Context Navigation

source: rtems-libbsd/freebsd/sys/contrib/pf/net/pf.c @ 2017a6d

55-freebsd-126-freebsd-12freebsd-9.3

Last change on this file since 2017a6d was 2017a6d, checked in by Sebastian Huber <sebastian.huber@…>, on 04/07/16 at 07:48:12
Directly use <sys/time.h> provided by Newlib
Property mode set to `100644`
File size: 185.7 KB

Line
1	#include <machine/rtems-bsd-kernel-space.h>
2
3	/* $OpenBSD: pf.c,v 1.634 2009/02/27 12:37:45 henning Exp $ */
4
5	/*
6	* Copyright (c) 2001 Daniel Hartmeier
7	* Copyright (c) 2002 - 2008 Henning Brauer
8	* All rights reserved.
9	*
10	* Redistribution and use in source and binary forms, with or without
11	* modification, are permitted provided that the following conditions
12	* are met:
13	*
14	* - Redistributions of source code must retain the above copyright
15	* notice, this list of conditions and the following disclaimer.
16	* - Redistributions in binary form must reproduce the above
17	* copyright notice, this list of conditions and the following
18	* disclaimer in the documentation and/or other materials provided
19	* with the distribution.
20	*
21	* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22	* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23	* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24	* FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
25	* COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26	* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
27	* BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28	* LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
29	* CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30	* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
31	* ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
32	* POSSIBILITY OF SUCH DAMAGE.
33	*
34	* Effort sponsored in part by the Defense Advanced Research Projects
35	* Agency (DARPA) and Air Force Research Laboratory, Air Force
36	* Materiel Command, USAF, under agreement number F30602-01-2-0537.
37	*
38	*/
39
40	#ifdef __FreeBSD__
41	#include <rtems/bsd/local/opt_inet.h>
42	#include <rtems/bsd/local/opt_inet6.h>
43
44	#include <sys/cdefs.h>
45	__FBSDID("$FreeBSD$");
46	#endif
47
48	#ifdef __FreeBSD__
49	#include <rtems/bsd/local/opt_bpf.h>
50	#include <rtems/bsd/local/opt_pf.h>
51
52	#define NPFSYNC 1
53
54	#ifdef DEV_PFLOW
55	#define NPFLOW DEV_PFLOW
56	#else
57	#define NPFLOW 0
58	#endif
59
60	#else
61	#include "bpfilter.h"
62	#include "pflog.h"
63	#include "pfsync.h"
64	#include "pflow.h"
65	#endif
66
67	#include <rtems/bsd/sys/param.h>
68	#include <sys/systm.h>
69	#include <sys/mbuf.h>
70	#include <sys/filio.h>
71	#include <sys/socket.h>
72	#include <sys/socketvar.h>
73	#include <sys/kernel.h>
74	#include <sys/time.h>
75	#ifdef __FreeBSD__
76	#include <sys/random.h>
77	#include <sys/sysctl.h>
78	#include <sys/endian.h>
79	#define betoh64 be64toh
80	#else
81	#include <sys/pool.h>
82	#endif
83	#include <sys/proc.h>
84	#ifdef __FreeBSD__
85	#include <sys/kthread.h>
86	#include <rtems/bsd/sys/lock.h>
87	#include <sys/sx.h>
88	#else
89	#include <sys/rwlock.h>
90	#endif
91
92	#ifdef __FreeBSD__
93	#include <sys/md5.h>
94	#else
95	#include <crypto/md5.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	#ifdef __FreeBSD__
103	#ifdef RADIX_MPATH
104	#include <net/radix_mpath.h>
105	#endif
106	#else
107	#include <net/radix_mpath.h>
108	#endif
109
110	#include <netinet/in.h>
111	#include <netinet/in_var.h>
112	#include <netinet/in_systm.h>
113	#include <netinet/ip.h>
114	#include <netinet/ip_var.h>
115	#include <netinet/tcp.h>
116	#include <netinet/tcp_seq.h>
117	#include <netinet/udp.h>
118	#include <netinet/ip_icmp.h>
119	#include <netinet/in_pcb.h>
120	#include <netinet/tcp_timer.h>
121	#include <netinet/tcp_var.h>
122	#include <netinet/udp_var.h>
123	#include <netinet/icmp_var.h>
124	#include <netinet/if_ether.h>
125	#ifdef __FreeBSD__
126	#include <netinet/ip_fw.h>
127	#include <netpfil/ipfw/ip_fw_private.h> /* XXX: only for DIR_IN/DIR_OUT */
128	#endif
129
130	#ifndef __FreeBSD__
131	#include <dev/rndvar.h>
132	#endif
133	#include <net/pfvar.h>
134	#include <net/if_pflog.h>
135	#include <net/if_pflow.h>
136	#include <net/if_pfsync.h>
137
138	#ifdef INET6
139	#include <netinet/ip6.h>
140	#include <netinet/in_pcb.h>
141	#include <netinet/icmp6.h>
142	#include <netinet6/nd6.h>
143	#ifdef __FreeBSD__
144	#include <netinet6/ip6_var.h>
145	#include <netinet6/in6_pcb.h>
146	#endif
147	#endif /* INET6 */
148
149	#ifdef __FreeBSD__
150	#include <machine/in_cksum.h>
151	#include <sys/limits.h>
152	#include <sys/ucred.h>
153	#include <security/mac/mac_framework.h>
154
155	extern int ip_optcopy(struct ip , struct ip );
156	#endif
157
158	#ifdef __FreeBSD__
159	#define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
160	#else
161	#define DPFPRINTF(n, x) if (pf_status.debug >= (n)) printf x
162	#endif
163
164	/*
165	* Global variables
166	*/
167
168	/* state tables */
169	#ifdef __FreeBSD__
170	VNET_DEFINE(struct pf_state_tree, pf_statetbl);
171
172	VNET_DEFINE(struct pf_altqqueue, pf_altqs[2]);
173	VNET_DEFINE(struct pf_palist, pf_pabuf);
174	VNET_DEFINE(struct pf_altqqueue *, pf_altqs_active);
175	VNET_DEFINE(struct pf_altqqueue *, pf_altqs_inactive);
176	VNET_DEFINE(struct pf_status, pf_status);
177
178	VNET_DEFINE(u_int32_t, ticket_altqs_active);
179	VNET_DEFINE(u_int32_t, ticket_altqs_inactive);
180	VNET_DEFINE(int, altqs_inactive_open);
181	VNET_DEFINE(u_int32_t, ticket_pabuf);
182
183	VNET_DEFINE(MD5_CTX, pf_tcp_secret_ctx);
184	#define V_pf_tcp_secret_ctx VNET(pf_tcp_secret_ctx)
185	VNET_DEFINE(u_char, pf_tcp_secret[16]);
186	#define V_pf_tcp_secret VNET(pf_tcp_secret)
187	VNET_DEFINE(int, pf_tcp_secret_init);
188	#define V_pf_tcp_secret_init VNET(pf_tcp_secret_init)
189	VNET_DEFINE(int, pf_tcp_iss_off);
190	#define V_pf_tcp_iss_off VNET(pf_tcp_iss_off)
191
192	struct pf_anchor_stackframe {
193	struct pf_ruleset *rs;
194	struct pf_rule *r;
195	struct pf_anchor_node *parent;
196	struct pf_anchor *child;
197	};
198	VNET_DEFINE(struct pf_anchor_stackframe, pf_anchor_stack[64]);
199	#define V_pf_anchor_stack VNET(pf_anchor_stack)
200
201	VNET_DEFINE(uma_zone_t, pf_src_tree_pl);
202	VNET_DEFINE(uma_zone_t, pf_rule_pl);
203	VNET_DEFINE(uma_zone_t, pf_pooladdr_pl);
204	VNET_DEFINE(uma_zone_t, pf_state_pl);
205	VNET_DEFINE(uma_zone_t, pf_state_key_pl);
206	VNET_DEFINE(uma_zone_t, pf_state_item_pl);
207	VNET_DEFINE(uma_zone_t, pf_altq_pl);
208	#else
209	struct pf_state_tree pf_statetbl;
210
211	struct pf_altqqueue pf_altqs[2];
212	struct pf_palist pf_pabuf;
213	struct pf_altqqueue *pf_altqs_active;
214	struct pf_altqqueue *pf_altqs_inactive;
215	struct pf_status pf_status;
216
217	u_int32_t ticket_altqs_active;
218	u_int32_t ticket_altqs_inactive;
219	int altqs_inactive_open;
220	u_int32_t ticket_pabuf;
221
222	MD5_CTX pf_tcp_secret_ctx;
223	u_char pf_tcp_secret[16];
224	int pf_tcp_secret_init;
225	int pf_tcp_iss_off;
226
227	struct pf_anchor_stackframe {
228	struct pf_ruleset *rs;
229	struct pf_rule *r;
230	struct pf_anchor_node *parent;
231	struct pf_anchor *child;
232	} pf_anchor_stack[64];
233
234	struct pool pf_src_tree_pl, pf_rule_pl, pf_pooladdr_pl;
235	struct pool pf_state_pl, pf_state_key_pl, pf_state_item_pl;
236	struct pool pf_altq_pl;
237	#endif
238
239	void pf_init_threshold(struct pf_threshold *, u_int32_t,
240	u_int32_t);
241	void pf_add_threshold(struct pf_threshold *);
242	int pf_check_threshold(struct pf_threshold *);
243
244	void pf_change_ap(struct pf_addr , u_int16_t ,
245	u_int16_t , u_int16_t , struct pf_addr *,
246	u_int16_t, u_int8_t, sa_family_t);
247	int pf_modulate_sack(struct mbuf , int, struct pf_pdesc ,
248	struct tcphdr , struct pf_state_peer );
249	#ifdef INET6
250	void pf_change_a6(struct pf_addr , u_int16_t ,
251	struct pf_addr *, u_int8_t);
252	#endif /* INET6 */
253	void pf_change_icmp(struct pf_addr , u_int16_t ,
254	struct pf_addr , struct pf_addr , u_int16_t,
255	u_int16_t , u_int16_t , u_int16_t *,
256	u_int16_t *, u_int8_t, sa_family_t);
257	#ifdef __FreeBSD__
258	void pf_send_tcp(struct mbuf *,
259	const struct pf_rule *, sa_family_t,
260	#else
261	void pf_send_tcp(const struct pf_rule *, sa_family_t,
262	#endif
263	const struct pf_addr , const struct pf_addr ,
264	u_int16_t, u_int16_t, u_int32_t, u_int32_t,
265	u_int8_t, u_int16_t, u_int16_t, u_int8_t, int,
266	u_int16_t, struct ether_header , struct ifnet );
267	static void pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t,
268	sa_family_t, struct pf_rule *);
269	void pf_detach_state(struct pf_state *);
270	void pf_state_key_detach(struct pf_state *, int);
271	u_int32_t pf_tcp_iss(struct pf_pdesc *);
272	int pf_test_rule(struct pf_rule , struct pf_state ,
273	int, struct pfi_kif , struct mbuf , int,
274	void , struct pf_pdesc , struct pf_rule **,
275	#ifdef __FreeBSD__
276	struct pf_ruleset *, struct ifqueue ,
277	struct inpcb *);
278	#else
279	struct pf_ruleset *, struct ifqueue );
280	#endif
281	static __inline int pf_create_state(struct pf_rule , struct pf_rule ,
282	struct pf_rule , struct pf_pdesc ,
283	struct pf_src_node , struct pf_state_key ,
284	struct pf_state_key , struct pf_state_key ,
285	struct pf_state_key , struct mbuf , int,
286	u_int16_t, u_int16_t, int , struct pfi_kif ,
287	struct pf_state **, int, u_int16_t, u_int16_t,
288	int);
289	int pf_test_fragment(struct pf_rule **, int,
290	struct pfi_kif , struct mbuf , void *,
291	struct pf_pdesc , struct pf_rule *,
292	struct pf_ruleset **);
293	int pf_tcp_track_full(struct pf_state_peer *,
294	struct pf_state_peer , struct pf_state *,
295	struct pfi_kif , struct mbuf , int,
296	struct pf_pdesc , u_short , int *);
297	int pf_tcp_track_sloppy(struct pf_state_peer *,
298	struct pf_state_peer , struct pf_state *,
299	struct pf_pdesc , u_short );
300	int pf_test_state_tcp(struct pf_state **, int,
301	struct pfi_kif , struct mbuf , int,
302	void , struct pf_pdesc , u_short *);
303	int pf_test_state_udp(struct pf_state **, int,
304	struct pfi_kif , struct mbuf , int,
305	void , struct pf_pdesc );
306	int pf_test_state_icmp(struct pf_state **, int,
307	struct pfi_kif , struct mbuf , int,
308	void , struct pf_pdesc , u_short *);
309	int pf_test_state_other(struct pf_state **, int,
310	struct pfi_kif , struct mbuf , struct pf_pdesc *);
311	void pf_route(struct mbuf *, struct pf_rule , int,
312	struct ifnet , struct pf_state ,
313	struct pf_pdesc *);
314	void pf_route6(struct mbuf *, struct pf_rule , int,
315	struct ifnet , struct pf_state ,
316	struct pf_pdesc *);
317	#ifndef __FreeBSD__
318	int pf_socket_lookup(int, struct pf_pdesc *);
319	#endif
320	u_int8_t pf_get_wscale(struct mbuf *, int, u_int16_t,
321	sa_family_t);
322	u_int16_t pf_get_mss(struct mbuf *, int, u_int16_t,
323	sa_family_t);
324	u_int16_t pf_calc_mss(struct pf_addr *, sa_family_t,
325	int, u_int16_t);
326	void pf_set_rt_ifp(struct pf_state *,
327	struct pf_addr *);
328	int pf_check_proto_cksum(struct mbuf *, int, int,
329	u_int8_t, sa_family_t);
330	#ifndef __FreeBSD__
331	struct pf_divert pf_get_divert(struct mbuf );
332	#endif
333	void pf_print_state_parts(struct pf_state *,
334	struct pf_state_key , struct pf_state_key );
335	int pf_addr_wrap_neq(struct pf_addr_wrap *,
336	struct pf_addr_wrap *);
337	int pf_compare_state_keys(struct pf_state_key *,
338	struct pf_state_key , struct pfi_kif , u_int);
339	#ifdef __FreeBSD__
340	struct pf_state pf_find_state(struct pfi_kif ,
341	struct pf_state_key_cmp , u_int, struct mbuf ,
342	struct pf_mtag *);
343	#else
344	struct pf_state pf_find_state(struct pfi_kif ,
345	struct pf_state_key_cmp , u_int, struct mbuf );
346	#endif
347	int pf_src_connlimit(struct pf_state **);
348	int pf_check_congestion(struct ifqueue *);
349
350	#ifdef __FreeBSD__
351	int in4_cksum(struct mbuf *m, u_int8_t nxt, int off, int len);
352
353	VNET_DECLARE(int, pf_end_threads);
354
355	VNET_DEFINE(struct pf_pool_limit, pf_pool_limits[PF_LIMIT_MAX]);
356	#else
357	extern struct pool pfr_ktable_pl;
358	extern struct pool pfr_kentry_pl;
359
360	struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] = {
361	{ &pf_state_pl, PFSTATE_HIWAT },
362	{ &pf_src_tree_pl, PFSNODE_HIWAT },
363	{ &pf_frent_pl, PFFRAG_FRENT_HIWAT },
364	{ &pfr_ktable_pl, PFR_KTABLE_HIWAT },
365	{ &pfr_kentry_pl, PFR_KENTRY_HIWAT }
366	};
367	#endif
368
369	#ifdef __FreeBSD__
370	#define PPACKET_LOOPED() \
371	(pd->pf_mtag->flags & PF_PACKET_LOOPED)
372
373	#define PACKET_LOOPED() \
374	(pd.pf_mtag->flags & PF_PACKET_LOOPED)
375
376	#define STATE_LOOKUP(i, k, d, s, m, pt) \
377	do { \
378	s = pf_find_state(i, k, d, m, pt); \
379	if (s == NULL \|\| (s)->timeout == PFTM_PURGE) \
380	return (PF_DROP); \
381	if (PPACKET_LOOPED()) \
382	return (PF_PASS); \
383	if (d == PF_OUT && \
384	(((s)->rule.ptr->rt == PF_ROUTETO && \
385	(s)->rule.ptr->direction == PF_OUT) \|\| \
386	((s)->rule.ptr->rt == PF_REPLYTO && \
387	(s)->rule.ptr->direction == PF_IN)) && \
388	(s)->rt_kif != NULL && \
389	(s)->rt_kif != i) \
390	return (PF_PASS); \
391	} while (0)
392	#else
393	#define STATE_LOOKUP(i, k, d, s, m) \
394	do { \
395	s = pf_find_state(i, k, d, m); \
396	if (s == NULL \|\| (s)->timeout == PFTM_PURGE) \
397	return (PF_DROP); \
398	if (d == PF_OUT && \
399	(((s)->rule.ptr->rt == PF_ROUTETO && \
400	(s)->rule.ptr->direction == PF_OUT) \|\| \
401	((s)->rule.ptr->rt == PF_REPLYTO && \
402	(s)->rule.ptr->direction == PF_IN)) && \
403	(s)->rt_kif != NULL && \
404	(s)->rt_kif != i) \
405	return (PF_PASS); \
406	} while (0)
407	#endif
408
409	#ifdef __FreeBSD__
410	#define BOUND_IFACE(r, k) \
411	((r)->rule_flag & PFRULE_IFBOUND) ? (k) : V_pfi_all
412	#else
413	#define BOUND_IFACE(r, k) \
414	((r)->rule_flag & PFRULE_IFBOUND) ? (k) : pfi_all
415	#endif
416
417	#define STATE_INC_COUNTERS(s) \
418	do { \
419	s->rule.ptr->states_cur++; \
420	s->rule.ptr->states_tot++; \
421	if (s->anchor.ptr != NULL) { \
422	s->anchor.ptr->states_cur++; \
423	s->anchor.ptr->states_tot++; \
424	} \
425	if (s->nat_rule.ptr != NULL) { \
426	s->nat_rule.ptr->states_cur++; \
427	s->nat_rule.ptr->states_tot++; \
428	} \
429	} while (0)
430
431	#define STATE_DEC_COUNTERS(s) \
432	do { \
433	if (s->nat_rule.ptr != NULL) \
434	s->nat_rule.ptr->states_cur--; \
435	if (s->anchor.ptr != NULL) \
436	s->anchor.ptr->states_cur--; \
437	s->rule.ptr->states_cur--; \
438	} while (0)
439
440	static __inline int pf_src_compare(struct pf_src_node , struct pf_src_node );
441	static __inline int pf_state_compare_key(struct pf_state_key *,
442	struct pf_state_key *);
443	static __inline int pf_state_compare_id(struct pf_state *,
444	struct pf_state *);
445
446	#ifdef __FreeBSD__
447	VNET_DEFINE(struct pf_src_tree, tree_src_tracking);
448
449	VNET_DEFINE(struct pf_state_tree_id, tree_id);
450	VNET_DEFINE(struct pf_state_queue, state_list);
451	#else
452	struct pf_src_tree tree_src_tracking;
453
454	struct pf_state_tree_id tree_id;
455	struct pf_state_queue state_list;
456	#endif
457
458	RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare);
459	RB_GENERATE(pf_state_tree, pf_state_key, entry, pf_state_compare_key);
460	RB_GENERATE(pf_state_tree_id, pf_state,
461	entry_id, pf_state_compare_id);
462
463	static __inline int
464	pf_src_compare(struct pf_src_node a, struct pf_src_node b)
465	{
466	int diff;
467
468	if (a->rule.ptr > b->rule.ptr)
469	return (1);
470	if (a->rule.ptr < b->rule.ptr)
471	return (-1);
472	if ((diff = a->af - b->af) != 0)
473	return (diff);
474	switch (a->af) {
475	#ifdef INET
476	case AF_INET:
477	if (a->addr.addr32[0] > b->addr.addr32[0])
478	return (1);
479	if (a->addr.addr32[0] < b->addr.addr32[0])
480	return (-1);
481	break;
482	#endif /* INET */
483	#ifdef INET6
484	case AF_INET6:
485	if (a->addr.addr32[3] > b->addr.addr32[3])
486	return (1);
487	if (a->addr.addr32[3] < b->addr.addr32[3])
488	return (-1);
489	if (a->addr.addr32[2] > b->addr.addr32[2])
490	return (1);
491	if (a->addr.addr32[2] < b->addr.addr32[2])
492	return (-1);
493	if (a->addr.addr32[1] > b->addr.addr32[1])
494	return (1);
495	if (a->addr.addr32[1] < b->addr.addr32[1])
496	return (-1);
497	if (a->addr.addr32[0] > b->addr.addr32[0])
498	return (1);
499	if (a->addr.addr32[0] < b->addr.addr32[0])
500	return (-1);
501	break;
502	#endif /* INET6 */
503	}
504	return (0);
505	}
506
507	#ifdef INET6
508	void
509	pf_addrcpy(struct pf_addr dst, struct pf_addr src, sa_family_t af)
510	{
511	switch (af) {
512	#ifdef INET
513	case AF_INET:
514	dst->addr32[0] = src->addr32[0];
515	break;
516	#endif /* INET */
517	case AF_INET6:
518	dst->addr32[0] = src->addr32[0];
519	dst->addr32[1] = src->addr32[1];
520	dst->addr32[2] = src->addr32[2];
521	dst->addr32[3] = src->addr32[3];
522	break;
523	}
524	}
525	#endif /* INET6 */
526
527	void
528	pf_init_threshold(struct pf_threshold *threshold,
529	u_int32_t limit, u_int32_t seconds)
530	{
531	threshold->limit = limit * PF_THRESHOLD_MULT;
532	threshold->seconds = seconds;
533	threshold->count = 0;
534	threshold->last = time_second;
535	}
536
537	void
538	pf_add_threshold(struct pf_threshold *threshold)
539	{
540	u_int32_t t = time_second, diff = t - threshold->last;
541
542	if (diff >= threshold->seconds)
543	threshold->count = 0;
544	else
545	threshold->count -= threshold->count * diff /
546	threshold->seconds;
547	threshold->count += PF_THRESHOLD_MULT;
548	threshold->last = t;
549	}
550
551	int
552	pf_check_threshold(struct pf_threshold *threshold)
553	{
554	return (threshold->count > threshold->limit);
555	}
556
557	int
558	pf_src_connlimit(struct pf_state **state)
559	{
560	int bad = 0;
561
562	(*state)->src_node->conn++;
563	(*state)->src.tcp_est = 1;
564	pf_add_threshold(&(*state)->src_node->conn_rate);
565
566	if ((*state)->rule.ptr->max_src_conn &&
567	(*state)->rule.ptr->max_src_conn <
568	(*state)->src_node->conn) {
569	#ifdef __FreeBSD__
570	V_pf_status.lcounters[LCNT_SRCCONN]++;
571	#else
572	pf_status.lcounters[LCNT_SRCCONN]++;
573	#endif
574	bad++;
575	}
576
577	if ((*state)->rule.ptr->max_src_conn_rate.limit &&
578	pf_check_threshold(&(*state)->src_node->conn_rate)) {
579	#ifdef __FreeBSD__
580	V_pf_status.lcounters[LCNT_SRCCONNRATE]++;
581	#else
582	pf_status.lcounters[LCNT_SRCCONNRATE]++;
583	#endif
584	bad++;
585	}
586
587	if (!bad)
588	return (0);
589
590	if ((*state)->rule.ptr->overload_tbl) {
591	struct pfr_addr p;
592	u_int32_t killed = 0;
593
594	#ifdef __FreeBSD__
595	V_pf_status.lcounters[LCNT_OVERLOAD_TABLE]++;
596	if (V_pf_status.debug >= PF_DEBUG_MISC) {
597	#else
598	pf_status.lcounters[LCNT_OVERLOAD_TABLE]++;
599	if (pf_status.debug >= PF_DEBUG_MISC) {
600	#endif
601	printf("pf_src_connlimit: blocking address ");
602	pf_print_host(&(*state)->src_node->addr, 0,
603	(*state)->key[PF_SK_WIRE]->af);
604	}
605
606	bzero(&p, sizeof(p));
607	p.pfra_af = (*state)->key[PF_SK_WIRE]->af;
608	switch ((*state)->key[PF_SK_WIRE]->af) {
609	#ifdef INET
610	case AF_INET:
611	p.pfra_net = 32;
612	p.pfra_ip4addr = (*state)->src_node->addr.v4;
613	break;
614	#endif /* INET */
615	#ifdef INET6
616	case AF_INET6:
617	p.pfra_net = 128;
618	p.pfra_ip6addr = (*state)->src_node->addr.v6;
619	break;
620	#endif /* INET6 */
621	}
622
623	pfr_insert_kentry((*state)->rule.ptr->overload_tbl,
624	&p, time_second);
625
626	/* kill existing states if that's required. */
627	if ((*state)->rule.ptr->flush) {
628	struct pf_state_key *sk;
629	struct pf_state *st;
630
631	#ifdef __FreeBSD__
632	V_pf_status.lcounters[LCNT_OVERLOAD_FLUSH]++;
633	RB_FOREACH(st, pf_state_tree_id, &V_tree_id) {
634	#else
635	pf_status.lcounters[LCNT_OVERLOAD_FLUSH]++;
636	RB_FOREACH(st, pf_state_tree_id, &tree_id) {
637	#endif
638	sk = st->key[PF_SK_WIRE];
639	/*
640	* Kill states from this source. (Only those
641	* from the same rule if PF_FLUSH_GLOBAL is not
642	* set)
643	*/
644	if (sk->af ==
645	(*state)->key[PF_SK_WIRE]->af &&
646	(((*state)->direction == PF_OUT &&
647	PF_AEQ(&(*state)->src_node->addr,
648	&sk->addr[1], sk->af)) \|\|
649	((*state)->direction == PF_IN &&
650	PF_AEQ(&(*state)->src_node->addr,
651	&sk->addr[0], sk->af))) &&
652	((*state)->rule.ptr->flush &
653	PF_FLUSH_GLOBAL \|\|
654	(*state)->rule.ptr == st->rule.ptr)) {
655	st->timeout = PFTM_PURGE;
656	st->src.state = st->dst.state =
657	TCPS_CLOSED;
658	killed++;
659	}
660	}
661	#ifdef __FreeBSD__
662	if (V_pf_status.debug >= PF_DEBUG_MISC)
663	#else
664	if (pf_status.debug >= PF_DEBUG_MISC)
665	#endif
666	printf(", %u states killed", killed);
667	}
668	#ifdef __FreeBSD__
669	if (V_pf_status.debug >= PF_DEBUG_MISC)
670	#else
671	if (pf_status.debug >= PF_DEBUG_MISC)
672	#endif
673	printf("\n");
674	}
675
676	/* kill this state */
677	(*state)->timeout = PFTM_PURGE;
678	(state)->src.state = (state)->dst.state = TCPS_CLOSED;
679	return (1);
680	}
681
682	int
683	pf_insert_src_node(struct pf_src_node *sn, struct pf_rule rule,
684	struct pf_addr *src, sa_family_t af)
685	{
686	struct pf_src_node k;
687
688	if (*sn == NULL) {
689	k.af = af;
690	PF_ACPY(&k.addr, src, af);
691	if (rule->rule_flag & PFRULE_RULESRCTRACK \|\|
692	rule->rpool.opts & PF_POOL_STICKYADDR)
693	k.rule.ptr = rule;
694	else
695	k.rule.ptr = NULL;
696	#ifdef __FreeBSD__
697	V_pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
698	*sn = RB_FIND(pf_src_tree, &V_tree_src_tracking, &k);
699	#else
700	pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
701	*sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
702	#endif
703	}
704	if (*sn == NULL) {
705	if (!rule->max_src_nodes \|\|
706	rule->src_nodes < rule->max_src_nodes)
707	#ifdef __FreeBSD__
708	(*sn) = pool_get(&V_pf_src_tree_pl, PR_NOWAIT \| PR_ZERO);
709	#else
710	(*sn) = pool_get(&pf_src_tree_pl, PR_NOWAIT \| PR_ZERO);
711	#endif
712	else
713	#ifdef __FreeBSD__
714	V_pf_status.lcounters[LCNT_SRCNODES]++;
715	#else
716	pf_status.lcounters[LCNT_SRCNODES]++;
717	#endif
718	if ((*sn) == NULL)
719	return (-1);
720
721	pf_init_threshold(&(*sn)->conn_rate,
722	rule->max_src_conn_rate.limit,
723	rule->max_src_conn_rate.seconds);
724
725	(*sn)->af = af;
726	if (rule->rule_flag & PFRULE_RULESRCTRACK \|\|
727	rule->rpool.opts & PF_POOL_STICKYADDR)
728	(*sn)->rule.ptr = rule;
729	else
730	(*sn)->rule.ptr = NULL;
731	PF_ACPY(&(*sn)->addr, src, af);
732	if (RB_INSERT(pf_src_tree,
733	#ifdef __FreeBSD__
734	&V_tree_src_tracking, *sn) != NULL) {
735	if (V_pf_status.debug >= PF_DEBUG_MISC) {
736	#else
737	&tree_src_tracking, *sn) != NULL) {
738	if (pf_status.debug >= PF_DEBUG_MISC) {
739	#endif
740	printf("pf: src_tree insert failed: ");
741	pf_print_host(&(*sn)->addr, 0, af);
742	printf("\n");
743	}
744	#ifdef __FreeBSD__
745	pool_put(&V_pf_src_tree_pl, *sn);
746	#else
747	pool_put(&pf_src_tree_pl, *sn);
748	#endif
749	return (-1);
750	}
751	(*sn)->creation = time_second;
752	(*sn)->ruletype = rule->action;
753	if ((*sn)->rule.ptr != NULL)
754	(*sn)->rule.ptr->src_nodes++;
755	#ifdef __FreeBSD__
756	V_pf_status.scounters[SCNT_SRC_NODE_INSERT]++;
757	V_pf_status.src_nodes++;
758	#else
759	pf_status.scounters[SCNT_SRC_NODE_INSERT]++;
760	pf_status.src_nodes++;
761	#endif
762	} else {
763	if (rule->max_src_states &&
764	(*sn)->states >= rule->max_src_states) {
765	#ifdef __FreeBSD__
766	V_pf_status.lcounters[LCNT_SRCSTATES]++;
767	#else
768	pf_status.lcounters[LCNT_SRCSTATES]++;
769	#endif
770	return (-1);
771	}
772	}
773	return (0);
774	}
775
776	/* state table stuff */
777
778	static __inline int
779	pf_state_compare_key(struct pf_state_key a, struct pf_state_key b)
780	{
781	int diff;
782
783	if ((diff = a->proto - b->proto) != 0)
784	return (diff);
785	if ((diff = a->af - b->af) != 0)
786	return (diff);
787	switch (a->af) {
788	#ifdef INET
789	case AF_INET:
790	if (a->addr[0].addr32[0] > b->addr[0].addr32[0])
791	return (1);
792	if (a->addr[0].addr32[0] < b->addr[0].addr32[0])
793	return (-1);
794	if (a->addr[1].addr32[0] > b->addr[1].addr32[0])
795	return (1);
796	if (a->addr[1].addr32[0] < b->addr[1].addr32[0])
797	return (-1);
798	break;
799	#endif /* INET */
800	#ifdef INET6
801	case AF_INET6:
802	if (a->addr[0].addr32[3] > b->addr[0].addr32[3])
803	return (1);
804	if (a->addr[0].addr32[3] < b->addr[0].addr32[3])
805	return (-1);
806	if (a->addr[1].addr32[3] > b->addr[1].addr32[3])
807	return (1);
808	if (a->addr[1].addr32[3] < b->addr[1].addr32[3])
809	return (-1);
810	if (a->addr[0].addr32[2] > b->addr[0].addr32[2])
811	return (1);
812	if (a->addr[0].addr32[2] < b->addr[0].addr32[2])
813	return (-1);
814	if (a->addr[1].addr32[2] > b->addr[1].addr32[2])
815	return (1);
816	if (a->addr[1].addr32[2] < b->addr[1].addr32[2])
817	return (-1);
818	if (a->addr[0].addr32[1] > b->addr[0].addr32[1])
819	return (1);
820	if (a->addr[0].addr32[1] < b->addr[0].addr32[1])
821	return (-1);
822	if (a->addr[1].addr32[1] > b->addr[1].addr32[1])
823	return (1);
824	if (a->addr[1].addr32[1] < b->addr[1].addr32[1])
825	return (-1);
826	if (a->addr[0].addr32[0] > b->addr[0].addr32[0])
827	return (1);
828	if (a->addr[0].addr32[0] < b->addr[0].addr32[0])
829	return (-1);
830	if (a->addr[1].addr32[0] > b->addr[1].addr32[0])
831	return (1);
832	if (a->addr[1].addr32[0] < b->addr[1].addr32[0])
833	return (-1);
834	break;
835	#endif /* INET6 */
836	}
837
838	if ((diff = a->port[0] - b->port[0]) != 0)
839	return (diff);
840	if ((diff = a->port[1] - b->port[1]) != 0)
841	return (diff);
842
843	return (0);
844	}
845
846	static __inline int
847	pf_state_compare_id(struct pf_state a, struct pf_state b)
848	{
849	if (a->id > b->id)
850	return (1);
851	if (a->id < b->id)
852	return (-1);
853	if (a->creatorid > b->creatorid)
854	return (1);
855	if (a->creatorid < b->creatorid)
856	return (-1);
857
858	return (0);
859	}
860
861	int
862	pf_state_key_attach(struct pf_state_key sk, struct pf_state s, int idx)
863	{
864	struct pf_state_item *si;
865	struct pf_state_key *cur;
866	struct pf_state *olds = NULL;
867
868	#ifdef __FreeBSD__
869	KASSERT(s->key[idx] == NULL, ("%s: key is null!", __FUNCTION__));
870	#else
871	KASSERT(s->key[idx] == NULL); /* XXX handle this? */
872	#endif
873
874	#ifdef __FreeBSD__
875	if ((cur = RB_INSERT(pf_state_tree, &V_pf_statetbl, sk)) != NULL) {
876	#else
877	if ((cur = RB_INSERT(pf_state_tree, &pf_statetbl, sk)) != NULL) {
878	#endif
879	/* key exists. check for same kif, if none, add to key */
880	TAILQ_FOREACH(si, &cur->states, entry)
881	if (si->s->kif == s->kif &&
882	si->s->direction == s->direction) {
883	if (sk->proto == IPPROTO_TCP &&
884	si->s->src.state >= TCPS_FIN_WAIT_2 &&
885	si->s->dst.state >= TCPS_FIN_WAIT_2) {
886	si->s->src.state = si->s->dst.state =
887	TCPS_CLOSED;
888	/* unlink late or sks can go away */
889	olds = si->s;
890	} else {
891	#ifdef __FreeBSD__
892	if (V_pf_status.debug >= PF_DEBUG_MISC) {
893	#else
894	if (pf_status.debug >= PF_DEBUG_MISC) {
895	#endif
896	printf("pf: %s key attach "
897	"failed on %s: ",
898	(idx == PF_SK_WIRE) ?
899	"wire" : "stack",
900	s->kif->pfik_name);
901	pf_print_state_parts(s,
902	(idx == PF_SK_WIRE) ?
903	sk : NULL,
904	(idx == PF_SK_STACK) ?
905	sk : NULL);
906	printf(", existing: ");
907	pf_print_state_parts(si->s,
908	(idx == PF_SK_WIRE) ?
909	sk : NULL,
910	(idx == PF_SK_STACK) ?
