1 /* $OpenBSD: pf.c,v 1.552.2.1 2007/11/27 16:37:57 henning Exp $ */
2 /* add $OpenBSD: pf.c,v 1.553 2007/08/23 11:15:49 dhartmei Exp $ */
3 /* add $OpenBSD: pf.c,v 1.554 2007/08/28 16:09:12 henning Exp $ */
7 * Copyright (c) 2004 The DragonFly Project. All rights reserved.
9 * Copyright (c) 2001 Daniel Hartmeier
10 * Copyright (c) 2002,2003 Henning Brauer
11 * All rights reserved.
13 * Redistribution and use in source and binary forms, with or without
14 * modification, are permitted provided that the following conditions
17 * - Redistributions of source code must retain the above copyright
18 * notice, this list of conditions and the following disclaimer.
19 * - Redistributions in binary form must reproduce the above
20 * copyright notice, this list of conditions and the following
21 * disclaimer in the documentation and/or other materials provided
22 * with the distribution.
24 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
25 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
26 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
27 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
28 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
29 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
30 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
31 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
32 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
34 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
35 * POSSIBILITY OF SUCH DAMAGE.
37 * Effort sponsored in part by the Defense Advanced Research Projects
38 * Agency (DARPA) and Air Force Research Laboratory, Air Force
39 * Materiel Command, USAF, under agreement number F30602-01-2-0537.
44 #include "opt_inet6.h"
45 #include "use_pfsync.h"
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/malloc.h>
51 #include <sys/filio.h>
52 #include <sys/socket.h>
53 #include <sys/socketvar.h>
54 #include <sys/kernel.h>
56 #include <sys/sysctl.h>
57 #include <sys/endian.h>
58 #include <vm/vm_zone.h>
60 #include <sys/kthread.h>
62 #include <sys/mplock2.h>
64 #include <machine/inttypes.h>
67 #include <net/if_types.h>
69 #include <net/netisr.h>
70 #include <net/route.h>
72 #include <netinet/in.h>
73 #include <netinet/in_var.h>
74 #include <netinet/in_systm.h>
75 #include <netinet/ip.h>
76 #include <netinet/ip_var.h>
77 #include <netinet/tcp.h>
78 #include <netinet/tcp_seq.h>
79 #include <netinet/udp.h>
80 #include <netinet/ip_icmp.h>
81 #include <netinet/in_pcb.h>
82 #include <netinet/tcp_timer.h>
83 #include <netinet/tcp_var.h>
84 #include <netinet/udp_var.h>
85 #include <netinet/icmp_var.h>
86 #include <netinet/if_ether.h>
88 #include <net/pf/pfvar.h>
89 #include <net/pf/if_pflog.h>
92 #include <net/pf/if_pfsync.h>
93 #endif /* NPFSYNC > 0 */
96 #include <netinet/ip6.h>
97 #include <netinet/in_pcb.h>
98 #include <netinet/icmp6.h>
99 #include <netinet6/nd6.h>
100 #include <netinet6/ip6_var.h>
101 #include <netinet6/in6_pcb.h>
104 #include <sys/in_cksum.h>
105 #include <sys/ucred.h>
106 #include <machine/limits.h>
107 #include <sys/msgport2.h>
108 #include <net/netmsg2.h>
110 extern int ip_optcopy(struct ip *, struct ip *);
111 extern int debug_pfugidhack;
113 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) kprintf x
120 struct pf_state_tree_lan_ext pf_statetbl_lan_ext;
121 struct pf_state_tree_ext_gwy pf_statetbl_ext_gwy;
123 struct pf_altqqueue pf_altqs[2];
124 struct pf_palist pf_pabuf;
125 struct pf_altqqueue *pf_altqs_active;
126 struct pf_altqqueue *pf_altqs_inactive;
127 struct pf_status pf_status;
129 u_int32_t ticket_altqs_active;
130 u_int32_t ticket_altqs_inactive;
131 int altqs_inactive_open;
132 u_int32_t ticket_pabuf;
134 struct pf_anchor_stackframe {
135 struct pf_ruleset *rs;
137 struct pf_anchor_node *parent;
138 struct pf_anchor *child;
139 } pf_anchor_stack[64];
141 vm_zone_t pf_src_tree_pl, pf_rule_pl, pf_pooladdr_pl;
142 vm_zone_t pf_state_pl, pf_state_key_pl;
143 vm_zone_t pf_altq_pl;
145 void pf_print_host(struct pf_addr *, u_int16_t, u_int8_t);
147 void pf_init_threshold(struct pf_threshold *, u_int32_t,
149 void pf_add_threshold(struct pf_threshold *);
150 int pf_check_threshold(struct pf_threshold *);
152 void pf_change_ap(struct pf_addr *, u_int16_t *,
153 u_int16_t *, u_int16_t *, struct pf_addr *,
154 u_int16_t, u_int8_t, sa_family_t);
155 int pf_modulate_sack(struct mbuf *, int, struct pf_pdesc *,
156 struct tcphdr *, struct pf_state_peer *);
158 void pf_change_a6(struct pf_addr *, u_int16_t *,
159 struct pf_addr *, u_int8_t);
161 void pf_change_icmp(struct pf_addr *, u_int16_t *,
162 struct pf_addr *, struct pf_addr *, u_int16_t,
163 u_int16_t *, u_int16_t *, u_int16_t *,
164 u_int16_t *, u_int8_t, sa_family_t);
165 void pf_send_tcp(const struct pf_rule *, sa_family_t,
166 const struct pf_addr *, const struct pf_addr *,
167 u_int16_t, u_int16_t, u_int32_t, u_int32_t,
168 u_int8_t, u_int16_t, u_int16_t, u_int8_t, int,
169 u_int16_t, struct ether_header *, struct ifnet *);
170 void pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t,
171 sa_family_t, struct pf_rule *);
172 struct pf_rule *pf_match_translation(struct pf_pdesc *, struct mbuf *,
173 int, int, struct pfi_kif *,
174 struct pf_addr *, u_int16_t, struct pf_addr *,
176 struct pf_rule *pf_get_translation(struct pf_pdesc *, struct mbuf *,
177 int, int, struct pfi_kif *, struct pf_src_node **,
178 struct pf_addr *, u_int16_t,
179 struct pf_addr *, u_int16_t,
180 struct pf_addr *, u_int16_t *);
181 void pf_attach_state(struct pf_state_key *,
182 struct pf_state *, int);
183 void pf_detach_state(struct pf_state *, int);
184 int pf_test_rule(struct pf_rule **, struct pf_state **,
185 int, struct pfi_kif *, struct mbuf *, int,
186 void *, struct pf_pdesc *, struct pf_rule **,
187 struct pf_ruleset **, struct ifqueue *, struct inpcb *);
188 int pf_test_fragment(struct pf_rule **, int,
189 struct pfi_kif *, struct mbuf *, void *,
190 struct pf_pdesc *, struct pf_rule **,
191 struct pf_ruleset **);
192 int pf_test_state_tcp(struct pf_state **, int,
193 struct pfi_kif *, struct mbuf *, int,
194 void *, struct pf_pdesc *, u_short *);
195 int pf_test_state_udp(struct pf_state **, int,
196 struct pfi_kif *, struct mbuf *, int,
197 void *, struct pf_pdesc *);
198 int pf_test_state_icmp(struct pf_state **, int,
199 struct pfi_kif *, struct mbuf *, int,
200 void *, struct pf_pdesc *, u_short *);
201 int pf_test_state_other(struct pf_state **, int,
202 struct pfi_kif *, struct pf_pdesc *);
203 int pf_match_tag(struct mbuf *, struct pf_rule *, int *);
204 void pf_step_into_anchor(int *, struct pf_ruleset **, int,
205 struct pf_rule **, struct pf_rule **, int *);
206 int pf_step_out_of_anchor(int *, struct pf_ruleset **,
207 int, struct pf_rule **, struct pf_rule **,
209 void pf_hash(struct pf_addr *, struct pf_addr *,
210 struct pf_poolhashkey *, sa_family_t);
211 int pf_map_addr(u_int8_t, struct pf_rule *,
212 struct pf_addr *, struct pf_addr *,
213 struct pf_addr *, struct pf_src_node **);
214 int pf_get_sport(sa_family_t, u_int8_t, struct pf_rule *,
215 struct pf_addr *, struct pf_addr *, u_int16_t,
216 struct pf_addr *, u_int16_t*, u_int16_t, u_int16_t,
217 struct pf_src_node **);
218 void pf_route(struct mbuf **, struct pf_rule *, int,
219 struct ifnet *, struct pf_state *,
221 void pf_route6(struct mbuf **, struct pf_rule *, int,
222 struct ifnet *, struct pf_state *,
224 u_int8_t pf_get_wscale(struct mbuf *, int, u_int16_t,
226 u_int16_t pf_get_mss(struct mbuf *, int, u_int16_t,
228 u_int16_t pf_calc_mss(struct pf_addr *, sa_family_t,
230 void pf_set_rt_ifp(struct pf_state *,
232 int pf_check_proto_cksum(struct mbuf *, int, int,
233 u_int8_t, sa_family_t);
234 int pf_addr_wrap_neq(struct pf_addr_wrap *,
235 struct pf_addr_wrap *);
236 struct pf_state *pf_find_state(struct pfi_kif *,
237 struct pf_state_key_cmp *, u_int8_t);
238 int pf_src_connlimit(struct pf_state **);
239 void pf_stateins_err(const char *, struct pf_state *,
241 int pf_check_congestion(struct ifqueue *);
243 extern int pf_end_threads;
245 struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] = {
246 { &pf_state_pl, PFSTATE_HIWAT },
247 { &pf_src_tree_pl, PFSNODE_HIWAT },
248 { &pf_frent_pl, PFFRAG_FRENT_HIWAT },
249 { &pfr_ktable_pl, PFR_KTABLE_HIWAT },
250 { &pfr_kentry_pl, PFR_KENTRY_HIWAT }
253 #define STATE_LOOKUP() \
255 if (direction == PF_IN) \
256 *state = pf_find_state(kif, &key, PF_EXT_GWY); \
258 *state = pf_find_state(kif, &key, PF_LAN_EXT); \
259 if (*state == NULL || (*state)->timeout == PFTM_PURGE) \
261 if (direction == PF_OUT && \
262 (((*state)->rule.ptr->rt == PF_ROUTETO && \
263 (*state)->rule.ptr->direction == PF_OUT) || \
264 ((*state)->rule.ptr->rt == PF_REPLYTO && \
265 (*state)->rule.ptr->direction == PF_IN)) && \
266 (*state)->rt_kif != NULL && \
267 (*state)->rt_kif != kif) \
271 #define STATE_TRANSLATE(sk) \
272 (sk)->lan.addr.addr32[0] != (sk)->gwy.addr.addr32[0] || \
273 ((sk)->af == AF_INET6 && \
274 ((sk)->lan.addr.addr32[1] != (sk)->gwy.addr.addr32[1] || \
275 (sk)->lan.addr.addr32[2] != (sk)->gwy.addr.addr32[2] || \
276 (sk)->lan.addr.addr32[3] != (sk)->gwy.addr.addr32[3])) || \
277 (sk)->lan.port != (sk)->gwy.port
279 #define BOUND_IFACE(r, k) \
280 ((r)->rule_flag & PFRULE_IFBOUND) ? (k) : pfi_all
282 #define STATE_INC_COUNTERS(s) \
284 s->rule.ptr->states++; \
285 if (s->anchor.ptr != NULL) \
286 s->anchor.ptr->states++; \
287 if (s->nat_rule.ptr != NULL) \
288 s->nat_rule.ptr->states++; \
291 #define STATE_DEC_COUNTERS(s) \
293 if (s->nat_rule.ptr != NULL) \
294 s->nat_rule.ptr->states--; \
295 if (s->anchor.ptr != NULL) \
296 s->anchor.ptr->states--; \
297 s->rule.ptr->states--; \
300 static __inline int pf_src_compare(struct pf_src_node *, struct pf_src_node *);
301 static __inline int pf_state_compare_lan_ext(struct pf_state_key *,
302 struct pf_state_key *);
303 static __inline int pf_state_compare_ext_gwy(struct pf_state_key *,
304 struct pf_state_key *);
305 static __inline int pf_state_compare_id(struct pf_state *,
308 struct pf_src_tree tree_src_tracking;
310 struct pf_state_tree_id tree_id;
311 struct pf_state_queue state_list;
313 RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare);
314 RB_GENERATE(pf_state_tree_lan_ext, pf_state_key,
315 entry_lan_ext, pf_state_compare_lan_ext);
316 RB_GENERATE(pf_state_tree_ext_gwy, pf_state_key,
317 entry_ext_gwy, pf_state_compare_ext_gwy);
318 RB_GENERATE(pf_state_tree_id, pf_state,
319 entry_id, pf_state_compare_id);
321 #define PF_DT_SKIP_LANEXT 0x01
322 #define PF_DT_SKIP_EXTGWY 0x02
325 pf_src_compare(struct pf_src_node *a, struct pf_src_node *b)
329 if (a->rule.ptr > b->rule.ptr)
331 if (a->rule.ptr < b->rule.ptr)
333 if ((diff = a->af - b->af) != 0)
338 if (a->addr.addr32[0] > b->addr.addr32[0])
340 if (a->addr.addr32[0] < b->addr.addr32[0])
346 if (a->addr.addr32[3] > b->addr.addr32[3])
348 if (a->addr.addr32[3] < b->addr.addr32[3])
350 if (a->addr.addr32[2] > b->addr.addr32[2])
352 if (a->addr.addr32[2] < b->addr.addr32[2])
354 if (a->addr.addr32[1] > b->addr.addr32[1])
356 if (a->addr.addr32[1] < b->addr.addr32[1])
358 if (a->addr.addr32[0] > b->addr.addr32[0])
360 if (a->addr.addr32[0] < b->addr.addr32[0])
369 pf_state_hash(struct pf_state_key *sk)
371 u_int32_t hv = (intptr_t)sk / sizeof(*sk);
373 hv ^= crc32(&sk->lan, sizeof(sk->lan));
374 hv ^= crc32(&sk->gwy, sizeof(sk->gwy));
375 hv ^= crc32(&sk->ext, sizeof(sk->ext));
376 if (hv == 0) /* disallow 0 */
382 pf_state_compare_lan_ext(struct pf_state_key *a, struct pf_state_key *b)
386 if ((diff = a->proto - b->proto) != 0)
388 if ((diff = a->af - b->af) != 0)
393 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
395 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
397 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
399 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
405 if (a->lan.addr.addr32[3] > b->lan.addr.addr32[3])
407 if (a->lan.addr.addr32[3] < b->lan.addr.addr32[3])
409 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
411 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
413 if (a->lan.addr.addr32[2] > b->lan.addr.addr32[2])
415 if (a->lan.addr.addr32[2] < b->lan.addr.addr32[2])
417 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
419 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
421 if (a->lan.addr.addr32[1] > b->lan.addr.addr32[1])
423 if (a->lan.addr.addr32[1] < b->lan.addr.addr32[1])
425 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
427 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
429 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
431 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
433 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
435 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
441 if ((diff = a->lan.port - b->lan.port) != 0)
443 if ((diff = a->ext.port - b->ext.port) != 0)
450 pf_state_compare_ext_gwy(struct pf_state_key *a, struct pf_state_key *b)
454 if ((diff = a->proto - b->proto) != 0)
456 if ((diff = a->af - b->af) != 0)
461 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
463 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
465 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
467 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
473 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
475 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
477 if (a->gwy.addr.addr32[3] > b->gwy.addr.addr32[3])
479 if (a->gwy.addr.addr32[3] < b->gwy.addr.addr32[3])
481 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
483 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
485 if (a->gwy.addr.addr32[2] > b->gwy.addr.addr32[2])
487 if (a->gwy.addr.addr32[2] < b->gwy.addr.addr32[2])
489 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
491 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
493 if (a->gwy.addr.addr32[1] > b->gwy.addr.addr32[1])
495 if (a->gwy.addr.addr32[1] < b->gwy.addr.addr32[1])
497 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
499 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
501 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
503 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
509 if ((diff = a->ext.port - b->ext.port) != 0)
511 if ((diff = a->gwy.port - b->gwy.port) != 0)
518 pf_state_compare_id(struct pf_state *a, struct pf_state *b)
524 if (a->creatorid > b->creatorid)
526 if (a->creatorid < b->creatorid)
534 pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af)
539 dst->addr32[0] = src->addr32[0];
543 dst->addr32[0] = src->addr32[0];
544 dst->addr32[1] = src->addr32[1];
545 dst->addr32[2] = src->addr32[2];
546 dst->addr32[3] = src->addr32[3];
553 pf_find_state_byid(struct pf_state_cmp *key)
555 pf_status.fcounters[FCNT_STATE_SEARCH]++;
557 return (RB_FIND(pf_state_tree_id, &tree_id, (struct pf_state *)key));
561 pf_find_state(struct pfi_kif *kif, struct pf_state_key_cmp *key, u_int8_t tree)
563 struct pf_state_key *sk;
566 pf_status.fcounters[FCNT_STATE_SEARCH]++;
570 sk = RB_FIND(pf_state_tree_lan_ext, &pf_statetbl_lan_ext,
571 (struct pf_state_key *)key);
574 sk = RB_FIND(pf_state_tree_ext_gwy, &pf_statetbl_ext_gwy,
575 (struct pf_state_key *)key);
578 panic("pf_find_state");
581 /* list is sorted, if-bound states before floating ones */
583 TAILQ_FOREACH(s, &sk->states, next)
584 if (s->kif == pfi_all || s->kif == kif)
591 pf_find_state_all(struct pf_state_key_cmp *key, u_int8_t tree, int *more)
593 struct pf_state_key *sk;
594 struct pf_state *s, *ret = NULL;
596 pf_status.fcounters[FCNT_STATE_SEARCH]++;
600 sk = RB_FIND(pf_state_tree_lan_ext,
601 &pf_statetbl_lan_ext, (struct pf_state_key *)key);
604 sk = RB_FIND(pf_state_tree_ext_gwy,
605 &pf_statetbl_ext_gwy, (struct pf_state_key *)key);
608 panic("pf_find_state_all");
612 ret = TAILQ_FIRST(&sk->states);
616 TAILQ_FOREACH(s, &sk->states, next)
624 pf_init_threshold(struct pf_threshold *threshold,
625 u_int32_t limit, u_int32_t seconds)
627 threshold->limit = limit * PF_THRESHOLD_MULT;
628 threshold->seconds = seconds;
629 threshold->count = 0;
630 threshold->last = time_second;
634 pf_add_threshold(struct pf_threshold *threshold)
636 u_int32_t t = time_second, diff = t - threshold->last;
638 if (diff >= threshold->seconds)
639 threshold->count = 0;
641 threshold->count -= threshold->count * diff /
643 threshold->count += PF_THRESHOLD_MULT;
648 pf_check_threshold(struct pf_threshold *threshold)
650 return (threshold->count > threshold->limit);
654 pf_src_connlimit(struct pf_state **state)
658 (*state)->src_node->conn++;
659 (*state)->src.tcp_est = 1;
660 pf_add_threshold(&(*state)->src_node->conn_rate);
662 if ((*state)->rule.ptr->max_src_conn &&
663 (*state)->rule.ptr->max_src_conn <
664 (*state)->src_node->conn) {
665 pf_status.lcounters[LCNT_SRCCONN]++;
669 if ((*state)->rule.ptr->max_src_conn_rate.limit &&
670 pf_check_threshold(&(*state)->src_node->conn_rate)) {
671 pf_status.lcounters[LCNT_SRCCONNRATE]++;
678 if ((*state)->rule.ptr->overload_tbl) {
680 u_int32_t killed = 0;
682 pf_status.lcounters[LCNT_OVERLOAD_TABLE]++;
683 if (pf_status.debug >= PF_DEBUG_MISC) {
684 kprintf("pf_src_connlimit: blocking address ");
685 pf_print_host(&(*state)->src_node->addr, 0,
