45#endif
46
47#ifdef __FreeBSD__
48#include "opt_bpf.h"
49#include "opt_pf.h"
50
51#ifdef DEV_BPF
52#define NBPFILTER DEV_BPF
53#else
54#define NBPFILTER 0
55#endif
56
57#ifdef DEV_PFLOG
58#define NPFLOG DEV_PFLOG
59#else
60#define NPFLOG 0
61#endif
62
63#ifdef DEV_PFSYNC
64#define NPFSYNC DEV_PFSYNC
65#else
66#define NPFSYNC 0
67#endif
68
69#else
70#include "bpfilter.h"
71#include "pflog.h"
72#include "pfsync.h"
73#endif
74
75#include <sys/param.h>
76#include <sys/systm.h>
77#include <sys/mbuf.h>
78#include <sys/filio.h>
79#include <sys/socket.h>
80#include <sys/socketvar.h>
81#include <sys/kernel.h>
82#include <sys/time.h>
83#ifdef __FreeBSD__
84#include <sys/sysctl.h>
85#include <sys/endian.h>
86#else
87#include <sys/pool.h>
88#endif
89#include <sys/proc.h>
90#ifdef __FreeBSD__
91#include <sys/kthread.h>
92#include <sys/lock.h>
93#include <sys/sx.h>
94#else
95#include <sys/rwlock.h>
96#endif
97
98#include <net/if.h>
99#include <net/if_types.h>
100#include <net/bpf.h>
101#include <net/route.h>
102#ifndef __FreeBSD__
103#include <net/radix_mpath.h>
104#endif
105
106#include <netinet/in.h>
107#include <netinet/in_var.h>
108#include <netinet/in_systm.h>
109#include <netinet/ip.h>
110#include <netinet/ip_var.h>
111#include <netinet/tcp.h>
112#include <netinet/tcp_seq.h>
113#include <netinet/udp.h>
114#include <netinet/ip_icmp.h>
115#include <netinet/in_pcb.h>
116#include <netinet/tcp_timer.h>
117#include <netinet/tcp_var.h>
118#include <netinet/udp_var.h>
119#include <netinet/icmp_var.h>
120#include <netinet/if_ether.h>
121
122#ifndef __FreeBSD__
123#include <dev/rndvar.h>
124#endif
125#include <net/pfvar.h>
126#include <net/if_pflog.h>
127
128#if NPFSYNC > 0
129#include <net/if_pfsync.h>
130#endif /* NPFSYNC > 0 */
131
132#ifdef INET6
133#include <netinet/ip6.h>
134#include <netinet/in_pcb.h>
135#include <netinet/icmp6.h>
136#include <netinet6/nd6.h>
137#ifdef __FreeBSD__
138#include <netinet6/ip6_var.h>
139#include <netinet6/in6_pcb.h>
140#endif
141#endif /* INET6 */
142
143#ifdef __FreeBSD__
144#include <machine/in_cksum.h>
145#include <sys/limits.h>
146#include <sys/ucred.h>
147#include <security/mac/mac_framework.h>
148
149extern int ip_optcopy(struct ip *, struct ip *);
150extern int debug_pfugidhack;
151#endif
152
153#define DPFPRINTF(n, x) if (pf_status.debug >= (n)) printf x
154
155/*
156 * Global variables
157 */
158
159struct pf_altqqueue pf_altqs[2];
160struct pf_palist pf_pabuf;
161struct pf_altqqueue *pf_altqs_active;
162struct pf_altqqueue *pf_altqs_inactive;
163struct pf_status pf_status;
164
165u_int32_t ticket_altqs_active;
166u_int32_t ticket_altqs_inactive;
167int altqs_inactive_open;
168u_int32_t ticket_pabuf;
169
170struct pf_anchor_stackframe {
171 struct pf_ruleset *rs;
172 struct pf_rule *r;
173 struct pf_anchor_node *parent;
174 struct pf_anchor *child;
175} pf_anchor_stack[64];
176
177#ifdef __FreeBSD__
178uma_zone_t pf_src_tree_pl, pf_rule_pl;
179uma_zone_t pf_state_pl, pf_altq_pl, pf_pooladdr_pl;
180#else
181struct pool pf_src_tree_pl, pf_rule_pl;
182struct pool pf_state_pl, pf_altq_pl, pf_pooladdr_pl;
183#endif
184
185void pf_print_host(struct pf_addr *, u_int16_t, u_int8_t);
186
187void pf_init_threshold(struct pf_threshold *, u_int32_t,
188 u_int32_t);
189void pf_add_threshold(struct pf_threshold *);
190int pf_check_threshold(struct pf_threshold *);
191
192void pf_change_ap(struct pf_addr *, u_int16_t *,
193 u_int16_t *, u_int16_t *, struct pf_addr *,
194 u_int16_t, u_int8_t, sa_family_t);
195int pf_modulate_sack(struct mbuf *, int, struct pf_pdesc *,
196 struct tcphdr *, struct pf_state_peer *);
197#ifdef INET6
198void pf_change_a6(struct pf_addr *, u_int16_t *,
199 struct pf_addr *, u_int8_t);
200#endif /* INET6 */
201void pf_change_icmp(struct pf_addr *, u_int16_t *,
202 struct pf_addr *, struct pf_addr *, u_int16_t,
203 u_int16_t *, u_int16_t *, u_int16_t *,
204 u_int16_t *, u_int8_t, sa_family_t);
205#ifdef __FreeBSD__
206void pf_send_tcp(struct mbuf *,
207 const struct pf_rule *, sa_family_t,
208#else
209void pf_send_tcp(const struct pf_rule *, sa_family_t,
210#endif
211 const struct pf_addr *, const struct pf_addr *,
212 u_int16_t, u_int16_t, u_int32_t, u_int32_t,
213 u_int8_t, u_int16_t, u_int16_t, u_int8_t, int,
214 u_int16_t, struct ether_header *, struct ifnet *);
215void pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t,
216 sa_family_t, struct pf_rule *);
217struct pf_rule *pf_match_translation(struct pf_pdesc *, struct mbuf *,
218 int, int, struct pfi_kif *,
219 struct pf_addr *, u_int16_t, struct pf_addr *,
220 u_int16_t, int);
221struct pf_rule *pf_get_translation(struct pf_pdesc *, struct mbuf *,
222 int, int, struct pfi_kif *, struct pf_src_node **,
223 struct pf_addr *, u_int16_t,
224 struct pf_addr *, u_int16_t,
225 struct pf_addr *, u_int16_t *);
226int pf_test_tcp(struct pf_rule **, struct pf_state **,
227 int, struct pfi_kif *, struct mbuf *, int,
228 void *, struct pf_pdesc *, struct pf_rule **,
229#ifdef __FreeBSD__
230 struct pf_ruleset **, struct ifqueue *,
231 struct inpcb *);
232#else
233 struct pf_ruleset **, struct ifqueue *);
234#endif
235int pf_test_udp(struct pf_rule **, struct pf_state **,
236 int, struct pfi_kif *, struct mbuf *, int,
237 void *, struct pf_pdesc *, struct pf_rule **,
238#ifdef __FreeBSD__
239 struct pf_ruleset **, struct ifqueue *,
240 struct inpcb *);
241#else
242 struct pf_ruleset **, struct ifqueue *);
243#endif
244int pf_test_icmp(struct pf_rule **, struct pf_state **,
245 int, struct pfi_kif *, struct mbuf *, int,
246 void *, struct pf_pdesc *, struct pf_rule **,
247 struct pf_ruleset **, struct ifqueue *);
248int pf_test_other(struct pf_rule **, struct pf_state **,
249 int, struct pfi_kif *, struct mbuf *, int, void *,
250 struct pf_pdesc *, struct pf_rule **,
251 struct pf_ruleset **, struct ifqueue *);
252int pf_test_fragment(struct pf_rule **, int,
253 struct pfi_kif *, struct mbuf *, void *,
254 struct pf_pdesc *, struct pf_rule **,
255 struct pf_ruleset **);
256int pf_test_state_tcp(struct pf_state **, int,
257 struct pfi_kif *, struct mbuf *, int,
258 void *, struct pf_pdesc *, u_short *);
259int pf_test_state_udp(struct pf_state **, int,
260 struct pfi_kif *, struct mbuf *, int,
261 void *, struct pf_pdesc *);
262int pf_test_state_icmp(struct pf_state **, int,
263 struct pfi_kif *, struct mbuf *, int,
264 void *, struct pf_pdesc *, u_short *);
265int pf_test_state_other(struct pf_state **, int,
266 struct pfi_kif *, struct pf_pdesc *);
267int pf_match_tag(struct mbuf *, struct pf_rule *,
268 struct pf_mtag *, int *);
269int pf_step_out_of_anchor(int *, struct pf_ruleset **,
270 int, struct pf_rule **, struct pf_rule **,
271 int *);
272void pf_hash(struct pf_addr *, struct pf_addr *,
273 struct pf_poolhashkey *, sa_family_t);
274int pf_map_addr(u_int8_t, struct pf_rule *,
275 struct pf_addr *, struct pf_addr *,
276 struct pf_addr *, struct pf_src_node **);
277int pf_get_sport(sa_family_t, u_int8_t, struct pf_rule *,
278 struct pf_addr *, struct pf_addr *, u_int16_t,
279 struct pf_addr *, u_int16_t*, u_int16_t, u_int16_t,
280 struct pf_src_node **);
281void pf_route(struct mbuf **, struct pf_rule *, int,
282 struct ifnet *, struct pf_state *,
283 struct pf_pdesc *);
284void pf_route6(struct mbuf **, struct pf_rule *, int,
285 struct ifnet *, struct pf_state *,
286 struct pf_pdesc *);
287#ifdef __FreeBSD__
288/* XXX: import */
289#else
290int pf_socket_lookup(int, struct pf_pdesc *);
291#endif
292u_int8_t pf_get_wscale(struct mbuf *, int, u_int16_t,
293 sa_family_t);
294u_int16_t pf_get_mss(struct mbuf *, int, u_int16_t,
295 sa_family_t);
296u_int16_t pf_calc_mss(struct pf_addr *, sa_family_t,
297 u_int16_t);
298void pf_set_rt_ifp(struct pf_state *,
299 struct pf_addr *);
300int pf_check_proto_cksum(struct mbuf *, int, int,
301 u_int8_t, sa_family_t);
302int pf_addr_wrap_neq(struct pf_addr_wrap *,
303 struct pf_addr_wrap *);
304struct pf_state *pf_find_state_recurse(struct pfi_kif *,
305 struct pf_state_cmp *, u_int8_t);
306int pf_src_connlimit(struct pf_state **);
307int pf_check_congestion(struct ifqueue *);
308
309#ifdef __FreeBSD__
310int in4_cksum(struct mbuf *m, u_int8_t nxt, int off, int len);
311
312extern int pf_end_threads;
313
314struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX];
315#else
316extern struct pool pfr_ktable_pl;
317extern struct pool pfr_kentry_pl;
318
319struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] = {
320 { &pf_state_pl, PFSTATE_HIWAT },
321 { &pf_src_tree_pl, PFSNODE_HIWAT },
322 { &pf_frent_pl, PFFRAG_FRENT_HIWAT },
323 { &pfr_ktable_pl, PFR_KTABLE_HIWAT },
324 { &pfr_kentry_pl, PFR_KENTRY_HIWAT }
325};
326#endif
327
328#define STATE_LOOKUP() \
329 do { \
330 if (direction == PF_IN) \
331 *state = pf_find_state_recurse( \
332 kif, &key, PF_EXT_GWY); \
333 else \
334 *state = pf_find_state_recurse( \
335 kif, &key, PF_LAN_EXT); \
336 if (*state == NULL || (*state)->timeout == PFTM_PURGE) \
337 return (PF_DROP); \
338 if (direction == PF_OUT && \
339 (((*state)->rule.ptr->rt == PF_ROUTETO && \
340 (*state)->rule.ptr->direction == PF_OUT) || \
341 ((*state)->rule.ptr->rt == PF_REPLYTO && \
342 (*state)->rule.ptr->direction == PF_IN)) && \
343 (*state)->rt_kif != NULL && \
344 (*state)->rt_kif != kif) \
345 return (PF_PASS); \
346 } while (0)
347
348#define STATE_TRANSLATE(s) \
349 (s)->lan.addr.addr32[0] != (s)->gwy.addr.addr32[0] || \
350 ((s)->af == AF_INET6 && \
351 ((s)->lan.addr.addr32[1] != (s)->gwy.addr.addr32[1] || \
352 (s)->lan.addr.addr32[2] != (s)->gwy.addr.addr32[2] || \
353 (s)->lan.addr.addr32[3] != (s)->gwy.addr.addr32[3])) || \
354 (s)->lan.port != (s)->gwy.port
355
356#define BOUND_IFACE(r, k) \
357 ((r)->rule_flag & PFRULE_IFBOUND) ? (k) : pfi_all
358
359#define STATE_INC_COUNTERS(s) \
360 do { \
361 s->rule.ptr->states++; \
362 if (s->anchor.ptr != NULL) \
363 s->anchor.ptr->states++; \
364 if (s->nat_rule.ptr != NULL) \
365 s->nat_rule.ptr->states++; \
366 } while (0)
367
368#define STATE_DEC_COUNTERS(s) \
369 do { \
370 if (s->nat_rule.ptr != NULL) \
371 s->nat_rule.ptr->states--; \
372 if (s->anchor.ptr != NULL) \
373 s->anchor.ptr->states--; \
374 s->rule.ptr->states--; \
375 } while (0)
376
377struct pf_src_tree tree_src_tracking;
378
379struct pf_state_tree_id tree_id;
380struct pf_state_queue state_list;
381
382#ifdef __FreeBSD__
383static int pf_src_compare(struct pf_src_node *, struct pf_src_node *);
384static int pf_state_compare_lan_ext(struct pf_state *, struct pf_state *);
385static int pf_state_compare_ext_gwy(struct pf_state *, struct pf_state *);
386static int pf_state_compare_id(struct pf_state *, struct pf_state *);
387#endif
388
389RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare);
390RB_GENERATE(pf_state_tree_lan_ext, pf_state,
391 u.s.entry_lan_ext, pf_state_compare_lan_ext);
392RB_GENERATE(pf_state_tree_ext_gwy, pf_state,
393 u.s.entry_ext_gwy, pf_state_compare_ext_gwy);
394RB_GENERATE(pf_state_tree_id, pf_state,
395 u.s.entry_id, pf_state_compare_id);
396
397#ifdef __FreeBSD__
398static int
399#else
400static __inline int
401#endif
402pf_src_compare(struct pf_src_node *a, struct pf_src_node *b)
403{
404 int diff;
405
406 if (a->rule.ptr > b->rule.ptr)
407 return (1);
408 if (a->rule.ptr < b->rule.ptr)
409 return (-1);
410 if ((diff = a->af - b->af) != 0)
411 return (diff);
412 switch (a->af) {
413#ifdef INET
414 case AF_INET:
415 if (a->addr.addr32[0] > b->addr.addr32[0])
416 return (1);
417 if (a->addr.addr32[0] < b->addr.addr32[0])
418 return (-1);
419 break;
420#endif /* INET */
421#ifdef INET6
422 case AF_INET6:
423 if (a->addr.addr32[3] > b->addr.addr32[3])
424 return (1);
425 if (a->addr.addr32[3] < b->addr.addr32[3])
426 return (-1);
427 if (a->addr.addr32[2] > b->addr.addr32[2])
428 return (1);
429 if (a->addr.addr32[2] < b->addr.addr32[2])
430 return (-1);
431 if (a->addr.addr32[1] > b->addr.addr32[1])
432 return (1);
433 if (a->addr.addr32[1] < b->addr.addr32[1])
434 return (-1);
435 if (a->addr.addr32[0] > b->addr.addr32[0])
436 return (1);
437 if (a->addr.addr32[0] < b->addr.addr32[0])
438 return (-1);
439 break;
440#endif /* INET6 */
441 }
442 return (0);
443}
444
445#ifdef __FreeBSD__
446static int
447#else
448static __inline int
449#endif
450pf_state_compare_lan_ext(struct pf_state *a, struct pf_state *b)
451{
452 int diff;
453
454 if ((diff = a->proto - b->proto) != 0)
455 return (diff);
456 if ((diff = a->af - b->af) != 0)
457 return (diff);
458 switch (a->af) {
459#ifdef INET
460 case AF_INET:
461 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
462 return (1);
463 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
464 return (-1);
465 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
466 return (1);
467 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
468 return (-1);
469 break;
470#endif /* INET */
471#ifdef INET6
472 case AF_INET6:
473 if (a->lan.addr.addr32[3] > b->lan.addr.addr32[3])
474 return (1);
475 if (a->lan.addr.addr32[3] < b->lan.addr.addr32[3])
476 return (-1);
477 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
478 return (1);
479 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
480 return (-1);
481 if (a->lan.addr.addr32[2] > b->lan.addr.addr32[2])
482 return (1);
483 if (a->lan.addr.addr32[2] < b->lan.addr.addr32[2])
484 return (-1);
485 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
486 return (1);
487 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
488 return (-1);
489 if (a->lan.addr.addr32[1] > b->lan.addr.addr32[1])
490 return (1);
491 if (a->lan.addr.addr32[1] < b->lan.addr.addr32[1])
492 return (-1);
493 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
494 return (1);
495 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
496 return (-1);
497 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
498 return (1);
499 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
500 return (-1);
501 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
502 return (1);
503 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
504 return (-1);
505 break;
506#endif /* INET6 */
507 }
508
509 if ((diff = a->lan.port - b->lan.port) != 0)
510 return (diff);
511 if ((diff = a->ext.port - b->ext.port) != 0)
512 return (diff);
513
514 return (0);
515}
516
517#ifdef __FreeBSD__
518static int
519#else
520static __inline int
521#endif
522pf_state_compare_ext_gwy(struct pf_state *a, struct pf_state *b)
523{
524 int diff;
525
526 if ((diff = a->proto - b->proto) != 0)
527 return (diff);
528 if ((diff = a->af - b->af) != 0)
529 return (diff);
530 switch (a->af) {
531#ifdef INET
532 case AF_INET:
533 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
534 return (1);
535 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
536 return (-1);
537 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
538 return (1);
539 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
540 return (-1);
541 break;
542#endif /* INET */
543#ifdef INET6
544 case AF_INET6:
545 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
546 return (1);
547 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
548 return (-1);
549 if (a->gwy.addr.addr32[3] > b->gwy.addr.addr32[3])
550 return (1);
551 if (a->gwy.addr.addr32[3] < b->gwy.addr.addr32[3])
552 return (-1);
553 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
554 return (1);
555 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
556 return (-1);
557 if (a->gwy.addr.addr32[2] > b->gwy.addr.addr32[2])
558 return (1);
559 if (a->gwy.addr.addr32[2] < b->gwy.addr.addr32[2])
560 return (-1);
561 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
562 return (1);
563 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
564 return (-1);
565 if (a->gwy.addr.addr32[1] > b->gwy.addr.addr32[1])
566 return (1);
567 if (a->gwy.addr.addr32[1] < b->gwy.addr.addr32[1])
568 return (-1);
569 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
570 return (1);
571 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
572 return (-1);
573 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
574 return (1);
575 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
576 return (-1);
577 break;
578#endif /* INET6 */
579 }
580
581 if ((diff = a->ext.port - b->ext.port) != 0)
582 return (diff);
583 if ((diff = a->gwy.port - b->gwy.port) != 0)
584 return (diff);
585
586 return (0);
587}
588
589#ifdef __FreeBSD__
590static int
591#else
592static __inline int
593#endif
594pf_state_compare_id(struct pf_state *a, struct pf_state *b)
595{
596 if (a->id > b->id)
597 return (1);
598 if (a->id < b->id)
599 return (-1);
600 if (a->creatorid > b->creatorid)
601 return (1);
602 if (a->creatorid < b->creatorid)
603 return (-1);
604
605 return (0);
606}
607
608#ifdef INET6
609void
610pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af)
611{
612 switch (af) {
613#ifdef INET
614 case AF_INET:
615 dst->addr32[0] = src->addr32[0];
616 break;
617#endif /* INET */
618 case AF_INET6:
619 dst->addr32[0] = src->addr32[0];
620 dst->addr32[1] = src->addr32[1];
621 dst->addr32[2] = src->addr32[2];
622 dst->addr32[3] = src->addr32[3];
623 break;
624 }
625}
626#endif /* INET6 */
627
628struct pf_state *
629pf_find_state_byid(struct pf_state_cmp *key)
630{
631 pf_status.fcounters[FCNT_STATE_SEARCH]++;
632 return (RB_FIND(pf_state_tree_id, &tree_id, (struct pf_state *)key));
633}
634
635struct pf_state *
636pf_find_state_recurse(struct pfi_kif *kif, struct pf_state_cmp *key, u_int8_t tree)
637{
638 struct pf_state *s;
639
640 pf_status.fcounters[FCNT_STATE_SEARCH]++;
641
642 switch (tree) {
643 case PF_LAN_EXT:
644 if ((s = RB_FIND(pf_state_tree_lan_ext, &kif->pfik_lan_ext,
645 (struct pf_state *)key)) != NULL)
646 return (s);
647 if ((s = RB_FIND(pf_state_tree_lan_ext, &pfi_all->pfik_lan_ext,
648 (struct pf_state *)key)) != NULL)
649 return (s);
650 return (NULL);
651 case PF_EXT_GWY:
652 if ((s = RB_FIND(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy,
653 (struct pf_state *)key)) != NULL)
654 return (s);
655 if ((s = RB_FIND(pf_state_tree_ext_gwy, &pfi_all->pfik_ext_gwy,
656 (struct pf_state *)key)) != NULL)
657 return (s);
658 return (NULL);
659 default:
660 panic("pf_find_state_recurse");
661 }
662}
663
664struct pf_state *
665pf_find_state_all(struct pf_state_cmp *key, u_int8_t tree, int *more)
666{
667 struct pf_state *s, *ss = NULL;
668 struct pfi_kif *kif;
669
670 pf_status.fcounters[FCNT_STATE_SEARCH]++;
671
672 switch (tree) {
673 case PF_LAN_EXT:
674 TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states) {
675 s = RB_FIND(pf_state_tree_lan_ext,
676 &kif->pfik_lan_ext, (struct pf_state *)key);
677 if (s == NULL)
678 continue;
679 if (more == NULL)
680 return (s);
681 ss = s;
682 (*more)++;
683 }
684 return (ss);
685 case PF_EXT_GWY:
686 TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states) {
687 s = RB_FIND(pf_state_tree_ext_gwy,
688 &kif->pfik_ext_gwy, (struct pf_state *)key);
689 if (s == NULL)
690 continue;
691 if (more == NULL)
692 return (s);
693 ss = s;
694 (*more)++;
695 }
696 return (ss);
697 default:
698 panic("pf_find_state_all");
699 }
700}
701
702void
703pf_init_threshold(struct pf_threshold *threshold,
704 u_int32_t limit, u_int32_t seconds)
705{
706 threshold->limit = limit * PF_THRESHOLD_MULT;
707 threshold->seconds = seconds;
708 threshold->count = 0;
709 threshold->last = time_second;
710}
711
712void
713pf_add_threshold(struct pf_threshold *threshold)
714{
715 u_int32_t t = time_second, diff = t - threshold->last;
716
717 if (diff >= threshold->seconds)
718 threshold->count = 0;
719 else
720 threshold->count -= threshold->count * diff /
721 threshold->seconds;
722 threshold->count += PF_THRESHOLD_MULT;
723 threshold->last = t;
724}
725
726int
727pf_check_threshold(struct pf_threshold *threshold)
728{
729 return (threshold->count > threshold->limit);
730}
731
732int
733pf_src_connlimit(struct pf_state **state)
734{
735 struct pf_state *s;
736 int bad = 0;
737
738 (*state)->src_node->conn++;
739 (*state)->src.tcp_est = 1;
740 pf_add_threshold(&(*state)->src_node->conn_rate);
741
742 if ((*state)->rule.ptr->max_src_conn &&
743 (*state)->rule.ptr->max_src_conn <
744 (*state)->src_node->conn) {
745 pf_status.lcounters[LCNT_SRCCONN]++;
746 bad++;
747 }
748
749 if ((*state)->rule.ptr->max_src_conn_rate.limit &&
750 pf_check_threshold(&(*state)->src_node->conn_rate)) {
751 pf_status.lcounters[LCNT_SRCCONNRATE]++;
752 bad++;
753 }
754
755 if (!bad)
756 return (0);
757
758 if ((*state)->rule.ptr->overload_tbl) {
759 struct pfr_addr p;
760 u_int32_t killed = 0;
761
762 pf_status.lcounters[LCNT_OVERLOAD_TABLE]++;
763 if (pf_status.debug >= PF_DEBUG_MISC) {
764 printf("pf_src_connlimit: blocking address ");
765 pf_print_host(&(*state)->src_node->addr, 0,
766 (*state)->af);
767 }
768
769 bzero(&p, sizeof(p));
770 p.pfra_af = (*state)->af;
771 switch ((*state)->af) {
772#ifdef INET
773 case AF_INET:
774 p.pfra_net = 32;
775 p.pfra_ip4addr = (*state)->src_node->addr.v4;
776 break;
777#endif /* INET */
778#ifdef INET6
779 case AF_INET6:
780 p.pfra_net = 128;
781 p.pfra_ip6addr = (*state)->src_node->addr.v6;
782 break;
783#endif /* INET6 */
784 }
785
786 pfr_insert_kentry((*state)->rule.ptr->overload_tbl,
787 &p, time_second);
788
789 /* kill existing states if that's required. */
790 if ((*state)->rule.ptr->flush) {
791 pf_status.lcounters[LCNT_OVERLOAD_FLUSH]++;
792
793 RB_FOREACH(s, pf_state_tree_id, &tree_id) {
794 /*
795 * Kill states from this source. (Only those
796 * from the same rule if PF_FLUSH_GLOBAL is not
797 * set)
798 */
799 if (s->af == (*state)->af &&
800 (((*state)->direction == PF_OUT &&
801 PF_AEQ(&(*state)->src_node->addr,
802 &s->lan.addr, s->af)) ||
803 ((*state)->direction == PF_IN &&
804 PF_AEQ(&(*state)->src_node->addr,
805 &s->ext.addr, s->af))) &&
806 ((*state)->rule.ptr->flush &
807 PF_FLUSH_GLOBAL ||
808 (*state)->rule.ptr == s->rule.ptr)) {
809 s->timeout = PFTM_PURGE;
810 s->src.state = s->dst.state =
811 TCPS_CLOSED;
812 killed++;
813 }
814 }
815 if (pf_status.debug >= PF_DEBUG_MISC)
816 printf(", %u states killed", killed);
817 }
818 if (pf_status.debug >= PF_DEBUG_MISC)
819 printf("\n");
820 }
821
822 /* kill this state */
823 (*state)->timeout = PFTM_PURGE;
824 (*state)->src.state = (*state)->dst.state = TCPS_CLOSED;
825 return (1);
826}
827
828int
829pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule,
830 struct pf_addr *src, sa_family_t af)
831{
832 struct pf_src_node k;
833
834 if (*sn == NULL) {
835 k.af = af;
836 PF_ACPY(&k.addr, src, af);
837 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
838 rule->rpool.opts & PF_POOL_STICKYADDR)
839 k.rule.ptr = rule;
840 else
841 k.rule.ptr = NULL;
842 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
843 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
844 }
845 if (*sn == NULL) {
846 if (!rule->max_src_nodes ||
847 rule->src_nodes < rule->max_src_nodes)
848 (*sn) = pool_get(&pf_src_tree_pl, PR_NOWAIT);
849 else
850 pf_status.lcounters[LCNT_SRCNODES]++;
851 if ((*sn) == NULL)
852 return (-1);
853 bzero(*sn, sizeof(struct pf_src_node));
854
855 pf_init_threshold(&(*sn)->conn_rate,
856 rule->max_src_conn_rate.limit,
857 rule->max_src_conn_rate.seconds);
858
859 (*sn)->af = af;
860 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
861 rule->rpool.opts & PF_POOL_STICKYADDR)
862 (*sn)->rule.ptr = rule;
863 else
864 (*sn)->rule.ptr = NULL;
865 PF_ACPY(&(*sn)->addr, src, af);
866 if (RB_INSERT(pf_src_tree,
867 &tree_src_tracking, *sn) != NULL) {
868 if (pf_status.debug >= PF_DEBUG_MISC) {
869 printf("pf: src_tree insert failed: ");
870 pf_print_host(&(*sn)->addr, 0, af);
871 printf("\n");
872 }
873 pool_put(&pf_src_tree_pl, *sn);
874 return (-1);
875 }
876 (*sn)->creation = time_second;
877 (*sn)->ruletype = rule->action;
878 if ((*sn)->rule.ptr != NULL)
879 (*sn)->rule.ptr->src_nodes++;
880 pf_status.scounters[SCNT_SRC_NODE_INSERT]++;
881 pf_status.src_nodes++;
882 } else {
883 if (rule->max_src_states &&
884 (*sn)->states >= rule->max_src_states) {
885 pf_status.lcounters[LCNT_SRCSTATES]++;
886 return (-1);
887 }
888 }
889 return (0);
890}
891
892int
893pf_insert_state(struct pfi_kif *kif, struct pf_state *state)
894{
895 /* Thou MUST NOT insert multiple duplicate keys */
896 state->u.s.kif = kif;
897 if (RB_INSERT(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state)) {
898 if (pf_status.debug >= PF_DEBUG_MISC) {
899 printf("pf: state insert failed: tree_lan_ext");
900 printf(" lan: ");
901 pf_print_host(&state->lan.addr, state->lan.port,
902 state->af);
903 printf(" gwy: ");
904 pf_print_host(&state->gwy.addr, state->gwy.port,
905 state->af);
906 printf(" ext: ");
907 pf_print_host(&state->ext.addr, state->ext.port,
908 state->af);
909 if (state->sync_flags & PFSTATE_FROMSYNC)
910 printf(" (from sync)");
911 printf("\n");
912 }
913 return (-1);
914 }
915
916 if (RB_INSERT(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, state)) {
917 if (pf_status.debug >= PF_DEBUG_MISC) {
918 printf("pf: state insert failed: tree_ext_gwy");
919 printf(" lan: ");
920 pf_print_host(&state->lan.addr, state->lan.port,
921 state->af);
922 printf(" gwy: ");
923 pf_print_host(&state->gwy.addr, state->gwy.port,
924 state->af);
925 printf(" ext: ");
926 pf_print_host(&state->ext.addr, state->ext.port,
927 state->af);
928 if (state->sync_flags & PFSTATE_FROMSYNC)
929 printf(" (from sync)");
930 printf("\n");
931 }
932 RB_REMOVE(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state);
933 return (-1);
934 }
935
936 if (state->id == 0 && state->creatorid == 0) {
937 state->id = htobe64(pf_status.stateid++);
938 state->creatorid = pf_status.hostid;
939 }
940 if (RB_INSERT(pf_state_tree_id, &tree_id, state) != NULL) {
941 if (pf_status.debug >= PF_DEBUG_MISC) {
942#ifdef __FreeBSD__
943 printf("pf: state insert failed: "
944 "id: %016llx creatorid: %08x",
945 (long long)be64toh(state->id),
946 ntohl(state->creatorid));
947#else
948 printf("pf: state insert failed: "
949 "id: %016llx creatorid: %08x",
950 betoh64(state->id), ntohl(state->creatorid));
951#endif
952 if (state->sync_flags & PFSTATE_FROMSYNC)
953 printf(" (from sync)");
954 printf("\n");
955 }
956 RB_REMOVE(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state);
957 RB_REMOVE(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, state);
958 return (-1);
959 }
960 TAILQ_INSERT_TAIL(&state_list, state, u.s.entry_list);
961 pf_status.fcounters[FCNT_STATE_INSERT]++;
962 pf_status.states++;
963 pfi_kif_ref(kif, PFI_KIF_REF_STATE);
964#if NPFSYNC
965 pfsync_insert_state(state);
966#endif
967 return (0);
968}
969
970void
971pf_purge_thread(void *v)
972{
973 int nloops = 0, s;
974
975 for (;;) {
976 tsleep(pf_purge_thread, PWAIT, "pftm", 1 * hz);
977
978#ifdef __FreeBSD__
979 sx_slock(&pf_consistency_lock);
980 PF_LOCK();
981
982 if (pf_end_threads) {
983 pf_purge_expired_states(pf_status.states);
984 pf_purge_expired_fragments();
985 pf_purge_expired_src_nodes(0);
986 pf_end_threads++;
987
988 sx_sunlock(&pf_consistency_lock);
989 PF_UNLOCK();
990 wakeup(pf_purge_thread);
991 kproc_exit(0);
992 }
993#endif
994 s = splsoftnet();
995
996 /* process a fraction of the state table every second */
997 pf_purge_expired_states(1 + (pf_status.states
998 / pf_default_rule.timeout[PFTM_INTERVAL]));
999
1000 /* purge other expired types every PFTM_INTERVAL seconds */
1001 if (++nloops >= pf_default_rule.timeout[PFTM_INTERVAL]) {
1002 pf_purge_expired_fragments();
1003 pf_purge_expired_src_nodes(0);
1004 nloops = 0;
1005 }
1006
1007 splx(s);
1008#ifdef __FreeBSD__
1009 PF_UNLOCK();
1010 sx_sunlock(&pf_consistency_lock);
1011#endif
1012 }
1013}
1014
1015u_int32_t
1016pf_state_expires(const struct pf_state *state)
1017{
1018 u_int32_t timeout;
1019 u_int32_t start;
1020 u_int32_t end;
1021 u_int32_t states;
1022
1023 /* handle all PFTM_* > PFTM_MAX here */
1024 if (state->timeout == PFTM_PURGE)
1025 return (time_second);
1026 if (state->timeout == PFTM_UNTIL_PACKET)
1027 return (0);
1028#ifdef __FreeBSD__
1029 KASSERT(state->timeout != PFTM_UNLINKED,
1030 ("pf_state_expires: timeout == PFTM_UNLINKED"));
1031 KASSERT((state->timeout < PFTM_MAX),
1032 ("pf_state_expires: timeout > PFTM_MAX"));
1033#else
1034 KASSERT(state->timeout != PFTM_UNLINKED);
1035 KASSERT(state->timeout < PFTM_MAX);
1036#endif
1037 timeout = state->rule.ptr->timeout[state->timeout];
1038 if (!timeout)
1039 timeout = pf_default_rule.timeout[state->timeout];
1040 start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START];
1041 if (start) {
1042 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END];
1043 states = state->rule.ptr->states;
1044 } else {
1045 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START];
1046 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END];
1047 states = pf_status.states;
1048 }
1049 if (end && states > start && start < end) {
1050 if (states < end)
1051 return (state->expire + timeout * (end - states) /
1052 (end - start));
1053 else
1054 return (time_second);
1055 }
1056 return (state->expire + timeout);
1057}
1058
1059void
1060pf_purge_expired_src_nodes(int waslocked)
1061{
1062 struct pf_src_node *cur, *next;
1063 int locked = waslocked;
1064
1065 for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) {
1066 next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur);
1067
1068 if (cur->states <= 0 && cur->expire <= time_second) {
1069 if (! locked) {
1070#ifdef __FreeBSD__
1071 if (!sx_try_upgrade(&pf_consistency_lock)) {
1072 PF_UNLOCK();
1073 sx_sunlock(&pf_consistency_lock);
1074 sx_xlock(&pf_consistency_lock);
1075 PF_LOCK();
1076 }
1077#else
1078 rw_enter_write(&pf_consistency_lock);
1079#endif
1080 next = RB_NEXT(pf_src_tree,
1081 &tree_src_tracking, cur);
1082 locked = 1;
1083 }
1084 if (cur->rule.ptr != NULL) {
1085 cur->rule.ptr->src_nodes--;
1086 if (cur->rule.ptr->states <= 0 &&
1087 cur->rule.ptr->max_src_nodes <= 0)
1088 pf_rm_rule(NULL, cur->rule.ptr);
1089 }
1090 RB_REMOVE(pf_src_tree, &tree_src_tracking, cur);
1091 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
1092 pf_status.src_nodes--;
1093 pool_put(&pf_src_tree_pl, cur);
1094 }
1095 }
1096
1097 if (locked && !waslocked)
1098#ifdef __FreeBSD__
1099 sx_downgrade(&pf_consistency_lock);
1100#else
1101 rw_exit_write(&pf_consistency_lock);
1102#endif
1103}
1104
1105void
1106pf_src_tree_remove_state(struct pf_state *s)
1107{
1108 u_int32_t timeout;
1109
1110 if (s->src_node != NULL) {
1111 if (s->proto == IPPROTO_TCP) {
1112 if (s->src.tcp_est)
1113 --s->src_node->conn;
1114 }
1115 if (--s->src_node->states <= 0) {
1116 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1117 if (!timeout)
1118 timeout =
1119 pf_default_rule.timeout[PFTM_SRC_NODE];
1120 s->src_node->expire = time_second + timeout;
1121 }
1122 }
1123 if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) {
1124 if (--s->nat_src_node->states <= 0) {
1125 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1126 if (!timeout)
1127 timeout =
1128 pf_default_rule.timeout[PFTM_SRC_NODE];
1129 s->nat_src_node->expire = time_second + timeout;
1130 }
1131 }
1132 s->src_node = s->nat_src_node = NULL;
1133}
1134
1135/* callers should be at splsoftnet */
1136void
1137pf_unlink_state(struct pf_state *cur)
1138{
1139#ifdef __FreeBSD__
1140 if (cur->local_flags & PFSTATE_EXPIRING)
1141 return;
1142 cur->local_flags |= PFSTATE_EXPIRING;
1143#endif
1144 if (cur->src.state == PF_TCPS_PROXY_DST) {
1145#ifdef __FreeBSD__
1146 pf_send_tcp(NULL, cur->rule.ptr, cur->af,
1147#else
1148 pf_send_tcp(cur->rule.ptr, cur->af,
1149#endif
1150 &cur->ext.addr, &cur->lan.addr,
1151 cur->ext.port, cur->lan.port,
1152 cur->src.seqhi, cur->src.seqlo + 1,
1153 TH_RST|TH_ACK, 0, 0, 0, 1, cur->tag, NULL, NULL);
1154 }
1155 RB_REMOVE(pf_state_tree_ext_gwy,
1156 &cur->u.s.kif->pfik_ext_gwy, cur);
1157 RB_REMOVE(pf_state_tree_lan_ext,
1158 &cur->u.s.kif->pfik_lan_ext, cur);
1159 RB_REMOVE(pf_state_tree_id, &tree_id, cur);
1160#if NPFSYNC
1161 if (cur->creatorid == pf_status.hostid)
1162 pfsync_delete_state(cur);
1163#endif
1164 cur->timeout = PFTM_UNLINKED;
1165 pf_src_tree_remove_state(cur);
1166}
1167
1168/* callers should be at splsoftnet and hold the
1169 * write_lock on pf_consistency_lock */
1170void
1171pf_free_state(struct pf_state *cur)
1172{
1173#if NPFSYNC
1174 if (pfsyncif != NULL &&
1175 (pfsyncif->sc_bulk_send_next == cur ||
1176 pfsyncif->sc_bulk_terminator == cur))
1177 return;
1178#endif
1179#ifdef __FreeBSD__
1180 KASSERT(cur->timeout == PFTM_UNLINKED,
1181 ("pf_free_state: cur->timeout != PFTM_UNLINKED"));
1182#else
1183 KASSERT(cur->timeout == PFTM_UNLINKED);
1184#endif
1185 if (--cur->rule.ptr->states <= 0 &&
1186 cur->rule.ptr->src_nodes <= 0)
1187 pf_rm_rule(NULL, cur->rule.ptr);
1188 if (cur->nat_rule.ptr != NULL)
1189 if (--cur->nat_rule.ptr->states <= 0 &&
1190 cur->nat_rule.ptr->src_nodes <= 0)
1191 pf_rm_rule(NULL, cur->nat_rule.ptr);
1192 if (cur->anchor.ptr != NULL)
1193 if (--cur->anchor.ptr->states <= 0)
1194 pf_rm_rule(NULL, cur->anchor.ptr);
1195 pf_normalize_tcp_cleanup(cur);
1196 pfi_kif_unref(cur->u.s.kif, PFI_KIF_REF_STATE);
1197 TAILQ_REMOVE(&state_list, cur, u.s.entry_list);
1198 if (cur->tag)
1199 pf_tag_unref(cur->tag);
1200 pool_put(&pf_state_pl, cur);
1201 pf_status.fcounters[FCNT_STATE_REMOVALS]++;
1202 pf_status.states--;
1203}
1204
1205void
1206pf_purge_expired_states(u_int32_t maxcheck)
1207{
1208 static struct pf_state *cur = NULL;
1209 struct pf_state *next;
1210 int locked = 0;
1211
1212 while (maxcheck--) {
1213 /* wrap to start of list when we hit the end */
1214 if (cur == NULL) {
1215 cur = TAILQ_FIRST(&state_list);
1216 if (cur == NULL)
1217 break; /* list empty */
1218 }
1219
1220 /* get next state, as cur may get deleted */
1221 next = TAILQ_NEXT(cur, u.s.entry_list);
1222
1223 if (cur->timeout == PFTM_UNLINKED) {
1224 /* free unlinked state */
1225 if (! locked) {
1226#ifdef __FreeBSD__
1227 if (!sx_try_upgrade(&pf_consistency_lock)) {
1228 PF_UNLOCK();
1229 sx_sunlock(&pf_consistency_lock);
1230 sx_xlock(&pf_consistency_lock);
1231 PF_LOCK();
1232 }
1233#else
1234 rw_enter_write(&pf_consistency_lock);
1235#endif
1236 locked = 1;
1237 }
1238 pf_free_state(cur);
1239 } else if (pf_state_expires(cur) <= time_second) {
1240 /* unlink and free expired state */
1241 pf_unlink_state(cur);
1242 if (! locked) {
1243#ifdef __FreeBSD__
1244 if (!sx_try_upgrade(&pf_consistency_lock)) {
1245 PF_UNLOCK();
1246 sx_sunlock(&pf_consistency_lock);
1247 sx_xlock(&pf_consistency_lock);
1248 PF_LOCK();
1249 }
1250#else
1251 rw_enter_write(&pf_consistency_lock);
1252#endif
1253 locked = 1;
1254 }
1255 pf_free_state(cur);
1256 }
1257 cur = next;
1258 }
1259
1260 if (locked)
1261#ifdef __FreeBSD__
1262 sx_downgrade(&pf_consistency_lock);
1263#else
1264 rw_exit_write(&pf_consistency_lock);
1265#endif
1266}
1267
1268int
1269pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw)
1270{
1271 if (aw->type != PF_ADDR_TABLE)
1272 return (0);
1273 if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL)
1274 return (1);
1275 return (0);
1276}
1277
1278void
1279pf_tbladdr_remove(struct pf_addr_wrap *aw)
1280{
1281 if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL)
1282 return;
1283 pfr_detach_table(aw->p.tbl);
1284 aw->p.tbl = NULL;
1285}
1286
1287void
1288pf_tbladdr_copyout(struct pf_addr_wrap *aw)
1289{
1290 struct pfr_ktable *kt = aw->p.tbl;
1291
1292 if (aw->type != PF_ADDR_TABLE || kt == NULL)
1293 return;
1294 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
1295 kt = kt->pfrkt_root;
1296 aw->p.tbl = NULL;
1297 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
1298 kt->pfrkt_cnt : -1;
1299}
1300
1301void
1302pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af)
1303{
1304 switch (af) {
1305#ifdef INET
1306 case AF_INET: {
1307 u_int32_t a = ntohl(addr->addr32[0]);
1308 printf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255,
1309 (a>>8)&255, a&255);
1310 if (p) {
1311 p = ntohs(p);
1312 printf(":%u", p);
1313 }
1314 break;
1315 }
1316#endif /* INET */
1317#ifdef INET6
1318 case AF_INET6: {
1319 u_int16_t b;
1320 u_int8_t i, curstart = 255, curend = 0,
1321 maxstart = 0, maxend = 0;
1322 for (i = 0; i < 8; i++) {
1323 if (!addr->addr16[i]) {
1324 if (curstart == 255)
1325 curstart = i;
1326 else
1327 curend = i;
1328 } else {
1329 if (curstart) {
1330 if ((curend - curstart) >
1331 (maxend - maxstart)) {
1332 maxstart = curstart;
1333 maxend = curend;
1334 curstart = 255;
1335 }
1336 }
1337 }
1338 }
1339 for (i = 0; i < 8; i++) {
1340 if (i >= maxstart && i <= maxend) {
1341 if (maxend != 7) {
1342 if (i == maxstart)
1343 printf(":");
1344 } else {
1345 if (i == maxend)
1346 printf(":");
1347 }
1348 } else {
1349 b = ntohs(addr->addr16[i]);
1350 printf("%x", b);
1351 if (i < 7)
1352 printf(":");
1353 }
1354 }
1355 if (p) {
1356 p = ntohs(p);
1357 printf("[%u]", p);
1358 }
1359 break;
1360 }
1361#endif /* INET6 */
1362 }
1363}
1364
1365void
1366pf_print_state(struct pf_state *s)
1367{
1368 switch (s->proto) {
1369 case IPPROTO_TCP:
1370 printf("TCP ");
1371 break;
1372 case IPPROTO_UDP:
1373 printf("UDP ");
1374 break;
1375 case IPPROTO_ICMP:
1376 printf("ICMP ");
1377 break;
1378 case IPPROTO_ICMPV6:
1379 printf("ICMPV6 ");
1380 break;
1381 default:
1382 printf("%u ", s->proto);
1383 break;
1384 }
1385 pf_print_host(&s->lan.addr, s->lan.port, s->af);
1386 printf(" ");
1387 pf_print_host(&s->gwy.addr, s->gwy.port, s->af);
1388 printf(" ");
1389 pf_print_host(&s->ext.addr, s->ext.port, s->af);
1390 printf(" [lo=%u high=%u win=%u modulator=%u", s->src.seqlo,
1391 s->src.seqhi, s->src.max_win, s->src.seqdiff);
1392 if (s->src.wscale && s->dst.wscale)
1393 printf(" wscale=%u", s->src.wscale & PF_WSCALE_MASK);
1394 printf("]");
1395 printf(" [lo=%u high=%u win=%u modulator=%u", s->dst.seqlo,
1396 s->dst.seqhi, s->dst.max_win, s->dst.seqdiff);
1397 if (s->src.wscale && s->dst.wscale)
1398 printf(" wscale=%u", s->dst.wscale & PF_WSCALE_MASK);
1399 printf("]");
1400 printf(" %u:%u", s->src.state, s->dst.state);
1401}
1402
1403void
1404pf_print_flags(u_int8_t f)
1405{
1406 if (f)
1407 printf(" ");
1408 if (f & TH_FIN)
1409 printf("F");
1410 if (f & TH_SYN)
1411 printf("S");
1412 if (f & TH_RST)
1413 printf("R");
1414 if (f & TH_PUSH)
1415 printf("P");
1416 if (f & TH_ACK)
1417 printf("A");
1418 if (f & TH_URG)
1419 printf("U");
1420 if (f & TH_ECE)
1421 printf("E");
1422 if (f & TH_CWR)
1423 printf("W");
1424}
1425
1426#define PF_SET_SKIP_STEPS(i) \
1427 do { \
1428 while (head[i] != cur) { \
1429 head[i]->skip[i].ptr = cur; \
1430 head[i] = TAILQ_NEXT(head[i], entries); \
1431 } \
1432 } while (0)
1433
1434void
1435pf_calc_skip_steps(struct pf_rulequeue *rules)
1436{
1437 struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT];
1438 int i;
1439
1440 cur = TAILQ_FIRST(rules);
1441 prev = cur;
1442 for (i = 0; i < PF_SKIP_COUNT; ++i)
1443 head[i] = cur;
1444 while (cur != NULL) {
1445
1446 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
1447 PF_SET_SKIP_STEPS(PF_SKIP_IFP);
1448 if (cur->direction != prev->direction)
1449 PF_SET_SKIP_STEPS(PF_SKIP_DIR);
1450 if (cur->af != prev->af)
1451 PF_SET_SKIP_STEPS(PF_SKIP_AF);
1452 if (cur->proto != prev->proto)
1453 PF_SET_SKIP_STEPS(PF_SKIP_PROTO);
1454 if (cur->src.neg != prev->src.neg ||
1455 pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr))
1456 PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR);
1457 if (cur->src.port[0] != prev->src.port[0] ||
1458 cur->src.port[1] != prev->src.port[1] ||
1459 cur->src.port_op != prev->src.port_op)
1460 PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
1461 if (cur->dst.neg != prev->dst.neg ||
1462 pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr))
1463 PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR);
1464 if (cur->dst.port[0] != prev->dst.port[0] ||
1465 cur->dst.port[1] != prev->dst.port[1] ||
1466 cur->dst.port_op != prev->dst.port_op)
1467 PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
1468
1469 prev = cur;
1470 cur = TAILQ_NEXT(cur, entries);
1471 }
1472 for (i = 0; i < PF_SKIP_COUNT; ++i)
1473 PF_SET_SKIP_STEPS(i);
1474}
1475
1476int
1477pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2)
1478{
1479 if (aw1->type != aw2->type)
1480 return (1);
1481 switch (aw1->type) {
1482 case PF_ADDR_ADDRMASK:
1483 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0))
1484 return (1);
1485 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0))
1486 return (1);
1487 return (0);
1488 case PF_ADDR_DYNIFTL:
1489 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt);
1490 case PF_ADDR_NOROUTE:
1491 case PF_ADDR_URPFFAILED:
1492 return (0);
1493 case PF_ADDR_TABLE:
1494 return (aw1->p.tbl != aw2->p.tbl);
1495 case PF_ADDR_RTLABEL:
1496 return (aw1->v.rtlabel != aw2->v.rtlabel);
1497 default:
1498 printf("invalid address type: %d\n", aw1->type);
1499 return (1);
1500 }
1501}
1502
1503u_int16_t
1504pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp)
1505{
1506 u_int32_t l;
1507
1508 if (udp && !cksum)
1509 return (0x0000);
1510 l = cksum + old - new;
1511 l = (l >> 16) + (l & 65535);
1512 l = l & 65535;
1513 if (udp && !l)
1514 return (0xFFFF);
1515 return (l);
1516}
1517
1518void
1519pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc,
1520 struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af)
1521{
1522 struct pf_addr ao;
1523 u_int16_t po = *p;
1524
1525 PF_ACPY(&ao, a, af);
1526 PF_ACPY(a, an, af);
1527
1528 *p = pn;
1529
1530 switch (af) {
1531#ifdef INET
1532 case AF_INET:
1533 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1534 ao.addr16[0], an->addr16[0], 0),
1535 ao.addr16[1], an->addr16[1], 0);
1536 *p = pn;
1537 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1538 ao.addr16[0], an->addr16[0], u),
1539 ao.addr16[1], an->addr16[1], u),
1540 po, pn, u);
1541 break;
1542#endif /* INET */
1543#ifdef INET6
1544 case AF_INET6:
1545 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1546 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1547 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1548 ao.addr16[0], an->addr16[0], u),
1549 ao.addr16[1], an->addr16[1], u),
1550 ao.addr16[2], an->addr16[2], u),
1551 ao.addr16[3], an->addr16[3], u),
1552 ao.addr16[4], an->addr16[4], u),
1553 ao.addr16[5], an->addr16[5], u),
1554 ao.addr16[6], an->addr16[6], u),
1555 ao.addr16[7], an->addr16[7], u),
1556 po, pn, u);
1557 break;
1558#endif /* INET6 */
1559 }
1560}
1561
1562
1563/* Changes a u_int32_t. Uses a void * so there are no align restrictions */
1564void
1565pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u)
1566{
1567 u_int32_t ao;
1568
1569 memcpy(&ao, a, sizeof(ao));
1570 memcpy(a, &an, sizeof(u_int32_t));
1571 *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u),
1572 ao % 65536, an % 65536, u);
1573}
1574
1575#ifdef INET6
1576void
1577pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u)
1578{
1579 struct pf_addr ao;
1580
1581 PF_ACPY(&ao, a, AF_INET6);
1582 PF_ACPY(a, an, AF_INET6);
1583
1584 *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1585 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1586 pf_cksum_fixup(pf_cksum_fixup(*c,
1587 ao.addr16[0], an->addr16[0], u),
1588 ao.addr16[1], an->addr16[1], u),
1589 ao.addr16[2], an->addr16[2], u),
1590 ao.addr16[3], an->addr16[3], u),
1591 ao.addr16[4], an->addr16[4], u),
1592 ao.addr16[5], an->addr16[5], u),
1593 ao.addr16[6], an->addr16[6], u),
1594 ao.addr16[7], an->addr16[7], u);
1595}
1596#endif /* INET6 */
1597
1598void
1599pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa,
1600 struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c,
1601 u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af)
1602{
1603 struct pf_addr oia, ooa;
1604
1605 PF_ACPY(&oia, ia, af);
1606 PF_ACPY(&ooa, oa, af);
1607
1608 /* Change inner protocol port, fix inner protocol checksum. */
1609 if (ip != NULL) {
1610 u_int16_t oip = *ip;
1611 u_int32_t opc = 0; /* make the compiler happy */
1612
1613 if (pc != NULL)
1614 opc = *pc;
1615 *ip = np;
1616 if (pc != NULL)
1617 *pc = pf_cksum_fixup(*pc, oip, *ip, u);
1618 *ic = pf_cksum_fixup(*ic, oip, *ip, 0);
1619 if (pc != NULL)
1620 *ic = pf_cksum_fixup(*ic, opc, *pc, 0);
1621 }
1622 /* Change inner ip address, fix inner ip and icmp checksums. */
1623 PF_ACPY(ia, na, af);
1624 switch (af) {
1625#ifdef INET
1626 case AF_INET: {
1627 u_int32_t oh2c = *h2c;
1628
1629 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c,
1630 oia.addr16[0], ia->addr16[0], 0),
1631 oia.addr16[1], ia->addr16[1], 0);
1632 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1633 oia.addr16[0], ia->addr16[0], 0),
1634 oia.addr16[1], ia->addr16[1], 0);
1635 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0);
1636 break;
1637 }
1638#endif /* INET */
1639#ifdef INET6
1640 case AF_INET6:
1641 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1642 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1643 pf_cksum_fixup(pf_cksum_fixup(*ic,
1644 oia.addr16[0], ia->addr16[0], u),
1645 oia.addr16[1], ia->addr16[1], u),
1646 oia.addr16[2], ia->addr16[2], u),
1647 oia.addr16[3], ia->addr16[3], u),
1648 oia.addr16[4], ia->addr16[4], u),
1649 oia.addr16[5], ia->addr16[5], u),
1650 oia.addr16[6], ia->addr16[6], u),
1651 oia.addr16[7], ia->addr16[7], u);
1652 break;
1653#endif /* INET6 */
1654 }
1655 /* Change outer ip address, fix outer ip or icmpv6 checksum. */
1656 PF_ACPY(oa, na, af);
1657 switch (af) {
1658#ifdef INET
1659 case AF_INET:
1660 *hc = pf_cksum_fixup(pf_cksum_fixup(*hc,
1661 ooa.addr16[0], oa->addr16[0], 0),
1662 ooa.addr16[1], oa->addr16[1], 0);
1663 break;
1664#endif /* INET */
1665#ifdef INET6
1666 case AF_INET6:
1667 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1668 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1669 pf_cksum_fixup(pf_cksum_fixup(*ic,
1670 ooa.addr16[0], oa->addr16[0], u),
1671 ooa.addr16[1], oa->addr16[1], u),
1672 ooa.addr16[2], oa->addr16[2], u),
1673 ooa.addr16[3], oa->addr16[3], u),
1674 ooa.addr16[4], oa->addr16[4], u),
1675 ooa.addr16[5], oa->addr16[5], u),
1676 ooa.addr16[6], oa->addr16[6], u),
1677 ooa.addr16[7], oa->addr16[7], u);
1678 break;
1679#endif /* INET6 */
1680 }
1681}
1682
1683
1684/*
1685 * Need to modulate the sequence numbers in the TCP SACK option
1686 * (credits to Krzysztof Pfaff for report and patch)
1687 */
1688int
1689pf_modulate_sack(struct mbuf *m, int off, struct pf_pdesc *pd,
1690 struct tcphdr *th, struct pf_state_peer *dst)
1691{
1692 int hlen = (th->th_off << 2) - sizeof(*th), thoptlen = hlen;
1693#ifdef __FreeBSD__
1694 u_int8_t opts[TCP_MAXOLEN], *opt = opts;
1695#else
1696 u_int8_t opts[MAX_TCPOPTLEN], *opt = opts;
1697#endif
1698 int copyback = 0, i, olen;
1699 struct sackblk sack;
1700
1701#define TCPOLEN_SACKLEN (TCPOLEN_SACK + 2)
1702 if (hlen < TCPOLEN_SACKLEN ||
1703 !pf_pull_hdr(m, off + sizeof(*th), opts, hlen, NULL, NULL, pd->af))
1704 return 0;
1705
1706 while (hlen >= TCPOLEN_SACKLEN) {
1707 olen = opt[1];
1708 switch (*opt) {
1709 case TCPOPT_EOL: /* FALLTHROUGH */
1710 case TCPOPT_NOP:
1711 opt++;
1712 hlen--;
1713 break;
1714 case TCPOPT_SACK:
1715 if (olen > hlen)
1716 olen = hlen;
1717 if (olen >= TCPOLEN_SACKLEN) {
1718 for (i = 2; i + TCPOLEN_SACK <= olen;
1719 i += TCPOLEN_SACK) {
1720 memcpy(&sack, &opt[i], sizeof(sack));
1721 pf_change_a(&sack.start, &th->th_sum,
1722 htonl(ntohl(sack.start) -
1723 dst->seqdiff), 0);
1724 pf_change_a(&sack.end, &th->th_sum,
1725 htonl(ntohl(sack.end) -
1726 dst->seqdiff), 0);
1727 memcpy(&opt[i], &sack, sizeof(sack));
1728 }
1729 copyback = 1;
1730 }
1731 /* FALLTHROUGH */
1732 default:
1733 if (olen < 2)
1734 olen = 2;
1735 hlen -= olen;
1736 opt += olen;
1737 }
1738 }
1739
1740 if (copyback)
1741#ifdef __FreeBSD__
1742 m_copyback(m, off + sizeof(*th), thoptlen, (caddr_t)opts);
1743#else
1744 m_copyback(m, off + sizeof(*th), thoptlen, opts);
1745#endif
1746 return (copyback);
1747}
1748
1749void
1750#ifdef __FreeBSD__
1751pf_send_tcp(struct mbuf *replyto, const struct pf_rule *r, sa_family_t af,
1752#else
1753pf_send_tcp(const struct pf_rule *r, sa_family_t af,
1754#endif
1755 const struct pf_addr *saddr, const struct pf_addr *daddr,
1756 u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack,
1757 u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl, int tag,
1758 u_int16_t rtag, struct ether_header *eh, struct ifnet *ifp)
1759{
1760 struct mbuf *m;
1761 int len, tlen;
1762#ifdef INET
1763 struct ip *h;
1764#endif /* INET */
1765#ifdef INET6
1766 struct ip6_hdr *h6;
1767#endif /* INET6 */
1768 struct tcphdr *th;
1769 char *opt;
1770 struct pf_mtag *pf_mtag;
1771
1772#ifdef __FreeBSD__
1773 KASSERT(
1774#ifdef INET
1775 af == AF_INET
1776#else
1777 0
1778#endif
1779 ||
1780#ifdef INET6
1781 af == AF_INET6
1782#else
1783 0
1784#endif
1785 , ("Unsupported AF %d", af));
1786 len = 0;
1787 th = NULL;
1788#ifdef INET
1789 h = NULL;
1790#endif
1791#ifdef INET6
1792 h6 = NULL;
1793#endif
1794#endif
1795
1796 /* maximum segment size tcp option */
1797 tlen = sizeof(struct tcphdr);
1798 if (mss)
1799 tlen += 4;
1800
1801 switch (af) {
1802#ifdef INET
1803 case AF_INET:
1804 len = sizeof(struct ip) + tlen;
1805 break;
1806#endif /* INET */
1807#ifdef INET6
1808 case AF_INET6:
1809 len = sizeof(struct ip6_hdr) + tlen;
1810 break;
1811#endif /* INET6 */
1812 }
1813
1814 /* create outgoing mbuf */
1815 m = m_gethdr(M_DONTWAIT, MT_HEADER);
1816 if (m == NULL)
1817 return;
1818#ifdef __FreeBSD__
1819#ifdef MAC
1820 if (replyto)
1821 mac_netinet_firewall_reply(replyto, m);
1822 else
1823 mac_netinet_firewall_send(m);
1824#else
1825 (void)replyto;
1826#endif
1827#endif
1828 if ((pf_mtag = pf_get_mtag(m)) == NULL) {
1829 m_freem(m);
1830 return;
1831 }
1832 if (tag)
1833#ifdef __FreeBSD__
1834 m->m_flags |= M_SKIP_FIREWALL;
1835#else
1836 pf_mtag->flags |= PF_TAG_GENERATED;
1837#endif
1838
1839 pf_mtag->tag = rtag;
1840
1841 if (r != NULL && r->rtableid >= 0)
1842#ifdef __FreeBSD__
1843 {
1844 M_SETFIB(m, r->rtableid);
1845#endif
1846 pf_mtag->rtableid = r->rtableid;
1847#ifdef __FreeBSD__
1848 }
1849#endif
1850#ifdef ALTQ
1851 if (r != NULL && r->qid) {
1852 pf_mtag->qid = r->qid;
1853 /* add hints for ecn */
1854 pf_mtag->af = af;
1855 pf_mtag->hdr = mtod(m, struct ip *);
1856 }
1857#endif /* ALTQ */
1858 m->m_data += max_linkhdr;
1859 m->m_pkthdr.len = m->m_len = len;
1860 m->m_pkthdr.rcvif = NULL;
1861 bzero(m->m_data, len);
1862 switch (af) {
1863#ifdef INET
1864 case AF_INET:
1865 h = mtod(m, struct ip *);
1866
1867 /* IP header fields included in the TCP checksum */
1868 h->ip_p = IPPROTO_TCP;
1869 h->ip_len = htons(tlen);
1870 h->ip_src.s_addr = saddr->v4.s_addr;
1871 h->ip_dst.s_addr = daddr->v4.s_addr;
1872
1873 th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip));
1874 break;
1875#endif /* INET */
1876#ifdef INET6
1877 case AF_INET6:
1878 h6 = mtod(m, struct ip6_hdr *);
1879
1880 /* IP header fields included in the TCP checksum */
1881 h6->ip6_nxt = IPPROTO_TCP;
1882 h6->ip6_plen = htons(tlen);
1883 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr));
1884 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr));
1885
1886 th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr));
1887 break;
1888#endif /* INET6 */
1889 }
1890
1891 /* TCP header */
1892 th->th_sport = sport;
1893 th->th_dport = dport;
1894 th->th_seq = htonl(seq);
1895 th->th_ack = htonl(ack);
1896 th->th_off = tlen >> 2;
1897 th->th_flags = flags;
1898 th->th_win = htons(win);
1899
1900 if (mss) {
1901 opt = (char *)(th + 1);
1902 opt[0] = TCPOPT_MAXSEG;
1903 opt[1] = 4;
1904 HTONS(mss);
1905 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2);
1906 }
1907
1908 switch (af) {
1909#ifdef INET
1910 case AF_INET:
1911 /* TCP checksum */
1912 th->th_sum = in_cksum(m, len);
1913
1914 /* Finish the IP header */
1915 h->ip_v = 4;
1916 h->ip_hl = sizeof(*h) >> 2;
1917 h->ip_tos = IPTOS_LOWDELAY;
1918#ifdef __FreeBSD__
1919 h->ip_off = V_path_mtu_discovery ? IP_DF : 0;
1920 h->ip_len = len;
1921#else
1922 h->ip_off = htons(ip_mtudisc ? IP_DF : 0);
1923 h->ip_len = htons(len);
1924#endif
1925 h->ip_ttl = ttl ? ttl : V_ip_defttl;
1926 h->ip_sum = 0;
1927 if (eh == NULL) {
1928#ifdef __FreeBSD__
1929 PF_UNLOCK();
1930 ip_output(m, (void *)NULL, (void *)NULL, 0,
1931 (void *)NULL, (void *)NULL);
1932 PF_LOCK();
1933#else /* ! __FreeBSD__ */
1934 ip_output(m, (void *)NULL, (void *)NULL, 0,
1935 (void *)NULL, (void *)NULL);
1936#endif
1937 } else {
1938 struct route ro;
1939 struct rtentry rt;
1940 struct ether_header *e = (void *)ro.ro_dst.sa_data;
1941
1942 if (ifp == NULL) {
1943 m_freem(m);
1944 return;
1945 }
1946 rt.rt_ifp = ifp;
1947 ro.ro_rt = &rt;
1948 ro.ro_dst.sa_len = sizeof(ro.ro_dst);
1949 ro.ro_dst.sa_family = pseudo_AF_HDRCMPLT;
1950 bcopy(eh->ether_dhost, e->ether_shost, ETHER_ADDR_LEN);
1951 bcopy(eh->ether_shost, e->ether_dhost, ETHER_ADDR_LEN);
1952 e->ether_type = eh->ether_type;
1953#ifdef __FreeBSD__
1954 PF_UNLOCK();
1955 /* XXX_IMPORT: later */
1956 ip_output(m, (void *)NULL, &ro, 0,
1957 (void *)NULL, (void *)NULL);
1958 PF_LOCK();
1959#else /* ! __FreeBSD__ */
1960 ip_output(m, (void *)NULL, &ro, IP_ROUTETOETHER,
1961 (void *)NULL, (void *)NULL);
1962#endif
1963 }
1964 break;
1965#endif /* INET */
1966#ifdef INET6
1967 case AF_INET6:
1968 /* TCP checksum */
1969 th->th_sum = in6_cksum(m, IPPROTO_TCP,
1970 sizeof(struct ip6_hdr), tlen);
1971
1972 h6->ip6_vfc |= IPV6_VERSION;
1973 h6->ip6_hlim = IPV6_DEFHLIM;
1974
1975#ifdef __FreeBSD__
1976 PF_UNLOCK();
1977 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
1978 PF_LOCK();
1979#else
1980 ip6_output(m, NULL, NULL, 0, NULL, NULL);
1981#endif
1982 break;
1983#endif /* INET6 */
1984 }
1985}
1986
1987void
1988pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af,
1989 struct pf_rule *r)
1990{
1991 struct pf_mtag *pf_mtag;
1992 struct mbuf *m0;
1993#ifdef __FreeBSD__
1994 struct ip *ip;
1995#endif
1996
1997#ifdef __FreeBSD__
1998 m0 = m_copypacket(m, M_DONTWAIT);
1999 if (m0 == NULL)
2000 return;
2001#else
2002 m0 = m_copy(m, 0, M_COPYALL);
2003#endif
2004 if ((pf_mtag = pf_get_mtag(m0)) == NULL)
2005 return;
2006#ifdef __FreeBSD__
2007 /* XXX: revisit */
2008 m0->m_flags |= M_SKIP_FIREWALL;
2009#else
2010 pf_mtag->flags |= PF_TAG_GENERATED;
2011#endif
2012
2013 if (r->rtableid >= 0)
2014#ifdef __FreeBSD__
2015 {
2016 M_SETFIB(m0, r->rtableid);
2017#endif
2018 pf_mtag->rtableid = r->rtableid;
2019#ifdef __FreeBSD__
2020 }
2021#endif
2022
2023#ifdef ALTQ
2024 if (r->qid) {
2025 pf_mtag->qid = r->qid;
2026 /* add hints for ecn */
2027 pf_mtag->af = af;
2028 pf_mtag->hdr = mtod(m0, struct ip *);
2029 }
2030#endif /* ALTQ */
2031
2032 switch (af) {
2033#ifdef INET
2034 case AF_INET:
2035#ifdef __FreeBSD__
2036 /* icmp_error() expects host byte ordering */
2037 ip = mtod(m0, struct ip *);
2038 NTOHS(ip->ip_len);
2039 NTOHS(ip->ip_off);
2040 PF_UNLOCK();
2041 icmp_error(m0, type, code, 0, 0);
2042 PF_LOCK();
2043#else
2044 icmp_error(m0, type, code, 0, 0);
2045#endif
2046 break;
2047#endif /* INET */
2048#ifdef INET6
2049 case AF_INET6:
2050#ifdef __FreeBSD__
2051 PF_UNLOCK();
2052#endif
2053 icmp6_error(m0, type, code, 0);
2054#ifdef __FreeBSD__
2055 PF_LOCK();
2056#endif
2057 break;
2058#endif /* INET6 */
2059 }
2060}
2061
2062/*
2063 * Return 1 if the addresses a and b match (with mask m), otherwise return 0.