911	sk : NULL);
912	printf("\n");
913	}
914	#ifdef __FreeBSD__
915	pool_put(&V_pf_state_key_pl, sk);
916	#else
917	pool_put(&pf_state_key_pl, sk);
918	#endif
919	return (-1); /* collision! */
920	}
921	}
922	#ifdef __FreeBSD__
923	pool_put(&V_pf_state_key_pl, sk);
924	#else
925	pool_put(&pf_state_key_pl, sk);
926	#endif
927	s->key[idx] = cur;
928	} else
929	s->key[idx] = sk;
930
931	#ifdef __FreeBSD__
932	if ((si = pool_get(&V_pf_state_item_pl, PR_NOWAIT)) == NULL) {
933	#else
934	if ((si = pool_get(&pf_state_item_pl, PR_NOWAIT)) == NULL) {
935	#endif
936	pf_state_key_detach(s, idx);
937	return (-1);
938	}
939	si->s = s;
940
941	/* list is sorted, if-bound states before floating */
942	#ifdef __FreeBSD__
943	if (s->kif == V_pfi_all)
944	#else
945	if (s->kif == pfi_all)
946	#endif
947	TAILQ_INSERT_TAIL(&s->key[idx]->states, si, entry);
948	else
949	TAILQ_INSERT_HEAD(&s->key[idx]->states, si, entry);
950
951	if (olds)
952	pf_unlink_state(olds);
953
954	return (0);
955	}
956
957	void
958	pf_detach_state(struct pf_state *s)
959	{
960	if (s->key[PF_SK_WIRE] == s->key[PF_SK_STACK])
961	s->key[PF_SK_WIRE] = NULL;
962
963	if (s->key[PF_SK_STACK] != NULL)
964	pf_state_key_detach(s, PF_SK_STACK);
965
966	if (s->key[PF_SK_WIRE] != NULL)
967	pf_state_key_detach(s, PF_SK_WIRE);
968	}
969
970	void
971	pf_state_key_detach(struct pf_state *s, int idx)
972	{
973	struct pf_state_item *si;
974
975	si = TAILQ_FIRST(&s->key[idx]->states);
976	while (si && si->s != s)
977	si = TAILQ_NEXT(si, entry);
978
979	if (si) {
980	TAILQ_REMOVE(&s->key[idx]->states, si, entry);
981	#ifdef __FreeBSD__
982	pool_put(&V_pf_state_item_pl, si);
983	#else
984	pool_put(&pf_state_item_pl, si);
985	#endif
986	}
987
988	if (TAILQ_EMPTY(&s->key[idx]->states)) {
989	#ifdef __FreeBSD__
990	RB_REMOVE(pf_state_tree, &V_pf_statetbl, s->key[idx]);
991	#else
992	RB_REMOVE(pf_state_tree, &pf_statetbl, s->key[idx]);
993	#endif
994	if (s->key[idx]->reverse)
995	s->key[idx]->reverse->reverse = NULL;
996	#ifdef __FreeBSD__
997	/* XXX: implement this */
998	#else
999	if (s->key[idx]->inp)
1000	s->key[idx]->inp->inp_pf_sk = NULL;
1001	#endif
1002	#ifdef __FreeBSD__
1003	pool_put(&V_pf_state_key_pl, s->key[idx]);
1004	#else
1005	pool_put(&pf_state_key_pl, s->key[idx]);
1006	#endif
1007	}
1008	s->key[idx] = NULL;
1009	}
1010
1011	struct pf_state_key *
1012	pf_alloc_state_key(int pool_flags)
1013	{
1014	struct pf_state_key *sk;
1015
1016	#ifdef __FreeBSD__
1017	if ((sk = pool_get(&V_pf_state_key_pl, pool_flags)) == NULL)
1018	#else
1019	if ((sk = pool_get(&pf_state_key_pl, pool_flags)) == NULL)
1020	#endif
1021	return (NULL);
1022	TAILQ_INIT(&sk->states);
1023
1024	return (sk);
1025	}
1026
1027	int
1028	pf_state_key_setup(struct pf_pdesc pd, struct pf_rule nr,
1029	struct pf_state_key skw, struct pf_state_key sks,
1030	struct pf_state_key skp, struct pf_state_key nkp,
1031	struct pf_addr saddr, struct pf_addr daddr,
1032	u_int16_t sport, u_int16_t dport)
1033	{
1034	#ifdef __FreeBSD__
1035	KASSERT((skp == NULL && nkp == NULL),
1036	("%s: skp == NULL && nkp == NULL", __FUNCTION__));
1037	#else
1038	KASSERT((skp == NULL && nkp == NULL));
1039	#endif
1040
1041	if ((*skp = pf_alloc_state_key(PR_NOWAIT \| PR_ZERO)) == NULL)
1042	return (ENOMEM);
1043
1044	PF_ACPY(&(*skp)->addr[pd->sidx], saddr, pd->af);
1045	PF_ACPY(&(*skp)->addr[pd->didx], daddr, pd->af);
1046	(*skp)->port[pd->sidx] = sport;
1047	(*skp)->port[pd->didx] = dport;
1048	(*skp)->proto = pd->proto;
1049	(*skp)->af = pd->af;
1050
1051	if (nr != NULL) {
1052	if ((*nkp = pf_alloc_state_key(PR_NOWAIT \| PR_ZERO)) == NULL)
1053	return (ENOMEM); /* caller must handle cleanup */
1054
1055	/* XXX maybe just bcopy and TAILQ_INIT(&(nkp)->states) /
1056	PF_ACPY(&(nkp)->addr[0], &(skp)->addr[0], pd->af);
1057	PF_ACPY(&(nkp)->addr[1], &(skp)->addr[1], pd->af);
1058	(nkp)->port[0] = (skp)->port[0];
1059	(nkp)->port[1] = (skp)->port[1];
1060	(*nkp)->proto = pd->proto;
1061	(*nkp)->af = pd->af;
1062	} else
1063	nkp = skp;
1064
1065	if (pd->dir == PF_IN) {
1066	skw = skp;
1067	sks = nkp;
1068	} else {
1069	sks = skp;
1070	skw = nkp;
1071	}
1072	return (0);
1073	}
1074
1075
1076	int
1077	pf_state_insert(struct pfi_kif kif, struct pf_state_key skw,
1078	struct pf_state_key sks, struct pf_state s)
1079	{
1080	#ifndef __FreeBSD__
1081	splassert(IPL_SOFTNET);
1082	#endif
1083
1084	s->kif = kif;
1085
1086	if (skw == sks) {
1087	if (pf_state_key_attach(skw, s, PF_SK_WIRE))
1088	return (-1);
1089	s->key[PF_SK_STACK] = s->key[PF_SK_WIRE];
1090	} else {
1091	if (pf_state_key_attach(skw, s, PF_SK_WIRE)) {
1092	#ifdef __FreeBSD__
1093	pool_put(&V_pf_state_key_pl, sks);
1094	#else
1095	pool_put(&pf_state_key_pl, sks);
1096	#endif
1097	return (-1);
1098	}
1099	if (pf_state_key_attach(sks, s, PF_SK_STACK)) {
1100	pf_state_key_detach(s, PF_SK_WIRE);
1101	return (-1);
1102	}
1103	}
1104
1105	if (s->id == 0 && s->creatorid == 0) {
1106	#ifdef __FreeBSD__
1107	s->id = htobe64(V_pf_status.stateid++);
1108	s->creatorid = V_pf_status.hostid;
1109	#else
1110	s->id = htobe64(pf_status.stateid++);
1111	s->creatorid = pf_status.hostid;
1112	#endif
1113	}
1114	#ifdef __FreeBSD__
1115	if (RB_INSERT(pf_state_tree_id, &V_tree_id, s) != NULL) {
1116	if (V_pf_status.debug >= PF_DEBUG_MISC) {
1117	#else
1118	if (RB_INSERT(pf_state_tree_id, &tree_id, s) != NULL) {
1119	if (pf_status.debug >= PF_DEBUG_MISC) {
1120	#endif
1121	printf("pf: state insert failed: "
1122	"id: %016llx creatorid: %08x",
1123	#ifdef __FreeBSD__
1124	(unsigned long long)betoh64(s->id), ntohl(s->creatorid));
1125	#else
1126	betoh64(s->id), ntohl(s->creatorid));
1127	#endif
1128	printf("\n");
1129	}
1130	pf_detach_state(s);
1131	return (-1);
1132	}
1133	#ifdef __FreeBSD__
1134	TAILQ_INSERT_TAIL(&V_state_list, s, entry_list);
1135	V_pf_status.fcounters[FCNT_STATE_INSERT]++;
1136	V_pf_status.states++;
1137	#else
1138	TAILQ_INSERT_TAIL(&state_list, s, entry_list);
1139	pf_status.fcounters[FCNT_STATE_INSERT]++;
1140	pf_status.states++;
1141	#endif
1142	pfi_kif_ref(kif, PFI_KIF_REF_STATE);
1143	#if NPFSYNC > 0
1144	#ifdef __FreeBSD__
1145	if (pfsync_insert_state_ptr != NULL)
1146	pfsync_insert_state_ptr(s);
1147	#else
1148	pfsync_insert_state(s);
1149	#endif
1150	#endif
1151	return (0);
1152	}
1153
1154	struct pf_state *
1155	pf_find_state_byid(struct pf_state_cmp *key)
1156	{
1157	#ifdef __FreeBSD__
1158	V_pf_status.fcounters[FCNT_STATE_SEARCH]++;
1159
1160	return (RB_FIND(pf_state_tree_id, &V_tree_id, (struct pf_state *)key));
1161	#else
1162	pf_status.fcounters[FCNT_STATE_SEARCH]++;
1163
1164	return (RB_FIND(pf_state_tree_id, &tree_id, (struct pf_state *)key));
1165	#endif
1166	}
1167
1168	/* XXX debug function, intended to be removed one day */
1169	int
1170	pf_compare_state_keys(struct pf_state_key a, struct pf_state_key b,
1171	struct pfi_kif *kif, u_int dir)
1172	{
1173	/* a (from hdr) and b (new) must be exact opposites of each other */
1174	if (a->af == b->af && a->proto == b->proto &&
1175	PF_AEQ(&a->addr[0], &b->addr[1], a->af) &&
1176	PF_AEQ(&a->addr[1], &b->addr[0], a->af) &&
1177	a->port[0] == b->port[1] &&
1178	a->port[1] == b->port[0])
1179	return (0);
1180	else {
1181	/* mismatch. must not happen. */
1182	printf("pf: state key linking mismatch! dir=%s, "
1183	"if=%s, stored af=%u, a0: ",
1184	dir == PF_OUT ? "OUT" : "IN", kif->pfik_name, a->af);
1185	pf_print_host(&a->addr[0], a->port[0], a->af);
1186	printf(", a1: ");
1187	pf_print_host(&a->addr[1], a->port[1], a->af);
1188	printf(", proto=%u", a->proto);
1189	printf(", found af=%u, a0: ", b->af);
1190	pf_print_host(&b->addr[0], b->port[0], b->af);
1191	printf(", a1: ");
1192	pf_print_host(&b->addr[1], b->port[1], b->af);
1193	printf(", proto=%u", b->proto);
1194	printf(".\n");
1195	return (-1);
1196	}
1197	}
1198
1199	struct pf_state *
1200	#ifdef __FreeBSD__
1201	pf_find_state(struct pfi_kif kif, struct pf_state_key_cmp key, u_int dir,
1202	struct mbuf m, struct pf_mtag pftag)
1203	#else
1204	pf_find_state(struct pfi_kif kif, struct pf_state_key_cmp key, u_int dir,
1205	struct mbuf *m)
1206	#endif
1207	{
1208	struct pf_state_key *sk;
1209	struct pf_state_item *si;
1210
1211	#ifdef __FreeBSD__
1212	V_pf_status.fcounters[FCNT_STATE_SEARCH]++;
1213	#else
1214	pf_status.fcounters[FCNT_STATE_SEARCH]++;
1215	#endif
1216
1217	#ifdef __FreeBSD__
1218	if (dir == PF_OUT && pftag->statekey &&
1219	((struct pf_state_key *)pftag->statekey)->reverse)
1220	sk = ((struct pf_state_key *)pftag->statekey)->reverse;
1221	else {
1222	#ifdef __FreeBSD__
1223	if ((sk = RB_FIND(pf_state_tree, &V_pf_statetbl,
1224	#else
1225	if ((sk = RB_FIND(pf_state_tree, &pf_statetbl,
1226	#endif
1227	(struct pf_state_key *)key)) == NULL)
1228	return (NULL);
1229	if (dir == PF_OUT && pftag->statekey &&
1230	pf_compare_state_keys(pftag->statekey, sk,
1231	kif, dir) == 0) {
1232	((struct pf_state_key *)
1233	pftag->statekey)->reverse = sk;
1234	sk->reverse = pftag->statekey;
1235	}
1236	}
1237	#else
1238	if (dir == PF_OUT && m->m_pkthdr.pf.statekey &&
1239	((struct pf_state_key *)m->m_pkthdr.pf.statekey)->reverse)
1240	sk = ((struct pf_state_key *)m->m_pkthdr.pf.statekey)->reverse;
1241	else {
1242	#ifdef __FreeBSD__
1243	if ((sk = RB_FIND(pf_state_tree, &V_pf_statetbl,
1244	#else
1245	if ((sk = RB_FIND(pf_state_tree, &pf_statetbl,
1246	#endif
1247	(struct pf_state_key *)key)) == NULL)
1248	return (NULL);
1249	if (dir == PF_OUT && m->m_pkthdr.pf.statekey &&
1250	pf_compare_state_keys(m->m_pkthdr.pf.statekey, sk,
1251	kif, dir) == 0) {
1252	((struct pf_state_key *)
1253	m->m_pkthdr.pf.statekey)->reverse = sk;
1254	sk->reverse = m->m_pkthdr.pf.statekey;
1255	}
1256	}
1257	#endif
1258
1259	if (dir == PF_OUT)
1260	#ifdef __FreeBSD__
1261	pftag->statekey = NULL;
1262	#else
1263	m->m_pkthdr.pf.statekey = NULL;
1264	#endif
1265
1266	/* list is sorted, if-bound states before floating ones */
1267	TAILQ_FOREACH(si, &sk->states, entry)
1268	#ifdef __FreeBSD__
1269	if ((si->s->kif == V_pfi_all \|\| si->s->kif == kif) &&
1270	#else
1271	if ((si->s->kif == pfi_all \|\| si->s->kif == kif) &&
1272	#endif
1273	sk == (dir == PF_IN ? si->s->key[PF_SK_WIRE] :
1274	si->s->key[PF_SK_STACK]))
1275	return (si->s);
1276
1277	return (NULL);
1278	}
1279
1280	struct pf_state *
1281	pf_find_state_all(struct pf_state_key_cmp key, u_int dir, int more)
1282	{
1283	struct pf_state_key *sk;
1284	struct pf_state_item si, ret = NULL;
1285
1286	#ifdef __FreeBSD__
1287	V_pf_status.fcounters[FCNT_STATE_SEARCH]++;
1288	#else
1289	pf_status.fcounters[FCNT_STATE_SEARCH]++;
1290	#endif
1291
1292	#ifdef __FreeBSD__
1293	sk = RB_FIND(pf_state_tree, &V_pf_statetbl, (struct pf_state_key *)key);
1294	#else
1295	sk = RB_FIND(pf_state_tree, &pf_statetbl, (struct pf_state_key *)key);
1296	#endif
1297	if (sk != NULL) {
1298	TAILQ_FOREACH(si, &sk->states, entry)
1299	if (dir == PF_INOUT \|\|
1300	(sk == (dir == PF_IN ? si->s->key[PF_SK_WIRE] :
1301	si->s->key[PF_SK_STACK]))) {
1302	if (more == NULL)
1303	return (si->s);
1304
1305	if (ret)
1306	(*more)++;
1307	else
1308	ret = si;
1309	}
1310	}
1311	return (ret ? ret->s : NULL);
1312	}
1313
1314	/* END state table stuff */
1315
1316
1317	void
1318	pf_purge_thread(void *v)
1319	{
1320	int nloops = 0, s;
1321	#ifdef __FreeBSD__
1322	int locked;
1323	#endif
1324
1325	CURVNET_SET((struct vnet *)v);
1326
1327	for (;;) {
1328	tsleep(pf_purge_thread, PWAIT, "pftm", 1 * hz);
1329
1330	#ifdef __FreeBSD__
1331	sx_slock(&V_pf_consistency_lock);
1332	PF_LOCK();
1333	locked = 0;
1334
1335	if (V_pf_end_threads) {
1336	PF_UNLOCK();
1337	sx_sunlock(&V_pf_consistency_lock);
1338	sx_xlock(&V_pf_consistency_lock);
1339	PF_LOCK();
1340
1341	pf_purge_expired_states(V_pf_status.states, 1);
1342	pf_purge_expired_fragments();
1343	pf_purge_expired_src_nodes(1);
1344	V_pf_end_threads++;
1345
1346	sx_xunlock(&V_pf_consistency_lock);
1347	PF_UNLOCK();
1348	wakeup(pf_purge_thread);
1349	kproc_exit(0);
1350	}
1351	#endif
1352	s = splsoftnet();
1353
1354	/* process a fraction of the state table every second */
1355	#ifdef __FreeBSD__
1356	if (!pf_purge_expired_states(1 + (V_pf_status.states /
1357	V_pf_default_rule.timeout[PFTM_INTERVAL]), 0)) {
1358	PF_UNLOCK();
1359	sx_sunlock(&V_pf_consistency_lock);
1360	sx_xlock(&V_pf_consistency_lock);
1361	PF_LOCK();
1362	locked = 1;
1363
1364	pf_purge_expired_states(1 + (V_pf_status.states /
1365	V_pf_default_rule.timeout[PFTM_INTERVAL]), 1);
1366	}
1367	#else
1368	pf_purge_expired_states(1 + (pf_status.states
1369	/ pf_default_rule.timeout[PFTM_INTERVAL]));
1370	#endif
1371
1372	/* purge other expired types every PFTM_INTERVAL seconds */
1373	#ifdef __FreeBSD__
1374	if (++nloops >= V_pf_default_rule.timeout[PFTM_INTERVAL]) {
1375	#else
1376	if (++nloops >= pf_default_rule.timeout[PFTM_INTERVAL]) {
1377	#endif
1378	pf_purge_expired_fragments();
1379	pf_purge_expired_src_nodes(0);
1380	nloops = 0;
1381	}
1382
1383	splx(s);
1384	#ifdef __FreeBSD__
1385	PF_UNLOCK();
1386	if (locked)
1387	sx_xunlock(&V_pf_consistency_lock);
1388	else
1389	sx_sunlock(&V_pf_consistency_lock);
1390	#endif
1391	}
1392	CURVNET_RESTORE();
1393	}
1394
1395	u_int32_t
1396	pf_state_expires(const struct pf_state *state)
1397	{
1398	u_int32_t timeout;
1399	u_int32_t start;
1400	u_int32_t end;
1401	u_int32_t states;
1402
1403	/* handle all PFTM_* > PFTM_MAX here */
1404	if (state->timeout == PFTM_PURGE)
1405	return (time_second);
1406	if (state->timeout == PFTM_UNTIL_PACKET)
1407	return (0);
1408	#ifdef __FreeBSD__
1409	KASSERT(state->timeout != PFTM_UNLINKED,
1410	("pf_state_expires: timeout == PFTM_UNLINKED"));
1411	KASSERT((state->timeout < PFTM_MAX),
1412	("pf_state_expires: timeout > PFTM_MAX"));
1413	#else
1414	KASSERT(state->timeout != PFTM_UNLINKED);
1415	KASSERT(state->timeout < PFTM_MAX);
1416	#endif
1417	timeout = state->rule.ptr->timeout[state->timeout];
1418	if (!timeout)
1419	#ifdef __FreeBSD__
1420	timeout = V_pf_default_rule.timeout[state->timeout];
1421	#else
1422	timeout = pf_default_rule.timeout[state->timeout];
1423	#endif
1424	start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START];
1425	if (start) {
1426	end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END];
1427	states = state->rule.ptr->states_cur;
1428	} else {
1429	#ifdef __FreeBSD__
1430	start = V_pf_default_rule.timeout[PFTM_ADAPTIVE_START];
1431	end = V_pf_default_rule.timeout[PFTM_ADAPTIVE_END];
1432	states = V_pf_status.states;
1433	#else
1434	start = pf_default_rule.timeout[PFTM_ADAPTIVE_START];
1435	end = pf_default_rule.timeout[PFTM_ADAPTIVE_END];
1436	states = pf_status.states;
1437	#endif
1438	}
1439	if (end && states > start && start < end) {
1440	if (states < end)
1441	return (state->expire + timeout * (end - states) /
1442	(end - start));
1443	else
1444	return (time_second);
1445	}
1446	return (state->expire + timeout);
1447	}
1448
1449	#ifdef __FreeBSD__
1450	int
1451	pf_purge_expired_src_nodes(int waslocked)
1452	#else
1453	void
1454	pf_purge_expired_src_nodes(int waslocked)
1455	#endif
1456	{
1457	struct pf_src_node cur, next;
1458	int locked = waslocked;
1459
1460	#ifdef __FreeBSD__
1461	for (cur = RB_MIN(pf_src_tree, &V_tree_src_tracking); cur; cur = next) {
1462	next = RB_NEXT(pf_src_tree, &V_tree_src_tracking, cur);
1463	#else
1464	for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) {
1465	next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur);
1466	#endif
1467
1468	if (cur->states <= 0 && cur->expire <= time_second) {
1469	if (! locked) {
1470	#ifdef __FreeBSD__
1471	if (!sx_try_upgrade(&V_pf_consistency_lock))
1472	return (0);
1473	#else
1474	rw_enter_write(&pf_consistency_lock);
1475	#endif
1476	next = RB_NEXT(pf_src_tree,
1477	#ifdef __FreeBSD__
1478	&V_tree_src_tracking, cur);
1479	#else
1480	&tree_src_tracking, cur);
1481	#endif
1482	locked = 1;
1483	}
1484	if (cur->rule.ptr != NULL) {
1485	cur->rule.ptr->src_nodes--;
1486	if (cur->rule.ptr->states_cur <= 0 &&
1487	cur->rule.ptr->max_src_nodes <= 0)
1488	pf_rm_rule(NULL, cur->rule.ptr);
1489	}
1490	#ifdef __FreeBSD__
1491	RB_REMOVE(pf_src_tree, &V_tree_src_tracking, cur);
1492	V_pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
1493	V_pf_status.src_nodes--;
1494	pool_put(&V_pf_src_tree_pl, cur);
1495	#else
1496	RB_REMOVE(pf_src_tree, &tree_src_tracking, cur);
1497	pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
1498	pf_status.src_nodes--;
1499	pool_put(&pf_src_tree_pl, cur);
1500	#endif
1501	}
1502	}
1503
1504	if (locked && !waslocked)
1505	#ifdef __FreeBSD__
1506	{
1507	sx_downgrade(&V_pf_consistency_lock);
1508	}
1509	return (1);
1510	#else
1511	rw_exit_write(&pf_consistency_lock);
1512	#endif
1513	}
1514
1515	void
1516	pf_src_tree_remove_state(struct pf_state *s)
1517	{
1518	u_int32_t timeout;
1519
1520	if (s->src_node != NULL) {
1521	if (s->src.tcp_est)
1522	--s->src_node->conn;
1523	if (--s->src_node->states <= 0) {
1524	timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1525	if (!timeout)
1526	timeout =
1527	#ifdef __FreeBSD__
1528	V_pf_default_rule.timeout[PFTM_SRC_NODE];
1529	#else
1530	pf_default_rule.timeout[PFTM_SRC_NODE];
1531	#endif
1532	s->src_node->expire = time_second + timeout;
1533	}
1534	}
1535	if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) {
1536	if (--s->nat_src_node->states <= 0) {
1537	timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1538	if (!timeout)
1539	timeout =
1540	#ifdef __FreeBSD__
1541	V_pf_default_rule.timeout[PFTM_SRC_NODE];
1542	#else
1543	pf_default_rule.timeout[PFTM_SRC_NODE];
1544	#endif
1545	s->nat_src_node->expire = time_second + timeout;
1546	}
1547	}
1548	s->src_node = s->nat_src_node = NULL;
1549	}
1550
1551	/* callers should be at splsoftnet */
1552	void
1553	pf_unlink_state(struct pf_state *cur)
1554	{
1555	#ifdef __FreeBSD__
1556	if (cur->local_flags & PFSTATE_EXPIRING)
1557	return;
1558	cur->local_flags \|= PFSTATE_EXPIRING;
1559	#else
1560	splassert(IPL_SOFTNET);
1561	#endif
1562
1563	if (cur->src.state == PF_TCPS_PROXY_DST) {
1564	/* XXX wire key the right one? */
1565	#ifdef __FreeBSD__
1566	pf_send_tcp(NULL, cur->rule.ptr, cur->key[PF_SK_WIRE]->af,
1567	#else
1568	pf_send_tcp(cur->rule.ptr, cur->key[PF_SK_WIRE]->af,
1569	#endif
1570	&cur->key[PF_SK_WIRE]->addr[1],
1571	&cur->key[PF_SK_WIRE]->addr[0],
1572	cur->key[PF_SK_WIRE]->port[1],
1573	cur->key[PF_SK_WIRE]->port[0],
1574	cur->src.seqhi, cur->src.seqlo + 1,
1575	TH_RST\|TH_ACK, 0, 0, 0, 1, cur->tag, NULL, NULL);
1576	}
1577	#ifdef __FreeBSD__
1578	RB_REMOVE(pf_state_tree_id, &V_tree_id, cur);
1579	#else
1580	RB_REMOVE(pf_state_tree_id, &tree_id, cur);
1581	#endif
1582	#if NPFLOW > 0
1583	if (cur->state_flags & PFSTATE_PFLOW)
1584	#ifdef __FreeBSD__
1585	if (export_pflow_ptr != NULL)
1586	export_pflow_ptr(cur);
1587	#else
1588	export_pflow(cur);
1589	#endif
1590	#endif
1591	#if NPFSYNC > 0
1592	#ifdef __FreeBSD__
1593	if (pfsync_delete_state_ptr != NULL)
1594	pfsync_delete_state_ptr(cur);
1595	#else
1596	pfsync_delete_state(cur);
1597	#endif
1598	#endif
1599	cur->timeout = PFTM_UNLINKED;
1600	pf_src_tree_remove_state(cur);
1601	pf_detach_state(cur);
1602	}
1603
1604	/* callers should be at splsoftnet and hold the
1605	* write_lock on pf_consistency_lock */
1606	void
1607	pf_free_state(struct pf_state *cur)
1608	{
1609	#ifndef __FreeBSD__
1610	splassert(IPL_SOFTNET);
1611	#endif
1612
1613	#if NPFSYNC > 0
1614	#ifdef __FreeBSD__
1615	if (pfsync_state_in_use_ptr != NULL &&
1616	pfsync_state_in_use_ptr(cur))
1617	#else
1618	if (pfsync_state_in_use(cur))
1619	#endif
1620	return;
1621	#endif
1622	#ifdef __FreeBSD__
1623	KASSERT(cur->timeout == PFTM_UNLINKED,
1624	("pf_free_state: cur->timeout != PFTM_UNLINKED"));
1625	#else
1626	KASSERT(cur->timeout == PFTM_UNLINKED);
1627	#endif
1628	if (--cur->rule.ptr->states_cur <= 0 &&
1629	cur->rule.ptr->src_nodes <= 0)
1630	pf_rm_rule(NULL, cur->rule.ptr);
1631	if (cur->nat_rule.ptr != NULL)
1632	if (--cur->nat_rule.ptr->states_cur <= 0 &&
1633	cur->nat_rule.ptr->src_nodes <= 0)
1634	pf_rm_rule(NULL, cur->nat_rule.ptr);
1635	if (cur->anchor.ptr != NULL)
1636	if (--cur->anchor.ptr->states_cur <= 0)
1637	pf_rm_rule(NULL, cur->anchor.ptr);
1638	pf_normalize_tcp_cleanup(cur);
1639	pfi_kif_unref(cur->kif, PFI_KIF_REF_STATE);
1640	#ifdef __FreeBSD__
1641	TAILQ_REMOVE(&V_state_list, cur, entry_list);
1642	#else
1643	TAILQ_REMOVE(&state_list, cur, entry_list);
1644	#endif
1645	if (cur->tag)
1646	pf_tag_unref(cur->tag);
1647	#ifdef __FreeBSD__
1648	pool_put(&V_pf_state_pl, cur);
1649	V_pf_status.fcounters[FCNT_STATE_REMOVALS]++;
1650	V_pf_status.states--;
1651	#else
1652	pool_put(&pf_state_pl, cur);
1653	pf_status.fcounters[FCNT_STATE_REMOVALS]++;
1654	pf_status.states--;
1655	#endif
1656	}
1657
1658	#ifdef __FreeBSD__
1659	int
1660	pf_purge_expired_states(u_int32_t maxcheck, int waslocked)
1661	#else
1662	void
1663	pf_purge_expired_states(u_int32_t maxcheck)
1664	#endif
1665	{
1666	static struct pf_state *cur = NULL;
1667	struct pf_state *next;
1668	#ifdef __FreeBSD__
1669	int locked = waslocked;
1670	#else
1671	int locked = 0;
1672	#endif
1673
1674	while (maxcheck--) {
1675	/* wrap to start of list when we hit the end */
1676	if (cur == NULL) {
1677	#ifdef __FreeBSD__
1678	cur = TAILQ_FIRST(&V_state_list);
1679	#else
1680	cur = TAILQ_FIRST(&state_list);
1681	#endif
1682	if (cur == NULL)
1683	break; /* list empty */
1684	}
1685
1686	/* get next state, as cur may get deleted */
1687	next = TAILQ_NEXT(cur, entry_list);
1688
1689	if (cur->timeout == PFTM_UNLINKED) {
1690	/* free unlinked state */
1691	if (! locked) {
1692	#ifdef __FreeBSD__
1693	if (!sx_try_upgrade(&V_pf_consistency_lock))
1694	return (0);
1695	#else
1696	rw_enter_write(&pf_consistency_lock);
1697	#endif
1698	locked = 1;
1699	}
1700	pf_free_state(cur);
1701	} else if (pf_state_expires(cur) <= time_second) {
1702	/* unlink and free expired state */
1703	pf_unlink_state(cur);
1704	if (! locked) {
1705	#ifdef __FreeBSD__
1706	if (!sx_try_upgrade(&V_pf_consistency_lock))
1707	return (0);
1708	#else
1709	rw_enter_write(&pf_consistency_lock);
1710	#endif
1711	locked = 1;
1712	}
1713	pf_free_state(cur);
1714	}
1715	cur = next;
1716	}
1717
1718	#ifdef __FreeBSD__
1719	if (!waslocked && locked)
1720	sx_downgrade(&V_pf_consistency_lock);
1721
1722	return (1);
1723	#else
1724	if (locked)
1725	rw_exit_write(&pf_consistency_lock);
1726	#endif
1727	}
1728
1729	int
1730	pf_tbladdr_setup(struct pf_ruleset rs, struct pf_addr_wrap aw)
1731	{
1732	if (aw->type != PF_ADDR_TABLE)
1733	return (0);
1734	if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname, 1)) == NULL)
1735	return (1);
1736	return (0);
1737	}
1738
1739	void
1740	pf_tbladdr_remove(struct pf_addr_wrap *aw)
1741	{
1742	if (aw->type != PF_ADDR_TABLE \|\| aw->p.tbl == NULL)
1743	return;
1744	pfr_detach_table(aw->p.tbl);
1745	aw->p.tbl = NULL;
1746	}
1747
1748	void
1749	pf_tbladdr_copyout(struct pf_addr_wrap *aw)
1750	{
1751	struct pfr_ktable *kt = aw->p.tbl;
1752
1753	if (aw->type != PF_ADDR_TABLE \|\| kt == NULL)
1754	return;
1755	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
1756	kt = kt->pfrkt_root;
1757	aw->p.tbl = NULL;
1758	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
1759	kt->pfrkt_cnt : -1;
1760	}
1761
1762	void
1763	pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af)
1764	{
1765	switch (af) {
1766	#ifdef INET
1767	case AF_INET: {
1768	u_int32_t a = ntohl(addr->addr32[0]);
1769	printf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255,
1770	(a>>8)&255, a&255);
1771	if (p) {
1772	p = ntohs(p);
1773	printf(":%u", p);
1774	}
1775	break;
1776	}
1777	#endif /* INET */
1778	#ifdef INET6
1779	case AF_INET6: {
1780	u_int16_t b;
1781	u_int8_t i, curstart, curend, maxstart, maxend;
1782	curstart = curend = maxstart = maxend = 255;
1783	for (i = 0; i < 8; i++) {
1784	if (!addr->addr16[i]) {
1785	if (curstart == 255)
1786	curstart = i;
1787	curend = i;
1788	} else {
1789	if ((curend - curstart) >
1790	(maxend - maxstart)) {
1791	maxstart = curstart;
1792	maxend = curend;
1793	}
1794	curstart = curend = 255;
1795	}
1796	}
1797	if ((curend - curstart) >
1798	(maxend - maxstart)) {
1799	maxstart = curstart;
1800	maxend = curend;
1801	}
1802	for (i = 0; i < 8; i++) {
1803	if (i >= maxstart && i <= maxend) {
1804	if (i == 0)
1805	printf(":");
1806	if (i == maxend)
1807	printf(":");
1808	} else {
1809	b = ntohs(addr->addr16[i]);
1810	printf("%x", b);
1811	if (i < 7)
1812	printf(":");
1813	}
1814	}
1815	if (p) {
1816	p = ntohs(p);
1817	printf("[%u]", p);
1818	}
1819	break;
1820	}
1821	#endif /* INET6 */
1822	}
1823	}
1824
1825	void
1826	pf_print_state(struct pf_state *s)
1827	{
1828	pf_print_state_parts(s, NULL, NULL);
1829	}
1830
1831	void
1832	pf_print_state_parts(struct pf_state *s,
1833	struct pf_state_key skwp, struct pf_state_key sksp)
1834	{
1835	struct pf_state_key skw, sks;
1836	u_int8_t proto, dir;
1837
1838	/* Do our best to fill these, but they're skipped if NULL */
1839	skw = skwp ? skwp : (s ? s->key[PF_SK_WIRE] : NULL);
1840	sks = sksp ? sksp : (s ? s->key[PF_SK_STACK] : NULL);
1841	proto = skw ? skw->proto : (sks ? sks->proto : 0);
1842	dir = s ? s->direction : 0;
1843
1844	switch (proto) {
1845	case IPPROTO_IPV4:
1846	printf("IPv4");
1847	break;
1848	case IPPROTO_IPV6:
1849	printf("IPv6");
1850	break;
1851	case IPPROTO_TCP:
1852	printf("TCP");
1853	break;
1854	case IPPROTO_UDP:
1855	printf("UDP");
1856	break;
1857	case IPPROTO_ICMP:
1858	printf("ICMP");
1859	break;
1860	case IPPROTO_ICMPV6:
1861	printf("ICMPv6");
1862	break;
1863	default:
1864	printf("%u", skw->proto);
1865	break;
1866	}
1867	switch (dir) {
1868	case PF_IN:
1869	printf(" in");
1870	break;
1871	case PF_OUT:
1872	printf(" out");
1873	break;
1874	}
1875	if (skw) {
1876	printf(" wire: ");
1877	pf_print_host(&skw->addr[0], skw->port[0], skw->af);
1878	printf(" ");
1879	pf_print_host(&skw->addr[1], skw->port[1], skw->af);
1880	}
1881	if (sks) {
1882	printf(" stack: ");
1883	if (sks != skw) {
1884	pf_print_host(&sks->addr[0], sks->port[0], sks->af);
1885	printf(" ");
1886	pf_print_host(&sks->addr[1], sks->port[1], sks->af);
1887	} else
1888	printf("-");
1889	}
1890	if (s) {
1891	if (proto == IPPROTO_TCP) {
1892	printf(" [lo=%u high=%u win=%u modulator=%u",
1893	s->src.seqlo, s->src.seqhi,
1894	s->src.max_win, s->src.seqdiff);
1895	if (s->src.wscale && s->dst.wscale)
1896	printf(" wscale=%u",
1897	s->src.wscale & PF_WSCALE_MASK);
1898	printf("]");
1899	printf(" [lo=%u high=%u win=%u modulator=%u",
1900	s->dst.seqlo, s->dst.seqhi,
1901	s->dst.max_win, s->dst.seqdiff);
1902	if (s->src.wscale && s->dst.wscale)
1903	printf(" wscale=%u",
1904	s->dst.wscale & PF_WSCALE_MASK);
1905	printf("]");
1906	}
1907	printf(" %u:%u", s->src.state, s->dst.state);
1908	}
1909	}
1910
1911	void
1912	pf_print_flags(u_int8_t f)
1913	{
1914	if (f)
1915	printf(" ");
1916	if (f & TH_FIN)
1917	printf("F");
1918	if (f & TH_SYN)
1919	printf("S");
1920	if (f & TH_RST)
1921	printf("R");
1922	if (f & TH_PUSH)
1923	printf("P");
1924	if (f & TH_ACK)
1925	printf("A");
1926	if (f & TH_URG)
1927	printf("U");
1928	if (f & TH_ECE)
1929	printf("E");
1930	if (f & TH_CWR)
1931	printf("W");
1932	}
1933
1934	#define PF_SET_SKIP_STEPS(i) \
1935	do { \
1936	while (head[i] != cur) { \
1937	head[i]->skip[i].ptr = cur; \
1938	head[i] = TAILQ_NEXT(head[i], entries); \
1939	} \
1940	} while (0)
1941
1942	void