686 (*state)->state_key->af);
689 bzero(&p, sizeof(p));
690 p.pfra_af = (*state)->state_key->af;
691 switch ((*state)->state_key->af) {
695 p.pfra_ip4addr = (*state)->src_node->addr.v4;
701 p.pfra_ip6addr = (*state)->src_node->addr.v6;
706 pfr_insert_kentry((*state)->rule.ptr->overload_tbl,
709 /* kill existing states if that's required. */
710 if ((*state)->rule.ptr->flush) {
711 struct pf_state_key *sk;
714 pf_status.lcounters[LCNT_OVERLOAD_FLUSH]++;
715 RB_FOREACH(st, pf_state_tree_id, &tree_id) {
718 * Kill states from this source. (Only those
719 * from the same rule if PF_FLUSH_GLOBAL is not
723 (*state)->state_key->af &&
724 (((*state)->state_key->direction ==
726 PF_AEQ(&(*state)->src_node->addr,
727 &sk->lan.addr, sk->af)) ||
728 ((*state)->state_key->direction == PF_IN &&
729 PF_AEQ(&(*state)->src_node->addr,
730 &sk->ext.addr, sk->af))) &&
731 ((*state)->rule.ptr->flush &
733 (*state)->rule.ptr == st->rule.ptr)) {
734 st->timeout = PFTM_PURGE;
735 st->src.state = st->dst.state =
740 if (pf_status.debug >= PF_DEBUG_MISC)
741 kprintf(", %u states killed", killed);
743 if (pf_status.debug >= PF_DEBUG_MISC)
747 /* kill this state */
748 (*state)->timeout = PFTM_PURGE;
749 (*state)->src.state = (*state)->dst.state = TCPS_CLOSED;
754 pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule,
755 struct pf_addr *src, sa_family_t af)
757 struct pf_src_node k;
761 PF_ACPY(&k.addr, src, af);
762 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
763 rule->rpool.opts & PF_POOL_STICKYADDR)
767 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
768 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
771 if (!rule->max_src_nodes ||
772 rule->src_nodes < rule->max_src_nodes)
773 (*sn) = pool_get(&pf_src_tree_pl, PR_NOWAIT);
775 pf_status.lcounters[LCNT_SRCNODES]++;
778 bzero(*sn, sizeof(struct pf_src_node));
780 pf_init_threshold(&(*sn)->conn_rate,
781 rule->max_src_conn_rate.limit,
782 rule->max_src_conn_rate.seconds);
785 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
786 rule->rpool.opts & PF_POOL_STICKYADDR)
787 (*sn)->rule.ptr = rule;
789 (*sn)->rule.ptr = NULL;
790 PF_ACPY(&(*sn)->addr, src, af);
791 if (RB_INSERT(pf_src_tree,
792 &tree_src_tracking, *sn) != NULL) {
793 if (pf_status.debug >= PF_DEBUG_MISC) {
794 kprintf("pf: src_tree insert failed: ");
795 pf_print_host(&(*sn)->addr, 0, af);
798 pool_put(&pf_src_tree_pl, *sn);
801 (*sn)->creation = time_second;
802 (*sn)->ruletype = rule->action;
803 if ((*sn)->rule.ptr != NULL)
804 (*sn)->rule.ptr->src_nodes++;
805 pf_status.scounters[SCNT_SRC_NODE_INSERT]++;
806 pf_status.src_nodes++;
808 if (rule->max_src_states &&
809 (*sn)->states >= rule->max_src_states) {
810 pf_status.lcounters[LCNT_SRCSTATES]++;
818 pf_stateins_err(const char *tree, struct pf_state *s, struct pfi_kif *kif)
820 struct pf_state_key *sk = s->state_key;
822 if (pf_status.debug >= PF_DEBUG_MISC) {
823 kprintf("pf: state insert failed: %s %s", tree, kif->pfik_name);
825 pf_print_host(&sk->lan.addr, sk->lan.port,
828 pf_print_host(&sk->gwy.addr, sk->gwy.port,
831 pf_print_host(&sk->ext.addr, sk->ext.port,
833 if (s->sync_flags & PFSTATE_FROMSYNC)
834 kprintf(" (from sync)");
840 pf_insert_state(struct pfi_kif *kif, struct pf_state *s)
842 struct pf_state_key *cur;
845 KKASSERT(s->state_key != NULL);
848 if ((cur = RB_INSERT(pf_state_tree_lan_ext, &pf_statetbl_lan_ext,
849 s->state_key)) != NULL) {
850 /* key exists. check for same kif, if none, add to key */
851 TAILQ_FOREACH(sp, &cur->states, next)
852 if (sp->kif == kif) { /* collision! */
853 pf_stateins_err("tree_lan_ext", s, kif);
855 PF_DT_SKIP_LANEXT|PF_DT_SKIP_EXTGWY);
858 pf_detach_state(s, PF_DT_SKIP_LANEXT|PF_DT_SKIP_EXTGWY);
859 pf_attach_state(cur, s, kif == pfi_all ? 1 : 0);
862 /* if cur != NULL, we already found a state key and attached to it */
863 if (cur == NULL && (cur = RB_INSERT(pf_state_tree_ext_gwy,
864 &pf_statetbl_ext_gwy, s->state_key)) != NULL) {
865 /* must not happen. we must have found the sk above! */
866 pf_stateins_err("tree_ext_gwy", s, kif);
867 pf_detach_state(s, PF_DT_SKIP_EXTGWY);
871 if (s->id == 0 && s->creatorid == 0) {
872 s->id = htobe64(pf_status.stateid++);
873 s->creatorid = pf_status.hostid;
875 if (RB_INSERT(pf_state_tree_id, &tree_id, s) != NULL) {
876 if (pf_status.debug >= PF_DEBUG_MISC) {
877 kprintf("pf: state insert failed: "
878 "id: %016llx creatorid: %08x",
879 be64toh(s->id), ntohl(s->creatorid));
880 if (s->sync_flags & PFSTATE_FROMSYNC)
881 kprintf(" (from sync)");
884 pf_detach_state(s, 0);
887 TAILQ_INSERT_TAIL(&state_list, s, entry_list);
888 pf_status.fcounters[FCNT_STATE_INSERT]++;
890 pfi_kif_ref(kif, PFI_KIF_REF_STATE);
892 pfsync_insert_state(s);
898 pf_purge_thread(void *v)
905 tsleep(pf_purge_thread, PWAIT, "pftm", 1 * hz);
907 lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
909 if (pf_end_threads) {
910 pf_purge_expired_states(pf_status.states, 1);
911 pf_purge_expired_fragments();
912 pf_purge_expired_src_nodes(1);
915 lockmgr(&pf_consistency_lock, LK_RELEASE);
916 wakeup(pf_purge_thread);
921 /* process a fraction of the state table every second */
922 if(!pf_purge_expired_states(1 + (pf_status.states
923 / pf_default_rule.timeout[PFTM_INTERVAL]), 0)) {
925 pf_purge_expired_states(1 + (pf_status.states
926 / pf_default_rule.timeout[PFTM_INTERVAL]), 1);
929 /* purge other expired types every PFTM_INTERVAL seconds */
930 if (++nloops >= pf_default_rule.timeout[PFTM_INTERVAL]) {
931 pf_purge_expired_fragments();
932 if (!pf_purge_expired_src_nodes(locked)) {
933 pf_purge_expired_src_nodes(1);
938 lockmgr(&pf_consistency_lock, LK_RELEASE);
944 pf_state_expires(const struct pf_state *state)
951 /* handle all PFTM_* > PFTM_MAX here */
952 if (state->timeout == PFTM_PURGE)
953 return (time_second);
954 if (state->timeout == PFTM_UNTIL_PACKET)
956 KKASSERT(state->timeout != PFTM_UNLINKED);
957 KASSERT((state->timeout < PFTM_MAX),
958 ("pf_state_expires: timeout > PFTM_MAX"));
959 timeout = state->rule.ptr->timeout[state->timeout];
961 timeout = pf_default_rule.timeout[state->timeout];
962 start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START];
964 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END];
965 states = state->rule.ptr->states;
967 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START];
968 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END];
969 states = pf_status.states;
971 if (end && states > start && start < end) {
973 return (state->expire + timeout * (end - states) /
976 return (time_second);
978 return (state->expire + timeout);
982 pf_purge_expired_src_nodes(int waslocked)
984 struct pf_src_node *cur, *next;
985 int locked = waslocked;
987 for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) {
988 next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur);
990 if (cur->states <= 0 && cur->expire <= time_second) {
992 lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
993 next = RB_NEXT(pf_src_tree,
994 &tree_src_tracking, cur);
997 if (cur->rule.ptr != NULL) {
998 cur->rule.ptr->src_nodes--;
999 if (cur->rule.ptr->states <= 0 &&
1000 cur->rule.ptr->max_src_nodes <= 0)
1001 pf_rm_rule(NULL, cur->rule.ptr);
1003 RB_REMOVE(pf_src_tree, &tree_src_tracking, cur);
1004 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
1005 pf_status.src_nodes--;
1006 pool_put(&pf_src_tree_pl, cur);
1010 if (locked && !waslocked)
1011 lockmgr(&pf_consistency_lock, LK_RELEASE);
1016 pf_src_tree_remove_state(struct pf_state *s)
1020 if (s->src_node != NULL) {
1022 --s->src_node->conn;
1023 if (--s->src_node->states <= 0) {
1024 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1027 pf_default_rule.timeout[PFTM_SRC_NODE];
1028 s->src_node->expire = time_second + timeout;
1031 if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) {
1032 if (--s->nat_src_node->states <= 0) {
1033 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1036 pf_default_rule.timeout[PFTM_SRC_NODE];
1037 s->nat_src_node->expire = time_second + timeout;
1040 s->src_node = s->nat_src_node = NULL;
1043 /* callers should be at crit_enter() */
1045 pf_unlink_state(struct pf_state *cur)
1047 if (cur->src.state == PF_TCPS_PROXY_DST) {
1048 pf_send_tcp(cur->rule.ptr, cur->state_key->af,
1049 &cur->state_key->ext.addr, &cur->state_key->lan.addr,
1050 cur->state_key->ext.port, cur->state_key->lan.port,
1051 cur->src.seqhi, cur->src.seqlo + 1,
1052 TH_RST|TH_ACK, 0, 0, 0, 1, cur->tag, NULL, NULL);
1054 RB_REMOVE(pf_state_tree_id, &tree_id, cur);
1056 if (cur->creatorid == pf_status.hostid)
1057 pfsync_delete_state(cur);
1059 cur->timeout = PFTM_UNLINKED;
1060 pf_src_tree_remove_state(cur);
1061 pf_detach_state(cur, 0);
1064 /* callers should be at crit_enter() and hold the
1065 * write_lock on pf_consistency_lock */
1067 pf_free_state(struct pf_state *cur)
1070 if (pfsyncif != NULL &&
1071 (pfsyncif->sc_bulk_send_next == cur ||
1072 pfsyncif->sc_bulk_terminator == cur))
1075 KKASSERT(cur->timeout == PFTM_UNLINKED);
1076 if (--cur->rule.ptr->states <= 0 &&
1077 cur->rule.ptr->src_nodes <= 0)
1078 pf_rm_rule(NULL, cur->rule.ptr);
1079 if (cur->nat_rule.ptr != NULL)
1080 if (--cur->nat_rule.ptr->states <= 0 &&
1081 cur->nat_rule.ptr->src_nodes <= 0)
1082 pf_rm_rule(NULL, cur->nat_rule.ptr);
1083 if (cur->anchor.ptr != NULL)
1084 if (--cur->anchor.ptr->states <= 0)
1085 pf_rm_rule(NULL, cur->anchor.ptr);
1086 pf_normalize_tcp_cleanup(cur);
1087 pfi_kif_unref(cur->kif, PFI_KIF_REF_STATE);
1088 TAILQ_REMOVE(&state_list, cur, entry_list);
1090 pf_tag_unref(cur->tag);
1091 pool_put(&pf_state_pl, cur);
1092 pf_status.fcounters[FCNT_STATE_REMOVALS]++;
1097 pf_purge_expired_states(u_int32_t maxcheck, int waslocked)
1099 static struct pf_state *cur = NULL;
1100 struct pf_state *next;
1101 int locked = waslocked;
1103 while (maxcheck--) {
1104 /* wrap to start of list when we hit the end */
1106 cur = TAILQ_FIRST(&state_list);
1108 break; /* list empty */
1111 /* get next state, as cur may get deleted */
1112 next = TAILQ_NEXT(cur, entry_list);
1114 if (cur->timeout == PFTM_UNLINKED) {
1115 /* free unlinked state */
1117 lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
1121 } else if (pf_state_expires(cur) <= time_second) {
1122 /* unlink and free expired state */
1123 pf_unlink_state(cur);
1125 if (!lockmgr(&pf_consistency_lock, LK_EXCLUSIVE))
1135 lockmgr(&pf_consistency_lock, LK_RELEASE);
1140 pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw)
1142 if (aw->type != PF_ADDR_TABLE)
1144 if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL)
1150 pf_tbladdr_remove(struct pf_addr_wrap *aw)
1152 if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL)
1154 pfr_detach_table(aw->p.tbl);
1159 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
1161 struct pfr_ktable *kt = aw->p.tbl;
1163 if (aw->type != PF_ADDR_TABLE || kt == NULL)
1165 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
1166 kt = kt->pfrkt_root;
1168 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
1173 pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af)
1178 u_int32_t a = ntohl(addr->addr32[0]);
1179 kprintf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255,
1191 u_int8_t i, curstart = 255, curend = 0,
1192 maxstart = 0, maxend = 0;
1193 for (i = 0; i < 8; i++) {
1194 if (!addr->addr16[i]) {
1195 if (curstart == 255)
1201 if ((curend - curstart) >
1202 (maxend - maxstart)) {
1203 maxstart = curstart;
1210 for (i = 0; i < 8; i++) {
1211 if (i >= maxstart && i <= maxend) {
1220 b = ntohs(addr->addr16[i]);
1237 pf_print_state(struct pf_state *s)
1239 struct pf_state_key *sk = s->state_key;
1240 switch (sk->proto) {
1250 case IPPROTO_ICMPV6:
1254 kprintf("%u ", sk->proto);
1257 pf_print_host(&sk->lan.addr, sk->lan.port, sk->af);
1259 pf_print_host(&sk->gwy.addr, sk->gwy.port, sk->af);
1261 pf_print_host(&sk->ext.addr, sk->ext.port, sk->af);
1262 kprintf(" [lo=%u high=%u win=%u modulator=%u", s->src.seqlo,
1263 s->src.seqhi, s->src.max_win, s->src.seqdiff);
1264 if (s->src.wscale && s->dst.wscale)
1265 kprintf(" wscale=%u", s->src.wscale & PF_WSCALE_MASK);
1267 kprintf(" [lo=%u high=%u win=%u modulator=%u", s->dst.seqlo,
1268 s->dst.seqhi, s->dst.max_win, s->dst.seqdiff);
1269 if (s->src.wscale && s->dst.wscale)
1270 kprintf(" wscale=%u", s->dst.wscale & PF_WSCALE_MASK);
1272 kprintf(" %u:%u", s->src.state, s->dst.state);
1276 pf_print_flags(u_int8_t f)
1298 #define PF_SET_SKIP_STEPS(i) \
1300 while (head[i] != cur) { \
1301 head[i]->skip[i].ptr = cur; \
1302 head[i] = TAILQ_NEXT(head[i], entries); \
1307 pf_calc_skip_steps(struct pf_rulequeue *rules)
1309 struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT];
1312 cur = TAILQ_FIRST(rules);
1314 for (i = 0; i < PF_SKIP_COUNT; ++i)
1316 while (cur != NULL) {
1318 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
1319 PF_SET_SKIP_STEPS(PF_SKIP_IFP);
1320 if (cur->direction != prev->direction)
1321 PF_SET_SKIP_STEPS(PF_SKIP_DIR);
1322 if (cur->af != prev->af)
1323 PF_SET_SKIP_STEPS(PF_SKIP_AF);
1324 if (cur->proto != prev->proto)
1325 PF_SET_SKIP_STEPS(PF_SKIP_PROTO);
1326 if (cur->src.neg != prev->src.neg ||
1327 pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr))
1328 PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR);
1329 if (cur->src.port[0] != prev->src.port[0] ||
1330 cur->src.port[1] != prev->src.port[1] ||
1331 cur->src.port_op != prev->src.port_op)
1332 PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
1333 if (cur->dst.neg != prev->dst.neg ||
1334 pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr))
1335 PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR);
1336 if (cur->dst.port[0] != prev->dst.port[0] ||
1337 cur->dst.port[1] != prev->dst.port[1] ||
1338 cur->dst.port_op != prev->dst.port_op)
1339 PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
1342 cur = TAILQ_NEXT(cur, entries);
1344 for (i = 0; i < PF_SKIP_COUNT; ++i)
1345 PF_SET_SKIP_STEPS(i);
1349 pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2)
1351 if (aw1->type != aw2->type)
1353 switch (aw1->type) {
1354 case PF_ADDR_ADDRMASK:
1355 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0))
1357 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0))
1360 case PF_ADDR_DYNIFTL:
1361 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt);
1362 case PF_ADDR_NOROUTE:
1363 case PF_ADDR_URPFFAILED:
1366 return (aw1->p.tbl != aw2->p.tbl);
1367 case PF_ADDR_RTLABEL:
1368 return (aw1->v.rtlabel != aw2->v.rtlabel);
1370 kprintf("invalid address type: %d\n", aw1->type);
1376 pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp)
1382 l = cksum + old - new;
1383 l = (l >> 16) + (l & 65535);
1391 pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc,
1392 struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af)
1397 PF_ACPY(&ao, a, af);
1405 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1406 ao.addr16[0], an->addr16[0], 0),
1407 ao.addr16[1], an->addr16[1], 0);
1409 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1410 ao.addr16[0], an->addr16[0], u),
1411 ao.addr16[1], an->addr16[1], u),
1417 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1418 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1419 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1420 ao.addr16[0], an->addr16[0], u),
1421 ao.addr16[1], an->addr16[1], u),
1422 ao.addr16[2], an->addr16[2], u),
1423 ao.addr16[3], an->addr16[3], u),
1424 ao.addr16[4], an->addr16[4], u),
1425 ao.addr16[5], an->addr16[5], u),
1426 ao.addr16[6], an->addr16[6], u),
1427 ao.addr16[7], an->addr16[7], u),
1435 /* Changes a u_int32_t. Uses a void * so there are no align restrictions */
1437 pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u)
1441 memcpy(&ao, a, sizeof(ao));
1442 memcpy(a, &an, sizeof(u_int32_t));
1443 *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u),
1444 ao % 65536, an % 65536, u);
1449 pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u)
1453 PF_ACPY(&ao, a, AF_INET6);
1454 PF_ACPY(a, an, AF_INET6);
1456 *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1457 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1458 pf_cksum_fixup(pf_cksum_fixup(*c,
1459 ao.addr16[0], an->addr16[0], u),
1460 ao.addr16[1], an->addr16[1], u),
1461 ao.addr16[2], an->addr16[2], u),
1462 ao.addr16[3], an->addr16[3], u),
1463 ao.addr16[4], an->addr16[4], u),
1464 ao.addr16[5], an->addr16[5], u),
1465 ao.addr16[6], an->addr16[6], u),
1466 ao.addr16[7], an->addr16[7], u);
1471 pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa,
1472 struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c,
1473 u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af)
1475 struct pf_addr oia, ooa;
1477 PF_ACPY(&oia, ia, af);
1478 PF_ACPY(&ooa, oa, af);
1480 /* Change inner protocol port, fix inner protocol checksum. */
1482 u_int16_t oip = *ip;
1489 *pc = pf_cksum_fixup(*pc, oip, *ip, u);
1490 *ic = pf_cksum_fixup(*ic, oip, *ip, 0);
1492 *ic = pf_cksum_fixup(*ic, opc, *pc, 0);
1494 /* Change inner ip address, fix inner ip and icmp checksums. */
1495 PF_ACPY(ia, na, af);
1499 u_int32_t oh2c = *h2c;
1501 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c,
1502 oia.addr16[0], ia->addr16[0], 0),
