1/* $FreeBSD: head/sys/contrib/pf/net/pf.c 130613 2004-06-16 23:24:02Z mlaier $ */
2/* $OpenBSD: pf.c,v 1.433 2004/03/26 22:20:57 dhartmei Exp $ */
3
4/*
5 * Copyright (c) 2001 Daniel Hartmeier
6 * Copyright (c) 2002,2003 Henning Brauer
7 * All rights reserved.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 *
13 * - Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * - Redistributions in binary form must reproduce the above
16 * copyright notice, this list of conditions and the following
17 * disclaimer in the documentation and/or other materials provided
18 * with the distribution.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
26 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
28 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
30 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
31 * POSSIBILITY OF SUCH DAMAGE.
32 *
33 * Effort sponsored in part by the Defense Advanced Research Projects
34 * Agency (DARPA) and Air Force Research Laboratory, Air Force
35 * Materiel Command, USAF, under agreement number F30602-01-2-0537.
36 *
37 */
38
39#ifdef __FreeBSD__
40#include "opt_inet.h"
41#include "opt_inet6.h"
42#endif
43
44#ifdef __FreeBSD__
45#include "opt_bpf.h"
46#include "opt_pf.h"
47#define NBPFILTER DEV_BPF
48#define NPFLOG DEV_PFLOG
49#define NPFSYNC DEV_PFSYNC
50#else
51#include "bpfilter.h"
52#include "pflog.h"
53#include "pfsync.h"
54#endif
55
56#include <sys/param.h>
57#include <sys/systm.h>
58#include <sys/mbuf.h>
59#include <sys/filio.h>
60#include <sys/socket.h>
61#include <sys/socketvar.h>
62#include <sys/kernel.h>
63#include <sys/time.h>
64#ifdef __FreeBSD__
65#include <sys/sysctl.h>
66#include <sys/endian.h>
67#else
68#include <sys/pool.h>
69#endif
70
71#include <net/if.h>
72#include <net/if_types.h>
73#include <net/bpf.h>
74#include <net/route.h>
75
76#include <netinet/in.h>
77#include <netinet/in_var.h>
78#include <netinet/in_systm.h>
79#include <netinet/ip.h>
80#include <netinet/ip_var.h>
81#include <netinet/tcp.h>
82#include <netinet/tcp_seq.h>
83#include <netinet/udp.h>
84#include <netinet/ip_icmp.h>
85#include <netinet/in_pcb.h>
86#include <netinet/tcp_timer.h>
87#include <netinet/tcp_var.h>
88#include <netinet/udp_var.h>
89#include <netinet/icmp_var.h>
90
91#ifndef __FreeBSD__
92#include <dev/rndvar.h>
93#endif
94#include <net/pfvar.h>
95#include <net/if_pflog.h>
96
97#if NPFSYNC > 0
98#include <net/if_pfsync.h>
99#endif /* NPFSYNC > 0 */
100
101#ifdef INET6
102#include <netinet/ip6.h>
103#include <netinet/in_pcb.h>
104#include <netinet/icmp6.h>
105#include <netinet6/nd6.h>
106#ifdef __FreeBSD__
107#include <netinet6/ip6_var.h>
108#include <netinet6/in6_pcb.h>
109#endif
110#endif /* INET6 */
111
112#ifdef __FreeBSD__
113#include <machine/in_cksum.h>
114#include <sys/limits.h>
115#include <sys/ucred.h>
116
117extern int ip_optcopy(struct ip *, struct ip *);
118#endif
119
120#define DPFPRINTF(n, x) if (pf_status.debug >= (n)) printf x
121
122/*
123 * Global variables
124 */
125
126struct pf_anchorqueue pf_anchors;
127struct pf_ruleset pf_main_ruleset;
128struct pf_altqqueue pf_altqs[2];
129struct pf_palist pf_pabuf;
130struct pf_altqqueue *pf_altqs_active;
131struct pf_altqqueue *pf_altqs_inactive;
132struct pf_status pf_status;
133
134u_int32_t ticket_altqs_active;
135u_int32_t ticket_altqs_inactive;
136int altqs_inactive_open;
137u_int32_t ticket_pabuf;
138
139#ifdef __FreeBSD__
140struct callout pf_expire_to; /* expire timeout */
141#else
142struct timeout pf_expire_to; /* expire timeout */
143#endif
144
145
146#ifdef __FreeBSD__
147uma_zone_t pf_src_tree_pl, pf_rule_pl;
148uma_zone_t pf_state_pl, pf_altq_pl, pf_pooladdr_pl;
149#else
150struct pool pf_src_tree_pl, pf_rule_pl;
151struct pool pf_state_pl, pf_altq_pl, pf_pooladdr_pl;
152#endif
153
154void pf_print_host(struct pf_addr *, u_int16_t, u_int8_t);
155void pf_print_state(struct pf_state *);
156void pf_print_flags(u_int8_t);
157
158u_int16_t pf_cksum_fixup(u_int16_t, u_int16_t, u_int16_t,
159 u_int8_t);
160void pf_change_ap(struct pf_addr *, u_int16_t *,
161 u_int16_t *, u_int16_t *, struct pf_addr *,
162 u_int16_t, u_int8_t, sa_family_t);
163#ifdef INET6
164void pf_change_a6(struct pf_addr *, u_int16_t *,
165 struct pf_addr *, u_int8_t);
166#endif /* INET6 */
167void pf_change_icmp(struct pf_addr *, u_int16_t *,
168 struct pf_addr *, struct pf_addr *, u_int16_t,
169 u_int16_t *, u_int16_t *, u_int16_t *,
170 u_int16_t *, u_int8_t, sa_family_t);
171void pf_send_tcp(const struct pf_rule *, sa_family_t,
172 const struct pf_addr *, const struct pf_addr *,
173 u_int16_t, u_int16_t, u_int32_t, u_int32_t,
174 u_int8_t, u_int16_t, u_int16_t, u_int8_t);
175void pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t,
176 sa_family_t, struct pf_rule *);
177struct pf_rule *pf_match_translation(struct pf_pdesc *, struct mbuf *,
178 int, int, struct pfi_kif *,
179 struct pf_addr *, u_int16_t, struct pf_addr *,
180 u_int16_t, int);
181struct pf_rule *pf_get_translation(struct pf_pdesc *, struct mbuf *,
182 int, int, struct pfi_kif *, struct pf_src_node **,
183 struct pf_addr *, u_int16_t,
184 struct pf_addr *, u_int16_t,
185 struct pf_addr *, u_int16_t *);
186int pf_test_tcp(struct pf_rule **, struct pf_state **,
187 int, struct pfi_kif *, struct mbuf *, int,
188 void *, struct pf_pdesc *, struct pf_rule **,
189 struct pf_ruleset **);
190int pf_test_udp(struct pf_rule **, struct pf_state **,
191 int, struct pfi_kif *, struct mbuf *, int,
192 void *, struct pf_pdesc *, struct pf_rule **,
193 struct pf_ruleset **);
194int pf_test_icmp(struct pf_rule **, struct pf_state **,
195 int, struct pfi_kif *, struct mbuf *, int,
196 void *, struct pf_pdesc *, struct pf_rule **,
197 struct pf_ruleset **);
198int pf_test_other(struct pf_rule **, struct pf_state **,
199 int, struct pfi_kif *, struct mbuf *, int, void *,
200 struct pf_pdesc *, struct pf_rule **,
201 struct pf_ruleset **);
202int pf_test_fragment(struct pf_rule **, int,
203 struct pfi_kif *, struct mbuf *, void *,
204 struct pf_pdesc *, struct pf_rule **,
205 struct pf_ruleset **);
206int pf_test_state_tcp(struct pf_state **, int,
207 struct pfi_kif *, struct mbuf *, int,
208 void *, struct pf_pdesc *, u_short *);
209int pf_test_state_udp(struct pf_state **, int,
210 struct pfi_kif *, struct mbuf *, int,
211 void *, struct pf_pdesc *);
212int pf_test_state_icmp(struct pf_state **, int,
213 struct pfi_kif *, struct mbuf *, int,
214 void *, struct pf_pdesc *);
215int pf_test_state_other(struct pf_state **, int,
216 struct pfi_kif *, struct pf_pdesc *);
217struct pf_tag *pf_get_tag(struct mbuf *);
218int pf_match_tag(struct mbuf *, struct pf_rule *,
219 struct pf_rule *, struct pf_tag *, int *);
220void pf_hash(struct pf_addr *, struct pf_addr *,
221 struct pf_poolhashkey *, sa_family_t);
222int pf_map_addr(u_int8_t, struct pf_rule *,
223 struct pf_addr *, struct pf_addr *,
224 struct pf_addr *, struct pf_src_node **);
225int pf_get_sport(sa_family_t, u_int8_t, struct pf_rule *,
226 struct pf_addr *, struct pf_addr *, u_int16_t,
227 struct pf_addr *, u_int16_t*, u_int16_t, u_int16_t,
228 struct pf_src_node **);
229void pf_route(struct mbuf **, struct pf_rule *, int,
230 struct ifnet *, struct pf_state *);
231void pf_route6(struct mbuf **, struct pf_rule *, int,
232 struct ifnet *, struct pf_state *);
233int pf_socket_lookup(uid_t *, gid_t *,
234 int, struct pf_pdesc *);
235u_int8_t pf_get_wscale(struct mbuf *, int, u_int16_t,
236 sa_family_t);
237u_int16_t pf_get_mss(struct mbuf *, int, u_int16_t,
238 sa_family_t);
239u_int16_t pf_calc_mss(struct pf_addr *, sa_family_t,
240 u_int16_t);
241void pf_set_rt_ifp(struct pf_state *,
242 struct pf_addr *);
243int pf_check_proto_cksum(struct mbuf *, int, int,
244 u_int8_t, sa_family_t);
245int pf_addr_wrap_neq(struct pf_addr_wrap *,
246 struct pf_addr_wrap *);
247static int pf_add_mbuf_tag(struct mbuf *, u_int);
248struct pf_state *pf_find_state_recurse(struct pfi_kif *,
249 struct pf_state *, u_int8_t);
250
251#ifdef __FreeBSD__
252int in4_cksum(struct mbuf *m, u_int8_t nxt, int off, int len);
253
254struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX];
255#else
256struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] = {
257 { &pf_state_pl, PFSTATE_HIWAT },
258 { &pf_src_tree_pl, PFSNODE_HIWAT },
259 { &pf_frent_pl, PFFRAG_FRENT_HIWAT }
260};
261#endif
262
263#define STATE_LOOKUP() \
264 do { \
265 if (direction == PF_IN) \
266 *state = pf_find_state_recurse( \
267 kif, &key, PF_EXT_GWY); \
268 else \
269 *state = pf_find_state_recurse( \
270 kif, &key, PF_LAN_EXT); \
271 if (*state == NULL) \
272 return (PF_DROP); \
273 if (direction == PF_OUT && \
274 (((*state)->rule.ptr->rt == PF_ROUTETO && \
275 (*state)->rule.ptr->direction == PF_OUT) || \
276 ((*state)->rule.ptr->rt == PF_REPLYTO && \
277 (*state)->rule.ptr->direction == PF_IN)) && \
278 (*state)->rt_kif != NULL && \
279 (*state)->rt_kif != kif) \
280 return (PF_PASS); \
281 } while (0)
282
283#define STATE_TRANSLATE(s) \
284 (s)->lan.addr.addr32[0] != (s)->gwy.addr.addr32[0] || \
285 ((s)->af == AF_INET6 && \
286 ((s)->lan.addr.addr32[1] != (s)->gwy.addr.addr32[1] || \
287 (s)->lan.addr.addr32[2] != (s)->gwy.addr.addr32[2] || \
288 (s)->lan.addr.addr32[3] != (s)->gwy.addr.addr32[3])) || \
289 (s)->lan.port != (s)->gwy.port
290
291#define BOUND_IFACE(r, k) (((r)->rule_flag & PFRULE_IFBOUND) ? (k) : \
292 ((r)->rule_flag & PFRULE_GRBOUND) ? (k)->pfik_parent : \
293 (k)->pfik_parent->pfik_parent)
294
295static __inline int pf_src_compare(struct pf_src_node *, struct pf_src_node *);
296static __inline int pf_state_compare_lan_ext(struct pf_state *,
297 struct pf_state *);
298static __inline int pf_state_compare_ext_gwy(struct pf_state *,
299 struct pf_state *);
300static __inline int pf_state_compare_id(struct pf_state *,
301 struct pf_state *);
302
303struct pf_src_tree tree_src_tracking;
304
305struct pf_state_tree_id tree_id;
306struct pf_state_queue state_updates;
307
308RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare);
309RB_GENERATE(pf_state_tree_lan_ext, pf_state,
310 u.s.entry_lan_ext, pf_state_compare_lan_ext);
311RB_GENERATE(pf_state_tree_ext_gwy, pf_state,
312 u.s.entry_ext_gwy, pf_state_compare_ext_gwy);
313RB_GENERATE(pf_state_tree_id, pf_state,
314 u.s.entry_id, pf_state_compare_id);
315
316#ifdef __FreeBSD__
317static int
318#else
319static __inline int
320#endif
321pf_src_compare(struct pf_src_node *a, struct pf_src_node *b)
322{
323 int diff;
324
325 if (a->rule.ptr > b->rule.ptr)
326 return (1);
327 if (a->rule.ptr < b->rule.ptr)
328 return (-1);
329 if ((diff = a->af - b->af) != 0)
330 return (diff);
331 switch (a->af) {
332#ifdef INET
333 case AF_INET:
334 if (a->addr.addr32[0] > b->addr.addr32[0])
335 return (1);
336 if (a->addr.addr32[0] < b->addr.addr32[0])
337 return (-1);
338 break;
339#endif /* INET */
340#ifdef INET6
341 case AF_INET6:
342 if (a->addr.addr32[3] > b->addr.addr32[3])
343 return (1);
344 if (a->addr.addr32[3] < b->addr.addr32[3])
345 return (-1);
346 if (a->addr.addr32[2] > b->addr.addr32[2])
347 return (1);
348 if (a->addr.addr32[2] < b->addr.addr32[2])
349 return (-1);
350 if (a->addr.addr32[1] > b->addr.addr32[1])
351 return (1);
352 if (a->addr.addr32[1] < b->addr.addr32[1])
353 return (-1);
354 if (a->addr.addr32[0] > b->addr.addr32[0])
355 return (1);
356 if (a->addr.addr32[0] < b->addr.addr32[0])
357 return (-1);
358 break;
359#endif /* INET6 */
360 }
361 return (0);
362}
363
364#ifdef __FreeBSD__
365static int
366#else
367static __inline int
368#endif
369pf_state_compare_lan_ext(struct pf_state *a, struct pf_state *b)
370{
371 int diff;
372
373 if ((diff = a->proto - b->proto) != 0)
374 return (diff);
375 if ((diff = a->af - b->af) != 0)
376 return (diff);
377 switch (a->af) {
378#ifdef INET
379 case AF_INET:
380 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
381 return (1);
382 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
383 return (-1);
384 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
385 return (1);
386 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
387 return (-1);
388 break;
389#endif /* INET */
390#ifdef INET6
391 case AF_INET6:
392 if (a->lan.addr.addr32[3] > b->lan.addr.addr32[3])
393 return (1);
394 if (a->lan.addr.addr32[3] < b->lan.addr.addr32[3])
395 return (-1);
396 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
397 return (1);
398 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
399 return (-1);
400 if (a->lan.addr.addr32[2] > b->lan.addr.addr32[2])
401 return (1);
402 if (a->lan.addr.addr32[2] < b->lan.addr.addr32[2])
403 return (-1);
404 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
405 return (1);
406 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
407 return (-1);
408 if (a->lan.addr.addr32[1] > b->lan.addr.addr32[1])
409 return (1);
410 if (a->lan.addr.addr32[1] < b->lan.addr.addr32[1])
411 return (-1);
412 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
413 return (1);
414 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
415 return (-1);
416 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
417 return (1);
418 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
419 return (-1);
420 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
421 return (1);
422 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
423 return (-1);
424 break;
425#endif /* INET6 */
426 }
427
428 if ((diff = a->lan.port - b->lan.port) != 0)
429 return (diff);
430 if ((diff = a->ext.port - b->ext.port) != 0)
431 return (diff);
432
433 return (0);
434}
435
436#ifdef __FreeBSD__
437static int
438#else
439static __inline int
440#endif
441pf_state_compare_ext_gwy(struct pf_state *a, struct pf_state *b)
442{
443 int diff;
444
445 if ((diff = a->proto - b->proto) != 0)
446 return (diff);
447 if ((diff = a->af - b->af) != 0)
448 return (diff);
449 switch (a->af) {
450#ifdef INET
451 case AF_INET:
452 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
453 return (1);
454 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
455 return (-1);
456 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
457 return (1);
458 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
459 return (-1);
460 break;
461#endif /* INET */
462#ifdef INET6
463 case AF_INET6:
464 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
465 return (1);
466 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
467 return (-1);
468 if (a->gwy.addr.addr32[3] > b->gwy.addr.addr32[3])
469 return (1);
470 if (a->gwy.addr.addr32[3] < b->gwy.addr.addr32[3])
471 return (-1);
472 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
473 return (1);
474 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
475 return (-1);
476 if (a->gwy.addr.addr32[2] > b->gwy.addr.addr32[2])
477 return (1);
478 if (a->gwy.addr.addr32[2] < b->gwy.addr.addr32[2])
479 return (-1);
480 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
481 return (1);
482 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
483 return (-1);
484 if (a->gwy.addr.addr32[1] > b->gwy.addr.addr32[1])
485 return (1);
486 if (a->gwy.addr.addr32[1] < b->gwy.addr.addr32[1])
487 return (-1);
488 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
489 return (1);
490 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
491 return (-1);
492 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
493 return (1);
494 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
495 return (-1);
496 break;
497#endif /* INET6 */
498 }
499
500 if ((diff = a->ext.port - b->ext.port) != 0)
501 return (diff);
502 if ((diff = a->gwy.port - b->gwy.port) != 0)
503 return (diff);
504
505 return (0);
506}
507
508#ifdef __FreeBSD__
509static int
510#else
511static __inline int
512#endif
513pf_state_compare_id(struct pf_state *a, struct pf_state *b)
514{
515 if (a->id > b->id)
516 return (1);
517 if (a->id < b->id)
518 return (-1);
519 if (a->creatorid > b->creatorid)
520 return (1);
521 if (a->creatorid < b->creatorid)
522 return (-1);
523
524 return (0);
525}
526
527#ifdef INET6
528void
529pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af)
530{
531 switch (af) {
532#ifdef INET
533 case AF_INET:
534 dst->addr32[0] = src->addr32[0];
535 break;
536#endif /* INET */
537 case AF_INET6:
538 dst->addr32[0] = src->addr32[0];
539 dst->addr32[1] = src->addr32[1];
540 dst->addr32[2] = src->addr32[2];
541 dst->addr32[3] = src->addr32[3];
542 break;
543 }
544}
545#endif
546
547struct pf_state *
548pf_find_state_byid(struct pf_state *key)
549{
550 pf_status.fcounters[FCNT_STATE_SEARCH]++;
551 return (RB_FIND(pf_state_tree_id, &tree_id, key));
552}
553
554struct pf_state *
555pf_find_state_recurse(struct pfi_kif *kif, struct pf_state *key, u_int8_t tree)
556{
557 struct pf_state *s;
558
559 pf_status.fcounters[FCNT_STATE_SEARCH]++;
560
561 switch (tree) {
562 case PF_LAN_EXT:
563 for (; kif != NULL; kif = kif->pfik_parent) {
564 s = RB_FIND(pf_state_tree_lan_ext,
565 &kif->pfik_lan_ext, key);
566 if (s != NULL)
567 return (s);
568 }
569 return (NULL);
570 case PF_EXT_GWY:
571 for (; kif != NULL; kif = kif->pfik_parent) {
572 s = RB_FIND(pf_state_tree_ext_gwy,
573 &kif->pfik_ext_gwy, key);
574 if (s != NULL)
575 return (s);
576 }
577 return (NULL);
578 default:
579 panic("pf_find_state_recurse");
580 }
581}
582
583struct pf_state *
584pf_find_state_all(struct pf_state *key, u_int8_t tree, int *more)
585{
586 struct pf_state *s, *ss = NULL;
587 struct pfi_kif *kif;
588
589 pf_status.fcounters[FCNT_STATE_SEARCH]++;
590
591 switch (tree) {
592 case PF_LAN_EXT:
593 TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states) {
594 s = RB_FIND(pf_state_tree_lan_ext,
595 &kif->pfik_lan_ext, key);
596 if (s == NULL)
597 continue;
598 if (more == NULL)
599 return (s);
600 ss = s;
601 (*more)++;
602 }
603 return (ss);
604 case PF_EXT_GWY:
605 TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states) {
606 s = RB_FIND(pf_state_tree_ext_gwy,
607 &kif->pfik_ext_gwy, key);
608 if (s == NULL)
609 continue;
610 if (more == NULL)
611 return (s);
612 ss = s;
613 (*more)++;
614 }
615 return (ss);
616 default:
617 panic("pf_find_state_all");
618 }
619}
620
621int
622pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule,
623 struct pf_addr *src, sa_family_t af)
624{
625 struct pf_src_node k;
626
627 if (*sn == NULL) {
628 k.af = af;
629 PF_ACPY(&k.addr, src, af);
630 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
631 rule->rpool.opts & PF_POOL_STICKYADDR)
632 k.rule.ptr = rule;
633 else
634 k.rule.ptr = NULL;
635 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
636 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
637 }
638 if (*sn == NULL) {
639 if (!rule->max_src_nodes ||
640 rule->src_nodes < rule->max_src_nodes)
641 (*sn) = pool_get(&pf_src_tree_pl, PR_NOWAIT);
642 if ((*sn) == NULL)
643 return (-1);
644 bzero(*sn, sizeof(struct pf_src_node));
645 (*sn)->af = af;
646 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
647 rule->rpool.opts & PF_POOL_STICKYADDR)
648 (*sn)->rule.ptr = rule;
649 else
650 (*sn)->rule.ptr = NULL;
651 PF_ACPY(&(*sn)->addr, src, af);
652 if (RB_INSERT(pf_src_tree,
653 &tree_src_tracking, *sn) != NULL) {
654 if (pf_status.debug >= PF_DEBUG_MISC) {
655 printf("pf: src_tree insert failed: ");
656 pf_print_host(&(*sn)->addr, 0, af);
657 printf("\n");
658 }
659 pool_put(&pf_src_tree_pl, *sn);
660 return (-1);
661 }
662#ifdef __FreeBSD__
663 (*sn)->creation = time_second;
664#else
665 (*sn)->creation = time.tv_sec;
666#endif
667 (*sn)->ruletype = rule->action;
668 if ((*sn)->rule.ptr != NULL)
669 (*sn)->rule.ptr->src_nodes++;
670 pf_status.scounters[SCNT_SRC_NODE_INSERT]++;
671 pf_status.src_nodes++;
672 } else {
673 if (rule->max_src_states &&
674 (*sn)->states >= rule->max_src_states)
675 return (-1);
676 }
677 return (0);
678}
679
680int
681pf_insert_state(struct pfi_kif *kif, struct pf_state *state)
682{
683 /* Thou MUST NOT insert multiple duplicate keys */
684 state->u.s.kif = kif;
685 if (RB_INSERT(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state)) {
686 if (pf_status.debug >= PF_DEBUG_MISC) {
687 printf("pf: state insert failed: tree_lan_ext");
688 printf(" lan: ");
689 pf_print_host(&state->lan.addr, state->lan.port,
690 state->af);
691 printf(" gwy: ");
692 pf_print_host(&state->gwy.addr, state->gwy.port,
693 state->af);
694 printf(" ext: ");
695 pf_print_host(&state->ext.addr, state->ext.port,
696 state->af);
697 if (state->sync_flags & PFSTATE_FROMSYNC)
698 printf(" (from sync)");
699 printf("\n");
700 }
701 return (-1);
702 }
703
704 if (RB_INSERT(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, state)) {
705 if (pf_status.debug >= PF_DEBUG_MISC) {
706 printf("pf: state insert failed: tree_ext_gwy");
707 printf(" lan: ");
708 pf_print_host(&state->lan.addr, state->lan.port,
709 state->af);
710 printf(" gwy: ");
711 pf_print_host(&state->gwy.addr, state->gwy.port,
712 state->af);
713 printf(" ext: ");
714 pf_print_host(&state->ext.addr, state->ext.port,
715 state->af);
716 if (state->sync_flags & PFSTATE_FROMSYNC)
717 printf(" (from sync)");
718 printf("\n");
719 }
720 RB_REMOVE(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state);
721 return (-1);
722 }
723
724 if (state->id == 0 && state->creatorid == 0) {
725 state->id = htobe64(pf_status.stateid++);
726 state->creatorid = pf_status.hostid;
727 }
728 if (RB_INSERT(pf_state_tree_id, &tree_id, state) != NULL) {
729 if (pf_status.debug >= PF_DEBUG_MISC) {
730#ifdef __FreeBSD__
731 printf("pf: state insert failed: "
732 "id: %016llx creatorid: %08x",
733 (long long)be64toh(state->id),
734 ntohl(state->creatorid));
735#else
736 printf("pf: state insert failed: "
737 "id: %016llx creatorid: %08x",
738 betoh64(state->id), ntohl(state->creatorid));
739#endif
740 if (state->sync_flags & PFSTATE_FROMSYNC)
741 printf(" (from sync)");
742 printf("\n");
743 }
744 RB_REMOVE(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state);
745 RB_REMOVE(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, state);
746 return (-1);
747 }
748 TAILQ_INSERT_HEAD(&state_updates, state, u.s.entry_updates);
749
750 pf_status.fcounters[FCNT_STATE_INSERT]++;
751 pf_status.states++;
752 pfi_attach_state(kif);
753#if NPFSYNC
754 pfsync_insert_state(state);
755#endif
756 return (0);
757}
758
759void
760pf_purge_timeout(void *arg)
761{
762#ifdef __FreeBSD__
763 struct callout *to = arg;
764#else
765 struct timeout *to = arg;
766#endif
767 int s;
768
769#ifdef __FreeBSD__
770 PF_LOCK();
771#endif
772 s = splsoftnet();