2064 * If n is 0, they match if they are equal. If n is != 0, they match if they
2065 * are different.
2066 */
2067int
2068pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m,
2069 struct pf_addr *b, sa_family_t af)
2070{
2071 int match = 0;
2072
2073 switch (af) {
2074#ifdef INET
2075 case AF_INET:
2076 if ((a->addr32[0] & m->addr32[0]) ==
2077 (b->addr32[0] & m->addr32[0]))
2078 match++;
2079 break;
2080#endif /* INET */
2081#ifdef INET6
2082 case AF_INET6:
2083 if (((a->addr32[0] & m->addr32[0]) ==
2084 (b->addr32[0] & m->addr32[0])) &&
2085 ((a->addr32[1] & m->addr32[1]) ==
2086 (b->addr32[1] & m->addr32[1])) &&
2087 ((a->addr32[2] & m->addr32[2]) ==
2088 (b->addr32[2] & m->addr32[2])) &&
2089 ((a->addr32[3] & m->addr32[3]) ==
2090 (b->addr32[3] & m->addr32[3])))
2091 match++;
2092 break;
2093#endif /* INET6 */
2094 }
2095 if (match) {
2096 if (n)
2097 return (0);
2098 else
2099 return (1);
2100 } else {
2101 if (n)
2102 return (1);
2103 else
2104 return (0);
2105 }
2106}
2107
2108int
2109pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p)
2110{
2111 switch (op) {
2112 case PF_OP_IRG:
2113 return ((p > a1) && (p < a2));
2114 case PF_OP_XRG:
2115 return ((p < a1) || (p > a2));
2116 case PF_OP_RRG:
2117 return ((p >= a1) && (p <= a2));
2118 case PF_OP_EQ:
2119 return (p == a1);
2120 case PF_OP_NE:
2121 return (p != a1);
2122 case PF_OP_LT:
2123 return (p < a1);
2124 case PF_OP_LE:
2125 return (p <= a1);
2126 case PF_OP_GT:
2127 return (p > a1);
2128 case PF_OP_GE:
2129 return (p >= a1);
2130 }
2131 return (0); /* never reached */
2132}
2133
2134int
2135pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p)
2136{
2137 NTOHS(a1);
2138 NTOHS(a2);
2139 NTOHS(p);
2140 return (pf_match(op, a1, a2, p));
2141}
2142
2143int
2144pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u)
2145{
2146 if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
2147 return (0);
2148 return (pf_match(op, a1, a2, u));
2149}
2150
2151int
2152pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g)
2153{
2154 if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
2155 return (0);
2156 return (pf_match(op, a1, a2, g));
2157}
2158
2159#ifndef __FreeBSD__
2160struct pf_mtag *
2161pf_find_mtag(struct mbuf *m)
2162{
2163 struct m_tag *mtag;
2164
2165 if ((mtag = m_tag_find(m, PACKET_TAG_PF, NULL)) == NULL)
2166 return (NULL);
2167
2168 return ((struct pf_mtag *)(mtag + 1));
2169}
2170
2171struct pf_mtag *
2172pf_get_mtag(struct mbuf *m)
2173{
2174 struct m_tag *mtag;
2175
2176 if ((mtag = m_tag_find(m, PACKET_TAG_PF, NULL)) == NULL) {
2177 mtag = m_tag_get(PACKET_TAG_PF, sizeof(struct pf_mtag),
2178 M_NOWAIT);
2179 if (mtag == NULL)
2180 return (NULL);
2181 bzero(mtag + 1, sizeof(struct pf_mtag));
2182 m_tag_prepend(m, mtag);
2183 }
2184
2185 return ((struct pf_mtag *)(mtag + 1));
2186}
2187#endif
2188
2189int
2190pf_match_tag(struct mbuf *m, struct pf_rule *r, struct pf_mtag *pf_mtag,
2191 int *tag)
2192{
2193 if (*tag == -1)
2194 *tag = pf_mtag->tag;
2195
2196 return ((!r->match_tag_not && r->match_tag == *tag) ||
2197 (r->match_tag_not && r->match_tag != *tag));
2198}
2199
2200int
2201pf_tag_packet(struct mbuf *m, struct pf_mtag *pf_mtag, int tag, int rtableid)
2202{
2203 if (tag <= 0 && rtableid < 0)
2204 return (0);
2205
2206 if (pf_mtag == NULL)
2207 if ((pf_mtag = pf_get_mtag(m)) == NULL)
2208 return (1);
2209 if (tag > 0)
2210 pf_mtag->tag = tag;
2211 if (rtableid >= 0)
2212#ifdef __FreeBSD__
2213 {
2214 M_SETFIB(m, rtableid);
2215#endif
2216 pf_mtag->rtableid = rtableid;
2217#ifdef __FreeBSD__
2218 }
2219#endif
2220
2221 return (0);
2222}
2223
2224static void
2225pf_step_into_anchor(int *depth, struct pf_ruleset **rs, int n,
2226 struct pf_rule **r, struct pf_rule **a, int *match)
2227{
2228 struct pf_anchor_stackframe *f;
2229
2230 (*r)->anchor->match = 0;
2231 if (match)
2232 *match = 0;
2233 if (*depth >= sizeof(pf_anchor_stack) /
2234 sizeof(pf_anchor_stack[0])) {
2235 printf("pf_step_into_anchor: stack overflow\n");
2236 *r = TAILQ_NEXT(*r, entries);
2237 return;
2238 } else if (*depth == 0 && a != NULL)
2239 *a = *r;
2240 f = pf_anchor_stack + (*depth)++;
2241 f->rs = *rs;
2242 f->r = *r;
2243 if ((*r)->anchor_wildcard) {
2244 f->parent = &(*r)->anchor->children;
2245 if ((f->child = RB_MIN(pf_anchor_node, f->parent)) ==
2246 NULL) {
2247 *r = NULL;
2248 return;
2249 }
2250 *rs = &f->child->ruleset;
2251 } else {
2252 f->parent = NULL;
2253 f->child = NULL;
2254 *rs = &(*r)->anchor->ruleset;
2255 }
2256 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
2257}
2258
2259int
2260pf_step_out_of_anchor(int *depth, struct pf_ruleset **rs, int n,
2261 struct pf_rule **r, struct pf_rule **a, int *match)
2262{
2263 struct pf_anchor_stackframe *f;
2264 int quick = 0;
2265
2266 do {
2267 if (*depth <= 0)
2268 break;
2269 f = pf_anchor_stack + *depth - 1;
2270 if (f->parent != NULL && f->child != NULL) {
2271 if (f->child->match ||
2272 (match != NULL && *match)) {
2273 f->r->anchor->match = 1;
2274 *match = 0;
2275 }
2276 f->child = RB_NEXT(pf_anchor_node, f->parent, f->child);
2277 if (f->child != NULL) {
2278 *rs = &f->child->ruleset;
2279 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
2280 if (*r == NULL)
2281 continue;
2282 else
2283 break;
2284 }
2285 }
2286 (*depth)--;
2287 if (*depth == 0 && a != NULL)
2288 *a = NULL;
2289 *rs = f->rs;
2290 if (f->r->anchor->match || (match != NULL && *match))
2291 quick = f->r->quick;
2292 *r = TAILQ_NEXT(f->r, entries);
2293 } while (*r == NULL);
2294
2295 return (quick);
2296}
2297
2298#ifdef INET6
2299void
2300pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr,
2301 struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af)
2302{
2303 switch (af) {
2304#ifdef INET
2305 case AF_INET:
2306 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2307 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2308 break;
2309#endif /* INET */
2310 case AF_INET6:
2311 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2312 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2313 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) |
2314 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]);
2315 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) |
2316 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]);
2317 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) |
2318 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]);
2319 break;
2320 }
2321}
2322
2323void
2324pf_addr_inc(struct pf_addr *addr, sa_family_t af)
2325{
2326 switch (af) {
2327#ifdef INET
2328 case AF_INET:
2329 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1);
2330 break;
2331#endif /* INET */
2332 case AF_INET6:
2333 if (addr->addr32[3] == 0xffffffff) {
2334 addr->addr32[3] = 0;
2335 if (addr->addr32[2] == 0xffffffff) {
2336 addr->addr32[2] = 0;
2337 if (addr->addr32[1] == 0xffffffff) {
2338 addr->addr32[1] = 0;
2339 addr->addr32[0] =
2340 htonl(ntohl(addr->addr32[0]) + 1);
2341 } else
2342 addr->addr32[1] =
2343 htonl(ntohl(addr->addr32[1]) + 1);
2344 } else
2345 addr->addr32[2] =
2346 htonl(ntohl(addr->addr32[2]) + 1);
2347 } else
2348 addr->addr32[3] =
2349 htonl(ntohl(addr->addr32[3]) + 1);
2350 break;
2351 }
2352}
2353#endif /* INET6 */
2354
2355#define mix(a,b,c) \
2356 do { \
2357 a -= b; a -= c; a ^= (c >> 13); \
2358 b -= c; b -= a; b ^= (a << 8); \
2359 c -= a; c -= b; c ^= (b >> 13); \
2360 a -= b; a -= c; a ^= (c >> 12); \
2361 b -= c; b -= a; b ^= (a << 16); \
2362 c -= a; c -= b; c ^= (b >> 5); \
2363 a -= b; a -= c; a ^= (c >> 3); \
2364 b -= c; b -= a; b ^= (a << 10); \
2365 c -= a; c -= b; c ^= (b >> 15); \
2366 } while (0)
2367
2368/*
2369 * hash function based on bridge_hash in if_bridge.c
2370 */
2371void
2372pf_hash(struct pf_addr *inaddr, struct pf_addr *hash,
2373 struct pf_poolhashkey *key, sa_family_t af)
2374{
2375 u_int32_t a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0];
2376
2377 switch (af) {
2378#ifdef INET
2379 case AF_INET:
2380 a += inaddr->addr32[0];
2381 b += key->key32[1];
2382 mix(a, b, c);
2383 hash->addr32[0] = c + key->key32[2];
2384 break;
2385#endif /* INET */
2386#ifdef INET6
2387 case AF_INET6:
2388 a += inaddr->addr32[0];
2389 b += inaddr->addr32[2];
2390 mix(a, b, c);
2391 hash->addr32[0] = c;
2392 a += inaddr->addr32[1];
2393 b += inaddr->addr32[3];
2394 c += key->key32[1];
2395 mix(a, b, c);
2396 hash->addr32[1] = c;
2397 a += inaddr->addr32[2];
2398 b += inaddr->addr32[1];
2399 c += key->key32[2];
2400 mix(a, b, c);
2401 hash->addr32[2] = c;
2402 a += inaddr->addr32[3];
2403 b += inaddr->addr32[0];
2404 c += key->key32[3];
2405 mix(a, b, c);
2406 hash->addr32[3] = c;
2407 break;
2408#endif /* INET6 */
2409 }
2410}
2411
2412int
2413pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr,
2414 struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn)
2415{
2416 unsigned char hash[16];
2417 struct pf_pool *rpool = &r->rpool;
2418 struct pf_addr *raddr = &rpool->cur->addr.v.a.addr;
2419 struct pf_addr *rmask = &rpool->cur->addr.v.a.mask;
2420 struct pf_pooladdr *acur = rpool->cur;
2421 struct pf_src_node k;
2422
2423 if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR &&
2424 (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2425 k.af = af;
2426 PF_ACPY(&k.addr, saddr, af);
2427 if (r->rule_flag & PFRULE_RULESRCTRACK ||
2428 r->rpool.opts & PF_POOL_STICKYADDR)
2429 k.rule.ptr = r;
2430 else
2431 k.rule.ptr = NULL;
2432 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
2433 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
2434 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) {
2435 PF_ACPY(naddr, &(*sn)->raddr, af);
2436 if (pf_status.debug >= PF_DEBUG_MISC) {
2437 printf("pf_map_addr: src tracking maps ");
2438 pf_print_host(&k.addr, 0, af);
2439 printf(" to ");
2440 pf_print_host(naddr, 0, af);
2441 printf("\n");
2442 }
2443 return (0);
2444 }
2445 }
2446
2447 if (rpool->cur->addr.type == PF_ADDR_NOROUTE)
2448 return (1);
2449 if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2450 switch (af) {
2451#ifdef INET
2452 case AF_INET:
2453 if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 &&
2454 (rpool->opts & PF_POOL_TYPEMASK) !=
2455 PF_POOL_ROUNDROBIN)
2456 return (1);
2457 raddr = &rpool->cur->addr.p.dyn->pfid_addr4;
2458 rmask = &rpool->cur->addr.p.dyn->pfid_mask4;
2459 break;
2460#endif /* INET */
2461#ifdef INET6
2462 case AF_INET6:
2463 if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 &&
2464 (rpool->opts & PF_POOL_TYPEMASK) !=
2465 PF_POOL_ROUNDROBIN)
2466 return (1);
2467 raddr = &rpool->cur->addr.p.dyn->pfid_addr6;
2468 rmask = &rpool->cur->addr.p.dyn->pfid_mask6;
2469 break;
2470#endif /* INET6 */
2471 }
2472 } else if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2473 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN)
2474 return (1); /* unsupported */
2475 } else {
2476 raddr = &rpool->cur->addr.v.a.addr;
2477 rmask = &rpool->cur->addr.v.a.mask;
2478 }
2479
2480 switch (rpool->opts & PF_POOL_TYPEMASK) {
2481 case PF_POOL_NONE:
2482 PF_ACPY(naddr, raddr, af);
2483 break;
2484 case PF_POOL_BITMASK:
2485 PF_POOLMASK(naddr, raddr, rmask, saddr, af);
2486 break;
2487 case PF_POOL_RANDOM:
2488 if (init_addr != NULL && PF_AZERO(init_addr, af)) {
2489 switch (af) {
2490#ifdef INET
2491 case AF_INET:
2492 rpool->counter.addr32[0] = htonl(arc4random());
2493 break;
2494#endif /* INET */
2495#ifdef INET6
2496 case AF_INET6:
2497 if (rmask->addr32[3] != 0xffffffff)
2498 rpool->counter.addr32[3] =
2499 htonl(arc4random());
2500 else
2501 break;
2502 if (rmask->addr32[2] != 0xffffffff)
2503 rpool->counter.addr32[2] =
2504 htonl(arc4random());
2505 else
2506 break;
2507 if (rmask->addr32[1] != 0xffffffff)
2508 rpool->counter.addr32[1] =
2509 htonl(arc4random());
2510 else
2511 break;
2512 if (rmask->addr32[0] != 0xffffffff)
2513 rpool->counter.addr32[0] =
2514 htonl(arc4random());
2515 break;
2516#endif /* INET6 */
2517 }
2518 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2519 PF_ACPY(init_addr, naddr, af);
2520
2521 } else {
2522 PF_AINC(&rpool->counter, af);
2523 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2524 }
2525 break;
2526 case PF_POOL_SRCHASH:
2527 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af);
2528 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af);
2529 break;
2530 case PF_POOL_ROUNDROBIN:
2531 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2532 if (!pfr_pool_get(rpool->cur->addr.p.tbl,
2533 &rpool->tblidx, &rpool->counter,
2534 &raddr, &rmask, af))
2535 goto get_addr;
2536 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2537 if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2538 &rpool->tblidx, &rpool->counter,
2539 &raddr, &rmask, af))
2540 goto get_addr;
2541 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af))
2542 goto get_addr;
2543
2544 try_next:
2545 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL)
2546 rpool->cur = TAILQ_FIRST(&rpool->list);
2547 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2548 rpool->tblidx = -1;
2549 if (pfr_pool_get(rpool->cur->addr.p.tbl,
2550 &rpool->tblidx, &rpool->counter,
2551 &raddr, &rmask, af)) {
2552 /* table contains no address of type 'af' */
2553 if (rpool->cur != acur)
2554 goto try_next;
2555 return (1);
2556 }
2557 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2558 rpool->tblidx = -1;
2559 if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2560 &rpool->tblidx, &rpool->counter,
2561 &raddr, &rmask, af)) {
2562 /* table contains no address of type 'af' */
2563 if (rpool->cur != acur)
2564 goto try_next;
2565 return (1);
2566 }
2567 } else {
2568 raddr = &rpool->cur->addr.v.a.addr;
2569 rmask = &rpool->cur->addr.v.a.mask;
2570 PF_ACPY(&rpool->counter, raddr, af);
2571 }
2572
2573 get_addr:
2574 PF_ACPY(naddr, &rpool->counter, af);
2575 if (init_addr != NULL && PF_AZERO(init_addr, af))
2576 PF_ACPY(init_addr, naddr, af);
2577 PF_AINC(&rpool->counter, af);
2578 break;
2579 }
2580 if (*sn != NULL)
2581 PF_ACPY(&(*sn)->raddr, naddr, af);
2582
2583 if (pf_status.debug >= PF_DEBUG_MISC &&
2584 (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2585 printf("pf_map_addr: selected address ");
2586 pf_print_host(naddr, 0, af);
2587 printf("\n");
2588 }
2589
2590 return (0);
2591}
2592
2593int
2594pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r,
2595 struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport,
2596 struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high,
2597 struct pf_src_node **sn)
2598{
2599 struct pf_state_cmp key;
2600 struct pf_addr init_addr;
2601 u_int16_t cut;
2602
2603 bzero(&init_addr, sizeof(init_addr));
2604 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2605 return (1);
2606
2607 if (proto == IPPROTO_ICMP) {
2608 low = 1;
2609 high = 65535;
2610 }
2611
2612 do {
2613 key.af = af;
2614 key.proto = proto;
2615 PF_ACPY(&key.ext.addr, daddr, key.af);
2616 PF_ACPY(&key.gwy.addr, naddr, key.af);
2617 key.ext.port = dport;
2618
2619 /*
2620 * port search; start random, step;
2621 * similar 2 portloop in in_pcbbind
2622 */
2623 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP ||
2624 proto == IPPROTO_ICMP)) {
2625 key.gwy.port = dport;
2626 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
2627 return (0);
2628 } else if (low == 0 && high == 0) {
2629 key.gwy.port = *nport;
2630 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
2631 return (0);
2632 } else if (low == high) {
2633 key.gwy.port = htons(low);
2634 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) {
2635 *nport = htons(low);
2636 return (0);
2637 }
2638 } else {
2639 u_int16_t tmp;
2640
2641 if (low > high) {
2642 tmp = low;
2643 low = high;
2644 high = tmp;
2645 }
2646 /* low < high */
2647 cut = htonl(arc4random()) % (1 + high - low) + low;
2648 /* low <= cut <= high */
2649 for (tmp = cut; tmp <= high; ++(tmp)) {
2650 key.gwy.port = htons(tmp);
2651 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
2652 NULL) {
2653 *nport = htons(tmp);
2654 return (0);
2655 }
2656 }
2657 for (tmp = cut - 1; tmp >= low; --(tmp)) {
2658 key.gwy.port = htons(tmp);
2659 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
2660 NULL) {
2661 *nport = htons(tmp);
2662 return (0);
2663 }
2664 }
2665 }
2666
2667 switch (r->rpool.opts & PF_POOL_TYPEMASK) {
2668 case PF_POOL_RANDOM:
2669 case PF_POOL_ROUNDROBIN:
2670 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2671 return (1);
2672 break;
2673 case PF_POOL_NONE:
2674 case PF_POOL_SRCHASH:
2675 case PF_POOL_BITMASK:
2676 default:
2677 return (1);
2678 }
2679 } while (! PF_AEQ(&init_addr, naddr, af) );
2680
2681 return (1); /* none available */
2682}
2683
2684struct pf_rule *
2685pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off,
2686 int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport,
2687 struct pf_addr *daddr, u_int16_t dport, int rs_num)
2688{
2689 struct pf_rule *r, *rm = NULL;
2690 struct pf_ruleset *ruleset = NULL;
2691 int tag = -1;
2692 int rtableid = -1;
2693 int asd = 0;
2694
2695 r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr);
2696 while (r && rm == NULL) {
2697 struct pf_rule_addr *src = NULL, *dst = NULL;
2698 struct pf_addr_wrap *xdst = NULL;
2699
2700 if (r->action == PF_BINAT && direction == PF_IN) {
2701 src = &r->dst;
2702 if (r->rpool.cur != NULL)
2703 xdst = &r->rpool.cur->addr;
2704 } else {
2705 src = &r->src;
2706 dst = &r->dst;
2707 }
2708
2709 r->evaluations++;
2710 if (pfi_kif_match(r->kif, kif) == r->ifnot)
2711 r = r->skip[PF_SKIP_IFP].ptr;
2712 else if (r->direction && r->direction != direction)
2713 r = r->skip[PF_SKIP_DIR].ptr;
2714 else if (r->af && r->af != pd->af)
2715 r = r->skip[PF_SKIP_AF].ptr;
2716 else if (r->proto && r->proto != pd->proto)
2717 r = r->skip[PF_SKIP_PROTO].ptr;
2718 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af,
2719 src->neg, kif))
2720 r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR :
2721 PF_SKIP_DST_ADDR].ptr;
2722 else if (src->port_op && !pf_match_port(src->port_op,
2723 src->port[0], src->port[1], sport))
2724 r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT :
2725 PF_SKIP_DST_PORT].ptr;
2726 else if (dst != NULL &&
2727 PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->neg, NULL))
2728 r = r->skip[PF_SKIP_DST_ADDR].ptr;
2729 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af,
2730 0, NULL))
2731 r = TAILQ_NEXT(r, entries);
2732 else if (dst != NULL && dst->port_op &&
2733 !pf_match_port(dst->port_op, dst->port[0],
2734 dst->port[1], dport))
2735 r = r->skip[PF_SKIP_DST_PORT].ptr;
2736 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag))
2737 r = TAILQ_NEXT(r, entries);
2738 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto !=
2739 IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m,
2740 off, pd->hdr.tcp), r->os_fingerprint)))
2741 r = TAILQ_NEXT(r, entries);
2742 else {
2743 if (r->tag)
2744 tag = r->tag;
2745 if (r->rtableid >= 0)
2746 rtableid = r->rtableid;
2747 if (r->anchor == NULL) {
2748 rm = r;
2749 } else
2750 pf_step_into_anchor(&asd, &ruleset, rs_num,
2751 &r, NULL, NULL);
2752 }
2753 if (r == NULL)
2754 pf_step_out_of_anchor(&asd, &ruleset, rs_num, &r,
2755 NULL, NULL);
2756 }
2757 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid))
2758 return (NULL);
2759 if (rm != NULL && (rm->action == PF_NONAT ||
2760 rm->action == PF_NORDR || rm->action == PF_NOBINAT))
2761 return (NULL);
2762 return (rm);
2763}
2764
2765struct pf_rule *
2766pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction,
2767 struct pfi_kif *kif, struct pf_src_node **sn,
2768 struct pf_addr *saddr, u_int16_t sport,
2769 struct pf_addr *daddr, u_int16_t dport,
2770 struct pf_addr *naddr, u_int16_t *nport)
2771{
2772 struct pf_rule *r = NULL;
2773
2774 if (direction == PF_OUT) {
2775 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2776 sport, daddr, dport, PF_RULESET_BINAT);
2777 if (r == NULL)
2778 r = pf_match_translation(pd, m, off, direction, kif,
2779 saddr, sport, daddr, dport, PF_RULESET_NAT);
2780 } else {
2781 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2782 sport, daddr, dport, PF_RULESET_RDR);
2783 if (r == NULL)
2784 r = pf_match_translation(pd, m, off, direction, kif,
2785 saddr, sport, daddr, dport, PF_RULESET_BINAT);
2786 }
2787
2788 if (r != NULL) {
2789 switch (r->action) {
2790 case PF_NONAT:
2791 case PF_NOBINAT:
2792 case PF_NORDR:
2793 return (NULL);
2794 case PF_NAT:
2795 if (pf_get_sport(pd->af, pd->proto, r, saddr,
2796 daddr, dport, naddr, nport, r->rpool.proxy_port[0],
2797 r->rpool.proxy_port[1], sn)) {
2798 DPFPRINTF(PF_DEBUG_MISC,
2799 ("pf: NAT proxy port allocation "
2800 "(%u-%u) failed\n",
2801 r->rpool.proxy_port[0],
2802 r->rpool.proxy_port[1]));
2803 return (NULL);
2804 }
2805 break;
2806 case PF_BINAT:
2807 switch (direction) {
2808 case PF_OUT:
2809 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){
2810 switch (pd->af) {
2811#ifdef INET
2812 case AF_INET:
2813 if (r->rpool.cur->addr.p.dyn->
2814 pfid_acnt4 < 1)
2815 return (NULL);
2816 PF_POOLMASK(naddr,
2817 &r->rpool.cur->addr.p.dyn->
2818 pfid_addr4,
2819 &r->rpool.cur->addr.p.dyn->
2820 pfid_mask4,
2821 saddr, AF_INET);
2822 break;
2823#endif /* INET */
2824#ifdef INET6
2825 case AF_INET6:
2826 if (r->rpool.cur->addr.p.dyn->
2827 pfid_acnt6 < 1)
2828 return (NULL);
2829 PF_POOLMASK(naddr,
2830 &r->rpool.cur->addr.p.dyn->
2831 pfid_addr6,
2832 &r->rpool.cur->addr.p.dyn->
2833 pfid_mask6,
2834 saddr, AF_INET6);
2835 break;
2836#endif /* INET6 */
2837 }
2838 } else
2839 PF_POOLMASK(naddr,
2840 &r->rpool.cur->addr.v.a.addr,
2841 &r->rpool.cur->addr.v.a.mask,
2842 saddr, pd->af);
2843 break;
2844 case PF_IN:
2845 if (r->src.addr.type == PF_ADDR_DYNIFTL) {
2846 switch (pd->af) {
2847#ifdef INET
2848 case AF_INET:
2849 if (r->src.addr.p.dyn->
2850 pfid_acnt4 < 1)
2851 return (NULL);
2852 PF_POOLMASK(naddr,
2853 &r->src.addr.p.dyn->
2854 pfid_addr4,
2855 &r->src.addr.p.dyn->
2856 pfid_mask4,
2857 daddr, AF_INET);
2858 break;
2859#endif /* INET */
2860#ifdef INET6
2861 case AF_INET6:
2862 if (r->src.addr.p.dyn->
2863 pfid_acnt6 < 1)
2864 return (NULL);
2865 PF_POOLMASK(naddr,
2866 &r->src.addr.p.dyn->
2867 pfid_addr6,
2868 &r->src.addr.p.dyn->
2869 pfid_mask6,
2870 daddr, AF_INET6);
2871 break;
2872#endif /* INET6 */
2873 }
2874 } else
2875 PF_POOLMASK(naddr,
2876 &r->src.addr.v.a.addr,
2877 &r->src.addr.v.a.mask, daddr,
2878 pd->af);
2879 break;
2880 }
2881 break;
2882 case PF_RDR: {
2883 if (pf_map_addr(pd->af, r, saddr, naddr, NULL, sn))
2884 return (NULL);
2885 if ((r->rpool.opts & PF_POOL_TYPEMASK) ==
2886 PF_POOL_BITMASK)
2887 PF_POOLMASK(naddr, naddr,
2888 &r->rpool.cur->addr.v.a.mask, daddr,
2889 pd->af);
2890
2891 if (r->rpool.proxy_port[1]) {
2892 u_int32_t tmp_nport;
2893
2894 tmp_nport = ((ntohs(dport) -
2895 ntohs(r->dst.port[0])) %
2896 (r->rpool.proxy_port[1] -
2897 r->rpool.proxy_port[0] + 1)) +
2898 r->rpool.proxy_port[0];
2899
2900 /* wrap around if necessary */
2901 if (tmp_nport > 65535)
2902 tmp_nport -= 65535;
2903 *nport = htons((u_int16_t)tmp_nport);
2904 } else if (r->rpool.proxy_port[0])
2905 *nport = htons(r->rpool.proxy_port[0]);
2906 break;
2907 }
2908 default:
2909 return (NULL);
2910 }
2911 }
2912
2913 return (r);
2914}
2915
2916int
2917#ifdef __FreeBSD__
2918pf_socket_lookup(int direction, struct pf_pdesc *pd, struct inpcb *inp_arg)
2919#else
2920pf_socket_lookup(int direction, struct pf_pdesc *pd)
2921#endif
2922{
2923 struct pf_addr *saddr, *daddr;
2924 u_int16_t sport, dport;
2925#ifdef __FreeBSD__
2926 struct inpcbinfo *pi;
2927#else
2928 struct inpcbtable *tb;
2929#endif
2930 struct inpcb *inp;
2931
2932 if (pd == NULL)
2933 return (-1);
2934 pd->lookup.uid = UID_MAX;
2935 pd->lookup.gid = GID_MAX;
2936 pd->lookup.pid = NO_PID; /* XXX: revisit */
2937#ifdef __FreeBSD__
2938 if (inp_arg != NULL) {
2939 INP_LOCK_ASSERT(inp_arg);
2940 pd->lookup.uid = inp_arg->inp_cred->cr_uid;
2941 pd->lookup.gid = inp_arg->inp_cred->cr_groups[0];
2942 return (1);
2943 }
2944#endif
2945 switch (pd->proto) {
2946 case IPPROTO_TCP:
2947 if (pd->hdr.tcp == NULL)
2948 return (-1);
2949 sport = pd->hdr.tcp->th_sport;
2950 dport = pd->hdr.tcp->th_dport;
2951#ifdef __FreeBSD__
2952 pi = &V_tcbinfo;
2953#else
2954 tb = &tcbtable;
2955#endif
2956 break;
2957 case IPPROTO_UDP:
2958 if (pd->hdr.udp == NULL)
2959 return (-1);
2960 sport = pd->hdr.udp->uh_sport;
2961 dport = pd->hdr.udp->uh_dport;
2962#ifdef __FreeBSD__
2963 pi = &V_udbinfo;
2964#else
2965 tb = &udbtable;
2966#endif
2967 break;
2968 default:
2969 return (-1);
2970 }
2971 if (direction == PF_IN) {
2972 saddr = pd->src;
2973 daddr = pd->dst;
2974 } else {
2975 u_int16_t p;
2976
2977 p = sport;
2978 sport = dport;
2979 dport = p;
2980 saddr = pd->dst;
2981 daddr = pd->src;
2982 }
2983 switch (pd->af) {
2984#ifdef INET
2985 case AF_INET:
2986#ifdef __FreeBSD__
2987 INP_INFO_RLOCK(pi); /* XXX LOR */
2988 inp = in_pcblookup_hash(pi, saddr->v4, sport, daddr->v4,
2989 dport, 0, NULL);
2990 if (inp == NULL) {
2991 inp = in_pcblookup_hash(pi, saddr->v4, sport,
2992 daddr->v4, dport, INPLOOKUP_WILDCARD, NULL);
2993 if(inp == NULL) {
2994 INP_INFO_RUNLOCK(pi);
2995 return (-1);
2996 }
2997 }
2998#else
2999 inp = in_pcbhashlookup(tb, saddr->v4, sport, daddr->v4, dport);
3000 if (inp == NULL) {
3001 inp = in_pcblookup_listen(tb, daddr->v4, dport, 0);
3002 if (inp == NULL)
3003 return (-1);
3004 }
3005#endif
3006 break;
3007#endif /* INET */
3008#ifdef INET6
3009 case AF_INET6:
3010#ifdef __FreeBSD__
3011 INP_INFO_RLOCK(pi);
3012 inp = in6_pcblookup_hash(pi, &saddr->v6, sport,
3013 &daddr->v6, dport, 0, NULL);
3014 if (inp == NULL) {
3015 inp = in6_pcblookup_hash(pi, &saddr->v6, sport,
3016 &daddr->v6, dport, INPLOOKUP_WILDCARD, NULL);
3017 if (inp == NULL) {
3018 INP_INFO_RUNLOCK(pi);
3019 return (-1);
3020 }
3021 }
3022#else
3023 inp = in6_pcbhashlookup(tb, &saddr->v6, sport, &daddr->v6,
3024 dport);
3025 if (inp == NULL) {
3026 inp = in6_pcblookup_listen(tb, &daddr->v6, dport, 0);
3027 if (inp == NULL)
3028 return (-1);
3029 }
3030#endif
3031 break;
3032#endif /* INET6 */
3033
3034 default:
3035 return (-1);
3036 }
3037#ifdef __FreeBSD__
3038 pd->lookup.uid = inp->inp_cred->cr_uid;
3039 pd->lookup.gid = inp->inp_cred->cr_groups[0];
3040 INP_INFO_RUNLOCK(pi);
3041#else
3042 pd->lookup.uid = inp->inp_socket->so_euid;
3043 pd->lookup.gid = inp->inp_socket->so_egid;
3044 pd->lookup.pid = inp->inp_socket->so_cpid;
3045#endif
3046 return (1);
3047}
3048
3049u_int8_t
3050pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
3051{
3052 int hlen;
3053 u_int8_t hdr[60];
3054 u_int8_t *opt, optlen;
3055 u_int8_t wscale = 0;
3056
3057 hlen = th_off << 2; /* hlen <= sizeof(hdr) */
3058 if (hlen <= sizeof(struct tcphdr))
3059 return (0);
3060 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
3061 return (0);
3062 opt = hdr + sizeof(struct tcphdr);
3063 hlen -= sizeof(struct tcphdr);
3064 while (hlen >= 3) {
3065 switch (*opt) {
3066 case TCPOPT_EOL:
3067 case TCPOPT_NOP:
3068 ++opt;
3069 --hlen;
3070 break;
3071 case TCPOPT_WINDOW:
3072 wscale = opt[2];
3073 if (wscale > TCP_MAX_WINSHIFT)
3074 wscale = TCP_MAX_WINSHIFT;
3075 wscale |= PF_WSCALE_FLAG;
3076 /* FALLTHROUGH */
3077 default:
3078 optlen = opt[1];
3079 if (optlen < 2)
3080 optlen = 2;
3081 hlen -= optlen;
3082 opt += optlen;
3083 break;
3084 }
3085 }
3086 return (wscale);
3087}
3088
3089u_int16_t
3090pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
3091{
3092 int hlen;
3093 u_int8_t hdr[60];
3094 u_int8_t *opt, optlen;
3095 u_int16_t mss = V_tcp_mssdflt;
3096
3097 hlen = th_off << 2; /* hlen <= sizeof(hdr) */
3098 if (hlen <= sizeof(struct tcphdr))
3099 return (0);
3100 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
3101 return (0);
3102 opt = hdr + sizeof(struct tcphdr);
3103 hlen -= sizeof(struct tcphdr);
3104 while (hlen >= TCPOLEN_MAXSEG) {
3105 switch (*opt) {
3106 case TCPOPT_EOL:
3107 case TCPOPT_NOP:
3108 ++opt;
3109 --hlen;
3110 break;
3111 case TCPOPT_MAXSEG:
3112 bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2);
3113 NTOHS(mss);
3114 /* FALLTHROUGH */
3115 default:
3116 optlen = opt[1];
3117 if (optlen < 2)
3118 optlen = 2;
3119 hlen -= optlen;
3120 opt += optlen;
3121 break;
3122 }
3123 }
3124 return (mss);
3125}
3126
3127u_int16_t
3128pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer)
3129{
3130#ifdef INET
3131 struct sockaddr_in *dst;
3132 struct route ro;
3133#endif /* INET */
3134#ifdef INET6
3135 struct sockaddr_in6 *dst6;
3136 struct route_in6 ro6;
3137#endif /* INET6 */
3138 struct rtentry *rt = NULL;
3139 int hlen = 0; /* make the compiler happy */
3140 u_int16_t mss = V_tcp_mssdflt;
3141
3142 switch (af) {
3143#ifdef INET
3144 case AF_INET:
3145 hlen = sizeof(struct ip);
3146 bzero(&ro, sizeof(ro));
3147 dst = (struct sockaddr_in *)&ro.ro_dst;
3148 dst->sin_family = AF_INET;
3149 dst->sin_len = sizeof(*dst);
3150 dst->sin_addr = addr->v4;
3151#ifdef __FreeBSD__
3152#ifdef RTF_PRCLONING
3153 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
3154#else /* !RTF_PRCLONING */
3155 in_rtalloc_ign(&ro, 0, 0);
3156#endif
3157#else /* ! __FreeBSD__ */
3158 rtalloc_noclone(&ro, NO_CLONING);
3159#endif
3160 rt = ro.ro_rt;
3161 break;
3162#endif /* INET */
3163#ifdef INET6
3164 case AF_INET6:
3165 hlen = sizeof(struct ip6_hdr);
3166 bzero(&ro6, sizeof(ro6));
3167 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst;
3168 dst6->sin6_family = AF_INET6;
3169 dst6->sin6_len = sizeof(*dst6);
3170 dst6->sin6_addr = addr->v6;
3171#ifdef __FreeBSD__
3172#ifdef RTF_PRCLONING
3173 rtalloc_ign((struct route *)&ro6,
3174 (RTF_CLONING | RTF_PRCLONING));
3175#else /* !RTF_PRCLONING */
3176 rtalloc_ign((struct route *)&ro6, 0);
3177#endif
3178#else /* ! __FreeBSD__ */
3179 rtalloc_noclone((struct route *)&ro6, NO_CLONING);
3180#endif
3181 rt = ro6.ro_rt;
3182 break;
3183#endif /* INET6 */
3184 }
3185
3186 if (rt && rt->rt_ifp) {
3187 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr);
3188 mss = max(V_tcp_mssdflt, mss);
3189 RTFREE(rt);
3190 }
3191 mss = min(mss, offer);
3192 mss = max(mss, 64); /* sanity - at least max opt space */
3193 return (mss);
3194}
3195
3196void
3197pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr)
3198{
3199 struct pf_rule *r = s->rule.ptr;
3200
3201 s->rt_kif = NULL;
3202 if (!r->rt || r->rt == PF_FASTROUTE)
3203 return;
3204 switch (s->af) {
3205#ifdef INET
3206 case AF_INET:
3207 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL,
3208 &s->nat_src_node);
3209 s->rt_kif = r->rpool.cur->kif;
3210 break;
3211#endif /* INET */
3212#ifdef INET6
3213 case AF_INET6:
3214 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL,
3215 &s->nat_src_node);
3216 s->rt_kif = r->rpool.cur->kif;
3217 break;
3218#endif /* INET6 */
3219 }
3220}
3221
3222int
3223pf_test_tcp(struct pf_rule **rm, struct pf_state **sm, int direction,
3224 struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3225#ifdef __FreeBSD__
3226 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
3227 struct ifqueue *ifq, struct inpcb *inp)
3228#else
3229 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
3230 struct ifqueue *ifq)
3231#endif
3232{
3233 struct pf_rule *nr = NULL;
3234 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
3235 struct tcphdr *th = pd->hdr.tcp;
3236 u_int16_t bport, nport = 0;
3237 sa_family_t af = pd->af;
3238 struct pf_rule *r, *a = NULL;
3239 struct pf_ruleset *ruleset = NULL;
3240 struct pf_src_node *nsn = NULL;
3241 u_short reason;
3242 int rewrite = 0;
3243 int tag = -1, rtableid = -1;
3244 u_int16_t mss = V_tcp_mssdflt;
3245 int asd = 0;
3246 int match = 0;
3247
3248 if (pf_check_congestion(ifq)) {
3249 REASON_SET(&reason, PFRES_CONGEST);
3250 return (PF_DROP);
3251 }
3252
3253#ifdef __FreeBSD__
3254 if (inp != NULL)
3255 pd->lookup.done = pf_socket_lookup(direction, pd, inp);
3256 else if (debug_pfugidhack) {
3257 PF_UNLOCK();
3258 DPFPRINTF(PF_DEBUG_MISC, ("pf: unlocked lookup\n"));
3259 pd->lookup.done = pf_socket_lookup(direction, pd, inp);
3260 PF_LOCK();
3261 }
3262#endif
3263
3264 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3265
3266 if (direction == PF_OUT) {
3267 bport = nport = th->th_sport;
3268 /* check outgoing packet for BINAT/NAT */
3269 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3270 saddr, th->th_sport, daddr, th->th_dport,
3271 &pd->naddr, &nport)) != NULL) {
3272 PF_ACPY(&pd->baddr, saddr, af);
3273 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
3274 &th->th_sum, &pd->naddr, nport, 0, af);
3275 rewrite++;
3276 if (nr->natpass)
3277 r = NULL;
3278 pd->nat_rule = nr;
3279 }
3280 } else {
3281 bport = nport = th->th_dport;
3282 /* check incoming packet for BINAT/RDR */
3283 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3284 saddr, th->th_sport, daddr, th->th_dport,
3285 &pd->naddr, &nport)) != NULL) {
3286 PF_ACPY(&pd->baddr, daddr, af);
3287 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
3288 &th->th_sum, &pd->naddr, nport, 0, af);
3289 rewrite++;
3290 if (nr->natpass)
3291 r = NULL;
3292 pd->nat_rule = nr;
3293 }
3294 }
3295
3296 while (r != NULL) {
3297 r->evaluations++;
3298 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3299 r = r->skip[PF_SKIP_IFP].ptr;
3300 else if (r->direction && r->direction != direction)
3301 r = r->skip[PF_SKIP_DIR].ptr;
3302 else if (r->af && r->af != af)
3303 r = r->skip[PF_SKIP_AF].ptr;
3304 else if (r->proto && r->proto != IPPROTO_TCP)
3305 r = r->skip[PF_SKIP_PROTO].ptr;
3306 else if (PF_MISMATCHAW(&r->src.addr, saddr, af,
3307 r->src.neg, kif))
3308 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3309 else if (r->src.port_op && !pf_match_port(r->src.port_op,
3310 r->src.port[0], r->src.port[1], th->th_sport))
3311 r = r->skip[PF_SKIP_SRC_PORT].ptr;
3312 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af,
3313 r->dst.neg, NULL))
3314 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3315 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
3316 r->dst.port[0], r->dst.port[1], th->th_dport))
3317 r = r->skip[PF_SKIP_DST_PORT].ptr;
3318 else if (r->tos && !(r->tos == pd->tos))
3319 r = TAILQ_NEXT(r, entries);
3320 else if (r->rule_flag & PFRULE_FRAGMENT)
3321 r = TAILQ_NEXT(r, entries);
3322 else if ((r->flagset & th->th_flags) != r->flags)
3323 r = TAILQ_NEXT(r, entries);
3324 else if (r->uid.op && (pd->lookup.done || (pd->lookup.done =
3325#ifdef __FreeBSD__
3326 pf_socket_lookup(direction, pd, inp), 1)) &&
3327#else
3328 pf_socket_lookup(direction, pd), 1)) &&
3329#endif
3330 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
3331 pd->lookup.uid))
3332 r = TAILQ_NEXT(r, entries);
3333 else if (r->gid.op && (pd->lookup.done || (pd->lookup.done =
3334#ifdef __FreeBSD__
3335 pf_socket_lookup(direction, pd, inp), 1)) &&
3336#else
3337 pf_socket_lookup(direction, pd), 1)) &&
3338#endif
3339 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
3340 pd->lookup.gid))
3341 r = TAILQ_NEXT(r, entries);
3342 else if (r->prob && r->prob <= arc4random())
3343 r = TAILQ_NEXT(r, entries);
3344 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag))
3345 r = TAILQ_NEXT(r, entries);
3346 else if (r->os_fingerprint != PF_OSFP_ANY && !pf_osfp_match(
3347 pf_osfp_fingerprint(pd, m, off, th), r->os_fingerprint))
3348 r = TAILQ_NEXT(r, entries);
3349 else {
3350 if (r->tag)
3351 tag = r->tag;
3352 if (r->rtableid >= 0)
3353 rtableid = r->rtableid;
3354 if (r->anchor == NULL) {
3355 match = 1;
3356 *rm = r;
3357 *am = a;
3358 *rsm = ruleset;
3359 if ((*rm)->quick)
3360 break;
3361 r = TAILQ_NEXT(r, entries);
3362 } else
3363 pf_step_into_anchor(&asd, &ruleset,
3364 PF_RULESET_FILTER, &r, &a, &match);
3365 }
3366 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3367 PF_RULESET_FILTER, &r, &a, &match))
3368 break;
3369 }
3370 r = *rm;
3371 a = *am;
3372 ruleset = *rsm;
3373
3374 REASON_SET(&reason, PFRES_MATCH);
3375
3376 if (r->log || (nr != NULL && nr->natpass && nr->log)) {
3377 if (rewrite)
3378#ifdef __FreeBSD__
3379 m_copyback(m, off, sizeof(*th), (caddr_t)th);
3380#else
3381 m_copyback(m, off, sizeof(*th), th);
3382#endif
3383 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr,
3384 a, ruleset, pd);
3385 }
3386
3387 if ((r->action == PF_DROP) &&
3388 ((r->rule_flag & PFRULE_RETURNRST) ||
3389 (r->rule_flag & PFRULE_RETURNICMP) ||
3390 (r->rule_flag & PFRULE_RETURN))) {
3391 /* undo NAT changes, if they have taken place */
3392 if (nr != NULL) {
3393 if (direction == PF_OUT) {
3394 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
3395 &th->th_sum, &pd->baddr, bport, 0, af);
3396 rewrite++;
3397 } else {
3398 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
3399 &th->th_sum, &pd->baddr, bport, 0, af);
3400 rewrite++;
3401 }
3402 }
3403 if (((r->rule_flag & PFRULE_RETURNRST) ||
3404 (r->rule_flag & PFRULE_RETURN)) &&
3405 !(th->th_flags & TH_RST)) {
3406 u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
3407
3408 if (th->th_flags & TH_SYN)
3409 ack++;
3410 if (th->th_flags & TH_FIN)
3411 ack++;
3412#ifdef __FreeBSD__
3413 pf_send_tcp(m, r, af, pd->dst,
3414#else
3415 pf_send_tcp(r, af, pd->dst,
3416#endif
3417 pd->src, th->th_dport, th->th_sport,
3418 ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0,
3419 r->return_ttl, 1, 0, pd->eh, kif->pfik_ifp);
3420 } else if ((af == AF_INET) && r->return_icmp)
3421 pf_send_icmp(m, r->return_icmp >> 8,
3422 r->return_icmp & 255, af, r);
3423 else if ((af == AF_INET6) && r->return_icmp6)
3424 pf_send_icmp(m, r->return_icmp6 >> 8,
3425 r->return_icmp6 & 255, af, r);
3426 }
3427
3428 if (r->action == PF_DROP)
3429 return (PF_DROP);
3430
3431 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) {
3432 REASON_SET(&reason, PFRES_MEMORY);
3433 return (PF_DROP);
3434 }
3435
3436 if (r->keep_state || nr != NULL ||
3437 (pd->flags & PFDESC_TCP_NORM)) {
3438 /* create new state */
3439 u_int16_t len;
3440 struct pf_state *s = NULL;
3441 struct pf_src_node *sn = NULL;
3442
3443 len = pd->tot_len - off - (th->th_off << 2);
3444
3445 /* check maximums */
3446 if (r->max_states && (r->states >= r->max_states)) {
3447 pf_status.lcounters[LCNT_STATES]++;
3448 REASON_SET(&reason, PFRES_MAXSTATES);
3449 goto cleanup;
3450 }
3451 /* src node for filter rule */
3452 if ((r->rule_flag & PFRULE_SRCTRACK ||
3453 r->rpool.opts & PF_POOL_STICKYADDR) &&
3454 pf_insert_src_node(&sn, r, saddr, af) != 0) {
3455 REASON_SET(&reason, PFRES_SRCLIMIT);
3456 goto cleanup;
3457 }
3458 /* src node for translation rule */
3459 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3460 ((direction == PF_OUT &&
3461 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3462 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) {
3463 REASON_SET(&reason, PFRES_SRCLIMIT);
3464 goto cleanup;
3465 }
3466 s = pool_get(&pf_state_pl, PR_NOWAIT);
3467 if (s == NULL) {
3468 REASON_SET(&reason, PFRES_MEMORY);
3469cleanup:
3470 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3471 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3472 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3473 pf_status.src_nodes--;
3474 pool_put(&pf_src_tree_pl, sn);
3475 }
3476 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3477 nsn->expire == 0) {
3478 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3479 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3480 pf_status.src_nodes--;
3481 pool_put(&pf_src_tree_pl, nsn);
3482 }
3483 return (PF_DROP);
3484 }
3485 bzero(s, sizeof(*s));
3486 s->rule.ptr = r;
3487 s->nat_rule.ptr = nr;
3488 s->anchor.ptr = a;
3489 STATE_INC_COUNTERS(s);
3490 s->allow_opts = r->allow_opts;
3491 s->log = r->log & PF_LOG_ALL;
3492 if (nr != NULL)
3493 s->log |= nr->log & PF_LOG_ALL;
3494 s->proto = IPPROTO_TCP;
3495 s->direction = direction;
3496 s->af = af;
3497 if (direction == PF_OUT) {
3498 PF_ACPY(&s->gwy.addr, saddr, af);
3499 s->gwy.port = th->th_sport; /* sport */
3500 PF_ACPY(&s->ext.addr, daddr, af);
3501 s->ext.port = th->th_dport;
3502 if (nr != NULL) {
3503 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3504 s->lan.port = bport;
3505 } else {
3506 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3507 s->lan.port = s->gwy.port;
3508 }
3509 } else {
3510 PF_ACPY(&s->lan.addr, daddr, af);
3511 s->lan.port = th->th_dport;
3512 PF_ACPY(&s->ext.addr, saddr, af);
3513 s->ext.port = th->th_sport;
3514 if (nr != NULL) {
3515 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3516 s->gwy.port = bport;
3517 } else {
3518 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3519 s->gwy.port = s->lan.port;
3520 }
3521 }
3522
3523 s->src.seqlo = ntohl(th->th_seq);
3524 s->src.seqhi = s->src.seqlo + len + 1;
3525 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
3526 r->keep_state == PF_STATE_MODULATE) {
3527 /* Generate sequence number modulator */
3528#ifdef __FreeBSD__
3529 while ((s->src.seqdiff =
3530 pf_new_isn(s) - s->src.seqlo) == 0)
3531 ;
3532#else
3533 while ((s->src.seqdiff =
3534 tcp_rndiss_next() - s->src.seqlo) == 0)
3535 ;
3536#endif
3537 pf_change_a(&th->th_seq, &th->th_sum,
3538 htonl(s->src.seqlo + s->src.seqdiff), 0);
3539 rewrite = 1;
3540 } else
3541 s->src.seqdiff = 0;
3542 if (th->th_flags & TH_SYN) {
3543 s->src.seqhi++;
3544 s->src.wscale = pf_get_wscale(m, off, th->th_off, af);
3545 }
3546 s->src.max_win = MAX(ntohs(th->th_win), 1);
