_key **, struct pf_state_key **,
 188                             struct pf_addr *, struct pf_addr *,
 189                             u_int16_t, u_int16_t);
 190 void                     pf_state_key_detach(struct pf_state *, int);
 191 u_int32_t                pf_tcp_iss(struct pf_pdesc *);
 192 int                      pf_test_rule(struct pf_rule **, struct pf_state **,
 193                             int, struct pfi_kif *, struct mbuf *, int,
 194                             void *, struct pf_pdesc *, struct pf_rule **,
 195                             struct pf_ruleset **, struct ifqueue *, struct inpcb *);
 196 static __inline int      pf_create_state(struct pf_rule *, struct pf_rule *,
 197                             struct pf_rule *, struct pf_pdesc *,
 198                             struct pf_src_node *, struct pf_state_key *,
 199                             struct pf_state_key *, struct pf_state_key *,
 200                             struct pf_state_key *, struct mbuf *, int,
 201                             u_int16_t, u_int16_t, int *, struct pfi_kif *,
 202                             struct pf_state **, int, u_int16_t, u_int16_t,
 203                             int);
 204 int                      pf_test_fragment(struct pf_rule **, int,
 205                             struct pfi_kif *, struct mbuf *, void *,
 206                             struct pf_pdesc *, struct pf_rule **,
 207                             struct pf_ruleset **);
 208 int                      pf_tcp_track_full(struct pf_state_peer *,
 209                             struct pf_state_peer *, struct pf_state **,
 210                             struct pfi_kif *, struct mbuf *, int,
 211                             struct pf_pdesc *, u_short *, int *);
 212 int                     pf_tcp_track_sloppy(struct pf_state_peer *,
 213                             struct pf_state_peer *, struct pf_state **,
 214                             struct pf_pdesc *, u_short *);
 215 int                      pf_test_state_tcp(struct pf_state **, int,
 216                             struct pfi_kif *, struct mbuf *, int,
 217                             void *, struct pf_pdesc *, u_short *);
 218 int                      pf_test_state_udp(struct pf_state **, int,
 219                             struct pfi_kif *, struct mbuf *, int,
 220                             void *, struct pf_pdesc *);
 221 int                      pf_test_state_icmp(struct pf_state **, int,
 222                             struct pfi_kif *, struct mbuf *, int,
 223                             void *, struct pf_pdesc *, u_short *);
 224 int                      pf_test_state_other(struct pf_state **, int,
 225                             struct pfi_kif *, struct mbuf *, struct pf_pdesc *);
 226 void                     pf_step_into_anchor(int *, struct pf_ruleset **, int,
 227                             struct pf_rule **, struct pf_rule **, int *);
 228 int                      pf_step_out_of_anchor(int *, struct pf_ruleset **,
 229                              int, struct pf_rule **, struct pf_rule **,
 230                              int *);
 231 void                     pf_hash(struct pf_addr *, struct pf_addr *,
 232                             struct pf_poolhashkey *, sa_family_t);
 233 int                      pf_map_addr(u_int8_t, struct pf_rule *,
 234                             struct pf_addr *, struct pf_addr *,
 235                             struct pf_addr *, struct pf_src_node **);
 236 int                      pf_get_sport(sa_family_t, u_int8_t, struct pf_rule *,
 237                             struct pf_addr *, struct pf_addr *, u_int16_t,
 238                             struct pf_addr *, u_int16_t*, u_int16_t, u_int16_t,
 239                             struct pf_src_node **);
 240 void                     pf_route(struct mbuf **, struct pf_rule *, int,
 241                             struct ifnet *, struct pf_state *,
 242                             struct pf_pdesc *);
 243 void                     pf_route6(struct mbuf **, struct pf_rule *, int,
 244                             struct ifnet *, struct pf_state *,
 245                             struct pf_pdesc *);
 246 u_int8_t                 pf_get_wscale(struct mbuf *, int, u_int16_t,
 247                             sa_family_t);
 248 u_int16_t                pf_get_mss(struct mbuf *, int, u_int16_t,
 249                             sa_family_t);
 250 u_int16_t                pf_calc_mss(struct pf_addr *, sa_family_t,
 251                                 u_int16_t);
 252 void                     pf_set_rt_ifp(struct pf_state *,
 253                             struct pf_addr *);
 254 int                      pf_check_proto_cksum(struct mbuf *, int, int,
 255                             u_int8_t, sa_family_t);
 256 struct pf_divert        *pf_get_divert(struct mbuf *);
 257 void                     pf_print_state_parts(struct pf_state *,
 258                             struct pf_state_key *, struct pf_state_key *);
 259 int                      pf_addr_wrap_neq(struct pf_addr_wrap *,
 260                             struct pf_addr_wrap *);
 261 struct pf_state         *pf_find_state(struct pfi_kif *,
 262                             struct pf_state_key_cmp *, u_int, struct mbuf *);
 263 int                      pf_src_connlimit(struct pf_state **);
 264 int                      pf_check_congestion(struct ifqueue *);
 265
 266 extern int pf_end_threads;
 267
 268 struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] = {
 269         { &pf_state_pl, PFSTATE_HIWAT },
 270         { &pf_src_tree_pl, PFSNODE_HIWAT },
 271         { &pf_frent_pl, PFFRAG_FRENT_HIWAT },
 272         { &pfr_ktable_pl, PFR_KTABLE_HIWAT },
 273         { &pfr_kentry_pl, PFR_KENTRY_HIWAT }
 274 };
 275
 276 #define STATE_LOOKUP(i, k, d, s, m)                                     \
 277         do {                                                            \
 278                 s = pf_find_state(i, k, d, m);                  \
 279                 if (s == NULL || (s)->timeout == PFTM_PURGE)            \
 280                         return (PF_DROP);                               \
 281                 if (d == PF_OUT &&                                      \
 282                     (((s)->rule.ptr->rt == PF_ROUTETO &&                \
 283                     (s)->rule.ptr->direction == PF_OUT) ||              \
 284                     ((s)->rule.ptr->rt == PF_REPLYTO &&                 \
 285                     (s)->rule.ptr->direction == PF_IN)) &&              \
 286                     (s)->rt_kif != NULL &&                              \
 287                     (s)->rt_kif != i)                                   \
 288                         return (PF_PASS);                               \
 289         } while (0)
 290
 291 #define BOUND_IFACE(r, k) \
 292         ((r)->rule_flag & PFRULE_IFBOUND) ? (k) : pfi_all
 293
 294 #define STATE_INC_COUNTERS(s)                           \
 295         do {                                            \
 296                 s->rule.ptr->states_cur++;              \
 297                 s->rule.ptr->states_tot++;              \
 298                 if (s->anchor.ptr != NULL) {            \
 299                         s->anchor.ptr->states_cur++;    \
 300                         s->anchor.ptr->states_tot++;    \
 301                 }                                       \
 302                 if (s->nat_rule.ptr != NULL) {          \
 303                         s->nat_rule.ptr->states_cur++;  \
 304                         s->nat_rule.ptr->states_tot++;  \
 305                 }                                       \
 306         } while (0)
 307
 308 #define STATE_DEC_COUNTERS(s)                           \
 309         do {                                            \
 310                 if (s->nat_rule.ptr != NULL)            \
 311                         s->nat_rule.ptr->states_cur--;  \
 312                 if (s->anchor.ptr != NULL)              \
 313                         s->anchor.ptr->states_cur--;    \
 314                 s->rule.ptr->states_cur--;              \
 315         } while (0)
 316
 317 static MALLOC_DEFINE(M_PFSTATEPL, "pfstatepl", "pf state pool list");
 318 static MALLOC_DEFINE(M_PFSRCTREEPL, "pfsrctpl", "pf source tree pool list");
 319 static MALLOC_DEFINE(M_PFSTATEKEYPL, "pfstatekeypl", "pf state key pool list");
 320 static MALLOC_DEFINE(M_PFSTATEITEMPL, "pfstateitempl", "pf state item pool list");
 321
 322 static __inline int pf_src_compare(struct pf_src_node *, struct pf_src_node *);
 323 static __inline int pf_state_compare_key(struct pf_state_key *,
 324         struct pf_state_key *);
 325 static __inline int pf_state_compare_id(struct pf_state *,
 326         struct pf_state *);
 327
 328 struct pf_src_tree tree_src_tracking;
 329
 330 struct pf_state_tree_id tree_id;
 331 struct pf_state_queue state_list;
 332
 333 RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare);
 334 RB_GENERATE(pf_state_tree, pf_state_key, entry, pf_state_compare_key);
 335 RB_GENERATE(pf_state_tree_id, pf_state,
 336     entry_id, pf_state_compare_id);
 337
 338 static __inline int
 339 pf_src_compare(struct pf_src_node *a, struct pf_src_node *b)
 340 {
 341         int     diff;
 342
 343         if (a->rule.ptr > b->rule.ptr)
 344                 return (1);
 345         if (a->rule.ptr < b->rule.ptr)
 346                 return (-1);
 347         if ((diff = a->af - b->af) != 0)
 348                 return (diff);
 349         switch (a->af) {
 350 #ifdef INET
 351         case AF_INET:
 352                 if (a->addr.addr32[0] > b->addr.addr32[0])
 353                         return (1);
 354                 if (a->addr.addr32[0] < b->addr.addr32[0])
 355                         return (-1);
 356                 break;
 357 #endif /* INET */
 358 #ifdef INET6
 359         case AF_INET6:
 360                 if (a->addr.addr32[3] > b->addr.addr32[3])
 361                         return (1);
 362                 if (a->addr.addr32[3] < b->addr.addr32[3])
 363                         return (-1);
 364                 if (a->addr.addr32[2] > b->addr.addr32[2])
 365                         return (1);
 366                 if (a->addr.addr32[2] < b->addr.addr32[2])
 367                         return (-1);
 368                 if (a->addr.addr32[1] > b->addr.addr32[1])
 369                         return (1);
 370                 if (a->addr.addr32[1] < b->addr.addr32[1])
 371                         return (-1);
 372                 if (a->addr.addr32[0] > b->addr.addr32[0])
 373                         return (1);
 374                 if (a->addr.addr32[0] < b->addr.addr32[0])
 375                         return (-1);
 376                 break;
 377 #endif /* INET6 */
 378         }
 379         return (0);
 380 }
 381
 382 u_int32_t
 383 pf_state_hash(struct pf_state_key *sk)
 384 {
 385         u_int32_t hv = (u_int32_t)(((intptr_t)sk >> 6) ^ ((intptr_t)sk >> 15));
 386         if (hv == 0)    /* disallow 0 */
 387                 hv = 1;
 388         return(hv);
 389 }
 390
 391 #ifdef INET6
 392 void
 393 pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af)
 394 {
 395         switch (af) {
 396 #ifdef INET
 397         case AF_INET:
 398                 dst->addr32[0] = src->addr32[0];
 399                 break;
 400 #endif /* INET */
 401         case AF_INET6:
 402                 dst->addr32[0] = src->addr32[0];
 403                 dst->addr32[1] = src->addr32[1];
 404                 dst->addr32[2] = src->addr32[2];
 405                 dst->addr32[3] = src->addr32[3];
 406                 break;
 407         }
 408 }
 409 #endif /* INET6 */
 410
 411 void
 412 pf_init_threshold(struct pf_threshold *threshold,
 413     u_int32_t limit, u_int32_t seconds)
 414 {
 415         threshold->limit = limit * PF_THRESHOLD_MULT;
 416         threshold->seconds = seconds;
 417         threshold->count = 0;
 418         threshold->last = time_second;
 419 }
 420
 421 void
 422 pf_add_threshold(struct pf_threshold *threshold)
 423 {
 424         u_int32_t t = time_second, diff = t - threshold->last;
 425
 426         if (diff >= threshold->seconds)
 427                 threshold->count = 0;
 428         else
 429                 threshold->count -= threshold->count * diff /
 430                     threshold->seconds;
 431         threshold->count += PF_THRESHOLD_MULT;
 432         threshold->last = t;
 433 }
 434
 435 int
 436 pf_check_threshold(struct pf_threshold *threshold)
 437 {
 438         return (threshold->count > threshold->limit);
 439 }
 440
 441 int
 442 pf_src_connlimit(struct pf_state **state)
 443 {
 444         int bad = 0;
 445
 446         (*state)->src_node->conn++;
 447         (*state)->src.tcp_est = 1;
 448         pf_add_threshold(&(*state)->src_node->conn_rate);
 449
 450         if ((*state)->rule.ptr->max_src_conn &&
 451             (*state)->rule.ptr->max_src_conn <
 452             (*state)->src_node->conn) {
 453                 pf_status.lcounters[LCNT_SRCCONN]++;
 454                 bad++;
 455         }
 456
 457         if ((*state)->rule.ptr->max_src_conn_rate.limit &&
 458             pf_check_threshold(&(*state)->src_node->conn_rate)) {
 459                 pf_status.lcounters[LCNT_SRCCONNRATE]++;
 460                 bad++;
 461         }
 462
 463         if (!bad)
 464                 return (0);
 465
 466         if ((*state)->rule.ptr->overload_tbl) {
 467                 struct pfr_addr p;
 468                 u_int32_t       killed = 0;
 469
 470                 pf_status.lcounters[LCNT_OVERLOAD_TABLE]++;
 471                 if (pf_status.debug >= PF_DEBUG_MISC) {
 472                         kprintf("pf_src_connlimit: blocking address ");
 473                         pf_print_host(&(*state)->src_node->addr, 0,
 474                             (*state)->key[PF_SK_WIRE]->af);
 475                 }
 476
 477                 bzero(&p, sizeof(p));
 478                 p.pfra_af = (*state)->key[PF_SK_WIRE]->af;
 479                 switch ((*state)->key[PF_SK_WIRE]->af) {
 480 #ifdef INET
 481                 case AF_INET:
 482                         p.pfra_net = 32;
 483                         p.pfra_ip4addr = (*state)->src_node->addr.v4;
 484                         break;
 485 #endif /* INET */
 486 #ifdef INET6
 487                 case AF_INET6:
 488                         p.pfra_net = 128;
 489                         p.pfra_ip6addr = (*state)->src_node->addr.v6;
 490                         break;
 491 #endif /* INET6 */
 492                 }
 493
 494                 pfr_insert_kentry((*state)->rule.ptr->overload_tbl,
 495                     &p, time_second);
 496
 497                 /* kill existing states if that's required. */
 498                 if ((*state)->rule.ptr->flush) {
 499                         struct pf_state_key *sk;
 500                         struct pf_state *st;
 501
 502                         pf_status.lcounters[LCNT_OVERLOAD_FLUSH]++;
 503                         RB_FOREACH(st, pf_state_tree_id, &tree_id) {
 504                                 sk = st->key[PF_SK_WIRE];
 505                                 /*
 506                                  * Kill states from this source.  (Only those
 507                                  * from the same rule if PF_FLUSH_GLOBAL is not
 508                                  * set)
 509                                  */
 510                                 if (sk->af ==
 511                                     (*state)->key[PF_SK_WIRE]->af &&
 512                                     (((*state)->direction == PF_OUT &&
 513                                     PF_AEQ(&(*state)->src_node->addr,
 514                                         &sk->addr[0], sk->af)) ||
 515                                     ((*state)->direction == PF_IN &&
 516                                     PF_AEQ(&(*state)->src_node->addr,
 517                                         &sk->addr[1], sk->af))) &&
 518                                     ((*state)->rule.ptr->flush &
 519                                     PF_FLUSH_GLOBAL ||
 520                                     (*state)->rule.ptr == st->rule.ptr)) {
 521                                         st->timeout = PFTM_PURGE;
 522                                         st->src.state = st->dst.state =
 523                                             TCPS_CLOSED;
 524                                         killed++;
 525                                 }
 526                         }
 527                         if (pf_status.debug >= PF_DEBUG_MISC)
 528                                 kprintf(", %u states killed", killed);
 529                 }
 530                 if (pf_status.debug >= PF_DEBUG_MISC)
 531                         kprintf("\n");
 532         }
 533
 534         /* kill this state */
 535         (*state)->timeout = PFTM_PURGE;
 536         (*state)->src.state = (*state)->dst.state = TCPS_CLOSED;
 537         return (1);
 538 }
 539
 540 int
 541 pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule,
 542     struct pf_addr *src, sa_family_t af)
 543 {
 544         struct pf_src_node      k;
 545
 546         if (*sn == NULL) {
 547                 k.af = af;
 548                 PF_ACPY(&k.addr, src, af);
 549                 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
 550                     rule->rpool.opts & PF_POOL_STICKYADDR)
 551                         k.rule.ptr = rule;
 552                 else
 553                         k.rule.ptr = NULL;
 554                 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
 555                 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
 556         }
 557         if (*sn == NULL) {
 558                 if (!rule->max_src_nodes ||
 559                     rule->src_nodes < rule->max_src_nodes)
 560                         (*sn) = kmalloc(sizeof(struct pf_src_node), M_PFSRCTREEPL, M_NOWAIT|M_ZERO);
 561                 else
 562                         pf_status.lcounters[LCNT_SRCNODES]++;
 563                 if ((*sn) == NULL)
 564                         return (-1);
 565
 566                 pf_init_threshold(&(*sn)->conn_rate,
 567                     rule->max_src_conn_rate.limit,
 568                     rule->max_src_conn_rate.seconds);
 569
 570                 (*sn)->af = af;
 571                 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
 572                     rule->rpool.opts & PF_POOL_STICKYADDR)
 573                         (*sn)->rule.ptr = rule;
 574                 else
 575                         (*sn)->rule.ptr = NULL;
 576                 PF_ACPY(&(*sn)->addr, src, af);
 577                 if (RB_INSERT(pf_src_tree,
 578                     &tree_src_tracking, *sn) != NULL) {
 579                         if (pf_status.debug >= PF_DEBUG_MISC) {
 580                                 kprintf("pf: src_tree insert failed: ");
 581                                 pf_print_host(&(*sn)->addr, 0, af);
 582                                 kprintf("\n");
 583                         }
 584                         kfree(*sn, M_PFSRCTREEPL);
 585                         return (-1);
 586                 }
 587                 (*sn)->creation = time_second;
 588                 (*sn)->ruletype = rule->action;
 589                 if ((*sn)->rule.ptr != NULL)
 590                         (*sn)->rule.ptr->src_nodes++;
 591                 pf_status.scounters[SCNT_SRC_NODE_INSERT]++;
 592                 pf_status.src_nodes++;
 593         } else {
 594                 if (rule->max_src_states &&
 595                     (*sn)->states >= rule->max_src_states) {
 596                         pf_status.lcounters[LCNT_SRCSTATES]++;
 597                         return (-1);
 598                 }
 599         }
 600         return (0);
 601 }
 602
 603 /* state table stuff */
 604
 605 static __inline int
 606 pf_state_compare_key(struct pf_state_key *a, struct pf_state_key *b)
 607 {
 608         int     diff;
 609
 610         if ((diff = a->proto - b->proto) != 0)
 611                 return (diff);
 612         if ((diff = a->af - b->af) != 0)
 613                 return (diff);
 614         switch (a->af) {
 615 #ifdef INET
 616         case AF_INET:
 617                 if (a->addr[0].addr32[0] > b->addr[0].addr32[0])
 618                         return (1);
 619                 if (a->addr[0].addr32[0] < b->addr[0].addr32[0])
 620                         return (-1);
 621                 if (a->addr[1].addr32[0] > b->addr[1].addr32[0])
 622                         return (1);
 623                 if (a->addr[1].addr32[0] < b->addr[1].addr32[0])
 624                         return (-1);
 625                 break;
 626 #endif /* INET */
 627 #ifdef INET6
 628         case AF_INET6:
 629                 if (a->addr[0].addr32[3] > b->addr[0].addr32[3])
 630                         return (1);
 631                 if (a->addr[0].addr32[3] < b->addr[0].addr32[3])
 632                         return (-1);
 633                 if (a->addr[1].addr32[3] > b->addr[1].addr32[3])
 634                         return (1);
 635                 if (a->addr[1].addr32[3] < b->addr[1].addr32[3])
 636                         return (-1);
 637                 if (a->addr[0].addr32[2] > b->addr[0].addr32[2])
 638                         return (1);
 639                 if (a->addr[0].addr32[2] < b->addr[0].addr32[2])
 640                         return (-1);
 641                 if (a->addr[1].addr32[2] > b->addr[1].addr32[2])
 642                         return (1);
 643                 if (a->addr[1].addr32[2] < b->addr[1].addr32[2])
 644                         return (-1);
 645                 if (a->addr[0].addr32[1] > b->addr[0].addr32[1])
 646                         return (1);
 647                 if (a->addr[0].addr32[1] < b->addr[0].addr32[1])
 648                         return (-1);
 649                 if (a->addr[1].addr32[1] > b->addr[1].addr32[1])
 650                         return (1);
 651                 if (a->addr[1].addr32[1] < b->addr[1].addr32[1])
 652                         return (-1);
 653                 if (a->addr[0].addr32[0] > b->addr[0].addr32[0])
 654                         return (1);
 655                 if (a->addr[0].addr32[0] < b->addr[0].addr32[0])
 656                         return (-1);
 657                 if (a->addr[1].addr32[0] > b->addr[1].addr32[0])
 658                         return (1);
 659                 if (a->addr[1].addr32[0] < b->addr[1].addr32[0])
 660                         return (-1);
 661                 break;
 662 #endif /* INET6 */
 663         }
 664
 665         if ((diff = a->port[0] - b->port[0]) != 0)
 666                 return (diff);
 667         if ((diff = a->port[1] - b->port[1]) != 0)
 668                 return (diff);
 669
 670         return (0);
 671 }
 672
 673 static __inline int
 674 pf_state_compare_id(struct pf_state *a, struct pf_state *b)
 675 {
 676         if (a->id > b->id)
 677                 return (1);
 678         if (a->id < b->id)
 679                 return (-1);
 680         if (a->creatorid > b->creatorid)
 681                 return (1);
 682         if (a->creatorid < b->creatorid)
 683                 return (-1);
 684
 685         return (0);
 686 }
 687
 688 int
 689 pf_state_key_attach(struct pf_state_key *sk, struct pf_state *s, int idx)
 690 {
 691         struct pf_state_item    *si;
 692         struct pf_state_key     *cur;
 693
 694         KKASSERT(s->key[idx] == NULL);  /* XXX handle this? */
 695
 696         if ((cur = RB_INSERT(pf_state_tree, &pf_statetbl, sk)) != NULL) {
 697                 /* key exists. check for same kif, if none, add to key */
 698                 TAILQ_FOREACH(si, &cur->states, entry)
 699                         if (si->s->kif == s->kif &&
 700                             si->s->direction == s->direction) {
 701                                 if (pf_status.debug >= PF_DEBUG_MISC) {
 702                                         kprintf(
 703                                             "pf: %s key attach failed on %s: ",
 704                                             (idx == PF_SK_WIRE) ?