1503 oia.addr16[1], ia->addr16[1], 0);
1504 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1505 oia.addr16[0], ia->addr16[0], 0),
1506 oia.addr16[1], ia->addr16[1], 0);
1507 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0);
1513 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1514 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1515 pf_cksum_fixup(pf_cksum_fixup(*ic,
1516 oia.addr16[0], ia->addr16[0], u),
1517 oia.addr16[1], ia->addr16[1], u),
1518 oia.addr16[2], ia->addr16[2], u),
1519 oia.addr16[3], ia->addr16[3], u),
1520 oia.addr16[4], ia->addr16[4], u),
1521 oia.addr16[5], ia->addr16[5], u),
1522 oia.addr16[6], ia->addr16[6], u),
1523 oia.addr16[7], ia->addr16[7], u);
1527 /* Change outer ip address, fix outer ip or icmpv6 checksum. */
1528 PF_ACPY(oa, na, af);
1532 *hc = pf_cksum_fixup(pf_cksum_fixup(*hc,
1533 ooa.addr16[0], oa->addr16[0], 0),
1534 ooa.addr16[1], oa->addr16[1], 0);
1539 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1540 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1541 pf_cksum_fixup(pf_cksum_fixup(*ic,
1542 ooa.addr16[0], oa->addr16[0], u),
1543 ooa.addr16[1], oa->addr16[1], u),
1544 ooa.addr16[2], oa->addr16[2], u),
1545 ooa.addr16[3], oa->addr16[3], u),
1546 ooa.addr16[4], oa->addr16[4], u),
1547 ooa.addr16[5], oa->addr16[5], u),
1548 ooa.addr16[6], oa->addr16[6], u),
1549 ooa.addr16[7], oa->addr16[7], u);
1557 * Need to modulate the sequence numbers in the TCP SACK option
1558 * (credits to Krzysztof Pfaff for report and patch)
1561 pf_modulate_sack(struct mbuf *m, int off, struct pf_pdesc *pd,
1562 struct tcphdr *th, struct pf_state_peer *dst)
1564 int hlen = (th->th_off << 2) - sizeof(*th), thoptlen = hlen;
1565 u_int8_t opts[TCP_MAXOLEN], *opt = opts;
1566 int copyback = 0, i, olen;
1567 struct raw_sackblock sack;
1569 #define TCPOLEN_SACKLEN (TCPOLEN_SACK + 2)
1570 if (hlen < TCPOLEN_SACKLEN ||
1571 !pf_pull_hdr(m, off + sizeof(*th), opts, hlen, NULL, NULL, pd->af))
1574 while (hlen >= TCPOLEN_SACKLEN) {
1577 case TCPOPT_EOL: /* FALLTHROUGH */
1585 if (olen >= TCPOLEN_SACKLEN) {
1586 for (i = 2; i + TCPOLEN_SACK <= olen;
1587 i += TCPOLEN_SACK) {
1588 memcpy(&sack, &opt[i], sizeof(sack));
1589 pf_change_a(&sack.rblk_start, &th->th_sum,
1590 htonl(sack.rblk_start -
1592 pf_change_a(&sack.rblk_end, &th->th_sum,
1593 htonl(sack.rblk_end -
1595 memcpy(&opt[i], &sack, sizeof(sack));
1609 m_copyback(m, off + sizeof(*th), thoptlen, opts);
1614 pf_send_tcp(const struct pf_rule *r, sa_family_t af,
1615 const struct pf_addr *saddr, const struct pf_addr *daddr,
1616 u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack,
1617 u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl, int tag,
1618 u_int16_t rtag, struct ether_header *eh, struct ifnet *ifp)
1623 struct ip *h = NULL;
1626 struct ip6_hdr *h6 = NULL;
1628 struct tcphdr *th = NULL;
1631 /* maximum segment size tcp option */
1632 tlen = sizeof(struct tcphdr);
1639 len = sizeof(struct ip) + tlen;
1644 len = sizeof(struct ip6_hdr) + tlen;
1649 /* create outgoing mbuf */
1650 m = m_gethdr(MB_DONTWAIT, MT_HEADER);
1654 m->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
1655 m->m_pkthdr.pf.tag = rtag;
1657 if (r != NULL && r->rtableid >= 0)
1658 m->m_pkthdr.pf.rtableid = r->rtableid;
1661 if (r != NULL && r->qid) {
1662 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
1663 m->m_pkthdr.pf.qid = r->qid;
1664 m->m_pkthdr.pf.ecn_af = af;
1665 m->m_pkthdr.pf.hdr = mtod(m, struct ip *);
1668 m->m_data += max_linkhdr;
1669 m->m_pkthdr.len = m->m_len = len;
1670 m->m_pkthdr.rcvif = NULL;
1671 bzero(m->m_data, len);
1675 h = mtod(m, struct ip *);
1677 /* IP header fields included in the TCP checksum */
1678 h->ip_p = IPPROTO_TCP;
1680 h->ip_src.s_addr = saddr->v4.s_addr;
1681 h->ip_dst.s_addr = daddr->v4.s_addr;
1683 th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip));
1688 h6 = mtod(m, struct ip6_hdr *);
1690 /* IP header fields included in the TCP checksum */
1691 h6->ip6_nxt = IPPROTO_TCP;
1692 h6->ip6_plen = htons(tlen);
1693 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr));
1694 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr));
1696 th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr));
1702 th->th_sport = sport;
1703 th->th_dport = dport;
1704 th->th_seq = htonl(seq);
1705 th->th_ack = htonl(ack);
1706 th->th_off = tlen >> 2;
1707 th->th_flags = flags;
1708 th->th_win = htons(win);
1711 opt = (char *)(th + 1);
1712 opt[0] = TCPOPT_MAXSEG;
1715 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2);
1722 th->th_sum = in_cksum(m, len);
1724 /* Finish the IP header */
1726 h->ip_hl = sizeof(*h) >> 2;
1727 h->ip_tos = IPTOS_LOWDELAY;
1729 h->ip_off = path_mtu_discovery ? IP_DF : 0;
1730 h->ip_ttl = ttl ? ttl : ip_defttl;
1733 ip_output(m, NULL, NULL, 0, NULL, NULL);
1737 struct ether_header *e = (void *)ro.ro_dst.sa_data;
1745 ro.ro_dst.sa_len = sizeof(ro.ro_dst);
1746 ro.ro_dst.sa_family = pseudo_AF_HDRCMPLT;
1747 bcopy(eh->ether_dhost, e->ether_shost, ETHER_ADDR_LEN);
1748 bcopy(eh->ether_shost, e->ether_dhost, ETHER_ADDR_LEN);
1749 e->ether_type = eh->ether_type;
1750 /* XXX_IMPORT: later */
1751 ip_output(m, (void *)NULL, &ro, 0,
1752 (void *)NULL, (void *)NULL);
1759 th->th_sum = in6_cksum(m, IPPROTO_TCP,
1760 sizeof(struct ip6_hdr), tlen);
1762 h6->ip6_vfc |= IPV6_VERSION;
1763 h6->ip6_hlim = IPV6_DEFHLIM;
1765 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
1772 pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af,
1777 m0 = m_copy(m, 0, M_COPYALL);
1778 m0->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
1780 if (r->rtableid >= 0)
1781 m0->m_pkthdr.pf.rtableid = r->rtableid;
1785 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
1786 m0->m_pkthdr.pf.qid = r->qid;
1787 m0->m_pkthdr.pf.ecn_af = af;
1788 m0->m_pkthdr.pf.hdr = mtod(m0, struct ip *);
1795 icmp_error(m0, type, code, 0, 0);
1800 icmp6_error(m0, type, code, 0);
1807 * Return 1 if the addresses a and b match (with mask m), otherwise return 0.
1808 * If n is 0, they match if they are equal. If n is != 0, they match if they
1812 pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m,
1813 struct pf_addr *b, sa_family_t af)
1820 if ((a->addr32[0] & m->addr32[0]) ==
1821 (b->addr32[0] & m->addr32[0]))
1827 if (((a->addr32[0] & m->addr32[0]) ==
1828 (b->addr32[0] & m->addr32[0])) &&
1829 ((a->addr32[1] & m->addr32[1]) ==
1830 (b->addr32[1] & m->addr32[1])) &&
1831 ((a->addr32[2] & m->addr32[2]) ==
1832 (b->addr32[2] & m->addr32[2])) &&
1833 ((a->addr32[3] & m->addr32[3]) ==
1834 (b->addr32[3] & m->addr32[3])))
1853 pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p)
1857 return ((p > a1) && (p < a2));
1859 return ((p < a1) || (p > a2));
1861 return ((p >= a1) && (p <= a2));
1875 return (0); /* never reached */
1879 pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p)
1884 return (pf_match(op, a1, a2, p));
1888 pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u)
1890 if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
1892 return (pf_match(op, a1, a2, u));
1896 pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g)
1898 if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
1900 return (pf_match(op, a1, a2, g));
1904 pf_match_tag(struct mbuf *m, struct pf_rule *r, int *tag)
1907 *tag = m->m_pkthdr.pf.tag;
1909 return ((!r->match_tag_not && r->match_tag == *tag) ||
1910 (r->match_tag_not && r->match_tag != *tag));
1914 pf_tag_packet(struct mbuf *m, int tag, int rtableid)
1916 if (tag <= 0 && rtableid < 0)
1920 m->m_pkthdr.pf.tag = tag;
1922 m->m_pkthdr.pf.rtableid = rtableid;
1928 pf_step_into_anchor(int *depth, struct pf_ruleset **rs, int n,
1929 struct pf_rule **r, struct pf_rule **a, int *match)
1931 struct pf_anchor_stackframe *f;
1933 (*r)->anchor->match = 0;
1936 if (*depth >= sizeof(pf_anchor_stack) /
1937 sizeof(pf_anchor_stack[0])) {
1938 kprintf("pf_step_into_anchor: stack overflow\n");
1939 *r = TAILQ_NEXT(*r, entries);
1941 } else if (*depth == 0 && a != NULL)
1943 f = pf_anchor_stack + (*depth)++;
1946 if ((*r)->anchor_wildcard) {
1947 f->parent = &(*r)->anchor->children;
1948 if ((f->child = RB_MIN(pf_anchor_node, f->parent)) ==
1953 *rs = &f->child->ruleset;
1957 *rs = &(*r)->anchor->ruleset;
1959 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
1963 pf_step_out_of_anchor(int *depth, struct pf_ruleset **rs, int n,
1964 struct pf_rule **r, struct pf_rule **a, int *match)
1966 struct pf_anchor_stackframe *f;
1972 f = pf_anchor_stack + *depth - 1;
1973 if (f->parent != NULL && f->child != NULL) {
1974 if (f->child->match ||
1975 (match != NULL && *match)) {
1976 f->r->anchor->match = 1;
1979 f->child = RB_NEXT(pf_anchor_node, f->parent, f->child);
1980 if (f->child != NULL) {
1981 *rs = &f->child->ruleset;
1982 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
1990 if (*depth == 0 && a != NULL)
1993 if (f->r->anchor->match || (match != NULL && *match))
1994 quick = f->r->quick;
1995 *r = TAILQ_NEXT(f->r, entries);
1996 } while (*r == NULL);
2003 pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr,
2004 struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af)
2009 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2010 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2014 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2015 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2016 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) |
2017 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]);
2018 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) |
2019 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]);
2020 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) |
2021 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]);
2027 pf_addr_inc(struct pf_addr *addr, sa_family_t af)
2032 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1);
2036 if (addr->addr32[3] == 0xffffffff) {
2037 addr->addr32[3] = 0;
2038 if (addr->addr32[2] == 0xffffffff) {
2039 addr->addr32[2] = 0;
2040 if (addr->addr32[1] == 0xffffffff) {
2041 addr->addr32[1] = 0;
2043 htonl(ntohl(addr->addr32[0]) + 1);
2046 htonl(ntohl(addr->addr32[1]) + 1);
2049 htonl(ntohl(addr->addr32[2]) + 1);
2052 htonl(ntohl(addr->addr32[3]) + 1);
2058 #define mix(a,b,c) \
2060 a -= b; a -= c; a ^= (c >> 13); \
2061 b -= c; b -= a; b ^= (a << 8); \
2062 c -= a; c -= b; c ^= (b >> 13); \
2063 a -= b; a -= c; a ^= (c >> 12); \
2064 b -= c; b -= a; b ^= (a << 16); \
2065 c -= a; c -= b; c ^= (b >> 5); \
2066 a -= b; a -= c; a ^= (c >> 3); \
2067 b -= c; b -= a; b ^= (a << 10); \
2068 c -= a; c -= b; c ^= (b >> 15); \
2072 * hash function based on bridge_hash in if_bridge.c
2075 pf_hash(struct pf_addr *inaddr, struct pf_addr *hash,
2076 struct pf_poolhashkey *key, sa_family_t af)
2078 u_int32_t a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0];
2083 a += inaddr->addr32[0];
2086 hash->addr32[0] = c + key->key32[2];
2091 a += inaddr->addr32[0];
2092 b += inaddr->addr32[2];
2094 hash->addr32[0] = c;
2095 a += inaddr->addr32[1];
2096 b += inaddr->addr32[3];
2099 hash->addr32[1] = c;
2100 a += inaddr->addr32[2];
2101 b += inaddr->addr32[1];
2104 hash->addr32[2] = c;
2105 a += inaddr->addr32[3];
2106 b += inaddr->addr32[0];
2109 hash->addr32[3] = c;
2116 pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr,
2117 struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn)
2119 unsigned char hash[16];
2120 struct pf_pool *rpool = &r->rpool;
2121 struct pf_addr *raddr = &rpool->cur->addr.v.a.addr;
2122 struct pf_addr *rmask = &rpool->cur->addr.v.a.mask;
2123 struct pf_pooladdr *acur = rpool->cur;
2124 struct pf_src_node k;
2126 if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR &&
2127 (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2129 PF_ACPY(&k.addr, saddr, af);
2130 if (r->rule_flag & PFRULE_RULESRCTRACK ||
2131 r->rpool.opts & PF_POOL_STICKYADDR)
2135 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
2136 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
2137 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) {
2138 PF_ACPY(naddr, &(*sn)->raddr, af);
2139 if (pf_status.debug >= PF_DEBUG_MISC) {
2140 kprintf("pf_map_addr: src tracking maps ");
2141 pf_print_host(&k.addr, 0, af);
2143 pf_print_host(naddr, 0, af);
2150 if (rpool->cur->addr.type == PF_ADDR_NOROUTE)
2152 if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2156 if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 &&
2157 (rpool->opts & PF_POOL_TYPEMASK) !=
2160 raddr = &rpool->cur->addr.p.dyn->pfid_addr4;
2161 rmask = &rpool->cur->addr.p.dyn->pfid_mask4;
2166 if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 &&
2167 (rpool->opts & PF_POOL_TYPEMASK) !=
2170 raddr = &rpool->cur->addr.p.dyn->pfid_addr6;
2171 rmask = &rpool->cur->addr.p.dyn->pfid_mask6;
2175 } else if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2176 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN)
2177 return (1); /* unsupported */
2179 raddr = &rpool->cur->addr.v.a.addr;
2180 rmask = &rpool->cur->addr.v.a.mask;
2183 switch (rpool->opts & PF_POOL_TYPEMASK) {
2185 PF_ACPY(naddr, raddr, af);
2187 case PF_POOL_BITMASK:
2188 PF_POOLMASK(naddr, raddr, rmask, saddr, af);
2190 case PF_POOL_RANDOM:
2191 if (init_addr != NULL && PF_AZERO(init_addr, af)) {
2195 rpool->counter.addr32[0] = karc4random();
2200 if (rmask->addr32[3] != 0xffffffff)
2201 rpool->counter.addr32[3] =
2205 if (rmask->addr32[2] != 0xffffffff)
2206 rpool->counter.addr32[2] =
2210 if (rmask->addr32[1] != 0xffffffff)
2211 rpool->counter.addr32[1] =
2215 if (rmask->addr32[0] != 0xffffffff)
2216 rpool->counter.addr32[0] =
2221 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2222 PF_ACPY(init_addr, naddr, af);
2225 PF_AINC(&rpool->counter, af);
2226 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2229 case PF_POOL_SRCHASH:
2230 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af);
2231 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af);
2233 case PF_POOL_ROUNDROBIN:
2234 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2235 if (!pfr_pool_get(rpool->cur->addr.p.tbl,
2236 &rpool->tblidx, &rpool->counter,
2237 &raddr, &rmask, af))
2239 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2240 if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2241 &rpool->tblidx, &rpool->counter,
2242 &raddr, &rmask, af))
2244 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af))
2248 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL)
2249 rpool->cur = TAILQ_FIRST(&rpool->list);
2250 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2252 if (pfr_pool_get(rpool->cur->addr.p.tbl,
2253 &rpool->tblidx, &rpool->counter,
2254 &raddr, &rmask, af)) {
2255 /* table contains no address of type 'af' */
2256 if (rpool->cur != acur)
2260 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2262 if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2263 &rpool->tblidx, &rpool->counter,
2264 &raddr, &rmask, af)) {
2265 /* table contains no address of type 'af' */
2266 if (rpool->cur != acur)
2271 raddr = &rpool->cur->addr.v.a.addr;
2272 rmask = &rpool->cur->addr.v.a.mask;
2273 PF_ACPY(&rpool->counter, raddr, af);
2277 PF_ACPY(naddr, &rpool->counter, af);
2278 if (init_addr != NULL && PF_AZERO(init_addr, af))
2279 PF_ACPY(init_addr, naddr, af);
2280 PF_AINC(&rpool->counter, af);
2284 PF_ACPY(&(*sn)->raddr, naddr, af);
2286 if (pf_status.debug >= PF_DEBUG_MISC &&
2287 (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2288 kprintf("pf_map_addr: selected address ");
2289 pf_print_host(naddr, 0, af);
2297 pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r,
2298 struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport,
2299 struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high,
2300 struct pf_src_node **sn)
2302 struct pf_state_key_cmp key;
2303 struct pf_addr init_addr;
2306 bzero(&init_addr, sizeof(init_addr));
2307 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2310 if (proto == IPPROTO_ICMP) {
2318 PF_ACPY(&key.ext.addr, daddr, key.af);
2319 PF_ACPY(&key.gwy.addr, naddr, key.af);
2320 key.ext.port = dport;
2323 * port search; start random, step;
2324 * similar 2 portloop in in_pcbbind
2326 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP ||
2327 proto == IPPROTO_ICMP)) {
2328 key.gwy.port = dport;
2329 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
2331 } else if (low == 0 && high == 0) {
2332 key.gwy.port = *nport;
2333 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
2335 } else if (low == high) {
2336 key.gwy.port = htons(low);
2337 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) {
2338 *nport = htons(low);
2350 cut = karc4random() % (1 + high - low) + low;
2351 /* low <= cut <= high */
2352 for (tmp = cut; tmp <= high; ++(tmp)) {
2353 key.gwy.port = htons(tmp);
2354 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
2356 *nport = htons(tmp);
2360 for (tmp = cut - 1; tmp >= low; --(tmp)) {
2361 key.gwy.port = htons(tmp);
2362 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
2364 *nport = htons(tmp);
2370 switch (r->rpool.opts & PF_POOL_TYPEMASK) {
2371 case PF_POOL_RANDOM:
2372 case PF_POOL_ROUNDROBIN:
