1 /* $FreeBSD: src/sys/contrib/pf/net/pf.c,v 1.19 2004/09/11 11:18:25 mlaier Exp $ */
2 /* $OpenBSD: pf.c,v 1.433.2.2 2004/07/17 03:22:34 brad Exp $ */
3 /* add $OpenBSD: pf.c,v 1.448 2004/05/11 07:34:11 dhartmei Exp $ */
4 /* $DragonFly: src/sys/net/pf/pf.c,v 1.20 2008/06/05 18:06:32 swildner Exp $ */
5 /* $OpenBSD: pf.c,v 1.527 2007/02/22 15:23:23 pyr Exp $ */
8 * Copyright (c) 2004 The DragonFly Project. All rights reserved.
10 * Copyright (c) 2001 Daniel Hartmeier
11 * Copyright (c) 2002,2003 Henning Brauer
12 * All rights reserved.
14 * Redistribution and use in source and binary forms, with or without
15 * modification, are permitted provided that the following conditions
18 * - Redistributions of source code must retain the above copyright
19 * notice, this list of conditions and the following disclaimer.
20 * - Redistributions in binary form must reproduce the above
21 * copyright notice, this list of conditions and the following
22 * disclaimer in the documentation and/or other materials provided
23 * with the distribution.
25 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
26 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
27 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
28 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
29 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
30 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
31 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
32 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
33 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
35 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36 * POSSIBILITY OF SUCH DAMAGE.
38 * Effort sponsored in part by the Defense Advanced Research Projects
39 * Agency (DARPA) and Air Force Research Laboratory, Air Force
40 * Materiel Command, USAF, under agreement number F30602-01-2-0537.
45 #include "opt_inet6.h"
46 #include "use_pfsync.h"
48 #include <sys/param.h>
49 #include <sys/systm.h>
50 #include <sys/malloc.h>
52 #include <sys/filio.h>
53 #include <sys/socket.h>
54 #include <sys/socketvar.h>
55 #include <sys/kernel.h>
57 #include <sys/sysctl.h>
58 #include <sys/endian.h>
59 #include <vm/vm_zone.h>
61 #include <sys/kthread.h>
63 #include <sys/mplock2.h>
65 #include <machine/inttypes.h>
68 #include <net/if_types.h>
70 #include <net/netisr.h>
71 #include <net/route.h>
73 #include <netinet/in.h>
74 #include <netinet/in_var.h>
75 #include <netinet/in_systm.h>
76 #include <netinet/ip.h>
77 #include <netinet/ip_var.h>
78 #include <netinet/tcp.h>
79 #include <netinet/tcp_seq.h>
80 #include <netinet/udp.h>
81 #include <netinet/ip_icmp.h>
82 #include <netinet/in_pcb.h>
83 #include <netinet/tcp_timer.h>
84 #include <netinet/tcp_var.h>
85 #include <netinet/udp_var.h>
86 #include <netinet/icmp_var.h>
87 #include <netinet/if_ether.h>
89 #include <net/pf/pfvar.h>
90 #include <net/pf/if_pflog.h>
93 #include <net/pf/if_pfsync.h>
94 #endif /* NPFSYNC > 0 */
97 #include <netinet/ip6.h>
98 #include <netinet/in_pcb.h>
99 #include <netinet/icmp6.h>
100 #include <netinet6/nd6.h>
101 #include <netinet6/ip6_var.h>
102 #include <netinet6/in6_pcb.h>
105 #include <sys/in_cksum.h>
106 #include <sys/ucred.h>
107 #include <machine/limits.h>
108 #include <sys/msgport2.h>
109 #include <net/netmsg2.h>
111 extern int ip_optcopy(struct ip *, struct ip *);
112 extern int debug_pfugidhack;
114 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) kprintf x
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;
139 vm_zone_t pf_state_pl, pf_altq_pl, pf_pooladdr_pl;
141 void pf_print_host(struct pf_addr *, u_int16_t, u_int8_t);
143 void pf_init_threshold(struct pf_threshold *, u_int32_t,
145 void pf_add_threshold(struct pf_threshold *);
146 int pf_check_threshold(struct pf_threshold *);
148 void pf_change_ap(struct pf_addr *, u_int16_t *,
149 u_int16_t *, u_int16_t *, struct pf_addr *,
150 u_int16_t, u_int8_t, sa_family_t);
151 int pf_modulate_sack(struct mbuf *, int, struct pf_pdesc *,
152 struct tcphdr *, struct pf_state_peer *);
154 void pf_change_a6(struct pf_addr *, u_int16_t *,
155 struct pf_addr *, u_int8_t);
157 void pf_change_icmp(struct pf_addr *, u_int16_t *,
158 struct pf_addr *, struct pf_addr *, u_int16_t,
159 u_int16_t *, u_int16_t *, u_int16_t *,
160 u_int16_t *, u_int8_t, sa_family_t);
161 void pf_send_tcp(const struct pf_rule *, sa_family_t,
162 const struct pf_addr *, const struct pf_addr *,
163 u_int16_t, u_int16_t, u_int32_t, u_int32_t,
164 u_int8_t, u_int16_t, u_int16_t, u_int8_t, int,
165 u_int16_t, struct ether_header *, struct ifnet *);
166 void pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t,
167 sa_family_t, struct pf_rule *);
168 struct pf_rule *pf_match_translation(struct pf_pdesc *, struct mbuf *,
169 int, int, struct pfi_kif *,
170 struct pf_addr *, u_int16_t, struct pf_addr *,
172 struct pf_rule *pf_get_translation(struct pf_pdesc *, struct mbuf *,
173 int, int, struct pfi_kif *, struct pf_src_node **,
174 struct pf_addr *, u_int16_t,
175 struct pf_addr *, u_int16_t,
176 struct pf_addr *, u_int16_t *);
177 int pf_test_tcp(struct pf_rule **, struct pf_state **,
178 int, struct pfi_kif *, struct mbuf *, int,
179 void *, struct pf_pdesc *, struct pf_rule **,
180 struct pf_ruleset **, struct ifqueue *, struct inpcb *);
181 int pf_test_udp(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_icmp(struct pf_rule **, struct pf_state **,
186 int, struct pfi_kif *, struct mbuf *, int,
187 void *, struct pf_pdesc *, struct pf_rule **,
188 struct pf_ruleset **, struct ifqueue *);
189 int pf_test_other(struct pf_rule **, struct pf_state **,
190 int, struct pfi_kif *, struct mbuf *, int, void *,
191 struct pf_pdesc *, struct pf_rule **,
192 struct pf_ruleset **, struct ifqueue *);
193 int pf_test_fragment(struct pf_rule **, int,
194 struct pfi_kif *, struct mbuf *, void *,
195 struct pf_pdesc *, struct pf_rule **,
196 struct pf_ruleset **);
197 int pf_test_state_tcp(struct pf_state **, int,
198 struct pfi_kif *, struct mbuf *, int,
199 void *, struct pf_pdesc *, u_short *);
200 int pf_test_state_udp(struct pf_state **, int,
201 struct pfi_kif *, struct mbuf *, int,
202 void *, struct pf_pdesc *);
203 int pf_test_state_icmp(struct pf_state **, int,
204 struct pfi_kif *, struct mbuf *, int,
205 void *, struct pf_pdesc *, u_short *);
206 int pf_test_state_other(struct pf_state **, int,
207 struct pfi_kif *, struct pf_pdesc *);
208 int pf_match_tag(struct mbuf *, struct pf_rule *,
209 struct pf_mtag *, int *);
210 int pf_step_out_of_anchor(int *, struct pf_ruleset **,
211 int, struct pf_rule **, struct pf_rule **,
213 void pf_hash(struct pf_addr *, struct pf_addr *,
214 struct pf_poolhashkey *, sa_family_t);
215 int pf_map_addr(u_int8_t, struct pf_rule *,
216 struct pf_addr *, struct pf_addr *,
217 struct pf_addr *, struct pf_src_node **);
218 int pf_get_sport(sa_family_t, u_int8_t, struct pf_rule *,
219 struct pf_addr *, struct pf_addr *, u_int16_t,
220 struct pf_addr *, u_int16_t*, u_int16_t, u_int16_t,
221 struct pf_src_node **);
222 void pf_route(struct mbuf **, struct pf_rule *, int,
223 struct ifnet *, struct pf_state *,
225 void pf_route6(struct mbuf **, struct pf_rule *, int,
226 struct ifnet *, struct pf_state *,
228 u_int8_t pf_get_wscale(struct mbuf *, int, u_int16_t,
230 u_int16_t pf_get_mss(struct mbuf *, int, u_int16_t,
232 u_int16_t pf_calc_mss(struct pf_addr *, sa_family_t,
234 void pf_set_rt_ifp(struct pf_state *,
236 int pf_check_proto_cksum(struct mbuf *, int, int,
237 u_int8_t, sa_family_t);
238 int pf_addr_wrap_neq(struct pf_addr_wrap *,
239 struct pf_addr_wrap *);
240 struct pf_state *pf_find_state_recurse(struct pfi_kif *,
241 struct pf_state_cmp *, u_int8_t);
242 int pf_src_connlimit(struct pf_state **);
243 int pf_check_congestion(struct ifqueue *);
245 extern int pf_end_threads;
247 struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] = {
248 { &pf_state_pl, PFSTATE_HIWAT },
249 { &pf_src_tree_pl, PFSNODE_HIWAT },
250 { &pf_frent_pl, PFFRAG_FRENT_HIWAT },
251 { &pfr_ktable_pl, PFR_KTABLE_HIWAT },
252 { &pfr_kentry_pl, PFR_KENTRY_HIWAT }
255 #define STATE_LOOKUP() \
257 if (direction == PF_IN) \
258 *state = pf_find_state_recurse( \
259 kif, &key, PF_EXT_GWY); \
261 *state = pf_find_state_recurse( \
262 kif, &key, PF_LAN_EXT); \
263 if (*state == NULL || (*state)->timeout == PFTM_PURGE) \
265 if (direction == PF_OUT && \
266 (((*state)->rule.ptr->rt == PF_ROUTETO && \
267 (*state)->rule.ptr->direction == PF_OUT) || \
268 ((*state)->rule.ptr->rt == PF_REPLYTO && \
269 (*state)->rule.ptr->direction == PF_IN)) && \
270 (*state)->rt_kif != NULL && \
271 (*state)->rt_kif != kif) \
275 #define STATE_TRANSLATE(s) \
276 (s)->lan.addr.addr32[0] != (s)->gwy.addr.addr32[0] || \
277 ((s)->af == AF_INET6 && \
278 ((s)->lan.addr.addr32[1] != (s)->gwy.addr.addr32[1] || \
279 (s)->lan.addr.addr32[2] != (s)->gwy.addr.addr32[2] || \
280 (s)->lan.addr.addr32[3] != (s)->gwy.addr.addr32[3])) || \
281 (s)->lan.port != (s)->gwy.port
283 #define BOUND_IFACE(r, k) \
284 ((r)->rule_flag & PFRULE_IFBOUND) ? (k) : pfi_all
286 #define STATE_INC_COUNTERS(s) \
288 s->rule.ptr->states++; \
289 if (s->anchor.ptr != NULL) \
290 s->anchor.ptr->states++; \
291 if (s->nat_rule.ptr != NULL) \
292 s->nat_rule.ptr->states++; \
295 #define STATE_DEC_COUNTERS(s) \
297 if (s->nat_rule.ptr != NULL) \
298 s->nat_rule.ptr->states--; \
299 if (s->anchor.ptr != NULL) \
300 s->anchor.ptr->states--; \
301 s->rule.ptr->states--; \
304 static __inline int pf_src_compare(struct pf_src_node *, struct pf_src_node *);
305 static __inline int pf_state_compare_lan_ext(struct pf_state *,
307 static __inline int pf_state_compare_ext_gwy(struct pf_state *,
309 static __inline int pf_state_compare_id(struct pf_state *,
312 struct pf_src_tree tree_src_tracking;
314 struct pf_state_tree_id tree_id;
315 struct pf_state_queue state_list;
317 RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare);
318 RB_GENERATE(pf_state_tree_lan_ext, pf_state,
319 u.s.entry_lan_ext, pf_state_compare_lan_ext);
320 RB_GENERATE(pf_state_tree_ext_gwy, pf_state,
321 u.s.entry_ext_gwy, pf_state_compare_ext_gwy);
322 RB_GENERATE(pf_state_tree_id, pf_state,
323 u.s.entry_id, pf_state_compare_id);
326 pf_src_compare(struct pf_src_node *a, struct pf_src_node *b)
330 if (a->rule.ptr > b->rule.ptr)
332 if (a->rule.ptr < b->rule.ptr)
334 if ((diff = a->af - b->af) != 0)
339 if (a->addr.addr32[0] > b->addr.addr32[0])
341 if (a->addr.addr32[0] < b->addr.addr32[0])
347 if (a->addr.addr32[3] > b->addr.addr32[3])
349 if (a->addr.addr32[3] < b->addr.addr32[3])
351 if (a->addr.addr32[2] > b->addr.addr32[2])
353 if (a->addr.addr32[2] < b->addr.addr32[2])
355 if (a->addr.addr32[1] > b->addr.addr32[1])
357 if (a->addr.addr32[1] < b->addr.addr32[1])
359 if (a->addr.addr32[0] > b->addr.addr32[0])
361 if (a->addr.addr32[0] < b->addr.addr32[0])
370 pf_state_hash(struct pf_state *s)
372 u_int32_t hv = (intptr_t)s / sizeof(*s);
374 hv ^= crc32(&s->lan, sizeof(s->lan));
375 hv ^= crc32(&s->gwy, sizeof(s->gwy));
376 hv ^= crc32(&s->ext, sizeof(s->ext));
377 if (hv == 0) /* disallow 0 */
383 pf_state_compare_lan_ext(struct pf_state *a, struct pf_state *b)
387 if ((diff = a->proto - b->proto) != 0)
389 if ((diff = a->af - b->af) != 0)
394 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
396 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
398 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
400 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
406 if (a->lan.addr.addr32[3] > b->lan.addr.addr32[3])
408 if (a->lan.addr.addr32[3] < b->lan.addr.addr32[3])
410 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
412 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
414 if (a->lan.addr.addr32[2] > b->lan.addr.addr32[2])
416 if (a->lan.addr.addr32[2] < b->lan.addr.addr32[2])
418 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
420 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
422 if (a->lan.addr.addr32[1] > b->lan.addr.addr32[1])
424 if (a->lan.addr.addr32[1] < b->lan.addr.addr32[1])
426 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
428 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
430 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
432 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
434 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
436 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
442 if ((diff = a->lan.port - b->lan.port) != 0)
444 if ((diff = a->ext.port - b->ext.port) != 0)
451 pf_state_compare_ext_gwy(struct pf_state *a, struct pf_state *b)
455 if ((diff = a->proto - b->proto) != 0)
457 if ((diff = a->af - b->af) != 0)
462 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
464 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
466 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
468 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
474 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
476 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
478 if (a->gwy.addr.addr32[3] > b->gwy.addr.addr32[3])
480 if (a->gwy.addr.addr32[3] < b->gwy.addr.addr32[3])
482 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
484 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
486 if (a->gwy.addr.addr32[2] > b->gwy.addr.addr32[2])
488 if (a->gwy.addr.addr32[2] < b->gwy.addr.addr32[2])
490 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
492 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
494 if (a->gwy.addr.addr32[1] > b->gwy.addr.addr32[1])
496 if (a->gwy.addr.addr32[1] < b->gwy.addr.addr32[1])
498 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
500 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
502 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
504 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
510 if ((diff = a->ext.port - b->ext.port) != 0)
512 if ((diff = a->gwy.port - b->gwy.port) != 0)
519 pf_state_compare_id(struct pf_state *a, struct pf_state *b)
525 if (a->creatorid > b->creatorid)
527 if (a->creatorid < b->creatorid)
535 pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af)
540 dst->addr32[0] = src->addr32[0];
544 dst->addr32[0] = src->addr32[0];
545 dst->addr32[1] = src->addr32[1];
546 dst->addr32[2] = src->addr32[2];
547 dst->addr32[3] = src->addr32[3];
554 pf_find_state_byid(struct pf_state_cmp *key)
556 pf_status.fcounters[FCNT_STATE_SEARCH]++;
557 return (RB_FIND(pf_state_tree_id, &tree_id, (struct pf_state *)key));
561 pf_find_state_recurse(struct pfi_kif *kif, struct pf_state_cmp *key, u_int8_t tree)
565 pf_status.fcounters[FCNT_STATE_SEARCH]++;
569 if ((s = RB_FIND(pf_state_tree_lan_ext, &kif->pfik_lan_ext,
570 (struct pf_state *)key)) != NULL)
572 if ((s = RB_FIND(pf_state_tree_lan_ext, &pfi_all->pfik_lan_ext,
573 (struct pf_state *)key)) != NULL)
577 if ((s = RB_FIND(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy,
578 (struct pf_state *)key)) != NULL)
580 if ((s = RB_FIND(pf_state_tree_ext_gwy, &pfi_all->pfik_ext_gwy,
581 (struct pf_state *)key)) != NULL)
585 panic("pf_find_state_recurse");
590 pf_find_state_all(struct pf_state_cmp *key, u_int8_t tree, int *more)
592 struct pf_state *s, *ss = NULL;
595 pf_status.fcounters[FCNT_STATE_SEARCH]++;
599 TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states) {
600 s = RB_FIND(pf_state_tree_lan_ext,
601 &kif->pfik_lan_ext, (struct pf_state *)key);
611 TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states) {
612 s = RB_FIND(pf_state_tree_ext_gwy,
613 &kif->pfik_ext_gwy, (struct pf_state *)key);
623 panic("pf_find_state_all");
628 pf_init_threshold(struct pf_threshold *threshold,
629 u_int32_t limit, u_int32_t seconds)
631 threshold->limit = limit * PF_THRESHOLD_MULT;
632 threshold->seconds = seconds;
633 threshold->count = 0;
634 threshold->last = time_second;
638 pf_add_threshold(struct pf_threshold *threshold)
640 u_int32_t t = time_second, diff = t - threshold->last;
642 if (diff >= threshold->seconds)
643 threshold->count = 0;
645 threshold->count -= threshold->count * diff /
647 threshold->count += PF_THRESHOLD_MULT;
652 pf_check_threshold(struct pf_threshold *threshold)
654 return (threshold->count > threshold->limit);
658 pf_src_connlimit(struct pf_state **state)
663 (*state)->src_node->conn++;
664 (*state)->src.tcp_est = 1;
665 pf_add_threshold(&(*state)->src_node->conn_rate);
667 if ((*state)->rule.ptr->max_src_conn &&
668 (*state)->rule.ptr->max_src_conn <
669 (*state)->src_node->conn) {
670 pf_status.lcounters[LCNT_SRCCONN]++;
674 if ((*state)->rule.ptr->max_src_conn_rate.limit &&
675 pf_check_threshold(&(*state)->src_node->conn_rate)) {
676 pf_status.lcounters[LCNT_SRCCONNRATE]++;
683 if ((*state)->rule.ptr->overload_tbl) {
685 u_int32_t killed = 0;
687 pf_status.lcounters[LCNT_OVERLOAD_TABLE]++;
688 if (pf_status.debug >= PF_DEBUG_MISC) {
689 kprintf("pf_src_connlimit: blocking address ");
690 pf_print_host(&(*state)->src_node->addr, 0,
694 bzero(&p, sizeof(p));
695 p.pfra_af = (*state)->af;
696 switch ((*state)->af) {
700 p.pfra_ip4addr = (*state)->src_node->addr.v4;
706 p.pfra_ip6addr = (*state)->src_node->addr.v6;
711 pfr_insert_kentry((*state)->rule.ptr->overload_tbl,
714 /* kill existing states if that's required. */
715 if ((*state)->rule.ptr->flush) {
716 pf_status.lcounters[LCNT_OVERLOAD_FLUSH]++;
718 RB_FOREACH(s, pf_state_tree_id, &tree_id) {
720 * Kill states from this source. (Only those
721 * from the same rule if PF_FLUSH_GLOBAL is not
724 if (s->af == (*state)->af &&
725 (((*state)->direction == PF_OUT &&
726 PF_AEQ(&(*state)->src_node->addr,
727 &s->lan.addr, s->af)) ||
728 ((*state)->direction == PF_IN &&
729 PF_AEQ(&(*state)->src_node->addr,
730 &s->ext.addr, s->af))) &&
731 ((*state)->rule.ptr->flush &
733 (*state)->rule.ptr == s->rule.ptr)) {
734 s->timeout = PFTM_PURGE;
735 s->src.state = s->dst.state =
740 if (pf_status.debug >= PF_DEBUG_MISC)
741 kprintf(", %u states killed", killed);
743 if (pf_status.debug >= PF_DEBUG_MISC)
747 /* kill this state */
748 (*state)->timeout = PFTM_PURGE;
749 (*state)->src.state = (*state)->dst.state = TCPS_CLOSED;
754 pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule,
755 struct pf_addr *src, sa_family_t af)
757 struct pf_src_node k;
761 PF_ACPY(&k.addr, src, af);
762 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
763 rule->rpool.opts & PF_POOL_STICKYADDR)
767 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
768 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
771 if (!rule->max_src_nodes ||
772 rule->src_nodes < rule->max_src_nodes)
773 (*sn) = pool_get(&pf_src_tree_pl, PR_NOWAIT);
775 pf_status.lcounters[LCNT_SRCNODES]++;
778 bzero(*sn, sizeof(struct pf_src_node));
780 pf_init_threshold(&(*sn)->conn_rate,
781 rule->max_src_conn_rate.limit,
782 rule->max_src_conn_rate.seconds);
785 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
786 rule->rpool.opts & PF_POOL_STICKYADDR)
787 (*sn)->rule.ptr = rule;
789 (*sn)->rule.ptr = NULL;
790 PF_ACPY(&(*sn)->addr, src, af);
791 if (RB_INSERT(pf_src_tree,
792 &tree_src_tracking, *sn) != NULL) {
793 if (pf_status.debug >= PF_DEBUG_MISC) {
794 kprintf("pf: src_tree insert failed: ");
795 pf_print_host(&(*sn)->addr, 0, af);
798 pool_put(&pf_src_tree_pl, *sn);
801 (*sn)->creation = time_second;
802 (*sn)->ruletype = rule->action;
803 if ((*sn)->rule.ptr != NULL)
804 (*sn)->rule.ptr->src_nodes++;
805 pf_status.scounters[SCNT_SRC_NODE_INSERT]++;
806 pf_status.src_nodes++;
808 if (rule->max_src_states &&
809 (*sn)->states >= rule->max_src_states) {
810 pf_status.lcounters[LCNT_SRCSTATES]++;
818 pf_insert_state(struct pfi_kif *kif, struct pf_state *state)
820 /* Thou MUST NOT insert multiple duplicate keys */
821 state->u.s.kif = kif;
822 if (RB_INSERT(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state)) {
823 if (pf_status.debug >= PF_DEBUG_MISC) {
824 kprintf("pf: state insert failed: tree_lan_ext");
