1 /* $NetBSD: pf.c,v 1.87 2022/11/04 09:01:53 ozaki-r Exp $ */
2 /* $OpenBSD: pf.c,v 1.552.2.1 2007/11/27 16:37:57 henning Exp $ */
3
4 /*
5 * Copyright (c) 2001 Daniel Hartmeier
6 * Copyright (c) 2002,2003 Henning Brauer
7 * All rights reserved.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 *
13 * - Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * - Redistributions in binary form must reproduce the above
16 * copyright notice, this list of conditions and the following
17 * disclaimer in the documentation and/or other materials provided
18 * with the distribution.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
26 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
28 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
30 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
31 * POSSIBILITY OF SUCH DAMAGE.
32 *
33 * Effort sponsored in part by the Defense Advanced Research Projects
34 * Agency (DARPA) and Air Force Research Laboratory, Air Force
35 * Materiel Command, USAF, under agreement number F30602-01-2-0537.
36 *
37 */
38
39 #include <sys/cdefs.h>
40 __KERNEL_RCSID(0, "$NetBSD: pf.c,v 1.87 2022/11/04 09:01:53 ozaki-r Exp $");
41
42 #include "pflog.h"
43
44 #include "pfsync.h"
45
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/mbuf.h>
49 #include <sys/filio.h>
50 #include <sys/socket.h>
51 #include <sys/socketvar.h>
52 #include <sys/kernel.h>
53 #include <sys/time.h>
54 #include <sys/pool.h>
55 #include <sys/proc.h>
56 #include <sys/rwlock.h>
57 #ifdef __NetBSD__
58 #include <sys/kthread.h>
59 #include <sys/kauth.h>
60 #endif /* __NetBSD__ */
61
62 #include <net/if.h>
63 #include <net/if_types.h>
64 #include <net/bpf.h>
65 #include <net/route.h>
66 #ifndef __NetBSD__
67 #include <net/radix_mpath.h>
68 #endif /* !__NetBSD__ */
69
70 #include <netinet/in.h>
71 #ifdef __NetBSD__
72 #include <netinet/in_offload.h>
73 #endif /* __NetBSD__ */
74 #include <netinet/in_var.h>
75 #include <netinet/in_systm.h>
76 #include <netinet/ip.h>
77 #include <netinet/ip_var.h>
78 #include <netinet/tcp.h>
79 #include <netinet/tcp_seq.h>
80 #include <netinet/udp.h>
81 #include <netinet/ip_icmp.h>
82 #include <netinet/in_pcb.h>
83 #include <netinet/tcp_timer.h>
84 #include <netinet/tcp_var.h>
85 #include <netinet/udp_var.h>
86 #include <netinet/icmp_var.h>
87 #ifndef __NetBSD__
88 #include <netinet/if_ether.h>
89 #else
90 #include <net/if_ether.h>
91 #endif /* __NetBSD__ */
92
93 #ifndef __NetBSD__
94 #include <dev/rndvar.h>
95 #else
96 #include <sys/cprng.h>
97 #endif /* __NetBSD__ */
98
99 #include <net/pfvar.h>
100 #include <net/if_pflog.h>
101
102 #if NPFSYNC > 0
103 #include <net/if_pfsync.h>
104 #endif /* NPFSYNC > 0 */
105
106 #ifdef INET6
107 #include <netinet/ip6.h>
108 #include <netinet6/ip6_var.h>
109 #ifdef __NetBSD__
110 #include <netinet6/in6_pcb.h>
111 #endif /* __NetBSD__ */
112 #include <netinet/icmp6.h>
113 #include <netinet6/nd6.h>
114 #endif /* INET6 */
115
116 #ifdef __NetBSD__
117 #include <netinet/tcp_rndiss.h>
118 #endif /* __NetBSD__ */
119
120
121 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) printf x
122
123 /*
124 * Global variables
125 */
126
127 /* state tables */
128 struct pf_state_tree_lan_ext pf_statetbl_lan_ext;
129 struct pf_state_tree_ext_gwy pf_statetbl_ext_gwy;
130
131 struct pf_altqqueue pf_altqs[2];
132 struct pf_palist pf_pabuf;
133 struct pf_altqqueue *pf_altqs_active;
134 struct pf_altqqueue *pf_altqs_inactive;
135 struct pf_status pf_status;
136
137 u_int32_t ticket_altqs_active;
138 u_int32_t ticket_altqs_inactive;
139 int altqs_inactive_open;
140 u_int32_t ticket_pabuf;
141
142 struct pf_anchor_stackframe {
143 struct pf_ruleset *rs;
144 struct pf_rule *r;
145 struct pf_anchor_node *parent;
146 struct pf_anchor *child;
147 } pf_anchor_stack[64];
148
149 struct pool pf_src_tree_pl, pf_rule_pl, pf_pooladdr_pl;
150 struct pool pf_state_pl, pf_state_key_pl;
151 struct pool pf_altq_pl;
152
153 void pf_print_host(struct pf_addr *, u_int16_t, u_int8_t);
154
155 void pf_init_threshold(struct pf_threshold *, u_int32_t,
156 u_int32_t);
157 void pf_add_threshold(struct pf_threshold *);
158 int pf_check_threshold(struct pf_threshold *);
159
160 void pf_change_ap(struct pf_addr *, u_int16_t *,
161 u_int16_t *, u_int16_t *, struct pf_addr *,
162 u_int16_t, u_int8_t, sa_family_t);
163 int pf_modulate_sack(struct mbuf *, int, struct pf_pdesc *,
164 struct tcphdr *, struct pf_state_peer *);
165 #ifdef INET6
166 void pf_change_a6(struct pf_addr *, u_int16_t *,
167 struct pf_addr *, u_int8_t);
168 #endif /* INET6 */
169 void pf_change_icmp(struct pf_addr *, u_int16_t *,
170 struct pf_addr *, struct pf_addr *, u_int16_t,
171 u_int16_t *, u_int16_t *, u_int16_t *,
172 u_int16_t *, u_int8_t, sa_family_t);
173 void pf_send_tcp(const struct pf_rule *, sa_family_t,
174 const struct pf_addr *, const struct pf_addr *,
175 u_int16_t, u_int16_t, u_int32_t, u_int32_t,
176 u_int8_t, u_int16_t, u_int16_t, u_int8_t, int,
177 u_int16_t, struct ether_header *, struct ifnet *);
178 void pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t,
179 sa_family_t, struct pf_rule *);
180 struct pf_rule *pf_match_translation(struct pf_pdesc *, struct mbuf *,
181 int, int, struct pfi_kif *,
182 struct pf_addr *, u_int16_t, struct pf_addr *,
183 u_int16_t, int);
184 struct pf_rule *pf_get_translation(struct pf_pdesc *, struct mbuf *,
185 int, int, struct pfi_kif *, struct pf_src_node **,
186 struct pf_addr *, u_int16_t,
187 struct pf_addr *, u_int16_t,
188 struct pf_addr *, u_int16_t *);
189 void pf_attach_state(struct pf_state_key *,
190 struct pf_state *, int);
191 void pf_detach_state(struct pf_state *, int);
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 *);
196 int pf_test_fragment(struct pf_rule **, int,
197 struct pfi_kif *, struct mbuf *, void *,
198 struct pf_pdesc *, struct pf_rule **,
199 struct pf_ruleset **);
200 int pf_test_state_tcp(struct pf_state **, int,
201 struct pfi_kif *, struct mbuf *, int,
202 void *, struct pf_pdesc *, u_short *);
203 int pf_test_state_udp(struct pf_state **, int,
204 struct pfi_kif *, struct mbuf *, int,
205 void *, struct pf_pdesc *);
206 int pf_test_state_icmp(struct pf_state **, int,
207 struct pfi_kif *, struct mbuf *, int,
208 void *, struct pf_pdesc *, u_short *);
209 int pf_test_state_other(struct pf_state **, int,
210 struct pfi_kif *, struct pf_pdesc *);
211 int pf_match_tag(struct mbuf *, struct pf_rule *, int *);
212 void pf_step_into_anchor(int *, struct pf_ruleset **, int,
213 struct pf_rule **, struct pf_rule **, int *);
214 int pf_step_out_of_anchor(int *, struct pf_ruleset **,
215 int, struct pf_rule **, struct pf_rule **,
216 int *);
217 void pf_hash(const struct pf_addr *, struct pf_addr *,
218 struct pf_poolhashkey *, sa_family_t);
219 int pf_map_addr(u_int8_t, struct pf_rule *,
220 const struct pf_addr *, struct pf_addr *,
221 struct pf_addr *, struct pf_src_node **);
222 int pf_get_sport(sa_family_t, u_int8_t, struct pf_rule *,
223 struct pf_addr *, struct pf_addr *, u_int16_t,
224 struct pf_addr *, u_int16_t*, u_int16_t, u_int16_t,
225 struct pf_src_node **);
226 void pf_route(struct mbuf **, struct pf_rule *, int,
227 struct ifnet *, struct pf_state *,
228 struct pf_pdesc *);
229 void pf_route6(struct mbuf **, struct pf_rule *, int,
230 struct ifnet *, struct pf_state *,
231 struct pf_pdesc *);
232 int pf_socket_lookup(int, struct pf_pdesc *);
233 u_int8_t pf_get_wscale(struct mbuf *, int, u_int16_t,
234 sa_family_t);
235 u_int16_t pf_get_mss(struct mbuf *, int, u_int16_t,
236 sa_family_t);
237 u_int16_t pf_calc_mss(struct pf_addr *, sa_family_t,
238 u_int16_t);
239 void pf_set_rt_ifp(struct pf_state *,
240 struct pf_addr *);
241 #ifdef __NetBSD__
242 int pf_check_proto_cksum(struct mbuf *, int, int, int,
243 u_int8_t, sa_family_t);
244 #else
245 int pf_check_proto_cksum(struct mbuf *, int, int,
246 u_int8_t, sa_family_t);
247 #endif /* !__NetBSD__ */
248 int pf_addr_wrap_neq(struct pf_addr_wrap *,
249 struct pf_addr_wrap *);
250 struct pf_state *pf_find_state(struct pfi_kif *,
251 struct pf_state_key_cmp *, u_int8_t);
252 int pf_src_connlimit(struct pf_state **);
253 void pf_stateins_err(const char *, struct pf_state *,
254 struct pfi_kif *);
255 int pf_check_congestion(struct ifqueue *);
256
257 extern struct pool pfr_ktable_pl;
258 extern struct pool pfr_kentry_pl;
259
260 extern int pf_state_lock;
261
262 struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] = {
263 { &pf_state_pl, PFSTATE_HIWAT },
264 { &pf_src_tree_pl, PFSNODE_HIWAT },
265 { &pf_frent_pl, PFFRAG_FRENT_HIWAT },
266 { &pfr_ktable_pl, PFR_KTABLE_HIWAT },
267 { &pfr_kentry_pl, PFR_KENTRY_HIWAT }
268 };
269
270 #define STATE_LOOKUP() \
271 do { \
272 if (pf_state_lock) { \
273 *state = NULL; \
274 return (PF_DROP); \
275 } \
276 if (direction == PF_IN) \
277 *state = pf_find_state(kif, &key, PF_EXT_GWY); \
278 else \
279 *state = pf_find_state(kif, &key, PF_LAN_EXT); \
280 if (*state == NULL || (*state)->timeout == PFTM_PURGE) \
281 return (PF_DROP); \
282 if (direction == PF_OUT && \
283 (((*state)->rule.ptr->rt == PF_ROUTETO && \
284 (*state)->rule.ptr->direction == PF_OUT) || \
285 ((*state)->rule.ptr->rt == PF_REPLYTO && \
286 (*state)->rule.ptr->direction == PF_IN)) && \
287 (*state)->rt_kif != NULL && \
288 (*state)->rt_kif != kif) \
289 return (PF_PASS); \
290 } while (0)
291
292 #define STATE_TRANSLATE(sk) \
293 (sk)->lan.addr.addr32[0] != (sk)->gwy.addr.addr32[0] || \
294 ((sk)->af == AF_INET6 && \
295 ((sk)->lan.addr.addr32[1] != (sk)->gwy.addr.addr32[1] || \
296 (sk)->lan.addr.addr32[2] != (sk)->gwy.addr.addr32[2] || \
297 (sk)->lan.addr.addr32[3] != (sk)->gwy.addr.addr32[3])) || \
298 (sk)->lan.port != (sk)->gwy.port
299
300 #define BOUND_IFACE(r, k) \
301 ((r)->rule_flag & PFRULE_IFBOUND) ? (k) : pfi_all
302
303 #define STATE_INC_COUNTERS(s) \
304 do { \
305 s->rule.ptr->states++; \
306 if (s->anchor.ptr != NULL) \
307 s->anchor.ptr->states++; \
308 if (s->nat_rule.ptr != NULL) \
309 s->nat_rule.ptr->states++; \
310 } while (0)
311
312 #define STATE_DEC_COUNTERS(s) \
313 do { \
314 if (s->nat_rule.ptr != NULL) \
315 s->nat_rule.ptr->states--; \
316 if (s->anchor.ptr != NULL) \
317 s->anchor.ptr->states--; \
318 s->rule.ptr->states--; \
319 } while (0)
320
321 static __inline int pf_src_compare(struct pf_src_node *, struct pf_src_node *);
322 static __inline int pf_state_compare_lan_ext(struct pf_state_key *,
323 struct pf_state_key *);
324 static __inline int pf_state_compare_ext_gwy(struct pf_state_key *,
325 struct pf_state_key *);
326 static __inline int pf_state_compare_id(struct pf_state *,
327 struct pf_state *);
328
329 struct pf_src_tree tree_src_tracking;
330
331 struct pf_state_tree_id tree_id;
332 struct pf_state_queue state_list;
333
334 RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare);
335 RB_GENERATE(pf_state_tree_lan_ext, pf_state_key,
336 entry_lan_ext, pf_state_compare_lan_ext);
337 RB_GENERATE(pf_state_tree_ext_gwy, pf_state_key,
338 entry_ext_gwy, pf_state_compare_ext_gwy);
339 RB_GENERATE(pf_state_tree_id, pf_state,
340 entry_id, pf_state_compare_id);
341
342 #define PF_DT_SKIP_LANEXT 0x01
343 #define PF_DT_SKIP_EXTGWY 0x02
344
345 #ifdef __NetBSD__
346 static __inline struct pfi_kif *
bound_iface(const struct pf_rule * r,const struct pf_rule * nr,struct pfi_kif * k)347 bound_iface(const struct pf_rule *r, const struct pf_rule *nr,
348 struct pfi_kif *k)
349 {
350 uint32_t rule_flag;
351
352 rule_flag = r->rule_flag;
353 if (nr != NULL)
354 rule_flag |= nr->rule_flag;
355
356 return ((rule_flag & PFRULE_IFBOUND) != 0) ? k : pfi_all;
357 }
358 #endif /* __NetBSD__ */
359
360 static __inline int
pf_src_compare(struct pf_src_node * a,struct pf_src_node * b)361 pf_src_compare(struct pf_src_node *a, struct pf_src_node *b)
362 {
363 int diff;
364
365 if (a->rule.ptr > b->rule.ptr)
366 return (1);
367 if (a->rule.ptr < b->rule.ptr)
368 return (-1);
369 if ((diff = a->af - b->af) != 0)
370 return (diff);
371 switch (a->af) {
372 #ifdef INET
373 case AF_INET:
374 if (a->addr.addr32[0] > b->addr.addr32[0])
375 return (1);
376 if (a->addr.addr32[0] < b->addr.addr32[0])
377 return (-1);
378 break;
379 #endif /* INET */
380 #ifdef INET6
381 case AF_INET6:
382 if (a->addr.addr32[3] > b->addr.addr32[3])
383 return (1);
384 if (a->addr.addr32[3] < b->addr.addr32[3])
385 return (-1);
386 if (a->addr.addr32[2] > b->addr.addr32[2])
387 return (1);
388 if (a->addr.addr32[2] < b->addr.addr32[2])
389 return (-1);
390 if (a->addr.addr32[1] > b->addr.addr32[1])
391 return (1);
392 if (a->addr.addr32[1] < b->addr.addr32[1])
393 return (-1);
394 if (a->addr.addr32[0] > b->addr.addr32[0])
395 return (1);
396 if (a->addr.addr32[0] < b->addr.addr32[0])
397 return (-1);
398 break;
399 #endif /* INET6 */
400 }
401 return (0);
402 }
403
404 static __inline int
pf_state_compare_lan_ext(struct pf_state_key * a,struct pf_state_key * b)405 pf_state_compare_lan_ext(struct pf_state_key *a, struct pf_state_key *b)
406 {
407 int diff;
408
409 if ((diff = a->proto - b->proto) != 0)
410 return (diff);
411 if ((diff = a->af - b->af) != 0)
412 return (diff);
413 switch (a->af) {
414 #ifdef INET
415 case AF_INET:
416 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
417 return (1);
418 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
419 return (-1);
420 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
421 return (1);
422 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
423 return (-1);
424 break;
425 #endif /* INET */
426 #ifdef INET6
427 case AF_INET6:
428 if (a->lan.addr.addr32[3] > b->lan.addr.addr32[3])
429 return (1);
430 if (a->lan.addr.addr32[3] < b->lan.addr.addr32[3])
431 return (-1);
432 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
433 return (1);
434 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
435 return (-1);
436 if (a->lan.addr.addr32[2] > b->lan.addr.addr32[2])
437 return (1);
438 if (a->lan.addr.addr32[2] < b->lan.addr.addr32[2])
439 return (-1);
440 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
441 return (1);
442 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
443 return (-1);
444 if (a->lan.addr.addr32[1] > b->lan.addr.addr32[1])
445 return (1);
446 if (a->lan.addr.addr32[1] < b->lan.addr.addr32[1])
447 return (-1);
448 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
449 return (1);
450 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
451 return (-1);
452 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
453 return (1);
454 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
455 return (-1);
456 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
457 return (1);
458 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
459 return (-1);
460 break;
461 #endif /* INET6 */
462 }
463
464 if ((diff = a->lan.port - b->lan.port) != 0)
465 return (diff);
466 if ((diff = a->ext.port - b->ext.port) != 0)
467 return (diff);
468
469 return (0);
470 }
471
472 static __inline int
pf_state_compare_ext_gwy(struct pf_state_key * a,struct pf_state_key * b)473 pf_state_compare_ext_gwy(struct pf_state_key *a, struct pf_state_key *b)
474 {
475 int diff;
476
477 if ((diff = a->proto - b->proto) != 0)
478 return (diff);
479 if ((diff = a->af - b->af) != 0)
480 return (diff);
481 switch (a->af) {
482 #ifdef INET
483 case AF_INET:
484 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
485 return (1);
486 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
487 return (-1);
488 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
489 return (1);
490 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
491 return (-1);
492 break;
493 #endif /* INET */
494 #ifdef INET6
495 case AF_INET6:
496 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
497 return (1);
498 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
499 return (-1);
500 if (a->gwy.addr.addr32[3] > b->gwy.addr.addr32[3])
501 return (1);
502 if (a->gwy.addr.addr32[3] < b->gwy.addr.addr32[3])
503 return (-1);
504 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
505 return (1);
506 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
507 return (-1);
508 if (a->gwy.addr.addr32[2] > b->gwy.addr.addr32[2])
509 return (1);
510 if (a->gwy.addr.addr32[2] < b->gwy.addr.addr32[2])
511 return (-1);
512 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
513 return (1);
514 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
515 return (-1);
516 if (a->gwy.addr.addr32[1] > b->gwy.addr.addr32[1])
517 return (1);
518 if (a->gwy.addr.addr32[1] < b->gwy.addr.addr32[1])
519 return (-1);
520 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
521 return (1);
522 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
523 return (-1);
524 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
525 return (1);
526 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
527 return (-1);
528 break;
529 #endif /* INET6 */
530 }
531
532 if ((diff = a->ext.port - b->ext.port) != 0)
533 return (diff);
534 if ((diff = a->gwy.port - b->gwy.port) != 0)
535 return (diff);
536
537 return (0);
538 }
539
540 static __inline int
pf_state_compare_id(struct pf_state * a,struct pf_state * b)541 pf_state_compare_id(struct pf_state *a, struct pf_state *b)
542 {
543 if (a->id > b->id)
544 return (1);
545 if (a->id < b->id)
546 return (-1);
547 if (a->creatorid > b->creatorid)
548 return (1);
549 if (a->creatorid < b->creatorid)
550 return (-1);
551
552 return (0);
553 }
554
555 #ifdef INET6
556 void
pf_addrcpy(struct pf_addr * dst,const struct pf_addr * src,sa_family_t af)557 pf_addrcpy(struct pf_addr *dst, const struct pf_addr *src, sa_family_t af)
558 {
559 switch (af) {
560 #ifdef INET
561 case AF_INET:
562 dst->addr32[0] = src->addr32[0];
563 break;
564 #endif /* INET */
565 case AF_INET6:
566 dst->addr32[0] = src->addr32[0];
567 dst->addr32[1] = src->addr32[1];
568 dst->addr32[2] = src->addr32[2];
569 dst->addr32[3] = src->addr32[3];
570 break;
571 }
572 }
573 #endif /* INET6 */
574
575 struct pf_state *
pf_find_state_byid(struct pf_state_cmp * key)576 pf_find_state_byid(struct pf_state_cmp *key)
577 {
578 pf_status.fcounters[FCNT_STATE_SEARCH]++;
579
580 return (RB_FIND(pf_state_tree_id, &tree_id, (struct pf_state *)key));
581 }
582
583 struct pf_state *
pf_find_state(struct pfi_kif * kif,struct pf_state_key_cmp * key,u_int8_t tree)584 pf_find_state(struct pfi_kif *kif, struct pf_state_key_cmp *key, u_int8_t tree)
585 {
586 struct pf_state_key *sk;
587 struct pf_state *s;
588
589 pf_status.fcounters[FCNT_STATE_SEARCH]++;
590
591 switch (tree) {
592 case PF_LAN_EXT:
593 sk = RB_FIND(pf_state_tree_lan_ext, &pf_statetbl_lan_ext,
594 (struct pf_state_key *)key);
595 break;
596 case PF_EXT_GWY:
597 sk = RB_FIND(pf_state_tree_ext_gwy, &pf_statetbl_ext_gwy,
598 (struct pf_state_key *)key);
599 break;
600 default:
601 panic("pf_find_state");
602 }
603
604 /* list is sorted, if-bound states before floating ones */
605 if (sk != NULL)
606 TAILQ_FOREACH(s, &sk->states, next)
607 if (s->kif == pfi_all || s->kif == kif)
608 return (s);
609
610 return (NULL);
611 }
612
613 struct pf_state *
pf_find_state_all(struct pf_state_key_cmp * key,u_int8_t tree,int * more)614 pf_find_state_all(struct pf_state_key_cmp *key, u_int8_t tree, int *more)
615 {
616 struct pf_state_key *sk;
617 struct pf_state *s, *ret = NULL;
618
619 pf_status.fcounters[FCNT_STATE_SEARCH]++;
620
621 switch (tree) {
622 case PF_LAN_EXT:
623 sk = RB_FIND(pf_state_tree_lan_ext,
624 &pf_statetbl_lan_ext, (struct pf_state_key *)key);
625 break;
626 case PF_EXT_GWY:
627 sk = RB_FIND(pf_state_tree_ext_gwy,
628 &pf_statetbl_ext_gwy, (struct pf_state_key *)key);
629 break;
630 default:
631 panic("pf_find_state_all");
632 }
633
634 if (sk != NULL) {
635 ret = TAILQ_FIRST(&sk->states);
636 if (more == NULL)
637 return (ret);
638
639 TAILQ_FOREACH(s, &sk->states, next)
640 (*more)++;
641 }
642
643 return (ret);
644 }
645
646 void
pf_init_threshold(struct pf_threshold * threshold,u_int32_t limit,u_int32_t seconds)647 pf_init_threshold(struct pf_threshold *threshold,
648 u_int32_t limit, u_int32_t seconds)
649 {
650 threshold->limit = limit * PF_THRESHOLD_MULT;
651 threshold->seconds = seconds;
652 threshold->count = 0;
653 threshold->last = time_second;
654 }
655
656 void
pf_add_threshold(struct pf_threshold * threshold)657 pf_add_threshold(struct pf_threshold *threshold)
658 {
659 u_int32_t t = time_second, diff = t - threshold->last;
660
661 if (diff >= threshold->seconds)
662 threshold->count = 0;
663 else
664 threshold->count -= threshold->count * diff /
665 threshold->seconds;
666 threshold->count += PF_THRESHOLD_MULT;
667 threshold->last = t;
668 }
669
670 int
pf_check_threshold(struct pf_threshold * threshold)671 pf_check_threshold(struct pf_threshold *threshold)
672 {
673 return (threshold->count > threshold->limit);
674 }
675
676 int
pf_src_connlimit(struct pf_state ** state)677 pf_src_connlimit(struct pf_state **state)
678 {
679 int bad = 0;
680
681 (*state)->src_node->conn++;
682 (*state)->src.tcp_est = 1;
683 pf_add_threshold(&(*state)->src_node->conn_rate);
684
685 if ((*state)->rule.ptr->max_src_conn &&
686 (*state)->rule.ptr->max_src_conn <
687 (*state)->src_node->conn) {
688 pf_status.lcounters[LCNT_SRCCONN]++;
689 bad++;
690 }
691
692 if ((*state)->rule.ptr->max_src_conn_rate.limit &&
693 pf_check_threshold(&(*state)->src_node->conn_rate)) {
694 pf_status.lcounters[LCNT_SRCCONNRATE]++;
695 bad++;
696 }
697
698 if (!bad)
699 return (0);
700
701 if ((*state)->rule.ptr->overload_tbl) {
702 struct pfr_addr p;
703 u_int32_t killed = 0;
704
705 pf_status.lcounters[LCNT_OVERLOAD_TABLE]++;
706 if (pf_status.debug >= PF_DEBUG_MISC) {
707 printf("pf_src_connlimit: blocking address ");
708 pf_print_host(&(*state)->src_node->addr, 0,
709 (*state)->state_key->af);
710 }
711
712 bzero(&p, sizeof(p));
713 p.pfra_af = (*state)->state_key->af;
714 switch ((*state)->state_key->af) {
715 #ifdef INET
716 case AF_INET:
717 p.pfra_net = 32;
718 p.pfra_ip4addr = (*state)->src_node->addr.v4;
719 break;
720 #endif /* INET */
721 #ifdef INET6
722 case AF_INET6:
723 p.pfra_net = 128;
724 p.pfra_ip6addr = (*state)->src_node->addr.v6;
725 break;
726 #endif /* INET6 */
727 }
728
729 pfr_insert_kentry((*state)->rule.ptr->overload_tbl,
730 &p, time_second);
731
732 /* kill existing states if that's required. */
733 if ((*state)->rule.ptr->flush) {
734 struct pf_state_key *sk;
735 struct pf_state *st;
736
737 pf_status.lcounters[LCNT_OVERLOAD_FLUSH]++;
738 RB_FOREACH(st, pf_state_tree_id, &tree_id) {
739 sk = st->state_key;
740 /*
741 * Kill states from this source. (Only those
742 * from the same rule if PF_FLUSH_GLOBAL is not
743 * set)
744 */
745 if (sk->af ==
746 (*state)->state_key->af &&
747 (((*state)->state_key->direction ==
748 PF_OUT &&
749 PF_AEQ(&(*state)->src_node->addr,
750 &sk->lan.addr, sk->af)) ||
751 ((*state)->state_key->direction == PF_IN &&
752 PF_AEQ(&(*state)->src_node->addr,
753 &sk->ext.addr, sk->af))) &&
754 ((*state)->rule.ptr->flush &
755 PF_FLUSH_GLOBAL ||
756 (*state)->rule.ptr == st->rule.ptr)) {
757 st->timeout = PFTM_PURGE;
758 st->src.state = st->dst.state =
759 TCPS_CLOSED;
760 killed++;
761 }
762 }
763 if (pf_status.debug >= PF_DEBUG_MISC)
764 printf(", %u states killed", killed);
765 }
766 if (pf_status.debug >= PF_DEBUG_MISC)
767 printf("\n");
768 }
769
770 /* kill this state */
771 (*state)->timeout = PFTM_PURGE;
772 (*state)->src.state = (*state)->dst.state = TCPS_CLOSED;
773 return (1);
774 }
775
776 int
pf_insert_src_node(struct pf_src_node ** sn,struct pf_rule * rule,struct pf_addr * src,sa_family_t af)777 pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule,
778 struct pf_addr *src, sa_family_t af)
779 {
780 struct pf_src_node k;
781
782 if (*sn == NULL) {
783 k.af = af;
784 PF_ACPY(&k.addr, src, af);
785 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
786 rule->rpool.opts & PF_POOL_STICKYADDR)
787 k.rule.ptr = rule;
788 else
789 k.rule.ptr = NULL;
790 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
791 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
792 }
793 if (*sn == NULL) {
794 if (!rule->max_src_nodes ||
795 rule->src_nodes < rule->max_src_nodes)
796 (*sn) = pool_get(&pf_src_tree_pl, PR_NOWAIT);
797 else
798 pf_status.lcounters[LCNT_SRCNODES]++;
799 if ((*sn) == NULL)
800 return (-1);
801 bzero(*sn, sizeof(struct pf_src_node));
802
803 pf_init_threshold(&(*sn)->conn_rate,
804 rule->max_src_conn_rate.limit,
805 rule->max_src_conn_rate.seconds);
806
807 (*sn)->af = af;
808 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
809 rule->rpool.opts & PF_POOL_STICKYADDR)
810 (*sn)->rule.ptr = rule;
811 else
812 (*sn)->rule.ptr = NULL;
813 PF_ACPY(&(*sn)->addr, src, af);
814 if (RB_INSERT(pf_src_tree,
815 &tree_src_tracking, *sn) != NULL) {
816 if (pf_status.debug >= PF_DEBUG_MISC) {
817 printf("pf: src_tree insert failed: ");
818 pf_print_host(&(*sn)->addr, 0, af);
819 printf("\n");
820 }
821 pool_put(&pf_src_tree_pl, *sn);
822 return (-1);
823 }
824 (*sn)->creation = time_second;
825 (*sn)->ruletype = rule->action;
826 if ((*sn)->rule.ptr != NULL)
827 (*sn)->rule.ptr->src_nodes++;
828 pf_status.scounters[SCNT_SRC_NODE_INSERT]++;
829 pf_status.src_nodes++;
830 } else {
831 if (rule->max_src_states &&
832 (*sn)->states >= rule->max_src_states) {
833 pf_status.lcounters[LCNT_SRCSTATES]++;
834 return (-1);
835 }
836 }
837 return (0);
838 }
839
840 void
pf_stateins_err(const char * tree,struct pf_state * s,struct pfi_kif * kif)841 pf_stateins_err(const char *tree, struct pf_state *s, struct pfi_kif *kif)
842 {
843 struct pf_state_key *sk = s->state_key;
844
845 if (pf_status.debug >= PF_DEBUG_MISC) {
846 printf("pf: state insert failed: %s %s", tree, kif->pfik_name);
847 printf(" lan: ");
848 pf_print_host(&sk->lan.addr, sk->lan.port,
849 sk->af);
850 printf(" gwy: ");
851 pf_print_host(&sk->gwy.addr, sk->gwy.port,
852 sk->af);
853 printf(" ext: ");
854 pf_print_host(&sk->ext.addr, sk->ext.port,
855 sk->af);
856 if (s->sync_flags & PFSTATE_FROMSYNC)
857 printf(" (from sync)");
858 printf("\n");
859 }
860 }
861
862 int
pf_insert_state(struct pfi_kif * kif,struct pf_state * s)863 pf_insert_state(struct pfi_kif *kif, struct pf_state *s)
864 {