1943	pf_calc_skip_steps(struct pf_rulequeue *rules)
1944	{
1945	struct pf_rule cur, prev, *head[PF_SKIP_COUNT];
1946	int i;
1947
1948	cur = TAILQ_FIRST(rules);
1949	prev = cur;
1950	for (i = 0; i < PF_SKIP_COUNT; ++i)
1951	head[i] = cur;
1952	while (cur != NULL) {
1953
1954	if (cur->kif != prev->kif \|\| cur->ifnot != prev->ifnot)
1955	PF_SET_SKIP_STEPS(PF_SKIP_IFP);
1956	if (cur->direction != prev->direction)
1957	PF_SET_SKIP_STEPS(PF_SKIP_DIR);
1958	if (cur->af != prev->af)
1959	PF_SET_SKIP_STEPS(PF_SKIP_AF);
1960	if (cur->proto != prev->proto)
1961	PF_SET_SKIP_STEPS(PF_SKIP_PROTO);
1962	if (cur->src.neg != prev->src.neg \|\|
1963	pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr))
1964	PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR);
1965	if (cur->src.port[0] != prev->src.port[0] \|\|
1966	cur->src.port[1] != prev->src.port[1] \|\|
1967	cur->src.port_op != prev->src.port_op)
1968	PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
1969	if (cur->dst.neg != prev->dst.neg \|\|
1970	pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr))
1971	PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR);
1972	if (cur->dst.port[0] != prev->dst.port[0] \|\|
1973	cur->dst.port[1] != prev->dst.port[1] \|\|
1974	cur->dst.port_op != prev->dst.port_op)
1975	PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
1976
1977	prev = cur;
1978	cur = TAILQ_NEXT(cur, entries);
1979	}
1980	for (i = 0; i < PF_SKIP_COUNT; ++i)
1981	PF_SET_SKIP_STEPS(i);
1982	}
1983
1984	int
1985	pf_addr_wrap_neq(struct pf_addr_wrap aw1, struct pf_addr_wrap aw2)
1986	{
1987	if (aw1->type != aw2->type)
1988	return (1);
1989	switch (aw1->type) {
1990	case PF_ADDR_ADDRMASK:
1991	case PF_ADDR_RANGE:
1992	if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0))
1993	return (1);
1994	if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0))
1995	return (1);
1996	return (0);
1997	case PF_ADDR_DYNIFTL:
1998	return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt);
1999	case PF_ADDR_NOROUTE:
2000	case PF_ADDR_URPFFAILED:
2001	return (0);
2002	case PF_ADDR_TABLE:
2003	return (aw1->p.tbl != aw2->p.tbl);
2004	case PF_ADDR_RTLABEL:
2005	return (aw1->v.rtlabel != aw2->v.rtlabel);
2006	default:
2007	printf("invalid address type: %d\n", aw1->type);
2008	return (1);
2009	}
2010	}
2011
2012	u_int16_t
2013	pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp)
2014	{
2015	u_int32_t l;
2016
2017	if (udp && !cksum)
2018	return (0x0000);
2019	l = cksum + old - new;
2020	l = (l >> 16) + (l & 65535);
2021	l = l & 65535;
2022	if (udp && !l)
2023	return (0xFFFF);
2024	return (l);
2025	}
2026
2027	void
2028	pf_change_ap(struct pf_addr a, u_int16_t p, u_int16_t ic, u_int16_t pc,
2029	struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af)
2030	{
2031	struct pf_addr ao;
2032	u_int16_t po = *p;
2033
2034	PF_ACPY(&ao, a, af);
2035	PF_ACPY(a, an, af);
2036
2037	*p = pn;
2038
2039	switch (af) {
2040	#ifdef INET
2041	case AF_INET:
2042	ic = pf_cksum_fixup(pf_cksum_fixup(ic,
2043	ao.addr16[0], an->addr16[0], 0),
2044	ao.addr16[1], an->addr16[1], 0);
2045	*p = pn;
2046	pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(pc,
2047	ao.addr16[0], an->addr16[0], u),
2048	ao.addr16[1], an->addr16[1], u),
2049	po, pn, u);
2050	break;
2051	#endif /* INET */
2052	#ifdef INET6
2053	case AF_INET6:
2054	*pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2055	pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2056	pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
2057	ao.addr16[0], an->addr16[0], u),
2058	ao.addr16[1], an->addr16[1], u),
2059	ao.addr16[2], an->addr16[2], u),
2060	ao.addr16[3], an->addr16[3], u),
2061	ao.addr16[4], an->addr16[4], u),
2062	ao.addr16[5], an->addr16[5], u),
2063	ao.addr16[6], an->addr16[6], u),
2064	ao.addr16[7], an->addr16[7], u),
2065	po, pn, u);
2066	break;
2067	#endif /* INET6 */
2068	}
2069	}
2070
2071
2072	/* Changes a u_int32_t. Uses a void * so there are no align restrictions */
2073	void
2074	pf_change_a(void a, u_int16_t c, u_int32_t an, u_int8_t u)
2075	{
2076	u_int32_t ao;
2077
2078	memcpy(&ao, a, sizeof(ao));
2079	memcpy(a, &an, sizeof(u_int32_t));
2080	c = pf_cksum_fixup(pf_cksum_fixup(c, ao / 65536, an / 65536, u),
2081	ao % 65536, an % 65536, u);
2082	}
2083
2084	#ifdef INET6
2085	void
2086	pf_change_a6(struct pf_addr a, u_int16_t c, struct pf_addr *an, u_int8_t u)
2087	{
2088	struct pf_addr ao;
2089
2090	PF_ACPY(&ao, a, AF_INET6);
2091	PF_ACPY(a, an, AF_INET6);
2092
2093	*c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2094	pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2095	pf_cksum_fixup(pf_cksum_fixup(*c,
2096	ao.addr16[0], an->addr16[0], u),
2097	ao.addr16[1], an->addr16[1], u),
2098	ao.addr16[2], an->addr16[2], u),
2099	ao.addr16[3], an->addr16[3], u),
2100	ao.addr16[4], an->addr16[4], u),
2101	ao.addr16[5], an->addr16[5], u),
2102	ao.addr16[6], an->addr16[6], u),
2103	ao.addr16[7], an->addr16[7], u);
2104	}
2105	#endif /* INET6 */
2106
2107	void
2108	pf_change_icmp(struct pf_addr ia, u_int16_t ip, struct pf_addr *oa,
2109	struct pf_addr na, u_int16_t np, u_int16_t pc, u_int16_t *h2c,
2110	u_int16_t ic, u_int16_t hc, u_int8_t u, sa_family_t af)
2111	{
2112	struct pf_addr oia, ooa;
2113
2114	PF_ACPY(&oia, ia, af);
2115	if (oa)
2116	PF_ACPY(&ooa, oa, af);
2117
2118	/* Change inner protocol port, fix inner protocol checksum. */
2119	if (ip != NULL) {
2120	u_int16_t oip = *ip;
2121	u_int32_t opc;
2122
2123	if (pc != NULL)
2124	opc = *pc;
2125	*ip = np;
2126	if (pc != NULL)
2127	pc = pf_cksum_fixup(pc, oip, *ip, u);
2128	ic = pf_cksum_fixup(ic, oip, *ip, 0);
2129	if (pc != NULL)
2130	ic = pf_cksum_fixup(ic, opc, *pc, 0);
2131	}
2132	/* Change inner ip address, fix inner ip and icmp checksums. */
2133	PF_ACPY(ia, na, af);
2134	switch (af) {
2135	#ifdef INET
2136	case AF_INET: {
2137	u_int32_t oh2c = *h2c;
2138
2139	h2c = pf_cksum_fixup(pf_cksum_fixup(h2c,
2140	oia.addr16[0], ia->addr16[0], 0),
2141	oia.addr16[1], ia->addr16[1], 0);
2142	ic = pf_cksum_fixup(pf_cksum_fixup(ic,
2143	oia.addr16[0], ia->addr16[0], 0),
2144	oia.addr16[1], ia->addr16[1], 0);
2145	ic = pf_cksum_fixup(ic, oh2c, *h2c, 0);
2146	break;
2147	}
2148	#endif /* INET */
2149	#ifdef INET6
2150	case AF_INET6:
2151	*ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2152	pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2153	pf_cksum_fixup(pf_cksum_fixup(*ic,
2154	oia.addr16[0], ia->addr16[0], u),
2155	oia.addr16[1], ia->addr16[1], u),
2156	oia.addr16[2], ia->addr16[2], u),
2157	oia.addr16[3], ia->addr16[3], u),
2158	oia.addr16[4], ia->addr16[4], u),
2159	oia.addr16[5], ia->addr16[5], u),
2160	oia.addr16[6], ia->addr16[6], u),
2161	oia.addr16[7], ia->addr16[7], u);
2162	break;
2163	#endif /* INET6 */
2164	}
2165	/* Outer ip address, fix outer ip or icmpv6 checksum, if necessary. */
2166	if (oa) {
2167	PF_ACPY(oa, na, af);
2168	switch (af) {
2169	#ifdef INET
2170	case AF_INET:
2171	hc = pf_cksum_fixup(pf_cksum_fixup(hc,
2172	ooa.addr16[0], oa->addr16[0], 0),
2173	ooa.addr16[1], oa->addr16[1], 0);
2174	break;
2175	#endif /* INET */
2176	#ifdef INET6
2177	case AF_INET6:
2178	*ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2179	pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2180	pf_cksum_fixup(pf_cksum_fixup(*ic,
2181	ooa.addr16[0], oa->addr16[0], u),
2182	ooa.addr16[1], oa->addr16[1], u),
2183	ooa.addr16[2], oa->addr16[2], u),
2184	ooa.addr16[3], oa->addr16[3], u),
2185	ooa.addr16[4], oa->addr16[4], u),
2186	ooa.addr16[5], oa->addr16[5], u),
2187	ooa.addr16[6], oa->addr16[6], u),
2188	ooa.addr16[7], oa->addr16[7], u);
2189	break;
2190	#endif /* INET6 */
2191	}
2192	}
2193	}
2194
2195
2196	/*
2197	* Need to modulate the sequence numbers in the TCP SACK option
2198	* (credits to Krzysztof Pfaff for report and patch)
2199	*/
2200	int
2201	pf_modulate_sack(struct mbuf m, int off, struct pf_pdesc pd,
2202	struct tcphdr th, struct pf_state_peer dst)
2203	{
2204	int hlen = (th->th_off << 2) - sizeof(*th), thoptlen = hlen;
2205	#ifdef __FreeBSD__
2206	u_int8_t opts[TCP_MAXOLEN], *opt = opts;
2207	#else
2208	u_int8_t opts[MAX_TCPOPTLEN], *opt = opts;
2209	#endif
2210	int copyback = 0, i, olen;
2211	struct sackblk sack;
2212
2213	#define TCPOLEN_SACKLEN (TCPOLEN_SACK + 2)
2214	if (hlen < TCPOLEN_SACKLEN \|\|
2215	!pf_pull_hdr(m, off + sizeof(*th), opts, hlen, NULL, NULL, pd->af))
2216	return 0;
2217
2218	while (hlen >= TCPOLEN_SACKLEN) {
2219	olen = opt[1];
2220	switch (*opt) {
2221	case TCPOPT_EOL: /* FALLTHROUGH */
2222	case TCPOPT_NOP:
2223	opt++;
2224	hlen--;
2225	break;
2226	case TCPOPT_SACK:
2227	if (olen > hlen)
2228	olen = hlen;
2229	if (olen >= TCPOLEN_SACKLEN) {
2230	for (i = 2; i + TCPOLEN_SACK <= olen;
2231	i += TCPOLEN_SACK) {
2232	memcpy(&sack, &opt[i], sizeof(sack));
2233	pf_change_a(&sack.start, &th->th_sum,
2234	htonl(ntohl(sack.start) -
2235	dst->seqdiff), 0);
2236	pf_change_a(&sack.end, &th->th_sum,
2237	htonl(ntohl(sack.end) -
2238	dst->seqdiff), 0);
2239	memcpy(&opt[i], &sack, sizeof(sack));
2240	}
2241	copyback = 1;
2242	}
2243	/* FALLTHROUGH */
2244	default:
2245	if (olen < 2)
2246	olen = 2;
2247	hlen -= olen;
2248	opt += olen;
2249	}
2250	}
2251
2252	if (copyback)
2253	#ifdef __FreeBSD__
2254	m_copyback(m, off + sizeof(*th), thoptlen, (caddr_t)opts);
2255	#else
2256	m_copyback(m, off + sizeof(*th), thoptlen, opts);
2257	#endif
2258	return (copyback);
2259	}
2260
2261	void
2262	#ifdef __FreeBSD__
2263	pf_send_tcp(struct mbuf replyto, const struct pf_rule r, sa_family_t af,
2264	#else
2265	pf_send_tcp(const struct pf_rule *r, sa_family_t af,
2266	#endif
2267	const struct pf_addr saddr, const struct pf_addr daddr,
2268	u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack,
2269	u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl, int tag,
2270	u_int16_t rtag, struct ether_header eh, struct ifnet ifp)
2271	{
2272	struct mbuf *m;
2273	int len, tlen;
2274	#ifdef INET
2275	struct ip *h;
2276	#endif /* INET */
2277	#ifdef INET6
2278	struct ip6_hdr *h6;
2279	#endif /* INET6 */
2280	struct tcphdr *th;
2281	char *opt;
2282	#ifdef __FreeBSD__
2283	struct pf_mtag *pf_mtag;
2284
2285	KASSERT(
2286	#ifdef INET
2287	af == AF_INET
2288	#else
2289	0
2290	#endif
2291	\|\|
2292	#ifdef INET6
2293	af == AF_INET6
2294	#else
2295	0
2296	#endif
2297	, ("Unsupported AF %d", af));
2298	len = 0;
2299	th = NULL;
2300	#ifdef INET
2301	h = NULL;
2302	#endif
2303	#ifdef INET6
2304	h6 = NULL;
2305	#endif
2306	#endif /* __FreeBSD__ */
2307
2308	/* maximum segment size tcp option */
2309	tlen = sizeof(struct tcphdr);
2310	if (mss)
2311	tlen += 4;
2312
2313	switch (af) {
2314	#ifdef INET
2315	case AF_INET:
2316	len = sizeof(struct ip) + tlen;
2317	break;
2318	#endif /* INET */
2319	#ifdef INET6
2320	case AF_INET6:
2321	len = sizeof(struct ip6_hdr) + tlen;
2322	break;
2323	#endif /* INET6 */
2324	}
2325
2326	/* create outgoing mbuf */
2327	m = m_gethdr(M_DONTWAIT, MT_HEADER);
2328	if (m == NULL)
2329	return;
2330	#ifdef __FreeBSD__
2331	#ifdef MAC
2332	mac_netinet_firewall_send(m);
2333	#endif
2334	if ((pf_mtag = pf_get_mtag(m)) == NULL) {
2335	m_freem(m);
2336	return;
2337	}
2338	#endif
2339	if (tag)
2340	#ifdef __FreeBSD__
2341	m->m_flags \|= M_SKIP_FIREWALL;
2342	pf_mtag->tag = rtag;
2343	#else
2344	m->m_pkthdr.pf.flags \|= PF_TAG_GENERATED;
2345	m->m_pkthdr.pf.tag = rtag;
2346	#endif
2347
2348	if (r != NULL && r->rtableid >= 0)
2349	#ifdef __FreeBSD__
2350	{
2351	M_SETFIB(m, r->rtableid);
2352	pf_mtag->rtableid = r->rtableid;
2353	#else
2354	m->m_pkthdr.pf.rtableid = r->rtableid;
2355	#endif
2356	#ifdef __FreeBSD__
2357	}
2358	#endif
2359
2360	#ifdef ALTQ
2361	if (r != NULL && r->qid) {
2362	#ifdef __FreeBSD__
2363	pf_mtag->qid = r->qid;
2364
2365	/* add hints for ecn */
2366	pf_mtag->hdr = mtod(m, struct ip *);
2367	#else
2368	m->m_pkthdr.pf.qid = r->qid;
2369	/* add hints for ecn */
2370	m->m_pkthdr.pf.hdr = mtod(m, struct ip *);
2371	#endif
2372	}
2373	#endif /* ALTQ */
2374	m->m_data += max_linkhdr;
2375	m->m_pkthdr.len = m->m_len = len;
2376	m->m_pkthdr.rcvif = NULL;
2377	bzero(m->m_data, len);
2378	switch (af) {
2379	#ifdef INET
2380	case AF_INET:
2381	h = mtod(m, struct ip *);
2382
2383	/* IP header fields included in the TCP checksum */
2384	h->ip_p = IPPROTO_TCP;
2385	h->ip_len = htons(tlen);
2386	h->ip_src.s_addr = saddr->v4.s_addr;
2387	h->ip_dst.s_addr = daddr->v4.s_addr;
2388
2389	th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip));
2390	break;
2391	#endif /* INET */
2392	#ifdef INET6
2393	case AF_INET6:
2394	h6 = mtod(m, struct ip6_hdr *);
2395
2396	/* IP header fields included in the TCP checksum */
2397	h6->ip6_nxt = IPPROTO_TCP;
2398	h6->ip6_plen = htons(tlen);
2399	memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr));
2400	memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr));
2401
2402	th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr));
2403	break;
2404	#endif /* INET6 */
2405	}
2406
2407	/* TCP header */
2408	th->th_sport = sport;
2409	th->th_dport = dport;
2410	th->th_seq = htonl(seq);
2411	th->th_ack = htonl(ack);
2412	th->th_off = tlen >> 2;
2413	th->th_flags = flags;
2414	th->th_win = htons(win);
2415
2416	if (mss) {
2417	opt = (char *)(th + 1);
2418	opt[0] = TCPOPT_MAXSEG;
2419	opt[1] = 4;
2420	HTONS(mss);
2421	bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2);
2422	}
2423
2424	switch (af) {
2425	#ifdef INET
2426	case AF_INET:
2427	/* TCP checksum */
2428	th->th_sum = in_cksum(m, len);
2429
2430	/* Finish the IP header */
2431	h->ip_v = 4;
2432	h->ip_hl = sizeof(*h) >> 2;
2433	h->ip_tos = IPTOS_LOWDELAY;
2434	#ifdef __FreeBSD__
2435	h->ip_off = V_path_mtu_discovery ? IP_DF : 0;
2436	h->ip_len = len;
2437	h->ip_ttl = ttl ? ttl : V_ip_defttl;
2438	#else
2439	h->ip_len = htons(len);
2440	h->ip_off = htons(ip_mtudisc ? IP_DF : 0);
2441	h->ip_ttl = ttl ? ttl : ip_defttl;
2442	#endif
2443	h->ip_sum = 0;
2444	if (eh == NULL) {
2445	#ifdef __FreeBSD__
2446	PF_UNLOCK();
2447	ip_output(m, (void )NULL, (void )NULL, 0,
2448	(void )NULL, (void )NULL);
2449	PF_LOCK();
2450	#else /* ! __FreeBSD__ */
2451	ip_output(m, (void )NULL, (void )NULL, 0,
2452	(void )NULL, (void )NULL);
2453	#endif
2454	} else {
2455	struct route ro;
2456	struct rtentry rt;
2457	struct ether_header e = (void )ro.ro_dst.sa_data;
2458
2459	if (ifp == NULL) {
2460	m_freem(m);
2461	return;
2462	}
2463	rt.rt_ifp = ifp;
2464	ro.ro_rt = &rt;
2465	ro.ro_dst.sa_len = sizeof(ro.ro_dst);
2466	ro.ro_dst.sa_family = pseudo_AF_HDRCMPLT;
2467	bcopy(eh->ether_dhost, e->ether_shost, ETHER_ADDR_LEN);
2468	bcopy(eh->ether_shost, e->ether_dhost, ETHER_ADDR_LEN);
2469	e->ether_type = eh->ether_type;
2470	#ifdef __FreeBSD__
2471	PF_UNLOCK();
2472	/* XXX_IMPORT: later */
2473	ip_output(m, (void *)NULL, &ro, 0,
2474	(void )NULL, (void )NULL);
2475	PF_LOCK();
2476	#else /* ! __FreeBSD__ */
2477	ip_output(m, (void *)NULL, &ro, IP_ROUTETOETHER,
2478	(void )NULL, (void )NULL);
2479	#endif
2480	}
2481	break;
2482	#endif /* INET */
2483	#ifdef INET6
2484	case AF_INET6:
2485	/* TCP checksum */
2486	th->th_sum = in6_cksum(m, IPPROTO_TCP,
2487	sizeof(struct ip6_hdr), tlen);
2488
2489	h6->ip6_vfc \|= IPV6_VERSION;
2490	h6->ip6_hlim = IPV6_DEFHLIM;
2491
2492	#ifdef __FreeBSD__
2493	PF_UNLOCK();
2494	ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
2495	PF_LOCK();
2496	#else
2497	ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
2498	#endif
2499	break;
2500	#endif /* INET6 */
2501	}
2502	}
2503
2504	static void
2505	pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af,
2506	struct pf_rule *r)
2507	{
2508	struct mbuf *m0;
2509	#ifdef __FreeBSD__
2510	#ifdef INET
2511	struct ip *ip;
2512	#endif
2513	struct pf_mtag *pf_mtag;
2514	#endif
2515
2516	#ifdef __FreeBSD__
2517	m0 = m_copypacket(m, M_DONTWAIT);
2518	if (m0 == NULL)
2519	return;
2520	#else
2521	if ((m0 = m_copy(m, 0, M_COPYALL)) == NULL)
2522	return;
2523	#endif
2524
2525	#ifdef __FreeBSD__
2526	if ((pf_mtag = pf_get_mtag(m0)) == NULL)
2527	return;
2528	/* XXX: revisit */
2529	m0->m_flags \|= M_SKIP_FIREWALL;
2530	#else
2531	m0->m_pkthdr.pf.flags \|= PF_TAG_GENERATED;
2532	#endif
2533
2534	if (r->rtableid >= 0)
2535	#ifdef __FreeBSD__
2536	{
2537	M_SETFIB(m0, r->rtableid);
2538	pf_mtag->rtableid = r->rtableid;
2539	#else
2540	m0->m_pkthdr.pf.rtableid = r->rtableid;
2541	#endif
2542	#ifdef __FreeBSD__
2543	}
2544	#endif
2545
2546	#ifdef ALTQ
2547	if (r->qid) {
2548	#ifdef __FreeBSD__
2549	pf_mtag->qid = r->qid;
2550	/* add hints for ecn */
2551	pf_mtag->hdr = mtod(m0, struct ip *);
2552	#else
2553	m0->m_pkthdr.pf.qid = r->qid;
2554	/* add hints for ecn */
2555	m0->m_pkthdr.pf.hdr = mtod(m0, struct ip *);
2556	#endif
2557	}
2558	#endif /* ALTQ */
2559
2560	switch (af) {
2561	#ifdef INET
2562	case AF_INET:
2563	#ifdef __FreeBSD__
2564	/* icmp_error() expects host byte ordering */
2565	ip = mtod(m0, struct ip *);
2566	NTOHS(ip->ip_len);
2567	NTOHS(ip->ip_off);
2568	PF_UNLOCK();
2569	icmp_error(m0, type, code, 0, 0);
2570	PF_LOCK();
2571	#else
2572	icmp_error(m0, type, code, 0, 0);
2573	#endif
2574	break;
2575	#endif /* INET */
2576	#ifdef INET6
2577	case AF_INET6:
2578	#ifdef __FreeBSD__
2579	PF_UNLOCK();
2580	#endif
2581	icmp6_error(m0, type, code, 0);
2582	#ifdef __FreeBSD__
2583	PF_LOCK();
2584	#endif
2585	break;
2586	#endif /* INET6 */
2587	}
2588	}
2589
2590	/*
2591	* Return 1 if the addresses a and b match (with mask m), otherwise return 0.
2592	* If n is 0, they match if they are equal. If n is != 0, they match if they
2593	* are different.
2594	*/
2595	int
2596	pf_match_addr(u_int8_t n, struct pf_addr a, struct pf_addr m,
2597	struct pf_addr *b, sa_family_t af)
2598	{
2599	int match = 0;
2600
2601	switch (af) {
2602	#ifdef INET
2603	case AF_INET:
2604	if ((a->addr32[0] & m->addr32[0]) ==
2605	(b->addr32[0] & m->addr32[0]))
2606	match++;
2607	break;
2608	#endif /* INET */
2609	#ifdef INET6
2610	case AF_INET6:
2611	if (((a->addr32[0] & m->addr32[0]) ==
2612	(b->addr32[0] & m->addr32[0])) &&
2613	((a->addr32[1] & m->addr32[1]) ==
2614	(b->addr32[1] & m->addr32[1])) &&
2615	((a->addr32[2] & m->addr32[2]) ==
2616	(b->addr32[2] & m->addr32[2])) &&
2617	((a->addr32[3] & m->addr32[3]) ==
2618	(b->addr32[3] & m->addr32[3])))
2619	match++;
2620	break;
2621	#endif /* INET6 */
2622	}
2623	if (match) {
2624	if (n)
2625	return (0);
2626	else
2627	return (1);
2628	} else {
2629	if (n)
2630	return (1);
2631	else
2632	return (0);
2633	}
2634	}
2635
2636	/*
2637	* Return 1 if b <= a <= e, otherwise return 0.
2638	*/
2639	int
2640	pf_match_addr_range(struct pf_addr b, struct pf_addr e,
2641	struct pf_addr *a, sa_family_t af)
2642	{
2643	switch (af) {
2644	#ifdef INET
2645	case AF_INET:
2646	if ((a->addr32[0] < b->addr32[0]) \|\|
2647	(a->addr32[0] > e->addr32[0]))
2648	return (0);
2649	break;
2650	#endif /* INET */
2651	#ifdef INET6
2652	case AF_INET6: {
2653	int i;
2654
2655	/* check a >= b */
2656	for (i = 0; i < 4; ++i)
2657	if (a->addr32[i] > b->addr32[i])
2658	break;
2659	else if (a->addr32[i] < b->addr32[i])
2660	return (0);
2661	/* check a <= e */
2662	for (i = 0; i < 4; ++i)
2663	if (a->addr32[i] < e->addr32[i])
2664	break;
2665	else if (a->addr32[i] > e->addr32[i])
2666	return (0);
2667	break;
2668	}
2669	#endif /* INET6 */
2670	}
2671	return (1);
2672	}
2673
2674	int
2675	pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p)
2676	{
2677	switch (op) {
2678	case PF_OP_IRG:
2679	return ((p > a1) && (p < a2));
2680	case PF_OP_XRG:
2681	return ((p < a1) \|\| (p > a2));
2682	case PF_OP_RRG:
2683	return ((p >= a1) && (p <= a2));
2684	case PF_OP_EQ:
2685	return (p == a1);
2686	case PF_OP_NE:
2687	return (p != a1);
2688	case PF_OP_LT:
2689	return (p < a1);
2690	case PF_OP_LE:
2691	return (p <= a1);
2692	case PF_OP_GT:
2693	return (p > a1);
2694	case PF_OP_GE:
2695	return (p >= a1);
2696	}
2697	return (0); /* never reached */
2698	}
2699
2700	int
2701	pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p)
2702	{
2703	NTOHS(a1);
2704	NTOHS(a2);
2705	NTOHS(p);
2706	return (pf_match(op, a1, a2, p));
2707	}
2708
2709	int
2710	pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u)
2711	{
2712	if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
2713	return (0);
2714	return (pf_match(op, a1, a2, u));
2715	}
2716
2717	int
2718	pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g)
2719	{
2720	if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
2721	return (0);
2722	return (pf_match(op, a1, a2, g));
2723	}
2724
2725	int
2726	#ifdef __FreeBSD__
2727	pf_match_tag(struct mbuf m, struct pf_rule r, int *tag,
2728	struct pf_mtag *pf_mtag)
2729	#else
2730	pf_match_tag(struct mbuf m, struct pf_rule r, int *tag)
2731	#endif
2732	{
2733	if (*tag == -1)
2734	#ifdef __FreeBSD__
2735	*tag = pf_mtag->tag;
2736	#else
2737	*tag = m->m_pkthdr.pf.tag;
2738	#endif
2739
2740	return ((!r->match_tag_not && r->match_tag == *tag) \|\|
2741	(r->match_tag_not && r->match_tag != *tag));
2742	}
2743
2744	int
2745	#ifdef __FreeBSD__
2746	pf_tag_packet(struct mbuf *m, int tag, int rtableid,
2747	struct pf_mtag *pf_mtag)
2748	#else
2749	pf_tag_packet(struct mbuf *m, int tag, int rtableid)
2750	#endif
2751	{
2752	if (tag <= 0 && rtableid < 0)
2753	return (0);
2754
2755	if (tag > 0)
2756	#ifdef __FreeBSD__
2757	pf_mtag->tag = tag;
2758	#else
2759	m->m_pkthdr.pf.tag = tag;
2760	#endif
2761	if (rtableid >= 0)
2762	#ifdef __FreeBSD__
2763	{
2764	M_SETFIB(m, rtableid);
2765	}
2766	#else
2767	m->m_pkthdr.pf.rtableid = rtableid;
2768	#endif
2769
2770	return (0);
2771	}
2772
2773	void
2774	pf_step_into_anchor(int depth, struct pf_ruleset *rs, int n,
2775	struct pf_rule r, struct pf_rule a, int *match)
2776	{
2777	struct pf_anchor_stackframe *f;
2778
2779	(*r)->anchor->match = 0;
2780	if (match)
2781	*match = 0;
2782	#ifdef __FreeBSD__
2783	if (*depth >= sizeof(V_pf_anchor_stack) /
2784	sizeof(V_pf_anchor_stack[0])) {
2785	#else
2786	if (*depth >= sizeof(pf_anchor_stack) /
2787	sizeof(pf_anchor_stack[0])) {
2788	#endif
2789	printf("pf_step_into_anchor: stack overflow\n");
2790	r = TAILQ_NEXT(r, entries);
2791	return;
2792	} else if (*depth == 0 && a != NULL)
2793	a = r;
2794	#ifdef __FreeBSD__
2795	f = V_pf_anchor_stack + (*depth)++;
2796	#else
2797	f = pf_anchor_stack + (*depth)++;
2798	#endif
2799	f->rs = *rs;
2800	f->r = *r;
2801	if ((*r)->anchor_wildcard) {
2802	f->parent = &(*r)->anchor->children;
2803	if ((f->child = RB_MIN(pf_anchor_node, f->parent)) ==
2804	NULL) {
2805	*r = NULL;
2806	return;
2807	}
2808	*rs = &f->child->ruleset;
2809	} else {
2810	f->parent = NULL;
2811	f->child = NULL;
2812	rs = &(r)->anchor->ruleset;
2813	}
2814	r = TAILQ_FIRST((rs)->rules[n].active.ptr);
2815	}
2816
2817	int
2818	pf_step_out_of_anchor(int depth, struct pf_ruleset *rs, int n,
2819	struct pf_rule r, struct pf_rule a, int *match)
2820	{
2821	struct pf_anchor_stackframe *f;
2822	int quick = 0;
2823
2824	do {
2825	if (*depth <= 0)
2826	break;
2827	#ifdef __FreeBSD__
2828	f = V_pf_anchor_stack + *depth - 1;
2829	#else
2830	f = pf_anchor_stack + *depth - 1;
2831	#endif
2832	if (f->parent != NULL && f->child != NULL) {
2833	if (f->child->match \|\|
2834	(match != NULL && *match)) {
2835	f->r->anchor->match = 1;
2836	*match = 0;
2837	}
2838	f->child = RB_NEXT(pf_anchor_node, f->parent, f->child);
2839	if (f->child != NULL) {
2840	*rs = &f->child->ruleset;
2841	r = TAILQ_FIRST((rs)->rules[n].active.ptr);
2842	if (*r == NULL)
2843	continue;
2844	else
2845	break;
2846	}
2847	}
2848	(*depth)--;
2849	if (*depth == 0 && a != NULL)
2850	*a = NULL;
2851	*rs = f->rs;
2852	if (f->r->anchor->match \|\| (match != NULL && *match))
2853	quick = f->r->quick;
2854	*r = TAILQ_NEXT(f->r, entries);
2855	} while (*r == NULL);
2856
2857	return (quick);
2858	}
2859
2860	#ifdef INET6
2861	void
2862	pf_poolmask(struct pf_addr naddr, struct pf_addr raddr,
2863	struct pf_addr rmask, struct pf_addr saddr, sa_family_t af)
2864	{
2865	switch (af) {
2866	#ifdef INET
2867	case AF_INET:
2868	naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) \|
2869	((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2870	break;
2871	#endif /* INET */
2872	case AF_INET6:
2873	naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) \|
2874	((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2875	naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) \|
2876	((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]);
2877	naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) \|
2878	((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]);
2879	naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) \|
2880	((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]);
2881	break;
2882	}
2883	}
2884
2885	void
2886	pf_addr_inc(struct pf_addr *addr, sa_family_t af)
2887	{
2888	switch (af) {
2889	#ifdef INET
2890	case AF_INET:
2891	addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1);
2892	break;
2893	#endif /* INET */
2894	case AF_INET6:
2895	if (addr->addr32[3] == 0xffffffff) {
2896	addr->addr32[3] = 0;
2897	if (addr->addr32[2] == 0xffffffff) {
2898	addr->addr32[2] = 0;
2899	if (addr->addr32[1] == 0xffffffff) {
2900	addr->addr32[1] = 0;
2901	addr->addr32[0] =
2902	htonl(ntohl(addr->addr32[0]) + 1);
2903	} else
2904	addr->addr32[1] =
2905	htonl(ntohl(addr->addr32[1]) + 1);
2906	} else
2907	addr->addr32[2] =
2908	htonl(ntohl(addr->addr32[2]) + 1);
2909	} else
2910	addr->addr32[3] =
2911	htonl(ntohl(addr->addr32[3]) + 1);
2912	break;
2913	}
2914	}
2915	#endif /* INET6 */
2916
2917	int
2918	#ifdef __FreeBSD__
2919	pf_socket_lookup(int direction, struct pf_pdesc pd, struct inpcb inp_arg)
2920	#else
2921	pf_socket_lookup(int direction, struct pf_pdesc *pd)
2922	#endif
2923	{
2924	struct pf_addr saddr, daddr;
2925	u_int16_t sport, dport;
2926	#ifdef __FreeBSD__
2927	struct inpcbinfo *pi;
2928	#else
2929	struct inpcbtable *tb;
2930	#endif
2931	struct inpcb *inp;
2932
2933	if (pd == NULL)
2934	return (-1);
2935	pd->lookup.uid = UID_MAX;
2936	pd->lookup.gid = GID_MAX;
2937	pd->lookup.pid = NO_PID;
2938
2939	#ifdef __FreeBSD__
2940	if (inp_arg != NULL) {
2941	INP_LOCK_ASSERT(inp_arg);
2942	#ifndef __rtems__
2943	pd->lookup.uid = inp_arg->inp_cred->cr_uid;
2944	pd->lookup.gid = inp_arg->inp_cred->cr_groups[0];
2945	#else /* __rtems__ */
2946	pd->lookup.uid = BSD_DEFAULT_UID;
2947	pd->lookup.gid = BSD_DEFAULT_GID;
2948	#endif /* __rtems__ */
2949	return (1);
2950	}
2951	#endif
2952
2953	switch (pd->proto) {
2954	case IPPROTO_TCP:
2955	if (pd->hdr.tcp == NULL)
2956	return (-1);
2957	sport = pd->hdr.tcp->th_sport;
2958	dport = pd->hdr.tcp->th_dport;
2959	#ifdef __FreeBSD__
2960	pi = &V_tcbinfo;
2961	#else
2962	tb = &tcbtable;
2963	#endif
2964	break;
2965	case IPPROTO_UDP:
2966	if (pd->hdr.udp == NULL)
2967	return (-1);
2968	sport = pd->hdr.udp->uh_sport;
2969	dport = pd->hdr.udp->uh_dport;
2970	#ifdef __FreeBSD__
2971	pi = &V_udbinfo;
2972	#else
2973	tb = &udbtable;
2974	#endif
2975	break;
2976	default:
2977	return (-1);
2978	}
2979	if (direction == PF_IN) {
2980	saddr = pd->src;
2981	daddr = pd->dst;
2982	} else {
2983	u_int16_t p;
2984
2985	p = sport;
2986	sport = dport;
2987	dport = p;
2988	saddr = pd->dst;
2989	daddr = pd->src;
2990	}
2991	switch (pd->af) {
2992	#ifdef INET
2993	case AF_INET:
2994	#ifdef __FreeBSD__
2995	/*
2996	* XXXRW: would be nice if we had an mbuf here so that we
2997	* could use in_pcblookup_mbuf().
