2/* $OpenBSD: pf.c,v 1.433.2.2 2004/07/17 03:22:34 brad Exp $ */
3/* add $OpenBSD: pf.c,v 1.448 2004/05/11 07:34:11 dhartmei Exp $ */
4
5/*
6 * Copyright (c) 2001 Daniel Hartmeier
7 * Copyright (c) 2002,2003 Henning Brauer
8 * All rights reserved.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 *
14 * - Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * - Redistributions in binary form must reproduce the above
17 * copyright notice, this list of conditions and the following
18 * disclaimer in the documentation and/or other materials provided
19 * with the distribution.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
25 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
27 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
29 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
31 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
32 * POSSIBILITY OF SUCH DAMAGE.
33 *
34 * Effort sponsored in part by the Defense Advanced Research Projects
35 * Agency (DARPA) and Air Force Research Laboratory, Air Force
36 * Materiel Command, USAF, under agreement number F30602-01-2-0537.
37 *
38 */
39
40#ifdef __FreeBSD__
41#include "opt_inet.h"
42#include "opt_inet6.h"
43#endif
44
45#ifdef __FreeBSD__
46#include "opt_bpf.h"
47#include "opt_pf.h"
48#define NBPFILTER DEV_BPF
49#define NPFLOG DEV_PFLOG
50#define NPFSYNC DEV_PFSYNC
51#else
52#include "bpfilter.h"
53#include "pflog.h"
54#include "pfsync.h"
55#endif
56
57#include <sys/param.h>
58#include <sys/systm.h>
59#include <sys/mbuf.h>
60#include <sys/filio.h>
61#include <sys/socket.h>
62#include <sys/socketvar.h>
63#include <sys/kernel.h>
64#include <sys/time.h>
65#ifdef __FreeBSD__
66#include <sys/sysctl.h>
67#include <sys/endian.h>
68#else
69#include <sys/pool.h>
70#endif
71
72#include <net/if.h>
73#include <net/if_types.h>
74#include <net/bpf.h>
75#include <net/route.h>
76
77#include <netinet/in.h>
78#include <netinet/in_var.h>
79#include <netinet/in_systm.h>
80#include <netinet/ip.h>
81#include <netinet/ip_var.h>
82#include <netinet/tcp.h>
83#include <netinet/tcp_seq.h>
84#include <netinet/udp.h>
85#include <netinet/ip_icmp.h>
86#include <netinet/in_pcb.h>
87#include <netinet/tcp_timer.h>
88#include <netinet/tcp_var.h>
89#include <netinet/udp_var.h>
90#include <netinet/icmp_var.h>
91
92#ifndef __FreeBSD__
93#include <dev/rndvar.h>
94#endif
95#include <net/pfvar.h>
96#include <net/if_pflog.h>
97
98#if NPFSYNC > 0
99#include <net/if_pfsync.h>
100#endif /* NPFSYNC > 0 */
101
102#ifdef INET6
103#include <netinet/ip6.h>
104#include <netinet/in_pcb.h>
105#include <netinet/icmp6.h>
106#include <netinet6/nd6.h>
107#ifdef __FreeBSD__
108#include <netinet6/ip6_var.h>
109#include <netinet6/in6_pcb.h>
110#endif
111#endif /* INET6 */
112
113#ifdef __FreeBSD__
114#include <machine/in_cksum.h>
115#include <sys/limits.h>
116#include <sys/ucred.h>
117
118extern int ip_optcopy(struct ip *, struct ip *);
119#endif
120
121#define DPFPRINTF(n, x) if (pf_status.debug >= (n)) printf x
122
123/*
124 * Global variables
125 */
126
127struct pf_anchorqueue pf_anchors;
128struct pf_ruleset pf_main_ruleset;
129struct pf_altqqueue pf_altqs[2];
130struct pf_palist pf_pabuf;
131struct pf_altqqueue *pf_altqs_active;
132struct pf_altqqueue *pf_altqs_inactive;
133struct pf_status pf_status;
134
135u_int32_t ticket_altqs_active;
136u_int32_t ticket_altqs_inactive;
137int altqs_inactive_open;
138u_int32_t ticket_pabuf;
139
140#ifdef __FreeBSD__
141struct callout pf_expire_to; /* expire timeout */
142#else
143struct timeout pf_expire_to; /* expire timeout */
144#endif
145
146
147#ifdef __FreeBSD__
148uma_zone_t pf_src_tree_pl, pf_rule_pl;
149uma_zone_t pf_state_pl, pf_altq_pl, pf_pooladdr_pl;
150#else
151struct pool pf_src_tree_pl, pf_rule_pl;
152struct pool pf_state_pl, pf_altq_pl, pf_pooladdr_pl;
153#endif
154
155void pf_print_host(struct pf_addr *, u_int16_t, u_int8_t);
156void pf_print_state(struct pf_state *);
157void pf_print_flags(u_int8_t);
158
159void pf_change_ap(struct pf_addr *, u_int16_t *,
160 u_int16_t *, u_int16_t *, struct pf_addr *,
161 u_int16_t, u_int8_t, sa_family_t);
162#ifdef INET6
163void pf_change_a6(struct pf_addr *, u_int16_t *,
164 struct pf_addr *, u_int8_t);
165#endif /* INET6 */
166void pf_change_icmp(struct pf_addr *, u_int16_t *,
167 struct pf_addr *, struct pf_addr *, u_int16_t,
168 u_int16_t *, u_int16_t *, u_int16_t *,
169 u_int16_t *, u_int8_t, sa_family_t);
170void pf_send_tcp(const struct pf_rule *, sa_family_t,
171 const struct pf_addr *, const struct pf_addr *,
172 u_int16_t, u_int16_t, u_int32_t, u_int32_t,
173 u_int8_t, u_int16_t, u_int16_t, u_int8_t);
174void pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t,
175 sa_family_t, struct pf_rule *);
176struct pf_rule *pf_match_translation(struct pf_pdesc *, struct mbuf *,
177 int, int, struct pfi_kif *,
178 struct pf_addr *, u_int16_t, struct pf_addr *,
179 u_int16_t, int);
180struct pf_rule *pf_get_translation(struct pf_pdesc *, struct mbuf *,
181 int, int, struct pfi_kif *, struct pf_src_node **,
182 struct pf_addr *, u_int16_t,
183 struct pf_addr *, u_int16_t,
184 struct pf_addr *, u_int16_t *);
185int pf_test_tcp(struct pf_rule **, struct pf_state **,
186 int, struct pfi_kif *, struct mbuf *, int,
187 void *, struct pf_pdesc *, struct pf_rule **,
188#ifdef __FreeBSD__
189 struct pf_ruleset **, struct inpcb *);
190#else
191 struct pf_ruleset **);
192#endif
193int pf_test_udp(struct pf_rule **, struct pf_state **,
194 int, struct pfi_kif *, struct mbuf *, int,
195 void *, struct pf_pdesc *, struct pf_rule **,
196#ifdef __FreeBSD__
197 struct pf_ruleset **, struct inpcb *);
198#else
199 struct pf_ruleset **);
200#endif
201int pf_test_icmp(struct pf_rule **, struct pf_state **,
202 int, struct pfi_kif *, struct mbuf *, int,
203 void *, struct pf_pdesc *, struct pf_rule **,
204 struct pf_ruleset **);
205int pf_test_other(struct pf_rule **, struct pf_state **,
206 int, struct pfi_kif *, struct mbuf *, int, void *,
207 struct pf_pdesc *, struct pf_rule **,
208 struct pf_ruleset **);
209int pf_test_fragment(struct pf_rule **, int,
210 struct pfi_kif *, struct mbuf *, void *,
211 struct pf_pdesc *, struct pf_rule **,
212 struct pf_ruleset **);
213int pf_test_state_tcp(struct pf_state **, int,
214 struct pfi_kif *, struct mbuf *, int,
215 void *, struct pf_pdesc *, u_short *);
216int pf_test_state_udp(struct pf_state **, int,
217 struct pfi_kif *, struct mbuf *, int,
218 void *, struct pf_pdesc *);
219int pf_test_state_icmp(struct pf_state **, int,
220 struct pfi_kif *, struct mbuf *, int,
221 void *, struct pf_pdesc *);
222int pf_test_state_other(struct pf_state **, int,
223 struct pfi_kif *, struct pf_pdesc *);
224struct pf_tag *pf_get_tag(struct mbuf *);
225int pf_match_tag(struct mbuf *, struct pf_rule *,
226 struct pf_rule *, struct pf_tag *, int *);
227void pf_hash(struct pf_addr *, struct pf_addr *,
228 struct pf_poolhashkey *, sa_family_t);
229int pf_map_addr(u_int8_t, struct pf_rule *,
230 struct pf_addr *, struct pf_addr *,
231 struct pf_addr *, struct pf_src_node **);
232int pf_get_sport(sa_family_t, u_int8_t, struct pf_rule *,
233 struct pf_addr *, struct pf_addr *, u_int16_t,
234 struct pf_addr *, u_int16_t*, u_int16_t, u_int16_t,
235 struct pf_src_node **);
236void pf_route(struct mbuf **, struct pf_rule *, int,
237 struct ifnet *, struct pf_state *);
238void pf_route6(struct mbuf **, struct pf_rule *, int,
239 struct ifnet *, struct pf_state *);
240#ifdef __FreeBSD__
241int pf_socket_lookup(uid_t *, gid_t *,
242 int, struct pf_pdesc *, struct inpcb *);
243#else
244int pf_socket_lookup(uid_t *, gid_t *,
245 int, struct pf_pdesc *);
246#endif
247u_int8_t pf_get_wscale(struct mbuf *, int, u_int16_t,
248 sa_family_t);
249u_int16_t pf_get_mss(struct mbuf *, int, u_int16_t,
250 sa_family_t);
251u_int16_t pf_calc_mss(struct pf_addr *, sa_family_t,
252 u_int16_t);
253void pf_set_rt_ifp(struct pf_state *,
254 struct pf_addr *);
255int pf_check_proto_cksum(struct mbuf *, int, int,
256 u_int8_t, sa_family_t);
257int pf_addr_wrap_neq(struct pf_addr_wrap *,
258 struct pf_addr_wrap *);
259static int pf_add_mbuf_tag(struct mbuf *, u_int);
260struct pf_state *pf_find_state_recurse(struct pfi_kif *,
261 struct pf_state *, u_int8_t);
262
263#ifdef __FreeBSD__
264int in4_cksum(struct mbuf *m, u_int8_t nxt, int off, int len);
265
266struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX];
267#else
268struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] = {
269 { &pf_state_pl, PFSTATE_HIWAT },
270 { &pf_src_tree_pl, PFSNODE_HIWAT },
271 { &pf_frent_pl, PFFRAG_FRENT_HIWAT }
272};
273#endif
274
275#define STATE_LOOKUP() \
276 do { \
277 if (direction == PF_IN) \
278 *state = pf_find_state_recurse( \
279 kif, &key, PF_EXT_GWY); \
280 else \
281 *state = pf_find_state_recurse( \
282 kif, &key, PF_LAN_EXT); \
283 if (*state == NULL) \
284 return (PF_DROP); \
285 if (direction == PF_OUT && \
286 (((*state)->rule.ptr->rt == PF_ROUTETO && \
287 (*state)->rule.ptr->direction == PF_OUT) || \
288 ((*state)->rule.ptr->rt == PF_REPLYTO && \
289 (*state)->rule.ptr->direction == PF_IN)) && \
290 (*state)->rt_kif != NULL && \
291 (*state)->rt_kif != kif) \
292 return (PF_PASS); \
293 } while (0)
294
295#define STATE_TRANSLATE(s) \
296 (s)->lan.addr.addr32[0] != (s)->gwy.addr.addr32[0] || \
297 ((s)->af == AF_INET6 && \
298 ((s)->lan.addr.addr32[1] != (s)->gwy.addr.addr32[1] || \
299 (s)->lan.addr.addr32[2] != (s)->gwy.addr.addr32[2] || \
300 (s)->lan.addr.addr32[3] != (s)->gwy.addr.addr32[3])) || \
301 (s)->lan.port != (s)->gwy.port
302
303#define BOUND_IFACE(r, k) (((r)->rule_flag & PFRULE_IFBOUND) ? (k) : \
304 ((r)->rule_flag & PFRULE_GRBOUND) ? (k)->pfik_parent : \
305 (k)->pfik_parent->pfik_parent)
306
307#ifndef __FreeBSD__
308static __inline int pf_src_compare(struct pf_src_node *, struct pf_src_node *);
309static __inline int pf_state_compare_lan_ext(struct pf_state *,
310 struct pf_state *);
311static __inline int pf_state_compare_ext_gwy(struct pf_state *,
312 struct pf_state *);
313static __inline int pf_state_compare_id(struct pf_state *,
314 struct pf_state *);
315#else
316static int pf_src_compare(struct pf_src_node *, struct pf_src_node *);
317static int pf_state_compare_lan_ext(struct pf_state *,
318 struct pf_state *);
319static int pf_state_compare_ext_gwy(struct pf_state *,
320 struct pf_state *);
321static int pf_state_compare_id(struct pf_state *,
322 struct pf_state *);
323#endif
324
325struct pf_src_tree tree_src_tracking;
326
327struct pf_state_tree_id tree_id;
328struct pf_state_queue state_updates;
329
330RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare);
331RB_GENERATE(pf_state_tree_lan_ext, pf_state,
332 u.s.entry_lan_ext, pf_state_compare_lan_ext);
333RB_GENERATE(pf_state_tree_ext_gwy, pf_state,
334 u.s.entry_ext_gwy, pf_state_compare_ext_gwy);
335RB_GENERATE(pf_state_tree_id, pf_state,
336 u.s.entry_id, pf_state_compare_id);
337
338#ifdef __FreeBSD__
339static int
340#else
341static __inline int
342#endif
343pf_src_compare(struct pf_src_node *a, struct pf_src_node *b)
344{
345 int diff;
346
347 if (a->rule.ptr > b->rule.ptr)
348 return (1);
349 if (a->rule.ptr < b->rule.ptr)
350 return (-1);
351 if ((diff = a->af - b->af) != 0)
352 return (diff);
353 switch (a->af) {
354#ifdef INET
355 case AF_INET:
356 if (a->addr.addr32[0] > b->addr.addr32[0])
357 return (1);
358 if (a->addr.addr32[0] < b->addr.addr32[0])
359 return (-1);
360 break;
361#endif /* INET */
362#ifdef INET6
363 case AF_INET6:
364 if (a->addr.addr32[3] > b->addr.addr32[3])
365 return (1);
366 if (a->addr.addr32[3] < b->addr.addr32[3])
367 return (-1);
368 if (a->addr.addr32[2] > b->addr.addr32[2])
369 return (1);
370 if (a->addr.addr32[2] < b->addr.addr32[2])
371 return (-1);
372 if (a->addr.addr32[1] > b->addr.addr32[1])
373 return (1);
374 if (a->addr.addr32[1] < b->addr.addr32[1])
375 return (-1);
376 if (a->addr.addr32[0] > b->addr.addr32[0])
377 return (1);
378 if (a->addr.addr32[0] < b->addr.addr32[0])
379 return (-1);
380 break;
381#endif /* INET6 */
382 }
383 return (0);
384}
385
386#ifdef __FreeBSD__
387static int
388#else
389static __inline int
390#endif
391pf_state_compare_lan_ext(struct pf_state *a, struct pf_state *b)
392{
393 int diff;
394
395 if ((diff = a->proto - b->proto) != 0)
396 return (diff);
397 if ((diff = a->af - b->af) != 0)
398 return (diff);
399 switch (a->af) {
400#ifdef INET
401 case AF_INET:
402 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
403 return (1);
404 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
405 return (-1);
406 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
407 return (1);
408 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
409 return (-1);
410 break;
411#endif /* INET */
412#ifdef INET6
413 case AF_INET6:
414 if (a->lan.addr.addr32[3] > b->lan.addr.addr32[3])
415 return (1);
416 if (a->lan.addr.addr32[3] < b->lan.addr.addr32[3])
417 return (-1);
418 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
419 return (1);
420 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
421 return (-1);
422 if (a->lan.addr.addr32[2] > b->lan.addr.addr32[2])
423 return (1);
424 if (a->lan.addr.addr32[2] < b->lan.addr.addr32[2])
425 return (-1);
426 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
427 return (1);
428 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
429 return (-1);
430 if (a->lan.addr.addr32[1] > b->lan.addr.addr32[1])
431 return (1);
432 if (a->lan.addr.addr32[1] < b->lan.addr.addr32[1])
433 return (-1);
434 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
435 return (1);
436 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
437 return (-1);
438 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
439 return (1);
440 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
441 return (-1);
442 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
443 return (1);
444 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
445 return (-1);
446 break;
447#endif /* INET6 */
448 }
449
450 if ((diff = a->lan.port - b->lan.port) != 0)
451 return (diff);
452 if ((diff = a->ext.port - b->ext.port) != 0)
453 return (diff);
454
455 return (0);
456}
457
458#ifdef __FreeBSD__
459static int
460#else
461static __inline int
462#endif
463pf_state_compare_ext_gwy(struct pf_state *a, struct pf_state *b)
464{
465 int diff;
466
467 if ((diff = a->proto - b->proto) != 0)
468 return (diff);
469 if ((diff = a->af - b->af) != 0)
470 return (diff);
471 switch (a->af) {
472#ifdef INET
473 case AF_INET:
474 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
475 return (1);
476 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
477 return (-1);
478 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
479 return (1);
480 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
481 return (-1);
482 break;
483#endif /* INET */
484#ifdef INET6
485 case AF_INET6:
486 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
487 return (1);
488 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
489 return (-1);
490 if (a->gwy.addr.addr32[3] > b->gwy.addr.addr32[3])
491 return (1);
492 if (a->gwy.addr.addr32[3] < b->gwy.addr.addr32[3])
493 return (-1);
494 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
495 return (1);
496 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
497 return (-1);
498 if (a->gwy.addr.addr32[2] > b->gwy.addr.addr32[2])
499 return (1);
500 if (a->gwy.addr.addr32[2] < b->gwy.addr.addr32[2])
501 return (-1);
502 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
503 return (1);
504 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
505 return (-1);
506 if (a->gwy.addr.addr32[1] > b->gwy.addr.addr32[1])
507 return (1);
508 if (a->gwy.addr.addr32[1] < b->gwy.addr.addr32[1])
509 return (-1);
510 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
511 return (1);
512 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
513 return (-1);
514 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
515 return (1);
516 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
517 return (-1);
518 break;
519#endif /* INET6 */
520 }
521
522 if ((diff = a->ext.port - b->ext.port) != 0)
523 return (diff);
524 if ((diff = a->gwy.port - b->gwy.port) != 0)
525 return (diff);
526
527 return (0);
528}
529
530#ifdef __FreeBSD__
531static int
532#else
533static __inline int
534#endif
535pf_state_compare_id(struct pf_state *a, struct pf_state *b)
536{
537 if (a->id > b->id)
538 return (1);
539 if (a->id < b->id)
540 return (-1);
541 if (a->creatorid > b->creatorid)
542 return (1);
543 if (a->creatorid < b->creatorid)
544 return (-1);
545
546 return (0);
547}
548
549#ifdef INET6
550void
551pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af)
552{
553 switch (af) {
554#ifdef INET
555 case AF_INET:
556 dst->addr32[0] = src->addr32[0];
557 break;
558#endif /* INET */
559 case AF_INET6:
560 dst->addr32[0] = src->addr32[0];
561 dst->addr32[1] = src->addr32[1];
562 dst->addr32[2] = src->addr32[2];
563 dst->addr32[3] = src->addr32[3];
564 break;
565 }
566}
567#endif
568
569struct pf_state *
570pf_find_state_byid(struct pf_state *key)
571{
572 pf_status.fcounters[FCNT_STATE_SEARCH]++;
573 return (RB_FIND(pf_state_tree_id, &tree_id, key));
574}
575
576struct pf_state *
577pf_find_state_recurse(struct pfi_kif *kif, struct pf_state *key, u_int8_t tree)
578{
579 struct pf_state *s;
580
581 pf_status.fcounters[FCNT_STATE_SEARCH]++;
582
583 switch (tree) {
584 case PF_LAN_EXT:
585 for (; kif != NULL; kif = kif->pfik_parent) {
586 s = RB_FIND(pf_state_tree_lan_ext,
587 &kif->pfik_lan_ext, key);
588 if (s != NULL)
589 return (s);
590 }
591 return (NULL);
592 case PF_EXT_GWY:
593 for (; kif != NULL; kif = kif->pfik_parent) {
594 s = RB_FIND(pf_state_tree_ext_gwy,
595 &kif->pfik_ext_gwy, key);
596 if (s != NULL)
597 return (s);
598 }
599 return (NULL);
600 default:
601 panic("pf_find_state_recurse");
602 }
603}
604
605struct pf_state *
606pf_find_state_all(struct pf_state *key, u_int8_t tree, int *more)
607{
608 struct pf_state *s, *ss = NULL;
609 struct pfi_kif *kif;
610
611 pf_status.fcounters[FCNT_STATE_SEARCH]++;
612
613 switch (tree) {
614 case PF_LAN_EXT:
615 TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states) {
616 s = RB_FIND(pf_state_tree_lan_ext,
617 &kif->pfik_lan_ext, key);
618 if (s == NULL)
619 continue;
620 if (more == NULL)
621 return (s);
622 ss = s;
623 (*more)++;
624 }
625 return (ss);
626 case PF_EXT_GWY:
627 TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states) {
628 s = RB_FIND(pf_state_tree_ext_gwy,
629 &kif->pfik_ext_gwy, key);
630 if (s == NULL)
631 continue;
632 if (more == NULL)
633 return (s);
634 ss = s;
635 (*more)++;
636 }
637 return (ss);
638 default:
639 panic("pf_find_state_all");
640 }
641}
642
643int
644pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule,
645 struct pf_addr *src, sa_family_t af)
646{
647 struct pf_src_node k;
648
649 if (*sn == NULL) {
650 k.af = af;
651 PF_ACPY(&k.addr, src, af);
652 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
653 rule->rpool.opts & PF_POOL_STICKYADDR)
654 k.rule.ptr = rule;
655 else
656 k.rule.ptr = NULL;
657 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
658 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
659 }
660 if (*sn == NULL) {
661 if (!rule->max_src_nodes ||
662 rule->src_nodes < rule->max_src_nodes)
663 (*sn) = pool_get(&pf_src_tree_pl, PR_NOWAIT);
664 if ((*sn) == NULL)
665 return (-1);
666 bzero(*sn, sizeof(struct pf_src_node));
667 (*sn)->af = af;
668 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
669 rule->rpool.opts & PF_POOL_STICKYADDR)
670 (*sn)->rule.ptr = rule;
671 else
672 (*sn)->rule.ptr = NULL;
673 PF_ACPY(&(*sn)->addr, src, af);
674 if (RB_INSERT(pf_src_tree,
675 &tree_src_tracking, *sn) != NULL) {
676 if (pf_status.debug >= PF_DEBUG_MISC) {
677 printf("pf: src_tree insert failed: ");
678 pf_print_host(&(*sn)->addr, 0, af);
679 printf("\n");
680 }
681 pool_put(&pf_src_tree_pl, *sn);
682 return (-1);
683 }
684#ifdef __FreeBSD__
685 (*sn)->creation = time_second;
686#else
687 (*sn)->creation = time.tv_sec;
688#endif
689 (*sn)->ruletype = rule->action;
690 if ((*sn)->rule.ptr != NULL)
691 (*sn)->rule.ptr->src_nodes++;
692 pf_status.scounters[SCNT_SRC_NODE_INSERT]++;
693 pf_status.src_nodes++;
694 } else {
695 if (rule->max_src_states &&
696 (*sn)->states >= rule->max_src_states)
697 return (-1);
698 }
699 return (0);
700}
701
702int
703pf_insert_state(struct pfi_kif *kif, struct pf_state *state)
704{
705 /* Thou MUST NOT insert multiple duplicate keys */
706 state->u.s.kif = kif;
707 if (RB_INSERT(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state)) {
708 if (pf_status.debug >= PF_DEBUG_MISC) {
709 printf("pf: state insert failed: tree_lan_ext");
710 printf(" lan: ");
711 pf_print_host(&state->lan.addr, state->lan.port,
712 state->af);
713 printf(" gwy: ");
714 pf_print_host(&state->gwy.addr, state->gwy.port,
715 state->af);
716 printf(" ext: ");
717 pf_print_host(&state->ext.addr, state->ext.port,
718 state->af);
719 if (state->sync_flags & PFSTATE_FROMSYNC)
720 printf(" (from sync)");
721 printf("\n");
722 }
723 return (-1);
724 }
725
726 if (RB_INSERT(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, state)) {
727 if (pf_status.debug >= PF_DEBUG_MISC) {
728 printf("pf: state insert failed: tree_ext_gwy");
729 printf(" lan: ");
730 pf_print_host(&state->lan.addr, state->lan.port,
731 state->af);
732 printf(" gwy: ");
733 pf_print_host(&state->gwy.addr, state->gwy.port,
734 state->af);
735 printf(" ext: ");
736 pf_print_host(&state->ext.addr, state->ext.port,
737 state->af);
738 if (state->sync_flags & PFSTATE_FROMSYNC)
739 printf(" (from sync)");
740 printf("\n");
741 }
742 RB_REMOVE(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state);
743 return (-1);
744 }
745
746 if (state->id == 0 && state->creatorid == 0) {
747 state->id = htobe64(pf_status.stateid++);
748 state->creatorid = pf_status.hostid;
749 }
750 if (RB_INSERT(pf_state_tree_id, &tree_id, state) != NULL) {
751 if (pf_status.debug >= PF_DEBUG_MISC) {
752#ifdef __FreeBSD__
753 printf("pf: state insert failed: "
754 "id: %016llx creatorid: %08x",
755 (long long)be64toh(state->id),
756 ntohl(state->creatorid));
757#else
758 printf("pf: state insert failed: "
759 "id: %016llx creatorid: %08x",
760 betoh64(state->id), ntohl(state->creatorid));
761#endif
762 if (state->sync_flags & PFSTATE_FROMSYNC)
763 printf(" (from sync)");
764 printf("\n");
765 }
766 RB_REMOVE(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state);
767 RB_REMOVE(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, state);
768 return (-1);
769 }
770 TAILQ_INSERT_HEAD(&state_updates, state, u.s.entry_updates);
771
772 pf_status.fcounters[FCNT_STATE_INSERT]++;
773 pf_status.states++;
774 pfi_attach_state(kif);
775#if NPFSYNC
776 pfsync_insert_state(state);
777#endif
778 return (0);
779}
780
781void
782pf_purge_timeout(void *arg)
783{
784#ifdef __FreeBSD__
785 struct callout *to = arg;
786#else
787 struct timeout *to = arg;
788#endif
789 int s;
790
791#ifdef __FreeBSD__
792 PF_LOCK();
793#endif
794 s = splsoftnet();
795 pf_purge_expired_states();
796 pf_purge_expired_fragments();
797 pf_purge_expired_src_nodes();
798 splx(s);
799#ifdef __FreeBSD__
800 PF_UNLOCK();
801#endif
802
803#ifdef __FreeBSD__
804 callout_reset(to, pf_default_rule.timeout[PFTM_INTERVAL] * hz,
805 pf_purge_timeout, to);
806#else
807 timeout_add(to, pf_default_rule.timeout[PFTM_INTERVAL] * hz);
808#endif
809}
810
811u_int32_t
812pf_state_expires(const struct pf_state *state)
813{
814 u_int32_t timeout;
815 u_int32_t start;
816 u_int32_t end;
817 u_int32_t states;
818
819 /* handle all PFTM_* > PFTM_MAX here */
820 if (state->timeout == PFTM_PURGE)
821#ifdef __FreeBSD__
822 return (time_second);
823#else
824 return (time.tv_sec);
825#endif
826 if (state->timeout == PFTM_UNTIL_PACKET)
827 return (0);
828#ifdef __FreeBSD__
829 KASSERT((state->timeout < PFTM_MAX),
830 ("pf_state_expires: timeout > PFTM_MAX"));
831#else
832 KASSERT(state->timeout < PFTM_MAX);
833#endif
834 timeout = state->rule.ptr->timeout[state->timeout];
835 if (!timeout)
836 timeout = pf_default_rule.timeout[state->timeout];
837 start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START];
838 if (start) {
839 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END];
840 states = state->rule.ptr->states;
841 } else {
842 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START];
843 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END];
844 states = pf_status.states;
845 }
846 if (end && states > start && start < end) {
847 if (states < end)
848 return (state->expire + timeout * (end - states) /
849 (end - start));
850 else
851#ifdef __FreeBSD__
852 return (time_second);
853#else
854 return (time.tv_sec);
855#endif
856 }
857 return (state->expire + timeout);
858}
859
860void
861pf_purge_expired_src_nodes(void)
862{
863 struct pf_src_node *cur, *next;
864
865 for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) {
866 next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur);
867
868#ifdef __FreeBSD__
869 if (cur->states <= 0 && cur->expire <= time_second) {
870#else
871 if (cur->states <= 0 && cur->expire <= time.tv_sec) {
872#endif
873 if (cur->rule.ptr != NULL) {
874 cur->rule.ptr->src_nodes--;
875 if (cur->rule.ptr->states <= 0 &&
876 cur->rule.ptr->max_src_nodes <= 0)
877 pf_rm_rule(NULL, cur->rule.ptr);
878 }
879 RB_REMOVE(pf_src_tree, &tree_src_tracking, cur);
880 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
881 pf_status.src_nodes--;
882 pool_put(&pf_src_tree_pl, cur);
883 }
884 }
885}
886
887void
888pf_src_tree_remove_state(struct pf_state *s)
889{
890 u_int32_t timeout;
891
892 if (s->src_node != NULL) {
893 if (--s->src_node->states <= 0) {
894 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
895 if (!timeout)
896 timeout =
897 pf_default_rule.timeout[PFTM_SRC_NODE];
898#ifdef __FreeBSD__
899 s->src_node->expire = time_second + timeout;
900#else
901 s->src_node->expire = time.tv_sec + timeout;
902#endif
903 }
904 }
905 if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) {
906 if (--s->nat_src_node->states <= 0) {
907 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
908 if (!timeout)
909 timeout =
910 pf_default_rule.timeout[PFTM_SRC_NODE];
911#ifdef __FreeBSD__
912 s->nat_src_node->expire = time_second + timeout;
913#else
914 s->nat_src_node->expire = time.tv_sec + timeout;
915#endif
916 }
917 }
918 s->src_node = s->nat_src_node = NULL;
919}
920
921void
922pf_purge_expired_states(void)
923{
924 struct pf_state *cur, *next;
925
926 for (cur = RB_MIN(pf_state_tree_id, &tree_id);
927 cur; cur = next) {
928 next = RB_NEXT(pf_state_tree_id, &tree_id, cur);
929
930#ifdef __FreeBSD__
931 if (pf_state_expires(cur) <= time_second) {
932#else
933 if (pf_state_expires(cur) <= time.tv_sec) {
934#endif
935 if (cur->src.state == PF_TCPS_PROXY_DST)
936 pf_send_tcp(cur->rule.ptr, cur->af,
937 &cur->ext.addr, &cur->lan.addr,
938 cur->ext.port, cur->lan.port,
939 cur->src.seqhi, cur->src.seqlo + 1, 0,
940 TH_RST|TH_ACK, 0, 0);
941 RB_REMOVE(pf_state_tree_ext_gwy,
942 &cur->u.s.kif->pfik_ext_gwy, cur);
943 RB_REMOVE(pf_state_tree_lan_ext,
944 &cur->u.s.kif->pfik_lan_ext, cur);
945 RB_REMOVE(pf_state_tree_id, &tree_id, cur);
946#if NPFSYNC
947 pfsync_delete_state(cur);
948#endif
949 pf_src_tree_remove_state(cur);
950 if (--cur->rule.ptr->states <= 0 &&
951 cur->rule.ptr->src_nodes <= 0)
952 pf_rm_rule(NULL, cur->rule.ptr);
953 if (cur->nat_rule.ptr != NULL)
954 if (--cur->nat_rule.ptr->states <= 0 &&
955 cur->nat_rule.ptr->src_nodes <= 0)
956 pf_rm_rule(NULL, cur->nat_rule.ptr);
957 if (cur->anchor.ptr != NULL)
958 if (--cur->anchor.ptr->states <= 0)
959 pf_rm_rule(NULL, cur->anchor.ptr);
960 pf_normalize_tcp_cleanup(cur);
961 pfi_detach_state(cur->u.s.kif);
962 TAILQ_REMOVE(&state_updates, cur, u.s.entry_updates);
963 pool_put(&pf_state_pl, cur);
964 pf_status.fcounters[FCNT_STATE_REMOVALS]++;
965 pf_status.states--;
966 }
967 }
968}
969
970int
971pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw)
972{
973 if (aw->type != PF_ADDR_TABLE)
974 return (0);
975 if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL)
976 return (1);
977 return (0);
978}
979
980void
981pf_tbladdr_remove(struct pf_addr_wrap *aw)
982{
983 if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL)
984 return;
985 pfr_detach_table(aw->p.tbl);
986 aw->p.tbl = NULL;
987}
988
989void
990pf_tbladdr_copyout(struct pf_addr_wrap *aw)
991{
992 struct pfr_ktable *kt = aw->p.tbl;
993
994 if (aw->type != PF_ADDR_TABLE || kt == NULL)
995 return;
996 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
997 kt = kt->pfrkt_root;
998 aw->p.tbl = NULL;
999 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
1000 kt->pfrkt_cnt : -1;
1001}
1002
1003void
1004pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af)
1005{
1006 switch (af) {
1007#ifdef INET
1008 case AF_INET: {
1009 u_int32_t a = ntohl(addr->addr32[0]);
1010 printf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255,
1011 (a>>8)&255, a&255);
1012 if (p) {
1013 p = ntohs(p);
1014 printf(":%u", p);
1015 }
1016 break;
1017 }
1018#endif /* INET */
1019#ifdef INET6
1020 case AF_INET6: {
1021 u_int16_t b;
1022 u_int8_t i, curstart = 255, curend = 0,
1023 maxstart = 0, maxend = 0;
1024 for (i = 0; i < 8; i++) {
1025 if (!addr->addr16[i]) {
1026 if (curstart == 255)
1027 curstart = i;
1028 else
1029 curend = i;
1030 } else {
1031 if (curstart) {
1032 if ((curend - curstart) >
1033 (maxend - maxstart)) {
1034 maxstart = curstart;
1035 maxend = curend;
1036 curstart = 255;
1037 }
1038 }
1039 }
1040 }
1041 for (i = 0; i < 8; i++) {
1042 if (i >= maxstart && i <= maxend) {
1043 if (maxend != 7) {
1044 if (i == maxstart)
1045 printf(":");
1046 } else {
1047 if (i == maxend)
1048 printf(":");
1049 }
1050 } else {
1051 b = ntohs(addr->addr16[i]);
1052 printf("%x", b);
1053 if (i < 7)
1054 printf(":");
1055 }
1056 }
1057 if (p) {
1058 p = ntohs(p);
1059 printf("[%u]", p);
1060 }
1061 break;
1062 }
1063#endif /* INET6 */
1064 }
1065}
1066
1067void
1068pf_print_state(struct pf_state *s)
1069{
1070 switch (s->proto) {
1071 case IPPROTO_TCP:
1072 printf("TCP ");
1073 break;
1074 case IPPROTO_UDP:
1075 printf("UDP ");
1076 break;
1077 case IPPROTO_ICMP:
1078 printf("ICMP ");
1079 break;
1080 case IPPROTO_ICMPV6:
1081 printf("ICMPV6 ");
1082 break;
1083 default:
1084 printf("%u ", s->proto);
1085 break;
1086 }
1087 pf_print_host(&s->lan.addr, s->lan.port, s->af);
1088 printf(" ");
1089 pf_print_host(&s->gwy.addr, s->gwy.port, s->af);
1090 printf(" ");
1091 pf_print_host(&s->ext.addr, s->ext.port, s->af);
1092 printf(" [lo=%u high=%u win=%u modulator=%u", s->src.seqlo,
1093 s->src.seqhi, s->src.max_win, s->src.seqdiff);
1094 if (s->src.wscale && s->dst.wscale)
1095 printf(" wscale=%u", s->src.wscale & PF_WSCALE_MASK);
1096 printf("]");
1097 printf(" [lo=%u high=%u win=%u modulator=%u", s->dst.seqlo,
1098 s->dst.seqhi, s->dst.max_win, s->dst.seqdiff);
1099 if (s->src.wscale && s->dst.wscale)
1100 printf(" wscale=%u", s->dst.wscale & PF_WSCALE_MASK);
1101 printf("]");
1102 printf(" %u:%u", s->src.state, s->dst.state);
1103}
1104
1105void
1106pf_print_flags(u_int8_t f)
1107{
1108 if (f)
1109 printf(" ");
1110 if (f & TH_FIN)
1111 printf("F");
1112 if (f & TH_SYN)
1113 printf("S");
1114 if (f & TH_RST)
1115 printf("R");
1116 if (f & TH_PUSH)
1117 printf("P");
1118 if (f & TH_ACK)
1119 printf("A");
1120 if (f & TH_URG)
1121 printf("U");
1122 if (f & TH_ECE)
1123 printf("E");
1124 if (f & TH_CWR)
1125 printf("W");
1126}
1127
1128#define PF_SET_SKIP_STEPS(i) \
1129 do { \
1130 while (head[i] != cur) { \
1131 head[i]->skip[i].ptr = cur; \
1132 head[i] = TAILQ_NEXT(head[i], entries); \
1133 } \
1134 } while (0)
1135
1136void
1137pf_calc_skip_steps(struct pf_rulequeue *rules)
1138{
1139 struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT];
1140 int i;
1141
1142 cur = TAILQ_FIRST(rules);
1143 prev = cur;
1144 for (i = 0; i < PF_SKIP_COUNT; ++i)
1145 head[i] = cur;
1146 while (cur != NULL) {
1147
1148 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
1149 PF_SET_SKIP_STEPS(PF_SKIP_IFP);
1150 if (cur->direction != prev->direction)
1151 PF_SET_SKIP_STEPS(PF_SKIP_DIR);
1152 if (cur->af != prev->af)
1153 PF_SET_SKIP_STEPS(PF_SKIP_AF);
1154 if (cur->proto != prev->proto)
1155 PF_SET_SKIP_STEPS(PF_SKIP_PROTO);
1156 if (cur->src.not != prev->src.not ||
1157 pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr))
1158 PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR);
1159 if (cur->src.port[0] != prev->src.port[0] ||
1160 cur->src.port[1] != prev->src.port[1] ||
1161 cur->src.port_op != prev->src.port_op)
1162 PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
1163 if (cur->dst.not != prev->dst.not ||
1164 pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr))
1165 PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR);
1166 if (cur->dst.port[0] != prev->dst.port[0] ||
1167 cur->dst.port[1] != prev->dst.port[1] ||
1168 cur->dst.port_op != prev->dst.port_op)
1169 PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
1170
1171 prev = cur;
1172 cur = TAILQ_NEXT(cur, entries);
1173 }
1174 for (i = 0; i < PF_SKIP_COUNT; ++i)
1175 PF_SET_SKIP_STEPS(i);
1176}
1177
1178int
1179pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2)
1180{
1181 if (aw1->type != aw2->type)
1182 return (1);
1183 switch (aw1->type) {
1184 case PF_ADDR_ADDRMASK:
1185 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0))
1186 return (1);
1187 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0))
1188 return (1);
1189 return (0);
1190 case PF_ADDR_DYNIFTL:
1191 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt);
1192 case PF_ADDR_NOROUTE:
1193 return (0);
1194 case PF_ADDR_TABLE:
1195 return (aw1->p.tbl != aw2->p.tbl);
1196 default:
1197 printf("invalid address type: %d\n", aw1->type);
1198 return (1);
1199 }
1200}
1201
1202void
1203pf_update_anchor_rules()
1204{
1205 struct pf_rule *rule;
1206 int i;
1207
1208 for (i = 0; i < PF_RULESET_MAX; ++i)
1209 TAILQ_FOREACH(rule, pf_main_ruleset.rules[i].active.ptr,
1210 entries)
1211 if (rule->anchorname[0])
1212 rule->anchor = pf_find_anchor(rule->anchorname);
1213 else
1214 rule->anchor = NULL;
1215}
1216
1217u_int16_t
1218pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp)
1219{
1220 u_int32_t l;
1221
1222 if (udp && !cksum)
1223 return (0x0000);
1224 l = cksum + old - new;
1225 l = (l >> 16) + (l & 65535);
1226 l = l & 65535;
1227 if (udp && !l)
1228 return (0xFFFF);
1229 return (l);
1230}
1231
1232void
1233pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc,
1234 struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af)
1235{
1236 struct pf_addr ao;
1237 u_int16_t po = *p;
1238
1239 PF_ACPY(&ao, a, af);
1240 PF_ACPY(a, an, af);
1241
1242 *p = pn;
1243
1244 switch (af) {
1245#ifdef INET
1246 case AF_INET:
1247 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1248 ao.addr16[0], an->addr16[0], 0),
1249 ao.addr16[1], an->addr16[1], 0);
1250 *p = pn;
1251 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1252 ao.addr16[0], an->addr16[0], u),
1253 ao.addr16[1], an->addr16[1], u),
1254 po, pn, u);
1255 break;
1256#endif /* INET */
1257#ifdef INET6
1258 case AF_INET6:
1259 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1260 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1261 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1262 ao.addr16[0], an->addr16[0], u),
1263 ao.addr16[1], an->addr16[1], u),
1264 ao.addr16[2], an->addr16[2], u),
1265 ao.addr16[3], an->addr16[3], u),
1266 ao.addr16[4], an->addr16[4], u),
1267 ao.addr16[5], an->addr16[5], u),
1268 ao.addr16[6], an->addr16[6], u),
1269 ao.addr16[7], an->addr16[7], u),
1270 po, pn, u);
1271 break;
1272#endif /* INET6 */
1273 }
1274}
1275
1276
1277/* Changes a u_int32_t. Uses a void * so there are no align restrictions */
1278void
1279pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u)
1280{
1281 u_int32_t ao;
1282
1283 memcpy(&ao, a, sizeof(ao));
1284 memcpy(a, &an, sizeof(u_int32_t));
1285 *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u),
1286 ao % 65536, an % 65536, u);
1287}
1288
1289#ifdef INET6
1290void
1291pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u)
1292{
1293 struct pf_addr ao;
1294
1295 PF_ACPY(&ao, a, AF_INET6);
1296 PF_ACPY(a, an, AF_INET6);
1297
1298 *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1299 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1300 pf_cksum_fixup(pf_cksum_fixup(*c,
1301 ao.addr16[0], an->addr16[0], u),
1302 ao.addr16[1], an->addr16[1], u),
1303 ao.addr16[2], an->addr16[2], u),
1304 ao.addr16[3], an->addr16[3], u),
1305 ao.addr16[4], an->addr16[4], u),
1306 ao.addr16[5], an->addr16[5], u),
1307 ao.addr16[6], an->addr16[6], u),
1308 ao.addr16[7], an->addr16[7], u);
1309}
1310#endif /* INET6 */
1311
1312void
1313pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa,
1314 struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c,
1315 u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af)
1316{
1317 struct pf_addr oia, ooa;
1318
1319 PF_ACPY(&oia, ia, af);
1320 PF_ACPY(&ooa, oa, af);
1321
1322 /* Change inner protocol port, fix inner protocol checksum. */
1323 if (ip != NULL) {
1324 u_int16_t oip = *ip;
1325 u_int32_t opc = 0; /* make the compiler happy */
1326
1327 if (pc != NULL)
1328 opc = *pc;
1329 *ip = np;
1330 if (pc != NULL)
1331 *pc = pf_cksum_fixup(*pc, oip, *ip, u);
1332 *ic = pf_cksum_fixup(*ic, oip, *ip, 0);
1333 if (pc != NULL)
1334 *ic = pf_cksum_fixup(*ic, opc, *pc, 0);
1335 }
1336 /* Change inner ip address, fix inner ip and icmp checksums. */
1337 PF_ACPY(ia, na, af);
1338 switch (af) {
1339#ifdef INET
1340 case AF_INET: {
1341 u_int32_t oh2c = *h2c;
1342
1343 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c,
1344 oia.addr16[0], ia->addr16[0], 0),
1345 oia.addr16[1], ia->addr16[1], 0);
1346 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1347 oia.addr16[0], ia->addr16[0], 0),
1348 oia.addr16[1], ia->addr16[1], 0);
1349 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0);
1350 break;
1351 }
1352#endif /* INET */
1353#ifdef INET6
1354 case AF_INET6:
1355 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1356 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1357 pf_cksum_fixup(pf_cksum_fixup(*ic,
1358 oia.addr16[0], ia->addr16[0], u),
1359 oia.addr16[1], ia->addr16[1], u),
1360 oia.addr16[2], ia->addr16[2], u),
1361 oia.addr16[3], ia->addr16[3], u),
1362 oia.addr16[4], ia->addr16[4], u),
1363 oia.addr16[5], ia->addr16[5], u),
1364 oia.addr16[6], ia->addr16[6], u),
1365 oia.addr16[7], ia->addr16[7], u);
1366 break;
1367#endif /* INET6 */
1368 }
1369 /* Change outer ip address, fix outer ip or icmpv6 checksum. */
1370 PF_ACPY(oa, na, af);
1371 switch (af) {
1372#ifdef INET
1373 case AF_INET:
1374 *hc = pf_cksum_fixup(pf_cksum_fixup(*hc,
1375 ooa.addr16[0], oa->addr16[0], 0),
1376 ooa.addr16[1], oa->addr16[1], 0);
1377 break;
1378#endif /* INET */
1379#ifdef INET6
1380 case AF_INET6:
1381 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1382 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1383 pf_cksum_fixup(pf_cksum_fixup(*ic,
1384 ooa.addr16[0], oa->addr16[0], u),
1385 ooa.addr16[1], oa->addr16[1], u),
1386 ooa.addr16[2], oa->addr16[2], u),
1387 ooa.addr16[3], oa->addr16[3], u),
1388 ooa.addr16[4], oa->addr16[4], u),
1389 ooa.addr16[5], oa->addr16[5], u),
1390 ooa.addr16[6], oa->addr16[6], u),
1391 ooa.addr16[7], oa->addr16[7], u);
1392 break;
1393#endif /* INET6 */
1394 }
1395}
1396
1397void
1398pf_send_tcp(const struct pf_rule *r, sa_family_t af,
1399 const struct pf_addr *saddr, const struct pf_addr *daddr,
1400 u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack,
1401 u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl)
1402{
1403 struct mbuf *m;
1404#ifdef ALTQ
1405 struct m_tag *mtag;
1406#endif
1407 int len = 0, tlen; /* make the compiler happy */
1408#ifdef INET
1409 struct ip *h = NULL; /* make the compiler happy */
1410#endif /* INET */
1411#ifdef INET6
1412 struct ip6_hdr *h6 = NULL; /* make the compiler happy */
1413#endif /* INET6 */
1414 struct tcphdr *th = NULL; /* make the compiler happy */
1415#ifdef __FreeBSD__
1416 struct ip *ip;
1417#endif
1418 char *opt;
1419
1420 /* maximum segment size tcp option */
1421 tlen = sizeof(struct tcphdr);
1422 if (mss)
1423 tlen += 4;
1424
1425 switch (af) {
1426#ifdef INET
1427 case AF_INET:
1428 len = sizeof(struct ip) + tlen;
1429 break;
1430#endif /* INET */
1431#ifdef INET6
1432 case AF_INET6:
1433 len = sizeof(struct ip6_hdr) + tlen;
1434 break;
1435#endif /* INET6 */
1436 }
1437
1438 /* create outgoing mbuf */
1439#ifdef __FreeBSD__
1440 m = m_gethdr(M_DONTWAIT, MT_HEADER);
1441 if (m == NULL)
1442 return;
1443 m->m_flags |= M_SKIP_FIREWALL;
1444#else
1445 mtag = m_tag_get(PACKET_TAG_PF_GENERATED, 0, M_NOWAIT);
1446 if (mtag == NULL)
1447 return;
1448 m = m_gethdr(M_DONTWAIT, MT_HEADER);
1449 if (m == NULL) {
1450 m_tag_free(mtag);
1451 return;
1452 }
1453 m_tag_prepend(m, mtag);
1454#endif
1455#ifdef ALTQ
1456 if (r != NULL && r->qid) {
1457 struct altq_tag *atag;
1458
1459 mtag = m_tag_get(PACKET_TAG_PF_QID, sizeof(*atag), M_NOWAIT);
1460 if (mtag != NULL) {
1461 atag = (struct altq_tag *)(mtag + 1);
1462 atag->qid = r->qid;
1463 /* add hints for ecn */
1464 atag->af = af;
1465 atag->hdr = mtod(m, struct ip *);
1466 m_tag_prepend(m, mtag);
1467 }
1468 }
1469#endif
1470 m->m_data += max_linkhdr;
1471 m->m_pkthdr.len = m->m_len = len;
1472 m->m_pkthdr.rcvif = NULL;
1473 bzero(m->m_data, len);
1474 switch (af) {
1475#ifdef INET
1476 case AF_INET:
1477 h = mtod(m, struct ip *);
1478
1479 /* IP header fields included in the TCP checksum */
1480 h->ip_p = IPPROTO_TCP;
1481 h->ip_len = htons(tlen);
1482 h->ip_src.s_addr = saddr->v4.s_addr;
1483 h->ip_dst.s_addr = daddr->v4.s_addr;
1484
1485 th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip));
1486 break;
1487#endif /* INET */
1488#ifdef INET6
1489 case AF_INET6:
1490 h6 = mtod(m, struct ip6_hdr *);
1491
1492 /* IP header fields included in the TCP checksum */
1493 h6->ip6_nxt = IPPROTO_TCP;
1494 h6->ip6_plen = htons(tlen);
1495 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr));
1496 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr));
1497
1498 th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr));
1499 break;
1500#endif /* INET6 */
1501 }
1502
1503 /* TCP header */
1504 th->th_sport = sport;
1505 th->th_dport = dport;
1506 th->th_seq = htonl(seq);
1507 th->th_ack = htonl(ack);
1508 th->th_off = tlen >> 2;
1509 th->th_flags = flags;
1510 th->th_win = htons(win);
1511
1512 if (mss) {
1513 opt = (char *)(th + 1);
1514 opt[0] = TCPOPT_MAXSEG;
1515 opt[1] = 4;
1516 HTONS(mss);
1517 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2);
1518 }
1519
1520 switch (af) {
1521#ifdef INET
1522 case AF_INET:
1523 /* TCP checksum */
1524 th->th_sum = in_cksum(m, len);
1525
1526 /* Finish the IP header */
1527 h->ip_v = 4;
1528 h->ip_hl = sizeof(*h) >> 2;
1529 h->ip_tos = IPTOS_LOWDELAY;
1530#ifdef __FreeBSD__
1531 h->ip_off = path_mtu_discovery ? IP_DF : 0;
1532 h->ip_len = len;
1533#else
1534 h->ip_off = htons(ip_mtudisc ? IP_DF : 0);
1535 h->ip_len = htons(len);
1536#endif
1537 h->ip_ttl = ttl ? ttl : ip_defttl;
1538 h->ip_sum = 0;
1539#ifdef __FreeBSD__
1540 ip = mtod(m, struct ip *);
1541 PF_UNLOCK();
1542 ip_output(m, (void *)NULL, (void *)NULL, 0, (void *)NULL,
1543 (void *)NULL);
1544 PF_LOCK();
1545#else /* ! __FreeBSD__ */
1546 ip_output(m, (void *)NULL, (void *)NULL, 0, (void *)NULL,
1547 (void *)NULL);
1548#endif
1549 break;
1550#endif /* INET */
1551#ifdef INET6
1552 case AF_INET6:
1553 /* TCP checksum */
1554 th->th_sum = in6_cksum(m, IPPROTO_TCP,
1555 sizeof(struct ip6_hdr), tlen);
1556
1557 h6->ip6_vfc |= IPV6_VERSION;
1558 h6->ip6_hlim = IPV6_DEFHLIM;
1559
1560#ifdef __FreeBSD__
1561 PF_UNLOCK();
1562 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
1563 PF_LOCK();
1564#else
1565 ip6_output(m, NULL, NULL, 0, NULL, NULL);
1566#endif
1567 break;
1568#endif /* INET6 */
1569 }
1570}
1571
1572void
1573pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af,
1574 struct pf_rule *r)
1575{
1576#ifdef ALTQ
1577 struct m_tag *mtag;
1578#endif
1579 struct mbuf *m0;
1580#ifdef __FreeBSD__
1581 struct ip *ip;
1582#endif
1583
1584#ifdef __FreeBSD__
1585 m0 = m_copypacket(m, M_DONTWAIT);
1586 if (m0 == NULL)
1587 return;
1588 m0->m_flags |= M_SKIP_FIREWALL;
1589#else
1590 mtag = m_tag_get(PACKET_TAG_PF_GENERATED, 0, M_NOWAIT);
1591 if (mtag == NULL)
1592 return;
1593 m0 = m_copy(m, 0, M_COPYALL);
1594 if (m0 == NULL) {
1595 m_tag_free(mtag);
1596 return;
1597 }
1598 m_tag_prepend(m0, mtag);
1599#endif
1600
1601#ifdef ALTQ
1602 if (r->qid) {
1603 struct altq_tag *atag;
1604
1605 mtag = m_tag_get(PACKET_TAG_PF_QID, sizeof(*atag), M_NOWAIT);
1606 if (mtag != NULL) {
1607 atag = (struct altq_tag *)(mtag + 1);
1608 atag->qid = r->qid;
1609 /* add hints for ecn */
1610 atag->af = af;
1611 atag->hdr = mtod(m0, struct ip *);
1612 m_tag_prepend(m0, mtag);
1613 }
1614 }
1615#endif
1616
1617 switch (af) {
1618#ifdef INET
1619 case AF_INET:
1620#ifdef __FreeBSD__
1621 /* icmp_error() expects host byte ordering */
1622 ip = mtod(m0, struct ip *);
1623 NTOHS(ip->ip_len);
1624 NTOHS(ip->ip_off);
1625 PF_UNLOCK();
1626#endif
1627 icmp_error(m0, type, code, 0, (void *)NULL);
1628#ifdef __FreeBSD__
1629 PF_LOCK();
1630#endif
1631 break;
1632#endif /* INET */
1633#ifdef INET6
1634 case AF_INET6:
1635#ifdef __FreeBSD__
1636 PF_UNLOCK();
1637#endif
1638 icmp6_error(m0, type, code, 0);
1639#ifdef __FreeBSD__
1640 PF_LOCK();
1641#endif
1642 break;
1643#endif /* INET6 */
1644 }
1645}
1646
1647/*
1648 * Return 1 if the addresses a and b match (with mask m), otherwise return 0.
