mbuf: Save linker layer, IP and TCP/UDP header length
[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 <sys/proc.h>
55 #include <sys/kthread.h>
56
57 #include <machine/inttypes.h>
58
59 #include <sys/md5.h>
60
61 #include <net/if.h>
62 #include <net/if_types.h>
63 #include <net/bpf.h>
64 #include <net/netisr.h>
65 #include <net/route.h>
66
67 #include <netinet/in.h>
68 #include <netinet/in_var.h>
69 #include <netinet/in_systm.h>
70 #include <netinet/ip.h>
71 #include <netinet/ip_var.h>
72 #include <netinet/tcp.h>
73 #include <netinet/tcp_seq.h>
74 #include <netinet/udp.h>
75 #include <netinet/ip_icmp.h>
76 #include <netinet/in_pcb.h>
77 #include <netinet/tcp_timer.h>
78 #include <netinet/tcp_var.h>
79 #include <netinet/udp_var.h>
80 #include <netinet/icmp_var.h>
81 #include <netinet/if_ether.h>
82
83 #include <net/pf/pfvar.h>
84 #include <net/pf/if_pflog.h>
85
86 #include <net/pf/if_pfsync.h>
87
88 #ifdef INET6
89 #include <netinet/ip6.h>
90 #include <netinet/in_pcb.h>
91 #include <netinet/icmp6.h>
92 #include <netinet6/nd6.h>
93 #include <netinet6/ip6_var.h>
94 #include <netinet6/in6_pcb.h>
95 #endif /* INET6 */
96
97 #include <sys/in_cksum.h>
98 #include <sys/ucred.h>
99 #include <machine/limits.h>
100 #include <sys/msgport2.h>
101 #include <net/netmsg2.h>
102
103 extern int ip_optcopy(struct ip *, struct ip *);
104 extern int debug_pfugidhack;
105
106 struct lwkt_token pf_token = LWKT_TOKEN_INITIALIZER(pf_token);
107
108 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) kprintf x
109
110 /*
111  * Global variables
112  */
113
114 /* mask radix tree */
115 struct radix_node_head  *pf_maskhead;
116
117 /* state tables */
118 struct pf_state_tree     pf_statetbl;
119
120 struct pf_altqqueue      pf_altqs[2];
121 struct pf_palist         pf_pabuf;
122 struct pf_altqqueue     *pf_altqs_active;
123 struct pf_altqqueue     *pf_altqs_inactive;
124 struct pf_status         pf_status;
125
126 u_int32_t                ticket_altqs_active;
127 u_int32_t                ticket_altqs_inactive;
128 int                      altqs_inactive_open;
129 u_int32_t                ticket_pabuf;
130
131 MD5_CTX                  pf_tcp_secret_ctx;
132 u_char                   pf_tcp_secret[16];
133 int                      pf_tcp_secret_init;
134 int                      pf_tcp_iss_off;
135
136 struct pf_anchor_stackframe {
137         struct pf_ruleset                       *rs;
138         struct pf_rule                          *r;
139         struct pf_anchor_node                   *parent;
140         struct pf_anchor                        *child;
141 } pf_anchor_stack[64];
142
143 struct malloc_type       *pf_src_tree_pl, *pf_rule_pl, *pf_pooladdr_pl;
144 struct malloc_type       *pf_state_pl, *pf_state_key_pl, *pf_state_item_pl;
145 struct malloc_type       *pf_altq_pl;
146
147 void                     pf_print_host(struct pf_addr *, u_int16_t, u_int8_t);
148
149 void                     pf_init_threshold(struct pf_threshold *, u_int32_t,
150                             u_int32_t);
151 void                     pf_add_threshold(struct pf_threshold *);
152 int                      pf_check_threshold(struct pf_threshold *);
153
154 void                     pf_change_ap(struct pf_addr *, u_int16_t *,
155                             u_int16_t *, u_int16_t *, struct pf_addr *,
156                             u_int16_t, u_int8_t, sa_family_t);
157 int                      pf_modulate_sack(struct mbuf *, int, struct pf_pdesc *,
158                             struct tcphdr *, struct pf_state_peer *);
159 #ifdef INET6
160 void                     pf_change_a6(struct pf_addr *, u_int16_t *,
161                             struct pf_addr *, u_int8_t);
162 #endif /* INET6 */
163 void                     pf_change_icmp(struct pf_addr *, u_int16_t *,
164                             struct pf_addr *, struct pf_addr *, u_int16_t,
165                             u_int16_t *, u_int16_t *, u_int16_t *,
166                             u_int16_t *, u_int8_t, sa_family_t);
167 void                     pf_send_tcp(const struct pf_rule *, sa_family_t,
168                             const struct pf_addr *, const struct pf_addr *,
169                             u_int16_t, u_int16_t, u_int32_t, u_int32_t,
170                             u_int8_t, u_int16_t, u_int16_t, u_int8_t, int,
171                             u_int16_t, struct ether_header *, struct ifnet *);
172 void                     pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t,
173                             sa_family_t, struct pf_rule *);
174 struct pf_rule          *pf_match_translation(struct pf_pdesc *, struct mbuf *,
175                             int, int, struct pfi_kif *,
176                             struct pf_addr *, u_int16_t, struct pf_addr *,
177                             u_int16_t, int);
178 struct pf_rule          *pf_get_translation(struct pf_pdesc *, struct mbuf *,
179                             int, int, struct pfi_kif *, struct pf_src_node **,
180                             struct pf_state_key **, struct pf_state_key **,
181                             struct pf_state_key **, struct pf_state_key **,
182                             struct pf_addr *, struct pf_addr *,
183                             u_int16_t, u_int16_t);
184 void                     pf_detach_state(struct pf_state *);
185 int                      pf_state_key_setup(struct pf_pdesc *, struct pf_rule *,
186                             struct pf_state_key **, struct pf_state_key **,
187                             struct pf_state_key **, struct pf_state_key **,
188                             struct pf_addr *, struct pf_addr *,
189                             u_int16_t, u_int16_t);
190 void                     pf_state_key_detach(struct pf_state *, int);
191 u_int32_t                pf_tcp_iss(struct pf_pdesc *);
192 int                      pf_test_rule(struct pf_rule **, struct pf_state **,
193                             int, struct pfi_kif *, struct mbuf *, int,
194                             void *, struct pf_pdesc *, struct pf_rule **,
195                             struct pf_ruleset **, struct ifqueue *, struct inpcb *);
196 static __inline int      pf_create_state(struct pf_rule *, struct pf_rule *,
197                             struct pf_rule *, struct pf_pdesc *,
198                             struct pf_src_node *, struct pf_state_key *,
199                             struct pf_state_key *, struct pf_state_key *,
200                             struct pf_state_key *, struct mbuf *, int,
201                             u_int16_t, u_int16_t, int *, struct pfi_kif *,
202                             struct pf_state **, int, u_int16_t, u_int16_t,
203                             int);
204 int                      pf_test_fragment(struct pf_rule **, int,
205                             struct pfi_kif *, struct mbuf *, void *,
206                             struct pf_pdesc *, struct pf_rule **,
207                             struct pf_ruleset **);
208 int                      pf_tcp_track_full(struct pf_state_peer *,
209                             struct pf_state_peer *, struct pf_state **,
210                             struct pfi_kif *, struct mbuf *, int,
211                             struct pf_pdesc *, u_short *, int *);
212 int                     pf_tcp_track_sloppy(struct pf_state_peer *,
213                             struct pf_state_peer *, struct pf_state **,
214                             struct pf_pdesc *, u_short *);
215 int                      pf_test_state_tcp(struct pf_state **, int,
216                             struct pfi_kif *, struct mbuf *, int,
217                             void *, struct pf_pdesc *, u_short *);
218 int                      pf_test_state_udp(struct pf_state **, int,
219                             struct pfi_kif *, struct mbuf *, int,
220                             void *, struct pf_pdesc *);
221 int                      pf_test_state_icmp(struct pf_state **, int,
222                             struct pfi_kif *, struct mbuf *, int,
223                             void *, struct pf_pdesc *, u_short *);
224 int                      pf_test_state_other(struct pf_state **, int,
225                             struct pfi_kif *, struct mbuf *, struct pf_pdesc *);
226 void                     pf_step_into_anchor(int *, struct pf_ruleset **, int,
227                             struct pf_rule **, struct pf_rule **, int *);
228 int                      pf_step_out_of_anchor(int *, struct pf_ruleset **,
229                              int, struct pf_rule **, struct pf_rule **,
230                              int *);
231 void                     pf_hash(struct pf_addr *, struct pf_addr *,
232                             struct pf_poolhashkey *, sa_family_t);
233 int                      pf_map_addr(u_int8_t, struct pf_rule *,
234                             struct pf_addr *, struct pf_addr *,
235                             struct pf_addr *, struct pf_src_node **);
236 int                      pf_get_sport(sa_family_t, u_int8_t, struct pf_rule *,
237                             struct pf_addr *, struct pf_addr *, u_int16_t,
238                             struct pf_addr *, u_int16_t*, u_int16_t, u_int16_t,
239                             struct pf_src_node **);
240 void                     pf_route(struct mbuf **, struct pf_rule *, int,
241                             struct ifnet *, struct pf_state *,
242                             struct pf_pdesc *);
243 void                     pf_route6(struct mbuf **, struct pf_rule *, int,
244                             struct ifnet *, struct pf_state *,
245                             struct pf_pdesc *);
246 u_int8_t                 pf_get_wscale(struct mbuf *, int, u_int16_t,
247                             sa_family_t);
248 u_int16_t                pf_get_mss(struct mbuf *, int, u_int16_t,
249                             sa_family_t);
250 u_int16_t                pf_calc_mss(struct pf_addr *, sa_family_t,
251                                 u_int16_t);
252 void                     pf_set_rt_ifp(struct pf_state *,
253                             struct pf_addr *);
254 int                      pf_check_proto_cksum(struct mbuf *, int, int,
255                             u_int8_t, sa_family_t);
256 struct pf_divert        *pf_get_divert(struct mbuf *);
257 void                     pf_print_state_parts(struct pf_state *,
258                             struct pf_state_key *, struct pf_state_key *);
259 int                      pf_addr_wrap_neq(struct pf_addr_wrap *,
260                             struct pf_addr_wrap *);
261 struct pf_state         *pf_find_state(struct pfi_kif *,
262                             struct pf_state_key_cmp *, u_int, struct mbuf *);
263 int                      pf_src_connlimit(struct pf_state **);
264 int                      pf_check_congestion(struct ifqueue *);
265
266 extern int pf_end_threads;
267
268 struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] = {
269         { &pf_state_pl, PFSTATE_HIWAT },
270         { &pf_src_tree_pl, PFSNODE_HIWAT },
271         { &pf_frent_pl, PFFRAG_FRENT_HIWAT },
272         { &pfr_ktable_pl, PFR_KTABLE_HIWAT },
273         { &pfr_kentry_pl, PFR_KENTRY_HIWAT }
274 };
275
276 #define STATE_LOOKUP(i, k, d, s, m)                                     \
277         do {                                                            \
278                 s = pf_find_state(i, k, d, m);                  \
279                 if (s == NULL || (s)->timeout == PFTM_PURGE)            \
280                         return (PF_DROP);                               \
281                 if (d == PF_OUT &&                                      \
282                     (((s)->rule.ptr->rt == PF_ROUTETO &&                \
283                     (s)->rule.ptr->direction == PF_OUT) ||              \
284                     ((s)->rule.ptr->rt == PF_REPLYTO &&                 \
285                     (s)->rule.ptr->direction == PF_IN)) &&              \
286                     (s)->rt_kif != NULL &&                              \
287                     (s)->rt_kif != i)                                   \
288                         return (PF_PASS);                               \
289         } while (0)
290
291 #define BOUND_IFACE(r, k) \
292         ((r)->rule_flag & PFRULE_IFBOUND) ? (k) : pfi_all
293
294 #define STATE_INC_COUNTERS(s)                           \
295         do {                                            \
296                 s->rule.ptr->states_cur++;              \
297                 s->rule.ptr->states_tot++;              \
298                 if (s->anchor.ptr != NULL) {            \
299                         s->anchor.ptr->states_cur++;    \
300                         s->anchor.ptr->states_tot++;    \
301                 }                                       \
302                 if (s->nat_rule.ptr != NULL) {          \
303                         s->nat_rule.ptr->states_cur++;  \
304                         s->nat_rule.ptr->states_tot++;  \
305                 }                                       \
306         } while (0)
307
308 #define STATE_DEC_COUNTERS(s)                           \
309         do {                                            \
310                 if (s->nat_rule.ptr != NULL)            \
311                         s->nat_rule.ptr->states_cur--;  \
312                 if (s->anchor.ptr != NULL)              \
313                         s->anchor.ptr->states_cur--;    \
314                 s->rule.ptr->states_cur--;              \
315         } while (0)
316
317 static MALLOC_DEFINE(M_PFSTATEPL, "pfstatepl", "pf state pool list");
318 static MALLOC_DEFINE(M_PFSRCTREEPL, "pfsrctpl", "pf source tree pool list");
319 static MALLOC_DEFINE(M_PFSTATEKEYPL, "pfstatekeypl", "pf state key pool list");
320 static MALLOC_DEFINE(M_PFSTATEITEMPL, "pfstateitempl", "pf state item pool list");
321
322 static __inline int pf_src_compare(struct pf_src_node *, struct pf_src_node *);
323 static __inline int pf_state_compare_key(struct pf_state_key *,
324         struct pf_state_key *);
325 static __inline int pf_state_compare_id(struct pf_state *,
326         struct pf_state *);
327
328 struct pf_src_tree tree_src_tracking;
329
330 struct pf_state_tree_id tree_id;
331 struct pf_state_queue state_list;
332
333 RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare);
334 RB_GENERATE(pf_state_tree, pf_state_key, entry, pf_state_compare_key);
335 RB_GENERATE(pf_state_tree_id, pf_state,
336     entry_id, pf_state_compare_id);
337
338 static __inline int
339 pf_src_compare(struct pf_src_node *a, struct pf_src_node *b)
340 {
341         int     diff;
342
343         if (a->rule.ptr > b->rule.ptr)
344                 return (1);
345         if (a->rule.ptr < b->rule.ptr)
346                 return (-1);
347         if ((diff = a->af - b->af) != 0)
348                 return (diff);
349         switch (a->af) {
350 #ifdef INET
351         case AF_INET:
352                 if (a->addr.addr32[0] > b->addr.addr32[0])
353                         return (1);
354                 if (a->addr.addr32[0] < b->addr.addr32[0])
355                         return (-1);
356                 break;
357 #endif /* INET */
358 #ifdef INET6
359         case AF_INET6:
360                 if (a->addr.addr32[3] > b->addr.addr32[3])
361                         return (1);
362                 if (a->addr.addr32[3] < b->addr.addr32[3])
363                         return (-1);
364                 if (a->addr.addr32[2] > b->addr.addr32[2])
365                         return (1);
366                 if (a->addr.addr32[2] < b->addr.addr32[2])
367                         return (-1);
368                 if (a->addr.addr32[1] > b->addr.addr32[1])
369                         return (1);
370                 if (a->addr.addr32[1] < b->addr.addr32[1])
371                         return (-1);
372                 if (a->addr.addr32[0] > b->addr.addr32[0])
373                         return (1);
374                 if (a->addr.addr32[0] < b->addr.addr32[0])
375                         return (-1);
376                 break;
377 #endif /* INET6 */
378         }
379         return (0);
380 }
381
382 u_int32_t
383 pf_state_hash(struct pf_state_key *sk)
384 {
385         u_int32_t hv = (u_int32_t)(((intptr_t)sk >> 6) ^ ((intptr_t)sk >> 15));
386         if (hv == 0)    /* disallow 0 */
387                 hv = 1;
388         return(hv);
389 }
390
391 #ifdef INET6
392 void
393 pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af)
394 {
395         switch (af) {
396 #ifdef INET
397         case AF_INET:
398                 dst->addr32[0] = src->addr32[0];
399                 break;
400 #endif /* INET */
401         case AF_INET6:
402                 dst->addr32[0] = src->addr32[0];
403                 dst->addr32[1] = src->addr32[1];
404                 dst->addr32[2] = src->addr32[2];
405                 dst->addr32[3] = src->addr32[3];
406                 break;
407         }
408 }
409 #endif /* INET6 */
410
411 void
412 pf_init_threshold(struct pf_threshold *threshold,
413     u_int32_t limit, u_int32_t seconds)
414 {
415         threshold->limit = limit * PF_THRESHOLD_MULT;
416         threshold->seconds = seconds;
417         threshold->count = 0;
418         threshold->last = time_second;
419 }
420
421 void
422 pf_add_threshold(struct pf_threshold *threshold)
423 {
424         u_int32_t t = time_second, diff = t - threshold->last;
425
426         if (diff >= threshold->seconds)
427                 threshold->count = 0;
428         else
429                 threshold->count -= threshold->count * diff /
430                     threshold->seconds;
431         threshold->count += PF_THRESHOLD_MULT;
432         threshold->last = t;
433 }
434
435 int
436 pf_check_threshold(struct pf_threshold *threshold)
437 {
438         return (threshold->count > threshold->limit);
439 }
440
441 int
442 pf_src_connlimit(struct pf_state **state)
443 {
444         int bad = 0;
445
446         (*state)->src_node->conn++;
447         (*state)->src.tcp_est = 1;
448         pf_add_threshold(&(*state)->src_node->conn_rate);
449
450         if ((*state)->rule.ptr->max_src_conn &&
451             (*state)->rule.ptr->max_src_conn <
452             (*state)->src_node->conn) {
453                 pf_status.lcounters[LCNT_SRCCONN]++;
454                 bad++;
455         }
456
457         if ((*state)->rule.ptr->max_src_conn_rate.limit &&
458             pf_check_threshold(&(*state)->src_node->conn_rate)) {
459                 pf_status.lcounters[LCNT_SRCCONNRATE]++;
460                 bad++;
461         }
462
463         if (!bad)
464                 return (0);
465
466         if ((*state)->rule.ptr->overload_tbl) {
467                 struct pfr_addr p;
468                 u_int32_t       killed = 0;
469
470                 pf_status.lcounters[LCNT_OVERLOAD_TABLE]++;
471                 if (pf_status.debug >= PF_DEBUG_MISC) {
472                         kprintf("pf_src_connlimit: blocking address ");
473                         pf_print_host(&(*state)->src_node->addr, 0,
474                             (*state)->key[PF_SK_WIRE]->af);
475                 }
476
477                 bzero(&p, sizeof(p));
478                 p.pfra_af = (*state)->key[PF_SK_WIRE]->af;
479                 switch ((*state)->key[PF_SK_WIRE]->af) {
480 #ifdef INET
481                 case AF_INET:
482                         p.pfra_net = 32;
483                         p.pfra_ip4addr = (*state)->src_node->addr.v4;
484                         break;
485 #endif /* INET */
486 #ifdef INET6