2998	*/
2999	inp = in_pcblookup(pi, saddr->v4, sport, daddr->v4,
3000	dport, INPLOOKUP_RLOCKPCB, NULL);
3001	if (inp == NULL) {
3002	inp = in_pcblookup(pi, saddr->v4, sport,
3003	daddr->v4, dport, INPLOOKUP_WILDCARD \|
3004	INPLOOKUP_RLOCKPCB, NULL);
3005	if (inp == NULL)
3006	return (-1);
3007	}
3008	#else
3009	inp = in_pcbhashlookup(tb, saddr->v4, sport, daddr->v4, dport);
3010	if (inp == NULL) {
3011	inp = in_pcblookup_listen(tb, daddr->v4, dport, 0,
3012	NULL);
3013	if (inp == NULL)
3014	return (-1);
3015	}
3016	#endif
3017	break;
3018	#endif /* INET */
3019	#ifdef INET6
3020	case AF_INET6:
3021	#ifdef __FreeBSD__
3022	/*
3023	* XXXRW: would be nice if we had an mbuf here so that we
3024	* could use in6_pcblookup_mbuf().
3025	*/
3026	inp = in6_pcblookup(pi, &saddr->v6, sport,
3027	&daddr->v6, dport, INPLOOKUP_RLOCKPCB, NULL);
3028	if (inp == NULL) {
3029	inp = in6_pcblookup(pi, &saddr->v6, sport,
3030	&daddr->v6, dport, INPLOOKUP_WILDCARD \|
3031	INPLOOKUP_RLOCKPCB, NULL);
3032	if (inp == NULL)
3033	return (-1);
3034	}
3035	#else
3036	inp = in6_pcbhashlookup(tb, &saddr->v6, sport, &daddr->v6,
3037	dport);
3038	if (inp == NULL) {
3039	inp = in6_pcblookup_listen(tb, &daddr->v6, dport, 0,
3040	NULL);
3041	if (inp == NULL)
3042	return (-1);
3043	}
3044	#endif
3045	break;
3046	#endif /* INET6 */
3047
3048	default:
3049	return (-1);
3050	}
3051	#ifdef __FreeBSD__
3052	INP_RLOCK_ASSERT(inp);
3053	#ifndef __rtems__
3054	pd->lookup.uid = inp->inp_cred->cr_uid;
3055	pd->lookup.gid = inp->inp_cred->cr_groups[0];
3056	#else /* __rtems__ */
3057	pd->lookup.uid = BSD_DEFAULT_UID;
3058	pd->lookup.gid = BSD_DEFAULT_GID;
3059	#endif /* __rtems__ */
3060	INP_RUNLOCK(inp);
3061	#else
3062	pd->lookup.uid = inp->inp_socket->so_euid;
3063	pd->lookup.gid = inp->inp_socket->so_egid;
3064	pd->lookup.pid = inp->inp_socket->so_cpid;
3065	#endif
3066	return (1);
3067	}
3068
3069	u_int8_t
3070	pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
3071	{
3072	int hlen;
3073	u_int8_t hdr[60];
3074	u_int8_t *opt, optlen;
3075	u_int8_t wscale = 0;
3076
3077	hlen = th_off << 2; /* hlen <= sizeof(hdr) */
3078	if (hlen <= sizeof(struct tcphdr))
3079	return (0);
3080	if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
3081	return (0);
3082	opt = hdr + sizeof(struct tcphdr);
3083	hlen -= sizeof(struct tcphdr);
3084	while (hlen >= 3) {
3085	switch (*opt) {
3086	case TCPOPT_EOL:
3087	case TCPOPT_NOP:
3088	++opt;
3089	--hlen;
3090	break;
3091	case TCPOPT_WINDOW:
3092	wscale = opt[2];
3093	if (wscale > TCP_MAX_WINSHIFT)
3094	wscale = TCP_MAX_WINSHIFT;
3095	wscale \|= PF_WSCALE_FLAG;
3096	/* FALLTHROUGH */
3097	default:
3098	optlen = opt[1];
3099	if (optlen < 2)
3100	optlen = 2;
3101	hlen -= optlen;
3102	opt += optlen;
3103	break;
3104	}
3105	}
3106	return (wscale);
3107	}
3108
3109	u_int16_t
3110	pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
3111	{
3112	int hlen;
3113	u_int8_t hdr[60];
3114	u_int8_t *opt, optlen;
3115	#ifdef __FreeBSD__
3116	u_int16_t mss = V_tcp_mssdflt;
3117	#else
3118	u_int16_t mss = tcp_mssdflt;
3119	#endif
3120
3121	hlen = th_off << 2; /* hlen <= sizeof(hdr) */
3122	if (hlen <= sizeof(struct tcphdr))
3123	return (0);
3124	if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
3125	return (0);
3126	opt = hdr + sizeof(struct tcphdr);
3127	hlen -= sizeof(struct tcphdr);
3128	while (hlen >= TCPOLEN_MAXSEG) {
3129	switch (*opt) {
3130	case TCPOPT_EOL:
3131	case TCPOPT_NOP:
3132	++opt;
3133	--hlen;
3134	break;
3135	case TCPOPT_MAXSEG:
3136	bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2);
3137	NTOHS(mss);
3138	/* FALLTHROUGH */
3139	default:
3140	optlen = opt[1];
3141	if (optlen < 2)
3142	optlen = 2;
3143	hlen -= optlen;
3144	opt += optlen;
3145	break;
3146	}
3147	}
3148	return (mss);
3149	}
3150
3151	u_int16_t
3152	pf_calc_mss(struct pf_addr *addr, sa_family_t af, int rtableid, u_int16_t offer)
3153	{
3154	#ifdef INET
3155	struct sockaddr_in *dst;
3156	struct route ro;
3157	#endif /* INET */
3158	#ifdef INET6
3159	struct sockaddr_in6 *dst6;
3160	struct route_in6 ro6;
3161	#endif /* INET6 */
3162	struct rtentry *rt = NULL;
3163	#ifdef __FreeBSD__
3164	int hlen = 0;
3165	u_int16_t mss = V_tcp_mssdflt;
3166	#else
3167	int hlen;
3168	u_int16_t mss = tcp_mssdflt;
3169	#endif
3170
3171	switch (af) {
3172	#ifdef INET
3173	case AF_INET:
3174	hlen = sizeof(struct ip);
3175	bzero(&ro, sizeof(ro));
3176	dst = (struct sockaddr_in *)&ro.ro_dst;
3177	dst->sin_family = AF_INET;
3178	dst->sin_len = sizeof(*dst);
3179	dst->sin_addr = addr->v4;
3180	#ifdef __FreeBSD__
3181	in_rtalloc_ign(&ro, 0, rtableid);
3182	#else /* ! __FreeBSD__ */
3183	rtalloc_noclone(&ro, NO_CLONING);
3184	#endif
3185	rt = ro.ro_rt;
3186	break;
3187	#endif /* INET */
3188	#ifdef INET6
3189	case AF_INET6:
3190	hlen = sizeof(struct ip6_hdr);
3191	bzero(&ro6, sizeof(ro6));
3192	dst6 = (struct sockaddr_in6 *)&ro6.ro_dst;
3193	dst6->sin6_family = AF_INET6;
3194	dst6->sin6_len = sizeof(*dst6);
3195	dst6->sin6_addr = addr->v6;
3196	#ifdef __FreeBSD__
3197	in6_rtalloc_ign(&ro6, 0, rtableid);
3198	#else /* ! __FreeBSD__ */
3199	rtalloc_noclone((struct route *)&ro6, NO_CLONING);
3200	#endif
3201	rt = ro6.ro_rt;
3202	break;
3203	#endif /* INET6 */
3204	}
3205
3206	if (rt && rt->rt_ifp) {
3207	mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr);
3208	#ifdef __FreeBSD__
3209	mss = max(V_tcp_mssdflt, mss);
3210	#else
3211	mss = max(tcp_mssdflt, mss);
3212	#endif
3213	RTFREE(rt);
3214	}
3215	mss = min(mss, offer);
3216	mss = max(mss, 64); /* sanity - at least max opt space */
3217	return (mss);
3218	}
3219
3220	void
3221	pf_set_rt_ifp(struct pf_state s, struct pf_addr saddr)
3222	{
3223	struct pf_rule *r = s->rule.ptr;
3224	struct pf_src_node *sn = NULL;
3225
3226	s->rt_kif = NULL;
3227	if (!r->rt \|\| r->rt == PF_FASTROUTE)
3228	return;
3229	switch (s->key[PF_SK_WIRE]->af) {
3230	#ifdef INET
3231	case AF_INET:
3232	pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL, &sn);
3233	s->rt_kif = r->rpool.cur->kif;
3234	break;
3235	#endif /* INET */
3236	#ifdef INET6
3237	case AF_INET6:
3238	pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL, &sn);
3239	s->rt_kif = r->rpool.cur->kif;
3240	break;
3241	#endif /* INET6 */
3242	}
3243	}
3244
3245	u_int32_t
3246	pf_tcp_iss(struct pf_pdesc *pd)
3247	{
3248	MD5_CTX ctx;
3249	u_int32_t digest[4];
3250
3251	#ifdef __FreeBSD__
3252	if (V_pf_tcp_secret_init == 0) {
3253	read_random(&V_pf_tcp_secret, sizeof(V_pf_tcp_secret));
3254	MD5Init(&V_pf_tcp_secret_ctx);
3255	MD5Update(&V_pf_tcp_secret_ctx, V_pf_tcp_secret,
3256	sizeof(V_pf_tcp_secret));
3257	V_pf_tcp_secret_init = 1;
3258	}
3259
3260	ctx = V_pf_tcp_secret_ctx;
3261	#else
3262	if (pf_tcp_secret_init == 0) {
3263	arc4random_buf(pf_tcp_secret, sizeof(pf_tcp_secret));
3264	MD5Init(&pf_tcp_secret_ctx);
3265	MD5Update(&pf_tcp_secret_ctx, pf_tcp_secret,
3266	sizeof(pf_tcp_secret));
3267	pf_tcp_secret_init = 1;
3268	}
3269
3270	ctx = pf_tcp_secret_ctx;
3271	#endif
3272
3273	MD5Update(&ctx, (char *)&pd->hdr.tcp->th_sport, sizeof(u_short));
3274	MD5Update(&ctx, (char *)&pd->hdr.tcp->th_dport, sizeof(u_short));
3275	if (pd->af == AF_INET6) {
3276	MD5Update(&ctx, (char *)&pd->src->v6, sizeof(struct in6_addr));
3277	MD5Update(&ctx, (char *)&pd->dst->v6, sizeof(struct in6_addr));
3278	} else {
3279	MD5Update(&ctx, (char *)&pd->src->v4, sizeof(struct in_addr));
3280	MD5Update(&ctx, (char *)&pd->dst->v4, sizeof(struct in_addr));
3281	}
3282	MD5Final((u_char *)digest, &ctx);
3283	#ifdef __FreeBSD__
3284	V_pf_tcp_iss_off += 4096;
3285	#define ISN_RANDOM_INCREMENT (4096 - 1)
3286	return (digest[0] + (arc4random() & ISN_RANDOM_INCREMENT) +
3287	V_pf_tcp_iss_off);
3288	#undef ISN_RANDOM_INCREMENT
3289	#else
3290	pf_tcp_iss_off += 4096;
3291	return (digest[0] + tcp_iss + pf_tcp_iss_off);
3292	#endif
3293	}
3294
3295	int
3296	pf_test_rule(struct pf_rule rm, struct pf_state sm, int direction,
3297	struct pfi_kif kif, struct mbuf m, int off, void *h,
3298	struct pf_pdesc pd, struct pf_rule am, struct pf_ruleset *rsm,
3299	#ifdef __FreeBSD__
3300	struct ifqueue ifq, struct inpcb inp)
3301	#else
3302	struct ifqueue *ifq)
3303	#endif
3304	{
3305	struct pf_rule *nr = NULL;
3306	struct pf_addr saddr = pd->src, daddr = pd->dst;
3307	sa_family_t af = pd->af;
3308	struct pf_rule r, a = NULL;
3309	struct pf_ruleset *ruleset = NULL;
3310	struct pf_src_node *nsn = NULL;
3311	struct tcphdr *th = pd->hdr.tcp;
3312	struct pf_state_key skw = NULL, sks = NULL;
3313	struct pf_state_key sk = NULL, nk = NULL;
3314	u_short reason;
3315	int rewrite = 0, hdrlen = 0;
3316	int tag = -1, rtableid = -1;
3317	int asd = 0;
3318	int match = 0;
3319	int state_icmp = 0;
3320	#ifdef __FreeBSD__
3321	u_int16_t sport = 0, dport = 0;
3322	u_int16_t bproto_sum = 0, bip_sum = 0;
3323	#else
3324	u_int16_t sport, dport;
3325	u_int16_t bproto_sum = 0, bip_sum;
3326	#endif
3327	u_int8_t icmptype = 0, icmpcode = 0;
3328
3329
3330	if (direction == PF_IN && pf_check_congestion(ifq)) {
3331	REASON_SET(&reason, PFRES_CONGEST);
3332	return (PF_DROP);
3333	}
3334
3335	#ifdef __FreeBSD__
3336	if (inp != NULL)
3337	pd->lookup.done = pf_socket_lookup(direction, pd, inp);
3338	else if (V_debug_pfugidhack) {
3339	PF_UNLOCK();
3340	DPFPRINTF(PF_DEBUG_MISC, ("pf: unlocked lookup\n"));
3341	pd->lookup.done = pf_socket_lookup(direction, pd, inp);
3342	PF_LOCK();
3343	}
3344	#endif
3345
3346	switch (pd->proto) {
3347	case IPPROTO_TCP:
3348	sport = th->th_sport;
3349	dport = th->th_dport;
3350	hdrlen = sizeof(*th);
3351	break;
3352	case IPPROTO_UDP:
3353	sport = pd->hdr.udp->uh_sport;
3354	dport = pd->hdr.udp->uh_dport;
3355	hdrlen = sizeof(*pd->hdr.udp);
3356	break;
3357	#ifdef INET
3358	case IPPROTO_ICMP:
3359	if (pd->af != AF_INET)
3360	break;
3361	sport = dport = pd->hdr.icmp->icmp_id;
3362	hdrlen = sizeof(*pd->hdr.icmp);
3363	icmptype = pd->hdr.icmp->icmp_type;
3364	icmpcode = pd->hdr.icmp->icmp_code;
3365
3366	if (icmptype == ICMP_UNREACH \|\|
3367	icmptype == ICMP_SOURCEQUENCH \|\|
3368	icmptype == ICMP_REDIRECT \|\|
3369	icmptype == ICMP_TIMXCEED \|\|
3370	icmptype == ICMP_PARAMPROB)
3371	state_icmp++;
3372	break;
3373	#endif /* INET */
3374	#ifdef INET6
3375	case IPPROTO_ICMPV6:
3376	if (af != AF_INET6)
3377	break;
3378	sport = dport = pd->hdr.icmp6->icmp6_id;
3379	hdrlen = sizeof(*pd->hdr.icmp6);
3380	icmptype = pd->hdr.icmp6->icmp6_type;
3381	icmpcode = pd->hdr.icmp6->icmp6_code;
3382
3383	if (icmptype == ICMP6_DST_UNREACH \|\|
3384	icmptype == ICMP6_PACKET_TOO_BIG \|\|
3385	icmptype == ICMP6_TIME_EXCEEDED \|\|
3386	icmptype == ICMP6_PARAM_PROB)
3387	state_icmp++;
3388	break;
3389	#endif /* INET6 */
3390	default:
3391	sport = dport = hdrlen = 0;
3392	break;
3393	}
3394
3395	r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3396
3397	/* check packet for BINAT/NAT/RDR */
3398	if ((nr = pf_get_translation(pd, m, off, direction, kif, &nsn,
3399	&skw, &sks, &sk, &nk, saddr, daddr, sport, dport)) != NULL) {
3400	if (nk == NULL \|\| sk == NULL) {
3401	REASON_SET(&reason, PFRES_MEMORY);
3402	goto cleanup;
3403	}
3404
3405	if (pd->ip_sum)
3406	bip_sum = *pd->ip_sum;
3407
3408	switch (pd->proto) {
3409	case IPPROTO_TCP:
3410	bproto_sum = th->th_sum;
3411	pd->proto_sum = &th->th_sum;
3412
3413	if (PF_ANEQ(saddr, &nk->addr[pd->sidx], af) \|\|
3414	nk->port[pd->sidx] != sport) {
3415	pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
3416	&th->th_sum, &nk->addr[pd->sidx],
3417	nk->port[pd->sidx], 0, af);
3418	pd->sport = &th->th_sport;
3419	sport = th->th_sport;
3420	}
3421
3422	if (PF_ANEQ(daddr, &nk->addr[pd->didx], af) \|\|
3423	nk->port[pd->didx] != dport) {
3424	pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
3425	&th->th_sum, &nk->addr[pd->didx],
3426	nk->port[pd->didx], 0, af);
3427	dport = th->th_dport;
3428	pd->dport = &th->th_dport;
3429	}
3430	rewrite++;
3431	break;
3432	case IPPROTO_UDP:
3433	bproto_sum = pd->hdr.udp->uh_sum;
3434	pd->proto_sum = &pd->hdr.udp->uh_sum;
3435
3436	if (PF_ANEQ(saddr, &nk->addr[pd->sidx], af) \|\|
3437	nk->port[pd->sidx] != sport) {
3438	pf_change_ap(saddr, &pd->hdr.udp->uh_sport,
3439	pd->ip_sum, &pd->hdr.udp->uh_sum,
3440	&nk->addr[pd->sidx],
3441	nk->port[pd->sidx], 1, af);
3442	sport = pd->hdr.udp->uh_sport;
3443	pd->sport = &pd->hdr.udp->uh_sport;
3444	}
3445
3446	if (PF_ANEQ(daddr, &nk->addr[pd->didx], af) \|\|
3447	nk->port[pd->didx] != dport) {
3448	pf_change_ap(daddr, &pd->hdr.udp->uh_dport,
3449	pd->ip_sum, &pd->hdr.udp->uh_sum,
3450	&nk->addr[pd->didx],
3451	nk->port[pd->didx], 1, af);
3452	dport = pd->hdr.udp->uh_dport;
3453	pd->dport = &pd->hdr.udp->uh_dport;
3454	}
3455	rewrite++;
3456	break;
3457	#ifdef INET
3458	case IPPROTO_ICMP:
3459	nk->port[0] = nk->port[1];
3460	if (PF_ANEQ(saddr, &nk->addr[pd->sidx], AF_INET))
3461	pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3462	nk->addr[pd->sidx].v4.s_addr, 0);
3463
3464	if (PF_ANEQ(daddr, &nk->addr[pd->didx], AF_INET))
3465	pf_change_a(&daddr->v4.s_addr, pd->ip_sum,
3466	nk->addr[pd->didx].v4.s_addr, 0);
3467
3468	if (nk->port[1] != pd->hdr.icmp->icmp_id) {
3469	pd->hdr.icmp->icmp_cksum = pf_cksum_fixup(
3470	pd->hdr.icmp->icmp_cksum, sport,
3471	nk->port[1], 0);
3472	pd->hdr.icmp->icmp_id = nk->port[1];
3473	pd->sport = &pd->hdr.icmp->icmp_id;
3474	}
3475	m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
3476	break;
3477	#endif /* INET */
3478	#ifdef INET6
3479	case IPPROTO_ICMPV6:
3480	nk->port[0] = nk->port[1];
3481	if (PF_ANEQ(saddr, &nk->addr[pd->sidx], AF_INET6))
3482	pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum,
3483	&nk->addr[pd->sidx], 0);
3484
3485	if (PF_ANEQ(daddr, &nk->addr[pd->didx], AF_INET6))
3486	pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum,
3487	&nk->addr[pd->didx], 0);
3488	rewrite++;
3489	break;
3490	#endif /* INET */
3491	default:
3492	switch (af) {
3493	#ifdef INET
3494	case AF_INET:
3495	if (PF_ANEQ(saddr,
3496	&nk->addr[pd->sidx], AF_INET))
3497	pf_change_a(&saddr->v4.s_addr,
3498	pd->ip_sum,
3499	nk->addr[pd->sidx].v4.s_addr, 0);
3500
3501	if (PF_ANEQ(daddr,
3502	&nk->addr[pd->didx], AF_INET))
3503	pf_change_a(&daddr->v4.s_addr,
3504	pd->ip_sum,
3505	nk->addr[pd->didx].v4.s_addr, 0);
3506	break;
3507	#endif /* INET */
3508	#ifdef INET6
3509	case AF_INET6:
3510	if (PF_ANEQ(saddr,
3511	&nk->addr[pd->sidx], AF_INET6))
3512	PF_ACPY(saddr, &nk->addr[pd->sidx], af);
3513
3514	if (PF_ANEQ(daddr,
3515	&nk->addr[pd->didx], AF_INET6))
3516	PF_ACPY(saddr, &nk->addr[pd->didx], af);
3517	break;
3518	#endif /* INET */
3519	}
3520	break;
3521	}
3522	if (nr->natpass)
3523	r = NULL;
3524	pd->nat_rule = nr;
3525	}
3526
3527	while (r != NULL) {
3528	r->evaluations++;
3529	if (pfi_kif_match(r->kif, kif) == r->ifnot)
3530	r = r->skip[PF_SKIP_IFP].ptr;
3531	else if (r->direction && r->direction != direction)
3532	r = r->skip[PF_SKIP_DIR].ptr;
3533	else if (r->af && r->af != af)
3534	r = r->skip[PF_SKIP_AF].ptr;
3535	else if (r->proto && r->proto != pd->proto)
3536	r = r->skip[PF_SKIP_PROTO].ptr;
3537	else if (PF_MISMATCHAW(&r->src.addr, saddr, af,
3538	r->src.neg, kif, M_GETFIB(m)))
3539	r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3540	/* tcp/udp only. port_op always 0 in other cases */
3541	else if (r->src.port_op && !pf_match_port(r->src.port_op,
3542	r->src.port[0], r->src.port[1], sport))
3543	r = r->skip[PF_SKIP_SRC_PORT].ptr;
3544	else if (PF_MISMATCHAW(&r->dst.addr, daddr, af,
3545	r->dst.neg, NULL, M_GETFIB(m)))
3546	r = r->skip[PF_SKIP_DST_ADDR].ptr;
3547	/* tcp/udp only. port_op always 0 in other cases */
3548	else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
3549	r->dst.port[0], r->dst.port[1], dport))
3550	r = r->skip[PF_SKIP_DST_PORT].ptr;
3551	/* icmp only. type always 0 in other cases */
3552	else if (r->type && r->type != icmptype + 1)
3553	r = TAILQ_NEXT(r, entries);
3554	/* icmp only. type always 0 in other cases */
3555	else if (r->code && r->code != icmpcode + 1)
3556	r = TAILQ_NEXT(r, entries);
3557	else if (r->tos && !(r->tos == pd->tos))
3558	r = TAILQ_NEXT(r, entries);
3559	else if (r->rule_flag & PFRULE_FRAGMENT)
3560	r = TAILQ_NEXT(r, entries);
3561	else if (pd->proto == IPPROTO_TCP &&
3562	(r->flagset & th->th_flags) != r->flags)
3563	r = TAILQ_NEXT(r, entries);
3564	/* tcp/udp only. uid.op always 0 in other cases */
3565	else if (r->uid.op && (pd->lookup.done \|\| (pd->lookup.done =
3566	#ifdef __FreeBSD__
3567	pf_socket_lookup(direction, pd, inp), 1)) &&
3568	#else
3569	pf_socket_lookup(direction, pd), 1)) &&
3570	#endif
3571	!pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
3572	pd->lookup.uid))
3573	r = TAILQ_NEXT(r, entries);
3574	/* tcp/udp only. gid.op always 0 in other cases */
3575	else if (r->gid.op && (pd->lookup.done \|\| (pd->lookup.done =
3576	#ifdef __FreeBSD__
3577	pf_socket_lookup(direction, pd, inp), 1)) &&
3578	#else
3579	pf_socket_lookup(direction, pd), 1)) &&
3580	#endif
3581	!pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
3582	pd->lookup.gid))
3583	r = TAILQ_NEXT(r, entries);
3584	else if (r->prob &&
3585	#ifdef __FreeBSD__
3586	r->prob <= arc4random())
3587	#else
3588	r->prob <= arc4random_uniform(UINT_MAX - 1) + 1)
3589	#endif
3590	r = TAILQ_NEXT(r, entries);
3591	#ifdef __FreeBSD__
3592	else if (r->match_tag && !pf_match_tag(m, r, &tag, pd->pf_mtag))
3593	#else
3594	else if (r->match_tag && !pf_match_tag(m, r, &tag))
3595	#endif
3596	r = TAILQ_NEXT(r, entries);
3597	else if (r->os_fingerprint != PF_OSFP_ANY &&
3598	(pd->proto != IPPROTO_TCP \|\| !pf_osfp_match(
3599	pf_osfp_fingerprint(pd, m, off, th),
3600	r->os_fingerprint)))
3601	r = TAILQ_NEXT(r, entries);
3602	else {
3603	if (r->tag)
3604	tag = r->tag;
3605	if (r->rtableid >= 0)
3606	rtableid = r->rtableid;
3607	if (r->anchor == NULL) {
3608	match = 1;
3609	*rm = r;
3610	*am = a;
3611	*rsm = ruleset;
3612	if ((*rm)->quick)
3613	break;
3614	r = TAILQ_NEXT(r, entries);
3615	} else
3616	pf_step_into_anchor(&asd, &ruleset,
3617	PF_RULESET_FILTER, &r, &a, &match);
3618	}
3619	if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3620	PF_RULESET_FILTER, &r, &a, &match))
3621	break;
3622	}
3623	r = *rm;
3624	a = *am;
3625	ruleset = *rsm;
3626
3627	REASON_SET(&reason, PFRES_MATCH);
3628
3629	if (r->log \|\| (nr != NULL && nr->log)) {
3630	if (rewrite)
3631	m_copyback(m, off, hdrlen, pd->hdr.any);
3632	PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr,
3633	a, ruleset, pd);
3634	}
3635
3636	if ((r->action == PF_DROP) &&
3637	((r->rule_flag & PFRULE_RETURNRST) \|\|
3638	(r->rule_flag & PFRULE_RETURNICMP) \|\|
3639	(r->rule_flag & PFRULE_RETURN))) {
3640	/* undo NAT changes, if they have taken place */
3641	if (nr != NULL) {
3642	PF_ACPY(saddr, &sk->addr[pd->sidx], af);
3643	PF_ACPY(daddr, &sk->addr[pd->didx], af);
3644	if (pd->sport)
3645	*pd->sport = sk->port[pd->sidx];
3646	if (pd->dport)
3647	*pd->dport = sk->port[pd->didx];
3648	if (pd->proto_sum)
3649	*pd->proto_sum = bproto_sum;
3650	if (pd->ip_sum)
3651	*pd->ip_sum = bip_sum;
3652	m_copyback(m, off, hdrlen, pd->hdr.any);
3653	}
3654	if (pd->proto == IPPROTO_TCP &&
3655	((r->rule_flag & PFRULE_RETURNRST) \|\|
3656	(r->rule_flag & PFRULE_RETURN)) &&
3657	!(th->th_flags & TH_RST)) {
3658	u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
3659	int len = 0;
3660	#ifdef INET
3661	struct ip *h4;
3662	#endif
3663	#ifdef INET6
3664	struct ip6_hdr *h6;
3665	#endif
3666
3667	switch (af) {
3668	#ifdef INET
3669	case AF_INET:
3670	h4 = mtod(m, struct ip *);
3671	len = ntohs(h4->ip_len) - off;
3672	break;
3673	#endif
3674	#ifdef INET6
3675	case AF_INET6:
3676	h6 = mtod(m, struct ip6_hdr *);
3677	len = ntohs(h6->ip6_plen) - (off - sizeof(*h6));
3678	break;
3679	#endif
3680	}
3681
3682	if (pf_check_proto_cksum(m, off, len, IPPROTO_TCP, af))
3683	REASON_SET(&reason, PFRES_PROTCKSUM);
3684	else {
3685	if (th->th_flags & TH_SYN)
3686	ack++;
3687	if (th->th_flags & TH_FIN)
3688	ack++;
3689	#ifdef __FreeBSD__
3690	pf_send_tcp(m, r, af, pd->dst,
3691	#else
3692	pf_send_tcp(r, af, pd->dst,
3693	#endif
3694	pd->src, th->th_dport, th->th_sport,
3695	ntohl(th->th_ack), ack, TH_RST\|TH_ACK, 0, 0,
3696	r->return_ttl, 1, 0, pd->eh, kif->pfik_ifp);
3697	}
3698	} else if (pd->proto != IPPROTO_ICMP && af == AF_INET &&
3699	r->return_icmp)
3700	pf_send_icmp(m, r->return_icmp >> 8,
3701	r->return_icmp & 255, af, r);
3702	else if (pd->proto != IPPROTO_ICMPV6 && af == AF_INET6 &&
3703	r->return_icmp6)
3704	pf_send_icmp(m, r->return_icmp6 >> 8,
3705	r->return_icmp6 & 255, af, r);
3706	}
3707
3708	if (r->action == PF_DROP)
3709	goto cleanup;
3710
3711	#ifdef __FreeBSD__
3712	if (pf_tag_packet(m, tag, rtableid, pd->pf_mtag)) {
3713	#else
3714	if (pf_tag_packet(m, tag, rtableid)) {
3715	#endif
3716	REASON_SET(&reason, PFRES_MEMORY);
3717	goto cleanup;
3718	}
3719
3720	if (!state_icmp && (r->keep_state \|\| nr != NULL \|\|
3721	(pd->flags & PFDESC_TCP_NORM))) {
3722	int action;
3723	action = pf_create_state(r, nr, a, pd, nsn, skw, sks, nk, sk, m,
3724	off, sport, dport, &rewrite, kif, sm, tag, bproto_sum,
3725	bip_sum, hdrlen);
3726	if (action != PF_PASS)
3727	return (action);
3728	} else {
3729	#ifdef __FreeBSD__
3730	if (sk != NULL)
3731	pool_put(&V_pf_state_key_pl, sk);
3732	if (nk != NULL)
3733	pool_put(&V_pf_state_key_pl, nk);
3734	#else
3735	if (sk != NULL)
3736	pool_put(&pf_state_key_pl, sk);
3737	if (nk != NULL)
3738	pool_put(&pf_state_key_pl, nk);
3739	#endif
3740	}
3741
3742	/* copy back packet headers if we performed NAT operations */
3743	if (rewrite)
3744	m_copyback(m, off, hdrlen, pd->hdr.any);
3745
3746	#if NPFSYNC > 0
3747	if (sm != NULL && !ISSET((sm)->state_flags, PFSTATE_NOSYNC) &&
3748	#ifdef __FreeBSD__
3749	direction == PF_OUT && pfsync_up_ptr != NULL && pfsync_up_ptr()) {
3750	#else
3751	direction == PF_OUT && pfsync_up()) {
3752	#endif
3753	/*
3754	* We want the state created, but we dont
3755	* want to send this in case a partner
3756	* firewall has to know about it to allow
3757	* replies through it.
