1 /* $OpenBSD: pf.c,v 1.552.2.1 2007/11/27 16:37:57 henning Exp $ */
4 * Copyright (c) 2004 The DragonFly Project. All rights reserved.
6 * Copyright (c) 2001 Daniel Hartmeier
7 * Copyright (c) 2002,2003 Henning Brauer
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
14 * - Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * - Redistributions in binary form must reproduce the above
17 * copyright notice, this list of conditions and the following
18 * disclaimer in the documentation and/or other materials provided
19 * with the distribution.
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
25 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
27 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
29 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
31 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
32 * POSSIBILITY OF SUCH DAMAGE.
34 * Effort sponsored in part by the Defense Advanced Research Projects
35 * Agency (DARPA) and Air Force Research Laboratory, Air Force
36 * Materiel Command, USAF, under agreement number F30602-01-2-0537.
41 #include "opt_inet6.h"
42 #include "use_pfsync.h"
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/malloc.h>
48 #include <sys/filio.h>
49 #include <sys/socket.h>
50 #include <sys/socketvar.h>
51 #include <sys/kernel.h>
53 #include <sys/sysctl.h>
54 #include <sys/endian.h>
55 #include <vm/vm_zone.h>
57 #include <sys/kthread.h>
59 #include <sys/mplock2.h>
61 #include <machine/inttypes.h>
64 #include <net/if_types.h>
66 #include <net/netisr.h>
67 #include <net/route.h>
69 #include <netinet/in.h>
70 #include <netinet/in_var.h>
71 #include <netinet/in_systm.h>
72 #include <netinet/ip.h>
73 #include <netinet/ip_var.h>
74 #include <netinet/tcp.h>
75 #include <netinet/tcp_seq.h>
76 #include <netinet/udp.h>
77 #include <netinet/ip_icmp.h>
78 #include <netinet/in_pcb.h>
79 #include <netinet/tcp_timer.h>
80 #include <netinet/tcp_var.h>
81 #include <netinet/udp_var.h>
82 #include <netinet/icmp_var.h>
83 #include <netinet/if_ether.h>
85 #include <net/pf/pfvar.h>
86 #include <net/pf/if_pflog.h>
89 #include <net/pf/if_pfsync.h>
90 #endif /* NPFSYNC > 0 */
93 #include <netinet/ip6.h>
94 #include <netinet/in_pcb.h>
95 #include <netinet/icmp6.h>
96 #include <netinet6/nd6.h>
97 #include <netinet6/ip6_var.h>
98 #include <netinet6/in6_pcb.h>
101 #include <sys/in_cksum.h>
102 #include <sys/ucred.h>
103 #include <machine/limits.h>
104 #include <sys/msgport2.h>
105 #include <net/netmsg2.h>
107 extern int ip_optcopy(struct ip *, struct ip *);
108 extern int debug_pfugidhack;
110 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) kprintf x
117 struct pf_state_tree_lan_ext pf_statetbl_lan_ext;
118 struct pf_state_tree_ext_gwy pf_statetbl_ext_gwy;
120 struct pf_altqqueue pf_altqs[2];
121 struct pf_palist pf_pabuf;
122 struct pf_altqqueue *pf_altqs_active;
123 struct pf_altqqueue *pf_altqs_inactive;
124 struct pf_status pf_status;
126 u_int32_t ticket_altqs_active;
127 u_int32_t ticket_altqs_inactive;
128 int altqs_inactive_open;
129 u_int32_t ticket_pabuf;
131 struct pf_anchor_stackframe {
132 struct pf_ruleset *rs;
134 struct pf_anchor_node *parent;
135 struct pf_anchor *child;
136 } pf_anchor_stack[64];
138 vm_zone_t pf_src_tree_pl, pf_rule_pl, pf_pooladdr_pl;
139 vm_zone_t pf_state_pl, pf_state_key_pl;
140 vm_zone_t pf_altq_pl;
142 void pf_print_host(struct pf_addr *, u_int16_t, u_int8_t);
144 void pf_init_threshold(struct pf_threshold *, u_int32_t,
146 void pf_add_threshold(struct pf_threshold *);
147 int pf_check_threshold(struct pf_threshold *);
149 void pf_change_ap(struct pf_addr *, u_int16_t *,
150 u_int16_t *, u_int16_t *, struct pf_addr *,
151 u_int16_t, u_int8_t, sa_family_t);
152 int pf_modulate_sack(struct mbuf *, int, struct pf_pdesc *,
153 struct tcphdr *, struct pf_state_peer *);
155 void pf_change_a6(struct pf_addr *, u_int16_t *,
156 struct pf_addr *, u_int8_t);
158 void pf_change_icmp(struct pf_addr *, u_int16_t *,
159 struct pf_addr *, struct pf_addr *, u_int16_t,
160 u_int16_t *, u_int16_t *, u_int16_t *,
161 u_int16_t *, u_int8_t, sa_family_t);
162 void pf_send_tcp(const struct pf_rule *, sa_family_t,
163 const struct pf_addr *, const struct pf_addr *,
164 u_int16_t, u_int16_t, u_int32_t, u_int32_t,
165 u_int8_t, u_int16_t, u_int16_t, u_int8_t, int,
166 u_int16_t, struct ether_header *, struct ifnet *);
167 void pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t,
168 sa_family_t, struct pf_rule *);
169 struct pf_rule *pf_match_translation(struct pf_pdesc *, struct mbuf *,
170 int, int, struct pfi_kif *,
171 struct pf_addr *, u_int16_t, struct pf_addr *,
173 struct pf_rule *pf_get_translation(struct pf_pdesc *, struct mbuf *,
174 int, int, struct pfi_kif *, struct pf_src_node **,
175 struct pf_addr *, u_int16_t,
176 struct pf_addr *, u_int16_t,
177 struct pf_addr *, u_int16_t *);
178 void pf_attach_state(struct pf_state_key *,
179 struct pf_state *, int);
180 void pf_detach_state(struct pf_state *, int);
181 int pf_test_rule(struct pf_rule **, struct pf_state **,
182 int, struct pfi_kif *, struct mbuf *, int,
183 void *, struct pf_pdesc *, struct pf_rule **,
184 struct pf_ruleset **, struct ifqueue *, struct inpcb *);
185 int pf_test_fragment(struct pf_rule **, int,
186 struct pfi_kif *, struct mbuf *, void *,
187 struct pf_pdesc *, struct pf_rule **,
188 struct pf_ruleset **);
189 int pf_test_state_tcp(struct pf_state **, int,
190 struct pfi_kif *, struct mbuf *, int,
191 void *, struct pf_pdesc *, u_short *);
192 int pf_test_state_udp(struct pf_state **, int,
193 struct pfi_kif *, struct mbuf *, int,
194 void *, struct pf_pdesc *);
195 int pf_test_state_icmp(struct pf_state **, int,
196 struct pfi_kif *, struct mbuf *, int,
197 void *, struct pf_pdesc *, u_short *);
198 int pf_test_state_other(struct pf_state **, int,
199 struct pfi_kif *, struct pf_pdesc *);
200 int pf_match_tag(struct mbuf *, struct pf_rule *, int *);
201 void pf_step_into_anchor(int *, struct pf_ruleset **, int,
202 struct pf_rule **, struct pf_rule **, int *);
203 int pf_step_out_of_anchor(int *, struct pf_ruleset **,
204 int, struct pf_rule **, struct pf_rule **,
206 void pf_hash(struct pf_addr *, struct pf_addr *,
207 struct pf_poolhashkey *, sa_family_t);
208 int pf_map_addr(u_int8_t, struct pf_rule *,
209 struct pf_addr *, struct pf_addr *,
210 struct pf_addr *, struct pf_src_node **);
211 int pf_get_sport(sa_family_t, u_int8_t, struct pf_rule *,
212 struct pf_addr *, struct pf_addr *, u_int16_t,
213 struct pf_addr *, u_int16_t*, u_int16_t, u_int16_t,
214 struct pf_src_node **);
215 void pf_route(struct mbuf **, struct pf_rule *, int,
216 struct ifnet *, struct pf_state *,
218 void pf_route6(struct mbuf **, struct pf_rule *, int,
219 struct ifnet *, struct pf_state *,
221 u_int8_t pf_get_wscale(struct mbuf *, int, u_int16_t,
223 u_int16_t pf_get_mss(struct mbuf *, int, u_int16_t,
225 u_int16_t pf_calc_mss(struct pf_addr *, sa_family_t,
227 void pf_set_rt_ifp(struct pf_state *,
229 int pf_check_proto_cksum(struct mbuf *, int, int,
230 u_int8_t, sa_family_t);
231 int pf_addr_wrap_neq(struct pf_addr_wrap *,
232 struct pf_addr_wrap *);
233 struct pf_state *pf_find_state(struct pfi_kif *,
234 struct pf_state_key_cmp *, u_int8_t);
235 int pf_src_connlimit(struct pf_state **);
236 void pf_stateins_err(const char *, struct pf_state *,
238 int pf_check_congestion(struct ifqueue *);
240 extern int pf_end_threads;
242 struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] = {
243 { &pf_state_pl, PFSTATE_HIWAT },
244 { &pf_src_tree_pl, PFSNODE_HIWAT },
245 { &pf_frent_pl, PFFRAG_FRENT_HIWAT },
246 { &pfr_ktable_pl, PFR_KTABLE_HIWAT },
247 { &pfr_kentry_pl, PFR_KENTRY_HIWAT }
250 #define STATE_LOOKUP() \
252 if (direction == PF_IN) \
253 *state = pf_find_state(kif, &key, PF_EXT_GWY); \
255 *state = pf_find_state(kif, &key, PF_LAN_EXT); \
256 if (*state == NULL || (*state)->timeout == PFTM_PURGE) \
258 if (direction == PF_OUT && \
259 (((*state)->rule.ptr->rt == PF_ROUTETO && \
260 (*state)->rule.ptr->direction == PF_OUT) || \
261 ((*state)->rule.ptr->rt == PF_REPLYTO && \
262 (*state)->rule.ptr->direction == PF_IN)) && \
263 (*state)->rt_kif != NULL && \
264 (*state)->rt_kif != kif) \
268 #define STATE_TRANSLATE(sk) \
269 (sk)->lan.addr.addr32[0] != (sk)->gwy.addr.addr32[0] || \
270 ((sk)->af == AF_INET6 && \
271 ((sk)->lan.addr.addr32[1] != (sk)->gwy.addr.addr32[1] || \
272 (sk)->lan.addr.addr32[2] != (sk)->gwy.addr.addr32[2] || \
273 (sk)->lan.addr.addr32[3] != (sk)->gwy.addr.addr32[3])) || \
274 (sk)->lan.port != (sk)->gwy.port
276 #define BOUND_IFACE(r, k) \
277 ((r)->rule_flag & PFRULE_IFBOUND) ? (k) : pfi_all
279 #define STATE_INC_COUNTERS(s) \
281 s->rule.ptr->states++; \
282 if (s->anchor.ptr != NULL) \
283 s->anchor.ptr->states++; \
284 if (s->nat_rule.ptr != NULL) \
285 s->nat_rule.ptr->states++; \
288 #define STATE_DEC_COUNTERS(s) \
290 if (s->nat_rule.ptr != NULL) \
291 s->nat_rule.ptr->states--; \
292 if (s->anchor.ptr != NULL) \
293 s->anchor.ptr->states--; \
294 s->rule.ptr->states--; \
297 static __inline int pf_src_compare(struct pf_src_node *, struct pf_src_node *);
298 static __inline int pf_state_compare_lan_ext(struct pf_state_key *,
299 struct pf_state_key *);
300 static __inline int pf_state_compare_ext_gwy(struct pf_state_key *,
301 struct pf_state_key *);
302 static __inline int pf_state_compare_id(struct pf_state *,
305 struct pf_src_tree tree_src_tracking;
307 struct pf_state_tree_id tree_id;
308 struct pf_state_queue state_list;
310 RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare);
311 RB_GENERATE(pf_state_tree_lan_ext, pf_state_key,
312 entry_lan_ext, pf_state_compare_lan_ext);
313 RB_GENERATE(pf_state_tree_ext_gwy, pf_state_key,
314 entry_ext_gwy, pf_state_compare_ext_gwy);
315 RB_GENERATE(pf_state_tree_id, pf_state,
316 entry_id, pf_state_compare_id);
318 #define PF_DT_SKIP_LANEXT 0x01
319 #define PF_DT_SKIP_EXTGWY 0x02
322 pf_src_compare(struct pf_src_node *a, struct pf_src_node *b)
326 if (a->rule.ptr > b->rule.ptr)
328 if (a->rule.ptr < b->rule.ptr)
330 if ((diff = a->af - b->af) != 0)
335 if (a->addr.addr32[0] > b->addr.addr32[0])
337 if (a->addr.addr32[0] < b->addr.addr32[0])
343 if (a->addr.addr32[3] > b->addr.addr32[3])
345 if (a->addr.addr32[3] < b->addr.addr32[3])
347 if (a->addr.addr32[2] > b->addr.addr32[2])
349 if (a->addr.addr32[2] < b->addr.addr32[2])
351 if (a->addr.addr32[1] > b->addr.addr32[1])
353 if (a->addr.addr32[1] < b->addr.addr32[1])
355 if (a->addr.addr32[0] > b->addr.addr32[0])
357 if (a->addr.addr32[0] < b->addr.addr32[0])
366 pf_state_hash(struct pf_state_key *sk)
368 u_int32_t hv = (intptr_t)sk / sizeof(*sk);
370 hv ^= crc32(&sk->lan, sizeof(sk->lan));
371 hv ^= crc32(&sk->gwy, sizeof(sk->gwy));
372 hv ^= crc32(&sk->ext, sizeof(sk->ext));
373 if (hv == 0) /* disallow 0 */
379 pf_state_compare_lan_ext(struct pf_state_key *a, struct pf_state_key *b)
383 if ((diff = a->proto - b->proto) != 0)
385 if ((diff = a->af - b->af) != 0)
390 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
392 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
394 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
396 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
402 if (a->lan.addr.addr32[3] > b->lan.addr.addr32[3])
404 if (a->lan.addr.addr32[3] < b->lan.addr.addr32[3])
406 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
408 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
410 if (a->lan.addr.addr32[2] > b->lan.addr.addr32[2])
412 if (a->lan.addr.addr32[2] < b->lan.addr.addr32[2])
414 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
416 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
418 if (a->lan.addr.addr32[1] > b->lan.addr.addr32[1])
420 if (a->lan.addr.addr32[1] < b->lan.addr.addr32[1])
422 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
424 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
426 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
428 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
430 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
432 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
438 if ((diff = a->lan.port - b->lan.port) != 0)
440 if ((diff = a->ext.port - b->ext.port) != 0)
447 pf_state_compare_ext_gwy(struct pf_state_key *a, struct pf_state_key *b)
451 if ((diff = a->proto - b->proto) != 0)
453 if ((diff = a->af - b->af) != 0)
458 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
460 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
462 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
464 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
470 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
472 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
474 if (a->gwy.addr.addr32[3] > b->gwy.addr.addr32[3])
476 if (a->gwy.addr.addr32[3] < b->gwy.addr.addr32[3])
478 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
480 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
482 if (a->gwy.addr.addr32[2] > b->gwy.addr.addr32[2])
484 if (a->gwy.addr.addr32[2] < b->gwy.addr.addr32[2])
486 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
488 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
490 if (a->gwy.addr.addr32[1] > b->gwy.addr.addr32[1])
492 if (a->gwy.addr.addr32[1] < b->gwy.addr.addr32[1])
494 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
496 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
498 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
500 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
506 if ((diff = a->ext.port - b->ext.port) != 0)
508 if ((diff = a->gwy.port - b->gwy.port) != 0)
515 pf_state_compare_id(struct pf_state *a, struct pf_state *b)
521 if (a->creatorid > b->creatorid)
523 if (a->creatorid < b->creatorid)
531 pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af)
536 dst->addr32[0] = src->addr32[0];
540 dst->addr32[0] = src->addr32[0];
541 dst->addr32[1] = src->addr32[1];
542 dst->addr32[2] = src->addr32[2];
543 dst->addr32[3] = src->addr32[3];
550 pf_find_state_byid(struct pf_state_cmp *key)
552 pf_status.fcounters[FCNT_STATE_SEARCH]++;
554 return (RB_FIND(pf_state_tree_id, &tree_id, (struct pf_state *)key));
558 pf_find_state(struct pfi_kif *kif, struct pf_state_key_cmp *key, u_int8_t tree)
560 struct pf_state_key *sk;
563 pf_status.fcounters[FCNT_STATE_SEARCH]++;
567 sk = RB_FIND(pf_state_tree_lan_ext, &pf_statetbl_lan_ext,
568 (struct pf_state_key *)key);
571 sk = RB_FIND(pf_state_tree_ext_gwy, &pf_statetbl_ext_gwy,
572 (struct pf_state_key *)key);
575 panic("pf_find_state");
578 /* list is sorted, if-bound states before floating ones */
580 TAILQ_FOREACH(s, &sk->states, next)
581 if (s->kif == pfi_all || s->kif == kif)
588 pf_find_state_all(struct pf_state_key_cmp *key, u_int8_t tree, int *more)
590 struct pf_state_key *sk;
591 struct pf_state *s, *ret = NULL;
593 pf_status.fcounters[FCNT_STATE_SEARCH]++;
597 sk = RB_FIND(pf_state_tree_lan_ext,
598 &pf_statetbl_lan_ext, (struct pf_state_key *)key);
601 sk = RB_FIND(pf_state_tree_ext_gwy,
602 &pf_statetbl_ext_gwy, (struct pf_state_key *)key);
605 panic("pf_find_state_all");
609 ret = TAILQ_FIRST(&sk->states);
613 TAILQ_FOREACH(s, &sk->states, next)
621 pf_init_threshold(struct pf_threshold *threshold,
622 u_int32_t limit, u_int32_t seconds)
624 threshold->limit = limit * PF_THRESHOLD_MULT;
625 threshold->seconds = seconds;
626 threshold->count = 0;
627 threshold->last = time_second;
631 pf_add_threshold(struct pf_threshold *threshold)
633 u_int32_t t = time_second, diff = t - threshold->last;
635 if (diff >= threshold->seconds)
636 threshold->count = 0;
638 threshold->count -= threshold->count * diff /
640 threshold->count += PF_THRESHOLD_MULT;
645 pf_check_threshold(struct pf_threshold *threshold)
647 return (threshold->count > threshold->limit);
651 pf_src_connlimit(struct pf_state **state)
655 (*state)->src_node->conn++;
656 (*state)->src.tcp_est = 1;
657 pf_add_threshold(&(*state)->src_node->conn_rate);
659 if ((*state)->rule.ptr->max_src_conn &&
660 (*state)->rule.ptr->max_src_conn <
661 (*state)->src_node->conn) {
662 pf_status.lcounters[LCNT_SRCCONN]++;
666 if ((*state)->rule.ptr->max_src_conn_rate.limit &&
667 pf_check_threshold(&(*state)->src_node->conn_rate)) {
668 pf_status.lcounters[LCNT_SRCCONNRATE]++;
675 if ((*state)->rule.ptr->overload_tbl) {
677 u_int32_t killed = 0;
679 pf_status.lcounters[LCNT_OVERLOAD_TABLE]++;
680 if (pf_status.debug >= PF_DEBUG_MISC) {
681 kprintf("pf_src_connlimit: blocking address ");
682 pf_print_host(&(*state)->src_node->addr, 0,
683 (*state)->state_key->af);
686 bzero(&p, sizeof(p));
687 p.pfra_af = (*state)->state_key->af;
688 switch ((*state)->state_key->af) {
692 p.pfra_ip4addr = (*state)->src_node->addr.v4;
698 p.pfra_ip6addr = (*state)->src_node->addr.v6;
703 pfr_insert_kentry((*state)->rule.ptr->overload_tbl,
706 /* kill existing states if that's required. */
707 if ((*state)->rule.ptr->flush) {
708 struct pf_state_key *sk;
711 pf_status.lcounters[LCNT_OVERLOAD_FLUSH]++;
712 RB_FOREACH(st, pf_state_tree_id, &tree_id) {
