1dc8d4db945d262564e33272afd744b3f25e48f2
[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_MP_INITIALIZER(pf_token);
108
109 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) kprintf x
110
111 /*
112  * Global variables
113  */
114
115 /* state tables */
116 struct pf_state_tree     pf_statetbl;
117
118 struct pf_altqqueue      pf_altqs[2];
119 struct pf_palist         pf_pabuf;
120 struct pf_altqqueue     *pf_altqs_active;
121 struct pf_altqqueue     *pf_altqs_inactive;
122 struct pf_status         pf_status;
123
124 u_int32_t                ticket_altqs_active;
125 u_int32_t                ticket_altqs_inactive;
126 int                      altqs_inactive_open;
127 u_int32_t                ticket_pabuf;
128
129 MD5_CTX                  pf_tcp_secret_ctx;
130 u_char                   pf_tcp_secret[16];
131 int                      pf_tcp_secret_init;
132 int                      pf_tcp_iss_off;
133
134 struct pf_anchor_stackframe {
135         struct pf_ruleset                       *rs;
136         struct pf_rule                          *r;
137         struct pf_anchor_node                   *parent;
138         struct pf_anchor                        *child;
139 } pf_anchor_stack[64];
140
141 vm_zone_t                pf_src_tree_pl, pf_rule_pl, pf_pooladdr_pl;
142 vm_zone_t                pf_state_pl, pf_state_key_pl, pf_state_item_pl;
143 vm_zone_t                pf_altq_pl;
144
145 void                     pf_print_host(struct pf_addr *, u_int16_t, u_int8_t);
146
147 void                     pf_init_threshold(struct pf_threshold *, u_int32_t,
148                             u_int32_t);
149 void                     pf_add_threshold(struct pf_threshold *);
150 int                      pf_check_threshold(struct pf_threshold *);
151
152 void                     pf_change_ap(struct pf_addr *, u_int16_t *,
153                             u_int16_t *, u_int16_t *, struct pf_addr *,
154                             u_int16_t, u_int8_t, sa_family_t);
155 int                      pf_modulate_sack(struct mbuf *, int, struct pf_pdesc *,
156                             struct tcphdr *, struct pf_state_peer *);
157 #ifdef INET6
158 void                     pf_change_a6(struct pf_addr *, u_int16_t *,
159                             struct pf_addr *, u_int8_t);
160 #endif /* INET6 */
161 void                     pf_change_icmp(struct pf_addr *, u_int16_t *,
162                             struct pf_addr *, struct pf_addr *, u_int16_t,
163                             u_int16_t *, u_int16_t *, u_int16_t *,
164                             u_int16_t *, u_int8_t, sa_family_t);
165 void                     pf_send_tcp(const struct pf_rule *, sa_family_t,
166                             const struct pf_addr *, const struct pf_addr *,
167                             u_int16_t, u_int16_t, u_int32_t, u_int32_t,
168                             u_int8_t, u_int16_t, u_int16_t, u_int8_t, int,
169                             u_int16_t, struct ether_header *, struct ifnet *);
170 void                     pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t,
171                             sa_family_t, struct pf_rule *);
172 struct pf_rule          *pf_match_translation(struct pf_pdesc *, struct mbuf *,
173                             int, int, struct pfi_kif *,
174                             struct pf_addr *, u_int16_t, struct pf_addr *,
175                             u_int16_t, int);
176 struct pf_rule          *pf_get_translation(struct pf_pdesc *, struct mbuf *,
177                             int, int, struct pfi_kif *, struct pf_src_node **,
178                             struct pf_state_key **, struct pf_state_key **,
179                             struct pf_state_key **, struct pf_state_key **,
180                             struct pf_addr *, struct pf_addr *,
181                             u_int16_t, u_int16_t);
182 void                     pf_detach_state(struct pf_state *);
183 int                      pf_state_key_setup(struct pf_pdesc *, struct pf_rule *,
184                             struct pf_state_key **, struct pf_state_key **,
185                             struct pf_state_key **, struct pf_state_key **,
186                             struct pf_addr *, struct pf_addr *,
187                             u_int16_t, u_int16_t);
188 void                     pf_state_key_detach(struct pf_state *, int);
189 u_int32_t                pf_tcp_iss(struct pf_pdesc *);
190 int                      pf_test_rule(struct pf_rule **, struct pf_state **,
191                             int, struct pfi_kif *, struct mbuf *, int,
192                             void *, struct pf_pdesc *, struct pf_rule **,
193                             struct pf_ruleset **, struct ifqueue *, struct inpcb *);
194 static __inline int      pf_create_state(struct pf_rule *, struct pf_rule *,
195                             struct pf_rule *, struct pf_pdesc *,
196                             struct pf_src_node *, struct pf_state_key *,
197                             struct pf_state_key *, struct pf_state_key *,
198                             struct pf_state_key *, struct mbuf *, int,
199                             u_int16_t, u_int16_t, int *, struct pfi_kif *,
200                             struct pf_state **, int, u_int16_t, u_int16_t,
201                             int);
202 int                      pf_test_fragment(struct pf_rule **, int,
203                             struct pfi_kif *, struct mbuf *, void *,
204                             struct pf_pdesc *, struct pf_rule **,
205                             struct pf_ruleset **);
206 int                      pf_tcp_track_full(struct pf_state_peer *,
207                             struct pf_state_peer *, struct pf_state **,
208                             struct pfi_kif *, struct mbuf *, int,
209                             struct pf_pdesc *, u_short *, int *);
210 int                     pf_tcp_track_sloppy(struct pf_state_peer *,
211                             struct pf_state_peer *, struct pf_state **,
212                             struct pf_pdesc *, u_short *);
213 int                      pf_test_state_tcp(struct pf_state **, int,
214                             struct pfi_kif *, struct mbuf *, int,
215                             void *, struct pf_pdesc *, u_short *);
216 int                      pf_test_state_udp(struct pf_state **, int,
217                             struct pfi_kif *, struct mbuf *, int,
218                             void *, struct pf_pdesc *);
219 int                      pf_test_state_icmp(struct pf_state **, int,
220                             struct pfi_kif *, struct mbuf *, int,
221                             void *, struct pf_pdesc *, u_short *);
222 int                      pf_test_state_other(struct pf_state **, int,
223                             struct pfi_kif *, struct mbuf *, struct pf_pdesc *);
224 void                     pf_step_into_anchor(int *, struct pf_ruleset **, int,
225                             struct pf_rule **, struct pf_rule **, int *);
226 int                      pf_step_out_of_anchor(int *, struct pf_ruleset **,
227                              int, struct pf_rule **, struct pf_rule **,
228                              int *);
229 void                     pf_hash(struct pf_addr *, struct pf_addr *,
230                             struct pf_poolhashkey *, sa_family_t);
231 int                      pf_map_addr(u_int8_t, struct pf_rule *,
232                             struct pf_addr *, struct pf_addr *,
233                             struct pf_addr *, struct pf_src_node **);
234 int                      pf_get_sport(sa_family_t, u_int8_t, struct pf_rule *,
235                             struct pf_addr *, struct pf_addr *, u_int16_t,
236                             struct pf_addr *, u_int16_t*, u_int16_t, u_int16_t,
237                             struct pf_src_node **);
238 void                     pf_route(struct mbuf **, struct pf_rule *, int,
239                             struct ifnet *, struct pf_state *,
240                             struct pf_pdesc *);
241 void                     pf_route6(struct mbuf **, struct pf_rule *, int,
242                             struct ifnet *, struct pf_state *,
243                             struct pf_pdesc *);
244 u_int8_t                 pf_get_wscale(struct mbuf *, int, u_int16_t,
245                             sa_family_t);
246 u_int16_t                pf_get_mss(struct mbuf *, int, u_int16_t,
247                             sa_family_t);
248 u_int16_t                pf_calc_mss(struct pf_addr *, sa_family_t,
249                                 u_int16_t);
250 void                     pf_set_rt_ifp(struct pf_state *,
251                             struct pf_addr *);
252 int                      pf_check_proto_cksum(struct mbuf *, int, int,
253                             u_int8_t, sa_family_t);
254 struct pf_divert        *pf_get_divert(struct mbuf *);
255 void                     pf_print_state_parts(struct pf_state *,
256                             struct pf_state_key *, struct pf_state_key *);
257 int                      pf_addr_wrap_neq(struct pf_addr_wrap *,
258                             struct pf_addr_wrap *);
259 struct pf_state         *pf_find_state(struct pfi_kif *,
260                             struct pf_state_key_cmp *, u_int, struct mbuf *);
261 int                      pf_src_connlimit(struct pf_state **);
262 int                      pf_check_congestion(struct ifqueue *);
263
264 extern int pf_end_threads;
265
266 struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] = {
267         { &pf_state_pl, PFSTATE_HIWAT },
268         { &pf_src_tree_pl, PFSNODE_HIWAT },
269         { &pf_frent_pl, PFFRAG_FRENT_HIWAT },
270         { &pfr_ktable_pl, PFR_KTABLE_HIWAT },
271         { &pfr_kentry_pl, PFR_KENTRY_HIWAT }
272 };
273
274 #define STATE_LOOKUP(i, k, d, s, m)                                     \
275         do {                                                            \
276                 s = pf_find_state(i, k, d, m);                  \
277                 if (s == NULL || (s)->timeout == PFTM_PURGE)            \
278                         return (PF_DROP);                               \
279                 if (d == PF_OUT &&                                      \
280                     (((s)->rule.ptr->rt == PF_ROUTETO &&                \
281                     (s)->rule.ptr->direction == PF_OUT) ||              \
282                     ((s)->rule.ptr->rt == PF_REPLYTO &&                 \
283                     (s)->rule.ptr->direction == PF_IN)) &&              \
284                     (s)->rt_kif != NULL &&                              \
285                     (s)->rt_kif != i)                                   \
286                         return (PF_PASS);                               \
287         } while (0)
288
289 #define BOUND_IFACE(r, k) \
290         ((r)->rule_flag & PFRULE_IFBOUND) ? (k) : pfi_all
291
292 #define STATE_INC_COUNTERS(s)                           \
293         do {                                            \
294                 s->rule.ptr->states_cur++;              \
295                 s->rule.ptr->states_tot++;              \
296                 if (s->anchor.ptr != NULL) {            \
297                         s->anchor.ptr->states_cur++;    \
298                         s->anchor.ptr->states_tot++;    \
299                 }                                       \
300                 if (s->nat_rule.ptr != NULL) {          \
301                         s->nat_rule.ptr->states_cur++;  \
302                         s->nat_rule.ptr->states_tot++;  \
303                 }                                       \
304         } while (0)
305
306 #define STATE_DEC_COUNTERS(s)                           \
307         do {                                            \
308                 if (s->nat_rule.ptr != NULL)            \
309                         s->nat_rule.ptr->states_cur--;  \
310                 if (s->anchor.ptr != NULL)              \
311                         s->anchor.ptr->states_cur--;    \
312                 s->rule.ptr->states_cur--;              \
313         } while (0)
314
315 static __inline int pf_src_compare(struct pf_src_node *, struct pf_src_node *);
316 static __inline int pf_state_compare_key(struct pf_state_key *,
317         struct pf_state_key *);
318 static __inline int pf_state_compare_id(struct pf_state *,
319         struct pf_state *);
320
321 struct pf_src_tree tree_src_tracking;
322
323 struct pf_state_tree_id tree_id;
324 struct pf_state_queue state_list;
325
326 RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare);
327 RB_GENERATE(pf_state_tree, pf_state_key, entry, pf_state_compare_key);
328 RB_GENERATE(pf_state_tree_id, pf_state,
329     entry_id, pf_state_compare_id);
330
331 static __inline int
332 pf_src_compare(struct pf_src_node *a, struct pf_src_node *b)
333 {
334         int     diff;
335
336         if (a->rule.ptr > b->rule.ptr)
337                 return (1);
338         if (a->rule.ptr < b->rule.ptr)
339                 return (-1);
340         if ((diff = a->af - b->af) != 0)
341                 return (diff);
342         switch (a->af) {
343 #ifdef INET
344         case AF_INET:
345                 if (a->addr.addr32[0] > b->addr.addr32[0])
346                         return (1);
347                 if (a->addr.addr32[0] < b->addr.addr32[0])
348                         return (-1);
349                 break;
350 #endif /* INET */
351 #ifdef INET6
352         case AF_INET6:
353                 if (a->addr.addr32[3] > b->addr.addr32[3])
354                         return (1);
355                 if (a->addr.addr32[3] < b->addr.addr32[3])
356                         return (-1);
357                 if (a->addr.addr32[2] > b->addr.addr32[2])
358                         return (1);
359                 if (a->addr.addr32[2] < b->addr.addr32[2])
360                         return (-1);
361                 if (a->addr.addr32[1] > b->addr.addr32[1])
362                         return (1);
363                 if (a->addr.addr32[1] < b->addr.addr32[1])
364                         return (-1);
365                 if (a->addr.addr32[0] > b->addr.addr32[0])
366                         return (1);
367                 if (a->addr.addr32[0] < b->addr.addr32[0])
368                         return (-1);
369                 break;
370 #endif /* INET6 */
371         }
372         return (0);
373 }
374
375 u_int32_t
376 pf_state_hash(struct pf_state_key *sk)
377 {
378         u_int32_t hv = (u_int32_t)(((intptr_t)sk >> 6) ^ ((intptr_t)sk >> 15));
379         if (hv == 0)    /* disallow 0 */
380                 hv = 1;
381         return(hv);
382 }
383
384 #ifdef INET6
385 void
386 pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af)
387 {
388         switch (af) {
389 #ifdef INET
390         case AF_INET:
391                 dst->addr32[0] = src->addr32[0];
392                 break;
393 #endif /* INET */
394         case AF_INET6:
395                 dst->addr32[0] = src->addr32[0];
396                 dst->addr32[1] = src->addr32[1];
397                 dst->addr32[2] = src->addr32[2];
398                 dst->addr32[3] = src->addr32[3];
399                 break;
400         }
401 }
402 #endif /* INET6 */
403
404 void
405 pf_init_threshold(struct pf_threshold *threshold,
406     u_int32_t limit, u_int32_t seconds)
407 {
408         threshold->limit = limit * PF_THRESHOLD_MULT;
409         threshold->seconds = seconds;
410         threshold->count = 0;
411         threshold->last = time_second;
412 }
413
414 void
415 pf_add_threshold(struct pf_threshold *threshold)
416 {
417         u_int32_t t = time_second, diff = t - threshold->last;
418
419         if (diff >= threshold->seconds)
420                 threshold->count = 0;
421         else
422                 threshold->count -= threshold->count * diff /
423                     threshold->seconds;
424         threshold->count += PF_THRESHOLD_MULT;
425         threshold->last = t;
426 }
427
428 int
429 pf_check_threshold(struct pf_threshold *threshold)
430 {
431         return (threshold->count > threshold->limit);
432 }
433
434 int
435 pf_src_connlimit(struct pf_state **state)
436 {
437         int bad = 0;
438
439         (*state)->src_node->conn++;
440         (*state)->src.tcp_est = 1;
441         pf_add_threshold(&(*state)->src_node->conn_rate);
442
443         if ((*state)->rule.ptr->max_src_conn &&
444             (*state)->rule.ptr->max_src_conn <
445             (*state)->src_node->conn) {
446                 pf_status.lcounters[LCNT_SRCCONN]++;
447                 bad++;
448         }
449
450         if ((*state)->rule.ptr->max_src_conn_rate.limit &&
451             pf_check_threshold(&(*state)->src_node->conn_rate)) {
452                 pf_status.lcounters[LCNT_SRCCONNRATE]++;
453                 bad++;
454         }
455
456         if (!bad)
457                 return (0);
458
459         if ((*state)->rule.ptr->overload_tbl) {
460                 struct pfr_addr p;
461                 u_int32_t       killed = 0;
462
463                 pf_status.lcounters[LCNT_OVERLOAD_TABLE]++;
464                 if (pf_status.debug >= PF_DEBUG_MISC) {
465                         kprintf("pf_src_connlimit: blocking address ");
466                         pf_print_host(&(*state)->src_node->addr, 0,
467                             (*state)->key[PF_SK_WIRE]->af);
468                 }
469
470                 bzero(&p, sizeof(p));
471                 p.pfra_af = (*state)->key[PF_SK_WIRE]->af;
472                 switch ((*state)->key[PF_SK_WIRE]->af) {
473 #ifdef INET
474                 case AF_INET:
475                         p.pfra_net = 32;
476                         p.pfra_ip4addr = (*state)->src_node->addr.v4;
477                         break;
478 #endif /* INET */
479 #ifdef INET6
480                 case AF_INET6:
481                         p.pfra_net = 128;
482                         p.pfra_ip6addr = (*state)->src_node->addr.v6;
483                         break;
484 #endif /* INET6 */
485                 }
486
487                 pfr_insert_kentry((*state)->rule.ptr->overload_tbl,
488                     &p, time_second);
489
490                 /* kill existing states if that's required. */
491                 if ((*state)->rule.ptr->flush) {
492                         struct pf_state_key *sk;
493                         struct pf_state *st;
494
495                         pf_status.lcounters[LCNT_OVERLOAD_FLUSH]++;
496                         RB_FOREACH(st, pf_state_tree_id, &tree_id) {
497                                 sk = st->key[PF_SK_WIRE];
498                                 /*
499                                  * Kill states from this source.  (Only those
500                                  * from the same rule if PF_FLUSH_GLOBAL is not
501                                  * set)
502                                  */
503                                 if (sk->af ==
504                                     (*state)->key[PF_SK_WIRE]->af &&
505                                     (((*state)->direction == PF_OUT &&
506                                     PF_AEQ(&(*state)->src_node->addr,
507                                         &sk->addr[0], sk->af)) ||
508                                     ((*state)->direction == PF_IN &&
509                                     PF_AEQ(&(*state)->src_node->addr,
510                                         &sk->addr[1], sk->af))) &&