487                 case AF_INET6:
488                         p.pfra_net = 128;
489                         p.pfra_ip6addr = (*state)->src_node->addr.v6;
490                         break;
491 #endif /* INET6 */
492                 }
493
494                 pfr_insert_kentry((*state)->rule.ptr->overload_tbl,
495                     &p, time_second);
496
497                 /* kill existing states if that's required. */
498                 if ((*state)->rule.ptr->flush) {
499                         struct pf_state_key *sk;
500                         struct pf_state *st;
501
502                         pf_status.lcounters[LCNT_OVERLOAD_FLUSH]++;
503                         RB_FOREACH(st, pf_state_tree_id, &tree_id) {
504                                 sk = st->key[PF_SK_WIRE];
505                                 /*
506                                  * Kill states from this source.  (Only those
507                                  * from the same rule if PF_FLUSH_GLOBAL is not
508                                  * set)
509                                  */
510                                 if (sk->af ==
511                                     (*state)->key[PF_SK_WIRE]->af &&
512                                     (((*state)->direction == PF_OUT &&
513                                     PF_AEQ(&(*state)->src_node->addr,
514                                         &sk->addr[0], sk->af)) ||
515                                     ((*state)->direction == PF_IN &&
516                                     PF_AEQ(&(*state)->src_node->addr,
517                                         &sk->addr[1], sk->af))) &&
518                                     ((*state)->rule.ptr->flush &
519                                     PF_FLUSH_GLOBAL ||
520                                     (*state)->rule.ptr == st->rule.ptr)) {
521                                         st->timeout = PFTM_PURGE;
522                                         st->src.state = st->dst.state =
523                                             TCPS_CLOSED;
524                                         killed++;
525                                 }
526                         }
527                         if (pf_status.debug >= PF_DEBUG_MISC)
528                                 kprintf(", %u states killed", killed);
529                 }
530                 if (pf_status.debug >= PF_DEBUG_MISC)
531                         kprintf("\n");
532         }
533
534         /* kill this state */
535         (*state)->timeout = PFTM_PURGE;
536         (*state)->src.state = (*state)->dst.state = TCPS_CLOSED;
537         return (1);
538 }
539
540 int
541 pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule,
542     struct pf_addr *src, sa_family_t af)
543 {
544         struct pf_src_node      k;
545
546         if (*sn == NULL) {
547                 k.af = af;
548                 PF_ACPY(&k.addr, src, af);
549                 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
550                     rule->rpool.opts & PF_POOL_STICKYADDR)
551                         k.rule.ptr = rule;
552                 else
553                         k.rule.ptr = NULL;
554                 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
555                 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
556         }
557         if (*sn == NULL) {
558                 if (!rule->max_src_nodes ||
559                     rule->src_nodes < rule->max_src_nodes)
560                         (*sn) = kmalloc(sizeof(struct pf_src_node), M_PFSRCTREEPL, M_NOWAIT|M_ZERO);
561                 else
562                         pf_status.lcounters[LCNT_SRCNODES]++;
563                 if ((*sn) == NULL)
564                         return (-1);
565
566                 pf_init_threshold(&(*sn)->conn_rate,
567                     rule->max_src_conn_rate.limit,
568                     rule->max_src_conn_rate.seconds);
569
570                 (*sn)->af = af;
571                 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
572                     rule->rpool.opts & PF_POOL_STICKYADDR)
573                         (*sn)->rule.ptr = rule;
574                 else
575                         (*sn)->rule.ptr = NULL;
576                 PF_ACPY(&(*sn)->addr, src, af);
577                 if (RB_INSERT(pf_src_tree,
578                     &tree_src_tracking, *sn) != NULL) {
579                         if (pf_status.debug >= PF_DEBUG_MISC) {
580                                 kprintf("pf: src_tree insert failed: ");
581                                 pf_print_host(&(*sn)->addr, 0, af);
582                                 kprintf("\n");
583                         }
584                         kfree(*sn, M_PFSRCTREEPL);
585                         return (-1);
586                 }
587                 (*sn)->creation = time_second;
588                 (*sn)->ruletype = rule->action;
589                 if ((*sn)->rule.ptr != NULL)
590                         (*sn)->rule.ptr->src_nodes++;
591                 pf_status.scounters[SCNT_SRC_NODE_INSERT]++;
592                 pf_status.src_nodes++;
593         } else {
594                 if (rule->max_src_states &&
595                     (*sn)->states >= rule->max_src_states) {
596                         pf_status.lcounters[LCNT_SRCSTATES]++;
597                         return (-1);
598                 }
599         }
600         return (0);
601 }
602
603 /* state table stuff */
604
605 static __inline int
606 pf_state_compare_key(struct pf_state_key *a, struct pf_state_key *b)
607 {
608         int     diff;
609
610         if ((diff = a->proto - b->proto) != 0)
611                 return (diff);
612         if ((diff = a->af - b->af) != 0)
613                 return (diff);
614         switch (a->af) {
615 #ifdef INET
616         case AF_INET:
617                 if (a->addr[0].addr32[0] > b->addr[0].addr32[0])
618                         return (1);
619                 if (a->addr[0].addr32[0] < b->addr[0].addr32[0])
620                         return (-1);
621                 if (a->addr[1].addr32[0] > b->addr[1].addr32[0])
622                         return (1);
623                 if (a->addr[1].addr32[0] < b->addr[1].addr32[0])
624                         return (-1);
625                 break;
626 #endif /* INET */
627 #ifdef INET6
628         case AF_INET6:
629                 if (a->addr[0].addr32[3] > b->addr[0].addr32[3])
630                         return (1);
631                 if (a->addr[0].addr32[3] < b->addr[0].addr32[3])
632                         return (-1);
633                 if (a->addr[1].addr32[3] > b->addr[1].addr32[3])
634                         return (1);
635                 if (a->addr[1].addr32[3] < b->addr[1].addr32[3])
636                         return (-1);
637                 if (a->addr[0].addr32[2] > b->addr[0].addr32[2])
638                         return (1);
639                 if (a->addr[0].addr32[2] < b->addr[0].addr32[2])
640                         return (-1);
641                 if (a->addr[1].addr32[2] > b->addr[1].addr32[2])
642                         return (1);
643                 if (a->addr[1].addr32[2] < b->addr[1].addr32[2])
644                         return (-1);
645                 if (a->addr[0].addr32[1] > b->addr[0].addr32[1])
646                         return (1);
647                 if (a->addr[0].addr32[1] < b->addr[0].addr32[1])
648                         return (-1);
649                 if (a->addr[1].addr32[1] > b->addr[1].addr32[1])
650                         return (1);
651                 if (a->addr[1].addr32[1] < b->addr[1].addr32[1])
652                         return (-1);
653                 if (a->addr[0].addr32[0] > b->addr[0].addr32[0])
654                         return (1);
655                 if (a->addr[0].addr32[0] < b->addr[0].addr32[0])
656                         return (-1);
657                 if (a->addr[1].addr32[0] > b->addr[1].addr32[0])
658                         return (1);
659                 if (a->addr[1].addr32[0] < b->addr[1].addr32[0])
660                         return (-1);
661                 break;
662 #endif /* INET6 */
663         }
664
665         if ((diff = a->port[0] - b->port[0]) != 0)
666                 return (diff);
667         if ((diff = a->port[1] - b->port[1]) != 0)
668                 return (diff);
669
670         return (0);
671 }
672
673 static __inline int
674 pf_state_compare_id(struct pf_state *a, struct pf_state *b)
675 {
676         if (a->id > b->id)
677                 return (1);
678         if (a->id < b->id)
679                 return (-1);
680         if (a->creatorid > b->creatorid)
681                 return (1);
682         if (a->creatorid < b->creatorid)
683                 return (-1);
684
685         return (0);
686 }
687
688 int
689 pf_state_key_attach(struct pf_state_key *sk, struct pf_state *s, int idx)
690 {
691         struct pf_state_item    *si;
692         struct pf_state_key     *cur;
693
694         KKASSERT(s->key[idx] == NULL);  /* XXX handle this? */
695
696         if ((cur = RB_INSERT(pf_state_tree, &pf_statetbl, sk)) != NULL) {
697                 /* key exists. check for same kif, if none, add to key */
698                 TAILQ_FOREACH(si, &cur->states, entry)
699                         if (si->s->kif == s->kif &&
700                             si->s->direction == s->direction) {
701                                 if (pf_status.debug >= PF_DEBUG_MISC) {
702                                         kprintf(
703                                             "pf: %s key attach failed on %s: ",
704                                             (idx == PF_SK_WIRE) ?
705                                             "wire" : "stack",
706                                             s->kif->pfik_name);
707                                         pf_print_state_parts(s,
708                                             (idx == PF_SK_WIRE) ? sk : NULL,
709                                             (idx == PF_SK_STACK) ? sk : NULL);
710                                         kprintf("\n");
711                                 }
712                                 kfree(sk, M_PFSTATEKEYPL);
713                                 return (-1);    /* collision! */
714                         }
715                 kfree(sk, M_PFSTATEKEYPL);
716
717                 s->key[idx] = cur;
718         } else
719                 s->key[idx] = sk;
720
721         if ((si = kmalloc(sizeof(struct pf_state_item), M_PFSTATEITEMPL, M_NOWAIT)) == NULL) {
722                 pf_state_key_detach(s, idx);
723                 return (-1);
724         }
725         si->s = s;
726
727         /* list is sorted, if-bound states before floating */
728         if (s->kif == pfi_all)
729                 TAILQ_INSERT_TAIL(&s->key[idx]->states, si, entry);
730         else
731                 TAILQ_INSERT_HEAD(&s->key[idx]->states, si, entry);
732         return (0);
733 }
734
735 void
736 pf_detach_state(struct pf_state *s)
737 {
738         if (s->key[PF_SK_WIRE] == s->key[PF_SK_STACK])
739                 s->key[PF_SK_WIRE] = NULL;
740
741         if (s->key[PF_SK_STACK] != NULL)
742                 pf_state_key_detach(s, PF_SK_STACK);
743
744         if (s->key[PF_SK_WIRE] != NULL)
745                 pf_state_key_detach(s, PF_SK_WIRE);
746 }
747
748 void
749 pf_state_key_detach(struct pf_state *s, int idx)
750 {
751         struct pf_state_item    *si;
752         si = TAILQ_FIRST(&s->key[idx]->states);
753         while (si && si->s != s)
754             si = TAILQ_NEXT(si, entry);
755
756         if (si) {
757                 TAILQ_REMOVE(&s->key[idx]->states, si, entry);
758                 kfree(si, M_PFSTATEITEMPL);
759         }
760
761         if (TAILQ_EMPTY(&s->key[idx]->states)) {
762                 RB_REMOVE(pf_state_tree, &pf_statetbl, s->key[idx]);
763                 if (s->key[idx]->reverse)
764                         s->key[idx]->reverse->reverse = NULL;
765                 if (s->key[idx]->inp)
766                         s->key[idx]->inp->inp_pf_sk = NULL;
767                 kfree(s->key[idx], M_PFSTATEKEYPL);
768         }
769         s->key[idx] = NULL;
770 }
771
772 struct pf_state_key *
773 pf_alloc_state_key(int pool_flags)
774 {
775         struct pf_state_key     *sk;
776
777         if ((sk = kmalloc(sizeof(struct pf_state_key), M_PFSTATEKEYPL, pool_flags)) == NULL)
778                         return (NULL);
779         TAILQ_INIT(&sk->states);
780
781         return (sk);
782 }
783
784 int
785 pf_state_key_setup(struct pf_pdesc *pd, struct pf_rule *nr,
786         struct pf_state_key **skw, struct pf_state_key **sks,
787         struct pf_state_key **skp, struct pf_state_key **nkp,
788         struct pf_addr *saddr, struct pf_addr *daddr,
789         u_int16_t sport, u_int16_t dport)
790 {
791         KKASSERT((*skp == NULL && *nkp == NULL));
792
793         if ((*skp = pf_alloc_state_key(M_NOWAIT | M_ZERO)) == NULL)
794                 return (ENOMEM);
795
796         PF_ACPY(&(*skp)->addr[pd->sidx], saddr, pd->af);
797         PF_ACPY(&(*skp)->addr[pd->didx], daddr, pd->af);
798         (*skp)->port[pd->sidx] = sport;
799         (*skp)->port[pd->didx] = dport;
800         (*skp)->proto = pd->proto;
801         (*skp)->af = pd->af;
802
803         if (nr != NULL) {
804                 if ((*nkp = pf_alloc_state_key(M_NOWAIT | M_ZERO)) == NULL)
805                         return (ENOMEM); /* caller must handle cleanup */
806
807                 /* XXX maybe just bcopy and TAILQ_INIT(&(*nkp)->states) */
808                 PF_ACPY(&(*nkp)->addr[0], &(*skp)->addr[0], pd->af);
809                 PF_ACPY(&(*nkp)->addr[1], &(*skp)->addr[1], pd->af);
810                 (*nkp)->port[0] = (*skp)->port[0];
811                 (*nkp)->port[1] = (*skp)->port[1];
812                 (*nkp)->proto = pd->proto;
813                 (*nkp)->af = pd->af;
814         } else
815                 *nkp = *skp;
816
817         if (pd->dir == PF_IN) {
818                 *skw = *skp;
819                 *sks = *nkp;
820         } else {
821                 *sks = *skp;
822                 *skw = *nkp;
823         }
824         return (0);
825 }
826
827
828 int
829 pf_state_insert(struct pfi_kif *kif, struct pf_state_key *skw,
830     struct pf_state_key *sks, struct pf_state *s)
831 {
832         s->kif = kif;
833
834         if (skw == sks) {
835                 if (pf_state_key_attach(skw, s, PF_SK_WIRE))
836                         return (-1);
837                 s->key[PF_SK_STACK] = s->key[PF_SK_WIRE];
838         } else {
839                 if (pf_state_key_attach(skw, s, PF_SK_WIRE)) {
840                         kfree(sks, M_PFSTATEKEYPL);
841                         return (-1);
842                 }
843                 if (pf_state_key_attach(sks, s, PF_SK_STACK)) {
844                         pf_state_key_detach(s, PF_SK_WIRE);
845                         return (-1);
846                 }
847         }
848
849         if (s->id == 0 && s->creatorid == 0) {
850                 s->id = htobe64(pf_status.stateid++);
851                 s->creatorid = pf_status.hostid;
852         }
853
854         /*
855          * Calculate hash code for altq
856          */
857         s->hash = crc32(s->key[PF_SK_WIRE], sizeof(*sks));
858
859         if (RB_INSERT(pf_state_tree_id, &tree_id, s) != NULL) {
860                 if (pf_status.debug >= PF_DEBUG_MISC) {
861                         kprintf("pf: state insert failed: "
862                             "id: %016jx creatorid: %08x",
863                               (uintmax_t)be64toh(s->id), ntohl(s->creatorid));
864                         if (s->sync_flags & PFSTATE_FROMSYNC)
865                                 kprintf(" (from sync)");
866                         kprintf("\n");
867                 }
868                 pf_detach_state(s);
869                 return (-1);
870         }
871         TAILQ_INSERT_TAIL(&state_list, s, entry_list);
872         pf_status.fcounters[FCNT_STATE_INSERT]++;
873         pf_status.states++;
874         pfi_kif_ref(kif, PFI_KIF_REF_STATE);
875         pfsync_insert_state(s);
876         return (0);
877 }
878
879 struct pf_state *
880 pf_find_state_byid(struct pf_state_cmp *key)
881 {
882         pf_status.fcounters[FCNT_STATE_SEARCH]++;
883
884         return (RB_FIND(pf_state_tree_id, &tree_id, (struct pf_state *)key));
885 }
886
887 struct pf_state *
888 pf_find_state(struct pfi_kif *kif, struct pf_state_key_cmp *key, u_int dir,
889     struct mbuf *m)
890 {
891         struct pf_state_key     *sk;
892         struct pf_state_item    *si;
893
894         pf_status.fcounters[FCNT_STATE_SEARCH]++;
895
896         if (dir == PF_OUT && m->m_pkthdr.pf.statekey &&
897             ((struct pf_state_key *)m->m_pkthdr.pf.statekey)->reverse)
898                 sk = ((struct pf_state_key *)m->m_pkthdr.pf.statekey)->reverse;
899         else {
900                 if ((sk = RB_FIND(pf_state_tree, &pf_statetbl,
901                     (struct pf_state_key *)key)) == NULL)
902                         return (NULL);
903                 if (dir == PF_OUT && m->m_pkthdr.pf.statekey) {
904                         ((struct pf_state_key *)
905                             m->m_pkthdr.pf.statekey)->reverse = sk;
906                         sk->reverse = m->m_pkthdr.pf.statekey;
907                 }
908         }
909
910         if (dir == PF_OUT)
911                 m->m_pkthdr.pf.statekey = NULL;
912
913         /* list is sorted, if-bound states before floating ones */
914         TAILQ_FOREACH(si, &sk->states, entry)
915                 if ((si->s->kif == pfi_all || si->s->kif == kif) &&
916                     sk == (dir == PF_IN ? si->s->key[PF_SK_WIRE] :
917                     si->s->key[PF_SK_STACK]))
918                         return (si->s);
919
920         return (NULL);
921 }
922
923 struct pf_state *
924 pf_find_state_all(struct pf_state_key_cmp *key, u_int dir, int *more)
925 {
926         struct pf_state_key     *sk;
927         struct pf_state_item    *si, *ret = NULL;
928
929         pf_status.fcounters[FCNT_STATE_SEARCH]++;
930
931         sk = RB_FIND(pf_state_tree, &pf_statetbl, (struct pf_state_key *)key);
932
933         if (sk != NULL) {
934                 TAILQ_FOREACH(si, &sk->states, entry)
935                         if (dir == PF_INOUT ||
936                             (sk == (dir == PF_IN ? si->s->key[PF_SK_WIRE] :
937                             si->s->key[PF_SK_STACK]))) {
938                                 if (more == NULL)
939                                         return (si->s);
940
941                                 if (ret)
942                                         (*more)++;
943                                 else
944                                         ret = si;
945                         }
946         }
947         return (ret ? ret->s : NULL);
948 }
949
950 /* END state table stuff */
951
952
953 void
954 pf_purge_thread(void *v)
955 {
956         int nloops = 0;
957         int locked = 0;
958
959         lwkt_gettoken(&pf_token);
960         for (;;) {
961                 tsleep(pf_purge_thread, PWAIT, "pftm", 1 * hz);
962
963                 lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
964
965                 if (pf_end_threads) {
966                         pf_purge_expired_states(pf_status.states, 0);
967                         pf_purge_expired_fragments();
968                         pf_purge_expired_src_nodes(1);
969                         pf_end_threads++;
970
971                         lockmgr(&pf_consistency_lock, LK_RELEASE);
972                         wakeup(pf_purge_thread);