715 * Kill states from this source. (Only those
716 * from the same rule if PF_FLUSH_GLOBAL is not
720 (*state)->state_key->af &&
721 (((*state)->state_key->direction ==
723 PF_AEQ(&(*state)->src_node->addr,
724 &sk->lan.addr, sk->af)) ||
725 ((*state)->state_key->direction == PF_IN &&
726 PF_AEQ(&(*state)->src_node->addr,
727 &sk->ext.addr, sk->af))) &&
728 ((*state)->rule.ptr->flush &
730 (*state)->rule.ptr == st->rule.ptr)) {
731 st->timeout = PFTM_PURGE;
732 st->src.state = st->dst.state =
737 if (pf_status.debug >= PF_DEBUG_MISC)
738 kprintf(", %u states killed", killed);
740 if (pf_status.debug >= PF_DEBUG_MISC)
744 /* kill this state */
745 (*state)->timeout = PFTM_PURGE;
746 (*state)->src.state = (*state)->dst.state = TCPS_CLOSED;
751 pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule,
752 struct pf_addr *src, sa_family_t af)
754 struct pf_src_node k;
758 PF_ACPY(&k.addr, src, af);
759 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
760 rule->rpool.opts & PF_POOL_STICKYADDR)
764 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
765 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
768 if (!rule->max_src_nodes ||
769 rule->src_nodes < rule->max_src_nodes)
770 (*sn) = pool_get(&pf_src_tree_pl, PR_NOWAIT);
772 pf_status.lcounters[LCNT_SRCNODES]++;
775 bzero(*sn, sizeof(struct pf_src_node));
777 pf_init_threshold(&(*sn)->conn_rate,
778 rule->max_src_conn_rate.limit,
779 rule->max_src_conn_rate.seconds);
782 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
783 rule->rpool.opts & PF_POOL_STICKYADDR)
784 (*sn)->rule.ptr = rule;
786 (*sn)->rule.ptr = NULL;
787 PF_ACPY(&(*sn)->addr, src, af);
788 if (RB_INSERT(pf_src_tree,
789 &tree_src_tracking, *sn) != NULL) {
790 if (pf_status.debug >= PF_DEBUG_MISC) {
791 kprintf("pf: src_tree insert failed: ");
792 pf_print_host(&(*sn)->addr, 0, af);
795 pool_put(&pf_src_tree_pl, *sn);
798 (*sn)->creation = time_second;
799 (*sn)->ruletype = rule->action;
800 if ((*sn)->rule.ptr != NULL)
801 (*sn)->rule.ptr->src_nodes++;
802 pf_status.scounters[SCNT_SRC_NODE_INSERT]++;
803 pf_status.src_nodes++;
805 if (rule->max_src_states &&
806 (*sn)->states >= rule->max_src_states) {
807 pf_status.lcounters[LCNT_SRCSTATES]++;
815 pf_stateins_err(const char *tree, struct pf_state *s, struct pfi_kif *kif)
817 struct pf_state_key *sk = s->state_key;
819 if (pf_status.debug >= PF_DEBUG_MISC) {
820 kprintf("pf: state insert failed: %s %s", tree, kif->pfik_name);
822 pf_print_host(&sk->lan.addr, sk->lan.port,
825 pf_print_host(&sk->gwy.addr, sk->gwy.port,
828 pf_print_host(&sk->ext.addr, sk->ext.port,
830 if (s->sync_flags & PFSTATE_FROMSYNC)
831 kprintf(" (from sync)");
837 pf_insert_state(struct pfi_kif *kif, struct pf_state *s)
839 struct pf_state_key *cur;
842 KKASSERT(s->state_key != NULL);
845 if ((cur = RB_INSERT(pf_state_tree_lan_ext, &pf_statetbl_lan_ext,
846 s->state_key)) != NULL) {
847 /* key exists. check for same kif, if none, add to key */
848 TAILQ_FOREACH(sp, &cur->states, next)
849 if (sp->kif == kif) { /* collision! */
850 pf_stateins_err("tree_lan_ext", s, kif);
852 PF_DT_SKIP_LANEXT|PF_DT_SKIP_EXTGWY);
855 pf_detach_state(s, PF_DT_SKIP_LANEXT|PF_DT_SKIP_EXTGWY);
856 pf_attach_state(cur, s, kif == pfi_all ? 1 : 0);
859 /* if cur != NULL, we already found a state key and attached to it */
860 if (cur == NULL && (cur = RB_INSERT(pf_state_tree_ext_gwy,
861 &pf_statetbl_ext_gwy, s->state_key)) != NULL) {
862 /* must not happen. we must have found the sk above! */
863 pf_stateins_err("tree_ext_gwy", s, kif);
864 pf_detach_state(s, PF_DT_SKIP_EXTGWY);
868 if (s->id == 0 && s->creatorid == 0) {
869 s->id = htobe64(pf_status.stateid++);
870 s->creatorid = pf_status.hostid;
872 if (RB_INSERT(pf_state_tree_id, &tree_id, s) != NULL) {
873 if (pf_status.debug >= PF_DEBUG_MISC) {
874 kprintf("pf: state insert failed: "
875 "id: %016llx creatorid: %08x",
876 be64toh(s->id), ntohl(s->creatorid));
877 if (s->sync_flags & PFSTATE_FROMSYNC)
878 kprintf(" (from sync)");
881 pf_detach_state(s, 0);
884 TAILQ_INSERT_TAIL(&state_list, s, entry_list);
885 pf_status.fcounters[FCNT_STATE_INSERT]++;
887 pfi_kif_ref(kif, PFI_KIF_REF_STATE);
889 pfsync_insert_state(s);
895 pf_purge_thread(void *v)
902 tsleep(pf_purge_thread, PWAIT, "pftm", 1 * hz);
904 lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
906 if (pf_end_threads) {
907 pf_purge_expired_states(pf_status.states, 1);
908 pf_purge_expired_fragments();
909 pf_purge_expired_src_nodes(1);
912 lockmgr(&pf_consistency_lock, LK_RELEASE);
913 wakeup(pf_purge_thread);
918 /* process a fraction of the state table every second */
919 if(!pf_purge_expired_states(1 + (pf_status.states
920 / pf_default_rule.timeout[PFTM_INTERVAL]), 0)) {
922 pf_purge_expired_states(1 + (pf_status.states
923 / pf_default_rule.timeout[PFTM_INTERVAL]), 1);
926 /* purge other expired types every PFTM_INTERVAL seconds */
927 if (++nloops >= pf_default_rule.timeout[PFTM_INTERVAL]) {
928 pf_purge_expired_fragments();
929 if (!pf_purge_expired_src_nodes(locked)) {
930 pf_purge_expired_src_nodes(1);
935 lockmgr(&pf_consistency_lock, LK_RELEASE);
941 pf_state_expires(const struct pf_state *state)
948 /* handle all PFTM_* > PFTM_MAX here */
949 if (state->timeout == PFTM_PURGE)
950 return (time_second);
951 if (state->timeout == PFTM_UNTIL_PACKET)
953 KKASSERT(state->timeout != PFTM_UNLINKED);
954 KASSERT((state->timeout < PFTM_MAX),
955 ("pf_state_expires: timeout > PFTM_MAX"));
956 timeout = state->rule.ptr->timeout[state->timeout];
958 timeout = pf_default_rule.timeout[state->timeout];
959 start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START];
961 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END];
962 states = state->rule.ptr->states;
964 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START];
965 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END];
966 states = pf_status.states;
968 if (end && states > start && start < end) {
970 return (state->expire + timeout * (end - states) /
973 return (time_second);
975 return (state->expire + timeout);
979 pf_purge_expired_src_nodes(int waslocked)
981 struct pf_src_node *cur, *next;
982 int locked = waslocked;
984 for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) {
985 next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur);
987 if (cur->states <= 0 && cur->expire <= time_second) {
989 lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
990 next = RB_NEXT(pf_src_tree,
991 &tree_src_tracking, cur);
994 if (cur->rule.ptr != NULL) {
995 cur->rule.ptr->src_nodes--;
996 if (cur->rule.ptr->states <= 0 &&
997 cur->rule.ptr->max_src_nodes <= 0)
998 pf_rm_rule(NULL, cur->rule.ptr);
1000 RB_REMOVE(pf_src_tree, &tree_src_tracking, cur);
1001 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
1002 pf_status.src_nodes--;
1003 pool_put(&pf_src_tree_pl, cur);
1007 if (locked && !waslocked)
1008 lockmgr(&pf_consistency_lock, LK_RELEASE);
1013 pf_src_tree_remove_state(struct pf_state *s)
1017 if (s->src_node != NULL) {
1019 --s->src_node->conn;
1020 if (--s->src_node->states <= 0) {
1021 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1024 pf_default_rule.timeout[PFTM_SRC_NODE];
1025 s->src_node->expire = time_second + timeout;
1028 if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) {
1029 if (--s->nat_src_node->states <= 0) {
1030 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1033 pf_default_rule.timeout[PFTM_SRC_NODE];
1034 s->nat_src_node->expire = time_second + timeout;
1037 s->src_node = s->nat_src_node = NULL;
1040 /* callers should be at crit_enter() */
1042 pf_unlink_state(struct pf_state *cur)
1044 if (cur->src.state == PF_TCPS_PROXY_DST) {
1045 pf_send_tcp(cur->rule.ptr, cur->state_key->af,
1046 &cur->state_key->ext.addr, &cur->state_key->lan.addr,
1047 cur->state_key->ext.port, cur->state_key->lan.port,
1048 cur->src.seqhi, cur->src.seqlo + 1,
1049 TH_RST|TH_ACK, 0, 0, 0, 1, cur->tag, NULL, NULL);
1051 RB_REMOVE(pf_state_tree_id, &tree_id, cur);
1053 if (cur->creatorid == pf_status.hostid)
1054 pfsync_delete_state(cur);
1056 cur->timeout = PFTM_UNLINKED;
1057 pf_src_tree_remove_state(cur);
1058 pf_detach_state(cur, 0);
1061 /* callers should be at crit_enter() and hold the
1062 * write_lock on pf_consistency_lock */
1064 pf_free_state(struct pf_state *cur)
1067 if (pfsyncif != NULL &&
1068 (pfsyncif->sc_bulk_send_next == cur ||
1069 pfsyncif->sc_bulk_terminator == cur))
1072 KKASSERT(cur->timeout == PFTM_UNLINKED);
1073 if (--cur->rule.ptr->states <= 0 &&
1074 cur->rule.ptr->src_nodes <= 0)
1075 pf_rm_rule(NULL, cur->rule.ptr);
1076 if (cur->nat_rule.ptr != NULL)
1077 if (--cur->nat_rule.ptr->states <= 0 &&
1078 cur->nat_rule.ptr->src_nodes <= 0)
1079 pf_rm_rule(NULL, cur->nat_rule.ptr);
1080 if (cur->anchor.ptr != NULL)
1081 if (--cur->anchor.ptr->states <= 0)
1082 pf_rm_rule(NULL, cur->anchor.ptr);
1083 pf_normalize_tcp_cleanup(cur);
1084 pfi_kif_unref(cur->kif, PFI_KIF_REF_STATE);
1085 TAILQ_REMOVE(&state_list, cur, entry_list);
1087 pf_tag_unref(cur->tag);
1088 pool_put(&pf_state_pl, cur);
1089 pf_status.fcounters[FCNT_STATE_REMOVALS]++;
1094 pf_purge_expired_states(u_int32_t maxcheck, int waslocked)
1096 static struct pf_state *cur = NULL;
1097 struct pf_state *next;
1098 int locked = waslocked;
1100 while (maxcheck--) {
1101 /* wrap to start of list when we hit the end */
1103 cur = TAILQ_FIRST(&state_list);
1105 break; /* list empty */
1108 /* get next state, as cur may get deleted */
1109 next = TAILQ_NEXT(cur, entry_list);
1111 if (cur->timeout == PFTM_UNLINKED) {
1112 /* free unlinked state */
1114 lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
1118 } else if (pf_state_expires(cur) <= time_second) {
1119 /* unlink and free expired state */
1120 pf_unlink_state(cur);
1122 if (!lockmgr(&pf_consistency_lock, LK_EXCLUSIVE))
1132 lockmgr(&pf_consistency_lock, LK_RELEASE);
1137 pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw)
1139 if (aw->type != PF_ADDR_TABLE)
1141 if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL)
1147 pf_tbladdr_remove(struct pf_addr_wrap *aw)
1149 if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL)
1151 pfr_detach_table(aw->p.tbl);
1156 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
1158 struct pfr_ktable *kt = aw->p.tbl;
1160 if (aw->type != PF_ADDR_TABLE || kt == NULL)
1162 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
1163 kt = kt->pfrkt_root;
1165 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
1170 pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af)
1175 u_int32_t a = ntohl(addr->addr32[0]);
1176 kprintf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255,
1188 u_int8_t i, curstart = 255, curend = 0,
1189 maxstart = 0, maxend = 0;
1190 for (i = 0; i < 8; i++) {
1191 if (!addr->addr16[i]) {
1192 if (curstart == 255)
1198 if ((curend - curstart) >
1199 (maxend - maxstart)) {
1200 maxstart = curstart;
1207 for (i = 0; i < 8; i++) {
1208 if (i >= maxstart && i <= maxend) {
1217 b = ntohs(addr->addr16[i]);
1234 pf_print_state(struct pf_state *s)
1236 struct pf_state_key *sk = s->state_key;
1237 switch (sk->proto) {
1247 case IPPROTO_ICMPV6:
1251 kprintf("%u ", sk->proto);
1254 pf_print_host(&sk->lan.addr, sk->lan.port, sk->af);
1256 pf_print_host(&sk->gwy.addr, sk->gwy.port, sk->af);
1258 pf_print_host(&sk->ext.addr, sk->ext.port, sk->af);
1259 kprintf(" [lo=%u high=%u win=%u modulator=%u", s->src.seqlo,
1260 s->src.seqhi, s->src.max_win, s->src.seqdiff);
1261 if (s->src.wscale && s->dst.wscale)
1262 kprintf(" wscale=%u", s->src.wscale & PF_WSCALE_MASK);
1264 kprintf(" [lo=%u high=%u win=%u modulator=%u", s->dst.seqlo,
1265 s->dst.seqhi, s->dst.max_win, s->dst.seqdiff);
1266 if (s->src.wscale && s->dst.wscale)
1267 kprintf(" wscale=%u", s->dst.wscale & PF_WSCALE_MASK);
1269 kprintf(" %u:%u", s->src.state, s->dst.state);
1273 pf_print_flags(u_int8_t f)
1295 #define PF_SET_SKIP_STEPS(i) \
1297 while (head[i] != cur) { \
1298 head[i]->skip[i].ptr = cur; \
1299 head[i] = TAILQ_NEXT(head[i], entries); \
1304 pf_calc_skip_steps(struct pf_rulequeue *rules)
1306 struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT];
1309 cur = TAILQ_FIRST(rules);
1311 for (i = 0; i < PF_SKIP_COUNT; ++i)
1313 while (cur != NULL) {
1315 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
1316 PF_SET_SKIP_STEPS(PF_SKIP_IFP);
1317 if (cur->direction != prev->direction)
1318 PF_SET_SKIP_STEPS(PF_SKIP_DIR);
1319 if (cur->af != prev->af)
1320 PF_SET_SKIP_STEPS(PF_SKIP_AF);
1321 if (cur->proto != prev->proto)
1322 PF_SET_SKIP_STEPS(PF_SKIP_PROTO);
1323 if (cur->src.neg != prev->src.neg ||
1324 pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr))
1325 PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR);
1326 if (cur->src.port[0] != prev->src.port[0] ||
1327 cur->src.port[1] != prev->src.port[1] ||
1328 cur->src.port_op != prev->src.port_op)
1329 PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
1330 if (cur->dst.neg != prev->dst.neg ||
1331 pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr))
1332 PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR);
1333 if (cur->dst.port[0] != prev->dst.port[0] ||
1334 cur->dst.port[1] != prev->dst.port[1] ||
1335 cur->dst.port_op != prev->dst.port_op)
1336 PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
1339 cur = TAILQ_NEXT(cur, entries);
1341 for (i = 0; i < PF_SKIP_COUNT; ++i)
1342 PF_SET_SKIP_STEPS(i);
1346 pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2)
1348 if (aw1->type != aw2->type)
1350 switch (aw1->type) {
1351 case PF_ADDR_ADDRMASK:
1352 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0))
1354 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0))
1357 case PF_ADDR_DYNIFTL:
1358 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt);
1359 case PF_ADDR_NOROUTE:
1360 case PF_ADDR_URPFFAILED:
1363 return (aw1->p.tbl != aw2->p.tbl);
1364 case PF_ADDR_RTLABEL:
1365 return (aw1->v.rtlabel != aw2->v.rtlabel);
1367 kprintf("invalid address type: %d\n", aw1->type);
1373 pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp)
1379 l = cksum + old - new;
1380 l = (l >> 16) + (l & 65535);
1388 pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc,
1389 struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af)
1394 PF_ACPY(&ao, a, af);
1402 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1403 ao.addr16[0], an->addr16[0], 0),
1404 ao.addr16[1], an->addr16[1], 0);
1406 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1407 ao.addr16[0], an->addr16[0], u),
1408 ao.addr16[1], an->addr16[1], u),
1414 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1415 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1416 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1417 ao.addr16[0], an->addr16[0], u),
1418 ao.addr16[1], an->addr16[1], u),
1419 ao.addr16[2], an->addr16[2], u),
1420 ao.addr16[3], an->addr16[3], u),
1421 ao.addr16[4], an->addr16[4], u),
1422 ao.addr16[5], an->addr16[5], u),
1423 ao.addr16[6], an->addr16[6], u),
1424 ao.addr16[7], an->addr16[7], u),
1432 /* Changes a u_int32_t. Uses a void * so there are no align restrictions */
1434 pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u)
1438 memcpy(&ao, a, sizeof(ao));
1439 memcpy(a, &an, sizeof(u_int32_t));
1440 *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u),
1441 ao % 65536, an % 65536, u);
1446 pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u)
1450 PF_ACPY(&ao, a, AF_INET6);
1451 PF_ACPY(a, an, AF_INET6);
1453 *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1454 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1455 pf_cksum_fixup(pf_cksum_fixup(*c,
1456 ao.addr16[0], an->addr16[0], u),
1457 ao.addr16[1], an->addr16[1], u),
1458 ao.addr16[2], an->addr16[2], u),
1459 ao.addr16[3], an->addr16[3], u),
1460 ao.addr16[4], an->addr16[4], u),
1461 ao.addr16[5], an->addr16[5], u),
1462 ao.addr16[6], an->addr16[6], u),
1463 ao.addr16[7], an->addr16[7], u);
1468 pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa,
1469 struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c,
1470 u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af)
1472 struct pf_addr oia, ooa;
1474 PF_ACPY(&oia, ia, af);
1475 PF_ACPY(&ooa, oa, af);
1477 /* Change inner protocol port, fix inner protocol checksum. */
1479 u_int16_t oip = *ip;
1486 *pc = pf_cksum_fixup(*pc, oip, *ip, u);
1487 *ic = pf_cksum_fixup(*ic, oip, *ip, 0);
1489 *ic = pf_cksum_fixup(*ic, opc, *pc, 0);
1491 /* Change inner ip address, fix inner ip and icmp checksums. */
1492 PF_ACPY(ia, na, af);
1496 u_int32_t oh2c = *h2c;
1498 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c,
1499 oia.addr16[0], ia->addr16[0], 0),
1500 oia.addr16[1], ia->addr16[1], 0);
1501 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1502 oia.addr16[0], ia->addr16[0], 0),
1503 oia.addr16[1], ia->addr16[1], 0);
1504 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0);
1510 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1511 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1512 pf_cksum_fixup(pf_cksum_fixup(*ic,
1513 oia.addr16[0], ia->addr16[0], u),