1649 * If n is 0, they match if they are equal. If n is != 0, they match if they
1650 * are different.
1651 */
1652int
1653pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m,
1654 struct pf_addr *b, sa_family_t af)
1655{
1656 int match = 0;
1657
1658 switch (af) {
1659#ifdef INET
1660 case AF_INET:
1661 if ((a->addr32[0] & m->addr32[0]) ==
1662 (b->addr32[0] & m->addr32[0]))
1663 match++;
1664 break;
1665#endif /* INET */
1666#ifdef INET6
1667 case AF_INET6:
1668 if (((a->addr32[0] & m->addr32[0]) ==
1669 (b->addr32[0] & m->addr32[0])) &&
1670 ((a->addr32[1] & m->addr32[1]) ==
1671 (b->addr32[1] & m->addr32[1])) &&
1672 ((a->addr32[2] & m->addr32[2]) ==
1673 (b->addr32[2] & m->addr32[2])) &&
1674 ((a->addr32[3] & m->addr32[3]) ==
1675 (b->addr32[3] & m->addr32[3])))
1676 match++;
1677 break;
1678#endif /* INET6 */
1679 }
1680 if (match) {
1681 if (n)
1682 return (0);
1683 else
1684 return (1);
1685 } else {
1686 if (n)
1687 return (1);
1688 else
1689 return (0);
1690 }
1691}
1692
1693int
1694pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p)
1695{
1696 switch (op) {
1697 case PF_OP_IRG:
1698 return ((p > a1) && (p < a2));
1699 case PF_OP_XRG:
1700 return ((p < a1) || (p > a2));
1701 case PF_OP_RRG:
1702 return ((p >= a1) && (p <= a2));
1703 case PF_OP_EQ:
1704 return (p == a1);
1705 case PF_OP_NE:
1706 return (p != a1);
1707 case PF_OP_LT:
1708 return (p < a1);
1709 case PF_OP_LE:
1710 return (p <= a1);
1711 case PF_OP_GT:
1712 return (p > a1);
1713 case PF_OP_GE:
1714 return (p >= a1);
1715 }
1716 return (0); /* never reached */
1717}
1718
1719int
1720pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p)
1721{
1722 NTOHS(a1);
1723 NTOHS(a2);
1724 NTOHS(p);
1725 return (pf_match(op, a1, a2, p));
1726}
1727
1728int
1729pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u)
1730{
1731 if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
1732 return (0);
1733 return (pf_match(op, a1, a2, u));
1734}
1735
1736int
1737pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g)
1738{
1739 if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
1740 return (0);
1741 return (pf_match(op, a1, a2, g));
1742}
1743
1744struct pf_tag *
1745pf_get_tag(struct mbuf *m)
1746{
1747 struct m_tag *mtag;
1748
1749 if ((mtag = m_tag_find(m, PACKET_TAG_PF_TAG, NULL)) != NULL)
1750 return ((struct pf_tag *)(mtag + 1));
1751 else
1752 return (NULL);
1753}
1754
1755int
1756pf_match_tag(struct mbuf *m, struct pf_rule *r, struct pf_rule *nat_rule,
1757 struct pf_tag *pftag, int *tag)
1758{
1759 if (*tag == -1) { /* find mbuf tag */
1760 pftag = pf_get_tag(m);
1761 if (pftag != NULL)
1762 *tag = pftag->tag;
1763 else
1764 *tag = 0;
1765 if (nat_rule != NULL && nat_rule->tag)
1766 *tag = nat_rule->tag;
1767 }
1768
1769 return ((!r->match_tag_not && r->match_tag == *tag) ||
1770 (r->match_tag_not && r->match_tag != *tag));
1771}
1772
1773int
1774pf_tag_packet(struct mbuf *m, struct pf_tag *pftag, int tag)
1775{
1776 struct m_tag *mtag;
1777
1778 if (tag <= 0)
1779 return (0);
1780
1781 if (pftag == NULL) {
1782 mtag = m_tag_get(PACKET_TAG_PF_TAG, sizeof(*pftag), M_NOWAIT);
1783 if (mtag == NULL)
1784 return (1);
1785 ((struct pf_tag *)(mtag + 1))->tag = tag;
1786 m_tag_prepend(m, mtag);
1787 } else
1788 pftag->tag = tag;
1789
1790 return (0);
1791}
1792
1793#define PF_STEP_INTO_ANCHOR(r, a, s, n) \
1794 do { \
1795 if ((r) == NULL || (r)->anchor == NULL || \
1796 (s) != NULL || (a) != NULL) \
1797 panic("PF_STEP_INTO_ANCHOR"); \
1798 (a) = (r); \
1799 (s) = TAILQ_FIRST(&(r)->anchor->rulesets); \
1800 (r) = NULL; \
1801 while ((s) != NULL && ((r) = \
1802 TAILQ_FIRST((s)->rules[n].active.ptr)) == NULL) \
1803 (s) = TAILQ_NEXT((s), entries); \
1804 if ((r) == NULL) { \
1805 (r) = TAILQ_NEXT((a), entries); \
1806 (a) = NULL; \
1807 } \
1808 } while (0)
1809
1810#define PF_STEP_OUT_OF_ANCHOR(r, a, s, n) \
1811 do { \
1812 if ((r) != NULL || (a) == NULL || (s) == NULL) \
1813 panic("PF_STEP_OUT_OF_ANCHOR"); \
1814 (s) = TAILQ_NEXT((s), entries); \
1815 while ((s) != NULL && ((r) = \
1816 TAILQ_FIRST((s)->rules[n].active.ptr)) == NULL) \
1817 (s) = TAILQ_NEXT((s), entries); \
1818 if ((r) == NULL) { \
1819 (r) = TAILQ_NEXT((a), entries); \
1820 (a) = NULL; \
1821 } \
1822 } while (0)
1823
1824#ifdef INET6
1825void
1826pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr,
1827 struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af)
1828{
1829 switch (af) {
1830#ifdef INET
1831 case AF_INET:
1832 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
1833 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
1834 break;
1835#endif /* INET */
1836 case AF_INET6:
1837 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
1838 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
1839 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) |
1840 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]);
1841 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) |
1842 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]);
1843 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) |
1844 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]);
1845 break;
1846 }
1847}
1848
1849void
1850pf_addr_inc(struct pf_addr *addr, sa_family_t af)
1851{
1852 switch (af) {
1853#ifdef INET
1854 case AF_INET:
1855 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1);
1856 break;
1857#endif /* INET */
1858 case AF_INET6:
1859 if (addr->addr32[3] == 0xffffffff) {
1860 addr->addr32[3] = 0;
1861 if (addr->addr32[2] == 0xffffffff) {
1862 addr->addr32[2] = 0;
1863 if (addr->addr32[1] == 0xffffffff) {
1864 addr->addr32[1] = 0;
1865 addr->addr32[0] =
1866 htonl(ntohl(addr->addr32[0]) + 1);
1867 } else
1868 addr->addr32[1] =
1869 htonl(ntohl(addr->addr32[1]) + 1);
1870 } else
1871 addr->addr32[2] =
1872 htonl(ntohl(addr->addr32[2]) + 1);
1873 } else
1874 addr->addr32[3] =
1875 htonl(ntohl(addr->addr32[3]) + 1);
1876 break;
1877 }
1878}
1879#endif /* INET6 */
1880
1881#define mix(a,b,c) \
1882 do { \
1883 a -= b; a -= c; a ^= (c >> 13); \
1884 b -= c; b -= a; b ^= (a << 8); \
1885 c -= a; c -= b; c ^= (b >> 13); \
1886 a -= b; a -= c; a ^= (c >> 12); \
1887 b -= c; b -= a; b ^= (a << 16); \
1888 c -= a; c -= b; c ^= (b >> 5); \
1889 a -= b; a -= c; a ^= (c >> 3); \
1890 b -= c; b -= a; b ^= (a << 10); \
1891 c -= a; c -= b; c ^= (b >> 15); \
1892 } while (0)
1893
1894/*
1895 * hash function based on bridge_hash in if_bridge.c
1896 */
1897void
1898pf_hash(struct pf_addr *inaddr, struct pf_addr *hash,
1899 struct pf_poolhashkey *key, sa_family_t af)
1900{
1901 u_int32_t a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0];
1902
1903 switch (af) {
1904#ifdef INET
1905 case AF_INET:
1906 a += inaddr->addr32[0];
1907 b += key->key32[1];
1908 mix(a, b, c);
1909 hash->addr32[0] = c + key->key32[2];
1910 break;
1911#endif /* INET */
1912#ifdef INET6
1913 case AF_INET6:
1914 a += inaddr->addr32[0];
1915 b += inaddr->addr32[2];
1916 mix(a, b, c);
1917 hash->addr32[0] = c;
1918 a += inaddr->addr32[1];
1919 b += inaddr->addr32[3];
1920 c += key->key32[1];
1921 mix(a, b, c);
1922 hash->addr32[1] = c;
1923 a += inaddr->addr32[2];
1924 b += inaddr->addr32[1];
1925 c += key->key32[2];
1926 mix(a, b, c);
1927 hash->addr32[2] = c;
1928 a += inaddr->addr32[3];
1929 b += inaddr->addr32[0];
1930 c += key->key32[3];
1931 mix(a, b, c);
1932 hash->addr32[3] = c;
1933 break;
1934#endif /* INET6 */
1935 }
1936}
1937
1938int
1939pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr,
1940 struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn)
1941{
1942 unsigned char hash[16];
1943 struct pf_pool *rpool = &r->rpool;
1944 struct pf_addr *raddr = &rpool->cur->addr.v.a.addr;
1945 struct pf_addr *rmask = &rpool->cur->addr.v.a.mask;
1946 struct pf_pooladdr *acur = rpool->cur;
1947 struct pf_src_node k;
1948
1949 if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR &&
1950 (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
1951 k.af = af;
1952 PF_ACPY(&k.addr, saddr, af);
1953 if (r->rule_flag & PFRULE_RULESRCTRACK ||
1954 r->rpool.opts & PF_POOL_STICKYADDR)
1955 k.rule.ptr = r;
1956 else
1957 k.rule.ptr = NULL;
1958 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
1959 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
1960 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) {
1961 PF_ACPY(naddr, &(*sn)->raddr, af);
1962 if (pf_status.debug >= PF_DEBUG_MISC) {
1963 printf("pf_map_addr: src tracking maps ");
1964 pf_print_host(&k.addr, 0, af);
1965 printf(" to ");
1966 pf_print_host(naddr, 0, af);
1967 printf("\n");
1968 }
1969 return (0);
1970 }
1971 }
1972
1973 if (rpool->cur->addr.type == PF_ADDR_NOROUTE)
1974 return (1);
1975 if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
1976 if (af == AF_INET) {
1977 if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 &&
1978 (rpool->opts & PF_POOL_TYPEMASK) !=
1979 PF_POOL_ROUNDROBIN)
1980 return (1);
1981 raddr = &rpool->cur->addr.p.dyn->pfid_addr4;
1982 rmask = &rpool->cur->addr.p.dyn->pfid_mask4;
1983 } else {
1984 if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 &&
1985 (rpool->opts & PF_POOL_TYPEMASK) !=
1986 PF_POOL_ROUNDROBIN)
1987 return (1);
1988 raddr = &rpool->cur->addr.p.dyn->pfid_addr6;
1989 rmask = &rpool->cur->addr.p.dyn->pfid_mask6;
1990 }
1991 } else if (rpool->cur->addr.type == PF_ADDR_TABLE) {
1992 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN)
1993 return (1); /* unsupported */
1994 } else {
1995 raddr = &rpool->cur->addr.v.a.addr;
1996 rmask = &rpool->cur->addr.v.a.mask;
1997 }
1998
1999 switch (rpool->opts & PF_POOL_TYPEMASK) {
2000 case PF_POOL_NONE:
2001 PF_ACPY(naddr, raddr, af);
2002 break;
2003 case PF_POOL_BITMASK:
2004 PF_POOLMASK(naddr, raddr, rmask, saddr, af);
2005 break;
2006 case PF_POOL_RANDOM:
2007 if (init_addr != NULL && PF_AZERO(init_addr, af)) {
2008 switch (af) {
2009#ifdef INET
2010 case AF_INET:
2011 rpool->counter.addr32[0] = arc4random();
2012 break;
2013#endif /* INET */
2014#ifdef INET6
2015 case AF_INET6:
2016 if (rmask->addr32[3] != 0xffffffff)
2017 rpool->counter.addr32[3] = arc4random();
2018 else
2019 break;
2020 if (rmask->addr32[2] != 0xffffffff)
2021 rpool->counter.addr32[2] = arc4random();
2022 else
2023 break;
2024 if (rmask->addr32[1] != 0xffffffff)
2025 rpool->counter.addr32[1] = arc4random();
2026 else
2027 break;
2028 if (rmask->addr32[0] != 0xffffffff)
2029 rpool->counter.addr32[0] = arc4random();
2030 break;
2031#endif /* INET6 */
2032 }
2033 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2034 PF_ACPY(init_addr, naddr, af);
2035
2036 } else {
2037 PF_AINC(&rpool->counter, af);
2038 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2039 }
2040 break;
2041 case PF_POOL_SRCHASH:
2042 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af);
2043 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af);
2044 break;
2045 case PF_POOL_ROUNDROBIN:
2046 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2047 if (!pfr_pool_get(rpool->cur->addr.p.tbl,
2048 &rpool->tblidx, &rpool->counter,
2049 &raddr, &rmask, af))
2050 goto get_addr;
2051 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2052 if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2053 &rpool->tblidx, &rpool->counter,
2054 &raddr, &rmask, af))
2055 goto get_addr;
2056 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af))
2057 goto get_addr;
2058
2059 try_next:
2060 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL)
2061 rpool->cur = TAILQ_FIRST(&rpool->list);
2062 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2063 rpool->tblidx = -1;
2064 if (pfr_pool_get(rpool->cur->addr.p.tbl,
2065 &rpool->tblidx, &rpool->counter,
2066 &raddr, &rmask, af)) {
2067 /* table contains no address of type 'af' */
2068 if (rpool->cur != acur)
2069 goto try_next;
2070 return (1);
2071 }
2072 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2073 rpool->tblidx = -1;
2074 if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2075 &rpool->tblidx, &rpool->counter,
2076 &raddr, &rmask, af)) {
2077 /* table contains no address of type 'af' */
2078 if (rpool->cur != acur)
2079 goto try_next;
2080 return (1);
2081 }
2082 } else {
2083 raddr = &rpool->cur->addr.v.a.addr;
2084 rmask = &rpool->cur->addr.v.a.mask;
2085 PF_ACPY(&rpool->counter, raddr, af);
2086 }
2087
2088 get_addr:
2089 PF_ACPY(naddr, &rpool->counter, af);
2090 PF_AINC(&rpool->counter, af);
2091 break;
2092 }
2093 if (*sn != NULL)
2094 PF_ACPY(&(*sn)->raddr, naddr, af);
2095
2096 if (pf_status.debug >= PF_DEBUG_MISC &&
2097 (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2098 printf("pf_map_addr: selected address ");
2099 pf_print_host(naddr, 0, af);
2100 printf("\n");
2101 }
2102
2103 return (0);
2104}
2105
2106int
2107pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r,
2108 struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport,
2109 struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high,
2110 struct pf_src_node **sn)
2111{
2112 struct pf_state key;
2113 struct pf_addr init_addr;
2114 u_int16_t cut;
2115
2116 bzero(&init_addr, sizeof(init_addr));
2117 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2118 return (1);
2119
2120 do {
2121 key.af = af;
2122 key.proto = proto;
2123 PF_ACPY(&key.ext.addr, daddr, key.af);
2124 PF_ACPY(&key.gwy.addr, naddr, key.af);
2125 key.ext.port = dport;
2126
2127 /*
2128 * port search; start random, step;
2129 * similar 2 portloop in in_pcbbind
2130 */
2131 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP)) {
2132 key.gwy.port = 0;
2133 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
2134 return (0);
2135 } else if (low == 0 && high == 0) {
2136 key.gwy.port = *nport;
2137 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
2138 return (0);
2139 } else if (low == high) {
2140 key.gwy.port = htons(low);
2141 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) {
2142 *nport = htons(low);
2143 return (0);
2144 }
2145 } else {
2146 u_int16_t tmp;
2147
2148 if (low > high) {
2149 tmp = low;
2150 low = high;
2151 high = tmp;
2152 }
2153 /* low < high */
2154 cut = arc4random() % (1 + high - low) + low;
2155 /* low <= cut <= high */
2156 for (tmp = cut; tmp <= high; ++(tmp)) {
2157 key.gwy.port = htons(tmp);
2158 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
2159 NULL) {
2160 *nport = htons(tmp);
2161 return (0);
2162 }
2163 }
2164 for (tmp = cut - 1; tmp >= low; --(tmp)) {
2165 key.gwy.port = htons(tmp);
2166 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
2167 NULL) {
2168 *nport = htons(tmp);
2169 return (0);
2170 }
2171 }
2172 }
2173
2174 switch (r->rpool.opts & PF_POOL_TYPEMASK) {
2175 case PF_POOL_RANDOM:
2176 case PF_POOL_ROUNDROBIN:
2177 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2178 return (1);
2179 break;
2180 case PF_POOL_NONE:
2181 case PF_POOL_SRCHASH:
2182 case PF_POOL_BITMASK:
2183 default:
2184 return (1);
2185 }
2186 } while (! PF_AEQ(&init_addr, naddr, af) );
2187
2188 return (1); /* none available */
2189}
2190
2191struct pf_rule *
2192pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off,
2193 int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport,
2194 struct pf_addr *daddr, u_int16_t dport, int rs_num)
2195{
2196 struct pf_rule *r, *rm = NULL, *anchorrule = NULL;
2197 struct pf_ruleset *ruleset = NULL;
2198
2199 r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr);
2200 while (r && rm == NULL) {
2201 struct pf_rule_addr *src = NULL, *dst = NULL;
2202 struct pf_addr_wrap *xdst = NULL;
2203
2204 if (r->action == PF_BINAT && direction == PF_IN) {
2205 src = &r->dst;
2206 if (r->rpool.cur != NULL)
2207 xdst = &r->rpool.cur->addr;
2208 } else {
2209 src = &r->src;
2210 dst = &r->dst;
2211 }
2212
2213 r->evaluations++;
2214 if (r->kif != NULL &&
2215 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
2216 r = r->skip[PF_SKIP_IFP].ptr;
2217 else if (r->direction && r->direction != direction)
2218 r = r->skip[PF_SKIP_DIR].ptr;
2219 else if (r->af && r->af != pd->af)
2220 r = r->skip[PF_SKIP_AF].ptr;
2221 else if (r->proto && r->proto != pd->proto)
2222 r = r->skip[PF_SKIP_PROTO].ptr;
2223 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af, src->not))
2224 r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR :
2225 PF_SKIP_DST_ADDR].ptr;
2226 else if (src->port_op && !pf_match_port(src->port_op,
2227 src->port[0], src->port[1], sport))
2228 r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT :
2229 PF_SKIP_DST_PORT].ptr;
2230 else if (dst != NULL &&
2231 PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->not))
2232 r = r->skip[PF_SKIP_DST_ADDR].ptr;
2233 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af, 0))
2234 r = TAILQ_NEXT(r, entries);
2235 else if (dst != NULL && dst->port_op &&
2236 !pf_match_port(dst->port_op, dst->port[0],
2237 dst->port[1], dport))
2238 r = r->skip[PF_SKIP_DST_PORT].ptr;
2239 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto !=
2240 IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m,
2241 off, pd->hdr.tcp), r->os_fingerprint)))
2242 r = TAILQ_NEXT(r, entries);
2243 else if (r->anchorname[0] && r->anchor == NULL)
2244 r = TAILQ_NEXT(r, entries);
2245 else if (r->anchor == NULL)
2246 rm = r;
2247 else
2248 PF_STEP_INTO_ANCHOR(r, anchorrule, ruleset, rs_num);
2249 if (r == NULL && anchorrule != NULL)
2250 PF_STEP_OUT_OF_ANCHOR(r, anchorrule, ruleset,
2251 rs_num);
2252 }
2253 if (rm != NULL && (rm->action == PF_NONAT ||
2254 rm->action == PF_NORDR || rm->action == PF_NOBINAT))
2255 return (NULL);
2256 return (rm);
2257}
2258
2259struct pf_rule *
2260pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction,
2261 struct pfi_kif *kif, struct pf_src_node **sn,
2262 struct pf_addr *saddr, u_int16_t sport,
2263 struct pf_addr *daddr, u_int16_t dport,
2264 struct pf_addr *naddr, u_int16_t *nport)
2265{
2266 struct pf_rule *r = NULL;
2267
2268 if (direction == PF_OUT) {
2269 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2270 sport, daddr, dport, PF_RULESET_BINAT);
2271 if (r == NULL)
2272 r = pf_match_translation(pd, m, off, direction, kif,
2273 saddr, sport, daddr, dport, PF_RULESET_NAT);
2274 } else {
2275 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2276 sport, daddr, dport, PF_RULESET_RDR);
2277 if (r == NULL)
2278 r = pf_match_translation(pd, m, off, direction, kif,
2279 saddr, sport, daddr, dport, PF_RULESET_BINAT);
2280 }
2281
2282 if (r != NULL) {
2283 switch (r->action) {
2284 case PF_NONAT:
2285 case PF_NOBINAT:
2286 case PF_NORDR:
2287 return (NULL);
2288 case PF_NAT:
2289 if (pf_get_sport(pd->af, pd->proto, r, saddr,
2290 daddr, dport, naddr, nport, r->rpool.proxy_port[0],
2291 r->rpool.proxy_port[1], sn)) {
2292 DPFPRINTF(PF_DEBUG_MISC,
2293 ("pf: NAT proxy port allocation "
2294 "(%u-%u) failed\n",
2295 r->rpool.proxy_port[0],
2296 r->rpool.proxy_port[1]));
2297 return (NULL);
2298 }
2299 break;
2300 case PF_BINAT:
2301 switch (direction) {
2302 case PF_OUT:
2303 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){
2304 if (pd->af == AF_INET) {
2305 if (r->rpool.cur->addr.p.dyn->
2306 pfid_acnt4 < 1)
2307 return (NULL);
2308 PF_POOLMASK(naddr,
2309 &r->rpool.cur->addr.p.dyn->
2310 pfid_addr4,
2311 &r->rpool.cur->addr.p.dyn->
2312 pfid_mask4,
2313 saddr, AF_INET);
2314 } else {
2315 if (r->rpool.cur->addr.p.dyn->
2316 pfid_acnt6 < 1)
2317 return (NULL);
2318 PF_POOLMASK(naddr,
2319 &r->rpool.cur->addr.p.dyn->
2320 pfid_addr6,
2321 &r->rpool.cur->addr.p.dyn->
2322 pfid_mask6,
2323 saddr, AF_INET6);
2324 }
2325 } else
2326 PF_POOLMASK(naddr,
2327 &r->rpool.cur->addr.v.a.addr,
2328 &r->rpool.cur->addr.v.a.mask,
2329 saddr, pd->af);
2330 break;
2331 case PF_IN:
2332 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){
2333 if (pd->af == AF_INET) {
2334 if (r->src.addr.p.dyn->
2335 pfid_acnt4 < 1)
2336 return (NULL);
2337 PF_POOLMASK(naddr,
2338 &r->src.addr.p.dyn->
2339 pfid_addr4,
2340 &r->src.addr.p.dyn->
2341 pfid_mask4,
2342 daddr, AF_INET);
2343 } else {
2344 if (r->src.addr.p.dyn->
2345 pfid_acnt6 < 1)
2346 return (NULL);
2347 PF_POOLMASK(naddr,
2348 &r->src.addr.p.dyn->
2349 pfid_addr6,
2350 &r->src.addr.p.dyn->
2351 pfid_mask6,
2352 daddr, AF_INET6);
2353 }
2354 } else
2355 PF_POOLMASK(naddr,
2356 &r->src.addr.v.a.addr,
2357 &r->src.addr.v.a.mask, daddr,
2358 pd->af);
2359 break;
2360 }
2361 break;
2362 case PF_RDR: {
2363 if (pf_map_addr(r->af, r, saddr, naddr, NULL, sn))
2364 return (NULL);
2365
2366 if (r->rpool.proxy_port[1]) {
2367 u_int32_t tmp_nport;
2368
2369 tmp_nport = ((ntohs(dport) -
2370 ntohs(r->dst.port[0])) %
2371 (r->rpool.proxy_port[1] -
2372 r->rpool.proxy_port[0] + 1)) +
2373 r->rpool.proxy_port[0];
2374
2375 /* wrap around if necessary */
2376 if (tmp_nport > 65535)
2377 tmp_nport -= 65535;
2378 *nport = htons((u_int16_t)tmp_nport);
2379 } else if (r->rpool.proxy_port[0])
2380 *nport = htons(r->rpool.proxy_port[0]);
2381 break;
2382 }
2383 default:
2384 return (NULL);
2385 }
2386 }
2387
2388 return (r);
2389}
2390
2391int
2392#ifdef __FreeBSD__
2393pf_socket_lookup(uid_t *uid, gid_t *gid, int direction, struct pf_pdesc *pd,
2394 struct inpcb *inp_arg)
2395#else
2396pf_socket_lookup(uid_t *uid, gid_t *gid, int direction, struct pf_pdesc *pd)
2397#endif
2398{
2399 struct pf_addr *saddr, *daddr;
2400 u_int16_t sport, dport;
2401#ifdef __FreeBSD__
2402 struct inpcbinfo *pi;
2403#else
2404 struct inpcbtable *tb;
2405#endif
2406 struct inpcb *inp;
2407
2408 *uid = UID_MAX;
2409 *gid = GID_MAX;
2410#ifdef __FreeBSD__
2411 if (inp_arg != NULL) {
2412 INP_LOCK_ASSERT(inp_arg);
2413 if (inp_arg->inp_socket) {
2414 *uid = inp_arg->inp_socket->so_cred->cr_uid;
2415 *gid = inp_arg->inp_socket->so_cred->cr_groups[0];
2416 return (1);
2417 } else
2418 return (0);
2419 }
2420#endif
2421 switch (pd->proto) {
2422 case IPPROTO_TCP:
2423 sport = pd->hdr.tcp->th_sport;
2424 dport = pd->hdr.tcp->th_dport;
2425#ifdef __FreeBSD__
2426 pi = &tcbinfo;
2427#else
2428 tb = &tcbtable;
2429#endif
2430 break;
2431 case IPPROTO_UDP:
2432 sport = pd->hdr.udp->uh_sport;
2433 dport = pd->hdr.udp->uh_dport;
2434#ifdef __FreeBSD__
2435 pi = &udbinfo;
2436#else
2437 tb = &udbtable;
2438#endif
2439 break;
2440 default:
2441 return (0);
2442 }
2443 if (direction == PF_IN) {
2444 saddr = pd->src;
2445 daddr = pd->dst;
2446 } else {
2447 u_int16_t p;
2448
2449 p = sport;
2450 sport = dport;
2451 dport = p;
2452 saddr = pd->dst;
2453 daddr = pd->src;
2454 }
2455 switch (pd->af) {
2456 case AF_INET:
2457#ifdef __FreeBSD__
2458 INP_INFO_RLOCK(pi); /* XXX LOR */
2459 inp = in_pcblookup_hash(pi, saddr->v4, sport, daddr->v4,
2460 dport, 0, NULL);
2461 if (inp == NULL) {
2462 inp = in_pcblookup_hash(pi, saddr->v4, sport,
2463 daddr->v4, dport, INPLOOKUP_WILDCARD, NULL);
2464 if(inp == NULL) {
2465 INP_INFO_RUNLOCK(pi);
2466 return (0);
2467 }
2468 }
2469#else
2470 inp = in_pcbhashlookup(tb, saddr->v4, sport, daddr->v4, dport);
2471 if (inp == NULL) {
2472 inp = in_pcblookup_listen(tb, daddr->v4, dport, 0);
2473 if (inp == NULL)
2474 return (0);
2475 }
2476#endif
2477 break;
2478#ifdef INET6
2479 case AF_INET6:
2480#ifdef __FreeBSD__
2481 INP_INFO_RLOCK(pi);
2482 inp = in6_pcblookup_hash(pi, &saddr->v6, sport,
2483 &daddr->v6, dport, 0, NULL);
2484 if (inp == NULL) {
2485 inp = in6_pcblookup_hash(pi, &saddr->v6, sport,
2486 &daddr->v6, dport, INPLOOKUP_WILDCARD, NULL);
2487 if (inp == NULL) {
2488 INP_INFO_RUNLOCK(pi);
2489 return (0);
2490 }
2491 }
2492#else
2493 inp = in6_pcbhashlookup(tb, &saddr->v6, sport, &daddr->v6,
2494 dport);
2495 if (inp == NULL) {
2496 inp = in6_pcblookup_listen(tb, &daddr->v6, dport, 0);
2497 if (inp == NULL)
2498 return (0);
2499 }
2500#endif
2501 break;
2502#endif /* INET6 */
2503
2504 default:
2505 return (0);
2506 }
2507#ifdef __FreeBSD__
2508 INP_LOCK(inp);
2509 if ((inp->inp_socket == NULL) || (inp->inp_socket->so_cred == NULL)) {
2510 INP_UNLOCK(inp);
2511 INP_INFO_RUNLOCK(pi);
2512 return (0);
2513 }
2514 *uid = inp->inp_socket->so_cred->cr_uid;
2515 *gid = inp->inp_socket->so_cred->cr_groups[0];
2516 INP_UNLOCK(inp);
2517 INP_INFO_RUNLOCK(pi);
2518#else
2519 *uid = inp->inp_socket->so_euid;
2520 *gid = inp->inp_socket->so_egid;
2521#endif
2522 return (1);
2523}
2524
2525u_int8_t
2526pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2527{
2528 int hlen;
2529 u_int8_t hdr[60];
2530 u_int8_t *opt, optlen;
2531 u_int8_t wscale = 0;
2532
2533 hlen = th_off << 2; /* hlen <= sizeof(hdr) */
2534 if (hlen <= sizeof(struct tcphdr))
2535 return (0);
2536 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2537 return (0);
2538 opt = hdr + sizeof(struct tcphdr);
2539 hlen -= sizeof(struct tcphdr);
2540 while (hlen >= 3) {
2541 switch (*opt) {
2542 case TCPOPT_EOL:
2543 case TCPOPT_NOP:
2544 ++opt;
2545 --hlen;
2546 break;
2547 case TCPOPT_WINDOW:
2548 wscale = opt[2];
2549 if (wscale > TCP_MAX_WINSHIFT)
2550 wscale = TCP_MAX_WINSHIFT;
2551 wscale |= PF_WSCALE_FLAG;
2552 /* FALLTHROUGH */
2553 default:
2554 optlen = opt[1];
2555 if (optlen < 2)
2556 optlen = 2;
2557 hlen -= optlen;
2558 opt += optlen;
2559 break;
2560 }
2561 }
2562 return (wscale);
2563}
2564
2565u_int16_t
2566pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2567{
2568 int hlen;
2569 u_int8_t hdr[60];
2570 u_int8_t *opt, optlen;
2571 u_int16_t mss = tcp_mssdflt;
2572
2573 hlen = th_off << 2; /* hlen <= sizeof(hdr) */
2574 if (hlen <= sizeof(struct tcphdr))
2575 return (0);
2576 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2577 return (0);
2578 opt = hdr + sizeof(struct tcphdr);
2579 hlen -= sizeof(struct tcphdr);
2580 while (hlen >= TCPOLEN_MAXSEG) {
2581 switch (*opt) {
2582 case TCPOPT_EOL:
2583 case TCPOPT_NOP:
2584 ++opt;
2585 --hlen;
2586 break;
2587 case TCPOPT_MAXSEG:
2588 bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2);
2589 /* FALLTHROUGH */
2590 default:
2591 optlen = opt[1];
2592 if (optlen < 2)
2593 optlen = 2;
2594 hlen -= optlen;
2595 opt += optlen;
2596 break;
2597 }
2598 }
2599 return (mss);
2600}
2601
2602u_int16_t
2603pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer)
2604{
2605#ifdef INET
2606 struct sockaddr_in *dst;
2607 struct route ro;
2608#endif /* INET */
2609#ifdef INET6
2610 struct sockaddr_in6 *dst6;
2611 struct route_in6 ro6;
2612#endif /* INET6 */
2613 struct rtentry *rt = NULL;
2614 int hlen = 0; /* make the compiler happy */
2615 u_int16_t mss = tcp_mssdflt;
2616
2617 switch (af) {
2618#ifdef INET
2619 case AF_INET:
2620 hlen = sizeof(struct ip);
2621 bzero(&ro, sizeof(ro));
2622 dst = (struct sockaddr_in *)&ro.ro_dst;
2623 dst->sin_family = AF_INET;
2624 dst->sin_len = sizeof(*dst);
2625 dst->sin_addr = addr->v4;
2626#ifdef __FreeBSD__
2627#ifdef RTF_PRCLONING
2628 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
2629#else /* !RTF_PRCLONING */
2630 rtalloc_ign(&ro, RTF_CLONING);
2631#endif
2632#else /* ! __FreeBSD__ */
2633 rtalloc_noclone(&ro, NO_CLONING);
2634#endif
2635 rt = ro.ro_rt;
2636 break;
2637#endif /* INET */
2638#ifdef INET6
2639 case AF_INET6:
2640 hlen = sizeof(struct ip6_hdr);
2641 bzero(&ro6, sizeof(ro6));
2642 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst;
2643 dst6->sin6_family = AF_INET6;
2644 dst6->sin6_len = sizeof(*dst6);
2645 dst6->sin6_addr = addr->v6;
2646#ifdef __FreeBSD__
2647#ifdef RTF_PRCLONING
2648 rtalloc_ign((struct route *)&ro6,
2649 (RTF_CLONING | RTF_PRCLONING));
2650#else /* !RTF_PRCLONING */
2651 rtalloc_ign((struct route *)&ro6, RTF_CLONING);
2652#endif
2653#else /* ! __FreeBSD__ */
2654 rtalloc_noclone((struct route *)&ro6, NO_CLONING);
2655#endif
2656 rt = ro6.ro_rt;
2657 break;
2658#endif /* INET6 */
2659 }
2660
2661 if (rt && rt->rt_ifp) {
2662 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr);
2663 mss = max(tcp_mssdflt, mss);
2664 RTFREE(rt);
2665 }
2666 mss = min(mss, offer);
2667 mss = max(mss, 64); /* sanity - at least max opt space */
2668 return (mss);
2669}
2670
2671void
2672pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr)
2673{
2674 struct pf_rule *r = s->rule.ptr;
2675
2676 s->rt_kif = NULL;
2677 if (!r->rt || r->rt == PF_FASTROUTE)
2678 return;
2679 switch (s->af) {
2680#ifdef INET
2681 case AF_INET:
2682 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL,
2683 &s->nat_src_node);
2684 s->rt_kif = r->rpool.cur->kif;
2685 break;
2686#endif /* INET */
2687#ifdef INET6
2688 case AF_INET6:
2689 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL,
2690 &s->nat_src_node);
2691 s->rt_kif = r->rpool.cur->kif;
2692 break;
2693#endif /* INET6 */
2694 }
2695}
2696
2697int
2698pf_test_tcp(struct pf_rule **rm, struct pf_state **sm, int direction,
2699 struct pfi_kif *kif, struct mbuf *m, int off, void *h,
2700#ifdef __FreeBSD__
2701 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
2702 struct inpcb *inp)
2703#else
2704 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
2705#endif
2706{
2707 struct pf_rule *nr = NULL;
2708 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
2709 struct tcphdr *th = pd->hdr.tcp;
2710 u_int16_t bport, nport = 0;
2711 sa_family_t af = pd->af;
2712 int lookup = -1;
2713 uid_t uid;
2714 gid_t gid;
2715 struct pf_rule *r, *a = NULL;
2716 struct pf_ruleset *ruleset = NULL;
2717 struct pf_src_node *nsn = NULL;
2718 u_short reason;
2719 int rewrite = 0;
2720 struct pf_tag *pftag = NULL;
2721 int tag = -1;
2722 u_int16_t mss = tcp_mssdflt;
2723
2724 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
2725
2726 if (direction == PF_OUT) {
2727 bport = nport = th->th_sport;
2728 /* check outgoing packet for BINAT/NAT */
2729 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
2730 saddr, th->th_sport, daddr, th->th_dport,
2731 &pd->naddr, &nport)) != NULL) {
2732 PF_ACPY(&pd->baddr, saddr, af);
2733 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
2734 &th->th_sum, &pd->naddr, nport, 0, af);
2735 rewrite++;
2736 if (nr->natpass)
2737 r = NULL;
2738 pd->nat_rule = nr;
2739 }
2740 } else {
2741 bport = nport = th->th_dport;
2742 /* check incoming packet for BINAT/RDR */
2743 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
2744 saddr, th->th_sport, daddr, th->th_dport,
2745 &pd->naddr, &nport)) != NULL) {
2746 PF_ACPY(&pd->baddr, daddr, af);
2747 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
2748 &th->th_sum, &pd->naddr, nport, 0, af);
2749 rewrite++;
2750 if (nr->natpass)
2751 r = NULL;
2752 pd->nat_rule = nr;
2753 }
2754 }
2755
2756 while (r != NULL) {
2757 r->evaluations++;
2758 if (r->kif != NULL &&
2759 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
2760 r = r->skip[PF_SKIP_IFP].ptr;
2761 else if (r->direction && r->direction != direction)
2762 r = r->skip[PF_SKIP_DIR].ptr;
2763 else if (r->af && r->af != af)
2764 r = r->skip[PF_SKIP_AF].ptr;
2765 else if (r->proto && r->proto != IPPROTO_TCP)
2766 r = r->skip[PF_SKIP_PROTO].ptr;
2767 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
2768 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
2769 else if (r->src.port_op && !pf_match_port(r->src.port_op,
2770 r->src.port[0], r->src.port[1], th->th_sport))
2771 r = r->skip[PF_SKIP_SRC_PORT].ptr;
2772 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
2773 r = r->skip[PF_SKIP_DST_ADDR].ptr;
2774 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
2775 r->dst.port[0], r->dst.port[1], th->th_dport))
2776 r = r->skip[PF_SKIP_DST_PORT].ptr;
2777 else if (r->tos && !(r->tos & pd->tos))
2778 r = TAILQ_NEXT(r, entries);
2779 else if (r->rule_flag & PFRULE_FRAGMENT)
2780 r = TAILQ_NEXT(r, entries);
2781 else if ((r->flagset & th->th_flags) != r->flags)
2782 r = TAILQ_NEXT(r, entries);
2783 else if (r->uid.op && (lookup != -1 || (lookup =
2784#ifdef __FreeBSD__
2785 pf_socket_lookup(&uid, &gid, direction, pd, inp), 1)) &&
2786#else
2787 pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2788#endif
2789 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
2790 uid))
2791 r = TAILQ_NEXT(r, entries);
2792 else if (r->gid.op && (lookup != -1 || (lookup =
2793#ifdef __FreeBSD__
2794 pf_socket_lookup(&uid, &gid, direction, pd, inp), 1)) &&
2795#else
2796 pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2797#endif
2798 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
2799 gid))
2800 r = TAILQ_NEXT(r, entries);
2801 else if (r->match_tag && !pf_match_tag(m, r, nr, pftag, &tag))
2802 r = TAILQ_NEXT(r, entries);
2803 else if (r->anchorname[0] && r->anchor == NULL)
2804 r = TAILQ_NEXT(r, entries);
2805 else if (r->os_fingerprint != PF_OSFP_ANY && !pf_osfp_match(
2806 pf_osfp_fingerprint(pd, m, off, th), r->os_fingerprint))
2807 r = TAILQ_NEXT(r, entries);
2808 else {
2809 if (r->tag)
2810 tag = r->tag;
2811 if (r->anchor == NULL) {
2812 *rm = r;
2813 *am = a;
2814 *rsm = ruleset;
2815 if ((*rm)->quick)
2816 break;
2817 r = TAILQ_NEXT(r, entries);
2818 } else
2819 PF_STEP_INTO_ANCHOR(r, a, ruleset,
2820 PF_RULESET_FILTER);
2821 }
2822 if (r == NULL && a != NULL)
2823 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
2824 PF_RULESET_FILTER);
2825 }
2826 r = *rm;
2827 a = *am;
2828 ruleset = *rsm;
2829
2830 REASON_SET(&reason, PFRES_MATCH);
2831
2832 if (r->log) {
2833 if (rewrite)
2834 m_copyback(m, off, sizeof(*th), (caddr_t)th);
2835 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
2836 }
2837
2838 if ((r->action == PF_DROP) &&
2839 ((r->rule_flag & PFRULE_RETURNRST) ||
2840 (r->rule_flag & PFRULE_RETURNICMP) ||
2841 (r->rule_flag & PFRULE_RETURN))) {
2842 /* undo NAT changes, if they have taken place */
2843 if (nr != NULL) {
2844 if (direction == PF_OUT) {
2845 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
2846 &th->th_sum, &pd->baddr, bport, 0, af);
2847 rewrite++;
2848 } else {
2849 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
2850 &th->th_sum, &pd->baddr, bport, 0, af);
2851 rewrite++;
2852 }
2853 }
2854 if (((r->rule_flag & PFRULE_RETURNRST) ||
2855 (r->rule_flag & PFRULE_RETURN)) &&
2856 !(th->th_flags & TH_RST)) {
2857 u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
2858
2859 if (th->th_flags & TH_SYN)
2860 ack++;
2861 if (th->th_flags & TH_FIN)
2862 ack++;
2863 pf_send_tcp(r, af, pd->dst,
2864 pd->src, th->th_dport, th->th_sport,
2865 ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0,
2866 r->return_ttl);
2867 } else if ((af == AF_INET) && r->return_icmp)
2868 pf_send_icmp(m, r->return_icmp >> 8,
2869 r->return_icmp & 255, af, r);
2870 else if ((af == AF_INET6) && r->return_icmp6)
2871 pf_send_icmp(m, r->return_icmp6 >> 8,
2872 r->return_icmp6 & 255, af, r);
2873 }
2874
2875 if (r->action == PF_DROP)
2876 return (PF_DROP);
2877
2878 if (pf_tag_packet(m, pftag, tag)) {
2879 REASON_SET(&reason, PFRES_MEMORY);
2880 return (PF_DROP);
2881 }
2882
2883 if (r->keep_state || nr != NULL ||
2884 (pd->flags & PFDESC_TCP_NORM)) {
2885 /* create new state */
2886 u_int16_t len;
2887 struct pf_state *s = NULL;
2888 struct pf_src_node *sn = NULL;
2889
2890 len = pd->tot_len - off - (th->th_off << 2);
2891
2892 /* check maximums */
2893 if (r->max_states && (r->states >= r->max_states))
2894 goto cleanup;
2895 /* src node for flter rule */
2896 if ((r->rule_flag & PFRULE_SRCTRACK ||
2897 r->rpool.opts & PF_POOL_STICKYADDR) &&
2898 pf_insert_src_node(&sn, r, saddr, af) != 0)
2899 goto cleanup;
2900 /* src node for translation rule */
2901 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
2902 ((direction == PF_OUT &&
2903 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
2904 (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
2905 goto cleanup;
2906 s = pool_get(&pf_state_pl, PR_NOWAIT);
2907 if (s == NULL) {
2908cleanup:
2909 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
2910 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
2911 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
2912 pf_status.src_nodes--;
2913 pool_put(&pf_src_tree_pl, sn);
2914 }
2915 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
2916 nsn->expire == 0) {
2917 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
2918 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
2919 pf_status.src_nodes--;
2920 pool_put(&pf_src_tree_pl, nsn);
2921 }
2922 REASON_SET(&reason, PFRES_MEMORY);
2923 return (PF_DROP);
2924 }
2925 bzero(s, sizeof(*s));
2926 r->states++;
2927 if (a != NULL)
2928 a->states++;
2929 s->rule.ptr = r;
2930 s->nat_rule.ptr = nr;
2931 if (s->nat_rule.ptr != NULL)
2932 s->nat_rule.ptr->states++;
2933 s->anchor.ptr = a;
2934 s->allow_opts = r->allow_opts;
2935 s->log = r->log & 2;
2936 s->proto = IPPROTO_TCP;
2937 s->direction = direction;
2938 s->af = af;
2939 if (direction == PF_OUT) {
2940 PF_ACPY(&s->gwy.addr, saddr, af);
2941 s->gwy.port = th->th_sport; /* sport */
2942 PF_ACPY(&s->ext.addr, daddr, af);
2943 s->ext.port = th->th_dport;
2944 if (nr != NULL) {
2945 PF_ACPY(&s->lan.addr, &pd->baddr, af);
2946 s->lan.port = bport;
2947 } else {
2948 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
2949 s->lan.port = s->gwy.port;
2950 }
2951 } else {
2952 PF_ACPY(&s->lan.addr, daddr, af);
2953 s->lan.port = th->th_dport;
2954 PF_ACPY(&s->ext.addr, saddr, af);
2955 s->ext.port = th->th_sport;
2956 if (nr != NULL) {
2957 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
2958 s->gwy.port = bport;
2959 } else {
2960 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
2961 s->gwy.port = s->lan.port;
2962 }
2963 }
2964
2965 s->src.seqlo = ntohl(th->th_seq);
2966 s->src.seqhi = s->src.seqlo + len + 1;
2967 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
2968 r->keep_state == PF_STATE_MODULATE) {
2969 /* Generate sequence number modulator */
2970 while ((s->src.seqdiff = arc4random()) == 0)
2971 ;
2972 pf_change_a(&th->th_seq, &th->th_sum,
2973 htonl(s->src.seqlo + s->src.seqdiff), 0);
2974 rewrite = 1;
2975 } else
2976 s->src.seqdiff = 0;
2977 if (th->th_flags & TH_SYN) {
2978 s->src.seqhi++;
2979 s->src.wscale = pf_get_wscale(m, off, th->th_off, af);
2980 }
2981 s->src.max_win = MAX(ntohs(th->th_win), 1);
2982 if (s->src.wscale & PF_WSCALE_MASK) {
2983 /* Remove scale factor from initial window */
2984 int win = s->src.max_win;
2985 win += 1 << (s->src.wscale & PF_WSCALE_MASK);
2986 s->src.max_win = (win - 1) >>
2987 (s->src.wscale & PF_WSCALE_MASK);
2988 }
2989 if (th->th_flags & TH_FIN)
2990 s->src.seqhi++;
2991 s->dst.seqhi = 1;
2992 s->dst.max_win = 1;
2993 s->src.state = TCPS_SYN_SENT;
2994 s->dst.state = TCPS_CLOSED;
2995#ifdef __FreeBSD__
2996 s->creation = time_second;
2997 s->expire = time_second;
2998#else
2999 s->creation = time.tv_sec;
3000 s->expire = time.tv_sec;
3001#endif
3002 s->timeout = PFTM_TCP_FIRST_PACKET;
3003 pf_set_rt_ifp(s, saddr);
3004 if (sn != NULL) {
3005 s->src_node = sn;
3006 s->src_node->states++;
3007 }
3008 if (nsn != NULL) {
3009 PF_ACPY(&nsn->raddr, &pd->naddr, af);
3010 s->nat_src_node = nsn;
3011 s->nat_src_node->states++;
3012 }
3013 if ((pd->flags & PFDESC_TCP_NORM) && pf_normalize_tcp_init(m,
