1/* $FreeBSD: head/sys/contrib/pf/net/pf.c 130397 2004-06-13 01:36:31Z mlaier $ */
2/* $OpenBSD: pf.c,v 1.389.2.4 2004/04/30 23:27:57 brad Exp $ */
3
4/*
5 * Copyright (c) 2001 Daniel Hartmeier
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 *
12 * - Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * - Redistributions in binary form must reproduce the above
15 * copyright notice, this list of conditions and the following
16 * disclaimer in the documentation and/or other materials provided
17 * with the distribution.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
20 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
21 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
22 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
23 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
24 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
25 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
26 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
27 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
29 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
30 * POSSIBILITY OF SUCH DAMAGE.
31 *
32 * Effort sponsored in part by the Defense Advanced Research Projects
33 * Agency (DARPA) and Air Force Research Laboratory, Air Force
34 * Materiel Command, USAF, under agreement number F30602-01-2-0537.
35 *
36 */
37
38#ifdef __FreeBSD__
39#include "opt_inet.h"
40#include "opt_inet6.h"
41#endif
42
43#ifdef __FreeBSD__
44#include "opt_bpf.h"
45#include "opt_pf.h"
46#define NBPFILTER DEV_BPF
47#define NPFLOG DEV_PFLOG
48#define NPFSYNC DEV_PFSYNC
49#else
50#include "bpfilter.h"
51#include "pflog.h"
52#include "pfsync.h"
53#endif
54
55#include <sys/param.h>
56#include <sys/systm.h>
57#include <sys/mbuf.h>
58#include <sys/filio.h>
59#include <sys/socket.h>
60#include <sys/socketvar.h>
61#include <sys/kernel.h>
62#include <sys/time.h>
63#ifdef __FreeBSD__
64#include <sys/sysctl.h>
65#else
66#include <sys/pool.h>
67#endif
68
69#include <net/if.h>
70#include <net/if_types.h>
71#include <net/bpf.h>
72#include <net/route.h>
73
74#include <netinet/in.h>
75#include <netinet/in_var.h>
76#include <netinet/in_systm.h>
77#include <netinet/ip.h>
78#include <netinet/ip_var.h>
79#include <netinet/tcp.h>
80#include <netinet/tcp_seq.h>
81#include <netinet/udp.h>
82#include <netinet/ip_icmp.h>
83#include <netinet/in_pcb.h>
84#include <netinet/tcp_timer.h>
85#include <netinet/tcp_var.h>
86#include <netinet/udp_var.h>
87#include <netinet/icmp_var.h>
88
89#ifndef __FreeBSD__
90#include <dev/rndvar.h>
91#endif
92#include <net/pfvar.h>
93#include <net/if_pflog.h>
94#include <net/if_pfsync.h>
95
96#ifdef INET6
97#include <netinet/ip6.h>
98#include <netinet/in_pcb.h>
99#include <netinet/icmp6.h>
100#include <netinet6/nd6.h>
101#ifdef __FreeBSD__
102#include <netinet6/ip6_var.h>
103#include <netinet6/in6_pcb.h>
104#endif
105#endif /* INET6 */
106
107#ifdef ALTQ
108#include <altq/if_altq.h>
109#endif
110
111#ifdef __FreeBSD__
112#include <machine/in_cksum.h>
113#if (__FreeBSD_version >= 500112)
114#include <sys/limits.h>
115#else
116#include <machine/limits.h>
117#endif
118#include <sys/ucred.h>
119#endif
120
121#ifdef __FreeBSD__
122extern int ip_optcopy(struct ip *, struct ip *);
123#if (__FreeBSD_version < 501105)
124int ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
125 u_long if_hwassist_flags, int sw_csum);
126#endif
127#endif
128
129#define DPFPRINTF(n, x) if (pf_status.debug >= (n)) printf x
130struct pf_state_tree;
131
132/*
133 * Global variables
134 */
135
136struct pf_anchorqueue pf_anchors;
137struct pf_ruleset pf_main_ruleset;
138struct pf_altqqueue pf_altqs[2];
139struct pf_palist pf_pabuf;
140struct pf_altqqueue *pf_altqs_active;
141struct pf_altqqueue *pf_altqs_inactive;
142struct pf_status pf_status;
143struct ifnet *status_ifp;
144
145u_int32_t ticket_altqs_active;
146u_int32_t ticket_altqs_inactive;
147u_int32_t ticket_pabuf;
148
149#ifdef __FreeBSD__
150struct callout pf_expire_to; /* expire timeout */
151#else
152struct timeout pf_expire_to; /* expire timeout */
153#endif
154
155
156#ifdef __FreeBSD__
157uma_zone_t pf_tree_pl, pf_rule_pl, pf_addr_pl;
158uma_zone_t pf_state_pl, pf_altq_pl, pf_pooladdr_pl;
159#else
160struct pool pf_tree_pl, pf_rule_pl, pf_addr_pl;
161struct pool pf_state_pl, pf_altq_pl, pf_pooladdr_pl;
162#endif
163
164void pf_dynaddr_update(void *);
165#ifdef __FreeBSD__
166void pf_dynaddr_update_event(void *arg, struct ifnet *ifp);
167#endif
168void pf_print_host(struct pf_addr *, u_int16_t, u_int8_t);
169void pf_print_state(struct pf_state *);
170void pf_print_flags(u_int8_t);
171
172u_int16_t pf_cksum_fixup(u_int16_t, u_int16_t, u_int16_t,
173 u_int8_t);
174void pf_change_ap(struct pf_addr *, u_int16_t *,
175 u_int16_t *, u_int16_t *, struct pf_addr *,
176 u_int16_t, u_int8_t, sa_family_t);
177#ifdef INET6
178void pf_change_a6(struct pf_addr *, u_int16_t *,
179 struct pf_addr *, u_int8_t);
180#endif /* INET6 */
181void pf_change_icmp(struct pf_addr *, u_int16_t *,
182 struct pf_addr *, struct pf_addr *, u_int16_t,
183 u_int16_t *, u_int16_t *, u_int16_t *,
184 u_int16_t *, u_int8_t, sa_family_t);
185void pf_send_tcp(const struct pf_rule *, sa_family_t,
186 const struct pf_addr *, const struct pf_addr *,
187 u_int16_t, u_int16_t, u_int32_t, u_int32_t,
188 u_int8_t, u_int16_t, u_int16_t, u_int8_t);
189void pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t,
190 sa_family_t, struct pf_rule *);
191struct pf_rule *pf_match_translation(struct pf_pdesc *, struct mbuf *,
192 int, int, struct ifnet *,
193 struct pf_addr *, u_int16_t, struct pf_addr *,
194 u_int16_t, int);
195struct pf_rule *pf_get_translation(struct pf_pdesc *, struct mbuf *,
196 int, int, struct ifnet *,
197 struct pf_addr *, u_int16_t,
198 struct pf_addr *, u_int16_t,
199 struct pf_addr *, u_int16_t *);
200int pf_test_tcp(struct pf_rule **, struct pf_state **,
201 int, struct ifnet *, struct mbuf *, int, int,
202 void *, struct pf_pdesc *, struct pf_rule **,
203 struct pf_ruleset **);
204int pf_test_udp(struct pf_rule **, struct pf_state **,
205 int, struct ifnet *, struct mbuf *, int, int,
206 void *, struct pf_pdesc *, struct pf_rule **,
207 struct pf_ruleset **);
208int pf_test_icmp(struct pf_rule **, struct pf_state **,
209 int, struct ifnet *, struct mbuf *, int, int,
210 void *, struct pf_pdesc *, struct pf_rule **,
211 struct pf_ruleset **);
212int pf_test_other(struct pf_rule **, struct pf_state **,
213 int, struct ifnet *, struct mbuf *, int, void *,
214 struct pf_pdesc *, struct pf_rule **,
215 struct pf_ruleset **);
216int pf_test_fragment(struct pf_rule **, int,
217 struct ifnet *, struct mbuf *, void *,
218 struct pf_pdesc *, struct pf_rule **,
219 struct pf_ruleset **);
220int pf_test_state_tcp(struct pf_state **, int,
221 struct ifnet *, struct mbuf *, int, int,
222 void *, struct pf_pdesc *, u_short *);
223int pf_test_state_udp(struct pf_state **, int,
224 struct ifnet *, struct mbuf *, int, int,
225 void *, struct pf_pdesc *);
226int pf_test_state_icmp(struct pf_state **, int,
227 struct ifnet *, struct mbuf *, int, int,
228 void *, struct pf_pdesc *);
229int pf_test_state_other(struct pf_state **, int,
230 struct ifnet *, struct pf_pdesc *);
231struct pf_tag *pf_get_tag(struct mbuf *);
232int pf_match_tag(struct mbuf *, struct pf_rule *,
233 struct pf_rule *, struct pf_rule *,
234 struct pf_tag *, int *);
235void pf_hash(struct pf_addr *, struct pf_addr *,
236 struct pf_poolhashkey *, sa_family_t);
237int pf_map_addr(u_int8_t, struct pf_pool *,
238 struct pf_addr *, struct pf_addr *,
239 struct pf_addr *);
240int pf_get_sport(sa_family_t, u_int8_t, struct pf_pool *,
241 struct pf_addr *, struct pf_addr *, u_int16_t,
242 struct pf_addr *, u_int16_t*, u_int16_t, u_int16_t);
243void pf_route(struct mbuf **, struct pf_rule *, int,
244 struct ifnet *, struct pf_state *);
245void pf_route6(struct mbuf **, struct pf_rule *, int,
246 struct ifnet *, struct pf_state *);
247int pf_socket_lookup(uid_t *, gid_t *, int, sa_family_t,
248 int, struct pf_pdesc *);
249u_int8_t pf_get_wscale(struct mbuf *, int, u_int16_t,
250 sa_family_t);
251u_int16_t pf_get_mss(struct mbuf *, int, u_int16_t,
252 sa_family_t);
253u_int16_t pf_calc_mss(struct pf_addr *, sa_family_t,
254 u_int16_t);
255void pf_set_rt_ifp(struct pf_state *,
256 struct pf_addr *);
257int pf_check_proto_cksum(struct mbuf *, int, int,
258 u_int8_t, sa_family_t);
259int pf_addr_wrap_neq(struct pf_addr_wrap *,
260 struct pf_addr_wrap *);
261
262#ifdef __FreeBSD__
263int in4_cksum(struct mbuf *m, u_int8_t nxt, int off, int len);
264
265struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX];
266#else
267struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] =
268 { { &pf_state_pl, PFSTATE_HIWAT }, { &pf_frent_pl, PFFRAG_FRENT_HIWAT } };
269#endif
270
271#define STATE_LOOKUP() \
272 do { \
273 if (direction == PF_IN) \
274 *state = pf_find_state(&tree_ext_gwy, &key); \
275 else \
276 *state = pf_find_state(&tree_lan_ext, &key); \
277 if (*state == NULL) \
278 return (PF_DROP); \
279 if (direction == PF_OUT && \
280 (((*state)->rule.ptr->rt == PF_ROUTETO && \
281 (*state)->rule.ptr->direction == PF_OUT) || \
282 ((*state)->rule.ptr->rt == PF_REPLYTO && \
283 (*state)->rule.ptr->direction == PF_IN)) && \
284 (*state)->rt_ifp != NULL && \
285 (*state)->rt_ifp != ifp) \
286 return (PF_PASS); \
287 } while (0)
288
289#define STATE_TRANSLATE(s) \
290 (s)->lan.addr.addr32[0] != (s)->gwy.addr.addr32[0] || \
291 ((s)->af == AF_INET6 && \
292 ((s)->lan.addr.addr32[1] != (s)->gwy.addr.addr32[1] || \
293 (s)->lan.addr.addr32[2] != (s)->gwy.addr.addr32[2] || \
294 (s)->lan.addr.addr32[3] != (s)->gwy.addr.addr32[3])) || \
295 (s)->lan.port != (s)->gwy.port
296
297static __inline int pf_state_compare(struct pf_tree_node *,
298 struct pf_tree_node *);
299
300struct pf_state_tree tree_lan_ext, tree_ext_gwy;
301RB_GENERATE(pf_state_tree, pf_tree_node, entry, pf_state_compare);
302
303#ifdef __FreeBSD__
304static int
305#else
306static __inline int
307#endif
308pf_state_compare(struct pf_tree_node *a, struct pf_tree_node *b)
309{
310 int diff;
311
312 if ((diff = a->proto - b->proto) != 0)
313 return (diff);
314 if ((diff = a->af - b->af) != 0)
315 return (diff);
316 switch (a->af) {
317#ifdef INET
318 case AF_INET:
319 if (a->addr[0].addr32[0] > b->addr[0].addr32[0])
320 return (1);
321 if (a->addr[0].addr32[0] < b->addr[0].addr32[0])
322 return (-1);
323 if (a->addr[1].addr32[0] > b->addr[1].addr32[0])
324 return (1);
325 if (a->addr[1].addr32[0] < b->addr[1].addr32[0])
326 return (-1);
327 break;
328#endif /* INET */
329#ifdef INET6
330 case AF_INET6:
331 if (a->addr[0].addr32[3] > b->addr[0].addr32[3])
332 return (1);
333 if (a->addr[0].addr32[3] < b->addr[0].addr32[3])
334 return (-1);
335 if (a->addr[1].addr32[3] > b->addr[1].addr32[3])
336 return (1);
337 if (a->addr[1].addr32[3] < b->addr[1].addr32[3])
338 return (-1);
339 if (a->addr[0].addr32[2] > b->addr[0].addr32[2])
340 return (1);
341 if (a->addr[0].addr32[2] < b->addr[0].addr32[2])
342 return (-1);
343 if (a->addr[1].addr32[2] > b->addr[1].addr32[2])
344 return (1);
345 if (a->addr[1].addr32[2] < b->addr[1].addr32[2])
346 return (-1);
347 if (a->addr[0].addr32[1] > b->addr[0].addr32[1])
348 return (1);
349 if (a->addr[0].addr32[1] < b->addr[0].addr32[1])
350 return (-1);
351 if (a->addr[1].addr32[1] > b->addr[1].addr32[1])
352 return (1);
353 if (a->addr[1].addr32[1] < b->addr[1].addr32[1])
354 return (-1);
355 if (a->addr[0].addr32[0] > b->addr[0].addr32[0])
356 return (1);
357 if (a->addr[0].addr32[0] < b->addr[0].addr32[0])
358 return (-1);
359 if (a->addr[1].addr32[0] > b->addr[1].addr32[0])
360 return (1);
361 if (a->addr[1].addr32[0] < b->addr[1].addr32[0])
362 return (-1);
363 break;
364#endif /* INET6 */
365 }
366
367 if ((diff = a->port[0] - b->port[0]) != 0)
368 return (diff);
369 if ((diff = a->port[1] - b->port[1]) != 0)
370 return (diff);
371
372 return (0);
373}
374
375#ifdef INET6
376void
377pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af)
378{
379 switch (af) {
380#ifdef INET
381 case AF_INET:
382 dst->addr32[0] = src->addr32[0];
383 break;
384#endif /* INET */
385 case AF_INET6:
386 dst->addr32[0] = src->addr32[0];
387 dst->addr32[1] = src->addr32[1];
388 dst->addr32[2] = src->addr32[2];
389 dst->addr32[3] = src->addr32[3];
390 break;
391 }
392}
393#endif
394
395struct pf_state *
396pf_find_state(struct pf_state_tree *tree, struct pf_tree_node *key)
397{
398 struct pf_tree_node *k;
399
400 pf_status.fcounters[FCNT_STATE_SEARCH]++;
401 k = RB_FIND(pf_state_tree, tree, key);
402 if (k)
403 return (k->state);
404 else
405 return (NULL);
406}
407
408int
409pf_insert_state(struct pf_state *state)
410{
411 struct pf_tree_node *keya, *keyb;
412
413 keya = pool_get(&pf_tree_pl, PR_NOWAIT);
414 if (keya == NULL)
415 return (-1);
416 keya->state = state;
417 keya->proto = state->proto;
418 keya->af = state->af;
419 PF_ACPY(&keya->addr[0], &state->lan.addr, state->af);
420 keya->port[0] = state->lan.port;
421 PF_ACPY(&keya->addr[1], &state->ext.addr, state->af);
422 keya->port[1] = state->ext.port;
423
424 /* Thou MUST NOT insert multiple duplicate keys */
425 if (RB_INSERT(pf_state_tree, &tree_lan_ext, keya) != NULL) {
426 if (pf_status.debug >= PF_DEBUG_MISC) {
427 printf("pf: state insert failed: tree_lan_ext");
428 printf(" lan: ");
429 pf_print_host(&state->lan.addr, state->lan.port,
430 state->af);
431 printf(" gwy: ");
432 pf_print_host(&state->gwy.addr, state->gwy.port,
433 state->af);
434 printf(" ext: ");
435 pf_print_host(&state->ext.addr, state->ext.port,
436 state->af);
437 printf("\n");
438 }
439 pool_put(&pf_tree_pl, keya);
440 return (-1);
441 }
442
443 keyb = pool_get(&pf_tree_pl, PR_NOWAIT);
444 if (keyb == NULL) {
445 /* Need to pull out the other state */
446 RB_REMOVE(pf_state_tree, &tree_lan_ext, keya);
447 pool_put(&pf_tree_pl, keya);
448 return (-1);
449 }
450 keyb->state = state;
451 keyb->proto = state->proto;
452 keyb->af = state->af;
453 PF_ACPY(&keyb->addr[0], &state->ext.addr, state->af);
454 keyb->port[0] = state->ext.port;
455 PF_ACPY(&keyb->addr[1], &state->gwy.addr, state->af);
456 keyb->port[1] = state->gwy.port;
457
458 if (RB_INSERT(pf_state_tree, &tree_ext_gwy, keyb) != NULL) {
459 if (pf_status.debug >= PF_DEBUG_MISC) {
460 printf("pf: state insert failed: tree_ext_gwy");
461 printf(" lan: ");
462 pf_print_host(&state->lan.addr, state->lan.port,
463 state->af);
464 printf(" gwy: ");
465 pf_print_host(&state->gwy.addr, state->gwy.port,
466 state->af);
467 printf(" ext: ");
468 pf_print_host(&state->ext.addr, state->ext.port,
469 state->af);
470 printf("\n");
471 }
472 RB_REMOVE(pf_state_tree, &tree_lan_ext, keya);
473 pool_put(&pf_tree_pl, keya);
474 pool_put(&pf_tree_pl, keyb);
475 return (-1);
476 }
477
478 pf_status.fcounters[FCNT_STATE_INSERT]++;
479 pf_status.states++;
480#if NPFSYNC
481 pfsync_insert_state(state);
482#endif
483 return (0);
484}
485
486void
487pf_purge_timeout(void *arg)
488{
489#ifdef __FreeBSD__
490 struct callout *to = arg;
491#else
492 struct timeout *to = arg;
493#endif
494 int s;
495
496#ifdef __FreeBSD__
497 PF_LOCK();
498#endif
499 s = splsoftnet();
500 pf_purge_expired_states();
501 pf_purge_expired_fragments();
502 splx(s);
503#ifdef __FreeBSD__
504 PF_UNLOCK();
505#endif
506
507#ifdef __FreeBSD__
508 callout_reset(to, pf_default_rule.timeout[PFTM_INTERVAL] * hz,
509 pf_purge_timeout, to);
510#else
511 timeout_add(to, pf_default_rule.timeout[PFTM_INTERVAL] * hz);
512#endif
513}
514
515u_int32_t
516pf_state_expires(const struct pf_state *state)
517{
518 u_int32_t timeout;
519 u_int32_t start;
520 u_int32_t end;
521 u_int32_t states;
522
523 /* handle all PFTM_* > PFTM_MAX here */
524 if (state->timeout == PFTM_PURGE)
525#ifdef __FreeBSD__
526 return (time_second);
527#else
528 return (time.tv_sec);
529#endif
530 if (state->timeout == PFTM_UNTIL_PACKET)
531 return (0);
532#ifdef __FreeBSD__
533 KASSERT((state->timeout < PFTM_MAX),
534 ("pf_state_expires: timeout > PFTM_MAX"));
535#else
536 KASSERT(state->timeout < PFTM_MAX);
537#endif
538 timeout = state->rule.ptr->timeout[state->timeout];
539 if (!timeout)
540 timeout = pf_default_rule.timeout[state->timeout];
541 start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START];
542 if (start) {
543 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END];
544 states = state->rule.ptr->states;
545 } else {
546 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START];
547 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END];
548 states = pf_status.states;
549 }
550 if (end && states > start && start < end) {
551 if (states < end)
552 return (state->expire + timeout * (end - states) /
553 (end - start));
554 else
555#ifdef __FreeBSD__
556 return (time_second);
557#else
558 return (time.tv_sec);
559#endif
560 }
561 return (state->expire + timeout);
562}
563
564void
565pf_purge_expired_states(void)
566{
567 struct pf_tree_node *cur, *peer, *next;
568 struct pf_tree_node key;
569
570 for (cur = RB_MIN(pf_state_tree, &tree_ext_gwy); cur; cur = next) {
571 next = RB_NEXT(pf_state_tree, &tree_ext_gwy, cur);
572
573#ifdef __FreeBSD__
574 if (pf_state_expires(cur->state) <= (u_int32_t)time_second) {
575#else
576 if (pf_state_expires(cur->state) <= time.tv_sec) {
577#endif
578 if (cur->state->src.state == PF_TCPS_PROXY_DST)
579 pf_send_tcp(cur->state->rule.ptr,
580 cur->state->af,
581 &cur->state->ext.addr,
582 &cur->state->lan.addr,
583 cur->state->ext.port,
584 cur->state->lan.port,
585 cur->state->src.seqhi,
586 cur->state->src.seqlo + 1,
587 0,
588 TH_RST|TH_ACK, 0, 0);
589 RB_REMOVE(pf_state_tree, &tree_ext_gwy, cur);
590
591 /* Need this key's peer (in the other tree) */
592 key.state = cur->state;
593 key.proto = cur->state->proto;
594 key.af = cur->state->af;
595 PF_ACPY(&key.addr[0], &cur->state->lan.addr,
596 cur->state->af);
597 key.port[0] = cur->state->lan.port;
598 PF_ACPY(&key.addr[1], &cur->state->ext.addr,
599 cur->state->af);
600 key.port[1] = cur->state->ext.port;
601
602 peer = RB_FIND(pf_state_tree, &tree_lan_ext, &key);
603#ifdef __FreeBSD__
604 KASSERT((peer), ("peer null :%s", __FUNCTION__));
605 KASSERT((peer->state == cur->state),
606 ("peer->state != cur->state: %s", __FUNCTION__));
607#else
608 KASSERT(peer);
609 KASSERT(peer->state == cur->state);
610#endif
611 RB_REMOVE(pf_state_tree, &tree_lan_ext, peer);
612
613#if NPFSYNC
614 pfsync_delete_state(cur->state);
615#endif
616 if (--cur->state->rule.ptr->states <= 0)
617 pf_rm_rule(NULL, cur->state->rule.ptr);
618 if (cur->state->nat_rule.ptr != NULL)
619 if (--cur->state->nat_rule.ptr->states <= 0)
620 pf_rm_rule(NULL,
621 cur->state->nat_rule.ptr);
622 if (cur->state->anchor.ptr != NULL)
623 if (--cur->state->anchor.ptr->states <= 0)
624 pf_rm_rule(NULL,
625 cur->state->anchor.ptr);
626 pf_normalize_tcp_cleanup(cur->state);
627 pool_put(&pf_state_pl, cur->state);
628 pool_put(&pf_tree_pl, cur);
629 pool_put(&pf_tree_pl, peer);
630 pf_status.fcounters[FCNT_STATE_REMOVALS]++;
631 pf_status.states--;
632 }
633 }
634}
635
636int
637pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw)
638{
639 if (aw->type != PF_ADDR_TABLE)
640 return (0);
641 if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL)
642 return (1);
643 return (0);
644}
645
646void
647pf_tbladdr_remove(struct pf_addr_wrap *aw)
648{
649 if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL)
650 return;
651 pfr_detach_table(aw->p.tbl);
652 aw->p.tbl = NULL;
653}
654
655void
656pf_tbladdr_copyout(struct pf_addr_wrap *aw)
657{
658 struct pfr_ktable *kt = aw->p.tbl;
659
660 if (aw->type != PF_ADDR_TABLE || kt == NULL)
661 return;
662 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
663 kt = kt->pfrkt_root;
664 aw->p.tbl = NULL;
665 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
666 kt->pfrkt_cnt : -1;
667}
668
669int
670pf_dynaddr_setup(struct pf_addr_wrap *aw, sa_family_t af)
671{
672 if (aw->type != PF_ADDR_DYNIFTL)
673 return (0);
674 aw->p.dyn = pool_get(&pf_addr_pl, PR_NOWAIT);
675 if (aw->p.dyn == NULL)
676 return (1);
677 bcopy(aw->v.ifname, aw->p.dyn->ifname, sizeof(aw->p.dyn->ifname));
678 aw->p.dyn->ifp = ifunit(aw->p.dyn->ifname);
679 if (aw->p.dyn->ifp == NULL) {
680 pool_put(&pf_addr_pl, aw->p.dyn);
681 aw->p.dyn = NULL;
682 return (1);
683 }
684 aw->p.dyn->addr = &aw->v.a.addr;
685 aw->p.dyn->af = af;
686 aw->p.dyn->undefined = 1;
687#ifndef __FreeBSD__
688 aw->p.dyn->hook_cookie = hook_establish(
689 aw->p.dyn->ifp->if_addrhooks, 1,
690 pf_dynaddr_update, aw->p.dyn);
691 if (aw->p.dyn->hook_cookie == NULL) {
692 pool_put(&pf_addr_pl, aw->p.dyn);
693 aw->p.dyn = NULL;
694 return (1);
695 }
696#else
697 PF_UNLOCK();
698 aw->p.dyn->hook_cookie = EVENTHANDLER_REGISTER(ifaddr_event,
699 pf_dynaddr_update_event, aw->p.dyn, EVENTHANDLER_PRI_ANY);
700 PF_LOCK();
701 if (aw->p.dyn->hook_cookie == NULL) {
702 pool_put(&pf_addr_pl, aw->p.dyn);
703 aw->p.dyn = NULL;
704 return (1);
705 }
706#endif
707 pf_dynaddr_update(aw->p.dyn);
708 return (0);
709}
710
711#ifdef __FreeBSD__
712void
713pf_dynaddr_update_event(void *arg, struct ifnet *ifp)
714{
715 PF_LOCK();
716 pf_dynaddr_update(arg);
717 PF_UNLOCK();
718}
719#endif
720
721void
722pf_dynaddr_update(void *p)
723{
724 struct pf_addr_dyn *ad = (struct pf_addr_dyn *)p;
725 struct ifaddr *ia;
726 int s, changed = 0;
727
728 if (ad == NULL || ad->ifp == NULL)
729 panic("pf_dynaddr_update");
730 s = splsoftnet();
731 TAILQ_FOREACH(ia, &ad->ifp->if_addrlist, ifa_list)
732 if (ia->ifa_addr != NULL &&
733 ia->ifa_addr->sa_family == ad->af) {
734 if (ad->af == AF_INET) {
735 struct in_addr *a, *b;
736
737 a = &ad->addr->v4;
738 b = &((struct sockaddr_in *)ia->ifa_addr)
739 ->sin_addr;
740 if (ad->undefined ||
741 memcmp(a, b, sizeof(*a))) {
742 bcopy(b, a, sizeof(*a));
743 changed = 1;
744 }
745 } else if (ad->af == AF_INET6) {
746 struct in6_addr *a, *b;
747
748 a = &ad->addr->v6;
749 b = &((struct sockaddr_in6 *)ia->ifa_addr)
750 ->sin6_addr;
751 if (ad->undefined ||
752 memcmp(a, b, sizeof(*a))) {
753 bcopy(b, a, sizeof(*a));
754 changed = 1;
755 }
756 }
757 if (changed)
758 ad->undefined = 0;
759 break;
760 }
761 if (ia == NULL)
762 ad->undefined = 1;
763 splx(s);
764}
765
766void
767pf_dynaddr_remove(struct pf_addr_wrap *aw)
768{
769 if (aw->type != PF_ADDR_DYNIFTL || aw->p.dyn == NULL)
770 return;
771#ifndef __FreeBSD__
772 hook_disestablish(aw->p.dyn->ifp->if_addrhooks,
773 aw->p.dyn->hook_cookie);
774#else
775 PF_UNLOCK();
776 EVENTHANDLER_DEREGISTER(ifaddr_event, aw->p.dyn->hook_cookie);
777 PF_LOCK();
778#endif
779 pool_put(&pf_addr_pl, aw->p.dyn);
780 aw->p.dyn = NULL;
781}
782
783void
784pf_dynaddr_copyout(struct pf_addr_wrap *aw)
785{
786 if (aw->type != PF_ADDR_DYNIFTL || aw->p.dyn == NULL)
787 return;
788 bcopy(aw->p.dyn->ifname, aw->v.ifname, sizeof(aw->v.ifname));
789 aw->p.dyn = (struct pf_addr_dyn *)1;
790}
791
792void
793pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af)
794{
795 switch (af) {
796#ifdef INET
797 case AF_INET: {
798 u_int32_t a = ntohl(addr->addr32[0]);
799 printf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255,
800 (a>>8)&255, a&255);
801 if (p) {
802 p = ntohs(p);
803 printf(":%u", p);
804 }
805 break;
806 }
807#endif /* INET */
808#ifdef INET6
809 case AF_INET6: {
810 u_int16_t b;
811 u_int8_t i, curstart = 255, curend = 0,
812 maxstart = 0, maxend = 0;
813 for (i = 0; i < 8; i++) {
814 if (!addr->addr16[i]) {
815 if (curstart == 255)
816 curstart = i;
817 else
818 curend = i;
819 } else {
820 if (curstart) {
821 if ((curend - curstart) >
822 (maxend - maxstart)) {
823 maxstart = curstart;
824 maxend = curend;
825 curstart = 255;
826 }
827 }
828 }
829 }
830 for (i = 0; i < 8; i++) {
831 if (i >= maxstart && i <= maxend) {
832 if (maxend != 7) {
833 if (i == maxstart)
834 printf(":");
835 } else {
836 if (i == maxend)
837 printf(":");
838 }
839 } else {
840 b = ntohs(addr->addr16[i]);
841 printf("%x", b);
842 if (i < 7)
843 printf(":");
844 }
845 }
846 if (p) {
847 p = ntohs(p);
848 printf("[%u]", p);
849 }
850 break;
851 }
852#endif /* INET6 */
853 }
854}
855
856void
857pf_print_state(struct pf_state *s)
858{
859 switch (s->proto) {
860 case IPPROTO_TCP:
861 printf("TCP ");
862 break;
863 case IPPROTO_UDP:
864 printf("UDP ");
865 break;
866 case IPPROTO_ICMP:
867 printf("ICMP ");
868 break;
869 case IPPROTO_ICMPV6:
870 printf("ICMPV6 ");
871 break;
872 default:
873 printf("%u ", s->proto);
874 break;
875 }
876 pf_print_host(&s->lan.addr, s->lan.port, s->af);
877 printf(" ");
878 pf_print_host(&s->gwy.addr, s->gwy.port, s->af);
879 printf(" ");
880 pf_print_host(&s->ext.addr, s->ext.port, s->af);
881 printf(" [lo=%u high=%u win=%u modulator=%u", s->src.seqlo,
882 s->src.seqhi, s->src.max_win, s->src.seqdiff);
883 if (s->src.wscale && s->dst.wscale)
884 printf(" wscale=%u", s->src.wscale & PF_WSCALE_MASK);
885 printf("]");
886 printf(" [lo=%u high=%u win=%u modulator=%u", s->dst.seqlo,
887 s->dst.seqhi, s->dst.max_win, s->dst.seqdiff);
888 if (s->src.wscale && s->dst.wscale)
889 printf(" wscale=%u", s->dst.wscale & PF_WSCALE_MASK);
890 printf("]");
891 printf(" %u:%u", s->src.state, s->dst.state);
892}
893
894void
895pf_print_flags(u_int8_t f)
896{
897 if (f)
898 printf(" ");
899 if (f & TH_FIN)
900 printf("F");
901 if (f & TH_SYN)
902 printf("S");
903 if (f & TH_RST)
904 printf("R");
905 if (f & TH_PUSH)
906 printf("P");
907 if (f & TH_ACK)
908 printf("A");
909 if (f & TH_URG)
910 printf("U");
911 if (f & TH_ECE)
912 printf("E");
913 if (f & TH_CWR)
914 printf("W");
915}
916
917#define PF_SET_SKIP_STEPS(i) \
918 do { \
919 while (head[i] != cur) { \
920 head[i]->skip[i].ptr = cur; \
921 head[i] = TAILQ_NEXT(head[i], entries); \
922 } \
923 } while (0)
924
925void
926pf_calc_skip_steps(struct pf_rulequeue *rules)
927{
928 struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT];
929 int i;
930
931 cur = TAILQ_FIRST(rules);
932 prev = cur;
933 for (i = 0; i < PF_SKIP_COUNT; ++i)
934 head[i] = cur;
935 while (cur != NULL) {
936
937 if (cur->ifp != prev->ifp || cur->ifnot != prev->ifnot)
938 PF_SET_SKIP_STEPS(PF_SKIP_IFP);
939 if (cur->direction != prev->direction)
940 PF_SET_SKIP_STEPS(PF_SKIP_DIR);
941 if (cur->af != prev->af)
942 PF_SET_SKIP_STEPS(PF_SKIP_AF);
943 if (cur->proto != prev->proto)
944 PF_SET_SKIP_STEPS(PF_SKIP_PROTO);
945 if (cur->src.not != prev->src.not ||
946 pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr))
947 PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR);
948 if (cur->src.port[0] != prev->src.port[0] ||
949 cur->src.port[1] != prev->src.port[1] ||
950 cur->src.port_op != prev->src.port_op)
951 PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
952 if (cur->dst.not != prev->dst.not ||
953 pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr))
954 PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR);
955 if (cur->dst.port[0] != prev->dst.port[0] ||
956 cur->dst.port[1] != prev->dst.port[1] ||
957 cur->dst.port_op != prev->dst.port_op)
958 PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
959
960 prev = cur;
961 cur = TAILQ_NEXT(cur, entries);
962 }
963 for (i = 0; i < PF_SKIP_COUNT; ++i)
964 PF_SET_SKIP_STEPS(i);
965}
966
967int
968pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2)
969{
970 if (aw1->type != aw2->type)
971 return (1);
972 switch (aw1->type) {
973 case PF_ADDR_ADDRMASK:
974 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0))
975 return (1);
976 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0))
977 return (1);
978 return (0);
979 case PF_ADDR_DYNIFTL:
980 if (aw1->p.dyn->ifp != aw2->p.dyn->ifp)
981 return (1);
982 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0))
983 return (1);
984 return (0);
985 case PF_ADDR_NOROUTE:
986 return (0);
987 case PF_ADDR_TABLE:
988 return (aw1->p.tbl != aw2->p.tbl);
989 default:
990 printf("invalid address type: %d\n", aw1->type);
991 return (1);
992 }
993}
994
995void
996pf_rule_set_qid(struct pf_rulequeue *rules)
997{
998 struct pf_rule *rule;
999
1000 TAILQ_FOREACH(rule, rules, entries)
1001 if (rule->qname[0] != 0) {
1002 rule->qid = pf_qname_to_qid(rule->qname);
1003 if (rule->pqname[0] != 0)
1004 rule->pqid = pf_qname_to_qid(rule->pqname);
1005 else
1006 rule->pqid = rule->qid;
1007 }
1008}
1009
1010u_int32_t
1011pf_qname_to_qid(char *qname)
1012{
1013 struct pf_altq *altq;
1014
1015 TAILQ_FOREACH(altq, pf_altqs_active, entries)
1016 if (!strcmp(altq->qname, qname))
1017 return (altq->qid);
1018
1019 return (0);
1020}
1021
1022void
1023pf_update_anchor_rules()
1024{
1025 struct pf_rule *rule;
1026 int i;
1027
1028 for (i = 0; i < PF_RULESET_MAX; ++i)
1029 TAILQ_FOREACH(rule, pf_main_ruleset.rules[i].active.ptr,
1030 entries)
1031 if (rule->anchorname[0])
1032 rule->anchor = pf_find_anchor(rule->anchorname);
1033 else
1034 rule->anchor = NULL;
1035}
1036
1037u_int16_t
1038pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp)
1039{
1040 u_int32_t l;
1041
1042 if (udp && !cksum)
1043 return (0x0000);
1044 l = cksum + old - new;
1045 l = (l >> 16) + (l & 65535);
1046 l = l & 65535;
1047 if (udp && !l)
1048 return (0xFFFF);
1049 return (l);
1050}
1051
1052void
1053pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc,
1054 struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af)
1055{
1056 struct pf_addr ao;
1057 u_int16_t po = *p;
1058
1059 PF_ACPY(&ao, a, af);
1060 PF_ACPY(a, an, af);
1061
1062 *p = pn;
1063
1064 switch (af) {
1065#ifdef INET
1066 case AF_INET:
1067 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1068 ao.addr16[0], an->addr16[0], 0),
1069 ao.addr16[1], an->addr16[1], 0);
1070 *p = pn;
1071 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1072 ao.addr16[0], an->addr16[0], u),
1073 ao.addr16[1], an->addr16[1], u),
1074 po, pn, u);
1075 break;
1076#endif /* INET */
1077#ifdef INET6
1078 case AF_INET6:
1079 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1080 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1081 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1082 ao.addr16[0], an->addr16[0], u),
1083 ao.addr16[1], an->addr16[1], u),
1084 ao.addr16[2], an->addr16[2], u),
1085 ao.addr16[3], an->addr16[3], u),
1086 ao.addr16[4], an->addr16[4], u),
1087 ao.addr16[5], an->addr16[5], u),
1088 ao.addr16[6], an->addr16[6], u),
1089 ao.addr16[7], an->addr16[7], u),
1090 po, pn, u);
1091 break;
1092#endif /* INET6 */
1093 }
1094}
1095
1096
1097/* Changes a u_int32_t. Uses a void * so there are no align restrictions */
1098void
1099pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u)
1100{
1101 u_int32_t ao;
1102
1103 memcpy(&ao, a, sizeof(ao));
1104 memcpy(a, &an, sizeof(u_int32_t));
1105 *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u),
1106 ao % 65536, an % 65536, u);
1107}
1108
1109#ifdef INET6
1110void
1111pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u)
1112{
1113 struct pf_addr ao;
1114
1115 PF_ACPY(&ao, a, AF_INET6);
1116 PF_ACPY(a, an, AF_INET6);
1117
1118 *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1119 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1120 pf_cksum_fixup(pf_cksum_fixup(*c,
1121 ao.addr16[0], an->addr16[0], u),
1122 ao.addr16[1], an->addr16[1], u),
1123 ao.addr16[2], an->addr16[2], u),
1124 ao.addr16[3], an->addr16[3], u),
1125 ao.addr16[4], an->addr16[4], u),
1126 ao.addr16[5], an->addr16[5], u),
1127 ao.addr16[6], an->addr16[6], u),
1128 ao.addr16[7], an->addr16[7], u);
1129}
1130#endif /* INET6 */
1131
1132void
1133pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa,
1134 struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c,
1135 u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af)
1136{
1137 struct pf_addr oia, ooa;
1138
1139 PF_ACPY(&oia, ia, af);
1140 PF_ACPY(&ooa, oa, af);
1141
1142 /* Change inner protocol port, fix inner protocol checksum. */
1143 if (ip != NULL) {
1144 u_int16_t oip = *ip;
1145 u_int32_t opc = 0; /* make the compiler happy */
1146
1147 if (pc != NULL)
1148 opc = *pc;
1149 *ip = np;
1150 if (pc != NULL)
1151 *pc = pf_cksum_fixup(*pc, oip, *ip, u);
1152 *ic = pf_cksum_fixup(*ic, oip, *ip, 0);
1153 if (pc != NULL)
1154 *ic = pf_cksum_fixup(*ic, opc, *pc, 0);
1155 }
1156 /* Change inner ip address, fix inner ip and icmp checksums. */
1157 PF_ACPY(ia, na, af);
1158 switch (af) {
1159#ifdef INET
1160 case AF_INET: {
1161 u_int32_t oh2c = *h2c;
1162
1163 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c,
1164 oia.addr16[0], ia->addr16[0], 0),
1165 oia.addr16[1], ia->addr16[1], 0);
1166 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1167 oia.addr16[0], ia->addr16[0], 0),
1168 oia.addr16[1], ia->addr16[1], 0);
1169 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0);
1170 break;
1171 }
1172#endif /* INET */
1173#ifdef INET6
1174 case AF_INET6:
1175 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1176 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1177 pf_cksum_fixup(pf_cksum_fixup(*ic,
1178 oia.addr16[0], ia->addr16[0], u),
1179 oia.addr16[1], ia->addr16[1], u),
1180 oia.addr16[2], ia->addr16[2], u),
1181 oia.addr16[3], ia->addr16[3], u),
1182 oia.addr16[4], ia->addr16[4], u),
1183 oia.addr16[5], ia->addr16[5], u),
1184 oia.addr16[6], ia->addr16[6], u),
1185 oia.addr16[7], ia->addr16[7], u);
1186 break;
1187#endif /* INET6 */
1188 }
1189 /* Change outer ip address, fix outer ip or icmpv6 checksum. */
1190 PF_ACPY(oa, na, af);
1191 switch (af) {
1192#ifdef INET
1193 case AF_INET:
1194 *hc = pf_cksum_fixup(pf_cksum_fixup(*hc,
1195 ooa.addr16[0], oa->addr16[0], 0),
1196 ooa.addr16[1], oa->addr16[1], 0);
1197 break;
1198#endif /* INET */
1199#ifdef INET6
1200 case AF_INET6:
1201 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1202 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1203 pf_cksum_fixup(pf_cksum_fixup(*ic,
1204 ooa.addr16[0], oa->addr16[0], u),
1205 ooa.addr16[1], oa->addr16[1], u),
1206 ooa.addr16[2], oa->addr16[2], u),
1207 ooa.addr16[3], oa->addr16[3], u),
1208 ooa.addr16[4], oa->addr16[4], u),
1209 ooa.addr16[5], oa->addr16[5], u),
1210 ooa.addr16[6], oa->addr16[6], u),
1211 ooa.addr16[7], oa->addr16[7], u);
1212 break;
1213#endif /* INET6 */
1214 }
1215}
1216
1217void
1218pf_send_tcp(const struct pf_rule *r, sa_family_t af,
1219 const struct pf_addr *saddr, const struct pf_addr *daddr,
1220 u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack,
1221 u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl)
1222{
1223 struct mbuf *m;
1224 struct m_tag *mtag;
1225 int len = 0, tlen; /* make the compiler happy */
1226#ifdef INET
1227 struct ip *h = NULL; /* make the compiler happy */
1228#endif /* INET */
1229#ifdef INET6
1230 struct ip6_hdr *h6 = NULL; /* make the compiler happy */
1231#endif /* INET6 */
1232 struct tcphdr *th = NULL; /* make the compiler happy */
1233#ifdef __FreeBSD__
1234 struct ip *ip;
1235#if (__FreeBSD_version < 501114)
1236 struct route ro;
1237#endif
1238#endif
1239 char *opt;
1240
1241 /* maximum segment size tcp option */
1242 tlen = sizeof(struct tcphdr);
1243 if (mss)
1244 tlen += 4;
1245
1246 switch (af) {
1247#ifdef INET
1248 case AF_INET:
1249 len = sizeof(struct ip) + tlen;
1250 break;
1251#endif /* INET */
1252#ifdef INET6
1253 case AF_INET6:
1254 len = sizeof(struct ip6_hdr) + tlen;
1255 break;
1256#endif /* INET6 */
1257 }
1258
1259 /* create outgoing mbuf */
1260 mtag = m_tag_get(PACKET_TAG_PF_GENERATED, 0, M_NOWAIT);
1261 if (mtag == NULL)
1262 return;
1263 m = m_gethdr(M_DONTWAIT, MT_HEADER);
1264 if (m == NULL) {
1265 m_tag_free(mtag);
1266 return;
1267 }
1268 m_tag_prepend(m, mtag);
1269#ifdef ALTQ
1270 if (r != NULL && r->qid) {
1271 struct altq_tag *atag;
1272
1273 mtag = m_tag_get(PACKET_TAG_PF_QID, sizeof(*atag), M_NOWAIT);
1274 if (mtag != NULL) {
1275 atag = (struct altq_tag *)(mtag + 1);
1276 atag->qid = r->qid;
1277 /* add hints for ecn */
1278 atag->af = af;
1279 atag->hdr = mtod(m, struct ip *);
1280 m_tag_prepend(m, mtag);
1281 }
1282 }
1283#endif
1284 m->m_data += max_linkhdr;
1285 m->m_pkthdr.len = m->m_len = len;
1286 m->m_pkthdr.rcvif = NULL;
1287 bzero(m->m_data, len);
1288 switch (af) {
1289#ifdef INET
1290 case AF_INET:
1291 h = mtod(m, struct ip *);
1292
1293 /* IP header fields included in the TCP checksum */
1294 h->ip_p = IPPROTO_TCP;
1295 h->ip_len = htons(tlen);
1296 h->ip_src.s_addr = saddr->v4.s_addr;
1297 h->ip_dst.s_addr = daddr->v4.s_addr;
1298
1299 th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip));
1300 break;
1301#endif /* INET */
1302#ifdef INET6
1303 case AF_INET6:
1304 h6 = mtod(m, struct ip6_hdr *);
1305
1306 /* IP header fields included in the TCP checksum */
1307 h6->ip6_nxt = IPPROTO_TCP;
1308 h6->ip6_plen = htons(tlen);
1309 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr));
1310 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr));
1311
1312 th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr));
1313 break;
1314#endif /* INET6 */
1315 }
1316
1317 /* TCP header */
1318 th->th_sport = sport;
1319 th->th_dport = dport;
1320 th->th_seq = htonl(seq);
1321 th->th_ack = htonl(ack);
1322 th->th_off = tlen >> 2;
1323 th->th_flags = flags;
1324 th->th_win = htons(win);
1325
1326 if (mss) {
1327 opt = (char *)(th + 1);
1328 opt[0] = TCPOPT_MAXSEG;
1329 opt[1] = 4;
1330 HTONS(mss);
1331 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2);
1332 }
1333
1334 switch (af) {
1335#ifdef INET
1336 case AF_INET:
1337 /* TCP checksum */
1338 th->th_sum = in_cksum(m, len);
1339
1340 /* Finish the IP header */
1341 h->ip_v = 4;
1342 h->ip_hl = sizeof(*h) >> 2;
1343 h->ip_tos = IPTOS_LOWDELAY;
1344#ifdef __FreeBSD__
1345 h->ip_off = htons(path_mtu_discovery ? IP_DF : 0);
1346#else
1347 h->ip_off = htons(ip_mtudisc ? IP_DF : 0);
1348#endif
1349 h->ip_len = htons(len);
1350 h->ip_ttl = ttl ? ttl : ip_defttl;
1351 h->ip_sum = 0;
1352#ifdef __FreeBSD__
1353 ip = mtod(m, struct ip *);
1354 /*
1355 * XXX
1356 * OpenBSD changed ip_len/ip_off byte ordering!
1357 * Because FreeBSD assumes host byte ordering we need to
1358 * change here.
1359 */
1360 NTOHS(ip->ip_len);
1361 NTOHS(ip->ip_off);
1362#if (__FreeBSD_version < 501114)
1363 bzero(&ro, sizeof(ro));
1364 ip_rtaddr(ip->ip_dst, &ro);
1365 PF_UNLOCK();
1366 ip_output(m, (void *)NULL, &ro, 0, (void *)NULL,
1367 (void *)NULL);
1368 PF_LOCK();
1369 if(ro.ro_rt) {
1370 RTFREE(ro.ro_rt);
1371 }
1372#else /* __FreeBSD_version >= 501114 */
1373 PF_UNLOCK();
1374 ip_output(m, (void *)NULL, (void *)NULL, 0, (void *)NULL,
1375 (void *)NULL);
1376 PF_LOCK();
1377#endif
1378#else /* ! __FreeBSD__ */
1379 ip_output(m, (void *)NULL, (void *)NULL, 0, (void *)NULL,
1380 (void *)NULL);
1381#endif
1382 break;
1383#endif /* INET */
1384#ifdef INET6
1385 case AF_INET6:
1386 /* TCP checksum */
1387 th->th_sum = in6_cksum(m, IPPROTO_TCP,
1388 sizeof(struct ip6_hdr), tlen);
1389
1390 h6->ip6_vfc |= IPV6_VERSION;
1391 h6->ip6_hlim = IPV6_DEFHLIM;
1392
1393#ifdef __FreeBSD__
1394 PF_UNLOCK();
1395 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
1396 PF_LOCK();
1397#else
1398 ip6_output(m, NULL, NULL, 0, NULL, NULL);
1399#endif
1400 break;
1401#endif /* INET6 */
1402 }
1403}
1404
1405void
1406pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af,
1407 struct pf_rule *r)
1408{
1409 struct m_tag *mtag;
1410 struct mbuf *m0;
1411#ifdef __FreeBSD__
1412 struct ip *ip;
1413#endif
1414
1415 mtag = m_tag_get(PACKET_TAG_PF_GENERATED, 0, M_NOWAIT);
1416 if (mtag == NULL)
1417 return;
1418#ifdef __FreeBSD__
1419 m0 = m_copypacket(m, M_DONTWAIT);
1420#else
1421 m0 = m_copy(m, 0, M_COPYALL);
1422#endif
1423 if (m0 == NULL) {
1424 m_tag_free(mtag);
1425 return;
1426 }
1427 m_tag_prepend(m0, mtag);
1428
1429#ifdef ALTQ
1430 if (r->qid) {
1431 struct altq_tag *atag;
1432
1433 mtag = m_tag_get(PACKET_TAG_PF_QID, sizeof(*atag), M_NOWAIT);
1434 if (mtag != NULL) {
1435 atag = (struct altq_tag *)(mtag + 1);
1436 atag->qid = r->qid;
1437 /* add hints for ecn */
1438 atag->af = af;
1439 atag->hdr = mtod(m0, struct ip *);
1440 m_tag_prepend(m0, mtag);
1441 }
1442 }
1443#endif
1444
1445 switch (af) {
1446#ifdef INET
1447 case AF_INET:
1448#ifdef __FreeBSD__
1449 /* icmp_error() expects host byte ordering */
1450 ip = mtod(m0, struct ip *);
1451 NTOHS(ip->ip_len);
1452 NTOHS(ip->ip_off);
1453 PF_UNLOCK();
1454#endif
1455 icmp_error(m0, type, code, 0, NULL);
1456#ifdef __FreeBSD__
1457 PF_LOCK();
1458#endif
1459 break;
1460#endif /* INET */
1461#ifdef INET6
1462 case AF_INET6:
1463#ifdef __FreeBSD__
1464 PF_UNLOCK();
1465#endif
1466 icmp6_error(m0, type, code, 0);
1467#ifdef __FreeBSD__
1468 PF_LOCK();
1469#endif
1470 break;
1471#endif /* INET6 */
1472 }
1473}
1474
1475/*
1476 * Return 1 if the addresses a and b match (with mask m), otherwise return 0.
1477 * If n is 0, they match if they are equal. If n is != 0, they match if they
1478 * are different.
1479 */
1480int
1481pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m,
1482 struct pf_addr *b, sa_family_t af)
1483{
1484 int match = 0;
1485
1486 switch (af) {
1487#ifdef INET
1488 case AF_INET:
1489 if ((a->addr32[0] & m->addr32[0]) ==
1490 (b->addr32[0] & m->addr32[0]))
1491 match++;
1492 break;
1493#endif /* INET */
1494#ifdef INET6
1495 case AF_INET6:
1496 if (((a->addr32[0] & m->addr32[0]) ==
1497 (b->addr32[0] & m->addr32[0])) &&
1498 ((a->addr32[1] & m->addr32[1]) ==
1499 (b->addr32[1] & m->addr32[1])) &&
1500 ((a->addr32[2] & m->addr32[2]) ==
1501 (b->addr32[2] & m->addr32[2])) &&
1502 ((a->addr32[3] & m->addr32[3]) ==
1503 (b->addr32[3] & m->addr32[3])))
1504 match++;
1505 break;
1506#endif /* INET6 */
1507 }
1508 if (match) {
1509 if (n)
1510 return (0);
1511 else
1512 return (1);
1513 } else {
1514 if (n)
1515 return (1);
1516 else
1517 return (0);
1518 }
1519}
1520
1521int
1522pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p)
1523{
1524 switch (op) {
1525 case PF_OP_IRG:
1526 return ((p > a1) && (p < a2));
1527 case PF_OP_XRG:
1528 return ((p < a1) || (p > a2));
1529 case PF_OP_RRG:
1530 return ((p >= a1) && (p <= a2));
1531 case PF_OP_EQ:
1532 return (p == a1);
1533 case PF_OP_NE:
1534 return (p != a1);
1535 case PF_OP_LT:
1536 return (p < a1);
1537 case PF_OP_LE:
1538 return (p <= a1);
1539 case PF_OP_GT:
1540 return (p > a1);
1541 case PF_OP_GE:
1542 return (p >= a1);
1543 }
1544 return (0); /* never reached */
1545}
1546
1547int
1548pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p)
1549{
1550 NTOHS(a1);
1551 NTOHS(a2);
1552 NTOHS(p);
1553 return (pf_match(op, a1, a2, p));
1554}
1555
1556int
1557pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u)
1558{
1559 if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
1560 return (0);
1561 return (pf_match(op, a1, a2, u));
1562}
1563
1564int
1565pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g)
1566{
1567 if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
1568 return (0);
1569 return (pf_match(op, a1, a2, g));
1570}
1571
1572struct pf_tag *
1573pf_get_tag(struct mbuf *m)
1574{
1575 struct m_tag *mtag;
1576
1577 if ((mtag = m_tag_find(m, PACKET_TAG_PF_TAG, NULL)) != NULL)
1578 return ((struct pf_tag *)(mtag + 1));
1579 else
1580 return (NULL);
1581}
1582
1583int
1584pf_match_tag(struct mbuf *m, struct pf_rule *r, struct pf_rule *nat,
1585 struct pf_rule *rdr, struct pf_tag *pftag, int *tag)
1586{
1587 if (*tag == -1) { /* find mbuf tag */
1588 pftag = pf_get_tag(m);
1589 if (pftag != NULL)
1590 *tag = pftag->tag;
1591 else
1592 *tag = 0;
1593 if (nat != NULL && nat->tag)
1594 *tag = nat->tag;
1595 if (rdr != NULL && rdr->tag)
1596 *tag = rdr->tag;
1597 }
1598
1599 return ((!r->match_tag_not && r->match_tag == *tag) ||
1600 (r->match_tag_not && r->match_tag != *tag));
1601}
1602
1603int
1604pf_tag_packet(struct mbuf *m, struct pf_tag *pftag, int tag)
1605{
1606 struct m_tag *mtag;
1607
1608 if (tag <= 0)
1609 return (0);
1610
1611 if (pftag == NULL) {
1612 mtag = m_tag_get(PACKET_TAG_PF_TAG, sizeof(*pftag), M_NOWAIT);
1613 if (mtag == NULL)
1614 return (1);
1615 ((struct pf_tag *)(mtag + 1))->tag = tag;
1616 m_tag_prepend(m, mtag);
1617 } else
1618 pftag->tag = tag;
1619
1620 return (0);
1621}
1622
1623#define PF_STEP_INTO_ANCHOR(r, a, s, n) \
1624 do { \
1625 if ((r) == NULL || (r)->anchor == NULL || \
1626 (s) != NULL || (a) != NULL) \
1627 panic("PF_STEP_INTO_ANCHOR"); \
1628 (a) = (r); \
1629 (s) = TAILQ_FIRST(&(r)->anchor->rulesets); \
1630 (r) = NULL; \
1631 while ((s) != NULL && ((r) = \
1632 TAILQ_FIRST((s)->rules[n].active.ptr)) == NULL) \
1633 (s) = TAILQ_NEXT((s), entries); \
1634 if ((r) == NULL) { \
1635 (r) = TAILQ_NEXT((a), entries); \
1636 (a) = NULL; \
1637 } \
1638 } while (0)
1639
1640#define PF_STEP_OUT_OF_ANCHOR(r, a, s, n) \
1641 do { \
1642 if ((r) != NULL || (a) == NULL || (s) == NULL) \
1643 panic("PF_STEP_OUT_OF_ANCHOR"); \
1644 (s) = TAILQ_NEXT((s), entries); \
1645 while ((s) != NULL && ((r) = \
1646 TAILQ_FIRST((s)->rules[n].active.ptr)) == NULL) \
1647 (s) = TAILQ_NEXT((s), entries); \
1648 if ((r) == NULL) { \
1649 (r) = TAILQ_NEXT((a), entries); \
1650 (a) = NULL; \
1651 } \
1652 } while (0)
1653
1654#ifdef INET6
1655void
1656pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr,
1657 struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af)
1658{
1659 switch (af) {
1660#ifdef INET
1661 case AF_INET:
1662 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
1663 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
1664 break;
1665#endif /* INET */
1666 case AF_INET6:
1667 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
1668 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
1669 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) |
1670 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]);
1671 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) |
1672 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]);
1673 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) |
1674 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]);
1675 break;
1676 }
1677}
1678
1679void
1680pf_addr_inc(struct pf_addr *addr, u_int8_t af)
1681{
1682 switch (af) {
1683#ifdef INET
1684 case AF_INET:
1685 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1);
1686 break;
1687#endif /* INET */
1688 case AF_INET6:
1689 if (addr->addr32[3] == 0xffffffff) {
1690 addr->addr32[3] = 0;
1691 if (addr->addr32[2] == 0xffffffff) {
1692 addr->addr32[2] = 0;
1693 if (addr->addr32[1] == 0xffffffff) {
1694 addr->addr32[1] = 0;
1695 addr->addr32[0] =
1696 htonl(ntohl(addr->addr32[0]) + 1);
1697 } else
1698 addr->addr32[1] =
1699 htonl(ntohl(addr->addr32[1]) + 1);
1700 } else
1701 addr->addr32[2] =
1702 htonl(ntohl(addr->addr32[2]) + 1);
1703 } else
1704 addr->addr32[3] =
1705 htonl(ntohl(addr->addr32[3]) + 1);
1706 break;
1707 }
1708}
1709#endif /* INET6 */
1710
1711#define mix(a,b,c) \
1712 do { \
1713 a -= b; a -= c; a ^= (c >> 13); \
1714 b -= c; b -= a; b ^= (a << 8); \
1715 c -= a; c -= b; c ^= (b >> 13); \
1716 a -= b; a -= c; a ^= (c >> 12); \
1717 b -= c; b -= a; b ^= (a << 16); \
1718 c -= a; c -= b; c ^= (b >> 5); \
1719 a -= b; a -= c; a ^= (c >> 3); \
1720 b -= c; b -= a; b ^= (a << 10); \
1721 c -= a; c -= b; c ^= (b >> 15); \
1722 } while (0)
1723
1724/*
1725 * hash function based on bridge_hash in if_bridge.c
1726 */
1727void
1728pf_hash(struct pf_addr *inaddr, struct pf_addr *hash,
1729 struct pf_poolhashkey *key, sa_family_t af)
1730{
1731 u_int32_t a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0];
1732
1733 switch (af) {
1734#ifdef INET
1735 case AF_INET:
1736 a += inaddr->addr32[0];
1737 b += key->key32[1];
1738 mix(a, b, c);
1739 hash->addr32[0] = c + key->key32[2];
1740 break;
1741#endif /* INET */
1742#ifdef INET6
1743 case AF_INET6:
1744 a += inaddr->addr32[0];
1745 b += inaddr->addr32[2];
1746 mix(a, b, c);
1747 hash->addr32[0] = c;
1748 a += inaddr->addr32[1];
1749 b += inaddr->addr32[3];
1750 c += key->key32[1];
1751 mix(a, b, c);
1752 hash->addr32[1] = c;
1753 a += inaddr->addr32[2];
1754 b += inaddr->addr32[1];
1755 c += key->key32[2];
1756 mix(a, b, c);
1757 hash->addr32[2] = c;
1758 a += inaddr->addr32[3];
1759 b += inaddr->addr32[0];
1760 c += key->key32[3];
1761 mix(a, b, c);
1762 hash->addr32[3] = c;
1763 break;
1764#endif /* INET6 */
1765 }
1766}
1767
1768int
1769pf_map_addr(u_int8_t af, struct pf_pool *rpool, struct pf_addr *saddr,
1770 struct pf_addr *naddr, struct pf_addr *init_addr)
1771{
1772 unsigned char hash[16];
1773 struct pf_addr *raddr;
1774 struct pf_addr *rmask;
1775 struct pf_pooladdr *acur = rpool->cur;
1776
1777 if (rpool->cur->addr.type == PF_ADDR_NOROUTE)
1778 return (1);
1779 if (rpool->cur->addr.type == PF_ADDR_DYNIFTL &&
1780 rpool->cur->addr.p.dyn->undefined)
1781 return (1);
1782 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
1783 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN)
1784 return (1); /* unsupported */
1785 } else {
1786 raddr = &rpool->cur->addr.v.a.addr;
1787 rmask = &rpool->cur->addr.v.a.mask;
1788 }
1789
1790 switch (rpool->opts & PF_POOL_TYPEMASK) {
1791 case PF_POOL_NONE:
1792 PF_ACPY(naddr, raddr, af);
1793 break;
1794 case PF_POOL_BITMASK:
1795 PF_POOLMASK(naddr, raddr, rmask, saddr, af);
1796 break;
1797 case PF_POOL_RANDOM:
1798 if (init_addr != NULL && PF_AZERO(init_addr, af)) {
1799 switch (af) {
1800#ifdef INET
1801 case AF_INET:
1802 rpool->counter.addr32[0] = arc4random();
1803 break;
1804#endif /* INET */
1805#ifdef INET6
1806 case AF_INET6:
1807 if (rmask->addr32[3] != 0xffffffff)
1808 rpool->counter.addr32[3] = arc4random();
1809 else
1810 break;
1811 if (rmask->addr32[2] != 0xffffffff)
1812 rpool->counter.addr32[2] = arc4random();
1813 else
1814 break;
1815 if (rmask->addr32[1] != 0xffffffff)
1816 rpool->counter.addr32[1] = arc4random();
1817 else
1818 break;
1819 if (rmask->addr32[0] != 0xffffffff)
1820 rpool->counter.addr32[0] = arc4random();
1821 break;
1822#endif /* INET6 */
1823 }
1824 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
1825 PF_ACPY(init_addr, naddr, af);
1826
1827 } else {
1828 PF_AINC(&rpool->counter, af);
1829 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
1830 }
1831 break;
1832 case PF_POOL_SRCHASH:
1833 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af);
1834 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af);
1835 break;
1836 case PF_POOL_ROUNDROBIN:
1837 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
1838 if (!pfr_pool_get(rpool->cur->addr.p.tbl,
1839 &rpool->tblidx, &rpool->counter,
1840 &raddr, &rmask, af))
1841 goto get_addr;
1842 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af))
1843 goto get_addr;
1844
1845 try_next:
1846 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL)
1847 rpool->cur = TAILQ_FIRST(&rpool->list);
1848 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
1849 rpool->tblidx = -1;
1850 if (pfr_pool_get(rpool->cur->addr.p.tbl,
1851 &rpool->tblidx, &rpool->counter,
1852 &raddr, &rmask, af)) {
1853 /* table contain no address of type 'af' */
1854 if (rpool->cur != acur)
1855 goto try_next;
1856 return (1);
1857 }
1858 } else {
1859 raddr = &rpool->cur->addr.v.a.addr;
1860 rmask = &rpool->cur->addr.v.a.mask;
1861 PF_ACPY(&rpool->counter, raddr, af);
1862 }
1863
1864 get_addr:
1865 PF_ACPY(naddr, &rpool->counter, af);
1866 PF_AINC(&rpool->counter, af);
1867 break;
1868 }
1869
1870 if (pf_status.debug >= PF_DEBUG_MISC &&
1871 (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
1872 printf("pf_map_addr: selected address: ");
1873 pf_print_host(naddr, 0, af);
1874 printf("\n");
1875 }
1876
1877 return (0);
1878}
1879
1880int
1881pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_pool *rpool,
1882 struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport,
1883 struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high)
1884{
1885 struct pf_tree_node key;
1886 struct pf_addr init_addr;
1887 u_int16_t cut;
1888
1889 bzero(&init_addr, sizeof(init_addr));
1890 if (pf_map_addr(af, rpool, saddr, naddr, &init_addr))
1891 return (1);
1892
1893 do {
1894 key.af = af;
1895 key.proto = proto;
1896 PF_ACPY(&key.addr[0], daddr, key.af);
1897 PF_ACPY(&key.addr[1], naddr, key.af);
1898 key.port[0] = dport;
1899
1900 /*
1901 * port search; start random, step;
1902 * similar 2 portloop in in_pcbbind
1903 */
1904 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP)) {
1905 key.port[1] = 0;
1906 if (pf_find_state(&tree_ext_gwy, &key) == NULL)
1907 return (0);
1908 } else if (low == 0 && high == 0) {
1909 key.port[1] = *nport;
1910 if (pf_find_state(&tree_ext_gwy, &key) == NULL) {
1911 return (0);
1912 }
1913 } else if (low == high) {
1914 key.port[1] = htons(low);
1915 if (pf_find_state(&tree_ext_gwy, &key) == NULL) {
1916 *nport = htons(low);
1917 return (0);
1918 }
1919 } else {
1920 u_int16_t tmp;
1921
1922 if (low > high) {
1923 tmp = low;
1924 low = high;
1925 high = tmp;
1926 }
1927 /* low < high */
1928 cut = arc4random() % (1 + high - low) + low;
1929 /* low <= cut <= high */
1930 for (tmp = cut; tmp <= high; ++(tmp)) {
1931 key.port[1] = htons(tmp);
1932 if (pf_find_state(&tree_ext_gwy, &key) ==
1933 NULL) {
1934 *nport = htons(tmp);
1935 return (0);
1936 }
1937 }
1938 for (tmp = cut - 1; tmp >= low; --(tmp)) {
1939 key.port[1] = htons(tmp);
1940 if (pf_find_state(&tree_ext_gwy, &key) ==
1941 NULL) {
1942 *nport = htons(tmp);
1943 return (0);
1944 }
1945 }
1946 }
1947
1948 switch (rpool->opts & PF_POOL_TYPEMASK) {
1949 case PF_POOL_RANDOM:
1950 case PF_POOL_ROUNDROBIN:
1951 if (pf_map_addr(af, rpool, saddr, naddr, &init_addr))
1952 return (1);
1953 break;
1954 case PF_POOL_NONE:
1955 case PF_POOL_SRCHASH:
1956 case PF_POOL_BITMASK:
1957 default:
1958 return (1);
1959 break;
1960 }
1961 } while (! PF_AEQ(&init_addr, naddr, af) );
1962
1963 return (1); /* none available */
1964}
1965
1966struct pf_rule *
1967pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off,
1968 int direction, struct ifnet *ifp, struct pf_addr *saddr, u_int16_t sport,
1969 struct pf_addr *daddr, u_int16_t dport, int rs_num)
1970{
1971 struct pf_rule *r, *rm = NULL, *anchorrule = NULL;
1972 struct pf_ruleset *ruleset = NULL;
1973
1974 r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr);
1975 while (r && rm == NULL) {
1976 struct pf_rule_addr *src = NULL, *dst = NULL;
1977 struct pf_addr_wrap *xdst = NULL;
1978
1979 if (r->action == PF_BINAT && direction == PF_IN) {
1980 src = &r->dst;
1981 if (r->rpool.cur != NULL)
1982 xdst = &r->rpool.cur->addr;
1983 } else {
1984 src = &r->src;
1985 dst = &r->dst;
1986 }
1987
1988 r->evaluations++;
1989 if (r->ifp != NULL && ((r->ifp != ifp && !r->ifnot) ||
1990 (r->ifp == ifp && r->ifnot)))
1991 r = r->skip[PF_SKIP_IFP].ptr;
1992 else if (r->direction && r->direction != direction)
1993 r = r->skip[PF_SKIP_DIR].ptr;
1994 else if (r->af && r->af != pd->af)
1995 r = r->skip[PF_SKIP_AF].ptr;
1996 else if (r->proto && r->proto != pd->proto)
1997 r = r->skip[PF_SKIP_PROTO].ptr;
1998 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af, src->not))
1999 r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR :
2000 PF_SKIP_DST_ADDR].ptr;
2001 else if (src->port_op && !pf_match_port(src->port_op,
2002 src->port[0], src->port[1], sport))
2003 r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT :
2004 PF_SKIP_DST_PORT].ptr;
2005 else if (dst != NULL &&
2006 PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->not))
2007 r = r->skip[PF_SKIP_DST_ADDR].ptr;
2008 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af, 0))
2009 r = TAILQ_NEXT(r, entries);
2010 else if (dst != NULL && dst->port_op &&
2011 !pf_match_port(dst->port_op, dst->port[0],
2012 dst->port[1], dport))
2013 r = r->skip[PF_SKIP_DST_PORT].ptr;
2014 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto !=
2015 IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m,
2016 off, pd->hdr.tcp), r->os_fingerprint)))
2017 r = TAILQ_NEXT(r, entries);
2018 else if (r->anchorname[0] && r->anchor == NULL)
2019 r = TAILQ_NEXT(r, entries);
2020 else if (r->anchor == NULL)
2021 rm = r;
2022 else
2023 PF_STEP_INTO_ANCHOR(r, anchorrule, ruleset, rs_num);
2024 if (r == NULL && anchorrule != NULL)
2025 PF_STEP_OUT_OF_ANCHOR(r, anchorrule, ruleset,
2026 rs_num);
2027 }
2028 if (rm != NULL && (rm->action == PF_NONAT ||
2029 rm->action == PF_NORDR || rm->action == PF_NOBINAT))
2030 return (NULL);
2031 return (rm);
2032}
2033
2034struct pf_rule *
2035pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction,
2036 struct ifnet *ifp,
2037 struct pf_addr *saddr, u_int16_t sport,
2038 struct pf_addr *daddr, u_int16_t dport,
2039 struct pf_addr *naddr, u_int16_t *nport)
2040{
2041 struct pf_rule *r = NULL;
2042
2043 if (direction == PF_OUT) {
2044 r = pf_match_translation(pd, m, off, direction, ifp, saddr,
2045 sport, daddr, dport, PF_RULESET_BINAT);
2046 if (r == NULL)
2047 r = pf_match_translation(pd, m, off, direction, ifp,
2048 saddr, sport, daddr, dport, PF_RULESET_NAT);
2049 } else {
2050 r = pf_match_translation(pd, m, off, direction, ifp, saddr,
2051 sport, daddr, dport, PF_RULESET_RDR);
2052 if (r == NULL)
2053 r = pf_match_translation(pd, m, off, direction, ifp,
2054 saddr, sport, daddr, dport, PF_RULESET_BINAT);
2055 }
2056
2057 if (r != NULL) {
2058 switch (r->action) {
2059 case PF_NONAT:
2060 case PF_NOBINAT:
2061 case PF_NORDR:
2062 return (NULL);
2063 break;
2064 case PF_NAT:
2065 if (pf_get_sport(pd->af, pd->proto, &r->rpool, saddr,
2066 daddr, dport, naddr, nport, r->rpool.proxy_port[0],
2067 r->rpool.proxy_port[1])) {
2068 DPFPRINTF(PF_DEBUG_MISC,
2069 ("pf: NAT proxy port allocation "
2070 "(%u-%u) failed\n",
2071 r->rpool.proxy_port[0],
2072 r->rpool.proxy_port[1]));
2073 return (NULL);
2074 }
2075 break;
2076 case PF_BINAT:
2077 switch (direction) {
2078 case PF_OUT:
2079 if (r->rpool.cur->addr.type ==
2080 PF_ADDR_DYNIFTL &&
2081 r->rpool.cur->addr.p.dyn->undefined)
2082 return (NULL);
2083 else
2084 PF_POOLMASK(naddr,
2085 &r->rpool.cur->addr.v.a.addr,
2086 &r->rpool.cur->addr.v.a.mask,
2087 saddr, pd->af);
2088 break;
2089 case PF_IN:
2090 if (r->src.addr.type == PF_ADDR_DYNIFTL &&
2091 r->src.addr.p.dyn->undefined)
2092 return (NULL);
2093 else
2094 PF_POOLMASK(naddr,
2095 &r->src.addr.v.a.addr,
2096 &r->src.addr.v.a.mask, daddr,
2097 pd->af);
2098 break;
2099 }
2100 break;
2101 case PF_RDR: {
2102 if (pf_map_addr(r->af, &r->rpool, saddr, naddr, NULL))
2103 return (NULL);
2104
2105 if (r->rpool.proxy_port[1]) {
2106 u_int32_t tmp_nport;
2107
2108 tmp_nport = ((ntohs(dport) -
2109 ntohs(r->dst.port[0])) %
2110 (r->rpool.proxy_port[1] -
2111 r->rpool.proxy_port[0] + 1)) +
2112 r->rpool.proxy_port[0];
2113
2114 /* wrap around if necessary */
2115 if (tmp_nport > 65535)
2116 tmp_nport -= 65535;
2117 *nport = htons((u_int16_t)tmp_nport);
2118 } else if (r->rpool.proxy_port[0])
2119 *nport = htons(r->rpool.proxy_port[0]);
2120 break;
2121 }
2122 default:
2123 return (NULL);
2124 break;
2125 }
2126 }
2127
2128 return (r);
2129}
2130
2131int
2132pf_socket_lookup(uid_t *uid, gid_t *gid, int direction, sa_family_t af,
2133 int proto, struct pf_pdesc *pd)
2134{
2135 struct pf_addr *saddr, *daddr;
2136 u_int16_t sport, dport;
2137#ifdef __FreeBSD__
2138 struct inpcbinfo *pi;
2139#else
2140 struct inpcbtable *tb;
2141#endif
2142 struct inpcb *inp;
2143
2144 *uid = UID_MAX;
2145 *gid = GID_MAX;
2146 switch (proto) {
2147 case IPPROTO_TCP:
2148 sport = pd->hdr.tcp->th_sport;
2149 dport = pd->hdr.tcp->th_dport;
2150#ifdef __FreeBSD__
2151 pi = &tcbinfo;
2152#else
2153 tb = &tcbtable;
2154#endif
2155 break;
2156 case IPPROTO_UDP:
2157 sport = pd->hdr.udp->uh_sport;
2158 dport = pd->hdr.udp->uh_dport;
2159#ifdef __FreeBSD__
2160 pi = &udbinfo;
2161#else
2162 tb = &udbtable;
2163#endif
2164 break;
2165 default:
2166 return (0);
2167 }
2168 if (direction == PF_IN) {
2169 saddr = pd->src;
2170 daddr = pd->dst;
2171 } else {
2172 u_int16_t p;
2173
2174 p = sport;
2175 sport = dport;
2176 dport = p;
2177 saddr = pd->dst;
2178 daddr = pd->src;
2179 }
2180 switch(af) {
2181 case AF_INET:
2182#ifdef __FreeBSD__
2183#if (__FreeBSD_version >= 500043)
2184 INP_INFO_RLOCK(pi); /* XXX LOR */
2185#endif
2186 inp = in_pcblookup_hash(pi, saddr->v4, sport, daddr->v4,
2187 dport, 0, NULL);
2188 if (inp == NULL) {
2189 inp = in_pcblookup_hash(pi, saddr->v4, sport,
2190 daddr->v4, dport, INPLOOKUP_WILDCARD, NULL);
2191 if(inp == NULL) {
2192#if (__FreeBSD_version >= 500043)
2193 INP_INFO_RUNLOCK(pi);
2194#endif
2195 return (0);
2196 }
2197 }
2198#else
2199 inp = in_pcbhashlookup(tb, saddr->v4, sport, daddr->v4, dport);
2200 if (inp == NULL) {
2201 inp = in_pcblookup(tb, &saddr->v4, sport, &daddr->v4,
2202 dport, INPLOOKUP_WILDCARD);
2203 if (inp == NULL)
2204 return (0);
2205 }
2206#endif
2207 break;
2208#ifdef INET6
2209 case AF_INET6:
2210#ifdef __FreeBSD__
2211#if (__FreeBSD_version >= 500043)
2212 INP_INFO_RLOCK(pi);
2213#endif
2214 inp = in6_pcblookup_hash(pi, &saddr->v6, sport,
2215 &daddr->v6, dport, 0, NULL);
2216 if (inp == NULL) {
2217 inp = in6_pcblookup_hash(pi, &saddr->v6, sport,
2218 &daddr->v6, dport, INPLOOKUP_WILDCARD, NULL);
2219 if (inp == NULL) {
2220#if (__FreeBSD_version >= 500043)
2221 INP_INFO_RUNLOCK(pi);
2222#endif
2223 return (0);
2224 }
2225 }
2226#else
2227 inp = in6_pcbhashlookup(tb, &saddr->v6, sport, &daddr->v6,
2228 dport);
2229 if (inp == NULL) {
2230 inp = in_pcblookup(tb, &saddr->v6, sport, &daddr->v6,
2231 dport, INPLOOKUP_WILDCARD | INPLOOKUP_IPV6);
2232 if (inp == NULL)
2233 return (0);
2234 }
2235#endif
2236 break;
2237#endif /* INET6 */
2238
2239 default:
2240 return (0);
2241 }
2242#ifdef __FreeBSD__
2243#if (__FreeBSD_version >= 500043)
2244 INP_LOCK(inp);
2245#endif
2246 *uid = inp->inp_socket->so_cred->cr_uid;
2247 *gid = inp->inp_socket->so_cred->cr_groups[0];
2248#if (__FreeBSD_version >= 500043)
2249 INP_UNLOCK(inp);
2250 INP_INFO_RUNLOCK(pi);
2251#endif
2252#else
2253 *uid = inp->inp_socket->so_euid;
2254 *gid = inp->inp_socket->so_egid;
2255#endif
2256 return (1);
2257}
2258
2259u_int8_t
2260pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2261{
2262 int hlen;
2263 u_int8_t hdr[60];
2264 u_int8_t *opt, optlen;
2265 u_int8_t wscale = 0;
2266
2267 hlen = th_off << 2; /* hlen <= sizeof(hdr) */
2268 if (hlen <= sizeof(struct tcphdr))
2269 return (0);
2270 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2271 return (0);
2272 opt = hdr + sizeof(struct tcphdr);
2273 hlen -= sizeof(struct tcphdr);
2274 while (hlen >= 3) {
2275 switch (*opt) {
2276 case TCPOPT_EOL:
2277 case TCPOPT_NOP:
2278 ++opt;
2279 --hlen;
2280 break;
2281 case TCPOPT_WINDOW:
2282 wscale = opt[2];
2283 if (wscale > TCP_MAX_WINSHIFT)
2284 wscale = TCP_MAX_WINSHIFT;
2285 wscale |= PF_WSCALE_FLAG;
2286 /* fallthrough */
2287 default:
2288 optlen = opt[1];
2289 if (optlen < 2)
2290 optlen = 2;
2291 hlen -= optlen;
2292 opt += optlen;
2293 }
2294 }
2295 return (wscale);
2296}
2297
2298u_int16_t
2299pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2300{
2301 int hlen;
2302 u_int8_t hdr[60];
2303 u_int8_t *opt, optlen;
2304 u_int16_t mss = tcp_mssdflt;
2305
2306 hlen = th_off << 2; /* hlen <= sizeof(hdr) */
2307 if (hlen <= sizeof(struct tcphdr))
2308 return (0);
2309 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2310 return (0);
2311 opt = hdr + sizeof(struct tcphdr);
2312 hlen -= sizeof(struct tcphdr);
2313 while (hlen >= TCPOLEN_MAXSEG) {
2314 switch (*opt) {
2315 case TCPOPT_EOL:
2316 case TCPOPT_NOP:
2317 ++opt;
2318 --hlen;
2319 break;
2320 case TCPOPT_MAXSEG:
2321 bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2);
2322 /* fallthrough */
2323 default:
2324 optlen = opt[1];
2325 if (optlen < 2)
2326 optlen = 2;
2327 hlen -= optlen;
2328 opt += optlen;
2329 }
2330 }
2331 return (mss);
2332}
2333
2334u_int16_t
2335pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer)
2336{
2337#ifdef INET
2338 struct sockaddr_in *dst;
2339 struct route ro;
2340#endif /* INET */
2341#ifdef INET6
2342 struct sockaddr_in6 *dst6;
2343 struct route_in6 ro6;
2344#endif /* INET6 */
2345 struct rtentry *rt = NULL;
2346 int hlen = 0; /* make the compiler happy */
2347 u_int16_t mss = tcp_mssdflt;
2348
2349 switch (af) {
2350#ifdef INET
2351 case AF_INET:
2352 hlen = sizeof(struct ip);
2353 bzero(&ro, sizeof(ro));
2354 dst = (struct sockaddr_in *)&ro.ro_dst;
2355 dst->sin_family = AF_INET;
2356 dst->sin_len = sizeof(*dst);
2357 dst->sin_addr = addr->v4;
2358#ifdef __FreeBSD__
2359#ifdef RTF_PRCLONING
2360 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
2361#else /* !RTF_PRCLONING */
2362 rtalloc_ign(&ro, RTF_CLONING);
2363#endif
2364#else /* ! __FreeBSD__ */
2365 rtalloc_noclone(&ro, NO_CLONING);
2366#endif
2367 rt = ro.ro_rt;
2368 break;
2369#endif /* INET */
2370#ifdef INET6
2371 case AF_INET6:
2372 hlen = sizeof(struct ip6_hdr);
2373 bzero(&ro6, sizeof(ro6));
2374 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst;
2375 dst6->sin6_family = AF_INET6;
2376 dst6->sin6_len = sizeof(*dst6);
2377 dst6->sin6_addr = addr->v6;
2378#ifdef __FreeBSD__
2379#ifdef RTF_PRCLONING
2380 rtalloc_ign((struct route *)&ro6,
2381 (RTF_CLONING | RTF_PRCLONING));
2382#else /* !RTF_PRCLONING */
2383 rtalloc_ign((struct route *)&ro6, RTF_CLONING);
2384#endif
2385#else /* ! __FreeBSD__ */
2386 rtalloc_noclone((struct route *)&ro6, NO_CLONING);
2387#endif
2388 rt = ro6.ro_rt;
2389 break;
2390#endif /* INET6 */
2391 }
2392
2393 if (rt && rt->rt_ifp) {
2394 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr);
2395 mss = max(tcp_mssdflt, mss);
2396 RTFREE(rt);
2397 }
2398 mss = min(mss, offer);
2399 mss = max(mss, 64); /* sanity - at least max opt space */
2400 return (mss);
2401}
2402
2403void
2404pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr)
2405{
2406 struct pf_rule *r = s->rule.ptr;
2407
2408 s->rt_ifp = NULL;
2409 if (!r->rt || r->rt == PF_FASTROUTE)
2410 return;
2411 switch (s->af) {
2412#ifdef INET
2413 case AF_INET:
2414 pf_map_addr(AF_INET, &r->rpool, saddr,
2415 &s->rt_addr, NULL);
2416 s->rt_ifp = r->rpool.cur->ifp;
2417 break;
2418#endif /* INET */
2419#ifdef INET6
2420 case AF_INET6:
2421 pf_map_addr(AF_INET6, &r->rpool, saddr,
2422 &s->rt_addr, NULL);
2423 s->rt_ifp = r->rpool.cur->ifp;
2424 break;
2425#endif /* INET6 */
2426 }
2427}
2428
2429int
2430pf_test_tcp(struct pf_rule **rm, struct pf_state **sm, int direction,
2431 struct ifnet *ifp, struct mbuf *m, int ipoff, int off, void *h,
2432 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
2433{
2434 struct pf_rule *nat = NULL, *rdr = NULL;
2435 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
2436 struct pf_addr baddr, naddr;
2437 struct tcphdr *th = pd->hdr.tcp;
2438 u_int16_t bport, nport = 0;
2439 sa_family_t af = pd->af;
2440 int lookup = -1;
2441 uid_t uid;
2442 gid_t gid;
2443 struct pf_rule *r, *a = NULL;
2444 struct pf_ruleset *ruleset = NULL;
2445 u_short reason;
2446 int rewrite = 0;
2447 struct pf_tag *pftag = NULL;
2448 int tag = -1;
2449 u_int16_t mss = tcp_mssdflt;
2450
2451 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
2452
2453 if (direction == PF_OUT) {
2454 bport = nport = th->th_sport;
2455 /* check outgoing packet for BINAT/NAT */
2456 if ((nat = pf_get_translation(pd, m, off, PF_OUT, ifp,
2457 saddr, th->th_sport, daddr, th->th_dport,
2458 &naddr, &nport)) != NULL) {
2459 PF_ACPY(&baddr, saddr, af);
2460 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
2461 &th->th_sum, &naddr, nport, 0, af);
2462 rewrite++;
2463 if (nat->natpass)
2464 r = NULL;
2465 }
2466 } else {
2467 bport = nport = th->th_dport;
2468 /* check incoming packet for BINAT/RDR */
2469 if ((rdr = pf_get_translation(pd, m, off, PF_IN, ifp, saddr,
2470 th->th_sport, daddr, th->th_dport,
2471 &naddr, &nport)) != NULL) {
2472 PF_ACPY(&baddr, daddr, af);
2473 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
2474 &th->th_sum, &naddr, nport, 0, af);
2475 rewrite++;
2476 if (rdr->natpass)
2477 r = NULL;
2478 }
2479 }
2480
2481 while (r != NULL) {
2482 r->evaluations++;
2483 if (r->ifp != NULL && ((r->ifp != ifp && !r->ifnot) ||
2484 (r->ifp == ifp && r->ifnot)))
2485 r = r->skip[PF_SKIP_IFP].ptr;
2486 else if (r->direction && r->direction != direction)
2487 r = r->skip[PF_SKIP_DIR].ptr;
2488 else if (r->af && r->af != af)
2489 r = r->skip[PF_SKIP_AF].ptr;
2490 else if (r->proto && r->proto != IPPROTO_TCP)
2491 r = r->skip[PF_SKIP_PROTO].ptr;
2492 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
2493 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
2494 else if (r->src.port_op && !pf_match_port(r->src.port_op,
2495 r->src.port[0], r->src.port[1], th->th_sport))
2496 r = r->skip[PF_SKIP_SRC_PORT].ptr;
2497 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
2498 r = r->skip[PF_SKIP_DST_ADDR].ptr;
2499 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
2500 r->dst.port[0], r->dst.port[1], th->th_dport))
2501 r = r->skip[PF_SKIP_DST_PORT].ptr;
2502 else if (r->tos && !(r->tos & pd->tos))
2503 r = TAILQ_NEXT(r, entries);
2504 else if (r->rule_flag & PFRULE_FRAGMENT)
2505 r = TAILQ_NEXT(r, entries);
2506 else if ((r->flagset & th->th_flags) != r->flags)
2507 r = TAILQ_NEXT(r, entries);
2508 else if (r->uid.op && (lookup != -1 || (lookup =
2509 pf_socket_lookup(&uid, &gid, direction, af, IPPROTO_TCP,
2510 pd), 1)) &&
2511 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
2512 uid))
2513 r = TAILQ_NEXT(r, entries);
2514 else if (r->gid.op && (lookup != -1 || (lookup =
2515 pf_socket_lookup(&uid, &gid, direction, af, IPPROTO_TCP,
2516 pd), 1)) &&
2517 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
2518 gid))
2519 r = TAILQ_NEXT(r, entries);
2520 else if (r->match_tag &&
2521 !pf_match_tag(m, r, nat, rdr, pftag, &tag))
2522 r = TAILQ_NEXT(r, entries);
2523 else if (r->anchorname[0] && r->anchor == NULL)
2524 r = TAILQ_NEXT(r, entries);
2525 else if (r->os_fingerprint != PF_OSFP_ANY && !pf_osfp_match(
2526 pf_osfp_fingerprint(pd, m, off, th), r->os_fingerprint))
2527 r = TAILQ_NEXT(r, entries);
2528 else {
2529 if (r->tag)
2530 tag = r->tag;
2531 if (r->anchor == NULL) {
2532 *rm = r;
2533 *am = a;
2534 *rsm = ruleset;
2535 if ((*rm)->quick)
2536 break;
2537 r = TAILQ_NEXT(r, entries);
2538 } else
2539 PF_STEP_INTO_ANCHOR(r, a, ruleset,
2540 PF_RULESET_FILTER);
2541 }
2542 if (r == NULL && a != NULL)
2543 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
2544 PF_RULESET_FILTER);
2545 }
2546 r = *rm;
2547 a = *am;
2548 ruleset = *rsm;
2549
2550 r->packets++;
2551 r->bytes += pd->tot_len;
2552 if (a != NULL) {
2553 a->packets++;
2554 a->bytes += pd->tot_len;
2555 }
2556 REASON_SET(&reason, PFRES_MATCH);
2557
2558 if (r->log) {
2559 if (rewrite)
2560 m_copyback(m, off, sizeof(*th), (caddr_t)th);
2561 PFLOG_PACKET(ifp, h, m, af, direction, reason, r, a, ruleset);
2562 }
2563
2564 if ((r->action == PF_DROP) &&
2565 ((r->rule_flag & PFRULE_RETURNRST) ||
2566 (r->rule_flag & PFRULE_RETURNICMP) ||
2567 (r->rule_flag & PFRULE_RETURN))) {
2568 /* undo NAT changes, if they have taken place */
2569 if (nat != NULL) {
2570 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
2571 &th->th_sum, &baddr, bport, 0, af);
2572 rewrite++;
2573 } else if (rdr != NULL) {
2574 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
2575 &th->th_sum, &baddr, bport, 0, af);
2576 rewrite++;
2577 }
2578 if (((r->rule_flag & PFRULE_RETURNRST) ||
2579 (r->rule_flag & PFRULE_RETURN)) &&
2580 !(th->th_flags & TH_RST)) {
2581 u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
2582
2583 if (th->th_flags & TH_SYN)
2584 ack++;
2585 if (th->th_flags & TH_FIN)
2586 ack++;
2587 pf_send_tcp(r, af, pd->dst,
2588 pd->src, th->th_dport, th->th_sport,
2589 ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0,
2590 r->return_ttl);
2591 } else if ((af == AF_INET) && r->return_icmp)
2592 pf_send_icmp(m, r->return_icmp >> 8,
2593 r->return_icmp & 255, af, r);
2594 else if ((af == AF_INET6) && r->return_icmp6)
2595 pf_send_icmp(m, r->return_icmp6 >> 8,
2596 r->return_icmp6 & 255, af, r);
2597 }
2598
2599 if (r->action == PF_DROP)
2600 return (PF_DROP);
2601
2602 if (pf_tag_packet(m, pftag, tag)) {
2603 REASON_SET(&reason, PFRES_MEMORY);
2604 return (PF_DROP);
2605 }
2606
2607 if (r->keep_state || nat != NULL || rdr != NULL ||
2608 (pd->flags & PFDESC_TCP_NORM)) {
2609 /* create new state */
2610 u_int16_t len;
2611 struct pf_state *s = NULL;
2612
2613 len = pd->tot_len - off - (th->th_off << 2);
2614 if (!r->max_states || r->states < r->max_states)
2615 s = pool_get(&pf_state_pl, PR_NOWAIT);
2616 if (s == NULL) {
2617 REASON_SET(&reason, PFRES_MEMORY);
2618 return (PF_DROP);
2619 }
2620 bzero(s, sizeof(*s));
2621 r->states++;
2622 if (a != NULL)
2623 a->states++;
2624 s->rule.ptr = r;
2625 if (nat != NULL)
2626 s->nat_rule.ptr = nat;
2627 else
2628 s->nat_rule.ptr = rdr;
2629 if (s->nat_rule.ptr != NULL)
2630 s->nat_rule.ptr->states++;
2631 s->anchor.ptr = a;
2632 s->allow_opts = r->allow_opts;
2633 s->log = r->log & 2;
2634 s->proto = IPPROTO_TCP;
2635 s->direction = direction;
2636 s->af = af;
2637 if (direction == PF_OUT) {
2638 PF_ACPY(&s->gwy.addr, saddr, af);
2639 s->gwy.port = th->th_sport; /* sport */
2640 PF_ACPY(&s->ext.addr, daddr, af);
2641 s->ext.port = th->th_dport;
2642 if (nat != NULL) {
2643 PF_ACPY(&s->lan.addr, &baddr, af);
2644 s->lan.port = bport;
2645 } else {
2646 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
2647 s->lan.port = s->gwy.port;
2648 }
2649 } else {
2650 PF_ACPY(&s->lan.addr, daddr, af);
2651 s->lan.port = th->th_dport;
2652 PF_ACPY(&s->ext.addr, saddr, af);
2653 s->ext.port = th->th_sport;
2654 if (rdr != NULL) {
2655 PF_ACPY(&s->gwy.addr, &baddr, af);
2656 s->gwy.port = bport;
2657 } else {
2658 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
2659 s->gwy.port = s->lan.port;
2660 }
2661 }
2662
2663 s->src.seqlo = ntohl(th->th_seq);
2664 s->src.seqhi = s->src.seqlo + len + 1;
2665 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
2666 r->keep_state == PF_STATE_MODULATE) {
2667 /* Generate sequence number modulator */
2668 while ((s->src.seqdiff = arc4random()) == 0)
2669 ;
2670 pf_change_a(&th->th_seq, &th->th_sum,
2671 htonl(s->src.seqlo + s->src.seqdiff), 0);
2672 rewrite = 1;
2673 } else
2674 s->src.seqdiff = 0;
2675 if (th->th_flags & TH_SYN) {
2676 s->src.seqhi++;
2677 s->src.wscale = pf_get_wscale(m, off, th->th_off, af);
2678 }
2679 s->src.max_win = MAX(ntohs(th->th_win), 1);
2680 if (s->src.wscale & PF_WSCALE_MASK) {
2681 /* Remove scale factor from initial window */
2682 int win = s->src.max_win;
2683 win += 1 << (s->src.wscale & PF_WSCALE_MASK);
2684 s->src.max_win = (win - 1) >>
2685 (s->src.wscale & PF_WSCALE_MASK);
2686 }
2687 if (th->th_flags & TH_FIN)
2688 s->src.seqhi++;
2689 s->dst.seqlo = 0; /* Haven't seen these yet */
2690 s->dst.seqhi = 1;
2691 s->dst.max_win = 1;
2692 s->dst.seqdiff = 0; /* Defer random generation */
2693 s->src.state = TCPS_SYN_SENT;
2694 s->dst.state = TCPS_CLOSED;
2695#ifdef __FreeBSD__
2696 s->creation = time_second;
2697 s->expire = time_second;
2698#else
2699 s->creation = time.tv_sec;
2700 s->expire = time.tv_sec;
2701#endif
2702 s->timeout = PFTM_TCP_FIRST_PACKET;
2703 s->packets[0] = 1;
2704 s->bytes[0] = pd->tot_len;
2705 pf_set_rt_ifp(s, saddr);
2706
2707 if ((pd->flags & PFDESC_TCP_NORM) && pf_normalize_tcp_init(m,
2708 off, pd, th, &s->src, &s->dst)) {
2709 REASON_SET(&reason, PFRES_MEMORY);
2710 pool_put(&pf_state_pl, s);
2711 return (PF_DROP);
2712 }
2713 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
2714 pf_normalize_tcp_stateful(m, off, pd, &reason, th, &s->src,
2715 &s->dst, &rewrite)) {
2716 pf_normalize_tcp_cleanup(s);
2717 pool_put(&pf_state_pl, s);
2718 return (PF_DROP);
2719 }
2720 if (pf_insert_state(s)) {
2721 pf_normalize_tcp_cleanup(s);
2722 REASON_SET(&reason, PFRES_MEMORY);
2723 pool_put(&pf_state_pl, s);
2724 return (PF_DROP);
2725 } else
2726 *sm = s;
2727 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
2728 r->keep_state == PF_STATE_SYNPROXY) {
2729 s->src.state = PF_TCPS_PROXY_SRC;
2730 if (nat != NULL)
2731 pf_change_ap(saddr, &th->th_sport,
2732 pd->ip_sum, &th->th_sum, &baddr,
2733 bport, 0, af);
2734 else if (rdr != NULL)
2735 pf_change_ap(daddr, &th->th_dport,
2736 pd->ip_sum, &th->th_sum, &baddr,
2737 bport, 0, af);
2738 s->src.seqhi = arc4random();
2739 /* Find mss option */
2740 mss = pf_get_mss(m, off, th->th_off, af);
2741 mss = pf_calc_mss(saddr, af, mss);
2742 mss = pf_calc_mss(daddr, af, mss);
2743 s->src.mss = mss;
2744 pf_send_tcp(r, af, daddr, saddr, th->th_dport,
2745 th->th_sport, s->src.seqhi,
2746 ntohl(th->th_seq) + 1, TH_SYN|TH_ACK, 0, s->src.mss, 0);
2747 return (PF_SYNPROXY_DROP);
2748 }
2749 }
2750
2751 /* copy back packet headers if we performed NAT operations */
2752 if (rewrite)
2753 m_copyback(m, off, sizeof(*th), (caddr_t)th);
2754
2755 return (PF_PASS);
2756}
2757
2758int
2759pf_test_udp(struct pf_rule **rm, struct pf_state **sm, int direction,
2760 struct ifnet *ifp, struct mbuf *m, int ipoff, int off, void *h,
2761 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
2762{
2763 struct pf_rule *nat = NULL, *rdr = NULL;
2764 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
2765 struct pf_addr baddr, naddr;
2766 struct udphdr *uh = pd->hdr.udp;
2767 u_int16_t bport, nport = 0;
2768 sa_family_t af = pd->af;
2769 int lookup = -1;
2770 uid_t uid;
2771 gid_t gid;
2772 struct pf_rule *r, *a = NULL;
2773 struct pf_ruleset *ruleset = NULL;
2774 u_short reason;
2775 int rewrite = 0;
2776 struct pf_tag *pftag = NULL;
2777 int tag = -1;
2778
2779 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
2780
2781 if (direction == PF_OUT) {
2782 bport = nport = uh->uh_sport;
2783 /* check outgoing packet for BINAT/NAT */
2784 if ((nat = pf_get_translation(pd, m, off, PF_OUT, ifp,
2785 saddr, uh->uh_sport, daddr, uh->uh_dport,
2786 &naddr, &nport)) != NULL) {
2787 PF_ACPY(&baddr, saddr, af);
2788 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum,
2789 &uh->uh_sum, &naddr, nport, 1, af);
2790 rewrite++;
2791 if (nat->natpass)
2792 r = NULL;
2793 }
2794 } else {
2795 bport = nport = uh->uh_dport;
2796 /* check incoming packet for BINAT/RDR */
2797 if ((rdr = pf_get_translation(pd, m, off, PF_IN, ifp, saddr,
2798 uh->uh_sport, daddr, uh->uh_dport, &naddr, &nport))
2799 != NULL) {
2800 PF_ACPY(&baddr, daddr, af);
2801 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum,
2802 &uh->uh_sum, &naddr, nport, 1, af);
2803 rewrite++;
2804 if (rdr->natpass)
2805 r = NULL;
2806 }
2807 }
2808
2809 while (r != NULL) {
2810 r->evaluations++;
2811 if (r->ifp != NULL && ((r->ifp != ifp && !r->ifnot) ||
2812 (r->ifp == ifp && r->ifnot)))
2813 r = r->skip[PF_SKIP_IFP].ptr;
2814 else if (r->direction && r->direction != direction)
2815 r = r->skip[PF_SKIP_DIR].ptr;
2816 else if (r->af && r->af != af)
2817 r = r->skip[PF_SKIP_AF].ptr;
2818 else if (r->proto && r->proto != IPPROTO_UDP)
2819 r = r->skip[PF_SKIP_PROTO].ptr;
2820 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
2821 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
2822 else if (r->src.port_op && !pf_match_port(r->src.port_op,
2823 r->src.port[0], r->src.port[1], uh->uh_sport))
2824 r = r->skip[PF_SKIP_SRC_PORT].ptr;
2825 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
2826 r = r->skip[PF_SKIP_DST_ADDR].ptr;
2827 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
2828 r->dst.port[0], r->dst.port[1], uh->uh_dport))
2829 r = r->skip[PF_SKIP_DST_PORT].ptr;
2830 else if (r->tos && !(r->tos & pd->tos))
2831 r = TAILQ_NEXT(r, entries);
2832 else if (r->rule_flag & PFRULE_FRAGMENT)
2833 r = TAILQ_NEXT(r, entries);
2834 else if (r->uid.op && (lookup != -1 || (lookup =
2835 pf_socket_lookup(&uid, &gid, direction, af, IPPROTO_UDP,
2836 pd), 1)) &&
2837 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
2838 uid))
2839 r = TAILQ_NEXT(r, entries);
2840 else if (r->gid.op && (lookup != -1 || (lookup =
2841 pf_socket_lookup(&uid, &gid, direction, af, IPPROTO_UDP,
2842 pd), 1)) &&
2843 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
2844 gid))
2845 r = TAILQ_NEXT(r, entries);
2846 else if (r->match_tag &&
2847 !pf_match_tag(m, r, nat, rdr, pftag, &tag))
2848 r = TAILQ_NEXT(r, entries);
2849 else if (r->anchorname[0] && r->anchor == NULL)
2850 r = TAILQ_NEXT(r, entries);
2851 else if (r->os_fingerprint != PF_OSFP_ANY)
2852 r = TAILQ_NEXT(r, entries);
2853 else {
2854 if (r->tag)
2855 tag = r->tag;
2856 if (r->anchor == NULL) {
2857 *rm = r;
2858 *am = a;
2859 *rsm = ruleset;
2860 if ((*rm)->quick)
2861 break;
2862 r = TAILQ_NEXT(r, entries);
2863 } else
2864 PF_STEP_INTO_ANCHOR(r, a, ruleset,
2865 PF_RULESET_FILTER);
2866 }
2867 if (r == NULL && a != NULL)
2868 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
2869 PF_RULESET_FILTER);
2870 }
2871 r = *rm;
2872 a = *am;
2873 ruleset = *rsm;
2874
2875 r->packets++;
2876 r->bytes += pd->tot_len;
2877 if (a != NULL) {
2878 a->packets++;
2879 a->bytes += pd->tot_len;
2880 }
2881 REASON_SET(&reason, PFRES_MATCH);
2882
2883 if (r->log) {
2884 if (rewrite)
2885 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
2886 PFLOG_PACKET(ifp, h, m, af, direction, reason, r, a, ruleset);
2887 }
2888
2889 if ((r->action == PF_DROP) &&
2890 ((r->rule_flag & PFRULE_RETURNICMP) ||
2891 (r->rule_flag & PFRULE_RETURN))) {
2892 /* undo NAT changes, if they have taken place */
2893 if (nat != NULL) {
2894 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum,
2895 &uh->uh_sum, &baddr, bport, 1, af);
2896 rewrite++;
2897 } else if (rdr != NULL) {
2898 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum,
2899 &uh->uh_sum, &baddr, bport, 1, af);
2900 rewrite++;
2901 }
2902 if ((af == AF_INET) && r->return_icmp)
2903 pf_send_icmp(m, r->return_icmp >> 8,
2904 r->return_icmp & 255, af, r);
2905 else if ((af == AF_INET6) && r->return_icmp6)
2906 pf_send_icmp(m, r->return_icmp6 >> 8,
2907 r->return_icmp6 & 255, af, r);
2908 }
2909
2910 if (r->action == PF_DROP)
2911 return (PF_DROP);
2912
2913 if (pf_tag_packet(m, pftag, tag)) {
2914 REASON_SET(&reason, PFRES_MEMORY);
2915 return (PF_DROP);
2916 }
2917
2918 if (r->keep_state || nat != NULL || rdr != NULL) {
2919 /* create new state */
2920 struct pf_state *s = NULL;
2921
2922 if (!r->max_states || r->states < r->max_states)
2923 s = pool_get(&pf_state_pl, PR_NOWAIT);
2924 if (s == NULL) {
2925 REASON_SET(&reason, PFRES_MEMORY);
2926 return (PF_DROP);
2927 }
2928 bzero(s, sizeof(*s));
2929 r->states++;
2930 if (a != NULL)
2931 a->states++;
2932 s->rule.ptr = r;
2933 if (nat != NULL)
2934 s->nat_rule.ptr = nat;
2935 else
2936 s->nat_rule.ptr = rdr;
2937 if (s->nat_rule.ptr != NULL)
2938 s->nat_rule.ptr->states++;
2939 s->anchor.ptr = a;
2940 s->allow_opts = r->allow_opts;
2941 s->log = r->log & 2;
2942 s->proto = IPPROTO_UDP;
2943 s->direction = direction;
2944 s->af = af;
2945 if (direction == PF_OUT) {
2946 PF_ACPY(&s->gwy.addr, saddr, af);
2947 s->gwy.port = uh->uh_sport;
2948 PF_ACPY(&s->ext.addr, daddr, af);
2949 s->ext.port = uh->uh_dport;
2950 if (nat != NULL) {
2951 PF_ACPY(&s->lan.addr, &baddr, af);
2952 s->lan.port = bport;
2953 } else {
2954 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
2955 s->lan.port = s->gwy.port;
2956 }
2957 } else {
2958 PF_ACPY(&s->lan.addr, daddr, af);
2959 s->lan.port = uh->uh_dport;
2960 PF_ACPY(&s->ext.addr, saddr, af);
2961 s->ext.port = uh->uh_sport;
2962 if (rdr != NULL) {
2963 PF_ACPY(&s->gwy.addr, &baddr, af);
2964 s->gwy.port = bport;
2965 } else {
2966 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
2967 s->gwy.port = s->lan.port;
2968 }
2969 }
2970 s->src.seqlo = 0;
2971 s->src.seqhi = 0;
2972 s->src.seqdiff = 0;
2973 s->src.max_win = 0;
2974 s->src.state = PFUDPS_SINGLE;
2975 s->dst.seqlo = 0;
2976 s->dst.seqhi = 0;
2977 s->dst.seqdiff = 0;
2978 s->dst.max_win = 0;
2979 s->dst.state = PFUDPS_NO_TRAFFIC;
2980#ifdef __FreeBSD__
2981 s->creation = time_second;
2982 s->expire = time_second;
2983#else
2984 s->creation = time.tv_sec;
2985 s->expire = time.tv_sec;
2986#endif
2987 s->timeout = PFTM_UDP_FIRST_PACKET;
2988 s->packets[0] = 1;
2989 s->bytes[0] = pd->tot_len;
2990 pf_set_rt_ifp(s, saddr);
2991 if (pf_insert_state(s)) {
2992 REASON_SET(&reason, PFRES_MEMORY);
2993 pool_put(&pf_state_pl, s);
2994 return (PF_DROP);
2995 } else
2996 *sm = s;
2997 }
2998
2999 /* copy back packet headers if we performed NAT operations */
3000 if (rewrite)
3001 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
3002
3003 return (PF_PASS);
3004}
3005
3006int
3007pf_test_icmp(struct pf_rule **rm, struct pf_state **sm, int direction,
3008 struct ifnet *ifp, struct mbuf *m, int ipoff, int off, void *h,
3009 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
3010{
3011 struct pf_rule *nat = NULL, *rdr = NULL;
3012 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
3013 struct pf_addr baddr, naddr;
3014 struct pf_rule *r, *a = NULL;
3015 struct pf_ruleset *ruleset = NULL;
3016 u_short reason;
3017 u_int16_t icmpid = 0; /* make the compiler happy */
3018 sa_family_t af = pd->af;
3019 u_int8_t icmptype = 0; /* make the compiler happy */
3020 u_int8_t icmpcode = 0; /* make the compiler happy */
3021 int state_icmp = 0;
3022 struct pf_tag *pftag = NULL;
3023 int tag = -1;
3024#ifdef INET6
3025 int rewrite = 0;
3026#endif /* INET6 */
3027
3028 switch (pd->proto) {
3029#ifdef INET
3030 case IPPROTO_ICMP:
3031 icmptype = pd->hdr.icmp->icmp_type;
3032 icmpcode = pd->hdr.icmp->icmp_code;
3033 icmpid = pd->hdr.icmp->icmp_id;
3034
3035 if (icmptype == ICMP_UNREACH ||
3036 icmptype == ICMP_SOURCEQUENCH ||
3037 icmptype == ICMP_REDIRECT ||
3038 icmptype == ICMP_TIMXCEED ||
3039 icmptype == ICMP_PARAMPROB)
3040 state_icmp++;
3041 break;
3042#endif /* INET */
3043#ifdef INET6
3044 case IPPROTO_ICMPV6:
3045 icmptype = pd->hdr.icmp6->icmp6_type;
3046 icmpcode = pd->hdr.icmp6->icmp6_code;
3047 icmpid = pd->hdr.icmp6->icmp6_id;
3048
3049 if (icmptype == ICMP6_DST_UNREACH ||
3050 icmptype == ICMP6_PACKET_TOO_BIG ||
3051 icmptype == ICMP6_TIME_EXCEEDED ||
3052 icmptype == ICMP6_PARAM_PROB)
3053 state_icmp++;
3054 break;
3055#endif /* INET6 */
3056 }
3057
3058 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3059
3060 if (direction == PF_OUT) {
3061 /* check outgoing packet for BINAT/NAT */
3062 if ((nat = pf_get_translation(pd, m, off, PF_OUT, ifp, saddr, 0,
3063 daddr, 0, &naddr, NULL)) != NULL) {
3064 PF_ACPY(&baddr, saddr, af);
3065 switch (af) {
3066#ifdef INET
3067 case AF_INET:
3068 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3069 naddr.v4.s_addr, 0);
3070 break;
3071#endif /* INET */
3072#ifdef INET6
3073 case AF_INET6:
3074 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum,
3075 &naddr, 0);
3076 rewrite++;
3077 break;
3078#endif /* INET6 */
3079 }
3080 if (nat->natpass)
3081 r = NULL;
3082 }
3083 } else {
3084 /* check incoming packet for BINAT/RDR */
3085 if ((rdr = pf_get_translation(pd, m, off, PF_IN, ifp, saddr, 0,
3086 daddr, 0, &naddr, NULL)) != NULL) {
3087 PF_ACPY(&baddr, daddr, af);
3088 switch (af) {
3089#ifdef INET
3090 case AF_INET:
3091 pf_change_a(&daddr->v4.s_addr,
3092 pd->ip_sum, naddr.v4.s_addr, 0);
3093 break;
3094#endif /* INET */
3095#ifdef INET6
3096 case AF_INET6:
3097 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum,
3098 &naddr, 0);
3099 rewrite++;
3100 break;
3101#endif /* INET6 */
3102 }
3103 if (rdr->natpass)
3104 r = NULL;
3105 }
3106 }
3107
3108 while (r != NULL) {
3109 r->evaluations++;
3110 if (r->ifp != NULL && ((r->ifp != ifp && !r->ifnot) ||
3111 (r->ifp == ifp && r->ifnot)))
3112 r = r->skip[PF_SKIP_IFP].ptr;
3113 else if (r->direction && r->direction != direction)
3114 r = r->skip[PF_SKIP_DIR].ptr;
3115 else if (r->af && r->af != af)
3116 r = r->skip[PF_SKIP_AF].ptr;
3117 else if (r->proto && r->proto != pd->proto)
3118 r = r->skip[PF_SKIP_PROTO].ptr;
3119 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
3120 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3121 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
3122 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3123 else if (r->type && r->type != icmptype + 1)
3124 r = TAILQ_NEXT(r, entries);
3125 else if (r->code && r->code != icmpcode + 1)
3126 r = TAILQ_NEXT(r, entries);
3127 else if (r->tos && !(r->tos & pd->tos))
3128 r = TAILQ_NEXT(r, entries);
3129 else if (r->rule_flag & PFRULE_FRAGMENT)
3130 r = TAILQ_NEXT(r, entries);
3131 else if (r->match_tag &&
3132 !pf_match_tag(m, r, nat, rdr, pftag, &tag))
3133 r = TAILQ_NEXT(r, entries);
3134 else if (r->anchorname[0] && r->anchor == NULL)
3135 r = TAILQ_NEXT(r, entries);
3136 else if (r->os_fingerprint != PF_OSFP_ANY)
3137 r = TAILQ_NEXT(r, entries);
3138 else {
3139 if (r->tag)
3140 tag = r->tag;
3141 if (r->anchor == NULL) {
3142 *rm = r;
3143 *am = a;
3144 *rsm = ruleset;
3145 if ((*rm)->quick)
3146 break;
3147 r = TAILQ_NEXT(r, entries);
3148 } else
3149 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3150 PF_RULESET_FILTER);
3151 }
3152 if (r == NULL && a != NULL)
3153 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3154 PF_RULESET_FILTER);
3155 }
3156 r = *rm;
3157 a = *am;
3158 ruleset = *rsm;
3159
3160 r->packets++;
3161 r->bytes += pd->tot_len;
3162 if (a != NULL) {
3163 a->packets++;
3164 a->bytes += pd->tot_len;
3165 }
3166 REASON_SET(&reason, PFRES_MATCH);
3167
3168 if (r->log) {
3169#ifdef INET6
3170 if (rewrite)
3171 m_copyback(m, off, sizeof(struct icmp6_hdr),
3172 (caddr_t)pd->hdr.icmp6);
3173#endif /* INET6 */
3174 PFLOG_PACKET(ifp, h, m, af, direction, reason, r, a, ruleset);
3175 }
3176
3177 if (r->action != PF_PASS)
3178 return (PF_DROP);
3179
3180 if (pf_tag_packet(m, pftag, tag)) {
3181 REASON_SET(&reason, PFRES_MEMORY);
3182 return (PF_DROP);
3183 }
3184
3185 if (!state_icmp && (r->keep_state ||
3186 nat != NULL || rdr != NULL)) {
3187 /* create new state */
3188 struct pf_state *s = NULL;
3189
3190 if (!r->max_states || r->states < r->max_states)
3191 s = pool_get(&pf_state_pl, PR_NOWAIT);
3192 if (s == NULL) {
3193 REASON_SET(&reason, PFRES_MEMORY);
3194 return (PF_DROP);
3195 }
3196 bzero(s, sizeof(*s));
3197 r->states++;
3198 if (a != NULL)
3199 a->states++;
3200 s->rule.ptr = r;
3201 if (nat != NULL)
3202 s->nat_rule.ptr = nat;
3203 else
3204 s->nat_rule.ptr = rdr;
3205 if (s->nat_rule.ptr != NULL)
3206 s->nat_rule.ptr->states++;
3207 s->anchor.ptr = a;
3208 s->allow_opts = r->allow_opts;
3209 s->log = r->log & 2;
3210 s->proto = pd->proto;
3211 s->direction = direction;
3212 s->af = af;
3213 if (direction == PF_OUT) {
3214 PF_ACPY(&s->gwy.addr, saddr, af);
3215 s->gwy.port = icmpid;
3216 PF_ACPY(&s->ext.addr, daddr, af);
3217 s->ext.port = icmpid;
3218 if (nat != NULL)
3219 PF_ACPY(&s->lan.addr, &baddr, af);
3220 else
3221 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3222 s->lan.port = icmpid;
3223 } else {
3224 PF_ACPY(&s->lan.addr, daddr, af);
3225 s->lan.port = icmpid;
3226 PF_ACPY(&s->ext.addr, saddr, af);
3227 s->ext.port = icmpid;
3228 if (rdr != NULL)
3229 PF_ACPY(&s->gwy.addr, &baddr, af);
3230 else
3231 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3232 s->gwy.port = icmpid;
3233 }
3234
3235 s->src.seqlo = 0;
3236 s->src.seqhi = 0;
3237 s->src.seqdiff = 0;
3238 s->src.max_win = 0;
3239 s->src.state = 0;
3240 s->dst.seqlo = 0;
3241 s->dst.seqhi = 0;
3242 s->dst.seqdiff = 0;
3243 s->dst.max_win = 0;
3244 s->dst.state = 0;
3245#ifdef __FreeBSD__
3246 s->creation = time_second;
3247 s->expire = time_second;
3248#else
3249 s->creation = time.tv_sec;
3250 s->expire = time.tv_sec;
3251#endif
3252 s->timeout = PFTM_ICMP_FIRST_PACKET;
3253 s->packets[0] = 1;
3254 s->bytes[0] = pd->tot_len;
3255 pf_set_rt_ifp(s, saddr);
3256 if (pf_insert_state(s)) {
3257 REASON_SET(&reason, PFRES_MEMORY);
3258 pool_put(&pf_state_pl, s);
3259 return (PF_DROP);
3260 } else
3261 *sm = s;
3262 }
3263
3264#ifdef INET6
3265 /* copy back packet headers if we performed IPv6 NAT operations */
3266 if (rewrite)
3267 m_copyback(m, off, sizeof(struct icmp6_hdr),
3268 (caddr_t)pd->hdr.icmp6);
3269#endif /* INET6 */
3270
3271 return (PF_PASS);
3272}
3273
3274int
3275pf_test_other(struct pf_rule **rm, struct pf_state **sm, int direction,
3276 struct ifnet *ifp, struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
3277 struct pf_rule **am, struct pf_ruleset **rsm)
3278{
3279 struct pf_rule *nat = NULL, *rdr = NULL;
3280 struct pf_rule *r, *a = NULL;
3281 struct pf_ruleset *ruleset = NULL;
3282 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
3283 struct pf_addr baddr, naddr;
3284 sa_family_t af = pd->af;
3285 u_short reason;
3286 struct pf_tag *pftag = NULL;
3287 int tag = -1;
3288
3289 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3290
3291 if (direction == PF_OUT) {
3292 /* check outgoing packet for BINAT/NAT */
3293 if ((nat = pf_get_translation(pd, m, off, PF_OUT, ifp, saddr, 0,
3294 daddr, 0, &naddr, NULL)) != NULL) {
3295 PF_ACPY(&baddr, saddr, af);
3296 switch (af) {
3297#ifdef INET
3298 case AF_INET:
3299 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3300 naddr.v4.s_addr, 0);
3301 break;
3302#endif /* INET */
3303#ifdef INET6
3304 case AF_INET6:
3305 PF_ACPY(saddr, &naddr, af);
3306 break;
3307#endif /* INET6 */
3308 }
3309 if (nat->natpass)
3310 r = NULL;
3311 }
3312 } else {
3313 /* check incoming packet for BINAT/RDR */
3314 if ((rdr = pf_get_translation(pd, m, off, PF_IN, ifp, saddr, 0,
3315 daddr, 0, &naddr, NULL)) != NULL) {
3316 PF_ACPY(&baddr, daddr, af);
3317 switch (af) {
3318#ifdef INET
3319 case AF_INET:
3320 pf_change_a(&daddr->v4.s_addr,
3321 pd->ip_sum, naddr.v4.s_addr, 0);
3322 break;
3323#endif /* INET */
3324#ifdef INET6
3325 case AF_INET6:
3326 PF_ACPY(daddr, &naddr, af);
3327 break;
3328#endif /* INET6 */
3329 }
3330 if (rdr->natpass)
3331 r = NULL;
3332 }
3333 }
3334
3335 while (r != NULL) {
3336 r->evaluations++;
3337 if (r->ifp != NULL && ((r->ifp != ifp && !r->ifnot) ||
3338 (r->ifp == ifp && r->ifnot)))
3339 r = r->skip[PF_SKIP_IFP].ptr;
3340 else if (r->direction && r->direction != direction)
3341 r = r->skip[PF_SKIP_DIR].ptr;
3342 else if (r->af && r->af != af)
3343 r = r->skip[PF_SKIP_AF].ptr;
3344 else if (r->proto && r->proto != pd->proto)
3345 r = r->skip[PF_SKIP_PROTO].ptr;
3346 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, r->src.not))
3347 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3348 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, r->dst.not))
3349 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3350 else if (r->tos && !(r->tos & pd->tos))
3351 r = TAILQ_NEXT(r, entries);
3352 else if (r->rule_flag & PFRULE_FRAGMENT)
3353 r = TAILQ_NEXT(r, entries);
3354 else if (r->match_tag &&
3355 !pf_match_tag(m, r, nat, rdr, pftag, &tag))
3356 r = TAILQ_NEXT(r, entries);
3357 else if (r->anchorname[0] && r->anchor == NULL)
3358 r = TAILQ_NEXT(r, entries);
3359 else if (r->os_fingerprint != PF_OSFP_ANY)
3360 r = TAILQ_NEXT(r, entries);
3361 else {
3362 if (r->tag)
3363 tag = r->tag;
3364 if (r->anchor == NULL) {
3365 *rm = r;
3366 *am = a;
3367 *rsm = ruleset;
3368 if ((*rm)->quick)
3369 break;
3370 r = TAILQ_NEXT(r, entries);
3371 } else
3372 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3373 PF_RULESET_FILTER);
3374 }
3375 if (r == NULL && a != NULL)
3376 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3377 PF_RULESET_FILTER);
3378 }
3379 r = *rm;
3380 a = *am;
3381 ruleset = *rsm;
3382
3383 r->packets++;
3384 r->bytes += pd->tot_len;
3385 if (a != NULL) {
3386 a->packets++;
3387 a->bytes += pd->tot_len;
3388 }
3389 REASON_SET(&reason, PFRES_MATCH);
3390 if (r->log)
3391 PFLOG_PACKET(ifp, h, m, af, direction, reason, r, a, ruleset);
3392
3393 if ((r->action == PF_DROP) &&
3394 ((r->rule_flag & PFRULE_RETURNICMP) ||
3395 (r->rule_flag & PFRULE_RETURN))) {
3396 struct pf_addr *a = NULL;
3397
3398 if (nat != NULL)
3399 a = saddr;
3400 else if (rdr != NULL)
3401 a = daddr;
3402 if (a != NULL) {
3403 switch (af) {
3404#ifdef INET
3405 case AF_INET:
3406 pf_change_a(&a->v4.s_addr, pd->ip_sum,
3407 baddr.v4.s_addr, 0);
3408 break;
3409#endif /* INET */
3410#ifdef INET6
3411 case AF_INET6:
3412 PF_ACPY(a, &baddr, af);
3413 break;
3414#endif /* INET6 */
3415 }
3416 }
3417 if ((af == AF_INET) && r->return_icmp)
3418 pf_send_icmp(m, r->return_icmp >> 8,
3419 r->return_icmp & 255, af, r);
3420 else if ((af == AF_INET6) && r->return_icmp6)
3421 pf_send_icmp(m, r->return_icmp6 >> 8,
3422 r->return_icmp6 & 255, af, r);
3423 }
3424
3425 if (r->action != PF_PASS)
3426 return (PF_DROP);
3427
3428 if (pf_tag_packet(m, pftag, tag)) {
3429 REASON_SET(&reason, PFRES_MEMORY);
3430 return (PF_DROP);
3431 }
3432
3433 if (r->keep_state || nat != NULL || rdr != NULL) {
3434 /* create new state */
3435 struct pf_state *s = NULL;
3436
3437 if (!r->max_states || r->states < r->max_states)
3438 s = pool_get(&pf_state_pl, PR_NOWAIT);
3439 if (s == NULL) {
3440 REASON_SET(&reason, PFRES_MEMORY);
3441 return (PF_DROP);
3442 }
3443 bzero(s, sizeof(*s));
3444 r->states++;
3445 if (a != NULL)
3446 a->states++;
3447 s->rule.ptr = r;
3448 if (nat != NULL)
3449 s->nat_rule.ptr = nat;
3450 else
3451 s->nat_rule.ptr = rdr;
3452 if (s->nat_rule.ptr != NULL)
3453 s->nat_rule.ptr->states++;
3454 s->anchor.ptr = a;
3455 s->allow_opts = r->allow_opts;
3456 s->log = r->log & 2;
3457 s->proto = pd->proto;
3458 s->direction = direction;
3459 s->af = af;
3460 if (direction == PF_OUT) {
3461 PF_ACPY(&s->gwy.addr, saddr, af);
3462 s->gwy.port = 0;
3463 PF_ACPY(&s->ext.addr, daddr, af);
3464 s->ext.port = 0;
3465 if (nat != NULL)
3466 PF_ACPY(&s->lan.addr, &baddr, af);
3467 else
3468 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3469 s->lan.port = 0;
3470 } else {
3471 PF_ACPY(&s->lan.addr, daddr, af);
3472 s->lan.port = 0;
3473 PF_ACPY(&s->ext.addr, saddr, af);
3474 s->ext.port = 0;
3475 if (rdr != NULL)
3476 PF_ACPY(&s->gwy.addr, &baddr, af);
3477 else
3478 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3479 s->gwy.port = 0;
3480 }
3481 s->src.seqlo = 0;
3482 s->src.seqhi = 0;
3483 s->src.seqdiff = 0;
3484 s->src.max_win = 0;
3485 s->src.state = PFOTHERS_SINGLE;
3486 s->dst.seqlo = 0;
3487 s->dst.seqhi = 0;
3488 s->dst.seqdiff = 0;
3489 s->dst.max_win = 0;
3490 s->dst.state = PFOTHERS_NO_TRAFFIC;
3491#ifdef __FreeBSD__
3492 s->creation = time_second;
3493 s->expire = time_second;
3494#else
3495 s->creation = time.tv_sec;
3496 s->expire = time.tv_sec;
3497#endif
3498 s->timeout = PFTM_OTHER_FIRST_PACKET;
3499 s->packets[0] = 1;
3500 s->bytes[0] = pd->tot_len;
3501 pf_set_rt_ifp(s, saddr);
3502 if (pf_insert_state(s)) {
3503 REASON_SET(&reason, PFRES_MEMORY);
3504 if (r->log)
3505 PFLOG_PACKET(ifp, h, m, af, direction, reason,
3506 r, a, ruleset);
3507 pool_put(&pf_state_pl, s);
3508 return (PF_DROP);
3509 } else
3510 *sm = s;
3511 }
3512
3513 return (PF_PASS);
3514}
3515
3516int
3517pf_test_fragment(struct pf_rule **rm, int direction, struct ifnet *ifp,
3518 struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am,
3519 struct pf_ruleset **rsm)
3520{
3521 struct pf_rule *r, *a = NULL;
3522 struct pf_ruleset *ruleset = NULL;
3523 sa_family_t af = pd->af;
3524 u_short reason;
3525 struct pf_tag *pftag = NULL;
3526 int tag = -1;
3527
3528 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3529 while (r != NULL) {
3530 r->evaluations++;
3531 if (r->ifp != NULL && ((r->ifp != ifp && !r->ifnot) ||
3532 (r->ifp == ifp && r->ifnot)))
3533 r = r->skip[PF_SKIP_IFP].ptr;
3534 else if (r->direction && r->direction != direction)
3535 r = r->skip[PF_SKIP_DIR].ptr;
3536 else if (r->af && r->af != af)
3537 r = r->skip[PF_SKIP_AF].ptr;
3538 else if (r->proto && r->proto != pd->proto)
3539 r = r->skip[PF_SKIP_PROTO].ptr;
3540 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, r->src.not))
3541 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3542 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, r->dst.not))
3543 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3544 else if (r->tos && !(r->tos & pd->tos))
3545 r = TAILQ_NEXT(r, entries);
3546 else if (r->src.port_op || r->dst.port_op ||
3547 r->flagset || r->type || r->code ||
3548 r->os_fingerprint != PF_OSFP_ANY)
3549 r = TAILQ_NEXT(r, entries);
3550 else if (r->match_tag &&
3551 !pf_match_tag(m, r, NULL, NULL, pftag, &tag))
3552 r = TAILQ_NEXT(r, entries);
3553 else if (r->anchorname[0] && r->anchor == NULL)
3554 r = TAILQ_NEXT(r, entries);
3555 else {
3556 if (r->anchor == NULL) {
3557 *rm = r;
3558 *am = a;
3559 *rsm = ruleset;
3560 if ((*rm)->quick)
3561 break;
3562 r = TAILQ_NEXT(r, entries);
3563 } else
3564 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3565 PF_RULESET_FILTER);
3566 }
3567 if (r == NULL && a != NULL)
3568 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3569 PF_RULESET_FILTER);
3570 }
3571 r = *rm;
3572 a = *am;
3573 ruleset = *rsm;
3574
3575 r->packets++;
3576 r->bytes += pd->tot_len;
3577 if (a != NULL) {
3578 a->packets++;
3579 a->bytes += pd->tot_len;
3580 }
3581 REASON_SET(&reason, PFRES_MATCH);
3582 if (r->log)
3583 PFLOG_PACKET(ifp, h, m, af, direction, reason, r, a, ruleset);
3584
3585 if (r->action != PF_PASS)
3586 return (PF_DROP);
3587
3588 if (pf_tag_packet(m, pftag, tag)) {
3589 REASON_SET(&reason, PFRES_MEMORY);
3590 return (PF_DROP);
3591 }
3592
3593 return (PF_PASS);
3594}
3595
3596int
3597pf_test_state_tcp(struct pf_state **state, int direction, struct ifnet *ifp,
3598 struct mbuf *m, int ipoff, int off, void *h, struct pf_pdesc *pd,
3599 u_short *reason)
3600{
3601 struct pf_tree_node key;
3602 struct tcphdr *th = pd->hdr.tcp;
3603 u_int16_t win = ntohs(th->th_win);
3604 u_int32_t ack, end, seq;
3605 u_int8_t sws, dws;
3606 int ackskew, dirndx;
3607 int copyback = 0;
3608 struct pf_state_peer *src, *dst;
3609
3610 key.af = pd->af;
3611 key.proto = IPPROTO_TCP;
3612 PF_ACPY(&key.addr[0], pd->src, key.af);
3613 PF_ACPY(&key.addr[1], pd->dst, key.af);
3614 key.port[0] = th->th_sport;
3615 key.port[1] = th->th_dport;
3616
3617 STATE_LOOKUP();
3618
3619 if (direction == (*state)->direction) {
3620 src = &(*state)->src;
3621 dst = &(*state)->dst;
3622 dirndx = 0;
3623 } else {
3624 src = &(*state)->dst;
3625 dst = &(*state)->src;
3626 dirndx = 1;
3627 }
3628
3629 if ((*state)->src.state == PF_TCPS_PROXY_SRC) {
3630 if (direction != (*state)->direction)
3631 return (PF_SYNPROXY_DROP);
3632 if (th->th_flags & TH_SYN) {
3633 if (ntohl(th->th_seq) != (*state)->src.seqlo)
3634 return (PF_DROP);
3635 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
3636 pd->src, th->th_dport, th->th_sport,
3637 (*state)->src.seqhi, ntohl(th->th_seq) + 1,
3638 TH_SYN|TH_ACK, 0, (*state)->src.mss, 0);
3639 return (PF_SYNPROXY_DROP);
3640 } else if (!(th->th_flags & TH_ACK) ||
3641 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
3642 (ntohl(th->th_seq) != (*state)->src.seqlo + 1))
3643 return (PF_DROP);
3644 else
3645 (*state)->src.state = PF_TCPS_PROXY_DST;
3646 }
3647 if ((*state)->src.state == PF_TCPS_PROXY_DST) {
3648 struct pf_state_host *src, *dst;
3649
3650 if (direction == PF_OUT) {
3651 src = &(*state)->gwy;
3652 dst = &(*state)->ext;
3653 } else {
3654 src = &(*state)->ext;
3655 dst = &(*state)->lan;
3656 }
3657 if (direction == (*state)->direction) {
3658 if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) ||
3659 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
3660 (ntohl(th->th_seq) != (*state)->src.seqlo + 1))
3661 return (PF_DROP);
3662 (*state)->src.max_win = MAX(ntohs(th->th_win), 1);
3663 if ((*state)->dst.seqhi == 1)
3664 (*state)->dst.seqhi = arc4random();
3665 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
3666 &dst->addr, src->port, dst->port,
3667 (*state)->dst.seqhi, 0, TH_SYN, 0, (*state)->src.mss, 0);
3668 return (PF_SYNPROXY_DROP);
3669 } else if (((th->th_flags & (TH_SYN|TH_ACK)) !=
3670 (TH_SYN|TH_ACK)) ||
3671 (ntohl(th->th_ack) != (*state)->dst.seqhi + 1))
3672 return (PF_DROP);
3673 else {
3674 (*state)->dst.max_win = MAX(ntohs(th->th_win), 1);
3675 (*state)->dst.seqlo = ntohl(th->th_seq);
3676 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
3677 pd->src, th->th_dport, th->th_sport,
3678 ntohl(th->th_ack), ntohl(th->th_seq) + 1,
3679 TH_ACK, (*state)->src.max_win, 0, 0);
3680 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
3681 &dst->addr, src->port, dst->port,
3682 (*state)->src.seqhi + 1, (*state)->src.seqlo + 1,
3683 TH_ACK, (*state)->dst.max_win, 0, 0);
3684 (*state)->src.seqdiff = (*state)->dst.seqhi -
3685 (*state)->src.seqlo;
3686 (*state)->dst.seqdiff = (*state)->src.seqhi -
3687 (*state)->dst.seqlo;
3688 (*state)->src.seqhi = (*state)->src.seqlo +
3689 (*state)->src.max_win;
3690 (*state)->dst.seqhi = (*state)->dst.seqlo +
3691 (*state)->dst.max_win;
3692 (*state)->src.wscale = (*state)->dst.wscale = 0;
3693 (*state)->src.state = (*state)->dst.state =
3694 TCPS_ESTABLISHED;
3695 return (PF_SYNPROXY_DROP);
3696 }
3697 }
3698
3699 if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) {
3700 sws = src->wscale & PF_WSCALE_MASK;
3701 dws = dst->wscale & PF_WSCALE_MASK;
3702 } else
3703 sws = dws = 0;
3704
3705 /*
3706 * Sequence tracking algorithm from Guido van Rooij's paper:
3707 * http://www.madison-gurkha.com/publications/tcp_filtering/
3708 * tcp_filtering.ps
3709 */
3710
3711 seq = ntohl(th->th_seq);
3712 if (src->seqlo == 0) {
3713 /* First packet from this end. Set its state */
3714
3715 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) &&
3716 src->scrub == NULL) {
3717 if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) {
3718 REASON_SET(reason, PFRES_MEMORY);
3719 return (PF_DROP);
3720 }
3721 }
3722
3723 /* Deferred generation of sequence number modulator */
3724 if (dst->seqdiff && !src->seqdiff) {
3725 while ((src->seqdiff = arc4random()) == 0)
3726 ;
3727 ack = ntohl(th->th_ack) - dst->seqdiff;
3728 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3729 src->seqdiff), 0);
3730 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3731 copyback = 1;
3732 } else {
3733 ack = ntohl(th->th_ack);
3734 }
3735
3736 end = seq + pd->p_len;
3737 if (th->th_flags & TH_SYN) {
3738 end++;
3739 if (dst->wscale & PF_WSCALE_FLAG) {
3740 src->wscale = pf_get_wscale(m, off, th->th_off,
3741 pd->af);
3742 if (src->wscale & PF_WSCALE_FLAG) {
3743 /* Remove scale factor from initial
3744 * window */
3745 sws = src->wscale & PF_WSCALE_MASK;
3746 win = ((u_int32_t)win + (1 << sws) - 1)
3747 >> sws;
3748 dws = dst->wscale & PF_WSCALE_MASK;
3749 } else {
3750 /* fixup other window */
3751 dst->max_win <<= dst->wscale &
3752 PF_WSCALE_MASK;
3753 /* in case of a retrans SYN|ACK */
3754 dst->wscale = 0;
3755 }
3756 }
3757 }
3758 if (th->th_flags & TH_FIN)
3759 end++;
3760
3761 src->seqlo = seq;
3762 if (src->state < TCPS_SYN_SENT)
3763 src->state = TCPS_SYN_SENT;
3764
3765 /*
3766 * May need to slide the window (seqhi may have been set by
3767 * the crappy stack check or if we picked up the connection
3768 * after establishment)
3769 */
3770 if (src->seqhi == 1 ||
3771 SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi))
3772 src->seqhi = end + MAX(1, dst->max_win << dws);
3773 if (win > src->max_win)
3774 src->max_win = win;
3775
3776 } else {
3777 ack = ntohl(th->th_ack) - dst->seqdiff;
3778 if (src->seqdiff) {
3779 /* Modulate sequence numbers */
3780 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3781 src->seqdiff), 0);
3782 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3783 copyback = 1;
3784 }
3785 end = seq + pd->p_len;
3786 if (th->th_flags & TH_SYN)
3787 end++;
3788 if (th->th_flags & TH_FIN)
3789 end++;
3790 }
3791
3792 if ((th->th_flags & TH_ACK) == 0) {
3793 /* Let it pass through the ack skew check */
3794 ack = dst->seqlo;
3795 } else if ((ack == 0 &&
3796 (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) ||
3797 /* broken tcp stacks do not set ack */
3798 (dst->state < TCPS_SYN_SENT)) {
3799 /*
3800 * Many stacks (ours included) will set the ACK number in an
3801 * FIN|ACK if the SYN times out -- no sequence to ACK.
3802 */
3803 ack = dst->seqlo;
3804 }
3805
3806 if (seq == end) {
3807 /* Ease sequencing restrictions on no data packets */
3808 seq = src->seqlo;
3809 end = seq;
3810 }
3811
3812 ackskew = dst->seqlo - ack;
3813
3814#define MAXACKWINDOW (0xffff + 1500) /* 1500 is an arbitrary fudge factor */
3815 if (SEQ_GEQ(src->seqhi, end) &&
3816 /* Last octet inside other's window space */
3817 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) &&
3818 /* Retrans: not more than one window back */
3819 (ackskew >= -MAXACKWINDOW) &&
3820 /* Acking not more than one reassembled fragment backwards */
3821 (ackskew <= (MAXACKWINDOW << sws))) {
3822 /* Acking not more than one window forward */
3823
3824 (*state)->packets[dirndx]++;
3825 (*state)->bytes[dirndx] += pd->tot_len;
3826
3827 /* update max window */
3828 if (src->max_win < win)
3829 src->max_win = win;
3830 /* synchronize sequencing */
3831 if (SEQ_GT(end, src->seqlo))
3832 src->seqlo = end;
3833 /* slide the window of what the other end can send */
3834 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
3835 dst->seqhi = ack + MAX((win << sws), 1);
3836
3837
3838 /* update states */
3839 if (th->th_flags & TH_SYN)
3840 if (src->state < TCPS_SYN_SENT)
3841 src->state = TCPS_SYN_SENT;
3842 if (th->th_flags & TH_FIN)
3843 if (src->state < TCPS_CLOSING)
3844 src->state = TCPS_CLOSING;
3845 if (th->th_flags & TH_ACK) {
3846 if (dst->state == TCPS_SYN_SENT)
3847 dst->state = TCPS_ESTABLISHED;
3848 else if (dst->state == TCPS_CLOSING)
3849 dst->state = TCPS_FIN_WAIT_2;
3850 }
3851 if (th->th_flags & TH_RST)
3852 src->state = dst->state = TCPS_TIME_WAIT;
3853
3854 /* update expire time */
3855#ifdef __FreeBSD__
3856 (*state)->expire = time_second;
3857#else
3858 (*state)->expire = time.tv_sec;
3859#endif
3860 if (src->state >= TCPS_FIN_WAIT_2 &&
3861 dst->state >= TCPS_FIN_WAIT_2)
3862 (*state)->timeout = PFTM_TCP_CLOSED;
3863 else if (src->state >= TCPS_FIN_WAIT_2 ||
3864 dst->state >= TCPS_FIN_WAIT_2)
3865 (*state)->timeout = PFTM_TCP_FIN_WAIT;
3866 else if (src->state < TCPS_ESTABLISHED ||
3867 dst->state < TCPS_ESTABLISHED)
3868 (*state)->timeout = PFTM_TCP_OPENING;
3869 else if (src->state >= TCPS_CLOSING ||
3870 dst->state >= TCPS_CLOSING)
3871 (*state)->timeout = PFTM_TCP_CLOSING;
3872 else
3873 (*state)->timeout = PFTM_TCP_ESTABLISHED;
3874
3875 /* Fall through to PASS packet */
3876
3877 } else if ((dst->state < TCPS_SYN_SENT ||
3878 dst->state >= TCPS_FIN_WAIT_2 ||
3879 src->state >= TCPS_FIN_WAIT_2) &&
3880 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) &&
3881 /* Within a window forward of the originating packet */
3882 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) {
3883 /* Within a window backward of the originating packet */
3884
3885 /*
3886 * This currently handles three situations:
3887 * 1) Stupid stacks will shotgun SYNs before their peer
3888 * replies.
3889 * 2) When PF catches an already established stream (the
3890 * firewall rebooted, the state table was flushed, routes
3891 * changed...)
3892 * 3) Packets get funky immediately after the connection
3893 * closes (this should catch Solaris spurious ACK|FINs
3894 * that web servers like to spew after a close)
3895 *
3896 * This must be a little more careful than the above code
3897 * since packet floods will also be caught here. We don't
3898 * update the TTL here to mitigate the damage of a packet
3899 * flood and so the same code can handle awkward establishment
3900 * and a loosened connection close.
3901 * In the establishment case, a correct peer response will
3902 * validate the connection, go through the normal state code
3903 * and keep updating the state TTL.
3904 */
3905
3906 if (pf_status.debug >= PF_DEBUG_MISC) {
3907 printf("pf: loose state match: ");
3908 pf_print_state(*state);
3909 pf_print_flags(th->th_flags);
3910 printf(" seq=%u ack=%u len=%u ackskew=%d pkts=%d:%d\n",
3911 seq, ack, pd->p_len, ackskew,
3912 (*state)->packets[0], (*state)->packets[1]);
3913 }
3914
3915 (*state)->packets[dirndx]++;
3916 (*state)->bytes[dirndx] += pd->tot_len;
3917
3918 /* update max window */
3919 if (src->max_win < win)
3920 src->max_win = win;
3921 /* synchronize sequencing */
3922 if (SEQ_GT(end, src->seqlo))
3923 src->seqlo = end;
3924 /* slide the window of what the other end can send */
3925 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
3926 dst->seqhi = ack + MAX((win << sws), 1);
3927
3928 /*
3929 * Cannot set dst->seqhi here since this could be a shotgunned
3930 * SYN and not an already established connection.
3931 */
3932
3933 if (th->th_flags & TH_FIN)
3934 if (src->state < TCPS_CLOSING)
3935 src->state = TCPS_CLOSING;
3936 if (th->th_flags & TH_RST)
3937 src->state = dst->state = TCPS_TIME_WAIT;
3938
3939 /* Fall through to PASS packet */
3940
3941 } else {
3942 if ((*state)->dst.state == TCPS_SYN_SENT &&
3943 (*state)->src.state == TCPS_SYN_SENT) {
3944 /* Send RST for state mismatches during handshake */
3945 if (!(th->th_flags & TH_RST)) {
3946 u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
3947
3948 if (th->th_flags & TH_SYN)
3949 ack++;
3950 if (th->th_flags & TH_FIN)
3951 ack++;
3952 pf_send_tcp((*state)->rule.ptr, pd->af,
3953 pd->dst, pd->src, th->th_dport,
3954 th->th_sport, ntohl(th->th_ack), ack,
3955 TH_RST|TH_ACK, 0, 0,
3956 (*state)->rule.ptr->return_ttl);
3957 }
3958 src->seqlo = 0;
3959 src->seqhi = 1;
3960 src->max_win = 1;
3961 } else if (pf_status.debug >= PF_DEBUG_MISC) {
3962 printf("pf: BAD state: ");
3963 pf_print_state(*state);
3964 pf_print_flags(th->th_flags);
3965 printf(" seq=%u ack=%u len=%u ackskew=%d pkts=%d:%d "
3966 "dir=%s,%s\n", seq, ack, pd->p_len, ackskew,
3967 (*state)->packets[0], (*state)->packets[1],
3968 direction == PF_IN ? "in" : "out",
3969 direction == (*state)->direction ? "fwd" : "rev");
3970 printf("pf: State failure on: %c %c %c %c | %c %c\n",
3971 SEQ_GEQ(src->seqhi, end) ? ' ' : '1',
3972 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ?
3973 ' ': '2',
3974 (ackskew >= -MAXACKWINDOW) ? ' ' : '3',
3975 (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4',
3976 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5',
3977 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6');
3978 }
3979 return (PF_DROP);
3980 }
3981
3982 if (dst->scrub || src->scrub) {
3983 if (pf_normalize_tcp_stateful(m, off, pd, reason, th, src, dst,
3984 ©back))
3985 return (PF_DROP);
3986 }
3987
3988 /* Any packets which have gotten here are to be passed */
3989
3990 /* translate source/destination address, if necessary */
3991 if (STATE_TRANSLATE(*state)) {
3992 if (direction == PF_OUT)
3993 pf_change_ap(pd->src, &th->th_sport, pd->ip_sum,
3994 &th->th_sum, &(*state)->gwy.addr,
3995 (*state)->gwy.port, 0, pd->af);
3996 else
3997 pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum,
3998 &th->th_sum, &(*state)->lan.addr,
3999 (*state)->lan.port, 0, pd->af);
4000 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4001 } else if (copyback) {
4002 /* Copyback sequence modulation or stateful scrub changes */
4003 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4004 }
4005
4006 (*state)->rule.ptr->packets++;
4007 (*state)->rule.ptr->bytes += pd->tot_len;
4008 if ((*state)->nat_rule.ptr != NULL) {
4009 (*state)->nat_rule.ptr->packets++;
4010 (*state)->nat_rule.ptr->bytes += pd->tot_len;
4011 }
4012 if ((*state)->anchor.ptr != NULL) {
4013 (*state)->anchor.ptr->packets++;
4014 (*state)->anchor.ptr->bytes += pd->tot_len;
4015 }
4016 return (PF_PASS);
4017}
4018
4019int
4020pf_test_state_udp(struct pf_state **state, int direction, struct ifnet *ifp,
4021 struct mbuf *m, int ipoff, int off, void *h, struct pf_pdesc *pd)
4022{
4023 struct pf_state_peer *src, *dst;
4024 struct pf_tree_node key;
4025 struct udphdr *uh = pd->hdr.udp;
4026 int dirndx;
4027
4028 key.af = pd->af;
4029 key.proto = IPPROTO_UDP;
4030 PF_ACPY(&key.addr[0], pd->src, key.af);
4031 PF_ACPY(&key.addr[1], pd->dst, key.af);
4032 key.port[0] = uh->uh_sport;
4033 key.port[1] = uh->uh_dport;
4034
4035 STATE_LOOKUP();
4036
4037 if (direction == (*state)->direction) {
4038 src = &(*state)->src;
4039 dst = &(*state)->dst;
4040 dirndx = 0;
4041 } else {
4042 src = &(*state)->dst;
4043 dst = &(*state)->src;
4044 dirndx = 1;
4045 }
4046
4047 (*state)->packets[dirndx]++;
4048 (*state)->bytes[dirndx] += pd->tot_len;
4049
4050 /* update states */
4051 if (src->state < PFUDPS_SINGLE)
4052 src->state = PFUDPS_SINGLE;
4053 if (dst->state == PFUDPS_SINGLE)
4054 dst->state = PFUDPS_MULTIPLE;
4055
4056 /* update expire time */
4057#ifdef __FreeBSD__
4058 (*state)->expire = time_second;
4059#else
4060 (*state)->expire = time.tv_sec;
4061#endif
4062 if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE)
4063 (*state)->timeout = PFTM_UDP_MULTIPLE;
4064 else
4065 (*state)->timeout = PFTM_UDP_SINGLE;
4066
4067 /* translate source/destination address, if necessary */
4068 if (STATE_TRANSLATE(*state)) {
4069 if (direction == PF_OUT)
4070 pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum,
4071 &uh->uh_sum, &(*state)->gwy.addr,
4072 (*state)->gwy.port, 1, pd->af);
4073 else
4074 pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum,
4075 &uh->uh_sum, &(*state)->lan.addr,
4076 (*state)->lan.port, 1, pd->af);
4077 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
4078 }
4079
4080 (*state)->rule.ptr->packets++;
4081 (*state)->rule.ptr->bytes += pd->tot_len;
4082 if ((*state)->nat_rule.ptr != NULL) {
4083 (*state)->nat_rule.ptr->packets++;
4084 (*state)->nat_rule.ptr->bytes += pd->tot_len;
4085 }
4086 if ((*state)->anchor.ptr != NULL) {
4087 (*state)->anchor.ptr->packets++;
4088 (*state)->anchor.ptr->bytes += pd->tot_len;
4089 }
4090 return (PF_PASS);
4091}
4092
4093int
4094pf_test_state_icmp(struct pf_state **state, int direction, struct ifnet *ifp,
4095 struct mbuf *m, int ipoff, int off, void *h, struct pf_pdesc *pd)
4096{
4097 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
4098 u_int16_t icmpid = 0; /* make the compiler happy */
4099 u_int16_t *icmpsum = NULL; /* make the compiler happy */
4100 u_int8_t icmptype = 0; /* make the compiler happy */
4101 int state_icmp = 0, dirndx;
4102
4103 switch (pd->proto) {
4104#ifdef INET
4105 case IPPROTO_ICMP:
4106 icmptype = pd->hdr.icmp->icmp_type;
4107 icmpid = pd->hdr.icmp->icmp_id;
4108 icmpsum = &pd->hdr.icmp->icmp_cksum;
4109
4110 if (icmptype == ICMP_UNREACH ||
4111 icmptype == ICMP_SOURCEQUENCH ||
4112 icmptype == ICMP_REDIRECT ||
4113 icmptype == ICMP_TIMXCEED ||
4114 icmptype == ICMP_PARAMPROB)
4115 state_icmp++;
4116 break;
4117#endif /* INET */
4118#ifdef INET6
4119 case IPPROTO_ICMPV6:
4120 icmptype = pd->hdr.icmp6->icmp6_type;
4121 icmpid = pd->hdr.icmp6->icmp6_id;
4122 icmpsum = &pd->hdr.icmp6->icmp6_cksum;
4123
4124 if (icmptype == ICMP6_DST_UNREACH ||
4125 icmptype == ICMP6_PACKET_TOO_BIG ||
4126 icmptype == ICMP6_TIME_EXCEEDED ||
4127 icmptype == ICMP6_PARAM_PROB)
4128 state_icmp++;
4129 break;
4130#endif /* INET6 */
4131 }
4132
4133 if (!state_icmp) {
4134
4135 /*
4136 * ICMP query/reply message not related to a TCP/UDP packet.
4137 * Search for an ICMP state.
4138 */
4139 struct pf_tree_node key;
4140
4141 key.af = pd->af;
4142 key.proto = pd->proto;
4143 PF_ACPY(&key.addr[0], saddr, key.af);
4144 PF_ACPY(&key.addr[1], daddr, key.af);
4145 key.port[0] = icmpid;
4146 key.port[1] = icmpid;
4147
4148 STATE_LOOKUP();
4149
4150 dirndx = (direction == (*state)->direction) ? 0 : 1;
4151 (*state)->packets[dirndx]++;
4152 (*state)->bytes[dirndx] += pd->tot_len;
4153#ifdef __FreeBSD__
4154 (*state)->expire = time_second;
4155#else
4156 (*state)->expire = time.tv_sec;
4157#endif
4158 (*state)->timeout = PFTM_ICMP_ERROR_REPLY;
4159
4160 /* translate source/destination address, if necessary */
4161 if (PF_ANEQ(&(*state)->lan.addr, &(*state)->gwy.addr, pd->af)) {
4162 if (direction == PF_OUT) {
4163 switch (pd->af) {
4164#ifdef INET
4165 case AF_INET:
4166 pf_change_a(&saddr->v4.s_addr,
4167 pd->ip_sum,
4168 (*state)->gwy.addr.v4.s_addr, 0);
4169 break;
4170#endif /* INET */
4171#ifdef INET6
4172 case AF_INET6:
4173 pf_change_a6(saddr,
4174 &pd->hdr.icmp6->icmp6_cksum,
4175 &(*state)->gwy.addr, 0);
4176 m_copyback(m, off,
4177 sizeof(struct icmp6_hdr),
4178 (caddr_t)pd->hdr.icmp6);
4179 break;
4180#endif /* INET6 */
4181 }
4182 } else {
4183 switch (pd->af) {
4184#ifdef INET
4185 case AF_INET:
4186 pf_change_a(&daddr->v4.s_addr,
4187 pd->ip_sum,
4188 (*state)->lan.addr.v4.s_addr, 0);
4189 break;
4190#endif /* INET */
4191#ifdef INET6
4192 case AF_INET6:
4193 pf_change_a6(daddr,
4194 &pd->hdr.icmp6->icmp6_cksum,
4195 &(*state)->lan.addr, 0);
4196 m_copyback(m, off,
4197 sizeof(struct icmp6_hdr),
4198 (caddr_t)pd->hdr.icmp6);
4199 break;
4200#endif /* INET6 */
4201 }
4202 }
4203 }
4204
4205 return (PF_PASS);
4206
4207 } else {
4208 /*
4209 * ICMP error message in response to a TCP/UDP packet.
4210 * Extract the inner TCP/UDP header and search for that state.
4211 */
4212
4213 struct pf_pdesc pd2;
4214#ifdef INET
4215 struct ip h2;
4216#endif /* INET */
4217#ifdef INET6
4218 struct ip6_hdr h2_6;
4219 int terminal = 0;
4220#endif /* INET6 */
4221 int ipoff2 = 0; /* make the compiler happy */
4222 int off2 = 0; /* make the compiler happy */
4223
4224 pd2.af = pd->af;
4225 switch (pd->af) {
4226#ifdef INET
4227 case AF_INET:
4228 /* offset of h2 in mbuf chain */
4229 ipoff2 = off + ICMP_MINLEN;
4230
4231 if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2),
4232 NULL, NULL, pd2.af)) {
4233 DPFPRINTF(PF_DEBUG_MISC,
4234 ("pf: ICMP error message too short "
4235 "(ip)\n"));
4236 return (PF_DROP);
4237 }
4238 /*
4239 * ICMP error messages don't refer to non-first
4240 * fragments
4241 */
4242 if (h2.ip_off & htons(IP_OFFMASK))
4243 return (PF_DROP);
4244
4245 /* offset of protocol header that follows h2 */
4246 off2 = ipoff2 + (h2.ip_hl << 2);
4247
4248 pd2.proto = h2.ip_p;
4249 pd2.src = (struct pf_addr *)&h2.ip_src;
4250 pd2.dst = (struct pf_addr *)&h2.ip_dst;
4251 pd2.ip_sum = &h2.ip_sum;
4252 break;
4253#endif /* INET */
4254#ifdef INET6
4255 case AF_INET6:
4256 ipoff2 = off + sizeof(struct icmp6_hdr);
4257
4258 if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6),
4259 NULL, NULL, pd2.af)) {
4260 DPFPRINTF(PF_DEBUG_MISC,
4261 ("pf: ICMP error message too short "
4262 "(ip6)\n"));
4263 return (PF_DROP);
4264 }
4265 pd2.proto = h2_6.ip6_nxt;
4266 pd2.src = (struct pf_addr *)&h2_6.ip6_src;
4267 pd2.dst = (struct pf_addr *)&h2_6.ip6_dst;
4268 pd2.ip_sum = NULL;
4269 off2 = ipoff2 + sizeof(h2_6);
4270 do {
4271 switch (pd2.proto) {
4272 case IPPROTO_FRAGMENT:
4273 /*
4274 * ICMPv6 error messages for
4275 * non-first fragments
4276 */
4277 return (PF_DROP);
4278 case IPPROTO_AH:
4279 case IPPROTO_HOPOPTS:
4280 case IPPROTO_ROUTING:
4281 case IPPROTO_DSTOPTS: {
4282 /* get next header and header length */
4283 struct ip6_ext opt6;
4284
4285 if (!pf_pull_hdr(m, off2, &opt6,
4286 sizeof(opt6), NULL, NULL, pd2.af)) {
4287 DPFPRINTF(PF_DEBUG_MISC,
4288 ("pf: ICMPv6 short opt\n"));
4289 return (PF_DROP);
4290 }
4291 if (pd2.proto == IPPROTO_AH)
4292 off2 += (opt6.ip6e_len + 2) * 4;
4293 else
4294 off2 += (opt6.ip6e_len + 1) * 8;
4295 pd2.proto = opt6.ip6e_nxt;
4296 /* goto the next header */
4297 break;
4298 }
4299 default:
4300 terminal++;
4301 break;
4302 }
4303 } while (!terminal);
4304 break;
4305#endif /* INET6 */
4306 }
4307
4308 switch (pd2.proto) {
4309 case IPPROTO_TCP: {
4310 struct tcphdr th;
4311 u_int32_t seq;
4312 struct pf_tree_node key;
4313 struct pf_state_peer *src, *dst;
4314 u_int8_t dws;
4315 int copyback = 0;
4316
4317 /*
4318 * Only the first 8 bytes of the TCP header can be
4319 * expected. Don't access any TCP header fields after
4320 * th_seq, an ackskew test is not possible.
4321 */
4322 if (!pf_pull_hdr(m, off2, &th, 8, NULL, NULL, pd2.af)) {
4323 DPFPRINTF(PF_DEBUG_MISC,
4324 ("pf: ICMP error message too short "
4325 "(tcp)\n"));
4326 return (PF_DROP);
4327 }
4328
4329 key.af = pd2.af;
4330 key.proto = IPPROTO_TCP;
4331 PF_ACPY(&key.addr[0], pd2.dst, pd2.af);
4332 key.port[0] = th.th_dport;
4333 PF_ACPY(&key.addr[1], pd2.src, pd2.af);
4334 key.port[1] = th.th_sport;
4335
4336 STATE_LOOKUP();
4337
4338 if (direction == (*state)->direction) {
4339 src = &(*state)->dst;
4340 dst = &(*state)->src;
4341 } else {
4342 src = &(*state)->src;
4343 dst = &(*state)->dst;
4344 }
4345
4346 if (src->wscale && dst->wscale && !(th.th_flags & TH_SYN))
4347 dws = dst->wscale & PF_WSCALE_MASK;
4348 else
4349 dws = 0;
4350
4351 /* Demodulate sequence number */
4352 seq = ntohl(th.th_seq) - src->seqdiff;
4353 if (src->seqdiff) {
4354 pf_change_a(&th.th_seq, icmpsum,
4355 htonl(seq), 0);
4356 copyback = 1;
4357 }
4358
4359 if (!SEQ_GEQ(src->seqhi, seq) ||
4360 !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws))) {
4361 if (pf_status.debug >= PF_DEBUG_MISC) {
4362 printf("pf: BAD ICMP %d:%d ",
4363 icmptype, pd->hdr.icmp->icmp_code);
4364 pf_print_host(pd->src, 0, pd->af);
4365 printf(" -> ");
4366 pf_print_host(pd->dst, 0, pd->af);
4367 printf(" state: ");
4368 pf_print_state(*state);
4369 printf(" seq=%u\n", seq);
4370 }
4371 return (PF_DROP);
4372 }
4373
4374 if (STATE_TRANSLATE(*state)) {
4375 if (direction == PF_IN) {
4376 pf_change_icmp(pd2.src, &th.th_sport,
4377 daddr, &(*state)->lan.addr,
4378 (*state)->lan.port, NULL,
4379 pd2.ip_sum, icmpsum,
4380 pd->ip_sum, 0, pd2.af);
4381 } else {
4382 pf_change_icmp(pd2.dst, &th.th_dport,
4383 saddr, &(*state)->gwy.addr,
4384 (*state)->gwy.port, NULL,
4385 pd2.ip_sum, icmpsum,
4386 pd->ip_sum, 0, pd2.af);
4387 }
4388 copyback = 1;
4389 }
4390
4391 if (copyback) {
4392 switch (pd2.af) {
4393#ifdef INET
4394 case AF_INET:
4395 m_copyback(m, off, ICMP_MINLEN,
4396 (caddr_t)pd->hdr.icmp);
4397 m_copyback(m, ipoff2, sizeof(h2),
4398 (caddr_t)&h2);
4399 break;
4400#endif /* INET */
4401#ifdef INET6
4402 case AF_INET6:
4403 m_copyback(m, off,
4404 sizeof(struct icmp6_hdr),
4405 (caddr_t)pd->hdr.icmp6);
4406 m_copyback(m, ipoff2, sizeof(h2_6),
4407 (caddr_t)&h2_6);
4408 break;
4409#endif /* INET6 */
4410 }
4411 m_copyback(m, off2, 8, (caddr_t)&th);
4412 }
4413
4414 return (PF_PASS);
4415 break;
4416 }
4417 case IPPROTO_UDP: {
4418 struct udphdr uh;
4419 struct pf_tree_node key;
4420
4421 if (!pf_pull_hdr(m, off2, &uh, sizeof(uh),
4422 NULL, NULL, pd2.af)) {
4423 DPFPRINTF(PF_DEBUG_MISC,
4424 ("pf: ICMP error message too short "
4425 "(udp)\n"));
4426 return (PF_DROP);
4427 }
4428
4429 key.af = pd2.af;
4430 key.proto = IPPROTO_UDP;
4431 PF_ACPY(&key.addr[0], pd2.dst, pd2.af);
4432 key.port[0] = uh.uh_dport;
4433 PF_ACPY(&key.addr[1], pd2.src, pd2.af);
4434 key.port[1] = uh.uh_sport;
4435
4436 STATE_LOOKUP();
4437
4438 if (STATE_TRANSLATE(*state)) {
4439 if (direction == PF_IN) {
4440 pf_change_icmp(pd2.src, &uh.uh_sport,
4441 daddr, &(*state)->lan.addr,
4442 (*state)->lan.port, &uh.uh_sum,
4443 pd2.ip_sum, icmpsum,
4444 pd->ip_sum, 1, pd2.af);
4445 } else {
4446 pf_change_icmp(pd2.dst, &uh.uh_dport,
4447 saddr, &(*state)->gwy.addr,
4448 (*state)->gwy.port, &uh.uh_sum,
4449 pd2.ip_sum, icmpsum,
4450 pd->ip_sum, 1, pd2.af);
4451 }
4452 switch (pd2.af) {
4453#ifdef INET
4454 case AF_INET:
4455 m_copyback(m, off, ICMP_MINLEN,
4456 (caddr_t)pd->hdr.icmp);
4457 m_copyback(m, ipoff2, sizeof(h2),
4458 (caddr_t)&h2);
4459 break;
4460#endif /* INET */
4461#ifdef INET6
4462 case AF_INET6:
4463 m_copyback(m, off,
4464 sizeof(struct icmp6_hdr),
4465 (caddr_t)pd->hdr.icmp6);
4466 m_copyback(m, ipoff2, sizeof(h2_6),
4467 (caddr_t)&h2_6);
4468 break;
4469#endif /* INET6 */
4470 }
4471 m_copyback(m, off2, sizeof(uh),
4472 (caddr_t)&uh);
4473 }
4474
4475 return (PF_PASS);
4476 break;
4477 }
4478#ifdef INET
4479 case IPPROTO_ICMP: {
4480 struct icmp iih;
4481 struct pf_tree_node key;
4482
4483 if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN,
4484 NULL, NULL, pd2.af)) {
4485 DPFPRINTF(PF_DEBUG_MISC,
4486 ("pf: ICMP error message too short i"
4487 "(icmp)\n"));
4488 return (PF_DROP);
4489 }
4490
4491 key.af = pd2.af;
4492 key.proto = IPPROTO_ICMP;
4493 PF_ACPY(&key.addr[0], pd2.dst, pd2.af);
4494 key.port[0] = iih.icmp_id;
4495 PF_ACPY(&key.addr[1], pd2.src, pd2.af);
4496 key.port[1] = iih.icmp_id;
4497
4498 STATE_LOOKUP();
4499
4500 if (STATE_TRANSLATE(*state)) {
4501 if (direction == PF_IN) {
4502 pf_change_icmp(pd2.src, &iih.icmp_id,
4503 daddr, &(*state)->lan.addr,
4504 (*state)->lan.port, NULL,
4505 pd2.ip_sum, icmpsum,
4506 pd->ip_sum, 0, AF_INET);
4507 } else {
4508 pf_change_icmp(pd2.dst, &iih.icmp_id,
4509 saddr, &(*state)->gwy.addr,
4510 (*state)->gwy.port, NULL,
4511 pd2.ip_sum, icmpsum,
4512 pd->ip_sum, 0, AF_INET);
4513 }
4514 m_copyback(m, off, ICMP_MINLEN,
4515 (caddr_t)pd->hdr.icmp);
4516 m_copyback(m, ipoff2, sizeof(h2),
4517 (caddr_t)&h2);
4518 m_copyback(m, off2, ICMP_MINLEN,
4519 (caddr_t)&iih);
4520 }
4521
4522 return (PF_PASS);
4523 break;
4524 }
4525#endif /* INET */
4526#ifdef INET6
4527 case IPPROTO_ICMPV6: {
4528 struct icmp6_hdr iih;
4529 struct pf_tree_node key;
4530
4531 if (!pf_pull_hdr(m, off2, &iih,
4532 sizeof(struct icmp6_hdr), NULL, NULL, pd2.af)) {
4533 DPFPRINTF(PF_DEBUG_MISC,
4534 ("pf: ICMP error message too short "
4535 "(icmp6)\n"));
4536 return (PF_DROP);
4537 }
4538
4539 key.af = pd2.af;
4540 key.proto = IPPROTO_ICMPV6;
4541 PF_ACPY(&key.addr[0], pd2.dst, pd2.af);
4542 key.port[0] = iih.icmp6_id;
4543 PF_ACPY(&key.addr[1], pd2.src, pd2.af);
4544 key.port[1] = iih.icmp6_id;
4545
4546 STATE_LOOKUP();
4547
4548 if (STATE_TRANSLATE(*state)) {
4549 if (direction == PF_IN) {
4550 pf_change_icmp(pd2.src, &iih.icmp6_id,
4551 daddr, &(*state)->lan.addr,
4552 (*state)->lan.port, NULL,
4553 pd2.ip_sum, icmpsum,
4554 pd->ip_sum, 0, AF_INET6);
4555 } else {
4556 pf_change_icmp(pd2.dst, &iih.icmp6_id,
4557 saddr, &(*state)->gwy.addr,
4558 (*state)->gwy.port, NULL,
4559 pd2.ip_sum, icmpsum,
4560 pd->ip_sum, 0, AF_INET6);
4561 }
4562 m_copyback(m, off, sizeof(struct icmp6_hdr),
4563 (caddr_t)pd->hdr.icmp6);
4564 m_copyback(m, ipoff2, sizeof(h2_6),
4565 (caddr_t)&h2_6);
4566 m_copyback(m, off2, sizeof(struct icmp6_hdr),
4567 (caddr_t)&iih);
4568 }
4569
4570 return (PF_PASS);
4571 break;
4572 }
4573#endif /* INET6 */
4574 default: {
4575 struct pf_tree_node key;
4576
4577 key.af = pd2.af;
4578 key.proto = pd2.proto;
4579 PF_ACPY(&key.addr[0], pd2.dst, pd2.af);
4580 key.port[0] = 0;
4581 PF_ACPY(&key.addr[1], pd2.src, pd2.af);
4582 key.port[1] = 0;
4583
4584 STATE_LOOKUP();
4585
4586 if (STATE_TRANSLATE(*state)) {
4587 if (direction == PF_IN) {
4588 pf_change_icmp(pd2.src, NULL,
4589 daddr, &(*state)->lan.addr,
4590 0, NULL,
4591 pd2.ip_sum, icmpsum,
4592 pd->ip_sum, 0, pd2.af);
4593 } else {
4594 pf_change_icmp(pd2.dst, NULL,
4595 saddr, &(*state)->gwy.addr,
4596 0, NULL,
4597 pd2.ip_sum, icmpsum,
4598 pd->ip_sum, 0, pd2.af);
4599 }
4600 switch (pd2.af) {
4601#ifdef INET
4602 case AF_INET:
4603 m_copyback(m, off, ICMP_MINLEN,
4604 (caddr_t)pd->hdr.icmp);
4605 m_copyback(m, ipoff2, sizeof(h2),
4606 (caddr_t)&h2);
4607 break;
4608#endif /* INET */
4609#ifdef INET6
4610 case AF_INET6:
4611 m_copyback(m, off,
4612 sizeof(struct icmp6_hdr),
4613 (caddr_t)pd->hdr.icmp6);
4614 m_copyback(m, ipoff2, sizeof(h2_6),
4615 (caddr_t)&h2_6);
4616 break;
4617#endif /* INET6 */
4618 }
4619 }
4620
4621 return (PF_PASS);
4622 break;
4623 }
4624 }
4625 }
4626}
4627
4628int
4629pf_test_state_other(struct pf_state **state, int direction, struct ifnet *ifp,
4630 struct pf_pdesc *pd)
4631{
4632 struct pf_state_peer *src, *dst;
4633 struct pf_tree_node key;
4634 int dirndx;
4635
4636 key.af = pd->af;
4637 key.proto = pd->proto;
4638 PF_ACPY(&key.addr[0], pd->src, key.af);
4639 PF_ACPY(&key.addr[1], pd->dst, key.af);
4640 key.port[0] = 0;
4641 key.port[1] = 0;
4642
4643 STATE_LOOKUP();
4644
4645 if (direction == (*state)->direction) {
4646 src = &(*state)->src;
4647 dst = &(*state)->dst;
4648 dirndx = 0;
4649 } else {
4650 src = &(*state)->dst;
4651 dst = &(*state)->src;
4652 dirndx = 1;
4653 }
4654
4655 (*state)->packets[dirndx]++;
4656 (*state)->bytes[dirndx] += pd->tot_len;
4657
4658 /* update states */
4659 if (src->state < PFOTHERS_SINGLE)
4660 src->state = PFOTHERS_SINGLE;
4661 if (dst->state == PFOTHERS_SINGLE)
4662 dst->state = PFOTHERS_MULTIPLE;
4663
4664 /* update expire time */
4665#ifdef __FreeBSD__
4666 (*state)->expire = time_second;
4667#else
4668 (*state)->expire = time.tv_sec;
4669#endif
4670 if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE)
4671 (*state)->timeout = PFTM_OTHER_MULTIPLE;
4672 else
4673 (*state)->timeout = PFTM_OTHER_SINGLE;
4674
4675 /* translate source/destination address, if necessary */
4676 if (STATE_TRANSLATE(*state)) {
4677 if (direction == PF_OUT)
4678 switch (pd->af) {
4679#ifdef INET
4680 case AF_INET:
4681 pf_change_a(&pd->src->v4.s_addr,
4682 pd->ip_sum, (*state)->gwy.addr.v4.s_addr,
4683 0);
4684 break;
4685#endif /* INET */
4686#ifdef INET6
4687 case AF_INET6:
4688 PF_ACPY(pd->src, &(*state)->gwy.addr, pd->af);
4689 break;
4690#endif /* INET6 */
4691 }
4692 else
4693 switch (pd->af) {
4694#ifdef INET
4695 case AF_INET:
4696 pf_change_a(&pd->dst->v4.s_addr,
4697 pd->ip_sum, (*state)->lan.addr.v4.s_addr,
4698 0);
4699 break;
4700#endif /* INET */
4701#ifdef INET6
4702 case AF_INET6:
4703 PF_ACPY(pd->dst, &(*state)->lan.addr, pd->af);
4704 break;
4705#endif /* INET6 */
4706 }
4707 }
4708
4709 (*state)->rule.ptr->packets++;
4710 (*state)->rule.ptr->bytes += pd->tot_len;
4711 if ((*state)->nat_rule.ptr != NULL) {
4712 (*state)->nat_rule.ptr->packets++;
4713 (*state)->nat_rule.ptr->bytes += pd->tot_len;
4714 }
4715 if ((*state)->anchor.ptr != NULL) {
4716 (*state)->anchor.ptr->packets++;
4717 (*state)->anchor.ptr->bytes += pd->tot_len;
4718 }
4719 return (PF_PASS);
4720}
4721
4722/*
4723 * ipoff and off are measured from the start of the mbuf chain.
4724 * h must be at "ipoff" on the mbuf chain.
4725 */
4726void *
4727pf_pull_hdr(struct mbuf *m, int off, void *p, int len,
4728 u_short *actionp, u_short *reasonp, sa_family_t af)
4729{
4730 switch (af) {
4731#ifdef INET
4732 case AF_INET: {
4733 struct ip *h = mtod(m, struct ip *);
4734 u_int16_t fragoff = (ntohs(h->ip_off) & IP_OFFMASK) << 3;
4735
4736 if (fragoff) {
4737 if (fragoff >= len)
4738 ACTION_SET(actionp, PF_PASS);
4739 else {
4740 ACTION_SET(actionp, PF_DROP);
4741 REASON_SET(reasonp, PFRES_FRAG);
4742 }
4743 return (NULL);
4744 }
4745 if (m->m_pkthdr.len < off + len || ntohs(h->ip_len) < off + len) {
4746 ACTION_SET(actionp, PF_DROP);
4747 REASON_SET(reasonp, PFRES_SHORT);
4748 return (NULL);
4749 }
4750 break;
4751 }
4752#endif /* INET */
4753#ifdef INET6
4754 case AF_INET6: {
4755 struct ip6_hdr *h = mtod(m, struct ip6_hdr *);
4756
4757 if (m->m_pkthdr.len < off + len ||
4758 (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) <
4759 (unsigned)(off + len)) {
4760 ACTION_SET(actionp, PF_DROP);
4761 REASON_SET(reasonp, PFRES_SHORT);
4762 return (NULL);
4763 }
4764 break;
4765 }
4766#endif /* INET6 */
4767 }
4768 m_copydata(m, off, len, p);
4769 return (p);
4770}
4771
4772int
4773pf_routable(struct pf_addr *addr, sa_family_t af)
4774{
4775 struct sockaddr_in *dst;
4776 struct route ro;
4777 int ret = 0;
4778
4779 bzero(&ro, sizeof(ro));
4780 dst = satosin(&ro.ro_dst);
4781 dst->sin_family = af;
4782 dst->sin_len = sizeof(*dst);
4783 dst->sin_addr = addr->v4;
4784#ifdef __FreeBSD__
4785#ifdef RTF_PRCLONING
4786 rtalloc_ign(&ro, (RTF_CLONING|RTF_PRCLONING));
4787#else /* !RTF_PRCLONING */
4788 rtalloc_ign(&ro, RTF_CLONING);
4789#endif
4790#else /* ! __FreeBSD__ */
4791 rtalloc_noclone(&ro, NO_CLONING);
4792#endif
4793
4794 if (ro.ro_rt != NULL) {
4795 ret = 1;
4796 RTFREE(ro.ro_rt);
4797 }
4798
4799 return (ret);
4800}
4801
4802#ifdef INET
4803
4804#if defined(__FreeBSD__) && (__FreeBSD_version < 501105)
4805int
4806ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
4807 u_long if_hwassist_flags, int sw_csum)
4808{
4809 int error = 0;
4810 int hlen = ip->ip_hl << 2;
4811 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
4812 int off;
4813 struct mbuf *m0 = *m_frag; /* the original packet */
4814 int firstlen;
4815 struct mbuf **mnext;
4816 int nfrags;
4817
4818 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */
4819 ipstat.ips_cantfrag++;
4820 return EMSGSIZE;
4821 }
4822
4823 /*
4824 * Must be able to put at least 8 bytes per fragment.
4825 */
4826 if (len < 8)
4827 return EMSGSIZE;
4828
4829 /*
4830 * If the interface will not calculate checksums on
4831 * fragmented packets, then do it here.
4832 */
4833 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA &&
4834 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) {
4835 in_delayed_cksum(m0);
4836 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
4837 }
4838
4839 if (len > PAGE_SIZE) {
4840 /*
4841 * Fragment large datagrams such that each segment
4842 * contains a multiple of PAGE_SIZE amount of data,
4843 * plus headers. This enables a receiver to perform
4844 * page-flipping zero-copy optimizations.
4845 *
4846 * XXX When does this help given that sender and receiver
4847 * could have different page sizes, and also mtu could
4848 * be less than the receiver's page size ?
4849 */
4850 int newlen;
4851 struct mbuf *m;
4852
4853 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
4854 off += m->m_len;
4855
4856 /*
4857 * firstlen (off - hlen) must be aligned on an
4858 * 8-byte boundary
4859 */
4860 if (off < hlen)
4861 goto smart_frag_failure;
4862 off = ((off - hlen) & ~7) + hlen;
4863 newlen = (~PAGE_MASK) & mtu;
4864 if ((newlen + sizeof (struct ip)) > mtu) {
4865 /* we failed, go back the default */
4866smart_frag_failure:
4867 newlen = len;
4868 off = hlen + len;
4869 }
4870 len = newlen;
4871
4872 } else {
4873 off = hlen + len;
4874 }
4875
4876 firstlen = off - hlen;
4877 mnext = &m0->m_nextpkt; /* pointer to next packet */
4878
4879 /*
4880 * Loop through length of segment after first fragment,
4881 * make new header and copy data of each part and link onto chain.
4882 * Here, m0 is the original packet, m is the fragment being created.
4883 * The fragments are linked off the m_nextpkt of the original
4884 * packet, which after processing serves as the first fragment.
4885 */
4886 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) {
4887 struct ip *mhip; /* ip header on the fragment */
4888 struct mbuf *m;
4889 int mhlen = sizeof (struct ip);
4890
4891 MGETHDR(m, M_DONTWAIT, MT_HEADER);
4892 if (m == 0) {
4893 error = ENOBUFS;
4894 ipstat.ips_odropped++;
4895 goto done;
4896 }
4897 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
4898 /*
4899 * In the first mbuf, leave room for the link header, then
4900 * copy the original IP header including options. The payload
4901 * goes into an additional mbuf chain returned by m_copy().
4902 */
4903 m->m_data += max_linkhdr;
4904 mhip = mtod(m, struct ip *);
4905 *mhip = *ip;
4906 if (hlen > sizeof (struct ip)) {
4907 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
4908 mhip->ip_v = IPVERSION;
4909 mhip->ip_hl = mhlen >> 2;
4910 }
4911 m->m_len = mhlen;
4912 /* XXX do we need to add ip->ip_off below ? */
4913 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off;
4914 if (off + len >= ip->ip_len) { /* last fragment */
4915 len = ip->ip_len - off;
4916 m->m_flags |= M_LASTFRAG;
4917 } else
4918 mhip->ip_off |= IP_MF;
4919 mhip->ip_len = htons((u_short)(len + mhlen));
4920 m->m_next = m_copy(m0, off, len);
4921 if (m->m_next == 0) { /* copy failed */
4922 m_free(m);
4923 error = ENOBUFS; /* ??? */
4924 ipstat.ips_odropped++;
4925 goto done;
4926 }
4927 m->m_pkthdr.len = mhlen + len;
4928 m->m_pkthdr.rcvif = (struct ifnet *)0;
4929#ifdef MAC
4930 mac_create_fragment(m0, m);
4931#endif
4932 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
4933 mhip->ip_off = htons(mhip->ip_off);
4934 mhip->ip_sum = 0;
4935 if (sw_csum & CSUM_DELAY_IP)
4936 mhip->ip_sum = in_cksum(m, mhlen);
4937 *mnext = m;
4938 mnext = &m->m_nextpkt;
4939 }
4940 ipstat.ips_ofragments += nfrags;
4941
4942 /* set first marker for fragment chain */
4943 m0->m_flags |= M_FIRSTFRAG | M_FRAG;
4944 m0->m_pkthdr.csum_data = nfrags;
4945
4946 /*
4947 * Update first fragment by trimming what's been copied out
4948 * and updating header.
4949 */
4950 m_adj(m0, hlen + firstlen - ip->ip_len);
4951 m0->m_pkthdr.len = hlen + firstlen;
4952 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
4953 ip->ip_off |= IP_MF;
4954 ip->ip_off = htons(ip->ip_off);
4955 ip->ip_sum = 0;
4956 if (sw_csum & CSUM_DELAY_IP)
4957 ip->ip_sum = in_cksum(m0, hlen);
4958
4959done:
4960 *m_frag = m0;
4961 return error;
4962}
4963#endif /* __FreeBSD__ && __FreeBSD_version > 501105 */
4964
4965void
4966pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
4967 struct pf_state *s)
4968{
4969 struct mbuf *m0, *m1;
4970 struct route iproute;
4971 struct route *ro = NULL; /* XXX: was uninitialized */
4972 struct sockaddr_in *dst;
4973 struct ip *ip;
4974 struct ifnet *ifp = NULL;
4975 struct m_tag *mtag;
4976 struct pf_addr naddr;
4977 int error = 0;
4978#ifdef __FreeBSD__
4979 int sw_csum;
4980#endif
4981
4982 if (m == NULL || *m == NULL || r == NULL ||
4983 (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
4984 panic("pf_route: invalid parameters");
4985
4986 if (r->rt == PF_DUPTO) {
4987 m0 = *m;
4988 mtag = m_tag_find(m0, PACKET_TAG_PF_ROUTED, NULL);
4989 if (mtag == NULL) {
4990 mtag = m_tag_get(PACKET_TAG_PF_ROUTED, 0, M_NOWAIT);
4991 if (mtag == NULL)
4992 goto bad;
4993 m_tag_prepend(m0, mtag);
4994 }
4995#ifdef __FreeBSD__
4996 m0 = m_dup(*m, M_DONTWAIT);
4997#else
4998 m0 = m_copym2(*m, 0, M_COPYALL, M_NOWAIT);
4999#endif
5000 if (m0 == NULL)
5001 return;
5002 } else {
5003 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5004 return;
5005 m0 = *m;
5006 }
5007
5008 if (m0->m_len < sizeof(struct ip))
5009 panic("pf_route: m0->m_len < sizeof(struct ip)");
5010 ip = mtod(m0, struct ip *);
5011
5012 ro = &iproute;
5013 bzero((caddr_t)ro, sizeof(*ro));
5014 dst = satosin(&ro->ro_dst);
5015 dst->sin_family = AF_INET;
5016 dst->sin_len = sizeof(*dst);
5017 dst->sin_addr = ip->ip_dst;
5018
5019 if (r->rt == PF_FASTROUTE) {
5020 rtalloc(ro);
5021 if (ro->ro_rt == 0) {
5022 ipstat.ips_noroute++;
5023 goto bad;
5024 }
5025
5026 ifp = ro->ro_rt->rt_ifp;
5027 ro->ro_rt->rt_use++;
5028
5029 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
5030 dst = satosin(ro->ro_rt->rt_gateway);
5031 } else {
5032 if (TAILQ_EMPTY(&r->rpool.list))
5033 panic("pf_route: TAILQ_EMPTY(&r->rpool.list)");
5034 if (s == NULL) {
5035 pf_map_addr(AF_INET, &r->rpool,
5036 (struct pf_addr *)&ip->ip_src,
5037 &naddr, NULL);
5038 if (!PF_AZERO(&naddr, AF_INET))
5039 dst->sin_addr.s_addr = naddr.v4.s_addr;
5040 ifp = r->rpool.cur->ifp;
5041 } else {
5042 if (!PF_AZERO(&s->rt_addr, AF_INET))
5043 dst->sin_addr.s_addr =
5044 s->rt_addr.v4.s_addr;
5045 ifp = s->rt_ifp;
5046 }
5047 }
5048
5049 if (ifp == NULL)
5050 goto bad;
5051
5052 if (m_tag_find(m0, PACKET_TAG_PF_ROUTED, NULL) != NULL)
5053 goto bad;
5054 mtag = m_tag_get(PACKET_TAG_PF_ROUTED, 0, M_NOWAIT);
5055 if (mtag == NULL)
5056 goto bad;
5057 m_tag_prepend(m0, mtag);
5058
5059 if (oifp != ifp) {
5060#ifdef __FreeBSD__
5061 PF_UNLOCK();
5062 if (pf_test(PF_OUT, ifp, &m0) != PF_PASS) {
5063 PF_LOCK();
5064 goto bad;
5065 } else if (m0 == NULL) {
5066 PF_LOCK();
5067 goto done;
5068 }
5069 PF_LOCK();
5070#else
5071 if (pf_test(PF_OUT, ifp, &m0) != PF_PASS)
5072 goto bad;
5073 else if (m0 == NULL)
5074 goto done;
5075#endif
5076 if (m0->m_len < sizeof(struct ip))
5077 panic("pf_route: m0->m_len < sizeof(struct ip)");
5078 ip = mtod(m0, struct ip *);
5079 }
5080
5081#ifdef __FreeBSD__
5082 /* Copied from FreeBSD 5.1-CURRENT ip_output. */
5083 m0->m_pkthdr.csum_flags |= CSUM_IP;
5084 sw_csum = m0->m_pkthdr.csum_flags & ~ifp->if_hwassist;
5085 if (sw_csum & CSUM_DELAY_DATA) {
5086 /*
5087 * XXX: in_delayed_cksum assumes HBO for ip->ip_len (at least)
5088 */
5089 NTOHS(ip->ip_len);
5090 NTOHS(ip->ip_off); /* XXX: needed? */
5091 in_delayed_cksum(m0);
5092 HTONS(ip->ip_len);
5093 HTONS(ip->ip_off);
5094 sw_csum &= ~CSUM_DELAY_DATA;
5095 }
5096 m0->m_pkthdr.csum_flags &= ifp->if_hwassist;
5097
5098 if (ntohs(ip->ip_len) <= ifp->if_mtu ||
5099 (ifp->if_hwassist & CSUM_FRAGMENT &&
5100 ((ip->ip_off & htons(IP_DF)) == 0))) {
5101 /*
5102 * ip->ip_len = htons(ip->ip_len);
5103 * ip->ip_off = htons(ip->ip_off);
5104 */
5105 ip->ip_sum = 0;
5106 if (sw_csum & CSUM_DELAY_IP) {
5107 /* From KAME */
5108 if (ip->ip_v == IPVERSION &&
5109 (ip->ip_hl << 2) == sizeof(*ip)) {
5110 ip->ip_sum = in_cksum_hdr(ip);
5111 } else {
5112 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
5113 }
5114 }
5115 PF_UNLOCK();
5116 error = (*ifp->if_output)(ifp, m0, sintosa(dst), ro->ro_rt);
5117 PF_LOCK();
5118 goto done;
5119 }
5120
5121#else
5122 /* Copied from ip_output. */
5123 if (ntohs(ip->ip_len) <= ifp->if_mtu) {
5124 if ((ifp->if_capabilities & IFCAP_CSUM_IPv4) &&
5125 ifp->if_bridge == NULL) {
5126 m0->m_pkthdr.csum |= M_IPV4_CSUM_OUT;
5127 ipstat.ips_outhwcsum++;
5128 } else {
5129 ip->ip_sum = 0;
5130 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
5131 }
5132 /* Update relevant hardware checksum stats for TCP/UDP */
5133 if (m0->m_pkthdr.csum & M_TCPV4_CSUM_OUT)
5134 tcpstat.tcps_outhwcsum++;
5135 else if (m0->m_pkthdr.csum & M_UDPV4_CSUM_OUT)
5136 udpstat.udps_outhwcsum++;
5137 error = (*ifp->if_output)(ifp, m0, sintosa(dst), NULL);
5138 goto done;
5139 }
5140#endif
5141 /*
5142 * Too large for interface; fragment if possible.
5143 * Must be able to put at least 8 bytes per fragment.
5144 */
5145 if (ip->ip_off & htons(IP_DF)) {
5146 ipstat.ips_cantfrag++;
5147 if (r->rt != PF_DUPTO) {
5148#ifdef __FreeBSD__
5149 /* icmp_error() expects host byte ordering */
5150 NTOHS(ip->ip_len);
5151 NTOHS(ip->ip_off);
5152 PF_UNLOCK();
5153#endif
5154 icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0,
5155 ifp);
5156#ifdef __FreeBSD__
5157 PF_LOCK();
5158#endif
5159 goto done;
5160 } else
5161 goto bad;
5162 }
5163
5164 m1 = m0;
5165#ifdef __FreeBSD__
5166 /*
5167 * XXX: is cheaper + less error prone than own function
5168 */
5169 NTOHS(ip->ip_len);
5170 NTOHS(ip->ip_off);
5171 error = ip_fragment(ip, &m0, ifp->if_mtu, ifp->if_hwassist, sw_csum);
5172#else
5173 error = ip_fragment(m0, ifp, ifp->if_mtu);
5174#endif
5175 if (error) {
5176#ifndef __FreeBSD__ /* ip_fragment does not do m_freem() on FreeBSD */
5177 m0 = NULL;
5178#endif
5179 goto bad;
5180 }
5181
5182 for (m0 = m1; m0; m0 = m1) {
5183 m1 = m0->m_nextpkt;
5184 m0->m_nextpkt = 0;
5185#ifdef __FreeBSD__
5186 if (error == 0) {
5187 PF_UNLOCK();
5188 error = (*ifp->if_output)(ifp, m0, sintosa(dst),
5189 NULL);
5190 PF_LOCK();
5191 } else
5192#else
5193 if (error == 0)
5194 error = (*ifp->if_output)(ifp, m0, sintosa(dst),
5195 NULL);
5196 else
5197#endif
5198 m_freem(m0);
5199 }
5200
5201 if (error == 0)
5202 ipstat.ips_fragmented++;
5203
5204done:
5205 if (r->rt != PF_DUPTO)
5206 *m = NULL;
5207 if (ro == &iproute && ro->ro_rt)
5208 RTFREE(ro->ro_rt);
5209 return;
5210
5211bad:
5212 m_freem(m0);
5213 goto done;
5214}
5215#endif /* INET */
5216
5217#ifdef INET6
5218void
5219pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5220 struct pf_state *s)
5221{
5222 struct mbuf *m0;
5223 struct m_tag *mtag;
5224 struct route_in6 ip6route;
5225 struct route_in6 *ro;
5226 struct sockaddr_in6 *dst;
5227 struct ip6_hdr *ip6;
5228 struct ifnet *ifp = NULL;
5229 struct pf_addr naddr;
5230 int error = 0;
5231
5232 if (m == NULL || *m == NULL || r == NULL ||
5233 (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5234 panic("pf_route6: invalid parameters");
5235
5236 if (r->rt == PF_DUPTO) {
5237 m0 = *m;
5238 mtag = m_tag_find(m0, PACKET_TAG_PF_ROUTED, NULL);
5239 if (mtag == NULL) {
5240 mtag = m_tag_get(PACKET_TAG_PF_ROUTED, 0, M_NOWAIT);
5241 if (mtag == NULL)
5242 goto bad;
5243 m_tag_prepend(m0, mtag);
5244 }
5245#ifdef __FreeBSD__
5246 m0 = m_dup(*m, M_DONTWAIT);
5247#else
5248 m0 = m_copym2(*m, 0, M_COPYALL, M_NOWAIT);
5249#endif
5250 if (m0 == NULL)
5251 return;
5252 } else {
5253 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5254 return;
5255 m0 = *m;
5256 }
5257
5258 if (m0->m_len < sizeof(struct ip6_hdr))
5259 panic("pf_route6: m0->m_len < sizeof(struct ip6_hdr)");
5260 ip6 = mtod(m0, struct ip6_hdr *);
5261
5262 ro = &ip6route;
5263 bzero((caddr_t)ro, sizeof(*ro));
5264 dst = (struct sockaddr_in6 *)&ro->ro_dst;
5265 dst->sin6_family = AF_INET6;
5266 dst->sin6_len = sizeof(*dst);
5267 dst->sin6_addr = ip6->ip6_dst;
5268
5269 /* Cheat. */
5270 if (r->rt == PF_FASTROUTE) {
5271 mtag = m_tag_get(PACKET_TAG_PF_GENERATED, 0, M_NOWAIT);
5272 if (mtag == NULL)
5273 goto bad;
5274 m_tag_prepend(m0, mtag);
5275#ifdef __FreeBSD__
5276 PF_UNLOCK();
5277 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
5278 PF_LOCK();
5279#else
5280 ip6_output(m0, NULL, NULL, 0, NULL, NULL);
5281#endif
5282 return;
5283 }
5284
5285 if (TAILQ_EMPTY(&r->rpool.list))
5286 panic("pf_route6: TAILQ_EMPTY(&r->rpool.list)");
5287 if (s == NULL) {
5288 pf_map_addr(AF_INET6, &r->rpool,
5289 (struct pf_addr *)&ip6->ip6_src, &naddr, NULL);
5290 if (!PF_AZERO(&naddr, AF_INET6))
5291 PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5292 &naddr, AF_INET6);
5293 ifp = r->rpool.cur->ifp;
5294 } else {
5295 if (!PF_AZERO(&s->rt_addr, AF_INET6))
5296 PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5297 &s->rt_addr, AF_INET6);
5298 ifp = s->rt_ifp;
5299 }
5300
5301 if (ifp == NULL)
5302 goto bad;
5303
5304 if (oifp != ifp) {
5305 mtag = m_tag_find(m0, PACKET_TAG_PF_ROUTED, NULL);
5306 if (mtag == NULL) {
5307 mtag = m_tag_get(PACKET_TAG_PF_ROUTED, 0, M_NOWAIT);
5308 if (mtag == NULL)
5309 goto bad;
5310 m_tag_prepend(m0, mtag);
5311#ifdef __FreeBSD__
5312 PF_UNLOCK();
5313 if (pf_test6(PF_OUT, ifp, &m0) != PF_PASS) {
5314 PF_LOCK();
5315 goto bad;
5316 } else if (m0 == NULL) {
5317 PF_LOCK();
5318 goto done;
5319 }
5320 PF_LOCK();
5321#else
5322 if (pf_test6(PF_OUT, ifp, &m0) != PF_PASS)
5323 goto bad;
5324 else if (m0 == NULL)
5325 goto done;
5326#endif
5327 }
5328 }
5329
5330 /*
5331 * If the packet is too large for the outgoing interface,
5332 * send back an icmp6 error.
5333 */
5334 if (IN6_IS_ADDR_LINKLOCAL(&dst->sin6_addr))
5335 dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
5336 if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) {
5337#ifdef __FreeBSD__
5338 PF_UNLOCK();
5339#endif
5340 error = nd6_output(ifp, ifp, m0, dst, NULL);
5341#ifdef __FreeBSD__
5342 PF_LOCK();
5343#endif
5344 } else {
5345 in6_ifstat_inc(ifp, ifs6_in_toobig);
5346#ifdef __FreeBSD__
5347 if (r->rt != PF_DUPTO) {
5348 PF_UNLOCK();
5349 icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
5350 PF_LOCK();
5351 } else
5352#else
5353 if (r->rt != PF_DUPTO)
5354 icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
5355 else
5356#endif
5357 goto bad;
5358 }
5359
5360done:
5361 if (r->rt != PF_DUPTO)
5362 *m = NULL;
5363 return;
5364
5365bad:
5366 m_freem(m0);
5367 goto done;
5368}
5369#endif /* INET6 */
5370
5371
5372#ifdef __FreeBSD__
5373/*
5374 * XXX
5375 * FreeBSD supports cksum offload for the following drivers.
5376 * em(4), gx(4), lge(4), nge(4), ti(4), xl(4)
5377 * If we can make full use of it we would outperform ipfw/ipfilter in
5378 * very heavy traffic.
5379 * I have not tested 'cause I don't have NICs that supports cksum offload.
5380 * (There might be problems. Typical phenomena would be
5381 * 1. No route message for UDP packet.
5382 * 2. No connection acceptance from external hosts regardless of rule set.)
5383 */
5384int
5385pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p, sa_family_t af)
5386{
5387 u_int16_t sum = 0;
5388 int hw_assist = 0;
5389 struct ip *ip;
5390
5391 if (off < sizeof(struct ip) || len < sizeof(struct udphdr))
5392 return (1);
5393 if (m->m_pkthdr.len < off + len)
5394 return (1);
5395
5396 switch (p) {
5397 case IPPROTO_TCP:
5398 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
5399 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) {
5400 sum = m->m_pkthdr.csum_data;
5401 } else {
5402 ip = mtod(m, struct ip *);
5403 sum = in_pseudo(ip->ip_src.s_addr,
5404 ip->ip_dst.s_addr,
5405 htonl(m->m_pkthdr.csum_data +
5406 IPPROTO_TCP) + ip->ip_len);
5407 }
5408 sum ^= 0xffff;
5409 ++hw_assist;
5410 }
5411 break;
5412 case IPPROTO_UDP:
5413 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
5414 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) {
5415 sum = m->m_pkthdr.csum_data;
5416 } else {
5417 ip = mtod(m, struct ip *);
5418 sum = in_pseudo(ip->ip_src.s_addr,
5419 ip->ip_dst.s_addr, htonl((u_short)len +
5420 m->m_pkthdr.csum_data + IPPROTO_UDP));
5421 }
5422 sum ^= 0xffff;
5423 ++hw_assist;
5424 }
5425 break;
5426 case IPPROTO_ICMP:
5427#ifdef INET6
5428 case IPPROTO_ICMPV6:
5429#endif /* INET6 */
5430 break;
5431 default:
5432 return (1);
5433 }
5434
5435 if (!hw_assist) {
5436 switch (af) {
5437 case AF_INET:
5438 if (p == IPPROTO_ICMP) {
5439 if (m->m_len < off)
5440 return (1);
5441 m->m_data += off;
5442 m->m_len -= off;
5443 sum = in_cksum(m, len);
5444 m->m_data -= off;
5445 m->m_len += off;
5446 } else {
5447 if (m->m_len < sizeof(struct ip))
5448 return (1);
5449 sum = in4_cksum(m, p, off, len);
5450 if (sum == 0) {
5451 m->m_pkthdr.csum_flags |=
5452 (CSUM_DATA_VALID |
5453 CSUM_PSEUDO_HDR);
5454 m->m_pkthdr.csum_data = 0xffff;
5455 }
5456 }
5457 break;
5458#ifdef INET6
5459 case AF_INET6:
5460 if (m->m_len < sizeof(struct ip6_hdr))
5461 return (1);
5462 sum = in6_cksum(m, p, off, len);
5463 /*
5464 * XXX
5465 * IPv6 H/W cksum off-load not supported yet!
5466 *
5467 * if (sum == 0) {
5468 * m->m_pkthdr.csum_flags |=
5469 * (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
5470 * m->m_pkthdr.csum_data = 0xffff;
5471 *}
5472 */
5473 break;
5474#endif /* INET6 */
5475 default:
5476 return (1);
5477 }
5478 }
5479 if (sum) {
5480 switch (p) {
5481 case IPPROTO_TCP:
5482 tcpstat.tcps_rcvbadsum++;
5483 break;
5484 case IPPROTO_UDP:
5485 udpstat.udps_badsum++;
5486 break;
5487 case IPPROTO_ICMP:
5488 icmpstat.icps_checksum++;
5489 break;
5490#ifdef INET6
5491 case IPPROTO_ICMPV6:
5492 icmp6stat.icp6s_checksum++;
5493 break;
5494#endif /* INET6 */
5495 }
5496 return (1);
5497 }
5498 return (0);
5499}
5500#else
5501/*
5502 * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag
5503 * off is the offset where the protocol header starts
5504 * len is the total length of protocol header plus payload
5505 * returns 0 when the checksum is valid, otherwise returns 1.
5506 */
5507int
5508pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p, sa_family_t af)
5509{
5510 u_int16_t flag_ok, flag_bad;
5511 u_int16_t sum;
5512
5513 switch (p) {
5514 case IPPROTO_TCP:
5515 flag_ok = M_TCP_CSUM_IN_OK;
5516 flag_bad = M_TCP_CSUM_IN_BAD;
5517 break;
5518 case IPPROTO_UDP:
5519 flag_ok = M_UDP_CSUM_IN_OK;
5520 flag_bad = M_UDP_CSUM_IN_BAD;
5521 break;
5522 case IPPROTO_ICMP:
5523#ifdef INET6
5524 case IPPROTO_ICMPV6:
5525#endif /* INET6 */
5526 flag_ok = flag_bad = 0;
5527 break;
5528 default:
5529 return (1);
5530 }
5531 if (m->m_pkthdr.csum & flag_ok)
5532 return (0);
5533 if (m->m_pkthdr.csum & flag_bad)
5534 return (1);
5535 if (off < sizeof(struct ip) || len < sizeof(struct udphdr))
5536 return (1);
5537 if (m->m_pkthdr.len < off + len)
5538 return (1);
5539 switch (af) {
5540 case AF_INET:
5541 if (p == IPPROTO_ICMP) {
5542 if (m->m_len < off)
5543 return (1);
5544 m->m_data += off;
5545 m->m_len -= off;
5546 sum = in_cksum(m, len);
5547 m->m_data -= off;
5548 m->m_len += off;
5549 } else {
5550 if (m->m_len < sizeof(struct ip))
5551 return (1);
5552 sum = in4_cksum(m, p, off, len);
5553 }
5554 break;
5555#ifdef INET6
5556 case AF_INET6:
5557 if (m->m_len < sizeof(struct ip6_hdr))
5558 return (1);
5559 sum = in6_cksum(m, p, off, len);
5560 break;
5561#endif /* INET6 */
5562 default:
5563 return (1);
5564 }
5565 if (sum) {
5566 m->m_pkthdr.csum |= flag_bad;
5567 switch (p) {
5568 case IPPROTO_TCP:
5569 tcpstat.tcps_rcvbadsum++;
5570 break;
5571 case IPPROTO_UDP:
5572 udpstat.udps_badsum++;
5573 break;
5574 case IPPROTO_ICMP:
5575 icmpstat.icps_checksum++;
5576 break;
5577#ifdef INET6
5578 case IPPROTO_ICMPV6:
5579 icmp6stat.icp6s_checksum++;
5580 break;
5581#endif /* INET6 */
5582 }
5583 return (1);
5584 }
5585 m->m_pkthdr.csum |= flag_ok;
5586 return (0);
5587}
5588#endif
5589
5590#ifdef INET
5591int
5592pf_test(int dir, struct ifnet *ifp, struct mbuf **m0)
5593{
5594 u_short action, reason = 0, log = 0;
5595 struct mbuf *m = *m0;
5596 struct ip *h = NULL; /* XXX: was uninitialized */
5597 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr;
5598 struct pf_state *s = NULL;
5599 struct pf_ruleset *ruleset = NULL;
5600 struct pf_pdesc pd;
5601 int off;
5602 int pqid = 0;
5603
5604#ifdef __FreeBSD__
5605 PF_LOCK();
5606#endif
5607 if (!pf_status.running ||
5608 (m_tag_find(m, PACKET_TAG_PF_GENERATED, NULL) != NULL)) {
5609#ifdef __FreeBSD__
5610 PF_UNLOCK();
5611#endif
5612 return (PF_PASS);
5613 }
5614
5615#if defined(__FreeBSD__) && (__FreeBSD_version >= 501000)
5616 M_ASSERTPKTHDR(m);
5617#else
5618#ifdef DIAGNOSTIC
5619 if ((m->m_flags & M_PKTHDR) == 0)
5620 panic("non-M_PKTHDR is passed to pf_test");
5621#endif
5622#endif
5623
5624 if (m->m_pkthdr.len < (int)sizeof(*h)) {
5625 action = PF_DROP;
5626 REASON_SET(&reason, PFRES_SHORT);
5627 log = 1;
5628 goto done;
5629 }
5630
5631 /* We do IP header normalization and packet reassembly here */
5632 if (pf_normalize_ip(m0, dir, ifp, &reason) != PF_PASS) {
5633 action = PF_DROP;
5634 goto done;
5635 }
5636 m = *m0;
5637 h = mtod(m, struct ip *);
5638
5639 off = h->ip_hl << 2;
5640 if (off < (int)sizeof(*h)) {
5641 action = PF_DROP;
5642 REASON_SET(&reason, PFRES_SHORT);
5643 log = 1;
5644 goto done;
5645 }
5646
5647 memset(&pd, 0, sizeof(pd));
5648 pd.src = (struct pf_addr *)&h->ip_src;
5649 pd.dst = (struct pf_addr *)&h->ip_dst;
5650 pd.ip_sum = &h->ip_sum;
5651 pd.proto = h->ip_p;
5652 pd.af = AF_INET;
5653 pd.tos = h->ip_tos;
5654 pd.tot_len = ntohs(h->ip_len);
5655
5656 /* handle fragments that didn't get reassembled by normalization */
5657 if (h->ip_off & htons(IP_MF | IP_OFFMASK)) {
5658 action = pf_test_fragment(&r, dir, ifp, m, h,
5659 &pd, &a, &ruleset);
5660 goto done;
5661 }
5662
5663 switch (h->ip_p) {
5664
5665 case IPPROTO_TCP: {
5666 struct tcphdr th;
5667
5668 pd.hdr.tcp = &th;
5669 if (!pf_pull_hdr(m, off, &th, sizeof(th),
5670 &action, &reason, AF_INET)) {
5671 log = action != PF_PASS;
5672 goto done;
5673 }
5674 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5675 ntohs(h->ip_len) - off, IPPROTO_TCP, AF_INET)) {
5676 action = PF_DROP;
5677 goto done;
5678 }
5679 pd.p_len = pd.tot_len - off - (th.th_off << 2);
5680 if ((th.th_flags & TH_ACK) && pd.p_len == 0)
5681 pqid = 1;
5682 action = pf_normalize_tcp(dir, ifp, m, 0, off, h, &pd);
5683 if (action == PF_DROP)
5684 break;
5685 action = pf_test_state_tcp(&s, dir, ifp, m, 0, off, h, &pd,
5686 &reason);
5687 if (action == PF_PASS) {
5688 r = s->rule.ptr;
5689 log = s->log;
5690 } else if (s == NULL)
5691 action = pf_test_tcp(&r, &s, dir, ifp,
5692 m, 0, off, h, &pd, &a, &ruleset);
5693 break;
5694 }
5695
5696 case IPPROTO_UDP: {
5697 struct udphdr uh;
5698
5699 pd.hdr.udp = &uh;
5700 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
5701 &action, &reason, AF_INET)) {
5702 log = action != PF_PASS;
5703 goto done;
5704 }
5705 if (dir == PF_IN && uh.uh_sum && pf_check_proto_cksum(m,
5706 off, ntohs(h->ip_len) - off, IPPROTO_UDP, AF_INET)) {
5707 action = PF_DROP;
5708 goto done;
5709 }
5710 action = pf_test_state_udp(&s, dir, ifp, m, 0, off, h, &pd);
5711 if (action == PF_PASS) {
5712 r = s->rule.ptr;
5713 a = s->anchor.ptr;
5714 log = s->log;
5715 } else if (s == NULL)
5716 action = pf_test_udp(&r, &s, dir, ifp,
5717 m, 0, off, h, &pd, &a, &ruleset);
5718 break;
5719 }
5720
5721 case IPPROTO_ICMP: {
5722 struct icmp ih;
5723
5724 pd.hdr.icmp = &ih;
5725 if (!pf_pull_hdr(m, off, &ih, ICMP_MINLEN,
5726 &action, &reason, AF_INET)) {
5727 log = action != PF_PASS;
5728 goto done;
5729 }
5730 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5731 ntohs(h->ip_len) - off, IPPROTO_ICMP, AF_INET)) {
5732 action = PF_DROP;
5733 goto done;
5734 }
5735 action = pf_test_state_icmp(&s, dir, ifp, m, 0, off, h, &pd);
5736 if (action == PF_PASS) {
5737 r = s->rule.ptr;
5738 r->packets++;
5739 r->bytes += ntohs(h->ip_len);
5740 a = s->anchor.ptr;
5741 if (a != NULL) {
5742 a->packets++;
5743 a->bytes += ntohs(h->ip_len);
5744 }
5745 log = s->log;
5746 } else if (s == NULL)
5747 action = pf_test_icmp(&r, &s, dir, ifp,
5748 m, 0, off, h, &pd, &a, &ruleset);
5749 break;
5750 }
5751
5752 default:
5753 action = pf_test_state_other(&s, dir, ifp, &pd);
5754 if (action == PF_PASS) {
5755 r = s->rule.ptr;
5756 a = s->anchor.ptr;
5757 log = s->log;
5758 } else if (s == NULL)
5759 action = pf_test_other(&r, &s, dir, ifp, m, off, h,
5760 &pd, &a, &ruleset);
5761 break;
5762 }
5763
5764 if (ifp == status_ifp) {
5765 pf_status.bcounters[0][dir == PF_OUT] += pd.tot_len;
5766 pf_status.pcounters[0][dir == PF_OUT][action != PF_PASS]++;
5767 }
5768
5769done:
5770 tr = r;
5771 if (r == &pf_default_rule && s != NULL && s->nat_rule.ptr != NULL)
5772 tr = s->nat_rule.ptr;
5773 if (tr->src.addr.type == PF_ADDR_TABLE)
5774 pfr_update_stats(tr->src.addr.p.tbl,
5775 (s == NULL || s->direction == dir) ? pd.src : pd.dst, pd.af,
5776 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5777 tr->src.not);
5778 if (tr->dst.addr.type == PF_ADDR_TABLE)
5779 pfr_update_stats(tr->dst.addr.p.tbl,
5780 (s == NULL || s->direction == dir) ? pd.dst : pd.src, pd.af,
5781 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5782 tr->dst.not);
5783
5784 if (action == PF_PASS && h->ip_hl > 5 &&
5785 !((s && s->allow_opts) || r->allow_opts)) {
5786 action = PF_DROP;
5787 REASON_SET(&reason, PFRES_SHORT);
5788 log = 1;
5789 DPFPRINTF(PF_DEBUG_MISC,
5790 ("pf: dropping packet with ip options\n"));
5791 }
5792
5793#ifdef ALTQ
5794 if (action == PF_PASS && r->qid) {
5795 struct m_tag *mtag;
5796 struct altq_tag *atag;
5797
5798 mtag = m_tag_get(PACKET_TAG_PF_QID, sizeof(*atag), M_NOWAIT);
5799 if (mtag != NULL) {
5800 atag = (struct altq_tag *)(mtag + 1);
5801 if (pqid || pd.tos == IPTOS_LOWDELAY)
5802 atag->qid = r->pqid;
5803 else
5804 atag->qid = r->qid;
5805 /* add hints for ecn */
5806 atag->af = AF_INET;
5807 atag->hdr = h;
5808 m_tag_prepend(m, mtag);
5809 }
5810 }
5811#endif
5812
5813 if (log)
5814 PFLOG_PACKET(ifp, h, m, AF_INET, dir, reason, r, a, ruleset);
5815
5816 if (action == PF_SYNPROXY_DROP) {
5817 m_freem(*m0);
5818 *m0 = NULL;
5819 action = PF_PASS;
5820 } else if (r->rt)
5821 /* pf_route can free the mbuf causing *m0 to become NULL */
5822 pf_route(m0, r, dir, ifp, s);
5823
5824#ifdef __FreeBSD__
5825 PF_UNLOCK();
5826#endif
5827
5828 return (action);
5829}
5830#endif /* INET */
5831
5832#ifdef INET6
5833int
5834pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0)
5835{
5836 u_short action, reason = 0, log = 0;
5837 struct mbuf *m = *m0;
5838 struct ip6_hdr *h = NULL; /* make the compiler happy */
5839 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr;
5840 struct pf_state *s = NULL;
5841 struct pf_ruleset *ruleset = NULL;
5842 struct pf_pdesc pd;
5843 int off, terminal = 0;
5844
5845#ifdef __FreeBSD__
5846 PF_LOCK();
5847#endif
5848
5849 if (!pf_status.running ||
5850 (m_tag_find(m, PACKET_TAG_PF_GENERATED, NULL) != NULL)) {
5851#ifdef __FreeBSD__
5852 PF_UNLOCK();
5853#endif
5854 return (PF_PASS);
5855 }
5856
5857#if defined(__FreeBSD__) && (__FreeBSD_version >= 501000)
5858 M_ASSERTPKTHDR(m);
5859#else
5860#ifdef DIAGNOSTIC
5861 if ((m->m_flags & M_PKTHDR) == 0)
5862 panic("non-M_PKTHDR is passed to pf_test");
5863#endif
5864#endif
5865
5866 if (m->m_pkthdr.len < (int)sizeof(*h)) {
5867 action = PF_DROP;
5868 REASON_SET(&reason, PFRES_SHORT);
5869 log = 1;
5870 goto done;
5871 }
5872
5873 /* We do IP header normalization and packet reassembly here */
5874 if (pf_normalize_ip6(m0, dir, ifp, &reason) != PF_PASS) {
5875 action = PF_DROP;
5876 goto done;
5877 }
5878 m = *m0;
5879 h = mtod(m, struct ip6_hdr *);
5880
5881 memset(&pd, 0, sizeof(pd));
5882 pd.src = (struct pf_addr *)&h->ip6_src;
5883 pd.dst = (struct pf_addr *)&h->ip6_dst;
5884 pd.ip_sum = NULL;
5885 pd.af = AF_INET6;
5886 pd.tos = 0;
5887 pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr);
5888
5889 off = ((caddr_t)h - m->m_data) + sizeof(struct ip6_hdr);
5890 pd.proto = h->ip6_nxt;
5891 do {
5892 switch (pd.proto) {
5893 case IPPROTO_FRAGMENT:
5894 action = pf_test_fragment(&r, dir, ifp, m, h,
5895 &pd, &a, &ruleset);
5896 if (action == PF_DROP)
5897 REASON_SET(&reason, PFRES_FRAG);
5898 goto done;
5899 case IPPROTO_AH:
5900 case IPPROTO_HOPOPTS:
5901 case IPPROTO_ROUTING:
5902 case IPPROTO_DSTOPTS: {
5903 /* get next header and header length */
5904 struct ip6_ext opt6;
5905
5906 if (!pf_pull_hdr(m, off, &opt6, sizeof(opt6),
5907 NULL, NULL, pd.af)) {
5908 DPFPRINTF(PF_DEBUG_MISC,
5909 ("pf: IPv6 short opt\n"));
5910 action = PF_DROP;
5911 REASON_SET(&reason, PFRES_SHORT);
5912 log = 1;
5913 goto done;
5914 }
5915 if (pd.proto == IPPROTO_AH)
5916 off += (opt6.ip6e_len + 2) * 4;
5917 else
5918 off += (opt6.ip6e_len + 1) * 8;
5919 pd.proto = opt6.ip6e_nxt;
5920 /* goto the next header */
5921 break;
5922 }
5923 default:
5924 terminal++;
5925 break;
5926 }
5927 } while (!terminal);
5928
5929 switch (pd.proto) {
5930
5931 case IPPROTO_TCP: {
5932 struct tcphdr th;
5933
5934 pd.hdr.tcp = &th;
5935 if (!pf_pull_hdr(m, off, &th, sizeof(th),
5936 &action, &reason, AF_INET6)) {
5937 log = action != PF_PASS;
5938 goto done;
5939 }
5940 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5941 ntohs(h->ip6_plen), IPPROTO_TCP, AF_INET6)) {
5942 action = PF_DROP;
5943 goto done;
5944 }
5945 pd.p_len = pd.tot_len - off - (th.th_off << 2);
5946 action = pf_normalize_tcp(dir, ifp, m, 0, off, h, &pd);
5947 if (action == PF_DROP)
5948 break;
5949 action = pf_test_state_tcp(&s, dir, ifp, m, 0, off, h, &pd,
5950 &reason);
5951 if (action == PF_PASS) {
5952 r = s->rule.ptr;
5953 log = s->log;
5954 } else if (s == NULL)
5955 action = pf_test_tcp(&r, &s, dir, ifp,
5956 m, 0, off, h, &pd, &a, &ruleset);
5957 break;
5958 }
5959
5960 case IPPROTO_UDP: {
5961 struct udphdr uh;
5962
5963 pd.hdr.udp = &uh;
5964 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
5965 &action, &reason, AF_INET6)) {
5966 log = action != PF_PASS;
5967 goto done;
5968 }
5969 if (dir == PF_IN && uh.uh_sum && pf_check_proto_cksum(m,
5970 off, ntohs(h->ip6_plen), IPPROTO_UDP, AF_INET6)) {
5971 action = PF_DROP;
5972 goto done;
5973 }
5974 action = pf_test_state_udp(&s, dir, ifp, m, 0, off, h, &pd);
5975 if (action == PF_PASS) {
5976 r = s->rule.ptr;
5977 log = s->log;
5978 } else if (s == NULL)
5979 action = pf_test_udp(&r, &s, dir, ifp,
5980 m, 0, off, h, &pd, &a, &ruleset);
5981 break;
5982 }
5983
5984 case IPPROTO_ICMPV6: {
5985 struct icmp6_hdr ih;
5986
5987 pd.hdr.icmp6 = &ih;
5988 if (!pf_pull_hdr(m, off, &ih, sizeof(ih),
5989 &action, &reason, AF_INET6)) {
5990 log = action != PF_PASS;
5991 goto done;
5992 }
5993 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5994 ntohs(h->ip6_plen), IPPROTO_ICMPV6, AF_INET6)) {
5995 action = PF_DROP;
5996 goto done;
5997 }
5998 action = pf_test_state_icmp(&s, dir, ifp,
5999 m, 0, off, h, &pd);
6000 if (action == PF_PASS) {
6001 r = s->rule.ptr;
6002 r->packets++;
6003 r->bytes += h->ip6_plen;
6004 log = s->log;
6005 } else if (s == NULL)
6006 action = pf_test_icmp(&r, &s, dir, ifp,
6007 m, 0, off, h, &pd, &a, &ruleset);
6008 break;
6009 }
6010
6011 default:
6012 action = pf_test_other(&r, &s, dir, ifp, m, off, h,
6013 &pd, &a, &ruleset);
6014 break;
6015 }
6016
6017 if (ifp == status_ifp) {
6018 pf_status.bcounters[1][dir == PF_OUT] += pd.tot_len;
6019 pf_status.pcounters[1][dir == PF_OUT][action != PF_PASS]++;
6020 }
6021
6022done:
6023 tr = r;
6024 if (r == &pf_default_rule && s != NULL && s->nat_rule.ptr != NULL)
6025 tr = s->nat_rule.ptr;
6026 if (tr->src.addr.type == PF_ADDR_TABLE)
6027 pfr_update_stats(tr->src.addr.p.tbl,
6028 (s == NULL || s->direction == dir) ? pd.src : pd.dst, pd.af,
6029 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
6030 tr->src.not);
6031 if (tr->dst.addr.type == PF_ADDR_TABLE)
6032 pfr_update_stats(tr->dst.addr.p.tbl,
6033 (s == NULL || s->direction == dir) ? pd.dst : pd.src, pd.af,
6034 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
6035 tr->dst.not);
6036
6037 /* XXX handle IPv6 options, if not allowed. not implemented. */
6038
6039#ifdef ALTQ
6040 if (action == PF_PASS && r->qid) {
6041 struct m_tag *mtag;
6042 struct altq_tag *atag;
6043
6044 mtag = m_tag_get(PACKET_TAG_PF_QID, sizeof(*atag), M_NOWAIT);
6045 if (mtag != NULL) {
6046 atag = (struct altq_tag *)(mtag + 1);
6047 if (pd.tos == IPTOS_LOWDELAY)
6048 atag->qid = r->pqid;
6049 else
6050 atag->qid = r->qid;
6051 /* add hints for ecn */
6052 atag->af = AF_INET6;
6053 atag->hdr = h;
6054 m_tag_prepend(m, mtag);
6055 }
6056 }
6057#endif
6058
6059 if (log)
6060 PFLOG_PACKET(ifp, h, m, AF_INET6, dir, reason, r, a, ruleset);
6061
6062 if (action == PF_SYNPROXY_DROP) {
6063 m_freem(*m0);
6064 *m0 = NULL;
6065 action = PF_PASS;
6066 } else if (r->rt)
6067 /* pf_route6 can free the mbuf causing *m0 to become NULL */
6068 pf_route6(m0, r, dir, ifp, s);
6069
6070#ifdef __FreeBSD__
6071 PF_UNLOCK();
6072#endif
6073 return (action);
6074}
6075#endif /* INET6 */