973                         kthread_exit();
974                 }
975                 crit_enter();
976
977                 /* process a fraction of the state table every second */
978                 if(!pf_purge_expired_states(1 + (pf_status.states
979                     / pf_default_rule.timeout[PFTM_INTERVAL]), 0)) {
980
981                         pf_purge_expired_states(1 + (pf_status.states
982                             / pf_default_rule.timeout[PFTM_INTERVAL]), 1);
983                 }
984
985                 /* purge other expired types every PFTM_INTERVAL seconds */
986                 if (++nloops >= pf_default_rule.timeout[PFTM_INTERVAL]) {
987                         pf_purge_expired_fragments();
988                         if (!pf_purge_expired_src_nodes(locked)) {
989                                 pf_purge_expired_src_nodes(1);
990                         }
991                         nloops = 0;
992                 }
993                 crit_exit();
994                 lockmgr(&pf_consistency_lock, LK_RELEASE);
995         }
996         lwkt_reltoken(&pf_token);
997 }
998
999 u_int32_t
1000 pf_state_expires(const struct pf_state *state)
1001 {
1002         u_int32_t       timeout;
1003         u_int32_t       start;
1004         u_int32_t       end;
1005         u_int32_t       states;
1006
1007         /* handle all PFTM_* > PFTM_MAX here */
1008         if (state->timeout == PFTM_PURGE)
1009                 return (time_second);
1010         if (state->timeout == PFTM_UNTIL_PACKET)
1011                 return (0);
1012         KKASSERT(state->timeout != PFTM_UNLINKED);
1013         KKASSERT(state->timeout < PFTM_MAX);
1014         timeout = state->rule.ptr->timeout[state->timeout];
1015         if (!timeout)
1016                 timeout = pf_default_rule.timeout[state->timeout];
1017         start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START];
1018         if (start) {
1019                 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END];
1020                 states = state->rule.ptr->states_cur;
1021         } else {
1022                 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START];
1023                 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END];
1024                 states = pf_status.states;
1025         }
1026         if (end && states > start && start < end) {
1027                 if (states < end)
1028                         return (state->expire + timeout * (end - states) /
1029                             (end - start));
1030                 else
1031                         return (time_second);
1032         }
1033         return (state->expire + timeout);
1034 }
1035
1036 int
1037 pf_purge_expired_src_nodes(int waslocked)
1038 {
1039          struct pf_src_node             *cur, *next;
1040          int                             locked = waslocked;
1041
1042          for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) {
1043                  next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur);
1044
1045                  if (cur->states <= 0 && cur->expire <= time_second) {
1046                          if (! locked) {
1047                                  lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
1048                                  next = RB_NEXT(pf_src_tree,
1049                                      &tree_src_tracking, cur);
1050                                  locked = 1;
1051                          }
1052                          if (cur->rule.ptr != NULL) {
1053                                  cur->rule.ptr->src_nodes--;
1054                                  if (cur->rule.ptr->states_cur <= 0 &&
1055                                      cur->rule.ptr->max_src_nodes <= 0)
1056                                          pf_rm_rule(NULL, cur->rule.ptr);
1057                          }
1058                          RB_REMOVE(pf_src_tree, &tree_src_tracking, cur);
1059                          pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
1060                          pf_status.src_nodes--;
1061                          kfree(cur, M_PFSRCTREEPL);
1062                  }
1063          }
1064
1065          if (locked && !waslocked)
1066                 lockmgr(&pf_consistency_lock, LK_RELEASE);
1067         return(1);
1068 }
1069
1070 void
1071 pf_src_tree_remove_state(struct pf_state *s)
1072 {
1073         u_int32_t timeout;
1074
1075         if (s->src_node != NULL) {
1076                 if (s->src.tcp_est)
1077                         --s->src_node->conn;
1078                 if (--s->src_node->states <= 0) {
1079                         timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1080                         if (!timeout)
1081                                 timeout =
1082                                     pf_default_rule.timeout[PFTM_SRC_NODE];
1083                         s->src_node->expire = time_second + timeout;
1084                 }
1085         }
1086         if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) {
1087                 if (--s->nat_src_node->states <= 0) {
1088                         timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1089                         if (!timeout)
1090                                 timeout =
1091                                     pf_default_rule.timeout[PFTM_SRC_NODE];
1092                         s->nat_src_node->expire = time_second + timeout;
1093                 }
1094         }
1095         s->src_node = s->nat_src_node = NULL;
1096 }
1097
1098 /* callers should be at crit_enter() */
1099 void
1100 pf_unlink_state(struct pf_state *cur)
1101 {
1102         if (cur->src.state == PF_TCPS_PROXY_DST) {
1103                 /* XXX wire key the right one? */
1104                 pf_send_tcp(cur->rule.ptr, cur->key[PF_SK_WIRE]->af,
1105                     &cur->key[PF_SK_WIRE]->addr[1],
1106                     &cur->key[PF_SK_WIRE]->addr[0],
1107                     cur->key[PF_SK_WIRE]->port[1],
1108                     cur->key[PF_SK_WIRE]->port[0],
1109                     cur->src.seqhi, cur->src.seqlo + 1,
1110                     TH_RST|TH_ACK, 0, 0, 0, 1, cur->tag, NULL, NULL);
1111         }
1112         RB_REMOVE(pf_state_tree_id, &tree_id, cur);
1113         if (cur->creatorid == pf_status.hostid)
1114                 pfsync_delete_state(cur);
1115         cur->timeout = PFTM_UNLINKED;
1116         pf_src_tree_remove_state(cur);
1117         pf_detach_state(cur);
1118 }
1119
1120 static struct pf_state  *purge_cur;
1121
1122 /* callers should be at crit_enter() and hold the
1123  * write_lock on pf_consistency_lock */
1124 void
1125 pf_free_state(struct pf_state *cur)
1126 {
1127         if (pfsyncif != NULL &&
1128             (pfsyncif->sc_bulk_send_next == cur ||
1129             pfsyncif->sc_bulk_terminator == cur))
1130                 return;
1131         KKASSERT(cur->timeout == PFTM_UNLINKED);
1132         if (--cur->rule.ptr->states_cur <= 0 &&
1133             cur->rule.ptr->src_nodes <= 0)
1134                 pf_rm_rule(NULL, cur->rule.ptr);
1135         if (cur->nat_rule.ptr != NULL)
1136                 if (--cur->nat_rule.ptr->states_cur <= 0 &&
1137                         cur->nat_rule.ptr->src_nodes <= 0)
1138                         pf_rm_rule(NULL, cur->nat_rule.ptr);
1139         if (cur->anchor.ptr != NULL)
1140                 if (--cur->anchor.ptr->states_cur <= 0)
1141                         pf_rm_rule(NULL, cur->anchor.ptr);
1142         pf_normalize_tcp_cleanup(cur);
1143         pfi_kif_unref(cur->kif, PFI_KIF_REF_STATE);
1144
1145         /*
1146          * We may be freeing pf_purge_expired_states()'s saved scan entry,
1147          * adjust it if necessary.
1148          */
1149         if (purge_cur == cur) {
1150                 kprintf("PURGE CONFLICT\n");
1151                 purge_cur = TAILQ_NEXT(purge_cur, entry_list);
1152         }
1153         TAILQ_REMOVE(&state_list, cur, entry_list);
1154         if (cur->tag)
1155                 pf_tag_unref(cur->tag);
1156         kfree(cur, M_PFSTATEPL);
1157         pf_status.fcounters[FCNT_STATE_REMOVALS]++;
1158         pf_status.states--;
1159 }
1160
1161 int
1162 pf_purge_expired_states(u_int32_t maxcheck, int waslocked)
1163 {
1164         struct pf_state         *cur;
1165         int                      locked = waslocked;
1166
1167         while (maxcheck--) {
1168                 /*
1169                  * Wrap to start of list when we hit the end
1170                  */
1171                 cur = purge_cur;
1172                 if (cur == NULL) {
1173                         cur = TAILQ_FIRST(&state_list);
1174                         if (cur == NULL)
1175                                 break;  /* list empty */
1176                 }
1177
1178                 /*
1179                  * Setup next (purge_cur) while we process this one.  If we block and
1180                  * something else deletes purge_cur, pf_free_state() will adjust it further
1181                  * ahead.
1182                  */
1183                 purge_cur = TAILQ_NEXT(cur, entry_list);
1184
1185                 if (cur->timeout == PFTM_UNLINKED) {
1186                         /* free unlinked state */
1187                         if (! locked) {
1188                                 lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
1189                                 locked = 1;
1190                         }
1191                         pf_free_state(cur);
1192                 } else if (pf_state_expires(cur) <= time_second) {
1193                         /* unlink and free expired state */
1194                         pf_unlink_state(cur);
1195                         if (! locked) {
1196                                 if (!lockmgr(&pf_consistency_lock, LK_EXCLUSIVE))
1197                                         return (0);
1198                                 locked = 1;
1199                         }
1200                         pf_free_state(cur);
1201                 }
1202         }
1203
1204         if (locked)
1205                 lockmgr(&pf_consistency_lock, LK_RELEASE);
1206         return (1);
1207 }
1208
1209 int
1210 pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw)
1211 {
1212         if (aw->type != PF_ADDR_TABLE)
1213                 return (0);
1214         if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL)
1215                 return (1);
1216         return (0);
1217 }
1218
1219 void
1220 pf_tbladdr_remove(struct pf_addr_wrap *aw)
1221 {
1222         if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL)
1223                 return;
1224         pfr_detach_table(aw->p.tbl);
1225         aw->p.tbl = NULL;
1226 }
1227
1228 void
1229 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
1230 {
1231         struct pfr_ktable *kt = aw->p.tbl;
1232
1233         if (aw->type != PF_ADDR_TABLE || kt == NULL)
1234                 return;
1235         if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
1236                 kt = kt->pfrkt_root;
1237         aw->p.tbl = NULL;
1238         aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
1239                 kt->pfrkt_cnt : -1;
1240 }
1241
1242 void
1243 pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af)
1244 {
1245         switch (af) {
1246 #ifdef INET
1247         case AF_INET: {
1248                 u_int32_t a = ntohl(addr->addr32[0]);
1249                 kprintf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255,
1250                     (a>>8)&255, a&255);
1251                 if (p) {
1252                         p = ntohs(p);
1253                         kprintf(":%u", p);
1254                 }
1255                 break;
1256         }
1257 #endif /* INET */
1258 #ifdef INET6
1259         case AF_INET6: {
1260                 u_int16_t b;
1261                 u_int8_t i, curstart = 255, curend = 0,
1262                     maxstart = 0, maxend = 0;
1263                 for (i = 0; i < 8; i++) {
1264                         if (!addr->addr16[i]) {
1265                                 if (curstart == 255)
1266                                         curstart = i;
1267                                 else
1268                                         curend = i;
1269                         } else {
1270                                 if (curstart) {
1271                                         if ((curend - curstart) >
1272                                             (maxend - maxstart)) {
1273                                                 maxstart = curstart;
1274                                                 maxend = curend;
1275                                                 curstart = 255;
1276                                         }
1277                                 }
1278                         }
1279                 }
1280                 for (i = 0; i < 8; i++) {
1281                         if (i >= maxstart && i <= maxend) {
1282                                 if (maxend != 7) {
1283                                         if (i == maxstart)
1284                                                 kprintf(":");
1285                                 } else {
1286                                         if (i == maxend)
1287                                                 kprintf(":");
1288                                 }
1289                         } else {
1290                                 b = ntohs(addr->addr16[i]);
1291                                 kprintf("%x", b);
1292                                 if (i < 7)
1293                                         kprintf(":");
1294                         }
1295                 }
1296                 if (p) {
1297                         p = ntohs(p);
1298                         kprintf("[%u]", p);
1299                 }
1300                 break;
1301         }
1302 #endif /* INET6 */
1303         }
1304 }
1305
1306 void
1307 pf_print_state(struct pf_state *s)
1308 {
1309         pf_print_state_parts(s, NULL, NULL);
1310 }
1311
1312 void
1313 pf_print_state_parts(struct pf_state *s,
1314     struct pf_state_key *skwp, struct pf_state_key *sksp)
1315 {
1316         struct pf_state_key *skw, *sks;
1317         u_int8_t proto, dir;
1318
1319         /* Do our best to fill these, but they're skipped if NULL */
1320         skw = skwp ? skwp : (s ? s->key[PF_SK_WIRE] : NULL);
1321         sks = sksp ? sksp : (s ? s->key[PF_SK_STACK] : NULL);
1322         proto = skw ? skw->proto : (sks ? sks->proto : 0);
1323         dir = s ? s->direction : 0;
1324
1325         switch (proto) {
1326         case IPPROTO_TCP:
1327                 kprintf("TCP ");
1328                 break;
1329         case IPPROTO_UDP:
1330                 kprintf("UDP ");
1331                 break;
1332         case IPPROTO_ICMP:
1333                 kprintf("ICMP ");
1334                 break;
1335         case IPPROTO_ICMPV6:
1336                 kprintf("ICMPV6 ");
1337                 break;
1338         default:
1339                 kprintf("%u ", skw->proto);
1340                 break;
1341         }
1342         switch (dir) {
1343         case PF_IN:
1344                 kprintf(" in");
1345                 break;
1346         case PF_OUT:
1347                 kprintf(" out");
1348                 break;
1349         }
1350         if (skw) {
1351                 kprintf(" wire: ");
1352                 pf_print_host(&skw->addr[0], skw->port[0], skw->af);
1353                 kprintf(" ");
1354                 pf_print_host(&skw->addr[1], skw->port[1], skw->af);
1355         }
1356         if (sks) {
1357                 kprintf(" stack: ");
1358                 if (sks != skw) {
1359                         pf_print_host(&sks->addr[0], sks->port[0], sks->af);
1360                         kprintf(" ");
1361                         pf_print_host(&sks->addr[1], sks->port[1], sks->af);
1362                 } else
1363                         kprintf("-");
1364         }
1365         if (s) {
1366                 if (proto == IPPROTO_TCP) {
1367                         kprintf(" [lo=%u high=%u win=%u modulator=%u",
1368                             s->src.seqlo, s->src.seqhi,
1369                             s->src.max_win, s->src.seqdiff);
1370                         if (s->src.wscale && s->dst.wscale)
1371                                 kprintf(" wscale=%u",
1372                                     s->src.wscale & PF_WSCALE_MASK);
1373                         kprintf("]");
1374                         kprintf(" [lo=%u high=%u win=%u modulator=%u",
1375                             s->dst.seqlo, s->dst.seqhi,
1376                             s->dst.max_win, s->dst.seqdiff);
1377                         if (s->src.wscale && s->dst.wscale)
1378                                 kprintf(" wscale=%u",
1379                                 s->dst.wscale & PF_WSCALE_MASK);
1380                         kprintf("]");
1381                 }
1382                 kprintf(" %u:%u", s->src.state, s->dst.state);
1383         }
1384 }
1385
1386 void
1387 pf_print_flags(u_int8_t f)
1388 {
1389         if (f)
1390                 kprintf(" ");
1391         if (f & TH_FIN)
1392                 kprintf("F");
1393         if (f & TH_SYN)
1394                 kprintf("S");
1395         if (f & TH_RST)
1396                 kprintf("R");
1397         if (f & TH_PUSH)
1398                 kprintf("P");
1399         if (f & TH_ACK)
1400                 kprintf("A");
1401         if (f & TH_URG)
1402                 kprintf("U");
1403         if (f & TH_ECE)
1404                 kprintf("E");
1405         if (f & TH_CWR)
1406                 kprintf("W");
1407 }
1408
1409 #define PF_SET_SKIP_STEPS(i)                                    \
1410         do {                                                    \
1411                 while (head[i] != cur) {                        \
1412                         head[i]->skip[i].ptr = cur;             \
1413                         head[i] = TAILQ_NEXT(head[i], entries); \
1414                 }                                               \
1415         } while (0)
1416
1417 void
1418 pf_calc_skip_steps(struct pf_rulequeue *rules)
1419 {
1420         struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT];
1421         int i;
1422
1423         cur = TAILQ_FIRST(rules);
1424         prev = cur;
1425         for (i = 0; i < PF_SKIP_COUNT; ++i)
1426                 head[i] = cur;
1427         while (cur != NULL) {
1428
1429                 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
1430                         PF_SET_SKIP_STEPS(PF_SKIP_IFP);
1431                 if (cur->direction != prev->direction)
1432                         PF_SET_SKIP_STEPS(PF_SKIP_DIR);
1433                 if (cur->af != prev->af)
1434                         PF_SET_SKIP_STEPS(PF_SKIP_AF);
1435                 if (cur->proto != prev->proto)
1436                         PF_SET_SKIP_STEPS(PF_SKIP_PROTO);
1437                 if (cur->src.neg != prev->src.neg ||
1438                     pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr))
1439                         PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR);
1440                 if (cur->src.port[0] != prev->src.port[0] ||
1441                     cur->src.port[1] != prev->src.port[1] ||
1442                     cur->src.port_op != prev->src.port_op)
1443                         PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
1444                 if (cur->dst.neg != prev->dst.neg ||
1445                     pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr))