2373 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2377 case PF_POOL_SRCHASH:
2378 case PF_POOL_BITMASK:
2382 } while (! PF_AEQ(&init_addr, naddr, af) );
2384 return (1); /* none available */
2388 pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off,
2389 int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport,
2390 struct pf_addr *daddr, u_int16_t dport, int rs_num)
2392 struct pf_rule *r, *rm = NULL;
2393 struct pf_ruleset *ruleset = NULL;
2398 r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr);
2399 while (r && rm == NULL) {
2400 struct pf_rule_addr *src = NULL, *dst = NULL;
2401 struct pf_addr_wrap *xdst = NULL;
2403 if (r->action == PF_BINAT && direction == PF_IN) {
2405 if (r->rpool.cur != NULL)
2406 xdst = &r->rpool.cur->addr;
2413 if (pfi_kif_match(r->kif, kif) == r->ifnot)
2414 r = r->skip[PF_SKIP_IFP].ptr;
2415 else if (r->direction && r->direction != direction)
2416 r = r->skip[PF_SKIP_DIR].ptr;
2417 else if (r->af && r->af != pd->af)
2418 r = r->skip[PF_SKIP_AF].ptr;
2419 else if (r->proto && r->proto != pd->proto)
2420 r = r->skip[PF_SKIP_PROTO].ptr;
2421 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af,
2423 r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR :
2424 PF_SKIP_DST_ADDR].ptr;
2425 else if (src->port_op && !pf_match_port(src->port_op,
2426 src->port[0], src->port[1], sport))
2427 r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT :
2428 PF_SKIP_DST_PORT].ptr;
2429 else if (dst != NULL &&
2430 PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->neg, NULL))
2431 r = r->skip[PF_SKIP_DST_ADDR].ptr;
2432 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af,
2434 r = TAILQ_NEXT(r, entries);
2435 else if (dst != NULL && dst->port_op &&
2436 !pf_match_port(dst->port_op, dst->port[0],
2437 dst->port[1], dport))
2438 r = r->skip[PF_SKIP_DST_PORT].ptr;
2439 else if (r->match_tag && !pf_match_tag(m, r, &tag))
2440 r = TAILQ_NEXT(r, entries);
2441 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto !=
2442 IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m,
2443 off, pd->hdr.tcp), r->os_fingerprint)))
2444 r = TAILQ_NEXT(r, entries);
2448 if (r->rtableid >= 0)
2449 rtableid = r->rtableid;
2450 if (r->anchor == NULL) {
2453 pf_step_into_anchor(&asd, &ruleset, rs_num,
2457 pf_step_out_of_anchor(&asd, &ruleset, rs_num, &r,
2460 if (pf_tag_packet(m, tag, rtableid))
2462 if (rm != NULL && (rm->action == PF_NONAT ||
2463 rm->action == PF_NORDR || rm->action == PF_NOBINAT))
2469 pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction,
2470 struct pfi_kif *kif, struct pf_src_node **sn,
2471 struct pf_addr *saddr, u_int16_t sport,
2472 struct pf_addr *daddr, u_int16_t dport,
2473 struct pf_addr *naddr, u_int16_t *nport)
2475 struct pf_rule *r = NULL;
2477 if (direction == PF_OUT) {
2478 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2479 sport, daddr, dport, PF_RULESET_BINAT);
2481 r = pf_match_translation(pd, m, off, direction, kif,
2482 saddr, sport, daddr, dport, PF_RULESET_NAT);
2484 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2485 sport, daddr, dport, PF_RULESET_RDR);
2487 r = pf_match_translation(pd, m, off, direction, kif,
2488 saddr, sport, daddr, dport, PF_RULESET_BINAT);
2492 switch (r->action) {
2498 if (pf_get_sport(pd->af, pd->proto, r, saddr,
2499 daddr, dport, naddr, nport, r->rpool.proxy_port[0],
2500 r->rpool.proxy_port[1], sn)) {
2501 DPFPRINTF(PF_DEBUG_MISC,
2502 ("pf: NAT proxy port allocation "
2504 r->rpool.proxy_port[0],
2505 r->rpool.proxy_port[1]));
2510 switch (direction) {
2512 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){
2516 if (r->rpool.cur->addr.p.dyn->
2520 &r->rpool.cur->addr.p.dyn->
2522 &r->rpool.cur->addr.p.dyn->
2529 if (r->rpool.cur->addr.p.dyn->
2533 &r->rpool.cur->addr.p.dyn->
2535 &r->rpool.cur->addr.p.dyn->
2543 &r->rpool.cur->addr.v.a.addr,
2544 &r->rpool.cur->addr.v.a.mask,
2548 if (r->src.addr.type == PF_ADDR_DYNIFTL) {
2552 if (r->src.addr.p.dyn->
2556 &r->src.addr.p.dyn->
2558 &r->src.addr.p.dyn->
2565 if (r->src.addr.p.dyn->
2569 &r->src.addr.p.dyn->
2571 &r->src.addr.p.dyn->
2579 &r->src.addr.v.a.addr,
2580 &r->src.addr.v.a.mask, daddr,
2586 if (pf_map_addr(pd->af, r, saddr, naddr, NULL, sn))
2588 if ((r->rpool.opts & PF_POOL_TYPEMASK) ==
2590 PF_POOLMASK(naddr, naddr,
2591 &r->rpool.cur->addr.v.a.mask, daddr,
2594 if (r->rpool.proxy_port[1]) {
2595 u_int32_t tmp_nport;
2597 tmp_nport = ((ntohs(dport) -
2598 ntohs(r->dst.port[0])) %
2599 (r->rpool.proxy_port[1] -
2600 r->rpool.proxy_port[0] + 1)) +
2601 r->rpool.proxy_port[0];
2603 /* wrap around if necessary */
2604 if (tmp_nport > 65535)
2606 *nport = htons((u_int16_t)tmp_nport);
2607 } else if (r->rpool.proxy_port[0])
2608 *nport = htons(r->rpool.proxy_port[0]);
2620 struct netmsg_hashlookup {
2621 struct netmsg nm_netmsg;
2622 struct inpcb **nm_pinp;
2623 struct inpcbinfo *nm_pcbinfo;
2624 struct pf_addr *nm_saddr;
2625 struct pf_addr *nm_daddr;
2632 in_pcblookup_hash_handler(struct netmsg *msg0)
2634 struct netmsg_hashlookup *msg = (struct netmsg_hashlookup *)msg0;
2636 if (msg->nm_af == AF_INET)
2637 *msg->nm_pinp = in_pcblookup_hash(msg->nm_pcbinfo,
2638 msg->nm_saddr->v4, msg->nm_sport, msg->nm_daddr->v4,
2639 msg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2642 *msg->nm_pinp = in6_pcblookup_hash(msg->nm_pcbinfo,
2643 &msg->nm_saddr->v6, msg->nm_sport, &msg->nm_daddr->v6,
2644 msg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2646 lwkt_replymsg(&msg->nm_netmsg.nm_lmsg, 0);
2651 pf_socket_lookup(int direction, struct pf_pdesc *pd)
2653 struct pf_addr *saddr, *daddr;
2654 u_int16_t sport, dport;
2655 struct inpcbinfo *pi;
2658 struct netmsg_hashlookup *msg = NULL;
2664 pd->lookup.uid = UID_MAX;
2665 pd->lookup.gid = GID_MAX;
2666 pd->lookup.pid = NO_PID;
2667 if (direction == PF_IN) {
2674 switch (pd->proto) {
2676 if (pd->hdr.tcp == NULL)
2678 sport = pd->hdr.tcp->th_sport;
2679 dport = pd->hdr.tcp->th_dport;
2681 pi_cpu = tcp_addrcpu(saddr->v4.s_addr, sport, daddr->v4.s_addr, dport);
2682 pi = &tcbinfo[pi_cpu];
2685 * Our netstack runs lockless on MP systems
2686 * (only for TCP connections at the moment).
2688 * As we are not allowed to read another CPU's tcbinfo,
2689 * we have to ask that CPU via remote call to search the
2692 * Prepare a msg iff data belongs to another CPU.
2694 if (pi_cpu != mycpu->gd_cpuid) {
2695 msg = kmalloc(sizeof(*msg), M_LWKTMSG, M_INTWAIT);
2696 netmsg_init(&msg->nm_netmsg, NULL, &netisr_afree_rport,
2697 0, in_pcblookup_hash_handler);
2698 msg->nm_pinp = &inp;
2699 msg->nm_pcbinfo = pi;
2700 msg->nm_saddr = saddr;
2701 msg->nm_sport = sport;
2702 msg->nm_daddr = daddr;
2703 msg->nm_dport = dport;
2704 msg->nm_af = pd->af;
2709 if (pd->hdr.udp == NULL)
2711 sport = pd->hdr.udp->uh_sport;
2712 dport = pd->hdr.udp->uh_dport;
2718 if (direction != PF_IN) {
2730 * Query other CPU, second part
2732 * msg only gets initialized when:
2734 * 2) the info belongs to another CPU
2736 * Use some switch/case magic to avoid code duplication.
2741 inp = in6_pcblookup_hash(pi, &saddr->v6, sport,
2742 &daddr->v6, dport, INPLOOKUP_WILDCARD, NULL);
2748 /* FALLTHROUGH if SMP and on other CPU */
2753 lwkt_sendmsg(tcp_cport(pi_cpu),
2754 &msg->nm_netmsg.nm_lmsg);
2758 inp = in_pcblookup_hash(pi, saddr->v4, sport, daddr->v4,
2759 dport, INPLOOKUP_WILDCARD, NULL);
2768 pd->lookup.uid = inp->inp_socket->so_cred->cr_uid;
2769 pd->lookup.gid = inp->inp_socket->so_cred->cr_groups[0];
2774 pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2778 u_int8_t *opt, optlen;
2779 u_int8_t wscale = 0;
2781 hlen = th_off << 2; /* hlen <= sizeof(hdr) */
2782 if (hlen <= sizeof(struct tcphdr))
2784 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2786 opt = hdr + sizeof(struct tcphdr);
2787 hlen -= sizeof(struct tcphdr);
2797 if (wscale > TCP_MAX_WINSHIFT)
2798 wscale = TCP_MAX_WINSHIFT;
2799 wscale |= PF_WSCALE_FLAG;
2814 pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2818 u_int8_t *opt, optlen;
2819 u_int16_t mss = tcp_mssdflt;
2821 hlen = th_off << 2; /* hlen <= sizeof(hdr) */
2822 if (hlen <= sizeof(struct tcphdr))
2824 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2826 opt = hdr + sizeof(struct tcphdr);
2827 hlen -= sizeof(struct tcphdr);
2828 while (hlen >= TCPOLEN_MAXSEG) {
2836 bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2);
2851 pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer)
2854 struct sockaddr_in *dst;
2858 struct sockaddr_in6 *dst6;
2859 struct route_in6 ro6;
2861 struct rtentry *rt = NULL;
2863 u_int16_t mss = tcp_mssdflt;
2868 hlen = sizeof(struct ip);
2869 bzero(&ro, sizeof(ro));
2870 dst = (struct sockaddr_in *)&ro.ro_dst;
2871 dst->sin_family = AF_INET;
2872 dst->sin_len = sizeof(*dst);
2873 dst->sin_addr = addr->v4;
2874 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
2880 hlen = sizeof(struct ip6_hdr);
2881 bzero(&ro6, sizeof(ro6));
2882 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst;
2883 dst6->sin6_family = AF_INET6;
2884 dst6->sin6_len = sizeof(*dst6);
2885 dst6->sin6_addr = addr->v6;
2886 rtalloc_ign((struct route *)&ro6, (RTF_CLONING | RTF_PRCLONING));
2892 if (rt && rt->rt_ifp) {
2893 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr);
2894 mss = max(tcp_mssdflt, mss);
2897 mss = min(mss, offer);
2898 mss = max(mss, 64); /* sanity - at least max opt space */
2903 pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr)
2905 struct pf_rule *r = s->rule.ptr;
2908 if (!r->rt || r->rt == PF_FASTROUTE)
2910 switch (s->state_key->af) {
2913 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL,
2915 s->rt_kif = r->rpool.cur->kif;
2920 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL,
2922 s->rt_kif = r->rpool.cur->kif;
2929 pf_attach_state(struct pf_state_key *sk, struct pf_state *s, int tail)
2934 /* list is sorted, if-bound states before floating */
2936 TAILQ_INSERT_TAIL(&sk->states, s, next);
2938 TAILQ_INSERT_HEAD(&sk->states, s, next);
2942 pf_detach_state(struct pf_state *s, int flags)
2944 struct pf_state_key *sk = s->state_key;
2949 s->state_key = NULL;
2950 TAILQ_REMOVE(&sk->states, s, next);
2951 if (--sk->refcnt == 0) {
2952 if (!(flags & PF_DT_SKIP_EXTGWY))
2953 RB_REMOVE(pf_state_tree_ext_gwy,
2954 &pf_statetbl_ext_gwy, sk);
2955 if (!(flags & PF_DT_SKIP_LANEXT))
2956 RB_REMOVE(pf_state_tree_lan_ext,
2957 &pf_statetbl_lan_ext, sk);
2958 pool_put(&pf_state_key_pl, sk);
2962 struct pf_state_key *
2963 pf_alloc_state_key(struct pf_state *s)
2965 struct pf_state_key *sk;
2967 if ((sk = pool_get(&pf_state_key_pl, PR_NOWAIT)) == NULL)
2969 bzero(sk, sizeof(*sk));
2970 TAILQ_INIT(&sk->states);
2971 pf_attach_state(sk, s, 0);
2977 pf_test_rule(struct pf_rule **rm, struct pf_state **sm, int direction,
2978 struct pfi_kif *kif, struct mbuf *m, int off, void *h,
2979 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
2980 struct ifqueue *ifq, struct inpcb *inp)
2982 struct pf_rule *nr = NULL;
2983 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
2984 u_int16_t bport, nport = 0;
2985 sa_family_t af = pd->af;
2986 struct pf_rule *r, *a = NULL;
2987 struct pf_ruleset *ruleset = NULL;
2988 struct pf_src_node *nsn = NULL;
2989 struct tcphdr *th = pd->hdr.tcp;
2991 int rewrite = 0, hdrlen = 0;
2992 int tag = -1, rtableid = -1;
2996 u_int16_t mss = tcp_mssdflt;
2997 u_int16_t sport, dport;
2998 u_int8_t icmptype = 0, icmpcode = 0;
3000 if (direction == PF_IN && pf_check_congestion(ifq)) {
3001 REASON_SET(&reason, PFRES_CONGEST);
3006 pd->lookup.done = pf_socket_lookup(direction, pd);
3007 else if (debug_pfugidhack) {
3009 DPFPRINTF(PF_DEBUG_MISC, ("pf: unlocked lookup\n"));
3010 pd->lookup.done = pf_socket_lookup(direction, pd);
3014 sport = dport = hdrlen = 0;
3016 switch (pd->proto) {
3018 sport = th->th_sport;
3019 dport = th->th_dport;
3020 hdrlen = sizeof(*th);
3023 sport = pd->hdr.udp->uh_sport;
3024 dport = pd->hdr.udp->uh_dport;
3025 hdrlen = sizeof(*pd->hdr.udp);
3029 if (pd->af != AF_INET)
3031 sport = dport = pd->hdr.icmp->icmp_id;
3032 icmptype = pd->hdr.icmp->icmp_type;
3033 icmpcode = pd->hdr.icmp->icmp_code;
3035 if (icmptype == ICMP_UNREACH ||
3036 icmptype == ICMP_SOURCEQUENCH ||
3037 icmptype == ICMP_REDIRECT ||
3038 icmptype == ICMP_TIMXCEED ||
3039 icmptype == ICMP_PARAMPROB)
3044 case IPPROTO_ICMPV6:
3045 if (pd->af != AF_INET6)
3047 sport = dport = pd->hdr.icmp6->icmp6_id;
3048 hdrlen = sizeof(*pd->hdr.icmp6);
3049 icmptype = pd->hdr.icmp6->icmp6_type;
3050 icmpcode = pd->hdr.icmp6->icmp6_code;
3052 if (icmptype == ICMP6_DST_UNREACH ||
3053 icmptype == ICMP6_PACKET_TOO_BIG ||
3054 icmptype == ICMP6_TIME_EXCEEDED ||
3055 icmptype == ICMP6_PARAM_PROB)
3061 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3063 if (direction == PF_OUT) {
3064 bport = nport = sport;
3065 /* check outgoing packet for BINAT/NAT */
3066 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3067 saddr, sport, daddr, dport, &pd->naddr, &nport)) != NULL) {
3068 PF_ACPY(&pd->baddr, saddr, af);
3069 switch (pd->proto) {
3071 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
3072 &th->th_sum, &pd->naddr, nport, 0, af);
3073 sport = th->th_sport;
3077 pf_change_ap(saddr, &pd->hdr.udp->uh_sport,
3078 pd->ip_sum, &pd->hdr.udp->uh_sum,
3079 &pd->naddr, nport, 1, af);
3080 sport = pd->hdr.udp->uh_sport;
3085 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3086 pd->naddr.v4.s_addr, 0);
3087 pd->hdr.icmp->icmp_cksum = pf_cksum_fixup(
3088 pd->hdr.icmp->icmp_cksum, sport, nport, 0);
3089 pd->hdr.icmp->icmp_id = nport;
3090 m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
3094 case IPPROTO_ICMPV6:
3095 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum,
3104 pf_change_a(&saddr->v4.s_addr,
3105 pd->ip_sum, pd->naddr.v4.s_addr, 0);
3110 PF_ACPY(saddr, &pd->naddr, af);
3122 bport = nport = dport;
3123 /* check incoming packet for BINAT/RDR */
3124 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3125 saddr, sport, daddr, dport, &pd->naddr, &nport)) != NULL) {
3126 PF_ACPY(&pd->baddr, daddr, af);
3127 switch (pd->proto) {
3129 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
3130 &th->th_sum, &pd->naddr, nport, 0, af);
3131 dport = th->th_dport;
3135 pf_change_ap(daddr, &pd->hdr.udp->uh_dport,
3136 pd->ip_sum, &pd->hdr.udp->uh_sum,
3137 &pd->naddr, nport, 1, af);
3138 dport = pd->hdr.udp->uh_dport;
3143 pf_change_a(&daddr->v4.s_addr, pd->ip_sum,
3144 pd->naddr.v4.s_addr, 0);
3148 case IPPROTO_ICMPV6:
3149 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum,
3158 pf_change_a(&daddr->v4.s_addr,
3159 pd->ip_sum, pd->naddr.v4.s_addr, 0);
3164 PF_ACPY(daddr, &pd->naddr, af);
3179 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3180 r = r->skip[PF_SKIP_IFP].ptr;
3181 else if (r->direction && r->direction != direction)
3182 r = r->skip[PF_SKIP_DIR].ptr;
3183 else if (r->af && r->af != af)
3184 r = r->skip[PF_SKIP_AF].ptr;
3185 else if (r->proto && r->proto != pd->proto)
3186 r = r->skip[PF_SKIP_PROTO].ptr;
3187 else if (PF_MISMATCHAW(&r->src.addr, saddr, af,
3189 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3190 /* tcp/udp only. port_op always 0 in other cases */
3191 else if (r->src.port_op && !pf_match_port(r->src.port_op,
3192 r->src.port[0], r->src.port[1], sport))
3193 r = r->skip[PF_SKIP_SRC_PORT].ptr;
3194 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af,
3196 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3197 /* tcp/udp only. port_op always 0 in other cases */
3198 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
3199 r->dst.port[0], r->dst.port[1], dport))
3200 r = r->skip[PF_SKIP_DST_PORT].ptr;
3201 /* icmp only. type always 0 in other cases */
3202 else if (r->type && r->type != icmptype + 1)
3203 r = TAILQ_NEXT(r, entries);
3204 /* icmp only. type always 0 in other cases */
3205 else if (r->code && r->code != icmpcode + 1)
3206 r = TAILQ_NEXT(r, entries);
3207 else if (r->tos && !(r->tos == pd->tos))
3208 r = TAILQ_NEXT(r, entries);
3209 else if (r->rule_flag & PFRULE_FRAGMENT)
3210 r = TAILQ_NEXT(r, entries);
3211 else if (pd->proto == IPPROTO_TCP &&
3212 (r->flagset & th->th_flags) != r->flags)
3213 r = TAILQ_NEXT(r, entries);
3214 /* tcp/udp only. uid.op always 0 in other cases */
3215 else if (r->uid.op && (pd->lookup.done || (pd->lookup.done =
3216 pf_socket_lookup(direction, pd), 1)) &&
3217 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
3219 r = TAILQ_NEXT(r, entries);
3220 /* tcp/udp only. gid.op always 0 in other cases */
3221 else if (r->gid.op && (pd->lookup.done || (pd->lookup.done =
3222 pf_socket_lookup(direction, pd), 1)) &&
3223 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
3225 r = TAILQ_NEXT(r, entries);
3226 else if (r->prob && r->prob <= karc4random())
3227 r = TAILQ_NEXT(r, entries);
3228 else if (r->match_tag && !pf_match_tag(m, r, &tag))
3229 r = TAILQ_NEXT(r, entries);
3230 else if (r->os_fingerprint != PF_OSFP_ANY &&
3231 (pd->proto != IPPROTO_TCP || !pf_osfp_match(
3232 pf_osfp_fingerprint(pd, m, off, th),
3233 r->os_fingerprint)))
3234 r = TAILQ_NEXT(r, entries);
3238 if (r->rtableid >= 0)
3239 rtableid = r->rtableid;
3240 if (r->anchor == NULL) {
3247 r = TAILQ_NEXT(r, entries);
3249 pf_step_into_anchor(&asd, &ruleset,