3758	*/
3759	#ifdef __FreeBSD__
3760	if (pfsync_defer_ptr != NULL &&
3761	pfsync_defer_ptr(*sm, m))
3762	#else
3763	if (pfsync_defer(*sm, m))
3764	#endif
3765	return (PF_DEFER);
3766	}
3767	#endif
3768
3769	return (PF_PASS);
3770
3771	cleanup:
3772	#ifdef __FreeBSD__
3773	if (sk != NULL)
3774	pool_put(&V_pf_state_key_pl, sk);
3775	if (nk != NULL)
3776	pool_put(&V_pf_state_key_pl, nk);
3777	#else
3778	if (sk != NULL)
3779	pool_put(&pf_state_key_pl, sk);
3780	if (nk != NULL)
3781	pool_put(&pf_state_key_pl, nk);
3782	#endif
3783	return (PF_DROP);
3784	}
3785
3786	static __inline int
3787	pf_create_state(struct pf_rule r, struct pf_rule nr, struct pf_rule *a,
3788	struct pf_pdesc pd, struct pf_src_node nsn, struct pf_state_key *skw,
3789	struct pf_state_key sks, struct pf_state_key nk, struct pf_state_key *sk,
3790	struct mbuf m, int off, u_int16_t sport, u_int16_t dport, int rewrite,
3791	struct pfi_kif kif, struct pf_state *sm, int tag, u_int16_t bproto_sum,
3792	u_int16_t bip_sum, int hdrlen)
3793	{
3794	struct pf_state *s = NULL;
3795	struct pf_src_node *sn = NULL;
3796	struct tcphdr *th = pd->hdr.tcp;
3797	#ifdef __FreeBSD__
3798	u_int16_t mss = V_tcp_mssdflt;
3799	#else
3800	u_int16_t mss = tcp_mssdflt;
3801	#endif
3802	u_short reason;
3803
3804	/* check maximums */
3805	if (r->max_states && (r->states_cur >= r->max_states)) {
3806	#ifdef __FreeBSD__
3807	V_pf_status.lcounters[LCNT_STATES]++;
3808	#else
3809	pf_status.lcounters[LCNT_STATES]++;
3810	#endif
3811	REASON_SET(&reason, PFRES_MAXSTATES);
3812	return (PF_DROP);
3813	}
3814	/* src node for filter rule */
3815	if ((r->rule_flag & PFRULE_SRCTRACK \|\|
3816	r->rpool.opts & PF_POOL_STICKYADDR) &&
3817	pf_insert_src_node(&sn, r, pd->src, pd->af) != 0) {
3818	REASON_SET(&reason, PFRES_SRCLIMIT);
3819	goto csfailed;
3820	}
3821	/* src node for translation rule */
3822	if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3823	pf_insert_src_node(&nsn, nr, &sk->addr[pd->sidx], pd->af)) {
3824	REASON_SET(&reason, PFRES_SRCLIMIT);
3825	goto csfailed;
3826	}
3827	#ifdef __FreeBSD__
3828	s = pool_get(&V_pf_state_pl, PR_NOWAIT \| PR_ZERO);
3829	#else
3830	s = pool_get(&pf_state_pl, PR_NOWAIT \| PR_ZERO);
3831	#endif
3832	if (s == NULL) {
3833	REASON_SET(&reason, PFRES_MEMORY);
3834	goto csfailed;
3835	}
3836	s->rule.ptr = r;
3837	s->nat_rule.ptr = nr;
3838	s->anchor.ptr = a;
3839	STATE_INC_COUNTERS(s);
3840	if (r->allow_opts)
3841	s->state_flags \|= PFSTATE_ALLOWOPTS;
3842	if (r->rule_flag & PFRULE_STATESLOPPY)
3843	s->state_flags \|= PFSTATE_SLOPPY;
3844	if (r->rule_flag & PFRULE_PFLOW)
3845	s->state_flags \|= PFSTATE_PFLOW;
3846	s->log = r->log & PF_LOG_ALL;
3847	s->sync_state = PFSYNC_S_NONE;
3848	if (nr != NULL)
3849	s->log \|= nr->log & PF_LOG_ALL;
3850	switch (pd->proto) {
3851	case IPPROTO_TCP:
3852	s->src.seqlo = ntohl(th->th_seq);
3853	s->src.seqhi = s->src.seqlo + pd->p_len + 1;
3854	if ((th->th_flags & (TH_SYN\|TH_ACK)) == TH_SYN &&
3855	r->keep_state == PF_STATE_MODULATE) {
3856	/* Generate sequence number modulator */
3857	if ((s->src.seqdiff = pf_tcp_iss(pd) - s->src.seqlo) ==
3858	0)
3859	s->src.seqdiff = 1;
3860	pf_change_a(&th->th_seq, &th->th_sum,
3861	htonl(s->src.seqlo + s->src.seqdiff), 0);
3862	*rewrite = 1;
3863	} else
3864	s->src.seqdiff = 0;
3865	if (th->th_flags & TH_SYN) {
3866	s->src.seqhi++;
3867	s->src.wscale = pf_get_wscale(m, off,
3868	th->th_off, pd->af);
3869	}
3870	s->src.max_win = MAX(ntohs(th->th_win), 1);
3871	if (s->src.wscale & PF_WSCALE_MASK) {
3872	/* Remove scale factor from initial window */
3873	int win = s->src.max_win;
3874	win += 1 << (s->src.wscale & PF_WSCALE_MASK);
3875	s->src.max_win = (win - 1) >>
3876	(s->src.wscale & PF_WSCALE_MASK);
3877	}
3878	if (th->th_flags & TH_FIN)
3879	s->src.seqhi++;
3880	s->dst.seqhi = 1;
3881	s->dst.max_win = 1;
3882	s->src.state = TCPS_SYN_SENT;
3883	s->dst.state = TCPS_CLOSED;
3884	s->timeout = PFTM_TCP_FIRST_PACKET;
3885	break;
3886	case IPPROTO_UDP:
3887	s->src.state = PFUDPS_SINGLE;
3888	s->dst.state = PFUDPS_NO_TRAFFIC;
3889	s->timeout = PFTM_UDP_FIRST_PACKET;
3890	break;
3891	case IPPROTO_ICMP:
3892	#ifdef INET6
3893	case IPPROTO_ICMPV6:
3894	#endif
3895	s->timeout = PFTM_ICMP_FIRST_PACKET;
3896	break;
3897	default:
3898	s->src.state = PFOTHERS_SINGLE;
3899	s->dst.state = PFOTHERS_NO_TRAFFIC;
3900	s->timeout = PFTM_OTHER_FIRST_PACKET;
3901	}
3902
3903	s->creation = time_second;
3904	s->expire = time_second;
3905
3906	if (sn != NULL) {
3907	s->src_node = sn;
3908	s->src_node->states++;
3909	}
3910	if (nsn != NULL) {
3911	/* XXX We only modify one side for now. */
3912	PF_ACPY(&nsn->raddr, &nk->addr[1], pd->af);
3913	s->nat_src_node = nsn;
3914	s->nat_src_node->states++;
3915	}
3916	if (pd->proto == IPPROTO_TCP) {
3917	if ((pd->flags & PFDESC_TCP_NORM) && pf_normalize_tcp_init(m,
3918	off, pd, th, &s->src, &s->dst)) {
3919	REASON_SET(&reason, PFRES_MEMORY);
3920	pf_src_tree_remove_state(s);
3921	STATE_DEC_COUNTERS(s);
3922	#ifdef __FreeBSD__
3923	pool_put(&V_pf_state_pl, s);
3924	#else
3925	pool_put(&pf_state_pl, s);
3926	#endif
3927	return (PF_DROP);
3928	}
3929	if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
3930	pf_normalize_tcp_stateful(m, off, pd, &reason, th, s,
3931	&s->src, &s->dst, rewrite)) {
3932	/* This really shouldn't happen!!! */
3933	DPFPRINTF(PF_DEBUG_URGENT,
3934	("pf_normalize_tcp_stateful failed on first pkt"));
3935	pf_normalize_tcp_cleanup(s);
3936	pf_src_tree_remove_state(s);
3937	STATE_DEC_COUNTERS(s);
3938	#ifdef __FreeBSD__
3939	pool_put(&V_pf_state_pl, s);
3940	#else
3941	pool_put(&pf_state_pl, s);
3942	#endif
3943	return (PF_DROP);
3944	}
3945	}
3946	s->direction = pd->dir;
3947
3948	if (sk == NULL && pf_state_key_setup(pd, nr, &skw, &sks, &sk, &nk,
3949	pd->src, pd->dst, sport, dport))
3950	goto csfailed;
3951
3952	if (pf_state_insert(BOUND_IFACE(r, kif), skw, sks, s)) {
3953	if (pd->proto == IPPROTO_TCP)
3954	pf_normalize_tcp_cleanup(s);
3955	REASON_SET(&reason, PFRES_STATEINS);
3956	pf_src_tree_remove_state(s);
3957	STATE_DEC_COUNTERS(s);
3958	#ifdef __FreeBSD__
3959	pool_put(&V_pf_state_pl, s);
3960	#else
3961	pool_put(&pf_state_pl, s);
3962	#endif
3963	return (PF_DROP);
3964	} else
3965	*sm = s;
3966
3967	pf_set_rt_ifp(s, pd->src); /* needs s->state_key set */
3968	if (tag > 0) {
3969	pf_tag_ref(tag);
3970	s->tag = tag;
3971	}
3972	if (pd->proto == IPPROTO_TCP && (th->th_flags & (TH_SYN\|TH_ACK)) ==
3973	TH_SYN && r->keep_state == PF_STATE_SYNPROXY) {
3974	s->src.state = PF_TCPS_PROXY_SRC;
3975	/* undo NAT changes, if they have taken place */
3976	if (nr != NULL) {
3977	struct pf_state_key *skt = s->key[PF_SK_WIRE];
3978	if (pd->dir == PF_OUT)
3979	skt = s->key[PF_SK_STACK];
3980	PF_ACPY(pd->src, &skt->addr[pd->sidx], pd->af);
3981	PF_ACPY(pd->dst, &skt->addr[pd->didx], pd->af);
3982	if (pd->sport)
3983	*pd->sport = skt->port[pd->sidx];
3984	if (pd->dport)
3985	*pd->dport = skt->port[pd->didx];
3986	if (pd->proto_sum)
3987	*pd->proto_sum = bproto_sum;
3988	if (pd->ip_sum)
3989	*pd->ip_sum = bip_sum;
3990	m_copyback(m, off, hdrlen, pd->hdr.any);
3991	}
3992	s->src.seqhi = htonl(arc4random());
3993	/* Find mss option */
3994	int rtid = M_GETFIB(m);
3995	mss = pf_get_mss(m, off, th->th_off, pd->af);
3996	mss = pf_calc_mss(pd->src, pd->af, rtid, mss);
3997	mss = pf_calc_mss(pd->dst, pd->af, rtid, mss);
3998	s->src.mss = mss;
3999	#ifdef __FreeBSD__
4000	pf_send_tcp(NULL, r, pd->af, pd->dst, pd->src, th->th_dport,
4001	#else
4002	pf_send_tcp(r, pd->af, pd->dst, pd->src, th->th_dport,
4003	#endif
4004	th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1,
4005	TH_SYN\|TH_ACK, 0, s->src.mss, 0, 1, 0, NULL, NULL);
4006	REASON_SET(&reason, PFRES_SYNPROXY);
4007	return (PF_SYNPROXY_DROP);
4008	}
4009
4010	return (PF_PASS);
4011
4012	csfailed:
4013	#ifdef __FreeBSD__
4014	if (sk != NULL)
4015	pool_put(&V_pf_state_key_pl, sk);
4016	if (nk != NULL)
4017	pool_put(&V_pf_state_key_pl, nk);
4018	#else
4019	if (sk != NULL)
4020	pool_put(&pf_state_key_pl, sk);
4021	if (nk != NULL)
4022	pool_put(&pf_state_key_pl, nk);
4023	#endif
4024
4025	if (sn != NULL && sn->states == 0 && sn->expire == 0) {
4026	#ifdef __FreeBSD__
4027	RB_REMOVE(pf_src_tree, &V_tree_src_tracking, sn);
4028	V_pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
4029	V_pf_status.src_nodes--;
4030	pool_put(&V_pf_src_tree_pl, sn);
4031	#else
4032	RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
4033	pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
4034	pf_status.src_nodes--;
4035	pool_put(&pf_src_tree_pl, sn);
4036	#endif
4037	}
4038	if (nsn != sn && nsn != NULL && nsn->states == 0 && nsn->expire == 0) {
4039	#ifdef __FreeBSD__
4040	RB_REMOVE(pf_src_tree, &V_tree_src_tracking, nsn);
4041	V_pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
4042	V_pf_status.src_nodes--;
4043	pool_put(&V_pf_src_tree_pl, nsn);
4044	#else
4045	RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
4046	pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
4047	pf_status.src_nodes--;
4048	pool_put(&pf_src_tree_pl, nsn);
4049	#endif
4050	}
4051	return (PF_DROP);
4052	}
4053
4054	int
4055	pf_test_fragment(struct pf_rule *rm, int direction, struct pfi_kif kif,
4056	struct mbuf m, void h, struct pf_pdesc pd, struct pf_rule *am,
4057	struct pf_ruleset **rsm)
4058	{
4059	struct pf_rule r, a = NULL;
4060	struct pf_ruleset *ruleset = NULL;
4061	sa_family_t af = pd->af;
4062	u_short reason;
4063	int tag = -1;
4064	int asd = 0;
4065	int match = 0;
4066
4067	r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
4068	while (r != NULL) {
4069	r->evaluations++;
4070	if (pfi_kif_match(r->kif, kif) == r->ifnot)
4071	r = r->skip[PF_SKIP_IFP].ptr;
4072	else if (r->direction && r->direction != direction)
4073	r = r->skip[PF_SKIP_DIR].ptr;
4074	else if (r->af && r->af != af)
4075	r = r->skip[PF_SKIP_AF].ptr;
4076	else if (r->proto && r->proto != pd->proto)
4077	r = r->skip[PF_SKIP_PROTO].ptr;
4078	else if (PF_MISMATCHAW(&r->src.addr, pd->src, af,
4079	r->src.neg, kif, M_GETFIB(m)))
4080	r = r->skip[PF_SKIP_SRC_ADDR].ptr;
4081	else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af,
4082	r->dst.neg, NULL, M_GETFIB(m)))
4083	r = r->skip[PF_SKIP_DST_ADDR].ptr;
4084	else if (r->tos && !(r->tos == pd->tos))
4085	r = TAILQ_NEXT(r, entries);
4086	else if (r->os_fingerprint != PF_OSFP_ANY)
4087	r = TAILQ_NEXT(r, entries);
4088	else if (pd->proto == IPPROTO_UDP &&
4089	(r->src.port_op \|\| r->dst.port_op))
4090	r = TAILQ_NEXT(r, entries);
4091	else if (pd->proto == IPPROTO_TCP &&
4092	(r->src.port_op \|\| r->dst.port_op \|\| r->flagset))
4093	r = TAILQ_NEXT(r, entries);
4094	else if ((pd->proto == IPPROTO_ICMP \|\|
4095	pd->proto == IPPROTO_ICMPV6) &&
4096	(r->type \|\| r->code))
4097	r = TAILQ_NEXT(r, entries);
4098	else if (r->prob && r->prob <=
4099	(arc4random() % (UINT_MAX - 1) + 1))
4100	r = TAILQ_NEXT(r, entries);
4101	#ifdef __FreeBSD__
4102	else if (r->match_tag && !pf_match_tag(m, r, &tag, pd->pf_mtag))
4103	#else
4104	else if (r->match_tag && !pf_match_tag(m, r, &tag))
4105	#endif
4106	r = TAILQ_NEXT(r, entries);
4107	else {
4108	if (r->anchor == NULL) {
4109	match = 1;
4110	*rm = r;
4111	*am = a;
4112	*rsm = ruleset;
4113	if ((*rm)->quick)
4114	break;
4115	r = TAILQ_NEXT(r, entries);
4116	} else
4117	pf_step_into_anchor(&asd, &ruleset,
4118	PF_RULESET_FILTER, &r, &a, &match);
4119	}
4120	if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
4121	PF_RULESET_FILTER, &r, &a, &match))
4122	break;
4123	}
4124	r = *rm;
4125	a = *am;
4126	ruleset = *rsm;
4127
4128	REASON_SET(&reason, PFRES_MATCH);
4129
4130	if (r->log)
4131	PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset,
4132	pd);
4133
4134	if (r->action != PF_PASS)
4135	return (PF_DROP);
4136
4137	#ifdef __FreeBSD__
4138	if (pf_tag_packet(m, tag, -1, pd->pf_mtag)) {
4139	#else
4140	if (pf_tag_packet(m, tag, -1)) {
4141	#endif
4142	REASON_SET(&reason, PFRES_MEMORY);
4143	return (PF_DROP);
4144	}
4145
4146	return (PF_PASS);
4147	}
4148
4149	int
4150	pf_tcp_track_full(struct pf_state_peer src, struct pf_state_peer dst,
4151	struct pf_state *state, struct pfi_kif kif, struct mbuf *m, int off,
4152	struct pf_pdesc pd, u_short reason, int *copyback)
4153	{
4154	struct tcphdr *th = pd->hdr.tcp;
4155	u_int16_t win = ntohs(th->th_win);
4156	u_int32_t ack, end, seq, orig_seq;
4157	u_int8_t sws, dws;
4158	int ackskew;
4159
4160	if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) {
4161	sws = src->wscale & PF_WSCALE_MASK;
4162	dws = dst->wscale & PF_WSCALE_MASK;
4163	} else
4164	sws = dws = 0;
4165
4166	/*
4167	* Sequence tracking algorithm from Guido van Rooij's paper:
4168	* http://www.madison-gurkha.com/publications/tcp_filtering/
4169	* tcp_filtering.ps
4170	*/
4171
4172	orig_seq = seq = ntohl(th->th_seq);
4173	if (src->seqlo == 0) {
4174	/* First packet from this end. Set its state */
4175
4176	if ((pd->flags & PFDESC_TCP_NORM \|\| dst->scrub) &&
4177	src->scrub == NULL) {
4178	if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) {
4179	REASON_SET(reason, PFRES_MEMORY);
4180	return (PF_DROP);
4181	}
4182	}
4183
4184	/* Deferred generation of sequence number modulator */
4185	if (dst->seqdiff && !src->seqdiff) {
4186	/* use random iss for the TCP server */
4187	while ((src->seqdiff = arc4random() - seq) == 0)
4188	;
4189	ack = ntohl(th->th_ack) - dst->seqdiff;
4190	pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
4191	src->seqdiff), 0);
4192	pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
4193	*copyback = 1;
4194	} else {
4195	ack = ntohl(th->th_ack);
4196	}
4197
4198	end = seq + pd->p_len;
4199	if (th->th_flags & TH_SYN) {
4200	end++;
4201	if (dst->wscale & PF_WSCALE_FLAG) {
4202	src->wscale = pf_get_wscale(m, off, th->th_off,
4203	pd->af);
4204	if (src->wscale & PF_WSCALE_FLAG) {
4205	/* Remove scale factor from initial
4206	* window */
4207	sws = src->wscale & PF_WSCALE_MASK;
4208	win = ((u_int32_t)win + (1 << sws) - 1)
4209	>> sws;
4210	dws = dst->wscale & PF_WSCALE_MASK;
4211	} else {
4212	/* fixup other window */
4213	dst->max_win <<= dst->wscale &
4214	PF_WSCALE_MASK;
4215	/* in case of a retrans SYN\|ACK */
4216	dst->wscale = 0;
4217	}
4218	}
4219	}
4220	if (th->th_flags & TH_FIN)
4221	end++;
4222
4223	src->seqlo = seq;
4224	if (src->state < TCPS_SYN_SENT)
4225	src->state = TCPS_SYN_SENT;
4226
4227	/*
4228	* May need to slide the window (seqhi may have been set by
4229	* the crappy stack check or if we picked up the connection
4230	* after establishment)
4231	*/
4232	if (src->seqhi == 1 \|\|
4233	SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi))
4234	src->seqhi = end + MAX(1, dst->max_win << dws);
4235	if (win > src->max_win)
4236	src->max_win = win;
4237
4238	} else {
4239	ack = ntohl(th->th_ack) - dst->seqdiff;
4240	if (src->seqdiff) {
4241	/* Modulate sequence numbers */
4242	pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
4243	src->seqdiff), 0);
4244	pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
4245	*copyback = 1;
4246	}
4247	end = seq + pd->p_len;
4248	if (th->th_flags & TH_SYN)
4249	end++;
4250	if (th->th_flags & TH_FIN)
4251	end++;
4252	}
4253
4254	if ((th->th_flags & TH_ACK) == 0) {
4255	/* Let it pass through the ack skew check */
4256	ack = dst->seqlo;
4257	} else if ((ack == 0 &&
4258	(th->th_flags & (TH_ACK\|TH_RST)) == (TH_ACK\|TH_RST)) \|\|
4259	/* broken tcp stacks do not set ack */
4260	(dst->state < TCPS_SYN_SENT)) {
4261	/*
4262	* Many stacks (ours included) will set the ACK number in an
4263	* FIN\|ACK if the SYN times out -- no sequence to ACK.
4264	*/
4265	ack = dst->seqlo;
4266	}
4267
4268	if (seq == end) {
4269	/* Ease sequencing restrictions on no data packets */
4270	seq = src->seqlo;
4271	end = seq;
4272	}
4273
4274	ackskew = dst->seqlo - ack;
4275
4276
4277	/*
4278	* Need to demodulate the sequence numbers in any TCP SACK options
4279	* (Selective ACK). We could optionally validate the SACK values
4280	* against the current ACK window, either forwards or backwards, but
4281	* I'm not confident that SACK has been implemented properly
4282	* everywhere. It wouldn't surprise me if several stacks accidently
4283	* SACK too far backwards of previously ACKed data. There really aren't
4284	* any security implications of bad SACKing unless the target stack
4285	* doesn't validate the option length correctly. Someone trying to
4286	* spoof into a TCP connection won't bother blindly sending SACK
4287	* options anyway.
4288	*/
4289	if (dst->seqdiff && (th->th_off << 2) > sizeof(struct tcphdr)) {
4290	if (pf_modulate_sack(m, off, pd, th, dst))
4291	*copyback = 1;
4292	}
4293
4294
4295	#define MAXACKWINDOW (0xffff + 1500) /* 1500 is an arbitrary fudge factor */
4296	if (SEQ_GEQ(src->seqhi, end) &&
4297	/* Last octet inside other's window space */
4298	SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) &&
4299	/* Retrans: not more than one window back */
4300	(ackskew >= -MAXACKWINDOW) &&
4301	/* Acking not more than one reassembled fragment backwards */
4302	(ackskew <= (MAXACKWINDOW << sws)) &&
4303	/* Acking not more than one window forward */
4304	((th->th_flags & TH_RST) == 0 \|\| orig_seq == src->seqlo \|\|
4305	(orig_seq == src->seqlo + 1) \|\| (orig_seq + 1 == src->seqlo) \|\|
4306	(pd->flags & PFDESC_IP_REAS) == 0)) {
4307	/* Require an exact/+1 sequence match on resets when possible */
4308
4309	if (dst->scrub \|\| src->scrub) {
4310	if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4311	*state, src, dst, copyback))
4312	return (PF_DROP);
4313	}
4314
4315	/* update max window */
4316	if (src->max_win < win)
4317	src->max_win = win;
4318	/* synchronize sequencing */
4319	if (SEQ_GT(end, src->seqlo))
4320	src->seqlo = end;
4321	/* slide the window of what the other end can send */
4322	if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4323	dst->seqhi = ack + MAX((win << sws), 1);
4324
4325
4326	/* update states */
4327	if (th->th_flags & TH_SYN)
4328	if (src->state < TCPS_SYN_SENT)
4329	src->state = TCPS_SYN_SENT;
4330	if (th->th_flags & TH_FIN)
4331	if (src->state < TCPS_CLOSING)
4332	src->state = TCPS_CLOSING;
4333	if (th->th_flags & TH_ACK) {
4334	if (dst->state == TCPS_SYN_SENT) {
4335	dst->state = TCPS_ESTABLISHED;
4336	if (src->state == TCPS_ESTABLISHED &&
4337	(*state)->src_node != NULL &&
4338	pf_src_connlimit(state)) {
4339	REASON_SET(reason, PFRES_SRCLIMIT);
4340	return (PF_DROP);
4341	}
4342	} else if (dst->state == TCPS_CLOSING)
4343	dst->state = TCPS_FIN_WAIT_2;
4344	}
4345	if (th->th_flags & TH_RST)
4346	src->state = dst->state = TCPS_TIME_WAIT;
4347
4348	/* update expire time */
4349	(*state)->expire = time_second;
4350	if (src->state >= TCPS_FIN_WAIT_2 &&
4351	dst->state >= TCPS_FIN_WAIT_2)
4352	(*state)->timeout = PFTM_TCP_CLOSED;
4353	else if (src->state >= TCPS_CLOSING &&
4354	dst->state >= TCPS_CLOSING)
4355	(*state)->timeout = PFTM_TCP_FIN_WAIT;
4356	else if (src->state < TCPS_ESTABLISHED \|\|
4357	dst->state < TCPS_ESTABLISHED)
4358	(*state)->timeout = PFTM_TCP_OPENING;
4359	else if (src->state >= TCPS_CLOSING \|\|
4360	dst->state >= TCPS_CLOSING)
4361	(*state)->timeout = PFTM_TCP_CLOSING;
4362	else
4363	(*state)->timeout = PFTM_TCP_ESTABLISHED;
4364
4365	/* Fall through to PASS packet */
4366
4367	} else if ((dst->state < TCPS_SYN_SENT \|\|
4368	dst->state >= TCPS_FIN_WAIT_2 \|\|
4369	src->state >= TCPS_FIN_WAIT_2) &&
4370	SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) &&
4371	/* Within a window forward of the originating packet */
4372	SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) {
4373	/* Within a window backward of the originating packet */
4374
4375	/*
4376	* This currently handles three situations:
4377	* 1) Stupid stacks will shotgun SYNs before their peer
4378	* replies.
4379	* 2) When PF catches an already established stream (the
4380	* firewall rebooted, the state table was flushed, routes
4381	* changed...)
4382	* 3) Packets get funky immediately after the connection
4383	* closes (this should catch Solaris spurious ACK\|FINs
4384	* that web servers like to spew after a close)
4385	*
4386	* This must be a little more careful than the above code
4387	* since packet floods will also be caught here. We don't
4388	* update the TTL here to mitigate the damage of a packet
4389	* flood and so the same code can handle awkward establishment
4390	* and a loosened connection close.
4391	* In the establishment case, a correct peer response will
4392	* validate the connection, go through the normal state code
4393	* and keep updating the state TTL.
4394	*/
4395
4396	#ifdef __FreeBSD__
4397	if (V_pf_status.debug >= PF_DEBUG_MISC) {
4398	#else
4399	if (pf_status.debug >= PF_DEBUG_MISC) {
4400	#endif
4401	printf("pf: loose state match: ");
4402	pf_print_state(*state);
4403	pf_print_flags(th->th_flags);
4404	printf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
4405	"pkts=%llu:%llu dir=%s,%s\n", seq, orig_seq, ack,
4406	#ifdef __FreeBSD__
4407	pd->p_len, ackskew, (unsigned long long)(*state)->packets[0],
4408	(unsigned long long)(*state)->packets[1],
4409	#else
4410	pd->p_len, ackskew, (*state)->packets[0],
4411	(*state)->packets[1],
4412	#endif
4413	pd->dir == PF_IN ? "in" : "out",
4414	pd->dir == (*state)->direction ? "fwd" : "rev");
4415	}
4416
4417	if (dst->scrub \|\| src->scrub) {
4418	if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4419	*state, src, dst, copyback))
4420	return (PF_DROP);
4421	}
4422
4423	/* update max window */
4424	if (src->max_win < win)
4425	src->max_win = win;
4426	/* synchronize sequencing */
4427	if (SEQ_GT(end, src->seqlo))
4428	src->seqlo = end;
4429	/* slide the window of what the other end can send */
4430	if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4431	dst->seqhi = ack + MAX((win << sws), 1);
4432
4433	/*
4434	* Cannot set dst->seqhi here since this could be a shotgunned
4435	* SYN and not an already established connection.
4436	*/
4437
4438	if (th->th_flags & TH_FIN)
4439	if (src->state < TCPS_CLOSING)
4440	src->state = TCPS_CLOSING;
4441	if (th->th_flags & TH_RST)
4442	src->state = dst->state = TCPS_TIME_WAIT;
4443
4444	/* Fall through to PASS packet */
4445
4446	} else {
4447	if ((*state)->dst.state == TCPS_SYN_SENT &&
4448	(*state)->src.state == TCPS_SYN_SENT) {
4449	/* Send RST for state mismatches during handshake */
4450	if (!(th->th_flags & TH_RST))
4451	#ifdef __FreeBSD__
4452	pf_send_tcp(NULL, (*state)->rule.ptr, pd->af,
4453	#else
4454	pf_send_tcp((*state)->rule.ptr, pd->af,
4455	#endif
4456	pd->dst, pd->src, th->th_dport,
4457	th->th_sport, ntohl(th->th_ack), 0,
4458	TH_RST, 0, 0,
4459	(*state)->rule.ptr->return_ttl, 1, 0,
4460	pd->eh, kif->pfik_ifp);
4461	src->seqlo = 0;
4462	src->seqhi = 1;
4463	src->max_win = 1;
4464	#ifdef __FreeBSD__
4465	} else if (V_pf_status.debug >= PF_DEBUG_MISC) {
4466	#else
4467	} else if (pf_status.debug >= PF_DEBUG_MISC) {
4468	#endif
4469	printf("pf: BAD state: ");
4470	pf_print_state(*state);
4471	pf_print_flags(th->th_flags);
4472	printf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
4473	"pkts=%llu:%llu dir=%s,%s\n",
4474	seq, orig_seq, ack, pd->p_len, ackskew,
4475	#ifdef __FreeBSD__
4476	(unsigned long long)(*state)->packets[0],
4477	(unsigned long long)(*state)->packets[1],
4478	#else
4479	(state)->packets[0], (state)->packets[1],
4480	#endif
4481	pd->dir == PF_IN ? "in" : "out",
4482	pd->dir == (*state)->direction ? "fwd" : "rev");
4483	printf("pf: State failure on: %c %c %c %c \| %c %c\n",
4484	SEQ_GEQ(src->seqhi, end) ? ' ' : '1',
4485	SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ?
4486	' ': '2',
4487	(ackskew >= -MAXACKWINDOW) ? ' ' : '3',
4488	(ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4',
4489	SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5',
4490	SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6');
4491	}
4492	REASON_SET(reason, PFRES_BADSTATE);
4493	return (PF_DROP);
4494	}
4495
4496	return (PF_PASS);
4497	}
4498
4499	int
4500	pf_tcp_track_sloppy(struct pf_state_peer src, struct pf_state_peer dst,
4501	struct pf_state *state, struct pf_pdesc pd, u_short *reason)
4502	{
4503	struct tcphdr *th = pd->hdr.tcp;
4504
4505	if (th->th_flags & TH_SYN)
4506	if (src->state < TCPS_SYN_SENT)
4507	src->state = TCPS_SYN_SENT;
4508	if (th->th_flags & TH_FIN)
4509	if (src->state < TCPS_CLOSING)
4510	src->state = TCPS_CLOSING;
4511	if (th->th_flags & TH_ACK) {
4512	if (dst->state == TCPS_SYN_SENT) {
4513	dst->state = TCPS_ESTABLISHED;
4514	if (src->state == TCPS_ESTABLISHED &&
4515	(*state)->src_node != NULL &&
4516	pf_src_connlimit(state)) {
4517	REASON_SET(reason, PFRES_SRCLIMIT);
4518	return (PF_DROP);
4519	}
4520	} else if (dst->state == TCPS_CLOSING) {
4521	dst->state = TCPS_FIN_WAIT_2;
4522	} else if (src->state == TCPS_SYN_SENT &&
4523	dst->state < TCPS_SYN_SENT) {
4524	/*
4525	* Handle a special sloppy case where we only see one
4526	* half of the connection. If there is a ACK after
4527	* the initial SYN without ever seeing a packet from
4528	* the destination, set the connection to established.
4529	*/
4530	dst->state = src->state = TCPS_ESTABLISHED;
4531	if ((*state)->src_node != NULL &&
4532	pf_src_connlimit(state)) {
4533	REASON_SET(reason, PFRES_SRCLIMIT);
4534	return (PF_DROP);
4535	}
4536	} else if (src->state == TCPS_CLOSING &&
4537	dst->state == TCPS_ESTABLISHED &&
4538	dst->seqlo == 0) {
4539	/*
4540	* Handle the closing of half connections where we
4541	* don't see the full bidirectional FIN/ACK+ACK
4542	* handshake.
4543	*/
4544	dst->state = TCPS_CLOSING;
4545	}
4546	}
4547	if (th->th_flags & TH_RST)
4548	src->state = dst->state = TCPS_TIME_WAIT;
4549
4550	/* update expire time */
4551	(*state)->expire = time_second;
4552	if (src->state >= TCPS_FIN_WAIT_2 &&
4553	dst->state >= TCPS_FIN_WAIT_2)
4554	(*state)->timeout = PFTM_TCP_CLOSED;
4555	else if (src->state >= TCPS_CLOSING &&
4556	dst->state >= TCPS_CLOSING)
4557	(*state)->timeout = PFTM_TCP_FIN_WAIT;
4558	else if (src->state < TCPS_ESTABLISHED \|\|
4559	dst->state < TCPS_ESTABLISHED)
4560	(*state)->timeout = PFTM_TCP_OPENING;
4561	else if (src->state >= TCPS_CLOSING \|\|
4562	dst->state >= TCPS_CLOSING)
4563	(*state)->timeout = PFTM_TCP_CLOSING;
4564	else
4565	(*state)->timeout = PFTM_TCP_ESTABLISHED;
4566
4567	return (PF_PASS);
4568	}
4569
4570	int
4571	pf_test_state_tcp(struct pf_state *state, int direction, struct pfi_kif kif,
4572	struct mbuf m, int off, void h, struct pf_pdesc *pd,
4573	u_short *reason)
4574	{
4575	struct pf_state_key_cmp key;
4576	struct tcphdr *th = pd->hdr.tcp;
4577	int copyback = 0;
4578	struct pf_state_peer src, dst;
4579	struct pf_state_key *sk;
4580
4581	key.af = pd->af;
4582	key.proto = IPPROTO_TCP;
4583	if (direction == PF_IN) { /* wire side, straight */
4584	PF_ACPY(&key.addr[0], pd->src, key.af);
4585	PF_ACPY(&key.addr[1], pd->dst, key.af);
4586	key.port[0] = th->th_sport;
4587	key.port[1] = th->th_dport;
4588	} else { /* stack side, reverse */
4589	PF_ACPY(&key.addr[1], pd->src, key.af);
4590	PF_ACPY(&key.addr[0], pd->dst, key.af);
4591	key.port[1] = th->th_sport;
4592	key.port[0] = th->th_dport;
4593	}
4594
4595	#ifdef __FreeBSD__
4596	STATE_LOOKUP(kif, &key, direction, *state, m, pd->pf_mtag);
4597	#else
4598	STATE_LOOKUP(kif, &key, direction, *state, m);
4599	#endif
4600
4601	if (direction == (*state)->direction) {
4602	src = &(*state)->src;
4603	dst = &(*state)->dst;
4604	} else {
4605	src = &(*state)->dst;
4606	dst = &(*state)->src;
4607	}
4608
4609	sk = (*state)->key[pd->didx];
4610
4611	if ((*state)->src.state == PF_TCPS_PROXY_SRC) {
4612	if (direction != (*state)->direction) {
4613	REASON_SET(reason, PFRES_SYNPROXY);
4614	return (PF_SYNPROXY_DROP);
4615	}
4616	if (th->th_flags & TH_SYN) {
4617	if (ntohl(th->th_seq) != (*state)->src.seqlo) {
4618	REASON_SET(reason, PFRES_SYNPROXY);
4619	return (PF_DROP);
4620	}
4621	#ifdef __FreeBSD__
4622	pf_send_tcp(NULL, (*state)->rule.ptr, pd->af, pd->dst,
4623	#else
4624	pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
4625	#endif
4626	pd->src, th->th_dport, th->th_sport,
4627	(*state)->src.seqhi, ntohl(th->th_seq) + 1,
4628	TH_SYN\|TH_ACK, 0, (*state)->src.mss, 0, 1,
4629	0, NULL, NULL);
4630	REASON_SET(reason, PFRES_SYNPROXY);
4631	return (PF_SYNPROXY_DROP);
4632	} else if (!(th->th_flags & TH_ACK) \|\|
4633	(ntohl(th->th_ack) != (*state)->src.seqhi + 1) \|\|
4634	(ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
4635	REASON_SET(reason, PFRES_SYNPROXY);
4636	return (PF_DROP);
4637	} else if ((*state)->src_node != NULL &&
4638	pf_src_connlimit(state)) {
4639	REASON_SET(reason, PFRES_SRCLIMIT);
4640	return (PF_DROP);
4641	} else
4642	(*state)->src.state = PF_TCPS_PROXY_DST;
4643	}
4644	if ((*state)->src.state == PF_TCPS_PROXY_DST) {
4645	if (direction == (*state)->direction) {
4646	if (((th->th_flags & (TH_SYN\|TH_ACK)) != TH_ACK) \|\|
4647	(ntohl(th->th_ack) != (*state)->src.seqhi + 1) \|\|
4648	(ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
4649	REASON_SET(reason, PFRES_SYNPROXY);
4650	return (PF_DROP);
4651	}
4652	(*state)->src.max_win = MAX(ntohs(th->th_win), 1);
4653	if ((*state)->dst.seqhi == 1)
4654	(*state)->dst.seqhi = htonl(arc4random());
4655	#ifdef __FreeBSD__
4656	pf_send_tcp(NULL, (*state)->rule.ptr, pd->af,
4657	#else
4658	pf_send_tcp((*state)->rule.ptr, pd->af,
4659	#endif
4660	&sk->addr[pd->sidx], &sk->addr[pd->didx],
4661	sk->port[pd->sidx], sk->port[pd->didx],
4662	(*state)->dst.seqhi, 0, TH_SYN, 0,
4663	(state)->src.mss, 0, 0, (state)->tag, NULL, NULL);
4664	REASON_SET(reason, PFRES_SYNPROXY);
4665	return (PF_SYNPROXY_DROP);
4666	} else if (((th->th_flags & (TH_SYN\|TH_ACK)) !=
4667	(TH_SYN\|TH_ACK)) \|\|
4668	(ntohl(th->th_ack) != (*state)->dst.seqhi + 1)) {
4669	REASON_SET(reason, PFRES_SYNPROXY);
4670	return (PF_DROP);
4671	} else {
4672	(*state)->dst.max_win = MAX(ntohs(th->th_win), 1);
4673	(*state)->dst.seqlo = ntohl(th->th_seq);
4674	#ifdef __FreeBSD__
4675	pf_send_tcp(NULL, (*state)->rule.ptr, pd->af, pd->dst,
4676	#else
4677	pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
4678	#endif
4679	pd->src, th->th_dport, th->th_sport,
4680	ntohl(th->th_ack), ntohl(th->th_seq) + 1,
4681	TH_ACK, (*state)->src.max_win, 0, 0, 0,
4682	(*state)->tag, NULL, NULL);
4683	#ifdef __FreeBSD__
4684	pf_send_tcp(NULL, (*state)->rule.ptr, pd->af,
4685	#else
4686	pf_send_tcp((*state)->rule.ptr, pd->af,
4687	#endif
4688	&sk->addr[pd->sidx], &sk->addr[pd->didx],
4689	sk->port[pd->sidx], sk->port[pd->didx],
4690	(state)->src.seqhi + 1, (state)->src.seqlo + 1,
4691	TH_ACK, (*state)->dst.max_win, 0, 0, 1,
4692	0, NULL, NULL);
4693	(state)->src.seqdiff = (state)->dst.seqhi -
4694	(*state)->src.seqlo;
4695	(state)->dst.seqdiff = (state)->src.seqhi -
4696	(*state)->dst.seqlo;
4697	(state)->src.seqhi = (state)->src.seqlo +
4698	(*state)->dst.max_win;
4699	(state)->dst.seqhi = (state)->dst.seqlo +
4700	(*state)->src.max_win;
4701	(state)->src.wscale = (state)->dst.wscale = 0;
4702	(state)->src.state = (state)->dst.state =
4703	TCPS_ESTABLISHED;
4704	REASON_SET(reason, PFRES_SYNPROXY);
4705	return (PF_SYNPROXY_DROP);
4706	}
4707	}
4708
4709	if (((th->th_flags & (TH_SYN\|TH_ACK)) == TH_SYN) &&
4710	dst->state >= TCPS_FIN_WAIT_2 &&
4711	src->state >= TCPS_FIN_WAIT_2) {
4712	#ifdef __FreeBSD__
4713	if (V_pf_status.debug >= PF_DEBUG_MISC) {
4714	#else
4715	if (pf_status.debug >= PF_DEBUG_MISC) {
4716	#endif
4717	printf("pf: state reuse ");
4718	pf_print_state(*state);
4719	pf_print_flags(th->th_flags);
4720	printf("\n");
4721	}
4722	/* XXX make sure it's the same direction ?? */
4723	(state)->src.state = (state)->dst.state = TCPS_CLOSED;
4724	pf_unlink_state(*state);
4725	*state = NULL;
4726	return (PF_DROP);
4727	}
4728
4729	if ((*state)->state_flags & PFSTATE_SLOPPY) {
4730	if (pf_tcp_track_sloppy(src, dst, state, pd, reason) == PF_DROP)
4731	return (PF_DROP);
4732	} else {
4733	if (pf_tcp_track_full(src, dst, state, kif, m, off, pd, reason,
4734	&copyback) == PF_DROP)
4735	return (PF_DROP);
4736	}
4737
4738	/* translate source/destination address, if necessary */
4739	if ((state)->key[PF_SK_WIRE] != (state)->key[PF_SK_STACK]) {
4740	struct pf_state_key nk = (state)->key[pd->didx];
4741
4742	if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], pd->af) \|\|
4743	nk->port[pd->sidx] != th->th_sport)
4744	pf_change_ap(pd->src, &th->th_sport, pd->ip_sum,
4745	&th->th_sum, &nk->addr[pd->sidx],
4746	nk->port[pd->sidx], 0, pd->af);
4747
4748	if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], pd->af) \|\|
4749	nk->port[pd->didx] != th->th_dport)
4750	pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum,
4751	&th->th_sum, &nk->addr[pd->didx],
4752	nk->port[pd->didx], 0, pd->af);
4753	copyback = 1;
4754	}
4755
4756	/* Copyback sequence modulation or stateful scrub changes if needed */
4757	if (copyback)
4758	#ifdef __FreeBSD__
4759	m_copyback(m, off, sizeof(*th), (caddr_t)th);
4760	#else
4761	m_copyback(m, off, sizeof(*th), th);
4762	#endif
4763
4764	return (PF_PASS);
4765	}
4766
4767	int
4768	pf_test_state_udp(struct pf_state *state, int direction, struct pfi_kif kif,
4769	struct mbuf m, int off, void h, struct pf_pdesc *pd)
4770	{
4771	struct pf_state_peer src, dst;
4772	struct pf_state_key_cmp key;
4773	struct udphdr *uh = pd->hdr.udp;
4774
4775	key.af = pd->af;
4776	key.proto = IPPROTO_UDP;
4777	if (direction == PF_IN) { /* wire side, straight */
4778	PF_ACPY(&key.addr[0], pd->src, key.af);
4779	PF_ACPY(&key.addr[1], pd->dst, key.af);
4780	key.port[0] = uh->uh_sport;
4781	key.port[1] = uh->uh_dport;
4782	} else { /* stack side, reverse */
4783	PF_ACPY(&key.addr[1], pd->src, key.af);
4784	PF_ACPY(&key.addr[0], pd->dst, key.af);
4785	key.port[1] = uh->uh_sport;
4786	key.port[0] = uh->uh_dport;
4787	}
4788
4789	#ifdef __FreeBSD__
4790	STATE_LOOKUP(kif, &key, direction, *state, m, pd->pf_mtag);
4791	#else
4792	STATE_LOOKUP(kif, &key, direction, *state, m);
4793	#endif
4794
4795	if (direction == (*state)->direction) {
4796	src = &(*state)->src;
4797	dst = &(*state)->dst;
4798	} else {
4799	src = &(*state)->dst;
4800	dst = &(*state)->src;
4801	}
4802
4803	/* update states */
4804	if (src->state < PFUDPS_SINGLE)
4805	src->state = PFUDPS_SINGLE;
4806	if (dst->state == PFUDPS_SINGLE)
4807	dst->state = PFUDPS_MULTIPLE;
4808
4809	/* update expire time */
4810	(*state)->expire = time_second;
4811	if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE)
4812	(*state)->timeout = PFTM_UDP_MULTIPLE;
4813	else
4814	(*state)->timeout = PFTM_UDP_SINGLE;
4815
4816	/* translate source/destination address, if necessary */
4817	if ((state)->key[PF_SK_WIRE] != (state)->key[PF_SK_STACK]) {
4818	struct pf_state_key nk = (state)->key[pd->didx];
4819
4820	if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], pd->af) \|\|
4821	nk->port[pd->sidx] != uh->uh_sport)
4822	pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum,
4823	&uh->uh_sum, &nk->addr[pd->sidx],
4824	nk->port[pd->sidx], 1, pd->af);
4825
4826	if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], pd->af) \|\|
4827	nk->port[pd->didx] != uh->uh_dport)
4828	pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum,
4829	&uh->uh_sum, &nk->addr[pd->didx],
4830	nk->port[pd->didx], 1, pd->af);
4831	#ifdef __FreeBSD__
4832	m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
4833	#else
4834	m_copyback(m, off, sizeof(*uh), uh);
4835	#endif
4836	}
4837
4838	return (PF_PASS);
4839	}
4840
4841	int
4842	pf_test_state_icmp(struct pf_state *state, int direction, struct pfi_kif kif,
4843	struct mbuf m, int off, void h, struct pf_pdesc pd, u_short reason)
4844	{
4845	struct pf_addr saddr = pd->src, daddr = pd->dst;
4846	#ifdef __FreeBSD__
4847	u_int16_t icmpid = 0, *icmpsum;
4848	#else
4849	u_int16_t icmpid, *icmpsum;
4850	#endif
4851	u_int8_t icmptype;
4852	int state_icmp = 0;
4853	struct pf_state_key_cmp key;
4854
4855	switch (pd->proto) {
4856	#ifdef INET
4857	case IPPROTO_ICMP:
4858	icmptype = pd->hdr.icmp->icmp_type;
4859	icmpid = pd->hdr.icmp->icmp_id;
4860	icmpsum = &pd->hdr.icmp->icmp_cksum;
4861
4862	if (icmptype == ICMP_UNREACH \|\|
4863	icmptype == ICMP_SOURCEQUENCH \|\|
4864	icmptype == ICMP_REDIRECT \|\|
4865	icmptype == ICMP_TIMXCEED \|\|
4866	icmptype == ICMP_PARAMPROB)
4867	state_icmp++;
4868	break;
4869	#endif /* INET */
4870	#ifdef INET6
4871	case IPPROTO_ICMPV6:
4872	icmptype = pd->hdr.icmp6->icmp6_type;
4873	icmpid = pd->hdr.icmp6->icmp6_id;
4874	icmpsum = &pd->hdr.icmp6->icmp6_cksum;
4875
4876	if (icmptype == ICMP6_DST_UNREACH \|\|
4877	icmptype == ICMP6_PACKET_TOO_BIG \|\|
4878	icmptype == ICMP6_TIME_EXCEEDED \|\|
4879	icmptype == ICMP6_PARAM_PROB)
4880	state_icmp++;
4881	break;
4882	#endif /* INET6 */
4883	}
4884
4885	if (!state_icmp) {
4886
4887	/*
4888	* ICMP query/reply message not related to a TCP/UDP packet.