3547 if (s->src.wscale & PF_WSCALE_MASK) {
3548 /* Remove scale factor from initial window */
3549 int win = s->src.max_win;
3550 win += 1 << (s->src.wscale & PF_WSCALE_MASK);
3551 s->src.max_win = (win - 1) >>
3552 (s->src.wscale & PF_WSCALE_MASK);
3553 }
3554 if (th->th_flags & TH_FIN)
3555 s->src.seqhi++;
3556 s->dst.seqhi = 1;
3557 s->dst.max_win = 1;
3558 s->src.state = TCPS_SYN_SENT;
3559 s->dst.state = TCPS_CLOSED;
3560 s->creation = time_second;
3561 s->expire = time_second;
3562 s->timeout = PFTM_TCP_FIRST_PACKET;
3563 pf_set_rt_ifp(s, saddr);
3564 if (sn != NULL) {
3565 s->src_node = sn;
3566 s->src_node->states++;
3567 }
3568 if (nsn != NULL) {
3569 PF_ACPY(&nsn->raddr, &pd->naddr, af);
3570 s->nat_src_node = nsn;
3571 s->nat_src_node->states++;
3572 }
3573 if ((pd->flags & PFDESC_TCP_NORM) && pf_normalize_tcp_init(m,
3574 off, pd, th, &s->src, &s->dst)) {
3575 REASON_SET(&reason, PFRES_MEMORY);
3576 pf_src_tree_remove_state(s);
3577 STATE_DEC_COUNTERS(s);
3578 pool_put(&pf_state_pl, s);
3579 return (PF_DROP);
3580 }
3581 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
3582 pf_normalize_tcp_stateful(m, off, pd, &reason, th, s,
3583 &s->src, &s->dst, &rewrite)) {
3584 /* This really shouldn't happen!!! */
3585 DPFPRINTF(PF_DEBUG_URGENT,
3586 ("pf_normalize_tcp_stateful failed on first pkt"));
3587 pf_normalize_tcp_cleanup(s);
3588 pf_src_tree_remove_state(s);
3589 STATE_DEC_COUNTERS(s);
3590 pool_put(&pf_state_pl, s);
3591 return (PF_DROP);
3592 }
3593 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3594 pf_normalize_tcp_cleanup(s);
3595 REASON_SET(&reason, PFRES_STATEINS);
3596 pf_src_tree_remove_state(s);
3597 STATE_DEC_COUNTERS(s);
3598 pool_put(&pf_state_pl, s);
3599 return (PF_DROP);
3600 } else
3601 *sm = s;
3602 if (tag > 0) {
3603 pf_tag_ref(tag);
3604 s->tag = tag;
3605 }
3606 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
3607 r->keep_state == PF_STATE_SYNPROXY) {
3608 s->src.state = PF_TCPS_PROXY_SRC;
3609 if (nr != NULL) {
3610 if (direction == PF_OUT) {
3611 pf_change_ap(saddr, &th->th_sport,
3612 pd->ip_sum, &th->th_sum, &pd->baddr,
3613 bport, 0, af);
3614 } else {
3615 pf_change_ap(daddr, &th->th_dport,
3616 pd->ip_sum, &th->th_sum, &pd->baddr,
3617 bport, 0, af);
3618 }
3619 }
3620 s->src.seqhi = htonl(arc4random());
3621 /* Find mss option */
3622 mss = pf_get_mss(m, off, th->th_off, af);
3623 mss = pf_calc_mss(saddr, af, mss);
3624 mss = pf_calc_mss(daddr, af, mss);
3625 s->src.mss = mss;
3626#ifdef __FreeBSD__
3627 pf_send_tcp(NULL, r, af, daddr, saddr, th->th_dport,
3628#else
3629 pf_send_tcp(r, af, daddr, saddr, th->th_dport,
3630#endif
3631 th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1,
3632 TH_SYN|TH_ACK, 0, s->src.mss, 0, 1, 0, NULL, NULL);
3633 REASON_SET(&reason, PFRES_SYNPROXY);
3634 return (PF_SYNPROXY_DROP);
3635 }
3636 }
3637
3638 /* copy back packet headers if we performed NAT operations */
3639 if (rewrite)
3640 m_copyback(m, off, sizeof(*th), (caddr_t)th);
3641
3642 return (PF_PASS);
3643}
3644
3645int
3646pf_test_udp(struct pf_rule **rm, struct pf_state **sm, int direction,
3647 struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3648#ifdef __FreeBSD__
3649 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
3650 struct ifqueue *ifq, struct inpcb *inp)
3651#else
3652 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
3653 struct ifqueue *ifq)
3654#endif
3655{
3656 struct pf_rule *nr = NULL;
3657 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
3658 struct udphdr *uh = pd->hdr.udp;
3659 u_int16_t bport, nport = 0;
3660 sa_family_t af = pd->af;
3661 struct pf_rule *r, *a = NULL;
3662 struct pf_ruleset *ruleset = NULL;
3663 struct pf_src_node *nsn = NULL;
3664 u_short reason;
3665 int rewrite = 0;
3666 int tag = -1, rtableid = -1;
3667 int asd = 0;
3668 int match = 0;
3669
3670 if (pf_check_congestion(ifq)) {
3671 REASON_SET(&reason, PFRES_CONGEST);
3672 return (PF_DROP);
3673 }
3674
3675#ifdef __FreeBSD__
3676 if (inp != NULL)
3677 pd->lookup.done = pf_socket_lookup(direction, pd, inp);
3678 else if (debug_pfugidhack) {
3679 PF_UNLOCK();
3680 DPFPRINTF(PF_DEBUG_MISC, ("pf: unlocked lookup\n"));
3681 pd->lookup.done = pf_socket_lookup(direction, pd, inp);
3682 PF_LOCK();
3683 }
3684#endif
3685
3686 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3687
3688 if (direction == PF_OUT) {
3689 bport = nport = uh->uh_sport;
3690 /* check outgoing packet for BINAT/NAT */
3691 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3692 saddr, uh->uh_sport, daddr, uh->uh_dport,
3693 &pd->naddr, &nport)) != NULL) {
3694 PF_ACPY(&pd->baddr, saddr, af);
3695 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum,
3696 &uh->uh_sum, &pd->naddr, nport, 1, af);
3697 rewrite++;
3698 if (nr->natpass)
3699 r = NULL;
3700 pd->nat_rule = nr;
3701 }
3702 } else {
3703 bport = nport = uh->uh_dport;
3704 /* check incoming packet for BINAT/RDR */
3705 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3706 saddr, uh->uh_sport, daddr, uh->uh_dport, &pd->naddr,
3707 &nport)) != NULL) {
3708 PF_ACPY(&pd->baddr, daddr, af);
3709 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum,
3710 &uh->uh_sum, &pd->naddr, nport, 1, af);
3711 rewrite++;
3712 if (nr->natpass)
3713 r = NULL;
3714 pd->nat_rule = nr;
3715 }
3716 }
3717
3718 while (r != NULL) {
3719 r->evaluations++;
3720 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3721 r = r->skip[PF_SKIP_IFP].ptr;
3722 else if (r->direction && r->direction != direction)
3723 r = r->skip[PF_SKIP_DIR].ptr;
3724 else if (r->af && r->af != af)
3725 r = r->skip[PF_SKIP_AF].ptr;
3726 else if (r->proto && r->proto != IPPROTO_UDP)
3727 r = r->skip[PF_SKIP_PROTO].ptr;
3728 else if (PF_MISMATCHAW(&r->src.addr, saddr, af,
3729 r->src.neg, kif))
3730 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3731 else if (r->src.port_op && !pf_match_port(r->src.port_op,
3732 r->src.port[0], r->src.port[1], uh->uh_sport))
3733 r = r->skip[PF_SKIP_SRC_PORT].ptr;
3734 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af,
3735 r->dst.neg, NULL))
3736 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3737 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
3738 r->dst.port[0], r->dst.port[1], uh->uh_dport))
3739 r = r->skip[PF_SKIP_DST_PORT].ptr;
3740 else if (r->tos && !(r->tos == pd->tos))
3741 r = TAILQ_NEXT(r, entries);
3742 else if (r->rule_flag & PFRULE_FRAGMENT)
3743 r = TAILQ_NEXT(r, entries);
3744 else if (r->uid.op && (pd->lookup.done || (pd->lookup.done =
3745#ifdef __FreeBSD__
3746 pf_socket_lookup(direction, pd, inp), 1)) &&
3747#else
3748 pf_socket_lookup(direction, pd), 1)) &&
3749#endif
3750 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
3751 pd->lookup.uid))
3752 r = TAILQ_NEXT(r, entries);
3753 else if (r->gid.op && (pd->lookup.done || (pd->lookup.done =
3754#ifdef __FreeBSD__
3755 pf_socket_lookup(direction, pd, inp), 1)) &&
3756#else
3757 pf_socket_lookup(direction, pd), 1)) &&
3758#endif
3759 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
3760 pd->lookup.gid))
3761 r = TAILQ_NEXT(r, entries);
3762 else if (r->prob && r->prob <= arc4random())
3763 r = TAILQ_NEXT(r, entries);
3764 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag))
3765 r = TAILQ_NEXT(r, entries);
3766 else if (r->os_fingerprint != PF_OSFP_ANY)
3767 r = TAILQ_NEXT(r, entries);
3768 else {
3769 if (r->tag)
3770 tag = r->tag;
3771 if (r->rtableid >= 0)
3772 rtableid = r->rtableid;
3773 if (r->anchor == NULL) {
3774 match = 1;
3775 *rm = r;
3776 *am = a;
3777 *rsm = ruleset;
3778 if ((*rm)->quick)
3779 break;
3780 r = TAILQ_NEXT(r, entries);
3781 } else
3782 pf_step_into_anchor(&asd, &ruleset,
3783 PF_RULESET_FILTER, &r, &a, &match);
3784 }
3785 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3786 PF_RULESET_FILTER, &r, &a, &match))
3787 break;
3788 }
3789 r = *rm;
3790 a = *am;
3791 ruleset = *rsm;
3792
3793 REASON_SET(&reason, PFRES_MATCH);
3794
3795 if (r->log || (nr != NULL && nr->natpass && nr->log)) {
3796 if (rewrite)
3797#ifdef __FreeBSD__
3798 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
3799#else
3800 m_copyback(m, off, sizeof(*uh), uh);
3801#endif
3802 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr,
3803 a, ruleset, pd);
3804 }
3805
3806 if ((r->action == PF_DROP) &&
3807 ((r->rule_flag & PFRULE_RETURNICMP) ||
3808 (r->rule_flag & PFRULE_RETURN))) {
3809 /* undo NAT changes, if they have taken place */
3810 if (nr != NULL) {
3811 if (direction == PF_OUT) {
3812 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum,
3813 &uh->uh_sum, &pd->baddr, bport, 1, af);
3814 rewrite++;
3815 } else {
3816 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum,
3817 &uh->uh_sum, &pd->baddr, bport, 1, af);
3818 rewrite++;
3819 }
3820 }
3821 if ((af == AF_INET) && r->return_icmp)
3822 pf_send_icmp(m, r->return_icmp >> 8,
3823 r->return_icmp & 255, af, r);
3824 else if ((af == AF_INET6) && r->return_icmp6)
3825 pf_send_icmp(m, r->return_icmp6 >> 8,
3826 r->return_icmp6 & 255, af, r);
3827 }
3828
3829 if (r->action == PF_DROP)
3830 return (PF_DROP);
3831
3832 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) {
3833 REASON_SET(&reason, PFRES_MEMORY);
3834 return (PF_DROP);
3835 }
3836
3837 if (r->keep_state || nr != NULL) {
3838 /* create new state */
3839 struct pf_state *s = NULL;
3840 struct pf_src_node *sn = NULL;
3841
3842 /* check maximums */
3843 if (r->max_states && (r->states >= r->max_states)) {
3844 pf_status.lcounters[LCNT_STATES]++;
3845 REASON_SET(&reason, PFRES_MAXSTATES);
3846 goto cleanup;
3847 }
3848 /* src node for filter rule */
3849 if ((r->rule_flag & PFRULE_SRCTRACK ||
3850 r->rpool.opts & PF_POOL_STICKYADDR) &&
3851 pf_insert_src_node(&sn, r, saddr, af) != 0) {
3852 REASON_SET(&reason, PFRES_SRCLIMIT);
3853 goto cleanup;
3854 }
3855 /* src node for translation rule */
3856 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3857 ((direction == PF_OUT &&
3858 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3859 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) {
3860 REASON_SET(&reason, PFRES_SRCLIMIT);
3861 goto cleanup;
3862 }
3863 s = pool_get(&pf_state_pl, PR_NOWAIT);
3864 if (s == NULL) {
3865 REASON_SET(&reason, PFRES_MEMORY);
3866cleanup:
3867 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3868 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3869 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3870 pf_status.src_nodes--;
3871 pool_put(&pf_src_tree_pl, sn);
3872 }
3873 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3874 nsn->expire == 0) {
3875 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3876 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3877 pf_status.src_nodes--;
3878 pool_put(&pf_src_tree_pl, nsn);
3879 }
3880 return (PF_DROP);
3881 }
3882 bzero(s, sizeof(*s));
3883 s->rule.ptr = r;
3884 s->nat_rule.ptr = nr;
3885 s->anchor.ptr = a;
3886 STATE_INC_COUNTERS(s);
3887 s->allow_opts = r->allow_opts;
3888 s->log = r->log & PF_LOG_ALL;
3889 if (nr != NULL)
3890 s->log |= nr->log & PF_LOG_ALL;
3891 s->proto = IPPROTO_UDP;
3892 s->direction = direction;
3893 s->af = af;
3894 if (direction == PF_OUT) {
3895 PF_ACPY(&s->gwy.addr, saddr, af);
3896 s->gwy.port = uh->uh_sport;
3897 PF_ACPY(&s->ext.addr, daddr, af);
3898 s->ext.port = uh->uh_dport;
3899 if (nr != NULL) {
3900 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3901 s->lan.port = bport;
3902 } else {
3903 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3904 s->lan.port = s->gwy.port;
3905 }
3906 } else {
3907 PF_ACPY(&s->lan.addr, daddr, af);
3908 s->lan.port = uh->uh_dport;
3909 PF_ACPY(&s->ext.addr, saddr, af);
3910 s->ext.port = uh->uh_sport;
3911 if (nr != NULL) {
3912 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3913 s->gwy.port = bport;
3914 } else {
3915 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3916 s->gwy.port = s->lan.port;
3917 }
3918 }
3919 s->src.state = PFUDPS_SINGLE;
3920 s->dst.state = PFUDPS_NO_TRAFFIC;
3921 s->creation = time_second;
3922 s->expire = time_second;
3923 s->timeout = PFTM_UDP_FIRST_PACKET;
3924 pf_set_rt_ifp(s, saddr);
3925 if (sn != NULL) {
3926 s->src_node = sn;
3927 s->src_node->states++;
3928 }
3929 if (nsn != NULL) {
3930 PF_ACPY(&nsn->raddr, &pd->naddr, af);
3931 s->nat_src_node = nsn;
3932 s->nat_src_node->states++;
3933 }
3934 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3935 REASON_SET(&reason, PFRES_STATEINS);
3936 pf_src_tree_remove_state(s);
3937 STATE_DEC_COUNTERS(s);
3938 pool_put(&pf_state_pl, s);
3939 return (PF_DROP);
3940 } else
3941 *sm = s;
3942 if (tag > 0) {
3943 pf_tag_ref(tag);
3944 s->tag = tag;
3945 }
3946 }
3947
3948 /* copy back packet headers if we performed NAT operations */
3949 if (rewrite)
3950 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
3951
3952 return (PF_PASS);
3953}
3954
3955int
3956pf_test_icmp(struct pf_rule **rm, struct pf_state **sm, int direction,
3957 struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3958 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
3959 struct ifqueue *ifq)
3960{
3961 struct pf_rule *nr = NULL;
3962 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
3963 struct pf_rule *r, *a = NULL;
3964 struct pf_ruleset *ruleset = NULL;
3965 struct pf_src_node *nsn = NULL;
3966 u_short reason;
3967 u_int16_t icmpid = 0, bport, nport = 0;
3968 sa_family_t af = pd->af;
3969 u_int8_t icmptype = 0; /* make the compiler happy */
3970 u_int8_t icmpcode = 0; /* make the compiler happy */
3971 int state_icmp = 0;
3972 int tag = -1, rtableid = -1;
3973#ifdef INET6
3974 int rewrite = 0;
3975#endif /* INET6 */
3976 int asd = 0;
3977 int match = 0;
3978
3979 if (pf_check_congestion(ifq)) {
3980 REASON_SET(&reason, PFRES_CONGEST);
3981 return (PF_DROP);
3982 }
3983
3984 switch (pd->proto) {
3985#ifdef INET
3986 case IPPROTO_ICMP:
3987 icmptype = pd->hdr.icmp->icmp_type;
3988 icmpcode = pd->hdr.icmp->icmp_code;
3989 icmpid = pd->hdr.icmp->icmp_id;
3990
3991 if (icmptype == ICMP_UNREACH ||
3992 icmptype == ICMP_SOURCEQUENCH ||
3993 icmptype == ICMP_REDIRECT ||
3994 icmptype == ICMP_TIMXCEED ||
3995 icmptype == ICMP_PARAMPROB)
3996 state_icmp++;
3997 break;
3998#endif /* INET */
3999#ifdef INET6
4000 case IPPROTO_ICMPV6:
4001 icmptype = pd->hdr.icmp6->icmp6_type;
4002 icmpcode = pd->hdr.icmp6->icmp6_code;
4003 icmpid = pd->hdr.icmp6->icmp6_id;
4004
4005 if (icmptype == ICMP6_DST_UNREACH ||
4006 icmptype == ICMP6_PACKET_TOO_BIG ||
4007 icmptype == ICMP6_TIME_EXCEEDED ||
4008 icmptype == ICMP6_PARAM_PROB)
4009 state_icmp++;
4010 break;
4011#endif /* INET6 */
4012 }
4013
4014 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
4015
4016 if (direction == PF_OUT) {
4017 bport = nport = icmpid;
4018 /* check outgoing packet for BINAT/NAT */
4019 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
4020 saddr, icmpid, daddr, icmpid, &pd->naddr, &nport)) !=
4021 NULL) {
4022 PF_ACPY(&pd->baddr, saddr, af);
4023 switch (af) {
4024#ifdef INET
4025 case AF_INET:
4026 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
4027 pd->naddr.v4.s_addr, 0);
4028 pd->hdr.icmp->icmp_cksum = pf_cksum_fixup(
4029 pd->hdr.icmp->icmp_cksum, icmpid, nport, 0);
4030 pd->hdr.icmp->icmp_id = nport;
4031 m_copyback(m, off, ICMP_MINLEN,
4032 (caddr_t)pd->hdr.icmp);
4033 break;
4034#endif /* INET */
4035#ifdef INET6
4036 case AF_INET6:
4037 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum,
4038 &pd->naddr, 0);
4039 rewrite++;
4040 break;
4041#endif /* INET6 */
4042 }
4043 if (nr->natpass)
4044 r = NULL;
4045 pd->nat_rule = nr;
4046 }
4047 } else {
4048 bport = nport = icmpid;
4049 /* check incoming packet for BINAT/RDR */
4050 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
4051 saddr, icmpid, daddr, icmpid, &pd->naddr, &nport)) !=
4052 NULL) {
4053 PF_ACPY(&pd->baddr, daddr, af);
4054 switch (af) {
4055#ifdef INET
4056 case AF_INET:
4057 pf_change_a(&daddr->v4.s_addr,
4058 pd->ip_sum, pd->naddr.v4.s_addr, 0);
4059 break;
4060#endif /* INET */
4061#ifdef INET6
4062 case AF_INET6:
4063 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum,
4064 &pd->naddr, 0);
4065 rewrite++;
4066 break;
4067#endif /* INET6 */
4068 }
4069 if (nr->natpass)
4070 r = NULL;
4071 pd->nat_rule = nr;
4072 }
4073 }
4074
4075 while (r != NULL) {
4076 r->evaluations++;
4077 if (pfi_kif_match(r->kif, kif) == r->ifnot)
4078 r = r->skip[PF_SKIP_IFP].ptr;
4079 else if (r->direction && r->direction != direction)
4080 r = r->skip[PF_SKIP_DIR].ptr;
4081 else if (r->af && r->af != af)
4082 r = r->skip[PF_SKIP_AF].ptr;
4083 else if (r->proto && r->proto != pd->proto)
4084 r = r->skip[PF_SKIP_PROTO].ptr;
4085 else if (PF_MISMATCHAW(&r->src.addr, saddr, af,
4086 r->src.neg, kif))
4087 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
4088 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af,
4089 r->dst.neg, NULL))
4090 r = r->skip[PF_SKIP_DST_ADDR].ptr;
4091 else if (r->type && r->type != icmptype + 1)
4092 r = TAILQ_NEXT(r, entries);
4093 else if (r->code && r->code != icmpcode + 1)
4094 r = TAILQ_NEXT(r, entries);
4095 else if (r->tos && !(r->tos == pd->tos))
4096 r = TAILQ_NEXT(r, entries);
4097 else if (r->rule_flag & PFRULE_FRAGMENT)
4098 r = TAILQ_NEXT(r, entries);
4099 else if (r->prob && r->prob <= arc4random())
4100 r = TAILQ_NEXT(r, entries);
4101 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag))
4102 r = TAILQ_NEXT(r, entries);
4103 else if (r->os_fingerprint != PF_OSFP_ANY)
4104 r = TAILQ_NEXT(r, entries);
4105 else {
4106 if (r->tag)
4107 tag = r->tag;
4108 if (r->rtableid >= 0)
4109 rtableid = r->rtableid;
4110 if (r->anchor == NULL) {
4111 match = 1;
4112 *rm = r;
4113 *am = a;
4114 *rsm = ruleset;
4115 if ((*rm)->quick)
4116 break;
4117 r = TAILQ_NEXT(r, entries);
4118 } else
4119 pf_step_into_anchor(&asd, &ruleset,
4120 PF_RULESET_FILTER, &r, &a, &match);
4121 }
4122 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
4123 PF_RULESET_FILTER, &r, &a, &match))
4124 break;
4125 }
4126 r = *rm;
4127 a = *am;
4128 ruleset = *rsm;
4129
4130 REASON_SET(&reason, PFRES_MATCH);
4131
4132 if (r->log || (nr != NULL && nr->natpass && nr->log)) {
4133#ifdef INET6
4134 if (rewrite)
4135 m_copyback(m, off, sizeof(struct icmp6_hdr),
4136 (caddr_t)pd->hdr.icmp6);
4137#endif /* INET6 */
4138 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr,
4139 a, ruleset, pd);
4140 }
4141
4142 if (r->action != PF_PASS)
4143 return (PF_DROP);
4144
4145 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) {
4146 REASON_SET(&reason, PFRES_MEMORY);
4147 return (PF_DROP);
4148 }
4149
4150 if (!state_icmp && (r->keep_state || nr != NULL)) {
4151 /* create new state */
4152 struct pf_state *s = NULL;
4153 struct pf_src_node *sn = NULL;
4154
4155 /* check maximums */
4156 if (r->max_states && (r->states >= r->max_states)) {
4157 pf_status.lcounters[LCNT_STATES]++;
4158 REASON_SET(&reason, PFRES_MAXSTATES);
4159 goto cleanup;
4160 }
4161 /* src node for filter rule */
4162 if ((r->rule_flag & PFRULE_SRCTRACK ||
4163 r->rpool.opts & PF_POOL_STICKYADDR) &&
4164 pf_insert_src_node(&sn, r, saddr, af) != 0) {
4165 REASON_SET(&reason, PFRES_SRCLIMIT);
4166 goto cleanup;
4167 }
4168 /* src node for translation rule */
4169 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
4170 ((direction == PF_OUT &&
4171 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
4172 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) {
4173 REASON_SET(&reason, PFRES_SRCLIMIT);
4174 goto cleanup;
4175 }
4176 s = pool_get(&pf_state_pl, PR_NOWAIT);
4177 if (s == NULL) {
4178 REASON_SET(&reason, PFRES_MEMORY);
4179cleanup:
4180 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
4181 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
4182 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
4183 pf_status.src_nodes--;
4184 pool_put(&pf_src_tree_pl, sn);
4185 }
4186 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
4187 nsn->expire == 0) {
4188 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
4189 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
4190 pf_status.src_nodes--;
4191 pool_put(&pf_src_tree_pl, nsn);
4192 }
4193 return (PF_DROP);
4194 }
4195 bzero(s, sizeof(*s));
4196 s->rule.ptr = r;
4197 s->nat_rule.ptr = nr;
4198 s->anchor.ptr = a;
4199 STATE_INC_COUNTERS(s);
4200 s->allow_opts = r->allow_opts;
4201 s->log = r->log & PF_LOG_ALL;
4202 if (nr != NULL)
4203 s->log |= nr->log & PF_LOG_ALL;
4204 s->proto = pd->proto;
4205 s->direction = direction;
4206 s->af = af;
4207 if (direction == PF_OUT) {
4208 PF_ACPY(&s->gwy.addr, saddr, af);
4209 s->gwy.port = nport;
4210 PF_ACPY(&s->ext.addr, daddr, af);
4211 s->ext.port = 0;
4212 if (nr != NULL) {
4213 PF_ACPY(&s->lan.addr, &pd->baddr, af);
4214 s->lan.port = bport;
4215 } else {
4216 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
4217 s->lan.port = s->gwy.port;
4218 }
4219 } else {
4220 PF_ACPY(&s->lan.addr, daddr, af);
4221 s->lan.port = nport;
4222 PF_ACPY(&s->ext.addr, saddr, af);
4223 s->ext.port = 0;
4224 if (nr != NULL) {
4225 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
4226 s->gwy.port = bport;
4227 } else {
4228 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
4229 s->gwy.port = s->lan.port;
4230 }
4231 }
4232 s->creation = time_second;
4233 s->expire = time_second;
4234 s->timeout = PFTM_ICMP_FIRST_PACKET;
4235 pf_set_rt_ifp(s, saddr);
4236 if (sn != NULL) {
4237 s->src_node = sn;
4238 s->src_node->states++;
4239 }
4240 if (nsn != NULL) {
4241 PF_ACPY(&nsn->raddr, &pd->naddr, af);
4242 s->nat_src_node = nsn;
4243 s->nat_src_node->states++;
4244 }
4245 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
4246 REASON_SET(&reason, PFRES_STATEINS);
4247 pf_src_tree_remove_state(s);
4248 STATE_DEC_COUNTERS(s);
4249 pool_put(&pf_state_pl, s);
4250 return (PF_DROP);
4251 } else
4252 *sm = s;
4253 if (tag > 0) {
4254 pf_tag_ref(tag);
4255 s->tag = tag;
4256 }
4257 }
4258
4259#ifdef INET6
4260 /* copy back packet headers if we performed IPv6 NAT operations */
4261 if (rewrite)
4262 m_copyback(m, off, sizeof(struct icmp6_hdr),
4263 (caddr_t)pd->hdr.icmp6);
4264#endif /* INET6 */
4265
4266 return (PF_PASS);
4267}
4268
4269int
4270pf_test_other(struct pf_rule **rm, struct pf_state **sm, int direction,
4271 struct pfi_kif *kif, struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
4272 struct pf_rule **am, struct pf_ruleset **rsm, struct ifqueue *ifq)
4273{
4274 struct pf_rule *nr = NULL;
4275 struct pf_rule *r, *a = NULL;
4276 struct pf_ruleset *ruleset = NULL;
4277 struct pf_src_node *nsn = NULL;
4278 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
4279 sa_family_t af = pd->af;
4280 u_short reason;
4281 int tag = -1, rtableid = -1;
4282 int asd = 0;
4283 int match = 0;
4284
4285 if (pf_check_congestion(ifq)) {
4286 REASON_SET(&reason, PFRES_CONGEST);
4287 return (PF_DROP);
4288 }
4289
4290 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
4291
4292 if (direction == PF_OUT) {
4293 /* check outgoing packet for BINAT/NAT */
4294 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
4295 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
4296 PF_ACPY(&pd->baddr, saddr, af);
4297 switch (af) {
4298#ifdef INET
4299 case AF_INET:
4300 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
4301 pd->naddr.v4.s_addr, 0);
4302 break;
4303#endif /* INET */
4304#ifdef INET6
4305 case AF_INET6:
4306 PF_ACPY(saddr, &pd->naddr, af);
4307 break;
4308#endif /* INET6 */
4309 }
4310 if (nr->natpass)
4311 r = NULL;
4312 pd->nat_rule = nr;
4313 }
4314 } else {
4315 /* check incoming packet for BINAT/RDR */
4316 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
4317 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
4318 PF_ACPY(&pd->baddr, daddr, af);
4319 switch (af) {
4320#ifdef INET
4321 case AF_INET:
4322 pf_change_a(&daddr->v4.s_addr,
4323 pd->ip_sum, pd->naddr.v4.s_addr, 0);
4324 break;
4325#endif /* INET */
4326#ifdef INET6
4327 case AF_INET6:
4328 PF_ACPY(daddr, &pd->naddr, af);
4329 break;
4330#endif /* INET6 */
4331 }
4332 if (nr->natpass)
4333 r = NULL;
4334 pd->nat_rule = nr;
4335 }
4336 }
4337
4338 while (r != NULL) {
4339 r->evaluations++;
4340 if (pfi_kif_match(r->kif, kif) == r->ifnot)
4341 r = r->skip[PF_SKIP_IFP].ptr;
4342 else if (r->direction && r->direction != direction)
4343 r = r->skip[PF_SKIP_DIR].ptr;
4344 else if (r->af && r->af != af)
4345 r = r->skip[PF_SKIP_AF].ptr;
4346 else if (r->proto && r->proto != pd->proto)
4347 r = r->skip[PF_SKIP_PROTO].ptr;
4348 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af,
4349 r->src.neg, kif))
4350 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
4351 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af,
4352 r->dst.neg, NULL))
4353 r = r->skip[PF_SKIP_DST_ADDR].ptr;
4354 else if (r->tos && !(r->tos == pd->tos))
4355 r = TAILQ_NEXT(r, entries);
4356 else if (r->rule_flag & PFRULE_FRAGMENT)
4357 r = TAILQ_NEXT(r, entries);
4358 else if (r->prob && r->prob <= arc4random())
4359 r = TAILQ_NEXT(r, entries);
4360 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag))
4361 r = TAILQ_NEXT(r, entries);
4362 else if (r->os_fingerprint != PF_OSFP_ANY)
4363 r = TAILQ_NEXT(r, entries);
4364 else {
4365 if (r->tag)
4366 tag = r->tag;
4367 if (r->rtableid >= 0)
4368 rtableid = r->rtableid;
4369 if (r->anchor == NULL) {
4370 match = 1;
4371 *rm = r;
4372 *am = a;
4373 *rsm = ruleset;
4374 if ((*rm)->quick)
4375 break;
4376 r = TAILQ_NEXT(r, entries);
4377 } else
4378 pf_step_into_anchor(&asd, &ruleset,
4379 PF_RULESET_FILTER, &r, &a, &match);
4380 }
4381 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
4382 PF_RULESET_FILTER, &r, &a, &match))
4383 break;
4384 }
4385 r = *rm;
4386 a = *am;
4387 ruleset = *rsm;
4388
4389 REASON_SET(&reason, PFRES_MATCH);
4390
4391 if (r->log || (nr != NULL && nr->natpass && nr->log))
4392 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr,
4393 a, ruleset, pd);
4394
4395 if ((r->action == PF_DROP) &&
4396 ((r->rule_flag & PFRULE_RETURNICMP) ||
4397 (r->rule_flag & PFRULE_RETURN))) {
4398 struct pf_addr *a = NULL;
4399
4400 if (nr != NULL) {
4401 if (direction == PF_OUT)
4402 a = saddr;
4403 else
4404 a = daddr;
4405 }
4406 if (a != NULL) {
4407 switch (af) {
4408#ifdef INET
4409 case AF_INET:
4410 pf_change_a(&a->v4.s_addr, pd->ip_sum,
4411 pd->baddr.v4.s_addr, 0);
4412 break;
4413#endif /* INET */
4414#ifdef INET6
4415 case AF_INET6:
4416 PF_ACPY(a, &pd->baddr, af);
4417 break;
4418#endif /* INET6 */
4419 }
4420 }
4421 if ((af == AF_INET) && r->return_icmp)
4422 pf_send_icmp(m, r->return_icmp >> 8,
4423 r->return_icmp & 255, af, r);
4424 else if ((af == AF_INET6) && r->return_icmp6)
4425 pf_send_icmp(m, r->return_icmp6 >> 8,
4426 r->return_icmp6 & 255, af, r);
4427 }
4428
4429 if (r->action != PF_PASS)
4430 return (PF_DROP);
4431
4432 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) {
4433 REASON_SET(&reason, PFRES_MEMORY);
4434 return (PF_DROP);
4435 }
4436
4437 if (r->keep_state || nr != NULL) {
4438 /* create new state */
4439 struct pf_state *s = NULL;
4440 struct pf_src_node *sn = NULL;
4441
4442 /* check maximums */
4443 if (r->max_states && (r->states >= r->max_states)) {
4444 pf_status.lcounters[LCNT_STATES]++;
4445 REASON_SET(&reason, PFRES_MAXSTATES);
4446 goto cleanup;
4447 }
4448 /* src node for filter rule */
4449 if ((r->rule_flag & PFRULE_SRCTRACK ||
4450 r->rpool.opts & PF_POOL_STICKYADDR) &&
4451 pf_insert_src_node(&sn, r, saddr, af) != 0) {
4452 REASON_SET(&reason, PFRES_SRCLIMIT);
4453 goto cleanup;
4454 }
4455 /* src node for translation rule */
4456 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
4457 ((direction == PF_OUT &&
4458 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
4459 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) {
4460 REASON_SET(&reason, PFRES_SRCLIMIT);
4461 goto cleanup;
4462 }
4463 s = pool_get(&pf_state_pl, PR_NOWAIT);
4464 if (s == NULL) {
4465 REASON_SET(&reason, PFRES_MEMORY);
4466cleanup:
4467 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
4468 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
4469 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
4470 pf_status.src_nodes--;
4471 pool_put(&pf_src_tree_pl, sn);
4472 }
4473 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
4474 nsn->expire == 0) {
4475 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
4476 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
4477 pf_status.src_nodes--;
4478 pool_put(&pf_src_tree_pl, nsn);
4479 }
4480 return (PF_DROP);
4481 }
4482 bzero(s, sizeof(*s));
4483 s->rule.ptr = r;
4484 s->nat_rule.ptr = nr;
4485 s->anchor.ptr = a;
4486 STATE_INC_COUNTERS(s);
4487 s->allow_opts = r->allow_opts;
4488 s->log = r->log & PF_LOG_ALL;
4489 if (nr != NULL)
4490 s->log |= nr->log & PF_LOG_ALL;
4491 s->proto = pd->proto;
4492 s->direction = direction;
4493 s->af = af;
4494 if (direction == PF_OUT) {
4495 PF_ACPY(&s->gwy.addr, saddr, af);
4496 PF_ACPY(&s->ext.addr, daddr, af);
4497 if (nr != NULL)
4498 PF_ACPY(&s->lan.addr, &pd->baddr, af);
4499 else
4500 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
4501 } else {
4502 PF_ACPY(&s->lan.addr, daddr, af);
4503 PF_ACPY(&s->ext.addr, saddr, af);
4504 if (nr != NULL)
4505 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
4506 else
4507 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
4508 }
4509 s->src.state = PFOTHERS_SINGLE;
4510 s->dst.state = PFOTHERS_NO_TRAFFIC;
4511 s->creation = time_second;
4512 s->expire = time_second;
4513 s->timeout = PFTM_OTHER_FIRST_PACKET;
4514 pf_set_rt_ifp(s, saddr);
4515 if (sn != NULL) {
4516 s->src_node = sn;
4517 s->src_node->states++;
4518 }
4519 if (nsn != NULL) {
4520 PF_ACPY(&nsn->raddr, &pd->naddr, af);
4521 s->nat_src_node = nsn;
4522 s->nat_src_node->states++;
4523 }
4524 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
4525 REASON_SET(&reason, PFRES_STATEINS);
4526 pf_src_tree_remove_state(s);
4527 STATE_DEC_COUNTERS(s);
4528 pool_put(&pf_state_pl, s);
4529 return (PF_DROP);
4530 } else
4531 *sm = s;
4532 if (tag > 0) {
4533 pf_tag_ref(tag);
4534 s->tag = tag;
4535 }
4536 }
4537
4538 return (PF_PASS);
4539}
4540
4541int
4542pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif,
4543 struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am,
4544 struct pf_ruleset **rsm)
4545{
4546 struct pf_rule *r, *a = NULL;
4547 struct pf_ruleset *ruleset = NULL;
4548 sa_family_t af = pd->af;
4549 u_short reason;
4550 int tag = -1;
4551 int asd = 0;
4552 int match = 0;
4553
4554 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
4555 while (r != NULL) {
4556 r->evaluations++;
4557 if (pfi_kif_match(r->kif, kif) == r->ifnot)
4558 r = r->skip[PF_SKIP_IFP].ptr;
4559 else if (r->direction && r->direction != direction)
4560 r = r->skip[PF_SKIP_DIR].ptr;
4561 else if (r->af && r->af != af)
4562 r = r->skip[PF_SKIP_AF].ptr;
4563 else if (r->proto && r->proto != pd->proto)
4564 r = r->skip[PF_SKIP_PROTO].ptr;
4565 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af,
4566 r->src.neg, kif))
4567 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
4568 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af,
4569 r->dst.neg, NULL))
4570 r = r->skip[PF_SKIP_DST_ADDR].ptr;
4571 else if (r->tos && !(r->tos == pd->tos))
4572 r = TAILQ_NEXT(r, entries);
4573 else if (r->os_fingerprint != PF_OSFP_ANY)
4574 r = TAILQ_NEXT(r, entries);
4575 else if (pd->proto == IPPROTO_UDP &&
4576 (r->src.port_op || r->dst.port_op))
4577 r = TAILQ_NEXT(r, entries);
4578 else if (pd->proto == IPPROTO_TCP &&
4579 (r->src.port_op || r->dst.port_op || r->flagset))
4580 r = TAILQ_NEXT(r, entries);
4581 else if ((pd->proto == IPPROTO_ICMP ||
4582 pd->proto == IPPROTO_ICMPV6) &&
4583 (r->type || r->code))
4584 r = TAILQ_NEXT(r, entries);
4585 else if (r->prob && r->prob <= arc4random())
4586 r = TAILQ_NEXT(r, entries);
4587 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag))
4588 r = TAILQ_NEXT(r, entries);
4589 else {
4590 if (r->anchor == NULL) {
4591 match = 1;
4592 *rm = r;
4593 *am = a;
4594 *rsm = ruleset;
4595 if ((*rm)->quick)
4596 break;
4597 r = TAILQ_NEXT(r, entries);
4598 } else
4599 pf_step_into_anchor(&asd, &ruleset,
4600 PF_RULESET_FILTER, &r, &a, &match);
4601 }
4602 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
4603 PF_RULESET_FILTER, &r, &a, &match))
4604 break;
4605 }
4606 r = *rm;
4607 a = *am;
4608 ruleset = *rsm;
4609
4610 REASON_SET(&reason, PFRES_MATCH);
4611
4612 if (r->log)
4613 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset,
4614 pd);
4615
4616 if (r->action != PF_PASS)
4617 return (PF_DROP);
4618
4619 if (pf_tag_packet(m, pd->pf_mtag, tag, -1)) {
4620 REASON_SET(&reason, PFRES_MEMORY);
4621 return (PF_DROP);
4622 }
4623
4624 return (PF_PASS);
4625}
4626
4627int
4628pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
4629 struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
4630 u_short *reason)
4631{
4632 struct pf_state_cmp key;
4633 struct tcphdr *th = pd->hdr.tcp;
4634 u_int16_t win = ntohs(th->th_win);
4635 u_int32_t ack, end, seq, orig_seq;
4636 u_int8_t sws, dws;
4637 int ackskew;
4638 int copyback = 0;
4639 struct pf_state_peer *src, *dst;
4640
4641 key.af = pd->af;
4642 key.proto = IPPROTO_TCP;
4643 if (direction == PF_IN) {
4644 PF_ACPY(&key.ext.addr, pd->src, key.af);
4645 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4646 key.ext.port = th->th_sport;
4647 key.gwy.port = th->th_dport;
4648 } else {
4649 PF_ACPY(&key.lan.addr, pd->src, key.af);
4650 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4651 key.lan.port = th->th_sport;
4652 key.ext.port = th->th_dport;
4653 }
4654
4655 STATE_LOOKUP();
4656
4657 if (direction == (*state)->direction) {
4658 src = &(*state)->src;
4659 dst = &(*state)->dst;
4660 } else {
4661 src = &(*state)->dst;
4662 dst = &(*state)->src;
4663 }
4664
4665 if ((*state)->src.state == PF_TCPS_PROXY_SRC) {
4666 if (direction != (*state)->direction) {
4667 REASON_SET(reason, PFRES_SYNPROXY);
4668 return (PF_SYNPROXY_DROP);
4669 }
4670 if (th->th_flags & TH_SYN) {
4671 if (ntohl(th->th_seq) != (*state)->src.seqlo) {
4672 REASON_SET(reason, PFRES_SYNPROXY);
4673 return (PF_DROP);
4674 }
4675#ifdef __FreeBSD__
4676 pf_send_tcp(NULL, (*state)->rule.ptr, pd->af, pd->dst,
4677#else
4678 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
4679#endif
4680 pd->src, th->th_dport, th->th_sport,
4681 (*state)->src.seqhi, ntohl(th->th_seq) + 1,
4682 TH_SYN|TH_ACK, 0, (*state)->src.mss, 0, 1,
4683 0, NULL, NULL);
4684 REASON_SET(reason, PFRES_SYNPROXY);
4685 return (PF_SYNPROXY_DROP);
4686 } else if (!(th->th_flags & TH_ACK) ||
4687 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
4688 (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
4689 REASON_SET(reason, PFRES_SYNPROXY);
4690 return (PF_DROP);
4691 } else if ((*state)->src_node != NULL &&
4692 pf_src_connlimit(state)) {
4693 REASON_SET(reason, PFRES_SRCLIMIT);
4694 return (PF_DROP);
4695 } else
4696 (*state)->src.state = PF_TCPS_PROXY_DST;
4697 }
4698 if ((*state)->src.state == PF_TCPS_PROXY_DST) {
4699 struct pf_state_host *src, *dst;
4700
4701 if (direction == PF_OUT) {
4702 src = &(*state)->gwy;
4703 dst = &(*state)->ext;
4704 } else {
4705 src = &(*state)->ext;
4706 dst = &(*state)->lan;
4707 }
4708 if (direction == (*state)->direction) {
4709 if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) ||
4710 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
4711 (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
4712 REASON_SET(reason, PFRES_SYNPROXY);
4713 return (PF_DROP);
4714 }
4715 (*state)->src.max_win = MAX(ntohs(th->th_win), 1);
4716 if ((*state)->dst.seqhi == 1)
4717 (*state)->dst.seqhi = htonl(arc4random());
4718#ifdef __FreeBSD__
4719 pf_send_tcp(NULL, (*state)->rule.ptr, pd->af,
4720 &src->addr,
4721#else
4722 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
4723#endif
4724 &dst->addr, src->port, dst->port,
4725 (*state)->dst.seqhi, 0, TH_SYN, 0,
4726 (*state)->src.mss, 0, 0, (*state)->tag, NULL, NULL);
4727 REASON_SET(reason, PFRES_SYNPROXY);
4728 return (PF_SYNPROXY_DROP);
4729 } else if (((th->th_flags & (TH_SYN|TH_ACK)) !=
4730 (TH_SYN|TH_ACK)) ||
4731 (ntohl(th->th_ack) != (*state)->dst.seqhi + 1)) {
4732 REASON_SET(reason, PFRES_SYNPROXY);
4733 return (PF_DROP);
4734 } else {
4735 (*state)->dst.max_win = MAX(ntohs(th->th_win), 1);
4736 (*state)->dst.seqlo = ntohl(th->th_seq);
4737#ifdef __FreeBSD__
4738 pf_send_tcp(NULL, (*state)->rule.ptr, pd->af, pd->dst,
4739#else
4740 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
4741#endif
4742 pd->src, th->th_dport, th->th_sport,
4743 ntohl(th->th_ack), ntohl(th->th_seq) + 1,
4744 TH_ACK, (*state)->src.max_win, 0, 0, 0,
4745 (*state)->tag, NULL, NULL);
4746#ifdef __FreeBSD__
4747 pf_send_tcp(NULL, (*state)->rule.ptr, pd->af,
4748 &src->addr,
4749#else
4750 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
4751#endif
4752 &dst->addr, src->port, dst->port,
4753 (*state)->src.seqhi + 1, (*state)->src.seqlo + 1,
4754 TH_ACK, (*state)->dst.max_win, 0, 0, 1,
4755 0, NULL, NULL);
4756 (*state)->src.seqdiff = (*state)->dst.seqhi -
4757 (*state)->src.seqlo;
4758 (*state)->dst.seqdiff = (*state)->src.seqhi -
4759 (*state)->dst.seqlo;
4760 (*state)->src.seqhi = (*state)->src.seqlo +
4761 (*state)->dst.max_win;
4762 (*state)->dst.seqhi = (*state)->dst.seqlo +
4763 (*state)->src.max_win;
4764 (*state)->src.wscale = (*state)->dst.wscale = 0;
4765 (*state)->src.state = (*state)->dst.state =
4766 TCPS_ESTABLISHED;
4767 REASON_SET(reason, PFRES_SYNPROXY);
4768 return (PF_SYNPROXY_DROP);
4769 }
4770 }
4771
4772 if (((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN) &&
4773 dst->state >= TCPS_FIN_WAIT_2 &&
4774 src->state >= TCPS_FIN_WAIT_2) {
4775 if (pf_status.debug >= PF_DEBUG_MISC) {
4776 printf("pf: state reuse ");
4777 pf_print_state(*state);
4778 pf_print_flags(th->th_flags);
4779 printf("\n");
4780 }
4781 /* XXX make sure it's the same direction ?? */
4782 (*state)->src.state = (*state)->dst.state = TCPS_CLOSED;
4783 pf_unlink_state(*state);
4784 *state = NULL;
4785 return (PF_DROP);
4786 }
4787
4788 if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) {
4789 sws = src->wscale & PF_WSCALE_MASK;
4790 dws = dst->wscale & PF_WSCALE_MASK;
4791 } else
4792 sws = dws = 0;
4793
4794 /*
4795 * Sequence tracking algorithm from Guido van Rooij's paper:
4796 * http://www.madison-gurkha.com/publications/tcp_filtering/
4797 * tcp_filtering.ps
4798 */
4799
4800 orig_seq = seq = ntohl(th->th_seq);
4801 if (src->seqlo == 0) {
4802 /* First packet from this end. Set its state */
4803
4804 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) &&
4805 src->scrub == NULL) {
4806 if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) {
4807 REASON_SET(reason, PFRES_MEMORY);
4808 return (PF_DROP);
4809 }
4810 }
4811
4812 /* Deferred generation of sequence number modulator */
4813 if (dst->seqdiff && !src->seqdiff) {
4814#ifdef __FreeBSD__
4815 while ((src->seqdiff = pf_new_isn(*state) - seq) == 0)
4816 ;
4817#else
4818 while ((src->seqdiff = tcp_rndiss_next() - seq) == 0)
4819 ;
4820#endif
4821 ack = ntohl(th->th_ack) - dst->seqdiff;
4822 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
4823 src->seqdiff), 0);
4824 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
4825 copyback = 1;
4826 } else {
4827 ack = ntohl(th->th_ack);
4828 }
4829
4830 end = seq + pd->p_len;
4831 if (th->th_flags & TH_SYN) {
4832 end++;
4833 if (dst->wscale & PF_WSCALE_FLAG) {
4834 src->wscale = pf_get_wscale(m, off, th->th_off,
4835 pd->af);
4836 if (src->wscale & PF_WSCALE_FLAG) {
4837 /* Remove scale factor from initial
4838 * window */
4839 sws = src->wscale & PF_WSCALE_MASK;
4840 win = ((u_int32_t)win + (1 << sws) - 1)
4841 >> sws;
4842 dws = dst->wscale & PF_WSCALE_MASK;
4843 } else {
4844 /* fixup other window */
4845 dst->max_win <<= dst->wscale &
4846 PF_WSCALE_MASK;
4847 /* in case of a retrans SYN|ACK */
4848 dst->wscale = 0;
4849 }
4850 }
4851 }
4852 if (th->th_flags & TH_FIN)
4853 end++;
4854
4855 src->seqlo = seq;
4856 if (src->state < TCPS_SYN_SENT)
4857 src->state = TCPS_SYN_SENT;
4858
4859 /*
4860 * May need to slide the window (seqhi may have been set by
4861 * the crappy stack check or if we picked up the connection
4862 * after establishment)
4863 */
4864 if (src->seqhi == 1 ||
4865 SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi))
4866 src->seqhi = end + MAX(1, dst->max_win << dws);
4867 if (win > src->max_win)
4868 src->max_win = win;
4869
4870 } else {
4871 ack = ntohl(th->th_ack) - dst->seqdiff;
4872 if (src->seqdiff) {
4873 /* Modulate sequence numbers */
4874 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
4875 src->seqdiff), 0);
4876 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
4877 copyback = 1;
4878 }
4879 end = seq + pd->p_len;
4880 if (th->th_flags & TH_SYN)
4881 end++;
4882 if (th->th_flags & TH_FIN)
4883 end++;
4884 }
4885
4886 if ((th->th_flags & TH_ACK) == 0) {
4887 /* Let it pass through the ack skew check */
4888 ack = dst->seqlo;
4889 } else if ((ack == 0 &&
4890 (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) ||
4891 /* broken tcp stacks do not set ack */
4892 (dst->state < TCPS_SYN_SENT)) {
4893 /*
4894 * Many stacks (ours included) will set the ACK number in an
4895 * FIN|ACK if the SYN times out -- no sequence to ACK.
