9c157d3fd48b809b0650f88bcd4238de70ade4a2
[dragonfly.git] / sys / net / pf / pf.c
1 /*      $OpenBSD: pf.c,v 1.614 2008/08/02 12:34:37 henning Exp $ */
2
3 /*
4  * Copyright (c) 2004 The DragonFly Project.  All rights reserved.
5  *
6  * Copyright (c) 2001 Daniel Hartmeier
7  * Copyright (c) 2002 - 2008 Henning Brauer
8  * All rights reserved.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  *
14  *    - Redistributions of source code must retain the above copyright
15  *      notice, this list of conditions and the following disclaimer.
16  *    - Redistributions in binary form must reproduce the above
17  *      copyright notice, this list of conditions and the following
18  *      disclaimer in the documentation and/or other materials provided
19  *      with the distribution.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
25  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
27  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
29  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
31  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
32  * POSSIBILITY OF SUCH DAMAGE.
33  *
34  * Effort sponsored in part by the Defense Advanced Research Projects
35  * Agency (DARPA) and Air Force Research Laboratory, Air Force
36  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
37  *
38  */
39
40 #include "opt_inet.h"
41 #include "opt_inet6.h"
42
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/malloc.h>
46 #include <sys/mbuf.h>
47 #include <sys/filio.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/kernel.h>
51 #include <sys/time.h>
52 #include <sys/sysctl.h>
53 #include <sys/endian.h>
54 #include <vm/vm_zone.h>
55 #include <sys/proc.h>
56 #include <sys/kthread.h>
57
58 #include <machine/inttypes.h>
59
60 #include <sys/md5.h>
61
62 #include <net/if.h>
63 #include <net/if_types.h>
64 #include <net/bpf.h>
65 #include <net/netisr.h>
66 #include <net/route.h>
67
68 #include <netinet/in.h>
69 #include <netinet/in_var.h>
70 #include <netinet/in_systm.h>
71 #include <netinet/ip.h>
72 #include <netinet/ip_var.h>
73 #include <netinet/tcp.h>
74 #include <netinet/tcp_seq.h>
75 #include <netinet/udp.h>
76 #include <netinet/ip_icmp.h>
77 #include <netinet/in_pcb.h>
78 #include <netinet/tcp_timer.h>
79 #include <netinet/tcp_var.h>
80 #include <netinet/udp_var.h>
81 #include <netinet/icmp_var.h>
82 #include <netinet/if_ether.h>
83
84 #include <net/pf/pfvar.h>
85 #include <net/pf/if_pflog.h>
86
87 #include <net/pf/if_pfsync.h>
88
89 #ifdef INET6
90 #include <netinet/ip6.h>
91 #include <netinet/in_pcb.h>
92 #include <netinet/icmp6.h>
93 #include <netinet6/nd6.h>
94 #include <netinet6/ip6_var.h>
95 #include <netinet6/in6_pcb.h>
96 #endif /* INET6 */
97
98 #include <sys/in_cksum.h>
99 #include <sys/ucred.h>
100 #include <machine/limits.h>
101 #include <sys/msgport2.h>
102 #include <net/netmsg2.h>
103
104 extern int ip_optcopy(struct ip *, struct ip *);
105 extern int debug_pfugidhack;
106
107 struct lwkt_token pf_token = LWKT_TOKEN_INITIALIZER(pf_token);
108
109 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) kprintf x
110
111 /*
112  * Global variables
113  */
114
115 /* mask radix tree */
116 struct radix_node_head  *pf_maskhead;
117
118 /* state tables */
119 struct pf_state_tree     pf_statetbl;
120
121 struct pf_altqqueue      pf_altqs[2];
122 struct pf_palist         pf_pabuf;
123 struct pf_altqqueue     *pf_altqs_active;
124 struct pf_altqqueue     *pf_altqs_inactive;
125 struct pf_status         pf_status;
126
127 u_int32_t                ticket_altqs_active;
128 u_int32_t                ticket_altqs_inactive;
129 int                      altqs_inactive_open;
130 u_int32_t                ticket_pabuf;
131
132 MD5_CTX                  pf_tcp_secret_ctx;
133 u_char                   pf_tcp_secret[16];
134 int                      pf_tcp_secret_init;
135 int                      pf_tcp_iss_off;
136
137 struct pf_anchor_stackframe {
138         struct pf_ruleset                       *rs;
139         struct pf_rule                          *r;
140         struct pf_anchor_node                   *parent;
141         struct pf_anchor                        *child;
142 } pf_anchor_stack[64];
143
144 vm_zone_t                pf_src_tree_pl, pf_rule_pl, pf_pooladdr_pl;
145 vm_zone_t                pf_state_pl, pf_state_key_pl, pf_state_item_pl;
146 vm_zone_t                pf_altq_pl;
147
148 void                     pf_print_host(struct pf_addr *, u_int16_t, u_int8_t);
149
150 void                     pf_init_threshold(struct pf_threshold *, u_int32_t,
151                             u_int32_t);
152 void                     pf_add_threshold(struct pf_threshold *);
153 int                      pf_check_threshold(struct pf_threshold *);
154
155 void                     pf_change_ap(struct pf_addr *, u_int16_t *,
156                             u_int16_t *, u_int16_t *, struct pf_addr *,
157                             u_int16_t, u_int8_t, sa_family_t);
158 int                      pf_modulate_sack(struct mbuf *, int, struct pf_pdesc *,
159                             struct tcphdr *, struct pf_state_peer *);
160 #ifdef INET6
161 void                     pf_change_a6(struct pf_addr *, u_int16_t *,
162                             struct pf_addr *, u_int8_t);
163 #endif /* INET6 */
164 void                     pf_change_icmp(struct pf_addr *, u_int16_t *,
165                             struct pf_addr *, struct pf_addr *, u_int16_t,
166                             u_int16_t *, u_int16_t *, u_int16_t *,
167                             u_int16_t *, u_int8_t, sa_family_t);
168 void                     pf_send_tcp(const struct pf_rule *, sa_family_t,
169                             const struct pf_addr *, const struct pf_addr *,
170                             u_int16_t, u_int16_t, u_int32_t, u_int32_t,
171                             u_int8_t, u_int16_t, u_int16_t, u_int8_t, int,
172                             u_int16_t, struct ether_header *, struct ifnet *);
173 void                     pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t,
174                             sa_family_t, struct pf_rule *);
175 struct pf_rule          *pf_match_translation(struct pf_pdesc *, struct mbuf *,
176                             int, int, struct pfi_kif *,
177                             struct pf_addr *, u_int16_t, struct pf_addr *,
178                             u_int16_t, int);
179 struct pf_rule          *pf_get_translation(struct pf_pdesc *, struct mbuf *,
180                             int, int, struct pfi_kif *, struct pf_src_node **,
181                             struct pf_state_key **, struct pf_state_key **,
182                             struct pf_state_key **, struct pf_state_key **,
183                             struct pf_addr *, struct pf_addr *,
184                             u_int16_t, u_int16_t);
185 void                     pf_detach_state(struct pf_state *);
186 int                      pf_state_key_setup(struct pf_pdesc *, struct pf_rule *,
187                             struct pf_state_key **, struct pf_state_key **,
188                             struct pf_state_key **, struct pf_state_key **,
189                             struct pf_addr *, struct pf_addr *,
190                             u_int16_t, u_int16_t);
191 void                     pf_state_key_detach(struct pf_state *, int);
192 u_int32_t                pf_tcp_iss(struct pf_pdesc *);
193 int                      pf_test_rule(struct pf_rule **, struct pf_state **,
194                             int, struct pfi_kif *, struct mbuf *, int,
195                             void *, struct pf_pdesc *, struct pf_rule **,
196                             struct pf_ruleset **, struct ifqueue *, struct inpcb *);
197 static __inline int      pf_create_state(struct pf_rule *, struct pf_rule *,
198                             struct pf_rule *, struct pf_pdesc *,
199                             struct pf_src_node *, struct pf_state_key *,
200                             struct pf_state_key *, struct pf_state_key *,
201                             struct pf_state_key *, struct mbuf *, int,
202                             u_int16_t, u_int16_t, int *, struct pfi_kif *,
203                             struct pf_state **, int, u_int16_t, u_int16_t,
204                             int);
205 int                      pf_test_fragment(struct pf_rule **, int,
206                             struct pfi_kif *, struct mbuf *, void *,
207                             struct pf_pdesc *, struct pf_rule **,
208                             struct pf_ruleset **);
209 int                      pf_tcp_track_full(struct pf_state_peer *,
210                             struct pf_state_peer *, struct pf_state **,
211                             struct pfi_kif *, struct mbuf *, int,
212                             struct pf_pdesc *, u_short *, int *);
213 int                     pf_tcp_track_sloppy(struct pf_state_peer *,
214                             struct pf_state_peer *, struct pf_state **,
215                             struct pf_pdesc *, u_short *);
216 int                      pf_test_state_tcp(struct pf_state **, int,
217                             struct pfi_kif *, struct mbuf *, int,
218                             void *, struct pf_pdesc *, u_short *);
219 int                      pf_test_state_udp(struct pf_state **, int,
220                             struct pfi_kif *, struct mbuf *, int,
221                             void *, struct pf_pdesc *);
222 int                      pf_test_state_icmp(struct pf_state **, int,
223                             struct pfi_kif *, struct mbuf *, int,
224                             void *, struct pf_pdesc *, u_short *);
225 int                      pf_test_state_other(struct pf_state **, int,
226                             struct pfi_kif *, struct mbuf *, struct pf_pdesc *);
227 void                     pf_step_into_anchor(int *, struct pf_ruleset **, int,
228                             struct pf_rule **, struct pf_rule **, int *);
229 int                      pf_step_out_of_anchor(int *, struct pf_ruleset **,
230                              int, struct pf_rule **, struct pf_rule **,
231                              int *);
232 void                     pf_hash(struct pf_addr *, struct pf_addr *,
233                             struct pf_poolhashkey *, sa_family_t);
234 int                      pf_map_addr(u_int8_t, struct pf_rule *,
235                             struct pf_addr *, struct pf_addr *,
236                             struct pf_addr *, struct pf_src_node **);
237 int                      pf_get_sport(sa_family_t, u_int8_t, struct pf_rule *,
238                             struct pf_addr *, struct pf_addr *, u_int16_t,
239                             struct pf_addr *, u_int16_t*, u_int16_t, u_int16_t,
240                             struct pf_src_node **);
241 void                     pf_route(struct mbuf **, struct pf_rule *, int,
242                             struct ifnet *, struct pf_state *,
243                             struct pf_pdesc *);
244 void                     pf_route6(struct mbuf **, struct pf_rule *, int,
245                             struct ifnet *, struct pf_state *,
246                             struct pf_pdesc *);
247 u_int8_t                 pf_get_wscale(struct mbuf *, int, u_int16_t,
248                             sa_family_t);
249 u_int16_t                pf_get_mss(struct mbuf *, int, u_int16_t,
250                             sa_family_t);
251 u_int16_t                pf_calc_mss(struct pf_addr *, sa_family_t,
252                                 u_int16_t);
253 void                     pf_set_rt_ifp(struct pf_state *,
254                             struct pf_addr *);
255 int                      pf_check_proto_cksum(struct mbuf *, int, int,
256                             u_int8_t, sa_family_t);
257 struct pf_divert        *pf_get_divert(struct mbuf *);
258 void                     pf_print_state_parts(struct pf_state *,
259                             struct pf_state_key *, struct pf_state_key *);
260 int                      pf_addr_wrap_neq(struct pf_addr_wrap *,
261                             struct pf_addr_wrap *);
262 struct pf_state         *pf_find_state(struct pfi_kif *,
263                             struct pf_state_key_cmp *, u_int, struct mbuf *);
264 int                      pf_src_connlimit(struct pf_state **);
265 int                      pf_check_congestion(struct ifqueue *);
266
267 extern int pf_end_threads;
268
269 struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] = {
270         { &pf_state_pl, PFSTATE_HIWAT },
271         { &pf_src_tree_pl, PFSNODE_HIWAT },
272         { &pf_frent_pl, PFFRAG_FRENT_HIWAT },
273         { &pfr_ktable_pl, PFR_KTABLE_HIWAT },
274         { &pfr_kentry_pl, PFR_KENTRY_HIWAT }
275 };
276
277 #define STATE_LOOKUP(i, k, d, s, m)                                     \
278         do {                                                            \
279                 s = pf_find_state(i, k, d, m);                  \
280                 if (s == NULL || (s)->timeout == PFTM_PURGE)            \
281                         return (PF_DROP);                               \
282                 if (d == PF_OUT &&                                      \
283                     (((s)->rule.ptr->rt == PF_ROUTETO &&                \
284                     (s)->rule.ptr->direction == PF_OUT) ||              \
285                     ((s)->rule.ptr->rt == PF_REPLYTO &&                 \
286                     (s)->rule.ptr->direction == PF_IN)) &&              \
287                     (s)->rt_kif != NULL &&                              \
288                     (s)->rt_kif != i)                                   \
289                         return (PF_PASS);                               \
290         } while (0)
291
292 #define BOUND_IFACE(r, k) \
293         ((r)->rule_flag & PFRULE_IFBOUND) ? (k) : pfi_all
294
295 #define STATE_INC_COUNTERS(s)                           \
296         do {                                            \
297                 s->rule.ptr->states_cur++;              \
298                 s->rule.ptr->states_tot++;              \
299                 if (s->anchor.ptr != NULL) {            \
300                         s->anchor.ptr->states_cur++;    \
301                         s->anchor.ptr->states_tot++;    \
302                 }                                       \
303                 if (s->nat_rule.ptr != NULL) {          \
304                         s->nat_rule.ptr->states_cur++;  \
305                         s->nat_rule.ptr->states_tot++;  \
306                 }                                       \
307         } while (0)
308
309 #define STATE_DEC_COUNTERS(s)                           \
310         do {                                            \
311                 if (s->nat_rule.ptr != NULL)            \
312                         s->nat_rule.ptr->states_cur--;  \
313                 if (s->anchor.ptr != NULL)              \
314                         s->anchor.ptr->states_cur--;    \
315                 s->rule.ptr->states_cur--;              \
316         } while (0)
317
318 static __inline int pf_src_compare(struct pf_src_node *, struct pf_src_node *);
319 static __inline int pf_state_compare_key(struct pf_state_key *,
320         struct pf_state_key *);
321 static __inline int pf_state_compare_id(struct pf_state *,
322         struct pf_state *);
323
324 struct pf_src_tree tree_src_tracking;
325
326 struct pf_state_tree_id tree_id;
327 struct pf_state_queue state_list;
328
329 RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare);
330 RB_GENERATE(pf_state_tree, pf_state_key, entry, pf_state_compare_key);
331 RB_GENERATE(pf_state_tree_id, pf_state,
332     entry_id, pf_state_compare_id);
333
334 static __inline int
335 pf_src_compare(struct pf_src_node *a, struct pf_src_node *b)
336 {
337         int     diff;
338
339         if (a->rule.ptr > b->rule.ptr)
340                 return (1);
341         if (a->rule.ptr < b->rule.ptr)
342                 return (-1);
343         if ((diff = a->af - b->af) != 0)
344                 return (diff);
345         switch (a->af) {
346 #ifdef INET
347         case AF_INET:
348                 if (a->addr.addr32[0] > b->addr.addr32[0])
349                         return (1);
350                 if (a->addr.addr32[0] < b->addr.addr32[0])
351                         return (-1);
352                 break;
353 #endif /* INET */
354 #ifdef INET6
355         case AF_INET6:
356                 if (a->addr.addr32[3] > b->addr.addr32[3])
357                         return (1);
358                 if (a->addr.addr32[3] < b->addr.addr32[3])
359                         return (-1);
360                 if (a->addr.addr32[2] > b->addr.addr32[2])
361                         return (1);
362                 if (a->addr.addr32[2] < b->addr.addr32[2])
363                         return (-1);
364                 if (a->addr.addr32[1] > b->addr.addr32[1])
365                         return (1);
366                 if (a->addr.addr32[1] < b->addr.addr32[1])
367                         return (-1);
368                 if (a->addr.addr32[0] > b->addr.addr32[0])
369                         return (1);
370                 if (a->addr.addr32[0] < b->addr.addr32[0])
371                         return (-1);
372                 break;
373 #endif /* INET6 */
374         }
375         return (0);
376 }
377
378 u_int32_t
379 pf_state_hash(struct pf_state_key *sk)
380 {
381         u_int32_t hv = (u_int32_t)(((intptr_t)sk >> 6) ^ ((intptr_t)sk >> 15));
382         if (hv == 0)    /* disallow 0 */
383                 hv = 1;
384         return(hv);
385 }
386
387 #ifdef INET6
388 void
389 pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af)
390 {
391         switch (af) {
392 #ifdef INET
393         case AF_INET:
394                 dst->addr32[0] = src->addr32[0];
395                 break;
396 #endif /* INET */
397         case AF_INET6:
398                 dst->addr32[0] = src->addr32[0];
399                 dst->addr32[1] = src->addr32[1];
400                 dst->addr32[2] = src->addr32[2];
401                 dst->addr32[3] = src->addr32[3];
402                 break;
403         }
404 }
405 #endif /* INET6 */
406
407 void
408 pf_init_threshold(struct pf_threshold *threshold,
409     u_int32_t limit, u_int32_t seconds)
410 {
411         threshold->limit = limit * PF_THRESHOLD_MULT;
412         threshold->seconds = seconds;
413         threshold->count = 0;
414         threshold->last = time_second;
415 }
416
417 void
418 pf_add_threshold(struct pf_threshold *threshold)
419 {
420         u_int32_t t = time_second, diff = t - threshold->last;
421
422         if (diff >= threshold->seconds)
423                 threshold->count = 0;
424         else
425                 threshold->count -= threshold->count * diff /
426                     threshold->seconds;
427         threshold->count += PF_THRESHOLD_MULT;
428         threshold->last = t;
429 }
430
431 int
432 pf_check_threshold(struct pf_threshold *threshold)
433 {
434         return (threshold->count > threshold->limit);
435 }
436
437 int
438 pf_src_connlimit(struct pf_state **state)
439 {
440         int bad = 0;
441
442         (*state)->src_node->conn++;
443         (*state)->src.tcp_est = 1;
444         pf_add_threshold(&(*state)->src_node->conn_rate);
445
446         if ((*state)->rule.ptr->max_src_conn &&
447             (*state)->rule.ptr->max_src_conn <
448             (*state)->src_node->conn) {
449                 pf_status.lcounters[LCNT_SRCCONN]++;
450                 bad++;
451         }
452
453         if ((*state)->rule.ptr->max_src_conn_rate.limit &&
454             pf_check_threshold(&(*state)->src_node->conn_rate)) {
455                 pf_status.lcounters[LCNT_SRCCONNRATE]++;
456                 bad++;
457         }
458
459         if (!bad)
460                 return (0);
461
462         if ((*state)->rule.ptr->overload_tbl) {
463                 struct pfr_addr p;
464                 u_int32_t       killed = 0;
465
466                 pf_status.lcounters[LCNT_OVERLOAD_TABLE]++;
467                 if (pf_status.debug >= PF_DEBUG_MISC) {
468                         kprintf("pf_src_connlimit: blocking address ");
469                         pf_print_host(&(*state)->src_node->addr, 0,
470                             (*state)->key[PF_SK_WIRE]->af);
471                 }
472
473                 bzero(&p, sizeof(p));
474                 p.pfra_af = (*state)->key[PF_SK_WIRE]->af;
475                 switch ((*state)->key[PF_SK_WIRE]->af) {
476 #ifdef INET
477                 case AF_INET:
478                         p.pfra_net = 32;
479                         p.pfra_ip4addr = (*state)->src_node->addr.v4;
480                         break;
481 #endif /* INET */
482 #ifdef INET6
483                 case AF_INET6:
484                         p.pfra_net = 128;
485                         p.pfra_ip6addr = (*state)->src_node->addr.v6;
486                         break;
487 #endif /* INET6 */
488                 }
489
490                 pfr_insert_kentry((*state)->rule.ptr->overload_tbl,
491                     &p, time_second);
492
493                 /* kill existing states if that's required. */
494                 if ((*state)->rule.ptr->flush) {
495                         struct pf_state_key *sk;
496                         struct pf_state *st;
497
498                         pf_status.lcounters[LCNT_OVERLOAD_FLUSH]++;
499                         RB_FOREACH(st, pf_state_tree_id, &tree_id) {
500                                 sk = st->key[PF_SK_WIRE];
501                                 /*
502                                  * Kill states from this source.  (Only those
503                                  * from the same rule if PF_FLUSH_GLOBAL is not
504                                  * set)
505                                  */
506                                 if (sk->af ==
507                                     (*state)->key[PF_SK_WIRE]->af &&
508                                     (((*state)->direction == PF_OUT &&
509                                     PF_AEQ(&(*state)->src_node->addr,
510                                         &sk->addr[0], sk->af)) ||
511                                     ((*state)->direction == PF_IN &&
512                                     PF_AEQ(&(*state)->src_node->addr,
513                                         &sk->addr[1], sk->af))) &&
514                                     ((*state)->rule.ptr->flush &
515                                     PF_FLUSH_GLOBAL ||
516                                     (*state)->rule.ptr == st->rule.ptr)) {
517                                         st->timeout = PFTM_PURGE;
518                                         st->src.state = st->dst.state =
519                                             TCPS_CLOSED;
520                                         killed++;
521                                 }
522                         }
523                         if (pf_status.debug >= PF_DEBUG_MISC)
524                                 kprintf(", %u states killed", killed);
525                 }
526                 if (pf_status.debug >= PF_DEBUG_MISC)
527                         kprintf("\n");
528         }
529
530         /* kill this state */
531         (*state)->timeout = PFTM_PURGE;
532         (*state)->src.state = (*state)->dst.state = TCPS_CLOSED;
533         return (1);
534 }
535
536 int
537 pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule,
538     struct pf_addr *src, sa_family_t af)
539 {
540         struct pf_src_node      k;
541
542         if (*sn == NULL) {
543                 k.af = af;
544                 PF_ACPY(&k.addr, src, af);
545                 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
546                     rule->rpool.opts & PF_POOL_STICKYADDR)
547                         k.rule.ptr = rule;
548                 else
549                         k.rule.ptr = NULL;
550                 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
551                 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
552         }
553         if (*sn == NULL) {
554                 if (!rule->max_src_nodes ||
555                     rule->src_nodes < rule->max_src_nodes)
556                         (*sn) = pool_get(&pf_src_tree_pl, PR_NOWAIT | PR_ZERO);
557                 else
558                         pf_status.lcounters[LCNT_SRCNODES]++;
559                 if ((*sn) == NULL)
560                         return (-1);
561
562                 pf_init_threshold(&(*sn)->conn_rate,
563                     rule->max_src_conn_rate.limit,
564                     rule->max_src_conn_rate.seconds);
565
566                 (*sn)->af = af;
567                 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
568                     rule->rpool.opts & PF_POOL_STICKYADDR)
569                         (*sn)->rule.ptr = rule;
570                 else
571                         (*sn)->rule.ptr = NULL;
572                 PF_ACPY(&(*sn)->addr, src, af);
573                 if (RB_INSERT(pf_src_tree,
574                     &tree_src_tracking, *sn) != NULL) {
575                         if (pf_status.debug >= PF_DEBUG_MISC) {
576                                 kprintf("pf: src_tree insert failed: ");
577                                 pf_print_host(&(*sn)->addr, 0, af);
578                                 kprintf("\n");
579                         }
580                         pool_put(&pf_src_tree_pl, *sn);
581                         return (-1);
582                 }
583                 (*sn)->creation = time_second;
584                 (*sn)->ruletype = rule->action;
585                 if ((*sn)->rule.ptr != NULL)
586                         (*sn)->rule.ptr->src_nodes++;
587                 pf_status.scounters[SCNT_SRC_NODE_INSERT]++;
588                 pf_status.src_nodes++;
589         } else {
590                 if (rule->max_src_states &&
591                     (*sn)->states >= rule->max_src_states) {
592                         pf_status.lcounters[LCNT_SRCSTATES]++;
593                         return (-1);
594                 }
595         }
596         return (0);
597 }
598
599 /* state table stuff */
600
601 static __inline int
602 pf_state_compare_key(struct pf_state_key *a, struct pf_state_key *b)
603 {
604         int     diff;
605
606         if ((diff = a->proto - b->proto) != 0)
607                 return (diff);
608         if ((diff = a->af - b->af) != 0)
609                 return (diff);
610         switch (a->af) {
611 #ifdef INET
612         case AF_INET:
613                 if (a->addr[0].addr32[0] > b->addr[0].addr32[0])
614                         return (1);
615                 if (a->addr[0].addr32[0] < b->addr[0].addr32[0])
616                         return (-1);
617                 if (a->addr[1].addr32[0] > b->addr[1].addr32[0])
618                         return (1);
619                 if (a->addr[1].addr32[0] < b->addr[1].addr32[0])
620                         return (-1);
621                 break;
622 #endif /* INET */
623 #ifdef INET6
624         case AF_INET6:
625                 if (a->addr[0].addr32[3] > b->addr[0].addr32[3])
626                         return (1);
627                 if (a->addr[0].addr32[3] < b->addr[0].addr32[3])
628                         return (-1);
629                 if (a->addr[1].addr32[3] > b->addr[1].addr32[3])
630                         return (1);
631                 if (a->addr[1].addr32[3] < b->addr[1].addr32[3])
632                         return (-1);
633                 if (a->addr[0].addr32[2] > b->addr[0].addr32[2])
634                         return (1);
635                 if (a->addr[0].addr32[2] < b->addr[0].addr32[2])
636                         return (-1);
637                 if (a->addr[1].addr32[2] > b->addr[1].addr32[2])
638                         return (1);
639                 if (a->addr[1].addr32[2] < b->addr[1].addr32[2])
640                         return (-1);
641                 if (a->addr[0].addr32[1] > b->addr[0].addr32[1])
642                         return (1);
643                 if (a->addr[0].addr32[1] < b->addr[0].addr32[1])
644                         return (-1);
645                 if (a->addr[1].addr32[1] > b->addr[1].addr32[1])
646                         return (1);
647                 if (a->addr[1].addr32[1] < b->addr[1].addr32[1])
648                         return (-1);
649                 if (a->addr[0].addr32[0] > b->addr[0].addr32[0])
650                         return (1);
651                 if (a->addr[0].addr32[0] < b->addr[0].addr32[0])
652                         return (-1);
653                 if (a->addr[1].addr32[0] > b->addr[1].addr32[0])
654                         return (1);
655                 if (a->addr[1].addr32[0] < b->addr[1].addr32[0])
656                         return (-1);
657                 break;
658 #endif /* INET6 */
659         }
660
661         if ((diff = a->port[0] - b->port[0]) != 0)
662                 return (diff);
663         if ((diff = a->port[1] - b->port[1]) != 0)
664                 return (diff);
665
666         return (0);
667 }
668
669 static __inline int
670 pf_state_compare_id(struct pf_state *a, struct pf_state *b)
671 {
672         if (a->id > b->id)
673                 return (1);
674         if (a->id < b->id)
675                 return (-1);
676         if (a->creatorid > b->creatorid)
677                 return (1);
678         if (a->creatorid < b->creatorid)
679                 return (-1);
680
681         return (0);
682 }
683
684 int
685 pf_state_key_attach(struct pf_state_key *sk, struct pf_state *s, int idx)
686 {
687         struct pf_state_item    *si;
688         struct pf_state_key     *cur;
689
690         KKASSERT(s->key[idx] == NULL);  /* XXX handle this? */
691
692         if ((cur = RB_INSERT(pf_state_tree, &pf_statetbl, sk)) != NULL) {
693                 /* key exists. check for same kif, if none, add to key */
694                 TAILQ_FOREACH(si, &cur->states, entry)
695                         if (si->s->kif == s->kif &&
696                             si->s->direction == s->direction) {
697                                 if (pf_status.debug >= PF_DEBUG_MISC) {
698                                         kprintf(
699                                             "pf: %s key attach failed on %s: ",
700                                             (idx == PF_SK_WIRE) ?
701                                             "wire" : "stack",
702                                             s->kif->pfik_name);
703                                         pf_print_state_parts(s,
704                                             (idx == PF_SK_WIRE) ? sk : NULL,
705                                             (idx == PF_SK_STACK) ? sk : NULL);
706                                         kprintf("\n");
707                                 }
708                                 pool_put(&pf_state_key_pl, sk);
709                                 return (-1);    /* collision! */
710                         }
711                 pool_put(&pf_state_key_pl, sk);
712                 s->key[idx] = cur;
713         } else
714                 s->key[idx] = sk;
715
716         if ((si = pool_get(&pf_state_item_pl, PR_NOWAIT)) == NULL) {
717                 pf_state_key_detach(s, idx);
718                 return (-1);
719         }
720         si->s = s;
721
722         /* list is sorted, if-bound states before floating */
723         if (s->kif == pfi_all)
724                 TAILQ_INSERT_TAIL(&s->key[idx]->states, si, entry);
725         else
726                 TAILQ_INSERT_HEAD(&s->key[idx]->states, si, entry);
727         return (0);
728 }
729
730 void
731 pf_detach_state(struct pf_state *s)
732 {
733         if (s->key[PF_SK_WIRE] == s->key[PF_SK_STACK])
734                 s->key[PF_SK_WIRE] = NULL;
735
736         if (s->key[PF_SK_STACK] != NULL)
737                 pf_state_key_detach(s, PF_SK_STACK);
738
739         if (s->key[PF_SK_WIRE] != NULL)
740                 pf_state_key_detach(s, PF_SK_WIRE);
741 }
742
743 void
744 pf_state_key_detach(struct pf_state *s, int idx)
745 {
746         struct pf_state_item    *si;
747
748         si = TAILQ_FIRST(&s->key[idx]->states);
749         while (si && si->s != s)
750             si = TAILQ_NEXT(si, entry);
751
752         if (si) {
753                 TAILQ_REMOVE(&s->key[idx]->states, si, entry);
754                 pool_put(&pf_state_item_pl, si);
755         }
756
757         if (TAILQ_EMPTY(&s->key[idx]->states)) {
758                 RB_REMOVE(pf_state_tree, &pf_statetbl, s->key[idx]);
759                 if (s->key[idx]->reverse)
760                         s->key[idx]->reverse->reverse = NULL;
761                 if (s->key[idx]->inp)
762                         s->key[idx]->inp->inp_pf_sk = NULL;
763                 pool_put(&pf_state_key_pl, s->key[idx]);
764         }
765         s->key[idx] = NULL;
766 }
767
768 struct pf_state_key *
769 pf_alloc_state_key(int pool_flags)
770 {
771         struct pf_state_key     *sk;
772
773         if ((sk = pool_get(&pf_state_key_pl, pool_flags)) == NULL)
774                 return (NULL);
775         TAILQ_INIT(&sk->states);
776
777         return (sk);
778 }
779
780 int
781 pf_state_key_setup(struct pf_pdesc *pd, struct pf_rule *nr,
782         struct pf_state_key **skw, struct pf_state_key **sks,
783         struct pf_state_key **skp, struct pf_state_key **nkp,
784         struct pf_addr *saddr, struct pf_addr *daddr,
785         u_int16_t sport, u_int16_t dport)
786 {
787         KKASSERT((*skp == NULL && *nkp == NULL));
788
789         if ((*skp = pf_alloc_state_key(PR_NOWAIT | PR_ZERO)) == NULL)
790                 return (ENOMEM);
791
792         PF_ACPY(&(*skp)->addr[pd->sidx], saddr, pd->af);
793         PF_ACPY(&(*skp)->addr[pd->didx], daddr, pd->af);
794         (*skp)->port[pd->sidx] = sport;
795         (*skp)->port[pd->didx] = dport;
796         (*skp)->proto = pd->proto;
797         (*skp)->af = pd->af;
798
799         if (nr != NULL) {
800                 if ((*nkp = pf_alloc_state_key(PR_NOWAIT | PR_ZERO)) == NULL)
801                         return (ENOMEM); /* caller must handle cleanup */
802
803                 /* XXX maybe just bcopy and TAILQ_INIT(&(*nkp)->states) */
804                 PF_ACPY(&(*nkp)->addr[0], &(*skp)->addr[0], pd->af);
805                 PF_ACPY(&(*nkp)->addr[1], &(*skp)->addr[1], pd->af);
806                 (*nkp)->port[0] = (*skp)->port[0];
807                 (*nkp)->port[1] = (*skp)->port[1];
808                 (*nkp)->proto = pd->proto;
809                 (*nkp)->af = pd->af;
810         } else
811                 *nkp = *skp;
812
813         if (pd->dir == PF_IN) {
814                 *skw = *skp;
815                 *sks = *nkp;
816         } else {
817                 *sks = *skp;
818                 *skw = *nkp;
819         }
820         return (0);
821 }
822
823
824 int
825 pf_state_insert(struct pfi_kif *kif, struct pf_state_key *skw,
826     struct pf_state_key *sks, struct pf_state *s)
827 {
828         s->kif = kif;
829
830         if (skw == sks) {
831                 if (pf_state_key_attach(skw, s, PF_SK_WIRE))
832                         return (-1);
833                 s->key[PF_SK_STACK] = s->key[PF_SK_WIRE];
834         } else {
835                 if (pf_state_key_attach(skw, s, PF_SK_WIRE)) {
836                         pool_put(&pf_state_key_pl, sks);
837                         return (-1);
838                 }
839                 if (pf_state_key_attach(sks, s, PF_SK_STACK)) {
840                         pf_state_key_detach(s, PF_SK_WIRE);
841                         return (-1);
842                 }
843         }
844
845         if (s->id == 0 && s->creatorid == 0) {
846                 s->id = htobe64(pf_status.stateid++);
847                 s->creatorid = pf_status.hostid;
848         }
849
850         /*
851          * Calculate hash code for altq
852          */
853         s->hash = crc32(s->key[PF_SK_WIRE], sizeof(*sks));
854
855         if (RB_INSERT(pf_state_tree_id, &tree_id, s) != NULL) {
856                 if (pf_status.debug >= PF_DEBUG_MISC) {
857                         kprintf("pf: state insert failed: "
858                             "id: %016jx creatorid: %08x",
859                               (uintmax_t)be64toh(s->id), ntohl(s->creatorid));
860                         if (s->sync_flags & PFSTATE_FROMSYNC)
861                                 kprintf(" (from sync)");
862                         kprintf("\n");
863                 }
864                 pf_detach_state(s);
865                 return (-1);
866         }
867         TAILQ_INSERT_TAIL(&state_list, s, entry_list);
868         pf_status.fcounters[FCNT_STATE_INSERT]++;
869         pf_status.states++;
870         pfi_kif_ref(kif, PFI_KIF_REF_STATE);
871         pfsync_insert_state(s);
872         return (0);
873 }
874
875 struct pf_state *
876 pf_find_state_byid(struct pf_state_cmp *key)
877 {
878         pf_status.fcounters[FCNT_STATE_SEARCH]++;
879
880         return (RB_FIND(pf_state_tree_id, &tree_id, (struct pf_state *)key));
881 }
882
883 struct pf_state *
884 pf_find_state(struct pfi_kif *kif, struct pf_state_key_cmp *key, u_int dir,
885     struct mbuf *m)
886 {
887         struct pf_state_key     *sk;
888         struct pf_state_item    *si;
889
890         pf_status.fcounters[FCNT_STATE_SEARCH]++;
891
892         if (dir == PF_OUT && m->m_pkthdr.pf.statekey &&
893             ((struct pf_state_key *)m->m_pkthdr.pf.statekey)->reverse)
894                 sk = ((struct pf_state_key *)m->m_pkthdr.pf.statekey)->reverse;
895         else {
896                 if ((sk = RB_FIND(pf_state_tree, &pf_statetbl,
897                     (struct pf_state_key *)key)) == NULL)
898                         return (NULL);
899                 if (dir == PF_OUT && m->m_pkthdr.pf.statekey) {
900                         ((struct pf_state_key *)
901                             m->m_pkthdr.pf.statekey)->reverse = sk;
902                         sk->reverse = m->m_pkthdr.pf.statekey;
903                 }
904         }
905
906         if (dir == PF_OUT)
907                 m->m_pkthdr.pf.statekey = NULL;
908
909         /* list is sorted, if-bound states before floating ones */
910         TAILQ_FOREACH(si, &sk->states, entry)
911                 if ((si->s->kif == pfi_all || si->s->kif == kif) &&
912                     sk == (dir == PF_IN ? si->s->key[PF_SK_WIRE] :
913                     si->s->key[PF_SK_STACK]))
914                         return (si->s);
915
916         return (NULL);
917 }
918
919 struct pf_state *
920 pf_find_state_all(struct pf_state_key_cmp *key, u_int dir, int *more)
921 {
922         struct pf_state_key     *sk;
923         struct pf_state_item    *si, *ret = NULL;
924
925         pf_status.fcounters[FCNT_STATE_SEARCH]++;
926
927         sk = RB_FIND(pf_state_tree, &pf_statetbl, (struct pf_state_key *)key);
928
929         if (sk != NULL) {
930                 TAILQ_FOREACH(si, &sk->states, entry)
931                         if (dir == PF_INOUT ||
932                             (sk == (dir == PF_IN ? si->s->key[PF_SK_WIRE] :
933                             si->s->key[PF_SK_STACK]))) {
934                                 if (more == NULL)
935                                         return (si->s);
936
937                                 if (ret)
938                                         (*more)++;
939                                 else
940                                         ret = si;
941                         }
942         }
943         return (ret ? ret->s : NULL);
944 }
945
946 /* END state table stuff */
947
948
949 void
950 pf_purge_thread(void *v)
951 {
952         int nloops = 0;
953         int locked = 0;
954
955         lwkt_gettoken(&pf_token);
956         for (;;) {
957                 tsleep(pf_purge_thread, PWAIT, "pftm", 1 * hz);
958
959                 lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
960
961                 if (pf_end_threads) {
962                         pf_purge_expired_states(pf_status.states, 1);
963                         pf_purge_expired_fragments();
964                         pf_purge_expired_src_nodes(1);
965                         pf_end_threads++;
966
967                         lockmgr(&pf_consistency_lock, LK_RELEASE);
968                         wakeup(pf_purge_thread);
969                         kthread_exit();
970                 }
971                 crit_enter();
972
973                 /* process a fraction of the state table every second */
974                 if(!pf_purge_expired_states(1 + (pf_status.states
975                     / pf_default_rule.timeout[PFTM_INTERVAL]), 0)) {
976
977                         pf_purge_expired_states(1 + (pf_status.states
978                             / pf_default_rule.timeout[PFTM_INTERVAL]), 1);
979                 }
980
981                 /* purge other expired types every PFTM_INTERVAL seconds */
982                 if (++nloops >= pf_default_rule.timeout[PFTM_INTERVAL]) {
983                         pf_purge_expired_fragments();
984                         if (!pf_purge_expired_src_nodes(locked)) {
985                                 pf_purge_expired_src_nodes(1);
986                         }
987                         nloops = 0;
988                 }
989                 crit_exit();
990                 lockmgr(&pf_consistency_lock, LK_RELEASE);
991         }
992         lwkt_reltoken(&pf_token);
993 }
994
995 u_int32_t
996 pf_state_expires(const struct pf_state *state)
997 {
998         u_int32_t       timeout;
999         u_int32_t       start;
1000         u_int32_t       end;
1001         u_int32_t       states;
1002
1003         /* handle all PFTM_* > PFTM_MAX here */
1004         if (state->timeout == PFTM_PURGE)
1005                 return (time_second);
1006         if (state->timeout == PFTM_UNTIL_PACKET)
1007                 return (0);
1008         KKASSERT(state->timeout != PFTM_UNLINKED);
1009         KKASSERT(state->timeout < PFTM_MAX);
1010         timeout = state->rule.ptr->timeout[state->timeout];
1011         if (!timeout)
1012                 timeout = pf_default_rule.timeout[state->timeout];
1013         start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START];
1014         if (start) {
1015                 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END];
1016                 states = state->rule.ptr->states_cur;
1017         } else {
1018                 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START];
1019                 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END];
1020                 states = pf_status.states;
1021         }
1022         if (end && states > start && start < end) {
1023                 if (states < end)
1024                         return (state->expire + timeout * (end - states) /
1025                             (end - start));
1026                 else
1027                         return (time_second);
1028         }
1029         return (state->expire + timeout);
1030 }
1031
1032 int
1033 pf_purge_expired_src_nodes(int waslocked)
1034 {
1035          struct pf_src_node             *cur, *next;
1036          int                             locked = waslocked;
1037
1038          for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) {
1039                  next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur);
1040
1041                  if (cur->states <= 0 && cur->expire <= time_second) {
1042                          if (! locked) {
1043                                  lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
1044                                  next = RB_NEXT(pf_src_tree,
1045                                      &tree_src_tracking, cur);
1046                                  locked = 1;
1047                          }
1048                          if (cur->rule.ptr != NULL) {
1049                                  cur->rule.ptr->src_nodes--;
1050                                  if (cur->rule.ptr->states_cur <= 0 &&
1051                                      cur->rule.ptr->max_src_nodes <= 0)
1052                                          pf_rm_rule(NULL, cur->rule.ptr);
1053                          }
1054                          RB_REMOVE(pf_src_tree, &tree_src_tracking, cur);
1055                          pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
1056                          pf_status.src_nodes--;
1057                          pool_put(&pf_src_tree_pl, cur);
1058                  }
1059          }
1060
1061          if (locked && !waslocked)
1062                 lockmgr(&pf_consistency_lock, LK_RELEASE);
1063         return(1);
1064 }
1065
1066 void
1067 pf_src_tree_remove_state(struct pf_state *s)
1068 {
1069         u_int32_t timeout;
1070
1071         if (s->src_node != NULL) {
1072                 if (s->src.tcp_est)
1073                         --s->src_node->conn;
1074                 if (--s->src_node->states <= 0) {
1075                         timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1076                         if (!timeout)
1077                                 timeout =
1078                                     pf_default_rule.timeout[PFTM_SRC_NODE];
1079                         s->src_node->expire = time_second + timeout;
1080                 }
1081         }
1082         if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) {
1083                 if (--s->nat_src_node->states <= 0) {
1084                         timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1085                         if (!timeout)
1086                                 timeout =
1087                                     pf_default_rule.timeout[PFTM_SRC_NODE];
1088                         s->nat_src_node->expire = time_second + timeout;
1089                 }
1090         }
1091         s->src_node = s->nat_src_node = NULL;
1092 }
1093
1094 /* callers should be at crit_enter() */
1095 void
1096 pf_unlink_state(struct pf_state *cur)
1097 {
1098         if (cur->src.state == PF_TCPS_PROXY_DST) {
1099                 /* XXX wire key the right one? */
1100                 pf_send_tcp(cur->rule.ptr, cur->key[PF_SK_WIRE]->af,
1101                     &cur->key[PF_SK_WIRE]->addr[1],
1102                     &cur->key[PF_SK_WIRE]->addr[0],
1103                     cur->key[PF_SK_WIRE]->port[1],
1104                     cur->key[PF_SK_WIRE]->port[0],
1105                     cur->src.seqhi, cur->src.seqlo + 1,
1106                     TH_RST|TH_ACK, 0, 0, 0, 1, cur->tag, NULL, NULL);
1107         }
1108         RB_REMOVE(pf_state_tree_id, &tree_id, cur);
1109         if (cur->creatorid == pf_status.hostid)
1110                 pfsync_delete_state(cur);
1111         cur->timeout = PFTM_UNLINKED;
1112         pf_src_tree_remove_state(cur);
1113         pf_detach_state(cur);
1114 }
1115
1116 static struct pf_state  *purge_cur;
1117
1118 /* callers should be at crit_enter() and hold the
1119  * write_lock on pf_consistency_lock */
1120 void
1121 pf_free_state(struct pf_state *cur)
1122 {
1123         if (pfsyncif != NULL &&
1124             (pfsyncif->sc_bulk_send_next == cur ||
1125             pfsyncif->sc_bulk_terminator == cur))
1126                 return;
1127         KKASSERT(cur->timeout == PFTM_UNLINKED);
1128         if (--cur->rule.ptr->states_cur <= 0 &&
1129             cur->rule.ptr->src_nodes <= 0)
1130                 pf_rm_rule(NULL, cur->rule.ptr);
1131         if (cur->nat_rule.ptr != NULL)
1132                 if (--cur->nat_rule.ptr->states_cur <= 0 &&
1133                         cur->nat_rule.ptr->src_nodes <= 0)
1134                         pf_rm_rule(NULL, cur->nat_rule.ptr);
1135         if (cur->anchor.ptr != NULL)
1136                 if (--cur->anchor.ptr->states_cur <= 0)
1137                         pf_rm_rule(NULL, cur->anchor.ptr);
1138         pf_normalize_tcp_cleanup(cur);
1139         pfi_kif_unref(cur->kif, PFI_KIF_REF_STATE);
1140
1141         /*
1142          * We may be freeing pf_purge_expired_states()'s saved scan entry,
1143          * adjust it if necessary.
1144          */
1145         if (purge_cur == cur) {
1146                 kprintf("PURGE CONFLICT\n");
1147                 purge_cur = TAILQ_NEXT(purge_cur, entry_list);
1148         }
1149         TAILQ_REMOVE(&state_list, cur, entry_list);
1150         if (cur->tag)
1151                 pf_tag_unref(cur->tag);
1152         pool_put(&pf_state_pl, cur);
1153         pf_status.fcounters[FCNT_STATE_REMOVALS]++;
1154         pf_status.states--;
1155 }
1156
1157 int
1158 pf_purge_expired_states(u_int32_t maxcheck, int waslocked)
1159 {
1160         struct pf_state         *cur;
1161         int                      locked = waslocked;
1162
1163         while (maxcheck--) {
1164                 /*
1165                  * Wrap to start of list when we hit the end
1166                  */
1167                 cur = purge_cur;
1168                 if (cur == NULL) {
1169                         cur = TAILQ_FIRST(&state_list);
1170                         if (cur == NULL)
1171                                 break;  /* list empty */
1172                 }
1173
1174                 /*
1175                  * Setup next (purge_cur) while we process this one.  If we block and
1176                  * something else deletes purge_cur, pf_free_state() will adjust it further
1177                  * ahead.
1178                  */
1179                 purge_cur = TAILQ_NEXT(cur, entry_list);
1180
1181                 if (cur->timeout == PFTM_UNLINKED) {
1182                         /* free unlinked state */
1183                         if (! locked) {
1184                                 lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
1185                                 locked = 1;
1186                         }
1187                         pf_free_state(cur);
1188                 } else if (pf_state_expires(cur) <= time_second) {
1189                         /* unlink and free expired state */
1190                         pf_unlink_state(cur);
1191                         if (! locked) {
1192                                 if (!lockmgr(&pf_consistency_lock, LK_EXCLUSIVE))
1193                                         return (0);
1194                                 locked = 1;
1195                         }
1196                         pf_free_state(cur);
1197                 }
1198         }
1199
1200         if (locked)
1201                 lockmgr(&pf_consistency_lock, LK_RELEASE);
1202         return (1);
1203 }
1204
1205 int
1206 pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw)
1207 {
1208         if (aw->type != PF_ADDR_TABLE)
1209                 return (0);
1210         if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL)
1211                 return (1);
1212         return (0);
1213 }
1214
1215 void
1216 pf_tbladdr_remove(struct pf_addr_wrap *aw)
1217 {
1218         if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL)
1219                 return;
1220         pfr_detach_table(aw->p.tbl);
1221         aw->p.tbl = NULL;
1222 }
1223
1224 void
1225 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
1226 {
1227         struct pfr_ktable *kt = aw->p.tbl;
1228
1229         if (aw->type != PF_ADDR_TABLE || kt == NULL)
1230                 return;
1231         if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
1232                 kt = kt->pfrkt_root;
1233         aw->p.tbl = NULL;
1234         aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
1235                 kt->pfrkt_cnt : -1;
1236 }
1237
1238 void
1239 pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af)
1240 {
1241         switch (af) {
1242 #ifdef INET
1243         case AF_INET: {
1244                 u_int32_t a = ntohl(addr->addr32[0]);
1245                 kprintf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255,
1246                     (a>>8)&255, a&255);
1247                 if (p) {
1248                         p = ntohs(p);
1249                         kprintf(":%u", p);
1250                 }
1251                 break;
1252         }
1253 #endif /* INET */
1254 #ifdef INET6
1255         case AF_INET6: {
1256                 u_int16_t b;
1257                 u_int8_t i, curstart = 255, curend = 0,
1258                     maxstart = 0, maxend = 0;
1259                 for (i = 0; i < 8; i++) {
1260                         if (!addr->addr16[i]) {
1261                                 if (curstart == 255)
1262                                         curstart = i;
1263                                 else
1264                                         curend = i;
1265                         } else {
1266                                 if (curstart) {
1267                                         if ((curend - curstart) >
1268                                             (maxend - maxstart)) {
1269                                                 maxstart = curstart;
1270                                                 maxend = curend;
1271                                                 curstart = 255;
1272                                         }
1273                                 }
1274                         }
1275                 }
1276                 for (i = 0; i < 8; i++) {
1277                         if (i >= maxstart && i <= maxend) {
1278                                 if (maxend != 7) {
1279                                         if (i == maxstart)
1280                                                 kprintf(":");
1281                                 } else {
1282                                         if (i == maxend)
1283                                                 kprintf(":");
1284                                 }
1285                         } else {
1286                                 b = ntohs(addr->addr16[i]);
1287                                 kprintf("%x", b);
1288                                 if (i < 7)
1289                                         kprintf(":");
1290                         }
1291                 }
1292                 if (p) {
1293                         p = ntohs(p);
1294                         kprintf("[%u]", p);
1295                 }
1296                 break;
1297         }
1298 #endif /* INET6 */
1299         }
1300 }
1301
1302 void
1303 pf_print_state(struct pf_state *s)
1304 {
1305         pf_print_state_parts(s, NULL, NULL);
1306 }
1307
1308 void
1309 pf_print_state_parts(struct pf_state *s,
1310     struct pf_state_key *skwp, struct pf_state_key *sksp)
1311 {
1312         struct pf_state_key *skw, *sks;
1313         u_int8_t proto, dir;
1314
1315         /* Do our best to fill these, but they're skipped if NULL */
1316         skw = skwp ? skwp : (s ? s->key[PF_SK_WIRE] : NULL);
1317         sks = sksp ? sksp : (s ? s->key[PF_SK_STACK] : NULL);
1318         proto = skw ? skw->proto : (sks ? sks->proto : 0);
1319         dir = s ? s->direction : 0;
1320
1321         switch (proto) {
1322         case IPPROTO_TCP:
1323                 kprintf("TCP ");
1324                 break;
1325         case IPPROTO_UDP:
1326                 kprintf("UDP ");
1327                 break;
1328         case IPPROTO_ICMP:
1329                 kprintf("ICMP ");
1330                 break;
1331         case IPPROTO_ICMPV6:
1332                 kprintf("ICMPV6 ");
1333                 break;
1334         default:
1335                 kprintf("%u ", skw->proto);
1336                 break;
1337         }
1338         switch (dir) {
1339         case PF_IN:
1340                 kprintf(" in");
1341                 break;
1342         case PF_OUT:
1343                 kprintf(" out");
1344                 break;
1345         }
1346         if (skw) {
1347                 kprintf(" wire: ");
1348                 pf_print_host(&skw->addr[0], skw->port[0], skw->af);
1349                 kprintf(" ");
1350                 pf_print_host(&skw->addr[1], skw->port[1], skw->af);
1351         }
1352         if (sks) {
1353                 kprintf(" stack: ");
1354                 if (sks != skw) {
1355                         pf_print_host(&sks->addr[0], sks->port[0], sks->af);
1356                         kprintf(" ");
1357                         pf_print_host(&sks->addr[1], sks->port[1], sks->af);
1358                 } else
1359                         kprintf("-");
1360         }
1361         if (s) {
1362                 if (proto == IPPROTO_TCP) {
1363                         kprintf(" [lo=%u high=%u win=%u modulator=%u",
1364                             s->src.seqlo, s->src.seqhi,
1365                             s->src.max_win, s->src.seqdiff);
1366                         if (s->src.wscale && s->dst.wscale)
1367                                 kprintf(" wscale=%u",
1368                                     s->src.wscale & PF_WSCALE_MASK);
1369                         kprintf("]");
1370                         kprintf(" [lo=%u high=%u win=%u modulator=%u",
1371                             s->dst.seqlo, s->dst.seqhi,
1372                             s->dst.max_win, s->dst.seqdiff);
1373                         if (s->src.wscale && s->dst.wscale)
1374                                 kprintf(" wscale=%u",
1375                                 s->dst.wscale & PF_WSCALE_MASK);
1376                         kprintf("]");
1377                 }
1378                 kprintf(" %u:%u", s->src.state, s->dst.state);
1379         }
1380 }
1381
1382 void
1383 pf_print_flags(u_int8_t f)
1384 {
1385         if (f)
1386                 kprintf(" ");
1387         if (f & TH_FIN)
1388                 kprintf("F");
1389         if (f & TH_SYN)
1390                 kprintf("S");
1391         if (f & TH_RST)
1392                 kprintf("R");
1393         if (f & TH_PUSH)
1394                 kprintf("P");
1395         if (f & TH_ACK)
1396                 kprintf("A");
1397         if (f & TH_URG)
1398                 kprintf("U");
1399         if (f & TH_ECE)
1400                 kprintf("E");
1401         if (f & TH_CWR)
1402                 kprintf("W");
1403 }
1404
1405 #define PF_SET_SKIP_STEPS(i)                                    \
1406         do {                                                    \
1407                 while (head[i] != cur) {                        \
1408                         head[i]->skip[i].ptr = cur;             \
1409                         head[i] = TAILQ_NEXT(head[i], entries); \
1410                 }                                               \
1411         } while (0)
1412
1413 void
1414 pf_calc_skip_steps(struct pf_rulequeue *rules)
1415 {
1416         struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT];
1417         int i;
1418
1419         cur = TAILQ_FIRST(rules);
1420         prev = cur;
1421         for (i = 0; i < PF_SKIP_COUNT; ++i)
1422                 head[i] = cur;
1423         while (cur != NULL) {
1424
1425                 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
1426                         PF_SET_SKIP_STEPS(PF_SKIP_IFP);
1427                 if (cur->direction != prev->direction)
1428                         PF_SET_SKIP_STEPS(PF_SKIP_DIR);
1429                 if (cur->af != prev->af)
1430                         PF_SET_SKIP_STEPS(PF_SKIP_AF);
1431                 if (cur->proto != prev->proto)
1432                         PF_SET_SKIP_STEPS(PF_SKIP_PROTO);
1433                 if (cur->src.neg != prev->src.neg ||
1434                     pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr))
1435                         PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR);
1436                 if (cur->src.port[0] != prev->src.port[0] ||
1437                     cur->src.port[1] != prev->src.port[1] ||
1438                     cur->src.port_op != prev->src.port_op)
1439                         PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
1440                 if (cur->dst.neg != prev->dst.neg ||
1441                     pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr))
1442                         PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR);
1443                 if (cur->dst.port[0] != prev->dst.port[0] ||
1444                     cur->dst.port[1] != prev->dst.port[1] ||
1445                     cur->dst.port_op != prev->dst.port_op)
1446                         PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
1447
1448                 prev = cur;
1449                 cur = TAILQ_NEXT(cur, entries);
1450         }
1451         for (i = 0; i < PF_SKIP_COUNT; ++i)
1452                 PF_SET_SKIP_STEPS(i);
1453 }
1454
1455 int
1456 pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2)
1457 {
1458         if (aw1->type != aw2->type)
1459                 return (1);
1460         switch (aw1->type) {
1461         case PF_ADDR_ADDRMASK:
1462         case PF_ADDR_RANGE:
1463                 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0))
1464                         return (1);
1465                 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0))
1466                         return (1);
1467                 return (0);
1468         case PF_ADDR_DYNIFTL:
1469                 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt);
1470         case PF_ADDR_NOROUTE:
1471         case PF_ADDR_URPFFAILED:
1472                 return (0);
1473         case PF_ADDR_TABLE:
1474                 return (aw1->p.tbl != aw2->p.tbl);
1475         case PF_ADDR_RTLABEL:
1476                 return (aw1->v.rtlabel != aw2->v.rtlabel);
1477         default:
1478                 kprintf("invalid address type: %d\n", aw1->type);
1479                 return (1);
1480         }
1481 }
1482
1483 u_int16_t
1484 pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp)
1485 {
1486         u_int32_t       l;
1487
1488         if (udp && !cksum)
1489                 return (0x0000);
1490         l = cksum + old - new;
1491         l = (l >> 16) + (l & 65535);
1492         l = l & 65535;
1493         if (udp && !l)
1494                 return (0xFFFF);
1495         return (l);
1496 }
1497
1498 void
1499 pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc,
1500     struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af)
1501 {
1502         struct pf_addr  ao;
1503         u_int16_t       po = *p;
1504
1505         PF_ACPY(&ao, a, af);
1506         PF_ACPY(a, an, af);
1507
1508         *p = pn;
1509
1510         switch (af) {
1511 #ifdef INET
1512         case AF_INET:
1513                 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1514                     ao.addr16[0], an->addr16[0], 0),
1515                     ao.addr16[1], an->addr16[1], 0);
1516                 *p = pn;
1517                 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1518                     ao.addr16[0], an->addr16[0], u),
1519                     ao.addr16[1], an->addr16[1], u),
1520                     po, pn, u);
1521                 break;
1522 #endif /* INET */
1523 #ifdef INET6
1524         case AF_INET6:
1525                 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1526                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1527                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1528                     ao.addr16[0], an->addr16[0], u),
1529                     ao.addr16[1], an->addr16[1], u),
1530                     ao.addr16[2], an->addr16[2], u),
1531                     ao.addr16[3], an->addr16[3], u),
1532                     ao.addr16[4], an->addr16[4], u),
1533                     ao.addr16[5], an->addr16[5], u),
1534                     ao.addr16[6], an->addr16[6], u),
1535                     ao.addr16[7], an->addr16[7], u),
1536                     po, pn, u);
1537                 break;
1538 #endif /* INET6 */
1539         }
1540 }
1541
1542
1543 /* Changes a u_int32_t.  Uses a void * so there are no align restrictions */
1544 void
1545 pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u)
1546 {
1547         u_int32_t       ao;
1548
1549         memcpy(&ao, a, sizeof(ao));
1550         memcpy(a, &an, sizeof(u_int32_t));
1551         *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u),
1552             ao % 65536, an % 65536, u);
1553 }
1554
1555 #ifdef INET6
1556 void
1557 pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u)
1558 {
1559         struct pf_addr  ao;
1560
1561         PF_ACPY(&ao, a, AF_INET6);
1562         PF_ACPY(a, an, AF_INET6);
1563
1564         *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1565             pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1566             pf_cksum_fixup(pf_cksum_fixup(*c,
1567             ao.addr16[0], an->addr16[0], u),
1568             ao.addr16[1], an->addr16[1], u),
1569             ao.addr16[2], an->addr16[2], u),
1570             ao.addr16[3], an->addr16[3], u),
1571             ao.addr16[4], an->addr16[4], u),
1572             ao.addr16[5], an->addr16[5], u),
1573             ao.addr16[6], an->addr16[6], u),
1574             ao.addr16[7], an->addr16[7], u);
1575 }
1576 #endif /* INET6 */
1577
1578 void
1579 pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa,
1580     struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c,
1581     u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af)
1582 {
1583         struct pf_addr  oia, ooa;
1584
1585         PF_ACPY(&oia, ia, af);
1586         if (oa)
1587                 PF_ACPY(&ooa, oa, af);
1588
1589         /* Change inner protocol port, fix inner protocol checksum. */
1590         if (ip != NULL) {
1591                 u_int16_t       oip = *ip;
1592                 u_int32_t       opc = 0;
1593
1594                 if (pc != NULL)
1595                         opc = *pc;
1596                 *ip = np;
1597                 if (pc != NULL)
1598                         *pc = pf_cksum_fixup(*pc, oip, *ip, u);
1599                 *ic = pf_cksum_fixup(*ic, oip, *ip, 0);
1600                 if (pc != NULL)
1601                         *ic = pf_cksum_fixup(*ic, opc, *pc, 0);
1602         }
1603         /* Change inner ip address, fix inner ip and icmp checksums. */
1604         PF_ACPY(ia, na, af);
1605         switch (af) {
1606 #ifdef INET
1607         case AF_INET: {
1608                 u_int32_t        oh2c = *h2c;
1609
1610                 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c,
1611                     oia.addr16[0], ia->addr16[0], 0),
1612                     oia.addr16[1], ia->addr16[1], 0);
1613                 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1614                     oia.addr16[0], ia->addr16[0], 0),
1615                     oia.addr16[1], ia->addr16[1], 0);
1616                 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0);
1617                 break;
1618         }
1619 #endif /* INET */
1620 #ifdef INET6
1621         case AF_INET6:
1622                 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1623                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1624                     pf_cksum_fixup(pf_cksum_fixup(*ic,
1625                     oia.addr16[0], ia->addr16[0], u),
1626                     oia.addr16[1], ia->addr16[1], u),
1627                     oia.addr16[2], ia->addr16[2], u),
1628                     oia.addr16[3], ia->addr16[3], u),
1629                     oia.addr16[4], ia->addr16[4], u),
1630                     oia.addr16[5], ia->addr16[5], u),
1631                     oia.addr16[6], ia->addr16[6], u),
1632                     oia.addr16[7], ia->addr16[7], u);
1633                 break;
1634 #endif /* INET6 */
1635         }
1636         /* Outer ip address, fix outer ip or icmpv6 checksum, if necessary. */
1637         if (oa) {
1638                 PF_ACPY(oa, na, af);
1639                 switch (af) {
1640 #ifdef INET
1641                 case AF_INET:
1642                         *hc = pf_cksum_fixup(pf_cksum_fixup(*hc,
1643                             ooa.addr16[0], oa->addr16[0], 0),
1644                             ooa.addr16[1], oa->addr16[1], 0);
1645                         break;
1646 #endif /* INET */
1647 #ifdef INET6
1648                 case AF_INET6:
1649                         *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1650                             pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1651                             pf_cksum_fixup(pf_cksum_fixup(*ic,
1652                             ooa.addr16[0], oa->addr16[0], u),
1653                             ooa.addr16[1], oa->addr16[1], u),
1654                             ooa.addr16[2], oa->addr16[2], u),
1655                             ooa.addr16[3], oa->addr16[3], u),
1656                             ooa.addr16[4], oa->addr16[4], u),
1657                             ooa.addr16[5], oa->addr16[5], u),
1658                             ooa.addr16[6], oa->addr16[6], u),
1659                             ooa.addr16[7], oa->addr16[7], u);
1660                         break;
1661 #endif /* INET6 */
1662                 }
1663         }
1664 }
1665
1666
1667 /*
1668  * Need to modulate the sequence numbers in the TCP SACK option
1669  * (credits to Krzysztof Pfaff for report and patch)
1670  */
1671 int
1672 pf_modulate_sack(struct mbuf *m, int off, struct pf_pdesc *pd,
1673     struct tcphdr *th, struct pf_state_peer *dst)
1674 {
1675         int hlen = (th->th_off << 2) - sizeof(*th), thoptlen = hlen;
1676         u_int8_t opts[TCP_MAXOLEN], *opt = opts;
1677         int copyback = 0, i, olen;
1678         struct raw_sackblock sack;
1679
1680 #define TCPOLEN_SACKLEN (TCPOLEN_SACK + 2)
1681         if (hlen < TCPOLEN_SACKLEN ||
1682             !pf_pull_hdr(m, off + sizeof(*th), opts, hlen, NULL, NULL, pd->af))
1683                 return 0;
1684
1685         while (hlen >= TCPOLEN_SACKLEN) {
1686                 olen = opt[1];
1687                 switch (*opt) {
1688                 case TCPOPT_EOL:        /* FALLTHROUGH */
1689                 case TCPOPT_NOP:
1690                         opt++;
1691                         hlen--;
1692                         break;
1693                 case TCPOPT_SACK:
1694                         if (olen > hlen)
1695                                 olen = hlen;
1696                         if (olen >= TCPOLEN_SACKLEN) {
1697                                 for (i = 2; i + TCPOLEN_SACK <= olen;
1698                                     i += TCPOLEN_SACK) {
1699                                         memcpy(&sack, &opt[i], sizeof(sack));
1700                                         pf_change_a(&sack.rblk_start, &th->th_sum,
1701                                             htonl(ntohl(sack.rblk_start) -
1702                                             dst->seqdiff), 0);
1703                                         pf_change_a(&sack.rblk_end, &th->th_sum,
1704                                             htonl(ntohl(sack.rblk_end) -
1705                                             dst->seqdiff), 0);
1706                                         memcpy(&opt[i], &sack, sizeof(sack));
1707                                 }
1708                                 copyback = 1;
1709                         }
1710                         /* FALLTHROUGH */
1711                 default:
1712                         if (olen < 2)
1713                                 olen = 2;
1714                         hlen -= olen;
1715                         opt += olen;
1716                 }
1717         }
1718
1719         if (copyback)
1720                 m_copyback(m, off + sizeof(*th), thoptlen, opts);
1721         return (copyback);
1722 }
1723
1724 void
1725 pf_send_tcp(const struct pf_rule *r, sa_family_t af,
1726     const struct pf_addr *saddr, const struct pf_addr *daddr,
1727     u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack,
1728     u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl, int tag,
1729     u_int16_t rtag, struct ether_header *eh, struct ifnet *ifp)
1730 {
1731         struct mbuf     *m;
1732         int              len = 0, tlen;
1733 #ifdef INET
1734         struct ip       *h = NULL;
1735 #endif /* INET */
1736 #ifdef INET6
1737         struct ip6_hdr  *h6 = NULL;
1738 #endif /* INET6 */
1739         struct tcphdr   *th = NULL;
1740         char            *opt;
1741
1742         ASSERT_LWKT_TOKEN_HELD(&pf_token);
1743
1744         /* maximum segment size tcp option */
1745         tlen = sizeof(struct tcphdr);
1746         if (mss)
1747                 tlen += 4;
1748
1749         switch (af) {
1750 #ifdef INET
1751         case AF_INET:
1752                 len = sizeof(struct ip) + tlen;
1753                 break;
1754 #endif /* INET */
1755 #ifdef INET6
1756         case AF_INET6:
1757                 len = sizeof(struct ip6_hdr) + tlen;
1758                 break;
1759 #endif /* INET6 */
1760         }
1761
1762         /*
1763          * Create outgoing mbuf.
1764          *
1765          * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags,
1766          * so make sure pf.flags is clear.
1767          */
1768         m = m_gethdr(MB_DONTWAIT, MT_HEADER);
1769         if (m == NULL) {
1770                 return;
1771         }
1772         if (tag)
1773                 m->m_pkthdr.fw_flags |= PF_MBUF_TAGGED;
1774         m->m_pkthdr.pf.flags = 0;
1775         m->m_pkthdr.pf.tag = rtag;
1776         /* XXX Recheck when upgrading to > 4.4 */
1777         m->m_pkthdr.pf.statekey = NULL;
1778         if (r != NULL && r->rtableid >= 0)
1779                 m->m_pkthdr.pf.rtableid = r->rtableid;
1780
1781 #ifdef ALTQ
1782         if (r != NULL && r->qid) {
1783                 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
1784                 m->m_pkthdr.pf.qid = r->qid;
1785                 m->m_pkthdr.pf.ecn_af = af;
1786                 m->m_pkthdr.pf.hdr = mtod(m, struct ip *);
1787         }
1788 #endif /* ALTQ */
1789         m->m_data += max_linkhdr;
1790         m->m_pkthdr.len = m->m_len = len;
1791         m->m_pkthdr.rcvif = NULL;
1792         bzero(m->m_data, len);
1793         switch (af) {
1794 #ifdef INET
1795         case AF_INET:
1796                 h = mtod(m, struct ip *);
1797
1798                 /* IP header fields included in the TCP checksum */
1799                 h->ip_p = IPPROTO_TCP;
1800                 h->ip_len = tlen;
1801                 h->ip_src.s_addr = saddr->v4.s_addr;
1802                 h->ip_dst.s_addr = daddr->v4.s_addr;
1803
1804                 th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip));
1805                 break;
1806 #endif /* INET */
1807 #ifdef INET6
1808         case AF_INET6:
1809                 h6 = mtod(m, struct ip6_hdr *);
1810
1811                 /* IP header fields included in the TCP checksum */
1812                 h6->ip6_nxt = IPPROTO_TCP;
1813                 h6->ip6_plen = htons(tlen);
1814                 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr));
1815                 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr));
1816
1817                 th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr));
1818                 break;
1819 #endif /* INET6 */
1820         }
1821
1822         /* TCP header */
1823         th->th_sport = sport;
1824         th->th_dport = dport;
1825         th->th_seq = htonl(seq);
1826         th->th_ack = htonl(ack);
1827         th->th_off = tlen >> 2;
1828         th->th_flags = flags;
1829         th->th_win = htons(win);
1830
1831         if (mss) {
1832                 opt = (char *)(th + 1);
1833                 opt[0] = TCPOPT_MAXSEG;
1834                 opt[1] = 4;
1835                 mss = htons(mss);
1836                 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2);
1837         }
1838
1839         switch (af) {
1840 #ifdef INET
1841         case AF_INET:
1842                 /* TCP checksum */
1843                 th->th_sum = in_cksum(m, len);
1844
1845                 /* Finish the IP header */
1846                 h->ip_v = 4;
1847                 h->ip_hl = sizeof(*h) >> 2;
1848                 h->ip_tos = IPTOS_LOWDELAY;
1849                 h->ip_len = len;
1850                 h->ip_off = path_mtu_discovery ? IP_DF : 0;
1851                 h->ip_ttl = ttl ? ttl : ip_defttl;
1852                 h->ip_sum = 0;
1853                 if (eh == NULL) {
1854                         lwkt_reltoken(&pf_token);
1855                         ip_output(m, NULL, NULL, 0, NULL, NULL);
1856                         lwkt_gettoken(&pf_token);
1857                 } else {
1858                         struct route             ro;
1859                         struct rtentry           rt;
1860                         struct ether_header     *e = (void *)ro.ro_dst.sa_data;
1861
1862                         if (ifp == NULL) {
1863                                 m_freem(m);
1864                                 return;
1865                         }
1866                         rt.rt_ifp = ifp;
1867                         ro.ro_rt = &rt;
1868                         ro.ro_dst.sa_len = sizeof(ro.ro_dst);
1869                         ro.ro_dst.sa_family = pseudo_AF_HDRCMPLT;
1870                         bcopy(eh->ether_dhost, e->ether_shost, ETHER_ADDR_LEN);
1871                         bcopy(eh->ether_shost, e->ether_dhost, ETHER_ADDR_LEN);
1872                         e->ether_type = eh->ether_type;
1873                         /* XXX_IMPORT: later */
1874                         lwkt_reltoken(&pf_token);
1875                         ip_output(m, (void *)NULL, &ro, 0,
1876                             (void *)NULL, (void *)NULL);
1877                         lwkt_gettoken(&pf_token);
1878                 }
1879                 break;
1880 #endif /* INET */
1881 #ifdef INET6
1882         case AF_INET6:
1883                 /* TCP checksum */
1884                 th->th_sum = in6_cksum(m, IPPROTO_TCP,
1885                     sizeof(struct ip6_hdr), tlen);
1886
1887                 h6->ip6_vfc |= IPV6_VERSION;
1888                 h6->ip6_hlim = IPV6_DEFHLIM;
1889
1890                 lwkt_reltoken(&pf_token);
1891                 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
1892                 lwkt_gettoken(&pf_token);
1893                 break;
1894 #endif /* INET6 */
1895         }
1896 }
1897
1898 void
1899 pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af,
1900     struct pf_rule *r)
1901 {
1902         struct mbuf     *m0;
1903
1904         /*
1905          * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags,
1906          * so make sure pf.flags is clear.
1907          */
1908         if ((m0 = m_copy(m, 0, M_COPYALL)) == NULL)
1909                 return;
1910
1911         m0->m_pkthdr.fw_flags |= PF_MBUF_TAGGED;
1912         m0->m_pkthdr.pf.flags = 0;
1913         /* XXX Re-Check when Upgrading to > 4.4 */
1914         m0->m_pkthdr.pf.statekey = NULL;
1915
1916         if (r->rtableid >= 0)
1917                 m0->m_pkthdr.pf.rtableid = r->rtableid;
1918
1919 #ifdef ALTQ
1920         if (r->qid) {
1921                 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
1922                 m0->m_pkthdr.pf.qid = r->qid;
1923                 m0->m_pkthdr.pf.ecn_af = af;
1924                 m0->m_pkthdr.pf.hdr = mtod(m0, struct ip *);
1925         }
1926 #endif /* ALTQ */
1927
1928         switch (af) {
1929 #ifdef INET
1930         case AF_INET:
1931                 icmp_error(m0, type, code, 0, 0);
1932                 break;
1933 #endif /* INET */
1934 #ifdef INET6
1935         case AF_INET6:
1936                 icmp6_error(m0, type, code, 0);
1937                 break;
1938 #endif /* INET6 */
1939         }
1940 }
1941
1942 /*
1943  * Return 1 if the addresses a and b match (with mask m), otherwise return 0.
1944  * If n is 0, they match if they are equal. If n is != 0, they match if they
1945  * are different.
1946  */
1947 int
1948 pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m,
1949     struct pf_addr *b, sa_family_t af)
1950 {
1951         int     match = 0;
1952
1953         switch (af) {
1954 #ifdef INET
1955         case AF_INET:
1956                 if ((a->addr32[0] & m->addr32[0]) ==
1957                     (b->addr32[0] & m->addr32[0]))
1958                         match++;
1959                 break;
1960 #endif /* INET */
1961 #ifdef INET6
1962         case AF_INET6:
1963                 if (((a->addr32[0] & m->addr32[0]) ==
1964                      (b->addr32[0] & m->addr32[0])) &&
1965                     ((a->addr32[1] & m->addr32[1]) ==
1966                      (b->addr32[1] & m->addr32[1])) &&
1967                     ((a->addr32[2] & m->addr32[2]) ==
1968                      (b->addr32[2] & m->addr32[2])) &&
1969                     ((a->addr32[3] & m->addr32[3]) ==
1970                      (b->addr32[3] & m->addr32[3])))
1971                         match++;
1972                 break;
1973 #endif /* INET6 */
1974         }
1975         if (match) {
1976                 if (n)
1977                         return (0);
1978                 else
1979                         return (1);
1980         } else {
1981                 if (n)
1982                         return (1);
1983                 else
1984                         return (0);
1985         }
1986 }
1987
1988 /*
1989  * Return 1 if b <= a <= e, otherwise return 0.
1990  */
1991 int
1992 pf_match_addr_range(struct pf_addr *b, struct pf_addr *e,
1993     struct pf_addr *a, sa_family_t af)
1994 {
1995         switch (af) {
1996 #ifdef INET
1997         case AF_INET:
1998                 if ((a->addr32[0] < b->addr32[0]) ||
1999                     (a->addr32[0] > e->addr32[0]))
2000                         return (0);
2001                 break;
2002 #endif /* INET */
2003 #ifdef INET6
2004         case AF_INET6: {
2005                 int     i;
2006
2007                 /* check a >= b */
2008                 for (i = 0; i < 4; ++i)
2009                         if (a->addr32[i] > b->addr32[i])
2010                                 break;
2011                         else if (a->addr32[i] < b->addr32[i])
2012                                 return (0);
2013                 /* check a <= e */
2014                 for (i = 0; i < 4; ++i)
2015                         if (a->addr32[i] < e->addr32[i])
2016                                 break;
2017                         else if (a->addr32[i] > e->addr32[i])
2018                                 return (0);
2019                 break;
2020         }
2021 #endif /* INET6 */
2022         }
2023         return (1);
2024 }
2025
2026 int
2027 pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p)
2028 {
2029         switch (op) {
2030         case PF_OP_IRG:
2031                 return ((p > a1) && (p < a2));
2032         case PF_OP_XRG:
2033                 return ((p < a1) || (p > a2));
2034         case PF_OP_RRG:
2035                 return ((p >= a1) && (p <= a2));
2036         case PF_OP_EQ:
2037                 return (p == a1);
2038         case PF_OP_NE:
2039                 return (p != a1);
2040         case PF_OP_LT:
2041                 return (p < a1);
2042         case PF_OP_LE:
2043                 return (p <= a1);
2044         case PF_OP_GT:
2045                 return (p > a1);
2046         case PF_OP_GE:
2047                 return (p >= a1);
2048         }
2049         return (0); /* never reached */
2050 }
2051
2052 int
2053 pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p)
2054 {
2055         a1 = ntohs(a1);
2056         a2 = ntohs(a2);
2057         p = ntohs(p);
2058         return (pf_match(op, a1, a2, p));
2059 }
2060
2061 int
2062 pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u)
2063 {
2064         if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
2065                 return (0);
2066         return (pf_match(op, a1, a2, u));
2067 }
2068
2069 int
2070 pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g)
2071 {
2072         if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
2073                 return (0);
2074         return (pf_match(op, a1, a2, g));
2075 }
2076
2077 int
2078 pf_match_tag(struct mbuf *m, struct pf_rule *r, int *tag)
2079 {
2080         if (*tag == -1)
2081                 *tag = m->m_pkthdr.pf.tag;
2082
2083         return ((!r->match_tag_not && r->match_tag == *tag) ||
2084             (r->match_tag_not && r->match_tag != *tag));
2085 }
2086
2087 int
2088 pf_tag_packet(struct mbuf *m, int tag, int rtableid)
2089 {
2090         if (tag <= 0 && rtableid < 0)
2091                 return (0);
2092
2093         if (tag > 0)
2094                 m->m_pkthdr.pf.tag = tag;
2095         if (rtableid >= 0)
2096                 m->m_pkthdr.pf.rtableid = rtableid;
2097
2098         return (0);
2099 }
2100
2101 void
2102 pf_step_into_anchor(int *depth, struct pf_ruleset **rs, int n,
2103     struct pf_rule **r, struct pf_rule **a, int *match)
2104 {
2105         struct pf_anchor_stackframe     *f;
2106
2107         (*r)->anchor->match = 0;
2108         if (match)
2109                 *match = 0;
2110         if (*depth >= sizeof(pf_anchor_stack) /
2111             sizeof(pf_anchor_stack[0])) {
2112                 kprintf("pf_step_into_anchor: stack overflow\n");
2113                 *r = TAILQ_NEXT(*r, entries);
2114                 return;
2115         } else if (*depth == 0 && a != NULL)
2116                 *a = *r;
2117         f = pf_anchor_stack + (*depth)++;
2118         f->rs = *rs;
2119         f->r = *r;
2120         if ((*r)->anchor_wildcard) {
2121                 f->parent = &(*r)->anchor->children;
2122                 if ((f->child = RB_MIN(pf_anchor_node, f->parent)) ==
2123                     NULL) {
2124                         *r = NULL;
2125                         return;
2126                 }
2127                 *rs = &f->child->ruleset;
2128         } else {
2129                 f->parent = NULL;
2130                 f->child = NULL;
2131                 *rs = &(*r)->anchor->ruleset;
2132         }
2133         *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
2134 }
2135
2136 int
2137 pf_step_out_of_anchor(int *depth, struct pf_ruleset **rs, int n,
2138     struct pf_rule **r, struct pf_rule **a, int *match)
2139 {
2140         struct pf_anchor_stackframe     *f;
2141         int quick = 0;
2142
2143         do {
2144                 if (*depth <= 0)
2145                         break;
2146                 f = pf_anchor_stack + *depth - 1;
2147                 if (f->parent != NULL && f->child != NULL) {
2148                         if (f->child->match ||
2149                             (match != NULL && *match)) {
2150                                 f->r->anchor->match = 1;
2151                                 *match = 0;
2152                         }
2153                         f->child = RB_NEXT(pf_anchor_node, f->parent, f->child);
2154                         if (f->child != NULL) {
2155                                 *rs = &f->child->ruleset;
2156                                 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
2157                                 if (*r == NULL)
2158                                         continue;
2159                                 else
2160                                         break;
2161                         }
2162                 }
2163                 (*depth)--;
2164                 if (*depth == 0 && a != NULL)
2165                         *a = NULL;
2166                 *rs = f->rs;
2167                 if (f->r->anchor->match || (match != NULL && *match))
2168                         quick = f->r->quick;
2169                 *r = TAILQ_NEXT(f->r, entries);
2170         } while (*r == NULL);
2171
2172         return (quick);
2173 }
2174
2175 #ifdef INET6
2176 void
2177 pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr,
2178     struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af)
2179 {
2180         switch (af) {
2181 #ifdef INET
2182         case AF_INET:
2183                 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2184                 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2185                 break;
2186 #endif /* INET */
2187         case AF_INET6:
2188                 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2189                 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2190                 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) |
2191                 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]);
2192                 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) |
2193                 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]);
2194                 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) |
2195                 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]);
2196                 break;
2197         }
2198 }
2199
2200 void
2201 pf_addr_inc(struct pf_addr *addr, sa_family_t af)
2202 {
2203         switch (af) {
2204 #ifdef INET
2205         case AF_INET:
2206                 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1);
2207                 break;
2208 #endif /* INET */
2209         case AF_INET6:
2210                 if (addr->addr32[3] == 0xffffffff) {
2211                         addr->addr32[3] = 0;
2212                         if (addr->addr32[2] == 0xffffffff) {
2213                                 addr->addr32[2] = 0;
2214                                 if (addr->addr32[1] == 0xffffffff) {
2215                                         addr->addr32[1] = 0;
2216                                         addr->addr32[0] =
2217                                             htonl(ntohl(addr->addr32[0]) + 1);
2218                                 } else
2219                                         addr->addr32[1] =
2220                                             htonl(ntohl(addr->addr32[1]) + 1);
2221                         } else
2222                                 addr->addr32[2] =
2223                                     htonl(ntohl(addr->addr32[2]) + 1);
2224                 } else
2225                         addr->addr32[3] =
2226                             htonl(ntohl(addr->addr32[3]) + 1);
2227                 break;
2228         }
2229 }
2230 #endif /* INET6 */
2231
2232 #define mix(a,b,c) \
2233         do {                                    \
2234                 a -= b; a -= c; a ^= (c >> 13); \
2235                 b -= c; b -= a; b ^= (a << 8);  \
2236                 c -= a; c -= b; c ^= (b >> 13); \
2237                 a -= b; a -= c; a ^= (c >> 12); \
2238                 b -= c; b -= a; b ^= (a << 16); \
2239                 c -= a; c -= b; c ^= (b >> 5);  \
2240                 a -= b; a -= c; a ^= (c >> 3);  \
2241                 b -= c; b -= a; b ^= (a << 10); \
2242                 c -= a; c -= b; c ^= (b >> 15); \
2243         } while (0)
2244
2245 /*
2246  * hash function based on bridge_hash in if_bridge.c
2247  */
2248 void
2249 pf_hash(struct pf_addr *inaddr, struct pf_addr *hash,
2250     struct pf_poolhashkey *key, sa_family_t af)
2251 {
2252         u_int32_t       a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0];
2253
2254         switch (af) {
2255 #ifdef INET
2256         case AF_INET:
2257                 a += inaddr->addr32[0];
2258                 b += key->key32[1];
2259                 mix(a, b, c);
2260                 hash->addr32[0] = c + key->key32[2];
2261                 break;
2262 #endif /* INET */
2263 #ifdef INET6
2264         case AF_INET6:
2265                 a += inaddr->addr32[0];
2266                 b += inaddr->addr32[2];
2267                 mix(a, b, c);
2268                 hash->addr32[0] = c;
2269                 a += inaddr->addr32[1];
2270                 b += inaddr->addr32[3];
2271                 c += key->key32[1];
2272                 mix(a, b, c);
2273                 hash->addr32[1] = c;
2274                 a += inaddr->addr32[2];
2275                 b += inaddr->addr32[1];
2276                 c += key->key32[2];
2277                 mix(a, b, c);
2278                 hash->addr32[2] = c;
2279                 a += inaddr->addr32[3];
2280                 b += inaddr->addr32[0];
2281                 c += key->key32[3];
2282                 mix(a, b, c);
2283                 hash->addr32[3] = c;
2284                 break;
2285 #endif /* INET6 */
2286         }
2287 }
2288
2289 int
2290 pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr,
2291     struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn)
2292 {
2293         unsigned char            hash[16];
2294         struct pf_pool          *rpool = &r->rpool;
2295         struct pf_addr          *raddr = &rpool->cur->addr.v.a.addr;
2296         struct pf_addr          *rmask = &rpool->cur->addr.v.a.mask;
2297         struct pf_pooladdr      *acur = rpool->cur;
2298         struct pf_src_node       k;
2299
2300         if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR &&
2301             (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2302                 k.af = af;
2303                 PF_ACPY(&k.addr, saddr, af);
2304                 if (r->rule_flag & PFRULE_RULESRCTRACK ||
2305                     r->rpool.opts & PF_POOL_STICKYADDR)
2306                         k.rule.ptr = r;
2307                 else
2308                         k.rule.ptr = NULL;
2309                 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
2310                 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
2311                 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) {
2312                         PF_ACPY(naddr, &(*sn)->raddr, af);
2313                         if (pf_status.debug >= PF_DEBUG_MISC) {
2314                                 kprintf("pf_map_addr: src tracking maps ");
2315                                 pf_print_host(&k.addr, 0, af);
2316                                 kprintf(" to ");
2317                                 pf_print_host(naddr, 0, af);
2318                                 kprintf("\n");
2319                         }
2320                         return (0);
2321                 }
2322         }
2323
2324         if (rpool->cur->addr.type == PF_ADDR_NOROUTE)
2325                 return (1);
2326         if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2327                 switch (af) {
2328 #ifdef INET
2329                 case AF_INET:
2330                         if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 &&
2331                             (rpool->opts & PF_POOL_TYPEMASK) !=
2332                             PF_POOL_ROUNDROBIN)
2333                                 return (1);
2334                          raddr = &rpool->cur->addr.p.dyn->pfid_addr4;
2335                          rmask = &rpool->cur->addr.p.dyn->pfid_mask4;
2336                         break;
2337 #endif /* INET */
2338 #ifdef INET6
2339                 case AF_INET6:
2340                         if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 &&
2341                             (rpool->opts & PF_POOL_TYPEMASK) !=
2342                             PF_POOL_ROUNDROBIN)
2343                                 return (1);
2344                         raddr = &rpool->cur->addr.p.dyn->pfid_addr6;
2345                         rmask = &rpool->cur->addr.p.dyn->pfid_mask6;
2346                         break;
2347 #endif /* INET6 */
2348                 }
2349         } else if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2350                 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN)
2351                         return (1); /* unsupported */
2352         } else {
2353                 raddr = &rpool->cur->addr.v.a.addr;
2354                 rmask = &rpool->cur->addr.v.a.mask;
2355         }
2356
2357         switch (rpool->opts & PF_POOL_TYPEMASK) {
2358         case PF_POOL_NONE:
2359                 PF_ACPY(naddr, raddr, af);
2360                 break;
2361         case PF_POOL_BITMASK:
2362                 PF_POOLMASK(naddr, raddr, rmask, saddr, af);
2363                 break;
2364         case PF_POOL_RANDOM:
2365                 if (init_addr != NULL && PF_AZERO(init_addr, af)) {
2366                         switch (af) {
2367 #ifdef INET
2368                         case AF_INET:
2369                                 rpool->counter.addr32[0] = htonl(karc4random());
2370                                 break;
2371 #endif /* INET */
2372 #ifdef INET6
2373                         case AF_INET6:
2374                                 if (rmask->addr32[3] != 0xffffffff)
2375                                         rpool->counter.addr32[3] =
2376                                             htonl(karc4random());
2377                                 else
2378                                         break;
2379                                 if (rmask->addr32[2] != 0xffffffff)
2380                                         rpool->counter.addr32[2] =
2381                                             htonl(karc4random());
2382                                 else
2383                                         break;
2384                                 if (rmask->addr32[1] != 0xffffffff)
2385                                         rpool->counter.addr32[1] =
2386                                             htonl(karc4random());
2387                                 else
2388                                         break;
2389                                 if (rmask->addr32[0] != 0xffffffff)
2390                                         rpool->counter.addr32[0] =
2391                                             htonl(karc4random());
2392                                 break;
2393 #endif /* INET6 */
2394                         }
2395                         PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2396                         PF_ACPY(init_addr, naddr, af);
2397
2398                 } else {
2399                         PF_AINC(&rpool->counter, af);
2400                         PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2401                 }
2402                 break;
2403         case PF_POOL_SRCHASH:
2404                 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af);
2405                 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af);
2406                 break;
2407         case PF_POOL_ROUNDROBIN:
2408                 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2409                         if (!pfr_pool_get(rpool->cur->addr.p.tbl,
2410                             &rpool->tblidx, &rpool->counter,
2411                             &raddr, &rmask, af))
2412                                 goto get_addr;
2413                 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2414                         if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2415                             &rpool->tblidx, &rpool->counter,
2416                             &raddr, &rmask, af))
2417                                 goto get_addr;
2418                 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af))
2419                         goto get_addr;
2420
2421         try_next:
2422                 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL)
2423                         rpool->cur = TAILQ_FIRST(&rpool->list);
2424                 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2425                         rpool->tblidx = -1;
2426                         if (pfr_pool_get(rpool->cur->addr.p.tbl,
2427                             &rpool->tblidx, &rpool->counter,
2428                             &raddr, &rmask, af)) {
2429                                 /* table contains no address of type 'af' */
2430                                 if (rpool->cur != acur)
2431                                         goto try_next;
2432                                 return (1);
2433                         }
2434                 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2435                         rpool->tblidx = -1;
2436                         if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2437                             &rpool->tblidx, &rpool->counter,
2438                             &raddr, &rmask, af)) {
2439                                 /* table contains no address of type 'af' */
2440                                 if (rpool->cur != acur)
2441                                         goto try_next;
2442                                 return (1);
2443                         }
2444                 } else {
2445                         raddr = &rpool->cur->addr.v.a.addr;
2446                         rmask = &rpool->cur->addr.v.a.mask;
2447                         PF_ACPY(&rpool->counter, raddr, af);
2448                 }
2449
2450         get_addr:
2451                 PF_ACPY(naddr, &rpool->counter, af);
2452                 if (init_addr != NULL && PF_AZERO(init_addr, af))
2453                         PF_ACPY(init_addr, naddr, af);
2454                 PF_AINC(&rpool->counter, af);
2455                 break;
2456         }
2457         if (*sn != NULL)
2458                 PF_ACPY(&(*sn)->raddr, naddr, af);
2459
2460         if (pf_status.debug >= PF_DEBUG_MISC &&
2461             (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2462                 kprintf("pf_map_addr: selected address ");
2463                 pf_print_host(naddr, 0, af);
2464                 kprintf("\n");
2465         }
2466
2467         return (0);
2468 }
2469
2470 int
2471 pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r,
2472     struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport,
2473     struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high,
2474     struct pf_src_node **sn)
2475 {
2476         struct pf_state_key_cmp key;
2477         struct pf_addr          init_addr;
2478         u_int16_t               cut;
2479
2480         bzero(&init_addr, sizeof(init_addr));
2481         if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2482                 return (1);
2483
2484         if (proto == IPPROTO_ICMP) {
2485                 low = 1;
2486                 high = 65535;
2487         }
2488
2489         do {
2490                 key.af = af;
2491                 key.proto = proto;
2492                 PF_ACPY(&key.addr[1], daddr, key.af);
2493                 PF_ACPY(&key.addr[0], naddr, key.af);
2494                 key.port[1] = dport;
2495
2496                 /*
2497                  * port search; start random, step;
2498                  * similar 2 portloop in in_pcbbind
2499                  */
2500                 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP ||
2501                     proto == IPPROTO_ICMP)) {
2502                         key.port[0] = dport;
2503                         if (pf_find_state_all(&key, PF_IN, NULL) == NULL)
2504                                 return (0);
2505                 } else if (low == 0 && high == 0) {
2506                         key.port[0] = *nport;
2507                         if (pf_find_state_all(&key, PF_IN, NULL) == NULL)
2508                                 return (0);
2509                 } else if (low == high) {
2510                         key.port[0] = htons(low);
2511                         if (pf_find_state_all(&key, PF_IN, NULL) == NULL) {
2512                                 *nport = htons(low);
2513                                 return (0);
2514                         }
2515                 } else {
2516                         u_int16_t tmp;
2517
2518                         if (low > high) {
2519                                 tmp = low;
2520                                 low = high;
2521                                 high = tmp;
2522                         }
2523                         /* low < high */
2524                         cut = htonl(karc4random()) % (1 + high - low) + low;
2525                         /* low <= cut <= high */
2526                         for (tmp = cut; tmp <= high; ++(tmp)) {
2527                                 key.port[0] = htons(tmp);
2528                                 if (pf_find_state_all(&key, PF_IN, NULL) ==
2529                                     NULL && !in_baddynamic(tmp, proto)) {
2530                                         *nport = htons(tmp);
2531                                         return (0);
2532                                 }
2533                         }
2534                         for (tmp = cut - 1; tmp >= low; --(tmp)) {
2535                                 key.port[0] = htons(tmp);
2536                                 if (pf_find_state_all(&key, PF_IN, NULL) ==
2537                                     NULL && !in_baddynamic(tmp, proto)) {
2538                                         *nport = htons(tmp);
2539                                         return (0);
2540                                 }
2541                         }
2542                 }
2543
2544                 switch (r->rpool.opts & PF_POOL_TYPEMASK) {
2545                 case PF_POOL_RANDOM:
2546                 case PF_POOL_ROUNDROBIN:
2547                         if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2548                                 return (1);
2549                         break;
2550                 case PF_POOL_NONE:
2551                 case PF_POOL_SRCHASH:
2552                 case PF_POOL_BITMASK:
2553                 default:
2554                         return (1);
2555                 }
2556         } while (! PF_AEQ(&init_addr, naddr, af) );
2557         return (1);                                     /* none available */
2558 }
2559
2560 struct pf_rule *
2561 pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off,
2562     int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport,
2563     struct pf_addr *daddr, u_int16_t dport, int rs_num)
2564 {
2565         struct pf_rule          *r, *rm = NULL;
2566         struct pf_ruleset       *ruleset = NULL;
2567         int                      tag = -1;
2568         int                      rtableid = -1;
2569         int                      asd = 0;
2570
2571         r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr);
2572         while (r && rm == NULL) {
2573                 struct pf_rule_addr     *src = NULL, *dst = NULL;
2574                 struct pf_addr_wrap     *xdst = NULL;
2575
2576                 if (r->action == PF_BINAT && direction == PF_IN) {
2577                         src = &r->dst;
2578                         if (r->rpool.cur != NULL)
2579                                 xdst = &r->rpool.cur->addr;
2580                 } else {
2581                         src = &r->src;
2582                         dst = &r->dst;
2583                 }
2584
2585                 r->evaluations++;
2586                 if (pfi_kif_match(r->kif, kif) == r->ifnot)
2587                         r = r->skip[PF_SKIP_IFP].ptr;
2588                 else if (r->direction && r->direction != direction)
2589                         r = r->skip[PF_SKIP_DIR].ptr;
2590                 else if (r->af && r->af != pd->af)
2591                         r = r->skip[PF_SKIP_AF].ptr;
2592                 else if (r->proto && r->proto != pd->proto)
2593                         r = r->skip[PF_SKIP_PROTO].ptr;
2594                 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af,
2595                     src->neg, kif))
2596                         r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR :
2597                             PF_SKIP_DST_ADDR].ptr;
2598                 else if (src->port_op && !pf_match_port(src->port_op,
2599                     src->port[0], src->port[1], sport))
2600                         r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT :
2601                             PF_SKIP_DST_PORT].ptr;
2602                 else if (dst != NULL &&
2603                     PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->neg, NULL))
2604                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
2605                 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af,
2606                     0, NULL))
2607                         r = TAILQ_NEXT(r, entries);
2608                 else if (dst != NULL && dst->port_op &&
2609                     !pf_match_port(dst->port_op, dst->port[0],
2610                     dst->port[1], dport))
2611                         r = r->skip[PF_SKIP_DST_PORT].ptr;
2612                 else if (r->match_tag && !pf_match_tag(m, r, &tag))
2613                         r = TAILQ_NEXT(r, entries);
2614                 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto !=
2615                     IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m,
2616                     off, pd->hdr.tcp), r->os_fingerprint)))
2617                         r = TAILQ_NEXT(r, entries);
2618                 else {
2619                         if (r->tag)
2620                                 tag = r->tag;
2621                         if (r->rtableid >= 0)
2622                                 rtableid = r->rtableid;
2623                         if (r->anchor == NULL) {
2624                                 rm = r;
2625                         } else
2626                                 pf_step_into_anchor(&asd, &ruleset, rs_num,
2627                                     &r, NULL, NULL);
2628                 }
2629                 if (r == NULL)
2630                         pf_step_out_of_anchor(&asd, &ruleset, rs_num, &r,
2631                             NULL, NULL);
2632         }
2633         if (pf_tag_packet(m, tag, rtableid))
2634                 return (NULL);
2635         if (rm != NULL && (rm->action == PF_NONAT ||
2636             rm->action == PF_NORDR || rm->action == PF_NOBINAT))
2637                 return (NULL);
2638         return (rm);
2639 }
2640
2641 struct pf_rule *
2642 pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction,
2643     struct pfi_kif *kif, struct pf_src_node **sn,
2644     struct pf_state_key **skw, struct pf_state_key **sks,
2645     struct pf_state_key **skp, struct pf_state_key **nkp,
2646     struct pf_addr *saddr, struct pf_addr *daddr,
2647     u_int16_t sport, u_int16_t dport)
2648 {
2649         struct pf_rule  *r = NULL;
2650
2651
2652         if (direction == PF_OUT) {
2653                 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2654                     sport, daddr, dport, PF_RULESET_BINAT);
2655                 if (r == NULL)
2656                         r = pf_match_translation(pd, m, off, direction, kif,
2657                             saddr, sport, daddr, dport, PF_RULESET_NAT);
2658         } else {
2659                 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2660                     sport, daddr, dport, PF_RULESET_RDR);
2661                 if (r == NULL)
2662                         r = pf_match_translation(pd, m, off, direction, kif,
2663                             saddr, sport, daddr, dport, PF_RULESET_BINAT);
2664         }
2665
2666         if (r != NULL) {
2667                 struct pf_addr  *naddr;
2668                 u_int16_t       *nport;
2669
2670                 if (pf_state_key_setup(pd, r, skw, sks, skp, nkp,
2671                     saddr, daddr, sport, dport))
2672                         return r;
2673
2674                 /* XXX We only modify one side for now. */
2675                 naddr = &(*nkp)->addr[1];
2676                 nport = &(*nkp)->port[1];
2677
2678                 /*
2679                  * NOTE: Currently all translations will clear
2680                  *       BRIDGE_MBUF_TAGGED, telling the bridge to
2681                  *       ignore the original input encapsulation.
2682                  */
2683                 switch (r->action) {
2684                 case PF_NONAT:
2685                 case PF_NOBINAT:
2686                 case PF_NORDR:
2687                         return (NULL);
2688                 case PF_NAT:
2689                         m->m_pkthdr.fw_flags &= ~BRIDGE_MBUF_TAGGED;
2690                         if (pf_get_sport(pd->af, pd->proto, r, saddr,
2691                             daddr, dport, naddr, nport, r->rpool.proxy_port[0],
2692                             r->rpool.proxy_port[1], sn)) {
2693                                 DPFPRINTF(PF_DEBUG_MISC,
2694                                     ("pf: NAT proxy port allocation "
2695                                     "(%u-%u) failed\n",
2696                                     r->rpool.proxy_port[0],
2697                                     r->rpool.proxy_port[1]));
2698                                 return (NULL);
2699                         }
2700                         break;
2701                 case PF_BINAT:
2702                         m->m_pkthdr.fw_flags &= ~BRIDGE_MBUF_TAGGED;
2703                         switch (direction) {
2704                         case PF_OUT:
2705                                 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){
2706                                         switch (pd->af) {
2707 #ifdef INET
2708                                         case AF_INET:
2709                                                 if (r->rpool.cur->addr.p.dyn->
2710                                                     pfid_acnt4 < 1)
2711                                                         return (NULL);
2712                                                 PF_POOLMASK(naddr,
2713                                                     &r->rpool.cur->addr.p.dyn->
2714                                                     pfid_addr4,
2715                                                     &r->rpool.cur->addr.p.dyn->
2716                                                     pfid_mask4,
2717                                                     saddr, AF_INET);
2718                                                 break;
2719 #endif /* INET */
2720 #ifdef INET6
2721                                         case AF_INET6:
2722                                                 if (r->rpool.cur->addr.p.dyn->
2723                                                     pfid_acnt6 < 1)
2724                                                         return (NULL);
2725                                                 PF_POOLMASK(naddr,
2726                                                     &r->rpool.cur->addr.p.dyn->
2727                                                     pfid_addr6,
2728                                                     &r->rpool.cur->addr.p.dyn->
2729                                                     pfid_mask6,
2730                                                     saddr, AF_INET6);
2731                                                 break;
2732 #endif /* INET6 */
2733                                         }
2734                                 } else
2735                                         PF_POOLMASK(naddr,
2736                                             &r->rpool.cur->addr.v.a.addr,
2737                                             &r->rpool.cur->addr.v.a.mask,
2738                                             saddr, pd->af);
2739                                 break;
2740                         case PF_IN:
2741                                 if (r->src.addr.type == PF_ADDR_DYNIFTL) {
2742                                         switch (pd->af) {
2743 #ifdef INET
2744                                         case AF_INET:
2745                                                 if (r->src.addr.p.dyn->
2746                                                     pfid_acnt4 < 1)
2747                                                         return (NULL);
2748                                                 PF_POOLMASK(naddr,
2749                                                     &r->src.addr.p.dyn->
2750                                                     pfid_addr4,
2751                                                     &r->src.addr.p.dyn->
2752                                                     pfid_mask4,
2753                                                     daddr, AF_INET);
2754                                                 break;
2755 #endif /* INET */
2756 #ifdef INET6
2757                                         case AF_INET6:
2758                                                 if (r->src.addr.p.dyn->
2759                                                     pfid_acnt6 < 1)
2760                                                         return (NULL);
2761                                                 PF_POOLMASK(naddr,
2762                                                     &r->src.addr.p.dyn->
2763                                                     pfid_addr6,
2764                                                     &r->src.addr.p.dyn->
2765                                                     pfid_mask6,
2766                                                     daddr, AF_INET6);
2767                                                 break;
2768 #endif /* INET6 */
2769                                         }
2770                                 } else
2771                                         PF_POOLMASK(naddr,
2772                                             &r->src.addr.v.a.addr,
2773                                             &r->src.addr.v.a.mask, daddr,
2774                                             pd->af);
2775                                 break;
2776                         }
2777                         break;
2778                 case PF_RDR: {
2779                         m->m_pkthdr.fw_flags &= ~BRIDGE_MBUF_TAGGED;
2780                         if (pf_map_addr(pd->af, r, saddr, naddr, NULL, sn))
2781                                 return (NULL);
2782                         if ((r->rpool.opts & PF_POOL_TYPEMASK) ==
2783                             PF_POOL_BITMASK)
2784                                 PF_POOLMASK(naddr, naddr,
2785                                     &r->rpool.cur->addr.v.a.mask, daddr,
2786                                     pd->af);
2787
2788                         if (r->rpool.proxy_port[1]) {
2789                                 u_int32_t       tmp_nport;
2790
2791                                 tmp_nport = ((ntohs(dport) -
2792                                     ntohs(r->dst.port[0])) %
2793                                     (r->rpool.proxy_port[1] -
2794                                     r->rpool.proxy_port[0] + 1)) +
2795                                     r->rpool.proxy_port[0];
2796
2797                                 /* wrap around if necessary */
2798                                 if (tmp_nport > 65535)
2799                                         tmp_nport -= 65535;
2800                                 *nport = htons((u_int16_t)tmp_nport);
2801                         } else if (r->rpool.proxy_port[0])
2802                                 *nport = htons(r->rpool.proxy_port[0]);
2803                         break;
2804                 }
2805                 default:
2806                         return (NULL);
2807                 }
2808         }
2809
2810         return (r);
2811 }
2812
2813 #ifdef SMP
2814 struct netmsg_hashlookup {
2815         struct netmsg_base      base;
2816         struct inpcb            **nm_pinp;
2817         struct inpcbinfo        *nm_pcbinfo;
2818         struct pf_addr          *nm_saddr;
2819         struct pf_addr          *nm_daddr;
2820         uint16_t                nm_sport;
2821         uint16_t                nm_dport;
2822         sa_family_t             nm_af;
2823 };
2824
2825 #ifdef PF_SOCKET_LOOKUP_DOMSG
2826 static void
2827 in_pcblookup_hash_handler(netmsg_t msg)
2828 {
2829         struct netmsg_hashlookup *rmsg = (struct netmsg_hashlookup *)msg;
2830
2831         if (rmsg->nm_af == AF_INET)
2832                 *rmsg->nm_pinp = in_pcblookup_hash(rmsg->nm_pcbinfo,
2833                     rmsg->nm_saddr->v4, rmsg->nm_sport, rmsg->nm_daddr->v4,
2834                     rmsg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2835 #ifdef INET6
2836         else
2837                 *rmsg->nm_pinp = in6_pcblookup_hash(rmsg->nm_pcbinfo,
2838                     &rmsg->nm_saddr->v6, rmsg->nm_sport, &rmsg->nm_daddr->v6,
2839                     rmsg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2840 #endif /* INET6 */
2841         lwkt_replymsg(&rmsg->base.lmsg, 0);
2842 }
2843 #endif  /* PF_SOCKET_LOOKUP_DOMSG */
2844
2845 #endif /* SMP */
2846
2847 int
2848 pf_socket_lookup(int direction, struct pf_pdesc *pd)
2849 {
2850         struct pf_addr          *saddr, *daddr;
2851         u_int16_t                sport, dport;
2852         struct inpcbinfo        *pi;
2853         struct inpcb            *inp;
2854 #ifdef SMP
2855         struct netmsg_hashlookup *msg = NULL;
2856 #ifdef PF_SOCKET_LOOKUP_DOMSG
2857         struct netmsg_hashlookup msg0;
2858 #endif
2859 #endif
2860         int                      pi_cpu = 0;
2861
2862         if (pd == NULL)
2863                 return (-1);
2864         pd->lookup.uid = UID_MAX;
2865         pd->lookup.gid = GID_MAX;
2866         pd->lookup.pid = NO_PID;
2867         if (direction == PF_IN) {
2868                 saddr = pd->src;
2869                 daddr = pd->dst;
2870         } else {
2871                 saddr = pd->dst;
2872                 daddr = pd->src;
2873         }
2874         switch (pd->proto) {
2875         case IPPROTO_TCP:
2876                 if (pd->hdr.tcp == NULL)
2877                         return (-1);
2878                 sport = pd->hdr.tcp->th_sport;
2879                 dport = pd->hdr.tcp->th_dport;
2880
2881                 pi_cpu = tcp_addrcpu(saddr->v4.s_addr, sport, daddr->v4.s_addr, dport);
2882                 pi = &tcbinfo[pi_cpu];
2883 #ifdef SMP
2884                 /*
2885                  * Our netstack runs lockless on MP systems
2886                  * (only for TCP connections at the moment).
2887                  * 
2888                  * As we are not allowed to read another CPU's tcbinfo,
2889                  * we have to ask that CPU via remote call to search the
2890                  * table for us.
2891                  * 
2892                  * Prepare a msg iff data belongs to another CPU.
2893                  */
2894                 if (pi_cpu != mycpu->gd_cpuid) {
2895 #ifdef PF_SOCKET_LOOKUP_DOMSG
2896                         /*
2897                          * NOTE:
2898                          *
2899                          * Following lwkt_domsg() is dangerous and could
2900                          * lockup the network system, e.g.
2901                          *
2902                          * On 2 CPU system:
2903                          * netisr0 domsg to netisr1 (due to lookup)
2904                          * netisr1 domsg to netisr0 (due to lookup)
2905                          *
2906                          * We simply return -1 here, since we are probably
2907                          * called before NAT, so the TCP packet should
2908                          * already be on the correct CPU.
2909                          */
2910                         msg = &msg0;
2911                         netmsg_init(&msg->base, NULL, &curthread->td_msgport,
2912                                     0, in_pcblookup_hash_handler);
2913                         msg->nm_pinp = &inp;
2914                         msg->nm_pcbinfo = pi;
2915                         msg->nm_saddr = saddr;
2916                         msg->nm_sport = sport;
2917                         msg->nm_daddr = daddr;
2918                         msg->nm_dport = dport;
2919                         msg->nm_af = pd->af;
2920 #else   /* !PF_SOCKET_LOOKUP_DOMSG */
2921                         kprintf("pf_socket_lookup: tcp packet not on the "
2922                                 "correct cpu %d, cur cpu %d\n",
2923                                 pi_cpu, mycpuid);
2924                         print_backtrace(-1);
2925                         return -1;
2926 #endif  /* PF_SOCKET_LOOKUP_DOMSG */
2927                 }
2928 #endif /* SMP */
2929                 break;
2930         case IPPROTO_UDP:
2931                 if (pd->hdr.udp == NULL)
2932                         return (-1);
2933                 sport = pd->hdr.udp->uh_sport;
2934                 dport = pd->hdr.udp->uh_dport;
2935                 pi = &udbinfo;
2936                 break;
2937         default:
2938                 return (-1);
2939         }
2940         if (direction != PF_IN) {
2941                 u_int16_t       p;
2942
2943                 p = sport;
2944                 sport = dport;
2945                 dport = p;
2946         }
2947         switch (pd->af) {
2948 #ifdef INET6
2949         case AF_INET6:
2950 #ifdef SMP
2951                 /*
2952                  * Query other CPU, second part
2953                  * 
2954                  * msg only gets initialized when:
2955                  * 1) packet is TCP
2956                  * 2) the info belongs to another CPU
2957                  *
2958                  * Use some switch/case magic to avoid code duplication.
2959                  */
2960                 if (msg == NULL)
2961 #endif /* SMP */
2962                 {
2963                         inp = in6_pcblookup_hash(pi, &saddr->v6, sport,
2964                             &daddr->v6, dport, INPLOOKUP_WILDCARD, NULL);
2965
2966                         if (inp == NULL)
2967                                 return (-1);
2968                         break;
2969                 }
2970                 /* FALLTHROUGH if SMP and on other CPU */
2971 #endif /* INET6 */
2972         case AF_INET:
2973 #ifdef SMP
2974                 if (msg != NULL) {
2975                         lwkt_domsg(cpu_portfn(pi_cpu),
2976                                      &msg->base.lmsg, 0);
2977                 } else
2978 #endif /* SMP */
2979                 {
2980                         inp = in_pcblookup_hash(pi, saddr->v4, sport, daddr->v4,
2981                             dport, INPLOOKUP_WILDCARD, NULL);
2982                 }
2983                 if (inp == NULL)
2984                         return (-1);
2985                 break;
2986
2987         default:
2988                 return (-1);
2989         }
2990         pd->lookup.uid = inp->inp_socket->so_cred->cr_uid;
2991         pd->lookup.gid = inp->inp_socket->so_cred->cr_groups[0];
2992         return (1);
2993 }
2994
2995 u_int8_t
2996 pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2997 {
2998         int              hlen;
2999         u_int8_t         hdr[60];
3000         u_int8_t        *opt, optlen;
3001         u_int8_t         wscale = 0;
3002
3003         hlen = th_off << 2;             /* hlen <= sizeof(hdr) */
3004         if (hlen <= sizeof(struct tcphdr))
3005                 return (0);
3006         if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
3007                 return (0);
3008         opt = hdr + sizeof(struct tcphdr);
3009         hlen -= sizeof(struct tcphdr);
3010         while (hlen >= 3) {
3011                 switch (*opt) {
3012                 case TCPOPT_EOL:
3013                 case TCPOPT_NOP:
3014                         ++opt;
3015                         --hlen;
3016                         break;
3017                 case TCPOPT_WINDOW:
3018                         wscale = opt[2];
3019                         if (wscale > TCP_MAX_WINSHIFT)
3020                                 wscale = TCP_MAX_WINSHIFT;
3021                         wscale |= PF_WSCALE_FLAG;
3022                         /* FALLTHROUGH */
3023                 default:
3024                         optlen = opt[1];
3025                         if (optlen < 2)
3026                                 optlen = 2;
3027                         hlen -= optlen;
3028                         opt += optlen;
3029                         break;
3030                 }
3031         }
3032         return (wscale);
3033 }
3034
3035 u_int16_t
3036 pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
3037 {
3038         int              hlen;
3039         u_int8_t         hdr[60];
3040         u_int8_t        *opt, optlen;
3041         u_int16_t        mss = tcp_mssdflt;
3042
3043         hlen = th_off << 2;     /* hlen <= sizeof(hdr) */
3044         if (hlen <= sizeof(struct tcphdr))
3045                 return (0);
3046         if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
3047                 return (0);
3048         opt = hdr + sizeof(struct tcphdr);
3049         hlen -= sizeof(struct tcphdr);
3050         while (hlen >= TCPOLEN_MAXSEG) {
3051                 switch (*opt) {
3052                 case TCPOPT_EOL:
3053                 case TCPOPT_NOP:
3054                         ++opt;
3055                         --hlen;
3056                         break;
3057                 case TCPOPT_MAXSEG:
3058                         bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2);
3059                         /* FALLTHROUGH */
3060                 default:
3061                         optlen = opt[1];
3062                         if (optlen < 2)
3063                                 optlen = 2;
3064                         hlen -= optlen;
3065                         opt += optlen;
3066                         break;
3067                 }
3068         }
3069         return (mss);
3070 }
3071
3072 u_int16_t
3073 pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer)
3074 {
3075 #ifdef INET
3076         struct sockaddr_in      *dst;
3077         struct route             ro;
3078 #endif /* INET */
3079 #ifdef INET6
3080         struct sockaddr_in6     *dst6;
3081         struct route_in6         ro6;
3082 #endif /* INET6 */
3083         struct rtentry          *rt = NULL;
3084         int                      hlen = 0;
3085         u_int16_t                mss = tcp_mssdflt;
3086
3087         switch (af) {
3088 #ifdef INET
3089         case AF_INET:
3090                 hlen = sizeof(struct ip);
3091                 bzero(&ro, sizeof(ro));
3092                 dst = (struct sockaddr_in *)&ro.ro_dst;
3093                 dst->sin_family = AF_INET;
3094                 dst->sin_len = sizeof(*dst);
3095                 dst->sin_addr = addr->v4;
3096                 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
3097                 rt = ro.ro_rt;
3098                 break;
3099 #endif /* INET */
3100 #ifdef INET6
3101         case AF_INET6:
3102                 hlen = sizeof(struct ip6_hdr);
3103                 bzero(&ro6, sizeof(ro6));
3104                 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst;
3105                 dst6->sin6_family = AF_INET6;
3106                 dst6->sin6_len = sizeof(*dst6);
3107                 dst6->sin6_addr = addr->v6;
3108                 rtalloc_ign((struct route *)&ro6, (RTF_CLONING | RTF_PRCLONING));
3109                 rt = ro6.ro_rt;
3110                 break;
3111 #endif /* INET6 */
3112         }
3113
3114         if (rt && rt->rt_ifp) {
3115                 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr);
3116                 mss = max(tcp_mssdflt, mss);
3117                 RTFREE(rt);
3118         }
3119         mss = min(mss, offer);
3120         mss = max(mss, 64);             /* sanity - at least max opt space */
3121         return (mss);
3122 }
3123
3124 void
3125 pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr)
3126 {
3127         struct pf_rule *r = s->rule.ptr;
3128
3129         s->rt_kif = NULL;
3130         if (!r->rt || r->rt == PF_FASTROUTE)
3131                 return;
3132         switch (s->key[PF_SK_WIRE]->af) {
3133 #ifdef INET
3134         case AF_INET:
3135                 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL,
3136                     &s->nat_src_node);
3137                 s->rt_kif = r->rpool.cur->kif;
3138                 break;
3139 #endif /* INET */
3140 #ifdef INET6
3141         case AF_INET6:
3142                 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL,
3143                     &s->nat_src_node);
3144                 s->rt_kif = r->rpool.cur->kif;
3145                 break;
3146 #endif /* INET6 */
3147         }
3148 }
3149
3150 u_int32_t
3151 pf_tcp_iss(struct pf_pdesc *pd)
3152 {
3153         MD5_CTX ctx;
3154         u_int32_t digest[4];
3155
3156         if (pf_tcp_secret_init == 0) {
3157                 karc4rand(pf_tcp_secret, sizeof(pf_tcp_secret));
3158                 MD5Init(&pf_tcp_secret_ctx);
3159                 MD5Update(&pf_tcp_secret_ctx, pf_tcp_secret,
3160                     sizeof(pf_tcp_secret));
3161                 pf_tcp_secret_init = 1;
3162         }
3163         ctx = pf_tcp_secret_ctx;
3164
3165         MD5Update(&ctx, (char *)&pd->hdr.tcp->th_sport, sizeof(u_short));
3166         MD5Update(&ctx, (char *)&pd->hdr.tcp->th_dport, sizeof(u_short));
3167         if (pd->af == AF_INET6) {
3168                 MD5Update(&ctx, (char *)&pd->src->v6, sizeof(struct in6_addr));
3169                 MD5Update(&ctx, (char *)&pd->dst->v6, sizeof(struct in6_addr));
3170         } else {
3171                 MD5Update(&ctx, (char *)&pd->src->v4, sizeof(struct in_addr));
3172                 MD5Update(&ctx, (char *)&pd->dst->v4, sizeof(struct in_addr));
3173         }
3174         MD5Final((u_char *)digest, &ctx);
3175         pf_tcp_iss_off += 4096;
3176         return (digest[0] + pd->hdr.tcp->th_seq + pf_tcp_iss_off);
3177 }
3178
3179 int
3180 pf_test_rule(struct pf_rule **rm, struct pf_state **sm, int direction,
3181     struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3182     struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
3183     struct ifqueue *ifq, struct inpcb *inp)
3184 {
3185         struct pf_rule          *nr = NULL;
3186         struct pf_addr          *saddr = pd->src, *daddr = pd->dst;
3187         sa_family_t              af = pd->af;
3188         struct pf_rule          *r, *a = NULL;
3189         struct pf_ruleset       *ruleset = NULL;
3190         struct pf_src_node      *nsn = NULL;
3191         struct tcphdr           *th = pd->hdr.tcp;
3192         struct pf_state_key     *skw = NULL, *sks = NULL;
3193         struct pf_state_key     *sk = NULL, *nk = NULL;
3194         u_short                  reason;
3195         int                      rewrite = 0, hdrlen = 0;
3196         int                      tag = -1, rtableid = -1;
3197         int                      asd = 0;
3198         int                      match = 0;
3199         int                      state_icmp = 0;
3200         u_int16_t                sport = 0, dport = 0;
3201         u_int16_t                nport = 0, bport = 0;
3202         u_int16_t                bproto_sum = 0, bip_sum = 0;
3203         u_int8_t                 icmptype = 0, icmpcode = 0;
3204
3205
3206         if (direction == PF_IN && pf_check_congestion(ifq)) {
3207                 REASON_SET(&reason, PFRES_CONGEST);
3208                 return (PF_DROP);
3209         }
3210
3211         if (inp != NULL)
3212                 pd->lookup.done = pf_socket_lookup(direction, pd);
3213         else if (debug_pfugidhack) { 
3214                 DPFPRINTF(PF_DEBUG_MISC, ("pf: unlocked lookup\n"));
3215                 pd->lookup.done = pf_socket_lookup(direction, pd);
3216         }
3217
3218         switch (pd->proto) {
3219         case IPPROTO_TCP:
3220                 sport = th->th_sport;
3221                 dport = th->th_dport;
3222                 hdrlen = sizeof(*th);
3223                 break;
3224         case IPPROTO_UDP:
3225                 sport = pd->hdr.udp->uh_sport;
3226                 dport = pd->hdr.udp->uh_dport;
3227                 hdrlen = sizeof(*pd->hdr.udp);
3228                 break;
3229 #ifdef INET
3230         case IPPROTO_ICMP:
3231                 if (pd->af != AF_INET)
3232                         break;
3233                 sport = dport = pd->hdr.icmp->icmp_id;
3234                 hdrlen = sizeof(*pd->hdr.icmp);
3235                 icmptype = pd->hdr.icmp->icmp_type;
3236                 icmpcode = pd->hdr.icmp->icmp_code;
3237
3238                 if (icmptype == ICMP_UNREACH ||
3239                     icmptype == ICMP_SOURCEQUENCH ||
3240                     icmptype == ICMP_REDIRECT ||
3241                     icmptype == ICMP_TIMXCEED ||
3242                     icmptype == ICMP_PARAMPROB)
3243                         state_icmp++;
3244                 break;
3245 #endif /* INET */
3246 #ifdef INET6
3247         case IPPROTO_ICMPV6:
3248                 if (af != AF_INET6)
3249                         break;
3250                 sport = dport = pd->hdr.icmp6->icmp6_id;
3251                 hdrlen = sizeof(*pd->hdr.icmp6);
3252                 icmptype = pd->hdr.icmp6->icmp6_type;
3253                 icmpcode = pd->hdr.icmp6->icmp6_code;
3254
3255                 if (icmptype == ICMP6_DST_UNREACH ||
3256                     icmptype == ICMP6_PACKET_TOO_BIG ||
3257                     icmptype == ICMP6_TIME_EXCEEDED ||
3258                     icmptype == ICMP6_PARAM_PROB)
3259                         state_icmp++;
3260                 break;
3261 #endif /* INET6 */
3262         default:
3263                 sport = dport = hdrlen = 0;
3264                 break;
3265         }
3266
3267         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3268
3269         bport = nport = sport;
3270         /* check packet for BINAT/NAT/RDR */
3271         if ((nr = pf_get_translation(pd, m, off, direction, kif, &nsn,
3272             &skw, &sks, &sk, &nk, saddr, daddr, sport, dport)) != NULL) {
3273                 if (nk == NULL || sk == NULL) {
3274                         REASON_SET(&reason, PFRES_MEMORY);
3275                         goto cleanup;
3276                 }
3277
3278                 if (pd->ip_sum)
3279                         bip_sum = *pd->ip_sum;
3280
3281                 switch (pd->proto) {
3282                 case IPPROTO_TCP:
3283                         bproto_sum = th->th_sum;
3284                         pd->proto_sum = &th->th_sum;
3285
3286                         if (PF_ANEQ(saddr, &nk->addr[pd->sidx], af) ||
3287                             nk->port[pd->sidx] != sport) {
3288                                 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
3289                                     &th->th_sum, &nk->addr[pd->sidx],
3290                                     nk->port[pd->sidx], 0, af);
3291                                 pd->sport = &th->th_sport;
3292                                 sport = th->th_sport;
3293                         }
3294
3295                         if (PF_ANEQ(daddr, &nk->addr[pd->didx], af) ||
3296                             nk->port[pd->didx] != dport) {
3297                                 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
3298                                     &th->th_sum, &nk->addr[pd->didx],
3299                                     nk->port[pd->didx], 0, af);
3300                                 dport = th->th_dport;
3301                                 pd->dport = &th->th_dport;
3302                         }
3303                         rewrite++;
3304                         break;
3305                 case IPPROTO_UDP:
3306                         bproto_sum = pd->hdr.udp->uh_sum;
3307                         pd->proto_sum = &pd->hdr.udp->uh_sum;
3308
3309                         if (PF_ANEQ(saddr, &nk->addr[pd->sidx], af) ||
3310                             nk->port[pd->sidx] != sport) {
3311                                 pf_change_ap(saddr, &pd->hdr.udp->uh_sport,
3312                                     pd->ip_sum, &pd->hdr.udp->uh_sum,
3313                                     &nk->addr[pd->sidx],
3314                                     nk->port[pd->sidx], 1, af);
3315                                 sport = pd->hdr.udp->uh_sport;
3316                                 pd->sport = &pd->hdr.udp->uh_sport;
3317                         }
3318
3319                         if (PF_ANEQ(daddr, &nk->addr[pd->didx], af) ||
3320                             nk->port[pd->didx] != dport) {
3321                                 pf_change_ap(daddr, &pd->hdr.udp->uh_dport,
3322                                     pd->ip_sum, &pd->hdr.udp->uh_sum,
3323                                     &nk->addr[pd->didx],
3324                                     nk->port[pd->didx], 1, af);
3325                                 dport = pd->hdr.udp->uh_dport;
3326                                 pd->dport = &pd->hdr.udp->uh_dport;
3327                         }
3328                         rewrite++;
3329                         break;
3330 #ifdef INET
3331                 case IPPROTO_ICMP:
3332                         nk->port[0] = nk->port[1];
3333                         if (PF_ANEQ(saddr, &nk->addr[pd->sidx], AF_INET))
3334                                 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3335                                     nk->addr[pd->sidx].v4.s_addr, 0);
3336
3337                         if (PF_ANEQ(daddr, &nk->addr[pd->didx], AF_INET))
3338                                 pf_change_a(&daddr->v4.s_addr, pd->ip_sum,
3339                                     nk->addr[pd->didx].v4.s_addr, 0);
3340
3341                         if (nk->port[1] != pd->hdr.icmp->icmp_id) {
3342                                 pd->hdr.icmp->icmp_cksum = pf_cksum_fixup(
3343                                     pd->hdr.icmp->icmp_cksum, sport,
3344                                     nk->port[1], 0);
3345                                 pd->hdr.icmp->icmp_id = nk->port[1];
3346                                 pd->sport = &pd->hdr.icmp->icmp_id;
3347                         }
3348                         m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
3349                         break;
3350 #endif /* INET */
3351 #ifdef INET6
3352                 case IPPROTO_ICMPV6:
3353                         nk->port[0] = nk->port[1];
3354                         if (PF_ANEQ(saddr, &nk->addr[pd->sidx], AF_INET6))
3355                                 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum,
3356                                     &nk->addr[pd->sidx], 0);
3357
3358                         if (PF_ANEQ(daddr, &nk->addr[pd->didx], AF_INET6))
3359                                 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum,
3360                                     &nk->addr[pd->didx], 0);
3361                         rewrite++;
3362                         break;
3363 #endif /* INET */
3364                 default:
3365                         switch (af) {
3366 #ifdef INET
3367                         case AF_INET:
3368                                 if (PF_ANEQ(saddr,
3369                                     &nk->addr[pd->sidx], AF_INET))
3370                                         pf_change_a(&saddr->v4.s_addr,
3371                                             pd->ip_sum,
3372                                             nk->addr[pd->sidx].v4.s_addr, 0);
3373
3374                                 if (PF_ANEQ(daddr,
3375                                     &nk->addr[pd->didx], AF_INET))
3376                                         pf_change_a(&daddr->v4.s_addr,
3377                                             pd->ip_sum,
3378                                             nk->addr[pd->didx].v4.s_addr, 0);
3379                                 break;
3380 #endif /* INET */
3381 #ifdef INET6
3382                         case AF_INET6:
3383                                 if (PF_ANEQ(saddr,
3384                                     &nk->addr[pd->sidx], AF_INET6))
3385                                         PF_ACPY(saddr, &nk->addr[pd->sidx], af);
3386
3387                                 if (PF_ANEQ(daddr,
3388                                     &nk->addr[pd->didx], AF_INET6))
3389                                         PF_ACPY(saddr, &nk->addr[pd->didx], af);
3390                                 break;
3391 #endif /* INET */
3392                         }
3393                         break;
3394                 }
3395                 if (nr->natpass)
3396                         r = NULL;
3397                 pd->nat_rule = nr;
3398         }
3399
3400         while (r != NULL) {
3401                 r->evaluations++;
3402                 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3403                         r = r->skip[PF_SKIP_IFP].ptr;
3404                 else if (r->direction && r->direction != direction)
3405                         r = r->skip[PF_SKIP_DIR].ptr;
3406                 else if (r->af && r->af != af)
3407                         r = r->skip[PF_SKIP_AF].ptr;
3408                 else if (r->proto && r->proto != pd->proto)
3409                         r = r->skip[PF_SKIP_PROTO].ptr;
3410                 else if (PF_MISMATCHAW(&r->src.addr, saddr, af,
3411                     r->src.neg, kif))
3412                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3413                 /* tcp/udp only. port_op always 0 in other cases */
3414                 else if (r->src.port_op && !pf_match_port(r->src.port_op,
3415                     r->src.port[0], r->src.port[1], sport))
3416                         r = r->skip[PF_SKIP_SRC_PORT].ptr;
3417                 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af,
3418                     r->dst.neg, NULL))
3419                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
3420                 /* tcp/udp only. port_op always 0 in other cases */
3421                 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
3422                     r->dst.port[0], r->dst.port[1], dport))
3423                         r = r->skip[PF_SKIP_DST_PORT].ptr;
3424                 /* icmp only. type always 0 in other cases */
3425                 else if (r->type && r->type != icmptype + 1)
3426                         r = TAILQ_NEXT(r, entries);
3427                 /* icmp only. type always 0 in other cases */
3428                 else if (r->code && r->code != icmpcode + 1)
3429                         r = TAILQ_NEXT(r, entries);
3430                 else if (r->tos && !(r->tos == pd->tos))
3431                         r = TAILQ_NEXT(r, entries);
3432                 else if (r->rule_flag & PFRULE_FRAGMENT)
3433                         r = TAILQ_NEXT(r, entries);
3434                 else if (pd->proto == IPPROTO_TCP &&
3435                     (r->flagset & th->th_flags) != r->flags)
3436                         r = TAILQ_NEXT(r, entries);
3437                 /* tcp/udp only. uid.op always 0 in other cases */
3438                 else if (r->uid.op && (pd->lookup.done || (pd->lookup.done =
3439                     pf_socket_lookup(direction, pd), 1)) &&
3440                     !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
3441                     pd->lookup.uid))
3442                         r = TAILQ_NEXT(r, entries);
3443                 /* tcp/udp only. gid.op always 0 in other cases */
3444                 else if (r->gid.op && (pd->lookup.done || (pd->lookup.done =
3445                     pf_socket_lookup(direction, pd), 1)) &&
3446                     !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
3447                     pd->lookup.gid))
3448                         r = TAILQ_NEXT(r, entries);
3449                 else if (r->prob &&
3450                   r->prob <= karc4random())
3451                         r = TAILQ_NEXT(r, entries);
3452                 else if (r->match_tag && !pf_match_tag(m, r, &tag))
3453                         r = TAILQ_NEXT(r, entries);
3454                 else if (r->os_fingerprint != PF_OSFP_ANY &&
3455                     (pd->proto != IPPROTO_TCP || !pf_osfp_match(
3456                     pf_osfp_fingerprint(pd, m, off, th),
3457                     r->os_fingerprint)))
3458                         r = TAILQ_NEXT(r, entries);
3459                 else {
3460                         if (r->tag)
3461                                 tag = r->tag;
3462                         if (r->rtableid >= 0)
3463                                 rtableid = r->rtableid;
3464                         if (r->anchor == NULL) {
3465                                 match = 1;
3466                                 *rm = r;
3467                                 *am = a;
3468                                 *rsm = ruleset;
3469                                 if ((*rm)->quick)
3470                                         break;
3471                                 r = TAILQ_NEXT(r, entries);
3472                         } else
3473                                 pf_step_into_anchor(&asd, &ruleset,
3474                                     PF_RULESET_FILTER, &r, &a, &match);
3475                 }
3476                 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3477                     PF_RULESET_FILTER, &r, &a, &match))
3478                         break;
3479         }
3480         r = *rm;
3481         a = *am;
3482         ruleset = *rsm;
3483
3484         REASON_SET(&reason, PFRES_MATCH);
3485
3486         if (r->log || (nr != NULL && nr->log)) {
3487                 if (rewrite)
3488                         m_copyback(m, off, hdrlen, pd->hdr.any);
3489                 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr,
3490                     a, ruleset, pd);
3491         }
3492
3493         if ((r->action == PF_DROP) &&
3494             ((r->rule_flag & PFRULE_RETURNRST) ||
3495             (r->rule_flag & PFRULE_RETURNICMP) ||
3496             (r->rule_flag & PFRULE_RETURN))) {
3497                 /* undo NAT changes, if they have taken place */
3498                 if (nr != NULL) {
3499                         PF_ACPY(saddr, &sk->addr[pd->sidx], af);
3500                         PF_ACPY(daddr, &sk->addr[pd->didx], af);
3501                         if (pd->sport)
3502                                 *pd->sport = sk->port[pd->sidx];
3503                         if (pd->dport)
3504                                 *pd->dport = sk->port[pd->didx];
3505                         if (pd->proto_sum)
3506                                 *pd->proto_sum = bproto_sum;
3507                         if (pd->ip_sum)
3508                                 *pd->ip_sum = bip_sum;
3509                         m_copyback(m, off, hdrlen, pd->hdr.any);
3510                 }
3511                 if (pd->proto == IPPROTO_TCP &&
3512                     ((r->rule_flag & PFRULE_RETURNRST) ||
3513                     (r->rule_flag & PFRULE_RETURN)) &&
3514                     !(th->th_flags & TH_RST)) {
3515                         u_int32_t        ack = ntohl(th->th_seq) + pd->p_len;
3516                         int              len = 0;
3517                         struct ip       *h4;
3518                         struct ip6_hdr  *h6;
3519
3520                         switch (af) {
3521                         case AF_INET:
3522                                 h4 = mtod(m, struct ip *);
3523                                 len = h4->ip_len - off;
3524                                 break;
3525 #ifdef INET6
3526                         case AF_INET6:
3527                                 h6 = mtod(m, struct ip6_hdr *);
3528                                 len = h6->ip6_plen - (off - sizeof(*h6));
3529                                 break;
3530 #endif
3531                         }
3532
3533                         if (pf_check_proto_cksum(m, off, len, IPPROTO_TCP, af))
3534                                 REASON_SET(&reason, PFRES_PROTCKSUM);
3535                         else {
3536                                 if (th->th_flags & TH_SYN)
3537                                         ack++;
3538                                 if (th->th_flags & TH_FIN)
3539                                         ack++;
3540                                 pf_send_tcp(r, af, pd->dst,
3541                                     pd->src, th->th_dport, th->th_sport,
3542                                     ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0,
3543                                     r->return_ttl, 1, 0, pd->eh, kif->pfik_ifp);
3544                         }
3545                 } else if (pd->proto != IPPROTO_ICMP && af == AF_INET &&
3546                     r->return_icmp)
3547                         pf_send_icmp(m, r->return_icmp >> 8,
3548                             r->return_icmp & 255, af, r);
3549                 else if (pd->proto != IPPROTO_ICMPV6 && af == AF_INET6 &&
3550                     r->return_icmp6)
3551                         pf_send_icmp(m, r->return_icmp6 >> 8,
3552                             r->return_icmp6 & 255, af, r);
3553         }
3554
3555         if (r->action == PF_DROP)
3556                 goto cleanup;
3557
3558         if (pf_tag_packet(m, tag, rtableid)) {
3559                 REASON_SET(&reason, PFRES_MEMORY);
3560                 goto cleanup;
3561         }
3562
3563         if (!state_icmp && (r->keep_state || nr != NULL ||
3564             (pd->flags & PFDESC_TCP_NORM))) {
3565                 int action;
3566                 action = pf_create_state(r, nr, a, pd, nsn, skw, sks, nk, sk, m,
3567                     off, sport, dport, &rewrite, kif, sm, tag, bproto_sum,
3568                     bip_sum, hdrlen);
3569                 if (action != PF_PASS)
3570                         return (action);
3571         }
3572
3573         /* copy back packet headers if we performed NAT operations */
3574         if (rewrite)
3575                 m_copyback(m, off, hdrlen, pd->hdr.any);
3576
3577         return (PF_PASS);
3578
3579 cleanup:
3580         if (sk != NULL)
3581                 pool_put(&pf_state_key_pl, sk);
3582         if (nk != NULL)
3583                 pool_put(&pf_state_key_pl, nk);
3584         return (PF_DROP);
3585 }
3586
3587 static __inline int
3588 pf_create_state(struct pf_rule *r, struct pf_rule *nr, struct pf_rule *a,
3589     struct pf_pdesc *pd, struct pf_src_node *nsn, struct pf_state_key *skw,
3590     struct pf_state_key *sks, struct pf_state_key *nk, struct pf_state_key *sk,
3591     struct mbuf *m, int off, u_int16_t sport, u_int16_t dport, int *rewrite,
3592     struct pfi_kif *kif, struct pf_state **sm, int tag, u_int16_t bproto_sum,
3593     u_int16_t bip_sum, int hdrlen)
3594 {
3595         struct pf_state         *s = NULL;
3596         struct pf_src_node      *sn = NULL;
3597         struct tcphdr           *th = pd->hdr.tcp;
3598         u_int16_t                mss = tcp_mssdflt;
3599         u_short                  reason;
3600
3601         /* check maximums */
3602         if (r->max_states && (r->states_cur >= r->max_states)) {
3603                 pf_status.lcounters[LCNT_STATES]++;
3604                 REASON_SET(&reason, PFRES_MAXSTATES);
3605                 return (PF_DROP);
3606         }
3607         /* src node for filter rule */
3608         if ((r->rule_flag & PFRULE_SRCTRACK ||
3609             r->rpool.opts & PF_POOL_STICKYADDR) &&
3610             pf_insert_src_node(&sn, r, pd->src, pd->af) != 0) {
3611                 REASON_SET(&reason, PFRES_SRCLIMIT);
3612                 goto csfailed;
3613         }
3614         /* src node for translation rule */
3615         if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3616             pf_insert_src_node(&nsn, nr, &sk->addr[pd->sidx], pd->af)) {
3617                 REASON_SET(&reason, PFRES_SRCLIMIT);
3618                 goto csfailed;
3619         }
3620         s = pool_get(&pf_state_pl, PR_NOWAIT | PR_ZERO);
3621         if (s == NULL) {
3622                 REASON_SET(&reason, PFRES_MEMORY);
3623                 goto csfailed;
3624         }
3625         s->id = 0; /* XXX Do we really need that? not in OpenBSD */
3626         s->creatorid = 0;
3627         s->rule.ptr = r;
3628         s->nat_rule.ptr = nr;
3629         s->anchor.ptr = a;
3630         STATE_INC_COUNTERS(s);
3631         if (r->allow_opts)
3632                 s->state_flags |= PFSTATE_ALLOWOPTS;
3633         if (r->rule_flag & PFRULE_STATESLOPPY)
3634                 s->state_flags |= PFSTATE_SLOPPY;
3635         s->log = r->log & PF_LOG_ALL;
3636         if (nr != NULL)
3637                 s->log |= nr->log & PF_LOG_ALL;
3638         switch (pd->proto) {
3639         case IPPROTO_TCP:
3640                 s->src.seqlo = ntohl(th->th_seq);
3641                 s->src.seqhi = s->src.seqlo + pd->p_len + 1;
3642                 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
3643                     r->keep_state == PF_STATE_MODULATE) {
3644                         /* Generate sequence number modulator */
3645                         if ((s->src.seqdiff = pf_tcp_iss(pd) - s->src.seqlo) ==
3646                             0)
3647                                 s->src.seqdiff = 1;
3648                         pf_change_a(&th->th_seq, &th->th_sum,
3649                             htonl(s->src.seqlo + s->src.seqdiff), 0);
3650                         *rewrite = 1;
3651                 } else
3652                         s->src.seqdiff = 0;
3653                 if (th->th_flags & TH_SYN) {
3654                         s->src.seqhi++;
3655                         s->src.wscale = pf_get_wscale(m, off,
3656                             th->th_off, pd->af);
3657                 }
3658                 s->src.max_win = MAX(ntohs(th->th_win), 1);
3659                 if (s->src.wscale & PF_WSCALE_MASK) {
3660                         /* Remove scale factor from initial window */
3661                         int win = s->src.max_win;
3662                         win += 1 << (s->src.wscale & PF_WSCALE_MASK);
3663                         s->src.max_win = (win - 1) >>
3664                             (s->src.wscale & PF_WSCALE_MASK);
3665                 }
3666                 if (th->th_flags & TH_FIN)
3667                         s->src.seqhi++;
3668                 s->dst.seqhi = 1;
3669                 s->dst.max_win = 1;
3670                 s->src.state = TCPS_SYN_SENT;
3671                 s->dst.state = TCPS_CLOSED;
3672                 s->timeout = PFTM_TCP_FIRST_PACKET;
3673                 break;
3674         case IPPROTO_UDP:
3675                 s->src.state = PFUDPS_SINGLE;
3676                 s->dst.state = PFUDPS_NO_TRAFFIC;
3677                 s->timeout = PFTM_UDP_FIRST_PACKET;
3678                 break;
3679         case IPPROTO_ICMP:
3680 #ifdef INET6
3681         case IPPROTO_ICMPV6:
3682 #endif
3683                 s->timeout = PFTM_ICMP_FIRST_PACKET;
3684                 break;
3685         default:
3686                 s->src.state = PFOTHERS_SINGLE;
3687                 s->dst.state = PFOTHERS_NO_TRAFFIC;
3688                 s->timeout = PFTM_OTHER_FIRST_PACKET;
3689         }
3690
3691         s->creation = time_second;
3692         s->expire = time_second;
3693
3694         if (sn != NULL) {
3695                 s->src_node = sn;
3696                 s->src_node->states++;
3697         }
3698         if (nsn != NULL) {
3699                 /* XXX We only modify one side for now. */
3700                 PF_ACPY(&nsn->raddr, &nk->addr[1], pd->af);
3701                 s->nat_src_node = nsn;
3702                 s->nat_src_node->states++;
3703         }
3704         if (pd->proto == IPPROTO_TCP) {
3705                 if ((pd->flags & PFDESC_TCP_NORM) && pf_normalize_tcp_init(m,
3706                     off, pd, th, &s->src, &s->dst)) {
3707                         REASON_SET(&reason, PFRES_MEMORY);
3708                         pf_src_tree_remove_state(s);
3709                         STATE_DEC_COUNTERS(s);
3710                         pool_put(&pf_state_pl, s);
3711                         return (PF_DROP);
3712                 }
3713                 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
3714                     pf_normalize_tcp_stateful(m, off, pd, &reason, th, s,
3715                     &s->src, &s->dst, rewrite)) {
3716                         /* This really shouldn't happen!!! */
3717                         DPFPRINTF(PF_DEBUG_URGENT,
3718                             ("pf_normalize_tcp_stateful failed on first pkt"));
3719                         pf_normalize_tcp_cleanup(s);
3720                         pf_src_tree_remove_state(s);
3721                         STATE_DEC_COUNTERS(s);
3722                         pool_put(&pf_state_pl, s);
3723                         return (PF_DROP);
3724                 }
3725         }
3726         s->direction = pd->dir;
3727
3728         if (sk == NULL && pf_state_key_setup(pd, nr, &skw, &sks, &sk, &nk,
3729             pd->src, pd->dst, sport, dport))
3730                 goto csfailed;
3731
3732         if (pf_state_insert(BOUND_IFACE(r, kif), skw, sks, s)) {
3733                 if (pd->proto == IPPROTO_TCP)
3734                         pf_normalize_tcp_cleanup(s);
3735                 REASON_SET(&reason, PFRES_STATEINS);
3736                 pf_src_tree_remove_state(s);
3737                 STATE_DEC_COUNTERS(s);
3738                 pool_put(&pf_state_pl, s);
3739                 return (PF_DROP);
3740         } else
3741                 *sm = s;
3742
3743         pf_set_rt_ifp(s, pd->src);      /* needs s->state_key set */
3744         if (tag > 0) {
3745                 pf_tag_ref(tag);
3746                 s->tag = tag;
3747         }
3748         if (pd->proto == IPPROTO_TCP && (th->th_flags & (TH_SYN|TH_ACK)) ==
3749             TH_SYN && r->keep_state == PF_STATE_SYNPROXY) {
3750                 s->src.state = PF_TCPS_PROXY_SRC;
3751                 /* undo NAT changes, if they have taken place */
3752                 if (nr != NULL) {
3753                         struct pf_state_key *skt = s->key[PF_SK_WIRE];
3754                         if (pd->dir == PF_OUT)
3755                                 skt = s->key[PF_SK_STACK];
3756                         PF_ACPY(pd->src, &skt->addr[pd->sidx], pd->af);
3757                         PF_ACPY(pd->dst, &skt->addr[pd->didx], pd->af);
3758                    &n