773 pf_purge_expired_states();
774 pf_purge_expired_fragments();
775 pf_purge_expired_src_nodes();
776 splx(s);
777#ifdef __FreeBSD__
778 PF_UNLOCK();
779#endif
780
781#ifdef __FreeBSD__
782 callout_reset(to, pf_default_rule.timeout[PFTM_INTERVAL] * hz,
783 pf_purge_timeout, to);
784#else
785 timeout_add(to, pf_default_rule.timeout[PFTM_INTERVAL] * hz);
786#endif
787}
788
789u_int32_t
790pf_state_expires(const struct pf_state *state)
791{
792 u_int32_t timeout;
793 u_int32_t start;
794 u_int32_t end;
795 u_int32_t states;
796
797 /* handle all PFTM_* > PFTM_MAX here */
798 if (state->timeout == PFTM_PURGE)
799#ifdef __FreeBSD__
800 return (time_second);
801#else
802 return (time.tv_sec);
803#endif
804 if (state->timeout == PFTM_UNTIL_PACKET)
805 return (0);
806#ifdef __FreeBSD__
807 KASSERT((state->timeout < PFTM_MAX),
808 ("pf_state_expires: timeout > PFTM_MAX"));
809#else
810 KASSERT(state->timeout < PFTM_MAX);
811#endif
812 timeout = state->rule.ptr->timeout[state->timeout];
813 if (!timeout)
814 timeout = pf_default_rule.timeout[state->timeout];
815 start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START];
816 if (start) {
817 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END];
818 states = state->rule.ptr->states;
819 } else {
820 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START];
821 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END];
822 states = pf_status.states;
823 }
824 if (end && states > start && start < end) {
825 if (states < end)
826 return (state->expire + timeout * (end - states) /
827 (end - start));
828 else
829#ifdef __FreeBSD__
830 return (time_second);
831#else
832 return (time.tv_sec);
833#endif
834 }
835 return (state->expire + timeout);
836}
837
838void
839pf_purge_expired_src_nodes(void)
840{
841 struct pf_src_node *cur, *next;
842
843 for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) {
844 next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur);
845
846#ifdef __FreeBSD__
847 if (cur->states <= 0 && cur->expire <= time_second) {
848#else
849 if (cur->states <= 0 && cur->expire <= time.tv_sec) {
850#endif
851 if (cur->rule.ptr != NULL) {
852 cur->rule.ptr->src_nodes--;
853 if (cur->rule.ptr->states <= 0 &&
854 cur->rule.ptr->max_src_nodes <= 0)
855 pf_rm_rule(NULL, cur->rule.ptr);
856 }
857 RB_REMOVE(pf_src_tree, &tree_src_tracking, cur);
858 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
859 pf_status.src_nodes--;
860 pool_put(&pf_src_tree_pl, cur);
861 }
862 }
863}
864
865void
866pf_src_tree_remove_state(struct pf_state *s)
867{
868 u_int32_t timeout;
869
870 if (s->src_node != NULL) {
871 if (--s->src_node->states <= 0) {
872 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
873 if (!timeout)
874 timeout =
875 pf_default_rule.timeout[PFTM_SRC_NODE];
876#ifdef __FreeBSD__
877 s->src_node->expire = time_second + timeout;
878#else
879 s->src_node->expire = time.tv_sec + timeout;
880#endif
881 }
882 }
883 if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) {
884 if (--s->nat_src_node->states <= 0) {
885 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
886 if (!timeout)
887 timeout =
888 pf_default_rule.timeout[PFTM_SRC_NODE];
889#ifdef __FreeBSD__
890 s->nat_src_node->expire = time_second + timeout;
891#else
892 s->nat_src_node->expire = time.tv_sec + timeout;
893#endif
894 }
895 }
896 s->src_node = s->nat_src_node = NULL;
897}
898
899void
900pf_purge_expired_states(void)
901{
902 struct pf_state *cur, *next;
903
904 for (cur = RB_MIN(pf_state_tree_id, &tree_id);
905 cur; cur = next) {
906 next = RB_NEXT(pf_state_tree_id, &tree_id, cur);
907
908#ifdef __FreeBSD__
909 if (pf_state_expires(cur) <= time_second) {
910#else
911 if (pf_state_expires(cur) <= time.tv_sec) {
912#endif
913 if (cur->src.state == PF_TCPS_PROXY_DST)
914 pf_send_tcp(cur->rule.ptr, cur->af,
915 &cur->ext.addr, &cur->lan.addr,
916 cur->ext.port, cur->lan.port,
917 cur->src.seqhi, cur->src.seqlo + 1, 0,
918 TH_RST|TH_ACK, 0, 0);
919 RB_REMOVE(pf_state_tree_ext_gwy,
920 &cur->u.s.kif->pfik_ext_gwy, cur);
921 RB_REMOVE(pf_state_tree_lan_ext,
922 &cur->u.s.kif->pfik_lan_ext, cur);
923 RB_REMOVE(pf_state_tree_id, &tree_id, cur);
924#if NPFSYNC
925 pfsync_delete_state(cur);
926#endif
927 pf_src_tree_remove_state(cur);
928 if (--cur->rule.ptr->states <= 0 &&
929 cur->rule.ptr->src_nodes <= 0)
930 pf_rm_rule(NULL, cur->rule.ptr);
931 if (cur->nat_rule.ptr != NULL)
932 if (--cur->nat_rule.ptr->states <= 0 &&
933 cur->nat_rule.ptr->src_nodes <= 0)
934 pf_rm_rule(NULL, cur->nat_rule.ptr);
935 if (cur->anchor.ptr != NULL)
936 if (--cur->anchor.ptr->states <= 0)
937 pf_rm_rule(NULL, cur->anchor.ptr);
938 pf_normalize_tcp_cleanup(cur);
939 pfi_detach_state(cur->u.s.kif);
940 TAILQ_REMOVE(&state_updates, cur, u.s.entry_updates);
941 pool_put(&pf_state_pl, cur);
942 pf_status.fcounters[FCNT_STATE_REMOVALS]++;
943 pf_status.states--;
944 }
945 }
946}
947
948int
949pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw)
950{
951 if (aw->type != PF_ADDR_TABLE)
952 return (0);
953 if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL)
954 return (1);
955 return (0);
956}
957
958void
959pf_tbladdr_remove(struct pf_addr_wrap *aw)
960{
961 if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL)
962 return;
963 pfr_detach_table(aw->p.tbl);
964 aw->p.tbl = NULL;
965}
966
967void
968pf_tbladdr_copyout(struct pf_addr_wrap *aw)
969{
970 struct pfr_ktable *kt = aw->p.tbl;
971
972 if (aw->type != PF_ADDR_TABLE || kt == NULL)
973 return;
974 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
975 kt = kt->pfrkt_root;
976 aw->p.tbl = NULL;
977 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
978 kt->pfrkt_cnt : -1;
979}
980
981void
982pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af)
983{
984 switch (af) {
985#ifdef INET
986 case AF_INET: {
987 u_int32_t a = ntohl(addr->addr32[0]);
988 printf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255,
989 (a>>8)&255, a&255);
990 if (p) {
991 p = ntohs(p);
992 printf(":%u", p);
993 }
994 break;
995 }
996#endif /* INET */
997#ifdef INET6
998 case AF_INET6: {
999 u_int16_t b;
1000 u_int8_t i, curstart = 255, curend = 0,
1001 maxstart = 0, maxend = 0;
1002 for (i = 0; i < 8; i++) {
1003 if (!addr->addr16[i]) {
1004 if (curstart == 255)
1005 curstart = i;
1006 else
1007 curend = i;
1008 } else {
1009 if (curstart) {
1010 if ((curend - curstart) >
1011 (maxend - maxstart)) {
1012 maxstart = curstart;
1013 maxend = curend;
1014 curstart = 255;
1015 }
1016 }
1017 }
1018 }
1019 for (i = 0; i < 8; i++) {
1020 if (i >= maxstart && i <= maxend) {
1021 if (maxend != 7) {
1022 if (i == maxstart)
1023 printf(":");
1024 } else {
1025 if (i == maxend)
1026 printf(":");
1027 }
1028 } else {
1029 b = ntohs(addr->addr16[i]);
1030 printf("%x", b);
1031 if (i < 7)
1032 printf(":");
1033 }
1034 }
1035 if (p) {
1036 p = ntohs(p);
1037 printf("[%u]", p);
1038 }
1039 break;
1040 }
1041#endif /* INET6 */
1042 }
1043}
1044
1045void
1046pf_print_state(struct pf_state *s)
1047{
1048 switch (s->proto) {
1049 case IPPROTO_TCP:
1050 printf("TCP ");
1051 break;
1052 case IPPROTO_UDP:
1053 printf("UDP ");
1054 break;
1055 case IPPROTO_ICMP:
1056 printf("ICMP ");
1057 break;
1058 case IPPROTO_ICMPV6:
1059 printf("ICMPV6 ");
1060 break;
1061 default:
1062 printf("%u ", s->proto);
1063 break;
1064 }
1065 pf_print_host(&s->lan.addr, s->lan.port, s->af);
1066 printf(" ");
1067 pf_print_host(&s->gwy.addr, s->gwy.port, s->af);
1068 printf(" ");
1069 pf_print_host(&s->ext.addr, s->ext.port, s->af);
1070 printf(" [lo=%u high=%u win=%u modulator=%u", s->src.seqlo,
1071 s->src.seqhi, s->src.max_win, s->src.seqdiff);
1072 if (s->src.wscale && s->dst.wscale)
1073 printf(" wscale=%u", s->src.wscale & PF_WSCALE_MASK);
1074 printf("]");
1075 printf(" [lo=%u high=%u win=%u modulator=%u", s->dst.seqlo,
1076 s->dst.seqhi, s->dst.max_win, s->dst.seqdiff);
1077 if (s->src.wscale && s->dst.wscale)
1078 printf(" wscale=%u", s->dst.wscale & PF_WSCALE_MASK);
1079 printf("]");
1080 printf(" %u:%u", s->src.state, s->dst.state);
1081}
1082
1083void
1084pf_print_flags(u_int8_t f)
1085{
1086 if (f)
1087 printf(" ");
1088 if (f & TH_FIN)
1089 printf("F");
1090 if (f & TH_SYN)
1091 printf("S");
1092 if (f & TH_RST)
1093 printf("R");
1094 if (f & TH_PUSH)
1095 printf("P");
1096 if (f & TH_ACK)
1097 printf("A");
1098 if (f & TH_URG)
1099 printf("U");
1100 if (f & TH_ECE)
1101 printf("E");
1102 if (f & TH_CWR)
1103 printf("W");
1104}
1105
1106#define PF_SET_SKIP_STEPS(i) \
1107 do { \
1108 while (head[i] != cur) { \
1109 head[i]->skip[i].ptr = cur; \
1110 head[i] = TAILQ_NEXT(head[i], entries); \
1111 } \
1112 } while (0)
1113
1114void
1115pf_calc_skip_steps(struct pf_rulequeue *rules)
1116{
1117 struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT];
1118 int i;
1119
1120 cur = TAILQ_FIRST(rules);
1121 prev = cur;
1122 for (i = 0; i < PF_SKIP_COUNT; ++i)
1123 head[i] = cur;
1124 while (cur != NULL) {
1125
1126 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
1127 PF_SET_SKIP_STEPS(PF_SKIP_IFP);
1128 if (cur->direction != prev->direction)
1129 PF_SET_SKIP_STEPS(PF_SKIP_DIR);
1130 if (cur->af != prev->af)
1131 PF_SET_SKIP_STEPS(PF_SKIP_AF);
1132 if (cur->proto != prev->proto)
1133 PF_SET_SKIP_STEPS(PF_SKIP_PROTO);
1134 if (cur->src.not != prev->src.not ||
1135 pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr))
1136 PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR);
1137 if (cur->src.port[0] != prev->src.port[0] ||
1138 cur->src.port[1] != prev->src.port[1] ||
1139 cur->src.port_op != prev->src.port_op)
1140 PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
1141 if (cur->dst.not != prev->dst.not ||
1142 pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr))
1143 PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR);
1144 if (cur->dst.port[0] != prev->dst.port[0] ||
1145 cur->dst.port[1] != prev->dst.port[1] ||
1146 cur->dst.port_op != prev->dst.port_op)
1147 PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
1148
1149 prev = cur;
1150 cur = TAILQ_NEXT(cur, entries);
1151 }
1152 for (i = 0; i < PF_SKIP_COUNT; ++i)
1153 PF_SET_SKIP_STEPS(i);
1154}
1155
1156int
1157pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2)
1158{
1159 if (aw1->type != aw2->type)
1160 return (1);
1161 switch (aw1->type) {
1162 case PF_ADDR_ADDRMASK:
1163 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0))
1164 return (1);
1165 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0))
1166 return (1);
1167 return (0);
1168 case PF_ADDR_DYNIFTL:
1169 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt);
1170 case PF_ADDR_NOROUTE:
1171 return (0);
1172 case PF_ADDR_TABLE:
1173 return (aw1->p.tbl != aw2->p.tbl);
1174 default:
1175 printf("invalid address type: %d\n", aw1->type);
1176 return (1);
1177 }
1178}
1179
1180void
1181pf_update_anchor_rules()
1182{
1183 struct pf_rule *rule;
1184 int i;
1185
1186 for (i = 0; i < PF_RULESET_MAX; ++i)
1187 TAILQ_FOREACH(rule, pf_main_ruleset.rules[i].active.ptr,
1188 entries)
1189 if (rule->anchorname[0])
1190 rule->anchor = pf_find_anchor(rule->anchorname);
1191 else
1192 rule->anchor = NULL;
1193}
1194
1195u_int16_t
1196pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp)
1197{
1198 u_int32_t l;
1199
1200 if (udp && !cksum)
1201 return (0x0000);
1202 l = cksum + old - new;
1203 l = (l >> 16) + (l & 65535);
1204 l = l & 65535;
1205 if (udp && !l)
1206 return (0xFFFF);
1207 return (l);
1208}
1209
1210void
1211pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc,
1212 struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af)
1213{
1214 struct pf_addr ao;
1215 u_int16_t po = *p;
1216
1217 PF_ACPY(&ao, a, af);
1218 PF_ACPY(a, an, af);
1219
1220 *p = pn;
1221
1222 switch (af) {
1223#ifdef INET
1224 case AF_INET:
1225 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1226 ao.addr16[0], an->addr16[0], 0),
1227 ao.addr16[1], an->addr16[1], 0);
1228 *p = pn;
1229 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1230 ao.addr16[0], an->addr16[0], u),
1231 ao.addr16[1], an->addr16[1], u),
1232 po, pn, u);
1233 break;
1234#endif /* INET */
1235#ifdef INET6
1236 case AF_INET6:
1237 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1238 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1239 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1240 ao.addr16[0], an->addr16[0], u),
1241 ao.addr16[1], an->addr16[1], u),
1242 ao.addr16[2], an->addr16[2], u),
1243 ao.addr16[3], an->addr16[3], u),
1244 ao.addr16[4], an->addr16[4], u),
1245 ao.addr16[5], an->addr16[5], u),
1246 ao.addr16[6], an->addr16[6], u),
1247 ao.addr16[7], an->addr16[7], u),
1248 po, pn, u);
1249 break;
1250#endif /* INET6 */
1251 }
1252}
1253
1254
1255/* Changes a u_int32_t. Uses a void * so there are no align restrictions */
1256void
1257pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u)
1258{
1259 u_int32_t ao;
1260
1261 memcpy(&ao, a, sizeof(ao));
1262 memcpy(a, &an, sizeof(u_int32_t));
1263 *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u),
1264 ao % 65536, an % 65536, u);
1265}
1266
1267#ifdef INET6
1268void
1269pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u)
1270{
1271 struct pf_addr ao;
1272
1273 PF_ACPY(&ao, a, AF_INET6);
1274 PF_ACPY(a, an, AF_INET6);
1275
1276 *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1277 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1278 pf_cksum_fixup(pf_cksum_fixup(*c,
1279 ao.addr16[0], an->addr16[0], u),
1280 ao.addr16[1], an->addr16[1], u),
1281 ao.addr16[2], an->addr16[2], u),
1282 ao.addr16[3], an->addr16[3], u),
1283 ao.addr16[4], an->addr16[4], u),
1284 ao.addr16[5], an->addr16[5], u),
1285 ao.addr16[6], an->addr16[6], u),
1286 ao.addr16[7], an->addr16[7], u);
1287}
1288#endif /* INET6 */
1289
1290void
1291pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa,
1292 struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c,
1293 u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af)
1294{
1295 struct pf_addr oia, ooa;
1296
1297 PF_ACPY(&oia, ia, af);
1298 PF_ACPY(&ooa, oa, af);
1299
1300 /* Change inner protocol port, fix inner protocol checksum. */
1301 if (ip != NULL) {
1302 u_int16_t oip = *ip;
1303 u_int32_t opc = 0; /* make the compiler happy */
1304
1305 if (pc != NULL)
1306 opc = *pc;
1307 *ip = np;
1308 if (pc != NULL)
1309 *pc = pf_cksum_fixup(*pc, oip, *ip, u);
1310 *ic = pf_cksum_fixup(*ic, oip, *ip, 0);
1311 if (pc != NULL)
1312 *ic = pf_cksum_fixup(*ic, opc, *pc, 0);
1313 }
1314 /* Change inner ip address, fix inner ip and icmp checksums. */
1315 PF_ACPY(ia, na, af);
1316 switch (af) {
1317#ifdef INET
1318 case AF_INET: {
1319 u_int32_t oh2c = *h2c;
1320
1321 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c,
1322 oia.addr16[0], ia->addr16[0], 0),
1323 oia.addr16[1], ia->addr16[1], 0);
1324 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1325 oia.addr16[0], ia->addr16[0], 0),
1326 oia.addr16[1], ia->addr16[1], 0);
1327 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0);
1328 break;
1329 }
1330#endif /* INET */
1331#ifdef INET6
1332 case AF_INET6:
1333 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1334 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1335 pf_cksum_fixup(pf_cksum_fixup(*ic,
1336 oia.addr16[0], ia->addr16[0], u),
1337 oia.addr16[1], ia->addr16[1], u),
1338 oia.addr16[2], ia->addr16[2], u),
1339 oia.addr16[3], ia->addr16[3], u),
1340 oia.addr16[4], ia->addr16[4], u),
1341 oia.addr16[5], ia->addr16[5], u),
1342 oia.addr16[6], ia->addr16[6], u),
1343 oia.addr16[7], ia->addr16[7], u);
1344 break;
1345#endif /* INET6 */
1346 }
1347 /* Change outer ip address, fix outer ip or icmpv6 checksum. */
1348 PF_ACPY(oa, na, af);
1349 switch (af) {
1350#ifdef INET
1351 case AF_INET:
1352 *hc = pf_cksum_fixup(pf_cksum_fixup(*hc,
1353 ooa.addr16[0], oa->addr16[0], 0),
1354 ooa.addr16[1], oa->addr16[1], 0);
1355 break;
1356#endif /* INET */
1357#ifdef INET6
1358 case AF_INET6:
1359 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1360 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1361 pf_cksum_fixup(pf_cksum_fixup(*ic,
1362 ooa.addr16[0], oa->addr16[0], u),
1363 ooa.addr16[1], oa->addr16[1], u),
1364 ooa.addr16[2], oa->addr16[2], u),
1365 ooa.addr16[3], oa->addr16[3], u),
1366 ooa.addr16[4], oa->addr16[4], u),
1367 ooa.addr16[5], oa->addr16[5], u),
1368 ooa.addr16[6], oa->addr16[6], u),
1369 ooa.addr16[7], oa->addr16[7], u);
1370 break;
1371#endif /* INET6 */
1372 }
1373}
1374
1375void
1376pf_send_tcp(const struct pf_rule *r, sa_family_t af,
1377 const struct pf_addr *saddr, const struct pf_addr *daddr,
1378 u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack,
1379 u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl)
1380{
1381 struct mbuf *m;
1382 struct m_tag *mtag;
1383 int len = 0, tlen; /* make the compiler happy */
1384#ifdef INET
1385 struct ip *h = NULL; /* make the compiler happy */
1386#endif /* INET */
1387#ifdef INET6
1388 struct ip6_hdr *h6 = NULL; /* make the compiler happy */
1389#endif /* INET6 */
1390 struct tcphdr *th = NULL; /* make the compiler happy */
1391#ifdef __FreeBSD__
1392 struct ip *ip;
1393#endif
1394 char *opt;
1395
1396 /* maximum segment size tcp option */
1397 tlen = sizeof(struct tcphdr);
1398 if (mss)
1399 tlen += 4;
1400
1401 switch (af) {
1402#ifdef INET
1403 case AF_INET:
1404 len = sizeof(struct ip) + tlen;
1405 break;
1406#endif /* INET */
1407#ifdef INET6
1408 case AF_INET6:
1409 len = sizeof(struct ip6_hdr) + tlen;
1410 break;
1411#endif /* INET6 */
1412 }
1413
1414 /* create outgoing mbuf */
1415 mtag = m_tag_get(PACKET_TAG_PF_GENERATED, 0, M_NOWAIT);
1416 if (mtag == NULL)
1417 return;
1418 m = m_gethdr(M_DONTWAIT, MT_HEADER);
1419 if (m == NULL) {
1420 m_tag_free(mtag);
1421 return;
1422 }
1423 m_tag_prepend(m, mtag);
1424#ifdef ALTQ
1425 if (r != NULL && r->qid) {
1426 struct altq_tag *atag;
1427
1428 mtag = m_tag_get(PACKET_TAG_PF_QID, sizeof(*atag), M_NOWAIT);
1429 if (mtag != NULL) {
1430 atag = (struct altq_tag *)(mtag + 1);
1431 atag->qid = r->qid;
1432 /* add hints for ecn */
1433 atag->af = af;
1434 atag->hdr = mtod(m, struct ip *);
1435 m_tag_prepend(m, mtag);
1436 }
1437 }
1438#endif
1439 m->m_data += max_linkhdr;
1440 m->m_pkthdr.len = m->m_len = len;
1441 m->m_pkthdr.rcvif = NULL;
1442 bzero(m->m_data, len);
1443 switch (af) {
1444#ifdef INET
1445 case AF_INET:
1446 h = mtod(m, struct ip *);
1447
1448 /* IP header fields included in the TCP checksum */
1449 h->ip_p = IPPROTO_TCP;
1450 h->ip_len = htons(tlen);
1451 h->ip_src.s_addr = saddr->v4.s_addr;
1452 h->ip_dst.s_addr = daddr->v4.s_addr;
1453
1454 th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip));
1455 break;
1456#endif /* INET */
1457#ifdef INET6
1458 case AF_INET6:
1459 h6 = mtod(m, struct ip6_hdr *);
1460
1461 /* IP header fields included in the TCP checksum */
1462 h6->ip6_nxt = IPPROTO_TCP;
1463 h6->ip6_plen = htons(tlen);
1464 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr));
1465 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr));
1466
1467 th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr));
1468 break;
1469#endif /* INET6 */
1470 }
1471
1472 /* TCP header */
1473 th->th_sport = sport;
1474 th->th_dport = dport;
1475 th->th_seq = htonl(seq);
1476 th->th_ack = htonl(ack);
1477 th->th_off = tlen >> 2;
1478 th->th_flags = flags;
1479 th->th_win = htons(win);
1480
1481 if (mss) {
1482 opt = (char *)(th + 1);
1483 opt[0] = TCPOPT_MAXSEG;
1484 opt[1] = 4;
1485 HTONS(mss);
1486 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2);
1487 }
1488
1489 switch (af) {
1490#ifdef INET
1491 case AF_INET:
1492 /* TCP checksum */
1493 th->th_sum = in_cksum(m, len);
1494
1495 /* Finish the IP header */
1496 h->ip_v = 4;
1497 h->ip_hl = sizeof(*h) >> 2;
1498 h->ip_tos = IPTOS_LOWDELAY;
1499#ifdef __FreeBSD__
1500 h->ip_off = path_mtu_discovery ? IP_DF : 0;
1501 h->ip_len = len;
1502#else
1503 h->ip_off = htons(ip_mtudisc ? IP_DF : 0);
1504 h->ip_len = htons(len);
1505#endif
1506 h->ip_ttl = ttl ? ttl : ip_defttl;
1507 h->ip_sum = 0;
1508#ifdef __FreeBSD__
1509 ip = mtod(m, struct ip *);
1510 PF_UNLOCK();
1511 ip_output(m, (void *)NULL, (void *)NULL, 0, (void *)NULL,
1512 (void *)NULL);
1513 PF_LOCK();
1514#else /* ! __FreeBSD__ */
1515 ip_output(m, (void *)NULL, (void *)NULL, 0, (void *)NULL,
1516 (void *)NULL);
1517#endif
1518 break;
1519#endif /* INET */
1520#ifdef INET6
1521 case AF_INET6:
1522 /* TCP checksum */
1523 th->th_sum = in6_cksum(m, IPPROTO_TCP,
1524 sizeof(struct ip6_hdr), tlen);
1525
1526 h6->ip6_vfc |= IPV6_VERSION;
1527 h6->ip6_hlim = IPV6_DEFHLIM;
1528
1529#ifdef __FreeBSD__
1530 PF_UNLOCK();
1531 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
1532 PF_LOCK();
1533#else
1534 ip6_output(m, NULL, NULL, 0, NULL, NULL);
1535#endif
1536 break;
1537#endif /* INET6 */
1538 }
1539}
1540
1541void
1542pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af,
1543 struct pf_rule *r)
1544{
1545 struct m_tag *mtag;
1546 struct mbuf *m0;
1547#ifdef __FreeBSD__
1548 struct ip *ip;
1549#endif
1550
1551 mtag = m_tag_get(PACKET_TAG_PF_GENERATED, 0, M_NOWAIT);
1552 if (mtag == NULL)
1553 return;
1554#ifdef __FreeBSD__
1555 m0 = m_copypacket(m, M_DONTWAIT);
1556#else
1557 m0 = m_copy(m, 0, M_COPYALL);
1558#endif
1559 if (m0 == NULL) {
1560 m_tag_free(mtag);
1561 return;
1562 }
1563 m_tag_prepend(m0, mtag);
1564
1565#ifdef ALTQ
1566 if (r->qid) {
1567 struct altq_tag *atag;
1568
1569 mtag = m_tag_get(PACKET_TAG_PF_QID, sizeof(*atag), M_NOWAIT);
1570 if (mtag != NULL) {
1571 atag = (struct altq_tag *)(mtag + 1);
1572 atag->qid = r->qid;
1573 /* add hints for ecn */
1574 atag->af = af;
1575 atag->hdr = mtod(m0, struct ip *);
1576 m_tag_prepend(m0, mtag);
1577 }
1578 }
1579#endif
1580
1581 switch (af) {
1582#ifdef INET
1583 case AF_INET:
1584#ifdef __FreeBSD__
1585 /* icmp_error() expects host byte ordering */
1586 ip = mtod(m0, struct ip *);
1587 NTOHS(ip->ip_len);
1588 NTOHS(ip->ip_off);
1589 PF_UNLOCK();
1590#endif
1591 icmp_error(m0, type, code, 0, (void *)NULL);
1592#ifdef __FreeBSD__
1593 PF_LOCK();
1594#endif
1595 break;
1596#endif /* INET */
1597#ifdef INET6
1598 case AF_INET6:
1599#ifdef __FreeBSD__
1600 PF_UNLOCK();
1601#endif
1602 icmp6_error(m0, type, code, 0);
1603#ifdef __FreeBSD__
1604 PF_LOCK();
1605#endif
1606 break;
1607#endif /* INET6 */
1608 }
1609}
1610
1611/*
1612 * Return 1 if the addresses a and b match (with mask m), otherwise return 0.
1613 * If n is 0, they match if they are equal. If n is != 0, they match if they
1614 * are different.
1615 */
1616int
1617pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m,
1618 struct pf_addr *b, sa_family_t af)
1619{
1620 int match = 0;
1621
1622 switch (af) {
1623#ifdef INET
1624 case AF_INET:
1625 if ((a->addr32[0] & m->addr32[0]) ==
1626 (b->addr32[0] & m->addr32[0]))
1627 match++;
1628 break;
1629#endif /* INET */
1630#ifdef INET6
1631 case AF_INET6:
1632 if (((a->addr32[0] & m->addr32[0]) ==
1633 (b->addr32[0] & m->addr32[0])) &&
1634 ((a->addr32[1] & m->addr32[1]) ==
1635 (b->addr32[1] & m->addr32[1])) &&
1636 ((a->addr32[2] & m->addr32[2]) ==
1637 (b->addr32[2] & m->addr32[2])) &&
1638 ((a->addr32[3] & m->addr32[3]) ==
1639 (b->addr32[3] & m->addr32[3])))
1640 match++;
1641 break;
1642#endif /* INET6 */
1643 }
1644 if (match) {
1645 if (n)
1646 return (0);
1647 else
1648 return (1);
1649 } else {
1650 if (n)
1651 return (1);
1652 else
1653 return (0);
1654 }
1655}
1656
1657int
1658pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p)
1659{
1660 switch (op) {
1661 case PF_OP_IRG:
1662 return ((p > a1) && (p < a2));
1663 case PF_OP_XRG:
1664 return ((p < a1) || (p > a2));
1665 case PF_OP_RRG:
1666 return ((p >= a1) && (p <= a2));
1667 case PF_OP_EQ:
1668 return (p == a1);
1669 case PF_OP_NE:
1670 return (p != a1);
1671 case PF_OP_LT:
1672 return (p < a1);
1673 case PF_OP_LE:
1674 return (p <= a1);
1675 case PF_OP_GT:
1676 return (p > a1);
1677 case PF_OP_GE:
1678 return (p >= a1);
1679 }
1680 return (0); /* never reached */
1681}
1682
1683int
1684pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p)
1685{
1686 NTOHS(a1);
1687 NTOHS(a2);
1688 NTOHS(p);
1689 return (pf_match(op, a1, a2, p));
1690}
1691
1692int
1693pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u)
1694{
1695 if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
1696 return (0);
1697 return (pf_match(op, a1, a2, u));
1698}
1699
1700int
1701pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g)
1702{
1703 if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
1704 return (0);
1705 return (pf_match(op, a1, a2, g));
1706}
1707
1708struct pf_tag *
1709pf_get_tag(struct mbuf *m)
1710{
1711 struct m_tag *mtag;
1712
1713 if ((mtag = m_tag_find(m, PACKET_TAG_PF_TAG, NULL)) != NULL)
1714 return ((struct pf_tag *)(mtag + 1));
1715 else
1716 return (NULL);
1717}
1718
1719int
1720pf_match_tag(struct mbuf *m, struct pf_rule *r, struct pf_rule *nat_rule,
1721 struct pf_tag *pftag, int *tag)
1722{
1723 if (*tag == -1) { /* find mbuf tag */
1724 pftag = pf_get_tag(m);
1725 if (pftag != NULL)
1726 *tag = pftag->tag;
1727 else
1728 *tag = 0;
1729 if (nat_rule != NULL && nat_rule->tag)
1730 *tag = nat_rule->tag;
1731 }
1732
1733 return ((!r->match_tag_not && r->match_tag == *tag) ||
1734 (r->match_tag_not && r->match_tag != *tag));
1735}
1736
1737int
1738pf_tag_packet(struct mbuf *m, struct pf_tag *pftag, int tag)
1739{
1740 struct m_tag *mtag;
1741
1742 if (tag <= 0)
1743 return (0);
1744
1745 if (pftag == NULL) {
1746 mtag = m_tag_get(PACKET_TAG_PF_TAG, sizeof(*pftag), M_NOWAIT);
1747 if (mtag == NULL)
1748 return (1);
1749 ((struct pf_tag *)(mtag + 1))->tag = tag;
1750 m_tag_prepend(m, mtag);
1751 } else
1752 pftag->tag = tag;
1753
1754 return (0);
1755}
1756
1757#define PF_STEP_INTO_ANCHOR(r, a, s, n) \
1758 do { \
1759 if ((r) == NULL || (r)->anchor == NULL || \
1760 (s) != NULL || (a) != NULL) \
1761 panic("PF_STEP_INTO_ANCHOR"); \
1762 (a) = (r); \
1763 (s) = TAILQ_FIRST(&(r)->anchor->rulesets); \
1764 (r) = NULL; \
1765 while ((s) != NULL && ((r) = \
1766 TAILQ_FIRST((s)->rules[n].active.ptr)) == NULL) \
1767 (s) = TAILQ_NEXT((s), entries); \
1768 if ((r) == NULL) { \
1769 (r) = TAILQ_NEXT((a), entries); \
1770 (a) = NULL; \
1771 } \
1772 } while (0)
1773
1774#define PF_STEP_OUT_OF_ANCHOR(r, a, s, n) \
1775 do { \
1776 if ((r) != NULL || (a) == NULL || (s) == NULL) \
1777 panic("PF_STEP_OUT_OF_ANCHOR"); \
1778 (s) = TAILQ_NEXT((s), entries); \
1779 while ((s) != NULL && ((r) = \
1780 TAILQ_FIRST((s)->rules[n].active.ptr)) == NULL) \
1781 (s) = TAILQ_NEXT((s), entries); \
1782 if ((r) == NULL) { \
1783 (r) = TAILQ_NEXT((a), entries); \
1784 (a) = NULL; \
1785 } \
1786 } while (0)
1787
1788#ifdef INET6
1789void
1790pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr,
1791 struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af)
1792{
1793 switch (af) {
1794#ifdef INET
1795 case AF_INET:
1796 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
1797 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
1798 break;
1799#endif /* INET */
1800 case AF_INET6:
1801 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
1802 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
1803 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) |
1804 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]);
1805 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) |
1806 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]);
1807 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) |
1808 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]);
1809 break;
1810 }
1811}
1812
1813void
1814pf_addr_inc(struct pf_addr *addr, sa_family_t af)
1815{
1816 switch (af) {
1817#ifdef INET
1818 case AF_INET:
1819 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1);
1820 break;
1821#endif /* INET */
1822 case AF_INET6:
1823 if (addr->addr32[3] == 0xffffffff) {
1824 addr->addr32[3] = 0;
1825 if (addr->addr32[2] == 0xffffffff) {
1826 addr->addr32[2] = 0;
1827 if (addr->addr32[1] == 0xffffffff) {
1828 addr->addr32[1] = 0;
1829 addr->addr32[0] =
1830 htonl(ntohl(addr->addr32[0]) + 1);
1831 } else
1832 addr->addr32[1] =
1833 htonl(ntohl(addr->addr32[1]) + 1);
1834 } else
1835 addr->addr32[2] =
1836 htonl(ntohl(addr->addr32[2]) + 1);
1837 } else
1838 addr->addr32[3] =
1839 htonl(ntohl(addr->addr32[3]) + 1);
1840 break;
1841 }
1842}
1843#endif /* INET6 */
1844
1845#define mix(a,b,c) \
1846 do { \
1847 a -= b; a -= c; a ^= (c >> 13); \
1848 b -= c; b -= a; b ^= (a << 8); \
1849 c -= a; c -= b; c ^= (b >> 13); \
1850 a -= b; a -= c; a ^= (c >> 12); \
1851 b -= c; b -= a; b ^= (a << 16); \
1852 c -= a; c -= b; c ^= (b >> 5); \
1853 a -= b; a -= c; a ^= (c >> 3); \
1854 b -= c; b -= a; b ^= (a << 10); \
1855 c -= a; c -= b; c ^= (b >> 15); \
1856 } while (0)
1857
1858/*
1859 * hash function based on bridge_hash in if_bridge.c
1860 */
1861void
1862pf_hash(struct pf_addr *inaddr, struct pf_addr *hash,
1863 struct pf_poolhashkey *key, sa_family_t af)
1864{
1865 u_int32_t a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0];
1866
1867 switch (af) {
1868#ifdef INET
1869 case AF_INET:
1870 a += inaddr->addr32[0];
1871 b += key->key32[1];
1872 mix(a, b, c);
1873 hash->addr32[0] = c + key->key32[2];
1874 break;
1875#endif /* INET */
1876#ifdef INET6
1877 case AF_INET6:
1878 a += inaddr->addr32[0];
1879 b += inaddr->addr32[2];
1880 mix(a, b, c);
1881 hash->addr32[0] = c;
1882 a += inaddr->addr32[1];
1883 b += inaddr->addr32[3];
1884 c += key->key32[1];
1885 mix(a, b, c);
1886 hash->addr32[1] = c;
1887 a += inaddr->addr32[2];
1888 b += inaddr->addr32[1];
1889 c += key->key32[2];
1890 mix(a, b, c);
1891 hash->addr32[2] = c;
1892 a += inaddr->addr32[3];
1893 b += inaddr->addr32[0];
1894 c += key->key32[3];
1895 mix(a, b, c);
1896 hash->addr32[3] = c;
1897 break;
1898#endif /* INET6 */
1899 }
1900}
1901
1902int
1903pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr,
1904 struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn)
1905{
1906 unsigned char hash[16];
1907 struct pf_pool *rpool = &r->rpool;
1908 struct pf_addr *raddr = &rpool->cur->addr.v.a.addr;
1909 struct pf_addr *rmask = &rpool->cur->addr.v.a.mask;
1910 struct pf_pooladdr *acur = rpool->cur;
1911 struct pf_src_node k;
1912
1913 if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR &&
1914 (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
1915 k.af = af;
1916 PF_ACPY(&k.addr, saddr, af);
1917 if (r->rule_flag & PFRULE_RULESRCTRACK ||
1918 r->rpool.opts & PF_POOL_STICKYADDR)
1919 k.rule.ptr = r;
1920 else
1921 k.rule.ptr = NULL;
1922 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
1923 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
1924 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) {
1925 PF_ACPY(naddr, &(*sn)->raddr, af);
1926 if (pf_status.debug >= PF_DEBUG_MISC) {
1927 printf("pf_map_addr: src tracking maps ");
1928 pf_print_host(&k.addr, 0, af);
1929 printf(" to ");
1930 pf_print_host(naddr, 0, af);
1931 printf("\n");
1932 }
1933 return (0);
1934 }
1935 }
1936
1937 if (rpool->cur->addr.type == PF_ADDR_NOROUTE)
1938 return (1);
1939 if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
1940 if (af == AF_INET) {
1941 if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 &&
1942 (rpool->opts & PF_POOL_TYPEMASK) !=
1943 PF_POOL_ROUNDROBIN)
1944 return (1);
1945 raddr = &rpool->cur->addr.p.dyn->pfid_addr4;
1946 rmask = &rpool->cur->addr.p.dyn->pfid_mask4;
1947 } else {
1948 if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 &&
1949 (rpool->opts & PF_POOL_TYPEMASK) !=
1950 PF_POOL_ROUNDROBIN)
1951 return (1);
1952 raddr = &rpool->cur->addr.p.dyn->pfid_addr6;
1953 rmask = &rpool->cur->addr.p.dyn->pfid_mask6;
1954 }
1955 } else if (rpool->cur->addr.type == PF_ADDR_TABLE) {
1956 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN)
1957 return (1); /* unsupported */
1958 } else {
1959 raddr = &rpool->cur->addr.v.a.addr;
1960 rmask = &rpool->cur->addr.v.a.mask;
1961 }
1962
1963 switch (rpool->opts & PF_POOL_TYPEMASK) {
1964 case PF_POOL_NONE:
1965 PF_ACPY(naddr, raddr, af);
1966 break;
1967 case PF_POOL_BITMASK:
1968 PF_POOLMASK(naddr, raddr, rmask, saddr, af);
1969 break;
1970 case PF_POOL_RANDOM:
1971 if (init_addr != NULL && PF_AZERO(init_addr, af)) {
1972 switch (af) {
1973#ifdef INET
1974 case AF_INET:
1975 rpool->counter.addr32[0] = arc4random();
1976 break;
1977#endif /* INET */
1978#ifdef INET6
1979 case AF_INET6:
1980 if (rmask->addr32[3] != 0xffffffff)
1981 rpool->counter.addr32[3] = arc4random();
1982 else
1983 break;
1984 if (rmask->addr32[2] != 0xffffffff)
1985 rpool->counter.addr32[2] = arc4random();
1986 else
1987 break;
1988 if (rmask->addr32[1] != 0xffffffff)
1989 rpool->counter.addr32[1] = arc4random();
1990 else
1991 break;
1992 if (rmask->addr32[0] != 0xffffffff)
1993 rpool->counter.addr32[0] = arc4random();
1994 break;
1995#endif /* INET6 */
1996 }
1997 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
1998 PF_ACPY(init_addr, naddr, af);
1999
2000 } else {
2001 PF_AINC(&rpool->counter, af);
2002 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2003 }
2004 break;
2005 case PF_POOL_SRCHASH:
2006 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af);
2007 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af);
2008 break;
2009 case PF_POOL_ROUNDROBIN:
2010 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2011 if (!pfr_pool_get(rpool->cur->addr.p.tbl,
2012 &rpool->tblidx, &rpool->counter,
2013 &raddr, &rmask, af))
2014 goto get_addr;
2015 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2016 if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2017 &rpool->tblidx, &rpool->counter,
2018 &raddr, &rmask, af))
2019 goto get_addr;
2020 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af))
2021 goto get_addr;
2022
2023 try_next:
2024 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL)
2025 rpool->cur = TAILQ_FIRST(&rpool->list);
2026 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2027 rpool->tblidx = -1;
2028 if (pfr_pool_get(rpool->cur->addr.p.tbl,
2029 &rpool->tblidx, &rpool->counter,
2030 &raddr, &rmask, af)) {
2031 /* table contains no address of type 'af' */
2032 if (rpool->cur != acur)
2033 goto try_next;
2034 return (1);
2035 }
2036 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2037 rpool->tblidx = -1;
2038 if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2039 &rpool->tblidx, &rpool->counter,
2040 &raddr, &rmask, af)) {
2041 /* table contains no address of type 'af' */
2042 if (rpool->cur != acur)
2043 goto try_next;
2044 return (1);
2045 }
2046 } else {
2047 raddr = &rpool->cur->addr.v.a.addr;
2048 rmask = &rpool->cur->addr.v.a.mask;
2049 PF_ACPY(&rpool->counter, raddr, af);
2050 }
2051
2052 get_addr:
2053 PF_ACPY(naddr, &rpool->counter, af);
2054 PF_AINC(&rpool->counter, af);
2055 break;
2056 }
2057 if (*sn != NULL)
2058 PF_ACPY(&(*sn)->raddr, naddr, af);
2059
2060 if (pf_status.debug >= PF_DEBUG_MISC &&
2061 (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2062 printf("pf_map_addr: selected address ");
2063 pf_print_host(naddr, 0, af);
2064 printf("\n");
2065 }
2066
2067 return (0);
2068}
2069
2070int
2071pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r,
2072 struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport,
2073 struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high,
2074 struct pf_src_node **sn)
2075{
2076 struct pf_state key;
2077 struct pf_addr init_addr;
2078 u_int16_t cut;
2079
2080 bzero(&init_addr, sizeof(init_addr));
2081 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2082 return (1);
2083
2084 do {
2085 key.af = af;
2086 key.proto = proto;
2087 PF_ACPY(&key.ext.addr, daddr, key.af);
2088 PF_ACPY(&key.gwy.addr, naddr, key.af);
2089 key.ext.port = dport;
2090
2091 /*
2092 * port search; start random, step;
2093 * similar 2 portloop in in_pcbbind
2094 */
2095 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP)) {
2096 key.gwy.port = 0;
2097 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
2098 return (0);
2099 } else if (low == 0 && high == 0) {
2100 key.gwy.port = *nport;
2101 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
2102 return (0);
2103 } else if (low == high) {
2104 key.gwy.port = htons(low);
2105 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) {
2106 *nport = htons(low);
2107 return (0);
2108 }
2109 } else {
2110 u_int16_t tmp;
2111
2112 if (low > high) {
2113 tmp = low;
2114 low = high;
2115 high = tmp;
2116 }
2117 /* low < high */
2118 cut = arc4random() % (1 + high - low) + low;
2119 /* low <= cut <= high */
2120 for (tmp = cut; tmp <= high; ++(tmp)) {
2121 key.gwy.port = htons(tmp);
2122 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
2123 NULL) {
2124 *nport = htons(tmp);
2125 return (0);
2126 }
2127 }
2128 for (tmp = cut - 1; tmp >= low; --(tmp)) {
2129 key.gwy.port = htons(tmp);
2130 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
2131 NULL) {
2132 *nport = htons(tmp);
2133 return (0);
2134 }
2135 }
2136 }
2137
2138 switch (r->rpool.opts & PF_POOL_TYPEMASK) {
2139 case PF_POOL_RANDOM:
2140 case PF_POOL_ROUNDROBIN:
2141 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2142 return (1);
2143 break;
2144 case PF_POOL_NONE:
2145 case PF_POOL_SRCHASH:
2146 case PF_POOL_BITMASK:
2147 default:
2148 return (1);
2149 }
2150 } while (! PF_AEQ(&init_addr, naddr, af) );
2151
2152 return (1); /* none available */
2153}
2154
2155struct pf_rule *
2156pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off,
2157 int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport,
2158 struct pf_addr *daddr, u_int16_t dport, int rs_num)
2159{
2160 struct pf_rule *r, *rm = NULL, *anchorrule = NULL;
2161 struct pf_ruleset *ruleset = NULL;
2162
2163 r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr);
2164 while (r && rm == NULL) {
2165 struct pf_rule_addr *src = NULL, *dst = NULL;
2166 struct pf_addr_wrap *xdst = NULL;
2167
2168 if (r->action == PF_BINAT && direction == PF_IN) {
2169 src = &r->dst;
2170 if (r->rpool.cur != NULL)
2171 xdst = &r->rpool.cur->addr;
2172 } else {
2173 src = &r->src;
2174 dst = &r->dst;
2175 }
2176
2177 r->evaluations++;
2178 if (r->kif != NULL &&
2179 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
2180 r = r->skip[PF_SKIP_IFP].ptr;
2181 else if (r->direction && r->direction != direction)
2182 r = r->skip[PF_SKIP_DIR].ptr;
2183 else if (r->af && r->af != pd->af)
2184 r = r->skip[PF_SKIP_AF].ptr;
2185 else if (r->proto && r->proto != pd->proto)
2186 r = r->skip[PF_SKIP_PROTO].ptr;
2187 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af, src->not))
2188 r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR :
2189 PF_SKIP_DST_ADDR].ptr;
2190 else if (src->port_op && !pf_match_port(src->port_op,
2191 src->port[0], src->port[1], sport))
2192 r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT :
2193 PF_SKIP_DST_PORT].ptr;
2194 else if (dst != NULL &&
2195 PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->not))
2196 r = r->skip[PF_SKIP_DST_ADDR].ptr;
2197 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af, 0))
2198 r = TAILQ_NEXT(r, entries);
2199 else if (dst != NULL && dst->port_op &&
2200 !pf_match_port(dst->port_op, dst->port[0],
2201 dst->port[1], dport))
2202 r = r->skip[PF_SKIP_DST_PORT].ptr;
2203 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto !=
2204 IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m,
2205 off, pd->hdr.tcp), r->os_fingerprint)))
2206 r = TAILQ_NEXT(r, entries);
2207 else if (r->anchorname[0] && r->anchor == NULL)
2208 r = TAILQ_NEXT(r, entries);
2209 else if (r->anchor == NULL)
2210 rm = r;
2211 else
2212 PF_STEP_INTO_ANCHOR(r, anchorrule, ruleset, rs_num);
2213 if (r == NULL && anchorrule != NULL)
2214 PF_STEP_OUT_OF_ANCHOR(r, anchorrule, ruleset,
2215 rs_num);
2216 }
2217 if (rm != NULL && (rm->action == PF_NONAT ||
2218 rm->action == PF_NORDR || rm->action == PF_NOBINAT))
2219 return (NULL);
2220 return (rm);
2221}
2222
2223struct pf_rule *
2224pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction,
2225 struct pfi_kif *kif, struct pf_src_node **sn,
2226 struct pf_addr *saddr, u_int16_t sport,
2227 struct pf_addr *daddr, u_int16_t dport,
2228 struct pf_addr *naddr, u_int16_t *nport)
2229{
2230 struct pf_rule *r = NULL;
2231
2232 if (direction == PF_OUT) {
2233 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2234 sport, daddr, dport, PF_RULESET_BINAT);
2235 if (r == NULL)
2236 r = pf_match_translation(pd, m, off, direction, kif,
2237 saddr, sport, daddr, dport, PF_RULESET_NAT);
2238 } else {
2239 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2240 sport, daddr, dport, PF_RULESET_RDR);
2241 if (r == NULL)
2242 r = pf_match_translation(pd, m, off, direction, kif,
2243 saddr, sport, daddr, dport, PF_RULESET_BINAT);
2244 }
2245
2246 if (r != NULL) {
2247 switch (r->action) {
2248 case PF_NONAT:
2249 case PF_NOBINAT:
2250 case PF_NORDR:
2251 return (NULL);
2252 case PF_NAT:
2253 if (pf_get_sport(pd->af, pd->proto, r, saddr,
2254 daddr, dport, naddr, nport, r->rpool.proxy_port[0],
2255 r->rpool.proxy_port[1], sn)) {
2256 DPFPRINTF(PF_DEBUG_MISC,
2257 ("pf: NAT proxy port allocation "
2258 "(%u-%u) failed\n",
2259 r->rpool.proxy_port[0],
2260 r->rpool.proxy_port[1]));
2261 return (NULL);
2262 }
2263 break;
2264 case PF_BINAT:
2265 switch (direction) {
2266 case PF_OUT:
2267 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){
2268 if (pd->af == AF_INET) {
2269 if (r->rpool.cur->addr.p.dyn->
2270 pfid_acnt4 < 1)
2271 return (NULL);
2272 PF_POOLMASK(naddr,
2273 &r->rpool.cur->addr.p.dyn->
2274 pfid_addr4,
2275 &r->rpool.cur->addr.p.dyn->
2276 pfid_mask4,
2277 saddr, AF_INET);
2278 } else {
2279 if (r->rpool.cur->addr.p.dyn->
2280 pfid_acnt6 < 1)
2281 return (NULL);
2282 PF_POOLMASK(naddr,
2283 &r->rpool.cur->addr.p.dyn->
2284 pfid_addr6,
2285 &r->rpool.cur->addr.p.dyn->
2286 pfid_mask6,
2287 saddr, AF_INET6);
2288 }
2289 } else
2290 PF_POOLMASK(naddr,
2291 &r->rpool.cur->addr.v.a.addr,
2292 &r->rpool.cur->addr.v.a.mask,
2293 saddr, pd->af);
2294 break;
2295 case PF_IN:
2296 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){
2297 if (pd->af == AF_INET) {
2298 if (r->src.addr.p.dyn->
2299 pfid_acnt4 < 1)
2300 return (NULL);
2301 PF_POOLMASK(naddr,
2302 &r->src.addr.p.dyn->
2303 pfid_addr4,
2304 &r->src.addr.p.dyn->
2305 pfid_mask4,
2306 daddr, AF_INET);
2307 } else {
2308 if (r->src.addr.p.dyn->
2309 pfid_acnt6 < 1)
2310 return (NULL);
2311 PF_POOLMASK(naddr,
2312 &r->src.addr.p.dyn->
2313 pfid_addr6,
2314 &r->src.addr.p.dyn->
2315 pfid_mask6,
2316 daddr, AF_INET6);
2317 }
2318 } else
2319 PF_POOLMASK(naddr,
2320 &r->src.addr.v.a.addr,
2321 &r->src.addr.v.a.mask, daddr,
2322 pd->af);
2323 break;
2324 }
2325 break;
2326 case PF_RDR: {
2327 if (pf_map_addr(r->af, r, saddr, naddr, NULL, sn))
2328 return (NULL);
2329
2330 if (r->rpool.proxy_port[1]) {
2331 u_int32_t tmp_nport;
2332
2333 tmp_nport = ((ntohs(dport) -
2334 ntohs(r->dst.port[0])) %
2335 (r->rpool.proxy_port[1] -
2336 r->rpool.proxy_port[0] + 1)) +
2337 r->rpool.proxy_port[0];
2338
2339 /* wrap around if necessary */
2340 if (tmp_nport > 65535)
2341 tmp_nport -= 65535;
2342 *nport = htons((u_int16_t)tmp_nport);
2343 } else if (r->rpool.proxy_port[0])
2344 *nport = htons(r->rpool.proxy_port[0]);
2345 break;
2346 }
2347 default:
2348 return (NULL);
2349 }
2350 }
2351
2352 return (r);
2353}
2354
2355int
2356pf_socket_lookup(uid_t *uid, gid_t *gid, int direction, struct pf_pdesc *pd)
2357{
2358 struct pf_addr *saddr, *daddr;
2359 u_int16_t sport, dport;
2360#ifdef __FreeBSD__
2361 struct inpcbinfo *pi;
2362#else
2363 struct inpcbtable *tb;
2364#endif
2365 struct inpcb *inp;
2366
2367 *uid = UID_MAX;
2368 *gid = GID_MAX;
2369 switch (pd->proto) {
2370 case IPPROTO_TCP:
2371 sport = pd->hdr.tcp->th_sport;
2372 dport = pd->hdr.tcp->th_dport;
2373#ifdef __FreeBSD__
2374 pi = &tcbinfo;
2375#else
2376 tb = &tcbtable;
2377#endif
2378 break;
2379 case IPPROTO_UDP:
2380 sport = pd->hdr.udp->uh_sport;
2381 dport = pd->hdr.udp->uh_dport;
2382#ifdef __FreeBSD__
2383 pi = &udbinfo;
2384#else
2385 tb = &udbtable;
2386#endif
2387 break;
2388 default:
2389 return (0);
2390 }
2391 if (direction == PF_IN) {
2392 saddr = pd->src;
2393 daddr = pd->dst;
2394 } else {
2395 u_int16_t p;
2396
2397 p = sport;
2398 sport = dport;
2399 dport = p;
2400 saddr = pd->dst;
2401 daddr = pd->src;
2402 }
2403 switch (pd->af) {
2404 case AF_INET:
2405#ifdef __FreeBSD__
2406 INP_INFO_RLOCK(pi); /* XXX LOR */
2407 inp = in_pcblookup_hash(pi, saddr->v4, sport, daddr->v4,
2408 dport, 0, NULL);
2409 if (inp == NULL) {
2410 inp = in_pcblookup_hash(pi, saddr->v4, sport,
2411 daddr->v4, dport, INPLOOKUP_WILDCARD, NULL);
2412 if(inp == NULL) {
2413 INP_INFO_RUNLOCK(pi);
2414 return (0);
2415 }
2416 }
2417#else
2418 inp = in_pcbhashlookup(tb, saddr->v4, sport, daddr->v4, dport);
2419 if (inp == NULL) {
2420 inp = in_pcblookup_listen(tb, daddr->v4, dport, 0);
2421 if (inp == NULL)
2422 return (0);
2423 }
2424#endif
2425 break;
2426#ifdef INET6
2427 case AF_INET6:
2428#ifdef __FreeBSD__
2429 INP_INFO_RLOCK(pi);
2430 inp = in6_pcblookup_hash(pi, &saddr->v6, sport,
2431 &daddr->v6, dport, 0, NULL);
2432 if (inp == NULL) {
2433 inp = in6_pcblookup_hash(pi, &saddr->v6, sport,
2434 &daddr->v6, dport, INPLOOKUP_WILDCARD, NULL);
2435 if (inp == NULL) {
2436 INP_INFO_RUNLOCK(pi);
2437 return (0);
2438 }
2439 }
2440#else
2441 inp = in6_pcbhashlookup(tb, &saddr->v6, sport, &daddr->v6,
2442 dport);
2443 if (inp == NULL) {
2444 inp = in6_pcblookup_listen(tb, &daddr->v6, dport, 0);
2445 if (inp == NULL)
2446 return (0);
2447 }
2448#endif
2449 break;
2450#endif /* INET6 */
2451
2452 default:
2453 return (0);
2454 }
2455#ifdef __FreeBSD__
2456 INP_LOCK(inp);
2457 *uid = inp->inp_socket->so_cred->cr_uid;
2458 *gid = inp->inp_socket->so_cred->cr_groups[0];
2459 INP_UNLOCK(inp);
2460 INP_INFO_RUNLOCK(pi);
2461#else
2462 *uid = inp->inp_socket->so_euid;
2463 *gid = inp->inp_socket->so_egid;
2464#endif
2465 return (1);
2466}
2467
2468u_int8_t
2469pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2470{
2471 int hlen;
2472 u_int8_t hdr[60];
2473 u_int8_t *opt, optlen;
2474 u_int8_t wscale = 0;
2475
2476 hlen = th_off << 2; /* hlen <= sizeof(hdr) */
2477 if (hlen <= sizeof(struct tcphdr))
2478 return (0);
2479 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2480 return (0);
2481 opt = hdr + sizeof(struct tcphdr);
2482 hlen -= sizeof(struct tcphdr);
2483 while (hlen >= 3) {
2484 switch (*opt) {
2485 case TCPOPT_EOL:
2486 case TCPOPT_NOP:
2487 ++opt;
2488 --hlen;
2489 break;
2490 case TCPOPT_WINDOW:
2491 wscale = opt[2];
2492 if (wscale > TCP_MAX_WINSHIFT)
2493 wscale = TCP_MAX_WINSHIFT;
2494 wscale |= PF_WSCALE_FLAG;
2495 /* FALLTHROUGH */
2496 default:
2497 optlen = opt[1];
2498 if (optlen < 2)
2499 optlen = 2;
2500 hlen -= optlen;
2501 opt += optlen;
2502 break;
2503 }
2504 }
2505 return (wscale);
2506}
2507
2508u_int16_t