1514 oia.addr16[1], ia->addr16[1], u),
1515 oia.addr16[2], ia->addr16[2], u),
1516 oia.addr16[3], ia->addr16[3], u),
1517 oia.addr16[4], ia->addr16[4], u),
1518 oia.addr16[5], ia->addr16[5], u),
1519 oia.addr16[6], ia->addr16[6], u),
1520 oia.addr16[7], ia->addr16[7], u);
1524 /* Change outer ip address, fix outer ip or icmpv6 checksum. */
1525 PF_ACPY(oa, na, af);
1529 *hc = pf_cksum_fixup(pf_cksum_fixup(*hc,
1530 ooa.addr16[0], oa->addr16[0], 0),
1531 ooa.addr16[1], oa->addr16[1], 0);
1536 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1537 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1538 pf_cksum_fixup(pf_cksum_fixup(*ic,
1539 ooa.addr16[0], oa->addr16[0], u),
1540 ooa.addr16[1], oa->addr16[1], u),
1541 ooa.addr16[2], oa->addr16[2], u),
1542 ooa.addr16[3], oa->addr16[3], u),
1543 ooa.addr16[4], oa->addr16[4], u),
1544 ooa.addr16[5], oa->addr16[5], u),
1545 ooa.addr16[6], oa->addr16[6], u),
1546 ooa.addr16[7], oa->addr16[7], u);
1554 * Need to modulate the sequence numbers in the TCP SACK option
1555 * (credits to Krzysztof Pfaff for report and patch)
1558 pf_modulate_sack(struct mbuf *m, int off, struct pf_pdesc *pd,
1559 struct tcphdr *th, struct pf_state_peer *dst)
1561 int hlen = (th->th_off << 2) - sizeof(*th), thoptlen = hlen;
1562 u_int8_t opts[TCP_MAXOLEN], *opt = opts;
1563 int copyback = 0, i, olen;
1564 struct raw_sackblock sack;
1566 #define TCPOLEN_SACKLEN (TCPOLEN_SACK + 2)
1567 if (hlen < TCPOLEN_SACKLEN ||
1568 !pf_pull_hdr(m, off + sizeof(*th), opts, hlen, NULL, NULL, pd->af))
1571 while (hlen >= TCPOLEN_SACKLEN) {
1574 case TCPOPT_EOL: /* FALLTHROUGH */
1582 if (olen >= TCPOLEN_SACKLEN) {
1583 for (i = 2; i + TCPOLEN_SACK <= olen;
1584 i += TCPOLEN_SACK) {
1585 memcpy(&sack, &opt[i], sizeof(sack));
1586 pf_change_a(&sack.rblk_start, &th->th_sum,
1587 htonl(ntohl(sack.rblk_start) -
1589 pf_change_a(&sack.rblk_end, &th->th_sum,
1590 htonl(ntohl(sack.rblk_end) -
1592 memcpy(&opt[i], &sack, sizeof(sack));
1606 m_copyback(m, off + sizeof(*th), thoptlen, opts);
1611 pf_send_tcp(const struct pf_rule *r, sa_family_t af,
1612 const struct pf_addr *saddr, const struct pf_addr *daddr,
1613 u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack,
1614 u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl, int tag,
1615 u_int16_t rtag, struct ether_header *eh, struct ifnet *ifp)
1620 struct ip *h = NULL;
1623 struct ip6_hdr *h6 = NULL;
1625 struct tcphdr *th = NULL;
1628 /* maximum segment size tcp option */
1629 tlen = sizeof(struct tcphdr);
1636 len = sizeof(struct ip) + tlen;
1641 len = sizeof(struct ip6_hdr) + tlen;
1646 /* create outgoing mbuf */
1647 m = m_gethdr(MB_DONTWAIT, MT_HEADER);
1651 m->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
1652 m->m_pkthdr.pf.tag = rtag;
1654 if (r != NULL && r->rtableid >= 0)
1655 m->m_pkthdr.pf.rtableid = m->m_pkthdr.pf.rtableid;
1658 if (r != NULL && r->qid) {
1659 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
1660 m->m_pkthdr.pf.qid = r->qid;
1661 m->m_pkthdr.pf.ecn_af = af;
1662 m->m_pkthdr.pf.hdr = mtod(m, struct ip *);
1665 m->m_data += max_linkhdr;
1666 m->m_pkthdr.len = m->m_len = len;
1667 m->m_pkthdr.rcvif = NULL;
1668 bzero(m->m_data, len);
1672 h = mtod(m, struct ip *);
1674 /* IP header fields included in the TCP checksum */
1675 h->ip_p = IPPROTO_TCP;
1677 h->ip_src.s_addr = saddr->v4.s_addr;
1678 h->ip_dst.s_addr = daddr->v4.s_addr;
1680 th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip));
1685 h6 = mtod(m, struct ip6_hdr *);
1687 /* IP header fields included in the TCP checksum */
1688 h6->ip6_nxt = IPPROTO_TCP;
1689 h6->ip6_plen = htons(tlen);
1690 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr));
1691 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr));
1693 th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr));
1699 th->th_sport = sport;
1700 th->th_dport = dport;
1701 th->th_seq = htonl(seq);
1702 th->th_ack = htonl(ack);
1703 th->th_off = tlen >> 2;
1704 th->th_flags = flags;
1705 th->th_win = htons(win);
1708 opt = (char *)(th + 1);
1709 opt[0] = TCPOPT_MAXSEG;
1712 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2);
1719 th->th_sum = in_cksum(m, len);
1721 /* Finish the IP header */
1723 h->ip_hl = sizeof(*h) >> 2;
1724 h->ip_tos = IPTOS_LOWDELAY;
1726 h->ip_off = path_mtu_discovery ? IP_DF : 0;
1727 h->ip_ttl = ttl ? ttl : ip_defttl;
1730 ip_output(m, NULL, NULL, 0, NULL, NULL);
1734 struct ether_header *e = (void *)ro.ro_dst.sa_data;
1742 ro.ro_dst.sa_len = sizeof(ro.ro_dst);
1743 ro.ro_dst.sa_family = pseudo_AF_HDRCMPLT;
1744 bcopy(eh->ether_dhost, e->ether_shost, ETHER_ADDR_LEN);
1745 bcopy(eh->ether_shost, e->ether_dhost, ETHER_ADDR_LEN);
1746 e->ether_type = eh->ether_type;
1747 /* XXX_IMPORT: later */
1748 ip_output(m, (void *)NULL, &ro, 0,
1749 (void *)NULL, (void *)NULL);
1756 th->th_sum = in6_cksum(m, IPPROTO_TCP,
1757 sizeof(struct ip6_hdr), tlen);
1759 h6->ip6_vfc |= IPV6_VERSION;
1760 h6->ip6_hlim = IPV6_DEFHLIM;
1762 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
1769 pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af,
1774 m0 = m_copy(m, 0, M_COPYALL);
1775 m0->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
1777 if (r->rtableid >= 0)
1778 m0->m_pkthdr.pf.rtableid = r->rtableid;
1782 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
1783 m0->m_pkthdr.pf.qid = r->qid;
1784 m0->m_pkthdr.pf.ecn_af = af;
1785 m0->m_pkthdr.pf.hdr = mtod(m0, struct ip *);
1792 icmp_error(m0, type, code, 0, 0);
1797 icmp6_error(m0, type, code, 0);
1804 * Return 1 if the addresses a and b match (with mask m), otherwise return 0.
1805 * If n is 0, they match if they are equal. If n is != 0, they match if they
1809 pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m,
1810 struct pf_addr *b, sa_family_t af)
1817 if ((a->addr32[0] & m->addr32[0]) ==
1818 (b->addr32[0] & m->addr32[0]))
1824 if (((a->addr32[0] & m->addr32[0]) ==
1825 (b->addr32[0] & m->addr32[0])) &&
1826 ((a->addr32[1] & m->addr32[1]) ==
1827 (b->addr32[1] & m->addr32[1])) &&
1828 ((a->addr32[2] & m->addr32[2]) ==
1829 (b->addr32[2] & m->addr32[2])) &&
1830 ((a->addr32[3] & m->addr32[3]) ==
1831 (b->addr32[3] & m->addr32[3])))
1850 pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p)
1854 return ((p > a1) && (p < a2));
1856 return ((p < a1) || (p > a2));
1858 return ((p >= a1) && (p <= a2));
1872 return (0); /* never reached */
1876 pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p)
1881 return (pf_match(op, a1, a2, p));
1885 pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u)
1887 if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
1889 return (pf_match(op, a1, a2, u));
1893 pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g)
1895 if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
1897 return (pf_match(op, a1, a2, g));
1901 pf_match_tag(struct mbuf *m, struct pf_rule *r, int *tag)
1904 *tag = m->m_pkthdr.pf.tag;
1906 return ((!r->match_tag_not && r->match_tag == *tag) ||
1907 (r->match_tag_not && r->match_tag != *tag));
1911 pf_tag_packet(struct mbuf *m, int tag, int rtableid)
1913 if (tag <= 0 && rtableid < 0)
1917 m->m_pkthdr.pf.tag = tag;
1919 m->m_pkthdr.pf.rtableid = rtableid;
1925 pf_step_into_anchor(int *depth, struct pf_ruleset **rs, int n,
1926 struct pf_rule **r, struct pf_rule **a, int *match)
1928 struct pf_anchor_stackframe *f;
1930 (*r)->anchor->match = 0;
1933 if (*depth >= sizeof(pf_anchor_stack) /
1934 sizeof(pf_anchor_stack[0])) {
1935 kprintf("pf_step_into_anchor: stack overflow\n");
1936 *r = TAILQ_NEXT(*r, entries);
1938 } else if (*depth == 0 && a != NULL)
1940 f = pf_anchor_stack + (*depth)++;
1943 if ((*r)->anchor_wildcard) {
1944 f->parent = &(*r)->anchor->children;
1945 if ((f->child = RB_MIN(pf_anchor_node, f->parent)) ==
1950 *rs = &f->child->ruleset;
1954 *rs = &(*r)->anchor->ruleset;
1956 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
1960 pf_step_out_of_anchor(int *depth, struct pf_ruleset **rs, int n,
1961 struct pf_rule **r, struct pf_rule **a, int *match)
1963 struct pf_anchor_stackframe *f;
1969 f = pf_anchor_stack + *depth - 1;
1970 if (f->parent != NULL && f->child != NULL) {
1971 if (f->child->match ||
1972 (match != NULL && *match)) {
1973 f->r->anchor->match = 1;
1976 f->child = RB_NEXT(pf_anchor_node, f->parent, f->child);
1977 if (f->child != NULL) {
1978 *rs = &f->child->ruleset;
1979 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
1987 if (*depth == 0 && a != NULL)
1990 if (f->r->anchor->match || (match != NULL && *match))
1991 quick = f->r->quick;
1992 *r = TAILQ_NEXT(f->r, entries);
1993 } while (*r == NULL);
2000 pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr,
2001 struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af)
2006 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2007 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2011 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2012 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2013 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) |
2014 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]);
2015 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) |
2016 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]);
2017 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) |
2018 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]);
2024 pf_addr_inc(struct pf_addr *addr, sa_family_t af)
2029 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1);
2033 if (addr->addr32[3] == 0xffffffff) {
2034 addr->addr32[3] = 0;
2035 if (addr->addr32[2] == 0xffffffff) {
2036 addr->addr32[2] = 0;
2037 if (addr->addr32[1] == 0xffffffff) {
2038 addr->addr32[1] = 0;
2040 htonl(ntohl(addr->addr32[0]) + 1);
2043 htonl(ntohl(addr->addr32[1]) + 1);
2046 htonl(ntohl(addr->addr32[2]) + 1);
2049 htonl(ntohl(addr->addr32[3]) + 1);
2055 #define mix(a,b,c) \
2057 a -= b; a -= c; a ^= (c >> 13); \
2058 b -= c; b -= a; b ^= (a << 8); \
2059 c -= a; c -= b; c ^= (b >> 13); \
2060 a -= b; a -= c; a ^= (c >> 12); \
2061 b -= c; b -= a; b ^= (a << 16); \
2062 c -= a; c -= b; c ^= (b >> 5); \
2063 a -= b; a -= c; a ^= (c >> 3); \
2064 b -= c; b -= a; b ^= (a << 10); \
2065 c -= a; c -= b; c ^= (b >> 15); \
2069 * hash function based on bridge_hash in if_bridge.c
2072 pf_hash(struct pf_addr *inaddr, struct pf_addr *hash,
2073 struct pf_poolhashkey *key, sa_family_t af)
2075 u_int32_t a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0];
2080 a += inaddr->addr32[0];
2083 hash->addr32[0] = c + key->key32[2];
2088 a += inaddr->addr32[0];
2089 b += inaddr->addr32[2];
2091 hash->addr32[0] = c;
2092 a += inaddr->addr32[1];
2093 b += inaddr->addr32[3];
2096 hash->addr32[1] = c;
2097 a += inaddr->addr32[2];
2098 b += inaddr->addr32[1];
2101 hash->addr32[2] = c;
2102 a += inaddr->addr32[3];
2103 b += inaddr->addr32[0];
2106 hash->addr32[3] = c;
2113 pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr,
2114 struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn)
2116 unsigned char hash[16];
2117 struct pf_pool *rpool = &r->rpool;
2118 struct pf_addr *raddr = &rpool->cur->addr.v.a.addr;
2119 struct pf_addr *rmask = &rpool->cur->addr.v.a.mask;
2120 struct pf_pooladdr *acur = rpool->cur;
2121 struct pf_src_node k;
2123 if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR &&
2124 (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2126 PF_ACPY(&k.addr, saddr, af);
2127 if (r->rule_flag & PFRULE_RULESRCTRACK ||
2128 r->rpool.opts & PF_POOL_STICKYADDR)
2132 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
2133 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
2134 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) {
2135 PF_ACPY(naddr, &(*sn)->raddr, af);
2136 if (pf_status.debug >= PF_DEBUG_MISC) {
2137 kprintf("pf_map_addr: src tracking maps ");
2138 pf_print_host(&k.addr, 0, af);
2140 pf_print_host(naddr, 0, af);
2147 if (rpool->cur->addr.type == PF_ADDR_NOROUTE)
2149 if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2153 if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 &&
2154 (rpool->opts & PF_POOL_TYPEMASK) !=
2157 raddr = &rpool->cur->addr.p.dyn->pfid_addr4;
2158 rmask = &rpool->cur->addr.p.dyn->pfid_mask4;
2163 if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 &&
2164 (rpool->opts & PF_POOL_TYPEMASK) !=
2167 raddr = &rpool->cur->addr.p.dyn->pfid_addr6;
2168 rmask = &rpool->cur->addr.p.dyn->pfid_mask6;
2172 } else if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2173 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN)
2174 return (1); /* unsupported */
2176 raddr = &rpool->cur->addr.v.a.addr;
2177 rmask = &rpool->cur->addr.v.a.mask;
2180 switch (rpool->opts & PF_POOL_TYPEMASK) {
2182 PF_ACPY(naddr, raddr, af);
2184 case PF_POOL_BITMASK:
2185 PF_POOLMASK(naddr, raddr, rmask, saddr, af);
2187 case PF_POOL_RANDOM:
2188 if (init_addr != NULL && PF_AZERO(init_addr, af)) {
2192 rpool->counter.addr32[0] = htonl(karc4random());
2197 if (rmask->addr32[3] != 0xffffffff)
2198 rpool->counter.addr32[3] =
2199 htonl(karc4random());
2202 if (rmask->addr32[2] != 0xffffffff)
2203 rpool->counter.addr32[2] =
2204 htonl(karc4random());
2207 if (rmask->addr32[1] != 0xffffffff)
2208 rpool->counter.addr32[1] =
2209 htonl(karc4random());
2212 if (rmask->addr32[0] != 0xffffffff)
2213 rpool->counter.addr32[0] =
2214 htonl(karc4random());
2218 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2219 PF_ACPY(init_addr, naddr, af);
2222 PF_AINC(&rpool->counter, af);
2223 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2226 case PF_POOL_SRCHASH:
2227 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af);
2228 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af);
2230 case PF_POOL_ROUNDROBIN:
2231 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2232 if (!pfr_pool_get(rpool->cur->addr.p.tbl,
2233 &rpool->tblidx, &rpool->counter,
2234 &raddr, &rmask, af))
2236 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2237 if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2238 &rpool->tblidx, &rpool->counter,
2239 &raddr, &rmask, af))
2241 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af))
2245 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL)
2246 rpool->cur = TAILQ_FIRST(&rpool->list);
2247 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2249 if (pfr_pool_get(rpool->cur->addr.p.tbl,
2250 &rpool->tblidx, &rpool->counter,
2251 &raddr, &rmask, af)) {
2252 /* table contains no address of type 'af' */
2253 if (rpool->cur != acur)
2257 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2259 if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2260 &rpool->tblidx, &rpool->counter,
2261 &raddr, &rmask, af)) {
2262 /* table contains no address of type 'af' */
2263 if (rpool->cur != acur)
2268 raddr = &rpool->cur->addr.v.a.addr;
2269 rmask = &rpool->cur->addr.v.a.mask;
2270 PF_ACPY(&rpool->counter, raddr, af);
2274 PF_ACPY(naddr, &rpool->counter, af);
2275 if (init_addr != NULL && PF_AZERO(init_addr, af))
2276 PF_ACPY(init_addr, naddr, af);
2277 PF_AINC(&rpool->counter, af);
2281 PF_ACPY(&(*sn)->raddr, naddr, af);
2283 if (pf_status.debug >= PF_DEBUG_MISC &&
2284 (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2285 kprintf("pf_map_addr: selected address ");
2286 pf_print_host(naddr, 0, af);
2294 pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r,
2295 struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport,
2296 struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high,
2297 struct pf_src_node **sn)
2299 struct pf_state_key_cmp key;
2300 struct pf_addr init_addr;
2303 bzero(&init_addr, sizeof(init_addr));
2304 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2307 if (proto == IPPROTO_ICMP) {
2315 PF_ACPY(&key.ext.addr, daddr, key.af);
2316 PF_ACPY(&key.gwy.addr, naddr, key.af);
2317 key.ext.port = dport;
2320 * port search; start random, step;
2321 * similar 2 portloop in in_pcbbind
2323 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP ||
2324 proto == IPPROTO_ICMP)) {
2325 key.gwy.port = dport;
2326 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
2328 } else if (low == 0 && high == 0) {
2329 key.gwy.port = *nport;
2330 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
2332 } else if (low == high) {
2333 key.gwy.port = htons(low);
2334 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) {
2335 *nport = htons(low);
2347 cut = htonl(karc4random()) % (1 + high - low) + low;
2348 /* low <= cut <= high */
2349 for (tmp = cut; tmp <= high; ++(tmp)) {
2350 key.gwy.port = htons(tmp);
2351 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
2353 *nport = htons(tmp);
2357 for (tmp = cut - 1; tmp >= low; --(tmp)) {
2358 key.gwy.port = htons(tmp);
2359 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
2361 *nport = htons(tmp);
2367 switch (r->rpool.opts & PF_POOL_TYPEMASK) {
2368 case PF_POOL_RANDOM:
2369 case PF_POOL_ROUNDROBIN:
2370 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2374 case PF_POOL_SRCHASH:
2375 case PF_POOL_BITMASK:
2379 } while (! PF_AEQ(&init_addr, naddr, af) );
2381 return (1); /* none available */
2385 pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off,