511                                     ((*state)->rule.ptr->flush &
512                                     PF_FLUSH_GLOBAL ||
513                                     (*state)->rule.ptr == st->rule.ptr)) {
514                                         st->timeout = PFTM_PURGE;
515                                         st->src.state = st->dst.state =
516                                             TCPS_CLOSED;
517                                         killed++;
518                                 }
519                         }
520                         if (pf_status.debug >= PF_DEBUG_MISC)
521                                 kprintf(", %u states killed", killed);
522                 }
523                 if (pf_status.debug >= PF_DEBUG_MISC)
524                         kprintf("\n");
525         }
526
527         /* kill this state */
528         (*state)->timeout = PFTM_PURGE;
529         (*state)->src.state = (*state)->dst.state = TCPS_CLOSED;
530         return (1);
531 }
532
533 int
534 pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule,
535     struct pf_addr *src, sa_family_t af)
536 {
537         struct pf_src_node      k;
538
539         if (*sn == NULL) {
540                 k.af = af;
541                 PF_ACPY(&k.addr, src, af);
542                 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
543                     rule->rpool.opts & PF_POOL_STICKYADDR)
544                         k.rule.ptr = rule;
545                 else
546                         k.rule.ptr = NULL;
547                 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
548                 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
549         }
550         if (*sn == NULL) {
551                 if (!rule->max_src_nodes ||
552                     rule->src_nodes < rule->max_src_nodes)
553                         (*sn) = pool_get(&pf_src_tree_pl, PR_NOWAIT | PR_ZERO);
554                 else
555                         pf_status.lcounters[LCNT_SRCNODES]++;
556                 if ((*sn) == NULL)
557                         return (-1);
558
559                 pf_init_threshold(&(*sn)->conn_rate,
560                     rule->max_src_conn_rate.limit,
561                     rule->max_src_conn_rate.seconds);
562
563                 (*sn)->af = af;
564                 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
565                     rule->rpool.opts & PF_POOL_STICKYADDR)
566                         (*sn)->rule.ptr = rule;
567                 else
568                         (*sn)->rule.ptr = NULL;
569                 PF_ACPY(&(*sn)->addr, src, af);
570                 if (RB_INSERT(pf_src_tree,
571                     &tree_src_tracking, *sn) != NULL) {
572                         if (pf_status.debug >= PF_DEBUG_MISC) {
573                                 kprintf("pf: src_tree insert failed: ");
574                                 pf_print_host(&(*sn)->addr, 0, af);
575                                 kprintf("\n");
576                         }
577                         pool_put(&pf_src_tree_pl, *sn);
578                         return (-1);
579                 }
580                 (*sn)->creation = time_second;
581                 (*sn)->ruletype = rule->action;
582                 if ((*sn)->rule.ptr != NULL)
583                         (*sn)->rule.ptr->src_nodes++;
584                 pf_status.scounters[SCNT_SRC_NODE_INSERT]++;
585                 pf_status.src_nodes++;
586         } else {
587                 if (rule->max_src_states &&
588                     (*sn)->states >= rule->max_src_states) {
589                         pf_status.lcounters[LCNT_SRCSTATES]++;
590                         return (-1);
591                 }
592         }
593         return (0);
594 }
595
596 /* state table stuff */
597
598 static __inline int
599 pf_state_compare_key(struct pf_state_key *a, struct pf_state_key *b)
600 {
601         int     diff;
602
603         if ((diff = a->proto - b->proto) != 0)
604                 return (diff);
605         if ((diff = a->af - b->af) != 0)
606                 return (diff);
607         switch (a->af) {
608 #ifdef INET
609         case AF_INET:
610                 if (a->addr[0].addr32[0] > b->addr[0].addr32[0])
611                         return (1);
612                 if (a->addr[0].addr32[0] < b->addr[0].addr32[0])
613                         return (-1);
614                 if (a->addr[1].addr32[0] > b->addr[1].addr32[0])
615                         return (1);
616                 if (a->addr[1].addr32[0] < b->addr[1].addr32[0])
617                         return (-1);
618                 break;
619 #endif /* INET */
620 #ifdef INET6
621         case AF_INET6:
622                 if (a->addr[0].addr32[3] > b->addr[0].addr32[3])
623                         return (1);
624                 if (a->addr[0].addr32[3] < b->addr[0].addr32[3])
625                         return (-1);
626                 if (a->addr[1].addr32[3] > b->addr[1].addr32[3])
627                         return (1);
628                 if (a->addr[1].addr32[3] < b->addr[1].addr32[3])
629                         return (-1);
630                 if (a->addr[0].addr32[2] > b->addr[0].addr32[2])
631                         return (1);
632                 if (a->addr[0].addr32[2] < b->addr[0].addr32[2])
633                         return (-1);
634                 if (a->addr[1].addr32[2] > b->addr[1].addr32[2])
635                         return (1);
636                 if (a->addr[1].addr32[2] < b->addr[1].addr32[2])
637                         return (-1);
638                 if (a->addr[0].addr32[1] > b->addr[0].addr32[1])
639                         return (1);
640                 if (a->addr[0].addr32[1] < b->addr[0].addr32[1])
641                         return (-1);
642                 if (a->addr[1].addr32[1] > b->addr[1].addr32[1])
643                         return (1);
644                 if (a->addr[1].addr32[1] < b->addr[1].addr32[1])
645                         return (-1);
646                 if (a->addr[0].addr32[0] > b->addr[0].addr32[0])
647                         return (1);
648                 if (a->addr[0].addr32[0] < b->addr[0].addr32[0])
649                         return (-1);
650                 if (a->addr[1].addr32[0] > b->addr[1].addr32[0])
651                         return (1);
652                 if (a->addr[1].addr32[0] < b->addr[1].addr32[0])
653                         return (-1);
654                 break;
655 #endif /* INET6 */
656         }
657
658         if ((diff = a->port[0] - b->port[0]) != 0)
659                 return (diff);
660         if ((diff = a->port[1] - b->port[1]) != 0)
661                 return (diff);
662
663         return (0);
664 }
665
666 static __inline int
667 pf_state_compare_id(struct pf_state *a, struct pf_state *b)
668 {
669         if (a->id > b->id)
670                 return (1);
671         if (a->id < b->id)
672                 return (-1);
673         if (a->creatorid > b->creatorid)
674                 return (1);
675         if (a->creatorid < b->creatorid)
676                 return (-1);
677
678         return (0);
679 }
680
681 int
682 pf_state_key_attach(struct pf_state_key *sk, struct pf_state *s, int idx)
683 {
684         struct pf_state_item    *si;
685         struct pf_state_key     *cur;
686
687         KKASSERT(s->key[idx] == NULL);  /* XXX handle this? */
688
689         if ((cur = RB_INSERT(pf_state_tree, &pf_statetbl, sk)) != NULL) {
690                 /* key exists. check for same kif, if none, add to key */
691                 TAILQ_FOREACH(si, &cur->states, entry)
692                         if (si->s->kif == s->kif &&
693                             si->s->direction == s->direction) {
694                                 if (pf_status.debug >= PF_DEBUG_MISC) {
695                                         kprintf(
696                                             "pf: %s key attach failed on %s: ",
697                                             (idx == PF_SK_WIRE) ?
698                                             "wire" : "stack",
699                                             s->kif->pfik_name);
700                                         pf_print_state_parts(s,
701                                             (idx == PF_SK_WIRE) ? sk : NULL,
702                                             (idx == PF_SK_STACK) ? sk : NULL);
703                                         kprintf("\n");
704                                 }
705                                 pool_put(&pf_state_key_pl, sk);
706                                 return (-1);    /* collision! */
707                         }
708                 pool_put(&pf_state_key_pl, sk);
709                 s->key[idx] = cur;
710         } else
711                 s->key[idx] = sk;
712
713         if ((si = pool_get(&pf_state_item_pl, PR_NOWAIT)) == NULL) {
714                 pf_state_key_detach(s, idx);
715                 return (-1);
716         }
717         si->s = s;
718
719         /* list is sorted, if-bound states before floating */
720         if (s->kif == pfi_all)
721                 TAILQ_INSERT_TAIL(&s->key[idx]->states, si, entry);
722         else
723                 TAILQ_INSERT_HEAD(&s->key[idx]->states, si, entry);
724         return (0);
725 }
726
727 void
728 pf_detach_state(struct pf_state *s)
729 {
730         if (s->key[PF_SK_WIRE] == s->key[PF_SK_STACK])
731                 s->key[PF_SK_WIRE] = NULL;
732
733         if (s->key[PF_SK_STACK] != NULL)
734                 pf_state_key_detach(s, PF_SK_STACK);
735
736         if (s->key[PF_SK_WIRE] != NULL)
737                 pf_state_key_detach(s, PF_SK_WIRE);
738 }
739
740 void
741 pf_state_key_detach(struct pf_state *s, int idx)
742 {
743         struct pf_state_item    *si;
744
745         si = TAILQ_FIRST(&s->key[idx]->states);
746         while (si && si->s != s)
747             si = TAILQ_NEXT(si, entry);
748
749         if (si) {
750                 TAILQ_REMOVE(&s->key[idx]->states, si, entry);
751                 pool_put(&pf_state_item_pl, si);
752         }
753
754         if (TAILQ_EMPTY(&s->key[idx]->states)) {
755                 RB_REMOVE(pf_state_tree, &pf_statetbl, s->key[idx]);
756                 if (s->key[idx]->reverse)
757                         s->key[idx]->reverse->reverse = NULL;
758                 if (s->key[idx]->inp)
759                         s->key[idx]->inp->inp_pf_sk = NULL;
760                 pool_put(&pf_state_key_pl, s->key[idx]);
761         }
762         s->key[idx] = NULL;
763 }
764
765 struct pf_state_key *
766 pf_alloc_state_key(int pool_flags)
767 {
768         struct pf_state_key     *sk;
769
770         if ((sk = pool_get(&pf_state_key_pl, pool_flags)) == NULL)
771                 return (NULL);
772         TAILQ_INIT(&sk->states);
773
774         return (sk);
775 }
776
777 int
778 pf_state_key_setup(struct pf_pdesc *pd, struct pf_rule *nr,
779         struct pf_state_key **skw, struct pf_state_key **sks,
780         struct pf_state_key **skp, struct pf_state_key **nkp,
781         struct pf_addr *saddr, struct pf_addr *daddr,
782         u_int16_t sport, u_int16_t dport)
783 {
784         KKASSERT((*skp == NULL && *nkp == NULL));
785
786         if ((*skp = pf_alloc_state_key(PR_NOWAIT | PR_ZERO)) == NULL)
787                 return (ENOMEM);
788
789         PF_ACPY(&(*skp)->addr[pd->sidx], saddr, pd->af);
790         PF_ACPY(&(*skp)->addr[pd->didx], daddr, pd->af);
791         (*skp)->port[pd->sidx] = sport;
792         (*skp)->port[pd->didx] = dport;
793         (*skp)->proto = pd->proto;
794         (*skp)->af = pd->af;
795
796         if (nr != NULL) {
797                 if ((*nkp = pf_alloc_state_key(PR_NOWAIT | PR_ZERO)) == NULL)
798                         return (ENOMEM); /* caller must handle cleanup */
799
800                 /* XXX maybe just bcopy and TAILQ_INIT(&(*nkp)->states) */
801                 PF_ACPY(&(*nkp)->addr[0], &(*skp)->addr[0], pd->af);
802                 PF_ACPY(&(*nkp)->addr[1], &(*skp)->addr[1], pd->af);
803                 (*nkp)->port[0] = (*skp)->port[0];
804                 (*nkp)->port[1] = (*skp)->port[1];
805                 (*nkp)->proto = pd->proto;
806                 (*nkp)->af = pd->af;
807         } else
808                 *nkp = *skp;
809
810         if (pd->dir == PF_IN) {
811                 *skw = *skp;
812                 *sks = *nkp;
813         } else {
814                 *sks = *skp;
815                 *skw = *nkp;
816         }
817         return (0);
818 }
819
820
821 int
822 pf_state_insert(struct pfi_kif *kif, struct pf_state_key *skw,
823     struct pf_state_key *sks, struct pf_state *s)
824 {
825         s->kif = kif;
826
827         if (skw == sks) {
828                 if (pf_state_key_attach(skw, s, PF_SK_WIRE))
829                         return (-1);
830                 s->key[PF_SK_STACK] = s->key[PF_SK_WIRE];
831         } else {
832                 if (pf_state_key_attach(skw, s, PF_SK_WIRE)) {
833                         pool_put(&pf_state_key_pl, sks);
834                         return (-1);
835                 }
836                 if (pf_state_key_attach(sks, s, PF_SK_STACK)) {
837                         pf_state_key_detach(s, PF_SK_WIRE);
838                         return (-1);
839                 }
840         }
841
842         if (s->id == 0 && s->creatorid == 0) {
843                 s->id = htobe64(pf_status.stateid++);
844                 s->creatorid = pf_status.hostid;
845         }
846         if (RB_INSERT(pf_state_tree_id, &tree_id, s) != NULL) {
847                 if (pf_status.debug >= PF_DEBUG_MISC) {
848                         kprintf("pf: state insert failed: "
849                             "id: %016jx creatorid: %08x",
850                               (uintmax_t)be64toh(s->id), ntohl(s->creatorid));
851                         if (s->sync_flags & PFSTATE_FROMSYNC)
852                                 kprintf(" (from sync)");
853                         kprintf("\n");
854                 }
855                 pf_detach_state(s);
856                 return (-1);
857         }
858         TAILQ_INSERT_TAIL(&state_list, s, entry_list);
859         pf_status.fcounters[FCNT_STATE_INSERT]++;
860         pf_status.states++;
861         pfi_kif_ref(kif, PFI_KIF_REF_STATE);
862         pfsync_insert_state(s);
863         return (0);
864 }
865
866 struct pf_state *
867 pf_find_state_byid(struct pf_state_cmp *key)
868 {
869         pf_status.fcounters[FCNT_STATE_SEARCH]++;
870
871         return (RB_FIND(pf_state_tree_id, &tree_id, (struct pf_state *)key));
872 }
873
874 struct pf_state *
875 pf_find_state(struct pfi_kif *kif, struct pf_state_key_cmp *key, u_int dir,
876     struct mbuf *m)
877 {
878         struct pf_state_key     *sk;
879         struct pf_state_item    *si;
880
881         pf_status.fcounters[FCNT_STATE_SEARCH]++;
882
883         if (dir == PF_OUT && m->m_pkthdr.pf.statekey &&
884             ((struct pf_state_key *)m->m_pkthdr.pf.statekey)->reverse)
885                 sk = ((struct pf_state_key *)m->m_pkthdr.pf.statekey)->reverse;
886         else {
887                 if ((sk = RB_FIND(pf_state_tree, &pf_statetbl,
888                     (struct pf_state_key *)key)) == NULL)
889                         return (NULL);
890                 if (dir == PF_OUT && m->m_pkthdr.pf.statekey) {
891                         ((struct pf_state_key *)
892                             m->m_pkthdr.pf.statekey)->reverse = sk;
893                         sk->reverse = m->m_pkthdr.pf.statekey;
894                 }
895         }
896
897         if (dir == PF_OUT)
898                 m->m_pkthdr.pf.statekey = NULL;
899
900         /* list is sorted, if-bound states before floating ones */
901         TAILQ_FOREACH(si, &sk->states, entry)
902                 if ((si->s->kif == pfi_all || si->s->kif == kif) &&
903                     sk == (dir == PF_IN ? si->s->key[PF_SK_WIRE] :
904                     si->s->key[PF_SK_STACK]))
905                         return (si->s);
906
907         return (NULL);
908 }
909
910 struct pf_state *
911 pf_find_state_all(struct pf_state_key_cmp *key, u_int dir, int *more)
912 {
913         struct pf_state_key     *sk;
914         struct pf_state_item    *si, *ret = NULL;
915
916         pf_status.fcounters[FCNT_STATE_SEARCH]++;
917
918         sk = RB_FIND(pf_state_tree, &pf_statetbl, (struct pf_state_key *)key);
919
920         if (sk != NULL) {
921                 TAILQ_FOREACH(si, &sk->states, entry)
922                         if (dir == PF_INOUT ||
923                             (sk == (dir == PF_IN ? si->s->key[PF_SK_WIRE] :
924                             si->s->key[PF_SK_STACK]))) {
925                                 if (more == NULL)
926                                         return (si->s);
927
928                                 if (ret)
929                                         (*more)++;
930                                 else
931                                         ret = si;
932                         }
933         }
934         return (ret ? ret->s : NULL);
935 }
936
937 /* END state table stuff */
938
939
940 void
941 pf_purge_thread(void *v)
942 {
943         int nloops = 0;
944         int locked = 0;
945
946         lwkt_gettoken(&pf_token);
947         for (;;) {
948                 tsleep(pf_purge_thread, PWAIT, "pftm", 1 * hz);
949
950                 lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
951
952                 if (pf_end_threads) {
953                         pf_purge_expired_states(pf_status.states, 1);
954                         pf_purge_expired_fragments();
955                         pf_purge_expired_src_nodes(1);
956                         pf_end_threads++;
957
958                         lockmgr(&pf_consistency_lock, LK_RELEASE);
959                         wakeup(pf_purge_thread);
960                         kthread_exit();
961                 }
962                 crit_enter();
963
964                 /* process a fraction of the state table every second */
965                 if(!pf_purge_expired_states(1 + (pf_status.states
966                     / pf_default_rule.timeout[PFTM_INTERVAL]), 0)) {
967
968                         pf_purge_expired_states(1 + (pf_status.states
969                             / pf_default_rule.timeout[PFTM_INTERVAL]), 1);
970                 }
971
972                 /* purge other expired types every PFTM_INTERVAL seconds */
973                 if (++nloops >= pf_default_rule.timeout[PFTM_INTERVAL]) {
974                         pf_purge_expired_fragments();
975                         if (!pf_purge_expired_src_nodes(locked)) {
976                                 pf_purge_expired_src_nodes(1);
977                         }
978                         nloops = 0;
979                 }
980                 crit_exit();
981                 lockmgr(&pf_consistency_lock, LK_RELEASE);
982         }
983         lwkt_reltoken(&pf_token);
984 }
985
986 u_int32_t
987 pf_state_expires(const struct pf_state *state)
988 {
989         u_int32_t       timeout;
990         u_int32_t       start;
991         u_int32_t       end;
992         u_int32_t       states;
993
994         /* handle all PFTM_* > PFTM_MAX here */
995         if (state->timeout == PFTM_PURGE)
996                 return (time_second);
997         if (state->timeout == PFTM_UNTIL_PACKET)
998                 return (0);
999         KKASSERT(state->timeout != PFTM_UNLINKED);
1000         KKASSERT(state->timeout < PFTM_MAX);
1001         timeout = state->rule.ptr->timeout[state->timeout];
1002         if (!timeout)