1446                         PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR);
1447                 if (cur->dst.port[0] != prev->dst.port[0] ||
1448                     cur->dst.port[1] != prev->dst.port[1] ||
1449                     cur->dst.port_op != prev->dst.port_op)
1450                         PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
1451
1452                 prev = cur;
1453                 cur = TAILQ_NEXT(cur, entries);
1454         }
1455         for (i = 0; i < PF_SKIP_COUNT; ++i)
1456                 PF_SET_SKIP_STEPS(i);
1457 }
1458
1459 int
1460 pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2)
1461 {
1462         if (aw1->type != aw2->type)
1463                 return (1);
1464         switch (aw1->type) {
1465         case PF_ADDR_ADDRMASK:
1466         case PF_ADDR_RANGE:
1467                 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0))
1468                         return (1);
1469                 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0))
1470                         return (1);
1471                 return (0);
1472         case PF_ADDR_DYNIFTL:
1473                 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt);
1474         case PF_ADDR_NOROUTE:
1475         case PF_ADDR_URPFFAILED:
1476                 return (0);
1477         case PF_ADDR_TABLE:
1478                 return (aw1->p.tbl != aw2->p.tbl);
1479         case PF_ADDR_RTLABEL:
1480                 return (aw1->v.rtlabel != aw2->v.rtlabel);
1481         default:
1482                 kprintf("invalid address type: %d\n", aw1->type);
1483                 return (1);
1484         }
1485 }
1486
1487 u_int16_t
1488 pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp)
1489 {
1490         u_int32_t       l;
1491
1492         if (udp && !cksum)
1493                 return (0x0000);
1494         l = cksum + old - new;
1495         l = (l >> 16) + (l & 65535);
1496         l = l & 65535;
1497         if (udp && !l)
1498                 return (0xFFFF);
1499         return (l);
1500 }
1501
1502 void
1503 pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc,
1504     struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af)
1505 {
1506         struct pf_addr  ao;
1507         u_int16_t       po = *p;
1508
1509         PF_ACPY(&ao, a, af);
1510         PF_ACPY(a, an, af);
1511
1512         *p = pn;
1513
1514         switch (af) {
1515 #ifdef INET
1516         case AF_INET:
1517                 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1518                     ao.addr16[0], an->addr16[0], 0),
1519                     ao.addr16[1], an->addr16[1], 0);
1520                 *p = pn;
1521                 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1522                     ao.addr16[0], an->addr16[0], u),
1523                     ao.addr16[1], an->addr16[1], u),
1524                     po, pn, u);
1525                 break;
1526 #endif /* INET */
1527 #ifdef INET6
1528         case AF_INET6:
1529                 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1530                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1531                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1532                     ao.addr16[0], an->addr16[0], u),
1533                     ao.addr16[1], an->addr16[1], u),
1534                     ao.addr16[2], an->addr16[2], u),
1535                     ao.addr16[3], an->addr16[3], u),
1536                     ao.addr16[4], an->addr16[4], u),
1537                     ao.addr16[5], an->addr16[5], u),
1538                     ao.addr16[6], an->addr16[6], u),
1539                     ao.addr16[7], an->addr16[7], u),
1540                     po, pn, u);
1541                 break;
1542 #endif /* INET6 */
1543         }
1544 }
1545
1546
1547 /* Changes a u_int32_t.  Uses a void * so there are no align restrictions */
1548 void
1549 pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u)
1550 {
1551         u_int32_t       ao;
1552
1553         memcpy(&ao, a, sizeof(ao));
1554         memcpy(a, &an, sizeof(u_int32_t));
1555         *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u),
1556             ao % 65536, an % 65536, u);
1557 }
1558
1559 #ifdef INET6
1560 void
1561 pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u)
1562 {
1563         struct pf_addr  ao;
1564
1565         PF_ACPY(&ao, a, AF_INET6);
1566         PF_ACPY(a, an, AF_INET6);
1567
1568         *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1569             pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1570             pf_cksum_fixup(pf_cksum_fixup(*c,
1571             ao.addr16[0], an->addr16[0], u),
1572             ao.addr16[1], an->addr16[1], u),
1573             ao.addr16[2], an->addr16[2], u),
1574             ao.addr16[3], an->addr16[3], u),
1575             ao.addr16[4], an->addr16[4], u),
1576             ao.addr16[5], an->addr16[5], u),
1577             ao.addr16[6], an->addr16[6], u),
1578             ao.addr16[7], an->addr16[7], u);
1579 }
1580 #endif /* INET6 */
1581
1582 void
1583 pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa,
1584     struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c,
1585     u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af)
1586 {
1587         struct pf_addr  oia, ooa;
1588
1589         PF_ACPY(&oia, ia, af);
1590         if (oa)
1591                 PF_ACPY(&ooa, oa, af);
1592
1593         /* Change inner protocol port, fix inner protocol checksum. */
1594         if (ip != NULL) {
1595                 u_int16_t       oip = *ip;
1596                 u_int32_t       opc = 0;
1597
1598                 if (pc != NULL)
1599                         opc = *pc;
1600                 *ip = np;
1601                 if (pc != NULL)
1602                         *pc = pf_cksum_fixup(*pc, oip, *ip, u);
1603                 *ic = pf_cksum_fixup(*ic, oip, *ip, 0);
1604                 if (pc != NULL)
1605                         *ic = pf_cksum_fixup(*ic, opc, *pc, 0);
1606         }
1607         /* Change inner ip address, fix inner ip and icmp checksums. */
1608         PF_ACPY(ia, na, af);
1609         switch (af) {
1610 #ifdef INET
1611         case AF_INET: {
1612                 u_int32_t        oh2c = *h2c;
1613
1614                 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c,
1615                     oia.addr16[0], ia->addr16[0], 0),
1616                     oia.addr16[1], ia->addr16[1], 0);
1617                 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1618                     oia.addr16[0], ia->addr16[0], 0),
1619                     oia.addr16[1], ia->addr16[1], 0);
1620                 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0);
1621                 break;
1622         }
1623 #endif /* INET */
1624 #ifdef INET6
1625         case AF_INET6:
1626                 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1627                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1628                     pf_cksum_fixup(pf_cksum_fixup(*ic,
1629                     oia.addr16[0], ia->addr16[0], u),
1630                     oia.addr16[1], ia->addr16[1], u),
1631                     oia.addr16[2], ia->addr16[2], u),
1632                     oia.addr16[3], ia->addr16[3], u),
1633                     oia.addr16[4], ia->addr16[4], u),
1634                     oia.addr16[5], ia->addr16[5], u),
1635                     oia.addr16[6], ia->addr16[6], u),
1636                     oia.addr16[7], ia->addr16[7], u);
1637                 break;
1638 #endif /* INET6 */
1639         }
1640         /* Outer ip address, fix outer ip or icmpv6 checksum, if necessary. */
1641         if (oa) {
1642                 PF_ACPY(oa, na, af);
1643                 switch (af) {
1644 #ifdef INET
1645                 case AF_INET:
1646                         *hc = pf_cksum_fixup(pf_cksum_fixup(*hc,
1647                             ooa.addr16[0], oa->addr16[0], 0),
1648                             ooa.addr16[1], oa->addr16[1], 0);
1649                         break;
1650 #endif /* INET */
1651 #ifdef INET6
1652                 case AF_INET6:
1653                         *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1654                             pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1655                             pf_cksum_fixup(pf_cksum_fixup(*ic,
1656                             ooa.addr16[0], oa->addr16[0], u),
1657                             ooa.addr16[1], oa->addr16[1], u),
1658                             ooa.addr16[2], oa->addr16[2], u),
1659                             ooa.addr16[3], oa->addr16[3], u),
1660                             ooa.addr16[4], oa->addr16[4], u),
1661                             ooa.addr16[5], oa->addr16[5], u),
1662                             ooa.addr16[6], oa->addr16[6], u),
1663                             ooa.addr16[7], oa->addr16[7], u);
1664                         break;
1665 #endif /* INET6 */
1666                 }
1667         }
1668 }
1669
1670
1671 /*
1672  * Need to modulate the sequence numbers in the TCP SACK option
1673  * (credits to Krzysztof Pfaff for report and patch)
1674  */
1675 int
1676 pf_modulate_sack(struct mbuf *m, int off, struct pf_pdesc *pd,
1677     struct tcphdr *th, struct pf_state_peer *dst)
1678 {
1679         int hlen = (th->th_off << 2) - sizeof(*th), thoptlen = hlen;
1680         u_int8_t opts[TCP_MAXOLEN], *opt = opts;
1681         int copyback = 0, i, olen;
1682         struct raw_sackblock sack;
1683
1684 #define TCPOLEN_SACKLEN (TCPOLEN_SACK + 2)
1685         if (hlen < TCPOLEN_SACKLEN ||
1686             !pf_pull_hdr(m, off + sizeof(*th), opts, hlen, NULL, NULL, pd->af))
1687                 return 0;
1688
1689         while (hlen >= TCPOLEN_SACKLEN) {
1690                 olen = opt[1];
1691                 switch (*opt) {
1692                 case TCPOPT_EOL:        /* FALLTHROUGH */
1693                 case TCPOPT_NOP:
1694                         opt++;
1695                         hlen--;
1696                         break;
1697                 case TCPOPT_SACK:
1698                         if (olen > hlen)
1699                                 olen = hlen;
1700                         if (olen >= TCPOLEN_SACKLEN) {
1701                                 for (i = 2; i + TCPOLEN_SACK <= olen;
1702                                     i += TCPOLEN_SACK) {
1703                                         memcpy(&sack, &opt[i], sizeof(sack));
1704                                         pf_change_a(&sack.rblk_start, &th->th_sum,
1705                                             htonl(ntohl(sack.rblk_start) -
1706                                             dst->seqdiff), 0);
1707                                         pf_change_a(&sack.rblk_end, &th->th_sum,
1708                                             htonl(ntohl(sack.rblk_end) -
1709                                             dst->seqdiff), 0);
1710                                         memcpy(&opt[i], &sack, sizeof(sack));
1711                                 }
1712                                 copyback = 1;
1713                         }
1714                         /* FALLTHROUGH */
1715                 default:
1716                         if (olen < 2)
1717                                 olen = 2;
1718                         hlen -= olen;
1719                         opt += olen;
1720                 }
1721         }
1722
1723         if (copyback)
1724                 m_copyback(m, off + sizeof(*th), thoptlen, opts);
1725         return (copyback);
1726 }
1727
1728 void
1729 pf_send_tcp(const struct pf_rule *r, sa_family_t af,
1730     const struct pf_addr *saddr, const struct pf_addr *daddr,
1731     u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack,
1732     u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl, int tag,
1733     u_int16_t rtag, struct ether_header *eh, struct ifnet *ifp)
1734 {
1735         struct mbuf     *m;
1736         int              len = 0, tlen;
1737 #ifdef INET
1738         struct ip       *h = NULL;
1739 #endif /* INET */
1740 #ifdef INET6
1741         struct ip6_hdr  *h6 = NULL;
1742 #endif /* INET6 */
1743         struct tcphdr   *th = NULL;
1744         char            *opt;
1745
1746         ASSERT_LWKT_TOKEN_HELD(&pf_token);
1747
1748         /* maximum segment size tcp option */
1749         tlen = sizeof(struct tcphdr);
1750         if (mss)
1751                 tlen += 4;
1752
1753         switch (af) {
1754 #ifdef INET
1755         case AF_INET:
1756                 len = sizeof(struct ip) + tlen;
1757                 break;
1758 #endif /* INET */
1759 #ifdef INET6
1760         case AF_INET6:
1761                 len = sizeof(struct ip6_hdr) + tlen;
1762                 break;
1763 #endif /* INET6 */
1764         }
1765
1766         /*
1767          * Create outgoing mbuf.
1768          *
1769          * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags,
1770          * so make sure pf.flags is clear.
1771          */
1772         m = m_gethdr(MB_DONTWAIT, MT_HEADER);
1773         if (m == NULL) {
1774                 return;
1775         }
1776         if (tag)
1777                 m->m_pkthdr.fw_flags |= PF_MBUF_TAGGED;
1778         m->m_pkthdr.pf.flags = 0;
1779         m->m_pkthdr.pf.tag = rtag;
1780         /* XXX Recheck when upgrading to > 4.4 */
1781         m->m_pkthdr.pf.statekey = NULL;
1782         if (r != NULL && r->rtableid >= 0)
1783                 m->m_pkthdr.pf.rtableid = r->rtableid;
1784
1785 #ifdef ALTQ
1786         if (r != NULL && r->qid) {
1787                 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
1788                 m->m_pkthdr.pf.qid = r->qid;
1789                 m->m_pkthdr.pf.ecn_af = af;
1790                 m->m_pkthdr.pf.hdr = mtod(m, struct ip *);
1791         }
1792 #endif /* ALTQ */
1793         m->m_data += max_linkhdr;
1794         m->m_pkthdr.len = m->m_len = len;
1795         m->m_pkthdr.rcvif = NULL;
1796         bzero(m->m_data, len);
1797         switch (af) {
1798 #ifdef INET
1799         case AF_INET:
1800                 h = mtod(m, struct ip *);
1801
1802                 /* IP header fields included in the TCP checksum */
1803                 h->ip_p = IPPROTO_TCP;
1804                 h->ip_len = tlen;
1805                 h->ip_src.s_addr = saddr->v4.s_addr;
1806                 h->ip_dst.s_addr = daddr->v4.s_addr;
1807
1808                 th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip));
1809                 break;
1810 #endif /* INET */
1811 #ifdef INET6
1812         case AF_INET6:
1813                 h6 = mtod(m, struct ip6_hdr *);
1814
1815                 /* IP header fields included in the TCP checksum */
1816                 h6->ip6_nxt = IPPROTO_TCP;
1817                 h6->ip6_plen = htons(tlen);
1818                 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr));
1819                 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr));
1820
1821                 th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr));
1822                 break;
1823 #endif /* INET6 */
1824         }
1825
1826         /* TCP header */
1827         th->th_sport = sport;
1828         th->th_dport = dport;
1829         th->th_seq = htonl(seq);
1830         th->th_ack = htonl(ack);
1831         th->th_off = tlen >> 2;
1832         th->th_flags = flags;
1833         th->th_win = htons(win);
1834
1835         if (mss) {
1836                 opt = (char *)(th + 1);
1837                 opt[0] = TCPOPT_MAXSEG;
1838                 opt[1] = 4;
1839                 mss = htons(mss);
1840                 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2);
1841         }
1842
1843         switch (af) {
1844 #ifdef INET
1845         case AF_INET:
1846                 /* TCP checksum */
1847                 th->th_sum = in_cksum(m, len);
1848
1849                 /* Finish the IP header */
1850                 h->ip_v = 4;
1851                 h->ip_hl = sizeof(*h) >> 2;
1852                 h->ip_tos = IPTOS_LOWDELAY;
1853                 h->ip_len = len;
1854                 h->ip_off = path_mtu_discovery ? IP_DF : 0;
1855                 h->ip_ttl = ttl ? ttl : ip_defttl;
1856                 h->ip_sum = 0;
1857                 if (eh == NULL) {
1858                         lwkt_reltoken(&pf_token);
1859                         ip_output(m, NULL, NULL, 0, NULL, NULL);
1860                         lwkt_gettoken(&pf_token);
1861                 } else {
1862                         struct route             ro;
1863                         struct rtentry           rt;
1864                         struct ether_header     *e = (void *)ro.ro_dst.sa_data;
1865
1866                         if (ifp == NULL) {
1867                                 m_freem(m);
1868                                 return;
1869                         }
1870                         rt.rt_ifp = ifp;
1871                         ro.ro_rt = &rt;
1872                         ro.ro_dst.sa_len = sizeof(ro.ro_dst);
1873                         ro.ro_dst.sa_family = pseudo_AF_HDRCMPLT;
1874                         bcopy(eh->ether_dhost, e->ether_shost, ETHER_ADDR_LEN);
1875                         bcopy(eh->ether_shost, e->ether_dhost, ETHER_ADDR_LEN);
1876                         e->ether_type = eh->ether_type;
1877                         /* XXX_IMPORT: later */
1878                         lwkt_reltoken(&pf_token);
1879                         ip_output(m, NULL, &ro, 0, NULL, NULL);
1880                         lwkt_gettoken(&pf_token);
1881                 }
1882                 break;
1883 #endif /* INET */
1884 #ifdef INET6
1885         case AF_INET6:
1886                 /* TCP checksum */
1887                 th->th_sum = in6_cksum(m, IPPROTO_TCP,
1888                     sizeof(struct ip6_hdr), tlen);
1889
1890                 h6->ip6_vfc |= IPV6_VERSION;
1891                 h6->ip6_hlim = IPV6_DEFHLIM;
1892
1893                 lwkt_reltoken(&pf_token);
1894                 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
1895                 lwkt_gettoken(&pf_token);
1896                 break;
1897 #endif /* INET6 */
1898         }
1899 }
1900
1901 void
1902 pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af,
1903     struct pf_rule *r)
1904 {
1905         struct mbuf     *m0;
1906
1907         /*
1908          * DragonFly doesn't zero the auxillary pkghdr fields, only fw_flags,
1909          * so make sure pf.flags is clear.
1910          */
1911         if ((m0 = m_copy(m, 0, M_COPYALL)) == NULL)
1912                 return;
1913
1914         m0->m_pkthdr.fw_flags |= PF_MBUF_TAGGED;
1915         m0->m_pkthdr.pf.flags = 0;
1916         /* XXX Re-Check when Upgrading to > 4.4 */
1917         m0->m_pkthdr.pf.statekey = NULL;
1918
1919         if (r->rtableid >= 0)
1920                 m0->m_pkthdr.pf.rtableid = r->rtableid;
1921
1922 #ifdef ALTQ
1923         if (r->qid) {
1924                 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
1925                 m0->m_pkthdr.pf.qid = r->qid;
1926                 m0->m_pkthdr.pf.ecn_af = af;
1927                 m0->m_pkthdr.pf.hdr = mtod(m0, struct ip *);
1928         }
1929 #endif /* ALTQ */
1930
1931         switch (af) {
1932 #ifdef INET
1933         case AF_INET:
1934                 icmp_error(m0, type, code, 0, 0);
1935                 break;
1936 #endif /* INET */
1937 #ifdef INET6
1938         case AF_INET6:
1939                 icmp6_error(m0, type, code, 0);
1940                 break;
1941 #endif /* INET6 */
1942         }
1943 }
1944
1945 /*
1946  * Return 1 if the addresses a and b match (with mask m), otherwise return 0.