2386 int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport,
2387 struct pf_addr *daddr, u_int16_t dport, int rs_num)
2389 struct pf_rule *r, *rm = NULL;
2390 struct pf_ruleset *ruleset = NULL;
2395 r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr);
2396 while (r && rm == NULL) {
2397 struct pf_rule_addr *src = NULL, *dst = NULL;
2398 struct pf_addr_wrap *xdst = NULL;
2400 if (r->action == PF_BINAT && direction == PF_IN) {
2402 if (r->rpool.cur != NULL)
2403 xdst = &r->rpool.cur->addr;
2410 if (pfi_kif_match(r->kif, kif) == r->ifnot)
2411 r = r->skip[PF_SKIP_IFP].ptr;
2412 else if (r->direction && r->direction != direction)
2413 r = r->skip[PF_SKIP_DIR].ptr;
2414 else if (r->af && r->af != pd->af)
2415 r = r->skip[PF_SKIP_AF].ptr;
2416 else if (r->proto && r->proto != pd->proto)
2417 r = r->skip[PF_SKIP_PROTO].ptr;
2418 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af,
2420 r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR :
2421 PF_SKIP_DST_ADDR].ptr;
2422 else if (src->port_op && !pf_match_port(src->port_op,
2423 src->port[0], src->port[1], sport))
2424 r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT :
2425 PF_SKIP_DST_PORT].ptr;
2426 else if (dst != NULL &&
2427 PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->neg, NULL))
2428 r = r->skip[PF_SKIP_DST_ADDR].ptr;
2429 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af,
2431 r = TAILQ_NEXT(r, entries);
2432 else if (dst != NULL && dst->port_op &&
2433 !pf_match_port(dst->port_op, dst->port[0],
2434 dst->port[1], dport))
2435 r = r->skip[PF_SKIP_DST_PORT].ptr;
2436 else if (r->match_tag && !pf_match_tag(m, r, &tag))
2437 r = TAILQ_NEXT(r, entries);
2438 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto !=
2439 IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m,
2440 off, pd->hdr.tcp), r->os_fingerprint)))
2441 r = TAILQ_NEXT(r, entries);
2445 if (r->rtableid >= 0)
2446 rtableid = r->rtableid;
2447 if (r->anchor == NULL) {
2450 pf_step_into_anchor(&asd, &ruleset, rs_num,
2454 pf_step_out_of_anchor(&asd, &ruleset, rs_num, &r,
2457 if (pf_tag_packet(m, tag, rtableid))
2459 if (rm != NULL && (rm->action == PF_NONAT ||
2460 rm->action == PF_NORDR || rm->action == PF_NOBINAT))
2466 pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction,
2467 struct pfi_kif *kif, struct pf_src_node **sn,
2468 struct pf_addr *saddr, u_int16_t sport,
2469 struct pf_addr *daddr, u_int16_t dport,
2470 struct pf_addr *naddr, u_int16_t *nport)
2472 struct pf_rule *r = NULL;
2474 if (direction == PF_OUT) {
2475 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2476 sport, daddr, dport, PF_RULESET_BINAT);
2478 r = pf_match_translation(pd, m, off, direction, kif,
2479 saddr, sport, daddr, dport, PF_RULESET_NAT);
2481 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2482 sport, daddr, dport, PF_RULESET_RDR);
2484 r = pf_match_translation(pd, m, off, direction, kif,
2485 saddr, sport, daddr, dport, PF_RULESET_BINAT);
2489 switch (r->action) {
2495 if (pf_get_sport(pd->af, pd->proto, r, saddr,
2496 daddr, dport, naddr, nport, r->rpool.proxy_port[0],
2497 r->rpool.proxy_port[1], sn)) {
2498 DPFPRINTF(PF_DEBUG_MISC,
2499 ("pf: NAT proxy port allocation "
2501 r->rpool.proxy_port[0],
2502 r->rpool.proxy_port[1]));
2507 switch (direction) {
2509 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){
2513 if (r->rpool.cur->addr.p.dyn->
2517 &r->rpool.cur->addr.p.dyn->
2519 &r->rpool.cur->addr.p.dyn->
2526 if (r->rpool.cur->addr.p.dyn->
2530 &r->rpool.cur->addr.p.dyn->
2532 &r->rpool.cur->addr.p.dyn->
2540 &r->rpool.cur->addr.v.a.addr,
2541 &r->rpool.cur->addr.v.a.mask,
2545 if (r->src.addr.type == PF_ADDR_DYNIFTL) {
2549 if (r->src.addr.p.dyn->
2553 &r->src.addr.p.dyn->
2555 &r->src.addr.p.dyn->
2562 if (r->src.addr.p.dyn->
2566 &r->src.addr.p.dyn->
2568 &r->src.addr.p.dyn->
2576 &r->src.addr.v.a.addr,
2577 &r->src.addr.v.a.mask, daddr,
2583 if (pf_map_addr(pd->af, r, saddr, naddr, NULL, sn))
2585 if ((r->rpool.opts & PF_POOL_TYPEMASK) ==
2587 PF_POOLMASK(naddr, naddr,
2588 &r->rpool.cur->addr.v.a.mask, daddr,
2591 if (r->rpool.proxy_port[1]) {
2592 u_int32_t tmp_nport;
2594 tmp_nport = ((ntohs(dport) -
2595 ntohs(r->dst.port[0])) %
2596 (r->rpool.proxy_port[1] -
2597 r->rpool.proxy_port[0] + 1)) +
2598 r->rpool.proxy_port[0];
2600 /* wrap around if necessary */
2601 if (tmp_nport > 65535)
2603 *nport = htons((u_int16_t)tmp_nport);
2604 } else if (r->rpool.proxy_port[0])
2605 *nport = htons(r->rpool.proxy_port[0]);
2617 struct netmsg_hashlookup {
2618 struct netmsg nm_netmsg;
2619 struct inpcb **nm_pinp;
2620 struct inpcbinfo *nm_pcbinfo;
2621 struct pf_addr *nm_saddr;
2622 struct pf_addr *nm_daddr;
2629 in_pcblookup_hash_handler(struct netmsg *msg0)
2631 struct netmsg_hashlookup *msg = (struct netmsg_hashlookup *)msg0;
2633 if (msg->nm_af == AF_INET)
2634 *msg->nm_pinp = in_pcblookup_hash(msg->nm_pcbinfo,
2635 msg->nm_saddr->v4, msg->nm_sport, msg->nm_daddr->v4,
2636 msg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2639 *msg->nm_pinp = in6_pcblookup_hash(msg->nm_pcbinfo,
2640 &msg->nm_saddr->v6, msg->nm_sport, &msg->nm_daddr->v6,
2641 msg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2643 lwkt_replymsg(&msg->nm_netmsg.nm_lmsg, 0);
2648 pf_socket_lookup(int direction, struct pf_pdesc *pd)
2650 struct pf_addr *saddr, *daddr;
2651 u_int16_t sport, dport;
2652 struct inpcbinfo *pi;
2655 struct netmsg_hashlookup *msg = NULL;
2661 pd->lookup.uid = UID_MAX;
2662 pd->lookup.gid = GID_MAX;
2663 pd->lookup.pid = NO_PID;
2664 if (direction == PF_IN) {
2671 switch (pd->proto) {
2673 if (pd->hdr.tcp == NULL)
2675 sport = pd->hdr.tcp->th_sport;
2676 dport = pd->hdr.tcp->th_dport;
2678 pi_cpu = tcp_addrcpu(saddr->v4.s_addr, sport, daddr->v4.s_addr, dport);
2679 pi = &tcbinfo[pi_cpu];
2682 * Our netstack runs lockless on MP systems
2683 * (only for TCP connections at the moment).
2685 * As we are not allowed to read another CPU's tcbinfo,
2686 * we have to ask that CPU via remote call to search the
2689 * Prepare a msg iff data belongs to another CPU.
2691 if (pi_cpu != mycpu->gd_cpuid) {
2692 msg = kmalloc(sizeof(*msg), M_LWKTMSG, M_INTWAIT);
2693 netmsg_init(&msg->nm_netmsg, NULL, &netisr_afree_rport,
2694 0, in_pcblookup_hash_handler);
2695 msg->nm_pinp = &inp;
2696 msg->nm_pcbinfo = pi;
2697 msg->nm_saddr = saddr;
2698 msg->nm_sport = sport;
2699 msg->nm_daddr = daddr;
2700 msg->nm_dport = dport;
2701 msg->nm_af = pd->af;
2706 if (pd->hdr.udp == NULL)
2708 sport = pd->hdr.udp->uh_sport;
2709 dport = pd->hdr.udp->uh_dport;
2715 if (direction != PF_IN) {
2727 * Query other CPU, second part
2729 * msg only gets initialized when:
2731 * 2) the info belongs to another CPU
2733 * Use some switch/case magic to avoid code duplication.
2738 inp = in6_pcblookup_hash(pi, &saddr->v6, sport,
2739 &daddr->v6, dport, INPLOOKUP_WILDCARD, NULL);
2745 /* FALLTHROUGH if SMP and on other CPU */
2750 lwkt_sendmsg(tcp_cport(pi_cpu),
2751 &msg->nm_netmsg.nm_lmsg);
2755 inp = in_pcblookup_hash(pi, saddr->v4, sport, daddr->v4,
2756 dport, INPLOOKUP_WILDCARD, NULL);
2765 pd->lookup.uid = inp->inp_socket->so_cred->cr_uid;
2766 pd->lookup.gid = inp->inp_socket->so_cred->cr_groups[0];
2771 pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2775 u_int8_t *opt, optlen;
2776 u_int8_t wscale = 0;
2778 hlen = th_off << 2; /* hlen <= sizeof(hdr) */
2779 if (hlen <= sizeof(struct tcphdr))
2781 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2783 opt = hdr + sizeof(struct tcphdr);
2784 hlen -= sizeof(struct tcphdr);
2794 if (wscale > TCP_MAX_WINSHIFT)
2795 wscale = TCP_MAX_WINSHIFT;
2796 wscale |= PF_WSCALE_FLAG;
2811 pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2815 u_int8_t *opt, optlen;
2816 u_int16_t mss = tcp_mssdflt;
2818 hlen = th_off << 2; /* hlen <= sizeof(hdr) */
2819 if (hlen <= sizeof(struct tcphdr))
2821 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2823 opt = hdr + sizeof(struct tcphdr);
2824 hlen -= sizeof(struct tcphdr);
2825 while (hlen >= TCPOLEN_MAXSEG) {
2833 bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2);
2848 pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer)
2851 struct sockaddr_in *dst;
2855 struct sockaddr_in6 *dst6;
2856 struct route_in6 ro6;
2858 struct rtentry *rt = NULL;
2860 u_int16_t mss = tcp_mssdflt;
2865 hlen = sizeof(struct ip);
2866 bzero(&ro, sizeof(ro));
2867 dst = (struct sockaddr_in *)&ro.ro_dst;
2868 dst->sin_family = AF_INET;
2869 dst->sin_len = sizeof(*dst);
2870 dst->sin_addr = addr->v4;
2871 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
2877 hlen = sizeof(struct ip6_hdr);
2878 bzero(&ro6, sizeof(ro6));
2879 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst;
2880 dst6->sin6_family = AF_INET6;
2881 dst6->sin6_len = sizeof(*dst6);
2882 dst6->sin6_addr = addr->v6;
2883 rtalloc_ign((struct route *)&ro6, (RTF_CLONING | RTF_PRCLONING));
2889 if (rt && rt->rt_ifp) {
2890 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr);
2891 mss = max(tcp_mssdflt, mss);
2894 mss = min(mss, offer);
2895 mss = max(mss, 64); /* sanity - at least max opt space */
2900 pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr)
2902 struct pf_rule *r = s->rule.ptr;
2905 if (!r->rt || r->rt == PF_FASTROUTE)
2907 switch (s->state_key->af) {
2910 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL,
2912 s->rt_kif = r->rpool.cur->kif;
2917 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL,
2919 s->rt_kif = r->rpool.cur->kif;
2926 pf_attach_state(struct pf_state_key *sk, struct pf_state *s, int tail)
2931 /* list is sorted, if-bound states before floating */
2933 TAILQ_INSERT_TAIL(&sk->states, s, next);
2935 TAILQ_INSERT_HEAD(&sk->states, s, next);
2939 pf_detach_state(struct pf_state *s, int flags)
2941 struct pf_state_key *sk = s->state_key;
2946 s->state_key = NULL;
2947 TAILQ_REMOVE(&sk->states, s, next);
2948 if (--sk->refcnt == 0) {
2949 if (!(flags & PF_DT_SKIP_EXTGWY))
2950 RB_REMOVE(pf_state_tree_ext_gwy,
2951 &pf_statetbl_ext_gwy, sk);
2952 if (!(flags & PF_DT_SKIP_LANEXT))
2953 RB_REMOVE(pf_state_tree_lan_ext,
2954 &pf_statetbl_lan_ext, sk);
2955 pool_put(&pf_state_key_pl, sk);
2959 struct pf_state_key *
2960 pf_alloc_state_key(struct pf_state *s)
2962 struct pf_state_key *sk;
2964 if ((sk = pool_get(&pf_state_key_pl, PR_NOWAIT)) == NULL)
2966 bzero(sk, sizeof(*sk));
2967 TAILQ_INIT(&sk->states);
2968 pf_attach_state(sk, s, 0);
2974 pf_test_rule(struct pf_rule **rm, struct pf_state **sm, int direction,
2975 struct pfi_kif *kif, struct mbuf *m, int off, void *h,
2976 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
2977 struct ifqueue *ifq, struct inpcb *inp)
2979 struct pf_rule *nr = NULL;
2980 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
2981 u_int16_t bport, nport = 0;
2982 sa_family_t af = pd->af;
2983 struct pf_rule *r, *a = NULL;
2984 struct pf_ruleset *ruleset = NULL;
2985 struct pf_src_node *nsn = NULL;
2986 struct tcphdr *th = pd->hdr.tcp;
2988 int rewrite = 0, hdrlen = 0;
2989 int tag = -1, rtableid = -1;
2993 u_int16_t mss = tcp_mssdflt;
2994 u_int16_t sport, dport;
2995 u_int8_t icmptype = 0, icmpcode = 0;
2997 if (direction == PF_IN && pf_check_congestion(ifq)) {
2998 REASON_SET(&reason, PFRES_CONGEST);
3003 pd->lookup.done = pf_socket_lookup(direction, pd);
3004 else if (debug_pfugidhack) {
3006 DPFPRINTF(PF_DEBUG_MISC, ("pf: unlocked lookup\n"));
3007 pd->lookup.done = pf_socket_lookup(direction, pd);
3011 sport = dport = hdrlen = 0;
3013 switch (pd->proto) {
3015 sport = th->th_sport;
3016 dport = th->th_dport;
3017 hdrlen = sizeof(*th);
3020 sport = pd->hdr.udp->uh_sport;
3021 dport = pd->hdr.udp->uh_dport;
3022 hdrlen = sizeof(*pd->hdr.udp);
3026 if (pd->af != AF_INET)
3028 sport = dport = pd->hdr.icmp->icmp_id;
3029 icmptype = pd->hdr.icmp->icmp_type;
3030 icmpcode = pd->hdr.icmp->icmp_code;
3032 if (icmptype == ICMP_UNREACH ||
3033 icmptype == ICMP_SOURCEQUENCH ||
3034 icmptype == ICMP_REDIRECT ||
3035 icmptype == ICMP_TIMXCEED ||
3036 icmptype == ICMP_PARAMPROB)
3041 case IPPROTO_ICMPV6:
3042 if (pd->af != AF_INET6)
3044 sport = dport = pd->hdr.icmp6->icmp6_id;
3045 hdrlen = sizeof(*pd->hdr.icmp6);
3046 icmptype = pd->hdr.icmp6->icmp6_type;
3047 icmpcode = pd->hdr.icmp6->icmp6_code;
3049 if (icmptype == ICMP6_DST_UNREACH ||
3050 icmptype == ICMP6_PACKET_TOO_BIG ||
3051 icmptype == ICMP6_TIME_EXCEEDED ||
3052 icmptype == ICMP6_PARAM_PROB)
3058 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3060 if (direction == PF_OUT) {
3061 bport = nport = sport;
3062 /* check outgoing packet for BINAT/NAT */
3063 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3064 saddr, sport, daddr, dport, &pd->naddr, &nport)) != NULL) {
3065 PF_ACPY(&pd->baddr, saddr, af);
3066 switch (pd->proto) {
3068 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
3069 &th->th_sum, &pd->naddr, nport, 0, af);
3070 sport = th->th_sport;
3074 pf_change_ap(saddr, &pd->hdr.udp->uh_sport,
3075 pd->ip_sum, &pd->hdr.udp->uh_sum,
3076 &pd->naddr, nport, 1, af);
3077 sport = pd->hdr.udp->uh_sport;
3082 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3083 pd->naddr.v4.s_addr, 0);
3084 pd->hdr.icmp->icmp_cksum = pf_cksum_fixup(
3085 pd->hdr.icmp->icmp_cksum, sport, nport, 0);
3086 pd->hdr.icmp->icmp_id = nport;
3087 m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
3091 case IPPROTO_ICMPV6:
3092 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum,
3101 pf_change_a(&saddr->v4.s_addr,
3102 pd->ip_sum, pd->naddr.v4.s_addr, 0);
3107 PF_ACPY(saddr, &pd->naddr, af);
3119 bport = nport = dport;
3120 /* check incoming packet for BINAT/RDR */
3121 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3122 saddr, sport, daddr, dport, &pd->naddr, &nport)) != NULL) {
3123 PF_ACPY(&pd->baddr, daddr, af);
3124 switch (pd->proto) {
3126 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
3127 &th->th_sum, &pd->naddr, nport, 0, af);
3128 dport = th->th_dport;
3132 pf_change_ap(daddr, &pd->hdr.udp->uh_dport,
3133 pd->ip_sum, &pd->hdr.udp->uh_sum,
3134 &pd->naddr, nport, 1, af);
3135 dport = pd->hdr.udp->uh_dport;
3140 pf_change_a(&daddr->v4.s_addr, pd->ip_sum,
3141 pd->naddr.v4.s_addr, 0);
3145 case IPPROTO_ICMPV6:
3146 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum,
3155 pf_change_a(&daddr->v4.s_addr,
3156 pd->ip_sum, pd->naddr.v4.s_addr, 0);
3161 PF_ACPY(daddr, &pd->naddr, af);
3176 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3177 r = r->skip[PF_SKIP_IFP].ptr;
3178 else if (r->direction && r->direction != direction)
3179 r = r->skip[PF_SKIP_DIR].ptr;
3180 else if (r->af && r->af != af)
3181 r = r->skip[PF_SKIP_AF].ptr;
3182 else if (r->proto && r->proto != pd->proto)
3183 r = r->skip[PF_SKIP_PROTO].ptr;
3184 else if (PF_MISMATCHAW(&r->src.addr, saddr, af,
3186 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3187 /* tcp/udp only. port_op always 0 in other cases */
3188 else if (r->src.port_op && !pf_match_port(r->src.port_op,
3189 r->src.port[0], r->src.port[1], sport))
3190 r = r->skip[PF_SKIP_SRC_PORT].ptr;
3191 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af,
3193 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3194 /* tcp/udp only. port_op always 0 in other cases */
3195 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
3196 r->dst.port[0], r->dst.port[1], dport))
3197 r = r->skip[PF_SKIP_DST_PORT].ptr;
3198 /* icmp only. type always 0 in other cases */
3199 else if (r->type && r->type != icmptype + 1)
3200 r = TAILQ_NEXT(r, entries);
3201 /* icmp only. type always 0 in other cases */
3202 else if (r->code && r->code != icmpcode + 1)
3203 r = TAILQ_NEXT(r, entries);
3204 else if (r->tos && !(r->tos == pd->tos))
3205 r = TAILQ_NEXT(r, entries);
3206 else if (r->rule_flag & PFRULE_FRAGMENT)
3207 r = TAILQ_NEXT(r, entries);
3208 else if (pd->proto == IPPROTO_TCP &&
3209 (r->flagset & th->th_flags) != r->flags)
3210 r = TAILQ_NEXT(r, entries);
3211 /* tcp/udp only. uid.op always 0 in other cases */
3212 else if (r->uid.op && (pd->lookup.done || (pd->lookup.done =
3213 pf_socket_lookup(direction, pd), 1)) &&
3214 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
3216 r = TAILQ_NEXT(r, entries);
3217 /* tcp/udp only. gid.op always 0 in other cases */
3218 else if (r->gid.op && (pd->lookup.done || (pd->lookup.done =
3219 pf_socket_lookup(direction, pd), 1)) &&
3220 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
3222 r = TAILQ_NEXT(r, entries);
3223 else if (r->prob && r->prob <= karc4random())
3224 r = TAILQ_NEXT(r, entries);
3225 else if (r->match_tag && !pf_match_tag(m, r, &tag))
3226 r = TAILQ_NEXT(r, entries);
3227 else if (r->os_fingerprint != PF_OSFP_ANY &&
3228 (pd->proto != IPPROTO_TCP || !pf_osfp_match(
3229 pf_osfp_fingerprint(pd, m, off, th),
3230 r->os_fingerprint)))
3231 r = TAILQ_NEXT(r, entries);
3235 if (r->rtableid >= 0)
3236 rtableid = r->rtableid;
3237 if (r->anchor == NULL) {
3244 r = TAILQ_NEXT(r, entries);
3246 pf_step_into_anchor(&asd, &ruleset,
3247 PF_RULESET_FILTER, &r, &a, &match);
3249 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3250 PF_RULESET_FILTER, &r, &a, &match))