 705                                             "wire" : "stack",
 706                                             s->kif->pfik_name);
 707                                         pf_print_state_parts(s,
 708                                             (idx == PF_SK_WIRE) ? sk : NULL,
 709                                             (idx == PF_SK_STACK) ? sk : NULL);
 710                                         kprintf("\n");
 711                                 }
 712                                 kfree(sk, M_PFSTATEKEYPL);
 713                                 return (-1);    /* collision! */
 714                         }
 715                 kfree(sk, M_PFSTATEKEYPL);
 716
 717                 s->key[idx] = cur;
 718         } else
 719                 s->key[idx] = sk;
 720
 721         if ((si = kmalloc(sizeof(struct pf_state_item), M_PFSTATEITEMPL, M_NOWAIT)) == NULL) {
 722                 pf_state_key_detach(s, idx);
 723                 return (-1);
 724         }
 725         si->s = s;
 726
 727         /* list is sorted, if-bound states before floating */
 728         if (s->kif == pfi_all)
 729                 TAILQ_INSERT_TAIL(&s->key[idx]->states, si, entry);
 730         else
 731                 TAILQ_INSERT_HEAD(&s->key[idx]->states, si, entry);
 732         return (0);
 733 }
 734
 735 void
 736 pf_detach_state(struct pf_state *s)
 737 {
 738         if (s->key[PF_SK_WIRE] == s->key[PF_SK_STACK])
 739                 s->key[PF_SK_WIRE] = NULL;
 740
 741         if (s->key[PF_SK_STACK] != NULL)
 742                 pf_state_key_detach(s, PF_SK_STACK);
 743
 744         if (s->key[PF_SK_WIRE] != NULL)
 745                 pf_state_key_detach(s, PF_SK_WIRE);
 746 }
 747
 748 void
 749 pf_state_key_detach(struct pf_state *s, int idx)
 750 {
 751         struct pf_state_item    *si;
 752         si = TAILQ_FIRST(&s->key[idx]->states);
 753         while (si && si->s != s)
 754             si = TAILQ_NEXT(si, entry);
 755
 756         if (si) {
 757                 TAILQ_REMOVE(&s->key[idx]->states, si, entry);
 758                 kfree(si, M_PFSTATEITEMPL);
 759         }
 760
 761         if (TAILQ_EMPTY(&s->key[idx]->states)) {
 762                 RB_REMOVE(pf_state_tree, &pf_statetbl, s->key[idx]);
 763                 if (s->key[idx]->reverse)
 764                         s->key[idx]->reverse->reverse = NULL;
 765                 if (s->key[idx]->inp)
 766                         s->key[idx]->inp->inp_pf_sk = NULL;
 767                 kfree(s->key[idx], M_PFSTATEKEYPL);
 768         }
 769         s->key[idx] = NULL;
 770 }
 771
 772 struct pf_state_key *
 773 pf_alloc_state_key(int pool_flags)
 774 {
 775         struct pf_state_key     *sk;
 776
 777         if ((sk = kmalloc(sizeof(struct pf_state_key), M_PFSTATEKEYPL, pool_flags)) == NULL)
 778                         return (NULL);
 779         TAILQ_INIT(&sk->states);
 780
 781         return (sk);
 782 }
 783
 784 int
 785 pf_state_key_setup(struct pf_pdesc *pd, struct pf_rule *nr,
 786         struct pf_state_key **skw, struct pf_state_key **sks,
 787         struct pf_state_key **skp, struct pf_state_key **nkp,
 788         struct pf_addr *saddr, struct pf_addr *daddr,
 789         u_int16_t sport, u_int16_t dport)
 790 {
 791         KKASSERT((*skp == NULL && *nkp == NULL));
 792
 793         if ((*skp = pf_alloc_state_key(M_NOWAIT | M_ZERO)) == NULL)
 794                 return (ENOMEM);
 795
 796         PF_ACPY(&(*skp)->addr[pd->sidx], saddr, pd->af);
 797         PF_ACPY(&(*skp)->addr[pd->didx], daddr, pd->af);
 798         (*skp)->port[pd->sidx] = sport;
 799         (*skp)->port[pd->didx] = dport;
 800         (*skp)->proto = pd->proto;
 801         (*skp)->af = pd->af;
 802
 803         if (nr != NULL) {
 804                 if ((*nkp = pf_alloc_state_key(M_NOWAIT | M_ZERO)) == NULL)
 805                         return (ENOMEM); /* caller must handle cleanup */
 806
 807                 /* XXX maybe just bcopy and TAILQ_INIT(&(*nkp)->states) */
 808                 PF_ACPY(&(*nkp)->addr[0], &(*skp)->addr[0], pd->af);
 809                 PF_ACPY(&(*nkp)->addr[1], &(*skp)->addr[1], pd->af);
 810                 (*nkp)->port[0] = (*skp)->port[0];
 811                 (*nkp)->port[1] = (*skp)->port[1];
 812                 (*nkp)->proto = pd->proto;
 813                 (*nkp)->af = pd->af;
 814         } else
 815                 *nkp = *skp;
 816
 817         if (pd->dir == PF_IN) {
 818                 *skw = *skp;
 819                 *sks = *nkp;
 820         } else {
 821                 *sks = *skp;
 822                 *skw = *nkp;
 823         }
 824         return (0);
 825 }
 826
 827
 828 int
 829 pf_state_insert(struct pfi_kif *kif, struct pf_state_key *skw,
 830     struct pf_state_key *sks, struct pf_state *s)
 831 {
 832         s->kif = kif;
 833
 834         if (skw == sks) {
 835                 if (pf_state_key_attach(skw, s, PF_SK_WIRE))
 836                         return (-1);
 837                 s->key[PF_SK_STACK] = s->key[PF_SK_WIRE];
 838         } else {
 839                 if (pf_state_key_attach(skw, s, PF_SK_WIRE)) {
 840                         kfree(sks, M_PFSTATEKEYPL);
 841                         return (-1);
 842                 }
 843                 if (pf_state_key_attach(sks, s, PF_SK_STACK)) {
 844                         pf_state_key_detach(s, PF_SK_WIRE);
 845                         return (-1);
 846                 }
 847         }
 848
 849         if (s->id == 0 && s->creatorid == 0) {
 850                 s->id = htobe64(pf_status.stateid++);
 851                 s->creatorid = pf_status.hostid;
 852         }
 853
 854         /*
 855          * Calculate hash code for altq
 856          */
 857         s->hash = crc32(s->key[PF_SK_WIRE], sizeof(*sks));
 858
 859         if (RB_INSERT(pf_state_tree_id, &tree_id, s) != NULL) {
 860                 if (pf_status.debug >= PF_DEBUG_MISC) {
 861                         kprintf("pf: state insert failed: "
 862                             "id: %016jx creatorid: %08x",
 863                               (uintmax_t)be64toh(s->id), ntohl(s->creatorid));
 864                         if (s->sync_flags & PFSTATE_FROMSYNC)
 865                                 kprintf(" (from sync)");
 866                         kprintf("\n");
 867                 }
 868                 pf_detach_state(s);
 869                 return (-1);
 870         }
 871         TAILQ_INSERT_TAIL(&state_list, s, entry_list);
 872         pf_status.fcounters[FCNT_STATE_INSERT]++;
 873         pf_status.states++;
 874         pfi_kif_ref(kif, PFI_KIF_REF_STATE);
 875         pfsync_insert_state(s);
 876         return (0);
 877 }
 878
 879 struct pf_state *
 880 pf_find_state_byid(struct pf_state_cmp *key)
 881 {
 882         pf_status.fcounters[FCNT_STATE_SEARCH]++;
 883
 884         return (RB_FIND(pf_state_tree_id, &tree_id, (struct pf_state *)key));
 885 }
 886
 887 struct pf_state *
 888 pf_find_state(struct pfi_kif *kif, struct pf_state_key_cmp *key, u_int dir,
 889     struct mbuf *m)
 890 {
 891         struct pf_state_key     *sk;
 892         struct pf_state_item    *si;
 893
 894         pf_status.fcounters[FCNT_STATE_SEARCH]++;
 895
 896         if (dir == PF_OUT && m->m_pkthdr.pf.statekey &&
 897             ((struct pf_state_key *)m->m_pkthdr.pf.statekey)->reverse)
 898                 sk = ((struct pf_state_key *)m->m_pkthdr.pf.statekey)->reverse;
 899         else {
 900                 if ((sk = RB_FIND(pf_state_tree, &pf_statetbl,
 901                     (struct pf_state_key *)key)) == NULL)
 902                         return (NULL);
 903                 if (dir == PF_OUT && m->m_pkthdr.pf.statekey) {
 904                         ((struct pf_state_key *)
 905                             m->m_pkthdr.pf.statekey)->reverse = sk;
 906                         sk->reverse = m->m_pkthdr.pf.statekey;
 907                 }
 908         }
 909
 910         if (dir == PF_OUT)
 911                 m->m_pkthdr.pf.statekey = NULL;
 912
 913         /* list is sorted, if-bound states before floating ones */
 914         TAILQ_FOREACH(si, &sk->states, entry)
 915                 if ((si->s->kif == pfi_all || si->s->kif == kif) &&
 916                     sk == (dir == PF_IN ? si->s->key[PF_SK_WIRE] :
 917                     si->s->key[PF_SK_STACK]))
 918                         return (si->s);
 919
 920         return (NULL);
 921 }
 922
 923 struct pf_state *
 924 pf_find_state_all(struct pf_state_key_cmp *key, u_int dir, int *more)
 925 {
 926         struct pf_state_key     *sk;
 927         struct pf_state_item    *si, *ret = NULL;
 928
 929         pf_status.fcounters[FCNT_STATE_SEARCH]++;
 930
 931         sk = RB_FIND(pf_state_tree, &pf_statetbl, (struct pf_state_key *)key);
 932
 933         if (sk != NULL) {
 934                 TAILQ_FOREACH(si, &sk->states, entry)
 935                         if (dir == PF_INOUT ||
 936                             (sk == (dir == PF_IN ? si->s->key[PF_SK_WIRE] :
 937                             si->s->key[PF_SK_STACK]))) {
 938                                 if (more == NULL)
 939                                         return (si->s);
 940
 941                                 if (ret)
 942                                         (*more)++;
 943                                 else
 944                                         ret = si;
 945                         }
 946         }
 947         return (ret ? ret->s : NULL);
 948 }
 949
 950 /* END state table stuff */
 951
 952
 953 void
 954 pf_purge_thread(void *v)
 955 {
 956         int nloops = 0;
 957         int locked = 0;
 958
 959         lwkt_gettoken(&pf_token);
 960         for (;;) {
 961                 tsleep(pf_purge_thread, PWAIT, "pftm", 1 * hz);
 962
 963                 lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
 964
 965                 if (pf_end_threads) {
 966                         pf_purge_expired_states(pf_status.states, 0);
 967                         pf_purge_expired_fragments();
 968                         pf_purge_expired_src_nodes(1);
 969                         pf_end_threads++;
 970
 971                         lockmgr(&pf_consistency_lock, LK_RELEASE);
 972                         wakeup(pf_purge_thread);
 973                         kthread_exit();
 974                 }
 975                 crit_enter();
 976
 977                 /* process a fraction of the state table every second */
 978                 if(!pf_purge_expired_states(1 + (pf_status.states
 979                     / pf_default_rule.timeout[PFTM_INTERVAL]), 0)) {
 980
 981                         pf_purge_expired_states(1 + (pf_status.states
 982                             / pf_default_rule.timeout[PFTM_INTERVAL]), 1);
 983                 }
 984
 985                 /* purge other expired types every PFTM_INTERVAL seconds */
 986                 if (++nloops >= pf_default_rule.timeout[PFTM_INTERVAL]) {
 987                         pf_purge_expired_fragments();
 988                         if (!pf_purge_expired_src_nodes(locked)) {
 989                                 pf_purge_expired_src_nodes(1);
 990                         }
 991                         nloops = 0;
 992                 }
 993                 crit_exit();
 994                 lockmgr(&pf_consistency_lock, LK_RELEASE);
 995         }
 996         lwkt_reltoken(&pf_token);
 997 }
 998
 999 u_int32_t
1000 pf_state_expires(const struct pf_state *state)
1001 {
1002         u_int32_t       timeout;
1003         u_int32_t       start;
1004         u_int32_t       end;
1005         u_int32_t       states;
1006
1007         /* handle all PFTM_* > PFTM_MAX here */
1008         if (state->timeout == PFTM_PURGE)
1009                 return (time_second);
1010         if (state->timeout == PFTM_UNTIL_PACKET)
1011                 return (0);
1012         KKASSERT(state->timeout != PFTM_UNLINKED);
1013         KKASSERT(state->timeout < PFTM_MAX);
1014         timeout = state->rule.ptr->timeout[state->timeout];
1015         if (!timeout)
1016                 timeout = pf_default_rule.timeout[state->timeout];
1017         start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START];
1018         if (start) {
1019                 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END];
1020                 states = state->rule.ptr->states_cur;
1021         } else {
1022                 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START];
1023                 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END];
1024                 states = pf_status.states;
1025         }
1026         if (end && states > start && start < end) {
1027                 if (states < end)
1028                         return (state->expire + timeout * (end - states) /
1029                             (end - start));
1030                 else
1031                         return (time_second);
1032         }
1033         return (state->expire + timeout);
1034 }
1035
1036 int
1037 pf_purge_expired_src_nodes(int waslocked)
1038 {
1039          struct pf_src_node             *cur, *next;
1040          int                             locked = waslocked;
1041
1042          for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) {
1043                  next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur);
1044
1045                  if (cur->states <= 0 && cur->expire <= time_second) {
1046                          if (! locked) {
1047                                  lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
1048                                  next = RB_NEXT(pf_src_tree,
1049                                      &tree_src_tracking, cur);
1050                                  locked = 1;
1051                          }
1052                          if (cur->rule.ptr != NULL) {
1053                                  cur->rule.ptr->src_nodes--;
1054                                  if (cur->rule.ptr->states_cur <= 0 &&
1055                                      cur->rule.ptr->max_src_nodes <= 0)
1056                                          pf_rm_rule(NULL, cur->rule.ptr);
1057                          }
1058                          RB_REMOVE(pf_src_tree, &tree_src_tracking, cur);
1059                          pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
1060                          pf_status.src_nodes--;
1061                          kfree(cur, M_PFSRCTREEPL);
1062                  }
1063          }
1064
1065          if (locked && !waslocked)
1066                 lockmgr(&pf_consistency_lock, LK_RELEASE);
1067         return(1);
1068 }
1069
1070 void
1071 pf_src_tree_remove_state(struct pf_state *s)
1072 {
1073         u_int32_t timeout;
1074
1075         if (s->src_node != NULL) {
1076                 if (s->src.tcp_est)
1077                         --s->src_node->conn;
1078                 if (--s->src_node->states <= 0) {
1079                         timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1080                         if (!timeout)
1081                                 timeout =
1082                                     pf_default_rule.timeout[PFTM_SRC_NODE];
1083                         s->src_node->expire = time_second + timeout;
1084                 }
1085         }
1086         if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) {
1087                 if (--s->nat_src_node->states <= 0) {
1088                         timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1089                         if (!timeout)
1090                                 timeout =
1091                                     pf_default_rule.timeout[PFTM_SRC_NODE];
1092                         s->nat_src_node->expire = time_second + timeout;
1093                 }
1094         }
1095         s->src_node = s->nat_src_node = NULL;
1096 }
1097
1098 /* callers should be at crit_enter() */
1099 void
1100 pf_unlink_state(struct pf_state *cur)
1101 {
1102         if (cur->src.state == PF_TCPS_PROXY_DST) {
1103                 /* XXX wire key the right one? */
1104                 pf_send_tcp(cur->rule.ptr, cur->key[PF_SK_WIRE]->af,
1105                     &cur->key[PF_SK_WIRE]->addr[1],
1106                     &cur->key[PF_SK_WIRE]->addr[0],
1107                     cur->key[PF_SK_WIRE]->port[1],
1108                     cur->key[PF_SK_WIRE]->port[0],
1109                     cur->src.seqhi, cur->src.seqlo + 1,
1110                     TH_RST|TH_ACK, 0, 0, 0, 1, cur->tag, NULL, NULL);
1111         }
1112         RB_REMOVE(pf_state_tree_id, &tree_id, cur);
1113         if (cur->creatorid == pf_status.hostid)
1114                 pfsync_delete_state(cur);
1115         cur->timeout = PFTM_UNLINKED;
1116         pf_src_tree_remove_state(cur);
1117         pf_detach_state(cur);
1118 }
1119
1120 static struct pf_state  *purge_cur;
1121
1122 /* callers should be at crit_enter() and hold the
1123  * write_lock on pf_consistency_lock */
1124 void
1125 pf_free_state(struct pf_state *cur)
1126 {
1127         if (pfsyncif != NULL &&
1128             (pfsyncif->sc_bulk_send_next == cur ||
1129             pfsyncif->sc_bulk_terminator == cur))
1130                 return;
1131         KKASSERT(cur->timeout == PFTM_UNLINKED);
1132         if (--cur->rule.ptr->states_cur <= 0 &&
1133             cur->rule.ptr->src_nodes <= 0)
1134                 pf_rm_rule(NULL, cur->rule.ptr);
1135         if (cur->nat_rule.ptr != NULL)
1136                 if (--cur->nat_rule.ptr->states_cur <= 0 &&
1137                         cur->nat_rule.ptr->src_nodes <= 0)
1138                         pf_rm_rule(NULL, cur->nat_rule.ptr);
1139         if (cur->anchor.ptr != NULL)
1140                 if (--cur->anchor.ptr->states_cur <= 0)
1141                         pf_rm_rule(NULL, cur->anchor.ptr);
1142         pf_normalize_tcp_cleanup(cur);
1143         pfi_kif_unref(cur->kif, PFI_KIF_REF_STATE);
1144
1145         /*
1146          * We may be freeing pf_purge_expired_states()'s saved scan entry,
1147          * adjust it if necessary.
1148          */
1149         if (purge_cur == cur) {
1150                 kprintf("PURGE CONFLICT\n");
1151                 purge_cur = TAILQ_NEXT(purge_cur, entry_list);
1152         }
1153         TAILQ_REMOVE(&state_list, cur, entry_list);
1154         if (cur->tag)
1155                 pf_tag_unref(cur->tag);
1156         kfree(cur, M_PFSTATEPL);
1157         pf_status.fcounters[FCNT_STATE_REMOVALS]++;
1158         pf_status.states--;
1159 }
1160
1161 int
1162 pf_purge_expired_states(u_int32_t maxcheck, int waslocked)
1163 {
1164         struct pf_state         *cur;
1165         int                      locked = waslocked;
1166
1167         while (maxcheck--) {
1168                 /*
1169                  * Wrap to start of list when we hit the end
1170                  */
1171                 cur = purge_cur;
1172                 if (cur == NULL) {
1173                         cur = TAILQ_FIRST(&state_list);
1174                         if (cur == NULL)
1175                                 break;  /* list empty */
1176                 }
1177
1178                 /*
1179                  * Setup next (purge_cur) while we process this one.  If we block and
1180                  * something else deletes purge_cur, pf_free_state() will adjust it further
1181                  * ahead.
1182                  */
1183                 purge_cur = TAILQ_NEXT(cur, entry_list);
1184
1185                 if (cur->timeout == PFTM_UNLINKED) {
1186                         /* free unlinked state */
1187                         if (! locked) {
1188                                 lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
1189                                 locked = 1;
1190                         }
1191                         pf_free_state(cur);
1192                 } else if (pf_state_expires(cur) <= time_second) {
1193                         /* unlink and free expired state */
1194                         pf_unlink_state(cur);
1195                         if (! locked) {
1196                                 if (!lockmgr(&pf_consistency_lock, LK_EXCLUSIVE))
1197                                         return (0);
1198                                 locked = 1;
1199                         }
1200                         pf_free_state(cur);
1201                 }
1202         }
1203
1204         if (locked)
1205                 lockmgr(&pf_consistency_lock, LK_RELEASE);
1206         return (1);
1207 }
1208
1209 int
1210 pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw)
1211 {
1212         if (aw->type != PF_ADDR_TABLE)
1213                 return (0);
1214         if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL)
1215                 return (1);
1216         return (0);
1217 }
1218
1219 void
1220 pf_tbladdr_remove(struct pf_addr_wrap *aw)
1221 {
1222         if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL)
1223                 return;
1224         pfr_detach_table(aw->p.tbl);
1225         aw->p.tbl = NULL;
1226 }
1227
1228 void
1229 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
1230 {
1231         struct pfr_ktable *kt = aw->p.tbl;
1232
1233         if (aw->type != PF_ADDR_TABLE || kt == NULL)
1234                 return;
1235         if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
1236                 kt = kt->pfrkt_root;
1237         aw->p.tbl = NULL;
1238         aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
1239                 kt->pfrkt_cnt : -1;
1240 }
1241
1242 void
1243 pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af)
1244 {
1245         switch (af) {
1246 #ifdef INET
1247         case AF_INET: {
1248                 u_int32_t a = ntohl(addr->addr32[0]);
1249                 kprintf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255,
1250                     (a>>8)&255, a&255);
1251                 if (p) {
1252                         p = ntohs(p);
1253                         kprintf(":%u", p);
1254                 }
1255                 break;
1256         }
1257 #endif /* INET */
1258 #ifdef INET6
1259         case AF_INET6: {
1260                 u_int16_t b;
1261                 u_int8_t i, curstart = 255, curend = 0,
1262                     maxstart = 0, maxend = 0;
1263                 for (i = 0; i < 8; i++) {
1264                         if (!addr->addr16[i]) {
1265                                 if (curstart == 255)
1266                                         curstart = i;
1267                                 else
1268                                         curend = i;
1269                         } else {
1270                                 if (curstart) {
1271                                         if ((curend - curstart) >
1272                                             (maxend - maxstart)) {
1273                                                 maxstart = curstart;
1274                                                 maxend = curend;
1275                                                 curstart = 255;
1276                                         }
1277                                 }
1278                         }
1279                 }
1280                 for (i = 0; i < 8; i++) {
1281                         if (i >= maxstart && i <= maxend) {
1282                                 if (maxend != 7) {
1283                                         if (i == maxstart)
1284                                                 kprintf(":");
1285                                 } else {
1286                                         if (i == maxend)
1287                                                 kprintf(":");
1288                                 }
1289                         } else {
1290                                 b = ntohs(addr->addr16[i]);
1291                                 kprintf("%x", b);
1292                                 if (i < 7)
1293                                         kprintf(":");
1294                         }
1295                 }
1296                 if (p) {
1297                         p = ntohs(p);
1298                         kprintf("[%u]", p);
1299                 }
1300                 break;
1301         }
1302 #endif /* INET6 */
1303         }
1304 }
1305
1306 void
1307 pf_print_state(struct pf_state *s)
1308 {
1309         pf_print_state_parts(s, NULL, NULL);
1310 }
1311
1312 void
1313 pf_print_state_parts(struct pf_state *s,
1314     struct pf_state_key *skwp, struct pf_state_key *sksp)
1315 {
1316         struct pf_state_key *skw, *sks;
1317         u_int8_t proto, dir;
1318
1319         /* Do our best to fill these, but they're skipped if NULL */
1320         skw = skwp ? skwp : (s ? s->key[PF_SK_WIRE] : NULL);
1321         sks = sksp ? sksp : (s ? s->key[PF_SK_STACK] : NULL);
1322         proto = skw ? skw->proto : (sks ? sks->proto : 0);
1323         dir = s ? s->direction : 0;
1324
1325         switch (proto) {
1326         case IPPROTO_TCP:
1327                 kprintf("TCP ");
1328                 break;
1329         case IPPROTO_UDP:
1330                 kprintf("UDP ");
1331                 break;
1332         case IPPROTO_ICMP:
1333                 kprintf("ICMP ");
1334                 break;
1335         case IPPROTO_ICMPV6:
1336                 kprintf("ICMPV6 ");
1337                 break;
1338         default:
1339                 kprintf("%u ", skw->proto);
1340                 break;
1341         }
1342         switch (dir) {
1343         case PF_IN:
1344                 kprintf(" in");
1345                 break;
1346         case PF_OUT:
1347                 kprintf(" out");
1348                 break;
1349         }
1350         if (skw) {
1351                 kprintf(" wire: ");
1352                 pf_print_host(&skw->addr[0], skw->port[0], skw->af);
1353                 kprintf(" ");
1354                 pf_print_host(&skw->addr[1], skw->port[1], skw->af);
1355         }
1356         if (sks) {
1357                 kprintf(" stack: ");
1358                 if (sks != skw) {
1359                         pf_print_host(&sks->addr[0], sks->port[0], sks->af);
1360                         kprintf(" ");
1361                         pf_print_host(&sks->addr[1], sks->port[1], sks->af);
1362                 } else
1363                         kprintf("-");
1364         }
1365         if (s) {
1366                 if (proto == IPPROTO_TCP) {
1367                         kprintf(" [lo=%u high=%u win=%u modulator=%u",
1368                             s->src.seqlo, s->src.seqhi,
1369                             s->src.max_win, s->src.seqdiff);
1370                         if (s->src.wscale && s->dst.wscale)
1371                                 kprintf(" wscale=%u",
1372                                     s->src.wscale & PF_WSCALE_MASK);
1373                         kprintf("]");
1374                         kprintf(" [lo=%u high=%u win=%u modulator=%u",
1375                             s->dst.seqlo, s->dst.seqhi,
1376                             s->dst.max_win, s->dst.seqdiff);
1377                         if (s->src.wscale && s->dst.wscale)
1378                                 kprintf(" wscale=%u",
1379                                 s->dst.wscale & PF_WSCALE_MASK);
1380                         kprintf("]");
1381                 }
1382                 kprintf(" %u:%u", s->src.state, s->dst.state);
1383         }
1384 }
1385
1386 void
1387 pf_print_flags(u_int8_t f)
1388 {
1389         if (f)
1390                 kprintf(" ");
1391         if (f & TH_FIN)
1392                 kprintf("F");
1393         if (f & TH_SYN)
1394                 kprintf("S");
1395         if (f & TH_RST)
1396                 kprintf("R");
1397         if (f & TH_PUSH)
1398                 kprintf("P");
1399         if (f & TH_ACK)
1400                 kprintf("A");
1401         if (f & TH_URG)
1402                 kprintf("U");
1403         if (f & TH_ECE)
1404                 kprintf("E");
1405         if (f & TH_CWR)
1406                 kprintf("W");
1407 }
1408
1409 #define PF_SET_SKIP_STEPS(i)                                    \
1410         do {                                                    \
1411                 while (head[i] != cur) {                        \
1412                         head[i]->skip[i].ptr = cur;             \
1413                         head[i] = TAILQ_NEXT(head[i], entries); \
1414                 }                                               \
1415         } while (0)
1416
1417 void
1418 pf_calc_skip_steps(struct pf_rulequeue *rules)
1419 {
1420         struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT];
1421         int i;
1422
1423         cur = TAILQ_FIRST(rules);
1424         prev = cur;
1425         for (i = 0; i < PF_SKIP_COUNT; ++i)
1426                 head[i] = cur;
1427         while (cur != NULL) {
1428
1429                 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
1430                         PF_SET_SKIP_STEPS(PF_SKIP_IFP);
1431                 if (cur->direction != prev->direction)
1432                         PF_SET_SKIP_STEPS(PF_SKIP_DIR);
1433                 if (cur->af != prev->af)
1434                         PF_SET_SKIP_STEPS(PF_SKIP_AF);
1435                 if (cur->proto != prev->proto)
1436                         PF_SET_SKIP_STEPS(PF_SKIP_PROTO);
1437                 if (cur->src.neg != prev->src.neg ||
1438                     pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr))
1439                         PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR);
1440                 if (cur->src.port[0] != prev->src.port[0] ||
1441                     cur->src.port[1] != prev->src.port[1] ||
1442                     cur->src.port_op != prev->src.port_op)
1443                         PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
1444                 if (cur->dst.neg != prev->dst.neg ||
1445                     pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr))
1446                         PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR);
1447                 if (cur->dst.port[0] != prev->dst.port[0] ||
1448                     cur->dst.port[1] != prev->dst.port[1] ||
1449                     cur->dst.port_op != prev->dst.port_op)
1450                         PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
1451
1452                 prev = cur;
1453                 cur = TAILQ_NEXT(cur, entries);
1454         }
1455         for (i = 0; i < PF_SKIP_COUNT; ++i)
1456                 PF_SET_SKIP_STEPS(i);
1457 }
1458
1459 int
1460 pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2)
1461 {
1462         if (aw1->type != aw2->type)
1463                 return (1);
1464         switch (aw1->type) {
1465         case PF_ADDR_ADDRMASK:
1466         case PF_ADDR_RANGE:
1467                 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0))
1468                         return (1);
1469                 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0))
1470                         return (1);
1471                 return (0);
1472         case PF_ADDR_DYNIFTL:
1473                 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt);
1474         case PF_ADDR_NOROUTE:
1475         case PF_ADDR_URPFFAILED:
1476                 return (0);
1477         case PF_ADDR_TABLE:
1478                 return (aw1->p.tbl != aw2->p.tbl);
1479         case PF_ADDR_RTLABEL:
1480                 return (aw1->v.rtlabel != aw2->v.rtlabel);
1481         default:
1482                 kprintf("invalid address type: %d\n", aw1->type);
1483                 return (1);
1484         }
1485 }
1486
1487 u_int16_t
1488 pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp)
1489 {
1490         u_int32_t       l;
1491
1492         if (udp && !cksum)
1493                 return (0x0000);
1494         l = cksum + old - new;
1495         l = (l >> 16) + (l & 65535);
1496         l = l & 65535;
1497         if (udp && !l)
1498                 return (0xFFFF);
1499         return (l);
1500 }
1501
1502 void
1503 pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc,
1504     struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af)
1505 {
1506         struct pf_addr  ao;
1507         u_int16_t       po = *p;
1508
1509         PF_ACPY(&ao, a, af);
1510         PF_ACPY(a, an, af);
1511
1512         *p = pn;
1513
1514         switch (af) {
1515 #ifdef INET
1516         case AF_INET:
1517                 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1518                     ao.addr16[0], an->addr16[0], 0),
1519                     ao.addr16[1], an->addr16[1], 0);
1520                 *p = pn;
1521                 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1522                     ao.addr16[0], an->addr16[0], u),
1523                     ao.addr16[1], an->addr16[1], u),
1524                     po, pn, u);
1525                 break;
1526 #endif /* INET */
1527 #ifdef INET6
1528         case AF_INET6:
1529                 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1530                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1531                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1532                     ao.addr16[0], an->addr16[0], u),
1533                     ao.addr16[1], an->addr16[1], u),
1534                     ao.addr16[2], an->addr16[2], u),
1535                     ao.addr16[3], an->addr16[3], u),
1536                     ao.addr16[4], an->addr16[4], u),
1537                     ao.addr16[5], an->addr16[5], u),
1538                     ao.addr16[6], an->addr16[6], u),
1539                     ao.addr16[7], an->addr16[7], u),
1540                     po, pn, u);
1541                 break;
1542 #endif /* INET6 */
1543         }
1544 }
1545
1546
1547 /* Changes a u_int32_t.  Uses a void * so there are no align restrictions */
1548 void
1549 pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u)
1550 {
1551         u_int32_t       ao;
1552
1553         memcpy(&ao, a, sizeof(ao));
1554         memcpy(a, &an, sizeof(u_int32_t));
1555         *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u),
1556             ao % 65536, an % 65536, u);
1557 }
1558
1559 #ifdef INET6
1560 void
1561 pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u)
1562 {
1563         struct pf_addr  ao;
1564
1565         PF_ACPY(&ao, a, AF_INET6);
1566         PF_ACPY(a, an, AF_INET6);
1567
1568         *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1569             pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1570             pf_cksum_fixup(pf_cksum_fixup(*c,
1571             ao.addr16[0], an->addr16[0], u),
1572             ao.addr16[1], an->addr16[1], u),
1573             ao.addr16[2], an->addr16[2], u),
1574             ao.addr16[3], an->addr16[3], u),
1575             ao.addr16[4], an->addr16[4], u),
1576             ao.addr16[5], an->addr16[5], u),
1577             ao.addr16[6], an->addr16[6], u),
1578             ao.addr16[7], an->addr16[7], u);
1579 }
1580 #endif /* INET6 */
1581
1582 void
1583 pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa,
1584     struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c,
1585     u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af)
1586 {
1587         struct pf_addr  oia, ooa;
1588
1589         PF_ACPY(&oia, ia, af);
1590         if (oa)
1591                 PF_ACPY(&ooa, oa, af);
1592
1593         /* Change inner protocol port, fix inner protocol checksum. */
1594         if (ip != NULL) {
1595                 u_int16_t       oip = *ip;
1596                 u_int32_t       opc = 0;
1597
1598                 if (pc != NULL)
1599                         opc = *pc;
1600                 *ip = np;
1601                 if (pc != NULL)
1602                         *pc = pf_cksum_fixup(*pc, oip, *ip, u);
1603                 *ic = pf_cksum_fixup(*ic, oip, *ip, 0);
1604                 if (pc != NULL)
1605                         *ic = pf_cksum_fixup(*ic, opc, *pc, 0);
1606         }
1607         /* Change inner ip address, fix inner ip and icmp checksums. */
1608         PF_ACPY(ia, na, af);
1609         switch (af) {
1610 #ifdef INET
1611         case AF_INET: {
1612                 u_int32_t        oh2c = *h2c;
1613
1614                 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c,
1615                     oia.addr16[0], ia->addr16[0], 0),
1616                     oia.addr16[1], ia->addr16[1], 0);
1617                 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1618                     oia.addr16[0], ia->addr16[0], 0),
1619                     oia.addr16[1], ia->addr16[1], 0);
1620                 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0);
1621                 break;
1622         }
1623 #endif /* INET */
1624 #ifdef INET6
1625         case AF_INET6:
1626                 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1627                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1628                     pf_cksum_fixup(pf_cksum_fixup(*ic,
1629                     oia.addr16[0], ia->addr16[0], u),
1630                     oia.addr16[1], ia->addr16[1], u),
1631                     oia.addr16[2], ia->addr16[2], u),
1632                     oia.addr16[3], ia->addr16[3], u),
1633                     oia.addr16[4], ia->addr16[4], u),
1634                     oia.addr16[5], ia->addr16[5], u),
1635                     oia.addr16[6], ia->addr16[6], u),
1636                     oia.addr16[7], ia->addr16[7], u);
1637                 break;
1638 #endif /* INET6 */
1639         }
1640         /* Outer ip address, fix outer ip or icmpv6 checksum, if necessary. */
1641         if (oa) {
1642                 PF_ACPY(oa, na, af);
1643                 switch (af) {
1644 #ifdef INET
1645                 case AF_INET:
1646                         *hc = pf_cksum_fixup(pf_cksum_fixup(*hc,
1647                             ooa.addr16[0], oa->addr16[0], 0),
1648                             ooa.addr16[1], oa->addr16[1], 0);
1649                         break;
1650 #endif /* INET */
1651 #ifdef INET6
1652                 case AF_INET6:
1653                         *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1654                             pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1655                             pf_cksum_fixup(pf_cksum_fixup(*ic,
1656                             ooa.addr16[0], oa->addr16[0], u),
1657                             ooa.addr16[1], oa->addr16[1], u),
1658                             ooa.addr16[2], oa->addr16[2], u),
1659                             ooa.addr16[3], oa->addr16[3], u),
1660                             ooa.addr16[4], oa->addr16[4], u),
1661                             ooa.addr16[5], oa->addr16[5], u),
1662                             ooa.addr16[6], oa->addr16[6], u),
1663                             ooa.addr16[7], oa->addr16[7], u);
1664                         break;
1665 #endif /* INET6 */
1666                 }
1667         }
1668 }
1669
1670
1671 /*
1672  * Need to modulate the sequence numbers in the TCP SACK option
1673  * (credits to Krzysztof Pfaff for report and patch)
1674  */
1675 int
1676 pf_modulate_sack(struct mbuf *m, int off, struct pf_pdesc *pd,
1677     struct tcphdr *th, struct pf_state_peer *dst)
1678 {
1679         int hlen = (th->th_off << 2) - sizeof(*th), thoptlen = hlen;
1680         u_int8_t opts[TCP_MAXOLEN], *opt = opts;
1681         int copyback = 0, i, olen;
1682         struct raw_sackblock sack;
1683
1684 #define TCPOLEN_SACKLEN (TCPOLEN_SACK + 2)
1685         if (hlen < TCPOLEN_SACKLEN ||
1686             !pf_pull_hdr(m, off + sizeof(*th), opts, hlen, NULL, NULL, pd->af))
1687                 return 0;
1688
1689         while (hlen >= TCPOLEN_SACKLEN) {
1690                 olen = opt[1];
1691                 switch (*opt) {
1692                 case TCPOPT_EOL:        /* FALLTHROUGH */
1693                 case TCPOPT_NOP:
1694                         opt++;
1695                         hlen--;
1696                         break;
1697                 case TCPOPT_SACK:
1698                         if (olen > hlen)
1699                                 olen = hlen;
1700                         if (olen >= TCPOLEN_SACKLEN) {
1701                                 for (i = 2; i + TCPOLEN_SACK <= olen;
1702                                     i += TCPOLEN_SACK) {
1703                                         memcpy(&sack, &opt[i], sizeof(sack));
1704                                         pf_change_a(&sack.rblk_start, &th->th_sum,
1705                                             htonl(ntohl(sack.rblk_start) -
1706                                             dst->seqdiff), 0);
1707                                         pf_change_a(&sack.rblk_end, &th->th_sum,
1708                                             htonl(ntohl(sack.rblk_end) -
1709                                             dst->seqdiff), 0);
1710                                         memcpy(&opt[i], &sack, sizeof(sack));
1711                                 }
1712                                 copyback = 1;
1713                         }
1714                         /* FALLTHROUGH */
1715                 default:
1716                         if (olen < 2)
1717                                 olen = 2;
1718                         hlen -= olen;
1719                         opt += olen;
1720                 }
1721         }
1722
1723         if (copyback)
1724                 m_copyback(m, off + sizeof(*th), thoptlen, opts);
1725         return (copyback);
1726 }
1727
1728 void
1729 pf_send_tcp(const struct pf_rule *r, sa_family_t af,
1730     const struct pf_addr *saddr, const struct pf_addr *daddr,
1731     u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack,
1732     u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl, int tag,
1733     u_int16_t rtag, struct ether_header *eh, struct ifnet *ifp)
1734 {
1735         struct mbuf     *m;
1736         int              len = 0, tlen;
1737 #ifdef INET
1738         struct ip       *h = NULL;
1739 #endif /* INET */
1740 #ifdef INET6
1741         struct ip6_hdr  *h6 = NULL;
1742 #endif /* INET6 */
1743         struct tcphdr   *th = NULL;
1744         char            *opt;
1745
1746         ASSERT_LWKT_TOKEN_HELD(&pf_token);
1747
1748         /* maximum segment size tcp option */
1749         tlen = sizeof(struct tcphdr);
1750         if (mss)
1751                 tlen += 4;
1752
1753         switch (af) {
1754 #ifdef INET
1755         case AF_INET:
1756                 len = sizeof(struct ip) + tlen;
1757                 break;
1758 #endif /* INET */
1759 #ifdef INET6
1760         case AF_INET6:
1761                 len = sizeof(struct ip6_hdr) + tlen;
1762                 break;
1763 #endif /* INET6 */
1764         }
1765
1766         /*
1767          * Create outgoing mbuf.
1768          *
1769          * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags,
1770          * so make sure pf.flags is clear.