3250 PF_RULESET_FILTER, &r, &a, &match);
3252 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3253 PF_RULESET_FILTER, &r, &a, &match))
3260 REASON_SET(&reason, PFRES_MATCH);
3262 if (r->log || (nr != NULL && nr->log)) {
3264 m_copyback(m, off, hdrlen, pd->hdr.any);
3265 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr,
3269 if ((r->action == PF_DROP) &&
3270 ((r->rule_flag & PFRULE_RETURNRST) ||
3271 (r->rule_flag & PFRULE_RETURNICMP) ||
3272 (r->rule_flag & PFRULE_RETURN))) {
3273 /* undo NAT changes, if they have taken place */
3275 if (direction == PF_OUT) {
3276 switch (pd->proto) {
3278 pf_change_ap(saddr, &th->th_sport,
3279 pd->ip_sum, &th->th_sum,
3280 &pd->baddr, bport, 0, af);
3281 sport = th->th_sport;
3286 &pd->hdr.udp->uh_sport, pd->ip_sum,
3287 &pd->hdr.udp->uh_sum, &pd->baddr,
3289 sport = pd->hdr.udp->uh_sport;
3294 case IPPROTO_ICMPV6:
3301 pf_change_a(&saddr->v4.s_addr,
3303 pd->baddr.v4.s_addr, 0);
3306 PF_ACPY(saddr, &pd->baddr, af);
3311 switch (pd->proto) {
3313 pf_change_ap(daddr, &th->th_dport,
3314 pd->ip_sum, &th->th_sum,
3315 &pd->baddr, bport, 0, af);
3316 dport = th->th_dport;
3321 &pd->hdr.udp->uh_dport, pd->ip_sum,
3322 &pd->hdr.udp->uh_sum, &pd->baddr,
3324 dport = pd->hdr.udp->uh_dport;
3329 case IPPROTO_ICMPV6:
3336 pf_change_a(&daddr->v4.s_addr,
3338 pd->baddr.v4.s_addr, 0);
3341 PF_ACPY(daddr, &pd->baddr, af);
3347 if (pd->proto == IPPROTO_TCP &&
3348 ((r->rule_flag & PFRULE_RETURNRST) ||
3349 (r->rule_flag & PFRULE_RETURN)) &&
3350 !(th->th_flags & TH_RST)) {
3351 u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
3352 struct ip *h = mtod(m, struct ip *);
3354 if (pf_check_proto_cksum(m, off,
3355 h->ip_len - off, IPPROTO_TCP, AF_INET))
3356 REASON_SET(&reason, PFRES_PROTCKSUM);
3358 if (th->th_flags & TH_SYN)
3360 if (th->th_flags & TH_FIN)
3362 pf_send_tcp(r, af, pd->dst,
3363 pd->src, th->th_dport, th->th_sport,
3364 ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0,
3365 r->return_ttl, 1, 0, pd->eh, kif->pfik_ifp);
3367 } else if ((af == AF_INET) && r->return_icmp)
3368 pf_send_icmp(m, r->return_icmp >> 8,
3369 r->return_icmp & 255, af, r);
3370 else if ((af == AF_INET6) && r->return_icmp6)
3371 pf_send_icmp(m, r->return_icmp6 >> 8,
3372 r->return_icmp6 & 255, af, r);
3375 if (r->action == PF_DROP)
3378 if (pf_tag_packet(m, tag, rtableid)) {
3379 REASON_SET(&reason, PFRES_MEMORY);
3383 if (!state_icmp && (r->keep_state || nr != NULL ||
3384 (pd->flags & PFDESC_TCP_NORM))) {
3385 /* create new state */
3387 struct pf_state *s = NULL;
3388 struct pf_state_key *sk = NULL;
3389 struct pf_src_node *sn = NULL;
3391 /* check maximums */
3392 if (r->max_states && (r->states >= r->max_states)) {
3393 pf_status.lcounters[LCNT_STATES]++;
3394 REASON_SET(&reason, PFRES_MAXSTATES);
3397 /* src node for filter rule */
3398 if ((r->rule_flag & PFRULE_SRCTRACK ||
3399 r->rpool.opts & PF_POOL_STICKYADDR) &&
3400 pf_insert_src_node(&sn, r, saddr, af) != 0) {
3401 REASON_SET(&reason, PFRES_SRCLIMIT);
3404 /* src node for translation rule */
3405 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3406 ((direction == PF_OUT &&
3407 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3408 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) {
3409 REASON_SET(&reason, PFRES_SRCLIMIT);
3412 s = pool_get(&pf_state_pl, PR_NOWAIT);
3414 REASON_SET(&reason, PFRES_MEMORY);
3416 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3417 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3418 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3419 pf_status.src_nodes--;
3420 pool_put(&pf_src_tree_pl, sn);
3422 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3424 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3425 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3426 pf_status.src_nodes--;
3427 pool_put(&pf_src_tree_pl, nsn);
3430 pool_put(&pf_state_key_pl, sk);
3434 bzero(s, sizeof(*s));
3436 s->nat_rule.ptr = nr;
3438 STATE_INC_COUNTERS(s);
3439 s->allow_opts = r->allow_opts;
3440 s->log = r->log & PF_LOG_ALL;
3442 s->log |= nr->log & PF_LOG_ALL;
3443 switch (pd->proto) {
3445 len = pd->tot_len - off - (th->th_off << 2);
3446 s->src.seqlo = ntohl(th->th_seq);
3447 s->src.seqhi = s->src.seqlo + len + 1;
3448 if ((th->th_flags & (TH_SYN|TH_ACK)) ==
3449 TH_SYN && r->keep_state == PF_STATE_MODULATE) {
3450 /* Generate sequence number modulator */
3451 while ((s->src.seqdiff =
3452 pf_new_isn(sk) - s->src.seqlo) == 0)
3454 pf_change_a(&th->th_seq, &th->th_sum,
3455 htonl(s->src.seqlo + s->src.seqdiff), 0);
3459 if (th->th_flags & TH_SYN) {
3461 s->src.wscale = pf_get_wscale(m, off,
3464 s->src.max_win = MAX(ntohs(th->th_win), 1);
3465 if (s->src.wscale & PF_WSCALE_MASK) {
3466 /* Remove scale factor from initial window */
3467 int win = s->src.max_win;
3468 win += 1 << (s->src.wscale & PF_WSCALE_MASK);
3469 s->src.max_win = (win - 1) >>
3470 (s->src.wscale & PF_WSCALE_MASK);
3472 if (th->th_flags & TH_FIN)
3476 s->src.state = TCPS_SYN_SENT;
3477 s->dst.state = TCPS_CLOSED;
3478 s->timeout = PFTM_TCP_FIRST_PACKET;
3481 s->src.state = PFUDPS_SINGLE;
3482 s->dst.state = PFUDPS_NO_TRAFFIC;
3483 s->timeout = PFTM_UDP_FIRST_PACKET;
3487 case IPPROTO_ICMPV6:
3489 s->timeout = PFTM_ICMP_FIRST_PACKET;
3492 s->src.state = PFOTHERS_SINGLE;
3493 s->dst.state = PFOTHERS_NO_TRAFFIC;
3494 s->timeout = PFTM_OTHER_FIRST_PACKET;
3497 s->creation = time_second;
3498 s->expire = time_second;
3502 s->src_node->states++;
3505 PF_ACPY(&nsn->raddr, &pd->naddr, af);
3506 s->nat_src_node = nsn;
3507 s->nat_src_node->states++;
3509 if (pd->proto == IPPROTO_TCP) {
3510 if ((pd->flags & PFDESC_TCP_NORM) &&
3511 pf_normalize_tcp_init(m, off, pd, th, &s->src,
3513 REASON_SET(&reason, PFRES_MEMORY);
3514 pf_src_tree_remove_state(s);
3515 STATE_DEC_COUNTERS(s);
3516 pool_put(&pf_state_pl, s);
3519 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
3520 pf_normalize_tcp_stateful(m, off, pd, &reason,
3521 th, s, &s->src, &s->dst, &rewrite)) {
3522 /* This really shouldn't happen!!! */
3523 DPFPRINTF(PF_DEBUG_URGENT,
3524 ("pf_normalize_tcp_stateful failed on "
3526 pf_normalize_tcp_cleanup(s);
3527 pf_src_tree_remove_state(s);
3528 STATE_DEC_COUNTERS(s);
3529 pool_put(&pf_state_pl, s);
3534 if ((sk = pf_alloc_state_key(s)) == NULL) {
3535 REASON_SET(&reason, PFRES_MEMORY);
3539 sk->proto = pd->proto;
3540 sk->direction = direction;
3542 if (direction == PF_OUT) {
3543 PF_ACPY(&sk->gwy.addr, saddr, af);
3544 PF_ACPY(&sk->ext.addr, daddr, af);
3545 switch (pd->proto) {
3548 case IPPROTO_ICMPV6:
3550 sk->gwy.port = nport;
3554 sk->gwy.port = sport;
3555 sk->ext.port = dport;
3558 PF_ACPY(&sk->lan.addr, &pd->baddr, af);
3559 sk->lan.port = bport;
3561 PF_ACPY(&sk->lan.addr, &sk->gwy.addr, af);
3562 sk->lan.port = sk->gwy.port;
3565 PF_ACPY(&sk->lan.addr, daddr, af);
3566 PF_ACPY(&sk->ext.addr, saddr, af);
3567 switch (pd->proto) {
3570 case IPPROTO_ICMPV6:
3572 sk->lan.port = nport;
3576 sk->lan.port = dport;
3577 sk->ext.port = sport;
3580 PF_ACPY(&sk->gwy.addr, &pd->baddr, af);
3581 sk->gwy.port = bport;
3583 PF_ACPY(&sk->gwy.addr, &sk->lan.addr, af);
3584 sk->gwy.port = sk->lan.port;
3588 s->hash = pf_state_hash(sk);
3589 s->pickup_mode = r->pickup_mode;
3591 pf_set_rt_ifp(s, saddr); /* needs s->state_key set */
3593 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3594 if (pd->proto == IPPROTO_TCP)
3595 pf_normalize_tcp_cleanup(s);
3596 REASON_SET(&reason, PFRES_STATEINS);
3597 pf_src_tree_remove_state(s);
3598 STATE_DEC_COUNTERS(s);
3599 pool_put(&pf_state_pl, s);
3607 if (pd->proto == IPPROTO_TCP &&
3608 (th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
3609 r->keep_state == PF_STATE_SYNPROXY) {
3610 s->src.state = PF_TCPS_PROXY_SRC;
3612 if (direction == PF_OUT) {
3613 pf_change_ap(saddr, &th->th_sport,
3614 pd->ip_sum, &th->th_sum, &pd->baddr,
3616 sport = th->th_sport;
3618 pf_change_ap(daddr, &th->th_dport,
3619 pd->ip_sum, &th->th_sum, &pd->baddr,
3621 sport = th->th_dport;
3624 s->src.seqhi = karc4random();
3625 /* Find mss option */
3626 mss = pf_get_mss(m, off, th->th_off, af);
3627 mss = pf_calc_mss(saddr, af, mss);
3628 mss = pf_calc_mss(daddr, af, mss);
3630 pf_send_tcp(r, af, daddr, saddr, th->th_dport,
3631 th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1,
3632 TH_SYN|TH_ACK, 0, s->src.mss, 0, 1, 0, NULL, NULL);
3633 REASON_SET(&reason, PFRES_SYNPROXY);
3634 return (PF_SYNPROXY_DROP);
3638 /* copy back packet headers if we performed NAT operations */
3640 m_copyback(m, off, hdrlen, pd->hdr.any);
3646 pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif,
3647 struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am,
3648 struct pf_ruleset **rsm)
3650 struct pf_rule *r, *a = NULL;
3651 struct pf_ruleset *ruleset = NULL;
3652 sa_family_t af = pd->af;
3658 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3661 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3662 r = r->skip[PF_SKIP_IFP].ptr;
3663 else if (r->direction && r->direction != direction)
3664 r = r->skip[PF_SKIP_DIR].ptr;
3665 else if (r->af && r->af != af)
3666 r = r->skip[PF_SKIP_AF].ptr;
3667 else if (r->proto && r->proto != pd->proto)
3668 r = r->skip[PF_SKIP_PROTO].ptr;
3669 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af,
3671 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3672 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af,
3674 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3675 else if (r->tos && !(r->tos == pd->tos))
3676 r = TAILQ_NEXT(r, entries);
3677 else if (r->os_fingerprint != PF_OSFP_ANY)
3678 r = TAILQ_NEXT(r, entries);
3679 else if (pd->proto == IPPROTO_UDP &&
3680 (r->src.port_op || r->dst.port_op))
3681 r = TAILQ_NEXT(r, entries);
3682 else if (pd->proto == IPPROTO_TCP &&
3683 (r->src.port_op || r->dst.port_op || r->flagset))
3684 r = TAILQ_NEXT(r, entries);
3685 else if ((pd->proto == IPPROTO_ICMP ||
3686 pd->proto == IPPROTO_ICMPV6) &&
3687 (r->type || r->code))
3688 r = TAILQ_NEXT(r, entries);
3689 else if (r->prob && r->prob <= karc4random())
3690 r = TAILQ_NEXT(r, entries);
3691 else if (r->match_tag && !pf_match_tag(m, r, &tag))
3692 r = TAILQ_NEXT(r, entries);
3694 if (r->anchor == NULL) {
3701 r = TAILQ_NEXT(r, entries);
3703 pf_step_into_anchor(&asd, &ruleset,
3704 PF_RULESET_FILTER, &r, &a, &match);
3706 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3707 PF_RULESET_FILTER, &r, &a, &match))
3714 REASON_SET(&reason, PFRES_MATCH);
3717 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset,
3720 if (r->action != PF_PASS)
3723 if (pf_tag_packet(m, tag, -1)) {
3724 REASON_SET(&reason, PFRES_MEMORY);
3732 pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
3733 struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
3736 struct pf_state_key_cmp key;
3737 struct tcphdr *th = pd->hdr.tcp;
3738 u_int16_t win = ntohs(th->th_win);
3739 u_int32_t ack, end, seq, orig_seq;
3743 struct pf_state_peer *src, *dst;
3746 key.proto = IPPROTO_TCP;
3747 if (direction == PF_IN) {
3748 PF_ACPY(&key.ext.addr, pd->src, key.af);
3749 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
3750 key.ext.port = th->th_sport;
3751 key.gwy.port = th->th_dport;
3753 PF_ACPY(&key.lan.addr, pd->src, key.af);
3754 PF_ACPY(&key.ext.addr, pd->dst, key.af);
3755 key.lan.port = th->th_sport;
3756 key.ext.port = th->th_dport;
3761 if (direction == (*state)->state_key->direction) {
3762 src = &(*state)->src;
3763 dst = &(*state)->dst;
3765 src = &(*state)->dst;
3766 dst = &(*state)->src;
3769 if ((*state)->src.state == PF_TCPS_PROXY_SRC) {
3770 if (direction != (*state)->state_key->direction) {
3771 REASON_SET(reason, PFRES_SYNPROXY);
3772 return (PF_SYNPROXY_DROP);
3774 if (th->th_flags & TH_SYN) {
3775 if (ntohl(th->th_seq) != (*state)->src.seqlo) {
3776 REASON_SET(reason, PFRES_SYNPROXY);
3779 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
3780 pd->src, th->th_dport, th->th_sport,
3781 (*state)->src.seqhi, ntohl(th->th_seq) + 1,
3782 TH_SYN|TH_ACK, 0, (*state)->src.mss, 0, 1,
3784 REASON_SET(reason, PFRES_SYNPROXY);
3785 return (PF_SYNPROXY_DROP);
3786 } else if (!(th->th_flags & TH_ACK) ||
3787 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
3788 (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
3789 REASON_SET(reason, PFRES_SYNPROXY);
3791 } else if ((*state)->src_node != NULL &&
3792 pf_src_connlimit(state)) {
3793 REASON_SET(reason, PFRES_SRCLIMIT);
3796 (*state)->src.state = PF_TCPS_PROXY_DST;
3798 if ((*state)->src.state == PF_TCPS_PROXY_DST) {
3799 struct pf_state_host *src, *dst;
3801 if (direction == PF_OUT) {
3802 src = &(*state)->state_key->gwy;
3803 dst = &(*state)->state_key->ext;
3805 src = &(*state)->state_key->ext;
3806 dst = &(*state)->state_key->lan;
3808 if (direction == (*state)->state_key->direction) {
3809 if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) ||
3810 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
3811 (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
3812 REASON_SET(reason, PFRES_SYNPROXY);
3815 (*state)->src.max_win = MAX(ntohs(th->th_win), 1);
3816 if ((*state)->dst.seqhi == 1)
3817 (*state)->dst.seqhi = karc4random();
3818 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
3819 &dst->addr, src->port, dst->port,
3820 (*state)->dst.seqhi, 0, TH_SYN, 0,
3821 (*state)->src.mss, 0, 0, (*state)->tag, NULL, NULL);
3822 REASON_SET(reason, PFRES_SYNPROXY);
3823 return (PF_SYNPROXY_DROP);
3824 } else if (((th->th_flags & (TH_SYN|TH_ACK)) !=
3826 (ntohl(th->th_ack) != (*state)->dst.seqhi + 1)) {
3827 REASON_SET(reason, PFRES_SYNPROXY);
3830 (*state)->dst.max_win = MAX(ntohs(th->th_win), 1);
3831 (*state)->dst.seqlo = ntohl(th->th_seq);
3832 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
3833 pd->src, th->th_dport, th->th_sport,
3834 ntohl(th->th_ack), ntohl(th->th_seq) + 1,
3835 TH_ACK, (*state)->src.max_win, 0, 0, 0,
3836 (*state)->tag, NULL, NULL);
3837 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
3838 &dst->addr, src->port, dst->port,
3839 (*state)->src.seqhi + 1, (*state)->src.seqlo + 1,
3840 TH_ACK, (*state)->dst.max_win, 0, 0, 1,
3842 (*state)->src.seqdiff = (*state)->dst.seqhi -
3843 (*state)->src.seqlo;
3844 (*state)->dst.seqdiff = (*state)->src.seqhi -
3845 (*state)->dst.seqlo;
3846 (*state)->src.seqhi = (*state)->src.seqlo +
3847 (*state)->dst.max_win;
3848 (*state)->dst.seqhi = (*state)->dst.seqlo +
3849 (*state)->src.max_win;
3850 (*state)->src.wscale = (*state)->dst.wscale = 0;
3851 (*state)->src.state = (*state)->dst.state =
3853 REASON_SET(reason, PFRES_SYNPROXY);
3854 return (PF_SYNPROXY_DROP);
3858 if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) {
3859 sws = src->wscale & PF_WSCALE_MASK;
3860 dws = dst->wscale & PF_WSCALE_MASK;
3865 * Sequence tracking algorithm from Guido van Rooij's paper:
3866 * http://www.madison-gurkha.com/publications/tcp_filtering/
3870 orig_seq = seq = ntohl(th->th_seq);
3871 if (src->seqlo == 0) {
3872 /* First packet from this end. Set its state */
3874 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) &&
3875 src->scrub == NULL) {
3876 if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) {
3877 REASON_SET(reason, PFRES_MEMORY);
3882 /* Deferred generation of sequence number modulator */
3883 if (dst->seqdiff && !src->seqdiff) {
3885 while ((src->seqdiff = pf_new_isn((struct pf_state_key *)&key) - seq) == 0)
3887 ack = ntohl(th->th_ack) - dst->seqdiff;
3888 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3890 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3893 ack = ntohl(th->th_ack);
3896 end = seq + pd->p_len;
3897 if (th->th_flags & TH_SYN) {
3899 (*state)->sync_flags |= PFSTATE_GOT_SYN2;
3900 if (dst->wscale & PF_WSCALE_FLAG) {
3901 src->wscale = pf_get_wscale(m, off, th->th_off,
3903 if (src->wscale & PF_WSCALE_FLAG) {
3904 /* Remove scale factor from initial
3906 sws = src->wscale & PF_WSCALE_MASK;
3907 win = ((u_int32_t)win + (1 << sws) - 1)
3909 dws = dst->wscale & PF_WSCALE_MASK;
3911 /* fixup other window */
3912 dst->max_win <<= dst->wscale &
3914 /* in case of a retrans SYN|ACK */
3919 if (th->th_flags & TH_FIN)
3923 if (src->state < TCPS_SYN_SENT)
3924 src->state = TCPS_SYN_SENT;
3927 * May need to slide the window (seqhi may have been set by
3928 * the crappy stack check or if we picked up the connection
3929 * after establishment)
3931 if (src->seqhi == 1 ||
3932 SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi))
3933 src->seqhi = end + MAX(1, dst->max_win << dws);
3934 if (win > src->max_win)
3938 ack = ntohl(th->th_ack) - dst->seqdiff;
3940 /* Modulate sequence numbers */
3941 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3943 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3946 end = seq + pd->p_len;
3947 if (th->th_flags & TH_SYN)
3949 if (th->th_flags & TH_FIN)
3953 if ((th->th_flags & TH_ACK) == 0) {
3954 /* Let it pass through the ack skew check */
3956 } else if ((ack == 0 &&
3957 (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) ||
3958 /* broken tcp stacks do not set ack */
3959 (dst->state < TCPS_SYN_SENT)) {
3961 * Many stacks (ours included) will set the ACK number in an
3962 * FIN|ACK if the SYN times out -- no sequence to ACK.