3014 off, pd, th, &s->src, &s->dst)) {
3015 REASON_SET(&reason, PFRES_MEMORY);
3016 pf_src_tree_remove_state(s);
3017 pool_put(&pf_state_pl, s);
3018 return (PF_DROP);
3019 }
3020 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
3021 pf_normalize_tcp_stateful(m, off, pd, &reason, th, &s->src,
3022 &s->dst, &rewrite)) {
3023 pf_normalize_tcp_cleanup(s);
3024 pf_src_tree_remove_state(s);
3025 pool_put(&pf_state_pl, s);
3026 return (PF_DROP);
3027 }
3028 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3029 pf_normalize_tcp_cleanup(s);
3030 REASON_SET(&reason, PFRES_MEMORY);
3031 pf_src_tree_remove_state(s);
3032 pool_put(&pf_state_pl, s);
3033 return (PF_DROP);
3034 } else
3035 *sm = s;
3036 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
3037 r->keep_state == PF_STATE_SYNPROXY) {
3038 s->src.state = PF_TCPS_PROXY_SRC;
3039 if (nr != NULL) {
3040 if (direction == PF_OUT) {
3041 pf_change_ap(saddr, &th->th_sport,
3042 pd->ip_sum, &th->th_sum, &pd->baddr,
3043 bport, 0, af);
3044 } else {
3045 pf_change_ap(daddr, &th->th_dport,
3046 pd->ip_sum, &th->th_sum, &pd->baddr,
3047 bport, 0, af);
3048 }
3049 }
3050 s->src.seqhi = arc4random();
3051 /* Find mss option */
3052 mss = pf_get_mss(m, off, th->th_off, af);
3053 mss = pf_calc_mss(saddr, af, mss);
3054 mss = pf_calc_mss(daddr, af, mss);
3055 s->src.mss = mss;
3056 pf_send_tcp(r, af, daddr, saddr, th->th_dport,
3057 th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1,
3058 TH_SYN|TH_ACK, 0, s->src.mss, 0);
3059 return (PF_SYNPROXY_DROP);
3060 }
3061 }
3062
3063 /* copy back packet headers if we performed NAT operations */
3064 if (rewrite)
3065 m_copyback(m, off, sizeof(*th), (caddr_t)th);
3066
3067 return (PF_PASS);
3068}
3069
3070int
3071pf_test_udp(struct pf_rule **rm, struct pf_state **sm, int direction,
3072 struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3073#ifdef __FreeBSD__
3074 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
3075 struct inpcb *inp)
3076#else
3077 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
3078#endif
3079{
3080 struct pf_rule *nr = NULL;
3081 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
3082 struct udphdr *uh = pd->hdr.udp;
3083 u_int16_t bport, nport = 0;
3084 sa_family_t af = pd->af;
3085 int lookup = -1;
3086 uid_t uid;
3087 gid_t gid;
3088 struct pf_rule *r, *a = NULL;
3089 struct pf_ruleset *ruleset = NULL;
3090 struct pf_src_node *nsn = NULL;
3091 u_short reason;
3092 int rewrite = 0;
3093 struct pf_tag *pftag = NULL;
3094 int tag = -1;
3095
3096 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3097
3098 if (direction == PF_OUT) {
3099 bport = nport = uh->uh_sport;
3100 /* check outgoing packet for BINAT/NAT */
3101 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3102 saddr, uh->uh_sport, daddr, uh->uh_dport,
3103 &pd->naddr, &nport)) != NULL) {
3104 PF_ACPY(&pd->baddr, saddr, af);
3105 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum,
3106 &uh->uh_sum, &pd->naddr, nport, 1, af);
3107 rewrite++;
3108 if (nr->natpass)
3109 r = NULL;
3110 pd->nat_rule = nr;
3111 }
3112 } else {
3113 bport = nport = uh->uh_dport;
3114 /* check incoming packet for BINAT/RDR */
3115 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3116 saddr, uh->uh_sport, daddr, uh->uh_dport, &pd->naddr,
3117 &nport)) != NULL) {
3118 PF_ACPY(&pd->baddr, daddr, af);
3119 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum,
3120 &uh->uh_sum, &pd->naddr, nport, 1, af);
3121 rewrite++;
3122 if (nr->natpass)
3123 r = NULL;
3124 pd->nat_rule = nr;
3125 }
3126 }
3127
3128 while (r != NULL) {
3129 r->evaluations++;
3130 if (r->kif != NULL &&
3131 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3132 r = r->skip[PF_SKIP_IFP].ptr;
3133 else if (r->direction && r->direction != direction)
3134 r = r->skip[PF_SKIP_DIR].ptr;
3135 else if (r->af && r->af != af)
3136 r = r->skip[PF_SKIP_AF].ptr;
3137 else if (r->proto && r->proto != IPPROTO_UDP)
3138 r = r->skip[PF_SKIP_PROTO].ptr;
3139 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
3140 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3141 else if (r->src.port_op && !pf_match_port(r->src.port_op,
3142 r->src.port[0], r->src.port[1], uh->uh_sport))
3143 r = r->skip[PF_SKIP_SRC_PORT].ptr;
3144 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
3145 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3146 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
3147 r->dst.port[0], r->dst.port[1], uh->uh_dport))
3148 r = r->skip[PF_SKIP_DST_PORT].ptr;
3149 else if (r->tos && !(r->tos & pd->tos))
3150 r = TAILQ_NEXT(r, entries);
3151 else if (r->rule_flag & PFRULE_FRAGMENT)
3152 r = TAILQ_NEXT(r, entries);
3153 else if (r->uid.op && (lookup != -1 || (lookup =
3154#ifdef __FreeBSD__
3155 pf_socket_lookup(&uid, &gid, direction, pd, inp), 1)) &&
3156#else
3157 pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
3158#endif
3159 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
3160 uid))
3161 r = TAILQ_NEXT(r, entries);
3162 else if (r->gid.op && (lookup != -1 || (lookup =
3163#ifdef __FreeBSD__
3164 pf_socket_lookup(&uid, &gid, direction, pd, inp), 1)) &&
3165#else
3166 pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
3167#endif
3168 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
3169 gid))
3170 r = TAILQ_NEXT(r, entries);
3171 else if (r->match_tag && !pf_match_tag(m, r, nr, pftag, &tag))
3172 r = TAILQ_NEXT(r, entries);
3173 else if (r->anchorname[0] && r->anchor == NULL)
3174 r = TAILQ_NEXT(r, entries);
3175 else if (r->os_fingerprint != PF_OSFP_ANY)
3176 r = TAILQ_NEXT(r, entries);
3177 else {
3178 if (r->tag)
3179 tag = r->tag;
3180 if (r->anchor == NULL) {
3181 *rm = r;
3182 *am = a;
3183 *rsm = ruleset;
3184 if ((*rm)->quick)
3185 break;
3186 r = TAILQ_NEXT(r, entries);
3187 } else
3188 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3189 PF_RULESET_FILTER);
3190 }
3191 if (r == NULL && a != NULL)
3192 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3193 PF_RULESET_FILTER);
3194 }
3195 r = *rm;
3196 a = *am;
3197 ruleset = *rsm;
3198
3199 REASON_SET(&reason, PFRES_MATCH);
3200
3201 if (r->log) {
3202 if (rewrite)
3203 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
3204 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3205 }
3206
3207 if ((r->action == PF_DROP) &&
3208 ((r->rule_flag & PFRULE_RETURNICMP) ||
3209 (r->rule_flag & PFRULE_RETURN))) {
3210 /* undo NAT changes, if they have taken place */
3211 if (nr != NULL) {
3212 if (direction == PF_OUT) {
3213 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum,
3214 &uh->uh_sum, &pd->baddr, bport, 1, af);
3215 rewrite++;
3216 } else {
3217 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum,
3218 &uh->uh_sum, &pd->baddr, bport, 1, af);
3219 rewrite++;
3220 }
3221 }
3222 if ((af == AF_INET) && r->return_icmp)
3223 pf_send_icmp(m, r->return_icmp >> 8,
3224 r->return_icmp & 255, af, r);
3225 else if ((af == AF_INET6) && r->return_icmp6)
3226 pf_send_icmp(m, r->return_icmp6 >> 8,
3227 r->return_icmp6 & 255, af, r);
3228 }
3229
3230 if (r->action == PF_DROP)
3231 return (PF_DROP);
3232
3233 if (pf_tag_packet(m, pftag, tag)) {
3234 REASON_SET(&reason, PFRES_MEMORY);
3235 return (PF_DROP);
3236 }
3237
3238 if (r->keep_state || nr != NULL) {
3239 /* create new state */
3240 struct pf_state *s = NULL;
3241 struct pf_src_node *sn = NULL;
3242
3243 /* check maximums */
3244 if (r->max_states && (r->states >= r->max_states))
3245 goto cleanup;
3246 /* src node for flter rule */
3247 if ((r->rule_flag & PFRULE_SRCTRACK ||
3248 r->rpool.opts & PF_POOL_STICKYADDR) &&
3249 pf_insert_src_node(&sn, r, saddr, af) != 0)
3250 goto cleanup;
3251 /* src node for translation rule */
3252 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3253 ((direction == PF_OUT &&
3254 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3255 (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
3256 goto cleanup;
3257 s = pool_get(&pf_state_pl, PR_NOWAIT);
3258 if (s == NULL) {
3259cleanup:
3260 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3261 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3262 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3263 pf_status.src_nodes--;
3264 pool_put(&pf_src_tree_pl, sn);
3265 }
3266 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3267 nsn->expire == 0) {
3268 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3269 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3270 pf_status.src_nodes--;
3271 pool_put(&pf_src_tree_pl, nsn);
3272 }
3273 REASON_SET(&reason, PFRES_MEMORY);
3274 return (PF_DROP);
3275 }
3276 bzero(s, sizeof(*s));
3277 r->states++;
3278 if (a != NULL)
3279 a->states++;
3280 s->rule.ptr = r;
3281 s->nat_rule.ptr = nr;
3282 if (s->nat_rule.ptr != NULL)
3283 s->nat_rule.ptr->states++;
3284 s->anchor.ptr = a;
3285 s->allow_opts = r->allow_opts;
3286 s->log = r->log & 2;
3287 s->proto = IPPROTO_UDP;
3288 s->direction = direction;
3289 s->af = af;
3290 if (direction == PF_OUT) {
3291 PF_ACPY(&s->gwy.addr, saddr, af);
3292 s->gwy.port = uh->uh_sport;
3293 PF_ACPY(&s->ext.addr, daddr, af);
3294 s->ext.port = uh->uh_dport;
3295 if (nr != NULL) {
3296 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3297 s->lan.port = bport;
3298 } else {
3299 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3300 s->lan.port = s->gwy.port;
3301 }
3302 } else {
3303 PF_ACPY(&s->lan.addr, daddr, af);
3304 s->lan.port = uh->uh_dport;
3305 PF_ACPY(&s->ext.addr, saddr, af);
3306 s->ext.port = uh->uh_sport;
3307 if (nr != NULL) {
3308 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3309 s->gwy.port = bport;
3310 } else {
3311 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3312 s->gwy.port = s->lan.port;
3313 }
3314 }
3315 s->src.state = PFUDPS_SINGLE;
3316 s->dst.state = PFUDPS_NO_TRAFFIC;
3317#ifdef __FreeBSD__
3318 s->creation = time_second;
3319 s->expire = time_second;
3320#else
3321 s->creation = time.tv_sec;
3322 s->expire = time.tv_sec;
3323#endif
3324 s->timeout = PFTM_UDP_FIRST_PACKET;
3325 pf_set_rt_ifp(s, saddr);
3326 if (sn != NULL) {
3327 s->src_node = sn;
3328 s->src_node->states++;
3329 }
3330 if (nsn != NULL) {
3331 PF_ACPY(&nsn->raddr, &pd->naddr, af);
3332 s->nat_src_node = nsn;
3333 s->nat_src_node->states++;
3334 }
3335 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3336 REASON_SET(&reason, PFRES_MEMORY);
3337 pf_src_tree_remove_state(s);
3338 pool_put(&pf_state_pl, s);
3339 return (PF_DROP);
3340 } else
3341 *sm = s;
3342 }
3343
3344 /* copy back packet headers if we performed NAT operations */
3345 if (rewrite)
3346 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
3347
3348 return (PF_PASS);
3349}
3350
3351int
3352pf_test_icmp(struct pf_rule **rm, struct pf_state **sm, int direction,
3353 struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3354 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
3355{
3356 struct pf_rule *nr = NULL;
3357 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
3358 struct pf_rule *r, *a = NULL;
3359 struct pf_ruleset *ruleset = NULL;
3360 struct pf_src_node *nsn = NULL;
3361 u_short reason;
3362 u_int16_t icmpid = 0; /* make the compiler happy */
3363 sa_family_t af = pd->af;
3364 u_int8_t icmptype = 0; /* make the compiler happy */
3365 u_int8_t icmpcode = 0; /* make the compiler happy */
3366 int state_icmp = 0;
3367 struct pf_tag *pftag = NULL;
3368 int tag = -1;
3369#ifdef INET6
3370 int rewrite = 0;
3371#endif /* INET6 */
3372
3373 switch (pd->proto) {
3374#ifdef INET
3375 case IPPROTO_ICMP:
3376 icmptype = pd->hdr.icmp->icmp_type;
3377 icmpcode = pd->hdr.icmp->icmp_code;
3378 icmpid = pd->hdr.icmp->icmp_id;
3379
3380 if (icmptype == ICMP_UNREACH ||
3381 icmptype == ICMP_SOURCEQUENCH ||
3382 icmptype == ICMP_REDIRECT ||
3383 icmptype == ICMP_TIMXCEED ||
3384 icmptype == ICMP_PARAMPROB)
3385 state_icmp++;
3386 break;
3387#endif /* INET */
3388#ifdef INET6
3389 case IPPROTO_ICMPV6:
3390 icmptype = pd->hdr.icmp6->icmp6_type;
3391 icmpcode = pd->hdr.icmp6->icmp6_code;
3392 icmpid = pd->hdr.icmp6->icmp6_id;
3393
3394 if (icmptype == ICMP6_DST_UNREACH ||
3395 icmptype == ICMP6_PACKET_TOO_BIG ||
3396 icmptype == ICMP6_TIME_EXCEEDED ||
3397 icmptype == ICMP6_PARAM_PROB)
3398 state_icmp++;
3399 break;
3400#endif /* INET6 */
3401 }
3402
3403 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3404
3405 if (direction == PF_OUT) {
3406 /* check outgoing packet for BINAT/NAT */
3407 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3408 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3409 PF_ACPY(&pd->baddr, saddr, af);
3410 switch (af) {
3411#ifdef INET
3412 case AF_INET:
3413 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3414 pd->naddr.v4.s_addr, 0);
3415 break;
3416#endif /* INET */
3417#ifdef INET6
3418 case AF_INET6:
3419 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum,
3420 &pd->naddr, 0);
3421 rewrite++;
3422 break;
3423#endif /* INET6 */
3424 }
3425 if (nr->natpass)
3426 r = NULL;
3427 pd->nat_rule = nr;
3428 }
3429 } else {
3430 /* check incoming packet for BINAT/RDR */
3431 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3432 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3433 PF_ACPY(&pd->baddr, daddr, af);
3434 switch (af) {
3435#ifdef INET
3436 case AF_INET:
3437 pf_change_a(&daddr->v4.s_addr,
3438 pd->ip_sum, pd->naddr.v4.s_addr, 0);
3439 break;
3440#endif /* INET */
3441#ifdef INET6
3442 case AF_INET6:
3443 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum,
3444 &pd->naddr, 0);
3445 rewrite++;
3446 break;
3447#endif /* INET6 */
3448 }
3449 if (nr->natpass)
3450 r = NULL;
3451 pd->nat_rule = nr;
3452 }
3453 }
3454
3455 while (r != NULL) {
3456 r->evaluations++;
3457 if (r->kif != NULL &&
3458 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3459 r = r->skip[PF_SKIP_IFP].ptr;
3460 else if (r->direction && r->direction != direction)
3461 r = r->skip[PF_SKIP_DIR].ptr;
3462 else if (r->af && r->af != af)
3463 r = r->skip[PF_SKIP_AF].ptr;
3464 else if (r->proto && r->proto != pd->proto)
3465 r = r->skip[PF_SKIP_PROTO].ptr;
3466 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
3467 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3468 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
3469 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3470 else if (r->type && r->type != icmptype + 1)
3471 r = TAILQ_NEXT(r, entries);
3472 else if (r->code && r->code != icmpcode + 1)
3473 r = TAILQ_NEXT(r, entries);
3474 else if (r->tos && !(r->tos & pd->tos))
3475 r = TAILQ_NEXT(r, entries);
3476 else if (r->rule_flag & PFRULE_FRAGMENT)
3477 r = TAILQ_NEXT(r, entries);
3478 else if (r->match_tag && !pf_match_tag(m, r, nr, pftag, &tag))
3479 r = TAILQ_NEXT(r, entries);
3480 else if (r->anchorname[0] && r->anchor == NULL)
3481 r = TAILQ_NEXT(r, entries);
3482 else if (r->os_fingerprint != PF_OSFP_ANY)
3483 r = TAILQ_NEXT(r, entries);
3484 else {
3485 if (r->tag)
3486 tag = r->tag;
3487 if (r->anchor == NULL) {
3488 *rm = r;
3489 *am = a;
3490 *rsm = ruleset;
3491 if ((*rm)->quick)
3492 break;
3493 r = TAILQ_NEXT(r, entries);
3494 } else
3495 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3496 PF_RULESET_FILTER);
3497 }
3498 if (r == NULL && a != NULL)
3499 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3500 PF_RULESET_FILTER);
3501 }
3502 r = *rm;
3503 a = *am;
3504 ruleset = *rsm;
3505
3506 REASON_SET(&reason, PFRES_MATCH);
3507
3508 if (r->log) {
3509#ifdef INET6
3510 if (rewrite)
3511 m_copyback(m, off, sizeof(struct icmp6_hdr),
3512 (caddr_t)pd->hdr.icmp6);
3513#endif /* INET6 */
3514 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3515 }
3516
3517 if (r->action != PF_PASS)
3518 return (PF_DROP);
3519
3520 if (pf_tag_packet(m, pftag, tag)) {
3521 REASON_SET(&reason, PFRES_MEMORY);
3522 return (PF_DROP);
3523 }
3524
3525 if (!state_icmp && (r->keep_state || nr != NULL)) {
3526 /* create new state */
3527 struct pf_state *s = NULL;
3528 struct pf_src_node *sn = NULL;
3529
3530 /* check maximums */
3531 if (r->max_states && (r->states >= r->max_states))
3532 goto cleanup;
3533 /* src node for flter rule */
3534 if ((r->rule_flag & PFRULE_SRCTRACK ||
3535 r->rpool.opts & PF_POOL_STICKYADDR) &&
3536 pf_insert_src_node(&sn, r, saddr, af) != 0)
3537 goto cleanup;
3538 /* src node for translation rule */
3539 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3540 ((direction == PF_OUT &&
3541 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3542 (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
3543 goto cleanup;
3544 s = pool_get(&pf_state_pl, PR_NOWAIT);
3545 if (s == NULL) {
3546cleanup:
3547 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3548 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3549 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3550 pf_status.src_nodes--;
3551 pool_put(&pf_src_tree_pl, sn);
3552 }
3553 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3554 nsn->expire == 0) {
3555 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3556 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3557 pf_status.src_nodes--;
3558 pool_put(&pf_src_tree_pl, nsn);
3559 }
3560 REASON_SET(&reason, PFRES_MEMORY);
3561 return (PF_DROP);
3562 }
3563 bzero(s, sizeof(*s));
3564 r->states++;
3565 if (a != NULL)
3566 a->states++;
3567 s->rule.ptr = r;
3568 s->nat_rule.ptr = nr;
3569 if (s->nat_rule.ptr != NULL)
3570 s->nat_rule.ptr->states++;
3571 s->anchor.ptr = a;
3572 s->allow_opts = r->allow_opts;
3573 s->log = r->log & 2;
3574 s->proto = pd->proto;
3575 s->direction = direction;
3576 s->af = af;
3577 if (direction == PF_OUT) {
3578 PF_ACPY(&s->gwy.addr, saddr, af);
3579 s->gwy.port = icmpid;
3580 PF_ACPY(&s->ext.addr, daddr, af);
3581 s->ext.port = icmpid;
3582 if (nr != NULL)
3583 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3584 else
3585 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3586 s->lan.port = icmpid;
3587 } else {
3588 PF_ACPY(&s->lan.addr, daddr, af);
3589 s->lan.port = icmpid;
3590 PF_ACPY(&s->ext.addr, saddr, af);
3591 s->ext.port = icmpid;
3592 if (nr != NULL)
3593 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3594 else
3595 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3596 s->gwy.port = icmpid;
3597 }
3598#ifdef __FreeBSD__
3599 s->creation = time_second;
3600 s->expire = time_second;
3601#else
3602 s->creation = time.tv_sec;
3603 s->expire = time.tv_sec;
3604#endif
3605 s->timeout = PFTM_ICMP_FIRST_PACKET;
3606 pf_set_rt_ifp(s, saddr);
3607 if (sn != NULL) {
3608 s->src_node = sn;
3609 s->src_node->states++;
3610 }
3611 if (nsn != NULL) {
3612 PF_ACPY(&nsn->raddr, &pd->naddr, af);
3613 s->nat_src_node = nsn;
3614 s->nat_src_node->states++;
3615 }
3616 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3617 REASON_SET(&reason, PFRES_MEMORY);
3618 pf_src_tree_remove_state(s);
3619 pool_put(&pf_state_pl, s);
3620 return (PF_DROP);
3621 } else
3622 *sm = s;
3623 }
3624
3625#ifdef INET6
3626 /* copy back packet headers if we performed IPv6 NAT operations */
3627 if (rewrite)
3628 m_copyback(m, off, sizeof(struct icmp6_hdr),
3629 (caddr_t)pd->hdr.icmp6);
3630#endif /* INET6 */
3631
3632 return (PF_PASS);
3633}
3634
3635int
3636pf_test_other(struct pf_rule **rm, struct pf_state **sm, int direction,
3637 struct pfi_kif *kif, struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
3638 struct pf_rule **am, struct pf_ruleset **rsm)
3639{
3640 struct pf_rule *nr = NULL;
3641 struct pf_rule *r, *a = NULL;
3642 struct pf_ruleset *ruleset = NULL;
3643 struct pf_src_node *nsn = NULL;
3644 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
3645 sa_family_t af = pd->af;
3646 u_short reason;
3647 struct pf_tag *pftag = NULL;
3648 int tag = -1;
3649
3650 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3651
3652 if (direction == PF_OUT) {
3653 /* check outgoing packet for BINAT/NAT */
3654 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3655 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3656 PF_ACPY(&pd->baddr, saddr, af);
3657 switch (af) {
3658#ifdef INET
3659 case AF_INET:
3660 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3661 pd->naddr.v4.s_addr, 0);
3662 break;
3663#endif /* INET */
3664#ifdef INET6
3665 case AF_INET6:
3666 PF_ACPY(saddr, &pd->naddr, af);
3667 break;
3668#endif /* INET6 */
3669 }
3670 if (nr->natpass)
3671 r = NULL;
3672 pd->nat_rule = nr;
3673 }
3674 } else {
3675 /* check incoming packet for BINAT/RDR */
3676 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3677 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3678 PF_ACPY(&pd->baddr, daddr, af);
3679 switch (af) {
3680#ifdef INET
3681 case AF_INET:
3682 pf_change_a(&daddr->v4.s_addr,
3683 pd->ip_sum, pd->naddr.v4.s_addr, 0);
3684 break;
3685#endif /* INET */
3686#ifdef INET6
3687 case AF_INET6:
3688 PF_ACPY(daddr, &pd->naddr, af);
3689 break;
3690#endif /* INET6 */
3691 }
3692 if (nr->natpass)
3693 r = NULL;
3694 pd->nat_rule = nr;
3695 }
3696 }
3697
3698 while (r != NULL) {
3699 r->evaluations++;
3700 if (r->kif != NULL &&
3701 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3702 r = r->skip[PF_SKIP_IFP].ptr;
3703 else if (r->direction && r->direction != direction)
3704 r = r->skip[PF_SKIP_DIR].ptr;
3705 else if (r->af && r->af != af)
3706 r = r->skip[PF_SKIP_AF].ptr;
3707 else if (r->proto && r->proto != pd->proto)
3708 r = r->skip[PF_SKIP_PROTO].ptr;
3709 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, r->src.not))
3710 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3711 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, r->dst.not))
3712 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3713 else if (r->tos && !(r->tos & pd->tos))
3714 r = TAILQ_NEXT(r, entries);
3715 else if (r->rule_flag & PFRULE_FRAGMENT)
3716 r = TAILQ_NEXT(r, entries);
3717 else if (r->match_tag && !pf_match_tag(m, r, nr, pftag, &tag))
3718 r = TAILQ_NEXT(r, entries);
3719 else if (r->anchorname[0] && r->anchor == NULL)
3720 r = TAILQ_NEXT(r, entries);
3721 else if (r->os_fingerprint != PF_OSFP_ANY)
3722 r = TAILQ_NEXT(r, entries);
3723 else {
3724 if (r->tag)
3725 tag = r->tag;
3726 if (r->anchor == NULL) {
3727 *rm = r;
3728 *am = a;
3729 *rsm = ruleset;
3730 if ((*rm)->quick)
3731 break;
3732 r = TAILQ_NEXT(r, entries);
3733 } else
3734 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3735 PF_RULESET_FILTER);
3736 }
3737 if (r == NULL && a != NULL)
3738 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3739 PF_RULESET_FILTER);
3740 }
3741 r = *rm;
3742 a = *am;
3743 ruleset = *rsm;
3744
3745 REASON_SET(&reason, PFRES_MATCH);
3746
3747 if (r->log)
3748 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3749
3750 if ((r->action == PF_DROP) &&
3751 ((r->rule_flag & PFRULE_RETURNICMP) ||
3752 (r->rule_flag & PFRULE_RETURN))) {
3753 struct pf_addr *a = NULL;
3754
3755 if (nr != NULL) {
3756 if (direction == PF_OUT)
3757 a = saddr;
3758 else
3759 a = daddr;
3760 }
3761 if (a != NULL) {
3762 switch (af) {
3763#ifdef INET
3764 case AF_INET:
3765 pf_change_a(&a->v4.s_addr, pd->ip_sum,
3766 pd->baddr.v4.s_addr, 0);
3767 break;
3768#endif /* INET */
3769#ifdef INET6
3770 case AF_INET6:
3771 PF_ACPY(a, &pd->baddr, af);
3772 break;
3773#endif /* INET6 */
3774 }
3775 }
3776 if ((af == AF_INET) && r->return_icmp)
3777 pf_send_icmp(m, r->return_icmp >> 8,
3778 r->return_icmp & 255, af, r);
3779 else if ((af == AF_INET6) && r->return_icmp6)
3780 pf_send_icmp(m, r->return_icmp6 >> 8,
3781 r->return_icmp6 & 255, af, r);
3782 }
3783
3784 if (r->action != PF_PASS)
3785 return (PF_DROP);
3786
3787 if (pf_tag_packet(m, pftag, tag)) {
3788 REASON_SET(&reason, PFRES_MEMORY);
3789 return (PF_DROP);
3790 }
3791
3792 if (r->keep_state || nr != NULL) {
3793 /* create new state */
3794 struct pf_state *s = NULL;
3795 struct pf_src_node *sn = NULL;
3796
3797 /* check maximums */
3798 if (r->max_states && (r->states >= r->max_states))
3799 goto cleanup;
3800 /* src node for flter rule */
3801 if ((r->rule_flag & PFRULE_SRCTRACK ||
3802 r->rpool.opts & PF_POOL_STICKYADDR) &&
3803 pf_insert_src_node(&sn, r, saddr, af) != 0)
3804 goto cleanup;
3805 /* src node for translation rule */
3806 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3807 ((direction == PF_OUT &&
3808 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3809 (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
3810 goto cleanup;
3811 s = pool_get(&pf_state_pl, PR_NOWAIT);
3812 if (s == NULL) {
3813cleanup:
3814 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3815 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3816 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3817 pf_status.src_nodes--;
3818 pool_put(&pf_src_tree_pl, sn);
3819 }
3820 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3821 nsn->expire == 0) {
3822 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3823 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3824 pf_status.src_nodes--;
3825 pool_put(&pf_src_tree_pl, nsn);
3826 }
3827 REASON_SET(&reason, PFRES_MEMORY);
3828 return (PF_DROP);
3829 }
3830 bzero(s, sizeof(*s));
3831 r->states++;
3832 if (a != NULL)
3833 a->states++;
3834 s->rule.ptr = r;
3835 s->nat_rule.ptr = nr;
3836 if (s->nat_rule.ptr != NULL)
3837 s->nat_rule.ptr->states++;
3838 s->anchor.ptr = a;
3839 s->allow_opts = r->allow_opts;
3840 s->log = r->log & 2;
3841 s->proto = pd->proto;
3842 s->direction = direction;
3843 s->af = af;
3844 if (direction == PF_OUT) {
3845 PF_ACPY(&s->gwy.addr, saddr, af);
3846 PF_ACPY(&s->ext.addr, daddr, af);
3847 if (nr != NULL)
3848 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3849 else
3850 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3851 } else {
3852 PF_ACPY(&s->lan.addr, daddr, af);
3853 PF_ACPY(&s->ext.addr, saddr, af);
3854 if (nr != NULL)
3855 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3856 else
3857 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3858 }
3859 s->src.state = PFOTHERS_SINGLE;
3860 s->dst.state = PFOTHERS_NO_TRAFFIC;
3861#ifdef __FreeBSD__
3862 s->creation = time_second;
3863 s->expire = time_second;
3864#else
3865 s->creation = time.tv_sec;
3866 s->expire = time.tv_sec;
3867#endif
3868 s->timeout = PFTM_OTHER_FIRST_PACKET;
3869 pf_set_rt_ifp(s, saddr);
3870 if (sn != NULL) {
3871 s->src_node = sn;
3872 s->src_node->states++;
3873 }
3874 if (nsn != NULL) {
3875 PF_ACPY(&nsn->raddr, &pd->naddr, af);
3876 s->nat_src_node = nsn;
3877 s->nat_src_node->states++;
3878 }
3879 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3880 REASON_SET(&reason, PFRES_MEMORY);
3881 pf_src_tree_remove_state(s);
3882 pool_put(&pf_state_pl, s);
3883 return (PF_DROP);
3884 } else
3885 *sm = s;
3886 }
3887
3888 return (PF_PASS);
3889}
3890
3891int
3892pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif,
3893 struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am,
3894 struct pf_ruleset **rsm)
3895{
3896 struct pf_rule *r, *a = NULL;
3897 struct pf_ruleset *ruleset = NULL;
3898 sa_family_t af = pd->af;
3899 u_short reason;
3900 struct pf_tag *pftag = NULL;
3901 int tag = -1;
3902
3903 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3904 while (r != NULL) {
3905 r->evaluations++;
3906 if (r->kif != NULL &&
3907 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3908 r = r->skip[PF_SKIP_IFP].ptr;
3909 else if (r->direction && r->direction != direction)
3910 r = r->skip[PF_SKIP_DIR].ptr;
3911 else if (r->af && r->af != af)
3912 r = r->skip[PF_SKIP_AF].ptr;
3913 else if (r->proto && r->proto != pd->proto)
3914 r = r->skip[PF_SKIP_PROTO].ptr;
3915 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, r->src.not))
3916 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3917 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, r->dst.not))
3918 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3919 else if (r->tos && !(r->tos & pd->tos))
3920 r = TAILQ_NEXT(r, entries);
3921 else if (r->src.port_op || r->dst.port_op ||
3922 r->flagset || r->type || r->code ||
3923 r->os_fingerprint != PF_OSFP_ANY)
3924 r = TAILQ_NEXT(r, entries);
3925 else if (r->match_tag && !pf_match_tag(m, r, NULL, pftag, &tag))
3926 r = TAILQ_NEXT(r, entries);
3927 else if (r->anchorname[0] && r->anchor == NULL)
3928 r = TAILQ_NEXT(r, entries);
3929 else {
3930 if (r->anchor == NULL) {
3931 *rm = r;
3932 *am = a;
3933 *rsm = ruleset;
3934 if ((*rm)->quick)
3935 break;
3936 r = TAILQ_NEXT(r, entries);
3937 } else
3938 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3939 PF_RULESET_FILTER);
3940 }
3941 if (r == NULL && a != NULL)
3942 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3943 PF_RULESET_FILTER);
3944 }
3945 r = *rm;
3946 a = *am;
3947 ruleset = *rsm;
3948
3949 REASON_SET(&reason, PFRES_MATCH);
3950
3951 if (r->log)
3952 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3953
3954 if (r->action != PF_PASS)
3955 return (PF_DROP);
3956
3957 if (pf_tag_packet(m, pftag, tag)) {
3958 REASON_SET(&reason, PFRES_MEMORY);
3959 return (PF_DROP);
3960 }
3961
3962 return (PF_PASS);
3963}
3964
3965int
3966pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
3967 struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
3968 u_short *reason)
3969{
3970 struct pf_state key;
3971 struct tcphdr *th = pd->hdr.tcp;
3972 u_int16_t win = ntohs(th->th_win);
3973 u_int32_t ack, end, seq;
3974 u_int8_t sws, dws;
3975 int ackskew;
3976 int copyback = 0;
3977 struct pf_state_peer *src, *dst;
3978
3979 key.af = pd->af;
3980 key.proto = IPPROTO_TCP;
3981 if (direction == PF_IN) {
3982 PF_ACPY(&key.ext.addr, pd->src, key.af);
3983 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
3984 key.ext.port = th->th_sport;
3985 key.gwy.port = th->th_dport;
3986 } else {
3987 PF_ACPY(&key.lan.addr, pd->src, key.af);
3988 PF_ACPY(&key.ext.addr, pd->dst, key.af);
3989 key.lan.port = th->th_sport;
3990 key.ext.port = th->th_dport;
3991 }
3992
3993 STATE_LOOKUP();
3994
3995 if (direction == (*state)->direction) {
3996 src = &(*state)->src;
3997 dst = &(*state)->dst;
3998 } else {
3999 src = &(*state)->dst;
4000 dst = &(*state)->src;
4001 }
4002
4003 if ((*state)->src.state == PF_TCPS_PROXY_SRC) {
4004 if (direction != (*state)->direction)
4005 return (PF_SYNPROXY_DROP);
4006 if (th->th_flags & TH_SYN) {
4007 if (ntohl(th->th_seq) != (*state)->src.seqlo)
4008 return (PF_DROP);
4009 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
4010 pd->src, th->th_dport, th->th_sport,
4011 (*state)->src.seqhi, ntohl(th->th_seq) + 1,
4012 TH_SYN|TH_ACK, 0, (*state)->src.mss, 0);
4013 return (PF_SYNPROXY_DROP);
4014 } else if (!(th->th_flags & TH_ACK) ||
4015 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
4016 (ntohl(th->th_seq) != (*state)->src.seqlo + 1))
4017 return (PF_DROP);
4018 else
4019 (*state)->src.state = PF_TCPS_PROXY_DST;
4020 }
4021 if ((*state)->src.state == PF_TCPS_PROXY_DST) {
4022 struct pf_state_host *src, *dst;
4023
4024 if (direction == PF_OUT) {
4025 src = &(*state)->gwy;
4026 dst = &(*state)->ext;
4027 } else {
4028 src = &(*state)->ext;
4029 dst = &(*state)->lan;
4030 }
4031 if (direction == (*state)->direction) {
4032 if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) ||
4033 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
4034 (ntohl(th->th_seq) != (*state)->src.seqlo + 1))
4035 return (PF_DROP);
4036 (*state)->src.max_win = MAX(ntohs(th->th_win), 1);
4037 if ((*state)->dst.seqhi == 1)
4038 (*state)->dst.seqhi = arc4random();
4039 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
4040 &dst->addr, src->port, dst->port,
4041 (*state)->dst.seqhi, 0, TH_SYN, 0,
4042 (*state)->src.mss, 0);
4043 return (PF_SYNPROXY_DROP);
4044 } else if (((th->th_flags & (TH_SYN|TH_ACK)) !=
4045 (TH_SYN|TH_ACK)) ||
4046 (ntohl(th->th_ack) != (*state)->dst.seqhi + 1))
4047 return (PF_DROP);
4048 else {
4049 (*state)->dst.max_win = MAX(ntohs(th->th_win), 1);
4050 (*state)->dst.seqlo = ntohl(th->th_seq);
4051 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
4052 pd->src, th->th_dport, th->th_sport,
4053 ntohl(th->th_ack), ntohl(th->th_seq) + 1,
4054 TH_ACK, (*state)->src.max_win, 0, 0);
4055 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
4056 &dst->addr, src->port, dst->port,
4057 (*state)->src.seqhi + 1, (*state)->src.seqlo + 1,
4058 TH_ACK, (*state)->dst.max_win, 0, 0);
4059 (*state)->src.seqdiff = (*state)->dst.seqhi -
4060 (*state)->src.seqlo;
4061 (*state)->dst.seqdiff = (*state)->src.seqhi -
4062 (*state)->dst.seqlo;
4063 (*state)->src.seqhi = (*state)->src.seqlo +
4064 (*state)->src.max_win;
4065 (*state)->dst.seqhi = (*state)->dst.seqlo +
4066 (*state)->dst.max_win;
4067 (*state)->src.wscale = (*state)->dst.wscale = 0;
4068 (*state)->src.state = (*state)->dst.state =
4069 TCPS_ESTABLISHED;
4070 return (PF_SYNPROXY_DROP);
4071 }
4072 }
4073
4074 if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) {
4075 sws = src->wscale & PF_WSCALE_MASK;
4076 dws = dst->wscale & PF_WSCALE_MASK;
4077 } else
4078 sws = dws = 0;
4079
4080 /*
4081 * Sequence tracking algorithm from Guido van Rooij's paper:
4082 * http://www.madison-gurkha.com/publications/tcp_filtering/
4083 * tcp_filtering.ps
4084 */
4085
4086 seq = ntohl(th->th_seq);
4087 if (src->seqlo == 0) {
4088 /* First packet from this end. Set its state */
4089
4090 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) &&
4091 src->scrub == NULL) {
4092 if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) {
4093 REASON_SET(reason, PFRES_MEMORY);
4094 return (PF_DROP);
4095 }
4096 }
4097
4098 /* Deferred generation of sequence number modulator */
4099 if (dst->seqdiff && !src->seqdiff) {
4100 while ((src->seqdiff = arc4random()) == 0)
4101 ;
4102 ack = ntohl(th->th_ack) - dst->seqdiff;
4103 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
4104 src->seqdiff), 0);
4105 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
4106 copyback = 1;
4107 } else {
4108 ack = ntohl(th->th_ack);
4109 }
4110
4111 end = seq + pd->p_len;
4112 if (th->th_flags & TH_SYN) {
4113 end++;
4114 if (dst->wscale & PF_WSCALE_FLAG) {
4115 src->wscale = pf_get_wscale(m, off, th->th_off,
4116 pd->af);
4117 if (src->wscale & PF_WSCALE_FLAG) {
4118 /* Remove scale factor from initial
4119 * window */
4120 sws = src->wscale & PF_WSCALE_MASK;
4121 win = ((u_int32_t)win + (1 << sws) - 1)
4122 >> sws;
4123 dws = dst->wscale & PF_WSCALE_MASK;
4124 } else {
4125 /* fixup other window */
4126 dst->max_win <<= dst->wscale &
4127 PF_WSCALE_MASK;
4128 /* in case of a retrans SYN|ACK */
4129 dst->wscale = 0;
4130 }
4131 }
4132 }
4133 if (th->th_flags & TH_FIN)
4134 end++;
4135
4136 src->seqlo = seq;
4137 if (src->state < TCPS_SYN_SENT)
4138 src->state = TCPS_SYN_SENT;
4139
4140 /*
4141 * May need to slide the window (seqhi may have been set by
4142 * the crappy stack check or if we picked up the connection
4143 * after establishment)
4144 */
4145 if (src->seqhi == 1 ||
4146 SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi))
4147 src->seqhi = end + MAX(1, dst->max_win << dws);
4148 if (win > src->max_win)
4149 src->max_win = win;
4150
4151 } else {
4152 ack = ntohl(th->th_ack) - dst->seqdiff;
4153 if (src->seqdiff) {
4154 /* Modulate sequence numbers */
4155 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
4156 src->seqdiff), 0);
4157 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
4158 copyback = 1;
4159 }
4160 end = seq + pd->p_len;
4161 if (th->th_flags & TH_SYN)
4162 end++;
4163 if (th->th_flags & TH_FIN)
4164 end++;
4165 }
4166
4167 if ((th->th_flags & TH_ACK) == 0) {
4168 /* Let it pass through the ack skew check */
4169 ack = dst->seqlo;
4170 } else if ((ack == 0 &&
4171 (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) ||
4172 /* broken tcp stacks do not set ack */
4173 (dst->state < TCPS_SYN_SENT)) {
4174 /*
4175 * Many stacks (ours included) will set the ACK number in an
4176 * FIN|ACK if the SYN times out -- no sequence to ACK.
4177 */
4178 ack = dst->seqlo;
4179 }
4180
4181 if (seq == end) {
4182 /* Ease sequencing restrictions on no data packets */
4183 seq = src->seqlo;
4184 end = seq;
4185 }
4186
4187 ackskew = dst->seqlo - ack;
4188
4189#define MAXACKWINDOW (0xffff + 1500) /* 1500 is an arbitrary fudge factor */
4190 if (SEQ_GEQ(src->seqhi, end) &&
4191 /* Last octet inside other's window space */
4192 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) &&
4193 /* Retrans: not more than one window back */
4194 (ackskew >= -MAXACKWINDOW) &&
4195 /* Acking not more than one reassembled fragment backwards */
4196 (ackskew <= (MAXACKWINDOW << sws))) {
4197 /* Acking not more than one window forward */
4198
4199 /* update max window */
4200 if (src->max_win < win)
4201 src->max_win = win;
4202 /* synchronize sequencing */
4203 if (SEQ_GT(end, src->seqlo))
4204 src->seqlo = end;
4205 /* slide the window of what the other end can send */
4206 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4207 dst->seqhi = ack + MAX((win << sws), 1);
4208
4209
4210 /* update states */
4211 if (th->th_flags & TH_SYN)
4212 if (src->state < TCPS_SYN_SENT)
4213 src->state = TCPS_SYN_SENT;
4214 if (th->th_flags & TH_FIN)
4215 if (src->state < TCPS_CLOSING)
4216 src->state = TCPS_CLOSING;
4217 if (th->th_flags & TH_ACK) {
4218 if (dst->state == TCPS_SYN_SENT)
4219 dst->state = TCPS_ESTABLISHED;
4220 else if (dst->state == TCPS_CLOSING)
4221 dst->state = TCPS_FIN_WAIT_2;
4222 }
4223 if (th->th_flags & TH_RST)
4224 src->state = dst->state = TCPS_TIME_WAIT;
4225
4226 /* update expire time */
4227#ifdef __FreeBSD__
4228 (*state)->expire = time_second;
4229#else
4230 (*state)->expire = time.tv_sec;
4231#endif
4232 if (src->state >= TCPS_FIN_WAIT_2 &&
4233 dst->state >= TCPS_FIN_WAIT_2)
4234 (*state)->timeout = PFTM_TCP_CLOSED;
4235 else if (src->state >= TCPS_FIN_WAIT_2 ||
4236 dst->state >= TCPS_FIN_WAIT_2)
4237 (*state)->timeout = PFTM_TCP_FIN_WAIT;
4238 else if (src->state < TCPS_ESTABLISHED ||
4239 dst->state < TCPS_ESTABLISHED)
4240 (*state)->timeout = PFTM_TCP_OPENING;
4241 else if (src->state >= TCPS_CLOSING ||
4242 dst->state >= TCPS_CLOSING)
4243 (*state)->timeout = PFTM_TCP_CLOSING;
4244 else
4245 (*state)->timeout = PFTM_TCP_ESTABLISHED;
4246
4247 /* Fall through to PASS packet */
4248
4249 } else if ((dst->state < TCPS_SYN_SENT ||
4250 dst->state >= TCPS_FIN_WAIT_2 ||
4251 src->state >= TCPS_FIN_WAIT_2) &&
4252 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) &&
4253 /* Within a window forward of the originating packet */
4254 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) {
4255 /* Within a window backward of the originating packet */
4256
4257 /*
4258 * This currently handles three situations:
4259 * 1) Stupid stacks will shotgun SYNs before their peer
4260 * replies.
4261 * 2) When PF catches an already established stream (the
4262 * firewall rebooted, the state table was flushed, routes
4263 * changed...)
4264 * 3) Packets get funky immediately after the connection
4265 * closes (this should catch Solaris spurious ACK|FINs
4266 * that web servers like to spew after a close)
4267 *
4268 * This must be a little more careful than the above code
4269 * since packet floods will also be caught here. We don't
4270 * update the TTL here to mitigate the damage of a packet
4271 * flood and so the same code can handle awkward establishment
4272 * and a loosened connection close.
4273 * In the establishment case, a correct peer response will
4274 * validate the connection, go through the normal state code
4275 * and keep updating the state TTL.
4276 */
4277
4278 if (pf_status.debug >= PF_DEBUG_MISC) {
4279 printf("pf: loose state match: ");
4280 pf_print_state(*state);
4281 pf_print_flags(th->th_flags);
4282 printf(" seq=%u ack=%u len=%u ackskew=%d pkts=%d:%d\n",
4283 seq, ack, pd->p_len, ackskew,
4284 (*state)->packets[0], (*state)->packets[1]);
4285 }
4286
4287 /* update max window */
4288 if (src->max_win < win)
4289 src->max_win = win;
4290 /* synchronize sequencing */
4291 if (SEQ_GT(end, src->seqlo))
4292 src->seqlo = end;
4293 /* slide the window of what the other end can send */
4294 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4295 dst->seqhi = ack + MAX((win << sws), 1);
4296
4297 /*
4298 * Cannot set dst->seqhi here since this could be a shotgunned
4299 * SYN and not an already established connection.
4300 */
4301
4302 if (th->th_flags & TH_FIN)
4303 if (src->state < TCPS_CLOSING)
4304 src->state = TCPS_CLOSING;
4305 if (th->th_flags & TH_RST)
4306 src->state = dst->state = TCPS_TIME_WAIT;
4307
4308 /* Fall through to PASS packet */
4309
4310 } else {
4311 if ((*state)->dst.state == TCPS_SYN_SENT &&
4312 (*state)->src.state == TCPS_SYN_SENT) {
4313 /* Send RST for state mismatches during handshake */
4314 if (!(th->th_flags & TH_RST)) {
4315 u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
4316
4317 if (th->th_flags & TH_SYN)
4318 ack++;
4319 if (th->th_flags & TH_FIN)
4320 ack++;
4321 pf_send_tcp((*state)->rule.ptr, pd->af,
4322 pd->dst, pd->src, th->th_dport,
4323 th->th_sport, ntohl(th->th_ack), ack,
4324 TH_RST|TH_ACK, 0, 0,
4325 (*state)->rule.ptr->return_ttl);
4326 }
4327 src->seqlo = 0;
4328 src->seqhi = 1;
4329 src->max_win = 1;
4330 } else if (pf_status.debug >= PF_DEBUG_MISC) {
4331 printf("pf: BAD state: ");
4332 pf_print_state(*state);
4333 pf_print_flags(th->th_flags);
4334 printf(" seq=%u ack=%u len=%u ackskew=%d pkts=%d:%d "
4335 "dir=%s,%s\n", seq, ack, pd->p_len, ackskew,
4336 (*state)->packets[0], (*state)->packets[1],
4337 direction == PF_IN ? "in" : "out",
4338 direction == (*state)->direction ? "fwd" : "rev");
4339 printf("pf: State failure on: %c %c %c %c | %c %c\n",
4340 SEQ_GEQ(src->seqhi, end) ? ' ' : '1',
4341 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ?
4342 ' ': '2',
4343 (ackskew >= -MAXACKWINDOW) ? ' ' : '3',
4344 (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4',
4345 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5',
4346 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6');
4347 }
4348 return (PF_DROP);
4349 }
4350
4351 if (dst->scrub || src->scrub) {
4352 if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4353 src, dst, ©back))
4354 return (PF_DROP);
4355 }
4356
4357 /* Any packets which have gotten here are to be passed */
4358
4359 /* translate source/destination address, if necessary */
4360 if (STATE_TRANSLATE(*state)) {
4361 if (direction == PF_OUT)
4362 pf_change_ap(pd->src, &th->th_sport, pd->ip_sum,
4363 &th->th_sum, &(*state)->gwy.addr,
4364 (*state)->gwy.port, 0, pd->af);
4365 else
4366 pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum,
4367 &th->th_sum, &(*state)->lan.addr,
4368 (*state)->lan.port, 0, pd->af);
4369 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4370 } else if (copyback) {
4371 /* Copyback sequence modulation or stateful scrub changes */
4372 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4373 }
4374
4375 return (PF_PASS);
4376}
4377
4378int
4379pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif,
4380 struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4381{
4382 struct pf_state_peer *src, *dst;
4383 struct pf_state key;
4384 struct udphdr *uh = pd->hdr.udp;
4385
4386 key.af = pd->af;
4387 key.proto = IPPROTO_UDP;
4388 if (direction == PF_IN) {
4389 PF_ACPY(&key.ext.addr, pd->src, key.af);
4390 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4391 key.ext.port = uh->uh_sport;
4392 key.gwy.port = uh->uh_dport;
4393 } else {
4394 PF_ACPY(&key.lan.addr, pd->src, key.af);
4395 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4396 key.lan.port = uh->uh_sport;
4397 key.ext.port = uh->uh_dport;
4398 }
4399
4400 STATE_LOOKUP();
4401
4402 if (direction == (*state)->direction) {
4403 src = &(*state)->src;
4404 dst = &(*state)->dst;
4405 } else {
4406 src = &(*state)->dst;
4407 dst = &(*state)->src;
4408 }
4409
4410 /* update states */
4411 if (src->state < PFUDPS_SINGLE)
4412 src->state = PFUDPS_SINGLE;
4413 if (dst->state == PFUDPS_SINGLE)
4414 dst->state = PFUDPS_MULTIPLE;
4415
4416 /* update expire time */
4417#ifdef __FreeBSD__
4418 (*state)->expire = time_second;
4419#else
4420 (*state)->expire = time.tv_sec;
4421#endif
4422 if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE)
4423 (*state)->timeout = PFTM_UDP_MULTIPLE;
4424 else
4425 (*state)->timeout = PFTM_UDP_SINGLE;
4426
4427 /* translate source/destination address, if necessary */
4428 if (STATE_TRANSLATE(*state)) {
4429 if (direction == PF_OUT)
4430 pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum,
4431 &uh->uh_sum, &(*state)->gwy.addr,
4432 (*state)->gwy.port, 1, pd->af);
4433 else
4434 pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum,
4435 &uh->uh_sum, &(*state)->lan.addr,
4436 (*state)->lan.port, 1, pd->af);
4437 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
4438 }
4439
4440 return (PF_PASS);
4441}
4442
4443int
4444pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif,
4445 struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4446{
4447 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
4448 u_int16_t icmpid = 0; /* make the compiler happy */
4449 u_int16_t *icmpsum = NULL; /* make the compiler happy */
4450 u_int8_t icmptype = 0; /* make the compiler happy */
4451 int state_icmp = 0;
4452
4453 switch (pd->proto) {
4454#ifdef INET
4455 case IPPROTO_ICMP:
4456 icmptype = pd->hdr.icmp->icmp_type;
4457 icmpid = pd->hdr.icmp->icmp_id;
4458 icmpsum = &pd->hdr.icmp->icmp_cksum;
4459
4460 if (icmptype == ICMP_UNREACH ||
4461 icmptype == ICMP_SOURCEQUENCH ||
4462 icmptype == ICMP_REDIRECT ||
4463 icmptype == ICMP_TIMXCEED ||
4464 icmptype == ICMP_PARAMPROB)
4465 state_icmp++;
4466 break;
4467#endif /* INET */
4468#ifdef INET6
4469 case IPPROTO_ICMPV6:
4470 icmptype = pd->hdr.icmp6->icmp6_type;
4471 icmpid = pd->hdr.icmp6->icmp6_id;
4472 icmpsum = &pd->hdr.icmp6->icmp6_cksum;
4473
4474 if (icmptype == ICMP6_DST_UNREACH ||
4475 icmptype == ICMP6_PACKET_TOO_BIG ||
4476 icmptype == ICMP6_TIME_EXCEEDED ||
4477 icmptype == ICMP6_PARAM_PROB)
4478 state_icmp++;
4479 break;
4480#endif /* INET6 */
4481 }
4482
4483 if (!state_icmp) {
4484
4485 /*
4486 * ICMP query/reply message not related to a TCP/UDP packet.
4487 * Search for an ICMP state.
4488 */
4489 struct pf_state key;
4490
4491 key.af = pd->af;
4492 key.proto = pd->proto;
4493 if (direction == PF_IN) {
4494 PF_ACPY(&key.ext.addr, pd->src, key.af);
4495 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4496 key.ext.port = icmpid;
4497 key.gwy.port = icmpid;
4498 } else {
4499 PF_ACPY(&key.lan.addr, pd->src, key.af);
4500 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4501 key.lan.port = icmpid;
4502 key.ext.port = icmpid;
4503 }
4504
4505 STATE_LOOKUP();
4506
4507#ifdef __FreeBSD__
4508 (*state)->expire = time_second;
4509#else
4510 (*state)->expire = time.tv_sec;
4511#endif
4512 (*state)->timeout = PFTM_ICMP_ERROR_REPLY;
4513
4514 /* translate source/destination address, if necessary */
4515 if (PF_ANEQ(&(*state)->lan.addr, &(*state)->gwy.addr, pd->af)) {
4516 if (direction == PF_OUT) {
4517 switch (pd->af) {
4518#ifdef INET
4519 case AF_INET:
4520 pf_change_a(&saddr->v4.s_addr,
4521 pd->ip_sum,
4522 (*state)->gwy.addr.v4.s_addr, 0);
4523 break;
4524#endif /* INET */
4525#ifdef INET6
4526 case AF_INET6:
4527 pf_change_a6(saddr,
4528 &pd->hdr.icmp6->icmp6_cksum,
4529 &(*state)->gwy.addr, 0);
4530 m_copyback(m, off,
4531 sizeof(struct icmp6_hdr),
4532 (caddr_t)pd->hdr.icmp6);
4533 break;
4534#endif /* INET6 */
4535 }
4536 } else {
4537 switch (pd->af) {
4538#ifdef INET
4539 case AF_INET:
4540 pf_change_a(&daddr->v4.s_addr,
4541 pd->ip_sum,
4542 (*state)->lan.addr.v4.s_addr, 0);
4543 break;
4544#endif /* INET */
4545#ifdef INET6
4546 case AF_INET6:
4547 pf_change_a6(daddr,
4548 &pd->hdr.icmp6->icmp6_cksum,
4549 &(*state)->lan.addr, 0);
4550 m_copyback(m, off,
4551 sizeof(struct icmp6_hdr),
4552 (caddr_t)pd->hdr.icmp6);
4553 break;
4554#endif /* INET6 */
4555 }
4556 }
4557 }
4558
4559 return (PF_PASS);
4560
4561 } else {
4562 /*
4563 * ICMP error message in response to a TCP/UDP packet.
4564 * Extract the inner TCP/UDP header and search for that state.
4565 */
4566
4567 struct pf_pdesc pd2;
4568#ifdef INET
4569 struct ip h2;
4570#endif /* INET */
4571#ifdef INET6
4572 struct ip6_hdr h2_6;
4573 int terminal = 0;
4574#endif /* INET6 */
4575 int ipoff2 = 0; /* make the compiler happy */
4576 int off2 = 0; /* make the compiler happy */
4577
4578 pd2.af = pd->af;
4579 switch (pd->af) {
4580#ifdef INET
4581 case AF_INET:
4582 /* offset of h2 in mbuf chain */
4583 ipoff2 = off + ICMP_MINLEN;
4584
4585 if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2),
4586 NULL, NULL, pd2.af)) {
4587 DPFPRINTF(PF_DEBUG_MISC,
4588 ("pf: ICMP error message too short "
4589 "(ip)\n"));
4590 return (PF_DROP);
4591 }
4592 /*
4593 * ICMP error messages don't refer to non-first
4594 * fragments
4595 */
4596 if (h2.ip_off & htons(IP_OFFMASK))
4597 return (PF_DROP);
4598
4599 /* offset of protocol header that follows h2 */
4600 off2 = ipoff2 + (h2.ip_hl << 2);
4601
4602 pd2.proto = h2.ip_p;
4603 pd2.src = (struct pf_addr *)&h2.ip_src;
4604 pd2.dst = (struct pf_addr *)&h2.ip_dst;
4605 pd2.ip_sum = &h2.ip_sum;
4606 break;
4607#endif /* INET */
4608#ifdef INET6
4609 case AF_INET6:
4610 ipoff2 = off + sizeof(struct icmp6_hdr);
4611
4612 if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6),
4613 NULL, NULL, pd2.af)) {
4614 DPFPRINTF(PF_DEBUG_MISC,
4615 ("pf: ICMP error message too short "
4616 "(ip6)\n"));
4617 return (PF_DROP);
4618 }
4619 pd2.proto = h2_6.ip6_nxt;
4620 pd2.src = (struct pf_addr *)&h2_6.ip6_src;
4621 pd2.dst = (struct pf_addr *)&h2_6.ip6_dst;
4622 pd2.ip_sum = NULL;
4623 off2 = ipoff2 + sizeof(h2_6);
4624 do {
4625 switch (pd2.proto) {
4626 case IPPROTO_FRAGMENT:
4627 /*
4628 * ICMPv6 error messages for
4629 * non-first fragments
4630 */
4631 return (PF_DROP);
4632 case IPPROTO_AH:
4633 case IPPROTO_HOPOPTS:
4634 case IPPROTO_ROUTING:
4635 case IPPROTO_DSTOPTS: {
4636 /* get next header and header length */
4637 struct ip6_ext opt6;
4638
4639 if (!pf_pull_hdr(m, off2, &opt6,
4640 sizeof(opt6), NULL, NULL, pd2.af)) {
4641 DPFPRINTF(PF_DEBUG_MISC,
4642 ("pf: ICMPv6 short opt\n"));
4643 return (PF_DROP);
4644 }
4645 if (pd2.proto == IPPROTO_AH)
4646 off2 += (opt6.ip6e_len + 2) * 4;
4647 else
4648 off2 += (opt6.ip6e_len + 1) * 8;
4649 pd2.proto = opt6.ip6e_nxt;
4650 /* goto the next header */
4651 break;
4652 }
4653 default:
4654 terminal++;
4655 break;
4656 }
4657 } while (!terminal);
4658 break;
4659#endif /* INET6 */
4660 }
4661
4662 switch (pd2.proto) {
4663 case IPPROTO_TCP: {
4664 struct tcphdr th;
4665 u_int32_t seq;
4666 struct pf_state key;
4667 struct pf_state_peer *src, *dst;
4668 u_int8_t dws;
4669 int copyback = 0;
4670
4671 /*
4672 * Only the first 8 bytes of the TCP header can be
4673 * expected. Don't access any TCP header fields after
4674 * th_seq, an ackskew test is not possible.
4675 */
4676 if (!pf_pull_hdr(m, off2, &th, 8, NULL, NULL, pd2.af)) {
4677 DPFPRINTF(PF_DEBUG_MISC,
4678 ("pf: ICMP error message too short "
4679 "(tcp)\n"));
4680 return (PF_DROP);
4681 }
4682
4683 key.af = pd2.af;
4684 key.proto = IPPROTO_TCP;
4685 if (direction == PF_IN) {
4686 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4687 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4688 key.ext.port = th.th_dport;
4689 key.gwy.port = th.th_sport;
4690 } else {
4691 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4692 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4693 key.lan.port = th.th_dport;
4694 key.ext.port = th.th_sport;
4695 }
4696
4697 STATE_LOOKUP();
4698
4699 if (direction == (*state)->direction) {
4700 src = &(*state)->dst;
4701 dst = &(*state)->src;
4702 } else {
4703 src = &(*state)->src;
4704 dst = &(*state)->dst;
4705 }
4706
4707 if (src->wscale && dst->wscale &&
4708 !(th.th_flags & TH_SYN))
4709 dws = dst->wscale & PF_WSCALE_MASK;
4710 else
4711 dws = 0;
4712
4713 /* Demodulate sequence number */
4714 seq = ntohl(th.th_seq) - src->seqdiff;
4715 if (src->seqdiff) {
4716 pf_change_a(&th.th_seq, icmpsum,
4717 htonl(seq), 0);
4718 copyback = 1;
4719 }
4720
4721 if (!SEQ_GEQ(src->seqhi, seq) ||
4722 !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws))) {
4723 if (pf_status.debug >= PF_DEBUG_MISC) {
4724 printf("pf: BAD ICMP %d:%d ",
4725 icmptype, pd->hdr.icmp->icmp_code);
4726 pf_print_host(pd->src, 0, pd->af);
4727 printf(" -> ");
4728 pf_print_host(pd->dst, 0, pd->af);
4729 printf(" state: ");
4730 pf_print_state(*state);
4731 printf(" seq=%u\n", seq);
4732 }
4733 return (PF_DROP);
4734 }
4735
4736 if (STATE_TRANSLATE(*state)) {
4737 if (direction == PF_IN) {
4738 pf_change_icmp(pd2.src, &th.th_sport,
4739 daddr, &(*state)->lan.addr,
4740 (*state)->lan.port, NULL,
4741 pd2.ip_sum, icmpsum,
4742 pd->ip_sum, 0, pd2.af);
4743 } else {
4744 pf_change_icmp(pd2.dst, &th.th_dport,
4745 saddr, &(*state)->gwy.addr,
4746 (*state)->gwy.port, NULL,
4747 pd2.ip_sum, icmpsum,
4748 pd->ip_sum, 0, pd2.af);
4749 }
4750 copyback = 1;
4751 }
4752
4753 if (copyback) {
4754 switch (pd2.af) {
4755#ifdef INET
4756 case AF_INET:
4757 m_copyback(m, off, ICMP_MINLEN,
4758 (caddr_t)pd->hdr.icmp);
4759 m_copyback(m, ipoff2, sizeof(h2),
4760 (caddr_t)&h2);
4761 break;
4762#endif /* INET */
4763#ifdef INET6
4764 case AF_INET6:
4765 m_copyback(m, off,
4766 sizeof(struct icmp6_hdr),
4767 (caddr_t)pd->hdr.icmp6);
4768 m_copyback(m, ipoff2, sizeof(h2_6),
4769 (caddr_t)&h2_6);
4770 break;
4771#endif /* INET6 */
4772 }
4773 m_copyback(m, off2, 8, (caddr_t)&th);
4774 }
4775
4776 return (PF_PASS);
4777 break;
4778 }
4779 case IPPROTO_UDP: {
4780 struct udphdr uh;
4781 struct pf_state key;
4782
4783 if (!pf_pull_hdr(m, off2, &uh, sizeof(uh),
4784 NULL, NULL, pd2.af)) {
4785 DPFPRINTF(PF_DEBUG_MISC,
4786 ("pf: ICMP error message too short "
4787 "(udp)\n"));
4788 return (PF_DROP);
4789 }
4790
4791 key.af = pd2.af;
4792 key.proto = IPPROTO_UDP;
4793 if (direction == PF_IN) {
4794 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4795 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4796 key.ext.port = uh.uh_dport;
4797 key.gwy.port = uh.uh_sport;
4798 } else {
4799 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4800 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4801 key.lan.port = uh.uh_dport;
4802 key.ext.port = uh.uh_sport;
4803 }
4804
4805 STATE_LOOKUP();
4806
4807 if (STATE_TRANSLATE(*state)) {
4808 if (direction == PF_IN) {
4809 pf_change_icmp(pd2.src, &uh.uh_sport,
4810 daddr, &(*state)->lan.addr,
4811 (*state)->lan.port, &uh.uh_sum,
4812 pd2.ip_sum, icmpsum,
4813 pd->ip_sum, 1, pd2.af);
4814 } else {
4815 pf_change_icmp(pd2.dst, &uh.uh_dport,
4816 saddr, &(*state)->gwy.addr,
4817 (*state)->gwy.port, &uh.uh_sum,
4818 pd2.ip_sum, icmpsum,
4819 pd->ip_sum, 1, pd2.af);
4820 }
4821 switch (pd2.af) {
4822#ifdef INET
4823 case AF_INET:
4824 m_copyback(m, off, ICMP_MINLEN,
4825 (caddr_t)pd->hdr.icmp);
4826 m_copyback(m, ipoff2, sizeof(h2),
4827 (caddr_t)&h2);
4828 break;
4829#endif /* INET */
4830#ifdef INET6
4831 case AF_INET6:
4832 m_copyback(m, off,
4833 sizeof(struct icmp6_hdr),
4834 (caddr_t)pd->hdr.icmp6);
4835 m_copyback(m, ipoff2, sizeof(h2_6),
4836 (caddr_t)&h2_6);
4837 break;
4838#endif /* INET6 */
4839 }
4840 m_copyback(m, off2, sizeof(uh),
4841 (caddr_t)&uh);
4842 }
4843
4844 return (PF_PASS);
4845 break;
4846 }
4847#ifdef INET
4848 case IPPROTO_ICMP: {
4849 struct icmp iih;
4850 struct pf_state key;
4851
4852 if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN,
4853 NULL, NULL, pd2.af)) {
4854 DPFPRINTF(PF_DEBUG_MISC,
4855 ("pf: ICMP error message too short i"
4856 "(icmp)\n"));
4857 return (PF_DROP);
4858 }
4859
4860 key.af = pd2.af;
4861 key.proto = IPPROTO_ICMP;
4862 if (direction == PF_IN) {
4863 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4864 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4865 key.ext.port = iih.icmp_id;
4866 key.gwy.port = iih.icmp_id;
4867 } else {
4868 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4869 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4870 key.lan.port = iih.icmp_id;
4871 key.ext.port = iih.icmp_id;
4872 }
4873
4874 STATE_LOOKUP();
4875
4876 if (STATE_TRANSLATE(*state)) {
4877 if (direction == PF_IN) {
4878 pf_change_icmp(pd2.src, &iih.icmp_id,
4879 daddr, &(*state)->lan.addr,
4880 (*state)->lan.port, NULL,
4881 pd2.ip_sum, icmpsum,
4882 pd->ip_sum, 0, AF_INET);
4883 } else {
4884 pf_change_icmp(pd2.dst, &iih.icmp_id,
4885 saddr, &(*state)->gwy.addr,
4886 (*state)->gwy.port, NULL,
4887 pd2.ip_sum, icmpsum,
4888 pd->ip_sum, 0, AF_INET);
4889 }
4890 m_copyback(m, off, ICMP_MINLEN,
4891 (caddr_t)pd->hdr.icmp);
4892 m_copyback(m, ipoff2, sizeof(h2),
4893 (caddr_t)&h2);
4894 m_copyback(m, off2, ICMP_MINLEN,
4895 (caddr_t)&iih);
4896 }
4897
4898 return (PF_PASS);
4899 break;
4900 }
4901#endif /* INET */
4902#ifdef INET6
4903 case IPPROTO_ICMPV6: {
4904 struct icmp6_hdr iih;
4905 struct pf_state key;
4906
4907 if (!pf_pull_hdr(m, off2, &iih,
4908 sizeof(struct icmp6_hdr), NULL, NULL, pd2.af)) {
4909 DPFPRINTF(PF_DEBUG_MISC,
4910 ("pf: ICMP error message too short "
4911 "(icmp6)\n"));
4912 return (PF_DROP);
4913 }
4914
4915 key.af = pd2.af;
4916 key.proto = IPPROTO_ICMPV6;
4917 if (direction == PF_IN) {
4918 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4919 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4920 key.ext.port = iih.icmp6_id;
4921 key.gwy.port = iih.icmp6_id;
4922 } else {
4923 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4924 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4925 key.lan.port = iih.icmp6_id;
4926 key.ext.port = iih.icmp6_id;
4927 }
4928
4929 STATE_LOOKUP();
4930
4931 if (STATE_TRANSLATE(*state)) {
4932 if (direction == PF_IN) {
4933 pf_change_icmp(pd2.src, &iih.icmp6_id,
4934 daddr, &(*state)->lan.addr,
4935 (*state)->lan.port, NULL,
4936 pd2.ip_sum, icmpsum,
4937 pd->ip_sum, 0, AF_INET6);
4938 } else {
4939 pf_change_icmp(pd2.dst, &iih.icmp6_id,
4940 saddr, &(*state)->gwy.addr,
4941 (*state)->gwy.port, NULL,
4942 pd2.ip_sum, icmpsum,
4943 pd->ip_sum, 0, AF_INET6);
4944 }
4945 m_copyback(m, off, sizeof(struct icmp6_hdr),
4946 (caddr_t)pd->hdr.icmp6);
4947 m_copyback(m, ipoff2, sizeof(h2_6),
4948 (caddr_t)&h2_6);
4949 m_copyback(m, off2, sizeof(struct icmp6_hdr),
4950 (caddr_t)&iih);
4951 }
4952
4953 return (PF_PASS);
4954 break;
4955 }
4956#endif /* INET6 */
4957 default: {
4958 struct pf_state key;
4959
4960 key.af = pd2.af;
4961 key.proto = pd2.proto;
4962 if (direction == PF_IN) {
4963 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4964 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4965 key.ext.port = 0;
4966 key.gwy.port = 0;
4967 } else {
4968 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4969 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4970 key.lan.port = 0;
4971 key.ext.port = 0;
4972 }
4973
4974 STATE_LOOKUP();
4975
4976 if (STATE_TRANSLATE(*state)) {
4977 if (direction == PF_IN) {
4978 pf_change_icmp(pd2.src, NULL,
4979 daddr, &(*state)->lan.addr,
4980 0, NULL,
4981 pd2.ip_sum, icmpsum,
4982 pd->ip_sum, 0, pd2.af);
4983 } else {
4984 pf_change_icmp(pd2.dst, NULL,
4985 saddr, &(*state)->gwy.addr,
4986 0, NULL,
4987 pd2.ip_sum, icmpsum,
4988 pd->ip_sum, 0, pd2.af);
4989 }
4990 switch (pd2.af) {
4991#ifdef INET
4992 case AF_INET:
4993 m_copyback(m, off, ICMP_MINLEN,
4994 (caddr_t)pd->hdr.icmp);
4995 m_copyback(m, ipoff2, sizeof(h2),
4996 (caddr_t)&h2);
4997 break;
4998#endif /* INET */
4999#ifdef INET6
5000 case AF_INET6:
5001 m_copyback(m, off,
5002 sizeof(struct icmp6_hdr),
5003 (caddr_t)pd->hdr.icmp6);
5004 m_copyback(m, ipoff2, sizeof(h2_6),
5005 (caddr_t)&h2_6);
5006 break;
5007#endif /* INET6 */
5008 }
5009 }
5010
5011 return (PF_PASS);
5012 break;
5013 }
5014 }
5015 }
5016}
5017
5018int
5019pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif,
5020 struct pf_pdesc *pd)
5021{
5022 struct pf_state_peer *src, *dst;
5023 struct pf_state key;
5024
5025 key.af = pd->af;
5026 key.proto = pd->proto;
5027 if (direction == PF_IN) {
5028 PF_ACPY(&key.ext.addr, pd->src, key.af);
5029 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
5030 key.ext.port = 0;
5031 key.gwy.port = 0;
5032 } else {
5033 PF_ACPY(&key.lan.addr, pd->src, key.af);
5034 PF_ACPY(&key.ext.addr, pd->dst, key.af);
5035 key.lan.port = 0;
5036 key.ext.port = 0;
5037 }
5038
5039 STATE_LOOKUP();
5040
5041 if (direction == (*state)->direction) {
5042 src = &(*state)->src;
5043 dst = &(*state)->dst;
5044 } else {
5045 src = &(*state)->dst;
5046 dst = &(*state)->src;
5047 }
5048
5049 /* update states */
5050 if (src->state < PFOTHERS_SINGLE)
5051 src->state = PFOTHERS_SINGLE;
5052 if (dst->state == PFOTHERS_SINGLE)
5053 dst->state = PFOTHERS_MULTIPLE;
5054
5055 /* update expire time */
5056#ifdef __FreeBSD__
5057 (*state)->expire = time_second;
5058#else
5059 (*state)->expire = time.tv_sec;
5060#endif
5061 if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE)
5062 (*state)->timeout = PFTM_OTHER_MULTIPLE;
5063 else
5064 (*state)->timeout = PFTM_OTHER_SINGLE;
5065
5066 /* translate source/destination address, if necessary */
5067 if (STATE_TRANSLATE(*state)) {
5068 if (direction == PF_OUT)
5069 switch (pd->af) {
5070#ifdef INET
5071 case AF_INET:
5072 pf_change_a(&pd->src->v4.s_addr,
5073 pd->ip_sum, (*state)->gwy.addr.v4.s_addr,
5074 0);
5075 break;
5076#endif /* INET */
5077#ifdef INET6
5078 case AF_INET6:
5079 PF_ACPY(pd->src, &(*state)->gwy.addr, pd->af);
5080 break;
5081#endif /* INET6 */
5082 }
5083 else
5084 switch (pd->af) {
5085#ifdef INET
5086 case AF_INET:
5087 pf_change_a(&pd->dst->v4.s_addr,
5088 pd->ip_sum, (*state)->lan.addr.v4.s_addr,
5089 0);
5090 break;
5091#endif /* INET */
5092#ifdef INET6
5093 case AF_INET6:
5094 PF_ACPY(pd->dst, &(*state)->lan.addr, pd->af);
5095 break;
5096#endif /* INET6 */
5097 }
5098 }
5099
5100 return (PF_PASS);
5101}
5102
5103/*
5104 * ipoff and off are measured from the start of the mbuf chain.
5105 * h must be at "ipoff" on the mbuf chain.
5106 */
5107void *
5108pf_pull_hdr(struct mbuf *m, int off, void *p, int len,
5109 u_short *actionp, u_short *reasonp, sa_family_t af)
5110{
5111 switch (af) {
5112#ifdef INET
5113 case AF_INET: {
5114 struct ip *h = mtod(m, struct ip *);
5115 u_int16_t fragoff = (ntohs(h->ip_off) & IP_OFFMASK) << 3;
5116
5117 if (fragoff) {
5118 if (fragoff >= len)
5119 ACTION_SET(actionp, PF_PASS);
5120 else {
5121 ACTION_SET(actionp, PF_DROP);
5122 REASON_SET(reasonp, PFRES_FRAG);
5123 }
5124 return (NULL);
5125 }
5126 if (m->m_pkthdr.len < off + len ||
5127 ntohs(h->ip_len) < off + len) {
5128 ACTION_SET(actionp, PF_DROP);
5129 REASON_SET(reasonp, PFRES_SHORT);
5130 return (NULL);
5131 }
5132 break;
5133 }
5134#endif /* INET */
5135#ifdef INET6
5136 case AF_INET6: {
5137 struct ip6_hdr *h = mtod(m, struct ip6_hdr *);
5138
5139 if (m->m_pkthdr.len < off + len ||
5140 (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) <
5141 (unsigned)(off + len)) {
5142 ACTION_SET(actionp, PF_DROP);
5143 REASON_SET(reasonp, PFRES_SHORT);
5144 return (NULL);
5145 }
5146 break;
5147 }
5148#endif /* INET6 */
5149 }
5150 m_copydata(m, off, len, p);
5151 return (p);
5152}
5153
5154int
5155pf_routable(struct pf_addr *addr, sa_family_t af)
5156{
5157 struct sockaddr_in *dst;
5158 struct route ro;
5159 int ret = 0;
5160
5161 bzero(&ro, sizeof(ro));
5162 dst = satosin(&ro.ro_dst);
5163 dst->sin_family = af;
5164 dst->sin_len = sizeof(*dst);
5165 dst->sin_addr = addr->v4;
5166#ifdef __FreeBSD__
5167#ifdef RTF_PRCLONING
5168 rtalloc_ign(&ro, (RTF_CLONING|RTF_PRCLONING));
5169#else /* !RTF_PRCLONING */
5170 rtalloc_ign(&ro, RTF_CLONING);
5171#endif
5172#else /* ! __FreeBSD__ */
5173 rtalloc_noclone(&ro, NO_CLONING);
5174#endif
5175
5176 if (ro.ro_rt != NULL) {
5177 ret = 1;
5178 RTFREE(ro.ro_rt);
5179 }
5180
5181 return (ret);
5182}
5183
5184#ifdef INET
5185
5186void
5187pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5188 struct pf_state *s)
5189{
5190 struct mbuf *m0, *m1;
5191 struct m_tag *mtag;
5192 struct route iproute;
5193 struct route *ro = NULL; /* XXX: was uninitialized */
5194 struct sockaddr_in *dst;
5195 struct ip *ip;
5196 struct ifnet *ifp = NULL;
5197 struct pf_addr naddr;
5198 struct pf_src_node *sn = NULL;
5199 int error = 0;
5200#ifdef __FreeBSD__
5201 int sw_csum;
5202#endif
5203
5204 if (m == NULL || *m == NULL || r == NULL ||
5205 (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5206 panic("pf_route: invalid parameters");
5207
5208 if ((mtag = m_tag_find(*m, PACKET_TAG_PF_ROUTED, NULL)) == NULL) {
5209 if ((mtag = m_tag_get(PACKET_TAG_PF_ROUTED, 1, M_NOWAIT)) ==
5210 NULL) {
5211 m0 = *m;
5212 *m = NULL;
5213 goto bad;
5214 }
5215 *(char *)(mtag + 1) = 1;
5216 m_tag_prepend(*m, mtag);
5217 } else {
5218 if (*(char *)(mtag + 1) > 3) {
5219 m0 = *m;
5220 *m = NULL;
5221 goto bad;
5222 }
5223 (*(char *)(mtag + 1))++;
5224 }
5225
5226 if (r->rt == PF_DUPTO) {
5227#ifdef __FreeBSD__
5228 if ((m0 = m_dup(*m, M_DONTWAIT)) == NULL)
5229#else
5230 if ((m0 = m_copym2(*m, 0, M_COPYALL, M_NOWAIT)) == NULL)
5231#endif
5232 return;
5233#ifdef __FreeBSD__
5234 if ((mtag = m_tag_copy(mtag, M_DONTWAIT)) == NULL)
5235#else
5236 if ((mtag = m_tag_copy(mtag)) == NULL)
5237#endif
5238 goto bad;
5239 m_tag_prepend(m0, mtag);
5240 } else {
5241 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5242 return;
5243 m0 = *m;
5244 }
5245
5246 if (m0->m_len < sizeof(struct ip))
5247 panic("pf_route: m0->m_len < sizeof(struct ip)");
5248 ip = mtod(m0, struct ip *);
5249
5250 ro = &iproute;
5251 bzero((caddr_t)ro, sizeof(*ro));
5252 dst = satosin(&ro->ro_dst);
5253 dst->sin_family = AF_INET;
5254 dst->sin_len = sizeof(*dst);
5255 dst->sin_addr = ip->ip_dst;
5256
5257 if (r->rt == PF_FASTROUTE) {
5258 rtalloc(ro);
5259 if (ro->ro_rt == 0) {
5260 ipstat.ips_noroute++;
5261 goto bad;
5262 }
5263
5264 ifp = ro->ro_rt->rt_ifp;
5265 ro->ro_rt->rt_use++;
5266
5267 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
5268 dst = satosin(ro->ro_rt->rt_gateway);
5269 } else {
5270 if (TAILQ_EMPTY(&r->rpool.list))
5271 panic("pf_route: TAILQ_EMPTY(&r->rpool.list)");
5272 if (s == NULL) {
5273 pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src,
5274 &naddr, NULL, &sn);
5275 if (!PF_AZERO(&naddr, AF_INET))
5276 dst->sin_addr.s_addr = naddr.v4.s_addr;
5277 ifp = r->rpool.cur->kif ?
5278 r->rpool.cur->kif->pfik_ifp : NULL;
5279 } else {
5280 if (!PF_AZERO(&s->rt_addr, AF_INET))
5281 dst->sin_addr.s_addr =
5282 s->rt_addr.v4.s_addr;
5283 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5284 }
5285 }
5286 if (ifp == NULL)
5287 goto bad;
5288
5289 if (oifp != ifp) {
5290#ifdef __FreeBSD__
5291 PF_UNLOCK();
5292 if (pf_test(PF_OUT, ifp, &m0, NULL) != PF_PASS) {
5293 PF_LOCK();
5294 goto bad;
5295 } else if (m0 == NULL) {
5296 PF_LOCK();
5297 goto done;
5298 }
5299 PF_LOCK();
5300#else
5301 if (pf_test(PF_OUT, ifp, &m0) != PF_PASS)
5302 goto bad;
5303 else if (m0 == NULL)
5304 goto done;
5305#endif
5306 if (m0->m_len < sizeof(struct ip))
5307 panic("pf_route: m0->m_len < sizeof(struct ip)");
5308 ip = mtod(m0, struct ip *);
5309 }
5310
5311#ifdef __FreeBSD__
5312 /* Copied from FreeBSD 5.1-CURRENT ip_output. */
5313 m0->m_pkthdr.csum_flags |= CSUM_IP;
5314 sw_csum = m0->m_pkthdr.csum_flags & ~ifp->if_hwassist;
5315 if (sw_csum & CSUM_DELAY_DATA) {
5316 /*
5317 * XXX: in_delayed_cksum assumes HBO for ip->ip_len (at least)
5318 */
5319 NTOHS(ip->ip_len);
5320 NTOHS(ip->ip_off); /* XXX: needed? */
5321 in_delayed_cksum(m0);
5322 HTONS(ip->ip_len);
5323 HTONS(ip->ip_off);
5324 sw_csum &= ~CSUM_DELAY_DATA;
5325 }
5326 m0->m_pkthdr.csum_flags &= ifp->if_hwassist;
5327
5328 if (ntohs(ip->ip_len) <= ifp->if_mtu ||
5329 (ifp->if_hwassist & CSUM_FRAGMENT &&
5330 ((ip->ip_off & htons(IP_DF)) == 0))) {
5331 /*
5332 * ip->ip_len = htons(ip->ip_len);
5333 * ip->ip_off = htons(ip->ip_off);
5334 */
5335 ip->ip_sum = 0;
5336 if (sw_csum & CSUM_DELAY_IP) {
5337 /* From KAME */
5338 if (ip->ip_v == IPVERSION &&
5339 (ip->ip_hl << 2) == sizeof(*ip)) {
5340 ip->ip_sum = in_cksum_hdr(ip);
5341 } else {
5342 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
5343 }
5344 }
5345 PF_UNLOCK();
5346 error = (*ifp->if_output)(ifp, m0, sintosa(dst), ro->ro_rt);
5347 PF_LOCK();
5348 goto done;
5349 }
5350
5351#else
5352 /* Copied from ip_output. */
5353#ifdef IPSEC
5354 /*
5355 * If deferred crypto processing is needed, check that the
5356 * interface supports it.
5357 */
5358 if ((mtag = m_tag_find(m0, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL))
5359 != NULL && (ifp->if_capabilities & IFCAP_IPSEC) == 0) {
5360 /* Notify IPsec to do its own crypto. */
5361 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1));
5362 goto bad;
5363 }
5364#endif /* IPSEC */
5365
5366 /* Catch routing changes wrt. hardware checksumming for TCP or UDP. */
5367 if (m0->m_pkthdr.csum & M_TCPV4_CSUM_OUT) {
5368 if (!(ifp->if_capabilities & IFCAP_CSUM_TCPv4) ||
5369 ifp->if_bridge != NULL) {
5370 in_delayed_cksum(m0);
5371 m0->m_pkthdr.csum &= ~M_TCPV4_CSUM_OUT; /* Clear */
5372 }
5373 } else if (m0->m_pkthdr.csum & M_UDPV4_CSUM_OUT) {
5374 if (!(ifp->if_capabilities & IFCAP_CSUM_UDPv4) ||
5375 ifp->if_bridge != NULL) {
5376 in_delayed_cksum(m0);
5377 m0->m_pkthdr.csum &= ~M_UDPV4_CSUM_OUT; /* Clear */
5378 }
5379 }
5380
5381 if (ntohs(ip->ip_len) <= ifp->if_mtu) {
5382 if ((ifp->if_capabilities & IFCAP_CSUM_IPv4) &&
5383 ifp->if_bridge == NULL) {
5384 m0->m_pkthdr.csum |= M_IPV4_CSUM_OUT;
5385 ipstat.ips_outhwcsum++;
5386 } else {
5387 ip->ip_sum = 0;
5388 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
5389 }
5390 /* Update relevant hardware checksum stats for TCP/UDP */
5391 if (m0->m_pkthdr.csum & M_TCPV4_CSUM_OUT)
5392 tcpstat.tcps_outhwcsum++;
5393 else if (m0->m_pkthdr.csum & M_UDPV4_CSUM_OUT)
5394 udpstat.udps_outhwcsum++;
5395 error = (*ifp->if_output)(ifp, m0, sintosa(dst), NULL);
5396 goto done;
5397 }
5398#endif
5399 /*
5400 * Too large for interface; fragment if possible.
5401 * Must be able to put at least 8 bytes per fragment.
5402 */
5403 if (ip->ip_off & htons(IP_DF)) {
5404 ipstat.ips_cantfrag++;
5405 if (r->rt != PF_DUPTO) {
5406#ifdef __FreeBSD__
5407 /* icmp_error() expects host byte ordering */
5408 NTOHS(ip->ip_len);
5409 NTOHS(ip->ip_off);
5410 PF_UNLOCK();
5411#endif
5412 icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0,
5413 ifp);
5414#ifdef __FreeBSD__
5415 PF_LOCK();
5416#endif
5417 goto done;
5418 } else
5419 goto bad;
5420 }
5421
5422 m1 = m0;
5423#ifdef __FreeBSD__
5424 /*
5425 * XXX: is cheaper + less error prone than own function
5426 */
5427 NTOHS(ip->ip_len);
5428 NTOHS(ip->ip_off);
5429 error = ip_fragment(ip, &m0, ifp->if_mtu, ifp->if_hwassist, sw_csum);
5430#else
5431 error = ip_fragment(m0, ifp, ifp->if_mtu);
5432#endif
5433 if (error) {
5434#ifndef __FreeBSD__ /* ip_fragment does not do m_freem() on FreeBSD */
5435 m0 = NULL;
5436#endif
5437 goto bad;
5438 }
5439
5440 for (m0 = m1; m0; m0 = m1) {
5441 m1 = m0->m_nextpkt;
5442 m0->m_nextpkt = 0;
5443#ifdef __FreeBSD__
5444 if (error == 0) {
5445 PF_UNLOCK();
5446 error = (*ifp->if_output)(ifp, m0, sintosa(dst),
5447 NULL);
5448 PF_LOCK();
5449 } else
5450#else
5451 if (error == 0)
5452 error = (*ifp->if_output)(ifp, m0, sintosa(dst),
5453 NULL);
5454 else
5455#endif
5456 m_freem(m0);
5457 }
5458
5459 if (error == 0)
5460 ipstat.ips_fragmented++;
5461
5462done:
5463 if (r->rt != PF_DUPTO)
5464 *m = NULL;
5465 if (ro == &iproute && ro->ro_rt)
5466 RTFREE(ro->ro_rt);
5467 return;
5468
5469bad:
5470 m_freem(m0);
5471 goto done;
5472}
5473#endif /* INET */
5474
5475#ifdef INET6
5476void
5477pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5478 struct pf_state *s)
5479{
5480 struct mbuf *m0;
5481 struct m_tag *mtag;
5482 struct route_in6 ip6route;
5483 struct route_in6 *ro;
5484 struct sockaddr_in6 *dst;
5485 struct ip6_hdr *ip6;
5486 struct ifnet *ifp = NULL;
5487 struct pf_addr naddr;
5488 struct pf_src_node *sn = NULL;
5489 int error = 0;
5490
5491 if (m == NULL || *m == NULL || r == NULL ||
5492 (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5493 panic("pf_route6: invalid parameters");
5494
5495 if ((mtag = m_tag_find(*m, PACKET_TAG_PF_ROUTED, NULL)) == NULL) {
5496 if ((mtag = m_tag_get(PACKET_TAG_PF_ROUTED, 1, M_NOWAIT)) ==
5497 NULL) {
5498 m0 = *m;
5499 *m = NULL;
5500 goto bad;
5501 }
5502 *(char *)(mtag + 1) = 1;
5503 m_tag_prepend(*m, mtag);
5504 } else {
5505 if (*(char *)(mtag + 1) > 3) {
5506 m0 = *m;
5507 *m = NULL;
5508 goto bad;
5509 }
5510 (*(char *)(mtag + 1))++;
5511 }
5512
5513 if (r->rt == PF_DUPTO) {
5514#ifdef __FreeBSD__
5515 if ((m0 = m_dup(*m, M_DONTWAIT)) == NULL)
5516#else
5517 if ((m0 = m_copym2(*m, 0, M_COPYALL, M_NOWAIT)) == NULL)
5518#endif
5519 return;
5520#ifdef __FreeBSD__
5521 if ((mtag = m_tag_copy(mtag, M_DONTWAIT)) == NULL)
5522#else
5523 if ((mtag = m_tag_copy(mtag)) == NULL)
5524#endif
5525 goto bad;
5526 m_tag_prepend(m0, mtag);
5527 } else {
5528 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5529 return;
5530 m0 = *m;
5531 }
5532
5533 if (m0->m_len < sizeof(struct ip6_hdr))
5534 panic("pf_route6: m0->m_len < sizeof(struct ip6_hdr)");
5535 ip6 = mtod(m0, struct ip6_hdr *);
5536
5537 ro = &ip6route;
5538 bzero((caddr_t)ro, sizeof(*ro));
5539 dst = (struct sockaddr_in6 *)&ro->ro_dst;
5540 dst->sin6_family = AF_INET6;
5541 dst->sin6_len = sizeof(*dst);
5542 dst->sin6_addr = ip6->ip6_dst;
5543
5544 /* Cheat. */
5545 if (r->rt == PF_FASTROUTE) {
5546#ifdef __FreeBSD__
5547 m0->m_flags |= M_SKIP_FIREWALL;
5548 PF_UNLOCK();
5549 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
5550 PF_LOCK();
5551#else
5552 mtag = m_tag_get(PACKET_TAG_PF_GENERATED, 0, M_NOWAIT);
5553 if (mtag == NULL)
5554 goto bad;
5555 m_tag_prepend(m0, mtag);
5556 ip6_output(m0, NULL, NULL, 0, NULL, NULL);
5557#endif
5558 return;
5559 }
5560
5561 if (TAILQ_EMPTY(&r->rpool.list))
5562 panic("pf_route6: TAILQ_EMPTY(&r->rpool.list)");
5563 if (s == NULL) {
5564 pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src,
5565 &naddr, NULL, &sn);
5566 if (!PF_AZERO(&naddr, AF_INET6))
5567 PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5568 &naddr, AF_INET6);
5569 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL;
5570 } else {
5571 if (!PF_AZERO(&s->rt_addr, AF_INET6))
5572 PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5573 &s->rt_addr, AF_INET6);
5574 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5575 }
5576 if (ifp == NULL)
5577 goto bad;
5578
5579 if (oifp != ifp) {
5580#ifdef __FreeBSD__
5581 PF_UNLOCK();
5582 if (pf_test6(PF_OUT, ifp, &m0, NULL) != PF_PASS) {
5583 PF_LOCK();
5584 goto bad;
5585 } else if (m0 == NULL) {
5586 PF_LOCK();
5587 goto done;
5588 }
5589 PF_LOCK();
5590#else
5591 if (pf_test6(PF_OUT, ifp, &m0) != PF_PASS)
5592 goto bad;
5593 else if (m0 == NULL)
5594 goto done;
5595#endif
5596 if (m0->m_len < sizeof(struct ip6_hdr))
5597 panic("pf_route6: m0->m_len < sizeof(struct ip6_hdr)");
5598 ip6 = mtod(m0, struct ip6_hdr *);
5599 }
5600
5601 /*
5602 * If the packet is too large for the outgoing interface,
5603 * send back an icmp6 error.
5604 */
5605 if (IN6_IS_ADDR_LINKLOCAL(&dst->sin6_addr))
5606 dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
5607 if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) {
5608#ifdef __FreeBSD__
5609 PF_UNLOCK();
5610#endif
5611 error = nd6_output(ifp, ifp, m0, dst, NULL);
5612#ifdef __FreeBSD__
5613 PF_LOCK();
5614#endif
5615 } else {
5616 in6_ifstat_inc(ifp, ifs6_in_toobig);
5617#ifdef __FreeBSD__
5618 if (r->rt != PF_DUPTO) {
5619 PF_UNLOCK();
5620 icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
5621 PF_LOCK();
5622 } else
5623#else
5624 if (r->rt != PF_DUPTO)
5625 icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
5626 else
5627#endif
5628 goto bad;
5629 }
5630
5631done:
5632 if (r->rt != PF_DUPTO)
5633 *m = NULL;
5634 return;
5635
5636bad:
5637 m_freem(m0);
5638 goto done;
5639}
5640#endif /* INET6 */
5641
5642
5643#ifdef __FreeBSD__
5644/*
5645 * FreeBSD supports cksum offloads for the following drivers.
| 2/* $OpenBSD: pf.c,v 1.433.2.2 2004/07/17 03:22:34 brad Exp $ */
3/* add $OpenBSD: pf.c,v 1.448 2004/05/11 07:34:11 dhartmei Exp $ */
4
5/*
6 * Copyright (c) 2001 Daniel Hartmeier
7 * Copyright (c) 2002,2003 Henning Brauer
8 * All rights reserved.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 *
14 * - Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * - Redistributions in binary form must reproduce the above
17 * copyright notice, this list of conditions and the following
18 * disclaimer in the documentation and/or other materials provided
19 * with the distribution.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
25 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
27 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
29 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
31 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
32 * POSSIBILITY OF SUCH DAMAGE.
33 *
34 * Effort sponsored in part by the Defense Advanced Research Projects
35 * Agency (DARPA) and Air Force Research Laboratory, Air Force
36 * Materiel Command, USAF, under agreement number F30602-01-2-0537.
37 *
38 */
39
40#ifdef __FreeBSD__
41#include "opt_inet.h"
42#include "opt_inet6.h"
43#endif
44
45#ifdef __FreeBSD__
46#include "opt_bpf.h"
47#include "opt_pf.h"
48#define NBPFILTER DEV_BPF
49#define NPFLOG DEV_PFLOG
50#define NPFSYNC DEV_PFSYNC
51#else
52#include "bpfilter.h"
53#include "pflog.h"
54#include "pfsync.h"
55#endif
56
57#include <sys/param.h>
58#include <sys/systm.h>
59#include <sys/mbuf.h>
60#include <sys/filio.h>
61#include <sys/socket.h>
62#include <sys/socketvar.h>
63#include <sys/kernel.h>
64#include <sys/time.h>
65#ifdef __FreeBSD__
66#include <sys/sysctl.h>
67#include <sys/endian.h>
68#else
69#include <sys/pool.h>
70#endif
71
72#include <net/if.h>
73#include <net/if_types.h>
74#include <net/bpf.h>
75#include <net/route.h>
76
77#include <netinet/in.h>
78#include <netinet/in_var.h>
79#include <netinet/in_systm.h>
80#include <netinet/ip.h>
81#include <netinet/ip_var.h>
82#include <netinet/tcp.h>
83#include <netinet/tcp_seq.h>
84#include <netinet/udp.h>
85#include <netinet/ip_icmp.h>
86#include <netinet/in_pcb.h>
87#include <netinet/tcp_timer.h>
88#include <netinet/tcp_var.h>
89#include <netinet/udp_var.h>
90#include <netinet/icmp_var.h>
91
92#ifndef __FreeBSD__
93#include <dev/rndvar.h>
94#endif
95#include <net/pfvar.h>
96#include <net/if_pflog.h>
97
98#if NPFSYNC > 0
99#include <net/if_pfsync.h>
100#endif /* NPFSYNC > 0 */
101
102#ifdef INET6
103#include <netinet/ip6.h>
104#include <netinet/in_pcb.h>
105#include <netinet/icmp6.h>
106#include <netinet6/nd6.h>
107#ifdef __FreeBSD__
108#include <netinet6/ip6_var.h>
109#include <netinet6/in6_pcb.h>
110#endif
111#endif /* INET6 */
112
113#ifdef __FreeBSD__
114#include <machine/in_cksum.h>
115#include <sys/limits.h>
116#include <sys/ucred.h>
117
118extern int ip_optcopy(struct ip *, struct ip *);
119#endif
120
121#define DPFPRINTF(n, x) if (pf_status.debug >= (n)) printf x
122
123/*
124 * Global variables
125 */
126
127struct pf_anchorqueue pf_anchors;
128struct pf_ruleset pf_main_ruleset;
129struct pf_altqqueue pf_altqs[2];
130struct pf_palist pf_pabuf;
131struct pf_altqqueue *pf_altqs_active;
132struct pf_altqqueue *pf_altqs_inactive;
133struct pf_status pf_status;
134
135u_int32_t ticket_altqs_active;
136u_int32_t ticket_altqs_inactive;
137int altqs_inactive_open;
138u_int32_t ticket_pabuf;
139
140#ifdef __FreeBSD__
141struct callout pf_expire_to; /* expire timeout */
142#else
143struct timeout pf_expire_to; /* expire timeout */
144#endif
145
146
147#ifdef __FreeBSD__
148uma_zone_t pf_src_tree_pl, pf_rule_pl;
149uma_zone_t pf_state_pl, pf_altq_pl, pf_pooladdr_pl;
150#else
151struct pool pf_src_tree_pl, pf_rule_pl;
152struct pool pf_state_pl, pf_altq_pl, pf_pooladdr_pl;
153#endif
154
155void pf_print_host(struct pf_addr *, u_int16_t, u_int8_t);
156void pf_print_state(struct pf_state *);
157void pf_print_flags(u_int8_t);
158
159void pf_change_ap(struct pf_addr *, u_int16_t *,
160 u_int16_t *, u_int16_t *, struct pf_addr *,
161 u_int16_t, u_int8_t, sa_family_t);
162#ifdef INET6
163void pf_change_a6(struct pf_addr *, u_int16_t *,
164 struct pf_addr *, u_int8_t);
165#endif /* INET6 */
166void pf_change_icmp(struct pf_addr *, u_int16_t *,
167 struct pf_addr *, struct pf_addr *, u_int16_t,
168 u_int16_t *, u_int16_t *, u_int16_t *,
169 u_int16_t *, u_int8_t, sa_family_t);
170void pf_send_tcp(const struct pf_rule *, sa_family_t,
171 const struct pf_addr *, const struct pf_addr *,
172 u_int16_t, u_int16_t, u_int32_t, u_int32_t,
173 u_int8_t, u_int16_t, u_int16_t, u_int8_t);
174void pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t,
175 sa_family_t, struct pf_rule *);
176struct pf_rule *pf_match_translation(struct pf_pdesc *, struct mbuf *,
177 int, int, struct pfi_kif *,
178 struct pf_addr *, u_int16_t, struct pf_addr *,
179 u_int16_t, int);
180struct pf_rule *pf_get_translation(struct pf_pdesc *, struct mbuf *,
181 int, int, struct pfi_kif *, struct pf_src_node **,
182 struct pf_addr *, u_int16_t,
183 struct pf_addr *, u_int16_t,
184 struct pf_addr *, u_int16_t *);
185int pf_test_tcp(struct pf_rule **, struct pf_state **,
186 int, struct pfi_kif *, struct mbuf *, int,
187 void *, struct pf_pdesc *, struct pf_rule **,
188#ifdef __FreeBSD__
189 struct pf_ruleset **, struct inpcb *);
190#else
191 struct pf_ruleset **);
192#endif
193int pf_test_udp(struct pf_rule **, struct pf_state **,
194 int, struct pfi_kif *, struct mbuf *, int,
195 void *, struct pf_pdesc *, struct pf_rule **,
196#ifdef __FreeBSD__
197 struct pf_ruleset **, struct inpcb *);
198#else
199 struct pf_ruleset **);
200#endif
201int pf_test_icmp(struct pf_rule **, struct pf_state **,
202 int, struct pfi_kif *, struct mbuf *, int,
203 void *, struct pf_pdesc *, struct pf_rule **,
204 struct pf_ruleset **);
205int pf_test_other(struct pf_rule **, struct pf_state **,
206 int, struct pfi_kif *, struct mbuf *, int, void *,
207 struct pf_pdesc *, struct pf_rule **,
208 struct pf_ruleset **);
209int pf_test_fragment(struct pf_rule **, int,
210 struct pfi_kif *, struct mbuf *, void *,
211 struct pf_pdesc *, struct pf_rule **,
212 struct pf_ruleset **);
213int pf_test_state_tcp(struct pf_state **, int,
214 struct pfi_kif *, struct mbuf *, int,
215 void *, struct pf_pdesc *, u_short *);
216int pf_test_state_udp(struct pf_state **, int,
217 struct pfi_kif *, struct mbuf *, int,
218 void *, struct pf_pdesc *);
219int pf_test_state_icmp(struct pf_state **, int,
220 struct pfi_kif *, struct mbuf *, int,
221 void *, struct pf_pdesc *);
222int pf_test_state_other(struct pf_state **, int,
223 struct pfi_kif *, struct pf_pdesc *);
224struct pf_tag *pf_get_tag(struct mbuf *);
225int pf_match_tag(struct mbuf *, struct pf_rule *,
226 struct pf_rule *, struct pf_tag *, int *);
227void pf_hash(struct pf_addr *, struct pf_addr *,
228 struct pf_poolhashkey *, sa_family_t);
229int pf_map_addr(u_int8_t, struct pf_rule *,
230 struct pf_addr *, struct pf_addr *,
231 struct pf_addr *, struct pf_src_node **);
232int pf_get_sport(sa_family_t, u_int8_t, struct pf_rule *,
233 struct pf_addr *, struct pf_addr *, u_int16_t,
234 struct pf_addr *, u_int16_t*, u_int16_t, u_int16_t,
235 struct pf_src_node **);
236void pf_route(struct mbuf **, struct pf_rule *, int,
237 struct ifnet *, struct pf_state *);
238void pf_route6(struct mbuf **, struct pf_rule *, int,
239 struct ifnet *, struct pf_state *);
240#ifdef __FreeBSD__
241int pf_socket_lookup(uid_t *, gid_t *,
242 int, struct pf_pdesc *, struct inpcb *);
243#else
244int pf_socket_lookup(uid_t *, gid_t *,
245 int, struct pf_pdesc *);
246#endif
247u_int8_t pf_get_wscale(struct mbuf *, int, u_int16_t,
248 sa_family_t);
249u_int16_t pf_get_mss(struct mbuf *, int, u_int16_t,
250 sa_family_t);
251u_int16_t pf_calc_mss(struct pf_addr *, sa_family_t,
252 u_int16_t);
253void pf_set_rt_ifp(struct pf_state *,
254 struct pf_addr *);
255int pf_check_proto_cksum(struct mbuf *, int, int,
256 u_int8_t, sa_family_t);
257int pf_addr_wrap_neq(struct pf_addr_wrap *,
258 struct pf_addr_wrap *);
259static int pf_add_mbuf_tag(struct mbuf *, u_int);
260struct pf_state *pf_find_state_recurse(struct pfi_kif *,
261 struct pf_state *, u_int8_t);
262
263#ifdef __FreeBSD__
264int in4_cksum(struct mbuf *m, u_int8_t nxt, int off, int len);
265
266struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX];
267#else
268struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] = {
269 { &pf_state_pl, PFSTATE_HIWAT },
270 { &pf_src_tree_pl, PFSNODE_HIWAT },
271 { &pf_frent_pl, PFFRAG_FRENT_HIWAT }
272};
273#endif
274
275#define STATE_LOOKUP() \
276 do { \
277 if (direction == PF_IN) \
278 *state = pf_find_state_recurse( \
279 kif, &key, PF_EXT_GWY); \
280 else \
281 *state = pf_find_state_recurse( \
282 kif, &key, PF_LAN_EXT); \
283 if (*state == NULL) \
284 return (PF_DROP); \
285 if (direction == PF_OUT && \
286 (((*state)->rule.ptr->rt == PF_ROUTETO && \
287 (*state)->rule.ptr->direction == PF_OUT) || \
288 ((*state)->rule.ptr->rt == PF_REPLYTO && \
289 (*state)->rule.ptr->direction == PF_IN)) && \
290 (*state)->rt_kif != NULL && \
291 (*state)->rt_kif != kif) \
292 return (PF_PASS); \
293 } while (0)
294
295#define STATE_TRANSLATE(s) \
296 (s)->lan.addr.addr32[0] != (s)->gwy.addr.addr32[0] || \
297 ((s)->af == AF_INET6 && \
298 ((s)->lan.addr.addr32[1] != (s)->gwy.addr.addr32[1] || \
299 (s)->lan.addr.addr32[2] != (s)->gwy.addr.addr32[2] || \
300 (s)->lan.addr.addr32[3] != (s)->gwy.addr.addr32[3])) || \
301 (s)->lan.port != (s)->gwy.port
302
303#define BOUND_IFACE(r, k) (((r)->rule_flag & PFRULE_IFBOUND) ? (k) : \
304 ((r)->rule_flag & PFRULE_GRBOUND) ? (k)->pfik_parent : \
305 (k)->pfik_parent->pfik_parent)
306
307#ifndef __FreeBSD__
308static __inline int pf_src_compare(struct pf_src_node *, struct pf_src_node *);
309static __inline int pf_state_compare_lan_ext(struct pf_state *,
310 struct pf_state *);
311static __inline int pf_state_compare_ext_gwy(struct pf_state *,
312 struct pf_state *);
313static __inline int pf_state_compare_id(struct pf_state *,
314 struct pf_state *);
315#else
316static int pf_src_compare(struct pf_src_node *, struct pf_src_node *);
317static int pf_state_compare_lan_ext(struct pf_state *,
318 struct pf_state *);
319static int pf_state_compare_ext_gwy(struct pf_state *,
320 struct pf_state *);
321static int pf_state_compare_id(struct pf_state *,
322 struct pf_state *);
323#endif
324
325struct pf_src_tree tree_src_tracking;
326
327struct pf_state_tree_id tree_id;
328struct pf_state_queue state_updates;
329
330RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare);
331RB_GENERATE(pf_state_tree_lan_ext, pf_state,
332 u.s.entry_lan_ext, pf_state_compare_lan_ext);
333RB_GENERATE(pf_state_tree_ext_gwy, pf_state,
334 u.s.entry_ext_gwy, pf_state_compare_ext_gwy);
335RB_GENERATE(pf_state_tree_id, pf_state,
336 u.s.entry_id, pf_state_compare_id);
337
338#ifdef __FreeBSD__
339static int
340#else
341static __inline int
342#endif
343pf_src_compare(struct pf_src_node *a, struct pf_src_node *b)
344{
345 int diff;
346
347 if (a->rule.ptr > b->rule.ptr)
348 return (1);
349 if (a->rule.ptr < b->rule.ptr)
350 return (-1);
351 if ((diff = a->af - b->af) != 0)
352 return (diff);
353 switch (a->af) {
354#ifdef INET
355 case AF_INET:
356 if (a->addr.addr32[0] > b->addr.addr32[0])
357 return (1);
358 if (a->addr.addr32[0] < b->addr.addr32[0])
359 return (-1);
360 break;
361#endif /* INET */
362#ifdef INET6
363 case AF_INET6:
364 if (a->addr.addr32[3] > b->addr.addr32[3])
365 return (1);
366 if (a->addr.addr32[3] < b->addr.addr32[3])
367 return (-1);
368 if (a->addr.addr32[2] > b->addr.addr32[2])
369 return (1);
370 if (a->addr.addr32[2] < b->addr.addr32[2])
371 return (-1);
372 if (a->addr.addr32[1] > b->addr.addr32[1])
373 return (1);
374 if (a->addr.addr32[1] < b->addr.addr32[1])
375 return (-1);
376 if (a->addr.addr32[0] > b->addr.addr32[0])
377 return (1);
378 if (a->addr.addr32[0] < b->addr.addr32[0])
379 return (-1);
380 break;
381#endif /* INET6 */
382 }
383 return (0);
384}
385
386#ifdef __FreeBSD__
387static int
388#else
389static __inline int
390#endif
391pf_state_compare_lan_ext(struct pf_state *a, struct pf_state *b)
392{
393 int diff;
394
395 if ((diff = a->proto - b->proto) != 0)
396 return (diff);
397 if ((diff = a->af - b->af) != 0)
398 return (diff);
399 switch (a->af) {
400#ifdef INET
401 case AF_INET:
402 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
403 return (1);
404 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
405 return (-1);
406 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
407 return (1);
408 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
409 return (-1);
410 break;
411#endif /* INET */
412#ifdef INET6
413 case AF_INET6:
414 if (a->lan.addr.addr32[3] > b->lan.addr.addr32[3])
415 return (1);
416 if (a->lan.addr.addr32[3] < b->lan.addr.addr32[3])
417 return (-1);
418 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
419 return (1);
420 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
421 return (-1);
422 if (a->lan.addr.addr32[2] > b->lan.addr.addr32[2])
423 return (1);
424 if (a->lan.addr.addr32[2] < b->lan.addr.addr32[2])
425 return (-1);
426 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
427 return (1);
428 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
429 return (-1);
430 if (a->lan.addr.addr32[1] > b->lan.addr.addr32[1])
431 return (1);
432 if (a->lan.addr.addr32[1] < b->lan.addr.addr32[1])
433 return (-1);
434 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
435 return (1);
436 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
437 return (-1);
438 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
439 return (1);
440 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
441 return (-1);
442 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
443 return (1);
444 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
445 return (-1);
446 break;
447#endif /* INET6 */
448 }
449
450 if ((diff = a->lan.port - b->lan.port) != 0)
451 return (diff);
452 if ((diff = a->ext.port - b->ext.port) != 0)
453 return (diff);
454
455 return (0);
456}
457
458#ifdef __FreeBSD__
459static int
460#else
461static __inline int
462#endif
463pf_state_compare_ext_gwy(struct pf_state *a, struct pf_state *b)
464{
465 int diff;
466
467 if ((diff = a->proto - b->proto) != 0)
468 return (diff);
469 if ((diff = a->af - b->af) != 0)
470 return (diff);
471 switch (a->af) {
472#ifdef INET
473 case AF_INET:
474 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
475 return (1);
476 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
477 return (-1);
478 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
479 return (1);
480 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
481 return (-1);
482 break;
483#endif /* INET */
484#ifdef INET6
485 case AF_INET6:
486 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
487 return (1);
488 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
489 return (-1);
490 if (a->gwy.addr.addr32[3] > b->gwy.addr.addr32[3])
491 return (1);
492 if (a->gwy.addr.addr32[3] < b->gwy.addr.addr32[3])
493 return (-1);
494 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
495 return (1);
496 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
497 return (-1);
498 if (a->gwy.addr.addr32[2] > b->gwy.addr.addr32[2])
499 return (1);
500 if (a->gwy.addr.addr32[2] < b->gwy.addr.addr32[2])
501 return (-1);
502 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
503 return (1);
504 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
505 return (-1);
506 if (a->gwy.addr.addr32[1] > b->gwy.addr.addr32[1])
507 return (1);
508 if (a->gwy.addr.addr32[1] < b->gwy.addr.addr32[1])
509 return (-1);
510 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
511 return (1);
512 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
513 return (-1);
514 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
515 return (1);
516 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
517 return (-1);
518 break;
519#endif /* INET6 */
520 }
521
522 if ((diff = a->ext.port - b->ext.port) != 0)
523 return (diff);
524 if ((diff = a->gwy.port - b->gwy.port) != 0)
525 return (diff);
526
527 return (0);
528}
529
530#ifdef __FreeBSD__
531static int
532#else
533static __inline int
534#endif
535pf_state_compare_id(struct pf_state *a, struct pf_state *b)
536{
537 if (a->id > b->id)
538 return (1);
539 if (a->id < b->id)
540 return (-1);
541 if (a->creatorid > b->creatorid)
542 return (1);
543 if (a->creatorid < b->creatorid)
544 return (-1);
545
546 return (0);
547}
548
549#ifdef INET6
550void
551pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af)
552{
553 switch (af) {
554#ifdef INET
555 case AF_INET:
556 dst->addr32[0] = src->addr32[0];
557 break;
558#endif /* INET */
559 case AF_INET6:
560 dst->addr32[0] = src->addr32[0];
561 dst->addr32[1] = src->addr32[1];
562 dst->addr32[2] = src->addr32[2];
563 dst->addr32[3] = src->addr32[3];
564 break;
565 }
566}
567#endif
568
569struct pf_state *
570pf_find_state_byid(struct pf_state *key)
571{
572 pf_status.fcounters[FCNT_STATE_SEARCH]++;
573 return (RB_FIND(pf_state_tree_id, &tree_id, key));
574}
575
576struct pf_state *
577pf_find_state_recurse(struct pfi_kif *kif, struct pf_state *key, u_int8_t tree)
578{
579 struct pf_state *s;
580
581 pf_status.fcounters[FCNT_STATE_SEARCH]++;
582
583 switch (tree) {
584 case PF_LAN_EXT:
585 for (; kif != NULL; kif = kif->pfik_parent) {
586 s = RB_FIND(pf_state_tree_lan_ext,
587 &kif->pfik_lan_ext, key);
588 if (s != NULL)
589 return (s);
590 }
591 return (NULL);
592 case PF_EXT_GWY:
593 for (; kif != NULL; kif = kif->pfik_parent) {
594 s = RB_FIND(pf_state_tree_ext_gwy,
595 &kif->pfik_ext_gwy, key);
596 if (s != NULL)
597 return (s);
598 }
599 return (NULL);
600 default:
601 panic("pf_find_state_recurse");
602 }
603}
604
605struct pf_state *
606pf_find_state_all(struct pf_state *key, u_int8_t tree, int *more)
607{
608 struct pf_state *s, *ss = NULL;
609 struct pfi_kif *kif;
610
611 pf_status.fcounters[FCNT_STATE_SEARCH]++;
612
613 switch (tree) {
614 case PF_LAN_EXT:
615 TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states) {
616 s = RB_FIND(pf_state_tree_lan_ext,
617 &kif->pfik_lan_ext, key);
618 if (s == NULL)
619 continue;
620 if (more == NULL)
621 return (s);
622 ss = s;
623 (*more)++;
624 }
625 return (ss);
626 case PF_EXT_GWY:
627 TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states) {
628 s = RB_FIND(pf_state_tree_ext_gwy,
629 &kif->pfik_ext_gwy, key);
630 if (s == NULL)
631 continue;
632 if (more == NULL)
633 return (s);
634 ss = s;
635 (*more)++;
636 }
637 return (ss);
638 default:
639 panic("pf_find_state_all");
640 }
641}
642
643int
644pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule,
645 struct pf_addr *src, sa_family_t af)
646{
647 struct pf_src_node k;
648
649 if (*sn == NULL) {
650 k.af = af;
651 PF_ACPY(&k.addr, src, af);
652 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
653 rule->rpool.opts & PF_POOL_STICKYADDR)
654 k.rule.ptr = rule;
655 else
656 k.rule.ptr = NULL;
657 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
658 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
659 }
660 if (*sn == NULL) {
661 if (!rule->max_src_nodes ||
662 rule->src_nodes < rule->max_src_nodes)
663 (*sn) = pool_get(&pf_src_tree_pl, PR_NOWAIT);
664 if ((*sn) == NULL)
665 return (-1);
666 bzero(*sn, sizeof(struct pf_src_node));
667 (*sn)->af = af;
668 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
669 rule->rpool.opts & PF_POOL_STICKYADDR)
670 (*sn)->rule.ptr = rule;
671 else
672 (*sn)->rule.ptr = NULL;
673 PF_ACPY(&(*sn)->addr, src, af);
674 if (RB_INSERT(pf_src_tree,
675 &tree_src_tracking, *sn) != NULL) {
676 if (pf_status.debug >= PF_DEBUG_MISC) {
677 printf("pf: src_tree insert failed: ");
678 pf_print_host(&(*sn)->addr, 0, af);
679 printf("\n");
680 }
681 pool_put(&pf_src_tree_pl, *sn);
682 return (-1);
683 }
684#ifdef __FreeBSD__
685 (*sn)->creation = time_second;
686#else
687 (*sn)->creation = time.tv_sec;
688#endif
689 (*sn)->ruletype = rule->action;
690 if ((*sn)->rule.ptr != NULL)
691 (*sn)->rule.ptr->src_nodes++;
692 pf_status.scounters[SCNT_SRC_NODE_INSERT]++;
693 pf_status.src_nodes++;
694 } else {
695 if (rule->max_src_states &&
696 (*sn)->states >= rule->max_src_states)
697 return (-1);
698 }
699 return (0);
700}
701
702int
703pf_insert_state(struct pfi_kif *kif, struct pf_state *state)
704{
705 /* Thou MUST NOT insert multiple duplicate keys */
706 state->u.s.kif = kif;
707 if (RB_INSERT(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state)) {
708 if (pf_status.debug >= PF_DEBUG_MISC) {
709 printf("pf: state insert failed: tree_lan_ext");
710 printf(" lan: ");
711 pf_print_host(&state->lan.addr, state->lan.port,
712 state->af);
713 printf(" gwy: ");
714 pf_print_host(&state->gwy.addr, state->gwy.port,
715 state->af);
716 printf(" ext: ");
717 pf_print_host(&state->ext.addr, state->ext.port,
718 state->af);
719 if (state->sync_flags & PFSTATE_FROMSYNC)
720 printf(" (from sync)");
721 printf("\n");
722 }
723 return (-1);
724 }
725
726 if (RB_INSERT(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, state)) {
727 if (pf_status.debug >= PF_DEBUG_MISC) {
728 printf("pf: state insert failed: tree_ext_gwy");
729 printf(" lan: ");
730 pf_print_host(&state->lan.addr, state->lan.port,
731 state->af);
732 printf(" gwy: ");
733 pf_print_host(&state->gwy.addr, state->gwy.port,
734 state->af);
735 printf(" ext: ");
736 pf_print_host(&state->ext.addr, state->ext.port,
737 state->af);
738 if (state->sync_flags & PFSTATE_FROMSYNC)
739 printf(" (from sync)");
740 printf("\n");
741 }
742 RB_REMOVE(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state);
743 return (-1);
744 }
745
746 if (state->id == 0 && state->creatorid == 0) {
747 state->id = htobe64(pf_status.stateid++);
748 state->creatorid = pf_status.hostid;
749 }
750 if (RB_INSERT(pf_state_tree_id, &tree_id, state) != NULL) {
751 if (pf_status.debug >= PF_DEBUG_MISC) {
752#ifdef __FreeBSD__
753 printf("pf: state insert failed: "
754 "id: %016llx creatorid: %08x",
755 (long long)be64toh(state->id),
756 ntohl(state->creatorid));
757#else
758 printf("pf: state insert failed: "
759 "id: %016llx creatorid: %08x",
760 betoh64(state->id), ntohl(state->creatorid));
761#endif
762 if (state->sync_flags & PFSTATE_FROMSYNC)
763 printf(" (from sync)");
764 printf("\n");
765 }
766 RB_REMOVE(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state);
767 RB_REMOVE(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, state);
768 return (-1);
769 }
770 TAILQ_INSERT_HEAD(&state_updates, state, u.s.entry_updates);
771
772 pf_status.fcounters[FCNT_STATE_INSERT]++;
773 pf_status.states++;
774 pfi_attach_state(kif);
775#if NPFSYNC
776 pfsync_insert_state(state);
777#endif
778 return (0);
779}
780
781void
782pf_purge_timeout(void *arg)
783{
784#ifdef __FreeBSD__
785 struct callout *to = arg;
786#else
787 struct timeout *to = arg;
788#endif
789 int s;
790
791#ifdef __FreeBSD__
792 PF_LOCK();
793#endif
794 s = splsoftnet();
795 pf_purge_expired_states();
796 pf_purge_expired_fragments();
797 pf_purge_expired_src_nodes();
798 splx(s);
799#ifdef __FreeBSD__
800 PF_UNLOCK();
801#endif
802
803#ifdef __FreeBSD__
804 callout_reset(to, pf_default_rule.timeout[PFTM_INTERVAL] * hz,
805 pf_purge_timeout, to);
806#else
807 timeout_add(to, pf_default_rule.timeout[PFTM_INTERVAL] * hz);
808#endif
809}
810
811u_int32_t
812pf_state_expires(const struct pf_state *state)
813{
814 u_int32_t timeout;
815 u_int32_t start;
816 u_int32_t end;
817 u_int32_t states;
818
819 /* handle all PFTM_* > PFTM_MAX here */
820 if (state->timeout == PFTM_PURGE)
821#ifdef __FreeBSD__
822 return (time_second);
823#else
824 return (time.tv_sec);
825#endif
826 if (state->timeout == PFTM_UNTIL_PACKET)
827 return (0);
828#ifdef __FreeBSD__
829 KASSERT((state->timeout < PFTM_MAX),
830 ("pf_state_expires: timeout > PFTM_MAX"));
831#else
832 KASSERT(state->timeout < PFTM_MAX);
833#endif
834 timeout = state->rule.ptr->timeout[state->timeout];
835 if (!timeout)
836 timeout = pf_default_rule.timeout[state->timeout];
837 start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START];
838 if (start) {
839 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END];
840 states = state->rule.ptr->states;
841 } else {
842 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START];
843 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END];
844 states = pf_status.states;
845 }
846 if (end && states > start && start < end) {
847 if (states < end)
848 return (state->expire + timeout * (end - states) /
849 (end - start));
850 else
851#ifdef __FreeBSD__
852 return (time_second);
853#else
854 return (time.tv_sec);
855#endif
856 }
857 return (state->expire + timeout);
858}
859
860void
861pf_purge_expired_src_nodes(void)
862{
863 struct pf_src_node *cur, *next;
864
865 for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) {
866 next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur);
867
868#ifdef __FreeBSD__
869 if (cur->states <= 0 && cur->expire <= time_second) {
870#else
871 if (cur->states <= 0 && cur->expire <= time.tv_sec) {
872#endif
873 if (cur->rule.ptr != NULL) {
874 cur->rule.ptr->src_nodes--;
875 if (cur->rule.ptr->states <= 0 &&
876 cur->rule.ptr->max_src_nodes <= 0)
877 pf_rm_rule(NULL, cur->rule.ptr);
878 }
879 RB_REMOVE(pf_src_tree, &tree_src_tracking, cur);
880 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
881 pf_status.src_nodes--;
882 pool_put(&pf_src_tree_pl, cur);
883 }
884 }
885}
886
887void
888pf_src_tree_remove_state(struct pf_state *s)
889{
890 u_int32_t timeout;
891
892 if (s->src_node != NULL) {
893 if (--s->src_node->states <= 0) {
894 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
895 if (!timeout)
896 timeout =
897 pf_default_rule.timeout[PFTM_SRC_NODE];
898#ifdef __FreeBSD__
899 s->src_node->expire = time_second + timeout;
900#else
901 s->src_node->expire = time.tv_sec + timeout;
902#endif
903 }
904 }
905 if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) {
906 if (--s->nat_src_node->states <= 0) {
907 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
908 if (!timeout)
909 timeout =
910 pf_default_rule.timeout[PFTM_SRC_NODE];
911#ifdef __FreeBSD__
912 s->nat_src_node->expire = time_second + timeout;
913#else
914 s->nat_src_node->expire = time.tv_sec + timeout;
915#endif
916 }
917 }
918 s->src_node = s->nat_src_node = NULL;
919}
920
921void
922pf_purge_expired_states(void)
923{
924 struct pf_state *cur, *next;
925
926 for (cur = RB_MIN(pf_state_tree_id, &tree_id);
927 cur; cur = next) {
928 next = RB_NEXT(pf_state_tree_id, &tree_id, cur);
929
930#ifdef __FreeBSD__
931 if (pf_state_expires(cur) <= time_second) {
932#else
933 if (pf_state_expires(cur) <= time.tv_sec) {
934#endif
935 if (cur->src.state == PF_TCPS_PROXY_DST)
936 pf_send_tcp(cur->rule.ptr, cur->af,
937 &cur->ext.addr, &cur->lan.addr,
938 cur->ext.port, cur->lan.port,
939 cur->src.seqhi, cur->src.seqlo + 1, 0,
940 TH_RST|TH_ACK, 0, 0);
941 RB_REMOVE(pf_state_tree_ext_gwy,
942 &cur->u.s.kif->pfik_ext_gwy, cur);
943 RB_REMOVE(pf_state_tree_lan_ext,
944 &cur->u.s.kif->pfik_lan_ext, cur);
945 RB_REMOVE(pf_state_tree_id, &tree_id, cur);
946#if NPFSYNC
947 pfsync_delete_state(cur);
948#endif
949 pf_src_tree_remove_state(cur);
950 if (--cur->rule.ptr->states <= 0 &&
951 cur->rule.ptr->src_nodes <= 0)
952 pf_rm_rule(NULL, cur->rule.ptr);
953 if (cur->nat_rule.ptr != NULL)
954 if (--cur->nat_rule.ptr->states <= 0 &&
955 cur->nat_rule.ptr->src_nodes <= 0)
956 pf_rm_rule(NULL, cur->nat_rule.ptr);
957 if (cur->anchor.ptr != NULL)
958 if (--cur->anchor.ptr->states <= 0)
959 pf_rm_rule(NULL, cur->anchor.ptr);
960 pf_normalize_tcp_cleanup(cur);
961 pfi_detach_state(cur->u.s.kif);
962 TAILQ_REMOVE(&state_updates, cur, u.s.entry_updates);
963 pool_put(&pf_state_pl, cur);
964 pf_status.fcounters[FCNT_STATE_REMOVALS]++;
965 pf_status.states--;
966 }
967 }
968}
969
970int
971pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw)
972{
973 if (aw->type != PF_ADDR_TABLE)
974 return (0);
975 if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL)
976 return (1);
977 return (0);
978}
979
980void
981pf_tbladdr_remove(struct pf_addr_wrap *aw)
982{
983 if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL)
984 return;
985 pfr_detach_table(aw->p.tbl);
986 aw->p.tbl = NULL;
987}
988
989void
990pf_tbladdr_copyout(struct pf_addr_wrap *aw)
991{
992 struct pfr_ktable *kt = aw->p.tbl;
993
994 if (aw->type != PF_ADDR_TABLE || kt == NULL)
995 return;
996 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
997 kt = kt->pfrkt_root;
998 aw->p.tbl = NULL;
999 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
1000 kt->pfrkt_cnt : -1;
1001}
1002
1003void
1004pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af)
1005{
1006 switch (af) {
1007#ifdef INET
1008 case AF_INET: {
1009 u_int32_t a = ntohl(addr->addr32[0]);
1010 printf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255,
1011 (a>>8)&255, a&255);
1012 if (p) {
1013 p = ntohs(p);
1014 printf(":%u", p);
1015 }
1016 break;
1017 }
1018#endif /* INET */
1019#ifdef INET6
1020 case AF_INET6: {
1021 u_int16_t b;
1022 u_int8_t i, curstart = 255, curend = 0,
1023 maxstart = 0, maxend = 0;
1024 for (i = 0; i < 8; i++) {
1025 if (!addr->addr16[i]) {
1026 if (curstart == 255)
1027 curstart = i;
1028 else
1029 curend = i;
1030 } else {
1031 if (curstart) {
1032 if ((curend - curstart) >
1033 (maxend - maxstart)) {
1034 maxstart = curstart;
1035 maxend = curend;
1036 curstart = 255;
1037 }
1038 }
1039 }
1040 }
1041 for (i = 0; i < 8; i++) {
1042 if (i >= maxstart && i <= maxend) {
1043 if (maxend != 7) {
1044 if (i == maxstart)
1045 printf(":");
1046 } else {
1047 if (i == maxend)
1048 printf(":");
1049 }
1050 } else {
1051 b = ntohs(addr->addr16[i]);
1052 printf("%x", b);
1053 if (i < 7)
1054 printf(":");
1055 }
1056 }
1057 if (p) {
1058 p = ntohs(p);
1059 printf("[%u]", p);
1060 }
1061 break;
1062 }
1063#endif /* INET6 */
1064 }
1065}
1066
1067void
1068pf_print_state(struct pf_state *s)
1069{
1070 switch (s->proto) {
1071 case IPPROTO_TCP:
1072 printf("TCP ");
1073 break;
1074 case IPPROTO_UDP:
1075 printf("UDP ");
1076 break;
1077 case IPPROTO_ICMP:
1078 printf("ICMP ");
1079 break;
1080 case IPPROTO_ICMPV6:
1081 printf("ICMPV6 ");
1082 break;
1083 default:
1084 printf("%u ", s->proto);
1085 break;
1086 }
1087 pf_print_host(&s->lan.addr, s->lan.port, s->af);
1088 printf(" ");
1089 pf_print_host(&s->gwy.addr, s->gwy.port, s->af);
1090 printf(" ");
1091 pf_print_host(&s->ext.addr, s->ext.port, s->af);
1092 printf(" [lo=%u high=%u win=%u modulator=%u", s->src.seqlo,
1093 s->src.seqhi, s->src.max_win, s->src.seqdiff);
1094 if (s->src.wscale && s->dst.wscale)
1095 printf(" wscale=%u", s->src.wscale & PF_WSCALE_MASK);
1096 printf("]");
1097 printf(" [lo=%u high=%u win=%u modulator=%u", s->dst.seqlo,
1098 s->dst.seqhi, s->dst.max_win, s->dst.seqdiff);
1099 if (s->src.wscale && s->dst.wscale)
1100 printf(" wscale=%u", s->dst.wscale & PF_WSCALE_MASK);
1101 printf("]");
1102 printf(" %u:%u", s->src.state, s->dst.state);
1103}
1104
1105void
1106pf_print_flags(u_int8_t f)
1107{
1108 if (f)
1109 printf(" ");
1110 if (f & TH_FIN)
1111 printf("F");
1112 if (f & TH_SYN)
1113 printf("S");
1114 if (f & TH_RST)
1115 printf("R");
1116 if (f & TH_PUSH)
1117 printf("P");
1118 if (f & TH_ACK)
1119 printf("A");
1120 if (f & TH_URG)
1121 printf("U");
1122 if (f & TH_ECE)
1123 printf("E");
1124 if (f & TH_CWR)
1125 printf("W");
1126}
1127
1128#define PF_SET_SKIP_STEPS(i) \
1129 do { \
1130 while (head[i] != cur) { \
1131 head[i]->skip[i].ptr = cur; \
1132 head[i] = TAILQ_NEXT(head[i], entries); \
1133 } \
1134 } while (0)
1135
1136void
1137pf_calc_skip_steps(struct pf_rulequeue *rules)
1138{
1139 struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT];
1140 int i;
1141
1142 cur = TAILQ_FIRST(rules);
1143 prev = cur;
1144 for (i = 0; i < PF_SKIP_COUNT; ++i)
1145 head[i] = cur;
1146 while (cur != NULL) {
1147
1148 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
1149 PF_SET_SKIP_STEPS(PF_SKIP_IFP);
1150 if (cur->direction != prev->direction)
1151 PF_SET_SKIP_STEPS(PF_SKIP_DIR);
1152 if (cur->af != prev->af)
1153 PF_SET_SKIP_STEPS(PF_SKIP_AF);
1154 if (cur->proto != prev->proto)
1155 PF_SET_SKIP_STEPS(PF_SKIP_PROTO);
1156 if (cur->src.not != prev->src.not ||
1157 pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr))
1158 PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR);
1159 if (cur->src.port[0] != prev->src.port[0] ||
1160 cur->src.port[1] != prev->src.port[1] ||
1161 cur->src.port_op != prev->src.port_op)
1162 PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
1163 if (cur->dst.not != prev->dst.not ||
1164 pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr))
1165 PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR);
1166 if (cur->dst.port[0] != prev->dst.port[0] ||
1167 cur->dst.port[1] != prev->dst.port[1] ||
1168 cur->dst.port_op != prev->dst.port_op)
1169 PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
1170
1171 prev = cur;
1172 cur = TAILQ_NEXT(cur, entries);
1173 }
1174 for (i = 0; i < PF_SKIP_COUNT; ++i)
1175 PF_SET_SKIP_STEPS(i);
1176}
1177
1178int
1179pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2)
1180{
1181 if (aw1->type != aw2->type)
1182 return (1);
1183 switch (aw1->type) {
1184 case PF_ADDR_ADDRMASK:
1185 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0))
1186 return (1);
1187 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0))
1188 return (1);
1189 return (0);
1190 case PF_ADDR_DYNIFTL:
1191 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt);
1192 case PF_ADDR_NOROUTE:
1193 return (0);
1194 case PF_ADDR_TABLE:
1195 return (aw1->p.tbl != aw2->p.tbl);
1196 default:
1197 printf("invalid address type: %d\n", aw1->type);
1198 return (1);
1199 }
1200}
1201
1202void
1203pf_update_anchor_rules()
1204{
1205 struct pf_rule *rule;
1206 int i;
1207
1208 for (i = 0; i < PF_RULESET_MAX; ++i)
1209 TAILQ_FOREACH(rule, pf_main_ruleset.rules[i].active.ptr,
1210 entries)
1211 if (rule->anchorname[0])
1212 rule->anchor = pf_find_anchor(rule->anchorname);
1213 else
1214 rule->anchor = NULL;
1215}
1216
1217u_int16_t
1218pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp)
1219{
1220 u_int32_t l;
1221
1222 if (udp && !cksum)
1223 return (0x0000);
1224 l = cksum + old - new;
1225 l = (l >> 16) + (l & 65535);
1226 l = l & 65535;
1227 if (udp && !l)
1228 return (0xFFFF);
1229 return (l);
1230}
1231
1232void
1233pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc,
1234 struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af)
1235{
1236 struct pf_addr ao;
1237 u_int16_t po = *p;
1238
1239 PF_ACPY(&ao, a, af);
1240 PF_ACPY(a, an, af);
1241
1242 *p = pn;
1243
1244 switch (af) {
1245#ifdef INET
1246 case AF_INET:
1247 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1248 ao.addr16[0], an->addr16[0], 0),
1249 ao.addr16[1], an->addr16[1], 0);
1250 *p = pn;
1251 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1252 ao.addr16[0], an->addr16[0], u),
1253 ao.addr16[1], an->addr16[1], u),
1254 po, pn, u);
1255 break;
1256#endif /* INET */
1257#ifdef INET6
1258 case AF_INET6:
1259 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1260 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1261 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1262 ao.addr16[0], an->addr16[0], u),
1263 ao.addr16[1], an->addr16[1], u),
1264 ao.addr16[2], an->addr16[2], u),
1265 ao.addr16[3], an->addr16[3], u),
1266 ao.addr16[4], an->addr16[4], u),
1267 ao.addr16[5], an->addr16[5], u),
1268 ao.addr16[6], an->addr16[6], u),
1269 ao.addr16[7], an->addr16[7], u),
1270 po, pn, u);
1271 break;
1272#endif /* INET6 */
1273 }
1274}
1275
1276
1277/* Changes a u_int32_t. Uses a void * so there are no align restrictions */
1278void
1279pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u)
1280{
1281 u_int32_t ao;
1282
1283 memcpy(&ao, a, sizeof(ao));
1284 memcpy(a, &an, sizeof(u_int32_t));
1285 *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u),
1286 ao % 65536, an % 65536, u);
1287}
1288
1289#ifdef INET6
1290void
1291pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u)
1292{
1293 struct pf_addr ao;
1294
1295 PF_ACPY(&ao, a, AF_INET6);
1296 PF_ACPY(a, an, AF_INET6);
1297
1298 *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1299 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1300 pf_cksum_fixup(pf_cksum_fixup(*c,
1301 ao.addr16[0], an->addr16[0], u),
1302 ao.addr16[1], an->addr16[1], u),
1303 ao.addr16[2], an->addr16[2], u),
1304 ao.addr16[3], an->addr16[3], u),
1305 ao.addr16[4], an->addr16[4], u),
1306 ao.addr16[5], an->addr16[5], u),
1307 ao.addr16[6], an->addr16[6], u),
1308 ao.addr16[7], an->addr16[7], u);
1309}
1310#endif /* INET6 */
1311
1312void
1313pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa,
1314 struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c,
1315 u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af)
1316{
1317 struct pf_addr oia, ooa;
1318
1319 PF_ACPY(&oia, ia, af);
1320 PF_ACPY(&ooa, oa, af);
1321
1322 /* Change inner protocol port, fix inner protocol checksum. */
1323 if (ip != NULL) {
1324 u_int16_t oip = *ip;
1325 u_int32_t opc = 0; /* make the compiler happy */
1326
1327 if (pc != NULL)
1328 opc = *pc;
1329 *ip = np;
1330 if (pc != NULL)
1331 *pc = pf_cksum_fixup(*pc, oip, *ip, u);
1332 *ic = pf_cksum_fixup(*ic, oip, *ip, 0);
1333 if (pc != NULL)
1334 *ic = pf_cksum_fixup(*ic, opc, *pc, 0);
1335 }
1336 /* Change inner ip address, fix inner ip and icmp checksums. */
1337 PF_ACPY(ia, na, af);
1338 switch (af) {
1339#ifdef INET
1340 case AF_INET: {
1341 u_int32_t oh2c = *h2c;
1342
1343 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c,
1344 oia.addr16[0], ia->addr16[0], 0),
1345 oia.addr16[1], ia->addr16[1], 0);
1346 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1347 oia.addr16[0], ia->addr16[0], 0),
1348 oia.addr16[1], ia->addr16[1], 0);
1349 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0);
1350 break;
1351 }
1352#endif /* INET */
1353#ifdef INET6
1354 case AF_INET6:
1355 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1356 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1357 pf_cksum_fixup(pf_cksum_fixup(*ic,
1358 oia.addr16[0], ia->addr16[0], u),
1359 oia.addr16[1], ia->addr16[1], u),
1360 oia.addr16[2], ia->addr16[2], u),
1361 oia.addr16[3], ia->addr16[3], u),
1362 oia.addr16[4], ia->addr16[4], u),
1363 oia.addr16[5], ia->addr16[5], u),
1364 oia.addr16[6], ia->addr16[6], u),
1365 oia.addr16[7], ia->addr16[7], u);
1366 break;
1367#endif /* INET6 */
1368 }
1369 /* Change outer ip address, fix outer ip or icmpv6 checksum. */
1370 PF_ACPY(oa, na, af);
1371 switch (af) {
1372#ifdef INET
1373 case AF_INET:
1374 *hc = pf_cksum_fixup(pf_cksum_fixup(*hc,
1375 ooa.addr16[0], oa->addr16[0], 0),
1376 ooa.addr16[1], oa->addr16[1], 0);
1377 break;
1378#endif /* INET */
1379#ifdef INET6
1380 case AF_INET6:
1381 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1382 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1383 pf_cksum_fixup(pf_cksum_fixup(*ic,
1384 ooa.addr16[0], oa->addr16[0], u),
1385 ooa.addr16[1], oa->addr16[1], u),
1386 ooa.addr16[2], oa->addr16[2], u),
1387 ooa.addr16[3], oa->addr16[3], u),
1388 ooa.addr16[4], oa->addr16[4], u),
1389 ooa.addr16[5], oa->addr16[5], u),
1390 ooa.addr16[6], oa->addr16[6], u),
1391 ooa.addr16[7], oa->addr16[7], u);
1392 break;
1393#endif /* INET6 */
1394 }
1395}
1396
1397void
1398pf_send_tcp(const struct pf_rule *r, sa_family_t af,
1399 const struct pf_addr *saddr, const struct pf_addr *daddr,
1400 u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack,
1401 u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl)
1402{
1403 struct mbuf *m;
1404#ifdef ALTQ
1405 struct m_tag *mtag;
1406#endif
1407 int len = 0, tlen; /* make the compiler happy */
1408#ifdef INET
1409 struct ip *h = NULL; /* make the compiler happy */
1410#endif /* INET */
1411#ifdef INET6
1412 struct ip6_hdr *h6 = NULL; /* make the compiler happy */
1413#endif /* INET6 */
1414 struct tcphdr *th = NULL; /* make the compiler happy */
1415#ifdef __FreeBSD__
1416 struct ip *ip;
1417#endif
1418 char *opt;
1419
1420 /* maximum segment size tcp option */
1421 tlen = sizeof(struct tcphdr);
1422 if (mss)
1423 tlen += 4;
1424
1425 switch (af) {
1426#ifdef INET
1427 case AF_INET:
1428 len = sizeof(struct ip) + tlen;
1429 break;
1430#endif /* INET */
1431#ifdef INET6
1432 case AF_INET6:
1433 len = sizeof(struct ip6_hdr) + tlen;
1434 break;
1435#endif /* INET6 */
1436 }
1437
1438 /* create outgoing mbuf */
1439#ifdef __FreeBSD__
1440 m = m_gethdr(M_DONTWAIT, MT_HEADER);
1441 if (m == NULL)
1442 return;
1443 m->m_flags |= M_SKIP_FIREWALL;
1444#else
1445 mtag = m_tag_get(PACKET_TAG_PF_GENERATED, 0, M_NOWAIT);
1446 if (mtag == NULL)
1447 return;
1448 m = m_gethdr(M_DONTWAIT, MT_HEADER);
1449 if (m == NULL) {
1450 m_tag_free(mtag);
1451 return;
1452 }
1453 m_tag_prepend(m, mtag);
1454#endif
1455#ifdef ALTQ
1456 if (r != NULL && r->qid) {
1457 struct altq_tag *atag;
1458
1459 mtag = m_tag_get(PACKET_TAG_PF_QID, sizeof(*atag), M_NOWAIT);
1460 if (mtag != NULL) {
1461 atag = (struct altq_tag *)(mtag + 1);
1462 atag->qid = r->qid;
1463 /* add hints for ecn */
1464 atag->af = af;
1465 atag->hdr = mtod(m, struct ip *);
1466 m_tag_prepend(m, mtag);
1467 }
1468 }
1469#endif
1470 m->m_data += max_linkhdr;
1471 m->m_pkthdr.len = m->m_len = len;
1472 m->m_pkthdr.rcvif = NULL;
1473 bzero(m->m_data, len);
1474 switch (af) {
1475#ifdef INET
1476 case AF_INET:
1477 h = mtod(m, struct ip *);
1478
1479 /* IP header fields included in the TCP checksum */
1480 h->ip_p = IPPROTO_TCP;
1481 h->ip_len = htons(tlen);
1482 h->ip_src.s_addr = saddr->v4.s_addr;
1483 h->ip_dst.s_addr = daddr->v4.s_addr;
1484
1485 th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip));
1486 break;
1487#endif /* INET */
1488#ifdef INET6
1489 case AF_INET6:
1490 h6 = mtod(m, struct ip6_hdr *);
1491
1492 /* IP header fields included in the TCP checksum */
1493 h6->ip6_nxt = IPPROTO_TCP;
1494 h6->ip6_plen = htons(tlen);
1495 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr));
1496 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr));
1497
1498 th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr));
1499 break;
1500#endif /* INET6 */
1501 }
1502
1503 /* TCP header */
1504 th->th_sport = sport;
1505 th->th_dport = dport;
1506 th->th_seq = htonl(seq);
1507 th->th_ack = htonl(ack);
1508 th->th_off = tlen >> 2;
1509 th->th_flags = flags;
1510 th->th_win = htons(win);
1511
1512 if (mss) {
1513 opt = (char *)(th + 1);
1514 opt[0] = TCPOPT_MAXSEG;
1515 opt[1] = 4;
1516 HTONS(mss);
1517 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2);
1518 }
1519
1520 switch (af) {
1521#ifdef INET
1522 case AF_INET:
1523 /* TCP checksum */
1524 th->th_sum = in_cksum(m, len);
1525
1526 /* Finish the IP header */
1527 h->ip_v = 4;
1528 h->ip_hl = sizeof(*h) >> 2;
1529 h->ip_tos = IPTOS_LOWDELAY;
1530#ifdef __FreeBSD__
1531 h->ip_off = path_mtu_discovery ? IP_DF : 0;
1532 h->ip_len = len;
1533#else
1534 h->ip_off = htons(ip_mtudisc ? IP_DF : 0);
1535 h->ip_len = htons(len);
1536#endif
1537 h->ip_ttl = ttl ? ttl : ip_defttl;
1538 h->ip_sum = 0;
1539#ifdef __FreeBSD__
1540 ip = mtod(m, struct ip *);
1541 PF_UNLOCK();
1542 ip_output(m, (void *)NULL, (void *)NULL, 0, (void *)NULL,
1543 (void *)NULL);
1544 PF_LOCK();
1545#else /* ! __FreeBSD__ */
1546 ip_output(m, (void *)NULL, (void *)NULL, 0, (void *)NULL,
1547 (void *)NULL);
1548#endif
1549 break;
1550#endif /* INET */
1551#ifdef INET6
1552 case AF_INET6:
1553 /* TCP checksum */
1554 th->th_sum = in6_cksum(m, IPPROTO_TCP,
1555 sizeof(struct ip6_hdr), tlen);
1556
1557 h6->ip6_vfc |= IPV6_VERSION;
1558 h6->ip6_hlim = IPV6_DEFHLIM;
1559
1560#ifdef __FreeBSD__
1561 PF_UNLOCK();
1562 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
1563 PF_LOCK();
1564#else
1565 ip6_output(m, NULL, NULL, 0, NULL, NULL);
1566#endif
1567 break;
1568#endif /* INET6 */
1569 }
1570}
1571
1572void
1573pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af,
1574 struct pf_rule *r)
1575{
1576#ifdef ALTQ
1577 struct m_tag *mtag;
1578#endif
1579 struct mbuf *m0;
1580#ifdef __FreeBSD__
1581 struct ip *ip;
1582#endif
1583
1584#ifdef __FreeBSD__
1585 m0 = m_copypacket(m, M_DONTWAIT);
1586 if (m0 == NULL)
1587 return;
1588 m0->m_flags |= M_SKIP_FIREWALL;
1589#else
1590 mtag = m_tag_get(PACKET_TAG_PF_GENERATED, 0, M_NOWAIT);
1591 if (mtag == NULL)
1592 return;
1593 m0 = m_copy(m, 0, M_COPYALL);
1594 if (m0 == NULL) {
1595 m_tag_free(mtag);
1596 return;
1597 }
1598 m_tag_prepend(m0, mtag);
1599#endif
1600
1601#ifdef ALTQ
1602 if (r->qid) {
1603 struct altq_tag *atag;
1604
1605 mtag = m_tag_get(PACKET_TAG_PF_QID, sizeof(*atag), M_NOWAIT);
1606 if (mtag != NULL) {
1607 atag = (struct altq_tag *)(mtag + 1);
1608 atag->qid = r->qid;
1609 /* add hints for ecn */
1610 atag->af = af;
1611 atag->hdr = mtod(m0, struct ip *);
1612 m_tag_prepend(m0, mtag);
1613 }
1614 }
1615#endif
1616
1617 switch (af) {
1618#ifdef INET
1619 case AF_INET:
1620#ifdef __FreeBSD__
1621 /* icmp_error() expects host byte ordering */
1622 ip = mtod(m0, struct ip *);
1623 NTOHS(ip->ip_len);
1624 NTOHS(ip->ip_off);
1625 PF_UNLOCK();
1626#endif
1627 icmp_error(m0, type, code, 0, (void *)NULL);
1628#ifdef __FreeBSD__
1629 PF_LOCK();
1630#endif
1631 break;
1632#endif /* INET */
1633#ifdef INET6
1634 case AF_INET6:
1635#ifdef __FreeBSD__
1636 PF_UNLOCK();
1637#endif
1638 icmp6_error(m0, type, code, 0);
1639#ifdef __FreeBSD__
1640 PF_LOCK();
1641#endif
1642 break;
1643#endif /* INET6 */
1644 }
1645}
1646
1647/*
1648 * Return 1 if the addresses a and b match (with mask m), otherwise return 0.
1649 * If n is 0, they match if they are equal. If n is != 0, they match if they
1650 * are different.
1651 */
1652int
1653pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m,
1654 struct pf_addr *b, sa_family_t af)
1655{
1656 int match = 0;
1657
1658 switch (af) {
1659#ifdef INET
1660 case AF_INET:
1661 if ((a->addr32[0] & m->addr32[0]) ==
1662 (b->addr32[0] & m->addr32[0]))
1663 match++;
1664 break;
1665#endif /* INET */
1666#ifdef INET6
1667 case AF_INET6:
1668 if (((a->addr32[0] & m->addr32[0]) ==
1669 (b->addr32[0] & m->addr32[0])) &&
1670 ((a->addr32[1] & m->addr32[1]) ==
1671 (b->addr32[1] & m->addr32[1])) &&
1672 ((a->addr32[2] & m->addr32[2]) ==
1673 (b->addr32[2] & m->addr32[2])) &&
1674 ((a->addr32[3] & m->addr32[3]) ==
1675 (b->addr32[3] & m->addr32[3])))
1676 match++;
1677 break;
1678#endif /* INET6 */
1679 }
1680 if (match) {
1681 if (n)
1682 return (0);
1683 else
1684 return (1);
1685 } else {
1686 if (n)
1687 return (1);
1688 else
1689 return (0);
1690 }
1691}
1692
1693int
1694pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p)
1695{
1696 switch (op) {
1697 case PF_OP_IRG:
1698 return ((p > a1) && (p < a2));
1699 case PF_OP_XRG:
1700 return ((p < a1) || (p > a2));
1701 case PF_OP_RRG:
1702 return ((p >= a1) && (p <= a2));
1703 case PF_OP_EQ:
1704 return (p == a1);
1705 case PF_OP_NE:
1706 return (p != a1);
1707 case PF_OP_LT:
1708 return (p < a1);
1709 case PF_OP_LE:
1710 return (p <= a1);
1711 case PF_OP_GT:
1712 return (p > a1);
1713 case PF_OP_GE:
1714 return (p >= a1);
1715 }
1716 return (0); /* never reached */
1717}
1718
1719int
1720pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p)
1721{
1722 NTOHS(a1);
1723 NTOHS(a2);
1724 NTOHS(p);
1725 return (pf_match(op, a1, a2, p));
1726}
1727
1728int
1729pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u)
1730{
1731 if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
1732 return (0);
1733 return (pf_match(op, a1, a2, u));
1734}
1735
1736int
1737pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g)
1738{
1739 if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
1740 return (0);
1741 return (pf_match(op, a1, a2, g));
1742}
1743
1744struct pf_tag *
1745pf_get_tag(struct mbuf *m)
1746{
1747 struct m_tag *mtag;
1748
1749 if ((mtag = m_tag_find(m, PACKET_TAG_PF_TAG, NULL)) != NULL)
1750 return ((struct pf_tag *)(mtag + 1));
1751 else
1752 return (NULL);
1753}
1754
1755int
1756pf_match_tag(struct mbuf *m, struct pf_rule *r, struct pf_rule *nat_rule,
1757 struct pf_tag *pftag, int *tag)
1758{
1759 if (*tag == -1) { /* find mbuf tag */
1760 pftag = pf_get_tag(m);
1761 if (pftag != NULL)
1762 *tag = pftag->tag;
1763 else
1764 *tag = 0;
1765 if (nat_rule != NULL && nat_rule->tag)
1766 *tag = nat_rule->tag;
1767 }
1768
1769 return ((!r->match_tag_not && r->match_tag == *tag) ||
1770 (r->match_tag_not && r->match_tag != *tag));
1771}
1772
1773int
1774pf_tag_packet(struct mbuf *m, struct pf_tag *pftag, int tag)
1775{
1776 struct m_tag *mtag;
1777
1778 if (tag <= 0)
1779 return (0);
1780
1781 if (pftag == NULL) {
1782 mtag = m_tag_get(PACKET_TAG_PF_TAG, sizeof(*pftag), M_NOWAIT);
1783 if (mtag == NULL)
1784 return (1);
1785 ((struct pf_tag *)(mtag + 1))->tag = tag;
1786 m_tag_prepend(m, mtag);
1787 } else
1788 pftag->tag = tag;
1789
1790 return (0);
1791}
1792
1793#define PF_STEP_INTO_ANCHOR(r, a, s, n) \
1794 do { \
1795 if ((r) == NULL || (r)->anchor == NULL || \
1796 (s) != NULL || (a) != NULL) \
1797 panic("PF_STEP_INTO_ANCHOR"); \
1798 (a) = (r); \
1799 (s) = TAILQ_FIRST(&(r)->anchor->rulesets); \
1800 (r) = NULL; \
1801 while ((s) != NULL && ((r) = \
1802 TAILQ_FIRST((s)->rules[n].active.ptr)) == NULL) \
1803 (s) = TAILQ_NEXT((s), entries); \
1804 if ((r) == NULL) { \
1805 (r) = TAILQ_NEXT((a), entries); \
1806 (a) = NULL; \
1807 } \
1808 } while (0)
1809
1810#define PF_STEP_OUT_OF_ANCHOR(r, a, s, n) \
1811 do { \
1812 if ((r) != NULL || (a) == NULL || (s) == NULL) \
1813 panic("PF_STEP_OUT_OF_ANCHOR"); \
1814 (s) = TAILQ_NEXT((s), entries); \
1815 while ((s) != NULL && ((r) = \
1816 TAILQ_FIRST((s)->rules[n].active.ptr)) == NULL) \
1817 (s) = TAILQ_NEXT((s), entries); \
1818 if ((r) == NULL) { \
1819 (r) = TAILQ_NEXT((a), entries); \
1820 (a) = NULL; \
1821 } \
1822 } while (0)
1823
1824#ifdef INET6
1825void
1826pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr,
1827 struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af)
1828{
1829 switch (af) {
1830#ifdef INET
1831 case AF_INET:
1832 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
1833 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
1834 break;
1835#endif /* INET */
1836 case AF_INET6:
1837 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
1838 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
1839 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) |
1840 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]);
1841 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) |
1842 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]);
1843 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) |
1844 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]);
1845 break;
1846 }
1847}
1848
1849void
1850pf_addr_inc(struct pf_addr *addr, sa_family_t af)
1851{
1852 switch (af) {
1853#ifdef INET
1854 case AF_INET:
1855 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1);
1856 break;
1857#endif /* INET */
1858 case AF_INET6:
1859 if (addr->addr32[3] == 0xffffffff) {
1860 addr->addr32[3] = 0;
1861 if (addr->addr32[2] == 0xffffffff) {
1862 addr->addr32[2] = 0;
1863 if (addr->addr32[1] == 0xffffffff) {
1864 addr->addr32[1] = 0;
1865 addr->addr32[0] =
1866 htonl(ntohl(addr->addr32[0]) + 1);
1867 } else
1868 addr->addr32[1] =
1869 htonl(ntohl(addr->addr32[1]) + 1);
1870 } else
1871 addr->addr32[2] =
1872 htonl(ntohl(addr->addr32[2]) + 1);
1873 } else
1874 addr->addr32[3] =
1875 htonl(ntohl(addr->addr32[3]) + 1);
1876 break;
1877 }
1878}
1879#endif /* INET6 */
1880
1881#define mix(a,b,c) \
1882 do { \
1883 a -= b; a -= c; a ^= (c >> 13); \
1884 b -= c; b -= a; b ^= (a << 8); \
1885 c -= a; c -= b; c ^= (b >> 13); \
1886 a -= b; a -= c; a ^= (c >> 12); \
1887 b -= c; b -= a; b ^= (a << 16); \
1888 c -= a; c -= b; c ^= (b >> 5); \
1889 a -= b; a -= c; a ^= (c >> 3); \
1890 b -= c; b -= a; b ^= (a << 10); \
1891 c -= a; c -= b; c ^= (b >> 15); \
1892 } while (0)
1893
1894/*
1895 * hash function based on bridge_hash in if_bridge.c
1896 */
1897void
1898pf_hash(struct pf_addr *inaddr, struct pf_addr *hash,
1899 struct pf_poolhashkey *key, sa_family_t af)
1900{
1901 u_int32_t a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0];
1902
1903 switch (af) {
1904#ifdef INET
1905 case AF_INET:
1906 a += inaddr->addr32[0];
1907 b += key->key32[1];
1908 mix(a, b, c);
1909 hash->addr32[0] = c + key->key32[2];
1910 break;
1911#endif /* INET */
1912#ifdef INET6
1913 case AF_INET6:
1914 a += inaddr->addr32[0];
1915 b += inaddr->addr32[2];
1916 mix(a, b, c);
1917 hash->addr32[0] = c;
1918 a += inaddr->addr32[1];
1919 b += inaddr->addr32[3];
1920 c += key->key32[1];
1921 mix(a, b, c);
1922 hash->addr32[1] = c;
1923 a += inaddr->addr32[2];
1924 b += inaddr->addr32[1];
1925 c += key->key32[2];
1926 mix(a, b, c);
1927 hash->addr32[2] = c;
1928 a += inaddr->addr32[3];
1929 b += inaddr->addr32[0];
1930 c += key->key32[3];
1931 mix(a, b, c);
1932 hash->addr32[3] = c;
1933 break;
1934#endif /* INET6 */
1935 }
1936}
1937
1938int
1939pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr,
1940 struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn)
1941{
1942 unsigned char hash[16];
1943 struct pf_pool *rpool = &r->rpool;
1944 struct pf_addr *raddr = &rpool->cur->addr.v.a.addr;
1945 struct pf_addr *rmask = &rpool->cur->addr.v.a.mask;
1946 struct pf_pooladdr *acur = rpool->cur;
1947 struct pf_src_node k;
1948
1949 if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR &&
1950 (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
1951 k.af = af;
1952 PF_ACPY(&k.addr, saddr, af);
1953 if (r->rule_flag & PFRULE_RULESRCTRACK ||
1954 r->rpool.opts & PF_POOL_STICKYADDR)
1955 k.rule.ptr = r;
1956 else
1957 k.rule.ptr = NULL;
1958 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
1959 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
1960 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) {
1961 PF_ACPY(naddr, &(*sn)->raddr, af);
1962 if (pf_status.debug >= PF_DEBUG_MISC) {
1963 printf("pf_map_addr: src tracking maps ");
1964 pf_print_host(&k.addr, 0, af);
1965 printf(" to ");
1966 pf_print_host(naddr, 0, af);
1967 printf("\n");
1968 }
1969 return (0);
1970 }
1971 }
1972
1973 if (rpool->cur->addr.type == PF_ADDR_NOROUTE)
1974 return (1);
1975 if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
1976 if (af == AF_INET) {
1977 if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 &&
1978 (rpool->opts & PF_POOL_TYPEMASK) !=
1979 PF_POOL_ROUNDROBIN)
1980 return (1);
1981 raddr = &rpool->cur->addr.p.dyn->pfid_addr4;
1982 rmask = &rpool->cur->addr.p.dyn->pfid_mask4;
1983 } else {
1984 if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 &&
1985 (rpool->opts & PF_POOL_TYPEMASK) !=
1986 PF_POOL_ROUNDROBIN)
1987 return (1);
1988 raddr = &rpool->cur->addr.p.dyn->pfid_addr6;
1989 rmask = &rpool->cur->addr.p.dyn->pfid_mask6;
1990 }
1991 } else if (rpool->cur->addr.type == PF_ADDR_TABLE) {
1992 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN)
1993 return (1); /* unsupported */
1994 } else {
1995 raddr = &rpool->cur->addr.v.a.addr;
1996 rmask = &rpool->cur->addr.v.a.mask;
1997 }
1998
1999 switch (rpool->opts & PF_POOL_TYPEMASK) {
2000 case PF_POOL_NONE:
2001 PF_ACPY(naddr, raddr, af);
2002 break;
2003 case PF_POOL_BITMASK:
2004 PF_POOLMASK(naddr, raddr, rmask, saddr, af);
2005 break;
2006 case PF_POOL_RANDOM:
2007 if (init_addr != NULL && PF_AZERO(init_addr, af)) {
2008 switch (af) {
2009#ifdef INET
2010 case AF_INET:
2011 rpool->counter.addr32[0] = arc4random();
2012 break;
2013#endif /* INET */
2014#ifdef INET6
2015 case AF_INET6:
2016 if (rmask->addr32[3] != 0xffffffff)
2017 rpool->counter.addr32[3] = arc4random();
2018 else
2019 break;
2020 if (rmask->addr32[2] != 0xffffffff)
2021 rpool->counter.addr32[2] = arc4random();
2022 else
2023 break;
2024 if (rmask->addr32[1] != 0xffffffff)
2025 rpool->counter.addr32[1] = arc4random();
2026 else
2027 break;
2028 if (rmask->addr32[0] != 0xffffffff)
2029 rpool->counter.addr32[0] = arc4random();
2030 break;
2031#endif /* INET6 */
2032 }
2033 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2034 PF_ACPY(init_addr, naddr, af);
2035
2036 } else {
2037 PF_AINC(&rpool->counter, af);
2038 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2039 }
2040 break;
2041 case PF_POOL_SRCHASH:
2042 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af);
2043 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af);
2044 break;
2045 case PF_POOL_ROUNDROBIN:
2046 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2047 if (!pfr_pool_get(rpool->cur->addr.p.tbl,
2048 &rpool->tblidx, &rpool->counter,
2049 &raddr, &rmask, af))
2050 goto get_addr;
2051 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2052 if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2053 &rpool->tblidx, &rpool->counter,
2054 &raddr, &rmask, af))
2055 goto get_addr;
2056 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af))
2057 goto get_addr;
2058
2059 try_next:
2060 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL)
2061 rpool->cur = TAILQ_FIRST(&rpool->list);
2062 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2063 rpool->tblidx = -1;
2064 if (pfr_pool_get(rpool->cur->addr.p.tbl,
2065 &rpool->tblidx, &rpool->counter,
2066 &raddr, &rmask, af)) {
2067 /* table contains no address of type 'af' */
2068 if (rpool->cur != acur)
2069 goto try_next;
2070 return (1);
2071 }
2072 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2073 rpool->tblidx = -1;
2074 if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2075 &rpool->tblidx, &rpool->counter,
2076 &raddr, &rmask, af)) {
2077 /* table contains no address of type 'af' */
2078 if (rpool->cur != acur)
2079 goto try_next;
2080 return (1);
2081 }
2082 } else {
2083 raddr = &rpool->cur->addr.v.a.addr;
2084 rmask = &rpool->cur->addr.v.a.mask;
2085 PF_ACPY(&rpool->counter, raddr, af);
2086 }
2087
2088 get_addr:
2089 PF_ACPY(naddr, &rpool->counter, af);
2090 PF_AINC(&rpool->counter, af);
2091 break;
2092 }
2093 if (*sn != NULL)
2094 PF_ACPY(&(*sn)->raddr, naddr, af);
2095
2096 if (pf_status.debug >= PF_DEBUG_MISC &&
2097 (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2098 printf("pf_map_addr: selected address ");
2099 pf_print_host(naddr, 0, af);
2100 printf("\n");
2101 }
2102
2103 return (0);
2104}
2105
2106int
2107pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r,
2108 struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport,
2109 struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high,
2110 struct pf_src_node **sn)
2111{
2112 struct pf_state key;
2113 struct pf_addr init_addr;
2114 u_int16_t cut;
2115
2116 bzero(&init_addr, sizeof(init_addr));
2117 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2118 return (1);
2119
2120 do {
2121 key.af = af;
2122 key.proto = proto;
2123 PF_ACPY(&key.ext.addr, daddr, key.af);
2124 PF_ACPY(&key.gwy.addr, naddr, key.af);
2125 key.ext.port = dport;
2126
2127 /*
2128 * port search; start random, step;
2129 * similar 2 portloop in in_pcbbind
2130 */
2131 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP)) {
2132 key.gwy.port = 0;
2133 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
2134 return (0);
2135 } else if (low == 0 && high == 0) {
2136 key.gwy.port = *nport;
2137 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
2138 return (0);
2139 } else if (low == high) {
2140 key.gwy.port = htons(low);
2141 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) {
2142 *nport = htons(low);
2143 return (0);
2144 }
2145 } else {
2146 u_int16_t tmp;
2147
2148 if (low > high) {
2149 tmp = low;
2150 low = high;
2151 high = tmp;
2152 }
2153 /* low < high */
2154 cut = arc4random() % (1 + high - low) + low;
2155 /* low <= cut <= high */
2156 for (tmp = cut; tmp <= high; ++(tmp)) {
2157 key.gwy.port = htons(tmp);
2158 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
2159 NULL) {
2160 *nport = htons(tmp);
2161 return (0);
2162 }
2163 }
2164 for (tmp = cut - 1; tmp >= low; --(tmp)) {
2165 key.gwy.port = htons(tmp);
2166 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
2167 NULL) {
2168 *nport = htons(tmp);
2169 return (0);
2170 }
2171 }
2172 }
2173
2174 switch (r->rpool.opts & PF_POOL_TYPEMASK) {
2175 case PF_POOL_RANDOM:
2176 case PF_POOL_ROUNDROBIN:
2177 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2178 return (1);
2179 break;
2180 case PF_POOL_NONE:
2181 case PF_POOL_SRCHASH:
2182 case PF_POOL_BITMASK:
2183 default:
2184 return (1);
2185 }
2186 } while (! PF_AEQ(&init_addr, naddr, af) );
2187
2188 return (1); /* none available */
2189}
2190
2191struct pf_rule *
2192pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off,
2193 int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport,
2194 struct pf_addr *daddr, u_int16_t dport, int rs_num)
2195{
2196 struct pf_rule *r, *rm = NULL, *anchorrule = NULL;
2197 struct pf_ruleset *ruleset = NULL;
2198
2199 r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr);
2200 while (r && rm == NULL) {
2201 struct pf_rule_addr *src = NULL, *dst = NULL;
2202 struct pf_addr_wrap *xdst = NULL;
2203
2204 if (r->action == PF_BINAT && direction == PF_IN) {
2205 src = &r->dst;
2206 if (r->rpool.cur != NULL)
2207 xdst = &r->rpool.cur->addr;
2208 } else {
2209 src = &r->src;
2210 dst = &r->dst;
2211 }
2212
2213 r->evaluations++;
2214 if (r->kif != NULL &&
2215 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
2216 r = r->skip[PF_SKIP_IFP].ptr;
2217 else if (r->direction && r->direction != direction)
2218 r = r->skip[PF_SKIP_DIR].ptr;
2219 else if (r->af && r->af != pd->af)
2220 r = r->skip[PF_SKIP_AF].ptr;
2221 else if (r->proto && r->proto != pd->proto)
2222 r = r->skip[PF_SKIP_PROTO].ptr;
2223 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af, src->not))
2224 r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR :
2225 PF_SKIP_DST_ADDR].ptr;
2226 else if (src->port_op && !pf_match_port(src->port_op,
2227 src->port[0], src->port[1], sport))
2228 r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT :
2229 PF_SKIP_DST_PORT].ptr;
2230 else if (dst != NULL &&
2231 PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->not))
2232 r = r->skip[PF_SKIP_DST_ADDR].ptr;
2233 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af, 0))
2234 r = TAILQ_NEXT(r, entries);
2235 else if (dst != NULL && dst->port_op &&
2236 !pf_match_port(dst->port_op, dst->port[0],
2237 dst->port[1], dport))
2238 r = r->skip[PF_SKIP_DST_PORT].ptr;
2239 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto !=
2240 IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m,
2241 off, pd->hdr.tcp), r->os_fingerprint)))
2242 r = TAILQ_NEXT(r, entries);
2243 else if (r->anchorname[0] && r->anchor == NULL)
2244 r = TAILQ_NEXT(r, entries);
2245 else if (r->anchor == NULL)
2246 rm = r;
2247 else
2248 PF_STEP_INTO_ANCHOR(r, anchorrule, ruleset, rs_num);
2249 if (r == NULL && anchorrule != NULL)
2250 PF_STEP_OUT_OF_ANCHOR(r, anchorrule, ruleset,
2251 rs_num);
2252 }
2253 if (rm != NULL && (rm->action == PF_NONAT ||
2254 rm->action == PF_NORDR || rm->action == PF_NOBINAT))
2255 return (NULL);
2256 return (rm);
2257}
2258
2259struct pf_rule *
2260pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction,
2261 struct pfi_kif *kif, struct pf_src_node **sn,
2262 struct pf_addr *saddr, u_int16_t sport,
2263 struct pf_addr *daddr, u_int16_t dport,
2264 struct pf_addr *naddr, u_int16_t *nport)
2265{
2266 struct pf_rule *r = NULL;
2267
2268 if (direction == PF_OUT) {
2269 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2270 sport, daddr, dport, PF_RULESET_BINAT);
2271 if (r == NULL)
2272 r = pf_match_translation(pd, m, off, direction, kif,
2273 saddr, sport, daddr, dport, PF_RULESET_NAT);
2274 } else {
2275 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2276 sport, daddr, dport, PF_RULESET_RDR);
2277 if (r == NULL)
2278 r = pf_match_translation(pd, m, off, direction, kif,
2279 saddr, sport, daddr, dport, PF_RULESET_BINAT);
2280 }
2281
2282 if (r != NULL) {
2283 switch (r->action) {
2284 case PF_NONAT:
2285 case PF_NOBINAT:
2286 case PF_NORDR:
2287 return (NULL);
2288 case PF_NAT:
2289 if (pf_get_sport(pd->af, pd->proto, r, saddr,
2290 daddr, dport, naddr, nport, r->rpool.proxy_port[0],
2291 r->rpool.proxy_port[1], sn)) {
2292 DPFPRINTF(PF_DEBUG_MISC,
2293 ("pf: NAT proxy port allocation "
2294 "(%u-%u) failed\n",
2295 r->rpool.proxy_port[0],
2296 r->rpool.proxy_port[1]));
2297 return (NULL);
2298 }
2299 break;
2300 case PF_BINAT:
2301 switch (direction) {
2302 case PF_OUT:
2303 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){
2304 if (pd->af == AF_INET) {
2305 if (r->rpool.cur->addr.p.dyn->
2306 pfid_acnt4 < 1)
2307 return (NULL);
2308 PF_POOLMASK(naddr,
2309 &r->rpool.cur->addr.p.dyn->
2310 pfid_addr4,
2311 &r->rpool.cur->addr.p.dyn->
2312 pfid_mask4,
2313 saddr, AF_INET);
2314 } else {
2315 if (r->rpool.cur->addr.p.dyn->
2316 pfid_acnt6 < 1)
2317 return (NULL);
2318 PF_POOLMASK(naddr,
2319 &r->rpool.cur->addr.p.dyn->
2320 pfid_addr6,
2321 &r->rpool.cur->addr.p.dyn->
2322 pfid_mask6,
2323 saddr, AF_INET6);
2324 }
2325 } else
2326 PF_POOLMASK(naddr,
2327 &r->rpool.cur->addr.v.a.addr,
2328 &r->rpool.cur->addr.v.a.mask,
2329 saddr, pd->af);
2330 break;
2331 case PF_IN:
2332 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){
2333 if (pd->af == AF_INET) {
2334 if (r->src.addr.p.dyn->
2335 pfid_acnt4 < 1)
2336 return (NULL);
2337 PF_POOLMASK(naddr,
2338 &r->src.addr.p.dyn->
2339 pfid_addr4,
2340 &r->src.addr.p.dyn->
2341 pfid_mask4,
2342 daddr, AF_INET);
2343 } else {
2344 if (r->src.addr.p.dyn->
2345 pfid_acnt6 < 1)
2346 return (NULL);
2347 PF_POOLMASK(naddr,
2348 &r->src.addr.p.dyn->
2349 pfid_addr6,
2350 &r->src.addr.p.dyn->
2351 pfid_mask6,
2352 daddr, AF_INET6);
2353 }
2354 } else
2355 PF_POOLMASK(naddr,
2356 &r->src.addr.v.a.addr,
2357 &r->src.addr.v.a.mask, daddr,
2358 pd->af);
2359 break;
2360 }
2361 break;
2362 case PF_RDR: {
2363 if (pf_map_addr(r->af, r, saddr, naddr, NULL, sn))
2364 return (NULL);
2365
2366 if (r->rpool.proxy_port[1]) {
2367 u_int32_t tmp_nport;
2368
2369 tmp_nport = ((ntohs(dport) -
2370 ntohs(r->dst.port[0])) %
2371 (r->rpool.proxy_port[1] -
2372 r->rpool.proxy_port[0] + 1)) +
2373 r->rpool.proxy_port[0];
2374
2375 /* wrap around if necessary */
2376 if (tmp_nport > 65535)
2377 tmp_nport -= 65535;
2378 *nport = htons((u_int16_t)tmp_nport);
2379 } else if (r->rpool.proxy_port[0])
2380 *nport = htons(r->rpool.proxy_port[0]);
2381 break;
2382 }
2383 default:
2384 return (NULL);
2385 }
2386 }
2387
2388 return (r);
2389}
2390
2391int
2392#ifdef __FreeBSD__
2393pf_socket_lookup(uid_t *uid, gid_t *gid, int direction, struct pf_pdesc *pd,
2394 struct inpcb *inp_arg)
2395#else
2396pf_socket_lookup(uid_t *uid, gid_t *gid, int direction, struct pf_pdesc *pd)
2397#endif
2398{
2399 struct pf_addr *saddr, *daddr;
2400 u_int16_t sport, dport;
2401#ifdef __FreeBSD__
2402 struct inpcbinfo *pi;
2403#else
2404 struct inpcbtable *tb;
2405#endif
2406 struct inpcb *inp;
2407
2408 *uid = UID_MAX;
2409 *gid = GID_MAX;
2410#ifdef __FreeBSD__
2411 if (inp_arg != NULL) {
2412 INP_LOCK_ASSERT(inp_arg);
2413 if (inp_arg->inp_socket) {
2414 *uid = inp_arg->inp_socket->so_cred->cr_uid;
2415 *gid = inp_arg->inp_socket->so_cred->cr_groups[0];
2416 return (1);
2417 } else
2418 return (0);
2419 }
2420#endif
2421 switch (pd->proto) {
2422 case IPPROTO_TCP:
2423 sport = pd->hdr.tcp->th_sport;
2424 dport = pd->hdr.tcp->th_dport;
2425#ifdef __FreeBSD__
2426 pi = &tcbinfo;
2427#else
2428 tb = &tcbtable;
2429#endif
2430 break;
2431 case IPPROTO_UDP:
2432 sport = pd->hdr.udp->uh_sport;
2433 dport = pd->hdr.udp->uh_dport;
2434#ifdef __FreeBSD__
2435 pi = &udbinfo;
2436#else
2437 tb = &udbtable;
2438#endif
2439 break;
2440 default:
2441 return (0);
2442 }
2443 if (direction == PF_IN) {
2444 saddr = pd->src;
2445 daddr = pd->dst;
2446 } else {
2447 u_int16_t p;
2448
2449 p = sport;
2450 sport = dport;
2451 dport = p;
2452 saddr = pd->dst;
2453 daddr = pd->src;
2454 }
2455 switch (pd->af) {
2456 case AF_INET:
2457#ifdef __FreeBSD__
2458 INP_INFO_RLOCK(pi); /* XXX LOR */
2459 inp = in_pcblookup_hash(pi, saddr->v4, sport, daddr->v4,
2460 dport, 0, NULL);
2461 if (inp == NULL) {
2462 inp = in_pcblookup_hash(pi, saddr->v4, sport,
2463 daddr->v4, dport, INPLOOKUP_WILDCARD, NULL);
2464 if(inp == NULL) {
2465 INP_INFO_RUNLOCK(pi);
2466 return (0);
2467 }
2468 }
2469#else
2470 inp = in_pcbhashlookup(tb, saddr->v4, sport, daddr->v4, dport);
2471 if (inp == NULL) {
2472 inp = in_pcblookup_listen(tb, daddr->v4, dport, 0);
2473 if (inp == NULL)
2474 return (0);
2475 }
2476#endif
2477 break;
2478#ifdef INET6
2479 case AF_INET6:
2480#ifdef __FreeBSD__
2481 INP_INFO_RLOCK(pi);
2482 inp = in6_pcblookup_hash(pi, &saddr->v6, sport,
2483 &daddr->v6, dport, 0, NULL);
2484 if (inp == NULL) {
2485 inp = in6_pcblookup_hash(pi, &saddr->v6, sport,
2486 &daddr->v6, dport, INPLOOKUP_WILDCARD, NULL);
2487 if (inp == NULL) {
2488 INP_INFO_RUNLOCK(pi);
2489 return (0);
2490 }
2491 }
2492#else
2493 inp = in6_pcbhashlookup(tb, &saddr->v6, sport, &daddr->v6,
2494 dport);
2495 if (inp == NULL) {
2496 inp = in6_pcblookup_listen(tb, &daddr->v6, dport, 0);
2497 if (inp == NULL)
2498 return (0);
2499 }
2500#endif
2501 break;
2502#endif /* INET6 */
2503
2504 default:
2505 return (0);
2506 }
2507#ifdef __FreeBSD__
2508 INP_LOCK(inp);
2509 if ((inp->inp_socket == NULL) || (inp->inp_socket->so_cred == NULL)) {
2510 INP_UNLOCK(inp);
2511 INP_INFO_RUNLOCK(pi);
2512 return (0);
2513 }
2514 *uid = inp->inp_socket->so_cred->cr_uid;
2515 *gid = inp->inp_socket->so_cred->cr_groups[0];
2516 INP_UNLOCK(inp);
2517 INP_INFO_RUNLOCK(pi);
2518#else
2519 *uid = inp->inp_socket->so_euid;
2520 *gid = inp->inp_socket->so_egid;
2521#endif
2522 return (1);
2523}
2524
2525u_int8_t
2526pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2527{
2528 int hlen;
2529 u_int8_t hdr[60];
2530 u_int8_t *opt, optlen;
2531 u_int8_t wscale = 0;
2532
2533 hlen = th_off << 2; /* hlen <= sizeof(hdr) */
2534 if (hlen <= sizeof(struct tcphdr))
2535 return (0);
2536 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2537 return (0);
2538 opt = hdr + sizeof(struct tcphdr);
2539 hlen -= sizeof(struct tcphdr);
2540 while (hlen >= 3) {
2541 switch (*opt) {
2542 case TCPOPT_EOL:
2543 case TCPOPT_NOP:
2544 ++opt;
2545 --hlen;
2546 break;
2547 case TCPOPT_WINDOW:
2548 wscale = opt[2];
2549 if (wscale > TCP_MAX_WINSHIFT)
2550 wscale = TCP_MAX_WINSHIFT;
2551 wscale |= PF_WSCALE_FLAG;
2552 /* FALLTHROUGH */
2553 default:
2554 optlen = opt[1];
2555 if (optlen < 2)
2556 optlen = 2;
2557 hlen -= optlen;
2558 opt += optlen;
2559 break;
2560 }
2561 }
2562 return (wscale);
2563}
2564
2565u_int16_t
2566pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2567{
2568 int hlen;
2569 u_int8_t hdr[60];
2570 u_int8_t *opt, optlen;
2571 u_int16_t mss = tcp_mssdflt;
2572
2573 hlen = th_off << 2; /* hlen <= sizeof(hdr) */
2574 if (hlen <= sizeof(struct tcphdr))
2575 return (0);
2576 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2577 return (0);
2578 opt = hdr + sizeof(struct tcphdr);
2579 hlen -= sizeof(struct tcphdr);
2580 while (hlen >= TCPOLEN_MAXSEG) {
2581 switch (*opt) {
2582 case TCPOPT_EOL:
2583 case TCPOPT_NOP:
2584 ++opt;
2585 --hlen;
2586 break;
2587 case TCPOPT_MAXSEG:
2588 bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2);
2589 /* FALLTHROUGH */
2590 default:
2591 optlen = opt[1];
2592 if (optlen < 2)
2593 optlen = 2;
2594 hlen -= optlen;
2595 opt += optlen;
2596 break;
2597 }
2598 }
2599 return (mss);
2600}
2601
2602u_int16_t
2603pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer)
2604{
2605#ifdef INET
2606 struct sockaddr_in *dst;
2607 struct route ro;
2608#endif /* INET */
2609#ifdef INET6
2610 struct sockaddr_in6 *dst6;
2611 struct route_in6 ro6;
2612#endif /* INET6 */
2613 struct rtentry *rt = NULL;
2614 int hlen = 0; /* make the compiler happy */
2615 u_int16_t mss = tcp_mssdflt;
2616
2617 switch (af) {
2618#ifdef INET
2619 case AF_INET:
2620 hlen = sizeof(struct ip);
2621 bzero(&ro, sizeof(ro));
2622 dst = (struct sockaddr_in *)&ro.ro_dst;
2623 dst->sin_family = AF_INET;
2624 dst->sin_len = sizeof(*dst);
2625 dst->sin_addr = addr->v4;
2626#ifdef __FreeBSD__
2627#ifdef RTF_PRCLONING
2628 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
2629#else /* !RTF_PRCLONING */
2630 rtalloc_ign(&ro, RTF_CLONING);
2631#endif
2632#else /* ! __FreeBSD__ */
2633 rtalloc_noclone(&ro, NO_CLONING);
2634#endif
2635 rt = ro.ro_rt;
2636 break;
2637#endif /* INET */
2638#ifdef INET6
2639 case AF_INET6:
2640 hlen = sizeof(struct ip6_hdr);
2641 bzero(&ro6, sizeof(ro6));
2642 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst;
2643 dst6->sin6_family = AF_INET6;
2644 dst6->sin6_len = sizeof(*dst6);
2645 dst6->sin6_addr = addr->v6;
2646#ifdef __FreeBSD__
2647#ifdef RTF_PRCLONING
2648 rtalloc_ign((struct route *)&ro6,
2649 (RTF_CLONING | RTF_PRCLONING));
2650#else /* !RTF_PRCLONING */
2651 rtalloc_ign((struct route *)&ro6, RTF_CLONING);
2652#endif
2653#else /* ! __FreeBSD__ */
2654 rtalloc_noclone((struct route *)&ro6, NO_CLONING);
2655#endif
2656 rt = ro6.ro_rt;
2657 break;
2658#endif /* INET6 */
2659 }
2660
2661 if (rt && rt->rt_ifp) {
2662 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr);
2663 mss = max(tcp_mssdflt, mss);
2664 RTFREE(rt);
2665 }
2666 mss = min(mss, offer);
2667 mss = max(mss, 64); /* sanity - at least max opt space */
2668 return (mss);
2669}
2670
2671void
2672pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr)
2673{
2674 struct pf_rule *r = s->rule.ptr;
2675
2676 s->rt_kif = NULL;
2677 if (!r->rt || r->rt == PF_FASTROUTE)
2678 return;
2679 switch (s->af) {
2680#ifdef INET
2681 case AF_INET:
2682 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL,
2683 &s->nat_src_node);
2684 s->rt_kif = r->rpool.cur->kif;
2685 break;
2686#endif /* INET */
2687#ifdef INET6
2688 case AF_INET6:
2689 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL,
2690 &s->nat_src_node);
2691 s->rt_kif = r->rpool.cur->kif;
2692 break;
2693#endif /* INET6 */
2694 }
2695}
2696
2697int
2698pf_test_tcp(struct pf_rule **rm, struct pf_state **sm, int direction,
2699 struct pfi_kif *kif, struct mbuf *m, int off, void *h,
2700#ifdef __FreeBSD__
2701 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
2702 struct inpcb *inp)
2703#else
2704 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
2705#endif
2706{
2707 struct pf_rule *nr = NULL;
2708 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
2709 struct tcphdr *th = pd->hdr.tcp;
2710 u_int16_t bport, nport = 0;
2711 sa_family_t af = pd->af;
2712 int lookup = -1;
2713 uid_t uid;
2714 gid_t gid;
2715 struct pf_rule *r, *a = NULL;
2716 struct pf_ruleset *ruleset = NULL;
2717 struct pf_src_node *nsn = NULL;
2718 u_short reason;
2719 int rewrite = 0;
2720 struct pf_tag *pftag = NULL;
2721 int tag = -1;
2722 u_int16_t mss = tcp_mssdflt;
2723
2724 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
2725
2726 if (direction == PF_OUT) {
2727 bport = nport = th->th_sport;
2728 /* check outgoing packet for BINAT/NAT */
2729 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
2730 saddr, th->th_sport, daddr, th->th_dport,
2731 &pd->naddr, &nport)) != NULL) {
2732 PF_ACPY(&pd->baddr, saddr, af);
2733 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
2734 &th->th_sum, &pd->naddr, nport, 0, af);
2735 rewrite++;
2736 if (nr->natpass)
2737 r = NULL;
2738 pd->nat_rule = nr;
2739 }
2740 } else {
2741 bport = nport = th->th_dport;
2742 /* check incoming packet for BINAT/RDR */
2743 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
2744 saddr, th->th_sport, daddr, th->th_dport,
2745 &pd->naddr, &nport)) != NULL) {
2746 PF_ACPY(&pd->baddr, daddr, af);
2747 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
2748 &th->th_sum, &pd->naddr, nport, 0, af);
2749 rewrite++;
2750 if (nr->natpass)
2751 r = NULL;
2752 pd->nat_rule = nr;
2753 }
2754 }
2755
2756 while (r != NULL) {
2757 r->evaluations++;
2758 if (r->kif != NULL &&
2759 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
2760 r = r->skip[PF_SKIP_IFP].ptr;
2761 else if (r->direction && r->direction != direction)
2762 r = r->skip[PF_SKIP_DIR].ptr;
2763 else if (r->af && r->af != af)
2764 r = r->skip[PF_SKIP_AF].ptr;
2765 else if (r->proto && r->proto != IPPROTO_TCP)
2766 r = r->skip[PF_SKIP_PROTO].ptr;
2767 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
2768 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
2769 else if (r->src.port_op && !pf_match_port(r->src.port_op,
2770 r->src.port[0], r->src.port[1], th->th_sport))
2771 r = r->skip[PF_SKIP_SRC_PORT].ptr;
2772 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
2773 r = r->skip[PF_SKIP_DST_ADDR].ptr;
2774 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
2775 r->dst.port[0], r->dst.port[1], th->th_dport))
2776 r = r->skip[PF_SKIP_DST_PORT].ptr;
2777 else if (r->tos && !(r->tos & pd->tos))
2778 r = TAILQ_NEXT(r, entries);
2779 else if (r->rule_flag & PFRULE_FRAGMENT)
2780 r = TAILQ_NEXT(r, entries);
2781 else if ((r->flagset & th->th_flags) != r->flags)
2782 r = TAILQ_NEXT(r, entries);
2783 else if (r->uid.op && (lookup != -1 || (lookup =
2784#ifdef __FreeBSD__
2785 pf_socket_lookup(&uid, &gid, direction, pd, inp), 1)) &&
2786#else
2787 pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2788#endif
2789 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
2790 uid))
2791 r = TAILQ_NEXT(r, entries);
2792 else if (r->gid.op && (lookup != -1 || (lookup =
2793#ifdef __FreeBSD__
2794 pf_socket_lookup(&uid, &gid, direction, pd, inp), 1)) &&
2795#else
2796 pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2797#endif
2798 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
2799 gid))
2800 r = TAILQ_NEXT(r, entries);
2801 else if (r->match_tag && !pf_match_tag(m, r, nr, pftag, &tag))
2802 r = TAILQ_NEXT(r, entries);
2803 else if (r->anchorname[0] && r->anchor == NULL)
2804 r = TAILQ_NEXT(r, entries);
2805 else if (r->os_fingerprint != PF_OSFP_ANY && !pf_osfp_match(
2806 pf_osfp_fingerprint(pd, m, off, th), r->os_fingerprint))
2807 r = TAILQ_NEXT(r, entries);
2808 else {
2809 if (r->tag)
2810 tag = r->tag;
2811 if (r->anchor == NULL) {
2812 *rm = r;
2813 *am = a;
2814 *rsm = ruleset;
2815 if ((*rm)->quick)
2816 break;
2817 r = TAILQ_NEXT(r, entries);
2818 } else
2819 PF_STEP_INTO_ANCHOR(r, a, ruleset,
2820 PF_RULESET_FILTER);
2821 }
2822 if (r == NULL && a != NULL)
2823 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
2824 PF_RULESET_FILTER);
2825 }
2826 r = *rm;
2827 a = *am;
2828 ruleset = *rsm;
2829
2830 REASON_SET(&reason, PFRES_MATCH);
2831
2832 if (r->log) {
2833 if (rewrite)
2834 m_copyback(m, off, sizeof(*th), (caddr_t)th);
2835 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
2836 }
2837
2838 if ((r->action == PF_DROP) &&
2839 ((r->rule_flag & PFRULE_RETURNRST) ||
2840 (r->rule_flag & PFRULE_RETURNICMP) ||
2841 (r->rule_flag & PFRULE_RETURN))) {
2842 /* undo NAT changes, if they have taken place */
2843 if (nr != NULL) {
2844 if (direction == PF_OUT) {
2845 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
2846 &th->th_sum, &pd->baddr, bport, 0, af);
2847 rewrite++;
2848 } else {
2849 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
2850 &th->th_sum, &pd->baddr, bport, 0, af);
2851 rewrite++;
2852 }
2853 }
2854 if (((r->rule_flag & PFRULE_RETURNRST) ||
2855 (r->rule_flag & PFRULE_RETURN)) &&
2856 !(th->th_flags & TH_RST)) {
2857 u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
2858
2859 if (th->th_flags & TH_SYN)
2860 ack++;
2861 if (th->th_flags & TH_FIN)
2862 ack++;
2863 pf_send_tcp(r, af, pd->dst,
2864 pd->src, th->th_dport, th->th_sport,
2865 ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0,
2866 r->return_ttl);
2867 } else if ((af == AF_INET) && r->return_icmp)
2868 pf_send_icmp(m, r->return_icmp >> 8,
2869 r->return_icmp & 255, af, r);
2870 else if ((af == AF_INET6) && r->return_icmp6)
2871 pf_send_icmp(m, r->return_icmp6 >> 8,
2872 r->return_icmp6 & 255, af, r);
2873 }
2874
2875 if (r->action == PF_DROP)
2876 return (PF_DROP);
2877
2878 if (pf_tag_packet(m, pftag, tag)) {
2879 REASON_SET(&reason, PFRES_MEMORY);
2880 return (PF_DROP);
2881 }
2882
2883 if (r->keep_state || nr != NULL ||
2884 (pd->flags & PFDESC_TCP_NORM)) {
2885 /* create new state */
2886 u_int16_t len;
2887 struct pf_state *s = NULL;
2888 struct pf_src_node *sn = NULL;
2889
2890 len = pd->tot_len - off - (th->th_off << 2);
2891
2892 /* check maximums */
2893 if (r->max_states && (r->states >= r->max_states))
2894 goto cleanup;
2895 /* src node for flter rule */
2896 if ((r->rule_flag & PFRULE_SRCTRACK ||
2897 r->rpool.opts & PF_POOL_STICKYADDR) &&
2898 pf_insert_src_node(&sn, r, saddr, af) != 0)
2899 goto cleanup;
2900 /* src node for translation rule */
2901 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
2902 ((direction == PF_OUT &&
2903 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
2904 (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
2905 goto cleanup;
2906 s = pool_get(&pf_state_pl, PR_NOWAIT);
2907 if (s == NULL) {
2908cleanup:
2909 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
2910 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
2911 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
2912 pf_status.src_nodes--;
2913 pool_put(&pf_src_tree_pl, sn);
2914 }
2915 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
2916 nsn->expire == 0) {
2917 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
2918 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
2919 pf_status.src_nodes--;
2920 pool_put(&pf_src_tree_pl, nsn);
2921 }
2922 REASON_SET(&reason, PFRES_MEMORY);
2923 return (PF_DROP);
2924 }
2925 bzero(s, sizeof(*s));
2926 r->states++;
2927 if (a != NULL)
2928 a->states++;
2929 s->rule.ptr = r;
2930 s->nat_rule.ptr = nr;
2931 if (s->nat_rule.ptr != NULL)
2932 s->nat_rule.ptr->states++;
2933 s->anchor.ptr = a;
2934 s->allow_opts = r->allow_opts;
2935 s->log = r->log & 2;
2936 s->proto = IPPROTO_TCP;
2937 s->direction = direction;
2938 s->af = af;
2939 if (direction == PF_OUT) {
2940 PF_ACPY(&s->gwy.addr, saddr, af);
2941 s->gwy.port = th->th_sport; /* sport */
2942 PF_ACPY(&s->ext.addr, daddr, af);
2943 s->ext.port = th->th_dport;
2944 if (nr != NULL) {
2945 PF_ACPY(&s->lan.addr, &pd->baddr, af);
2946 s->lan.port = bport;
2947 } else {
2948 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
2949 s->lan.port = s->gwy.port;
2950 }
2951 } else {
2952 PF_ACPY(&s->lan.addr, daddr, af);
2953 s->lan.port = th->th_dport;
2954 PF_ACPY(&s->ext.addr, saddr, af);
2955 s->ext.port = th->th_sport;
2956 if (nr != NULL) {
2957 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
2958 s->gwy.port = bport;
2959 } else {
2960 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
2961 s->gwy.port = s->lan.port;
2962 }
2963 }
2964
2965 s->src.seqlo = ntohl(th->th_seq);
2966 s->src.seqhi = s->src.seqlo + len + 1;
2967 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
2968 r->keep_state == PF_STATE_MODULATE) {
2969 /* Generate sequence number modulator */
2970 while ((s->src.seqdiff = arc4random()) == 0)
2971 ;
2972 pf_change_a(&th->th_seq, &th->th_sum,
2973 htonl(s->src.seqlo + s->src.seqdiff), 0);
2974 rewrite = 1;
2975 } else
2976 s->src.seqdiff = 0;
2977 if (th->th_flags & TH_SYN) {
2978 s->src.seqhi++;
2979 s->src.wscale = pf_get_wscale(m, off, th->th_off, af);
2980 }
2981 s->src.max_win = MAX(ntohs(th->th_win), 1);
2982 if (s->src.wscale & PF_WSCALE_MASK) {
2983 /* Remove scale factor from initial window */
2984 int win = s->src.max_win;
2985 win += 1 << (s->src.wscale & PF_WSCALE_MASK);
2986 s->src.max_win = (win - 1) >>
2987 (s->src.wscale & PF_WSCALE_MASK);
2988 }
2989 if (th->th_flags & TH_FIN)
2990 s->src.seqhi++;
2991 s->dst.seqhi = 1;
2992 s->dst.max_win = 1;
2993 s->src.state = TCPS_SYN_SENT;
2994 s->dst.state = TCPS_CLOSED;
2995#ifdef __FreeBSD__
2996 s->creation = time_second;
2997 s->expire = time_second;
2998#else
2999 s->creation = time.tv_sec;
3000 s->expire = time.tv_sec;
3001#endif
3002 s->timeout = PFTM_TCP_FIRST_PACKET;
3003 pf_set_rt_ifp(s, saddr);
3004 if (sn != NULL) {
3005 s->src_node = sn;
3006 s->src_node->states++;
3007 }
3008 if (nsn != NULL) {
3009 PF_ACPY(&nsn->raddr, &pd->naddr, af);
3010 s->nat_src_node = nsn;
3011 s->nat_src_node->states++;
3012 }
3013 if ((pd->flags & PFDESC_TCP_NORM) && pf_normalize_tcp_init(m,
3014 off, pd, th, &s->src, &s->dst)) {
3015 REASON_SET(&reason, PFRES_MEMORY);
3016 pf_src_tree_remove_state(s);
3017 pool_put(&pf_state_pl, s);
3018 return (PF_DROP);
3019 }
3020 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
3021 pf_normalize_tcp_stateful(m, off, pd, &reason, th, &s->src,
3022 &s->dst, &rewrite)) {
3023 pf_normalize_tcp_cleanup(s);
3024 pf_src_tree_remove_state(s);
3025 pool_put(&pf_state_pl, s);
3026 return (PF_DROP);
3027 }
3028 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3029 pf_normalize_tcp_cleanup(s);
3030 REASON_SET(&reason, PFRES_MEMORY);
3031 pf_src_tree_remove_state(s);
3032 pool_put(&pf_state_pl, s);
3033 return (PF_DROP);
3034 } else
3035 *sm = s;
3036 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
3037 r->keep_state == PF_STATE_SYNPROXY) {
3038 s->src.state = PF_TCPS_PROXY_SRC;
3039 if (nr != NULL) {
3040 if (direction == PF_OUT) {
3041 pf_change_ap(saddr, &th->th_sport,
3042 pd->ip_sum, &th->th_sum, &pd->baddr,
3043 bport, 0, af);
3044 } else {
3045 pf_change_ap(daddr, &th->th_dport,
3046 pd->ip_sum, &th->th_sum, &pd->baddr,
3047 bport, 0, af);
3048 }
3049 }
3050 s->src.seqhi = arc4random();
3051 /* Find mss option */
3052 mss = pf_get_mss(m, off, th->th_off, af);
3053 mss = pf_calc_mss(saddr, af, mss);
3054 mss = pf_calc_mss(daddr, af, mss);
3055 s->src.mss = mss;
3056 pf_send_tcp(r, af, daddr, saddr, th->th_dport,
3057 th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1,
3058 TH_SYN|TH_ACK, 0, s->src.mss, 0);
3059 return (PF_SYNPROXY_DROP);
3060 }
3061 }
3062
3063 /* copy back packet headers if we performed NAT operations */
3064 if (rewrite)
3065 m_copyback(m, off, sizeof(*th), (caddr_t)th);
3066
3067 return (PF_PASS);
3068}
3069
3070int
3071pf_test_udp(struct pf_rule **rm, struct pf_state **sm, int direction,
3072 struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3073#ifdef __FreeBSD__
3074 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
3075 struct inpcb *inp)
3076#else
3077 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
3078#endif
3079{
3080 struct pf_rule *nr = NULL;
3081 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
3082 struct udphdr *uh = pd->hdr.udp;
3083 u_int16_t bport, nport = 0;
3084 sa_family_t af = pd->af;
3085 int lookup = -1;
3086 uid_t uid;
3087 gid_t gid;
3088 struct pf_rule *r, *a = NULL;
3089 struct pf_ruleset *ruleset = NULL;
3090 struct pf_src_node *nsn = NULL;
3091 u_short reason;
3092 int rewrite = 0;
3093 struct pf_tag *pftag = NULL;
3094 int tag = -1;
3095
3096 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3097
3098 if (direction == PF_OUT) {
3099 bport = nport = uh->uh_sport;
3100 /* check outgoing packet for BINAT/NAT */
3101 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3102 saddr, uh->uh_sport, daddr, uh->uh_dport,
3103 &pd->naddr, &nport)) != NULL) {
3104 PF_ACPY(&pd->baddr, saddr, af);
3105 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum,
3106 &uh->uh_sum, &pd->naddr, nport, 1, af);
3107 rewrite++;
3108 if (nr->natpass)
3109 r = NULL;
3110 pd->nat_rule = nr;
3111 }
3112 } else {
3113 bport = nport = uh->uh_dport;
3114 /* check incoming packet for BINAT/RDR */
3115 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3116 saddr, uh->uh_sport, daddr, uh->uh_dport, &pd->naddr,
3117 &nport)) != NULL) {
3118 PF_ACPY(&pd->baddr, daddr, af);
3119 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum,
3120 &uh->uh_sum, &pd->naddr, nport, 1, af);
3121 rewrite++;
3122 if (nr->natpass)
3123 r = NULL;
3124 pd->nat_rule = nr;
3125 }
3126 }
3127
3128 while (r != NULL) {
3129 r->evaluations++;
3130 if (r->kif != NULL &&
3131 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3132 r = r->skip[PF_SKIP_IFP].ptr;
3133 else if (r->direction && r->direction != direction)
3134 r = r->skip[PF_SKIP_DIR].ptr;
3135 else if (r->af && r->af != af)
3136 r = r->skip[PF_SKIP_AF].ptr;
3137 else if (r->proto && r->proto != IPPROTO_UDP)
3138 r = r->skip[PF_SKIP_PROTO].ptr;
3139 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
3140 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3141 else if (r->src.port_op && !pf_match_port(r->src.port_op,
3142 r->src.port[0], r->src.port[1], uh->uh_sport))
3143 r = r->skip[PF_SKIP_SRC_PORT].ptr;
3144 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
3145 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3146 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
3147 r->dst.port[0], r->dst.port[1], uh->uh_dport))
3148 r = r->skip[PF_SKIP_DST_PORT].ptr;
3149 else if (r->tos && !(r->tos & pd->tos))
3150 r = TAILQ_NEXT(r, entries);
3151 else if (r->rule_flag & PFRULE_FRAGMENT)
3152 r = TAILQ_NEXT(r, entries);
3153 else if (r->uid.op && (lookup != -1 || (lookup =
3154#ifdef __FreeBSD__
3155 pf_socket_lookup(&uid, &gid, direction, pd, inp), 1)) &&
3156#else
3157 pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
3158#endif
3159 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
3160 uid))
3161 r = TAILQ_NEXT(r, entries);
3162 else if (r->gid.op && (lookup != -1 || (lookup =
3163#ifdef __FreeBSD__
3164 pf_socket_lookup(&uid, &gid, direction, pd, inp), 1)) &&
3165#else
3166 pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
3167#endif
3168 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
3169 gid))
3170 r = TAILQ_NEXT(r, entries);
3171 else if (r->match_tag && !pf_match_tag(m, r, nr, pftag, &tag))
3172 r = TAILQ_NEXT(r, entries);
3173 else if (r->anchorname[0] && r->anchor == NULL)
3174 r = TAILQ_NEXT(r, entries);
3175 else if (r->os_fingerprint != PF_OSFP_ANY)
3176 r = TAILQ_NEXT(r, entries);
3177 else {
3178 if (r->tag)
3179 tag = r->tag;
3180 if (r->anchor == NULL) {
3181 *rm = r;
3182 *am = a;
3183 *rsm = ruleset;
3184 if ((*rm)->quick)
3185 break;
3186 r = TAILQ_NEXT(r, entries);
3187 } else
3188 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3189 PF_RULESET_FILTER);
3190 }
3191 if (r == NULL && a != NULL)
3192 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3193 PF_RULESET_FILTER);
3194 }
3195 r = *rm;
3196 a = *am;
3197 ruleset = *rsm;
3198
3199 REASON_SET(&reason, PFRES_MATCH);
3200
3201 if (r->log) {
3202 if (rewrite)
3203 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
3204 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3205 }
3206
3207 if ((r->action == PF_DROP) &&
3208 ((r->rule_flag & PFRULE_RETURNICMP) ||
3209 (r->rule_flag & PFRULE_RETURN))) {
3210 /* undo NAT changes, if they have taken place */
3211 if (nr != NULL) {
3212 if (direction == PF_OUT) {
3213 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum,
3214 &uh->uh_sum, &pd->baddr, bport, 1, af);
3215 rewrite++;
3216 } else {
3217 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum,
3218 &uh->uh_sum, &pd->baddr, bport, 1, af);
3219 rewrite++;
3220 }
3221 }
3222 if ((af == AF_INET) && r->return_icmp)
3223 pf_send_icmp(m, r->return_icmp >> 8,
3224 r->return_icmp & 255, af, r);
3225 else if ((af == AF_INET6) && r->return_icmp6)
3226 pf_send_icmp(m, r->return_icmp6 >> 8,
3227 r->return_icmp6 & 255, af, r);
3228 }
3229
3230 if (r->action == PF_DROP)
3231 return (PF_DROP);
3232
3233 if (pf_tag_packet(m, pftag, tag)) {
3234 REASON_SET(&reason, PFRES_MEMORY);
3235 return (PF_DROP);
3236 }
3237
3238 if (r->keep_state || nr != NULL) {
3239 /* create new state */
3240 struct pf_state *s = NULL;
3241 struct pf_src_node *sn = NULL;
3242
3243 /* check maximums */
3244 if (r->max_states && (r->states >= r->max_states))
3245 goto cleanup;
3246 /* src node for flter rule */
3247 if ((r->rule_flag & PFRULE_SRCTRACK ||
3248 r->rpool.opts & PF_POOL_STICKYADDR) &&
3249 pf_insert_src_node(&sn, r, saddr, af) != 0)
3250 goto cleanup;
3251 /* src node for translation rule */
3252 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3253 ((direction == PF_OUT &&
3254 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3255 (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
3256 goto cleanup;
3257 s = pool_get(&pf_state_pl, PR_NOWAIT);
3258 if (s == NULL) {
3259cleanup:
3260 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3261 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3262 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3263 pf_status.src_nodes--;
3264 pool_put(&pf_src_tree_pl, sn);
3265 }
3266 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3267 nsn->expire == 0) {
3268 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3269 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3270 pf_status.src_nodes--;
3271 pool_put(&pf_src_tree_pl, nsn);
3272 }
3273 REASON_SET(&reason, PFRES_MEMORY);
3274 return (PF_DROP);
3275 }
3276 bzero(s, sizeof(*s));
3277 r->states++;
3278 if (a != NULL)
3279 a->states++;
3280 s->rule.ptr = r;
3281 s->nat_rule.ptr = nr;
3282 if (s->nat_rule.ptr != NULL)
3283 s->nat_rule.ptr->states++;
3284 s->anchor.ptr = a;
3285 s->allow_opts = r->allow_opts;
3286 s->log = r->log & 2;
3287 s->proto = IPPROTO_UDP;
3288 s->direction = direction;
3289 s->af = af;
3290 if (direction == PF_OUT) {
3291 PF_ACPY(&s->gwy.addr, saddr, af);
3292 s->gwy.port = uh->uh_sport;
3293 PF_ACPY(&s->ext.addr, daddr, af);
3294 s->ext.port = uh->uh_dport;
3295 if (nr != NULL) {
3296 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3297 s->lan.port = bport;
3298 } else {
3299 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3300 s->lan.port = s->gwy.port;
3301 }
3302 } else {
3303 PF_ACPY(&s->lan.addr, daddr, af);
3304 s->lan.port = uh->uh_dport;
3305 PF_ACPY(&s->ext.addr, saddr, af);
3306 s->ext.port = uh->uh_sport;
3307 if (nr != NULL) {
3308 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3309 s->gwy.port = bport;
3310 } else {
3311 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3312 s->gwy.port = s->lan.port;
3313 }
3314 }
3315 s->src.state = PFUDPS_SINGLE;
3316 s->dst.state = PFUDPS_NO_TRAFFIC;
3317#ifdef __FreeBSD__
3318 s->creation = time_second;
3319 s->expire = time_second;
3320#else
3321 s->creation = time.tv_sec;
3322 s->expire = time.tv_sec;
3323#endif
3324 s->timeout = PFTM_UDP_FIRST_PACKET;
3325 pf_set_rt_ifp(s, saddr);
3326 if (sn != NULL) {
3327 s->src_node = sn;
3328 s->src_node->states++;
3329 }
3330 if (nsn != NULL) {
3331 PF_ACPY(&nsn->raddr, &pd->naddr, af);
3332 s->nat_src_node = nsn;
3333 s->nat_src_node->states++;
3334 }
3335 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3336 REASON_SET(&reason, PFRES_MEMORY);
3337 pf_src_tree_remove_state(s);
3338 pool_put(&pf_state_pl, s);
3339 return (PF_DROP);
3340 } else
3341 *sm = s;
3342 }
3343
3344 /* copy back packet headers if we performed NAT operations */
3345 if (rewrite)
3346 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
3347
3348 return (PF_PASS);
3349}
3350
3351int
3352pf_test_icmp(struct pf_rule **rm, struct pf_state **sm, int direction,
3353 struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3354 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
3355{
3356 struct pf_rule *nr = NULL;
3357 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
3358 struct pf_rule *r, *a = NULL;
3359 struct pf_ruleset *ruleset = NULL;
3360 struct pf_src_node *nsn = NULL;
3361 u_short reason;
3362 u_int16_t icmpid = 0; /* make the compiler happy */
3363 sa_family_t af = pd->af;
3364 u_int8_t icmptype = 0; /* make the compiler happy */
3365 u_int8_t icmpcode = 0; /* make the compiler happy */
3366 int state_icmp = 0;
3367 struct pf_tag *pftag = NULL;
3368 int tag = -1;
3369#ifdef INET6
3370 int rewrite = 0;
3371#endif /* INET6 */
3372
3373 switch (pd->proto) {
3374#ifdef INET
3375 case IPPROTO_ICMP:
3376 icmptype = pd->hdr.icmp->icmp_type;
3377 icmpcode = pd->hdr.icmp->icmp_code;
3378 icmpid = pd->hdr.icmp->icmp_id;
3379
3380 if (icmptype == ICMP_UNREACH ||
3381 icmptype == ICMP_SOURCEQUENCH ||
3382 icmptype == ICMP_REDIRECT ||
3383 icmptype == ICMP_TIMXCEED ||
3384 icmptype == ICMP_PARAMPROB)
3385 state_icmp++;
3386 break;
3387#endif /* INET */
3388#ifdef INET6
3389 case IPPROTO_ICMPV6:
3390 icmptype = pd->hdr.icmp6->icmp6_type;
3391 icmpcode = pd->hdr.icmp6->icmp6_code;
3392 icmpid = pd->hdr.icmp6->icmp6_id;
3393
3394 if (icmptype == ICMP6_DST_UNREACH ||
3395 icmptype == ICMP6_PACKET_TOO_BIG ||
3396 icmptype == ICMP6_TIME_EXCEEDED ||
3397 icmptype == ICMP6_PARAM_PROB)
3398 state_icmp++;
3399 break;
3400#endif /* INET6 */
3401 }
3402
3403 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3404
3405 if (direction == PF_OUT) {
3406 /* check outgoing packet for BINAT/NAT */
3407 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3408 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3409 PF_ACPY(&pd->baddr, saddr, af);
3410 switch (af) {
3411#ifdef INET
3412 case AF_INET:
3413 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3414 pd->naddr.v4.s_addr, 0);
3415 break;
3416#endif /* INET */
3417#ifdef INET6
3418 case AF_INET6:
3419 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum,
3420 &pd->naddr, 0);
3421 rewrite++;
3422 break;
3423#endif /* INET6 */
3424 }
3425 if (nr->natpass)
3426 r = NULL;
3427 pd->nat_rule = nr;
3428 }
3429 } else {
3430 /* check incoming packet for BINAT/RDR */
3431 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3432 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3433 PF_ACPY(&pd->baddr, daddr, af);
3434 switch (af) {
3435#ifdef INET
3436 case AF_INET:
3437 pf_change_a(&daddr->v4.s_addr,
3438 pd->ip_sum, pd->naddr.v4.s_addr, 0);
3439 break;
3440#endif /* INET */
3441#ifdef INET6
3442 case AF_INET6:
3443 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum,
3444 &pd->naddr, 0);
3445 rewrite++;
3446 break;
3447#endif /* INET6 */
3448 }
3449 if (nr->natpass)
3450 r = NULL;
3451 pd->nat_rule = nr;
3452 }
3453 }
3454
3455 while (r != NULL) {
3456 r->evaluations++;
3457 if (r->kif != NULL &&
3458 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3459 r = r->skip[PF_SKIP_IFP].ptr;
3460 else if (r->direction && r->direction != direction)
3461 r = r->skip[PF_SKIP_DIR].ptr;
3462 else if (r->af && r->af != af)
3463 r = r->skip[PF_SKIP_AF].ptr;
3464 else if (r->proto && r->proto != pd->proto)
3465 r = r->skip[PF_SKIP_PROTO].ptr;
3466 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
3467 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3468 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
3469 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3470 else if (r->type && r->type != icmptype + 1)
3471 r = TAILQ_NEXT(r, entries);
3472 else if (r->code && r->code != icmpcode + 1)
3473 r = TAILQ_NEXT(r, entries);
3474 else if (r->tos && !(r->tos & pd->tos))
3475 r = TAILQ_NEXT(r, entries);
3476 else if (r->rule_flag & PFRULE_FRAGMENT)
3477 r = TAILQ_NEXT(r, entries);
3478 else if (r->match_tag && !pf_match_tag(m, r, nr, pftag, &tag))
3479 r = TAILQ_NEXT(r, entries);
3480 else if (r->anchorname[0] && r->anchor == NULL)
3481 r = TAILQ_NEXT(r, entries);
3482 else if (r->os_fingerprint != PF_OSFP_ANY)
3483 r = TAILQ_NEXT(r, entries);
3484 else {
3485 if (r->tag)
3486 tag = r->tag;
3487 if (r->anchor == NULL) {
3488 *rm = r;
3489 *am = a;
3490 *rsm = ruleset;
3491 if ((*rm)->quick)
3492 break;
3493 r = TAILQ_NEXT(r, entries);
3494 } else
3495 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3496 PF_RULESET_FILTER);
3497 }
3498 if (r == NULL && a != NULL)
3499 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3500 PF_RULESET_FILTER);
3501 }
3502 r = *rm;
3503 a = *am;
3504 ruleset = *rsm;
3505
3506 REASON_SET(&reason, PFRES_MATCH);
3507
3508 if (r->log) {
3509#ifdef INET6
3510 if (rewrite)
3511 m_copyback(m, off, sizeof(struct icmp6_hdr),
3512 (caddr_t)pd->hdr.icmp6);
3513#endif /* INET6 */
3514 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3515 }
3516
3517 if (r->action != PF_PASS)
3518 return (PF_DROP);
3519
3520 if (pf_tag_packet(m, pftag, tag)) {
3521 REASON_SET(&reason, PFRES_MEMORY);
3522 return (PF_DROP);
3523 }
3524
3525 if (!state_icmp && (r->keep_state || nr != NULL)) {
3526 /* create new state */
3527 struct pf_state *s = NULL;
3528 struct pf_src_node *sn = NULL;
3529
3530 /* check maximums */
3531 if (r->max_states && (r->states >= r->max_states))
3532 goto cleanup;
3533 /* src node for flter rule */
3534 if ((r->rule_flag & PFRULE_SRCTRACK ||
3535 r->rpool.opts & PF_POOL_STICKYADDR) &&
3536 pf_insert_src_node(&sn, r, saddr, af) != 0)
3537 goto cleanup;
3538 /* src node for translation rule */
3539 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3540 ((direction == PF_OUT &&
3541 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3542 (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
3543 goto cleanup;
3544 s = pool_get(&pf_state_pl, PR_NOWAIT);
3545 if (s == NULL) {
3546cleanup:
3547 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3548 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3549 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3550 pf_status.src_nodes--;
3551 pool_put(&pf_src_tree_pl, sn);
3552 }
3553 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3554 nsn->expire == 0) {
3555 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3556 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3557 pf_status.src_nodes--;
3558 pool_put(&pf_src_tree_pl, nsn);
3559 }
3560 REASON_SET(&reason, PFRES_MEMORY);
3561 return (PF_DROP);
3562 }
3563 bzero(s, sizeof(*s));
3564 r->states++;
3565 if (a != NULL)
3566 a->states++;
3567 s->rule.ptr = r;
3568 s->nat_rule.ptr = nr;
3569 if (s->nat_rule.ptr != NULL)
3570 s->nat_rule.ptr->states++;
3571 s->anchor.ptr = a;
3572 s->allow_opts = r->allow_opts;
3573 s->log = r->log & 2;
3574 s->proto = pd->proto;
3575 s->direction = direction;
3576 s->af = af;
3577 if (direction == PF_OUT) {
3578 PF_ACPY(&s->gwy.addr, saddr, af);
3579 s->gwy.port = icmpid;
3580 PF_ACPY(&s->ext.addr, daddr, af);
3581 s->ext.port = icmpid;
3582 if (nr != NULL)
3583 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3584 else
3585 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3586 s->lan.port = icmpid;
3587 } else {
3588 PF_ACPY(&s->lan.addr, daddr, af);
3589 s->lan.port = icmpid;
3590 PF_ACPY(&s->ext.addr, saddr, af);
3591 s->ext.port = icmpid;
3592 if (nr != NULL)
3593 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3594 else
3595 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3596 s->gwy.port = icmpid;
3597 }
3598#ifdef __FreeBSD__
3599 s->creation = time_second;
3600 s->expire = time_second;
3601#else
3602 s->creation = time.tv_sec;
3603 s->expire = time.tv_sec;
3604#endif
3605 s->timeout = PFTM_ICMP_FIRST_PACKET;
3606 pf_set_rt_ifp(s, saddr);
3607 if (sn != NULL) {
3608 s->src_node = sn;
3609 s->src_node->states++;
3610 }
3611 if (nsn != NULL) {
3612 PF_ACPY(&nsn->raddr, &pd->naddr, af);
3613 s->nat_src_node = nsn;
3614 s->nat_src_node->states++;
3615 }
3616 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3617 REASON_SET(&reason, PFRES_MEMORY);
3618 pf_src_tree_remove_state(s);
3619 pool_put(&pf_state_pl, s);
3620 return (PF_DROP);
3621 } else
3622 *sm = s;
3623 }
3624
3625#ifdef INET6
3626 /* copy back packet headers if we performed IPv6 NAT operations */
3627 if (rewrite)
3628 m_copyback(m, off, sizeof(struct icmp6_hdr),
3629 (caddr_t)pd->hdr.icmp6);
3630#endif /* INET6 */
3631
3632 return (PF_PASS);
3633}
3634
3635int
3636pf_test_other(struct pf_rule **rm, struct pf_state **sm, int direction,
3637 struct pfi_kif *kif, struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
3638 struct pf_rule **am, struct pf_ruleset **rsm)
3639{
3640 struct pf_rule *nr = NULL;
3641 struct pf_rule *r, *a = NULL;
3642 struct pf_ruleset *ruleset = NULL;
3643 struct pf_src_node *nsn = NULL;
3644 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
3645 sa_family_t af = pd->af;
3646 u_short reason;
3647 struct pf_tag *pftag = NULL;
3648 int tag = -1;
3649
3650 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3651
3652 if (direction == PF_OUT) {
3653 /* check outgoing packet for BINAT/NAT */
3654 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3655 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3656 PF_ACPY(&pd->baddr, saddr, af);
3657 switch (af) {
3658#ifdef INET
3659 case AF_INET:
3660 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3661 pd->naddr.v4.s_addr, 0);
3662 break;
3663#endif /* INET */
3664#ifdef INET6
3665 case AF_INET6:
3666 PF_ACPY(saddr, &pd->naddr, af);
3667 break;
3668#endif /* INET6 */
3669 }
3670 if (nr->natpass)
3671 r = NULL;
3672 pd->nat_rule = nr;
3673 }
3674 } else {
3675 /* check incoming packet for BINAT/RDR */
3676 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3677 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3678 PF_ACPY(&pd->baddr, daddr, af);
3679 switch (af) {
3680#ifdef INET
3681 case AF_INET:
3682 pf_change_a(&daddr->v4.s_addr,
3683 pd->ip_sum, pd->naddr.v4.s_addr, 0);
3684 break;
3685#endif /* INET */
3686#ifdef INET6
3687 case AF_INET6:
3688 PF_ACPY(daddr, &pd->naddr, af);
3689 break;
3690#endif /* INET6 */
3691 }
3692 if (nr->natpass)
3693 r = NULL;
3694 pd->nat_rule = nr;
3695 }
3696 }
3697
3698 while (r != NULL) {
3699 r->evaluations++;
3700 if (r->kif != NULL &&
3701 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3702 r = r->skip[PF_SKIP_IFP].ptr;
3703 else if (r->direction && r->direction != direction)
3704 r = r->skip[PF_SKIP_DIR].ptr;
3705 else if (r->af && r->af != af)
3706 r = r->skip[PF_SKIP_AF].ptr;
3707 else if (r->proto && r->proto != pd->proto)
3708 r = r->skip[PF_SKIP_PROTO].ptr;
3709 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, r->src.not))
3710 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3711 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, r->dst.not))
3712 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3713 else if (r->tos && !(r->tos & pd->tos))
3714 r = TAILQ_NEXT(r, entries);
3715 else if (r->rule_flag & PFRULE_FRAGMENT)
3716 r = TAILQ_NEXT(r, entries);
3717 else if (r->match_tag && !pf_match_tag(m, r, nr, pftag, &tag))
3718 r = TAILQ_NEXT(r, entries);
3719 else if (r->anchorname[0] && r->anchor == NULL)
3720 r = TAILQ_NEXT(r, entries);
3721 else if (r->os_fingerprint != PF_OSFP_ANY)
3722 r = TAILQ_NEXT(r, entries);
3723 else {
3724 if (r->tag)
3725 tag = r->tag;
3726 if (r->anchor == NULL) {
3727 *rm = r;
3728 *am = a;
3729 *rsm = ruleset;
3730 if ((*rm)->quick)
3731 break;
3732 r = TAILQ_NEXT(r, entries);
3733 } else
3734 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3735 PF_RULESET_FILTER);
3736 }
3737 if (r == NULL && a != NULL)
3738 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3739 PF_RULESET_FILTER);
3740 }
3741 r = *rm;
3742 a = *am;
3743 ruleset = *rsm;
3744
3745 REASON_SET(&reason, PFRES_MATCH);
3746
3747 if (r->log)
3748 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3749
3750 if ((r->action == PF_DROP) &&
3751 ((r->rule_flag & PFRULE_RETURNICMP) ||
3752 (r->rule_flag & PFRULE_RETURN))) {
3753 struct pf_addr *a = NULL;
3754
3755 if (nr != NULL) {
3756 if (direction == PF_OUT)
3757 a = saddr;
3758 else
3759 a = daddr;
3760 }
3761 if (a != NULL) {
3762 switch (af) {
3763#ifdef INET
3764 case AF_INET:
3765 pf_change_a(&a->v4.s_addr, pd->ip_sum,
3766 pd->baddr.v4.s_addr, 0);
3767 break;
3768#endif /* INET */
3769#ifdef INET6
3770 case AF_INET6:
3771 PF_ACPY(a, &pd->baddr, af);
3772 break;
3773#endif /* INET6 */
3774 }
3775 }
3776 if ((af == AF_INET) && r->return_icmp)
3777 pf_send_icmp(m, r->return_icmp >> 8,
3778 r->return_icmp & 255, af, r);
3779 else if ((af == AF_INET6) && r->return_icmp6)
3780 pf_send_icmp(m, r->return_icmp6 >> 8,
3781 r->return_icmp6 & 255, af, r);
3782 }
3783
3784 if (r->action != PF_PASS)
3785 return (PF_DROP);
3786
3787 if (pf_tag_packet(m, pftag, tag)) {
3788 REASON_SET(&reason, PFRES_MEMORY);
3789 return (PF_DROP);
3790 }
3791
3792 if (r->keep_state || nr != NULL) {
3793 /* create new state */
3794 struct pf_state *s = NULL;
3795 struct pf_src_node *sn = NULL;
3796
3797 /* check maximums */
3798 if (r->max_states && (r->states >= r->max_states))
3799 goto cleanup;
3800 /* src node for flter rule */
3801 if ((r->rule_flag & PFRULE_SRCTRACK ||
3802 r->rpool.opts & PF_POOL_STICKYADDR) &&
3803 pf_insert_src_node(&sn, r, saddr, af) != 0)
3804 goto cleanup;
3805 /* src node for translation rule */
3806 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3807 ((direction == PF_OUT &&
3808 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3809 (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
3810 goto cleanup;
3811 s = pool_get(&pf_state_pl, PR_NOWAIT);
3812 if (s == NULL) {
3813cleanup:
3814 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3815 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3816 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3817 pf_status.src_nodes--;
3818 pool_put(&pf_src_tree_pl, sn);
3819 }
3820 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3821 nsn->expire == 0) {
3822 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3823 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3824 pf_status.src_nodes--;
3825 pool_put(&pf_src_tree_pl, nsn);
3826 }
3827 REASON_SET(&reason, PFRES_MEMORY);
3828 return (PF_DROP);
3829 }
3830 bzero(s, sizeof(*s));
3831 r->states++;
3832 if (a != NULL)
3833 a->states++;
3834 s->rule.ptr = r;
3835 s->nat_rule.ptr = nr;
3836 if (s->nat_rule.ptr != NULL)
3837 s->nat_rule.ptr->states++;
3838 s->anchor.ptr = a;
3839 s->allow_opts = r->allow_opts;
3840 s->log = r->log & 2;
3841 s->proto = pd->proto;
3842 s->direction = direction;
3843 s->af = af;
3844 if (direction == PF_OUT) {
3845 PF_ACPY(&s->gwy.addr, saddr, af);
3846 PF_ACPY(&s->ext.addr, daddr, af);
3847 if (nr != NULL)
3848 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3849 else
3850 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3851 } else {
3852 PF_ACPY(&s->lan.addr, daddr, af);
3853 PF_ACPY(&s->ext.addr, saddr, af);
3854 if (nr != NULL)
3855 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3856 else
3857 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3858 }
3859 s->src.state = PFOTHERS_SINGLE;
3860 s->dst.state = PFOTHERS_NO_TRAFFIC;
3861#ifdef __FreeBSD__
3862 s->creation = time_second;
3863 s->expire = time_second;
3864#else
3865 s->creation = time.tv_sec;
3866 s->expire = time.tv_sec;
3867#endif
3868 s->timeout = PFTM_OTHER_FIRST_PACKET;
3869 pf_set_rt_ifp(s, saddr);
3870 if (sn != NULL) {
3871 s->src_node = sn;
3872 s->src_node->states++;
3873 }
3874 if (nsn != NULL) {
3875 PF_ACPY(&nsn->raddr, &pd->naddr, af);
3876 s->nat_src_node = nsn;
3877 s->nat_src_node->states++;
3878 }
3879 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3880 REASON_SET(&reason, PFRES_MEMORY);
3881 pf_src_tree_remove_state(s);
3882 pool_put(&pf_state_pl, s);
3883 return (PF_DROP);
3884 } else
3885 *sm = s;
3886 }
3887
3888 return (PF_PASS);
3889}
3890
3891int
3892pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif,
3893 struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am,
3894 struct pf_ruleset **rsm)
3895{
3896 struct pf_rule *r, *a = NULL;
3897 struct pf_ruleset *ruleset = NULL;
3898 sa_family_t af = pd->af;
3899 u_short reason;
3900 struct pf_tag *pftag = NULL;
3901 int tag = -1;
3902
3903 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3904 while (r != NULL) {
3905 r->evaluations++;
3906 if (r->kif != NULL &&
3907 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3908 r = r->skip[PF_SKIP_IFP].ptr;
3909 else if (r->direction && r->direction != direction)
3910 r = r->skip[PF_SKIP_DIR].ptr;
3911 else if (r->af && r->af != af)
3912 r = r->skip[PF_SKIP_AF].ptr;
3913 else if (r->proto && r->proto != pd->proto)
3914 r = r->skip[PF_SKIP_PROTO].ptr;
3915 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, r->src.not))
3916 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3917 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, r->dst.not))
3918 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3919 else if (r->tos && !(r->tos & pd->tos))
3920 r = TAILQ_NEXT(r, entries);
3921 else if (r->src.port_op || r->dst.port_op ||
3922 r->flagset || r->type || r->code ||
3923 r->os_fingerprint != PF_OSFP_ANY)
3924 r = TAILQ_NEXT(r, entries);
3925 else if (r->match_tag && !pf_match_tag(m, r, NULL, pftag, &tag))
3926 r = TAILQ_NEXT(r, entries);
3927 else if (r->anchorname[0] && r->anchor == NULL)
3928 r = TAILQ_NEXT(r, entries);
3929 else {
3930 if (r->anchor == NULL) {
3931 *rm = r;
3932 *am = a;
3933 *rsm = ruleset;
3934 if ((*rm)->quick)
3935 break;
3936 r = TAILQ_NEXT(r, entries);
3937 } else
3938 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3939 PF_RULESET_FILTER);
3940 }
3941 if (r == NULL && a != NULL)
3942 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3943 PF_RULESET_FILTER);
3944 }
3945 r = *rm;
3946 a = *am;
3947 ruleset = *rsm;
3948
3949 REASON_SET(&reason, PFRES_MATCH);
3950
3951 if (r->log)
3952 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3953
3954 if (r->action != PF_PASS)
3955 return (PF_DROP);
3956
3957 if (pf_tag_packet(m, pftag, tag)) {
3958 REASON_SET(&reason, PFRES_MEMORY);
3959 return (PF_DROP);
3960 }
3961
3962 return (PF_PASS);
3963}
3964
3965int
3966pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
3967 struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
3968 u_short *reason)
3969{
3970 struct pf_state key;
3971 struct tcphdr *th = pd->hdr.tcp;
3972 u_int16_t win = ntohs(th->th_win);
3973 u_int32_t ack, end, seq;
3974 u_int8_t sws, dws;
3975 int ackskew;
3976 int copyback = 0;
3977 struct pf_state_peer *src, *dst;
3978
3979 key.af = pd->af;
3980 key.proto = IPPROTO_TCP;
3981 if (direction == PF_IN) {
3982 PF_ACPY(&key.ext.addr, pd->src, key.af);
3983 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
3984 key.ext.port = th->th_sport;
3985 key.gwy.port = th->th_dport;
3986 } else {
3987 PF_ACPY(&key.lan.addr, pd->src, key.af);
3988 PF_ACPY(&key.ext.addr, pd->dst, key.af);
3989 key.lan.port = th->th_sport;
3990 key.ext.port = th->th_dport;
3991 }
3992
3993 STATE_LOOKUP();
3994
3995 if (direction == (*state)->direction) {
3996 src = &(*state)->src;
3997 dst = &(*state)->dst;
3998 } else {
3999 src = &(*state)->dst;
4000 dst = &(*state)->src;
4001 }
4002
4003 if ((*state)->src.state == PF_TCPS_PROXY_SRC) {
4004 if (direction != (*state)->direction)
4005 return (PF_SYNPROXY_DROP);
4006 if (th->th_flags & TH_SYN) {
4007 if (ntohl(th->th_seq) != (*state)->src.seqlo)
4008 return (PF_DROP);
4009 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
4010 pd->src, th->th_dport, th->th_sport,
4011 (*state)->src.seqhi, ntohl(th->th_seq) + 1,
4012 TH_SYN|TH_ACK, 0, (*state)->src.mss, 0);
4013 return (PF_SYNPROXY_DROP);
4014 } else if (!(th->th_flags & TH_ACK) ||
4015 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
4016 (ntohl(th->th_seq) != (*state)->src.seqlo + 1))
4017 return (PF_DROP);
4018 else
4019 (*state)->src.state = PF_TCPS_PROXY_DST;
4020 }
4021 if ((*state)->src.state == PF_TCPS_PROXY_DST) {
4022 struct pf_state_host *src, *dst;
4023
4024 if (direction == PF_OUT) {
4025 src = &(*state)->gwy;
4026 dst = &(*state)->ext;
4027 } else {
4028 src = &(*state)->ext;
4029 dst = &(*state)->lan;
4030 }
4031 if (direction == (*state)->direction) {
4032 if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) ||
4033 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
4034 (ntohl(th->th_seq) != (*state)->src.seqlo + 1))
4035 return (PF_DROP);
4036 (*state)->src.max_win = MAX(ntohs(th->th_win), 1);
4037 if ((*state)->dst.seqhi == 1)
4038 (*state)->dst.seqhi = arc4random();
4039 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
4040 &dst->addr, src->port, dst->port,
4041 (*state)->dst.seqhi, 0, TH_SYN, 0,
4042 (*state)->src.mss, 0);
4043 return (PF_SYNPROXY_DROP);
4044 } else if (((th->th_flags & (TH_SYN|TH_ACK)) !=
4045 (TH_SYN|TH_ACK)) ||
4046 (ntohl(th->th_ack) != (*state)->dst.seqhi + 1))
4047 return (PF_DROP);
4048 else {
4049 (*state)->dst.max_win = MAX(ntohs(th->th_win), 1);
4050 (*state)->dst.seqlo = ntohl(th->th_seq);
4051 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
4052 pd->src, th->th_dport, th->th_sport,
4053 ntohl(th->th_ack), ntohl(th->th_seq) + 1,
4054 TH_ACK, (*state)->src.max_win, 0, 0);
4055 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
4056 &dst->addr, src->port, dst->port,
4057 (*state)->src.seqhi + 1, (*state)->src.seqlo + 1,
4058 TH_ACK, (*state)->dst.max_win, 0, 0);
4059 (*state)->src.seqdiff = (*state)->dst.seqhi -
4060 (*state)->src.seqlo;
4061 (*state)->dst.seqdiff = (*state)->src.seqhi -
4062 (*state)->dst.seqlo;
4063 (*state)->src.seqhi = (*state)->src.seqlo +
4064 (*state)->src.max_win;
4065 (*state)->dst.seqhi = (*state)->dst.seqlo +
4066 (*state)->dst.max_win;
4067 (*state)->src.wscale = (*state)->dst.wscale = 0;
4068 (*state)->src.state = (*state)->dst.state =
4069 TCPS_ESTABLISHED;
4070 return (PF_SYNPROXY_DROP);
4071 }
4072 }
4073
4074 if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) {
4075 sws = src->wscale & PF_WSCALE_MASK;
4076 dws = dst->wscale & PF_WSCALE_MASK;
4077 } else
4078 sws = dws = 0;
4079
4080 /*
4081 * Sequence tracking algorithm from Guido van Rooij's paper:
4082 * http://www.madison-gurkha.com/publications/tcp_filtering/
4083 * tcp_filtering.ps
4084 */
4085
4086 seq = ntohl(th->th_seq);
4087 if (src->seqlo == 0) {
4088 /* First packet from this end. Set its state */
4089
4090 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) &&
4091 src->scrub == NULL) {
4092 if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) {
4093 REASON_SET(reason, PFRES_MEMORY);
4094 return (PF_DROP);
4095 }
4096 }
4097
4098 /* Deferred generation of sequence number modulator */
4099 if (dst->seqdiff && !src->seqdiff) {
4100 while ((src->seqdiff = arc4random()) == 0)
4101 ;
4102 ack = ntohl(th->th_ack) - dst->seqdiff;
4103 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
4104 src->seqdiff), 0);
4105 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
4106 copyback = 1;
4107 } else {
4108 ack = ntohl(th->th_ack);
4109 }
4110
4111 end = seq + pd->p_len;
4112 if (th->th_flags & TH_SYN) {
4113 end++;
4114 if (dst->wscale & PF_WSCALE_FLAG) {
4115 src->wscale = pf_get_wscale(m, off, th->th_off,
4116 pd->af);
4117 if (src->wscale & PF_WSCALE_FLAG) {
4118 /* Remove scale factor from initial
4119 * window */
4120 sws = src->wscale & PF_WSCALE_MASK;
4121 win = ((u_int32_t)win + (1 << sws) - 1)
4122 >> sws;
4123 dws = dst->wscale & PF_WSCALE_MASK;
4124 } else {
4125 /* fixup other window */
4126 dst->max_win <<= dst->wscale &
4127 PF_WSCALE_MASK;
4128 /* in case of a retrans SYN|ACK */
4129 dst->wscale = 0;
4130 }
4131 }
4132 }
4133 if (th->th_flags & TH_FIN)
4134 end++;
4135
4136 src->seqlo = seq;
4137 if (src->state < TCPS_SYN_SENT)
4138 src->state = TCPS_SYN_SENT;
4139
4140 /*
4141 * May need to slide the window (seqhi may have been set by
4142 * the crappy stack check or if we picked up the connection
4143 * after establishment)
4144 */
4145 if (src->seqhi == 1 ||
4146 SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi))
4147 src->seqhi = end + MAX(1, dst->max_win << dws);
4148 if (win > src->max_win)
4149 src->max_win = win;
4150
4151 } else {
4152 ack = ntohl(th->th_ack) - dst->seqdiff;
4153 if (src->seqdiff) {
4154 /* Modulate sequence numbers */
4155 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
4156 src->seqdiff), 0);
4157 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
4158 copyback = 1;
4159 }
4160 end = seq + pd->p_len;
4161 if (th->th_flags & TH_SYN)
4162 end++;
4163 if (th->th_flags & TH_FIN)
4164 end++;
4165 }
4166
4167 if ((th->th_flags & TH_ACK) == 0) {
4168 /* Let it pass through the ack skew check */
4169 ack = dst->seqlo;
4170 } else if ((ack == 0 &&
4171 (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) ||
4172 /* broken tcp stacks do not set ack */
4173 (dst->state < TCPS_SYN_SENT)) {
4174 /*
4175 * Many stacks (ours included) will set the ACK number in an
4176 * FIN|ACK if the SYN times out -- no sequence to ACK.
4177 */
4178 ack = dst->seqlo;
4179 }
4180
4181 if (seq == end) {
4182 /* Ease sequencing restrictions on no data packets */
4183 seq = src->seqlo;
4184 end = seq;
4185 }
4186
4187 ackskew = dst->seqlo - ack;
4188
4189#define MAXACKWINDOW (0xffff + 1500) /* 1500 is an arbitrary fudge factor */
4190 if (SEQ_GEQ(src->seqhi, end) &&
4191 /* Last octet inside other's window space */
4192 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) &&
4193 /* Retrans: not more than one window back */
4194 (ackskew >= -MAXACKWINDOW) &&
4195 /* Acking not more than one reassembled fragment backwards */
4196 (ackskew <= (MAXACKWINDOW << sws))) {
4197 /* Acking not more than one window forward */
4198
4199 /* update max window */
4200 if (src->max_win < win)
4201 src->max_win = win;
4202 /* synchronize sequencing */
4203 if (SEQ_GT(end, src->seqlo))
4204 src->seqlo = end;
4205 /* slide the window of what the other end can send */
4206 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4207 dst->seqhi = ack + MAX((win << sws), 1);
4208
4209
4210 /* update states */
4211 if (th->th_flags & TH_SYN)
4212 if (src->state < TCPS_SYN_SENT)
4213 src->state = TCPS_SYN_SENT;
4214 if (th->th_flags & TH_FIN)
4215 if (src->state < TCPS_CLOSING)
4216 src->state = TCPS_CLOSING;
4217 if (th->th_flags & TH_ACK) {
4218 if (dst->state == TCPS_SYN_SENT)
4219 dst->state = TCPS_ESTABLISHED;
4220 else if (dst->state == TCPS_CLOSING)
4221 dst->state = TCPS_FIN_WAIT_2;
4222 }
4223 if (th->th_flags & TH_RST)
4224 src->state = dst->state = TCPS_TIME_WAIT;
4225
4226 /* update expire time */
4227#ifdef __FreeBSD__
4228 (*state)->expire = time_second;
4229#else
4230 (*state)->expire = time.tv_sec;
4231#endif
4232 if (src->state >= TCPS_FIN_WAIT_2 &&
4233 dst->state >= TCPS_FIN_WAIT_2)
4234 (*state)->timeout = PFTM_TCP_CLOSED;
4235 else if (src->state >= TCPS_FIN_WAIT_2 ||
4236 dst->state >= TCPS_FIN_WAIT_2)
4237 (*state)->timeout = PFTM_TCP_FIN_WAIT;
4238 else if (src->state < TCPS_ESTABLISHED ||
4239 dst->state < TCPS_ESTABLISHED)
4240 (*state)->timeout = PFTM_TCP_OPENING;
4241 else if (src->state >= TCPS_CLOSING ||
4242 dst->state >= TCPS_CLOSING)
4243 (*state)->timeout = PFTM_TCP_CLOSING;
4244 else
4245 (*state)->timeout = PFTM_TCP_ESTABLISHED;
4246
4247 /* Fall through to PASS packet */
4248
4249 } else if ((dst->state < TCPS_SYN_SENT ||
4250 dst->state >= TCPS_FIN_WAIT_2 ||
4251 src->state >= TCPS_FIN_WAIT_2) &&
4252 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) &&
4253 /* Within a window forward of the originating packet */
4254 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) {
4255 /* Within a window backward of the originating packet */
4256
4257 /*
4258 * This currently handles three situations:
4259 * 1) Stupid stacks will shotgun SYNs before their peer
4260 * replies.
4261 * 2) When PF catches an already established stream (the
4262 * firewall rebooted, the state table was flushed, routes
4263 * changed...)
4264 * 3) Packets get funky immediately after the connection
4265 * closes (this should catch Solaris spurious ACK|FINs
4266 * that web servers like to spew after a close)
4267 *
4268 * This must be a little more careful than the above code
4269 * since packet floods will also be caught here. We don't
4270 * update the TTL here to mitigate the damage of a packet
4271 * flood and so the same code can handle awkward establishment
4272 * and a loosened connection close.
4273 * In the establishment case, a correct peer response will
4274 * validate the connection, go through the normal state code
4275 * and keep updating the state TTL.
4276 */
4277
4278 if (pf_status.debug >= PF_DEBUG_MISC) {
4279 printf("pf: loose state match: ");
4280 pf_print_state(*state);
4281 pf_print_flags(th->th_flags);
4282 printf(" seq=%u ack=%u len=%u ackskew=%d pkts=%d:%d\n",
4283 seq, ack, pd->p_len, ackskew,
4284 (*state)->packets[0], (*state)->packets[1]);
4285 }
4286
4287 /* update max window */
4288 if (src->max_win < win)
4289 src->max_win = win;
4290 /* synchronize sequencing */
4291 if (SEQ_GT(end, src->seqlo))
4292 src->seqlo = end;
4293 /* slide the window of what the other end can send */
4294 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4295 dst->seqhi = ack + MAX((win << sws), 1);
4296
4297 /*
4298 * Cannot set dst->seqhi here since this could be a shotgunned
4299 * SYN and not an already established connection.
4300 */
4301
4302 if (th->th_flags & TH_FIN)
4303 if (src->state < TCPS_CLOSING)
4304 src->state = TCPS_CLOSING;
4305 if (th->th_flags & TH_RST)
4306 src->state = dst->state = TCPS_TIME_WAIT;
4307
4308 /* Fall through to PASS packet */
4309
4310 } else {
4311 if ((*state)->dst.state == TCPS_SYN_SENT &&
4312 (*state)->src.state == TCPS_SYN_SENT) {
4313 /* Send RST for state mismatches during handshake */
4314 if (!(th->th_flags & TH_RST)) {
4315 u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
4316
4317 if (th->th_flags & TH_SYN)
4318 ack++;
4319 if (th->th_flags & TH_FIN)
4320 ack++;
4321 pf_send_tcp((*state)->rule.ptr, pd->af,
4322 pd->dst, pd->src, th->th_dport,
4323 th->th_sport, ntohl(th->th_ack), ack,
4324 TH_RST|TH_ACK, 0, 0,
4325 (*state)->rule.ptr->return_ttl);
4326 }
4327 src->seqlo = 0;
4328 src->seqhi = 1;
4329 src->max_win = 1;
4330 } else if (pf_status.debug >= PF_DEBUG_MISC) {
4331 printf("pf: BAD state: ");
4332 pf_print_state(*state);
4333 pf_print_flags(th->th_flags);
4334 printf(" seq=%u ack=%u len=%u ackskew=%d pkts=%d:%d "
4335 "dir=%s,%s\n", seq, ack, pd->p_len, ackskew,
4336 (*state)->packets[0], (*state)->packets[1],
4337 direction == PF_IN ? "in" : "out",
4338 direction == (*state)->direction ? "fwd" : "rev");
4339 printf("pf: State failure on: %c %c %c %c | %c %c\n",
4340 SEQ_GEQ(src->seqhi, end) ? ' ' : '1',
4341 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ?
4342 ' ': '2',
4343 (ackskew >= -MAXACKWINDOW) ? ' ' : '3',
4344 (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4',
4345 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5',
4346 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6');
4347 }
4348 return (PF_DROP);
4349 }
4350
4351 if (dst->scrub || src->scrub) {
4352 if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4353 src, dst, ©back))
4354 return (PF_DROP);
4355 }
4356
4357 /* Any packets which have gotten here are to be passed */
4358
4359 /* translate source/destination address, if necessary */
4360 if (STATE_TRANSLATE(*state)) {
4361 if (direction == PF_OUT)
4362 pf_change_ap(pd->src, &th->th_sport, pd->ip_sum,
4363 &th->th_sum, &(*state)->gwy.addr,
4364 (*state)->gwy.port, 0, pd->af);
4365 else
4366 pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum,
4367 &th->th_sum, &(*state)->lan.addr,
4368 (*state)->lan.port, 0, pd->af);
4369 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4370 } else if (copyback) {
4371 /* Copyback sequence modulation or stateful scrub changes */
4372 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4373 }
4374
4375 return (PF_PASS);
4376}
4377
4378int
4379pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif,
4380 struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4381{
4382 struct pf_state_peer *src, *dst;
4383 struct pf_state key;
4384 struct udphdr *uh = pd->hdr.udp;
4385
4386 key.af = pd->af;
4387 key.proto = IPPROTO_UDP;
4388 if (direction == PF_IN) {
4389 PF_ACPY(&key.ext.addr, pd->src, key.af);
4390 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4391 key.ext.port = uh->uh_sport;
4392 key.gwy.port = uh->uh_dport;
4393 } else {
4394 PF_ACPY(&key.lan.addr, pd->src, key.af);
4395 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4396 key.lan.port = uh->uh_sport;
4397 key.ext.port = uh->uh_dport;
4398 }
4399
4400 STATE_LOOKUP();
4401
4402 if (direction == (*state)->direction) {
4403 src = &(*state)->src;
4404 dst = &(*state)->dst;
4405 } else {
4406 src = &(*state)->dst;
4407 dst = &(*state)->src;
4408 }
4409
4410 /* update states */
4411 if (src->state < PFUDPS_SINGLE)
4412 src->state = PFUDPS_SINGLE;
4413 if (dst->state == PFUDPS_SINGLE)
4414 dst->state = PFUDPS_MULTIPLE;
4415
4416 /* update expire time */
4417#ifdef __FreeBSD__
4418 (*state)->expire = time_second;
4419#else
4420 (*state)->expire = time.tv_sec;
4421#endif
4422 if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE)
4423 (*state)->timeout = PFTM_UDP_MULTIPLE;
4424 else
4425 (*state)->timeout = PFTM_UDP_SINGLE;
4426
4427 /* translate source/destination address, if necessary */
4428 if (STATE_TRANSLATE(*state)) {
4429 if (direction == PF_OUT)
4430 pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum,
4431 &uh->uh_sum, &(*state)->gwy.addr,
4432 (*state)->gwy.port, 1, pd->af);
4433 else
4434 pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum,
4435 &uh->uh_sum, &(*state)->lan.addr,
4436 (*state)->lan.port, 1, pd->af);
4437 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
4438 }
4439
4440 return (PF_PASS);
4441}
4442
4443int
4444pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif,
4445 struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4446{
4447 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
4448 u_int16_t icmpid = 0; /* make the compiler happy */
4449 u_int16_t *icmpsum = NULL; /* make the compiler happy */
4450 u_int8_t icmptype = 0; /* make the compiler happy */
4451 int state_icmp = 0;
4452
4453 switch (pd->proto) {
4454#ifdef INET
4455 case IPPROTO_ICMP:
4456 icmptype = pd->hdr.icmp->icmp_type;
4457 icmpid = pd->hdr.icmp->icmp_id;
4458 icmpsum = &pd->hdr.icmp->icmp_cksum;
4459
4460 if (icmptype == ICMP_UNREACH ||
4461 icmptype == ICMP_SOURCEQUENCH ||
4462 icmptype == ICMP_REDIRECT ||
4463 icmptype == ICMP_TIMXCEED ||
4464 icmptype == ICMP_PARAMPROB)
4465 state_icmp++;
4466 break;
4467#endif /* INET */
4468#ifdef INET6
4469 case IPPROTO_ICMPV6:
4470 icmptype = pd->hdr.icmp6->icmp6_type;
4471 icmpid = pd->hdr.icmp6->icmp6_id;
4472 icmpsum = &pd->hdr.icmp6->icmp6_cksum;
4473
4474 if (icmptype == ICMP6_DST_UNREACH ||
4475 icmptype == ICMP6_PACKET_TOO_BIG ||
4476 icmptype == ICMP6_TIME_EXCEEDED ||
4477 icmptype == ICMP6_PARAM_PROB)
4478 state_icmp++;
4479 break;
4480#endif /* INET6 */
4481 }
4482
4483 if (!state_icmp) {
4484
4485 /*
4486 * ICMP query/reply message not related to a TCP/UDP packet.
4487 * Search for an ICMP state.
4488 */
4489 struct pf_state key;
4490
4491 key.af = pd->af;
4492 key.proto = pd->proto;
4493 if (direction == PF_IN) {
4494 PF_ACPY(&key.ext.addr, pd->src, key.af);
4495 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4496 key.ext.port = icmpid;
4497 key.gwy.port = icmpid;
4498 } else {
4499 PF_ACPY(&key.lan.addr, pd->src, key.af);
4500 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4501 key.lan.port = icmpid;
4502 key.ext.port = icmpid;
4503 }
4504
4505 STATE_LOOKUP();
4506
4507#ifdef __FreeBSD__
4508 (*state)->expire = time_second;
4509#else
4510 (*state)->expire = time.tv_sec;
4511#endif
4512 (*state)->timeout = PFTM_ICMP_ERROR_REPLY;
4513
4514 /* translate source/destination address, if necessary */
4515 if (PF_ANEQ(&(*state)->lan.addr, &(*state)->gwy.addr, pd->af)) {
4516 if (direction == PF_OUT) {
4517 switch (pd->af) {
4518#ifdef INET
4519 case AF_INET:
4520 pf_change_a(&saddr->v4.s_addr,
4521 pd->ip_sum,
4522 (*state)->gwy.addr.v4.s_addr, 0);
4523 break;
4524#endif /* INET */
4525#ifdef INET6
4526 case AF_INET6:
4527 pf_change_a6(saddr,
4528 &pd->hdr.icmp6->icmp6_cksum,
4529 &(*state)->gwy.addr, 0);
4530 m_copyback(m, off,
4531 sizeof(struct icmp6_hdr),
4532 (caddr_t)pd->hdr.icmp6);
4533 break;
4534#endif /* INET6 */
4535 }
4536 } else {
4537 switch (pd->af) {
4538#ifdef INET
4539 case AF_INET:
4540 pf_change_a(&daddr->v4.s_addr,
4541 pd->ip_sum,
4542 (*state)->lan.addr.v4.s_addr, 0);
4543 break;
4544#endif /* INET */
4545#ifdef INET6
4546 case AF_INET6:
4547 pf_change_a6(daddr,
4548 &pd->hdr.icmp6->icmp6_cksum,
4549 &(*state)->lan.addr, 0);
4550 m_copyback(m, off,
4551 sizeof(struct icmp6_hdr),
4552 (caddr_t)pd->hdr.icmp6);
4553 break;
4554#endif /* INET6 */
4555 }
4556 }
4557 }
4558
4559 return (PF_PASS);
4560
4561 } else {
4562 /*
4563 * ICMP error message in response to a TCP/UDP packet.
4564 * Extract the inner TCP/UDP header and search for that state.
4565 */
4566
4567 struct pf_pdesc pd2;
4568#ifdef INET
4569 struct ip h2;
4570#endif /* INET */
4571#ifdef INET6
4572 struct ip6_hdr h2_6;
4573 int terminal = 0;
4574#endif /* INET6 */
4575 int ipoff2 = 0; /* make the compiler happy */
4576 int off2 = 0; /* make the compiler happy */
4577
4578 pd2.af = pd->af;
4579 switch (pd->af) {
4580#ifdef INET
4581 case AF_INET:
4582 /* offset of h2 in mbuf chain */
4583 ipoff2 = off + ICMP_MINLEN;
4584
4585 if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2),
4586 NULL, NULL, pd2.af)) {
4587 DPFPRINTF(PF_DEBUG_MISC,
4588 ("pf: ICMP error message too short "
4589 "(ip)\n"));
4590 return (PF_DROP);
4591 }
4592 /*
4593 * ICMP error messages don't refer to non-first
4594 * fragments
4595 */
4596 if (h2.ip_off & htons(IP_OFFMASK))
4597 return (PF_DROP);
4598
4599 /* offset of protocol header that follows h2 */
4600 off2 = ipoff2 + (h2.ip_hl << 2);
4601
4602 pd2.proto = h2.ip_p;
4603 pd2.src = (struct pf_addr *)&h2.ip_src;
4604 pd2.dst = (struct pf_addr *)&h2.ip_dst;
4605 pd2.ip_sum = &h2.ip_sum;
4606 break;
4607#endif /* INET */
4608#ifdef INET6
4609 case AF_INET6:
4610 ipoff2 = off + sizeof(struct icmp6_hdr);
4611
4612 if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6),
4613 NULL, NULL, pd2.af)) {
4614 DPFPRINTF(PF_DEBUG_MISC,
4615 ("pf: ICMP error message too short "
4616 "(ip6)\n"));
4617 return (PF_DROP);
4618 }
4619 pd2.proto = h2_6.ip6_nxt;
4620 pd2.src = (struct pf_addr *)&h2_6.ip6_src;
4621 pd2.dst = (struct pf_addr *)&h2_6.ip6_dst;
4622 pd2.ip_sum = NULL;
4623 off2 = ipoff2 + sizeof(h2_6);
4624 do {
4625 switch (pd2.proto) {
4626 case IPPROTO_FRAGMENT:
4627 /*
4628 * ICMPv6 error messages for
4629 * non-first fragments
4630 */
4631 return (PF_DROP);
4632 case IPPROTO_AH:
4633 case IPPROTO_HOPOPTS:
4634 case IPPROTO_ROUTING:
4635 case IPPROTO_DSTOPTS: {
4636 /* get next header and header length */
4637 struct ip6_ext opt6;
4638
4639 if (!pf_pull_hdr(m, off2, &opt6,
4640 sizeof(opt6), NULL, NULL, pd2.af)) {
4641 DPFPRINTF(PF_DEBUG_MISC,
4642 ("pf: ICMPv6 short opt\n"));
4643 return (PF_DROP);
4644 }
4645 if (pd2.proto == IPPROTO_AH)
4646 off2 += (opt6.ip6e_len + 2) * 4;
4647 else
4648 off2 += (opt6.ip6e_len + 1) * 8;
4649 pd2.proto = opt6.ip6e_nxt;
4650 /* goto the next header */
4651 break;
4652 }
4653 default:
4654 terminal++;
4655 break;
4656 }
4657 } while (!terminal);
4658 break;
4659#endif /* INET6 */
4660 }
4661
4662 switch (pd2.proto) {
4663 case IPPROTO_TCP: {
4664 struct tcphdr th;
4665 u_int32_t seq;
4666 struct pf_state key;
4667 struct pf_state_peer *src, *dst;
4668 u_int8_t dws;
4669 int copyback = 0;
4670
4671 /*
4672 * Only the first 8 bytes of the TCP header can be
4673 * expected. Don't access any TCP header fields after
4674 * th_seq, an ackskew test is not possible.
4675 */
4676 if (!pf_pull_hdr(m, off2, &th, 8, NULL, NULL, pd2.af)) {
4677 DPFPRINTF(PF_DEBUG_MISC,
4678 ("pf: ICMP error message too short "
4679 "(tcp)\n"));
4680 return (PF_DROP);
4681 }
4682
4683 key.af = pd2.af;
4684 key.proto = IPPROTO_TCP;
4685 if (direction == PF_IN) {
4686 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4687 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4688 key.ext.port = th.th_dport;
4689 key.gwy.port = th.th_sport;
4690 } else {
4691 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4692 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4693 key.lan.port = th.th_dport;
4694 key.ext.port = th.th_sport;
4695 }
4696
4697 STATE_LOOKUP();
4698
4699 if (direction == (*state)->direction) {
4700 src = &(*state)->dst;
4701 dst = &(*state)->src;
4702 } else {
4703 src = &(*state)->src;
4704 dst = &(*state)->dst;
4705 }
4706
4707 if (src->wscale && dst->wscale &&
4708 !(th.th_flags & TH_SYN))
4709 dws = dst->wscale & PF_WSCALE_MASK;
4710 else
4711 dws = 0;
4712
4713 /* Demodulate sequence number */
4714 seq = ntohl(th.th_seq) - src->seqdiff;
4715 if (src->seqdiff) {
4716 pf_change_a(&th.th_seq, icmpsum,
4717 htonl(seq), 0);
4718 copyback = 1;
4719 }
4720
4721 if (!SEQ_GEQ(src->seqhi, seq) ||
4722 !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws))) {
4723 if (pf_status.debug >= PF_DEBUG_MISC) {
4724 printf("pf: BAD ICMP %d:%d ",
4725 icmptype, pd->hdr.icmp->icmp_code);
4726 pf_print_host(pd->src, 0, pd->af);
4727 printf(" -> ");
4728 pf_print_host(pd->dst, 0, pd->af);
4729 printf(" state: ");
4730 pf_print_state(*state);
4731 printf(" seq=%u\n", seq);
4732 }
4733 return (PF_DROP);
4734 }
4735
4736 if (STATE_TRANSLATE(*state)) {
4737 if (direction == PF_IN) {
4738 pf_change_icmp(pd2.src, &th.th_sport,
4739 daddr, &(*state)->lan.addr,
4740 (*state)->lan.port, NULL,
4741 pd2.ip_sum, icmpsum,
4742 pd->ip_sum, 0, pd2.af);
4743 } else {
4744 pf_change_icmp(pd2.dst, &th.th_dport,
4745 saddr, &(*state)->gwy.addr,
4746 (*state)->gwy.port, NULL,
4747 pd2.ip_sum, icmpsum,
4748 pd->ip_sum, 0, pd2.af);
4749 }
4750 copyback = 1;
4751 }
4752
4753 if (copyback) {
4754 switch (pd2.af) {
4755#ifdef INET
4756 case AF_INET:
4757 m_copyback(m, off, ICMP_MINLEN,
4758 (caddr_t)pd->hdr.icmp);
4759 m_copyback(m, ipoff2, sizeof(h2),
4760 (caddr_t)&h2);
4761 break;
4762#endif /* INET */
4763#ifdef INET6
4764 case AF_INET6:
4765 m_copyback(m, off,
4766 sizeof(struct icmp6_hdr),
4767 (caddr_t)pd->hdr.icmp6);
4768 m_copyback(m, ipoff2, sizeof(h2_6),
4769 (caddr_t)&h2_6);
4770 break;
4771#endif /* INET6 */
4772 }
4773 m_copyback(m, off2, 8, (caddr_t)&th);
4774 }
4775
4776 return (PF_PASS);
4777 break;
4778 }
4779 case IPPROTO_UDP: {
4780 struct udphdr uh;
4781 struct pf_state key;
4782
4783 if (!pf_pull_hdr(m, off2, &uh, sizeof(uh),
4784 NULL, NULL, pd2.af)) {
4785 DPFPRINTF(PF_DEBUG_MISC,
4786 ("pf: ICMP error message too short "
4787 "(udp)\n"));
4788 return (PF_DROP);
4789 }
4790
4791 key.af = pd2.af;
4792 key.proto = IPPROTO_UDP;
4793 if (direction == PF_IN) {
4794 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4795 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4796 key.ext.port = uh.uh_dport;
4797 key.gwy.port = uh.uh_sport;
4798 } else {
4799 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4800 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4801 key.lan.port = uh.uh_dport;
4802 key.ext.port = uh.uh_sport;
4803 }
4804
4805 STATE_LOOKUP();
4806
4807 if (STATE_TRANSLATE(*state)) {
4808 if (direction == PF_IN) {
4809 pf_change_icmp(pd2.src, &uh.uh_sport,
4810 daddr, &(*state)->lan.addr,
4811 (*state)->lan.port, &uh.uh_sum,
4812 pd2.ip_sum, icmpsum,
4813 pd->ip_sum, 1, pd2.af);
4814 } else {
4815 pf_change_icmp(pd2.dst, &uh.uh_dport,
4816 saddr, &(*state)->gwy.addr,
4817 (*state)->gwy.port, &uh.uh_sum,
4818 pd2.ip_sum, icmpsum,
4819 pd->ip_sum, 1, pd2.af);
4820 }
4821 switch (pd2.af) {
4822#ifdef INET
4823 case AF_INET:
4824 m_copyback(m, off, ICMP_MINLEN,
4825 (caddr_t)pd->hdr.icmp);
4826 m_copyback(m, ipoff2, sizeof(h2),
4827 (caddr_t)&h2);
4828 break;
4829#endif /* INET */
4830#ifdef INET6
4831 case AF_INET6:
4832 m_copyback(m, off,
4833 sizeof(struct icmp6_hdr),
4834 (caddr_t)pd->hdr.icmp6);
4835 m_copyback(m, ipoff2, sizeof(h2_6),
4836 (caddr_t)&h2_6);
4837 break;
4838#endif /* INET6 */
4839 }
4840 m_copyback(m, off2, sizeof(uh),
4841 (caddr_t)&uh);
4842 }
4843
4844 return (PF_PASS);
4845 break;
4846 }
4847#ifdef INET
4848 case IPPROTO_ICMP: {
4849 struct icmp iih;
4850 struct pf_state key;
4851
4852 if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN,
4853 NULL, NULL, pd2.af)) {
4854 DPFPRINTF(PF_DEBUG_MISC,
4855 ("pf: ICMP error message too short i"
4856 "(icmp)\n"));
4857 return (PF_DROP);
4858 }
4859
4860 key.af = pd2.af;
4861 key.proto = IPPROTO_ICMP;
4862 if (direction == PF_IN) {
4863 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4864 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4865 key.ext.port = iih.icmp_id;
4866 key.gwy.port = iih.icmp_id;
4867 } else {
4868 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4869 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4870 key.lan.port = iih.icmp_id;
4871 key.ext.port = iih.icmp_id;
4872 }
4873
4874 STATE_LOOKUP();
4875
4876 if (STATE_TRANSLATE(*state)) {
4877 if (direction == PF_IN) {
4878 pf_change_icmp(pd2.src, &iih.icmp_id,
4879 daddr, &(*state)->lan.addr,
4880 (*state)->lan.port, NULL,
4881 pd2.ip_sum, icmpsum,
4882 pd->ip_sum, 0, AF_INET);
4883 } else {
4884 pf_change_icmp(pd2.dst, &iih.icmp_id,
4885 saddr, &(*state)->gwy.addr,
4886 (*state)->gwy.port, NULL,
4887 pd2.ip_sum, icmpsum,
4888 pd->ip_sum, 0, AF_INET);
4889 }
4890 m_copyback(m, off, ICMP_MINLEN,
4891 (caddr_t)pd->hdr.icmp);
4892 m_copyback(m, ipoff2, sizeof(h2),
4893 (caddr_t)&h2);
4894 m_copyback(m, off2, ICMP_MINLEN,
4895 (caddr_t)&iih);
4896 }
4897
4898 return (PF_PASS);
4899 break;
4900 }
4901#endif /* INET */
4902#ifdef INET6
4903 case IPPROTO_ICMPV6: {
4904 struct icmp6_hdr iih;
4905 struct pf_state key;
4906
4907 if (!pf_pull_hdr(m, off2, &iih,
4908 sizeof(struct icmp6_hdr), NULL, NULL, pd2.af)) {
4909 DPFPRINTF(PF_DEBUG_MISC,
4910 ("pf: ICMP error message too short "
4911 "(icmp6)\n"));
4912 return (PF_DROP);
4913 }
4914
4915 key.af = pd2.af;
4916 key.proto = IPPROTO_ICMPV6;
4917 if (direction == PF_IN) {
4918 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4919 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4920 key.ext.port = iih.icmp6_id;
4921 key.gwy.port = iih.icmp6_id;
4922 } else {
4923 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4924 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4925 key.lan.port = iih.icmp6_id;
4926 key.ext.port = iih.icmp6_id;
4927 }
4928
4929 STATE_LOOKUP();
4930
4931 if (STATE_TRANSLATE(*state)) {
4932 if (direction == PF_IN) {
4933 pf_change_icmp(pd2.src, &iih.icmp6_id,
4934 daddr, &(*state)->lan.addr,
4935 (*state)->lan.port, NULL,
4936 pd2.ip_sum, icmpsum,
4937 pd->ip_sum, 0, AF_INET6);
4938 } else {
4939 pf_change_icmp(pd2.dst, &iih.icmp6_id,
4940 saddr, &(*state)->gwy.addr,
4941 (*state)->gwy.port, NULL,
4942 pd2.ip_sum, icmpsum,
4943 pd->ip_sum, 0, AF_INET6);
4944 }
4945 m_copyback(m, off, sizeof(struct icmp6_hdr),
4946 (caddr_t)pd->hdr.icmp6);
4947 m_copyback(m, ipoff2, sizeof(h2_6),
4948 (caddr_t)&h2_6);
4949 m_copyback(m, off2, sizeof(struct icmp6_hdr),
4950 (caddr_t)&iih);
4951 }
4952
4953 return (PF_PASS);
4954 break;
4955 }
4956#endif /* INET6 */
4957 default: {
4958 struct pf_state key;
4959
4960 key.af = pd2.af;
4961 key.proto = pd2.proto;
4962 if (direction == PF_IN) {
4963 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4964 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4965 key.ext.port = 0;
4966 key.gwy.port = 0;
4967 } else {
4968 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4969 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4970 key.lan.port = 0;
4971 key.ext.port = 0;
4972 }
4973
4974 STATE_LOOKUP();
4975
4976 if (STATE_TRANSLATE(*state)) {
4977 if (direction == PF_IN) {
4978 pf_change_icmp(pd2.src, NULL,
4979 daddr, &(*state)->lan.addr,
4980 0, NULL,
4981 pd2.ip_sum, icmpsum,
4982 pd->ip_sum, 0, pd2.af);
4983 } else {
4984 pf_change_icmp(pd2.dst, NULL,
4985 saddr, &(*state)->gwy.addr,
4986 0, NULL,
4987 pd2.ip_sum, icmpsum,
4988 pd->ip_sum, 0, pd2.af);
4989 }
4990 switch (pd2.af) {
4991#ifdef INET
4992 case AF_INET:
4993 m_copyback(m, off, ICMP_MINLEN,
4994 (caddr_t)pd->hdr.icmp);
4995 m_copyback(m, ipoff2, sizeof(h2),
4996 (caddr_t)&h2);
4997 break;
4998#endif /* INET */
4999#ifdef INET6
5000 case AF_INET6:
5001 m_copyback(m, off,
5002 sizeof(struct icmp6_hdr),
5003 (caddr_t)pd->hdr.icmp6);
5004 m_copyback(m, ipoff2, sizeof(h2_6),
5005 (caddr_t)&h2_6);
5006 break;
5007#endif /* INET6 */
5008 }
5009 }
5010
5011 return (PF_PASS);
5012 break;
5013 }
5014 }
5015 }
5016}
5017
5018int
5019pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif,
5020 struct pf_pdesc *pd)
5021{
5022 struct pf_state_peer *src, *dst;
5023 struct pf_state key;
5024
5025 key.af = pd->af;
5026 key.proto = pd->proto;
5027 if (direction == PF_IN) {
5028 PF_ACPY(&key.ext.addr, pd->src, key.af);
5029 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
5030 key.ext.port = 0;
5031 key.gwy.port = 0;
5032 } else {
5033 PF_ACPY(&key.lan.addr, pd->src, key.af);
5034 PF_ACPY(&key.ext.addr, pd->dst, key.af);
5035 key.lan.port = 0;
5036 key.ext.port = 0;
5037 }
5038
5039 STATE_LOOKUP();
5040
5041 if (direction == (*state)->direction) {
5042 src = &(*state)->src;
5043 dst = &(*state)->dst;
5044 } else {
5045 src = &(*state)->dst;
5046 dst = &(*state)->src;
5047 }
5048
5049 /* update states */
5050 if (src->state < PFOTHERS_SINGLE)
5051 src->state = PFOTHERS_SINGLE;
5052 if (dst->state == PFOTHERS_SINGLE)
5053 dst->state = PFOTHERS_MULTIPLE;
5054
5055 /* update expire time */
5056#ifdef __FreeBSD__
5057 (*state)->expire = time_second;
5058#else
5059 (*state)->expire = time.tv_sec;
5060#endif
5061 if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE)
5062 (*state)->timeout = PFTM_OTHER_MULTIPLE;
5063 else
5064 (*state)->timeout = PFTM_OTHER_SINGLE;
5065
5066 /* translate source/destination address, if necessary */
5067 if (STATE_TRANSLATE(*state)) {
5068 if (direction == PF_OUT)
5069 switch (pd->af) {
5070#ifdef INET
5071 case AF_INET:
5072 pf_change_a(&pd->src->v4.s_addr,
5073 pd->ip_sum, (*state)->gwy.addr.v4.s_addr,
5074 0);
5075 break;
5076#endif /* INET */
5077#ifdef INET6
5078 case AF_INET6:
5079 PF_ACPY(pd->src, &(*state)->gwy.addr, pd->af);
5080 break;
5081#endif /* INET6 */
5082 }
5083 else
5084 switch (pd->af) {
5085#ifdef INET
5086 case AF_INET:
5087 pf_change_a(&pd->dst->v4.s_addr,
5088 pd->ip_sum, (*state)->lan.addr.v4.s_addr,
5089 0);
5090 break;
5091#endif /* INET */
5092#ifdef INET6
5093 case AF_INET6:
5094 PF_ACPY(pd->dst, &(*state)->lan.addr, pd->af);
5095 break;
5096#endif /* INET6 */
5097 }
5098 }
5099
5100 return (PF_PASS);
5101}
5102
5103/*
5104 * ipoff and off are measured from the start of the mbuf chain.
5105 * h must be at "ipoff" on the mbuf chain.
5106 */
5107void *
5108pf_pull_hdr(struct mbuf *m, int off, void *p, int len,
5109 u_short *actionp, u_short *reasonp, sa_family_t af)
5110{
5111 switch (af) {
5112#ifdef INET
5113 case AF_INET: {
5114 struct ip *h = mtod(m, struct ip *);
5115 u_int16_t fragoff = (ntohs(h->ip_off) & IP_OFFMASK) << 3;
5116
5117 if (fragoff) {
5118 if (fragoff >= len)
5119 ACTION_SET(actionp, PF_PASS);
5120 else {
5121 ACTION_SET(actionp, PF_DROP);
5122 REASON_SET(reasonp, PFRES_FRAG);
5123 }
5124 return (NULL);
5125 }
5126 if (m->m_pkthdr.len < off + len ||
5127 ntohs(h->ip_len) < off + len) {
5128 ACTION_SET(actionp, PF_DROP);
5129 REASON_SET(reasonp, PFRES_SHORT);
5130 return (NULL);
5131 }
5132 break;
5133 }
5134#endif /* INET */
5135#ifdef INET6
5136 case AF_INET6: {
5137 struct ip6_hdr *h = mtod(m, struct ip6_hdr *);
5138
5139 if (m->m_pkthdr.len < off + len ||
5140 (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) <
5141 (unsigned)(off + len)) {
5142 ACTION_SET(actionp, PF_DROP);
5143 REASON_SET(reasonp, PFRES_SHORT);
5144 return (NULL);
5145 }
5146 break;
5147 }
5148#endif /* INET6 */
5149 }
5150 m_copydata(m, off, len, p);
5151 return (p);
5152}
5153
5154int
5155pf_routable(struct pf_addr *addr, sa_family_t af)
5156{
5157 struct sockaddr_in *dst;
5158 struct route ro;
5159 int ret = 0;
5160
5161 bzero(&ro, sizeof(ro));
5162 dst = satosin(&ro.ro_dst);
5163 dst->sin_family = af;
5164 dst->sin_len = sizeof(*dst);
5165 dst->sin_addr = addr->v4;
5166#ifdef __FreeBSD__
5167#ifdef RTF_PRCLONING
5168 rtalloc_ign(&ro, (RTF_CLONING|RTF_PRCLONING));
5169#else /* !RTF_PRCLONING */
5170 rtalloc_ign(&ro, RTF_CLONING);
5171#endif
5172#else /* ! __FreeBSD__ */
5173 rtalloc_noclone(&ro, NO_CLONING);
5174#endif
5175
5176 if (ro.ro_rt != NULL) {
5177 ret = 1;
5178 RTFREE(ro.ro_rt);
5179 }
5180
5181 return (ret);
5182}
5183
5184#ifdef INET
5185
5186void
5187pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5188 struct pf_state *s)
5189{
5190 struct mbuf *m0, *m1;
5191 struct m_tag *mtag;
5192 struct route iproute;
5193 struct route *ro = NULL; /* XXX: was uninitialized */
5194 struct sockaddr_in *dst;
5195 struct ip *ip;
5196 struct ifnet *ifp = NULL;
5197 struct pf_addr naddr;
5198 struct pf_src_node *sn = NULL;
5199 int error = 0;
5200#ifdef __FreeBSD__
5201 int sw_csum;
5202#endif
5203
5204 if (m == NULL || *m == NULL || r == NULL ||
5205 (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5206 panic("pf_route: invalid parameters");
5207
5208 if ((mtag = m_tag_find(*m, PACKET_TAG_PF_ROUTED, NULL)) == NULL) {
5209 if ((mtag = m_tag_get(PACKET_TAG_PF_ROUTED, 1, M_NOWAIT)) ==
5210 NULL) {
5211 m0 = *m;
5212 *m = NULL;
5213 goto bad;
5214 }
5215 *(char *)(mtag + 1) = 1;
5216 m_tag_prepend(*m, mtag);
5217 } else {
5218 if (*(char *)(mtag + 1) > 3) {
5219 m0 = *m;
5220 *m = NULL;
5221 goto bad;
5222 }
5223 (*(char *)(mtag + 1))++;
5224 }
5225
5226 if (r->rt == PF_DUPTO) {
5227#ifdef __FreeBSD__
5228 if ((m0 = m_dup(*m, M_DONTWAIT)) == NULL)
5229#else
5230 if ((m0 = m_copym2(*m, 0, M_COPYALL, M_NOWAIT)) == NULL)
5231#endif
5232 return;
5233#ifdef __FreeBSD__
5234 if ((mtag = m_tag_copy(mtag, M_DONTWAIT)) == NULL)
5235#else
5236 if ((mtag = m_tag_copy(mtag)) == NULL)
5237#endif
5238 goto bad;
5239 m_tag_prepend(m0, mtag);
5240 } else {
5241 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5242 return;
5243 m0 = *m;
5244 }
5245
5246 if (m0->m_len < sizeof(struct ip))
5247 panic("pf_route: m0->m_len < sizeof(struct ip)");
5248 ip = mtod(m0, struct ip *);
5249
5250 ro = &iproute;
5251 bzero((caddr_t)ro, sizeof(*ro));
5252 dst = satosin(&ro->ro_dst);
5253 dst->sin_family = AF_INET;
5254 dst->sin_len = sizeof(*dst);
5255 dst->sin_addr = ip->ip_dst;
5256
5257 if (r->rt == PF_FASTROUTE) {
5258 rtalloc(ro);
5259 if (ro->ro_rt == 0) {
5260 ipstat.ips_noroute++;
5261 goto bad;
5262 }
5263
5264 ifp = ro->ro_rt->rt_ifp;
5265 ro->ro_rt->rt_use++;
5266
5267 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
5268 dst = satosin(ro->ro_rt->rt_gateway);
5269 } else {
5270 if (TAILQ_EMPTY(&r->rpool.list))
5271 panic("pf_route: TAILQ_EMPTY(&r->rpool.list)");
5272 if (s == NULL) {
5273 pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src,
5274 &naddr, NULL, &sn);
5275 if (!PF_AZERO(&naddr, AF_INET))
5276 dst->sin_addr.s_addr = naddr.v4.s_addr;
5277 ifp = r->rpool.cur->kif ?
5278 r->rpool.cur->kif->pfik_ifp : NULL;
5279 } else {
5280 if (!PF_AZERO(&s->rt_addr, AF_INET))
5281 dst->sin_addr.s_addr =
5282 s->rt_addr.v4.s_addr;
5283 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5284 }
5285 }
5286 if (ifp == NULL)
5287 goto bad;
5288
5289 if (oifp != ifp) {
5290#ifdef __FreeBSD__
5291 PF_UNLOCK();
5292 if (pf_test(PF_OUT, ifp, &m0, NULL) != PF_PASS) {
5293 PF_LOCK();
5294 goto bad;
5295 } else if (m0 == NULL) {
5296 PF_LOCK();
5297 goto done;
5298 }
5299 PF_LOCK();
5300#else
5301 if (pf_test(PF_OUT, ifp, &m0) != PF_PASS)
5302 goto bad;
5303 else if (m0 == NULL)
5304 goto done;
5305#endif
5306 if (m0->m_len < sizeof(struct ip))
5307 panic("pf_route: m0->m_len < sizeof(struct ip)");
5308 ip = mtod(m0, struct ip *);
5309 }
5310
5311#ifdef __FreeBSD__
5312 /* Copied from FreeBSD 5.1-CURRENT ip_output. */
5313 m0->m_pkthdr.csum_flags |= CSUM_IP;
5314 sw_csum = m0->m_pkthdr.csum_flags & ~ifp->if_hwassist;
5315 if (sw_csum & CSUM_DELAY_DATA) {
5316 /*
5317 * XXX: in_delayed_cksum assumes HBO for ip->ip_len (at least)
5318 */
5319 NTOHS(ip->ip_len);
5320 NTOHS(ip->ip_off); /* XXX: needed? */
5321 in_delayed_cksum(m0);
5322 HTONS(ip->ip_len);
5323 HTONS(ip->ip_off);
5324 sw_csum &= ~CSUM_DELAY_DATA;
5325 }
5326 m0->m_pkthdr.csum_flags &= ifp->if_hwassist;
5327
5328 if (ntohs(ip->ip_len) <= ifp->if_mtu ||
5329 (ifp->if_hwassist & CSUM_FRAGMENT &&
5330 ((ip->ip_off & htons(IP_DF)) == 0))) {
5331 /*
5332 * ip->ip_len = htons(ip->ip_len);
5333 * ip->ip_off = htons(ip->ip_off);
5334 */
5335 ip->ip_sum = 0;
5336 if (sw_csum & CSUM_DELAY_IP) {
5337 /* From KAME */
5338 if (ip->ip_v == IPVERSION &&
5339 (ip->ip_hl << 2) == sizeof(*ip)) {
5340 ip->ip_sum = in_cksum_hdr(ip);
5341 } else {
5342 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
5343 }
5344 }
5345 PF_UNLOCK();
5346 error = (*ifp->if_output)(ifp, m0, sintosa(dst), ro->ro_rt);
5347 PF_LOCK();
5348 goto done;
5349 }
5350
5351#else
5352 /* Copied from ip_output. */
5353#ifdef IPSEC
5354 /*
5355 * If deferred crypto processing is needed, check that the
5356 * interface supports it.
5357 */
5358 if ((mtag = m_tag_find(m0, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL))
5359 != NULL && (ifp->if_capabilities & IFCAP_IPSEC) == 0) {
5360 /* Notify IPsec to do its own crypto. */
5361 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1));
5362 goto bad;
5363 }
5364#endif /* IPSEC */
5365
5366 /* Catch routing changes wrt. hardware checksumming for TCP or UDP. */
5367 if (m0->m_pkthdr.csum & M_TCPV4_CSUM_OUT) {
5368 if (!(ifp->if_capabilities & IFCAP_CSUM_TCPv4) ||
5369 ifp->if_bridge != NULL) {
5370 in_delayed_cksum(m0);
5371 m0->m_pkthdr.csum &= ~M_TCPV4_CSUM_OUT; /* Clear */
5372 }
5373 } else if (m0->m_pkthdr.csum & M_UDPV4_CSUM_OUT) {
5374 if (!(ifp->if_capabilities & IFCAP_CSUM_UDPv4) ||
5375 ifp->if_bridge != NULL) {
5376 in_delayed_cksum(m0);
5377 m0->m_pkthdr.csum &= ~M_UDPV4_CSUM_OUT; /* Clear */
5378 }
5379 }
5380
5381 if (ntohs(ip->ip_len) <= ifp->if_mtu) {
5382 if ((ifp->if_capabilities & IFCAP_CSUM_IPv4) &&
5383 ifp->if_bridge == NULL) {
5384 m0->m_pkthdr.csum |= M_IPV4_CSUM_OUT;
5385 ipstat.ips_outhwcsum++;
5386 } else {
5387 ip->ip_sum = 0;
5388 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
5389 }
5390 /* Update relevant hardware checksum stats for TCP/UDP */
5391 if (m0->m_pkthdr.csum & M_TCPV4_CSUM_OUT)
5392 tcpstat.tcps_outhwcsum++;
5393 else if (m0->m_pkthdr.csum & M_UDPV4_CSUM_OUT)
5394 udpstat.udps_outhwcsum++;
5395 error = (*ifp->if_output)(ifp, m0, sintosa(dst), NULL);
5396 goto done;
5397 }
5398#endif
5399 /*
5400 * Too large for interface; fragment if possible.
5401 * Must be able to put at least 8 bytes per fragment.
5402 */
5403 if (ip->ip_off & htons(IP_DF)) {
5404 ipstat.ips_cantfrag++;
5405 if (r->rt != PF_DUPTO) {
5406#ifdef __FreeBSD__
5407 /* icmp_error() expects host byte ordering */
5408 NTOHS(ip->ip_len);
5409 NTOHS(ip->ip_off);
5410 PF_UNLOCK();
5411#endif
5412 icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0,
5413 ifp);
5414#ifdef __FreeBSD__
5415 PF_LOCK();
5416#endif
5417 goto done;
5418 } else
5419 goto bad;
5420 }
5421
5422 m1 = m0;
5423#ifdef __FreeBSD__
5424 /*
5425 * XXX: is cheaper + less error prone than own function
5426 */
5427 NTOHS(ip->ip_len);
5428 NTOHS(ip->ip_off);
5429 error = ip_fragment(ip, &m0, ifp->if_mtu, ifp->if_hwassist, sw_csum);
5430#else
5431 error = ip_fragment(m0, ifp, ifp->if_mtu);
5432#endif
5433 if (error) {
5434#ifndef __FreeBSD__ /* ip_fragment does not do m_freem() on FreeBSD */
5435 m0 = NULL;
5436#endif
5437 goto bad;
5438 }
5439
5440 for (m0 = m1; m0; m0 = m1) {
5441 m1 = m0->m_nextpkt;
5442 m0->m_nextpkt = 0;
5443#ifdef __FreeBSD__
5444 if (error == 0) {
5445 PF_UNLOCK();
5446 error = (*ifp->if_output)(ifp, m0, sintosa(dst),
5447 NULL);
5448 PF_LOCK();
5449 } else
5450#else
5451 if (error == 0)
5452 error = (*ifp->if_output)(ifp, m0, sintosa(dst),
5453 NULL);
5454 else
5455#endif
5456 m_freem(m0);
5457 }
5458
5459 if (error == 0)
5460 ipstat.ips_fragmented++;
5461
5462done:
5463 if (r->rt != PF_DUPTO)
5464 *m = NULL;
5465 if (ro == &iproute && ro->ro_rt)
5466 RTFREE(ro->ro_rt);
5467 return;
5468
5469bad:
5470 m_freem(m0);
5471 goto done;
5472}
5473#endif /* INET */
5474
5475#ifdef INET6
5476void
5477pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5478 struct pf_state *s)
5479{
5480 struct mbuf *m0;
5481 struct m_tag *mtag;
5482 struct route_in6 ip6route;
5483 struct route_in6 *ro;
5484 struct sockaddr_in6 *dst;
5485 struct ip6_hdr *ip6;
5486 struct ifnet *ifp = NULL;
5487 struct pf_addr naddr;
5488 struct pf_src_node *sn = NULL;
5489 int error = 0;
5490
5491 if (m == NULL || *m == NULL || r == NULL ||
5492 (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5493 panic("pf_route6: invalid parameters");
5494
5495 if ((mtag = m_tag_find(*m, PACKET_TAG_PF_ROUTED, NULL)) == NULL) {
5496 if ((mtag = m_tag_get(PACKET_TAG_PF_ROUTED, 1, M_NOWAIT)) ==
5497 NULL) {
5498 m0 = *m;
5499 *m = NULL;
5500 goto bad;
5501 }
5502 *(char *)(mtag + 1) = 1;
5503 m_tag_prepend(*m, mtag);
5504 } else {
5505 if (*(char *)(mtag + 1) > 3) {
5506 m0 = *m;
5507 *m = NULL;
5508 goto bad;
5509 }
5510 (*(char *)(mtag + 1))++;
5511 }
5512
5513 if (r->rt == PF_DUPTO) {
5514#ifdef __FreeBSD__
5515 if ((m0 = m_dup(*m, M_DONTWAIT)) == NULL)
5516#else
5517 if ((m0 = m_copym2(*m, 0, M_COPYALL, M_NOWAIT)) == NULL)
5518#endif
5519 return;
5520#ifdef __FreeBSD__
5521 if ((mtag = m_tag_copy(mtag, M_DONTWAIT)) == NULL)
5522#else
5523 if ((mtag = m_tag_copy(mtag)) == NULL)
5524#endif
5525 goto bad;
5526 m_tag_prepend(m0, mtag);
5527 } else {
5528 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5529 return;
5530 m0 = *m;
5531 }
5532
5533 if (m0->m_len < sizeof(struct ip6_hdr))
5534 panic("pf_route6: m0->m_len < sizeof(struct ip6_hdr)");
5535 ip6 = mtod(m0, struct ip6_hdr *);
5536
5537 ro = &ip6route;
5538 bzero((caddr_t)ro, sizeof(*ro));
5539 dst = (struct sockaddr_in6 *)&ro->ro_dst;
5540 dst->sin6_family = AF_INET6;
5541 dst->sin6_len = sizeof(*dst);
5542 dst->sin6_addr = ip6->ip6_dst;
5543
5544 /* Cheat. */
5545 if (r->rt == PF_FASTROUTE) {
5546#ifdef __FreeBSD__
5547 m0->m_flags |= M_SKIP_FIREWALL;
5548 PF_UNLOCK();
5549 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
5550 PF_LOCK();
5551#else
5552 mtag = m_tag_get(PACKET_TAG_PF_GENERATED, 0, M_NOWAIT);
5553 if (mtag == NULL)
5554 goto bad;
5555 m_tag_prepend(m0, mtag);
5556 ip6_output(m0, NULL, NULL, 0, NULL, NULL);
5557#endif
5558 return;
5559 }
5560
5561 if (TAILQ_EMPTY(&r->rpool.list))
5562 panic("pf_route6: TAILQ_EMPTY(&r->rpool.list)");
5563 if (s == NULL) {
5564 pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src,
5565 &naddr, NULL, &sn);
5566 if (!PF_AZERO(&naddr, AF_INET6))
5567 PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5568 &naddr, AF_INET6);
5569 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL;
5570 } else {
5571 if (!PF_AZERO(&s->rt_addr, AF_INET6))
5572 PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5573 &s->rt_addr, AF_INET6);
5574 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5575 }
5576 if (ifp == NULL)
5577 goto bad;
5578
5579 if (oifp != ifp) {
5580#ifdef __FreeBSD__
5581 PF_UNLOCK();
5582 if (pf_test6(PF_OUT, ifp, &m0, NULL) != PF_PASS) {
5583 PF_LOCK();
5584 goto bad;
5585 } else if (m0 == NULL) {
5586 PF_LOCK();
5587 goto done;
5588 }
5589 PF_LOCK();
5590#else
5591 if (pf_test6(PF_OUT, ifp, &m0) != PF_PASS)
5592 goto bad;
5593 else if (m0 == NULL)
5594 goto done;
5595#endif
5596 if (m0->m_len < sizeof(struct ip6_hdr))
5597 panic("pf_route6: m0->m_len < sizeof(struct ip6_hdr)");
5598 ip6 = mtod(m0, struct ip6_hdr *);
5599 }
5600
5601 /*
5602 * If the packet is too large for the outgoing interface,
5603 * send back an icmp6 error.
5604 */
5605 if (IN6_IS_ADDR_LINKLOCAL(&dst->sin6_addr))
5606 dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
5607 if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) {
5608#ifdef __FreeBSD__
5609 PF_UNLOCK();
5610#endif
5611 error = nd6_output(ifp, ifp, m0, dst, NULL);
5612#ifdef __FreeBSD__
5613 PF_LOCK();
5614#endif
5615 } else {
5616 in6_ifstat_inc(ifp, ifs6_in_toobig);
5617#ifdef __FreeBSD__
5618 if (r->rt != PF_DUPTO) {
5619 PF_UNLOCK();
5620 icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
5621 PF_LOCK();
5622 } else
5623#else
5624 if (r->rt != PF_DUPTO)
5625 icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
5626 else
5627#endif
5628 goto bad;
5629 }
5630
5631done:
5632 if (r->rt != PF_DUPTO)
5633 *m = NULL;
5634 return;
5635
5636bad:
5637 m_freem(m0);
5638 goto done;
5639}
5640#endif /* INET6 */
5641
5642
5643#ifdef __FreeBSD__
5644/*
5645 * FreeBSD supports cksum offloads for the following drivers.
|