1771          */
1772         m = m_gethdr(MB_DONTWAIT, MT_HEADER);
1773         if (m == NULL) {
1774                 return;
1775         }
1776         if (tag)
1777                 m->m_pkthdr.fw_flags |= PF_MBUF_TAGGED;
1778         m->m_pkthdr.pf.flags = 0;
1779         m->m_pkthdr.pf.tag = rtag;
1780         /* XXX Recheck when upgrading to > 4.4 */
1781         m->m_pkthdr.pf.statekey = NULL;
1782         if (r != NULL && r->rtableid >= 0)
1783                 m->m_pkthdr.pf.rtableid = r->rtableid;
1784
1785 #ifdef ALTQ
1786         if (r != NULL && r->qid) {
1787                 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
1788                 m->m_pkthdr.pf.qid = r->qid;
1789                 m->m_pkthdr.pf.ecn_af = af;
1790                 m->m_pkthdr.pf.hdr = mtod(m, struct ip *);
1791         }
1792 #endif /* ALTQ */
1793         m->m_data += max_linkhdr;
1794         m->m_pkthdr.len = m->m_len = len;
1795         m->m_pkthdr.rcvif = NULL;
1796         bzero(m->m_data, len);
1797         switch (af) {
1798 #ifdef INET
1799         case AF_INET:
1800                 h = mtod(m, struct ip *);
1801
1802                 /* IP header fields included in the TCP checksum */
1803                 h->ip_p = IPPROTO_TCP;
1804                 h->ip_len = tlen;
1805                 h->ip_src.s_addr = saddr->v4.s_addr;
1806                 h->ip_dst.s_addr = daddr->v4.s_addr;
1807
1808                 th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip));
1809                 break;
1810 #endif /* INET */
1811 #ifdef INET6
1812         case AF_INET6:
1813                 h6 = mtod(m, struct ip6_hdr *);
1814
1815                 /* IP header fields included in the TCP checksum */
1816                 h6->ip6_nxt = IPPROTO_TCP;
1817                 h6->ip6_plen = htons(tlen);
1818                 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr));
1819                 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr));
1820
1821                 th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr));
1822                 break;
1823 #endif /* INET6 */
1824         }
1825
1826         /* TCP header */
1827         th->th_sport = sport;
1828         th->th_dport = dport;
1829         th->th_seq = htonl(seq);
1830         th->th_ack = htonl(ack);
1831         th->th_off = tlen >> 2;
1832         th->th_flags = flags;
1833         th->th_win = htons(win);
1834
1835         if (mss) {
1836                 opt = (char *)(th + 1);
1837                 opt[0] = TCPOPT_MAXSEG;
1838                 opt[1] = 4;
1839                 mss = htons(mss);
1840                 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2);
1841         }
1842
1843         switch (af) {
1844 #ifdef INET
1845         case AF_INET:
1846                 /* TCP checksum */
1847                 th->th_sum = in_cksum(m, len);
1848
1849                 /* Finish the IP header */
1850                 h->ip_v = 4;
1851                 h->ip_hl = sizeof(*h) >> 2;
1852                 h->ip_tos = IPTOS_LOWDELAY;
1853                 h->ip_len = len;
1854                 h->ip_off = path_mtu_discovery ? IP_DF : 0;
1855                 h->ip_ttl = ttl ? ttl : ip_defttl;
1856                 h->ip_sum = 0;
1857                 if (eh == NULL) {
1858                         lwkt_reltoken(&pf_token);
1859                         ip_output(m, NULL, NULL, 0, NULL, NULL);
1860                         lwkt_gettoken(&pf_token);
1861                 } else {
1862                         struct route             ro;
1863                         struct rtentry           rt;
1864                         struct ether_header     *e = (void *)ro.ro_dst.sa_data;
1865
1866                         if (ifp == NULL) {
1867                                 m_freem(m);
1868                                 return;
1869                         }
1870                         rt.rt_ifp = ifp;
1871                         ro.ro_rt = &rt;
1872                         ro.ro_dst.sa_len = sizeof(ro.ro_dst);
1873                         ro.ro_dst.sa_family = pseudo_AF_HDRCMPLT;
1874                         bcopy(eh->ether_dhost, e->ether_shost, ETHER_ADDR_LEN);
1875                         bcopy(eh->ether_shost, e->ether_dhost, ETHER_ADDR_LEN);
1876                         e->ether_type = eh->ether_type;
1877                         /* XXX_IMPORT: later */
1878                         lwkt_reltoken(&pf_token);
1879                         ip_output(m, (void *)NULL, &ro, 0,
1880                             (void *)NULL, (void *)NULL);
1881                         lwkt_gettoken(&pf_token);
1882                 }
1883                 break;
1884 #endif /* INET */
1885 #ifdef INET6
1886         case AF_INET6:
1887                 /* TCP checksum */
1888                 th->th_sum = in6_cksum(m, IPPROTO_TCP,
1889                     sizeof(struct ip6_hdr), tlen);
1890
1891                 h6->ip6_vfc |= IPV6_VERSION;
1892                 h6->ip6_hlim = IPV6_DEFHLIM;
1893
1894                 lwkt_reltoken(&pf_token);
1895                 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
1896                 lwkt_gettoken(&pf_token);
1897                 break;
1898 #endif /* INET6 */
1899         }
1900 }
1901
1902 void
1903 pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af,
1904     struct pf_rule *r)
1905 {
1906         struct mbuf     *m0;
1907
1908         /*
1909          * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags,
1910          * so make sure pf.flags is clear.
1911          */
1912         if ((m0 = m_copy(m, 0, M_COPYALL)) == NULL)
1913                 return;
1914
1915         m0->m_pkthdr.fw_flags |= PF_MBUF_TAGGED;
1916         m0->m_pkthdr.pf.flags = 0;
1917         /* XXX Re-Check when Upgrading to > 4.4 */
1918         m0->m_pkthdr.pf.statekey = NULL;
1919
1920         if (r->rtableid >= 0)
1921                 m0->m_pkthdr.pf.rtableid = r->rtableid;
1922
1923 #ifdef ALTQ
1924         if (r->qid) {
1925                 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
1926                 m0->m_pkthdr.pf.qid = r->qid;
1927                 m0->m_pkthdr.pf.ecn_af = af;
1928                 m0->m_pkthdr.pf.hdr = mtod(m0, struct ip *);
1929         }
1930 #endif /* ALTQ */
1931
1932         switch (af) {
1933 #ifdef INET
1934         case AF_INET:
1935                 icmp_error(m0, type, code, 0, 0);
1936                 break;
1937 #endif /* INET */
1938 #ifdef INET6
1939         case AF_INET6:
1940                 icmp6_error(m0, type, code, 0);
1941                 break;
1942 #endif /* INET6 */
1943         }
1944 }
1945
1946 /*
1947  * Return 1 if the addresses a and b match (with mask m), otherwise return 0.
1948  * If n is 0, they match if they are equal. If n is != 0, they match if they
1949  * are different.
1950  */
1951 int
1952 pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m,
1953     struct pf_addr *b, sa_family_t af)
1954 {
1955         int     match = 0;
1956
1957         switch (af) {
1958 #ifdef INET
1959         case AF_INET:
1960                 if ((a->addr32[0] & m->addr32[0]) ==
1961                     (b->addr32[0] & m->addr32[0]))
1962                         match++;
1963                 break;
1964 #endif /* INET */
1965 #ifdef INET6
1966         case AF_INET6:
1967                 if (((a->addr32[0] & m->addr32[0]) ==
1968                      (b->addr32[0] & m->addr32[0])) &&
1969                     ((a->addr32[1] & m->addr32[1]) ==
1970                      (b->addr32[1] & m->addr32[1])) &&
1971                     ((a->addr32[2] & m->addr32[2]) ==
1972                      (b->addr32[2] & m->addr32[2])) &&
1973                     ((a->addr32[3] & m->addr32[3]) ==
1974                      (b->addr32[3] & m->addr32[3])))
1975                         match++;
1976                 break;
1977 #endif /* INET6 */
1978         }
1979         if (match) {
1980                 if (n)
1981                         return (0);
1982                 else
1983                         return (1);
1984         } else {
1985                 if (n)
1986                         return (1);
1987                 else
1988                         return (0);
1989         }
1990 }
1991
1992 /*
1993  * Return 1 if b <= a <= e, otherwise return 0.
1994  */
1995 int
1996 pf_match_addr_range(struct pf_addr *b, struct pf_addr *e,
1997     struct pf_addr *a, sa_family_t af)
1998 {
1999         switch (af) {
2000 #ifdef INET
2001         case AF_INET:
2002                 if ((a->addr32[0] < b->addr32[0]) ||
2003                     (a->addr32[0] > e->addr32[0]))
2004                         return (0);
2005                 break;
2006 #endif /* INET */
2007 #ifdef INET6
2008         case AF_INET6: {
2009                 int     i;
2010
2011                 /* check a >= b */
2012                 for (i = 0; i < 4; ++i)
2013                         if (a->addr32[i] > b->addr32[i])
2014                                 break;
2015                         else if (a->addr32[i] < b->addr32[i])
2016                                 return (0);
2017                 /* check a <= e */
2018                 for (i = 0; i < 4; ++i)
2019                         if (a->addr32[i] < e->addr32[i])
2020                                 break;
2021                         else if (a->addr32[i] > e->addr32[i])
2022                                 return (0);
2023                 break;
2024         }
2025 #endif /* INET6 */
2026         }
2027         return (1);
2028 }
2029
2030 int
2031 pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p)
2032 {
2033         switch (op) {
2034         case PF_OP_IRG:
2035                 return ((p > a1) && (p < a2));
2036         case PF_OP_XRG:
2037                 return ((p < a1) || (p > a2));
2038         case PF_OP_RRG:
2039                 return ((p >= a1) && (p <= a2));
2040         case PF_OP_EQ:
2041                 return (p == a1);
2042         case PF_OP_NE:
2043                 return (p != a1);
2044         case PF_OP_LT:
2045                 return (p < a1);
2046         case PF_OP_LE:
2047                 return (p <= a1);
2048         case PF_OP_GT:
2049                 return (p > a1);
2050         case PF_OP_GE:
2051                 return (p >= a1);
2052         }
2053         return (0); /* never reached */
2054 }
2055
2056 int
2057 pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p)
2058 {
2059         a1 = ntohs(a1);
2060         a2 = ntohs(a2);
2061         p = ntohs(p);
2062         return (pf_match(op, a1, a2, p));
2063 }
2064
2065 int
2066 pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u)
2067 {
2068         if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
2069                 return (0);
2070         return (pf_match(op, a1, a2, u));
2071 }
2072
2073 int
2074 pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g)
2075 {
2076         if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
2077                 return (0);
2078         return (pf_match(op, a1, a2, g));
2079 }
2080
2081 int
2082 pf_match_tag(struct mbuf *m, struct pf_rule *r, int *tag)
2083 {
2084         if (*tag == -1)
2085                 *tag = m->m_pkthdr.pf.tag;
2086
2087         return ((!r->match_tag_not && r->match_tag == *tag) ||
2088             (r->match_tag_not && r->match_tag != *tag));
2089 }
2090
2091 int
2092 pf_tag_packet(struct mbuf *m, int tag, int rtableid)
2093 {
2094         if (tag <= 0 && rtableid < 0)
2095                 return (0);
2096
2097         if (tag > 0)
2098                 m->m_pkthdr.pf.tag = tag;
2099         if (rtableid >= 0)
2100                 m->m_pkthdr.pf.rtableid = rtableid;
2101
2102         return (0);
2103 }
2104
2105 void
2106 pf_step_into_anchor(int *depth, struct pf_ruleset **rs, int n,
2107     struct pf_rule **r, struct pf_rule **a, int *match)
2108 {
2109         struct pf_anchor_stackframe     *f;
2110
2111         (*r)->anchor->match = 0;
2112         if (match)
2113                 *match = 0;
2114         if (*depth >= NELEM(pf_anchor_stack)) {
2115                 kprintf("pf_step_into_anchor: stack overflow\n");
2116                 *r = TAILQ_NEXT(*r, entries);
2117                 return;
2118         } else if (*depth == 0 && a != NULL)
2119                 *a = *r;
2120         f = pf_anchor_stack + (*depth)++;
2121         f->rs = *rs;
2122         f->r = *r;
2123         if ((*r)->anchor_wildcard) {
2124                 f->parent = &(*r)->anchor->children;
2125                 if ((f->child = RB_MIN(pf_anchor_node, f->parent)) ==
2126                     NULL) {
2127                         *r = NULL;
2128                         return;
2129                 }
2130                 *rs = &f->child->ruleset;
2131         } else {
2132                 f->parent = NULL;
2133                 f->child = NULL;
2134                 *rs = &(*r)->anchor->ruleset;
2135         }
2136         *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
2137 }
2138
2139 int
2140 pf_step_out_of_anchor(int *depth, struct pf_ruleset **rs, int n,
2141     struct pf_rule **r, struct pf_rule **a, int *match)
2142 {
2143         struct pf_anchor_stackframe     *f;
2144         int quick = 0;
2145
2146         do {
2147                 if (*depth <= 0)
2148                         break;
2149                 f = pf_anchor_stack + *depth - 1;
2150                 if (f->parent != NULL && f->child != NULL) {
2151                         if (f->child->match ||
2152                             (match != NULL && *match)) {
2153                                 f->r->anchor->match = 1;
2154                                 *match = 0;
2155                         }
2156                         f->child = RB_NEXT(pf_anchor_node, f->parent, f->child);
2157                         if (f->child != NULL) {
2158                                 *rs = &f->child->ruleset;
2159                                 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
2160                                 if (*r == NULL)
2161                                         continue;
2162                                 else
2163                                         break;
2164                         }
2165                 }
2166                 (*depth)--;
2167                 if (*depth == 0 && a != NULL)
2168                         *a = NULL;
2169                 *rs = f->rs;
2170                 if (f->r->anchor->match || (match != NULL && *match))
2171                         quick = f->r->quick;
2172                 *r = TAILQ_NEXT(f->r, entries);
2173         } while (*r == NULL);
2174
2175         return (quick);
2176 }
2177
2178 #ifdef INET6
2179 void
2180 pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr,
2181     struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af)
2182 {
2183         switch (af) {
2184 #ifdef INET
2185         case AF_INET:
2186                 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2187                 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2188                 break;
2189 #endif /* INET */
2190         case AF_INET6:
2191                 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2192                 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2193                 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) |
2194                 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]);
2195                 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) |
2196                 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]);
2197                 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) |
2198                 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]);
2199                 break;
2200         }
2201 }
2202
2203 void
2204 pf_addr_inc(struct pf_addr *addr, sa_family_t af)
2205 {
2206         switch (af) {
2207 #ifdef INET
2208         case AF_INET:
2209                 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1);
2210                 break;
2211 #endif /* INET */
2212         case AF_INET6:
2213                 if (addr->addr32[3] == 0xffffffff) {
2214                         addr->addr32[3] = 0;
2215                         if (addr->addr32[2] == 0xffffffff) {
2216                                 addr->addr32[2] = 0;
2217                                 if (addr->addr32[1] == 0xffffffff) {
2218                                         addr->addr32[1] = 0;
2219                                         addr->addr32[0] =
2220                                             htonl(ntohl(addr->addr32[0]) + 1);
2221                                 } else
2222                                         addr->addr32[1] =
2223                                             htonl(ntohl(addr->addr32[1]) + 1);
2224                         } else
2225                                 addr->addr32[2] =
2226                                     htonl(ntohl(addr->addr32[2]) + 1);
2227                 } else
2228                         addr->addr32[3] =
2229                             htonl(ntohl(addr->addr32[3]) + 1);
2230                 break;
2231         }
2232 }
2233 #endif /* INET6 */
2234
2235 #define mix(a,b,c) \
2236         do {                                    \
2237                 a -= b; a -= c; a ^= (c >> 13); \
2238                 b -= c; b -= a; b ^= (a << 8);  \
2239                 c -= a; c -= b; c ^= (b >> 13); \
2240                 a -= b; a -= c; a ^= (c >> 12); \
2241                 b -= c; b -= a; b ^= (a << 16); \
2242                 c -= a; c -= b; c ^= (b >> 5);  \
2243                 a -= b; a -= c; a ^= (c >> 3);  \
2244                 b -= c; b -= a; b ^= (a << 10); \
2245                 c -= a; c -= b; c ^= (b >> 15); \
2246         } while (0)
2247
2248 /*
2249  * hash function based on bridge_hash in if_bridge.c
2250  */
2251 void
2252 pf_hash(struct pf_addr *inaddr, struct pf_addr *hash,
2253     struct pf_poolhashkey *key, sa_family_t af)
2254 {
2255         u_int32_t       a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0];
2256
2257         switch (af) {
2258 #ifdef INET
2259         case AF_INET:
2260                 a += inaddr->addr32[0];
2261                 b += key->key32[1];
2262                 mix(a, b, c);
2263                 hash->addr32[0] = c + key->key32[2];
2264                 break;
2265 #endif /* INET */
2266 #ifdef INET6
2267         case AF_INET6:
2268                 a += inaddr->addr32[0];
2269                 b += inaddr->addr32[2];
2270                 mix(a, b, c);
2271                 hash->addr32[0] = c;
2272                 a += inaddr->addr32[1];
2273                 b += inaddr->addr32[3];
2274                 c += key->key32[1];
2275                 mix(a, b, c);
2276                 hash->addr32[1] = c;
2277                 a += inaddr->addr32[2];
2278                 b += inaddr->addr32[1];
2279                 c += key->key32[2];
2280                 mix(a, b, c);
2281                 hash->addr32[2] = c;
2282                 a += inaddr->addr32[3];
2283                 b += inaddr->addr32[0];
2284                 c += key->key32[3];
2285                 mix(a, b, c);
2286                 hash->addr32[3] = c;
2287                 break;
2288 #endif /* INET6 */
2289         }
2290 }
2291
2292 int
2293 pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr,
2294     struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn)
2295 {
2296         unsigned char            hash[16];
2297         struct pf_pool          *rpool = &r->rpool;
2298         struct pf_addr          *raddr = &rpool->cur->addr.v.a.addr;
2299         struct pf_addr          *rmask = &rpool->cur->addr.v.a.mask;
2300         struct pf_pooladdr      *acur = rpool->cur;
2301         struct pf_src_node       k;
2302
2303         if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR &&
2304             (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2305                 k.af = af;
2306                 PF_ACPY(&k.addr, saddr, af);
2307                 if (r->rule_flag & PFRULE_RULESRCTRACK ||
2308                     r->rpool.opts & PF_POOL_STICKYADDR)
2309                         k.rule.ptr = r;
2310                 else
2311                         k.rule.ptr = NULL;
2312                 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
2313                 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
2314                 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) {
2315                         PF_ACPY(naddr, &(*sn)->raddr, af);
2316                         if (pf_status.debug >= PF_DEBUG_MISC) {
2317                                 kprintf("pf_map_addr: src tracking maps ");
2318                                 pf_print_host(&k.addr, 0, af);
2319                                 kprintf(" to ");
2320                                 pf_print_host(naddr, 0, af);
2321                                 kprintf("\n");
2322                         }
2323                         return (0);
2324                 }
2325         }
2326
2327         if (rpool->cur->addr.type == PF_ADDR_NOROUTE)
2328                 return (1);
2329         if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2330                 switch (af) {
2331 #ifdef INET
2332                 case AF_INET:
2333                         if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 &&
2334                             (rpool->opts & PF_POOL_TYPEMASK) !=
2335                             PF_POOL_ROUNDROBIN)
2336                                 return (1);
2337                          raddr = &rpool->cur->addr.p.dyn->pfid_addr4;
2338                          rmask = &rpool->cur->addr.p.dyn->pfid_mask4;
2339                         break;
2340 #endif /* INET */
2341 #ifdef INET6
2342                 case AF_INET6:
2343                         if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 &&
2344                             (rpool->opts & PF_POOL_TYPEMASK) !=
2345                             PF_POOL_ROUNDROBIN)
2346                                 return (1);
2347                         raddr = &rpool->cur->addr.p.dyn->pfid_addr6;
2348                         rmask = &rpool->cur->addr.p.dyn->pfid_mask6;
2349                         break;
2350 #endif /* INET6 */
2351                 }
2352         } else if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2353                 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN)
2354                         return (1); /* unsupported */
2355         } else {
2356                 raddr = &rpool->cur->addr.v.a.addr;
2357                 rmask = &rpool->cur->addr.v.a.mask;
2358         }
2359
2360         switch (rpool->opts & PF_POOL_TYPEMASK) {
2361         case PF_POOL_NONE:
2362                 PF_ACPY(naddr, raddr, af);
2363                 break;
2364         case PF_POOL_BITMASK:
2365                 PF_POOLMASK(naddr, raddr, rmask, saddr, af);
2366                 break;
2367         case PF_POOL_RANDOM:
2368                 if (init_addr != NULL && PF_AZERO(init_addr, af)) {
2369                         switch (af) {
2370 #ifdef INET
2371                         case AF_INET:
2372                                 rpool->counter.addr32[0] = htonl(karc4random());
2373                                 break;
2374 #endif /* INET */
2375 #ifdef INET6
2376                         case AF_INET6:
2377                                 if (rmask->addr32[3] != 0xffffffff)
2378                                         rpool->counter.addr32[3] =
2379                                             htonl(karc4random());
2380                                 else
2381                                         break;
2382                                 if (rmask->addr32[2] != 0xffffffff)
2383                                         rpool->counter.addr32[2] =
2384                                             htonl(karc4random());
2385                                 else
2386                                         break;
2387                                 if (rmask->addr32[1] != 0xffffffff)
2388                                         rpool->counter.addr32[1] =
2389                                             htonl(karc4random());
2390                                 else
2391                                         break;
2392                                 if (rmask->addr32[0] != 0xffffffff)
2393                                         rpool->counter.addr32[0] =
2394                                             htonl(karc4random());
2395                                 break;
2396 #endif /* INET6 */
2397                         }
2398                         PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2399                         PF_ACPY(init_addr, naddr, af);
2400
2401                 } else {
2402                         PF_AINC(&rpool->counter, af);
2403                         PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2404                 }
2405                 break;
2406         case PF_POOL_SRCHASH:
2407                 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af);
2408                 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af);
2409                 break;
2410         case PF_POOL_ROUNDROBIN:
2411                 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2412                         if (!pfr_pool_get(rpool->cur->addr.p.tbl,
2413                             &rpool->tblidx, &rpool->counter,
2414                             &raddr, &rmask, af))
2415                                 goto get_addr;
2416                 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2417                         if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2418                             &rpool->tblidx, &rpool->counter,
2419                             &raddr, &rmask, af))
2420                                 goto get_addr;
2421                 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af))
2422                         goto get_addr;
2423
2424         try_next:
2425                 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL)
2426                         rpool->cur = TAILQ_FIRST(&rpool->list);
2427                 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2428                         rpool->tblidx = -1;
2429                         if (pfr_pool_get(rpool->cur->addr.p.tbl,
2430                             &rpool->tblidx, &rpool->counter,
2431                             &raddr, &rmask, af)) {
2432                                 /* table contains no address of type 'af' */
2433                                 if (rpool->cur != acur)
2434                                         goto try_next;
2435                                 return (1);
2436                         }
2437                 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2438                         rpool->tblidx = -1;
2439                         if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2440                             &rpool->tblidx, &rpool->counter,
2441                             &raddr, &rmask, af)) {
2442                                 /* table contains no address of type 'af' */
2443                                 if (rpool->cur != acur)
2444                                         goto try_next;
2445                                 return (1);
2446                         }
2447                 } else {
2448                         raddr = &rpool->cur->addr.v.a.addr;
2449                         rmask = &rpool->cur->addr.v.a.mask;
2450                         PF_ACPY(&rpool->counter, raddr, af);
2451                 }
2452
2453         get_addr:
2454                 PF_ACPY(naddr, &rpool->counter, af);
2455                 if (init_addr != NULL && PF_AZERO(init_addr, af))
2456                         PF_ACPY(init_addr, naddr, af);
2457                 PF_AINC(&rpool->counter, af);
2458                 break;
2459         }
2460         if (*sn != NULL)
2461                 PF_ACPY(&(*sn)->raddr, naddr, af);
2462
2463         if (pf_status.debug >= PF_DEBUG_MISC &&
2464             (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2465                 kprintf("pf_map_addr: selected address ");
2466                 pf_print_host(naddr, 0, af);
2467                 kprintf("\n");
2468         }
2469
2470         return (0);
2471 }
2472
2473 int
2474 pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r,
2475     struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport,
2476     struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high,
2477     struct pf_src_node **sn)
2478 {
2479         struct pf_state_key_cmp key;
2480         struct pf_addr          init_addr;
2481         u_int16_t               cut;
2482
2483         bzero(&init_addr, sizeof(init_addr));
2484         if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2485                 return (1);
2486
2487         if (proto == IPPROTO_ICMP) {
2488                 low = 1;
2489                 high = 65535;
2490         }
2491
2492         do {
2493                 key.af = af;
2494                 key.proto = proto;
2495                 PF_ACPY(&key.addr[1], daddr, key.af);
2496                 PF_ACPY(&key.addr[0], naddr, key.af);
2497                 key.port[1] = dport;
2498
2499                 /*
2500                  * port search; start random, step;
2501                  * similar 2 portloop in in_pcbbind
2502                  */
2503                 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP ||
2504                     proto == IPPROTO_ICMP)) {
2505                         key.port[0] = dport;
2506                         if (pf_find_state_all(&key, PF_IN, NULL) == NULL)
2507                                 return (0);
2508                 } else if (low == 0 && high == 0) {
2509                         key.port[0] = *nport;
2510                         if (pf_find_state_all(&key, PF_IN, NULL) == NULL)
2511                                 return (0);
2512                 } else if (low == high) {
2513                         key.port[0] = htons(low);
2514                         if (pf_find_state_all(&key, PF_IN, NULL) == NULL) {
2515                                 *nport = htons(low);
2516                                 return (0);
2517                         }
2518                 } else {
2519                         u_int16_t tmp;
2520
2521                         if (low > high) {
2522                                 tmp = low;
2523                                 low = high;
2524                                 high = tmp;
2525                         }
2526                         /* low < high */
2527                         cut = htonl(karc4random()) % (1 + high - low) + low;
2528                         /* low <= cut <= high */
2529                         for (tmp = cut; tmp <= high; ++(tmp)) {
2530                                 key.port[0] = htons(tmp);
2531                                 if (pf_find_state_all(&key, PF_IN, NULL) ==
2532                                     NULL && !in_baddynamic(tmp, proto)) {
2533                                         *nport = htons(tmp);
2534                                         return (0);
2535                                 }
2536                         }
2537                         for (tmp = cut - 1; tmp >= low; --(tmp)) {
2538                                 key.port[0] = htons(tmp);
2539                                 if (pf_find_state_all(&key, PF_IN, NULL) ==
2540                                     NULL && !in_baddynamic(tmp, proto)) {
2541                                         *nport = htons(tmp);
2542                                         return (0);
2543                                 }
2544                         }
2545                 }
2546
2547                 switch (r->rpool.opts & PF_POOL_TYPEMASK) {
2548                 case PF_POOL_RANDOM:
2549                 case PF_POOL_ROUNDROBIN:
2550                         if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2551                                 return (1);
2552                         break;
2553                 case PF_POOL_NONE:
2554                 case PF_POOL_SRCHASH:
2555                 case PF_POOL_BITMASK:
2556                 default:
2557                         return (1);
2558                 }
2559         } while (! PF_AEQ(&init_addr, naddr, af) );
2560         return (1);                                     /* none available */
2561 }
2562
2563 struct pf_rule *
2564 pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off,
2565     int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport,
2566     struct pf_addr *daddr, u_int16_t dport, int rs_num)
2567 {
2568         struct pf_rule          *r, *rm = NULL;
2569         struct pf_ruleset       *ruleset = NULL;
2570         int                      tag = -1;
2571         int                      rtableid = -1;
2572         int                      asd = 0;
2573
2574         r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr);
2575         while (r && rm == NULL) {
2576                 struct pf_rule_addr     *src = NULL, *dst = NULL;
2577                 struct pf_addr_wrap     *xdst = NULL;
2578
2579                 if (r->action == PF_BINAT && direction == PF_IN) {
2580                         src = &r->dst;
2581                         if (r->rpool.cur != NULL)
2582                                 xdst = &r->rpool.cur->addr;
2583                 } else {
2584                         src = &r->src;
2585                         dst = &r->dst;
2586                 }
2587
2588                 r->evaluations++;
2589                 if (pfi_kif_match(r->kif, kif) == r->ifnot)
2590                         r = r->skip[PF_SKIP_IFP].ptr;
2591                 else if (r->direction && r->direction != direction)
2592                         r = r->skip[PF_SKIP_DIR].ptr;
2593                 else if (r->af && r->af != pd->af)
2594                         r = r->skip[PF_SKIP_AF].ptr;
2595                 else if (r->proto && r->proto != pd->proto)
2596                         r = r->skip[PF_SKIP_PROTO].ptr;
2597                 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af,
2598                     src->neg, kif))
2599                         r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR :
2600                             PF_SKIP_DST_ADDR].ptr;
2601                 else if (src->port_op && !pf_match_port(src->port_op,
2602                     src->port[0], src->port[1], sport))
2603                         r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT :
2604                             PF_SKIP_DST_PORT].ptr;
2605                 else if (dst != NULL &&
2606                     PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->neg, NULL))
2607                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
2608                 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af,
2609                     0, NULL))
2610                         r = TAILQ_NEXT(r, entries);
2611                 else if (dst != NULL && dst->port_op &&
2612                     !pf_match_port(dst->port_op, dst->port[0],
2613                     dst->port[1], dport))
2614                         r = r->skip[PF_SKIP_DST_PORT].ptr;
2615                 else if (r->match_tag && !pf_match_tag(m, r, &tag))
2616                         r = TAILQ_NEXT(r, entries);
2617                 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto !=
2618                     IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m,
2619                     off, pd->hdr.tcp), r->os_fingerprint)))
2620                         r = TAILQ_NEXT(r, entries);
2621                 else {
2622                         if (r->tag)
2623                                 tag = r->tag;
2624                         if (r->rtableid >= 0)
2625                                 rtableid = r->rtableid;
2626                         if (r->anchor == NULL) {
2627                                 rm = r;
2628                         } else
2629                                 pf_step_into_anchor(&asd, &ruleset, rs_num,
2630                                     &r, NULL, NULL);
2631                 }
2632                 if (r == NULL)
2633                         pf_step_out_of_anchor(&asd, &ruleset, rs_num, &r,
2634                             NULL, NULL);
2635         }
2636         if (pf_tag_packet(m, tag, rtableid))
2637                 return (NULL);
2638         if (rm != NULL && (rm->action == PF_NONAT ||
2639             rm->action == PF_NORDR || rm->action == PF_NOBINAT))
2640                 return (NULL);
2641         return (rm);
2642 }
2643
2644 struct pf_rule *
2645 pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction,
2646     struct pfi_kif *kif, struct pf_src_node **sn,
2647     struct pf_state_key **skw, struct pf_state_key **sks,
2648     struct pf_state_key **skp, struct pf_state_key **nkp,
2649     struct pf_addr *saddr, struct pf_addr *daddr,
2650     u_int16_t sport, u_int16_t dport)
2651 {
2652         struct pf_rule  *r = NULL;
2653
2654
2655         if (direction == PF_OUT) {
2656                 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2657                     sport, daddr, dport, PF_RULESET_BINAT);
2658                 if (r == NULL)
2659                         r = pf_match_translation(pd, m, off, direction, kif,
2660                             saddr, sport, daddr, dport, PF_RULESET_NAT);
2661         } else {
2662                 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2663                     sport, daddr, dport, PF_RULESET_RDR);
2664                 if (r == NULL)
2665                         r = pf_match_translation(pd, m, off, direction, kif,
2666                             saddr, sport, daddr, dport, PF_RULESET_BINAT);
2667         }
2668
2669         if (r != NULL) {
2670                 struct pf_addr  *naddr;
2671                 u_int16_t       *nport;
2672
2673                 if (pf_state_key_setup(pd, r, skw, sks, skp, nkp,
2674                     saddr, daddr, sport, dport))
2675                         return r;
2676
2677                 /* XXX We only modify one side for now. */
2678                 naddr = &(*nkp)->addr[1];
2679                 nport = &(*nkp)->port[1];
2680
2681                 /*
2682                  * NOTE: Currently all translations will clear
2683                  *       BRIDGE_MBUF_TAGGED, telling the bridge to
2684                  *       ignore the original input encapsulation.