3257 REASON_SET(&reason, PFRES_MATCH);
3259 if (r->log || (nr != NULL && nr->log)) {
3261 m_copyback(m, off, hdrlen, pd->hdr.any);
3262 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr,
3266 if ((r->action == PF_DROP) &&
3267 ((r->rule_flag & PFRULE_RETURNRST) ||
3268 (r->rule_flag & PFRULE_RETURNICMP) ||
3269 (r->rule_flag & PFRULE_RETURN))) {
3270 /* undo NAT changes, if they have taken place */
3272 if (direction == PF_OUT) {
3273 switch (pd->proto) {
3275 pf_change_ap(saddr, &th->th_sport,
3276 pd->ip_sum, &th->th_sum,
3277 &pd->baddr, bport, 0, af);
3278 sport = th->th_sport;
3283 &pd->hdr.udp->uh_sport, pd->ip_sum,
3284 &pd->hdr.udp->uh_sum, &pd->baddr,
3286 sport = pd->hdr.udp->uh_sport;
3291 case IPPROTO_ICMPV6:
3298 pf_change_a(&saddr->v4.s_addr,
3300 pd->baddr.v4.s_addr, 0);
3303 PF_ACPY(saddr, &pd->baddr, af);
3308 switch (pd->proto) {
3310 pf_change_ap(daddr, &th->th_dport,
3311 pd->ip_sum, &th->th_sum,
3312 &pd->baddr, bport, 0, af);
3313 dport = th->th_dport;
3318 &pd->hdr.udp->uh_dport, pd->ip_sum,
3319 &pd->hdr.udp->uh_sum, &pd->baddr,
3321 dport = pd->hdr.udp->uh_dport;
3326 case IPPROTO_ICMPV6:
3333 pf_change_a(&daddr->v4.s_addr,
3335 pd->baddr.v4.s_addr, 0);
3338 PF_ACPY(daddr, &pd->baddr, af);
3344 if (pd->proto == IPPROTO_TCP &&
3345 ((r->rule_flag & PFRULE_RETURNRST) ||
3346 (r->rule_flag & PFRULE_RETURN)) &&
3347 !(th->th_flags & TH_RST)) {
3348 u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
3349 struct ip *h = mtod(m, struct ip *);
3351 if (pf_check_proto_cksum(m, off,
3352 h->ip_len - off, IPPROTO_TCP, AF_INET))
3353 REASON_SET(&reason, PFRES_PROTCKSUM);
3355 if (th->th_flags & TH_SYN)
3357 if (th->th_flags & TH_FIN)
3359 pf_send_tcp(r, af, pd->dst,
3360 pd->src, th->th_dport, th->th_sport,
3361 ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0,
3362 r->return_ttl, 1, 0, pd->eh, kif->pfik_ifp);
3364 } else if ((af == AF_INET) && r->return_icmp)
3365 pf_send_icmp(m, r->return_icmp >> 8,
3366 r->return_icmp & 255, af, r);
3367 else if ((af == AF_INET6) && r->return_icmp6)
3368 pf_send_icmp(m, r->return_icmp6 >> 8,
3369 r->return_icmp6 & 255, af, r);
3372 if (r->action == PF_DROP)
3375 if (pf_tag_packet(m, tag, rtableid)) {
3376 REASON_SET(&reason, PFRES_MEMORY);
3380 if (!state_icmp && (r->keep_state || nr != NULL ||
3381 (pd->flags & PFDESC_TCP_NORM))) {
3382 /* create new state */
3384 struct pf_state *s = NULL;
3385 struct pf_state_key *sk = NULL;
3386 struct pf_src_node *sn = NULL;
3388 /* check maximums */
3389 if (r->max_states && (r->states >= r->max_states)) {
3390 pf_status.lcounters[LCNT_STATES]++;
3391 REASON_SET(&reason, PFRES_MAXSTATES);
3394 /* src node for filter rule */
3395 if ((r->rule_flag & PFRULE_SRCTRACK ||
3396 r->rpool.opts & PF_POOL_STICKYADDR) &&
3397 pf_insert_src_node(&sn, r, saddr, af) != 0) {
3398 REASON_SET(&reason, PFRES_SRCLIMIT);
3401 /* src node for translation rule */
3402 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3403 ((direction == PF_OUT &&
3404 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3405 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) {
3406 REASON_SET(&reason, PFRES_SRCLIMIT);
3409 s = pool_get(&pf_state_pl, PR_NOWAIT);
3411 REASON_SET(&reason, PFRES_MEMORY);
3413 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3414 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3415 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3416 pf_status.src_nodes--;
3417 pool_put(&pf_src_tree_pl, sn);
3419 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3421 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3422 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3423 pf_status.src_nodes--;
3424 pool_put(&pf_src_tree_pl, nsn);
3427 pool_put(&pf_state_key_pl, sk);
3431 bzero(s, sizeof(*s));
3433 s->nat_rule.ptr = nr;
3435 STATE_INC_COUNTERS(s);
3436 s->allow_opts = r->allow_opts;
3437 s->log = r->log & PF_LOG_ALL;
3439 s->log |= nr->log & PF_LOG_ALL;
3440 switch (pd->proto) {
3442 len = pd->tot_len - off - (th->th_off << 2);
3443 s->src.seqlo = ntohl(th->th_seq);
3444 s->src.seqhi = s->src.seqlo + len + 1;
3445 if ((th->th_flags & (TH_SYN|TH_ACK)) ==
3446 TH_SYN && r->keep_state == PF_STATE_MODULATE) {
3447 /* Generate sequence number modulator */
3448 while ((s->src.seqdiff =
3449 pf_new_isn(sk) - s->src.seqlo) == 0)
3451 pf_change_a(&th->th_seq, &th->th_sum,
3452 htonl(s->src.seqlo + s->src.seqdiff), 0);
3456 if (th->th_flags & TH_SYN) {
3458 s->src.wscale = pf_get_wscale(m, off,
3461 s->src.max_win = MAX(ntohs(th->th_win), 1);
3462 if (s->src.wscale & PF_WSCALE_MASK) {
3463 /* Remove scale factor from initial window */
3464 int win = s->src.max_win;
3465 win += 1 << (s->src.wscale & PF_WSCALE_MASK);
3466 s->src.max_win = (win - 1) >>
3467 (s->src.wscale & PF_WSCALE_MASK);
3469 if (th->th_flags & TH_FIN)
3473 s->src.state = TCPS_SYN_SENT;
3474 s->dst.state = TCPS_CLOSED;
3475 s->timeout = PFTM_TCP_FIRST_PACKET;
3478 s->src.state = PFUDPS_SINGLE;
3479 s->dst.state = PFUDPS_NO_TRAFFIC;
3480 s->timeout = PFTM_UDP_FIRST_PACKET;
3484 case IPPROTO_ICMPV6:
3486 s->timeout = PFTM_ICMP_FIRST_PACKET;
3489 s->src.state = PFOTHERS_SINGLE;
3490 s->dst.state = PFOTHERS_NO_TRAFFIC;
3491 s->timeout = PFTM_OTHER_FIRST_PACKET;
3494 s->creation = time_second;
3495 s->expire = time_second;
3499 s->src_node->states++;
3502 PF_ACPY(&nsn->raddr, &pd->naddr, af);
3503 s->nat_src_node = nsn;
3504 s->nat_src_node->states++;
3506 if (pd->proto == IPPROTO_TCP) {
3507 if ((pd->flags & PFDESC_TCP_NORM) &&
3508 pf_normalize_tcp_init(m, off, pd, th, &s->src,
3510 REASON_SET(&reason, PFRES_MEMORY);
3511 pf_src_tree_remove_state(s);
3512 STATE_DEC_COUNTERS(s);
3513 pool_put(&pf_state_pl, s);
3516 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
3517 pf_normalize_tcp_stateful(m, off, pd, &reason,
3518 th, s, &s->src, &s->dst, &rewrite)) {
3519 /* This really shouldn't happen!!! */
3520 DPFPRINTF(PF_DEBUG_URGENT,
3521 ("pf_normalize_tcp_stateful failed on "
3523 pf_normalize_tcp_cleanup(s);
3524 pf_src_tree_remove_state(s);
3525 STATE_DEC_COUNTERS(s);
3526 pool_put(&pf_state_pl, s);
3531 if ((sk = pf_alloc_state_key(s)) == NULL) {
3532 REASON_SET(&reason, PFRES_MEMORY);
3536 sk->proto = pd->proto;
3537 sk->direction = direction;
3539 if (direction == PF_OUT) {
3540 PF_ACPY(&sk->gwy.addr, saddr, af);
3541 PF_ACPY(&sk->ext.addr, daddr, af);
3542 switch (pd->proto) {
3545 case IPPROTO_ICMPV6:
3547 sk->gwy.port = nport;
3551 sk->gwy.port = sport;
3552 sk->ext.port = dport;
3555 PF_ACPY(&sk->lan.addr, &pd->baddr, af);
3556 sk->lan.port = bport;
3558 PF_ACPY(&sk->lan.addr, &sk->gwy.addr, af);
3559 sk->lan.port = sk->gwy.port;
3562 PF_ACPY(&sk->lan.addr, daddr, af);
3563 PF_ACPY(&sk->ext.addr, saddr, af);
3564 switch (pd->proto) {
3567 case IPPROTO_ICMPV6:
3569 sk->lan.port = nport;
3573 sk->lan.port = dport;
3574 sk->ext.port = sport;
3577 PF_ACPY(&sk->gwy.addr, &pd->baddr, af);
3578 sk->gwy.port = bport;
3580 PF_ACPY(&sk->gwy.addr, &sk->lan.addr, af);
3581 sk->gwy.port = sk->lan.port;
3585 s->hash = pf_state_hash(sk);
3586 s->pickup_mode = r->pickup_mode;
3588 pf_set_rt_ifp(s, saddr); /* needs s->state_key set */
3590 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3591 if (pd->proto == IPPROTO_TCP)
3592 pf_normalize_tcp_cleanup(s);
3593 REASON_SET(&reason, PFRES_STATEINS);
3594 pf_src_tree_remove_state(s);
3595 STATE_DEC_COUNTERS(s);
3596 pool_put(&pf_state_pl, s);
3604 if (pd->proto == IPPROTO_TCP &&
3605 (th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
3606 r->keep_state == PF_STATE_SYNPROXY) {
3607 s->src.state = PF_TCPS_PROXY_SRC;
3609 if (direction == PF_OUT) {
3610 pf_change_ap(saddr, &th->th_sport,
3611 pd->ip_sum, &th->th_sum, &pd->baddr,
3613 sport = th->th_sport;
3615 pf_change_ap(daddr, &th->th_dport,
3616 pd->ip_sum, &th->th_sum, &pd->baddr,
3618 sport = th->th_dport;
3621 s->src.seqhi = htonl(karc4random());
3622 /* Find mss option */
3623 mss = pf_get_mss(m, off, th->th_off, af);
3624 mss = pf_calc_mss(saddr, af, mss);
3625 mss = pf_calc_mss(daddr, af, mss);
3627 pf_send_tcp(r, af, daddr, saddr, th->th_dport,
3628 th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1,
3629 TH_SYN|TH_ACK, 0, s->src.mss, 0, 1, 0, NULL, NULL);
3630 REASON_SET(&reason, PFRES_SYNPROXY);
3631 return (PF_SYNPROXY_DROP);
3635 /* copy back packet headers if we performed NAT operations */
3637 m_copyback(m, off, hdrlen, pd->hdr.any);
3643 pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif,
3644 struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am,
3645 struct pf_ruleset **rsm)
3647 struct pf_rule *r, *a = NULL;
3648 struct pf_ruleset *ruleset = NULL;
3649 sa_family_t af = pd->af;
3655 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3658 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3659 r = r->skip[PF_SKIP_IFP].ptr;
3660 else if (r->direction && r->direction != direction)
3661 r = r->skip[PF_SKIP_DIR].ptr;
3662 else if (r->af && r->af != af)
3663 r = r->skip[PF_SKIP_AF].ptr;
3664 else if (r->proto && r->proto != pd->proto)
3665 r = r->skip[PF_SKIP_PROTO].ptr;
3666 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af,
3668 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3669 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af,
3671 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3672 else if (r->tos && !(r->tos == pd->tos))
3673 r = TAILQ_NEXT(r, entries);
3674 else if (r->os_fingerprint != PF_OSFP_ANY)
3675 r = TAILQ_NEXT(r, entries);
3676 else if (pd->proto == IPPROTO_UDP &&
3677 (r->src.port_op || r->dst.port_op))
3678 r = TAILQ_NEXT(r, entries);
3679 else if (pd->proto == IPPROTO_TCP &&
3680 (r->src.port_op || r->dst.port_op || r->flagset))
3681 r = TAILQ_NEXT(r, entries);
3682 else if ((pd->proto == IPPROTO_ICMP ||
3683 pd->proto == IPPROTO_ICMPV6) &&
3684 (r->type || r->code))
3685 r = TAILQ_NEXT(r, entries);
3686 else if (r->prob && r->prob <= karc4random())
3687 r = TAILQ_NEXT(r, entries);
3688 else if (r->match_tag && !pf_match_tag(m, r, &tag))
3689 r = TAILQ_NEXT(r, entries);
3691 if (r->anchor == NULL) {
3698 r = TAILQ_NEXT(r, entries);
3700 pf_step_into_anchor(&asd, &ruleset,
3701 PF_RULESET_FILTER, &r, &a, &match);
3703 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3704 PF_RULESET_FILTER, &r, &a, &match))
3711 REASON_SET(&reason, PFRES_MATCH);
3714 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset,
3717 if (r->action != PF_PASS)
3720 if (pf_tag_packet(m, tag, -1)) {
3721 REASON_SET(&reason, PFRES_MEMORY);
3729 pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
3730 struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
3733 struct pf_state_key_cmp key;
3734 struct tcphdr *th = pd->hdr.tcp;
3735 u_int16_t win = ntohs(th->th_win);
3736 u_int32_t ack, end, seq, orig_seq;
3740 struct pf_state_peer *src, *dst;
3743 key.proto = IPPROTO_TCP;
3744 if (direction == PF_IN) {
3745 PF_ACPY(&key.ext.addr, pd->src, key.af);
3746 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
3747 key.ext.port = th->th_sport;
3748 key.gwy.port = th->th_dport;
3750 PF_ACPY(&key.lan.addr, pd->src, key.af);
3751 PF_ACPY(&key.ext.addr, pd->dst, key.af);
3752 key.lan.port = th->th_sport;
3753 key.ext.port = th->th_dport;
3758 if (direction == (*state)->state_key->direction) {
3759 src = &(*state)->src;
3760 dst = &(*state)->dst;
3762 src = &(*state)->dst;
3763 dst = &(*state)->src;
3766 if ((*state)->src.state == PF_TCPS_PROXY_SRC) {
3767 if (direction != (*state)->state_key->direction) {
3768 REASON_SET(reason, PFRES_SYNPROXY);
3769 return (PF_SYNPROXY_DROP);
3771 if (th->th_flags & TH_SYN) {
3772 if (ntohl(th->th_seq) != (*state)->src.seqlo) {
3773 REASON_SET(reason, PFRES_SYNPROXY);
3776 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
3777 pd->src, th->th_dport, th->th_sport,
3778 (*state)->src.seqhi, ntohl(th->th_seq) + 1,
3779 TH_SYN|TH_ACK, 0, (*state)->src.mss, 0, 1,
3781 REASON_SET(reason, PFRES_SYNPROXY);
3782 return (PF_SYNPROXY_DROP);
3783 } else if (!(th->th_flags & TH_ACK) ||
3784 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
3785 (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
3786 REASON_SET(reason, PFRES_SYNPROXY);
3788 } else if ((*state)->src_node != NULL &&
3789 pf_src_connlimit(state)) {
3790 REASON_SET(reason, PFRES_SRCLIMIT);
3793 (*state)->src.state = PF_TCPS_PROXY_DST;
3795 if ((*state)->src.state == PF_TCPS_PROXY_DST) {
3796 struct pf_state_host *src, *dst;
3798 if (direction == PF_OUT) {
3799 src = &(*state)->state_key->gwy;
3800 dst = &(*state)->state_key->ext;
3802 src = &(*state)->state_key->ext;
3803 dst = &(*state)->state_key->lan;
3805 if (direction == (*state)->state_key->direction) {
3806 if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) ||
3807 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
3808 (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
3809 REASON_SET(reason, PFRES_SYNPROXY);
3812 (*state)->src.max_win = MAX(ntohs(th->th_win), 1);
3813 if ((*state)->dst.seqhi == 1)
3814 (*state)->dst.seqhi = htonl(karc4random());
3815 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
3816 &dst->addr, src->port, dst->port,
3817 (*state)->dst.seqhi, 0, TH_SYN, 0,
3818 (*state)->src.mss, 0, 0, (*state)->tag, NULL, NULL);
3819 REASON_SET(reason, PFRES_SYNPROXY);
3820 return (PF_SYNPROXY_DROP);
3821 } else if (((th->th_flags & (TH_SYN|TH_ACK)) !=
3823 (ntohl(th->th_ack) != (*state)->dst.seqhi + 1)) {
3824 REASON_SET(reason, PFRES_SYNPROXY);
3827 (*state)->dst.max_win = MAX(ntohs(th->th_win), 1);
3828 (*state)->dst.seqlo = ntohl(th->th_seq);
3829 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
3830 pd->src, th->th_dport, th->th_sport,
3831 ntohl(th->th_ack), ntohl(th->th_seq) + 1,
3832 TH_ACK, (*state)->src.max_win, 0, 0, 0,
3833 (*state)->tag, NULL, NULL);
3834 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
3835 &dst->addr, src->port, dst->port,
3836 (*state)->src.seqhi + 1, (*state)->src.seqlo + 1,
3837 TH_ACK, (*state)->dst.max_win, 0, 0, 1,
3839 (*state)->src.seqdiff = (*state)->dst.seqhi -
3840 (*state)->src.seqlo;
3841 (*state)->dst.seqdiff = (*state)->src.seqhi -
3842 (*state)->dst.seqlo;
3843 (*state)->src.seqhi = (*state)->src.seqlo +
3844 (*state)->dst.max_win;
3845 (*state)->dst.seqhi = (*state)->dst.seqlo +
3846 (*state)->src.max_win;
3847 (*state)->src.wscale = (*state)->dst.wscale = 0;
3848 (*state)->src.state = (*state)->dst.state =
3850 REASON_SET(reason, PFRES_SYNPROXY);
3851 return (PF_SYNPROXY_DROP);
3855 if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) {
3856 sws = src->wscale & PF_WSCALE_MASK;
3857 dws = dst->wscale & PF_WSCALE_MASK;
3862 * Sequence tracking algorithm from Guido van Rooij's paper:
3863 * http://www.madison-gurkha.com/publications/tcp_filtering/
3867 orig_seq = seq = ntohl(th->th_seq);
3868 if (src->seqlo == 0) {
3869 /* First packet from this end. Set its state */
3871 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) &&
3872 src->scrub == NULL) {
3873 if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) {
3874 REASON_SET(reason, PFRES_MEMORY);
3879 /* Deferred generation of sequence number modulator */
3880 if (dst->seqdiff && !src->seqdiff) {
3882 while ((src->seqdiff = pf_new_isn((struct pf_state_key *)&key) - seq) == 0)
3884 ack = ntohl(th->th_ack) - dst->seqdiff;
3885 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3887 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3890 ack = ntohl(th->th_ack);
3893 end = seq + pd->p_len;
3894 if (th->th_flags & TH_SYN) {
3896 (*state)->sync_flags |= PFSTATE_GOT_SYN2;
3897 if (dst->wscale & PF_WSCALE_FLAG) {
3898 src->wscale = pf_get_wscale(m, off, th->th_off,
3900 if (src->wscale & PF_WSCALE_FLAG) {
3901 /* Remove scale factor from initial
3903 sws = src->wscale & PF_WSCALE_MASK;
3904 win = ((u_int32_t)win + (1 << sws) - 1)
3906 dws = dst->wscale & PF_WSCALE_MASK;
3908 /* fixup other window */
3909 dst->max_win <<= dst->wscale &
3911 /* in case of a retrans SYN|ACK */
3916 if (th->th_flags & TH_FIN)
3920 if (src->state < TCPS_SYN_SENT)
3921 src->state = TCPS_SYN_SENT;
3924 * May need to slide the window (seqhi may have been set by
3925 * the crappy stack check or if we picked up the connection
3926 * after establishment)
3928 if (src->seqhi == 1 ||
3929 SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi))
3930 src->seqhi = end + MAX(1, dst->max_win << dws);
3931 if (win > src->max_win)
3935 ack = ntohl(th->th_ack) - dst->seqdiff;
3937 /* Modulate sequence numbers */
3938 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3940 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3943 end = seq + pd->p_len;
3944 if (th->th_flags & TH_SYN)
3946 if (th->th_flags & TH_FIN)
3950 if ((th->th_flags & TH_ACK) == 0) {
3951 /* Let it pass through the ack skew check */
3953 } else if ((ack == 0 &&
3954 (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) ||
3955 /* broken tcp stacks do not set ack */
3956 (dst->state < TCPS_SYN_SENT)) {
3958 * Many stacks (ours included) will set the ACK number in an
3959 * FIN|ACK if the SYN times out -- no sequence to ACK.
3965 /* Ease sequencing restrictions on no data packets */
3970 ackskew = dst->seqlo - ack;
3974 * Need to demodulate the sequence numbers in any TCP SACK options
3975 * (Selective ACK). We could optionally validate the SACK values
3976 * against the current ACK window, either forwards or backwards, but
3977 * I'm not confident that SACK has been implemented properly
3978 * everywhere. It wouldn't surprise me if several stacks accidently
3979 * SACK too far backwards of previously ACKed data. There really aren't
3980 * any security implications of bad SACKing unless the target stack
3981 * doesn't validate the option length correctly. Someone trying to
3982 * spoof into a TCP connection won't bother blindly sending SACK
3985 if (dst->seqdiff && (th->th_off << 2) > sizeof(struct tcphdr)) {
3986 if (pf_modulate_sack(m, off, pd, th, dst))
3991 #define MAXACKWINDOW (0xffff + 1500) /* 1500 is an arbitrary fudge factor */
3992 if (SEQ_GEQ(src->seqhi, end) &&
3993 /* Last octet inside other's window space */
3994 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) &&
3995 /* Retrans: not more than one window back */
3996 (ackskew >= -MAXACKWINDOW) &&
3997 /* Acking not more than one reassembled fragment backwards */
3998 (ackskew <= (MAXACKWINDOW << sws)) &&
3999 /* Acking not more than one window forward */
4000 ((th->th_flags & TH_RST) == 0 || orig_seq == src->seqlo ||
4001 (orig_seq == src->seqlo + 1) || (pd->flags & PFDESC_IP_REAS) == 0)) {
4002 /* Require an exact/+1 sequence match on resets when possible */
4004 if (dst->scrub || src->scrub) {
4005 if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4006 *state, src, dst, ©back))
4010 /* update max window */
4011 if (src->max_win < win)
4013 /* synchronize sequencing */
4014 if (SEQ_GT(end, src->seqlo))
4016 /* slide the window of what the other end can send */
4017 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4018 dst->seqhi = ack + MAX((win << sws), 1);
4022 if (th->th_flags & TH_SYN)
4023 if (src->state < TCPS_SYN_SENT)
4024 src->state = TCPS_SYN_SENT;
4025 if (th->th_flags & TH_FIN)
4026 if (src->state < TCPS_CLOSING)
4027 src->state = TCPS_CLOSING;
4028 if (th->th_flags & TH_ACK) {
4029 if (dst->state == TCPS_SYN_SENT) {
4030 dst->state = TCPS_ESTABLISHED;
4031 if (src->state == TCPS_ESTABLISHED &&
4032 (*state)->src_node != NULL &&
4033 pf_src_connlimit(state)) {
4034 REASON_SET(reason, PFRES_SRCLIMIT);
4037 } else if (dst->state == TCPS_CLOSING)
4038 dst->state = TCPS_FIN_WAIT_2;
4040 if (th->th_flags & TH_RST)
4041 src->state = dst->state = TCPS_TIME_WAIT;
4043 /* update expire time */
4044 (*state)->expire = time_second;
4045 if (src->state >= TCPS_FIN_WAIT_2 &&
4046 dst->state >= TCPS_FIN_WAIT_2)
4047 (*state)->timeout = PFTM_TCP_CLOSED;
4048 else if (src->state >= TCPS_CLOSING &&
4049 dst->state >= TCPS_CLOSING)
4050 (*state)->timeout = PFTM_TCP_FIN_WAIT;
4051 else if (src->state < TCPS_ESTABLISHED ||
4052 dst->state < TCPS_ESTABLISHED)
4053 (*state)->timeout = PFTM_TCP_OPENING;
4054 else if (src->state >= TCPS_CLOSING ||
4055 dst->state >= TCPS_CLOSING)
4056 (*state)->timeout = PFTM_TCP_CLOSING;
4058 (*state)->timeout = PFTM_TCP_ESTABLISHED;
4060 /* Fall through to PASS packet */
4062 } else if ((dst->state < TCPS_SYN_SENT ||
4063 dst->state >= TCPS_FIN_WAIT_2 ||
4064 src->state >= TCPS_FIN_WAIT_2) &&
4065 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) &&
4066 /* Within a window forward of the originating packet */
4067 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) {
4068 /* Within a window backward of the originating packet */
4071 * This currently handles three situations:
4072 * 1) Stupid stacks will shotgun SYNs before their peer
4074 * 2) When PF catches an already established stream (the
4075 * firewall rebooted, the state table was flushed, routes
4077 * 3) Packets get funky immediately after the connection
4078 * closes (this should catch Solaris spurious ACK|FINs
4079 * that web servers like to spew after a close)
4081 * This must be a little more careful than the above code
4082 * since packet floods will also be caught here. We don't
4083 * update the TTL here to mitigate the damage of a packet
4084 * flood and so the same code can handle awkward establishment
4085 * and a loosened connection close.
4086 * In the establishment case, a correct peer response will
4087 * validate the connection, go through the normal state code
4088 * and keep updating the state TTL.
4091 if (pf_status.debug >= PF_DEBUG_MISC) {
4092 kprintf("pf: loose state match: ");
4093 pf_print_state(*state);
4094 pf_print_flags(th->th_flags);
4095 kprintf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
4096 "pkts=%llu:%llu\n", seq, orig_seq, ack, pd->p_len,
4097 ackskew, (unsigned long long)(*state)->packets[0],
4098 (unsigned long long)(*state)->packets[1]);
4101 if (dst->scrub || src->scrub) {
4102 if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4103 *state, src, dst, ©back))
4107 /* update max window */
4108 if (src->max_win < win)
4110 /* synchronize sequencing */
4111 if (SEQ_GT(end, src->seqlo))
4113 /* slide the window of what the other end can send */
4114 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4115 dst->seqhi = ack + MAX((win << sws), 1);
4118 * Cannot set dst->seqhi here since this could be a shotgunned
4119 * SYN and not an already established connection.
4122 if (th->th_flags & TH_FIN)
4123 if (src->state < TCPS_CLOSING)
4124 src->state = TCPS_CLOSING;
4125 if (th->th_flags & TH_RST)
4126 src->state = dst->state = TCPS_TIME_WAIT;
4128 /* Fall through to PASS packet */
4130 } else if ((*state)->pickup_mode == PF_PICKUPS_HASHONLY ||
4131 ((*state)->pickup_mode == PF_PICKUPS_ENABLED &&
4132 ((*state)->sync_flags & PFSTATE_GOT_SYN_MASK) !=
4133 PFSTATE_GOT_SYN_MASK)) {
4135 * If pickup mode is hash only, do not fail on sequence checks.
4137 * If pickup mode is enabled and we did not see the SYN in
4138 * both direction, do not fail on sequence checks because
4139 * we do not have complete information on window scale.
4141 * Adjust expiration and fall through to PASS packet.
4142 * XXX Add a FIN check to reduce timeout?
4144 (*state)->expire = time_second;
4147 * Failure processing
4149 if ((*state)->dst.state == TCPS_SYN_SENT &&
4150 (*state)->src.state == TCPS_SYN_SENT) {
4151 /* Send RST for state mismatches during handshake */
4152 if (!(th->th_flags & TH_RST))
4153 pf_send_tcp((*state)->rule.ptr, pd->af,
4154 pd->dst, pd->src, th->th_dport,
4155 th->th_sport, ntohl(th->th_ack), 0,
4157 (*state)->rule.ptr->return_ttl, 1, 0,
4158 pd->eh, kif->pfik_ifp);
4162 } else if (pf_status.debug >= PF_DEBUG_MISC) {
4163 kprintf("pf: BAD state: ");
4164 pf_print_state(*state);
4165 pf_print_flags(th->th_flags);
4166 kprintf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
4167 "pkts=%llu:%llu dir=%s,%s\n",
4168 seq, orig_seq, ack, pd->p_len, ackskew,
4169 (unsigned long long)(*state)->packets[0],
4170 (unsigned long long)(*state)->packets[1],
4171 direction == PF_IN ? "in" : "out",
4172 direction == (*state)->state_key->direction ?
4174 kprintf("pf: State failure on: %c %c %c %c | %c %c\n",
4175 SEQ_GEQ(src->seqhi, end) ? ' ' : '1',
4176 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ?
4178 (ackskew >= -MAXACKWINDOW) ? ' ' : '3',
4179 (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4',
4180 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5',
4181 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6');
4183 REASON_SET(reason, PFRES_BADSTATE);
4187 /* Any packets which have gotten here are to be passed */
4189 /* translate source/destination address, if necessary */
4190 if (STATE_TRANSLATE((*state)->state_key)) {
4191 if (direction == PF_OUT) {
4192 pf_change_ap(pd->src, &th->th_sport, pd->ip_sum,
4193 &th->th_sum, &(*state)->state_key->gwy.addr,
4194 (*state)->state_key->gwy.port, 0, pd->af);
4197 * If we don't redispatch the packet will go into
4198 * the protocol stack on the wrong cpu for the
4199 * post-translated address.
4201 m->m_pkthdr.fw_flags |= FW_MBUF_REDISPATCH;
4202 pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum,
4203 &th->th_sum, &(*state)->state_key->lan.addr,
4204 (*state)->state_key->lan.port, 0, pd->af);
4206 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4207 } else if (copyback) {
4208 /* Copyback sequence modulation or stateful scrub changes */
4209 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4216 pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif,
4217 struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4219 struct pf_state_peer *src, *dst;
4220 struct pf_state_key_cmp key;
4221 struct udphdr *uh = pd->hdr.udp;
4224 key.proto = IPPROTO_UDP;
4225 if (direction == PF_IN) {
4226 PF_ACPY(&key.ext.addr, pd->src, key.af);
4227 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4228 key.ext.port = uh->uh_sport;
4229 key.gwy.port = uh->uh_dport;
4231 PF_ACPY(&key.lan.addr, pd->src, key.af);
4232 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4233 key.lan.port = uh->uh_sport;
4234 key.ext.port = uh->uh_dport;
4239 if (direction == (*state)->state_key->direction) {
4240 src = &(*state)->src;
4241 dst = &(*state)->dst;
4243 src = &(*state)->dst;
4244 dst = &(*state)->src;
4248 if (src->state < PFUDPS_SINGLE)
4249 src->state = PFUDPS_SINGLE;
4250 if (dst->state == PFUDPS_SINGLE)
4251 dst->state = PFUDPS_MULTIPLE;
4253 /* update expire time */
4254 (*state)->expire = time_second;
4255 if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE)
4256 (*state)->timeout = PFTM_UDP_MULTIPLE;
4258 (*state)->timeout = PFTM_UDP_SINGLE;
4260 /* translate source/destination address, if necessary */
4261 if (STATE_TRANSLATE((*state)->state_key)) {
4262 if (direction == PF_OUT) {
4263 pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum,
4264 &uh->uh_sum, &(*state)->state_key->gwy.addr,
4265 (*state)->state_key->gwy.port, 1, pd->af);
4268 * If we don't redispatch the packet will go into
4269 * the protocol stack on the wrong cpu for the
4270 * post-translated address.
4272 m->m_pkthdr.fw_flags |= FW_MBUF_REDISPATCH;
4273 pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum,
4274 &uh->uh_sum, &(*state)->state_key->lan.addr,
4275 (*state)->state_key->lan.port, 1, pd->af);
4277 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
4284 pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif,
4285 struct mbuf *m, int off, void *h, struct pf_pdesc *pd, u_short *reason)
4287 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
4288 u_int16_t icmpid = 0, *icmpsum;
4291 struct pf_state_key_cmp key;
4293 switch (pd->proto) {
4296 icmptype = pd->hdr.icmp->icmp_type;
4297 icmpid = pd->hdr.icmp->icmp_id;
4298 icmpsum = &pd->hdr.icmp->icmp_cksum;
4300 if (icmptype == ICMP_UNREACH ||
4301 icmptype == ICMP_SOURCEQUENCH ||
4302 icmptype == ICMP_REDIRECT ||
4303 icmptype == ICMP_TIMXCEED ||
4304 icmptype == ICMP_PARAMPROB)
4309 case IPPROTO_ICMPV6:
4310 icmptype = pd->hdr.icmp6->icmp6_type;
4311 icmpid = pd->hdr.icmp6->icmp6_id;
4312 icmpsum = &pd->hdr.icmp6->icmp6_cksum;
4314 if (icmptype == ICMP6_DST_UNREACH ||
4315 icmptype == ICMP6_PACKET_TOO_BIG ||
4316 icmptype == ICMP6_TIME_EXCEEDED ||
4317 icmptype == ICMP6_PARAM_PROB)
4326 * ICMP query/reply message not related to a TCP/UDP packet.
4327 * Search for an ICMP state.
4330 key.proto = pd->proto;
4331 if (direction == PF_IN) {
4332 PF_ACPY(&key.ext.addr, pd->src, key.af);
4333 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4335 key.gwy.port = icmpid;
4337 PF_ACPY(&key.lan.addr, pd->src, key.af);
4338 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4339 key.lan.port = icmpid;
4345 (*state)->expire = time_second;
4346 (*state)->timeout = PFTM_ICMP_ERROR_REPLY;
4348 /* translate source/destination address, if necessary */
4349 if (STATE_TRANSLATE((*state)->state_key)) {
4350 if (direction == PF_OUT) {
4354 pf_change_a(&saddr->v4.s_addr,
4356 (*state)->state_key->gwy.addr.v4.s_addr, 0);
4357 pd->hdr.icmp->icmp_cksum =
4359 pd->hdr.icmp->icmp_cksum, icmpid,
4360 (*state)->state_key->gwy.port, 0);
4361 pd->hdr.icmp->icmp_id =
4362 (*state)->state_key->gwy.port;
4363 m_copyback(m, off, ICMP_MINLEN,
4364 (caddr_t)pd->hdr.icmp);
4370 &pd->hdr.icmp6->icmp6_cksum,
4371 &(*state)->state_key->gwy.addr, 0);
4373 sizeof(struct icmp6_hdr),
4374 (caddr_t)pd->hdr.icmp6);
4382 pf_change_a(&daddr->v4.s_addr,
4384 (*state)->state_key->lan.addr.v4.s_addr, 0);
4385 pd->hdr.icmp->icmp_cksum =
4387 pd->hdr.icmp->icmp_cksum, icmpid,
4388 (*state)->state_key->lan.port, 0);
4389 pd->hdr.icmp->icmp_id =
4390 (*state)->state_key->lan.port;
4391 m_copyback(m, off, ICMP_MINLEN,
4392 (caddr_t)pd->hdr.icmp);
4398 &pd->hdr.icmp6->icmp6_cksum,
4399 &(*state)->state_key->lan.addr, 0);
4401 sizeof(struct icmp6_hdr),
4402 (caddr_t)pd->hdr.icmp6);
4413 * ICMP error message in response to a TCP/UDP packet.
4414 * Extract the inner TCP/UDP header and search for that state.
4417 struct pf_pdesc pd2;
4422 struct ip6_hdr h2_6;
4432 /* offset of h2 in mbuf chain */
4433 ipoff2 = off + ICMP_MINLEN;
4435 if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2),
4436 NULL, reason, pd2.af)) {
4437 DPFPRINTF(PF_DEBUG_MISC,
4438 ("pf: ICMP error message too short "
4443 * ICMP error messages don't refer to non-first
4446 if (h2.ip_off & htons(IP_OFFMASK)) {
4447 REASON_SET(reason, PFRES_FRAG);
4451 /* offset of protocol header that follows h2 */
4452 off2 = ipoff2 + (h2.ip_hl << 2);
4454 pd2.proto = h2.ip_p;
4455 pd2.src = (struct pf_addr *)&h2.ip_src;
4456 pd2.dst = (struct pf_addr *)&h2.ip_dst;
4457 pd2.ip_sum = &h2.ip_sum;
4462 ipoff2 = off + sizeof(struct icmp6_hdr);
4464 if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6),
4465 NULL, reason, pd2.af)) {
4466 DPFPRINTF(PF_DEBUG_MISC,
4467 ("pf: ICMP error message too short "
4471 pd2.proto = h2_6.ip6_nxt;
4472 pd2.src = (struct pf_addr *)&h2_6.ip6_src;
4473 pd2.dst = (struct pf_addr *)&h2_6.ip6_dst;
4475 off2 = ipoff2 + sizeof(h2_6);
4477 switch (pd2.proto) {
4478 case IPPROTO_FRAGMENT:
4480 * ICMPv6 error messages for
4481 * non-first fragments
4483 REASON_SET(reason, PFRES_FRAG);
4486 case IPPROTO_HOPOPTS:
4487 case IPPROTO_ROUTING:
4488 case IPPROTO_DSTOPTS: {
4489 /* get next header and header length */
4490 struct ip6_ext opt6;
4492 if (!pf_pull_hdr(m, off2, &opt6,
4493 sizeof(opt6), NULL, reason,
4495 DPFPRINTF(PF_DEBUG_MISC,
4496 ("pf: ICMPv6 short opt\n"));
4499 if (pd2.proto == IPPROTO_AH)
4500 off2 += (opt6.ip6e_len + 2) * 4;
4502 off2 += (opt6.ip6e_len + 1) * 8;
4503 pd2.proto = opt6.ip6e_nxt;
4504 /* goto the next header */
4511 } while (!terminal);
4515 DPFPRINTF(PF_DEBUG_MISC,
4516 ("pf: ICMP AF %d unknown (ip6)\n", pd->af));
4521 switch (pd2.proto) {
4525 struct pf_state_peer *src, *dst;
4530 * Only the first 8 bytes of the TCP header can be
4531 * expected. Don't access any TCP header fields after
4532 * th_seq, an ackskew test is not possible.
4534 if (!pf_pull_hdr(m, off2, &th, 8, NULL, reason,
4536 DPFPRINTF(PF_DEBUG_MISC,
4537 ("pf: ICMP error message too short "
4543 key.proto = IPPROTO_TCP;
4544 if (direction == PF_IN) {
4545 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4546 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4547 key.ext.port = th.th_dport;
4548 key.gwy.port = th.th_sport;
4550 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4551 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4552 key.lan.port = th.th_dport;
4553 key.ext.port = th.th_sport;
4558 if (direction == (*state)->state_key->direction) {
4559 src = &(*state)->dst;
4560 dst = &(*state)->src;
4562 src = &(*state)->src;
4563 dst = &(*state)->dst;
4566 if (src->wscale && dst->wscale)
4567 dws = dst->wscale & PF_WSCALE_MASK;
4571 /* Demodulate sequence number */
4572 seq = ntohl(th.th_seq) - src->seqdiff;
4574 pf_change_a(&th.th_seq, icmpsum,
4579 if (!SEQ_GEQ(src->seqhi, seq) ||
4580 !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws))) {
4581 if (pf_status.debug >= PF_DEBUG_MISC) {
4582 kprintf("pf: BAD ICMP %d:%d ",
4583 icmptype, pd->hdr.icmp->icmp_code);
4584 pf_print_host(pd->src, 0, pd->af);
4586 pf_print_host(pd->dst, 0, pd->af);
4587 kprintf(" state: ");
4588 pf_print_state(*state);
4589 kprintf(" seq=%u\n", seq);
4591 REASON_SET(reason, PFRES_BADSTATE);
4595 if (STATE_TRANSLATE((*state)->state_key)) {
4596 if (direction == PF_IN) {
4597 pf_change_icmp(pd2.src, &th.th_sport,
4598 daddr, &(*state)->state_key->lan.addr,
4599 (*state)->state_key->lan.port, NULL,
4600 pd2.ip_sum, icmpsum,
4601 pd->ip_sum, 0, pd2.af);
4603 pf_change_icmp(pd2.dst, &th.th_dport,
4604 saddr, &(*state)->state_key->gwy.addr,
4605 (*state)->state_key->gwy.port, NULL,
4606 pd2.ip_sum, icmpsum,
4607 pd->ip_sum, 0, pd2.af);
4616 m_copyback(m, off, ICMP_MINLEN,
4617 (caddr_t)pd->hdr.icmp);
4618 m_copyback(m, ipoff2, sizeof(h2),
4625 sizeof(struct icmp6_hdr),
4626 (caddr_t)pd->hdr.icmp6);
4627 m_copyback(m, ipoff2, sizeof(h2_6),
4632 m_copyback(m, off2, 8, (caddr_t)&th);
4641 if (!pf_pull_hdr(m, off2, &uh, sizeof(uh),
4642 NULL, reason, pd2.af)) {
4643 DPFPRINTF(PF_DEBUG_MISC,
4644 ("pf: ICMP error message too short "
4650 key.proto = IPPROTO_UDP;
4651 if (direction == PF_IN) {
4652 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4653 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4654 key.ext.port = uh.uh_dport;
4655 key.gwy.port = uh.uh_sport;
4657 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4658 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4659 key.lan.port = uh.uh_dport;
4660 key.ext.port = uh.uh_sport;
4665 if (STATE_TRANSLATE((*state)->state_key)) {
4666 if (direction == PF_IN) {
4667 pf_change_icmp(pd2.src, &uh.uh_sport,
4669 &(*state)->state_key->lan.addr,
4670 (*state)->state_key->lan.port,
4672 pd2.ip_sum, icmpsum,
4673 pd->ip_sum, 1, pd2.af);
4675 pf_change_icmp(pd2.dst, &uh.uh_dport,
4677 &(*state)->state_key->gwy.addr,
4678 (*state)->state_key->gwy.port, &uh.uh_sum,
4679 pd2.ip_sum, icmpsum,
4680 pd->ip_sum, 1, pd2.af);
4685 m_copyback(m, off, ICMP_MINLEN,
4686 (caddr_t)pd->hdr.icmp);
4687 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4693 sizeof(struct icmp6_hdr),
4694 (caddr_t)pd->hdr.icmp6);
4695 m_copyback(m, ipoff2, sizeof(h2_6),
4700 m_copyback(m, off2, sizeof(uh), (caddr_t)&uh);
4707 case IPPROTO_ICMP: {
4710 if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN,
4711 NULL, reason, pd2.af)) {
4712 DPFPRINTF(PF_DEBUG_MISC,
4713 ("pf: ICMP error message too short i"
4719 key.proto = IPPROTO_ICMP;
4720 if (direction == PF_IN) {
4721 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4722 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4724 key.gwy.port = iih.icmp_id;
4726 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4727 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4728 key.lan.port = iih.icmp_id;
4734 if (STATE_TRANSLATE((*state)->state_key)) {
4735 if (direction == PF_IN) {
4736 pf_change_icmp(pd2.src, &iih.icmp_id,
4738 &(*state)->state_key->lan.addr,
4739 (*state)->state_key->lan.port, NULL,
4740 pd2.ip_sum, icmpsum,
4741 pd->ip_sum, 0, AF_INET);
4743 pf_change_icmp(pd2.dst, &iih.icmp_id,
4745 &(*state)->state_key->gwy.addr,
4746 (*state)->state_key->gwy.port, NULL,
4747 pd2.ip_sum, icmpsum,
4748 pd->ip_sum, 0, AF_INET);
4750 m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
4751 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4752 m_copyback(m, off2, ICMP_MINLEN, (caddr_t)&iih);
4760 case IPPROTO_ICMPV6: {
4761 struct icmp6_hdr iih;
4763 if (!pf_pull_hdr(m, off2, &iih,
4764 sizeof(struct icmp6_hdr), NULL, reason, pd2.af)) {
4765 DPFPRINTF(PF_DEBUG_MISC,
4766 ("pf: ICMP error message too short "
4772 key.proto = IPPROTO_ICMPV6;
4773 if (direction == PF_IN) {
4774 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4775 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4777 key.gwy.port = iih.icmp6_id;
4779 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4780 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4781 key.lan.port = iih.icmp6_id;
4787 if (STATE_TRANSLATE((*state)->state_key)) {
4788 if (direction == PF_IN) {
4789 pf_change_icmp(pd2.src, &iih.icmp6_id,
4791 &(*state)->state_key->lan.addr,
4792 (*state)->state_key->lan.port, NULL,
4793 pd2.ip_sum, icmpsum,
4794 pd->ip_sum, 0, AF_INET6);
4796 pf_change_icmp(pd2.dst, &iih.icmp6_id,
4797 saddr, &(*state)->state_key->gwy.addr,
4798 (*state)->state_key->gwy.port, NULL,
4799 pd2.ip_sum, icmpsum,
4800 pd->ip_sum, 0, AF_INET6);
4802 m_copyback(m, off, sizeof(struct icmp6_hdr),
4803 (caddr_t)pd->hdr.icmp6);
4804 m_copyback(m, ipoff2, sizeof(h2_6), (caddr_t)&h2_6);
4805 m_copyback(m, off2, sizeof(struct icmp6_hdr),
4815 key.proto = pd2.proto;
4816 if (direction == PF_IN) {
4817 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4818 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4822 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4823 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4830 if (STATE_TRANSLATE((*state)->state_key)) {
4831 if (direction == PF_IN) {
4832 pf_change_icmp(pd2.src, NULL,
4834 &(*state)->state_key->lan.addr,
4836 pd2.ip_sum, icmpsum,
4837 pd->ip_sum, 0, pd2.af);
4839 pf_change_icmp(pd2.dst, NULL,
4841 &(*state)->state_key->gwy.addr,
4843 pd2.ip_sum, icmpsum,
4844 pd->ip_sum, 0, pd2.af);
4849 m_copyback(m, off, ICMP_MINLEN,
4850 (caddr_t)pd->hdr.icmp);
4851 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4857 sizeof(struct icmp6_hdr),
4858 (caddr_t)pd->hdr.icmp6);
4859 m_copyback(m, ipoff2, sizeof(h2_6),
4874 pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif,
4875 struct pf_pdesc *pd)
4877 struct pf_state_peer *src, *dst;
4878 struct pf_state_key_cmp key;
4881 key.proto = pd->proto;
4882 if (direction == PF_IN) {
4883 PF_ACPY(&key.ext.addr, pd->src, key.af);
4884 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4888 PF_ACPY(&key.lan.addr, pd->src, key.af);
4889 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4896 if (direction == (*state)->state_key->direction) {
4897 src = &(*state)->src;
4898 dst = &(*state)->dst;
4900 src = &(*state)->dst;
4901 dst = &(*state)->src;
4905 if (src->state < PFOTHERS_SINGLE)
4906 src->state = PFOTHERS_SINGLE;
4907 if (dst->state == PFOTHERS_SINGLE)
4908 dst->state = PFOTHERS_MULTIPLE;
4910 /* update expire time */
4911 (*state)->expire = time_second;
4912 if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE)
4913 (*state)->timeout = PFTM_OTHER_MULTIPLE;
4915 (*state)->timeout = PFTM_OTHER_SINGLE;
4917 /* translate source/destination address, if necessary */
4918 if (STATE_TRANSLATE((*state)->state_key)) {
4919 if (direction == PF_OUT)
4923 pf_change_a(&pd->src->v4.s_addr,
4925 (*state)->state_key->gwy.addr.v4.s_addr,
4932 &(*state)->state_key->gwy.addr, pd->af);
4940 pf_change_a(&pd->dst->v4.s_addr,
4942 (*state)->state_key->lan.addr.v4.s_addr,
4949 &(*state)->state_key->lan.addr, pd->af);
4959 * ipoff and off are measured from the start of the mbuf chain.
4960 * h must be at "ipoff" on the mbuf chain.
4963 pf_pull_hdr(struct mbuf *m, int off, void *p, int len,
4964 u_short *actionp, u_short *reasonp, sa_family_t af)
4969 struct ip *h = mtod(m, struct ip *);
4970 u_int16_t fragoff = (h->ip_off & IP_OFFMASK) << 3;
4974 ACTION_SET(actionp, PF_PASS);
4976 ACTION_SET(actionp, PF_DROP);
4977 REASON_SET(reasonp, PFRES_FRAG);
4981 if (m->m_pkthdr.len < off + len ||
4982 h->ip_len < off + len) {
4983 ACTION_SET(actionp, PF_DROP);
4984 REASON_SET(reasonp, PFRES_SHORT);
4992 struct ip6_hdr *h = mtod(m, struct ip6_hdr *);
4994 if (m->m_pkthdr.len < off + len ||
4995 (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) <
4996 (unsigned)(off + len)) {
4997 ACTION_SET(actionp, PF_DROP);
4998 REASON_SET(reasonp, PFRES_SHORT);
5005 m_copydata(m, off, len, p);
5010 pf_routable(struct pf_addr *addr, sa_family_t af, struct pfi_kif *kif)
5012 struct sockaddr_in *dst;
5016 struct sockaddr_in6 *dst6;
5017 struct route_in6 ro;
5021 struct radix_node *rn;
5026 bzero(&ro, sizeof(ro));
5029 dst = satosin(&ro.ro_dst);
5030 dst->sin_family = AF_INET;
5031 dst->sin_len = sizeof(*dst);
5032 dst->sin_addr = addr->v4;
5036 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
5037 dst6->sin6_family = AF_INET6;
5038 dst6->sin6_len = sizeof(*dst6);
5039 dst6->sin6_addr = addr->v6;
5046 /* Skip checks for ipsec interfaces */
5047 if (kif != NULL && kif->pfik_ifp->if_type == IFT_ENC)
5050 rtalloc_ign((struct route *)&ro, 0);
5052 if (ro.ro_rt != NULL) {
5053 /* No interface given, this is a no-route check */
5057 if (kif->pfik_ifp == NULL) {
5062 /* Perform uRPF check if passed input interface */
5064 rn = (struct radix_node *)ro.ro_rt;
5066 rt = (struct rtentry *)rn;
5069 if (kif->pfik_ifp == ifp)
5072 } while (check_mpath == 1 && rn != NULL && ret == 0);
5076 if (ro.ro_rt != NULL)
5082 pf_rtlabel_match(struct pf_addr *addr, sa_family_t af, struct pf_addr_wrap *aw)
5084 struct sockaddr_in *dst;
5086 struct sockaddr_in6 *dst6;
5087 struct route_in6 ro;
5093 bzero(&ro, sizeof(ro));
5096 dst = satosin(&ro.ro_dst);
5097 dst->sin_family = AF_INET;
5098 dst->sin_len = sizeof(*dst);
5099 dst->sin_addr = addr->v4;
5103 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
5104 dst6->sin6_family = AF_INET6;
5105 dst6->sin6_len = sizeof(*dst6);
5106 dst6->sin6_addr = addr->v6;
5113 rtalloc_ign((struct route *)&ro, (RTF_CLONING | RTF_PRCLONING));
5115 if (ro.ro_rt != NULL) {
5124 pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5125 struct pf_state *s, struct pf_pdesc *pd)
5127 struct mbuf *m0, *m1;
5128 struct route iproute;
5129 struct route *ro = NULL;
5130 struct sockaddr_in *dst;
5132 struct ifnet *ifp = NULL;
5133 struct pf_addr naddr;
5134 struct pf_src_node *sn = NULL;
5141 if (m == NULL || *m == NULL || r == NULL ||
5142 (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5143 panic("pf_route: invalid parameters");
5145 if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
5146 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
5147 (*m)->m_pkthdr.pf.routed = 1;
5149 if ((*m)->m_pkthdr.pf.routed++ > 3) {
5156 if (r->rt == PF_DUPTO) {
5157 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL)
5160 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5165 if (m0->m_len < sizeof(struct ip)) {
5166 DPFPRINTF(PF_DEBUG_URGENT,
5167 ("pf_route: m0->m_len < sizeof(struct ip)\n"));
5171 ip = mtod(m0, struct ip *);
5174 bzero((caddr_t)ro, sizeof(*ro));
5175 dst = satosin(&ro->ro_dst);
5176 dst->sin_family = AF_INET;
5177 dst->sin_len = sizeof(*dst);
5178 dst->sin_addr = ip->ip_dst;
5180 if (r->rt == PF_FASTROUTE) {
5182 if (ro->ro_rt == 0) {
5183 ipstat.ips_noroute++;
5187 ifp = ro->ro_rt->rt_ifp;
5188 ro->ro_rt->rt_use++;
5190 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
5191 dst = satosin(ro->ro_rt->rt_gateway);
5193 if (TAILQ_EMPTY(&r->rpool.list)) {
5194 DPFPRINTF(PF_DEBUG_URGENT,
5195 ("pf_route: TAILQ_EMPTY(&r->rpool.list)\n"));
5199 pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src,
5201 if (!PF_AZERO(&naddr, AF_INET))
5202 dst->sin_addr.s_addr = naddr.v4.s_addr;
5203 ifp = r->rpool.cur->kif ?
5204 r->rpool.cur->kif->pfik_ifp : NULL;
5206 if (!PF_AZERO(&s->rt_addr, AF_INET))
5207 dst->sin_addr.s_addr =
5208 s->rt_addr.v4.s_addr;
5209 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5217 if (pf_test(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) {
5220 } else if (m0 == NULL) {
5225 if (m0->m_len < sizeof(struct ip)) {
5226 DPFPRINTF(PF_DEBUG_URGENT,
5227 ("pf_route: m0->m_len < sizeof(struct ip)\n"));
5230 ip = mtod(m0, struct ip *);
5233 /* Copied from FreeBSD 5.1-CURRENT ip_output. */
5234 m0->m_pkthdr.csum_flags |= CSUM_IP;
5235 sw_csum = m0->m_pkthdr.csum_flags & ~ifp->if_hwassist;
5236 if (sw_csum & CSUM_DELAY_DATA) {
5237 in_delayed_cksum(m0);
5238 sw_csum &= ~CSUM_DELAY_DATA;
5240 m0->m_pkthdr.csum_flags &= ifp->if_hwassist;
5242 if (ip->ip_len <= ifp->if_mtu ||
5243 (ifp->if_hwassist & CSUM_FRAGMENT &&
5244 (ip->ip_off & IP_DF) == 0)) {
5245 ip->ip_len = htons(ip->ip_len);
5246 ip->ip_off = htons(ip->ip_off);
5248 if (sw_csum & CSUM_DELAY_IP) {
5250 if (ip->ip_v == IPVERSION &&
5251 (ip->ip_hl << 2) == sizeof(*ip)) {
5252 ip->ip_sum = in_cksum_hdr(ip);
5254 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
5258 error = ifp->if_output(ifp, m0, sintosa(dst), ro->ro_rt);
5264 * Too large for interface; fragment if possible.
5265 * Must be able to put at least 8 bytes per fragment.
5267 if (ip->ip_off & IP_DF) {
5268 ipstat.ips_cantfrag++;
5269 if (r->rt != PF_DUPTO) {
5271 icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0,
5280 error = ip_fragment(ip, &m0, ifp->if_mtu, ifp->if_hwassist, sw_csum);
5285 for (m0 = m1; m0; m0 = m1) {
5290 error = (*ifp->if_output)(ifp, m0, sintosa(dst),
5298 ipstat.ips_fragmented++;
5301 if (r->rt != PF_DUPTO)
5303 if (ro == &iproute && ro->ro_rt)
5315 pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5316 struct pf_state *s, struct pf_pdesc *pd)
5319 struct route_in6 ip6route;
5320 struct route_in6 *ro;
5321 struct sockaddr_in6 *dst;
5322 struct ip6_hdr *ip6;
5323 struct ifnet *ifp = NULL;
5324 struct pf_addr naddr;
5325 struct pf_src_node *sn = NULL;
5328 if (m == NULL || *m == NULL || r == NULL ||
5329 (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5330 panic("pf_route6: invalid parameters");
5332 if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
5333 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
5334 (*m)->m_pkthdr.pf.routed = 1;
5336 if ((*m)->m_pkthdr.pf.routed++ > 3) {
5343 if (r->rt == PF_DUPTO) {
5344 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL)
5347 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5352 if (m0->m_len < sizeof(struct ip6_hdr)) {
5353 DPFPRINTF(PF_DEBUG_URGENT,
5354 ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n"));
5357 ip6 = mtod(m0, struct ip6_hdr *);
5360 bzero((caddr_t)ro, sizeof(*ro));
5361 dst = (struct sockaddr_in6 *)&ro->ro_dst;
5362 dst->sin6_family = AF_INET6;
5363 dst->sin6_len = sizeof(*dst);
5364 dst->sin6_addr = ip6->ip6_dst;
5366 /* Cheat. XXX why only in the v6 case??? */
5367 if (r->rt == PF_FASTROUTE) {
5368 m0->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
5369 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
5374 if (TAILQ_EMPTY(&r->rpool.list)) {
5375 DPFPRINTF(PF_DEBUG_URGENT,
5376 ("pf_route6: TAILQ_EMPTY(&r->rpool.list)\n"));
5380 pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src,
5382 if (!PF_AZERO(&naddr, AF_INET6))
5383 PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5385 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL;
5387 if (!PF_AZERO(&s->rt_addr, AF_INET6))
5388 PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5389 &s->rt_addr, AF_INET6);
5390 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5397 if (pf_test6(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) {
5400 } else if (m0 == NULL) {
5405 if (m0->m_len < sizeof(struct ip6_hdr)) {
5406 DPFPRINTF(PF_DEBUG_URGENT,
5407 ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n"));
5410 ip6 = mtod(m0, struct ip6_hdr *);
5414 * If the packet is too large for the outgoing interface,
5415 * send back an icmp6 error.
5417 if (IN6_IS_ADDR_LINKLOCAL(&dst->sin6_addr))
5418 dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
5419 if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) {
5421 error = nd6_output(ifp, ifp, m0, dst, NULL);
5424 in6_ifstat_inc(ifp, ifs6_in_toobig);
5425 if (r->rt != PF_DUPTO) {
5427 icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
5434 if (r->rt != PF_DUPTO)
5447 * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag
5448 * off is the offset where the protocol header starts
5449 * len is the total length of protocol header plus payload
5450 * returns 0 when the checksum is valid, otherwise returns 1.
5454 * FreeBSD supports cksum offload for the following drivers.
5455 * em(4), gx(4), lge(4), nge(4), ti(4), xl(4)
5456 * If we can make full use of it we would outperform ipfw/ipfilter in
5457 * very heavy traffic.
5458 * I have not tested 'cause I don't have NICs that supports cksum offload.
5459 * (There might be problems. Typical phenomena would be
5460 * 1. No route message for UDP packet.
5461 * 2. No connection acceptance from external hosts regardless of rule set.)
5464 pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p,
5471 if (off < sizeof(struct ip) || len < sizeof(struct udphdr))
5473 if (m->m_pkthdr.len < off + len)
5479 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
5480 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) {
5481 sum = m->m_pkthdr.csum_data;
5483 ip = mtod(m, struct ip *);
5484 sum = in_pseudo(ip->ip_src.s_addr,
5485 ip->ip_dst.s_addr, htonl((u_short)len +
5486 m->m_pkthdr.csum_data + p));
5494 case IPPROTO_ICMPV6:
5504 if (p == IPPROTO_ICMP) {
5509 sum = in_cksum(m, len);
5513 if (m->m_len < sizeof(struct ip))
5515 sum = in_cksum_range(m, p, off, len);
5517 m->m_pkthdr.csum_flags |=
5520 m->m_pkthdr.csum_data = 0xffff;
5526 if (m->m_len < sizeof(struct ip6_hdr))
5528 sum = in6_cksum(m, p, off, len);
5531 * IPv6 H/W cksum off-load not supported yet!
5534 * m->m_pkthdr.csum_flags |=
5535 * (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
5536 * m->m_pkthdr.csum_data = 0xffff;
5548 tcpstat.tcps_rcvbadsum++;
5551 udpstat.udps_badsum++;
5554 icmpstat.icps_checksum++;
5557 case IPPROTO_ICMPV6:
5558 icmp6stat.icp6s_checksum++;
5569 pf_test(int dir, struct ifnet *ifp, struct mbuf **m0,
5570 struct ether_header *eh, struct inpcb *inp)
5572 struct pfi_kif *kif;
5573 u_short action, reason = 0, log = 0;
5574 struct mbuf *m = *m0;
5575 struct ip *h = NULL;
5576 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr;
5577 struct pf_state *s = NULL;
5578 struct pf_state_key *sk = NULL;
5579 struct pf_ruleset *ruleset = NULL;
5581 int off, dirndx, pqid = 0;
5583 if (!pf_status.running)
5586 memset(&pd, 0, sizeof(pd));
5587 if (ifp->if_type == IFT_CARP && ifp->if_carpdev)
5588 kif = (struct pfi_kif *)ifp->if_carpdev->if_pf_kif;
5590 kif = (struct pfi_kif *)ifp->if_pf_kif;
5593 DPFPRINTF(PF_DEBUG_URGENT,
5594 ("pf_test: kif == NULL, if_xname %s\n", ifp->if_xname));
5597 if (kif->pfik_flags & PFI_IFLAG_SKIP)
5601 if ((m->m_flags & M_PKTHDR) == 0)
5602 panic("non-M_PKTHDR is passed to pf_test");
5603 #endif /* DIAGNOSTIC */
5605 if (m->m_pkthdr.len < (int)sizeof(*h)) {
5607 REASON_SET(&reason, PFRES_SHORT);
5612 if (m->m_pkthdr.pf.flags & PF_TAG_GENERATED)
5615 /* We do IP header normalization and packet reassembly here */
5616 if (pf_normalize_ip(m0, dir, kif, &reason, &pd) != PF_PASS) {
5620 m = *m0; /* pf_normalize messes with m0 */
5621 h = mtod(m, struct ip *);
5623 off = h->ip_hl << 2;
5624 if (off < (int)sizeof(*h)) {
5626 REASON_SET(&reason, PFRES_SHORT);
5631 pd.src = (struct pf_addr *)&h->ip_src;
5632 pd.dst = (struct pf_addr *)&h->ip_dst;
5633 PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET);
5634 pd.ip_sum = &h->ip_sum;
5638 pd.tot_len = h->ip_len;
5641 /* handle fragments that didn't get reassembled by normalization */
5642 if (h->ip_off & (IP_MF | IP_OFFMASK)) {
5643 action = pf_test_fragment(&r, dir, kif, m, h,
5654 if (!pf_pull_hdr(m, off, &th, sizeof(th),
5655 &action, &reason, AF_INET)) {
5656 log = action != PF_PASS;
5659 pd.p_len = pd.tot_len - off - (th.th_off << 2);
5660 if ((th.th_flags & TH_ACK) && pd.p_len == 0)
5662 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
5663 if (action == PF_DROP)
5665 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
5667 if (action == PF_PASS) {
5669 pfsync_update_state(s);
5670 #endif /* NPFSYNC */
5674 } else if (s == NULL)
5675 action = pf_test_rule(&r, &s, dir, kif,
5676 m, off, h, &pd, &a, &ruleset, NULL, inp);
5684 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
5685 &action, &reason, AF_INET)) {
5686 log = action != PF_PASS;
5689 if (uh.uh_dport == 0 ||
5690 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
5691 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
5693 REASON_SET(&reason, PFRES_SHORT);
5696 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
5697 if (action == PF_PASS) {
5699 pfsync_update_state(s);
5700 #endif /* NPFSYNC */
5704 } else if (s == NULL)
5705 action = pf_test_rule(&r, &s, dir, kif,
5706 m, off, h, &pd, &a, &ruleset, NULL, inp);
5710 case IPPROTO_ICMP: {
5714 if (!pf_pull_hdr(m, off, &ih, ICMP_MINLEN,
5715 &action, &reason, AF_INET)) {
5716 log = action != PF_PASS;
5719 action = pf_test_state_icmp(&s, dir, kif, m, off, h, &pd,
5721 if (action == PF_PASS) {
5723 pfsync_update_state(s);
5724 #endif /* NPFSYNC */
5728 } else if (s == NULL)
5729 action = pf_test_rule(&r, &s, dir, kif,
5730 m, off, h, &pd, &a, &ruleset, NULL, inp);
5735 action = pf_test_state_other(&s, dir, kif, &pd);
5736 if (action == PF_PASS) {
5738 pfsync_update_state(s);
5739 #endif /* NPFSYNC */
5743 } else if (s == NULL)
5744 action = pf_test_rule(&r, &s, dir, kif, m, off, h,
5745 &pd, &a, &ruleset, NULL, inp);
5750 if (action == PF_PASS && h->ip_hl > 5 &&
5751 !((s && s->allow_opts) || r->allow_opts)) {
5753 REASON_SET(&reason, PFRES_IPOPTIONS);
5755 DPFPRINTF(PF_DEBUG_MISC,
5756 ("pf: dropping packet with ip options\n"));
5759 if ((s && s->tag) || r->rtableid)
5760 pf_tag_packet(m, s ? s->tag : 0, r->rtableid);
5763 if (action == PF_PASS && r->qid) {
5764 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
5765 if (pqid || (pd.tos & IPTOS_LOWDELAY))
5766 m->m_pkthdr.pf.qid = r->pqid;
5768 m->m_pkthdr.pf.qid = r->qid;
5769 m->m_pkthdr.pf.ecn_af = AF_INET;
5770 m->m_pkthdr.pf.hdr = h;
5771 /* add connection hash for fairq */
5774 m->m_pkthdr.pf.state_hash = s->hash;
5775 m->m_pkthdr.pf.flags |= PF_TAG_STATE_HASHED;
5781 * connections redirected to loopback should not match sockets
5782 * bound specifically to loopback due to security implications,
5783 * see tcp_input() and in_pcblookup_listen().
5785 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
5786 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
5787 (s->nat_rule.ptr->action == PF_RDR ||
5788 s->nat_rule.ptr->action == PF_BINAT) &&
5789 (ntohl(pd.dst->v4.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET)
5790 m->m_pkthdr.pf.flags |= PF_TAG_TRANSLATE_LOCALHOST;
5795 if (s != NULL && s->nat_rule.ptr != NULL &&
5796 s->nat_rule.ptr->log & PF_LOG_ALL)
5797 lr = s->nat_rule.ptr;
5800 PFLOG_PACKET(kif, h, m, AF_INET, dir, reason, lr, a, ruleset,
5804 kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
5805 kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++;
5807 if (action == PF_PASS || r->action == PF_DROP) {
5808 dirndx = (dir == PF_OUT);
5809 r->packets[dirndx]++;
5810 r->bytes[dirndx] += pd.tot_len;
5812 a->packets[dirndx]++;
5813 a->bytes[dirndx] += pd.tot_len;
5817 if (s->nat_rule.ptr != NULL) {
5818 s->nat_rule.ptr->packets[dirndx]++;
5819 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len;
5821 if (s->src_node != NULL) {
5822 s->src_node->packets[dirndx]++;
5823 s->src_node->bytes[dirndx] += pd.tot_len;
5825 if (s->nat_src_node != NULL) {
5826 s->nat_src_node->packets[dirndx]++;
5827 s->nat_src_node->bytes[dirndx] += pd.tot_len;
5829 dirndx = (dir == sk->direction) ? 0 : 1;
5830 s->packets[dirndx]++;
5831 s->bytes[dirndx] += pd.tot_len;
5834 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
5838 * XXX: we need to make sure that the addresses
5839 * passed to pfr_update_stats() are the same than
5840 * the addresses used during matching (pfr_match)
5842 if (r == &pf_default_rule) {
5844 x = (sk == NULL || sk->direction == dir) ?
5845 &pd.baddr : &pd.naddr;
5847 x = (sk == NULL || sk->direction == dir) ?
5848 &pd.naddr : &pd.baddr;
5849 if (x == &pd.baddr || s == NULL) {
5850 /* we need to change the address */
5857 if (tr->src.addr.type == PF_ADDR_TABLE)
5858 pfr_update_stats(tr->src.addr.p.tbl, (sk == NULL ||
5859 sk->direction == dir) ?
5860 pd.src : pd.dst, pd.af,
5861 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5863 if (tr->dst.addr.type == PF_ADDR_TABLE)
5864 pfr_update_stats(tr->dst.addr.p.tbl, (sk == NULL ||
5865 sk->direction == dir) ? pd.dst : pd.src, pd.af,
5866 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5871 if (action == PF_SYNPROXY_DROP) {
5876 /* pf_route can free the mbuf causing *m0 to become NULL */
5877 pf_route(m0, r, dir, kif->pfik_ifp, s, &pd);
5885 pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0,
5886 struct ether_header *eh, struct inpcb *inp)
5888 struct pfi_kif *kif;
5889 u_short action, reason = 0, log = 0;
5890 struct mbuf *m = *m0, *n = NULL;
5891 struct ip6_hdr *h = NULL;
5892 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr;
5893 struct pf_state *s = NULL;
5894 struct pf_state_key *sk = NULL;
5895 struct pf_ruleset *ruleset = NULL;
5897 int off, terminal = 0, dirndx, rh_cnt = 0;
5899 if (!pf_status.running)
5902 memset(&pd, 0, sizeof(pd));
5903 if (ifp->if_type == IFT_CARP && ifp->if_carpdev)
5904 kif = (struct pfi_kif *)ifp->if_carpdev->if_pf_kif;
5906 kif = (struct pfi_kif *)ifp->if_pf_kif;
5909 DPFPRINTF(PF_DEBUG_URGENT,
5910 ("pf_test6: kif == NULL, if_xname %s\n", ifp->if_xname));
5913 if (kif->pfik_flags & PFI_IFLAG_SKIP)
5917 if ((m->m_flags & M_PKTHDR) == 0)
5918 panic("non-M_PKTHDR is passed to pf_test6");
5919 #endif /* DIAGNOSTIC */
5921 if (m->m_pkthdr.len < (int)sizeof(*h)) {
5923 REASON_SET(&reason, PFRES_SHORT);
5928 if (m->m_pkthdr.pf.flags & PF_TAG_GENERATED)
5931 /* We do IP header normalization and packet reassembly here */
5932 if (pf_normalize_ip6(m0, dir, kif, &reason, &pd) != PF_PASS) {
5936 m = *m0; /* pf_normalize messes with m0 */
5937 h = mtod(m, struct ip6_hdr *);
5941 * we do not support jumbogram yet. if we keep going, zero ip6_plen
5942 * will do something bad, so drop the packet for now.
5944 if (htons(h->ip6_plen) == 0) {
5946 REASON_SET(&reason, PFRES_NORM); /*XXX*/
5951 pd.src = (struct pf_addr *)&h->ip6_src;
5952 pd.dst = (struct pf_addr *)&h->ip6_dst;
5953 PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET6);
5957 pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr);
5960 off = ((caddr_t)h - m->m_data) + sizeof(struct ip6_hdr);
5961 pd.proto = h->ip6_nxt;
5964 case IPPROTO_FRAGMENT:
5965 action = pf_test_fragment(&r, dir, kif, m, h,
5967 if (action == PF_DROP)
5968 REASON_SET(&reason, PFRES_FRAG);
5970 case IPPROTO_ROUTING: {
5971 struct ip6_rthdr rthdr;
5974 DPFPRINTF(PF_DEBUG_MISC,
5975 ("pf: IPv6 more than one rthdr\n"));
5977 REASON_SET(&reason, PFRES_IPOPTIONS);
5981 if (!pf_pull_hdr(m, off, &rthdr, sizeof(rthdr), NULL,
5983 DPFPRINTF(PF_DEBUG_MISC,
5984 ("pf: IPv6 short rthdr\n"));
5986 REASON_SET(&reason, PFRES_SHORT);
5990 if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) {
5991 DPFPRINTF(PF_DEBUG_MISC,
5992 ("pf: IPv6 rthdr0\n"));
5994 REASON_SET(&reason, PFRES_IPOPTIONS);
6001 case IPPROTO_HOPOPTS:
6002 case IPPROTO_DSTOPTS: {
6003 /* get next header and header length */
6004 struct ip6_ext opt6;
6006 if (!pf_pull_hdr(m, off, &opt6, sizeof(opt6),
6007 NULL, &reason, pd.af)) {
6008 DPFPRINTF(PF_DEBUG_MISC,
6009 ("pf: IPv6 short opt\n"));
6014 if (pd.proto == IPPROTO_AH)
6015 off += (opt6.ip6e_len + 2) * 4;
6017 off += (opt6.ip6e_len + 1) * 8;
6018 pd.proto = opt6.ip6e_nxt;
6019 /* goto the next header */
6026 } while (!terminal);
6028 /* if there's no routing header, use unmodified mbuf for checksumming */
6038 if (!pf_pull_hdr(m, off, &th, sizeof(th),
6039 &action, &reason, AF_INET6)) {
6040 log = action != PF_PASS;
6043 pd.p_len = pd.tot_len - off - (th.th_off << 2);
6044 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
6045 if (action == PF_DROP)
6047 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
6049 if (action == PF_PASS) {
6051 pfsync_update_state(s);
6052 #endif /* NPFSYNC */
6056 } else if (s == NULL)
6057 action = pf_test_rule(&r, &s, dir, kif,
6058 m, off, h, &pd, &a, &ruleset, NULL, inp);
6066 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
6067 &action, &reason, AF_INET6)) {
6068 log = action != PF_PASS;
6071 if (uh.uh_dport == 0 ||
6072 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
6073 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
6075 REASON_SET(&reason, PFRES_SHORT);
6078 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
6079 if (action == PF_PASS) {
6081 pfsync_update_state(s);
6082 #endif /* NPFSYNC */
6086 } else if (s == NULL)
6087 action = pf_test_rule(&r, &s, dir, kif,
6088 m, off, h, &pd, &a, &ruleset, NULL, inp);
6092 case IPPROTO_ICMPV6: {
6093 struct icmp6_hdr ih;
6096 if (!pf_pull_hdr(m, off, &ih, sizeof(ih),
6097 &action, &reason, AF_INET6)) {
6098 log = action != PF_PASS;
6101 action = pf_test_state_icmp(&s, dir, kif,
6102 m, off, h, &pd, &reason);
6103 if (action == PF_PASS) {
6105 pfsync_update_state(s);
6106 #endif /* NPFSYNC */
6110 } else if (s == NULL)
6111 action = pf_test_rule(&r, &s, dir, kif,
6112 m, off, h, &pd, &a, &ruleset, NULL, inp);
6117 action = pf_test_state_other(&s, dir, kif, &pd);
6118 if (action == PF_PASS) {
6120 pfsync_update_state(s);
6121 #endif /* NPFSYNC */
6125 } else if (s == NULL)
6126 action = pf_test_rule(&r, &s, dir, kif, m, off, h,
6127 &pd, &a, &ruleset, NULL, inp);
6137 /* handle dangerous IPv6 extension headers. */
6138 if (action == PF_PASS && rh_cnt &&
6139 !((s && s->allow_opts) || r->allow_opts)) {
6141 REASON_SET(&reason, PFRES_IPOPTIONS);
6143 DPFPRINTF(PF_DEBUG_MISC,
6144 ("pf: dropping packet with dangerous v6 headers\n"));
6147 if ((s && s->tag) || r->rtableid)
6148 pf_tag_packet(m, s ? s->tag : 0, r->rtableid);
6151 if (action == PF_PASS && r->qid) {
6152 m->m_pkthdr.fw_flags |= PF_MBUF_STRUCTURE;
6153 if (pd.tos & IPTOS_LOWDELAY)
6154 m->m_pkthdr.pf.qid = r->pqid;
6156 m->m_pkthdr.pf.qid = r->qid;
6157 m->m_pkthdr.pf.ecn_af = AF_INET6;
6158 m->m_pkthdr.pf.hdr = h;
6161 m->m_pkthdr.pf.state_hash = s->hash;
6162 m->m_pkthdr.pf.flags |= PF_TAG_STATE_HASHED;
6167 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
6168 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
6169 (s->nat_rule.ptr->action == PF_RDR ||
6170 s->nat_rule.ptr->action == PF_BINAT) &&
6171 IN6_IS_ADDR_LOOPBACK(&pd.dst->v6))
6172 m->m_pkthdr.pf.flags |= PF_TAG_TRANSLATE_LOCALHOST;
6177 if (s != NULL && s->nat_rule.ptr != NULL &&
6178 s->nat_rule.ptr->log & PF_LOG_ALL)
6179 lr = s->nat_rule.ptr;
6182 PFLOG_PACKET(kif, h, m, AF_INET6, dir, reason, lr, a, ruleset,
6186 kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
6187 kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++;
6189 if (action == PF_PASS || r->action == PF_DROP) {
6190 dirndx = (dir == PF_OUT);
6191 r->packets[dirndx]++;
6192 r->bytes[dirndx] += pd.tot_len;
6194 a->packets[dirndx]++;
6195 a->bytes[dirndx] += pd.tot_len;
6199 if (s->nat_rule.ptr != NULL) {
6200 s->nat_rule.ptr->packets[dirndx]++;
6201 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len;
6203 if (s->src_node != NULL) {
6204 s->src_node->packets[dirndx]++;
6205 s->src_node->bytes[dirndx] += pd.tot_len;
6207 if (s->nat_src_node != NULL) {
6208 s->nat_src_node->packets[dirndx]++;
6209 s->nat_src_node->bytes[dirndx] += pd.tot_len;
6211 dirndx = (dir == sk->direction) ? 0 : 1;
6212 s->packets[dirndx]++;
6213 s->bytes[dirndx] += pd.tot_len;
6216 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
6220 * XXX: we need to make sure that the addresses
6221 * passed to pfr_update_stats() are the same than
6222 * the addresses used during matching (pfr_match)
6224 if (r == &pf_default_rule) {
6226 x = (s == NULL || sk->direction == dir) ?
6227 &pd.baddr : &pd.naddr;
6229 x = (s == NULL || sk->direction == dir) ?
6230 &pd.naddr : &pd.baddr;
6232 if (x == &pd.baddr || s == NULL) {
6239 if (tr->src.addr.type == PF_ADDR_TABLE)
6240 pfr_update_stats(tr->src.addr.p.tbl, (sk == NULL ||
6241 sk->direction == dir) ? pd.src : pd.dst, pd.af,
6242 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
6244 if (tr->dst.addr.type == PF_ADDR_TABLE)
6245 pfr_update_stats(tr->dst.addr.p.tbl, (sk == NULL ||
6246 sk->direction == dir) ? pd.dst : pd.src, pd.af,
6247 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
6252 if (action == PF_SYNPROXY_DROP) {
6257 /* pf_route6 can free the mbuf causing *m0 to become NULL */
6258 pf_route6(m0, r, dir, kif->pfik_ifp, s, &pd);
6265 pf_check_congestion(struct ifqueue *ifq)