826 pf_print_host(&state->lan.addr, state->lan.port,
829 pf_print_host(&state->gwy.addr, state->gwy.port,
832 pf_print_host(&state->ext.addr, state->ext.port,
834 if (state->sync_flags & PFSTATE_FROMSYNC)
835 kprintf(" (from sync)");
841 if (RB_INSERT(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, state)) {
842 if (pf_status.debug >= PF_DEBUG_MISC) {
843 kprintf("pf: state insert failed: tree_ext_gwy");
845 pf_print_host(&state->lan.addr, state->lan.port,
848 pf_print_host(&state->gwy.addr, state->gwy.port,
851 pf_print_host(&state->ext.addr, state->ext.port,
853 if (state->sync_flags & PFSTATE_FROMSYNC)
854 kprintf(" (from sync)");
857 RB_REMOVE(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state);
861 if (state->id == 0 && state->creatorid == 0) {
862 state->id = htobe64(pf_status.stateid++);
863 state->creatorid = pf_status.hostid;
865 if (RB_INSERT(pf_state_tree_id, &tree_id, state) != NULL) {
866 if (pf_status.debug >= PF_DEBUG_MISC) {
867 kprintf("pf: state insert failed: "
868 "id: %016" PRIx64 " creatorid: %08" PRIx32,
869 be64toh(state->id), ntohl(state->creatorid));
870 if (state->sync_flags & PFSTATE_FROMSYNC)
871 kprintf(" (from sync)");
874 RB_REMOVE(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state);
875 RB_REMOVE(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, state);
878 TAILQ_INSERT_TAIL(&state_list, state, u.s.entry_list);
879 pf_status.fcounters[FCNT_STATE_INSERT]++;
881 pfi_kif_ref(kif, PFI_KIF_REF_STATE);
883 pfsync_insert_state(state);
889 pf_purge_thread(void *v)
896 tsleep(pf_purge_thread, PWAIT, "pftm", 1 * hz);
898 lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
900 if (pf_end_threads) {
901 pf_purge_expired_states(pf_status.states, 1);
902 pf_purge_expired_fragments();
903 pf_purge_expired_src_nodes(1);
906 lockmgr(&pf_consistency_lock, LK_RELEASE);
907 wakeup(pf_purge_thread);
912 /* process a fraction of the state table every second */
913 if(!pf_purge_expired_states(1 + (pf_status.states
914 / pf_default_rule.timeout[PFTM_INTERVAL]), 0)) {
916 pf_purge_expired_states(1 + (pf_status.states
917 / pf_default_rule.timeout[PFTM_INTERVAL]), 1);
920 /* purge other expired types every PFTM_INTERVAL seconds */
921 if (++nloops >= pf_default_rule.timeout[PFTM_INTERVAL]) {
922 pf_purge_expired_fragments();
923 if (!pf_purge_expired_src_nodes(locked)) {
924 pf_purge_expired_src_nodes(1);
929 lockmgr(&pf_consistency_lock, LK_RELEASE);
935 pf_state_expires(const struct pf_state *state)
942 /* handle all PFTM_* > PFTM_MAX here */
943 if (state->timeout == PFTM_PURGE)
944 return (time_second);
945 if (state->timeout == PFTM_UNTIL_PACKET)
947 KKASSERT(state->timeout != PFTM_UNLINKED);
948 KASSERT((state->timeout < PFTM_MAX),
949 ("pf_state_expires: timeout > PFTM_MAX"));
950 timeout = state->rule.ptr->timeout[state->timeout];
952 timeout = pf_default_rule.timeout[state->timeout];
953 start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START];
955 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END];
956 states = state->rule.ptr->states;
958 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START];
959 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END];
960 states = pf_status.states;
962 if (end && states > start && start < end) {
964 return (state->expire + timeout * (end - states) /
967 return (time_second);
969 return (state->expire + timeout);
973 pf_purge_expired_src_nodes(int waslocked)
975 struct pf_src_node *cur, *next;
976 int locked = waslocked;
978 for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) {
979 next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur);
981 if (cur->states <= 0 && cur->expire <= time_second) {
983 lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
984 next = RB_NEXT(pf_src_tree,
985 &tree_src_tracking, cur);
988 if (cur->rule.ptr != NULL) {
989 cur->rule.ptr->src_nodes--;
990 if (cur->rule.ptr->states <= 0 &&
991 cur->rule.ptr->max_src_nodes <= 0)
992 pf_rm_rule(NULL, cur->rule.ptr);
994 RB_REMOVE(pf_src_tree, &tree_src_tracking, cur);
995 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
996 pf_status.src_nodes--;
997 pool_put(&pf_src_tree_pl, cur);
1001 if (locked && !waslocked)
1002 lockmgr(&pf_consistency_lock, LK_RELEASE);
1007 pf_src_tree_remove_state(struct pf_state *s)
1011 if (s->src_node != NULL) {
1012 if (s->proto == IPPROTO_TCP) {
1014 --s->src_node->conn;
1016 if (--s->src_node->states <= 0) {
1017 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1020 pf_default_rule.timeout[PFTM_SRC_NODE];
1021 s->src_node->expire = time_second + timeout;
1024 if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) {
1025 if (--s->nat_src_node->states <= 0) {
1026 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
1029 pf_default_rule.timeout[PFTM_SRC_NODE];
1030 s->nat_src_node->expire = time_second + timeout;
1033 s->src_node = s->nat_src_node = NULL;
1036 /* callers should be at crit_enter() */
1038 pf_unlink_state(struct pf_state *cur)
1040 if (cur->src.state == PF_TCPS_PROXY_DST) {
1041 pf_send_tcp(cur->rule.ptr, cur->af,
1042 &cur->ext.addr, &cur->lan.addr,
1043 cur->ext.port, cur->lan.port,
1044 cur->src.seqhi, cur->src.seqlo + 1,
1045 TH_RST|TH_ACK, 0, 0, 0, 1, cur->tag, NULL, NULL);
1047 RB_REMOVE(pf_state_tree_ext_gwy,
1048 &cur->u.s.kif->pfik_ext_gwy, cur);
1049 RB_REMOVE(pf_state_tree_lan_ext,
1050 &cur->u.s.kif->pfik_lan_ext, cur);
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);
1060 /* callers should be at crit_enter() and hold the
1061 * write_lock on pf_consistency_lock */
1063 pf_free_state(struct pf_state *cur)
1066 if (pfsyncif != NULL &&
1067 (pfsyncif->sc_bulk_send_next == cur ||
1068 pfsyncif->sc_bulk_terminator == cur))
1071 KKASSERT(cur->timeout == PFTM_UNLINKED);
1072 if (--cur->rule.ptr->states <= 0 &&
1073 cur->rule.ptr->src_nodes <= 0)
1074 pf_rm_rule(NULL, cur->rule.ptr);
1075 if (cur->nat_rule.ptr != NULL)
1076 if (--cur->nat_rule.ptr->states <= 0 &&
1077 cur->nat_rule.ptr->src_nodes <= 0)
1078 pf_rm_rule(NULL, cur->nat_rule.ptr);
1079 if (cur->anchor.ptr != NULL)
1080 if (--cur->anchor.ptr->states <= 0)
1081 pf_rm_rule(NULL, cur->anchor.ptr);
1082 pf_normalize_tcp_cleanup(cur);
1083 pfi_kif_unref(cur->u.s.kif, PFI_KIF_REF_STATE);
1084 TAILQ_REMOVE(&state_list, cur, u.s.entry_list);
1086 pf_tag_unref(cur->tag);
1087 pool_put(&pf_state_pl, cur);
1088 pf_status.fcounters[FCNT_STATE_REMOVALS]++;
1093 pf_purge_expired_states(u_int32_t maxcheck, int waslocked)
1095 static struct pf_state *cur = NULL;
1096 struct pf_state *next;
1097 int locked = waslocked;
1099 while (maxcheck--) {
1100 /* wrap to start of list when we hit the end */
1102 cur = TAILQ_FIRST(&state_list);
1104 break; /* list empty */
1107 /* get next state, as cur may get deleted */
1108 next = TAILQ_NEXT(cur, u.s.entry_list);
1110 if (cur->timeout == PFTM_UNLINKED) {
1111 /* free unlinked state */
1113 lockmgr(&pf_consistency_lock, LK_EXCLUSIVE);
1117 } else if (pf_state_expires(cur) <= time_second) {
1118 /* unlink and free expired state */
1119 pf_unlink_state(cur);
1121 if (!lockmgr(&pf_consistency_lock, LK_EXCLUSIVE))
1131 lockmgr(&pf_consistency_lock, LK_RELEASE);
1136 pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw)
1138 if (aw->type != PF_ADDR_TABLE)
1140 if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL)
1146 pf_tbladdr_remove(struct pf_addr_wrap *aw)
1148 if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL)
1150 pfr_detach_table(aw->p.tbl);
1155 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
1157 struct pfr_ktable *kt = aw->p.tbl;
1159 if (aw->type != PF_ADDR_TABLE || kt == NULL)
1161 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
1162 kt = kt->pfrkt_root;
1164 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
1169 pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af)
1174 u_int32_t a = ntohl(addr->addr32[0]);
1175 kprintf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255,
1187 u_int8_t i, curstart = 255, curend = 0,
1188 maxstart = 0, maxend = 0;
1189 for (i = 0; i < 8; i++) {
1190 if (!addr->addr16[i]) {
1191 if (curstart == 255)
1197 if ((curend - curstart) >
1198 (maxend - maxstart)) {
1199 maxstart = curstart;
1206 for (i = 0; i < 8; i++) {
1207 if (i >= maxstart && i <= maxend) {
1216 b = ntohs(addr->addr16[i]);
1233 pf_print_state(struct pf_state *s)
1245 case IPPROTO_ICMPV6:
1249 kprintf("%u ", s->proto);
1252 pf_print_host(&s->lan.addr, s->lan.port, s->af);
1254 pf_print_host(&s->gwy.addr, s->gwy.port, s->af);
1256 pf_print_host(&s->ext.addr, s->ext.port, s->af);
1257 kprintf(" [lo=%u high=%u win=%u modulator=%u", s->src.seqlo,
1258 s->src.seqhi, s->src.max_win, s->src.seqdiff);
1259 if (s->src.wscale && s->dst.wscale)
1260 kprintf(" wscale=%u", s->src.wscale & PF_WSCALE_MASK);
1262 kprintf(" [lo=%u high=%u win=%u modulator=%u", s->dst.seqlo,
1263 s->dst.seqhi, s->dst.max_win, s->dst.seqdiff);
1264 if (s->src.wscale && s->dst.wscale)
1265 kprintf(" wscale=%u", s->dst.wscale & PF_WSCALE_MASK);
1267 kprintf(" %u:%u", s->src.state, s->dst.state);
1271 pf_print_flags(u_int8_t f)
1293 #define PF_SET_SKIP_STEPS(i) \
1295 while (head[i] != cur) { \
1296 head[i]->skip[i].ptr = cur; \
1297 head[i] = TAILQ_NEXT(head[i], entries); \
1302 pf_calc_skip_steps(struct pf_rulequeue *rules)
1304 struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT];
1307 cur = TAILQ_FIRST(rules);
1309 for (i = 0; i < PF_SKIP_COUNT; ++i)
1311 while (cur != NULL) {
1313 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
1314 PF_SET_SKIP_STEPS(PF_SKIP_IFP);
1315 if (cur->direction != prev->direction)
1316 PF_SET_SKIP_STEPS(PF_SKIP_DIR);
1317 if (cur->af != prev->af)
1318 PF_SET_SKIP_STEPS(PF_SKIP_AF);
1319 if (cur->proto != prev->proto)
1320 PF_SET_SKIP_STEPS(PF_SKIP_PROTO);
1321 if (cur->src.neg != prev->src.neg ||
1322 pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr))
1323 PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR);
1324 if (cur->src.port[0] != prev->src.port[0] ||
1325 cur->src.port[1] != prev->src.port[1] ||
1326 cur->src.port_op != prev->src.port_op)
1327 PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
1328 if (cur->dst.neg != prev->dst.neg ||
1329 pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr))
1330 PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR);
1331 if (cur->dst.port[0] != prev->dst.port[0] ||
1332 cur->dst.port[1] != prev->dst.port[1] ||
1333 cur->dst.port_op != prev->dst.port_op)
1334 PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
1337 cur = TAILQ_NEXT(cur, entries);
1339 for (i = 0; i < PF_SKIP_COUNT; ++i)
1340 PF_SET_SKIP_STEPS(i);
1344 pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2)
1346 if (aw1->type != aw2->type)
1348 switch (aw1->type) {
1349 case PF_ADDR_ADDRMASK:
1350 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0))
1352 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0))
1355 case PF_ADDR_DYNIFTL:
1356 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt);
1357 case PF_ADDR_NOROUTE:
1358 case PF_ADDR_URPFFAILED:
1361 return (aw1->p.tbl != aw2->p.tbl);
1362 case PF_ADDR_RTLABEL:
1363 return (aw1->v.rtlabel != aw2->v.rtlabel);
1365 kprintf("invalid address type: %d\n", aw1->type);
1371 pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp)
1377 l = cksum + old - new;
1378 l = (l >> 16) + (l & 65535);
1386 pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc,
1387 struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af)
1392 PF_ACPY(&ao, a, af);
1400 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1401 ao.addr16[0], an->addr16[0], 0),
1402 ao.addr16[1], an->addr16[1], 0);
1404 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1405 ao.addr16[0], an->addr16[0], u),
1406 ao.addr16[1], an->addr16[1], u),
1412 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1413 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1414 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1415 ao.addr16[0], an->addr16[0], u),
1416 ao.addr16[1], an->addr16[1], u),
1417 ao.addr16[2], an->addr16[2], u),
1418 ao.addr16[3], an->addr16[3], u),
1419 ao.addr16[4], an->addr16[4], u),
1420 ao.addr16[5], an->addr16[5], u),
1421 ao.addr16[6], an->addr16[6], u),
1422 ao.addr16[7], an->addr16[7], u),
1430 /* Changes a u_int32_t. Uses a void * so there are no align restrictions */
1432 pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u)
1436 memcpy(&ao, a, sizeof(ao));
1437 memcpy(a, &an, sizeof(u_int32_t));
1438 *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u),
1439 ao % 65536, an % 65536, u);
1444 pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u)
1448 PF_ACPY(&ao, a, AF_INET6);
1449 PF_ACPY(a, an, AF_INET6);
1451 *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1452 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1453 pf_cksum_fixup(pf_cksum_fixup(*c,
1454 ao.addr16[0], an->addr16[0], u),
1455 ao.addr16[1], an->addr16[1], u),
1456 ao.addr16[2], an->addr16[2], u),
1457 ao.addr16[3], an->addr16[3], u),
1458 ao.addr16[4], an->addr16[4], u),
1459 ao.addr16[5], an->addr16[5], u),
1460 ao.addr16[6], an->addr16[6], u),
1461 ao.addr16[7], an->addr16[7], u);
1466 pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa,
1467 struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c,
1468 u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af)
1470 struct pf_addr oia, ooa;
1472 PF_ACPY(&oia, ia, af);
1473 PF_ACPY(&ooa, oa, af);
1475 /* Change inner protocol port, fix inner protocol checksum. */
1477 u_int16_t oip = *ip;
1484 *pc = pf_cksum_fixup(*pc, oip, *ip, u);
1485 *ic = pf_cksum_fixup(*ic, oip, *ip, 0);
1487 *ic = pf_cksum_fixup(*ic, opc, *pc, 0);
1489 /* Change inner ip address, fix inner ip and icmp checksums. */
1490 PF_ACPY(ia, na, af);
1494 u_int32_t oh2c = *h2c;
1496 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c,
1497 oia.addr16[0], ia->addr16[0], 0),
1498 oia.addr16[1], ia->addr16[1], 0);
1499 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1500 oia.addr16[0], ia->addr16[0], 0),
1501 oia.addr16[1], ia->addr16[1], 0);
1502 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0);
1508 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1509 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1510 pf_cksum_fixup(pf_cksum_fixup(*ic,
1511 oia.addr16[0], ia->addr16[0], u),
1512 oia.addr16[1], ia->addr16[1], u),
1513 oia.addr16[2], ia->addr16[2], u),
1514 oia.addr16[3], ia->addr16[3], u),
1515 oia.addr16[4], ia->addr16[4], u),
1516 oia.addr16[5], ia->addr16[5], u),
1517 oia.addr16[6], ia->addr16[6], u),
1518 oia.addr16[7], ia->addr16[7], u);
1522 /* Change outer ip address, fix outer ip or icmpv6 checksum. */
1523 PF_ACPY(oa, na, af);
1527 *hc = pf_cksum_fixup(pf_cksum_fixup(*hc,
1528 ooa.addr16[0], oa->addr16[0], 0),
1529 ooa.addr16[1], oa->addr16[1], 0);
1534 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1535 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1536 pf_cksum_fixup(pf_cksum_fixup(*ic,
1537 ooa.addr16[0], oa->addr16[0], u),
1538 ooa.addr16[1], oa->addr16[1], u),
1539 ooa.addr16[2], oa->addr16[2], u),
1540 ooa.addr16[3], oa->addr16[3], u),
1541 ooa.addr16[4], oa->addr16[4], u),
1542 ooa.addr16[5], oa->addr16[5], u),
1543 ooa.addr16[6], oa->addr16[6], u),
1544 ooa.addr16[7], oa->addr16[7], u);
1552 * Need to modulate the sequence numbers in the TCP SACK option
1553 * (credits to Krzysztof Pfaff for report and patch)
1556 pf_modulate_sack(struct mbuf *m, int off, struct pf_pdesc *pd,
1557 struct tcphdr *th, struct pf_state_peer *dst)
1559 int hlen = (th->th_off << 2) - sizeof(*th), thoptlen = hlen;
1560 u_int8_t opts[TCP_MAXOLEN], *opt = opts;
1561 int copyback = 0, i, olen;
1562 struct raw_sackblock sack;
1564 #define TCPOLEN_SACKLEN (TCPOLEN_SACK + 2)
1565 if (hlen < TCPOLEN_SACKLEN ||
1566 !pf_pull_hdr(m, off + sizeof(*th), opts, hlen, NULL, NULL, pd->af))
1569 while (hlen >= TCPOLEN_SACKLEN) {
1572 case TCPOPT_EOL: /* FALLTHROUGH */
1580 if (olen >= TCPOLEN_SACKLEN) {
1581 for (i = 2; i + TCPOLEN_SACK <= olen;
1582 i += TCPOLEN_SACK) {
1583 memcpy(&sack, &opt[i], sizeof(sack));
1584 pf_change_a(&sack.rblk_start, &th->th_sum,
1585 htonl(ntohl(sack.rblk_start) -
1587 pf_change_a(&sack.rblk_end, &th->th_sum,
1588 htonl(ntohl(sack.rblk_end) -
1590 memcpy(&opt[i], &sack, sizeof(sack));
1604 m_copyback(m, off + sizeof(*th), thoptlen, opts);
1609 pf_send_tcp(const struct pf_rule *r, sa_family_t af,
1610 const struct pf_addr *saddr, const struct pf_addr *daddr,
1611 u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack,
1612 u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl, int tag,
1613 u_int16_t rtag, struct ether_header *eh, struct ifnet *ifp)
1618 struct ip *h = NULL;
1621 struct ip6_hdr *h6 = NULL;
1623 struct tcphdr *th = NULL;
1625 struct pf_mtag *pf_mtag;
1627 /* maximum segment size tcp option */
1628 tlen = sizeof(struct tcphdr);
1635 len = sizeof(struct ip) + tlen;
1640 len = sizeof(struct ip6_hdr) + tlen;
1645 /* create outgoing mbuf */
1646 m = m_gethdr(MB_DONTWAIT, MT_HEADER);
1649 if ((pf_mtag = pf_get_mtag(m)) == NULL) {
1654 pf_mtag->flags |= PF_TAG_GENERATED;
1656 pf_mtag->tag = rtag;
1658 if (r != NULL && r->rtableid >= 0)
1659 pf_mtag->rtableid = r->rtableid;
1662 if (r != NULL && r->qid) {
1663 pf_mtag->qid = r->qid;
1664 /* add hints for ecn */
1666 pf_mtag->hdr = mtod(m, struct ip *);
1669 m->m_data += max_linkhdr;
1670 m->m_pkthdr.len = m->m_len = len;
1671 m->m_pkthdr.rcvif = NULL;
1672 bzero(m->m_data, len);
1676 h = mtod(m, struct ip *);
1678 /* IP header fields included in the TCP checksum */
1679 h->ip_p = IPPROTO_TCP;
1681 h->ip_src.s_addr = saddr->v4.s_addr;
1682 h->ip_dst.s_addr = daddr->v4.s_addr;
1684 th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip));
1689 h6 = mtod(m, struct ip6_hdr *);
1691 /* IP header fields included in the TCP checksum */
1692 h6->ip6_nxt = IPPROTO_TCP;
1693 h6->ip6_plen = htons(tlen);
1694 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr));
1695 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr));
1697 th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr));
1703 th->th_sport = sport;
1704 th->th_dport = dport;
1705 th->th_seq = htonl(seq);
1706 th->th_ack = htonl(ack);
1707 th->th_off = tlen >> 2;
1708 th->th_flags = flags;
1709 th->th_win = htons(win);
1712 opt = (char *)(th + 1);
1713 opt[0] = TCPOPT_MAXSEG;
1716 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2);
1723 th->th_sum = in_cksum(m, len);
1725 /* Finish the IP header */
1727 h->ip_hl = sizeof(*h) >> 2;
1728 h->ip_tos = IPTOS_LOWDELAY;
1730 h->ip_off = path_mtu_discovery ? IP_DF : 0;
1731 h->ip_ttl = ttl ? ttl : ip_defttl;
1734 ip_output(m, NULL, NULL, 0, NULL, NULL);
1738 struct ether_header *e = (void *)ro.ro_dst.sa_data;
1746 ro.ro_dst.sa_len = sizeof(ro.ro_dst);
1747 ro.ro_dst.sa_family = pseudo_AF_HDRCMPLT;
1748 bcopy(eh->ether_dhost, e->ether_shost, ETHER_ADDR_LEN);
1749 bcopy(eh->ether_shost, e->ether_dhost, ETHER_ADDR_LEN);
1750 e->ether_type = eh->ether_type;
1751 /* XXX_IMPORT: later */
1752 ip_output(m, (void *)NULL, &ro, 0,
1753 (void *)NULL, (void *)NULL);
1760 th->th_sum = in6_cksum(m, IPPROTO_TCP,
1761 sizeof(struct ip6_hdr), tlen);
1763 h6->ip6_vfc |= IPV6_VERSION;
1764 h6->ip6_hlim = IPV6_DEFHLIM;
1766 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
1773 pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af,
1776 struct pf_mtag *pf_mtag;
1779 m0 = m_copy(m, 0, M_COPYALL);
1781 if ((pf_mtag = pf_get_mtag(m0)) == NULL)
1783 pf_mtag->flags |= PF_TAG_GENERATED;
1785 if (r->rtableid >= 0)
1786 pf_mtag->rtableid = r->rtableid;
1790 pf_mtag->qid = r->qid;
1791 /* add hints for ecn */
1793 pf_mtag->hdr = mtod(m0, struct ip *);
1800 icmp_error(m0, type, code, 0, 0);
1805 icmp6_error(m0, type, code, 0);
1812 * Return 1 if the addresses a and b match (with mask m), otherwise return 0.