2685                  */
2686                 switch (r->action) {
2687                 case PF_NONAT:
2688                 case PF_NOBINAT:
2689                 case PF_NORDR:
2690                         return (NULL);
2691                 case PF_NAT:
2692                         m->m_pkthdr.fw_flags &= ~BRIDGE_MBUF_TAGGED;
2693                         if (pf_get_sport(pd->af, pd->proto, r, saddr,
2694                             daddr, dport, naddr, nport, r->rpool.proxy_port[0],
2695                             r->rpool.proxy_port[1], sn)) {
2696                                 DPFPRINTF(PF_DEBUG_MISC,
2697                                     ("pf: NAT proxy port allocation "
2698                                     "(%u-%u) failed\n",
2699                                     r->rpool.proxy_port[0],
2700                                     r->rpool.proxy_port[1]));
2701                                 return (NULL);
2702                         }
2703                         break;
2704                 case PF_BINAT:
2705                         m->m_pkthdr.fw_flags &= ~BRIDGE_MBUF_TAGGED;
2706                         switch (direction) {
2707                         case PF_OUT:
2708                                 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){
2709                                         switch (pd->af) {
2710 #ifdef INET
2711                                         case AF_INET:
2712                                                 if (r->rpool.cur->addr.p.dyn->
2713                                                     pfid_acnt4 < 1)
2714                                                         return (NULL);
2715                                                 PF_POOLMASK(naddr,
2716                                                     &r->rpool.cur->addr.p.dyn->
2717                                                     pfid_addr4,
2718                                                     &r->rpool.cur->addr.p.dyn->
2719                                                     pfid_mask4,
2720                                                     saddr, AF_INET);
2721                                                 break;
2722 #endif /* INET */
2723 #ifdef INET6
2724                                         case AF_INET6:
2725                                                 if (r->rpool.cur->addr.p.dyn->
2726                                                     pfid_acnt6 < 1)
2727                                                         return (NULL);
2728                                                 PF_POOLMASK(naddr,
2729                                                     &r->rpool.cur->addr.p.dyn->
2730                                                     pfid_addr6,
2731                                                     &r->rpool.cur->addr.p.dyn->
2732                                                     pfid_mask6,
2733                                                     saddr, AF_INET6);
2734                                                 break;
2735 #endif /* INET6 */
2736                                         }
2737                                 } else
2738                                         PF_POOLMASK(naddr,
2739                                             &r->rpool.cur->addr.v.a.addr,
2740                                             &r->rpool.cur->addr.v.a.mask,
2741                                             saddr, pd->af);
2742                                 break;
2743                         case PF_IN:
2744                                 if (r->src.addr.type == PF_ADDR_DYNIFTL) {
2745                                         switch (pd->af) {
2746 #ifdef INET
2747                                         case AF_INET:
2748                                                 if (r->src.addr.p.dyn->
2749                                                     pfid_acnt4 < 1)
2750                                                         return (NULL);
2751                                                 PF_POOLMASK(naddr,
2752                                                     &r->src.addr.p.dyn->
2753                                                     pfid_addr4,
2754                                                     &r->src.addr.p.dyn->
2755                                                     pfid_mask4,
2756                                                     daddr, AF_INET);
2757                                                 break;
2758 #endif /* INET */
2759 #ifdef INET6
2760                                         case AF_INET6:
2761                                                 if (r->src.addr.p.dyn->
2762                                                     pfid_acnt6 < 1)
2763                                                         return (NULL);
2764                                                 PF_POOLMASK(naddr,
2765                                                     &r->src.addr.p.dyn->
2766                                                     pfid_addr6,
2767                                                     &r->src.addr.p.dyn->
2768                                                     pfid_mask6,
2769                                                     daddr, AF_INET6);
2770                                                 break;
2771 #endif /* INET6 */
2772                                         }
2773                                 } else
2774                                         PF_POOLMASK(naddr,
2775                                             &r->src.addr.v.a.addr,
2776                                             &r->src.addr.v.a.mask, daddr,
2777                                             pd->af);
2778                                 break;
2779                         }
2780                         break;
2781                 case PF_RDR: {
2782                         m->m_pkthdr.fw_flags &= ~BRIDGE_MBUF_TAGGED;
2783                         if (pf_map_addr(pd->af, r, saddr, naddr, NULL, sn))
2784                                 return (NULL);
2785                         if ((r->rpool.opts & PF_POOL_TYPEMASK) ==
2786                             PF_POOL_BITMASK)
2787                                 PF_POOLMASK(naddr, naddr,
2788                                     &r->rpool.cur->addr.v.a.mask, daddr,
2789                                     pd->af);
2790
2791                         if (r->rpool.proxy_port[1]) {
2792                                 u_int32_t       tmp_nport;
2793
2794                                 tmp_nport = ((ntohs(dport) -
2795                                     ntohs(r->dst.port[0])) %
2796                                     (r->rpool.proxy_port[1] -
2797                                     r->rpool.proxy_port[0] + 1)) +
2798                                     r->rpool.proxy_port[0];
2799
2800                                 /* wrap around if necessary */
2801                                 if (tmp_nport > 65535)
2802                                         tmp_nport -= 65535;
2803                                 *nport = htons((u_int16_t)tmp_nport);
2804                         } else if (r->rpool.proxy_port[0])
2805                                 *nport = htons(r->rpool.proxy_port[0]);
2806                         break;
2807                 }
2808                 default:
2809                         return (NULL);
2810                 }
2811         }
2812
2813         return (r);
2814 }
2815
2816 #ifdef SMP
2817 struct netmsg_hashlookup {
2818         struct netmsg_base      base;
2819         struct inpcb            **nm_pinp;
2820         struct inpcbinfo        *nm_pcbinfo;
2821         struct pf_addr          *nm_saddr;
2822         struct pf_addr          *nm_daddr;
2823         uint16_t                nm_sport;
2824         uint16_t                nm_dport;
2825         sa_family_t             nm_af;
2826 };
2827
2828 #ifdef PF_SOCKET_LOOKUP_DOMSG
2829 static void
2830 in_pcblookup_hash_handler(netmsg_t msg)
2831 {
2832         struct netmsg_hashlookup *rmsg = (struct netmsg_hashlookup *)msg;
2833
2834         if (rmsg->nm_af == AF_INET)
2835                 *rmsg->nm_pinp = in_pcblookup_hash(rmsg->nm_pcbinfo,
2836                     rmsg->nm_saddr->v4, rmsg->nm_sport, rmsg->nm_daddr->v4,
2837                     rmsg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2838 #ifdef INET6
2839         else
2840                 *rmsg->nm_pinp = in6_pcblookup_hash(rmsg->nm_pcbinfo,
2841                     &rmsg->nm_saddr->v6, rmsg->nm_sport, &rmsg->nm_daddr->v6,
2842                     rmsg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2843 #endif /* INET6 */
2844         lwkt_replymsg(&rmsg->base.lmsg, 0);
2845 }
2846 #endif  /* PF_SOCKET_LOOKUP_DOMSG */
2847
2848 #endif /* SMP */
2849
2850 int
2851 pf_socket_lookup(int direction, struct pf_pdesc *pd)
2852 {
2853         struct pf_addr          *saddr, *daddr;
2854         u_int16_t                sport, dport;
2855         struct inpcbinfo        *pi;
2856         struct inpcb            *inp;
2857 #ifdef SMP
2858         struct netmsg_hashlookup *msg = NULL;
2859 #ifdef PF_SOCKET_LOOKUP_DOMSG
2860         struct netmsg_hashlookup msg0;
2861 #endif
2862 #endif
2863         int                      pi_cpu = 0;
2864
2865         if (pd == NULL)
2866                 return (-1);
2867         pd->lookup.uid = UID_MAX;
2868         pd->lookup.gid = GID_MAX;
2869         pd->lookup.pid = NO_PID;
2870         if (direction == PF_IN) {
2871                 saddr = pd->src;
2872                 daddr = pd->dst;
2873         } else {
2874                 saddr = pd->dst;
2875                 daddr = pd->src;
2876         }
2877         switch (pd->proto) {
2878         case IPPROTO_TCP:
2879                 if (pd->hdr.tcp == NULL)
2880                         return (-1);
2881                 sport = pd->hdr.tcp->th_sport;
2882                 dport = pd->hdr.tcp->th_dport;
2883
2884                 pi_cpu = tcp_addrcpu(saddr->v4.s_addr, sport, daddr->v4.s_addr, dport);
2885                 pi = &tcbinfo[pi_cpu];
2886 #ifdef SMP
2887                 /*
2888                  * Our netstack runs lockless on MP systems
2889                  * (only for TCP connections at the moment).
2890                  *
2891                  * As we are not allowed to read another CPU's tcbinfo,
2892                  * we have to ask that CPU via remote call to search the
2893                  * table for us.
2894                  *
2895                  * Prepare a msg iff data belongs to another CPU.
2896                  */
2897                 if (pi_cpu != mycpu->gd_cpuid) {
2898 #ifdef PF_SOCKET_LOOKUP_DOMSG
2899                         /*
2900                          * NOTE:
2901                          *
2902                          * Following lwkt_domsg() is dangerous and could
2903                          * lockup the network system, e.g.
2904                          *
2905                          * On 2 CPU system:
2906                          * netisr0 domsg to netisr1 (due to lookup)
2907                          * netisr1 domsg to netisr0 (due to lookup)
2908                          *
2909                          * We simply return -1 here, since we are probably
2910                          * called before NAT, so the TCP packet should
2911                          * already be on the correct CPU.
2912                          */
2913                         msg = &msg0;
2914                         netmsg_init(&msg->base, NULL, &curthread->td_msgport,
2915                                     0, in_pcblookup_hash_handler);
2916                         msg->nm_pinp = &inp;
2917                         msg->nm_pcbinfo = pi;
2918                         msg->nm_saddr = saddr;
2919                         msg->nm_sport = sport;
2920                         msg->nm_daddr = daddr;
2921                         msg->nm_dport = dport;
2922                         msg->nm_af = pd->af;
2923 #else   /* !PF_SOCKET_LOOKUP_DOMSG */
2924                         kprintf("pf_socket_lookup: tcp packet not on the "
2925                                 "correct cpu %d, cur cpu %d\n",
2926                                 pi_cpu, mycpuid);
2927                         print_backtrace(-1);
2928                         return -1;
2929 #endif  /* PF_SOCKET_LOOKUP_DOMSG */
2930                 }
2931 #endif /* SMP */
2932                 break;
2933         case IPPROTO_UDP:
2934                 if (pd->hdr.udp == NULL)
2935                         return (-1);
2936                 sport = pd->hdr.udp->uh_sport;
2937                 dport = pd->hdr.udp->uh_dport;
2938                 pi = &udbinfo;
2939                 break;
2940         default:
2941                 return (-1);
2942         }
2943         if (direction != PF_IN) {
2944                 u_int16_t       p;
2945
2946                 p = sport;
2947                 sport = dport;
2948                 dport = p;
2949         }
2950         switch (pd->af) {
2951 #ifdef INET6
2952         case AF_INET6:
2953 #ifdef SMP
2954                 /*
2955                  * Query other CPU, second part
2956                  *
2957                  * msg only gets initialized when:
2958                  * 1) packet is TCP
2959                  * 2) the info belongs to another CPU
2960                  *
2961                  * Use some switch/case magic to avoid code duplication.
2962                  */
2963                 if (msg == NULL)
2964 #endif /* SMP */
2965                 {
2966                         inp = in6_pcblookup_hash(pi, &saddr->v6, sport,
2967                             &daddr->v6, dport, INPLOOKUP_WILDCARD, NULL);
2968
2969                         if (inp == NULL)
2970                                 return (-1);
2971                         break;
2972                 }
2973                 /* FALLTHROUGH if SMP and on other CPU */
2974 #endif /* INET6 */
2975         case AF_INET:
2976 #ifdef SMP
2977                 if (msg != NULL) {
2978                         lwkt_domsg(cpu_portfn(pi_cpu),
2979                                      &msg->base.lmsg, 0);
2980                 } else
2981 #endif /* SMP */
2982                 {
2983                         inp = in_pcblookup_hash(pi, saddr->v4, sport, daddr->v4,
2984                             dport, INPLOOKUP_WILDCARD, NULL);
2985                 }
2986                 if (inp == NULL)
2987                         return (-1);
2988                 break;
2989
2990         default:
2991                 return (-1);
2992         }
2993         pd->lookup.uid = inp->inp_socket->so_cred->cr_uid;
2994         pd->lookup.gid = inp->inp_socket->so_cred->cr_groups[0];
2995         return (1);
2996 }
2997
2998 u_int8_t
2999 pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
3000 {
3001         int              hlen;
3002         u_int8_t         hdr[60];
3003         u_int8_t        *opt, optlen;
3004         u_int8_t         wscale = 0;
3005
3006         hlen = th_off << 2;             /* hlen <= sizeof(hdr) */
3007         if (hlen <= sizeof(struct tcphdr))
3008                 return (0);
3009         if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
3010                 return (0);
3011         opt = hdr + sizeof(struct tcphdr);
3012         hlen -= sizeof(struct tcphdr);
3013         while (hlen >= 3) {
3014                 switch (*opt) {
3015                 case TCPOPT_EOL:
3016                 case TCPOPT_NOP:
3017                         ++opt;
3018                         --hlen;
3019                         break;
3020                 case TCPOPT_WINDOW:
3021                         wscale = opt[2];
3022                         if (wscale > TCP_MAX_WINSHIFT)
3023                                 wscale = TCP_MAX_WINSHIFT;
3024                         wscale |= PF_WSCALE_FLAG;
3025                         /* FALLTHROUGH */
3026                 default:
3027                         optlen = opt[1];
3028                         if (optlen < 2)
3029                                 optlen = 2;
3030                         hlen -= optlen;
3031                         opt += optlen;
3032                         break;
3033                 }
3034         }
3035         return (wscale);
3036 }
3037
3038 u_int16_t
3039 pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
3040 {
3041         int              hlen;
3042         u_int8_t         hdr[60];
3043         u_int8_t        *opt, optlen;
3044         u_int16_t        mss = tcp_mssdflt;
3045
3046         hlen = th_off << 2;     /* hlen <= sizeof(hdr) */
3047         if (hlen <= sizeof(struct tcphdr))
3048                 return (0);
3049         if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
3050                 return (0);
3051         opt = hdr + sizeof(struct tcphdr);
3052         hlen -= sizeof(struct tcphdr);
3053         while (hlen >= TCPOLEN_MAXSEG) {
3054                 switch (*opt) {
3055                 case TCPOPT_EOL:
3056                 case TCPOPT_NOP:
3057                         ++opt;
3058                         --hlen;
3059                         break;
3060                 case TCPOPT_MAXSEG:
3061                         bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2);
3062                         /* FALLTHROUGH */
3063                 default:
3064                         optlen = opt[1];
3065                         if (optlen < 2)
3066                                 optlen = 2;
3067                         hlen -= optlen;
3068                         opt += optlen;
3069                         break;
3070                 }
3071         }
3072         return (mss);
3073 }
3074
3075 u_int16_t
3076 pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer)
3077 {
3078 #ifdef INET
3079         struct sockaddr_in      *dst;
3080         struct route             ro;
3081 #endif /* INET */
3082 #ifdef INET6
3083         struct sockaddr_in6     *dst6;
3084         struct route_in6         ro6;
3085 #endif /* INET6 */
3086         struct rtentry          *rt = NULL;
3087         int                      hlen = 0;
3088         u_int16_t                mss = tcp_mssdflt;
3089
3090         switch (af) {
3091 #ifdef INET
3092         case AF_INET:
3093                 hlen = sizeof(struct ip);
3094                 bzero(&ro, sizeof(ro));
3095                 dst = (struct sockaddr_in *)&ro.ro_dst;
3096                 dst->sin_family = AF_INET;
3097                 dst->sin_len = sizeof(*dst);
3098                 dst->sin_addr = addr->v4;
3099                 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
3100                 rt = ro.ro_rt;
3101                 break;
3102 #endif /* INET */
3103 #ifdef INET6
3104         case AF_INET6:
3105                 hlen = sizeof(struct ip6_hdr);
3106                 bzero(&ro6, sizeof(ro6));
3107                 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst;
3108                 dst6->sin6_family = AF_INET6;
3109                 dst6->sin6_len = sizeof(*dst6);
3110                 dst6->sin6_addr = addr->v6;
3111                 rtalloc_ign((struct route *)&ro6, (RTF_CLONING | RTF_PRCLONING));
3112                 rt = ro6.ro_rt;
3113                 break;
3114 #endif /* INET6 */
3115         }
3116
3117         if (rt && rt->rt_ifp) {
3118                 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr);
3119                 mss = max(tcp_mssdflt, mss);
3120                 RTFREE(rt);
3121         }
3122         mss = min(mss, offer);
3123         mss = max(mss, 64);             /* sanity - at least max opt space */
3124         return (mss);
3125 }
3126
3127 void
3128 pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr)
3129 {
3130         struct pf_rule *r = s->rule.ptr;
3131
3132         s->rt_kif = NULL;
3133         if (!r->rt || r->rt == PF_FASTROUTE)
3134                 return;
3135         switch (s->key[PF_SK_WIRE]->af) {
3136 #ifdef INET
3137         case AF_INET:
3138                 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL,
3139                     &s->nat_src_node);
3140                 s->rt_kif = r->rpool.cur->kif;
3141                 break;
3142 #endif /* INET */
3143 #ifdef INET6
3144         case AF_INET6:
3145                 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL,
3146                     &s->nat_src_node);
3147                 s->rt_kif = r->rpool.cur->kif;
3148                 break;
3149 #endif /* INET6 */
3150         }
3151 }
3152
3153 u_int32_t
3154 pf_tcp_iss(struct pf_pdesc *pd)
3155 {
3156         MD5_CTX ctx;
3157         u_int32_t digest[4];
3158
3159         if (pf_tcp_secret_init == 0) {
3160                 karc4rand(pf_tcp_secret, sizeof(pf_tcp_secret));
3161                 MD5Init(&pf_tcp_secret_ctx);
3162                 MD5Update(&pf_tcp_secret_ctx, pf_tcp_secret,
3163                     sizeof(pf_tcp_secret));
3164                 pf_tcp_secret_init = 1;
3165         }
3166         ctx = pf_tcp_secret_ctx;
3167
3168         MD5Update(&ctx, (char *)&pd->hdr.tcp->th_sport, sizeof(u_short));
3169         MD5Update(&ctx, (char *)&pd->hdr.tcp->th_dport, sizeof(u_short));
3170         if (pd->af == AF_INET6) {
3171                 MD5Update(&ctx, (char *)&pd->src->v6, sizeof(struct in6_addr));
3172                 MD5Update(&ctx, (char *)&pd->dst->v6, sizeof(struct in6_addr));
3173         } else {
3174                 MD5Update(&ctx, (char *)&pd->src->v4, sizeof(struct in_addr));
3175                 MD5Update(&ctx, (char *)&pd->dst->v4, sizeof(struct in_addr));
3176         }
3177         MD5Final((u_char *)digest, &ctx);
3178         pf_tcp_iss_off += 4096;
3179         return (digest[0] + pd->hdr.tcp->th_seq + pf_tcp_iss_off);
3180 }
3181
3182 int
3183 pf_test_rule(struct pf_rule **rm, struct pf_state **sm, int direction,
3184     struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3185     struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
3186     struct ifqueue *ifq, struct inpcb *inp)
3187 {
3188         struct pf_rule          *nr = NULL;
3189         struct pf_addr          *saddr = pd->src, *daddr = pd->dst;
3190         sa_family_t              af = pd->af;
3191         struct pf_rule          *r, *a = NULL;
3192         struct pf_ruleset       *ruleset = NULL;
3193         struct pf_src_node      *nsn = NULL;
3194         struct tcphdr           *th = pd->hdr.tcp;
3195         struct pf_state_key     *skw = NULL, *sks = NULL;
3196         struct pf_state_key     *sk = NULL, *nk = NULL;
3197         u_short                  reason;
3198         int                      rewrite = 0, hdrlen = 0;
3199         int                      tag = -1, rtableid = -1;
3200         int                      asd = 0;
3201         int                      match = 0;
3202         int                      state_icmp = 0;
3203         u_int16_t                sport = 0, dport = 0;
3204         u_int16_t                nport = 0, bport = 0;
3205         u_int16_t                bproto_sum = 0, bip_sum = 0;
3206         u_int8_t                 icmptype = 0, icmpcode = 0;
3207
3208
3209         if (direction == PF_IN && pf_check_congestion(ifq)) {
3210                 REASON_SET(&reason, PFRES_CONGEST);
3211                 return (PF_DROP);
3212         }
3213
3214         if (inp != NULL)
3215                 pd->lookup.done = pf_socket_lookup(direction, pd);
3216         else if (debug_pfugidhack) {
3217                 DPFPRINTF(PF_DEBUG_MISC, ("pf: unlocked lookup\n"));
3218                 pd->lookup.done = pf_socket_lookup(direction, pd);
3219         }
3220
3221         switch (pd->proto) {
3222         case IPPROTO_TCP:
3223                 sport = th->th_sport;
3224                 dport = th->th_dport;
3225                 hdrlen = sizeof(*th);
3226                 break;
3227         case IPPROTO_UDP:
3228                 sport = pd->hdr.udp->uh_sport;
3229                 dport = pd->hdr.udp->uh_dport;
3230                 hdrlen = sizeof(*pd->hdr.udp);
3231                 break;
3232 #ifdef INET
3233         case IPPROTO_ICMP:
3234                 if (pd->af != AF_INET)
3235                         break;
3236                 sport = dport = pd->hdr.icmp->icmp_id;
3237                 hdrlen = sizeof(*pd->hdr.icmp);
3238                 icmptype = pd->hdr.icmp->icmp_type;
3239                 icmpcode = pd->hdr.icmp->icmp_code;
3240
3241                 if (icmptype == ICMP_UNREACH ||
3242                     icmptype == ICMP_SOURCEQUENCH ||
3243                     icmptype == ICMP_REDIRECT ||
3244                     icmptype == ICMP_TIMXCEED ||
3245                     icmptype == ICMP_PARAMPROB)
3246                         state_icmp++;
3247                 break;
3248 #endif /* INET */
3249 #ifdef INET6
3250         case IPPROTO_ICMPV6:
3251                 if (af != AF_INET6)
3252                         break;
3253                 sport = dport = pd->hdr.icmp6->icmp6_id;
3254                 hdrlen = sizeof(*pd->hdr.icmp6);
3255                 icmptype = pd->hdr.icmp6->icmp6_type;
3256                 icmpcode = pd->hdr.icmp6->icmp6_code;
3257
3258                 if (icmptype == ICMP6_DST_UNREACH ||
3259                     icmptype == ICMP6_PACKET_TOO_BIG ||
3260                     icmptype == ICMP6_TIME_EXCEEDED ||
3261                     icmptype == ICMP6_PARAM_PROB)
3262                         state_icmp++;
3263                 break;
3264 #endif /* INET6 */
3265         default:
3266                 sport = dport = hdrlen = 0;
3267                 break;
3268         }
3269
3270         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3271
3272         bport = nport = sport;
3273         /* check packet for BINAT/NAT/RDR */
3274         if ((nr = pf_get_translation(pd, m, off, direction, kif, &nsn,
3275             &skw, &sks, &sk, &nk, saddr, daddr, sport, dport)) != NULL) {
3276                 if (nk == NULL || sk == NULL) {
3277                         REASON_SET(&reason, PFRES_MEMORY);
3278                         goto cleanup;
3279                 }
3280
3281                 if (pd->ip_sum)
3282                         bip_sum = *pd->ip_sum;
3283
3284                 switch (pd->proto) {
3285                 case IPPROTO_TCP:
3286                         bproto_sum = th->th_sum;
3287                         pd->proto_sum = &th->th_sum;
3288
3289                         if (PF_ANEQ(saddr, &nk->addr[pd->sidx], af) ||
3290                             nk->port[pd->sidx] != sport) {
3291                                 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
3292                                     &th->th_sum, &nk->addr[pd->sidx],
3293                                     nk->port[pd->sidx], 0, af);
3294                                 pd->sport = &th->th_sport;
3295                                 sport = th->th_sport;
3296                         }
3297
3298                         if (PF_ANEQ(daddr, &nk->addr[pd->didx], af) ||
3299                             nk->port[pd->didx] != dport) {
3300                                 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
3301                                     &th->th_sum, &nk->addr[pd->didx],
3302                                     nk->port[pd->didx], 0, af);
3303                                 dport = th->th_dport;
3304                                 pd->dport = &th->th_dport;
3305                         }
3306                         rewrite++;
3307                         break;
3308                 case IPPROTO_UDP:
3309                         bproto_sum = pd->hdr.udp->uh_sum;
3310                         pd->proto_sum = &pd->hdr.udp->uh_sum;
3311
3312                         if (PF_ANEQ(saddr, &nk->addr[pd->sidx], af) ||
3313                             nk->port[pd->sidx] != sport) {
3314                                 pf_change_ap(saddr, &pd->hdr.udp->uh_sport,
3315                                     pd->ip_sum, &pd->hdr.udp->uh_sum,
3316                                     &nk->addr[pd->sidx],
3317                                     nk->port[pd->sidx], 1, af);
3318                                 sport = pd->hdr.udp->uh_sport;
3319                                 pd->sport = &pd->hdr.udp->uh_sport;
3320                         }
3321
3322                         if (PF_ANEQ(daddr, &nk->addr[pd->didx], af) ||
3323                             nk->port[pd->didx] != dport) {
3324                                 pf_change_ap(daddr, &pd->hdr.udp->uh_dport,
3325                                     pd->ip_sum, &pd->hdr.udp->uh_sum,
3326                                     &nk->addr[pd->didx],
3327                                     nk->port[pd->didx], 1, af);
3328                                 dport = pd->hdr.udp->uh_dport;
3329                                 pd->dport = &pd->hdr.udp->uh_dport;
3330                         }
3331                         rewrite++;
3332                         break;
3333 #ifdef INET
3334                 case IPPROTO_ICMP:
3335                         nk->port[0] = nk->port[1];
3336                         if (PF_ANEQ(saddr, &nk->addr[pd->sidx], AF_INET))
3337                                 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3338                                     nk->addr[pd->sidx].v4.s_addr, 0);
3339
3340                         if (PF_ANEQ(daddr, &nk->addr[pd->didx], AF_INET))
3341                                 pf_change_a(&daddr->v4.s_addr, pd->ip_sum,
3342                                     nk->addr[pd->didx].v4.s_addr, 0);
3343
3344                         if (nk->port[1] != pd->hdr.icmp->icmp_id) {
3345                                 pd->hdr.icmp->icmp_cksum = pf_cksum_fixup(
3346                                     pd->hdr.icmp->icmp_cksum, sport,
3347                                     nk->port[1], 0);
3348                                 pd->hdr.icmp->icmp_id = nk->port[1];
3349                                 pd->sport = &pd->hdr.icmp->icmp_id;
3350                         }
3351                         m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
3352                         break;
3353 #endif /* INET */
3354 #ifdef INET6
3355                 case IPPROTO_ICMPV6:
3356                         nk->port[0] = nk->port[1];
3357                         if (PF_ANEQ(saddr, &nk->addr[pd->sidx], AF_INET6))
3358                                 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum,
3359                                     &nk->addr[pd->sidx], 0);
3360
3361                         if (PF_ANEQ(daddr, &nk->addr[pd->didx], AF_INET6))
3362                                 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum,
3363                                     &nk->addr[pd->didx], 0);
3364                         rewrite++;
3365                         break;
3366 #endif /* INET */
3367                 default:
3368                         switch (af) {
3369 #ifdef INET
3370                         case AF_INET:
3371                                 if (PF_ANEQ(saddr,
3372                                     &nk->addr[pd->sidx], AF_INET))
3373                                         pf_change_a(&saddr->v4.s_addr,
3374                                             pd->ip_sum,
3375                                             nk->addr[pd->sidx].v4.s_addr, 0);
3376
3377                                 if (PF_ANEQ(daddr,
3378                                     &nk->addr[pd->didx], AF_INET))
3379                                         pf_change_a(&daddr->v4.s_addr,
3380                                             pd->ip_sum,
3381                                             nk->addr[pd->didx].v4.s_addr, 0);
3382                                 break;
3383 #endif /* INET */
3384 #ifdef INET6
3385                         case AF_INET6:
3386                                 if (PF_ANEQ(saddr,
3387                                     &nk->addr[pd->sidx], AF_INET6))
3388                                         PF_ACPY(saddr, &nk->addr[pd->sidx], af);
3389
3390                                 if (PF_ANEQ(daddr,
3391                                     &nk->addr[pd->didx], AF_INET6))
3392                                         PF_ACPY(saddr, &nk->addr[pd->didx], af);
3393                                 break;
3394 #endif /* INET */
3395                         }
3396                         break;
3397                 }
3398                 if (nr->natpass)
3399                         r = NULL;
3400                 pd->nat_rule = nr;
3401         }
3402
3403         while (r != NULL) {
3404                 r->evaluations++;
3405                 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3406                         r = r->skip[PF_SKIP_IFP].ptr;
3407                 else if (r->direction && r->direction != direction)
3408                         r = r->skip[PF_SKIP_DIR].ptr;
3409                 else if (r->af && r->af != af)
3410                         r = r->skip[PF_SKIP_AF].ptr;
3411                 else if (r->proto && r->proto != pd->proto)
3412                         r = r->skip[PF_SKIP_PROTO].ptr;
3413                 else if (PF_MISMATCHAW(&r->src.addr, saddr, af,
3414                     r->src.neg, kif))
3415                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3416                 /* tcp/udp only. port_op always 0 in other cases */
3417                 else if (r->src.port_op && !pf_match_port(r->src.port_op,
3418                     r->src.port[0], r->src.port[1], sport))
3419                         r = r->skip[PF_SKIP_SRC_PORT].ptr;
3420                 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af,
3421                     r->dst.neg, NULL))
3422                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
3423                 /* tcp/udp only. port_op always 0 in other cases */
3424                 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
3425                     r->dst.port[0], r->dst.port[1], dport))
3426                         r = r->skip[PF_SKIP_DST_PORT].ptr;
3427                 /* icmp only. type always 0 in other cases */
3428                 else if (r->type && r->type != icmptype + 1)
3429                         r = TAILQ_NEXT(r, entries);
3430                 /* icmp only. type always 0 in other cases */
3431                 else if (r->code && r->code != icmpcode + 1)
3432                         r = TAILQ_NEXT(r, entries);
3433                 else if (r->tos && !(r->tos == pd->tos))
3434                         r = TAILQ_NEXT(r, entries);
3435                 else if (r->rule_flag & PFRULE_FRAGMENT)
3436                         r = TAILQ_NEXT(r, entries);
3437                 else if (pd->proto == IPPROTO_TCP &&
3438                     (r->flagset & th->th_flags) != r->flags)
3439                         r = TAILQ_NEXT(r, entries);
3440                 /* tcp/udp only. uid.op always 0 in other cases */
3441                 else if (r->uid.op && (pd->lookup.done || (pd->lookup.done =
3442                     pf_socket_lookup(direction, pd), 1)) &&
3443                     !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
3444                     pd->lookup.uid))
3445                         r = TAILQ_NEXT(r, entries);
3446                 /* tcp/udp only. gid.op always 0 in other cases */
3447                 else if (r->gid.op && (pd->lookup.done || (pd->lookup.done =
3448                     pf_socket_lookup(direction, pd), 1)) &&
3449                     !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
3450                     pd->lookup.gid))
3451                         r = TAILQ_NEXT(r, entries);
3452                 else if (r->prob &&
3453                   r->prob <= karc4random())
3454                         r = TAILQ_NEXT(r, entries);
3455                 else if (r->match_tag && !pf_match_tag(m, r, &tag))
3456                         r = TAILQ_NEXT(r, entries);
3457                 else if (r->os_fingerprint != PF_OSFP_ANY &&