3968 /* Ease sequencing restrictions on no data packets */
3973 ackskew = dst->seqlo - ack;
3977 * Need to demodulate the sequence numbers in any TCP SACK options
3978 * (Selective ACK). We could optionally validate the SACK values
3979 * against the current ACK window, either forwards or backwards, but
3980 * I'm not confident that SACK has been implemented properly
3981 * everywhere. It wouldn't surprise me if several stacks accidently
3982 * SACK too far backwards of previously ACKed data. There really aren't
3983 * any security implications of bad SACKing unless the target stack
3984 * doesn't validate the option length correctly. Someone trying to
3985 * spoof into a TCP connection won't bother blindly sending SACK
3988 if (dst->seqdiff && (th->th_off << 2) > sizeof(struct tcphdr)) {
3989 if (pf_modulate_sack(m, off, pd, th, dst))
3994 #define MAXACKWINDOW (0xffff + 1500) /* 1500 is an arbitrary fudge factor */
3995 if (SEQ_GEQ(src->seqhi, end) &&
3996 /* Last octet inside other's window space */
3997 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) &&
3998 /* Retrans: not more than one window back */
3999 (ackskew >= -MAXACKWINDOW) &&
4000 /* Acking not more than one reassembled fragment backwards */
4001 (ackskew <= (MAXACKWINDOW << sws)) &&
4002 /* Acking not more than one window forward */
4003 ((th->th_flags & TH_RST) == 0 || orig_seq == src->seqlo ||
4004 (orig_seq == src->seqlo + 1) || (orig_seq + 1 == src->seqlo) ||
4005 (pd->flags & PFDESC_IP_REAS) == 0)) {
4006 /* Require an exact/+1 sequence match on resets when possible */
4008 if (dst->scrub || src->scrub) {
4009 if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4010 *state, src, dst, ©back))
4014 /* update max window */
4015 if (src->max_win < win)
4017 /* synchronize sequencing */
4018 if (SEQ_GT(end, src->seqlo))
4020 /* slide the window of what the other end can send */
4021 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4022 dst->seqhi = ack + MAX((win << sws), 1);
4026 if (th->th_flags & TH_SYN)
4027 if (src->state < TCPS_SYN_SENT)
4028 src->state = TCPS_SYN_SENT;
4029 if (th->th_flags & TH_FIN)
4030 if (src->state < TCPS_CLOSING)
4031 src->state = TCPS_CLOSING;
4032 if (th->th_flags & TH_ACK) {
4033 if (dst->state == TCPS_SYN_SENT) {
4034 dst->state = TCPS_ESTABLISHED;
4035 if (src->state == TCPS_ESTABLISHED &&
4036 (*state)->src_node != NULL &&
4037 pf_src_connlimit(state)) {
4038 REASON_SET(reason, PFRES_SRCLIMIT);
4041 } else if (dst->state == TCPS_CLOSING)
4042 dst->state = TCPS_FIN_WAIT_2;
4044 if (th->th_flags & TH_RST)
4045 src->state = dst->state = TCPS_TIME_WAIT;
4047 /* update expire time */
4048 (*state)->expire = time_second;
4049 if (src->state >= TCPS_FIN_WAIT_2 &&
4050 dst->state >= TCPS_FIN_WAIT_2)
4051 (*state)->timeout = PFTM_TCP_CLOSED;
4052 else if (src->state >= TCPS_CLOSING &&
4053 dst->state >= TCPS_CLOSING)
4054 (*state)->timeout = PFTM_TCP_FIN_WAIT;
4055 else if (src->state < TCPS_ESTABLISHED ||
4056 dst->state < TCPS_ESTABLISHED)
4057 (*state)->timeout = PFTM_TCP_OPENING;
4058 else if (src->state >= TCPS_CLOSING ||
4059 dst->state >= TCPS_CLOSING)
4060 (*state)->timeout = PFTM_TCP_CLOSING;
4062 (*state)->timeout = PFTM_TCP_ESTABLISHED;
4064 /* Fall through to PASS packet */
4066 } else if ((dst->state < TCPS_SYN_SENT ||
4067 dst->state >= TCPS_FIN_WAIT_2 ||
4068 src->state >= TCPS_FIN_WAIT_2) &&
4069 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) &&
4070 /* Within a window forward of the originating packet */
4071 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) {
4072 /* Within a window backward of the originating packet */
4075 * This currently handles three situations:
4076 * 1) Stupid stacks will shotgun SYNs before their peer
4078 * 2) When PF catches an already established stream (the
4079 * firewall rebooted, the state table was flushed, routes
4081 * 3) Packets get funky immediately after the connection
4082 * closes (this should catch Solaris spurious ACK|FINs
4083 * that web servers like to spew after a close)
4085 * This must be a little more careful than the above code
4086 * since packet floods will also be caught here. We don't
4087 * update the TTL here to mitigate the damage of a packet
4088 * flood and so the same code can handle awkward establishment
4089 * and a loosened connection close.
4090 * In the establishment case, a correct peer response will
4091 * validate the connection, go through the normal state code
4092 * and keep updating the state TTL.
4095 if (pf_status.debug >= PF_DEBUG_MISC) {
4096 kprintf("pf: loose state match: ");
4097 pf_print_state(*state);
4098 pf_print_flags(th->th_flags);
4099 kprintf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
4100 "pkts=%llu:%llu dir=%s,%s\n", seq, orig_seq, ack, pd->p_len,
4101 ackskew, (unsigned long long)(*state)->packets[0],
4102 (unsigned long long)(*state)->packets[1],
4103 direction == PF_IN ? "in" : "out",
4104 direction == (*state)->state_key->direction ?
4108 if (dst->scrub || src->scrub) {
4109 if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4110 *state, src, dst, ©back))
4114 /* update max window */
4115 if (src->max_win < win)
4117 /* synchronize sequencing */
4118 if (SEQ_GT(end, src->seqlo))
4120 /* slide the window of what the other end can send */
4121 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4122 dst->seqhi = ack + MAX((win << sws), 1);
4125 * Cannot set dst->seqhi here since this could be a shotgunned
4126 * SYN and not an already established connection.
4129 if (th->th_flags & TH_FIN)
4130 if (src->state < TCPS_CLOSING)
4131 src->state = TCPS_CLOSING;
4132 if (th->th_flags & TH_RST)
4133 src->state = dst->state = TCPS_TIME_WAIT;
4135 /* Fall through to PASS packet */
4137 } else if ((*state)->pickup_mode == PF_PICKUPS_HASHONLY ||
4138 ((*state)->pickup_mode == PF_PICKUPS_ENABLED &&
4139 ((*state)->sync_flags & PFSTATE_GOT_SYN_MASK) !=
4140 PFSTATE_GOT_SYN_MASK)) {
4142 * If pickup mode is hash only, do not fail on sequence checks.
4144 * If pickup mode is enabled and we did not see the SYN in
4145 * both direction, do not fail on sequence checks because
4146 * we do not have complete information on window scale.
4148 * Adjust expiration and fall through to PASS packet.
4149 * XXX Add a FIN check to reduce timeout?
4151 (*state)->expire = time_second;
4154 * Failure processing
4156 if ((*state)->dst.state == TCPS_SYN_SENT &&
4157 (*state)->src.state == TCPS_SYN_SENT) {
4158 /* Send RST for state mismatches during handshake */
4159 if (!(th->th_flags & TH_RST))
4160 pf_send_tcp((*state)->rule.ptr, pd->af,
4161 pd->dst, pd->src, th->th_dport,
4162 th->th_sport, ntohl(th->th_ack), 0,
4164 (*state)->rule.ptr->return_ttl, 1, 0,
4165 pd->eh, kif->pfik_ifp);
4169 } else if (pf_status.debug >= PF_DEBUG_MISC) {
4170 kprintf("pf: BAD state: ");
4171 pf_print_state(*state);
4172 pf_print_flags(th->th_flags);
4173 kprintf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
4174 "pkts=%llu:%llu dir=%s,%s\n",
4175 seq, orig_seq, ack, pd->p_len, ackskew,
4176 (unsigned long long)(*state)->packets[0],
4177 (unsigned long long)(*state)->packets[1],
4178 direction == PF_IN ? "in" : "out",
4179 direction == (*state)->state_key->direction ?
4181 kprintf("pf: State failure on: %c %c %c %c | %c %c\n",
4182 SEQ_GEQ(src->seqhi, end) ? ' ' : '1',
4183 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ?
4185 (ackskew >= -MAXACKWINDOW) ? ' ' : '3',
4186 (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4',
4187 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5',
4188 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6');
4190 REASON_SET(reason, PFRES_BADSTATE);
4194 /* Any packets which have gotten here are to be passed */
4196 /* translate source/destination address, if necessary */
4197 if (STATE_TRANSLATE((*state)->state_key)) {
4198 if (direction == PF_OUT) {
4199 pf_change_ap(pd->src, &th->th_sport, pd->ip_sum,
4200 &th->th_sum, &(*state)->state_key->gwy.addr,
4201 (*state)->state_key->gwy.port, 0, pd->af);
4204 * If we don't redispatch the packet will go into
4205 * the protocol stack on the wrong cpu for the
4206 * post-translated address.
4208 m->m_pkthdr.fw_flags |= FW_MBUF_REDISPATCH;
4209 pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum,
4210 &th->th_sum, &(*state)->state_key->lan.addr,
4211 (*state)->state_key->lan.port, 0, pd->af);
4213 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4214 } else if (copyback) {
4215 /* Copyback sequence modulation or stateful scrub changes */
4216 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4223 pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif,
4224 struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4226 struct pf_state_peer *src, *dst;
4227 struct pf_state_key_cmp key;
4228 struct udphdr *uh = pd->hdr.udp;
4231 key.proto = IPPROTO_UDP;
4232 if (direction == PF_IN) {
4233 PF_ACPY(&key.ext.addr, pd->src, key.af);
4234 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4235 key.ext.port = uh->uh_sport;
4236 key.gwy.port = uh->uh_dport;
4238 PF_ACPY(&key.lan.addr, pd->src, key.af);
4239 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4240 key.lan.port = uh->uh_sport;
4241 key.ext.port = uh->uh_dport;
4246 if (direction == (*state)->state_key->direction) {
4247 src = &(*state)->src;
4248 dst = &(*state)->dst;
4250 src = &(*state)->dst;
4251 dst = &(*state)->src;
4255 if (src->state < PFUDPS_SINGLE)
4256 src->state = PFUDPS_SINGLE;
4257 if (dst->state == PFUDPS_SINGLE)
4258 dst->state = PFUDPS_MULTIPLE;
4260 /* update expire time */
4261 (*state)->expire = time_second;
4262 if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE)
4263 (*state)->timeout = PFTM_UDP_MULTIPLE;
4265 (*state)->timeout = PFTM_UDP_SINGLE;
4267 /* translate source/destination address, if necessary */
4268 if (STATE_TRANSLATE((*state)->state_key)) {
4269 if (direction == PF_OUT) {
4270 pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum,
4271 &uh->uh_sum, &(*state)->state_key->gwy.addr,
4272 (*state)->state_key->gwy.port, 1, pd->af);
4275 * If we don't redispatch the packet will go into
4276 * the protocol stack on the wrong cpu for the
4277 * post-translated address.
4279 m->m_pkthdr.fw_flags |= FW_MBUF_REDISPATCH;
4280 pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum,
4281 &uh->uh_sum, &(*state)->state_key->lan.addr,
4282 (*state)->state_key->lan.port, 1, pd->af);
4284 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
4291 pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif,
4292 struct mbuf *m, int off, void *h, struct pf_pdesc *pd, u_short *reason)
4294 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
4295 u_int16_t icmpid = 0, *icmpsum;
4298 struct pf_state_key_cmp key;
4300 switch (pd->proto) {
4303 icmptype = pd->hdr.icmp->icmp_type;
4304 icmpid = pd->hdr.icmp->icmp_id;
4305 icmpsum = &pd->hdr.icmp->icmp_cksum;
4307 if (icmptype == ICMP_UNREACH ||
4308 icmptype == ICMP_SOURCEQUENCH ||
4309 icmptype == ICMP_REDIRECT ||
4310 icmptype == ICMP_TIMXCEED ||
4311 icmptype == ICMP_PARAMPROB)
4316 case IPPROTO_ICMPV6:
4317 icmptype = pd->hdr.icmp6->icmp6_type;
4318 icmpid = pd->hdr.icmp6->icmp6_id;
4319 icmpsum = &pd->hdr.icmp6->icmp6_cksum;
4321 if (icmptype == ICMP6_DST_UNREACH ||
4322 icmptype == ICMP6_PACKET_TOO_BIG ||
4323 icmptype == ICMP6_TIME_EXCEEDED ||
4324 icmptype == ICMP6_PARAM_PROB)
4333 * ICMP query/reply message not related to a TCP/UDP packet.
4334 * Search for an ICMP state.
4337 key.proto = pd->proto;
4338 if (direction == PF_IN) {
4339 PF_ACPY(&key.ext.addr, pd->src, key.af);
4340 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4342 key.gwy.port = icmpid;
4344 PF_ACPY(&key.lan.addr, pd->src, key.af);
4345 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4346 key.lan.port = icmpid;
4352 (*state)->expire = time_second;
4353 (*state)->timeout = PFTM_ICMP_ERROR_REPLY;
4355 /* translate source/destination address, if necessary */
4356 if (STATE_TRANSLATE((*state)->state_key)) {
4357 if (direction == PF_OUT) {
4361 pf_change_a(&saddr->v4.s_addr,
4363 (*state)->state_key->gwy.addr.v4.s_addr, 0);
4364 pd->hdr.icmp->icmp_cksum =
4366 pd->hdr.icmp->icmp_cksum, icmpid,
4367 (*state)->state_key->gwy.port, 0);
4368 pd->hdr.icmp->icmp_id =
4369 (*state)->state_key->gwy.port;
4370 m_copyback(m, off, ICMP_MINLEN,
4371 (caddr_t)pd->hdr.icmp);
4377 &pd->hdr.icmp6->icmp6_cksum,
4378 &(*state)->state_key->gwy.addr, 0);
4380 sizeof(struct icmp6_hdr),
4381 (caddr_t)pd->hdr.icmp6);
4389 pf_change_a(&daddr->v4.s_addr,
4391 (*state)->state_key->lan.addr.v4.s_addr, 0);
4392 pd->hdr.icmp->icmp_cksum =
4394 pd->hdr.icmp->icmp_cksum, icmpid,
4395 (*state)->state_key->lan.port, 0);
4396 pd->hdr.icmp->icmp_id =
4397 (*state)->state_key->lan.port;
4398 m_copyback(m, off, ICMP_MINLEN,
4399 (caddr_t)pd->hdr.icmp);
4405 &pd->hdr.icmp6->icmp6_cksum,
4406 &(*state)->state_key->lan.addr, 0);
4408 sizeof(struct icmp6_hdr),
4409 (caddr_t)pd->hdr.icmp6);
4420 * ICMP error message in response to a TCP/UDP packet.
4421 * Extract the inner TCP/UDP header and search for that state.
4424 struct pf_pdesc pd2;
4429 struct ip6_hdr h2_6;
4439 /* offset of h2 in mbuf chain */
4440 ipoff2 = off + ICMP_MINLEN;
4442 if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2),
4443 NULL, reason, pd2.af)) {
4444 DPFPRINTF(PF_DEBUG_MISC,
4445 ("pf: ICMP error message too short "
4450 * ICMP error messages don't refer to non-first
4453 if (h2.ip_off & htons(IP_OFFMASK)) {
4454 REASON_SET(reason, PFRES_FRAG);
4458 /* offset of protocol header that follows h2 */
4459 off2 = ipoff2 + (h2.ip_hl << 2);
4461 pd2.proto = h2.ip_p;
4462 pd2.src = (struct pf_addr *)&h2.ip_src;
4463 pd2.dst = (struct pf_addr *)&h2.ip_dst;
4464 pd2.ip_sum = &h2.ip_sum;
4469 ipoff2 = off + sizeof(struct icmp6_hdr);
4471 if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6),
4472 NULL, reason, pd2.af)) {
4473 DPFPRINTF(PF_DEBUG_MISC,
4474 ("pf: ICMP error message too short "
4478 pd2.proto = h2_6.ip6_nxt;
4479 pd2.src = (struct pf_addr *)&h2_6.ip6_src;
4480 pd2.dst = (struct pf_addr *)&h2_6.ip6_dst;
4482 off2 = ipoff2 + sizeof(h2_6);
4484 switch (pd2.proto) {
4485 case IPPROTO_FRAGMENT:
4487 * ICMPv6 error messages for
4488 * non-first fragments
4490 REASON_SET(reason, PFRES_FRAG);
4493 case IPPROTO_HOPOPTS:
4494 case IPPROTO_ROUTING:
4495 case IPPROTO_DSTOPTS: {
4496 /* get next header and header length */
4497 struct ip6_ext opt6;
4499 if (!pf_pull_hdr(m, off2, &opt6,
4500 sizeof(opt6), NULL, reason,
4502 DPFPRINTF(PF_DEBUG_MISC,
4503 ("pf: ICMPv6 short opt\n"));
4506 if (pd2.proto == IPPROTO_AH)
4507 off2 += (opt6.ip6e_len + 2) * 4;
4509 off2 += (opt6.ip6e_len + 1) * 8;
4510 pd2.proto = opt6.ip6e_nxt;
4511 /* goto the next header */
4518 } while (!terminal);
4522 DPFPRINTF(PF_DEBUG_MISC,
4523 ("pf: ICMP AF %d unknown (ip6)\n", pd->af));
4528 switch (pd2.proto) {
4532 struct pf_state_peer *src, *dst;
4537 * Only the first 8 bytes of the TCP header can be
4538 * expected. Don't access any TCP header fields after
4539 * th_seq, an ackskew test is not possible.
4541 if (!pf_pull_hdr(m, off2, &th, 8, NULL, reason,
4543 DPFPRINTF(PF_DEBUG_MISC,
4544 ("pf: ICMP error message too short "
4550 key.proto = IPPROTO_TCP;
4551 if (direction == PF_IN) {
4552 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4553 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4554 key.ext.port = th.th_dport;
4555 key.gwy.port = th.th_sport;
4557 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4558 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4559 key.lan.port = th.th_dport;
4560 key.ext.port = th.th_sport;
4565 if (direction == (*state)->state_key->direction) {
4566 src = &(*state)->dst;
4567 dst = &(*state)->src;
4569 src = &(*state)->src;
4570 dst = &(*state)->dst;
4573 if (src->wscale && dst->wscale)
4574 dws = dst->wscale & PF_WSCALE_MASK;
4578 /* Demodulate sequence number */
4579 seq = ntohl(th.th_seq) - src->seqdiff;
4581 pf_change_a(&th.th_seq, icmpsum,
4586 if (!SEQ_GEQ(src->seqhi, seq) ||
4587 !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws))) {
4588 if (pf_status.debug >= PF_DEBUG_MISC) {
4589 kprintf("pf: BAD ICMP %d:%d ",
4590 icmptype, pd->hdr.icmp->icmp_code);
4591 pf_print_host(pd->src, 0, pd->af);
4593 pf_print_host(pd->dst, 0, pd->af);
4594 kprintf(" state: ");
4595 pf_print_state(*state);
4596 kprintf(" seq=%u\n", seq);
4598 REASON_SET(reason, PFRES_BADSTATE);
4602 if (STATE_TRANSLATE((*state)->state_key)) {
4603 if (direction == PF_IN) {
4604 pf_change_icmp(pd2.src, &th.th_sport,
4605 daddr, &(*state)->state_key->lan.addr,
4606 (*state)->state_key->lan.port, NULL,
4607 pd2.ip_sum, icmpsum,
4608 pd->ip_sum, 0, pd2.af);
4610 pf_change_icmp(pd2.dst, &th.th_dport,
4611 saddr, &(*state)->state_key->gwy.addr,
4612 (*state)->state_key->gwy.port, NULL,
4613 pd2.ip_sum, icmpsum,
4614 pd->ip_sum, 0, pd2.af);
4623 m_copyback(m, off, ICMP_MINLEN,
4624 (caddr_t)pd->hdr.icmp);
4625 m_copyback(m, ipoff2, sizeof(h2),
4632 sizeof(struct icmp6_hdr),
4633 (caddr_t)pd->hdr.icmp6);
4634 m_copyback(m, ipoff2, sizeof(h2_6),
4639 m_copyback(m, off2, 8, (caddr_t)&th);
4648 if (!pf_pull_hdr(m, off2, &uh, sizeof(uh),
4649 NULL, reason, pd2.af)) {
4650 DPFPRINTF(PF_DEBUG_MISC,
4651 ("pf: ICMP error message too short "
4657 key.proto = IPPROTO_UDP;
4658 if (direction == PF_IN) {
4659 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4660 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4661 key.ext.port = uh.uh_dport;
4662 key.gwy.port = uh.uh_sport;
4664 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4665 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4666 key.lan.port = uh.uh_dport;
4667 key.ext.port = uh.uh_sport;
4672 if (STATE_TRANSLATE((*state)->state_key)) {
4673 if (direction == PF_IN) {
4674 pf_change_icmp(pd2.src, &uh.uh_sport,
4676 &(*state)->state_key->lan.addr,
4677 (*state)->state_key->lan.port,
4679 pd2.ip_sum, icmpsum,
4680 pd->ip_sum, 1, pd2.af);
4682 pf_change_icmp(pd2.dst, &uh.uh_dport,
4684 &(*state)->state_key->gwy.addr,
4685 (*state)->state_key->gwy.port, &uh.uh_sum,
4686 pd2.ip_sum, icmpsum,
4687 pd->ip_sum, 1, pd2.af);
4692 m_copyback(m, off, ICMP_MINLEN,
4693 (caddr_t)pd->hdr.icmp);
4694 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4700 sizeof(struct icmp6_hdr),
4701 (caddr_t)pd->hdr.icmp6);
4702 m_copyback(m, ipoff2, sizeof(h2_6),
4707 m_copyback(m, off2, sizeof(uh), (caddr_t)&uh);
4714 case IPPROTO_ICMP: {
4717 if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN,
4718 NULL, reason, pd2.af)) {
4719 DPFPRINTF(PF_DEBUG_MISC,
4720 ("pf: ICMP error message too short i"
4726 key.proto = IPPROTO_ICMP;
4727 if (direction == PF_IN) {
4728 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4729 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4731 key.gwy.port = iih.icmp_id;
4733 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4734 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4735 key.lan.port = iih.icmp_id;
4741 if (STATE_TRANSLATE((*state)->state_key)) {
4742 if (direction == PF_IN) {
4743 pf_change_icmp(pd2.src, &iih.icmp_id,
4745 &(*state)->state_key->lan.addr,
4746 (*state)->state_key->lan.port, NULL,
4747 pd2.ip_sum, icmpsum,
4748 pd->ip_sum, 0, AF_INET);
4750 pf_change_icmp(pd2.dst, &iih.icmp_id,
4752 &(*state)->state_key->gwy.addr,
4753 (*state)->state_key->gwy.port, NULL,
4754 pd2.ip_sum, icmpsum,
4755 pd->ip_sum, 0, AF_INET);
4757 m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
4758 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4759 m_copyback(m, off2, ICMP_MINLEN, (caddr_t)&iih);
4767 case IPPROTO_ICMPV6: {
4768 struct icmp6_hdr iih;
4770 if (!pf_pull_hdr(m, off2, &iih,
4771 sizeof(struct icmp6_hdr), NULL, reason, pd2.af)) {
4772 DPFPRINTF(PF_DEBUG_MISC,
4773 ("pf: ICMP error message too short "
4779 key.proto = IPPROTO_ICMPV6;
4780 if (direction == PF_IN) {
4781 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4782 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4784 key.gwy.port = iih.icmp6_id;
4786 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4787 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4788 key.lan.port = iih.icmp6_id;
4794 if (STATE_TRANSLATE((*state)->state_key)) {
4795 if (direction == PF_IN) {
4796 pf_change_icmp(pd2.src, &iih.icmp6_id,
4798 &(*state)->state_key->lan.addr,
4799 (*state)->state_key->lan.port, NULL,
4800 pd2.ip_sum, icmpsum,
4801 pd->ip_sum, 0, AF_INET6);
4803 pf_change_icmp(pd2.dst, &iih.icmp6_id,
4804 saddr, &(*state)->state_key->gwy.addr,
4805 (*state)->state_key->gwy.port, NULL,
4806 pd2.ip_sum, icmpsum,
4807 pd->ip_sum, 0, AF_INET6);
4809 m_copyback(m, off, sizeof(struct icmp6_hdr),
4810 (caddr_t)pd->hdr.icmp6);
4811 m_copyback(m, ipoff2, sizeof(h2_6), (caddr_t)&h2_6);
4812 m_copyback(m, off2, sizeof(struct icmp6_hdr),
4822 key.proto = pd2.proto;
4823 if (direction == PF_IN) {
4824 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4825 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4829 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4830 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4837 if (STATE_TRANSLATE((*state)->state_key)) {
4838 if (direction == PF_IN) {
4839 pf_change_icmp(pd2.src, NULL,
4841 &(*state)->state_key->lan.addr,
4843 pd2.ip_sum, icmpsum,
4844 pd->ip_sum, 0, pd2.af);
4846 pf_change_icmp(pd2.dst, NULL,
4848 &(*state)->state_key->gwy.addr,
4850 pd2.ip_sum, icmpsum,
4851 pd->ip_sum, 0, pd2.af);
4856 m_copyback(m, off, ICMP_MINLEN,
4857 (caddr_t)pd->hdr.icmp);
4858 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4864 sizeof(struct icmp6_hdr),
4865 (caddr_t)pd->hdr.icmp6);
4866 m_copyback(m, ipoff2, sizeof(h2_6),
4881 pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif,
4882 struct pf_pdesc *pd)
4884 struct pf_state_peer *src, *dst;
4885 struct pf_state_key_cmp key;
4888 key.proto = pd->proto;
4889 if (direction == PF_IN) {
4890 PF_ACPY(&key.ext.addr, pd->src, key.af);
4891 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4895 PF_ACPY(&key.lan.addr, pd->src, key.af);
4896 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4903 if (direction == (*state)->state_key->direction) {
4904 src = &(*state)->src;
4905 dst = &(*state)->dst;
4907 src = &(*state)->dst;
4908 dst = &(*state)->src;
4912 if (src->state < PFOTHERS_SINGLE)
4913 src->state = PFOTHERS_SINGLE;
4914 if (dst->state == PFOTHERS_SINGLE)
4915 dst->state = PFOTHERS_MULTIPLE;
4917 /* update expire time */
4918 (*state)->expire = time_second;
4919 if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE)
4920 (*state)->timeout = PFTM_OTHER_MULTIPLE;
4922 (*state)->timeout = PFTM_OTHER_SINGLE;
4924 /* translate source/destination address, if necessary */
4925 if (STATE_TRANSLATE((*state)->state_key)) {
4926 if (direction == PF_OUT)
4930 pf_change_a(&pd->src->v4.s_addr,
4932 (*state)->state_key->gwy.addr.v4.s_addr,
4939 &(*state)->state_key->gwy.addr, pd->af);
4947 pf_change_a(&pd->dst->v4.s_addr,
4949 (*state)->state_key->lan.addr.v4.s_addr,
4956 &(*state)->state_key->lan.addr, pd->af);
4966 * ipoff and off are measured from the start of the mbuf chain.
4967 * h must be at "ipoff" on the mbuf chain.
4970 pf_pull_hdr(struct mbuf *m, int off, void *p, int len,
4971 u_short *actionp, u_short *reasonp, sa_family_t af)
4976 struct ip *h = mtod(m, struct ip *);
4977 u_int16_t fragoff = (h->ip_off & IP_OFFMASK) << 3;
4981 ACTION_SET(actionp, PF_PASS);
4983 ACTION_SET(actionp, PF_DROP);
4984 REASON_SET(reasonp, PFRES_FRAG);
4988 if (m->m_pkthdr.len < off + len ||
4989 h->ip_len < off + len) {
4990 ACTION_SET(actionp, PF_DROP);
4991 REASON_SET(reasonp, PFRES_SHORT);
4999 struct ip6_hdr *h = mtod(m, struct ip6_hdr *);
5001 if (m->m_pkthdr.len < off + len ||
5002 (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) <
5003 (unsigned)(off + len)) {
5004 ACTION_SET(actionp, PF_DROP);
5005 REASON_SET(reasonp, PFRES_SHORT);
5012 m_copydata(m, off, len, p);
5017 pf_routable(struct pf_addr *addr, sa_family_t af, struct pfi_kif *kif)
5019 struct sockaddr_in *dst;
5023 struct sockaddr_in6 *dst6;
5024 struct route_in6 ro;
5028 struct radix_node *rn;
5033 bzero(&ro, sizeof(ro));
5036 dst = satosin(&ro.ro_dst);
5037 dst->sin_family = AF_INET;
5038 dst->sin_len = sizeof(*dst);
5039 dst->sin_addr = addr->v4;
5043 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
5044 dst6->sin6_family = AF_INET6;
5045 dst6->sin6_len = sizeof(*dst6);
5046 dst6->sin6_addr = addr->v6;
5053 /* Skip checks for ipsec interfaces */
5054 if (kif != NULL && kif->pfik_ifp->if_type == IFT_ENC)
5057 rtalloc_ign((struct route *)&ro, 0);
5059 if (ro.ro_rt != NULL) {
5060 /* No interface given, this is a no-route check */
5064 if (kif->pfik_ifp == NULL) {
5069 /* Perform uRPF check if passed input interface */
5071 rn = (struct radix_node *)ro.ro_rt;
5073 rt = (struct rtentry *)rn;
5076 if (kif->pfik_ifp == ifp)
5079 } while (check_mpath == 1 && rn != NULL && ret == 0);
5083 if (ro.ro_rt != NULL)
5089 pf_rtlabel_match(struct pf_addr *addr, sa_family_t af, struct pf_addr_wrap *aw)
5091 struct sockaddr_in *dst;
5093 struct sockaddr_in6 *dst6;
5094 struct route_in6 ro;
5100 bzero(&ro, sizeof(ro));
5103 dst = satosin(&ro.ro_dst);
5104 dst->sin_family = AF_INET;
5105 dst->sin_len = sizeof(*dst);
5106 dst->sin_addr = addr->v4;
5110 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
5111 dst6->sin6_family = AF_INET6;
5112 dst6->sin6_len = sizeof(*dst6);
5113 dst6->sin6_addr = addr->v6;
5120 rtalloc_ign((struct route *)&ro, (RTF_CLONING | RTF_PRCLONING));
5122 if (ro.ro_rt != NULL) {
5131 pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5132 struct pf_state *s, struct pf_pdesc *pd)
5134 struct mbuf *m0, *m1;
5135 struct route iproute;
5136 struct route *ro = NULL;
5137 struct sockaddr_in *dst;
5139 struct ifnet *ifp = NULL;
5140 struct pf_addr naddr;
5141 struct pf_src_node *sn = NULL;
5148 if (m == NULL || *m == NULL || r == NULL ||
5149 (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5150 panic("pf_route: invalid parameters");
5152 if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
5153 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
5154 (*m)->m_pkthdr.pf.routed = 1;
5156 if ((*m)->m_pkthdr.pf.routed++ > 3) {
5163 if (r->rt == PF_DUPTO) {
5164 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL)
5167 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5172 if (m0->m_len < sizeof(struct ip)) {
5173 DPFPRINTF(PF_DEBUG_URGENT,
5174 ("pf_route: m0->m_len < sizeof(struct ip)\n"));
5178 ip = mtod(m0, struct ip *);
5181 bzero((caddr_t)ro, sizeof(*ro));
5182 dst = satosin(&ro->ro_dst);
5183 dst->sin_family = AF_INET;
5184 dst->sin_len = sizeof(*dst);
5185 dst->sin_addr = ip->ip_dst;
5187 if (r->rt == PF_FASTROUTE) {
5189 if (ro->ro_rt == 0) {
5190 ipstat.ips_noroute++;
5194 ifp = ro->ro_rt->rt_ifp;
5195 ro->ro_rt->rt_use++;
5197 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
5198 dst = satosin(ro->ro_rt->rt_gateway);
5200 if (TAILQ_EMPTY(&r->rpool.list)) {
5201 DPFPRINTF(PF_DEBUG_URGENT,
5202 ("pf_route: TAILQ_EMPTY(&r->rpool.list)\n"));
5206 pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src,
5208 if (!PF_AZERO(&naddr, AF_INET))
5209 dst->sin_addr.s_addr = naddr.v4.s_addr;
5210 ifp = r->rpool.cur->kif ?
5211 r->rpool.cur->kif->pfik_ifp : NULL;
5213 if (!PF_AZERO(&s->rt_addr, AF_INET))
5214 dst->sin_addr.s_addr =
5215 s->rt_addr.v4.s_addr;
5216 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5224 if (pf_test(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) {
5227 } else if (m0 == NULL) {
5232 if (m0->m_len < sizeof(struct ip)) {
5233 DPFPRINTF(PF_DEBUG_URGENT,
5234 ("pf_route: m0->m_len < sizeof(struct ip)\n"));
5237 ip = mtod(m0, struct ip *);
5240 /* Copied from FreeBSD 5.1-CURRENT ip_output. */
5241 m0->m_pkthdr.csum_flags |= CSUM_IP;
5242 sw_csum = m0->m_pkthdr.csum_flags & ~ifp->if_hwassist;
5243 if (sw_csum & CSUM_DELAY_DATA) {
5244 in_delayed_cksum(m0);
5245 sw_csum &= ~CSUM_DELAY_DATA;
5247 m0->m_pkthdr.csum_flags &= ifp->if_hwassist;
5249 if (ip->ip_len <= ifp->if_mtu ||
5250 (ifp->if_hwassist & CSUM_FRAGMENT &&
5251 (ip->ip_off & IP_DF) == 0)) {
5252 ip->ip_len = htons(ip->ip_len);
5253 ip->ip_off = htons(ip->ip_off);
5255 if (sw_csum & CSUM_DELAY_IP) {
5257 if (ip->ip_v == IPVERSION &&
5258 (ip->ip_hl << 2) == sizeof(*ip)) {
5259 ip->ip_sum = in_cksum_hdr(ip);
5261 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
5265 error = ifp->if_output(ifp, m0, sintosa(dst), ro->ro_rt);
5271 * Too large for interface; fragment if possible.
5272 * Must be able to put at least 8 bytes per fragment.
5274 if (ip->ip_off & IP_DF) {
5275 ipstat.ips_cantfrag++;
5276 if (r->rt != PF_DUPTO) {
5278 icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0,
5287 error = ip_fragment(ip, &m0, ifp->if_mtu, ifp->if_hwassist, sw_csum);
5292 for (m0 = m1; m0; m0 = m1) {
5297 error = (*ifp->if_output)(ifp, m0, sintosa(dst),
5305 ipstat.ips_fragmented++;
5308 if (r->rt != PF_DUPTO)
5310 if (ro == &iproute && ro->ro_rt)
5322 pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5323 struct pf_state *s, struct pf_pdesc *pd)
5326 struct route_in6 ip6route;
5327 struct route_in6 *ro;
5328 struct sockaddr_in6 *dst;
5329 struct ip6_hdr *ip6;
5330 struct ifnet *ifp = NULL;
5331 struct pf_addr naddr;
5332 struct pf_src_node *sn = NULL;
5335 if (m == NULL || *m == NULL || r == NULL ||
5336 (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5337 panic("pf_route6: invalid parameters");
5339 if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
5340 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
5341 (*m)->m_pkthdr.pf.routed = 1;
5343 if ((*m)->m_pkthdr.pf.routed++ > 3) {
5350 if (r->rt == PF_DUPTO) {
5351 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL)
5354 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5359 if (m0->m_len < sizeof(struct ip6_hdr)) {
5360 DPFPRINTF(PF_DEBUG_URGENT,
5361 ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n"));
5364 ip6 = mtod(m0, struct ip6_hdr *);
5367 bzero((caddr_t)ro, sizeof(*ro));
5368 dst = (struct sockaddr_in6 *)&ro->ro_dst;
5369 dst->sin6_family = AF_INET6;
5370 dst->sin6_len = sizeof(*dst);
5371 dst->sin6_addr = ip6->ip6_dst;
5373 /* Cheat. XXX why only in the v6 case??? */
5374 if (r->rt == PF_FASTROUTE) {
5375 m0->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
5376 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
5381 if (TAILQ_EMPTY(&r->rpool.list)) {
5382 DPFPRINTF(PF_DEBUG_URGENT,
5383 ("pf_route6: TAILQ_EMPTY(&r->rpool.list)\n"));
5387 pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src,
5389 if (!PF_AZERO(&naddr, AF_INET6))
5390 PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5392 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL;
5394 if (!PF_AZERO(&s->rt_addr, AF_INET6))
5395 PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5396 &s->rt_addr, AF_INET6);
5397 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5404 if (pf_test6(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) {
5407 } else if (m0 == NULL) {
5412 if (m0->m_len < sizeof(struct ip6_hdr)) {
5413 DPFPRINTF(PF_DEBUG_URGENT,
5414 ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n"));
5417 ip6 = mtod(m0, struct ip6_hdr *);
5421 * If the packet is too large for the outgoing interface,
5422 * send back an icmp6 error.
5424 if (IN6_IS_ADDR_LINKLOCAL(&dst->sin6_addr))
5425 dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
5426 if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) {
5428 error = nd6_output(ifp, ifp, m0, dst, NULL);
5431 in6_ifstat_inc(ifp, ifs6_in_toobig);
5432 if (r->rt != PF_DUPTO) {
5434 icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
5441 if (r->rt != PF_DUPTO)
5454 * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag
5455 * off is the offset where the protocol header starts
5456 * len is the total length of protocol header plus payload
5457 * returns 0 when the checksum is valid, otherwise returns 1.
5461 * FreeBSD supports cksum offload for the following drivers.
5462 * em(4), gx(4), lge(4), nge(4), ti(4), xl(4)
5463 * If we can make full use of it we would outperform ipfw/ipfilter in
5464 * very heavy traffic.
5465 * I have not tested 'cause I don't have NICs that supports cksum offload.
5466 * (There might be problems. Typical phenomena would be
5467 * 1. No route message for UDP packet.
5468 * 2. No connection acceptance from external hosts regardless of rule set.)
5471 pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p,
5478 if (off < sizeof(struct ip) || len < sizeof(struct udphdr))
5480 if (m->m_pkthdr.len < off + len)
5486 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
5487 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) {
5488 sum = m->m_pkthdr.csum_data;
5490 ip = mtod(m, struct ip *);
5491 sum = in_pseudo(ip->ip_src.s_addr,
5492 ip->ip_dst.s_addr, htonl((u_short)len +
5493 m->m_pkthdr.csum_data + p));
5501 case IPPROTO_ICMPV6:
5511 if (p == IPPROTO_ICMP) {
5516 sum = in_cksum(m, len);
5520 if (m->m_len < sizeof(struct ip))
5522 sum = in_cksum_range(m, p, off, len);
5524 m->m_pkthdr.csum_flags |=
5527 m->m_pkthdr.csum_data = 0xffff;
5533 if (m->m_len < sizeof(struct ip6_hdr))
5535 sum = in6_cksum(m, p, off, len);
5538 * IPv6 H/W cksum off-load not supported yet!
5541 * m->m_pkthdr.csum_flags |=
5542 * (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
5543 * m->m_pkthdr.csum_data = 0xffff;
5555 tcpstat.tcps_rcvbadsum++;
5558 udpstat.udps_badsum++;
5561 icmpstat.icps_checksum++;
5564 case IPPROTO_ICMPV6:
5565 icmp6stat.icp6s_checksum++;
5576 pf_test(int dir, struct ifnet *ifp, struct mbuf **m0,
5577 struct ether_header *eh, struct inpcb *inp)
5579 struct pfi_kif *kif;
5580 u_short action, reason = 0, log = 0;
5581 struct mbuf *m = *m0;
5582 struct ip *h = NULL;
5583 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr;
5584 struct pf_state *s = NULL;
5585 struct pf_state_key *sk = NULL;
5586 struct pf_ruleset *ruleset = NULL;
5588 int off, dirndx, pqid = 0;
5590 if (!pf_status.running)
5593 memset(&pd, 0, sizeof(pd));
5594 if (ifp->if_type == IFT_CARP && ifp->if_carpdev)
5595 kif = (struct pfi_kif *)ifp->if_carpdev->if_pf_kif;
5597 kif = (struct pfi_kif *)ifp->if_pf_kif;
5600 DPFPRINTF(PF_DEBUG_URGENT,
5601 ("pf_test: kif == NULL, if_xname %s\n", ifp->if_xname));
5604 if (kif->pfik_flags & PFI_IFLAG_SKIP)
5608 if ((m->m_flags & M_PKTHDR) == 0)
5609 panic("non-M_PKTHDR is passed to pf_test");
5610 #endif /* DIAGNOSTIC */
5612 if (m->m_pkthdr.len < (int)sizeof(*h)) {
5614 REASON_SET(&reason, PFRES_SHORT);
5619 if (m->m_pkthdr.pf.flags & PF_TAG_GENERATED)
5622 /* We do IP header normalization and packet reassembly here */
5623 if (pf_normalize_ip(m0, dir, kif, &reason, &pd) != PF_PASS) {
5627 m = *m0; /* pf_normalize messes with m0 */
5628 h = mtod(m, struct ip *);
5630 off = h->ip_hl << 2;
5631 if (off < (int)sizeof(*h)) {
5633 REASON_SET(&reason, PFRES_SHORT);
5638 pd.src = (struct pf_addr *)&h->ip_src;
5639 pd.dst = (struct pf_addr *)&h->ip_dst;
5640 PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET);
5641 pd.ip_sum = &h->ip_sum;
5645 pd.tot_len = h->ip_len;
5648 /* handle fragments that didn't get reassembled by normalization */
5649 if (h->ip_off & (IP_MF | IP_OFFMASK)) {
5650 action = pf_test_fragment(&r, dir, kif, m, h,
5661 if (!pf_pull_hdr(m, off, &th, sizeof(th),
5662 &action, &reason, AF_INET)) {
5663 log = action != PF_PASS;
5666 pd.p_len = pd.tot_len - off - (th.th_off << 2);
5667 if ((th.th_flags & TH_ACK) && pd.p_len == 0)
5669 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
5670 if (action == PF_DROP)
5672 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
5674 if (action == PF_PASS) {
5676 pfsync_update_state(s);
5677 #endif /* NPFSYNC */
5681 } else if (s == NULL)
5682 action = pf_test_rule(&r, &s, dir, kif,
5683 m, off, h, &pd, &a, &ruleset, NULL, inp);
5691 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
5692 &action, &reason, AF_INET)) {
5693 log = action != PF_PASS;
5696 if (uh.uh_dport == 0 ||
5697 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
5698 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
5700 REASON_SET(&reason, PFRES_SHORT);
5703 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
5704 if (action == PF_PASS) {
5706 pfsync_update_state(s);
5707 #endif /* NPFSYNC */
5711 } else if (s == NULL)
5712 action = pf_test_rule(&r, &s, dir, kif,
5713 m, off, h, &pd, &a, &ruleset, NULL, inp);
5717 case IPPROTO_ICMP: {
5721 if (!pf_pull_hdr(m, off, &ih, ICMP_MINLEN,
5722 &action, &reason, AF_INET)) {
5723 log = action != PF_PASS;
5726 action = pf_test_state_icmp(&s, dir, kif, m, off, h, &pd,
5728 if (action == PF_PASS) {
5730 pfsync_update_state(s);
5731 #endif /* NPFSYNC */
5735 } else if (s == NULL)
5736 action = pf_test_rule(&r, &s, dir, kif,
5737 m, off, h, &pd, &a, &ruleset, NULL, inp);
5742 action = pf_test_state_other(&s, dir, kif, &pd);
5743 if (action == PF_PASS) {
5745 pfsync_update_state(s);
5746 #endif /* NPFSYNC */
5750 } else if (s == NULL)
5751 action = pf_test_rule(&r, &s, dir, kif, m, off, h,
5752 &pd, &a, &ruleset, NULL, inp);
5757 if (action == PF_PASS && h->ip_hl > 5 &&
5758 !((s && s->allow_opts) || r->allow_opts)) {
5760 REASON_SET(&reason, PFRES_IPOPTIONS);
5762 DPFPRINTF(PF_DEBUG_MISC,
5763 ("pf: dropping packet with ip options\n"));
5766 if ((s && s->tag) || r->rtableid)
5767 pf_tag_packet(m, s ? s->tag : 0, r->rtableid);
5770 if (action == PF_PASS && r->qid) {
5771 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
5772 if (pqid || (pd.tos & IPTOS_LOWDELAY))
5773 m->m_pkthdr.pf.qid = r->pqid;
5775 m->m_pkthdr.pf.qid = r->qid;
5776 m->m_pkthdr.pf.ecn_af = AF_INET;
5777 m->m_pkthdr.pf.hdr = h;
5778 /* add connection hash for fairq */
5781 m->m_pkthdr.pf.state_hash = s->hash;
5782 m->m_pkthdr.pf.flags |= PF_TAG_STATE_HASHED;
5788 * connections redirected to loopback should not match sockets
5789 * bound specifically to loopback due to security implications,
5790 * see tcp_input() and in_pcblookup_listen().
5792 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
5793 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
5794 (s->nat_rule.ptr->action == PF_RDR ||
5795 s->nat_rule.ptr->action == PF_BINAT) &&
5796 (ntohl(pd.dst->v4.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET)
5797 m->m_pkthdr.pf.flags |= PF_TAG_TRANSLATE_LOCALHOST;
5802 if (s != NULL && s->nat_rule.ptr != NULL &&
5803 s->nat_rule.ptr->log & PF_LOG_ALL)
5804 lr = s->nat_rule.ptr;
5807 PFLOG_PACKET(kif, h, m, AF_INET, dir, reason, lr, a, ruleset,
5811 kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
5812 kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++;
5814 if (action == PF_PASS || r->action == PF_DROP) {
5815 dirndx = (dir == PF_OUT);
5816 r->packets[dirndx]++;
5817 r->bytes[dirndx] += pd.tot_len;
5819 a->packets[dirndx]++;
5820 a->bytes[dirndx] += pd.tot_len;
5824 if (s->nat_rule.ptr != NULL) {
5825 s->nat_rule.ptr->packets[dirndx]++;
5826 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len;
5828 if (s->src_node != NULL) {
5829 s->src_node->packets[dirndx]++;
5830 s->src_node->bytes[dirndx] += pd.tot_len;
5832 if (s->nat_src_node != NULL) {
5833 s->nat_src_node->packets[dirndx]++;
5834 s->nat_src_node->bytes[dirndx] += pd.tot_len;
5836 dirndx = (dir == sk->direction) ? 0 : 1;
5837 s->packets[dirndx]++;
5838 s->bytes[dirndx] += pd.tot_len;
5841 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
5845 * XXX: we need to make sure that the addresses
5846 * passed to pfr_update_stats() are the same than
5847 * the addresses used during matching (pfr_match)
5849 if (r == &pf_default_rule) {
5851 x = (sk == NULL || sk->direction == dir) ?
5852 &pd.baddr : &pd.naddr;
5854 x = (sk == NULL || sk->direction == dir) ?
5855 &pd.naddr : &pd.baddr;
5856 if (x == &pd.baddr || s == NULL) {
5857 /* we need to change the address */
5864 if (tr->src.addr.type == PF_ADDR_TABLE)
5865 pfr_update_stats(tr->src.addr.p.tbl, (sk == NULL ||
5866 sk->direction == dir) ?
5867 pd.src : pd.dst, pd.af,
5868 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5870 if (tr->dst.addr.type == PF_ADDR_TABLE)
5871 pfr_update_stats(tr->dst.addr.p.tbl, (sk == NULL ||
5872 sk->direction == dir) ? pd.dst : pd.src, pd.af,
5873 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5878 if (action == PF_SYNPROXY_DROP) {
5883 /* pf_route can free the mbuf causing *m0 to become NULL */
5884 pf_route(m0, r, dir, kif->pfik_ifp, s, &pd);
5892 pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0,
5893 struct ether_header *eh, struct inpcb *inp)
5895 struct pfi_kif *kif;
5896 u_short action, reason = 0, log = 0;
5897 struct mbuf *m = *m0, *n = NULL;
5898 struct ip6_hdr *h = NULL;
5899 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr;
5900 struct pf_state *s = NULL;
5901 struct pf_state_key *sk = NULL;
5902 struct pf_ruleset *ruleset = NULL;
5904 int off, terminal = 0, dirndx, rh_cnt = 0;
5906 if (!pf_status.running)
5909 memset(&pd, 0, sizeof(pd));
5910 if (ifp->if_type == IFT_CARP && ifp->if_carpdev)
5911 kif = (struct pfi_kif *)ifp->if_carpdev->if_pf_kif;
5913 kif = (struct pfi_kif *)ifp->if_pf_kif;
5916 DPFPRINTF(PF_DEBUG_URGENT,
5917 ("pf_test6: kif == NULL, if_xname %s\n", ifp->if_xname));
5920 if (kif->pfik_flags & PFI_IFLAG_SKIP)
5924 if ((m->m_flags & M_PKTHDR) == 0)
5925 panic("non-M_PKTHDR is passed to pf_test6");
5926 #endif /* DIAGNOSTIC */
5928 if (m->m_pkthdr.len < (int)sizeof(*h)) {
5930 REASON_SET(&reason, PFRES_SHORT);
5935 if (m->m_pkthdr.pf.flags & PF_TAG_GENERATED)
5938 /* We do IP header normalization and packet reassembly here */
5939 if (pf_normalize_ip6(m0, dir, kif, &reason, &pd) != PF_PASS) {
5943 m = *m0; /* pf_normalize messes with m0 */
5944 h = mtod(m, struct ip6_hdr *);
5948 * we do not support jumbogram yet. if we keep going, zero ip6_plen
5949 * will do something bad, so drop the packet for now.
5951 if (htons(h->ip6_plen) == 0) {
5953 REASON_SET(&reason, PFRES_NORM); /*XXX*/
5958 pd.src = (struct pf_addr *)&h->ip6_src;
5959 pd.dst = (struct pf_addr *)&h->ip6_dst;
5960 PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET6);
5964 pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr);
5967 off = ((caddr_t)h - m->m_data) + sizeof(struct ip6_hdr);
5968 pd.proto = h->ip6_nxt;
5971 case IPPROTO_FRAGMENT:
5972 action = pf_test_fragment(&r, dir, kif, m, h,
5974 if (action == PF_DROP)
5975 REASON_SET(&reason, PFRES_FRAG);
5977 case IPPROTO_ROUTING: {
5978 struct ip6_rthdr rthdr;
5981 DPFPRINTF(PF_DEBUG_MISC,
5982 ("pf: IPv6 more than one rthdr\n"));
5984 REASON_SET(&reason, PFRES_IPOPTIONS);
5988 if (!pf_pull_hdr(m, off, &rthdr, sizeof(rthdr), NULL,
5990 DPFPRINTF(PF_DEBUG_MISC,
5991 ("pf: IPv6 short rthdr\n"));
5993 REASON_SET(&reason, PFRES_SHORT);
5997 if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) {
5998 DPFPRINTF(PF_DEBUG_MISC,
5999 ("pf: IPv6 rthdr0\n"));
6001 REASON_SET(&reason, PFRES_IPOPTIONS);
6008 case IPPROTO_HOPOPTS:
6009 case IPPROTO_DSTOPTS: {
6010 /* get next header and header length */
6011 struct ip6_ext opt6;
6013 if (!pf_pull_hdr(m, off, &opt6, sizeof(opt6),
6014 NULL, &reason, pd.af)) {
6015 DPFPRINTF(PF_DEBUG_MISC,
6016 ("pf: IPv6 short opt\n"));
6021 if (pd.proto == IPPROTO_AH)
6022 off += (opt6.ip6e_len + 2) * 4;
6024 off += (opt6.ip6e_len + 1) * 8;
6025 pd.proto = opt6.ip6e_nxt;
6026 /* goto the next header */
6033 } while (!terminal);
6035 /* if there's no routing header, use unmodified mbuf for checksumming */
6045 if (!pf_pull_hdr(m, off, &th, sizeof(th),
6046 &action, &reason, AF_INET6)) {
6047 log = action != PF_PASS;
6050 pd.p_len = pd.tot_len - off - (th.th_off << 2);
6051 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
6052 if (action == PF_DROP)
6054 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
6056 if (action == PF_PASS) {
6058 pfsync_update_state(s);
6059 #endif /* NPFSYNC */
6063 } else if (s == NULL)
6064 action = pf_test_rule(&r, &s, dir, kif,
6065 m, off, h, &pd, &a, &ruleset, NULL, inp);
6073 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
6074 &action, &reason, AF_INET6)) {
6075 log = action != PF_PASS;
6078 if (uh.uh_dport == 0 ||
6079 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
6080 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
6082 REASON_SET(&reason, PFRES_SHORT);
6085 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
6086 if (action == PF_PASS) {
6088 pfsync_update_state(s);
6089 #endif /* NPFSYNC */
6093 } else if (s == NULL)
6094 action = pf_test_rule(&r, &s, dir, kif,
6095 m, off, h, &pd, &a, &ruleset, NULL, inp);
6099 case IPPROTO_ICMPV6: {
6100 struct icmp6_hdr ih;
6103 if (!pf_pull_hdr(m, off, &ih, sizeof(ih),
6104 &action, &reason, AF_INET6)) {
6105 log = action != PF_PASS;
6108 action = pf_test_state_icmp(&s, dir, kif,
6109 m, off, h, &pd, &reason);
6110 if (action == PF_PASS) {
6112 pfsync_update_state(s);
6113 #endif /* NPFSYNC */
6117 } else if (s == NULL)
6118 action = pf_test_rule(&r, &s, dir, kif,
6119 m, off, h, &pd, &a, &ruleset, NULL, inp);
6124 action = pf_test_state_other(&s, dir, kif, &pd);
6125 if (action == PF_PASS) {
6127 pfsync_update_state(s);
6128 #endif /* NPFSYNC */
6132 } else if (s == NULL)
6133 action = pf_test_rule(&r, &s, dir, kif, m, off, h,
6134 &pd, &a, &ruleset, NULL, inp);
6144 /* handle dangerous IPv6 extension headers. */
6145 if (action == PF_PASS && rh_cnt &&
6146 !((s && s->allow_opts) || r->allow_opts)) {
6148 REASON_SET(&reason, PFRES_IPOPTIONS);
6150 DPFPRINTF(PF_DEBUG_MISC,
6151 ("pf: dropping packet with dangerous v6 headers\n"));
6154 if ((s && s->tag) || r->rtableid)
6155 pf_tag_packet(m, s ? s->tag : 0, r->rtableid);
6158 if (action == PF_PASS && r->qid) {
6159 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
6160 if (pd.tos & IPTOS_LOWDELAY)
6161 m->m_pkthdr.pf.qid = r->pqid;
6163 m->m_pkthdr.pf.qid = r->qid;
6164 m->m_pkthdr.pf.ecn_af = AF_INET6;
6165 m->m_pkthdr.pf.hdr = h;
6168 m->m_pkthdr.pf.state_hash = s->hash;
6169 m->m_pkthdr.pf.flags |= PF_TAG_STATE_HASHED;
6174 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
6175 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
6176 (s->nat_rule.ptr->action == PF_RDR ||
6177 s->nat_rule.ptr->action == PF_BINAT) &&
6178 IN6_IS_ADDR_LOOPBACK(&pd.dst->v6))
6179 m->m_pkthdr.pf.flags |= PF_TAG_TRANSLATE_LOCALHOST;
6184 if (s != NULL && s->nat_rule.ptr != NULL &&
6185 s->nat_rule.ptr->log & PF_LOG_ALL)
6186 lr = s->nat_rule.ptr;
6189 PFLOG_PACKET(kif, h, m, AF_INET6, dir, reason, lr, a, ruleset,
6193 kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
6194 kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++;
6196 if (action == PF_PASS || r->action == PF_DROP) {
6197 dirndx = (dir == PF_OUT);
6198 r->packets[dirndx]++;
6199 r->bytes[dirndx] += pd.tot_len;
6201 a->packets[dirndx]++;
6202 a->bytes[dirndx] += pd.tot_len;
6206 if (s->nat_rule.ptr != NULL) {
6207 s->nat_rule.ptr->packets[dirndx]++;
6208 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len;
6210 if (s->src_node != NULL) {
6211 s->src_node->packets[dirndx]++;
6212 s->src_node->bytes[dirndx] += pd.tot_len;
6214 if (s->nat_src_node != NULL) {
6215 s->nat_src_node->packets[dirndx]++;
6216 s->nat_src_node->bytes[dirndx] += pd.tot_len;
6218 dirndx = (dir == sk->direction) ? 0 : 1;
6219 s->packets[dirndx]++;
6220 s->bytes[dirndx] += pd.tot_len;
6223 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
6227 * XXX: we need to make sure that the addresses
6228 * passed to pfr_update_stats() are the same than
6229 * the addresses used during matching (pfr_match)
6231 if (r == &pf_default_rule) {
6233 x = (s == NULL || sk->direction == dir) ?
6234 &pd.baddr : &pd.naddr;
6236 x = (s == NULL || sk->direction == dir) ?
6237 &pd.naddr : &pd.baddr;
6239 if (x == &pd.baddr || s == NULL) {
6246 if (tr->src.addr.type == PF_ADDR_TABLE)
6247 pfr_update_stats(tr->src.addr.p.tbl, (sk == NULL ||
6248 sk->direction == dir) ? pd.src : pd.dst, pd.af,
6249 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
6251 if (tr->dst.addr.type == PF_ADDR_TABLE)
6252 pfr_update_stats(tr->dst.addr.p.tbl, (sk == NULL ||
6253 sk->direction == dir) ? pd.dst : pd.src, pd.af,
6254 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
6259 if (action == PF_SYNPROXY_DROP) {
6264 /* pf_route6 can free the mbuf causing *m0 to become NULL */
6265 pf_route6(m0, r, dir, kif->pfik_ifp, s, &pd);
6272 pf_check_congestion(struct ifqueue *ifq)