4889	* Search for an ICMP state.
4890	*/
4891	key.af = pd->af;
4892	key.proto = pd->proto;
4893	key.port[0] = key.port[1] = icmpid;
4894	if (direction == PF_IN) { /* wire side, straight */
4895	PF_ACPY(&key.addr[0], pd->src, key.af);
4896	PF_ACPY(&key.addr[1], pd->dst, key.af);
4897	} else { /* stack side, reverse */
4898	PF_ACPY(&key.addr[1], pd->src, key.af);
4899	PF_ACPY(&key.addr[0], pd->dst, key.af);
4900	}
4901
4902	#ifdef __FreeBSD__
4903	STATE_LOOKUP(kif, &key, direction, *state, m, pd->pf_mtag);
4904	#else
4905	STATE_LOOKUP(kif, &key, direction, *state, m);
4906	#endif
4907
4908	(*state)->expire = time_second;
4909	(*state)->timeout = PFTM_ICMP_ERROR_REPLY;
4910
4911	/* translate source/destination address, if necessary */
4912	if ((state)->key[PF_SK_WIRE] != (state)->key[PF_SK_STACK]) {
4913	struct pf_state_key nk = (state)->key[pd->didx];
4914
4915	switch (pd->af) {
4916	#ifdef INET
4917	case AF_INET:
4918	if (PF_ANEQ(pd->src,
4919	&nk->addr[pd->sidx], AF_INET))
4920	pf_change_a(&saddr->v4.s_addr,
4921	pd->ip_sum,
4922	nk->addr[pd->sidx].v4.s_addr, 0);
4923
4924	if (PF_ANEQ(pd->dst, &nk->addr[pd->didx],
4925	AF_INET))
4926	pf_change_a(&daddr->v4.s_addr,
4927	pd->ip_sum,
4928	nk->addr[pd->didx].v4.s_addr, 0);
4929
4930	if (nk->port[0] !=
4931	pd->hdr.icmp->icmp_id) {
4932	pd->hdr.icmp->icmp_cksum =
4933	pf_cksum_fixup(
4934	pd->hdr.icmp->icmp_cksum, icmpid,
4935	nk->port[pd->sidx], 0);
4936	pd->hdr.icmp->icmp_id =
4937	nk->port[pd->sidx];
4938	}
4939
4940	m_copyback(m, off, ICMP_MINLEN,
4941	#ifdef __FreeBSD__
4942	(caddr_t)
4943	#endif
4944	pd->hdr.icmp);
4945	break;
4946	#endif /* INET */
4947	#ifdef INET6
4948	case AF_INET6:
4949	if (PF_ANEQ(pd->src,
4950	&nk->addr[pd->sidx], AF_INET6))
4951	pf_change_a6(saddr,
4952	&pd->hdr.icmp6->icmp6_cksum,
4953	&nk->addr[pd->sidx], 0);
4954
4955	if (PF_ANEQ(pd->dst,
4956	&nk->addr[pd->didx], AF_INET6))
4957	pf_change_a6(daddr,
4958	&pd->hdr.icmp6->icmp6_cksum,
4959	&nk->addr[pd->didx], 0);
4960
4961	m_copyback(m, off,
4962	sizeof(struct icmp6_hdr),
4963	#ifdef __FreeBSD__
4964	(caddr_t)
4965	#endif
4966	pd->hdr.icmp6);
4967	break;
4968	#endif /* INET6 */
4969	}
4970	}
4971	return (PF_PASS);
4972
4973	} else {
4974	/*
4975	* ICMP error message in response to a TCP/UDP packet.
4976	* Extract the inner TCP/UDP header and search for that state.
4977	*/
4978
4979	struct pf_pdesc pd2;
4980	#ifdef __FreeBSD__
4981	bzero(&pd2, sizeof pd2);
4982	#endif
4983	#ifdef INET
4984	struct ip h2;
4985	#endif /* INET */
4986	#ifdef INET6
4987	struct ip6_hdr h2_6;
4988	int terminal = 0;
4989	#endif /* INET6 */
4990	#ifdef __FreeBSD__
4991	int ipoff2 = 0;
4992	int off2 = 0;
4993	#else
4994	int ipoff2;
4995	int off2;
4996	#endif
4997
4998	pd2.af = pd->af;
4999	/* Payload packet is from the opposite direction. */
5000	pd2.sidx = (direction == PF_IN) ? 1 : 0;
5001	pd2.didx = (direction == PF_IN) ? 0 : 1;
5002	switch (pd->af) {
5003	#ifdef INET
5004	case AF_INET:
5005	/* offset of h2 in mbuf chain */
5006	ipoff2 = off + ICMP_MINLEN;
5007
5008	if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2),
5009	NULL, reason, pd2.af)) {
5010	DPFPRINTF(PF_DEBUG_MISC,
5011	("pf: ICMP error message too short "
5012	"(ip)\n"));
5013	return (PF_DROP);
5014	}
5015	/*
5016	* ICMP error messages don't refer to non-first
5017	* fragments
5018	*/
5019	if (h2.ip_off & htons(IP_OFFMASK)) {
5020	REASON_SET(reason, PFRES_FRAG);
5021	return (PF_DROP);
5022	}
5023
5024	/* offset of protocol header that follows h2 */
5025	off2 = ipoff2 + (h2.ip_hl << 2);
5026
5027	pd2.proto = h2.ip_p;
5028	pd2.src = (struct pf_addr *)&h2.ip_src;
5029	pd2.dst = (struct pf_addr *)&h2.ip_dst;
5030	pd2.ip_sum = &h2.ip_sum;
5031	break;
5032	#endif /* INET */
5033	#ifdef INET6
5034	case AF_INET6:
5035	ipoff2 = off + sizeof(struct icmp6_hdr);
5036
5037	if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6),
5038	NULL, reason, pd2.af)) {
5039	DPFPRINTF(PF_DEBUG_MISC,
5040	("pf: ICMP error message too short "
5041	"(ip6)\n"));
5042	return (PF_DROP);
5043	}
5044	pd2.proto = h2_6.ip6_nxt;
5045	pd2.src = (struct pf_addr *)&h2_6.ip6_src;
5046	pd2.dst = (struct pf_addr *)&h2_6.ip6_dst;
5047	pd2.ip_sum = NULL;
5048	off2 = ipoff2 + sizeof(h2_6);
5049	do {
5050	switch (pd2.proto) {
5051	case IPPROTO_FRAGMENT:
5052	/*
5053	* ICMPv6 error messages for
5054	* non-first fragments
5055	*/
5056	REASON_SET(reason, PFRES_FRAG);
5057	return (PF_DROP);
5058	case IPPROTO_AH:
5059	case IPPROTO_HOPOPTS:
5060	case IPPROTO_ROUTING:
5061	case IPPROTO_DSTOPTS: {
5062	/* get next header and header length */
5063	struct ip6_ext opt6;
5064
5065	if (!pf_pull_hdr(m, off2, &opt6,
5066	sizeof(opt6), NULL, reason,
5067	pd2.af)) {
5068	DPFPRINTF(PF_DEBUG_MISC,
5069	("pf: ICMPv6 short opt\n"));
5070	return (PF_DROP);
5071	}
5072	if (pd2.proto == IPPROTO_AH)
5073	off2 += (opt6.ip6e_len + 2) * 4;
5074	else
5075	off2 += (opt6.ip6e_len + 1) * 8;
5076	pd2.proto = opt6.ip6e_nxt;
5077	/* goto the next header */
5078	break;
5079	}
5080	default:
5081	terminal++;
5082	break;
5083	}
5084	} while (!terminal);
5085	break;
5086	#endif /* INET6 */
5087	}
5088
5089	switch (pd2.proto) {
5090	case IPPROTO_TCP: {
5091	struct tcphdr th;
5092	u_int32_t seq;
5093	struct pf_state_peer src, dst;
5094	u_int8_t dws;
5095	int copyback = 0;
5096
5097	/*
5098	* Only the first 8 bytes of the TCP header can be
5099	* expected. Don't access any TCP header fields after
5100	* th_seq, an ackskew test is not possible.
5101	*/
5102	if (!pf_pull_hdr(m, off2, &th, 8, NULL, reason,
5103	pd2.af)) {
5104	DPFPRINTF(PF_DEBUG_MISC,
5105	("pf: ICMP error message too short "
5106	"(tcp)\n"));
5107	return (PF_DROP);
5108	}
5109
5110	key.af = pd2.af;
5111	key.proto = IPPROTO_TCP;
5112	PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
5113	PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
5114	key.port[pd2.sidx] = th.th_sport;
5115	key.port[pd2.didx] = th.th_dport;
5116
5117	#ifdef __FreeBSD__
5118	STATE_LOOKUP(kif, &key, direction, *state, m, pd->pf_mtag);
5119	#else
5120	STATE_LOOKUP(kif, &key, direction, *state, m);
5121	#endif
5122
5123	if (direction == (*state)->direction) {
5124	src = &(*state)->dst;
5125	dst = &(*state)->src;
5126	} else {
5127	src = &(*state)->src;
5128	dst = &(*state)->dst;
5129	}
5130
5131	if (src->wscale && dst->wscale)
5132	dws = dst->wscale & PF_WSCALE_MASK;
5133	else
5134	dws = 0;
5135
5136	/* Demodulate sequence number */
5137	seq = ntohl(th.th_seq) - src->seqdiff;
5138	if (src->seqdiff) {
5139	pf_change_a(&th.th_seq, icmpsum,
5140	htonl(seq), 0);
5141	copyback = 1;
5142	}
5143
5144	if (!((*state)->state_flags & PFSTATE_SLOPPY) &&
5145	(!SEQ_GEQ(src->seqhi, seq) \|\|
5146	!SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)))) {
5147	#ifdef __FreeBSD__
5148	if (V_pf_status.debug >= PF_DEBUG_MISC) {
5149	#else
5150	if (pf_status.debug >= PF_DEBUG_MISC) {
5151	#endif
5152	printf("pf: BAD ICMP %d:%d ",
5153	icmptype, pd->hdr.icmp->icmp_code);
5154	pf_print_host(pd->src, 0, pd->af);
5155	printf(" -> ");
5156	pf_print_host(pd->dst, 0, pd->af);
5157	printf(" state: ");
5158	pf_print_state(*state);
5159	printf(" seq=%u\n", seq);
5160	}
5161	REASON_SET(reason, PFRES_BADSTATE);
5162	return (PF_DROP);
5163	} else {
5164	#ifdef __FreeBSD__
5165	if (V_pf_status.debug >= PF_DEBUG_MISC) {
5166	#else
5167	if (pf_status.debug >= PF_DEBUG_MISC) {
5168	#endif
5169	printf("pf: OK ICMP %d:%d ",
5170	icmptype, pd->hdr.icmp->icmp_code);
5171	pf_print_host(pd->src, 0, pd->af);
5172	printf(" -> ");
5173	pf_print_host(pd->dst, 0, pd->af);
5174	printf(" state: ");
5175	pf_print_state(*state);
5176	printf(" seq=%u\n", seq);
5177	}
5178	}
5179
5180	/* translate source/destination address, if necessary */
5181	if ((*state)->key[PF_SK_WIRE] !=
5182	(*state)->key[PF_SK_STACK]) {
5183	struct pf_state_key *nk =
5184	(*state)->key[pd->didx];
5185
5186	if (PF_ANEQ(pd2.src,
5187	&nk->addr[pd2.sidx], pd2.af) \|\|
5188	nk->port[pd2.sidx] != th.th_sport)
5189	pf_change_icmp(pd2.src, &th.th_sport,
5190	daddr, &nk->addr[pd2.sidx],
5191	nk->port[pd2.sidx], NULL,
5192	pd2.ip_sum, icmpsum,
5193	pd->ip_sum, 0, pd2.af);
5194
5195	if (PF_ANEQ(pd2.dst,
5196	&nk->addr[pd2.didx], pd2.af) \|\|
5197	nk->port[pd2.didx] != th.th_dport)
5198	pf_change_icmp(pd2.dst, &th.th_dport,
5199	NULL, /* XXX Inbound NAT? */
5200	&nk->addr[pd2.didx],
5201	nk->port[pd2.didx], NULL,
5202	pd2.ip_sum, icmpsum,
5203	pd->ip_sum, 0, pd2.af);
5204	copyback = 1;
5205	}
5206
5207	if (copyback) {
5208	switch (pd2.af) {
5209	#ifdef INET
5210	case AF_INET:
5211	m_copyback(m, off, ICMP_MINLEN,
5212	#ifdef __FreeBSD__
5213	(caddr_t)
5214	#endif
5215	pd->hdr.icmp);
5216	m_copyback(m, ipoff2, sizeof(h2),
5217	#ifdef __FreeBSD__
5218	(caddr_t)
5219	#endif
5220	&h2);
5221	break;
5222	#endif /* INET */
5223	#ifdef INET6
5224	case AF_INET6:
5225	m_copyback(m, off,
5226	sizeof(struct icmp6_hdr),
5227	#ifdef __FreeBSD__
5228	(caddr_t)
5229	#endif
5230	pd->hdr.icmp6);
5231	m_copyback(m, ipoff2, sizeof(h2_6),
5232	#ifdef __FreeBSD__
5233	(caddr_t)
5234	#endif
5235	&h2_6);
5236	break;
5237	#endif /* INET6 */
5238	}
5239	#ifdef __FreeBSD__
5240	m_copyback(m, off2, 8, (caddr_t)&th);
5241	#else
5242	m_copyback(m, off2, 8, &th);
5243	#endif
5244	}
5245
5246	return (PF_PASS);
5247	break;
5248	}
5249	case IPPROTO_UDP: {
5250	struct udphdr uh;
5251
5252	if (!pf_pull_hdr(m, off2, &uh, sizeof(uh),
5253	NULL, reason, pd2.af)) {
5254	DPFPRINTF(PF_DEBUG_MISC,
5255	("pf: ICMP error message too short "
5256	"(udp)\n"));
5257	return (PF_DROP);
5258	}
5259
5260	key.af = pd2.af;
5261	key.proto = IPPROTO_UDP;
5262	PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
5263	PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
5264	key.port[pd2.sidx] = uh.uh_sport;
5265	key.port[pd2.didx] = uh.uh_dport;
5266
5267	#ifdef __FreeBSD__
5268	STATE_LOOKUP(kif, &key, direction, *state, m, pd->pf_mtag);
5269	#else
5270	STATE_LOOKUP(kif, &key, direction, *state, m);
5271	#endif
5272
5273	/* translate source/destination address, if necessary */
5274	if ((*state)->key[PF_SK_WIRE] !=
5275	(*state)->key[PF_SK_STACK]) {
5276	struct pf_state_key *nk =
5277	(*state)->key[pd->didx];
5278
5279	if (PF_ANEQ(pd2.src,
5280	&nk->addr[pd2.sidx], pd2.af) \|\|
5281	nk->port[pd2.sidx] != uh.uh_sport)
5282	pf_change_icmp(pd2.src, &uh.uh_sport,
5283	daddr, &nk->addr[pd2.sidx],
5284	nk->port[pd2.sidx], &uh.uh_sum,
5285	pd2.ip_sum, icmpsum,
5286	pd->ip_sum, 1, pd2.af);
5287
5288	if (PF_ANEQ(pd2.dst,
5289	&nk->addr[pd2.didx], pd2.af) \|\|
5290	nk->port[pd2.didx] != uh.uh_dport)
5291	pf_change_icmp(pd2.dst, &uh.uh_dport,
5292	NULL, /* XXX Inbound NAT? */
5293	&nk->addr[pd2.didx],
5294	nk->port[pd2.didx], &uh.uh_sum,
5295	pd2.ip_sum, icmpsum,
5296	pd->ip_sum, 1, pd2.af);
5297
5298	switch (pd2.af) {
5299	#ifdef INET
5300	case AF_INET:
5301	m_copyback(m, off, ICMP_MINLEN,
5302	#ifdef __FreeBSD__
5303	(caddr_t)
5304	#endif
5305	pd->hdr.icmp);
5306	#ifdef __FreeBSD__
5307	m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
5308	#else
5309	m_copyback(m, ipoff2, sizeof(h2), &h2);
5310	#endif
5311	break;
5312	#endif /* INET */
5313	#ifdef INET6
5314	case AF_INET6:
5315	m_copyback(m, off,
5316	sizeof(struct icmp6_hdr),
5317	#ifdef __FreeBSD__
5318	(caddr_t)
5319	#endif
5320	pd->hdr.icmp6);
5321	m_copyback(m, ipoff2, sizeof(h2_6),
5322	#ifdef __FreeBSD__
5323	(caddr_t)
5324	#endif
5325	&h2_6);
5326	break;
5327	#endif /* INET6 */
5328	}
5329	#ifdef __FreeBSD__
5330	m_copyback(m, off2, sizeof(uh), (caddr_t)&uh);
5331	#else
5332	m_copyback(m, off2, sizeof(uh), &uh);
5333	#endif
5334	}
5335	return (PF_PASS);
5336	break;
5337	}
5338	#ifdef INET
5339	case IPPROTO_ICMP: {
5340	struct icmp iih;
5341
5342	if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN,
5343	NULL, reason, pd2.af)) {
5344	DPFPRINTF(PF_DEBUG_MISC,
5345	("pf: ICMP error message too short i"
5346	"(icmp)\n"));
5347	return (PF_DROP);
5348	}
5349
5350	key.af = pd2.af;
5351	key.proto = IPPROTO_ICMP;
5352	PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
5353	PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
5354	key.port[0] = key.port[1] = iih.icmp_id;
5355
5356	#ifdef __FreeBSD__
5357	STATE_LOOKUP(kif, &key, direction, *state, m, pd->pf_mtag);
5358	#else
5359	STATE_LOOKUP(kif, &key, direction, *state, m);
5360	#endif
5361
5362	/* translate source/destination address, if necessary */
5363	if ((*state)->key[PF_SK_WIRE] !=
5364	(*state)->key[PF_SK_STACK]) {
5365	struct pf_state_key *nk =
5366	(*state)->key[pd->didx];
5367
5368	if (PF_ANEQ(pd2.src,
5369	&nk->addr[pd2.sidx], pd2.af) \|\|
5370	nk->port[pd2.sidx] != iih.icmp_id)
5371	pf_change_icmp(pd2.src, &iih.icmp_id,
5372	daddr, &nk->addr[pd2.sidx],
5373	nk->port[pd2.sidx], NULL,
5374	pd2.ip_sum, icmpsum,
5375	pd->ip_sum, 0, AF_INET);
5376
5377	if (PF_ANEQ(pd2.dst,
5378	&nk->addr[pd2.didx], pd2.af) \|\|
5379	nk->port[pd2.didx] != iih.icmp_id)
5380	pf_change_icmp(pd2.dst, &iih.icmp_id,
5381	NULL, /* XXX Inbound NAT? */
5382	&nk->addr[pd2.didx],
5383	nk->port[pd2.didx], NULL,
5384	pd2.ip_sum, icmpsum,
5385	pd->ip_sum, 0, AF_INET);
5386
5387	#ifdef __FreeBSD__
5388	m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
5389	m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
5390	m_copyback(m, off2, ICMP_MINLEN, (caddr_t)&iih);
5391	#else
5392	m_copyback(m, off, ICMP_MINLEN, pd->hdr.icmp);
5393	m_copyback(m, ipoff2, sizeof(h2), &h2);
5394	m_copyback(m, off2, ICMP_MINLEN, &iih);
5395	#endif
5396	}
5397	return (PF_PASS);
5398	break;
5399	}
5400	#endif /* INET */
5401	#ifdef INET6
5402	case IPPROTO_ICMPV6: {
5403	struct icmp6_hdr iih;
5404
5405	if (!pf_pull_hdr(m, off2, &iih,
5406	sizeof(struct icmp6_hdr), NULL, reason, pd2.af)) {
5407	DPFPRINTF(PF_DEBUG_MISC,
5408	("pf: ICMP error message too short "
5409	"(icmp6)\n"));
5410	return (PF_DROP);
5411	}
5412
5413	key.af = pd2.af;
5414	key.proto = IPPROTO_ICMPV6;
5415	PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
5416	PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
5417	key.port[0] = key.port[1] = iih.icmp6_id;
5418
5419	#ifdef __FreeBSD__
5420	STATE_LOOKUP(kif, &key, direction, *state, m, pd->pf_mtag);
5421	#else
5422	STATE_LOOKUP(kif, &key, direction, *state, m);
5423	#endif
5424
5425	/* translate source/destination address, if necessary */
5426	if ((*state)->key[PF_SK_WIRE] !=
5427	(*state)->key[PF_SK_STACK]) {
5428	struct pf_state_key *nk =
5429	(*state)->key[pd->didx];
5430
5431	if (PF_ANEQ(pd2.src,
5432	&nk->addr[pd2.sidx], pd2.af) \|\|
5433	nk->port[pd2.sidx] != iih.icmp6_id)
5434	pf_change_icmp(pd2.src, &iih.icmp6_id,
5435	daddr, &nk->addr[pd2.sidx],
5436	nk->port[pd2.sidx], NULL,
5437	pd2.ip_sum, icmpsum,
5438	pd->ip_sum, 0, AF_INET6);
5439
5440	if (PF_ANEQ(pd2.dst,
5441	&nk->addr[pd2.didx], pd2.af) \|\|
5442	nk->port[pd2.didx] != iih.icmp6_id)
5443	pf_change_icmp(pd2.dst, &iih.icmp6_id,
5444	NULL, /* XXX Inbound NAT? */
5445	&nk->addr[pd2.didx],
5446	nk->port[pd2.didx], NULL,
5447	pd2.ip_sum, icmpsum,
5448	pd->ip_sum, 0, AF_INET6);
5449
5450	#ifdef __FreeBSD__
5451	m_copyback(m, off, sizeof(struct icmp6_hdr),
5452	(caddr_t)pd->hdr.icmp6);
5453	m_copyback(m, ipoff2, sizeof(h2_6), (caddr_t)&h2_6);
5454	m_copyback(m, off2, sizeof(struct icmp6_hdr),
5455	(caddr_t)&iih);
5456	#else
5457	m_copyback(m, off, sizeof(struct icmp6_hdr),
5458	pd->hdr.icmp6);
5459	m_copyback(m, ipoff2, sizeof(h2_6), &h2_6);
5460	m_copyback(m, off2, sizeof(struct icmp6_hdr),
5461	&iih);
5462	#endif
5463	}
5464	return (PF_PASS);
5465	break;
5466	}
5467	#endif /* INET6 */
5468	default: {
5469	key.af = pd2.af;
5470	key.proto = pd2.proto;
5471	PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af);
5472	PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af);
5473	key.port[0] = key.port[1] = 0;
5474
5475	#ifdef __FreeBSD__
5476	STATE_LOOKUP(kif, &key, direction, *state, m, pd->pf_mtag);
5477	#else
5478	STATE_LOOKUP(kif, &key, direction, *state, m);
5479	#endif
5480
5481	/* translate source/destination address, if necessary */
5482	if ((*state)->key[PF_SK_WIRE] !=
5483	(*state)->key[PF_SK_STACK]) {
5484	struct pf_state_key *nk =
5485	(*state)->key[pd->didx];
5486
5487	if (PF_ANEQ(pd2.src,
5488	&nk->addr[pd2.sidx], pd2.af))
5489	pf_change_icmp(pd2.src, NULL, daddr,
5490	&nk->addr[pd2.sidx], 0, NULL,
5491	pd2.ip_sum, icmpsum,
5492	pd->ip_sum, 0, pd2.af);
5493
5494	if (PF_ANEQ(pd2.dst,
5495	&nk->addr[pd2.didx], pd2.af))
5496	pf_change_icmp(pd2.src, NULL,
5497	NULL, /* XXX Inbound NAT? */
5498	&nk->addr[pd2.didx], 0, NULL,
5499	pd2.ip_sum, icmpsum,
5500	pd->ip_sum, 0, pd2.af);
5501
5502	switch (pd2.af) {
5503	#ifdef INET
5504	case AF_INET:
5505	#ifdef __FreeBSD__
5506	m_copyback(m, off, ICMP_MINLEN,
5507	(caddr_t)pd->hdr.icmp);
5508	m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
5509	#else
5510	m_copyback(m, off, ICMP_MINLEN,
5511	pd->hdr.icmp);
5512	m_copyback(m, ipoff2, sizeof(h2), &h2);
5513	#endif
5514	break;
5515	#endif /* INET */
5516	#ifdef INET6
5517	case AF_INET6:
5518	m_copyback(m, off,
5519	sizeof(struct icmp6_hdr),
5520	#ifdef __FreeBSD__
5521	(caddr_t)
5522	#endif
5523	pd->hdr.icmp6);
5524	m_copyback(m, ipoff2, sizeof(h2_6),
5525	#ifdef __FreeBSD__
5526	(caddr_t)
5527	#endif
5528	&h2_6);
5529	break;
5530	#endif /* INET6 */
5531	}
5532	}
5533	return (PF_PASS);
5534	break;
5535	}
5536	}
5537	}
5538	}
5539
5540	int
5541	pf_test_state_other(struct pf_state *state, int direction, struct pfi_kif kif,
5542	struct mbuf m, struct pf_pdesc pd)
5543	{
5544	struct pf_state_peer src, dst;
5545	struct pf_state_key_cmp key;
5546
5547	key.af = pd->af;
5548	key.proto = pd->proto;
5549	if (direction == PF_IN) {
5550	PF_ACPY(&key.addr[0], pd->src, key.af);
5551	PF_ACPY(&key.addr[1], pd->dst, key.af);
5552	key.port[0] = key.port[1] = 0;
5553	} else {
5554	PF_ACPY(&key.addr[1], pd->src, key.af);
5555	PF_ACPY(&key.addr[0], pd->dst, key.af);
5556	key.port[1] = key.port[0] = 0;
5557	}
5558
5559	#ifdef __FreeBSD__
5560	STATE_LOOKUP(kif, &key, direction, *state, m, pd->pf_mtag);
5561	#else
5562	STATE_LOOKUP(kif, &key, direction, *state, m);
5563	#endif
5564
5565	if (direction == (*state)->direction) {
5566	src = &(*state)->src;
5567	dst = &(*state)->dst;
5568	} else {
5569	src = &(*state)->dst;
5570	dst = &(*state)->src;
5571	}
5572
5573	/* update states */
5574	if (src->state < PFOTHERS_SINGLE)
5575	src->state = PFOTHERS_SINGLE;
5576	if (dst->state == PFOTHERS_SINGLE)
5577	dst->state = PFOTHERS_MULTIPLE;
5578
5579	/* update expire time */
5580	(*state)->expire = time_second;
5581	if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE)
5582	(*state)->timeout = PFTM_OTHER_MULTIPLE;
5583	else
5584	(*state)->timeout = PFTM_OTHER_SINGLE;
5585
5586	/* translate source/destination address, if necessary */
5587	if ((state)->key[PF_SK_WIRE] != (state)->key[PF_SK_STACK]) {
5588	struct pf_state_key nk = (state)->key[pd->didx];
5589
5590	#ifdef __FreeBSD__
5591	KASSERT(nk, ("%s: nk is null", __FUNCTION__));
5592	KASSERT(pd, ("%s: pd is null", __FUNCTION__));
5593	KASSERT(pd->src, ("%s: pd->src is null", __FUNCTION__));
5594	KASSERT(pd->dst, ("%s: pd->dst is null", __FUNCTION__));
5595	#else
5596	KASSERT(nk);
5597	KASSERT(pd);
5598	KASSERT(pd->src);
5599	KASSERT(pd->dst);
5600	#endif
5601	switch (pd->af) {
5602	#ifdef INET
5603	case AF_INET:
5604	if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], AF_INET))
5605	pf_change_a(&pd->src->v4.s_addr,
5606	pd->ip_sum,
5607	nk->addr[pd->sidx].v4.s_addr,
5608	0);
5609
5610
5611	if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], AF_INET))
5612	pf_change_a(&pd->dst->v4.s_addr,
5613	pd->ip_sum,
5614	nk->addr[pd->didx].v4.s_addr,
5615	0);
5616
5617	break;
5618	#endif /* INET */
5619	#ifdef INET6
5620	case AF_INET6:
5621	if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], AF_INET))
5622	PF_ACPY(pd->src, &nk->addr[pd->sidx], pd->af);
5623
5624	if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], AF_INET))
5625	PF_ACPY(pd->dst, &nk->addr[pd->didx], pd->af);
5626	#endif /* INET6 */
5627	}
5628	}
5629	return (PF_PASS);
5630	}
5631
5632	/*
5633	* ipoff and off are measured from the start of the mbuf chain.
5634	* h must be at "ipoff" on the mbuf chain.