3458                     (pd->proto != IPPROTO_TCP || !pf_osfp_match(
3459                     pf_osfp_fingerprint(pd, m, off, th),
3460                     r->os_fingerprint)))
3461                         r = TAILQ_NEXT(r, entries);
3462                 else {
3463                         if (r->tag)
3464                                 tag = r->tag;
3465                         if (r->rtableid >= 0)
3466                                 rtableid = r->rtableid;
3467                         if (r->anchor == NULL) {
3468                                 match = 1;
3469                                 *rm = r;
3470                                 *am = a;
3471                                 *rsm = ruleset;
3472                                 if ((*rm)->quick)
3473                                         break;
3474                                 r = TAILQ_NEXT(r, entries);
3475                         } else
3476                                 pf_step_into_anchor(&asd, &ruleset,
3477                                     PF_RULESET_FILTER, &r, &a, &match);
3478                 }
3479                 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3480                     PF_RULESET_FILTER, &r, &a, &match))
3481                         break;
3482         }
3483         r = *rm;
3484         a = *am;
3485         ruleset = *rsm;
3486
3487         REASON_SET(&reason, PFRES_MATCH);
3488
3489         if (r->log || (nr != NULL && nr->log)) {
3490                 if (rewrite)
3491                         m_copyback(m, off, hdrlen, pd->hdr.any);
3492                 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr,
3493                     a, ruleset, pd);
3494         }
3495
3496         if ((r->action == PF_DROP) &&
3497             ((r->rule_flag & PFRULE_RETURNRST) ||
3498             (r->rule_flag & PFRULE_RETURNICMP) ||
3499             (r->rule_flag & PFRULE_RETURN))) {
3500                 /* undo NAT changes, if they have taken place */
3501                 if (nr != NULL) {
3502                         PF_ACPY(saddr, &sk->addr[pd->sidx], af);
3503                         PF_ACPY(daddr, &sk->addr[pd->didx], af);
3504                         if (pd->sport)
3505                                 *pd->sport = sk->port[pd->sidx];
3506                         if (pd->dport)
3507                                 *pd->dport = sk->port[pd->didx];
3508                         if (pd->proto_sum)
3509                                 *pd->proto_sum = bproto_sum;
3510                         if (pd->ip_sum)
3511                                 *pd->ip_sum = bip_sum;
3512                         m_copyback(m, off, hdrlen, pd->hdr.any);
3513                 }
3514                 if (pd->proto == IPPROTO_TCP &&
3515                     ((r->rule_flag & PFRULE_RETURNRST) ||
3516                     (r->rule_flag & PFRULE_RETURN)) &&
3517                     !(th->th_flags & TH_RST)) {
3518                         u_int32_t        ack = ntohl(th->th_seq) + pd->p_len;
3519                         int              len = 0;
3520                         struct ip       *h4;
3521 #ifdef INET6
3522                         struct ip6_hdr  *h6;
3523 #endif
3524                         switch (af) {
3525                         case AF_INET:
3526                                 h4 = mtod(m, struct ip *);
3527                                 len = h4->ip_len - off;
3528                                 break;
3529 #ifdef INET6
3530                         case AF_INET6:
3531                                 h6 = mtod(m, struct ip6_hdr *);
3532                                 len = h6->ip6_plen - (off - sizeof(*h6));
3533                                 break;
3534 #endif
3535                         }
3536
3537                         if (pf_check_proto_cksum(m, off, len, IPPROTO_TCP, af))
3538                                 REASON_SET(&reason, PFRES_PROTCKSUM);
3539                         else {
3540                                 if (th->th_flags & TH_SYN)
3541                                         ack++;
3542                                 if (th->th_flags & TH_FIN)
3543                                         ack++;
3544                                 pf_send_tcp(r, af, pd->dst,
3545                                     pd->src, th->th_dport, th->th_sport,
3546                                     ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0,
3547                                     r->return_ttl, 1, 0, pd->eh, kif->pfik_ifp);
3548                         }
3549                 } else if (pd->proto != IPPROTO_ICMP && af == AF_INET &&
3550                     r->return_icmp)
3551                         pf_send_icmp(m, r->return_icmp >> 8,
3552                             r->return_icmp & 255, af, r);
3553                 else if (pd->proto != IPPROTO_ICMPV6 && af == AF_INET6 &&
3554                     r->return_icmp6)
3555                         pf_send_icmp(m, r->return_icmp6 >> 8,
3556                             r->return_icmp6 & 255, af, r);
3557         }
3558
3559         if (r->action == PF_DROP)
3560                 goto cleanup;
3561
3562         if (pf_tag_packet(m, tag, rtableid)) {
3563                 REASON_SET(&reason, PFRES_MEMORY);
3564                 goto cleanup;
3565         }
3566
3567         if (!state_icmp && (r->keep_state || nr != NULL ||
3568             (pd->flags & PFDESC_TCP_NORM))) {
3569                 int action;
3570                 action = pf_create_state(r, nr, a, pd, nsn, skw, sks, nk, sk, m,
3571                     off, sport, dport, &rewrite, kif, sm, tag, bproto_sum,
3572                     bip_sum, hdrlen);
3573                 if (action != PF_PASS)
3574                         return (action);
3575         }
3576
3577         /* copy back packet headers if we performed NAT operations */
3578         if (rewrite)
3579                 m_copyback(m, off, hdrlen, pd->hdr.any);
3580
3581         return (PF_PASS);
3582
3583 cleanup:
3584         if (sk != NULL)
3585                 kfree(sk, M_PFSTATEKEYPL);
3586         if (nk != NULL)
3587                 kfree(nk, M_PFSTATEKEYPL);
3588         return (PF_DROP);
3589 }
3590
3591 static __inline int
3592 pf_create_state(struct pf_rule *r, struct pf_rule *nr, struct pf_rule *a,
3593     struct pf_pdesc *pd, struct pf_src_node *nsn, struct pf_state_key *skw,
3594     struct pf_state_key *sks, struct pf_state_key *nk, struct pf_state_key *sk,
3595     struct mbuf *m, int off, u_int16_t sport, u_int16_t dport, int *rewrite,
3596     struct pfi_kif *kif, struct pf_state **sm, int tag, u_int16_t bproto_sum,
3597     u_int16_t bip_sum, int hdrlen)
3598 {
3599         struct pf_state         *s = NULL;
3600         struct pf_src_node      *sn = NULL;
3601         struct tcphdr           *th = pd->hdr.tcp;
3602         u_int16_t                mss = tcp_mssdflt;
3603         u_short                  reason;
3604
3605         /* check maximums */
3606         if (r->max_states && (r->states_cur >= r->max_states)) {
3607                 pf_status.lcounters[LCNT_STATES]++;
3608                 REASON_SET(&reason, PFRES_MAXSTATES);
3609                 return (PF_DROP);
3610         }
3611         /* src node for filter rule */
3612         if ((r->rule_flag & PFRULE_SRCTRACK ||
3613             r->rpool.opts & PF_POOL_STICKYADDR) &&
3614             pf_insert_src_node(&sn, r, pd->src, pd->af) != 0) {
3615                 REASON_SET(&reason, PFRES_SRCLIMIT);
3616                 goto csfailed;
3617         }
3618         /* src node for translation rule */
3619         if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3620             pf_insert_src_node(&nsn, nr, &sk->addr[pd->sidx], pd->af)) {
3621                 REASON_SET(&reason, PFRES_SRCLIMIT);
3622                 goto csfailed;
3623         }
3624         s = kmalloc(sizeof(struct pf_state), M_PFSTATEPL, M_NOWAIT|M_ZERO);
3625         if (s == NULL) {
3626                 REASON_SET(&reason, PFRES_MEMORY);
3627                 goto csfailed;
3628         }
3629         s->id = 0; /* XXX Do we really need that? not in OpenBSD */
3630         s->creatorid = 0;
3631         s->rule.ptr = r;
3632         s->nat_rule.ptr = nr;
3633         s->anchor.ptr = a;
3634         STATE_INC_COUNTERS(s);
3635         if (r->allow_opts)
3636                 s->state_flags |= PFSTATE_ALLOWOPTS;
3637         if (r->rule_flag & PFRULE_STATESLOPPY)
3638                 s->state_flags |= PFSTATE_SLOPPY;
3639         s->log = r->log & PF_LOG_ALL;
3640         if (nr != NULL)
3641                 s->log |= nr->log & PF_LOG_ALL;
3642         switch (pd->proto) {
3643         case IPPROTO_TCP:
3644                 s->src.seqlo = ntohl(th->th_seq);
3645                 s->src.seqhi = s->src.seqlo + pd->p_len + 1;
3646                 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
3647                     r->keep_state == PF_STATE_MODULATE) {
3648                         /* Generate sequence number modulator */
3649                         if ((s->src.seqdiff = pf_tcp_iss(pd) - s->src.seqlo) ==
3650                             0)
3651                                 s->src.seqdiff = 1;
3652                         pf_change_a(&th->th_seq, &th->th_sum,
3653                             htonl(s->src.seqlo + s->src.seqdiff), 0);
3654                         *rewrite = 1;
3655                 } else
3656                         s->src.seqdiff = 0;
3657                 if (th->th_flags & TH_SYN) {
3658                         s->src.seqhi++;
3659                         s->src.wscale = pf_get_wscale(m, off,
3660                             th->th_off, pd->af);
3661                 }
3662                 s->src.max_win = MAX(ntohs(th->th_win), 1);
3663                 if (s->src.wscale & PF_WSCALE_MASK) {
3664                         /* Remove scale factor from initial window */
3665                         int win = s->src.max_win;
3666                         win += 1 << (s->src.wscale & PF_WSCALE_MASK);
3667                         s->src.max_win = (win - 1) >>
3668                             (s->src.wscale & PF_WSCALE_MASK);
3669                 }
3670                 if (th->th_flags & TH_FIN)
3671                         s->src.seqhi++;
3672                 s->dst.seqhi = 1;
3673                 s->dst.max_win = 1;
3674                 s->src.state = TCPS_SYN_SENT;
3675                 s->dst.state = TCPS_CLOSED;
3676                 s->timeout = PFTM_TCP_FIRST_PACKET;
3677                 break;
3678         case IPPROTO_UDP:
3679                 s->src.state = PFUDPS_SINGLE;
3680                 s->dst.state = PFUDPS_NO_TRAFFIC;
3681                 s->timeout = PFTM_UDP_FIRST_PACKET;
3682                 break;
3683         case IPPROTO_ICMP:
3684 #ifdef INET6
3685         case IPPROTO_ICMPV6:
3686 #endif
3687                 s->timeout = PFTM_ICMP_FIRST_PACKET;
3688                 break;
3689         default:
3690                 s->src.state = PFOTHERS_SINGLE;
3691                 s->dst.state = PFOTHERS_NO_TRAFFIC;
3692                 s->timeout = PFTM_OTHER_FIRST_PACKET;
3693         }
3694
3695         s->creation = time_second;
3696         s->expire = time_second;
3697
3698         if (sn != NULL) {
3699                 s->src_node = sn;
3700                 s->src_node->states++;
3701         }
3702         if (nsn != NULL) {
3703                 /* XXX We only modify one side for now. */
3704                 PF_ACPY(&nsn->raddr, &nk->addr[1], pd->af);
3705                 s->nat_src_node = nsn;
3706                 s->nat_src_node->states++;
3707         }
3708         if (pd->proto == IPPROTO_TCP) {
3709                 if ((pd->flags & PFDESC_TCP_NORM) && pf_normalize_tcp_init(m,
3710                     off, pd, th, &s->src, &s->dst)) {
3711                         REASON_SET(&reason, PFRES_MEMORY);
3712                         pf_src_tree_remove_state(s);
3713                         STATE_DEC_COUNTERS(s);
3714                         kfree(s, M_PFSTATEPL);
3715                         return (PF_DROP);
3716                 }
3717                 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
3718                     pf_normalize_tcp_stateful(m, off, pd, &reason, th, s,
3719                     &s->src, &s->dst, rewrite)) {
3720                         /* This really shouldn't happen!!! */
3721                         DPFPRINTF(PF_DEBUG_URGENT,
3722                             ("pf_normalize_tcp_stateful failed on first pkt"));
3723                         pf_normalize_tcp_cleanup(s);
3724                         pf_src_tree_remove_state(s);
3725                         STATE_DEC_COUNTERS(s);
3726                         kfree(s, M_PFSTATEPL);
3727                         return (PF_DROP);
3728                 }
3729         }
3730         s->direction = pd->dir;
3731
3732         if (sk == NULL && pf_state_key_setup(pd, nr, &skw, &sks, &sk, &nk,
3733             pd->src, pd->dst, sport, dport))
3734                 goto csfailed;
3735
3736         if (pf_state_insert(BOUND_IFACE(r, kif), skw, sks, s)) {
3737                 if (pd->proto == IPPROTO_TCP)
3738                         pf_normalize_tcp_cleanup(s);
3739                 REASON_SET(&reason, PFRES_STATEINS);
3740                 pf_src_tree_remove_state(s);
3741                 STATE_DEC_COUNTERS(s);
3742                 kfree(s, M_PFSTATEPL);
3743                 return (PF_DROP);
3744         } else
3745                 *sm = s;
3746
3747         pf_set_rt_ifp(s, pd->src);      /* needs s->state_key set */
3748         if (tag > 0) {
3749                 pf_tag_ref(tag);
3750                 s->tag = tag;
3751         }
3752         if (pd->proto == IPPROTO_TCP && (th->th_flags & (TH_SYN|TH_ACK)) ==
3753             TH_SYN && r->keep_state == PF_STATE_SYNPROXY) {
3754                 s->src.state = PF_TCPS_PROXY_SRC;
3755                 /* undo NAT changes, if they have taken place */
3756                 if (nr != NULL) {
3757                         struct pf_state_key *skt = s->key[PF_SK_WIRE];
3758                         if (pd->dir == PF_OUT)
3759                                 skt = s->key[PF_SK_STACK];
3760                         PF_ACPY(pd->src, &skt->addr[pd->sidx], pd->af);
3761                         PF_ACPY(pd->dst, &skt->addr[pd->didx], pd->af);
3762                         if (pd->sport)
3763                                 *pd->sport = skt->port[pd->sidx];
3764                         if (pd->dport)
3765                                 *pd->dport = skt->port[pd->didx];
3766                         if (pd->proto_sum)
3767                                 *pd->proto_sum = bproto_sum;
3768                         if (pd->ip_sum)
3769                                 *pd->ip_sum = bip_sum;
3770                         m_copyback(m, off, hdrlen, pd->hdr.any);
3771                 }
3772                 s->src.seqhi = htonl(karc4random());
3773                 /* Find mss option */
3774                 mss = pf_get_mss(m, off, th->th_off, pd->af);
3775                 mss = pf_calc_mss(pd->src, pd->af, mss);
3776                 mss = pf_calc_mss(pd->dst, pd->af, mss);
3777                 s->src.mss = mss;
3778                 pf_send_tcp(r, pd->af, pd->dst, pd->src, th->th_dport,
3779                     th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1,
3780                     TH_SYN|TH_ACK, 0, s->src.mss, 0, 1, 0, NULL, NULL);
3781                 REASON_SET(&reason, PFRES_SYNPROXY);
3782                 return (PF_SYNPROXY_DROP);
3783         }
3784
3785         return (PF_PASS);
3786
3787 csfailed:
3788         if (sk != NULL)
3789                 kfree(sk, M_PFSTATEKEYPL);
3790         if (nk != NULL)
3791                 kfree(nk, M_PFSTATEKEYPL);
3792
3793         if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3794                 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3795                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3796                 pf_status.src_nodes--;
3797                 kfree(sn, M_PFSRCTREEPL);
3798         }
3799         if (nsn != sn && nsn != NULL && nsn->states == 0 && nsn->expire == 0) {
3800                 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3801                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3802                 pf_status.src_nodes--;
3803                 kfree(nsn, M_PFSRCTREEPL);
3804         }
3805         return (PF_DROP);
3806 }
3807
3808 int
3809 pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif,
3810     struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am,
3811     struct pf_ruleset **rsm)
3812 {
3813         struct pf_rule          *r, *a = NULL;
3814         struct pf_ruleset       *ruleset = NULL;
3815         sa_family_t              af = pd->af;
3816         u_short                  reason;
3817         int                      tag = -1;
3818         int                      asd = 0;
3819         int                      match = 0;
3820
3821         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3822         while (r != NULL) {
3823                 r->evaluations++;
3824                 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3825                         r = r->skip[PF_SKIP_IFP].ptr;
3826                 else if (r->direction && r->direction != direction)
3827                         r = r->skip[PF_SKIP_DIR].ptr;
3828                 else if (r->af && r->af != af)
3829                         r = r->skip[PF_SKIP_AF].ptr;
3830                 else if (r->proto && r->proto != pd->proto)
3831                         r = r->skip[PF_SKIP_PROTO].ptr;
3832                 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af,
3833                     r->src.neg, kif))
3834                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3835                 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af,
3836                     r->dst.neg, NULL))
3837                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
3838                 else if (r->tos && !(r->tos == pd->tos))
3839                         r = TAILQ_NEXT(r, entries);
3840                 else if (r->os_fingerprint != PF_OSFP_ANY)
3841                         r = TAILQ_NEXT(r, entries);
3842                 else if (pd->proto == IPPROTO_UDP &&
3843                     (r->src.port_op || r->dst.port_op))
3844                         r = TAILQ_NEXT(r, entries);
3845                 else if (pd->proto == IPPROTO_TCP &&
3846                     (r->src.port_op || r->dst.port_op || r->flagset))
3847                         r = TAILQ_NEXT(r, entries);
3848                 else if ((pd->proto == IPPROTO_ICMP ||
3849                     pd->proto == IPPROTO_ICMPV6) &&
3850                     (r->type || r->code))
3851                         r = TAILQ_NEXT(r, entries);
3852                 else if (r->prob && r->prob <= karc4random())
3853                         r = TAILQ_NEXT(r, entries);
3854                 else if (r->match_tag && !pf_match_tag(m, r, &tag))
3855                         r = TAILQ_NEXT(r, entries);
3856                 else {
3857                         if (r->anchor == NULL) {
3858                                 match = 1;
3859                                 *rm = r;
3860                                 *am = a;
3861                                 *rsm = ruleset;
3862                                 if ((*rm)->quick)
3863                                         break;
3864                                 r = TAILQ_NEXT(r, entries);
3865                         } else
3866                                 pf_step_into_anchor(&asd, &ruleset,
3867                                     PF_RULESET_FILTER, &r, &a, &match);
3868                 }
3869                 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3870                     PF_RULESET_FILTER, &r, &a, &match))
3871                         break;
3872         }
3873         r = *rm;
3874         a = *am;
3875         ruleset = *rsm;
3876
3877         REASON_SET(&reason, PFRES_MATCH);
3878
3879         if (r->log)
3880                 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset,
3881                     pd);
3882
3883         if (r->action != PF_PASS)
3884                 return (PF_DROP);
3885
3886         if (pf_tag_packet(m, tag, -1)) {
3887                 REASON_SET(&reason, PFRES_MEMORY);
3888                 return (PF_DROP);
3889         }
3890
3891         return (PF_PASS);
3892 }
3893
3894 int
3895 pf_tcp_track_full(struct pf_state_peer *src, struct pf_state_peer *dst,
3896         struct pf_state **state, struct pfi_kif *kif, struct mbuf *m, int off,
3897         struct pf_pdesc *pd, u_short *reason, int *copyback)
3898 {
3899         struct tcphdr           *th = pd->hdr.tcp;
3900         u_int16_t                win = ntohs(th->th_win);
3901         u_int32_t                ack, end, seq, orig_seq;
3902         u_int8_t                 sws, dws;
3903         int                      ackskew;
3904
3905         if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) {
3906                 sws = src->wscale & PF_WSCALE_MASK;
3907                 dws = dst->wscale & PF_WSCALE_MASK;
3908         } else
3909                 sws = dws = 0;
3910
3911         /*
3912          * Sequence tracking algorithm from Guido van Rooij's paper:
3913          *   http://www.madison-gurkha.com/publications/tcp_filtering/
3914          *      tcp_filtering.ps
3915          */
3916
3917         orig_seq = seq = ntohl(th->th_seq);
3918         if (src->seqlo == 0) {
3919                 /* First packet from this end. Set its state */
3920
3921                 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) &&
3922                     src->scrub == NULL) {
3923                         if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) {
3924                                 REASON_SET(reason, PFRES_MEMORY);
3925                                 return (PF_DROP);
3926                         }
3927                 }
3928
3929                 /* Deferred generation of sequence number modulator */
3930                 if (dst->seqdiff && !src->seqdiff) {
3931                         /* use random iss for the TCP server */
3932                         while ((src->seqdiff = karc4random() - seq) == 0)
3933                                 ;
3934                         ack = ntohl(th->th_ack) - dst->seqdiff;
3935                         pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3936                             src->seqdiff), 0);
3937                         pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3938                         *copyback = 1;
3939                 } else {
3940                         ack = ntohl(th->th_ack);
3941                 }
3942
3943                 end = seq + pd->p_len;
3944                 if (th->th_flags & TH_SYN) {
3945                         end++;
3946                         (*state)->sync_flags |= PFSTATE_GOT_SYN2;
3947                         if (dst->wscale & PF_WSCALE_FLAG) {
3948                                 src->wscale = pf_get_wscale(m, off, th->th_off,
3949                                     pd->af);
3950                                 if (src->wscale & PF_WSCALE_FLAG) {
3951                                         /* Remove scale factor from initial
3952                                          * window */
3953                                         sws = src->wscale & PF_WSCALE_MASK;
3954                                         win = ((u_int32_t)win + (1 << sws) - 1)
3955                                             >> sws;
3956                                         dws = dst->wscale & PF_WSCALE_MASK;
3957                                 } else {
3958                                         /* fixup other window */
3959                                         dst->max_win <<= dst->wscale &
3960                                             PF_WSCALE_MASK;
3961                                         /* in case of a retrans SYN|ACK */
3962                                         dst->wscale = 0;
3963                                 }
3964                         }
3965                 }
3966                 if (th->th_flags & TH_FIN)
3967                         end++;
3968
3969                 src->seqlo = seq;
3970                 if (src->state < TCPS_SYN_SENT)
3971                         src->state = TCPS_SYN_SENT;
3972
3973                 /*
3974                  * May need to slide the window (seqhi may have been set by
3975                  * the crappy stack check or if we picked up the connection
3976                  * after establishment)
3977                  */
3978                 if (src->seqhi == 1 ||
3979                     SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi))
3980                         src->seqhi = end + MAX(1, dst->max_win << dws);
3981                 if (win > src->max_win)
3982                         src->max_win = win;
3983
3984         } else {
3985                 ack = ntohl(th->th_ack) - dst->seqdiff;
3986                 if (src->seqdiff) {
3987                         /* Modulate sequence numbers */
3988                         pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3989                             src->seqdiff), 0);
3990                         pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3991                         *copyback = 1;
3992                 }
3993                 end = seq + pd->p_len;
3994                 if (th->th_flags & TH_SYN)
3995                         end++;
3996                 if (th->th_flags & TH_FIN)
3997                         end++;
3998         }
3999
4000         if ((th->th_flags & TH_ACK) == 0) {
4001                 /* Let it pass through the ack skew check */
4002                 ack = dst->seqlo;
4003         } else if ((ack == 0 &&
4004             (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) ||
4005             /* broken tcp stacks do not set ack */
4006             (dst->state < TCPS_SYN_SENT)) {
4007                 /*
4008                  * Many stacks (ours included) will set the ACK number in an
4009                  * FIN|ACK if the SYN times out -- no sequence to ACK.
4010                  */
4011                 ack = dst->seqlo;
4012         }
4013
4014         if (seq == end) {
4015                 /* Ease sequencing restrictions on no data packets */
4016                 seq = src->seqlo;
4017                 end = seq;
4018         }
4019
4020         ackskew = dst->seqlo - ack;
4021
4022
4023         /*
4024          * Need to demodulate the sequence numbers in any TCP SACK options
4025          * (Selective ACK). We could optionally validate the SACK values
4026          * against the current ACK window, either forwards or backwards, but
4027          * I'm not confident that SACK has been implemented properly
4028          * everywhere. It wouldn't surprise me if several stacks accidently
4029          * SACK too far backwards of previously ACKed data. There really aren't
4030          * any security implications of bad SACKing unless the target stack
4031          * doesn't validate the option length correctly. Someone trying to
4032          * spoof into a TCP connection won't bother blindly sending SACK
4033          * options anyway.
4034          */
4035         if (dst->seqdiff && (th->th_off << 2) > sizeof(struct tcphdr)) {
4036                 if (pf_modulate_sack(m, off, pd, th, dst))
4037                         *copyback = 1;
4038         }
4039
4040
4041 #define MAXACKWINDOW (0xffff + 1500)    /* 1500 is an arbitrary fudge factor */
4042         if (SEQ_GEQ(src->seqhi, end) &&
4043             /* Last octet inside other's window space */
4044             SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) &&
4045             /* Retrans: not more than one window back */
4046             (ackskew >= -MAXACKWINDOW) &&
4047             /* Acking not more than one reassembled fragment backwards */
4048             (ackskew <= (MAXACKWINDOW << sws)) &&
4049             /* Acking not more than one window forward */
4050             ((th->th_flags & TH_RST) == 0 || orig_seq == src->seqlo ||
4051             (orig_seq == src->seqlo + 1) || (orig_seq + 1 == src->seqlo) ||
4052             (pd->flags & PFDESC_IP_REAS) == 0)) {
4053             /* Require an exact/+1 sequence match on resets when possible */
4054
4055                 if (dst->scrub || src->scrub) {
4056                         if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4057                             *state, src, dst, copyback))
4058                                 return (PF_DROP);
4059                 }
4060
4061                 /* update max window */
4062                 if (src->max_win < win)
4063                         src->max_win = win;
4064                 /* synchronize sequencing */
4065                 if (SEQ_GT(end, src->seqlo))
4066                         src->seqlo = end;
4067                 /* slide the window of what the other end can send */
4068                 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4069                         dst->seqhi = ack + MAX((win << sws), 1);
4070
4071
4072                 /* update states */
4073                 if (th->th_flags & TH_SYN)
4074                         if (src->state < TCPS_SYN_SENT)
4075                                 src->state = TCPS_SYN_SENT;
4076                 if (th->th_flags & TH_FIN)
4077                         if (src->state < TCPS_CLOSING)
4078                                 src->state = TCPS_CLOSING;
4079                 if (th->th_flags & TH_ACK) {
4080                         if (dst->state == TCPS_SYN_SENT) {
4081                                 dst->state = TCPS_ESTABLISHED;
4082                                 if (src->state == TCPS_ESTABLISHED &&
4083                                     (*state)->src_node != NULL &&
4084                                     pf_src_connlimit(state)) {
4085                                         REASON_SET(reason, PFRES_SRCLIMIT);
4086                                         return (PF_DROP);
4087                                 }
4088                         } else if (dst->state == TCPS_CLOSING)
4089                                 dst->state = TCPS_FIN_WAIT_2;
4090                 }
4091                 if (th->th_flags & TH_RST)
4092                         src->state = dst->state = TCPS_TIME_WAIT;
4093
4094                 /* update expire time */
4095                 (*state)->expire = time_second;
4096                 if (src->state >= TCPS_FIN_WAIT_2 &&
4097                     dst->state >= TCPS_FIN_WAIT_2)
4098                         (*state)->timeout = PFTM_TCP_CLOSED;
4099                 else if (src->state >= TCPS_CLOSING &&
4100                     dst->state >= TCPS_CLOSING)
4101                         (*state)->timeout = PFTM_TCP_FIN_WAIT;
4102                 else if (src->state < TCPS_ESTABLISHED ||
4103                     dst->state < TCPS_ESTABLISHED)
4104                         (*state)->timeout = PFTM_TCP_OPENING;
4105                 else if (src->state >= TCPS_CLOSING ||
4106                     dst->state >= TCPS_CLOSING)
4107                         (*state)->timeout = PFTM_TCP_CLOSING;
4108                 else
4109                         (*state)->timeout = PFTM_TCP_ESTABLISHED;
4110
4111                 /* Fall through to PASS packet */
4112
4113         } else if ((dst->state < TCPS_SYN_SENT ||
4114                 dst->state >= TCPS_FIN_WAIT_2 ||
4115                 src->state >= TCPS_FIN_WAIT_2) &&
4116             SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) &&
4117             /* Within a window forward of the originating packet */
4118             SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) {
4119             /* Within a window backward of the originating packet */
4120
4121                 /*
4122                  * This currently handles three situations:
4123                  *  1) Stupid stacks will shotgun SYNs before their peer
4124                  *     replies.
4125                  *  2) When PF catches an already established stream (the
4126                  *     firewall rebooted, the state table was flushed, routes
4127                  *     changed...)
4128                  *  3) Packets get funky immediately after the connection
4129                  *     closes (this should catch Solaris spurious ACK|FINs
4130                  *     that web servers like to spew after a close)
4131                  *
4132                  * This must be a little more careful than the above code
4133                  * since packet floods will also be caught here. We don't
4134                  * update the TTL here to mitigate the damage of a packet
4135                  * flood and so the same code can handle awkward establishment
4136                  * and a loosened connection close.
4137                  * In the establishment case, a correct peer response will
4138                  * validate the connection, go through the normal state code
4139                  * and keep updating the state TTL.
4140                  */
4141
4142                 if (pf_status.debug >= PF_DEBUG_MISC) {
4143                         kprintf("pf: loose state match: ");
4144                         pf_print_state(*state);
4145                         pf_print_flags(th->th_flags);
4146                         kprintf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
4147                             "pkts=%llu:%llu dir=%s,%s\n", seq, orig_seq, ack, pd->p_len,
4148                             ackskew, (unsigned long long)(*state)->packets[0],
4149                             (unsigned long long)(*state)->packets[1],
4150                             pd->dir == PF_IN ? "in" : "out",
4151                             pd->dir == (*state)->direction ? "fwd" : "rev");
4152                 }
4153
4154                 if (dst->scrub || src->scrub) {
4155                         if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4156                             *state, src, dst, copyback))
4157                                 return (PF_DROP);
4158                 }
4159
4160                 /* update max window */
4161                 if (src->max_win < win)
4162                         src->max_win = win;
4163                 /* synchronize sequencing */
4164                 if (SEQ_GT(end, src->seqlo))
4165                         src->seqlo = end;
4166                 /* slide the window of what the other end can send */
4167                 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4168                         dst->seqhi = ack + MAX((win << sws), 1);
4169
4170                 /*
4171                  * Cannot set dst->seqhi here since this could be a shotgunned
4172                  * SYN and not an already established connection.
4173                  */
4174
4175                 if (th->th_flags & TH_FIN)
4176                         if (src->state < TCPS_CLOSING)
4177                                 src->state = TCPS_CLOSING;
4178                 if (th->th_flags & TH_RST)
4179                         src->state = dst->state = TCPS_TIME_WAIT;
4180
4181                 /* Fall through to PASS packet */
4182
4183         } else if ((*state)->pickup_mode == PF_PICKUPS_HASHONLY ||
4184                     ((*state)->pickup_mode == PF_PICKUPS_ENABLED &&
4185                      ((*state)->sync_flags & PFSTATE_GOT_SYN_MASK) !=
4186                       PFSTATE_GOT_SYN_MASK)) {
4187                 /*
4188                  * If pickup mode is hash only, do not fail on sequence checks.
4189                  *
4190                  * If pickup mode is enabled and we did not see the SYN in
4191                  * both direction, do not fail on sequence checks because
4192                  * we do not have complete information on window scale.