4896 */
4897 ack = dst->seqlo;
4898 }
4899
4900 if (seq == end) {
4901 /* Ease sequencing restrictions on no data packets */
4902 seq = src->seqlo;
4903 end = seq;
4904 }
4905
4906 ackskew = dst->seqlo - ack;
4907
4908
4909 /*
4910 * Need to demodulate the sequence numbers in any TCP SACK options
4911 * (Selective ACK). We could optionally validate the SACK values
4912 * against the current ACK window, either forwards or backwards, but
4913 * I'm not confident that SACK has been implemented properly
4914 * everywhere. It wouldn't surprise me if several stacks accidently
4915 * SACK too far backwards of previously ACKed data. There really aren't
4916 * any security implications of bad SACKing unless the target stack
4917 * doesn't validate the option length correctly. Someone trying to
4918 * spoof into a TCP connection won't bother blindly sending SACK
4919 * options anyway.
4920 */
4921 if (dst->seqdiff && (th->th_off << 2) > sizeof(struct tcphdr)) {
4922 if (pf_modulate_sack(m, off, pd, th, dst))
4923 copyback = 1;
4924 }
4925
4926
4927#define MAXACKWINDOW (0xffff + 1500) /* 1500 is an arbitrary fudge factor */
4928 if (SEQ_GEQ(src->seqhi, end) &&
4929 /* Last octet inside other's window space */
4930 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) &&
4931 /* Retrans: not more than one window back */
4932 (ackskew >= -MAXACKWINDOW) &&
4933 /* Acking not more than one reassembled fragment backwards */
4934 (ackskew <= (MAXACKWINDOW << sws)) &&
4935 /* Acking not more than one window forward */
4936 ((th->th_flags & TH_RST) == 0 || orig_seq == src->seqlo ||
4937 (orig_seq == src->seqlo + 1) || (pd->flags & PFDESC_IP_REAS) == 0)) {
4938 /* Require an exact/+1 sequence match on resets when possible */
4939
4940 if (dst->scrub || src->scrub) {
4941 if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4942 *state, src, dst, ©back))
4943 return (PF_DROP);
4944 }
4945
4946 /* update max window */
4947 if (src->max_win < win)
4948 src->max_win = win;
4949 /* synchronize sequencing */
4950 if (SEQ_GT(end, src->seqlo))
4951 src->seqlo = end;
4952 /* slide the window of what the other end can send */
4953 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4954 dst->seqhi = ack + MAX((win << sws), 1);
4955
4956
4957 /* update states */
4958 if (th->th_flags & TH_SYN)
4959 if (src->state < TCPS_SYN_SENT)
4960 src->state = TCPS_SYN_SENT;
4961 if (th->th_flags & TH_FIN)
4962 if (src->state < TCPS_CLOSING)
4963 src->state = TCPS_CLOSING;
4964 if (th->th_flags & TH_ACK) {
4965 if (dst->state == TCPS_SYN_SENT) {
4966 dst->state = TCPS_ESTABLISHED;
4967 if (src->state == TCPS_ESTABLISHED &&
4968 (*state)->src_node != NULL &&
4969 pf_src_connlimit(state)) {
4970 REASON_SET(reason, PFRES_SRCLIMIT);
4971 return (PF_DROP);
4972 }
4973 } else if (dst->state == TCPS_CLOSING)
4974 dst->state = TCPS_FIN_WAIT_2;
4975 }
4976 if (th->th_flags & TH_RST)
4977 src->state = dst->state = TCPS_TIME_WAIT;
4978
4979 /* update expire time */
4980 (*state)->expire = time_second;
4981 if (src->state >= TCPS_FIN_WAIT_2 &&
4982 dst->state >= TCPS_FIN_WAIT_2)
4983 (*state)->timeout = PFTM_TCP_CLOSED;
4984 else if (src->state >= TCPS_CLOSING &&
4985 dst->state >= TCPS_CLOSING)
4986 (*state)->timeout = PFTM_TCP_FIN_WAIT;
4987 else if (src->state < TCPS_ESTABLISHED ||
4988 dst->state < TCPS_ESTABLISHED)
4989 (*state)->timeout = PFTM_TCP_OPENING;
4990 else if (src->state >= TCPS_CLOSING ||
4991 dst->state >= TCPS_CLOSING)
4992 (*state)->timeout = PFTM_TCP_CLOSING;
4993 else
4994 (*state)->timeout = PFTM_TCP_ESTABLISHED;
4995
4996 /* Fall through to PASS packet */
4997
4998 } else if ((dst->state < TCPS_SYN_SENT ||
4999 dst->state >= TCPS_FIN_WAIT_2 ||
5000 src->state >= TCPS_FIN_WAIT_2) &&
5001 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) &&
5002 /* Within a window forward of the originating packet */
5003 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) {
5004 /* Within a window backward of the originating packet */
5005
5006 /*
5007 * This currently handles three situations:
5008 * 1) Stupid stacks will shotgun SYNs before their peer
5009 * replies.
5010 * 2) When PF catches an already established stream (the
5011 * firewall rebooted, the state table was flushed, routes
5012 * changed...)
5013 * 3) Packets get funky immediately after the connection
5014 * closes (this should catch Solaris spurious ACK|FINs
5015 * that web servers like to spew after a close)
5016 *
5017 * This must be a little more careful than the above code
5018 * since packet floods will also be caught here. We don't
5019 * update the TTL here to mitigate the damage of a packet
5020 * flood and so the same code can handle awkward establishment
5021 * and a loosened connection close.
5022 * In the establishment case, a correct peer response will
5023 * validate the connection, go through the normal state code
5024 * and keep updating the state TTL.
5025 */
5026
5027 if (pf_status.debug >= PF_DEBUG_MISC) {
5028 printf("pf: loose state match: ");
5029 pf_print_state(*state);
5030 pf_print_flags(th->th_flags);
5031 printf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
5032 "pkts=%llu:%llu\n", seq, orig_seq, ack, pd->p_len,
5033#ifdef __FreeBSD__
5034 ackskew, (unsigned long long)(*state)->packets[0],
5035 (unsigned long long)(*state)->packets[1]);
5036#else
5037 ackskew, (*state)->packets[0],
5038 (*state)->packets[1]);
5039#endif
5040 }
5041
5042 if (dst->scrub || src->scrub) {
5043 if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
5044 *state, src, dst, ©back))
5045 return (PF_DROP);
5046 }
5047
5048 /* update max window */
5049 if (src->max_win < win)
5050 src->max_win = win;
5051 /* synchronize sequencing */
5052 if (SEQ_GT(end, src->seqlo))
5053 src->seqlo = end;
5054 /* slide the window of what the other end can send */
5055 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
5056 dst->seqhi = ack + MAX((win << sws), 1);
5057
5058 /*
5059 * Cannot set dst->seqhi here since this could be a shotgunned
5060 * SYN and not an already established connection.
5061 */
5062
5063 if (th->th_flags & TH_FIN)
5064 if (src->state < TCPS_CLOSING)
5065 src->state = TCPS_CLOSING;
5066 if (th->th_flags & TH_RST)
5067 src->state = dst->state = TCPS_TIME_WAIT;
5068
5069 /* Fall through to PASS packet */
5070
5071 } else {
5072 if ((*state)->dst.state == TCPS_SYN_SENT &&
5073 (*state)->src.state == TCPS_SYN_SENT) {
5074 /* Send RST for state mismatches during handshake */
5075 if (!(th->th_flags & TH_RST))
5076#ifdef __FreeBSD__
5077 pf_send_tcp(m, (*state)->rule.ptr, pd->af,
5078#else
5079 pf_send_tcp((*state)->rule.ptr, pd->af,
5080#endif
5081 pd->dst, pd->src, th->th_dport,
5082 th->th_sport, ntohl(th->th_ack), 0,
5083 TH_RST, 0, 0,
5084 (*state)->rule.ptr->return_ttl, 1, 0,
5085 pd->eh, kif->pfik_ifp);
5086 src->seqlo = 0;
5087 src->seqhi = 1;
5088 src->max_win = 1;
5089 } else if (pf_status.debug >= PF_DEBUG_MISC) {
5090 printf("pf: BAD state: ");
5091 pf_print_state(*state);
5092 pf_print_flags(th->th_flags);
5093 printf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
5094 "pkts=%llu:%llu dir=%s,%s\n",
5095 seq, orig_seq, ack, pd->p_len, ackskew,
5096#ifdef __FreeBSD__
5097 (unsigned long long)(*state)->packets[0],
5098 (unsigned long long)(*state)->packets[1],
5099#else
5100 (*state)->packets[0], (*state)->packets[1],
5101#endif
5102 direction == PF_IN ? "in" : "out",
5103 direction == (*state)->direction ? "fwd" : "rev");
5104 printf("pf: State failure on: %c %c %c %c | %c %c\n",
5105 SEQ_GEQ(src->seqhi, end) ? ' ' : '1',
5106 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ?
5107 ' ': '2',
5108 (ackskew >= -MAXACKWINDOW) ? ' ' : '3',
5109 (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4',
5110 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5',
5111 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6');
5112 }
5113 REASON_SET(reason, PFRES_BADSTATE);
5114 return (PF_DROP);
5115 }
5116
5117 /* Any packets which have gotten here are to be passed */
5118
5119 /* translate source/destination address, if necessary */
5120 if (STATE_TRANSLATE(*state)) {
5121 if (direction == PF_OUT)
5122 pf_change_ap(pd->src, &th->th_sport, pd->ip_sum,
5123 &th->th_sum, &(*state)->gwy.addr,
5124 (*state)->gwy.port, 0, pd->af);
5125 else
5126 pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum,
5127 &th->th_sum, &(*state)->lan.addr,
5128 (*state)->lan.port, 0, pd->af);
5129 m_copyback(m, off, sizeof(*th), (caddr_t)th);
5130 } else if (copyback) {
5131 /* Copyback sequence modulation or stateful scrub changes */
5132 m_copyback(m, off, sizeof(*th), (caddr_t)th);
5133 }
5134
5135 return (PF_PASS);
5136}
5137
5138int
5139pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif,
5140 struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
5141{
5142 struct pf_state_peer *src, *dst;
5143 struct pf_state_cmp key;
5144 struct udphdr *uh = pd->hdr.udp;
5145
5146 key.af = pd->af;
5147 key.proto = IPPROTO_UDP;
5148 if (direction == PF_IN) {
5149 PF_ACPY(&key.ext.addr, pd->src, key.af);
5150 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
5151 key.ext.port = uh->uh_sport;
5152 key.gwy.port = uh->uh_dport;
5153 } else {
5154 PF_ACPY(&key.lan.addr, pd->src, key.af);
5155 PF_ACPY(&key.ext.addr, pd->dst, key.af);
5156 key.lan.port = uh->uh_sport;
5157 key.ext.port = uh->uh_dport;
5158 }
5159
5160 STATE_LOOKUP();
5161
5162 if (direction == (*state)->direction) {
5163 src = &(*state)->src;
5164 dst = &(*state)->dst;
5165 } else {
5166 src = &(*state)->dst;
5167 dst = &(*state)->src;
5168 }
5169
5170 /* update states */
5171 if (src->state < PFUDPS_SINGLE)
5172 src->state = PFUDPS_SINGLE;
5173 if (dst->state == PFUDPS_SINGLE)
5174 dst->state = PFUDPS_MULTIPLE;
5175
5176 /* update expire time */
5177 (*state)->expire = time_second;
5178 if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE)
5179 (*state)->timeout = PFTM_UDP_MULTIPLE;
5180 else
5181 (*state)->timeout = PFTM_UDP_SINGLE;
5182
5183 /* translate source/destination address, if necessary */
5184 if (STATE_TRANSLATE(*state)) {
5185 if (direction == PF_OUT)
5186 pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum,
5187 &uh->uh_sum, &(*state)->gwy.addr,
5188 (*state)->gwy.port, 1, pd->af);
5189 else
5190 pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum,
5191 &uh->uh_sum, &(*state)->lan.addr,
5192 (*state)->lan.port, 1, pd->af);
5193 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
5194 }
5195
5196 return (PF_PASS);
5197}
5198
5199int
5200pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif,
5201 struct mbuf *m, int off, void *h, struct pf_pdesc *pd, u_short *reason)
5202{
5203 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
5204 u_int16_t icmpid = 0; /* make the compiler happy */
5205 u_int16_t *icmpsum = NULL; /* make the compiler happy */
5206 u_int8_t icmptype = 0; /* make the compiler happy */
5207 int state_icmp = 0;
5208 struct pf_state_cmp key;
5209
5210 switch (pd->proto) {
5211#ifdef INET
5212 case IPPROTO_ICMP:
5213 icmptype = pd->hdr.icmp->icmp_type;
5214 icmpid = pd->hdr.icmp->icmp_id;
5215 icmpsum = &pd->hdr.icmp->icmp_cksum;
5216
5217 if (icmptype == ICMP_UNREACH ||
5218 icmptype == ICMP_SOURCEQUENCH ||
5219 icmptype == ICMP_REDIRECT ||
5220 icmptype == ICMP_TIMXCEED ||
5221 icmptype == ICMP_PARAMPROB)
5222 state_icmp++;
5223 break;
5224#endif /* INET */
5225#ifdef INET6
5226 case IPPROTO_ICMPV6:
5227 icmptype = pd->hdr.icmp6->icmp6_type;
5228 icmpid = pd->hdr.icmp6->icmp6_id;
5229 icmpsum = &pd->hdr.icmp6->icmp6_cksum;
5230
5231 if (icmptype == ICMP6_DST_UNREACH ||
5232 icmptype == ICMP6_PACKET_TOO_BIG ||
5233 icmptype == ICMP6_TIME_EXCEEDED ||
5234 icmptype == ICMP6_PARAM_PROB)
5235 state_icmp++;
5236 break;
5237#endif /* INET6 */
5238 }
5239
5240 if (!state_icmp) {
5241
5242 /*
5243 * ICMP query/reply message not related to a TCP/UDP packet.
5244 * Search for an ICMP state.
5245 */
5246 key.af = pd->af;
5247 key.proto = pd->proto;
5248 if (direction == PF_IN) {
5249 PF_ACPY(&key.ext.addr, pd->src, key.af);
5250 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
5251 key.ext.port = 0;
5252 key.gwy.port = icmpid;
5253 } else {
5254 PF_ACPY(&key.lan.addr, pd->src, key.af);
5255 PF_ACPY(&key.ext.addr, pd->dst, key.af);
5256 key.lan.port = icmpid;
5257 key.ext.port = 0;
5258 }
5259
5260 STATE_LOOKUP();
5261
5262 (*state)->expire = time_second;
5263 (*state)->timeout = PFTM_ICMP_ERROR_REPLY;
5264
5265 /* translate source/destination address, if necessary */
5266 if (STATE_TRANSLATE(*state)) {
5267 if (direction == PF_OUT) {
5268 switch (pd->af) {
5269#ifdef INET
5270 case AF_INET:
5271 pf_change_a(&saddr->v4.s_addr,
5272 pd->ip_sum,
5273 (*state)->gwy.addr.v4.s_addr, 0);
5274 pd->hdr.icmp->icmp_cksum =
5275 pf_cksum_fixup(
5276 pd->hdr.icmp->icmp_cksum, icmpid,
5277 (*state)->gwy.port, 0);
5278 pd->hdr.icmp->icmp_id =
5279 (*state)->gwy.port;
5280 m_copyback(m, off, ICMP_MINLEN,
5281 (caddr_t)pd->hdr.icmp);
5282 break;
5283#endif /* INET */
5284#ifdef INET6
5285 case AF_INET6:
5286 pf_change_a6(saddr,
5287 &pd->hdr.icmp6->icmp6_cksum,
5288 &(*state)->gwy.addr, 0);
5289 m_copyback(m, off,
5290 sizeof(struct icmp6_hdr),
5291 (caddr_t)pd->hdr.icmp6);
5292 break;
5293#endif /* INET6 */
5294 }
5295 } else {
5296 switch (pd->af) {
5297#ifdef INET
5298 case AF_INET:
5299 pf_change_a(&daddr->v4.s_addr,
5300 pd->ip_sum,
5301 (*state)->lan.addr.v4.s_addr, 0);
5302 pd->hdr.icmp->icmp_cksum =
5303 pf_cksum_fixup(
5304 pd->hdr.icmp->icmp_cksum, icmpid,
5305 (*state)->lan.port, 0);
5306 pd->hdr.icmp->icmp_id =
5307 (*state)->lan.port;
5308 m_copyback(m, off, ICMP_MINLEN,
5309 (caddr_t)pd->hdr.icmp);
5310 break;
5311#endif /* INET */
5312#ifdef INET6
5313 case AF_INET6:
5314 pf_change_a6(daddr,
5315 &pd->hdr.icmp6->icmp6_cksum,
5316 &(*state)->lan.addr, 0);
5317 m_copyback(m, off,
5318 sizeof(struct icmp6_hdr),
5319 (caddr_t)pd->hdr.icmp6);
5320 break;
5321#endif /* INET6 */
5322 }
5323 }
5324 }
5325
5326 return (PF_PASS);
5327
5328 } else {
5329 /*
5330 * ICMP error message in response to a TCP/UDP packet.
5331 * Extract the inner TCP/UDP header and search for that state.
5332 */
5333
5334 struct pf_pdesc pd2;
5335#ifdef INET
5336 struct ip h2;
5337#endif /* INET */
5338#ifdef INET6
5339 struct ip6_hdr h2_6;
5340 int terminal = 0;
5341#endif /* INET6 */
5342 int ipoff2 = 0; /* make the compiler happy */
5343 int off2 = 0; /* make the compiler happy */
5344
5345 pd2.af = pd->af;
5346 switch (pd->af) {
5347#ifdef INET
5348 case AF_INET:
5349 /* offset of h2 in mbuf chain */
5350 ipoff2 = off + ICMP_MINLEN;
5351
5352 if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2),
5353 NULL, reason, pd2.af)) {
5354 DPFPRINTF(PF_DEBUG_MISC,
5355 ("pf: ICMP error message too short "
5356 "(ip)\n"));
5357 return (PF_DROP);
5358 }
5359 /*
5360 * ICMP error messages don't refer to non-first
5361 * fragments
5362 */
5363 if (h2.ip_off & htons(IP_OFFMASK)) {
5364 REASON_SET(reason, PFRES_FRAG);
5365 return (PF_DROP);
5366 }
5367
5368 /* offset of protocol header that follows h2 */
5369 off2 = ipoff2 + (h2.ip_hl << 2);
5370
5371 pd2.proto = h2.ip_p;
5372 pd2.src = (struct pf_addr *)&h2.ip_src;
5373 pd2.dst = (struct pf_addr *)&h2.ip_dst;
5374 pd2.ip_sum = &h2.ip_sum;
5375 break;
5376#endif /* INET */
5377#ifdef INET6
5378 case AF_INET6:
5379 ipoff2 = off + sizeof(struct icmp6_hdr);
5380
5381 if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6),
5382 NULL, reason, pd2.af)) {
5383 DPFPRINTF(PF_DEBUG_MISC,
5384 ("pf: ICMP error message too short "
5385 "(ip6)\n"));
5386 return (PF_DROP);
5387 }
5388 pd2.proto = h2_6.ip6_nxt;
5389 pd2.src = (struct pf_addr *)&h2_6.ip6_src;
5390 pd2.dst = (struct pf_addr *)&h2_6.ip6_dst;
5391 pd2.ip_sum = NULL;
5392 off2 = ipoff2 + sizeof(h2_6);
5393 do {
5394 switch (pd2.proto) {
5395 case IPPROTO_FRAGMENT:
5396 /*
5397 * ICMPv6 error messages for
5398 * non-first fragments
5399 */
5400 REASON_SET(reason, PFRES_FRAG);
5401 return (PF_DROP);
5402 case IPPROTO_AH:
5403 case IPPROTO_HOPOPTS:
5404 case IPPROTO_ROUTING:
5405 case IPPROTO_DSTOPTS: {
5406 /* get next header and header length */
5407 struct ip6_ext opt6;
5408
5409 if (!pf_pull_hdr(m, off2, &opt6,
5410 sizeof(opt6), NULL, reason,
5411 pd2.af)) {
5412 DPFPRINTF(PF_DEBUG_MISC,
5413 ("pf: ICMPv6 short opt\n"));
5414 return (PF_DROP);
5415 }
5416 if (pd2.proto == IPPROTO_AH)
5417 off2 += (opt6.ip6e_len + 2) * 4;
5418 else
5419 off2 += (opt6.ip6e_len + 1) * 8;
5420 pd2.proto = opt6.ip6e_nxt;
5421 /* goto the next header */
5422 break;
5423 }
5424 default:
5425 terminal++;
5426 break;
5427 }
5428 } while (!terminal);
5429 break;
5430#endif /* INET6 */
5431#ifdef __FreeBSD__
5432 default:
5433 panic("AF not supported: %d", pd->af);
5434#endif
5435 }
5436
5437 switch (pd2.proto) {
5438 case IPPROTO_TCP: {
5439 struct tcphdr th;
5440 u_int32_t seq;
5441 struct pf_state_peer *src, *dst;
5442 u_int8_t dws;
5443 int copyback = 0;
5444
5445 /*
5446 * Only the first 8 bytes of the TCP header can be
5447 * expected. Don't access any TCP header fields after
5448 * th_seq, an ackskew test is not possible.
5449 */
5450 if (!pf_pull_hdr(m, off2, &th, 8, NULL, reason,
5451 pd2.af)) {
5452 DPFPRINTF(PF_DEBUG_MISC,
5453 ("pf: ICMP error message too short "
5454 "(tcp)\n"));
5455 return (PF_DROP);
5456 }
5457
5458 key.af = pd2.af;
5459 key.proto = IPPROTO_TCP;
5460 if (direction == PF_IN) {
5461 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
5462 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
5463 key.ext.port = th.th_dport;
5464 key.gwy.port = th.th_sport;
5465 } else {
5466 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
5467 PF_ACPY(&key.ext.addr, pd2.src, key.af);
5468 key.lan.port = th.th_dport;
5469 key.ext.port = th.th_sport;
5470 }
5471
5472 STATE_LOOKUP();
5473
5474 if (direction == (*state)->direction) {
5475 src = &(*state)->dst;
5476 dst = &(*state)->src;
5477 } else {
5478 src = &(*state)->src;
5479 dst = &(*state)->dst;
5480 }
5481
5482 if (src->wscale && dst->wscale)
5483 dws = dst->wscale & PF_WSCALE_MASK;
5484 else
5485 dws = 0;
5486
5487 /* Demodulate sequence number */
5488 seq = ntohl(th.th_seq) - src->seqdiff;
5489 if (src->seqdiff) {
5490 pf_change_a(&th.th_seq, icmpsum,
5491 htonl(seq), 0);
5492 copyback = 1;
5493 }
5494
5495 if (!SEQ_GEQ(src->seqhi, seq) ||
5496 !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws))) {
5497 if (pf_status.debug >= PF_DEBUG_MISC) {
5498 printf("pf: BAD ICMP %d:%d ",
5499 icmptype, pd->hdr.icmp->icmp_code);
5500 pf_print_host(pd->src, 0, pd->af);
5501 printf(" -> ");
5502 pf_print_host(pd->dst, 0, pd->af);
5503 printf(" state: ");
5504 pf_print_state(*state);
5505 printf(" seq=%u\n", seq);
5506 }
5507 REASON_SET(reason, PFRES_BADSTATE);
5508 return (PF_DROP);
5509 }
5510
5511 if (STATE_TRANSLATE(*state)) {
5512 if (direction == PF_IN) {
5513 pf_change_icmp(pd2.src, &th.th_sport,
5514 daddr, &(*state)->lan.addr,
5515 (*state)->lan.port, NULL,
5516 pd2.ip_sum, icmpsum,
5517 pd->ip_sum, 0, pd2.af);
5518 } else {
5519 pf_change_icmp(pd2.dst, &th.th_dport,
5520 saddr, &(*state)->gwy.addr,
5521 (*state)->gwy.port, NULL,
5522 pd2.ip_sum, icmpsum,
5523 pd->ip_sum, 0, pd2.af);
5524 }
5525 copyback = 1;
5526 }
5527
5528 if (copyback) {
5529 switch (pd2.af) {
5530#ifdef INET
5531 case AF_INET:
5532 m_copyback(m, off, ICMP_MINLEN,
5533 (caddr_t)pd->hdr.icmp);
5534 m_copyback(m, ipoff2, sizeof(h2),
5535 (caddr_t)&h2);
5536 break;
5537#endif /* INET */
5538#ifdef INET6
5539 case AF_INET6:
5540 m_copyback(m, off,
5541 sizeof(struct icmp6_hdr),
5542 (caddr_t)pd->hdr.icmp6);
5543 m_copyback(m, ipoff2, sizeof(h2_6),
5544 (caddr_t)&h2_6);
5545 break;
5546#endif /* INET6 */
5547 }
5548 m_copyback(m, off2, 8, (caddr_t)&th);
5549 }
5550
5551 return (PF_PASS);
5552 break;
5553 }
5554 case IPPROTO_UDP: {
5555 struct udphdr uh;
5556
5557 if (!pf_pull_hdr(m, off2, &uh, sizeof(uh),
5558 NULL, reason, pd2.af)) {
5559 DPFPRINTF(PF_DEBUG_MISC,
5560 ("pf: ICMP error message too short "
5561 "(udp)\n"));
5562 return (PF_DROP);
5563 }
5564
5565 key.af = pd2.af;
5566 key.proto = IPPROTO_UDP;
5567 if (direction == PF_IN) {
5568 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
5569 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
5570 key.ext.port = uh.uh_dport;
5571 key.gwy.port = uh.uh_sport;
5572 } else {
5573 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
5574 PF_ACPY(&key.ext.addr, pd2.src, key.af);
5575 key.lan.port = uh.uh_dport;
5576 key.ext.port = uh.uh_sport;
5577 }
5578
5579 STATE_LOOKUP();
5580
5581 if (STATE_TRANSLATE(*state)) {
5582 if (direction == PF_IN) {
5583 pf_change_icmp(pd2.src, &uh.uh_sport,
5584 daddr, &(*state)->lan.addr,
5585 (*state)->lan.port, &uh.uh_sum,
5586 pd2.ip_sum, icmpsum,
5587 pd->ip_sum, 1, pd2.af);
5588 } else {
5589 pf_change_icmp(pd2.dst, &uh.uh_dport,
5590 saddr, &(*state)->gwy.addr,
5591 (*state)->gwy.port, &uh.uh_sum,
5592 pd2.ip_sum, icmpsum,
5593 pd->ip_sum, 1, pd2.af);
5594 }
5595 switch (pd2.af) {
5596#ifdef INET
5597 case AF_INET:
5598 m_copyback(m, off, ICMP_MINLEN,
5599 (caddr_t)pd->hdr.icmp);
5600 m_copyback(m, ipoff2, sizeof(h2),
5601 (caddr_t)&h2);
5602 break;
5603#endif /* INET */
5604#ifdef INET6
5605 case AF_INET6:
5606 m_copyback(m, off,
5607 sizeof(struct icmp6_hdr),
5608 (caddr_t)pd->hdr.icmp6);
5609 m_copyback(m, ipoff2, sizeof(h2_6),
5610 (caddr_t)&h2_6);
5611 break;
5612#endif /* INET6 */
5613 }
5614 m_copyback(m, off2, sizeof(uh),
5615 (caddr_t)&uh);
5616 }
5617
5618 return (PF_PASS);
5619 break;
5620 }
5621#ifdef INET
5622 case IPPROTO_ICMP: {
5623 struct icmp iih;
5624
5625 if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN,
5626 NULL, reason, pd2.af)) {
5627 DPFPRINTF(PF_DEBUG_MISC,
5628 ("pf: ICMP error message too short i"
5629 "(icmp)\n"));
5630 return (PF_DROP);
5631 }
5632
5633 key.af = pd2.af;
5634 key.proto = IPPROTO_ICMP;
5635 if (direction == PF_IN) {
5636 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
5637 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
5638 key.ext.port = 0;
5639 key.gwy.port = iih.icmp_id;
5640 } else {
5641 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
5642 PF_ACPY(&key.ext.addr, pd2.src, key.af);
5643 key.lan.port = iih.icmp_id;
5644 key.ext.port = 0;
5645 }
5646
5647 STATE_LOOKUP();
5648
5649 if (STATE_TRANSLATE(*state)) {
5650 if (direction == PF_IN) {
5651 pf_change_icmp(pd2.src, &iih.icmp_id,
5652 daddr, &(*state)->lan.addr,
5653 (*state)->lan.port, NULL,
5654 pd2.ip_sum, icmpsum,
5655 pd->ip_sum, 0, AF_INET);
5656 } else {
5657 pf_change_icmp(pd2.dst, &iih.icmp_id,
5658 saddr, &(*state)->gwy.addr,
5659 (*state)->gwy.port, NULL,
5660 pd2.ip_sum, icmpsum,
5661 pd->ip_sum, 0, AF_INET);
5662 }
5663 m_copyback(m, off, ICMP_MINLEN,
5664 (caddr_t)pd->hdr.icmp);
5665 m_copyback(m, ipoff2, sizeof(h2),
5666 (caddr_t)&h2);
5667 m_copyback(m, off2, ICMP_MINLEN,
5668 (caddr_t)&iih);
5669 }
5670
5671 return (PF_PASS);
5672 break;
5673 }
5674#endif /* INET */
5675#ifdef INET6
5676 case IPPROTO_ICMPV6: {
5677 struct icmp6_hdr iih;
5678
5679 if (!pf_pull_hdr(m, off2, &iih,
5680 sizeof(struct icmp6_hdr), NULL, reason, pd2.af)) {
5681 DPFPRINTF(PF_DEBUG_MISC,
5682 ("pf: ICMP error message too short "
5683 "(icmp6)\n"));
5684 return (PF_DROP);
5685 }
5686
5687 key.af = pd2.af;
5688 key.proto = IPPROTO_ICMPV6;
5689 if (direction == PF_IN) {
5690 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
5691 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
5692 key.ext.port = 0;
5693 key.gwy.port = iih.icmp6_id;
5694 } else {
5695 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
5696 PF_ACPY(&key.ext.addr, pd2.src, key.af);
5697 key.lan.port = iih.icmp6_id;
5698 key.ext.port = 0;
5699 }
5700
5701 STATE_LOOKUP();
5702
5703 if (STATE_TRANSLATE(*state)) {
5704 if (direction == PF_IN) {
5705 pf_change_icmp(pd2.src, &iih.icmp6_id,
5706 daddr, &(*state)->lan.addr,
5707 (*state)->lan.port, NULL,
5708 pd2.ip_sum, icmpsum,
5709 pd->ip_sum, 0, AF_INET6);
5710 } else {
5711 pf_change_icmp(pd2.dst, &iih.icmp6_id,
5712 saddr, &(*state)->gwy.addr,
5713 (*state)->gwy.port, NULL,
5714 pd2.ip_sum, icmpsum,
5715 pd->ip_sum, 0, AF_INET6);
5716 }
5717 m_copyback(m, off, sizeof(struct icmp6_hdr),
5718 (caddr_t)pd->hdr.icmp6);
5719 m_copyback(m, ipoff2, sizeof(h2_6),
5720 (caddr_t)&h2_6);
5721 m_copyback(m, off2, sizeof(struct icmp6_hdr),
5722 (caddr_t)&iih);
5723 }
5724
5725 return (PF_PASS);
5726 break;
5727 }
5728#endif /* INET6 */
5729 default: {
5730 key.af = pd2.af;
5731 key.proto = pd2.proto;
5732 if (direction == PF_IN) {
5733 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
5734 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
5735 key.ext.port = 0;
5736 key.gwy.port = 0;
5737 } else {
5738 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
5739 PF_ACPY(&key.ext.addr, pd2.src, key.af);
5740 key.lan.port = 0;
5741 key.ext.port = 0;
5742 }
5743
5744 STATE_LOOKUP();
5745
5746 if (STATE_TRANSLATE(*state)) {
5747 if (direction == PF_IN) {
5748 pf_change_icmp(pd2.src, NULL,
5749 daddr, &(*state)->lan.addr,
5750 0, NULL,
5751 pd2.ip_sum, icmpsum,
5752 pd->ip_sum, 0, pd2.af);
5753 } else {
5754 pf_change_icmp(pd2.dst, NULL,
5755 saddr, &(*state)->gwy.addr,
5756 0, NULL,
5757 pd2.ip_sum, icmpsum,
5758 pd->ip_sum, 0, pd2.af);
5759 }
5760 switch (pd2.af) {
5761#ifdef INET
5762 case AF_INET:
5763 m_copyback(m, off, ICMP_MINLEN,
5764 (caddr_t)pd->hdr.icmp);
5765 m_copyback(m, ipoff2, sizeof(h2),
5766 (caddr_t)&h2);
5767 break;
5768#endif /* INET */
5769#ifdef INET6
5770 case AF_INET6:
5771 m_copyback(m, off,
5772 sizeof(struct icmp6_hdr),
5773 (caddr_t)pd->hdr.icmp6);
5774 m_copyback(m, ipoff2, sizeof(h2_6),
5775 (caddr_t)&h2_6);
5776 break;
5777#endif /* INET6 */
5778 }
5779 }
5780
5781 return (PF_PASS);
5782 break;
5783 }
5784 }
5785 }
5786}
5787
5788int
5789pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif,
5790 struct pf_pdesc *pd)
5791{
5792 struct pf_state_peer *src, *dst;
5793 struct pf_state_cmp key;
5794
5795 key.af = pd->af;
5796 key.proto = pd->proto;
5797 if (direction == PF_IN) {
5798 PF_ACPY(&key.ext.addr, pd->src, key.af);
5799 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
5800 key.ext.port = 0;
5801 key.gwy.port = 0;
5802 } else {
5803 PF_ACPY(&key.lan.addr, pd->src, key.af);
5804 PF_ACPY(&key.ext.addr, pd->dst, key.af);
5805 key.lan.port = 0;
5806 key.ext.port = 0;
5807 }
5808
5809 STATE_LOOKUP();
5810
5811 if (direction == (*state)->direction) {
5812 src = &(*state)->src;
5813 dst = &(*state)->dst;
5814 } else {
5815 src = &(*state)->dst;
5816 dst = &(*state)->src;
5817 }
5818
5819 /* update states */
5820 if (src->state < PFOTHERS_SINGLE)
5821 src->state = PFOTHERS_SINGLE;
5822 if (dst->state == PFOTHERS_SINGLE)
5823 dst->state = PFOTHERS_MULTIPLE;
5824
5825 /* update expire time */
5826 (*state)->expire = time_second;
5827 if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE)
5828 (*state)->timeout = PFTM_OTHER_MULTIPLE;
5829 else
5830 (*state)->timeout = PFTM_OTHER_SINGLE;
5831
5832 /* translate source/destination address, if necessary */
5833 if (STATE_TRANSLATE(*state)) {
5834 if (direction == PF_OUT)
5835 switch (pd->af) {
5836#ifdef INET
5837 case AF_INET:
5838 pf_change_a(&pd->src->v4.s_addr,
5839 pd->ip_sum, (*state)->gwy.addr.v4.s_addr,
5840 0);
5841 break;
5842#endif /* INET */
5843#ifdef INET6
5844 case AF_INET6:
5845 PF_ACPY(pd->src, &(*state)->gwy.addr, pd->af);
5846 break;
5847#endif /* INET6 */
5848 }
5849 else
5850 switch (pd->af) {
5851#ifdef INET
5852 case AF_INET:
5853 pf_change_a(&pd->dst->v4.s_addr,
5854 pd->ip_sum, (*state)->lan.addr.v4.s_addr,
5855 0);
5856 break;
5857#endif /* INET */
5858#ifdef INET6
5859 case AF_INET6:
5860 PF_ACPY(pd->dst, &(*state)->lan.addr, pd->af);
5861 break;
5862#endif /* INET6 */
5863 }
5864 }
5865
5866 return (PF_PASS);
5867}
5868
5869/*
5870 * ipoff and off are measured from the start of the mbuf chain.
5871 * h must be at "ipoff" on the mbuf chain.
5872 */
5873void *
5874pf_pull_hdr(struct mbuf *m, int off, void *p, int len,
5875 u_short *actionp, u_short *reasonp, sa_family_t af)
5876{
5877 switch (af) {
5878#ifdef INET
5879 case AF_INET: {
5880 struct ip *h = mtod(m, struct ip *);
5881 u_int16_t fragoff = (ntohs(h->ip_off) & IP_OFFMASK) << 3;
5882
5883 if (fragoff) {
5884 if (fragoff >= len)
5885 ACTION_SET(actionp, PF_PASS);
5886 else {
5887 ACTION_SET(actionp, PF_DROP);
5888 REASON_SET(reasonp, PFRES_FRAG);
5889 }
5890 return (NULL);
5891 }
5892 if (m->m_pkthdr.len < off + len ||
5893 ntohs(h->ip_len) < off + len) {
5894 ACTION_SET(actionp, PF_DROP);
5895 REASON_SET(reasonp, PFRES_SHORT);
5896 return (NULL);
5897 }
5898 break;
5899 }
5900#endif /* INET */
5901#ifdef INET6
5902 case AF_INET6: {
5903 struct ip6_hdr *h = mtod(m, struct ip6_hdr *);
5904
5905 if (m->m_pkthdr.len < off + len ||
5906 (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) <
5907 (unsigned)(off + len)) {
5908 ACTION_SET(actionp, PF_DROP);
5909 REASON_SET(reasonp, PFRES_SHORT);
5910 return (NULL);
5911 }
5912 break;
5913 }
5914#endif /* INET6 */
5915 }
5916 m_copydata(m, off, len, p);
5917 return (p);
5918}
5919
5920int
5921pf_routable(struct pf_addr *addr, sa_family_t af, struct pfi_kif *kif)
5922{
5923 struct sockaddr_in *dst;
5924 int ret = 1;
5925 int check_mpath;
5926#ifndef __FreeBSD__
5927 extern int ipmultipath;
5928#endif
5929#ifdef INET6
5930#ifndef __FreeBSD__
5931 extern int ip6_multipath;
5932#endif
5933 struct sockaddr_in6 *dst6;
5934 struct route_in6 ro;
5935#else
5936 struct route ro;
5937#endif
5938 struct radix_node *rn;
5939 struct rtentry *rt;
5940 struct ifnet *ifp;
5941
5942 check_mpath = 0;
5943 bzero(&ro, sizeof(ro));
5944 switch (af) {
5945 case AF_INET:
5946 dst = satosin(&ro.ro_dst);
5947 dst->sin_family = AF_INET;
5948 dst->sin_len = sizeof(*dst);
5949 dst->sin_addr = addr->v4;
5950#ifndef __FreeBSD__ /* MULTIPATH_ROUTING */
5951 if (ipmultipath)
5952 check_mpath = 1;
5953#endif
5954 break;
5955#ifdef INET6
5956 case AF_INET6:
5957 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
5958 dst6->sin6_family = AF_INET6;
5959 dst6->sin6_len = sizeof(*dst6);
5960 dst6->sin6_addr = addr->v6;
5961#ifndef __FreeBSD__ /* MULTIPATH_ROUTING */
5962 if (ip6_multipath)
5963 check_mpath = 1;
5964#endif
5965 break;
5966#endif /* INET6 */
5967 default:
5968 return (0);
5969 }
5970
5971 /* Skip checks for ipsec interfaces */
5972 if (kif != NULL && kif->pfik_ifp->if_type == IFT_ENC)
5973 goto out;
5974
5975#ifdef __FreeBSD__
5976/* XXX MRT not always INET */ /* stick with table 0 though */
5977 if (af == AF_INET)
5978 in_rtalloc_ign((struct route *)&ro, 0, 0);
5979 else
5980 rtalloc_ign((struct route *)&ro, 0);
5981#else /* ! __FreeBSD__ */
5982 rtalloc_noclone((struct route *)&ro, NO_CLONING);
5983#endif
5984
5985 if (ro.ro_rt != NULL) {
5986 /* No interface given, this is a no-route check */
5987 if (kif == NULL)
5988 goto out;
5989
5990 if (kif->pfik_ifp == NULL) {
5991 ret = 0;
5992 goto out;
5993 }
5994
5995 /* Perform uRPF check if passed input interface */
5996 ret = 0;
5997 rn = (struct radix_node *)ro.ro_rt;
5998 do {
5999 rt = (struct rtentry *)rn;
6000#ifndef __FreeBSD__ /* CARPDEV */
6001 if (rt->rt_ifp->if_type == IFT_CARP)
6002 ifp = rt->rt_ifp->if_carpdev;
6003 else
6004#endif
6005 ifp = rt->rt_ifp;
6006
6007 if (kif->pfik_ifp == ifp)
6008 ret = 1;
6009#ifdef __FreeBSD__ /* MULTIPATH_ROUTING */
6010 rn = NULL;
6011#else
6012 rn = rn_mpath_next(rn);
6013#endif
6014 } while (check_mpath == 1 && rn != NULL && ret == 0);
6015 } else
6016 ret = 0;
6017out:
6018 if (ro.ro_rt != NULL)
6019 RTFREE(ro.ro_rt);
6020 return (ret);
6021}
6022
6023int
6024pf_rtlabel_match(struct pf_addr *addr, sa_family_t af, struct pf_addr_wrap *aw)
6025{
6026 struct sockaddr_in *dst;
6027#ifdef INET6
6028 struct sockaddr_in6 *dst6;
6029 struct route_in6 ro;
6030#else
6031 struct route ro;
6032#endif
6033 int ret = 0;
6034
6035 bzero(&ro, sizeof(ro));
6036 switch (af) {
6037 case AF_INET:
6038 dst = satosin(&ro.ro_dst);
6039 dst->sin_family = AF_INET;
6040 dst->sin_len = sizeof(*dst);
6041 dst->sin_addr = addr->v4;
6042 break;
6043#ifdef INET6
6044 case AF_INET6:
6045 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
6046 dst6->sin6_family = AF_INET6;
6047 dst6->sin6_len = sizeof(*dst6);
6048 dst6->sin6_addr = addr->v6;
6049 break;
6050#endif /* INET6 */
6051 default:
6052 return (0);
6053 }
6054
6055#ifdef __FreeBSD__
6056# ifdef RTF_PRCLONING
6057 rtalloc_ign((struct route *)&ro, (RTF_CLONING|RTF_PRCLONING));
6058# else /* !RTF_PRCLONING */
6059 if (af == AF_INET)
6060 in_rtalloc_ign((struct route *)&ro, 0, 0);
6061 else
6062 rtalloc_ign((struct route *)&ro, 0);
6063# endif
6064#else /* ! __FreeBSD__ */
6065 rtalloc_noclone((struct route *)&ro, NO_CLONING);
6066#endif
6067
6068 if (ro.ro_rt != NULL) {
6069#ifdef __FreeBSD__
6070 /* XXX_IMPORT: later */
6071#else
6072 if (ro.ro_rt->rt_labelid == aw->v.rtlabel)
6073 ret = 1;
6074#endif
6075 RTFREE(ro.ro_rt);
6076 }
6077
6078 return (ret);
6079}
6080
6081#ifdef INET
6082
6083void
6084pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
6085 struct pf_state *s, struct pf_pdesc *pd)
6086{
6087 struct mbuf *m0, *m1;
6088 struct route iproute;
6089 struct route *ro = NULL;
6090 struct sockaddr_in *dst;
6091 struct ip *ip;
6092 struct ifnet *ifp = NULL;
6093 struct pf_addr naddr;
6094 struct pf_src_node *sn = NULL;
6095 int error = 0;
6096#ifdef __FreeBSD__
6097 int sw_csum;
6098#endif
6099#ifdef IPSEC
6100 struct m_tag *mtag;
6101#endif /* IPSEC */
6102
6103 if (m == NULL || *m == NULL || r == NULL ||
6104 (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
6105 panic("pf_route: invalid parameters");
6106
6107 if (pd->pf_mtag->routed++ > 3) {
6108 m0 = *m;
6109 *m = NULL;
6110 goto bad;
6111 }
6112
6113 if (r->rt == PF_DUPTO) {
6114#ifdef __FreeBSD__
6115 if ((m0 = m_dup(*m, M_DONTWAIT)) == NULL)
6116#else
6117 if ((m0 = m_copym2(*m, 0, M_COPYALL, M_NOWAIT)) == NULL)
6118#endif
6119 return;
6120 } else {
6121 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
6122 return;
6123 m0 = *m;
6124 }
6125
6126 if (m0->m_len < sizeof(struct ip)) {
6127 DPFPRINTF(PF_DEBUG_URGENT,
6128 ("pf_route: m0->m_len < sizeof(struct ip)\n"));
6129 goto bad;
6130 }
6131
6132 ip = mtod(m0, struct ip *);
6133
6134 ro = &iproute;
6135 bzero((caddr_t)ro, sizeof(*ro));
6136 dst = satosin(&ro->ro_dst);
6137 dst->sin_family = AF_INET;
6138 dst->sin_len = sizeof(*dst);
6139 dst->sin_addr = ip->ip_dst;
6140
6141 if (r->rt == PF_FASTROUTE) {
6142 in_rtalloc(ro, 0);
6143 if (ro->ro_rt == 0) {
| 45#endif
46
47#ifdef __FreeBSD__
48#include "opt_bpf.h"
49#include "opt_pf.h"
50
51#ifdef DEV_BPF
52#define NBPFILTER DEV_BPF
53#else
54#define NBPFILTER 0
55#endif
56
57#ifdef DEV_PFLOG
58#define NPFLOG DEV_PFLOG
59#else
60#define NPFLOG 0
61#endif
62
63#ifdef DEV_PFSYNC
64#define NPFSYNC DEV_PFSYNC
65#else
66#define NPFSYNC 0
67#endif
68
69#else
70#include "bpfilter.h"
71#include "pflog.h"
72#include "pfsync.h"
73#endif
74
75#include <sys/param.h>
76#include <sys/systm.h>
77#include <sys/mbuf.h>
78#include <sys/filio.h>
79#include <sys/socket.h>
80#include <sys/socketvar.h>
81#include <sys/kernel.h>
82#include <sys/time.h>
83#ifdef __FreeBSD__
84#include <sys/sysctl.h>
85#include <sys/endian.h>
86#else
87#include <sys/pool.h>
88#endif
89#include <sys/proc.h>
90#ifdef __FreeBSD__
91#include <sys/kthread.h>
92#include <sys/lock.h>
93#include <sys/sx.h>
94#else
95#include <sys/rwlock.h>
96#endif
97
98#include <net/if.h>
99#include <net/if_types.h>
100#include <net/bpf.h>
101#include <net/route.h>
102#ifndef __FreeBSD__
103#include <net/radix_mpath.h>
104#endif
105
106#include <netinet/in.h>
107#include <netinet/in_var.h>
108#include <netinet/in_systm.h>
109#include <netinet/ip.h>
110#include <netinet/ip_var.h>
111#include <netinet/tcp.h>
112#include <netinet/tcp_seq.h>
113#include <netinet/udp.h>
114#include <netinet/ip_icmp.h>
115#include <netinet/in_pcb.h>
116#include <netinet/tcp_timer.h>
117#include <netinet/tcp_var.h>
118#include <netinet/udp_var.h>
119#include <netinet/icmp_var.h>
120#include <netinet/if_ether.h>
121
122#ifndef __FreeBSD__
123#include <dev/rndvar.h>
124#endif
125#include <net/pfvar.h>
126#include <net/if_pflog.h>
127
128#if NPFSYNC > 0
129#include <net/if_pfsync.h>
130#endif /* NPFSYNC > 0 */
131
132#ifdef INET6
133#include <netinet/ip6.h>
134#include <netinet/in_pcb.h>
135#include <netinet/icmp6.h>
136#include <netinet6/nd6.h>
137#ifdef __FreeBSD__
138#include <netinet6/ip6_var.h>
139#include <netinet6/in6_pcb.h>
140#endif
141#endif /* INET6 */
142
143#ifdef __FreeBSD__
144#include <machine/in_cksum.h>
145#include <sys/limits.h>
146#include <sys/ucred.h>
147#include <security/mac/mac_framework.h>
148
149extern int ip_optcopy(struct ip *, struct ip *);
150extern int debug_pfugidhack;
151#endif
152
153#define DPFPRINTF(n, x) if (pf_status.debug >= (n)) printf x
154
155/*
156 * Global variables
157 */
158
159struct pf_altqqueue pf_altqs[2];
160struct pf_palist pf_pabuf;
161struct pf_altqqueue *pf_altqs_active;
162struct pf_altqqueue *pf_altqs_inactive;
163struct pf_status pf_status;
164
165u_int32_t ticket_altqs_active;
166u_int32_t ticket_altqs_inactive;
167int altqs_inactive_open;
168u_int32_t ticket_pabuf;
169
170struct pf_anchor_stackframe {
171 struct pf_ruleset *rs;
172 struct pf_rule *r;
173 struct pf_anchor_node *parent;
174 struct pf_anchor *child;
175} pf_anchor_stack[64];
176
177#ifdef __FreeBSD__
178uma_zone_t pf_src_tree_pl, pf_rule_pl;
179uma_zone_t pf_state_pl, pf_altq_pl, pf_pooladdr_pl;
180#else
181struct pool pf_src_tree_pl, pf_rule_pl;
182struct pool pf_state_pl, pf_altq_pl, pf_pooladdr_pl;
183#endif
184
185void pf_print_host(struct pf_addr *, u_int16_t, u_int8_t);
186
187void pf_init_threshold(struct pf_threshold *, u_int32_t,
188 u_int32_t);
189void pf_add_threshold(struct pf_threshold *);
190int pf_check_threshold(struct pf_threshold *);
191
192void pf_change_ap(struct pf_addr *, u_int16_t *,
193 u_int16_t *, u_int16_t *, struct pf_addr *,
194 u_int16_t, u_int8_t, sa_family_t);
195int pf_modulate_sack(struct mbuf *, int, struct pf_pdesc *,
196 struct tcphdr *, struct pf_state_peer *);
197#ifdef INET6
198void pf_change_a6(struct pf_addr *, u_int16_t *,
199 struct pf_addr *, u_int8_t);
200#endif /* INET6 */
201void pf_change_icmp(struct pf_addr *, u_int16_t *,
202 struct pf_addr *, struct pf_addr *, u_int16_t,
203 u_int16_t *, u_int16_t *, u_int16_t *,
204 u_int16_t *, u_int8_t, sa_family_t);
205#ifdef __FreeBSD__
206void pf_send_tcp(struct mbuf *,
207 const struct pf_rule *, sa_family_t,
208#else
209void pf_send_tcp(const struct pf_rule *, sa_family_t,
210#endif
211 const struct pf_addr *, const struct pf_addr *,
212 u_int16_t, u_int16_t, u_int32_t, u_int32_t,
213 u_int8_t, u_int16_t, u_int16_t, u_int8_t, int,
214 u_int16_t, struct ether_header *, struct ifnet *);
215void pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t,
216 sa_family_t, struct pf_rule *);
217struct pf_rule *pf_match_translation(struct pf_pdesc *, struct mbuf *,
218 int, int, struct pfi_kif *,
219 struct pf_addr *, u_int16_t, struct pf_addr *,
220 u_int16_t, int);
221struct pf_rule *pf_get_translation(struct pf_pdesc *, struct mbuf *,
222 int, int, struct pfi_kif *, struct pf_src_node **,
223 struct pf_addr *, u_int16_t,
224 struct pf_addr *, u_int16_t,
225 struct pf_addr *, u_int16_t *);
226int pf_test_tcp(struct pf_rule **, struct pf_state **,
227 int, struct pfi_kif *, struct mbuf *, int,
228 void *, struct pf_pdesc *, struct pf_rule **,
229#ifdef __FreeBSD__
230 struct pf_ruleset **, struct ifqueue *,
231 struct inpcb *);
232#else
233 struct pf_ruleset **, struct ifqueue *);
234#endif
235int pf_test_udp(struct pf_rule **, struct pf_state **,
236 int, struct pfi_kif *, struct mbuf *, int,
237 void *, struct pf_pdesc *, struct pf_rule **,
238#ifdef __FreeBSD__
239 struct pf_ruleset **, struct ifqueue *,
240 struct inpcb *);
241#else
242 struct pf_ruleset **, struct ifqueue *);
243#endif
244int pf_test_icmp(struct pf_rule **, struct pf_state **,
245 int, struct pfi_kif *, struct mbuf *, int,
246 void *, struct pf_pdesc *, struct pf_rule **,
247 struct pf_ruleset **, struct ifqueue *);
248int pf_test_other(struct pf_rule **, struct pf_state **,
249 int, struct pfi_kif *, struct mbuf *, int, void *,
250 struct pf_pdesc *, struct pf_rule **,
251 struct pf_ruleset **, struct ifqueue *);
252int pf_test_fragment(struct pf_rule **, int,
253 struct pfi_kif *, struct mbuf *, void *,
254 struct pf_pdesc *, struct pf_rule **,
255 struct pf_ruleset **);
256int pf_test_state_tcp(struct pf_state **, int,
257 struct pfi_kif *, struct mbuf *, int,
258 void *, struct pf_pdesc *, u_short *);
259int pf_test_state_udp(struct pf_state **, int,
260 struct pfi_kif *, struct mbuf *, int,
261 void *, struct pf_pdesc *);
262int pf_test_state_icmp(struct pf_state **, int,
263 struct pfi_kif *, struct mbuf *, int,
264 void *, struct pf_pdesc *, u_short *);
265int pf_test_state_other(struct pf_state **, int,
266 struct pfi_kif *, struct pf_pdesc *);
267int pf_match_tag(struct mbuf *, struct pf_rule *,
268 struct pf_mtag *, int *);
269int pf_step_out_of_anchor(int *, struct pf_ruleset **,
270 int, struct pf_rule **, struct pf_rule **,
271 int *);
272void pf_hash(struct pf_addr *, struct pf_addr *,
273 struct pf_poolhashkey *, sa_family_t);
274int pf_map_addr(u_int8_t, struct pf_rule *,
275 struct pf_addr *, struct pf_addr *,
276 struct pf_addr *, struct pf_src_node **);
277int pf_get_sport(sa_family_t, u_int8_t, struct pf_rule *,
278 struct pf_addr *, struct pf_addr *, u_int16_t,
279 struct pf_addr *, u_int16_t*, u_int16_t, u_int16_t,
280 struct pf_src_node **);
281void pf_route(struct mbuf **, struct pf_rule *, int,
282 struct ifnet *, struct pf_state *,
283 struct pf_pdesc *);
284void pf_route6(struct mbuf **, struct pf_rule *, int,
285 struct ifnet *, struct pf_state *,
286 struct pf_pdesc *);
287#ifdef __FreeBSD__
288/* XXX: import */
289#else
290int pf_socket_lookup(int, struct pf_pdesc *);
291#endif
292u_int8_t pf_get_wscale(struct mbuf *, int, u_int16_t,
293 sa_family_t);
294u_int16_t pf_get_mss(struct mbuf *, int, u_int16_t,
295 sa_family_t);
296u_int16_t pf_calc_mss(struct pf_addr *, sa_family_t,
297 u_int16_t);
298void pf_set_rt_ifp(struct pf_state *,
299 struct pf_addr *);
300int pf_check_proto_cksum(struct mbuf *, int, int,
301 u_int8_t, sa_family_t);
302int pf_addr_wrap_neq(struct pf_addr_wrap *,
303 struct pf_addr_wrap *);
304struct pf_state *pf_find_state_recurse(struct pfi_kif *,
305 struct pf_state_cmp *, u_int8_t);
306int pf_src_connlimit(struct pf_state **);
307int pf_check_congestion(struct ifqueue *);
308
309#ifdef __FreeBSD__
310int in4_cksum(struct mbuf *m, u_int8_t nxt, int off, int len);
311
312extern int pf_end_threads;
313
314struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX];
315#else
316extern struct pool pfr_ktable_pl;
317extern struct pool pfr_kentry_pl;
318
319struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] = {
320 { &pf_state_pl, PFSTATE_HIWAT },
321 { &pf_src_tree_pl, PFSNODE_HIWAT },
322 { &pf_frent_pl, PFFRAG_FRENT_HIWAT },
323 { &pfr_ktable_pl, PFR_KTABLE_HIWAT },
324 { &pfr_kentry_pl, PFR_KENTRY_HIWAT }
325};
326#endif
327
328#define STATE_LOOKUP() \
329 do { \
330 if (direction == PF_IN) \
331 *state = pf_find_state_recurse( \
332 kif, &key, PF_EXT_GWY); \
333 else \
334 *state = pf_find_state_recurse( \
335 kif, &key, PF_LAN_EXT); \
336 if (*state == NULL || (*state)->timeout == PFTM_PURGE) \
337 return (PF_DROP); \
338 if (direction == PF_OUT && \
339 (((*state)->rule.ptr->rt == PF_ROUTETO && \
340 (*state)->rule.ptr->direction == PF_OUT) || \
341 ((*state)->rule.ptr->rt == PF_REPLYTO && \
342 (*state)->rule.ptr->direction == PF_IN)) && \
343 (*state)->rt_kif != NULL && \
344 (*state)->rt_kif != kif) \
345 return (PF_PASS); \
346 } while (0)
347
348#define STATE_TRANSLATE(s) \
349 (s)->lan.addr.addr32[0] != (s)->gwy.addr.addr32[0] || \
350 ((s)->af == AF_INET6 && \
351 ((s)->lan.addr.addr32[1] != (s)->gwy.addr.addr32[1] || \
352 (s)->lan.addr.addr32[2] != (s)->gwy.addr.addr32[2] || \
353 (s)->lan.addr.addr32[3] != (s)->gwy.addr.addr32[3])) || \
354 (s)->lan.port != (s)->gwy.port
355
356#define BOUND_IFACE(r, k) \
357 ((r)->rule_flag & PFRULE_IFBOUND) ? (k) : pfi_all
358
359#define STATE_INC_COUNTERS(s) \
360 do { \
361 s->rule.ptr->states++; \
362 if (s->anchor.ptr != NULL) \
363 s->anchor.ptr->states++; \
364 if (s->nat_rule.ptr != NULL) \
365 s->nat_rule.ptr->states++; \
366 } while (0)
367
368#define STATE_DEC_COUNTERS(s) \
369 do { \
370 if (s->nat_rule.ptr != NULL) \
371 s->nat_rule.ptr->states--; \
372 if (s->anchor.ptr != NULL) \
373 s->anchor.ptr->states--; \
374 s->rule.ptr->states--; \
375 } while (0)
376
377struct pf_src_tree tree_src_tracking;
378
379struct pf_state_tree_id tree_id;
380struct pf_state_queue state_list;
381
382#ifdef __FreeBSD__
383static int pf_src_compare(struct pf_src_node *, struct pf_src_node *);
384static int pf_state_compare_lan_ext(struct pf_state *, struct pf_state *);
385static int pf_state_compare_ext_gwy(struct pf_state *, struct pf_state *);
386static int pf_state_compare_id(struct pf_state *, struct pf_state *);
387#endif
388
389RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare);
390RB_GENERATE(pf_state_tree_lan_ext, pf_state,
391 u.s.entry_lan_ext, pf_state_compare_lan_ext);
392RB_GENERATE(pf_state_tree_ext_gwy, pf_state,
393 u.s.entry_ext_gwy, pf_state_compare_ext_gwy);
394RB_GENERATE(pf_state_tree_id, pf_state,
395 u.s.entry_id, pf_state_compare_id);
396
397#ifdef __FreeBSD__
398static int
399#else
400static __inline int
401#endif
402pf_src_compare(struct pf_src_node *a, struct pf_src_node *b)
403{
404 int diff;
405
406 if (a->rule.ptr > b->rule.ptr)
407 return (1);
408 if (a->rule.ptr < b->rule.ptr)
409 return (-1);
410 if ((diff = a->af - b->af) != 0)
411 return (diff);
412 switch (a->af) {
413#ifdef INET
414 case AF_INET:
415 if (a->addr.addr32[0] > b->addr.addr32[0])
416 return (1);
417 if (a->addr.addr32[0] < b->addr.addr32[0])
418 return (-1);
419 break;
420#endif /* INET */
421#ifdef INET6
422 case AF_INET6:
423 if (a->addr.addr32[3] > b->addr.addr32[3])
424 return (1);
425 if (a->addr.addr32[3] < b->addr.addr32[3])
426 return (-1);
427 if (a->addr.addr32[2] > b->addr.addr32[2])
428 return (1);
429 if (a->addr.addr32[2] < b->addr.addr32[2])
430 return (-1);
431 if (a->addr.addr32[1] > b->addr.addr32[1])
432 return (1);
433 if (a->addr.addr32[1] < b->addr.addr32[1])
434 return (-1);
435 if (a->addr.addr32[0] > b->addr.addr32[0])
436 return (1);
437 if (a->addr.addr32[0] < b->addr.addr32[0])
438 return (-1);
439 break;
440#endif /* INET6 */
441 }
442 return (0);
443}
444
445#ifdef __FreeBSD__
446static int
447#else
448static __inline int
449#endif
450pf_state_compare_lan_ext(struct pf_state *a, struct pf_state *b)
451{
452 int diff;
453
454 if ((diff = a->proto - b->proto) != 0)
455 return (diff);
456 if ((diff = a->af - b->af) != 0)
457 return (diff);
458 switch (a->af) {
459#ifdef INET
460 case AF_INET:
461 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
462 return (1);
463 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
464 return (-1);
465 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
466 return (1);
467 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
468 return (-1);
469 break;
470#endif /* INET */
471#ifdef INET6
472 case AF_INET6:
473 if (a->lan.addr.addr32[3] > b->lan.addr.addr32[3])
474 return (1);
475 if (a->lan.addr.addr32[3] < b->lan.addr.addr32[3])
476 return (-1);
477 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
478 return (1);
479 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
480 return (-1);
481 if (a->lan.addr.addr32[2] > b->lan.addr.addr32[2])
482 return (1);
483 if (a->lan.addr.addr32[2] < b->lan.addr.addr32[2])
484 return (-1);
485 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
486 return (1);
487 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
488 return (-1);
489 if (a->lan.addr.addr32[1] > b->lan.addr.addr32[1])
490 return (1);
491 if (a->lan.addr.addr32[1] < b->lan.addr.addr32[1])
492 return (-1);
493 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
494 return (1);
495 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
496 return (-1);
497 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
498 return (1);
499 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
500 return (-1);
501 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
502 return (1);
503 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
504 return (-1);
505 break;
506#endif /* INET6 */
507 }
508
509 if ((diff = a->lan.port - b->lan.port) != 0)
510 return (diff);
511 if ((diff = a->ext.port - b->ext.port) != 0)
512 return (diff);
513
514 return (0);
515}
516
517#ifdef __FreeBSD__
518static int
519#else
520static __inline int
521#endif
522pf_state_compare_ext_gwy(struct pf_state *a, struct pf_state *b)
523{
524 int diff;
525
526 if ((diff = a->proto - b->proto) != 0)
527 return (diff);
528 if ((diff = a->af - b->af) != 0)
529 return (diff);
530 switch (a->af) {
531#ifdef INET
532 case AF_INET:
533 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
534 return (1);
535 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
536 return (-1);
537 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
538 return (1);
539 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
540 return (-1);
541 break;
542#endif /* INET */
543#ifdef INET6
544 case AF_INET6:
545 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
546 return (1);
547 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
548 return (-1);
549 if (a->gwy.addr.addr32[3] > b->gwy.addr.addr32[3])
550 return (1);
551 if (a->gwy.addr.addr32[3] < b->gwy.addr.addr32[3])
552 return (-1);
553 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
554 return (1);
555 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
556 return (-1);
557 if (a->gwy.addr.addr32[2] > b->gwy.addr.addr32[2])
558 return (1);
559 if (a->gwy.addr.addr32[2] < b->gwy.addr.addr32[2])
560 return (-1);
561 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
562 return (1);
563 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
564 return (-1);
565 if (a->gwy.addr.addr32[1] > b->gwy.addr.addr32[1])
566 return (1);
567 if (a->gwy.addr.addr32[1] < b->gwy.addr.addr32[1])
568 return (-1);
569 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
570 return (1);
571 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
572 return (-1);
573 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
574 return (1);
575 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
576 return (-1);
577 break;
578#endif /* INET6 */
579 }
580
581 if ((diff = a->ext.port - b->ext.port) != 0)
582 return (diff);
583 if ((diff = a->gwy.port - b->gwy.port) != 0)
584 return (diff);
585
586 return (0);
587}
588
589#ifdef __FreeBSD__
590static int
591#else
592static __inline int
593#endif
594pf_state_compare_id(struct pf_state *a, struct pf_state *b)
595{
596 if (a->id > b->id)
597 return (1);
598 if (a->id < b->id)
599 return (-1);
600 if (a->creatorid > b->creatorid)
601 return (1);
602 if (a->creatorid < b->creatorid)
603 return (-1);
604
605 return (0);
606}
607
608#ifdef INET6
609void
610pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af)
611{
612 switch (af) {
613#ifdef INET
614 case AF_INET:
615 dst->addr32[0] = src->addr32[0];
616 break;
617#endif /* INET */
618 case AF_INET6:
619 dst->addr32[0] = src->addr32[0];
620 dst->addr32[1] = src->addr32[1];
621 dst->addr32[2] = src->addr32[2];
622 dst->addr32[3] = src->addr32[3];
623 break;
624 }
625}
626#endif /* INET6 */
627
628struct pf_state *
629pf_find_state_byid(struct pf_state_cmp *key)
630{
631 pf_status.fcounters[FCNT_STATE_SEARCH]++;
632 return (RB_FIND(pf_state_tree_id, &tree_id, (struct pf_state *)key));
633}
634
635struct pf_state *
636pf_find_state_recurse(struct pfi_kif *kif, struct pf_state_cmp *key, u_int8_t tree)
637{
638 struct pf_state *s;
639
640 pf_status.fcounters[FCNT_STATE_SEARCH]++;
641
642 switch (tree) {
643 case PF_LAN_EXT:
644 if ((s = RB_FIND(pf_state_tree_lan_ext, &kif->pfik_lan_ext,
645 (struct pf_state *)key)) != NULL)
646 return (s);
647 if ((s = RB_FIND(pf_state_tree_lan_ext, &pfi_all->pfik_lan_ext,
648 (struct pf_state *)key)) != NULL)
649 return (s);
650 return (NULL);
651 case PF_EXT_GWY:
652 if ((s = RB_FIND(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy,
653 (struct pf_state *)key)) != NULL)
654 return (s);
655 if ((s = RB_FIND(pf_state_tree_ext_gwy, &pfi_all->pfik_ext_gwy,
656 (struct pf_state *)key)) != NULL)
657 return (s);
658 return (NULL);
659 default:
660 panic("pf_find_state_recurse");
661 }
662}
663
664struct pf_state *
665pf_find_state_all(struct pf_state_cmp *key, u_int8_t tree, int *more)
666{
667 struct pf_state *s, *ss = NULL;
668 struct pfi_kif *kif;
669
670 pf_status.fcounters[FCNT_STATE_SEARCH]++;
671
672 switch (tree) {
673 case PF_LAN_EXT:
674 TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states) {
675 s = RB_FIND(pf_state_tree_lan_ext,
676 &kif->pfik_lan_ext, (struct pf_state *)key);
677 if (s == NULL)
678 continue;
679 if (more == NULL)
680 return (s);
681 ss = s;
682 (*more)++;
683 }
684 return (ss);
685 case PF_EXT_GWY:
686 TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states) {
687 s = RB_FIND(pf_state_tree_ext_gwy,
688 &kif->pfik_ext_gwy, (struct pf_state *)key);
689 if (s == NULL)
690 continue;
691 if (more == NULL)
692 return (s);
693 ss = s;
694 (*more)++;
695 }
696 return (ss);
697 default:
698 panic("pf_find_state_all");
699 }
700}
701
702void
703pf_init_threshold(struct pf_threshold *threshold,
704 u_int32_t limit, u_int32_t seconds)
705{
706 threshold->limit = limit * PF_THRESHOLD_MULT;
707 threshold->seconds = seconds;
708 threshold->count = 0;
709 threshold->last = time_second;
710}
711
712void
713pf_add_threshold(struct pf_threshold *threshold)
714{
715 u_int32_t t = time_second, diff = t - threshold->last;
716
717 if (diff >= threshold->seconds)
718 threshold->count = 0;
719 else
720 threshold->count -= threshold->count * diff /
721 threshold->seconds;
722 threshold->count += PF_THRESHOLD_MULT;
723 threshold->last = t;
724}
725
726int
727pf_check_threshold(struct pf_threshold *threshold)
728{
729 return (threshold->count > threshold->limit);
730}
731
732int
733pf_src_connlimit(struct pf_state **state)
734{
735 struct pf_state *s;
736 int bad = 0;
737
738 (*state)->src_node->conn++;
739 (*state)->src.tcp_est = 1;
740 pf_add_threshold(&(*state)->src_node->conn_rate);
741
742 if ((*state)->rule.ptr->max_src_conn &&
743 (*state)->rule.ptr->max_src_conn <
744 (*state)->src_node->conn) {
745 pf_status.lcounters[LCNT_SRCCONN]++;
746 bad++;
747 }
748
749 if ((*state)->rule.ptr->max_src_conn_rate.limit &&
750 pf_check_threshold(&(*state)->src_node->conn_rate)) {
751 pf_status.lcounters[LCNT_SRCCONNRATE]++;
752 bad++;
753 }
754
755 if (!bad)
756 return (0);
757
758 if ((*state)->rule.ptr->overload_tbl) {
759 struct pfr_addr p;
760 u_int32_t killed = 0;
761
762 pf_status.lcounters[LCNT_OVERLOAD_TABLE]++;
763 if (pf_status.debug >= PF_DEBUG_MISC) {
764 printf("pf_src_connlimit: blocking address ");
765 pf_print_host(&(*state)->src_node->addr, 0,
766 (*state)->af);
767 }
768
769 bzero(&p, sizeof(p));
770 p.pfra_af = (*state)->af;
771 switch ((*state)->af) {
772#ifdef INET
773 case AF_INET:
774 p.pfra_net = 32;
775 p.pfra_ip4addr = (*state)->src_node->addr.v4;
776 break;
777#endif /* INET */
778#ifdef INET6
779 case AF_INET6:
780 p.pfra_net = 128;
781 p.pfra_ip6addr = (*state)->src_node->addr.v6;
782 break;
783#endif /* INET6 */
784 }
785
786 pfr_insert_kentry((*state)->rule.ptr->overload_tbl,
787 &p, time_second);
788
789 /* kill existing states if that's required. */
790 if ((*state)->rule.ptr->flush) {
791 pf_status.lcounters[LCNT_OVERLOAD_FLUSH]++;
792
793 RB_FOREACH(s, pf_state_tree_id, &tree_id) {
794 /*
795 * Kill states from this source. (Only those
796 * from the same rule if PF_FLUSH_GLOBAL is not
797 * set)
798 */
799 if (s->af == (*state)->af &&
800 (((*state)->direction == PF_OUT &&
801 PF_AEQ(&(*state)->src_node->addr,
802 &s->lan.addr, s->af)) ||
803 ((*state)->direction == PF_IN &&
804 PF_AEQ(&(*state)->src_node->addr,
805 &s->ext.addr, s->af))) &&
806 ((*state)->rule.ptr->flush &
807 PF_FLUSH_GLOBAL ||
808 (*state)->rule.ptr == s->rule.ptr)) {
809 s->timeout = PFTM_PURGE;
810 s->src.state = s->dst.state =
811 TCPS_CLOSED;
812 killed++;
813 }
814 }
815 if (pf_status.debug >= PF_DEBUG_MISC)
816 printf(", %u states killed", killed);
817 }
818 if (pf_status.debug >= PF_DEBUG_MISC)
819 printf("\n");
820 }
821
822 /* kill this state */
823 (*state)->timeout = PFTM_PURGE;
824 (*state)->src.state = (*state)->dst.state = TCPS_CLOSED;
825 return (1);
826}
827
828int
829pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule,
830 struct pf_addr *src, sa_family_t af)
831{
832 struct pf_src_node k;
833
834 if (*sn == NULL) {
835 k.af = af;
836 PF_ACPY(&k.addr, src, af);
837 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
838 rule->rpool.opts & PF_POOL_STICKYADDR)
839 k.rule.ptr = rule;
840 else
841 k.rule.ptr = NULL;
842 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
843 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
844 }
845 if (*sn == NULL) {
846 if (!rule->max_src_nodes ||
847 rule->src_nodes < rule->max_src_nodes)
848 (*sn) = pool_get(&pf_src_tree_pl, PR_NOWAIT);
849 else
850 pf_status.lcounters[LCNT_SRCNODES]++;
851 if ((*sn) == NULL)
852 return (-1);
853 bzero(*sn, sizeof(struct pf_src_node));
854
855 pf_init_threshold(&(*sn)->conn_rate,
856 rule->max_src_conn_rate.limit,
857 rule->max_src_conn_rate.seconds);
858
859 (*sn)->af = af;
860 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
861 rule->rpool.opts & PF_POOL_STICKYADDR)
862 (*sn)->rule.ptr = rule;
863 else
864 (*sn)->rule.ptr = NULL;
865 PF_ACPY(&(*sn)->addr, src, af);
866 if (RB_INSERT(pf_src_tree,
867 &tree_src_tracking, *sn) != NULL) {
868 if (pf_status.debug >= PF_DEBUG_MISC) {
869 printf("pf: src_tree insert failed: ");
870 pf_print_host(&(*sn)->addr, 0, af);
871 printf("\n");
872 }
873 pool_put(&pf_src_tree_pl, *sn);
874 return (-1);
875 }
876 (*sn)->creation = time_second;
877 (*sn)->ruletype = rule->action;
878 if ((*sn)->rule.ptr != NULL)
879 (*sn)->rule.ptr->src_nodes++;
880 pf_status.scounters[SCNT_SRC_NODE_INSERT]++;
881 pf_status.src_nodes++;
882 } else {
883 if (rule->max_src_states &&
884 (*sn)->states >= rule->max_src_states) {
885 pf_status.lcounters[LCNT_SRCSTATES]++;
886 return (-1);
887 }
888 }
889 return (0);
890}
891
892int
893pf_insert_state(struct pfi_kif *kif, struct pf_state *state)
894{
895 /* Thou MUST NOT insert multiple duplicate keys */
896 state->u.s.kif = kif;
897 if (RB_INSERT(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state)) {
898 if (pf_status.debug >= PF_DEBUG_MISC) {
899 printf("pf: state insert failed: tree_lan_ext");
900 printf(" lan: ");
901 pf_print_host(&state->lan.addr, state->lan.port,
902 state->af);
903 printf(" gwy: ");
904 pf_print_host(&state->gwy.addr, state->gwy.port,
905 state->af);
906 printf(" ext: ");
907 pf_print_host(&state->ext.addr, state->ext.port,
908 state->af);
909 if (state->sync_flags & PFSTATE_FROMSYNC)
910 printf(" (from sync)");
911 printf("\n");
912 }
913 return (-1);
914 }
915
916 if (RB_INSERT(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, state)) {
917 if (pf_status.debug >= PF_DEBUG_MISC) {
918 printf("pf: state insert failed: tree_ext_gwy");
919 printf(" lan: ");
920 pf_print_host(&state->lan.addr, state->lan.port,
921 state->af);
922 printf(" gwy: ");
923 pf_print_host(&state->gwy.addr, state->gwy.port,
924 state->af);
925 printf(" ext: ");
926 pf_print_host(&state->ext.addr, state->ext.port,
927 state->af);
928 if (state->sync_flags & PFSTATE_FROMSYNC)
929 printf(" (from sync)");
930 printf("\n");
931 }
932 RB_REMOVE(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state);
933 return (-1);
934 }
935
936 if (state->id == 0 && state->creatorid == 0) {
937 state->id = htobe64(pf_status.stateid++);
938 state->creatorid = pf_status.hostid;
939 }
940 if (RB_INSERT(pf_state_tree_id, &tree_id, state) != NULL) {
941 if (pf_status.debug >= PF_DEBUG_MISC) {
942#ifdef __FreeBSD__
943 printf("pf: state insert failed: "
944 "id: %016llx creatorid: %08x",
945 (long long)be64toh(state->id),
946 ntohl(state->creatorid));
947#else
948 printf("pf: state insert failed: "
949 "id: %016llx creatorid: %08x",
950 betoh64(state->id), ntohl(state->creatorid));
951#endif
952 if (state->sync_flags & PFSTATE_FROMSYNC)
953 printf(" (from sync)");
954 printf("\n");
955 }
956 RB_REMOVE(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state);
957 RB_REMOVE(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, state);
958 return (-1);
959 }
960 TAILQ_INSERT_TAIL(&state_list, state, u.s.entry_list);
961 pf_status.fcounters[FCNT_STATE_INSERT]++;
962 pf_status.states++;
963 pfi_kif_ref(kif, PFI_KIF_REF_STATE);
964#if NPFSYNC
965 pfsync_insert_state(state);
966#endif
967 return (0);
968}
969
970void
971pf_purge_thread(void *v)
972{
973 int nloops = 0, s;
974
975 for (;;) {
976 tsleep(pf_purge_thread, PWAIT, "pftm", 1 * hz);
977
978#ifdef __FreeBSD__
979 sx_slock(&pf_consistency_lock);
980 PF_LOCK();
981
982 if (pf_end_threads) {
983 pf_purge_expired_states(pf_status.states);
984 pf_purge_expired_fragments();
985 pf_purge_expired_src_nodes(0);
986 pf_end_threads++;
987
988 sx_sunlock(&pf_consistency_lock);
989 PF_UNLOCK();
990 wakeup(pf_purge_thread);
991 kproc_exit(0);
992 }
993#endif
994 s = splsoftnet();
995
996 /* process a fraction of the state table every second */
997 pf_purge_expired_states(1 + (pf_status.states
998 / pf_default_rule.timeout[PFTM_INTERVAL]));
999
1000 /* purge other expired types every PFTM_INTERVAL seconds */
1001 if (++nloops >= pf_default_rule.timeout[PFTM_INTERVAL]) {
1002 pf_purge_expired_fragments();
1003 pf_purge_expired_src_nodes(0);
1004 nloops = 0;
1005 }
1006
1007 splx(s);
1008#ifdef __FreeBSD__
1009 PF_UNLOCK();
1010 sx_sunlock(&pf_consistency_lock);
1011#endif
1012 }
1013}
1014
1015u_int32_t
1016pf_state_expires(const struct pf_state *state)
1017{
1018 u_int32_t timeout;
1019 u_int32_t start;
1020 u_int32_t end;
1021 u_int32_t states;
1022
1023 /* handle all PFTM_* > PFTM_MAX here */
1024 if (state->timeout == PFTM_PURGE)
1025 return (time_second);
1026 if (state->timeout == PFTM_UNTIL_PACKET)
1027 return (0);
1028#ifdef __FreeBSD__
1029 KASSERT(state->timeout != PFTM_UNLINKED,
1030 ("pf_state_expires: timeout == PFTM_UNLINKED"));
1031 KASSERT((state->timeout < PFTM_MAX),
1032 ("pf_state_expires: timeout > PFTM_MAX"));
1033#else
1034 KASSERT(state->timeout != PFTM_UNLINKED);
1035 KASSERT(state->timeout < PFTM_MAX);
1036#endif
1037 timeout = state->rule.ptr->timeout[state->timeout];
1038 if (!timeout)
1039 timeout = pf_default_rule.timeout[state->timeout];
1040 start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START];
1041 if (start) {
1042 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END];
1043 states = state->rule.ptr->states;
1044 } else {
1045 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START];
1046 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END];
1047 states = pf_status.states;
1048 }
1049 if (end && states > start && start < end) {
1050 if (states < end)
1051 return (state->expire + timeout * (end - states) /
1052 (end - start));
1053 else
1054 return (time_second);
1055 }
1056 return (state->expire + timeout);
1057}
1058
1059void
1060pf_purge_expired_src_nodes(int waslocked)
1061{
1062 struct pf_src_node *cur, *next;
1063 int locked = waslocked;
1064
1065 for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) {
1066 next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur);
1067
1068 if (cur->states <= 0 && cur->expire <= time_second) {
1069 if (! locked) {
1070#ifdef __FreeBSD__
1071 if (!sx_try_upgrade(&pf_consistency_lock)) {
1072 PF_UNLOCK();
1073 sx_sunlock(&pf_consistency_lock);
1074 sx_xlock(&pf_consistency_lock);
1075 PF_LOCK();
1076 }
1077#else
1078 rw_enter_write(&pf_consistency_lock);
1079#endif
1080 next = RB_NEXT(pf_src_tree,
1081 &tree_src_tracking, cur);
1082 locked = 1;
1083 }
1084 if (cur->rule.ptr != NULL) {
1085 cur->rule.ptr->src_nodes--;
1086 if (cur->rule.ptr->states <= 0 &&
1087 cur->rule.ptr->max_src_nodes <= 0)
1088 pf_rm_rule(NULL, cur->rule.ptr);
1089 }
1090 RB_REMOVE(pf_src_tree, &tree_src_tracking, cur);
1091 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
1092 pf_status.src_nodes--;
1093 pool_put(&pf_src_tree_pl, cur);
1094 }
1095 }
1096
1097 if (locked && !waslocked)
1098#ifdef __FreeBSD__
1099 sx_downgrade(&pf_consistency_lock);
1100#else
1101 rw_exit_write(&pf_consistency_lock);
1102#endif
1103}
1104
1105void
1106pf_src_tree_remove_state(struct pf_state *s)
1107{
1108 u_int32_t timeout;
1109
1110 if (s->src_node != NULL) {
1111 if (s->proto == IPPROTO_TCP) {
1112 if (s->src.tcp_est)
1113 --s->src_node->conn;
1114 }
1115 if (--s->src_node->states <= 0) {
1116 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1117 if (!timeout)
1118 timeout =
1119 pf_default_rule.timeout[PFTM_SRC_NODE];
1120 s->src_node->expire = time_second + timeout;
1121 }
1122 }
1123 if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) {
1124 if (--s->nat_src_node->states <= 0) {
1125 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1126 if (!timeout)
1127 timeout =
1128 pf_default_rule.timeout[PFTM_SRC_NODE];
1129 s->nat_src_node->expire = time_second + timeout;
1130 }
1131 }
1132 s->src_node = s->nat_src_node = NULL;
1133}
1134
1135/* callers should be at splsoftnet */
1136void
1137pf_unlink_state(struct pf_state *cur)
1138{
1139#ifdef __FreeBSD__
1140 if (cur->local_flags & PFSTATE_EXPIRING)
1141 return;
1142 cur->local_flags |= PFSTATE_EXPIRING;
1143#endif
1144 if (cur->src.state == PF_TCPS_PROXY_DST) {
1145#ifdef __FreeBSD__
1146 pf_send_tcp(NULL, cur->rule.ptr, cur->af,
1147#else
1148 pf_send_tcp(cur->rule.ptr, cur->af,
1149#endif
1150 &cur->ext.addr, &cur->lan.addr,
1151 cur->ext.port, cur->lan.port,
1152 cur->src.seqhi, cur->src.seqlo + 1,
1153 TH_RST|TH_ACK, 0, 0, 0, 1, cur->tag, NULL, NULL);
1154 }
1155 RB_REMOVE(pf_state_tree_ext_gwy,
1156 &cur->u.s.kif->pfik_ext_gwy, cur);
1157 RB_REMOVE(pf_state_tree_lan_ext,
1158 &cur->u.s.kif->pfik_lan_ext, cur);
1159 RB_REMOVE(pf_state_tree_id, &tree_id, cur);
1160#if NPFSYNC
1161 if (cur->creatorid == pf_status.hostid)
1162 pfsync_delete_state(cur);
1163#endif
1164 cur->timeout = PFTM_UNLINKED;
1165 pf_src_tree_remove_state(cur);
1166}
1167
1168/* callers should be at splsoftnet and hold the
1169 * write_lock on pf_consistency_lock */
1170void
1171pf_free_state(struct pf_state *cur)
1172{
1173#if NPFSYNC
1174 if (pfsyncif != NULL &&
1175 (pfsyncif->sc_bulk_send_next == cur ||
1176 pfsyncif->sc_bulk_terminator == cur))
1177 return;
1178#endif
1179#ifdef __FreeBSD__
1180 KASSERT(cur->timeout == PFTM_UNLINKED,
1181 ("pf_free_state: cur->timeout != PFTM_UNLINKED"));
1182#else
1183 KASSERT(cur->timeout == PFTM_UNLINKED);
1184#endif
1185 if (--cur->rule.ptr->states <= 0 &&
1186 cur->rule.ptr->src_nodes <= 0)
1187 pf_rm_rule(NULL, cur->rule.ptr);
1188 if (cur->nat_rule.ptr != NULL)
1189 if (--cur->nat_rule.ptr->states <= 0 &&
1190 cur->nat_rule.ptr->src_nodes <= 0)
1191 pf_rm_rule(NULL, cur->nat_rule.ptr);
1192 if (cur->anchor.ptr != NULL)
1193 if (--cur->anchor.ptr->states <= 0)
1194 pf_rm_rule(NULL, cur->anchor.ptr);
1195 pf_normalize_tcp_cleanup(cur);
1196 pfi_kif_unref(cur->u.s.kif, PFI_KIF_REF_STATE);
1197 TAILQ_REMOVE(&state_list, cur, u.s.entry_list);
1198 if (cur->tag)
1199 pf_tag_unref(cur->tag);
1200 pool_put(&pf_state_pl, cur);
1201 pf_status.fcounters[FCNT_STATE_REMOVALS]++;
1202 pf_status.states--;
1203}
1204
1205void
1206pf_purge_expired_states(u_int32_t maxcheck)
1207{
1208 static struct pf_state *cur = NULL;
1209 struct pf_state *next;
1210 int locked = 0;
1211
1212 while (maxcheck--) {
1213 /* wrap to start of list when we hit the end */
1214 if (cur == NULL) {
1215 cur = TAILQ_FIRST(&state_list);
1216 if (cur == NULL)
1217 break; /* list empty */
1218 }
1219
1220 /* get next state, as cur may get deleted */
1221 next = TAILQ_NEXT(cur, u.s.entry_list);
1222
1223 if (cur->timeout == PFTM_UNLINKED) {
1224 /* free unlinked state */
1225 if (! locked) {
1226#ifdef __FreeBSD__
1227 if (!sx_try_upgrade(&pf_consistency_lock)) {
1228 PF_UNLOCK();
1229 sx_sunlock(&pf_consistency_lock);
1230 sx_xlock(&pf_consistency_lock);
1231 PF_LOCK();
1232 }
1233#else
1234 rw_enter_write(&pf_consistency_lock);
1235#endif
1236 locked = 1;
1237 }
1238 pf_free_state(cur);
1239 } else if (pf_state_expires(cur) <= time_second) {
1240 /* unlink and free expired state */
1241 pf_unlink_state(cur);
1242 if (! locked) {
1243#ifdef __FreeBSD__
1244 if (!sx_try_upgrade(&pf_consistency_lock)) {
1245 PF_UNLOCK();
1246 sx_sunlock(&pf_consistency_lock);
1247 sx_xlock(&pf_consistency_lock);
1248 PF_LOCK();
1249 }
1250#else
1251 rw_enter_write(&pf_consistency_lock);
1252#endif
1253 locked = 1;
1254 }
1255 pf_free_state(cur);
1256 }
1257 cur = next;
1258 }
1259
1260 if (locked)
1261#ifdef __FreeBSD__
1262 sx_downgrade(&pf_consistency_lock);
1263#else
1264 rw_exit_write(&pf_consistency_lock);
1265#endif
1266}
1267
1268int
1269pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw)
1270{
1271 if (aw->type != PF_ADDR_TABLE)
1272 return (0);
1273 if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL)
1274 return (1);
1275 return (0);
1276}
1277
1278void
1279pf_tbladdr_remove(struct pf_addr_wrap *aw)
1280{
1281 if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL)
1282 return;
1283 pfr_detach_table(aw->p.tbl);
1284 aw->p.tbl = NULL;
1285}
1286
1287void
1288pf_tbladdr_copyout(struct pf_addr_wrap *aw)
1289{
1290 struct pfr_ktable *kt = aw->p.tbl;
1291
1292 if (aw->type != PF_ADDR_TABLE || kt == NULL)
1293 return;
1294 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
1295 kt = kt->pfrkt_root;
1296 aw->p.tbl = NULL;
1297 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
1298 kt->pfrkt_cnt : -1;
1299}
1300
1301void
1302pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af)
1303{
1304 switch (af) {
1305#ifdef INET
1306 case AF_INET: {
1307 u_int32_t a = ntohl(addr->addr32[0]);
1308 printf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255,
1309 (a>>8)&255, a&255);
1310 if (p) {
1311 p = ntohs(p);
1312 printf(":%u", p);
1313 }
1314 break;
1315 }
1316#endif /* INET */
1317#ifdef INET6
1318 case AF_INET6: {
1319 u_int16_t b;
1320 u_int8_t i, curstart = 255, curend = 0,
1321 maxstart = 0, maxend = 0;
1322 for (i = 0; i < 8; i++) {
1323 if (!addr->addr16[i]) {
1324 if (curstart == 255)
1325 curstart = i;
1326 else
1327 curend = i;
1328 } else {
1329 if (curstart) {
1330 if ((curend - curstart) >
1331 (maxend - maxstart)) {
1332 maxstart = curstart;
1333 maxend = curend;
1334 curstart = 255;
1335 }
1336 }
1337 }
1338 }
1339 for (i = 0; i < 8; i++) {
1340 if (i >= maxstart && i <= maxend) {
1341 if (maxend != 7) {
1342 if (i == maxstart)
1343 printf(":");
1344 } else {
1345 if (i == maxend)
1346 printf(":");
1347 }
1348 } else {
1349 b = ntohs(addr->addr16[i]);
1350 printf("%x", b);
1351 if (i < 7)
1352 printf(":");
1353 }
1354 }
1355 if (p) {
1356 p = ntohs(p);
1357 printf("[%u]", p);
1358 }
1359 break;
1360 }
1361#endif /* INET6 */
1362 }
1363}
1364
1365void
1366pf_print_state(struct pf_state *s)
1367{
1368 switch (s->proto) {
1369 case IPPROTO_TCP:
1370 printf("TCP ");
1371 break;
1372 case IPPROTO_UDP:
1373 printf("UDP ");
1374 break;
1375 case IPPROTO_ICMP:
1376 printf("ICMP ");
1377 break;
1378 case IPPROTO_ICMPV6:
1379 printf("ICMPV6 ");
1380 break;
1381 default:
1382 printf("%u ", s->proto);
1383 break;
1384 }
1385 pf_print_host(&s->lan.addr, s->lan.port, s->af);
1386 printf(" ");
1387 pf_print_host(&s->gwy.addr, s->gwy.port, s->af);
1388 printf(" ");
1389 pf_print_host(&s->ext.addr, s->ext.port, s->af);
1390 printf(" [lo=%u high=%u win=%u modulator=%u", s->src.seqlo,
1391 s->src.seqhi, s->src.max_win, s->src.seqdiff);
1392 if (s->src.wscale && s->dst.wscale)
1393 printf(" wscale=%u", s->src.wscale & PF_WSCALE_MASK);
1394 printf("]");
1395 printf(" [lo=%u high=%u win=%u modulator=%u", s->dst.seqlo,
1396 s->dst.seqhi, s->dst.max_win, s->dst.seqdiff);
1397 if (s->src.wscale && s->dst.wscale)
1398 printf(" wscale=%u", s->dst.wscale & PF_WSCALE_MASK);
1399 printf("]");
1400 printf(" %u:%u", s->src.state, s->dst.state);
1401}
1402
1403void
1404pf_print_flags(u_int8_t f)
1405{
1406 if (f)
1407 printf(" ");
1408 if (f & TH_FIN)
1409 printf("F");
1410 if (f & TH_SYN)
1411 printf("S");
1412 if (f & TH_RST)
1413 printf("R");
1414 if (f & TH_PUSH)
1415 printf("P");
1416 if (f & TH_ACK)
1417 printf("A");
1418 if (f & TH_URG)
1419 printf("U");
1420 if (f & TH_ECE)
1421 printf("E");
1422 if (f & TH_CWR)
1423 printf("W");
1424}
1425
1426#define PF_SET_SKIP_STEPS(i) \
1427 do { \
1428 while (head[i] != cur) { \
1429 head[i]->skip[i].ptr = cur; \
1430 head[i] = TAILQ_NEXT(head[i], entries); \
1431 } \
1432 } while (0)
1433
1434void
1435pf_calc_skip_steps(struct pf_rulequeue *rules)
1436{
1437 struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT];
1438 int i;
1439
1440 cur = TAILQ_FIRST(rules);
1441 prev = cur;
1442 for (i = 0; i < PF_SKIP_COUNT; ++i)
1443 head[i] = cur;
1444 while (cur != NULL) {
1445
1446 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
1447 PF_SET_SKIP_STEPS(PF_SKIP_IFP);
1448 if (cur->direction != prev->direction)
1449 PF_SET_SKIP_STEPS(PF_SKIP_DIR);
1450 if (cur->af != prev->af)
1451 PF_SET_SKIP_STEPS(PF_SKIP_AF);
1452 if (cur->proto != prev->proto)
1453 PF_SET_SKIP_STEPS(PF_SKIP_PROTO);
1454 if (cur->src.neg != prev->src.neg ||
1455 pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr))
1456 PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR);
1457 if (cur->src.port[0] != prev->src.port[0] ||
1458 cur->src.port[1] != prev->src.port[1] ||
1459 cur->src.port_op != prev->src.port_op)
1460 PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
1461 if (cur->dst.neg != prev->dst.neg ||
1462 pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr))
1463 PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR);
1464 if (cur->dst.port[0] != prev->dst.port[0] ||
1465 cur->dst.port[1] != prev->dst.port[1] ||
1466 cur->dst.port_op != prev->dst.port_op)
1467 PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
1468
1469 prev = cur;
1470 cur = TAILQ_NEXT(cur, entries);
1471 }
1472 for (i = 0; i < PF_SKIP_COUNT; ++i)
1473 PF_SET_SKIP_STEPS(i);
1474}
1475
1476int
1477pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2)
1478{
1479 if (aw1->type != aw2->type)
1480 return (1);
1481 switch (aw1->type) {
1482 case PF_ADDR_ADDRMASK:
1483 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0))
1484 return (1);
1485 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0))
1486 return (1);
1487 return (0);
1488 case PF_ADDR_DYNIFTL:
1489 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt);
1490 case PF_ADDR_NOROUTE:
1491 case PF_ADDR_URPFFAILED:
1492 return (0);
1493 case PF_ADDR_TABLE:
1494 return (aw1->p.tbl != aw2->p.tbl);
1495 case PF_ADDR_RTLABEL:
1496 return (aw1->v.rtlabel != aw2->v.rtlabel);
1497 default:
1498 printf("invalid address type: %d\n", aw1->type);
1499 return (1);
1500 }
1501}
1502
1503u_int16_t
1504pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp)
1505{
1506 u_int32_t l;
1507
1508 if (udp && !cksum)
1509 return (0x0000);
1510 l = cksum + old - new;
1511 l = (l >> 16) + (l & 65535);
1512 l = l & 65535;
1513 if (udp && !l)
1514 return (0xFFFF);
1515 return (l);
1516}
1517
1518void
1519pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc,
1520 struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af)
1521{
1522 struct pf_addr ao;
1523 u_int16_t po = *p;
1524
1525 PF_ACPY(&ao, a, af);
1526 PF_ACPY(a, an, af);
1527
1528 *p = pn;
1529
1530 switch (af) {
1531#ifdef INET
1532 case AF_INET:
1533 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1534 ao.addr16[0], an->addr16[0], 0),
1535 ao.addr16[1], an->addr16[1], 0);
1536 *p = pn;
1537 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1538 ao.addr16[0], an->addr16[0], u),
1539 ao.addr16[1], an->addr16[1], u),
1540 po, pn, u);
1541 break;
1542#endif /* INET */
1543#ifdef INET6
1544 case AF_INET6:
1545 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1546 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1547 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1548 ao.addr16[0], an->addr16[0], u),
1549 ao.addr16[1], an->addr16[1], u),
1550 ao.addr16[2], an->addr16[2], u),
1551 ao.addr16[3], an->addr16[3], u),
1552 ao.addr16[4], an->addr16[4], u),
1553 ao.addr16[5], an->addr16[5], u),
1554 ao.addr16[6], an->addr16[6], u),
1555 ao.addr16[7], an->addr16[7], u),
1556 po, pn, u);
1557 break;
1558#endif /* INET6 */
1559 }
1560}
1561
1562
1563/* Changes a u_int32_t. Uses a void * so there are no align restrictions */
1564void
1565pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u)
1566{
1567 u_int32_t ao;
1568
1569 memcpy(&ao, a, sizeof(ao));
1570 memcpy(a, &an, sizeof(u_int32_t));
1571 *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u),
1572 ao % 65536, an % 65536, u);
1573}
1574
1575#ifdef INET6
1576void
1577pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u)
1578{
1579 struct pf_addr ao;
1580
1581 PF_ACPY(&ao, a, AF_INET6);
1582 PF_ACPY(a, an, AF_INET6);
1583
1584 *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1585 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1586 pf_cksum_fixup(pf_cksum_fixup(*c,
1587 ao.addr16[0], an->addr16[0], u),
1588 ao.addr16[1], an->addr16[1], u),
1589 ao.addr16[2], an->addr16[2], u),
1590 ao.addr16[3], an->addr16[3], u),
1591 ao.addr16[4], an->addr16[4], u),
1592 ao.addr16[5], an->addr16[5], u),
1593 ao.addr16[6], an->addr16[6], u),
1594 ao.addr16[7], an->addr16[7], u);
1595}
1596#endif /* INET6 */
1597
1598void
1599pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa,
1600 struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c,
1601 u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af)
1602{
1603 struct pf_addr oia, ooa;
1604
1605 PF_ACPY(&oia, ia, af);
1606 PF_ACPY(&ooa, oa, af);
1607
1608 /* Change inner protocol port, fix inner protocol checksum. */
1609 if (ip != NULL) {
1610 u_int16_t oip = *ip;
1611 u_int32_t opc = 0; /* make the compiler happy */
1612
1613 if (pc != NULL)
1614 opc = *pc;
1615 *ip = np;
1616 if (pc != NULL)
1617 *pc = pf_cksum_fixup(*pc, oip, *ip, u);
1618 *ic = pf_cksum_fixup(*ic, oip, *ip, 0);
1619 if (pc != NULL)
1620 *ic = pf_cksum_fixup(*ic, opc, *pc, 0);
1621 }
1622 /* Change inner ip address, fix inner ip and icmp checksums. */
1623 PF_ACPY(ia, na, af);
1624 switch (af) {
1625#ifdef INET
1626 case AF_INET: {
1627 u_int32_t oh2c = *h2c;
1628
1629 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c,
1630 oia.addr16[0], ia->addr16[0], 0),
1631 oia.addr16[1], ia->addr16[1], 0);
1632 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1633 oia.addr16[0], ia->addr16[0], 0),
1634 oia.addr16[1], ia->addr16[1], 0);
1635 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0);
1636 break;
1637 }
1638#endif /* INET */
1639#ifdef INET6
1640 case AF_INET6:
1641 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1642 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1643 pf_cksum_fixup(pf_cksum_fixup(*ic,
1644 oia.addr16[0], ia->addr16[0], u),
1645 oia.addr16[1], ia->addr16[1], u),
1646 oia.addr16[2], ia->addr16[2], u),
1647 oia.addr16[3], ia->addr16[3], u),
1648 oia.addr16[4], ia->addr16[4], u),
1649 oia.addr16[5], ia->addr16[5], u),
1650 oia.addr16[6], ia->addr16[6], u),
1651 oia.addr16[7], ia->addr16[7], u);
1652 break;
1653#endif /* INET6 */
1654 }
1655 /* Change outer ip address, fix outer ip or icmpv6 checksum. */
1656 PF_ACPY(oa, na, af);
1657 switch (af) {
1658#ifdef INET
1659 case AF_INET:
1660 *hc = pf_cksum_fixup(pf_cksum_fixup(*hc,
1661 ooa.addr16[0], oa->addr16[0], 0),
1662 ooa.addr16[1], oa->addr16[1], 0);
1663 break;
1664#endif /* INET */
1665#ifdef INET6
1666 case AF_INET6:
1667 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1668 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1669 pf_cksum_fixup(pf_cksum_fixup(*ic,
1670 ooa.addr16[0], oa->addr16[0], u),
1671 ooa.addr16[1], oa->addr16[1], u),
1672 ooa.addr16[2], oa->addr16[2], u),
1673 ooa.addr16[3], oa->addr16[3], u),
1674 ooa.addr16[4], oa->addr16[4], u),
1675 ooa.addr16[5], oa->addr16[5], u),
1676 ooa.addr16[6], oa->addr16[6], u),
1677 ooa.addr16[7], oa->addr16[7], u);
1678 break;
1679#endif /* INET6 */
1680 }
1681}
1682
1683
1684/*
1685 * Need to modulate the sequence numbers in the TCP SACK option
1686 * (credits to Krzysztof Pfaff for report and patch)
1687 */
1688int
1689pf_modulate_sack(struct mbuf *m, int off, struct pf_pdesc *pd,
1690 struct tcphdr *th, struct pf_state_peer *dst)
1691{
1692 int hlen = (th->th_off << 2) - sizeof(*th), thoptlen = hlen;
1693#ifdef __FreeBSD__
1694 u_int8_t opts[TCP_MAXOLEN], *opt = opts;
1695#else
1696 u_int8_t opts[MAX_TCPOPTLEN], *opt = opts;
1697#endif
1698 int copyback = 0, i, olen;
1699 struct sackblk sack;
1700
1701#define TCPOLEN_SACKLEN (TCPOLEN_SACK + 2)
1702 if (hlen < TCPOLEN_SACKLEN ||
1703 !pf_pull_hdr(m, off + sizeof(*th), opts, hlen, NULL, NULL, pd->af))
1704 return 0;
1705
1706 while (hlen >= TCPOLEN_SACKLEN) {
1707 olen = opt[1];
1708 switch (*opt) {
1709 case TCPOPT_EOL: /* FALLTHROUGH */
1710 case TCPOPT_NOP:
1711 opt++;
1712 hlen--;
1713 break;
1714 case TCPOPT_SACK:
1715 if (olen > hlen)
1716 olen = hlen;
1717 if (olen >= TCPOLEN_SACKLEN) {
1718 for (i = 2; i + TCPOLEN_SACK <= olen;
1719 i += TCPOLEN_SACK) {
1720 memcpy(&sack, &opt[i], sizeof(sack));
1721 pf_change_a(&sack.start, &th->th_sum,
1722 htonl(ntohl(sack.start) -
1723 dst->seqdiff), 0);
1724 pf_change_a(&sack.end, &th->th_sum,
1725 htonl(ntohl(sack.end) -
1726 dst->seqdiff), 0);
1727 memcpy(&opt[i], &sack, sizeof(sack));
1728 }
1729 copyback = 1;
1730 }
1731 /* FALLTHROUGH */
1732 default:
1733 if (olen < 2)
1734 olen = 2;
1735 hlen -= olen;
1736 opt += olen;
1737 }
1738 }
1739
1740 if (copyback)
1741#ifdef __FreeBSD__
1742 m_copyback(m, off + sizeof(*th), thoptlen, (caddr_t)opts);
1743#else
1744 m_copyback(m, off + sizeof(*th), thoptlen, opts);
1745#endif
1746 return (copyback);
1747}
1748
1749void
1750#ifdef __FreeBSD__
1751pf_send_tcp(struct mbuf *replyto, const struct pf_rule *r, sa_family_t af,
1752#else
1753pf_send_tcp(const struct pf_rule *r, sa_family_t af,
1754#endif
1755 const struct pf_addr *saddr, const struct pf_addr *daddr,
1756 u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack,
1757 u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl, int tag,
1758 u_int16_t rtag, struct ether_header *eh, struct ifnet *ifp)
1759{
1760 struct mbuf *m;
1761 int len, tlen;
1762#ifdef INET
1763 struct ip *h;
1764#endif /* INET */
1765#ifdef INET6
1766 struct ip6_hdr *h6;
1767#endif /* INET6 */
1768 struct tcphdr *th;
1769 char *opt;
1770 struct pf_mtag *pf_mtag;
1771
1772#ifdef __FreeBSD__
1773 KASSERT(
1774#ifdef INET
1775 af == AF_INET
1776#else
1777 0
1778#endif
1779 ||
1780#ifdef INET6
1781 af == AF_INET6
1782#else
1783 0
1784#endif
1785 , ("Unsupported AF %d", af));
1786 len = 0;
1787 th = NULL;
1788#ifdef INET
1789 h = NULL;
1790#endif
1791#ifdef INET6
1792 h6 = NULL;
1793#endif
1794#endif
1795
1796 /* maximum segment size tcp option */
1797 tlen = sizeof(struct tcphdr);
1798 if (mss)
1799 tlen += 4;
1800
1801 switch (af) {
1802#ifdef INET
1803 case AF_INET:
1804 len = sizeof(struct ip) + tlen;
1805 break;
1806#endif /* INET */
1807#ifdef INET6
1808 case AF_INET6:
1809 len = sizeof(struct ip6_hdr) + tlen;
1810 break;
1811#endif /* INET6 */
1812 }
1813
1814 /* create outgoing mbuf */
1815 m = m_gethdr(M_DONTWAIT, MT_HEADER);
1816 if (m == NULL)
1817 return;
1818#ifdef __FreeBSD__
1819#ifdef MAC
1820 if (replyto)
1821 mac_netinet_firewall_reply(replyto, m);
1822 else
1823 mac_netinet_firewall_send(m);
1824#else
1825 (void)replyto;
1826#endif
1827#endif
1828 if ((pf_mtag = pf_get_mtag(m)) == NULL) {
1829 m_freem(m);
1830 return;
1831 }
1832 if (tag)
1833#ifdef __FreeBSD__
1834 m->m_flags |= M_SKIP_FIREWALL;
1835#else
1836 pf_mtag->flags |= PF_TAG_GENERATED;
1837#endif
1838
1839 pf_mtag->tag = rtag;
1840
1841 if (r != NULL && r->rtableid >= 0)
1842#ifdef __FreeBSD__
1843 {
1844 M_SETFIB(m, r->rtableid);
1845#endif
1846 pf_mtag->rtableid = r->rtableid;
1847#ifdef __FreeBSD__
1848 }
1849#endif
1850#ifdef ALTQ
1851 if (r != NULL && r->qid) {
1852 pf_mtag->qid = r->qid;
1853 /* add hints for ecn */
1854 pf_mtag->af = af;
1855 pf_mtag->hdr = mtod(m, struct ip *);
1856 }
1857#endif /* ALTQ */
1858 m->m_data += max_linkhdr;
1859 m->m_pkthdr.len = m->m_len = len;
1860 m->m_pkthdr.rcvif = NULL;
1861 bzero(m->m_data, len);
1862 switch (af) {
1863#ifdef INET
1864 case AF_INET:
1865 h = mtod(m, struct ip *);
1866
1867 /* IP header fields included in the TCP checksum */
1868 h->ip_p = IPPROTO_TCP;
1869 h->ip_len = htons(tlen);
1870 h->ip_src.s_addr = saddr->v4.s_addr;
1871 h->ip_dst.s_addr = daddr->v4.s_addr;
1872
1873 th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip));
1874 break;
1875#endif /* INET */
1876#ifdef INET6
1877 case AF_INET6:
1878 h6 = mtod(m, struct ip6_hdr *);
1879
1880 /* IP header fields included in the TCP checksum */
1881 h6->ip6_nxt = IPPROTO_TCP;
1882 h6->ip6_plen = htons(tlen);
1883 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr));
1884 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr));
1885
1886 th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr));
1887 break;
1888#endif /* INET6 */
1889 }
1890
1891 /* TCP header */
1892 th->th_sport = sport;
1893 th->th_dport = dport;
1894 th->th_seq = htonl(seq);
1895 th->th_ack = htonl(ack);
1896 th->th_off = tlen >> 2;
1897 th->th_flags = flags;
1898 th->th_win = htons(win);
1899
1900 if (mss) {
1901 opt = (char *)(th + 1);
1902 opt[0] = TCPOPT_MAXSEG;
1903 opt[1] = 4;
1904 HTONS(mss);
1905 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2);
1906 }
1907
1908 switch (af) {
1909#ifdef INET
1910 case AF_INET:
1911 /* TCP checksum */
1912 th->th_sum = in_cksum(m, len);
1913
1914 /* Finish the IP header */
1915 h->ip_v = 4;
1916 h->ip_hl = sizeof(*h) >> 2;
1917 h->ip_tos = IPTOS_LOWDELAY;
1918#ifdef __FreeBSD__
1919 h->ip_off = V_path_mtu_discovery ? IP_DF : 0;
1920 h->ip_len = len;
1921#else
1922 h->ip_off = htons(ip_mtudisc ? IP_DF : 0);
1923 h->ip_len = htons(len);
1924#endif
1925 h->ip_ttl = ttl ? ttl : V_ip_defttl;
1926 h->ip_sum = 0;
1927 if (eh == NULL) {
1928#ifdef __FreeBSD__
1929 PF_UNLOCK();
1930 ip_output(m, (void *)NULL, (void *)NULL, 0,
1931 (void *)NULL, (void *)NULL);
1932 PF_LOCK();
1933#else /* ! __FreeBSD__ */
1934 ip_output(m, (void *)NULL, (void *)NULL, 0,
1935 (void *)NULL, (void *)NULL);
1936#endif
1937 } else {
1938 struct route ro;
1939 struct rtentry rt;
1940 struct ether_header *e = (void *)ro.ro_dst.sa_data;
1941
1942 if (ifp == NULL) {
1943 m_freem(m);
1944 return;
1945 }
1946 rt.rt_ifp = ifp;
1947 ro.ro_rt = &rt;
1948 ro.ro_dst.sa_len = sizeof(ro.ro_dst);
1949 ro.ro_dst.sa_family = pseudo_AF_HDRCMPLT;
1950 bcopy(eh->ether_dhost, e->ether_shost, ETHER_ADDR_LEN);
1951 bcopy(eh->ether_shost, e->ether_dhost, ETHER_ADDR_LEN);
1952 e->ether_type = eh->ether_type;
1953#ifdef __FreeBSD__
1954 PF_UNLOCK();
1955 /* XXX_IMPORT: later */
1956 ip_output(m, (void *)NULL, &ro, 0,
1957 (void *)NULL, (void *)NULL);
1958 PF_LOCK();
1959#else /* ! __FreeBSD__ */
1960 ip_output(m, (void *)NULL, &ro, IP_ROUTETOETHER,
1961 (void *)NULL, (void *)NULL);
1962#endif
1963 }
1964 break;
1965#endif /* INET */
1966#ifdef INET6
1967 case AF_INET6:
1968 /* TCP checksum */
1969 th->th_sum = in6_cksum(m, IPPROTO_TCP,
1970 sizeof(struct ip6_hdr), tlen);
1971
1972 h6->ip6_vfc |= IPV6_VERSION;
1973 h6->ip6_hlim = IPV6_DEFHLIM;
1974
1975#ifdef __FreeBSD__
1976 PF_UNLOCK();
1977 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
1978 PF_LOCK();
1979#else
1980 ip6_output(m, NULL, NULL, 0, NULL, NULL);
1981#endif
1982 break;
1983#endif /* INET6 */
1984 }
1985}
1986
1987void
1988pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af,
1989 struct pf_rule *r)
1990{
1991 struct pf_mtag *pf_mtag;
1992 struct mbuf *m0;
1993#ifdef __FreeBSD__
1994 struct ip *ip;
1995#endif
1996
1997#ifdef __FreeBSD__
1998 m0 = m_copypacket(m, M_DONTWAIT);
1999 if (m0 == NULL)
2000 return;
2001#else
2002 m0 = m_copy(m, 0, M_COPYALL);
2003#endif
2004 if ((pf_mtag = pf_get_mtag(m0)) == NULL)
2005 return;
2006#ifdef __FreeBSD__
2007 /* XXX: revisit */
2008 m0->m_flags |= M_SKIP_FIREWALL;
2009#else
2010 pf_mtag->flags |= PF_TAG_GENERATED;
2011#endif
2012
2013 if (r->rtableid >= 0)
2014#ifdef __FreeBSD__
2015 {
2016 M_SETFIB(m0, r->rtableid);
2017#endif
2018 pf_mtag->rtableid = r->rtableid;
2019#ifdef __FreeBSD__
2020 }
2021#endif
2022
2023#ifdef ALTQ
2024 if (r->qid) {
2025 pf_mtag->qid = r->qid;
2026 /* add hints for ecn */
2027 pf_mtag->af = af;
2028 pf_mtag->hdr = mtod(m0, struct ip *);
2029 }
2030#endif /* ALTQ */
2031
2032 switch (af) {
2033#ifdef INET
2034 case AF_INET:
2035#ifdef __FreeBSD__
2036 /* icmp_error() expects host byte ordering */
2037 ip = mtod(m0, struct ip *);
2038 NTOHS(ip->ip_len);
2039 NTOHS(ip->ip_off);
2040 PF_UNLOCK();
2041 icmp_error(m0, type, code, 0, 0);
2042 PF_LOCK();
2043#else
2044 icmp_error(m0, type, code, 0, 0);
2045#endif
2046 break;
2047#endif /* INET */
2048#ifdef INET6
2049 case AF_INET6:
2050#ifdef __FreeBSD__
2051 PF_UNLOCK();
2052#endif
2053 icmp6_error(m0, type, code, 0);
2054#ifdef __FreeBSD__
2055 PF_LOCK();
2056#endif
2057 break;
2058#endif /* INET6 */
2059 }
2060}
2061
2062/*
2063 * Return 1 if the addresses a and b match (with mask m), otherwise return 0.