1947  * If n is 0, they match if they are equal. If n is != 0, they match if they
1948  * are different.
1949  */
1950 int
1951 pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m,
1952     struct pf_addr *b, sa_family_t af)
1953 {
1954         int     match = 0;
1955
1956         switch (af) {
1957 #ifdef INET
1958         case AF_INET:
1959                 if ((a->addr32[0] & m->addr32[0]) ==
1960                     (b->addr32[0] & m->addr32[0]))
1961                         match++;
1962                 break;
1963 #endif /* INET */
1964 #ifdef INET6
1965         case AF_INET6:
1966                 if (((a->addr32[0] & m->addr32[0]) ==
1967                      (b->addr32[0] & m->addr32[0])) &&
1968                     ((a->addr32[1] & m->addr32[1]) ==
1969                      (b->addr32[1] & m->addr32[1])) &&
1970                     ((a->addr32[2] & m->addr32[2]) ==
1971                      (b->addr32[2] & m->addr32[2])) &&
1972                     ((a->addr32[3] & m->addr32[3]) ==
1973                      (b->addr32[3] & m->addr32[3])))
1974                         match++;
1975                 break;
1976 #endif /* INET6 */
1977         }
1978         if (match) {
1979                 if (n)
1980                         return (0);
1981                 else
1982                         return (1);
1983         } else {
1984                 if (n)
1985                         return (1);
1986                 else
1987                         return (0);
1988         }
1989 }
1990
1991 /*
1992  * Return 1 if b <= a <= e, otherwise return 0.
1993  */
1994 int
1995 pf_match_addr_range(struct pf_addr *b, struct pf_addr *e,
1996     struct pf_addr *a, sa_family_t af)
1997 {
1998         switch (af) {
1999 #ifdef INET
2000         case AF_INET:
2001                 if ((a->addr32[0] < b->addr32[0]) ||
2002                     (a->addr32[0] > e->addr32[0]))
2003                         return (0);
2004                 break;
2005 #endif /* INET */
2006 #ifdef INET6
2007         case AF_INET6: {
2008                 int     i;
2009
2010                 /* check a >= b */
2011                 for (i = 0; i < 4; ++i)
2012                         if (a->addr32[i] > b->addr32[i])
2013                                 break;
2014                         else if (a->addr32[i] < b->addr32[i])
2015                                 return (0);
2016                 /* check a <= e */
2017                 for (i = 0; i < 4; ++i)
2018                         if (a->addr32[i] < e->addr32[i])
2019                                 break;
2020                         else if (a->addr32[i] > e->addr32[i])
2021                                 return (0);
2022                 break;
2023         }
2024 #endif /* INET6 */
2025         }
2026         return (1);
2027 }
2028
2029 int
2030 pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p)
2031 {
2032         switch (op) {
2033         case PF_OP_IRG:
2034                 return ((p > a1) && (p < a2));
2035         case PF_OP_XRG:
2036                 return ((p < a1) || (p > a2));
2037         case PF_OP_RRG:
2038                 return ((p >= a1) && (p <= a2));
2039         case PF_OP_EQ:
2040                 return (p == a1);
2041         case PF_OP_NE:
2042                 return (p != a1);
2043         case PF_OP_LT:
2044                 return (p < a1);
2045         case PF_OP_LE:
2046                 return (p <= a1);
2047         case PF_OP_GT:
2048                 return (p > a1);
2049         case PF_OP_GE:
2050                 return (p >= a1);
2051         }
2052         return (0); /* never reached */
2053 }
2054
2055 int
2056 pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p)
2057 {
2058         a1 = ntohs(a1);
2059         a2 = ntohs(a2);
2060         p = ntohs(p);
2061         return (pf_match(op, a1, a2, p));
2062 }
2063
2064 int
2065 pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u)
2066 {
2067         if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
2068                 return (0);
2069         return (pf_match(op, a1, a2, u));
2070 }
2071
2072 int
2073 pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g)
2074 {
2075         if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
2076                 return (0);
2077         return (pf_match(op, a1, a2, g));
2078 }
2079
2080 int
2081 pf_match_tag(struct mbuf *m, struct pf_rule *r, int *tag)
2082 {
2083         if (*tag == -1)
2084                 *tag = m->m_pkthdr.pf.tag;
2085
2086         return ((!r->match_tag_not && r->match_tag == *tag) ||
2087             (r->match_tag_not && r->match_tag != *tag));
2088 }
2089
2090 int
2091 pf_tag_packet(struct mbuf *m, int tag, int rtableid)
2092 {
2093         if (tag <= 0 && rtableid < 0)
2094                 return (0);
2095
2096         if (tag > 0)
2097                 m->m_pkthdr.pf.tag = tag;
2098         if (rtableid >= 0)
2099                 m->m_pkthdr.pf.rtableid = rtableid;
2100
2101         return (0);
2102 }
2103
2104 void
2105 pf_step_into_anchor(int *depth, struct pf_ruleset **rs, int n,
2106     struct pf_rule **r, struct pf_rule **a, int *match)
2107 {
2108         struct pf_anchor_stackframe     *f;
2109
2110         (*r)->anchor->match = 0;
2111         if (match)
2112                 *match = 0;
2113         if (*depth >= NELEM(pf_anchor_stack)) {
2114                 kprintf("pf_step_into_anchor: stack overflow\n");
2115                 *r = TAILQ_NEXT(*r, entries);
2116                 return;
2117         } else if (*depth == 0 && a != NULL)
2118                 *a = *r;
2119         f = pf_anchor_stack + (*depth)++;
2120         f->rs = *rs;
2121         f->r = *r;
2122         if ((*r)->anchor_wildcard) {
2123                 f->parent = &(*r)->anchor->children;
2124                 if ((f->child = RB_MIN(pf_anchor_node, f->parent)) ==
2125                     NULL) {
2126                         *r = NULL;
2127                         return;
2128                 }
2129                 *rs = &f->child->ruleset;
2130         } else {
2131                 f->parent = NULL;
2132                 f->child = NULL;
2133                 *rs = &(*r)->anchor->ruleset;
2134         }
2135         *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
2136 }
2137
2138 int
2139 pf_step_out_of_anchor(int *depth, struct pf_ruleset **rs, int n,
2140     struct pf_rule **r, struct pf_rule **a, int *match)
2141 {
2142         struct pf_anchor_stackframe     *f;
2143         int quick = 0;
2144
2145         do {
2146                 if (*depth <= 0)
2147                         break;
2148                 f = pf_anchor_stack + *depth - 1;
2149                 if (f->parent != NULL && f->child != NULL) {
2150                         if (f->child->match ||
2151                             (match != NULL && *match)) {
2152                                 f->r->anchor->match = 1;
2153                                 *match = 0;
2154                         }
2155                         f->child = RB_NEXT(pf_anchor_node, f->parent, f->child);
2156                         if (f->child != NULL) {
2157                                 *rs = &f->child->ruleset;
2158                                 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
2159                                 if (*r == NULL)
2160                                         continue;
2161                                 else
2162                                         break;
2163                         }
2164                 }
2165                 (*depth)--;
2166                 if (*depth == 0 && a != NULL)
2167                         *a = NULL;
2168                 *rs = f->rs;
2169                 if (f->r->anchor->match || (match != NULL && *match))
2170                         quick = f->r->quick;
2171                 *r = TAILQ_NEXT(f->r, entries);
2172         } while (*r == NULL);
2173
2174         return (quick);
2175 }
2176
2177 #ifdef INET6
2178 void
2179 pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr,
2180     struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af)
2181 {
2182         switch (af) {
2183 #ifdef INET
2184         case AF_INET:
2185                 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2186                 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2187                 break;
2188 #endif /* INET */
2189         case AF_INET6:
2190                 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2191                 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2192                 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) |
2193                 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]);
2194                 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) |
2195                 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]);
2196                 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) |
2197                 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]);
2198                 break;
2199         }
2200 }
2201
2202 void
2203 pf_addr_inc(struct pf_addr *addr, sa_family_t af)
2204 {
2205         switch (af) {
2206 #ifdef INET
2207         case AF_INET:
2208                 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1);
2209                 break;
2210 #endif /* INET */
2211         case AF_INET6:
2212                 if (addr->addr32[3] == 0xffffffff) {
2213                         addr->addr32[3] = 0;
2214                         if (addr->addr32[2] == 0xffffffff) {
2215                                 addr->addr32[2] = 0;
2216                                 if (addr->addr32[1] == 0xffffffff) {
2217                                         addr->addr32[1] = 0;
2218                                         addr->addr32[0] =
2219                                             htonl(ntohl(addr->addr32[0]) + 1);
2220                                 } else
2221                                         addr->addr32[1] =
2222                                             htonl(ntohl(addr->addr32[1]) + 1);
2223                         } else
2224                                 addr->addr32[2] =
2225                                     htonl(ntohl(addr->addr32[2]) + 1);
2226                 } else
2227                         addr->addr32[3] =
2228                             htonl(ntohl(addr->addr32[3]) + 1);
2229                 break;
2230         }
2231 }
2232 #endif /* INET6 */
2233
2234 #define mix(a,b,c) \
2235         do {                                    \
2236                 a -= b; a -= c; a ^= (c >> 13); \
2237                 b -= c; b -= a; b ^= (a << 8);  \
2238                 c -= a; c -= b; c ^= (b >> 13); \
2239                 a -= b; a -= c; a ^= (c >> 12); \
2240                 b -= c; b -= a; b ^= (a << 16); \
2241                 c -= a; c -= b; c ^= (b >> 5);  \
2242                 a -= b; a -= c; a ^= (c >> 3);  \
2243                 b -= c; b -= a; b ^= (a << 10); \
2244                 c -= a; c -= b; c ^= (b >> 15); \
2245         } while (0)
2246
2247 /*
2248  * hash function based on bridge_hash in if_bridge.c
2249  */
2250 void
2251 pf_hash(struct pf_addr *inaddr, struct pf_addr *hash,
2252     struct pf_poolhashkey *key, sa_family_t af)
2253 {
2254         u_int32_t       a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0];
2255
2256         switch (af) {
2257 #ifdef INET
2258         case AF_INET:
2259                 a += inaddr->addr32[0];
2260                 b += key->key32[1];
2261                 mix(a, b, c);
2262                 hash->addr32[0] = c + key->key32[2];
2263                 break;
2264 #endif /* INET */
2265 #ifdef INET6
2266         case AF_INET6:
2267                 a += inaddr->addr32[0];
2268                 b += inaddr->addr32[2];
2269                 mix(a, b, c);
2270                 hash->addr32[0] = c;
2271                 a += inaddr->addr32[1];
2272                 b += inaddr->addr32[3];
2273                 c += key->key32[1];
2274                 mix(a, b, c);
2275                 hash->addr32[1] = c;
2276                 a += inaddr->addr32[2];
2277                 b += inaddr->addr32[1];
2278                 c += key->key32[2];
2279                 mix(a, b, c);
2280                 hash->addr32[2] = c;
2281                 a += inaddr->addr32[3];
2282                 b += inaddr->addr32[0];
2283                 c += key->key32[3];
2284                 mix(a, b, c);
2285                 hash->addr32[3] = c;
2286                 break;
2287 #endif /* INET6 */
2288         }
2289 }
2290
2291 int
2292 pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr,
2293     struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn)
2294 {
2295         unsigned char            hash[16];
2296         struct pf_pool          *rpool = &r->rpool;
2297         struct pf_addr          *raddr = &rpool->cur->addr.v.a.addr;
2298         struct pf_addr          *rmask = &rpool->cur->addr.v.a.mask;
2299         struct pf_pooladdr      *acur = rpool->cur;
2300         struct pf_src_node       k;
2301
2302         if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR &&
2303             (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2304                 k.af = af;
2305                 PF_ACPY(&k.addr, saddr, af);
2306                 if (r->rule_flag & PFRULE_RULESRCTRACK ||
2307                     r->rpool.opts & PF_POOL_STICKYADDR)
2308                         k.rule.ptr = r;
2309                 else
2310                         k.rule.ptr = NULL;
2311                 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
2312                 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
2313                 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) {
2314                         PF_ACPY(naddr, &(*sn)->raddr, af);
2315                         if (pf_status.debug >= PF_DEBUG_MISC) {
2316                                 kprintf("pf_map_addr: src tracking maps ");
2317                                 pf_print_host(&k.addr, 0, af);
2318                                 kprintf(" to ");
2319                                 pf_print_host(naddr, 0, af);
2320                                 kprintf("\n");
2321                         }
2322                         return (0);
2323                 }
2324         }
2325
2326         if (rpool->cur->addr.type == PF_ADDR_NOROUTE)
2327                 return (1);
2328         if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2329                 switch (af) {
2330 #ifdef INET
2331                 case AF_INET:
2332                         if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 &&
2333                             (rpool->opts & PF_POOL_TYPEMASK) !=
2334                             PF_POOL_ROUNDROBIN)
2335                                 return (1);
2336                         raddr = &rpool->cur->addr.p.dyn->pfid_addr4;
2337                         rmask = &rpool->cur->addr.p.dyn->pfid_mask4;
2338                         break;
2339 #endif /* INET */
2340 #ifdef INET6
2341                 case AF_INET6:
2342                         if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 &&
2343                             (rpool->opts & PF_POOL_TYPEMASK) !=
2344                             PF_POOL_ROUNDROBIN)
2345                                 return (1);
2346                         raddr = &rpool->cur->addr.p.dyn->pfid_addr6;
2347                         rmask = &rpool->cur->addr.p.dyn->pfid_mask6;
2348                         break;
2349 #endif /* INET6 */
2350                 }
2351         } else if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2352                 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN)
2353                         return (1); /* unsupported */
2354         } else {
2355                 raddr = &rpool->cur->addr.v.a.addr;
2356                 rmask = &rpool->cur->addr.v.a.mask;
2357         }
2358
2359         switch (rpool->opts & PF_POOL_TYPEMASK) {
2360         case PF_POOL_NONE:
2361                 PF_ACPY(naddr, raddr, af);
2362                 break;
2363         case PF_POOL_BITMASK:
2364                 PF_POOLMASK(naddr, raddr, rmask, saddr, af);
2365                 break;
2366         case PF_POOL_RANDOM:
2367                 if (init_addr != NULL && PF_AZERO(init_addr, af)) {
2368                         switch (af) {
2369 #ifdef INET
2370                         case AF_INET:
2371                                 rpool->counter.addr32[0] = htonl(karc4random());
2372                                 break;
2373 #endif /* INET */
2374 #ifdef INET6
2375                         case AF_INET6:
2376                                 if (rmask->addr32[3] != 0xffffffff)
2377                                         rpool->counter.addr32[3] =
2378                                             htonl(karc4random());
2379                                 else
2380                                         break;
2381                                 if (rmask->addr32[2] != 0xffffffff)
2382                                         rpool->counter.addr32[2] =
2383                                             htonl(karc4random());
2384                                 else
2385                                         break;
2386                                 if (rmask->addr32[1] != 0xffffffff)
2387                                         rpool->counter.addr32[1] =
2388                                             htonl(karc4random());
2389                                 else
2390                                         break;
2391                                 if (rmask->addr32[0] != 0xffffffff)
2392                                         rpool->counter.addr32[0] =
2393                                             htonl(karc4random());
2394                                 break;
2395 #endif /* INET6 */
2396                         }
2397                         PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2398                         PF_ACPY(init_addr, naddr, af);
2399
2400                 } else {
2401                         PF_AINC(&rpool->counter, af);
2402                         PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2403                 }
2404                 break;
2405         case PF_POOL_SRCHASH:
2406                 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af);
2407                 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af);
2408                 break;
2409         case PF_POOL_ROUNDROBIN:
2410                 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2411                         if (!pfr_pool_get(rpool->cur->addr.p.tbl,
2412                             &rpool->tblidx, &rpool->counter,
2413                             &raddr, &rmask, af))
2414                                 goto get_addr;
2415                 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2416                         if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2417                             &rpool->tblidx, &rpool->counter,
2418                             &raddr, &rmask, af))
2419                                 goto get_addr;
2420                 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af))
2421                         goto get_addr;
2422
2423         try_next:
2424                 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL)
2425                         rpool->cur = TAILQ_FIRST(&rpool->list);
2426                 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2427                         rpool->tblidx = -1;
2428                         if (pfr_pool_get(rpool->cur->addr.p.tbl,
2429                             &rpool->tblidx, &rpool->counter,
2430                             &raddr, &rmask, af)) {
2431                                 /* table contains no address of type 'af' */
2432                                 if (rpool->cur != acur)
2433                                         goto try_next;
2434                                 return (1);
2435                         }
2436                 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2437                         rpool->tblidx = -1;
2438                         if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2439                             &rpool->tblidx, &rpool->counter,
2440                             &raddr, &rmask, af)) {
2441                                 /* table contains no address of type 'af' */
2442                                 if (rpool->cur != acur)
2443                                         goto try_next;
2444                                 return (1);
2445                         }
2446                 } else {
2447                         raddr = &rpool->cur->addr.v.a.addr;
2448                         rmask = &rpool->cur->addr.v.a.mask;
2449                         PF_ACPY(&rpool->counter, raddr, af);
2450                 }
2451
2452         get_addr:
2453                 PF_ACPY(naddr, &rpool->counter, af);
2454                 if (init_addr != NULL && PF_AZERO(init_addr, af))
2455                         PF_ACPY(init_addr, naddr, af);
2456                 PF_AINC(&rpool->counter, af);
2457                 break;
2458         }
2459         if (*sn != NULL)
2460                 PF_ACPY(&(*sn)->raddr, naddr, af);
2461
2462         if (pf_status.debug >= PF_DEBUG_MISC &&
2463             (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2464                 kprintf("pf_map_addr: selected address ");
2465                 pf_print_host(naddr, 0, af);
2466                 kprintf("\n");
2467         }
2468
2469         return (0);
2470 }
2471
2472 int
2473 pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r,
2474     struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport,
2475     struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high,
2476     struct pf_src_node **sn)
2477 {
2478         struct pf_state_key_cmp key;
2479         struct pf_addr          init_addr;
2480         u_int16_t               cut;
2481
2482         bzero(&init_addr, sizeof(init_addr));
2483         if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2484                 return (1);
2485
2486         if (proto == IPPROTO_ICMP) {
2487                 low = 1;
2488                 high = 65535;
2489         }
2490
2491         do {
2492                 key.af = af;
2493                 key.proto = proto;
2494                 PF_ACPY(&key.addr[1], daddr, key.af);
2495                 PF_ACPY(&key.addr[0], naddr, key.af);
2496                 key.port[1] = dport;
2497
2498                 /*
2499                  * port search; start random, step;
2500                  * similar 2 portloop in in_pcbbind
2501                  */
2502                 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP ||
2503                     proto == IPPROTO_ICMP)) {
2504                         key.port[0] = dport;
2505                         if (pf_find_state_all(&key, PF_IN, NULL) == NULL)
2506                                 return (0);
2507                 } else if (low == 0 && high == 0) {
2508                         key.port[0] = *nport;
2509                         if (pf_find_state_all(&key, PF_IN, NULL) == NULL)
2510                                 return (0);
2511                 } else if (low == high) {
2512                         key.port[0] = htons(low);
2513                         if (pf_find_state_all(&key, PF_IN, NULL) == NULL) {
2514                                 *nport = htons(low);
2515                                 return (0);
2516                         }
2517                 } else {
2518                         u_int16_t tmp;
2519
2520                         if (low > high) {
2521                                 tmp = low;
2522                                 low = high;
2523                                 high = tmp;
2524                         }
2525                         /* low < high */
2526                         cut = htonl(karc4random()) % (1 + high - low) + low;
2527                         /* low <= cut <= high */
2528                         for (tmp = cut; tmp <= high; ++(tmp)) {
2529                                 key.port[0] = htons(tmp);
2530                                 if (pf_find_state_all(&key, PF_IN, NULL) ==
2531                                     NULL && !in_baddynamic(tmp, proto)) {
2532                                         *nport = htons(tmp);
2533                                         return (0);
2534                                 }
2535                         }
2536                         for (tmp = cut - 1; tmp >= low; --(tmp)) {
2537                                 key.port[0] = htons(tmp);
2538                                 if (pf_find_state_all(&key, PF_IN, NULL) ==
2539                                     NULL && !in_baddynamic(tmp, proto)) {
2540                                         *nport = htons(tmp);
2541                                         return (0);
2542                                 }
2543                         }
2544                 }
2545
2546                 switch (r->rpool.opts & PF_POOL_TYPEMASK) {
2547                 case PF_POOL_RANDOM:
2548                 case PF_POOL_ROUNDROBIN:
2549                         if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2550                                 return (1);
2551                         break;
2552                 case PF_POOL_NONE:
2553                 case PF_POOL_SRCHASH:
2554                 case PF_POOL_BITMASK:
2555                 default:
2556                         return (1);
2557                 }
2558         } while (! PF_AEQ(&init_addr, naddr, af) );
2559         return (1);                                     /* none available */
2560 }
2561
2562 struct pf_rule *
2563 pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off,
2564     int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport,
2565     struct pf_addr *daddr, u_int16_t dport, int rs_num)
2566 {
2567         struct pf_rule          *r, *rm = NULL;
2568         struct pf_ruleset       *ruleset = NULL;
2569         int                      tag = -1;
2570         int                      rtableid = -1;
2571         int                      asd = 0;
2572
2573         r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr);
2574         while (r && rm == NULL) {
2575                 struct pf_rule_addr     *src = NULL, *dst = NULL;
2576                 struct pf_addr_wrap     *xdst = NULL;
2577
2578                 if (r->action == PF_BINAT && direction == PF_IN) {
2579                         src = &r->dst;
2580                         if (r->rpool.cur != NULL)
2581                                 xdst = &r->rpool.cur->addr;
2582                 } else {
2583                         src = &r->src;
2584                         dst = &r->dst;
2585                 }
2586
2587                 r->evaluations++;
2588                 if (pfi_kif_match(r->kif, kif) == r->ifnot)
2589                         r = r->skip[PF_SKIP_IFP].ptr;
2590                 else if (r->direction && r->direction != direction)
2591                         r = r->skip[PF_SKIP_DIR].ptr;
2592                 else if (r->af && r->af != pd->af)
2593                         r = r->skip[PF_SKIP_AF].ptr;
2594                 else if (r->proto && r->proto != pd->proto)
2595                         r = r->skip[PF_SKIP_PROTO].ptr;
2596                 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af,
2597                     src->neg, kif))
2598                         r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR :
2599                             PF_SKIP_DST_ADDR].ptr;
2600                 else if (src->port_op && !pf_match_port(src->port_op,
2601                     src->port[0], src->port[1], sport))
2602                         r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT :
2603                             PF_SKIP_DST_PORT].ptr;
2604                 else if (dst != NULL &&
2605                     PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->neg, NULL))
2606                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
2607                 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af,
2608                     0, NULL))
2609                         r = TAILQ_NEXT(r, entries);
2610                 else if (dst != NULL && dst->port_op &&
2611                     !pf_match_port(dst->port_op, dst->port[0],
2612                     dst->port[1], dport))
2613                         r = r->skip[PF_SKIP_DST_PORT].ptr;
2614                 else if (r->match_tag && !pf_match_tag(m, r, &tag))
2615                         r = TAILQ_NEXT(r, entries);
2616                 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto !=
2617                     IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m,
2618                     off, pd->hdr.tcp), r->os_fingerprint)))
2619                         r = TAILQ_NEXT(r, entries);
2620                 else {
2621                         if (r->tag)
2622                                 tag = r->tag;
2623                         if (r->rtableid >= 0)
2624                                 rtableid = r->rtableid;
2625                         if (r->anchor == NULL) {
2626                                 rm = r;
2627                         } else
2628                                 pf_step_into_anchor(&asd, &ruleset, rs_num,
2629                                     &r, NULL, NULL);
2630                 }
2631                 if (r == NULL)
2632                         pf_step_out_of_anchor(&asd, &ruleset, rs_num, &r,
2633                             NULL, NULL);
2634         }
2635         if (pf_tag_packet(m, tag, rtableid))
2636                 return (NULL);
2637         if (rm != NULL && (rm->action == PF_NONAT ||
2638             rm->action == PF_NORDR || rm->action == PF_NOBINAT))
2639                 return (NULL);
2640         return (rm);
2641 }
2642
2643 struct pf_rule *
2644 pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction,
2645     struct pfi_kif *kif, struct pf_src_node **sn,
2646     struct pf_state_key **skw, struct pf_state_key **sks,
2647     struct pf_state_key **skp, struct pf_state_key **nkp,
2648     struct pf_addr *saddr, struct pf_addr *daddr,
2649     u_int16_t sport, u_int16_t dport)
2650 {
2651         struct pf_rule  *r = NULL;
2652
2653
2654         if (direction == PF_OUT) {
2655                 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2656                     sport, daddr, dport, PF_RULESET_BINAT);
2657                 if (r == NULL)
2658                         r = pf_match_translation(pd, m, off, direction, kif,
2659                             saddr, sport, daddr, dport, PF_RULESET_NAT);
2660         } else {
2661                 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2662                     sport, daddr, dport, PF_RULESET_RDR);
2663                 if (r == NULL)
2664                         r = pf_match_translation(pd, m, off, direction, kif,
2665                             saddr, sport, daddr, dport, PF_RULESET_BINAT);
2666         }
2667
2668         if (r != NULL) {
2669                 struct pf_addr  *naddr;
2670                 u_int16_t       *nport;
2671
2672                 if (pf_state_key_setup(pd, r, skw, sks, skp, nkp,
2673                     saddr, daddr, sport, dport))
2674                         return r;
2675
2676                 /* XXX We only modify one side for now. */
2677                 naddr = &(*nkp)->addr[1];
2678                 nport = &(*nkp)->port[1];
2679
2680                 /*
2681                  * NOTE: Currently all translations will clear
2682                  *       BRIDGE_MBUF_TAGGED, telling the bridge to
2683                  *       ignore the original input encapsulation.
2684                  */
2685                 switch (r->action) {
2686                 case PF_NONAT:
2687                 case PF_NOBINAT:
2688                 case PF_NORDR:
2689                         return (NULL);
2690                 case PF_NAT:
2691                         m->m_pkthdr.fw_flags &= ~BRIDGE_MBUF_TAGGED;
2692                         if (pf_get_sport(pd->af, pd->proto, r, saddr,
2693                             daddr, dport, naddr, nport, r->rpool.proxy_port[0],
2694                             r->rpool.proxy_port[1], sn)) {
2695                                 DPFPRINTF(PF_DEBUG_MISC,
2696                                     ("pf: NAT proxy port allocation "
2697                                     "(%u-%u) failed\n",
2698                                     r->rpool.proxy_port[0],
2699                                     r->rpool.proxy_port[1]));
2700                                 return (NULL);
2701                         }
2702                         break;
2703                 case PF_BINAT:
2704                         m->m_pkthdr.fw_flags &= ~BRIDGE_MBUF_TAGGED;
2705                         switch (direction) {
2706                         case PF_OUT:
2707                                 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){
2708                                         switch (pd->af) {
2709 #ifdef INET
2710                                         case AF_INET:
2711                                                 if (r->rpool.cur->addr.p.dyn->
2712                                                     pfid_acnt4 < 1)
2713                                                         return (NULL);
2714                                                 PF_POOLMASK(naddr,
2715                                                     &r->rpool.cur->addr.p.dyn->
2716                                                     pfid_addr4,
2717                                                     &r->rpool.cur->addr.p.dyn->
2718                                                     pfid_mask4,
2719                                                     saddr, AF_INET);
2720                                                 break;
2721 #endif /* INET */
2722 #ifdef INET6
2723                                         case AF_INET6:
2724                                                 if (r->rpool.cur->addr.p.dyn->
2725                                                     pfid_acnt6 < 1)
2726                                                         return (NULL);
2727                                                 PF_POOLMASK(naddr,
2728                                                     &r->rpool.cur->addr.p.dyn->
2729                                                     pfid_addr6,
2730                                                     &r->rpool.cur->addr.p.dyn->
2731                                                     pfid_mask6,
2732                                                     saddr, AF_INET6);
2733                                                 break;
2734 #endif /* INET6 */
2735                                         }
2736                                 } else
2737                                         PF_POOLMASK(naddr,
2738                                             &r->rpool.cur->addr.v.a.addr,
2739                                             &r->rpool.cur->addr.v.a.mask,
2740                                             saddr, pd->af);
2741                                 break;
2742                         case PF_IN:
2743                                 if (r->src.addr.type == PF_ADDR_DYNIFTL) {
2744                                         switch (pd->af) {
2745 #ifdef INET
2746                                         case AF_INET:
2747                                                 if (r->src.addr.p.dyn->
2748                                                     pfid_acnt4 < 1)
2749                                                         return (NULL);
2750                                                 PF_POOLMASK(naddr,
2751                                                     &r->src.addr.p.dyn->
2752                                                     pfid_addr4,
2753                                                     &r->src.addr.p.dyn->
2754                                                     pfid_mask4,
2755                                                     daddr, AF_INET);
2756                                                 break;
2757 #endif /* INET */
2758 #ifdef INET6
2759                                         case AF_INET6:
2760                                                 if (r->src.addr.p.dyn->
2761                                                     pfid_acnt6 < 1)
2762                                                         return (NULL);
2763                                                 PF_POOLMASK(naddr,
2764                                                     &r->src.addr.p.dyn->
2765                                                     pfid_addr6,
2766                                                     &r->src.addr.p.dyn->
2767                                                     pfid_mask6,
2768                                                     daddr, AF_INET6);
2769                                                 break;
2770 #endif /* INET6 */
2771                                         }
2772                                 } else
2773                                         PF_POOLMASK(naddr,
2774                                             &r->src.addr.v.a.addr,
2775                                             &r->src.addr.v.a.mask, daddr,
2776                                             pd->af);
2777                                 break;
2778                         }
2779                         break;
2780                 case PF_RDR: {
2781                         m->m_pkthdr.fw_flags &= ~BRIDGE_MBUF_TAGGED;
2782                         if (pf_map_addr(pd->af, r, saddr, naddr, NULL, sn))
2783                                 return (NULL);
2784                         if ((r->rpool.opts & PF_POOL_TYPEMASK) ==
2785                             PF_POOL_BITMASK)
2786                                 PF_POOLMASK(naddr, naddr,
2787                                     &r->rpool.cur->addr.v.a.mask, daddr,
2788                                     pd->af);
2789
2790                         if (r->rpool.proxy_port[1]) {
2791                                 u_int32_t       tmp_nport;
2792
2793                                 tmp_nport = ((ntohs(dport) -
2794                                     ntohs(r->dst.port[0])) %
2795                                     (r->rpool.proxy_port[1] -
2796                                     r->rpool.proxy_port[0] + 1)) +
2797                                     r->rpool.proxy_port[0];
2798
2799                                 /* wrap around if necessary */
2800                                 if (tmp_nport > 65535)
2801                                         tmp_nport -= 65535;
2802                                 *nport = htons((u_int16_t)tmp_nport);
2803                         } else if (r->rpool.proxy_port[0])
2804                                 *nport = htons(r->rpool.proxy_port[0]);
2805                         break;
2806                 }
2807                 default:
2808                         return (NULL);
2809                 }
2810         }
2811
2812         return (r);
2813 }
2814
2815 #ifdef SMP
2816 struct netmsg_hashlookup {
2817         struct netmsg_base      base;
2818         struct inpcb            **nm_pinp;
2819         struct inpcbinfo        *nm_pcbinfo;
2820         struct pf_addr          *nm_saddr;
2821         struct pf_addr          *nm_daddr;
2822         uint16_t                nm_sport;
2823         uint16_t                nm_dport;
2824         sa_family_t             nm_af;
2825 };
2826
2827 #ifdef PF_SOCKET_LOOKUP_DOMSG
2828 static void
2829 in_pcblookup_hash_handler(netmsg_t msg)
2830 {
2831         struct netmsg_hashlookup *rmsg = (struct netmsg_hashlookup *)msg;
2832
2833         if (rmsg->nm_af == AF_INET)
2834                 *rmsg->nm_pinp = in_pcblookup_hash(rmsg->nm_pcbinfo,
2835                     rmsg->nm_saddr->v4, rmsg->nm_sport, rmsg->nm_daddr->v4,
2836                     rmsg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2837 #ifdef INET6
2838         else
2839                 *rmsg->nm_pinp = in6_pcblookup_hash(rmsg->nm_pcbinfo,
2840                     &rmsg->nm_saddr->v6, rmsg->nm_sport, &rmsg->nm_daddr->v6,
2841                     rmsg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2842 #endif /* INET6 */
2843         lwkt_replymsg(&rmsg->base.lmsg, 0);
2844 }
2845 #endif  /* PF_SOCKET_LOOKUP_DOMSG */
2846
2847 #endif /* SMP */
2848
2849 int
2850 pf_socket_lookup(int direction, struct pf_pdesc *pd)
2851 {
2852         struct pf_addr          *saddr, *daddr;
2853         u_int16_t                sport, dport;
2854         struct inpcbinfo        *pi;
2855         struct inpcb            *inp;
2856 #ifdef SMP
2857         struct netmsg_hashlookup *msg = NULL;
2858 #ifdef PF_SOCKET_LOOKUP_DOMSG
2859         struct netmsg_hashlookup msg0;
2860 #endif
2861 #endif
2862         int                      pi_cpu = 0;
2863
2864         if (pd == NULL)
2865                 return (-1);
2866         pd->lookup.uid = UID_MAX;
2867         pd->lookup.gid = GID_MAX;
2868         pd->lookup.pid = NO_PID;
2869         if (direction == PF_IN) {
2870                 saddr = pd->src;
2871                 daddr = pd->dst;
2872         } else {
2873                 saddr = pd->dst;
2874                 daddr = pd->src;
2875         }
2876         switch (pd->proto) {
2877         case IPPROTO_TCP:
2878                 if (pd->hdr.tcp == NULL)
2879                         return (-1);
2880                 sport = pd->hdr.tcp->th_sport;
2881                 dport = pd->hdr.tcp->th_dport;
2882
2883                 pi_cpu = tcp_addrcpu(saddr->v4.s_addr, sport, daddr->v4.s_addr, dport);
2884                 pi = &tcbinfo[pi_cpu];
2885 #ifdef SMP
2886                 /*
2887                  * Our netstack runs lockless on MP systems
2888                  * (only for TCP connections at the moment).
2889                  * 
2890                  * As we are not allowed to read another CPU's tcbinfo,
2891                  * we have to ask that CPU via remote call to search the
2892                  * table for us.
2893                  * 
2894                  * Prepare a msg iff data belongs to another CPU.
2895                  */
2896                 if (pi_cpu != mycpu->gd_cpuid) {
2897 #ifdef PF_SOCKET_LOOKUP_DOMSG
2898                         /*
2899                          * NOTE:
2900                          *
2901                          * Following lwkt_domsg() is dangerous and could
2902                          * lockup the network system, e.g.
2903                          *
2904                          * On 2 CPU system:
2905                          * netisr0 domsg to netisr1 (due to lookup)
2906                          * netisr1 domsg to netisr0 (due to lookup)
2907                          *
2908                          * We simply return -1 here, since we are probably
2909                          * called before NAT, so the TCP packet should
2910                          * already be on the correct CPU.
2911                          */
2912                         msg = &msg0;
2913                         netmsg_init(&msg->base, NULL, &curthread->td_msgport,
2914                                     0, in_pcblookup_hash_handler);
2915                         msg->nm_pinp = &inp;
2916                         msg->nm_pcbinfo = pi;
2917                         msg->nm_saddr = saddr;
2918                         msg->nm_sport = sport;
2919                         msg->nm_daddr = daddr;
2920                         msg->nm_dport = dport;
2921                         msg->nm_af = pd->af;
2922 #else   /* !PF_SOCKET_LOOKUP_DOMSG */
2923                         kprintf("pf_socket_lookup: tcp packet not on the "
2924                                 "correct cpu %d, cur cpu %d\n",
2925                                 pi_cpu, mycpuid);
2926                         print_backtrace(-1);
2927                         return -1;
2928 #endif  /* PF_SOCKET_LOOKUP_DOMSG */
2929                 }
2930 #endif /* SMP */
2931                 break;
2932         case IPPROTO_UDP:
2933                 if (pd->hdr.udp == NULL)
2934                         return (-1);
2935                 sport = pd->hdr.udp->uh_sport;
2936                 dport = pd->hdr.udp->uh_dport;
2937                 pi = &udbinfo;
2938                 break;
2939         default:
2940                 return (-1);
2941         }
2942         if (direction != PF_IN) {
2943                 u_int16_t       p;
2944
2945                 p = sport;
2946                 sport = dport;
2947                 dport = p;
2948         }
2949         switch (pd->af) {
2950 #ifdef INET6
2951         case AF_INET6:
2952 #ifdef SMP
2953                 /*
2954                  * Query other CPU, second part
2955                  * 
2956                  * msg only gets initialized when:
2957                  * 1) packet is TCP
2958                  * 2) the info belongs to another CPU
2959                  *
2960                  * Use some switch/case magic to avoid code duplication.
2961                  */
2962                 if (msg == NULL)
2963 #endif /* SMP */
2964                 {
2965                         inp = in6_pcblookup_hash(pi, &saddr->v6, sport,
2966                             &daddr->v6, dport, INPLOOKUP_WILDCARD, NULL);
2967
2968                         if (inp == NULL)
2969                                 return (-1);
2970                         break;
2971                 }
2972                 /* FALLTHROUGH if SMP and on other CPU */
2973 #endif /* INET6 */
2974         case AF_INET:
2975 #ifdef SMP
2976                 if (msg != NULL) {
2977                         lwkt_domsg(cpu_portfn(pi_cpu),
2978                                      &msg->base.lmsg, 0);
2979                 } else
2980 #endif /* SMP */
2981                 {
2982                         inp = in_pcblookup_hash(pi, saddr->v4, sport, daddr->v4,
2983                             dport, INPLOOKUP_WILDCARD, NULL);
2984                 }
2985                 if (inp == NULL)
2986                         return (-1);
2987                 break;
2988
2989         default:
2990                 return (-1);
2991         }
2992         pd->lookup.uid = inp->inp_socket->so_cred->cr_uid;
2993         pd->lookup.gid = inp->inp_socket->so_cred->cr_groups[0];
2994         return (1);
2995 }
2996
2997 u_int8_t
2998 pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2999 {
3000         int              hlen;
3001         u_int8_t         hdr[60];
3002         u_int8_t        *opt, optlen;
3003         u_int8_t         wscale = 0;
3004
3005         hlen = th_off << 2;             /* hlen <= sizeof(hdr) */
3006         if (hlen <= sizeof(struct tcphdr))
3007                 return (0);
3008         if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
3009                 return (0);
3010         opt = hdr + sizeof(struct tcphdr);
3011         hlen -= sizeof(struct tcphdr);
3012         while (hlen >= 3) {
3013                 switch (*opt) {
3014                 case TCPOPT_EOL:
3015                 case TCPOPT_NOP:
3016                         ++opt;
3017                         --hlen;
3018                         break;
3019                 case TCPOPT_WINDOW:
3020                         wscale = opt[2];
3021                         if (wscale > TCP_MAX_WINSHIFT)
3022                                 wscale = TCP_MAX_WINSHIFT;
3023                         wscale |= PF_WSCALE_FLAG;
3024                         /* FALLTHROUGH */
3025                 default:
3026                         optlen = opt[1];
3027                         if (optlen < 2)
3028                                 optlen = 2;
3029                         hlen -= optlen;
3030                         opt += optlen;
3031                         break;
3032                 }
3033         }
3034         return (wscale);
3035 }
3036
3037 u_int16_t
3038 pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
3039 {
3040         int              hlen;
3041         u_int8_t         hdr[60];
3042         u_int8_t        *opt, optlen;
3043         u_int16_t        mss = tcp_mssdflt;
3044
3045         hlen = th_off << 2;     /* hlen <= sizeof(hdr) */
3046         if (hlen <= sizeof(struct tcphdr))
3047                 return (0);
3048         if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
3049                 return (0);
3050         opt = hdr + sizeof(struct tcphdr);
3051         hlen -= sizeof(struct tcphdr);
3052         while (hlen >= TCPOLEN_MAXSEG) {
3053                 switch (*opt) {
3054                 case TCPOPT_EOL:
3055                 case TCPOPT_NOP:
3056                         ++opt;
3057                         --hlen;
3058                         break;
3059                 case TCPOPT_MAXSEG:
3060                         bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2);
3061                         /* FALLTHROUGH */
3062                 default:
3063                         optlen = opt[1];
3064                         if (optlen < 2)
3065                                 optlen = 2;
3066                         hlen -= optlen;
3067                         opt += optlen;
3068                         break;
3069                 }
3070         }
3071         return (mss);
3072 }
3073
3074 u_int16_t
3075 pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer)
3076 {
3077 #ifdef INET
3078         struct sockaddr_in      *dst;
3079         struct route             ro;
3080 #endif /* INET */
3081 #ifdef INET6
3082         struct sockaddr_in6     *dst6;
3083         struct route_in6         ro6;
3084 #endif /* INET6 */
3085         struct rtentry          *rt = NULL;
3086         int                      hlen = 0;
3087         u_int16_t                mss = tcp_mssdflt;
3088
3089         switch (af) {
3090 #ifdef INET
3091         case AF_INET:
3092                 hlen = sizeof(struct ip);
3093                 bzero(&ro, sizeof(ro));
3094                 dst = (struct sockaddr_in *)&ro.ro_dst;
3095                 dst->sin_family = AF_INET;
3096                 dst->sin_len = sizeof(*dst);
3097                 dst->sin_addr = addr->v4;
3098                 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
3099                 rt = ro.ro_rt;
3100                 break;
3101 #endif /* INET */
3102 #ifdef INET6
3103         case AF_INET6:
3104                 hlen = sizeof(struct ip6_hdr);
3105                 bzero(&ro6, sizeof(ro6));
3106                 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst;
3107                 dst6->sin6_family = AF_INET6;
3108                 dst6->sin6_len = sizeof(*dst6);
3109                 dst6->sin6_addr = addr->v6;
3110                 rtalloc_ign((struct route *)&ro6, (RTF_CLONING | RTF_PRCLONING));
3111                 rt = ro6.ro_rt;
3112                 break;
3113 #endif /* INET6 */
3114         }
3115
3116         if (rt && rt->rt_ifp) {
3117                 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr);
3118                 mss = max(tcp_mssdflt, mss);
3119                 RTFREE(rt);
3120         }
3121         mss = min(mss, offer);
3122         mss = max(mss, 64);             /* sanity - at least max opt space */
3123         return (mss);
3124 }
3125
3126 void
3127 pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr)
3128 {
3129         struct pf_rule *r = s->rule.ptr;
3130
3131         s->rt_kif = NULL;
3132         if (!r->rt || r->rt == PF_FASTROUTE)
3133                 return;
3134         switch (s->key[PF_SK_WIRE]->af) {
3135 #ifdef INET
3136         case AF_INET:
3137                 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL,
3138                     &s->nat_src_node);
3139                 s->rt_kif = r->rpool.cur->kif;
3140                 break;
3141 #endif /* INET */
3142 #ifdef INET6
3143         case AF_INET6:
3144                 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL,
3145                     &s->nat_src_node);
3146                 s->rt_kif = r->rpool.cur->kif;
3147                 break;
3148 #endif /* INET6 */
3149         }
3150 }
3151
3152 u_int32_t
3153 pf_tcp_iss(struct pf_pdesc *pd)
3154 {
3155         MD5_CTX ctx;
3156         u_int32_t digest[4];
3157
3158         if (pf_tcp_secret_init == 0) {
3159                 karc4rand(pf_tcp_secret, sizeof(pf_tcp_secret));
3160                 MD5Init(&pf_tcp_secret_ctx);
3161                 MD5Update(&pf_tcp_secret_ctx, pf_tcp_secret,
3162                     sizeof(pf_tcp_secret));
3163                 pf_tcp_secret_init = 1;
3164         }
3165         ctx = pf_tcp_secret_ctx;
3166
3167         MD5Update(&ctx, (char *)&pd->hdr.tcp->th_sport, sizeof(u_short));
3168         MD5Update(&ctx, (char *)&pd->hdr.tcp->th_dport, sizeof(u_short));
3169         if (pd->af == AF_INET6) {
3170                 MD5Update(&ctx, (char *)&pd->src->v6, sizeof(struct in6_addr));
3171                 MD5Update(&ctx, (char *)&pd->dst->v6, sizeof(struct in6_addr));
3172         } else {
3173                 MD5Update(&ctx, (char *)&pd->src->v4, sizeof(struct in_addr));
3174                 MD5Update(&ctx, (char *)&pd->dst->v4, sizeof(struct in_addr));
3175         }
3176         MD5Final((u_char *)digest, &ctx);
3177         pf_tcp_iss_off += 4096;
3178         return (digest[0] + pd->hdr.tcp->th_seq + pf_tcp_iss_off);
3179 }
3180
3181 int
3182 pf_test_rule(struct pf_rule **rm, struct pf_state **sm, int direction,
3183     struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3184     struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
3185     struct ifqueue *ifq, struct inpcb *inp)
3186 {
3187         struct pf_rule          *nr = NULL;
3188         struct pf_addr          *saddr = pd->src, *daddr = pd->dst;
3189         sa_family_t              af = pd->af;
3190         struct pf_rule          *r, *a = NULL;
3191         struct pf_ruleset       *ruleset = NULL;
3192         struct pf_src_node      *nsn = NULL;
3193         struct tcphdr           *th = pd->hdr.tcp;
3194         struct pf_state_key     *skw = NULL, *sks = NULL;
3195         struct pf_state_key     *sk = NULL, *nk = NULL;
3196         u_short                  reason;
3197         int                      rewrite = 0, hdrlen = 0;
3198         int                      tag = -1, rtableid = -1;
3199         int                      asd = 0;
3200         int                      match = 0;
3201         int                      state_icmp = 0;
3202         u_int16_t                sport = 0, dport = 0;
3203         u_int16_t                nport = 0, bport = 0;
3204         u_int16_t                bproto_sum = 0, bip_sum = 0;
3205         u_int8_t                 icmptype = 0, icmpcode = 0;
3206
3207
3208         if (direction == PF_IN && pf_check_congestion(ifq)) {
3209                 REASON_SET(&reason, PFRES_CONGEST);
3210                 return (PF_DROP);
3211         }
3212
3213         if (inp != NULL)
3214                 pd->lookup.done = pf_socket_lookup(direction, pd);
3215         else if (debug_pfugidhack) { 
3216                 DPFPRINTF(PF_DEBUG_MISC, ("pf: unlocked lookup\n"));
3217                 pd->lookup.done = pf_socket_lookup(direction, pd);
3218         }
3219
3220         switch (pd->proto) {
3221         case IPPROTO_TCP:
3222                 sport = th->th_sport;
3223                 dport = th->th_dport;
3224                 hdrlen = sizeof(*th);
3225                 break;
3226         case IPPROTO_UDP:
3227                 sport = pd->hdr.udp->uh_sport;
3228                 dport = pd->hdr.udp->uh_dport;
3229                 hdrlen = sizeof(*pd->hdr.udp);
3230                 break;
3231 #ifdef INET
3232         case IPPROTO_ICMP:
3233                 if (pd->af != AF_INET)
3234                         break;
3235                 sport = dport = pd->hdr.icmp->icmp_id;
3236                 hdrlen = sizeof(*pd->hdr.icmp);
3237                 icmptype = pd->hdr.icmp->icmp_type;
3238                 icmpcode = pd->hdr.icmp->icmp_code;
3239
3240                 if (icmptype == ICMP_UNREACH ||
3241                     icmptype == ICMP_SOURCEQUENCH ||
3242                     icmptype == ICMP_REDIRECT ||
3243                     icmptype == ICMP_TIMXCEED ||
3244                     icmptype == ICMP_PARAMPROB)
3245                         state_icmp++;
3246                 break;
3247 #endif /* INET */
3248 #ifdef INET6
3249         case IPPROTO_ICMPV6:
3250                 if (af != AF_INET6)
3251                         break;
3252                 sport = dport = pd->hdr.icmp6->icmp6_id;
3253                 hdrlen = sizeof(*pd->hdr.icmp6);
3254                 icmptype = pd->hdr.icmp6->icmp6_type;
3255                 icmpcode = pd->hdr.icmp6->icmp6_code;
3256
3257                 if (icmptype == ICMP6_DST_UNREACH ||
3258                     icmptype == ICMP6_PACKET_TOO_BIG ||
3259                     icmptype == ICMP6_TIME_EXCEEDED ||
3260                     icmptype == ICMP6_PARAM_PROB)
3261                         state_icmp++;
3262                 break;
3263 #endif /* INET6 */
3264         default:
3265                 sport = dport = hdrlen = 0;
3266                 break;
3267         }
3268
3269         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3270
3271         bport = nport = sport;
3272         /* check packet for BINAT/NAT/RDR */
3273         if ((nr = pf_get_translation(pd, m, off, direction, kif, &nsn,
3274             &skw, &sks, &sk, &nk, saddr, daddr, sport, dport)) != NULL) {
3275                 if (nk == NULL || sk == NULL) {
3276                         REASON_SET(&reason, PFRES_MEMORY);
3277                         goto cleanup;
3278                 }
3279
3280                 if (pd->ip_sum)
3281                         bip_sum = *pd->ip_sum;
3282
3283                 switch (pd->proto) {
3284                 case IPPROTO_TCP:
3285                         bproto_sum = th->th_sum;
3286                         pd->proto_sum = &th->th_sum;
3287
3288                         if (PF_ANEQ(saddr, &nk->addr[pd->sidx], af) ||
3289                             nk->port[pd->sidx] != sport) {
3290                                 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
3291                                     &th->th_sum, &nk->addr[pd->sidx],
3292                                     nk->port[pd->sidx], 0, af);
3293                                 pd->sport = &th->th_sport;
3294                                 sport = th->th_sport;
3295                         }
3296
3297                         if (PF_ANEQ(daddr, &nk->addr[pd->didx], af) ||
3298                             nk->port[pd->didx] != dport) {
3299                                 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
3300                                     &th->th_sum, &nk->addr[pd->didx],
3301                                     nk->port[pd->didx], 0, af);
3302                                 dport = th->th_dport;
3303                                 pd->dport = &th->th_dport;
3304                         }
3305                         rewrite++;
3306                         break;
3307                 case IPPROTO_UDP:
3308                         bproto_sum = pd->hdr.udp->uh_sum;
3309                         pd->proto_sum = &pd->hdr.udp->uh_sum;
3310
3311                         if (PF_ANEQ(saddr, &nk->addr[pd->sidx], af) ||
3312                             nk->port[pd->sidx] != sport) {
3313                                 pf_change_ap(saddr, &pd->hdr.udp->uh_sport,
3314                                     pd->ip_sum, &pd->hdr.udp->uh_sum,
3315                                     &nk->addr[pd->sidx],
3316                                     nk->port[pd->sidx], 1, af);
3317                                 sport = pd->hdr.udp->uh_sport;
3318                                 pd->sport = &pd->hdr.udp->uh_sport;
3319                         }
3320
3321                         if (PF_ANEQ(daddr, &nk->addr[pd->didx], af) ||
3322                             nk->port[pd->didx] != dport) {
3323                                 pf_change_ap(daddr, &pd->hdr.udp->uh_dport,
3324                                     pd->ip_sum, &pd->hdr.udp->uh_sum,
3325                                     &nk->addr[pd->didx],
3326                                     nk->port[pd->didx], 1, af);
3327                                 dport = pd->hdr.udp->uh_dport;
3328                                 pd->dport = &pd->hdr.udp->uh_dport;
3329                         }
3330                         rewrite++;
3331                         break;
3332 #ifdef INET
3333                 case IPPROTO_ICMP:
3334                         nk->port[0] = nk->port[1];
3335                         if (PF_ANEQ(saddr, &nk->addr[pd->sidx], AF_INET))
3336                                 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3337                                     nk->addr[pd->sidx].v4.s_addr, 0);
3338
3339                         if (PF_ANEQ(daddr, &nk->addr[pd->didx], AF_INET))
3340                                 pf_change_a(&daddr->v4.s_addr, pd->ip_sum,
3341                                     nk->addr[pd->didx].v4.s_addr, 0);
3342
3343                         if (nk->port[1] != pd->hdr.icmp->icmp_id) {
3344                                 pd->hdr.icmp->icmp_cksum = pf_cksum_fixup(
3345                                     pd->hdr.icmp->icmp_cksum, sport,
3346                                     nk->port[1], 0);
3347                                 pd->hdr.icmp->icmp_id = nk->port[1];
3348                                 pd->sport = &pd->hdr.icmp->icmp_id;
3349                         }
3350                         m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
3351                         break;
3352 #endif /* INET */
3353 #ifdef INET6
3354                 case IPPROTO_ICMPV6:
3355                         nk->port[0] = nk->port[1];
3356                         if (PF_ANEQ(saddr, &nk->addr[pd->sidx], AF_INET6))
3357                                 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum,
3358                                     &nk->addr[pd->sidx], 0);
3359
3360                         if (PF_ANEQ(daddr, &nk->addr[pd->didx], AF_INET6))
3361                                 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum,
3362                                     &nk->addr[pd->didx], 0);
3363                         rewrite++;
3364                         break;
3365 #endif /* INET */
3366                 default:
3367                         switch (af) {
3368 #ifdef INET
3369                         case AF_INET:
3370                                 if (PF_ANEQ(saddr,