4193                  *
4194                  * Adjust expiration and fall through to PASS packet.
4195                  * XXX Add a FIN check to reduce timeout?
4196                  */
4197                 (*state)->expire = time_second;
4198         } else  {
4199                 /*
4200                  * Failure processing
4201                  */
4202                 if ((*state)->dst.state == TCPS_SYN_SENT &&
4203                     (*state)->src.state == TCPS_SYN_SENT) {
4204                         /* Send RST for state mismatches during handshake */
4205                         if (!(th->th_flags & TH_RST))
4206                                 pf_send_tcp((*state)->rule.ptr, pd->af,
4207                                     pd->dst, pd->src, th->th_dport,
4208                                     th->th_sport, ntohl(th->th_ack), 0,
4209                                     TH_RST, 0, 0,
4210                                     (*state)->rule.ptr->return_ttl, 1, 0,
4211                                     pd->eh, kif->pfik_ifp);
4212                         src->seqlo = 0;
4213                         src->seqhi = 1;
4214                         src->max_win = 1;
4215                 } else if (pf_status.debug >= PF_DEBUG_MISC) {
4216                         kprintf("pf: BAD state: ");
4217                         pf_print_state(*state);
4218                         pf_print_flags(th->th_flags);
4219                         kprintf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
4220                             "pkts=%llu:%llu dir=%s,%s\n",
4221                             seq, orig_seq, ack, pd->p_len, ackskew,
4222                             (unsigned long long)(*state)->packets[0],
4223                                 (unsigned long long)(*state)->packets[1],
4224                             pd->dir == PF_IN ? "in" : "out",
4225                             pd->dir == (*state)->direction ? "fwd" : "rev");
4226                         kprintf("pf: State failure on: %c %c %c %c | %c %c\n",
4227                             SEQ_GEQ(src->seqhi, end) ? ' ' : '1',
4228                             SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ?
4229                             ' ': '2',
4230                             (ackskew >= -MAXACKWINDOW) ? ' ' : '3',
4231                             (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4',
4232                             SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5',
4233                             SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6');
4234                 }
4235                 REASON_SET(reason, PFRES_BADSTATE);
4236                 return (PF_DROP);
4237         }
4238
4239         return (PF_PASS);
4240 }
4241
4242 int
4243 pf_tcp_track_sloppy(struct pf_state_peer *src, struct pf_state_peer *dst,
4244         struct pf_state **state, struct pf_pdesc *pd, u_short *reason)
4245 {
4246         struct tcphdr           *th = pd->hdr.tcp;
4247
4248         if (th->th_flags & TH_SYN)
4249                 if (src->state < TCPS_SYN_SENT)
4250                         src->state = TCPS_SYN_SENT;
4251         if (th->th_flags & TH_FIN)
4252                 if (src->state < TCPS_CLOSING)
4253                         src->state = TCPS_CLOSING;
4254         if (th->th_flags & TH_ACK) {
4255                 if (dst->state == TCPS_SYN_SENT) {
4256                         dst->state = TCPS_ESTABLISHED;
4257                         if (src->state == TCPS_ESTABLISHED &&
4258                             (*state)->src_node != NULL &&
4259                             pf_src_connlimit(state)) {
4260                                 REASON_SET(reason, PFRES_SRCLIMIT);
4261                                 return (PF_DROP);
4262                         }
4263                 } else if (dst->state == TCPS_CLOSING) {
4264                         dst->state = TCPS_FIN_WAIT_2;
4265                 } else if (src->state == TCPS_SYN_SENT &&
4266                     dst->state < TCPS_SYN_SENT) {
4267                         /*
4268                          * Handle a special sloppy case where we only see one
4269                          * half of the connection. If there is a ACK after
4270                          * the initial SYN without ever seeing a packet from
4271                          * the destination, set the connection to established.
4272                          */
4273                         dst->state = src->state = TCPS_ESTABLISHED;
4274                         if ((*state)->src_node != NULL &&
4275                             pf_src_connlimit(state)) {
4276                                 REASON_SET(reason, PFRES_SRCLIMIT);
4277                                 return (PF_DROP);
4278                         }
4279                 } else if (src->state == TCPS_CLOSING &&
4280                     dst->state == TCPS_ESTABLISHED &&
4281                     dst->seqlo == 0) {
4282                         /*
4283                          * Handle the closing of half connections where we
4284                          * don't see the full bidirectional FIN/ACK+ACK
4285                          * handshake.
4286                          */
4287                         dst->state = TCPS_CLOSING;
4288                 }
4289         }
4290         if (th->th_flags & TH_RST)
4291                 src->state = dst->state = TCPS_TIME_WAIT;
4292
4293         /* update expire time */
4294         (*state)->expire = time_second;
4295         if (src->state >= TCPS_FIN_WAIT_2 &&
4296             dst->state >= TCPS_FIN_WAIT_2)
4297                 (*state)->timeout = PFTM_TCP_CLOSED;
4298         else if (src->state >= TCPS_CLOSING &&
4299             dst->state >= TCPS_CLOSING)
4300                 (*state)->timeout = PFTM_TCP_FIN_WAIT;
4301         else if (src->state < TCPS_ESTABLISHED ||
4302             dst->state < TCPS_ESTABLISHED)
4303                 (*state)->timeout = PFTM_TCP_OPENING;
4304         else if (src->state >= TCPS_CLOSING ||
4305             dst->state >= TCPS_CLOSING)
4306                 (*state)->timeout = PFTM_TCP_CLOSING;
4307         else
4308                 (*state)->timeout = PFTM_TCP_ESTABLISHED;
4309
4310         return (PF_PASS);
4311 }
4312
4313 int
4314 pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
4315     struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
4316     u_short *reason)
4317 {
4318         struct pf_state_key_cmp  key;
4319         struct tcphdr           *th = pd->hdr.tcp;
4320         int                      copyback = 0;
4321         struct pf_state_peer    *src, *dst;
4322         struct pf_state_key     *sk;
4323
4324         key.af = pd->af;
4325         key.proto = IPPROTO_TCP;
4326         if (direction == PF_IN) {       /* wire side, straight */
4327                 PF_ACPY(&key.addr[0], pd->src, key.af);
4328                 PF_ACPY(&key.addr[1], pd->dst, key.af);
4329                 key.port[0] = th->th_sport;
4330                 key.port[1] = th->th_dport;
4331         } else {                        /* stack side, reverse */
4332                 PF_ACPY(&key.addr[1], pd->src, key.af);
4333                 PF_ACPY(&key.addr[0], pd->dst, key.af);
4334                 key.port[1] = th->th_sport;
4335                 key.port[0] = th->th_dport;
4336         }
4337
4338         STATE_LOOKUP(kif, &key, direction, *state, m);
4339
4340         if (direction == (*state)->direction) {
4341                 src = &(*state)->src;
4342                 dst = &(*state)->dst;
4343         } else {
4344                 src = &(*state)->dst;
4345                 dst = &(*state)->src;
4346         }
4347
4348         sk = (*state)->key[pd->didx];
4349
4350         if ((*state)->src.state == PF_TCPS_PROXY_SRC) {
4351                 if (direction != (*state)->direction) {
4352                         REASON_SET(reason, PFRES_SYNPROXY);
4353                         return (PF_SYNPROXY_DROP);
4354                 }
4355                 if (th->th_flags & TH_SYN) {
4356                         if (ntohl(th->th_seq) != (*state)->src.seqlo) {
4357                                 REASON_SET(reason, PFRES_SYNPROXY);
4358                                 return (PF_DROP);
4359                         }
4360                         pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
4361                             pd->src, th->th_dport, th->th_sport,
4362                             (*state)->src.seqhi, ntohl(th->th_seq) + 1,
4363                             TH_SYN|TH_ACK, 0, (*state)->src.mss, 0, 1,
4364                             0, NULL, NULL);
4365                         REASON_SET(reason, PFRES_SYNPROXY);
4366                         return (PF_SYNPROXY_DROP);
4367                 } else if (!(th->th_flags & TH_ACK) ||
4368                     (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
4369                     (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
4370                         REASON_SET(reason, PFRES_SYNPROXY);
4371                         return (PF_DROP);
4372                 } else if ((*state)->src_node != NULL &&
4373                     pf_src_connlimit(state)) {
4374                         REASON_SET(reason, PFRES_SRCLIMIT);
4375                         return (PF_DROP);
4376                 } else
4377                         (*state)->src.state = PF_TCPS_PROXY_DST;
4378         }
4379         if ((*state)->src.state == PF_TCPS_PROXY_DST) {
4380                 if (direction == (*state)->direction) {
4381                         if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) ||
4382                             (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
4383                             (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
4384                                 REASON_SET(reason, PFRES_SYNPROXY);
4385                                 return (PF_DROP);
4386                         }
4387                         (*state)->src.max_win = MAX(ntohs(th->th_win), 1);
4388                         if ((*state)->dst.seqhi == 1)
4389                                 (*state)->dst.seqhi = htonl(karc4random());
4390                         pf_send_tcp((*state)->rule.ptr, pd->af,
4391                             &sk->addr[pd->sidx], &sk->addr[pd->didx],
4392                             sk->port[pd->sidx], sk->port[pd->didx],
4393                             (*state)->dst.seqhi, 0, TH_SYN, 0,
4394                             (*state)->src.mss, 0, 0, (*state)->tag, NULL, NULL);
4395                         REASON_SET(reason, PFRES_SYNPROXY);
4396                         return (PF_SYNPROXY_DROP);
4397                 } else if (((th->th_flags & (TH_SYN|TH_ACK)) !=
4398                     (TH_SYN|TH_ACK)) ||
4399                     (ntohl(th->th_ack) != (*state)->dst.seqhi + 1)) {
4400                         REASON_SET(reason, PFRES_SYNPROXY);
4401                         return (PF_DROP);
4402                 } else {
4403                         (*state)->dst.max_win = MAX(ntohs(th->th_win), 1);
4404                         (*state)->dst.seqlo = ntohl(th->th_seq);
4405                         pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
4406                             pd->src, th->th_dport, th->th_sport,
4407                             ntohl(th->th_ack), ntohl(th->th_seq) + 1,
4408                             TH_ACK, (*state)->src.max_win, 0, 0, 0,
4409                             (*state)->tag, NULL, NULL);
4410                         pf_send_tcp((*state)->rule.ptr, pd->af,
4411                             &sk->addr[pd->sidx], &sk->addr[pd->didx],
4412                             sk->port[pd->sidx], sk->port[pd->didx],
4413                             (*state)->src.seqhi + 1, (*state)->src.seqlo + 1,
4414                             TH_ACK, (*state)->dst.max_win, 0, 0, 1,
4415                             0, NULL, NULL);
4416                         (*state)->src.seqdiff = (*state)->dst.seqhi -
4417                             (*state)->src.seqlo;
4418                         (*state)->dst.seqdiff = (*state)->src.seqhi -
4419                             (*state)->dst.seqlo;
4420                         (*state)->src.seqhi = (*state)->src.seqlo +
4421                             (*state)->dst.max_win;
4422                         (*state)->dst.seqhi = (*state)->dst.seqlo +
4423                             (*state)->src.max_win;
4424                         (*state)->src.wscale = (*state)->dst.wscale = 0;
4425                         (*state)->src.state = (*state)->dst.state =
4426                             TCPS_ESTABLISHED;
4427                         REASON_SET(reason, PFRES_SYNPROXY);
4428                         return (PF_SYNPROXY_DROP);
4429                 }
4430         }
4431
4432         if (((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN) &&
4433             dst->state >= TCPS_FIN_WAIT_2 &&
4434             src->state >= TCPS_FIN_WAIT_2) {
4435                 if (pf_status.debug >= PF_DEBUG_MISC) {
4436                         kprintf("pf: state reuse ");
4437                         pf_print_state(*state);
4438                         pf_print_flags(th->th_flags);
4439                         kprintf("\n");
4440                 }
4441                 /* XXX make sure it's the same direction ?? */
4442                 (*state)->src.state = (*state)->dst.state = TCPS_CLOSED;
4443                 pf_unlink_state(*state);
4444                 *state = NULL;
4445                 return (PF_DROP);
4446         }
4447
4448         if ((*state)->state_flags & PFSTATE_SLOPPY) {
4449                 if (pf_tcp_track_sloppy(src, dst, state, pd, reason) == PF_DROP)
4450                         return (PF_DROP);
4451         } else {
4452                 if (pf_tcp_track_full(src, dst, state, kif, m, off, pd, reason,
4453                     &copyback) == PF_DROP)
4454                         return (PF_DROP);
4455         }
4456
4457         /* translate source/destination address, if necessary */
4458         if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) {
4459                 struct pf_state_key *nk = (*state)->key[pd->didx];
4460
4461                 if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], pd->af) ||
4462                     nk->port[pd->sidx] != th->th_sport)  {
4463                         /*
4464                          * The translated source address may be completely
4465                          * unrelated to the saved link header, make sure
4466                          * a bridge doesn't try to use it.
4467                          */
4468                         m->m_pkthdr.fw_flags &= ~BRIDGE_MBUF_TAGGED;
4469                         m->m_flags &= ~M_HASH;
4470                         pf_change_ap(pd->src, &th->th_sport, pd->ip_sum,
4471                             &th->th_sum, &nk->addr[pd->sidx],
4472                             nk->port[pd->sidx], 0, pd->af);
4473                 }
4474
4475                 if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], pd->af) ||
4476                     nk->port[pd->didx] != th->th_dport) {
4477                         /*
4478                          * If we don't redispatch the packet will go into
4479                          * the protocol stack on the wrong cpu for the
4480                          * post-translated address.
4481                          */
4482                         m->m_flags &= ~M_HASH;
4483                         pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum,
4484                             &th->th_sum, &nk->addr[pd->didx],
4485                             nk->port[pd->didx], 0, pd->af);
4486                 }
4487                 copyback = 1;
4488         }
4489
4490         /* Copyback sequence modulation or stateful scrub changes if needed */
4491         if (copyback)
4492                 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4493
4494         return (PF_PASS);
4495 }
4496
4497 int
4498 pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif,
4499     struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4500 {
4501         struct pf_state_peer    *src, *dst;
4502         struct pf_state_key_cmp  key;
4503         struct udphdr           *uh = pd->hdr.udp;
4504
4505         key.af = pd->af;
4506         key.proto = IPPROTO_UDP;
4507         if (direction == PF_IN) {       /* wire side, straight */
4508                 PF_ACPY(&key.addr[0], pd->src, key.af);
4509                 PF_ACPY(&key.addr[1], pd->dst, key.af);
4510                 key.port[0] = uh->uh_sport;
4511                 key.port[1] = uh->uh_dport;
4512         } else {                        /* stack side, reverse */
4513                 PF_ACPY(&key.addr[1], pd->src, key.af);
4514                 PF_ACPY(&key.addr[0], pd->dst, key.af);
4515                 key.port[1] = uh->uh_sport;
4516                 key.port[0] = uh->uh_dport;
4517         }
4518
4519         STATE_LOOKUP(kif, &key, direction, *state, m);
4520
4521         if (direction == (*state)->direction) {
4522                 src = &(*state)->src;
4523                 dst = &(*state)->dst;
4524         } else {
4525                 src = &(*state)->dst;
4526                 dst = &(*state)->src;
4527         }
4528
4529         /* update states */
4530         if (src->state < PFUDPS_SINGLE)
4531                 src->state = PFUDPS_SINGLE;
4532         if (dst->state == PFUDPS_SINGLE)
4533                 dst->state = PFUDPS_MULTIPLE;
4534
4535         /* update expire time */
4536         (*state)->expire = time_second;
4537         if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE)
4538                 (*state)->timeout = PFTM_UDP_MULTIPLE;
4539         else
4540                 (*state)->timeout = PFTM_UDP_SINGLE;
4541
4542         /* translate source/destination address, if necessary */
4543         if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) {
4544                 struct pf_state_key *nk = (*state)->key[pd->didx];
4545
4546                 if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], pd->af) ||
4547                     nk->port[pd->sidx] != uh->uh_sport) {
4548                         /*
4549                          * The translated source address may be completely
4550                          * unrelated to the saved link header, make sure
4551                          * a bridge doesn't try to use it.
4552                          */
4553                         m->m_pkthdr.fw_flags &= ~BRIDGE_MBUF_TAGGED;
4554                         m->m_flags &= ~M_HASH;
4555                         pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum,
4556                             &uh->uh_sum, &nk->addr[pd->sidx],
4557                             nk->port[pd->sidx], 1, pd->af);
4558                 }
4559
4560                 if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], pd->af) ||
4561                     nk->port[pd->didx] != uh->uh_dport) {
4562                         /*
4563                          * If we don't redispatch the packet will go into
4564                          * the protocol stack on the wrong cpu for the
4565                          * post-translated address.
4566                          */
4567                         m->m_flags &= ~M_HASH;
4568                         pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum,
4569                             &uh->uh_sum, &nk->addr[pd->didx],
4570                             nk->port[pd->didx], 1, pd->af);
4571                 }
4572                 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
4573         }
4574
4575         return (PF_PASS);
4576 }
4577
4578 int
4579 pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif,
4580     struct mbuf *m, int off, void *h, struct pf_pdesc *pd, u_short *reason)
4581 {
4582         struct pf_addr  *saddr = pd->src, *daddr = pd->dst;
4583         u_int16_t        icmpid = 0, *icmpsum;
4584         u_int8_t         icmptype;
4585         int              state_icmp = 0;
4586         struct pf_state_key_cmp key;
4587
4588         switch (pd->proto) {
4589 #ifdef INET
4590         case IPPROTO_ICMP:
4591                 icmptype = pd->hdr.icmp->icmp_type;
4592                 icmpid = pd->hdr.icmp->icmp_id;
4593                 icmpsum = &pd->hdr.icmp->icmp_cksum;
4594
4595                 if (icmptype == ICMP_UNREACH ||
4596                     icmptype == ICMP_SOURCEQUENCH ||
4597                     icmptype == ICMP_REDIRECT ||
4598                     icmptype == ICMP_TIMXCEED ||
4599                     icmptype == ICMP_PARAMPROB)
4600                         state_icmp++;
4601                 break;
4602 #endif /* INET */
4603 #ifdef INET6
4604         case IPPROTO_ICMPV6:
4605                 icmptype = pd->hdr.icmp6->icmp6_type;
4606                 icmpid = pd->hdr.icmp6->icmp6_id;
4607                 icmpsum = &pd->hdr.icmp6->icmp6_cksum;
4608
4609                 if (icmptype == ICMP6_DST_UNREACH ||
4610                     icmptype == ICMP6_PACKET_TOO_BIG ||
4611                     icmptype == ICMP6_TIME_EXCEEDED ||
4612                     icmptype == ICMP6_PARAM_PROB)
4613                         state_icmp++;
4614                 break;
4615 #endif /* INET6 */
4616         }
4617
4618         if (!state_icmp) {
4619
4620                 /*
4621                  * ICMP query/reply message not related to a TCP/UDP packet.
4622                  * Search for an ICMP state.
4623                  */
4624                 key.af = pd->af;
4625                 key.proto = pd->proto;
4626                 key.port[0] = key.port[1] = icmpid;
4627                 if (direction == PF_IN) {       /* wire side, straight */
4628                         PF_ACPY(&key.addr[0], pd->src, key.af);
4629                         PF_ACPY(&key.addr[1], pd->dst, key.af);
4630                 } else {                        /* stack side, reverse */
4631                         PF_ACPY(&key.addr[1], pd->src, key.af);
4632                         PF_ACPY(&key.addr[0], pd->dst, key.af);
4633                 }
4634
4635                 STATE_LOOKUP(kif, &key, direction, *state, m);
4636
4637                 (*state)->expire = time_second;
4638                 (*state)->timeout = PFTM_ICMP_ERROR_REPLY;
4639
4640                 /* translate source/destination address, if necessary */
4641                 if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) {
4642                         struct pf_state_key *nk = (*state)->key[pd->didx];
4643
4644                         switch (pd->af) {
4645 #ifdef INET
4646                         case AF_INET:
4647                                 if (PF_ANEQ(pd->src,
4648                                     &nk->addr[pd->sidx], AF_INET))
4649                                         pf_change_a(&saddr->v4.s_addr,
4650                                             pd->ip_sum,
4651                                             nk->addr[pd->sidx].v4.s_addr, 0);
4652
4653                                 if (PF_ANEQ(pd->dst, &nk->addr[pd->didx],
4654                                     AF_INET))
4655                                         pf_change_a(&daddr->v4.s_addr,
4656                                             pd->ip_sum,
4657                                             nk->addr[pd->didx].v4.s_addr, 0);
4658
4659                                 if (nk->port[0] !=
4660                                     pd->hdr.icmp->icmp_id) {
4661                                         pd->hdr.icmp->icmp_cksum =
4662                                             pf_cksum_fixup(
4663                                             pd->hdr.icmp->icmp_cksum, icmpid,
4664                                             nk->port[pd->sidx], 0);
4665                                         pd->hdr.icmp->icmp_id =
4666                                             nk->port[pd->sidx];
4667                                 }
4668
4669                                 m_copyback(m, off, ICMP_MINLEN,
4670                                     (caddr_t)pd->hdr.icmp);
4671                                 break;
4672 #endif /* INET */
4673 #ifdef INET6
4674                         case AF_INET6:
4675                                 if (PF_ANEQ(pd->src,
4676                                     &nk->addr[pd->sidx], AF_INET6))
4677                                         pf_change_a6(saddr,
4678                                             &pd->hdr.icmp6->icmp6_cksum,
4679                                             &nk->addr[pd->sidx], 0);
4680
4681                                 if (PF_ANEQ(pd->dst,
4682                                     &nk->addr[pd->didx], AF_INET6))
4683                                         pf_change_a6(daddr,
4684                                             &pd->hdr.icmp6->icmp6_cksum,
4685                                             &nk->addr[pd->didx], 0);
4686
4687                                 m_copyback(m, off,
4688                                         sizeof(struct icmp6_hdr),
4689                                         (caddr_t)pd->hdr.icmp6);
4690                                 break;
4691 #endif /* INET6 */
4692                         }
4693                 }
4694                 return (PF_PASS);
4695
4696         } else {
4697                 /*
4698                  * ICMP error message in response to a TCP/UDP packet.
4699                  * Extract the inner TCP/UDP header and search for that state.
4700                  */
4701
4702                 struct pf_pdesc pd2;
4703 #ifdef INET
4704                 struct ip       h2;
4705 #endif /* INET */
4706 #ifdef INET6
4707                 struct ip6_hdr  h2_6;
4708                 int             terminal = 0;
4709 #endif /* INET6 */
4710                 int             ipoff2;
4711                 int             off2;
4712
4713                 pd2.af = pd->af;
4714                 /* Payload packet is from the opposite direction. */
4715                 pd2.sidx = (direction == PF_IN) ? 1 : 0;
4716                 pd2.didx = (direction == PF_IN) ? 0 : 1;
4717                 switch (pd->af) {
4718 #ifdef INET
4719                 case AF_INET:
4720                         /* offset of h2 in mbuf chain */
4721                         ipoff2 = off + ICMP_MINLEN;
4722
4723                         if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2),
4724                             NULL, reason, pd2.af)) {
4725                                 DPFPRINTF(PF_DEBUG_MISC,
4726                                     ("pf: ICMP error message too short "
4727                                     "(ip)\n"));
4728                                 return (PF_DROP);
4729                         }
4730                         /*
4731                          * ICMP error messages don't refer to non-first
4732                          * fragments
4733                          */
4734                         if (h2.ip_off & htons(IP_OFFMASK)) {
4735                                 REASON_SET(reason, PFRES_FRAG);
4736                                 return (PF_DROP);
4737                         }
4738
4739                         /* offset of protocol header that follows h2 */
4740                         off2 = ipoff2 + (h2.ip_hl << 2);
4741
4742                         pd2.proto = h2.ip_p;
4743                         pd2.src = (struct pf_addr *)&h2.ip_src;
4744                         pd2.dst = (struct pf_addr *)&h2.ip_dst;
4745                         pd2.ip_sum = &h2.ip_sum;
4746                         break;
4747 #endif /* INET */
4748 #ifdef INET6
4749                 case AF_INET6:
4750                         ipoff2 = off + sizeof(struct icmp6_hdr);
4751
4752                         if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6),
4753                             NULL, reason, pd2.af)) {
4754                                 DPFPRINTF(PF_DEBUG_MISC,
4755                                     ("pf: ICMP error message too short "
4756                                     "(ip6)\n"));
4757                                 return (PF_DROP);
4758                         }
4759                         pd2.proto = h2_6.ip6_nxt;
4760                         pd2.src = (struct pf_addr *)&h2_6.ip6_src;
4761                         pd2.dst = (struct pf_addr *)&h2_6.ip6_dst;
4762                         pd2.ip_sum = NULL;
4763                         off2 = ipoff2 + sizeof(h2_6);
4764                         do {
4765                                 switch (pd2.proto) {
4766                                 case IPPROTO_FRAGMENT:
4767                                         /*
4768                                          * ICMPv6 error messages for
4769                                          * non-first fragments
4770                                          */
4771                                         REASON_SET(reason, PFRES_FRAG);
4772                                         return (PF_DROP);
4773                                 case IPPROTO_AH:
4774                                 case IPPROTO_HOPOPTS:
4775                                 case IPPROTO_ROUTING:
4776                                 case IPPROTO_DSTOPTS: {
4777                                         /* get next header and header length */
4778                                         struct ip6_ext opt6;
4779
4780                                         if (!pf_pull_hdr(m, off2, &opt6,
4781                                             sizeof(opt6), NULL, reason,
4782                                             pd2.af)) {
4783                                                 DPFPRINTF(PF_DEBUG_MISC,
4784                                                     ("pf: ICMPv6 short opt\n"));
4785                                                 return (PF_DROP);
4786                                         }
4787                                         if (pd2.proto == IPPROTO_AH)
4788                                                 off2 += (opt6.ip6e_len + 2) * 4;
4789                                         else
4790                                                 off2 += (opt6.ip6e_len + 1) * 8;
4791                                         pd2.proto = opt6.ip6e_nxt;
4792                                         /* goto the next header */
4793                                         break;
4794                                 }
4795                                 default:
4796                                         terminal++;
4797                                         break;
4798                                 }
4799                         } while (!terminal);
4800                         break;
4801 #endif /* INET6 */
4802                 default:
4803                         DPFPRINTF(PF_DEBUG_MISC,
4804                             ("pf: ICMP AF %d unknown (ip6)\n", pd->af));
4805                         return (PF_DROP);
4806                         break;
4807                 }
4808
4809                 switch (pd2.proto) {
4810                 case IPPROTO_TCP: {
4811                         struct tcphdr            th;
4812                         u_int32_t                seq;
4813                         struct pf_state_peer    *src, *dst;
4814                         u_int8_t                 dws;
4815                         int                      copyback = 0;
4816
4817                         /*
4818                          * Only the first 8 bytes of the TCP header can be
4819                          * expected. Don't access any TCP header fields after
4820                          * th_seq, an ackskew test is not possible.