5635	*/
5636	void *
5637	pf_pull_hdr(struct mbuf m, int off, void p, int len,
5638	u_short actionp, u_short reasonp, sa_family_t af)
5639	{
5640	switch (af) {
5641	#ifdef INET
5642	case AF_INET: {
5643	struct ip h = mtod(m, struct ip );
5644	u_int16_t fragoff = (ntohs(h->ip_off) & IP_OFFMASK) << 3;
5645
5646	if (fragoff) {
5647	if (fragoff >= len)
5648	ACTION_SET(actionp, PF_PASS);
5649	else {
5650	ACTION_SET(actionp, PF_DROP);
5651	REASON_SET(reasonp, PFRES_FRAG);
5652	}
5653	return (NULL);
5654	}
5655	if (m->m_pkthdr.len < off + len \|\|
5656	ntohs(h->ip_len) < off + len) {
5657	ACTION_SET(actionp, PF_DROP);
5658	REASON_SET(reasonp, PFRES_SHORT);
5659	return (NULL);
5660	}
5661	break;
5662	}
5663	#endif /* INET */
5664	#ifdef INET6
5665	case AF_INET6: {
5666	struct ip6_hdr h = mtod(m, struct ip6_hdr );
5667
5668	if (m->m_pkthdr.len < off + len \|\|
5669	(ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) <
5670	(unsigned)(off + len)) {
5671	ACTION_SET(actionp, PF_DROP);
5672	REASON_SET(reasonp, PFRES_SHORT);
5673	return (NULL);
5674	}
5675	break;
5676	}
5677	#endif /* INET6 */
5678	}
5679	m_copydata(m, off, len, p);
5680	return (p);
5681	}
5682
5683	int
5684	pf_routable(struct pf_addr addr, sa_family_t af, struct pfi_kif kif,
5685	int rtableid)
5686	{
5687	#ifdef __FreeBSD__
5688	#ifdef RADIX_MPATH
5689	struct radix_node_head *rnh;
5690	#endif
5691	#endif
5692	struct sockaddr_in *dst;
5693	int ret = 1;
5694	int check_mpath;
5695	#ifndef __FreeBSD__
5696	extern int ipmultipath;
5697	#endif
5698	#ifdef INET6
5699	#ifndef __FreeBSD__
5700	extern int ip6_multipath;
5701	#endif
5702	struct sockaddr_in6 *dst6;
5703	struct route_in6 ro;
5704	#else
5705	struct route ro;
5706	#endif
5707	struct radix_node *rn;
5708	struct rtentry *rt;
5709	struct ifnet *ifp;
5710
5711	check_mpath = 0;
5712	#ifdef __FreeBSD__
5713	#ifdef RADIX_MPATH
5714	/* XXX: stick to table 0 for now */
5715	rnh = rt_tables_get_rnh(0, af);
5716	if (rnh != NULL && rn_mpath_capable(rnh))
5717	check_mpath = 1;
5718	#endif
5719	#endif
5720	bzero(&ro, sizeof(ro));
5721	switch (af) {
5722	case AF_INET:
5723	dst = satosin(&ro.ro_dst);
5724	dst->sin_family = AF_INET;
5725	dst->sin_len = sizeof(*dst);
5726	dst->sin_addr = addr->v4;
5727	#ifndef __FreeBSD__
5728	if (ipmultipath)
5729	check_mpath = 1;
5730	#endif
5731	break;
5732	#ifdef INET6
5733	case AF_INET6:
5734	/*
5735	* Skip check for addresses with embedded interface scope,
5736	* as they would always match anyway.
5737	*/
5738	if (IN6_IS_SCOPE_EMBED(&addr->v6))
5739	goto out;
5740	dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
5741	dst6->sin6_family = AF_INET6;
5742	dst6->sin6_len = sizeof(*dst6);
5743	dst6->sin6_addr = addr->v6;
5744	#ifndef __FreeBSD__
5745	if (ip6_multipath)
5746	check_mpath = 1;
5747	#endif
5748	break;
5749	#endif /* INET6 */
5750	default:
5751	return (0);
5752	}
5753
5754	/* Skip checks for ipsec interfaces */
5755	if (kif != NULL && kif->pfik_ifp->if_type == IFT_ENC)
5756	goto out;
5757
5758	#ifdef __FreeBSD__
5759	switch (af) {
5760	#ifdef INET6
5761	case AF_INET6:
5762	in6_rtalloc_ign(&ro, 0, rtableid);
5763	break;
5764	#endif
5765	#ifdef INET
5766	case AF_INET:
5767	in_rtalloc_ign((struct route *)&ro, 0, rtableid);
5768	break;
5769	#endif
5770	default:
5771	rtalloc_ign((struct route )&ro, 0); / No/default FIB. */
5772	break;
5773	}
5774	#else /* ! __FreeBSD__ */
5775	rtalloc_noclone((struct route *)&ro, NO_CLONING);
5776	#endif
5777
5778	if (ro.ro_rt != NULL) {
5779	/* No interface given, this is a no-route check */
5780	if (kif == NULL)
5781	goto out;
5782
5783	if (kif->pfik_ifp == NULL) {
5784	ret = 0;
5785	goto out;
5786	}
5787
5788	/* Perform uRPF check if passed input interface */
5789	ret = 0;
5790	rn = (struct radix_node *)ro.ro_rt;
5791	do {
5792	rt = (struct rtentry *)rn;
5793	#ifndef __FreeBSD__ /* CARPDEV */
5794	if (rt->rt_ifp->if_type == IFT_CARP)
5795	ifp = rt->rt_ifp->if_carpdev;
5796	else
5797	#endif
5798	ifp = rt->rt_ifp;
5799
5800	if (kif->pfik_ifp == ifp)
5801	ret = 1;
5802	#ifdef __FreeBSD__
5803	#ifdef RADIX_MPATH
5804	rn = rn_mpath_next(rn);
5805	#endif
5806	#else
5807	rn = rn_mpath_next(rn, 0);
5808	#endif
5809	} while (check_mpath == 1 && rn != NULL && ret == 0);
5810	} else
5811	ret = 0;
5812	out:
5813	if (ro.ro_rt != NULL)
5814	RTFREE(ro.ro_rt);
5815	return (ret);
5816	}
5817
5818	int
5819	pf_rtlabel_match(struct pf_addr addr, sa_family_t af, struct pf_addr_wrap aw,
5820	int rtableid)
5821	{
5822	struct sockaddr_in *dst;
5823	#ifdef INET6
5824	struct sockaddr_in6 *dst6;
5825	struct route_in6 ro;
5826	#else
5827	struct route ro;
5828	#endif
5829	int ret = 0;
5830
5831	bzero(&ro, sizeof(ro));
5832	switch (af) {
5833	case AF_INET:
5834	dst = satosin(&ro.ro_dst);
5835	dst->sin_family = AF_INET;
5836	dst->sin_len = sizeof(*dst);
5837	dst->sin_addr = addr->v4;
5838	break;
5839	#ifdef INET6
5840	case AF_INET6:
5841	dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
5842	dst6->sin6_family = AF_INET6;
5843	dst6->sin6_len = sizeof(*dst6);
5844	dst6->sin6_addr = addr->v6;
5845	break;
5846	#endif /* INET6 */
5847	default:
5848	return (0);
5849	}
5850
5851	#ifdef __FreeBSD__
5852	switch (af) {
5853	#ifdef INET6
5854	case AF_INET6:
5855	in6_rtalloc_ign(&ro, 0, rtableid);
5856	break;
5857	#endif
5858	#ifdef INET
5859	case AF_INET:
5860	in_rtalloc_ign((struct route *)&ro, 0, rtableid);
5861	break;
5862	#endif
5863	default:
5864	rtalloc_ign((struct route *)&ro, 0);
5865	break;
5866	}
5867	#else /* ! __FreeBSD__ */
5868	rtalloc_noclone((struct route *)&ro, NO_CLONING);
5869	#endif
5870
5871	if (ro.ro_rt != NULL) {
5872	#ifdef __FreeBSD__
5873	/* XXX_IMPORT: later */
5874	#else
5875	if (ro.ro_rt->rt_labelid == aw->v.rtlabel)
5876	ret = 1;
5877	#endif
5878	RTFREE(ro.ro_rt);
5879	}
5880
5881	return (ret);
5882	}
5883
5884	#ifdef INET
5885	void
5886	pf_route(struct mbuf *m, struct pf_rule r, int dir, struct ifnet *oifp,
5887	struct pf_state s, struct pf_pdesc pd)
5888	{
5889	struct mbuf m0, m1;
5890	struct route iproute;
5891	struct route *ro = NULL;
5892	struct sockaddr_in *dst;
5893	struct ip *ip;
5894	struct ifnet *ifp = NULL;
5895	struct pf_addr naddr;
5896	struct pf_src_node *sn = NULL;
5897	int error = 0;
5898	#ifdef __FreeBSD__
5899	int sw_csum;
5900	#endif
5901	#ifdef IPSEC
5902	struct m_tag *mtag;
5903	#endif /* IPSEC */
5904
5905	if (m == NULL \|\| *m == NULL \|\| r == NULL \|\|
5906	(dir != PF_IN && dir != PF_OUT) \|\| oifp == NULL)
5907	panic("pf_route: invalid parameters");
5908
5909	#ifdef __FreeBSD__
5910	if (pd->pf_mtag->routed++ > 3) {
5911	#else
5912	if ((*m)->m_pkthdr.pf.routed++ > 3) {
5913	#endif
5914	m0 = *m;
5915	*m = NULL;
5916	goto bad;
5917	}
5918
5919	if (r->rt == PF_DUPTO) {
5920	#ifdef __FreeBSD__
5921	if ((m0 = m_dup(*m, M_DONTWAIT)) == NULL)
5922	#else
5923	if ((m0 = m_copym2(*m, 0, M_COPYALL, M_NOWAIT)) == NULL)
5924	#endif
5925	return;
5926	} else {
5927	if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5928	return;
5929	m0 = *m;
5930	}
5931
5932	if (m0->m_len < sizeof(struct ip)) {
5933	DPFPRINTF(PF_DEBUG_URGENT,
5934	("pf_route: m0->m_len < sizeof(struct ip)\n"));
5935	goto bad;
5936	}
5937
5938	ip = mtod(m0, struct ip *);
5939
5940	ro = &iproute;
5941	bzero((caddr_t)ro, sizeof(*ro));
5942	dst = satosin(&ro->ro_dst);
5943	dst->sin_family = AF_INET;
5944	dst->sin_len = sizeof(*dst);
5945	dst->sin_addr = ip->ip_dst;
5946
5947	if (r->rt == PF_FASTROUTE) {
5948	#ifdef __FreeBSD__
5949	in_rtalloc_ign(ro, 0, M_GETFIB(m0));
5950	#else
5951	rtalloc(ro);
5952	#endif
5953	if (ro->ro_rt == 0) {
5954	#ifdef __FreeBSD__
5955	KMOD_IPSTAT_INC(ips_noroute);
5956	#else
5957	ipstat.ips_noroute++;
5958	#endif
5959	goto bad;
5960	}
5961
5962	ifp = ro->ro_rt->rt_ifp;
5963	ro->ro_rt->rt_use++;
5964
5965	if (ro->ro_rt->rt_flags & RTF_GATEWAY)
5966	dst = satosin(ro->ro_rt->rt_gateway);
5967	} else {
5968	if (TAILQ_EMPTY(&r->rpool.list)) {
5969	DPFPRINTF(PF_DEBUG_URGENT,
5970	("pf_route: TAILQ_EMPTY(&r->rpool.list)\n"));
5971	goto bad;
5972	}
5973	if (s == NULL) {
5974	pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src,
5975	&naddr, NULL, &sn);
5976	if (!PF_AZERO(&naddr, AF_INET))
5977	dst->sin_addr.s_addr = naddr.v4.s_addr;
5978	ifp = r->rpool.cur->kif ?
5979	r->rpool.cur->kif->pfik_ifp : NULL;
5980	} else {
5981	if (!PF_AZERO(&s->rt_addr, AF_INET))
5982	dst->sin_addr.s_addr =
5983	s->rt_addr.v4.s_addr;
5984	ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5985	}
5986	}
5987	if (ifp == NULL)
5988	goto bad;
5989
5990	if (oifp != ifp) {
5991	#ifdef __FreeBSD__
5992	PF_UNLOCK();
5993	if (pf_test(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) {
5994	PF_LOCK();
5995	goto bad;
5996	} else if (m0 == NULL) {
5997	PF_LOCK();
5998	goto done;
5999	}
6000	PF_LOCK();
6001	#else
6002	if (pf_test(PF_OUT, ifp, &m0, NULL) != PF_PASS)
6003	goto bad;
6004	else if (m0 == NULL)
6005	goto done;
6006	#endif
6007	if (m0->m_len < sizeof(struct ip)) {
6008	DPFPRINTF(PF_DEBUG_URGENT,
6009	("pf_route: m0->m_len < sizeof(struct ip)\n"));
6010	goto bad;
6011	}
6012	ip = mtod(m0, struct ip *);
6013	}
6014
6015	#ifdef __FreeBSD__
6016	/* Copied from FreeBSD 5.1-CURRENT ip_output. */
6017	m0->m_pkthdr.csum_flags \|= CSUM_IP;
6018	sw_csum = m0->m_pkthdr.csum_flags & ~ifp->if_hwassist;
6019	if (sw_csum & CSUM_DELAY_DATA) {
6020	/*
6021	* XXX: in_delayed_cksum assumes HBO for ip->ip_len (at least)
6022	*/
6023	NTOHS(ip->ip_len);
6024	NTOHS(ip->ip_off); /* XXX: needed? */
6025	in_delayed_cksum(m0);
6026	HTONS(ip->ip_len);
6027	HTONS(ip->ip_off);
6028	sw_csum &= ~CSUM_DELAY_DATA;
6029	}
6030	m0->m_pkthdr.csum_flags &= ifp->if_hwassist;
6031
6032	if (ntohs(ip->ip_len) <= ifp->if_mtu \|\|
6033	(ifp->if_hwassist & CSUM_FRAGMENT &&
6034	((ip->ip_off & htons(IP_DF)) == 0))) {
6035	/*
6036	* ip->ip_len = htons(ip->ip_len);
6037	* ip->ip_off = htons(ip->ip_off);
6038	*/
6039	ip->ip_sum = 0;
6040	if (sw_csum & CSUM_DELAY_IP) {
6041	/* From KAME */
6042	if (ip->ip_v == IPVERSION &&
6043	(ip->ip_hl << 2) == sizeof(*ip)) {
6044	ip->ip_sum = in_cksum_hdr(ip);
6045	} else {
6046	ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
6047	}
6048	}
6049	PF_UNLOCK();
6050	error = (*ifp->if_output)(ifp, m0, sintosa(dst), ro);
6051	PF_LOCK();
6052	goto done;
6053	}
6054	#else
6055	/* Copied from ip_output. */
6056	#ifdef IPSEC
6057	/*
6058	* If deferred crypto processing is needed, check that the
6059	* interface supports it.
6060	*/
6061	if ((mtag = m_tag_find(m0, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL))
6062	!= NULL && (ifp->if_capabilities & IFCAP_IPSEC) == 0) {
6063	/* Notify IPsec to do its own crypto. */
6064	ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1));
6065	goto bad;
6066	}
6067	#endif /* IPSEC */
6068
6069	/* Catch routing changes wrt. hardware checksumming for TCP or UDP. */
6070	if (m0->m_pkthdr.csum_flags & M_TCPV4_CSUM_OUT) {
6071	if (!(ifp->if_capabilities & IFCAP_CSUM_TCPv4) \|\|
6072	ifp->if_bridge != NULL) {
6073	in_delayed_cksum(m0);
6074	m0->m_pkthdr.csum_flags &= ~M_TCPV4_CSUM_OUT; /* Clr */
6075	}
6076	} else if (m0->m_pkthdr.csum_flags & M_UDPV4_CSUM_OUT) {
6077	if (!(ifp->if_capabilities & IFCAP_CSUM_UDPv4) \|\|
6078	ifp->if_bridge != NULL) {
6079	in_delayed_cksum(m0);
6080	m0->m_pkthdr.csum_flags &= ~M_UDPV4_CSUM_OUT; /* Clr */
6081	}
6082	}
6083
6084	if (ntohs(ip->ip_len) <= ifp->if_mtu) {
6085	ip->ip_sum = 0;
6086	if ((ifp->if_capabilities & IFCAP_CSUM_IPv4) &&
6087	ifp->if_bridge == NULL) {
6088	m0->m_pkthdr.csum_flags \|= M_IPV4_CSUM_OUT;
6089	#ifdef __FreeBSD__
6090	KMOD_IPSTAT_INC(ips_outhwcsum);
6091	#else
6092	ipstat.ips_outhwcsum++;
6093	#endif
6094	} else
6095	ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
6096	/* Update relevant hardware checksum stats for TCP/UDP */
6097	if (m0->m_pkthdr.csum_flags & M_TCPV4_CSUM_OUT)
6098	KMOD_TCPSTAT_INC(tcps_outhwcsum);
6099	else if (m0->m_pkthdr.csum_flags & M_UDPV4_CSUM_OUT)
6100	KMOD_UDPSTAT_INC(udps_outhwcsum);
6101	error = (*ifp->if_output)(ifp, m0, sintosa(dst), NULL);
6102	goto done;
6103	}
6104	#endif
6105
6106	/*
6107	* Too large for interface; fragment if possible.
6108	* Must be able to put at least 8 bytes per fragment.
6109	*/
6110	if (ip->ip_off & htons(IP_DF)) {
6111	#ifdef __FreeBSD__
6112	KMOD_IPSTAT_INC(ips_cantfrag);
6113	#else
6114	ipstat.ips_cantfrag++;
6115	#endif
6116	if (r->rt != PF_DUPTO) {
6117	#ifdef __FreeBSD__
6118	/* icmp_error() expects host byte ordering */
6119	NTOHS(ip->ip_len);
6120	NTOHS(ip->ip_off);
6121	PF_UNLOCK();
6122	icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0,
6123	ifp->if_mtu);
6124	PF_LOCK();
6125	#else
6126	icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0,
6127	ifp->if_mtu);
6128	#endif
6129	goto done;
6130	} else
6131	goto bad;
6132	}
6133
6134	m1 = m0;
6135	#ifdef __FreeBSD__
6136	/*
6137	* XXX: is cheaper + less error prone than own function
6138	*/
6139	NTOHS(ip->ip_len);
6140	NTOHS(ip->ip_off);
6141	error = ip_fragment(ip, &m0, ifp->if_mtu, ifp->if_hwassist, sw_csum);
6142	#else
6143	error = ip_fragment(m0, ifp, ifp->if_mtu);
6144	#endif
6145	if (error) {
6146	#ifndef __FreeBSD__ /* ip_fragment does not do m_freem() on FreeBSD */
6147	m0 = NULL;
6148	#endif
6149	goto bad;
6150	}
6151
6152	for (m0 = m1; m0; m0 = m1) {
6153	m1 = m0->m_nextpkt;
6154	m0->m_nextpkt = 0;
6155	#ifdef __FreeBSD__
6156	if (error == 0) {
6157	PF_UNLOCK();
6158	error = (*ifp->if_output)(ifp, m0, sintosa(dst),
6159	NULL);
6160	PF_LOCK();
6161	} else
6162	#else
6163	if (error == 0)
6164	error = (*ifp->if_output)(ifp, m0, sintosa(dst),
6165	NULL);
6166	else
6167	#endif
6168	m_freem(m0);
6169	}
6170
6171	if (error == 0)
6172	#ifdef __FreeBSD__
6173	KMOD_IPSTAT_INC(ips_fragmented);
6174	#else
6175	ipstat.ips_fragmented++;
6176	#endif
6177
6178	done:
6179	if (r->rt != PF_DUPTO)
6180	*m = NULL;
6181	if (ro == &iproute && ro->ro_rt)
6182	RTFREE(ro->ro_rt);
6183	return;
6184
6185	bad:
6186	m_freem(m0);
6187	goto done;
6188	}
6189	#endif /* INET */
6190
6191	#ifdef INET6
6192	void
6193	pf_route6(struct mbuf *m, struct pf_rule r, int dir, struct ifnet *oifp,
6194	struct pf_state s, struct pf_pdesc pd)
6195	{
6196	struct mbuf *m0;
6197	struct route_in6 ip6route;
6198	struct route_in6 *ro;
6199	struct sockaddr_in6 *dst;
6200	struct ip6_hdr *ip6;
6201	struct ifnet *ifp = NULL;
6202	struct pf_addr naddr;
6203	struct pf_src_node *sn = NULL;
6204
6205	if (m == NULL \|\| *m == NULL \|\| r == NULL \|\|
6206	(dir != PF_IN && dir != PF_OUT) \|\| oifp == NULL)
6207	panic("pf_route6: invalid parameters");
6208
6209	#ifdef __FreeBSD__
6210	if (pd->pf_mtag->routed++ > 3) {
6211	#else
6212	if ((*m)->m_pkthdr.pf.routed++ > 3) {
6213	#endif
6214	m0 = *m;
6215	*m = NULL;
6216	goto bad;
6217	}
6218
6219	if (r->rt == PF_DUPTO) {
6220	#ifdef __FreeBSD__
6221	if ((m0 = m_dup(*m, M_DONTWAIT)) == NULL)
6222	#else
6223	if ((m0 = m_copym2(*m, 0, M_COPYALL, M_NOWAIT)) == NULL)
6224	#endif
6225	return;
6226	} else {
6227	if ((r->rt == PF_REPLYTO) == (r->direction == dir))
6228	return;
6229	m0 = *m;
6230	}
6231
6232	if (m0->m_len < sizeof(struct ip6_hdr)) {
6233	DPFPRINTF(PF_DEBUG_URGENT,
6234	("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n"));
6235	goto bad;
6236	}
6237	ip6 = mtod(m0, struct ip6_hdr *);
6238
6239	ro = &ip6route;
6240	bzero((caddr_t)ro, sizeof(*ro));
6241	dst = (struct sockaddr_in6 *)&ro->ro_dst;
6242	dst->sin6_family = AF_INET6;
6243	dst->sin6_len = sizeof(*dst);
6244	dst->sin6_addr = ip6->ip6_dst;
6245
6246	/* Cheat. XXX why only in the v6 case??? */
6247	if (r->rt == PF_FASTROUTE) {
6248	#ifdef __FreeBSD__
6249	m0->m_flags \|= M_SKIP_FIREWALL;
6250	PF_UNLOCK();
6251	ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
6252	#else
6253	m0->m_pkthdr.pf.flags \|= PF_TAG_GENERATED;
6254	ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
6255	#endif
6256	return;
6257	}
6258
6259	if (TAILQ_EMPTY(&r->rpool.list)) {
6260	DPFPRINTF(PF_DEBUG_URGENT,
6261	("pf_route6: TAILQ_EMPTY(&r->rpool.list)\n"));
6262	goto bad;
6263	}
6264	if (s == NULL) {
6265	pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src,
6266	&naddr, NULL, &sn);
6267	if (!PF_AZERO(&naddr, AF_INET6))
6268	PF_ACPY((struct pf_addr *)&dst->sin6_addr,
6269	&naddr, AF_INET6);
6270	ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL;
6271	} else {
6272	if (!PF_AZERO(&s->rt_addr, AF_INET6))
6273	PF_ACPY((struct pf_addr *)&dst->sin6_addr,
6274	&s->rt_addr, AF_INET6);
6275	ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
6276	}
6277	if (ifp == NULL)
6278	goto bad;
6279
6280	if (oifp != ifp) {
6281	#ifdef __FreeBSD__
6282	PF_UNLOCK();
6283	if (pf_test6(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) {
6284	PF_LOCK();
6285	goto bad;
6286	} else if (m0 == NULL) {
6287	PF_LOCK();
6288	goto done;
6289	}
6290	PF_LOCK();
6291	#else
6292	if (pf_test6(PF_OUT, ifp, &m0, NULL) != PF_PASS)
6293	goto bad;
6294	else if (m0 == NULL)
6295	goto done;
6296	#endif
6297	if (m0->m_len < sizeof(struct ip6_hdr)) {
6298	DPFPRINTF(PF_DEBUG_URGENT,
6299	("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n"));
6300	goto bad;
6301	}
6302	ip6 = mtod(m0, struct ip6_hdr *);
6303	}
6304
6305	/*
6306	* If the packet is too large for the outgoing interface,
6307	* send back an icmp6 error.
6308	*/
6309	if (IN6_IS_SCOPE_EMBED(&dst->sin6_addr))
6310	dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
6311	if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) {
6312	#ifdef __FreeBSD__
6313	PF_UNLOCK();
6314	#endif
6315	nd6_output(ifp, ifp, m0, dst, NULL);
6316	#ifdef __FreeBSD__
6317	PF_LOCK();
6318	#endif
6319	} else {
6320	in6_ifstat_inc(ifp, ifs6_in_toobig);
6321	#ifdef __FreeBSD__
6322	if (r->rt != PF_DUPTO) {
6323	PF_UNLOCK();
6324	icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
6325	PF_LOCK();
6326	} else
6327	#else
6328	if (r->rt != PF_DUPTO)
6329	icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
6330	else
6331	#endif
6332	goto bad;
6333	}
6334
6335	done:
6336	if (r->rt != PF_DUPTO)
6337	*m = NULL;
6338	return;
6339
6340	bad:
6341	m_freem(m0);
6342	goto done;
6343	}
6344	#endif /* INET6 */
6345
6346	#ifdef __FreeBSD__
6347	/*
6348	* FreeBSD supports cksum offloads for the following drivers.
6349	* em(4), fxp(4), ixgb(4), lge(4), ndis(4), nge(4), re(4),
6350	* ti(4), txp(4), xl(4)
6351	*
6352	* CSUM_DATA_VALID \| CSUM_PSEUDO_HDR :
6353	* network driver performed cksum including pseudo header, need to verify
6354	* csum_data
6355	* CSUM_DATA_VALID :
6356	* network driver performed cksum, needs to additional pseudo header
6357	* cksum computation with partial csum_data(i.e. lack of H/W support for
6358	* pseudo header, for instance hme(4), sk(4) and possibly gem(4))
6359	*
6360	* After validating the cksum of packet, set both flag CSUM_DATA_VALID and
6361	* CSUM_PSEUDO_HDR in order to avoid recomputation of the cksum in upper
6362	* TCP/UDP layer.
6363	* Also, set csum_data to 0xffff to force cksum validation.
6364	*/
6365	int
6366	pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p, sa_family_t af)
6367	{
6368	u_int16_t sum = 0;
6369	int hw_assist = 0;
6370	struct ip *ip;
6371
6372	if (off < sizeof(struct ip) \|\| len < sizeof(struct udphdr))
6373	return (1);
6374	if (m->m_pkthdr.len < off + len)
6375	return (1);
6376
6377	switch (p) {
6378	case IPPROTO_TCP:
6379	if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
6380	if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) {
6381	sum = m->m_pkthdr.csum_data;
6382	} else {
6383	ip = mtod(m, struct ip *);
6384	sum = in_pseudo(ip->ip_src.s_addr,
6385	ip->ip_dst.s_addr, htonl((u_short)len +
6386	m->m_pkthdr.csum_data + IPPROTO_TCP));
6387	}
6388	sum ^= 0xffff;
6389	++hw_assist;
6390	}
6391	break;
6392	case IPPROTO_UDP:
6393	if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
6394	if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) {
6395	sum = m->m_pkthdr.csum_data;
6396	} else {
6397	ip = mtod(m, struct ip *);
6398	sum = in_pseudo(ip->ip_src.s_addr,
6399	ip->ip_dst.s_addr, htonl((u_short)len +
6400	m->m_pkthdr.csum_data + IPPROTO_UDP));
6401	}
6402	sum ^= 0xffff;
6403	++hw_assist;
6404	}
6405	break;
6406	case IPPROTO_ICMP:
6407	#ifdef INET6
6408	case IPPROTO_ICMPV6:
6409	#endif /* INET6 */
6410	break;
6411	default:
6412	return (1);
6413	}
6414
6415	if (!hw_assist) {
6416	switch (af) {
6417	case AF_INET:
6418	if (p == IPPROTO_ICMP) {
6419	if (m->m_len < off)
6420	return (1);
6421	m->m_data += off;
6422	m->m_len -= off;
6423	sum = in_cksum(m, len);
6424	m->m_data -= off;
6425	m->m_len += off;
6426	} else {
6427	if (m->m_len < sizeof(struct ip))
6428	return (1);
6429	sum = in4_cksum(m, p, off, len);
6430	}
6431	break;
6432	#ifdef INET6
6433	case AF_INET6:
6434	if (m->m_len < sizeof(struct ip6_hdr))
6435	return (1);
6436	sum = in6_cksum(m, p, off, len);
6437	break;
6438	#endif /* INET6 */
6439	default:
6440	return (1);
6441	}
6442	}
6443	if (sum) {
6444	switch (p) {
6445	case IPPROTO_TCP:
6446	{
6447	KMOD_TCPSTAT_INC(tcps_rcvbadsum);
6448	break;
6449	}
6450	case IPPROTO_UDP:
6451	{
6452	KMOD_UDPSTAT_INC(udps_badsum);
6453	break;
6454	}
6455	#ifdef INET
6456	case IPPROTO_ICMP:
6457	{
6458	KMOD_ICMPSTAT_INC(icps_checksum);
6459	break;
6460	}
6461	#endif
6462	#ifdef INET6
6463	case IPPROTO_ICMPV6:
6464	{
6465	KMOD_ICMP6STAT_INC(icp6s_checksum);
6466	break;
6467	}
6468	#endif /* INET6 */
6469	}
6470	return (1);
6471	} else {
6472	if (p == IPPROTO_TCP \|\| p == IPPROTO_UDP) {
6473	m->m_pkthdr.csum_flags \|=
6474	(CSUM_DATA_VALID \| CSUM_PSEUDO_HDR);
6475	m->m_pkthdr.csum_data = 0xffff;
6476	}
6477	}
6478	return (0);
6479	}
6480	#else /* !__FreeBSD__ */
6481
6482	/*
6483	* check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag
6484	* off is the offset where the protocol header starts
6485	* len is the total length of protocol header plus payload
6486	* returns 0 when the checksum is valid, otherwise returns 1.