1813 * If n is 0, they match if they are equal. If n is != 0, they match if they
1817 pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m,
1818 struct pf_addr *b, sa_family_t af)
1825 if ((a->addr32[0] & m->addr32[0]) ==
1826 (b->addr32[0] & m->addr32[0]))
1832 if (((a->addr32[0] & m->addr32[0]) ==
1833 (b->addr32[0] & m->addr32[0])) &&
1834 ((a->addr32[1] & m->addr32[1]) ==
1835 (b->addr32[1] & m->addr32[1])) &&
1836 ((a->addr32[2] & m->addr32[2]) ==
1837 (b->addr32[2] & m->addr32[2])) &&
1838 ((a->addr32[3] & m->addr32[3]) ==
1839 (b->addr32[3] & m->addr32[3])))
1858 pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p)
1862 return ((p > a1) && (p < a2));
1864 return ((p < a1) || (p > a2));
1866 return ((p >= a1) && (p <= a2));
1880 return (0); /* never reached */
1884 pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p)
1889 return (pf_match(op, a1, a2, p));
1893 pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u)
1895 if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
1897 return (pf_match(op, a1, a2, u));
1901 pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g)
1903 if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
1905 return (pf_match(op, a1, a2, g));
1909 pf_find_mtag(struct mbuf *m)
1913 if ((mtag = m_tag_find(m, PF_MBUF_TAGGED, NULL)) == NULL)
1916 return ((struct pf_mtag *)(mtag + 1));
1920 pf_get_mtag(struct mbuf *m)
1924 if ((mtag = m_tag_find(m, PF_MBUF_TAGGED, NULL)) == NULL) {
1925 mtag = m_tag_get(PF_MBUF_TAGGED, sizeof(struct pf_mtag),
1929 bzero(mtag + 1, sizeof(struct pf_mtag));
1930 m_tag_prepend(m, mtag);
1933 return ((struct pf_mtag *)(mtag + 1));
1937 pf_match_tag(struct mbuf *m, struct pf_rule *r, struct pf_mtag *pf_mtag,
1941 *tag = pf_mtag->tag;
1943 return ((!r->match_tag_not && r->match_tag == *tag) ||
1944 (r->match_tag_not && r->match_tag != *tag));
1948 pf_tag_packet(struct mbuf *m, struct pf_mtag *pf_mtag, int tag, int rtableid)
1950 if (tag <= 0 && rtableid < 0)
1953 if (pf_mtag == NULL)
1954 if ((pf_mtag = pf_get_mtag(m)) == NULL)
1959 pf_mtag->rtableid = rtableid;
1965 pf_step_into_anchor(int *depth, struct pf_ruleset **rs, int n,
1966 struct pf_rule **r, struct pf_rule **a, int *match)
1968 struct pf_anchor_stackframe *f;
1970 (*r)->anchor->match = 0;
1973 if (*depth >= sizeof(pf_anchor_stack) /
1974 sizeof(pf_anchor_stack[0])) {
1975 kprintf("pf_step_into_anchor: stack overflow\n");
1976 *r = TAILQ_NEXT(*r, entries);
1978 } else if (*depth == 0 && a != NULL)
1980 f = pf_anchor_stack + (*depth)++;
1983 if ((*r)->anchor_wildcard) {
1984 f->parent = &(*r)->anchor->children;
1985 if ((f->child = RB_MIN(pf_anchor_node, f->parent)) ==
1990 *rs = &f->child->ruleset;
1994 *rs = &(*r)->anchor->ruleset;
1996 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
2000 pf_step_out_of_anchor(int *depth, struct pf_ruleset **rs, int n,
2001 struct pf_rule **r, struct pf_rule **a, int *match)
2003 struct pf_anchor_stackframe *f;
2009 f = pf_anchor_stack + *depth - 1;
2010 if (f->parent != NULL && f->child != NULL) {
2011 if (f->child->match ||
2012 (match != NULL && *match)) {
2013 f->r->anchor->match = 1;
2016 f->child = RB_NEXT(pf_anchor_node, f->parent, f->child);
2017 if (f->child != NULL) {
2018 *rs = &f->child->ruleset;
2019 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
2027 if (*depth == 0 && a != NULL)
2030 if (f->r->anchor->match || (match != NULL && *match))
2031 quick = f->r->quick;
2032 *r = TAILQ_NEXT(f->r, entries);
2033 } while (*r == NULL);
2040 pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr,
2041 struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af)
2046 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2047 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2051 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2052 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
2053 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) |
2054 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]);
2055 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) |
2056 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]);
2057 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) |
2058 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]);
2064 pf_addr_inc(struct pf_addr *addr, sa_family_t af)
2069 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1);
2073 if (addr->addr32[3] == 0xffffffff) {
2074 addr->addr32[3] = 0;
2075 if (addr->addr32[2] == 0xffffffff) {
2076 addr->addr32[2] = 0;
2077 if (addr->addr32[1] == 0xffffffff) {
2078 addr->addr32[1] = 0;
2080 htonl(ntohl(addr->addr32[0]) + 1);
2083 htonl(ntohl(addr->addr32[1]) + 1);
2086 htonl(ntohl(addr->addr32[2]) + 1);
2089 htonl(ntohl(addr->addr32[3]) + 1);
2095 #define mix(a,b,c) \
2097 a -= b; a -= c; a ^= (c >> 13); \
2098 b -= c; b -= a; b ^= (a << 8); \
2099 c -= a; c -= b; c ^= (b >> 13); \
2100 a -= b; a -= c; a ^= (c >> 12); \
2101 b -= c; b -= a; b ^= (a << 16); \
2102 c -= a; c -= b; c ^= (b >> 5); \
2103 a -= b; a -= c; a ^= (c >> 3); \
2104 b -= c; b -= a; b ^= (a << 10); \
2105 c -= a; c -= b; c ^= (b >> 15); \
2109 * hash function based on bridge_hash in if_bridge.c
2112 pf_hash(struct pf_addr *inaddr, struct pf_addr *hash,
2113 struct pf_poolhashkey *key, sa_family_t af)
2115 u_int32_t a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0];
2120 a += inaddr->addr32[0];
2123 hash->addr32[0] = c + key->key32[2];
2128 a += inaddr->addr32[0];
2129 b += inaddr->addr32[2];
2131 hash->addr32[0] = c;
2132 a += inaddr->addr32[1];
2133 b += inaddr->addr32[3];
2136 hash->addr32[1] = c;
2137 a += inaddr->addr32[2];
2138 b += inaddr->addr32[1];
2141 hash->addr32[2] = c;
2142 a += inaddr->addr32[3];
2143 b += inaddr->addr32[0];
2146 hash->addr32[3] = c;
2153 pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr,
2154 struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn)
2156 unsigned char hash[16];
2157 struct pf_pool *rpool = &r->rpool;
2158 struct pf_addr *raddr = &rpool->cur->addr.v.a.addr;
2159 struct pf_addr *rmask = &rpool->cur->addr.v.a.mask;
2160 struct pf_pooladdr *acur = rpool->cur;
2161 struct pf_src_node k;
2163 if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR &&
2164 (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2166 PF_ACPY(&k.addr, saddr, af);
2167 if (r->rule_flag & PFRULE_RULESRCTRACK ||
2168 r->rpool.opts & PF_POOL_STICKYADDR)
2172 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
2173 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
2174 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) {
2175 PF_ACPY(naddr, &(*sn)->raddr, af);
2176 if (pf_status.debug >= PF_DEBUG_MISC) {
2177 kprintf("pf_map_addr: src tracking maps ");
2178 pf_print_host(&k.addr, 0, af);
2180 pf_print_host(naddr, 0, af);
2187 if (rpool->cur->addr.type == PF_ADDR_NOROUTE)
2189 if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2193 if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 &&
2194 (rpool->opts & PF_POOL_TYPEMASK) !=
2197 raddr = &rpool->cur->addr.p.dyn->pfid_addr4;
2198 rmask = &rpool->cur->addr.p.dyn->pfid_mask4;
2203 if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 &&
2204 (rpool->opts & PF_POOL_TYPEMASK) !=
2207 raddr = &rpool->cur->addr.p.dyn->pfid_addr6;
2208 rmask = &rpool->cur->addr.p.dyn->pfid_mask6;
2212 } else if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2213 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN)
2214 return (1); /* unsupported */
2216 raddr = &rpool->cur->addr.v.a.addr;
2217 rmask = &rpool->cur->addr.v.a.mask;
2220 switch (rpool->opts & PF_POOL_TYPEMASK) {
2222 PF_ACPY(naddr, raddr, af);
2224 case PF_POOL_BITMASK:
2225 PF_POOLMASK(naddr, raddr, rmask, saddr, af);
2227 case PF_POOL_RANDOM:
2228 if (init_addr != NULL && PF_AZERO(init_addr, af)) {
2232 rpool->counter.addr32[0] = htonl(karc4random());
2237 if (rmask->addr32[3] != 0xffffffff)
2238 rpool->counter.addr32[3] =
2239 htonl(karc4random());
2242 if (rmask->addr32[2] != 0xffffffff)
2243 rpool->counter.addr32[2] =
2244 htonl(karc4random());
2247 if (rmask->addr32[1] != 0xffffffff)
2248 rpool->counter.addr32[1] =
2249 htonl(karc4random());
2252 if (rmask->addr32[0] != 0xffffffff)
2253 rpool->counter.addr32[0] =
2254 htonl(karc4random());
2258 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2259 PF_ACPY(init_addr, naddr, af);
2262 PF_AINC(&rpool->counter, af);
2263 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
2266 case PF_POOL_SRCHASH:
2267 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af);
2268 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af);
2270 case PF_POOL_ROUNDROBIN:
2271 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2272 if (!pfr_pool_get(rpool->cur->addr.p.tbl,
2273 &rpool->tblidx, &rpool->counter,
2274 &raddr, &rmask, af))
2276 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2277 if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2278 &rpool->tblidx, &rpool->counter,
2279 &raddr, &rmask, af))
2281 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af))
2285 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL)
2286 rpool->cur = TAILQ_FIRST(&rpool->list);
2287 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
2289 if (pfr_pool_get(rpool->cur->addr.p.tbl,
2290 &rpool->tblidx, &rpool->counter,
2291 &raddr, &rmask, af)) {
2292 /* table contains no address of type 'af' */
2293 if (rpool->cur != acur)
2297 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
2299 if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
2300 &rpool->tblidx, &rpool->counter,
2301 &raddr, &rmask, af)) {
2302 /* table contains no address of type 'af' */
2303 if (rpool->cur != acur)
2308 raddr = &rpool->cur->addr.v.a.addr;
2309 rmask = &rpool->cur->addr.v.a.mask;
2310 PF_ACPY(&rpool->counter, raddr, af);
2314 PF_ACPY(naddr, &rpool->counter, af);
2315 if (init_addr != NULL && PF_AZERO(init_addr, af))
2316 PF_ACPY(init_addr, naddr, af);
2317 PF_AINC(&rpool->counter, af);
2321 PF_ACPY(&(*sn)->raddr, naddr, af);
2323 if (pf_status.debug >= PF_DEBUG_MISC &&
2324 (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
2325 kprintf("pf_map_addr: selected address ");
2326 pf_print_host(naddr, 0, af);
2334 pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r,
2335 struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport,
2336 struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high,
2337 struct pf_src_node **sn)
2339 struct pf_state_cmp key;
2340 struct pf_addr init_addr;
2343 bzero(&init_addr, sizeof(init_addr));
2344 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2347 if (proto == IPPROTO_ICMP) {
2355 PF_ACPY(&key.ext.addr, daddr, key.af);
2356 PF_ACPY(&key.gwy.addr, naddr, key.af);
2357 key.ext.port = dport;
2360 * port search; start random, step;
2361 * similar 2 portloop in in_pcbbind
2363 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP ||
2364 proto == IPPROTO_ICMP)) {
2365 key.gwy.port = dport;
2366 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
2368 } else if (low == 0 && high == 0) {
2369 key.gwy.port = *nport;
2370 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
2372 } else if (low == high) {
2373 key.gwy.port = htons(low);
2374 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) {
2375 *nport = htons(low);
2387 cut = htonl(karc4random()) % (1 + high - low) + low;
2388 /* low <= cut <= high */
2389 for (tmp = cut; tmp <= high; ++(tmp)) {
2390 key.gwy.port = htons(tmp);
2391 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
2393 *nport = htons(tmp);
2397 for (tmp = cut - 1; tmp >= low; --(tmp)) {
2398 key.gwy.port = htons(tmp);
2399 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
2401 *nport = htons(tmp);
2407 switch (r->rpool.opts & PF_POOL_TYPEMASK) {
2408 case PF_POOL_RANDOM:
2409 case PF_POOL_ROUNDROBIN:
2410 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
2414 case PF_POOL_SRCHASH:
2415 case PF_POOL_BITMASK:
2419 } while (! PF_AEQ(&init_addr, naddr, af) );
2421 return (1); /* none available */
2425 pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off,
2426 int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport,
2427 struct pf_addr *daddr, u_int16_t dport, int rs_num)
2429 struct pf_rule *r, *rm = NULL;
2430 struct pf_ruleset *ruleset = NULL;
2435 r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr);
2436 while (r && rm == NULL) {
2437 struct pf_rule_addr *src = NULL, *dst = NULL;
2438 struct pf_addr_wrap *xdst = NULL;
2440 if (r->action == PF_BINAT && direction == PF_IN) {
2442 if (r->rpool.cur != NULL)
2443 xdst = &r->rpool.cur->addr;
2450 if (pfi_kif_match(r->kif, kif) == r->ifnot)
2451 r = r->skip[PF_SKIP_IFP].ptr;
2452 else if (r->direction && r->direction != direction)
2453 r = r->skip[PF_SKIP_DIR].ptr;
2454 else if (r->af && r->af != pd->af)
2455 r = r->skip[PF_SKIP_AF].ptr;
2456 else if (r->proto && r->proto != pd->proto)
2457 r = r->skip[PF_SKIP_PROTO].ptr;
2458 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af,
2460 r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR :
2461 PF_SKIP_DST_ADDR].ptr;
2462 else if (src->port_op && !pf_match_port(src->port_op,
2463 src->port[0], src->port[1], sport))
2464 r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT :
2465 PF_SKIP_DST_PORT].ptr;
2466 else if (dst != NULL &&
2467 PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->neg, NULL))
2468 r = r->skip[PF_SKIP_DST_ADDR].ptr;
2469 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af,
2471 r = TAILQ_NEXT(r, entries);
2472 else if (dst != NULL && dst->port_op &&
2473 !pf_match_port(dst->port_op, dst->port[0],
2474 dst->port[1], dport))
2475 r = r->skip[PF_SKIP_DST_PORT].ptr;
2476 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag))
2477 r = TAILQ_NEXT(r, entries);
2478 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto !=
2479 IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m,
2480 off, pd->hdr.tcp), r->os_fingerprint)))
2481 r = TAILQ_NEXT(r, entries);
2485 if (r->rtableid >= 0)
2486 rtableid = r->rtableid;
2487 if (r->anchor == NULL) {
2490 pf_step_into_anchor(&asd, &ruleset, rs_num,
2494 pf_step_out_of_anchor(&asd, &ruleset, rs_num, &r,
2497 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid))
2499 if (rm != NULL && (rm->action == PF_NONAT ||
2500 rm->action == PF_NORDR || rm->action == PF_NOBINAT))
2506 pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction,
2507 struct pfi_kif *kif, struct pf_src_node **sn,
2508 struct pf_addr *saddr, u_int16_t sport,
2509 struct pf_addr *daddr, u_int16_t dport,
2510 struct pf_addr *naddr, u_int16_t *nport)
2512 struct pf_rule *r = NULL;
2514 if (direction == PF_OUT) {
2515 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2516 sport, daddr, dport, PF_RULESET_BINAT);
2518 r = pf_match_translation(pd, m, off, direction, kif,
2519 saddr, sport, daddr, dport, PF_RULESET_NAT);
2521 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2522 sport, daddr, dport, PF_RULESET_RDR);
2524 r = pf_match_translation(pd, m, off, direction, kif,
2525 saddr, sport, daddr, dport, PF_RULESET_BINAT);
2529 switch (r->action) {
2535 if (pf_get_sport(pd->af, pd->proto, r, saddr,
2536 daddr, dport, naddr, nport, r->rpool.proxy_port[0],
2537 r->rpool.proxy_port[1], sn)) {
2538 DPFPRINTF(PF_DEBUG_MISC,
2539 ("pf: NAT proxy port allocation "
2541 r->rpool.proxy_port[0],
2542 r->rpool.proxy_port[1]));
2547 switch (direction) {
2549 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){
2553 if (r->rpool.cur->addr.p.dyn->
2557 &r->rpool.cur->addr.p.dyn->
2559 &r->rpool.cur->addr.p.dyn->
2566 if (r->rpool.cur->addr.p.dyn->
2570 &r->rpool.cur->addr.p.dyn->
2572 &r->rpool.cur->addr.p.dyn->
2580 &r->rpool.cur->addr.v.a.addr,
2581 &r->rpool.cur->addr.v.a.mask,
2585 if (r->src.addr.type == PF_ADDR_DYNIFTL) {
2589 if (r->src.addr.p.dyn->
2593 &r->src.addr.p.dyn->
2595 &r->src.addr.p.dyn->
2602 if (r->src.addr.p.dyn->
2606 &r->src.addr.p.dyn->
2608 &r->src.addr.p.dyn->
2616 &r->src.addr.v.a.addr,
2617 &r->src.addr.v.a.mask, daddr,
2623 if (pf_map_addr(pd->af, r, saddr, naddr, NULL, sn))
2625 if ((r->rpool.opts & PF_POOL_TYPEMASK) ==
2627 PF_POOLMASK(naddr, naddr,
2628 &r->rpool.cur->addr.v.a.mask, daddr,
2630 if (r->rpool.proxy_port[1]) {
2631 u_int32_t tmp_nport;
2633 tmp_nport = ((ntohs(dport) -
2634 ntohs(r->dst.port[0])) %
2635 (r->rpool.proxy_port[1] -
2636 r->rpool.proxy_port[0] + 1)) +
2637 r->rpool.proxy_port[0];
2639 /* wrap around if necessary */
2640 if (tmp_nport > 65535)
2642 *nport = htons((u_int16_t)tmp_nport);
2643 } else if (r->rpool.proxy_port[0])
2644 *nport = htons(r->rpool.proxy_port[0]);
2656 struct netmsg_hashlookup {
2657 struct netmsg nm_netmsg;
2658 struct inpcb **nm_pinp;
2659 struct inpcbinfo *nm_pcbinfo;
2660 struct pf_addr *nm_saddr;
2661 struct pf_addr *nm_daddr;
2668 in_pcblookup_hash_handler(struct netmsg *msg0)
2670 struct netmsg_hashlookup *msg = (struct netmsg_hashlookup *)msg0;
2672 if (msg->nm_af == AF_INET)
2673 *msg->nm_pinp = in_pcblookup_hash(msg->nm_pcbinfo,
2674 msg->nm_saddr->v4, msg->nm_sport, msg->nm_daddr->v4,
2675 msg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2678 *msg->nm_pinp = in6_pcblookup_hash(msg->nm_pcbinfo,
2679 &msg->nm_saddr->v6, msg->nm_sport, &msg->nm_daddr->v6,
2680 msg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2682 lwkt_replymsg(&msg->nm_netmsg.nm_lmsg, 0);
2687 pf_socket_lookup(int direction, struct pf_pdesc *pd, struct inpcb *inp_arg)
2689 struct pf_addr *saddr, *daddr;
2690 u_int16_t sport, dport;
2691 struct inpcbinfo *pi;
2694 struct netmsg_hashlookup *msg = NULL;
2700 pd->lookup.uid = UID_MAX;
2701 pd->lookup.gid = GID_MAX;
2702 pd->lookup.pid = NO_PID;
2703 if (direction == PF_IN) {
2710 switch (pd->proto) {
2712 sport = pd->hdr.tcp->th_sport;
2713 dport = pd->hdr.tcp->th_dport;
2715 pi_cpu = tcp_addrcpu(saddr->v4.s_addr, sport, daddr->v4.s_addr, dport);
2716 pi = &tcbinfo[pi_cpu];
2719 * Our netstack runs lockless on MP systems
2720 * (only for TCP connections at the moment).
2722 * As we are not allowed to read another CPU's tcbinfo,
2723 * we have to ask that CPU via remote call to search the
2726 * Prepare a msg iff data belongs to another CPU.
2728 if (pi_cpu != mycpu->gd_cpuid) {
2729 msg = kmalloc(sizeof(*msg), M_LWKTMSG, M_INTWAIT);
2730 netmsg_init(&msg->nm_netmsg, NULL, &netisr_afree_rport,
2731 0, in_pcblookup_hash_handler);
2732 msg->nm_pinp = &inp;
2733 msg->nm_pcbinfo = pi;
2734 msg->nm_saddr = saddr;
2735 msg->nm_sport = sport;
2736 msg->nm_daddr = daddr;
2737 msg->nm_dport = dport;
2738 msg->nm_af = pd->af;
2743 sport = pd->hdr.udp->uh_sport;
2744 dport = pd->hdr.udp->uh_dport;
2750 if (direction != PF_IN) {
2762 * Query other CPU, second part
2764 * msg only gets initialized when:
2766 * 2) the info belongs to another CPU
2768 * Use some switch/case magic to avoid code duplication.
2773 inp = in6_pcblookup_hash(pi, &saddr->v6, sport,
2774 &daddr->v6, dport, INPLOOKUP_WILDCARD, NULL);
2780 /* FALLTHROUGH if SMP and on other CPU */
2785 lwkt_sendmsg(tcp_cport(pi_cpu),
2786 &msg->nm_netmsg.nm_lmsg);
2790 inp = in_pcblookup_hash(pi, saddr->v4, sport, daddr->v4,
2791 dport, INPLOOKUP_WILDCARD, NULL);
2800 pd->lookup.uid = inp->inp_socket->so_cred->cr_uid;
2801 pd->lookup.gid = inp->inp_socket->so_cred->cr_groups[0];
2806 pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2810 u_int8_t *opt, optlen;
2811 u_int8_t wscale = 0;
2813 hlen = th_off << 2; /* hlen <= sizeof(hdr) */
2814 if (hlen <= sizeof(struct tcphdr))
2816 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2818 opt = hdr + sizeof(struct tcphdr);
2819 hlen -= sizeof(struct tcphdr);
2829 if (wscale > TCP_MAX_WINSHIFT)
2830 wscale = TCP_MAX_WINSHIFT;
2831 wscale |= PF_WSCALE_FLAG;
2846 pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2850 u_int8_t *opt, optlen;
2851 u_int16_t mss = tcp_mssdflt;
2853 hlen = th_off << 2; /* hlen <= sizeof(hdr) */
2854 if (hlen <= sizeof(struct tcphdr))
2856 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2858 opt = hdr + sizeof(struct tcphdr);
2859 hlen -= sizeof(struct tcphdr);
2860 while (hlen >= TCPOLEN_MAXSEG) {
2868 bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2);
2884 pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer)
2887 struct sockaddr_in *dst;
2891 struct sockaddr_in6 *dst6;
2892 struct route_in6 ro6;
2894 struct rtentry *rt = NULL;
2896 u_int16_t mss = tcp_mssdflt;
2901 hlen = sizeof(struct ip);
2902 bzero(&ro, sizeof(ro));
2903 dst = (struct sockaddr_in *)&ro.ro_dst;
2904 dst->sin_family = AF_INET;
2905 dst->sin_len = sizeof(*dst);
2906 dst->sin_addr = addr->v4;
2907 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
2913 hlen = sizeof(struct ip6_hdr);
2914 bzero(&ro6, sizeof(ro6));
2915 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst;
2916 dst6->sin6_family = AF_INET6;
2917 dst6->sin6_len = sizeof(*dst6);
2918 dst6->sin6_addr = addr->v6;
2919 rtalloc_ign((struct route *)&ro6, (RTF_CLONING | RTF_PRCLONING));
2925 if (rt && rt->rt_ifp) {
2926 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr);
2927 mss = max(tcp_mssdflt, mss);
2930 mss = min(mss, offer);
2931 mss = max(mss, 64); /* sanity - at least max opt space */
2936 pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr)
2938 struct pf_rule *r = s->rule.ptr;
2941 if (!r->rt || r->rt == PF_FASTROUTE)
2946 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL,
2948 s->rt_kif = r->rpool.cur->kif;
2953 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL,
2955 s->rt_kif = r->rpool.cur->kif;
2962 pf_test_tcp(struct pf_rule **rm, struct pf_state **sm, int direction,
2963 struct pfi_kif *kif, struct mbuf *m, int off, void *h,
2964 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
2965 struct ifqueue *ifq, struct inpcb *inp)
2967 struct pf_rule *nr = NULL;
2968 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
2969 struct tcphdr *th = pd->hdr.tcp;
2970 u_int16_t bport, nport = 0;
2971 sa_family_t af = pd->af;
2972 struct pf_rule *r, *a = NULL;
2973 struct pf_ruleset *ruleset = NULL;
2974 struct pf_src_node *nsn = NULL;
2977 int tag = -1, rtableid = -1;
2978 u_int16_t mss = tcp_mssdflt;
2982 if (pf_check_congestion(ifq)) {
2983 REASON_SET(&reason, PFRES_CONGEST);
2988 pd->lookup.done = pf_socket_lookup(direction, pd, inp);
2989 else if (debug_pfugidhack) {
2991 DPFPRINTF(PF_DEBUG_MISC, ("pf: unlocked lookup\n"));
2992 pd->lookup.done = pf_socket_lookup(direction, pd, inp);
2997 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
2999 if (direction == PF_OUT) {
3000 bport = nport = th->th_sport;
3001 /* check outgoing packet for BINAT/NAT */
3002 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3003 saddr, th->th_sport, daddr, th->th_dport,
3004 &pd->naddr, &nport)) != NULL) {
3005 PF_ACPY(&pd->baddr, saddr, af);
3006 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
3007 &th->th_sum, &pd->naddr, nport, 0, af);
3014 bport = nport = th->th_dport;
3015 /* check incoming packet for BINAT/RDR */
3016 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3017 saddr, th->th_sport, daddr, th->th_dport,
3018 &pd->naddr, &nport)) != NULL) {
3019 PF_ACPY(&pd->baddr, daddr, af);
3020 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
3021 &th->th_sum, &pd->naddr, nport, 0, af);
3031 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3032 r = r->skip[PF_SKIP_IFP].ptr;
3033 else if (r->direction && r->direction != direction)
3034 r = r->skip[PF_SKIP_DIR].ptr;
3035 else if (r->af && r->af != af)
3036 r = r->skip[PF_SKIP_AF].ptr;
3037 else if (r->proto && r->proto != IPPROTO_TCP)
3038 r = r->skip[PF_SKIP_PROTO].ptr;
3039 else if (PF_MISMATCHAW(&r->src.addr, saddr, af,
3041 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3042 else if (r->src.port_op && !pf_match_port(r->src.port_op,
3043 r->src.port[0], r->src.port[1], th->th_sport))
3044 r = r->skip[PF_SKIP_SRC_PORT].ptr;
3045 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af,
3047 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3048 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
3049 r->dst.port[0], r->dst.port[1], th->th_dport))
3050 r = r->skip[PF_SKIP_DST_PORT].ptr;
3051 else if (r->tos && !(r->tos == pd->tos))
3052 r = TAILQ_NEXT(r, entries);
3053 else if (r->rule_flag & PFRULE_FRAGMENT)
3054 r = TAILQ_NEXT(r, entries);
3055 else if ((r->flagset & th->th_flags) != r->flags)
3056 r = TAILQ_NEXT(r, entries);
3057 else if (r->uid.op && (pd->lookup.done || (pd->lookup.done =
3058 pf_socket_lookup(direction, pd, inp), 1)) &&
3059 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
3061 r = TAILQ_NEXT(r, entries);
3062 else if (r->gid.op && (pd->lookup.done || (pd->lookup.done =
3063 pf_socket_lookup(direction, pd, inp), 1)) &&
3064 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
3066 r = TAILQ_NEXT(r, entries);
3067 else if (r->prob && r->prob <= karc4random())
3068 r = TAILQ_NEXT(r, entries);
3069 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag))
3070 r = TAILQ_NEXT(r, entries);
3071 else if (r->os_fingerprint != PF_OSFP_ANY && !pf_osfp_match(
3072 pf_osfp_fingerprint(pd, m, off, th), r->os_fingerprint))
3073 r = TAILQ_NEXT(r, entries);
3077 if (r->rtableid >= 0)
3078 rtableid = r->rtableid;
3079 if (r->anchor == NULL) {
3086 r = TAILQ_NEXT(r, entries);
3088 pf_step_into_anchor(&asd, &ruleset,
3089 PF_RULESET_FILTER, &r, &a, &match);
3091 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3092 PF_RULESET_FILTER, &r, &a, &match))
3099 REASON_SET(&reason, PFRES_MATCH);
3101 if (r->log || (nr != NULL && nr->natpass && nr->log)) {
3103 m_copyback(m, off, sizeof(*th), (caddr_t)th);
3104 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr,
3108 if ((r->action == PF_DROP) &&
3109 ((r->rule_flag & PFRULE_RETURNRST) ||
3110 (r->rule_flag & PFRULE_RETURNICMP) ||
3111 (r->rule_flag & PFRULE_RETURN))) {
3112 /* undo NAT changes, if they have taken place */
3114 if (direction == PF_OUT) {
3115 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
3116 &th->th_sum, &pd->baddr, bport, 0, af);
3119 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
3120 &th->th_sum, &pd->baddr, bport, 0, af);
3124 if (((r->rule_flag & PFRULE_RETURNRST) ||
3125 (r->rule_flag & PFRULE_RETURN)) &&
3126 !(th->th_flags & TH_RST)) {
3127 u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
3129 if (th->th_flags & TH_SYN)
3131 if (th->th_flags & TH_FIN)
3133 pf_send_tcp(r, af, pd->dst,
3134 pd->src, th->th_dport, th->th_sport,
3135 ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0,
3136 r->return_ttl, 1, 0, pd->eh, kif->pfik_ifp);
3137 } else if ((af == AF_INET) && r->return_icmp)
3138 pf_send_icmp(m, r->return_icmp >> 8,
3139 r->return_icmp & 255, af, r);
3140 else if ((af == AF_INET6) && r->return_icmp6)
3141 pf_send_icmp(m, r->return_icmp6 >> 8,
3142 r->return_icmp6 & 255, af, r);
3145 if (r->action == PF_DROP) {
3149 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) {
3150 REASON_SET(&reason, PFRES_MEMORY);
3154 if (r->keep_state || nr != NULL ||
3155 (pd->flags & PFDESC_TCP_NORM)) {
3156 /* create new state */
3158 struct pf_state *s = NULL;
3159 struct pf_src_node *sn = NULL;
3161 len = pd->tot_len - off - (th->th_off << 2);
3163 /* check maximums */
3164 if (r->max_states && (r->states >= r->max_states)) {
3165 pf_status.lcounters[LCNT_STATES]++;
3166 REASON_SET(&reason, PFRES_MAXSTATES);
3169 /* src node for filter rule */
3170 if ((r->rule_flag & PFRULE_SRCTRACK ||
3171 r->rpool.opts & PF_POOL_STICKYADDR) &&
3172 pf_insert_src_node(&sn, r, saddr, af) != 0) {
3173 REASON_SET(&reason, PFRES_SRCLIMIT);
3176 /* src node for translation rule */
3177 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3178 ((direction == PF_OUT &&
3179 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3180 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) {
3181 REASON_SET(&reason, PFRES_SRCLIMIT);
3184 s = pool_get(&pf_state_pl, PR_NOWAIT);
3186 REASON_SET(&reason, PFRES_MEMORY);
3188 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3189 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3190 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3191 pf_status.src_nodes--;
3192 pool_put(&pf_src_tree_pl, sn);
3194 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3196 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3197 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3198 pf_status.src_nodes--;
3199 pool_put(&pf_src_tree_pl, nsn);
3203 bzero(s, sizeof(*s));
3205 s->nat_rule.ptr = nr;
3207 STATE_INC_COUNTERS(s);
3208 s->allow_opts = r->allow_opts;
3209 s->log = r->log & PF_LOG_ALL;
3211 s->log |= nr->log & PF_LOG_ALL;
3212 s->proto = IPPROTO_TCP;
3213 s->direction = direction;
3215 if (direction == PF_OUT) {
3216 PF_ACPY(&s->gwy.addr, saddr, af);
3217 s->gwy.port = th->th_sport; /* sport */
3218 PF_ACPY(&s->ext.addr, daddr, af);
3219 s->ext.port = th->th_dport;
3221 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3222 s->lan.port = bport;
3224 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3225 s->lan.port = s->gwy.port;
3228 PF_ACPY(&s->lan.addr, daddr, af);
3229 s->lan.port = th->th_dport;
3230 PF_ACPY(&s->ext.addr, saddr, af);
3231 s->ext.port = th->th_sport;
3233 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3234 s->gwy.port = bport;
3236 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3237 s->gwy.port = s->lan.port;
3241 s->hash = pf_state_hash(s);
3242 s->src.seqlo = ntohl(th->th_seq);
3243 s->src.seqhi = s->src.seqlo + len + 1;
3244 s->pickup_mode = r->pickup_mode;
3246 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
3247 r->keep_state == PF_STATE_MODULATE) {
3248 /* Generate sequence number modulator */
3249 while ((s->src.seqdiff =
3250 pf_new_isn(s) - s->src.seqlo) == 0)
3252 pf_change_a(&th->th_seq, &th->th_sum,
3253 htonl(s->src.seqlo + s->src.seqdiff), 0);
3257 if (th->th_flags & TH_SYN) {
3259 s->src.wscale = pf_get_wscale(m, off, th->th_off, af);
3260 s->sync_flags |= PFSTATE_GOT_SYN1;
3262 s->src.max_win = MAX(ntohs(th->th_win), 1);
3263 if (s->src.wscale & PF_WSCALE_MASK) {
3264 /* Remove scale factor from initial window */
3265 u_int win = s->src.max_win;
3266 win += 1 << (s->src.wscale & PF_WSCALE_MASK);
3267 s->src.max_win = (win - 1) >>
3268 (s->src.wscale & PF_WSCALE_MASK);
3270 if (th->th_flags & TH_FIN)
3274 s->src.state = TCPS_SYN_SENT;
3275 s->dst.state = TCPS_CLOSED;
3276 s->creation = time_second;
3277 s->expire = time_second;
3278 s->timeout = PFTM_TCP_FIRST_PACKET;
3279 pf_set_rt_ifp(s, saddr);
3282 s->src_node->states++;
3285 PF_ACPY(&nsn->raddr, &pd->naddr, af);
3286 s->nat_src_node = nsn;
3287 s->nat_src_node->states++;
3289 if ((pd->flags & PFDESC_TCP_NORM) && pf_normalize_tcp_init(m,
3290 off, pd, th, &s->src, &s->dst)) {
3291 REASON_SET(&reason, PFRES_MEMORY);
3292 pf_src_tree_remove_state(s);
3293 STATE_DEC_COUNTERS(s);
3294 pool_put(&pf_state_pl, s);
3297 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
3298 pf_normalize_tcp_stateful(m, off, pd, &reason, th, s,
3299 &s->src, &s->dst, &rewrite)) {
3300 /* This really shouldn't happen!!! */
3301 DPFPRINTF(PF_DEBUG_URGENT,
3302 ("pf_normalize_tcp_stateful failed on first pkt"));
3303 pf_normalize_tcp_cleanup(s);
3304 pf_src_tree_remove_state(s);
3305 STATE_DEC_COUNTERS(s);
3306 pool_put(&pf_state_pl, s);
3309 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3310 pf_normalize_tcp_cleanup(s);
3311 REASON_SET(&reason, PFRES_STATEINS);
3312 pf_src_tree_remove_state(s);
3313 STATE_DEC_COUNTERS(s);
3314 pool_put(&pf_state_pl, s);
3322 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
3323 r->keep_state == PF_STATE_SYNPROXY) {
3324 s->src.state = PF_TCPS_PROXY_SRC;
3326 if (direction == PF_OUT) {
3327 pf_change_ap(saddr, &th->th_sport,
3328 pd->ip_sum, &th->th_sum, &pd->baddr,
3331 pf_change_ap(daddr, &th->th_dport,
3332 pd->ip_sum, &th->th_sum, &pd->baddr,
3336 s->src.seqhi = htonl(karc4random());
3337 /* Find mss option */
3338 mss = pf_get_mss(m, off, th->th_off, af);
3339 mss = pf_calc_mss(saddr, af, mss);
3340 mss = pf_calc_mss(daddr, af, mss);
3342 pf_send_tcp(r, af, daddr, saddr, th->th_dport,
3343 th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1,
3344 TH_SYN|TH_ACK, 0, s->src.mss, 0, 1, 0, NULL, NULL);
3345 REASON_SET(&reason, PFRES_SYNPROXY);
3346 return (PF_SYNPROXY_DROP);
3350 /* copy back packet headers if we performed NAT operations */
3352 m_copyback(m, off, sizeof(*th), (caddr_t)th);
3358 pf_test_udp(struct pf_rule **rm, struct pf_state **sm, int direction,
3359 struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3360 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
3361 struct ifqueue *ifq, struct inpcb *inp)
3363 struct pf_rule *nr = NULL;
3364 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
3365 struct udphdr *uh = pd->hdr.udp;
3366 u_int16_t bport, nport = 0;
3367 sa_family_t af = pd->af;
3368 struct pf_rule *r, *a = NULL;
3369 struct pf_ruleset *ruleset = NULL;
3370 struct pf_src_node *nsn = NULL;
3373 int tag = -1, rtableid = -1;
3377 if (pf_check_congestion(ifq)) {
3378 REASON_SET(&reason, PFRES_CONGEST);
3382 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3384 if (direction == PF_OUT) {
3385 bport = nport = uh->uh_sport;
3386 /* check outgoing packet for BINAT/NAT */
3387 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3388 saddr, uh->uh_sport, daddr, uh->uh_dport,
3389 &pd->naddr, &nport)) != NULL) {
3390 PF_ACPY(&pd->baddr, saddr, af);
3391 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum,
3392 &uh->uh_sum, &pd->naddr, nport, 1, af);
3399 bport = nport = uh->uh_dport;
3400 /* check incoming packet for BINAT/RDR */
3401 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3402 saddr, uh->uh_sport, daddr, uh->uh_dport, &pd->naddr,
3404 PF_ACPY(&pd->baddr, daddr, af);
3405 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum,
3406 &uh->uh_sum, &pd->naddr, nport, 1, af);
3416 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3417 r = r->skip[PF_SKIP_IFP].ptr;
3418 else if (r->direction && r->direction != direction)
3419 r = r->skip[PF_SKIP_DIR].ptr;
3420 else if (r->af && r->af != af)
3421 r = r->skip[PF_SKIP_AF].ptr;
3422 else if (r->proto && r->proto != IPPROTO_UDP)
3423 r = r->skip[PF_SKIP_PROTO].ptr;
3424 else if (PF_MISMATCHAW(&r->src.addr, saddr, af,
3426 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3427 else if (r->src.port_op && !pf_match_port(r->src.port_op,
3428 r->src.port[0], r->src.port[1], uh->uh_sport))
3429 r = r->skip[PF_SKIP_SRC_PORT].ptr;
3430 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af,
3432 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3433 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
3434 r->dst.port[0], r->dst.port[1], uh->uh_dport))
3435 r = r->skip[PF_SKIP_DST_PORT].ptr;
3436 else if (r->tos && !(r->tos == pd->tos))
3437 r = TAILQ_NEXT(r, entries);
3438 else if (r->rule_flag & PFRULE_FRAGMENT)
3439 r = TAILQ_NEXT(r, entries);
3440 else if (r->uid.op && (pd->lookup.done || (pd->lookup.done =
3441 pf_socket_lookup(direction, pd, inp), 1)) &&
3442 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
3444 r = TAILQ_NEXT(r, entries);
3445 else if (r->gid.op && (pd->lookup.done || (pd->lookup.done =
3446 pf_socket_lookup(direction, pd, inp), 1)) &&
3447 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
3449 r = TAILQ_NEXT(r, entries);
3450 else if (r->prob && r->prob <= karc4random())
3451 r = TAILQ_NEXT(r, entries);
3452 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag))
3453 r = TAILQ_NEXT(r, entries);
3454 else if (r->os_fingerprint != PF_OSFP_ANY)
3455 r = TAILQ_NEXT(r, entries);
3459 if (r->rtableid >= 0)
3460 rtableid = r->rtableid;
3461 if (r->anchor == NULL) {
3468 r = TAILQ_NEXT(r, entries);
3470 pf_step_into_anchor(&asd, &ruleset,
3471 PF_RULESET_FILTER, &r, &a, &match);
3473 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3474 PF_RULESET_FILTER, &r, &a, &match))
3481 REASON_SET(&reason, PFRES_MATCH);
3483 if (r->log || (nr != NULL && nr->natpass && nr->log)) {
3485 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
3486 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr,
3490 if ((r->action == PF_DROP) &&
3491 ((r->rule_flag & PFRULE_RETURNICMP) ||
3492 (r->rule_flag & PFRULE_RETURN))) {
3493 /* undo NAT changes, if they have taken place */
3495 if (direction == PF_OUT) {
3496 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum,
3497 &uh->uh_sum, &pd->baddr, bport, 1, af);
3500 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum,
3501 &uh->uh_sum, &pd->baddr, bport, 1, af);
3505 if ((af == AF_INET) && r->return_icmp)
3506 pf_send_icmp(m, r->return_icmp >> 8,
3507 r->return_icmp & 255, af, r);
3508 else if ((af == AF_INET6) && r->return_icmp6)
3509 pf_send_icmp(m, r->return_icmp6 >> 8,
3510 r->return_icmp6 & 255, af, r);
3513 if (r->action == PF_DROP)
3516 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) {
3517 REASON_SET(&reason, PFRES_MEMORY);
3521 if (r->keep_state || nr != NULL) {
3522 /* create new state */
3523 struct pf_state *s = NULL;
3524 struct pf_src_node *sn = NULL;
3526 /* check maximums */
3527 if (r->max_states && (r->states >= r->max_states)) {
3528 pf_status.lcounters[LCNT_STATES]++;
3529 REASON_SET(&reason, PFRES_MAXSTATES);
3532 /* src node for filter rule */
3533 if ((r->rule_flag & PFRULE_SRCTRACK ||
3534 r->rpool.opts & PF_POOL_STICKYADDR) &&
3535 pf_insert_src_node(&sn, r, saddr, af) != 0) {
3536 REASON_SET(&reason, PFRES_SRCLIMIT);
3539 /* src node for translation rule */
3540 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3541 ((direction == PF_OUT &&
3542 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3543 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) {
3544 REASON_SET(&reason, PFRES_SRCLIMIT);
3547 s = pool_get(&pf_state_pl, PR_NOWAIT);
3549 REASON_SET(&reason, PFRES_MEMORY);
3551 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3552 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3553 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3554 pf_status.src_nodes--;
3555 pool_put(&pf_src_tree_pl, sn);
3557 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3559 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3560 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3561 pf_status.src_nodes--;
3562 pool_put(&pf_src_tree_pl, nsn);
3566 bzero(s, sizeof(*s));
3568 s->nat_rule.ptr = nr;
3570 STATE_INC_COUNTERS(s);
3571 s->allow_opts = r->allow_opts;
3572 s->log = r->log & PF_LOG_ALL;
3574 s->log |= nr->log & PF_LOG_ALL;
3575 s->proto = IPPROTO_UDP;
3576 s->direction = direction;
3578 if (direction == PF_OUT) {
3579 PF_ACPY(&s->gwy.addr, saddr, af);
3580 s->gwy.port = uh->uh_sport;
3581 PF_ACPY(&s->ext.addr, daddr, af);
3582 s->ext.port = uh->uh_dport;
3584 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3585 s->lan.port = bport;
3587 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3588 s->lan.port = s->gwy.port;
3591 PF_ACPY(&s->lan.addr, daddr, af);
3592 s->lan.port = uh->uh_dport;
3593 PF_ACPY(&s->ext.addr, saddr, af);
3594 s->ext.port = uh->uh_sport;
3596 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3597 s->gwy.port = bport;
3599 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3600 s->gwy.port = s->lan.port;
3603 s->hash = pf_state_hash(s);
3604 s->src.state = PFUDPS_SINGLE;
3605 s->dst.state = PFUDPS_NO_TRAFFIC;
3606 s->creation = time_second;
3607 s->expire = time_second;
3608 s->timeout = PFTM_UDP_FIRST_PACKET;
3609 pf_set_rt_ifp(s, saddr);
3612 s->src_node->states++;
3615 PF_ACPY(&nsn->raddr, &pd->naddr, af);
3616 s->nat_src_node = nsn;
3617 s->nat_src_node->states++;
3619 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3620 REASON_SET(&reason, PFRES_STATEINS);
3621 pf_src_tree_remove_state(s);
3622 STATE_DEC_COUNTERS(s);
3623 pool_put(&pf_state_pl, s);
3633 /* copy back packet headers if we performed NAT operations */
3635 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
3641 pf_test_icmp(struct pf_rule **rm, struct pf_state **sm, int direction,
3642 struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3643 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
3644 struct ifqueue *ifq)
3646 struct pf_rule *nr = NULL;
3647 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
3648 struct pf_rule *r, *a = NULL;
3649 struct pf_ruleset *ruleset = NULL;
3650 struct pf_src_node *nsn = NULL;
3652 u_int16_t icmpid = 0, bport, nport = 0;
3653 sa_family_t af = pd->af;
3654 u_int8_t icmptype = 0, icmpcode = 0;
3656 int tag = -1, rtableid = -1;
3663 if (pf_check_congestion(ifq)) {
3664 REASON_SET(&reason, PFRES_CONGEST);
3668 switch (pd->proto) {
3671 icmptype = pd->hdr.icmp->icmp_type;
3672 icmpcode = pd->hdr.icmp->icmp_code;
3673 icmpid = pd->hdr.icmp->icmp_id;
3675 if (icmptype == ICMP_UNREACH ||
3676 icmptype == ICMP_SOURCEQUENCH ||
3677 icmptype == ICMP_REDIRECT ||
3678 icmptype == ICMP_TIMXCEED ||
3679 icmptype == ICMP_PARAMPROB)
3684 case IPPROTO_ICMPV6:
3685 icmptype = pd->hdr.icmp6->icmp6_type;
3686 icmpcode = pd->hdr.icmp6->icmp6_code;
3687 icmpid = pd->hdr.icmp6->icmp6_id;
3689 if (icmptype == ICMP6_DST_UNREACH ||
3690 icmptype == ICMP6_PACKET_TOO_BIG ||
3691 icmptype == ICMP6_TIME_EXCEEDED ||
3692 icmptype == ICMP6_PARAM_PROB)
3698 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3700 if (direction == PF_OUT) {
3701 bport = nport = icmpid;
3702 /* check outgoing packet for BINAT/NAT */
3703 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3704 saddr, icmpid, daddr, icmpid, &pd->naddr, &nport)) !=
3706 PF_ACPY(&pd->baddr, saddr, af);
3710 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3711 pd->naddr.v4.s_addr, 0);
3712 pd->hdr.icmp->icmp_cksum = pf_cksum_fixup(
3713 pd->hdr.icmp->icmp_cksum, icmpid, nport, 0);
3714 pd->hdr.icmp->icmp_id = nport;
3715 m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
3720 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum,
3731 bport = nport = icmpid;
3732 /* check incoming packet for BINAT/RDR */
3733 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3734 saddr, icmpid, daddr, icmpid, &pd->naddr, &nport)) !=
3736 PF_ACPY(&pd->baddr, daddr, af);
3740 pf_change_a(&daddr->v4.s_addr,
3741 pd->ip_sum, pd->naddr.v4.s_addr, 0);
3746 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum,
3760 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3761 r = r->skip[PF_SKIP_IFP].ptr;
3762 else if (r->direction && r->direction != direction)
3763 r = r->skip[PF_SKIP_DIR].ptr;
3764 else if (r->af && r->af != af)
3765 r = r->skip[PF_SKIP_AF].ptr;
3766 else if (r->proto && r->proto != pd->proto)
3767 r = r->skip[PF_SKIP_PROTO].ptr;
3768 else if (PF_MISMATCHAW(&r->src.addr, saddr, af,
3770 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3771 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af,
3773 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3774 else if (r->type && r->type != icmptype + 1)
3775 r = TAILQ_NEXT(r, entries);
3776 else if (r->code && r->code != icmpcode + 1)
3777 r = TAILQ_NEXT(r, entries);
3778 else if (r->tos && !(r->tos == pd->tos))
3779 r = TAILQ_NEXT(r, entries);
3780 else if (r->rule_flag & PFRULE_FRAGMENT)
3781 r = TAILQ_NEXT(r, entries);
3782 else if (r->prob && r->prob <= karc4random())
3783 r = TAILQ_NEXT(r, entries);
3784 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag))
3785 r = TAILQ_NEXT(r, entries);
3786 else if (r->os_fingerprint != PF_OSFP_ANY)
3787 r = TAILQ_NEXT(r, entries);
3791 if (r->rtableid >= 0)
3792 rtableid = r->rtableid;
3793 if (r->anchor == NULL) {
3800 r = TAILQ_NEXT(r, entries);
3802 pf_step_into_anchor(&asd, &ruleset,
3803 PF_RULESET_FILTER, &r, &a, &match);
3805 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
3806 PF_RULESET_FILTER, &r, &a, &match))
3813 REASON_SET(&reason, PFRES_MATCH);
3815 if (r->log || (nr != NULL && nr->natpass && nr->log)) {
3818 m_copyback(m, off, sizeof(struct icmp6_hdr),
3819 (caddr_t)pd->hdr.icmp6);
3821 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr,
3825 if (r->action != PF_PASS)
3828 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) {
3829 REASON_SET(&reason, PFRES_MEMORY);
3833 if (!state_icmp && (r->keep_state || nr != NULL)) {
3834 /* create new state */
3835 struct pf_state *s = NULL;
3836 struct pf_src_node *sn = NULL;
3838 /* check maximums */
3839 if (r->max_states && (r->states >= r->max_states)) {
3840 pf_status.lcounters[LCNT_STATES]++;
3841 REASON_SET(&reason, PFRES_MAXSTATES);
3844 /* src node for filter rule */
3845 if ((r->rule_flag & PFRULE_SRCTRACK ||
3846 r->rpool.opts & PF_POOL_STICKYADDR) &&
3847 pf_insert_src_node(&sn, r, saddr, af) != 0) {
3848 REASON_SET(&reason, PFRES_SRCLIMIT);
3851 /* src node for translation rule */
3852 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3853 ((direction == PF_OUT &&
3854 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3855 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) {
3856 REASON_SET(&reason, PFRES_SRCLIMIT);
3859 s = pool_get(&pf_state_pl, PR_NOWAIT);
3861 REASON_SET(&reason, PFRES_MEMORY);
3863 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3864 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3865 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3866 pf_status.src_nodes--;
3867 pool_put(&pf_src_tree_pl, sn);
3869 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3871 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3872 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3873 pf_status.src_nodes--;
3874 pool_put(&pf_src_tree_pl, nsn);
3878 bzero(s, sizeof(*s));
3880 s->nat_rule.ptr = nr;
3882 STATE_INC_COUNTERS(s);
3883 s->allow_opts = r->allow_opts;
3884 s->log = r->log & PF_LOG_ALL;
3886 s->log |= nr->log & PF_LOG_ALL;
3887 s->proto = pd->proto;
3888 s->direction = direction;
3890 if (direction == PF_OUT) {
3891 PF_ACPY(&s->gwy.addr, saddr, af);
3892 s->gwy.port = nport;
3893 PF_ACPY(&s->ext.addr, daddr, af);
3896 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3897 s->lan.port = bport;
3899 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3900 s->lan.port = s->gwy.port;
3903 PF_ACPY(&s->lan.addr, daddr, af);
3904 s->lan.port = nport;
3905 PF_ACPY(&s->ext.addr, saddr, af);
3908 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3909 s->gwy.port = bport;
3911 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3912 s->gwy.port = s->lan.port;
3915 s->hash = pf_state_hash(s);
3916 s->creation = time_second;
3917 s->expire = time_second;
3918 s->timeout = PFTM_ICMP_FIRST_PACKET;
3919 pf_set_rt_ifp(s, saddr);
3922 s->src_node->states++;
3925 PF_ACPY(&nsn->raddr, &pd->naddr, af);
3926 s->nat_src_node = nsn;
3927 s->nat_src_node->states++;
3929 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3930 REASON_SET(&reason, PFRES_STATEINS);
3931 pf_src_tree_remove_state(s);
3932 STATE_DEC_COUNTERS(s);
3933 pool_put(&pf_state_pl, s);
3944 /* copy back packet headers if we performed IPv6 NAT operations */
3946 m_copyback(m, off, sizeof(struct icmp6_hdr),
3947 (caddr_t)pd->hdr.icmp6);
3954 pf_test_other(struct pf_rule **rm, struct pf_state **sm, int direction,
3955 struct pfi_kif *kif, struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
3956 struct pf_rule **am, struct pf_ruleset **rsm, struct ifqueue *ifq)
3958 struct pf_rule *nr = NULL;
3959 struct pf_rule *r, *a = NULL;
3960 struct pf_ruleset *ruleset = NULL;
3961 struct pf_src_node *nsn = NULL;
3962 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
3963 sa_family_t af = pd->af;
3965 int tag = -1, rtableid = -1;
3969 if (pf_check_congestion(ifq)) {
3970 REASON_SET(&reason, PFRES_CONGEST);
3974 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3976 if (direction == PF_OUT) {
3977 /* check outgoing packet for BINAT/NAT */
3978 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3979 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3980 PF_ACPY(&pd->baddr, saddr, af);
3984 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3985 pd->naddr.v4.s_addr, 0);
3990 PF_ACPY(saddr, &pd->naddr, af);
3999 /* check incoming packet for BINAT/RDR */
4000 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
4001 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
4002 PF_ACPY(&pd->baddr, daddr, af);
4006 pf_change_a(&daddr->v4.s_addr,
4007 pd->ip_sum, pd->naddr.v4.s_addr, 0);
4012 PF_ACPY(daddr, &pd->naddr, af);
4024 if (pfi_kif_match(r->kif, kif) == r->ifnot)
4025 r = r->skip[PF_SKIP_IFP].ptr;
4026 else if (r->direction && r->direction != direction)
4027 r = r->skip[PF_SKIP_DIR].ptr;
4028 else if (r->af && r->af != af)
4029 r = r->skip[PF_SKIP_AF].ptr;
4030 else if (r->proto && r->proto != pd->proto)
4031 r = r->skip[PF_SKIP_PROTO].ptr;
4032 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af,
4034 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
4035 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af,
4037 r = r->skip[PF_SKIP_DST_ADDR].ptr;
4038 else if (r->tos && !(r->tos == pd->tos))
4039 r = TAILQ_NEXT(r, entries);
4040 else if (r->rule_flag & PFRULE_FRAGMENT)
4041 r = TAILQ_NEXT(r, entries);
4042 else if (r->prob && r->prob <= karc4random())
4043 r = TAILQ_NEXT(r, entries);
4044 else if (r->match_tag && !pf_match_tag(m, r, pd->pf_mtag, &tag))
4045 r = TAILQ_NEXT(r, entries);
4046 else if (r->os_fingerprint != PF_OSFP_ANY)
4047 r = TAILQ_NEXT(r, entries);
4051 if (r->rtableid >= 0)
4052 rtableid = r->rtableid;
4053 if (r->anchor == NULL) {
4060 r = TAILQ_NEXT(r, entries);
4062 pf_step_into_anchor(&asd, &ruleset,
4063 PF_RULESET_FILTER, &r, &a, &match);
4065 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
4066 PF_RULESET_FILTER, &r, &a, &match))
4073 REASON_SET(&reason, PFRES_MATCH);
4075 if (r->log || (nr != NULL && nr->natpass && nr->log))
4076 PFLOG_PACKET(kif, h, m, af, direction, reason, r->log ? r : nr,
4079 if ((r->action == PF_DROP) &&
4080 ((r->rule_flag & PFRULE_RETURNICMP) ||
4081 (r->rule_flag & PFRULE_RETURN))) {
4082 struct pf_addr *a = NULL;
4085 if (direction == PF_OUT)
4094 pf_change_a(&a->v4.s_addr, pd->ip_sum,
4095 pd->baddr.v4.s_addr, 0);
4100 PF_ACPY(a, &pd->baddr, af);
4105 if ((af == AF_INET) && r->return_icmp)
4106 pf_send_icmp(m, r->return_icmp >> 8,
4107 r->return_icmp & 255, af, r);
4108 else if ((af == AF_INET6) && r->return_icmp6)
4109 pf_send_icmp(m, r->return_icmp6 >> 8,
4110 r->return_icmp6 & 255, af, r);
4113 if (r->action != PF_PASS)
4116 if (pf_tag_packet(m, pd->pf_mtag, tag, rtableid)) {
4117 REASON_SET(&reason, PFRES_MEMORY);
4121 if (r->keep_state || nr != NULL) {
4122 /* create new state */
4123 struct pf_state *s = NULL;
4124 struct pf_src_node *sn = NULL;
4126 /* check maximums */
4127 if (r->max_states && (r->states >= r->max_states)) {
4128 pf_status.lcounters[LCNT_STATES]++;
4129 REASON_SET(&reason, PFRES_MAXSTATES);
4132 /* src node for filter rule */
4133 if ((r->rule_flag & PFRULE_SRCTRACK ||
4134 r->rpool.opts & PF_POOL_STICKYADDR) &&
4135 pf_insert_src_node(&sn, r, saddr, af) != 0) {
4136 REASON_SET(&reason, PFRES_SRCLIMIT);
4139 /* src node for translation rule */
4140 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
4141 ((direction == PF_OUT &&
4142 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
4143 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) {
4144 REASON_SET(&reason, PFRES_SRCLIMIT);
4147 s = pool_get(&pf_state_pl, PR_NOWAIT);
4149 REASON_SET(&reason, PFRES_MEMORY);
4151 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
4152 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
4153 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
4154 pf_status.src_nodes--;
4155 pool_put(&pf_src_tree_pl, sn);
4157 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
4159 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
4160 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
4161 pf_status.src_nodes--;
4162 pool_put(&pf_src_tree_pl, nsn);
4166 bzero(s, sizeof(*s));
4168 s->nat_rule.ptr = nr;
4170 STATE_INC_COUNTERS(s);
4171 s->allow_opts = r->allow_opts;
4172 s->log = r->log & PF_LOG_ALL;
4174 s->log |= nr->log & PF_LOG_ALL;
4175 s->proto = pd->proto;
4176 s->direction = direction;
4178 if (direction == PF_OUT) {
4179 PF_ACPY(&s->gwy.addr, saddr, af);
4180 PF_ACPY(&s->ext.addr, daddr, af);
4182 PF_ACPY(&s->lan.addr, &pd->baddr, af);
4184 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
4186 PF_ACPY(&s->lan.addr, daddr, af);
4187 PF_ACPY(&s->ext.addr, saddr, af);
4189 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
4191 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
4193 s->hash = pf_state_hash(s);
4194 s->src.state = PFOTHERS_SINGLE;
4195 s->dst.state = PFOTHERS_NO_TRAFFIC;
4196 s->creation = time_second;
4197 s->expire = time_second;
4198 s->timeout = PFTM_OTHER_FIRST_PACKET;
4199 pf_set_rt_ifp(s, saddr);
4202 s->src_node->states++;
4205 PF_ACPY(&nsn->raddr, &pd->naddr, af);
4206 s->nat_src_node = nsn;
4207 s->nat_src_node->states++;
4209 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
4210 REASON_SET(&reason, PFRES_STATEINS);
4211 pf_src_tree_remove_state(s);
4212 STATE_DEC_COUNTERS(s);
4213 pool_put(&pf_state_pl, s);
4227 pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif,
4228 struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am,
4229 struct pf_ruleset **rsm)
4231 struct pf_rule *r, *a = NULL;
4232 struct pf_ruleset *ruleset = NULL;
4233 sa_family_t af = pd->af;
4239 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
4242 if (pfi_kif_match(r->kif, kif) == r->ifnot)
4243 r = r->skip[PF_SKIP_IFP].ptr;
4244 else if (r->direction && r->direction != direction)
4245 r = r->skip[PF_SKIP_DIR].ptr;
4246 else if (r->af && r->af != af)
4247 r = r->skip[PF_SKIP_AF].ptr;
4248 else if (r->proto && r->proto != pd->proto)
4249 r = r->skip[PF_SKIP_PROTO].ptr;
4250 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af,
4252 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
4253 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af,
4255 r = r->skip[PF_SKIP_DST_ADDR].ptr;
4256 else if (r->tos && !(r->tos == pd->tos))
4257 r = TAILQ_NEXT(r, entries);
4258 else if (r->os_fingerprint != PF_OSFP_ANY)
4259 r = TAILQ_NEXT(r, entries);
4260 else if (pd->proto == IPPROTO_UDP &&
4261 (r->src.port_op || r->dst.port_op))
4262 r = TAILQ_NEXT(r, entries);
4263 else if (pd->proto == IPPROTO_TCP &&
4264 (r->src.port_op || r->dst.port_op || r->flagset))
4265 r = TAILQ_NEXT(r, entries);
4266 else if ((pd->proto == IPPROTO_ICMP ||
4267 pd->proto == IPPROTO_ICMPV6) &&
4268 (r->type || r->code))
4269 r = TAILQ_NEXT(r, entries);
4270 else if (r->prob && r->prob <= karc4random())
4271 r = TAILQ_NEXT(r, entries);
4272 else if (r->match_tag && !pf_match_tag(m, r, NULL, &tag))
4273 r = TAILQ_NEXT(r, entries);
4275 if (r->anchor == NULL) {
4282 r = TAILQ_NEXT(r, entries);
4284 pf_step_into_anchor(&asd, &ruleset,
4285 PF_RULESET_FILTER, &r, &a, &match);
4287 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
4288 PF_RULESET_FILTER, &r, &a, &match))
4295 REASON_SET(&reason, PFRES_MATCH);
4298 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset,
4301 if (r->action != PF_PASS)
4304 if (pf_tag_packet(m, pd->pf_mtag, tag, -1)) {
4305 REASON_SET(&reason, PFRES_MEMORY);
4313 pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
4314 struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
4317 struct pf_state_cmp key;
4318 struct tcphdr *th = pd->hdr.tcp;
4319 u_int16_t win = ntohs(th->th_win);
4320 u_int32_t ack, end, seq, orig_seq;
4324 struct pf_state_peer *src, *dst;
4327 key.proto = IPPROTO_TCP;
4328 if (direction == PF_IN) {
4329 PF_ACPY(&key.ext.addr, pd->src, key.af);
4330 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4331 key.ext.port = th->th_sport;
4332 key.gwy.port = th->th_dport;
4334 PF_ACPY(&key.lan.addr, pd->src, key.af);
4335 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4336 key.lan.port = th->th_sport;
4337 key.ext.port = th->th_dport;
4342 if (direction == (*state)->direction) {
4343 src = &(*state)->src;
4344 dst = &(*state)->dst;
4346 src = &(*state)->dst;
4347 dst = &(*state)->src;
4350 if ((*state)->src.state == PF_TCPS_PROXY_SRC) {
4351 if (direction != (*state)->direction) {
4352 REASON_SET(reason, PFRES_SYNPROXY);
4353 return (PF_SYNPROXY_DROP);
4355 if (th->th_flags & TH_SYN) {
4356 if (ntohl(th->th_seq) != (*state)->src.seqlo) {
4357 REASON_SET(reason, PFRES_SYNPROXY);
4360 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
4361 pd->src, th->th_dport, th->th_sport,
4362 (*state)->src.seqhi, ntohl(th->th_seq) + 1,
4363 TH_SYN|TH_ACK, 0, (*state)->src.mss, 0, 1,
4365 REASON_SET(reason, PFRES_SYNPROXY);
4366 return (PF_SYNPROXY_DROP);
4367 } else if (!(th->th_flags & TH_ACK) ||
4368 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
4369 (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
4370 REASON_SET(reason, PFRES_SYNPROXY);
4372 } else if ((*state)->src_node != NULL &&
4373 pf_src_connlimit(state)) {
4374 REASON_SET(reason, PFRES_SRCLIMIT);
4377 (*state)->src.state = PF_TCPS_PROXY_DST;
4379 if ((*state)->src.state == PF_TCPS_PROXY_DST) {
4380 struct pf_state_host *src, *dst;
4382 if (direction == PF_OUT) {
4383 src = &(*state)->gwy;
4384 dst = &(*state)->ext;
4386 src = &(*state)->ext;
4387 dst = &(*state)->lan;
4389 if (direction == (*state)->direction) {
4390 if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) ||
4391 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
4392 (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) {
4393 REASON_SET(reason, PFRES_SYNPROXY);
4396 (*state)->src.max_win = MAX(ntohs(th->th_win), 1);
4397 if ((*state)->dst.seqhi == 1)
4398 (*state)->dst.seqhi = htonl(karc4random());
4399 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
4400 &dst->addr, src->port, dst->port,
4401 (*state)->dst.seqhi, 0, TH_SYN, 0,
4402 (*state)->src.mss, 0, 0, (*state)->tag, NULL, NULL);
4403 REASON_SET(reason, PFRES_SYNPROXY);
4404 return (PF_SYNPROXY_DROP);
4405 } else if (((th->th_flags & (TH_SYN|TH_ACK)) !=
4407 (ntohl(th->th_ack) != (*state)->dst.seqhi + 1)) {
4408 REASON_SET(reason, PFRES_SYNPROXY);
4411 (*state)->dst.max_win = MAX(ntohs(th->th_win), 1);
4412 (*state)->dst.seqlo = ntohl(th->th_seq);
4413 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
4414 pd->src, th->th_dport, th->th_sport,
4415 ntohl(th->th_ack), ntohl(th->th_seq) + 1,
4416 TH_ACK, (*state)->src.max_win, 0, 0, 0,
4417 (*state)->tag, NULL, NULL);
4418 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
4419 &dst->addr, src->port, dst->port,
4420 (*state)->src.seqhi + 1, (*state)->src.seqlo + 1,
4421 TH_ACK, (*state)->dst.max_win, 0, 0, 1,
4423 (*state)->src.seqdiff = (*state)->dst.seqhi -
4424 (*state)->src.seqlo;
4425 (*state)->dst.seqdiff = (*state)->src.seqhi -
4426 (*state)->dst.seqlo;
4427 (*state)->src.seqhi = (*state)->src.seqlo +
4428 (*state)->dst.max_win;
4429 (*state)->dst.seqhi = (*state)->dst.seqlo +
4430 (*state)->src.max_win;
4431 (*state)->src.wscale = (*state)->dst.wscale = 0;
4432 (*state)->src.state = (*state)->dst.state =
4434 REASON_SET(reason, PFRES_SYNPROXY);
4435 return (PF_SYNPROXY_DROP);
4439 if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) {
4440 sws = src->wscale & PF_WSCALE_MASK;
4441 dws = dst->wscale & PF_WSCALE_MASK;
4446 * Sequence tracking algorithm from Guido van Rooij's paper:
4447 * http://www.madison-gurkha.com/publications/tcp_filtering/
4451 orig_seq = seq = ntohl(th->th_seq);
4452 if (src->seqlo == 0) {
4453 /* First packet from this end. Set its state */
4455 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) &&
4456 src->scrub == NULL) {
4457 if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) {
4458 REASON_SET(reason, PFRES_MEMORY);
4463 /* Deferred generation of sequence number modulator */
4464 if (dst->seqdiff && !src->seqdiff) {
4465 while ((src->seqdiff = pf_new_isn(*state) - seq) == 0)
4467 ack = ntohl(th->th_ack) - dst->seqdiff;
4468 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
4470 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
4473 ack = ntohl(th->th_ack);
4476 end = seq + pd->p_len;
4477 if (th->th_flags & TH_SYN) {
4479 (*state)->sync_flags |= PFSTATE_GOT_SYN2;
4480 if (dst->wscale & PF_WSCALE_FLAG) {
4481 src->wscale = pf_get_wscale(m, off, th->th_off,
4483 if (src->wscale & PF_WSCALE_FLAG) {
4484 /* Remove scale factor from initial
4486 sws = src->wscale & PF_WSCALE_MASK;
4487 win = ((u_int32_t)win + (1 << sws) - 1)
4489 dws = dst->wscale & PF_WSCALE_MASK;
4491 /* fixup other window */
4492 dst->max_win <<= dst->wscale &
4494 /* in case of a retrans SYN|ACK */
4499 if (th->th_flags & TH_FIN)
4503 if (src->state < TCPS_SYN_SENT)
4504 src->state = TCPS_SYN_SENT;
4507 * May need to slide the window (seqhi may have been set by
4508 * the crappy stack check or if we picked up the connection
4509 * after establishment)
4511 if (src->seqhi == 1 ||
4512 SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi))
4513 src->seqhi = end + MAX(1, dst->max_win << dws);
4514 if (win > src->max_win)
4518 ack = ntohl(th->th_ack) - dst->seqdiff;
4520 /* Modulate sequence numbers */
4521 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
4523 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
4526 end = seq + pd->p_len;
4527 if (th->th_flags & TH_SYN)
4529 if (th->th_flags & TH_FIN)
4533 if ((th->th_flags & TH_ACK) == 0) {
4534 /* Let it pass through the ack skew check */
4536 } else if ((ack == 0 &&
4537 (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) ||
4538 /* broken tcp stacks do not set ack */
4539 (dst->state < TCPS_SYN_SENT)) {
4541 * Many stacks (ours included) will set the ACK number in an
4542 * FIN|ACK if the SYN times out -- no sequence to ACK.
4548 /* Ease sequencing restrictions on no data packets */
4553 ackskew = dst->seqlo - ack;
4557 * Need to demodulate the sequence numbers in any TCP SACK options
4558 * (Selective ACK). We could optionally validate the SACK values
4559 * against the current ACK window, either forwards or backwards, but
4560 * I'm not confident that SACK has been implemented properly
4561 * everywhere. It wouldn't surprise me if several stacks accidently
4562 * SACK too far backwards of previously ACKed data. There really aren't
4563 * any security implications of bad SACKing unless the target stack
4564 * doesn't validate the option length correctly. Someone trying to
4565 * spoof into a TCP connection won't bother blindly sending SACK
4568 if (dst->seqdiff && (th->th_off << 2) > sizeof(struct tcphdr)) {
4569 if (pf_modulate_sack(m, off, pd, th, dst))
4574 #define MAXACKWINDOW (0xffff + 1500) /* 1500 is an arbitrary fudge factor */
4575 if (SEQ_GEQ(src->seqhi, end) &&
4576 /* Last octet inside other's window space */
4577 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) &&
4578 /* Retrans: not more than one window back */
4579 (ackskew >= -MAXACKWINDOW) &&
4580 /* Acking not more than one reassembled fragment backwards */
4581 (ackskew <= (MAXACKWINDOW << sws)) &&
4582 /* Acking not more than one window forward */
4583 ((th->th_flags & TH_RST) == 0 || orig_seq == src->seqlo ||
4584 (orig_seq == src->seqlo + 1) || (pd->flags & PFDESC_IP_REAS) == 0)) {
4585 /* Require an exact/+1 sequence match on resets when possible */
4587 if (dst->scrub || src->scrub) {
4588 if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4589 *state, src, dst, ©back))
4593 /* update max window */
4594 if (src->max_win < win)
4596 /* synchronize sequencing */
4597 if (SEQ_GT(end, src->seqlo))
4599 /* slide the window of what the other end can send */
4600 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4601 dst->seqhi = ack + MAX((win << sws), 1);
4605 if (th->th_flags & TH_SYN)
4606 if (src->state < TCPS_SYN_SENT)
4607 src->state = TCPS_SYN_SENT;
4608 if (th->th_flags & TH_FIN)
4609 if (src->state < TCPS_CLOSING)
4610 src->state = TCPS_CLOSING;
4611 if (th->th_flags & TH_ACK) {
4612 if (dst->state == TCPS_SYN_SENT) {
4613 dst->state = TCPS_ESTABLISHED;
4614 if (src->state == TCPS_ESTABLISHED &&
4615 (*state)->src_node != NULL &&
4616 pf_src_connlimit(state)) {
4617 REASON_SET(reason, PFRES_SRCLIMIT);
4620 } else if (dst->state == TCPS_CLOSING)
4621 dst->state = TCPS_FIN_WAIT_2;
4623 if (th->th_flags & TH_RST)
4624 src->state = dst->state = TCPS_TIME_WAIT;
4626 /* update expire time */
4627 (*state)->expire = time_second;
4628 if (src->state >= TCPS_FIN_WAIT_2 &&
4629 dst->state >= TCPS_FIN_WAIT_2)
4630 (*state)->timeout = PFTM_TCP_CLOSED;
4631 else if (src->state >= TCPS_CLOSING &&
4632 dst->state >= TCPS_CLOSING)
4633 (*state)->timeout = PFTM_TCP_FIN_WAIT;
4634 else if (src->state < TCPS_ESTABLISHED ||
4635 dst->state < TCPS_ESTABLISHED)
4636 (*state)->timeout = PFTM_TCP_OPENING;
4637 else if (src->state >= TCPS_CLOSING ||
4638 dst->state >= TCPS_CLOSING)
4639 (*state)->timeout = PFTM_TCP_CLOSING;
4641 (*state)->timeout = PFTM_TCP_ESTABLISHED;
4643 /* Fall through to PASS packet */
4645 } else if ((dst->state < TCPS_SYN_SENT ||
4646 dst->state >= TCPS_FIN_WAIT_2 ||
4647 src->state >= TCPS_FIN_WAIT_2) &&
4648 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) &&
4649 /* Within a window forward of the originating packet */
4650 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) {
4651 /* Within a window backward of the originating packet */
4654 * This currently handles three situations:
4655 * 1) Stupid stacks will shotgun SYNs before their peer
4657 * 2) When PF catches an already established stream (the
4658 * firewall rebooted, the state table was flushed, routes
4660 * 3) Packets get funky immediately after the connection
4661 * closes (this should catch Solaris spurious ACK|FINs
4662 * that web servers like to spew after a close)
4664 * This must be a little more careful than the above code
4665 * since packet floods will also be caught here. We don't
4666 * update the TTL here to mitigate the damage of a packet
4667 * flood and so the same code can handle awkward establishment
4668 * and a loosened connection close.
4669 * In the establishment case, a correct peer response will
4670 * validate the connection, go through the normal state code
4671 * and keep updating the state TTL.
4674 if (pf_status.debug >= PF_DEBUG_MISC) {
4675 kprintf("pf: loose state match: ");
4676 pf_print_state(*state);
4677 pf_print_flags(th->th_flags);
4678 kprintf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
4679 "pkts=%llu:%llu\n", seq, orig_seq, ack, pd->p_len,
4680 ackskew, (unsigned long long)(*state)->packets[0],
4681 (unsigned long long)(*state)->packets[1]);
4684 if (dst->scrub || src->scrub) {
4685 if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4686 *state, src, dst, ©back))
4690 /* update max window */
4691 if (src->max_win < win)
4693 /* synchronize sequencing */
4694 if (SEQ_GT(end, src->seqlo))
4696 /* slide the window of what the other end can send */
4697 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4698 dst->seqhi = ack + MAX((win << sws), 1);
4701 * Cannot set dst->seqhi here since this could be a shotgunned
4702 * SYN and not an already established connection.
4705 if (th->th_flags & TH_FIN)
4706 if (src->state < TCPS_CLOSING)
4707 src->state = TCPS_CLOSING;
4708 if (th->th_flags & TH_RST)
4709 src->state = dst->state = TCPS_TIME_WAIT;
4711 /* Fall through to PASS packet */
4713 } else if ((*state)->pickup_mode == PF_PICKUPS_HASHONLY ||
4714 ((*state)->pickup_mode == PF_PICKUPS_ENABLED &&
4715 ((*state)->sync_flags & PFSTATE_GOT_SYN_MASK) !=
4716 PFSTATE_GOT_SYN_MASK)) {
4718 * If pickup mode is hash only, do not fail on sequence checks.
4720 * If pickup mode is enabled and we did not see the SYN in
4721 * both direction, do not fail on sequence checks because
4722 * we do not have complete information on window scale.
4724 * Adjust expiration and fall through to PASS packet.
4725 * XXX Add a FIN check to reduce timeout?
4727 (*state)->expire = time_second;
4730 * Failure processing
4732 if ((*state)->dst.state == TCPS_SYN_SENT &&
4733 (*state)->src.state == TCPS_SYN_SENT) {
4734 /* Send RST for state mismatches during handshake */
4735 if (!(th->th_flags & TH_RST))
4736 pf_send_tcp((*state)->rule.ptr, pd->af,
4737 pd->dst, pd->src, th->th_dport,
4738 th->th_sport, ntohl(th->th_ack), 0,
4740 (*state)->rule.ptr->return_ttl, 1, 0,
4741 pd->eh, kif->pfik_ifp);
4745 } else if (pf_status.debug >= PF_DEBUG_MISC) {
4746 kprintf("pf: BAD state: ");
4747 pf_print_state(*state);
4748 pf_print_flags(th->th_flags);
4749 kprintf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
4750 "pkts=%llu:%llu dir=%s,%s\n",
4751 seq, orig_seq, ack, pd->p_len, ackskew,
4752 (unsigned long long)(*state)->packets[0],
4753 (unsigned long long)(*state)->packets[1],
4754 direction == PF_IN ? "in" : "out",
4755 direction == (*state)->direction ? "fwd" : "rev");
4756 kprintf("pf: State failure on: %c %c %c %c | %c %c\n",
4757 SEQ_GEQ(src->seqhi, end) ? ' ' : '1',
4758 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ?
4760 (ackskew >= -MAXACKWINDOW) ? ' ' : '3',
4761 (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4',
4762 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5',
4763 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6');
4765 REASON_SET(reason, PFRES_BADSTATE);
4769 /* Any packets which have gotten here are to be passed */
4771 /* translate source/destination address, if necessary */
4772 if (STATE_TRANSLATE(*state)) {
4773 if (direction == PF_OUT) {
4774 pf_change_ap(pd->src, &th->th_sport, pd->ip_sum,
4775 &th->th_sum, &(*state)->gwy.addr,
4776 (*state)->gwy.port, 0, pd->af);
4779 * If we don't redispatch the packet will go into
4780 * the protocol stack on the wrong cpu for the
4781 * post-translated address.
4783 m->m_pkthdr.fw_flags |= FW_MBUF_REDISPATCH;
4784 pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum,
4785 &th->th_sum, &(*state)->lan.addr,
4786 (*state)->lan.port, 0, pd->af);
4788 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4789 } else if (copyback) {
4790 /* Copyback sequence modulation or stateful scrub changes */
4791 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4798 pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif,
4799 struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4801 struct pf_state_peer *src, *dst;
4802 struct pf_state_cmp key;
4803 struct udphdr *uh = pd->hdr.udp;
4806 key.proto = IPPROTO_UDP;
4807 if (direction == PF_IN) {
4808 PF_ACPY(&key.ext.addr, pd->src, key.af);
4809 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4810 key.ext.port = uh->uh_sport;
4811 key.gwy.port = uh->uh_dport;
4813 PF_ACPY(&key.lan.addr, pd->src, key.af);
4814 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4815 key.lan.port = uh->uh_sport;
4816 key.ext.port = uh->uh_dport;
4821 if (direction == (*state)->direction) {
4822 src = &(*state)->src;
4823 dst = &(*state)->dst;
4825 src = &(*state)->dst;
4826 dst = &(*state)->src;
4830 if (src->state < PFUDPS_SINGLE)
4831 src->state = PFUDPS_SINGLE;
4832 if (dst->state == PFUDPS_SINGLE)
4833 dst->state = PFUDPS_MULTIPLE;
4835 /* update expire time */
4836 (*state)->expire = time_second;
4837 if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE)
4838 (*state)->timeout = PFTM_UDP_MULTIPLE;
4840 (*state)->timeout = PFTM_UDP_SINGLE;
4842 /* translate source/destination address, if necessary */
4843 if (STATE_TRANSLATE(*state)) {
4844 if (direction == PF_OUT) {
4845 pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum,
4846 &uh->uh_sum, &(*state)->gwy.addr,
4847 (*state)->gwy.port, 1, pd->af);
4850 * If we don't redispatch the packet will go into
4851 * the protocol stack on the wrong cpu for the
4852 * post-translated address.
4854 m->m_pkthdr.fw_flags |= FW_MBUF_REDISPATCH;
4855 pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum,
4856 &uh->uh_sum, &(*state)->lan.addr,
4857 (*state)->lan.port, 1, pd->af);
4859 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
4866 pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif,
4867 struct mbuf *m, int off, void *h, struct pf_pdesc *pd, u_short *reason)
4869 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
4870 u_int16_t icmpid = 0, *icmpsum;
4873 struct pf_state_cmp key;
4875 switch (pd->proto) {
4878 icmptype = pd->hdr.icmp->icmp_type;
4879 icmpid = pd->hdr.icmp->icmp_id;
4880 icmpsum = &pd->hdr.icmp->icmp_cksum;
4882 if (icmptype == ICMP_UNREACH ||
4883 icmptype == ICMP_SOURCEQUENCH ||
4884 icmptype == ICMP_REDIRECT ||
4885 icmptype == ICMP_TIMXCEED ||
4886 icmptype == ICMP_PARAMPROB)
4891 case IPPROTO_ICMPV6:
4892 icmptype = pd->hdr.icmp6->icmp6_type;
4893 icmpid = pd->hdr.icmp6->icmp6_id;
4894 icmpsum = &pd->hdr.icmp6->icmp6_cksum;
4896 if (icmptype == ICMP6_DST_UNREACH ||
4897 icmptype == ICMP6_PACKET_TOO_BIG ||
4898 icmptype == ICMP6_TIME_EXCEEDED ||
4899 icmptype == ICMP6_PARAM_PROB)
4908 * ICMP query/reply message not related to a TCP/UDP packet.
4909 * Search for an ICMP state.
4912 key.proto = pd->proto;
4913 if (direction == PF_IN) {
4914 PF_ACPY(&key.ext.addr, pd->src, key.af);
4915 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4917 key.gwy.port = icmpid;
4919 PF_ACPY(&key.lan.addr, pd->src, key.af);
4920 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4921 key.lan.port = icmpid;
4927 (*state)->expire = time_second;
4928 (*state)->timeout = PFTM_ICMP_ERROR_REPLY;
4930 /* translate source/destination address, if necessary */
4931 if (STATE_TRANSLATE(*state)) {
4932 if (direction == PF_OUT) {
4936 pf_change_a(&saddr->v4.s_addr,
4938 (*state)->gwy.addr.v4.s_addr, 0);
4939 pd->hdr.icmp->icmp_cksum =
4941 pd->hdr.icmp->icmp_cksum, icmpid,
4942 (*state)->gwy.port, 0);
4943 pd->hdr.icmp->icmp_id =
4945 m_copyback(m, off, ICMP_MINLEN,
4946 (caddr_t)pd->hdr.icmp);
4952 &pd->hdr.icmp6->icmp6_cksum,
4953 &(*state)->gwy.addr, 0);
4955 sizeof(struct icmp6_hdr),
4956 (caddr_t)pd->hdr.icmp6);
4964 pf_change_a(&daddr->v4.s_addr,
4966 (*state)->lan.addr.v4.s_addr, 0);
4967 pd->hdr.icmp->icmp_cksum =
4969 pd->hdr.icmp->icmp_cksum, icmpid,
4970 (*state)->lan.port, 0);
4971 pd->hdr.icmp->icmp_id =
4973 m_copyback(m, off, ICMP_MINLEN,
4974 (caddr_t)pd->hdr.icmp);
4980 &pd->hdr.icmp6->icmp6_cksum,
4981 &(*state)->lan.addr, 0);
4983 sizeof(struct icmp6_hdr),
4984 (caddr_t)pd->hdr.icmp6);
4995 * ICMP error message in response to a TCP/UDP packet.
4996 * Extract the inner TCP/UDP header and search for that state.
4999 struct pf_pdesc pd2;
5004 struct ip6_hdr h2_6;
5014 /* offset of h2 in mbuf chain */
5015 ipoff2 = off + ICMP_MINLEN;
5017 if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2),
5018 NULL, reason, pd2.af)) {
5019 DPFPRINTF(PF_DEBUG_MISC,
5020 ("pf: ICMP error message too short "
5025 * ICMP error messages don't refer to non-first
5028 if (h2.ip_off & htons(IP_OFFMASK)) {
5029 REASON_SET(reason, PFRES_FRAG);
5033 /* offset of protocol header that follows h2 */
5034 off2 = ipoff2 + (h2.ip_hl << 2);
5036 pd2.proto = h2.ip_p;
5037 pd2.src = (struct pf_addr *)&h2.ip_src;
5038 pd2.dst = (struct pf_addr *)&h2.ip_dst;
5039 pd2.ip_sum = &h2.ip_sum;
5044 ipoff2 = off + sizeof(struct icmp6_hdr);
5046 if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6),
5047 NULL, reason, pd2.af)) {
5048 DPFPRINTF(PF_DEBUG_MISC,
5049 ("pf: ICMP error message too short "
5053 pd2.proto = h2_6.ip6_nxt;
5054 pd2.src = (struct pf_addr *)&h2_6.ip6_src;
5055 pd2.dst = (struct pf_addr *)&h2_6.ip6_dst;
5057 off2 = ipoff2 + sizeof(h2_6);
5059 switch (pd2.proto) {
5060 case IPPROTO_FRAGMENT:
5062 * ICMPv6 error messages for
5063 * non-first fragments
5065 REASON_SET(reason, PFRES_FRAG);
5068 case IPPROTO_HOPOPTS:
5069 case IPPROTO_ROUTING:
5070 case IPPROTO_DSTOPTS: {
5071 /* get next header and header length */
5072 struct ip6_ext opt6;
5074 if (!pf_pull_hdr(m, off2, &opt6,
5075 sizeof(opt6), NULL, reason,
5077 DPFPRINTF(PF_DEBUG_MISC,
5078 ("pf: ICMPv6 short opt\n"));
5081 if (pd2.proto == IPPROTO_AH)
5082 off2 += (opt6.ip6e_len + 2) * 4;
5084 off2 += (opt6.ip6e_len + 1) * 8;
5085 pd2.proto = opt6.ip6e_nxt;
5086 /* goto the next header */
5093 } while (!terminal);
5097 DPFPRINTF(PF_DEBUG_MISC,
5098 ("pf: ICMP AF %d unknown (ip6)\n", pd->af));
5103 switch (pd2.proto) {
5107 struct pf_state_cmp key;
5108 struct pf_state_peer *src, *dst;
5113 * Only the first 8 bytes of the TCP header can be
5114 * expected. Don't access any TCP header fields after
5115 * th_seq, an ackskew test is not possible.
5117 if (!pf_pull_hdr(m, off2, &th, 8, NULL, reason,
5119 DPFPRINTF(PF_DEBUG_MISC,
5120 ("pf: ICMP error message too short "
5126 key.proto = IPPROTO_TCP;
5127 if (direction == PF_IN) {
5128 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
5129 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
5130 key.ext.port = th.th_dport;
5131 key.gwy.port = th.th_sport;
5133 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
5134 PF_ACPY(&key.ext.addr, pd2.src, key.af);
5135 key.lan.port = th.th_dport;
5136 key.ext.port = th.th_sport;
5141 if (direction == (*state)->direction) {
5142 src = &(*state)->dst;
5143 dst = &(*state)->src;
5145 src = &(*state)->src;
5146 dst = &(*state)->dst;
5149 if (src->wscale && dst->wscale &&
5150 !(th.th_flags & TH_SYN))
5151 dws = dst->wscale & PF_WSCALE_MASK;
5155 /* Demodulate sequence number */
5156 seq = ntohl(th.th_seq) - src->seqdiff;
5158 pf_change_a(&th.th_seq, icmpsum,
5163 if (!SEQ_GEQ(src->seqhi, seq) ||
5164 !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws))) {
5165 if (pf_status.debug >= PF_DEBUG_MISC) {
5166 kprintf("pf: BAD ICMP %d:%d ",
5167 icmptype, pd->hdr.icmp->icmp_code);
5168 pf_print_host(pd->src, 0, pd->af);
5170 pf_print_host(pd->dst, 0, pd->af);
5171 kprintf(" state: ");
5172 pf_print_state(*state);
5173 kprintf(" seq=%u\n", seq);
5175 REASON_SET(reason, PFRES_BADSTATE);
5179 if (STATE_TRANSLATE(*state)) {
5180 if (direction == PF_IN) {
5181 pf_change_icmp(pd2.src, &th.th_sport,
5182 daddr, &(*state)->lan.addr,
5183 (*state)->lan.port, NULL,
5184 pd2.ip_sum, icmpsum,
5185 pd->ip_sum, 0, pd2.af);
5187 pf_change_icmp(pd2.dst, &th.th_dport,
5188 saddr, &(*state)->gwy.addr,
5189 (*state)->gwy.port, NULL,
5190 pd2.ip_sum, icmpsum,
5191 pd->ip_sum, 0, pd2.af);
5200 m_copyback(m, off, ICMP_MINLEN,
5201 (caddr_t)pd->hdr.icmp);
5202 m_copyback(m, ipoff2, sizeof(h2),
5209 sizeof(struct icmp6_hdr),
5210 (caddr_t)pd->hdr.icmp6);
5211 m_copyback(m, ipoff2, sizeof(h2_6),
5216 m_copyback(m, off2, 8, (caddr_t)&th);
5225 if (!pf_pull_hdr(m, off2, &uh, sizeof(uh),
5226 NULL, reason, pd2.af)) {
5227 DPFPRINTF(PF_DEBUG_MISC,
5228 ("pf: ICMP error message too short "
5234 key.proto = IPPROTO_UDP;
5235 if (direction == PF_IN) {
5236 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
5237 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
5238 key.ext.port = uh.uh_dport;
5239 key.gwy.port = uh.uh_sport;
5241 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
5242 PF_ACPY(&key.ext.addr, pd2.src, key.af);
5243 key.lan.port = uh.uh_dport;
5244 key.ext.port = uh.uh_sport;
5249 if (STATE_TRANSLATE(*state)) {
5250 if (direction == PF_IN) {
5251 pf_change_icmp(pd2.src, &uh.uh_sport,
5252 daddr, &(*state)->lan.addr,
5253 (*state)->lan.port, &uh.uh_sum,
5254 pd2.ip_sum, icmpsum,
5255 pd->ip_sum, 1, pd2.af);
5257 pf_change_icmp(pd2.dst, &uh.uh_dport,
5258 saddr, &(*state)->gwy.addr,
5259 (*state)->gwy.port, &uh.uh_sum,
5260 pd2.ip_sum, icmpsum,
5261 pd->ip_sum, 1, pd2.af);
5266 m_copyback(m, off, ICMP_MINLEN,
5267 (caddr_t)pd->hdr.icmp);
5268 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
5274 sizeof(struct icmp6_hdr),
5275 (caddr_t)pd->hdr.icmp6);
5276 m_copyback(m, ipoff2, sizeof(h2_6),
5281 m_copyback(m, off2, sizeof(uh), (caddr_t)&uh);
5288 case IPPROTO_ICMP: {
5291 if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN,
5292 NULL, reason, pd2.af)) {
5293 DPFPRINTF(PF_DEBUG_MISC,
5294 ("pf: ICMP error message too short i"
5300 key.proto = IPPROTO_ICMP;
5301 if (direction == PF_IN) {
5302 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
5303 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
5305 key.gwy.port = iih.icmp_id;
5307 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
5308 PF_ACPY(&key.ext.addr, pd2.src, key.af);
5309 key.lan.port = iih.icmp_id;
5315 if (STATE_TRANSLATE(*state)) {
5316 if (direction == PF_IN) {
5317 pf_change_icmp(pd2.src, &iih.icmp_id,
5318 daddr, &(*state)->lan.addr,
5319 (*state)->lan.port, NULL,
5320 pd2.ip_sum, icmpsum,
5321 pd->ip_sum, 0, AF_INET);
5323 pf_change_icmp(pd2.dst, &iih.icmp_id,
5324 saddr, &(*state)->gwy.addr,
5325 (*state)->gwy.port, NULL,
5326 pd2.ip_sum, icmpsum,
5327 pd->ip_sum, 0, AF_INET);
5329 m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
5330 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
5331 m_copyback(m, off2, ICMP_MINLEN, (caddr_t)&iih);
5339 case IPPROTO_ICMPV6: {
5340 struct icmp6_hdr iih;
5342 if (!pf_pull_hdr(m, off2, &iih,
5343 sizeof(struct icmp6_hdr), NULL, reason, pd2.af)) {
5344 DPFPRINTF(PF_DEBUG_MISC,
5345 ("pf: ICMP error message too short "
5351 key.proto = IPPROTO_ICMPV6;
5352 if (direction == PF_IN) {
5353 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
5354 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
5356 key.gwy.port = iih.icmp6_id;
5358 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
5359 PF_ACPY(&key.ext.addr, pd2.src, key.af);
5360 key.lan.port = iih.icmp6_id;
5366 if (STATE_TRANSLATE(*state)) {
5367 if (direction == PF_IN) {
5368 pf_change_icmp(pd2.src, &iih.icmp6_id,
5369 daddr, &(*state)->lan.addr,
5370 (*state)->lan.port, NULL,
5371 pd2.ip_sum, icmpsum,
5372 pd->ip_sum, 0, AF_INET6);
5374 pf_change_icmp(pd2.dst, &iih.icmp6_id,
5375 saddr, &(*state)->gwy.addr,
5376 (*state)->gwy.port, NULL,
5377 pd2.ip_sum, icmpsum,
5378 pd->ip_sum, 0, AF_INET6);
5380 m_copyback(m, off, sizeof(struct icmp6_hdr),
5381 (caddr_t)pd->hdr.icmp6);
5382 m_copyback(m, ipoff2, sizeof(h2_6), (caddr_t)&h2_6);
5383 m_copyback(m, off2, sizeof(struct icmp6_hdr),
5393 key.proto = pd2.proto;
5394 if (direction == PF_IN) {
5395 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
5396 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
5400 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
5401 PF_ACPY(&key.ext.addr, pd2.src, key.af);
5408 if (STATE_TRANSLATE(*state)) {
5409 if (direction == PF_IN) {
5410 pf_change_icmp(pd2.src, NULL,
5411 daddr, &(*state)->lan.addr,
5413 pd2.ip_sum, icmpsum,
5414 pd->ip_sum, 0, pd2.af);
5416 pf_change_icmp(pd2.dst, NULL,
5417 saddr, &(*state)->gwy.addr,
5419 pd2.ip_sum, icmpsum,
5420 pd->ip_sum, 0, pd2.af);
5425 m_copyback(m, off, ICMP_MINLEN,
5426 (caddr_t)pd->hdr.icmp);
5427 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
5433 sizeof(struct icmp6_hdr),
5434 (caddr_t)pd->hdr.icmp6);
5435 m_copyback(m, ipoff2, sizeof(h2_6),
5450 pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif,
5451 struct pf_pdesc *pd)
5453 struct pf_state_peer *src, *dst;
5454 struct pf_state_cmp key;
5457 key.proto = pd->proto;
5458 if (direction == PF_IN) {
5459 PF_ACPY(&key.ext.addr, pd->src, key.af);
5460 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
5464 PF_ACPY(&key.lan.addr, pd->src, key.af);
5465 PF_ACPY(&key.ext.addr, pd->dst, key.af);
5472 if (direction == (*state)->direction) {
5473 src = &(*state)->src;
5474 dst = &(*state)->dst;
5476 src = &(*state)->dst;
5477 dst = &(*state)->src;
5481 if (src->state < PFOTHERS_SINGLE)
5482 src->state = PFOTHERS_SINGLE;
5483 if (dst->state == PFOTHERS_SINGLE)
5484 dst->state = PFOTHERS_MULTIPLE;
5486 /* update expire time */
5487 (*state)->expire = time_second;
5488 if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE)
5489 (*state)->timeout = PFTM_OTHER_MULTIPLE;
5491 (*state)->timeout = PFTM_OTHER_SINGLE;
5493 /* translate source/destination address, if necessary */
5494 if (STATE_TRANSLATE(*state)) {
5495 if (direction == PF_OUT)
5499 pf_change_a(&pd->src->v4.s_addr,
5500 pd->ip_sum, (*state)->gwy.addr.v4.s_addr,
5506 PF_ACPY(pd->src, &(*state)->gwy.addr, pd->af);
5514 pf_change_a(&pd->dst->v4.s_addr,
5515 pd->ip_sum, (*state)->lan.addr.v4.s_addr,
5521 PF_ACPY(pd->dst, &(*state)->lan.addr, pd->af);
5531 * ipoff and off are measured from the start of the mbuf chain.
5532 * h must be at "ipoff" on the mbuf chain.
5535 pf_pull_hdr(struct mbuf *m, int off, void *p, int len,
5536 u_short *actionp, u_short *reasonp, sa_family_t af)
5541 struct ip *h = mtod(m, struct ip *);
5542 u_int16_t fragoff = (h->ip_off & IP_OFFMASK) << 3;
5546 ACTION_SET(actionp, PF_PASS);
5548 ACTION_SET(actionp, PF_DROP);
5549 REASON_SET(reasonp, PFRES_FRAG);
5553 if (m->m_pkthdr.len < off + len ||
5554 h->ip_len < off + len) {
5555 ACTION_SET(actionp, PF_DROP);
5556 REASON_SET(reasonp, PFRES_SHORT);
5564 struct ip6_hdr *h = mtod(m, struct ip6_hdr *);
5566 if (m->m_pkthdr.len < off + len ||
5567 (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) <
5568 (unsigned)(off + len)) {
5569 ACTION_SET(actionp, PF_DROP);
5570 REASON_SET(reasonp, PFRES_SHORT);
5577 m_copydata(m, off, len, p);
5582 pf_routable(struct pf_addr *addr, sa_family_t af, struct pfi_kif *kif)
5584 struct sockaddr_in *dst;
5588 struct sockaddr_in6 *dst6;
5589 struct route_in6 ro;
5593 struct radix_node *rn;
5598 bzero(&ro, sizeof(ro));
5601 dst = satosin(&ro.ro_dst);
5602 dst->sin_family = AF_INET;
5603 dst->sin_len = sizeof(*dst);
5604 dst->sin_addr = addr->v4;
5608 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
5609 dst6->sin6_family = AF_INET6;
5610 dst6->sin6_len = sizeof(*dst6);
5611 dst6->sin6_addr = addr->v6;
5618 /* Skip checks for ipsec interfaces */
5619 if (kif != NULL && kif->pfik_ifp->if_type == IFT_ENC)
5622 rtalloc_ign((struct route *)&ro, 0);
5624 if (ro.ro_rt != NULL) {
5625 /* No interface given, this is a no-route check */
5629 if (kif->pfik_ifp == NULL) {
5634 /* Perform uRPF check if passed input interface */
5636 rn = (struct radix_node *)ro.ro_rt;
5638 rt = (struct rtentry *)rn;
5641 if (kif->pfik_ifp == ifp)
5644 } while (check_mpath == 1 && rn != NULL && ret == 0);
5648 if (ro.ro_rt != NULL)
5654 pf_rtlabel_match(struct pf_addr *addr, sa_family_t af, struct pf_addr_wrap *aw)
5656 struct sockaddr_in *dst;
5658 struct sockaddr_in6 *dst6;
5659 struct route_in6 ro;
5665 bzero(&ro, sizeof(ro));
5668 dst = satosin(&ro.ro_dst);
5669 dst->sin_family = AF_INET;
5670 dst->sin_len = sizeof(*dst);
5671 dst->sin_addr = addr->v4;
5675 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
5676 dst6->sin6_family = AF_INET6;
5677 dst6->sin6_len = sizeof(*dst6);
5678 dst6->sin6_addr = addr->v6;
5685 rtalloc_ign((struct route *)&ro, (RTF_CLONING | RTF_PRCLONING));
5687 if (ro.ro_rt != NULL) {
5696 pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5697 struct pf_state *s, struct pf_pdesc *pd)
5699 struct mbuf *m0, *m1;
5700 struct route iproute;
5701 struct route *ro = NULL;
5702 struct sockaddr_in *dst;
5704 struct ifnet *ifp = NULL;
5705 struct pf_addr naddr;
5706 struct pf_src_node *sn = NULL;
5713 if (m == NULL || *m == NULL || r == NULL ||
5714 (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5715 panic("pf_route: invalid parameters");
5717 if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
5718 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
5719 (*m)->m_pkthdr.pf_routed = 1;
5721 if ((*m)->m_pkthdr.pf_routed > 3) {
5726 (*m)->m_pkthdr.pf_routed++;
5729 if (r->rt == PF_DUPTO) {
5730 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL)
5733 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5738 if (m0->m_len < sizeof(struct ip)) {
5739 DPFPRINTF(PF_DEBUG_URGENT,
5740 ("pf_route: m0->m_len < sizeof(struct ip)\n"));
5744 ip = mtod(m0, struct ip *);
5747 bzero((caddr_t)ro, sizeof(*ro));
5748 dst = satosin(&ro->ro_dst);
5749 dst->sin_family = AF_INET;
5750 dst->sin_len = sizeof(*dst);
5751 dst->sin_addr = ip->ip_dst;
5753 if (r->rt == PF_FASTROUTE) {
5755 if (ro->ro_rt == 0) {
5756 ipstat.ips_noroute++;
5760 ifp = ro->ro_rt->rt_ifp;
5761 ro->ro_rt->rt_use++;
5763 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
5764 dst = satosin(ro->ro_rt->rt_gateway);
5766 if (TAILQ_EMPTY(&r->rpool.list)) {
5767 DPFPRINTF(PF_DEBUG_URGENT,
5768 ("pf_route: TAILQ_EMPTY(&r->rpool.list)\n"));
5772 pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src,
5774 if (!PF_AZERO(&naddr, AF_INET))
5775 dst->sin_addr.s_addr = naddr.v4.s_addr;
5776 ifp = r->rpool.cur->kif ?
5777 r->rpool.cur->kif->pfik_ifp : NULL;
5779 if (!PF_AZERO(&s->rt_addr, AF_INET))
5780 dst->sin_addr.s_addr =
5781 s->rt_addr.v4.s_addr;
5782 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5790 if (pf_test(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) {
5793 } else if (m0 == NULL) {
5798 if (m0->m_len < sizeof(struct ip)) {
5799 DPFPRINTF(PF_DEBUG_URGENT,
5800 ("pf_route: m0->m_len < sizeof(struct ip)\n"));
5803 ip = mtod(m0, struct ip *);
5806 /* Copied from FreeBSD 5.1-CURRENT ip_output. */
5807 m0->m_pkthdr.csum_flags |= CSUM_IP;
5808 sw_csum = m0->m_pkthdr.csum_flags & ~ifp->if_hwassist;
5809 if (sw_csum & CSUM_DELAY_DATA) {
5811 * XXX: in_delayed_cksum assumes HBO for ip->ip_len (at least)
5814 NTOHS(ip->ip_off); /* XXX: needed? */
5815 in_delayed_cksum(m0);
5818 sw_csum &= ~CSUM_DELAY_DATA;
5820 m0->m_pkthdr.csum_flags &= ifp->if_hwassist;
5822 if (ntohs(ip->ip_len) <= ifp->if_mtu ||
5823 (ifp->if_hwassist & CSUM_FRAGMENT &&
5824 ((ip->ip_off & htons(IP_DF)) == 0))) {
5826 * ip->ip_len = htons(ip->ip_len);
5827 * ip->ip_off = htons(ip->ip_off);
5830 if (sw_csum & CSUM_DELAY_IP) {
5832 if (ip->ip_v == IPVERSION &&
5833 (ip->ip_hl << 2) == sizeof(*ip)) {
5834 ip->ip_sum = in_cksum_hdr(ip);
5836 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
5840 error = (*ifp->if_output)(ifp, m0, sintosa(dst), ro->ro_rt);
5846 * Too large for interface; fragment if possible.
5847 * Must be able to put at least 8 bytes per fragment.
5849 if (ip->ip_off & htons(IP_DF)) {
5850 ipstat.ips_cantfrag++;
5851 if (r->rt != PF_DUPTO) {
5852 /* icmp_error() expects host byte ordering */
5856 icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0,
5866 * XXX: is cheaper + less error prone than own function
5870 error = ip_fragment(ip, &m0, ifp->if_mtu, ifp->if_hwassist, sw_csum);
5875 for (m0 = m1; m0; m0 = m1) {
5880 error = (*ifp->if_output)(ifp, m0, sintosa(dst),
5888 ipstat.ips_fragmented++;
5891 if (r->rt != PF_DUPTO)
5893 if (ro == &iproute && ro->ro_rt)
5905 pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5906 struct pf_state *s, struct pf_pdesc *pd)
5909 struct route_in6 ip6route;
5910 struct route_in6 *ro;
5911 struct sockaddr_in6 *dst;
5912 struct ip6_hdr *ip6;
5913 struct ifnet *ifp = NULL;
5914 struct pf_addr naddr;
5915 struct pf_src_node *sn = NULL;
5918 if (m == NULL || *m == NULL || r == NULL ||
5919 (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5920 panic("pf_route6: invalid parameters");
5922 if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
5923 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
5924 (*m)->m_pkthdr.pf_routed = 1;
5926 if ((*m)->m_pkthdr.pf_routed > 3) {
5931 (*m)->m_pkthdr.pf_routed++;
5934 if (r->rt == PF_DUPTO) {
5935 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL)
5938 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5943 if (m0->m_len < sizeof(struct ip6_hdr)) {
5944 DPFPRINTF(PF_DEBUG_URGENT,
5945 ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n"));
5948 ip6 = mtod(m0, struct ip6_hdr *);
5951 bzero((caddr_t)ro, sizeof(*ro));
5952 dst = (struct sockaddr_in6 *)&ro->ro_dst;
5953 dst->sin6_family = AF_INET6;
5954 dst->sin6_len = sizeof(*dst);
5955 dst->sin6_addr = ip6->ip6_dst;
5957 /* Cheat. XXX why only in the v6 case??? */
5958 if (r->rt == PF_FASTROUTE) {
5959 pd->pf_mtag->flags |= PF_TAG_GENERATED;
5961 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
5966 if (TAILQ_EMPTY(&r->rpool.list)) {
5967 DPFPRINTF(PF_DEBUG_URGENT,
5968 ("pf_route6: TAILQ_EMPTY(&r->rpool.list)\n"));
5972 pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src,
5974 if (!PF_AZERO(&naddr, AF_INET6))
5975 PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5977 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL;
5979 if (!PF_AZERO(&s->rt_addr, AF_INET6))
5980 PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5981 &s->rt_addr, AF_INET6);
5982 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5989 if (pf_test6(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) {
5992 } else if (m0 == NULL) {
5997 if (m0->m_len < sizeof(struct ip6_hdr)) {
5998 DPFPRINTF(PF_DEBUG_URGENT,
5999 ("pf_route6: m0->m_len < sizeof(struct ip6_hdr)\n"));
6002 ip6 = mtod(m0, struct ip6_hdr *);
6006 * If the packet is too large for the outgoing interface,
6007 * send back an icmp6 error.
6009 if (IN6_IS_ADDR_LINKLOCAL(&dst->sin6_addr))
6010 dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
6011 if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) {
6013 error = nd6_output(ifp, ifp, m0, dst, NULL);
6016 in6_ifstat_inc(ifp, ifs6_in_toobig);
6017 if (r->rt != PF_DUPTO) {
6019 icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
6026 if (r->rt != PF_DUPTO)
6039 * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag
6040 * off is the offset where the protocol header starts
6041 * len is the total length of protocol header plus payload
6042 * returns 0 when the checksum is valid, otherwise returns 1.
6046 * FreeBSD supports cksum offload for the following drivers.
6047 * em(4), gx(4), lge(4), nge(4), ti(4), xl(4)
6048 * If we can make full use of it we would outperform ipfw/ipfilter in
6049 * very heavy traffic.
6050 * I have not tested 'cause I don't have NICs that supports cksum offload.
6051 * (There might be problems. Typical phenomena would be
6052 * 1. No route message for UDP packet.
6053 * 2. No connection acceptance from external hosts regardless of rule set.)
6056 pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p,
6063 if (off < sizeof(struct ip) || len < sizeof(struct udphdr))
6065 if (m->m_pkthdr.len < off + len)
6071 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
6072 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) {
6073 sum = m->m_pkthdr.csum_data;
6075 ip = mtod(m, struct ip *);
6076 sum = in_pseudo(ip->ip_src.s_addr,
6077 ip->ip_dst.s_addr, htonl((u_short)len +
6078 m->m_pkthdr.csum_data + p));
6086 case IPPROTO_ICMPV6:
6096 if (p == IPPROTO_ICMP) {
6101 sum = in_cksum(m, len);
6105 if (m->m_len < sizeof(struct ip))
6107 sum = in_cksum_range(m, p, off, len);
6109 m->m_pkthdr.csum_flags |=
6112 m->m_pkthdr.csum_data = 0xffff;
6118 if (m->m_len < sizeof(struct ip6_hdr))
6120 sum = in6_cksum(m, p, off, len);
6123 * IPv6 H/W cksum off-load not supported yet!
6126 * m->m_pkthdr.csum_flags |=
6127 * (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
6128 * m->m_pkthdr.csum_data = 0xffff;
6140 tcpstat.tcps_rcvbadsum++;
6143 udpstat.udps_badsum++;
6146 icmpstat.icps_checksum++;
6149 case IPPROTO_ICMPV6:
6150 icmp6stat.icp6s_checksum++;
6161 pf_test(int dir, struct ifnet *ifp, struct mbuf **m0,
6162 struct ether_header *eh, struct inpcb *inp)
6164 struct pfi_kif *kif;
6165 u_short action, reason = 0, log = 0;
6166 struct mbuf *m = *m0;
6167 struct ip *h = NULL;
6168 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr;
6169 struct pf_state *s = NULL;
6170 struct pf_ruleset *ruleset = NULL;
6172 int off, dirndx, pqid = 0;
6174 if (!pf_status.running)
6177 memset(&pd, 0, sizeof(pd));
6178 if ((pd.pf_mtag = pf_get_mtag(m)) == NULL) {
6179 DPFPRINTF(PF_DEBUG_URGENT,
6180 ("pf_test: pf_get_mtag returned NULL\n"));
6183 if (pd.pf_mtag->flags & PF_TAG_GENERATED)
6185 kif = (struct pfi_kif *)ifp->if_pf_kif;
6187 DPFPRINTF(PF_DEBUG_URGENT,
6188 ("pf_test: kif == NULL, if_xname %s\n", ifp->if_xname));
6191 if (kif->pfik_flags & PFI_IFLAG_SKIP)
6195 if ((m->m_flags & M_PKTHDR) == 0)
6196 panic("non-M_PKTHDR is passed to pf_test");
6197 #endif /* DIAGNOSTIC */
6199 if (m->m_pkthdr.len < (int)sizeof(*h)) {
6201 REASON_SET(&reason, PFRES_SHORT);
6207 /* We do IP header normalization and packet reassembly here */
6208 if (pf_normalize_ip(m0, dir, kif, &reason, &pd) != PF_PASS) {
6213 h = mtod(m, struct ip *);
6215 off = h->ip_hl << 2;
6216 if (off < (int)sizeof(*h)) {
6218 REASON_SET(&reason, PFRES_SHORT);
6224 pd.src = (struct pf_addr *)&h->ip_src;
6225 pd.dst = (struct pf_addr *)&h->ip_dst;
6226 PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET);
6227 pd.ip_sum = &h->ip_sum;
6231 pd.tot_len = h->ip_len;
6234 /* handle fragments that didn't get reassembled by normalization */
6235 if (h->ip_off & (IP_MF | IP_OFFMASK)) {
6236 action = pf_test_fragment(&r, dir, kif, m, h,
6246 if (!pf_pull_hdr(m, off, &th, sizeof(th),
6247 &action, &reason, AF_INET)) {
6248 log = action != PF_PASS;
6251 if (dir == PF_IN && pf_check_proto_cksum(m, off,
6252 h->ip_len - off, IPPROTO_TCP, AF_INET)) {
6253 REASON_SET(&reason, PFRES_PROTCKSUM);
6257 pd.p_len = pd.tot_len - off - (th.th_off << 2);
6258 if ((th.th_flags & TH_ACK) && pd.p_len == 0)
6260 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
6261 if (action == PF_DROP)
6263 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
6265 if (action == PF_PASS) {
6267 pfsync_update_state(s);
6268 #endif /* NPFSYNC */
6272 } else if (s == NULL) {
6273 action = pf_test_tcp(&r, &s, dir, kif,
6274 m, off, h, &pd, &a, &ruleset, NULL, inp);
6283 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
6284 &action, &reason, AF_INET)) {
6285 log = action != PF_PASS;
6288 if (dir == PF_IN && uh.uh_sum && pf_check_proto_cksum(m,
6289 off, h->ip_len - off, IPPROTO_UDP, AF_INET)) {
6291 REASON_SET(&reason, PFRES_PROTCKSUM);
6294 if (uh.uh_dport == 0 ||
6295 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
6296 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
6298 REASON_SET(&reason, PFRES_SHORT);
6301 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
6302 if (action == PF_PASS) {
6304 pfsync_update_state(s);
6305 #endif /* NPFSYNC */
6309 } else if (s == NULL)
6310 action = pf_test_udp(&r, &s, dir, kif,
6311 m, off, h, &pd, &a, &ruleset, NULL, inp);
6315 case IPPROTO_ICMP: {
6319 if (!pf_pull_hdr(m, off, &ih, ICMP_MINLEN,
6320 &action, &reason, AF_INET)) {
6321 log = action != PF_PASS;
6324 if (dir == PF_IN && pf_check_proto_cksum(m, off,
6325 h->ip_len - off, IPPROTO_ICMP, AF_INET)) {
6327 REASON_SET(&reason, PFRES_PROTCKSUM);
6330 action = pf_test_state_icmp(&s, dir, kif, m, off, h, &pd,
6332 if (action == PF_PASS) {
6334 pfsync_update_state(s);
6335 #endif /* NPFSYNC */
6339 } else if (s == NULL)
6340 action = pf_test_icmp(&r, &s, dir, kif,
6341 m, off, h, &pd, &a, &ruleset, NULL);
6346 action = pf_test_state_other(&s, dir, kif, &pd);
6347 if (action == PF_PASS) {
6349 pfsync_update_state(s);
6350 #endif /* NPFSYNC */
6354 } else if (s == NULL)
6355 action = pf_test_other(&r, &s, dir, kif, m, off, h,
6356 &pd, &a, &ruleset, NULL);
6361 if (action == PF_PASS && h->ip_hl > 5 &&
6362 !((s && s->allow_opts) || r->allow_opts)) {
6364 REASON_SET(&reason, PFRES_IPOPTIONS);
6366 DPFPRINTF(PF_DEBUG_MISC,
6367 ("pf: dropping packet with ip options\n"));
6370 if ((s && s->tag) || r->rtableid)
6371 pf_tag_packet(m, pd.pf_mtag, s ? s->tag : 0, r->rtableid);
6374 if (action == PF_PASS && r->qid) {
6375 if (pqid || (pd.tos & IPTOS_LOWDELAY))
6376 pd.pf_mtag->qid = r->pqid;
6378 pd.pf_mtag->qid = r->qid;
6379 /* add hints for ecn */
6380 pd.pf_mtag->af = AF_INET;
6381 pd.pf_mtag->hdr = h;
6382 /* add connection hash for fairq */
6384 KKASSERT(s->hash != 0);
6385 pd.pf_mtag->flags |= PF_TAG_STATE_HASHED;
6386 pd.pf_mtag->state_hash = s->hash;
6392 * connections redirected to loopback should not match sockets
6393 * bound specifically to loopback due to security implications,
6394 * see tcp_input() and in_pcblookup_listen().
6396 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
6397 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
6398 (s->nat_rule.ptr->action == PF_RDR ||
6399 s->nat_rule.ptr->action == PF_BINAT) &&
6400 (ntohl(pd.dst->v4.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET)
6401 pd.pf_mtag->flags |= PF_TAG_TRANSLATE_LOCALHOST;
6406 if (s != NULL && s->nat_rule.ptr != NULL &&
6407 s->nat_rule.ptr->log & PF_LOG_ALL)
6408 lr = s->nat_rule.ptr;
6411 PFLOG_PACKET(kif, h, m, AF_INET, dir, reason, lr, a, ruleset,
6415 kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
6416 kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++;
6418 if (action == PF_PASS || r->action == PF_DROP) {
6419 dirndx = (dir == PF_OUT);
6420 r->packets[dirndx]++;
6421 r->bytes[dirndx] += pd.tot_len;
6423 a->packets[dirndx]++;
6424 a->bytes[dirndx] += pd.tot_len;
6427 if (s->nat_rule.ptr != NULL) {
6428 s->nat_rule.ptr->packets[dirndx]++;
6429 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len;
6431 if (s->src_node != NULL) {
6432 s->src_node->packets[dirndx]++;
6433 s->src_node->bytes[dirndx] += pd.tot_len;
6435 if (s->nat_src_node != NULL) {
6436 s->nat_src_node->packets[dirndx]++;
6437 s->nat_src_node->bytes[dirndx] += pd.tot_len;
6439 dirndx = (dir == s->direction) ? 0 : 1;
6440 s->packets[dirndx]++;
6441 s->bytes[dirndx] += pd.tot_len;
6444 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
6448 * XXX: we need to make sure that the addresses
6449 * passed to pfr_update_stats() are the same than
6450 * the addresses used during matching (pfr_match)
6452 if (r == &pf_default_rule) {
6454 x = (s == NULL || s->direction == dir) ?
6455 &pd.baddr : &pd.naddr;
6457 x = (s == NULL || s->direction == dir) ?
6458 &pd.naddr : &pd.baddr;
6459 if (x == &pd.baddr || s == NULL) {
6460 /* we need to change the address */
6467 if (tr->src.addr.type == PF_ADDR_TABLE)
6468 pfr_update_stats(tr->src.addr.p.tbl, (s == NULL ||
6469 s->direction == dir) ? pd.src : pd.dst, pd.af,
6470 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
6472 if (tr->dst.addr.type == PF_ADDR_TABLE)
6473 pfr_update_stats(tr->dst.addr.p.tbl, (s == NULL ||
6474 s->direction == dir) ? pd.dst : pd.src, pd.af,
6475 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
6480 if (action == PF_SYNPROXY_DROP) {
6485 /* pf_route can free the mbuf causing *m0 to become NULL */
6486 pf_route(m0, r, dir, ifp, s, &pd);
6493 pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0,
6494 struct ether_header *eh, struct inpcb *inp)
6496 struct pfi_kif *kif;
6497 u_short action, reason = 0, log = 0;
6498 struct mbuf *m = *m0, *n = NULL;
6499 struct ip6_hdr *h = NULL;
6500 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr;
6501 struct pf_state *s = NULL;
6502 struct pf_ruleset *ruleset = NULL;
6504 int off, terminal = 0, dirndx;
6506 if (!pf_status.running)
6509 memset(&pd, 0, sizeof(pd));
6510 if ((pd.pf_mtag = pf_get_mtag(m)) == NULL) {
6511 DPFPRINTF(PF_DEBUG_URGENT,
6512 ("pf_test6: pf_get_mtag returned NULL\n"));
6515 if (pd.pf_mtag->flags & PF_TAG_GENERATED)
6518 kif = (struct pfi_kif *)ifp->if_pf_kif;
6520 DPFPRINTF(PF_DEBUG_URGENT,
6521 ("pf_test6: kif == NULL, if_xname %s\n", ifp->if_xname));
6524 if (kif->pfik_flags & PFI_IFLAG_SKIP)
6528 if ((m->m_flags & M_PKTHDR) == 0)
6529 panic("non-M_PKTHDR is passed to pf_test6");
6530 #endif /* DIAGNOSTIC */
6532 if (m->m_pkthdr.len < (int)sizeof(*h)) {
6534 REASON_SET(&reason, PFRES_SHORT);
6539 /* We do IP header normalization and packet reassembly here */
6540 if (pf_normalize_ip6(m0, dir, kif, &reason, &pd) != PF_PASS) {
6545 h = mtod(m, struct ip6_hdr *);
6549 * we do not support jumbogram yet. if we keep going, zero ip6_plen
6550 * will do something bad, so drop the packet for now.
6552 if (htons(h->ip6_plen) == 0) {
6554 REASON_SET(&reason, PFRES_NORM); /*XXX*/
6559 pd.src = (struct pf_addr *)&h->ip6_src;
6560 pd.dst = (struct pf_addr *)&h->ip6_dst;
6561 PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET6);
6565 pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr);
6568 off = ((caddr_t)h - m->m_data) + sizeof(struct ip6_hdr);
6569 pd.proto = h->ip6_nxt;
6572 case IPPROTO_FRAGMENT:
6573 action = pf_test_fragment(&r, dir, kif, m, h,
6575 if (action == PF_DROP)
6576 REASON_SET(&reason, PFRES_FRAG);
6578 case IPPROTO_ROUTING: {
6579 struct ip6_rthdr rthdr;
6581 if (!pf_pull_hdr(m, off, &rthdr, sizeof(rthdr), NULL,
6583 DPFPRINTF(PF_DEBUG_MISC,
6584 ("pf: IPv6 short rthdr\n"));
6589 if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) {
6590 DPFPRINTF(PF_DEBUG_MISC,
6591 ("pf: IPv6 rthdr0\n"));
6593 REASON_SET(&reason, PFRES_IPOPTIONS);
6600 case IPPROTO_HOPOPTS:
6601 case IPPROTO_DSTOPTS: {
6602 /* get next header and header length */
6603 struct ip6_ext opt6;
6605 if (!pf_pull_hdr(m, off, &opt6, sizeof(opt6),
6606 NULL, &reason, pd.af)) {
6607 DPFPRINTF(PF_DEBUG_MISC,
6608 ("pf: IPv6 short opt\n"));
6613 if (pd.proto == IPPROTO_AH)
6614 off += (opt6.ip6e_len + 2) * 4;
6616 off += (opt6.ip6e_len + 1) * 8;
6617 pd.proto = opt6.ip6e_nxt;
6618 /* goto the next header */
6625 } while (!terminal);
6627 /* if there's no routing header, use unmodified mbuf for checksumming */
6637 if (!pf_pull_hdr(m, off, &th, sizeof(th),
6638 &action, &reason, AF_INET6)) {
6639 log = action != PF_PASS;
6642 if (dir == PF_IN && pf_check_proto_cksum(n, off,
6643 ntohs(h->ip6_plen) - (off - sizeof(struct ip6_hdr)),
6644 IPPROTO_TCP, AF_INET6)) {
6646 REASON_SET(&reason, PFRES_PROTCKSUM);
6649 pd.p_len = pd.tot_len - off - (th.th_off << 2);
6650 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
6651 if (action == PF_DROP)
6653 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
6655 if (action == PF_PASS) {
6657 pfsync_update_state(s);
6658 #endif /* NPFSYNC */
6662 } else if (s == NULL)
6663 action = pf_test_tcp(&r, &s, dir, kif,
6664 m, off, h, &pd, &a, &ruleset, NULL, inp);
6672 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
6673 &action, &reason, AF_INET6)) {
6674 log = action != PF_PASS;
6677 if (dir == PF_IN && uh.uh_sum && pf_check_proto_cksum(n,
6678 off, ntohs(h->ip6_plen) - (off - sizeof(struct ip6_hdr)),
6679 IPPROTO_UDP, AF_INET6)) {
6681 REASON_SET(&reason, PFRES_PROTCKSUM);
6684 if (uh.uh_dport == 0 ||
6685 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
6686 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
6688 REASON_SET(&reason, PFRES_SHORT);
6691 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
6692 if (action == PF_PASS) {
6694 pfsync_update_state(s);
6695 #endif /* NPFSYNC */
6699 } else if (s == NULL)
6700 action = pf_test_udp(&r, &s, dir, kif,
6701 m, off, h, &pd, &a, &ruleset, NULL, inp);
6705 case IPPROTO_ICMPV6: {
6706 struct icmp6_hdr ih;
6709 if (!pf_pull_hdr(m, off, &ih, sizeof(ih),
6710 &action, &reason, AF_INET6)) {
6711 log = action != PF_PASS;
6714 if (dir == PF_IN && pf_check_proto_cksum(n, off,
6715 ntohs(h->ip6_plen) - (off - sizeof(struct ip6_hdr)),
6716 IPPROTO_ICMPV6, AF_INET6)) {
6718 REASON_SET(&reason, PFRES_PROTCKSUM);
6721 action = pf_test_state_icmp(&s, dir, kif,
6722 m, off, h, &pd, &reason);
6723 if (action == PF_PASS) {
6725 pfsync_update_state(s);
6726 #endif /* NPFSYNC */
6730 } else if (s == NULL)
6731 action = pf_test_icmp(&r, &s, dir, kif,
6732 m, off, h, &pd, &a, &ruleset, NULL);
6737 action = pf_test_state_other(&s, dir, kif, &pd);
6738 if (action == PF_PASS) {
6740 pfsync_update_state(s);
6741 #endif /* NPFSYNC */
6745 } else if (s == NULL)
6746 action = pf_test_other(&r, &s, dir, kif, m, off, h,
6747 &pd, &a, &ruleset, NULL);
6757 /* XXX handle IPv6 options, if not allowed. not implemented. */
6759 if ((s && s->tag) || r->rtableid)
6760 pf_tag_packet(m, pd.pf_mtag, s ? s->tag : 0, r->rtableid);
6763 if (action == PF_PASS && r->qid) {
6764 if (pd.tos & IPTOS_LOWDELAY)
6765 pd.pf_mtag->qid = r->pqid;
6767 pd.pf_mtag->qid = r->qid;
6768 /* add hints for ecn */
6769 pd.pf_mtag->af = AF_INET6;
6770 pd.pf_mtag->hdr = h;
6771 /* add connection hash for fairq */
6773 KKASSERT(s->hash != 0);
6774 pd.pf_mtag->flags |= PF_TAG_STATE_HASHED;
6775 pd.pf_mtag->state_hash = s->hash;
6780 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
6781 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
6782 (s->nat_rule.ptr->action == PF_RDR ||
6783 s->nat_rule.ptr->action == PF_BINAT) &&
6784 IN6_IS_ADDR_LOOPBACK(&pd.dst->v6))
6785 pd.pf_mtag->flags |= PF_TAG_TRANSLATE_LOCALHOST;
6790 if (s != NULL && s->nat_rule.ptr != NULL &&
6791 s->nat_rule.ptr->log & PF_LOG_ALL)
6792 lr = s->nat_rule.ptr;
6795 PFLOG_PACKET(kif, h, m, AF_INET6, dir, reason, lr, a, ruleset,
6799 kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
6800 kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++;
6802 if (action == PF_PASS || r->action == PF_DROP) {
6803 dirndx = (dir == PF_OUT);
6804 r->packets[dirndx]++;
6805 r->bytes[dirndx] += pd.tot_len;
6807 a->packets[dirndx]++;
6808 a->bytes[dirndx] += pd.tot_len;
6811 if (s->nat_rule.ptr != NULL) {
6812 s->nat_rule.ptr->packets[dirndx]++;
6813 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len;
6815 if (s->src_node != NULL) {
6816 s->src_node->packets[dirndx]++;
6817 s->src_node->bytes[dirndx] += pd.tot_len;
6819 if (s->nat_src_node != NULL) {
6820 s->nat_src_node->packets[dirndx]++;
6821 s->nat_src_node->bytes[dirndx] += pd.tot_len;
6823 dirndx = (dir == s->direction) ? 0 : 1;
6824 s->packets[dirndx]++;
6825 s->bytes[dirndx] += pd.tot_len;
6828 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
6832 * XXX: we need to make sure that the addresses
6833 * passed to pfr_update_stats() are the same than
6834 * the addresses used during matching (pfr_match)
6836 if (r == &pf_default_rule) {
6838 x = (s == NULL || s->direction == dir) ?
6839 &pd.baddr : &pd.naddr;
6841 x = (s == NULL || s->direction == dir) ?
6842 &pd.naddr : &pd.baddr;
6844 if (x == &pd.baddr || s == NULL) {
6851 if (tr->src.addr.type == PF_ADDR_TABLE)
6852 pfr_update_stats(tr->src.addr.p.tbl, (s == NULL ||
6853 s->direction == dir) ? pd.src : pd.dst, pd.af,
6854 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
6856 if (tr->dst.addr.type == PF_ADDR_TABLE)
6857 pfr_update_stats(tr->dst.addr.p.tbl, (s == NULL ||
6858 s->direction == dir) ? pd.dst : pd.src, pd.af,
6859 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
6864 if (action == PF_SYNPROXY_DROP) {
6869 /* pf_route6 can free the mbuf causing *m0 to become NULL */
6870 pf_route6(m0, r, dir, ifp, s, &pd);
6877 pf_check_congestion(struct ifqueue *ifq)