2509pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2510{
2511 int hlen;
2512 u_int8_t hdr[60];
2513 u_int8_t *opt, optlen;
2514 u_int16_t mss = tcp_mssdflt;
2515
2516 hlen = th_off << 2; /* hlen <= sizeof(hdr) */
2517 if (hlen <= sizeof(struct tcphdr))
2518 return (0);
2519 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2520 return (0);
2521 opt = hdr + sizeof(struct tcphdr);
2522 hlen -= sizeof(struct tcphdr);
2523 while (hlen >= TCPOLEN_MAXSEG) {
2524 switch (*opt) {
2525 case TCPOPT_EOL:
2526 case TCPOPT_NOP:
2527 ++opt;
2528 --hlen;
2529 break;
2530 case TCPOPT_MAXSEG:
2531 bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2);
2532 /* FALLTHROUGH */
2533 default:
2534 optlen = opt[1];
2535 if (optlen < 2)
2536 optlen = 2;
2537 hlen -= optlen;
2538 opt += optlen;
2539 break;
2540 }
2541 }
2542 return (mss);
2543}
2544
2545u_int16_t
2546pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer)
2547{
2548#ifdef INET
2549 struct sockaddr_in *dst;
2550 struct route ro;
2551#endif /* INET */
2552#ifdef INET6
2553 struct sockaddr_in6 *dst6;
2554 struct route_in6 ro6;
2555#endif /* INET6 */
2556 struct rtentry *rt = NULL;
2557 int hlen = 0; /* make the compiler happy */
2558 u_int16_t mss = tcp_mssdflt;
2559
2560 switch (af) {
2561#ifdef INET
2562 case AF_INET:
2563 hlen = sizeof(struct ip);
2564 bzero(&ro, sizeof(ro));
2565 dst = (struct sockaddr_in *)&ro.ro_dst;
2566 dst->sin_family = AF_INET;
2567 dst->sin_len = sizeof(*dst);
2568 dst->sin_addr = addr->v4;
2569#ifdef __FreeBSD__
2570#ifdef RTF_PRCLONING
2571 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
2572#else /* !RTF_PRCLONING */
2573 rtalloc_ign(&ro, RTF_CLONING);
2574#endif
2575#else /* ! __FreeBSD__ */
2576 rtalloc_noclone(&ro, NO_CLONING);
2577#endif
2578 rt = ro.ro_rt;
2579 break;
2580#endif /* INET */
2581#ifdef INET6
2582 case AF_INET6:
2583 hlen = sizeof(struct ip6_hdr);
2584 bzero(&ro6, sizeof(ro6));
2585 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst;
2586 dst6->sin6_family = AF_INET6;
2587 dst6->sin6_len = sizeof(*dst6);
2588 dst6->sin6_addr = addr->v6;
2589#ifdef __FreeBSD__
2590#ifdef RTF_PRCLONING
2591 rtalloc_ign((struct route *)&ro6,
2592 (RTF_CLONING | RTF_PRCLONING));
2593#else /* !RTF_PRCLONING */
2594 rtalloc_ign((struct route *)&ro6, RTF_CLONING);
2595#endif
2596#else /* ! __FreeBSD__ */
2597 rtalloc_noclone((struct route *)&ro6, NO_CLONING);
2598#endif
2599 rt = ro6.ro_rt;
2600 break;
2601#endif /* INET6 */
2602 }
2603
2604 if (rt && rt->rt_ifp) {
2605 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr);
2606 mss = max(tcp_mssdflt, mss);
2607 RTFREE(rt);
2608 }
2609 mss = min(mss, offer);
2610 mss = max(mss, 64); /* sanity - at least max opt space */
2611 return (mss);
2612}
2613
2614void
2615pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr)
2616{
2617 struct pf_rule *r = s->rule.ptr;
2618
2619 s->rt_kif = NULL;
2620 if (!r->rt || r->rt == PF_FASTROUTE)
2621 return;
2622 switch (s->af) {
2623#ifdef INET
2624 case AF_INET:
2625 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL,
2626 &s->nat_src_node);
2627 s->rt_kif = r->rpool.cur->kif;
2628 break;
2629#endif /* INET */
2630#ifdef INET6
2631 case AF_INET6:
2632 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL,
2633 &s->nat_src_node);
2634 s->rt_kif = r->rpool.cur->kif;
2635 break;
2636#endif /* INET6 */
2637 }
2638}
2639
2640int
2641pf_test_tcp(struct pf_rule **rm, struct pf_state **sm, int direction,
2642 struct pfi_kif *kif, struct mbuf *m, int off, void *h,
2643 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
2644{
2645 struct pf_rule *nr = NULL;
2646 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
2647 struct tcphdr *th = pd->hdr.tcp;
2648 u_int16_t bport, nport = 0;
2649 sa_family_t af = pd->af;
2650 int lookup = -1;
2651 uid_t uid;
2652 gid_t gid;
2653 struct pf_rule *r, *a = NULL;
2654 struct pf_ruleset *ruleset = NULL;
2655 struct pf_src_node *nsn = NULL;
2656 u_short reason;
2657 int rewrite = 0;
2658 struct pf_tag *pftag = NULL;
2659 int tag = -1;
2660 u_int16_t mss = tcp_mssdflt;
2661
2662 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
2663
2664 if (direction == PF_OUT) {
2665 bport = nport = th->th_sport;
2666 /* check outgoing packet for BINAT/NAT */
2667 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
2668 saddr, th->th_sport, daddr, th->th_dport,
2669 &pd->naddr, &nport)) != NULL) {
2670 PF_ACPY(&pd->baddr, saddr, af);
2671 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
2672 &th->th_sum, &pd->naddr, nport, 0, af);
2673 rewrite++;
2674 if (nr->natpass)
2675 r = NULL;
2676 pd->nat_rule = nr;
2677 }
2678 } else {
2679 bport = nport = th->th_dport;
2680 /* check incoming packet for BINAT/RDR */
2681 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
2682 saddr, th->th_sport, daddr, th->th_dport,
2683 &pd->naddr, &nport)) != NULL) {
2684 PF_ACPY(&pd->baddr, daddr, af);
2685 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
2686 &th->th_sum, &pd->naddr, nport, 0, af);
2687 rewrite++;
2688 if (nr->natpass)
2689 r = NULL;
2690 pd->nat_rule = nr;
2691 }
2692 }
2693
2694 while (r != NULL) {
2695 r->evaluations++;
2696 if (r->kif != NULL &&
2697 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
2698 r = r->skip[PF_SKIP_IFP].ptr;
2699 else if (r->direction && r->direction != direction)
2700 r = r->skip[PF_SKIP_DIR].ptr;
2701 else if (r->af && r->af != af)
2702 r = r->skip[PF_SKIP_AF].ptr;
2703 else if (r->proto && r->proto != IPPROTO_TCP)
2704 r = r->skip[PF_SKIP_PROTO].ptr;
2705 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
2706 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
2707 else if (r->src.port_op && !pf_match_port(r->src.port_op,
2708 r->src.port[0], r->src.port[1], th->th_sport))
2709 r = r->skip[PF_SKIP_SRC_PORT].ptr;
2710 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
2711 r = r->skip[PF_SKIP_DST_ADDR].ptr;
2712 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
2713 r->dst.port[0], r->dst.port[1], th->th_dport))
2714 r = r->skip[PF_SKIP_DST_PORT].ptr;
2715 else if (r->tos && !(r->tos & pd->tos))
2716 r = TAILQ_NEXT(r, entries);
2717 else if (r->rule_flag & PFRULE_FRAGMENT)
2718 r = TAILQ_NEXT(r, entries);
2719 else if ((r->flagset & th->th_flags) != r->flags)
2720 r = TAILQ_NEXT(r, entries);
2721 else if (r->uid.op && (lookup != -1 || (lookup =
2722 pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2723 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
2724 uid))
2725 r = TAILQ_NEXT(r, entries);
2726 else if (r->gid.op && (lookup != -1 || (lookup =
2727 pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2728 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
2729 gid))
2730 r = TAILQ_NEXT(r, entries);
2731 else if (r->match_tag && !pf_match_tag(m, r, nr, pftag, &tag))
2732 r = TAILQ_NEXT(r, entries);
2733 else if (r->anchorname[0] && r->anchor == NULL)
2734 r = TAILQ_NEXT(r, entries);
2735 else if (r->os_fingerprint != PF_OSFP_ANY && !pf_osfp_match(
2736 pf_osfp_fingerprint(pd, m, off, th), r->os_fingerprint))
2737 r = TAILQ_NEXT(r, entries);
2738 else {
2739 if (r->tag)
2740 tag = r->tag;
2741 if (r->anchor == NULL) {
2742 *rm = r;
2743 *am = a;
2744 *rsm = ruleset;
2745 if ((*rm)->quick)
2746 break;
2747 r = TAILQ_NEXT(r, entries);
2748 } else
2749 PF_STEP_INTO_ANCHOR(r, a, ruleset,
2750 PF_RULESET_FILTER);
2751 }
2752 if (r == NULL && a != NULL)
2753 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
2754 PF_RULESET_FILTER);
2755 }
2756 r = *rm;
2757 a = *am;
2758 ruleset = *rsm;
2759
2760 REASON_SET(&reason, PFRES_MATCH);
2761
2762 if (r->log) {
2763 if (rewrite)
2764 m_copyback(m, off, sizeof(*th), (caddr_t)th);
2765 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
2766 }
2767
2768 if ((r->action == PF_DROP) &&
2769 ((r->rule_flag & PFRULE_RETURNRST) ||
2770 (r->rule_flag & PFRULE_RETURNICMP) ||
2771 (r->rule_flag & PFRULE_RETURN))) {
2772 /* undo NAT changes, if they have taken place */
2773 if (nr != NULL) {
2774 if (direction == PF_OUT) {
2775 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
2776 &th->th_sum, &pd->baddr, bport, 0, af);
2777 rewrite++;
2778 } else {
2779 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
2780 &th->th_sum, &pd->baddr, bport, 0, af);
2781 rewrite++;
2782 }
2783 }
2784 if (((r->rule_flag & PFRULE_RETURNRST) ||
2785 (r->rule_flag & PFRULE_RETURN)) &&
2786 !(th->th_flags & TH_RST)) {
2787 u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
2788
2789 if (th->th_flags & TH_SYN)
2790 ack++;
2791 if (th->th_flags & TH_FIN)
2792 ack++;
2793 pf_send_tcp(r, af, pd->dst,
2794 pd->src, th->th_dport, th->th_sport,
2795 ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0,
2796 r->return_ttl);
2797 } else if ((af == AF_INET) && r->return_icmp)
2798 pf_send_icmp(m, r->return_icmp >> 8,
2799 r->return_icmp & 255, af, r);
2800 else if ((af == AF_INET6) && r->return_icmp6)
2801 pf_send_icmp(m, r->return_icmp6 >> 8,
2802 r->return_icmp6 & 255, af, r);
2803 }
2804
2805 if (r->action == PF_DROP)
2806 return (PF_DROP);
2807
2808 if (pf_tag_packet(m, pftag, tag)) {
2809 REASON_SET(&reason, PFRES_MEMORY);
2810 return (PF_DROP);
2811 }
2812
2813 if (r->keep_state || nr != NULL ||
2814 (pd->flags & PFDESC_TCP_NORM)) {
2815 /* create new state */
2816 u_int16_t len;
2817 struct pf_state *s = NULL;
2818 struct pf_src_node *sn = NULL;
2819
2820 len = pd->tot_len - off - (th->th_off << 2);
2821
2822 /* check maximums */
2823 if (r->max_states && (r->states >= r->max_states))
2824 goto cleanup;
2825 /* src node for flter rule */
2826 if ((r->rule_flag & PFRULE_SRCTRACK ||
2827 r->rpool.opts & PF_POOL_STICKYADDR) &&
2828 pf_insert_src_node(&sn, r, saddr, af) != 0)
2829 goto cleanup;
2830 /* src node for translation rule */
2831 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
2832 ((direction == PF_OUT &&
2833 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
2834 (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
2835 goto cleanup;
2836 s = pool_get(&pf_state_pl, PR_NOWAIT);
2837 if (s == NULL) {
2838cleanup:
2839 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
2840 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
2841 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
2842 pf_status.src_nodes--;
2843 pool_put(&pf_src_tree_pl, sn);
2844 }
2845 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
2846 nsn->expire == 0) {
2847 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
2848 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
2849 pf_status.src_nodes--;
2850 pool_put(&pf_src_tree_pl, nsn);
2851 }
2852 REASON_SET(&reason, PFRES_MEMORY);
2853 return (PF_DROP);
2854 }
2855 bzero(s, sizeof(*s));
2856 r->states++;
2857 if (a != NULL)
2858 a->states++;
2859 s->rule.ptr = r;
2860 s->nat_rule.ptr = nr;
2861 if (s->nat_rule.ptr != NULL)
2862 s->nat_rule.ptr->states++;
2863 s->anchor.ptr = a;
2864 s->allow_opts = r->allow_opts;
2865 s->log = r->log & 2;
2866 s->proto = IPPROTO_TCP;
2867 s->direction = direction;
2868 s->af = af;
2869 if (direction == PF_OUT) {
2870 PF_ACPY(&s->gwy.addr, saddr, af);
2871 s->gwy.port = th->th_sport; /* sport */
2872 PF_ACPY(&s->ext.addr, daddr, af);
2873 s->ext.port = th->th_dport;
2874 if (nr != NULL) {
2875 PF_ACPY(&s->lan.addr, &pd->baddr, af);
2876 s->lan.port = bport;
2877 } else {
2878 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
2879 s->lan.port = s->gwy.port;
2880 }
2881 } else {
2882 PF_ACPY(&s->lan.addr, daddr, af);
2883 s->lan.port = th->th_dport;
2884 PF_ACPY(&s->ext.addr, saddr, af);
2885 s->ext.port = th->th_sport;
2886 if (nr != NULL) {
2887 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
2888 s->gwy.port = bport;
2889 } else {
2890 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
2891 s->gwy.port = s->lan.port;
2892 }
2893 }
2894
2895 s->src.seqlo = ntohl(th->th_seq);
2896 s->src.seqhi = s->src.seqlo + len + 1;
2897 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
2898 r->keep_state == PF_STATE_MODULATE) {
2899 /* Generate sequence number modulator */
2900 while ((s->src.seqdiff = arc4random()) == 0)
2901 ;
2902 pf_change_a(&th->th_seq, &th->th_sum,
2903 htonl(s->src.seqlo + s->src.seqdiff), 0);
2904 rewrite = 1;
2905 } else
2906 s->src.seqdiff = 0;
2907 if (th->th_flags & TH_SYN) {
2908 s->src.seqhi++;
2909 s->src.wscale = pf_get_wscale(m, off, th->th_off, af);
2910 }
2911 s->src.max_win = MAX(ntohs(th->th_win), 1);
2912 if (s->src.wscale & PF_WSCALE_MASK) {
2913 /* Remove scale factor from initial window */
2914 int win = s->src.max_win;
2915 win += 1 << (s->src.wscale & PF_WSCALE_MASK);
2916 s->src.max_win = (win - 1) >>
2917 (s->src.wscale & PF_WSCALE_MASK);
2918 }
2919 if (th->th_flags & TH_FIN)
2920 s->src.seqhi++;
2921 s->dst.seqhi = 1;
2922 s->dst.max_win = 1;
2923 s->src.state = TCPS_SYN_SENT;
2924 s->dst.state = TCPS_CLOSED;
2925#ifdef __FreeBSD__
2926 s->creation = time_second;
2927 s->expire = time_second;
2928#else
2929 s->creation = time.tv_sec;
2930 s->expire = time.tv_sec;
2931#endif
2932 s->timeout = PFTM_TCP_FIRST_PACKET;
2933 pf_set_rt_ifp(s, saddr);
2934 if (sn != NULL) {
2935 s->src_node = sn;
2936 s->src_node->states++;
2937 }
2938 if (nsn != NULL) {
2939 PF_ACPY(&nsn->raddr, &pd->naddr, af);
2940 s->nat_src_node = nsn;
2941 s->nat_src_node->states++;
2942 }
2943 if ((pd->flags & PFDESC_TCP_NORM) && pf_normalize_tcp_init(m,
2944 off, pd, th, &s->src, &s->dst)) {
2945 REASON_SET(&reason, PFRES_MEMORY);
2946 pf_src_tree_remove_state(s);
2947 pool_put(&pf_state_pl, s);
2948 return (PF_DROP);
2949 }
2950 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
2951 pf_normalize_tcp_stateful(m, off, pd, &reason, th, &s->src,
2952 &s->dst, &rewrite)) {
2953 pf_normalize_tcp_cleanup(s);
2954 pf_src_tree_remove_state(s);
2955 pool_put(&pf_state_pl, s);
2956 return (PF_DROP);
2957 }
2958 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
2959 pf_normalize_tcp_cleanup(s);
2960 REASON_SET(&reason, PFRES_MEMORY);
2961 pf_src_tree_remove_state(s);
2962 pool_put(&pf_state_pl, s);
2963 return (PF_DROP);
2964 } else
2965 *sm = s;
2966 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
2967 r->keep_state == PF_STATE_SYNPROXY) {
2968 s->src.state = PF_TCPS_PROXY_SRC;
2969 if (nr != NULL) {
2970 if (direction == PF_OUT) {
2971 pf_change_ap(saddr, &th->th_sport,
2972 pd->ip_sum, &th->th_sum, &pd->baddr,
2973 bport, 0, af);
2974 } else {
2975 pf_change_ap(daddr, &th->th_dport,
2976 pd->ip_sum, &th->th_sum, &pd->baddr,
2977 bport, 0, af);
2978 }
2979 }
2980 s->src.seqhi = arc4random();
2981 /* Find mss option */
2982 mss = pf_get_mss(m, off, th->th_off, af);
2983 mss = pf_calc_mss(saddr, af, mss);
2984 mss = pf_calc_mss(daddr, af, mss);
2985 s->src.mss = mss;
2986 pf_send_tcp(r, af, daddr, saddr, th->th_dport,
2987 th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1,
2988 TH_SYN|TH_ACK, 0, s->src.mss, 0);
2989 return (PF_SYNPROXY_DROP);
2990 }
2991 }
2992
2993 /* copy back packet headers if we performed NAT operations */
2994 if (rewrite)
2995 m_copyback(m, off, sizeof(*th), (caddr_t)th);
2996
2997 return (PF_PASS);
2998}
2999
3000int
3001pf_test_udp(struct pf_rule **rm, struct pf_state **sm, int direction,
3002 struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3003 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
3004{
3005 struct pf_rule *nr = NULL;
3006 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
3007 struct udphdr *uh = pd->hdr.udp;
3008 u_int16_t bport, nport = 0;
3009 sa_family_t af = pd->af;
3010 int lookup = -1;
3011 uid_t uid;
3012 gid_t gid;
3013 struct pf_rule *r, *a = NULL;
3014 struct pf_ruleset *ruleset = NULL;
3015 struct pf_src_node *nsn = NULL;
3016 u_short reason;
3017 int rewrite = 0;
3018 struct pf_tag *pftag = NULL;
3019 int tag = -1;
3020
3021 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3022
3023 if (direction == PF_OUT) {
3024 bport = nport = uh->uh_sport;
3025 /* check outgoing packet for BINAT/NAT */
3026 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3027 saddr, uh->uh_sport, daddr, uh->uh_dport,
3028 &pd->naddr, &nport)) != NULL) {
3029 PF_ACPY(&pd->baddr, saddr, af);
3030 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum,
3031 &uh->uh_sum, &pd->naddr, nport, 1, af);
3032 rewrite++;
3033 if (nr->natpass)
3034 r = NULL;
3035 pd->nat_rule = nr;
3036 }
3037 } else {
3038 bport = nport = uh->uh_dport;
3039 /* check incoming packet for BINAT/RDR */
3040 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3041 saddr, uh->uh_sport, daddr, uh->uh_dport, &pd->naddr,
3042 &nport)) != NULL) {
3043 PF_ACPY(&pd->baddr, daddr, af);
3044 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum,
3045 &uh->uh_sum, &pd->naddr, nport, 1, af);
3046 rewrite++;
3047 if (nr->natpass)
3048 r = NULL;
3049 pd->nat_rule = nr;
3050 }
3051 }
3052
3053 while (r != NULL) {
3054 r->evaluations++;
3055 if (r->kif != NULL &&
3056 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3057 r = r->skip[PF_SKIP_IFP].ptr;
3058 else if (r->direction && r->direction != direction)
3059 r = r->skip[PF_SKIP_DIR].ptr;
3060 else if (r->af && r->af != af)
3061 r = r->skip[PF_SKIP_AF].ptr;
3062 else if (r->proto && r->proto != IPPROTO_UDP)
3063 r = r->skip[PF_SKIP_PROTO].ptr;
3064 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
3065 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3066 else if (r->src.port_op && !pf_match_port(r->src.port_op,
3067 r->src.port[0], r->src.port[1], uh->uh_sport))
3068 r = r->skip[PF_SKIP_SRC_PORT].ptr;
3069 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
3070 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3071 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
3072 r->dst.port[0], r->dst.port[1], uh->uh_dport))
3073 r = r->skip[PF_SKIP_DST_PORT].ptr;
3074 else if (r->tos && !(r->tos & pd->tos))
3075 r = TAILQ_NEXT(r, entries);
3076 else if (r->rule_flag & PFRULE_FRAGMENT)
3077 r = TAILQ_NEXT(r, entries);
3078 else if (r->uid.op && (lookup != -1 || (lookup =
3079 pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
3080 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
3081 uid))
3082 r = TAILQ_NEXT(r, entries);
3083 else if (r->gid.op && (lookup != -1 || (lookup =
3084 pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
3085 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
3086 gid))
3087 r = TAILQ_NEXT(r, entries);
3088 else if (r->match_tag && !pf_match_tag(m, r, nr, pftag, &tag))
3089 r = TAILQ_NEXT(r, entries);
3090 else if (r->anchorname[0] && r->anchor == NULL)
3091 r = TAILQ_NEXT(r, entries);
3092 else if (r->os_fingerprint != PF_OSFP_ANY)
3093 r = TAILQ_NEXT(r, entries);
3094 else {
3095 if (r->tag)
3096 tag = r->tag;
3097 if (r->anchor == NULL) {
3098 *rm = r;
3099 *am = a;
3100 *rsm = ruleset;
3101 if ((*rm)->quick)
3102 break;
3103 r = TAILQ_NEXT(r, entries);
3104 } else
3105 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3106 PF_RULESET_FILTER);
3107 }
3108 if (r == NULL && a != NULL)
3109 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3110 PF_RULESET_FILTER);
3111 }
3112 r = *rm;
3113 a = *am;
3114 ruleset = *rsm;
3115
3116 REASON_SET(&reason, PFRES_MATCH);
3117
3118 if (r->log) {
3119 if (rewrite)
3120 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
3121 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3122 }
3123
3124 if ((r->action == PF_DROP) &&
3125 ((r->rule_flag & PFRULE_RETURNICMP) ||
3126 (r->rule_flag & PFRULE_RETURN))) {
3127 /* undo NAT changes, if they have taken place */
3128 if (nr != NULL) {
3129 if (direction == PF_OUT) {
3130 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum,
3131 &uh->uh_sum, &pd->baddr, bport, 1, af);
3132 rewrite++;
3133 } else {
3134 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum,
3135 &uh->uh_sum, &pd->baddr, bport, 1, af);
3136 rewrite++;
3137 }
3138 }
3139 if ((af == AF_INET) && r->return_icmp)
3140 pf_send_icmp(m, r->return_icmp >> 8,
3141 r->return_icmp & 255, af, r);
3142 else if ((af == AF_INET6) && r->return_icmp6)
3143 pf_send_icmp(m, r->return_icmp6 >> 8,
3144 r->return_icmp6 & 255, af, r);
3145 }
3146
3147 if (r->action == PF_DROP)
3148 return (PF_DROP);
3149
3150 if (pf_tag_packet(m, pftag, tag)) {
3151 REASON_SET(&reason, PFRES_MEMORY);
3152 return (PF_DROP);
3153 }
3154
3155 if (r->keep_state || nr != NULL) {
3156 /* create new state */
3157 struct pf_state *s = NULL;
3158 struct pf_src_node *sn = NULL;
3159
3160 /* check maximums */
3161 if (r->max_states && (r->states >= r->max_states))
3162 goto cleanup;
3163 /* src node for flter rule */
3164 if ((r->rule_flag & PFRULE_SRCTRACK ||
3165 r->rpool.opts & PF_POOL_STICKYADDR) &&
3166 pf_insert_src_node(&sn, r, saddr, af) != 0)
3167 goto cleanup;
3168 /* src node for translation rule */
3169 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3170 ((direction == PF_OUT &&
3171 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3172 (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
3173 goto cleanup;
3174 s = pool_get(&pf_state_pl, PR_NOWAIT);
3175 if (s == NULL) {
3176cleanup:
3177 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3178 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3179 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3180 pf_status.src_nodes--;
3181 pool_put(&pf_src_tree_pl, sn);
3182 }
3183 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3184 nsn->expire == 0) {
3185 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3186 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3187 pf_status.src_nodes--;
3188 pool_put(&pf_src_tree_pl, nsn);
3189 }
3190 REASON_SET(&reason, PFRES_MEMORY);
3191 return (PF_DROP);
3192 }
3193 bzero(s, sizeof(*s));
3194 r->states++;
3195 if (a != NULL)
3196 a->states++;
3197 s->rule.ptr = r;
3198 s->nat_rule.ptr = nr;
3199 if (s->nat_rule.ptr != NULL)
3200 s->nat_rule.ptr->states++;
3201 s->anchor.ptr = a;
3202 s->allow_opts = r->allow_opts;
3203 s->log = r->log & 2;
3204 s->proto = IPPROTO_UDP;
3205 s->direction = direction;
3206 s->af = af;
3207 if (direction == PF_OUT) {
3208 PF_ACPY(&s->gwy.addr, saddr, af);
3209 s->gwy.port = uh->uh_sport;
3210 PF_ACPY(&s->ext.addr, daddr, af);
3211 s->ext.port = uh->uh_dport;
3212 if (nr != NULL) {
3213 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3214 s->lan.port = bport;
3215 } else {
3216 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3217 s->lan.port = s->gwy.port;
3218 }
3219 } else {
3220 PF_ACPY(&s->lan.addr, daddr, af);
3221 s->lan.port = uh->uh_dport;
3222 PF_ACPY(&s->ext.addr, saddr, af);
3223 s->ext.port = uh->uh_sport;
3224 if (nr != NULL) {
3225 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3226 s->gwy.port = bport;
3227 } else {
3228 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3229 s->gwy.port = s->lan.port;
3230 }
3231 }
3232 s->src.state = PFUDPS_SINGLE;
3233 s->dst.state = PFUDPS_NO_TRAFFIC;
3234#ifdef __FreeBSD__
3235 s->creation = time_second;
3236 s->expire = time_second;
3237#else
3238 s->creation = time.tv_sec;
3239 s->expire = time.tv_sec;
3240#endif
3241 s->timeout = PFTM_UDP_FIRST_PACKET;
3242 pf_set_rt_ifp(s, saddr);
3243 if (sn != NULL) {
3244 s->src_node = sn;
3245 s->src_node->states++;
3246 }
3247 if (nsn != NULL) {
3248 PF_ACPY(&nsn->raddr, &pd->naddr, af);
3249 s->nat_src_node = nsn;
3250 s->nat_src_node->states++;
3251 }
3252 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3253 REASON_SET(&reason, PFRES_MEMORY);
3254 pf_src_tree_remove_state(s);
3255 pool_put(&pf_state_pl, s);
3256 return (PF_DROP);
3257 } else
3258 *sm = s;
3259 }
3260
3261 /* copy back packet headers if we performed NAT operations */
3262 if (rewrite)
3263 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
3264
3265 return (PF_PASS);
3266}
3267
3268int
3269pf_test_icmp(struct pf_rule **rm, struct pf_state **sm, int direction,
3270 struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3271 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
3272{
3273 struct pf_rule *nr = NULL;
3274 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
3275 struct pf_rule *r, *a = NULL;
3276 struct pf_ruleset *ruleset = NULL;
3277 struct pf_src_node *nsn = NULL;
3278 u_short reason;
3279 u_int16_t icmpid = 0; /* make the compiler happy */
3280 sa_family_t af = pd->af;
3281 u_int8_t icmptype = 0; /* make the compiler happy */
3282 u_int8_t icmpcode = 0; /* make the compiler happy */
3283 int state_icmp = 0;
3284 struct pf_tag *pftag = NULL;
3285 int tag = -1;
3286#ifdef INET6
3287 int rewrite = 0;
3288#endif /* INET6 */
3289
3290 switch (pd->proto) {
3291#ifdef INET
3292 case IPPROTO_ICMP:
3293 icmptype = pd->hdr.icmp->icmp_type;
3294 icmpcode = pd->hdr.icmp->icmp_code;
3295 icmpid = pd->hdr.icmp->icmp_id;
3296
3297 if (icmptype == ICMP_UNREACH ||
3298 icmptype == ICMP_SOURCEQUENCH ||
3299 icmptype == ICMP_REDIRECT ||
3300 icmptype == ICMP_TIMXCEED ||
3301 icmptype == ICMP_PARAMPROB)
3302 state_icmp++;
3303 break;
3304#endif /* INET */
3305#ifdef INET6
3306 case IPPROTO_ICMPV6:
3307 icmptype = pd->hdr.icmp6->icmp6_type;
3308 icmpcode = pd->hdr.icmp6->icmp6_code;
3309 icmpid = pd->hdr.icmp6->icmp6_id;
3310
3311 if (icmptype == ICMP6_DST_UNREACH ||
3312 icmptype == ICMP6_PACKET_TOO_BIG ||
3313 icmptype == ICMP6_TIME_EXCEEDED ||
3314 icmptype == ICMP6_PARAM_PROB)
3315 state_icmp++;
3316 break;
3317#endif /* INET6 */
3318 }
3319
3320 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3321
3322 if (direction == PF_OUT) {
3323 /* check outgoing packet for BINAT/NAT */
3324 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3325 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3326 PF_ACPY(&pd->baddr, saddr, af);
3327 switch (af) {
3328#ifdef INET
3329 case AF_INET:
3330 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3331 pd->naddr.v4.s_addr, 0);
3332 break;
3333#endif /* INET */
3334#ifdef INET6
3335 case AF_INET6:
3336 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum,
3337 &pd->naddr, 0);
3338 rewrite++;
3339 break;
3340#endif /* INET6 */
3341 }
3342 if (nr->natpass)
3343 r = NULL;
3344 pd->nat_rule = nr;
3345 }
3346 } else {
3347 /* check incoming packet for BINAT/RDR */
3348 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3349 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3350 PF_ACPY(&pd->baddr, daddr, af);
3351 switch (af) {
3352#ifdef INET
3353 case AF_INET:
3354 pf_change_a(&daddr->v4.s_addr,
3355 pd->ip_sum, pd->naddr.v4.s_addr, 0);
3356 break;
3357#endif /* INET */
3358#ifdef INET6
3359 case AF_INET6:
3360 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum,
3361 &pd->naddr, 0);
3362 rewrite++;
3363 break;
3364#endif /* INET6 */
3365 }
3366 if (nr->natpass)
3367 r = NULL;
3368 pd->nat_rule = nr;
3369 }
3370 }
3371
3372 while (r != NULL) {
3373 r->evaluations++;
3374 if (r->kif != NULL &&
3375 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3376 r = r->skip[PF_SKIP_IFP].ptr;
3377 else if (r->direction && r->direction != direction)
3378 r = r->skip[PF_SKIP_DIR].ptr;
3379 else if (r->af && r->af != af)
3380 r = r->skip[PF_SKIP_AF].ptr;
3381 else if (r->proto && r->proto != pd->proto)
3382 r = r->skip[PF_SKIP_PROTO].ptr;
3383 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
3384 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3385 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
3386 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3387 else if (r->type && r->type != icmptype + 1)
3388 r = TAILQ_NEXT(r, entries);
3389 else if (r->code && r->code != icmpcode + 1)
3390 r = TAILQ_NEXT(r, entries);
3391 else if (r->tos && !(r->tos & pd->tos))
3392 r = TAILQ_NEXT(r, entries);
3393 else if (r->rule_flag & PFRULE_FRAGMENT)
3394 r = TAILQ_NEXT(r, entries);
3395 else if (r->match_tag && !pf_match_tag(m, r, nr, pftag, &tag))
3396 r = TAILQ_NEXT(r, entries);
3397 else if (r->anchorname[0] && r->anchor == NULL)
3398 r = TAILQ_NEXT(r, entries);
3399 else if (r->os_fingerprint != PF_OSFP_ANY)
3400 r = TAILQ_NEXT(r, entries);
3401 else {
3402 if (r->tag)
3403 tag = r->tag;
3404 if (r->anchor == NULL) {
3405 *rm = r;
3406 *am = a;
3407 *rsm = ruleset;
3408 if ((*rm)->quick)
3409 break;
3410 r = TAILQ_NEXT(r, entries);
3411 } else
3412 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3413 PF_RULESET_FILTER);
3414 }
3415 if (r == NULL && a != NULL)
3416 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3417 PF_RULESET_FILTER);
3418 }
3419 r = *rm;
3420 a = *am;
3421 ruleset = *rsm;
3422
3423 REASON_SET(&reason, PFRES_MATCH);
3424
3425 if (r->log) {
3426#ifdef INET6
3427 if (rewrite)
3428 m_copyback(m, off, sizeof(struct icmp6_hdr),
3429 (caddr_t)pd->hdr.icmp6);
3430#endif /* INET6 */
3431 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3432 }
3433
3434 if (r->action != PF_PASS)
3435 return (PF_DROP);
3436
3437 if (pf_tag_packet(m, pftag, tag)) {
3438 REASON_SET(&reason, PFRES_MEMORY);
3439 return (PF_DROP);
3440 }
3441
3442 if (!state_icmp && (r->keep_state || nr != NULL)) {
3443 /* create new state */
3444 struct pf_state *s = NULL;
3445 struct pf_src_node *sn = NULL;
3446
3447 /* check maximums */
3448 if (r->max_states && (r->states >= r->max_states))
3449 goto cleanup;
3450 /* src node for flter rule */
3451 if ((r->rule_flag & PFRULE_SRCTRACK ||
3452 r->rpool.opts & PF_POOL_STICKYADDR) &&
3453 pf_insert_src_node(&sn, r, saddr, af) != 0)
3454 goto cleanup;
3455 /* src node for translation rule */
3456 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3457 ((direction == PF_OUT &&
3458 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3459 (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
3460 goto cleanup;
3461 s = pool_get(&pf_state_pl, PR_NOWAIT);
3462 if (s == NULL) {
3463cleanup:
3464 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3465 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3466 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3467 pf_status.src_nodes--;
3468 pool_put(&pf_src_tree_pl, sn);
3469 }
3470 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3471 nsn->expire == 0) {
3472 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3473 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3474 pf_status.src_nodes--;
3475 pool_put(&pf_src_tree_pl, nsn);
3476 }
3477 REASON_SET(&reason, PFRES_MEMORY);
3478 return (PF_DROP);
3479 }
3480 bzero(s, sizeof(*s));
3481 r->states++;
3482 if (a != NULL)
3483 a->states++;
3484 s->rule.ptr = r;
3485 s->nat_rule.ptr = nr;
3486 if (s->nat_rule.ptr != NULL)
3487 s->nat_rule.ptr->states++;
3488 s->anchor.ptr = a;
3489 s->allow_opts = r->allow_opts;
3490 s->log = r->log & 2;
3491 s->proto = pd->proto;
3492 s->direction = direction;
3493 s->af = af;
3494 if (direction == PF_OUT) {
3495 PF_ACPY(&s->gwy.addr, saddr, af);
3496 s->gwy.port = icmpid;
3497 PF_ACPY(&s->ext.addr, daddr, af);
3498 s->ext.port = icmpid;
3499 if (nr != NULL)
3500 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3501 else
3502 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3503 s->lan.port = icmpid;
3504 } else {
3505 PF_ACPY(&s->lan.addr, daddr, af);
3506 s->lan.port = icmpid;
3507 PF_ACPY(&s->ext.addr, saddr, af);
3508 s->ext.port = icmpid;
3509 if (nr != NULL)
3510 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3511 else
3512 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3513 s->gwy.port = icmpid;
3514 }
3515#ifdef __FreeBSD__
3516 s->creation = time_second;
3517 s->expire = time_second;
3518#else
3519 s->creation = time.tv_sec;
3520 s->expire = time.tv_sec;
3521#endif
3522 s->timeout = PFTM_ICMP_FIRST_PACKET;
3523 pf_set_rt_ifp(s, saddr);
3524 if (sn != NULL) {
3525 s->src_node = sn;
3526 s->src_node->states++;
3527 }
3528 if (nsn != NULL) {
3529 PF_ACPY(&nsn->raddr, &pd->naddr, af);
3530 s->nat_src_node = nsn;
3531 s->nat_src_node->states++;
3532 }
3533 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3534 REASON_SET(&reason, PFRES_MEMORY);
3535 pf_src_tree_remove_state(s);
3536 pool_put(&pf_state_pl, s);
3537 return (PF_DROP);
3538 } else
3539 *sm = s;
3540 }
3541
3542#ifdef INET6
3543 /* copy back packet headers if we performed IPv6 NAT operations */
3544 if (rewrite)
3545 m_copyback(m, off, sizeof(struct icmp6_hdr),
3546 (caddr_t)pd->hdr.icmp6);
3547#endif /* INET6 */
3548
3549 return (PF_PASS);
3550}
3551
3552int
3553pf_test_other(struct pf_rule **rm, struct pf_state **sm, int direction,
3554 struct pfi_kif *kif, struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
3555 struct pf_rule **am, struct pf_ruleset **rsm)
3556{
3557 struct pf_rule *nr = NULL;
3558 struct pf_rule *r, *a = NULL;
3559 struct pf_ruleset *ruleset = NULL;
3560 struct pf_src_node *nsn = NULL;
3561 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
3562 sa_family_t af = pd->af;
3563 u_short reason;
3564 struct pf_tag *pftag = NULL;
3565 int tag = -1;
3566
3567 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3568
3569 if (direction == PF_OUT) {
3570 /* check outgoing packet for BINAT/NAT */
3571 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3572 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3573 PF_ACPY(&pd->baddr, saddr, af);
3574 switch (af) {
3575#ifdef INET
3576 case AF_INET:
3577 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3578 pd->naddr.v4.s_addr, 0);
3579 break;
3580#endif /* INET */
3581#ifdef INET6
3582 case AF_INET6:
3583 PF_ACPY(saddr, &pd->naddr, af);
3584 break;
3585#endif /* INET6 */
3586 }
3587 if (nr->natpass)
3588 r = NULL;
3589 pd->nat_rule = nr;
3590 }
3591 } else {
3592 /* check incoming packet for BINAT/RDR */
3593 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3594 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3595 PF_ACPY(&pd->baddr, daddr, af);
3596 switch (af) {
3597#ifdef INET
3598 case AF_INET:
3599 pf_change_a(&daddr->v4.s_addr,
3600 pd->ip_sum, pd->naddr.v4.s_addr, 0);
3601 break;
3602#endif /* INET */
3603#ifdef INET6
3604 case AF_INET6:
3605 PF_ACPY(daddr, &pd->naddr, af);
3606 break;
3607#endif /* INET6 */
3608 }
3609 if (nr->natpass)
3610 r = NULL;
3611 pd->nat_rule = nr;
3612 }
3613 }
3614
3615 while (r != NULL) {
3616 r->evaluations++;
3617 if (r->kif != NULL &&
3618 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3619 r = r->skip[PF_SKIP_IFP].ptr;
3620 else if (r->direction && r->direction != direction)
3621 r = r->skip[PF_SKIP_DIR].ptr;
3622 else if (r->af && r->af != af)
3623 r = r->skip[PF_SKIP_AF].ptr;
3624 else if (r->proto && r->proto != pd->proto)
3625 r = r->skip[PF_SKIP_PROTO].ptr;
3626 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, r->src.not))
3627 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3628 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, r->dst.not))
3629 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3630 else if (r->tos && !(r->tos & pd->tos))
3631 r = TAILQ_NEXT(r, entries);
3632 else if (r->rule_flag & PFRULE_FRAGMENT)
3633 r = TAILQ_NEXT(r, entries);
3634 else if (r->match_tag && !pf_match_tag(m, r, nr, pftag, &tag))
3635 r = TAILQ_NEXT(r, entries);
3636 else if (r->anchorname[0] && r->anchor == NULL)
3637 r = TAILQ_NEXT(r, entries);
3638 else if (r->os_fingerprint != PF_OSFP_ANY)
3639 r = TAILQ_NEXT(r, entries);
3640 else {
3641 if (r->tag)
3642 tag = r->tag;
3643 if (r->anchor == NULL) {
3644 *rm = r;
3645 *am = a;
3646 *rsm = ruleset;
3647 if ((*rm)->quick)
3648 break;
3649 r = TAILQ_NEXT(r, entries);
3650 } else
3651 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3652 PF_RULESET_FILTER);
3653 }
3654 if (r == NULL && a != NULL)
3655 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3656 PF_RULESET_FILTER);
3657 }
3658 r = *rm;
3659 a = *am;
3660 ruleset = *rsm;
3661
3662 REASON_SET(&reason, PFRES_MATCH);
3663
3664 if (r->log)
3665 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3666
3667 if ((r->action == PF_DROP) &&
3668 ((r->rule_flag & PFRULE_RETURNICMP) ||
3669 (r->rule_flag & PFRULE_RETURN))) {
3670 struct pf_addr *a = NULL;
3671
3672 if (nr != NULL) {
3673 if (direction == PF_OUT)
3674 a = saddr;
3675 else
3676 a = daddr;
3677 }
3678 if (a != NULL) {
3679 switch (af) {
3680#ifdef INET
3681 case AF_INET:
3682 pf_change_a(&a->v4.s_addr, pd->ip_sum,
3683 pd->baddr.v4.s_addr, 0);
3684 break;
3685#endif /* INET */
3686#ifdef INET6
3687 case AF_INET6:
3688 PF_ACPY(a, &pd->baddr, af);
3689 break;
3690#endif /* INET6 */
3691 }
3692 }
3693 if ((af == AF_INET) && r->return_icmp)
3694 pf_send_icmp(m, r->return_icmp >> 8,
3695 r->return_icmp & 255, af, r);
3696 else if ((af == AF_INET6) && r->return_icmp6)
3697 pf_send_icmp(m, r->return_icmp6 >> 8,
3698 r->return_icmp6 & 255, af, r);
3699 }
3700
3701 if (r->action != PF_PASS)
3702 return (PF_DROP);
3703
3704 if (pf_tag_packet(m, pftag, tag)) {
3705 REASON_SET(&reason, PFRES_MEMORY);
3706 return (PF_DROP);
3707 }
3708
3709 if (r->keep_state || nr != NULL) {
3710 /* create new state */
3711 struct pf_state *s = NULL;
3712 struct pf_src_node *sn = NULL;
3713
3714 /* check maximums */
3715 if (r->max_states && (r->states >= r->max_states))
3716 goto cleanup;
3717 /* src node for flter rule */
3718 if ((r->rule_flag & PFRULE_SRCTRACK ||
3719 r->rpool.opts & PF_POOL_STICKYADDR) &&
3720 pf_insert_src_node(&sn, r, saddr, af) != 0)
3721 goto cleanup;
3722 /* src node for translation rule */
3723 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3724 ((direction == PF_OUT &&
3725 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3726 (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
3727 goto cleanup;
3728 s = pool_get(&pf_state_pl, PR_NOWAIT);
3729 if (s == NULL) {
3730cleanup:
3731 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3732 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3733 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3734 pf_status.src_nodes--;
3735 pool_put(&pf_src_tree_pl, sn);
3736 }
3737 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3738 nsn->expire == 0) {
3739 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3740 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3741 pf_status.src_nodes--;
3742 pool_put(&pf_src_tree_pl, nsn);
3743 }
3744 REASON_SET(&reason, PFRES_MEMORY);
3745 return (PF_DROP);
3746 }
3747 bzero(s, sizeof(*s));
3748 r->states++;
3749 if (a != NULL)
3750 a->states++;
3751 s->rule.ptr = r;
3752 s->nat_rule.ptr = nr;
3753 if (s->nat_rule.ptr != NULL)
3754 s->nat_rule.ptr->states++;
3755 s->anchor.ptr = a;
3756 s->allow_opts = r->allow_opts;
3757 s->log = r->log & 2;
3758 s->proto = pd->proto;
3759 s->direction = direction;
3760 s->af = af;
3761 if (direction == PF_OUT) {
3762 PF_ACPY(&s->gwy.addr, saddr, af);
3763 PF_ACPY(&s->ext.addr, daddr, af);
3764 if (nr != NULL)
3765 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3766 else
3767 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3768 } else {
3769 PF_ACPY(&s->lan.addr, daddr, af);
3770 PF_ACPY(&s->ext.addr, saddr, af);
3771 if (nr != NULL)
3772 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3773 else
3774 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3775 }
3776 s->src.state = PFOTHERS_SINGLE;
3777 s->dst.state = PFOTHERS_NO_TRAFFIC;
3778#ifdef __FreeBSD__
3779 s->creation = time_second;
3780 s->expire = time_second;
3781#else
3782 s->creation = time.tv_sec;
3783 s->expire = time.tv_sec;
3784#endif
3785 s->timeout = PFTM_OTHER_FIRST_PACKET;
3786 pf_set_rt_ifp(s, saddr);
3787 if (sn != NULL) {
3788 s->src_node = sn;
3789 s->src_node->states++;
3790 }
3791 if (nsn != NULL) {
3792 PF_ACPY(&nsn->raddr, &pd->naddr, af);
3793 s->nat_src_node = nsn;
3794 s->nat_src_node->states++;
3795 }
3796 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3797 REASON_SET(&reason, PFRES_MEMORY);
3798 pf_src_tree_remove_state(s);
3799 pool_put(&pf_state_pl, s);
3800 return (PF_DROP);
3801 } else
3802 *sm = s;
3803 }
3804
3805 return (PF_PASS);
3806}
3807
3808int
3809pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif,
3810 struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am,
3811 struct pf_ruleset **rsm)
3812{
3813 struct pf_rule *r, *a = NULL;
3814 struct pf_ruleset *ruleset = NULL;
3815 sa_family_t af = pd->af;
3816 u_short reason;
3817 struct pf_tag *pftag = NULL;
3818 int tag = -1;
3819
3820 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3821 while (r != NULL) {
3822 r->evaluations++;
3823 if (r->kif != NULL &&
3824 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3825 r = r->skip[PF_SKIP_IFP].ptr;
3826 else if (r->direction && r->direction != direction)
3827 r = r->skip[PF_SKIP_DIR].ptr;
3828 else if (r->af && r->af != af)
3829 r = r->skip[PF_SKIP_AF].ptr;
3830 else if (r->proto && r->proto != pd->proto)
3831 r = r->skip[PF_SKIP_PROTO].ptr;
3832 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, r->src.not))
3833 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3834 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, r->dst.not))
3835 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3836 else if (r->tos && !(r->tos & pd->tos))
3837 r = TAILQ_NEXT(r, entries);
3838 else if (r->src.port_op || r->dst.port_op ||
3839 r->flagset || r->type || r->code ||
3840 r->os_fingerprint != PF_OSFP_ANY)
3841 r = TAILQ_NEXT(r, entries);
3842 else if (r->match_tag && !pf_match_tag(m, r, NULL, pftag, &tag))
3843 r = TAILQ_NEXT(r, entries);
3844 else if (r->anchorname[0] && r->anchor == NULL)
3845 r = TAILQ_NEXT(r, entries);
3846 else {
3847 if (r->anchor == NULL) {
3848 *rm = r;
3849 *am = a;
3850 *rsm = ruleset;
3851 if ((*rm)->quick)
3852 break;
3853 r = TAILQ_NEXT(r, entries);
3854 } else
3855 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3856 PF_RULESET_FILTER);
3857 }
3858 if (r == NULL && a != NULL)
3859 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3860 PF_RULESET_FILTER);
3861 }
3862 r = *rm;
3863 a = *am;
3864 ruleset = *rsm;
3865
3866 REASON_SET(&reason, PFRES_MATCH);
3867
3868 if (r->log)
3869 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3870
3871 if (r->action != PF_PASS)
3872 return (PF_DROP);
3873
3874 if (pf_tag_packet(m, pftag, tag)) {
3875 REASON_SET(&reason, PFRES_MEMORY);
3876 return (PF_DROP);
3877 }
3878
3879 return (PF_PASS);
3880}
3881
3882int
3883pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
3884 struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
3885 u_short *reason)
3886{
3887 struct pf_state key;
3888 struct tcphdr *th = pd->hdr.tcp;
3889 u_int16_t win = ntohs(th->th_win);
3890 u_int32_t ack, end, seq;
3891 u_int8_t sws, dws;
3892 int ackskew;
3893 int copyback = 0;
3894 struct pf_state_peer *src, *dst;
3895
3896 key.af = pd->af;
3897 key.proto = IPPROTO_TCP;
3898 if (direction == PF_IN) {
3899 PF_ACPY(&key.ext.addr, pd->src, key.af);
3900 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
3901 key.ext.port = th->th_sport;
3902 key.gwy.port = th->th_dport;
3903 } else {
3904 PF_ACPY(&key.lan.addr, pd->src, key.af);
3905 PF_ACPY(&key.ext.addr, pd->dst, key.af);
3906 key.lan.port = th->th_sport;
3907 key.ext.port = th->th_dport;
3908 }
3909
3910 STATE_LOOKUP();
3911
3912 if (direction == (*state)->direction) {
3913 src = &(*state)->src;
3914 dst = &(*state)->dst;
3915 } else {
3916 src = &(*state)->dst;
3917 dst = &(*state)->src;
3918 }
3919
3920 if ((*state)->src.state == PF_TCPS_PROXY_SRC) {
3921 if (direction != (*state)->direction)
3922 return (PF_SYNPROXY_DROP);
3923 if (th->th_flags & TH_SYN) {
3924 if (ntohl(th->th_seq) != (*state)->src.seqlo)
3925 return (PF_DROP);
3926 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
3927 pd->src, th->th_dport, th->th_sport,
3928 (*state)->src.seqhi, ntohl(th->th_seq) + 1,
3929 TH_SYN|TH_ACK, 0, (*state)->src.mss, 0);
3930 return (PF_SYNPROXY_DROP);
3931 } else if (!(th->th_flags & TH_ACK) ||
3932 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
3933 (ntohl(th->th_seq) != (*state)->src.seqlo + 1))
3934 return (PF_DROP);
3935 else
3936 (*state)->src.state = PF_TCPS_PROXY_DST;
3937 }
3938 if ((*state)->src.state == PF_TCPS_PROXY_DST) {
3939 struct pf_state_host *src, *dst;
3940
3941 if (direction == PF_OUT) {
3942 src = &(*state)->gwy;
3943 dst = &(*state)->ext;
3944 } else {
3945 src = &(*state)->ext;
3946 dst = &(*state)->lan;
3947 }
3948 if (direction == (*state)->direction) {
3949 if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) ||
3950 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
3951 (ntohl(th->th_seq) != (*state)->src.seqlo + 1))
3952 return (PF_DROP);
3953 (*state)->src.max_win = MAX(ntohs(th->th_win), 1);
3954 if ((*state)->dst.seqhi == 1)
3955 (*state)->dst.seqhi = arc4random();
3956 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
3957 &dst->addr, src->port, dst->port,
3958 (*state)->dst.seqhi, 0, TH_SYN, 0,
3959 (*state)->src.mss, 0);
3960 return (PF_SYNPROXY_DROP);
3961 } else if (((th->th_flags & (TH_SYN|TH_ACK)) !=
3962 (TH_SYN|TH_ACK)) ||
3963 (ntohl(th->th_ack) != (*state)->dst.seqhi + 1))
3964 return (PF_DROP);
3965 else {
3966 (*state)->dst.max_win = MAX(ntohs(th->th_win), 1);
3967 (*state)->dst.seqlo = ntohl(th->th_seq);
3968 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
3969 pd->src, th->th_dport, th->th_sport,
3970 ntohl(th->th_ack), ntohl(th->th_seq) + 1,
3971 TH_ACK, (*state)->src.max_win, 0, 0);
3972 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
3973 &dst->addr, src->port, dst->port,
3974 (*state)->src.seqhi + 1, (*state)->src.seqlo + 1,
3975 TH_ACK, (*state)->dst.max_win, 0, 0);
3976 (*state)->src.seqdiff = (*state)->dst.seqhi -
3977 (*state)->src.seqlo;
3978 (*state)->dst.seqdiff = (*state)->src.seqhi -
3979 (*state)->dst.seqlo;
3980 (*state)->src.seqhi = (*state)->src.seqlo +
3981 (*state)->src.max_win;
3982 (*state)->dst.seqhi = (*state)->dst.seqlo +
3983 (*state)->dst.max_win;
3984 (*state)->src.wscale = (*state)->dst.wscale = 0;
3985 (*state)->src.state = (*state)->dst.state =
3986 TCPS_ESTABLISHED;
3987 return (PF_SYNPROXY_DROP);
3988 }
3989 }
3990
3991 if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) {
3992 sws = src->wscale & PF_WSCALE_MASK;
3993 dws = dst->wscale & PF_WSCALE_MASK;
3994 } else
3995 sws = dws = 0;
3996
3997 /*
3998 * Sequence tracking algorithm from Guido van Rooij's paper:
3999 * http://www.madison-gurkha.com/publications/tcp_filtering/
4000 * tcp_filtering.ps
4001 */
4002
4003 seq = ntohl(th->th_seq);
4004 if (src->seqlo == 0) {
4005 /* First packet from this end. Set its state */
4006
4007 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) &&
4008 src->scrub == NULL) {
4009 if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) {
4010 REASON_SET(reason, PFRES_MEMORY);
4011 return (PF_DROP);
4012 }
4013 }
4014
4015 /* Deferred generation of sequence number modulator */
4016 if (dst->seqdiff && !src->seqdiff) {
4017 while ((src->seqdiff = arc4random()) == 0)
4018 ;
4019 ack = ntohl(th->th_ack) - dst->seqdiff;
4020 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
4021 src->seqdiff), 0);
4022 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
4023 copyback = 1;
4024 } else {
4025 ack = ntohl(th->th_ack);
4026 }
4027
4028 end = seq + pd->p_len;
4029 if (th->th_flags & TH_SYN) {
4030 end++;
4031 if (dst->wscale & PF_WSCALE_FLAG) {
4032 src->wscale = pf_get_wscale(m, off, th->th_off,
4033 pd->af);
4034 if (src->wscale & PF_WSCALE_FLAG) {
4035 /* Remove scale factor from initial
4036 * window */
4037 sws = src->wscale & PF_WSCALE_MASK;
4038 win = ((u_int32_t)win + (1 << sws) - 1)
4039 >> sws;
4040 dws = dst->wscale & PF_WSCALE_MASK;
4041 } else {
4042 /* fixup other window */
4043 dst->max_win <<= dst->wscale &
4044 PF_WSCALE_MASK;
4045 /* in case of a retrans SYN|ACK */
4046 dst->wscale = 0;
4047 }
4048 }
4049 }
4050 if (th->th_flags & TH_FIN)
4051 end++;
4052
4053 src->seqlo = seq;
4054 if (src->state < TCPS_SYN_SENT)
4055 src->state = TCPS_SYN_SENT;
4056
4057 /*
4058 * May need to slide the window (seqhi may have been set by
4059 * the crappy stack check or if we picked up the connection
4060 * after establishment)
4061 */
4062 if (src->seqhi == 1 ||
4063 SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi))
4064 src->seqhi = end + MAX(1, dst->max_win << dws);
4065 if (win > src->max_win)
4066 src->max_win = win;
4067
4068 } else {
4069 ack = ntohl(th->th_ack) - dst->seqdiff;
4070 if (src->seqdiff) {
4071 /* Modulate sequence numbers */
4072 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
4073 src->seqdiff), 0);
4074 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
4075 copyback = 1;
4076 }
4077 end = seq + pd->p_len;
4078 if (th->th_flags & TH_SYN)
4079 end++;
4080 if (th->th_flags & TH_FIN)
4081 end++;
4082 }
4083
4084 if ((th->th_flags & TH_ACK) == 0) {
4085 /* Let it pass through the ack skew check */
4086 ack = dst->seqlo;
4087 } else if ((ack == 0 &&
4088 (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) ||
4089 /* broken tcp stacks do not set ack */
4090 (dst->state < TCPS_SYN_SENT)) {
4091 /*
4092 * Many stacks (ours included) will set the ACK number in an
4093 * FIN|ACK if the SYN times out -- no sequence to ACK.
4094 */
4095 ack = dst->seqlo;
4096 }
4097
4098 if (seq == end) {
4099 /* Ease sequencing restrictions on no data packets */
4100 seq = src->seqlo;
4101 end = seq;
4102 }
4103
4104 ackskew = dst->seqlo - ack;
4105
4106#define MAXACKWINDOW (0xffff + 1500) /* 1500 is an arbitrary fudge factor */
4107 if (SEQ_GEQ(src->seqhi, end) &&
4108 /* Last octet inside other's window space */
4109 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) &&
4110 /* Retrans: not more than one window back */
4111 (ackskew >= -MAXACKWINDOW) &&
4112 /* Acking not more than one reassembled fragment backwards */
4113 (ackskew <= (MAXACKWINDOW << sws))) {
4114 /* Acking not more than one window forward */
4115
4116 /* update max window */
4117 if (src->max_win < win)
4118 src->max_win = win;
4119 /* synchronize sequencing */
4120 if (SEQ_GT(end, src->seqlo))
4121 src->seqlo = end;
4122 /* slide the window of what the other end can send */
4123 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4124 dst->seqhi = ack + MAX((win << sws), 1);
4125
4126
4127 /* update states */
4128 if (th->th_flags & TH_SYN)
4129 if (src->state < TCPS_SYN_SENT)
4130 src->state = TCPS_SYN_SENT;
4131 if (th->th_flags & TH_FIN)
4132 if (src->state < TCPS_CLOSING)
4133 src->state = TCPS_CLOSING;
4134 if (th->th_flags & TH_ACK) {
4135 if (dst->state == TCPS_SYN_SENT)
4136 dst->state = TCPS_ESTABLISHED;
4137 else if (dst->state == TCPS_CLOSING)
4138 dst->state = TCPS_FIN_WAIT_2;
4139 }
4140 if (th->th_flags & TH_RST)
4141 src->state = dst->state = TCPS_TIME_WAIT;
4142
4143 /* update expire time */
4144#ifdef __FreeBSD__
4145 (*state)->expire = time_second;
4146#else
4147 (*state)->expire = time.tv_sec;
4148#endif
4149 if (src->state >= TCPS_FIN_WAIT_2 &&
4150 dst->state >= TCPS_FIN_WAIT_2)
4151 (*state)->timeout = PFTM_TCP_CLOSED;
4152 else if (src->state >= TCPS_FIN_WAIT_2 ||
4153 dst->state >= TCPS_FIN_WAIT_2)
4154 (*state)->timeout = PFTM_TCP_FIN_WAIT;
4155 else if (src->state < TCPS_ESTABLISHED ||
4156 dst->state < TCPS_ESTABLISHED)
4157 (*state)->timeout = PFTM_TCP_OPENING;
4158 else if (src->state >= TCPS_CLOSING ||
4159 dst->state >= TCPS_CLOSING)
4160 (*state)->timeout = PFTM_TCP_CLOSING;
4161 else
4162 (*state)->timeout = PFTM_TCP_ESTABLISHED;
4163
4164 /* Fall through to PASS packet */
4165
4166 } else if ((dst->state < TCPS_SYN_SENT ||
4167 dst->state >= TCPS_FIN_WAIT_2 ||
4168 src->state >= TCPS_FIN_WAIT_2) &&
4169 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) &&
4170 /* Within a window forward of the originating packet */
4171 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) {
4172 /* Within a window backward of the originating packet */
4173
4174 /*
4175 * This currently handles three situations:
4176 * 1) Stupid stacks will shotgun SYNs before their peer
4177 * replies.
4178 * 2) When PF catches an already established stream (the
4179 * firewall rebooted, the state table was flushed, routes
4180 * changed...)
4181 * 3) Packets get funky immediately after the connection
4182 * closes (this should catch Solaris spurious ACK|FINs
4183 * that web servers like to spew after a close)
4184 *
4185 * This must be a little more careful than the above code
4186 * since packet floods will also be caught here. We don't
4187 * update the TTL here to mitigate the damage of a packet
4188 * flood and so the same code can handle awkward establishment
4189 * and a loosened connection close.
4190 * In the establishment case, a correct peer response will
4191 * validate the connection, go through the normal state code
4192 * and keep updating the state TTL.
4193 */
4194
4195 if (pf_status.debug >= PF_DEBUG_MISC) {
4196 printf("pf: loose state match: ");
4197 pf_print_state(*state);
4198 pf_print_flags(th->th_flags);
4199 printf(" seq=%u ack=%u len=%u ackskew=%d pkts=%d:%d\n",
4200 seq, ack, pd->p_len, ackskew,
4201 (*state)->packets[0], (*state)->packets[1]);
4202 }
4203
4204 /* update max window */
4205 if (src->max_win < win)
4206 src->max_win = win;
4207 /* synchronize sequencing */
4208 if (SEQ_GT(end, src->seqlo))
4209 src->seqlo = end;
4210 /* slide the window of what the other end can send */
4211 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4212 dst->seqhi = ack + MAX((win << sws), 1);
4213
4214 /*
4215 * Cannot set dst->seqhi here since this could be a shotgunned
4216 * SYN and not an already established connection.
4217 */
4218
4219 if (th->th_flags & TH_FIN)
4220 if (src->state < TCPS_CLOSING)
4221 src->state = TCPS_CLOSING;
4222 if (th->th_flags & TH_RST)
4223 src->state = dst->state = TCPS_TIME_WAIT;
4224
4225 /* Fall through to PASS packet */
4226
4227 } else {
4228 if ((*state)->dst.state == TCPS_SYN_SENT &&
4229 (*state)->src.state == TCPS_SYN_SENT) {
4230 /* Send RST for state mismatches during handshake */
4231 if (!(th->th_flags & TH_RST)) {
4232 u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
4233
4234 if (th->th_flags & TH_SYN)
4235 ack++;
4236 if (th->th_flags & TH_FIN)
4237 ack++;
4238 pf_send_tcp((*state)->rule.ptr, pd->af,
4239 pd->dst, pd->src, th->th_dport,
4240 th->th_sport, ntohl(th->th_ack), ack,
4241 TH_RST|TH_ACK, 0, 0,
4242 (*state)->rule.ptr->return_ttl);
4243 }
4244 src->seqlo = 0;
4245 src->seqhi = 1;
4246 src->max_win = 1;
4247 } else if (pf_status.debug >= PF_DEBUG_MISC) {
4248 printf("pf: BAD state: ");
4249 pf_print_state(*state);
4250 pf_print_flags(th->th_flags);
4251 printf(" seq=%u ack=%u len=%u ackskew=%d pkts=%d:%d "
4252 "dir=%s,%s\n", seq, ack, pd->p_len, ackskew,
4253 (*state)->packets[0], (*state)->packets[1],
4254 direction == PF_IN ? "in" : "out",
4255 direction == (*state)->direction ? "fwd" : "rev");
4256 printf("pf: State failure on: %c %c %c %c | %c %c\n",
4257 SEQ_GEQ(src->seqhi, end) ? ' ' : '1',
4258 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ?
4259 ' ': '2',
4260 (ackskew >= -MAXACKWINDOW) ? ' ' : '3',
4261 (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4',
4262 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5',
4263 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6');
4264 }
4265 return (PF_DROP);
4266 }
4267
4268 if (dst->scrub || src->scrub) {
4269 if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4270 src, dst, &copyback))
4271 return (PF_DROP);
4272 }
4273
4274 /* Any packets which have gotten here are to be passed */
4275
4276 /* translate source/destination address, if necessary */
4277 if (STATE_TRANSLATE(*state)) {
4278 if (direction == PF_OUT)
4279 pf_change_ap(pd->src, &th->th_sport, pd->ip_sum,
4280 &th->th_sum, &(*state)->gwy.addr,
4281 (*state)->gwy.port, 0, pd->af);
4282 else
4283 pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum,
4284 &th->th_sum, &(*state)->lan.addr,
4285 (*state)->lan.port, 0, pd->af);
4286 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4287 } else if (copyback) {
4288 /* Copyback sequence modulation or stateful scrub changes */
4289 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4290 }
4291
4292 return (PF_PASS);
4293}
4294
4295int
4296pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif,
4297 struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4298{
4299 struct pf_state_peer *src, *dst;
4300 struct pf_state key;
4301 struct udphdr *uh = pd->hdr.udp;
4302
4303 key.af = pd->af;
4304 key.proto = IPPROTO_UDP;
4305 if (direction == PF_IN) {
4306 PF_ACPY(&key.ext.addr, pd->src, key.af);
4307 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4308 key.ext.port = uh->uh_sport;
4309 key.gwy.port = uh->uh_dport;
4310 } else {
4311 PF_ACPY(&key.lan.addr, pd->src, key.af);
4312 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4313 key.lan.port = uh->uh_sport;
4314 key.ext.port = uh->uh_dport;
4315 }
4316
4317 STATE_LOOKUP();
4318
4319 if (direction == (*state)->direction) {
4320 src = &(*state)->src;
4321 dst = &(*state)->dst;
4322 } else {
4323 src = &(*state)->dst;
4324 dst = &(*state)->src;
4325 }
4326
4327 /* update states */
4328 if (src->state < PFUDPS_SINGLE)
4329 src->state = PFUDPS_SINGLE;
4330 if (dst->state == PFUDPS_SINGLE)
4331 dst->state = PFUDPS_MULTIPLE;
4332
4333 /* update expire time */
4334#ifdef __FreeBSD__
4335 (*state)->expire = time_second;
4336#else
4337 (*state)->expire = time.tv_sec;
4338#endif
4339 if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE)
4340 (*state)->timeout = PFTM_UDP_MULTIPLE;
4341 else
4342 (*state)->timeout = PFTM_UDP_SINGLE;
4343
4344 /* translate source/destination address, if necessary */
4345 if (STATE_TRANSLATE(*state)) {
4346 if (direction == PF_OUT)
4347 pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum,
4348 &uh->uh_sum, &(*state)->gwy.addr,
4349 (*state)->gwy.port, 1, pd->af);
4350 else
4351 pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum,
4352 &uh->uh_sum, &(*state)->lan.addr,
4353 (*state)->lan.port, 1, pd->af);
4354 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
4355 }
4356
4357 return (PF_PASS);
4358}
4359
4360int
4361pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif,
4362 struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4363{
4364 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
4365 u_int16_t icmpid = 0; /* make the compiler happy */
4366 u_int16_t *icmpsum = NULL; /* make the compiler happy */
4367 u_int8_t icmptype = 0; /* make the compiler happy */
4368 int state_icmp = 0;
4369
4370 switch (pd->proto) {
4371#ifdef INET
4372 case IPPROTO_ICMP:
4373 icmptype = pd->hdr.icmp->icmp_type;
4374 icmpid = pd->hdr.icmp->icmp_id;
4375 icmpsum = &pd->hdr.icmp->icmp_cksum;
4376
4377 if (icmptype == ICMP_UNREACH ||
4378 icmptype == ICMP_SOURCEQUENCH ||
4379 icmptype == ICMP_REDIRECT ||
4380 icmptype == ICMP_TIMXCEED ||
4381 icmptype == ICMP_PARAMPROB)
4382 state_icmp++;
4383 break;
4384#endif /* INET */
4385#ifdef INET6
4386 case IPPROTO_ICMPV6:
4387 icmptype = pd->hdr.icmp6->icmp6_type;
4388 icmpid = pd->hdr.icmp6->icmp6_id;
4389 icmpsum = &pd->hdr.icmp6->icmp6_cksum;
4390
4391 if (icmptype == ICMP6_DST_UNREACH ||
4392 icmptype == ICMP6_PACKET_TOO_BIG ||
4393 icmptype == ICMP6_TIME_EXCEEDED ||
4394 icmptype == ICMP6_PARAM_PROB)
4395 state_icmp++;
4396 break;
4397#endif /* INET6 */
4398 }
4399
4400 if (!state_icmp) {
4401
4402 /*
4403 * ICMP query/reply message not related to a TCP/UDP packet.
4404 * Search for an ICMP state.
4405 */
4406 struct pf_state key;
4407
4408 key.af = pd->af;
4409 key.proto = pd->proto;
4410 if (direction == PF_IN) {
4411 PF_ACPY(&key.ext.addr, pd->src, key.af);
4412 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4413 key.ext.port = icmpid;
4414 key.gwy.port = icmpid;
4415 } else {
4416 PF_ACPY(&key.lan.addr, pd->src, key.af);
4417 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4418 key.lan.port = icmpid;
4419 key.ext.port = icmpid;
4420 }
4421
4422 STATE_LOOKUP();
4423
4424#ifdef __FreeBSD__
4425 (*state)->expire = time_second;
4426#else
4427 (*state)->expire = time.tv_sec;
4428#endif
4429 (*state)->timeout = PFTM_ICMP_ERROR_REPLY;
4430
4431 /* translate source/destination address, if necessary */
4432 if (PF_ANEQ(&(*state)->lan.addr, &(*state)->gwy.addr, pd->af)) {
4433 if (direction == PF_OUT) {
4434 switch (pd->af) {
4435#ifdef INET
4436 case AF_INET:
4437 pf_change_a(&saddr->v4.s_addr,
4438 pd->ip_sum,
4439 (*state)->gwy.addr.v4.s_addr, 0);
4440 break;
4441#endif /* INET */
4442#ifdef INET6
4443 case AF_INET6:
4444 pf_change_a6(saddr,
4445 &pd->hdr.icmp6->icmp6_cksum,
4446 &(*state)->gwy.addr, 0);
4447 m_copyback(m, off,
4448 sizeof(struct icmp6_hdr),
4449 (caddr_t)pd->hdr.icmp6);
4450 break;
4451#endif /* INET6 */
4452 }
4453 } else {
4454 switch (pd->af) {
4455#ifdef INET
4456 case AF_INET:
4457 pf_change_a(&daddr->v4.s_addr,
4458 pd->ip_sum,
4459 (*state)->lan.addr.v4.s_addr, 0);
4460 break;
4461#endif /* INET */
4462#ifdef INET6
4463 case AF_INET6:
4464 pf_change_a6(daddr,
4465 &pd->hdr.icmp6->icmp6_cksum,
4466 &(*state)->lan.addr, 0);
4467 m_copyback(m, off,
4468 sizeof(struct icmp6_hdr),
4469 (caddr_t)pd->hdr.icmp6);
4470 break;
4471#endif /* INET6 */
4472 }
4473 }
4474 }
4475
4476 return (PF_PASS);
4477
4478 } else {
4479 /*
4480 * ICMP error message in response to a TCP/UDP packet.
4481 * Extract the inner TCP/UDP header and search for that state.
4482 */
4483
4484 struct pf_pdesc pd2;
4485#ifdef INET
4486 struct ip h2;
4487#endif /* INET */
4488#ifdef INET6
4489 struct ip6_hdr h2_6;
4490 int terminal = 0;
4491#endif /* INET6 */
4492 int ipoff2 = 0; /* make the compiler happy */
4493 int off2 = 0; /* make the compiler happy */
4494
4495 pd2.af = pd->af;
4496 switch (pd->af) {
4497#ifdef INET
4498 case AF_INET:
4499 /* offset of h2 in mbuf chain */
4500 ipoff2 = off + ICMP_MINLEN;
4501
4502 if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2),
4503 NULL, NULL, pd2.af)) {
4504 DPFPRINTF(PF_DEBUG_MISC,
4505 ("pf: ICMP error message too short "
4506 "(ip)\n"));
4507 return (PF_DROP);
4508 }
4509 /*
4510 * ICMP error messages don't refer to non-first
4511 * fragments
4512 */
4513 if (h2.ip_off & htons(IP_OFFMASK))
4514 return (PF_DROP);
4515
4516 /* offset of protocol header that follows h2 */
4517 off2 = ipoff2 + (h2.ip_hl << 2);
4518
4519 pd2.proto = h2.ip_p;
4520 pd2.src = (struct pf_addr *)&h2.ip_src;
4521 pd2.dst = (struct pf_addr *)&h2.ip_dst;
4522 pd2.ip_sum = &h2.ip_sum;
4523 break;
4524#endif /* INET */
4525#ifdef INET6
4526 case AF_INET6:
4527 ipoff2 = off + sizeof(struct icmp6_hdr);
4528
4529 if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6),
4530 NULL, NULL, pd2.af)) {
4531 DPFPRINTF(PF_DEBUG_MISC,
4532 ("pf: ICMP error message too short "
4533 "(ip6)\n"));
4534 return (PF_DROP);
4535 }
4536 pd2.proto = h2_6.ip6_nxt;
4537 pd2.src = (struct pf_addr *)&h2_6.ip6_src;
4538 pd2.dst = (struct pf_addr *)&h2_6.ip6_dst;
4539 pd2.ip_sum = NULL;
4540 off2 = ipoff2 + sizeof(h2_6);
4541 do {
4542 switch (pd2.proto) {
4543 case IPPROTO_FRAGMENT:
4544 /*
4545 * ICMPv6 error messages for
4546 * non-first fragments
4547 */
4548 return (PF_DROP);
4549 case IPPROTO_AH:
4550 case IPPROTO_HOPOPTS:
4551 case IPPROTO_ROUTING:
4552 case IPPROTO_DSTOPTS: {
4553 /* get next header and header length */
4554 struct ip6_ext opt6;
4555
4556 if (!pf_pull_hdr(m, off2, &opt6,
4557 sizeof(opt6), NULL, NULL, pd2.af)) {
4558 DPFPRINTF(PF_DEBUG_MISC,
4559 ("pf: ICMPv6 short opt\n"));
4560 return (PF_DROP);
4561 }
4562 if (pd2.proto == IPPROTO_AH)
4563 off2 += (opt6.ip6e_len + 2) * 4;
4564 else
4565 off2 += (opt6.ip6e_len + 1) * 8;
4566 pd2.proto = opt6.ip6e_nxt;
4567 /* goto the next header */
4568 break;
4569 }
4570 default:
4571 terminal++;
4572 break;
4573 }
4574 } while (!terminal);
4575 break;
4576#endif /* INET6 */
4577 }
4578
4579 switch (pd2.proto) {
4580 case IPPROTO_TCP: {
4581 struct tcphdr th;
4582 u_int32_t seq;
4583 struct pf_state key;
4584 struct pf_state_peer *src, *dst;
4585 u_int8_t dws;
4586 int copyback = 0;
4587
4588 /*
4589 * Only the first 8 bytes of the TCP header can be
4590 * expected. Don't access any TCP header fields after
4591 * th_seq, an ackskew test is not possible.
4592 */
4593 if (!pf_pull_hdr(m, off2, &th, 8, NULL, NULL, pd2.af)) {
4594 DPFPRINTF(PF_DEBUG_MISC,
4595 ("pf: ICMP error message too short "
4596 "(tcp)\n"));
4597 return (PF_DROP);
4598 }
4599
4600 key.af = pd2.af;
4601 key.proto = IPPROTO_TCP;
4602 if (direction == PF_IN) {
4603 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4604 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4605 key.ext.port = th.th_dport;
4606 key.gwy.port = th.th_sport;
4607 } else {
4608 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4609 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4610 key.lan.port = th.th_dport;
4611 key.ext.port = th.th_sport;
4612 }
4613
4614 STATE_LOOKUP();
4615
4616 if (direction == (*state)->direction) {
4617 src = &(*state)->dst;
4618 dst = &(*state)->src;
4619 } else {
4620 src = &(*state)->src;
4621 dst = &(*state)->dst;
4622 }
4623
4624 if (src->wscale && dst->wscale &&
4625 !(th.th_flags & TH_SYN))
4626 dws = dst->wscale & PF_WSCALE_MASK;
4627 else
4628 dws = 0;
4629
4630 /* Demodulate sequence number */
4631 seq = ntohl(th.th_seq) - src->seqdiff;
4632 if (src->seqdiff) {
4633 pf_change_a(&th.th_seq, icmpsum,
4634 htonl(seq), 0);
4635 copyback = 1;
4636 }
4637
4638 if (!SEQ_GEQ(src->seqhi, seq) ||
4639 !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws))) {
4640 if (pf_status.debug >= PF_DEBUG_MISC) {
4641 printf("pf: BAD ICMP %d:%d ",
4642 icmptype, pd->hdr.icmp->icmp_code);
4643 pf_print_host(pd->src, 0, pd->af);
4644 printf(" -> ");
4645 pf_print_host(pd->dst, 0, pd->af);
4646 printf(" state: ");
4647 pf_print_state(*state);
4648 printf(" seq=%u\n", seq);
4649 }
4650 return (PF_DROP);
4651 }
4652
4653 if (STATE_TRANSLATE(*state)) {
4654 if (direction == PF_IN) {
4655 pf_change_icmp(pd2.src, &th.th_sport,
4656 daddr, &(*state)->lan.addr,
4657 (*state)->lan.port, NULL,
4658 pd2.ip_sum, icmpsum,
4659 pd->ip_sum, 0, pd2.af);
4660 } else {
4661 pf_change_icmp(pd2.dst, &th.th_dport,
4662 saddr, &(*state)->gwy.addr,
4663 (*state)->gwy.port, NULL,
4664 pd2.ip_sum, icmpsum,
4665 pd->ip_sum, 0, pd2.af);
4666 }
4667 copyback = 1;
4668 }
4669
4670 if (copyback) {
4671 switch (pd2.af) {
4672#ifdef INET
4673 case AF_INET:
4674 m_copyback(m, off, ICMP_MINLEN,
4675 (caddr_t)pd->hdr.icmp);
4676 m_copyback(m, ipoff2, sizeof(h2),
4677 (caddr_t)&h2);
4678 break;
4679#endif /* INET */
4680#ifdef INET6
4681 case AF_INET6:
4682 m_copyback(m, off,
4683 sizeof(struct icmp6_hdr),
4684 (caddr_t)pd->hdr.icmp6);
4685 m_copyback(m, ipoff2, sizeof(h2_6),
4686 (caddr_t)&h2_6);
4687 break;
4688#endif /* INET6 */
4689 }
4690 m_copyback(m, off2, 8, (caddr_t)&th);
4691 }
4692
4693 return (PF_PASS);
4694 break;
4695 }
4696 case IPPROTO_UDP: {
4697 struct udphdr uh;
4698 struct pf_state key;
4699
4700 if (!pf_pull_hdr(m, off2, &uh, sizeof(uh),
4701 NULL, NULL, pd2.af)) {
4702 DPFPRINTF(PF_DEBUG_MISC,
4703 ("pf: ICMP error message too short "
4704 "(udp)\n"));
4705 return (PF_DROP);
4706 }
4707
4708 key.af = pd2.af;
4709 key.proto = IPPROTO_UDP;
4710 if (direction == PF_IN) {
4711 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4712 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4713 key.ext.port = uh.uh_dport;
4714 key.gwy.port = uh.uh_sport;
4715 } else {
4716 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4717 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4718 key.lan.port = uh.uh_dport;
4719 key.ext.port = uh.uh_sport;
4720 }
4721
4722 STATE_LOOKUP();
4723
4724 if (STATE_TRANSLATE(*state)) {
4725 if (direction == PF_IN) {
4726 pf_change_icmp(pd2.src, &uh.uh_sport,
4727 daddr, &(*state)->lan.addr,
4728 (*state)->lan.port, &uh.uh_sum,
4729 pd2.ip_sum, icmpsum,
4730 pd->ip_sum, 1, pd2.af);
4731 } else {
4732 pf_change_icmp(pd2.dst, &uh.uh_dport,
4733 saddr, &(*state)->gwy.addr,
4734 (*state)->gwy.port, &uh.uh_sum,
4735 pd2.ip_sum, icmpsum,
4736 pd->ip_sum, 1, pd2.af);
4737 }
4738 switch (pd2.af) {
4739#ifdef INET
4740 case AF_INET:
4741 m_copyback(m, off, ICMP_MINLEN,
4742 (caddr_t)pd->hdr.icmp);
4743 m_copyback(m, ipoff2, sizeof(h2),
4744 (caddr_t)&h2);
4745 break;
4746#endif /* INET */
4747#ifdef INET6
4748 case AF_INET6:
4749 m_copyback(m, off,
4750 sizeof(struct icmp6_hdr),
4751 (caddr_t)pd->hdr.icmp6);
4752 m_copyback(m, ipoff2, sizeof(h2_6),
4753 (caddr_t)&h2_6);
4754 break;
4755#endif /* INET6 */
4756 }
4757 m_copyback(m, off2, sizeof(uh),
4758 (caddr_t)&uh);
4759 }
4760
4761 return (PF_PASS);
4762 break;
4763 }
4764#ifdef INET
4765 case IPPROTO_ICMP: {
4766 struct icmp iih;
4767 struct pf_state key;
4768
4769 if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN,
4770 NULL, NULL, pd2.af)) {
4771 DPFPRINTF(PF_DEBUG_MISC,
4772 ("pf: ICMP error message too short i"
4773 "(icmp)\n"));
4774 return (PF_DROP);
4775 }
4776
4777 key.af = pd2.af;
4778 key.proto = IPPROTO_ICMP;
4779 if (direction == PF_IN) {
4780 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4781 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4782 key.ext.port = iih.icmp_id;
4783 key.gwy.port = iih.icmp_id;
4784 } else {
4785 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4786 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4787 key.lan.port = iih.icmp_id;
4788 key.ext.port = iih.icmp_id;
4789 }
4790
4791 STATE_LOOKUP();
4792
4793 if (STATE_TRANSLATE(*state)) {
4794 if (direction == PF_IN) {
4795 pf_change_icmp(pd2.src, &iih.icmp_id,
4796 daddr, &(*state)->lan.addr,
4797 (*state)->lan.port, NULL,
4798 pd2.ip_sum, icmpsum,
4799 pd->ip_sum, 0, AF_INET);
4800 } else {
4801 pf_change_icmp(pd2.dst, &iih.icmp_id,
4802 saddr, &(*state)->gwy.addr,
4803 (*state)->gwy.port, NULL,
4804 pd2.ip_sum, icmpsum,
4805 pd->ip_sum, 0, AF_INET);
4806 }
4807 m_copyback(m, off, ICMP_MINLEN,
4808 (caddr_t)pd->hdr.icmp);
4809 m_copyback(m, ipoff2, sizeof(h2),
4810 (caddr_t)&h2);
4811 m_copyback(m, off2, ICMP_MINLEN,
4812 (caddr_t)&iih);
4813 }
4814
4815 return (PF_PASS);
4816 break;
4817 }
4818#endif /* INET */
4819#ifdef INET6
4820 case IPPROTO_ICMPV6: {
4821 struct icmp6_hdr iih;
4822 struct pf_state key;
4823
4824 if (!pf_pull_hdr(m, off2, &iih,
4825 sizeof(struct icmp6_hdr), NULL, NULL, pd2.af)) {
4826 DPFPRINTF(PF_DEBUG_MISC,
4827 ("pf: ICMP error message too short "
4828 "(icmp6)\n"));
4829 return (PF_DROP);
4830 }
4831
4832 key.af = pd2.af;
4833 key.proto = IPPROTO_ICMPV6;
4834 if (direction == PF_IN) {
4835 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4836 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4837 key.ext.port = iih.icmp6_id;
4838 key.gwy.port = iih.icmp6_id;
4839 } else {
4840 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4841 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4842 key.lan.port = iih.icmp6_id;
4843 key.ext.port = iih.icmp6_id;
4844 }
4845
4846 STATE_LOOKUP();
4847
4848 if (STATE_TRANSLATE(*state)) {
4849 if (direction == PF_IN) {
4850 pf_change_icmp(pd2.src, &iih.icmp6_id,
4851 daddr, &(*state)->lan.addr,
4852 (*state)->lan.port, NULL,
4853 pd2.ip_sum, icmpsum,
4854 pd->ip_sum, 0, AF_INET6);
4855 } else {
4856 pf_change_icmp(pd2.dst, &iih.icmp6_id,
4857 saddr, &(*state)->gwy.addr,
4858 (*state)->gwy.port, NULL,
4859 pd2.ip_sum, icmpsum,
4860 pd->ip_sum, 0, AF_INET6);
4861 }
4862 m_copyback(m, off, sizeof(struct icmp6_hdr),
4863 (caddr_t)pd->hdr.icmp6);
4864 m_copyback(m, ipoff2, sizeof(h2_6),
4865 (caddr_t)&h2_6);
4866 m_copyback(m, off2, sizeof(struct icmp6_hdr),
4867 (caddr_t)&iih);
4868 }
4869
4870 return (PF_PASS);
4871 break;
4872 }
4873#endif /* INET6 */
4874 default: {
4875 struct pf_state key;
4876
4877 key.af = pd2.af;
4878 key.proto = pd2.proto;
4879 if (direction == PF_IN) {
4880 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4881 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4882 key.ext.port = 0;
4883 key.gwy.port = 0;
4884 } else {
4885 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4886 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4887 key.lan.port = 0;
4888 key.ext.port = 0;
4889 }
4890
4891 STATE_LOOKUP();
4892
4893 if (STATE_TRANSLATE(*state)) {
4894 if (direction == PF_IN) {
4895 pf_change_icmp(pd2.src, NULL,
4896 daddr, &(*state)->lan.addr,
4897 0, NULL,
4898 pd2.ip_sum, icmpsum,
4899 pd->ip_sum, 0, pd2.af);
4900 } else {
4901 pf_change_icmp(pd2.dst, NULL,
4902 saddr, &(*state)->gwy.addr,
4903 0, NULL,
4904 pd2.ip_sum, icmpsum,
4905 pd->ip_sum, 0, pd2.af);
4906 }
4907 switch (pd2.af) {
4908#ifdef INET
4909 case AF_INET:
4910 m_copyback(m, off, ICMP_MINLEN,
4911 (caddr_t)pd->hdr.icmp);
4912 m_copyback(m, ipoff2, sizeof(h2),
4913 (caddr_t)&h2);
4914 break;
4915#endif /* INET */
4916#ifdef INET6
4917 case AF_INET6:
4918 m_copyback(m, off,
4919 sizeof(struct icmp6_hdr),
4920 (caddr_t)pd->hdr.icmp6);
4921 m_copyback(m, ipoff2, sizeof(h2_6),
4922 (caddr_t)&h2_6);
4923 break;
4924#endif /* INET6 */
4925 }
4926 }
4927
4928 return (PF_PASS);
4929 break;
4930 }
4931 }
4932 }
4933}
4934
4935int
4936pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif,
4937 struct pf_pdesc *pd)
4938{
4939 struct pf_state_peer *src, *dst;
4940 struct pf_state key;
4941
4942 key.af = pd->af;
4943 key.proto = pd->proto;
4944 if (direction == PF_IN) {
4945 PF_ACPY(&key.ext.addr, pd->src, key.af);
4946 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4947 key.ext.port = 0;
4948 key.gwy.port = 0;
4949 } else {
4950 PF_ACPY(&key.lan.addr, pd->src, key.af);
4951 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4952 key.lan.port = 0;
4953 key.ext.port = 0;
4954 }
4955
4956 STATE_LOOKUP();
4957
4958 if (direction == (*state)->direction) {
4959 src = &(*state)->src;
4960 dst = &(*state)->dst;
4961 } else {
4962 src = &(*state)->dst;
4963 dst = &(*state)->src;
4964 }
4965
4966 /* update states */
4967 if (src->state < PFOTHERS_SINGLE)
4968 src->state = PFOTHERS_SINGLE;
4969 if (dst->state == PFOTHERS_SINGLE)
4970 dst->state = PFOTHERS_MULTIPLE;
4971
4972 /* update expire time */
4973#ifdef __FreeBSD__
4974 (*state)->expire = time_second;
4975#else
4976 (*state)->expire = time.tv_sec;
4977#endif
4978 if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE)
4979 (*state)->timeout = PFTM_OTHER_MULTIPLE;
4980 else
4981 (*state)->timeout = PFTM_OTHER_SINGLE;
4982
4983 /* translate source/destination address, if necessary */
4984 if (STATE_TRANSLATE(*state)) {
4985 if (direction == PF_OUT)
4986 switch (pd->af) {
4987#ifdef INET
4988 case AF_INET:
4989 pf_change_a(&pd->src->v4.s_addr,
4990 pd->ip_sum, (*state)->gwy.addr.v4.s_addr,
4991 0);
4992 break;
4993#endif /* INET */
4994#ifdef INET6
4995 case AF_INET6:
4996 PF_ACPY(pd->src, &(*state)->gwy.addr, pd->af);
4997 break;
4998#endif /* INET6 */
4999 }
5000 else
5001 switch (pd->af) {
5002#ifdef INET
5003 case AF_INET:
5004 pf_change_a(&pd->dst->v4.s_addr,
5005 pd->ip_sum, (*state)->lan.addr.v4.s_addr,
5006 0);
5007 break;
5008#endif /* INET */
5009#ifdef INET6
5010 case AF_INET6:
5011 PF_ACPY(pd->dst, &(*state)->lan.addr, pd->af);
5012 break;
5013#endif /* INET6 */
5014 }
5015 }
5016
5017 return (PF_PASS);
5018}
5019
5020/*
5021 * ipoff and off are measured from the start of the mbuf chain.
5022 * h must be at "ipoff" on the mbuf chain.
5023 */
5024void *
5025pf_pull_hdr(struct mbuf *m, int off, void *p, int len,
5026 u_short *actionp, u_short *reasonp, sa_family_t af)
5027{
5028 switch (af) {
5029#ifdef INET
5030 case AF_INET: {
5031 struct ip *h = mtod(m, struct ip *);
5032 u_int16_t fragoff = (ntohs(h->ip_off) & IP_OFFMASK) << 3;
5033
5034 if (fragoff) {
5035 if (fragoff >= len)
5036 ACTION_SET(actionp, PF_PASS);
5037 else {
5038 ACTION_SET(actionp, PF_DROP);
5039 REASON_SET(reasonp, PFRES_FRAG);
5040 }
5041 return (NULL);
5042 }
5043 if (m->m_pkthdr.len < off + len ||
5044 ntohs(h->ip_len) < off + len) {
5045 ACTION_SET(actionp, PF_DROP);
5046 REASON_SET(reasonp, PFRES_SHORT);
5047 return (NULL);
5048 }
5049 break;
5050 }
5051#endif /* INET */
5052#ifdef INET6
5053 case AF_INET6: {
5054 struct ip6_hdr *h = mtod(m, struct ip6_hdr *);
5055
5056 if (m->m_pkthdr.len < off + len ||
5057 (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) <
5058 (unsigned)(off + len)) {
5059 ACTION_SET(actionp, PF_DROP);
5060 REASON_SET(reasonp, PFRES_SHORT);
5061 return (NULL);
5062 }
5063 break;
5064 }
5065#endif /* INET6 */
5066 }
5067 m_copydata(m, off, len, p);
5068 return (p);
5069}
5070
5071int
5072pf_routable(struct pf_addr *addr, sa_family_t af)
5073{
5074 struct sockaddr_in *dst;
5075 struct route ro;
5076 int ret = 0;
5077
5078 bzero(&ro, sizeof(ro));
5079 dst = satosin(&ro.ro_dst);
5080 dst->sin_family = af;
5081 dst->sin_len = sizeof(*dst);
5082 dst->sin_addr = addr->v4;
5083#ifdef __FreeBSD__
5084#ifdef RTF_PRCLONING
5085 rtalloc_ign(&ro, (RTF_CLONING|RTF_PRCLONING));
5086#else /* !RTF_PRCLONING */
5087 rtalloc_ign(&ro, RTF_CLONING);
5088#endif
5089#else /* ! __FreeBSD__ */
5090 rtalloc_noclone(&ro, NO_CLONING);
5091#endif
5092
5093 if (ro.ro_rt != NULL) {
5094 ret = 1;
5095 RTFREE(ro.ro_rt);
5096 }
5097
5098 return (ret);
5099}
5100
5101#ifdef INET
5102
5103void
5104pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5105 struct pf_state *s)
5106{
5107 struct mbuf *m0, *m1;
5108 struct route iproute;
5109 struct route *ro = NULL; /* XXX: was uninitialized */
5110 struct sockaddr_in *dst;
5111 struct ip *ip;
5112 struct ifnet *ifp = NULL;
5113 struct m_tag *mtag;
5114 struct pf_addr naddr;
5115 struct pf_src_node *sn = NULL;
5116 int error = 0;
5117#ifdef __FreeBSD__
5118 int sw_csum;
5119#endif
5120
5121 if (m == NULL || *m == NULL || r == NULL ||
5122 (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5123 panic("pf_route: invalid parameters");
5124
5125 if (r->rt == PF_DUPTO) {
5126 m0 = *m;
5127 mtag = m_tag_find(m0, PACKET_TAG_PF_ROUTED, NULL);
5128 if (mtag == NULL) {
5129 mtag = m_tag_get(PACKET_TAG_PF_ROUTED, 0, M_NOWAIT);
5130 if (mtag == NULL)
5131 goto bad;
5132 m_tag_prepend(m0, mtag);
5133 }
5134#ifdef __FreeBSD__
5135 m0 = m_dup(*m, M_DONTWAIT);
5136#else
5137 m0 = m_copym2(*m, 0, M_COPYALL, M_NOWAIT);
5138#endif
5139 if (m0 == NULL)
5140 return;
5141 } else {
5142 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5143 return;
5144 m0 = *m;
5145 }
5146
5147 if (m0->m_len < sizeof(struct ip))
5148 panic("pf_route: m0->m_len < sizeof(struct ip)");
5149 ip = mtod(m0, struct ip *);
5150
5151 ro = &iproute;
5152 bzero((caddr_t)ro, sizeof(*ro));
5153 dst = satosin(&ro->ro_dst);
5154 dst->sin_family = AF_INET;
5155 dst->sin_len = sizeof(*dst);
5156 dst->sin_addr = ip->ip_dst;
5157
5158 if (r->rt == PF_FASTROUTE) {
5159 rtalloc(ro);
5160 if (ro->ro_rt == 0) {
5161 ipstat.ips_noroute++;
5162 goto bad;
5163 }
5164
5165 ifp = ro->ro_rt->rt_ifp;
5166 ro->ro_rt->rt_use++;
5167
5168 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
5169 dst = satosin(ro->ro_rt->rt_gateway);
5170 } else {
5171 if (TAILQ_EMPTY(&r->rpool.list))
5172 panic("pf_route: TAILQ_EMPTY(&r->rpool.list)");
5173 if (s == NULL) {
5174 pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src,
5175 &naddr, NULL, &sn);
5176 if (!PF_AZERO(&naddr, AF_INET))
5177 dst->sin_addr.s_addr = naddr.v4.s_addr;
5178 ifp = r->rpool.cur->kif ?
5179 r->rpool.cur->kif->pfik_ifp : NULL;
5180 } else {
5181 if (!PF_AZERO(&s->rt_addr, AF_INET))
5182 dst->sin_addr.s_addr =
5183 s->rt_addr.v4.s_addr;
5184 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5185 }
5186 }
5187
5188 if (ifp == NULL)
5189 goto bad;
5190
5191 mtag = m_tag_find(m0, PACKET_TAG_PF_ROUTED, NULL);
5192 if (mtag == NULL) {
5193 struct m_tag *mtag;
5194
5195 mtag = m_tag_get(PACKET_TAG_PF_ROUTED, 0, M_NOWAIT);
5196 if (mtag == NULL)
5197 goto bad;
5198 m_tag_prepend(m0, mtag);
5199 }
5200
5201 if (oifp != ifp && mtag == NULL) {
5202#ifdef __FreeBSD__
5203 PF_UNLOCK();
5204 if (pf_test(PF_OUT, ifp, &m0) != PF_PASS) {
5205 PF_LOCK();
5206 goto bad;
5207 } else if (m0 == NULL) {
5208 PF_LOCK();
5209 goto done;
5210 }
5211 PF_LOCK();
5212#else
5213 if (pf_test(PF_OUT, ifp, &m0) != PF_PASS)
5214 goto bad;
5215 else if (m0 == NULL)
5216 goto done;
5217#endif
5218 if (m0->m_len < sizeof(struct ip))
5219 panic("pf_route: m0->m_len < sizeof(struct ip)");
5220 ip = mtod(m0, struct ip *);
5221 }
5222
5223#ifdef __FreeBSD__
5224 /* Copied from FreeBSD 5.1-CURRENT ip_output. */
5225 m0->m_pkthdr.csum_flags |= CSUM_IP;
5226 sw_csum = m0->m_pkthdr.csum_flags & ~ifp->if_hwassist;
5227 if (sw_csum & CSUM_DELAY_DATA) {
5228 /*
5229 * XXX: in_delayed_cksum assumes HBO for ip->ip_len (at least)
5230 */
5231 NTOHS(ip->ip_len);
5232 NTOHS(ip->ip_off); /* XXX: needed? */
5233 in_delayed_cksum(m0);
5234 HTONS(ip->ip_len);
5235 HTONS(ip->ip_off);
5236 sw_csum &= ~CSUM_DELAY_DATA;
5237 }
5238 m0->m_pkthdr.csum_flags &= ifp->if_hwassist;
5239
5240 if (ntohs(ip->ip_len) <= ifp->if_mtu ||
5241 (ifp->if_hwassist & CSUM_FRAGMENT &&
5242 ((ip->ip_off & htons(IP_DF)) == 0))) {
5243 /*
5244 * ip->ip_len = htons(ip->ip_len);
5245 * ip->ip_off = htons(ip->ip_off);
5246 */
5247 ip->ip_sum = 0;
5248 if (sw_csum & CSUM_DELAY_IP) {
5249 /* From KAME */
5250 if (ip->ip_v == IPVERSION &&
5251 (ip->ip_hl << 2) == sizeof(*ip)) {
5252 ip->ip_sum = in_cksum_hdr(ip);
5253 } else {
5254 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
5255 }
5256 }
5257 PF_UNLOCK();
5258 error = (*ifp->if_output)(ifp, m0, sintosa(dst), ro->ro_rt);
5259 PF_LOCK();
5260 goto done;
5261 }
5262
5263#else
5264 /* Copied from ip_output. */
5265#ifdef IPSEC
5266 /*
5267 * If deferred crypto processing is needed, check that the
5268 * interface supports it.
5269 */
5270 if ((mtag = m_tag_find(m0, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL))
5271 != NULL && (ifp->if_capabilities & IFCAP_IPSEC) == 0) {
5272 /* Notify IPsec to do its own crypto. */
5273 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1));
5274 goto bad;
5275 }
5276#endif /* IPSEC */
5277
5278 /* Catch routing changes wrt. hardware checksumming for TCP or UDP. */
5279 if (m0->m_pkthdr.csum & M_TCPV4_CSUM_OUT) {
5280 if (!(ifp->if_capabilities & IFCAP_CSUM_TCPv4) ||
5281 ifp->if_bridge != NULL) {
5282 in_delayed_cksum(m0);
5283 m0->m_pkthdr.csum &= ~M_TCPV4_CSUM_OUT; /* Clear */
5284 }
5285 } else if (m0->m_pkthdr.csum & M_UDPV4_CSUM_OUT) {
5286 if (!(ifp->if_capabilities & IFCAP_CSUM_UDPv4) ||
5287 ifp->if_bridge != NULL) {
5288 in_delayed_cksum(m0);
5289 m0->m_pkthdr.csum &= ~M_UDPV4_CSUM_OUT; /* Clear */
5290 }
5291 }
5292
5293 if (ntohs(ip->ip_len) <= ifp->if_mtu) {
5294 if ((ifp->if_capabilities & IFCAP_CSUM_IPv4) &&
5295 ifp->if_bridge == NULL) {
5296 m0->m_pkthdr.csum |= M_IPV4_CSUM_OUT;
5297 ipstat.ips_outhwcsum++;
5298 } else {
5299 ip->ip_sum = 0;
5300 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
5301 }
5302 /* Update relevant hardware checksum stats for TCP/UDP */
5303 if (m0->m_pkthdr.csum & M_TCPV4_CSUM_OUT)
5304 tcpstat.tcps_outhwcsum++;
5305 else if (m0->m_pkthdr.csum & M_UDPV4_CSUM_OUT)
5306 udpstat.udps_outhwcsum++;
5307 error = (*ifp->if_output)(ifp, m0, sintosa(dst), NULL);
5308 goto done;
5309 }
5310#endif
5311 /*
5312 * Too large for interface; fragment if possible.
5313 * Must be able to put at least 8 bytes per fragment.
5314 */
5315 if (ip->ip_off & htons(IP_DF)) {
5316 ipstat.ips_cantfrag++;
5317 if (r->rt != PF_DUPTO) {
5318#ifdef __FreeBSD__
5319 /* icmp_error() expects host byte ordering */
5320 NTOHS(ip->ip_len);
5321 NTOHS(ip->ip_off);
5322 PF_UNLOCK();
5323#endif
5324 icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0,
5325 ifp);
5326#ifdef __FreeBSD__
5327 PF_LOCK();
5328#endif
5329 goto done;
5330 } else
5331 goto bad;
5332 }
5333
5334 m1 = m0;
5335#ifdef __FreeBSD__
5336 /*
5337 * XXX: is cheaper + less error prone than own function
5338 */
5339 NTOHS(ip->ip_len);
5340 NTOHS(ip->ip_off);
5341 error = ip_fragment(ip, &m0, ifp->if_mtu, ifp->if_hwassist, sw_csum);
5342#else
5343 error = ip_fragment(m0, ifp, ifp->if_mtu);
5344#endif
5345 if (error) {
5346#ifndef __FreeBSD__ /* ip_fragment does not do m_freem() on FreeBSD */
5347 m0 = NULL;
5348#endif
5349 goto bad;
5350 }
5351
5352 for (m0 = m1; m0; m0 = m1) {
5353 m1 = m0->m_nextpkt;
5354 m0->m_nextpkt = 0;
5355#ifdef __FreeBSD__
5356 if (error == 0) {
5357 PF_UNLOCK();
5358 error = (*ifp->if_output)(ifp, m0, sintosa(dst),
5359 NULL);
5360 PF_LOCK();
5361 } else
5362#else
5363 if (error == 0)
5364 error = (*ifp->if_output)(ifp, m0, sintosa(dst),
5365 NULL);
5366 else
5367#endif
5368 m_freem(m0);
5369 }
5370
5371 if (error == 0)
5372 ipstat.ips_fragmented++;
5373
5374done:
5375 if (r->rt != PF_DUPTO)
5376 *m = NULL;
5377 if (ro == &iproute && ro->ro_rt)
5378 RTFREE(ro->ro_rt);
5379 return;
5380
5381bad:
5382 m_freem(m0);
5383 goto done;
5384}
5385#endif /* INET */
5386
5387#ifdef INET6
5388void
5389pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5390 struct pf_state *s)
5391{
5392 struct mbuf *m0;
5393 struct m_tag *mtag;
5394 struct route_in6 ip6route;
5395 struct route_in6 *ro;
5396 struct sockaddr_in6 *dst;
5397 struct ip6_hdr *ip6;
5398 struct ifnet *ifp = NULL;
5399 struct pf_addr naddr;
5400 struct pf_src_node *sn = NULL;
5401 int error = 0;
5402
5403 if (m == NULL || *m == NULL || r == NULL ||
5404 (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5405 panic("pf_route6: invalid parameters");
5406
5407 if (r->rt == PF_DUPTO) {
5408 m0 = *m;
5409 mtag = m_tag_find(m0, PACKET_TAG_PF_ROUTED, NULL);
5410 if (mtag == NULL) {
5411 mtag = m_tag_get(PACKET_TAG_PF_ROUTED, 0, M_NOWAIT);
5412 if (mtag == NULL)
5413 goto bad;
5414 m_tag_prepend(m0, mtag);
5415 }
5416#ifdef __FreeBSD__
5417 m0 = m_dup(*m, M_DONTWAIT);
5418#else
5419 m0 = m_copym2(*m, 0, M_COPYALL, M_NOWAIT);
5420#endif
5421 if (m0 == NULL)
5422 return;
5423 } else {
5424 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5425 return;
5426 m0 = *m;
5427 }
5428
5429 if (m0->m_len < sizeof(struct ip6_hdr))
5430 panic("pf_route6: m0->m_len < sizeof(struct ip6_hdr)");
5431 ip6 = mtod(m0, struct ip6_hdr *);
5432
5433 ro = &ip6route;
5434 bzero((caddr_t)ro, sizeof(*ro));
5435 dst = (struct sockaddr_in6 *)&ro->ro_dst;
5436 dst->sin6_family = AF_INET6;
5437 dst->sin6_len = sizeof(*dst);
5438 dst->sin6_addr = ip6->ip6_dst;
5439
5440 /* Cheat. */
5441 if (r->rt == PF_FASTROUTE) {
5442 mtag = m_tag_get(PACKET_TAG_PF_GENERATED, 0, M_NOWAIT);
5443 if (mtag == NULL)
5444 goto bad;
5445 m_tag_prepend(m0, mtag);
5446#ifdef __FreeBSD__
5447 PF_UNLOCK();
5448 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
5449 PF_LOCK();
5450#else
5451 ip6_output(m0, NULL, NULL, 0, NULL, NULL);
5452#endif
5453 return;
5454 }
5455
5456 if (TAILQ_EMPTY(&r->rpool.list))
5457 panic("pf_route6: TAILQ_EMPTY(&r->rpool.list)");
5458 if (s == NULL) {
5459 pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src,
5460 &naddr, NULL, &sn);
5461 if (!PF_AZERO(&naddr, AF_INET6))
5462 PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5463 &naddr, AF_INET6);
5464 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL;
5465 } else {
5466 if (!PF_AZERO(&s->rt_addr, AF_INET6))
5467 PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5468 &s->rt_addr, AF_INET6);
5469 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5470 }
5471
5472 if (ifp == NULL)
5473 goto bad;
5474
5475 if (oifp != ifp) {
5476 mtag = m_tag_find(m0, PACKET_TAG_PF_ROUTED, NULL);
5477 if (mtag == NULL) {
5478 mtag = m_tag_get(PACKET_TAG_PF_ROUTED, 0, M_NOWAIT);
5479 if (mtag == NULL)
5480 goto bad;
5481 m_tag_prepend(m0, mtag);
5482#ifdef __FreeBSD__
5483 PF_UNLOCK();
5484 if (pf_test6(PF_OUT, ifp, &m0) != PF_PASS) {
5485 PF_LOCK();
5486 goto bad;
5487 } else if (m0 == NULL) {
5488 PF_LOCK();
5489 goto done;
5490 }
5491 PF_LOCK();
5492#else
5493 if (pf_test6(PF_OUT, ifp, &m0) != PF_PASS)
5494 goto bad;
5495 else if (m0 == NULL)
5496 goto done;
5497#endif
5498 }
5499 }
5500
5501 /*
5502 * If the packet is too large for the outgoing interface,
5503 * send back an icmp6 error.
5504 */
5505 if (IN6_IS_ADDR_LINKLOCAL(&dst->sin6_addr))
5506 dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
5507 if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) {
5508#ifdef __FreeBSD__
5509 PF_UNLOCK();
5510#endif
5511 error = nd6_output(ifp, ifp, m0, dst, NULL);
5512#ifdef __FreeBSD__
5513 PF_LOCK();
5514#endif
5515 } else {
5516 in6_ifstat_inc(ifp, ifs6_in_toobig);
5517#ifdef __FreeBSD__
5518 if (r->rt != PF_DUPTO) {
5519 PF_UNLOCK();
5520 icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
5521 PF_LOCK();
5522 } else
5523#else
5524 if (r->rt != PF_DUPTO)
5525 icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
5526 else
5527#endif
5528 goto bad;
5529 }
5530
5531done:
5532 if (r->rt != PF_DUPTO)
5533 *m = NULL;
5534 return;
5535
5536bad:
5537 m_freem(m0);
5538 goto done;
5539}
5540#endif /* INET6 */
5541
5542
5543#ifdef __FreeBSD__
5544/*
5545 * XXX
5546 * FreeBSD supports cksum offload for the following drivers.
5547 * em(4), gx(4), lge(4), nge(4), ti(4), xl(4)
5548 * If we can make full use of it we would outperform ipfw/ipfilter in
5549 * very heavy traffic.
5550 * I have not tested 'cause I don't have NICs that supports cksum offload.
5551 * (There might be problems. Typical phenomena would be
5552 * 1. No route message for UDP packet.
5553 * 2. No connection acceptance from external hosts regardless of rule set.)
5554 */
5555int
5556pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p, sa_family_t af)
5557{
5558 u_int16_t sum = 0;
5559 int hw_assist = 0;
5560 struct ip *ip;
5561
5562 if (off < sizeof(struct ip) || len < sizeof(struct udphdr))
5563 return (1);
5564 if (m->m_pkthdr.len < off + len)
5565 return (1);
5566
5567 switch (p) {
5568 case IPPROTO_TCP:
5569 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
5570 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) {
5571 sum = m->m_pkthdr.csum_data;
5572 } else {
5573 ip = mtod(m, struct ip *);
5574 sum = in_pseudo(ip->ip_src.s_addr,
5575 ip->ip_dst.s_addr,
5576 htonl(m->m_pkthdr.csum_data +
5577 IPPROTO_TCP) + ip->ip_len);
5578 }
5579 sum ^= 0xffff;
5580 ++hw_assist;
5581 }
5582 break;
5583 case IPPROTO_UDP:
5584 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
5585 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) {
5586 sum = m->m_pkthdr.csum_data;
5587 } else {
5588 ip = mtod(m, struct ip *);
5589 sum = in_pseudo(ip->ip_src.s_addr,
5590 ip->ip_dst.s_addr, htonl((u_short)len +
5591 m->m_pkthdr.csum_data + IPPROTO_UDP));
5592 }
5593 sum ^= 0xffff;
5594 ++hw_assist;
5595 }
5596 break;
5597 case IPPROTO_ICMP:
5598#ifdef INET6
5599 case IPPROTO_ICMPV6:
5600#endif /* INET6 */
5601 break;
5602 default:
5603 return (1);
5604 }
5605
5606 if (!hw_assist) {
5607 switch (af) {
5608 case AF_INET:
5609 if (p == IPPROTO_ICMP) {
5610 if (m->m_len < off)
5611 return (1);
5612 m->m_data += off;
5613 m->m_len -= off;
5614 sum = in_cksum(m, len);
5615 m->m_data -= off;
5616 m->m_len += off;
5617 } else {
5618 if (m->m_len < sizeof(struct ip))
5619 return (1);
5620 sum = in4_cksum(m, p, off, len);
5621 if (sum == 0) {
5622 m->m_pkthdr.csum_flags |=
5623 (CSUM_DATA_VALID |
5624 CSUM_PSEUDO_HDR);
5625 m->m_pkthdr.csum_data = 0xffff;
5626 }
5627 }
5628 break;
5629#ifdef INET6
5630 case AF_INET6:
5631 if (m->m_len < sizeof(struct ip6_hdr))
5632 return (1);
5633 sum = in6_cksum(m, p, off, len);
5634 /*
5635 * XXX
5636 * IPv6 H/W cksum off-load not supported yet!
5637 *
5638 * if (sum == 0) {
5639 * m->m_pkthdr.csum_flags |=
5640 * (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
5641 * m->m_pkthdr.csum_data = 0xffff;
5642 *}
5643 */
5644 break;
5645#endif /* INET6 */
5646 default:
5647 return (1);
5648 }
5649 }
5650 if (sum) {
5651 switch (p) {
5652 case IPPROTO_TCP:
5653 tcpstat.tcps_rcvbadsum++;
5654 break;
5655 case IPPROTO_UDP:
5656 udpstat.udps_badsum++;
5657 break;
5658 case IPPROTO_ICMP:
5659 icmpstat.icps_checksum++;
5660 break;
5661#ifdef INET6
5662 case IPPROTO_ICMPV6:
5663 icmp6stat.icp6s_checksum++;
5664 break;
5665#endif /* INET6 */
5666 }
5667 return (1);
5668 }
5669 return (0);
5670}
5671#else
5672/*
5673 * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag
5674 * off is the offset where the protocol header starts
5675 * len is the total length of protocol header plus payload
5676 * returns 0 when the checksum is valid, otherwise returns 1.
5677 */
5678int
5679pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p,
5680 sa_family_t af)
5681{
5682 u_int16_t flag_ok, flag_bad;
5683 u_int16_t sum;
5684
5685 switch (p) {
5686 case IPPROTO_TCP:
5687 flag_ok = M_TCP_CSUM_IN_OK;
5688 flag_bad = M_TCP_CSUM_IN_BAD;
5689 break;
5690 case IPPROTO_UDP:
5691 flag_ok = M_UDP_CSUM_IN_OK;
5692 flag_bad = M_UDP_CSUM_IN_BAD;
5693 break;
5694 case IPPROTO_ICMP:
5695#ifdef INET6
5696 case IPPROTO_ICMPV6:
5697#endif /* INET6 */
5698 flag_ok = flag_bad = 0;
5699 break;
5700 default:
5701 return (1);
5702 }
5703 if (m->m_pkthdr.csum & flag_ok)
5704 return (0);
5705 if (m->m_pkthdr.csum & flag_bad)
5706 return (1);
5707 if (off < sizeof(struct ip) || len < sizeof(struct udphdr))
5708 return (1);
5709 if (m->m_pkthdr.len < off + len)
5710 return (1);
5711 switch (af) {
5712 case AF_INET:
5713 if (p == IPPROTO_ICMP) {
5714 if (m->m_len < off)
5715 return (1);
5716 m->m_data += off;
5717 m->m_len -= off;
5718 sum = in_cksum(m, len);
5719 m->m_data -= off;
5720 m->m_len += off;
5721 } else {
5722 if (m->m_len < sizeof(struct ip))
5723 return (1);
5724 sum = in4_cksum(m, p, off, len);
5725 }
5726 break;
5727#ifdef INET6
5728 case AF_INET6:
5729 if (m->m_len < sizeof(struct ip6_hdr))
5730 return (1);
5731 sum = in6_cksum(m, p, off, len);
5732 break;
5733#endif /* INET6 */
5734 default:
5735 return (1);
5736 }
5737 if (sum) {
5738 m->m_pkthdr.csum |= flag_bad;
5739 switch (p) {
5740 case IPPROTO_TCP:
5741 tcpstat.tcps_rcvbadsum++;
5742 break;
5743 case IPPROTO_UDP:
5744 udpstat.udps_badsum++;
5745 break;
5746 case IPPROTO_ICMP:
5747 icmpstat.icps_checksum++;
5748 break;
5749#ifdef INET6
5750 case IPPROTO_ICMPV6:
5751 icmp6stat.icp6s_checksum++;
5752 break;
5753#endif /* INET6 */
5754 }
5755 return (1);
5756 }
5757 m->m_pkthdr.csum |= flag_ok;
5758 return (0);
5759}
5760#endif
5761
5762static int
5763pf_add_mbuf_tag(struct mbuf *m, u_int tag)
5764{
5765 struct m_tag *mtag;
5766
5767 if (m_tag_find(m, tag, NULL) != NULL)
5768 return (0);
5769 mtag = m_tag_get(tag, 0, M_NOWAIT);
5770 if (mtag == NULL)
5771 return (1);
5772 m_tag_prepend(m, mtag);
5773 return (0);
5774}
5775
5776#ifdef INET
5777int
5778pf_test(int dir, struct ifnet *ifp, struct mbuf **m0)
5779{
5780 struct pfi_kif *kif;
5781 u_short action, reason = 0, log = 0;
5782 struct mbuf *m = *m0;
5783 struct ip *h = NULL; /* make the compiler happy */
5784 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr;
5785 struct pf_state *s = NULL;
5786 struct pf_ruleset *ruleset = NULL;
5787 struct pf_pdesc pd;
5788 int off, dirndx, pqid = 0;
5789
5790#ifdef __FreeBSD__
5791 PF_LOCK();
5792#endif
5793 if (!pf_status.running ||
5794 (m_tag_find(m, PACKET_TAG_PF_GENERATED, NULL) != NULL)) {
5795#ifdef __FreeBSD__
5796 PF_UNLOCK();
5797#endif
5798 return (PF_PASS);
5799 }
5800
5801 kif = pfi_index2kif[ifp->if_index];
5802 if (kif == NULL) {
5803#ifdef __FreeBSD__
5804 PF_UNLOCK();
5805#endif
5806 return (PF_DROP);
5807 }
5808
5809#ifdef __FreeBSD__
5810 M_ASSERTPKTHDR(m);
5811#else
5812#ifdef DIAGNOSTIC
5813 if ((m->m_flags & M_PKTHDR) == 0)
5814 panic("non-M_PKTHDR is passed to pf_test");
5815#endif
5816#endif
5817
5818 memset(&pd, 0, sizeof(pd));
5819 if (m->m_pkthdr.len < (int)sizeof(*h)) {
5820 action = PF_DROP;
5821 REASON_SET(&reason, PFRES_SHORT);
5822 log = 1;
5823 goto done;
5824 }
5825
5826 /* We do IP header normalization and packet reassembly here */
5827 if (pf_normalize_ip(m0, dir, kif, &reason) != PF_PASS) {
5828 action = PF_DROP;
5829 goto done;
5830 }
5831 m = *m0;
5832 h = mtod(m, struct ip *);
5833
5834 off = h->ip_hl << 2;
5835 if (off < (int)sizeof(*h)) {
5836 action = PF_DROP;
5837 REASON_SET(&reason, PFRES_SHORT);
5838 log = 1;
5839 goto done;
5840 }
5841
5842 pd.src = (struct pf_addr *)&h->ip_src;
5843 pd.dst = (struct pf_addr *)&h->ip_dst;
5844 PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET);
5845 pd.ip_sum = &h->ip_sum;
5846 pd.proto = h->ip_p;
5847 pd.af = AF_INET;
5848 pd.tos = h->ip_tos;
5849 pd.tot_len = ntohs(h->ip_len);
5850
5851 /* handle fragments that didn't get reassembled by normalization */
5852 if (h->ip_off & htons(IP_MF | IP_OFFMASK)) {
5853 action = pf_test_fragment(&r, dir, kif, m, h,
5854 &pd, &a, &ruleset);
5855 goto done;
5856 }
5857
5858 switch (h->ip_p) {
5859
5860 case IPPROTO_TCP: {
5861 struct tcphdr th;
5862
5863 pd.hdr.tcp = &th;
5864 if (!pf_pull_hdr(m, off, &th, sizeof(th),
5865 &action, &reason, AF_INET)) {
5866 log = action != PF_PASS;
5867 goto done;
5868 }
5869 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5870 ntohs(h->ip_len) - off, IPPROTO_TCP, AF_INET)) {
5871 action = PF_DROP;
5872 goto done;
5873 }
5874 pd.p_len = pd.tot_len - off - (th.th_off << 2);
5875 if ((th.th_flags & TH_ACK) && pd.p_len == 0)
5876 pqid = 1;
5877 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
5878 if (action == PF_DROP)
5879 goto done;
5880 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
5881 &reason);
5882 if (action == PF_PASS) {
5883#if NPFSYNC
5884 pfsync_update_state(s);
5885#endif
5886 r = s->rule.ptr;
5887 a = s->anchor.ptr;
5888 log = s->log;
5889 } else if (s == NULL)
5890 action = pf_test_tcp(&r, &s, dir, kif,
5891 m, off, h, &pd, &a, &ruleset);
5892 break;
5893 }
5894
5895 case IPPROTO_UDP: {
5896 struct udphdr uh;
5897
5898 pd.hdr.udp = &uh;
5899 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
5900 &action, &reason, AF_INET)) {
5901 log = action != PF_PASS;
5902 goto done;
5903 }
5904 if (dir == PF_IN && uh.uh_sum && pf_check_proto_cksum(m,
5905 off, ntohs(h->ip_len) - off, IPPROTO_UDP, AF_INET)) {
5906 action = PF_DROP;
5907 goto done;
5908 }
5909 if (uh.uh_dport == 0 ||
5910 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
5911 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
5912 action = PF_DROP;
5913 goto done;
5914 }
5915 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
5916 if (action == PF_PASS) {
5917#if NPFSYNC
5918 pfsync_update_state(s);
5919#endif
5920 r = s->rule.ptr;
5921 a = s->anchor.ptr;
5922 log = s->log;
5923 } else if (s == NULL)
5924 action = pf_test_udp(&r, &s, dir, kif,
5925 m, off, h, &pd, &a, &ruleset);
5926 break;
5927 }
5928
5929 case IPPROTO_ICMP: {
5930 struct icmp ih;
5931
5932 pd.hdr.icmp = &ih;
5933 if (!pf_pull_hdr(m, off, &ih, ICMP_MINLEN,
5934 &action, &reason, AF_INET)) {
5935 log = action != PF_PASS;
5936 goto done;
5937 }
5938 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5939 ntohs(h->ip_len) - off, IPPROTO_ICMP, AF_INET)) {
5940 action = PF_DROP;
5941 goto done;
5942 }
5943 action = pf_test_state_icmp(&s, dir, kif, m, off, h, &pd);
5944 if (action == PF_PASS) {
5945#if NPFSYNC
5946 pfsync_update_state(s);
5947#endif
5948 r = s->rule.ptr;
5949 a = s->anchor.ptr;
5950 log = s->log;
5951 } else if (s == NULL)
5952 action = pf_test_icmp(&r, &s, dir, kif,
5953 m, off, h, &pd, &a, &ruleset);
5954 break;
5955 }
5956
5957 default:
5958 action = pf_test_state_other(&s, dir, kif, &pd);
5959 if (action == PF_PASS) {
5960#if NPFSYNC
5961 pfsync_update_state(s);
5962#endif
5963 r = s->rule.ptr;
5964 a = s->anchor.ptr;
5965 log = s->log;
5966 } else if (s == NULL)
5967 action = pf_test_other(&r, &s, dir, kif, m, off, h,
5968 &pd, &a, &ruleset);
5969 break;
5970 }
5971
5972done:
5973 if (action == PF_PASS && h->ip_hl > 5 &&
5974 !((s && s->allow_opts) || r->allow_opts)) {
5975 action = PF_DROP;
5976 REASON_SET(&reason, PFRES_SHORT);
5977 log = 1;
5978 DPFPRINTF(PF_DEBUG_MISC,
5979 ("pf: dropping packet with ip options\n"));
5980 }
5981
5982#ifdef ALTQ
5983 if (action == PF_PASS && r->qid) {
5984 struct m_tag *mtag;
5985 struct altq_tag *atag;
5986
5987 mtag = m_tag_get(PACKET_TAG_PF_QID, sizeof(*atag), M_NOWAIT);
5988 if (mtag != NULL) {
5989 atag = (struct altq_tag *)(mtag + 1);
5990 if (pqid || pd.tos == IPTOS_LOWDELAY)
5991 atag->qid = r->pqid;
5992 else
5993 atag->qid = r->qid;
5994 /* add hints for ecn */
5995 atag->af = AF_INET;
5996 atag->hdr = h;
5997 m_tag_prepend(m, mtag);
5998 }
5999 }
6000#endif
6001
6002 /*
6003 * connections redirected to loopback should not match sockets
6004 * bound specifically to loopback due to security implications,
6005 * see tcp_input() and in_pcblookup_listen().
6006 */
6007 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
6008 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
6009 (s->nat_rule.ptr->action == PF_RDR ||
6010 s->nat_rule.ptr->action == PF_BINAT) &&
6011 (ntohl(pd.dst->v4.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET &&
6012 pf_add_mbuf_tag(m, PACKET_TAG_PF_TRANSLATE_LOCALHOST)) {
6013 action = PF_DROP;
6014 REASON_SET(&reason, PFRES_MEMORY);
6015 }
6016
6017 if (log)
6018 PFLOG_PACKET(kif, h, m, AF_INET, dir, reason, r, a, ruleset);
6019
6020 kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
6021 kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++;
6022
6023 if (action == PF_PASS || r->action == PF_DROP) {
6024 r->packets++;
6025 r->bytes += pd.tot_len;
6026 if (a != NULL) {
6027 a->packets++;
6028 a->bytes += pd.tot_len;
6029 }
6030 if (s != NULL) {
6031 dirndx = (dir == s->direction) ? 0 : 1;
6032 s->packets[dirndx]++;
6033 s->bytes[dirndx] += pd.tot_len;
6034 if (s->nat_rule.ptr != NULL) {
6035 s->nat_rule.ptr->packets++;
6036 s->nat_rule.ptr->bytes += pd.tot_len;
6037 }
6038 if (s->src_node != NULL) {
6039 s->src_node->packets++;
6040 s->src_node->bytes += pd.tot_len;
6041 }
6042 if (s->nat_src_node != NULL) {
6043 s->nat_src_node->packets++;
6044 s->nat_src_node->bytes += pd.tot_len;
6045 }
6046 }
6047 tr = r;
6048 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
6049 if (nr != NULL) {
6050 struct pf_addr *x;
6051 /*
6052 * XXX: we need to make sure that the addresses
6053 * passed to pfr_update_stats() are the same than
6054 * the addresses used during matching (pfr_match)
6055 */
6056 if (r == &pf_default_rule) {
6057 tr = nr;
6058 x = (s == NULL || s->direction == dir) ?
6059 &pd.baddr : &pd.naddr;
6060 } else
6061 x = (s == NULL || s->direction == dir) ?
6062 &pd.naddr : &pd.baddr;
6063 if (x == &pd.baddr || s == NULL) {
6064 /* we need to change the address */
6065 if (dir == PF_OUT)
6066 pd.src = x;
6067 else
6068 pd.dst = x;
6069 }
6070 }
6071 if (tr->src.addr.type == PF_ADDR_TABLE)
6072 pfr_update_stats(tr->src.addr.p.tbl, (s == NULL ||
6073 s->direction == dir) ? pd.src : pd.dst, pd.af,
6074 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
6075 tr->src.not);
6076 if (tr->dst.addr.type == PF_ADDR_TABLE)
6077 pfr_update_stats(tr->dst.addr.p.tbl, (s == NULL ||
6078 s->direction == dir) ? pd.dst : pd.src, pd.af,
6079 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
6080 tr->dst.not);
6081 }
6082
6083
6084 if (action == PF_SYNPROXY_DROP) {
6085 m_freem(*m0);
6086 *m0 = NULL;
6087 action = PF_PASS;
6088 } else if (r->rt)
6089 /* pf_route can free the mbuf causing *m0 to become NULL */
6090 pf_route(m0, r, dir, ifp, s);
6091
6092#ifdef __FreeBSD__
6093 PF_UNLOCK();
6094#endif
6095
6096 return (action);
6097}
6098#endif /* INET */
6099
6100#ifdef INET6
6101int
6102pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0)
6103{
6104 struct pfi_kif *kif;
6105 u_short action, reason = 0, log = 0;
6106 struct mbuf *m = *m0;
6107 struct ip6_hdr *h = NULL; /* make the compiler happy */
6108 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr;
6109 struct pf_state *s = NULL;
6110 struct pf_ruleset *ruleset = NULL;
6111 struct pf_pdesc pd;
6112 int off, terminal = 0, dirndx;
6113
6114#ifdef __FreeBSD__
6115 PF_LOCK();
6116#endif
6117
6118 if (!pf_status.running ||
6119 (m_tag_find(m, PACKET_TAG_PF_GENERATED, NULL) != NULL)) {
6120#ifdef __FreeBSD__
6121 PF_UNLOCK();
6122#endif
6123 return (PF_PASS);
6124 }
6125
6126 kif = pfi_index2kif[ifp->if_index];
6127 if (kif == NULL) {
6128#ifdef __FreeBSD__
6129 PF_UNLOCK();
6130#endif
6131 return (PF_DROP);
6132 }
6133
6134#ifdef __FreeBSD__
6135 M_ASSERTPKTHDR(m);
6136#else
6137#ifdef DIAGNOSTIC
6138 if ((m->m_flags & M_PKTHDR) == 0)
6139 panic("non-M_PKTHDR is passed to pf_test");
6140#endif
6141#endif
6142
6143 memset(&pd, 0, sizeof(pd));
6144 if (m->m_pkthdr.len < (int)sizeof(*h)) {
6145 action = PF_DROP;
6146 REASON_SET(&reason, PFRES_SHORT);
6147 log = 1;
6148 goto done;
6149 }
6150
6151 /* We do IP header normalization and packet reassembly here */
6152 if (pf_normalize_ip6(m0, dir, kif, &reason) != PF_PASS) {
6153 action = PF_DROP;
6154 goto done;
6155 }
6156 m = *m0;
6157 h = mtod(m, struct ip6_hdr *);
6158
6159 pd.src = (struct pf_addr *)&h->ip6_src;
6160 pd.dst = (struct pf_addr *)&h->ip6_dst;
6161 PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET6);
6162 pd.ip_sum = NULL;
6163 pd.af = AF_INET6;
6164 pd.tos = 0;
6165 pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr);
6166
6167 off = ((caddr_t)h - m->m_data) + sizeof(struct ip6_hdr);
6168 pd.proto = h->ip6_nxt;
6169 do {
6170 switch (pd.proto) {
6171 case IPPROTO_FRAGMENT:
6172 action = pf_test_fragment(&r, dir, kif, m, h,
6173 &pd, &a, &ruleset);
6174 if (action == PF_DROP)
6175 REASON_SET(&reason, PFRES_FRAG);
6176 goto done;
6177 case IPPROTO_AH:
6178 case IPPROTO_HOPOPTS:
6179 case IPPROTO_ROUTING:
6180 case IPPROTO_DSTOPTS: {
6181 /* get next header and header length */
6182 struct ip6_ext opt6;
6183
6184 if (!pf_pull_hdr(m, off, &opt6, sizeof(opt6),
6185 NULL, NULL, pd.af)) {
6186 DPFPRINTF(PF_DEBUG_MISC,
6187 ("pf: IPv6 short opt\n"));
6188 action = PF_DROP;
6189 REASON_SET(&reason, PFRES_SHORT);
6190 log = 1;
6191 goto done;
6192 }
6193 if (pd.proto == IPPROTO_AH)
6194 off += (opt6.ip6e_len + 2) * 4;
6195 else
6196 off += (opt6.ip6e_len + 1) * 8;
6197 pd.proto = opt6.ip6e_nxt;
6198 /* goto the next header */
6199 break;
6200 }
6201 default:
6202 terminal++;
6203 break;
6204 }
6205 } while (!terminal);
6206
6207 switch (pd.proto) {
6208
6209 case IPPROTO_TCP: {
6210 struct tcphdr th;
6211
6212 pd.hdr.tcp = &th;
6213 if (!pf_pull_hdr(m, off, &th, sizeof(th),
6214 &action, &reason, AF_INET6)) {
6215 log = action != PF_PASS;
6216 goto done;
6217 }
6218 if (dir == PF_IN && pf_check_proto_cksum(m, off,
6219 ntohs(h->ip6_plen), IPPROTO_TCP, AF_INET6)) {
6220 action = PF_DROP;
6221 goto done;
6222 }
6223 pd.p_len = pd.tot_len - off - (th.th_off << 2);
6224 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
6225 if (action == PF_DROP)
6226 goto done;
6227 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
6228 &reason);
6229 if (action == PF_PASS) {
6230#if NPFSYNC
6231 pfsync_update_state(s);
6232#endif
6233 r = s->rule.ptr;
6234 a = s->anchor.ptr;
6235 log = s->log;
6236 } else if (s == NULL)
6237 action = pf_test_tcp(&r, &s, dir, kif,
6238 m, off, h, &pd, &a, &ruleset);
6239 break;
6240 }
6241
6242 case IPPROTO_UDP: {
6243 struct udphdr uh;
6244
6245 pd.hdr.udp = &uh;
6246 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
6247 &action, &reason, AF_INET6)) {
6248 log = action != PF_PASS;
6249 goto done;
6250 }
6251 if (dir == PF_IN && uh.uh_sum && pf_check_proto_cksum(m,
6252 off, ntohs(h->ip6_plen), IPPROTO_UDP, AF_INET6)) {
6253 action = PF_DROP;
6254 goto done;
6255 }
6256 if (uh.uh_dport == 0 ||
6257 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
6258 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
6259 action = PF_DROP;
6260 goto done;
6261 }
6262 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
6263 if (action == PF_PASS) {
6264#if NPFSYNC
6265 pfsync_update_state(s);
6266#endif
6267 r = s->rule.ptr;
6268 a = s->anchor.ptr;
6269 log = s->log;
6270 } else if (s == NULL)
6271 action = pf_test_udp(&r, &s, dir, kif,
6272 m, off, h, &pd, &a, &ruleset);
6273 break;
6274 }
6275
6276 case IPPROTO_ICMPV6: {
6277 struct icmp6_hdr ih;
6278
6279 pd.hdr.icmp6 = &ih;
6280 if (!pf_pull_hdr(m, off, &ih, sizeof(ih),
6281 &action, &reason, AF_INET6)) {
6282 log = action != PF_PASS;
6283 goto done;
6284 }
6285 if (dir == PF_IN && pf_check_proto_cksum(m, off,
6286 ntohs(h->ip6_plen), IPPROTO_ICMPV6, AF_INET6)) {
6287 action = PF_DROP;
6288 goto done;
6289 }
6290 action = pf_test_state_icmp(&s, dir, kif,
6291 m, off, h, &pd);
6292 if (action == PF_PASS) {
6293#if NPFSYNC
6294 pfsync_update_state(s);
6295#endif
6296 r = s->rule.ptr;
6297 a = s->anchor.ptr;
6298 log = s->log;
6299 } else if (s == NULL)
6300 action = pf_test_icmp(&r, &s, dir, kif,
6301 m, off, h, &pd, &a, &ruleset);
6302 break;
6303 }
6304
6305 default:
6306 action = pf_test_state_other(&s, dir, kif, &pd);
6307 if (action == PF_PASS) {
6308 r = s->rule.ptr;
6309 a = s->anchor.ptr;
6310 log = s->log;
6311 } else if (s == NULL)
6312 action = pf_test_other(&r, &s, dir, kif, m, off, h,
6313 &pd, &a, &ruleset);
6314 break;
6315 }
6316
6317done:
6318 /* XXX handle IPv6 options, if not allowed. not implemented. */
6319
6320#ifdef ALTQ
6321 if (action == PF_PASS && r->qid) {
6322 struct m_tag *mtag;
6323 struct altq_tag *atag;
6324
6325 mtag = m_tag_get(PACKET_TAG_PF_QID, sizeof(*atag), M_NOWAIT);
6326 if (mtag != NULL) {
6327 atag = (struct altq_tag *)(mtag + 1);
6328 if (pd.tos == IPTOS_LOWDELAY)
6329 atag->qid = r->pqid;
6330 else
6331 atag->qid = r->qid;
6332 /* add hints for ecn */
6333 atag->af = AF_INET6;
6334 atag->hdr = h;
6335 m_tag_prepend(m, mtag);
6336 }
6337 }
6338#endif
6339
6340 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
6341 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
6342 (s->nat_rule.ptr->action == PF_RDR ||
6343 s->nat_rule.ptr->action == PF_BINAT) &&
6344 IN6_IS_ADDR_LOOPBACK(&pd.dst->v6) &&
6345 pf_add_mbuf_tag(m, PACKET_TAG_PF_TRANSLATE_LOCALHOST)) {
6346 action = PF_DROP;
6347 REASON_SET(&reason, PFRES_MEMORY);
6348 }
6349
6350 if (log)
6351 PFLOG_PACKET(kif, h, m, AF_INET6, dir, reason, r, a, ruleset);
6352
6353 kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
6354 kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++;
6355
6356 if (action == PF_PASS || r->action == PF_DROP) {
6357 r->packets++;
6358 r->bytes += pd.tot_len;
6359 if (a != NULL) {
6360 a->packets++;
6361 a->bytes += pd.tot_len;
6362 }
6363 if (s != NULL) {
6364 dirndx = (dir == s->direction) ? 0 : 1;
6365 s->packets[dirndx]++;
6366 s->bytes[dirndx] += pd.tot_len;
6367 if (s->nat_rule.ptr != NULL) {
6368 s->nat_rule.ptr->packets++;
6369 s->nat_rule.ptr->bytes += pd.tot_len;
6370 }
6371 if (s->src_node != NULL) {
6372 s->src_node->packets++;
6373 s->src_node->bytes += pd.tot_len;
6374 }
6375 if (s->nat_src_node != NULL) {
6376 s->nat_src_node->packets++;
6377 s->nat_src_node->bytes += pd.tot_len;
6378 }
6379 }
6380 tr = r;
6381 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
6382 if (nr != NULL) {
6383 struct pf_addr *x;
6384 /*
6385 * XXX: we need to make sure that the addresses
6386 * passed to pfr_update_stats() are the same than
6387 * the addresses used during matching (pfr_match)
6388 */
6389 if (r == &pf_default_rule) {
6390 tr = nr;
6391 x = (s == NULL || s->direction == dir) ?
6392 &pd.baddr : &pd.naddr;
6393 } else {
6394 x = (s == NULL || s->direction == dir) ?
6395 &pd.naddr : &pd.baddr;
6396 }
6397 if (x == &pd.baddr || s == NULL) {
6398 if (dir == PF_OUT)
6399 pd.src = x;
6400 else
6401 pd.dst = x;
6402 }
6403 }
6404 if (tr->src.addr.type == PF_ADDR_TABLE)
6405 pfr_update_stats(tr->src.addr.p.tbl, (s == NULL ||
6406 s->direction == dir) ? pd.src : pd.dst, pd.af,
6407 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
6408 tr->src.not);
6409 if (tr->dst.addr.type == PF_ADDR_TABLE)
6410 pfr_update_stats(tr->dst.addr.p.tbl, (s == NULL ||
6411 s->direction == dir) ? pd.dst : pd.src, pd.af,
6412 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
6413 tr->dst.not);
6414 }
6415
6416
6417 if (action == PF_SYNPROXY_DROP) {
6418 m_freem(*m0);
6419 *m0 = NULL;
6420 action = PF_PASS;
6421 } else if (r->rt)
6422 /* pf_route6 can free the mbuf causing *m0 to become NULL */
6423 pf_route6(m0, r, dir, ifp, s);
6424
6425#ifdef __FreeBSD__
6426 PF_UNLOCK();
6427#endif
6428 return (action);
6429}
6430#endif /* INET6 */