2064 * If n is 0, they match if they are equal. If n is != 0, they match if they
2065 * are different.
2066 */
2067int
2068pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m,
2069 struct pf_addr *b, sa_family_t af)
2070{
2071 int match = 0;
2072
2073 switch (af) {
2074#ifdef INET
2075 case AF_INET:
2076 if ((a->addr32[0] & m->addr32[0]) ==
2077 (b->addr32[0] & m->addr32[0]))
2078 match++;
2079 break;
2080#endif /* INET */
2081#ifdef INET6
2082 case AF_INET6:
2083 if (((a->addr32[0] & m->addr32[0]) ==
2084 (b->addr32[0] & m->addr32[0])) &&
2085 ((a->addr32[1] & m->addr32[1]) ==
2086 (b->addr32[1] & m->addr32[1])) &&
2087 ((a->addr32[2] & m->addr32[2]) ==
2088 (b->addr32[2] & m->addr32[2])) &&
2089 ((a->addr32[3] & m->addr32[3]) ==
2090 (b->addr32[3] & m->addr32[3])))
2091 match++;
2092 break;
2093#endif /* INET6 */
2094 }
2095 if (match) {
2096 if (n)
2097 return (0);
2098 else
2099 return (1);
2100 } else {
2101 if (n)
2102 return (1);
2103 else
2104 return (0);
2105 }
2106}
2107
2108int
2109pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p)
2110{
2111 switch (op) {
2112 case PF_OP_IRG:
2113 return ((p > a1) && (p < a2));
2114 case PF_OP_XRG:
2115 return ((p < a1) || (p > a2));
2116 case PF_OP_RRG:
2117 return ((p >= a1) && (p <= a2));
2118 case PF_OP_EQ:
2119 return (p == a1);
2120 case PF_OP_NE:
2121 return (p != a1);
2122 case PF_OP_LT:
2123 return (p < a1);
2124 case PF_OP_LE:
2125 return (p <= a1);
2126 case PF_OP_GT:
2127 return (p > a1);
2128 case PF_OP_GE:
2129 return (p >= a1);
2130 }
2131 return (0); /* never reached */
2132}
2133
2134int
2135pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p)
2136{
2137 NTOHS(a1);
2138 NTOHS(a2);
2139 NTOHS(p);
2140 return (pf_match(op, a1, a2, p));
2141}
2142
2143int
2144pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u)
2145{
2146 if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
2147 return (0);
2148 return (pf_match(op, a1, a2, u));
2149}
2150
2151int
2152pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g)
2153{
2154 if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
2155 return (0);
2156 return (pf_match(op, a1, a2, g));
2157}
2158
2159#ifndef __FreeBSD__
2160struct pf_mtag *
2161pf_find_mtag(struct mbuf *m)
2162{
2163 struct m_tag *mtag;
2164
2165 if ((mtag = m_tag_find(m, PACKET_TAG_PF, NULL)) == NULL)
2166 return (NULL);
2167
2168 return ((struct pf_mtag *)(mtag + 1));
2169}
2170
2171struct pf_mtag *
2172pf_get_mtag(struct mbuf *m)
2173{
2174 struct m_tag *mtag;
2175
2176 if ((mtag = m_tag_find(m, PACKET_TAG_PF, NULL)) == NULL) {
2177 mtag = m_tag_get(PACKET_TAG_PF, sizeof(struct pf_mtag),
2178 M_NOWAIT);
2179 if (mtag == NULL)
2180 return (NULL);
2181 bzero(mtag + 1, sizeof(struct pf_mtag));
2182 m_tag_prepend(m, mtag);
2183 }
2184
2185 return ((struct pf_mtag *)(mtag + 1));
2186}
2187#endif
2188
2189int
2190pf_match_tag(struct mbuf *m, struct pf_rule *r, struct pf_mtag *pf_mtag,
2191 int *tag)
2192{
2193 if (*tag == -1)
2194 *tag = pf_mtag->tag;
2195
2196 return ((!r->match_tag_not && r->match_tag == *tag) ||
2197 (r->match_tag_not && r->match_tag != *tag));
2198}
2199
2200int
2201pf_tag_packet(struct mbuf *m, struct pf_mtag *pf_mtag, int tag, int rtableid)
2202{
2203 if (tag <= 0 && rtableid < 0)
2204 return (0);
2205
2206 if (pf_mtag == NULL)
2207 if ((pf_mtag = pf_get_mtag(m)) == NULL)
2208 return (1);
2209 if (tag > 0)
2210 pf_mtag->tag = tag;
2211 if (rtableid >= 0)
2212#ifdef __FreeBSD__
2213 {
2214 M_SETFIB(m, rtableid);
2215#endif
2216 pf_mtag->rtableid = rtableid;
2217#ifdef __FreeBSD__
2218 }
2219#endif
2220
2221 return (0);
2222}
2223
2224static void
2225pf_step_into_anchor(int *depth, struct pf_ruleset **rs, int n,
2226 struct pf_rule **r, struct pf_rule **a, int *match)
2227{
2228 struct pf_anchor_stackframe *f;
2229
2230 (*r)->anchor->match = 0;
2231 if (match)
2232 *match = 0;
2233 if (*depth >= sizeof(pf_anchor_stack) /
2234 sizeof(pf_anchor_stack[0])) {
2235 printf("pf_step_into_anchor: stack overflow\n");
2236 *r = TAILQ_NEXT(*r, entries);
2237 return;
2238 } else if (*depth == 0 && a != NULL)
2239 *a = *r;
2240 f = pf_anchor_stack + (*depth)++;
2241 f->rs = *rs;
2242 f->r = *r;
2243 if ((*r)->anchor_wildcard) {
2244 f->parent = &(*r)->anchor->children;
2245 if ((f->child = RB_MIN(pf_anchor_node, f->parent)) ==
2246 NULL) {
2247 *r = NULL;
2248 return;
2249 }
2250 *rs = &f->child->ruleset;
2251 } else {
2252 f->parent = NULL;
2253 f->child = NULL;
2254 *rs = &(*r)->anchor->ruleset;
2255 }
2256 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
2257}
2258
2259int
2260pf_step_out_of_anchor(int *depth, struct pf_ruleset **rs, int n,
2261 struct pf_rule **r, struct pf_rule **a, int *match)
2262{
2263 struct pf_anchor_stackframe *f;
2264 int quick = 0;
2265
2266 do {
2267 if (*depth <= 0)
2268 break;
2269 f = pf_anchor_stack + *depth - 1;
2270 if (f->parent != NULL && f->child != NULL) {
2271 if (f->child->match ||
2272 (match != NULL && *match)) {
2273 f->r->anchor->match = 1;
2274 *match = 0;
2275 }
2276 f->child = RB_NEXT(pf_anchor_node, f->parent, f->child);
2277 if (f->child != NULL) {
2278 *rs = &f->child->ruleset;
2279 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
2280 if (*r == NULL)
2281 continue;
2282 else
2283 break;
2284 }
2285 }
2286 (*depth)--;
2287 if (*depth == 0 && a != NULL)
2288 *a = NULL;
2289 *rs = f->rs;
2290 if (f->r->anchor->match || (match != NULL && *match))
2291 quick = f->r->quick;
2292 *r = TAILQ_NEXT(f->r, entries);
2293 } while (*r == NULL);
2294
2295 return (quick);
2296}
2297
2298#ifdef INET6
2299void
2300pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr,
2301 struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af)
2302{
2303 switch (af) {
2304#ifdef INET
2305 case AF_INET:
2306 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2307 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2308 break;
2309#endif /* INET */
2310 case AF_INET6:
2311 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2312 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2313 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) |
2314 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]);
2315 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) |
2316 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]);
2317 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) |
2318 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]);
2319 break;
2320 }
2321}
2322
2323void
2324pf_addr_inc(struct pf_addr *addr, sa_family_t af)
2325{
2326 switch (af) {
2327#ifdef INET
2328 case AF_INET:
2329 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1);
2330 break;
2331#endif /* INET */
2332 case AF_INET6:
2333 if (addr->addr32[3] == 0xffffffff) {
2334 addr->addr32[3] = 0;
2335 if (addr->addr32[2] == 0xffffffff) {
2336 addr->addr32[2] = 0;
2337 if (addr->addr32[1] == 0xffffffff) {
2338 addr->addr32[1] = 0;
2339 addr->addr32[0] =
2340 htonl(ntohl(addr->addr32[0]) + 1);
2341 } else
2342 addr->addr32[1] =
2343 htonl(ntohl(addr->addr32[1]) + 1);
2344 } else
2345 addr->addr32[2] =
2346 htonl(ntohl(addr->addr32[2]) + 1);
2347 } else
2348 addr->addr32[3] =
2349 htonl(ntohl(addr->addr32[3]) + 1);
2350 break;
2351 }
2352}
2353#endif /* INET6 */
2354
2355#define mix(a,b,c) \
2356 do { \
2357 a -= b; a -= c; a ^= (c >> 13); \
2358 b -= c; b -= a; b ^= (a << 8); \
2359 c -= a; c -= b; c ^= (b >> 13); \
2360 a -= b; a -= c; a ^= (c >> 12); \
2361 b -= c; b -= a; b ^= (a << 16); \
2362 c -= a; c -= b; c ^= (b >> 5); \
2363 a -= b; a -= c; a ^= (c >> 3); \
2364 b -= c; b -= a; b ^= (a << 10); \
2365 c -= a; c -= b; c ^= (b >> 15); \
2366 } while (0)
2367
2368/*
2369 * hash function based on bridge_hash in if_bridge.c
2370 */
2371void
2372pf_hash(struct pf_addr *inaddr, struct pf_addr *hash,
2373 struct pf_poolhashkey *key, sa_family_t af)
2374{
2375 u_int32_t a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0];
2376
2377 switch (af) {
2378#ifdef INET
2379 case AF_INET:
2380 a += inaddr->addr32[0];
2381 b += key->key32[1];
2382 mix(a, b, c);
2383 hash->addr32[0] = c + key->key32[2];
2384 break;
2385#endif /* INET */
2386#ifdef INET6
2387 case AF_INET6:
2388 a += inaddr->addr32[0];
2389 b += inaddr->addr32[2];
2390 mix(a, b, c);
2391 hash->addr32[0] = c;
2392 a += inaddr->addr32[1];
2393 b += inaddr->addr32[3];
2394 c += key->key32[1];
2395 mix(a, b, c);
2396 hash->addr32[1] = c;
2397 a += inaddr->addr32[2];
2398 b += inaddr->addr32[1];
2399 c += key->key32[2];
2400 mix(a, b, c);
2401 hash->addr32[2] = c;
2402 a += inaddr->addr32[3];
2403 b += inaddr->addr32[0];
2404 c += key->key32[3];
2405 mix(a, b, c);
2406 hash->addr32[3] = c;
2407 break;
2408#endif /* INET6 */
2409 }
2410}
2411
2412int
2413pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr,
2414 struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn)
2415{
2416 unsigned char hash[16];
2417 struct pf_pool *rpool = &r->rpool;
2418 struct pf_addr *raddr = &rpool->cur->addr.v.a.addr;
2419 struct pf_addr *rmask = &rpool->cur->addr.v.a.mask;
2420 struct pf_pooladdr *acur = rpool->cur;
2421 struct pf_src_node k;
2422
2423 if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR &&
2424 (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2425 k.af = af;
2426 PF_ACPY(&k.addr, saddr, af);
2427 if (r->rule_flag & PFRULE_RULESRCTRACK ||
2428 r->rpool.opts & PF_POOL_STICKYADDR)
2429 k.rule.ptr = r;
2430 else
2431 k.rule.ptr = NULL;
2432 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
2433 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
2434 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) {
2435 PF_ACPY(naddr, &(*sn)->raddr, af);
2436 if (pf_status.debug >= PF_DEBUG_MISC) {
2437 printf("pf_map_addr: src tracking maps ");
2438 pf_print_host(&k.addr, 0, af);
2439 printf(" to ");
2440 pf_print_host(naddr, 0, af);
2441 printf("\n");
2442 }
2443 return (0);
2444 }
2445 }
2446
2447 if (rpool->cur->addr.type == PF_ADDR_NOROUTE)
2448 return (1);
2449 if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2450 switch (af) {
2451#ifdef INET
2452 case AF_INET:
2453 if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 &&
2454 (rpool->opts & PF_POOL_TYPEMASK) !=
2455 PF_POOL_ROUNDROBIN)
2456 return (1);
2457 raddr = &rpool->cur->addr.p.dyn->pfid_addr4;
2458 rmask = &rpool->cur->addr.p.dyn->pfid_mask4;
2459 break;
2460#endif /* INET */
2461#ifdef INET6
2462 case AF_INET6:
2463 if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 &&
2464 (rpool->opts & PF_POOL_TYPEMASK) !=
2465 PF_POOL_ROUNDROBIN)
2466 return (1);
2467 raddr = &rpool->cur->addr.p.dyn->pfid_addr6;
2468 rmask = &rpool->cur->addr.p.dyn->pfid_mask6;
2469 break;
2470#endif /* INET6 */
2471 }
2472 } else if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2473 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN)
2474 return (1); /* unsupported */
2475 } else {
2476 raddr = &rpool->cur->addr.v.a.addr;
2477 rmask = &rpool->cur->addr.v.a.mask;
2478 }
2479
2480 switch (rpool->opts & PF_POOL_TYPEMASK) {
2481 case PF_POOL_NONE:
2482 PF_ACPY(naddr, raddr, af);
2483 break;
2484 case PF_POOL_BITMASK:
2485 PF_POOLMASK(naddr, raddr, rmask, saddr, af);
2486 break;
2487 case PF_POOL_RANDOM:
2488 if (init_addr != NULL && PF_AZERO(init_addr, af)) {
2489 switch (af) {
2490#ifdef INET
2491 case AF_INET:
2492 rpool->counter.addr32[0] = htonl(arc4random());
2493 break;
2494#endif /* INET */
2495#ifdef INET6
2496 case AF_INET6:
2497 if (rmask->addr32[3] != 0xffffffff)
2498 rpool->counter.addr32[3] =
2499 htonl(arc4random());
2500 else
2501 break;
2502 if (rmask->addr32[2] != 0xffffffff)
2503 rpool->counter.addr32[2] =
2504 htonl(arc4random());
2505 else
2506 break;
2507 if (rmask->addr32[1] != 0xffffffff)
2508 rpool->counter.addr32[1] =
2509 htonl(arc4random());
2510 else
2511 break;
2512 if (rmask->addr32[0] != 0xffffffff)
2513 rpool->counter.addr32[0] =
2514 htonl(arc4random());
2515 break;
2516#endif /* INET6 */
2517 }
2518 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2519 PF_ACPY(init_addr, naddr, af);
2520
2521 } else {
2522 PF_AINC(&rpool->counter, af);
2523 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2524 }
2525 break;
2526 case PF_POOL_SRCHASH:
2527 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af);
2528 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af);
2529 break;
2530 case PF_POOL_ROUNDROBIN:
2531 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2532 if (!pfr_pool_get(rpool->cur->addr.p.tbl,
2533 &rpool->tblidx, &rpool->counter,
2534 &raddr, &rmask, af))
2535 goto get_addr;
2536 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2537 if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2538 &rpool->tblidx, &rpool->counter,
2539 &raddr, &rmask, af))
2540 goto get_addr;
2541 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af))
2542 goto get_addr;
2543
2544 try_next:
2545 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL)
2546 rpool->cur = TAILQ_FIRST(&rpool->list);
2547 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2548 rpool->tblidx = -1;
2549 if (pfr_pool_get(rpool->cur->addr.p.tbl,
2550 &rpool->tblidx, &rpool->counter,
2551 &raddr, &rmask, af)) {
2552 /* table contains no address of type 'af' */
2553 if (rpool->cur != acur)
2554 goto try_next;
2555 return (1);
2556 }
2557 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2558 rpool->tblidx = -1;
2559 if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2560 &rpool->tblidx, &rpool->counter,
2561 &raddr, &rmask, af)) {
2562 /* table contains no address of type 'af' */
2563 if (rpool->cur != acur)
2564 goto try_next;
2565 return (1);
2566 }
2567 } else {
2568 raddr = &rpool->cur->addr.v.a.addr;
2569 rmask = &rpool->cur->addr.v.a.mask;
2570 PF_ACPY(&rpool->counter, raddr, af);
2571 }
2572
2573 get_addr:
2574 PF_ACPY(naddr, &rpool->counter, af);
2575 if (init_addr != NULL && PF_AZERO(init_addr, af))
2576 PF_ACPY(init_addr, naddr, af);
2577 PF_AINC(&rpool->counter, af);
2578 break;
2579 }
2580 if (*sn != NULL)
2581 PF_ACPY(&(*sn)->raddr, naddr, af);
2582
2583 if (pf_status.debug >= PF_DEBUG_MISC &&
2584 (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2585 printf("pf_map_addr: selected address ");
2586 pf_print_host(naddr, 0, af);
2587 printf("\n");
2588 }
2589
2590 return (0);
2591}
2592
2593int
2594pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r,
2595 struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport,
2596 struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high,
2597 struct pf_src_node **sn)
2598{
2599 struct pf_state_cmp key;
2600 struct pf_addr init_addr;
2601 u_int16_t cut;
2602
2603 bzero(&init_addr, sizeof(init_addr));
2604 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2605 return (1);
2606
2607 if (proto == IPPROTO_ICMP) {
2608 low = 1;
2609 high = 65535;
2610 }
2611
2612 do {
2613 key.af = af;
2614 key.proto = proto;
2615 PF_ACPY(&key.ext.addr, daddr, key.af);
2616 PF_ACPY(&key.gwy.addr, naddr, key.af);
2617 key.ext.port = dport;
2618
2619 /*
2620 * port search; start random, step;
2621 * similar 2 portloop in in_pcbbind
2622 */
2623 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP ||
2624 proto == IPPROTO_ICMP)) {
2625 key.gwy.port = dport;
2626 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
2627 return (0);
2628 } else if (low == 0 && high == 0) {
2629 key.gwy.port = *nport;
2630 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
2631 return (0);
2632 } else if (low == high) {
2633 key.gwy.port = htons(low);
2634 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) {
2635 *nport = htons(low);
2636 return (0);
2637 }
2638 } else {
2639 u_int16_t tmp;
2640
2641 if (low > high) {
2642 tmp = low;
2643 low = high;
2644 high = tmp;
2645 }
2646 /* low < high */
2647 cut = htonl(arc4random()) % (1 + high - low) + low;
2648 /* low <= cut <= high */
2649 for (tmp = cut; tmp <= high; ++(tmp)) {
2650 key.gwy.port = htons(tmp);
2651 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
2652 NULL) {
2653 *nport = htons(tmp);
2654 return (0);
2655 }
2656 }
2657 for (tmp = cut - 1; tmp >= low; --(tmp)) {
2658 key.gwy.port = htons(tmp);
2659 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
2660 NULL) {
2661 *nport = htons(tmp);
2662 return (0);
2663 }
2664 }
2665 }
2666
2667 switch (r->rpool.opts & PF_POOL_TYPEMASK) {
2668 case PF_POOL_RANDOM:
2669 case PF_POOL_ROUNDROBIN:
2670 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2671 return (1);
2672 break;
2673 case PF_POOL_NONE:
2674 case PF_POOL_SRCHASH:
2675 case PF_POOL_BITMASK:
2676 default:
2677 return (1);
2678 }
2679 } while (! PF_AEQ(&init_addr, naddr, af) );
2680
2681 return (1); /* none available */
2682}
2683
2684struct pf_rule *
2685pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off,
2686 int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport,
2687 struct pf_addr *daddr, u_int16_t dport, int rs_num)
2688{
2689 struct pf_rule *r, *rm = NULL;
2690 struct pf_ruleset *ruleset = NULL;
2691 int tag = -1;
2692 int rtableid = -1;
2693 int asd = 0;
2694
2695 r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr);
2696 while (r && rm == NULL) {
2697 struct pf_rule_addr *src = NULL, *dst = NULL;
2698 struct pf_addr_wrap *xdst = NULL;
2699
2700 if (r->action == PF_BINAT && direction == PF_IN) {
2701 src = &r->dst;
2702 if (r->rpool.cur != NULL)
2703 xdst = &r->rpool.cur->addr;
2704 } else {
2705 src = &r->src;
2706 dst = &r->dst;
2707 }
2708
2709 r->evaluations++;
2710 if (pfi_kif_match(r->kif, kif) == r->ifnot)
2711 r = r->skip[PF_SKIP_IFP].ptr;
2712 else if (r->direction && r->direction != direction)
2713 r = r->skip[PF_SKIP_DIR].ptr;
2714 else if (r->af && r->af != pd->af)
2715 r = r->skip[PF_SKIP_AF].ptr;
2716 else if (r->proto && r->proto != pd->proto)
2717 r = r->skip[PF_SKIP_PROTO].ptr;
2718 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af,
2719 src->neg, kif))
2720 r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR :
2721 PF_SKIP_DST_ADDR].ptr;
2722 else if (src->port_op && !pf_match_port(src->port_op,
2723 src->port[0], src->port[1], sport))
2724 r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT :
2725 PF_SKIP_DST_PORT].ptr;
2726 else if (dst != NULL &&
2727 PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->neg, NULL))
2728 r = r->skip[PF_SKIP_DST_ADDR].ptr;
2729 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af,
2730 0, NULL))
2731 r = TAILQ_NEXT(r, entries);
2732 else if (dst != NULL && dst->port_op &&
2733 !pf_match_port(dst->port_op, dst->port[0],
2734 dst->port[1], dport))
2735 r = r->skip[PF_SKIP_DST_PORT].ptr;
2736 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag))
2737 r = TAILQ_NEXT(r, entries);
2738 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto !=
2739 IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m,
2740 off, pd->hdr.tcp), r->os_fingerprint)))
2741 r = TAILQ_NEXT(r, entries);
2742 else {
2743 if (r->tag)
2744 tag = r->tag;
2745 if (r->rtableid >= 0)
2746 rtableid = r->rtableid;
2747 if (r->anchor == NULL) {
2748 rm = r;
2749 } else
2750 pf_step_into_anchor(&asd, &ruleset, rs_num,
2751 &r, NULL, NULL);
2752 }
2753 if (r == NULL)
2754 pf_step_out_of_anchor(&asd, &ruleset, rs_num, &r,
2755 NULL, NULL);
2756 }
2757 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid))
2758 return (NULL);
2759 if (rm != NULL && (rm->action == PF_NONAT ||
2760 rm->action == PF_NORDR || rm->action == PF_NOBINAT))
2761 return (NULL);
2762 return (rm);
2763}
2764
2765struct pf_rule *
2766pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction,
2767 struct pfi_kif *kif, struct pf_src_node **sn,
2768 struct pf_addr *saddr, u_int16_t sport,
2769 struct pf_addr *daddr, u_int16_t dport,
2770 struct pf_addr *naddr, u_int16_t *nport)
2771{
2772 struct pf_rule *r = NULL;
2773
2774 if (direction == PF_OUT) {
2775 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2776 sport, daddr, dport, PF_RULESET_BINAT);
2777 if (r == NULL)
2778 r = pf_match_translation(pd, m, off, direction, kif,
2779 saddr, sport, daddr, dport, PF_RULESET_NAT);
2780 } else {
2781 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2782 sport, daddr, dport, PF_RULESET_RDR);
2783 if (r == NULL)
2784 r = pf_match_translation(pd, m, off, direction, kif,
2785 saddr, sport, daddr, dport, PF_RULESET_BINAT);
2786 }
2787
2788 if (r != NULL) {
2789 switch (r->action) {
2790 case PF_NONAT:
2791 case PF_NOBINAT:
2792 case PF_NORDR:
2793 return (NULL);
2794 case PF_NAT:
2795 if (pf_get_sport(pd->af, pd->proto, r, saddr,
2796 daddr, dport, naddr, nport, r->rpool.proxy_port[0],
2797 r->rpool.proxy_port[1], sn)) {
2798 DPFPRINTF(PF_DEBUG_MISC,
2799 ("pf: NAT proxy port allocation "
2800 "(%u-%u) failed\n",
2801 r->rpool.proxy_port[0],
2802 r->rpool.proxy_port[1]));
2803 return (NULL);
2804 }
2805 break;
2806 case PF_BINAT:
2807 switch (direction) {
2808 case PF_OUT:
2809 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){
2810 switch (pd->af) {
2811#ifdef INET
2812 case AF_INET:
2813 if (r->rpool.cur->addr.p.dyn->
2814 pfid_acnt4 < 1)
2815 return (NULL);
2816 PF_POOLMASK(naddr,
2817 &r->rpool.cur->addr.p.dyn->
2818 pfid_addr4,
2819 &r->rpool.cur->addr.p.dyn->
2820 pfid_mask4,
2821 saddr, AF_INET);
2822 break;
2823#endif /* INET */
2824#ifdef INET6
2825 case AF_INET6:
2826 if (r->rpool.cur->addr.p.dyn->
2827 pfid_acnt6 < 1)
2828 return (NULL);
2829 PF_POOLMASK(naddr,
2830 &r->rpool.cur->addr.p.dyn->
2831 pfid_addr6,
2832 &r->rpool.cur->addr.p.dyn->
2833 pfid_mask6,
2834 saddr, AF_INET6);
2835 break;
2836#endif /* INET6 */
2837 }
2838 } else
2839 PF_POOLMASK(naddr,
2840 &r->rpool.cur->addr.v.a.addr,
2841 &r->rpool.cur->addr.v.a.mask,
2842 saddr, pd->af);
2843 break;
2844 case PF_IN:
2845 if (r->src.addr.type == PF_ADDR_DYNIFTL) {
2846 switch (pd->af) {
2847#ifdef INET
2848 case AF_INET:
2849 if (r->src.addr.p.dyn->
2850 pfid_acnt4 < 1)
2851 return (NULL);
2852 PF_POOLMASK(naddr,
2853 &r->src.addr.p.dyn->
2854 pfid_addr4,
2855 &r->src.addr.p.dyn->
2856 pfid_mask4,
2857 daddr, AF_INET);
2858 break;
2859#endif /* INET */
2860#ifdef INET6
2861 case AF_INET6:
2862 if (r->src.addr.p.dyn->
2863 pfid_acnt6 < 1)
2864 return (NULL);
2865 PF_POOLMASK(naddr,
2866 &r->src.addr.p.dyn->
2867 pfid_addr6,
2868 &r->src.addr.p.dyn->
2869 pfid_mask6,
2870 daddr, AF_INET6);
2871 break;
2872#endif /* INET6 */
2873 }
2874 } else
2875 PF_POOLMASK(naddr,
2876 &r->src.addr.v.a.addr,
2877 &r->src.addr.v.a.mask, daddr,
2878 pd->af);
2879 break;
2880 }
2881 break;
2882 case PF_RDR: {
2883 if (pf_map_addr(pd->af, r, saddr, naddr, NULL, sn))
2884 return (NULL);
2885 if ((r->rpool.opts & PF_POOL_TYPEMASK) ==
2886 PF_POOL_BITMASK)
2887 PF_POOLMASK(naddr, naddr,
2888 &r->rpool.cur->addr.v.a.mask, daddr,
2889 pd->af);
2890
2891 if (r->rpool.proxy_port[1]) {
2892 u_int32_t tmp_nport;
2893
2894 tmp_nport = ((ntohs(dport) -
2895 ntohs(r->dst.port[0])) %
2896 (r->rpool.proxy_port[1] -
2897 r->rpool.proxy_port[0] + 1)) +
2898 r->rpool.proxy_port[0];
2899
2900 /* wrap around if necessary */
2901 if (tmp_nport > 65535)
2902 tmp_nport -= 65535;
2903 *nport = htons((u_int16_t)tmp_nport);
2904 } else if (r->rpool.proxy_port[0])
2905 *nport = htons(r->rpool.proxy_port[0]);
2906 break;
2907 }
2908 default:
2909 return (NULL);
2910 }
2911 }
2912
2913 return (r);
2914}
2915
2916int
2917#ifdef __FreeBSD__
2918pf_socket_lookup(int direction, struct pf_pdesc *pd, struct inpcb *inp_arg)
2919#else
2920pf_socket_lookup(int direction, struct pf_pdesc *pd)
2921#endif
2922{
2923 struct pf_addr *saddr, *daddr;
2924 u_int16_t sport, dport;
2925#ifdef __FreeBSD__
2926 struct inpcbinfo *pi;
2927#else
2928 struct inpcbtable *tb;
2929#endif
2930 struct inpcb *inp;
2931
2932 if (pd == NULL)
2933 return (-1);
2934 pd->lookup.uid = UID_MAX;
2935 pd->lookup.gid = GID_MAX;
2936 pd->lookup.pid = NO_PID; /* XXX: revisit */
2937#ifdef __FreeBSD__
2938 if (inp_arg != NULL) {
2939 INP_LOCK_ASSERT(inp_arg);
2940 pd->lookup.uid = inp_arg->inp_cred->cr_uid;
2941 pd->lookup.gid = inp_arg->inp_cred->cr_groups[0];
2942 return (1);
2943 }
2944#endif
2945 switch (pd->proto) {
2946 case IPPROTO_TCP:
2947 if (pd->hdr.tcp == NULL)
2948 return (-1);
2949 sport = pd->hdr.tcp->th_sport;
2950 dport = pd->hdr.tcp->th_dport;
2951#ifdef __FreeBSD__
2952 pi = &V_tcbinfo;
2953#else
2954 tb = &tcbtable;
2955#endif
2956 break;
2957 case IPPROTO_UDP:
2958 if (pd->hdr.udp == NULL)
2959 return (-1);
2960 sport = pd->hdr.udp->uh_sport;
2961 dport = pd->hdr.udp->uh_dport;
2962#ifdef __FreeBSD__
2963 pi = &V_udbinfo;
2964#else
2965 tb = &udbtable;
2966#endif
2967 break;
2968 default:
2969 return (-1);
2970 }
2971 if (direction == PF_IN) {
2972 saddr = pd->src;
2973 daddr = pd->dst;
2974 } else {
2975 u_int16_t p;
2976
2977 p = sport;
2978 sport = dport;
2979 dport = p;
2980 saddr = pd->dst;
2981 daddr = pd->src;
2982 }
2983 switch (pd->af) {
2984#ifdef INET
2985 case AF_INET:
2986#ifdef __FreeBSD__
2987 INP_INFO_RLOCK(pi); /* XXX LOR */
2988 inp = in_pcblookup_hash(pi, saddr->v4, sport, daddr->v4,
2989 dport, 0, NULL);
2990 if (inp == NULL) {
2991 inp = in_pcblookup_hash(pi, saddr->v4, sport,
2992 daddr->v4, dport, INPLOOKUP_WILDCARD, NULL);
2993 if(inp == NULL) {
2994 INP_INFO_RUNLOCK(pi);
2995 return (-1);
2996 }
2997 }
2998#else
2999 inp = in_pcbhashlookup(tb, saddr->v4, sport, daddr->v4, dport);
3000 if (inp == NULL) {
3001 inp = in_pcblookup_listen(tb, daddr->v4, dport, 0);
3002 if (inp == NULL)
3003 return (-1);
3004 }
3005#endif
3006 break;
3007#endif /* INET */
3008#ifdef INET6
3009 case AF_INET6:
3010#ifdef __FreeBSD__
3011 INP_INFO_RLOCK(pi);
3012 inp = in6_pcblookup_hash(pi, &saddr->v6, sport,
3013 &daddr->v6, dport, 0, NULL);
3014 if (inp == NULL) {
3015 inp = in6_pcblookup_hash(pi, &saddr->v6, sport,
3016 &daddr->v6, dport, INPLOOKUP_WILDCARD, NULL);
3017 if (inp == NULL) {
3018 INP_INFO_RUNLOCK(pi);
3019 return (-1);
3020 }
3021 }
3022#else
3023 inp = in6_pcbhashlookup(tb, &saddr->v6, sport, &daddr->v6,
3024 dport);
3025 if (inp == NULL) {
3026 inp = in6_pcblookup_listen(tb, &daddr->v6, dport, 0);
3027 if (inp == NULL)
3028 return (-1);
3029 }
3030#endif
3031 break;
3032#endif /* INET6 */
3033
3034 default:
3035 return (-1);
3036 }
3037#ifdef __FreeBSD__
3038 pd->lookup.uid = inp->inp_cred->cr_uid;
3039 pd->lookup.gid = inp->inp_cred->cr_groups[0];
3040 INP_INFO_RUNLOCK(pi);
3041#else
3042 pd->lookup.uid = inp->inp_socket->so_euid;
3043 pd->lookup.gid = inp->inp_socket->so_egid;
3044 pd->lookup.pid = inp->inp_socket->so_cpid;
3045#endif
3046 return (1);
3047}
3048
3049u_int8_t
3050pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
3051{
3052 int hlen;
3053 u_int8_t hdr[60];
3054 u_int8_t *opt, optlen;
3055 u_int8_t wscale = 0;
3056
3057 hlen = th_off << 2; /* hlen <= sizeof(hdr) */
3058 if (hlen <= sizeof(struct tcphdr))
3059 return (0);
3060 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
3061 return (0);
3062 opt = hdr + sizeof(struct tcphdr);
3063 hlen -= sizeof(struct tcphdr);
3064 while (hlen >= 3) {
3065 switch (*opt) {
3066 case TCPOPT_EOL:
3067 case TCPOPT_NOP:
3068 ++opt;
3069 --hlen;
3070 break;
3071 case TCPOPT_WINDOW:
3072 wscale = opt[2];
3073 if (wscale > TCP_MAX_WINSHIFT)
3074 wscale = TCP_MAX_WINSHIFT;
3075 wscale |= PF_WSCALE_FLAG;
3076 /* FALLTHROUGH */
3077 default:
3078 optlen = opt[1];
3079 if (optlen < 2)
3080 optlen = 2;
3081 hlen -= optlen;
3082 opt += optlen;
3083 break;
3084 }
3085 }
3086 return (wscale);
3087}
3088
3089u_int16_t
3090pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
3091{
3092 int hlen;
3093 u_int8_t hdr[60];
3094 u_int8_t *opt, optlen;
3095 u_int16_t mss = V_tcp_mssdflt;
3096
3097 hlen = th_off << 2; /* hlen <= sizeof(hdr) */
3098 if (hlen <= sizeof(struct tcphdr))
3099 return (0);
3100 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
3101 return (0);
3102 opt = hdr + sizeof(struct tcphdr);
3103 hlen -= sizeof(struct tcphdr);
3104 while (hlen >= TCPOLEN_MAXSEG) {
3105 switch (*opt) {
3106 case TCPOPT_EOL:
3107 case TCPOPT_NOP:
3108 ++opt;
3109 --hlen;
3110 break;
3111 case TCPOPT_MAXSEG:
3112 bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2);
3113 NTOHS(mss);
3114 /* FALLTHROUGH */
3115 default:
3116 optlen = opt[1];
3117 if (optlen < 2)
3118 optlen = 2;
3119 hlen -= optlen;
3120 opt += optlen;
3121 break;
3122 }
3123 }
3124 return (mss);
3125}
3126
3127u_int16_t
3128pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer)
3129{
3130#ifdef INET
3131 struct sockaddr_in *dst;
3132 struct route ro;
3133#endif /* INET */
3134#ifdef INET6
3135 struct sockaddr_in6 *dst6;
3136 struct route_in6 ro6;
3137#endif /* INET6 */
3138 struct rtentry *rt = NULL;
3139 int hlen = 0; /* make the compiler happy */
3140 u_int16_t mss = V_tcp_mssdflt;
3141
3142 switch (af) {
3143#ifdef INET
3144 case AF_INET:
3145 hlen = sizeof(struct ip);
3146 bzero(&ro, sizeof(ro));
3147 dst = (struct sockaddr_in *)&ro.ro_dst;
3148 dst->sin_family = AF_INET;
3149 dst->sin_len = sizeof(*dst);
3150 dst->sin_addr = addr->v4;
3151#ifdef __FreeBSD__
3152#ifdef RTF_PRCLONING
3153 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
3154#else /* !RTF_PRCLONING */
3155 in_rtalloc_ign(&ro, 0, 0);
3156#endif
3157#else /* ! __FreeBSD__ */
3158 rtalloc_noclone(&ro, NO_CLONING);
3159#endif
3160 rt = ro.ro_rt;
3161 break;
3162#endif /* INET */
3163#ifdef INET6
3164 case AF_INET6:
3165 hlen = sizeof(struct ip6_hdr);
3166 bzero(&ro6, sizeof(ro6));
3167 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst;
3168 dst6->sin6_family = AF_INET6;
3169 dst6->sin6_len = sizeof(*dst6);
3170 dst6->sin6_addr = addr->v6;
3171#ifdef __FreeBSD__
3172#ifdef RTF_PRCLONING
3173 rtalloc_ign((struct route *)&ro6,
3174 (RTF_CLONING | RTF_PRCLONING));
3175#else /* !RTF_PRCLONING */
3176 rtalloc_ign((struct route *)&ro6, 0);
3177#endif
3178#else /* ! __FreeBSD__ */
3179 rtalloc_noclone((struct route *)&ro6, NO_CLONING);
3180#endif
3181 rt = ro6.ro_rt;
3182 break;
3183#endif /* INET6 */
3184 }
3185
3186 if (rt && rt->rt_ifp) {
3187 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr);
3188 mss = max(V_tcp_mssdflt, mss);
3189 RTFREE(rt);
3190 }
3191 mss = min(mss, offer);
3192 mss = max(mss, 64); /* sanity - at least max opt space */
3193 return (mss);
3194}
3195
3196void
3197pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr)
3198{
3199 struct pf_rule *r = s->rule.ptr;
3200
3201 s->rt_kif = NULL;
3202 if (!r->rt || r->rt == PF_FASTROUTE)
3203 return;
3204 switch (s->af) {
3205#ifdef INET
3206 case AF_INET:
3207 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL,
3208 &s->nat_src_node);
3209 s->rt_kif = r->rpool.cur->kif;
3210 break;
3211#endif /* INET */
3212#ifdef INET6
3213 case AF_INET6:
3214 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL,
3215 &s->nat_src_node);
3216 s->rt_kif = r->rpool.cur->kif;
3217 break;
3218#endif /* INET6 */
3219 }
3220}
3221
3222int
3223pf_test_tcp(struct pf_rule **rm, struct pf_state **sm, int direction,
3224 struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3225#ifdef __FreeBSD__
3226 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
3227 struct ifqueue *ifq, struct inpcb *inp)
3228#else
3229 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
3230 struct ifqueue *ifq)
3231#endif
3232{
3233 struct pf_rule *nr = NULL;
3234 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
3235 struct tcphdr *th = pd->hdr.tcp;
3236 u_int16_t bport, nport = 0;
3237 sa_family_t af = pd->af;
3238 struct pf_rule *r, *a = NULL;
3239 struct pf_ruleset *ruleset = NULL;
3240 struct pf_src_node *nsn = NULL;
3241 u_short reason;
3242 int rewrite = 0;
3243 int tag = -1, rtableid = -1;
3244 u_int16_t mss = V_tcp_mssdflt;
3245 int asd = 0;
3246 int match = 0;
3247
3248 if (pf_check_congestion(ifq)) {
3249 REASON_SET(&reason, PFRES_CONGEST);
3250 return (PF_DROP);
3251 }
3252
3253#ifdef __FreeBSD__
3254 if (inp != NULL)
3255 pd->lookup.done = pf_socket_lookup(direction, pd, inp);
3256 else if (debug_pfugidhack) {
3257 PF_UNLOCK();
3258 DPFPRINTF(PF_DEBUG_MISC, ("pf: unlocked lookup\n"));
3259 pd->lookup.done = pf_socket_lookup(direction, pd, inp);
3260 PF_LOCK();
3261 }
3262#endif
3263
3264 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3265
3266 if (direction == PF_OUT) {
3267 bport = nport = th->th_sport;
3268 /* check outgoing packet for BINAT/NAT */
3269 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3270 saddr, th->th_sport, daddr, th->th_dport,
3271 &pd->naddr, &nport)) != NULL) {
3272 PF_ACPY(&pd->baddr, saddr, af);
3273 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
3274 &th->th_sum, &pd->naddr, nport, 0, af);
3275 rewrite++;
3276 if (nr->natpass)
3277 r = NULL;
3278 pd->nat_rule = nr;
3279 }
3280 } else {
3281 bport = nport = th->th_dport;
3282 /* check incoming packet for BINAT/RDR */
3283 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3284 saddr, th->th_sport, daddr, th->th_dport,
3285 &pd->naddr, &nport)) != NULL) {
3286 PF_ACPY(&pd->baddr, daddr, af);
3287 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
3288 &th->th_sum, &pd->naddr, nport, 0, af);
3289 rewrite++;
3290 if (nr->natpass)
3291 r = NULL;
3292 pd->nat_rule = nr;
3293 }
3294 }
3295
3296 while (r != NULL) {
3297 r->evaluations++;
3298 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3299 r = r->skip[PF_SKIP_IFP].ptr;
3300 else if (r->direction && r->direction != direction)
3301 r = r->skip[PF_SKIP_DIR].ptr;
3302 else if (r->af && r->af != af)
3303 r = r->skip[PF_SKIP_AF].ptr;
3304 else if (r->proto && r->proto != IPPROTO_TCP)
3305 r = r->skip[PF_SKIP_PROTO].ptr;
3306 else if (PF_MISMATCHAW(&r->src.addr, saddr, af,
3307 r->src.neg, kif))
3308 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3309 else if (r->src.port_op && !pf_match_port(r->src.port_op,
3310 r->src.port[0], r->src.port[1], th->th_sport))
3311 r = r->skip[PF_SKIP_SRC_PORT].ptr;
3312 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af,
3313 r->dst.neg, NULL))
3314 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3315 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
3316 r->dst.port[0], r->dst.port[1], th->th_dport))
3317 r = r->skip[PF_SKIP_DST_PORT].ptr;
3318 else if (r->tos && !(r->tos == pd->tos))
3319 r = TAILQ_NEXT(r, entries);
3320 else if (r->rule_flag & PFRULE_FRAGMENT)
3321 r = TAILQ_NEXT(r, entries);
3322 else if ((r->flagset & th->th_flags) != r->flags)
3323 r = TAILQ_NEXT(r, entries);
3324 else if (r->uid.op && (pd->lookup.done || (pd->lookup.done =
3325#ifdef __FreeBSD__
3326 pf_socket_lookup(direction, pd, inp), 1)) &&
3327#else
3328 pf_socket_lookup(direction, pd), 1)) &&
3329#endif
3330 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
3331 pd->lookup.uid))
3332 r = TAILQ_NEXT(r, entries);
3333 else if (r->gid.op && (pd->lookup.done || (pd->lookup.done =
3334#ifdef __FreeBSD__
3335 pf_socket_lookup(direction, pd, inp), 1)) &&
3336#else
3337 pf_socket_lookup(direction, pd), 1)) &&
3338#endif
3339 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
3340 pd->lookup.gid))
3341 r = TAILQ_NEXT(r, entries);
3342 else if (r->prob && r->prob <= arc4random())
3343 r = TAILQ_NEXT(r, entries);
3344 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag))
3345 r = TAILQ_NEXT(r, entries);
3346 else if (r->os_fingerprint != PF_OSFP_ANY && !pf_osfp_match(
3347 pf_osfp_fingerprint(pd, m, off, th), r->os_fingerprint))
3348 r = TAILQ_NEXT(r, entries);
3349 else {
3350 if (r->tag)
3351 tag = r->tag;
3352 if (r->rtableid >= 0)
3353 rtableid = r->rtableid;
3354 if (r->anchor == NULL) {
3355 match = 1;
3356 *rm = r;
3357 *am = a;
3358 *rsm = ruleset;
3359 if ((*rm)->quick)
3360 break;
3361 r = TAILQ_NEXT(r, entries);
3362 } else
3363 pf_step_into_anchor(&asd, &ruleset,
3364 PF_RULESET_FILTER, &r, &a, &match);
3365 }
3366 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3367 PF_RULESET_FILTER, &r, &a, &match))
3368 break;
3369 }
3370 r = *rm;
3371 a = *am;
3372 ruleset = *rsm;
3373
3374 REASON_SET(&reason, PFRES_MATCH);
3375
3376 if (r->log || (nr != NULL && nr->natpass && nr->log)) {
3377 if (rewrite)
3378#ifdef __FreeBSD__
3379 m_copyback(m, off, sizeof(*th), (caddr_t)th);
3380#else
3381 m_copyback(m, off, sizeof(*th), th);
3382#endif
3383 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr,
3384 a, ruleset, pd);
3385 }
3386
3387 if ((r->action == PF_DROP) &&
3388 ((r->rule_flag & PFRULE_RETURNRST) ||
3389 (r->rule_flag & PFRULE_RETURNICMP) ||
3390 (r->rule_flag & PFRULE_RETURN))) {
3391 /* undo NAT changes, if they have taken place */
3392 if (nr != NULL) {
3393 if (direction == PF_OUT) {
3394 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
3395 &th->th_sum, &pd->baddr, bport, 0, af);
3396 rewrite++;
3397 } else {
3398 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
3399 &th->th_sum, &pd->baddr, bport, 0, af);
3400 rewrite++;
3401 }
3402 }
3403 if (((r->rule_flag & PFRULE_RETURNRST) ||
3404 (r->rule_flag & PFRULE_RETURN)) &&
3405 !(th->th_flags & TH_RST)) {
3406 u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
3407
3408 if (th->th_flags & TH_SYN)
3409 ack++;
3410 if (th->th_flags & TH_FIN)
3411 ack++;
3412#ifdef __FreeBSD__
3413 pf_send_tcp(m, r, af, pd->dst,
3414#else
3415 pf_send_tcp(r, af, pd->dst,
3416#endif
3417 pd->src, th->th_dport, th->th_sport,
3418 ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0,
3419 r->return_ttl, 1, 0, pd->eh, kif->pfik_ifp);
3420 } else if ((af == AF_INET) && r->return_icmp)
3421 pf_send_icmp(m, r->return_icmp >> 8,
3422 r->return_icmp & 255, af, r);
3423 else if ((af == AF_INET6) && r->return_icmp6)
3424 pf_send_icmp(m, r->return_icmp6 >> 8,
3425 r->return_icmp6 & 255, af, r);
3426 }
3427
3428 if (r->action == PF_DROP)
3429 return (PF_DROP);
3430
3431 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) {
3432 REASON_SET(&reason, PFRES_MEMORY);
3433 return (PF_DROP);
3434 }
3435
3436 if (r->keep_state || nr != NULL ||
3437 (pd->flags & PFDESC_TCP_NORM)) {
3438 /* create new state */
3439 u_int16_t len;
3440 struct pf_state *s = NULL;
3441 struct pf_src_node *sn = NULL;
3442
3443 len = pd->tot_len - off - (th->th_off << 2);
3444
3445 /* check maximums */
3446 if (r->max_states && (r->states >= r->max_states)) {
3447 pf_status.lcounters[LCNT_STATES]++;
3448 REASON_SET(&reason, PFRES_MAXSTATES);
3449 goto cleanup;
3450 }
3451 /* src node for filter rule */
3452 if ((r->rule_flag & PFRULE_SRCTRACK ||
3453 r->rpool.opts & PF_POOL_STICKYADDR) &&
3454 pf_insert_src_node(&sn, r, saddr, af) != 0) {
3455 REASON_SET(&reason, PFRES_SRCLIMIT);
3456 goto cleanup;
3457 }
3458 /* src node for translation rule */
3459 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3460 ((direction == PF_OUT &&
3461 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3462 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) {
3463 REASON_SET(&reason, PFRES_SRCLIMIT);
3464 goto cleanup;
3465 }
3466 s = pool_get(&pf_state_pl, PR_NOWAIT);
3467 if (s == NULL) {
3468 REASON_SET(&reason, PFRES_MEMORY);
3469cleanup:
3470 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3471 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3472 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3473 pf_status.src_nodes--;
3474 pool_put(&pf_src_tree_pl, sn);
3475 }
3476 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3477 nsn->expire == 0) {
3478 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3479 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3480 pf_status.src_nodes--;
3481 pool_put(&pf_src_tree_pl, nsn);
3482 }
3483 return (PF_DROP);
3484 }
3485 bzero(s, sizeof(*s));
3486 s->rule.ptr = r;
3487 s->nat_rule.ptr = nr;
3488 s->anchor.ptr = a;
3489 STATE_INC_COUNTERS(s);
3490 s->allow_opts = r->allow_opts;
3491 s->log = r->log & PF_LOG_ALL;
3492 if (nr != NULL)
3493 s->log |= nr->log & PF_LOG_ALL;
3494 s->proto = IPPROTO_TCP;
3495 s->direction = direction;
3496 s->af = af;
3497 if (direction == PF_OUT) {
3498 PF_ACPY(&s->gwy.addr, saddr, af);
3499 s->gwy.port = th->th_sport; /* sport */
3500 PF_ACPY(&s->ext.addr, daddr, af);
3501 s->ext.port = th->th_dport;
3502 if (nr != NULL) {
3503 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3504 s->lan.port = bport;
3505 } else {
3506 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3507 s->lan.port = s->gwy.port;
3508 }
3509 } else {
3510 PF_ACPY(&s->lan.addr, daddr, af);
3511 s->lan.port = th->th_dport;
3512 PF_ACPY(&s->ext.addr, saddr, af);
3513 s->ext.port = th->th_sport;
3514 if (nr != NULL) {
3515 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3516 s->gwy.port = bport;
3517 } else {
3518 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3519 s->gwy.port = s->lan.port;
3520 }
3521 }
3522
3523 s->src.seqlo = ntohl(th->th_seq);
3524 s->src.seqhi = s->src.seqlo + len + 1;
3525 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
3526 r->keep_state == PF_STATE_MODULATE) {
3527 /* Generate sequence number modulator */
3528#ifdef __FreeBSD__
3529 while ((s->src.seqdiff =
3530 pf_new_isn(s) - s->src.seqlo) == 0)
3531 ;
3532#else
3533 while ((s->src.seqdiff =
3534 tcp_rndiss_next() - s->src.seqlo) == 0)
3535 ;
3536#endif
3537 pf_change_a(&th->th_seq, &th->th_sum,
3538 htonl(s->src.seqlo + s->src.seqdiff), 0);
3539 rewrite = 1;
3540 } else
3541 s->src.seqdiff = 0;
3542 if (th->th_flags & TH_SYN) {
3543 s->src.seqhi++;
3544 s->src.wscale = pf_get_wscale(m, off, th->th_off, af);
3545 }
3546 s->src.max_win = MAX(ntohs(th->th_win), 1);