4821                          */
4822                         if (!pf_pull_hdr(m, off2, &th, 8, NULL, reason,
4823                             pd2.af)) {
4824                                 DPFPRINTF(PF_DEBUG_MISC,
4825                                     ("pf: ICMP error message too short "
4826                                     "(tcp)\n"));
4827                                 return (PF_DROP);
4828                         }
4829
4830                         key.af = pd2.af;
4831                         key.proto = IPPROTO_TCP;
4832                         PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
4833                         PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
4834                         key.port[pd2.sidx] = th.th_sport;
4835                         key.port[pd2.didx] = th.th_dport;
4836
4837                         STATE_LOOKUP(kif, &key, direction, *state, m);
4838
4839                         if (direction == (*state)->direction) {
4840                                 src = &(*state)->dst;
4841                                 dst = &(*state)->src;
4842                         } else {
4843                                 src = &(*state)->src;
4844                                 dst = &(*state)->dst;
4845                         }
4846
4847                         if (src->wscale && dst->wscale)
4848                                 dws = dst->wscale & PF_WSCALE_MASK;
4849                         else
4850                                 dws = 0;
4851
4852                         /* Demodulate sequence number */
4853                         seq = ntohl(th.th_seq) - src->seqdiff;
4854                         if (src->seqdiff) {
4855                                 pf_change_a(&th.th_seq, icmpsum,
4856                                     htonl(seq), 0);
4857                                 copyback = 1;
4858                         }
4859
4860                         if (!((*state)->state_flags & PFSTATE_SLOPPY) &&
4861                             (!SEQ_GEQ(src->seqhi, seq) ||
4862                             !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)))) {
4863                                 if (pf_status.debug >= PF_DEBUG_MISC) {
4864                                         kprintf("pf: BAD ICMP %d:%d ",
4865                                             icmptype, pd->hdr.icmp->icmp_code);
4866                                         pf_print_host(pd->src, 0, pd->af);
4867                                         kprintf(" -> ");
4868                                         pf_print_host(pd->dst, 0, pd->af);
4869                                         kprintf(" state: ");
4870                                         pf_print_state(*state);
4871                                         kprintf(" seq=%u\n", seq);
4872                                 }
4873                                 REASON_SET(reason, PFRES_BADSTATE);
4874                                 return (PF_DROP);
4875                         } else {
4876                                 if (pf_status.debug >= PF_DEBUG_MISC) {
4877                                         kprintf("pf: OK ICMP %d:%d ",
4878                                             icmptype, pd->hdr.icmp->icmp_code);
4879                                         pf_print_host(pd->src, 0, pd->af);
4880                                         kprintf(" -> ");
4881                                         pf_print_host(pd->dst, 0, pd->af);
4882                                         kprintf(" state: ");
4883                                         pf_print_state(*state);
4884                                         kprintf(" seq=%u\n", seq);
4885                                 }
4886                         }
4887
4888                         /* translate source/destination address, if necessary */
4889                         if ((*state)->key[PF_SK_WIRE] !=
4890                             (*state)->key[PF_SK_STACK]) {
4891                                 struct pf_state_key *nk =
4892                                     (*state)->key[pd->didx];
4893
4894                                 if (PF_ANEQ(pd2.src,
4895                                     &nk->addr[pd2.sidx], pd2.af) ||
4896                                     nk->port[pd2.sidx] != th.th_sport)
4897                                         pf_change_icmp(pd2.src, &th.th_sport,
4898                                             daddr, &nk->addr[pd2.sidx],
4899                                             nk->port[pd2.sidx], NULL,
4900                                             pd2.ip_sum, icmpsum,
4901                                             pd->ip_sum, 0, pd2.af);
4902
4903                                 if (PF_ANEQ(pd2.dst,
4904                                     &nk->addr[pd2.didx], pd2.af) ||
4905                                     nk->port[pd2.didx] != th.th_dport)
4906                                         pf_change_icmp(pd2.dst, &th.th_dport,
4907                                             NULL, /* XXX Inbound NAT? */
4908                                             &nk->addr[pd2.didx],
4909                                             nk->port[pd2.didx], NULL,
4910                                             pd2.ip_sum, icmpsum,
4911                                             pd->ip_sum, 0, pd2.af);
4912                                 copyback = 1;
4913                         }
4914
4915                         if (copyback) {
4916                                 switch (pd2.af) {
4917 #ifdef INET
4918                                 case AF_INET:
4919                                         m_copyback(m, off, ICMP_MINLEN,
4920                                             (caddr_t)pd->hdr.icmp);
4921                                         m_copyback(m, ipoff2, sizeof(h2),
4922                                             (caddr_t)&h2);
4923                                         break;
4924 #endif /* INET */
4925 #ifdef INET6
4926                                 case AF_INET6:
4927                                         m_copyback(m, off,
4928                                             sizeof(struct icmp6_hdr),
4929                                             (caddr_t)pd->hdr.icmp6);
4930                                         m_copyback(m, ipoff2, sizeof(h2_6),
4931                                             (caddr_t)&h2_6);
4932                                         break;
4933 #endif /* INET6 */
4934                                 }
4935                                 m_copyback(m, off2, 8, (caddr_t)&th);
4936                         }
4937
4938                         return (PF_PASS);
4939                         break;
4940                 }
4941                 case IPPROTO_UDP: {
4942                         struct udphdr           uh;
4943
4944                         if (!pf_pull_hdr(m, off2, &uh, sizeof(uh),
4945                             NULL, reason, pd2.af)) {
4946                                 DPFPRINTF(PF_DEBUG_MISC,
4947                                     ("pf: ICMP error message too short "
4948                                     "(udp)\n"));
4949                                 return (PF_DROP);
4950                         }
4951
4952                         key.af = pd2.af;
4953                         key.proto = IPPROTO_UDP;
4954                         PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
4955                         PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
4956                         key.port[pd2.sidx] = uh.uh_sport;
4957                         key.port[pd2.didx] = uh.uh_dport;
4958
4959                         STATE_LOOKUP(kif, &key, direction, *state, m);
4960
4961                         /* translate source/destination address, if necessary */
4962                         if ((*state)->key[PF_SK_WIRE] !=
4963                             (*state)->key[PF_SK_STACK]) {
4964                                 struct pf_state_key *nk =
4965                                     (*state)->key[pd->didx];
4966
4967                                 if (PF_ANEQ(pd2.src,
4968                                     &nk->addr[pd2.sidx], pd2.af) ||
4969                                     nk->port[pd2.sidx] != uh.uh_sport)
4970                                         pf_change_icmp(pd2.src, &uh.uh_sport,
4971                                             daddr, &nk->addr[pd2.sidx],
4972                                             nk->port[pd2.sidx], &uh.uh_sum,
4973                                             pd2.ip_sum, icmpsum,
4974                                             pd->ip_sum, 1, pd2.af);
4975
4976                                 if (PF_ANEQ(pd2.dst,
4977                                     &nk->addr[pd2.didx], pd2.af) ||
4978                                     nk->port[pd2.didx] != uh.uh_dport)
4979                                         pf_change_icmp(pd2.dst, &uh.uh_dport,
4980                                             NULL, /* XXX Inbound NAT? */
4981                                             &nk->addr[pd2.didx],
4982                                             nk->port[pd2.didx], &uh.uh_sum,
4983                                             pd2.ip_sum, icmpsum,
4984                                             pd->ip_sum, 1, pd2.af);
4985
4986                                 switch (pd2.af) {
4987 #ifdef INET
4988                                 case AF_INET:
4989                                         m_copyback(m, off, ICMP_MINLEN,
4990                                             (caddr_t)pd->hdr.icmp);
4991                                         m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4992                                         break;
4993 #endif /* INET */
4994 #ifdef INET6
4995                                 case AF_INET6:
4996                                         m_copyback(m, off,
4997                                             sizeof(struct icmp6_hdr),
4998                                             (caddr_t)pd->hdr.icmp6);
4999                                         m_copyback(m, ipoff2, sizeof(h2_6),
5000                                             (caddr_t)&h2_6);
5001                                         break;
5002 #endif /* INET6 */
5003                                 }
5004                                 m_copyback(m, off2, sizeof(uh), (caddr_t)&uh);
5005                         }
5006
5007                         return (PF_PASS);
5008                         break;
5009                 }
5010 #ifdef INET
5011                 case IPPROTO_ICMP: {
5012                         struct icmp             iih;
5013
5014                         if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN,
5015                             NULL, reason, pd2.af)) {
5016                                 DPFPRINTF(PF_DEBUG_MISC,
5017                                     ("pf: ICMP error message too short i"
5018                                     "(icmp)\n"));
5019                                 return (PF_DROP);
5020                         }
5021
5022                         key.af = pd2.af;
5023                         key.proto = IPPROTO_ICMP;
5024                         PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
5025                         PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
5026                         key.port[0] = key.port[1] = iih.icmp_id;
5027
5028                         STATE_LOOKUP(kif, &key, direction, *state, m);
5029
5030                         /* translate source/destination address, if necessary */
5031                         if ((*state)->key[PF_SK_WIRE] !=
5032                             (*state)->key[PF_SK_STACK]) {
5033                                 struct pf_state_key *nk =
5034                                     (*state)->key[pd->didx];
5035
5036                                 if (PF_ANEQ(pd2.src,
5037                                     &nk->addr[pd2.sidx], pd2.af) ||
5038                                     nk->port[pd2.sidx] != iih.icmp_id)
5039                                         pf_change_icmp(pd2.src, &iih.icmp_id,
5040                                             daddr, &nk->addr[pd2.sidx],
5041                                             nk->port[pd2.sidx], NULL,
5042                                             pd2.ip_sum, icmpsum,
5043                                             pd->ip_sum, 0, AF_INET);
5044
5045                                 if (PF_ANEQ(pd2.dst,
5046                                     &nk->addr[pd2.didx], pd2.af) ||
5047                                     nk->port[pd2.didx] != iih.icmp_id)
5048                                         pf_change_icmp(pd2.dst, &iih.icmp_id,
5049                                             NULL, /* XXX Inbound NAT? */
5050                                             &nk->addr[pd2.didx],
5051                                             nk->port[pd2.didx], NULL,
5052                                             pd2.ip_sum, icmpsum,
5053                                             pd->ip_sum, 0, AF_INET);
5054
5055                                 m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
5056                                 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
5057                                 m_copyback(m, off2, ICMP_MINLEN, (caddr_t)&iih);
5058                         }
5059                         return (PF_PASS);
5060                         break;
5061                 }
5062 #endif /* INET */
5063 #ifdef INET6
5064                 case IPPROTO_ICMPV6: {
5065                         struct icmp6_hdr        iih;
5066
5067                         if (!pf_pull_hdr(m, off2, &iih,
5068                             sizeof(struct icmp6_hdr), NULL, reason, pd2.af)) {
5069                                 DPFPRINTF(PF_DEBUG_MISC,
5070                                     ("pf: ICMP error message too short "
5071                                     "(icmp6)\n"));
5072                                 return (PF_DROP);
5073                         }
5074
5075                         key.af = pd2.af;
5076                         key.proto = IPPROTO_ICMPV6;
5077                         PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
5078                         PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
5079                         key.port[0] = key.port[1] = iih.icmp6_id;
5080
5081                         STATE_LOOKUP(kif, &key, direction, *state, m);
5082
5083                         /* translate source/destination address, if necessary */
5084                         if ((*state)->key[PF_SK_WIRE] !=
5085                             (*state)->key[PF_SK_STACK]) {
5086                                 struct pf_state_key *nk =
5087                                     (*state)->key[pd->didx];
5088
5089                                 if (PF_ANEQ(pd2.src,
5090                                     &nk->addr[pd2.sidx], pd2.af) ||
5091                                     nk->port[pd2.sidx] != iih.icmp6_id)
5092                                         pf_change_icmp(pd2.src, &iih.icmp6_id,
5093                                             daddr, &nk->addr[pd2.sidx],
5094                                             nk->port[pd2.sidx], NULL,
5095                                             pd2.ip_sum, icmpsum,
5096                                             pd->ip_sum, 0, AF_INET6);
5097
5098                                 if (PF_ANEQ(pd2.dst,
5099                                     &nk->addr[pd2.didx], pd2.af) ||
5100                                     nk->port[pd2.didx] != iih.icmp6_id)
5101                                         pf_change_icmp(pd2.dst, &iih.icmp6_id,
5102                                             NULL, /* XXX Inbound NAT? */
5103                                             &nk->addr[pd2.didx],
5104                                             nk->port[pd2.didx], NULL,
5105                                             pd2.ip_sum, icmpsum,
5106                                             pd->ip_sum, 0, AF_INET6);
5107
5108                                 m_copyback(m, off, sizeof(struct icmp6_hdr),
5109                                     (caddr_t)pd->hdr.icmp6);
5110                                 m_copyback(m, ipoff2, sizeof(h2_6), (caddr_t)&h2_6);
5111                                 m_copyback(m, off2, sizeof(struct icmp6_hdr),
5112                                     (caddr_t)&iih);
5113                         }
5114
5115                         return (PF_PASS);
5116                         break;
5117                 }
5118 #endif /* INET6 */
5119                 default: {
5120                         key.af = pd2.af;
5121                         key.proto = pd2.proto;
5122                         PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
5123                         PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
5124                         key.port[0] = key.port[1] = 0;
5125
5126                         STATE_LOOKUP(kif, &key, direction, *state, m);
5127
5128                         /* translate source/destination address, if necessary */
5129                         if ((*state)->key[PF_SK_WIRE] !=
5130                             (*state)->key[PF_SK_STACK]) {
5131                                 struct pf_state_key *nk =
5132                                     (*state)->key[pd->didx];
5133
5134                                 if (PF_ANEQ(pd2.src,
5135                                     &nk->addr[pd2.sidx], pd2.af))
5136                                         pf_change_icmp(pd2.src, NULL, daddr,
5137                                             &nk->addr[pd2.sidx], 0, NULL,
5138                                             pd2.ip_sum, icmpsum,
5139                                             pd->ip_sum, 0, pd2.af);
5140
5141                                 if (PF_ANEQ(pd2.dst,
5142                                     &nk->addr[pd2.didx], pd2.af))
5143                                         pf_change_icmp(pd2.src, NULL,
5144                                             NULL, /* XXX Inbound NAT? */
5145                                             &nk->addr[pd2.didx], 0, NULL,
5146                                             pd2.ip_sum, icmpsum,
5147                                             pd->ip_sum, 0, pd2.af);
5148
5149                                 switch (pd2.af) {
5150 #ifdef INET
5151                                 case AF_INET:
5152                                         m_copyback(m, off, ICMP_MINLEN,
5153                                             (caddr_t)pd->hdr.icmp);
5154                                         m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
5155                                         break;
5156 #endif /* INET */
5157 #ifdef INET6
5158                                 case AF_INET6:
5159                                         m_copyback(m, off,
5160                                             sizeof(struct icmp6_hdr),
5161                                             (caddr_t)pd->hdr.icmp6);
5162                                         m_copyback(m, ipoff2, sizeof(h2_6),
5163                                             (caddr_t)&h2_6);
5164                                         break;
5165 #endif /* INET6 */
5166                                 }
5167                         }
5168                         return (PF_PASS);
5169                         break;
5170                 }
5171                 }
5172         }
5173 }
5174
5175 int
5176 pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif,
5177     struct mbuf *m, struct pf_pdesc *pd)
5178 {
5179         struct pf_state_peer    *src, *dst;
5180         struct pf_state_key_cmp  key;
5181
5182         key.af = pd->af;
5183         key.proto = pd->proto;
5184         if (direction == PF_IN) {
5185                 PF_ACPY(&key.addr[0], pd->src, key.af);
5186                 PF_ACPY(&key.addr[1], pd->dst, key.af);
5187                 key.port[0] = key.port[1] = 0;
5188         } else {
5189                 PF_ACPY(&key.addr[1], pd->src, key.af);
5190                 PF_ACPY(&key.addr[0], pd->dst, key.af);
5191                 key.port[1] = key.port[0] = 0;
5192         }
5193
5194         STATE_LOOKUP(kif, &key, direction, *state, m);
5195
5196         if (direction == (*state)->direction) {
5197                 src = &(*state)->src;
5198                 dst = &(*state)->dst;
5199         } else {
5200                 src = &(*state)->dst;
5201                 dst = &(*state)->src;
5202         }
5203
5204         /* update states */
5205         if (src->state < PFOTHERS_SINGLE)
5206                 src->state = PFOTHERS_SINGLE;
5207         if (dst->state == PFOTHERS_SINGLE)
5208                 dst->state = PFOTHERS_MULTIPLE;
5209
5210         /* update expire time */
5211         (*state)->expire = time_second;
5212         if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE)
5213                 (*state)->timeout = PFTM_OTHER_MULTIPLE;
5214         else
5215                 (*state)->timeout = PFTM_OTHER_SINGLE;
5216
5217         /* translate source/destination address, if necessary */
5218         if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) {
5219                 struct pf_state_key *nk = (*state)->key[pd->didx];
5220
5221                 KKASSERT(nk);
5222                 KKASSERT(pd);
5223                 KKASSERT(pd->src);
5224                 KKASSERT(pd->dst);
5225                 switch (pd->af) {
5226 #ifdef INET
5227                 case AF_INET:
5228                         if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], AF_INET))
5229                                 pf_change_a(&pd->src->v4.s_addr,
5230                                     pd->ip_sum,
5231                                     nk->addr[pd->sidx].v4.s_addr,
5232                                     0);
5233
5234
5235                         if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], AF_INET))
5236                                 pf_change_a(&pd->dst->v4.s_addr,
5237                                     pd->ip_sum,
5238                                     nk->addr[pd->didx].v4.s_addr,
5239                                     0);
5240
5241                                 break;
5242 #endif /* INET */
5243 #ifdef INET6
5244                 case AF_INET6:
5245                         if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], AF_INET))
5246                                 PF_ACPY(pd->src, &nk->addr[pd->sidx], pd->af);
5247
5248                         if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], AF_INET))
5249                                 PF_ACPY(pd->dst, &nk->addr[pd->didx], pd->af);
5250 #endif /* INET6 */
5251                 }
5252         }
5253         return (PF_PASS);
5254 }
5255
5256 /*
5257  * ipoff and off are measured from the start of the mbuf chain.
5258  * h must be at "ipoff" on the mbuf chain.
5259  */
5260 void *
5261 pf_pull_hdr(struct mbuf *m, int off, void *p, int len,
5262     u_short *actionp, u_short *reasonp, sa_family_t af)
5263 {
5264         switch (af) {
5265 #ifdef INET
5266         case AF_INET: {
5267                 struct ip       *h = mtod(m, struct ip *);
5268                 u_int16_t        fragoff = (h->ip_off & IP_OFFMASK) << 3;
5269
5270                 if (fragoff) {
5271                         if (fragoff >= len)
5272                                 ACTION_SET(actionp, PF_PASS);
5273                         else {
5274                                 ACTION_SET(actionp, PF_DROP);
5275                                 REASON_SET(reasonp, PFRES_FRAG);
5276                         }
5277                         return (NULL);
5278                 }
5279                 if (m->m_pkthdr.len < off + len ||
5280                     h->ip_len < off + len) {
5281                         ACTION_SET(actionp, PF_DROP);
5282                         REASON_SET(reasonp, PFRES_SHORT);
5283                         return (NULL);
5284                 }
5285                 break;
5286         }
5287 #endif /* INET */
5288 #ifdef INET6
5289         case AF_INET6: {
5290                 struct ip6_hdr  *h = mtod(m, struct ip6_hdr *);
5291
5292                 if (m->m_pkthdr.len < off + len ||
5293                     (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) <
5294                     (unsigned)(off + len)) {
5295                         ACTION_SET(actionp, PF_DROP);
5296                         REASON_SET(reasonp, PFRES_SHORT);
5297                         return (NULL);
5298                 }
5299                 break;
5300         }
5301 #endif /* INET6 */
5302         }
5303         m_copydata(m, off, len, p);
5304         return (p);
5305 }
5306
5307 int
5308 pf_routable(struct pf_addr *addr, sa_family_t af, struct pfi_kif *kif)
5309 {
5310         struct sockaddr_in      *dst;
5311         int                      ret = 1;
5312         int                      check_mpath;
5313 #ifdef INET6
5314         struct sockaddr_in6     *dst6;
5315         struct route_in6         ro;
5316 #else
5317         struct route             ro;
5318 #endif
5319         struct radix_node       *rn;
5320         struct rtentry          *rt;
5321         struct ifnet            *ifp;
5322
5323         check_mpath = 0;
5324         bzero(&ro, sizeof(ro));
5325         switch (af) {
5326         case AF_INET:
5327                 dst = satosin(&ro.ro_dst);
5328                 dst->sin_family = AF_INET;
5329                 dst->sin_len = sizeof(*dst);
5330                 dst->sin_addr = addr->v4;
5331                 break;
5332 #ifdef INET6
5333         case AF_INET6:
5334                 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
5335                 dst6->sin6_family = AF_INET6;
5336                 dst6->sin6_len = sizeof(*dst6);
5337                 dst6->sin6_addr = addr->v6;
5338                 break;
5339 #endif /* INET6 */
5340         default:
5341                 return (0);
5342         }
5343
5344         /* Skip checks for ipsec interfaces */
5345         if (kif != NULL && kif->pfik_ifp->if_type == IFT_ENC)
5346                 goto out;
5347
5348         rtalloc_ign((struct route *)&ro, 0);
5349
5350         if (ro.ro_rt != NULL) {
5351                 /* No interface given, this is a no-route check */
5352                 if (kif == NULL)
5353                         goto out;
5354
5355                 if (kif->pfik_ifp == NULL) {
5356                         ret = 0;
5357                         goto out;
5358                 }
5359
5360                 /* Perform uRPF check if passed input interface */
5361                 ret = 0;
5362                 rn = (struct radix_node *)ro.ro_rt;
5363                 do {
5364                         rt = (struct rtentry *)rn;
5365                         ifp = rt->rt_ifp;
5366
5367                         if (kif->pfik_ifp == ifp)
5368                                 ret = 1;
5369                         rn = NULL;
5370                 } while (check_mpath == 1 && rn != NULL && ret == 0);
5371         } else
5372                 ret = 0;
5373 out:
5374         if (ro.ro_rt != NULL)
5375                 RTFREE(ro.ro_rt);
5376         return (ret);
5377 }
5378
5379 int
5380 pf_rtlabel_match(struct pf_addr *addr, sa_family_t af, struct pf_addr_wrap *aw)
5381 {
5382         struct sockaddr_in      *dst;
5383 #ifdef INET6
5384         struct sockaddr_in6     *dst6;
5385         struct route_in6         ro;
5386 #else
5387         struct route             ro;
5388 #endif
5389         int                      ret = 0;
5390
5391         ASSERT_LWKT_TOKEN_HELD(&pf_token);
5392
5393         bzero(&ro, sizeof(ro));
5394         switch (af) {
5395         case AF_INET:
5396                 dst = satosin(&ro.ro_dst);
5397                 dst->sin_family = AF_INET;
5398                 dst->sin_len = sizeof(*dst);
5399                 dst->sin_addr = addr->v4;
5400                 break;
5401 #ifdef INET6
5402         case AF_INET6:
5403                 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
5404                 dst6->sin6_family = AF_INET6;
5405                 dst6->sin6_len = sizeof(*dst6);
5406                 dst6->sin6_addr = addr->v6;
5407                 break;
5408 #endif /* INET6 */
5409         default:
5410                 return (0);
5411         }
5412
5413 rtalloc_ign((struct route *)&ro, (RTF_CLONING | RTF_PRCLONING));
5414
5415         if (ro.ro_rt != NULL) {
5416                 RTFREE(ro.ro_rt);
5417         }
5418
5419         return (ret);
5420 }
5421
5422 #ifdef INET
5423 void
5424 pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5425     struct pf_state *s, struct pf_pdesc *pd)
5426 {
5427         struct mbuf             *m0, *m1;
5428         struct route             iproute;
5429         struct route            *ro = NULL;
5430         struct sockaddr_in      *dst;
5431         struct ip               *ip;
5432         struct ifnet            *ifp = NULL;
5433         struct pf_addr           naddr;
5434         struct pf_src_node      *sn = NULL;
5435         int                      error = 0;
5436         int sw_csum;
5437 #ifdef IPSEC
5438         struct m_tag            *mtag;
5439 #endif /* IPSEC */
5440
5441         ASSERT_LWKT_TOKEN_HELD(&pf_token);
5442
5443         if (m == NULL || *m == NULL || r == NULL ||
5444             (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5445                 panic("pf_route: invalid parameters");
5446
5447         if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
5448                 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
5449                 (*m)->m_pkthdr.pf.routed = 1;
5450         } else {
5451                 if ((*m)->m_pkthdr.pf.routed++ > 3) {
5452                         m0 = *m;
5453                         *m = NULL;
5454                         goto bad;
5455                 }
5456         }
5457
5458         if (r->rt == PF_DUPTO) {
5459                 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL) {
5460                         return;
5461                 }
5462         } else {
5463                 if ((r->rt == PF_REPLYTO) == (r->direction == dir)) {
5464                         return;
5465                 }
5466                 m0 = *m;
5467         }
5468
5469         if (m0->m_len < sizeof(struct ip)) {
5470                 DPFPRINTF(PF_DEBUG_URGENT,
5471                     ("pf_route: m0->m_len < sizeof(struct ip)\n"));
5472                 goto bad;
5473         }
5474
5475         ip = mtod(m0, struct ip *);
5476
5477         ro = &iproute;
5478         bzero((caddr_t)ro, sizeof(*ro));
5479         dst = satosin(&ro->ro_dst);
5480         dst->sin_family = AF_INET;
5481         dst->sin_len = sizeof(*dst);
5482         dst->sin_addr = ip->ip_dst;
5483
5484         if (r->rt == PF_FASTROUTE) {
5485                 rtalloc(ro);
5486                 if (ro->ro_rt == 0) {
5487                         ipstat.ips_noroute++;
5488                         goto bad;
5489                 }
5490
5491                 ifp = ro->ro_rt->rt_ifp;
5492                 ro->ro_rt->rt_use++;
5493
5494                 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
5495                         dst = satosin(ro->ro_rt->rt_gateway);
5496         } else {
5497                 if (TAILQ_EMPTY(&r->rpool.list)) {
5498                         DPFPRINTF(PF_DEBUG_URGENT,
5499                             ("pf_route: TAILQ_EMPTY(&r->rpool.list)\n"));
5500                         goto bad;
5501                 }
5502                 if (s == NULL) {
5503                         pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src,
5504                             &naddr, NULL, &sn);
5505                         if (!PF_AZERO(&naddr, AF_INET))
5506                                 dst->sin_addr.s_addr = naddr.v4.s_addr;
5507                         ifp = r->rpool.cur->kif ?
5508                             r->rpool.cur->kif->pfik_ifp : NULL;
5509                 } else {
5510                         if (!PF_AZERO(&s->rt_addr, AF_INET))
5511                                 dst->sin_addr.s_addr =
5512                                     s->rt_addr.v4.s_addr;
5513                         ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5514                 }
5515         }
5516         if (ifp == NULL)
5517                 goto bad;
5518
5519         if (oifp != ifp) {
5520                 if (pf_test(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) {
5521                         goto bad;
5522                 } else if (m0 == NULL) {
5523                         goto done;
5524                 }
5525                 if (m0->m_len < sizeof(struct ip)) {
5526                         DPFPRINTF(PF_DEBUG_URGENT,
5527                             ("pf_route: m0->m_len < sizeof(struct ip)\n"));
5528                         goto bad;
5529                 }
5530                 ip = mtod(m0, struct ip *);
5531         }
5532
5533         /* Copied from FreeBSD 5.1-CURRENT ip_output. */
5534         m0->m_pkthdr.csum_flags |= CSUM_IP;
5535         sw_csum = m0->m_pkthdr.csum_flags & ~ifp->if_hwassist;
5536         if (sw_csum & CSUM_DELAY_DATA) {
5537                 in_delayed_cksum(m0);
5538                 sw_csum &= ~CSUM_DELAY_DATA;
5539         }
5540         m0->m_pkthdr.csum_flags &= ifp->if_hwassist;
5541
5542         if (ip->ip_len <= ifp->if_mtu ||
5543             (ifp->if_hwassist & CSUM_FRAGMENT &&
5544                 (ip->ip_off & IP_DF) == 0)) {
5545                 ip->ip_len = htons(ip->ip_len);
5546                 ip->ip_off = htons(ip->ip_off);
5547                 ip->ip_sum = 0;
5548                 if (sw_csum & CSUM_DELAY_IP) {
5549                         /* From KAME */
5550                         if (ip->ip_v == IPVERSION &&
5551                             (ip->ip_hl << 2) == sizeof(*ip)) {
5552                                 ip->ip_sum = in_cksum_hdr(ip);
5553                         } else {
5554                                 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
5555                         }
5556                 }
5557                 lwkt_reltoken(&pf_token);
5558                 error = ifp->if_output(ifp, m0, sintosa(dst), ro->ro_rt);
5559                 lwkt_gettoken(&pf_token);
5560                 goto done;
5561         }
5562
5563         /*
5564          * Too large for interface; fragment if possible.
5565          * Must be able to put at least 8 bytes per fragment.
5566          */
5567         if (ip->ip_off & IP_DF) {
5568                 ipstat.ips_cantfrag++;
5569                 if (r->rt != PF_DUPTO) {
5570                         icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0,
5571                             ifp->if_mtu);
5572                         goto done;
5573                 } else
5574                         goto bad;
5575         }
5576
5577         m1 = m0;
5578         error = ip_fragment(ip, &m0, ifp->if_mtu, ifp->if_hwassist, sw_csum);
5579         if (error) {
5580                 goto bad;
5581         }
5582
5583         for (m0 = m1; m0; m0 = m1) {
5584                 m1 = m0->m_nextpkt;
5585                 m0->m_nextpkt = 0;
5586                 if (error == 0) {
5587                         lwkt_reltoken(&pf_token);
5588                         error = (*ifp->if_output)(ifp, m0, sintosa(dst),
5589                                                   NULL);
5590                         lwkt_gettoken(&pf_token);
5591                 } else
5592                         m_freem(m0);
5593         }
5594
5595         if (error == 0)
5596                 ipstat.ips_fragmented++;
5597
5598 done:
5599         if (r->rt != PF_DUPTO)
5600                 *m = NULL;
5601         if (ro == &iproute && ro->ro_rt)
5602                 RTFREE(ro->ro_rt);
5603         return;
5604
5605 bad:
5606         m_freem(m0);
5607         goto done;
5608 }
5609 #endif /* INET */
5610
5611 #ifdef INET6
5612 void
5613 pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5614     struct pf_state *s, struct pf_pdesc *pd)
5615 {
5616         struct mbuf             *m0;
5617         struct route_in6         ip6route;
5618         struct route_in6        *ro;
5619         struct sockaddr_in6     *dst;
5620         struct ip6_hdr          *ip6;
5621         struct ifnet            *ifp = NULL;
5622         struct pf_addr           naddr;
5623         struct pf_src_node      *sn = NULL;
5624         int                      error = 0;
5625
5626         if (m == NULL || *m == NULL || r == NULL ||
5627             (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5628                 panic("pf_route6: invalid parameters");
5629
5630         if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
5631                 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
5632                 (*m)->m_pkthdr.pf.routed = 1;
5633         } else {
5634                 if ((*m)->m_pkthdr.pf.routed++ > 3) {
5635                         m0 = *m;
5636                         *m = NULL;
5637                         goto bad;
5638                 }
5639         }
5640
5641         if (r->rt == PF_DUPTO) {
5642                 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL)
5643                         return;
5644         } else {
5645                 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5646                         return;
5647                 m0 = *m;
5648         }
5649
5650         if (m0->m_len < sizeof(struct ip6_hdr)) {
5651                 DPFPRINTF(PF_DEBUG_URGENT,
5652                     ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n"));
5653                 goto bad;
5654         }
5655         ip6 = mtod(m0, struct ip6_hdr *);
5656
5657         ro = &ip6route;
5658         bzero((caddr_t)ro, sizeof(*ro));
5659         dst = (struct sockaddr_in6 *)&ro->ro_dst;
5660         dst->sin6_family = AF_INET6;
5661         dst->sin6_len = sizeof(*dst);
5662         dst->sin6_addr = ip6->ip6_dst;
5663
5664         /*
5665          * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags,
5666          * so make sure pf.flags is clear.
5667          *
5668          * Cheat. XXX why only in the v6 case???
5669          */
5670         if (r->rt == PF_FASTROUTE) {
5671                 m0->m_pkthdr.fw_flags |= PF_MBUF_TAGGED;
5672                 m0->m_pkthdr.pf.flags = 0;
5673                 /* XXX Re-Check when Upgrading to > 4.4 */
5674                 m0->m_pkthdr.pf.statekey = NULL;
5675                 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
5676                 return;
5677         }
5678
5679         if (TAILQ_EMPTY(&r->rpool.list)) {
5680                 DPFPRINTF(PF_DEBUG_URGENT,
5681                     ("pf_route6: TAILQ_EMPTY(&r->rpool.list)\n"));
5682                 goto bad;
5683         }
5684         if (s == NULL) {
5685                 pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src,
5686                     &naddr, NULL, &sn);
5687                 if (!PF_AZERO(&naddr, AF_INET6))
5688                         PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5689                             &naddr, AF_INET6);
5690                 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL;
5691         } else {
5692                 if (!PF_AZERO(&s->rt_addr, AF_INET6))
5693                         PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5694                             &s->rt_addr, AF_INET6);
5695                 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5696         }
5697         if (ifp == NULL)
5698                 goto bad;
5699
5700         if (oifp != ifp) {
5701                 if (pf_test6(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) {
5702                         goto bad;
5703                 } else if (m0 == NULL) {
5704                         goto done;
5705                 }
5706                 if (m0->m_len < sizeof(struct ip6_hdr)) {
5707                         DPFPRINTF(PF_DEBUG_URGENT,
5708                             ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n"));
5709                         goto bad;
5710                 }
5711                 ip6 = mtod(m0, struct ip6_hdr *);
5712         }
5713
5714         /*
5715          * If the packet is too large for the outgoing interface,
5716          * send back an icmp6 error.
5717          */
5718         if (IN6_IS_ADDR_LINKLOCAL(&dst->sin6_addr))
5719                 dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
5720         if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) {
5721                 error = nd6_output(ifp, ifp, m0, dst, NULL);
5722         } else {
5723                 in6_ifstat_inc(ifp, ifs6_in_toobig);
5724                 if (r->rt != PF_DUPTO)
5725                         icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
5726                 else
5727                         goto bad;
5728         }
5729
5730 done:
5731         if (r->rt != PF_DUPTO)
5732                 *m = NULL;
5733         return;
5734
5735 bad:
5736         m_freem(m0);
5737         goto done;
5738 }
5739 #endif /* INET6 */
5740
5741
5742 /*
5743  * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag
5744  *   off is the offset where the protocol header starts
5745  *   len is the total length of protocol header plus payload
5746  * returns 0 when the checksum is valid, otherwise returns 1.
5747  */
5748 /*
5749  * XXX
5750  * FreeBSD supports cksum offload for the following drivers.
5751  * em(4), gx(4), lge(4), nge(4), ti(4), xl(4)
5752  * If we can make full use of it we would outperform ipfw/ipfilter in
5753  * very heavy traffic.
5754  * I have not tested 'cause I don't have NICs that supports cksum offload.
5755  * (There might be problems. Typical phenomena would be
5756  *   1. No route message for UDP packet.
5757  *   2. No connection acceptance from external hosts regardless of rule set.)