865 struct pf_state_key *cur;
866 struct pf_state *sp;
867
868 KASSERT(s->state_key != NULL);
869 s->kif = kif;
870
871 if ((cur = RB_INSERT(pf_state_tree_lan_ext, &pf_statetbl_lan_ext,
872 s->state_key)) != NULL) {
873 /* key exists. check for same kif, if none, add to key */
874 TAILQ_FOREACH(sp, &cur->states, next)
875 if (sp->kif == kif) { /* collision! */
876 pf_stateins_err("tree_lan_ext", s, kif);
877 pf_detach_state(s,
878 PF_DT_SKIP_LANEXT|PF_DT_SKIP_EXTGWY);
879 return (-1);
880 }
881 pf_detach_state(s, PF_DT_SKIP_LANEXT|PF_DT_SKIP_EXTGWY);
882 pf_attach_state(cur, s, kif == pfi_all ? 1 : 0);
883 }
884
885 /* if cur != NULL, we already found a state key and attached to it */
886 if (cur == NULL && (cur = RB_INSERT(pf_state_tree_ext_gwy,
887 &pf_statetbl_ext_gwy, s->state_key)) != NULL) {
888 /* must not happen. we must have found the sk above! */
889 pf_stateins_err("tree_ext_gwy", s, kif);
890 pf_detach_state(s, PF_DT_SKIP_EXTGWY);
891 return (-1);
892 }
893
894 if (s->id == 0 && s->creatorid == 0) {
895 s->id = htobe64(pf_status.stateid++);
896 s->creatorid = pf_status.hostid;
897 }
898 if (RB_INSERT(pf_state_tree_id, &tree_id, s) != NULL) {
899 if (pf_status.debug >= PF_DEBUG_MISC) {
900 #ifdef __NetBSD__
901 printf("pf: state insert failed: "
902 "id: %016" PRIx64 " creatorid: %08x",
903 be64toh(s->id), ntohl(s->creatorid));
904 #else
905 printf("pf: state insert failed: "
906 "id: %016llx creatorid: %08x",
907 betoh64(s->id), ntohl(s->creatorid));
908 #endif /* !__NetBSD__ */
909 if (s->sync_flags & PFSTATE_FROMSYNC)
910 printf(" (from sync)");
911 printf("\n");
912 }
913 pf_detach_state(s, 0);
914 return (-1);
915 }
916 TAILQ_INSERT_TAIL(&state_list, s, entry_list);
917 pf_status.fcounters[FCNT_STATE_INSERT]++;
918 pf_status.states++;
919 pfi_kif_ref(kif, PFI_KIF_REF_STATE);
920 #if NPFSYNC
921 pfsync_insert_state(s);
922 #endif
923 return (0);
924 }
925
926 #ifdef _MODULE
927 volatile int pf_purge_thread_stop;
928 volatile int pf_purge_thread_running;
929 #endif
930
931 void
pf_purge_thread(void * v)932 pf_purge_thread(void *v)
933 {
934 int nloops = 0, s;
935
936 #ifdef _MODULE
937 pf_purge_thread_running = 1;
938 pf_purge_thread_stop = 0;
939
940 while (!pf_purge_thread_stop) {
941 #else
942 for (;;) {
943 #endif /* !_MODULE */
944 tsleep(pf_purge_thread, PWAIT, "pftm", 1 * hz);
945
946 s = splsoftnet();
947
948 /* process a fraction of the state table every second */
949 if (! pf_state_lock)
950 pf_purge_expired_states(1 + (pf_status.states
951 / pf_default_rule.timeout[PFTM_INTERVAL]));
952
953 /* purge other expired types every PFTM_INTERVAL seconds */
954 if (++nloops >= pf_default_rule.timeout[PFTM_INTERVAL]) {
955 pf_purge_expired_fragments();
956 pf_purge_expired_src_nodes(0);
957 nloops = 0;
958 }
959
960 splx(s);
961 }
962
963 #ifdef _MODULE
964 pf_purge_thread_running = 0;
965 wakeup(&pf_purge_thread_running);
966 kthread_exit(0);
967 #endif /* _MODULE */
968 }
969
970 u_int32_t
971 pf_state_expires(const struct pf_state *state)
972 {
973 u_int32_t timeout;
974 u_int32_t start;
975 u_int32_t end;
976 u_int32_t states;
977
978 /* handle all PFTM_* > PFTM_MAX here */
979 if (state->timeout == PFTM_PURGE)
980 return (time_second);
981 if (state->timeout == PFTM_UNTIL_PACKET)
982 return (0);
983 KASSERT(state->timeout != PFTM_UNLINKED);
984 KASSERT(state->timeout < PFTM_MAX);
985 timeout = state->rule.ptr->timeout[state->timeout];
986 if (!timeout)
987 timeout = pf_default_rule.timeout[state->timeout];
988 start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START];
989 if (start) {
990 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END];
991 states = state->rule.ptr->states;
992 } else {
993 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START];
994 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END];
995 states = pf_status.states;
996 }
997 if (end && states > start && start < end) {
998 if (states < end)
999 return (state->expire + timeout * (end - states) /
1000 (end - start));
1001 else
1002 return (time_second);
1003 }
1004 return (state->expire + timeout);
1005 }
1006
1007 void
1008 pf_purge_expired_src_nodes(int waslocked)
1009 {
1010 struct pf_src_node *cur, *next;
1011 int locked = waslocked;
1012
1013 for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) {
1014 next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur);
1015
1016 if (cur->states <= 0 && cur->expire <= time_second) {
1017 if (! locked) {
1018 rw_enter_write(&pf_consistency_lock);
1019 next = RB_NEXT(pf_src_tree,
1020 &tree_src_tracking, cur);
1021 locked = 1;
1022 }
1023 if (cur->rule.ptr != NULL) {
1024 cur->rule.ptr->src_nodes--;
1025 if (cur->rule.ptr->states <= 0 &&
1026 cur->rule.ptr->max_src_nodes <= 0)
1027 pf_rm_rule(NULL, cur->rule.ptr);
1028 }
1029 RB_REMOVE(pf_src_tree, &tree_src_tracking, cur);
1030 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
1031 pf_status.src_nodes--;
1032 pool_put(&pf_src_tree_pl, cur);
1033 }
1034 }
1035
1036 if (locked && !waslocked)
1037 rw_exit_write(&pf_consistency_lock);
1038 }
1039
1040 void
1041 pf_src_tree_remove_state(struct pf_state *s)
1042 {
1043 u_int32_t timeout;
1044
1045 if (s->src_node != NULL) {
1046 if (s->src.tcp_est)
1047 --s->src_node->conn;
1048 if (--s->src_node->states <= 0) {
1049 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1050 if (!timeout)
1051 timeout =
1052 pf_default_rule.timeout[PFTM_SRC_NODE];
1053 s->src_node->expire = time_second + timeout;
1054 }
1055 }
1056 if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) {
1057 if (--s->nat_src_node->states <= 0) {
1058 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1059 if (!timeout)
1060 timeout =
1061 pf_default_rule.timeout[PFTM_SRC_NODE];
1062 s->nat_src_node->expire = time_second + timeout;
1063 }
1064 }
1065 s->src_node = s->nat_src_node = NULL;
1066 }
1067
1068 /* callers should be at splsoftnet */
1069 void
1070 pf_unlink_state(struct pf_state *cur)
1071 {
1072 if (cur->src.state == PF_TCPS_PROXY_DST) {
1073 pf_send_tcp(cur->rule.ptr, cur->state_key->af,
1074 &cur->state_key->ext.addr, &cur->state_key->lan.addr,
1075 cur->state_key->ext.port, cur->state_key->lan.port,
1076 cur->src.seqhi, cur->src.seqlo + 1,
1077 TH_RST|TH_ACK, 0, 0, 0, 1, cur->tag, NULL, NULL);
1078 }
1079 RB_REMOVE(pf_state_tree_id, &tree_id, cur);
1080 #if NPFSYNC
1081 if (cur->creatorid == pf_status.hostid)
1082 pfsync_delete_state(cur);
1083 #endif
1084 cur->timeout = PFTM_UNLINKED;
1085 pf_src_tree_remove_state(cur);
1086 pf_detach_state(cur, 0);
1087 }
1088
1089 /* callers should be at splsoftnet and hold the
1090 * write_lock on pf_consistency_lock */
1091 void
1092 pf_free_state(struct pf_state *cur)
1093 {
1094 #if NPFSYNC
1095 if (pfsyncif != NULL &&
1096 (pfsyncif->sc_bulk_send_next == cur ||
1097 pfsyncif->sc_bulk_terminator == cur))
1098 return;
1099 #endif
1100 KASSERT(cur->timeout == PFTM_UNLINKED);
1101 if (--cur->rule.ptr->states <= 0 &&
1102 cur->rule.ptr->src_nodes <= 0)
1103 pf_rm_rule(NULL, cur->rule.ptr);
1104 if (cur->nat_rule.ptr != NULL)
1105 if (--cur->nat_rule.ptr->states <= 0 &&
1106 cur->nat_rule.ptr->src_nodes <= 0)
1107 pf_rm_rule(NULL, cur->nat_rule.ptr);
1108 if (cur->anchor.ptr != NULL)
1109 if (--cur->anchor.ptr->states <= 0)
1110 pf_rm_rule(NULL, cur->anchor.ptr);
1111 pf_normalize_tcp_cleanup(cur);
1112 pfi_kif_unref(cur->kif, PFI_KIF_REF_STATE);
1113 TAILQ_REMOVE(&state_list, cur, entry_list);
1114 if (cur->tag)
1115 pf_tag_unref(cur->tag);
1116 pool_put(&pf_state_pl, cur);
1117 pf_status.fcounters[FCNT_STATE_REMOVALS]++;
1118 pf_status.states--;
1119 }
1120
1121 void
1122 pf_purge_expired_states(u_int32_t maxcheck)
1123 {
1124 static struct pf_state *cur = NULL;
1125 struct pf_state *next;
1126 int locked = 0;
1127
1128 while (maxcheck--) {
1129 /* wrap to start of list when we hit the end */
1130 if (cur == NULL) {
1131 cur = TAILQ_FIRST(&state_list);
1132 if (cur == NULL)
1133 break; /* list empty */
1134 }
1135
1136 /* get next state, as cur may get deleted */
1137 next = TAILQ_NEXT(cur, entry_list);
1138
1139 if (cur->timeout == PFTM_UNLINKED) {
1140 /* free unlinked state */
1141 if (! locked) {
1142 rw_enter_write(&pf_consistency_lock);
1143 locked = 1;
1144 }
1145 pf_free_state(cur);
1146 } else if (pf_state_expires(cur) <= time_second) {
1147 /* unlink and free expired state */
1148 pf_unlink_state(cur);
1149 if (! locked) {
1150 rw_enter_write(&pf_consistency_lock);
1151 locked = 1;
1152 }
1153 pf_free_state(cur);
1154 }
1155 cur = next;
1156 }
1157
1158 if (locked)
1159 rw_exit_write(&pf_consistency_lock);
1160 }
1161
1162 int
1163 pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw)
1164 {
1165 if (aw->type != PF_ADDR_TABLE)
1166 return (0);
1167 if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL)
1168 return (1);
1169 return (0);
1170 }
1171
1172 void
1173 pf_tbladdr_remove(struct pf_addr_wrap *aw)
1174 {
1175 if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL)
1176 return;
1177 pfr_detach_table(aw->p.tbl);
1178 aw->p.tbl = NULL;
1179 }
1180
1181 void
1182 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
1183 {
1184 struct pfr_ktable *kt = aw->p.tbl;
1185
1186 if (aw->type != PF_ADDR_TABLE || kt == NULL)
1187 return;
1188 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
1189 kt = kt->pfrkt_root;
1190 aw->p.tbl = NULL;
1191 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
1192 kt->pfrkt_cnt : -1;
1193 }
1194
1195 void
1196 pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af)
1197 {
1198 switch (af) {
1199 #ifdef INET
1200 case AF_INET: {
1201 u_int32_t a = ntohl(addr->addr32[0]);
1202 printf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255,
1203 (a>>8)&255, a&255);
1204 if (p) {
1205 p = ntohs(p);
1206 printf(":%u", p);
1207 }
1208 break;
1209 }
1210 #endif /* INET */
1211 #ifdef INET6
1212 case AF_INET6: {
1213 u_int16_t b;
1214 u_int8_t i, curstart = 255, curend = 0,
1215 maxstart = 0, maxend = 0;
1216 for (i = 0; i < 8; i++) {
1217 if (!addr->addr16[i]) {
1218 if (curstart == 255)
1219 curstart = i;
1220 else
1221 curend = i;
1222 } else {
1223 if (curstart) {
1224 if ((curend - curstart) >
1225 (maxend - maxstart)) {
1226 maxstart = curstart;
1227 maxend = curend;
1228 curstart = 255;
1229 }
1230 }
1231 }
1232 }
1233 for (i = 0; i < 8; i++) {
1234 if (i >= maxstart && i <= maxend) {
1235 if (maxend != 7) {
1236 if (i == maxstart)
1237 printf(":");
1238 } else {
1239 if (i == maxend)
1240 printf(":");
1241 }
1242 } else {
1243 b = ntohs(addr->addr16[i]);
1244 printf("%x", b);
1245 if (i < 7)
1246 printf(":");
1247 }
1248 }
1249 if (p) {
1250 p = ntohs(p);
1251 printf("[%u]", p);
1252 }
1253 break;
1254 }
1255 #endif /* INET6 */
1256 }
1257 }
1258
1259 void
1260 pf_print_state(struct pf_state *s)
1261 {
1262 struct pf_state_key *sk = s->state_key;
1263 switch (sk->proto) {
1264 case IPPROTO_TCP:
1265 printf("TCP ");
1266 break;
1267 case IPPROTO_UDP:
1268 printf("UDP ");
1269 break;
1270 case IPPROTO_ICMP:
1271 printf("ICMP ");
1272 break;
1273 case IPPROTO_ICMPV6:
1274 printf("ICMPV6 ");
1275 break;
1276 default:
1277 printf("%u ", sk->proto);
1278 break;
1279 }
1280 pf_print_host(&sk->lan.addr, sk->lan.port, sk->af);
1281 printf(" ");
1282 pf_print_host(&sk->gwy.addr, sk->gwy.port, sk->af);
1283 printf(" ");
1284 pf_print_host(&sk->ext.addr, sk->ext.port, sk->af);
1285 printf(" [lo=%u high=%u win=%u modulator=%u", s->src.seqlo,
1286 s->src.seqhi, s->src.max_win, s->src.seqdiff);
1287 if (s->src.wscale && s->dst.wscale)
1288 printf(" wscale=%u", s->src.wscale & PF_WSCALE_MASK);
1289 printf("]");
1290 printf(" [lo=%u high=%u win=%u modulator=%u", s->dst.seqlo,
1291 s->dst.seqhi, s->dst.max_win, s->dst.seqdiff);
1292 if (s->src.wscale && s->dst.wscale)
1293 printf(" wscale=%u", s->dst.wscale & PF_WSCALE_MASK);
1294 printf("]");
1295 printf(" %u:%u", s->src.state, s->dst.state);
1296 }
1297
1298 void
1299 pf_print_flags(u_int8_t f)
1300 {
1301 if (f)
1302 printf(" ");
1303 if (f & TH_FIN)
1304 printf("F");
1305 if (f & TH_SYN)
1306 printf("S");
1307 if (f & TH_RST)
1308 printf("R");
1309 if (f & TH_PUSH)
1310 printf("P");
1311 if (f & TH_ACK)
1312 printf("A");
1313 if (f & TH_URG)
1314 printf("U");
1315 if (f & TH_ECE)
1316 printf("E");
1317 if (f & TH_CWR)
1318 printf("W");
1319 }
1320
1321 #define PF_SET_SKIP_STEPS(i) \
1322 do { \
1323 while (head[i] != cur) { \
1324 head[i]->skip[i].ptr = cur; \
1325 head[i] = TAILQ_NEXT(head[i], entries); \
1326 } \
1327 } while (0)
1328
1329 void
1330 pf_calc_skip_steps(struct pf_rulequeue *rules)
1331 {
1332 struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT];
1333 int i;
1334
1335 cur = TAILQ_FIRST(rules);
1336 prev = cur;
1337 for (i = 0; i < PF_SKIP_COUNT; ++i)
1338 head[i] = cur;
1339 while (cur != NULL) {
1340
1341 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
1342 PF_SET_SKIP_STEPS(PF_SKIP_IFP);
1343 if (cur->direction != prev->direction)
1344 PF_SET_SKIP_STEPS(PF_SKIP_DIR);
1345 if (cur->af != prev->af)
1346 PF_SET_SKIP_STEPS(PF_SKIP_AF);
1347 if (cur->proto != prev->proto)
1348 PF_SET_SKIP_STEPS(PF_SKIP_PROTO);
1349 if (cur->src.neg != prev->src.neg ||
1350 pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr))
1351 PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR);
1352 if (cur->src.port[0] != prev->src.port[0] ||
1353 cur->src.port[1] != prev->src.port[1] ||
1354 cur->src.port_op != prev->src.port_op)
1355 PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
1356 if (cur->dst.neg != prev->dst.neg ||
1357 pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr))
1358 PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR);
1359 if (cur->dst.port[0] != prev->dst.port[0] ||
1360 cur->dst.port[1] != prev->dst.port[1] ||
1361 cur->dst.port_op != prev->dst.port_op)
1362 PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
1363
1364 prev = cur;
1365 cur = TAILQ_NEXT(cur, entries);
1366 }
1367 for (i = 0; i < PF_SKIP_COUNT; ++i)
1368 PF_SET_SKIP_STEPS(i);
1369 }
1370
1371 int
1372 pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2)
1373 {
1374 if (aw1->type != aw2->type)
1375 return (1);
1376 switch (aw1->type) {
1377 case PF_ADDR_ADDRMASK:
1378 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0))
1379 return (1);
1380 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0))
1381 return (1);
1382 return (0);
1383 case PF_ADDR_DYNIFTL:
1384 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt);
1385 case PF_ADDR_NOROUTE:
1386 case PF_ADDR_URPFFAILED:
1387 return (0);
1388 case PF_ADDR_TABLE:
1389 return (aw1->p.tbl != aw2->p.tbl);
1390 case PF_ADDR_RTLABEL:
1391 return (aw1->v.rtlabel != aw2->v.rtlabel);
1392 default:
1393 printf("invalid address type: %d\n", aw1->type);
1394 return (1);
1395 }
1396 }
1397
1398 u_int16_t
1399 pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp)
1400 {
1401 u_int32_t l;
1402
1403 if (udp && !cksum)
1404 return (0x0000);
1405 l = cksum + old - new;
1406 l = (l >> 16) + (l & 65535);
1407 l = l & 65535;
1408 if (udp && !l)
1409 return (0xFFFF);
1410 return (l);
1411 }
1412
1413 void
1414 pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc,
1415 struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af)
1416 {
1417 struct pf_addr ao;
1418 u_int16_t po = *p;
1419
1420 PF_ACPY(&ao, a, af);
1421 PF_ACPY(a, an, af);
1422
1423 *p = pn;
1424
1425 switch (af) {
1426 #ifdef INET
1427 case AF_INET:
1428 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1429 ao.addr16[0], an->addr16[0], 0),
1430 ao.addr16[1], an->addr16[1], 0);
1431 *p = pn;
1432 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1433 ao.addr16[0], an->addr16[0], u),
1434 ao.addr16[1], an->addr16[1], u),
1435 po, pn, u);
1436 break;
1437 #endif /* INET */
1438 #ifdef INET6
1439 case AF_INET6:
1440 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1441 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1442 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1443 ao.addr16[0], an->addr16[0], u),
1444 ao.addr16[1], an->addr16[1], u),
1445 ao.addr16[2], an->addr16[2], u),
1446 ao.addr16[3], an->addr16[3], u),
1447 ao.addr16[4], an->addr16[4], u),
1448 ao.addr16[5], an->addr16[5], u),
1449 ao.addr16[6], an->addr16[6], u),
1450 ao.addr16[7], an->addr16[7], u),
1451 po, pn, u);
1452 break;
1453 #endif /* INET6 */
1454 }
1455 }
1456
1457
1458 /* Changes a u_int32_t. Uses a void * so there are no align restrictions */
1459 void
1460 pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u)
1461 {
1462 u_int32_t ao;
1463
1464 memcpy(&ao, a, sizeof(ao));
1465 memcpy(a, &an, sizeof(u_int32_t));
1466 *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u),
1467 ao % 65536, an % 65536, u);
1468 }
1469
1470 #ifdef INET6
1471 void
1472 pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u)
1473 {
1474 struct pf_addr ao;
1475
1476 PF_ACPY(&ao, a, AF_INET6);
1477 PF_ACPY(a, an, AF_INET6);
1478
1479 *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1480 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1481 pf_cksum_fixup(pf_cksum_fixup(*c,
1482 ao.addr16[0], an->addr16[0], u),
1483 ao.addr16[1], an->addr16[1], u),
1484 ao.addr16[2], an->addr16[2], u),
1485 ao.addr16[3], an->addr16[3], u),
1486 ao.addr16[4], an->addr16[4], u),
1487 ao.addr16[5], an->addr16[5], u),
1488 ao.addr16[6], an->addr16[6], u),
1489 ao.addr16[7], an->addr16[7], u);
1490 }
1491 #endif /* INET6 */
1492
1493 void
1494 pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa,
1495 struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c,
1496 u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af)
1497 {
1498 struct pf_addr oia, ooa;
1499
1500 PF_ACPY(&oia, ia, af);
1501 PF_ACPY(&ooa, oa, af);
1502
1503 /* Change inner protocol port, fix inner protocol checksum. */
1504 if (ip != NULL) {
1505 u_int16_t oip = *ip;
1506 u_int32_t opc = 0;
1507
1508 if (pc != NULL)
1509 opc = *pc;
1510 *ip = np;
1511 if (pc != NULL)
1512 *pc = pf_cksum_fixup(*pc, oip, *ip, u);
1513 *ic = pf_cksum_fixup(*ic, oip, *ip, 0);
1514 if (pc != NULL)
1515 *ic = pf_cksum_fixup(*ic, opc, *pc, 0);
1516 }
1517 /* Change inner ip address, fix inner ip and icmp checksums. */
1518 PF_ACPY(ia, na, af);
1519 switch (af) {
1520 #ifdef INET
1521 case AF_INET: {
1522 u_int32_t oh2c = *h2c;
1523
1524 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c,
1525 oia.addr16[0], ia->addr16[0], 0),
1526 oia.addr16[1], ia->addr16[1], 0);
1527 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1528 oia.addr16[0], ia->addr16[0], 0),
1529 oia.addr16[1], ia->addr16[1], 0);
1530 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0);
1531 break;
1532 }
1533 #endif /* INET */
1534 #ifdef INET6
1535 case AF_INET6:
1536 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1537 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1538 pf_cksum_fixup(pf_cksum_fixup(*ic,
1539 oia.addr16[0], ia->addr16[0], u),
1540 oia.addr16[1], ia->addr16[1], u),
1541 oia.addr16[2], ia->addr16[2], u),
1542 oia.addr16[3], ia->addr16[3], u),
1543 oia.addr16[4], ia->addr16[4], u),
1544 oia.addr16[5], ia->addr16[5], u),
1545 oia.addr16[6], ia->addr16[6], u),
1546 oia.addr16[7], ia->addr16[7], u);
1547 break;
1548 #endif /* INET6 */
1549 }
1550 /* Change outer ip address, fix outer ip or icmpv6 checksum. */
1551 PF_ACPY(oa, na, af);
1552 switch (af) {
1553 #ifdef INET
1554 case AF_INET:
1555 *hc = pf_cksum_fixup(pf_cksum_fixup(*hc,
1556 ooa.addr16[0], oa->addr16[0], 0),
1557 ooa.addr16[1], oa->addr16[1], 0);
1558 break;
1559 #endif /* INET */
1560 #ifdef INET6
1561 case AF_INET6:
1562 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1563 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1564 pf_cksum_fixup(pf_cksum_fixup(*ic,
1565 ooa.addr16[0], oa->addr16[0], u),
1566 ooa.addr16[1], oa->addr16[1], u),
1567 ooa.addr16[2], oa->addr16[2], u),
1568 ooa.addr16[3], oa->addr16[3], u),
1569 ooa.addr16[4], oa->addr16[4], u),
1570 ooa.addr16[5], oa->addr16[5], u),
1571 ooa.addr16[6], oa->addr16[6], u),
1572 ooa.addr16[7], oa->addr16[7], u);
1573 break;
1574 #endif /* INET6 */
1575 }
1576 }
1577
1578
1579 /*
1580 * Need to modulate the sequence numbers in the TCP SACK option
1581 * (credits to Krzysztof Pfaff for report and patch)
1582 */
1583 int
1584 pf_modulate_sack(struct mbuf *m, int off, struct pf_pdesc *pd,
1585 struct tcphdr *th, struct pf_state_peer *dst)
1586 {
1587 int hlen = (th->th_off << 2) - sizeof(*th), thoptlen = hlen;
1588 u_int8_t opts[MAX_TCPOPTLEN], *opt = opts;
1589 int copyback = 0, i, olen;
1590 struct sackblk sack;
1591
1592 #ifdef __NetBSD__
1593 #define TCPOLEN_SACK 8 /* 2*sizeof(tcp_seq) */
1594 #endif
1595
1596 #define TCPOLEN_SACKLEN (TCPOLEN_SACK + 2)
1597 if (hlen < TCPOLEN_SACKLEN ||
1598 !pf_pull_hdr(m, off + sizeof(*th), opts, hlen, NULL, NULL, pd->af))
1599 return 0;
1600
1601 while (hlen >= TCPOLEN_SACKLEN) {
1602 olen = opt[1];
1603 switch (*opt) {
1604 case TCPOPT_EOL: /* FALLTHROUGH */
1605 case TCPOPT_NOP:
1606 opt++;
1607 hlen--;
1608 break;
1609 case TCPOPT_SACK:
1610 if (olen > hlen)
1611 olen = hlen;
1612 if (olen >= TCPOLEN_SACKLEN) {
1613 for (i = 2; i + TCPOLEN_SACK <= olen;
1614 i += TCPOLEN_SACK) {
1615 memcpy(&sack, &opt[i], sizeof(sack));
1616 #ifdef __NetBSD__
1617 #define SACK_START sack.left
1618 #define SACK_END sack.right
1619 #else
1620 #define SACK_START sack.start
1621 #define SACK_END sack.end
1622 #endif
1623 pf_change_a(&SACK_START, &th->th_sum,
1624 htonl(ntohl(SACK_START) -
1625 dst->seqdiff), 0);
1626 pf_change_a(&SACK_END, &th->th_sum,
1627 htonl(ntohl(SACK_END) -
1628 dst->seqdiff), 0);
1629 #undef SACK_START
1630 #undef SACK_END
1631 memcpy(&opt[i], &sack, sizeof(sack));
1632 }
1633 copyback = 1;
1634 }
1635 /* FALLTHROUGH */
1636 default:
1637 if (olen < 2)
1638 olen = 2;
1639 hlen -= olen;
1640 opt += olen;
1641 }
1642 }
1643
1644 if (copyback)
1645 m_copyback(m, off + sizeof(*th), thoptlen, opts);
1646 return (copyback);
1647 }
1648
1649 void
1650 pf_send_tcp(const struct pf_rule *r, sa_family_t af,
1651 const struct pf_addr *saddr, const struct pf_addr *daddr,
1652 u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack,
1653 u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl, int tag,
1654 u_int16_t rtag, struct ether_header *eh, struct ifnet *ifp)
1655 {
1656 struct mbuf *m;
1657 int len, tlen;
1658 #ifdef INET
1659 struct ip *h = NULL;
1660 #endif /* INET */
1661 #ifdef INET6
1662 struct ip6_hdr *h6 = NULL;
1663 #endif /* INET6 */
1664 struct tcphdr *th;
1665 char *opt;
1666 #ifdef __NetBSD__
1667 struct pf_mtag *pf_mtag;
1668 #endif /* __NetBSD__ */
1669
1670 /* maximum segment size tcp option */
1671 tlen = sizeof(struct tcphdr);
1672 if (mss)
1673 tlen += 4;
1674
1675 switch (af) {
1676 #ifdef INET
1677 case AF_INET:
1678 len = sizeof(struct ip) + tlen;
1679 break;
1680 #endif /* INET */
1681 #ifdef INET6
1682 case AF_INET6:
1683 len = sizeof(struct ip6_hdr) + tlen;
1684 break;
1685 #endif /* INET6 */
1686 default:
1687 return;
1688 }
1689
1690 /* create outgoing mbuf */
1691 m = m_gethdr(M_DONTWAIT, MT_HEADER);
1692 if (m == NULL)
1693 return;
1694 #ifdef __NetBSD__
1695 if ((pf_mtag = pf_get_mtag(m)) == NULL) {
1696 m_freem(m);
1697 return;
1698 }
1699 if (tag)
1700 pf_mtag->flags |= PF_TAG_GENERATED;
1701 pf_mtag->tag = rtag;
1702
1703 if (r != NULL && r->rtableid >= 0)
1704 pf_mtag->rtableid = r->rtableid;
1705 #else
1706 if (tag)
1707 m->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
1708 m->m_pkthdr.pf.tag = rtag;
1709
1710 if (r != NULL && r->rtableid >= 0)
1711 m->m_pkthdr.pf.rtableid = r->rtableid;
1712 #endif /* !__NetBSD__ */
1713
1714 #ifdef ALTQ
1715 if (r != NULL && r->qid) {
1716 #ifdef __NetBSD__
1717 struct m_tag *mtag;
1718 struct altq_tag *atag;
1719
1720 mtag = m_tag_get(PACKET_TAG_ALTQ_QID, sizeof(*atag), M_NOWAIT);
1721 if (mtag != NULL) {
1722 atag = (struct altq_tag *)(mtag + 1);
1723 atag->qid = r->qid;
1724 /* add hints for ecn */
1725 atag->af = af;
1726 atag->hdr = mtod(m, struct ip *);
1727 m_tag_prepend(m, mtag);
1728 }
1729 #else
1730 m->m_pkthdr.pf.qid = r->qid;
1731 /* add hints for ecn */
1732 m->m_pkthdr.pf.hdr = mtod(m, struct ip *);
1733 #endif /* !__NetBSD__ */
1734 }
1735 #endif /* ALTQ */
1736 m->m_data += max_linkhdr;
1737 m->m_pkthdr.len = m->m_len = len;
1738 m_reset_rcvif(m);
1739 bzero(m->m_data, len);
1740 switch (af) {
1741 #ifdef INET
1742 case AF_INET:
1743 h = mtod(m, struct ip *);
1744
1745 /* IP header fields included in the TCP checksum */
1746 h->ip_p = IPPROTO_TCP;
1747 h->ip_len = htons(tlen);
1748 h->ip_src.s_addr = saddr->v4.s_addr;
1749 h->ip_dst.s_addr = daddr->v4.s_addr;
1750
1751 th = (struct tcphdr *)((char *)h + sizeof(struct ip));
1752 break;
1753 #endif /* INET */
1754 #ifdef INET6
1755 case AF_INET6:
1756 h6 = mtod(m, struct ip6_hdr *);
1757
1758 /* IP header fields included in the TCP checksum */
1759 h6->ip6_nxt = IPPROTO_TCP;
1760 h6->ip6_plen = htons(tlen);
1761 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr));
1762 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr));
1763
1764 th = (struct tcphdr *)((char *)h6 + sizeof(struct ip6_hdr));
1765 break;
1766 #endif /* INET6 */
1767 default:
1768 m_freem(m);
1769 return;
1770 }
1771
1772 /* TCP header */
1773 th->th_sport = sport;
1774 th->th_dport = dport;
1775 th->th_seq = htonl(seq);
1776 th->th_ack = htonl(ack);
1777 th->th_off = tlen >> 2;
1778 th->th_flags = flags;
1779 th->th_win = htons(win);
1780
1781 if (mss) {
1782 opt = (char *)(th + 1);
1783 opt[0] = TCPOPT_MAXSEG;
1784 opt[1] = 4;
1785 HTONS(mss);
1786 bcopy((void *)&mss, (void *)(opt + 2), 2);
1787 }
1788
1789 switch (af) {
1790 #ifdef INET
1791 case AF_INET:
1792 /* TCP checksum */
1793 th->th_sum = in_cksum(m, len);
1794
1795 /* Finish the IP header */
1796 h->ip_v = 4;
1797 h->ip_hl = sizeof(*h) >> 2;
1798 h->ip_tos = IPTOS_LOWDELAY;
1799 h->ip_len = htons(len);
1800 h->ip_off = htons(ip_mtudisc ? IP_DF : 0);
1801 h->ip_ttl = ttl ? ttl : ip_defttl;
1802 h->ip_sum = 0;
1803 if (eh == NULL) {
1804 ip_output(m, (void *)NULL, (void *)NULL, 0,
1805 (void *)NULL, (void *)NULL);
1806 } else {
1807 #ifdef __NetBSD__
1808 /*
1809 * On netbsd, pf_test and pf_test6 are always called
1810 * with eh == NULL.
1811 */
1812 panic("pf_send_tcp: eh != NULL");
1813 #else
1814 struct route ro;
1815 struct rtentry rt;
1816 struct ether_header *e = (void *)ro.ro_dst.sa_data;
1817
1818 if (ifp == NULL) {
1819 m_freem(m);
1820 return;
1821 }
1822 rt.rt_ifp = ifp;
1823 ro.ro_rt = &rt;
1824 ro.ro_dst.sa_len = sizeof(ro.ro_dst);
1825 ro.ro_dst.sa_family = pseudo_AF_HDRCMPLT;
1826 bcopy(eh->ether_dhost, e->ether_shost, ETHER_ADDR_LEN);
1827 bcopy(eh->ether_shost, e->ether_dhost, ETHER_ADDR_LEN);
1828 e->ether_type = eh->ether_type;
1829 ip_output(m, (void *)NULL, &ro, IP_ROUTETOETHER,
1830 (void *)NULL, (void *)NULL);
1831 #endif /* !__NetBSD__ */
1832 }
1833 break;
1834 #endif /* INET */
1835 #ifdef INET6
1836 case AF_INET6:
1837 /* TCP checksum */
1838 th->th_sum = in6_cksum(m, IPPROTO_TCP,
1839 sizeof(struct ip6_hdr), tlen);
1840
1841 h6->ip6_vfc |= IPV6_VERSION;
1842 h6->ip6_hlim = IPV6_DEFHLIM;
1843
1844 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
1845 break;
1846 #endif /* INET6 */
1847 }
1848 }
1849
1850 void
1851 pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af,
1852 struct pf_rule *r)
1853 {
1854 struct mbuf *m0;
1855 #ifdef __NetBSD__
1856 struct pf_mtag *pf_mtag;
1857 #endif /* __NetBSD__ */
1858
1859 #ifdef __NetBSD__
1860 m0 = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
1861 #else
1862 m0 = m_copy(m, 0, M_COPYALL);
1863 #endif
1864
1865 #ifdef __NetBSD__
1866 if ((pf_mtag = pf_get_mtag(m0)) == NULL)
1867 return;
1868 pf_mtag->flags |= PF_TAG_GENERATED;
1869
1870 if (r->rtableid >= 0)
1871 pf_mtag->rtableid = r->rtableid;
1872 #else
1873 m0->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
1874
1875 if (r->rtableid >= 0)
1876 m0->m_pkthdr.pf.rtableid = r->rtableid;
1877 #endif /* !__NetBSD__ */
1878
1879 #ifdef ALTQ
1880 if (r->qid) {
1881 #ifdef __NetBSD__
1882 struct m_tag *mtag;
1883 struct altq_tag *atag;
1884
1885 mtag = m_tag_get(PACKET_TAG_ALTQ_QID, sizeof(*atag), M_NOWAIT);
1886 if (mtag != NULL) {
1887 atag = (struct altq_tag *)(mtag + 1);
1888 atag->qid = r->qid;
1889 /* add hints for ecn */
1890 atag->af = af;
1891 atag->hdr = mtod(m0, struct ip *);
1892 m_tag_prepend(m0, mtag);
1893 }
1894 #else
1895 m0->m_pkthdr.pf.qid = r->qid;
1896 /* add hints for ecn */
1897 m0->m_pkthdr.pf.hdr = mtod(m0, struct ip *);
1898 #endif /* !__NetBSD__ */
1899 }
1900 #endif /* ALTQ */
1901
1902 switch (af) {
1903 #ifdef INET
1904 case AF_INET:
1905 icmp_error(m0, type, code, 0, 0);
1906 break;
1907 #endif /* INET */
1908 #ifdef INET6
1909 case AF_INET6:
1910 icmp6_error(m0, type, code, 0);
1911 break;
1912 #endif /* INET6 */
1913 }
1914 }
1915
1916 /*
1917 * Return 1 if the addresses a and b match (with mask m), otherwise return 0.
1918 * If n is 0, they match if they are equal. If n is != 0, they match if they
1919 * are different.
1920 */
1921 int
1922 pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m,
1923 struct pf_addr *b, sa_family_t af)
1924 {
1925 int match = 0;
1926
1927 switch (af) {
1928 #ifdef INET
1929 case AF_INET:
1930 if ((a->addr32[0] & m->addr32[0]) ==
1931 (b->addr32[0] & m->addr32[0]))
1932 match++;
1933 break;
1934 #endif /* INET */
1935 #ifdef INET6
1936 case AF_INET6:
1937 if (((a->addr32[0] & m->addr32[0]) ==
1938 (b->addr32[0] & m->addr32[0])) &&
1939 ((a->addr32[1] & m->addr32[1]) ==
1940 (b->addr32[1] & m->addr32[1])) &&
1941 ((a->addr32[2] & m->addr32[2]) ==
1942 (b->addr32[2] & m->addr32[2])) &&
1943 ((a->addr32[3] & m->addr32[3]) ==
1944 (b->addr32[3] & m->addr32[3])))
1945 match++;
1946 break;
1947 #endif /* INET6 */
1948 }
1949 if (match) {
1950 if (n)
1951 return (0);
1952 else
1953 return (1);
1954 } else {
1955 if (n)
1956 return (1);
1957 else
1958 return (0);
1959 }
1960 }
1961
1962 int
1963 pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p)
1964 {
1965 switch (op) {
1966 case PF_OP_IRG:
1967 return ((p > a1) && (p < a2));
1968 case PF_OP_XRG:
1969 return ((p < a1) || (p > a2));
1970 case PF_OP_RRG:
1971 return ((p >= a1) && (p <= a2));
1972 case PF_OP_EQ:
1973 return (p == a1);
1974 case PF_OP_NE:
1975 return (p != a1);
1976 case PF_OP_LT:
1977 return (p < a1);
1978 case PF_OP_LE:
1979 return (p <= a1);
1980 case PF_OP_GT:
1981 return (p > a1);
1982 case PF_OP_GE:
1983 return (p >= a1);
1984 }
1985 return (0); /* never reached */
1986 }
1987
1988 int
1989 pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p)
1990 {
1991 NTOHS(a1);
1992 NTOHS(a2);
1993 NTOHS(p);
1994 return (pf_match(op, a1, a2, p));
1995 }
1996
1997 int
1998 pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u)
1999 {
2000 if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
2001 return (0);
2002 return (pf_match(op, a1, a2, u));
2003 }
2004
2005 int
2006 pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g)
2007 {
2008 if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
2009 return (0);
2010 return (pf_match(op, a1, a2, g));
2011 }
2012
2013 int
2014 pf_match_tag(struct mbuf *m, struct pf_rule *r, int *tag)
2015 {
2016 #ifdef __NetBSD__
2017 if (*tag == -1) {
2018 struct pf_mtag *pf_mtag = pf_get_mtag(m);
2019 if (pf_mtag == NULL)
2020 return (0);
2021
2022 *tag = pf_mtag->tag;
2023 }
2024 #else
2025 if (*tag == -1)
2026 *tag = m->m_pkthdr.pf.tag;
2027 #endif /* !__NetBSD__ */
2028
2029 return ((!r->match_tag_not && r->match_tag == *tag) ||
2030 (r->match_tag_not && r->match_tag != *tag));
2031 }
2032
2033 int
2034 pf_tag_packet(struct mbuf *m, int tag, int rtableid)
2035 {
2036 if (tag <= 0 && rtableid < 0)
2037 return (0);
2038
2039 #ifdef __NetBSD__
2040 if (tag > 0 || rtableid > 0) {
2041 struct pf_mtag *pf_mtag = pf_get_mtag(m);
2042 if (pf_mtag == NULL)
2043 return (1);
2044
2045 if (tag > 0)
2046 pf_mtag->tag = tag;
2047 if (rtableid > 0)
2048 pf_mtag->rtableid = rtableid;
2049 }
2050 #else
2051 if (tag > 0)
2052 m->m_pkthdr.pf.tag = tag;
2053 if (rtableid >= 0)
2054 m->m_pkthdr.pf.rtableid = rtableid;
2055 #endif /* !__NetBSD__ */
2056
2057 return (0);
2058 }
2059
2060 void
2061 pf_step_into_anchor(int *depth, struct pf_ruleset **rs, int n,
2062 struct pf_rule **r, struct pf_rule **a, int *match)
2063 {
2064 struct pf_anchor_stackframe *f;
2065
2066 (*r)->anchor->match = 0;
2067 if (match)
2068 *match = 0;
2069 if (*depth >= sizeof(pf_anchor_stack) /
2070 sizeof(pf_anchor_stack[0])) {
2071 printf("pf_step_into_anchor: stack overflow\n");
2072 *r = TAILQ_NEXT(*r, entries);
2073 return;
2074 } else if (*depth == 0 && a != NULL)
2075 *a = *r;
2076 f = pf_anchor_stack + (*depth)++;
2077 f->rs = *rs;
2078 f->r = *r;
2079 if ((*r)->anchor_wildcard) {
2080 f->parent = &(*r)->anchor->children;
2081 if ((f->child = RB_MIN(pf_anchor_node, f->parent)) ==
2082 NULL) {
2083 *r = NULL;
2084 return;
2085 }
2086 *rs = &f->child->ruleset;
2087 } else {
2088 f->parent = NULL;
2089 f->child = NULL;
2090 *rs = &(*r)->anchor->ruleset;
2091 }
2092 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
2093 }
2094
2095 int
2096 pf_step_out_of_anchor(int *depth, struct pf_ruleset **rs, int n,
2097 struct pf_rule **r, struct pf_rule **a, int *match)
2098 {
2099 struct pf_anchor_stackframe *f;
2100 int quick = 0;
2101
2102 do {
2103 if (*depth <= 0)
2104 break;
2105 f = pf_anchor_stack + *depth - 1;
2106 if (f->parent != NULL && f->child != NULL) {
2107 if (f->child->match ||
2108 (match != NULL && *match)) {
2109 f->r->anchor->match = 1;
2110 *match = 0;
2111 }
2112 f->child = RB_NEXT(pf_anchor_node, f->parent, f->child);
2113 if (f->child != NULL) {
2114 *rs = &f->child->ruleset;
2115 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
2116 if (*r == NULL)
2117 continue;
2118 else
2119 break;
2120 }
2121 }
2122 (*depth)--;
2123 if (*depth == 0 && a != NULL)
2124 *a = NULL;
2125 *rs = f->rs;
2126 if (f->r->anchor->match || (match != NULL && *match))
2127 quick = f->r->quick;
2128 *r = TAILQ_NEXT(f->r, entries);
2129 } while (*r == NULL);
2130
2131 return (quick);
2132 }
2133
2134 #ifdef INET6
2135 void
2136 pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr,
2137 struct pf_addr *rmask, const struct pf_addr *saddr, sa_family_t af)
2138 {
2139 switch (af) {
2140 #ifdef INET
2141 case AF_INET:
2142 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2143 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2144 break;
2145 #endif /* INET */
2146 case AF_INET6:
2147 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2148 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2149 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) |
2150 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]);
2151 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) |
2152 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]);
2153 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) |
2154 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]);
2155 break;
2156 }
2157 }
2158
2159 void
2160 pf_addr_inc(struct pf_addr *addr, sa_family_t af)
2161 {
2162 switch (af) {
2163 #ifdef INET
2164 case AF_INET:
2165 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1);
2166 break;
2167 #endif /* INET */
2168 case AF_INET6:
2169 if (addr->addr32[3] == 0xffffffff) {
2170 addr->addr32[3] = 0;
2171 if (addr->addr32[2] == 0xffffffff) {
2172 addr->addr32[2] = 0;
2173 if (addr->addr32[1] == 0xffffffff) {
2174 addr->addr32[1] = 0;
2175 addr->addr32[0] =
2176 htonl(ntohl(addr->addr32[0]) + 1);
2177 } else
2178 addr->addr32[1] =
2179 htonl(ntohl(addr->addr32[1]) + 1);
2180 } else
2181 addr->addr32[2] =
2182 htonl(ntohl(addr->addr32[2]) + 1);
2183 } else
2184 addr->addr32[3] =
2185 htonl(ntohl(addr->addr32[3]) + 1);
2186 break;
2187 }
2188 }
2189 #endif /* INET6 */
2190
2191 #define mix(a,b,c) \
2192 do { \
2193 a -= b; a -= c; a ^= (c >> 13); \
2194 b -= c; b -= a; b ^= (a << 8); \
2195 c -= a; c -= b; c ^= (b >> 13); \
2196 a -= b; a -= c; a ^= (c >> 12); \
2197 b -= c; b -= a; b ^= (a << 16); \
2198 c -= a; c -= b; c ^= (b >> 5); \
2199 a -= b; a -= c; a ^= (c >> 3); \
2200 b -= c; b -= a; b ^= (a << 10); \
2201 c -= a; c -= b; c ^= (b >> 15); \
2202 } while (0)
2203
2204 /*
2205 * hash function based on bridge_hash in if_bridge.c
2206 */
2207 void
2208 pf_hash(const struct pf_addr *inaddr, struct pf_addr *hash,
2209 struct pf_poolhashkey *key, sa_family_t af)
2210 {
2211 u_int32_t a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0];
2212
2213 switch (af) {
2214 #ifdef INET
2215 case AF_INET:
2216 a += inaddr->addr32[0];
2217 b += key->key32[1];
2218 mix(a, b, c);
2219 hash->addr32[0] = c + key->key32[2];
2220 break;
2221 #endif /* INET */
2222 #ifdef INET6
2223 case AF_INET6:
2224 a += inaddr->addr32[0];
2225 b += inaddr->addr32[2];
2226 mix(a, b, c);
2227 hash->addr32[0] = c;
2228 a += inaddr->addr32[1];
2229 b += inaddr->addr32[3];
2230 c += key->key32[1];
2231 mix(a, b, c);
2232 hash->addr32[1] = c;
2233 a += inaddr->addr32[2];
2234 b += inaddr->addr32[1];
2235 c += key->key32[2];
2236 mix(a, b, c);
2237 hash->addr32[2] = c;
2238 a += inaddr->addr32[3];
2239 b += inaddr->addr32[0];
2240 c += key->key32[3];
2241 mix(a, b, c);
2242 hash->addr32[3] = c;
2243 break;
2244 #endif /* INET6 */
2245 }
2246 }
2247
2248 int
2249 pf_map_addr(sa_family_t af, struct pf_rule *r, const struct pf_addr *saddr,
2250 struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn)
2251 {
2252 unsigned char hash[16];
2253 struct pf_pool *rpool = &r->rpool;
2254 struct pf_addr *raddr = &rpool->cur->addr.v.a.addr;
2255 struct pf_addr *rmask = &rpool->cur->addr.v.a.mask;
2256 struct pf_pooladdr *acur = rpool->cur;
2257 struct pf_src_node k;
2258
2259 if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR &&
2260 (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2261 k.af = af;
2262 PF_ACPY(&k.addr, saddr, af);
2263 if (r->rule_flag & PFRULE_RULESRCTRACK ||
2264 r->rpool.opts & PF_POOL_STICKYADDR)
2265 k.rule.ptr = r;
2266 else
2267 k.rule.ptr = NULL;
2268 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
2269 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
2270 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) {
2271 PF_ACPY(naddr, &(*sn)->raddr, af);
2272 if (pf_status.debug >= PF_DEBUG_MISC) {
2273 printf("pf_map_addr: src tracking maps ");
2274 pf_print_host(&k.addr, 0, af);
2275 printf(" to ");
2276 pf_print_host(naddr, 0, af);
2277 printf("\n");
2278 }
2279 return (0);
2280 }
2281 }
2282
2283 if (rpool->cur->addr.type == PF_ADDR_NOROUTE)
2284 return (1);
2285 if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2286 switch (af) {
2287 #ifdef INET
2288 case AF_INET:
2289 if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 &&
2290 (rpool->opts & PF_POOL_TYPEMASK) !=
2291 PF_POOL_ROUNDROBIN)
2292 return (1);
2293 raddr = &rpool->cur->addr.p.dyn->pfid_addr4;
2294 rmask = &rpool->cur->addr.p.dyn->pfid_mask4;
2295 break;
2296 #endif /* INET */
2297 #ifdef INET6
2298 case AF_INET6:
2299 if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 &&
2300 (rpool->opts & PF_POOL_TYPEMASK) !=
2301 PF_POOL_ROUNDROBIN)
2302 return (1);
2303 raddr = &rpool->cur->addr.p.dyn->pfid_addr6;
2304 rmask = &rpool->cur->addr.p.dyn->pfid_mask6;
2305 break;
2306 #endif /* INET6 */
2307 }
2308 } else if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2309 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN)
2310 return (1); /* unsupported */
2311 } else {
2312 raddr = &rpool->cur->addr.v.a.addr;
2313 rmask = &rpool->cur->addr.v.a.mask;
2314 }
2315
2316 switch (rpool->opts & PF_POOL_TYPEMASK) {
2317 case PF_POOL_NONE:
2318 PF_ACPY(naddr, raddr, af);
2319 break;
2320 case PF_POOL_BITMASK:
2321 PF_POOLMASK(naddr, raddr, rmask, saddr, af);
2322 break;
2323 case PF_POOL_RANDOM:
2324 if (init_addr != NULL && PF_AZERO(init_addr, af)) {
2325 switch (af) {
2326 #ifdef INET
2327 case AF_INET:
2328 rpool->counter.addr32[0] =
2329 htonl(cprng_fast32());
2330 break;
2331 #endif /* INET */
2332 #ifdef INET6
2333 case AF_INET6:
2334 if (rmask->addr32[3] != 0xffffffff)
2335 rpool->counter.addr32[3] =
2336 htonl(cprng_fast32());
2337 else
2338 break;
2339 if (rmask->addr32[2] != 0xffffffff)
2340 rpool->counter.addr32[2] =
2341 htonl(cprng_fast32());
2342 else
2343 break;
2344 if (rmask->addr32[1] != 0xffffffff)
2345 rpool->counter.addr32[1] =
2346 htonl(cprng_fast32());
2347 else
2348 break;
2349 if (rmask->addr32[0] != 0xffffffff)
2350 rpool->counter.addr32[0] =
2351 htonl(cprng_fast32());
2352 break;
2353 #endif /* INET6 */
2354 }
2355 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2356 PF_ACPY(init_addr, naddr, af);
2357
2358 } else {
2359 PF_AINC(&rpool->counter, af);
2360 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2361 }
2362 break;
2363 case PF_POOL_SRCHASH:
2364 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af);
2365 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af);
2366 break;
2367 case PF_POOL_ROUNDROBIN:
2368 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2369 if (!pfr_pool_get(rpool->cur->addr.p.tbl,
2370 &rpool->tblidx, &rpool->counter,
2371 &raddr, &rmask, af))
2372 goto get_addr;
2373 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2374 if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2375 &rpool->tblidx, &rpool->counter,
2376 &raddr, &rmask, af))
2377 goto get_addr;
2378 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af))
2379 goto get_addr;
2380
2381 try_next:
2382 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL)
2383 rpool->cur = TAILQ_FIRST(&rpool->list);
2384 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2385 rpool->tblidx = -1;
2386 if (pfr_pool_get(rpool->cur->addr.p.tbl,
2387 &rpool->tblidx, &rpool->counter,
2388 &raddr, &rmask, af)) {
2389 /* table contains no address of type 'af' */
2390 if (rpool->cur != acur)
2391 goto try_next;
2392 return (1);
2393 }
2394 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2395 rpool->tblidx = -1;
2396 if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2397 &rpool->tblidx, &rpool->counter,
2398 &raddr, &rmask, af)) {
2399 /* table contains no address of type 'af' */
2400 if (rpool->cur != acur)
2401 goto try_next;
2402 return (1);
2403 }
2404 } else {
2405 raddr = &rpool->cur->addr.v.a.addr;
2406 rmask = &rpool->cur->addr.v.a.mask;
2407 PF_ACPY(&rpool->counter, raddr, af);
2408 }
2409
2410 get_addr:
2411 PF_ACPY(naddr, &rpool->counter, af);
2412 if (init_addr != NULL && PF_AZERO(init_addr, af))
2413 PF_ACPY(init_addr, naddr, af);
2414 PF_AINC(&rpool->counter, af);
2415 break;
2416 }
2417 if (*sn != NULL)
2418 PF_ACPY(&(*sn)->raddr, naddr, af);
2419
2420 if (pf_status.debug >= PF_DEBUG_MISC &&
2421 (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2422 printf("pf_map_addr: selected address ");
2423 pf_print_host(naddr, 0, af);
2424 printf("\n");
2425 }
2426
2427 return (0);
2428 }
2429
2430 int
2431 pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r,
2432 struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport,
2433 struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high,
2434 struct pf_src_node **sn)
2435 {
2436 struct pf_state_key_cmp key;
2437 struct pf_addr init_addr;
2438 u_int16_t cut;
2439
2440 bzero(&init_addr, sizeof(init_addr));
2441 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2442 return (1);
2443
2444 if (proto == IPPROTO_ICMP) {
2445 low = 1;
2446 high = 65535;
2447 }
2448
2449 do {
2450 key.af = af;
2451 key.proto = proto;
2452 PF_ACPY(&key.ext.addr, daddr, key.af);
2453 PF_ACPY(&key.gwy.addr, naddr, key.af);
2454 key.ext.port = dport;
2455
2456 /*
2457 * port search; start random, step;
2458 * similar 2 portloop in inpcb_bind
2459 */
2460 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP ||
2461 proto == IPPROTO_ICMP)) {
2462 key.gwy.port = dport;
2463 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
2464 return (0);
2465 } else if (low == 0 && high == 0) {
2466 key.gwy.port = *nport;
2467 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
2468 return (0);
2469 } else if (low == high) {
2470 key.gwy.port = htons(low);
2471 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) {
2472 *nport = htons(low);
2473 return (0);
2474 }
2475 } else {
2476 u_int16_t tmp;
2477
2478 if (low > high) {
2479 tmp = low;
2480 low = high;
2481 high = tmp;
2482 }
2483 /* low < high */
2484 cut = htonl(cprng_fast32()) % (1 + high - low) + low;
2485 /* low <= cut <= high */
2486 for (tmp = cut; tmp <= high; ++(tmp)) {
2487 key.gwy.port = htons(tmp);
2488 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
2489 NULL) {
2490 *nport = htons(tmp);
2491 return (0);
2492 }
2493 }
2494 for (tmp = cut - 1; tmp >= low; --(tmp)) {
2495 key.gwy.port = htons(tmp);
2496 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
2497 NULL) {
2498 *nport = htons(tmp);
2499 return (0);
2500 }
2501 }
2502 }
2503
2504 switch (r->rpool.opts & PF_POOL_TYPEMASK) {
2505 case PF_POOL_RANDOM:
2506 case PF_POOL_ROUNDROBIN:
2507 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2508 return (1);
2509 break;
2510 case PF_POOL_NONE:
2511 case PF_POOL_SRCHASH:
2512 case PF_POOL_BITMASK:
2513 default:
2514 return (1);
2515 }
2516 } while (! PF_AEQ(&init_addr, naddr, af) );
2517
2518 return (1); /* none available */
2519 }
2520
2521 struct pf_rule *
2522 pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off,
2523 int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport,
2524 struct pf_addr *daddr, u_int16_t dport, int rs_num)
2525 {
2526 struct pf_rule *r, *rm = NULL;
2527 struct pf_ruleset *ruleset = NULL;
2528 int tag = -1;
2529 int rtableid = -1;
2530 int asd = 0;
2531
2532 r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr);
2533 while (r && rm == NULL) {
2534 struct pf_rule_addr *src = NULL, *dst = NULL;
2535 struct pf_addr_wrap *xdst = NULL;
2536
2537 if (r->action == PF_BINAT && direction == PF_IN) {
2538 src = &r->dst;
2539 if (r->rpool.cur != NULL)
2540 xdst = &r->rpool.cur->addr;
2541 } else {
2542 src = &r->src;
2543 dst = &r->dst;
2544 }
2545
2546 r->evaluations++;
2547 if (pfi_kif_match(r->kif, kif) == r->ifnot)
2548 r = r->skip[PF_SKIP_IFP].ptr;
2549 else if (r->direction && r->direction != direction)
2550 r = r->skip[PF_SKIP_DIR].ptr;
2551 else if (r->af && r->af != pd->af)
2552 r = r->skip[PF_SKIP_AF].ptr;
2553 else if (r->proto && r->proto != pd->proto)
2554 r = r->skip[PF_SKIP_PROTO].ptr;
2555 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af,
2556 src->neg, kif))
2557 r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR :
2558 PF_SKIP_DST_ADDR].ptr;
2559 else if (src->port_op && !pf_match_port(src->port_op,
2560 src->port[0], src->port[1], sport))
2561 r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT :
2562 PF_SKIP_DST_PORT].ptr;
2563 else if (dst != NULL &&
2564 PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->neg, NULL))
2565 r = r->skip[PF_SKIP_DST_ADDR].ptr;
2566 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af,
2567 0, NULL))
2568 r = TAILQ_NEXT(r, entries);
2569 else if (dst != NULL && dst->port_op &&
2570 !pf_match_port(dst->port_op, dst->port[0],
2571 dst->port[1], dport))
2572 r = r->skip[PF_SKIP_DST_PORT].ptr;
2573 else if (r->match_tag && !pf_match_tag(m, r, &tag))
2574 r = TAILQ_NEXT(r, entries);
2575 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto !=
2576 IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m,
2577 off, pd->hdr.tcp), r->os_fingerprint)))
2578 r = TAILQ_NEXT(r, entries);
2579 else {
2580 if (r->tag)
2581 tag = r->tag;
2582 if (r->rtableid >= 0)
2583 rtableid = r->rtableid;
2584 if (r->anchor == NULL) {
2585 rm = r;
2586 } else
2587 pf_step_into_anchor(&asd, &ruleset, rs_num,
2588 &r, NULL, NULL);
2589 }
2590 if (r == NULL)
2591 pf_step_out_of_anchor(&asd, &ruleset, rs_num, &r,
2592 NULL, NULL);
2593 }
2594 if (pf_tag_packet(m, tag, rtableid))
2595 return (NULL);
2596 if (rm != NULL && (rm->action == PF_NONAT ||
2597 rm->action == PF_NORDR || rm->action == PF_NOBINAT))
2598 return (NULL);
2599 return (rm);
2600 }
2601
2602 struct pf_rule *
2603 pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction,
2604 struct pfi_kif *kif, struct pf_src_node **sn,
2605 struct pf_addr *saddr, u_int16_t sport,
2606 struct pf_addr *daddr, u_int16_t dport,
2607 struct pf_addr *naddr, u_int16_t *nport)
2608 {
2609 struct pf_rule *r = NULL;
2610
2611 if (direction == PF_OUT) {
2612 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2613 sport, daddr, dport, PF_RULESET_BINAT);
2614 if (r == NULL)
2615 r = pf_match_translation(pd, m, off, direction, kif,
2616 saddr, sport, daddr, dport, PF_RULESET_NAT);
2617 } else {
2618 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2619 sport, daddr, dport, PF_RULESET_RDR);
2620 if (r == NULL)
2621 r = pf_match_translation(pd, m, off, direction, kif,
2622 saddr, sport, daddr, dport, PF_RULESET_BINAT);
2623 }
2624
2625 if (r != NULL) {
2626 switch (r->action) {
2627 case PF_NONAT:
2628 case PF_NOBINAT:
2629 case PF_NORDR:
2630 return (NULL);
2631 case PF_NAT:
2632 if (pf_get_sport(pd->af, pd->proto, r, saddr,
2633 daddr, dport, naddr, nport, r->rpool.proxy_port[0],
2634 r->rpool.proxy_port[1], sn)) {
2635 DPFPRINTF(PF_DEBUG_MISC,
2636 ("pf: NAT proxy port allocation "
2637 "(%u-%u) failed\n",
2638 r->rpool.proxy_port[0],
2639 r->rpool.proxy_port[1]));
2640 return (NULL);
2641 }
2642 break;
2643 case PF_BINAT:
2644 switch (direction) {
2645 case PF_OUT:
2646 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){
2647 switch (pd->af) {
2648 #ifdef INET
2649 case AF_INET:
2650 if (r->rpool.cur->addr.p.dyn->
2651 pfid_acnt4 < 1)
2652 return (NULL);
2653 PF_POOLMASK(naddr,
2654 &r->rpool.cur->addr.p.dyn->
2655 pfid_addr4,
2656 &r->rpool.cur->addr.p.dyn->
2657 pfid_mask4,
2658 saddr, AF_INET);
2659 break;
2660 #endif /* INET */
2661 #ifdef INET6
2662 case AF_INET6:
2663 if (r->rpool.cur->addr.p.dyn->
2664 pfid_acnt6 < 1)
2665 return (NULL);
2666 PF_POOLMASK(naddr,
2667 &r->rpool.cur->addr.p.dyn->
2668 pfid_addr6,
2669 &r->rpool.cur->addr.p.dyn->
2670 pfid_mask6,
2671 saddr, AF_INET6);
2672 break;
2673 #endif /* INET6 */
2674 }
2675 } else
2676 PF_POOLMASK(naddr,
2677 &r->rpool.cur->addr.v.a.addr,
2678 &r->rpool.cur->addr.v.a.mask,
2679 saddr, pd->af);
2680 break;
2681 case PF_IN:
2682 if (r->src.addr.type == PF_ADDR_DYNIFTL) {
2683 switch (pd->af) {
2684 #ifdef INET
2685 case AF_INET:
2686 if (r->src.addr.p.dyn->
2687 pfid_acnt4 < 1)
2688 return (NULL);
2689 PF_POOLMASK(naddr,
2690 &r->src.addr.p.dyn->
2691 pfid_addr4,
2692 &r->src.addr.p.dyn->
2693 pfid_mask4,
2694 daddr, AF_INET);
2695 break;
2696 #endif /* INET */
2697 #ifdef INET6
2698 case AF_INET6:
2699 if (r->src.addr.p.dyn->
2700 pfid_acnt6 < 1)
2701 return (NULL);
2702 PF_POOLMASK(naddr,
2703 &r->src.addr.p.dyn->
2704 pfid_addr6,
2705 &r->src.addr.p.dyn->
2706 pfid_mask6,
2707 daddr, AF_INET6);
2708 break;
2709 #endif /* INET6 */
2710 }
2711 } else
2712 PF_POOLMASK(naddr,
2713 &r->src.addr.v.a.addr,
2714 &r->src.addr.v.a.mask, daddr,
2715 pd->af);
2716 break;
2717 }
2718 break;
2719 case PF_RDR: {
2720 if (pf_map_addr(pd->af, r, saddr, naddr, NULL, sn))
2721 return (NULL);
2722 if ((r->rpool.opts & PF_POOL_TYPEMASK) ==
2723 PF_POOL_BITMASK)
2724 PF_POOLMASK(naddr, naddr,
2725 &r->rpool.cur->addr.v.a.mask, daddr,
2726 pd->af);
2727
2728 if (r->rpool.proxy_port[1]) {
2729 u_int32_t tmp_nport;
2730
2731 tmp_nport = ((ntohs(dport) -
2732 ntohs(r->dst.port[0])) %
2733 (r->rpool.proxy_port[1] -
2734 r->rpool.proxy_port[0] + 1)) +
2735 r->rpool.proxy_port[0];
2736
2737 /* wrap around if necessary */
2738 if (tmp_nport > 65535)
2739 tmp_nport -= 65535;
2740 *nport = htons((u_int16_t)tmp_nport);
2741 } else if (r->rpool.proxy_port[0])
2742 *nport = htons(r->rpool.proxy_port[0]);
2743 break;
2744 }
2745 default:
2746 return (NULL);
2747 }
2748 }
2749
2750 return (r);
2751 }
2752
2753 int
2754 pf_socket_lookup(int direction, struct pf_pdesc *pd)
2755 {
2756 struct pf_addr *saddr, *daddr;
2757 u_int16_t sport, dport;
2758 struct inpcbtable *tb;
2759 struct inpcb *inp = NULL;
2760 struct socket *so = NULL;
2761 #define in6p inp
2762 #define in6p_socket inp_socket
2763
2764 if (pd == NULL)
2765 return (-1);
2766 pd->lookup.uid = UID_MAX;
2767 pd->lookup.gid = GID_MAX;
2768 pd->lookup.pid = NO_PID;
2769 switch (pd->proto) {
2770 case IPPROTO_TCP:
2771 if (pd->hdr.tcp == NULL)
2772 return (-1);
2773 sport = pd->hdr.tcp->th_sport;
2774 dport = pd->hdr.tcp->th_dport;
2775 tb = &tcbtable;
2776 break;
2777 case IPPROTO_UDP:
2778 if (pd->hdr.udp == NULL)
2779 return (-1);
2780 sport = pd->hdr.udp->uh_sport;
2781 dport = pd->hdr.udp->uh_dport;
2782 tb = &udbtable;
2783 break;
2784 default:
2785 return (-1);
2786 }
2787 if (direction == PF_IN) {
2788 saddr = pd->src;
2789 daddr = pd->dst;
2790 } else {
2791 u_int16_t p;
2792
2793 p = sport;
2794 sport = dport;
2795 dport = p;
2796 saddr = pd->dst;
2797 daddr = pd->src;
2798 }
2799 switch (pd->af) {
2800
2801 #ifdef __NetBSD__
2802 #define in_pcbhashlookup(tbl, saddr, sport, daddr, dport) \
2803 inpcb_lookup(tbl, saddr, sport, daddr, dport, NULL)
2804 #define in6_pcbhashlookup(tbl, saddr, sport, daddr, dport) \
2805 in6pcb_lookup(tbl, saddr, sport, daddr, dport, 0, NULL)
2806 #define in_pcblookup_listen(tbl, addr, port, zero) \
2807 inpcb_lookup_bound(tbl, addr, port)
2808 #define in6_pcblookup_listen(tbl, addr, port, zero) \
2809 in6pcb_lookup_bound(tbl, addr, port, zero)
2810 #endif
2811
2812 #ifdef INET
2813 case AF_INET:
2814 inp = in_pcbhashlookup(tb, saddr->v4, sport, daddr->v4, dport);
2815 if (inp == NULL) {
2816 inp = in_pcblookup_listen(tb, daddr->v4, dport, 0);
2817 if (inp == NULL)
2818 return (-1);
2819 }
2820 break;
2821 #endif /* INET */
2822 #ifdef INET6
2823 case AF_INET6:
2824 in6p = in6_pcbhashlookup(tb, &saddr->v6, sport, &daddr->v6,
2825 dport);
2826 if (inp == NULL) {
2827 in6p = in6_pcblookup_listen(tb, &daddr->v6, dport, 0);
2828 if (inp == NULL)
2829 return (-1);
2830 }
2831 break;
2832 #endif /* INET6 */
2833
2834 default:
2835 return (-1);
2836 }
2837
2838 #ifdef __NetBSD__
2839 switch (pd->af) {
2840 #ifdef INET
2841 case AF_INET:
2842 so = inp->inp_socket;
2843 break;
2844 #endif
2845 #ifdef INET6
2846 case AF_INET6:
2847 so = in6p->in6p_socket;
2848 break;
2849 #endif /* INET6 */
2850 }
2851 if (so == NULL || so->so_cred == NULL)
2852 return -1;
2853 pd->lookup.uid = kauth_cred_geteuid(so->so_cred);
2854 pd->lookup.gid = kauth_cred_getegid(so->so_cred);
2855 #else
2856 so = inp->inp_socket;
2857 pd->lookup.uid = so->so_euid;
2858 pd->lookup.gid = so->so_egid;
2859 #endif /* !__NetBSD__ */
2860 pd->lookup.pid = so->so_cpid;
2861 return (1);
2862 }
2863
2864 u_int8_t
2865 pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2866 {
2867 int hlen;
2868 u_int8_t hdr[60];
2869 u_int8_t *opt, optlen;
2870 u_int8_t wscale = 0;
2871
2872 hlen = th_off << 2; /* hlen <= sizeof(hdr) */
2873 if (hlen <= sizeof(struct tcphdr))
2874 return (0);
2875 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2876 return (0);
2877 opt = hdr + sizeof(struct tcphdr);
2878 hlen -= sizeof(struct tcphdr);
2879 while (hlen >= 3) {
2880 switch (*opt) {
2881 case TCPOPT_EOL:
2882 case TCPOPT_NOP:
2883 ++opt;
2884 --hlen;
2885 break;
2886 case TCPOPT_WINDOW:
2887 wscale = opt[2];
2888 if (wscale > TCP_MAX_WINSHIFT)
2889 wscale = TCP_MAX_WINSHIFT;
2890 wscale |= PF_WSCALE_FLAG;
2891 /* FALLTHROUGH */
2892 default:
2893 optlen = opt[1];
2894 if (optlen < 2)
2895 optlen = 2;
2896 hlen -= optlen;
2897 opt += optlen;
2898 break;
2899 }
2900 }
2901 return (wscale);
2902 }
2903
2904 u_int16_t
2905 pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2906 {
2907 int hlen;
2908 u_int8_t hdr[60];
2909 u_int8_t *opt, optlen;
2910 u_int16_t mss = tcp_mssdflt;
2911
2912 hlen = th_off << 2; /* hlen <= sizeof(hdr) */
2913 if (hlen <= sizeof(struct tcphdr))
2914 return (0);
2915 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2916 return (0);
2917 opt = hdr + sizeof(struct tcphdr);
2918 hlen -= sizeof(struct tcphdr);
2919 while (hlen >= TCPOLEN_MAXSEG) {
2920 switch (*opt) {
2921 case TCPOPT_EOL:
2922 case TCPOPT_NOP:
2923 ++opt;
2924 --hlen;
2925 break;
2926 case TCPOPT_MAXSEG:
2927 bcopy((void *)(opt + 2), (void *)&mss, 2);
2928 NTOHS(mss);
2929 /* FALLTHROUGH */
2930 default:
2931 optlen = opt[1];
2932 if (optlen < 2)
2933 optlen = 2;
2934 hlen -= optlen;
2935 opt += optlen;
2936 break;
2937 }
2938 }
2939 return (mss);
2940 }
2941
2942 u_int16_t
2943 pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer)
2944 {
2945 union {
2946 struct sockaddr dst;
2947 struct sockaddr_in dst4;
2948 struct sockaddr_in6 dst6;
2949 } u;
2950 struct route ro;
2951 struct route *rop = &ro;
2952 struct rtentry *rt;
2953 int hlen;
2954 u_int16_t mss = tcp_mssdflt;
2955
2956 hlen = 0; /* XXXGCC -Wuninitialized m68k */
2957
2958 memset(&ro, 0, sizeof(ro));
2959 switch (af) {
2960 #ifdef INET
2961 case AF_INET:
2962 hlen = sizeof(struct ip);
2963 sockaddr_in_init(&u.dst4, &addr->v4, 0);
2964 rtcache_setdst(rop, &u.dst);
2965 break;
2966 #endif /* INET */
2967 #ifdef INET6
2968 case AF_INET6:
2969 hlen = sizeof(struct ip6_hdr);
2970 sockaddr_in6_init(&u.dst6, &addr->v6, 0, 0, 0);
2971 rtcache_setdst(rop, &u.dst);
2972 break;
2973 #endif /* INET6 */
2974 }
2975
2976 #ifndef __NetBSD__
2977 rtalloc_noclone(rop, NO_CLONING);
2978 if ((rt = ro->ro_rt) != NULL) {
2979 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr);
2980 mss = uimax(tcp_mssdflt, mss);
2981 }
2982 #else
2983 if ((rt = rtcache_init_noclone(rop)) != NULL) {
2984 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr);
2985 mss = uimax(tcp_mssdflt, mss);
2986 rtcache_unref(rt, rop);
2987 }
2988 rtcache_free(rop);
2989 #endif
2990 mss = uimin(mss, offer);
2991 mss = uimax(mss, 64); /* sanity - at least max opt space */
2992 return (mss);
2993 }
2994
2995 void
2996 pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr)
2997 {
2998 struct pf_rule *r = s->rule.ptr;
2999
3000 s->rt_kif = NULL;
3001 if (!r->rt || r->rt == PF_FASTROUTE)
3002 return;
3003 switch (s->state_key->af) {
3004 #ifdef INET
3005 case AF_INET:
3006 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL,
3007 &s->nat_src_node);
3008 s->rt_kif = r->rpool.cur->kif;
3009 break;
3010 #endif /* INET */
3011 #ifdef INET6
3012 case AF_INET6:
3013 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL,
3014 &s->nat_src_node);
3015 s->rt_kif = r->rpool.cur->kif;
3016 break;
3017 #endif /* INET6 */
3018 }
3019 }
3020
3021 void
3022 pf_attach_state(struct pf_state_key *sk, struct pf_state *s, int tail)
3023 {
3024 s->state_key = sk;
3025 sk->refcnt++;
3026
3027 /* list is sorted, if-bound states before floating */
3028 if (tail)
3029 TAILQ_INSERT_TAIL(&sk->states, s, next);
3030 else
3031 TAILQ_INSERT_HEAD(&sk->states, s, next);
3032 }
3033
3034 void
3035 pf_detach_state(struct pf_state *s, int flags)
3036 {
3037 struct pf_state_key *sk = s->state_key;
3038
3039 if (sk == NULL)
3040 return;
3041
3042 s->state_key = NULL;
3043 TAILQ_REMOVE(&sk->states, s, next);
3044 if (--sk->refcnt == 0) {
3045 if (!(flags & PF_DT_SKIP_EXTGWY))
3046 RB_REMOVE(pf_state_tree_ext_gwy,
3047 &pf_statetbl_ext_gwy, sk);
3048 if (!(flags & PF_DT_SKIP_LANEXT))
3049 RB_REMOVE(pf_state_tree_lan_ext,
3050 &pf_statetbl_lan_ext, sk);
3051 pool_put(&pf_state_key_pl, sk);
3052 }
3053 }
3054
3055 struct pf_state_key *
3056 pf_alloc_state_key(struct pf_state *s)
3057 {
3058 struct pf_state_key *sk;
3059
3060 if ((sk = pool_get(&pf_state_key_pl, PR_NOWAIT)) == NULL)
3061 return (NULL);
3062 bzero(sk, sizeof(*sk));
3063 TAILQ_INIT(&sk->states);
3064 pf_attach_state(sk, s, 0);
3065
3066 return (sk);
3067 }
3068
3069 int
3070 pf_test_rule(struct pf_rule **rm, struct pf_state **sm, int direction,
3071 struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3072 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
3073 struct ifqueue *ifq)
3074 {
3075 struct pf_rule *nr = NULL;
3076 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
3077 u_int16_t bport, nport = 0;
3078 sa_family_t af = pd->af;
3079 struct pf_rule *r, *a = NULL;
3080 struct pf_ruleset *ruleset = NULL;
3081 struct pf_src_node *nsn = NULL;
3082 struct tcphdr *th = pd->hdr.tcp;
3083 u_short reason;
3084 int rewrite = 0, hdrlen = 0;
3085 int tag = -1, rtableid = -1;
3086 int asd = 0;
3087 int match = 0;
3088 int state_icmp = 0;
3089 u_int16_t mss = tcp_mssdflt;
3090 u_int16_t sport, dport;
3091 u_int8_t icmptype = 0, icmpcode = 0;
3092
3093 if (direction == PF_IN && pf_check_congestion(ifq)) {
3094 REASON_SET_NOPTR(&reason, PFRES_CONGEST);
3095 return (PF_DROP);
3096 }
3097
3098 sport = dport = hdrlen = 0;
3099
3100 switch (pd->proto) {
3101 case IPPROTO_TCP:
3102 sport = th->th_sport;
3103 dport = th->th_dport;
3104 hdrlen = sizeof(*th);
3105 break;
3106 case IPPROTO_UDP:
3107 sport = pd->hdr.udp->uh_sport;
3108 dport = pd->hdr.udp->uh_dport;
3109 hdrlen = sizeof(*pd->hdr.udp);
3110 break;
3111 #ifdef INET
3112 case IPPROTO_ICMP:
3113 if (pd->af != AF_INET)
3114 break;
3115 sport = dport = pd->hdr.icmp->icmp_id;
3116 icmptype = pd->hdr.icmp->icmp_type;
3117 icmpcode = pd->hdr.icmp->icmp_code;
3118
3119 if (icmptype == ICMP_UNREACH ||
3120 icmptype == ICMP_SOURCEQUENCH ||
3121 icmptype == ICMP_REDIRECT ||
3122 icmptype == ICMP_TIMXCEED ||
3123 icmptype == ICMP_PARAMPROB)
3124 state_icmp++;
3125 break;
3126 #endif /* INET */
3127
3128 #ifdef INET6
3129 case IPPROTO_ICMPV6:
3130 if (pd->af != AF_INET6)
3131 break;
3132 sport = dport = pd->hdr.icmp6->icmp6_id;
3133 hdrlen = sizeof(*pd->hdr.icmp6);
3134 icmptype = pd->hdr.icmp6->icmp6_type;
3135 icmpcode = pd->hdr.icmp6->icmp6_code;
3136
3137 if (icmptype == ICMP6_DST_UNREACH ||
3138 icmptype == ICMP6_PACKET_TOO_BIG ||
3139 icmptype == ICMP6_TIME_EXCEEDED ||
3140 icmptype == ICMP6_PARAM_PROB)
3141 state_icmp++;
3142 break;
3143 #endif /* INET6 */
3144 }
3145
3146 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3147
3148 if (direction == PF_OUT) {
3149 bport = nport = sport;
3150 /* check outgoing packet for BINAT/NAT */
3151 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3152 saddr, sport, daddr, dport, &pd->naddr, &nport)) != NULL) {
3153 PF_ACPY(&pd->baddr, saddr, af);
3154 switch (pd->proto) {
3155 case IPPROTO_TCP:
3156 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
3157 &th->th_sum, &pd->naddr, nport, 0, af);
3158 sport = th->th_sport;
3159 rewrite++;
3160 break;
3161 case IPPROTO_UDP:
3162 pf_change_ap(saddr, &pd->hdr.udp->uh_sport,
3163 pd->ip_sum, &pd->hdr.udp->uh_sum,
3164 &pd->naddr, nport, 1, af);
3165 sport = pd->hdr.udp->uh_sport;
3166 rewrite++;
3167 break;
3168 #ifdef INET
3169 case IPPROTO_ICMP:
3170 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3171 pd->naddr.v4.s_addr, 0);
3172 pd->hdr.icmp->icmp_cksum = pf_cksum_fixup(
3173 pd->hdr.icmp->icmp_cksum, sport, nport, 0);
3174 pd->hdr.icmp->icmp_id = nport;
3175 m_copyback(m, off, ICMP_MINLEN, pd->hdr.icmp);
3176 break;
3177 #endif /* INET */
3178 #ifdef INET6
3179 case IPPROTO_ICMPV6:
3180 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum,
3181 &pd->naddr, 0);
3182 rewrite++;
3183 break;
3184 #endif /* INET */
3185 default:
3186 switch (af) {
3187 #ifdef INET
3188 case AF_INET:
3189 pf_change_a(&saddr->v4.s_addr,
3190 pd->ip_sum, pd->naddr.v4.s_addr, 0);
3191 break;
3192 #endif /* INET */
3193 #ifdef INET6
3194 case AF_INET6:
3195 PF_ACPY(saddr, &pd->naddr, af);
3196 break;
3197 #endif /* INET */
3198 }
3199 break;
3200 }
3201
3202 if (nr->natpass)
3203 r = NULL;
3204 pd->nat_rule = nr;
3205 }
3206 } else {
3207 bport = nport = dport;
3208 /* check incoming packet for BINAT/RDR */
3209 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3210 saddr, sport, daddr, dport, &pd->naddr, &nport)) != NULL) {
3211 PF_ACPY(&pd->baddr, daddr, af);
3212 switch (pd->proto) {
3213 case IPPROTO_TCP:
3214 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
3215 &th->th_sum, &pd->naddr, nport, 0, af);
3216 dport = th->th_dport;
3217 rewrite++;
3218 break;
3219 case IPPROTO_UDP:
3220 pf_change_ap(daddr, &pd->hdr.udp->uh_dport,
3221 pd->ip_sum, &pd->hdr.udp->uh_sum,
3222 &pd->naddr, nport, 1, af);
3223 dport = pd->hdr.udp->uh_dport;
3224 rewrite++;
3225 break;
3226 #ifdef INET
3227 case IPPROTO_ICMP:
3228 pf_change_a(&daddr->v4.s_addr, pd->ip_sum,
3229 pd->naddr.v4.s_addr, 0);
3230 break;
3231 #endif /* INET */
3232 #ifdef INET6
3233 case IPPROTO_ICMPV6:
3234 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum,
3235 &pd->naddr, 0);
3236 rewrite++;
3237 break;
3238 #endif /* INET6 */
3239 default:
3240 switch (af) {
3241 #ifdef INET
3242 case AF_INET:
3243 pf_change_a(&daddr->v4.s_addr,
3244 pd->ip_sum, pd->naddr.v4.s_addr, 0);
3245 break;
3246 #endif /* INET */
3247 #ifdef INET6
3248 case AF_INET6:
3249 PF_ACPY(daddr, &pd->naddr, af);
3250 break;
3251 #endif /* INET */
3252 }
3253 break;
3254 }
3255
3256 if (nr->natpass)
3257 r = NULL;
3258 pd->nat_rule = nr;
3259 }
3260 }
3261
3262 while (r != NULL) {
3263 r->evaluations++;
3264 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3265 r = r->skip[PF_SKIP_IFP].ptr;
3266 else if (r->direction && r->direction != direction)
3267 r = r->skip[PF_SKIP_DIR].ptr;
3268 else if (r->af && r->af != af)
3269 r = r->skip[PF_SKIP_AF].ptr;
3270 else if (r->proto && r->proto != pd->proto)
3271 r = r->skip[PF_SKIP_PROTO].ptr;
3272 else if (PF_MISMATCHAW(&r->src.addr, saddr, af,
3273 r->src.neg, kif))
3274 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3275 /* tcp/udp only. port_op always 0 in other cases */
3276 else if (r->src.port_op && !pf_match_port(r->src.port_op,
3277 r->src.port[0], r->src.port[1], sport))
3278 r = r->skip[PF_SKIP_SRC_PORT].ptr;
3279 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af,
3280 r->dst.neg, NULL))
3281 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3282 /* tcp/udp only. port_op always 0 in other cases */
3283 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
3284 r->dst.port[0], r->dst.port[1], dport))
3285 r = r->skip[PF_SKIP_DST_PORT].ptr;
3286 /* icmp only. type always 0 in other cases */
3287 else if (r->type && r->type != icmptype + 1)
3288 r = TAILQ_NEXT(r, entries);
3289 /* icmp only. type always 0 in other cases */
3290 else if (r->code && r->code != icmpcode + 1)
3291 r = TAILQ_NEXT(r, entries);
3292 else if (r->tos && !(r->tos == pd->tos))
3293 r = TAILQ_NEXT(r, entries);
3294 else if (r->rule_flag & PFRULE_FRAGMENT)
3295 r = TAILQ_NEXT(r, entries);
3296 else if (pd->proto == IPPROTO_TCP &&
3297 (r->flagset & th->th_flags) != r->flags)
3298 r = TAILQ_NEXT(r, entries);
3299 /* tcp/udp only. uid.op always 0 in other cases */
3300 else if (r->uid.op && (pd->lookup.done || (pd->lookup.done =
3301 pf_socket_lookup(direction, pd), 1)) &&
3302 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
3303 pd->lookup.uid))
3304 r = TAILQ_NEXT(r, entries);
3305 /* tcp/udp only. gid.op always 0 in other cases */
3306 else if (r->gid.op && (pd->lookup.done || (pd->lookup.done =
3307 pf_socket_lookup(direction, pd), 1)) &&
3308 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
3309 pd->lookup.gid))
3310 r = TAILQ_NEXT(r, entries);
3311 else if (r->prob && r->prob <= cprng_fast32())
3312 r = TAILQ_NEXT(r, entries);
3313 else if (r->match_tag && !pf_match_tag(m, r, &tag))
3314 r = TAILQ_NEXT(r, entries);
3315 else if (r->os_fingerprint != PF_OSFP_ANY &&
3316 (pd->proto != IPPROTO_TCP || !pf_osfp_match(
3317 pf_osfp_fingerprint(pd, m, off, th),
3318 r->os_fingerprint)))
3319 r = TAILQ_NEXT(r, entries);
3320 else {
3321 if (r->tag)
3322 tag = r->tag;
3323 if (r->rtableid >= 0)
3324 rtableid = r->rtableid;
3325 if (r->anchor == NULL) {
3326 match = 1;
3327 *rm = r;
3328 *am = a;
3329 *rsm = ruleset;
3330 if ((*rm)->quick)
3331 break;
3332 r = TAILQ_NEXT(r, entries);
3333 } else
3334 pf_step_into_anchor(&asd, &ruleset,
3335 PF_RULESET_FILTER, &r, &a, &match);
3336 }
3337 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3338 PF_RULESET_FILTER, &r, &a, &match))
3339 break;
3340 }
3341 r = *rm;
3342 a = *am;
3343 ruleset = *rsm;
3344
3345 REASON_SET_NOPTR(&reason, PFRES_MATCH);
3346
3347 if (r->log || (nr != NULL && nr->log)) {
3348 if (rewrite)
3349 m_copyback(m, off, hdrlen, pd->hdr.any);
3350 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr,
3351 a, ruleset, pd);
3352 }
3353
3354 if (r->keep_state && pf_state_lock) {
3355 REASON_SET_NOPTR(&reason, PFRES_STATELOCKED);
3356 return PF_DROP;
3357 }
3358
3359 if ((r->action == PF_DROP) &&
3360 ((r->rule_flag & PFRULE_RETURNRST) ||
3361 (r->rule_flag & PFRULE_RETURNICMP) ||
3362 (r->rule_flag & PFRULE_RETURN))) {
3363 /* undo NAT changes, if they have taken place */
3364 if (nr != NULL) {
3365 if (direction == PF_OUT) {
3366 switch (pd->proto) {
3367 case IPPROTO_TCP:
3368 pf_change_ap(saddr, &th->th_sport,
3369 pd->ip_sum, &th->th_sum,
3370 &pd->baddr, bport, 0, af);
3371 sport = th->th_sport;
3372 rewrite++;
3373 break;
3374 case IPPROTO_UDP:
3375 pf_change_ap(saddr,
3376 &pd->hdr.udp->uh_sport, pd->ip_sum,
3377 &pd->hdr.udp->uh_sum, &pd->baddr,
3378 bport, 1, af);
3379 sport = pd->hdr.udp->uh_sport;
3380 rewrite++;
3381 break;
3382 case IPPROTO_ICMP:
3383 #ifdef INET6
3384 case IPPROTO_ICMPV6:
3385 #endif
3386 /* nothing! */
3387 break;
3388 default:
3389 switch (af) {
3390 case AF_INET:
3391 pf_change_a(&saddr->v4.s_addr,
3392 pd->ip_sum,
3393 pd->baddr.v4.s_addr, 0);
3394 break;
3395 case AF_INET6:
3396 PF_ACPY(saddr, &pd->baddr, af);
3397 break;
3398 }
3399 }
3400 } else {
3401 switch (pd->proto) {
3402 case IPPROTO_TCP:
3403 pf_change_ap(daddr, &th->th_dport,
3404 pd->ip_sum, &th->th_sum,
3405 &pd->baddr, bport, 0, af);
3406 dport = th->th_dport;
3407 rewrite++;
3408 break;
3409 case IPPROTO_UDP:
3410 pf_change_ap(daddr,
3411 &pd->hdr.udp->uh_dport, pd->ip_sum,
3412 &pd->hdr.udp->uh_sum, &pd->baddr,
3413 bport, 1, af);
3414 dport = pd->hdr.udp->uh_dport;
3415 rewrite++;
3416 break;
3417 case IPPROTO_ICMP:
3418 #ifdef INET6
3419 case IPPROTO_ICMPV6:
3420 #endif
3421 /* nothing! */
3422 break;
3423 default:
3424 switch (af) {
3425 case AF_INET:
3426 pf_change_a(&daddr->v4.s_addr,
3427 pd->ip_sum,
3428 pd->baddr.v4.s_addr, 0);
3429 break;
3430 case AF_INET6:
3431 PF_ACPY(daddr, &pd->baddr, af);
3432 break;
3433 }
3434 }
3435 }
3436 }
3437 if (pd->proto == IPPROTO_TCP &&
3438 ((r->rule_flag & PFRULE_RETURNRST) ||
3439 (r->rule_flag & PFRULE_RETURN)) &&
3440 !(th->th_flags & TH_RST)) {
3441 u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
3442 struct ip *hip = mtod(m, struct ip *);
3443
3444 #ifdef __NetBSD__
3445 if (pf_check_proto_cksum(m, direction, off,
3446 ntohs(hip->ip_len) - off, IPPROTO_TCP, AF_INET))
3447 #else
3448 if (pf_check_proto_cksum(m, off,
3449 ntohs(hip->ip_len) - off, IPPROTO_TCP, AF_INET))
3450 #endif /* !__NetBSD__ */
3451 REASON_SET_NOPTR(&reason, PFRES_PROTCKSUM);
3452 else {
3453 if (th->th_flags & TH_SYN)
3454 ack++;
3455 if (th->th_flags & TH_FIN)
3456 ack++;
3457 pf_send_tcp(r, af, pd->dst,
3458 pd->src, th->th_dport, th->th_sport,
3459 ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0,
3460 r->return_ttl, 1, 0, pd->eh, kif->pfik_ifp);
3461 }
3462 } else if ((af == AF_INET) && r->return_icmp)
3463 pf_send_icmp(m, r->return_icmp >> 8,
3464 r->return_icmp & 255, af, r);
3465 else if ((af == AF_INET6) && r->return_icmp6)
3466 pf_send_icmp(m, r->return_icmp6 >> 8,
3467 r->return_icmp6 & 255, af, r);
3468 }
3469
3470 if (r->action == PF_DROP)
3471 return (PF_DROP);
3472
3473 if (pf_tag_packet(m, tag, rtableid)) {
3474 REASON_SET_NOPTR(&reason, PFRES_MEMORY);
3475 return (PF_DROP);
3476 }
3477
3478 if (!state_icmp && (r->keep_state || nr != NULL ||
3479 (pd->flags & PFDESC_TCP_NORM))) {
3480 /* create new state */
3481 u_int16_t len;
3482 struct pf_state *s = NULL;
3483 struct pf_state_key *sk = NULL;
3484 struct pf_src_node *sn = NULL;
3485
3486 /* check maximums */
3487 if (r->max_states && (r->states >= r->max_states)) {
3488 pf_status.lcounters[LCNT_STATES]++;
3489 REASON_SET_NOPTR(&reason, PFRES_MAXSTATES);
3490 goto cleanup;
3491 }
3492 /* src node for filter rule */
3493 if ((r->rule_flag & PFRULE_SRCTRACK ||
3494 r->rpool.opts & PF_POOL_STICKYADDR) &&
3495 pf_insert_src_node(&sn, r, saddr, af) != 0) {
3496 REASON_SET_NOPTR(&reason, PFRES_SRCLIMIT);
3497 goto cleanup;
3498 }
3499 /* src node for translation rule */
3500 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3501 ((direction == PF_OUT &&
3502 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3503 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) {
3504 REASON_SET_NOPTR(&reason, PFRES_SRCLIMIT);
3505 goto cleanup;
3506 }
3507 s = pool_get(&pf_state_pl, PR_NOWAIT);
3508 if (s == NULL) {
3509 REASON_SET_NOPTR(&reason, PFRES_MEMORY);
3510 cleanup:
3511 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3512 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3513 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3514 pf_status.src_nodes--;
3515 pool_put(&pf_src_tree_pl, sn);
3516 }
3517 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3518 nsn->expire == 0) {
3519 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3520 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3521 pf_status.src_nodes--;
3522 pool_put(&pf_src_tree_pl, nsn);
3523 }
3524 if (sk != NULL) {
3525 pool_put(&pf_state_key_pl, sk);
3526 }
3527 return (PF_DROP);
3528 }
3529 bzero(s, sizeof(*s));
3530 s->rule.ptr = r;
3531 s->nat_rule.ptr = nr;
3532 s->anchor.ptr = a;
3533 STATE_INC_COUNTERS(s);
3534 s->allow_opts = r->allow_opts;
3535 s->log = r->log & PF_LOG_ALL;
3536 if (nr != NULL)
3537 s->log |= nr->log & PF_LOG_ALL;
3538 switch (pd->proto) {
3539 case IPPROTO_TCP:
3540 len = pd->tot_len - off - (th->th_off << 2);
3541 s->src.seqlo = ntohl(th->th_seq);
3542 s->src.seqhi = s->src.seqlo + len + 1;
3543 if ((th->th_flags & (TH_SYN|TH_ACK)) ==
3544 TH_SYN && r->keep_state == PF_STATE_MODULATE) {
3545 /* Generate sequence number modulator */
3546 while ((s->src.seqdiff =
3547 tcp_rndiss_next() - s->src.seqlo) == 0)
3548 ;
3549 pf_change_a(&th->th_seq, &th->th_sum,
3550 htonl(s->src.seqlo + s->src.seqdiff), 0);
3551 rewrite = 1;
3552 } else
3553 s->src.seqdiff = 0;
3554 if (th->th_flags & TH_SYN) {
3555 s->src.seqhi++;
3556 s->src.wscale = pf_get_wscale(m, off,
3557 th->th_off, af);
3558 }
3559 s->src.max_win = MAX(ntohs(th->th_win), 1);
3560 if (s->src.wscale & PF_WSCALE_MASK) {
3561 /* Remove scale factor from initial window */
3562 int win = s->src.max_win;
3563 win += 1 << (s->src.wscale & PF_WSCALE_MASK);
3564 s->src.max_win = (win - 1) >>
3565 (s->src.wscale & PF_WSCALE_MASK);
3566 }
3567 if (th->th_flags & TH_FIN)
3568 s->src.seqhi++;
3569 s->dst.seqhi = 1;
3570 s->dst.max_win = 1;
3571 s->src.state = TCPS_SYN_SENT;
3572 s->dst.state = TCPS_CLOSED;
3573 s->timeout = PFTM_TCP_FIRST_PACKET;
3574 break;
3575 case IPPROTO_UDP:
3576 s->src.state = PFUDPS_SINGLE;
3577 s->dst.state = PFUDPS_NO_TRAFFIC;
3578 s->timeout = PFTM_UDP_FIRST_PACKET;
3579 break;
3580 case IPPROTO_ICMP:
3581 #ifdef INET6
3582 case IPPROTO_ICMPV6:
3583 #endif
3584 s->timeout = PFTM_ICMP_FIRST_PACKET;
3585 break;
3586 default:
3587 s->src.state = PFOTHERS_SINGLE;
3588 s->dst.state = PFOTHERS_NO_TRAFFIC;
3589 s->timeout = PFTM_OTHER_FIRST_PACKET;
3590 }
3591
3592 s->creation = time_second;
3593 s->expire = time_second;
3594
3595 if (sn != NULL) {
3596 s->src_node = sn;
3597 s->src_node->states++;
3598 }
3599 if (nsn != NULL) {
3600 PF_ACPY(&nsn->raddr, &pd->naddr, af);
3601 s->nat_src_node = nsn;
3602 s->nat_src_node->states++;
3603 }
3604 if (pd->proto == IPPROTO_TCP) {
3605 if ((pd->flags & PFDESC_TCP_NORM) &&
3606 pf_normalize_tcp_init(m, off, pd, th, &s->src,
3607 &s->dst)) {
3608 REASON_SET_NOPTR(&reason, PFRES_MEMORY);
3609 pf_src_tree_remove_state(s);
3610 STATE_DEC_COUNTERS(s);
3611 pool_put(&pf_state_pl, s);
3612 return (PF_DROP);
3613 }
3614 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
3615 pf_normalize_tcp_stateful(m, off, pd, &reason,
3616 th, s, &s->src, &s->dst, &rewrite)) {
3617 /* This really shouldn't happen!!! */
3618 DPFPRINTF(PF_DEBUG_URGENT,
3619 ("pf_normalize_tcp_stateful failed on "
3620 "first pkt"));
3621 pf_normalize_tcp_cleanup(s);
3622 pf_src_tree_remove_state(s);
3623 STATE_DEC_COUNTERS(s);
3624 pool_put(&pf_state_pl, s);
3625 return (PF_DROP);
3626 }
3627 }
3628
3629 if ((sk = pf_alloc_state_key(s)) == NULL) {
3630 REASON_SET_NOPTR(&reason, PFRES_MEMORY);
3631 goto cleanup;
3632 }
3633
3634 sk->proto = pd->proto;
3635 sk->direction = direction;
3636 sk->af = af;
3637 if (direction == PF_OUT) {
3638 PF_ACPY(&sk->gwy.addr, saddr, af);
3639 PF_ACPY(&sk->ext.addr, daddr, af);
3640 switch (pd->proto) {
3641 case IPPROTO_ICMP:
3642 #ifdef INET6
3643 case IPPROTO_ICMPV6:
3644 #endif
3645 sk->gwy.port = nport;
3646 sk->ext.port = 0;
3647 break;
3648 default:
3649 sk->gwy.port = sport;
3650 sk->ext.port = dport;
3651 }
3652 if (nr != NULL) {
3653 PF_ACPY(&sk->lan.addr, &pd->baddr, af);
3654 sk->lan.port = bport;
3655 } else {
3656 PF_ACPY(&sk->lan.addr, &sk->gwy.addr, af);
3657 sk->lan.port = sk->gwy.port;
3658 }
3659 } else {
3660 PF_ACPY(&sk->lan.addr, daddr, af);
3661 PF_ACPY(&sk->ext.addr, saddr, af);
3662 switch (pd->proto) {
3663 case IPPROTO_ICMP:
3664 #ifdef INET6
3665 case IPPROTO_ICMPV6:
3666 #endif
3667 sk->lan.port = nport;
3668 sk->ext.port = 0;
3669 break;
3670 default:
3671 sk->lan.port = dport;
3672 sk->ext.port = sport;
3673 }
3674 if (nr != NULL) {
3675 PF_ACPY(&sk->gwy.addr, &pd->baddr, af);
3676 sk->gwy.port = bport;
3677 } else {
3678 PF_ACPY(&sk->gwy.addr, &sk->lan.addr, af);
3679 sk->gwy.port = sk->lan.port;
3680 }
3681 }
3682
3683 pf_set_rt_ifp(s, saddr); /* needs s->state_key set */
3684
3685 if (pf_insert_state(bound_iface(r, nr, kif), s)) {
3686 if (pd->proto == IPPROTO_TCP)
3687 pf_normalize_tcp_cleanup(s);
3688 REASON_SET_NOPTR(&reason, PFRES_STATEINS);
3689 pf_src_tree_remove_state(s);
3690 STATE_DEC_COUNTERS(s);
3691 pool_put(&pf_state_pl, s);
3692 return (PF_DROP);
3693 } else
3694 *sm = s;
3695 if (tag > 0) {
3696 pf_tag_ref(tag);
3697 s->tag = tag;
3698 }
3699 if (pd->proto == IPPROTO_TCP &&
3700 (th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
3701 r->keep_state == PF_STATE_SYNPROXY) {
3702 s->src.state = PF_TCPS_PROXY_SRC;
3703 if (nr != NULL) {
3704 if (direction == PF_OUT) {
3705 pf_change_ap(saddr, &th->th_sport,
3706 pd->ip_sum, &th->th_sum, &pd->baddr,
3707 bport, 0, af);
3708 sport = th->th_sport;
3709 } else {
3710 pf_change_ap(daddr, &th->th_dport,
3711 pd->ip_sum, &th->th_sum, &pd->baddr,
3712 bport, 0, af);
3713 sport = th->th_dport;
3714 }
3715 }
3716 s->src.seqhi = htonl(cprng_fast32());
3717 /* Find mss option */
3718 mss = pf_get_mss(m, off, th->th_off, af);
3719 mss = pf_calc_mss(saddr, af, mss);
3720 mss = pf_calc_mss(daddr, af, mss);
3721 s->src.mss = mss;
3722 pf_send_tcp(r, af, daddr, saddr, th->th_dport,
3723 th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1,
3724 TH_SYN|TH_ACK, 0, s->src.mss, 0, 1, 0, NULL, NULL);
3725 REASON_SET_NOPTR(&reason, PFRES_SYNPROXY);
3726 return (PF_SYNPROXY_DROP);
3727 }
3728 }
3729
3730 /* copy back packet headers if we performed NAT operations */
3731 if (rewrite)
3732 m_copyback(m, off, hdrlen, pd->hdr.any);
3733
3734 return (PF_PASS);
3735 }
3736
3737 int
3738 pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif,
3739 struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am,
3740 struct pf_ruleset **rsm)
3741 {
3742 struct pf_rule *r, *a = NULL;
3743 struct pf_ruleset *ruleset = NULL;
3744 sa_family_t af = pd->af;
3745 u_short reason;
3746 int tag = -1;
3747 int asd = 0;
3748 int match = 0;
3749
3750 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3751 while (r != NULL) {
3752 r->evaluations++;
3753 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3754 r = r->skip[PF_SKIP_IFP].ptr;
3755 else if (r->direction && r->direction != direction)
3756 r = r->skip[PF_SKIP_DIR].ptr;
3757 else if (r->af && r->af != af)
3758 r = r->skip[PF_SKIP_AF].ptr;
3759 else if (r->proto && r->proto != pd->proto)
3760 r = r->skip[PF_SKIP_PROTO].ptr;
3761 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af,
3762 r->src.neg, kif))
3763 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3764 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af,
3765 r->dst.neg, NULL))
3766 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3767 else if (r->tos && !(r->tos == pd->tos))
3768 r = TAILQ_NEXT(r, entries);
3769 else if (r->src.port_op || r->dst.port_op ||
3770 r->flagset || r->type || r->code ||
3771 r->os_fingerprint != PF_OSFP_ANY)
3772 r = TAILQ_NEXT(r, entries);
3773 else if (r->prob && r->prob <= cprng_fast32())
3774 r = TAILQ_NEXT(r, entries);
3775 else if (r->match_tag && !pf_match_tag(m, r, &tag))
3776 r = TAILQ_NEXT(r, entries);
3777 else {
3778 if (r->anchor == NULL) {
3779 match = 1;
3780 *rm = r;
3781 *am = a;
3782 *rsm = ruleset;
3783 if ((*rm)->quick)
3784 break;
3785 r = TAILQ_NEXT(r, entries);
3786 } else
3787 pf_step_into_anchor(&asd, &ruleset,
3788 PF_RULESET_FILTER, &r, &a, &match);
3789 }
3790 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3791 PF_RULESET_FILTER, &r, &a, &match))
3792 break;
3793 }
3794 r = *rm;
3795 a = *am;
3796 ruleset = *rsm;
3797
3798 REASON_SET_NOPTR(&reason, PFRES_MATCH);
3799
3800 if (r->log)
3801 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset,
3802 pd);
3803
3804 if (r->action != PF_PASS)
3805 return (PF_DROP);
3806
3807 if (pf_tag_packet(m, tag, -1)) {
3808 REASON_SET_NOPTR(&reason, PFRES_MEMORY);
3809 return (PF_DROP);
3810 }
3811
3812 return (PF_PASS);
3813 }
3814
3815 int
3816 pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
3817 struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
3818 u_short *reason)
3819 {
3820 struct pf_state_key_cmp key;
3821 struct tcphdr *th = pd->hdr.tcp;
3822 u_int16_t win = ntohs(th->th_win);
3823 u_int32_t ack, end, seq, orig_seq;
3824 u_int8_t sws, dws;
3825 int ackskew;
3826 int copyback = 0;
3827 struct pf_state_peer *src, *dst;
3828
3829 key.af = pd->af;
3830 key.proto = IPPROTO_TCP;
3831 if (direction == PF_IN) {
3832 PF_ACPY(&key.ext.addr, pd->src, key.af);
3833 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
3834 key.ext.port = th->th_sport;
3835 key.gwy.port = th->th_dport;
3836 } else {
3837 PF_ACPY(&key.lan.addr, pd->src, key.af);
3838 PF_ACPY(&key.ext.addr, pd->dst, key.af);
3839 key.lan.port = th->th_sport;
3840 key.ext.port = th->th_dport;
3841 }
3842
3843 STATE_LOOKUP();
3844
3845 if (direction == (*state)->state_key->direction) {
3846 src = &(*state)->src;
3847 dst = &(*state)->dst;
3848 } else {
3849 src = &(*state)->dst;
3850 dst = &(*state)->src;
3851 }
3852
3853 if ((*state)->src.state == PF_TCPS_PROXY_SRC) {
3854 if (direction != (*state)->state_key->direction) {
3855 REASON_SET(reason, PFRES_SYNPROXY);
3856 return (PF_SYNPROXY_DROP);
3857 }
3858 if (th->th_flags & TH_SYN) {
3859 if (ntohl(th->th_seq) != (*state)->src.seqlo) {
3860 REASON_SET(reason, PFRES_SYNPROXY);
3861 return (PF_DROP);
3862 }
3863 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
3864 pd->src, th->th_dport, th->th_sport,
3865 (*state)->src.seqhi, ntohl(th->th_seq) + 1,
3866 TH_SYN|TH_ACK, 0, (*state)->src.mss, 0, 1,
3867 0, NULL, NULL);
3868 REASON_SET(reason, PFRES_SYNPROXY);
3869 return (PF_SYNPROXY_DROP);
3870 } else if (!(th->th_flags & TH_ACK) ||
3871 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
3872 (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
3873 REASON_SET(reason, PFRES_SYNPROXY);
3874 return (PF_DROP);
3875 } else if ((*state)->src_node != NULL &&
3876 pf_src_connlimit(state)) {
3877 REASON_SET(reason, PFRES_SRCLIMIT);
3878 return (PF_DROP);
3879 } else
3880 (*state)->src.state = PF_TCPS_PROXY_DST;
3881 }
3882 if ((*state)->src.state == PF_TCPS_PROXY_DST) {
3883 struct pf_state_host *psrc, *pdst;
3884
3885 if (direction == PF_OUT) {
3886 psrc = &(*state)->state_key->gwy;
3887 pdst = &(*state)->state_key->ext;
3888 } else {
3889 psrc = &(*state)->state_key->ext;
3890 pdst = &(*state)->state_key->lan;
3891 }
3892 if (direction == (*state)->state_key->direction) {
3893 if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) ||
3894 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
3895 (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
3896 REASON_SET(reason, PFRES_SYNPROXY);
3897 return (PF_DROP);
3898 }
3899 (*state)->src.max_win = MAX(ntohs(th->th_win), 1);
3900 if ((*state)->dst.seqhi == 1)
3901 (*state)->dst.seqhi = htonl(cprng_fast32());
3902 pf_send_tcp((*state)->rule.ptr, pd->af, &psrc->addr,
3903 &pdst->addr, psrc->port, pdst->port,
3904 (*state)->dst.seqhi, 0, TH_SYN, 0,
3905 (*state)->src.mss, 0, 0, (*state)->tag, NULL, NULL);
3906 REASON_SET(reason, PFRES_SYNPROXY);
3907 return (PF_SYNPROXY_DROP);
3908 } else if (((th->th_flags & (TH_SYN|TH_ACK)) !=
3909 (TH_SYN|TH_ACK)) ||
3910 (ntohl(th->th_ack) != (*state)->dst.seqhi + 1)) {
3911 REASON_SET(reason, PFRES_SYNPROXY);
3912 return (PF_DROP);
3913 } else {
3914 (*state)->dst.max_win = MAX(ntohs(th->th_win), 1);
3915 (*state)->dst.seqlo = ntohl(th->th_seq);
3916 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
3917 pd->src, th->th_dport, th->th_sport,
3918 ntohl(th->th_ack), ntohl(th->th_seq) + 1,
3919 TH_ACK, (*state)->src.max_win, 0, 0, 0,
3920 (*state)->tag, NULL, NULL);
3921 pf_send_tcp((*state)->rule.ptr, pd->af, &psrc->addr,
3922 &pdst->addr, psrc->port, pdst->port,
3923 (*state)->src.seqhi + 1, (*state)->src.seqlo + 1,
3924 TH_ACK, (*state)->dst.max_win, 0, 0, 1,
3925 0, NULL, NULL);
3926 (*state)->src.seqdiff = (*state)->dst.seqhi -
3927 (*state)->src.seqlo;
3928 (*state)->dst.seqdiff = (*state)->src.seqhi -
3929 (*state)->dst.seqlo;
3930 (*state)->src.seqhi = (*state)->src.seqlo +
3931 (*state)->dst.max_win;
3932 (*state)->dst.seqhi = (*state)->dst.seqlo +
3933 (*state)->src.max_win;
3934 (*state)->src.wscale = (*state)->dst.wscale = 0;
3935 (*state)->src.state = (*state)->dst.state =
3936 TCPS_ESTABLISHED;
3937 REASON_SET(reason, PFRES_SYNPROXY);
3938 return (PF_SYNPROXY_DROP);
3939 }
3940 }
3941
3942 if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) {
3943 sws = src->wscale & PF_WSCALE_MASK;
3944 dws = dst->wscale & PF_WSCALE_MASK;
3945 } else
3946 sws = dws = 0;
3947
3948 /*
3949 * Sequence tracking algorithm from Guido van Rooij's paper:
3950 * http://www.madison-gurkha.com/publications/tcp_filtering/
3951 * tcp_filtering.ps
3952 */
3953
3954 orig_seq = seq = ntohl(th->th_seq);
3955 if (src->seqlo == 0) {
3956 /* First packet from this end. Set its state */
3957
3958 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) &&
3959 src->scrub == NULL) {
3960 if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) {
3961 REASON_SET(reason, PFRES_MEMORY);
3962 return (PF_DROP);
3963 }
3964 }
3965
3966 /* Deferred generation of sequence number modulator */
3967 if (dst->seqdiff && !src->seqdiff) {
3968 while ((src->seqdiff = tcp_rndiss_next() - seq) == 0)
3969 ;
3970 ack = ntohl(th->th_ack) - dst->seqdiff;
3971 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3972 src->seqdiff), 0);
3973 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3974 copyback = 1;
3975 } else {
3976 ack = ntohl(th->th_ack);
3977 }
3978
3979 end = seq + pd->p_len;
3980 if (th->th_flags & TH_SYN) {
3981 end++;
3982 if (dst->wscale & PF_WSCALE_FLAG) {
3983 src->wscale = pf_get_wscale(m, off, th->th_off,
3984 pd->af);
3985 if (src->wscale & PF_WSCALE_FLAG) {
3986 /* Remove scale factor from initial
3987 * window */
3988 sws = src->wscale & PF_WSCALE_MASK;
3989 win = ((u_int32_t)win + (1 << sws) - 1)
3990 >> sws;
3991 dws = dst->wscale & PF_WSCALE_MASK;
3992 } else {
3993 /* fixup other window */
3994 dst->max_win <<= dst->wscale &
3995 PF_WSCALE_MASK;
3996 /* in case of a retrans SYN|ACK */
3997 dst->wscale = 0;
3998 }
3999 }
4000 }
4001 if (th->th_flags & TH_FIN)
4002 end++;
4003
4004 src->seqlo = seq;
4005 if (src->state < TCPS_SYN_SENT)
4006 src->state = TCPS_SYN_SENT;
4007
4008 /*
4009 * May need to slide the window (seqhi may have been set by
4010 * the crappy stack check or if we picked up the connection
4011 * after establishment)
4012 */
4013 if (src->seqhi == 1 ||
4014 SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi))
4015 src->seqhi = end + MAX(1, dst->max_win << dws);
4016 if (win > src->max_win)
4017 src->max_win = win;
4018
4019 } else {
4020 ack = ntohl(th->th_ack) - dst->seqdiff;
4021 if (src->seqdiff) {
4022 /* Modulate sequence numbers */
4023 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
4024 src->seqdiff), 0);
4025 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
4026 copyback = 1;
4027 }
4028 end = seq + pd->p_len;
4029 if (th->th_flags & TH_SYN)
4030 end++;
4031 if (th->th_flags & TH_FIN)
4032 end++;
4033 }
4034
4035 if ((th->th_flags & TH_ACK) == 0) {
4036 /* Let it pass through the ack skew check */
4037 ack = dst->seqlo;
4038 } else if ((ack == 0 &&
4039 (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) ||
4040 /* broken tcp stacks do not set ack */
4041 (dst->state < TCPS_SYN_SENT)) {
4042 /*
4043 * Many stacks (ours included) will set the ACK number in an
4044 * FIN|ACK if the SYN times out -- no sequence to ACK.
4045 */
4046 ack = dst->seqlo;
4047 }
4048
4049 if (seq == end) {
4050 /* Ease sequencing restrictions on no data packets */
4051 seq = src->seqlo;
4052 end = seq;
4053 }
4054
4055 ackskew = dst->seqlo - ack;
4056
4057
4058 /*
4059 * Need to demodulate the sequence numbers in any TCP SACK options
4060 * (Selective ACK). We could optionally validate the SACK values
4061 * against the current ACK window, either forwards or backwards, but
4062 * I'm not confident that SACK has been implemented properly
4063 * everywhere. It wouldn't surprise me if several stacks accidently
4064 * SACK too far backwards of previously ACKed data. There really aren't
4065 * any security implications of bad SACKing unless the target stack
4066 * doesn't validate the option length correctly. Someone trying to
4067 * spoof into a TCP connection won't bother blindly sending SACK
4068 * options anyway.
4069 */
4070 if (dst->seqdiff && (th->th_off << 2) > sizeof(struct tcphdr)) {
4071 if (pf_modulate_sack(m, off, pd, th, dst))
4072 copyback = 1;
4073 }
4074
4075
4076 #define MAXACKWINDOW (0xffff + 1500) /* 1500 is an arbitrary fudge factor */
4077 if (SEQ_GEQ(src->seqhi, end) &&
4078 /* Last octet inside other's window space */
4079 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) &&
4080 /* Retrans: not more than one window back */
4081 (ackskew >= -MAXACKWINDOW) &&
4082 /* Acking not more than one reassembled fragment backwards */
4083 (ackskew <= (MAXACKWINDOW << sws)) &&
4084 /* Acking not more than one window forward */
4085 ((th->th_flags & TH_RST) == 0 || orig_seq == src->seqlo ||
4086 (orig_seq == src->seqlo + 1) || (pd->flags & PFDESC_IP_REAS) == 0)) {
4087 /* Require an exact/+1 sequence match on resets when possible */
4088
4089 if (dst->scrub || src->scrub) {
4090 if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4091 *state, src, dst, ©back))
4092 return (PF_DROP);
4093 }
4094
4095 /* update max window */
4096 if (src->max_win < win)
4097 src->max_win = win;
4098 /* synchronize sequencing */
4099 if (SEQ_GT(end, src->seqlo))
4100 src->seqlo = end;
4101 /* slide the window of what the other end can send */
4102 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4103 dst->seqhi = ack + MAX((win << sws), 1);
4104
4105
4106 /* update states */
4107 if (th->th_flags & TH_SYN)
4108 if (src->state < TCPS_SYN_SENT)
4109 src->state = TCPS_SYN_SENT;
4110 if (th->th_flags & TH_FIN)
4111 if (src->state < TCPS_CLOSING)
4112 src->state = TCPS_CLOSING;
4113 if (th->th_flags & TH_ACK) {
4114 if (dst->state == TCPS_SYN_SENT) {
4115 dst->state = TCPS_ESTABLISHED;
4116 if (src->state == TCPS_ESTABLISHED &&
4117 (*state)->src_node != NULL &&
4118 pf_src_connlimit(state)) {
4119 REASON_SET(reason, PFRES_SRCLIMIT);
4120 return (PF_DROP);
4121 }
4122 } else if (dst->state == TCPS_CLOSING)
4123 dst->state = TCPS_FIN_WAIT_2;
4124 }
4125 if (th->th_flags & TH_RST)
4126 src->state = dst->state = TCPS_TIME_WAIT;
4127
4128 /* update expire time */
4129 (*state)->expire = time_second;
4130 if (src->state >= TCPS_FIN_WAIT_2 &&
4131 dst->state >= TCPS_FIN_WAIT_2)
4132 (*state)->timeout = PFTM_TCP_CLOSED;
4133 else if (src->state >= TCPS_CLOSING &&
4134 dst->state >= TCPS_CLOSING)
4135 (*state)->timeout = PFTM_TCP_FIN_WAIT;
4136 else if (src->state < TCPS_ESTABLISHED ||
4137 dst->state < TCPS_ESTABLISHED)
4138 (*state)->timeout = PFTM_TCP_OPENING;
4139 else if (src->state >= TCPS_CLOSING ||
4140 dst->state >= TCPS_CLOSING)
4141 (*state)->timeout = PFTM_TCP_CLOSING;
4142 else
4143 (*state)->timeout = PFTM_TCP_ESTABLISHED;
4144
4145 /* Fall through to PASS packet */
4146
4147 } else if ((dst->state < TCPS_SYN_SENT ||
4148 dst->state >= TCPS_FIN_WAIT_2 ||
4149 src->state >= TCPS_FIN_WAIT_2) &&
4150 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) &&
4151 /* Within a window forward of the originating packet */
4152 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) {
4153 /* Within a window backward of the originating packet */
4154
4155 /*
4156 * This currently handles three situations:
4157 * 1) Stupid stacks will shotgun SYNs before their peer
4158 * replies.
4159 * 2) When PF catches an already established stream (the
4160 * firewall rebooted, the state table was flushed, routes
4161 * changed...)
4162 * 3) Packets get funky immediately after the connection
4163 * closes (this should catch Solaris spurious ACK|FINs
4164 * that web servers like to spew after a close)
4165 *
4166 * This must be a little more careful than the above code
4167 * since packet floods will also be caught here. We don't
4168 * update the TTL here to mitigate the damage of a packet
4169 * flood and so the same code can handle awkward establishment
4170 * and a loosened connection close.
4171 * In the establishment case, a correct peer response will
4172 * validate the connection, go through the normal state code
4173 * and keep updating the state TTL.
4174 */
4175
4176 if (pf_status.debug >= PF_DEBUG_MISC) {
4177 printf("pf: loose state match: ");
4178 pf_print_state(*state);
4179 pf_print_flags(th->th_flags);
4180 printf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
4181 "pkts=%llu:%llu\n", seq, orig_seq, ack, pd->p_len,
4182 ackskew,
4183 (unsigned long long int)(*state)->packets[0],
4184 (unsigned long long int)(*state)->packets[1]);
4185 }
4186
4187 if (dst->scrub || src->scrub) {
4188 if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4189 *state, src, dst, ©back))
4190 return (PF_DROP);
4191 }
4192
4193 /* update max window */
4194 if (src->max_win < win)
4195 src->max_win = win;
4196 /* synchronize sequencing */
4197 if (SEQ_GT(end, src->seqlo))
4198 src->seqlo = end;
4199 /* slide the window of what the other end can send */
4200 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4201 dst->seqhi = ack + MAX((win << sws), 1);
4202
4203 /*
4204 * Cannot set dst->seqhi here since this could be a shotgunned
4205 * SYN and not an already established connection.
4206 */
4207
4208 if (th->th_flags & TH_FIN)
4209 if (src->state < TCPS_CLOSING)
4210 src->state = TCPS_CLOSING;
4211 if (th->th_flags & TH_RST)
4212 src->state = dst->state = TCPS_TIME_WAIT;
4213
4214 /* Fall through to PASS packet */
4215
4216 } else {
4217 if ((*state)->dst.state == TCPS_SYN_SENT &&
4218 (*state)->src.state == TCPS_SYN_SENT) {
4219 /* Send RST for state mismatches during handshake */
4220 if (!(th->th_flags & TH_RST))
4221 pf_send_tcp((*state)->rule.ptr, pd->af,
4222 pd->dst, pd->src, th->th_dport,
4223 th->th_sport, ntohl(th->th_ack), 0,
4224 TH_RST, 0, 0,
4225 (*state)->rule.ptr->return_ttl, 1, 0,
4226 pd->eh, kif->pfik_ifp);
4227 src->seqlo = 0;
4228 src->seqhi = 1;
4229 src->max_win = 1;
4230 } else if (pf_status.debug >= PF_DEBUG_MISC) {
4231 printf("pf: BAD state: ");
4232 pf_print_state(*state);
4233 pf_print_flags(th->th_flags);
4234 printf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
4235 "pkts=%llu:%llu dir=%s,%s\n",
4236 seq, orig_seq, ack, pd->p_len, ackskew,
4237 (unsigned long long int)(*state)->packets[0],
4238 (unsigned long long int)(*state)->packets[1],
4239 direction == PF_IN ? "in" : "out",
4240 direction == (*state)->state_key->direction ?
4241 "fwd" : "rev");
4242 printf("pf: State failure on: %c %c %c %c | %c %c\n",
4243 SEQ_GEQ(src->seqhi, end) ? ' ' : '1',
4244 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ?
4245 ' ': '2',
4246 (ackskew >= -MAXACKWINDOW) ? ' ' : '3',
4247 (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4',
4248 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5',
4249 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6');
4250 }
4251 REASON_SET(reason, PFRES_BADSTATE);
4252 return (PF_DROP);
4253 }
4254
4255 /* Any packets which have gotten here are to be passed */
4256
4257 /* translate source/destination address, if necessary */
4258 if (STATE_TRANSLATE((*state)->state_key)) {
4259 if (direction == PF_OUT)
4260 pf_change_ap(pd->src, &th->th_sport, pd->ip_sum,
4261 &th->th_sum, &(*state)->state_key->gwy.addr,
4262 (*state)->state_key->gwy.port, 0, pd->af);
4263 else
4264 pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum,
4265 &th->th_sum, &(*state)->state_key->lan.addr,
4266 (*state)->state_key->lan.port, 0, pd->af);
4267 m_copyback(m, off, sizeof(*th), th);
4268 } else if (copyback) {
4269 /* Copyback sequence modulation or stateful scrub changes */
4270 m_copyback(m, off, sizeof(*th), th);
4271 }
4272
4273 return (PF_PASS);
4274 }
4275
4276 int
4277 pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif,
4278 struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4279 {
4280 struct pf_state_peer *src, *dst;
4281 struct pf_state_key_cmp key;
4282 struct udphdr *uh = pd->hdr.udp;
4283
4284 key.af = pd->af;
4285 key.proto = IPPROTO_UDP;
4286 if (direction == PF_IN) {
4287 PF_ACPY(&key.ext.addr, pd->src, key.af);
4288 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4289 key.ext.port = uh->uh_sport;
4290 key.gwy.port = uh->uh_dport;
4291 } else {
4292 PF_ACPY(&key.lan.addr, pd->src, key.af);
4293 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4294 key.lan.port = uh->uh_sport;
4295 key.ext.port = uh->uh_dport;
4296 }
4297
4298 STATE_LOOKUP();
4299
4300 if (direction == (*state)->state_key->direction) {
4301 src = &(*state)->src;
4302 dst = &(*state)->dst;
4303 } else {
4304 src = &(*state)->dst;
4305 dst = &(*state)->src;
4306 }
4307
4308 /* update states */
4309 if (src->state < PFUDPS_SINGLE)
4310 src->state = PFUDPS_SINGLE;
4311 if (dst->state == PFUDPS_SINGLE)
4312 dst->state = PFUDPS_MULTIPLE;
4313
4314 /* update expire time */
4315 (*state)->expire = time_second;
4316 if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE)
4317 (*state)->timeout = PFTM_UDP_MULTIPLE;
4318 else
4319 (*state)->timeout = PFTM_UDP_SINGLE;
4320
4321 /* translate source/destination address, if necessary */
4322 if (STATE_TRANSLATE((*state)->state_key)) {
4323 if (direction == PF_OUT)
4324 pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum,
4325 &uh->uh_sum, &(*state)->state_key->gwy.addr,
4326 (*state)->state_key->gwy.port, 1, pd->af);
4327 else
4328 pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum,
4329 &uh->uh_sum, &(*state)->state_key->lan.addr,
4330 (*state)->state_key->lan.port, 1, pd->af);
4331 m_copyback(m, off, sizeof(*uh), uh);
4332 }
4333
4334 return (PF_PASS);
4335 }
4336
4337 int
4338 pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif,
4339 struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
4340 u_short *reason)
4341 {
4342 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
4343 u_int16_t icmpid = 0, *icmpsum;
4344 u_int8_t icmptype;
4345 int state_icmp = 0;
4346 struct pf_state_key_cmp key;
4347
4348 icmpsum = NULL; /* XXXGCC -Wuninitialized m68k */
4349 icmptype = 0; /* XXXGCC -Wuninitialized m68k */
4350
4351 switch (pd->proto) {
4352 #ifdef INET
4353 case IPPROTO_ICMP:
4354 icmptype = pd->hdr.icmp->icmp_type;
4355 icmpid = pd->hdr.icmp->icmp_id;
4356 icmpsum = &pd->hdr.icmp->icmp_cksum;
4357
4358 if (icmptype == ICMP_UNREACH ||
4359 icmptype == ICMP_SOURCEQUENCH ||
4360 icmptype == ICMP_REDIRECT ||
4361 icmptype == ICMP_TIMXCEED ||
4362 icmptype == ICMP_PARAMPROB)
4363 state_icmp++;
4364 break;
4365 #endif /* INET */
4366 #ifdef INET6
4367 case IPPROTO_ICMPV6:
4368 icmptype = pd->hdr.icmp6->icmp6_type;
4369 icmpid = pd->hdr.icmp6->icmp6_id;
4370 icmpsum = &pd->hdr.icmp6->icmp6_cksum;
4371
4372 if (icmptype == ICMP6_DST_UNREACH ||
4373 icmptype == ICMP6_PACKET_TOO_BIG ||
4374 icmptype == ICMP6_TIME_EXCEEDED ||
4375 icmptype == ICMP6_PARAM_PROB)
4376 state_icmp++;
4377 break;
4378 #endif /* INET6 */
4379 }
4380
4381 if (!state_icmp) {
4382
4383 /*
4384 * ICMP query/reply message not related to a TCP/UDP packet.
4385 * Search for an ICMP state.
4386 */
4387 key.af = pd->af;
4388 key.proto = pd->proto;
4389 if (direction == PF_IN) {
4390 PF_ACPY(&key.ext.addr, pd->src, key.af);
4391 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4392 key.ext.port = 0;
4393 key.gwy.port = icmpid;
4394 } else {
4395 PF_ACPY(&key.lan.addr, pd->src, key.af);
4396 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4397 key.lan.port = icmpid;
4398 key.ext.port = 0;
4399 }
4400
4401 STATE_LOOKUP();
4402
4403 (*state)->expire = time_second;
4404 (*state)->timeout = PFTM_ICMP_ERROR_REPLY;
4405
4406 /* translate source/destination address, if necessary */
4407 if (STATE_TRANSLATE((*state)->state_key)) {
4408 if (direction == PF_OUT) {
4409 switch (pd->af) {
4410 #ifdef INET
4411 case AF_INET:
4412 pf_change_a(&saddr->v4.s_addr,
4413 pd->ip_sum,
4414 (*state)->state_key->gwy.addr.v4.s_addr, 0);
4415 pd->hdr.icmp->icmp_cksum =
4416 pf_cksum_fixup(
4417 pd->hdr.icmp->icmp_cksum, icmpid,
4418 (*state)->state_key->gwy.port, 0);
4419 pd->hdr.icmp->icmp_id =
4420 (*state)->state_key->gwy.port;
4421 m_copyback(m, off, ICMP_MINLEN,
4422 pd->hdr.icmp);
4423 break;
4424 #endif /* INET */
4425 #ifdef INET6
4426 case AF_INET6:
4427 pf_change_a6(saddr,
4428 &pd->hdr.icmp6->icmp6_cksum,
4429 &(*state)->state_key->gwy.addr, 0);
4430 m_copyback(m, off,
4431 sizeof(struct icmp6_hdr),
4432 pd->hdr.icmp6);
4433 break;
4434 #endif /* INET6 */
4435 }
4436 } else {
4437 switch (pd->af) {
4438 #ifdef INET
4439 case AF_INET:
4440 pf_change_a(&daddr->v4.s_addr,
4441 pd->ip_sum,
4442 (*state)->state_key->lan.addr.v4.s_addr, 0);
4443 pd->hdr.icmp->icmp_cksum =
4444 pf_cksum_fixup(
4445 pd->hdr.icmp->icmp_cksum, icmpid,
4446 (*state)->state_key->lan.port, 0);
4447 pd->hdr.icmp->icmp_id =
4448 (*state)->state_key->lan.port;
4449 m_copyback(m, off, ICMP_MINLEN,
4450 pd->hdr.icmp);
4451 break;
4452 #endif /* INET */
4453 #ifdef INET6
4454 case AF_INET6:
4455 pf_change_a6(daddr,
4456 &pd->hdr.icmp6->icmp6_cksum,
4457 &(*state)->state_key->lan.addr, 0);
4458 m_copyback(m, off,
4459 sizeof(struct icmp6_hdr),
4460 pd->hdr.icmp6);
4461 break;
4462 #endif /* INET6 */
4463 }
4464 }
4465 }
4466
4467 return (PF_PASS);
4468
4469 } else {
4470 /*
4471 * ICMP error message in response to a TCP/UDP packet.
4472 * Extract the inner TCP/UDP header and search for that state.
4473 */
4474
4475 struct pf_pdesc pd2;
4476 #ifdef INET
4477 struct ip h2;
4478 #endif /* INET */
4479 #ifdef INET6
4480 struct ip6_hdr h2_6;
4481 int terminal = 0;
4482 #endif /* INET6 */
4483 int ipoff2 = 0;
4484 int off2 = 0;
4485
4486 memset(&pd2, 0, sizeof pd2); /* XXX gcc */
4487
4488 pd2.af = pd->af;
4489 switch (pd->af) {
4490 #ifdef INET
4491 case AF_INET:
4492 /* offset of h2 in mbuf chain */
4493 ipoff2 = off + ICMP_MINLEN;
4494
4495 if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2),
4496 NULL, reason, pd2.af)) {
4497 DPFPRINTF(PF_DEBUG_MISC,
4498 ("pf: ICMP error message too short "
4499 "(ip)\n"));
4500 return (PF_DROP);
4501 }
4502 /*
4503 * ICMP error messages don't refer to non-first
4504 * fragments
4505 */
4506 if (h2.ip_off & htons(IP_OFFMASK)) {
4507 REASON_SET(reason, PFRES_FRAG);
4508 return (PF_DROP);
4509 }
4510
4511 /* offset of protocol header that follows h2 */
4512 off2 = ipoff2 + (h2.ip_hl << 2);
4513
4514 pd2.proto = h2.ip_p;
4515 pd2.src = (struct pf_addr *)&h2.ip_src;
4516 pd2.dst = (struct pf_addr *)&h2.ip_dst;
4517 pd2.ip_sum = &h2.ip_sum;
4518 break;
4519 #endif /* INET */
4520 #ifdef INET6
4521 case AF_INET6:
4522 ipoff2 = off + sizeof(struct icmp6_hdr);
4523
4524 if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6),
4525 NULL, reason, pd2.af)) {
4526 DPFPRINTF(PF_DEBUG_MISC,
4527 ("pf: ICMP error message too short "
4528 "(ip6)\n"));
4529 return (PF_DROP);
4530 }
4531 pd2.proto = h2_6.ip6_nxt;
4532 pd2.src = (struct pf_addr *)&h2_6.ip6_src;
4533 pd2.dst = (struct pf_addr *)&h2_6.ip6_dst;
4534 pd2.ip_sum = NULL;
4535 off2 = ipoff2 + sizeof(h2_6);
4536 do {
4537 switch (pd2.proto) {
4538 case IPPROTO_FRAGMENT:
4539 /*
4540 * ICMPv6 error messages for
4541 * non-first fragments
4542 */
4543 REASON_SET(reason, PFRES_FRAG);
4544 return (PF_DROP);
4545 case IPPROTO_AH:
4546 case IPPROTO_HOPOPTS:
4547 case IPPROTO_ROUTING:
4548 case IPPROTO_DSTOPTS: {
4549 /* get next header and header length */
4550 struct ip6_ext opt6;
4551
4552 if (!pf_pull_hdr(m, off2, &opt6,
4553 sizeof(opt6), NULL, reason,
4554 pd2.af)) {
4555 DPFPRINTF(PF_DEBUG_MISC,
4556 ("pf: ICMPv6 short opt\n"));
4557 return (PF_DROP);
4558 }
4559 if (pd2.proto == IPPROTO_AH)
4560 off2 += (opt6.ip6e_len + 2) * 4;
4561 else
4562 off2 += (opt6.ip6e_len + 1) * 8;
4563 pd2.proto = opt6.ip6e_nxt;
4564 /* goto the next header */
4565 break;
4566 }
4567 default:
4568 terminal++;
4569 break;
4570 }
4571 } while (!terminal);
4572 break;
4573 #endif /* INET6 */
4574 }
4575
4576 switch (pd2.proto) {
4577 case IPPROTO_TCP: {
4578 struct tcphdr th;
4579 u_int32_t seq;
4580 struct pf_state_peer *src, *dst;
4581 u_int8_t dws;
4582 int copyback = 0;
4583
4584 /*
4585 * Only the first 8 bytes of the TCP header can be
4586 * expected. Don't access any TCP header fields after
4587 * th_seq, an ackskew test is not possible.
4588 */
4589 if (!pf_pull_hdr(m, off2, &th, 8, NULL, reason,
4590 pd2.af)) {
4591 DPFPRINTF(PF_DEBUG_MISC,
4592 ("pf: ICMP error message too short "
4593 "(tcp)\n"));
4594 return (PF_DROP);
4595 }
4596
4597 key.af = pd2.af;
4598 key.proto = IPPROTO_TCP;
4599 if (direction == PF_IN) {
4600 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4601 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4602 key.ext.port = th.th_dport;
4603 key.gwy.port = th.th_sport;
4604 } else {
4605 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4606 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4607 key.lan.port = th.th_dport;
4608 key.ext.port = th.th_sport;
4609 }
4610
4611 STATE_LOOKUP();
4612
4613 if (direction == (*state)->state_key->direction) {
4614 src = &(*state)->dst;
4615 dst = &(*state)->src;
4616 } else {
4617 src = &(*state)->src;
4618 dst = &(*state)->dst;
4619 }
4620
4621 if (src->wscale && dst->wscale)
4622 dws = dst->wscale & PF_WSCALE_MASK;
4623 else
4624 dws = 0;
4625
4626 /* Demodulate sequence number */
4627 seq = ntohl(th.th_seq) - src->seqdiff;
4628 if (src->seqdiff) {
4629 pf_change_a(&th.th_seq, icmpsum,
4630 htonl(seq), 0);
4631 copyback = 1;
4632 }
4633
4634 if (!SEQ_GEQ(src->seqhi, seq) ||
4635 !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws))) {
4636 if (pf_status.debug >= PF_DEBUG_MISC) {
4637 printf("pf: BAD ICMP %d:%d ",
4638 icmptype, pd->hdr.icmp->icmp_code);
4639 pf_print_host(pd->src, 0, pd->af);
4640 printf(" -> ");
4641 pf_print_host(pd->dst, 0, pd->af);
4642 printf(" state: ");
4643 pf_print_state(*state);
4644 printf(" seq=%u\n", seq);
4645 }
4646 REASON_SET(reason, PFRES_BADSTATE);
4647 return (PF_DROP);
4648 }
4649
4650 if (STATE_TRANSLATE((*state)->state_key)) {
4651 if (direction == PF_IN) {
4652 pf_change_icmp(pd2.src, &th.th_sport,
4653 daddr, &(*state)->state_key->lan.addr,
4654 (*state)->state_key->lan.port, NULL,
4655 pd2.ip_sum, icmpsum,
4656 pd->ip_sum, 0, pd2.af);
4657 } else {
4658 pf_change_icmp(pd2.dst, &th.th_dport,
4659 saddr, &(*state)->state_key->gwy.addr,
4660 (*state)->state_key->gwy.port, NULL,
4661 pd2.ip_sum, icmpsum,
4662 pd->ip_sum, 0, pd2.af);
4663 }
4664 copyback = 1;
4665 }
4666
4667 if (copyback) {
4668 switch (pd2.af) {
4669 #ifdef INET
4670 case AF_INET:
4671 m_copyback(m, off, ICMP_MINLEN,
4672 pd->hdr.icmp);
4673 m_copyback(m, ipoff2, sizeof(h2),
4674 &h2);
4675 break;
4676 #endif /* INET */
4677 #ifdef INET6
4678 case AF_INET6:
4679 m_copyback(m, off,
4680 sizeof(struct icmp6_hdr),
4681 pd->hdr.icmp6);
4682 m_copyback(m, ipoff2, sizeof(h2_6),
4683 &h2_6);
4684 break;
4685 #endif /* INET6 */
4686 }
4687 m_copyback(m, off2, 8, &th);
4688 }
4689
4690 return (PF_PASS);
4691 break;
4692 }
4693 case IPPROTO_UDP: {
4694 struct udphdr uh;
4695
4696 if (!pf_pull_hdr(m, off2, &uh, sizeof(uh),
4697 NULL, reason, pd2.af)) {
4698 DPFPRINTF(PF_DEBUG_MISC,
4699 ("pf: ICMP error message too short "
4700 "(udp)\n"));
4701 return (PF_DROP);
4702 }
4703
4704 key.af = pd2.af;
4705 key.proto = IPPROTO_UDP;
4706 if (direction == PF_IN) {
4707 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4708 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4709 key.ext.port = uh.uh_dport;
4710 key.gwy.port = uh.uh_sport;
4711 } else {
4712 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4713 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4714 key.lan.port = uh.uh_dport;
4715 key.ext.port = uh.uh_sport;
4716 }
4717
4718 STATE_LOOKUP();
4719
4720 if (STATE_TRANSLATE((*state)->state_key)) {
4721 if (direction == PF_IN) {
4722 pf_change_icmp(pd2.src, &uh.uh_sport,
4723 daddr,
4724 &(*state)->state_key->lan.addr,
4725 (*state)->state_key->lan.port,
4726 &uh.uh_sum,
4727 pd2.ip_sum, icmpsum,
4728 pd->ip_sum, 1, pd2.af);
4729 } else {
4730 pf_change_icmp(pd2.dst, &uh.uh_dport,
4731 saddr,
4732 &(*state)->state_key->gwy.addr,
4733 (*state)->state_key->gwy.port, &uh.uh_sum,
4734 pd2.ip_sum, icmpsum,
4735 pd->ip_sum, 1, pd2.af);
4736 }
4737 switch (pd2.af) {
4738 #ifdef INET
4739 case AF_INET:
4740 m_copyback(m, off, ICMP_MINLEN,
4741 pd->hdr.icmp);
4742 m_copyback(m, ipoff2, sizeof(h2), &h2);
4743 break;
4744 #endif /* INET */
4745 #ifdef INET6
4746 case AF_INET6:
4747 m_copyback(m, off,
4748 sizeof(struct icmp6_hdr),
4749 pd->hdr.icmp6);
4750 m_copyback(m, ipoff2, sizeof(h2_6),
4751 &h2_6);
4752 break;
4753 #endif /* INET6 */
4754 }
4755 m_copyback(m, off2, sizeof(uh), &uh);
4756 }
4757
4758 return (PF_PASS);
4759 break;
4760 }
4761 #ifdef INET
4762 case IPPROTO_ICMP: {
4763 struct icmp iih;
4764
4765 if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN,
4766 NULL, reason, pd2.af)) {
4767 DPFPRINTF(PF_DEBUG_MISC,
4768 ("pf: ICMP error message too short i"
4769 "(icmp)\n"));
4770 return (PF_DROP);
4771 }
4772
4773 key.af = pd2.af;
4774 key.proto = IPPROTO_ICMP;
4775 if (direction == PF_IN) {
4776 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4777 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4778 key.ext.port = 0;
4779 key.gwy.port = iih.icmp_id;
4780 } else {
4781 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4782 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4783 key.lan.port = iih.icmp_id;
4784 key.ext.port = 0;
4785 }
4786
4787 STATE_LOOKUP();
4788
4789 if (STATE_TRANSLATE((*state)->state_key)) {
4790 if (direction == PF_IN) {
4791 pf_change_icmp(pd2.src, &iih.icmp_id,
4792 daddr,
4793 &(*state)->state_key->lan.addr,
4794 (*state)->state_key->lan.port, NULL,
4795 pd2.ip_sum, icmpsum,
4796 pd->ip_sum, 0, AF_INET);
4797 } else {
4798 pf_change_icmp(pd2.dst, &iih.icmp_id,
4799 saddr,
4800 &(*state)->state_key->gwy.addr,
4801 (*state)->state_key->gwy.port, NULL,
4802 pd2.ip_sum, icmpsum,
4803 pd->ip_sum, 0, AF_INET);
4804 }
4805 m_copyback(m, off, ICMP_MINLEN, pd->hdr.icmp);
4806 m_copyback(m, ipoff2, sizeof(h2), &h2);
4807 m_copyback(m, off2, ICMP_MINLEN, &iih);
4808 }
4809
4810 return (PF_PASS);
4811 break;
4812 }
4813 #endif /* INET */
4814 #ifdef INET6
4815 case IPPROTO_ICMPV6: {
4816 struct icmp6_hdr iih;
4817
4818 if (!pf_pull_hdr(m, off2, &iih,
4819 sizeof(struct icmp6_hdr), NULL, reason, pd2.af)) {
4820 DPFPRINTF(PF_DEBUG_MISC,
4821 ("pf: ICMP error message too short "
4822 "(icmp6)\n"));
4823 return (PF_DROP);
4824 }
4825
4826 key.af = pd2.af;
4827 key.proto = IPPROTO_ICMPV6;
4828 if (direction == PF_IN) {
4829 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4830 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4831 key.ext.port = 0;
4832 key.gwy.port = iih.icmp6_id;
4833 } else {
4834 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4835 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4836 key.lan.port = iih.icmp6_id;
4837 key.ext.port = 0;
4838 }
4839
4840 STATE_LOOKUP();
4841
4842 if (STATE_TRANSLATE((*state)->state_key)) {
4843 if (direction == PF_IN) {
4844 pf_change_icmp(pd2.src, &iih.icmp6_id,
4845 daddr,
4846 &(*state)->state_key->lan.addr,
4847 (*state)->state_key->lan.port, NULL,
4848 pd2.ip_sum, icmpsum,
4849 pd->ip_sum, 0, AF_INET6);
4850 } else {
4851 pf_change_icmp(pd2.dst, &iih.icmp6_id,
4852 saddr, &(*state)->state_key->gwy.addr,
4853 (*state)->state_key->gwy.port, NULL,
4854 pd2.ip_sum, icmpsum,
4855 pd->ip_sum, 0, AF_INET6);
4856 }
4857 m_copyback(m, off, sizeof(struct icmp6_hdr),
4858 pd->hdr.icmp6);
4859 m_copyback(m, ipoff2, sizeof(h2_6), &h2_6);
4860 m_copyback(m, off2, sizeof(struct icmp6_hdr),
4861 &iih);
4862 }
4863
4864 return (PF_PASS);
4865 break;
4866 }
4867 #endif /* INET6 */
4868 default: {
4869 key.af = pd2.af;
4870 key.proto = pd2.proto;
4871 if (direction == PF_IN) {
4872 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4873 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4874 key.ext.port = 0;
4875 key.gwy.port = 0;
4876 } else {
4877 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4878 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4879 key.lan.port = 0;
4880 key.ext.port = 0;
4881 }
4882
4883 STATE_LOOKUP();
4884
4885 if (STATE_TRANSLATE((*state)->state_key)) {
4886 if (direction == PF_IN) {
4887 pf_change_icmp(pd2.src, NULL,
4888 daddr,
4889 &(*state)->state_key->lan.addr,
4890 0, NULL,
4891 pd2.ip_sum, icmpsum,
4892 pd->ip_sum, 0, pd2.af);
4893 } else {
4894 pf_change_icmp(pd2.dst, NULL,
4895 saddr,
4896 &(*state)->state_key->gwy.addr,
4897 0, NULL,
4898 pd2.ip_sum, icmpsum,
4899 pd->ip_sum, 0, pd2.af);
4900 }
4901 switch (pd2.af) {
4902 #ifdef INET
4903 case AF_INET:
4904 m_copyback(m, off, ICMP_MINLEN,
4905 pd->hdr.icmp);
4906 m_copyback(m, ipoff2, sizeof(h2), &h2);
4907 break;
4908 #endif /* INET */
4909 #ifdef INET6
4910 case AF_INET6:
4911 m_copyback(m, off,
4912 sizeof(struct icmp6_hdr),
4913 pd->hdr.icmp6);
4914 m_copyback(m, ipoff2, sizeof(h2_6),
4915 &h2_6);
4916 break;
4917 #endif /* INET6 */
4918 }
4919 }
4920
4921 return (PF_PASS);
4922 break;
4923 }
4924 }
4925 }
4926 }
4927
4928 int
4929 pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif,
4930 struct pf_pdesc *pd)
4931 {
4932 struct pf_state_peer *src, *dst;
4933 struct pf_state_key_cmp key;
4934
4935 key.af = pd->af;
4936 key.proto = pd->proto;
4937 if (direction == PF_IN) {
4938 PF_ACPY(&key.ext.addr, pd->src, key.af);
4939 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4940 key.ext.port = 0;
4941 key.gwy.port = 0;
4942 } else {
4943 PF_ACPY(&key.lan.addr, pd->src, key.af);
4944 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4945 key.lan.port = 0;
4946 key.ext.port = 0;
4947 }
4948
4949 STATE_LOOKUP();
4950
4951 if (direction == (*state)->state_key->direction) {
4952 src = &(*state)->src;
4953 dst = &(*state)->dst;
4954 } else {
4955 src = &(*state)->dst;
4956 dst = &(*state)->src;
4957 }
4958
4959 /* update states */
4960 if (src->state < PFOTHERS_SINGLE)
4961 src->state = PFOTHERS_SINGLE;
4962 if (dst->state == PFOTHERS_SINGLE)
4963 dst->state = PFOTHERS_MULTIPLE;
4964
4965 /* update expire time */
4966 (*state)->expire = time_second;
4967 if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE)
4968 (*state)->timeout = PFTM_OTHER_MULTIPLE;
4969 else
4970 (*state)->timeout = PFTM_OTHER_SINGLE;
4971
4972 /* translate source/destination address, if necessary */
4973 if (STATE_TRANSLATE((*state)->state_key)) {
4974 if (direction == PF_OUT)
4975 switch (pd->af) {
4976 #ifdef INET
4977 case AF_INET:
4978 pf_change_a(&pd->src->v4.s_addr,
4979 pd->ip_sum,
4980 (*state)->state_key->gwy.addr.v4.s_addr,
4981 0);
4982 break;
4983 #endif /* INET */
4984 #ifdef INET6
4985 case AF_INET6:
4986 PF_ACPY(pd->src,
4987 &(*state)->state_key->gwy.addr, pd->af);
4988 break;
4989 #endif /* INET6 */
4990 }
4991 else
4992 switch (pd->af) {
4993 #ifdef INET
4994 case AF_INET:
4995 pf_change_a(&pd->dst->v4.s_addr,
4996 pd->ip_sum,
4997 (*state)->state_key->lan.addr.v4.s_addr,
4998 0);
4999 break;
5000 #endif /* INET */
5001 #ifdef INET6
5002 case AF_INET6:
5003 PF_ACPY(pd->dst,
5004 &(*state)->state_key->lan.addr, pd->af);
5005 break;
5006 #endif /* INET6 */
5007 }
5008 }
5009
5010 return (PF_PASS);
5011 }
5012
5013 /*
5014 * ipoff and off are measured from the start of the mbuf chain.
5015 * h must be at "ipoff" on the mbuf chain.
5016 */
5017 void *
5018 pf_pull_hdr(struct mbuf *m, int off, void *p, int len,
5019 u_short *actionp, u_short *reasonp, sa_family_t af)
5020 {
5021 switch (af) {
5022 #ifdef INET
5023 case AF_INET: {
5024 struct ip *h = mtod(m, struct ip *);
5025 u_int16_t fragoff = (ntohs(h->ip_off) & IP_OFFMASK) << 3;
5026
5027 if (fragoff) {
5028 if (fragoff >= len)
5029 ACTION_SET(actionp, PF_PASS);
5030 else {
5031 ACTION_SET(actionp, PF_DROP);
5032 REASON_SET(reasonp, PFRES_FRAG);
5033 }
5034 return (NULL);
5035 }
5036 if (m->m_pkthdr.len < off + len ||
5037 ntohs(h->ip_len) < off + len) {
5038 ACTION_SET(actionp, PF_DROP);
5039 REASON_SET(reasonp, PFRES_SHORT);
5040 return (NULL);
5041 }
5042 break;
5043 }
5044 #endif /* INET */
5045 #ifdef INET6
5046 case AF_INET6: {
5047 struct ip6_hdr *h = mtod(m, struct ip6_hdr *);
5048
5049 if (m->m_pkthdr.len < off + len ||
5050 (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) <
5051 (unsigned)(off + len)) {
5052 ACTION_SET(actionp, PF_DROP);
5053 REASON_SET(reasonp, PFRES_SHORT);
5054 return (NULL);
5055 }
5056 break;
5057 }
5058 #endif /* INET6 */
5059 }
5060 m_copydata(m, off, len, p);
5061 return (p);
5062 }
5063
5064 int
5065 pf_routable(struct pf_addr *addr, sa_family_t af, struct pfi_kif *kif)
5066 {
5067 #ifdef __NetBSD__
5068 union {
5069 struct sockaddr dst;
5070 struct sockaddr_in dst4;
5071 struct sockaddr_in6 dst6;
5072 } u;
5073 struct route ro;
5074 int ret = 1;
5075 struct rtentry *rt;
5076
5077 bzero(&ro, sizeof(ro));
5078 switch (af) {
5079 case AF_INET:
5080 sockaddr_in_init(&u.dst4, &addr->v4, 0);
5081 break;
5082 #ifdef INET6
5083 case AF_INET6:
5084 sockaddr_in6_init(&u.dst6, &addr->v6, 0, 0, 0);
5085 break;
5086 #endif /* INET6 */
5087 default:
5088 return (0);
5089 }
5090 rtcache_setdst(&ro, &u.dst);
5091
5092 rt = rtcache_init(&ro);
5093 ret = rt != NULL ? 1 : 0;
5094 if (rt != NULL)
5095 rtcache_unref(rt, &ro);
5096 rtcache_free(&ro);
5097
5098 return (ret);
5099 #else /* !__NetBSD__ */
5100 struct sockaddr_in *dst;
5101 int ret = 1;
5102 int check_mpath;
5103 extern int ipmultipath;
5104 #ifdef INET6
5105 extern int ip6_multipath;
5106 struct sockaddr_in6 *dst6;
5107 struct route_in6 ro;
5108 #else
5109 struct route ro;
5110 #endif
5111 struct radix_node *rn;
5112 struct rtentry *rt;
5113 struct ifnet *ifp;
5114
5115 check_mpath = 0;
5116 bzero(&ro, sizeof(ro));
5117 switch (af) {
5118 case AF_INET:
5119 dst = satosin(&ro.ro_dst);
5120 dst->sin_family = AF_INET;
5121 dst->sin_len = sizeof(*dst);
5122 dst->sin_addr = addr->v4;
5123 if (ipmultipath)
5124 check_mpath = 1;
5125 break;
5126 #ifdef INET6
5127 case AF_INET6:
5128 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
5129 dst6->sin6_family = AF_INET6;
5130 dst6->sin6_len = sizeof(*dst6);
5131 dst6->sin6_addr = addr->v6;
5132 if (ip6_multipath)
5133 check_mpath = 1;
5134 break;
5135 #endif /* INET6 */
5136 default:
5137 return (0);
5138 }
5139
5140 /* Skip checks for ipsec interfaces */
5141 if (kif != NULL && kif->pfik_ifp->if_type == IFT_ENC)
5142 goto out;
5143
5144 rtalloc_noclone((struct route *)&ro, NO_CLONING);
5145
5146 if (ro.ro_rt != NULL) {
5147 /* No interface given, this is a no-route check */
5148 if (kif == NULL)
5149 goto out;
5150
5151 if (kif->pfik_ifp == NULL) {
5152 ret = 0;
5153 goto out;
5154 }
5155
5156 /* Perform uRPF check if passed input interface */
5157 ret = 0;
5158 rn = (struct radix_node *)ro.ro_rt;
5159 do {
5160 rt = (struct rtentry *)rn;
5161 if (rt->rt_ifp->if_type == IFT_CARP)
5162 ifp = rt->rt_ifp->if_carpdev;
5163 else
5164 ifp = rt->rt_ifp;
5165
5166 if (kif->pfik_ifp == ifp)
5167 ret = 1;
5168 rn = rn_mpath_next(rn);
5169 } while (check_mpath == 1 && rn != NULL && ret == 0);
5170 } else
5171 ret = 0;
5172 out:
5173 if (ro.ro_rt != NULL)
5174 RTFREE(ro.ro_rt);
5175 return (ret);
5176 #endif /* !__NetBSD__ */
5177 }
5178
5179 int
5180 pf_rtlabel_match(struct pf_addr *addr, sa_family_t af, struct pf_addr_wrap *aw)
5181 {
5182 #ifdef __NetBSD__
5183 /* NetBSD doesn't have route labels. */
5184
5185 return (0);
5186 #else
5187 struct sockaddr_in *dst;
5188 #ifdef INET6
5189 struct sockaddr_in6 *dst6;
5190 struct route_in6 ro;
5191 #else
5192 struct route ro;
5193 #endif
5194 int ret = 0;
5195
5196 bzero(&ro, sizeof(ro));
5197 switch (af) {
5198 case AF_INET:
5199 dst = satosin(&ro.ro_dst);
5200 dst->sin_family = AF_INET;
5201 dst->sin_len = sizeof(*dst);
5202 dst->sin_addr = addr->v4;
5203 break;
5204 #ifdef INET6
5205 case AF_INET6:
5206 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
5207 dst6->sin6_family = AF_INET6;
5208 dst6->sin6_len = sizeof(*dst6);
5209 dst6->sin6_addr = addr->v6;
5210 break;
5211 #endif /* INET6 */
5212 default:
5213 return (0);
5214 }
5215
5216 rtalloc_noclone((struct route *)&ro, NO_CLONING);
5217
5218 if (ro.ro_rt != NULL) {
5219 if (ro.ro_rt->rt_labelid == aw->v.rtlabel)
5220 ret = 1;
5221 RTFREE(ro.ro_rt);
5222 }
5223
5224 return (ret);
5225 #endif /* !__NetBSD__ */
5226 }
5227
5228 #ifdef INET
5229 void
5230 pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5231 struct pf_state *s, struct pf_pdesc *pd)
5232 {
5233 struct mbuf *m0, *m1;
5234 struct route iproute;
5235 struct route *ro = NULL;
5236 const struct sockaddr *dst;
5237 union {
5238 struct sockaddr dst;
5239 struct sockaddr_in dst4;
5240 } u;
5241 struct ip *ip;
5242 struct ifnet *ifp = NULL;
5243 struct pf_addr naddr;
5244 struct pf_src_node *sn = NULL;
5245 int error = 0;
5246 #ifdef __NetBSD__
5247 struct pf_mtag *pf_mtag;
5248 #endif /* __NetBSD__ */
5249
5250 if (m == NULL || *m == NULL || r == NULL ||
5251 (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5252 panic("pf_route: invalid parameters");
5253
5254 #ifdef __NetBSD__
5255 if ((pf_mtag = pf_get_mtag(*m)) == NULL) {
5256 m0 = *m;
5257 *m = NULL;
5258 goto bad;
5259 }
5260 if (pf_mtag->routed++ > 3) {
5261 m0 = *m;
5262 *m = NULL;
5263 goto bad;
5264 }
5265 #else
5266 if ((*m)->m_pkthdr.pf.routed++ > 3) {
5267 m0 = *m;
5268 *m = NULL;
5269 goto bad;
5270 }
5271 #endif /* !__NetBSD__ */
5272
5273 if (r->rt == PF_DUPTO) {
5274 if ((m0 = m_dup(*m, 0, M_COPYALL, M_NOWAIT)) == NULL)
5275 return;
5276 } else {
5277 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5278 return;
5279 m0 = *m;
5280 }
5281
5282 if (m0->m_len < sizeof(struct ip)) {
5283 DPFPRINTF(PF_DEBUG_URGENT,
5284 ("pf_route: m0->m_len < sizeof(struct ip)\n"));
5285 goto bad;
5286 }
5287
5288 ip = mtod(m0, struct ip *);
5289
5290 ro = &iproute;
5291 memset(ro, 0, sizeof(*ro));
5292 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0);
5293 dst = &u.dst;
5294 rtcache_setdst(ro, dst);
5295
5296 if (r->rt == PF_FASTROUTE) {
5297 struct rtentry *rt;
5298
5299 rt = rtcache_init(ro);
5300
5301 if (rt == NULL) {
5302 ip_statinc(IP_STAT_NOROUTE);
5303 goto bad;
5304 }
5305
5306 ifp = rt->rt_ifp;
5307 rt->rt_use++;
5308
5309 if (rt->rt_flags & RTF_GATEWAY)
5310 dst = rt->rt_gateway;
5311 rtcache_unref(rt, ro); /* FIXME dst is NOMPSAFE */
5312 } else {
5313 if (TAILQ_EMPTY(&r->rpool.list)) {
5314 DPFPRINTF(PF_DEBUG_URGENT,
5315 ("pf_route: TAILQ_EMPTY(&r->rpool.list)\n"));
5316 goto bad;
5317 }
5318 if (s == NULL) {
5319 pf_map_addr(AF_INET, r,
5320 (const struct pf_addr *)&ip->ip_src,
5321 &naddr, NULL, &sn);
5322 if (!PF_AZERO(&naddr, AF_INET))
5323 u.dst4.sin_addr.s_addr = naddr.v4.s_addr;
5324 ifp = r->rpool.cur->kif ?
5325 r->rpool.cur->kif->pfik_ifp : NULL;
5326 } else {
5327 if (!PF_AZERO(&s->rt_addr, AF_INET))
5328 u.dst4.sin_addr.s_addr = s->rt_addr.v4.s_addr;
5329 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5330 }
5331 }
5332 if (ifp == NULL)
5333 goto bad;
5334
5335 if (oifp != ifp) {
5336 if (pf_test(PF_OUT, ifp, &m0, NULL) != PF_PASS)
5337 goto bad;
5338 else if (m0 == NULL)
5339 goto done;
5340 if (m0->m_len < sizeof(struct ip)) {
5341 DPFPRINTF(PF_DEBUG_URGENT,
5342 ("pf_route: m0->m_len < sizeof(struct ip)\n"));
5343 goto bad;
5344 }
5345 ip = mtod(m0, struct ip *);
5346 }
5347
5348 /* Copied from ip_output. */
5349
5350 /* Catch routing changes wrt. hardware checksumming for TCP or UDP. */
5351 #ifdef __NetBSD__
5352 if (m0->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
5353 in_undefer_cksum_tcpudp(m0);
5354 m0->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
5355 }
5356 #else
5357 if (m0->m_pkthdr.csum_flags & M_TCPV4_CSUM_OUT) {
5358 if (!(ifp->if_capabilities & IFCAP_CSUM_TCPv4) ||
5359 ifp->if_bridge != NULL) {
5360 in_delayed_cksum(m0);
5361 m0->m_pkthdr.csum_flags &= ~M_TCPV4_CSUM_OUT; /* Clear */
5362 }
5363 } else if (m0->m_pkthdr.csum_flags & M_UDPV4_CSUM_OUT) {
5364 if (!(ifp->if_capabilities & IFCAP_CSUM_UDPv4) ||
5365 ifp->if_bridge != NULL) {
5366 in_delayed_cksum(m0);
5367 m0->m_pkthdr.csum_flags &= ~M_UDPV4_CSUM_OUT; /* Clear */
5368 }
5369 }
5370 #endif /* !__NetBSD__ */
5371
5372 if (ntohs(ip->ip_len) <= ifp->if_mtu) {
5373 #ifdef __NetBSD__
5374 ip->ip_sum = 0;
5375 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
5376
5377 m0->m_pkthdr.csum_flags &= ~M_CSUM_IPv4;
5378 #else
5379 if ((ifp->if_capabilities & IFCAP_CSUM_IPv4) &&
5380 ifp->if_bridge == NULL) {
5381 m0->m_pkthdr.csum_flags |= M_IPV4_CSUM_OUT;
5382 ipstat.ips_outhwcsum++;
5383 } else {
5384 ip->ip_sum = 0;
5385 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
5386 }
5387 /* Update relevant hardware checksum stats for TCP/UDP */
5388 if (m0->m_pkthdr.csum_flags & M_TCPV4_CSUM_OUT)
5389 tcpstat.tcps_outhwcsum++;
5390 else if (m0->m_pkthdr.csum_flags & M_UDPV4_CSUM_OUT)
5391 udpstat.udps_outhwcsum++;
5392 #endif /* !__NetBSD__ */
5393 error = if_output_lock(ifp, ifp, m0, dst, NULL);
5394 goto done;
5395 }
5396
5397 /*
5398 * Too large for interface; fragment if possible.
5399 * Must be able to put at least 8 bytes per fragment.
5400 */
5401 if (ip->ip_off & htons(IP_DF)) {
5402 ip_statinc(IP_STAT_CANTFRAG);
5403 if (r->rt != PF_DUPTO) {
5404 icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0,
5405 ifp->if_mtu);
5406 goto done;
5407 } else
5408 goto bad;
5409 }
5410
5411 #ifdef __NetBSD__
5412 /* Make ip_fragment re-compute checksums. */
5413 if (IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) {
5414 m0->m_pkthdr.csum_flags |= M_CSUM_IPv4;
5415 }
5416 #endif /* __NetBSD__ */
5417 m1 = m0;
5418 error = ip_fragment(m0, ifp, ifp->if_mtu);
5419 if (error) {
5420 m0 = NULL;
5421 goto bad;
5422 }
5423
5424 for (m0 = m1; m0; m0 = m1) {
5425 m1 = m0->m_nextpkt;
5426 m0->m_nextpkt = 0;
5427 if (error == 0)
5428 error = (*ifp->if_output)(ifp, m0, dst, NULL);
5429 else
5430 m_freem(m0);
5431 }
5432
5433 if (error == 0)
5434 ip_statinc(IP_STAT_FRAGMENTED);
5435
5436 done:
5437 if (r->rt != PF_DUPTO)
5438 *m = NULL;
5439 if (ro == &iproute)
5440 rtcache_free(ro);
5441 return;
5442
5443 bad:
5444 m_freem(m0);
5445 goto done;
5446 }
5447 #endif /* INET */
5448
5449 #ifdef INET6
5450 void
5451 pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5452 struct pf_state *s, struct pf_pdesc *pd)
5453 {
5454 struct mbuf *m0;
5455 struct sockaddr_in6 dst;
5456 struct ip6_hdr *ip6;
5457 struct ifnet *ifp = NULL;
5458 struct pf_addr naddr;
5459 struct pf_src_node *sn = NULL;
5460 #ifdef __NetBSD__
5461 struct pf_mtag *pf_mtag;
5462 #endif /* __NetBSD__ */
5463
5464 if (m == NULL || *m == NULL || r == NULL ||
5465 (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5466 panic("pf_route6: invalid parameters");
5467
5468 #ifdef __NetBSD__
5469 if ((pf_mtag = pf_get_mtag(*m)) == NULL) {
5470 m0 = *m;
5471 *m = NULL;
5472 goto bad;
5473 }
5474 if (pf_mtag->routed++ > 3) {
5475 m0 = *m;
5476 *m = NULL;
5477 goto bad;
5478 }
5479 #else
5480 if ((*m)->m_pkthdr.pf.routed++ > 3) {
5481 m0 = *m;
5482 *m = NULL;
5483 goto bad;
5484 }
5485 #endif /* !__NetBSD__ */
5486
5487 if (r->rt == PF_DUPTO) {
5488 if ((m0 = m_dup(*m, 0, M_COPYALL, M_NOWAIT)) == NULL)
5489 return;
5490 } else {
5491 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5492 return;
5493 m0 = *m;
5494 }
5495
5496 if (m0->m_len < sizeof(struct ip6_hdr)) {
5497 DPFPRINTF(PF_DEBUG_URGENT,
5498 ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n"));
5499 goto bad;
5500 }
5501 ip6 = mtod(m0, struct ip6_hdr *);
5502
5503 dst.sin6_family = AF_INET6;
5504 dst.sin6_len = sizeof(dst);
5505 dst.sin6_addr = ip6->ip6_dst;
5506
5507 /* Cheat. XXX why only in the v6 case??? */
5508 if (r->rt == PF_FASTROUTE) {
5509 #ifdef __NetBSD__
5510 pf_mtag->flags |= PF_TAG_GENERATED;
5511 #else
5512 m0->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
5513 #endif /* !__NetBSD__ */
5514 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
5515 return;
5516 }
5517
5518 if (TAILQ_EMPTY(&r->rpool.list)) {
5519 DPFPRINTF(PF_DEBUG_URGENT,
5520 ("pf_route6: TAILQ_EMPTY(&r->rpool.list)\n"));
5521 goto bad;
5522 }
5523 if (s == NULL) {
5524 pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src,
5525 &naddr, NULL, &sn);
5526 if (!PF_AZERO(&naddr, AF_INET6))
5527 PF_ACPY((struct pf_addr *)&dst.sin6_addr,
5528 &naddr, AF_INET6);
5529 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL;
5530 } else {
5531 if (!PF_AZERO(&s->rt_addr, AF_INET6))
5532 PF_ACPY((struct pf_addr *)&dst.sin6_addr,
5533 &s->rt_addr, AF_INET6);
5534 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5535 }
5536 if (ifp == NULL)
5537 goto bad;
5538
5539 if (oifp != ifp) {
5540 if (pf_test6(PF_OUT, ifp, &m0, NULL) != PF_PASS)
5541 goto bad;
5542 else if (m0 == NULL)
5543 goto done;
5544 if (m0->m_len < sizeof(struct ip6_hdr)) {
5545 DPFPRINTF(PF_DEBUG_URGENT,
5546 ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n"));
5547 goto bad;
5548 }
5549 ip6 = mtod(m0, struct ip6_hdr *);
5550 }
5551
5552 /*
5553 * If the packet is too large for the outgoing interface,
5554 * send back an icmp6 error.
5555 */
5556 if (IN6_IS_SCOPE_EMBEDDABLE(&dst.sin6_addr))
5557 dst.sin6_addr.s6_addr16[1] = htons(ifp->if_index);
5558 if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) {
5559 (void)ip6_if_output(ifp, ifp, m0, &dst, NULL);
5560 } else {
5561 in6_ifstat_inc(ifp, ifs6_in_toobig);
5562 if (r->rt != PF_DUPTO)
5563 icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
5564 else
5565 goto bad;
5566 }
5567
5568 done:
5569 if (r->rt != PF_DUPTO)
5570 *m = NULL;
5571 return;
5572
5573 bad:
5574 m_freem(m0);
5575 goto done;
5576 }
5577 #endif /* INET6 */
5578
5579
5580 /*
5581 * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag
5582 * off is the offset where the protocol header starts
5583 * len is the total length of protocol header plus payload
5584 * returns 0 when the checksum is valid, otherwise returns 1.
5585 */
5586 #ifdef __NetBSD__
5587 int
5588 pf_check_proto_cksum(struct mbuf *m, int direction, int off, int len,
5589 u_int8_t p, sa_family_t af)
5590 #else
5591 int
5592 pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p,
5593 sa_family_t af)
5594 #endif /* !__NetBSD__ */
5595 {
5596 #ifndef __NetBSD__
5597 u_int16_t flag_ok, flag_bad;
5598 #endif /* !__NetBSD__ */
5599 u_int16_t sum;
5600
5601 #ifndef __NetBSD__
5602 switch (p) {
5603 case IPPROTO_TCP:
5604 flag_ok = M_TCP_CSUM_IN_OK;
5605 flag_bad = M_TCP_CSUM_IN_BAD;
5606 break;
5607 case IPPROTO_UDP:
5608 flag_ok = M_UDP_CSUM_IN_OK;
5609 flag_bad = M_UDP_CSUM_IN_BAD;
5610 break;
5611 case IPPROTO_ICMP:
5612 #ifdef INET6
5613 case IPPROTO_ICMPV6:
5614 #endif /* INET6 */
5615 flag_ok = flag_bad = 0;
5616 break;
5617 default:
5618 return (1);
5619 }
5620 if (m->m_pkthdr.csum_flags & flag_ok)
5621 return (0);
5622 if (m->m_pkthdr.csum_flags & flag_bad)
5623 return (1);
5624 #endif /* !__NetBSD__ */
5625 if (off < sizeof(struct ip) || len < sizeof(struct udphdr))
5626 return (1);
5627 if (m->m_pkthdr.len < off + len)
5628 return (1);
5629 #ifdef __NetBSD__
5630 if (direction == PF_IN) {
5631 switch (p) {
5632 case IPPROTO_TCP: {
5633 struct tcphdr th; /* XXX */
5634 int thlen;
5635
5636 m_copydata(m, off, sizeof(th), &th); /* XXX */
5637 thlen = th.th_off << 2;
5638 return tcp_input_checksum(af, m, &th, off,
5639 thlen, len - thlen) != 0;
5640 }
5641
5642 case IPPROTO_UDP: {
5643 struct udphdr uh; /* XXX */
5644
5645 m_copydata(m, off, sizeof(uh), &uh); /* XXX */
5646 return udp_input_checksum(af, m, &uh, off, len) != 0;
5647 }
5648 }
5649 }
5650 #endif /* __NetBSD__ */
5651 switch (af) {
5652 #ifdef INET
5653 case AF_INET:
5654 if (p == IPPROTO_ICMP) {
5655 if (m->m_len < off)
5656 return (1);
5657 m->m_data += off;
5658 m->m_len -= off;
5659 sum = in_cksum(m, len);
5660 m->m_data -= off;
5661 m->m_len += off;
5662 } else {
5663 if (m->m_len < sizeof(struct ip))
5664 return (1);
5665 sum = in4_cksum(m, p, off, len);
5666 }
5667 break;
5668 #endif /* INET */
5669 #ifdef INET6
5670 case AF_INET6:
5671 if (m->m_len < sizeof(struct ip6_hdr))
5672 return (1);
5673 sum = in6_cksum(m, p, off, len);
5674 break;
5675 #endif /* INET6 */
5676 default:
5677 return (1);
5678 }
5679 if (sum) {
5680 #ifndef __NetBSD__
5681 m->m_pkthdr.csum_flags |= flag_bad;
5682 #endif /* !__NetBSD__ */
5683 switch (p) {
5684 case IPPROTO_TCP:
5685 tcp_statinc(TCP_STAT_RCVBADSUM);
5686 break;
5687 case IPPROTO_UDP:
5688 udp_statinc(UDP_STAT_BADSUM);
5689 break;
5690 case IPPROTO_ICMP:
5691 icmp_statinc(ICMP_STAT_CHECKSUM);
5692 break;
5693 #ifdef INET6
5694 case IPPROTO_ICMPV6:
5695 icmp6_statinc(ICMP6_STAT_CHECKSUM);
5696 break;
5697 #endif /* INET6 */
5698 }
5699 return (1);
5700 }
5701 #ifndef __NetBSD__
5702 m->m_pkthdr.csum_flags |= flag_ok;
5703 #endif /* !__NetBSD__ */
5704 return (0);
5705 }
5706
5707 #ifdef INET
5708 int
5709 pf_test(int dir, struct ifnet *ifp, struct mbuf **m0,
5710 struct ether_header *eh)
5711 {
5712 struct pfi_kif *kif;
5713 u_short action, reason = 0, log = 0;
5714 struct mbuf *m = *m0;
5715 struct ip *h = NULL;
5716 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr;
5717 struct pf_state *s = NULL;
5718 struct pf_state_key *sk = NULL;
5719 struct pf_ruleset *ruleset = NULL;
5720 struct pf_pdesc pd;
5721 int off, dirndx;
5722 #ifdef __NetBSD__
5723 struct pf_mtag *pf_mtag = NULL; /* XXX gcc */
5724 #if defined(ALTQ)
5725 int pqid = 0;
5726 #endif
5727 #endif /* __NetBSD__ */
5728
5729 if (!pf_status.running)
5730 return (PF_PASS);
5731
5732 memset(&pd, 0, sizeof(pd));
5733 if (ifp->if_type == IFT_CARP && ifp->if_carpdev)
5734 kif = (struct pfi_kif *)ifp->if_carpdev->if_pf_kif;
5735 else
5736 kif = (struct pfi_kif *)ifp->if_pf_kif;
5737
5738 if (kif == NULL) {
5739 DPFPRINTF(PF_DEBUG_URGENT,
5740 ("pf_test: kif == NULL, if_xname %s\n", ifp->if_xname));
5741 return (PF_DROP);
5742 }
5743 if (kif->pfik_flags & PFI_IFLAG_SKIP)
5744 return (PF_PASS);
5745
5746 #ifdef DIAGNOSTIC
5747 if ((m->m_flags & M_PKTHDR) == 0)
5748 panic("non-M_PKTHDR is passed to pf_test");
5749 #endif /* DIAGNOSTIC */
5750
5751 if (m->m_pkthdr.len < (int)sizeof(*h)) {
5752 action = PF_DROP;
5753 REASON_SET_NOPTR(&reason, PFRES_SHORT);
5754 log = 1;
5755 goto done;
5756 }
5757
5758 #ifdef __NetBSD__
5759 if ((pf_mtag = pf_get_mtag(m)) == NULL) {
5760 DPFPRINTF(PF_DEBUG_URGENT,
5761 ("pf_test: pf_get_mtag returned NULL\n"));
5762 return (PF_DROP);
5763 }
5764 if (pf_mtag->flags & PF_TAG_GENERATED)
5765 return (PF_PASS);
5766 #else
5767 if (m->m_pkthdr.pf.flags & PF_TAG_GENERATED)
5768 return (PF_PASS);
5769 #endif /* !__NetBSD__ */
5770
5771 /* We do IP header normalization and packet reassembly here */
5772 if (pf_normalize_ip(m0, dir, kif, &reason, &pd) != PF_PASS) {
5773 action = PF_DROP;
5774 goto done;
5775 }
5776 m = *m0; /* pf_normalize messes with m0 */
5777 h = mtod(m, struct ip *);
5778
5779 off = h->ip_hl << 2;
5780 if (off < (int)sizeof(*h)) {
5781 action = PF_DROP;
5782 REASON_SET_NOPTR(&reason, PFRES_SHORT);
5783 log = 1;
5784 goto done;
5785 }
5786
5787 pd.src = (struct pf_addr *)&h->ip_src;
5788 pd.dst = (struct pf_addr *)&h->ip_dst;
5789 PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET);
5790 pd.ip_sum = &h->ip_sum;
5791 pd.proto = h->ip_p;
5792 pd.af = AF_INET;
5793 pd.tos = h->ip_tos;
5794 pd.tot_len = ntohs(h->ip_len);
5795 pd.eh = eh;
5796
5797 /* handle fragments that didn't get reassembled by normalization */
5798 if (h->ip_off & htons(IP_MF | IP_OFFMASK)) {
5799 action = pf_test_fragment(&r, dir, kif, m, h,
5800 &pd, &a, &ruleset);
5801 goto done;
5802 }
5803
5804 switch (h->ip_p) {
5805
5806 case IPPROTO_TCP: {
5807 struct tcphdr th;
5808
5809 pd.hdr.tcp = &th;
5810 if (!pf_pull_hdr(m, off, &th, sizeof(th),
5811 &action, &reason, AF_INET)) {
5812 log = action != PF_PASS;
5813 goto done;
5814 }
5815 pd.p_len = pd.tot_len - off - (th.th_off << 2);
5816 #if defined(ALTQ) && defined(__NetBSD__)
5817 if ((th.th_flags & TH_ACK) && pd.p_len == 0)
5818 pqid = 1;
5819 #endif
5820 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
5821 if (action == PF_DROP)
5822 goto done;
5823 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
5824 &reason);
5825 if (action == PF_PASS) {
5826 #if NPFSYNC
5827 pfsync_update_state(s);
5828 #endif /* NPFSYNC */
5829 r = s->rule.ptr;
5830 a = s->anchor.ptr;
5831 log = s->log;
5832 } else if (s == NULL)
5833 action = pf_test_rule(&r, &s, dir, kif,
5834 m, off, h, &pd, &a, &ruleset, NULL);
5835 break;
5836 }
5837
5838 case IPPROTO_UDP: {
5839 struct udphdr uh;
5840
5841 pd.hdr.udp = &uh;
5842 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
5843 &action, &reason, AF_INET)) {
5844 log = action != PF_PASS;
5845 goto done;
5846 }
5847 if (uh.uh_dport == 0 ||
5848 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
5849 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
5850 action = PF_DROP;
5851 REASON_SET_NOPTR(&reason, PFRES_SHORT);
5852 goto done;
5853 }
5854 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
5855 if (action == PF_PASS) {
5856 #if NPFSYNC
5857 pfsync_update_state(s);
5858 #endif /* NPFSYNC */
5859 r = s->rule.ptr;
5860 a = s->anchor.ptr;
5861 log = s->log;
5862 } else if (s == NULL)
5863 action = pf_test_rule(&r, &s, dir, kif,
5864 m, off, h, &pd, &a, &ruleset, NULL);
5865 break;
5866 }
5867
5868 case IPPROTO_ICMP: {
5869 struct icmp ih;
5870
5871 pd.hdr.icmp = &ih;
5872 if (!pf_pull_hdr(m, off, &ih, ICMP_MINLEN,
5873 &action, &reason, AF_INET)) {
5874 log = action != PF_PASS;
5875 goto done;
5876 }
5877 action = pf_test_state_icmp(&s, dir, kif, m, off, h, &pd,
5878 &reason);
5879 if (action == PF_PASS) {
5880 #if NPFSYNC
5881 pfsync_update_state(s);
5882 #endif /* NPFSYNC */
5883 r = s->rule.ptr;
5884 a = s->anchor.ptr;
5885 log = s->log;
5886 } else if (s == NULL)
5887 action = pf_test_rule(&r, &s, dir, kif,
5888 m, off, h, &pd, &a, &ruleset, NULL);
5889 break;
5890 }
5891
5892 #ifdef INET6
5893 case IPPROTO_ICMPV6: {
5894 action = PF_DROP;
5895 DPFPRINTF(PF_DEBUG_MISC,
5896 ("pf: dropping IPv4 packet with ICMPv6 payload\n"));
5897 goto done;
5898 }
5899 #endif
5900
5901 default:
5902 action = pf_test_state_other(&s, dir, kif, &pd);
5903 if (action == PF_PASS) {
5904 #if NPFSYNC
5905 pfsync_update_state(s);
5906 #endif /* NPFSYNC */
5907 r = s->rule.ptr;
5908 a = s->anchor.ptr;
5909 log = s->log;
5910 } else if (s == NULL)
5911 action = pf_test_rule(&r, &s, dir, kif, m, off, h,
5912 &pd, &a, &ruleset, NULL);
5913 break;
5914 }
5915
5916 done:
5917 if (action == PF_PASS && h->ip_hl > 5 &&
5918 !((s && s->allow_opts) || r->allow_opts)) {
5919 action = PF_DROP;
5920 REASON_SET_NOPTR(&reason, PFRES_IPOPTIONS);
5921 log = 1;
5922 DPFPRINTF(PF_DEBUG_MISC,
5923 ("pf: dropping packet with ip options\n"));
5924 }
5925
5926 if ((s && s->tag) || r->rtableid)
5927 pf_tag_packet(m, s ? s->tag : 0, r->rtableid);
5928
5929 #ifdef ALTQ
5930 if (action == PF_PASS && r->qid) {
5931 #ifdef __NetBSD__
5932 struct m_tag *mtag;
5933 struct altq_tag *atag;
5934
5935 mtag = m_tag_get(PACKET_TAG_ALTQ_QID, sizeof(*atag), M_NOWAIT);
5936 if (mtag != NULL) {
5937 atag = (struct altq_tag *)(mtag + 1);
5938 if (pqid || (pd.tos & IPTOS_LOWDELAY))
5939 atag->qid = r->pqid;
5940 else
5941 atag->qid = r->qid;
5942 /* add hints for ecn */
5943 atag->af = AF_INET;
5944 atag->hdr = h;
5945 m_tag_prepend(m, mtag);
5946 }
5947 #else
5948 if (pqid || (pd.tos & IPTOS_LOWDELAY))
5949 m->m_pkthdr.pf.qid = r->pqid;
5950 else
5951 m->m_pkthdr.pf.qid = r->qid;
5952 /* add hints for ecn */
5953 m->m_pkthdr.pf.hdr = h;
5954 #endif /* !__NetBSD__ */
5955 }
5956 #endif /* ALTQ */
5957
5958 /*
5959 * connections redirected to loopback should not match sockets
5960 * bound specifically to loopback due to security implications,
5961 * see tcp_input() and in_pcblookup_listen().
5962 */
5963 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
5964 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
5965 (s->nat_rule.ptr->action == PF_RDR ||
5966 s->nat_rule.ptr->action == PF_BINAT) &&
5967 (ntohl(pd.dst->v4.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET)
5968 #ifdef __NetBSD__
5969 pf_mtag->flags |= PF_TAG_TRANSLATE_LOCALHOST;
5970 #else
5971 m->m_pkthdr.pf.flags |= PF_TAG_TRANSLATE_LOCALHOST;
5972 #endif /* !__NetBSD__ */
5973
5974 if (log) {
5975 #if NPFLOG > 0
5976 struct pf_rule *lr;
5977
5978 if (s != NULL && s->nat_rule.ptr != NULL &&
5979 s->nat_rule.ptr->log & PF_LOG_ALL)
5980 lr = s->nat_rule.ptr;
5981 else
5982 lr = r;
5983 PFLOG_PACKET(kif, h, m, AF_INET, dir, reason, lr, a, ruleset,
5984 &pd);
5985 #endif
5986 }
5987
5988 kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
5989 kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++;
5990
5991 if (action == PF_PASS || r->action == PF_DROP) {
5992 dirndx = (dir == PF_OUT);
5993 r->packets[dirndx]++;
5994 r->bytes[dirndx] += pd.tot_len;
5995 if (a != NULL) {
5996 a->packets[dirndx]++;
5997 a->bytes[dirndx] += pd.tot_len;
5998 }
5999 if (s != NULL) {
6000 sk = s->state_key;
6001 if (s->nat_rule.ptr != NULL) {
6002 s->nat_rule.ptr->packets[dirndx]++;
6003 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len;
6004 }
6005 if (s->src_node != NULL) {
6006 s->src_node->packets[dirndx]++;
6007 s->src_node->bytes[dirndx] += pd.tot_len;
6008 }
6009 if (s->nat_src_node != NULL) {
6010 s->nat_src_node->packets[dirndx]++;
6011 s->nat_src_node->bytes[dirndx] += pd.tot_len;
6012 }
6013 dirndx = (dir == sk->direction) ? 0 : 1;
6014 s->packets[dirndx]++;
6015 s->bytes[dirndx] += pd.tot_len;
6016 }
6017 tr = r;
6018 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
6019 if (nr != NULL) {
6020 struct pf_addr *x;
6021 /*
6022 * XXX: we need to make sure that the addresses
6023 * passed to pfr_update_stats() are the same than
6024 * the addresses used during matching (pfr_match)
6025 */
6026 if (r == &pf_default_rule) {
6027 tr = nr;
6028 x = (sk == NULL || sk->direction == dir) ?
6029 &pd.baddr : &pd.naddr;
6030 } else
6031 x = (sk == NULL || sk->direction == dir) ?
6032 &pd.naddr : &pd.baddr;
6033 if (x == &pd.baddr || s == NULL) {
6034 /* we need to change the address */
6035 if (dir == PF_OUT)
6036 pd.src = x;
6037 else
6038 pd.dst = x;
6039 }
6040 }
6041 if (tr->src.addr.type == PF_ADDR_TABLE)
6042 pfr_update_stats(tr->src.addr.p.tbl, (sk == NULL ||
6043 sk->direction == dir) ?
6044 pd.src : pd.dst, pd.af,
6045 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
6046 tr->src.neg);
6047 if (tr->dst.addr.type == PF_ADDR_TABLE)
6048 pfr_update_stats(tr->dst.addr.p.tbl, (sk == NULL ||
6049 sk->direction == dir) ? pd.dst : pd.src, pd.af,
6050 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
6051 tr->dst.neg);
6052 }
6053
6054
6055 if (action == PF_SYNPROXY_DROP) {
6056 m_freem(*m0);
6057 *m0 = NULL;
6058 action = PF_PASS;
6059 } else if (r->rt)
6060 /* pf_route can free the mbuf causing *m0 to become NULL */
6061 pf_route(m0, r, dir, kif->pfik_ifp, s, &pd);
6062
6063 return (action);
6064 }
6065 #endif /* INET */
6066
6067 #ifdef INET6
6068 int
6069 pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0,
6070 struct ether_header *eh)
6071 {
6072 struct pfi_kif *kif;
6073 u_short action, reason = 0, log = 0;
6074 struct mbuf *m = *m0, *n = NULL;
6075 struct ip6_hdr *h = NULL; /* XXX gcc */
6076 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr;
6077 struct pf_state *s = NULL;
6078 struct pf_state_key *sk = NULL;
6079 struct pf_ruleset *ruleset = NULL;
6080 struct pf_pdesc pd;
6081 int off, terminal = 0, dirndx, rh_cnt = 0;
6082 #ifdef __NetBSD__
6083 struct pf_mtag *pf_mtag = NULL; /* XXX gcc */
6084 #endif /* __NetBSD__ */
6085
6086 if (!pf_status.running)
6087 return (PF_PASS);
6088
6089 memset(&pd, 0, sizeof(pd));
6090 if (ifp->if_type == IFT_CARP && ifp->if_carpdev)
6091 kif = (struct pfi_kif *)ifp->if_carpdev->if_pf_kif;
6092 else
6093 kif = (struct pfi_kif *)ifp->if_pf_kif;
6094
6095 if (kif == NULL) {
6096 DPFPRINTF(PF_DEBUG_URGENT,
6097 ("pf_test6: kif == NULL, if_xname %s\n", ifp->if_xname));
6098 return (PF_DROP);
6099 }
6100 if (kif->pfik_flags & PFI_IFLAG_SKIP)
6101 return (PF_PASS);
6102
6103 #ifdef DIAGNOSTIC
6104 if ((m->m_flags & M_PKTHDR) == 0)
6105 panic("non-M_PKTHDR is passed to pf_test6");
6106 #endif /* DIAGNOSTIC */
6107
6108 if (m->m_pkthdr.len < (int)sizeof(*h)) {
6109 action = PF_DROP;
6110 REASON_SET_NOPTR(&reason, PFRES_SHORT);
6111 log = 1;
6112 goto done;
6113 }
6114
6115 #ifdef __NetBSD__
6116 if ((pf_mtag = pf_get_mtag(m)) == NULL) {
6117 DPFPRINTF(PF_DEBUG_URGENT,
6118 ("pf_test6: pf_get_mtag returned NULL\n"));
6119 return (PF_DROP);
6120 }
6121 if (pf_mtag->flags & PF_TAG_GENERATED)
6122 return (PF_PASS);
6123 #else
6124 if (m->m_pkthdr.pf.flags & PF_TAG_GENERATED)
6125 return (PF_PASS);
6126 #endif /* !__NetBSD__ */
6127
6128 /* We do IP header normalization and packet reassembly here */
6129 if (pf_normalize_ip6(m0, dir, kif, &reason, &pd) != PF_PASS) {
6130 action = PF_DROP;
6131 goto done;
6132 }
6133 m = *m0; /* pf_normalize messes with m0 */
6134 h = mtod(m, struct ip6_hdr *);
6135
6136 #if 1
6137 /*
6138 * we do not support jumbogram yet. if we keep going, zero ip6_plen
6139 * will do something bad, so drop the packet for now.
6140 */
6141 if (htons(h->ip6_plen) == 0) {
6142 action = PF_DROP;
6143 REASON_SET_NOPTR(&reason, PFRES_NORM); /*XXX*/
6144 goto done;
6145 }
6146 #endif
6147
6148 pd.src = (struct pf_addr *)&h->ip6_src;
6149 pd.dst = (struct pf_addr *)&h->ip6_dst;
6150 PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET6);
6151 pd.ip_sum = NULL;
6152 pd.af = AF_INET6;
6153 pd.tos = 0;
6154 pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr);
6155 pd.eh = eh;
6156
6157 off = ((char *)h - m->m_data) + sizeof(struct ip6_hdr);
6158 pd.proto = h->ip6_nxt;
6159 do {
6160 switch (pd.proto) {
6161 case IPPROTO_FRAGMENT:
6162 action = pf_test_fragment(&r, dir, kif, m, h,
6163 &pd, &a, &ruleset);
6164 if (action == PF_DROP)
6165 REASON_SET_NOPTR(&reason, PFRES_FRAG);
6166 goto done;
6167 case IPPROTO_ROUTING: {
6168 struct ip6_rthdr rthdr;
6169
6170 if (rh_cnt++) {
6171 DPFPRINTF(PF_DEBUG_MISC,
6172 ("pf: IPv6 more than one rthdr\n"));
6173 action = PF_DROP;
6174 REASON_SET_NOPTR(&reason, PFRES_IPOPTIONS);
6175 log = 1;
6176 goto done;
6177 }
6178 if (!pf_pull_hdr(m, off, &rthdr, sizeof(rthdr), NULL,
6179 &reason, pd.af)) {
6180 DPFPRINTF(PF_DEBUG_MISC,
6181 ("pf: IPv6 short rthdr\n"));
6182 action = PF_DROP;
6183 REASON_SET_NOPTR(&reason, PFRES_SHORT);
6184 log = 1;
6185 goto done;
6186 }
6187 if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) {
6188 DPFPRINTF(PF_DEBUG_MISC,
6189 ("pf: IPv6 rthdr0\n"));
6190 action = PF_DROP;
6191 REASON_SET_NOPTR(&reason, PFRES_IPOPTIONS);
6192 log = 1;
6193 goto done;
6194 }
6195 /* FALLTHROUGH */
6196 }
6197 case IPPROTO_AH:
6198 case IPPROTO_HOPOPTS:
6199 case IPPROTO_DSTOPTS: {
6200 /* get next header and header length */
6201 struct ip6_ext opt6;
6202
6203 if (!pf_pull_hdr(m, off, &opt6, sizeof(opt6),
6204 NULL, &reason, pd.af)) {
6205 DPFPRINTF(PF_DEBUG_MISC,
6206 ("pf: IPv6 short opt\n"));
6207 action = PF_DROP;
6208 log = 1;
6209 goto done;
6210 }
6211 if (pd.proto == IPPROTO_AH)
6212 off += (opt6.ip6e_len + 2) * 4;
6213 else
6214 off += (opt6.ip6e_len + 1) * 8;
6215 pd.proto = opt6.ip6e_nxt;
6216 /* goto the next header */
6217 break;
6218 }
6219 default:
6220 terminal++;
6221 break;
6222 }
6223 } while (!terminal);
6224
6225 /* if there's no routing header, use unmodified mbuf for checksumming */
6226 if (!n)
6227 n = m;
6228
6229 switch (pd.proto) {
6230
6231 case IPPROTO_TCP: {
6232 struct tcphdr th;
6233
6234 pd.hdr.tcp = &th;
6235 if (!pf_pull_hdr(m, off, &th, sizeof(th),
6236 &action, &reason, AF_INET6)) {
6237 log = action != PF_PASS;
6238 goto done;
6239 }
6240 pd.p_len = pd.tot_len - off - (th.th_off << 2);
6241 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
6242 if (action == PF_DROP)
6243 goto done;
6244 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
6245 &reason);
6246 if (action == PF_PASS) {
6247 #if NPFSYNC
6248 pfsync_update_state(s);
6249 #endif /* NPFSYNC */
6250 r = s->rule.ptr;
6251 a = s->anchor.ptr;
6252 log = s->log;
6253 } else if (s == NULL)
6254 action = pf_test_rule(&r, &s, dir, kif,
6255 m, off, h, &pd, &a, &ruleset, NULL);
6256 break;
6257 }
6258
6259 case IPPROTO_UDP: {
6260 struct udphdr uh;
6261
6262 pd.hdr.udp = &uh;
6263 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
6264 &action, &reason, AF_INET6)) {
6265 log = action != PF_PASS;
6266 goto done;
6267 }
6268 if (uh.uh_dport == 0 ||
6269 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
6270 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
6271 action = PF_DROP;
6272 REASON_SET_NOPTR(&reason, PFRES_SHORT);
6273 goto done;
6274 }
6275 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
6276 if (action == PF_PASS) {
6277 #if NPFSYNC
6278 pfsync_update_state(s);
6279 #endif /* NPFSYNC */
6280 r = s->rule.ptr;
6281 a = s->anchor.ptr;
6282 log = s->log;
6283 } else if (s == NULL)
6284 action = pf_test_rule(&r, &s, dir, kif,
6285 m, off, h, &pd, &a, &ruleset, NULL);
6286 break;
6287 }
6288
6289 #ifdef INET
6290 case IPPROTO_ICMP: {
6291 action = PF_DROP;
6292 DPFPRINTF(PF_DEBUG_MISC,
6293 ("pf: dropping IPv6 packet with ICMPv4 payload\n"));
6294 goto done;
6295 }
6296 #endif
6297
6298 case IPPROTO_ICMPV6: {
6299 struct icmp6_hdr ih;
6300
6301 pd.hdr.icmp6 = &ih;
6302 if (!pf_pull_hdr(m, off, &ih, sizeof(ih),
6303 &action, &reason, AF_INET6)) {
6304 log = action != PF_PASS;
6305 goto done;
6306 }
6307 action = pf_test_state_icmp(&s, dir, kif,
6308 m, off, h, &pd, &reason);
6309 if (action == PF_PASS) {
6310 #if NPFSYNC
6311 pfsync_update_state(s);
6312 #endif /* NPFSYNC */
6313 r = s->rule.ptr;
6314 a = s->anchor.ptr;
6315 log = s->log;
6316 } else if (s == NULL)
6317 action = pf_test_rule(&r, &s, dir, kif,
6318 m, off, h, &pd, &a, &ruleset, NULL);
6319 break;
6320 }
6321
6322 default:
6323 action = pf_test_state_other(&s, dir, kif, &pd);
6324 if (action == PF_PASS) {
6325 #if NPFSYNC
6326 pfsync_update_state(s);
6327 #endif /* NPFSYNC */
6328 r = s->rule.ptr;
6329 a = s->anchor.ptr;
6330 log = s->log;
6331 } else if (s == NULL)
6332 action = pf_test_rule(&r, &s, dir, kif, m, off, h,
6333 &pd, &a, &ruleset, NULL);
6334 break;
6335 }
6336
6337 done:
6338 if (n != m) {
6339 m_freem(n);
6340 n = NULL;
6341 }
6342
6343 /* handle dangerous IPv6 extension headers. */
6344 if (action == PF_PASS && rh_cnt &&
6345 !((s && s->allow_opts) || r->allow_opts)) {
6346 action = PF_DROP;
6347 REASON_SET_NOPTR(&reason, PFRES_IPOPTIONS);
6348 log = 1;
6349 DPFPRINTF(PF_DEBUG_MISC,
6350 ("pf: dropping packet with dangerous v6 headers\n"));
6351 }
6352
6353 if ((s && s->tag) || r->rtableid)
6354 pf_tag_packet(m, s ? s->tag : 0, r->rtableid);
6355
6356 #ifdef ALTQ
6357 if (action == PF_PASS && r->qid) {
6358 #ifdef __NetBSD__
6359 struct m_tag *mtag;
6360 struct altq_tag *atag;
6361
6362 mtag = m_tag_get(PACKET_TAG_ALTQ_QID, sizeof(*atag), M_NOWAIT);
6363 if (mtag != NULL) {
6364 atag = (struct altq_tag *)(mtag + 1);
6365 if (pd.tos & IPTOS_LOWDELAY)
6366 atag->qid = r->pqid;
6367 else
6368 atag->qid = r->qid;
6369 /* add hints for ecn */
6370 atag->af = AF_INET6;
6371 atag->hdr = h;
6372 m_tag_prepend(m, mtag);
6373 }
6374 #else
6375 if (pd.tos & IPTOS_LOWDELAY)
6376 m->m_pkthdr.pf.qid = r->pqid;
6377 else
6378 m->m_pkthdr.pf.qid = r->qid;
6379 /* add hints for ecn */
6380 m->m_pkthdr.pf.hdr = h;
6381 #endif /* !__NetBSD__ */
6382 }
6383 #endif /* ALTQ */
6384
6385 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
6386 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
6387 (s->nat_rule.ptr->action == PF_RDR ||
6388 s->nat_rule.ptr->action == PF_BINAT) &&
6389 IN6_IS_ADDR_LOOPBACK(&pd.dst->v6))
6390 #ifdef __NetBSD__
6391 pf_mtag->flags |= PF_TAG_TRANSLATE_LOCALHOST;
6392 #else
6393 m->m_pkthdr.pf.flags |= PF_TAG_TRANSLATE_LOCALHOST;
6394 #endif /* !__NetBSD__ */
6395
6396 if (log) {
6397 #if NPFLOG > 0
6398 struct pf_rule *lr;
6399
6400 if (s != NULL && s->nat_rule.ptr != NULL &&
6401 s->nat_rule.ptr->log & PF_LOG_ALL)
6402 lr = s->nat_rule.ptr;
6403 else
6404 lr = r;
6405 PFLOG_PACKET(kif, h, m, AF_INET6, dir, reason, lr, a, ruleset,
6406 &pd);
6407 #endif
6408 }
6409
6410 kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
6411 kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++;
6412
6413 if (action == PF_PASS || r->action == PF_DROP) {
6414 dirndx = (dir == PF_OUT);
6415 r->packets[dirndx]++;
6416 r->bytes[dirndx] += pd.tot_len;
6417 if (a != NULL) {
6418 a->packets[dirndx]++;
6419 a->bytes[dirndx] += pd.tot_len;
6420 }
6421 if (s != NULL) {
6422 sk = s->state_key;
6423 if (s->nat_rule.ptr != NULL) {
6424 s->nat_rule.ptr->packets[dirndx]++;
6425 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len;
6426 }
6427 if (s->src_node != NULL) {
6428 s->src_node->packets[dirndx]++;
6429 s->src_node->bytes[dirndx] += pd.tot_len;
6430 }
6431 if (s->nat_src_node != NULL) {
6432 s->nat_src_node->packets[dirndx]++;
6433 s->nat_src_node->bytes[dirndx] += pd.tot_len;
6434 }
6435 dirndx = (dir == sk->direction) ? 0 : 1;
6436 s->packets[dirndx]++;
6437 s->bytes[dirndx] += pd.tot_len;
6438 }
6439 tr = r;
6440 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
6441 if (nr != NULL) {
6442 struct pf_addr *x;
6443 /*
6444 * XXX: we need to make sure that the addresses
6445 * passed to pfr_update_stats() are the same than
6446 * the addresses used during matching (pfr_match)
6447 */
6448 if (r == &pf_default_rule) {
6449 tr = nr;
6450 x = (s == NULL || sk->direction == dir) ?
6451 &pd.baddr : &pd.naddr;
6452 } else {
6453 x = (s == NULL || sk->direction == dir) ?
6454 &pd.naddr : &pd.baddr;
6455 }
6456 if (x == &pd.baddr || s == NULL) {
6457 if (dir == PF_OUT)
6458 pd.src = x;
6459 else
6460 pd.dst = x;
6461 }
6462 }
6463 if (tr->src.addr.type == PF_ADDR_TABLE)
6464 pfr_update_stats(tr->src.addr.p.tbl, (sk == NULL ||
6465 sk->direction == dir) ? pd.src : pd.dst, pd.af,
6466 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
6467 tr->src.neg);
6468 if (tr->dst.addr.type == PF_ADDR_TABLE)
6469 pfr_update_stats(tr->dst.addr.p.tbl, (sk == NULL ||
6470 sk->direction == dir) ? pd.dst : pd.src, pd.af,
6471 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
6472 tr->dst.neg);
6473 }
6474
6475
6476 if (action == PF_SYNPROXY_DROP) {
6477 m_freem(*m0);
6478 *m0 = NULL;
6479 action = PF_PASS;
6480 } else if (r->rt)
6481 /* pf_route6 can free the mbuf causing *m0 to become NULL */
6482 pf_route6(m0, r, dir, kif->pfik_ifp, s, &pd);
6483
6484 return (action);
6485 }
6486 #endif /* INET6 */
6487
6488 int
6489 pf_check_congestion(struct ifqueue *ifq)
6490 {
6491 #ifdef __NetBSD__
6492 // XXX: not handled anyway
6493 KASSERT(ifq == NULL);
6494 return (0);
6495 #else
6496 if (ifq->ifq_congestion)
6497 return (1);
6498 else
6499 return (0);
6500 #endif /* !__NetBSD__ */
6501 }
6502