3547 if (s->src.wscale & PF_WSCALE_MASK) {
3548 /* Remove scale factor from initial window */
3549 int win = s->src.max_win;
3550 win += 1 << (s->src.wscale & PF_WSCALE_MASK);
3551 s->src.max_win = (win - 1) >>
3552 (s->src.wscale & PF_WSCALE_MASK);
3553 }
3554 if (th->th_flags & TH_FIN)
3555 s->src.seqhi++;
3556 s->dst.seqhi = 1;
3557 s->dst.max_win = 1;
3558 s->src.state = TCPS_SYN_SENT;
3559 s->dst.state = TCPS_CLOSED;
3560 s->creation = time_second;
3561 s->expire = time_second;
3562 s->timeout = PFTM_TCP_FIRST_PACKET;
3563 pf_set_rt_ifp(s, saddr);
3564 if (sn != NULL) {
3565 s->src_node = sn;
3566 s->src_node->states++;
3567 }
3568 if (nsn != NULL) {
3569 PF_ACPY(&nsn->raddr, &pd->naddr, af);
3570 s->nat_src_node = nsn;
3571 s->nat_src_node->states++;
3572 }
3573 if ((pd->flags & PFDESC_TCP_NORM) && pf_normalize_tcp_init(m,
3574 off, pd, th, &s->src, &s->dst)) {
3575 REASON_SET(&reason, PFRES_MEMORY);
3576 pf_src_tree_remove_state(s);
3577 STATE_DEC_COUNTERS(s);
3578 pool_put(&pf_state_pl, s);
3579 return (PF_DROP);
3580 }
3581 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
3582 pf_normalize_tcp_stateful(m, off, pd, &reason, th, s,
3583 &s->src, &s->dst, &rewrite)) {
3584 /* This really shouldn't happen!!! */
3585 DPFPRINTF(PF_DEBUG_URGENT,
3586 ("pf_normalize_tcp_stateful failed on first pkt"));
3587 pf_normalize_tcp_cleanup(s);
3588 pf_src_tree_remove_state(s);
3589 STATE_DEC_COUNTERS(s);
3590 pool_put(&pf_state_pl, s);
3591 return (PF_DROP);
3592 }
3593 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3594 pf_normalize_tcp_cleanup(s);
3595 REASON_SET(&reason, PFRES_STATEINS);
3596 pf_src_tree_remove_state(s);
3597 STATE_DEC_COUNTERS(s);
3598 pool_put(&pf_state_pl, s);
3599 return (PF_DROP);
3600 } else
3601 *sm = s;
3602 if (tag > 0) {
3603 pf_tag_ref(tag);
3604 s->tag = tag;
3605 }
3606 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
3607 r->keep_state == PF_STATE_SYNPROXY) {
3608 s->src.state = PF_TCPS_PROXY_SRC;
3609 if (nr != NULL) {
3610 if (direction == PF_OUT) {
3611 pf_change_ap(saddr, &th->th_sport,
3612 pd->ip_sum, &th->th_sum, &pd->baddr,
3613 bport, 0, af);
3614 } else {
3615 pf_change_ap(daddr, &th->th_dport,
3616 pd->ip_sum, &th->th_sum, &pd->baddr,
3617 bport, 0, af);
3618 }
3619 }
3620 s->src.seqhi = htonl(arc4random());
3621 /* Find mss option */
3622 mss = pf_get_mss(m, off, th->th_off, af);
3623 mss = pf_calc_mss(saddr, af, mss);
3624 mss = pf_calc_mss(daddr, af, mss);
3625 s->src.mss = mss;
3626#ifdef __FreeBSD__
3627 pf_send_tcp(NULL, r, af, daddr, saddr, th->th_dport,
3628#else
3629 pf_send_tcp(r, af, daddr, saddr, th->th_dport,
3630#endif
3631 th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1,
3632 TH_SYN|TH_ACK, 0, s->src.mss, 0, 1, 0, NULL, NULL);
3633 REASON_SET(&reason, PFRES_SYNPROXY);
3634 return (PF_SYNPROXY_DROP);
3635 }
3636 }
3637
3638 /* copy back packet headers if we performed NAT operations */
3639 if (rewrite)
3640 m_copyback(m, off, sizeof(*th), (caddr_t)th);
3641
3642 return (PF_PASS);
3643}
3644
3645int
3646pf_test_udp(struct pf_rule **rm, struct pf_state **sm, int direction,
3647 struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3648#ifdef __FreeBSD__
3649 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
3650 struct ifqueue *ifq, struct inpcb *inp)
3651#else
3652 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
3653 struct ifqueue *ifq)
3654#endif
3655{
3656 struct pf_rule *nr = NULL;
3657 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
3658 struct udphdr *uh = pd->hdr.udp;
3659 u_int16_t bport, nport = 0;
3660 sa_family_t af = pd->af;
3661 struct pf_rule *r, *a = NULL;
3662 struct pf_ruleset *ruleset = NULL;
3663 struct pf_src_node *nsn = NULL;
3664 u_short reason;
3665 int rewrite = 0;
3666 int tag = -1, rtableid = -1;
3667 int asd = 0;
3668 int match = 0;
3669
3670 if (pf_check_congestion(ifq)) {
3671 REASON_SET(&reason, PFRES_CONGEST);
3672 return (PF_DROP);
3673 }
3674
3675#ifdef __FreeBSD__
3676 if (inp != NULL)
3677 pd->lookup.done = pf_socket_lookup(direction, pd, inp);
3678 else if (debug_pfugidhack) {
3679 PF_UNLOCK();
3680 DPFPRINTF(PF_DEBUG_MISC, ("pf: unlocked lookup\n"));
3681 pd->lookup.done = pf_socket_lookup(direction, pd, inp);
3682 PF_LOCK();
3683 }
3684#endif
3685
3686 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3687
3688 if (direction == PF_OUT) {
3689 bport = nport = uh->uh_sport;
3690 /* check outgoing packet for BINAT/NAT */
3691 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3692 saddr, uh->uh_sport, daddr, uh->uh_dport,
3693 &pd->naddr, &nport)) != NULL) {
3694 PF_ACPY(&pd->baddr, saddr, af);
3695 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum,
3696 &uh->uh_sum, &pd->naddr, nport, 1, af);
3697 rewrite++;
3698 if (nr->natpass)
3699 r = NULL;
3700 pd->nat_rule = nr;
3701 }
3702 } else {
3703 bport = nport = uh->uh_dport;
3704 /* check incoming packet for BINAT/RDR */
3705 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3706 saddr, uh->uh_sport, daddr, uh->uh_dport, &pd->naddr,
3707 &nport)) != NULL) {
3708 PF_ACPY(&pd->baddr, daddr, af);
3709 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum,
3710 &uh->uh_sum, &pd->naddr, nport, 1, af);
3711 rewrite++;
3712 if (nr->natpass)
3713 r = NULL;
3714 pd->nat_rule = nr;
3715 }
3716 }
3717
3718 while (r != NULL) {
3719 r->evaluations++;
3720 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3721 r = r->skip[PF_SKIP_IFP].ptr;
3722 else if (r->direction && r->direction != direction)
3723 r = r->skip[PF_SKIP_DIR].ptr;
3724 else if (r->af && r->af != af)
3725 r = r->skip[PF_SKIP_AF].ptr;
3726 else if (r->proto && r->proto != IPPROTO_UDP)
3727 r = r->skip[PF_SKIP_PROTO].ptr;
3728 else if (PF_MISMATCHAW(&r->src.addr, saddr, af,
3729 r->src.neg, kif))
3730 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3731 else if (r->src.port_op && !pf_match_port(r->src.port_op,
3732 r->src.port[0], r->src.port[1], uh->uh_sport))
3733 r = r->skip[PF_SKIP_SRC_PORT].ptr;
3734 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af,
3735 r->dst.neg, NULL))
3736 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3737 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
3738 r->dst.port[0], r->dst.port[1], uh->uh_dport))
3739 r = r->skip[PF_SKIP_DST_PORT].ptr;
3740 else if (r->tos && !(r->tos == pd->tos))
3741 r = TAILQ_NEXT(r, entries);
3742 else if (r->rule_flag & PFRULE_FRAGMENT)
3743 r = TAILQ_NEXT(r, entries);
3744 else if (r->uid.op && (pd->lookup.done || (pd->lookup.done =
3745#ifdef __FreeBSD__
3746 pf_socket_lookup(direction, pd, inp), 1)) &&
3747#else
3748 pf_socket_lookup(direction, pd), 1)) &&
3749#endif
3750 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
3751 pd->lookup.uid))
3752 r = TAILQ_NEXT(r, entries);
3753 else if (r->gid.op && (pd->lookup.done || (pd->lookup.done =
3754#ifdef __FreeBSD__
3755 pf_socket_lookup(direction, pd, inp), 1)) &&
3756#else
3757 pf_socket_lookup(direction, pd), 1)) &&
3758#endif
3759 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
3760 pd->lookup.gid))
3761 r = TAILQ_NEXT(r, entries);
3762 else if (r->prob && r->prob <= arc4random())
3763 r = TAILQ_NEXT(r, entries);
3764 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag))
3765 r = TAILQ_NEXT(r, entries);
3766 else if (r->os_fingerprint != PF_OSFP_ANY)
3767 r = TAILQ_NEXT(r, entries);
3768 else {
3769 if (r->tag)
3770 tag = r->tag;
3771 if (r->rtableid >= 0)
3772 rtableid = r->rtableid;
3773 if (r->anchor == NULL) {
3774 match = 1;
3775 *rm = r;
3776 *am = a;
3777 *rsm = ruleset;
3778 if ((*rm)->quick)
3779 break;
3780 r = TAILQ_NEXT(r, entries);
3781 } else
3782 pf_step_into_anchor(&asd, &ruleset,
3783 PF_RULESET_FILTER, &r, &a, &match);
3784 }
3785 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3786 PF_RULESET_FILTER, &r, &a, &match))
3787 break;
3788 }
3789 r = *rm;
3790 a = *am;
3791 ruleset = *rsm;
3792
3793 REASON_SET(&reason, PFRES_MATCH);
3794
3795 if (r->log || (nr != NULL && nr->natpass && nr->log)) {
3796 if (rewrite)
3797#ifdef __FreeBSD__
3798 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
3799#else
3800 m_copyback(m, off, sizeof(*uh), uh);
3801#endif
3802 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr,
3803 a, ruleset, pd);
3804 }
3805
3806 if ((r->action == PF_DROP) &&
3807 ((r->rule_flag & PFRULE_RETURNICMP) ||
3808 (r->rule_flag & PFRULE_RETURN))) {
3809 /* undo NAT changes, if they have taken place */
3810 if (nr != NULL) {
3811 if (direction == PF_OUT) {
3812 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum,
3813 &uh->uh_sum, &pd->baddr, bport, 1, af);
3814 rewrite++;
3815 } else {
3816 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum,
3817 &uh->uh_sum, &pd->baddr, bport, 1, af);
3818 rewrite++;
3819 }
3820 }
3821 if ((af == AF_INET) && r->return_icmp)
3822 pf_send_icmp(m, r->return_icmp >> 8,
3823 r->return_icmp & 255, af, r);
3824 else if ((af == AF_INET6) && r->return_icmp6)
3825 pf_send_icmp(m, r->return_icmp6 >> 8,
3826 r->return_icmp6 & 255, af, r);
3827 }
3828
3829 if (r->action == PF_DROP)
3830 return (PF_DROP);
3831
3832 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) {
3833 REASON_SET(&reason, PFRES_MEMORY);
3834 return (PF_DROP);
3835 }
3836
3837 if (r->keep_state || nr != NULL) {
3838 /* create new state */
3839 struct pf_state *s = NULL;
3840 struct pf_src_node *sn = NULL;
3841
3842 /* check maximums */
3843 if (r->max_states && (r->states >= r->max_states)) {
3844 pf_status.lcounters[LCNT_STATES]++;
3845 REASON_SET(&reason, PFRES_MAXSTATES);
3846 goto cleanup;
3847 }
3848 /* src node for filter rule */
3849 if ((r->rule_flag & PFRULE_SRCTRACK ||
3850 r->rpool.opts & PF_POOL_STICKYADDR) &&
3851 pf_insert_src_node(&sn, r, saddr, af) != 0) {
3852 REASON_SET(&reason, PFRES_SRCLIMIT);
3853 goto cleanup;
3854 }
3855 /* src node for translation rule */
3856 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3857 ((direction == PF_OUT &&
3858 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3859 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) {
3860 REASON_SET(&reason, PFRES_SRCLIMIT);
3861 goto cleanup;
3862 }
3863 s = pool_get(&pf_state_pl, PR_NOWAIT);
3864 if (s == NULL) {
3865 REASON_SET(&reason, PFRES_MEMORY);
3866cleanup:
3867 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3868 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3869 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3870 pf_status.src_nodes--;
3871 pool_put(&pf_src_tree_pl, sn);
3872 }
3873 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3874 nsn->expire == 0) {
3875 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3876 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3877 pf_status.src_nodes--;
3878 pool_put(&pf_src_tree_pl, nsn);
3879 }
3880 return (PF_DROP);
3881 }
3882 bzero(s, sizeof(*s));
3883 s->rule.ptr = r;
3884 s->nat_rule.ptr = nr;
3885 s->anchor.ptr = a;
3886 STATE_INC_COUNTERS(s);
3887 s->allow_opts = r->allow_opts;
3888 s->log = r->log & PF_LOG_ALL;
3889 if (nr != NULL)
3890 s->log |= nr->log & PF_LOG_ALL;
3891 s->proto = IPPROTO_UDP;
3892 s->direction = direction;
3893 s->af = af;
3894 if (direction == PF_OUT) {
3895 PF_ACPY(&s->gwy.addr, saddr, af);
3896 s->gwy.port = uh->uh_sport;
3897 PF_ACPY(&s->ext.addr, daddr, af);
3898 s->ext.port = uh->uh_dport;
3899 if (nr != NULL) {
3900 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3901 s->lan.port = bport;
3902 } else {
3903 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3904 s->lan.port = s->gwy.port;
3905 }
3906 } else {
3907 PF_ACPY(&s->lan.addr, daddr, af);
3908 s->lan.port = uh->uh_dport;
3909 PF_ACPY(&s->ext.addr, saddr, af);
3910 s->ext.port = uh->uh_sport;
3911 if (nr != NULL) {
3912 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3913 s->gwy.port = bport;
3914 } else {
3915 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3916 s->gwy.port = s->lan.port;
3917 }
3918 }
3919 s->src.state = PFUDPS_SINGLE;
3920 s->dst.state = PFUDPS_NO_TRAFFIC;
3921 s->creation = time_second;
3922 s->expire = time_second;
3923 s->timeout = PFTM_UDP_FIRST_PACKET;
3924 pf_set_rt_ifp(s, saddr);
3925 if (sn != NULL) {
3926 s->src_node = sn;
3927 s->src_node->states++;
3928 }
3929 if (nsn != NULL) {
3930 PF_ACPY(&nsn->raddr, &pd->naddr, af);
3931 s->nat_src_node = nsn;
3932 s->nat_src_node->states++;
3933 }
3934 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3935 REASON_SET(&reason, PFRES_STATEINS);
3936 pf_src_tree_remove_state(s);
3937 STATE_DEC_COUNTERS(s);
3938 pool_put(&pf_state_pl, s);
3939 return (PF_DROP);
3940 } else
3941 *sm = s;
3942 if (tag > 0) {
3943 pf_tag_ref(tag);
3944 s->tag = tag;
3945 }
3946 }
3947
3948 /* copy back packet headers if we performed NAT operations */
3949 if (rewrite)
3950 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
3951
3952 return (PF_PASS);
3953}
3954
3955int
3956pf_test_icmp(struct pf_rule **rm, struct pf_state **sm, int direction,
3957 struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3958 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
3959 struct ifqueue *ifq)
3960{
3961 struct pf_rule *nr = NULL;
3962 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
3963 struct pf_rule *r, *a = NULL;
3964 struct pf_ruleset *ruleset = NULL;
3965 struct pf_src_node *nsn = NULL;
3966 u_short reason;
3967 u_int16_t icmpid = 0, bport, nport = 0;
3968 sa_family_t af = pd->af;
3969 u_int8_t icmptype = 0; /* make the compiler happy */
3970 u_int8_t icmpcode = 0; /* make the compiler happy */
3971 int state_icmp = 0;
3972 int tag = -1, rtableid = -1;
3973#ifdef INET6
3974 int rewrite = 0;
3975#endif /* INET6 */
3976 int asd = 0;
3977 int match = 0;
3978
3979 if (pf_check_congestion(ifq)) {
3980 REASON_SET(&reason, PFRES_CONGEST);
3981 return (PF_DROP);
3982 }
3983
3984 switch (pd->proto) {
3985#ifdef INET
3986 case IPPROTO_ICMP:
3987 icmptype = pd->hdr.icmp->icmp_type;
3988 icmpcode = pd->hdr.icmp->icmp_code;
3989 icmpid = pd->hdr.icmp->icmp_id;
3990
3991 if (icmptype == ICMP_UNREACH ||
3992 icmptype == ICMP_SOURCEQUENCH ||
3993 icmptype == ICMP_REDIRECT ||
3994 icmptype == ICMP_TIMXCEED ||
3995 icmptype == ICMP_PARAMPROB)
3996 state_icmp++;
3997 break;
3998#endif /* INET */
3999#ifdef INET6
4000 case IPPROTO_ICMPV6:
4001 icmptype = pd->hdr.icmp6->icmp6_type;
4002 icmpcode = pd->hdr.icmp6->icmp6_code;
4003 icmpid = pd->hdr.icmp6->icmp6_id;
4004
4005 if (icmptype == ICMP6_DST_UNREACH ||
4006 icmptype == ICMP6_PACKET_TOO_BIG ||
4007 icmptype == ICMP6_TIME_EXCEEDED ||
4008 icmptype == ICMP6_PARAM_PROB)
4009 state_icmp++;
4010 break;
4011#endif /* INET6 */
4012 }
4013
4014 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
4015
4016 if (direction == PF_OUT) {
4017 bport = nport = icmpid;
4018 /* check outgoing packet for BINAT/NAT */
4019 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
4020 saddr, icmpid, daddr, icmpid, &pd->naddr, &nport)) !=
4021 NULL) {
4022 PF_ACPY(&pd->baddr, saddr, af);
4023 switch (af) {
4024#ifdef INET
4025 case AF_INET:
4026 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
4027 pd->naddr.v4.s_addr, 0);
4028 pd->hdr.icmp->icmp_cksum = pf_cksum_fixup(
4029 pd->hdr.icmp->icmp_cksum, icmpid, nport, 0);
4030 pd->hdr.icmp->icmp_id = nport;
4031 m_copyback(m, off, ICMP_MINLEN,
4032 (caddr_t)pd->hdr.icmp);
4033 break;
4034#endif /* INET */
4035#ifdef INET6
4036 case AF_INET6:
4037 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum,
4038 &pd->naddr, 0);
4039 rewrite++;
4040 break;
4041#endif /* INET6 */
4042 }
4043 if (nr->natpass)
4044 r = NULL;
4045 pd->nat_rule = nr;
4046 }
4047 } else {
4048 bport = nport = icmpid;
4049 /* check incoming packet for BINAT/RDR */
4050 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
4051 saddr, icmpid, daddr, icmpid, &pd->naddr, &nport)) !=
4052 NULL) {
4053 PF_ACPY(&pd->baddr, daddr, af);
4054 switch (af) {
4055#ifdef INET
4056 case AF_INET:
4057 pf_change_a(&daddr->v4.s_addr,
4058 pd->ip_sum, pd->naddr.v4.s_addr, 0);
4059 break;
4060#endif /* INET */
4061#ifdef INET6
4062 case AF_INET6:
4063 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum,
4064 &pd->naddr, 0);
4065 rewrite++;
4066 break;
4067#endif /* INET6 */
4068 }
4069 if (nr->natpass)
4070 r = NULL;
4071 pd->nat_rule = nr;
4072 }
4073 }
4074
4075 while (r != NULL) {
4076 r->evaluations++;
4077 if (pfi_kif_match(r->kif, kif) == r->ifnot)
4078 r = r->skip[PF_SKIP_IFP].ptr;
4079 else if (r->direction && r->direction != direction)
4080 r = r->skip[PF_SKIP_DIR].ptr;
4081 else if (r->af && r->af != af)
4082 r = r->skip[PF_SKIP_AF].ptr;
4083 else if (r->proto && r->proto != pd->proto)
4084 r = r->skip[PF_SKIP_PROTO].ptr;
4085 else if (PF_MISMATCHAW(&r->src.addr, saddr, af,
4086 r->src.neg, kif))
4087 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
4088 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af,
4089 r->dst.neg, NULL))
4090 r = r->skip[PF_SKIP_DST_ADDR].ptr;
4091 else if (r->type && r->type != icmptype + 1)
4092 r = TAILQ_NEXT(r, entries);
4093 else if (r->code && r->code != icmpcode + 1)
4094 r = TAILQ_NEXT(r, entries);
4095 else if (r->tos && !(r->tos == pd->tos))
4096 r = TAILQ_NEXT(r, entries);
4097 else if (r->rule_flag & PFRULE_FRAGMENT)
4098 r = TAILQ_NEXT(r, entries);
4099 else if (r->prob && r->prob <= arc4random())
4100 r = TAILQ_NEXT(r, entries);
4101 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag))
4102 r = TAILQ_NEXT(r, entries);
4103 else if (r->os_fingerprint != PF_OSFP_ANY)
4104 r = TAILQ_NEXT(r, entries);
4105 else {
4106 if (r->tag)
4107 tag = r->tag;
4108 if (r->rtableid >= 0)
4109 rtableid = r->rtableid;
4110 if (r->anchor == NULL) {
4111 match = 1;
4112 *rm = r;
4113 *am = a;
4114 *rsm = ruleset;
4115 if ((*rm)->quick)
4116 break;
4117 r = TAILQ_NEXT(r, entries);
4118 } else
4119 pf_step_into_anchor(&asd, &ruleset,
4120 PF_RULESET_FILTER, &r, &a, &match);
4121 }
4122 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
4123 PF_RULESET_FILTER, &r, &a, &match))
4124 break;
4125 }
4126 r = *rm;
4127 a = *am;
4128 ruleset = *rsm;
4129
4130 REASON_SET(&reason, PFRES_MATCH);
4131
4132 if (r->log || (nr != NULL && nr->natpass && nr->log)) {
4133#ifdef INET6
4134 if (rewrite)
4135 m_copyback(m, off, sizeof(struct icmp6_hdr),
4136 (caddr_t)pd->hdr.icmp6);
4137#endif /* INET6 */
4138 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr,
4139 a, ruleset, pd);
4140 }
4141
4142 if (r->action != PF_PASS)
4143 return (PF_DROP);
4144
4145 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) {
4146 REASON_SET(&reason, PFRES_MEMORY);
4147 return (PF_DROP);
4148 }
4149
4150 if (!state_icmp && (r->keep_state || nr != NULL)) {
4151 /* create new state */
4152 struct pf_state *s = NULL;
4153 struct pf_src_node *sn = NULL;
4154
4155 /* check maximums */
4156 if (r->max_states && (r->states >= r->max_states)) {
4157 pf_status.lcounters[LCNT_STATES]++;
4158 REASON_SET(&reason, PFRES_MAXSTATES);
4159 goto cleanup;
4160 }
4161 /* src node for filter rule */
4162 if ((r->rule_flag & PFRULE_SRCTRACK ||
4163 r->rpool.opts & PF_POOL_STICKYADDR) &&
4164 pf_insert_src_node(&sn, r, saddr, af) != 0) {
4165 REASON_SET(&reason, PFRES_SRCLIMIT);
4166 goto cleanup;
4167 }
4168 /* src node for translation rule */
4169 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
4170 ((direction == PF_OUT &&
4171 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
4172 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) {
4173 REASON_SET(&reason, PFRES_SRCLIMIT);
4174 goto cleanup;
4175 }
4176 s = pool_get(&pf_state_pl, PR_NOWAIT);
4177 if (s == NULL) {
4178 REASON_SET(&reason, PFRES_MEMORY);
4179cleanup:
4180 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
4181 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
4182 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
4183 pf_status.src_nodes--;
4184 pool_put(&pf_src_tree_pl, sn);
4185 }
4186 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
4187 nsn->expire == 0) {
4188 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
4189 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
4190 pf_status.src_nodes--;
4191 pool_put(&pf_src_tree_pl, nsn);
4192 }
4193 return (PF_DROP);
4194 }
4195 bzero(s, sizeof(*s));
4196 s->rule.ptr = r;
4197 s->nat_rule.ptr = nr;
4198 s->anchor.ptr = a;
4199 STATE_INC_COUNTERS(s);
4200 s->allow_opts = r->allow_opts;
4201 s->log = r->log & PF_LOG_ALL;
4202 if (nr != NULL)
4203 s->log |= nr->log & PF_LOG_ALL;
4204 s->proto = pd->proto;
4205 s->direction = direction;
4206 s->af = af;
4207 if (direction == PF_OUT) {
4208 PF_ACPY(&s->gwy.addr, saddr, af);
4209 s->gwy.port = nport;
4210 PF_ACPY(&s->ext.addr, daddr, af);
4211 s->ext.port = 0;
4212 if (nr != NULL) {
4213 PF_ACPY(&s->lan.addr, &pd->baddr, af);
4214 s->lan.port = bport;
4215 } else {
4216 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
4217 s->lan.port = s->gwy.port;
4218 }
4219 } else {
4220 PF_ACPY(&s->lan.addr, daddr, af);
4221 s->lan.port = nport;
4222 PF_ACPY(&s->ext.addr, saddr, af);
4223 s->ext.port = 0;
4224 if (nr != NULL) {
4225 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
4226 s->gwy.port = bport;
4227 } else {
4228 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
4229 s->gwy.port = s->lan.port;
4230 }
4231 }
4232 s->creation = time_second;
4233 s->expire = time_second;
4234 s->timeout = PFTM_ICMP_FIRST_PACKET;
4235 pf_set_rt_ifp(s, saddr);
4236 if (sn != NULL) {
4237 s->src_node = sn;
4238 s->src_node->states++;
4239 }
4240 if (nsn != NULL) {
4241 PF_ACPY(&nsn->raddr, &pd->naddr, af);
4242 s->nat_src_node = nsn;
4243 s->nat_src_node->states++;
4244 }
4245 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
4246 REASON_SET(&reason, PFRES_STATEINS);
4247 pf_src_tree_remove_state(s);
4248 STATE_DEC_COUNTERS(s);
4249 pool_put(&pf_state_pl, s);
4250 return (PF_DROP);
4251 } else
4252 *sm = s;
4253 if (tag > 0) {
4254 pf_tag_ref(tag);
4255 s->tag = tag;
4256 }
4257 }
4258
4259#ifdef INET6
4260 /* copy back packet headers if we performed IPv6 NAT operations */
4261 if (rewrite)
4262 m_copyback(m, off, sizeof(struct icmp6_hdr),
4263 (caddr_t)pd->hdr.icmp6);
4264#endif /* INET6 */
4265
4266 return (PF_PASS);
4267}
4268
4269int
4270pf_test_other(struct pf_rule **rm, struct pf_state **sm, int direction,
4271 struct pfi_kif *kif, struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
4272 struct pf_rule **am, struct pf_ruleset **rsm, struct ifqueue *ifq)
4273{
4274 struct pf_rule *nr = NULL;
4275 struct pf_rule *r, *a = NULL;
4276 struct pf_ruleset *ruleset = NULL;
4277 struct pf_src_node *nsn = NULL;
4278 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
4279 sa_family_t af = pd->af;
4280 u_short reason;
4281 int tag = -1, rtableid = -1;
4282 int asd = 0;
4283 int match = 0;
4284
4285 if (pf_check_congestion(ifq)) {
4286 REASON_SET(&reason, PFRES_CONGEST);
4287 return (PF_DROP);
4288 }
4289
4290 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
4291
4292 if (direction == PF_OUT) {
4293 /* check outgoing packet for BINAT/NAT */
4294 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
4295 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
4296 PF_ACPY(&pd->baddr, saddr, af);
4297 switch (af) {
4298#ifdef INET
4299 case AF_INET:
4300 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
4301 pd->naddr.v4.s_addr, 0);
4302 break;
4303#endif /* INET */
4304#ifdef INET6
4305 case AF_INET6:
4306 PF_ACPY(saddr, &pd->naddr, af);
4307 break;
4308#endif /* INET6 */
4309 }
4310 if (nr->natpass)
4311 r = NULL;
4312 pd->nat_rule = nr;
4313 }
4314 } else {
4315 /* check incoming packet for BINAT/RDR */
4316 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
4317 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
4318 PF_ACPY(&pd->baddr, daddr, af);
4319 switch (af) {
4320#ifdef INET
4321 case AF_INET:
4322 pf_change_a(&daddr->v4.s_addr,
4323 pd->ip_sum, pd->naddr.v4.s_addr, 0);
4324 break;
4325#endif /* INET */
4326#ifdef INET6
4327 case AF_INET6:
4328 PF_ACPY(daddr, &pd->naddr, af);
4329 break;
4330#endif /* INET6 */
4331 }
4332 if (nr->natpass)
4333 r = NULL;
4334 pd->nat_rule = nr;
4335 }
4336 }
4337
4338 while (r != NULL) {
4339 r->evaluations++;
4340 if (pfi_kif_match(r->kif, kif) == r->ifnot)
4341 r = r->skip[PF_SKIP_IFP].ptr;
4342 else if (r->direction && r->direction != direction)
4343 r = r->skip[PF_SKIP_DIR].ptr;
4344 else if (r->af && r->af != af)
4345 r = r->skip[PF_SKIP_AF].ptr;
4346 else if (r->proto && r->proto != pd->proto)
4347 r = r->skip[PF_SKIP_PROTO].ptr;
4348 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af,
4349 r->src.neg, kif))
4350 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
4351 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af,
4352 r->dst.neg, NULL))
4353 r = r->skip[PF_SKIP_DST_ADDR].ptr;
4354 else if (r->tos && !(r->tos == pd->tos))
4355 r = TAILQ_NEXT(r, entries);
4356 else if (r->rule_flag & PFRULE_FRAGMENT)
4357 r = TAILQ_NEXT(r, entries);
4358 else if (r->prob && r->prob <= arc4random())
4359 r = TAILQ_NEXT(r, entries);
4360 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag))
4361 r = TAILQ_NEXT(r, entries);
4362 else if (r->os_fingerprint != PF_OSFP_ANY)
4363 r = TAILQ_NEXT(r, entries);
4364 else {
4365 if (r->tag)
4366 tag = r->tag;
4367 if (r->rtableid >= 0)
4368 rtableid = r->rtableid;
4369 if (r->anchor == NULL) {
4370 match = 1;
4371 *rm = r;
4372 *am = a;
4373 *rsm = ruleset;
4374 if ((*rm)->quick)
4375 break;
4376 r = TAILQ_NEXT(r, entries);
4377 } else
4378 pf_step_into_anchor(&asd, &ruleset,
4379 PF_RULESET_FILTER, &r, &a, &match);
4380 }
4381 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
4382 PF_RULESET_FILTER, &r, &a, &match))
4383 break;
4384 }
4385 r = *rm;
4386 a = *am;
4387 ruleset = *rsm;
4388
4389 REASON_SET(&reason, PFRES_MATCH);
4390
4391 if (r->log || (nr != NULL && nr->natpass && nr->log))
4392 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr,
4393 a, ruleset, pd);
4394
4395 if ((r->action == PF_DROP) &&
4396 ((r->rule_flag & PFRULE_RETURNICMP) ||
4397 (r->rule_flag & PFRULE_RETURN))) {
4398 struct pf_addr *a = NULL;
4399
4400 if (nr != NULL) {
4401 if (direction == PF_OUT)
4402 a = saddr;
4403 else
4404 a = daddr;
4405 }
4406 if (a != NULL) {
4407 switch (af) {
4408#ifdef INET
4409 case AF_INET:
4410 pf_change_a(&a->v4.s_addr, pd->ip_sum,
4411 pd->baddr.v4.s_addr, 0);
4412 break;
4413#endif /* INET */
4414#ifdef INET6
4415 case AF_INET6:
4416 PF_ACPY(a, &pd->baddr, af);
4417 break;
4418#endif /* INET6 */
4419 }
4420 }
4421 if ((af == AF_INET) && r->return_icmp)
4422 pf_send_icmp(m, r->return_icmp >> 8,
4423 r->return_icmp & 255, af, r);
4424 else if ((af == AF_INET6) && r->return_icmp6)
4425 pf_send_icmp(m, r->return_icmp6 >> 8,
4426 r->return_icmp6 & 255, af, r);
4427 }
4428
4429 if (r->action != PF_PASS)
4430 return (PF_DROP);
4431
4432 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) {
4433 REASON_SET(&reason, PFRES_MEMORY);
4434 return (PF_DROP);
4435 }
4436
4437 if (r->keep_state || nr != NULL) {
4438 /* create new state */
4439 struct pf_state *s = NULL;
4440 struct pf_src_node *sn = NULL;
4441
4442 /* check maximums */
4443 if (r->max_states && (r->states >= r->max_states)) {
4444 pf_status.lcounters[LCNT_STATES]++;
4445 REASON_SET(&reason, PFRES_MAXSTATES);
4446 goto cleanup;
4447 }
4448 /* src node for filter rule */
4449 if ((r->rule_flag & PFRULE_SRCTRACK ||
4450 r->rpool.opts & PF_POOL_STICKYADDR) &&
4451 pf_insert_src_node(&sn, r, saddr, af) != 0) {
4452 REASON_SET(&reason, PFRES_SRCLIMIT);
4453 goto cleanup;
4454 }
4455 /* src node for translation rule */
4456 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
4457 ((direction == PF_OUT &&
4458 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
4459 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) {
4460 REASON_SET(&reason, PFRES_SRCLIMIT);
4461 goto cleanup;
4462 }
4463 s = pool_get(&pf_state_pl, PR_NOWAIT);
4464 if (s == NULL) {
4465 REASON_SET(&reason, PFRES_MEMORY);
4466cleanup:
4467 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
4468 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
4469 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
4470 pf_status.src_nodes--;
4471 pool_put(&pf_src_tree_pl, sn);
4472 }
4473 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
4474 nsn->expire == 0) {
4475 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
4476 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
4477 pf_status.src_nodes--;
4478 pool_put(&pf_src_tree_pl, nsn);
4479 }
4480 return (PF_DROP);
4481 }
4482 bzero(s, sizeof(*s));
4483 s->rule.ptr = r;
4484 s->nat_rule.ptr = nr;
4485 s->anchor.ptr = a;
4486 STATE_INC_COUNTERS(s);
4487 s->allow_opts = r->allow_opts;
4488 s->log = r->log & PF_LOG_ALL;
4489 if (nr != NULL)
4490 s->log |= nr->log & PF_LOG_ALL;
4491 s->proto = pd->proto;
4492 s->direction = direction;
4493 s->af = af;
4494 if (direction == PF_OUT) {
4495 PF_ACPY(&s->gwy.addr, saddr, af);
4496 PF_ACPY(&s->ext.addr, daddr, af);
4497 if (nr != NULL)
4498 PF_ACPY(&s->lan.addr, &pd->baddr, af);
4499 else
4500 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
4501 } else {
4502 PF_ACPY(&s->lan.addr, daddr, af);
4503 PF_ACPY(&s->ext.addr, saddr, af);
4504 if (nr != NULL)
4505 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
4506 else
4507 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
4508 }
4509 s->src.state = PFOTHERS_SINGLE;
4510 s->dst.state = PFOTHERS_NO_TRAFFIC;
4511 s->creation = time_second;
4512 s->expire = time_second;
4513 s->timeout = PFTM_OTHER_FIRST_PACKET;
4514 pf_set_rt_ifp(s, saddr);
4515 if (sn != NULL) {
4516 s->src_node = sn;
4517 s->src_node->states++;
4518 }
4519 if (nsn != NULL) {
4520 PF_ACPY(&nsn->raddr, &pd->naddr, af);
4521 s->nat_src_node = nsn;
4522 s->nat_src_node->states++;
4523 }
4524 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
4525 REASON_SET(&reason, PFRES_STATEINS);
4526 pf_src_tree_remove_state(s);
4527 STATE_DEC_COUNTERS(s);
4528 pool_put(&pf_state_pl, s);
4529 return (PF_DROP);
4530 } else
4531 *sm = s;
4532 if (tag > 0) {
4533 pf_tag_ref(tag);
4534 s->tag = tag;
4535 }
4536 }
4537
4538 return (PF_PASS);
4539}
4540
4541int
4542pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif,
4543 struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am,
4544 struct pf_ruleset **rsm)
4545{
4546 struct pf_rule *r, *a = NULL;
4547 struct pf_ruleset *ruleset = NULL;
4548 sa_family_t af = pd->af;
4549 u_short reason;
4550 int tag = -1;
4551 int asd = 0;
4552 int match = 0;
4553
4554 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
4555 while (r != NULL) {
4556 r->evaluations++;
4557 if (pfi_kif_match(r->kif, kif) == r->ifnot)
4558 r = r->skip[PF_SKIP_IFP].ptr;
4559 else if (r->direction && r->direction != direction)
4560 r = r->skip[PF_SKIP_DIR].ptr;
4561 else if (r->af && r->af != af)
4562 r = r->skip[PF_SKIP_AF].ptr;
4563 else if (r->proto && r->proto != pd->proto)
4564 r = r->skip[PF_SKIP_PROTO].ptr;
4565 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af,
4566 r->src.neg, kif))
4567 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
4568 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af,
4569 r->dst.neg, NULL))
4570 r = r->skip[PF_SKIP_DST_ADDR].ptr;
4571 else if (r->tos && !(r->tos == pd->tos))
4572 r = TAILQ_NEXT(r, entries);
4573 else if (r->os_fingerprint != PF_OSFP_ANY)
4574 r = TAILQ_NEXT(r, entries);
4575 else if (pd->proto == IPPROTO_UDP &&
4576 (r->src.port_op || r->dst.port_op))
4577 r = TAILQ_NEXT(r, entries);
4578 else if (pd->proto == IPPROTO_TCP &&
4579 (r->src.port_op || r->dst.port_op || r->flagset))
4580 r = TAILQ_NEXT(r, entries);
4581 else if ((pd->proto == IPPROTO_ICMP ||
4582 pd->proto == IPPROTO_ICMPV6) &&
4583 (r->type || r->code))
4584 r = TAILQ_NEXT(r, entries);
4585 else if (r->prob && r->prob <= arc4random())
4586 r = TAILQ_NEXT(r, entries);
4587 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag))
4588 r = TAILQ_NEXT(r, entries);
4589 else {
4590 if (r->anchor == NULL) {
4591 match = 1;
4592 *rm = r;
4593 *am = a;
4594 *rsm = ruleset;
4595 if ((*rm)->quick)
4596 break;
4597 r = TAILQ_NEXT(r, entries);
4598 } else
4599 pf_step_into_anchor(&asd, &ruleset,
4600 PF_RULESET_FILTER, &r, &a, &match);
4601 }
4602 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
4603 PF_RULESET_FILTER, &r, &a, &match))
4604 break;
4605 }
4606 r = *rm;
4607 a = *am;
4608 ruleset = *rsm;
4609
4610 REASON_SET(&reason, PFRES_MATCH);
4611
4612 if (r->log)
4613 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset,
4614 pd);
4615
4616 if (r->action != PF_PASS)
4617 return (PF_DROP);
4618
4619 if (pf_tag_packet(m, pd->pf_mtag, tag, -1)) {
4620 REASON_SET(&reason, PFRES_MEMORY);
4621 return (PF_DROP);
4622 }
4623
4624 return (PF_PASS);
4625}
4626
4627int
4628pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
4629 struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
4630 u_short *reason)
4631{
4632 struct pf_state_cmp key;
4633 struct tcphdr *th = pd->hdr.tcp;
4634 u_int16_t win = ntohs(th->th_win);
4635 u_int32_t ack, end, seq, orig_seq;
4636 u_int8_t sws, dws;
4637 int ackskew;
4638 int copyback = 0;
4639 struct pf_state_peer *src, *dst;
4640
4641 key.af = pd->af;
4642 key.proto = IPPROTO_TCP;
4643 if (direction == PF_IN) {
4644 PF_ACPY(&key.ext.addr, pd->src, key.af);
4645 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4646 key.ext.port = th->th_sport;
4647 key.gwy.port = th->th_dport;
4648 } else {
4649 PF_ACPY(&key.lan.addr, pd->src, key.af);
4650 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4651 key.lan.port = th->th_sport;
4652 key.ext.port = th->th_dport;
4653 }
4654
4655 STATE_LOOKUP();
4656
4657 if (direction == (*state)->direction) {
4658 src = &(*state)->src;
4659 dst = &(*state)->dst;
4660 } else {
4661 src = &(*state)->dst;
4662 dst = &(*state)->src;
4663 }
4664
4665 if ((*state)->src.state == PF_TCPS_PROXY_SRC) {
4666 if (direction != (*state)->direction) {
4667 REASON_SET(reason, PFRES_SYNPROXY);
4668 return (PF_SYNPROXY_DROP);
4669 }
4670 if (th->th_flags & TH_SYN) {
4671 if (ntohl(th->th_seq) != (*state)->src.seqlo) {
4672 REASON_SET(reason, PFRES_SYNPROXY);
4673 return (PF_DROP);
4674 }
4675#ifdef __FreeBSD__
4676 pf_send_tcp(NULL, (*state)->rule.ptr, pd->af, pd->dst,
4677#else
4678 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
4679#endif
4680 pd->src, th->th_dport, th->th_sport,
4681 (*state)->src.seqhi, ntohl(th->th_seq) + 1,
4682 TH_SYN|TH_ACK, 0, (*state)->src.mss, 0, 1,
4683 0, NULL, NULL);
4684 REASON_SET(reason, PFRES_SYNPROXY);
4685 return (PF_SYNPROXY_DROP);
4686 } else if (!(th->th_flags & TH_ACK) ||
4687 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
4688 (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
4689 REASON_SET(reason, PFRES_SYNPROXY);
4690 return (PF_DROP);
4691 } else if ((*state)->src_node != NULL &&
4692 pf_src_connlimit(state)) {
4693 REASON_SET(reason, PFRES_SRCLIMIT);
4694 return (PF_DROP);
4695 } else
4696 (*state)->src.state = PF_TCPS_PROXY_DST;
4697 }
4698 if ((*state)->src.state == PF_TCPS_PROXY_DST) {
4699 struct pf_state_host *src, *dst;
4700
4701 if (direction == PF_OUT) {
4702 src = &(*state)->gwy;
4703 dst = &(*state)->ext;
4704 } else {
4705 src = &(*state)->ext;
4706 dst = &(*state)->lan;
4707 }
4708 if (direction == (*state)->direction) {
4709 if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) ||
4710 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
4711 (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
4712 REASON_SET(reason, PFRES_SYNPROXY);
4713 return (PF_DROP);
4714 }
4715 (*state)->src.max_win = MAX(ntohs(th->th_win), 1);
4716 if ((*state)->dst.seqhi == 1)
4717 (*state)->dst.seqhi = htonl(arc4random());
4718#ifdef __FreeBSD__
4719 pf_send_tcp(NULL, (*state)->rule.ptr, pd->af,
4720 &src->addr,
4721#else
4722 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
4723#endif
4724 &dst->addr, src->port, dst->port,
4725 (*state)->dst.seqhi, 0, TH_SYN, 0,
4726 (*state)->src.mss, 0, 0, (*state)->tag, NULL, NULL);
4727 REASON_SET(reason, PFRES_SYNPROXY);
4728 return (PF_SYNPROXY_DROP);
4729 } else if (((th->th_flags & (TH_SYN|TH_ACK)) !=
4730 (TH_SYN|TH_ACK)) ||
4731 (ntohl(th->th_ack) != (*state)->dst.seqhi + 1)) {
4732 REASON_SET(reason, PFRES_SYNPROXY);
4733 return (PF_DROP);
4734 } else {
4735 (*state)->dst.max_win = MAX(ntohs(th->th_win), 1);
4736 (*state)->dst.seqlo = ntohl(th->th_seq);
4737#ifdef __FreeBSD__
4738 pf_send_tcp(NULL, (*state)->rule.ptr, pd->af, pd->dst,
4739#else
4740 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
4741#endif
4742 pd->src, th->th_dport, th->th_sport,
4743 ntohl(th->th_ack), ntohl(th->th_seq) + 1,
4744 TH_ACK, (*state)->src.max_win, 0, 0, 0,
4745 (*state)->tag, NULL, NULL);
4746#ifdef __FreeBSD__
4747 pf_send_tcp(NULL, (*state)->rule.ptr, pd->af,
4748 &src->addr,
4749#else
4750 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
4751#endif
4752 &dst->addr, src->port, dst->port,
4753 (*state)->src.seqhi + 1, (*state)->src.seqlo + 1,
4754 TH_ACK, (*state)->dst.max_win, 0, 0, 1,
4755 0, NULL, NULL);
4756 (*state)->src.seqdiff = (*state)->dst.seqhi -
4757 (*state)->src.seqlo;
4758 (*state)->dst.seqdiff = (*state)->src.seqhi -
4759 (*state)->dst.seqlo;
4760 (*state)->src.seqhi = (*state)->src.seqlo +
4761 (*state)->dst.max_win;
4762 (*state)->dst.seqhi = (*state)->dst.seqlo +
4763 (*state)->src.max_win;
4764 (*state)->src.wscale = (*state)->dst.wscale = 0;
4765 (*state)->src.state = (*state)->dst.state =
4766 TCPS_ESTABLISHED;
4767 REASON_SET(reason, PFRES_SYNPROXY);
4768 return (PF_SYNPROXY_DROP);
4769 }
4770 }
4771
4772 if (((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN) &&
4773 dst->state >= TCPS_FIN_WAIT_2 &&
4774 src->state >= TCPS_FIN_WAIT_2) {
4775 if (pf_status.debug >= PF_DEBUG_MISC) {
4776 printf("pf: state reuse ");
4777 pf_print_state(*state);
4778 pf_print_flags(th->th_flags);
4779 printf("\n");
4780 }
4781 /* XXX make sure it's the same direction ?? */
4782 (*state)->src.state = (*state)->dst.state = TCPS_CLOSED;
4783 pf_unlink_state(*state);
4784 *state = NULL;
4785 return (PF_DROP);
4786 }
4787
4788 if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) {
4789 sws = src->wscale & PF_WSCALE_MASK;
4790 dws = dst->wscale & PF_WSCALE_MASK;
4791 } else
4792 sws = dws = 0;
4793
4794 /*
4795 * Sequence tracking algorithm from Guido van Rooij's paper:
4796 * http://www.madison-gurkha.com/publications/tcp_filtering/
4797 * tcp_filtering.ps
4798 */
4799
4800 orig_seq = seq = ntohl(th->th_seq);
4801 if (src->seqlo == 0) {
4802 /* First packet from this end. Set its state */
4803
4804 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) &&
4805 src->scrub == NULL) {
4806 if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) {
4807 REASON_SET(reason, PFRES_MEMORY);
4808 return (PF_DROP);
4809 }
4810 }
4811
4812 /* Deferred generation of sequence number modulator */
4813 if (dst->seqdiff && !src->seqdiff) {
4814#ifdef __FreeBSD__
4815 while ((src->seqdiff = pf_new_isn(*state) - seq) == 0)
4816 ;
4817#else
4818 while ((src->seqdiff = tcp_rndiss_next() - seq) == 0)
4819 ;
4820#endif
4821 ack = ntohl(th->th_ack) - dst->seqdiff;
4822 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
4823 src->seqdiff), 0);
4824 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
4825 copyback = 1;
4826 } else {
4827 ack = ntohl(th->th_ack);
4828 }
4829
4830 end = seq + pd->p_len;
4831 if (th->th_flags & TH_SYN) {
4832 end++;
4833 if (dst->wscale & PF_WSCALE_FLAG) {
4834 src->wscale = pf_get_wscale(m, off, th->th_off,
4835 pd->af);
4836 if (src->wscale & PF_WSCALE_FLAG) {
4837 /* Remove scale factor from initial
4838 * window */
4839 sws = src->wscale & PF_WSCALE_MASK;
4840 win = ((u_int32_t)win + (1 << sws) - 1)
4841 >> sws;
4842 dws = dst->wscale & PF_WSCALE_MASK;
4843 } else {
4844 /* fixup other window */
4845 dst->max_win <<= dst->wscale &
4846 PF_WSCALE_MASK;
4847 /* in case of a retrans SYN|ACK */
4848 dst->wscale = 0;
4849 }
4850 }
4851 }
4852 if (th->th_flags & TH_FIN)
4853 end++;
4854
4855 src->seqlo = seq;
4856 if (src->state < TCPS_SYN_SENT)
4857 src->state = TCPS_SYN_SENT;
4858
4859 /*
4860 * May need to slide the window (seqhi may have been set by
4861 * the crappy stack check or if we picked up the connection
4862 * after establishment)
4863 */
4864 if (src->seqhi == 1 ||
4865 SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi))
4866 src->seqhi = end + MAX(1, dst->max_win << dws);
4867 if (win > src->max_win)
4868 src->max_win = win;
4869
4870 } else {
4871 ack = ntohl(th->th_ack) - dst->seqdiff;
4872 if (src->seqdiff) {
4873 /* Modulate sequence numbers */
4874 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
4875 src->seqdiff), 0);
4876 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
4877 copyback = 1;
4878 }
4879 end = seq + pd->p_len;
4880 if (th->th_flags & TH_SYN)
4881 end++;
4882 if (th->th_flags & TH_FIN)
4883 end++;
4884 }
4885
4886 if ((th->th_flags & TH_ACK) == 0) {
4887 /* Let it pass through the ack skew check */
4888 ack = dst->seqlo;
4889 } else if ((ack == 0 &&
4890 (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) ||
4891 /* broken tcp stacks do not set ack */
4892 (dst->state < TCPS_SYN_SENT)) {
4893 /*
4894 * Many stacks (ours included) will set the ACK number in an
4895 * FIN|ACK if the SYN times out -- no sequence to ACK.
4896 */
4897 ack = dst->seqlo;
4898 }
4899
4900 if (seq == end) {
4901 /* Ease sequencing restrictions on no data packets */
4902 seq = src->seqlo;
4903 end = seq;
4904 }
4905
4906 ackskew = dst->seqlo - ack;
4907
4908
4909 /*
4910 * Need to demodulate the sequence numbers in any TCP SACK options
4911 * (Selective ACK). We could optionally validate the SACK values
4912 * against the current ACK window, either forwards or backwards, but
4913 * I'm not confident that SACK has been implemented properly
4914 * everywhere. It wouldn't surprise me if several stacks accidently
4915 * SACK too far backwards of previously ACKed data. There really aren't
4916 * any security implications of bad SACKing unless the target stack
4917 * doesn't validate the option length correctly. Someone trying to
4918 * spoof into a TCP connection won't bother blindly sending SACK
4919 * options anyway.
4920 */
4921 if (dst->seqdiff && (th->th_off << 2) > sizeof(struct tcphdr)) {
4922 if (pf_modulate_sack(m, off, pd, th, dst))
4923 copyback = 1;
4924 }
4925
4926
4927#define MAXACKWINDOW (0xffff + 1500) /* 1500 is an arbitrary fudge factor */
4928 if (SEQ_GEQ(src->seqhi, end) &&
4929 /* Last octet inside other's window space */
4930 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) &&
4931 /* Retrans: not more than one window back */
4932 (ackskew >= -MAXACKWINDOW) &&
4933 /* Acking not more than one reassembled fragment backwards */
4934 (ackskew <= (MAXACKWINDOW << sws)) &&
4935 /* Acking not more than one window forward */
4936 ((th->th_flags & TH_RST) == 0 || orig_seq == src->seqlo ||
4937 (orig_seq == src->seqlo + 1) || (pd->flags & PFDESC_IP_REAS) == 0)) {
4938 /* Require an exact/+1 sequence match on resets when possible */
4939
4940 if (dst->scrub || src->scrub) {
4941 if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4942 *state, src, dst, ©back))
4943 return (PF_DROP);
4944 }
4945
4946 /* update max window */
4947 if (src->max_win < win)
4948 src->max_win = win;
4949 /* synchronize sequencing */
4950 if (SEQ_GT(end, src->seqlo))
4951 src->seqlo = end;
4952 /* slide the window of what the other end can send */
4953 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4954 dst->seqhi = ack + MAX((win << sws), 1);
4955
4956
4957 /* update states */
4958 if (th->th_flags & TH_SYN)
4959 if (src->state < TCPS_SYN_SENT)
4960 src->state = TCPS_SYN_SENT;
4961 if (th->th_flags & TH_FIN)
4962 if (src->state < TCPS_CLOSING)
4963 src->state = TCPS_CLOSING;
4964 if (th->th_flags & TH_ACK) {
4965 if (dst->state == TCPS_SYN_SENT) {
4966 dst->state = TCPS_ESTABLISHED;
4967 if (src->state == TCPS_ESTABLISHED &&
4968 (*state)->src_node != NULL &&
4969 pf_src_connlimit(state)) {
4970 REASON_SET(reason, PFRES_SRCLIMIT);
4971 return (PF_DROP);
4972 }
4973 } else if (dst->state == TCPS_CLOSING)
4974 dst->state = TCPS_FIN_WAIT_2;
4975 }
4976 if (th->th_flags & TH_RST)
4977 src->state = dst->state = TCPS_TIME_WAIT;
4978
4979 /* update expire time */
4980 (*state)->expire = time_second;
4981 if (src->state >= TCPS_FIN_WAIT_2 &&
4982 dst->state >= TCPS_FIN_WAIT_2)
4983 (*state)->timeout = PFTM_TCP_CLOSED;
4984 else if (src->state >= TCPS_CLOSING &&
4985 dst->state >= TCPS_CLOSING)
4986 (*state)->timeout = PFTM_TCP_FIN_WAIT;
4987 else if (src->state < TCPS_ESTABLISHED ||
4988 dst->state < TCPS_ESTABLISHED)
4989 (*state)->timeout = PFTM_TCP_OPENING;
4990 else if (src->state >= TCPS_CLOSING ||
4991 dst->state >= TCPS_CLOSING)
4992 (*state)->timeout = PFTM_TCP_CLOSING;
4993 else
4994 (*state)->timeout = PFTM_TCP_ESTABLISHED;
4995
4996 /* Fall through to PASS packet */
4997
4998 } else if ((dst->state < TCPS_SYN_SENT ||
4999 dst->state >= TCPS_FIN_WAIT_2 ||
5000 src->state >= TCPS_FIN_WAIT_2) &&
5001 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) &&
5002 /* Within a window forward of the originating packet */
5003 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) {
5004 /* Within a window backward of the originating packet */
5005
5006 /*
5007 * This currently handles three situations:
5008 * 1) Stupid stacks will shotgun SYNs before their peer
5009 * replies.
5010 * 2) When PF catches an already established stream (the
5011 * firewall rebooted, the state table was flushed, routes
5012 * changed...)
5013 * 3) Packets get funky immediately after the connection
5014 * closes (this should catch Solaris spurious ACK|FINs
5015 * that web servers like to spew after a close)
5016 *
5017 * This must be a little more careful than the above code
5018 * since packet floods will also be caught here. We don't
5019 * update the TTL here to mitigate the damage of a packet
5020 * flood and so the same code can handle awkward establishment
5021 * and a loosened connection close.
5022 * In the establishment case, a correct peer response will
5023 * validate the connection, go through the normal state code
5024 * and keep updating the state TTL.
5025 */
5026
5027 if (pf_status.debug >= PF_DEBUG_MISC) {
5028 printf("pf: loose state match: ");
5029 pf_print_state(*state);
5030 pf_print_flags(th->th_flags);
5031 printf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
5032 "pkts=%llu:%llu\n", seq, orig_seq, ack, pd->p_len,
5033#ifdef __FreeBSD__
5034 ackskew, (unsigned long long)(*state)->packets[0],
5035 (unsigned long long)(*state)->packets[1]);
5036#else
5037 ackskew, (*state)->packets[0],
5038 (*state)->packets[1]);
5039#endif
5040 }
5041
5042 if (dst->scrub || src->scrub) {
5043 if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
5044 *state, src, dst, ©back))
5045 return (PF_DROP);
5046 }
5047
5048 /* update max window */
5049 if (src->max_win < win)
5050 src->max_win = win;
5051 /* synchronize sequencing */
5052 if (SEQ_GT(end, src->seqlo))
5053 src->seqlo = end;
5054 /* slide the window of what the other end can send */
5055 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
5056 dst->seqhi = ack + MAX((win << sws), 1);
5057
5058 /*
5059 * Cannot set dst->seqhi here since this could be a shotgunned
5060 * SYN and not an already established connection.
5061 */
5062
5063 if (th->th_flags & TH_FIN)
5064 if (src->state < TCPS_CLOSING)
5065 src->state = TCPS_CLOSING;
5066 if (th->th_flags & TH_RST)
5067 src->state = dst->state = TCPS_TIME_WAIT;
5068
5069 /* Fall through to PASS packet */
5070
5071 } else {
5072 if ((*state)->dst.state == TCPS_SYN_SENT &&
5073 (*state)->src.state == TCPS_SYN_SENT) {
5074 /* Send RST for state mismatches during handshake */
5075 if (!(th->th_flags & TH_RST))
5076#ifdef __FreeBSD__
5077 pf_send_tcp(m, (*state)->rule.ptr, pd->af,
5078#else
5079 pf_send_tcp((*state)->rule.ptr, pd->af,
5080#endif
5081 pd->dst, pd->src, th->th_dport,
5082 th->th_sport, ntohl(th->th_ack), 0,
5083 TH_RST, 0, 0,
5084 (*state)->rule.ptr->return_ttl, 1, 0,
5085 pd->eh, kif->pfik_ifp);
5086 src->seqlo = 0;
5087 src->seqhi = 1;
5088 src->max_win = 1;
5089 } else if (pf_status.debug >= PF_DEBUG_MISC) {
5090 printf("pf: BAD state: ");
5091 pf_print_state(*state);
5092 pf_print_flags(th->th_flags);
5093 printf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
5094 "pkts=%llu:%llu dir=%s,%s\n",
5095 seq, orig_seq, ack, pd->p_len, ackskew,
5096#ifdef __FreeBSD__
5097 (unsigned long long)(*state)->packets[0],
5098 (unsigned long long)(*state)->packets[1],
5099#else
5100 (*state)->packets[0], (*state)->packets[1],
5101#endif
5102 direction == PF_IN ? "in" : "out",
5103 direction == (*state)->direction ? "fwd" : "rev");
5104 printf("pf: State failure on: %c %c %c %c | %c %c\n",
5105 SEQ_GEQ(src->seqhi, end) ? ' ' : '1',
5106 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ?
5107 ' ': '2',
5108 (ackskew >= -MAXACKWINDOW) ? ' ' : '3',
5109 (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4',
5110 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5',
5111 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6');
5112 }
5113 REASON_SET(reason, PFRES_BADSTATE);
5114 return (PF_DROP);
5115 }
5116
5117 /* Any packets which have gotten here are to be passed */
5118
5119 /* translate source/destination address, if necessary */
5120 if (STATE_TRANSLATE(*state)) {
5121 if (direction == PF_OUT)
5122 pf_change_ap(pd->src, &th->th_sport, pd->ip_sum,
5123 &th->th_sum, &(*state)->gwy.addr,
5124 (*state)->gwy.port, 0, pd->af);
5125 else
5126 pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum,
5127 &th->th_sum, &(*state)->lan.addr,
5128 (*state)->lan.port, 0, pd->af);
5129 m_copyback(m, off, sizeof(*th), (caddr_t)th);
5130 } else if (copyback) {
5131 /* Copyback sequence modulation or stateful scrub changes */
5132 m_copyback(m, off, sizeof(*th), (caddr_t)th);
5133 }
5134
5135 return (PF_PASS);
5136}
5137
5138int
5139pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif,
5140 struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
5141{
5142 struct pf_state_peer *src, *dst;
5143 struct pf_state_cmp key;
5144 struct udphdr *uh = pd->hdr.udp;
5145
5146 key.af = pd->af;
5147 key.proto = IPPROTO_UDP;
5148 if (direction == PF_IN) {
5149 PF_ACPY(&key.ext.addr, pd->src, key.af);
5150 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
5151 key.ext.port = uh->uh_sport;
5152 key.gwy.port = uh->uh_dport;
5153 } else {
5154 PF_ACPY(&key.lan.addr, pd->src, key.af);
5155 PF_ACPY(&key.ext.addr, pd->dst, key.af);
5156 key.lan.port = uh->uh_sport;
5157 key.ext.port = uh->uh_dport;
5158 }
5159
5160 STATE_LOOKUP();
5161
5162 if (direction == (*state)->direction) {
5163 src = &(*state)->src;
5164 dst = &(*state)->dst;
5165 } else {
5166 src = &(*state)->dst;
5167 dst = &(*state)->src;
5168 }
5169
5170 /* update states */
5171 if (src->state < PFUDPS_SINGLE)
5172 src->state = PFUDPS_SINGLE;
5173 if (dst->state == PFUDPS_SINGLE)
5174 dst->state = PFUDPS_MULTIPLE;
5175
5176 /* update expire time */
5177 (*state)->expire = time_second;
5178 if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE)
5179 (*state)->timeout = PFTM_UDP_MULTIPLE;
5180 else
5181 (*state)->timeout = PFTM_UDP_SINGLE;
5182
5183 /* translate source/destination address, if necessary */
5184 if (STATE_TRANSLATE(*state)) {
5185 if (direction == PF_OUT)
5186 pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum,
5187 &uh->uh_sum, &(*state)->gwy.addr,
5188 (*state)->gwy.port, 1, pd->af);
5189 else
5190 pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum,
5191 &uh->uh_sum, &(*state)->lan.addr,
5192 (*state)->lan.port, 1, pd->af);
5193 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
5194 }
5195
5196 return (PF_PASS);
5197}
5198
5199int
5200pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif,
5201 struct mbuf *m, int off, void *h, struct pf_pdesc *pd, u_short *reason)
5202{
5203 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
5204 u_int16_t icmpid = 0; /* make the compiler happy */
5205 u_int16_t *icmpsum = NULL; /* make the compiler happy */
5206 u_int8_t icmptype = 0; /* make the compiler happy */
5207 int state_icmp = 0;
5208 struct pf_state_cmp key;
5209
5210 switch (pd->proto) {
5211#ifdef INET
5212 case IPPROTO_ICMP:
5213 icmptype = pd->hdr.icmp->icmp_type;
5214 icmpid = pd->hdr.icmp->icmp_id;
5215 icmpsum = &pd->hdr.icmp->icmp_cksum;
5216
5217 if (icmptype == ICMP_UNREACH ||
5218 icmptype == ICMP_SOURCEQUENCH ||
5219 icmptype == ICMP_REDIRECT ||
5220 icmptype == ICMP_TIMXCEED ||
5221 icmptype == ICMP_PARAMPROB)
5222 state_icmp++;
5223 break;
5224#endif /* INET */
5225#ifdef INET6
5226 case IPPROTO_ICMPV6:
5227 icmptype = pd->hdr.icmp6->icmp6_type;
5228 icmpid = pd->hdr.icmp6->icmp6_id;
5229 icmpsum = &pd->hdr.icmp6->icmp6_cksum;
5230
5231 if (icmptype == ICMP6_DST_UNREACH ||
5232 icmptype == ICMP6_PACKET_TOO_BIG ||
5233 icmptype == ICMP6_TIME_EXCEEDED ||
5234 icmptype == ICMP6_PARAM_PROB)
5235 state_icmp++;
5236 break;
5237#endif /* INET6 */
5238 }
5239
5240 if (!state_icmp) {
5241
5242 /*
5243 * ICMP query/reply message not related to a TCP/UDP packet.
5244 * Search for an ICMP state.
5245 */
5246 key.af = pd->af;
5247 key.proto = pd->proto;
5248 if (direction == PF_IN) {
5249 PF_ACPY(&key.ext.addr, pd->src, key.af);
5250 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
5251 key.ext.port = 0;
5252 key.gwy.port = icmpid;
5253 } else {
5254 PF_ACPY(&key.lan.addr, pd->src, key.af);
5255 PF_ACPY(&key.ext.addr, pd->dst, key.af);
5256 key.lan.port = icmpid;
5257 key.ext.port = 0;
5258 }
5259
5260 STATE_LOOKUP();
5261
5262 (*state)->expire = time_second;
5263 (*state)->timeout = PFTM_ICMP_ERROR_REPLY;
5264
5265 /* translate source/destination address, if necessary */
5266 if (STATE_TRANSLATE(*state)) {
5267 if (direction == PF_OUT) {
5268 switch (pd->af) {
5269#ifdef INET
5270 case AF_INET:
5271 pf_change_a(&saddr->v4.s_addr,
5272 pd->ip_sum,
5273 (*state)->gwy.addr.v4.s_addr, 0);
5274 pd->hdr.icmp->icmp_cksum =
5275 pf_cksum_fixup(
5276 pd->hdr.icmp->icmp_cksum, icmpid,
5277 (*state)->gwy.port, 0);
5278 pd->hdr.icmp->icmp_id =
5279 (*state)->gwy.port;
5280 m_copyback(m, off, ICMP_MINLEN,
5281 (caddr_t)pd->hdr.icmp);
5282 break;
5283#endif /* INET */
5284#ifdef INET6
5285 case AF_INET6:
5286 pf_change_a6(saddr,
5287 &pd->hdr.icmp6->icmp6_cksum,
5288 &(*state)->gwy.addr, 0);
5289 m_copyback(m, off,
5290 sizeof(struct icmp6_hdr),
5291 (caddr_t)pd->hdr.icmp6);
5292 break;
5293#endif /* INET6 */
5294 }
5295 } else {
5296 switch (pd->af) {
5297#ifdef INET
5298 case AF_INET:
5299 pf_change_a(&daddr->v4.s_addr,
5300 pd->ip_sum,
5301 (*state)->lan.addr.v4.s_addr, 0);
5302 pd->hdr.icmp->icmp_cksum =
5303 pf_cksum_fixup(
5304 pd->hdr.icmp->icmp_cksum, icmpid,
5305 (*state)->lan.port, 0);
5306 pd->hdr.icmp->icmp_id =
5307 (*state)->lan.port;
5308 m_copyback(m, off, ICMP_MINLEN,
5309 (caddr_t)pd->hdr.icmp);
5310 break;
5311#endif /* INET */
5312#ifdef INET6
5313 case AF_INET6:
5314 pf_change_a6(daddr,
5315 &pd->hdr.icmp6->icmp6_cksum,
5316 &(*state)->lan.addr, 0);
5317 m_copyback(m, off,
5318 sizeof(struct icmp6_hdr),
5319 (caddr_t)pd->hdr.icmp6);
5320 break;
5321#endif /* INET6 */
5322 }
5323 }
5324 }
5325
5326 return (PF_PASS);
5327
5328 } else {
5329 /*
5330 * ICMP error message in response to a TCP/UDP packet.
5331 * Extract the inner TCP/UDP header and search for that state.
5332 */
5333
5334 struct pf_pdesc pd2;
5335#ifdef INET
5336 struct ip h2;
5337#endif /* INET */
5338#ifdef INET6
5339 struct ip6_hdr h2_6;
5340 int terminal = 0;
5341#endif /* INET6 */
5342 int ipoff2 = 0; /* make the compiler happy */
5343 int off2 = 0; /* make the compiler happy */
5344
5345 pd2.af = pd->af;
5346 switch (pd->af) {
5347#ifdef INET
5348 case AF_INET:
5349 /* offset of h2 in mbuf chain */
5350 ipoff2 = off + ICMP_MINLEN;
5351
5352 if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2),
5353 NULL, reason, pd2.af)) {
5354 DPFPRINTF(PF_DEBUG_MISC,
5355 ("pf: ICMP error message too short "
5356 "(ip)\n"));
5357 return (PF_DROP);
5358 }
5359 /*
5360 * ICMP error messages don't refer to non-first
5361 * fragments
5362 */
5363 if (h2.ip_off & htons(IP_OFFMASK)) {
5364 REASON_SET(reason, PFRES_FRAG);
5365 return (PF_DROP);
5366 }
5367
5368 /* offset of protocol header that follows h2 */
5369 off2 = ipoff2 + (h2.ip_hl << 2);
5370
5371 pd2.proto = h2.ip_p;
5372 pd2.src = (struct pf_addr *)&h2.ip_src;
5373 pd2.dst = (struct pf_addr *)&h2.ip_dst;
5374 pd2.ip_sum = &h2.ip_sum;
5375 break;
5376#endif /* INET */
5377#ifdef INET6
5378 case AF_INET6:
5379 ipoff2 = off + sizeof(struct icmp6_hdr);
5380
5381 if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6),
5382 NULL, reason, pd2.af)) {
5383 DPFPRINTF(PF_DEBUG_MISC,
5384 ("pf: ICMP error message too short "
5385 "(ip6)\n"));
5386 return (PF_DROP);
5387 }
5388 pd2.proto = h2_6.ip6_nxt;
5389 pd2.src = (struct pf_addr *)&h2_6.ip6_src;
5390 pd2.dst = (struct pf_addr *)&h2_6.ip6_dst;
5391 pd2.ip_sum = NULL;
5392 off2 = ipoff2 + sizeof(h2_6);
5393 do {
5394 switch (pd2.proto) {
5395 case IPPROTO_FRAGMENT:
5396 /*
5397 * ICMPv6 error messages for
5398 * non-first fragments
5399 */
5400 REASON_SET(reason, PFRES_FRAG);
5401 return (PF_DROP);
5402 case IPPROTO_AH:
5403 case IPPROTO_HOPOPTS:
5404 case IPPROTO_ROUTING:
5405 case IPPROTO_DSTOPTS: {
5406 /* get next header and header length */
5407 struct ip6_ext opt6;
5408
5409 if (!pf_pull_hdr(m, off2, &opt6,
5410 sizeof(opt6), NULL, reason,
5411 pd2.af)) {
5412 DPFPRINTF(PF_DEBUG_MISC,
5413 ("pf: ICMPv6 short opt\n"));
5414 return (PF_DROP);
5415 }
5416 if (pd2.proto == IPPROTO_AH)
5417 off2 += (opt6.ip6e_len + 2) * 4;
5418 else
5419 off2 += (opt6.ip6e_len + 1) * 8;
5420 pd2.proto = opt6.ip6e_nxt;
5421 /* goto the next header */
5422 break;
5423 }
5424 default:
5425 terminal++;
5426 break;
5427 }
5428 } while (!terminal);
5429 break;
5430#endif /* INET6 */
5431#ifdef __FreeBSD__
5432 default:
5433 panic("AF not supported: %d", pd->af);
5434#endif
5435 }
5436
5437 switch (pd2.proto) {
5438 case IPPROTO_TCP: {
5439 struct tcphdr th;
5440 u_int32_t seq;
5441 struct pf_state_peer *src, *dst;
5442 u_int8_t dws;
5443 int copyback = 0;
5444
5445 /*
5446 * Only the first 8 bytes of the TCP header can be
5447 * expected. Don't access any TCP header fields after
5448 * th_seq, an ackskew test is not possible.
5449 */
5450 if (!pf_pull_hdr(m, off2, &th, 8, NULL, reason,
5451 pd2.af)) {
5452 DPFPRINTF(PF_DEBUG_MISC,
5453 ("pf: ICMP error message too short "
5454 "(tcp)\n"));
5455 return (PF_DROP);
5456 }
5457
5458 key.af = pd2.af;
5459 key.proto = IPPROTO_TCP;
5460 if (direction == PF_IN) {
5461 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
5462 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
5463 key.ext.port = th.th_dport;
5464 key.gwy.port = th.th_sport;
5465 } else {
5466 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
5467 PF_ACPY(&key.ext.addr, pd2.src, key.af);
5468 key.lan.port = th.th_dport;
5469 key.ext.port = th.th_sport;
5470 }
5471
5472 STATE_LOOKUP();
5473
5474 if (direction == (*state)->direction) {
5475 src = &(*state)->dst;
5476 dst = &(*state)->src;
5477 } else {
5478 src = &(*state)->src;
5479 dst = &(*state)->dst;
5480 }
5481
5482 if (src->wscale && dst->wscale)
5483 dws = dst->wscale & PF_WSCALE_MASK;
5484 else
5485 dws = 0;
5486
5487 /* Demodulate sequence number */
5488 seq = ntohl(th.th_seq) - src->seqdiff;
5489 if (src->seqdiff) {
5490 pf_change_a(&th.th_seq, icmpsum,
5491 htonl(seq), 0);
5492 copyback = 1;
5493 }
5494
5495 if (!SEQ_GEQ(src->seqhi, seq) ||
5496 !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws))) {
5497 if (pf_status.debug >= PF_DEBUG_MISC) {
5498 printf("pf: BAD ICMP %d:%d ",
5499 icmptype, pd->hdr.icmp->icmp_code);
5500 pf_print_host(pd->src, 0, pd->af);
5501 printf(" -> ");
5502 pf_print_host(pd->dst, 0, pd->af);
5503 printf(" state: ");
5504 pf_print_state(*state);
5505 printf(" seq=%u\n", seq);
5506 }
5507 REASON_SET(reason, PFRES_BADSTATE);
5508 return (PF_DROP);
5509 }
5510
5511 if (STATE_TRANSLATE(*state)) {
5512 if (direction == PF_IN) {
5513 pf_change_icmp(pd2.src, &th.th_sport,
5514 daddr, &(*state)->lan.addr,
5515 (*state)->lan.port, NULL,
5516 pd2.ip_sum, icmpsum,
5517 pd->ip_sum, 0, pd2.af);
5518 } else {
5519 pf_change_icmp(pd2.dst, &th.th_dport,
5520 saddr, &(*state)->gwy.addr,
5521 (*state)->gwy.port, NULL,
5522 pd2.ip_sum, icmpsum,
5523 pd->ip_sum, 0, pd2.af);
5524 }
5525 copyback = 1;
5526 }
5527
5528 if (copyback) {
5529 switch (pd2.af) {
5530#ifdef INET
5531 case AF_INET:
5532 m_copyback(m, off, ICMP_MINLEN,
5533 (caddr_t)pd->hdr.icmp);
5534 m_copyback(m, ipoff2, sizeof(h2),
5535 (caddr_t)&h2);
5536 break;
5537#endif /* INET */
5538#ifdef INET6
5539 case AF_INET6:
5540 m_copyback(m, off,
5541 sizeof(struct icmp6_hdr),
5542 (caddr_t)pd->hdr.icmp6);
5543 m_copyback(m, ipoff2, sizeof(h2_6),
5544 (caddr_t)&h2_6);
5545 break;
5546#endif /* INET6 */
5547 }
5548 m_copyback(m, off2, 8, (caddr_t)&th);
5549 }
5550
5551 return (PF_PASS);
5552 break;
5553 }
5554 case IPPROTO_UDP: {
5555 struct udphdr uh;
5556
5557 if (!pf_pull_hdr(m, off2, &uh, sizeof(uh),
5558 NULL, reason, pd2.af)) {
5559 DPFPRINTF(PF_DEBUG_MISC,
5560 ("pf: ICMP error message too short "
5561 "(udp)\n"));
5562 return (PF_DROP);
5563 }
5564
5565 key.af = pd2.af;
5566 key.proto = IPPROTO_UDP;
5567 if (direction == PF_IN) {
5568 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
5569 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
5570 key.ext.port = uh.uh_dport;
5571 key.gwy.port = uh.uh_sport;
5572 } else {
5573 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
5574 PF_ACPY(&key.ext.addr, pd2.src, key.af);
5575 key.lan.port = uh.uh_dport;
5576 key.ext.port = uh.uh_sport;
5577 }
5578
5579 STATE_LOOKUP();
5580
5581 if (STATE_TRANSLATE(*state)) {
5582 if (direction == PF_IN) {
5583 pf_change_icmp(pd2.src, &uh.uh_sport,
5584 daddr, &(*state)->lan.addr,
5585 (*state)->lan.port, &uh.uh_sum,
5586 pd2.ip_sum, icmpsum,
5587 pd->ip_sum, 1, pd2.af);
5588 } else {
5589 pf_change_icmp(pd2.dst, &uh.uh_dport,
5590 saddr, &(*state)->gwy.addr,
5591 (*state)->gwy.port, &uh.uh_sum,
5592 pd2.ip_sum, icmpsum,
5593 pd->ip_sum, 1, pd2.af);
5594 }
5595 switch (pd2.af) {
5596#ifdef INET
5597 case AF_INET:
5598 m_copyback(m, off, ICMP_MINLEN,
5599 (caddr_t)pd->hdr.icmp);
5600 m_copyback(m, ipoff2, sizeof(h2),
5601 (caddr_t)&h2);
5602 break;
5603#endif /* INET */
5604#ifdef INET6
5605 case AF_INET6:
5606 m_copyback(m, off,
5607 sizeof(struct icmp6_hdr),
5608 (caddr_t)pd->hdr.icmp6);
5609 m_copyback(m, ipoff2, sizeof(h2_6),
5610 (caddr_t)&h2_6);
5611 break;
5612#endif /* INET6 */
5613 }
5614 m_copyback(m, off2, sizeof(uh),
5615 (caddr_t)&uh);
5616 }
5617
5618 return (PF_PASS);
5619 break;
5620 }
5621#ifdef INET
5622 case IPPROTO_ICMP: {
5623 struct icmp iih;
5624
5625 if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN,
5626 NULL, reason, pd2.af)) {
5627 DPFPRINTF(PF_DEBUG_MISC,
5628 ("pf: ICMP error message too short i"
5629 "(icmp)\n"));
5630 return (PF_DROP);
5631 }
5632
5633 key.af = pd2.af;
5634 key.proto = IPPROTO_ICMP;
5635 if (direction == PF_IN) {
5636 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
5637 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
5638 key.ext.port = 0;
5639 key.gwy.port = iih.icmp_id;
5640 } else {
5641 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
5642 PF_ACPY(&key.ext.addr, pd2.src, key.af);
5643 key.lan.port = iih.icmp_id;
5644 key.ext.port = 0;
5645 }
5646
5647 STATE_LOOKUP();
5648
5649 if (STATE_TRANSLATE(*state)) {
5650 if (direction == PF_IN) {
5651 pf_change_icmp(pd2.src, &iih.icmp_id,
5652 daddr, &(*state)->lan.addr,
5653 (*state)->lan.port, NULL,
5654 pd2.ip_sum, icmpsum,
5655 pd->ip_sum, 0, AF_INET);
5656 } else {
5657 pf_change_icmp(pd2.dst, &iih.icmp_id,
5658 saddr, &(*state)->gwy.addr,
5659 (*state)->gwy.port, NULL,
5660 pd2.ip_sum, icmpsum,
5661 pd->ip_sum, 0, AF_INET);
5662 }
5663 m_copyback(m, off, ICMP_MINLEN,
5664 (caddr_t)pd->hdr.icmp);
5665 m_copyback(m, ipoff2, sizeof(h2),
5666 (caddr_t)&h2);
5667 m_copyback(m, off2, ICMP_MINLEN,
5668 (caddr_t)&iih);
5669 }
5670
5671 return (PF_PASS);
5672 break;
5673 }
5674#endif /* INET */
5675#ifdef INET6
5676 case IPPROTO_ICMPV6: {
5677 struct icmp6_hdr iih;
5678
5679 if (!pf_pull_hdr(m, off2, &iih,
5680 sizeof(struct icmp6_hdr), NULL, reason, pd2.af)) {
5681 DPFPRINTF(PF_DEBUG_MISC,
5682 ("pf: ICMP error message too short "
5683 "(icmp6)\n"));
5684 return (PF_DROP);
5685 }
5686
5687 key.af = pd2.af;
5688 key.proto = IPPROTO_ICMPV6;
5689 if (direction == PF_IN) {
5690 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
5691 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
5692 key.ext.port = 0;
5693 key.gwy.port = iih.icmp6_id;
5694 } else {
5695 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
5696 PF_ACPY(&key.ext.addr, pd2.src, key.af);
5697 key.lan.port = iih.icmp6_id;
5698 key.ext.port = 0;
5699 }
5700
5701 STATE_LOOKUP();
5702
5703 if (STATE_TRANSLATE(*state)) {
5704 if (direction == PF_IN) {
5705 pf_change_icmp(pd2.src, &iih.icmp6_id,
5706 daddr, &(*state)->lan.addr,
5707 (*state)->lan.port, NULL,
5708 pd2.ip_sum, icmpsum,
5709 pd->ip_sum, 0, AF_INET6);
5710 } else {
5711 pf_change_icmp(pd2.dst, &iih.icmp6_id,
5712 saddr, &(*state)->gwy.addr,
5713 (*state)->gwy.port, NULL,
5714 pd2.ip_sum, icmpsum,
5715 pd->ip_sum, 0, AF_INET6);
5716 }
5717 m_copyback(m, off, sizeof(struct icmp6_hdr),
5718 (caddr_t)pd->hdr.icmp6);
5719 m_copyback(m, ipoff2, sizeof(h2_6),
5720 (caddr_t)&h2_6);
5721 m_copyback(m, off2, sizeof(struct icmp6_hdr),
5722 (caddr_t)&iih);
5723 }
5724
5725 return (PF_PASS);
5726 break;
5727 }
5728#endif /* INET6 */
5729 default: {
5730 key.af = pd2.af;
5731 key.proto = pd2.proto;
5732 if (direction == PF_IN) {
5733 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
5734 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
5735 key.ext.port = 0;
5736 key.gwy.port = 0;
5737 } else {
5738 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
5739 PF_ACPY(&key.ext.addr, pd2.src, key.af);
5740 key.lan.port = 0;
5741 key.ext.port = 0;
5742 }
5743
5744 STATE_LOOKUP();
5745
5746 if (STATE_TRANSLATE(*state)) {
5747 if (direction == PF_IN) {
5748 pf_change_icmp(pd2.src, NULL,
5749 daddr, &(*state)->lan.addr,
5750 0, NULL,
5751 pd2.ip_sum, icmpsum,
5752 pd->ip_sum, 0, pd2.af);
5753 } else {
5754 pf_change_icmp(pd2.dst, NULL,
5755 saddr, &(*state)->gwy.addr,
5756 0, NULL,
5757 pd2.ip_sum, icmpsum,
5758 pd->ip_sum, 0, pd2.af);
5759 }
5760 switch (pd2.af) {
5761#ifdef INET
5762 case AF_INET:
5763 m_copyback(m, off, ICMP_MINLEN,
5764 (caddr_t)pd->hdr.icmp);
5765 m_copyback(m, ipoff2, sizeof(h2),
5766 (caddr_t)&h2);
5767 break;
5768#endif /* INET */
5769#ifdef INET6
5770 case AF_INET6:
5771 m_copyback(m, off,
5772 sizeof(struct icmp6_hdr),
5773 (caddr_t)pd->hdr.icmp6);
5774 m_copyback(m, ipoff2, sizeof(h2_6),
5775 (caddr_t)&h2_6);
5776 break;
5777#endif /* INET6 */
5778 }
5779 }
5780
5781 return (PF_PASS);
5782 break;
5783 }
5784 }
5785 }
5786}
5787
5788int
5789pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif,
5790 struct pf_pdesc *pd)
5791{
5792 struct pf_state_peer *src, *dst;
5793 struct pf_state_cmp key;
5794
5795 key.af = pd->af;
5796 key.proto = pd->proto;
5797 if (direction == PF_IN) {
5798 PF_ACPY(&key.ext.addr, pd->src, key.af);
5799 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
5800 key.ext.port = 0;
5801 key.gwy.port = 0;
5802 } else {
5803 PF_ACPY(&key.lan.addr, pd->src, key.af);
5804 PF_ACPY(&key.ext.addr, pd->dst, key.af);
5805 key.lan.port = 0;
5806 key.ext.port = 0;
5807 }
5808
5809 STATE_LOOKUP();
5810
5811 if (direction == (*state)->direction) {
5812 src = &(*state)->src;
5813 dst = &(*state)->dst;
5814 } else {
5815 src = &(*state)->dst;
5816 dst = &(*state)->src;
5817 }
5818
5819 /* update states */
5820 if (src->state < PFOTHERS_SINGLE)
5821 src->state = PFOTHERS_SINGLE;
5822 if (dst->state == PFOTHERS_SINGLE)
5823 dst->state = PFOTHERS_MULTIPLE;
5824
5825 /* update expire time */
5826 (*state)->expire = time_second;
5827 if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE)
5828 (*state)->timeout = PFTM_OTHER_MULTIPLE;
5829 else
5830 (*state)->timeout = PFTM_OTHER_SINGLE;
5831
5832 /* translate source/destination address, if necessary */
5833 if (STATE_TRANSLATE(*state)) {
5834 if (direction == PF_OUT)
5835 switch (pd->af) {
5836#ifdef INET
5837 case AF_INET:
5838 pf_change_a(&pd->src->v4.s_addr,
5839 pd->ip_sum, (*state)->gwy.addr.v4.s_addr,
5840 0);
5841 break;
5842#endif /* INET */
5843#ifdef INET6
5844 case AF_INET6:
5845 PF_ACPY(pd->src, &(*state)->gwy.addr, pd->af);
5846 break;
5847#endif /* INET6 */
5848 }
5849 else
5850 switch (pd->af) {
5851#ifdef INET
5852 case AF_INET:
5853 pf_change_a(&pd->dst->v4.s_addr,
5854 pd->ip_sum, (*state)->lan.addr.v4.s_addr,
5855 0);
5856 break;
5857#endif /* INET */
5858#ifdef INET6
5859 case AF_INET6:
5860 PF_ACPY(pd->dst, &(*state)->lan.addr, pd->af);
5861 break;
5862#endif /* INET6 */
5863 }
5864 }
5865
5866 return (PF_PASS);
5867}
5868
5869/*
5870 * ipoff and off are measured from the start of the mbuf chain.
5871 * h must be at "ipoff" on the mbuf chain.
5872 */
5873void *
5874pf_pull_hdr(struct mbuf *m, int off, void *p, int len,
5875 u_short *actionp, u_short *reasonp, sa_family_t af)
5876{
5877 switch (af) {
5878#ifdef INET
5879 case AF_INET: {
5880 struct ip *h = mtod(m, struct ip *);
5881 u_int16_t fragoff = (ntohs(h->ip_off) & IP_OFFMASK) << 3;
5882
5883 if (fragoff) {
5884 if (fragoff >= len)
5885 ACTION_SET(actionp, PF_PASS);
5886 else {
5887 ACTION_SET(actionp, PF_DROP);
5888 REASON_SET(reasonp, PFRES_FRAG);
5889 }
5890 return (NULL);
5891 }
5892 if (m->m_pkthdr.len < off + len ||
5893 ntohs(h->ip_len) < off + len) {
5894 ACTION_SET(actionp, PF_DROP);
5895 REASON_SET(reasonp, PFRES_SHORT);
5896 return (NULL);
5897 }
5898 break;
5899 }
5900#endif /* INET */
5901#ifdef INET6
5902 case AF_INET6: {
5903 struct ip6_hdr *h = mtod(m, struct ip6_hdr *);
5904
5905 if (m->m_pkthdr.len < off + len ||
5906 (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) <
5907 (unsigned)(off + len)) {
5908 ACTION_SET(actionp, PF_DROP);
5909 REASON_SET(reasonp, PFRES_SHORT);
5910 return (NULL);
5911 }
5912 break;
5913 }
5914#endif /* INET6 */
5915 }
5916 m_copydata(m, off, len, p);
5917 return (p);
5918}
5919
5920int
5921pf_routable(struct pf_addr *addr, sa_family_t af, struct pfi_kif *kif)
5922{
5923 struct sockaddr_in *dst;
5924 int ret = 1;
5925 int check_mpath;
5926#ifndef __FreeBSD__
5927 extern int ipmultipath;
5928#endif
5929#ifdef INET6
5930#ifndef __FreeBSD__
5931 extern int ip6_multipath;
5932#endif
5933 struct sockaddr_in6 *dst6;
5934 struct route_in6 ro;
5935#else
5936 struct route ro;
5937#endif
5938 struct radix_node *rn;
5939 struct rtentry *rt;
5940 struct ifnet *ifp;
5941
5942 check_mpath = 0;
5943 bzero(&ro, sizeof(ro));
5944 switch (af) {
5945 case AF_INET:
5946 dst = satosin(&ro.ro_dst);
5947 dst->sin_family = AF_INET;
5948 dst->sin_len = sizeof(*dst);
5949 dst->sin_addr = addr->v4;
5950#ifndef __FreeBSD__ /* MULTIPATH_ROUTING */
5951 if (ipmultipath)
5952 check_mpath = 1;
5953#endif
5954 break;
5955#ifdef INET6
5956 case AF_INET6:
5957 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
5958 dst6->sin6_family = AF_INET6;
5959 dst6->sin6_len = sizeof(*dst6);
5960 dst6->sin6_addr = addr->v6;
5961#ifndef __FreeBSD__ /* MULTIPATH_ROUTING */
5962 if (ip6_multipath)
5963 check_mpath = 1;
5964#endif
5965 break;
5966#endif /* INET6 */
5967 default:
5968 return (0);
5969 }
5970
5971 /* Skip checks for ipsec interfaces */
5972 if (kif != NULL && kif->pfik_ifp->if_type == IFT_ENC)
5973 goto out;
5974
5975#ifdef __FreeBSD__
5976/* XXX MRT not always INET */ /* stick with table 0 though */
5977 if (af == AF_INET)
5978 in_rtalloc_ign((struct route *)&ro, 0, 0);
5979 else
5980 rtalloc_ign((struct route *)&ro, 0);
5981#else /* ! __FreeBSD__ */
5982 rtalloc_noclone((struct route *)&ro, NO_CLONING);
5983#endif
5984
5985 if (ro.ro_rt != NULL) {
5986 /* No interface given, this is a no-route check */
5987 if (kif == NULL)
5988 goto out;
5989
5990 if (kif->pfik_ifp == NULL) {
5991 ret = 0;
5992 goto out;
5993 }
5994
5995 /* Perform uRPF check if passed input interface */
5996 ret = 0;
5997 rn = (struct radix_node *)ro.ro_rt;
5998 do {
5999 rt = (struct rtentry *)rn;
6000#ifndef __FreeBSD__ /* CARPDEV */
6001 if (rt->rt_ifp->if_type == IFT_CARP)
6002 ifp = rt->rt_ifp->if_carpdev;
6003 else
6004#endif
6005 ifp = rt->rt_ifp;
6006
6007 if (kif->pfik_ifp == ifp)
6008 ret = 1;
6009#ifdef __FreeBSD__ /* MULTIPATH_ROUTING */
6010 rn = NULL;
6011#else
6012 rn = rn_mpath_next(rn);
6013#endif
6014 } while (check_mpath == 1 && rn != NULL && ret == 0);
6015 } else
6016 ret = 0;
6017out:
6018 if (ro.ro_rt != NULL)
6019 RTFREE(ro.ro_rt);
6020 return (ret);
6021}
6022
6023int
6024pf_rtlabel_match(struct pf_addr *addr, sa_family_t af, struct pf_addr_wrap *aw)
6025{
6026 struct sockaddr_in *dst;
6027#ifdef INET6
6028 struct sockaddr_in6 *dst6;
6029 struct route_in6 ro;
6030#else
6031 struct route ro;
6032#endif
6033 int ret = 0;
6034
6035 bzero(&ro, sizeof(ro));
6036 switch (af) {
6037 case AF_INET:
6038 dst = satosin(&ro.ro_dst);
6039 dst->sin_family = AF_INET;
6040 dst->sin_len = sizeof(*dst);
6041 dst->sin_addr = addr->v4;
6042 break;
6043#ifdef INET6
6044 case AF_INET6:
6045 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
6046 dst6->sin6_family = AF_INET6;
6047 dst6->sin6_len = sizeof(*dst6);
6048 dst6->sin6_addr = addr->v6;
6049 break;
6050#endif /* INET6 */
6051 default:
6052 return (0);
6053 }
6054
6055#ifdef __FreeBSD__
6056# ifdef RTF_PRCLONING
6057 rtalloc_ign((struct route *)&ro, (RTF_CLONING|RTF_PRCLONING));
6058# else /* !RTF_PRCLONING */
6059 if (af == AF_INET)
6060 in_rtalloc_ign((struct route *)&ro, 0, 0);
6061 else
6062 rtalloc_ign((struct route *)&ro, 0);
6063# endif
6064#else /* ! __FreeBSD__ */
6065 rtalloc_noclone((struct route *)&ro, NO_CLONING);
6066#endif
6067
6068 if (ro.ro_rt != NULL) {
6069#ifdef __FreeBSD__
6070 /* XXX_IMPORT: later */
6071#else
6072 if (ro.ro_rt->rt_labelid == aw->v.rtlabel)
6073 ret = 1;
6074#endif
6075 RTFREE(ro.ro_rt);
6076 }
6077
6078 return (ret);
6079}
6080
6081#ifdef INET
6082
6083void
6084pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
6085 struct pf_state *s, struct pf_pdesc *pd)
6086{
6087 struct mbuf *m0, *m1;
6088 struct route iproute;
6089 struct route *ro = NULL;
6090 struct sockaddr_in *dst;
6091 struct ip *ip;
6092 struct ifnet *ifp = NULL;
6093 struct pf_addr naddr;
6094 struct pf_src_node *sn = NULL;
6095 int error = 0;
6096#ifdef __FreeBSD__
6097 int sw_csum;
6098#endif
6099#ifdef IPSEC
6100 struct m_tag *mtag;
6101#endif /* IPSEC */
6102
6103 if (m == NULL || *m == NULL || r == NULL ||
6104 (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
6105 panic("pf_route: invalid parameters");
6106
6107 if (pd->pf_mtag->routed++ > 3) {
6108 m0 = *m;
6109 *m = NULL;
6110 goto bad;
6111 }
6112
6113 if (r->rt == PF_DUPTO) {
6114#ifdef __FreeBSD__
6115 if ((m0 = m_dup(*m, M_DONTWAIT)) == NULL)
6116#else
6117 if ((m0 = m_copym2(*m, 0, M_COPYALL, M_NOWAIT)) == NULL)
6118#endif
6119 return;
6120 } else {
6121 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
6122 return;
6123 m0 = *m;
6124 }
6125
6126 if (m0->m_len < sizeof(struct ip)) {
6127 DPFPRINTF(PF_DEBUG_URGENT,
6128 ("pf_route: m0->m_len < sizeof(struct ip)\n"));
6129 goto bad;
6130 }
6131
6132 ip = mtod(m0, struct ip *);
6133
6134 ro = &iproute;
6135 bzero((caddr_t)ro, sizeof(*ro));
6136 dst = satosin(&ro->ro_dst);
6137 dst->sin_family = AF_INET;
6138 dst->sin_len = sizeof(*dst);
6139 dst->sin_addr = ip->ip_dst;
6140
6141 if (r->rt == PF_FASTROUTE) {
6142 in_rtalloc(ro, 0);
6143 if (ro->ro_rt == 0) {
|