5758  */
5759 int
5760 pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p,
5761     sa_family_t af)
5762 {
5763         u_int16_t sum = 0;
5764         int hw_assist = 0;
5765         struct ip *ip;
5766
5767         if (off < sizeof(struct ip) || len < sizeof(struct udphdr))
5768                 return (1);
5769         if (m->m_pkthdr.len < off + len)
5770                 return (1);
5771
5772         switch (p) {
5773         case IPPROTO_TCP:
5774         case IPPROTO_UDP:
5775                 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
5776                         if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) {
5777                                 sum = m->m_pkthdr.csum_data;
5778                         } else {
5779                                 ip = mtod(m, struct ip *);
5780                                 sum = in_pseudo(ip->ip_src.s_addr,
5781                                         ip->ip_dst.s_addr, htonl((u_short)len +
5782                                         m->m_pkthdr.csum_data + p));
5783                         }
5784                         sum ^= 0xffff;
5785                         ++hw_assist;
5786                 }
5787                 break;
5788         case IPPROTO_ICMP:
5789 #ifdef INET6
5790         case IPPROTO_ICMPV6:
5791 #endif /* INET6 */
5792                 break;
5793         default:
5794                 return (1);
5795         }
5796
5797         if (!hw_assist) {
5798                 switch (af) {
5799                 case AF_INET:
5800                         if (p == IPPROTO_ICMP) {
5801                                 if (m->m_len < off)
5802                                         return (1);
5803                                 m->m_data += off;
5804                                 m->m_len -= off;
5805                                 sum = in_cksum(m, len);
5806                                 m->m_data -= off;
5807                                 m->m_len += off;
5808                         } else {
5809                                 if (m->m_len < sizeof(struct ip))
5810                                         return (1);
5811                                 sum = in_cksum_range(m, p, off, len);
5812                                 if (sum == 0) {
5813                                         m->m_pkthdr.csum_flags |=
5814                                             (CSUM_DATA_VALID |
5815                                              CSUM_PSEUDO_HDR);
5816                                         m->m_pkthdr.csum_data = 0xffff;
5817                                 }
5818                         }
5819                         break;
5820 #ifdef INET6
5821                 case AF_INET6:
5822                         if (m->m_len < sizeof(struct ip6_hdr))
5823                                 return (1);
5824                         sum = in6_cksum(m, p, off, len);
5825                         /*
5826                          * XXX
5827                          * IPv6 H/W cksum off-load not supported yet!
5828                          *
5829                          * if (sum == 0) {
5830                          *      m->m_pkthdr.csum_flags |=
5831                          *          (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
5832                          *      m->m_pkthdr.csum_data = 0xffff;
5833                          *}
5834                          */
5835                         break;
5836 #endif /* INET6 */
5837                 default:
5838                         return (1);
5839                 }
5840         }
5841         if (sum) {
5842                 switch (p) {
5843                 case IPPROTO_TCP:
5844                         tcpstat.tcps_rcvbadsum++;
5845                         break;
5846                 case IPPROTO_UDP:
5847                         udpstat.udps_badsum++;
5848                         break;
5849                 case IPPROTO_ICMP:
5850                         icmpstat.icps_checksum++;
5851                         break;
5852 #ifdef INET6
5853                 case IPPROTO_ICMPV6:
5854                         icmp6stat.icp6s_checksum++;
5855                         break;
5856 #endif /* INET6 */
5857                 }
5858                 return (1);
5859         }
5860         return (0);
5861 }
5862
5863 struct pf_divert *
5864 pf_find_divert(struct mbuf *m)
5865 {
5866         struct m_tag    *mtag;
5867
5868         if ((mtag = m_tag_find(m, PACKET_TAG_PF_DIVERT, NULL)) == NULL)
5869                 return (NULL);
5870
5871         return ((struct pf_divert *)(mtag + 1));
5872 }
5873
5874 struct pf_divert *
5875 pf_get_divert(struct mbuf *m)
5876 {
5877         struct m_tag    *mtag;
5878
5879         if ((mtag = m_tag_find(m, PACKET_TAG_PF_DIVERT, NULL)) == NULL) {
5880                 mtag = m_tag_get(PACKET_TAG_PF_DIVERT, sizeof(struct pf_divert),
5881                     M_NOWAIT);
5882                 if (mtag == NULL)
5883                         return (NULL);
5884                 bzero(mtag + 1, sizeof(struct pf_divert));
5885                 m_tag_prepend(m, mtag);
5886         }
5887
5888         return ((struct pf_divert *)(mtag + 1));
5889 }
5890
5891 #ifdef INET
5892 int
5893 pf_test(int dir, struct ifnet *ifp, struct mbuf **m0,
5894     struct ether_header *eh, struct inpcb *inp)
5895 {
5896         struct pfi_kif          *kif;
5897         u_short                  action, reason = 0, log = 0;
5898         struct mbuf             *m = *m0;
5899         struct ip               *h = NULL;
5900         struct pf_rule          *a = NULL, *r = &pf_default_rule, *tr, *nr;
5901         struct pf_state         *s = NULL;
5902         struct pf_ruleset       *ruleset = NULL;
5903         struct pf_pdesc          pd;
5904         int                      off, dirndx, pqid = 0;
5905
5906         if (!pf_status.running)
5907                 return (PF_PASS);
5908
5909         memset(&pd, 0, sizeof(pd));
5910         if (ifp->if_type == IFT_CARP && ifp->if_carpdev)
5911                 kif = (struct pfi_kif *)ifp->if_carpdev->if_pf_kif;
5912         else
5913                 kif = (struct pfi_kif *)ifp->if_pf_kif;
5914
5915         if (kif == NULL) {
5916                 DPFPRINTF(PF_DEBUG_URGENT,
5917                     ("pf_test: kif == NULL, if_xname %s\n", ifp->if_xname));
5918                 return (PF_DROP);
5919         }
5920         if (kif->pfik_flags & PFI_IFLAG_SKIP)
5921                 return (PF_PASS);
5922
5923 #ifdef DIAGNOSTIC
5924         if ((m->m_flags & M_PKTHDR) == 0)
5925                 panic("non-M_PKTHDR is passed to pf_test");
5926 #endif /* DIAGNOSTIC */
5927
5928         if (m->m_pkthdr.len < (int)sizeof(*h)) {
5929                 action = PF_DROP;
5930                 REASON_SET(&reason, PFRES_SHORT);
5931                 log = 1;
5932                 goto done;
5933         }
5934
5935         /*
5936          * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags,
5937          * so make sure pf.flags is clear.
5938          */
5939         if (m->m_pkthdr.fw_flags & PF_MBUF_TAGGED)
5940                 return (PF_PASS);
5941         m->m_pkthdr.pf.flags = 0;
5942         /* Re-Check when updating to > 4.4 */
5943         m->m_pkthdr.pf.statekey = NULL;
5944
5945         /* We do IP header normalization and packet reassembly here */
5946         if (pf_normalize_ip(m0, dir, kif, &reason, &pd) != PF_PASS) {
5947                 action = PF_DROP;
5948                 goto done;
5949         }
5950         m = *m0;        /* pf_normalize messes with m0 */
5951         h = mtod(m, struct ip *);
5952
5953         off = h->ip_hl << 2;
5954         if (off < (int)sizeof(*h)) {
5955                 action = PF_DROP;
5956                 REASON_SET(&reason, PFRES_SHORT);
5957                 log = 1;
5958                 goto done;
5959         }
5960
5961         pd.src = (struct pf_addr *)&h->ip_src;
5962         pd.dst = (struct pf_addr *)&h->ip_dst;
5963         pd.sport = pd.dport = NULL;
5964         pd.ip_sum = &h->ip_sum;
5965         pd.proto_sum = NULL;
5966         pd.proto = h->ip_p;
5967         pd.dir = dir;
5968         pd.sidx = (dir == PF_IN) ? 0 : 1;
5969         pd.didx = (dir == PF_IN) ? 1 : 0;
5970         pd.af = AF_INET;
5971         pd.tos = h->ip_tos;
5972         pd.tot_len = h->ip_len;
5973         pd.eh = eh;
5974
5975         /* handle fragments that didn't get reassembled by normalization */
5976         if (h->ip_off & (IP_MF | IP_OFFMASK)) {
5977                 action = pf_test_fragment(&r, dir, kif, m, h,
5978                     &pd, &a, &ruleset);
5979                 goto done;
5980         }
5981
5982         switch (h->ip_p) {
5983
5984         case IPPROTO_TCP: {
5985                 struct tcphdr   th;
5986
5987                 pd.hdr.tcp = &th;
5988                 if (!pf_pull_hdr(m, off, &th, sizeof(th),
5989                     &action, &reason, AF_INET)) {
5990                         log = action != PF_PASS;
5991                         goto done;
5992                 }
5993                 pd.p_len = pd.tot_len - off - (th.th_off << 2);
5994                 if ((th.th_flags & TH_ACK) && pd.p_len == 0)
5995                         pqid = 1;
5996                 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
5997                 if (action == PF_DROP)
5998                         goto done;
5999                 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
6000                     &reason);
6001                 if (action == PF_PASS) {
6002                         pfsync_update_state(s);
6003                         r = s->rule.ptr;
6004                         a = s->anchor.ptr;
6005                         log = s->log;
6006                 } else if (s == NULL)
6007                         action = pf_test_rule(&r, &s, dir, kif,
6008                             m, off, h, &pd, &a, &ruleset, NULL, inp);
6009                 break;
6010         }
6011
6012         case IPPROTO_UDP: {
6013                 struct udphdr   uh;
6014
6015                 pd.hdr.udp = &uh;
6016                 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
6017                     &action, &reason, AF_INET)) {
6018                         log = action != PF_PASS;
6019                         goto done;
6020                 }
6021                 if (uh.uh_dport == 0 ||
6022                     ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
6023                     ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
6024                         action = PF_DROP;
6025                         REASON_SET(&reason, PFRES_SHORT);
6026                         goto done;
6027                 }
6028                 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
6029                 if (action == PF_PASS) {
6030                         pfsync_update_state(s);
6031                         r = s->rule.ptr;
6032                         a = s->anchor.ptr;
6033                         log = s->log;
6034                 } else if (s == NULL)
6035                         action = pf_test_rule(&r, &s, dir, kif,
6036                             m, off, h, &pd, &a, &ruleset, NULL, inp);
6037                 break;
6038         }
6039
6040         case IPPROTO_ICMP: {
6041                 struct icmp     ih;
6042
6043                 pd.hdr.icmp = &ih;
6044                 if (!pf_pull_hdr(m, off, &ih, ICMP_MINLEN,
6045                     &action, &reason, AF_INET)) {
6046                         log = action != PF_PASS;
6047                         goto done;
6048                 }
6049                 action = pf_test_state_icmp(&s, dir, kif, m, off, h, &pd,
6050                     &reason);
6051                 if (action == PF_PASS) {
6052                         pfsync_update_state(s);
6053                         r = s->rule.ptr;
6054                         a = s->anchor.ptr;
6055                         log = s->log;
6056                 } else if (s == NULL)
6057                         action = pf_test_rule(&r, &s, dir, kif,
6058                             m, off, h, &pd, &a, &ruleset, NULL, inp);
6059                 break;
6060         }
6061
6062         default:
6063                 action = pf_test_state_other(&s, dir, kif, m, &pd);
6064                 if (action == PF_PASS) {
6065                         pfsync_update_state(s);
6066                         r = s->rule.ptr;
6067                         a = s->anchor.ptr;
6068                         log = s->log;
6069                 } else if (s == NULL)
6070                         action = pf_test_rule(&r, &s, dir, kif, m, off, h,
6071                             &pd, &a, &ruleset, NULL, inp);
6072                 break;
6073         }
6074
6075 done:
6076         if (action == PF_PASS && h->ip_hl > 5 &&
6077             !((s && s->state_flags & PFSTATE_ALLOWOPTS) || r->allow_opts)) {
6078                 action = PF_DROP;
6079                 REASON_SET(&reason, PFRES_IPOPTIONS);
6080                 log = 1;
6081                 DPFPRINTF(PF_DEBUG_MISC,
6082                     ("pf: dropping packet with ip options\n"));
6083         }
6084
6085         if ((s && s->tag) || r->rtableid)
6086                 pf_tag_packet(m, s ? s->tag : 0, r->rtableid);
6087
6088 #if 0
6089         if (dir == PF_IN && s && s->key[PF_SK_STACK])
6090                 m->m_pkthdr.pf.statekey = s->key[PF_SK_STACK];
6091 #endif
6092
6093 #ifdef ALTQ
6094         if (action == PF_PASS && r->qid) {
6095                 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
6096                 if (pqid || (pd.tos & IPTOS_LOWDELAY))
6097                         m->m_pkthdr.pf.qid = r->pqid;
6098                 else
6099                         m->m_pkthdr.pf.qid = r->qid;
6100                 m->m_pkthdr.pf.ecn_af = AF_INET;
6101                 m->m_pkthdr.pf.hdr = h;
6102                 /* add connection hash for fairq */
6103                 if (s) {
6104                         /* for fairq */
6105                         m->m_pkthdr.pf.state_hash = s->hash;
6106                         m->m_pkthdr.pf.flags |= PF_TAG_STATE_HASHED;
6107                 }
6108         }
6109 #endif /* ALTQ */
6110
6111         /*
6112          * connections redirected to loopback should not match sockets
6113          * bound specifically to loopback due to security implications,
6114          * see tcp_input() and in_pcblookup_listen().
6115          */
6116         if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
6117             pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
6118             (s->nat_rule.ptr->action == PF_RDR ||
6119             s->nat_rule.ptr->action == PF_BINAT) &&
6120             (ntohl(pd.dst->v4.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET)
6121                 m->m_pkthdr.pf.flags |= PF_TAG_TRANSLATE_LOCALHOST;
6122
6123         if (dir == PF_IN && action == PF_PASS && r->divert.port) {
6124                 struct pf_divert *divert;
6125
6126                 if ((divert = pf_get_divert(m))) {
6127                         m->m_pkthdr.pf.flags |= PF_TAG_DIVERTED;
6128                         divert->port = r->divert.port;
6129                         divert->addr.ipv4 = r->divert.addr.v4;
6130                 }
6131         }
6132
6133         if (log) {
6134                 struct pf_rule *lr;
6135
6136                 if (s != NULL && s->nat_rule.ptr != NULL &&
6137                     s->nat_rule.ptr->log & PF_LOG_ALL)
6138                         lr = s->nat_rule.ptr;
6139                 else
6140                         lr = r;
6141                 PFLOG_PACKET(kif, h, m, AF_INET, dir, reason, lr, a, ruleset,
6142                     &pd);
6143         }
6144
6145         kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
6146         kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++;
6147
6148         if (action == PF_PASS || r->action == PF_DROP) {
6149                 dirndx = (dir == PF_OUT);
6150                 r->packets[dirndx]++;
6151                 r->bytes[dirndx] += pd.tot_len;
6152                 if (a != NULL) {
6153                         a->packets[dirndx]++;
6154                         a->bytes[dirndx] += pd.tot_len;
6155                 }
6156                 if (s != NULL) {
6157                         if (s->nat_rule.ptr != NULL) {
6158                                 s->nat_rule.ptr->packets[dirndx]++;
6159                                 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len;
6160                         }
6161                         if (s->src_node != NULL) {
6162                                 s->src_node->packets[dirndx]++;
6163                                 s->src_node->bytes[dirndx] += pd.tot_len;
6164                         }
6165                         if (s->nat_src_node != NULL) {
6166                                 s->nat_src_node->packets[dirndx]++;
6167                                 s->nat_src_node->bytes[dirndx] += pd.tot_len;
6168                         }
6169                         dirndx = (dir == s->direction) ? 0 : 1;
6170                         s->packets[dirndx]++;
6171                         s->bytes[dirndx] += pd.tot_len;
6172                 }
6173                 tr = r;
6174                 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
6175                 if (nr != NULL && r == &pf_default_rule)
6176                         tr = nr;
6177                 if (tr->src.addr.type == PF_ADDR_TABLE)
6178                         pfr_update_stats(tr->src.addr.p.tbl,
6179                             (s == NULL) ? pd.src :
6180                             &s->key[(s->direction == PF_IN)]->
6181                                 addr[(s->direction == PF_OUT)],
6182                             pd.af, pd.tot_len, dir == PF_OUT,
6183                             r->action == PF_PASS, tr->src.neg);
6184                 if (tr->dst.addr.type == PF_ADDR_TABLE)
6185                         pfr_update_stats(tr->dst.addr.p.tbl,
6186                             (s == NULL) ? pd.dst :
6187                             &s->key[(s->direction == PF_IN)]->
6188                                 addr[(s->direction == PF_IN)],
6189                             pd.af, pd.tot_len, dir == PF_OUT,
6190                             r->action == PF_PASS, tr->dst.neg);
6191         }
6192
6193
6194         if (action == PF_SYNPROXY_DROP) {
6195                 m_freem(*m0);
6196                 *m0 = NULL;
6197                 action = PF_PASS;
6198         } else if (r->rt)
6199                 /* pf_route can free the mbuf causing *m0 to become NULL */
6200                 pf_route(m0, r, dir, kif->pfik_ifp, s, &pd);
6201
6202         return (action);
6203 }
6204 #endif /* INET */
6205
6206 #ifdef INET6
6207 int
6208 pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0,
6209     struct ether_header *eh, struct inpcb *inp)
6210 {
6211         struct pfi_kif          *kif;
6212         u_short                  action, reason = 0, log = 0;
6213         struct mbuf             *m = *m0, *n = NULL;
6214         struct ip6_hdr          *h = NULL;
6215         struct pf_rule          *a = NULL, *r = &pf_default_rule, *tr, *nr;
6216         struct pf_state         *s = NULL;
6217         struct pf_ruleset       *ruleset = NULL;
6218         struct pf_pdesc          pd;
6219         int                      off, terminal = 0, dirndx, rh_cnt = 0;
6220
6221         if (!pf_status.running)
6222                 return (PF_PASS);
6223
6224         memset(&pd, 0, sizeof(pd));
6225         if (ifp->if_type == IFT_CARP && ifp->if_carpdev)
6226                 kif = (struct pfi_kif *)ifp->if_carpdev->if_pf_kif;
6227         else
6228                 kif = (struct pfi_kif *)ifp->if_pf_kif;
6229
6230         if (kif == NULL) {
6231                 DPFPRINTF(PF_DEBUG_URGENT,
6232                     ("pf_test6: kif == NULL, if_xname %s\n", ifp->if_xname));
6233                 return (PF_DROP);
6234         }
6235         if (kif->pfik_flags & PFI_IFLAG_SKIP)
6236                 return (PF_PASS);
6237
6238 #ifdef DIAGNOSTIC
6239         if ((m->m_flags & M_PKTHDR) == 0)
6240                 panic("non-M_PKTHDR is passed to pf_test6");
6241 #endif /* DIAGNOSTIC */
6242
6243         if (m->m_pkthdr.len < (int)sizeof(*h)) {
6244                 action = PF_DROP;
6245                 REASON_SET(&reason, PFRES_SHORT);
6246                 log = 1;
6247                 goto done;
6248         }
6249
6250         /*
6251          * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags,
6252          * so make sure pf.flags is clear.
6253          */
6254         if (m->m_pkthdr.fw_flags & PF_MBUF_TAGGED)
6255                 return (PF_PASS);
6256         m->m_pkthdr.pf.flags = 0;
6257         /* Re-Check when updating to > 4.4 */
6258         m->m_pkthdr.pf.statekey = NULL;
6259
6260         /* We do IP header normalization and packet reassembly here */
6261         if (pf_normalize_ip6(m0, dir, kif, &reason, &pd) != PF_PASS) {
6262                 action = PF_DROP;
6263                 goto done;
6264         }
6265         m = *m0;        /* pf_normalize messes with m0 */
6266         h = mtod(m, struct ip6_hdr *);
6267
6268 #if 1
6269         /*
6270          * we do not support jumbogram yet.  if we keep going, zero ip6_plen
6271          * will do something bad, so drop the packet for now.
6272          */
6273         if (htons(h->ip6_plen) == 0) {
6274                 action = PF_DROP;
6275                 REASON_SET(&reason, PFRES_NORM);        /*XXX*/
6276                 goto done;
6277         }
6278 #endif
6279
6280         pd.src = (struct pf_addr *)&h->ip6_src;
6281         pd.dst = (struct pf_addr *)&h->ip6_dst;
6282         pd.sport = pd.dport = NULL;
6283         pd.ip_sum = NULL;
6284         pd.proto_sum = NULL;
6285         pd.dir = dir;
6286         pd.sidx = (dir == PF_IN) ? 0 : 1;
6287         pd.didx = (dir == PF_IN) ? 1 : 0;
6288         pd.af = AF_INET6;
6289         pd.tos = 0;
6290         pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr);
6291         pd.eh = eh;
6292
6293         off = ((caddr_t)h - m->m_data) + sizeof(struct ip6_hdr);
6294         pd.proto = h->ip6_nxt;
6295         do {
6296                 switch (pd.proto) {
6297                 case IPPROTO_FRAGMENT:
6298                         action = pf_test_fragment(&r, dir, kif, m, h,
6299                             &pd, &a, &ruleset);
6300                         if (action == PF_DROP)
6301                                 REASON_SET(&reason, PFRES_FRAG);
6302                         goto done;
6303                 case IPPROTO_ROUTING: {
6304                         struct ip6_rthdr rthdr;
6305
6306                         if (rh_cnt++) {
6307                                 DPFPRINTF(PF_DEBUG_MISC,
6308                                     ("pf: IPv6 more than one rthdr\n"));
6309                                 action = PF_DROP;
6310                                 REASON_SET(&reason, PFRES_IPOPTIONS);
6311                                 log = 1;
6312                                 goto done;
6313                         }
6314                         if (!pf_pull_hdr(m, off, &rthdr, sizeof(rthdr), NULL,
6315                             &reason, pd.af)) {
6316                                 DPFPRINTF(PF_DEBUG_MISC,
6317                                     ("pf: IPv6 short rthdr\n"));
6318                                 action = PF_DROP;
6319                                 REASON_SET(&reason, PFRES_SHORT);
6320                                 log = 1;
6321                                 goto done;
6322                         }
6323                         if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) {
6324                                 DPFPRINTF(PF_DEBUG_MISC,
6325                                     ("pf: IPv6 rthdr0\n"));
6326                                 action = PF_DROP;
6327                                 REASON_SET(&reason, PFRES_IPOPTIONS);
6328                                 log = 1;
6329                                 goto done;
6330                         }
6331                         /* FALLTHROUGH */
6332                 }
6333                 case IPPROTO_AH:
6334                 case IPPROTO_HOPOPTS:
6335                 case IPPROTO_DSTOPTS: {
6336                         /* get next header and header length */
6337                         struct ip6_ext  opt6;
6338
6339                         if (!pf_pull_hdr(m, off, &opt6, sizeof(opt6),
6340                             NULL, &reason, pd.af)) {
6341                                 DPFPRINTF(PF_DEBUG_MISC,
6342                                     ("pf: IPv6 short opt\n"));
6343                                 action = PF_DROP;
6344                                 log = 1;
6345                                 goto done;
6346                         }
6347                         if (pd.proto == IPPROTO_AH)
6348                                 off += (opt6.ip6e_len + 2) * 4;
6349                         else
6350                                 off += (opt6.ip6e_len + 1) * 8;
6351                         pd.proto = opt6.ip6e_nxt;
6352                         /* goto the next header */
6353                         break;
6354                 }
6355                 default:
6356                         terminal++;
6357                         break;
6358                 }
6359         } while (!terminal);
6360
6361         /* if there's no routing header, use unmodified mbuf for checksumming */
6362         if (!n)
6363                 n = m;
6364
6365         switch (pd.proto) {
6366
6367         case IPPROTO_TCP: {
6368                 struct tcphdr   th;
6369
6370                 pd.hdr.tcp = &th;
6371                 if (!pf_pull_hdr(m, off, &th, sizeof(th),
6372                     &action, &reason, AF_INET6)) {
6373                         log = action != PF_PASS;
6374                         goto done;
6375                 }
6376                 pd.p_len = pd.tot_len - off - (th.th_off << 2);
6377                 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
6378                 if (action == PF_DROP)
6379                         goto done;
6380                 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
6381                     &reason);
6382                 if (action == PF_PASS) {
6383                         pfsync_update_state(s);
6384                         r = s->rule.ptr;
6385                         a = s->anchor.ptr;
6386                         log = s->log;
6387                 } else if (s == NULL)
6388                         action = pf_test_rule(&r, &s, dir, kif,
6389                             m, off, h, &pd, &a, &ruleset, NULL, inp);
6390                 break;
6391         }
6392
6393         case IPPROTO_UDP: {
6394                 struct udphdr   uh;
6395
6396                 pd.hdr.udp = &uh;
6397                 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
6398                     &action, &reason, AF_INET6)) {
6399                         log = action != PF_PASS;
6400                         goto done;
6401                 }
6402                 if (uh.uh_dport == 0 ||
6403                     ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
6404                     ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
6405                         action = PF_DROP;
6406                         REASON_SET(&reason, PFRES_SHORT);
6407                         goto done;
6408                 }
6409                 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
6410                 if (action == PF_PASS) {
6411                         pfsync_update_state(s);
6412                         r = s->rule.ptr;
6413                         a = s->anchor.ptr;
6414                         log = s->log;
6415                 } else if (s == NULL)
6416                         action = pf_test_rule(&r, &s, dir, kif,
6417                             m, off, h, &pd, &a, &ruleset, NULL, inp);
6418                 break;
6419         }
6420
6421         case IPPROTO_ICMPV6: {
6422                 struct icmp6_hdr        ih;
6423
6424                 pd.hdr.icmp6 = &ih;
6425                 if (!pf_pull_hdr(m, off, &ih, sizeof(ih),
6426                     &action, &reason, AF_INET6)) {
6427                         log = action != PF_PASS;
6428                         goto done;
6429                 }
6430                 action = pf_test_state_icmp(&s, dir, kif,
6431                     m, off, h, &pd, &reason);
6432                 if (action == PF_PASS) {
6433                         pfsync_update_state(s);
6434                         r = s->rule.ptr;
6435                         a = s->anchor.ptr;
6436                         log = s->log;
6437                 } else if (s == NULL)
6438                         action = pf_test_rule(&r, &s, dir, kif,
6439                             m, off, h, &pd, &a, &ruleset, NULL, inp);
6440                 break;
6441         }
6442
6443         default:
6444                 action = pf_test_state_other(&s, dir, kif, m, &pd);
6445                 if (action == PF_PASS) {
6446                         pfsync_update_state(s);
6447                         r = s->rule.ptr;
6448                         a = s->anchor.ptr;
6449                         log = s->log;
6450                 } else if (s == NULL)
6451                         action = pf_test_rule(&r, &s, dir, kif, m, off, h,
6452                             &pd, &a, &ruleset, NULL, inp);
6453                 break;
6454         }
6455
6456 done:
6457         if (n != m) {
6458                 m_freem(n);
6459                 n = NULL;
6460         }
6461
6462         /* handle dangerous IPv6 extension headers. */
6463         if (action == PF_PASS && rh_cnt &&
6464             !((s && s->state_flags & PFSTATE_ALLOWOPTS) || r->allow_opts)) {
6465                 action = PF_DROP;
6466                 REASON_SET(&reason, PFRES_IPOPTIONS);
6467                 log = 1;
6468                 DPFPRINTF(PF_DEBUG_MISC,
6469                     ("pf: dropping packet with dangerous v6 headers\n"));
6470         }
6471
6472         if ((s && s->tag) || r->rtableid)
6473                 pf_tag_packet(m, s ? s->tag : 0, r->rtableid);
6474
6475 #if 0
6476         if (dir == PF_IN && s && s->key[PF_SK_STACK])
6477                 m->m_pkthdr.pf.statekey = s->key[PF_SK_STACK];
6478 #endif
6479
6480 #ifdef ALTQ
6481         if (action == PF_PASS && r->qid) {
6482                 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
6483                 if (pd.tos & IPTOS_LOWDELAY)
6484                         m->m_pkthdr.pf.qid = r->pqid;
6485                 else
6486                         m->m_pkthdr.pf.qid = r->qid;
6487                 m->m_pkthdr.pf.ecn_af = AF_INET6;
6488                 m->m_pkthdr.pf.hdr = h;
6489                 if (s) {
6490                         /* for fairq */
6491                         m->m_pkthdr.pf.state_hash = s->hash;
6492                         m->m_pkthdr.pf.flags |= PF_TAG_STATE_HASHED;
6493                 }
6494         }
6495 #endif /* ALTQ */
6496
6497         if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
6498             pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
6499             (s->nat_rule.ptr->action == PF_RDR ||
6500             s->nat_rule.ptr->action == PF_BINAT) &&
6501             IN6_IS_ADDR_LOOPBACK(&pd.dst->v6))
6502                 m->m_pkthdr.pf.flags |= PF_TAG_TRANSLATE_LOCALHOST;
6503
6504         if (dir == PF_IN && action == PF_PASS && r->divert.port) {
6505                 struct pf_divert *divert;
6506
6507                 if ((divert = pf_get_divert(m))) {
6508                         m->m_pkthdr.pf.flags |= PF_TAG_DIVERTED;
6509                         divert->port = r->divert.port;
6510                         divert->addr.ipv6 = r->divert.addr.v6;
6511                 }
6512         }
6513
6514         if (log) {
6515                 struct pf_rule *lr;
6516
6517                 if (s != NULL && s->nat_rule.ptr != NULL &&
6518                     s->nat_rule.ptr->log & PF_LOG_ALL)
6519                         lr = s->nat_rule.ptr;
6520                 else
6521                         lr = r;
6522                 PFLOG_PACKET(kif, h, m, AF_INET6, dir, reason, lr, a, ruleset,
6523                     &pd);
6524         }
6525
6526         kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
6527         kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++;
6528
6529         if (action == PF_PASS || r->action == PF_DROP) {
6530                 dirndx = (dir == PF_OUT);
6531                 r->packets[dirndx]++;
6532                 r->bytes[dirndx] += pd.tot_len;
6533                 if (a != NULL) {
6534                         a->packets[dirndx]++;
6535                         a->bytes[dirndx] += pd.tot_len;
6536                 }
6537                 if (s != NULL) {
6538                         if (s->nat_rule.ptr != NULL) {
6539                                 s->nat_rule.ptr->packets[dirndx]++;
6540                                 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len;
6541                         }
6542                         if (s->src_node != NULL) {
6543                                 s->src_node->packets[dirndx]++;
6544                                 s->src_node->bytes[dirndx] += pd.tot_len;
6545                         }
6546                         if (s->nat_src_node != NULL) {
6547                                 s->nat_src_node->packets[dirndx]++;
6548                                 s->nat_src_node->bytes[dirndx] += pd.tot_len;
6549                         }
6550                         dirndx = (dir == s->direction) ? 0 : 1;
6551                         s->packets[dirndx]++;
6552                         s->bytes[dirndx] += pd.tot_len;
6553                 }
6554                 tr = r;
6555                 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
6556                 if (nr != NULL && r == &pf_default_rule)
6557                         tr = nr;
6558                 if (tr->src.addr.type == PF_ADDR_TABLE)
6559                         pfr_update_stats(tr->src.addr.p.tbl,
6560                             (s == NULL) ? pd.src :
6561                             &s->key[(s->direction == PF_IN)]->addr[0],
6562                             pd.af, pd.tot_len, dir == PF_OUT,
6563                             r->action == PF_PASS, tr->src.neg);
6564                 if (tr->dst.addr.type == PF_ADDR_TABLE)
6565                         pfr_update_stats(tr->dst.addr.p.tbl,
6566                             (s == NULL) ? pd.dst :
6567                             &s->key[(s->direction == PF_IN)]->addr[1],
6568                             pd.af, pd.tot_len, dir == PF_OUT,
6569                             r->action == PF_PASS, tr->dst.neg);
6570         }
6571
6572
6573         if (action == PF_SYNPROXY_DROP) {
6574                 m_freem(*m0);
6575                 *m0 = NULL;
6576                 action = PF_PASS;
6577         } else if (r->rt)
6578                 /* pf_route6 can free the mbuf causing *m0 to become NULL */
6579                 pf_route6(m0, r, dir, kif->pfik_ifp, s, &pd);
6580
6581         return (action);
6582 }
6583 #endif /* INET6 */
6584
6585 int
6586 pf_check_congestion(struct ifqueue *ifq)
6587 {
6588                 return (0);
6589 }