3371                                     &nk->addr[pd->sidx], AF_INET))
3372                                         pf_change_a(&saddr->v4.s_addr,
3373                                             pd->ip_sum,
3374                                             nk->addr[pd->sidx].v4.s_addr, 0);
3375
3376                                 if (PF_ANEQ(daddr,
3377                                     &nk->addr[pd->didx], AF_INET))
3378                                         pf_change_a(&daddr->v4.s_addr,
3379                                             pd->ip_sum,
3380                                             nk->addr[pd->didx].v4.s_addr, 0);
3381                                 break;
3382 #endif /* INET */
3383 #ifdef INET6
3384                         case AF_INET6:
3385                                 if (PF_ANEQ(saddr,
3386                                     &nk->addr[pd->sidx], AF_INET6))
3387                                         PF_ACPY(saddr, &nk->addr[pd->sidx], af);
3388
3389                                 if (PF_ANEQ(daddr,
3390                                     &nk->addr[pd->didx], AF_INET6))
3391                                         PF_ACPY(saddr, &nk->addr[pd->didx], af);
3392                                 break;
3393 #endif /* INET */
3394                         }
3395                         break;
3396                 }
3397                 if (nr->natpass)
3398                         r = NULL;
3399                 pd->nat_rule = nr;
3400         }
3401
3402         while (r != NULL) {
3403                 r->evaluations++;
3404                 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3405                         r = r->skip[PF_SKIP_IFP].ptr;
3406                 else if (r->direction && r->direction != direction)
3407                         r = r->skip[PF_SKIP_DIR].ptr;
3408                 else if (r->af && r->af != af)
3409                         r = r->skip[PF_SKIP_AF].ptr;
3410                 else if (r->proto && r->proto != pd->proto)
3411                         r = r->skip[PF_SKIP_PROTO].ptr;
3412                 else if (PF_MISMATCHAW(&r->src.addr, saddr, af,
3413                     r->src.neg, kif))
3414                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3415                 /* tcp/udp only. port_op always 0 in other cases */
3416                 else if (r->src.port_op && !pf_match_port(r->src.port_op,
3417                     r->src.port[0], r->src.port[1], sport))
3418                         r = r->skip[PF_SKIP_SRC_PORT].ptr;
3419                 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af,
3420                     r->dst.neg, NULL))
3421                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
3422                 /* tcp/udp only. port_op always 0 in other cases */
3423                 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
3424                     r->dst.port[0], r->dst.port[1], dport))
3425                         r = r->skip[PF_SKIP_DST_PORT].ptr;
3426                 /* icmp only. type always 0 in other cases */
3427                 else if (r->type && r->type != icmptype + 1)
3428                         r = TAILQ_NEXT(r, entries);
3429                 /* icmp only. type always 0 in other cases */
3430                 else if (r->code && r->code != icmpcode + 1)
3431                         r = TAILQ_NEXT(r, entries);
3432                 else if (r->tos && !(r->tos == pd->tos))
3433                         r = TAILQ_NEXT(r, entries);
3434                 else if (r->rule_flag & PFRULE_FRAGMENT)
3435                         r = TAILQ_NEXT(r, entries);
3436                 else if (pd->proto == IPPROTO_TCP &&
3437                     (r->flagset & th->th_flags) != r->flags)
3438                         r = TAILQ_NEXT(r, entries);
3439                 /* tcp/udp only. uid.op always 0 in other cases */
3440                 else if (r->uid.op && (pd->lookup.done || (pd->lookup.done =
3441                     pf_socket_lookup(direction, pd), 1)) &&
3442                     !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
3443                     pd->lookup.uid))
3444                         r = TAILQ_NEXT(r, entries);
3445                 /* tcp/udp only. gid.op always 0 in other cases */
3446                 else if (r->gid.op && (pd->lookup.done || (pd->lookup.done =
3447                     pf_socket_lookup(direction, pd), 1)) &&
3448                     !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
3449                     pd->lookup.gid))
3450                         r = TAILQ_NEXT(r, entries);
3451                 else if (r->prob &&
3452                   r->prob <= karc4random())
3453                         r = TAILQ_NEXT(r, entries);
3454                 else if (r->match_tag && !pf_match_tag(m, r, &tag))
3455                         r = TAILQ_NEXT(r, entries);
3456                 else if (r->os_fingerprint != PF_OSFP_ANY &&
3457                     (pd->proto != IPPROTO_TCP || !pf_osfp_match(
3458                     pf_osfp_fingerprint(pd, m, off, th),
3459                     r->os_fingerprint)))
3460                         r = TAILQ_NEXT(r, entries);
3461                 else {
3462                         if (r->tag)
3463                                 tag = r->tag;
3464                         if (r->rtableid >= 0)
3465                                 rtableid = r->rtableid;
3466                         if (r->anchor == NULL) {
3467                                 match = 1;
3468                                 *rm = r;
3469                                 *am = a;
3470                                 *rsm = ruleset;
3471                                 if ((*rm)->quick)
3472                                         break;
3473                                 r = TAILQ_NEXT(r, entries);
3474                         } else
3475                                 pf_step_into_anchor(&asd, &ruleset,
3476                                     PF_RULESET_FILTER, &r, &a, &match);
3477                 }
3478                 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3479                     PF_RULESET_FILTER, &r, &a, &match))
3480                         break;
3481         }
3482         r = *rm;
3483         a = *am;
3484         ruleset = *rsm;
3485
3486         REASON_SET(&reason, PFRES_MATCH);
3487
3488         if (r->log || (nr != NULL && nr->log)) {
3489                 if (rewrite)
3490                         m_copyback(m, off, hdrlen, pd->hdr.any);
3491                 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr,
3492                     a, ruleset, pd);
3493         }
3494
3495         if ((r->action == PF_DROP) &&
3496             ((r->rule_flag & PFRULE_RETURNRST) ||
3497             (r->rule_flag & PFRULE_RETURNICMP) ||
3498             (r->rule_flag & PFRULE_RETURN))) {
3499                 /* undo NAT changes, if they have taken place */
3500                 if (nr != NULL) {
3501                         PF_ACPY(saddr, &sk->addr[pd->sidx], af);
3502                         PF_ACPY(daddr, &sk->addr[pd->didx], af);
3503                         if (pd->sport)
3504                                 *pd->sport = sk->port[pd->sidx];
3505                         if (pd->dport)
3506                                 *pd->dport = sk->port[pd->didx];
3507                         if (pd->proto_sum)
3508                                 *pd->proto_sum = bproto_sum;
3509                         if (pd->ip_sum)
3510                                 *pd->ip_sum = bip_sum;
3511                         m_copyback(m, off, hdrlen, pd->hdr.any);
3512                 }
3513                 if (pd->proto == IPPROTO_TCP &&
3514                     ((r->rule_flag & PFRULE_RETURNRST) ||
3515                     (r->rule_flag & PFRULE_RETURN)) &&
3516                     !(th->th_flags & TH_RST)) {
3517                         u_int32_t        ack = ntohl(th->th_seq) + pd->p_len;
3518                         int              len = 0;
3519                         struct ip       *h4;
3520 #ifdef INET6
3521                         struct ip6_hdr  *h6;
3522 #endif
3523                         switch (af) {
3524                         case AF_INET:
3525                                 h4 = mtod(m, struct ip *);
3526                                 len = h4->ip_len - off;
3527                                 break;
3528 #ifdef INET6
3529                         case AF_INET6:
3530                                 h6 = mtod(m, struct ip6_hdr *);
3531                                 len = h6->ip6_plen - (off - sizeof(*h6));
3532                                 break;
3533 #endif
3534                         }
3535
3536                         if (pf_check_proto_cksum(m, off, len, IPPROTO_TCP, af))
3537                                 REASON_SET(&reason, PFRES_PROTCKSUM);
3538                         else {
3539                                 if (th->th_flags & TH_SYN)
3540                                         ack++;
3541                                 if (th->th_flags & TH_FIN)
3542                                         ack++;
3543                                 pf_send_tcp(r, af, pd->dst,
3544                                     pd->src, th->th_dport, th->th_sport,
3545                                     ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0,
3546                                     r->return_ttl, 1, 0, pd->eh, kif->pfik_ifp);
3547                         }
3548                 } else if (pd->proto != IPPROTO_ICMP && af == AF_INET &&
3549                     r->return_icmp)
3550                         pf_send_icmp(m, r->return_icmp >> 8,
3551                             r->return_icmp & 255, af, r);
3552                 else if (pd->proto != IPPROTO_ICMPV6 && af == AF_INET6 &&
3553                     r->return_icmp6)
3554                         pf_send_icmp(m, r->return_icmp6 >> 8,
3555                             r->return_icmp6 & 255, af, r);
3556         }
3557
3558         if (r->action == PF_DROP)
3559                 goto cleanup;
3560
3561         if (pf_tag_packet(m, tag, rtableid)) {
3562                 REASON_SET(&reason, PFRES_MEMORY);
3563                 goto cleanup;
3564         }
3565
3566         if (!state_icmp && (r->keep_state || nr != NULL ||
3567             (pd->flags & PFDESC_TCP_NORM))) {
3568                 int action;
3569                 action = pf_create_state(r, nr, a, pd, nsn, skw, sks, nk, sk, m,
3570                     off, sport, dport, &rewrite, kif, sm, tag, bproto_sum,
3571                     bip_sum, hdrlen);
3572                 if (action != PF_PASS)
3573                         return (action);
3574         }
3575
3576         /* copy back packet headers if we performed NAT operations */
3577         if (rewrite)
3578                 m_copyback(m, off, hdrlen, pd->hdr.any);
3579
3580         return (PF_PASS);
3581
3582 cleanup:
3583         if (sk != NULL)
3584                 kfree(sk, M_PFSTATEKEYPL);
3585         if (nk != NULL)
3586                 kfree(nk, M_PFSTATEKEYPL);
3587         return (PF_DROP);
3588 }
3589
3590 static __inline int
3591 pf_create_state(struct pf_rule *r, struct pf_rule *nr, struct pf_rule *a,
3592     struct pf_pdesc *pd, struct pf_src_node *nsn, struct pf_state_key *skw,
3593     struct pf_state_key *sks, struct pf_state_key *nk, struct pf_state_key *sk,
3594     struct mbuf *m, int off, u_int16_t sport, u_int16_t dport, int *rewrite,
3595     struct pfi_kif *kif, struct pf_state **sm, int tag, u_int16_t bproto_sum,
3596     u_int16_t bip_sum, int hdrlen)
3597 {
3598         struct pf_state         *s = NULL;
3599         struct pf_src_node      *sn = NULL;
3600         struct tcphdr           *th = pd->hdr.tcp;
3601         u_int16_t                mss = tcp_mssdflt;
3602         u_short                  reason;
3603
3604         /* check maximums */
3605         if (r->max_states && (r->states_cur >= r->max_states)) {
3606                 pf_status.lcounters[LCNT_STATES]++;
3607                 REASON_SET(&reason, PFRES_MAXSTATES);
3608                 return (PF_DROP);
3609         }
3610         /* src node for filter rule */
3611         if ((r->rule_flag & PFRULE_SRCTRACK ||
3612             r->rpool.opts & PF_POOL_STICKYADDR) &&
3613             pf_insert_src_node(&sn, r, pd->src, pd->af) != 0) {
3614                 REASON_SET(&reason, PFRES_SRCLIMIT);
3615                 goto csfailed;
3616         }
3617         /* src node for translation rule */
3618         if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3619             pf_insert_src_node(&nsn, nr, &sk->addr[pd->sidx], pd->af)) {
3620                 REASON_SET(&reason, PFRES_SRCLIMIT);
3621                 goto csfailed;
3622         }
3623         s = kmalloc(sizeof(struct pf_state), M_PFSTATEPL, M_NOWAIT|M_ZERO);
3624         if (s == NULL) {
3625                 REASON_SET(&reason, PFRES_MEMORY);
3626                 goto csfailed;
3627         }
3628         s->id = 0; /* XXX Do we really need that? not in OpenBSD */
3629         s->creatorid = 0;
3630         s->rule.ptr = r;
3631         s->nat_rule.ptr = nr;
3632         s->anchor.ptr = a;
3633         STATE_INC_COUNTERS(s);
3634         if (r->allow_opts)
3635                 s->state_flags |= PFSTATE_ALLOWOPTS;
3636         if (r->rule_flag & PFRULE_STATESLOPPY)
3637                 s->state_flags |= PFSTATE_SLOPPY;
3638         s->log = r->log & PF_LOG_ALL;
3639         if (nr != NULL)
3640                 s->log |= nr->log & PF_LOG_ALL;
3641         switch (pd->proto) {
3642         case IPPROTO_TCP:
3643                 s->src.seqlo = ntohl(th->th_seq);
3644                 s->src.seqhi = s->src.seqlo + pd->p_len + 1;
3645                 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
3646                     r->keep_state == PF_STATE_MODULATE) {
3647                         /* Generate sequence number modulator */
3648                         if ((s->src.seqdiff = pf_tcp_iss(pd) - s->src.seqlo) ==
3649                             0)
3650                                 s->src.seqdiff = 1;
3651                         pf_change_a(&th->th_seq, &th->th_sum,
3652                             htonl(s->src.seqlo + s->src.seqdiff), 0);
3653                         *rewrite = 1;
3654                 } else
3655                         s->src.seqdiff = 0;
3656                 if (th->th_flags & TH_SYN) {
3657                         s->src.seqhi++;
3658                         s->src.wscale = pf_get_wscale(m, off,
3659                             th->th_off, pd->af);
3660                 }
3661                 s->src.max_win = MAX(ntohs(th->th_win), 1);
3662                 if (s->src.wscale & PF_WSCALE_MASK) {
3663                         /* Remove scale factor from initial window */
3664                         int win = s->src.max_win;
3665                         win += 1 << (s->src.wscale & PF_WSCALE_MASK);
3666                         s->src.max_win = (win - 1) >>
3667                             (s->src.wscale & PF_WSCALE_MASK);
3668                 }
3669                 if (th->th_flags & TH_FIN)
3670                         s->src.seqhi++;
3671                 s->dst.seqhi = 1;
3672                 s->dst.max_win = 1;
3673                 s->src.state = TCPS_SYN_SENT;
3674                 s->dst.state = TCPS_CLOSED;
3675                 s->timeout = PFTM_TCP_FIRST_PACKET;
3676                 break;
3677         case IPPROTO_UDP:
3678                 s->src.state = PFUDPS_SINGLE;
3679                 s->dst.state = PFUDPS_NO_TRAFFIC;
3680                 s->timeout = PFTM_UDP_FIRST_PACKET;
3681                 break;
3682         case IPPROTO_ICMP:
3683 #ifdef INET6
3684         case IPPROTO_ICMPV6:
3685 #endif
3686                 s->timeout = PFTM_ICMP_FIRST_PACKET;
3687                 break;
3688         default:
3689                 s->src.state = PFOTHERS_SINGLE;
3690                 s->dst.state = PFOTHERS_NO_TRAFFIC;
3691                 s->timeout = PFTM_OTHER_FIRST_PACKET;
3692         }
3693
3694         s->creation = time_second;
3695         s->expire = time_second;
3696
3697         if (sn != NULL) {
3698                 s->src_node = sn;
3699                 s->src_node->states++;
3700         }
3701         if (nsn != NULL) {
3702                 /* XXX We only modify one side for now. */
3703                 PF_ACPY(&nsn->raddr, &nk->addr[1], pd->af);
3704                 s->nat_src_node = nsn;
3705                 s->nat_src_node->states++;
3706         }
3707         if (pd->proto == IPPROTO_TCP) {
3708                 if ((pd->flags & PFDESC_TCP_NORM) && pf_normalize_tcp_init(m,
3709                     off, pd, th, &s->src, &s->dst)) {
3710                         REASON_SET(&reason, PFRES_MEMORY);
3711                         pf_src_tree_remove_state(s);
3712                         STATE_DEC_COUNTERS(s);
3713                         kfree(s, M_PFSTATEPL);
3714                         return (PF_DROP);
3715                 }
3716                 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
3717                     pf_normalize_tcp_stateful(m, off, pd, &reason, th, s,
3718                     &s->src, &s->dst, rewrite)) {
3719                         /* This really shouldn't happen!!! */
3720                         DPFPRINTF(PF_DEBUG_URGENT,
3721                             ("pf_normalize_tcp_stateful failed on first pkt"));
3722                         pf_normalize_tcp_cleanup(s);
3723                         pf_src_tree_remove_state(s);
3724                         STATE_DEC_COUNTERS(s);
3725                         kfree(s, M_PFSTATEPL);
3726                         return (PF_DROP);
3727                 }
3728         }
3729         s->direction = pd->dir;
3730
3731         if (sk == NULL && pf_state_key_setup(pd, nr, &skw, &sks, &sk, &nk,
3732             pd->src, pd->dst, sport, dport))
3733                 goto csfailed;
3734
3735         if (pf_state_insert(BOUND_IFACE(r, kif), skw, sks, s)) {
3736                 if (pd->proto == IPPROTO_TCP)
3737                         pf_normalize_tcp_cleanup(s);
3738                 REASON_SET(&reason, PFRES_STATEINS);
3739                 pf_src_tree_remove_state(s);
3740                 STATE_DEC_COUNTERS(s);
3741                 kfree(s, M_PFSTATEPL);
3742                 return (PF_DROP);
3743         } else
3744                 *sm = s;
3745
3746         pf_set_rt_ifp(s, pd->src);      /* needs s->state_key set */
3747         if (tag > 0) {
3748                 pf_tag_ref(tag);
3749                 s->tag = tag;
3750         }
3751         if (pd->proto == IPPROTO_TCP && (th->th_flags & (TH_SYN|TH_ACK)) ==
3752             TH_SYN && r->keep_state == PF_STATE_SYNPROXY) {
3753                 s->src.state = PF_TCPS_PROXY_SRC;
3754                 /* undo NAT changes, if they have taken place */
3755                 if (nr != NULL) {
3756                         struct pf_state_key *skt = s->key[PF_SK_WIRE];
3757                         if (pd->dir == PF_OUT)
3758                                 skt = s->key[PF_SK_STACK];
3759                         PF_ACPY(pd->src, &skt->addr[pd->sidx], pd->af);
3760                         PF_ACPY(pd->dst, &skt->addr[pd->didx], pd->af);
3761                         if (pd->sport)
3762                                 *pd->sport = skt->port[pd->sidx];
3763                         if (pd->dport)
3764                                 *pd->dport = skt->port[pd->didx];
3765                         if (pd->proto_sum)
3766                                 *pd->proto_sum = bproto_sum;
3767                  &