1003                 timeout = pf_default_rule.timeout[state->timeout];
1004         start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START];
1005         if (start) {
1006                 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END];
1007                 states = state->rule.ptr->states_cur;
1008         } else {
1009                 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START];
1010                 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END];
1011                 states = pf_status.states;
1012         }
1013         if (end && states > start && start < end) {
1014                 if (states < end)
1015                         return (state->expire + timeout * (end - states) /
1016                             (end - start));
1017                 else
1018                         return (time_second);
1019         }
1020         return (state->expire + timeout);
1021 }
1022
1023 int
1024 pf_purge_expired_src_nodes(int waslocked)
1025 {
1026          struct pf_src_node             *cur, *next;
1027          int                             locked = waslocked;
1028
1029          for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) {
1030                  next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur);
1031
1032                  if (cur->states <= 0 && cur->expire <= time_second) {
1033                          if (! locked) {
1034                                  lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
1035                                  next = RB_NEXT(pf_src_tree,
1036                                      &tree_src_tracking, cur);
1037                                  locked = 1;
1038                          }
1039                          if (cur->rule.ptr != NULL) {
1040                                  cur->rule.ptr->src_nodes--;
1041                                  if (cur->rule.ptr->states_cur <= 0 &&
1042                                      cur->rule.ptr->max_src_nodes <= 0)
1043                                          pf_rm_rule(NULL, cur->rule.ptr);
1044                          }
1045                          RB_REMOVE(pf_src_tree, &tree_src_tracking, cur);
1046                          pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
1047                          pf_status.src_nodes--;
1048                          pool_put(&pf_src_tree_pl, cur);
1049                  }
1050          }
1051
1052          if (locked && !waslocked)
1053                 lockmgr(&pf_consistency_lock, LK_RELEASE);
1054         return(1);
1055 }
1056
1057 void
1058 pf_src_tree_remove_state(struct pf_state *s)
1059 {
1060         u_int32_t timeout;
1061
1062         if (s->src_node != NULL) {
1063                 if (s->src.tcp_est)
1064                         --s->src_node->conn;
1065                 if (--s->src_node->states <= 0) {
1066                         timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1067                         if (!timeout)
1068                                 timeout =
1069                                     pf_default_rule.timeout[PFTM_SRC_NODE];
1070                         s->src_node->expire = time_second + timeout;
1071                 }
1072         }
1073         if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) {
1074                 if (--s->nat_src_node->states <= 0) {
1075                         timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1076                         if (!timeout)
1077                                 timeout =
1078                                     pf_default_rule.timeout[PFTM_SRC_NODE];
1079                         s->nat_src_node->expire = time_second + timeout;
1080                 }
1081         }
1082         s->src_node = s->nat_src_node = NULL;
1083 }
1084
1085 /* callers should be at crit_enter() */
1086 void
1087 pf_unlink_state(struct pf_state *cur)
1088 {
1089         if (cur->src.state == PF_TCPS_PROXY_DST) {
1090                 /* XXX wire key the right one? */
1091                 pf_send_tcp(cur->rule.ptr, cur->key[PF_SK_WIRE]->af,
1092                     &cur->key[PF_SK_WIRE]->addr[1],
1093                     &cur->key[PF_SK_WIRE]->addr[0],
1094                     cur->key[PF_SK_WIRE]->port[1],
1095                     cur->key[PF_SK_WIRE]->port[0],
1096                     cur->src.seqhi, cur->src.seqlo + 1,
1097                     TH_RST|TH_ACK, 0, 0, 0, 1, cur->tag, NULL, NULL);
1098         }
1099         RB_REMOVE(pf_state_tree_id, &tree_id, cur);
1100         if (cur->creatorid == pf_status.hostid)
1101                 pfsync_delete_state(cur);
1102         cur->timeout = PFTM_UNLINKED;
1103         pf_src_tree_remove_state(cur);
1104         pf_detach_state(cur);
1105 }
1106
1107 static struct pf_state  *purge_cur;
1108
1109 /* callers should be at crit_enter() and hold the
1110  * write_lock on pf_consistency_lock */
1111 void
1112 pf_free_state(struct pf_state *cur)
1113 {
1114         if (pfsyncif != NULL &&
1115             (pfsyncif->sc_bulk_send_next == cur ||
1116             pfsyncif->sc_bulk_terminator == cur))
1117                 return;
1118         KKASSERT(cur->timeout == PFTM_UNLINKED);
1119         if (--cur->rule.ptr->states_cur <= 0 &&
1120             cur->rule.ptr->src_nodes <= 0)
1121                 pf_rm_rule(NULL, cur->rule.ptr);
1122         if (cur->nat_rule.ptr != NULL)
1123                 if (--cur->nat_rule.ptr->states_cur <= 0 &&
1124                         cur->nat_rule.ptr->src_nodes <= 0)
1125                         pf_rm_rule(NULL, cur->nat_rule.ptr);
1126         if (cur->anchor.ptr != NULL)
1127                 if (--cur->anchor.ptr->states_cur <= 0)
1128                         pf_rm_rule(NULL, cur->anchor.ptr);
1129         pf_normalize_tcp_cleanup(cur);
1130         pfi_kif_unref(cur->kif, PFI_KIF_REF_STATE);
1131
1132         /*
1133          * We may be freeing pf_purge_expired_states()'s saved scan entry,
1134          * adjust it if necessary.
1135          */
1136         if (purge_cur == cur) {
1137                 kprintf("PURGE CONFLICT\n");
1138                 purge_cur = TAILQ_NEXT(purge_cur, entry_list);
1139         }
1140         TAILQ_REMOVE(&state_list, cur, entry_list);
1141         if (cur->tag)
1142                 pf_tag_unref(cur->tag);
1143         pool_put(&pf_state_pl, cur);
1144         pf_status.fcounters[FCNT_STATE_REMOVALS]++;
1145         pf_status.states--;
1146 }
1147
1148 int
1149 pf_purge_expired_states(u_int32_t maxcheck, int waslocked)
1150 {
1151         struct pf_state         *cur;
1152         int                      locked = waslocked;
1153
1154         while (maxcheck--) {
1155                 /*
1156                  * Wrap to start of list when we hit the end
1157                  */
1158                 cur = purge_cur;
1159                 if (cur == NULL) {
1160                         cur = TAILQ_FIRST(&state_list);
1161                         if (cur == NULL)
1162                                 break;  /* list empty */
1163                 }
1164
1165                 /*
1166                  * Setup next (purge_cur) while we process this one.  If we block and
1167                  * something else deletes purge_cur, pf_free_state() will adjust it further
1168                  * ahead.
1169                  */
1170                 purge_cur = TAILQ_NEXT(cur, entry_list);
1171
1172                 if (cur->timeout == PFTM_UNLINKED) {
1173                         /* free unlinked state */
1174                         if (! locked) {
1175                                 lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
1176                                 locked = 1;
1177                         }
1178                         pf_free_state(cur);
1179                 } else if (pf_state_expires(cur) <= time_second) {
1180                         /* unlink and free expired state */
1181                         pf_unlink_state(cur);
1182                         if (! locked) {
1183                                 if (!lockmgr(&pf_consistency_lock, LK_EXCLUSIVE))
1184                                         return (0);
1185                                 locked = 1;
1186                         }
1187                         pf_free_state(cur);
1188                 }
1189         }
1190
1191         if (locked)
1192                 lockmgr(&pf_consistency_lock, LK_RELEASE);
1193         return (1);
1194 }
1195
1196 int
1197 pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw)
1198 {
1199         if (aw->type != PF_ADDR_TABLE)
1200                 return (0);
1201         if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL)
1202                 return (1);
1203         return (0);
1204 }
1205
1206 void
1207 pf_tbladdr_remove(struct pf_addr_wrap *aw)
1208 {
1209         if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL)
1210                 return;
1211         pfr_detach_table(aw->p.tbl);
1212         aw->p.tbl = NULL;
1213 }
1214
1215 void
1216 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
1217 {
1218         struct pfr_ktable *kt = aw->p.tbl;
1219
1220         if (aw->type != PF_ADDR_TABLE || kt == NULL)
1221                 return;
1222         if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
1223                 kt = kt->pfrkt_root;
1224         aw->p.tbl = NULL;
1225         aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
1226                 kt->pfrkt_cnt : -1;
1227 }
1228
1229 void
1230 pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af)
1231 {
1232         switch (af) {
1233 #ifdef INET
1234         case AF_INET: {
1235                 u_int32_t a = ntohl(addr->addr32[0]);
1236                 kprintf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255,
1237                     (a>>8)&255, a&255);
1238                 if (p) {
1239                         p = ntohs(p);
1240                         kprintf(":%u", p);
1241                 }
1242                 break;
1243         }
1244 #endif /* INET */
1245 #ifdef INET6
1246         case AF_INET6: {
1247                 u_int16_t b;
1248                 u_int8_t i, curstart = 255, curend = 0,
1249                     maxstart = 0, maxend = 0;
1250                 for (i = 0; i < 8; i++) {
1251                         if (!addr->addr16[i]) {
1252                                 if (curstart == 255)
1253                                         curstart = i;
1254                                 else
1255                                         curend = i;
1256                         } else {
1257                                 if (curstart) {
1258                                         if ((curend - curstart) >
1259                                             (maxend - maxstart)) {
1260                                                 maxstart = curstart;
1261                                                 maxend = curend;
1262                                                 curstart = 255;
1263                                         }
1264                                 }
1265                         }
1266                 }
1267                 for (i = 0; i < 8; i++) {
1268                         if (i >= maxstart && i <= maxend) {
1269                                 if (maxend != 7) {
1270                                         if (i == maxstart)
1271                                                 kprintf(":");
1272                                 } else {
1273                                         if (i == maxend)
1274                                                 kprintf(":");
1275                                 }
1276                         } else {
1277                                 b = ntohs(addr->addr16[i]);
1278                                 kprintf("%x", b);
1279                                 if (i < 7)
1280                                         kprintf(":");
1281                         }
1282                 }
1283                 if (p) {
1284                         p = ntohs(p);
1285                         kprintf("[%u]", p);
1286                 }
1287                 break;
1288         }
1289 #endif /* INET6 */
1290         }
1291 }
1292
1293 void
1294 pf_print_state(struct pf_state *s)
1295 {
1296         pf_print_state_parts(s, NULL, NULL);
1297 }
1298
1299 void
1300 pf_print_state_parts(struct pf_state *s,
1301     struct pf_state_key *skwp, struct pf_state_key *sksp)
1302 {
1303         struct pf_state_key *skw, *sks;
1304         u_int8_t proto, dir;
1305
1306         /* Do our best to fill these, but they're skipped if NULL */
1307         skw = skwp ? skwp : (s ? s->key[PF_SK_WIRE] : NULL);
1308         sks = sksp ? sksp : (s ? s->key[PF_SK_STACK] : NULL);
1309         proto = skw ? skw->proto : (sks ? sks->proto : 0);
1310         dir = s ? s->direction : 0;
1311
1312         switch (proto) {
1313         case IPPROTO_TCP:
1314                 kprintf("TCP ");
1315                 break;
1316         case IPPROTO_UDP:
1317                 kprintf("UDP ");
1318                 break;
1319         case IPPROTO_ICMP:
1320                 kprintf("ICMP ");
1321                 break;
1322         case IPPROTO_ICMPV6:
1323                 kprintf("ICMPV6 ");
1324                 break;
1325         default:
1326                 kprintf("%u ", skw->proto);
1327                 break;
1328         }
1329         switch (dir) {
1330         case PF_IN:
1331                 kprintf(" in");
1332                 break;
1333         case PF_OUT:
1334                 kprintf(" out");
1335                 break;
1336         }
1337         if (skw) {
1338                 kprintf(" wire: ");
1339                 pf_print_host(&skw->addr[0], skw->port[0], skw->af);
1340                 kprintf(" ");
1341                 pf_print_host(&skw->addr[1], skw->port[1], skw->af);
1342         }
1343         if (sks) {
1344                 kprintf(" stack: ");
1345                 if (sks != skw) {
1346                         pf_print_host(&sks->addr[0], sks->port[0], sks->af);
1347                         kprintf(" ");
1348                         pf_print_host(&sks->addr[1], sks->port[1], sks->af);
1349                 } else
1350                         kprintf("-");
1351         }
1352         if (s) {
1353                 if (proto == IPPROTO_TCP) {
1354                         kprintf(" [lo=%u high=%u win=%u modulator=%u",
1355                             s->src.seqlo, s->src.seqhi,
1356                             s->src.max_win, s->src.seqdiff);
1357                         if (s->src.wscale && s->dst.wscale)
1358                                 kprintf(" wscale=%u",
1359                                     s->src.wscale & PF_WSCALE_MASK);
1360                         kprintf("]");
1361                         kprintf(" [lo=%u high=%u win=%u modulator=%u",
1362                             s->dst.seqlo, s->dst.seqhi,
1363                             s->dst.max_win, s->dst.seqdiff);
1364                         if (s->src.wscale && s->dst.wscale)
1365                                 kprintf(" wscale=%u",
1366                                 s->dst.wscale & PF_WSCALE_MASK);
1367                         kprintf("]");
1368                 }
1369                 kprintf(" %u:%u", s->src.state, s->dst.state);
1370         }
1371 }
1372
1373 void
1374 pf_print_flags(u_int8_t f)
1375 {
1376         if (f)
1377                 kprintf(" ");
1378         if (f & TH_FIN)
1379                 kprintf("F");
1380         if (f & TH_SYN)
1381                 kprintf("S");
1382         if (f & TH_RST)
1383                 kprintf("R");
1384         if (f & TH_PUSH)
1385                 kprintf("P");
1386         if (f & TH_ACK)
1387                 kprintf("A");
1388         if (f & TH_URG)
1389                 kprintf("U");
1390         if (f & TH_ECE)
1391                 kprintf("E");
1392         if (f & TH_CWR)
1393                 kprintf("W");
1394 }
1395
1396 #define PF_SET_SKIP_STEPS(i)                                    \
1397         do {                                                    \
1398                 while (head[i] != cur) {                        \
1399                         head[i]->skip[i].ptr = cur;             \
1400                         head[i] = TAILQ_NEXT(head[i], entries); \
1401                 }                                               \
1402         } while (0)
1403
1404 void
1405 pf_calc_skip_steps(struct pf_rulequeue *rules)
1406 {
1407         struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT];
1408         int i;
1409
1410         cur = TAILQ_FIRST(rules);
1411         prev = cur;
1412         for (i = 0; i < PF_SKIP_COUNT; ++i)
1413                 head[i] = cur;
1414         while (cur != NULL) {
1415
1416                 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
1417                         PF_SET_SKIP_STEPS(PF_SKIP_IFP);
1418                 if (cur->direction != prev->direction)
1419                         PF_SET_SKIP_STEPS(PF_SKIP_DIR);
1420                 if (cur->af != prev->af)
1421                         PF_SET_SKIP_STEPS(PF_SKIP_AF);
1422                 if (cur->proto != prev->proto)
1423                         PF_SET_SKIP_STEPS(PF_SKIP_PROTO);
1424                 if (cur->src.neg != prev->src.neg ||
1425                     pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr))
1426                         PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR);
1427                 if (cur->src.port[0] != prev->src.port[0] ||
1428                     cur->src.port[1] != prev->src.port[1] ||
1429                     cur->src.port_op != prev->src.port_op)
1430                         PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
1431                 if (cur->dst.neg != prev->dst.neg ||
1432                     pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr))
1433                         PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR);
1434                 if (cur->dst.port[0] != prev->dst.port[0] ||
1435                     cur->dst.port[1] != prev->dst.port[1] ||
1436                     cur->dst.port_op != prev->dst.port_op)
1437                         PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
1438
1439                 prev = cur;
1440                 cur = TAILQ_NEXT(cur, entries);
1441         }
1442         for (i = 0; i < PF_SKIP_COUNT; ++i)
1443                 PF_SET_SKIP_STEPS(i);
1444 }
1445
1446 int
1447 pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2)
1448 {
1449         if (aw1->type != aw2->type)
1450                 return (1);
1451         switch (aw1->type) {
1452         case PF_ADDR_ADDRMASK:
1453         case PF_ADDR_RANGE:
1454                 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0))
1455                         return (1);
1456                 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0))
1457                         return (1);
1458                 return (0);
1459         case PF_ADDR_DYNIFTL:
1460                 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt);
1461         case PF_ADDR_NOROUTE:
1462         case PF_ADDR_URPFFAILED:
1463                 return (0);
1464         case PF_ADDR_TABLE:
1465                 return (aw1->p.tbl != aw2->p.tbl);
1466         case PF_ADDR_RTLABEL:
1467                 return (aw1->v.rtlabel != aw2->v.rtlabel);
1468         default:
1469                 kprintf("invalid address type: %d\n", aw1->type);
1470                 return (1);
1471         }
1472 }
1473
1474 u_int16_t
1475 pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp)
1476 {
1477         u_int32_t       l;
1478
1479         if (udp && !cksum)
1480                 return (0x0000);
1481         l = cksum + old - new;
1482         l = (l >> 16) + (l & 65535);
1483         l = l & 65535;
1484         if (udp && !l)
1485                 return (0xFFFF);
1486         return (l);
1487 }
1488
1489 void
1490 pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc,
1491     struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af)
1492 {
1493         struct pf_addr  ao;
1494         u_int16_t       po = *p;
1495
1496         PF_ACPY(&ao, a, af);
1497         PF_ACPY(a, an, af);
1498
1499         *p = pn;
1500
1501         switch (af) {
1502 #ifdef INET
1503         case AF_INET:
1504                 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1505                     ao.addr16[0], an->addr16[0], 0),
1506                     ao.addr16[1], an->addr16[1], 0);
1507                 *p = pn;
1508                 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1509                     ao.addr16[0], an->addr16[0], u),
1510                     ao.addr16[1], an->addr16[1], u),
1511                     po, pn, u);
1512                 break;
1513 #endif /* INET */
1514 #ifdef INET6
1515         case AF_INET6:
1516                 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1517                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1518                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1519                     ao.addr16[0], an->addr16[0], u),
1520                     ao.addr16[1], an->addr16[1], u),
1521                     ao.addr16[2], an->addr16[2], u),
1522                     ao.addr16[3], an->addr16[3], u),
1523                     ao.addr16[4], an->addr16[4], u),
1524                     ao.addr16[5], an->addr16[5], u),
1525                     ao.addr16[6], an->addr16[6], u),
1526                     ao.addr16[7], an->addr16[7], u),
1527                     po, pn, u);
1528                 break;
1529 #endif /* INET6 */
1530         }
1531 }
1532
1533
1534 /* Changes a u_int32_t.  Uses a void * so there are no align restrictions */
1535 void
1536 pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u)
1537 {
1538         u_int32_t       ao;
1539
1540         memcpy(&ao, a, sizeof(ao));
1541         memcpy(a, &an, sizeof(u_int32_t));
1542         *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u),
1543             ao % 65536, an % 65536, u);
1544 }
1545
1546 #ifdef INET6
1547 void
1548 pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u)
1549 {
1550         struct pf_addr  ao;
1551
1552         PF_ACPY(&ao, a, AF_INET6);
1553         PF_ACPY(a, an, AF_INET6);
1554
1555         *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1556             pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1557             pf_cksum_fixup(pf_cksum_fixup(*c,
1558             ao.addr16[0], an->addr16[0], u),
1559             ao.addr16[1], an->addr16[1], u),
1560             ao.addr16[2], an->addr16[2], u),
1561             ao.addr16[3], an->addr16[3], u),
1562             ao.addr16[4], an->addr16[4], u),
1563             ao.addr16[5], an->addr16[5], u),
1564             ao.addr16[6], an->addr16[6], u),
1565             ao.addr16[7], an->addr16[7], u);
1566 }
1567 #endif /* INET6 */
1568
1569 void
1570 pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa,
1571     struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c,
1572     u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af)
1573 {
1574         struct pf_addr  oia, ooa;
1575
1576         PF_ACPY(&oia, ia, af);
1577         if (oa)
1578                 PF_ACPY(&ooa, oa, af);
1579
1580         /* Change inner protocol port, fix inner protocol checksum. */
1581         if (ip != NULL) {
1582                 u_int16_t       oip = *ip;
1583                 u_int32_t       opc = 0;
1584
1585                 if (pc != NULL)
1586                         opc = *pc;
1587                 *ip = np;
1588                 if (pc != NULL)
1589                         *pc = pf_cksum_fixup(*pc, oip, *ip, u);
1590                 *ic = pf_cksum_fixup(*ic, oip, *ip, 0);
1591                 if (pc != NULL)
1592                         *ic = pf_cksum_fixup(*ic, opc, *pc, 0);
1593         }
1594         /* Change inner ip address, fix inner ip and icmp checksums. */
1595         PF_ACPY(ia, na, af);
1596         switch (af) {
1597 #ifdef INET
1598         case AF_INET: {
1599                 u_int32_t        oh2c = *h2c;
1600
1601                 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c,
1602                     oia.addr16[0], ia->addr16[0], 0),
1603                     oia.addr16[1], ia->addr16[1], 0);
1604                 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1605                     oia.addr16[0], ia->addr16[0], 0),
1606                     oia.addr16[1], ia->addr16[1], 0);
1607                 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0);
1608                 break;
1609         }
1610 #endif /* INET */
1611 #ifdef INET6
1612         case AF_INET6:
1613                 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1614                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1615                     pf_cksum_fixup(pf_cksum_fixup(*ic,
1616                     oia.addr16[0], ia->addr16[0], u),
1617                     oia.addr16[1], ia->addr16[1], u),
1618                     oia.addr16[2], ia->addr16[2], u),
1619                     oia.addr16[3], ia->addr16[3], u),
1620                     oia.addr16[4], ia->addr16[4], u),
1621                     oia.addr16[5], ia->addr16[5], u),
1622                     oia.addr16[6], ia->addr16[6], u),
1623                     oia.addr16[7], ia->addr16[7], u);
1624                 break;
1625 #endif /* INET6 */
1626         }
1627         /* Outer ip address, fix outer ip or icmpv6 checksum, if necessary. */
1628         if (oa) {
1629                 PF_ACPY(oa, na, af);
1630                 switch (af) {
1631 #ifdef INET
1632                 case AF_INET:
1633                         *hc = pf_cksum_fixup(pf_cksum_fixup(*hc,
1634                             ooa.addr16[0], oa->addr16[0], 0),
1635                             ooa.addr16[1], oa->addr16[1], 0);
1636                         break;
1637 #endif /* INET */
1638 #ifdef INET6
1639                 case AF_INET6:
1640                         *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1641                             pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1642                             pf_cksum_fixup(pf_cksum_fixup(*ic,
1643                             ooa.addr16[0], oa->addr16[0], u),
1644                             ooa.addr16[1], oa->addr16[1], u),
1645                             ooa.addr16[2], oa->addr16[2], u),
1646                             ooa.addr16[3], oa->addr16[3], u),
1647                             ooa.addr16[4], oa->addr16[4], u),
1648                             ooa.addr16[5], oa->addr16[5], u),
1649                             ooa.addr16[6], oa->addr16[6], u),
1650                             ooa.addr16[7], oa->addr16[7], u);
1651                         break;
1652 #endif /* INET6 */
1653                 }
1654         }
1655 }
1656
1657
1658 /*
1659  * Need to modulate the sequence numbers in the TCP SACK option
1660  * (credits to Krzysztof Pfaff for report and patch)
1661  */
1662 int
1663 pf_modulate_sack(struct mbuf *m, int off, struct pf_pdesc *pd,
1664     struct tcphdr *th, struct pf_state_peer *dst)
1665 {
1666         int hlen = (th->th_off << 2) - sizeof(*th), thoptlen = hlen;
1667         u_int8_t opts[TCP_MAXOLEN], *opt = opts;
1668         int copyback = 0, i, olen;
1669         struct raw_sackblock sack;
1670
1671 #define TCPOLEN_SACKLEN (TCPOLEN_SACK + 2)
1672         if (hlen < TCPOLEN_SACKLEN ||
1673             !pf_pull_hdr(m, off + sizeof(*th), opts, hlen, NULL, NULL, pd->af))
1674                 return 0;
1675
1676         while (hlen >= TCPOLEN_SACKLEN) {
1677                 olen = opt[1];
1678                 switch (*opt) {
1679                 case TCPOPT_EOL:        /* FALLTHROUGH */
1680                 case TCPOPT_NOP:
1681                         opt++;
1682                         hlen--;
1683                         break;
1684                 case TCPOPT_SACK:
1685                         if (olen > hlen)
1686                                 olen = hlen;
1687                         if (olen >= TCPOLEN_SACKLEN) {
1688                                 for (i = 2; i + TCPOLEN_SACK <= olen;
1689                                     i += TCPOLEN_SACK) {
1690                                         memcpy(&sack, &opt[i], sizeof(sack));
1691                                         pf_change_a(&sack.rblk_start, &th->th_sum,
1692                                             htonl(ntohl(sack.rblk_start) -
1693                                             dst->seqdiff), 0);
1694                                         pf_change_a(&sack.rblk_end, &th->th_sum,
1695                                             htonl(ntohl(sack.rblk_end) -
1696                                             dst->seqdiff), 0);
1697                                         memcpy(&opt[i], &sack, sizeof(sack));
1698                                 }
1699                                 copyback = 1;
1700                         }
1701                         /* FALLTHROUGH */
1702                 default:
1703                         if (olen < 2)
1704                                 olen = 2;
1705                         hlen -= olen;
1706                         opt += olen;
1707                 }
1708         }
1709
1710         if (copyback)
1711                 m_copyback(m, off + sizeof(*th), thoptlen, opts);
1712         return (copyback);
1713 }
1714
1715 void
1716 pf_send_tcp(const struct pf_rule *r, sa_family_t af,
1717     const struct pf_addr *saddr, const struct pf_addr *daddr,
1718     u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack,
1719     u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl, int tag,
1720     u_int16_t rtag, struct ether_header *eh, struct ifnet *ifp)
1721 {
1722         struct mbuf     *m;
1723         int              len = 0, tlen;
1724 #ifdef INET
1725         struct ip       *h = NULL;
1726 #endif /* INET */
1727 #ifdef INET6
1728         struct ip6_hdr  *h6 = NULL;
1729 #endif /* INET6 */
1730         struct tcphdr   *th = NULL;
1731         char            *opt;
1732
1733         ASSERT_LWKT_TOKEN_HELD(&pf_token);
1734
1735         /* maximum segment size tcp option */
1736         tlen = sizeof(struct tcphdr);
1737         if (mss)
1738                 tlen += 4;
1739
1740         switch (af) {
1741 #ifdef INET
1742         case AF_INET:
1743                 len = sizeof(struct ip) + tlen;
1744                 break;
1745 #endif /* INET */
1746 #ifdef INET6
1747         case AF_INET6:
1748                 len = sizeof(struct ip6_hdr) + tlen;
1749                 break;
1750 #endif /* INET6 */
1751         }
1752
1753         /*
1754          * Create outgoing mbuf.
1755          *
1756          * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags,
1757          * so make sure pf.flags is clear.
1758          */
1759         m = m_gethdr(MB_DONTWAIT, MT_HEADER);
1760         if (m == NULL) {
1761                 return;
1762         }
1763         if (tag)
1764                 m->m_pkthdr.fw_flags |= PF_MBUF_TAGGED;
1765         m->m_pkthdr.pf.flags = 0;
1766         m->m_pkthdr.pf.tag = rtag;
1767         /* XXX Recheck when upgrading to > 4.4 */
1768         m->m_pkthdr.pf.statekey = NULL;
1769         if (r != NULL && r->rtableid >= 0)
1770                 m->m_pkthdr.pf.rtableid = r->rtableid;
1771
1772 #ifdef ALTQ
1773         if (r != NULL && r->qid) {
1774                 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
1775                 m->m_pkthdr.pf.qid = r->qid;
1776                 m->m_pkthdr.pf.ecn_af = af;
1777                 m->m_pkthdr.pf.hdr = mtod(m, struct ip *);
1778         }
1779 #endif /* ALTQ */
1780         m->m_data += max_linkhdr;
1781         m->m_pkthdr.len = m->m_len = len;
1782         m->m_pkthdr.rcvif = NULL;
1783         bzero(m->m_data, len);
1784         switch (af) {
1785 #ifdef INET
1786         case AF_INET:
1787                 h = mtod(m, struct ip *);
1788
1789                 /* IP header fields included in the TCP checksum */
1790                 h->ip_p = IPPROTO_TCP;
1791                 h->ip_len = tlen;
1792                 h->ip_src.s_addr = saddr->v4.s_addr;
1793                 h->ip_dst.s_addr = daddr->v4.s_addr;
1794
1795                 th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip));
1796                 break;
1797 #endif /* INET */
1798 #ifdef INET6
1799         case AF_INET6:
1800                 h6 = mtod(m, struct ip6_hdr *);
1801
1802                 /* IP header fields included in the TCP checksum */
1803                 h6->ip6_nxt = IPPROTO_TCP;
1804                 h6->ip6_plen = htons(tlen);
1805                 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr));
1806                 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr));
1807
1808                 th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr));
1809                 break;
1810 #endif /* INET6 */
1811         }
1812
1813         /* TCP header */
1814         th->th_sport = sport;
1815         th->th_dport = dport;
1816         th->th_seq = htonl(seq);
1817         th->th_ack = htonl(ack);
1818         th->th_off = tlen >> 2;
1819         th->th_flags = flags;
1820         th->th_win = htons(win);
1821
1822         if (mss) {
1823                 opt = (char *)(th + 1);
1824                 opt[0] = TCPOPT_MAXSEG;
1825                 opt[1] = 4;
1826                 mss = htons(mss);
1827                 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2);
1828         }
1829
1830         switch (af) {
1831 #ifdef INET
1832         case AF_INET:
1833                 /* TCP checksum */
1834                 th->th_sum = in_cksum(m, len);
1835
1836                 /* Finish the IP header */
1837                 h->ip_v = 4;
1838                 h->ip_hl = sizeof(*h) >> 2;
1839                 h->ip_tos = IPTOS_LOWDELAY;
1840                 h->ip_len = len;
1841                 h->ip_off = path_mtu_discovery ? IP_DF : 0;
1842                 h->ip_ttl = ttl ? ttl : ip_defttl;
1843                 h->ip_sum = 0;
1844                 if (eh == NULL) {
1845                         lwkt_reltoken(&pf_token);
1846                         ip_output(m, NULL, NULL, 0, NULL, NULL);
1847                         lwkt_gettoken(&pf_token);
1848                 } else {
1849                         struct route             ro;
1850                         struct rtentry           rt;
1851                         struct ether_header     *e = (void *)ro.ro_dst.sa_data;
1852
1853                         if (ifp == NULL) {
1854                                 m_freem(m);
1855                                 return;
1856                         }
1857                         rt.rt_ifp = ifp;
1858                         ro.ro_rt = &rt;
1859                         ro.ro_dst.sa_len = sizeof(ro.ro_dst);
1860                         ro.ro_dst.sa_family = pseudo_AF_HDRCMPLT;
1861                         bcopy(eh->ether_dhost, e->ether_shost, ETHER_ADDR_LEN);
1862                         bcopy(eh->ether_shost, e->ether_dhost, ETHER_ADDR_LEN);
1863                         e->ether_type = eh->ether_type;
1864                         /* XXX_IMPORT: later */
1865                         lwkt_reltoken(&pf_token);
1866                         ip_output(m, (void *)NULL, &ro, 0,
1867                             (void *)NULL, (void *)NULL);
1868                         lwkt_gettoken(&pf_token);
1869                 }
1870                 break;
1871 #endif /* INET */
1872 #ifdef INET6
1873         case AF_INET6:
1874                 /* TCP checksum */
1875                 th->th_sum = in6_cksum(m, IPPROTO_TCP,
1876                     sizeof(struct ip6_hdr), tlen);
1877
1878                 h6->ip6_vfc |= IPV6_VERSION;
1879                 h6->ip6_hlim = IPV6_DEFHLIM;
1880
1881                 lwkt_reltoken(&pf_token);
1882                 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
1883                 lwkt_gettoken(&pf_token);
1884                 break;
1885 #endif /* INET6 */
1886         }
1887 }
1888
1889 void
1890 pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af,
1891     struct pf_rule *r)
1892 {
1893         struct mbuf     *m0;
1894
1895         /*
1896          * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags,
1897          * so make sure pf.flags is clear.
1898          */
1899         if ((m0 = m_copy(m, 0, M_COPYALL)) == NULL)
1900                 return;
1901
1902         m0->m_pkthdr.fw_flags |= PF_MBUF_TAGGED;
1903         m0->m_pkthdr.pf.flags = 0;
1904         /* XXX Re-Check when Upgrading to > 4.4 */
1905         m0->m_pkthdr.pf.statekey = NULL;
1906
1907         if (r->rtableid >= 0)
1908                 m0->m_pkthdr.pf.rtableid = r->rtableid;
1909
1910 #ifdef ALTQ
1911         if (r->qid) {
1912                 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
1913                 m0->m_pkthdr.pf.qid = r->qid;
1914                 m0->m_pkthdr.pf.ecn_af = af;
1915                 m0->m_pkthdr.pf.hdr = mtod(m0, struct ip *);
1916         }
1917 #endif /* ALTQ */
1918
1919         switch (af) {
1920 #ifdef INET
1921         case AF_INET:
1922                 icmp_error(m0, type, code, 0, 0);
1923                 break;
1924 #endif /* INET */
1925 #ifdef INET6
1926         case AF_INET6:
1927                 icmp6_error(m0, type, code, 0);
1928                 break;
1929 #endif /* INET6 */
1930         }
1931 }
1932
1933 /*
1934  * Return 1 if the addresses a and b match (with mask m), otherwise return 0.
1935  * If n is 0, they match if they are equal. If n is != 0, they match if they
1936  * are different.
1937  */
1938 int
1939 pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m,
1940     struct pf_addr *b, sa_family_t af)
1941 {
1942         int     match = 0;
1943
1944         switch (af) {
1945 #ifdef INET
1946         case AF_INET:
1947                 if ((a->addr32[0] & m->addr32[0]) ==
1948                     (b->addr32[0] & m->addr32[0]))
1949                         match++;
1950                 break;
1951 #endif /* INET */
1952 #ifdef INET6
1953         case AF_INET6:
1954                 if (((a->addr32[0] & m->addr32[0]) ==
1955                      (b->addr32[0] & m->addr32[0])) &&
1956                     ((a->addr32[1] & m->addr32[1]) ==
1957                      (b->addr32[1] & m->addr32[1])) &&
1958                     ((a->addr32[2] & m->addr32[2]) ==
1959                      (b->addr32[2] & m->addr32[2])) &&
1960                     ((a->addr32[3] & m->addr32[3]) ==
1961                      (b->addr32[3] & m->addr32[3])))
1962                         match++;
1963                 break;
1964 #endif /* INET6 */
1965         }
1966         if (match) {
1967                 if (n)
1968                         return (0);
1969                 else
1970                         return (1);
1971         } else {
1972                 if (n)
1973                         return (1);
1974                 else
1975                         return (0);
1976         }
1977 }
1978
1979 /*
1980  * Return 1 if b <= a <= e, otherwise return 0.
1981  */
1982 int
1983 pf_match_addr_range(struct pf_addr *b, struct pf_addr *e,
1984     struct pf_addr *a, sa_family_t af)
1985 {
1986         switch (af) {
1987 #ifdef INET
1988         case AF_INET:
1989                 if ((a->addr32[0] < b->addr32[0]) ||
1990                     (a->addr32[0] > e->addr32[0]))
1991                         return (0);
1992                 break;
1993 #endif /* INET */
1994 #ifdef INET6
1995         case AF_INET6: {
1996                 int     i;
1997
1998                 /* check a >= b */
1999                 for (i = 0; i < 4; ++i)
2000                         if (a->addr32[i] > b->addr32[i])
2001                                 break;
2002                         else if (a->addr32[i] < b->addr32[i])
2003                                 return (0);
2004                 /* check a <= e */
2005                 for (i = 0; i < 4; ++i)
2006                         if (a->addr32[i] < e->addr32[i])
2007                                 break;
2008                         else if (a->addr32[i] > e->addr32[i])
2009                                 return (0);
2010                 break;
2011         }
2012 #endif /* INET6 */
2013         }
2014         return (1);
2015 }
2016
2017 int
2018 pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p)
2019 {
2020         switch (op) {
2021         case PF_OP_IRG:
2022                 return ((p > a1) && (p < a2));
2023         case PF_OP_XRG:
2024                 return ((p < a1) || (p > a2));
2025         case PF_OP_RRG:
2026                 return ((p >= a1) && (p <= a2));
2027         case PF_OP_EQ:
2028                 return (p == a1);
2029         case PF_OP_NE:
2030                 return (p != a1);
2031         case PF_OP_LT:
2032                 return (p < a1);
2033         case PF_OP_LE:
2034                 return (p <= a1);
2035         case PF_OP_GT:
2036                 return (p > a1);
2037         case PF_OP_GE:
2038                 return (p >= a1);
2039         }
2040         return (0); /* never reached */
2041 }
2042
2043 int
2044 pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p)
2045 {
2046         a1 = ntohs(a1);
2047         a2 = ntohs(a2);
2048         p = ntohs(p);
2049         return (pf_match(op, a1, a2, p));
2050 }
2051
2052 int
2053 pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u)
2054 {
2055         if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
2056                 return (0);
2057         return (pf_match(op, a1, a2, u));
2058 }
2059
2060 int
2061 pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g)
2062 {
2063         if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
2064                 return (0);
2065         return (pf_match(op, a1, a2, g));
2066 }
2067
2068 int
2069 pf_match_tag(struct mbuf *m, struct pf_rule *r, int *tag)
2070 {
2071         if (*tag == -1)
2072                 *tag = m->m_pkthdr.pf.tag;
2073
2074         return ((!r->match_tag_not && r->match_tag == *tag) ||
2075             (r->match_tag_not && r->match_tag != *tag));
2076 }
2077
2078 int
2079 pf_tag_packet(struct mbuf *m, int tag, int rtableid)
2080 {
2081         if (tag <= 0 && rtableid < 0)
2082                 return (0);
2083
2084         if (tag > 0)
2085                 m->m_pkthdr.pf.tag = tag;
2086         if (rtableid >= 0)
2087                 m->m_pkthdr.pf.rtableid = rtableid;
2088
2089         return (0);
2090 }
2091
2092 void
2093 pf_step_into_anchor(int *depth, struct pf_ruleset **rs, int n,
2094     struct pf_rule **r, struct pf_rule **a, int *match)
2095 {
2096         struct pf_anchor_stackframe     *f;
2097
2098         (*r)->anchor->match = 0;
2099         if (match)
2100                 *match = 0;
2101         if (*depth >= sizeof(pf_anchor_stack) /
2102             sizeof(pf_anchor_stack[0])) {
2103                 kprintf("pf_step_into_anchor: stack overflow\n");
2104                 *r = TAILQ_NEXT(*r, entries);
2105                 return;
2106         } else if (*depth == 0 && a != NULL)
2107                 *a = *r;
2108         f = pf_anchor_stack + (*depth)++;
2109         f->rs = *rs;
2110         f->r = *r;
2111         if ((*r)->anchor_wildcard) {
2112                 f->parent = &(*r)->anchor->children;
2113                 if ((f->child = RB_MIN(pf_anchor_node, f->parent)) ==
2114                     NULL) {
2115                         *r = NULL;
2116                         return;
2117                 }
2118                 *rs = &f->child->ruleset;
2119         } else {
2120                 f->parent = NULL;
2121                 f->child = NULL;
2122                 *rs = &(*r)->anchor->ruleset;
2123         }
2124         *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
2125 }
2126
2127 int
2128 pf_step_out_of_anchor(int *depth, struct pf_ruleset **rs, int n,
2129     struct pf_rule **r, struct pf_rule **a, int *match)
2130 {
2131         struct pf_anchor_stackframe     *f;
2132         int quick = 0;
2133
2134         do {
2135                 if (*depth <= 0)
2136                         break;
2137                 f = pf_anchor_stack + *depth - 1;
2138                 if (f->parent != NULL && f->child != NULL) {
2139                         if (f->child->match ||
2140                             (match != NULL && *match)) {
2141                                 f->r->anchor->match = 1;
2142                                 *match = 0;
2143                         }
2144                         f->child = RB_NEXT(pf_anchor_node, f->parent, f->child);
2145                         if (f->child != NULL) {
2146                                 *rs = &f->child->ruleset;
2147                                 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
2148                                 if (*r == NULL)
2149                                         continue;
2150                                 else
2151                                         break;
2152                         }
2153                 }
2154                 (*depth)--;
2155                 if (*depth == 0 && a != NULL)
2156                         *a = NULL;
2157                 *rs = f->rs;
2158                 if (f->r->anchor->match || (match != NULL && *match))
2159                         quick = f->r->quick;
2160                 *r = TAILQ_NEXT(f->r, entries);
2161         } while (*r == NULL);
2162
2163         return (quick);
2164 }
2165
2166 #ifdef INET6
2167 void
2168 pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr,
2169     struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af)
2170 {
2171         switch (af) {
2172 #ifdef INET
2173         case AF_INET:
2174                 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2175                 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2176                 break;
2177 #endif /* INET */
2178         case AF_INET6:
2179                 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2180                 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2181                 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) |
2182                 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]);
2183                 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) |
2184                 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]);
2185                 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) |
2186                 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]);
2187                 break;
2188         }
2189 }
2190
2191 void
2192 pf_addr_inc(struct pf_addr *addr, sa_family_t af)
2193 {
2194         switch (af) {
2195 #ifdef INET
2196         case AF_INET:
2197                 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1);
2198                 break;
2199 #endif /* INET */
2200         case AF_INET6:
2201                 if (addr->addr32[3] == 0xffffffff) {
2202                         addr->addr32[3] = 0;
2203                         if (addr->addr32[2] == 0xffffffff) {
2204                                 addr->addr32[2] = 0;
2205                                 if (addr->addr32[1] == 0xffffffff) {
2206                                         addr->addr32[1] = 0;
2207                                         addr->addr32[0] =
2208                                             htonl(ntohl(addr->addr32[0]) + 1);
2209                                 } else
2210                                         addr->addr32[1] =
2211                                             htonl(ntohl(addr->addr32[1]) + 1);
2212                         } else
2213                                 addr->addr32[2] =
2214                                     htonl(ntohl(addr->addr32[2]) + 1);
2215                 } else
2216                         addr->addr32[3] =
2217                             htonl(ntohl(addr->addr32[3]) + 1);
2218                 break;
2219         }
2220 }
2221 #endif /* INET6 */
2222
2223 #define mix(a,b,c) \
2224         do {                                    \
2225                 a -= b; a -= c; a ^= (c >> 13); \
2226                 b -= c; b -= a; b ^= (a << 8);  \
2227                 c -= a; c -= b; c ^= (b >> 13); \
2228                 a -= b; a -= c; a ^= (c >> 12); \
2229                 b -= c; b -= a; b ^= (a << 16); \
2230                 c -= a; c -= b; c ^= (b >> 5);  \
2231                 a -= b; a -= c; a ^= (c >> 3);  \
2232                 b -= c; b -= a; b ^= (a << 10); \
2233                 c -= a; c -= b; c ^= (b >> 15); \
2234         } while (0)
2235
2236 /*
2237  * hash function based on bridge_hash in if_bridge.c
2238  */
2239 void
2240 pf_hash(struct pf_addr *inaddr, struct pf_addr *hash,
2241     struct pf_poolhashkey *key, sa_family_t af)
2242 {
2243         u_int32_t       a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0];
2244
2245         switch (af) {
2246 #ifdef INET
2247         case AF_INET:
2248                 a += inaddr->addr32[0];
2249                 b += key->key32[1];
2250                 mix(a, b, c);
2251                 hash->addr32[0] = c + key->key32[2];
2252                 break;
2253 #endif /* INET */
2254 #ifdef INET6
2255         case AF_INET6:
2256                 a += inaddr->addr32[0];
2257                 b += inaddr->addr32[2];
2258                 mix(a, b, c);
2259                 hash->addr32[0] = c;
2260                 a += inaddr->addr32[1];
2261                 b += inaddr->addr32[3];
2262                 c += key->key32[1];
2263                 mix(a, b, c);
2264                 hash->addr32[1] = c;
2265                 a += inaddr->addr32[2];
2266                 b += inaddr->addr32[1];
2267                 c += key->key32[2];
2268                 mix(a, b, c);
2269                 hash->addr32[2] = c;
2270                 a += inaddr->addr32[3];
2271                 b += inaddr->addr32[0];
2272                 c += key->key32[3];
2273                 mix(a, b, c);
2274                 hash->addr32[3] = c;
2275                 break;
2276 #endif /* INET6 */
2277         }
2278 }
2279
2280 int
2281 pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr,
2282     struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn)
2283 {
2284         unsigned char            hash[16];
2285         struct pf_pool          *rpool = &r->rpool;
2286         struct pf_addr          *raddr = &rpool->cur->addr.v.a.addr;
2287         struct pf_addr          *rmask = &rpool->cur->addr.v.a.mask;
2288         struct pf_pooladdr      *acur = rpool->cur;
2289         struct pf_src_node       k;
2290
2291         if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR &&
2292             (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2293                 k.af = af;
2294                 PF_ACPY(&k.addr, saddr, af);
2295                 if (r->rule_flag & PFRULE_RULESRCTRACK ||
2296                     r->rpool.opts & PF_POOL_STICKYADDR)
2297                         k.rule.ptr = r;
2298                 else
2299                         k.rule.ptr = NULL;
2300                 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
2301                 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
2302                 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) {
2303                         PF_ACPY(naddr, &(*sn)->raddr, af);
2304                         if (pf_status.debug >= PF_DEBUG_MISC) {
2305                                 kprintf("pf_map_addr: src tracking maps ");
2306                                 pf_print_host(&k.addr, 0, af);
2307                                 kprintf(" to ");
2308                                 pf_print_host(naddr, 0, af);
2309                                 kprintf("\n");
2310                         }
2311                         return (0);
2312                 }
2313         }
2314
2315         if (rpool->cur->addr.type == PF_ADDR_NOROUTE)
2316                 return (1);
2317         if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2318                 switch (af) {
2319 #ifdef INET
2320                 case AF_INET:
2321                         if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 &&
2322                             (rpool->opts & PF_POOL_TYPEMASK) !=
2323                             PF_POOL_ROUNDROBIN)
2324                                 return (1);
2325                          raddr = &rpool->cur->addr.p.dyn->pfid_addr4;
2326                          rmask = &rpool->cur->addr.p.dyn->pfid_mask4;
2327                         break;
2328 #endif /* INET */
2329 #ifdef INET6
2330                 case AF_INET6:
2331                         if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 &&
2332                             (rpool->opts & PF_POOL_TYPEMASK) !=
2333                             PF_POOL_ROUNDROBIN)
2334                                 return (1);
2335                         raddr = &rpool->cur->addr.p.dyn->pfid_addr6;
2336                         rmask = &rpool->cur->addr.p.dyn->pfid_mask6;
2337                         break;
2338 #endif /* INET6 */
2339                 }
2340         } else if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2341                 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN)
2342                         return (1); /* unsupported */
2343         } else {
2344                 raddr = &rpool->cur->addr.v.a.addr;
2345                 rmask = &rpool->cur->addr.v.a.mask;
2346         }
2347
2348         switch (rpool->opts & PF_POOL_TYPEMASK) {
2349         case PF_POOL_NONE:
2350                 PF_ACPY(naddr, raddr, af);
2351                 break;
2352         case PF_POOL_BITMASK:
2353                 PF_POOLMASK(naddr, raddr, rmask, saddr, af);
2354                 break;
2355         case PF_POOL_RANDOM:
2356                 if (init_addr != NULL && PF_AZERO(init_addr, af)) {
2357                         switch (af) {
2358 #ifdef INET
2359                         case AF_INET:
2360                                 rpool->counter.addr32[0] = htonl(karc4random());
2361                                 break;
2362 #endif /* INET */
2363 #ifdef INET6
2364                         case AF_INET6:
2365                                 if (rmask->addr32[3] != 0xffffffff)
2366                                         rpool->counter.addr32[3] =
2367                                             htonl(karc4random());
2368                                 else
2369                                         break;
2370                                 if (rmask->addr32[2] != 0xffffffff)
2371                                         rpool->counter.addr32[2] =
2372                                             htonl(karc4random());
2373                                 else
2374                                         break;
2375                                 if (rmask->addr32[1] != 0xffffffff)
2376                                         rpool->counter.addr32[1] =
2377                                             htonl(karc4random());
2378                                 else
2379                                         break;
2380                                 if (rmask->addr32[0] != 0xffffffff)
2381                                         rpool->counter.addr32[0] =
2382                                             htonl(karc4random());
2383                                 break;
2384 #endif /* INET6 */
2385                         }
2386                         PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2387                         PF_ACPY(init_addr, naddr, af);
2388
2389                 } else {
2390                         PF_AINC(&rpool->counter, af);
2391                         PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2392                 }
2393                 break;
2394         case PF_POOL_SRCHASH:
2395                 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af);
2396                 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af);
2397                 break;
2398         case PF_POOL_ROUNDROBIN:
2399                 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2400                         if (!pfr_pool_get(rpool->cur->addr.p.tbl,
2401                             &rpool->tblidx, &rpool->counter,
2402                             &raddr, &rmask, af))
2403                                 goto get_addr;
2404                 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2405                         if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2406                             &rpool->tblidx, &rpool->counter,
2407                             &raddr, &rmask, af))
2408                                 goto get_addr;
2409                 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af))
2410                         goto get_addr;
2411
2412         try_next:
2413                 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL)
2414                         rpool->cur = TAILQ_FIRST(&rpool->list);
2415                 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2416                         rpool->tblidx = -1;
2417                         if (pfr_pool_get(rpool->cur->addr.p.tbl,
2418                             &rpool->tblidx, &rpool->counter,
2419                             &raddr, &rmask, af)) {
2420                                 /* table contains no address of type 'af' */
2421                                 if (rpool->cur != acur)
2422                                         goto try_next;
2423                                 return (1);
2424                         }
2425                 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2426                         rpool->tblidx = -1;
2427                         if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2428                             &rpool->tblidx, &rpool->counter,
2429                             &raddr, &rmask, af)) {
2430                                 /* table contains no address of type 'af' */
2431                                 if (rpool->cur != acur)
2432                                         goto try_next;
2433                                 return (1);
2434                         }
2435                 } else {
2436                         raddr = &rpool->cur->addr.v.a.addr;
2437                         rmask = &rpool->cur->addr.v.a.mask;
2438                         PF_ACPY(&rpool->counter, raddr, af);
2439                 }
2440
2441         get_addr:
2442                 PF_ACPY(naddr, &rpool->counter, af);
2443                 if (init_addr != NULL && PF_AZERO(init_addr, af))
2444                         PF_ACPY(init_addr, naddr, af);
2445                 PF_AINC(&rpool->counter, af);
2446                 break;
2447         }
2448         if (*sn != NULL)
2449                 PF_ACPY(&(*sn)->raddr, naddr, af);
2450
2451         if (pf_status.debug >= PF_DEBUG_MISC &&
2452             (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2453                 kprintf("pf_map_addr: selected address ");
2454                 pf_print_host(naddr, 0, af);
2455                 kprintf("\n");
2456         }
2457
2458         return (0);
2459 }
2460
2461 int
2462 pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r,
2463     struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport,
2464     struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high,
2465     struct pf_src_node **sn)
2466 {
2467         struct pf_state_key_cmp key;
2468         struct pf_addr          init_addr;
2469         u_int16_t               cut;
2470
2471         bzero(&init_addr, sizeof(init_addr));
2472         if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2473                 return (1);
2474
2475         if (proto == IPPROTO_ICMP) {
2476                 low = 1;
2477                 high = 65535;
2478         }
2479
2480         do {
2481                 key.af = af;
2482                 key.proto = proto;
2483                 PF_ACPY(&key.addr[1], daddr, key.af);
2484                 PF_ACPY(&key.addr[0], naddr, key.af);
2485                 key.port[1] = dport;
2486
2487                 /*
2488                  * port search; start random, step;
2489                  * similar 2 portloop in in_pcbbind
2490                  */
2491                 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP ||
2492                     proto == IPPROTO_ICMP)) {
2493                         key.port[0] = dport;
2494                         if (pf_find_state_all(&key, PF_IN, NULL) == NULL)
2495                                 return (0);
2496                 } else if (low == 0 && high == 0) {
2497                         key.port[0] = *nport;
2498                         if (pf_find_state_all(&key, PF_IN, NULL) == NULL)
2499                                 return (0);
2500                 } else if (low == high) {
2501                         key.port[0] = htons(low);
2502                         if (pf_find_state_all(&key, PF_IN, NULL) == NULL) {
2503                                 *nport = htons(low);
2504                                 return (0);
2505                         }
2506                 } else {
2507                         u_int16_t tmp;
2508
2509                         if (low > high) {
2510                                 tmp = low;
2511                                 low = high;
2512                                 high = tmp;
2513                         }
2514                         /* low < high */
2515                         cut = htonl(karc4random()) % (1 + high - low) + low;
2516                         /* low <= cut <= high */
2517                         for (tmp = cut; tmp <= high; ++(tmp)) {
2518                                 key.port[0] = htons(tmp);
2519                                 if (pf_find_state_all(&key, PF_IN, NULL) ==
2520                                     NULL && !in_baddynamic(tmp, proto)) {
2521                                         *nport = htons(tmp);
2522                                         return (0);
2523                                 }
2524                         }
2525                         for (tmp = cut - 1; tmp >= low; --(tmp)) {
2526                                 key.port[0] = htons(tmp);
2527                                 if (pf_find_state_all(&key, PF_IN, NULL) ==
2528                                     NULL && !in_baddynamic(tmp, proto)) {
2529                                         *nport = htons(tmp);
2530                                         return (0);
2531                                 }
2532                         }
2533                 }
2534
2535                 switch (r->rpool.opts & PF_POOL_TYPEMASK) {
2536                 case PF_POOL_RANDOM:
2537                 case PF_POOL_ROUNDROBIN:
2538                         if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2539                                 return (1);
2540                         break;
2541                 case PF_POOL_NONE:
2542                 case PF_POOL_SRCHASH:
2543                 case PF_POOL_BITMASK:
2544                 default:
2545                         return (1);
2546                 }
2547         } while (! PF_AEQ(&init_addr, naddr, af) );
2548         return (1);                                     /* none available */
2549 }
2550
2551 struct pf_rule *
2552 pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off,
2553     int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport,
2554     struct pf_addr *daddr, u_int16_t dport, int rs_num)
2555 {
2556         struct pf_rule          *r, *rm = NULL;
2557         struct pf_ruleset       *ruleset = NULL;
2558         int                      tag = -1;
2559         int                      rtableid = -1;
2560         int                      asd = 0;
2561
2562         r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr);
2563         while (r && rm == NULL) {
2564                 struct pf_rule_addr     *src = NULL, *dst = NULL;
2565                 struct pf_addr_wrap     *xdst = NULL;
2566
2567                 if (r->action == PF_BINAT && direction == PF_IN) {
2568                         src = &r->dst;
2569                         if (r->rpool.cur != NULL)
2570                                 xdst = &r->rpool.cur->addr;
2571                 } else {
2572                         src = &r->src;
2573                         dst = &r->dst;
2574                 }
2575
2576                 r->evaluations++;
2577                 if (pfi_kif_match(r->kif, kif) == r->ifnot)
2578                         r = r->skip[PF_SKIP_IFP].ptr;
2579                 else if (r->direction && r->direction != direction)
2580                         r = r->skip[PF_SKIP_DIR].ptr;
2581                 else if (r->af && r->af != pd->af)
2582                         r = r->skip[PF_SKIP_AF].ptr;
2583                 else if (r->proto && r->proto != pd->proto)
2584                         r = r->skip[PF_SKIP_PROTO].ptr;
2585                 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af,
2586                     src->neg, kif))
2587                         r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR :
2588                             PF_SKIP_DST_ADDR].ptr;
2589                 else if (src->port_op && !pf_match_port(src->port_op,
2590                     src->port[0], src->port[1], sport))
2591                         r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT :
2592                             PF_SKIP_DST_PORT].ptr;
2593                 else if (dst != NULL &&
2594                     PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->neg, NULL))
2595                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
2596                 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af,
2597                     0, NULL))
2598                         r = TAILQ_NEXT(r, entries);
2599                 else if (dst != NULL && dst->port_op &&
2600                     !pf_match_port(dst->port_op, dst->port[0],
2601                     dst->port[1], dport))
2602                         r = r->skip[PF_SKIP_DST_PORT].ptr;
2603                 else if (r->match_tag && !pf_match_tag(m, r, &tag))
2604                         r = TAILQ_NEXT(r, entries);
2605                 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto !=
2606                     IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m,
2607                     off, pd->hdr.tcp), r->os_fingerprint)))
2608                         r = TAILQ_NEXT(r, entries);
2609                 else {
2610                         if (r->tag)
2611                                 tag = r->tag;
2612                         if (r->rtableid >= 0)
2613                                 rtableid = r->rtableid;
2614                         if (r->anchor == NULL) {
2615                                 rm = r;
2616                         } else
2617                                 pf_step_into_anchor(&asd, &ruleset, rs_num,
2618                                     &r, NULL, NULL);
2619                 }
2620                 if (r == NULL)
2621                         pf_step_out_of_anchor(&asd, &ruleset, rs_num, &r,
2622                             NULL, NULL);
2623         }
2624         if (pf_tag_packet(m, tag, rtableid))
2625                 return (NULL);
2626         if (rm != NULL && (rm->action == PF_NONAT ||
2627             rm->action == PF_NORDR || rm->action == PF_NOBINAT))
2628                 return (NULL);
2629         return (rm);
2630 }
2631
2632 struct pf_rule *
2633 pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction,
2634     struct pfi_kif *kif, struct pf_src_node **sn,
2635     struct pf_state_key **skw, struct pf_state_key **sks,
2636     struct pf_state_key **skp, struct pf_state_key **nkp,
2637     struct pf_addr *saddr, struct pf_addr *daddr,
2638     u_int16_t sport, u_int16_t dport)
2639 {
2640         struct pf_rule  *r = NULL;
2641
2642
2643         if (direction == PF_OUT) {
2644                 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2645                     sport, daddr, dport, PF_RULESET_BINAT);
2646                 if (r == NULL)
2647                         r = pf_match_translation(pd, m, off, direction, kif,
2648                             saddr, sport, daddr, dport, PF_RULESET_NAT);
2649         } else {
2650                 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2651                     sport, daddr, dport, PF_RULESET_RDR);
2652                 if (r == NULL)
2653                         r = pf_match_translation(pd, m, off, direction, kif,
2654                             saddr, sport, daddr, dport, PF_RULESET_BINAT);
2655         }
2656
2657         if (r != NULL) {
2658                 struct pf_addr  *naddr;
2659                 u_int16_t       *nport;
2660
2661                 if (pf_state_key_setup(pd, r, skw, sks, skp, nkp,
2662                     saddr, daddr, sport, dport))
2663                         return r;
2664
2665                 /* XXX We only modify one side for now. */
2666                 naddr = &(*nkp)->addr[1];
2667                 nport = &(*nkp)->port[1];
2668
2669                 switch (r->action) {
2670                 case PF_NONAT:
2671                 case PF_NOBINAT:
2672                 case PF_NORDR:
2673                         return (NULL);
2674                 case PF_NAT:
2675                         if (pf_get_sport(pd->af, pd->proto, r, saddr,
2676                             daddr, dport, naddr, nport, r->rpool.proxy_port[0],
2677                             r->rpool.proxy_port[1], sn)) {
2678                                 DPFPRINTF(PF_DEBUG_MISC,
2679                                     ("pf: NAT proxy port allocation "
2680                                     "(%u-%u) failed\n",
2681                                     r->rpool.proxy_port[0],
2682                                     r->rpool.proxy_port[1]));
2683                                 return (NULL);
2684                         }
2685                         break;
2686                 case PF_BINAT:
2687                         switch (direction) {
2688                         case PF_OUT:
2689                                 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){
2690                                         switch (pd->af) {
2691 #ifdef INET
2692                                         case AF_INET:
2693                                                 if (r->rpool.cur->addr.p.dyn->
2694                                                     pfid_acnt4 < 1)
2695                                                         return (NULL);
2696                                                 PF_POOLMASK(naddr,
2697                                                     &r->rpool.cur->addr.p.dyn->
2698                                                     pfid_addr4,
2699                                                     &r->rpool.cur->addr.p.dyn->
2700                                                     pfid_mask4,
2701                                                     saddr, AF_INET);
2702                                                 break;
2703 #endif /* INET */
2704 #ifdef INET6
2705                                         case AF_INET6:
2706                                                 if (r->rpool.cur->addr.p.dyn->
2707                                                     pfid_acnt6 < 1)
2708                                                         return (NULL);
2709                                                 PF_POOLMASK(naddr,
2710                                                     &r->rpool.cur->addr.p.dyn->
2711                                                     pfid_addr6,
2712                                                     &r->rpool.cur->addr.p.dyn->
2713                                                     pfid_mask6,
2714                                                     saddr, AF_INET6);
2715                                                 break;
2716 #endif /* INET6 */
2717                                         }
2718                                 } else
2719                                         PF_POOLMASK(naddr,
2720                                             &r->rpool.cur->addr.v.a.addr,
2721                                             &r->rpool.cur->addr.v.a.mask,
2722                                             saddr, pd->af);
2723                                 break;
2724                         case PF_IN:
2725                                 if (r->src.addr.type == PF_ADDR_DYNIFTL) {
2726                                         switch (pd->af) {
2727 #ifdef INET
2728                                         case AF_INET:
2729                                                 if (r->src.addr.p.dyn->
2730                                                     pfid_acnt4 < 1)
2731                                                         return (NULL);
2732                                                 PF_POOLMASK(naddr,
2733                                                     &r->src.addr.p.dyn->
2734                                                     pfid_addr4,
2735                                                     &r->src.addr.p.dyn->
2736                                                     pfid_mask4,
2737                                                     daddr, AF_INET);
2738                                                 break;
2739 #endif /* INET */
2740 #ifdef INET6
2741                                         case AF_INET6:
2742                                                 if (r->src.addr.p.dyn->
2743                                                     pfid_acnt6 < 1)
2744                                                         return (NULL);
2745                                                 PF_POOLMASK(naddr,
2746                                                     &r->src.addr.p.dyn->
2747                                                     pfid_addr6,
2748                                                     &r->src.addr.p.dyn->
2749                                                     pfid_mask6,
2750                                                     daddr, AF_INET6);
2751                                                 break;
2752 #endif /* INET6 */
2753                                         }
2754                                 } else
2755                                         PF_POOLMASK(naddr,
2756                                             &r->src.addr.v.a.addr,
2757                                             &r->src.addr.v.a.mask, daddr,
2758                                             pd->af);
2759                                 break;
2760                         }
2761                         break;
2762                 case PF_RDR: {
2763                         if (pf_map_addr(pd->af, r, saddr, naddr, NULL, sn))
2764                                 return (NULL);
2765                         if ((r->rpool.opts & PF_POOL_TYPEMASK) ==
2766                             PF_POOL_BITMASK)
2767                                 PF_POOLMASK(naddr, naddr,
2768                                     &r->rpool.cur->addr.v.a.mask, daddr,
2769                                     pd->af);
2770
2771                         if (r->rpool.proxy_port[1]) {
2772                                 u_int32_t       tmp_nport;
2773
2774                                 tmp_nport = ((ntohs(dport) -
2775                                     ntohs(r->dst.port[0])) %
2776                                     (r->rpool.proxy_port[1] -
2777                                     r->rpool.proxy_port[0] + 1)) +
2778                                     r->rpool.proxy_port[0];
2779
2780                                 /* wrap around if necessary */
2781                                 if (tmp_nport > 65535)
2782                                         tmp_nport -= 65535;
2783                                 *nport = htons((u_int16_t)tmp_nport);
2784                         } else if (r->rpool.proxy_port[0])
2785                                 *nport = htons(r->rpool.proxy_port[0]);
2786                         break;
2787                 }
2788                 default:
2789                         return (NULL);
2790                 }
2791         }
2792
2793         return (r);
2794 }
2795
2796 #ifdef SMP
2797 struct netmsg_hashlookup {
2798         struct netmsg_base      base;
2799         struct inpcb            **nm_pinp;
2800         struct inpcbinfo        *nm_pcbinfo;
2801         struct pf_addr          *nm_saddr;
2802         struct pf_addr          *nm_daddr;
2803         uint16_t                nm_sport;
2804         uint16_t                nm_dport;
2805         sa_family_t             nm_af;
2806 };
2807
2808 static void
2809 in_pcblookup_hash_handler(netmsg_t msg)
2810 {
2811         struct netmsg_hashlookup *rmsg = (struct netmsg_hashlookup *)msg;
2812
2813         if (rmsg->nm_af == AF_INET)
2814                 *rmsg->nm_pinp = in_pcblookup_hash(rmsg->nm_pcbinfo,
2815                     rmsg->nm_saddr->v4, rmsg->nm_sport, rmsg->nm_daddr->v4,
2816                     rmsg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2817 #ifdef INET6
2818         else
2819                 *rmsg->nm_pinp = in6_pcblookup_hash(rmsg->nm_pcbinfo,
2820                     &rmsg->nm_saddr->v6, rmsg->nm_sport, &rmsg->nm_daddr->v6,
2821                     rmsg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2822 #endif /* INET6 */
2823         lwkt_replymsg(&rmsg->base.lmsg, 0);
2824 }
2825 #endif /* SMP */
2826
2827 int
2828 pf_socket_lookup(int direction, struct pf_pdesc *pd)
2829 {
2830         struct pf_addr          *saddr, *daddr;
2831         u_int16_t                sport, dport;
2832         struct inpcbinfo        *pi;
2833         struct inpcb            *inp;
2834 #ifdef SMP
2835         struct netmsg_hashlookup *msg = NULL;
2836 #endif
2837         int                      pi_cpu = 0;
2838
2839         if (pd == NULL)
2840                 return (-1);
2841         pd->lookup.uid = UID_MAX;
2842         pd->lookup.gid = GID_MAX;
2843         pd->lookup.pid = NO_PID;
2844         if (direction == PF_IN) {
2845                 saddr = pd->src;
2846                 daddr = pd->dst;
2847         } else {
2848                 saddr = pd->dst;
2849                 daddr = pd->src;
2850         }
2851         switch (pd->proto) {
2852         case IPPROTO_TCP:
2853                 if (pd->hdr.tcp == NULL)
2854                         return (-1);
2855                 sport = pd->hdr.tcp->th_sport;
2856                 dport = pd->hdr.tcp->th_dport;
2857
2858                 pi_cpu = tcp_addrcpu(saddr->v4.s_addr, sport, daddr->v4.s_addr, dport);
2859                 pi = &tcbinfo[pi_cpu];
2860 #ifdef SMP
2861                 /*
2862                  * Our netstack runs lockless on MP systems
2863                  * (only for TCP connections at the moment).
2864                  * 
2865                  * As we are not allowed to read another CPU's tcbinfo,
2866                  * we have to ask that CPU via remote call to search the
2867                  * table for us.
2868                  * 
2869                  * Prepare a msg iff data belongs to another CPU.
2870                  */
2871                 if (pi_cpu != mycpu->gd_cpuid) {
2872                         msg = kmalloc(sizeof(*msg), M_LWKTMSG, M_INTWAIT);
2873                         netmsg_init(&msg->base, NULL, &netisr_afree_rport,
2874                                     0, in_pcblookup_hash_handler);
2875                         msg->nm_pinp = &inp;
2876                         msg->nm_pcbinfo = pi;
2877                         msg->nm_saddr = saddr;
2878                         msg->nm_sport = sport;
2879                         msg->nm_daddr = daddr;
2880                         msg->nm_dport = dport;
2881                         msg->nm_af = pd->af;
2882                 }
2883 #endif /* SMP */
2884                 break;
2885         case IPPROTO_UDP:
2886                 if (pd->hdr.udp == NULL)
2887                         return (-1);
2888                 sport = pd->hdr.udp->uh_sport;
2889                 dport = pd->hdr.udp->uh_dport;
2890                 pi = &udbinfo;
2891                 break;
2892         default:
2893                 return (-1);
2894         }
2895         if (direction != PF_IN) {
2896                 u_int16_t       p;
2897
2898                 p = sport;
2899                 sport = dport;
2900                 dport = p;
2901         }
2902         switch (pd->af) {
2903 #ifdef INET6
2904         case AF_INET6:
2905 #ifdef SMP
2906                 /*
2907                  * Query other CPU, second part
2908                  * 
2909                  * msg only gets initialized when:
2910                  * 1) packet is TCP
2911                  * 2) the info belongs to another CPU
2912                  *
2913                  * Use some switch/case magic to avoid code duplication.
2914                  */
2915                 if (msg == NULL)
2916 #endif /* SMP */
2917                 {
2918                         inp = in6_pcblookup_hash(pi, &saddr->v6, sport,
2919                             &daddr->v6, dport, INPLOOKUP_WILDCARD, NULL);
2920
2921                         if (inp == NULL)
2922                                 return (-1);
2923                         break;
2924                 }
2925                 /* FALLTHROUGH if SMP and on other CPU */
2926 #endif /* INET6 */
2927         case AF_INET:
2928 #ifdef SMP
2929                 if (msg != NULL) {
2930                         lwkt_domsg(cpu_portfn(pi_cpu),
2931                                      &msg->base.lmsg, 0);
2932                 } else
2933 #endif /* SMP */
2934                 {
2935                         inp = in_pcblookup_hash(pi, saddr->v4, sport, daddr->v4,
2936                             dport, INPLOOKUP_WILDCARD, NULL);
2937                 }
2938                 if (inp == NULL)
2939                         return (-1);
2940                 break;
2941
2942         default:
2943                 return (-1);
2944         }
2945         pd->lookup.uid = inp->inp_socket->so_cred->cr_uid;
2946         pd->lookup.gid = inp->inp_socket->so_cred->cr_groups[0];
2947         return (1);
2948 }
2949
2950 u_int8_t
2951 pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2952 {
2953         int              hlen;
2954         u_int8_t         hdr[60];
2955         u_int8_t        *opt, optlen;
2956         u_int8_t         wscale = 0;
2957
2958         hlen = th_off << 2;             /* hlen <= sizeof(hdr) */
2959         if (hlen <= sizeof(struct tcphdr))
2960                 return (0);
2961         if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2962                 return (0);
2963         opt = hdr + sizeof(struct tcphdr);
2964         hlen -= sizeof(struct tcphdr);
2965         while (hlen >= 3) {
2966                 switch (*opt) {
2967                 case TCPOPT_EOL:
2968                 case TCPOPT_NOP:
2969                         ++opt;
2970                         --hlen;
2971                         break;
2972                 case TCPOPT_WINDOW:
2973                         wscale = opt[2];
2974                         if (wscale > TCP_MAX_WINSHIFT)
2975                                 wscale = TCP_MAX_WINSHIFT;
2976                         wscale |= PF_WSCALE_FLAG;
2977                         /* FALLTHROUGH */
2978                 default:
2979                         optlen = opt[1];
2980                         if (optlen < 2)
2981                                 optlen = 2;
2982                         hlen -= optlen;
2983                         opt += optlen;
2984                         break;
2985                 }
2986         }
2987         return (wscale);
2988 }
2989
2990 u_int16_t
2991 pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2992 {
2993         int              hlen;
2994         u_int8_t         hdr[60];
2995         u_int8_t        *opt, optlen;
2996         u_int16_t        mss = tcp_mssdflt;
2997
2998         hlen = th_off << 2;     /* hlen <= sizeof(hdr) */
2999         if (hlen <= sizeof(struct tcphdr))
3000                 return (0);
3001         if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
3002                 return (0);
3003         opt = hdr + sizeof(struct tcphdr);
3004         hlen -= sizeof(struct tcphdr);
3005         while (hlen >= TCPOLEN_MAXSEG) {
3006                 switch (*opt) {
3007                 case TCPOPT_EOL:
3008                 case TCPOPT_NOP:
3009                         ++opt;
3010                         --hlen;
3011                         break;
3012                 case TCPOPT_MAXSEG:
3013                         bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2);
3014                         /* FALLTHROUGH */
3015                 default:
3016                         optlen = opt[1];
3017                         if (optlen < 2)
3018                                 optlen = 2;
3019                         hlen -= optlen;
3020                         opt += optlen;
3021                         break;
3022                 }
3023         }
3024         return (mss);
3025 }
3026
3027 u_int16_t
3028 pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer)
3029 {
3030 #ifdef INET
3031         struct sockaddr_in      *dst;
3032         struct route             ro;
3033 #endif /* INET */
3034 #ifdef INET6
3035         struct sockaddr_in6     *dst6;
3036         struct route_in6         ro6;
3037 #endif /* INET6 */
3038         struct rtentry          *rt = NULL;
3039         int                      hlen = 0;
3040         u_int16_t                mss = tcp_mssdflt;
3041
3042         switch (af) {
3043 #ifdef INET
3044         case AF_INET:
3045                 hlen = sizeof(struct ip);
3046                 bzero(&ro, sizeof(ro));
3047                 dst = (struct sockaddr_in *)&ro.ro_dst;
3048                 dst->sin_family = AF_INET;
3049                 dst->sin_len = sizeof(*dst);
3050                 dst->sin_addr = addr->v4;
3051                 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
3052                 rt = ro.ro_rt;
3053                 break;
3054 #endif /* INET */
3055 #ifdef INET6
3056         case AF_INET6:
3057                 hlen = sizeof(struct ip6_hdr);
3058                 bzero(&ro6, sizeof(ro6));
3059                 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst;
3060                 dst6->sin6_family = AF_INET6;
3061                 dst6->sin6_len = sizeof(*dst6);
3062                 dst6->sin6_addr = addr->v6;
3063                 rtalloc_ign((struct route *)&ro6, (RTF_CLONING | RTF_PRCLONING));
3064                 rt = ro6.ro_rt;
3065                 break;
3066 #endif /* INET6 */
3067         }
3068
3069         if (rt && rt->rt_ifp) {
3070                 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr);
3071                 mss = max(tcp_mssdflt, mss);
3072                 RTFREE(rt);
3073         }
3074         mss = min(mss, offer);
3075         mss = max(mss, 64);             /* sanity - at least max opt space */
3076         return (mss);
3077 }
3078
3079 void
3080 pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr)
3081 {
3082         struct pf_rule *r = s->rule.ptr;
3083
3084         s->rt_kif = NULL;
3085         if (!r->rt || r->rt == PF_FASTROUTE)
3086                 return;
3087         switch (s->key[PF_SK_WIRE]->af) {
3088 #ifdef INET
3089         case AF_INET:
3090                 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL,
3091                     &s->nat_src_node);
3092                 s->rt_kif = r->rpool.cur->kif;
3093                 break;
3094 #endif /* INET */
3095 #ifdef INET6
3096         case AF_INET6:
3097                 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL,
3098                     &s->nat_src_node);
3099                 s->rt_kif = r->rpool.cur->kif;
3100                 break;
3101 #endif /* INET6 */
3102         }
3103 }
3104
3105 u_int32_t
3106 pf_tcp_iss(struct pf_pdesc *pd)
3107 {
3108         MD5_CTX ctx;
3109         u_int32_t digest[4];
3110
3111         if (pf_tcp_secret_init == 0) {
3112                 karc4rand(pf_tcp_secret, sizeof(pf_tcp_secret));
3113                 MD5Init(&pf_tcp_secret_ctx);
3114                 MD5Update(&pf_tcp_secret_ctx, pf_tcp_secret,
3115                     sizeof(pf_tcp_secret));
3116                 pf_tcp_secret_init = 1;
3117         }
3118         ctx = pf_tcp_secret_ctx;
3119
3120         MD5Update(&ctx, (char *)&pd->hdr.tcp->th_sport, sizeof(u_short));
3121         MD5Update(&ctx, (char *)&pd->hdr.tcp->th_dport, sizeof(u_short));
3122         if (pd->af == AF_INET6) {
3123                 MD5Update(&ctx, (char *)&pd->src->v6, sizeof(struct in6_addr));
3124                 MD5Update(&ctx, (char *)&pd->dst->v6, sizeof(struct in6_addr));
3125         } else {
3126                 MD5Update(&ctx, (char *)&pd->src->v4, sizeof(struct in_addr));
3127                 MD5Update(&ctx, (char *)&pd->dst->v4, sizeof(struct in_addr));
3128         }
3129         MD5Final((u_char *)digest, &ctx);
3130         pf_tcp_iss_off += 4096;
3131         return (digest[0] + pd->hdr.tcp->th_seq + pf_tcp_iss_off);
3132 }
3133
3134 int
3135 pf_test_rule(struct pf_rule **rm, struct pf_state **sm, int direction,
3136     struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3137     struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
3138     struct ifqueue *ifq, struct inpcb *inp)
3139 {
3140         struct pf_rule          *nr = NULL;
3141         struct pf_addr          *saddr = pd->src, *daddr = pd->dst;
3142         sa_family_t              af = pd->af;
3143         struct pf_rule          *r, *a = NULL;
3144         struct pf_ruleset       *ruleset = NULL;
3145         struct pf_src_node      *nsn = NULL;
3146         struct tcphdr           *th = pd->hdr.tcp;
3147         struct pf_state_key     *skw = NULL, *sks = NULL;
3148         struct pf_state_key     *sk = NULL, *nk = NULL;
3149         u_short                  reason;
3150         int                      rewrite = 0, hdrlen = 0;
3151         int                      tag = -1, rtableid = -1;
3152         int                      asd = 0;
3153         int                      match = 0;
3154         int                      state_icmp = 0;
3155         u_int16_t                sport = 0, dport = 0;
3156         u_int16_t                nport = 0, bport = 0;
3157         u_int16_t                bproto_sum = 0, bip_sum = 0;
3158         u_int8_t                 icmptype = 0, icmpcode = 0;
3159
3160
3161         if (direction == PF_IN && pf_check_congestion(ifq)) {
3162                 REASON_SET(&reason, PFRES_CONGEST);
3163                 return (PF_DROP);
3164         }
3165
3166         if (inp != NULL)
3167                 pd->lookup.done = pf_socket_lookup(direction, pd);
3168         else if (debug_pfugidhack) { 
3169                 DPFPRINTF(PF_DEBUG_MISC, ("pf: unlocked lookup\n"));
3170                 pd->lookup.done = pf_socket_lookup(direction, pd);
3171         }
3172
3173         switch (pd->proto) {
3174         case IPPROTO_TCP:
3175                 sport = th->th_sport;
3176                 dport = th->th_dport;
3177                 hdrlen = sizeof(*th);
3178                 break;
3179         case IPPROTO_UDP:
3180                 sport = pd->hdr.udp->uh_sport;
3181                 dport = pd->hdr.udp->uh_dport;
3182                 hdrlen = sizeof(*pd->hdr.udp);
3183                 break;
3184 #ifdef INET
3185         case IPPROTO_ICMP:
3186                 if (pd->af != AF_INET)
3187                         break;
3188                 sport = dport = pd->hdr.icmp->icmp_id;
3189                 hdrlen = sizeof(*pd->hdr.icmp);
3190                 icmptype = pd->hdr.icmp->icmp_type;
3191                 icmpcode = pd->hdr.icmp->icmp_code;
3192
3193                 if (icmptype == ICMP_UNREACH ||
3194                     icmptype == ICMP_SOURCEQUENCH ||
3195                     icmptype == ICMP_REDIRECT ||
3196                     icmptype == ICMP_TIMXCEED ||
3197                     icmptype == ICMP_PARAMPROB)
3198                         state_icmp++;
3199                 break;
3200 #endif /* INET */
3201 #ifdef INET6
3202         case IPPROTO_ICMPV6:
3203                 if (af != AF_INET6)
3204                         break;
3205                 sport = dport = pd->hdr.icmp6->icmp6_id;
3206                 hdrlen = sizeof(*pd->hdr.icmp6);
3207                 icmptype = pd->hdr.icmp6->icmp6_type;
3208                 icmpcode = pd->hdr.icmp6->icmp6_code;
3209
3210                 if (icmptype == ICMP6_DST_UNREACH ||
3211                     icmptype == ICMP6_PACKET_TOO_BIG ||
3212                     icmptype == ICMP6_TIME_EXCEEDED ||
3213                     icmptype == ICMP6_PARAM_PROB)
3214                         state_icmp++;
3215                 break;
3216 #endif /* INET6 */
3217         default:
3218                 sport = dport = hdrlen = 0;
3219                 break;
3220         }
3221
3222         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3223
3224         bport = nport = sport;
3225         /* check packet for BINAT/NAT/RDR */
3226         if ((nr = pf_get_translation(pd, m, off, direction, kif, &nsn,
3227             &skw, &sks, &sk, &nk, saddr, daddr, sport, dport)) != NULL) {
3228                 if (nk == NULL || sk == NULL) {
3229                         REASON_SET(&reason, PFRES_MEMORY);
3230                         goto cleanup;
3231                 }
3232
3233                 if (pd->ip_sum)
3234                         bip_sum = *pd->ip_sum;
3235
3236                 switch (pd->proto) {
3237                 case IPPROTO_TCP:
3238                         bproto_sum = th->th_sum;
3239                         pd->proto_sum = &th->th_sum;
3240
3241                         if (PF_ANEQ(saddr, &nk->addr[pd->sidx], af) ||
3242                             nk->port[pd->sidx] != sport) {
3243                                 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
3244                                     &th->th_sum, &nk->addr[pd->sidx],
3245                                     nk->port[pd->sidx], 0, af);
3246                                 pd->sport = &th->th_sport;
3247                                 sport = th->th_sport;
3248                         }
3249
3250                         if (PF_ANEQ(daddr, &nk->addr[pd->didx], af) ||
3251                             nk->port[pd->didx] != dport) {
3252                                 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
3253                                     &th->th_sum, &nk->addr[pd->didx],
3254                                     nk->port[pd->didx], 0, af);
3255                                 dport = th->th_dport;
3256                                 pd->dport = &th->th_dport;
3257                         }
3258                         rewrite++;
3259                         break;
3260                 case IPPROTO_UDP:
3261                         bproto_sum = pd->hdr.udp->uh_sum;
3262                         pd->proto_sum = &pd->hdr.udp->uh_sum;
3263
3264                         if (PF_ANEQ(saddr, &nk->addr[pd->sidx], af) ||
3265                             nk->port[pd->sidx] != sport) {
3266                                 pf_change_ap(saddr, &pd->hdr.udp->uh_sport,
3267                                     pd->ip_sum, &pd->hdr.udp->uh_sum,
3268                                     &nk->addr[pd->sidx],
3269                                     nk->port[pd->sidx], 1, af);
3270                                 sport = pd->hdr.udp->uh_sport;
3271                                 pd->sport = &pd->hdr.udp->uh_sport;
3272                         }
3273
3274                         if (PF_ANEQ(daddr, &nk->addr[pd->didx], af) ||
3275                             nk->port[pd->didx] != dport) {
3276                                 pf_change_ap(daddr, &pd->hdr.udp->uh_dport,
3277                                     pd->ip_sum, &pd->hdr.udp->uh_sum,
3278                                     &nk->addr[pd->didx],
3279                                     nk->port[pd->didx], 1, af);
3280                                 dport = pd->hdr.udp->uh_dport;
3281                                 pd->dport = &pd->hdr.udp->uh_dport;
3282                         }
3283                         rewrite++;
3284                         break;
3285 #ifdef INET
3286                 case IPPROTO_ICMP:
3287                         nk->port[0] = nk->port[1];
3288                         if (PF_ANEQ(saddr, &nk->addr[pd->sidx], AF_INET))
3289                                 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3290                                     nk->addr[pd->sidx].v4.s_addr, 0);
3291
3292                         if (PF_ANEQ(daddr, &nk->addr[pd->didx], AF_INET))
3293                                 pf_change_a(&daddr->v4.s_addr, pd->ip_sum,
3294                                     nk->addr[pd->didx].v4.s_addr, 0);
3295
3296                         if (nk->port[1] != pd->hdr.icmp->icmp_id) {
3297                                 pd->hdr.icmp->icmp_cksum = pf_cksum_fixup(
3298                                     pd->hdr.icmp->icmp_cksum, sport,
3299                                     nk->port[1], 0);
3300                                 pd->hdr.icmp->icmp_id = nk->port[1];
3301                                 pd->sport = &pd->hdr.icmp->icmp_id;
3302                         }
3303                         m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
3304                         break;
3305 #endif /* INET */
3306 #ifdef INET6
3307                 case IPPROTO_ICMPV6:
3308                         nk->port[0] = nk->port[1];
3309                         if (PF_ANEQ(saddr, &nk->addr[pd->sidx], AF_INET6))
3310                                 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum,
3311                                     &nk->addr[pd->sidx], 0);
3312
3313                         if (PF_ANEQ(daddr, &nk->addr[pd->didx], AF_INET6))
3314                                 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum,
3315                                     &nk->addr[pd->didx], 0);
3316                         rewrite++;
3317                         break;
3318 #endif /* INET */
3319                 default:
3320                         switch (af) {
3321 #ifdef INET
3322                         case AF_INET:
3323                                 if (PF_ANEQ(saddr,
3324                                     &nk->addr[pd->sidx], AF_INET))
3325                                         pf_change_a(&saddr->v4.s_addr,
3326                                             pd->ip_sum,
3327                                             nk->addr[pd->sidx].v4.s_addr, 0);
3328
3329                                 if (PF_ANEQ(daddr,
3330                                     &nk->addr[pd->didx], AF_INET))
3331                                         pf_change_a(&daddr->v4.s_addr,
3332                                             pd->ip_sum,
3333                                             nk->addr[pd->didx].v4.s_addr, 0);
3334                                 break;
3335 #endif /* INET */
3336 #ifdef INET6
3337                         case AF_INET6:
3338                                 if (PF_ANEQ(saddr,
3339                                     &nk->addr[pd->sidx], AF_INET6))
3340                                         PF_ACPY(saddr, &nk->addr[pd->sidx], af);
3341
3342                                 if (PF_ANEQ(daddr,
3343                                     &nk->addr[pd->didx], AF_INET6))
3344                                         PF_ACPY(saddr, &nk->addr[pd->didx], af);
3345                                 break;
3346 #endif /* INET */
3347                         }
3348                         break;
3349                 }
3350                 if (nr->natpass)
3351                         r = NULL;
3352                 pd->nat_rule = nr;
3353         }
3354
3355         while (r != NULL) {
3356                 r->evaluations++;
3357                 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3358                         r = r->skip[PF_SKIP_IFP].ptr;
3359                 else if (r->direction && r->direction != direction)
3360                         r = r->skip[PF_SKIP_DIR].ptr;
3361                 else if (r->af && r->af != af)
3362                         r = r->skip[PF_SKIP_AF].ptr;
3363                 else if (r->proto && r->proto != pd->proto)
3364                         r = r->skip[PF_SKIP_PROTO].ptr;
3365                 else if (PF_MISMATCHAW(&r->src.addr, saddr, af,
3366                     r->src.neg, kif))
3367                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3368                 /* tcp/udp only. port_op always 0 in other cases */
3369                 else if (r->src.port_op && !pf_match_port(r->src.port_op,
3370                     r->src.port[0], r->src.port[1], sport))
3371                         r = r->skip[PF_SKIP_SRC_PORT].ptr;
3372                 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af,
3373                     r->dst.neg, NULL))
3374                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
3375                 /* tcp/udp only. port_op always 0 in other cases */
3376                 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
3377                     r->dst.port[0], r->dst.port[1], dport))
3378                         r = r->skip[PF_SKIP_DST_PORT].ptr;
3379                 /* icmp only. type always 0 in other cases */
3380                 else if (r->type && r->type != icmptype + 1)
3381                         r = TAILQ_NEXT(r, entries);
3382                 /* icmp only. type always 0 in other cases */
3383                 else if (r->code && r->code != icmpcode + 1)
3384                         r = TAILQ_NEXT(r, entries);
3385                 else if (r->tos && !(r->tos == pd->tos))
3386                         r = TAILQ_NEXT(r, entries);
3387                 else if (r->rule_flag & PFRULE_FRAGMENT)
3388                         r = TAILQ_NEXT(r, entries);
3389                 else if (pd->proto == IPPROTO_TCP &&
3390                     (r->flagset & th->th_flags) != r->flags)
3391                         r = TAILQ_NEXT(r, entries);
3392                 /* tcp/udp only. uid.op always 0 in other cases */
3393                 else if (r->uid.op && (pd->lookup.done || (pd->lookup.done =
3394                     pf_socket_lookup(direction, pd), 1)) &&
3395                     !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
3396                     pd->lookup.uid))
3397                         r = TAILQ_NEXT(r, entries);
3398                 /* tcp/udp only. gid.op always 0 in other cases */
3399                 else if (r->gid.op && (pd->lookup.done || (pd->lookup.done =
3400                     pf_socket_lookup(direction, pd), 1)) &&
3401                     !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
3402                     pd->lookup.gid))
3403                         r = TAILQ_NEXT(r, entries);
3404                 else if (r->prob &&
3405                   r->prob <= karc4random())
3406                         r = TAILQ_NEXT(r, entries);
3407                 else if (r->match_tag && !pf_match_tag(m, r, &tag))
3408                         r = TAILQ_NEXT(r, entries);
3409                 else if (r->os_fingerprint != PF_OSFP_ANY &&
3410                     (pd->proto != IPPROTO_TCP || !pf_osfp_match(
3411                     pf_osfp_fingerprint(pd, m, off, th),
3412                     r->os_fingerprint)))
3413                         r = TAILQ_NEXT(r, entries);
3414                 else {
3415                         if (r->tag)
3416                                 tag = r->tag;
3417                         if (r->rtableid >= 0)
3418                                 rtableid = r->rtableid;
3419                         if (r->anchor == NULL) {
3420                                 match = 1;
3421                                 *rm = r;
3422                                 *am = a;
3423                                 *rsm = ruleset;
3424                                 if ((*rm)->quick)
3425                                         break;
3426                                 r = TAILQ_NEXT(r, entries);
3427                         } else
3428                                 pf_step_into_anchor(&asd, &ruleset,
3429                                     PF_RULESET_FILTER, &r, &a, &match);
3430                 }
3431                 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3432                     PF_RULESET_FILTER, &r, &a, &match))
3433                         break;
3434         }
3435         r = *rm;
3436         a = *am;
3437         ruleset = *rsm;
3438
3439         REASON_SET(&reason, PFRES_MATCH);
3440
3441         if (r->log || (nr != NULL && nr->log)) {
3442                 if (rewrite)
3443                         m_copyback(m, off, hdrlen, pd->hdr.any);
3444                 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr,
3445                     a, ruleset, pd);
3446         }
3447
3448         if ((r->action == PF_DROP) &&
3449             ((r->rule_flag & PFRULE_RETURNRST) ||
3450             (r->rule_flag & PFRULE_RETURNICMP) ||
3451             (r->rule_flag & PFRULE_RETURN))) {
3452                 /* undo NAT changes, if they have taken place */
3453                 if (nr != NULL) {
3454                         PF_ACPY(saddr, &sk->addr[pd->sidx], af);
3455                         PF_ACPY(daddr, &sk->addr[pd->didx], af);
3456                         if (pd->sport)
3457                                 *pd->sport = sk->port[pd->sidx];
3458                         if (pd->dport)
3459                                 *pd->dport = sk->port[pd->didx];
3460                         if (pd->proto_sum)
3461                                 *pd->proto_sum = bproto_sum;
3462                         if (pd->ip_sum)
3463                                 *pd->ip_sum = bip_sum;
3464                         m_copyback(m, off, hdrlen, pd->hdr.any);
3465                 }
3466                 if (pd->proto == IPPROTO_TCP &&
3467                     ((r->rule_flag & PFRULE_RETURNRST) ||
3468                     (r->rule_flag & PFRULE_RETURN)) &&
3469                     !(th->th_flags & TH_RST)) {
3470                         u_int32_t        ack = ntohl(th->th_seq) + pd->p_len;
3471                         int              len = 0;
3472                         struct ip       *h4;
3473                         struct ip6_hdr  *h6;
3474
3475                         switch (af) {
3476                         case AF_INET:
3477                                 h4 = mtod(m, struct ip *);
3478                                 len = h4->ip_len - off;
3479                                 break;
3480                         case AF_INET6:
3481                                 h6 = mtod(m, struct ip6_hdr *);
3482                                 len = h6->ip6_plen - (off - sizeof(*h6));
3483                                 break;
3484                         }
3485
3486                         if (pf_check_proto_cksum(m, off, len, IPPROTO_TCP, af))
3487                                 REASON_SET(&reason, PFRES_PROTCKSUM);
3488                         else {
3489                                 if (th->th_flags & TH_SYN)
3490                                         ack++;
3491                                 if (th->th_flags & TH_FIN)
3492                                         ack++;
3493                                 pf_send_tcp(r, af, pd->dst,
3494                                     pd->src, th->th_dport, th->th_sport,
3495                                     ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0,
3496                                     r->return_ttl, 1, 0, pd->eh, kif->pfik_ifp);
3497                         }
3498                 } else if (pd->proto != IPPROTO_ICMP && af == AF_INET &&
3499                     r->return_icmp)
3500                         pf_send_icmp(m, r->return_icmp >> 8,
3501                             r->return_icmp & 255, af, r);
3502                 else if (pd->proto != IPPROTO_ICMPV6 && af == AF_INET6 &&
3503                     r->return_icmp6)
3504                         pf_send_icmp(m, r->return_icmp6 >> 8,
3505                             r->return_icmp6 & 255, af, r);
3506         }
3507
3508         if (r->action == PF_DROP)
3509                 goto cleanup;
3510
3511         if (pf_tag_packet(m, tag, rtableid)) {
3512                 REASON_SET(&reason, PFRES_MEMORY);
3513                 goto cleanup;
3514         }
3515
3516         if (!state_icmp && (r->keep_state || nr != NULL ||
3517             (pd->flags & PFDESC_TCP_NORM))) {
3518                 int action;
3519                 action = pf_create_state(r, nr, a, pd, nsn, skw, sks, nk, sk, m,
3520                     off, sport, dport, &rewrite, kif, sm, tag, bproto_sum,
3521                     bip_sum, hdrlen);
3522                 if (action != PF_PASS)
3523                         return (action);
3524         }
3525
3526         /* copy back packet headers if we performed NAT operations */
3527         if (rewrite)
3528                 m_copyback(m, off, hdrlen, pd->hdr.any);
3529
3530         return (PF_PASS);
3531
3532 cleanup:
3533         if (sk != NULL)
3534                 pool_put(&pf_state_key_pl, sk);
3535         if (nk != NULL)
3536                 pool_put(&pf_state_key_pl, nk);
3537         return (PF_DROP);
3538 }
3539
3540 static __inline int
3541 pf_create_state(struct pf_rule *r, struct pf_rule *nr, struct pf_rule *a,
3542     struct pf_pdesc *pd, struct pf_src_node *nsn, struct pf_state_key *skw,
3543     struct pf_state_key *sks, struct pf_state_key *nk, struct pf_state_key *sk,
3544     struct mbuf *m, int off, u_int16_t sport, u_int16_t dport, int *rewrite,
3545     struct pfi_kif *kif, struct pf_state **sm, int tag, u_int16_t bproto_sum,
3546     u_int16_t bip_sum, int hdrlen)
3547 {
3548         struct pf_state         *s = NULL;
3549         struct pf_src_node      *sn = NULL;
3550         struct tcphdr           *th = pd->hdr.tcp;
3551         u_int16_t                mss = tcp_mssdflt;
3552         u_short                  reason;
3553
3554         /* check maximums */
3555         if (r->max_states && (r->states_cur >= r->max_states)) {
3556                 pf_status.lcounters[LCNT_STATES]++;
3557                 REASON_SET(&reason, PFRES_MAXSTATES);
3558                 return (PF_DROP);
3559         }
3560         /* src node for filter rule */
3561         if ((r->rule_flag & PFRULE_SRCTRACK ||
3562             r->rpool.opts & PF_POOL_STICKYADDR) &&
3563             pf_insert_src_node(&sn, r, pd->src, pd->af) != 0) {
3564                 REASON_SET(&reason, PFRES_SRCLIMIT);
3565                 goto csfailed;
3566         }
3567         /* src node for translation rule */
3568         if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3569             pf_insert_src_node(&nsn, nr, &sk->addr[pd->sidx], pd->af)) {
3570                 REASON_SET(&reason, PFRES_SRCLIMIT);
3571                 goto csfailed;
3572         }
3573         s = pool_get(&pf_state_pl, PR_NOWAIT | PR_ZERO);
3574         if (s == NULL) {
3575                 REASON_SET(&reason, PFRES_MEMORY);
3576                 goto csfailed;
3577         }
3578         s->id = 0; /* XXX Do we really need that? not in OpenBSD */
3579         s->creatorid = 0;
3580         s->rule.ptr = r;
3581         s->nat_rule.ptr = nr;
3582         s->anchor.ptr = a;
3583         STATE_INC_COUNTERS(s);
3584         if (r->allow_opts)
3585                 s->state_flags |= PFSTATE_ALLOWOPTS;
3586         if (r->rule_flag & PFRULE_STATESLOPPY)
3587                 s->state_flags |= PFSTATE_SLOPPY;
3588         s->log = r->log & PF_LOG_ALL;
3589         if (nr != NULL)
3590                 s->log |= nr->log & PF_LOG_ALL;
3591         switch (pd->proto) {
3592         case IPPROTO_TCP:
3593                 s->src.seqlo = ntohl(th->th_seq);
3594                 s->src.seqhi = s->src.seqlo + pd->p_len + 1;
3595                 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
3596                     r->keep_state == PF_STATE_MODULATE) {
3597                         /* Generate sequence number modulator */
3598                         if ((s->src.seqdiff = pf_tcp_iss(pd) - s->src.seqlo) ==
3599                             0)
3600                                 s->src.seqdiff = 1;
3601                         pf_change_a(&th->th_seq, &th->th_sum,
3602                             htonl(s->src.seqlo + s->src.seqdiff), 0);
3603                         *rewrite = 1;
3604                 } else
3605                         s->src.seqdiff = 0;
3606                 if (th->th_flags & TH_SYN) {
3607                         s->src.seqhi++;
3608                         s->src.wscale = pf_get_wscale(m, off,
3609                             th->th_off, pd->af);
3610                 }
3611                 s->src.max_win = MAX(ntohs(th->th_win), 1);
3612                 if (s->src.wscale & PF_WSCALE_MASK) {
3613                         /* Remove scale factor from initial window */
3614                         int win = s->src.max_win;
3615                         win += 1 << (s->src.wscale & PF_WSCALE_MASK);
3616                         s->src.max_win = (win - 1) >>
3617                             (s->src.wscale & PF_WSCALE_MASK);
3618                 }
3619                 if (th->th_flags & TH_FIN)
3620                         s->src.seqhi++;
3621                 s->dst.seqhi = 1;
3622                 s->dst.max_win = 1;
3623                 s->src.state = TCPS_SYN_SENT;
3624                 s->dst.state = TCPS_CLOSED;
3625                 s->timeout = PFTM_TCP_FIRST_PACKET;
3626                 break;
3627         case IPPROTO_UDP:
3628                 s->src.state = PFUDPS_SINGLE;
3629                 s->dst.state = PFUDPS_NO_TRAFFIC;
3630                 s->timeout = PFTM_UDP_FIRST_PACKET;
3631                 break;
3632         case IPPROTO_ICMP:
3633 #ifdef INET6
3634         case IPPROTO_ICMPV6:
3635 #endif
3636                 s->timeout = PFTM_ICMP_FIRST_PACKET;
3637                 break;
3638         default:
3639                 s->src.state = PFOTHERS_SINGLE;
3640                 s->dst.state = PFOTHERS_NO_TRAFFIC;
3641                 s->timeout = PFTM_OTHER_FIRST_PACKET;
3642         }
3643
3644         s->creation = time_second;
3645         s->expire = time_second;
3646
3647         if (sn != NULL) {
3648                 s->src_node = sn;
3649                 s->src_node->states++;
3650         }
3651         if (nsn != NULL) {
3652                 /* XXX We only modify one side for now. */
3653                 PF_ACPY(&nsn->raddr, &nk->addr[1], pd->af);
3654                 s->nat_src_node = nsn;
3655                 s->nat_src_node->states++;
3656         }
3657         if (pd->proto == IPPROTO_TCP) {
3658                 if ((pd->flags & PFDESC_TCP_NORM) && pf_normalize_tcp_init(m,
3659                     off, pd, th, &s->src, &s->dst)) {
3660                         REASON_SET(&reason, PFRES_MEMORY);
3661                         pf_src_tree_remove_state(s);
3662                         STATE_DEC_COUNTERS(s);
3663                         pool_put(&pf_state_pl, s);
3664                         return (PF_DROP);
3665                 }
3666                 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
3667                     pf_normalize_tcp_stateful(m, off, pd, &reason, th, s,
3668                     &s->src, &s->dst, rewrite)) {
3669                         /* This really shouldn't happen!!! */
3670                         DPFPRINTF(PF_DEBUG_URGENT,
3671                             ("pf_normalize_tcp_stateful failed on first pkt"));
3672                         pf_normalize_tcp_cleanup(s);
3673                         pf_src_tree_remove_state(s);
3674                         STATE_DEC_COUNTERS(s);
3675                         pool_put(&pf_state_pl, s);
3676                         return (PF_DROP);
3677                 }
3678         }
3679         s->direction = pd->dir;
3680
3681         if (sk == NULL && pf_state_key_setup(pd, nr, &skw, &sks, &sk, &nk,
3682             pd->src, pd->dst, sport, dport))
3683                 goto csfailed;
3684
3685         if (pf_state_insert(BOUND_IFACE(r, kif), skw, sks, s)) {
3686                 if (pd->proto == IPPROTO_TCP)
3687                         pf_normalize_tcp_cleanup(s);
3688                 REASON_SET(&reason, PFRES_STATEINS);
3689                 pf_src_tree_remove_state(s);
3690                 STATE_DEC_COUNTERS(s);
3691                 pool_put(&pf_state_pl, s);
3692                 return (PF_DROP);
3693         } else
3694                 *sm = s;
3695
3696         pf_set_rt_ifp(s, pd->src);      /* needs s->state_key set */
3697         if (tag > 0) {
3698                 pf_tag_ref(tag);
3699                 s->tag = tag;
3700         }
3701         if (pd->proto == IPPROTO_TCP && (th->th_flags & (TH_SYN|TH_ACK)) ==
3702             TH_SYN && r->keep_state == PF_STATE_SYNPROXY) {
3703                 s->src.state = PF_TCPS_PROXY_SRC;
3704                 /* undo NAT changes, if they have taken place */
3705                 if (nr != NULL) {
3706                         struct pf_state_key *skt = s->key[PF_SK_WIRE];
3707                         if (pd->dir == PF_OUT)
3708                                 skt = s->key[PF_SK_STACK];
3709                         PF_ACPY(pd->src, &skt->addr[pd->sidx], pd->af);
3710                         PF_ACPY(pd->dst, &skt->addr[pd->didx], pd->af);
3711                         if (pd->sport)
3712                                 *pd->sport = skt->port[pd->sidx];
3713                         if (pd->dport)
3714                                 *pd->dport = skt->port[pd->didx];
3715                         if (pd->proto_sum)
3716                                 *pd->proto_sum = bproto_sum;
3717                         if (pd->ip_sum)
3718                                 *pd->ip_sum = bip_sum;
3719                         m_copyback(m, off, hdrlen, pd->hdr.any);
3720                 }
3721                 s->src.seqhi = htonl(karc4random());
3722                 /* Find mss option */
3723                 mss = pf_get_mss(m, off, th->th_off, pd->af);
3724                 mss = pf_calc_mss(pd->src, pd->af, mss);
3725                 mss = pf_calc_mss(pd->dst, pd->af, mss);
3726                 s->src.mss = mss;
3727                 pf_send_tcp(r, pd->af, pd->dst, pd->src, th->th_dport,
3728                     th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1,
3729                     TH_SYN|TH_ACK, 0, s->src.mss, 0, 1, 0, NULL, NULL);
3730                 REASON_SET(&reason, PFRES_SYNPROXY);
3731                 return (PF_SYNPROXY_DROP);
3732         }
3733
3734         return (PF_PASS);
3735
3736 csfailed:
3737         if (sk != NULL)
3738                 pool_put(&pf_state_key_pl, sk);
3739         if (nk != NULL)
3740                 pool_put(&pf_state_key_pl, nk);
3741
3742         if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3743                 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3744                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3745                 pf_status.src_nodes--;
3746                 pool_put(&pf_src_tree_pl, sn);
3747         }
3748         if (nsn != sn && nsn != NULL && nsn->states == 0 && nsn->expire == 0) {
3749                 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3750                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3751                 pf_status.src_nodes--;
3752                 pool_put(&pf_src_tree_pl, nsn);
3753         }
3754         return (PF_DROP);
3755 }
3756
3757 int
3758 pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif,