1 /* $OpenBSD: pf.c,v 1.634 2009/02/27 12:37:45 henning Exp $ */
2
3 /*
4 * Copyright (c) 2001 Daniel Hartmeier
5 * Copyright (c) 2002 - 2008 Henning Brauer
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
42 #include <sys/cdefs.h>
43 __FBSDID("$FreeBSD: stable/9/sys/contrib/pf/net/pf.c 304466 2016-08-19 13:39:36Z kp $");
44 #endif
45
46 #ifdef __FreeBSD__
47 #include "opt_bpf.h"
48 #include "opt_pf.h"
49
50 #define NPFSYNC 1
51
52 #ifdef DEV_PFLOW
53 #define NPFLOW DEV_PFLOW
54 #else
55 #define NPFLOW 0
56 #endif
57
58 #else
59 #include "bpfilter.h"
60 #include "pflog.h"
61 #include "pfsync.h"
62 #include "pflow.h"
63 #endif
64
65 #include <sys/param.h>
66 #include <sys/systm.h>
67 #include <sys/mbuf.h>
68 #include <sys/filio.h>
69 #include <sys/socket.h>
70 #include <sys/socketvar.h>
71 #include <sys/kernel.h>
72 #include <sys/time.h>
73 #ifdef __FreeBSD__
74 #include <sys/random.h>
75 #include <sys/sysctl.h>
76 #include <sys/endian.h>
77 #define betoh64 be64toh
78 #else
79 #include <sys/pool.h>
80 #endif
81 #include <sys/proc.h>
82 #ifdef __FreeBSD__
83 #include <sys/kthread.h>
84 #include <sys/lock.h>
85 #include <sys/sx.h>
86 #else
87 #include <sys/rwlock.h>
88 #endif
89
90 #ifdef __FreeBSD__
91 #include <sys/md5.h>
92 #else
93 #include <crypto/md5.h>
94 #endif
95
96 #include <net/if.h>
97 #include <net/if_types.h>
98 #include <net/bpf.h>
99 #include <net/route.h>
100 #ifdef __FreeBSD__
101 #ifdef RADIX_MPATH
102 #include <net/radix_mpath.h>
103 #endif
104 #else
105 #include <net/radix_mpath.h>
106 #endif
107
108 #include <netinet/in.h>
109 #include <netinet/in_var.h>
110 #include <netinet/in_systm.h>
111 #include <netinet/ip.h>
112 #include <netinet/ip_var.h>
113 #include <netinet/tcp.h>
114 #include <netinet/tcp_seq.h>
115 #include <netinet/udp.h>
116 #include <netinet/ip_icmp.h>
117 #include <netinet/in_pcb.h>
118 #include <netinet/tcp_timer.h>
119 #include <netinet/tcp_var.h>
120 #include <netinet/udp_var.h>
121 #include <netinet/icmp_var.h>
122 #include <netinet/if_ether.h>
123 #ifdef __FreeBSD__
124 #include <netinet/ip_fw.h>
125 #include <netpfil/ipfw/ip_fw_private.h> /* XXX: only for DIR_IN/DIR_OUT */
126 #endif
127
128 #ifndef __FreeBSD__
129 #include <dev/rndvar.h>
130 #endif
131 #include <net/pfvar.h>
132 #include <net/if_pflog.h>
133 #include <net/if_pflow.h>
134 #include <net/if_pfsync.h>
135
136 #ifdef INET6
137 #include <netinet/ip6.h>
138 #include <netinet/in_pcb.h>
139 #include <netinet/icmp6.h>
140 #include <netinet6/nd6.h>
141 #ifdef __FreeBSD__
142 #include <netinet6/ip6_var.h>
143 #include <netinet6/in6_pcb.h>
144 #endif
145 #endif /* INET6 */
146
147 #ifdef __FreeBSD__
148 #include <machine/in_cksum.h>
149 #include <sys/limits.h>
150 #include <sys/ucred.h>
151 #include <security/mac/mac_framework.h>
152
153 extern int ip_optcopy(struct ip *, struct ip *);
154 #endif
155
156 #ifdef __FreeBSD__
157 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
158 #else
159 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) printf x
160 #endif
161
162 /*
163 * Global variables
164 */
165
166 /* state tables */
167 #ifdef __FreeBSD__
168 VNET_DEFINE(struct pf_state_tree, pf_statetbl);
169
170 VNET_DEFINE(struct pf_altqqueue, pf_altqs[2]);
171 VNET_DEFINE(struct pf_palist, pf_pabuf);
172 VNET_DEFINE(struct pf_altqqueue *, pf_altqs_active);
173 VNET_DEFINE(struct pf_altqqueue *, pf_altqs_inactive);
174 VNET_DEFINE(struct pf_status, pf_status);
175
176 VNET_DEFINE(u_int32_t, ticket_altqs_active);
177 VNET_DEFINE(u_int32_t, ticket_altqs_inactive);
178 VNET_DEFINE(int, altqs_inactive_open);
179 VNET_DEFINE(u_int32_t, ticket_pabuf);
180
181 VNET_DEFINE(MD5_CTX, pf_tcp_secret_ctx);
182 #define V_pf_tcp_secret_ctx VNET(pf_tcp_secret_ctx)
183 VNET_DEFINE(u_char, pf_tcp_secret[16]);
184 #define V_pf_tcp_secret VNET(pf_tcp_secret)
185 VNET_DEFINE(int, pf_tcp_secret_init);
186 #define V_pf_tcp_secret_init VNET(pf_tcp_secret_init)
187 VNET_DEFINE(int, pf_tcp_iss_off);
188 #define V_pf_tcp_iss_off VNET(pf_tcp_iss_off)
189
190 struct pf_anchor_stackframe {
191 struct pf_ruleset *rs;
192 struct pf_rule *r;
193 struct pf_anchor_node *parent;
194 struct pf_anchor *child;
195 };
196 VNET_DEFINE(struct pf_anchor_stackframe, pf_anchor_stack[64]);
197 #define V_pf_anchor_stack VNET(pf_anchor_stack)
198
199 VNET_DEFINE(uma_zone_t, pf_src_tree_pl);
200 VNET_DEFINE(uma_zone_t, pf_rule_pl);
201 VNET_DEFINE(uma_zone_t, pf_pooladdr_pl);
202 VNET_DEFINE(uma_zone_t, pf_state_pl);
203 VNET_DEFINE(uma_zone_t, pf_state_key_pl);
204 VNET_DEFINE(uma_zone_t, pf_state_item_pl);
205 VNET_DEFINE(uma_zone_t, pf_altq_pl);
206 #else
207 struct pf_state_tree pf_statetbl;
208
209 struct pf_altqqueue pf_altqs[2];
210 struct pf_palist pf_pabuf;
211 struct pf_altqqueue *pf_altqs_active;
212 struct pf_altqqueue *pf_altqs_inactive;
213 struct pf_status pf_status;
214
215 u_int32_t ticket_altqs_active;
216 u_int32_t ticket_altqs_inactive;
217 int altqs_inactive_open;
218 u_int32_t ticket_pabuf;
219
220 MD5_CTX pf_tcp_secret_ctx;
221 u_char pf_tcp_secret[16];
222 int pf_tcp_secret_init;
223 int pf_tcp_iss_off;
224
225 struct pf_anchor_stackframe {
226 struct pf_ruleset *rs;
227 struct pf_rule *r;
228 struct pf_anchor_node *parent;
229 struct pf_anchor *child;
230 } pf_anchor_stack[64];
231
232 struct pool pf_src_tree_pl, pf_rule_pl, pf_pooladdr_pl;
233 struct pool pf_state_pl, pf_state_key_pl, pf_state_item_pl;
234 struct pool pf_altq_pl;
235 #endif
236
237 void pf_init_threshold(struct pf_threshold *, u_int32_t,
238 u_int32_t);
239 void pf_add_threshold(struct pf_threshold *);
240 int pf_check_threshold(struct pf_threshold *);
241
242 void pf_change_ap(struct mbuf *, struct pf_addr *, u_int16_t *,
243 u_int16_t *, u_int16_t *, struct pf_addr *,
244 u_int16_t, u_int8_t, sa_family_t);
245 int pf_modulate_sack(struct mbuf *, int, struct pf_pdesc *,
246 struct tcphdr *, struct pf_state_peer *);
247 #ifdef INET6
248 void pf_change_a6(struct pf_addr *, u_int16_t *,
249 struct pf_addr *, u_int8_t);
250 #endif /* INET6 */
251 void pf_change_icmp(struct pf_addr *, u_int16_t *,
252 struct pf_addr *, struct pf_addr *, u_int16_t,
253 u_int16_t *, u_int16_t *, u_int16_t *,
254 u_int16_t *, u_int8_t, sa_family_t);
255 #ifdef __FreeBSD__
256 void pf_send_tcp(struct mbuf *,
257 const struct pf_rule *, sa_family_t,
258 #else
259 void pf_send_tcp(const struct pf_rule *, sa_family_t,
260 #endif
261 const struct pf_addr *, const struct pf_addr *,
262 u_int16_t, u_int16_t, u_int32_t, u_int32_t,
263 u_int8_t, u_int16_t, u_int16_t, u_int8_t, int,
264 u_int16_t, struct ether_header *, struct ifnet *);
265 static void pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t,
266 sa_family_t, struct pf_rule *);
267 void pf_detach_state(struct pf_state *);
268 void pf_state_key_detach(struct pf_state *, int);
269 u_int32_t pf_tcp_iss(struct pf_pdesc *);
270 int pf_test_rule(struct pf_rule **, struct pf_state **,
271 int, struct pfi_kif *, struct mbuf *, int,
272 void *, struct pf_pdesc *, struct pf_rule **,
273 #ifdef __FreeBSD__
274 struct pf_ruleset **, struct ifqueue *,
275 struct inpcb *);
276 #else
277 struct pf_ruleset **, struct ifqueue *);
278 #endif
279 static __inline int pf_create_state(struct pf_rule *, struct pf_rule *,
280 struct pf_rule *, struct pf_pdesc *,
281 struct pf_src_node *, struct pf_state_key *,
282 struct pf_state_key *, struct pf_state_key *,
283 struct pf_state_key *, struct mbuf *, int,
284 u_int16_t, u_int16_t, int *, struct pfi_kif *,
285 struct pf_state **, int, u_int16_t, u_int16_t,
286 int);
287 int pf_test_fragment(struct pf_rule **, int,
288 struct pfi_kif *, struct mbuf *, void *,
289 struct pf_pdesc *, struct pf_rule **,
290 struct pf_ruleset **);
291 int pf_tcp_track_full(struct pf_state_peer *,
292 struct pf_state_peer *, struct pf_state **,
293 struct pfi_kif *, struct mbuf *, int,
294 struct pf_pdesc *, u_short *, int *);
295 int pf_tcp_track_sloppy(struct pf_state_peer *,
296 struct pf_state_peer *, struct pf_state **,
297 struct pf_pdesc *, u_short *);
298 int pf_test_state_tcp(struct pf_state **, int,
299 struct pfi_kif *, struct mbuf *, int,
300 void *, struct pf_pdesc *, u_short *);
301 int pf_test_state_udp(struct pf_state **, int,
302 struct pfi_kif *, struct mbuf *, int,
303 void *, struct pf_pdesc *);
304 int pf_test_state_icmp(struct pf_state **, int,
305 struct pfi_kif *, struct mbuf *, int,
306 void *, struct pf_pdesc *, u_short *);
307 int pf_test_state_other(struct pf_state **, int,
308 struct pfi_kif *, struct mbuf *, struct pf_pdesc *);
309 void pf_route(struct mbuf **, struct pf_rule *, int,
310 struct ifnet *, struct pf_state *,
311 struct pf_pdesc *);
312 void pf_route6(struct mbuf **, struct pf_rule *, int,
313 struct ifnet *, struct pf_state *,
314 struct pf_pdesc *);
315 #ifndef __FreeBSD__
316 int pf_socket_lookup(int, struct pf_pdesc *);
317 #endif
318 u_int8_t pf_get_wscale(struct mbuf *, int, u_int16_t,
319 sa_family_t);
320 u_int16_t pf_get_mss(struct mbuf *, int, u_int16_t,
321 sa_family_t);
322 u_int16_t pf_calc_mss(struct pf_addr *, sa_family_t,
323 int, u_int16_t);
324 void pf_set_rt_ifp(struct pf_state *,
325 struct pf_addr *);
326 int pf_check_proto_cksum(struct mbuf *, int, int,
327 u_int8_t, sa_family_t);
328 #ifndef __FreeBSD__
329 struct pf_divert *pf_get_divert(struct mbuf *);
330 #endif
331 void pf_print_state_parts(struct pf_state *,
332 struct pf_state_key *, struct pf_state_key *);
333 int pf_addr_wrap_neq(struct pf_addr_wrap *,
334 struct pf_addr_wrap *);
335 int pf_compare_state_keys(struct pf_state_key *,
336 struct pf_state_key *, struct pfi_kif *, u_int);
337 #ifdef __FreeBSD__
338 struct pf_state *pf_find_state(struct pfi_kif *,
339 struct pf_state_key_cmp *, u_int, struct mbuf *,
340 struct pf_mtag *);
341 #else
342 struct pf_state *pf_find_state(struct pfi_kif *,
343 struct pf_state_key_cmp *, u_int, struct mbuf *);
344 #endif
345 int pf_src_connlimit(struct pf_state **);
346 int pf_check_congestion(struct ifqueue *);
347
348 #ifdef __FreeBSD__
349 int in4_cksum(struct mbuf *m, u_int8_t nxt, int off, int len);
350
351 VNET_DECLARE(int, pf_end_threads);
352
353 VNET_DEFINE(struct pf_pool_limit, pf_pool_limits[PF_LIMIT_MAX]);
354 #else
355 extern struct pool pfr_ktable_pl;
356 extern struct pool pfr_kentry_pl;
357
358 struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] = {
359 { &pf_state_pl, PFSTATE_HIWAT },
360 { &pf_src_tree_pl, PFSNODE_HIWAT },
361 { &pf_frent_pl, PFFRAG_FRENT_HIWAT },
362 { &pfr_ktable_pl, PFR_KTABLE_HIWAT },
363 { &pfr_kentry_pl, PFR_KENTRY_HIWAT }
364 };
365 #endif
366
367 #ifdef __FreeBSD__
368 #define PPACKET_LOOPED() \
369 (pd->pf_mtag->flags & PF_PACKET_LOOPED)
370
371 #define PACKET_LOOPED() \
372 (pd.pf_mtag->flags & PF_PACKET_LOOPED)
373
374 #define STATE_LOOKUP(i, k, d, s, m, pt) \
375 do { \
376 s = pf_find_state(i, k, d, m, pt); \
377 if (s == NULL || (s)->timeout == PFTM_PURGE) \
378 return (PF_DROP); \
379 if (PPACKET_LOOPED()) \
380 return (PF_PASS); \
381 if (d == PF_OUT && \
382 (((s)->rule.ptr->rt == PF_ROUTETO && \
383 (s)->rule.ptr->direction == PF_OUT) || \
384 ((s)->rule.ptr->rt == PF_REPLYTO && \
385 (s)->rule.ptr->direction == PF_IN)) && \
386 (s)->rt_kif != NULL && \
387 (s)->rt_kif != i) \
388 return (PF_PASS); \
389 } while (0)
390 #else
391 #define STATE_LOOKUP(i, k, d, s, m) \
392 do { \
393 s = pf_find_state(i, k, d, m); \
394 if (s == NULL || (s)->timeout == PFTM_PURGE) \
395 return (PF_DROP); \
396 if (d == PF_OUT && \
397 (((s)->rule.ptr->rt == PF_ROUTETO && \
398 (s)->rule.ptr->direction == PF_OUT) || \
399 ((s)->rule.ptr->rt == PF_REPLYTO && \
400 (s)->rule.ptr->direction == PF_IN)) && \
401 (s)->rt_kif != NULL && \
402 (s)->rt_kif != i) \
403 return (PF_PASS); \
404 } while (0)
405 #endif
406
407 #ifdef __FreeBSD__
408 #define BOUND_IFACE(r, k) \
409 ((r)->rule_flag & PFRULE_IFBOUND) ? (k) : V_pfi_all
410 #else
411 #define BOUND_IFACE(r, k) \
412 ((r)->rule_flag & PFRULE_IFBOUND) ? (k) : pfi_all
413 #endif
414
415 #define STATE_INC_COUNTERS(s) \
416 do { \
417 s->rule.ptr->states_cur++; \
418 s->rule.ptr->states_tot++; \
419 if (s->anchor.ptr != NULL) { \
420 s->anchor.ptr->states_cur++; \
421 s->anchor.ptr->states_tot++; \
422 } \
423 if (s->nat_rule.ptr != NULL) { \
424 s->nat_rule.ptr->states_cur++; \
425 s->nat_rule.ptr->states_tot++; \
426 } \
427 } while (0)
428
429 #define STATE_DEC_COUNTERS(s) \
430 do { \
431 if (s->nat_rule.ptr != NULL) \
432 s->nat_rule.ptr->states_cur--; \
433 if (s->anchor.ptr != NULL) \
434 s->anchor.ptr->states_cur--; \
435 s->rule.ptr->states_cur--; \
436 } while (0)
437
438 static __inline int pf_src_compare(struct pf_src_node *, struct pf_src_node *);
439 static __inline int pf_state_compare_key(struct pf_state_key *,
440 struct pf_state_key *);
441 static __inline int pf_state_compare_id(struct pf_state *,
442 struct pf_state *);
443
444 #ifdef __FreeBSD__
445 VNET_DEFINE(struct pf_src_tree, tree_src_tracking);
446
447 VNET_DEFINE(struct pf_state_tree_id, tree_id);
448 VNET_DEFINE(struct pf_state_queue, state_list);
449 #else
450 struct pf_src_tree tree_src_tracking;
451
452 struct pf_state_tree_id tree_id;
453 struct pf_state_queue state_list;
454 #endif
455
456 RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare);
457 RB_GENERATE(pf_state_tree, pf_state_key, entry, pf_state_compare_key);
458 RB_GENERATE(pf_state_tree_id, pf_state,
459 entry_id, pf_state_compare_id);
460
461 static __inline int
pf_src_compare(struct pf_src_node * a,struct pf_src_node * b)462 pf_src_compare(struct pf_src_node *a, struct pf_src_node *b)
463 {
464 int diff;
465
466 if (a->rule.ptr > b->rule.ptr)
467 return (1);
468 if (a->rule.ptr < b->rule.ptr)
469 return (-1);
470 if ((diff = a->af - b->af) != 0)
471 return (diff);
472 switch (a->af) {
473 #ifdef INET
474 case AF_INET:
475 if (a->addr.addr32[0] > b->addr.addr32[0])
476 return (1);
477 if (a->addr.addr32[0] < b->addr.addr32[0])
478 return (-1);
479 break;
480 #endif /* INET */
481 #ifdef INET6
482 case AF_INET6:
483 if (a->addr.addr32[3] > b->addr.addr32[3])
484 return (1);
485 if (a->addr.addr32[3] < b->addr.addr32[3])
486 return (-1);
487 if (a->addr.addr32[2] > b->addr.addr32[2])
488 return (1);
489 if (a->addr.addr32[2] < b->addr.addr32[2])
490 return (-1);
491 if (a->addr.addr32[1] > b->addr.addr32[1])
492 return (1);
493 if (a->addr.addr32[1] < b->addr.addr32[1])
494 return (-1);
495 if (a->addr.addr32[0] > b->addr.addr32[0])
496 return (1);
497 if (a->addr.addr32[0] < b->addr.addr32[0])
498 return (-1);
499 break;
500 #endif /* INET6 */
501 }
502 return (0);
503 }
504
505 #ifdef INET6
506 void
pf_addrcpy(struct pf_addr * dst,struct pf_addr * src,sa_family_t af)507 pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af)
508 {
509 switch (af) {
510 #ifdef INET
511 case AF_INET:
512 dst->addr32[0] = src->addr32[0];
513 break;
514 #endif /* INET */
515 case AF_INET6:
516 dst->addr32[0] = src->addr32[0];
517 dst->addr32[1] = src->addr32[1];
518 dst->addr32[2] = src->addr32[2];
519 dst->addr32[3] = src->addr32[3];
520 break;
521 }
522 }
523 #endif /* INET6 */
524
525 void
pf_init_threshold(struct pf_threshold * threshold,u_int32_t limit,u_int32_t seconds)526 pf_init_threshold(struct pf_threshold *threshold,
527 u_int32_t limit, u_int32_t seconds)
528 {
529 threshold->limit = limit * PF_THRESHOLD_MULT;
530 threshold->seconds = seconds;
531 threshold->count = 0;
532 threshold->last = time_second;
533 }
534
535 void
pf_add_threshold(struct pf_threshold * threshold)536 pf_add_threshold(struct pf_threshold *threshold)
537 {
538 u_int32_t t = time_second, diff = t - threshold->last;
539
540 if (diff >= threshold->seconds)
541 threshold->count = 0;
542 else
543 threshold->count -= threshold->count * diff /
544 threshold->seconds;
545 threshold->count += PF_THRESHOLD_MULT;
546 threshold->last = t;
547 }
548
549 int
pf_check_threshold(struct pf_threshold * threshold)550 pf_check_threshold(struct pf_threshold *threshold)
551 {
552 return (threshold->count > threshold->limit);
553 }
554
555 int
pf_src_connlimit(struct pf_state ** state)556 pf_src_connlimit(struct pf_state **state)
557 {
558 int bad = 0;
559
560 (*state)->src_node->conn++;
561 (*state)->src.tcp_est = 1;
562 pf_add_threshold(&(*state)->src_node->conn_rate);
563
564 if ((*state)->rule.ptr->max_src_conn &&
565 (*state)->rule.ptr->max_src_conn <
566 (*state)->src_node->conn) {
567 #ifdef __FreeBSD__
568 V_pf_status.lcounters[LCNT_SRCCONN]++;
569 #else
570 pf_status.lcounters[LCNT_SRCCONN]++;
571 #endif
572 bad++;
573 }
574
575 if ((*state)->rule.ptr->max_src_conn_rate.limit &&
576 pf_check_threshold(&(*state)->src_node->conn_rate)) {
577 #ifdef __FreeBSD__
578 V_pf_status.lcounters[LCNT_SRCCONNRATE]++;
579 #else
580 pf_status.lcounters[LCNT_SRCCONNRATE]++;
581 #endif
582 bad++;
583 }
584
585 if (!bad)
586 return (0);
587
588 if ((*state)->rule.ptr->overload_tbl) {
589 struct pfr_addr p;
590 u_int32_t killed = 0;
591
592 #ifdef __FreeBSD__
593 V_pf_status.lcounters[LCNT_OVERLOAD_TABLE]++;
594 if (V_pf_status.debug >= PF_DEBUG_MISC) {
595 #else
596 pf_status.lcounters[LCNT_OVERLOAD_TABLE]++;
597 if (pf_status.debug >= PF_DEBUG_MISC) {
598 #endif
599 printf("pf_src_connlimit: blocking address ");
600 pf_print_host(&(*state)->src_node->addr, 0,
601 (*state)->key[PF_SK_WIRE]->af);
602 }
603
604 bzero(&p, sizeof(p));
605 p.pfra_af = (*state)->key[PF_SK_WIRE]->af;
606 switch ((*state)->key[PF_SK_WIRE]->af) {
607 #ifdef INET
608 case AF_INET:
609 p.pfra_net = 32;
610 p.pfra_ip4addr = (*state)->src_node->addr.v4;
611 break;
612 #endif /* INET */
613 #ifdef INET6
614 case AF_INET6:
615 p.pfra_net = 128;
616 p.pfra_ip6addr = (*state)->src_node->addr.v6;
617 break;
618 #endif /* INET6 */
619 }
620
621 pfr_insert_kentry((*state)->rule.ptr->overload_tbl,
622 &p, time_second);
623
624 /* kill existing states if that's required. */
625 if ((*state)->rule.ptr->flush) {
626 struct pf_state_key *sk;
627 struct pf_state *st;
628
629 #ifdef __FreeBSD__
630 V_pf_status.lcounters[LCNT_OVERLOAD_FLUSH]++;
631 RB_FOREACH(st, pf_state_tree_id, &V_tree_id) {
632 #else
633 pf_status.lcounters[LCNT_OVERLOAD_FLUSH]++;
634 RB_FOREACH(st, pf_state_tree_id, &tree_id) {
635 #endif
636 sk = st->key[PF_SK_WIRE];
637 /*
638 * Kill states from this source. (Only those
639 * from the same rule if PF_FLUSH_GLOBAL is not
640 * set)
641 */
642 if (sk->af ==
643 (*state)->key[PF_SK_WIRE]->af &&
644 (((*state)->direction == PF_OUT &&
645 PF_AEQ(&(*state)->src_node->addr,
646 &sk->addr[1], sk->af)) ||
647 ((*state)->direction == PF_IN &&
648 PF_AEQ(&(*state)->src_node->addr,
649 &sk->addr[0], sk->af))) &&
650 ((*state)->rule.ptr->flush &
651 PF_FLUSH_GLOBAL ||
652 (*state)->rule.ptr == st->rule.ptr)) {
653 st->timeout = PFTM_PURGE;
654 st->src.state = st->dst.state =
655 TCPS_CLOSED;
656 killed++;
657 }
658 }
659 #ifdef __FreeBSD__
660 if (V_pf_status.debug >= PF_DEBUG_MISC)
661 #else
662 if (pf_status.debug >= PF_DEBUG_MISC)
663 #endif
664 printf(", %u states killed", killed);
665 }
666 #ifdef __FreeBSD__
667 if (V_pf_status.debug >= PF_DEBUG_MISC)
668 #else
669 if (pf_status.debug >= PF_DEBUG_MISC)
670 #endif
671 printf("\n");
672 }
673
674 /* kill this state */
675 (*state)->timeout = PFTM_PURGE;
676 (*state)->src.state = (*state)->dst.state = TCPS_CLOSED;
677 return (1);
678 }
679
680 int
681 pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule,
682 struct pf_addr *src, sa_family_t af)
683 {
684 struct pf_src_node k;
685
686 if (*sn == NULL) {
687 k.af = af;
688 PF_ACPY(&k.addr, src, af);
689 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
690 rule->rpool.opts & PF_POOL_STICKYADDR)
691 k.rule.ptr = rule;
692 else
693 k.rule.ptr = NULL;
694 #ifdef __FreeBSD__
695 V_pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
696 *sn = RB_FIND(pf_src_tree, &V_tree_src_tracking, &k);
697 #else
698 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
699 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
700 #endif
701 }
702 if (*sn == NULL) {
703 if (!rule->max_src_nodes ||
704 rule->src_nodes < rule->max_src_nodes)
705 #ifdef __FreeBSD__
706 (*sn) = pool_get(&V_pf_src_tree_pl, PR_NOWAIT | PR_ZERO);
707 #else
708 (*sn) = pool_get(&pf_src_tree_pl, PR_NOWAIT | PR_ZERO);
709 #endif
710 else
711 #ifdef __FreeBSD__
712 V_pf_status.lcounters[LCNT_SRCNODES]++;
713 #else
714 pf_status.lcounters[LCNT_SRCNODES]++;
715 #endif
716 if ((*sn) == NULL)
717 return (-1);
718
719 pf_init_threshold(&(*sn)->conn_rate,
720 rule->max_src_conn_rate.limit,
721 rule->max_src_conn_rate.seconds);
722
723 (*sn)->af = af;
724 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
725 rule->rpool.opts & PF_POOL_STICKYADDR)
726 (*sn)->rule.ptr = rule;
727 else
728 (*sn)->rule.ptr = NULL;
729 PF_ACPY(&(*sn)->addr, src, af);
730 if (RB_INSERT(pf_src_tree,
731 #ifdef __FreeBSD__
732 &V_tree_src_tracking, *sn) != NULL) {
733 if (V_pf_status.debug >= PF_DEBUG_MISC) {
734 #else
735 &tree_src_tracking, *sn) != NULL) {
736 if (pf_status.debug >= PF_DEBUG_MISC) {
737 #endif
738 printf("pf: src_tree insert failed: ");
739 pf_print_host(&(*sn)->addr, 0, af);
740 printf("\n");
741 }
742 #ifdef __FreeBSD__
743 pool_put(&V_pf_src_tree_pl, *sn);
744 #else
745 pool_put(&pf_src_tree_pl, *sn);
746 #endif
747 return (-1);
748 }
749 (*sn)->creation = time_second;
750 (*sn)->ruletype = rule->action;
751 if ((*sn)->rule.ptr != NULL)
752 (*sn)->rule.ptr->src_nodes++;
753 #ifdef __FreeBSD__
754 V_pf_status.scounters[SCNT_SRC_NODE_INSERT]++;
755 V_pf_status.src_nodes++;
756 #else
757 pf_status.scounters[SCNT_SRC_NODE_INSERT]++;
758 pf_status.src_nodes++;
759 #endif
760 } else {
761 if (rule->max_src_states &&
762 (*sn)->states >= rule->max_src_states) {
763 #ifdef __FreeBSD__
764 V_pf_status.lcounters[LCNT_SRCSTATES]++;
765 #else
766 pf_status.lcounters[LCNT_SRCSTATES]++;
767 #endif
768 return (-1);
769 }
770 }
771 return (0);
772 }
773
774 /* state table stuff */
775
776 static __inline int
777 pf_state_compare_key(struct pf_state_key *a, struct pf_state_key *b)
778 {
779 int diff;
780
781 if ((diff = a->proto - b->proto) != 0)
782 return (diff);
783 if ((diff = a->af - b->af) != 0)
784 return (diff);
785 switch (a->af) {
786 #ifdef INET
787 case AF_INET:
788 if (a->addr[0].addr32[0] > b->addr[0].addr32[0])
789 return (1);
790 if (a->addr[0].addr32[0] < b->addr[0].addr32[0])
791 return (-1);
792 if (a->addr[1].addr32[0] > b->addr[1].addr32[0])
793 return (1);
794 if (a->addr[1].addr32[0] < b->addr[1].addr32[0])
795 return (-1);
796 break;
797 #endif /* INET */
798 #ifdef INET6
799 case AF_INET6:
800 if (a->addr[0].addr32[3] > b->addr[0].addr32[3])
801 return (1);
802 if (a->addr[0].addr32[3] < b->addr[0].addr32[3])
803 return (-1);
804 if (a->addr[1].addr32[3] > b->addr[1].addr32[3])
805 return (1);
806 if (a->addr[1].addr32[3] < b->addr[1].addr32[3])
807 return (-1);
808 if (a->addr[0].addr32[2] > b->addr[0].addr32[2])
809 return (1);
810 if (a->addr[0].addr32[2] < b->addr[0].addr32[2])
811 return (-1);
812 if (a->addr[1].addr32[2] > b->addr[1].addr32[2])
813 return (1);
814 if (a->addr[1].addr32[2] < b->addr[1].addr32[2])
815 return (-1);
816 if (a->addr[0].addr32[1] > b->addr[0].addr32[1])
817 return (1);
818 if (a->addr[0].addr32[1] < b->addr[0].addr32[1])
819 return (-1);
820 if (a->addr[1].addr32[1] > b->addr[1].addr32[1])
821 return (1);
822 if (a->addr[1].addr32[1] < b->addr[1].addr32[1])
823 return (-1);
824 if (a->addr[0].addr32[0] > b->addr[0].addr32[0])
825 return (1);
826 if (a->addr[0].addr32[0] < b->addr[0].addr32[0])
827 return (-1);
828 if (a->addr[1].addr32[0] > b->addr[1].addr32[0])
829 return (1);
830 if (a->addr[1].addr32[0] < b->addr[1].addr32[0])
831 return (-1);
832 break;
833 #endif /* INET6 */
834 }
835
836 if ((diff = a->port[0] - b->port[0]) != 0)
837 return (diff);
838 if ((diff = a->port[1] - b->port[1]) != 0)
839 return (diff);
840
841 return (0);
842 }
843
844 static __inline int
845 pf_state_compare_id(struct pf_state *a, struct pf_state *b)
846 {
847 if (a->id > b->id)
848 return (1);
849 if (a->id < b->id)
850 return (-1);
851 if (a->creatorid > b->creatorid)
852 return (1);
853 if (a->creatorid < b->creatorid)
854 return (-1);
855
856 return (0);
857 }
858
859 int
860 pf_state_key_attach(struct pf_state_key *sk, struct pf_state *s, int idx)
861 {
862 struct pf_state_item *si;
863 struct pf_state_key *cur;
864 struct pf_state *olds = NULL;
865
866 #ifdef __FreeBSD__
867 KASSERT(s->key[idx] == NULL, ("%s: key is null!", __FUNCTION__));
868 #else
869 KASSERT(s->key[idx] == NULL); /* XXX handle this? */
870 #endif
871
872 #ifdef __FreeBSD__
873 if ((cur = RB_INSERT(pf_state_tree, &V_pf_statetbl, sk)) != NULL) {
874 #else
875 if ((cur = RB_INSERT(pf_state_tree, &pf_statetbl, sk)) != NULL) {
876 #endif
877 /* key exists. check for same kif, if none, add to key */
878 TAILQ_FOREACH(si, &cur->states, entry)
879 if (si->s->kif == s->kif &&
880 si->s->direction == s->direction) {
881 if (sk->proto == IPPROTO_TCP &&
882 si->s->src.state >= TCPS_FIN_WAIT_2 &&
883 si->s->dst.state >= TCPS_FIN_WAIT_2) {
884 si->s->src.state = si->s->dst.state =
885 TCPS_CLOSED;
886 /* unlink late or sks can go away */
887 olds = si->s;
888 } else {
889 #ifdef __FreeBSD__
890 if (V_pf_status.debug >= PF_DEBUG_MISC) {
891 #else
892 if (pf_status.debug >= PF_DEBUG_MISC) {
893 #endif
894 printf("pf: %s key attach "
895 "failed on %s: ",
896 (idx == PF_SK_WIRE) ?
897 "wire" : "stack",
898 s->kif->pfik_name);
899 pf_print_state_parts(s,
900 (idx == PF_SK_WIRE) ?
901 sk : NULL,
902 (idx == PF_SK_STACK) ?
903 sk : NULL);
904 printf(", existing: ");
905 pf_print_state_parts(si->s,
906 (idx == PF_SK_WIRE) ?
907 sk : NULL,
908 (idx == PF_SK_STACK) ?
909 sk : NULL);
910 printf("\n");
911 }
912 #ifdef __FreeBSD__
913 pool_put(&V_pf_state_key_pl, sk);
914 #else
915 pool_put(&pf_state_key_pl, sk);
916 #endif
917 return (-1); /* collision! */
918 }
919 }
920 #ifdef __FreeBSD__
921 pool_put(&V_pf_state_key_pl, sk);
922 #else
923 pool_put(&pf_state_key_pl, sk);
924 #endif
925 s->key[idx] = cur;
926 } else
927 s->key[idx] = sk;
928
929 #ifdef __FreeBSD__
930 if ((si = pool_get(&V_pf_state_item_pl, PR_NOWAIT)) == NULL) {
931 #else
932 if ((si = pool_get(&pf_state_item_pl, PR_NOWAIT)) == NULL) {
933 #endif
934 pf_state_key_detach(s, idx);
935 return (-1);
936 }
937 si->s = s;
938
939 /* list is sorted, if-bound states before floating */
940 #ifdef __FreeBSD__
941 if (s->kif == V_pfi_all)
942 #else
943 if (s->kif == pfi_all)
944 #endif
945 TAILQ_INSERT_TAIL(&s->key[idx]->states, si, entry);
946 else
947 TAILQ_INSERT_HEAD(&s->key[idx]->states, si, entry);
948
949 if (olds)
950 pf_unlink_state(olds);
951
952 return (0);
953 }
954
955 void
956 pf_detach_state(struct pf_state *s)
957 {
958 if (s->key[PF_SK_WIRE] == s->key[PF_SK_STACK])
959 s->key[PF_SK_WIRE] = NULL;
960
961 if (s->key[PF_SK_STACK] != NULL)
962 pf_state_key_detach(s, PF_SK_STACK);
963
964 if (s->key[PF_SK_WIRE] != NULL)
965 pf_state_key_detach(s, PF_SK_WIRE);
966 }
967
968 void
969 pf_state_key_detach(struct pf_state *s, int idx)
970 {
971 struct pf_state_item *si;
972
973 si = TAILQ_FIRST(&s->key[idx]->states);
974 while (si && si->s != s)
975 si = TAILQ_NEXT(si, entry);
976
977 if (si) {
978 TAILQ_REMOVE(&s->key[idx]->states, si, entry);
979 #ifdef __FreeBSD__
980 pool_put(&V_pf_state_item_pl, si);
981 #else
982 pool_put(&pf_state_item_pl, si);
983 #endif
984 }
985
986 if (TAILQ_EMPTY(&s->key[idx]->states)) {
987 #ifdef __FreeBSD__
988 RB_REMOVE(pf_state_tree, &V_pf_statetbl, s->key[idx]);
989 #else
990 RB_REMOVE(pf_state_tree, &pf_statetbl, s->key[idx]);
991 #endif
992 if (s->key[idx]->reverse)
993 s->key[idx]->reverse->reverse = NULL;
994 #ifdef __FreeBSD__
995 /* XXX: implement this */
996 #else
997 if (s->key[idx]->inp)
998 s->key[idx]->inp->inp_pf_sk = NULL;
999 #endif
1000 #ifdef __FreeBSD__
1001 pool_put(&V_pf_state_key_pl, s->key[idx]);
1002 #else
1003 pool_put(&pf_state_key_pl, s->key[idx]);
1004 #endif
1005 }
1006 s->key[idx] = NULL;
1007 }
1008
1009 struct pf_state_key *
1010 pf_alloc_state_key(int pool_flags)
1011 {
1012 struct pf_state_key *sk;
1013
1014 #ifdef __FreeBSD__
1015 if ((sk = pool_get(&V_pf_state_key_pl, pool_flags)) == NULL)
1016 #else
1017 if ((sk = pool_get(&pf_state_key_pl, pool_flags)) == NULL)
1018 #endif
1019 return (NULL);
1020 TAILQ_INIT(&sk->states);
1021
1022 return (sk);
1023 }
1024
1025 int
1026 pf_state_key_setup(struct pf_pdesc *pd, struct pf_rule *nr,
1027 struct pf_state_key **skw, struct pf_state_key **sks,
1028 struct pf_state_key **skp, struct pf_state_key **nkp,
1029 struct pf_addr *saddr, struct pf_addr *daddr,
1030 u_int16_t sport, u_int16_t dport)
1031 {
1032 #ifdef __FreeBSD__
1033 KASSERT((*skp == NULL && *nkp == NULL),
1034 ("%s: skp == NULL && nkp == NULL", __FUNCTION__));
1035 #else
1036 KASSERT((*skp == NULL && *nkp == NULL));
1037 #endif
1038
1039 if ((*skp = pf_alloc_state_key(PR_NOWAIT | PR_ZERO)) == NULL)
1040 return (ENOMEM);
1041
1042 PF_ACPY(&(*skp)->addr[pd->sidx], saddr, pd->af);
1043 PF_ACPY(&(*skp)->addr[pd->didx], daddr, pd->af);
1044 (*skp)->port[pd->sidx] = sport;
1045 (*skp)->port[pd->didx] = dport;
1046 (*skp)->proto = pd->proto;
1047 (*skp)->af = pd->af;
1048
1049 if (nr != NULL) {
1050 if ((*nkp = pf_alloc_state_key(PR_NOWAIT | PR_ZERO)) == NULL)
1051 return (ENOMEM); /* caller must handle cleanup */
1052
1053 /* XXX maybe just bcopy and TAILQ_INIT(&(*nkp)->states) */
1054 PF_ACPY(&(*nkp)->addr[0], &(*skp)->addr[0], pd->af);
1055 PF_ACPY(&(*nkp)->addr[1], &(*skp)->addr[1], pd->af);
1056 (*nkp)->port[0] = (*skp)->port[0];
1057 (*nkp)->port[1] = (*skp)->port[1];
1058 (*nkp)->proto = pd->proto;
1059 (*nkp)->af = pd->af;
1060 } else
1061 *nkp = *skp;
1062
1063 if (pd->dir == PF_IN) {
1064 *skw = *skp;
1065 *sks = *nkp;
1066 } else {
1067 *sks = *skp;
1068 *skw = *nkp;
1069 }
1070 return (0);
1071 }
1072
1073
1074 int
1075 pf_state_insert(struct pfi_kif *kif, struct pf_state_key *skw,
1076 struct pf_state_key *sks, struct pf_state *s)
1077 {
1078 #ifndef __FreeBSD__
1079 splassert(IPL_SOFTNET);
1080 #endif
1081
1082 s->kif = kif;
1083
1084 if (skw == sks) {
1085 if (pf_state_key_attach(skw, s, PF_SK_WIRE))
1086 return (-1);
1087 s->key[PF_SK_STACK] = s->key[PF_SK_WIRE];
1088 } else {
1089 if (pf_state_key_attach(skw, s, PF_SK_WIRE)) {
1090 #ifdef __FreeBSD__
1091 pool_put(&V_pf_state_key_pl, sks);
1092 #else
1093 pool_put(&pf_state_key_pl, sks);
1094 #endif
1095 return (-1);
1096 }
1097 if (pf_state_key_attach(sks, s, PF_SK_STACK)) {
1098 pf_state_key_detach(s, PF_SK_WIRE);
1099 return (-1);
1100 }
1101 }
1102
1103 if (s->id == 0 && s->creatorid == 0) {
1104 #ifdef __FreeBSD__
1105 s->id = htobe64(V_pf_status.stateid++);
1106 s->creatorid = V_pf_status.hostid;
1107 #else
1108 s->id = htobe64(pf_status.stateid++);
1109 s->creatorid = pf_status.hostid;
1110 #endif
1111 }
1112 #ifdef __FreeBSD__
1113 if (RB_INSERT(pf_state_tree_id, &V_tree_id, s) != NULL) {
1114 if (V_pf_status.debug >= PF_DEBUG_MISC) {
1115 #else
1116 if (RB_INSERT(pf_state_tree_id, &tree_id, s) != NULL) {
1117 if (pf_status.debug >= PF_DEBUG_MISC) {
1118 #endif
1119 printf("pf: state insert failed: "
1120 "id: %016llx creatorid: %08x",
1121 #ifdef __FreeBSD__
1122 (unsigned long long)betoh64(s->id), ntohl(s->creatorid));
1123 #else
1124 betoh64(s->id), ntohl(s->creatorid));
1125 #endif
1126 printf("\n");
1127 }
1128 pf_detach_state(s);
1129 return (-1);
1130 }
1131 #ifdef __FreeBSD__
1132 TAILQ_INSERT_TAIL(&V_state_list, s, entry_list);
1133 V_pf_status.fcounters[FCNT_STATE_INSERT]++;
1134 V_pf_status.states++;
1135 #else
1136 TAILQ_INSERT_TAIL(&state_list, s, entry_list);
1137 pf_status.fcounters[FCNT_STATE_INSERT]++;
1138 pf_status.states++;
1139 #endif
1140 pfi_kif_ref(kif, PFI_KIF_REF_STATE);
1141 #if NPFSYNC > 0
1142 #ifdef __FreeBSD__
1143 if (pfsync_insert_state_ptr != NULL)
1144 pfsync_insert_state_ptr(s);
1145 #else
1146 pfsync_insert_state(s);
1147 #endif
1148 #endif
1149 return (0);
1150 }
1151
1152 struct pf_state *
1153 pf_find_state_byid(struct pf_state_cmp *key)
1154 {
1155 #ifdef __FreeBSD__
1156 V_pf_status.fcounters[FCNT_STATE_SEARCH]++;
1157
1158 return (RB_FIND(pf_state_tree_id, &V_tree_id, (struct pf_state *)key));
1159 #else
1160 pf_status.fcounters[FCNT_STATE_SEARCH]++;
1161
1162 return (RB_FIND(pf_state_tree_id, &tree_id, (struct pf_state *)key));
1163 #endif
1164 }
1165
1166 /* XXX debug function, intended to be removed one day */
1167 int
1168 pf_compare_state_keys(struct pf_state_key *a, struct pf_state_key *b,
1169 struct pfi_kif *kif, u_int dir)
1170 {
1171 /* a (from hdr) and b (new) must be exact opposites of each other */
1172 if (a->af == b->af && a->proto == b->proto &&
1173 PF_AEQ(&a->addr[0], &b->addr[1], a->af) &&
1174 PF_AEQ(&a->addr[1], &b->addr[0], a->af) &&
1175 a->port[0] == b->port[1] &&
1176 a->port[1] == b->port[0])
1177 return (0);
1178 else {
1179 /* mismatch. must not happen. */
1180 printf("pf: state key linking mismatch! dir=%s, "
1181 "if=%s, stored af=%u, a0: ",
1182 dir == PF_OUT ? "OUT" : "IN", kif->pfik_name, a->af);
1183 pf_print_host(&a->addr[0], a->port[0], a->af);
1184 printf(", a1: ");
1185 pf_print_host(&a->addr[1], a->port[1], a->af);
1186 printf(", proto=%u", a->proto);
1187 printf(", found af=%u, a0: ", b->af);
1188 pf_print_host(&b->addr[0], b->port[0], b->af);
1189 printf(", a1: ");
1190 pf_print_host(&b->addr[1], b->port[1], b->af);
1191 printf(", proto=%u", b->proto);
1192 printf(".\n");
1193 return (-1);
1194 }
1195 }
1196
1197 struct pf_state *
1198 #ifdef __FreeBSD__
1199 pf_find_state(struct pfi_kif *kif, struct pf_state_key_cmp *key, u_int dir,
1200 struct mbuf *m, struct pf_mtag *pftag)
1201 #else
1202 pf_find_state(struct pfi_kif *kif, struct pf_state_key_cmp *key, u_int dir,
1203 struct mbuf *m)
1204 #endif
1205 {
1206 struct pf_state_key *sk;
1207 struct pf_state_item *si;
1208
1209 #ifdef __FreeBSD__
1210 V_pf_status.fcounters[FCNT_STATE_SEARCH]++;
1211 #else
1212 pf_status.fcounters[FCNT_STATE_SEARCH]++;
1213 #endif
1214
1215 #ifdef __FreeBSD__
1216 if (dir == PF_OUT && pftag->statekey &&
1217 ((struct pf_state_key *)pftag->statekey)->reverse)
1218 sk = ((struct pf_state_key *)pftag->statekey)->reverse;
1219 else {
1220 #ifdef __FreeBSD__
1221 if ((sk = RB_FIND(pf_state_tree, &V_pf_statetbl,
1222 #else
1223 if ((sk = RB_FIND(pf_state_tree, &pf_statetbl,
1224 #endif
1225 (struct pf_state_key *)key)) == NULL)
1226 return (NULL);
1227 if (dir == PF_OUT && pftag->statekey &&
1228 pf_compare_state_keys(pftag->statekey, sk,
1229 kif, dir) == 0) {
1230 ((struct pf_state_key *)
1231 pftag->statekey)->reverse = sk;
1232 sk->reverse = pftag->statekey;
1233 }
1234 }
1235 #else
1236 if (dir == PF_OUT && m->m_pkthdr.pf.statekey &&
1237 ((struct pf_state_key *)m->m_pkthdr.pf.statekey)->reverse)
1238 sk = ((struct pf_state_key *)m->m_pkthdr.pf.statekey)->reverse;
1239 else {
1240 #ifdef __FreeBSD__
1241 if ((sk = RB_FIND(pf_state_tree, &V_pf_statetbl,
1242 #else
1243 if ((sk = RB_FIND(pf_state_tree, &pf_statetbl,
1244 #endif
1245 (struct pf_state_key *)key)) == NULL)
1246 return (NULL);
1247 if (dir == PF_OUT && m->m_pkthdr.pf.statekey &&
1248 pf_compare_state_keys(m->m_pkthdr.pf.statekey, sk,
1249 kif, dir) == 0) {
1250 ((struct pf_state_key *)
1251 m->m_pkthdr.pf.statekey)->reverse = sk;
1252 sk->reverse = m->m_pkthdr.pf.statekey;
1253 }
1254 }
1255 #endif
1256
1257 if (dir == PF_OUT)
1258 #ifdef __FreeBSD__
1259 pftag->statekey = NULL;
1260 #else
1261 m->m_pkthdr.pf.statekey = NULL;
1262 #endif
1263
1264 /* list is sorted, if-bound states before floating ones */
1265 TAILQ_FOREACH(si, &sk->states, entry)
1266 #ifdef __FreeBSD__
1267 if ((si->s->kif == V_pfi_all || si->s->kif == kif) &&
1268 #else
1269 if ((si->s->kif == pfi_all || si->s->kif == kif) &&
1270 #endif
1271 sk == (dir == PF_IN ? si->s->key[PF_SK_WIRE] :
1272 si->s->key[PF_SK_STACK]))
1273 return (si->s);
1274
1275 return (NULL);
1276 }
1277
1278 struct pf_state *
1279 pf_find_state_all(struct pf_state_key_cmp *key, u_int dir, int *more)
1280 {
1281 struct pf_state_key *sk;
1282 struct pf_state_item *si, *ret = NULL;
1283
1284 #ifdef __FreeBSD__
1285 V_pf_status.fcounters[FCNT_STATE_SEARCH]++;
1286 #else
1287 pf_status.fcounters[FCNT_STATE_SEARCH]++;
1288 #endif
1289
1290 #ifdef __FreeBSD__
1291 sk = RB_FIND(pf_state_tree, &V_pf_statetbl, (struct pf_state_key *)key);
1292 #else
1293 sk = RB_FIND(pf_state_tree, &pf_statetbl, (struct pf_state_key *)key);
1294 #endif
1295 if (sk != NULL) {
1296 TAILQ_FOREACH(si, &sk->states, entry)
1297 if (dir == PF_INOUT ||
1298 (sk == (dir == PF_IN ? si->s->key[PF_SK_WIRE] :
1299 si->s->key[PF_SK_STACK]))) {
1300 if (more == NULL)
1301 return (si->s);
1302
1303 if (ret)
1304 (*more)++;
1305 else
1306 ret = si;
1307 }
1308 }
1309 return (ret ? ret->s : NULL);
1310 }
1311
1312 /* END state table stuff */
1313
1314
1315 void
1316 pf_purge_thread(void *v)
1317 {
1318 int nloops = 0, s;
1319 #ifdef __FreeBSD__
1320 int locked;
1321 #endif
1322
1323 CURVNET_SET((struct vnet *)v);
1324
1325 for (;;) {
1326 tsleep(pf_purge_thread, PWAIT, "pftm", 1 * hz);
1327
1328 #ifdef __FreeBSD__
1329 sx_slock(&V_pf_consistency_lock);
1330 PF_LOCK();
1331 locked = 0;
1332
1333 if (V_pf_end_threads) {
1334 PF_UNLOCK();
1335 sx_sunlock(&V_pf_consistency_lock);
1336 sx_xlock(&V_pf_consistency_lock);
1337 PF_LOCK();
1338
1339 pf_purge_expired_states(V_pf_status.states, 1);
1340 pf_purge_expired_fragments();
1341 pf_purge_expired_src_nodes(1);
1342 V_pf_end_threads++;
1343
1344 sx_xunlock(&V_pf_consistency_lock);
1345 PF_UNLOCK();
1346 wakeup(pf_purge_thread);
1347 kproc_exit(0);
1348 }
1349 #endif
1350 s = splsoftnet();
1351
1352 /* process a fraction of the state table every second */
1353 #ifdef __FreeBSD__
1354 if (!pf_purge_expired_states(1 + (V_pf_status.states /
1355 V_pf_default_rule.timeout[PFTM_INTERVAL]), 0)) {
1356 PF_UNLOCK();
1357 sx_sunlock(&V_pf_consistency_lock);
1358 sx_xlock(&V_pf_consistency_lock);
1359 PF_LOCK();
1360 locked = 1;
1361
1362 pf_purge_expired_states(1 + (V_pf_status.states /
1363 V_pf_default_rule.timeout[PFTM_INTERVAL]), 1);
1364 }
1365 #else
1366 pf_purge_expired_states(1 + (pf_status.states
1367 / pf_default_rule.timeout[PFTM_INTERVAL]));
1368 #endif
1369
1370 /* purge other expired types every PFTM_INTERVAL seconds */
1371 #ifdef __FreeBSD__
1372 if (++nloops >= V_pf_default_rule.timeout[PFTM_INTERVAL]) {
1373 #else
1374 if (++nloops >= pf_default_rule.timeout[PFTM_INTERVAL]) {
1375 #endif
1376 pf_purge_expired_fragments();
1377 pf_purge_expired_src_nodes(0);
1378 nloops = 0;
1379 }
1380
1381 splx(s);
1382 #ifdef __FreeBSD__
1383 PF_UNLOCK();
1384 if (locked)
1385 sx_xunlock(&V_pf_consistency_lock);
1386 else
1387 sx_sunlock(&V_pf_consistency_lock);
1388 #endif
1389 }
1390 CURVNET_RESTORE();
1391 }
1392
1393 u_int32_t
1394 pf_state_expires(const struct pf_state *state)
1395 {
1396 u_int32_t timeout;
1397 u_int32_t start;
1398 u_int32_t end;
1399 u_int32_t states;
1400
1401 /* handle all PFTM_* > PFTM_MAX here */
1402 if (state->timeout == PFTM_PURGE)
1403 return (time_second);
1404 if (state->timeout == PFTM_UNTIL_PACKET)
1405 return (0);
1406 #ifdef __FreeBSD__
1407 KASSERT(state->timeout != PFTM_UNLINKED,
1408 ("pf_state_expires: timeout == PFTM_UNLINKED"));
1409 KASSERT((state->timeout < PFTM_MAX),
1410 ("pf_state_expires: timeout > PFTM_MAX"));
1411 #else
1412 KASSERT(state->timeout != PFTM_UNLINKED);
1413 KASSERT(state->timeout < PFTM_MAX);
1414 #endif
1415 timeout = state->rule.ptr->timeout[state->timeout];
1416 if (!timeout)
1417 #ifdef __FreeBSD__
1418 timeout = V_pf_default_rule.timeout[state->timeout];
1419 #else
1420 timeout = pf_default_rule.timeout[state->timeout];
1421 #endif
1422 start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START];
1423 if (start) {
1424 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END];
1425 states = state->rule.ptr->states_cur;
1426 } else {
1427 #ifdef __FreeBSD__
1428 start = V_pf_default_rule.timeout[PFTM_ADAPTIVE_START];
1429 end = V_pf_default_rule.timeout[PFTM_ADAPTIVE_END];
1430 states = V_pf_status.states;
1431 #else
1432 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START];
1433 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END];
1434 states = pf_status.states;
1435 #endif
1436 }
1437 if (end && states > start && start < end) {
1438 if (states < end)
1439 return (state->expire + timeout * (end - states) /
1440 (end - start));
1441 else
1442 return (time_second);
1443 }
1444 return (state->expire + timeout);
1445 }
1446
1447 #ifdef __FreeBSD__
1448 int
1449 pf_purge_expired_src_nodes(int waslocked)
1450 #else
1451 void
1452 pf_purge_expired_src_nodes(int waslocked)
1453 #endif
1454 {
1455 struct pf_src_node *cur, *next;
1456 int locked = waslocked;
1457
1458 #ifdef __FreeBSD__
1459 for (cur = RB_MIN(pf_src_tree, &V_tree_src_tracking); cur; cur = next) {
1460 next = RB_NEXT(pf_src_tree, &V_tree_src_tracking, cur);
1461 #else
1462 for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) {
1463 next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur);
1464 #endif
1465
1466 if (cur->states <= 0 && cur->expire <= time_second) {
1467 if (! locked) {
1468 #ifdef __FreeBSD__
1469 if (!sx_try_upgrade(&V_pf_consistency_lock))
1470 return (0);
1471 #else
1472 rw_enter_write(&pf_consistency_lock);
1473 #endif
1474 next = RB_NEXT(pf_src_tree,
1475 #ifdef __FreeBSD__
1476 &V_tree_src_tracking, cur);
1477 #else
1478 &tree_src_tracking, cur);
1479 #endif
1480 locked = 1;
1481 }
1482 if (cur->rule.ptr != NULL) {
1483 cur->rule.ptr->src_nodes--;
1484 if (cur->rule.ptr->states_cur <= 0 &&
1485 cur->rule.ptr->max_src_nodes <= 0)
1486 pf_rm_rule(NULL, cur->rule.ptr);
1487 }
1488 #ifdef __FreeBSD__
1489 RB_REMOVE(pf_src_tree, &V_tree_src_tracking, cur);
1490 V_pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
1491 V_pf_status.src_nodes--;
1492 pool_put(&V_pf_src_tree_pl, cur);
1493 #else
1494 RB_REMOVE(pf_src_tree, &tree_src_tracking, cur);
1495 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
1496 pf_status.src_nodes--;
1497 pool_put(&pf_src_tree_pl, cur);
1498 #endif
1499 }
1500 }
1501
1502 if (locked && !waslocked)
1503 #ifdef __FreeBSD__
1504 {
1505 sx_downgrade(&V_pf_consistency_lock);
1506 }
1507 return (1);
1508 #else
1509 rw_exit_write(&pf_consistency_lock);
1510 #endif
1511 }
1512
1513 void
1514 pf_src_tree_remove_state(struct pf_state *s)
1515 {
1516 u_int32_t timeout;
1517
1518 if (s->src_node != NULL) {
1519 if (s->src.tcp_est)
1520 --s->src_node->conn;
1521 if (--s->src_node->states <= 0) {
1522 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1523 if (!timeout)
1524 timeout =
1525 #ifdef __FreeBSD__
1526 V_pf_default_rule.timeout[PFTM_SRC_NODE];
1527 #else
1528 pf_default_rule.timeout[PFTM_SRC_NODE];
1529 #endif
1530 s->src_node->expire = time_second + timeout;
1531 }
1532 }
1533 if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) {
1534 if (--s->nat_src_node->states <= 0) {
1535 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1536 if (!timeout)
1537 timeout =
1538 #ifdef __FreeBSD__
1539 V_pf_default_rule.timeout[PFTM_SRC_NODE];
1540 #else
1541 pf_default_rule.timeout[PFTM_SRC_NODE];
1542 #endif
1543 s->nat_src_node->expire = time_second + timeout;
1544 }
1545 }
1546 s->src_node = s->nat_src_node = NULL;
1547 }
1548
1549 /* callers should be at splsoftnet */
1550 void
1551 pf_unlink_state(struct pf_state *cur)
1552 {
1553 #ifdef __FreeBSD__
1554 if (cur->local_flags & PFSTATE_EXPIRING)
1555 return;
1556 cur->local_flags |= PFSTATE_EXPIRING;
1557 #else
1558 splassert(IPL_SOFTNET);
1559 #endif
1560
1561 if (cur->src.state == PF_TCPS_PROXY_DST) {
1562 /* XXX wire key the right one? */
1563 #ifdef __FreeBSD__
1564 pf_send_tcp(NULL, cur->rule.ptr, cur->key[PF_SK_WIRE]->af,
1565 #else
1566 pf_send_tcp(cur->rule.ptr, cur->key[PF_SK_WIRE]->af,
1567 #endif
1568 &cur->key[PF_SK_WIRE]->addr[1],
1569 &cur->key[PF_SK_WIRE]->addr[0],
1570 cur->key[PF_SK_WIRE]->port[1],
1571 cur->key[PF_SK_WIRE]->port[0],
1572 cur->src.seqhi, cur->src.seqlo + 1,
1573 TH_RST|TH_ACK, 0, 0, 0, 1, cur->tag, NULL, NULL);
1574 }
1575 #ifdef __FreeBSD__
1576 RB_REMOVE(pf_state_tree_id, &V_tree_id, cur);
1577 #else
1578 RB_REMOVE(pf_state_tree_id, &tree_id, cur);
1579 #endif
1580 #if NPFLOW > 0
1581 if (cur->state_flags & PFSTATE_PFLOW)
1582 #ifdef __FreeBSD__
1583 if (export_pflow_ptr != NULL)
1584 export_pflow_ptr(cur);
1585 #else
1586 export_pflow(cur);
1587 #endif
1588 #endif
1589 #if NPFSYNC > 0
1590 #ifdef __FreeBSD__
1591 if (pfsync_delete_state_ptr != NULL)
1592 pfsync_delete_state_ptr(cur);
1593 #else
1594 pfsync_delete_state(cur);
1595 #endif
1596 #endif
1597 cur->timeout = PFTM_UNLINKED;
1598 pf_src_tree_remove_state(cur);
1599 pf_detach_state(cur);
1600 }
1601
1602 /* callers should be at splsoftnet and hold the
1603 * write_lock on pf_consistency_lock */
1604 void
1605 pf_free_state(struct pf_state *cur)
1606 {
1607 #ifndef __FreeBSD__
1608 splassert(IPL_SOFTNET);
1609 #endif
1610
1611 #if NPFSYNC > 0
1612 #ifdef __FreeBSD__
1613 if (pfsync_state_in_use_ptr != NULL &&
1614 pfsync_state_in_use_ptr(cur))
1615 #else
1616 if (pfsync_state_in_use(cur))
1617 #endif
1618 return;
1619 #endif
1620 #ifdef __FreeBSD__
1621 KASSERT(cur->timeout == PFTM_UNLINKED,
1622 ("pf_free_state: cur->timeout != PFTM_UNLINKED"));
1623 #else
1624 KASSERT(cur->timeout == PFTM_UNLINKED);
1625 #endif
1626 if (--cur->rule.ptr->states_cur <= 0 &&
1627 cur->rule.ptr->src_nodes <= 0)
1628 pf_rm_rule(NULL, cur->rule.ptr);
1629 if (cur->nat_rule.ptr != NULL)
1630 if (--cur->nat_rule.ptr->states_cur <= 0 &&
1631 cur->nat_rule.ptr->src_nodes <= 0)
1632 pf_rm_rule(NULL, cur->nat_rule.ptr);
1633 if (cur->anchor.ptr != NULL)
1634 if (--cur->anchor.ptr->states_cur <= 0)
1635 pf_rm_rule(NULL, cur->anchor.ptr);
1636 pf_normalize_tcp_cleanup(cur);
1637 pfi_kif_unref(cur->kif, PFI_KIF_REF_STATE);
1638 #ifdef __FreeBSD__
1639 TAILQ_REMOVE(&V_state_list, cur, entry_list);
1640 #else
1641 TAILQ_REMOVE(&state_list, cur, entry_list);
1642 #endif
1643 if (cur->tag)
1644 pf_tag_unref(cur->tag);
1645 #ifdef __FreeBSD__
1646 pool_put(&V_pf_state_pl, cur);
1647 V_pf_status.fcounters[FCNT_STATE_REMOVALS]++;
1648 V_pf_status.states--;
1649 #else
1650 pool_put(&pf_state_pl, cur);
1651 pf_status.fcounters[FCNT_STATE_REMOVALS]++;
1652 pf_status.states--;
1653 #endif
1654 }
1655
1656 #ifdef __FreeBSD__
1657 int
1658 pf_purge_expired_states(u_int32_t maxcheck, int waslocked)
1659 #else
1660 void
1661 pf_purge_expired_states(u_int32_t maxcheck)
1662 #endif
1663 {
1664 static struct pf_state *cur = NULL;
1665 struct pf_state *next;
1666 #ifdef __FreeBSD__
1667 int locked = waslocked;
1668 #else
1669 int locked = 0;
1670 #endif
1671
1672 while (maxcheck--) {
1673 /* wrap to start of list when we hit the end */
1674 if (cur == NULL) {
1675 #ifdef __FreeBSD__
1676 cur = TAILQ_FIRST(&V_state_list);
1677 #else
1678 cur = TAILQ_FIRST(&state_list);
1679 #endif
1680 if (cur == NULL)
1681 break; /* list empty */
1682 }
1683
1684 /* get next state, as cur may get deleted */
1685 next = TAILQ_NEXT(cur, entry_list);
1686
1687 if (cur->timeout == PFTM_UNLINKED) {
1688 /* free unlinked state */
1689 if (! locked) {
1690 #ifdef __FreeBSD__
1691 if (!sx_try_upgrade(&V_pf_consistency_lock))
1692 return (0);
1693 #else
1694 rw_enter_write(&pf_consistency_lock);
1695 #endif
1696 locked = 1;
1697 }
1698 pf_free_state(cur);
1699 } else if (pf_state_expires(cur) <= time_second) {
1700 /* unlink and free expired state */
1701 pf_unlink_state(cur);
1702 if (! locked) {
1703 #ifdef __FreeBSD__
1704 if (!sx_try_upgrade(&V_pf_consistency_lock))
1705 return (0);
1706 #else
1707 rw_enter_write(&pf_consistency_lock);
1708 #endif
1709 locked = 1;
1710 }
1711 pf_free_state(cur);
1712 }
1713 cur = next;
1714 }
1715
1716 #ifdef __FreeBSD__
1717 if (!waslocked && locked)
1718 sx_downgrade(&V_pf_consistency_lock);
1719
1720 return (1);
1721 #else
1722 if (locked)
1723 rw_exit_write(&pf_consistency_lock);
1724 #endif
1725 }
1726
1727 int
1728 pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw)
1729 {
1730 if (aw->type != PF_ADDR_TABLE)
1731 return (0);
1732 if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname, 1)) == NULL)
1733 return (1);
1734 return (0);
1735 }
1736
1737 void
1738 pf_tbladdr_remove(struct pf_addr_wrap *aw)
1739 {
1740 if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL)
1741 return;
1742 pfr_detach_table(aw->p.tbl);
1743 aw->p.tbl = NULL;
1744 }
1745
1746 void
1747 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
1748 {
1749 struct pfr_ktable *kt = aw->p.tbl;
1750
1751 if (aw->type != PF_ADDR_TABLE || kt == NULL)
1752 return;
1753 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
1754 kt = kt->pfrkt_root;
1755 aw->p.tbl = NULL;
1756 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
1757 kt->pfrkt_cnt : -1;
1758 }
1759
1760 void
1761 pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af)
1762 {
1763 switch (af) {
1764 #ifdef INET
1765 case AF_INET: {
1766 u_int32_t a = ntohl(addr->addr32[0]);
1767 printf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255,
1768 (a>>8)&255, a&255);
1769 if (p) {
1770 p = ntohs(p);
1771 printf(":%u", p);
1772 }
1773 break;
1774 }
1775 #endif /* INET */
1776 #ifdef INET6
1777 case AF_INET6: {
1778 u_int16_t b;
1779 u_int8_t i, curstart, curend, maxstart, maxend;
1780 curstart = curend = maxstart = maxend = 255;
1781 for (i = 0; i < 8; i++) {
1782 if (!addr->addr16[i]) {
1783 if (curstart == 255)
1784 curstart = i;
1785 curend = i;
1786 } else {
1787 if ((curend - curstart) >
1788 (maxend - maxstart)) {
1789 maxstart = curstart;
1790 maxend = curend;
1791 }
1792 curstart = curend = 255;
1793 }
1794 }
1795 if ((curend - curstart) >
1796 (maxend - maxstart)) {
1797 maxstart = curstart;
1798 maxend = curend;
1799 }
1800 for (i = 0; i < 8; i++) {
1801 if (i >= maxstart && i <= maxend) {
1802 if (i == 0)
1803 printf(":");
1804 if (i == maxend)
1805 printf(":");
1806 } else {
1807 b = ntohs(addr->addr16[i]);
1808 printf("%x", b);
1809 if (i < 7)
1810 printf(":");
1811 }
1812 }
1813 if (p) {
1814 p = ntohs(p);
1815 printf("[%u]", p);
1816 }
1817 break;
1818 }
1819 #endif /* INET6 */
1820 }
1821 }
1822
1823 void
1824 pf_print_state(struct pf_state *s)
1825 {
1826 pf_print_state_parts(s, NULL, NULL);
1827 }
1828
1829 void
1830 pf_print_state_parts(struct pf_state *s,
1831 struct pf_state_key *skwp, struct pf_state_key *sksp)
1832 {
1833 struct pf_state_key *skw, *sks;
1834 u_int8_t proto, dir;
1835
1836 /* Do our best to fill these, but they're skipped if NULL */
1837 skw = skwp ? skwp : (s ? s->key[PF_SK_WIRE] : NULL);
1838 sks = sksp ? sksp : (s ? s->key[PF_SK_STACK] : NULL);
1839 proto = skw ? skw->proto : (sks ? sks->proto : 0);
1840 dir = s ? s->direction : 0;
1841
1842 switch (proto) {
1843 case IPPROTO_IPV4:
1844 printf("IPv4");
1845 break;
1846 case IPPROTO_IPV6:
1847 printf("IPv6");
1848 break;
1849 case IPPROTO_TCP:
1850 printf("TCP");
1851 break;
1852 case IPPROTO_UDP:
1853 printf("UDP");
1854 break;
1855 case IPPROTO_ICMP:
1856 printf("ICMP");
1857 break;
1858 case IPPROTO_ICMPV6:
1859 printf("ICMPv6");
1860 break;
1861 default:
1862 printf("%u", skw->proto);
1863 break;
1864 }
1865 switch (dir) {
1866 case PF_IN:
1867 printf(" in");
1868 break;
1869 case PF_OUT:
1870 printf(" out");
1871 break;
1872 }
1873 if (skw) {
1874 printf(" wire: ");
1875 pf_print_host(&skw->addr[0], skw->port[0], skw->af);
1876 printf(" ");
1877 pf_print_host(&skw->addr[1], skw->port[1], skw->af);
1878 }
1879 if (sks) {
1880 printf(" stack: ");
1881 if (sks != skw) {
1882 pf_print_host(&sks->addr[0], sks->port[0], sks->af);
1883 printf(" ");
1884 pf_print_host(&sks->addr[1], sks->port[1], sks->af);
1885 } else
1886 printf("-");
1887 }
1888 if (s) {
1889 if (proto == IPPROTO_TCP) {
1890 printf(" [lo=%u high=%u win=%u modulator=%u",
1891 s->src.seqlo, s->src.seqhi,
1892 s->src.max_win, s->src.seqdiff);
1893 if (s->src.wscale && s->dst.wscale)
1894 printf(" wscale=%u",
1895 s->src.wscale & PF_WSCALE_MASK);
1896 printf("]");
1897 printf(" [lo=%u high=%u win=%u modulator=%u",
1898 s->dst.seqlo, s->dst.seqhi,
1899 s->dst.max_win, s->dst.seqdiff);
1900 if (s->src.wscale && s->dst.wscale)
1901 printf(" wscale=%u",
1902 s->dst.wscale & PF_WSCALE_MASK);
1903 printf("]");
1904 }
1905 printf(" %u:%u", s->src.state, s->dst.state);
1906 }
1907 }
1908
1909 void
1910 pf_print_flags(u_int8_t f)
1911 {
1912 if (f)
1913 printf(" ");
1914 if (f & TH_FIN)
1915 printf("F");
1916 if (f & TH_SYN)
1917 printf("S");
1918 if (f & TH_RST)
1919 printf("R");
1920 if (f & TH_PUSH)
1921 printf("P");
1922 if (f & TH_ACK)
1923 printf("A");
1924 if (f & TH_URG)
1925 printf("U");
1926 if (f & TH_ECE)
1927 printf("E");
1928 if (f & TH_CWR)
1929 printf("W");
1930 }
1931
1932 #define PF_SET_SKIP_STEPS(i) \
1933 do { \
1934 while (head[i] != cur) { \
1935 head[i]->skip[i].ptr = cur; \
1936 head[i] = TAILQ_NEXT(head[i], entries); \
1937 } \
1938 } while (0)
1939
1940 void
1941 pf_calc_skip_steps(struct pf_rulequeue *rules)
1942 {
1943 struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT];
1944 int i;
1945
1946 cur = TAILQ_FIRST(rules);
1947 prev = cur;
1948 for (i = 0; i < PF_SKIP_COUNT; ++i)
1949 head[i] = cur;
1950 while (cur != NULL) {
1951
1952 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
1953 PF_SET_SKIP_STEPS(PF_SKIP_IFP);
1954 if (cur->direction != prev->direction)
1955 PF_SET_SKIP_STEPS(PF_SKIP_DIR);
1956 if (cur->af != prev->af)
1957 PF_SET_SKIP_STEPS(PF_SKIP_AF);
1958 if (cur->proto != prev->proto)
1959 PF_SET_SKIP_STEPS(PF_SKIP_PROTO);
1960 if (cur->src.neg != prev->src.neg ||
1961 pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr))
1962 PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR);
1963 if (cur->src.port[0] != prev->src.port[0] ||
1964 cur->src.port[1] != prev->src.port[1] ||
1965 cur->src.port_op != prev->src.port_op)
1966 PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
1967 if (cur->dst.neg != prev->dst.neg ||
1968 pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr))
1969 PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR);
1970 if (cur->dst.port[0] != prev->dst.port[0] ||
1971 cur->dst.port[1] != prev->dst.port[1] ||
1972 cur->dst.port_op != prev->dst.port_op)
1973 PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
1974
1975 prev = cur;
1976 cur = TAILQ_NEXT(cur, entries);
1977 }
1978 for (i = 0; i < PF_SKIP_COUNT; ++i)
1979 PF_SET_SKIP_STEPS(i);
1980 }
1981
1982 int
1983 pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2)
1984 {
1985 if (aw1->type != aw2->type)
1986 return (1);
1987 switch (aw1->type) {
1988 case PF_ADDR_ADDRMASK:
1989 case PF_ADDR_RANGE:
1990 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0))
1991 return (1);
1992 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0))
1993 return (1);
1994 return (0);
1995 case PF_ADDR_DYNIFTL:
1996 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt);
1997 case PF_ADDR_NOROUTE:
1998 case PF_ADDR_URPFFAILED:
1999 return (0);
2000 case PF_ADDR_TABLE:
2001 return (aw1->p.tbl != aw2->p.tbl);
2002 case PF_ADDR_RTLABEL:
2003 return (aw1->v.rtlabel != aw2->v.rtlabel);
2004 default:
2005 printf("invalid address type: %d\n", aw1->type);
2006 return (1);
2007 }
2008 }
2009
2010 /**
2011 * Checksum updates are a little complicated because the checksum in the TCP/UDP
2012 * header isn't always a full checksum. In some cases (i.e. output) it's a
2013 * pseudo-header checksum, which is a partial checksum over src/dst IP
2014 * addresses, protocol number and length.
2015 *
2016 * That means we have the following cases:
2017 * * Input or forwarding: we don't have TSO, the checksum fields are full
2018 * checksums, we need to update the checksum whenever we change anything.
2019 * * Output (i.e. the checksum is a pseudo-header checksum):
2020 * x The field being updated is src/dst address or affects the length of
2021 * the packet. We need to update the pseudo-header checksum (note that this
2022 * checksum is not ones' complement).
2023 * x Some other field is being modified (e.g. src/dst port numbers): We
2024 * don't have to update anything.
2025 **/
2026 u_int16_t
2027 pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp)
2028 {
2029 u_int32_t l;
2030
2031 if (udp && !cksum)
2032 return (0x0000);
2033 l = cksum + old - new;
2034 l = (l >> 16) + (l & 65535);
2035 l = l & 65535;
2036 if (udp && !l)
2037 return (0xFFFF);
2038 return (l);
2039 }
2040
2041 u_int16_t
2042 pf_proto_cksum_fixup(struct mbuf *m, u_int16_t cksum, u_int16_t old,
2043 u_int16_t new, u_int8_t udp)
2044 {
2045 if (m->m_pkthdr.csum_flags & (CSUM_DELAY_DATA | CSUM_DELAY_DATA_IPV6))
2046 return (cksum);
2047
2048 return (pf_cksum_fixup(cksum, old, new, udp));
2049 }
2050
2051 void
2052 pf_change_ap(struct mbuf *m, struct pf_addr *a, u_int16_t *p, u_int16_t *ic,
2053 u_int16_t *pc, struct pf_addr *an, u_int16_t pn, u_int8_t u,
2054 sa_family_t af)
2055 {
2056 struct pf_addr ao;
2057 u_int16_t po = *p;
2058
2059 PF_ACPY(&ao, a, af);
2060 PF_ACPY(a, an, af);
2061
2062 if (m->m_pkthdr.csum_flags & (CSUM_DELAY_DATA | CSUM_DELAY_DATA_IPV6))
2063 *pc = ~*pc;
2064
2065 *p = pn;
2066
2067 switch (af) {
2068 #ifdef INET
2069 case AF_INET:
2070 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
2071 ao.addr16[0], an->addr16[0], 0),
2072 ao.addr16[1], an->addr16[1], 0);
2073 *p = pn;
2074
2075 *pc = pf_cksum_fixup(pf_cksum_fixup(*pc,
2076 ao.addr16[0], an->addr16[0], u),
2077 ao.addr16[1], an->addr16[1], u);
2078
2079 *pc = pf_proto_cksum_fixup(m, *pc, po, pn, u);
2080 break;
2081 #endif /* INET */
2082 #ifdef INET6
2083 case AF_INET6:
2084 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2085 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2086 pf_cksum_fixup(pf_cksum_fixup(*pc,
2087 ao.addr16[0], an->addr16[0], u),
2088 ao.addr16[1], an->addr16[1], u),
2089 ao.addr16[2], an->addr16[2], u),
2090 ao.addr16[3], an->addr16[3], u),
2091 ao.addr16[4], an->addr16[4], u),
2092 ao.addr16[5], an->addr16[5], u),
2093 ao.addr16[6], an->addr16[6], u),
2094 ao.addr16[7], an->addr16[7], u);
2095
2096 *pc = pf_proto_cksum_fixup(m, *pc, po, pn, u);
2097 break;
2098 #endif /* INET6 */
2099 }
2100
2101 if (m->m_pkthdr.csum_flags & (CSUM_DELAY_DATA |
2102 CSUM_DELAY_DATA_IPV6)) {
2103 *pc = ~*pc;
2104 if (! *pc)
2105 *pc = 0xffff;
2106 }
2107 }
2108
2109 /* Changes a u_int32_t. Uses a void * so there are no align restrictions */
2110 void
2111 pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u)
2112 {
2113 u_int32_t ao;
2114
2115 memcpy(&ao, a, sizeof(ao));
2116 memcpy(a, &an, sizeof(u_int32_t));
2117 *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u),
2118 ao % 65536, an % 65536, u);
2119 }
2120
2121 void
2122 pf_change_proto_a(struct mbuf *m, void *a, u_int16_t *c, u_int32_t an, u_int8_t udp)
2123 {
2124 u_int32_t ao;
2125
2126 memcpy(&ao, a, sizeof(ao));
2127 memcpy(a, &an, sizeof(u_int32_t));
2128
2129 *c = pf_proto_cksum_fixup(m,
2130 pf_proto_cksum_fixup(m, *c, ao / 65536, an / 65536, udp),
2131 ao % 65536, an % 65536, udp);
2132 }
2133
2134 #ifdef INET6
2135 void
2136 pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u)
2137 {
2138 struct pf_addr ao;
2139
2140 PF_ACPY(&ao, a, AF_INET6);
2141 PF_ACPY(a, an, AF_INET6);
2142
2143 *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2144 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2145 pf_cksum_fixup(pf_cksum_fixup(*c,
2146 ao.addr16[0], an->addr16[0], u),
2147 ao.addr16[1], an->addr16[1], u),
2148 ao.addr16[2], an->addr16[2], u),
2149 ao.addr16[3], an->addr16[3], u),
2150 ao.addr16[4], an->addr16[4], u),
2151 ao.addr16[5], an->addr16[5], u),
2152 ao.addr16[6], an->addr16[6], u),
2153 ao.addr16[7], an->addr16[7], u);
2154 }
2155 #endif /* INET6 */
2156
2157 void
2158 pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa,
2159 struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c,
2160 u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af)
2161 {
2162 struct pf_addr oia, ooa;
2163
2164 PF_ACPY(&oia, ia, af);
2165 if (oa)
2166 PF_ACPY(&ooa, oa, af);
2167
2168 /* Change inner protocol port, fix inner protocol checksum. */
2169 if (ip != NULL) {
2170 u_int16_t oip = *ip;
2171 u_int32_t opc;
2172
2173 if (pc != NULL)
2174 opc = *pc;
2175 *ip = np;
2176 if (pc != NULL)
2177 *pc = pf_cksum_fixup(*pc, oip, *ip, u);
2178 *ic = pf_cksum_fixup(*ic, oip, *ip, 0);
2179 if (pc != NULL)
2180 *ic = pf_cksum_fixup(*ic, opc, *pc, 0);
2181 }
2182 /* Change inner ip address, fix inner ip and icmp checksums. */
2183 PF_ACPY(ia, na, af);
2184 switch (af) {
2185 #ifdef INET
2186 case AF_INET: {
2187 u_int32_t oh2c = *h2c;
2188
2189 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c,
2190 oia.addr16[0], ia->addr16[0], 0),
2191 oia.addr16[1], ia->addr16[1], 0);
2192 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
2193 oia.addr16[0], ia->addr16[0], 0),
2194 oia.addr16[1], ia->addr16[1], 0);
2195 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0);
2196 break;
2197 }
2198 #endif /* INET */
2199 #ifdef INET6
2200 case AF_INET6:
2201 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2202 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2203 pf_cksum_fixup(pf_cksum_fixup(*ic,
2204 oia.addr16[0], ia->addr16[0], u),
2205 oia.addr16[1], ia->addr16[1], u),
2206 oia.addr16[2], ia->addr16[2], u),
2207 oia.addr16[3], ia->addr16[3], u),
2208 oia.addr16[4], ia->addr16[4], u),
2209 oia.addr16[5], ia->addr16[5], u),
2210 oia.addr16[6], ia->addr16[6], u),
2211 oia.addr16[7], ia->addr16[7], u);
2212 break;
2213 #endif /* INET6 */
2214 }
2215 /* Outer ip address, fix outer ip or icmpv6 checksum, if necessary. */
2216 if (oa) {
2217 PF_ACPY(oa, na, af);
2218 switch (af) {
2219 #ifdef INET
2220 case AF_INET:
2221 *hc = pf_cksum_fixup(pf_cksum_fixup(*hc,
2222 ooa.addr16[0], oa->addr16[0], 0),
2223 ooa.addr16[1], oa->addr16[1], 0);
2224 break;
2225 #endif /* INET */
2226 #ifdef INET6
2227 case AF_INET6:
2228 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2229 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2230 pf_cksum_fixup(pf_cksum_fixup(*ic,
2231 ooa.addr16[0], oa->addr16[0], u),
2232 ooa.addr16[1], oa->addr16[1], u),
2233 ooa.addr16[2], oa->addr16[2], u),
2234 ooa.addr16[3], oa->addr16[3], u),
2235 ooa.addr16[4], oa->addr16[4], u),
2236 ooa.addr16[5], oa->addr16[5], u),
2237 ooa.addr16[6], oa->addr16[6], u),
2238 ooa.addr16[7], oa->addr16[7], u);
2239 break;
2240 #endif /* INET6 */
2241 }
2242 }
2243 }
2244
2245
2246 /*
2247 * Need to modulate the sequence numbers in the TCP SACK option
2248 * (credits to Krzysztof Pfaff for report and patch)
2249 */
2250 int
2251 pf_modulate_sack(struct mbuf *m, int off, struct pf_pdesc *pd,
2252 struct tcphdr *th, struct pf_state_peer *dst)
2253 {
2254 int hlen = (th->th_off << 2) - sizeof(*th), thoptlen = hlen;
2255 #ifdef __FreeBSD__
2256 u_int8_t opts[TCP_MAXOLEN], *opt = opts;
2257 #else
2258 u_int8_t opts[MAX_TCPOPTLEN], *opt = opts;
2259 #endif
2260 int copyback = 0, i, olen;
2261 struct sackblk sack;
2262
2263 #define TCPOLEN_SACKLEN (TCPOLEN_SACK + 2)
2264 if (hlen < TCPOLEN_SACKLEN ||
2265 !pf_pull_hdr(m, off + sizeof(*th), opts, hlen, NULL, NULL, pd->af))
2266 return 0;
2267
2268 while (hlen >= TCPOLEN_SACKLEN) {
2269 olen = opt[1];
2270 switch (*opt) {
2271 case TCPOPT_EOL: /* FALLTHROUGH */
2272 case TCPOPT_NOP:
2273 opt++;
2274 hlen--;
2275 break;
2276 case TCPOPT_SACK:
2277 if (olen > hlen)
2278 olen = hlen;
2279 if (olen >= TCPOLEN_SACKLEN) {
2280 for (i = 2; i + TCPOLEN_SACK <= olen;
2281 i += TCPOLEN_SACK) {
2282 memcpy(&sack, &opt[i], sizeof(sack));
2283 pf_change_proto_a(m, &sack.start, &th->th_sum,
2284 htonl(ntohl(sack.start) - dst->seqdiff), 0);
2285 pf_change_proto_a(m, &sack.end, &th->th_sum,
2286 htonl(ntohl(sack.end) - dst->seqdiff), 0);
2287 memcpy(&opt[i], &sack, sizeof(sack));
2288 }
2289 copyback = 1;
2290 }
2291 /* FALLTHROUGH */
2292 default:
2293 if (olen < 2)
2294 olen = 2;
2295 hlen -= olen;
2296 opt += olen;
2297 }
2298 }
2299
2300 if (copyback)
2301 #ifdef __FreeBSD__
2302 m_copyback(m, off + sizeof(*th), thoptlen, (caddr_t)opts);
2303 #else
2304 m_copyback(m, off + sizeof(*th), thoptlen, opts);
2305 #endif
2306 return (copyback);
2307 }
2308
2309 void
2310 #ifdef __FreeBSD__
2311 pf_send_tcp(struct mbuf *replyto, const struct pf_rule *r, sa_family_t af,
2312 #else
2313 pf_send_tcp(const struct pf_rule *r, sa_family_t af,
2314 #endif
2315 const struct pf_addr *saddr, const struct pf_addr *daddr,
2316 u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack,
2317 u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl, int tag,
2318 u_int16_t rtag, struct ether_header *eh, struct ifnet *ifp)
2319 {
2320 struct mbuf *m;
2321 int len, tlen;
2322 #ifdef INET
2323 struct ip *h;
2324 #endif /* INET */
2325 #ifdef INET6
2326 struct ip6_hdr *h6;
2327 #endif /* INET6 */
2328 struct tcphdr *th;
2329 char *opt;
2330 #ifdef __FreeBSD__
2331 struct pf_mtag *pf_mtag;
2332
2333 KASSERT(
2334 #ifdef INET
2335 af == AF_INET
2336 #else
2337 0
2338 #endif
2339 ||
2340 #ifdef INET6
2341 af == AF_INET6
2342 #else
2343 0
2344 #endif
2345 , ("Unsupported AF %d", af));
2346 len = 0;
2347 th = NULL;
2348 #ifdef INET
2349 h = NULL;
2350 #endif
2351 #ifdef INET6
2352 h6 = NULL;
2353 #endif
2354 #endif /* __FreeBSD__ */
2355
2356 /* maximum segment size tcp option */
2357 tlen = sizeof(struct tcphdr);
2358 if (mss)
2359 tlen += 4;
2360
2361 switch (af) {
2362 #ifdef INET
2363 case AF_INET:
2364 len = sizeof(struct ip) + tlen;
2365 break;
2366 #endif /* INET */
2367 #ifdef INET6
2368 case AF_INET6:
2369 len = sizeof(struct ip6_hdr) + tlen;
2370 break;
2371 #endif /* INET6 */
2372 }
2373
2374 /* create outgoing mbuf */
2375 m = m_gethdr(M_DONTWAIT, MT_HEADER);
2376 if (m == NULL)
2377 return;
2378 #ifdef __FreeBSD__
2379 #ifdef MAC
2380 mac_netinet_firewall_send(m);
2381 #endif
2382 if ((pf_mtag = pf_get_mtag(m)) == NULL) {
2383 m_freem(m);
2384 return;
2385 }
2386 #endif
2387 if (tag)
2388 #ifdef __FreeBSD__
2389 m->m_flags |= M_SKIP_FIREWALL;
2390 pf_mtag->tag = rtag;
2391 #else
2392 m->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
2393 m->m_pkthdr.pf.tag = rtag;
2394 #endif
2395
2396 if (r != NULL && r->rtableid >= 0)
2397 #ifdef __FreeBSD__
2398 {
2399 M_SETFIB(m, r->rtableid);
2400 pf_mtag->rtableid = r->rtableid;
2401 #else
2402 m->m_pkthdr.pf.rtableid = r->rtableid;
2403 #endif
2404 #ifdef __FreeBSD__
2405 }
2406 #endif
2407
2408 #ifdef ALTQ
2409 if (r != NULL && r->qid) {
2410 #ifdef __FreeBSD__
2411 pf_mtag->qid = r->qid;
2412
2413 /* add hints for ecn */
2414 pf_mtag->hdr = mtod(m, struct ip *);
2415 #else
2416 m->m_pkthdr.pf.qid = r->qid;
2417 /* add hints for ecn */
2418 m->m_pkthdr.pf.hdr = mtod(m, struct ip *);
2419 #endif
2420 }
2421 #endif /* ALTQ */
2422 m->m_data += max_linkhdr;
2423 m->m_pkthdr.len = m->m_len = len;
2424 m->m_pkthdr.rcvif = NULL;
2425 bzero(m->m_data, len);
2426 switch (af) {
2427 #ifdef INET
2428 case AF_INET:
2429 h = mtod(m, struct ip *);
2430
2431 /* IP header fields included in the TCP checksum */
2432 h->ip_p = IPPROTO_TCP;
2433 h->ip_len = htons(tlen);
2434 h->ip_src.s_addr = saddr->v4.s_addr;
2435 h->ip_dst.s_addr = daddr->v4.s_addr;
2436
2437 th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip));
2438 break;
2439 #endif /* INET */
2440 #ifdef INET6
2441 case AF_INET6:
2442 h6 = mtod(m, struct ip6_hdr *);
2443
2444 /* IP header fields included in the TCP checksum */
2445 h6->ip6_nxt = IPPROTO_TCP;
2446 h6->ip6_plen = htons(tlen);
2447 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr));
2448 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr));
2449
2450 th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr));
2451 break;
2452 #endif /* INET6 */
2453 }
2454
2455 /* TCP header */
2456 th->th_sport = sport;
2457 th->th_dport = dport;
2458 th->th_seq = htonl(seq);
2459 th->th_ack = htonl(ack);
2460 th->th_off = tlen >> 2;
2461 th->th_flags = flags;
2462 th->th_win = htons(win);
2463
2464 if (mss) {
2465 opt = (char *)(th + 1);
2466 opt[0] = TCPOPT_MAXSEG;
2467 opt[1] = 4;
2468 HTONS(mss);
2469 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2);
2470 }
2471
2472 switch (af) {
2473 #ifdef INET
2474 case AF_INET:
2475 /* TCP checksum */
2476 th->th_sum = in_cksum(m, len);
2477
2478 /* Finish the IP header */
2479 h->ip_v = 4;
2480 h->ip_hl = sizeof(*h) >> 2;
2481 h->ip_tos = IPTOS_LOWDELAY;
2482 #ifdef __FreeBSD__
2483 h->ip_off = V_path_mtu_discovery ? IP_DF : 0;
2484 h->ip_len = len;
2485 h->ip_ttl = ttl ? ttl : V_ip_defttl;
2486 #else
2487 h->ip_len = htons(len);
2488 h->ip_off = htons(ip_mtudisc ? IP_DF : 0);
2489 h->ip_ttl = ttl ? ttl : ip_defttl;
2490 #endif
2491 h->ip_sum = 0;
2492 if (eh == NULL) {
2493 #ifdef __FreeBSD__
2494 PF_UNLOCK();
2495 ip_output(m, (void *)NULL, (void *)NULL, 0,
2496 (void *)NULL, (void *)NULL);
2497 PF_LOCK();
2498 #else /* ! __FreeBSD__ */
2499 ip_output(m, (void *)NULL, (void *)NULL, 0,
2500 (void *)NULL, (void *)NULL);
2501 #endif
2502 } else {
2503 struct route ro;
2504 struct rtentry rt;
2505 struct ether_header *e = (void *)ro.ro_dst.sa_data;
2506
2507 if (ifp == NULL) {
2508 m_freem(m);
2509 return;
2510 }
2511 rt.rt_ifp = ifp;
2512 ro.ro_rt = &rt;
2513 ro.ro_dst.sa_len = sizeof(ro.ro_dst);
2514 ro.ro_dst.sa_family = pseudo_AF_HDRCMPLT;
2515 bcopy(eh->ether_dhost, e->ether_shost, ETHER_ADDR_LEN);
2516 bcopy(eh->ether_shost, e->ether_dhost, ETHER_ADDR_LEN);
2517 e->ether_type = eh->ether_type;
2518 #ifdef __FreeBSD__
2519 PF_UNLOCK();
2520 /* XXX_IMPORT: later */
2521 ip_output(m, (void *)NULL, &ro, 0,
2522 (void *)NULL, (void *)NULL);
2523 PF_LOCK();
2524 #else /* ! __FreeBSD__ */
2525 ip_output(m, (void *)NULL, &ro, IP_ROUTETOETHER,
2526 (void *)NULL, (void *)NULL);
2527 #endif
2528 }
2529 break;
2530 #endif /* INET */
2531 #ifdef INET6
2532 case AF_INET6:
2533 /* TCP checksum */
2534 th->th_sum = in6_cksum(m, IPPROTO_TCP,
2535 sizeof(struct ip6_hdr), tlen);
2536
2537 h6->ip6_vfc |= IPV6_VERSION;
2538 h6->ip6_hlim = IPV6_DEFHLIM;
2539
2540 #ifdef __FreeBSD__
2541 PF_UNLOCK();
2542 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
2543 PF_LOCK();
2544 #else
2545 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
2546 #endif
2547 break;
2548 #endif /* INET6 */
2549 }
2550 }
2551
2552 static void
2553 pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af,
2554 struct pf_rule *r)
2555 {
2556 struct mbuf *m0;
2557 #ifdef __FreeBSD__
2558 #ifdef INET
2559 struct ip *ip;
2560 #endif
2561 struct pf_mtag *pf_mtag;
2562 #endif
2563
2564 #ifdef __FreeBSD__
2565 m0 = m_copypacket(m, M_DONTWAIT);
2566 if (m0 == NULL)
2567 return;
2568 #else
2569 if ((m0 = m_copy(m, 0, M_COPYALL)) == NULL)
2570 return;
2571 #endif
2572
2573 #ifdef __FreeBSD__
2574 if ((pf_mtag = pf_get_mtag(m0)) == NULL)
2575 return;
2576 /* XXX: revisit */
2577 m0->m_flags |= M_SKIP_FIREWALL;
2578 #else
2579 m0->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
2580 #endif
2581
2582 if (r->rtableid >= 0)
2583 #ifdef __FreeBSD__
2584 {
2585 M_SETFIB(m0, r->rtableid);
2586 pf_mtag->rtableid = r->rtableid;
2587 #else
2588 m0->m_pkthdr.pf.rtableid = r->rtableid;
2589 #endif
2590 #ifdef __FreeBSD__
2591 }
2592 #endif
2593
2594 #ifdef ALTQ
2595 if (r->qid) {
2596 #ifdef __FreeBSD__
2597 pf_mtag->qid = r->qid;
2598 /* add hints for ecn */
2599 pf_mtag->hdr = mtod(m0, struct ip *);
2600 #else
2601 m0->m_pkthdr.pf.qid = r->qid;
2602 /* add hints for ecn */
2603 m0->m_pkthdr.pf.hdr = mtod(m0, struct ip *);
2604 #endif
2605 }
2606 #endif /* ALTQ */
2607
2608 switch (af) {
2609 #ifdef INET
2610 case AF_INET:
2611 #ifdef __FreeBSD__
2612 /* icmp_error() expects host byte ordering */
2613 ip = mtod(m0, struct ip *);
2614 NTOHS(ip->ip_len);
2615 NTOHS(ip->ip_off);
2616 PF_UNLOCK();
2617 icmp_error(m0, type, code, 0, 0);
2618 PF_LOCK();
2619 #else
2620 icmp_error(m0, type, code, 0, 0);
2621 #endif
2622 break;
2623 #endif /* INET */
2624 #ifdef INET6
2625 case AF_INET6:
2626 #ifdef __FreeBSD__
2627 PF_UNLOCK();
2628 #endif
2629 icmp6_error(m0, type, code, 0);
2630 #ifdef __FreeBSD__
2631 PF_LOCK();
2632 #endif
2633 break;
2634 #endif /* INET6 */
2635 }
2636 }
2637
2638 /*
2639 * Return 1 if the addresses a and b match (with mask m), otherwise return 0.
2640 * If n is 0, they match if they are equal. If n is != 0, they match if they
2641 * are different.
2642 */
2643 int
2644 pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m,
2645 struct pf_addr *b, sa_family_t af)
2646 {
2647 int match = 0;
2648
2649 switch (af) {
2650 #ifdef INET
2651 case AF_INET:
2652 if ((a->addr32[0] & m->addr32[0]) ==
2653 (b->addr32[0] & m->addr32[0]))
2654 match++;
2655 break;
2656 #endif /* INET */
2657 #ifdef INET6
2658 case AF_INET6:
2659 if (((a->addr32[0] & m->addr32[0]) ==
2660 (b->addr32[0] & m->addr32[0])) &&
2661 ((a->addr32[1] & m->addr32[1]) ==
2662 (b->addr32[1] & m->addr32[1])) &&
2663 ((a->addr32[2] & m->addr32[2]) ==
2664 (b->addr32[2] & m->addr32[2])) &&
2665 ((a->addr32[3] & m->addr32[3]) ==
2666 (b->addr32[3] & m->addr32[3])))
2667 match++;
2668 break;
2669 #endif /* INET6 */
2670 }
2671 if (match) {
2672 if (n)
2673 return (0);
2674 else
2675 return (1);
2676 } else {
2677 if (n)
2678 return (1);
2679 else
2680 return (0);
2681 }
2682 }
2683
2684 /*
2685 * Return 1 if b <= a <= e, otherwise return 0.
2686 */
2687 int
2688 pf_match_addr_range(struct pf_addr *b, struct pf_addr *e,
2689 struct pf_addr *a, sa_family_t af)
2690 {
2691 switch (af) {
2692 #ifdef INET
2693 case AF_INET:
2694 if ((ntohl(a->addr32[0]) < ntohl(b->addr32[0])) ||
2695 (ntohl(a->addr32[0]) > ntohl(e->addr32[0])))
2696 return (0);
2697 break;
2698 #endif /* INET */
2699 #ifdef INET6
2700 case AF_INET6: {
2701 int i;
2702
2703 /* check a >= b */
2704 for (i = 0; i < 4; ++i)
2705 if (ntohl(a->addr32[i]) > ntohl(b->addr32[i]))
2706 break;
2707 else if (ntohl(a->addr32[i]) < ntohl(b->addr32[i]))
2708 return (0);
2709 /* check a <= e */
2710 for (i = 0; i < 4; ++i)
2711 if (ntohl(a->addr32[i]) < ntohl(e->addr32[i]))
2712 break;
2713 else if (ntohl(a->addr32[i]) > ntohl(e->addr32[i]))
2714 return (0);
2715 break;
2716 }
2717 #endif /* INET6 */
2718 }
2719 return (1);
2720 }
2721
2722 int
2723 pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p)
2724 {
2725 switch (op) {
2726 case PF_OP_IRG:
2727 return ((p > a1) && (p < a2));
2728 case PF_OP_XRG:
2729 return ((p < a1) || (p > a2));
2730 case PF_OP_RRG:
2731 return ((p >= a1) && (p <= a2));
2732 case PF_OP_EQ:
2733 return (p == a1);
2734 case PF_OP_NE:
2735 return (p != a1);
2736 case PF_OP_LT:
2737 return (p < a1);
2738 case PF_OP_LE:
2739 return (p <= a1);
2740 case PF_OP_GT:
2741 return (p > a1);
2742 case PF_OP_GE:
2743 return (p >= a1);
2744 }
2745 return (0); /* never reached */
2746 }
2747
2748 int
2749 pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p)
2750 {
2751 NTOHS(a1);
2752 NTOHS(a2);
2753 NTOHS(p);
2754 return (pf_match(op, a1, a2, p));
2755 }
2756
2757 int
2758 pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u)
2759 {
2760 if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
2761 return (0);
2762 return (pf_match(op, a1, a2, u));
2763 }
2764
2765 int
2766 pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g)
2767 {
2768 if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
2769 return (0);
2770 return (pf_match(op, a1, a2, g));
2771 }
2772
2773 int
2774 #ifdef __FreeBSD__
2775 pf_match_tag(struct mbuf *m, struct pf_rule *r, int *tag,
2776 struct pf_mtag *pf_mtag)
2777 #else
2778 pf_match_tag(struct mbuf *m, struct pf_rule *r, int *tag)
2779 #endif
2780 {
2781 if (*tag == -1)
2782 #ifdef __FreeBSD__
2783 *tag = pf_mtag->tag;
2784 #else
2785 *tag = m->m_pkthdr.pf.tag;
2786 #endif
2787
2788 return ((!r->match_tag_not && r->match_tag == *tag) ||
2789 (r->match_tag_not && r->match_tag != *tag));
2790 }
2791
2792 int
2793 #ifdef __FreeBSD__
2794 pf_tag_packet(struct mbuf *m, int tag, int rtableid,
2795 struct pf_mtag *pf_mtag)
2796 #else
2797 pf_tag_packet(struct mbuf *m, int tag, int rtableid)
2798 #endif
2799 {
2800 if (tag <= 0 && rtableid < 0)
2801 return (0);
2802
2803 if (tag > 0)
2804 #ifdef __FreeBSD__
2805 pf_mtag->tag = tag;
2806 #else
2807 m->m_pkthdr.pf.tag = tag;
2808 #endif
2809 if (rtableid >= 0)
2810 #ifdef __FreeBSD__
2811 {
2812 M_SETFIB(m, rtableid);
2813 }
2814 #else
2815 m->m_pkthdr.pf.rtableid = rtableid;
2816 #endif
2817
2818 return (0);
2819 }
2820
2821 void
2822 pf_step_into_anchor(int *depth, struct pf_ruleset **rs, int n,
2823 struct pf_rule **r, struct pf_rule **a, int *match)
2824 {
2825 struct pf_anchor_stackframe *f;
2826
2827 (*r)->anchor->match = 0;
2828 if (match)
2829 *match = 0;
2830 #ifdef __FreeBSD__
2831 if (*depth >= sizeof(V_pf_anchor_stack) /
2832 sizeof(V_pf_anchor_stack[0])) {
2833 #else
2834 if (*depth >= sizeof(pf_anchor_stack) /
2835 sizeof(pf_anchor_stack[0])) {
2836 #endif
2837 printf("pf_step_into_anchor: stack overflow\n");
2838 *r = TAILQ_NEXT(*r, entries);
2839 return;
2840 } else if (*depth == 0 && a != NULL)
2841 *a = *r;
2842 #ifdef __FreeBSD__
2843 f = V_pf_anchor_stack + (*depth)++;
2844 #else
2845 f = pf_anchor_stack + (*depth)++;
2846 #endif
2847 f->rs = *rs;
2848 f->r = *r;
2849 if ((*r)->anchor_wildcard) {
2850 f->parent = &(*r)->anchor->children;
2851 if ((f->child = RB_MIN(pf_anchor_node, f->parent)) ==
2852 NULL) {
2853 *r = NULL;
2854 return;
2855 }
2856 *rs = &f->child->ruleset;
2857 } else {
2858 f->parent = NULL;
2859 f->child = NULL;
2860 *rs = &(*r)->anchor->ruleset;
2861 }
2862 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
2863 }
2864
2865 int
2866 pf_step_out_of_anchor(int *depth, struct pf_ruleset **rs, int n,
2867 struct pf_rule **r, struct pf_rule **a, int *match)
2868 {
2869 struct pf_anchor_stackframe *f;
2870 int quick = 0;
2871
2872 do {
2873 if (*depth <= 0)
2874 break;
2875 #ifdef __FreeBSD__
2876 f = V_pf_anchor_stack + *depth - 1;
2877 #else
2878 f = pf_anchor_stack + *depth - 1;
2879 #endif
2880 if (f->parent != NULL && f->child != NULL) {
2881 if (f->child->match ||
2882 (match != NULL && *match)) {
2883 f->r->anchor->match = 1;
2884 *match = 0;
2885 }
2886 f->child = RB_NEXT(pf_anchor_node, f->parent, f->child);
2887 if (f->child != NULL) {
2888 *rs = &f->child->ruleset;
2889 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
2890 if (*r == NULL)
2891 continue;
2892 else
2893 break;
2894 }
2895 }
2896 (*depth)--;
2897 if (*depth == 0 && a != NULL)
2898 *a = NULL;
2899 *rs = f->rs;
2900 if (f->r->anchor->match || (match != NULL && *match))
2901 quick = f->r->quick;
2902 *r = TAILQ_NEXT(f->r, entries);
2903 } while (*r == NULL);
2904
2905 return (quick);
2906 }
2907
2908 #ifdef INET6
2909 void
2910 pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr,
2911 struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af)
2912 {
2913 switch (af) {
2914 #ifdef INET
2915 case AF_INET:
2916 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2917 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2918 break;
2919 #endif /* INET */
2920 case AF_INET6:
2921 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2922 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2923 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) |
2924 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]);
2925 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) |
2926 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]);
2927 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) |
2928 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]);
2929 break;
2930 }
2931 }
2932
2933 void
2934 pf_addr_inc(struct pf_addr *addr, sa_family_t af)
2935 {
2936 switch (af) {
2937 #ifdef INET
2938 case AF_INET:
2939 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1);
2940 break;
2941 #endif /* INET */
2942 case AF_INET6:
2943 if (addr->addr32[3] == 0xffffffff) {
2944 addr->addr32[3] = 0;
2945 if (addr->addr32[2] == 0xffffffff) {
2946 addr->addr32[2] = 0;
2947 if (addr->addr32[1] == 0xffffffff) {
2948 addr->addr32[1] = 0;
2949 addr->addr32[0] =
2950 htonl(ntohl(addr->addr32[0]) + 1);
2951 } else
2952 addr->addr32[1] =
2953 htonl(ntohl(addr->addr32[1]) + 1);
2954 } else
2955 addr->addr32[2] =
2956 htonl(ntohl(addr->addr32[2]) + 1);
2957 } else
2958 addr->addr32[3] =
2959 htonl(ntohl(addr->addr32[3]) + 1);
2960 break;
2961 }
2962 }
2963 #endif /* INET6 */
2964
2965 int
2966 #ifdef __FreeBSD__
2967 pf_socket_lookup(int direction, struct pf_pdesc *pd, struct inpcb *inp_arg)
2968 #else
2969 pf_socket_lookup(int direction, struct pf_pdesc *pd)
2970 #endif
2971 {
2972 struct pf_addr *saddr, *daddr;
2973 u_int16_t sport, dport;
2974 #ifdef __FreeBSD__
2975 struct inpcbinfo *pi;
2976 #else
2977 struct inpcbtable *tb;
2978 #endif
2979 struct inpcb *inp;
2980
2981 if (pd == NULL)
2982 return (-1);
2983 pd->lookup.uid = UID_MAX;
2984 pd->lookup.gid = GID_MAX;
2985 pd->lookup.pid = NO_PID;
2986
2987 #ifdef __FreeBSD__
2988 if (inp_arg != NULL) {
2989 INP_LOCK_ASSERT(inp_arg);
2990 pd->lookup.uid = inp_arg->inp_cred->cr_uid;
2991 pd->lookup.gid = inp_arg->inp_cred->cr_groups[0];
2992 return (1);
2993 }
2994 #endif
2995
2996 switch (pd->proto) {
2997 case IPPROTO_TCP:
2998 if (pd->hdr.tcp == NULL)
2999 return (-1);
3000 sport = pd->hdr.tcp->th_sport;
3001 dport = pd->hdr.tcp->th_dport;
3002 #ifdef __FreeBSD__
3003 pi = &V_tcbinfo;
3004 #else
3005 tb = &tcbtable;
3006 #endif
3007 break;
3008 case IPPROTO_UDP:
3009 if (pd->hdr.udp == NULL)
3010 return (-1);
3011 sport = pd->hdr.udp->uh_sport;
3012 dport = pd->hdr.udp->uh_dport;
3013 #ifdef __FreeBSD__
3014 pi = &V_udbinfo;
3015 #else
3016 tb = &udbtable;
3017 #endif
3018 break;
3019 default:
3020 return (-1);
3021 }
3022 if (direction == PF_IN) {
3023 saddr = pd->src;
3024 daddr = pd->dst;
3025 } else {
3026 u_int16_t p;
3027
3028 p = sport;
3029 sport = dport;
3030 dport = p;
3031 saddr = pd->dst;
3032 daddr = pd->src;
3033 }
3034 switch (pd->af) {
3035 #ifdef INET
3036 case AF_INET:
3037 #ifdef __FreeBSD__
3038 /*
3039 * XXXRW: would be nice if we had an mbuf here so that we
3040 * could use in_pcblookup_mbuf().
3041 */
3042 inp = in_pcblookup(pi, saddr->v4, sport, daddr->v4,
3043 dport, INPLOOKUP_RLOCKPCB, NULL);
3044 if (inp == NULL) {
3045 inp = in_pcblookup(pi, saddr->v4, sport,
3046 daddr->v4, dport, INPLOOKUP_WILDCARD |
3047 INPLOOKUP_RLOCKPCB, NULL);
3048 if (inp == NULL)
3049 return (-1);
3050 }
3051 #else
3052 inp = in_pcbhashlookup(tb, saddr->v4, sport, daddr->v4, dport);
3053 if (inp == NULL) {
3054 inp = in_pcblookup_listen(tb, daddr->v4, dport, 0,
3055 NULL);
3056 if (inp == NULL)
3057 return (-1);
3058 }
3059 #endif
3060 break;
3061 #endif /* INET */
3062 #ifdef INET6
3063 case AF_INET6:
3064 #ifdef __FreeBSD__
3065 /*
3066 * XXXRW: would be nice if we had an mbuf here so that we
3067 * could use in6_pcblookup_mbuf().
3068 */
3069 inp = in6_pcblookup(pi, &saddr->v6, sport,
3070 &daddr->v6, dport, INPLOOKUP_RLOCKPCB, NULL);
3071 if (inp == NULL) {
3072 inp = in6_pcblookup(pi, &saddr->v6, sport,
3073 &daddr->v6, dport, INPLOOKUP_WILDCARD |
3074 INPLOOKUP_RLOCKPCB, NULL);
3075 if (inp == NULL)
3076 return (-1);
3077 }
3078 #else
3079 inp = in6_pcbhashlookup(tb, &saddr->v6, sport, &daddr->v6,
3080 dport);
3081 if (inp == NULL) {
3082 inp = in6_pcblookup_listen(tb, &daddr->v6, dport, 0,
3083 NULL);
3084 if (inp == NULL)
3085 return (-1);
3086 }
3087 #endif
3088 break;
3089 #endif /* INET6 */
3090
3091 default:
3092 return (-1);
3093 }
3094 #ifdef __FreeBSD__
3095 INP_RLOCK_ASSERT(inp);
3096 pd->lookup.uid = inp->inp_cred->cr_uid;
3097 pd->lookup.gid = inp->inp_cred->cr_groups[0];
3098 INP_RUNLOCK(inp);
3099 #else
3100 pd->lookup.uid = inp->inp_socket->so_euid;
3101 pd->lookup.gid = inp->inp_socket->so_egid;
3102 pd->lookup.pid = inp->inp_socket->so_cpid;
3103 #endif
3104 return (1);
3105 }
3106
3107 u_int8_t
3108 pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
3109 {
3110 int hlen;
3111 u_int8_t hdr[60];
3112 u_int8_t *opt, optlen;
3113 u_int8_t wscale = 0;
3114
3115 hlen = th_off << 2; /* hlen <= sizeof(hdr) */
3116 if (hlen <= sizeof(struct tcphdr))
3117 return (0);
3118 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
3119 return (0);
3120 opt = hdr + sizeof(struct tcphdr);
3121 hlen -= sizeof(struct tcphdr);
3122 while (hlen >= 3) {
3123 switch (*opt) {
3124 case TCPOPT_EOL:
3125 case TCPOPT_NOP:
3126 ++opt;
3127 --hlen;
3128 break;
3129 case TCPOPT_WINDOW:
3130 wscale = opt[2];
3131 if (wscale > TCP_MAX_WINSHIFT)
3132 wscale = TCP_MAX_WINSHIFT;
3133 wscale |= PF_WSCALE_FLAG;
3134 /* FALLTHROUGH */
3135 default:
3136 optlen = opt[1];
3137 if (optlen < 2)
3138 optlen = 2;
3139 hlen -= optlen;
3140 opt += optlen;
3141 break;
3142 }
3143 }
3144 return (wscale);
3145 }
3146
3147 u_int16_t
3148 pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
3149 {
3150 int hlen;
3151 u_int8_t hdr[60];
3152 u_int8_t *opt, optlen;
3153 #ifdef __FreeBSD__
3154 u_int16_t mss = V_tcp_mssdflt;
3155 #else
3156 u_int16_t mss = tcp_mssdflt;
3157 #endif
3158
3159 hlen = th_off << 2; /* hlen <= sizeof(hdr) */
3160 if (hlen <= sizeof(struct tcphdr))
3161 return (0);
3162 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
3163 return (0);
3164 opt = hdr + sizeof(struct tcphdr);
3165 hlen -= sizeof(struct tcphdr);
3166 while (hlen >= TCPOLEN_MAXSEG) {
3167 switch (*opt) {
3168 case TCPOPT_EOL:
3169 case TCPOPT_NOP:
3170 ++opt;
3171 --hlen;
3172 break;
3173 case TCPOPT_MAXSEG:
3174 bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2);
3175 NTOHS(mss);
3176 /* FALLTHROUGH */
3177 default:
3178 optlen = opt[1];
3179 if (optlen < 2)
3180 optlen = 2;
3181 hlen -= optlen;
3182 opt += optlen;
3183 break;
3184 }
3185 }
3186 return (mss);
3187 }
3188
3189 u_int16_t
3190 pf_calc_mss(struct pf_addr *addr, sa_family_t af, int rtableid, u_int16_t offer)
3191 {
3192 #ifdef INET
3193 struct sockaddr_in *dst;
3194 struct route ro;
3195 #endif /* INET */
3196 #ifdef INET6
3197 struct sockaddr_in6 *dst6;
3198 struct route_in6 ro6;
3199 #endif /* INET6 */
3200 struct rtentry *rt = NULL;
3201 #ifdef __FreeBSD__
3202 int hlen = 0;
3203 u_int16_t mss = V_tcp_mssdflt;
3204 #else
3205 int hlen;
3206 u_int16_t mss = tcp_mssdflt;
3207 #endif
3208
3209 switch (af) {
3210 #ifdef INET
3211 case AF_INET:
3212 hlen = sizeof(struct ip);
3213 bzero(&ro, sizeof(ro));
3214 dst = (struct sockaddr_in *)&ro.ro_dst;
3215 dst->sin_family = AF_INET;
3216 dst->sin_len = sizeof(*dst);
3217 dst->sin_addr = addr->v4;
3218 #ifdef __FreeBSD__
3219 in_rtalloc_ign(&ro, 0, rtableid);
3220 #else /* ! __FreeBSD__ */
3221 rtalloc_noclone(&ro, NO_CLONING);
3222 #endif
3223 rt = ro.ro_rt;
3224 break;
3225 #endif /* INET */
3226 #ifdef INET6
3227 case AF_INET6:
3228 hlen = sizeof(struct ip6_hdr);
3229 bzero(&ro6, sizeof(ro6));
3230 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst;
3231 dst6->sin6_family = AF_INET6;
3232 dst6->sin6_len = sizeof(*dst6);
3233 dst6->sin6_addr = addr->v6;
3234 #ifdef __FreeBSD__
3235 in6_rtalloc_ign(&ro6, 0, rtableid);
3236 #else /* ! __FreeBSD__ */
3237 rtalloc_noclone((struct route *)&ro6, NO_CLONING);
3238 #endif
3239 rt = ro6.ro_rt;
3240 break;
3241 #endif /* INET6 */
3242 }
3243
3244 if (rt && rt->rt_ifp) {
3245 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr);
3246 #ifdef __FreeBSD__
3247 mss = max(V_tcp_mssdflt, mss);
3248 #else
3249 mss = max(tcp_mssdflt, mss);
3250 #endif
3251 RTFREE(rt);
3252 }
3253 mss = min(mss, offer);
3254 mss = max(mss, 64); /* sanity - at least max opt space */
3255 return (mss);
3256 }
3257
3258 void
3259 pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr)
3260 {
3261 struct pf_rule *r = s->rule.ptr;
3262 struct pf_src_node *sn = NULL;
3263
3264 s->rt_kif = NULL;
3265 if (!r->rt || r->rt == PF_FASTROUTE)
3266 return;
3267 switch (s->key[PF_SK_WIRE]->af) {
3268 #ifdef INET
3269 case AF_INET:
3270 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL, &sn);
3271 s->rt_kif = r->rpool.cur->kif;
3272 break;
3273 #endif /* INET */
3274 #ifdef INET6
3275 case AF_INET6:
3276 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL, &sn);
3277 s->rt_kif = r->rpool.cur->kif;
3278 break;
3279 #endif /* INET6 */
3280 }
3281 }
3282
3283 u_int32_t
3284 pf_tcp_iss(struct pf_pdesc *pd)
3285 {
3286 MD5_CTX ctx;
3287 u_int32_t digest[4];
3288
3289 #ifdef __FreeBSD__
3290 if (V_pf_tcp_secret_init == 0) {
3291 read_random(&V_pf_tcp_secret, sizeof(V_pf_tcp_secret));
3292 MD5Init(&V_pf_tcp_secret_ctx);
3293 MD5Update(&V_pf_tcp_secret_ctx, V_pf_tcp_secret,
3294 sizeof(V_pf_tcp_secret));
3295 V_pf_tcp_secret_init = 1;
3296 }
3297
3298 ctx = V_pf_tcp_secret_ctx;
3299 #else
3300 if (pf_tcp_secret_init == 0) {
3301 arc4random_buf(pf_tcp_secret, sizeof(pf_tcp_secret));
3302 MD5Init(&pf_tcp_secret_ctx);
3303 MD5Update(&pf_tcp_secret_ctx, pf_tcp_secret,
3304 sizeof(pf_tcp_secret));
3305 pf_tcp_secret_init = 1;
3306 }
3307
3308 ctx = pf_tcp_secret_ctx;
3309 #endif
3310
3311 MD5Update(&ctx, (char *)&pd->hdr.tcp->th_sport, sizeof(u_short));
3312 MD5Update(&ctx, (char *)&pd->hdr.tcp->th_dport, sizeof(u_short));
3313 if (pd->af == AF_INET6) {
3314 MD5Update(&ctx, (char *)&pd->src->v6, sizeof(struct in6_addr));
3315 MD5Update(&ctx, (char *)&pd->dst->v6, sizeof(struct in6_addr));
3316 } else {
3317 MD5Update(&ctx, (char *)&pd->src->v4, sizeof(struct in_addr));
3318 MD5Update(&ctx, (char *)&pd->dst->v4, sizeof(struct in_addr));
3319 }
3320 MD5Final((u_char *)digest, &ctx);
3321 #ifdef __FreeBSD__
3322 V_pf_tcp_iss_off += 4096;
3323 #define ISN_RANDOM_INCREMENT (4096 - 1)
3324 return (digest[0] + (arc4random() & ISN_RANDOM_INCREMENT) +
3325 V_pf_tcp_iss_off);
3326 #undef ISN_RANDOM_INCREMENT
3327 #else
3328 pf_tcp_iss_off += 4096;
3329 return (digest[0] + tcp_iss + pf_tcp_iss_off);
3330 #endif
3331 }
3332
3333 int
3334 pf_test_rule(struct pf_rule **rm, struct pf_state **sm, int direction,
3335 struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3336 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
3337 #ifdef __FreeBSD__
3338 struct ifqueue *ifq, struct inpcb *inp)
3339 #else
3340 struct ifqueue *ifq)
3341 #endif
3342 {
3343 struct pf_rule *nr = NULL;
3344 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
3345 sa_family_t af = pd->af;
3346 struct pf_rule *r, *a = NULL;
3347 struct pf_ruleset *ruleset = NULL;
3348 struct pf_src_node *nsn = NULL;
3349 struct tcphdr *th = pd->hdr.tcp;
3350 struct pf_state_key *skw = NULL, *sks = NULL;
3351 struct pf_state_key *sk = NULL, *nk = NULL;
3352 u_short reason;
3353 int rewrite = 0, hdrlen = 0;
3354 int tag = -1, rtableid = -1;
3355 int asd = 0;
3356 int match = 0;
3357 int state_icmp = 0;
3358 #ifdef __FreeBSD__
3359 u_int16_t sport = 0, dport = 0;
3360 u_int16_t bproto_sum = 0, bip_sum = 0;
3361 #else
3362 u_int16_t sport, dport;
3363 u_int16_t bproto_sum = 0, bip_sum;
3364 #endif
3365 u_int8_t icmptype = 0, icmpcode = 0;
3366
3367
3368 if (direction == PF_IN && pf_check_congestion(ifq)) {
3369 REASON_SET(&reason, PFRES_CONGEST);
3370 return (PF_DROP);
3371 }
3372
3373 #ifdef __FreeBSD__
3374 if (inp != NULL)
3375 pd->lookup.done = pf_socket_lookup(direction, pd, inp);
3376 else if (V_debug_pfugidhack) {
3377 PF_UNLOCK();
3378 DPFPRINTF(PF_DEBUG_MISC, ("pf: unlocked lookup\n"));
3379 pd->lookup.done = pf_socket_lookup(direction, pd, inp);
3380 PF_LOCK();
3381 }
3382 #endif
3383
3384 switch (pd->proto) {
3385 case IPPROTO_TCP:
3386 sport = th->th_sport;
3387 dport = th->th_dport;
3388 hdrlen = sizeof(*th);
3389 break;
3390 case IPPROTO_UDP:
3391 sport = pd->hdr.udp->uh_sport;
3392 dport = pd->hdr.udp->uh_dport;
3393 hdrlen = sizeof(*pd->hdr.udp);
3394 break;
3395 #ifdef INET
3396 case IPPROTO_ICMP:
3397 if (pd->af != AF_INET)
3398 break;
3399 sport = dport = pd->hdr.icmp->icmp_id;
3400 hdrlen = sizeof(*pd->hdr.icmp);
3401 icmptype = pd->hdr.icmp->icmp_type;
3402 icmpcode = pd->hdr.icmp->icmp_code;
3403
3404 if (icmptype == ICMP_UNREACH ||
3405 icmptype == ICMP_SOURCEQUENCH ||
3406 icmptype == ICMP_REDIRECT ||
3407 icmptype == ICMP_TIMXCEED ||
3408 icmptype == ICMP_PARAMPROB)
3409 state_icmp++;
3410 break;
3411 #endif /* INET */
3412 #ifdef INET6
3413 case IPPROTO_ICMPV6:
3414 if (af != AF_INET6)
3415 break;
3416 sport = dport = pd->hdr.icmp6->icmp6_id;
3417 hdrlen = sizeof(*pd->hdr.icmp6);
3418 icmptype = pd->hdr.icmp6->icmp6_type;
3419 icmpcode = pd->hdr.icmp6->icmp6_code;
3420
3421 if (icmptype == ICMP6_DST_UNREACH ||
3422 icmptype == ICMP6_PACKET_TOO_BIG ||
3423 icmptype == ICMP6_TIME_EXCEEDED ||
3424 icmptype == ICMP6_PARAM_PROB)
3425 state_icmp++;
3426 break;
3427 #endif /* INET6 */
3428 default:
3429 sport = dport = hdrlen = 0;
3430 break;
3431 }
3432
3433 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3434
3435 /* check packet for BINAT/NAT/RDR */
3436 if ((nr = pf_get_translation(pd, m, off, direction, kif, &nsn,
3437 &skw, &sks, &sk, &nk, saddr, daddr, sport, dport)) != NULL) {
3438 if (nk == NULL || sk == NULL) {
3439 REASON_SET(&reason, PFRES_MEMORY);
3440 goto cleanup;
3441 }
3442
3443 if (pd->ip_sum)
3444 bip_sum = *pd->ip_sum;
3445
3446 switch (pd->proto) {
3447 case IPPROTO_TCP:
3448 bproto_sum = th->th_sum;
3449 pd->proto_sum = &th->th_sum;
3450
3451 if (PF_ANEQ(saddr, &nk->addr[pd->sidx], af) ||
3452 nk->port[pd->sidx] != sport) {
3453 pf_change_ap(m, saddr, &th->th_sport, pd->ip_sum,
3454 &th->th_sum, &nk->addr[pd->sidx],
3455 nk->port[pd->sidx], 0, af);
3456 pd->sport = &th->th_sport;
3457 sport = th->th_sport;
3458 }
3459
3460 if (PF_ANEQ(daddr, &nk->addr[pd->didx], af) ||
3461 nk->port[pd->didx] != dport) {
3462 pf_change_ap(m, daddr, &th->th_dport, pd->ip_sum,
3463 &th->th_sum, &nk->addr[pd->didx],
3464 nk->port[pd->didx], 0, af);
3465 dport = th->th_dport;
3466 pd->dport = &th->th_dport;
3467 }
3468 rewrite++;
3469 break;
3470 case IPPROTO_UDP:
3471 bproto_sum = pd->hdr.udp->uh_sum;
3472 pd->proto_sum = &pd->hdr.udp->uh_sum;
3473
3474 if (PF_ANEQ(saddr, &nk->addr[pd->sidx], af) ||
3475 nk->port[pd->sidx] != sport) {
3476 pf_change_ap(m, saddr, &pd->hdr.udp->uh_sport,
3477 pd->ip_sum, &pd->hdr.udp->uh_sum,
3478 &nk->addr[pd->sidx],
3479 nk->port[pd->sidx], 1, af);
3480 sport = pd->hdr.udp->uh_sport;
3481 pd->sport = &pd->hdr.udp->uh_sport;
3482 }
3483
3484 if (PF_ANEQ(daddr, &nk->addr[pd->didx], af) ||
3485 nk->port[pd->didx] != dport) {
3486 pf_change_ap(m, daddr, &pd->hdr.udp->uh_dport,
3487 pd->ip_sum, &pd->hdr.udp->uh_sum,
3488 &nk->addr[pd->didx],
3489 nk->port[pd->didx], 1, af);
3490 dport = pd->hdr.udp->uh_dport;
3491 pd->dport = &pd->hdr.udp->uh_dport;
3492 }
3493 rewrite++;
3494 break;
3495 #ifdef INET
3496 case IPPROTO_ICMP:
3497 nk->port[0] = nk->port[1];
3498 if (PF_ANEQ(saddr, &nk->addr[pd->sidx], AF_INET))
3499 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3500 nk->addr[pd->sidx].v4.s_addr, 0);
3501
3502 if (PF_ANEQ(daddr, &nk->addr[pd->didx], AF_INET))
3503 pf_change_a(&daddr->v4.s_addr, pd->ip_sum,
3504 nk->addr[pd->didx].v4.s_addr, 0);
3505
3506 if (nk->port[1] != pd->hdr.icmp->icmp_id) {
3507 pd->hdr.icmp->icmp_cksum = pf_cksum_fixup(
3508 pd->hdr.icmp->icmp_cksum, sport,
3509 nk->port[1], 0);
3510 pd->hdr.icmp->icmp_id = nk->port[1];
3511 pd->sport = &pd->hdr.icmp->icmp_id;
3512 }
3513 m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
3514 break;
3515 #endif /* INET */
3516 #ifdef INET6
3517 case IPPROTO_ICMPV6:
3518 nk->port[0] = nk->port[1];
3519 if (PF_ANEQ(saddr, &nk->addr[pd->sidx], AF_INET6))
3520 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum,
3521 &nk->addr[pd->sidx], 0);
3522
3523 if (PF_ANEQ(daddr, &nk->addr[pd->didx], AF_INET6))
3524 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum,
3525 &nk->addr[pd->didx], 0);
3526 rewrite++;
3527 break;
3528 #endif /* INET */
3529 default:
3530 switch (af) {
3531 #ifdef INET
3532 case AF_INET:
3533 if (PF_ANEQ(saddr,
3534 &nk->addr[pd->sidx], AF_INET))
3535 pf_change_a(&saddr->v4.s_addr,
3536 pd->ip_sum,
3537 nk->addr[pd->sidx].v4.s_addr, 0);
3538
3539 if (PF_ANEQ(daddr,
3540 &nk->addr[pd->didx], AF_INET))
3541 pf_change_a(&daddr->v4.s_addr,
3542 pd->ip_sum,
3543 nk->addr[pd->didx].v4.s_addr, 0);
3544 break;
3545 #endif /* INET */
3546 #ifdef INET6
3547 case AF_INET6:
3548 if (PF_ANEQ(saddr,
3549 &nk->addr[pd->sidx], AF_INET6))
3550 PF_ACPY(saddr, &nk->addr[pd->sidx], af);
3551
3552 if (PF_ANEQ(daddr,
3553 &nk->addr[pd->didx], AF_INET6))
3554 PF_ACPY(saddr, &nk->addr[pd->didx], af);
3555 break;
3556 #endif /* INET */
3557 }
3558 break;
3559 }
3560 if (nr->natpass)
3561 r = NULL;
3562 pd->nat_rule = nr;
3563 }
3564
3565 while (r != NULL) {
3566 r->evaluations++;
3567 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3568 r = r->skip[PF_SKIP_IFP].ptr;
3569 else if (r->direction && r->direction != direction)
3570 r = r->skip[PF_SKIP_DIR].ptr;
3571 else if (r->af && r->af != af)
3572 r = r->skip[PF_SKIP_AF].ptr;
3573 else if (r->proto && r->proto != pd->proto)
3574 r = r->skip[PF_SKIP_PROTO].ptr;
3575 else if (PF_MISMATCHAW(&r->src.addr, saddr, af,
3576 r->src.neg, kif, M_GETFIB(m)))
3577 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3578 /* tcp/udp only. port_op always 0 in other cases */
3579 else if (r->src.port_op && !pf_match_port(r->src.port_op,
3580 r->src.port[0], r->src.port[1], sport))
3581 r = r->skip[PF_SKIP_SRC_PORT].ptr;
3582 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af,
3583 r->dst.neg, NULL, M_GETFIB(m)))
3584 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3585 /* tcp/udp only. port_op always 0 in other cases */
3586 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
3587 r->dst.port[0], r->dst.port[1], dport))
3588 r = r->skip[PF_SKIP_DST_PORT].ptr;
3589 /* icmp only. type always 0 in other cases */
3590 else if (r->type && r->type != icmptype + 1)
3591 r = TAILQ_NEXT(r, entries);
3592 /* icmp only. type always 0 in other cases */
3593 else if (r->code && r->code != icmpcode + 1)
3594 r = TAILQ_NEXT(r, entries);
3595 else if (r->tos && !(r->tos == pd->tos))
3596 r = TAILQ_NEXT(r, entries);
3597 else if (r->rule_flag & PFRULE_FRAGMENT)
3598 r = TAILQ_NEXT(r, entries);
3599 else if (pd->proto == IPPROTO_TCP &&
3600 (r->flagset & th->th_flags) != r->flags)
3601 r = TAILQ_NEXT(r, entries);
3602 /* tcp/udp only. uid.op always 0 in other cases */
3603 else if (r->uid.op && (pd->lookup.done || (pd->lookup.done =
3604 #ifdef __FreeBSD__
3605 pf_socket_lookup(direction, pd, inp), 1)) &&
3606 #else
3607 pf_socket_lookup(direction, pd), 1)) &&
3608 #endif
3609 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
3610 pd->lookup.uid))
3611 r = TAILQ_NEXT(r, entries);
3612 /* tcp/udp only. gid.op always 0 in other cases */
3613 else if (r->gid.op && (pd->lookup.done || (pd->lookup.done =
3614 #ifdef __FreeBSD__
3615 pf_socket_lookup(direction, pd, inp), 1)) &&
3616 #else
3617 pf_socket_lookup(direction, pd), 1)) &&
3618 #endif
3619 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
3620 pd->lookup.gid))
3621 r = TAILQ_NEXT(r, entries);
3622 else if (r->prob &&
3623 #ifdef __FreeBSD__
3624 r->prob <= arc4random())
3625 #else
3626 r->prob <= arc4random_uniform(UINT_MAX - 1) + 1)
3627 #endif
3628 r = TAILQ_NEXT(r, entries);
3629 #ifdef __FreeBSD__
3630 else if (r->match_tag && !pf_match_tag(m, r, &tag, pd->pf_mtag))
3631 #else
3632 else if (r->match_tag && !pf_match_tag(m, r, &tag))
3633 #endif
3634 r = TAILQ_NEXT(r, entries);
3635 else if (r->os_fingerprint != PF_OSFP_ANY &&
3636 (pd->proto != IPPROTO_TCP || !pf_osfp_match(
3637 pf_osfp_fingerprint(pd, m, off, th),
3638 r->os_fingerprint)))
3639 r = TAILQ_NEXT(r, entries);
3640 else {
3641 if (r->tag)
3642 tag = r->tag;
3643 if (r->rtableid >= 0)
3644 rtableid = r->rtableid;
3645 if (r->anchor == NULL) {
3646 match = 1;
3647 *rm = r;
3648 *am = a;
3649 *rsm = ruleset;
3650 if ((*rm)->quick)
3651 break;
3652 r = TAILQ_NEXT(r, entries);
3653 } else
3654 pf_step_into_anchor(&asd, &ruleset,
3655 PF_RULESET_FILTER, &r, &a, &match);
3656 }
3657 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3658 PF_RULESET_FILTER, &r, &a, &match))
3659 break;
3660 }
3661 r = *rm;
3662 a = *am;
3663 ruleset = *rsm;
3664
3665 REASON_SET(&reason, PFRES_MATCH);
3666
3667 if (r->log || (nr != NULL && nr->log)) {
3668 if (rewrite)
3669 m_copyback(m, off, hdrlen, pd->hdr.any);
3670 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr,
3671 a, ruleset, pd);
3672 }
3673
3674 if ((r->action == PF_DROP) &&
3675 ((r->rule_flag & PFRULE_RETURNRST) ||
3676 (r->rule_flag & PFRULE_RETURNICMP) ||
3677 (r->rule_flag & PFRULE_RETURN))) {
3678 /* undo NAT changes, if they have taken place */
3679 if (nr != NULL) {
3680 PF_ACPY(saddr, &sk->addr[pd->sidx], af);
3681 PF_ACPY(daddr, &sk->addr[pd->didx], af);
3682 if (pd->sport)
3683 *pd->sport = sk->port[pd->sidx];
3684 if (pd->dport)
3685 *pd->dport = sk->port[pd->didx];
3686 if (pd->proto_sum)
3687 *pd->proto_sum = bproto_sum;
3688 if (pd->ip_sum)
3689 *pd->ip_sum = bip_sum;
3690 m_copyback(m, off, hdrlen, pd->hdr.any);
3691 }
3692 if (pd->proto == IPPROTO_TCP &&
3693 ((r->rule_flag & PFRULE_RETURNRST) ||
3694 (r->rule_flag & PFRULE_RETURN)) &&
3695 !(th->th_flags & TH_RST)) {
3696 u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
3697 int len = 0;
3698 #ifdef INET
3699 struct ip *h4;
3700 #endif
3701 #ifdef INET6
3702 struct ip6_hdr *h6;
3703 #endif
3704
3705 switch (af) {
3706 #ifdef INET
3707 case AF_INET:
3708 h4 = mtod(m, struct ip *);
3709 len = ntohs(h4->ip_len) - off;
3710 break;
3711 #endif
3712 #ifdef INET6
3713 case AF_INET6:
3714 h6 = mtod(m, struct ip6_hdr *);
3715 len = ntohs(h6->ip6_plen) - (off - sizeof(*h6));
3716 break;
3717 #endif
3718 }
3719
3720 if (pf_check_proto_cksum(m, off, len, IPPROTO_TCP, af))
3721 REASON_SET(&reason, PFRES_PROTCKSUM);
3722 else {
3723 if (th->th_flags & TH_SYN)
3724 ack++;
3725 if (th->th_flags & TH_FIN)
3726 ack++;
3727 #ifdef __FreeBSD__
3728 pf_send_tcp(m, r, af, pd->dst,
3729 #else
3730 pf_send_tcp(r, af, pd->dst,
3731 #endif
3732 pd->src, th->th_dport, th->th_sport,
3733 ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0,
3734 r->return_ttl, 1, 0, pd->eh, kif->pfik_ifp);
3735 }
3736 } else if (pd->proto != IPPROTO_ICMP && af == AF_INET &&
3737 r->return_icmp)
3738 pf_send_icmp(m, r->return_icmp >> 8,
3739 r->return_icmp & 255, af, r);
3740 else if (pd->proto != IPPROTO_ICMPV6 && af == AF_INET6 &&
3741 r->return_icmp6)
3742 pf_send_icmp(m, r->return_icmp6 >> 8,
3743 r->return_icmp6 & 255, af, r);
3744 }
3745
3746 if (r->action == PF_DROP)
3747 goto cleanup;
3748
3749 #ifdef __FreeBSD__
3750 if (pf_tag_packet(m, tag, rtableid, pd->pf_mtag)) {
3751 #else
3752 if (pf_tag_packet(m, tag, rtableid)) {
3753 #endif
3754 REASON_SET(&reason, PFRES_MEMORY);
3755 goto cleanup;
3756 }
3757
3758 if (!state_icmp && (r->keep_state || nr != NULL ||
3759 (pd->flags & PFDESC_TCP_NORM))) {
3760 int action;
3761 action = pf_create_state(r, nr, a, pd, nsn, skw, sks, nk, sk, m,
3762 off, sport, dport, &rewrite, kif, sm, tag, bproto_sum,
3763 bip_sum, hdrlen);
3764 if (action != PF_PASS)
3765 return (action);
3766 } else {
3767 #ifdef __FreeBSD__
3768 if (sk != NULL)
3769 pool_put(&V_pf_state_key_pl, sk);
3770 if (nk != NULL)
3771 pool_put(&V_pf_state_key_pl, nk);
3772 #else
3773 if (sk != NULL)
3774 pool_put(&pf_state_key_pl, sk);
3775 if (nk != NULL)
3776 pool_put(&pf_state_key_pl, nk);
3777 #endif
3778 }
3779
3780 /* copy back packet headers if we performed NAT operations */
3781 if (rewrite)
3782 m_copyback(m, off, hdrlen, pd->hdr.any);
3783
3784 #if NPFSYNC > 0
3785 if (*sm != NULL && !ISSET((*sm)->state_flags, PFSTATE_NOSYNC) &&
3786 #ifdef __FreeBSD__
3787 direction == PF_OUT && pfsync_up_ptr != NULL && pfsync_up_ptr()) {
3788 #else
3789 direction == PF_OUT && pfsync_up()) {
3790 #endif
3791 /*
3792 * We want the state created, but we dont
3793 * want to send this in case a partner
3794 * firewall has to know about it to allow
3795 * replies through it.
3796 */
3797 #ifdef __FreeBSD__
3798 if (pfsync_defer_ptr != NULL &&
3799 pfsync_defer_ptr(*sm, m))
3800 #else
3801 if (pfsync_defer(*sm, m))
3802 #endif
3803 return (PF_DEFER);
3804 }
3805 #endif
3806
3807 return (PF_PASS);
3808
3809 cleanup:
3810 #ifdef __FreeBSD__
3811 if (sk != NULL)
3812 pool_put(&V_pf_state_key_pl, sk);
3813 if (nk != NULL)
3814 pool_put(&V_pf_state_key_pl, nk);
3815 #else
3816 if (sk != NULL)
3817 pool_put(&pf_state_key_pl, sk);
3818 if (nk != NULL)
3819 pool_put(&pf_state_key_pl, nk);
3820 #endif
3821 return (PF_DROP);
3822 }
3823
3824 static __inline int
3825 pf_create_state(struct pf_rule *r, struct pf_rule *nr, struct pf_rule *a,
3826 struct pf_pdesc *pd, struct pf_src_node *nsn, struct pf_state_key *skw,
3827 struct pf_state_key *sks, struct pf_state_key *nk, struct pf_state_key *sk,
3828 struct mbuf *m, int off, u_int16_t sport, u_int16_t dport, int *rewrite,
3829 struct pfi_kif *kif, struct pf_state **sm, int tag, u_int16_t bproto_sum,
3830 u_int16_t bip_sum, int hdrlen)
3831 {
3832 struct pf_state *s = NULL;
3833 struct pf_src_node *sn = NULL;
3834 struct tcphdr *th = pd->hdr.tcp;
3835 #ifdef __FreeBSD__
3836 u_int16_t mss = V_tcp_mssdflt;
3837 #else
3838 u_int16_t mss = tcp_mssdflt;
3839 #endif
3840 u_short reason;
3841
3842 /* check maximums */
3843 if (r->max_states && (r->states_cur >= r->max_states)) {
3844 #ifdef __FreeBSD__
3845 V_pf_status.lcounters[LCNT_STATES]++;
3846 #else
3847 pf_status.lcounters[LCNT_STATES]++;
3848 #endif
3849 REASON_SET(&reason, PFRES_MAXSTATES);
3850 return (PF_DROP);
3851 }
3852 /* src node for filter rule */
3853 if ((r->rule_flag & PFRULE_SRCTRACK ||
3854 r->rpool.opts & PF_POOL_STICKYADDR) &&
3855 pf_insert_src_node(&sn, r, pd->src, pd->af) != 0) {
3856 REASON_SET(&reason, PFRES_SRCLIMIT);
3857 goto csfailed;
3858 }
3859 /* src node for translation rule */
3860 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3861 pf_insert_src_node(&nsn, nr, &sk->addr[pd->sidx], pd->af)) {
3862 REASON_SET(&reason, PFRES_SRCLIMIT);
3863 goto csfailed;
3864 }
3865 #ifdef __FreeBSD__
3866 s = pool_get(&V_pf_state_pl, PR_NOWAIT | PR_ZERO);
3867 #else
3868 s = pool_get(&pf_state_pl, PR_NOWAIT | PR_ZERO);
3869 #endif
3870 if (s == NULL) {
3871 REASON_SET(&reason, PFRES_MEMORY);
3872 goto csfailed;
3873 }
3874 s->rule.ptr = r;
3875 s->nat_rule.ptr = nr;
3876 s->anchor.ptr = a;
3877 STATE_INC_COUNTERS(s);
3878 if (r->allow_opts)
3879 s->state_flags |= PFSTATE_ALLOWOPTS;
3880 if (r->rule_flag & PFRULE_STATESLOPPY)
3881 s->state_flags |= PFSTATE_SLOPPY;
3882 if (r->rule_flag & PFRULE_PFLOW)
3883 s->state_flags |= PFSTATE_PFLOW;
3884 s->log = r->log & PF_LOG_ALL;
3885 s->sync_state = PFSYNC_S_NONE;
3886 if (nr != NULL)
3887 s->log |= nr->log & PF_LOG_ALL;
3888 switch (pd->proto) {
3889 case IPPROTO_TCP:
3890 s->src.seqlo = ntohl(th->th_seq);
3891 s->src.seqhi = s->src.seqlo + pd->p_len + 1;
3892 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
3893 r->keep_state == PF_STATE_MODULATE) {
3894 /* Generate sequence number modulator */
3895 if ((s->src.seqdiff = pf_tcp_iss(pd) - s->src.seqlo) ==
3896 0)
3897 s->src.seqdiff = 1;
3898 pf_change_proto_a(m, &th->th_seq, &th->th_sum,
3899 htonl(s->src.seqlo + s->src.seqdiff), 0);
3900 *rewrite = 1;
3901 } else
3902 s->src.seqdiff = 0;
3903 if (th->th_flags & TH_SYN) {
3904 s->src.seqhi++;
3905 s->src.wscale = pf_get_wscale(m, off,
3906 th->th_off, pd->af);
3907 }
3908 s->src.max_win = MAX(ntohs(th->th_win), 1);
3909 if (s->src.wscale & PF_WSCALE_MASK) {
3910 /* Remove scale factor from initial window */
3911 int win = s->src.max_win;
3912 win += 1 << (s->src.wscale & PF_WSCALE_MASK);
3913 s->src.max_win = (win - 1) >>
3914 (s->src.wscale & PF_WSCALE_MASK);
3915 }
3916 if (th->th_flags & TH_FIN)
3917 s->src.seqhi++;
3918 s->dst.seqhi = 1;
3919 s->dst.max_win = 1;
3920 s->src.state = TCPS_SYN_SENT;
3921 s->dst.state = TCPS_CLOSED;
3922 s->timeout = PFTM_TCP_FIRST_PACKET;
3923 break;
3924 case IPPROTO_UDP:
3925 s->src.state = PFUDPS_SINGLE;
3926 s->dst.state = PFUDPS_NO_TRAFFIC;
3927 s->timeout = PFTM_UDP_FIRST_PACKET;
3928 break;
3929 case IPPROTO_ICMP:
3930 #ifdef INET6
3931 case IPPROTO_ICMPV6:
3932 #endif
3933 s->timeout = PFTM_ICMP_FIRST_PACKET;
3934 break;
3935 default:
3936 s->src.state = PFOTHERS_SINGLE;
3937 s->dst.state = PFOTHERS_NO_TRAFFIC;
3938 s->timeout = PFTM_OTHER_FIRST_PACKET;
3939 }
3940
3941 s->creation = time_second;
3942 s->expire = time_second;
3943
3944 if (sn != NULL) {
3945 s->src_node = sn;
3946 s->src_node->states++;
3947 }
3948 if (nsn != NULL) {
3949 /* XXX We only modify one side for now. */
3950 PF_ACPY(&nsn->raddr, &nk->addr[1], pd->af);
3951 s->nat_src_node = nsn;
3952 s->nat_src_node->states++;
3953 }
3954 if (pd->proto == IPPROTO_TCP) {
3955 if ((pd->flags & PFDESC_TCP_NORM) && pf_normalize_tcp_init(m,
3956 off, pd, th, &s->src, &s->dst)) {
3957 REASON_SET(&reason, PFRES_MEMORY);
3958 pf_src_tree_remove_state(s);
3959 STATE_DEC_COUNTERS(s);
3960 #ifdef __FreeBSD__
3961 pool_put(&V_pf_state_pl, s);
3962 #else
3963 pool_put(&pf_state_pl, s);
3964 #endif
3965 return (PF_DROP);
3966 }
3967 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
3968 pf_normalize_tcp_stateful(m, off, pd, &reason, th, s,
3969 &s->src, &s->dst, rewrite)) {
3970 /* This really shouldn't happen!!! */
3971 DPFPRINTF(PF_DEBUG_URGENT,
3972 ("pf_normalize_tcp_stateful failed on first pkt"));
3973 pf_normalize_tcp_cleanup(s);
3974 pf_src_tree_remove_state(s);
3975 STATE_DEC_COUNTERS(s);
3976 #ifdef __FreeBSD__
3977 pool_put(&V_pf_state_pl, s);
3978 #else
3979 pool_put(&pf_state_pl, s);
3980 #endif
3981 return (PF_DROP);
3982 }
3983 }
3984 s->direction = pd->dir;
3985
3986 if (sk == NULL && pf_state_key_setup(pd, nr, &skw, &sks, &sk, &nk,
3987 pd->src, pd->dst, sport, dport))
3988 goto csfailed;
3989
3990 if (pf_state_insert(BOUND_IFACE(r, kif), skw, sks, s)) {
3991 if (pd->proto == IPPROTO_TCP)
3992 pf_normalize_tcp_cleanup(s);
3993 REASON_SET(&reason, PFRES_STATEINS);
3994 pf_src_tree_remove_state(s);
3995 STATE_DEC_COUNTERS(s);
3996 #ifdef __FreeBSD__
3997 pool_put(&V_pf_state_pl, s);
3998 #else
3999 pool_put(&pf_state_pl, s);
4000 #endif
4001 return (PF_DROP);
4002 } else
4003 *sm = s;
4004
4005 pf_set_rt_ifp(s, pd->src); /* needs s->state_key set */
4006 if (tag > 0) {
4007 pf_tag_ref(tag);
4008 s->tag = tag;
4009 }
4010 if (pd->proto == IPPROTO_TCP && (th->th_flags & (TH_SYN|TH_ACK)) ==
4011 TH_SYN && r->keep_state == PF_STATE_SYNPROXY) {
4012 s->src.state = PF_TCPS_PROXY_SRC;
4013 /* undo NAT changes, if they have taken place */
4014 if (nr != NULL) {
4015 struct pf_state_key *skt = s->key[PF_SK_WIRE];
4016 if (pd->dir == PF_OUT)
4017 skt = s->key[PF_SK_STACK];
4018 PF_ACPY(pd->src, &skt->addr[pd->sidx], pd->af);
4019 PF_ACPY(pd->dst, &skt->addr[pd->didx], pd->af);
4020 if (pd->sport)
4021 *pd->sport = skt->port[pd->sidx];
4022 if (pd->dport)
4023 *pd->dport = skt->port[pd->didx];
4024 if (pd->proto_sum)
4025 *pd->proto_sum = bproto_sum;
4026 if (pd->ip_sum)
4027 *pd->ip_sum = bip_sum;
4028 m_copyback(m, off, hdrlen, pd->hdr.any);
4029 }
4030 s->src.seqhi = htonl(arc4random());
4031 /* Find mss option */
4032 int rtid = M_GETFIB(m);
4033 mss = pf_get_mss(m, off, th->th_off, pd->af);
4034 mss = pf_calc_mss(pd->src, pd->af, rtid, mss);
4035 mss = pf_calc_mss(pd->dst, pd->af, rtid, mss);
4036 s->src.mss = mss;
4037 #ifdef __FreeBSD__
4038 pf_send_tcp(NULL, r, pd->af, pd->dst, pd->src, th->th_dport,
4039 #else
4040 pf_send_tcp(r, pd->af, pd->dst, pd->src, th->th_dport,
4041 #endif
4042 th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1,
4043 TH_SYN|TH_ACK, 0, s->src.mss, 0, 1, 0, NULL, NULL);
4044 REASON_SET(&reason, PFRES_SYNPROXY);
4045 return (PF_SYNPROXY_DROP);
4046 }
4047
4048 return (PF_PASS);
4049
4050 csfailed:
4051 #ifdef __FreeBSD__
4052 if (sk != NULL)
4053 pool_put(&V_pf_state_key_pl, sk);
4054 if (nk != NULL)
4055 pool_put(&V_pf_state_key_pl, nk);
4056 #else
4057 if (sk != NULL)
4058 pool_put(&pf_state_key_pl, sk);
4059 if (nk != NULL)
4060 pool_put(&pf_state_key_pl, nk);
4061 #endif
4062
4063 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
4064 #ifdef __FreeBSD__
4065 RB_REMOVE(pf_src_tree, &V_tree_src_tracking, sn);
4066 V_pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
4067 V_pf_status.src_nodes--;
4068 pool_put(&V_pf_src_tree_pl, sn);
4069 #else
4070 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
4071 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
4072 pf_status.src_nodes--;
4073 pool_put(&pf_src_tree_pl, sn);
4074 #endif
4075 }
4076 if (nsn != sn && nsn != NULL && nsn->states == 0 && nsn->expire == 0) {
4077 #ifdef __FreeBSD__
4078 RB_REMOVE(pf_src_tree, &V_tree_src_tracking, nsn);
4079 V_pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
4080 V_pf_status.src_nodes--;
4081 pool_put(&V_pf_src_tree_pl, nsn);
4082 #else
4083 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
4084 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
4085 pf_status.src_nodes--;
4086 pool_put(&pf_src_tree_pl, nsn);
4087 #endif
4088 }
4089 return (PF_DROP);
4090 }
4091
4092 int
4093 pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif,
4094 struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am,
4095 struct pf_ruleset **rsm)
4096 {
4097 struct pf_rule *r, *a = NULL;
4098 struct pf_ruleset *ruleset = NULL;
4099 sa_family_t af = pd->af;
4100 u_short reason;
4101 int tag = -1;
4102 int asd = 0;
4103 int match = 0;
4104
4105 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
4106 while (r != NULL) {
4107 r->evaluations++;
4108 if (pfi_kif_match(r->kif, kif) == r->ifnot)
4109 r = r->skip[PF_SKIP_IFP].ptr;
4110 else if (r->direction && r->direction != direction)
4111 r = r->skip[PF_SKIP_DIR].ptr;
4112 else if (r->af && r->af != af)
4113 r = r->skip[PF_SKIP_AF].ptr;
4114 else if (r->proto && r->proto != pd->proto)
4115 r = r->skip[PF_SKIP_PROTO].ptr;
4116 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af,
4117 r->src.neg, kif, M_GETFIB(m)))
4118 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
4119 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af,
4120 r->dst.neg, NULL, M_GETFIB(m)))
4121 r = r->skip[PF_SKIP_DST_ADDR].ptr;
4122 else if (r->tos && !(r->tos == pd->tos))
4123 r = TAILQ_NEXT(r, entries);
4124 else if (r->os_fingerprint != PF_OSFP_ANY)
4125 r = TAILQ_NEXT(r, entries);
4126 else if (pd->proto == IPPROTO_UDP &&
4127 (r->src.port_op || r->dst.port_op))
4128 r = TAILQ_NEXT(r, entries);
4129 else if (pd->proto == IPPROTO_TCP &&
4130 (r->src.port_op || r->dst.port_op || r->flagset))
4131 r = TAILQ_NEXT(r, entries);
4132 else if ((pd->proto == IPPROTO_ICMP ||
4133 pd->proto == IPPROTO_ICMPV6) &&
4134 (r->type || r->code))
4135 r = TAILQ_NEXT(r, entries);
4136 else if (r->prob && r->prob <=
4137 (arc4random() % (UINT_MAX - 1) + 1))
4138 r = TAILQ_NEXT(r, entries);
4139 #ifdef __FreeBSD__
4140 else if (r->match_tag && !pf_match_tag(m, r, &tag, pd->pf_mtag))
4141 #else
4142 else if (r->match_tag && !pf_match_tag(m, r, &tag))
4143 #endif
4144 r = TAILQ_NEXT(r, entries);
4145 else {
4146 if (r->anchor == NULL) {
4147 match = 1;
4148 *rm = r;
4149 *am = a;
4150 *rsm = ruleset;
4151 if ((*rm)->quick)
4152 break;
4153 r = TAILQ_NEXT(r, entries);
4154 } else
4155 pf_step_into_anchor(&asd, &ruleset,
4156 PF_RULESET_FILTER, &r, &a, &match);
4157 }
4158 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
4159 PF_RULESET_FILTER, &r, &a, &match))
4160 break;
4161 }
4162 r = *rm;
4163 a = *am;
4164 ruleset = *rsm;
4165
4166 REASON_SET(&reason, PFRES_MATCH);
4167
4168 if (r->log)
4169 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset,
4170 pd);
4171
4172 if (r->action != PF_PASS)
4173 return (PF_DROP);
4174
4175 #ifdef __FreeBSD__
4176 if (pf_tag_packet(m, tag, -1, pd->pf_mtag)) {
4177 #else
4178 if (pf_tag_packet(m, tag, -1)) {
4179 #endif
4180 REASON_SET(&reason, PFRES_MEMORY);
4181 return (PF_DROP);
4182 }
4183
4184 return (PF_PASS);
4185 }
4186
4187 int
4188 pf_tcp_track_full(struct pf_state_peer *src, struct pf_state_peer *dst,
4189 struct pf_state **state, struct pfi_kif *kif, struct mbuf *m, int off,
4190 struct pf_pdesc *pd, u_short *reason, int *copyback)
4191 {
4192 struct tcphdr *th = pd->hdr.tcp;
4193 u_int16_t win = ntohs(th->th_win);
4194 u_int32_t ack, end, seq, orig_seq;
4195 u_int8_t sws, dws;
4196 int ackskew;
4197
4198 if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) {
4199 sws = src->wscale & PF_WSCALE_MASK;
4200 dws = dst->wscale & PF_WSCALE_MASK;
4201 } else
4202 sws = dws = 0;
4203
4204 /*
4205 * Sequence tracking algorithm from Guido van Rooij's paper:
4206 * http://www.madison-gurkha.com/publications/tcp_filtering/
4207 * tcp_filtering.ps
4208 */
4209
4210 orig_seq = seq = ntohl(th->th_seq);
4211 if (src->seqlo == 0) {
4212 /* First packet from this end. Set its state */
4213
4214 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) &&
4215 src->scrub == NULL) {
4216 if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) {
4217 REASON_SET(reason, PFRES_MEMORY);
4218 return (PF_DROP);
4219 }
4220 }
4221
4222 /* Deferred generation of sequence number modulator */
4223 if (dst->seqdiff && !src->seqdiff) {
4224 /* use random iss for the TCP server */
4225 while ((src->seqdiff = arc4random() - seq) == 0)
4226 ;
4227 ack = ntohl(th->th_ack) - dst->seqdiff;
4228 pf_change_proto_a(m, &th->th_seq, &th->th_sum, htonl(seq +
4229 src->seqdiff), 0);
4230 pf_change_proto_a(m, &th->th_ack, &th->th_sum, htonl(ack), 0);
4231 *copyback = 1;
4232 } else {
4233 ack = ntohl(th->th_ack);
4234 }
4235
4236 end = seq + pd->p_len;
4237 if (th->th_flags & TH_SYN) {
4238 end++;
4239 if (dst->wscale & PF_WSCALE_FLAG) {
4240 src->wscale = pf_get_wscale(m, off, th->th_off,
4241 pd->af);
4242 if (src->wscale & PF_WSCALE_FLAG) {
4243 /* Remove scale factor from initial
4244 * window */
4245 sws = src->wscale & PF_WSCALE_MASK;
4246 win = ((u_int32_t)win + (1 << sws) - 1)
4247 >> sws;
4248 dws = dst->wscale & PF_WSCALE_MASK;
4249 } else {
4250 /* fixup other window */
4251 dst->max_win <<= dst->wscale &
4252 PF_WSCALE_MASK;
4253 /* in case of a retrans SYN|ACK */
4254 dst->wscale = 0;
4255 }
4256 }
4257 }
4258 if (th->th_flags & TH_FIN)
4259 end++;
4260
4261 src->seqlo = seq;
4262 if (src->state < TCPS_SYN_SENT)
4263 src->state = TCPS_SYN_SENT;
4264
4265 /*
4266 * May need to slide the window (seqhi may have been set by
4267 * the crappy stack check or if we picked up the connection
4268 * after establishment)
4269 */
4270 if (src->seqhi == 1 ||
4271 SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi))
4272 src->seqhi = end + MAX(1, dst->max_win << dws);
4273 if (win > src->max_win)
4274 src->max_win = win;
4275
4276 } else {
4277 ack = ntohl(th->th_ack) - dst->seqdiff;
4278 if (src->seqdiff) {
4279 /* Modulate sequence numbers */
4280 pf_change_proto_a(m, &th->th_seq, &th->th_sum, htonl(seq +
4281 src->seqdiff), 0);
4282 pf_change_proto_a(m, &th->th_ack, &th->th_sum, htonl(ack), 0);
4283 *copyback = 1;
4284 }
4285 end = seq + pd->p_len;
4286 if (th->th_flags & TH_SYN)
4287 end++;
4288 if (th->th_flags & TH_FIN)
4289 end++;
4290 }
4291
4292 if ((th->th_flags & TH_ACK) == 0) {
4293 /* Let it pass through the ack skew check */
4294 ack = dst->seqlo;
4295 } else if ((ack == 0 &&
4296 (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) ||
4297 /* broken tcp stacks do not set ack */
4298 (dst->state < TCPS_SYN_SENT)) {
4299 /*
4300 * Many stacks (ours included) will set the ACK number in an
4301 * FIN|ACK if the SYN times out -- no sequence to ACK.
4302 */
4303 ack = dst->seqlo;
4304 }
4305
4306 if (seq == end) {
4307 /* Ease sequencing restrictions on no data packets */
4308 seq = src->seqlo;
4309 end = seq;
4310 }
4311
4312 ackskew = dst->seqlo - ack;
4313
4314
4315 /*
4316 * Need to demodulate the sequence numbers in any TCP SACK options
4317 * (Selective ACK). We could optionally validate the SACK values
4318 * against the current ACK window, either forwards or backwards, but
4319 * I'm not confident that SACK has been implemented properly
4320 * everywhere. It wouldn't surprise me if several stacks accidently
4321 * SACK too far backwards of previously ACKed data. There really aren't
4322 * any security implications of bad SACKing unless the target stack
4323 * doesn't validate the option length correctly. Someone trying to
4324 * spoof into a TCP connection won't bother blindly sending SACK
4325 * options anyway.
4326 */
4327 if (dst->seqdiff && (th->th_off << 2) > sizeof(struct tcphdr)) {
4328 if (pf_modulate_sack(m, off, pd, th, dst))
4329 *copyback = 1;
4330 }
4331
4332
4333 #define MAXACKWINDOW (0xffff + 1500) /* 1500 is an arbitrary fudge factor */
4334 if (SEQ_GEQ(src->seqhi, end) &&
4335 /* Last octet inside other's window space */
4336 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) &&
4337 /* Retrans: not more than one window back */
4338 (ackskew >= -MAXACKWINDOW) &&
4339 /* Acking not more than one reassembled fragment backwards */
4340 (ackskew <= (MAXACKWINDOW << sws)) &&
4341 /* Acking not more than one window forward */
4342 ((th->th_flags & TH_RST) == 0 || orig_seq == src->seqlo ||
4343 (orig_seq == src->seqlo + 1) || (orig_seq + 1 == src->seqlo) ||
4344 (pd->flags & PFDESC_IP_REAS) == 0)) {
4345 /* Require an exact/+1 sequence match on resets when possible */
4346
4347 if (dst->scrub || src->scrub) {
4348 if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4349 *state, src, dst, copyback))
4350 return (PF_DROP);
4351 }
4352
4353 /* update max window */
4354 if (src->max_win < win)
4355 src->max_win = win;
4356 /* synchronize sequencing */
4357 if (SEQ_GT(end, src->seqlo))
4358 src->seqlo = end;
4359 /* slide the window of what the other end can send */
4360 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4361 dst->seqhi = ack + MAX((win << sws), 1);
4362
4363
4364 /* update states */
4365 if (th->th_flags & TH_SYN)
4366 if (src->state < TCPS_SYN_SENT)
4367 src->state = TCPS_SYN_SENT;
4368 if (th->th_flags & TH_FIN)
4369 if (src->state < TCPS_CLOSING)
4370 src->state = TCPS_CLOSING;
4371 if (th->th_flags & TH_ACK) {
4372 if (dst->state == TCPS_SYN_SENT) {
4373 dst->state = TCPS_ESTABLISHED;
4374 if (src->state == TCPS_ESTABLISHED &&
4375 (*state)->src_node != NULL &&
4376 pf_src_connlimit(state)) {
4377 REASON_SET(reason, PFRES_SRCLIMIT);
4378 return (PF_DROP);
4379 }
4380 } else if (dst->state == TCPS_CLOSING)
4381 dst->state = TCPS_FIN_WAIT_2;
4382 }
4383 if (th->th_flags & TH_RST)
4384 src->state = dst->state = TCPS_TIME_WAIT;
4385
4386 /* update expire time */
4387 (*state)->expire = time_second;
4388 if (src->state >= TCPS_FIN_WAIT_2 &&
4389 dst->state >= TCPS_FIN_WAIT_2)
4390 (*state)->timeout = PFTM_TCP_CLOSED;
4391 else if (src->state >= TCPS_CLOSING &&
4392 dst->state >= TCPS_CLOSING)
4393 (*state)->timeout = PFTM_TCP_FIN_WAIT;
4394 else if (src->state < TCPS_ESTABLISHED ||
4395 dst->state < TCPS_ESTABLISHED)
4396 (*state)->timeout = PFTM_TCP_OPENING;
4397 else if (src->state >= TCPS_CLOSING ||
4398 dst->state >= TCPS_CLOSING)
4399 (*state)->timeout = PFTM_TCP_CLOSING;
4400 else
4401 (*state)->timeout = PFTM_TCP_ESTABLISHED;
4402
4403 /* Fall through to PASS packet */
4404
4405 } else if ((dst->state < TCPS_SYN_SENT ||
4406 dst->state >= TCPS_FIN_WAIT_2 ||
4407 src->state >= TCPS_FIN_WAIT_2) &&
4408 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) &&
4409 /* Within a window forward of the originating packet */
4410 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) {
4411 /* Within a window backward of the originating packet */
4412
4413 /*
4414 * This currently handles three situations:
4415 * 1) Stupid stacks will shotgun SYNs before their peer
4416 * replies.
4417 * 2) When PF catches an already established stream (the
4418 * firewall rebooted, the state table was flushed, routes
4419 * changed...)
4420 * 3) Packets get funky immediately after the connection
4421 * closes (this should catch Solaris spurious ACK|FINs
4422 * that web servers like to spew after a close)
4423 *
4424 * This must be a little more careful than the above code
4425 * since packet floods will also be caught here. We don't
4426 * update the TTL here to mitigate the damage of a packet
4427 * flood and so the same code can handle awkward establishment
4428 * and a loosened connection close.
4429 * In the establishment case, a correct peer response will
4430 * validate the connection, go through the normal state code
4431 * and keep updating the state TTL.
4432 */
4433
4434 #ifdef __FreeBSD__
4435 if (V_pf_status.debug >= PF_DEBUG_MISC) {
4436 #else
4437 if (pf_status.debug >= PF_DEBUG_MISC) {
4438 #endif
4439 printf("pf: loose state match: ");
4440 pf_print_state(*state);
4441 pf_print_flags(th->th_flags);
4442 printf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
4443 "pkts=%llu:%llu dir=%s,%s\n", seq, orig_seq, ack,
4444 #ifdef __FreeBSD__
4445 pd->p_len, ackskew, (unsigned long long)(*state)->packets[0],
4446 (unsigned long long)(*state)->packets[1],
4447 #else
4448 pd->p_len, ackskew, (*state)->packets[0],
4449 (*state)->packets[1],
4450 #endif
4451 pd->dir == PF_IN ? "in" : "out",
4452 pd->dir == (*state)->direction ? "fwd" : "rev");
4453 }
4454
4455 if (dst->scrub || src->scrub) {
4456 if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4457 *state, src, dst, copyback))
4458 return (PF_DROP);
4459 }
4460
4461 /* update max window */
4462 if (src->max_win < win)
4463 src->max_win = win;
4464 /* synchronize sequencing */
4465 if (SEQ_GT(end, src->seqlo))
4466 src->seqlo = end;
4467 /* slide the window of what the other end can send */
4468 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4469 dst->seqhi = ack + MAX((win << sws), 1);
4470
4471 /*
4472 * Cannot set dst->seqhi here since this could be a shotgunned
4473 * SYN and not an already established connection.
4474 */
4475
4476 if (th->th_flags & TH_FIN)
4477 if (src->state < TCPS_CLOSING)
4478 src->state = TCPS_CLOSING;
4479 if (th->th_flags & TH_RST)
4480 src->state = dst->state = TCPS_TIME_WAIT;
4481
4482 /* Fall through to PASS packet */
4483
4484 } else {
4485 if ((*state)->dst.state == TCPS_SYN_SENT &&
4486 (*state)->src.state == TCPS_SYN_SENT) {
4487 /* Send RST for state mismatches during handshake */
4488 if (!(th->th_flags & TH_RST))
4489 #ifdef __FreeBSD__
4490 pf_send_tcp(NULL, (*state)->rule.ptr, pd->af,
4491 #else
4492 pf_send_tcp((*state)->rule.ptr, pd->af,
4493 #endif
4494 pd->dst, pd->src, th->th_dport,
4495 th->th_sport, ntohl(th->th_ack), 0,
4496 TH_RST, 0, 0,
4497 (*state)->rule.ptr->return_ttl, 1, 0,
4498 pd->eh, kif->pfik_ifp);
4499 src->seqlo = 0;
4500 src->seqhi = 1;
4501 src->max_win = 1;
4502 #ifdef __FreeBSD__
4503 } else if (V_pf_status.debug >= PF_DEBUG_MISC) {
4504 #else
4505 } else if (pf_status.debug >= PF_DEBUG_MISC) {
4506 #endif
4507 printf("pf: BAD state: ");
4508 pf_print_state(*state);
4509 pf_print_flags(th->th_flags);
4510 printf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
4511 "pkts=%llu:%llu dir=%s,%s\n",
4512 seq, orig_seq, ack, pd->p_len, ackskew,
4513 #ifdef __FreeBSD__
4514 (unsigned long long)(*state)->packets[0],
4515 (unsigned long long)(*state)->packets[1],
4516 #else
4517 (*state)->packets[0], (*state)->packets[1],
4518 #endif
4519 pd->dir == PF_IN ? "in" : "out",
4520 pd->dir == (*state)->direction ? "fwd" : "rev");
4521 printf("pf: State failure on: %c %c %c %c | %c %c\n",
4522 SEQ_GEQ(src->seqhi, end) ? ' ' : '1',
4523 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ?
4524 ' ': '2',
4525 (ackskew >= -MAXACKWINDOW) ? ' ' : '3',
4526 (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4',
4527 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5',
4528 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6');
4529 }
4530 REASON_SET(reason, PFRES_BADSTATE);
4531 return (PF_DROP);
4532 }
4533
4534 return (PF_PASS);
4535 }
4536
4537 int
4538 pf_tcp_track_sloppy(struct pf_state_peer *src, struct pf_state_peer *dst,
4539 struct pf_state **state, struct pf_pdesc *pd, u_short *reason)
4540 {
4541 struct tcphdr *th = pd->hdr.tcp;
4542
4543 if (th->th_flags & TH_SYN)
4544 if (src->state < TCPS_SYN_SENT)
4545 src->state = TCPS_SYN_SENT;
4546 if (th->th_flags & TH_FIN)
4547 if (src->state < TCPS_CLOSING)
4548 src->state = TCPS_CLOSING;
4549 if (th->th_flags & TH_ACK) {
4550 if (dst->state == TCPS_SYN_SENT) {
4551 dst->state = TCPS_ESTABLISHED;
4552 if (src->state == TCPS_ESTABLISHED &&
4553 (*state)->src_node != NULL &&
4554 pf_src_connlimit(state)) {
4555 REASON_SET(reason, PFRES_SRCLIMIT);
4556 return (PF_DROP);
4557 }
4558 } else if (dst->state == TCPS_CLOSING) {
4559 dst->state = TCPS_FIN_WAIT_2;
4560 } else if (src->state == TCPS_SYN_SENT &&
4561 dst->state < TCPS_SYN_SENT) {
4562 /*
4563 * Handle a special sloppy case where we only see one
4564 * half of the connection. If there is a ACK after
4565 * the initial SYN without ever seeing a packet from
4566 * the destination, set the connection to established.
4567 */
4568 dst->state = src->state = TCPS_ESTABLISHED;
4569 if ((*state)->src_node != NULL &&
4570 pf_src_connlimit(state)) {
4571 REASON_SET(reason, PFRES_SRCLIMIT);
4572 return (PF_DROP);
4573 }
4574 } else if (src->state == TCPS_CLOSING &&
4575 dst->state == TCPS_ESTABLISHED &&
4576 dst->seqlo == 0) {
4577 /*
4578 * Handle the closing of half connections where we
4579 * don't see the full bidirectional FIN/ACK+ACK
4580 * handshake.
4581 */
4582 dst->state = TCPS_CLOSING;
4583 }
4584 }
4585 if (th->th_flags & TH_RST)
4586 src->state = dst->state = TCPS_TIME_WAIT;
4587
4588 /* update expire time */
4589 (*state)->expire = time_second;
4590 if (src->state >= TCPS_FIN_WAIT_2 &&
4591 dst->state >= TCPS_FIN_WAIT_2)
4592 (*state)->timeout = PFTM_TCP_CLOSED;
4593 else if (src->state >= TCPS_CLOSING &&
4594 dst->state >= TCPS_CLOSING)
4595 (*state)->timeout = PFTM_TCP_FIN_WAIT;
4596 else if (src->state < TCPS_ESTABLISHED ||
4597 dst->state < TCPS_ESTABLISHED)
4598 (*state)->timeout = PFTM_TCP_OPENING;
4599 else if (src->state >= TCPS_CLOSING ||
4600 dst->state >= TCPS_CLOSING)
4601 (*state)->timeout = PFTM_TCP_CLOSING;
4602 else
4603 (*state)->timeout = PFTM_TCP_ESTABLISHED;
4604
4605 return (PF_PASS);
4606 }
4607
4608 int
4609 pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
4610 struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
4611 u_short *reason)
4612 {
4613 struct pf_state_key_cmp key;
4614 struct tcphdr *th = pd->hdr.tcp;
4615 int copyback = 0;
4616 struct pf_state_peer *src, *dst;
4617 struct pf_state_key *sk;
4618
4619 key.af = pd->af;
4620 key.proto = IPPROTO_TCP;
4621 if (direction == PF_IN) { /* wire side, straight */
4622 PF_ACPY(&key.addr[0], pd->src, key.af);
4623 PF_ACPY(&key.addr[1], pd->dst, key.af);
4624 key.port[0] = th->th_sport;
4625 key.port[1] = th->th_dport;
4626 } else { /* stack side, reverse */
4627 PF_ACPY(&key.addr[1], pd->src, key.af);
4628 PF_ACPY(&key.addr[0], pd->dst, key.af);
4629 key.port[1] = th->th_sport;
4630 key.port[0] = th->th_dport;
4631 }
4632
4633 #ifdef __FreeBSD__
4634 STATE_LOOKUP(kif, &key, direction, *state, m, pd->pf_mtag);
4635 #else
4636 STATE_LOOKUP(kif, &key, direction, *state, m);
4637 #endif
4638
4639 if (direction == (*state)->direction) {
4640 src = &(*state)->src;
4641 dst = &(*state)->dst;
4642 } else {
4643 src = &(*state)->dst;
4644 dst = &(*state)->src;
4645 }
4646
4647 sk = (*state)->key[pd->didx];
4648
4649 if ((*state)->src.state == PF_TCPS_PROXY_SRC) {
4650 if (direction != (*state)->direction) {
4651 REASON_SET(reason, PFRES_SYNPROXY);
4652 return (PF_SYNPROXY_DROP);
4653 }
4654 if (th->th_flags & TH_SYN) {
4655 if (ntohl(th->th_seq) != (*state)->src.seqlo) {
4656 REASON_SET(reason, PFRES_SYNPROXY);
4657 return (PF_DROP);
4658 }
4659 #ifdef __FreeBSD__
4660 pf_send_tcp(NULL, (*state)->rule.ptr, pd->af, pd->dst,
4661 #else
4662 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
4663 #endif
4664 pd->src, th->th_dport, th->th_sport,
4665 (*state)->src.seqhi, ntohl(th->th_seq) + 1,
4666 TH_SYN|TH_ACK, 0, (*state)->src.mss, 0, 1,
4667 0, NULL, NULL);
4668 REASON_SET(reason, PFRES_SYNPROXY);
4669 return (PF_SYNPROXY_DROP);
4670 } else if (!(th->th_flags & TH_ACK) ||
4671 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
4672 (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
4673 REASON_SET(reason, PFRES_SYNPROXY);
4674 return (PF_DROP);
4675 } else if ((*state)->src_node != NULL &&
4676 pf_src_connlimit(state)) {
4677 REASON_SET(reason, PFRES_SRCLIMIT);
4678 return (PF_DROP);
4679 } else
4680 (*state)->src.state = PF_TCPS_PROXY_DST;
4681 }
4682 if ((*state)->src.state == PF_TCPS_PROXY_DST) {
4683 if (direction == (*state)->direction) {
4684 if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) ||
4685 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
4686 (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
4687 REASON_SET(reason, PFRES_SYNPROXY);
4688 return (PF_DROP);
4689 }
4690 (*state)->src.max_win = MAX(ntohs(th->th_win), 1);
4691 if ((*state)->dst.seqhi == 1)
4692 (*state)->dst.seqhi = htonl(arc4random());
4693 #ifdef __FreeBSD__
4694 pf_send_tcp(NULL, (*state)->rule.ptr, pd->af,
4695 #else
4696 pf_send_tcp((*state)->rule.ptr, pd->af,
4697 #endif
4698 &sk->addr[pd->sidx], &sk->addr[pd->didx],
4699 sk->port[pd->sidx], sk->port[pd->didx],
4700 (*state)->dst.seqhi, 0, TH_SYN, 0,
4701 (*state)->src.mss, 0, 0, (*state)->tag, NULL, NULL);
4702 REASON_SET(reason, PFRES_SYNPROXY);
4703 return (PF_SYNPROXY_DROP);
4704 } else if (((th->th_flags & (TH_SYN|TH_ACK)) !=
4705 (TH_SYN|TH_ACK)) ||
4706 (ntohl(th->th_ack) != (*state)->dst.seqhi + 1)) {
4707 REASON_SET(reason, PFRES_SYNPROXY);
4708 return (PF_DROP);
4709 } else {
4710 (*state)->dst.max_win = MAX(ntohs(th->th_win), 1);
4711 (*state)->dst.seqlo = ntohl(th->th_seq);
4712 #ifdef __FreeBSD__
4713 pf_send_tcp(NULL, (*state)->rule.ptr, pd->af, pd->dst,
4714 #else
4715 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
4716 #endif
4717 pd->src, th->th_dport, th->th_sport,
4718 ntohl(th->th_ack), ntohl(th->th_seq) + 1,
4719 TH_ACK, (*state)->src.max_win, 0, 0, 0,
4720 (*state)->tag, NULL, NULL);
4721 #ifdef __FreeBSD__
4722 pf_send_tcp(NULL, (*state)->rule.ptr, pd->af,
4723 #else
4724 pf_send_tcp((*state)->rule.ptr, pd->af,
4725 #endif
4726 &sk->addr[pd->sidx], &sk->addr[pd->didx],
4727 sk->port[pd->sidx], sk->port[pd->didx],
4728 (*state)->src.seqhi + 1, (*state)->src.seqlo + 1,
4729 TH_ACK, (*state)->dst.max_win, 0, 0, 1,
4730 0, NULL, NULL);
4731 (*state)->src.seqdiff = (*state)->dst.seqhi -
4732 (*state)->src.seqlo;
4733 (*state)->dst.seqdiff = (*state)->src.seqhi -
4734 (*state)->dst.seqlo;
4735 (*state)->src.seqhi = (*state)->src.seqlo +
4736 (*state)->dst.max_win;
4737 (*state)->dst.seqhi = (*state)->dst.seqlo +
4738 (*state)->src.max_win;
4739 (*state)->src.wscale = (*state)->dst.wscale = 0;
4740 (*state)->src.state = (*state)->dst.state =
4741 TCPS_ESTABLISHED;
4742 REASON_SET(reason, PFRES_SYNPROXY);
4743 return (PF_SYNPROXY_DROP);
4744 }
4745 }
4746
4747 if (((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN) &&
4748 dst->state >= TCPS_FIN_WAIT_2 &&
4749 src->state >= TCPS_FIN_WAIT_2) {
4750 #ifdef __FreeBSD__
4751 if (V_pf_status.debug >= PF_DEBUG_MISC) {
4752 #else
4753 if (pf_status.debug >= PF_DEBUG_MISC) {
4754 #endif
4755 printf("pf: state reuse ");
4756 pf_print_state(*state);
4757 pf_print_flags(th->th_flags);
4758 printf("\n");
4759 }
4760 /* XXX make sure it's the same direction ?? */
4761 (*state)->src.state = (*state)->dst.state = TCPS_CLOSED;
4762 pf_unlink_state(*state);
4763 *state = NULL;
4764 return (PF_DROP);
4765 }
4766
4767 if ((*state)->state_flags & PFSTATE_SLOPPY) {
4768 if (pf_tcp_track_sloppy(src, dst, state, pd, reason) == PF_DROP)
4769 return (PF_DROP);
4770 } else {
4771 if (pf_tcp_track_full(src, dst, state, kif, m, off, pd, reason,
4772 ©back) == PF_DROP)
4773 return (PF_DROP);
4774 }
4775
4776 /* translate source/destination address, if necessary */
4777 if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) {
4778 struct pf_state_key *nk = (*state)->key[pd->didx];
4779
4780 if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], pd->af) ||
4781 nk->port[pd->sidx] != th->th_sport)
4782 pf_change_ap(m, pd->src, &th->th_sport,
4783 pd->ip_sum, &th->th_sum, &nk->addr[pd->sidx],
4784 nk->port[pd->sidx], 0, pd->af);
4785
4786 if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], pd->af) ||
4787 nk->port[pd->didx] != th->th_dport)
4788 pf_change_ap(m, pd->dst, &th->th_dport,
4789 pd->ip_sum, &th->th_sum, &nk->addr[pd->didx],
4790 nk->port[pd->didx], 0, pd->af);
4791 copyback = 1;
4792 }
4793
4794 /* Copyback sequence modulation or stateful scrub changes if needed */
4795 if (copyback)
4796 #ifdef __FreeBSD__
4797 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4798 #else
4799 m_copyback(m, off, sizeof(*th), th);
4800 #endif
4801
4802 return (PF_PASS);
4803 }
4804
4805 int
4806 pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif,
4807 struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4808 {
4809 struct pf_state_peer *src, *dst;
4810 struct pf_state_key_cmp key;
4811 struct udphdr *uh = pd->hdr.udp;
4812
4813 key.af = pd->af;
4814 key.proto = IPPROTO_UDP;
4815 if (direction == PF_IN) { /* wire side, straight */
4816 PF_ACPY(&key.addr[0], pd->src, key.af);
4817 PF_ACPY(&key.addr[1], pd->dst, key.af);
4818 key.port[0] = uh->uh_sport;
4819 key.port[1] = uh->uh_dport;
4820 } else { /* stack side, reverse */
4821 PF_ACPY(&key.addr[1], pd->src, key.af);
4822 PF_ACPY(&key.addr[0], pd->dst, key.af);
4823 key.port[1] = uh->uh_sport;
4824 key.port[0] = uh->uh_dport;
4825 }
4826
4827 #ifdef __FreeBSD__
4828 STATE_LOOKUP(kif, &key, direction, *state, m, pd->pf_mtag);
4829 #else
4830 STATE_LOOKUP(kif, &key, direction, *state, m);
4831 #endif
4832
4833 if (direction == (*state)->direction) {
4834 src = &(*state)->src;
4835 dst = &(*state)->dst;
4836 } else {
4837 src = &(*state)->dst;
4838 dst = &(*state)->src;
4839 }
4840
4841 /* update states */
4842 if (src->state < PFUDPS_SINGLE)
4843 src->state = PFUDPS_SINGLE;
4844 if (dst->state == PFUDPS_SINGLE)
4845 dst->state = PFUDPS_MULTIPLE;
4846
4847 /* update expire time */
4848 (*state)->expire = time_second;
4849 if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE)
4850 (*state)->timeout = PFTM_UDP_MULTIPLE;
4851 else
4852 (*state)->timeout = PFTM_UDP_SINGLE;
4853
4854 /* translate source/destination address, if necessary */
4855 if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) {
4856 struct pf_state_key *nk = (*state)->key[pd->didx];
4857
4858 if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], pd->af) ||
4859 nk->port[pd->sidx] != uh->uh_sport)
4860 pf_change_ap(m, pd->src, &uh->uh_sport, pd->ip_sum,
4861 &uh->uh_sum, &nk->addr[pd->sidx],
4862 nk->port[pd->sidx], 1, pd->af);
4863
4864 if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], pd->af) ||
4865 nk->port[pd->didx] != uh->uh_dport)
4866 pf_change_ap(m, pd->dst, &uh->uh_dport, pd->ip_sum,
4867 &uh->uh_sum, &nk->addr[pd->didx],
4868 nk->port[pd->didx], 1, pd->af);
4869 #ifdef __FreeBSD__
4870 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
4871 #else
4872 m_copyback(m, off, sizeof(*uh), uh);
4873 #endif
4874 }
4875
4876 return (PF_PASS);
4877 }
4878
4879 int
4880 pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif,
4881 struct mbuf *m, int off, void *h, struct pf_pdesc *pd, u_short *reason)
4882 {
4883 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
4884 #ifdef __FreeBSD__
4885 u_int16_t icmpid = 0, *icmpsum;
4886 #else
4887 u_int16_t icmpid, *icmpsum;
4888 #endif
4889 u_int8_t icmptype;
4890 int state_icmp = 0;
4891 struct pf_state_key_cmp key;
4892
4893 switch (pd->proto) {
4894 #ifdef INET
4895 case IPPROTO_ICMP:
4896 icmptype = pd->hdr.icmp->icmp_type;
4897 icmpid = pd->hdr.icmp->icmp_id;
4898 icmpsum = &pd->hdr.icmp->icmp_cksum;
4899
4900 if (icmptype == ICMP_UNREACH ||
4901 icmptype == ICMP_SOURCEQUENCH ||
4902 icmptype == ICMP_REDIRECT ||
4903 icmptype == ICMP_TIMXCEED ||
4904 icmptype == ICMP_PARAMPROB)
4905 state_icmp++;
4906 break;
4907 #endif /* INET */
4908 #ifdef INET6
4909 case IPPROTO_ICMPV6:
4910 icmptype = pd->hdr.icmp6->icmp6_type;
4911 icmpid = pd->hdr.icmp6->icmp6_id;
4912 icmpsum = &pd->hdr.icmp6->icmp6_cksum;
4913
4914 if (icmptype == ICMP6_DST_UNREACH ||
4915 icmptype == ICMP6_PACKET_TOO_BIG ||
4916 icmptype == ICMP6_TIME_EXCEEDED ||
4917 icmptype == ICMP6_PARAM_PROB)
4918 state_icmp++;
4919 break;
4920 #endif /* INET6 */
4921 }
4922
4923 if (!state_icmp) {
4924
4925 /*
4926 * ICMP query/reply message not related to a TCP/UDP packet.
4927 * Search for an ICMP state.
4928 */
4929 key.af = pd->af;
4930 key.proto = pd->proto;
4931 key.port[0] = key.port[1] = icmpid;
4932 if (direction == PF_IN) { /* wire side, straight */
4933 PF_ACPY(&key.addr[0], pd->src, key.af);
4934 PF_ACPY(&key.addr[1], pd->dst, key.af);
4935 } else { /* stack side, reverse */
4936 PF_ACPY(&key.addr[1], pd->src, key.af);
4937 PF_ACPY(&key.addr[0], pd->dst, key.af);
4938 }
4939
4940 #ifdef __FreeBSD__
4941 STATE_LOOKUP(kif, &key, direction, *state, m, pd->pf_mtag);
4942 #else
4943 STATE_LOOKUP(kif, &key, direction, *state, m);
4944 #endif
4945
4946 (*state)->expire = time_second;
4947 (*state)->timeout = PFTM_ICMP_ERROR_REPLY;
4948
4949 /* translate source/destination address, if necessary */
4950 if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) {
4951 struct pf_state_key *nk = (*state)->key[pd->didx];
4952
4953 switch (pd->af) {
4954 #ifdef INET
4955 case AF_INET:
4956 if (PF_ANEQ(pd->src,
4957 &nk->addr[pd->sidx], AF_INET))
4958 pf_change_a(&saddr->v4.s_addr,
4959 pd->ip_sum,
4960 nk->addr[pd->sidx].v4.s_addr, 0);
4961
4962 if (PF_ANEQ(pd->dst, &nk->addr[pd->didx],
4963 AF_INET))
4964 pf_change_a(&daddr->v4.s_addr,
4965 pd->ip_sum,
4966 nk->addr[pd->didx].v4.s_addr, 0);
4967
4968 if (nk->port[0] !=
4969 pd->hdr.icmp->icmp_id) {
4970 pd->hdr.icmp->icmp_cksum =
4971 pf_cksum_fixup(
4972 pd->hdr.icmp->icmp_cksum, icmpid,
4973 nk->port[pd->sidx], 0);
4974 pd->hdr.icmp->icmp_id =
4975 nk->port[pd->sidx];
4976 }
4977
4978 m_copyback(m, off, ICMP_MINLEN,
4979 #ifdef __FreeBSD__
4980 (caddr_t)
4981 #endif
4982 pd->hdr.icmp);
4983 break;
4984 #endif /* INET */
4985 #ifdef INET6
4986 case AF_INET6:
4987 if (PF_ANEQ(pd->src,
4988 &nk->addr[pd->sidx], AF_INET6))
4989 pf_change_a6(saddr,
4990 &pd->hdr.icmp6->icmp6_cksum,
4991 &nk->addr[pd->sidx], 0);
4992
4993 if (PF_ANEQ(pd->dst,
4994 &nk->addr[pd->didx], AF_INET6))
4995 pf_change_a6(daddr,
4996 &pd->hdr.icmp6->icmp6_cksum,
4997 &nk->addr[pd->didx], 0);
4998
4999 m_copyback(m, off,
5000 sizeof(struct icmp6_hdr),
5001 #ifdef __FreeBSD__
5002 (caddr_t)
5003 #endif
5004 pd->hdr.icmp6);
5005 break;
5006 #endif /* INET6 */
5007 }
5008 }
5009 return (PF_PASS);
5010
5011 } else {
5012 /*
5013 * ICMP error message in response to a TCP/UDP packet.
5014 * Extract the inner TCP/UDP header and search for that state.
5015 */
5016
5017 struct pf_pdesc pd2;
5018 #ifdef __FreeBSD__
5019 bzero(&pd2, sizeof pd2);
5020 #endif
5021 #ifdef INET
5022 struct ip h2;
5023 #endif /* INET */
5024 #ifdef INET6
5025 struct ip6_hdr h2_6;
5026 int terminal = 0;
5027 #endif /* INET6 */
5028 #ifdef __FreeBSD__
5029 int ipoff2 = 0;
5030 int off2 = 0;
5031 #else
5032 int ipoff2;
5033 int off2;
5034 #endif
5035
5036 pd2.af = pd->af;
5037 /* Payload packet is from the opposite direction. */
5038 pd2.sidx = (direction == PF_IN) ? 1 : 0;
5039 pd2.didx = (direction == PF_IN) ? 0 : 1;
5040 switch (pd->af) {
5041 #ifdef INET
5042 case AF_INET:
5043 /* offset of h2 in mbuf chain */
5044 ipoff2 = off + ICMP_MINLEN;
5045
5046 if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2),
5047 NULL, reason, pd2.af)) {
5048 DPFPRINTF(PF_DEBUG_MISC,
5049 ("pf: ICMP error message too short "
5050 "(ip)\n"));
5051 return (PF_DROP);
5052 }
5053 /*
5054 * ICMP error messages don't refer to non-first
5055 * fragments
5056 */
5057 if (h2.ip_off & htons(IP_OFFMASK)) {
5058 REASON_SET(reason, PFRES_FRAG);
5059 return (PF_DROP);
5060 }
5061
5062 /* offset of protocol header that follows h2 */
5063 off2 = ipoff2 + (h2.ip_hl << 2);
5064
5065 pd2.proto = h2.ip_p;
5066 pd2.src = (struct pf_addr *)&h2.ip_src;
5067 pd2.dst = (struct pf_addr *)&h2.ip_dst;
5068 pd2.ip_sum = &h2.ip_sum;
5069 break;
5070 #endif /* INET */
5071 #ifdef INET6
5072 case AF_INET6:
5073 ipoff2 = off + sizeof(struct icmp6_hdr);
5074
5075 if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6),
5076 NULL, reason, pd2.af)) {
5077 DPFPRINTF(PF_DEBUG_MISC,
5078 ("pf: ICMP error message too short "
5079 "(ip6)\n"));
5080 return (PF_DROP);
5081 }
5082 pd2.proto = h2_6.ip6_nxt;
5083 pd2.src = (struct pf_addr *)&h2_6.ip6_src;
5084 pd2.dst = (struct pf_addr *)&h2_6.ip6_dst;
5085 pd2.ip_sum = NULL;
5086 off2 = ipoff2 + sizeof(h2_6);
5087 do {
5088 switch (pd2.proto) {
5089 case IPPROTO_FRAGMENT:
5090 /*
5091 * ICMPv6 error messages for
5092 * non-first fragments
5093 */
5094 REASON_SET(reason, PFRES_FRAG);
5095 return (PF_DROP);
5096 case IPPROTO_AH:
5097 case IPPROTO_HOPOPTS:
5098 case IPPROTO_ROUTING:
5099 case IPPROTO_DSTOPTS: {
5100 /* get next header and header length */
5101 struct ip6_ext opt6;
5102
5103 if (!pf_pull_hdr(m, off2, &opt6,
5104 sizeof(opt6), NULL, reason,
5105 pd2.af)) {
5106 DPFPRINTF(PF_DEBUG_MISC,
5107 ("pf: ICMPv6 short opt\n"));
5108 return (PF_DROP);
5109 }
5110 if (pd2.proto == IPPROTO_AH)
5111 off2 += (opt6.ip6e_len + 2) * 4;
5112 else
5113 off2 += (opt6.ip6e_len + 1) * 8;
5114 pd2.proto = opt6.ip6e_nxt;
5115 /* goto the next header */
5116 break;
5117 }
5118 default:
5119 terminal++;
5120 break;
5121 }
5122 } while (!terminal);
5123 break;
5124 #endif /* INET6 */
5125 }
5126
5127 switch (pd2.proto) {
5128 case IPPROTO_TCP: {
5129 struct tcphdr th;
5130 u_int32_t seq;
5131 struct pf_state_peer *src, *dst;
5132 u_int8_t dws;
5133 int copyback = 0;
5134
5135 /*
5136 * Only the first 8 bytes of the TCP header can be
5137 * expected. Don't access any TCP header fields after
5138 * th_seq, an ackskew test is not possible.
5139 */
5140 if (!pf_pull_hdr(m, off2, &th, 8, NULL, reason,
5141 pd2.af)) {
5142 DPFPRINTF(PF_DEBUG_MISC,
5143 ("pf: ICMP error message too short "
5144 "(tcp)\n"));
5145 return (PF_DROP);
5146 }
5147
5148 key.af = pd2.af;
5149 key.proto = IPPROTO_TCP;
5150 PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
5151 PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
5152 key.port[pd2.sidx] = th.th_sport;
5153 key.port[pd2.didx] = th.th_dport;
5154
5155 #ifdef __FreeBSD__
5156 STATE_LOOKUP(kif, &key, direction, *state, m, pd->pf_mtag);
5157 #else
5158 STATE_LOOKUP(kif, &key, direction, *state, m);
5159 #endif
5160
5161 if (direction == (*state)->direction) {
5162 src = &(*state)->dst;
5163 dst = &(*state)->src;
5164 } else {
5165 src = &(*state)->src;
5166 dst = &(*state)->dst;
5167 }
5168
5169 if (src->wscale && dst->wscale)
5170 dws = dst->wscale & PF_WSCALE_MASK;
5171 else
5172 dws = 0;
5173
5174 /* Demodulate sequence number */
5175 seq = ntohl(th.th_seq) - src->seqdiff;
5176 if (src->seqdiff) {
5177 pf_change_a(&th.th_seq, icmpsum,
5178 htonl(seq), 0);
5179 copyback = 1;
5180 }
5181
5182 if (!((*state)->state_flags & PFSTATE_SLOPPY) &&
5183 (!SEQ_GEQ(src->seqhi, seq) ||
5184 !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)))) {
5185 #ifdef __FreeBSD__
5186 if (V_pf_status.debug >= PF_DEBUG_MISC) {
5187 #else
5188 if (pf_status.debug >= PF_DEBUG_MISC) {
5189 #endif
5190 printf("pf: BAD ICMP %d:%d ",
5191 icmptype, pd->hdr.icmp->icmp_code);
5192 pf_print_host(pd->src, 0, pd->af);
5193 printf(" -> ");
5194 pf_print_host(pd->dst, 0, pd->af);
5195 printf(" state: ");
5196 pf_print_state(*state);
5197 printf(" seq=%u\n", seq);
5198 }
5199 REASON_SET(reason, PFRES_BADSTATE);
5200 return (PF_DROP);
5201 } else {
5202 #ifdef __FreeBSD__
5203 if (V_pf_status.debug >= PF_DEBUG_MISC) {
5204 #else
5205 if (pf_status.debug >= PF_DEBUG_MISC) {
5206 #endif
5207 printf("pf: OK ICMP %d:%d ",
5208 icmptype, pd->hdr.icmp->icmp_code);
5209 pf_print_host(pd->src, 0, pd->af);
5210 printf(" -> ");
5211 pf_print_host(pd->dst, 0, pd->af);
5212 printf(" state: ");
5213 pf_print_state(*state);
5214 printf(" seq=%u\n", seq);
5215 }
5216 }
5217
5218 /* translate source/destination address, if necessary */
5219 if ((*state)->key[PF_SK_WIRE] !=
5220 (*state)->key[PF_SK_STACK]) {
5221 struct pf_state_key *nk =
5222 (*state)->key[pd->didx];
5223
5224 if (PF_ANEQ(pd2.src,
5225 &nk->addr[pd2.sidx], pd2.af) ||
5226 nk->port[pd2.sidx] != th.th_sport)
5227 pf_change_icmp(pd2.src, &th.th_sport,
5228 daddr, &nk->addr[pd2.sidx],
5229 nk->port[pd2.sidx], NULL,
5230 pd2.ip_sum, icmpsum,
5231 pd->ip_sum, 0, pd2.af);
5232
5233 if (PF_ANEQ(pd2.dst,
5234 &nk->addr[pd2.didx], pd2.af) ||
5235 nk->port[pd2.didx] != th.th_dport)
5236 pf_change_icmp(pd2.dst, &th.th_dport,
5237 NULL, /* XXX Inbound NAT? */
5238 &nk->addr[pd2.didx],
5239 nk->port[pd2.didx], NULL,
5240 pd2.ip_sum, icmpsum,
5241 pd->ip_sum, 0, pd2.af);
5242 copyback = 1;
5243 }
5244
5245 if (copyback) {
5246 switch (pd2.af) {
5247 #ifdef INET
5248 case AF_INET:
5249 m_copyback(m, off, ICMP_MINLEN,
5250 #ifdef __FreeBSD__
5251 (caddr_t)
5252 #endif
5253 pd->hdr.icmp);
5254 m_copyback(m, ipoff2, sizeof(h2),
5255 #ifdef __FreeBSD__
5256 (caddr_t)
5257 #endif
5258 &h2);
5259 break;
5260 #endif /* INET */
5261 #ifdef INET6
5262 case AF_INET6:
5263 m_copyback(m, off,
5264 sizeof(struct icmp6_hdr),
5265 #ifdef __FreeBSD__
5266 (caddr_t)
5267 #endif
5268 pd->hdr.icmp6);
5269 m_copyback(m, ipoff2, sizeof(h2_6),
5270 #ifdef __FreeBSD__
5271 (caddr_t)
5272 #endif
5273 &h2_6);
5274 break;
5275 #endif /* INET6 */
5276 }
5277 #ifdef __FreeBSD__
5278 m_copyback(m, off2, 8, (caddr_t)&th);
5279 #else
5280 m_copyback(m, off2, 8, &th);
5281 #endif
5282 }
5283
5284 return (PF_PASS);
5285 break;
5286 }
5287 case IPPROTO_UDP: {
5288 struct udphdr uh;
5289
5290 if (!pf_pull_hdr(m, off2, &uh, sizeof(uh),
5291 NULL, reason, pd2.af)) {
5292 DPFPRINTF(PF_DEBUG_MISC,
5293 ("pf: ICMP error message too short "
5294 "(udp)\n"));
5295 return (PF_DROP);
5296 }
5297
5298 key.af = pd2.af;
5299 key.proto = IPPROTO_UDP;
5300 PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
5301 PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
5302 key.port[pd2.sidx] = uh.uh_sport;
5303 key.port[pd2.didx] = uh.uh_dport;
5304
5305 #ifdef __FreeBSD__
5306 STATE_LOOKUP(kif, &key, direction, *state, m, pd->pf_mtag);
5307 #else
5308 STATE_LOOKUP(kif, &key, direction, *state, m);
5309 #endif
5310
5311 /* translate source/destination address, if necessary */
5312 if ((*state)->key[PF_SK_WIRE] !=
5313 (*state)->key[PF_SK_STACK]) {
5314 struct pf_state_key *nk =
5315 (*state)->key[pd->didx];
5316
5317 if (PF_ANEQ(pd2.src,
5318 &nk->addr[pd2.sidx], pd2.af) ||
5319 nk->port[pd2.sidx] != uh.uh_sport)
5320 pf_change_icmp(pd2.src, &uh.uh_sport,
5321 daddr, &nk->addr[pd2.sidx],
5322 nk->port[pd2.sidx], &uh.uh_sum,
5323 pd2.ip_sum, icmpsum,
5324 pd->ip_sum, 1, pd2.af);
5325
5326 if (PF_ANEQ(pd2.dst,
5327 &nk->addr[pd2.didx], pd2.af) ||
5328 nk->port[pd2.didx] != uh.uh_dport)
5329 pf_change_icmp(pd2.dst, &uh.uh_dport,
5330 NULL, /* XXX Inbound NAT? */
5331 &nk->addr[pd2.didx],
5332 nk->port[pd2.didx], &uh.uh_sum,
5333 pd2.ip_sum, icmpsum,
5334 pd->ip_sum, 1, pd2.af);
5335
5336 switch (pd2.af) {
5337 #ifdef INET
5338 case AF_INET:
5339 m_copyback(m, off, ICMP_MINLEN,
5340 #ifdef __FreeBSD__
5341 (caddr_t)
5342 #endif
5343 pd->hdr.icmp);
5344 #ifdef __FreeBSD__
5345 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
5346 #else
5347 m_copyback(m, ipoff2, sizeof(h2), &h2);
5348 #endif
5349 break;
5350 #endif /* INET */
5351 #ifdef INET6
5352 case AF_INET6:
5353 m_copyback(m, off,
5354 sizeof(struct icmp6_hdr),
5355 #ifdef __FreeBSD__
5356 (caddr_t)
5357 #endif
5358 pd->hdr.icmp6);
5359 m_copyback(m, ipoff2, sizeof(h2_6),
5360 #ifdef __FreeBSD__
5361 (caddr_t)
5362 #endif
5363 &h2_6);
5364 break;
5365 #endif /* INET6 */
5366 }
5367 #ifdef __FreeBSD__
5368 m_copyback(m, off2, sizeof(uh), (caddr_t)&uh);
5369 #else
5370 m_copyback(m, off2, sizeof(uh), &uh);
5371 #endif
5372 }
5373 return (PF_PASS);
5374 break;
5375 }
5376 #ifdef INET
5377 case IPPROTO_ICMP: {
5378 struct icmp iih;
5379
5380 if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN,
5381 NULL, reason, pd2.af)) {
5382 DPFPRINTF(PF_DEBUG_MISC,
5383 ("pf: ICMP error message too short i"
5384 "(icmp)\n"));
5385 return (PF_DROP);
5386 }
5387
5388 key.af = pd2.af;
5389 key.proto = IPPROTO_ICMP;
5390 PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
5391 PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
5392 key.port[0] = key.port[1] = iih.icmp_id;
5393
5394 #ifdef __FreeBSD__
5395 STATE_LOOKUP(kif, &key, direction, *state, m, pd->pf_mtag);
5396 #else
5397 STATE_LOOKUP(kif, &key, direction, *state, m);
5398 #endif
5399
5400 /* translate source/destination address, if necessary */
5401 if ((*state)->key[PF_SK_WIRE] !=
5402 (*state)->key[PF_SK_STACK]) {
5403 struct pf_state_key *nk =
5404 (*state)->key[pd->didx];
5405
5406 if (PF_ANEQ(pd2.src,
5407 &nk->addr[pd2.sidx], pd2.af) ||
5408 nk->port[pd2.sidx] != iih.icmp_id)
5409 pf_change_icmp(pd2.src, &iih.icmp_id,
5410 daddr, &nk->addr[pd2.sidx],
5411 nk->port[pd2.sidx], NULL,
5412 pd2.ip_sum, icmpsum,
5413 pd->ip_sum, 0, AF_INET);
5414
5415 if (PF_ANEQ(pd2.dst,
5416 &nk->addr[pd2.didx], pd2.af) ||
5417 nk->port[pd2.didx] != iih.icmp_id)
5418 pf_change_icmp(pd2.dst, &iih.icmp_id,
5419 NULL, /* XXX Inbound NAT? */
5420 &nk->addr[pd2.didx],
5421 nk->port[pd2.didx], NULL,
5422 pd2.ip_sum, icmpsum,
5423 pd->ip_sum, 0, AF_INET);
5424
5425 #ifdef __FreeBSD__
5426 m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
5427 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
5428 m_copyback(m, off2, ICMP_MINLEN, (caddr_t)&iih);
5429 #else
5430 m_copyback(m, off, ICMP_MINLEN, pd->hdr.icmp);
5431 m_copyback(m, ipoff2, sizeof(h2), &h2);
5432 m_copyback(m, off2, ICMP_MINLEN, &iih);
5433 #endif
5434 }
5435 return (PF_PASS);
5436 break;
5437 }
5438 #endif /* INET */
5439 #ifdef INET6
5440 case IPPROTO_ICMPV6: {
5441 struct icmp6_hdr iih;
5442
5443 if (!pf_pull_hdr(m, off2, &iih,
5444 sizeof(struct icmp6_hdr), NULL, reason, pd2.af)) {
5445 DPFPRINTF(PF_DEBUG_MISC,
5446 ("pf: ICMP error message too short "
5447 "(icmp6)\n"));
5448 return (PF_DROP);
5449 }
5450
5451 key.af = pd2.af;
5452 key.proto = IPPROTO_ICMPV6;
5453 PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
5454 PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
5455 key.port[0] = key.port[1] = iih.icmp6_id;
5456
5457 #ifdef __FreeBSD__
5458 STATE_LOOKUP(kif, &key, direction, *state, m, pd->pf_mtag);
5459 #else
5460 STATE_LOOKUP(kif, &key, direction, *state, m);
5461 #endif
5462
5463 /* translate source/destination address, if necessary */
5464 if ((*state)->key[PF_SK_WIRE] !=
5465 (*state)->key[PF_SK_STACK]) {
5466 struct pf_state_key *nk =
5467 (*state)->key[pd->didx];
5468
5469 if (PF_ANEQ(pd2.src,
5470 &nk->addr[pd2.sidx], pd2.af) ||
5471 nk->port[pd2.sidx] != iih.icmp6_id)
5472 pf_change_icmp(pd2.src, &iih.icmp6_id,
5473 daddr, &nk->addr[pd2.sidx],
5474 nk->port[pd2.sidx], NULL,
5475 pd2.ip_sum, icmpsum,
5476 pd->ip_sum, 0, AF_INET6);
5477
5478 if (PF_ANEQ(pd2.dst,
5479 &nk->addr[pd2.didx], pd2.af) ||
5480 nk->port[pd2.didx] != iih.icmp6_id)
5481 pf_change_icmp(pd2.dst, &iih.icmp6_id,
5482 NULL, /* XXX Inbound NAT? */
5483 &nk->addr[pd2.didx],
5484 nk->port[pd2.didx], NULL,
5485 pd2.ip_sum, icmpsum,
5486 pd->ip_sum, 0, AF_INET6);
5487
5488 #ifdef __FreeBSD__
5489 m_copyback(m, off, sizeof(struct icmp6_hdr),
5490 (caddr_t)pd->hdr.icmp6);
5491 m_copyback(m, ipoff2, sizeof(h2_6), (caddr_t)&h2_6);
5492 m_copyback(m, off2, sizeof(struct icmp6_hdr),
5493 (caddr_t)&iih);
5494 #else
5495 m_copyback(m, off, sizeof(struct icmp6_hdr),
5496 pd->hdr.icmp6);
5497 m_copyback(m, ipoff2, sizeof(h2_6), &h2_6);
5498 m_copyback(m, off2, sizeof(struct icmp6_hdr),
5499 &iih);
5500 #endif
5501 }
5502 return (PF_PASS);
5503 break;
5504 }
5505 #endif /* INET6 */
5506 default: {
5507 key.af = pd2.af;
5508 key.proto = pd2.proto;
5509 PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
5510 PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
5511 key.port[0] = key.port[1] = 0;
5512
5513 #ifdef __FreeBSD__
5514 STATE_LOOKUP(kif, &key, direction, *state, m, pd->pf_mtag);
5515 #else
5516 STATE_LOOKUP(kif, &key, direction, *state, m);
5517 #endif
5518
5519 /* translate source/destination address, if necessary */
5520 if ((*state)->key[PF_SK_WIRE] !=
5521 (*state)->key[PF_SK_STACK]) {
5522 struct pf_state_key *nk =
5523 (*state)->key[pd->didx];
5524
5525 if (PF_ANEQ(pd2.src,
5526 &nk->addr[pd2.sidx], pd2.af))
5527 pf_change_icmp(pd2.src, NULL, daddr,
5528 &nk->addr[pd2.sidx], 0, NULL,
5529 pd2.ip_sum, icmpsum,
5530 pd->ip_sum, 0, pd2.af);
5531
5532 if (PF_ANEQ(pd2.dst,
5533 &nk->addr[pd2.didx], pd2.af))
5534 pf_change_icmp(pd2.src, NULL,
5535 NULL, /* XXX Inbound NAT? */
5536 &nk->addr[pd2.didx], 0, NULL,
5537 pd2.ip_sum, icmpsum,
5538 pd->ip_sum, 0, pd2.af);
5539
5540 switch (pd2.af) {
5541 #ifdef INET
5542 case AF_INET:
5543 #ifdef __FreeBSD__
5544 m_copyback(m, off, ICMP_MINLEN,
5545 (caddr_t)pd->hdr.icmp);
5546 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
5547 #else
5548 m_copyback(m, off, ICMP_MINLEN,
5549 pd->hdr.icmp);
5550 m_copyback(m, ipoff2, sizeof(h2), &h2);
5551 #endif
5552 break;
5553 #endif /* INET */
5554 #ifdef INET6
5555 case AF_INET6:
5556 m_copyback(m, off,
5557 sizeof(struct icmp6_hdr),
5558 #ifdef __FreeBSD__
5559 (caddr_t)
5560 #endif
5561 pd->hdr.icmp6);
5562 m_copyback(m, ipoff2, sizeof(h2_6),
5563 #ifdef __FreeBSD__
5564 (caddr_t)
5565 #endif
5566 &h2_6);
5567 break;
5568 #endif /* INET6 */
5569 }
5570 }
5571 return (PF_PASS);
5572 break;
5573 }
5574 }
5575 }
5576 }
5577
5578 int
5579 pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif,
5580 struct mbuf *m, struct pf_pdesc *pd)
5581 {
5582 struct pf_state_peer *src, *dst;
5583 struct pf_state_key_cmp key;
5584
5585 key.af = pd->af;
5586 key.proto = pd->proto;
5587 if (direction == PF_IN) {
5588 PF_ACPY(&key.addr[0], pd->src, key.af);
5589 PF_ACPY(&key.addr[1], pd->dst, key.af);
5590 key.port[0] = key.port[1] = 0;
5591 } else {
5592 PF_ACPY(&key.addr[1], pd->src, key.af);
5593 PF_ACPY(&key.addr[0], pd->dst, key.af);
5594 key.port[1] = key.port[0] = 0;
5595 }
5596
5597 #ifdef __FreeBSD__
5598 STATE_LOOKUP(kif, &key, direction, *state, m, pd->pf_mtag);
5599 #else
5600 STATE_LOOKUP(kif, &key, direction, *state, m);
5601 #endif
5602
5603 if (direction == (*state)->direction) {
5604 src = &(*state)->src;
5605 dst = &(*state)->dst;
5606 } else {
5607 src = &(*state)->dst;
5608 dst = &(*state)->src;
5609 }
5610
5611 /* update states */
5612 if (src->state < PFOTHERS_SINGLE)
5613 src->state = PFOTHERS_SINGLE;
5614 if (dst->state == PFOTHERS_SINGLE)
5615 dst->state = PFOTHERS_MULTIPLE;
5616
5617 /* update expire time */
5618 (*state)->expire = time_second;
5619 if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE)
5620 (*state)->timeout = PFTM_OTHER_MULTIPLE;
5621 else
5622 (*state)->timeout = PFTM_OTHER_SINGLE;
5623
5624 /* translate source/destination address, if necessary */
5625 if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) {
5626 struct pf_state_key *nk = (*state)->key[pd->didx];
5627
5628 #ifdef __FreeBSD__
5629 KASSERT(nk, ("%s: nk is null", __FUNCTION__));
5630 KASSERT(pd, ("%s: pd is null", __FUNCTION__));
5631 KASSERT(pd->src, ("%s: pd->src is null", __FUNCTION__));
5632 KASSERT(pd->dst, ("%s: pd->dst is null", __FUNCTION__));
5633 #else
5634 KASSERT(nk);
5635 KASSERT(pd);
5636 KASSERT(pd->src);
5637 KASSERT(pd->dst);
5638 #endif
5639 switch (pd->af) {
5640 #ifdef INET
5641 case AF_INET:
5642 if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], AF_INET))
5643 pf_change_a(&pd->src->v4.s_addr,
5644 pd->ip_sum,
5645 nk->addr[pd->sidx].v4.s_addr,
5646 0);
5647
5648
5649 if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], AF_INET))
5650 pf_change_a(&pd->dst->v4.s_addr,
5651 pd->ip_sum,
5652 nk->addr[pd->didx].v4.s_addr,
5653 0);
5654
5655 break;
5656 #endif /* INET */
5657 #ifdef INET6
5658 case AF_INET6:
5659 if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], AF_INET))
5660 PF_ACPY(pd->src, &nk->addr[pd->sidx], pd->af);
5661
5662 if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], AF_INET))
5663 PF_ACPY(pd->dst, &nk->addr[pd->didx], pd->af);
5664 #endif /* INET6 */
5665 }
5666 }
5667 return (PF_PASS);
5668 }
5669
5670 /*
5671 * ipoff and off are measured from the start of the mbuf chain.
5672 * h must be at "ipoff" on the mbuf chain.
5673 */
5674 void *
5675 pf_pull_hdr(struct mbuf *m, int off, void *p, int len,
5676 u_short *actionp, u_short *reasonp, sa_family_t af)
5677 {
5678 switch (af) {
5679 #ifdef INET
5680 case AF_INET: {
5681 struct ip *h = mtod(m, struct ip *);
5682 u_int16_t fragoff = (ntohs(h->ip_off) & IP_OFFMASK) << 3;
5683
5684 if (fragoff) {
5685 if (fragoff >= len)
5686 ACTION_SET(actionp, PF_PASS);
5687 else {
5688 ACTION_SET(actionp, PF_DROP);
5689 REASON_SET(reasonp, PFRES_FRAG);
5690 }
5691 return (NULL);
5692 }
5693 if (m->m_pkthdr.len < off + len ||
5694 ntohs(h->ip_len) < off + len) {
5695 ACTION_SET(actionp, PF_DROP);
5696 REASON_SET(reasonp, PFRES_SHORT);
5697 return (NULL);
5698 }
5699 break;
5700 }
5701 #endif /* INET */
5702 #ifdef INET6
5703 case AF_INET6: {
5704 struct ip6_hdr *h = mtod(m, struct ip6_hdr *);
5705
5706 if (m->m_pkthdr.len < off + len ||
5707 (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) <
5708 (unsigned)(off + len)) {
5709 ACTION_SET(actionp, PF_DROP);
5710 REASON_SET(reasonp, PFRES_SHORT);
5711 return (NULL);
5712 }
5713 break;
5714 }
5715 #endif /* INET6 */
5716 }
5717 m_copydata(m, off, len, p);
5718 return (p);
5719 }
5720
5721 int
5722 pf_routable(struct pf_addr *addr, sa_family_t af, struct pfi_kif *kif,
5723 int rtableid)
5724 {
5725 #ifdef __FreeBSD__
5726 #ifdef RADIX_MPATH
5727 struct radix_node_head *rnh;
5728 #endif
5729 #endif
5730 struct sockaddr_in *dst;
5731 int ret = 1;
5732 int check_mpath;
5733 #ifndef __FreeBSD__
5734 extern int ipmultipath;
5735 #endif
5736 #ifdef INET6
5737 #ifndef __FreeBSD__
5738 extern int ip6_multipath;
5739 #endif
5740 struct sockaddr_in6 *dst6;
5741 struct route_in6 ro;
5742 #else
5743 struct route ro;
5744 #endif
5745 struct radix_node *rn;
5746 struct rtentry *rt;
5747 struct ifnet *ifp;
5748
5749 check_mpath = 0;
5750 #ifdef __FreeBSD__
5751 #ifdef RADIX_MPATH
5752 /* XXX: stick to table 0 for now */
5753 rnh = rt_tables_get_rnh(0, af);
5754 if (rnh != NULL && rn_mpath_capable(rnh))
5755 check_mpath = 1;
5756 #endif
5757 #endif
5758 bzero(&ro, sizeof(ro));
5759 switch (af) {
5760 case AF_INET:
5761 dst = satosin(&ro.ro_dst);
5762 dst->sin_family = AF_INET;
5763 dst->sin_len = sizeof(*dst);
5764 dst->sin_addr = addr->v4;
5765 #ifndef __FreeBSD__
5766 if (ipmultipath)
5767 check_mpath = 1;
5768 #endif
5769 break;
5770 #ifdef INET6
5771 case AF_INET6:
5772 /*
5773 * Skip check for addresses with embedded interface scope,
5774 * as they would always match anyway.
5775 */
5776 if (IN6_IS_SCOPE_EMBED(&addr->v6))
5777 goto out;
5778 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
5779 dst6->sin6_family = AF_INET6;
5780 dst6->sin6_len = sizeof(*dst6);
5781 dst6->sin6_addr = addr->v6;
5782 #ifndef __FreeBSD__
5783 if (ip6_multipath)
5784 check_mpath = 1;
5785 #endif
5786 break;
5787 #endif /* INET6 */
5788 default:
5789 return (0);
5790 }
5791
5792 /* Skip checks for ipsec interfaces */
5793 if (kif != NULL && kif->pfik_ifp->if_type == IFT_ENC)
5794 goto out;
5795
5796 #ifdef __FreeBSD__
5797 switch (af) {
5798 #ifdef INET6
5799 case AF_INET6:
5800 in6_rtalloc_ign(&ro, 0, rtableid);
5801 break;
5802 #endif
5803 #ifdef INET
5804 case AF_INET:
5805 in_rtalloc_ign((struct route *)&ro, 0, rtableid);
5806 break;
5807 #endif
5808 default:
5809 rtalloc_ign((struct route *)&ro, 0); /* No/default FIB. */
5810 break;
5811 }
5812 #else /* ! __FreeBSD__ */
5813 rtalloc_noclone((struct route *)&ro, NO_CLONING);
5814 #endif
5815
5816 if (ro.ro_rt != NULL) {
5817 /* No interface given, this is a no-route check */
5818 if (kif == NULL)
5819 goto out;
5820
5821 if (kif->pfik_ifp == NULL) {
5822 ret = 0;
5823 goto out;
5824 }
5825
5826 /* Perform uRPF check if passed input interface */
5827 ret = 0;
5828 rn = (struct radix_node *)ro.ro_rt;
5829 do {
5830 rt = (struct rtentry *)rn;
5831 #ifndef __FreeBSD__ /* CARPDEV */
5832 if (rt->rt_ifp->if_type == IFT_CARP)
5833 ifp = rt->rt_ifp->if_carpdev;
5834 else
5835 #endif
5836 ifp = rt->rt_ifp;
5837
5838 if (kif->pfik_ifp == ifp)
5839 ret = 1;
5840 #ifdef __FreeBSD__
5841 #ifdef RADIX_MPATH
5842 rn = rn_mpath_next(rn);
5843 #endif
5844 #else
5845 rn = rn_mpath_next(rn, 0);
5846 #endif
5847 } while (check_mpath == 1 && rn != NULL && ret == 0);
5848 } else
5849 ret = 0;
5850 out:
5851 if (ro.ro_rt != NULL)
5852 RTFREE(ro.ro_rt);
5853 return (ret);
5854 }
5855
5856 int
5857 pf_rtlabel_match(struct pf_addr *addr, sa_family_t af, struct pf_addr_wrap *aw,
5858 int rtableid)
5859 {
5860 struct sockaddr_in *dst;
5861 #ifdef INET6
5862 struct sockaddr_in6 *dst6;
5863 struct route_in6 ro;
5864 #else
5865 struct route ro;
5866 #endif
5867 int ret = 0;
5868
5869 bzero(&ro, sizeof(ro));
5870 switch (af) {
5871 case AF_INET:
5872 dst = satosin(&ro.ro_dst);
5873 dst->sin_family = AF_INET;
5874 dst->sin_len = sizeof(*dst);
5875 dst->sin_addr = addr->v4;
5876 break;
5877 #ifdef INET6
5878 case AF_INET6:
5879 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
5880 dst6->sin6_family = AF_INET6;
5881 dst6->sin6_len = sizeof(*dst6);
5882 dst6->sin6_addr = addr->v6;
5883 break;
5884 #endif /* INET6 */
5885 default:
5886 return (0);
5887 }
5888
5889 #ifdef __FreeBSD__
5890 switch (af) {
5891 #ifdef INET6
5892 case AF_INET6:
5893 in6_rtalloc_ign(&ro, 0, rtableid);
5894 break;
5895 #endif
5896 #ifdef INET
5897 case AF_INET:
5898 in_rtalloc_ign((struct route *)&ro, 0, rtableid);
5899 break;
5900 #endif
5901 default:
5902 rtalloc_ign((struct route *)&ro, 0);
5903 break;
5904 }
5905 #else /* ! __FreeBSD__ */
5906 rtalloc_noclone((struct route *)&ro, NO_CLONING);
5907 #endif
5908
5909 if (ro.ro_rt != NULL) {
5910 #ifdef __FreeBSD__
5911 /* XXX_IMPORT: later */
5912 #else
5913 if (ro.ro_rt->rt_labelid == aw->v.rtlabel)
5914 ret = 1;
5915 #endif
5916 RTFREE(ro.ro_rt);
5917 }
5918
5919 return (ret);
5920 }
5921
5922 #ifdef INET
5923 void
5924 pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5925 struct pf_state *s, struct pf_pdesc *pd)
5926 {
5927 struct mbuf *m0, *m1;
5928 struct route iproute;
5929 struct route *ro = NULL;
5930 struct sockaddr_in *dst;
5931 struct ip *ip;
5932 struct ifnet *ifp = NULL;
5933 struct pf_addr naddr;
5934 struct pf_src_node *sn = NULL;
5935 int error = 0;
5936 #ifdef __FreeBSD__
5937 int sw_csum;
5938 #endif
5939 #ifdef IPSEC
5940 struct m_tag *mtag;
5941 #endif /* IPSEC */
5942
5943 if (m == NULL || *m == NULL || r == NULL ||
5944 (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5945 panic("pf_route: invalid parameters");
5946
5947 #ifdef __FreeBSD__
5948 if (pd->pf_mtag->routed++ > 3) {
5949 #else
5950 if ((*m)->m_pkthdr.pf.routed++ > 3) {
5951 #endif
5952 m0 = *m;
5953 *m = NULL;
5954 goto bad;
5955 }
5956
5957 if (r->rt == PF_DUPTO) {
5958 #ifdef __FreeBSD__
5959 if ((m0 = m_dup(*m, M_DONTWAIT)) == NULL)
5960 #else
5961 if ((m0 = m_copym2(*m, 0, M_COPYALL, M_NOWAIT)) == NULL)
5962 #endif
5963 return;
5964 } else {
5965 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5966 return;
5967 m0 = *m;
5968 }
5969
5970 if (m0->m_len < sizeof(struct ip)) {
5971 DPFPRINTF(PF_DEBUG_URGENT,
5972 ("pf_route: m0->m_len < sizeof(struct ip)\n"));
5973 goto bad;
5974 }
5975
5976 ip = mtod(m0, struct ip *);
5977
5978 ro = &iproute;
5979 bzero((caddr_t)ro, sizeof(*ro));
5980 dst = satosin(&ro->ro_dst);
5981 dst->sin_family = AF_INET;
5982 dst->sin_len = sizeof(*dst);
5983 dst->sin_addr = ip->ip_dst;
5984
5985 if (r->rt == PF_FASTROUTE) {
5986 #ifdef __FreeBSD__
5987 in_rtalloc_ign(ro, 0, M_GETFIB(m0));
5988 #else
5989 rtalloc(ro);
5990 #endif
5991 if (ro->ro_rt == 0) {
5992 #ifdef __FreeBSD__
5993 KMOD_IPSTAT_INC(ips_noroute);
5994 #else
5995 ipstat.ips_noroute++;
5996 #endif
5997 goto bad;
5998 }
5999
6000 ifp = ro->ro_rt->rt_ifp;
6001 ro->ro_rt->rt_use++;
6002
6003 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
6004 dst = satosin(ro->ro_rt->rt_gateway);
6005 } else {
6006 if (TAILQ_EMPTY(&r->rpool.list)) {
6007 DPFPRINTF(PF_DEBUG_URGENT,
6008 ("pf_route: TAILQ_EMPTY(&r->rpool.list)\n"));
6009 goto bad;
6010 }
6011 if (s == NULL) {
6012 pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src,
6013 &naddr, NULL, &sn);
6014 if (!PF_AZERO(&naddr, AF_INET))
6015 dst->sin_addr.s_addr = naddr.v4.s_addr;
6016 ifp = r->rpool.cur->kif ?
6017 r->rpool.cur->kif->pfik_ifp : NULL;
6018 } else {
6019 if (!PF_AZERO(&s->rt_addr, AF_INET))
6020 dst->sin_addr.s_addr =
6021 s->rt_addr.v4.s_addr;
6022 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
6023 }
6024 }
6025 if (ifp == NULL)
6026 goto bad;
6027
6028 if (oifp != ifp) {
6029 #ifdef __FreeBSD__
6030 PF_UNLOCK();
6031 if (pf_test(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) {
6032 PF_LOCK();
6033 goto bad;
6034 } else if (m0 == NULL) {
6035 PF_LOCK();
6036 goto done;
6037 }
6038 PF_LOCK();
6039 #else
6040 if (pf_test(PF_OUT, ifp, &m0, NULL) != PF_PASS)
6041 goto bad;
6042 else if (m0 == NULL)
6043 goto done;
6044 #endif
6045 if (m0->m_len < sizeof(struct ip)) {
6046 DPFPRINTF(PF_DEBUG_URGENT,
6047 ("pf_route: m0->m_len < sizeof(struct ip)\n"));
6048 goto bad;
6049 }
6050 ip = mtod(m0, struct ip *);
6051 }
6052
6053 #ifdef __FreeBSD__
6054 /* Copied from FreeBSD 5.1-CURRENT ip_output. */
6055 m0->m_pkthdr.csum_flags |= CSUM_IP;
6056 sw_csum = m0->m_pkthdr.csum_flags & ~ifp->if_hwassist;
6057 if (sw_csum & CSUM_DELAY_DATA) {
6058 /*
6059 * XXX: in_delayed_cksum assumes HBO for ip->ip_len (at least)
6060 */
6061 NTOHS(ip->ip_len);
6062 NTOHS(ip->ip_off); /* XXX: needed? */
6063 in_delayed_cksum(m0);
6064 HTONS(ip->ip_len);
6065 HTONS(ip->ip_off);
6066 sw_csum &= ~CSUM_DELAY_DATA;
6067 }
6068 m0->m_pkthdr.csum_flags &= ifp->if_hwassist;
6069
6070 if (ntohs(ip->ip_len) <= ifp->if_mtu ||
6071 (ifp->if_hwassist & CSUM_FRAGMENT &&
6072 ((ip->ip_off & htons(IP_DF)) == 0))) {
6073 /*
6074 * ip->ip_len = htons(ip->ip_len);
6075 * ip->ip_off = htons(ip->ip_off);
6076 */
6077 ip->ip_sum = 0;
6078 if (sw_csum & CSUM_DELAY_IP) {
6079 /* From KAME */
6080 if (ip->ip_v == IPVERSION &&
6081 (ip->ip_hl << 2) == sizeof(*ip)) {
6082 ip->ip_sum = in_cksum_hdr(ip);
6083 } else {
6084 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
6085 }
6086 }
6087 PF_UNLOCK();
6088 error = (*ifp->if_output)(ifp, m0, sintosa(dst), ro);
6089 PF_LOCK();
6090 goto done;
6091 }
6092 #else
6093 /* Copied from ip_output. */
6094 #ifdef IPSEC
6095 /*
6096 * If deferred crypto processing is needed, check that the
6097 * interface supports it.
6098 */
6099 if ((mtag = m_tag_find(m0, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL))
6100 != NULL && (ifp->if_capabilities & IFCAP_IPSEC) == 0) {
6101 /* Notify IPsec to do its own crypto. */
6102 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1));
6103 goto bad;
6104 }
6105 #endif /* IPSEC */
6106
6107 /* Catch routing changes wrt. hardware checksumming for TCP or UDP. */
6108 if (m0->m_pkthdr.csum_flags & M_TCPV4_CSUM_OUT) {
6109 if (!(ifp->if_capabilities & IFCAP_CSUM_TCPv4) ||
6110 ifp->if_bridge != NULL) {
6111 in_delayed_cksum(m0);
6112 m0->m_pkthdr.csum_flags &= ~M_TCPV4_CSUM_OUT; /* Clr */
6113 }
6114 } else if (m0->m_pkthdr.csum_flags & M_UDPV4_CSUM_OUT) {
6115 if (!(ifp->if_capabilities & IFCAP_CSUM_UDPv4) ||
6116 ifp->if_bridge != NULL) {
6117 in_delayed_cksum(m0);
6118 m0->m_pkthdr.csum_flags &= ~M_UDPV4_CSUM_OUT; /* Clr */
6119 }
6120 }
6121
6122 if (ntohs(ip->ip_len) <= ifp->if_mtu) {
6123 ip->ip_sum = 0;
6124 if ((ifp->if_capabilities & IFCAP_CSUM_IPv4) &&
6125 ifp->if_bridge == NULL) {
6126 m0->m_pkthdr.csum_flags |= M_IPV4_CSUM_OUT;
6127 #ifdef __FreeBSD__
6128 KMOD_IPSTAT_INC(ips_outhwcsum);
6129 #else
6130 ipstat.ips_outhwcsum++;
6131 #endif
6132 } else
6133 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
6134 /* Update relevant hardware checksum stats for TCP/UDP */
6135 if (m0->m_pkthdr.csum_flags & M_TCPV4_CSUM_OUT)
6136 KMOD_TCPSTAT_INC(tcps_outhwcsum);
6137 else if (m0->m_pkthdr.csum_flags & M_UDPV4_CSUM_OUT)
6138 KMOD_UDPSTAT_INC(udps_outhwcsum);
6139 error = (*ifp->if_output)(ifp, m0, sintosa(dst), NULL);
6140 goto done;
6141 }
6142 #endif
6143
6144 /*
6145 * Too large for interface; fragment if possible.
6146 * Must be able to put at least 8 bytes per fragment.
6147 */
6148 if (ip->ip_off & htons(IP_DF)) {
6149 #ifdef __FreeBSD__
6150 KMOD_IPSTAT_INC(ips_cantfrag);
6151 #else
6152 ipstat.ips_cantfrag++;
6153 #endif
6154 if (r->rt != PF_DUPTO) {
6155 #ifdef __FreeBSD__
6156 /* icmp_error() expects host byte ordering */
6157 NTOHS(ip->ip_len);
6158 NTOHS(ip->ip_off);
6159 PF_UNLOCK();
6160 icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0,
6161 ifp->if_mtu);
6162 PF_LOCK();
6163 #else
6164 icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0,
6165 ifp->if_mtu);
6166 #endif
6167 goto done;
6168 } else
6169 goto bad;
6170 }
6171
6172 m1 = m0;
6173 #ifdef __FreeBSD__
6174 /*
6175 * XXX: is cheaper + less error prone than own function
6176 */
6177 NTOHS(ip->ip_len);
6178 NTOHS(ip->ip_off);
6179 error = ip_fragment(ip, &m0, ifp->if_mtu, ifp->if_hwassist, sw_csum);
6180 #else
6181 error = ip_fragment(m0, ifp, ifp->if_mtu);
6182 #endif
6183 if (error) {
6184 #ifndef __FreeBSD__ /* ip_fragment does not do m_freem() on FreeBSD */
6185 m0 = NULL;
6186 #endif
6187 goto bad;
6188 }
6189
6190 for (m0 = m1; m0; m0 = m1) {
6191 m1 = m0->m_nextpkt;
6192 m0->m_nextpkt = 0;
6193 #ifdef __FreeBSD__
6194 if (error == 0) {
6195 PF_UNLOCK();
6196 error = (*ifp->if_output)(ifp, m0, sintosa(dst),
6197 NULL);
6198 PF_LOCK();
6199 } else
6200 #else
6201 if (error == 0)
6202 error = (*ifp->if_output)(ifp, m0, sintosa(dst),
6203 NULL);
6204 else
6205 #endif
6206 m_freem(m0);
6207 }
6208
6209 if (error == 0)
6210 #ifdef __FreeBSD__
6211 KMOD_IPSTAT_INC(ips_fragmented);
6212 #else
6213 ipstat.ips_fragmented++;
6214 #endif
6215
6216 done:
6217 if (r->rt != PF_DUPTO)
6218 *m = NULL;
6219 if (ro == &iproute && ro->ro_rt)
6220 RTFREE(ro->ro_rt);
6221 return;
6222
6223 bad:
6224 m_freem(m0);
6225 goto done;
6226 }
6227 #endif /* INET */
6228
6229 #ifdef INET6
6230 void
6231 pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
6232 struct pf_state *s, struct pf_pdesc *pd)
6233 {
6234 struct mbuf *m0;
6235 struct route_in6 ip6route;
6236 struct route_in6 *ro;
6237 struct sockaddr_in6 *dst;
6238 struct ip6_hdr *ip6;
6239 struct ifnet *ifp = NULL;
6240 struct pf_addr naddr;
6241 struct pf_src_node *sn = NULL;
6242
6243 if (m == NULL || *m == NULL || r == NULL ||
6244 (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
6245 panic("pf_route6: invalid parameters");
6246
6247 #ifdef __FreeBSD__
6248 if (pd->pf_mtag->routed++ > 3) {
6249 #else
6250 if ((*m)->m_pkthdr.pf.routed++ > 3) {
6251 #endif
6252 m0 = *m;
6253 *m = NULL;
6254 goto bad;
6255 }
6256
6257 if (r->rt == PF_DUPTO) {
6258 #ifdef __FreeBSD__
6259 if ((m0 = m_dup(*m, M_DONTWAIT)) == NULL)
6260 #else
6261 if ((m0 = m_copym2(*m, 0, M_COPYALL, M_NOWAIT)) == NULL)
6262 #endif
6263 return;
6264 } else {
6265 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
6266 return;
6267 m0 = *m;
6268 }
6269
6270 if (m0->m_len < sizeof(struct ip6_hdr)) {
6271 DPFPRINTF(PF_DEBUG_URGENT,
6272 ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n"));
6273 goto bad;
6274 }
6275 ip6 = mtod(m0, struct ip6_hdr *);
6276
6277 ro = &ip6route;
6278 bzero((caddr_t)ro, sizeof(*ro));
6279 dst = (struct sockaddr_in6 *)&ro->ro_dst;
6280 dst->sin6_family = AF_INET6;
6281 dst->sin6_len = sizeof(*dst);
6282 dst->sin6_addr = ip6->ip6_dst;
6283
6284 /* Cheat. XXX why only in the v6 case??? */
6285 if (r->rt == PF_FASTROUTE) {
6286 #ifdef __FreeBSD__
6287 m0->m_flags |= M_SKIP_FIREWALL;
6288 PF_UNLOCK();
6289 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
6290 #else
6291 m0->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
6292 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
6293 #endif
6294 return;
6295 }
6296
6297 if (TAILQ_EMPTY(&r->rpool.list)) {
6298 DPFPRINTF(PF_DEBUG_URGENT,
6299 ("pf_route6: TAILQ_EMPTY(&r->rpool.list)\n"));
6300 goto bad;
6301 }
6302 if (s == NULL) {
6303 pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src,
6304 &naddr, NULL, &sn);
6305 if (!PF_AZERO(&naddr, AF_INET6))
6306 PF_ACPY((struct pf_addr *)&dst->sin6_addr,
6307 &naddr, AF_INET6);
6308 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL;
6309 } else {
6310 if (!PF_AZERO(&s->rt_addr, AF_INET6))
6311 PF_ACPY((struct pf_addr *)&dst->sin6_addr,
6312 &s->rt_addr, AF_INET6);
6313 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
6314 }
6315 if (ifp == NULL)
6316 goto bad;
6317
6318 if (oifp != ifp) {
6319 #ifdef __FreeBSD__
6320 PF_UNLOCK();
6321 if (pf_test6(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) {
6322 PF_LOCK();
6323 goto bad;
6324 } else if (m0 == NULL) {
6325 PF_LOCK();
6326 goto done;
6327 }
6328 PF_LOCK();
6329 #else
6330 if (pf_test6(PF_OUT, ifp, &m0, NULL) != PF_PASS)
6331 goto bad;
6332 else if (m0 == NULL)
6333 goto done;
6334 #endif
6335 if (m0->m_len < sizeof(struct ip6_hdr)) {
6336 DPFPRINTF(PF_DEBUG_URGENT,
6337 ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n"));
6338 goto bad;
6339 }
6340 ip6 = mtod(m0, struct ip6_hdr *);
6341 }
6342
6343 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6 &
6344 ~ifp->if_hwassist) {
6345 uint32_t plen = m0->m_pkthdr.len - sizeof(*ip6);
6346 in6_delayed_cksum(m0, plen, sizeof(struct ip6_hdr));
6347 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
6348 }
6349
6350 /*
6351 * If the packet is too large for the outgoing interface,
6352 * send back an icmp6 error.
6353 */
6354 if (IN6_IS_SCOPE_EMBED(&dst->sin6_addr))
6355 dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
6356 if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) {
6357 #ifdef __FreeBSD__
6358 PF_UNLOCK();
6359 #endif
6360 nd6_output(ifp, ifp, m0, dst, NULL);
6361 #ifdef __FreeBSD__
6362 PF_LOCK();
6363 #endif
6364 } else {
6365 in6_ifstat_inc(ifp, ifs6_in_toobig);
6366 #ifdef __FreeBSD__
6367 if (r->rt != PF_DUPTO) {
6368 PF_UNLOCK();
6369 icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
6370 PF_LOCK();
6371 } else
6372 #else
6373 if (r->rt != PF_DUPTO)
6374 icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
6375 else
6376 #endif
6377 goto bad;
6378 }
6379
6380 done:
6381 if (r->rt != PF_DUPTO)
6382 *m = NULL;
6383 return;
6384
6385 bad:
6386 m_freem(m0);
6387 goto done;
6388 }
6389 #endif /* INET6 */
6390
6391 #ifdef __FreeBSD__
6392 /*
6393 * FreeBSD supports cksum offloads for the following drivers.
6394 * em(4), fxp(4), ixgb(4), lge(4), ndis(4), nge(4), re(4),
6395 * ti(4), txp(4), xl(4)
6396 *
6397 * CSUM_DATA_VALID | CSUM_PSEUDO_HDR :
6398 * network driver performed cksum including pseudo header, need to verify
6399 * csum_data
6400 * CSUM_DATA_VALID :
6401 * network driver performed cksum, needs to additional pseudo header
6402 * cksum computation with partial csum_data(i.e. lack of H/W support for
6403 * pseudo header, for instance hme(4), sk(4) and possibly gem(4))
6404 *
6405 * After validating the cksum of packet, set both flag CSUM_DATA_VALID and
6406 * CSUM_PSEUDO_HDR in order to avoid recomputation of the cksum in upper
6407 * TCP/UDP layer.
6408 * Also, set csum_data to 0xffff to force cksum validation.
6409 */
6410 int
6411 pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p, sa_family_t af)
6412 {
6413 u_int16_t sum = 0;
6414 int hw_assist = 0;
6415 struct ip *ip;
6416
6417 if (off < sizeof(struct ip) || len < sizeof(struct udphdr))
6418 return (1);
6419 if (m->m_pkthdr.len < off + len)
6420 return (1);
6421
6422 switch (p) {
6423 case IPPROTO_TCP:
6424 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
6425 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) {
6426 sum = m->m_pkthdr.csum_data;
6427 } else {
6428 ip = mtod(m, struct ip *);
6429 sum = in_pseudo(ip->ip_src.s_addr,
6430 ip->ip_dst.s_addr, htonl((u_short)len +
6431 m->m_pkthdr.csum_data + IPPROTO_TCP));
6432 }
6433 sum ^= 0xffff;
6434 ++hw_assist;
6435 }
6436 break;
6437 case IPPROTO_UDP:
6438 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
6439 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) {
6440 sum = m->m_pkthdr.csum_data;
6441 } else {
6442 ip = mtod(m, struct ip *);
6443 sum = in_pseudo(ip->ip_src.s_addr,
6444 ip->ip_dst.s_addr, htonl((u_short)len +
6445 m->m_pkthdr.csum_data + IPPROTO_UDP));
6446 }
6447 sum ^= 0xffff;
6448 ++hw_assist;
6449 }
6450 break;
6451 case IPPROTO_ICMP:
6452 #ifdef INET6
6453 case IPPROTO_ICMPV6:
6454 #endif /* INET6 */
6455 break;
6456 default:
6457 return (1);
6458 }
6459
6460 if (!hw_assist) {
6461 switch (af) {
6462 case AF_INET:
6463 if (p == IPPROTO_ICMP) {
6464 if (m->m_len < off)
6465 return (1);
6466 m->m_data += off;
6467 m->m_len -= off;
6468 sum = in_cksum(m, len);
6469 m->m_data -= off;
6470 m->m_len += off;
6471 } else {
6472 if (m->m_len < sizeof(struct ip))
6473 return (1);
6474 sum = in4_cksum(m, p, off, len);
6475 }
6476 break;
6477 #ifdef INET6
6478 case AF_INET6:
6479 if (m->m_len < sizeof(struct ip6_hdr))
6480 return (1);
6481 sum = in6_cksum(m, p, off, len);
6482 break;
6483 #endif /* INET6 */
6484 default:
6485 return (1);
6486 }
6487 }
6488 if (sum) {
6489 switch (p) {
6490 case IPPROTO_TCP:
6491 {
6492 KMOD_TCPSTAT_INC(tcps_rcvbadsum);
6493 break;
6494 }
6495 case IPPROTO_UDP:
6496 {
6497 KMOD_UDPSTAT_INC(udps_badsum);
6498 break;
6499 }
6500 #ifdef INET
6501 case IPPROTO_ICMP:
6502 {
6503 KMOD_ICMPSTAT_INC(icps_checksum);
6504 break;
6505 }
6506 #endif
6507 #ifdef INET6
6508 case IPPROTO_ICMPV6:
6509 {
6510 KMOD_ICMP6STAT_INC(icp6s_checksum);
6511 break;
6512 }
6513 #endif /* INET6 */
6514 }
6515 return (1);
6516 } else {
6517 if (p == IPPROTO_TCP || p == IPPROTO_UDP) {
6518 m->m_pkthdr.csum_flags |=
6519 (CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
6520 m->m_pkthdr.csum_data = 0xffff;
6521 }
6522 }
6523 return (0);
6524 }
6525 #else /* !__FreeBSD__ */
6526
6527 /*
6528 * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag
6529 * off is the offset where the protocol header starts
6530 * len is the total length of protocol header plus payload
6531 * returns 0 when the checksum is valid, otherwise returns 1.
6532 */
6533 int
6534 pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p,
6535 sa_family_t af)
6536 {
6537 u_int16_t flag_ok, flag_bad;
6538 u_int16_t sum;
6539
6540 switch (p) {
6541 case IPPROTO_TCP:
6542 flag_ok = M_TCP_CSUM_IN_OK;
6543 flag_bad = M_TCP_CSUM_IN_BAD;
6544 break;
6545 case IPPROTO_UDP:
6546 flag_ok = M_UDP_CSUM_IN_OK;
6547 flag_bad = M_UDP_CSUM_IN_BAD;
6548 break;
6549 case IPPROTO_ICMP:
6550 #ifdef INET6
6551 case IPPROTO_ICMPV6:
6552 #endif /* INET6 */
6553 flag_ok = flag_bad = 0;
6554 break;
6555 default:
6556 return (1);
6557 }
6558 if (m->m_pkthdr.csum_flags & flag_ok)
6559 return (0);
6560 if (m->m_pkthdr.csum_flags & flag_bad)
6561 return (1);
6562 if (off < sizeof(struct ip) || len < sizeof(struct udphdr))
6563 return (1);
6564 if (m->m_pkthdr.len < off + len)
6565 return (1);
6566 switch (af) {
6567 #ifdef INET
6568 case AF_INET:
6569 if (p == IPPROTO_ICMP) {
6570 if (m->m_len < off)
6571 return (1);
6572 m->m_data += off;
6573 m->m_len -= off;
6574 sum = in_cksum(m, len);
6575 m->m_data -= off;
6576 m->m_len += off;
6577 } else {
6578 if (m->m_len < sizeof(struct ip))
6579 return (1);
6580 sum = in4_cksum(m, p, off, len);
6581 }
6582 break;
6583 #endif /* INET */
6584 #ifdef INET6
6585 case AF_INET6:
6586 if (m->m_len < sizeof(struct ip6_hdr))
6587 return (1);
6588 sum = in6_cksum(m, p, off, len);
6589 break;
6590 #endif /* INET6 */
6591 default:
6592 return (1);
6593 }
6594 if (sum) {
6595 m->m_pkthdr.csum_flags |= flag_bad;
6596 switch (p) {
6597 case IPPROTO_TCP:
6598 KMOD_TCPSTAT_INC(tcps_rcvbadsum);
6599 break;
6600 case IPPROTO_UDP:
6601 KMOD_UDPSTAT_INC(udps_badsum);
6602 break;
6603 #ifdef INET
6604 case IPPROTO_ICMP:
6605 KMOD_ICMPSTAT_INC(icps_checksum);
6606 break;
6607 #endif
6608 #ifdef INET6
6609 case IPPROTO_ICMPV6:
6610 KMOD_ICMP6STAT_INC(icp6s_checksum);
6611 break;
6612 #endif /* INET6 */
6613 }
6614 return (1);
6615 }
6616 m->m_pkthdr.csum_flags |= flag_ok;
6617 return (0);
6618 }
6619 #endif
6620
6621 #ifndef __FreeBSD__
6622 struct pf_divert *
6623 pf_find_divert(struct mbuf *m)
6624 {
6625 struct m_tag *mtag;
6626
6627 if ((mtag = m_tag_find(m, PACKET_TAG_PF_DIVERT, NULL)) == NULL)
6628 return (NULL);
6629
6630 return ((struct pf_divert *)(mtag + 1));
6631 }
6632
6633 struct pf_divert *
6634 pf_get_divert(struct mbuf *m)
6635 {
6636 struct m_tag *mtag;
6637
6638 if ((mtag = m_tag_find(m, PACKET_TAG_PF_DIVERT, NULL)) == NULL) {
6639 mtag = m_tag_get(PACKET_TAG_PF_DIVERT, sizeof(struct pf_divert),
6640 M_NOWAIT);
6641 if (mtag == NULL)
6642 return (NULL);
6643 bzero(mtag + 1, sizeof(struct pf_divert));
6644 m_tag_prepend(m, mtag);
6645 }
6646
6647 return ((struct pf_divert *)(mtag + 1));
6648 }
6649 #endif
6650
6651 #ifdef INET
6652 int
6653 #ifdef __FreeBSD__
6654 pf_test(int dir, struct ifnet *ifp, struct mbuf **m0,
6655 struct ether_header *eh, struct inpcb *inp)
6656 #else
6657 pf_test(int dir, struct ifnet *ifp, struct mbuf **m0,
6658 struct ether_header *eh)
6659 #endif
6660 {
6661 struct pfi_kif *kif;
6662 u_short action, reason = 0, log = 0;
6663 struct mbuf *m = *m0;
6664 #ifdef __FreeBSD__
6665 struct ip *h = NULL;
6666 struct m_tag *ipfwtag;
6667 struct pf_rule *a = NULL, *r = &V_pf_default_rule, *tr, *nr;
6668 #else
6669 struct ip *h;
6670 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr;
6671 #endif
6672 struct pf_state *s = NULL;
6673 struct pf_ruleset *ruleset = NULL;
6674 struct pf_pdesc pd;
6675 int off, dirndx, pqid = 0;
6676
6677 #ifdef __FreeBSD__
6678 PF_LOCK();
6679 if (!V_pf_status.running)
6680 {
6681 PF_UNLOCK();
6682 return (PF_PASS);
6683 }
6684 #else
6685 if (!pf_status.running)
6686 return (PF_PASS);
6687 #endif
6688
6689 memset(&pd, 0, sizeof(pd));
6690 #ifdef __FreeBSD__
6691 if ((pd.pf_mtag = pf_get_mtag(m)) == NULL) {
6692 PF_UNLOCK();
6693 DPFPRINTF(PF_DEBUG_URGENT,
6694 ("pf_test: pf_get_mtag returned NULL\n"));
6695 return (PF_DROP);
6696 }
6697 #endif
6698 #ifndef __FreeBSD__
6699 if (ifp->if_type == IFT_CARP && ifp->if_carpdev)
6700 kif = (struct pfi_kif *)ifp->if_carpdev->if_pf_kif;
6701 else
6702 #endif
6703 kif = (struct pfi_kif *)ifp->if_pf_kif;
6704
6705 if (kif == NULL) {
6706 #ifdef __FreeBSD__
6707 PF_UNLOCK();
6708 #endif
6709 DPFPRINTF(PF_DEBUG_URGENT,
6710 ("pf_test: kif == NULL, if_xname %s\n", ifp->if_xname));
6711 return (PF_DROP);
6712 }
6713 if (kif->pfik_flags & PFI_IFLAG_SKIP)
6714 #ifdef __FreeBSD__
6715 {
6716 PF_UNLOCK();
6717 #endif
6718 return (PF_PASS);
6719 #ifdef __FreeBSD__
6720 }
6721 #endif
6722
6723 #ifdef __FreeBSD__
6724 M_ASSERTPKTHDR(m);
6725 #else
6726 #ifdef DIAGNOSTIC
6727 if ((m->m_flags & M_PKTHDR) == 0)
6728 panic("non-M_PKTHDR is passed to pf_test");
6729 #endif /* DIAGNOSTIC */
6730 #endif
6731
6732 if (m->m_pkthdr.len < (int)sizeof(*h)) {
6733 action = PF_DROP;
6734 REASON_SET(&reason, PFRES_SHORT);
6735 log = 1;
6736 goto done;
6737 }
6738
6739 #ifdef __FreeBSD__
6740 if (m->m_flags & M_SKIP_FIREWALL) {
6741 PF_UNLOCK();
6742 return (PF_PASS);
6743 }
6744 #else
6745 if (m->m_pkthdr.pf.flags & PF_TAG_GENERATED)
6746 return (PF_PASS);
6747 #endif
6748
6749 #ifdef __FreeBSD__
6750 if (ip_divert_ptr != NULL &&
6751 ((ipfwtag = m_tag_locate(m, MTAG_IPFW_RULE, 0, NULL)) != NULL)) {
6752 struct ipfw_rule_ref *rr = (struct ipfw_rule_ref *)(ipfwtag+1);
6753 if (rr->info & IPFW_IS_DIVERT && rr->rulenum == 0) {
6754 pd.pf_mtag->flags |= PF_PACKET_LOOPED;
6755 m_tag_delete(m, ipfwtag);
6756 }
6757 if (pd.pf_mtag->flags & PF_FASTFWD_OURS_PRESENT) {
6758 m->m_flags |= M_FASTFWD_OURS;
6759 pd.pf_mtag->flags &= ~PF_FASTFWD_OURS_PRESENT;
6760 }
6761 } else
6762 #endif
6763 /* We do IP header normalization and packet reassembly here */
6764 if (pf_normalize_ip(m0, dir, kif, &reason, &pd) != PF_PASS) {
6765 action = PF_DROP;
6766 goto done;
6767 }
6768 m = *m0; /* pf_normalize messes with m0 */
6769 h = mtod(m, struct ip *);
6770
6771 off = h->ip_hl << 2;
6772 if (off < (int)sizeof(*h)) {
6773 action = PF_DROP;
6774 REASON_SET(&reason, PFRES_SHORT);
6775 log = 1;
6776 goto done;
6777 }
6778
6779 pd.src = (struct pf_addr *)&h->ip_src;
6780 pd.dst = (struct pf_addr *)&h->ip_dst;
6781 pd.sport = pd.dport = NULL;
6782 pd.ip_sum = &h->ip_sum;
6783 pd.proto_sum = NULL;
6784 pd.proto = h->ip_p;
6785 pd.dir = dir;
6786 pd.sidx = (dir == PF_IN) ? 0 : 1;
6787 pd.didx = (dir == PF_IN) ? 1 : 0;
6788 pd.af = AF_INET;
6789 pd.tos = h->ip_tos;
6790 pd.tot_len = ntohs(h->ip_len);
6791 pd.eh = eh;
6792
6793 /* handle fragments that didn't get reassembled by normalization */
6794 if (h->ip_off & htons(IP_MF | IP_OFFMASK)) {
6795 action = pf_test_fragment(&r, dir, kif, m, h,
6796 &pd, &a, &ruleset);
6797 goto done;
6798 }
6799
6800 switch (h->ip_p) {
6801
6802 case IPPROTO_TCP: {
6803 struct tcphdr th;
6804
6805 pd.hdr.tcp = &th;
6806 if (!pf_pull_hdr(m, off, &th, sizeof(th),
6807 &action, &reason, AF_INET)) {
6808 log = action != PF_PASS;
6809 goto done;
6810 }
6811 pd.p_len = pd.tot_len - off - (th.th_off << 2);
6812 if ((th.th_flags & TH_ACK) && pd.p_len == 0)
6813 pqid = 1;
6814 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
6815 if (action == PF_DROP)
6816 goto done;
6817 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
6818 &reason);
6819 if (action == PF_PASS) {
6820 #if NPFSYNC > 0
6821 #ifdef __FreeBSD__
6822 if (pfsync_update_state_ptr != NULL)
6823 pfsync_update_state_ptr(s);
6824 #else
6825 pfsync_update_state(s);
6826 #endif
6827 #endif /* NPFSYNC */
6828 r = s->rule.ptr;
6829 a = s->anchor.ptr;
6830 log = s->log;
6831 } else if (s == NULL)
6832 #ifdef __FreeBSD__
6833 action = pf_test_rule(&r, &s, dir, kif,
6834 m, off, h, &pd, &a, &ruleset, NULL, inp);
6835 #else
6836 action = pf_test_rule(&r, &s, dir, kif,
6837 m, off, h, &pd, &a, &ruleset, &ipintrq);
6838 #endif
6839 break;
6840 }
6841
6842 case IPPROTO_UDP: {
6843 struct udphdr uh;
6844
6845 pd.hdr.udp = &uh;
6846 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
6847 &action, &reason, AF_INET)) {
6848 log = action != PF_PASS;
6849 goto done;
6850 }
6851 if (uh.uh_dport == 0 ||
6852 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
6853 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
6854 action = PF_DROP;
6855 REASON_SET(&reason, PFRES_SHORT);
6856 goto done;
6857 }
6858 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
6859 if (action == PF_PASS) {
6860 #if NPFSYNC > 0
6861 #ifdef __FreeBSD__
6862 if (pfsync_update_state_ptr != NULL)
6863 pfsync_update_state_ptr(s);
6864 #else
6865 pfsync_update_state(s);
6866 #endif
6867 #endif /* NPFSYNC */
6868 r = s->rule.ptr;
6869 a = s->anchor.ptr;
6870 log = s->log;
6871 } else if (s == NULL)
6872 #ifdef __FreeBSD__
6873 action = pf_test_rule(&r, &s, dir, kif,
6874 m, off, h, &pd, &a, &ruleset, NULL, inp);
6875 #else
6876 action = pf_test_rule(&r, &s, dir, kif,
6877 m, off, h, &pd, &a, &ruleset, &ipintrq);
6878 #endif
6879 break;
6880 }
6881
6882 case IPPROTO_ICMP: {
6883 struct icmp ih;
6884
6885 pd.hdr.icmp = &ih;
6886 if (!pf_pull_hdr(m, off, &ih, ICMP_MINLEN,
6887 &action, &reason, AF_INET)) {
6888 log = action != PF_PASS;
6889 goto done;
6890 }
6891 action = pf_test_state_icmp(&s, dir, kif, m, off, h, &pd,
6892 &reason);
6893 if (action == PF_PASS) {
6894 #if NPFSYNC > 0
6895 #ifdef __FreeBSD__
6896 if (pfsync_update_state_ptr != NULL)
6897 pfsync_update_state_ptr(s);
6898 #else
6899 pfsync_update_state(s);
6900 #endif
6901 #endif /* NPFSYNC */
6902 r = s->rule.ptr;
6903 a = s->anchor.ptr;
6904 log = s->log;
6905 } else if (s == NULL)
6906 #ifdef __FreeBSD__
6907 action = pf_test_rule(&r, &s, dir, kif,
6908 m, off, h, &pd, &a, &ruleset, NULL, inp);
6909 #else
6910 action = pf_test_rule(&r, &s, dir, kif,
6911 m, off, h, &pd, &a, &ruleset, &ipintrq);
6912 #endif
6913 break;
6914 }
6915
6916 #ifdef INET6
6917 case IPPROTO_ICMPV6: {
6918 action = PF_DROP;
6919 DPFPRINTF(PF_DEBUG_MISC,
6920 ("pf: dropping IPv4 packet with ICMPv6 payload\n"));
6921 goto done;
6922 }
6923 #endif
6924
6925 default:
6926 action = pf_test_state_other(&s, dir, kif, m, &pd);
6927 if (action == PF_PASS) {
6928 #if NPFSYNC > 0
6929 #ifdef __FreeBSD__
6930 if (pfsync_update_state_ptr != NULL)
6931 pfsync_update_state_ptr(s);
6932 #else
6933 pfsync_update_state(s);
6934 #endif
6935 #endif /* NPFSYNC */
6936 r = s->rule.ptr;
6937 a = s->anchor.ptr;
6938 log = s->log;
6939 } else if (s == NULL)
6940 #ifdef __FreeBSD__
6941 action = pf_test_rule(&r, &s, dir, kif, m, off, h,
6942 &pd, &a, &ruleset, NULL, inp);
6943 #else
6944 action = pf_test_rule(&r, &s, dir, kif, m, off, h,
6945 &pd, &a, &ruleset, &ipintrq);
6946 #endif
6947 break;
6948 }
6949
6950 done:
6951 if (action == PF_PASS && h->ip_hl > 5 &&
6952 !((s && s->state_flags & PFSTATE_ALLOWOPTS) || r->allow_opts)) {
6953 action = PF_DROP;
6954 REASON_SET(&reason, PFRES_IPOPTIONS);
6955 log = 1;
6956 DPFPRINTF(PF_DEBUG_MISC,
6957 ("pf: dropping packet with ip options\n"));
6958 }
6959
6960 if ((s && s->tag) || r->rtableid >= 0)
6961 #ifdef __FreeBSD__
6962 pf_tag_packet(m, s ? s->tag : 0, r->rtableid, pd.pf_mtag);
6963 #else
6964 pf_tag_packet(m, s ? s->tag : 0, r->rtableid);
6965 #endif
6966
6967 if (dir == PF_IN && s && s->key[PF_SK_STACK])
6968 #ifdef __FreeBSD__
6969 pd.pf_mtag->statekey = s->key[PF_SK_STACK];
6970 #else
6971 m->m_pkthdr.pf.statekey = s->key[PF_SK_STACK];
6972 #endif
6973
6974 #ifdef ALTQ
6975 if (action == PF_PASS && r->qid) {
6976 #ifdef __FreeBSD__
6977 if (pqid || (pd.tos & IPTOS_LOWDELAY))
6978 pd.pf_mtag->qid = r->pqid;
6979 else
6980 pd.pf_mtag->qid = r->qid;
6981 /* add hints for ecn */
6982 pd.pf_mtag->hdr = h;
6983
6984 #else
6985 if (pqid || (pd.tos & IPTOS_LOWDELAY))
6986 m->m_pkthdr.pf.qid = r->pqid;
6987 else
6988 m->m_pkthdr.pf.qid = r->qid;
6989 /* add hints for ecn */
6990 m->m_pkthdr.pf.hdr = h;
6991 #endif
6992 }
6993 #endif /* ALTQ */
6994
6995 /*
6996 * connections redirected to loopback should not match sockets
6997 * bound specifically to loopback due to security implications,
6998 * see tcp_input() and in_pcblookup_listen().
6999 */
7000 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
7001 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
7002 (s->nat_rule.ptr->action == PF_RDR ||
7003 s->nat_rule.ptr->action == PF_BINAT) &&
7004 (ntohl(pd.dst->v4.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET)
7005 #ifdef __FreeBSD__
7006 m->m_flags |= M_SKIP_FIREWALL;
7007 #else
7008 m->m_pkthdr.pf.flags |= PF_TAG_TRANSLATE_LOCALHOST;
7009 #endif
7010
7011 #ifdef __FreeBSD__
7012 if (action == PF_PASS && r->divert.port &&
7013 ip_divert_ptr != NULL && !PACKET_LOOPED()) {
7014
7015 ipfwtag = m_tag_alloc(MTAG_IPFW_RULE, 0,
7016 sizeof(struct ipfw_rule_ref), M_NOWAIT | M_ZERO);
7017 if (ipfwtag != NULL) {
7018 ((struct ipfw_rule_ref *)(ipfwtag+1))->info =
7019 ntohs(r->divert.port);
7020 ((struct ipfw_rule_ref *)(ipfwtag+1))->rulenum = dir;
7021
7022 m_tag_prepend(m, ipfwtag);
7023
7024 PF_UNLOCK();
7025
7026 if (m->m_flags & M_FASTFWD_OURS) {
7027 pd.pf_mtag->flags |= PF_FASTFWD_OURS_PRESENT;
7028 m->m_flags &= ~M_FASTFWD_OURS;
7029 }
7030
7031 ip_divert_ptr(*m0,
7032 dir == PF_IN ? DIR_IN : DIR_OUT);
7033 *m0 = NULL;
7034 return (action);
7035 } else {
7036 /* XXX: ipfw has the same behaviour! */
7037 action = PF_DROP;
7038 REASON_SET(&reason, PFRES_MEMORY);
7039 log = 1;
7040 DPFPRINTF(PF_DEBUG_MISC,
7041 ("pf: failed to allocate divert tag\n"));
7042 }
7043 }
7044 #else
7045 if (dir == PF_IN && action == PF_PASS && r->divert.port) {
7046 struct pf_divert *divert;
7047
7048 if ((divert = pf_get_divert(m))) {
7049 m->m_pkthdr.pf.flags |= PF_TAG_DIVERTED;
7050 divert->port = r->divert.port;
7051 divert->addr.ipv4 = r->divert.addr.v4;
7052 }
7053 }
7054 #endif
7055
7056 if (log) {
7057 struct pf_rule *lr;
7058
7059 if (s != NULL && s->nat_rule.ptr != NULL &&
7060 s->nat_rule.ptr->log & PF_LOG_ALL)
7061 lr = s->nat_rule.ptr;
7062 else
7063 lr = r;
7064 PFLOG_PACKET(kif, h, m, AF_INET, dir, reason, lr, a, ruleset,
7065 &pd);
7066 }
7067
7068 kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
7069 kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++;
7070
7071 if (action == PF_PASS || r->action == PF_DROP) {
7072 dirndx = (dir == PF_OUT);
7073 r->packets[dirndx]++;
7074 r->bytes[dirndx] += pd.tot_len;
7075 if (a != NULL) {
7076 a->packets[dirndx]++;
7077 a->bytes[dirndx] += pd.tot_len;
7078 }
7079 if (s != NULL) {
7080 if (s->nat_rule.ptr != NULL) {
7081 s->nat_rule.ptr->packets[dirndx]++;
7082 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len;
7083 }
7084 if (s->src_node != NULL) {
7085 s->src_node->packets[dirndx]++;
7086 s->src_node->bytes[dirndx] += pd.tot_len;
7087 }
7088 if (s->nat_src_node != NULL) {
7089 s->nat_src_node->packets[dirndx]++;
7090 s->nat_src_node->bytes[dirndx] += pd.tot_len;
7091 }
7092 dirndx = (dir == s->direction) ? 0 : 1;
7093 s->packets[dirndx]++;
7094 s->bytes[dirndx] += pd.tot_len;
7095 }
7096 tr = r;
7097 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
7098 #ifdef __FreeBSD__
7099 if (nr != NULL && r == &V_pf_default_rule)
7100 #else
7101 if (nr != NULL && r == &pf_default_rule)
7102 #endif
7103 tr = nr;
7104 if (tr->src.addr.type == PF_ADDR_TABLE)
7105 pfr_update_stats(tr->src.addr.p.tbl,
7106 (s == NULL) ? pd.src :
7107 &s->key[(s->direction == PF_IN)]->
7108 addr[(s->direction == PF_OUT)],
7109 pd.af, pd.tot_len, dir == PF_OUT,
7110 r->action == PF_PASS, tr->src.neg);
7111 if (tr->dst.addr.type == PF_ADDR_TABLE)
7112 pfr_update_stats(tr->dst.addr.p.tbl,
7113 (s == NULL) ? pd.dst :
7114 &s->key[(s->direction == PF_IN)]->
7115 addr[(s->direction == PF_IN)],
7116 pd.af, pd.tot_len, dir == PF_OUT,
7117 r->action == PF_PASS, tr->dst.neg);
7118 }
7119
7120 switch (action) {
7121 case PF_SYNPROXY_DROP:
7122 m_freem(*m0);
7123 case PF_DEFER:
7124 *m0 = NULL;
7125 action = PF_PASS;
7126 break;
7127 default:
7128 /* pf_route can free the mbuf causing *m0 to become NULL */
7129 if (r->rt)
7130 pf_route(m0, r, dir, kif->pfik_ifp, s, &pd);
7131 break;
7132 }
7133 #ifdef __FreeBSD__
7134 PF_UNLOCK();
7135 #endif
7136 return (action);
7137 }
7138 #endif /* INET */
7139
7140 #ifdef INET6
7141 int
7142 #ifdef __FreeBSD__
7143 pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0,
7144 struct ether_header *eh, struct inpcb *inp)
7145 #else
7146 pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0,
7147 struct ether_header *eh)
7148 #endif
7149 {
7150 struct pfi_kif *kif;
7151 u_short action, reason = 0, log = 0;
7152 struct mbuf *m = *m0, *n = NULL;
7153 #ifdef __FreeBSD__
7154 struct ip6_hdr *h = NULL;
7155 struct pf_rule *a = NULL, *r = &V_pf_default_rule, *tr, *nr;
7156 #else
7157 struct ip6_hdr *h;
7158 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr;
7159 #endif
7160 struct pf_state *s = NULL;
7161 struct pf_ruleset *ruleset = NULL;
7162 struct pf_pdesc pd;
7163 int off, terminal = 0, dirndx, rh_cnt = 0;
7164
7165 #ifdef __FreeBSD__
7166 PF_LOCK();
7167 if (!V_pf_status.running) {
7168 PF_UNLOCK();
7169 return (PF_PASS);
7170 }
7171 #else
7172 if (!pf_status.running)
7173 return (PF_PASS);
7174 #endif
7175
7176 memset(&pd, 0, sizeof(pd));
7177 #ifdef __FreeBSD__
7178 if ((pd.pf_mtag = pf_get_mtag(m)) == NULL) {
7179 PF_UNLOCK();
7180 DPFPRINTF(PF_DEBUG_URGENT,
7181 ("pf_test: pf_get_mtag returned NULL\n"));
7182 return (PF_DROP);
7183 }
7184 #endif
7185 #ifndef __FreeBSD__
7186 if (ifp->if_type == IFT_CARP && ifp->if_carpdev)
7187 kif = (struct pfi_kif *)ifp->if_carpdev->if_pf_kif;
7188 else
7189 #endif
7190 kif = (struct pfi_kif *)ifp->if_pf_kif;
7191
7192 if (kif == NULL) {
7193 #ifdef __FreeBSD__
7194 PF_UNLOCK();
7195 #endif
7196 DPFPRINTF(PF_DEBUG_URGENT,
7197 ("pf_test6: kif == NULL, if_xname %s\n", ifp->if_xname));
7198 return (PF_DROP);
7199 }
7200 if (kif->pfik_flags & PFI_IFLAG_SKIP)
7201 #ifdef __FreeBSD__
7202 {
7203 PF_UNLOCK();
7204 #endif
7205 return (PF_PASS);
7206 #ifdef __FreeBSD__
7207 }
7208 #endif
7209
7210 #ifdef __FreeBSD__
7211 M_ASSERTPKTHDR(m);
7212 #else
7213 #ifdef DIAGNOSTIC
7214 if ((m->m_flags & M_PKTHDR) == 0)
7215 panic("non-M_PKTHDR is passed to pf_test6");
7216 #endif /* DIAGNOSTIC */
7217 #endif
7218
7219 if (m->m_pkthdr.len < (int)sizeof(*h)) {
7220 action = PF_DROP;
7221 REASON_SET(&reason, PFRES_SHORT);
7222 log = 1;
7223 goto done;
7224 }
7225
7226 #ifdef __FreeBSD__
7227 if (pd.pf_mtag->flags & PF_TAG_GENERATED) {
7228 PF_UNLOCK();
7229 #else
7230 if (m->m_pkthdr.pf.flags & PF_TAG_GENERATED)
7231 #endif
7232 return (PF_PASS);
7233 #ifdef __FreeBSD__
7234 }
7235 #endif
7236
7237 /* We do IP header normalization and packet reassembly here */
7238 if (pf_normalize_ip6(m0, dir, kif, &reason, &pd) != PF_PASS) {
7239 action = PF_DROP;
7240 goto done;
7241 }
7242 m = *m0; /* pf_normalize messes with m0 */
7243 h = mtod(m, struct ip6_hdr *);
7244
7245 #if 1
7246 /*
7247 * we do not support jumbogram yet. if we keep going, zero ip6_plen
7248 * will do something bad, so drop the packet for now.
7249 */
7250 if (htons(h->ip6_plen) == 0) {
7251 action = PF_DROP;
7252 REASON_SET(&reason, PFRES_NORM); /*XXX*/
7253 goto done;
7254 }
7255 #endif
7256
7257 pd.src = (struct pf_addr *)&h->ip6_src;
7258 pd.dst = (struct pf_addr *)&h->ip6_dst;
7259 pd.sport = pd.dport = NULL;
7260 pd.ip_sum = NULL;
7261 pd.proto_sum = NULL;
7262 pd.dir = dir;
7263 pd.sidx = (dir == PF_IN) ? 0 : 1;
7264 pd.didx = (dir == PF_IN) ? 1 : 0;
7265 pd.af = AF_INET6;
7266 pd.tos = 0;
7267 pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr);
7268 pd.eh = eh;
7269
7270 off = ((caddr_t)h - m->m_data) + sizeof(struct ip6_hdr);
7271 pd.proto = h->ip6_nxt;
7272 do {
7273 switch (pd.proto) {
7274 case IPPROTO_FRAGMENT:
7275 action = pf_test_fragment(&r, dir, kif, m, h,
7276 &pd, &a, &ruleset);
7277 if (action == PF_DROP)
7278 REASON_SET(&reason, PFRES_FRAG);
7279 goto done;
7280 case IPPROTO_ROUTING: {
7281 struct ip6_rthdr rthdr;
7282
7283 if (rh_cnt++) {
7284 DPFPRINTF(PF_DEBUG_MISC,
7285 ("pf: IPv6 more than one rthdr\n"));
7286 action = PF_DROP;
7287 REASON_SET(&reason, PFRES_IPOPTIONS);
7288 log = 1;
7289 goto done;
7290 }
7291 if (!pf_pull_hdr(m, off, &rthdr, sizeof(rthdr), NULL,
7292 &reason, pd.af)) {
7293 DPFPRINTF(PF_DEBUG_MISC,
7294 ("pf: IPv6 short rthdr\n"));
7295 action = PF_DROP;
7296 REASON_SET(&reason, PFRES_SHORT);
7297 log = 1;
7298 goto done;
7299 }
7300 if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) {
7301 DPFPRINTF(PF_DEBUG_MISC,
7302 ("pf: IPv6 rthdr0\n"));
7303 action = PF_DROP;
7304 REASON_SET(&reason, PFRES_IPOPTIONS);
7305 log = 1;
7306 goto done;
7307 }
7308 /* FALLTHROUGH */
7309 }
7310 case IPPROTO_AH:
7311 case IPPROTO_HOPOPTS:
7312 case IPPROTO_DSTOPTS: {
7313 /* get next header and header length */
7314 struct ip6_ext opt6;
7315
7316 if (!pf_pull_hdr(m, off, &opt6, sizeof(opt6),
7317 NULL, &reason, pd.af)) {
7318 DPFPRINTF(PF_DEBUG_MISC,
7319 ("pf: IPv6 short opt\n"));
7320 action = PF_DROP;
7321 log = 1;
7322 goto done;
7323 }
7324 if (pd.proto == IPPROTO_AH)
7325 off += (opt6.ip6e_len + 2) * 4;
7326 else
7327 off += (opt6.ip6e_len + 1) * 8;
7328 pd.proto = opt6.ip6e_nxt;
7329 /* goto the next header */
7330 break;
7331 }
7332 default:
7333 terminal++;
7334 break;
7335 }
7336 } while (!terminal);
7337
7338 /* if there's no routing header, use unmodified mbuf for checksumming */
7339 if (!n)
7340 n = m;
7341
7342 switch (pd.proto) {
7343
7344 case IPPROTO_TCP: {
7345 struct tcphdr th;
7346
7347 pd.hdr.tcp = &th;
7348 if (!pf_pull_hdr(m, off, &th, sizeof(th),
7349 &action, &reason, AF_INET6)) {
7350 log = action != PF_PASS;
7351 goto done;
7352 }
7353 pd.p_len = pd.tot_len - off - (th.th_off << 2);
7354 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
7355 if (action == PF_DROP)
7356 goto done;
7357 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
7358 &reason);
7359 if (action == PF_PASS) {
7360 #if NPFSYNC > 0
7361 #ifdef __FreeBSD__
7362 if (pfsync_update_state_ptr != NULL)
7363 pfsync_update_state_ptr(s);
7364 #else
7365 pfsync_update_state(s);
7366 #endif
7367 #endif /* NPFSYNC */
7368 r = s->rule.ptr;
7369 a = s->anchor.ptr;
7370 log = s->log;
7371 } else if (s == NULL)
7372 #ifdef __FreeBSD__
7373 action = pf_test_rule(&r, &s, dir, kif,
7374 m, off, h, &pd, &a, &ruleset, NULL, inp);
7375 #else
7376 action = pf_test_rule(&r, &s, dir, kif,
7377 m, off, h, &pd, &a, &ruleset, &ip6intrq);
7378 #endif
7379 break;
7380 }
7381
7382 case IPPROTO_UDP: {
7383 struct udphdr uh;
7384
7385 pd.hdr.udp = &uh;
7386 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
7387 &action, &reason, AF_INET6)) {
7388 log = action != PF_PASS;
7389 goto done;
7390 }
7391 if (uh.uh_dport == 0 ||
7392 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
7393 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
7394 action = PF_DROP;
7395 REASON_SET(&reason, PFRES_SHORT);
7396 goto done;
7397 }
7398 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
7399 if (action == PF_PASS) {
7400 #if NPFSYNC > 0
7401 #ifdef __FreeBSD__
7402 if (pfsync_update_state_ptr != NULL)
7403 pfsync_update_state_ptr(s);
7404 #else
7405 pfsync_update_state(s);
7406 #endif
7407 #endif /* NPFSYNC */
7408 r = s->rule.ptr;
7409 a = s->anchor.ptr;
7410 log = s->log;
7411 } else if (s == NULL)
7412 #ifdef __FreeBSD__
7413 action = pf_test_rule(&r, &s, dir, kif,
7414 m, off, h, &pd, &a, &ruleset, NULL, inp);
7415 #else
7416 action = pf_test_rule(&r, &s, dir, kif,
7417 m, off, h, &pd, &a, &ruleset, &ip6intrq);
7418 #endif
7419 break;
7420 }
7421
7422 case IPPROTO_ICMP: {
7423 action = PF_DROP;
7424 DPFPRINTF(PF_DEBUG_MISC,
7425 ("pf: dropping IPv6 packet with ICMPv4 payload\n"));
7426 goto done;
7427 }
7428
7429 case IPPROTO_ICMPV6: {
7430 struct icmp6_hdr ih;
7431
7432 pd.hdr.icmp6 = &ih;
7433 if (!pf_pull_hdr(m, off, &ih, sizeof(ih),
7434 &action, &reason, AF_INET6)) {
7435 log = action != PF_PASS;
7436 goto done;
7437 }
7438 action = pf_test_state_icmp(&s, dir, kif,
7439 m, off, h, &pd, &reason);
7440 if (action == PF_PASS) {
7441 #if NPFSYNC > 0
7442 #ifdef __FreeBSD__
7443 if (pfsync_update_state_ptr != NULL)
7444 pfsync_update_state_ptr(s);
7445 #else
7446 pfsync_update_state(s);
7447 #endif
7448 #endif /* NPFSYNC */
7449 r = s->rule.ptr;
7450 a = s->anchor.ptr;
7451 log = s->log;
7452 } else if (s == NULL)
7453 #ifdef __FreeBSD__
7454 action = pf_test_rule(&r, &s, dir, kif,
7455 m, off, h, &pd, &a, &ruleset, NULL, inp);
7456 #else
7457 action = pf_test_rule(&r, &s, dir, kif,
7458 m, off, h, &pd, &a, &ruleset, &ip6intrq);
7459 #endif
7460 break;
7461 }
7462
7463 default:
7464 action = pf_test_state_other(&s, dir, kif, m, &pd);
7465 if (action == PF_PASS) {
7466 #if NPFSYNC > 0
7467 #ifdef __FreeBSD__
7468 if (pfsync_update_state_ptr != NULL)
7469 pfsync_update_state_ptr(s);
7470 #else
7471 pfsync_update_state(s);
7472 #endif
7473 #endif /* NPFSYNC */
7474 r = s->rule.ptr;
7475 a = s->anchor.ptr;
7476 log = s->log;
7477 } else if (s == NULL)
7478 #ifdef __FreeBSD__
7479 action = pf_test_rule(&r, &s, dir, kif, m, off, h,
7480 &pd, &a, &ruleset, NULL, inp);
7481 #else
7482 action = pf_test_rule(&r, &s, dir, kif, m, off, h,
7483 &pd, &a, &ruleset, &ip6intrq);
7484 #endif
7485 break;
7486 }
7487
7488 done:
7489 if (n != m) {
7490 m_freem(n);
7491 n = NULL;
7492 }
7493
7494 /* handle dangerous IPv6 extension headers. */
7495 if (action == PF_PASS && rh_cnt &&
7496 !((s && s->state_flags & PFSTATE_ALLOWOPTS) || r->allow_opts)) {
7497 action = PF_DROP;
7498 REASON_SET(&reason, PFRES_IPOPTIONS);
7499 log = 1;
7500 DPFPRINTF(PF_DEBUG_MISC,
7501 ("pf: dropping packet with dangerous v6 headers\n"));
7502 }
7503
7504 if ((s && s->tag) || r->rtableid >= 0)
7505 #ifdef __FreeBSD__
7506 pf_tag_packet(m, s ? s->tag : 0, r->rtableid, pd.pf_mtag);
7507 #else
7508 pf_tag_packet(m, s ? s->tag : 0, r->rtableid);
7509 #endif
7510
7511 if (dir == PF_IN && s && s->key[PF_SK_STACK])
7512 #ifdef __FreeBSD__
7513 pd.pf_mtag->statekey = s->key[PF_SK_STACK];
7514 #else
7515 m->m_pkthdr.pf.statekey = s->key[PF_SK_STACK];
7516 #endif
7517
7518 #ifdef ALTQ
7519 if (action == PF_PASS && r->qid) {
7520 #ifdef __FreeBSD__
7521 if (pd.tos & IPTOS_LOWDELAY)
7522 pd.pf_mtag->qid = r->pqid;
7523 else
7524 pd.pf_mtag->qid = r->qid;
7525 /* add hints for ecn */
7526 pd.pf_mtag->hdr = h;
7527 #else
7528 if (pd.tos & IPTOS_LOWDELAY)
7529 m->m_pkthdr.pf.qid = r->pqid;
7530 else
7531 m->m_pkthdr.pf.qid = r->qid;
7532 /* add hints for ecn */
7533 m->m_pkthdr.pf.hdr = h;
7534 #endif
7535 }
7536 #endif /* ALTQ */
7537
7538 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
7539 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
7540 (s->nat_rule.ptr->action == PF_RDR ||
7541 s->nat_rule.ptr->action == PF_BINAT) &&
7542 IN6_IS_ADDR_LOOPBACK(&pd.dst->v6))
7543 #ifdef __FreeBSD__
7544 m->m_flags |= M_SKIP_FIREWALL;
7545 #else
7546 m->m_pkthdr.pf.flags |= PF_TAG_TRANSLATE_LOCALHOST;
7547 #endif
7548
7549 #ifdef __FreeBSD__
7550 /* XXX: Anybody working on it?! */
7551 if (r->divert.port)
7552 printf("pf: divert(9) is not supported for IPv6\n");
7553 #else
7554 if (dir == PF_IN && action == PF_PASS && r->divert.port) {
7555 struct pf_divert *divert;
7556
7557 if ((divert = pf_get_divert(m))) {
7558 m->m_pkthdr.pf.flags |= PF_TAG_DIVERTED;
7559 divert->port = r->divert.port;
7560 divert->addr.ipv6 = r->divert.addr.v6;
7561 }
7562 }
7563 #endif
7564
7565 if (log) {
7566 struct pf_rule *lr;
7567
7568 if (s != NULL && s->nat_rule.ptr != NULL &&
7569 s->nat_rule.ptr->log & PF_LOG_ALL)
7570 lr = s->nat_rule.ptr;
7571 else
7572 lr = r;
7573 PFLOG_PACKET(kif, h, m, AF_INET6, dir, reason, lr, a, ruleset,
7574 &pd);
7575 }
7576
7577 kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
7578 kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++;
7579
7580 if (action == PF_PASS || r->action == PF_DROP) {
7581 dirndx = (dir == PF_OUT);
7582 r->packets[dirndx]++;
7583 r->bytes[dirndx] += pd.tot_len;
7584 if (a != NULL) {
7585 a->packets[dirndx]++;
7586 a->bytes[dirndx] += pd.tot_len;
7587 }
7588 if (s != NULL) {
7589 if (s->nat_rule.ptr != NULL) {
7590 s->nat_rule.ptr->packets[dirndx]++;
7591 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len;
7592 }
7593 if (s->src_node != NULL) {
7594 s->src_node->packets[dirndx]++;
7595 s->src_node->bytes[dirndx] += pd.tot_len;
7596 }
7597 if (s->nat_src_node != NULL) {
7598 s->nat_src_node->packets[dirndx]++;
7599 s->nat_src_node->bytes[dirndx] += pd.tot_len;
7600 }
7601 dirndx = (dir == s->direction) ? 0 : 1;
7602 s->packets[dirndx]++;
7603 s->bytes[dirndx] += pd.tot_len;
7604 }
7605 tr = r;
7606 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
7607 #ifdef __FreeBSD__
7608 if (nr != NULL && r == &V_pf_default_rule)
7609 #else
7610 if (nr != NULL && r == &pf_default_rule)
7611 #endif
7612 tr = nr;
7613 if (tr->src.addr.type == PF_ADDR_TABLE)
7614 pfr_update_stats(tr->src.addr.p.tbl,
7615 (s == NULL) ? pd.src :
7616 &s->key[(s->direction == PF_IN)]->addr[0],
7617 pd.af, pd.tot_len, dir == PF_OUT,
7618 r->action == PF_PASS, tr->src.neg);
7619 if (tr->dst.addr.type == PF_ADDR_TABLE)
7620 pfr_update_stats(tr->dst.addr.p.tbl,
7621 (s == NULL) ? pd.dst :
7622 &s->key[(s->direction == PF_IN)]->addr[1],
7623 pd.af, pd.tot_len, dir == PF_OUT,
7624 r->action == PF_PASS, tr->dst.neg);
7625 }
7626
7627 switch (action) {
7628 case PF_SYNPROXY_DROP:
7629 m_freem(*m0);
7630 case PF_DEFER:
7631 *m0 = NULL;
7632 action = PF_PASS;
7633 break;
7634 default:
7635 /* pf_route6 can free the mbuf causing *m0 to become NULL */
7636 if (r->rt)
7637 pf_route6(m0, r, dir, kif->pfik_ifp, s, &pd);
7638 break;
7639 }
7640
7641 #ifdef __FreeBSD__
7642 PF_UNLOCK();
7643 #endif
7644 return (action);
7645 }
7646 #endif /* INET6 */
7647
7648 int
7649 pf_check_congestion(struct ifqueue *ifq)
7650 {
7651 #ifdef __FreeBSD__
7652 /* XXX_IMPORT: later */
7653 return (0);
7654 #else
7655 if (ifq->ifq_congestion)
7656 return (1);
7657 else
7658 return (0);
7659 #endif
7660 }
7661
7662 /*
7663 * must be called whenever any addressing information such as
7664 * address, port, protocol has changed
7665 */
7666 void
7667 pf_pkt_addr_changed(struct mbuf *m)
7668 {
7669 #ifdef __FreeBSD__
7670 struct pf_mtag *pf_tag;
7671
7672 if ((pf_tag = pf_find_mtag(m)) != NULL)
7673 pf_tag->statekey = NULL;
7674 #else
7675 m->m_pkthdr.pf.statekey = NULL;
7676 #endif
7677 }
7678