6487	*/
6488	int
6489	pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p,
6490	sa_family_t af)
6491	{
6492	u_int16_t flag_ok, flag_bad;
6493	u_int16_t sum;
6494
6495	switch (p) {
6496	case IPPROTO_TCP:
6497	flag_ok = M_TCP_CSUM_IN_OK;
6498	flag_bad = M_TCP_CSUM_IN_BAD;
6499	break;
6500	case IPPROTO_UDP:
6501	flag_ok = M_UDP_CSUM_IN_OK;
6502	flag_bad = M_UDP_CSUM_IN_BAD;
6503	break;
6504	case IPPROTO_ICMP:
6505	#ifdef INET6
6506	case IPPROTO_ICMPV6:
6507	#endif /* INET6 */
6508	flag_ok = flag_bad = 0;
6509	break;
6510	default:
6511	return (1);
6512	}
6513	if (m->m_pkthdr.csum_flags & flag_ok)
6514	return (0);
6515	if (m->m_pkthdr.csum_flags & flag_bad)
6516	return (1);
6517	if (off < sizeof(struct ip) \|\| len < sizeof(struct udphdr))
6518	return (1);
6519	if (m->m_pkthdr.len < off + len)
6520	return (1);
6521	switch (af) {
6522	#ifdef INET
6523	case AF_INET:
6524	if (p == IPPROTO_ICMP) {
6525	if (m->m_len < off)
6526	return (1);
6527	m->m_data += off;
6528	m->m_len -= off;
6529	sum = in_cksum(m, len);
6530	m->m_data -= off;
6531	m->m_len += off;
6532	} else {
6533	if (m->m_len < sizeof(struct ip))
6534	return (1);
6535	sum = in4_cksum(m, p, off, len);
6536	}
6537	break;
6538	#endif /* INET */
6539	#ifdef INET6
6540	case AF_INET6:
6541	if (m->m_len < sizeof(struct ip6_hdr))
6542	return (1);
6543	sum = in6_cksum(m, p, off, len);
6544	break;
6545	#endif /* INET6 */
6546	default:
6547	return (1);
6548	}
6549	if (sum) {
6550	m->m_pkthdr.csum_flags \|= flag_bad;
6551	switch (p) {
6552	case IPPROTO_TCP:
6553	KMOD_TCPSTAT_INC(tcps_rcvbadsum);
6554	break;
6555	case IPPROTO_UDP:
6556	KMOD_UDPSTAT_INC(udps_badsum);
6557	break;
6558	#ifdef INET
6559	case IPPROTO_ICMP:
6560	KMOD_ICMPSTAT_INC(icps_checksum);
6561	break;
6562	#endif
6563	#ifdef INET6
6564	case IPPROTO_ICMPV6:
6565	KMOD_ICMP6STAT_INC(icp6s_checksum);
6566	break;
6567	#endif /* INET6 */
6568	}
6569	return (1);
6570	}
6571	m->m_pkthdr.csum_flags \|= flag_ok;
6572	return (0);
6573	}
6574	#endif
6575
6576	#ifndef __FreeBSD__
6577	struct pf_divert *
6578	pf_find_divert(struct mbuf *m)
6579	{
6580	struct m_tag *mtag;
6581
6582	if ((mtag = m_tag_find(m, PACKET_TAG_PF_DIVERT, NULL)) == NULL)
6583	return (NULL);
6584
6585	return ((struct pf_divert *)(mtag + 1));
6586	}
6587
6588	struct pf_divert *
6589	pf_get_divert(struct mbuf *m)
6590	{
6591	struct m_tag *mtag;
6592
6593	if ((mtag = m_tag_find(m, PACKET_TAG_PF_DIVERT, NULL)) == NULL) {
6594	mtag = m_tag_get(PACKET_TAG_PF_DIVERT, sizeof(struct pf_divert),
6595	M_NOWAIT);
6596	if (mtag == NULL)
6597	return (NULL);
6598	bzero(mtag + 1, sizeof(struct pf_divert));
6599	m_tag_prepend(m, mtag);
6600	}
6601
6602	return ((struct pf_divert *)(mtag + 1));
6603	}
6604	#endif
6605
6606	#ifdef INET
6607	int
6608	#ifdef __FreeBSD__
6609	pf_test(int dir, struct ifnet ifp, struct mbuf *m0,
6610	struct ether_header eh, struct inpcb inp)
6611	#else
6612	pf_test(int dir, struct ifnet ifp, struct mbuf *m0,
6613	struct ether_header *eh)
6614	#endif
6615	{
6616	struct pfi_kif *kif;
6617	u_short action, reason = 0, log = 0;
6618	struct mbuf m = m0;
6619	#ifdef __FreeBSD__
6620	struct ip *h = NULL;
6621	struct m_tag *ipfwtag;
6622	struct pf_rule a = NULL, r = &V_pf_default_rule, tr, nr;
6623	#else
6624	struct ip *h;
6625	struct pf_rule a = NULL, r = &pf_default_rule, tr, nr;
6626	#endif
6627	struct pf_state *s = NULL;
6628	struct pf_ruleset *ruleset = NULL;
6629	struct pf_pdesc pd;
6630	int off, dirndx, pqid = 0;
6631
6632	#ifdef __FreeBSD__
6633	PF_LOCK();
6634	if (!V_pf_status.running)
6635	{
6636	PF_UNLOCK();
6637	return (PF_PASS);
6638	}
6639	#else
6640	if (!pf_status.running)
6641	return (PF_PASS);
6642	#endif
6643
6644	memset(&pd, 0, sizeof(pd));
6645	#ifdef __FreeBSD__
6646	if ((pd.pf_mtag = pf_get_mtag(m)) == NULL) {
6647	PF_UNLOCK();
6648	DPFPRINTF(PF_DEBUG_URGENT,
6649	("pf_test: pf_get_mtag returned NULL\n"));
6650	return (PF_DROP);
6651	}
6652	#endif
6653	#ifndef __FreeBSD__
6654	if (ifp->if_type == IFT_CARP && ifp->if_carpdev)
6655	kif = (struct pfi_kif *)ifp->if_carpdev->if_pf_kif;
6656	else
6657	#endif
6658	kif = (struct pfi_kif *)ifp->if_pf_kif;
6659
6660	if (kif == NULL) {
6661	#ifdef __FreeBSD__
6662	PF_UNLOCK();
6663	#endif
6664	DPFPRINTF(PF_DEBUG_URGENT,
6665	("pf_test: kif == NULL, if_xname %s\n", ifp->if_xname));
6666	return (PF_DROP);
6667	}
6668	if (kif->pfik_flags & PFI_IFLAG_SKIP)
6669	#ifdef __FreeBSD__
6670	{
6671	PF_UNLOCK();
6672	#endif
6673	return (PF_PASS);
6674	#ifdef __FreeBSD__
6675	}
6676	#endif
6677
6678	#ifdef __FreeBSD__
6679	M_ASSERTPKTHDR(m);
6680	#else
6681	#ifdef DIAGNOSTIC
6682	if ((m->m_flags & M_PKTHDR) == 0)
6683	panic("non-M_PKTHDR is passed to pf_test");
6684	#endif /* DIAGNOSTIC */
6685	#endif
6686
6687	if (m->m_pkthdr.len < (int)sizeof(*h)) {
6688	action = PF_DROP;
6689	REASON_SET(&reason, PFRES_SHORT);
6690	log = 1;
6691	goto done;
6692	}
6693
6694	#ifdef __FreeBSD__
6695	if (m->m_flags & M_SKIP_FIREWALL) {
6696	PF_UNLOCK();
6697	return (PF_PASS);
6698	}
6699	#else
6700	if (m->m_pkthdr.pf.flags & PF_TAG_GENERATED)
6701	return (PF_PASS);
6702	#endif
6703
6704	#ifdef __FreeBSD__
6705	if (ip_divert_ptr != NULL &&
6706	((ipfwtag = m_tag_locate(m, MTAG_IPFW_RULE, 0, NULL)) != NULL)) {
6707	struct ipfw_rule_ref rr = (struct ipfw_rule_ref )(ipfwtag+1);
6708	if (rr->info & IPFW_IS_DIVERT && rr->rulenum == 0) {
6709	pd.pf_mtag->flags \|= PF_PACKET_LOOPED;
6710	m_tag_delete(m, ipfwtag);
6711	}
6712	if (pd.pf_mtag->flags & PF_FASTFWD_OURS_PRESENT) {
6713	m->m_flags \|= M_FASTFWD_OURS;
6714	pd.pf_mtag->flags &= ~PF_FASTFWD_OURS_PRESENT;
6715	}
6716	} else
6717	#endif
6718	/* We do IP header normalization and packet reassembly here */
6719	if (pf_normalize_ip(m0, dir, kif, &reason, &pd) != PF_PASS) {
6720	action = PF_DROP;
6721	goto done;
6722	}
6723	m = m0; / pf_normalize messes with m0 */
6724	h = mtod(m, struct ip *);
6725
6726	off = h->ip_hl << 2;
6727	if (off < (int)sizeof(*h)) {
6728	action = PF_DROP;
6729	REASON_SET(&reason, PFRES_SHORT);
6730	log = 1;
6731	goto done;
6732	}
6733
6734	pd.src = (struct pf_addr *)&h->ip_src;
6735	pd.dst = (struct pf_addr *)&h->ip_dst;
6736	pd.sport = pd.dport = NULL;
6737	pd.ip_sum = &h->ip_sum;
6738	pd.proto_sum = NULL;
6739	pd.proto = h->ip_p;
6740	pd.dir = dir;
6741	pd.sidx = (dir == PF_IN) ? 0 : 1;
6742	pd.didx = (dir == PF_IN) ? 1 : 0;
6743	pd.af = AF_INET;
6744	pd.tos = h->ip_tos;
6745	pd.tot_len = ntohs(h->ip_len);
6746	pd.eh = eh;
6747
6748	/* handle fragments that didn't get reassembled by normalization */
6749	if (h->ip_off & htons(IP_MF \| IP_OFFMASK)) {
6750	action = pf_test_fragment(&r, dir, kif, m, h,
6751	&pd, &a, &ruleset);
6752	goto done;
6753	}
6754
6755	switch (h->ip_p) {
6756
6757	case IPPROTO_TCP: {
6758	struct tcphdr th;
6759
6760	pd.hdr.tcp = &th;
6761	if (!pf_pull_hdr(m, off, &th, sizeof(th),
6762	&action, &reason, AF_INET)) {
6763	log = action != PF_PASS;
6764	goto done;
6765	}
6766	pd.p_len = pd.tot_len - off - (th.th_off << 2);
6767	if ((th.th_flags & TH_ACK) && pd.p_len == 0)
6768	pqid = 1;
6769	action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
6770	if (action == PF_DROP)
6771	goto done;
6772	action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
6773	&reason);
6774	if (action == PF_PASS) {
6775	#if NPFSYNC > 0
6776	#ifdef __FreeBSD__
6777	if (pfsync_update_state_ptr != NULL)
6778	pfsync_update_state_ptr(s);
6779	#else
6780	pfsync_update_state(s);
6781	#endif
6782	#endif /* NPFSYNC */
6783	r = s->rule.ptr;
6784	a = s->anchor.ptr;
6785	log = s->log;
6786	} else if (s == NULL)
6787	#ifdef __FreeBSD__
6788	action = pf_test_rule(&r, &s, dir, kif,
6789	m, off, h, &pd, &a, &ruleset, NULL, inp);
6790	#else
6791	action = pf_test_rule(&r, &s, dir, kif,
6792	m, off, h, &pd, &a, &ruleset, &ipintrq);
6793	#endif
6794	break;
6795	}
6796
6797	case IPPROTO_UDP: {
6798	struct udphdr uh;
6799
6800	pd.hdr.udp = &uh;
6801	if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
6802	&action, &reason, AF_INET)) {
6803	log = action != PF_PASS;
6804	goto done;
6805	}
6806	if (uh.uh_dport == 0 \|\|
6807	ntohs(uh.uh_ulen) > m->m_pkthdr.len - off \|\|
6808	ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
6809	action = PF_DROP;
6810	REASON_SET(&reason, PFRES_SHORT);
6811	goto done;
6812	}
6813	action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
6814	if (action == PF_PASS) {
6815	#if NPFSYNC > 0
6816	#ifdef __FreeBSD__
6817	if (pfsync_update_state_ptr != NULL)
6818	pfsync_update_state_ptr(s);
6819	#else
6820	pfsync_update_state(s);
6821	#endif
6822	#endif /* NPFSYNC */
6823	r = s->rule.ptr;
6824	a = s->anchor.ptr;
6825	log = s->log;
6826	} else if (s == NULL)
6827	#ifdef __FreeBSD__
6828	action = pf_test_rule(&r, &s, dir, kif,
6829	m, off, h, &pd, &a, &ruleset, NULL, inp);
6830	#else
6831	action = pf_test_rule(&r, &s, dir, kif,
6832	m, off, h, &pd, &a, &ruleset, &ipintrq);
6833	#endif
6834	break;
6835	}
6836
6837	case IPPROTO_ICMP: {
6838	struct icmp ih;
6839
6840	pd.hdr.icmp = &ih;
6841	if (!pf_pull_hdr(m, off, &ih, ICMP_MINLEN,
6842	&action, &reason, AF_INET)) {
6843	log = action != PF_PASS;
6844	goto done;
6845	}
6846	action = pf_test_state_icmp(&s, dir, kif, m, off, h, &pd,
6847	&reason);
6848	if (action == PF_PASS) {
6849	#if NPFSYNC > 0
6850	#ifdef __FreeBSD__
6851	if (pfsync_update_state_ptr != NULL)
6852	pfsync_update_state_ptr(s);
6853	#else
6854	pfsync_update_state(s);
6855	#endif
6856	#endif /* NPFSYNC */
6857	r = s->rule.ptr;
6858	a = s->anchor.ptr;
6859	log = s->log;
6860	} else if (s == NULL)
6861	#ifdef __FreeBSD__
6862	action = pf_test_rule(&r, &s, dir, kif,
6863	m, off, h, &pd, &a, &ruleset, NULL, inp);
6864	#else
6865	action = pf_test_rule(&r, &s, dir, kif,
6866	m, off, h, &pd, &a, &ruleset, &ipintrq);
6867	#endif
6868	break;
6869	}
6870
6871	#ifdef INET6
6872	case IPPROTO_ICMPV6: {
6873	action = PF_DROP;
6874	DPFPRINTF(PF_DEBUG_MISC,
6875	("pf: dropping IPv4 packet with ICMPv6 payload\n"));
6876	goto done;
6877	}
6878	#endif
6879
6880	default:
6881	action = pf_test_state_other(&s, dir, kif, m, &pd);
6882	if (action == PF_PASS) {
6883	#if NPFSYNC > 0
6884	#ifdef __FreeBSD__
6885	if (pfsync_update_state_ptr != NULL)
6886	pfsync_update_state_ptr(s);
6887	#else
6888	pfsync_update_state(s);
6889	#endif
6890	#endif /* NPFSYNC */
6891	r = s->rule.ptr;
6892	a = s->anchor.ptr;
6893	log = s->log;
6894	} else if (s == NULL)
6895	#ifdef __FreeBSD__
6896	action = pf_test_rule(&r, &s, dir, kif, m, off, h,
6897	&pd, &a, &ruleset, NULL, inp);
6898	#else
6899	action = pf_test_rule(&r, &s, dir, kif, m, off, h,
6900	&pd, &a, &ruleset, &ipintrq);
6901	#endif
6902	break;
6903	}
6904
6905	done:
6906	if (action == PF_PASS && h->ip_hl > 5 &&
6907	!((s && s->state_flags & PFSTATE_ALLOWOPTS) \|\| r->allow_opts)) {
6908	action = PF_DROP;
6909	REASON_SET(&reason, PFRES_IPOPTIONS);
6910	log = 1;
6911	DPFPRINTF(PF_DEBUG_MISC,
6912	("pf: dropping packet with ip options\n"));
6913	}
6914
6915	if ((s && s->tag) \|\| r->rtableid >= 0)
6916	#ifdef __FreeBSD__
6917	pf_tag_packet(m, s ? s->tag : 0, r->rtableid, pd.pf_mtag);
6918	#else
6919	pf_tag_packet(m, s ? s->tag : 0, r->rtableid);
6920	#endif
6921
6922	if (dir == PF_IN && s && s->key[PF_SK_STACK])
6923	#ifdef __FreeBSD__
6924	pd.pf_mtag->statekey = s->key[PF_SK_STACK];
6925	#else
6926	m->m_pkthdr.pf.statekey = s->key[PF_SK_STACK];
6927	#endif
6928
6929	#ifdef ALTQ
6930	if (action == PF_PASS && r->qid) {
6931	#ifdef __FreeBSD__
6932	if (pqid \|\| (pd.tos & IPTOS_LOWDELAY))
6933	pd.pf_mtag->qid = r->pqid;
6934	else
6935	pd.pf_mtag->qid = r->qid;
6936	/* add hints for ecn */
6937	pd.pf_mtag->hdr = h;
6938
6939	#else
6940	if (pqid \|\| (pd.tos & IPTOS_LOWDELAY))
6941	m->m_pkthdr.pf.qid = r->pqid;
6942	else
6943	m->m_pkthdr.pf.qid = r->qid;
6944	/* add hints for ecn */
6945	m->m_pkthdr.pf.hdr = h;
6946	#endif
6947	}
6948	#endif /* ALTQ */
6949
6950	/*
6951	* connections redirected to loopback should not match sockets
6952	* bound specifically to loopback due to security implications,
6953	* see tcp_input() and in_pcblookup_listen().
6954	*/
6955	if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP \|\|
6956	pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
6957	(s->nat_rule.ptr->action == PF_RDR \|\|
6958	s->nat_rule.ptr->action == PF_BINAT) &&
6959	(ntohl(pd.dst->v4.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET)
6960	#ifdef __FreeBSD__
6961	m->m_flags \|= M_SKIP_FIREWALL;
6962	#else
6963	m->m_pkthdr.pf.flags \|= PF_TAG_TRANSLATE_LOCALHOST;
6964	#endif
6965
6966	#ifdef __FreeBSD__
6967	if (action == PF_PASS && r->divert.port &&
6968	ip_divert_ptr != NULL && !PACKET_LOOPED()) {
6969
6970	ipfwtag = m_tag_alloc(MTAG_IPFW_RULE, 0,
6971	sizeof(struct ipfw_rule_ref), M_NOWAIT \| M_ZERO);
6972	if (ipfwtag != NULL) {
6973	((struct ipfw_rule_ref *)(ipfwtag+1))->info =
6974	ntohs(r->divert.port);
6975	((struct ipfw_rule_ref *)(ipfwtag+1))->rulenum = dir;
6976
6977	m_tag_prepend(m, ipfwtag);
6978
6979	PF_UNLOCK();
6980
6981	if (m->m_flags & M_FASTFWD_OURS) {
6982	pd.pf_mtag->flags \|= PF_FASTFWD_OURS_PRESENT;
6983	m->m_flags &= ~M_FASTFWD_OURS;
6984	}
6985
6986	ip_divert_ptr(*m0,
6987	dir == PF_IN ? DIR_IN : DIR_OUT);
6988	*m0 = NULL;
6989	return (action);
6990	} else {
6991	/* XXX: ipfw has the same behaviour! */
6992	action = PF_DROP;
6993	REASON_SET(&reason, PFRES_MEMORY);
6994	log = 1;
6995	DPFPRINTF(PF_DEBUG_MISC,
6996	("pf: failed to allocate divert tag\n"));
6997	}
6998	}
6999	#else
7000	if (dir == PF_IN && action == PF_PASS && r->divert.port) {
7001	struct pf_divert *divert;
7002
7003	if ((divert = pf_get_divert(m))) {
7004	m->m_pkthdr.pf.flags \|= PF_TAG_DIVERTED;
7005	divert->port = r->divert.port;
7006	divert->addr.ipv4 = r->divert.addr.v4;
7007	}
7008	}
7009	#endif
7010
7011	if (log) {
7012	struct pf_rule *lr;
7013
7014	if (s != NULL && s->nat_rule.ptr != NULL &&
7015	s->nat_rule.ptr->log & PF_LOG_ALL)
7016	lr = s->nat_rule.ptr;
7017	else
7018	lr = r;
7019	PFLOG_PACKET(kif, h, m, AF_INET, dir, reason, lr, a, ruleset,
7020	&pd);
7021	}
7022
7023	kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
7024	kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++;
7025
7026	if (action == PF_PASS \|\| r->action == PF_DROP) {
7027	dirndx = (dir == PF_OUT);
7028	r->packets[dirndx]++;
7029	r->bytes[dirndx] += pd.tot_len;
7030	if (a != NULL) {
7031	a->packets[dirndx]++;
7032	a->bytes[dirndx] += pd.tot_len;
7033	}
7034	if (s != NULL) {
7035	if (s->nat_rule.ptr != NULL) {
7036	s->nat_rule.ptr->packets[dirndx]++;
7037	s->nat_rule.ptr->bytes[dirndx] += pd.tot_len;
7038	}
7039	if (s->src_node != NULL) {
7040	s->src_node->packets[dirndx]++;
7041	s->src_node->bytes[dirndx] += pd.tot_len;
7042	}
7043	if (s->nat_src_node != NULL) {
7044	s->nat_src_node->packets[dirndx]++;
7045	s->nat_src_node->bytes[dirndx] += pd.tot_len;
7046	}
7047	dirndx = (dir == s->direction) ? 0 : 1;
7048	s->packets[dirndx]++;
7049	s->bytes[dirndx] += pd.tot_len;
7050	}
7051	tr = r;
7052	nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
7053	#ifdef __FreeBSD__
7054	if (nr != NULL && r == &V_pf_default_rule)
7055	#else
7056	if (nr != NULL && r == &pf_default_rule)
7057	#endif
7058	tr = nr;
7059	if (tr->src.addr.type == PF_ADDR_TABLE)
7060	pfr_update_stats(tr->src.addr.p.tbl,
7061	(s == NULL) ? pd.src :
7062	&s->key[(s->direction == PF_IN)]->
7063	addr[(s->direction == PF_OUT)],
7064	pd.af, pd.tot_len, dir == PF_OUT,
7065	r->action == PF_PASS, tr->src.neg);
7066	if (tr->dst.addr.type == PF_ADDR_TABLE)
7067	pfr_update_stats(tr->dst.addr.p.tbl,
7068	(s == NULL) ? pd.dst :
7069	&s->key[(s->direction == PF_IN)]->
7070	addr[(s->direction == PF_IN)],
7071	pd.af, pd.tot_len, dir == PF_OUT,
7072	r->action == PF_PASS, tr->dst.neg);
7073	}
7074
7075	switch (action) {
7076	case PF_SYNPROXY_DROP:
7077	m_freem(*m0);
7078	case PF_DEFER:
7079	*m0 = NULL;
7080	action = PF_PASS;
7081	break;
7082	default:
7083	/* pf_route can free the mbuf causing m0 to become NULL /
7084	if (r->rt)
7085	pf_route(m0, r, dir, kif->pfik_ifp, s, &pd);
7086	break;
7087	}
7088	#ifdef __FreeBSD__
7089	PF_UNLOCK();
7090	#endif
7091	return (action);
7092	}
7093	#endif /* INET */
7094
7095	#ifdef INET6
7096	int
7097	#ifdef __FreeBSD__
7098	pf_test6(int dir, struct ifnet ifp, struct mbuf *m0,
7099	struct ether_header eh, struct inpcb inp)
7100	#else
7101	pf_test6(int dir, struct ifnet ifp, struct mbuf *m0,
7102	struct ether_header *eh)
7103	#endif
7104	{
7105	struct pfi_kif *kif;
7106	u_short action, reason = 0, log = 0;
7107	struct mbuf m = m0, *n = NULL;
7108	#ifdef __FreeBSD__
7109	struct ip6_hdr *h = NULL;
7110	struct pf_rule a = NULL, r = &V_pf_default_rule, tr, nr;
7111	#else
7112	struct ip6_hdr *h;
7113	struct pf_rule a = NULL, r = &pf_default_rule, tr, nr;
7114	#endif
7115	struct pf_state *s = NULL;
7116	struct pf_ruleset *ruleset = NULL;
7117	struct pf_pdesc pd;
7118	int off, terminal = 0, dirndx, rh_cnt = 0;
7119
7120	#ifdef __FreeBSD__
7121	PF_LOCK();
7122	if (!V_pf_status.running) {
7123	PF_UNLOCK();
7124	return (PF_PASS);
7125	}
7126	#else
7127	if (!pf_status.running)
7128	return (PF_PASS);
7129	#endif
7130
7131	memset(&pd, 0, sizeof(pd));
7132	#ifdef __FreeBSD__
7133	if ((pd.pf_mtag = pf_get_mtag(m)) == NULL) {
7134	PF_UNLOCK();
7135	DPFPRINTF(PF_DEBUG_URGENT,
7136	("pf_test: pf_get_mtag returned NULL\n"));
7137	return (PF_DROP);
7138	}
7139	#endif
7140	#ifndef __FreeBSD__
7141	if (ifp->if_type == IFT_CARP && ifp->if_carpdev)
7142	kif = (struct pfi_kif *)ifp->if_carpdev->if_pf_kif;
7143	else
7144	#endif
7145	kif = (struct pfi_kif *)ifp->if_pf_kif;
7146
7147	if (kif == NULL) {
7148	#ifdef __FreeBSD__
7149	PF_UNLOCK();
7150	#endif
7151	DPFPRINTF(PF_DEBUG_URGENT,
7152	("pf_test6: kif == NULL, if_xname %s\n", ifp->if_xname));
7153	return (PF_DROP);
7154	}
7155	if (kif->pfik_flags & PFI_IFLAG_SKIP)
7156	#ifdef __FreeBSD__
7157	{
7158	PF_UNLOCK();
7159	#endif
7160	return (PF_PASS);
7161	#ifdef __FreeBSD__
7162	}
7163	#endif
7164
7165	#ifdef __FreeBSD__
7166	M_ASSERTPKTHDR(m);
7167	#else
7168	#ifdef DIAGNOSTIC
7169	if ((m->m_flags & M_PKTHDR) == 0)
7170	panic("non-M_PKTHDR is passed to pf_test6");
7171	#endif /* DIAGNOSTIC */
7172	#endif
7173
7174	if (m->m_pkthdr.len < (int)sizeof(*h)) {
7175	action = PF_DROP;
7176	REASON_SET(&reason, PFRES_SHORT);
7177	log = 1;
7178	goto done;
7179	}
7180
7181	#ifdef __FreeBSD__
7182	if (pd.pf_mtag->flags & PF_TAG_GENERATED) {
7183	PF_UNLOCK();
7184	#else
7185	if (m->m_pkthdr.pf.flags & PF_TAG_GENERATED)
7186	#endif
7187	return (PF_PASS);
7188	#ifdef __FreeBSD__
7189	}
7190	#endif
7191
7192	/* We do IP header normalization and packet reassembly here */
7193	if (pf_normalize_ip6(m0, dir, kif, &reason, &pd) != PF_PASS) {
7194	action = PF_DROP;
7195	goto done;
7196	}
7197	m = m0; / pf_normalize messes with m0 */
7198	h = mtod(m, struct ip6_hdr *);
7199
7200	#if 1
7201	/*
7202	* we do not support jumbogram yet. if we keep going, zero ip6_plen
7203	* will do something bad, so drop the packet for now.
7204	*/
7205	if (htons(h->ip6_plen) == 0) {
7206	action = PF_DROP;
7207	REASON_SET(&reason, PFRES_NORM); /XXX/
7208	goto done;
7209	}
7210	#endif
7211
7212	pd.src = (struct pf_addr *)&h->ip6_src;
7213	pd.dst = (struct pf_addr *)&h->ip6_dst;
7214	pd.sport = pd.dport = NULL;
7215	pd.ip_sum = NULL;
7216	pd.proto_sum = NULL;
7217	pd.dir = dir;
7218	pd.sidx = (dir == PF_IN) ? 0 : 1;
7219	pd.didx = (dir == PF_IN) ? 1 : 0;
7220	pd.af = AF_INET6;
7221	pd.tos = 0;
7222	pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr);
7223	pd.eh = eh;
7224
7225	off = ((caddr_t)h - m->m_data) + sizeof(struct ip6_hdr);
7226	pd.proto = h->ip6_nxt;
7227	do {
7228	switch (pd.proto) {
7229	case IPPROTO_FRAGMENT:
7230	action = pf_test_fragment(&r, dir, kif, m, h,
7231	&pd, &a, &ruleset);
7232	if (action == PF_DROP)
7233	REASON_SET(&reason, PFRES_FRAG);
7234	goto done;
7235	case IPPROTO_ROUTING: {
7236	struct ip6_rthdr rthdr;
7237
7238	if (rh_cnt++) {
7239	DPFPRINTF(PF_DEBUG_MISC,
7240	("pf: IPv6 more than one rthdr\n"));
7241	action = PF_DROP;
7242	REASON_SET(&reason, PFRES_IPOPTIONS);
7243	log = 1;
7244	goto done;
7245	}
7246	if (!pf_pull_hdr(m, off, &rthdr, sizeof(rthdr), NULL,
7247	&reason, pd.af)) {
7248	DPFPRINTF(PF_DEBUG_MISC,
7249	("pf: IPv6 short rthdr\n"));
7250	action = PF_DROP;
7251	REASON_SET(&reason, PFRES_SHORT);
7252	log = 1;
7253	goto done;
7254	}
7255	if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) {
7256	DPFPRINTF(PF_DEBUG_MISC,
7257	("pf: IPv6 rthdr0\n"));
7258	action = PF_DROP;
7259	REASON_SET(&reason, PFRES_IPOPTIONS);
7260	log = 1;
7261	goto done;
7262	}
7263	/* FALLTHROUGH */
7264	}
7265	case IPPROTO_AH:
7266	case IPPROTO_HOPOPTS:
7267	case IPPROTO_DSTOPTS: {
7268	/* get next header and header length */
7269	struct ip6_ext opt6;
7270
7271	if (!pf_pull_hdr(m, off, &opt6, sizeof(opt6),
7272	NULL, &reason, pd.af)) {
7273	DPFPRINTF(PF_DEBUG_MISC,
7274	("pf: IPv6 short opt\n"));
7275	action = PF_DROP;
7276	log = 1;
7277	goto done;
7278	}
7279	if (pd.proto == IPPROTO_AH)
7280	off += (opt6.ip6e_len + 2) * 4;
7281	else
7282	off += (opt6.ip6e_len + 1) * 8;
7283	pd.proto = opt6.ip6e_nxt;
7284	/* goto the next header */
7285	break;
7286	}
7287	default:
7288	terminal++;
7289	break;
7290	}
7291	} while (!terminal);
7292
7293	/* if there's no routing header, use unmodified mbuf for checksumming */
7294	if (!n)
7295	n = m;
7296
7297	switch (pd.proto) {
7298
7299	case IPPROTO_TCP: {
7300	struct tcphdr th;
7301
7302	pd.hdr.tcp = &th;
7303	if (!pf_pull_hdr(m, off, &th, sizeof(th),
7304	&action, &reason, AF_INET6)) {
7305	log = action != PF_PASS;
7306	goto done;
7307	}
7308	pd.p_len = pd.tot_len - off - (th.th_off << 2);
7309	action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
7310	if (action == PF_DROP)
7311	goto done;
7312	action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
7313	&reason);
7314	if (action == PF_PASS) {
7315	#if NPFSYNC > 0
7316	#ifdef __FreeBSD__
7317	if (pfsync_update_state_ptr != NULL)
7318	pfsync_update_state_ptr(s);
7319	#else
7320	pfsync_update_state(s);
7321	#endif
7322	#endif /* NPFSYNC */
7323	r = s->rule.ptr;
7324	a = s->anchor.ptr;
7325	log = s->log;
7326	} else if (s == NULL)
7327	#ifdef __FreeBSD__
7328	action = pf_test_rule(&r, &s, dir, kif,
7329	m, off, h, &pd, &a, &ruleset, NULL, inp);
7330	#else
7331	action = pf_test_rule(&r, &s, dir, kif,
7332	m, off, h, &pd, &a, &ruleset, &ip6intrq);
7333	#endif
7334	break;
7335	}
7336
7337	case IPPROTO_UDP: {
7338	struct udphdr uh;
7339
7340	pd.hdr.udp = &uh;
7341	if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
7342	&action, &reason, AF_INET6)) {
7343	log = action != PF_PASS;
7344	goto done;
7345	}
7346	if (uh.uh_dport == 0 \|\|
7347	ntohs(uh.uh_ulen) > m->m_pkthdr.len - off \|\|
7348	ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
7349	action = PF_DROP;
7350	REASON_SET(&reason, PFRES_SHORT);
7351	goto done;
7352	}
7353	action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
7354	if (action == PF_PASS) {
7355	#if NPFSYNC > 0
7356	#ifdef __FreeBSD__
7357	if (pfsync_update_state_ptr != NULL)
7358	pfsync_update_state_ptr(s);
7359	#else
7360	pfsync_update_state(s);
7361	#endif
7362	#endif /* NPFSYNC */
7363	r = s->rule.ptr;
7364	a = s->anchor.ptr;
7365	log = s->log;
7366	} else if (s == NULL)
7367	#ifdef __FreeBSD__
7368	action = pf_test_rule(&r, &s, dir, kif,
7369	m, off, h, &pd, &a, &ruleset, NULL, inp);
7370	#else
7371	action = pf_test_rule(&r, &s, dir, kif,
7372	m, off, h, &pd, &a, &ruleset, &ip6intrq);
7373	#endif
7374	break;
7375	}
7376
7377	case IPPROTO_ICMP: {
7378	action = PF_DROP;
7379	DPFPRINTF(PF_DEBUG_MISC,
7380	("pf: dropping IPv6 packet with ICMPv4 payload\n"));
7381	goto done;
7382	}
7383
7384	case IPPROTO_ICMPV6: {
7385	struct icmp6_hdr ih;
7386
7387	pd.hdr.icmp6 = &ih;
7388	if (!pf_pull_hdr(m, off, &ih, sizeof(ih),
7389	&action, &reason, AF_INET6)) {
7390	log = action != PF_PASS;
7391	goto done;
7392	}
7393	action = pf_test_state_icmp(&s, dir, kif,
7394	m, off, h, &pd, &reason);
7395	if (action == PF_PASS) {
7396	#if NPFSYNC > 0
7397	#ifdef __FreeBSD__
7398	if (pfsync_update_state_ptr != NULL)
7399	pfsync_update_state_ptr(s);
7400	#else
7401	pfsync_update_state(s);
7402	#endif
7403	#endif /* NPFSYNC */
7404	r = s->rule.ptr;
7405	a = s->anchor.ptr;
7406	log = s->log;
7407	} else if (s == NULL)
7408	#ifdef __FreeBSD__
7409	action = pf_test_rule(&r, &s, dir, kif,
7410	m, off, h, &pd, &a, &ruleset, NULL, inp);
7411	#else
7412	action = pf_test_rule(&r, &s, dir, kif,
7413	m, off, h, &pd, &a, &ruleset, &ip6intrq);
7414	#endif
7415	break;
7416	}
7417
7418	default:
7419	action = pf_test_state_other(&s, dir, kif, m, &pd);
7420	if (action == PF_PASS) {
7421	#if NPFSYNC > 0
7422	#ifdef __FreeBSD__
7423	if (pfsync_update_state_ptr != NULL)
7424	pfsync_update_state_ptr(s);
7425	#else
7426	pfsync_update_state(s);
7427	#endif
7428	#endif /* NPFSYNC */
7429	r = s->rule.ptr;
7430	a = s->anchor.ptr;
7431	log = s->log;
7432	} else if (s == NULL)
7433	#ifdef __FreeBSD__
7434	action = pf_test_rule(&r, &s, dir, kif, m, off, h,
7435	&pd, &a, &ruleset, NULL, inp);
7436	#else
7437	action = pf_test_rule(&r, &s, dir, kif, m, off, h,
7438	&pd, &a, &ruleset, &ip6intrq);
7439	#endif
7440	break;
7441	}
7442
7443	done:
7444	if (n != m) {
7445	m_freem(n);
7446	n = NULL;
7447	}
7448
7449	/* handle dangerous IPv6 extension headers. */
7450	if (action == PF_PASS && rh_cnt &&
7451	!((s && s->state_flags & PFSTATE_ALLOWOPTS) \|\| r->allow_opts)) {
7452	action = PF_DROP;
7453	REASON_SET(&reason, PFRES_IPOPTIONS);
7454	log = 1;
7455	DPFPRINTF(PF_DEBUG_MISC,
7456	("pf: dropping packet with dangerous v6 headers\n"));
7457	}
7458
7459	if ((s && s->tag) \|\| r->rtableid >= 0)
7460	#ifdef __FreeBSD__
7461	pf_tag_packet(m, s ? s->tag : 0, r->rtableid, pd.pf_mtag);
7462	#else
7463	pf_tag_packet(m, s ? s->tag : 0, r->rtableid);
7464	#endif
7465
7466	if (dir == PF_IN && s && s->key[PF_SK_STACK])
7467	#ifdef __FreeBSD__
7468	pd.pf_mtag->statekey = s->key[PF_SK_STACK];
7469	#else
7470	m->m_pkthdr.pf.statekey = s->key[PF_SK_STACK];
7471	#endif
7472
7473	#ifdef ALTQ
7474	if (action == PF_PASS && r->qid) {
7475	#ifdef __FreeBSD__
7476	if (pd.tos & IPTOS_LOWDELAY)
7477	pd.pf_mtag->qid = r->pqid;
7478	else
7479	pd.pf_mtag->qid = r->qid;
7480	/* add hints for ecn */
7481	pd.pf_mtag->hdr = h;
7482	#else
7483	if (pd.tos & IPTOS_LOWDELAY)
7484	m->m_pkthdr.pf.qid = r->pqid;
7485	else
7486	m->m_pkthdr.pf.qid = r->qid;
7487	/* add hints for ecn */
7488	m->m_pkthdr.pf.hdr = h;
7489	#endif
7490	}
7491	#endif /* ALTQ */
7492
7493	if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP \|\|
7494	pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
7495	(s->nat_rule.ptr->action == PF_RDR \|\|
7496	s->nat_rule.ptr->action == PF_BINAT) &&
7497	IN6_IS_ADDR_LOOPBACK(&pd.dst->v6))
7498	#ifdef __FreeBSD__
7499	m->m_flags \|= M_SKIP_FIREWALL;
7500	#else
7501	m->m_pkthdr.pf.flags \|= PF_TAG_TRANSLATE_LOCALHOST;
7502	#endif
7503
7504	#ifdef __FreeBSD__
7505	/* XXX: Anybody working on it?! */
7506	if (r->divert.port)
7507	printf("pf: divert(9) is not supported for IPv6\n");
7508	#else
7509	if (dir == PF_IN && action == PF_PASS && r->divert.port) {
7510	struct pf_divert *divert;
7511
7512	if ((divert = pf_get_divert(m))) {
7513	m->m_pkthdr.pf.flags \|= PF_TAG_DIVERTED;
7514	divert->port = r->divert.port;
7515	divert->addr.ipv6 = r->divert.addr.v6;
7516	}
7517	}
7518	#endif
7519
7520	if (log) {
7521	struct pf_rule *lr;
7522
7523	if (s != NULL && s->nat_rule.ptr != NULL &&
7524	s->nat_rule.ptr->log & PF_LOG_ALL)
7525	lr = s->nat_rule.ptr;
7526	else
7527	lr = r;
7528	PFLOG_PACKET(kif, h, m, AF_INET6, dir, reason, lr, a, ruleset,
7529	&pd);
7530	}
7531
7532	kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
7533	kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++;
7534
7535	if (action == PF_PASS \|\| r->action == PF_DROP) {
7536	dirndx = (dir == PF_OUT);
7537	r->packets[dirndx]++;
7538	r->bytes[dirndx] += pd.tot_len;
7539	if (a != NULL) {
7540	a->packets[dirndx]++;
7541	a->bytes[dirndx] += pd.tot_len;
7542	}
7543	if (s != NULL) {
7544	if (s->nat_rule.ptr != NULL) {
7545	s->nat_rule.ptr->packets[dirndx]++;
7546	s->nat_rule.ptr->bytes[dirndx] += pd.tot_len;
7547	}
7548	if (s->src_node != NULL) {
7549	s->src_node->packets[dirndx]++;
7550	s->src_node->bytes[dirndx] += pd.tot_len;
7551	}
7552	if (s->nat_src_node != NULL) {
7553	s->nat_src_node->packets[dirndx]++;
7554	s->nat_src_node->bytes[dirndx] += pd.tot_len;
7555	}
7556	dirndx = (dir == s->direction) ? 0 : 1;
7557	s->packets[dirndx]++;
7558	s->bytes[dirndx] += pd.tot_len;
7559	}
7560	tr = r;
7561	nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
7562	#ifdef __FreeBSD__
7563	if (nr != NULL && r == &V_pf_default_rule)
7564	#else
7565	if (nr != NULL && r == &pf_default_rule)
7566	#endif
7567	tr = nr;
7568	if (tr->src.addr.type == PF_ADDR_TABLE)
7569	pfr_update_stats(tr->src.addr.p.tbl,
7570	(s == NULL) ? pd.src :
7571	&s->key[(s->direction == PF_IN)]->addr[0],
7572	pd.af, pd.tot_len, dir == PF_OUT,
7573	r->action == PF_PASS, tr->src.neg);
7574	if (tr->dst.addr.type == PF_ADDR_TABLE)
7575	pfr_update_stats(tr->dst.addr.p.tbl,
7576	(s == NULL) ? pd.dst :
7577	&s->key[(s->direction == PF_IN)]->addr[1],
7578	pd.af, pd.tot_len, dir == PF_OUT,
7579	r->action == PF_PASS, tr->dst.neg);
7580	}
7581
7582	switch (action) {
7583	case PF_SYNPROXY_DROP:
7584	m_freem(*m0);
7585	case PF_DEFER:
7586	*m0 = NULL;
7587	action = PF_PASS;
7588	break;
7589	default:
7590	/* pf_route6 can free the mbuf causing m0 to become NULL /
7591	if (r->rt)
7592	pf_route6(m0, r, dir, kif->pfik_ifp, s, &pd);
7593	break;
7594	}
7595
7596	#ifdef __FreeBSD__
7597	PF_UNLOCK();
7598	#endif
7599	return (action);
7600	}
7601	#endif /* INET6 */
7602
7603	int
7604	pf_check_congestion(struct ifqueue *ifq)
7605	{
7606	#ifdef __FreeBSD__
7607	/* XXX_IMPORT: later */
7608	return (0);
7609	#else
7610	if (ifq->ifq_congestion)
7611	return (1);
7612	else
7613	return (0);
7614	#endif
7615	}
7616
7617	/*
7618	* must be called whenever any addressing information such as
7619	* address, port, protocol has changed
7620	*/
7621	void
7622	pf_pkt_addr_changed(struct mbuf *m)
7623	{
7624	#ifdef __FreeBSD__
7625	struct pf_mtag *pf_tag;
7626
7627	if ((pf_tag = pf_find_mtag(m)) != NULL)
7628	pf_tag->statekey = NULL;
7629	#else
7630	m->m_pkthdr.pf.statekey = NULL;
7631	#endif
7632	}

Note: See TracBrowser for help on using the repository browser.

Download in other formats: