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 $ */
7 * Copyright (c) 2004 The DragonFly Project. All rights reserved.
9 * Copyright (c) 2001 Daniel Hartmeier
10 * Copyright (c) 2002,2003 Henning Brauer
11 * All rights reserved.
13 * Redistribution and use in source and binary forms, with or without
14 * modification, are permitted provided that the following conditions
17 * - Redistributions of source code must retain the above copyright
18 * notice, this list of conditions and the following disclaimer.
19 * - Redistributions in binary form must reproduce the above
20 * copyright notice, this list of conditions and the following
21 * disclaimer in the documentation and/or other materials provided
22 * with the distribution.
24 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
25 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
26 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
27 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
28 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
29 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
30 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
31 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
32 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
34 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
35 * POSSIBILITY OF SUCH DAMAGE.
37 * Effort sponsored in part by the Defense Advanced Research Projects
38 * Agency (DARPA) and Air Force Research Laboratory, Air Force
39 * Materiel Command, USAF, under agreement number F30602-01-2-0537.
44 #include "opt_inet6.h"
45 #include "use_pfsync.h"
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/malloc.h>
51 #include <sys/filio.h>
52 #include <sys/socket.h>
53 #include <sys/socketvar.h>
54 #include <sys/kernel.h>
56 #include <sys/sysctl.h>
57 #include <sys/endian.h>
58 #include <vm/vm_zone.h>
60 #include <machine/inttypes.h>
63 #include <net/if_types.h>
65 #include <net/netisr.h>
66 #include <net/route.h>
68 #include <netinet/in.h>
69 #include <netinet/in_var.h>
70 #include <netinet/in_systm.h>
71 #include <netinet/ip.h>
72 #include <netinet/ip_var.h>
73 #include <netinet/tcp.h>
74 #include <netinet/tcp_seq.h>
75 #include <netinet/udp.h>
76 #include <netinet/ip_icmp.h>
77 #include <netinet/in_pcb.h>
78 #include <netinet/tcp_timer.h>
79 #include <netinet/tcp_var.h>
80 #include <netinet/udp_var.h>
81 #include <netinet/icmp_var.h>
83 #include <net/pf/pfvar.h>
84 #include <net/pf/if_pflog.h>
87 #include <net/pf/if_pfsync.h>
88 #endif /* NPFSYNC > 0 */
91 #include <netinet/ip6.h>
92 #include <netinet/in_pcb.h>
93 #include <netinet/icmp6.h>
94 #include <netinet6/nd6.h>
95 #include <netinet6/ip6_var.h>
96 #include <netinet6/in6_pcb.h>
99 #include <sys/in_cksum.h>
100 #include <sys/ucred.h>
101 #include <machine/limits.h>
102 #include <sys/msgport2.h>
103 #include <net/netmsg2.h>
105 extern int ip_optcopy(struct ip *, struct ip *);
107 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) kprintf x
113 struct pf_anchorqueue pf_anchors;
114 struct pf_ruleset pf_main_ruleset;
115 struct pf_altqqueue pf_altqs[2];
116 struct pf_palist pf_pabuf;
117 struct pf_altqqueue *pf_altqs_active;
118 struct pf_altqqueue *pf_altqs_inactive;
119 struct pf_status pf_status;
121 u_int32_t ticket_altqs_active;
122 u_int32_t ticket_altqs_inactive;
123 int altqs_inactive_open;
124 u_int32_t ticket_pabuf;
126 struct callout pf_expire_to; /* expire timeout */
128 vm_zone_t pf_src_tree_pl, pf_rule_pl;
129 vm_zone_t pf_state_pl, pf_altq_pl, pf_pooladdr_pl;
131 void pf_print_host(struct pf_addr *, u_int16_t, u_int8_t);
132 void pf_print_state(struct pf_state *);
133 void pf_print_flags(u_int8_t);
135 u_int16_t pf_cksum_fixup(u_int16_t, u_int16_t, u_int16_t,
137 void pf_change_ap(struct pf_addr *, u_int16_t *,
138 u_int16_t *, u_int16_t *, struct pf_addr *,
139 u_int16_t, u_int8_t, sa_family_t);
141 void pf_change_a6(struct pf_addr *, u_int16_t *,
142 struct pf_addr *, u_int8_t);
144 void pf_change_icmp(struct pf_addr *, u_int16_t *,
145 struct pf_addr *, struct pf_addr *, u_int16_t,
146 u_int16_t *, u_int16_t *, u_int16_t *,
147 u_int16_t *, u_int8_t, sa_family_t);
148 void pf_send_tcp(const struct pf_rule *, sa_family_t,
149 const struct pf_addr *, const struct pf_addr *,
150 u_int16_t, u_int16_t, u_int32_t, u_int32_t,
151 u_int8_t, u_int16_t, u_int16_t, u_int8_t);
152 void pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t,
153 sa_family_t, struct pf_rule *);
154 struct pf_rule *pf_match_translation(struct pf_pdesc *, struct mbuf *,
155 int, int, struct pfi_kif *,
156 struct pf_addr *, u_int16_t, struct pf_addr *,
158 struct pf_rule *pf_get_translation(struct pf_pdesc *, struct mbuf *,
159 int, int, struct pfi_kif *, struct pf_src_node **,
160 struct pf_addr *, u_int16_t,
161 struct pf_addr *, u_int16_t,
162 struct pf_addr *, u_int16_t *);
163 int pf_test_tcp(struct pf_rule **, struct pf_state **,
164 int, struct pfi_kif *, struct mbuf *, int,
165 void *, struct pf_pdesc *, struct pf_rule **,
166 struct pf_ruleset **);
167 int pf_test_udp(struct pf_rule **, struct pf_state **,
168 int, struct pfi_kif *, struct mbuf *, int,
169 void *, struct pf_pdesc *, struct pf_rule **,
170 struct pf_ruleset **);
171 int pf_test_icmp(struct pf_rule **, struct pf_state **,
172 int, struct pfi_kif *, struct mbuf *, int,
173 void *, struct pf_pdesc *, struct pf_rule **,
174 struct pf_ruleset **);
175 int pf_test_other(struct pf_rule **, struct pf_state **,
176 int, struct pfi_kif *, struct mbuf *, int, void *,
177 struct pf_pdesc *, struct pf_rule **,
178 struct pf_ruleset **);
179 int pf_test_fragment(struct pf_rule **, int,
180 struct pfi_kif *, struct mbuf *, void *,
181 struct pf_pdesc *, struct pf_rule **,
182 struct pf_ruleset **);
183 int pf_test_state_tcp(struct pf_state **, int,
184 struct pfi_kif *, struct mbuf *, int,
185 void *, struct pf_pdesc *, u_short *);
186 int pf_test_state_udp(struct pf_state **, int,
187 struct pfi_kif *, struct mbuf *, int,
188 void *, struct pf_pdesc *);
189 int pf_test_state_icmp(struct pf_state **, int,
190 struct pfi_kif *, struct mbuf *, int,
191 void *, struct pf_pdesc *);
192 int pf_test_state_other(struct pf_state **, int,
193 struct pfi_kif *, struct pf_pdesc *);
194 static int pf_match_tag(struct mbuf *, struct pf_rule *,
195 struct pf_rule *, int *);
196 void pf_hash(struct pf_addr *, struct pf_addr *,
197 struct pf_poolhashkey *, sa_family_t);
198 int pf_map_addr(u_int8_t, struct pf_rule *,
199 struct pf_addr *, struct pf_addr *,
200 struct pf_addr *, struct pf_src_node **);
201 int pf_get_sport(sa_family_t, u_int8_t, struct pf_rule *,
202 struct pf_addr *, struct pf_addr *, u_int16_t,
203 struct pf_addr *, u_int16_t*, u_int16_t, u_int16_t,
204 struct pf_src_node **);
205 void pf_route(struct mbuf **, struct pf_rule *, int,
206 struct ifnet *, struct pf_state *);
207 void pf_route6(struct mbuf **, struct pf_rule *, int,
208 struct ifnet *, struct pf_state *);
209 int pf_socket_lookup(uid_t *, gid_t *,
210 int, struct pf_pdesc *);
211 u_int8_t pf_get_wscale(struct mbuf *, int, u_int16_t,
213 u_int16_t pf_get_mss(struct mbuf *, int, u_int16_t,
215 u_int16_t pf_calc_mss(struct pf_addr *, sa_family_t,
217 void pf_set_rt_ifp(struct pf_state *,
219 int pf_check_proto_cksum(struct mbuf *, int, int,
220 u_int8_t, sa_family_t);
221 int pf_addr_wrap_neq(struct pf_addr_wrap *,
222 struct pf_addr_wrap *);
223 struct pf_state *pf_find_state_recurse(struct pfi_kif *,
224 struct pf_state *, u_int8_t);
226 struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX];
228 #define STATE_LOOKUP() \
230 if (direction == PF_IN) \
231 *state = pf_find_state_recurse( \
232 kif, &key, PF_EXT_GWY); \
234 *state = pf_find_state_recurse( \
235 kif, &key, PF_LAN_EXT); \
236 if (*state == NULL) \
238 if (direction == PF_OUT && \
239 (((*state)->rule.ptr->rt == PF_ROUTETO && \
240 (*state)->rule.ptr->direction == PF_OUT) || \
241 ((*state)->rule.ptr->rt == PF_REPLYTO && \
242 (*state)->rule.ptr->direction == PF_IN)) && \
243 (*state)->rt_kif != NULL && \
244 (*state)->rt_kif != kif) \
248 #define STATE_TRANSLATE(s) \
249 (s)->lan.addr.addr32[0] != (s)->gwy.addr.addr32[0] || \
250 ((s)->af == AF_INET6 && \
251 ((s)->lan.addr.addr32[1] != (s)->gwy.addr.addr32[1] || \
252 (s)->lan.addr.addr32[2] != (s)->gwy.addr.addr32[2] || \
253 (s)->lan.addr.addr32[3] != (s)->gwy.addr.addr32[3])) || \
254 (s)->lan.port != (s)->gwy.port
256 #define BOUND_IFACE(r, k) (((r)->rule_flag & PFRULE_IFBOUND) ? (k) : \
257 ((r)->rule_flag & PFRULE_GRBOUND) ? (k)->pfik_parent : \
258 (k)->pfik_parent->pfik_parent)
260 static int pf_src_compare(struct pf_src_node *, struct pf_src_node *);
261 static int pf_state_compare_lan_ext(struct pf_state *,
263 static int pf_state_compare_ext_gwy(struct pf_state *,
265 static int pf_state_compare_id(struct pf_state *,
268 struct pf_src_tree tree_src_tracking;
270 struct pf_state_tree_id tree_id;
271 struct pf_state_queue state_updates;
273 RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare);
274 RB_GENERATE(pf_state_tree_lan_ext, pf_state,
275 u.s.entry_lan_ext, pf_state_compare_lan_ext);
276 RB_GENERATE(pf_state_tree_ext_gwy, pf_state,
277 u.s.entry_ext_gwy, pf_state_compare_ext_gwy);
278 RB_GENERATE(pf_state_tree_id, pf_state,
279 u.s.entry_id, pf_state_compare_id);
282 pf_src_compare(struct pf_src_node *a, struct pf_src_node *b)
286 if (a->rule.ptr > b->rule.ptr)
288 if (a->rule.ptr < b->rule.ptr)
290 if ((diff = a->af - b->af) != 0)
295 if (a->addr.addr32[0] > b->addr.addr32[0])
297 if (a->addr.addr32[0] < b->addr.addr32[0])
303 if (a->addr.addr32[3] > b->addr.addr32[3])
305 if (a->addr.addr32[3] < b->addr.addr32[3])
307 if (a->addr.addr32[2] > b->addr.addr32[2])
309 if (a->addr.addr32[2] < b->addr.addr32[2])
311 if (a->addr.addr32[1] > b->addr.addr32[1])
313 if (a->addr.addr32[1] < b->addr.addr32[1])
315 if (a->addr.addr32[0] > b->addr.addr32[0])
317 if (a->addr.addr32[0] < b->addr.addr32[0])
326 pf_state_hash(struct pf_state *s)
328 u_int32_t hv = (intptr_t)s / sizeof(*s);
330 hv ^= crc32(&s->lan, sizeof(s->lan));
331 hv ^= crc32(&s->gwy, sizeof(s->gwy));
332 hv ^= crc32(&s->ext, sizeof(s->ext));
333 if (hv == 0) /* disallow 0 */
339 pf_state_compare_lan_ext(struct pf_state *a, struct pf_state *b)
343 if ((diff = a->proto - b->proto) != 0)
345 if ((diff = a->af - b->af) != 0)
350 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
352 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
354 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
356 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
362 if (a->lan.addr.addr32[3] > b->lan.addr.addr32[3])
364 if (a->lan.addr.addr32[3] < b->lan.addr.addr32[3])
366 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
368 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
370 if (a->lan.addr.addr32[2] > b->lan.addr.addr32[2])
372 if (a->lan.addr.addr32[2] < b->lan.addr.addr32[2])
374 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
376 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
378 if (a->lan.addr.addr32[1] > b->lan.addr.addr32[1])
380 if (a->lan.addr.addr32[1] < b->lan.addr.addr32[1])
382 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
384 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
386 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
388 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
390 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
392 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
398 if ((diff = a->lan.port - b->lan.port) != 0)
400 if ((diff = a->ext.port - b->ext.port) != 0)
407 pf_state_compare_ext_gwy(struct pf_state *a, struct pf_state *b)
411 if ((diff = a->proto - b->proto) != 0)
413 if ((diff = a->af - b->af) != 0)
418 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
420 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
422 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
424 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
430 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
432 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
434 if (a->gwy.addr.addr32[3] > b->gwy.addr.addr32[3])
436 if (a->gwy.addr.addr32[3] < b->gwy.addr.addr32[3])
438 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
440 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
442 if (a->gwy.addr.addr32[2] > b->gwy.addr.addr32[2])
444 if (a->gwy.addr.addr32[2] < b->gwy.addr.addr32[2])
446 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
448 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
450 if (a->gwy.addr.addr32[1] > b->gwy.addr.addr32[1])
452 if (a->gwy.addr.addr32[1] < b->gwy.addr.addr32[1])
454 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
456 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
458 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
460 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
466 if ((diff = a->ext.port - b->ext.port) != 0)
468 if ((diff = a->gwy.port - b->gwy.port) != 0)
475 pf_state_compare_id(struct pf_state *a, struct pf_state *b)
481 if (a->creatorid > b->creatorid)
483 if (a->creatorid < b->creatorid)
491 pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af)
496 dst->addr32[0] = src->addr32[0];
500 dst->addr32[0] = src->addr32[0];
501 dst->addr32[1] = src->addr32[1];
502 dst->addr32[2] = src->addr32[2];
503 dst->addr32[3] = src->addr32[3];
510 pf_find_state_byid(struct pf_state *key)
512 pf_status.fcounters[FCNT_STATE_SEARCH]++;
513 return (RB_FIND(pf_state_tree_id, &tree_id, key));
517 pf_find_state_recurse(struct pfi_kif *kif, struct pf_state *key, u_int8_t tree)
521 pf_status.fcounters[FCNT_STATE_SEARCH]++;
525 for (; kif != NULL; kif = kif->pfik_parent) {
526 s = RB_FIND(pf_state_tree_lan_ext,
527 &kif->pfik_lan_ext, key);
533 for (; kif != NULL; kif = kif->pfik_parent) {
534 s = RB_FIND(pf_state_tree_ext_gwy,
535 &kif->pfik_ext_gwy, key);
541 panic("pf_find_state_recurse");
546 pf_find_state_all(struct pf_state *key, u_int8_t tree, int *more)
548 struct pf_state *s, *ss = NULL;
551 pf_status.fcounters[FCNT_STATE_SEARCH]++;
555 TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states) {
556 s = RB_FIND(pf_state_tree_lan_ext,
557 &kif->pfik_lan_ext, key);
567 TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states) {
568 s = RB_FIND(pf_state_tree_ext_gwy,
569 &kif->pfik_ext_gwy, key);
579 panic("pf_find_state_all");
584 pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule,
585 struct pf_addr *src, sa_family_t af)
587 struct pf_src_node k;
591 PF_ACPY(&k.addr, src, af);
592 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
593 rule->rpool.opts & PF_POOL_STICKYADDR)
597 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
598 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
601 if (!rule->max_src_nodes ||
602 rule->src_nodes < rule->max_src_nodes)
603 (*sn) = pool_get(&pf_src_tree_pl, PR_NOWAIT);
606 bzero(*sn, sizeof(struct pf_src_node));
608 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
609 rule->rpool.opts & PF_POOL_STICKYADDR)
610 (*sn)->rule.ptr = rule;
612 (*sn)->rule.ptr = NULL;
613 PF_ACPY(&(*sn)->addr, src, af);
614 if (RB_INSERT(pf_src_tree,
615 &tree_src_tracking, *sn) != NULL) {
616 if (pf_status.debug >= PF_DEBUG_MISC) {
617 kprintf("pf: src_tree insert failed: ");
618 pf_print_host(&(*sn)->addr, 0, af);
621 pool_put(&pf_src_tree_pl, *sn);
624 (*sn)->creation = time_second;
625 (*sn)->ruletype = rule->action;
626 if ((*sn)->rule.ptr != NULL)
627 (*sn)->rule.ptr->src_nodes++;
628 pf_status.scounters[SCNT_SRC_NODE_INSERT]++;
629 pf_status.src_nodes++;
631 if (rule->max_src_states &&
632 (*sn)->states >= rule->max_src_states)
639 pf_insert_state(struct pfi_kif *kif, struct pf_state *state)
641 /* Thou MUST NOT insert multiple duplicate keys */
642 state->u.s.kif = kif;
643 if (RB_INSERT(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state)) {
644 if (pf_status.debug >= PF_DEBUG_MISC) {
645 kprintf("pf: state insert failed: tree_lan_ext");
647 pf_print_host(&state->lan.addr, state->lan.port,
650 pf_print_host(&state->gwy.addr, state->gwy.port,
653 pf_print_host(&state->ext.addr, state->ext.port,
655 if (state->sync_flags & PFSTATE_FROMSYNC)
656 kprintf(" (from sync)");
662 if (RB_INSERT(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, state)) {
663 if (pf_status.debug >= PF_DEBUG_MISC) {
664 kprintf("pf: state insert failed: tree_ext_gwy");
666 pf_print_host(&state->lan.addr, state->lan.port,
669 pf_print_host(&state->gwy.addr, state->gwy.port,
672 pf_print_host(&state->ext.addr, state->ext.port,
674 if (state->sync_flags & PFSTATE_FROMSYNC)
675 kprintf(" (from sync)");
678 RB_REMOVE(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state);
682 if (state->id == 0 && state->creatorid == 0) {
683 state->id = htobe64(pf_status.stateid++);
684 state->creatorid = pf_status.hostid;
686 if (RB_INSERT(pf_state_tree_id, &tree_id, state) != NULL) {
687 if (pf_status.debug >= PF_DEBUG_MISC) {
688 kprintf("pf: state insert failed: "
689 "id: %016" PRIx64 " creatorid: %08" PRIx32,
690 be64toh(state->id), ntohl(state->creatorid));
691 if (state->sync_flags & PFSTATE_FROMSYNC)
692 kprintf(" (from sync)");
695 RB_REMOVE(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state);
696 RB_REMOVE(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, state);
699 TAILQ_INSERT_HEAD(&state_updates, state, u.s.entry_updates);
701 pf_status.fcounters[FCNT_STATE_INSERT]++;
703 pfi_attach_state(kif);
705 pfsync_insert_state(state);
711 pf_purge_timeout(void *arg)
713 struct callout *to = arg;
716 pf_purge_expired_states();
717 pf_purge_expired_fragments();
718 pf_purge_expired_src_nodes();
721 callout_reset(to, pf_default_rule.timeout[PFTM_INTERVAL] * hz,
722 pf_purge_timeout, to);
726 pf_state_expires(const struct pf_state *state)
733 /* handle all PFTM_* > PFTM_MAX here */
734 if (state->timeout == PFTM_PURGE)
735 return (time_second);
736 if (state->timeout == PFTM_UNTIL_PACKET)
738 KASSERT((state->timeout < PFTM_MAX),
739 ("pf_state_expires: timeout > PFTM_MAX"));
740 timeout = state->rule.ptr->timeout[state->timeout];
742 timeout = pf_default_rule.timeout[state->timeout];
743 start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START];
745 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END];
746 states = state->rule.ptr->states;
748 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START];
749 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END];
750 states = pf_status.states;
752 if (end && states > start && start < end) {
754 return (state->expire + timeout * (end - states) /
757 return (time_second);
759 return (state->expire + timeout);
763 pf_purge_expired_src_nodes(void)
765 struct pf_src_node *cur, *next;
767 for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) {
768 next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur);
770 if (cur->states <= 0 && cur->expire <= time_second) {
771 if (cur->rule.ptr != NULL) {
772 cur->rule.ptr->src_nodes--;
773 if (cur->rule.ptr->states <= 0 &&
774 cur->rule.ptr->max_src_nodes <= 0)
775 pf_rm_rule(NULL, cur->rule.ptr);
777 RB_REMOVE(pf_src_tree, &tree_src_tracking, cur);
778 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
779 pf_status.src_nodes--;
780 pool_put(&pf_src_tree_pl, cur);
786 pf_src_tree_remove_state(struct pf_state *s)
790 if (s->src_node != NULL) {
791 if (--s->src_node->states <= 0) {
792 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
795 pf_default_rule.timeout[PFTM_SRC_NODE];
796 s->src_node->expire = time_second + timeout;
799 if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) {
800 if (--s->nat_src_node->states <= 0) {
801 timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
804 pf_default_rule.timeout[PFTM_SRC_NODE];
805 s->nat_src_node->expire = time_second + timeout;
808 s->src_node = s->nat_src_node = NULL;
812 pf_purge_expired_states_callback(struct pf_state *cur, void *data __unused)
814 if (pf_state_expires(cur) <= time_second) {
815 RB_REMOVE(pf_state_tree_ext_gwy,
816 &cur->u.s.kif->pfik_ext_gwy, cur);
817 RB_REMOVE(pf_state_tree_lan_ext,
818 &cur->u.s.kif->pfik_lan_ext, cur);
819 RB_REMOVE(pf_state_tree_id, &tree_id, cur);
820 if (cur->src.state == PF_TCPS_PROXY_DST) {
821 pf_send_tcp(cur->rule.ptr, cur->af,
822 &cur->ext.addr, &cur->lan.addr,
823 cur->ext.port, cur->lan.port,
824 cur->src.seqhi, cur->src.seqlo + 1, 0,
825 TH_RST|TH_ACK, 0, 0);
828 pfsync_delete_state(cur);
830 pf_src_tree_remove_state(cur);
831 if (--cur->rule.ptr->states <= 0 &&
832 cur->rule.ptr->src_nodes <= 0)
833 pf_rm_rule(NULL, cur->rule.ptr);
834 if (cur->nat_rule.ptr != NULL)
835 if (--cur->nat_rule.ptr->states <= 0 &&
836 cur->nat_rule.ptr->src_nodes <= 0)
837 pf_rm_rule(NULL, cur->nat_rule.ptr);
838 if (cur->anchor.ptr != NULL)
839 if (--cur->anchor.ptr->states <= 0)
840 pf_rm_rule(NULL, cur->anchor.ptr);
841 pf_normalize_tcp_cleanup(cur);
842 pfi_detach_state(cur->u.s.kif);
843 TAILQ_REMOVE(&state_updates, cur, u.s.entry_updates);
844 pool_put(&pf_state_pl, cur);
845 pf_status.fcounters[FCNT_STATE_REMOVALS]++;
852 pf_purge_expired_states(void)
854 RB_SCAN(pf_state_tree_id, &tree_id, NULL,
855 pf_purge_expired_states_callback, NULL);
860 pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw)
862 if (aw->type != PF_ADDR_TABLE)
864 if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL)
870 pf_tbladdr_remove(struct pf_addr_wrap *aw)
872 if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL)
874 pfr_detach_table(aw->p.tbl);
879 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
881 struct pfr_ktable *kt = aw->p.tbl;
883 if (aw->type != PF_ADDR_TABLE || kt == NULL)
885 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
888 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
893 pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af)
898 u_int32_t a = ntohl(addr->addr32[0]);
899 kprintf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255,
911 u_int8_t i, curstart = 255, curend = 0,
912 maxstart = 0, maxend = 0;
913 for (i = 0; i < 8; i++) {
914 if (!addr->addr16[i]) {
921 if ((curend - curstart) >
922 (maxend - maxstart)) {
930 for (i = 0; i < 8; i++) {
931 if (i >= maxstart && i <= maxend) {
940 b = ntohs(addr->addr16[i]);
957 pf_print_state(struct pf_state *s)
973 kprintf("%u ", s->proto);
976 pf_print_host(&s->lan.addr, s->lan.port, s->af);
978 pf_print_host(&s->gwy.addr, s->gwy.port, s->af);
980 pf_print_host(&s->ext.addr, s->ext.port, s->af);
981 kprintf(" [lo=%u high=%u win=%u modulator=%u", s->src.seqlo,
982 s->src.seqhi, s->src.max_win, s->src.seqdiff);
983 if (s->src.wscale && s->dst.wscale)
984 kprintf(" wscale=%u", s->src.wscale & PF_WSCALE_MASK);
986 kprintf(" [lo=%u high=%u win=%u modulator=%u", s->dst.seqlo,
987 s->dst.seqhi, s->dst.max_win, s->dst.seqdiff);
988 if (s->src.wscale && s->dst.wscale)
989 kprintf(" wscale=%u", s->dst.wscale & PF_WSCALE_MASK);
991 kprintf(" %u:%u", s->src.state, s->dst.state);
995 pf_print_flags(u_int8_t f)
1017 #define PF_SET_SKIP_STEPS(i) \
1019 while (head[i] != cur) { \
1020 head[i]->skip[i].ptr = cur; \
1021 head[i] = TAILQ_NEXT(head[i], entries); \
1026 pf_calc_skip_steps(struct pf_rulequeue *rules)
1028 struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT];
1031 cur = TAILQ_FIRST(rules);
1033 for (i = 0; i < PF_SKIP_COUNT; ++i)
1035 while (cur != NULL) {
1037 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
1038 PF_SET_SKIP_STEPS(PF_SKIP_IFP);
1039 if (cur->direction != prev->direction)
1040 PF_SET_SKIP_STEPS(PF_SKIP_DIR);
1041 if (cur->af != prev->af)
1042 PF_SET_SKIP_STEPS(PF_SKIP_AF);
1043 if (cur->proto != prev->proto)
1044 PF_SET_SKIP_STEPS(PF_SKIP_PROTO);
1045 if (cur->src.not != prev->src.not ||
1046 pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr))
1047 PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR);
1048 if (cur->src.port[0] != prev->src.port[0] ||
1049 cur->src.port[1] != prev->src.port[1] ||
1050 cur->src.port_op != prev->src.port_op)
1051 PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
1052 if (cur->dst.not != prev->dst.not ||
1053 pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr))
1054 PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR);
1055 if (cur->dst.port[0] != prev->dst.port[0] ||
1056 cur->dst.port[1] != prev->dst.port[1] ||
1057 cur->dst.port_op != prev->dst.port_op)
1058 PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
1061 cur = TAILQ_NEXT(cur, entries);
1063 for (i = 0; i < PF_SKIP_COUNT; ++i)
1064 PF_SET_SKIP_STEPS(i);
1068 pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2)
1070 if (aw1->type != aw2->type)
1072 switch (aw1->type) {
1073 case PF_ADDR_ADDRMASK:
1074 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0))
1076 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0))
1079 case PF_ADDR_DYNIFTL:
1080 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt);
1081 case PF_ADDR_NOROUTE:
1084 return (aw1->p.tbl != aw2->p.tbl);
1086 kprintf("invalid address type: %d\n", aw1->type);
1092 pf_update_anchor_rules(void)
1094 struct pf_rule *rule;
1097 for (i = 0; i < PF_RULESET_MAX; ++i)
1098 TAILQ_FOREACH(rule, pf_main_ruleset.rules[i].active.ptr,
1100 if (rule->anchorname[0])
1101 rule->anchor = pf_find_anchor(rule->anchorname);
1103 rule->anchor = NULL;
1107 pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp)
1113 l = cksum + old - new;
1114 l = (l >> 16) + (l & 65535);
1122 pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc,
1123 struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af)
1128 PF_ACPY(&ao, a, af);
1136 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1137 ao.addr16[0], an->addr16[0], 0),
1138 ao.addr16[1], an->addr16[1], 0);
1140 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1141 ao.addr16[0], an->addr16[0], u),
1142 ao.addr16[1], an->addr16[1], u),
1148 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1149 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1150 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1151 ao.addr16[0], an->addr16[0], u),
1152 ao.addr16[1], an->addr16[1], u),
1153 ao.addr16[2], an->addr16[2], u),
1154 ao.addr16[3], an->addr16[3], u),
1155 ao.addr16[4], an->addr16[4], u),
1156 ao.addr16[5], an->addr16[5], u),
1157 ao.addr16[6], an->addr16[6], u),
1158 ao.addr16[7], an->addr16[7], u),
1166 /* Changes a u_int32_t. Uses a void * so there are no align restrictions */
1168 pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u)
1172 memcpy(&ao, a, sizeof(ao));
1173 memcpy(a, &an, sizeof(u_int32_t));
1174 *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u),
1175 ao % 65536, an % 65536, u);
1180 pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u)
1184 PF_ACPY(&ao, a, AF_INET6);
1185 PF_ACPY(a, an, AF_INET6);
1187 *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1188 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1189 pf_cksum_fixup(pf_cksum_fixup(*c,
1190 ao.addr16[0], an->addr16[0], u),
1191 ao.addr16[1], an->addr16[1], u),
1192 ao.addr16[2], an->addr16[2], u),
1193 ao.addr16[3], an->addr16[3], u),
1194 ao.addr16[4], an->addr16[4], u),
1195 ao.addr16[5], an->addr16[5], u),
1196 ao.addr16[6], an->addr16[6], u),
1197 ao.addr16[7], an->addr16[7], u);
1202 pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa,
1203 struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c,
1204 u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af)
1206 struct pf_addr oia, ooa;
1208 PF_ACPY(&oia, ia, af);
1209 PF_ACPY(&ooa, oa, af);
1211 /* Change inner protocol port, fix inner protocol checksum. */
1213 u_int16_t oip = *ip;
1220 *pc = pf_cksum_fixup(*pc, oip, *ip, u);
1221 *ic = pf_cksum_fixup(*ic, oip, *ip, 0);
1223 *ic = pf_cksum_fixup(*ic, opc, *pc, 0);
1225 /* Change inner ip address, fix inner ip and icmp checksums. */
1226 PF_ACPY(ia, na, af);
1230 u_int32_t oh2c = *h2c;
1232 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c,
1233 oia.addr16[0], ia->addr16[0], 0),
1234 oia.addr16[1], ia->addr16[1], 0);
1235 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1236 oia.addr16[0], ia->addr16[0], 0),
1237 oia.addr16[1], ia->addr16[1], 0);
1238 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0);
1244 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1245 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1246 pf_cksum_fixup(pf_cksum_fixup(*ic,
1247 oia.addr16[0], ia->addr16[0], u),
1248 oia.addr16[1], ia->addr16[1], u),
1249 oia.addr16[2], ia->addr16[2], u),
1250 oia.addr16[3], ia->addr16[3], u),
1251 oia.addr16[4], ia->addr16[4], u),
1252 oia.addr16[5], ia->addr16[5], u),
1253 oia.addr16[6], ia->addr16[6], u),
1254 oia.addr16[7], ia->addr16[7], u);
1258 /* Change outer ip address, fix outer ip or icmpv6 checksum. */
1259 PF_ACPY(oa, na, af);
1263 *hc = pf_cksum_fixup(pf_cksum_fixup(*hc,
1264 ooa.addr16[0], oa->addr16[0], 0),
1265 ooa.addr16[1], oa->addr16[1], 0);
1270 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1271 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1272 pf_cksum_fixup(pf_cksum_fixup(*ic,
1273 ooa.addr16[0], oa->addr16[0], u),
1274 ooa.addr16[1], oa->addr16[1], u),
1275 ooa.addr16[2], oa->addr16[2], u),
1276 ooa.addr16[3], oa->addr16[3], u),
1277 ooa.addr16[4], oa->addr16[4], u),
1278 ooa.addr16[5], oa->addr16[5], u),
1279 ooa.addr16[6], oa->addr16[6], u),
1280 ooa.addr16[7], oa->addr16[7], u);
1287 pf_send_tcp(const struct pf_rule *r, sa_family_t af,
1288 const struct pf_addr *saddr, const struct pf_addr *daddr,
1289 u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack,
1290 u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl)
1295 struct ip *h = NULL;
1298 struct ip6_hdr *h6 = NULL;
1300 struct tcphdr *th = NULL;
1303 /* maximum segment size tcp option */
1304 tlen = sizeof(struct tcphdr);
1311 len = sizeof(struct ip) + tlen;
1316 len = sizeof(struct ip6_hdr) + tlen;
1321 /* create outgoing mbuf */
1322 m = m_gethdr(MB_DONTWAIT, MT_HEADER);
1325 m->m_pkthdr.fw_flags = PF_MBUF_GENERATED;
1327 if (r != NULL && r->qid) {
1328 m->m_pkthdr.fw_flags |= ALTQ_MBUF_TAGGED;
1329 m->m_pkthdr.altq_qid = r->qid;
1330 m->m_pkthdr.ecn_af = af;
1331 m->m_pkthdr.header = mtod(m, struct ip *);
1334 m->m_data += max_linkhdr;
1335 m->m_pkthdr.len = m->m_len = len;
1336 m->m_pkthdr.rcvif = NULL;
1337 bzero(m->m_data, len);
1341 h = mtod(m, struct ip *);
1343 /* IP header fields included in the TCP checksum */
1344 h->ip_p = IPPROTO_TCP;
1346 h->ip_src.s_addr = saddr->v4.s_addr;
1347 h->ip_dst.s_addr = daddr->v4.s_addr;
1349 th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip));
1354 h6 = mtod(m, struct ip6_hdr *);
1356 /* IP header fields included in the TCP checksum */
1357 h6->ip6_nxt = IPPROTO_TCP;
1358 h6->ip6_plen = htons(tlen);
1359 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr));
1360 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr));
1362 th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr));
1368 th->th_sport = sport;
1369 th->th_dport = dport;
1370 th->th_seq = htonl(seq);
1371 th->th_ack = htonl(ack);
1372 th->th_off = tlen >> 2;
1373 th->th_flags = flags;
1374 th->th_win = htons(win);
1377 opt = (char *)(th + 1);
1378 opt[0] = TCPOPT_MAXSEG;
1381 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2);
1388 th->th_sum = in_cksum(m, len);
1390 /* Finish the IP header */
1392 h->ip_hl = sizeof(*h) >> 2;
1393 h->ip_tos = IPTOS_LOWDELAY;
1395 h->ip_off = path_mtu_discovery ? IP_DF : 0;
1396 h->ip_ttl = ttl ? ttl : ip_defttl;
1398 ip_output(m, NULL, NULL, 0, NULL, NULL);
1404 th->th_sum = in6_cksum(m, IPPROTO_TCP,
1405 sizeof(struct ip6_hdr), tlen);
1407 h6->ip6_vfc |= IPV6_VERSION;
1408 h6->ip6_hlim = IPV6_DEFHLIM;
1410 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
1417 pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af,
1422 m0 = m_copypacket(m, MB_DONTWAIT);
1425 m0->m_pkthdr.fw_flags |= PF_MBUF_GENERATED;
1429 m->m_pkthdr.fw_flags |= ALTQ_MBUF_TAGGED;
1430 m->m_pkthdr.altq_qid = r->qid;
1431 m->m_pkthdr.ecn_af = af;
1432 m->m_pkthdr.header = mtod(m0, struct ip *);
1439 icmp_error(m0, type, code, 0, 0);
1444 icmp6_error(m0, type, code, 0);
1451 * Return 1 if the addresses a and b match (with mask m), otherwise return 0.
1452 * If n is 0, they match if they are equal. If n is != 0, they match if they
1456 pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m,
1457 struct pf_addr *b, sa_family_t af)
1464 if ((a->addr32[0] & m->addr32[0]) ==
1465 (b->addr32[0] & m->addr32[0]))
1471 if (((a->addr32[0] & m->addr32[0]) ==
1472 (b->addr32[0] & m->addr32[0])) &&
1473 ((a->addr32[1] & m->addr32[1]) ==
1474 (b->addr32[1] & m->addr32[1])) &&
1475 ((a->addr32[2] & m->addr32[2]) ==
1476 (b->addr32[2] & m->addr32[2])) &&
1477 ((a->addr32[3] & m->addr32[3]) ==
1478 (b->addr32[3] & m->addr32[3])))
1497 pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p)
1501 return ((p > a1) && (p < a2));
1503 return ((p < a1) || (p > a2));
1505 return ((p >= a1) && (p <= a2));
1519 return (0); /* never reached */
1523 pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p)
1528 return (pf_match(op, a1, a2, p));
1532 pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u)
1534 if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
1536 return (pf_match(op, a1, a2, u));
1540 pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g)
1542 if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
1544 return (pf_match(op, a1, a2, g));
1548 pf_match_tag(struct mbuf *m, struct pf_rule *r, struct pf_rule *nat_rule,
1551 if (*tag == -1) { /* find mbuf tag */
1552 if (nat_rule != NULL && nat_rule->tag)
1553 *tag = nat_rule->tag;
1554 else if (m->m_pkthdr.fw_flags & PF_MBUF_TAGGED)
1555 *tag = m->m_pkthdr.pf_tag;
1560 return ((!r->match_tag_not && r->match_tag == *tag) ||
1561 (r->match_tag_not && r->match_tag != *tag));
1565 pf_tag_packet(struct mbuf *m, int tag)
1570 m->m_pkthdr.fw_flags |= PF_MBUF_TAGGED;
1571 m->m_pkthdr.pf_tag = tag;
1574 #define PF_STEP_INTO_ANCHOR(r, a, s, n) \
1576 if ((r) == NULL || (r)->anchor == NULL || \
1577 (s) != NULL || (a) != NULL) \
1578 panic("PF_STEP_INTO_ANCHOR"); \
1580 (s) = TAILQ_FIRST(&(r)->anchor->rulesets); \
1582 while ((s) != NULL && ((r) = \
1583 TAILQ_FIRST((s)->rules[n].active.ptr)) == NULL) \
1584 (s) = TAILQ_NEXT((s), entries); \
1585 if ((r) == NULL) { \
1586 (r) = TAILQ_NEXT((a), entries); \
1591 #define PF_STEP_OUT_OF_ANCHOR(r, a, s, n) \
1593 if ((r) != NULL || (a) == NULL || (s) == NULL) \
1594 panic("PF_STEP_OUT_OF_ANCHOR"); \
1595 (s) = TAILQ_NEXT((s), entries); \
1596 while ((s) != NULL && ((r) = \
1597 TAILQ_FIRST((s)->rules[n].active.ptr)) == NULL) \
1598 (s) = TAILQ_NEXT((s), entries); \
1599 if ((r) == NULL) { \
1600 (r) = TAILQ_NEXT((a), entries); \
1607 pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr,
1608 struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af)
1613 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
1614 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
1618 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
1619 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
1620 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) |
1621 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]);
1622 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) |
1623 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]);
1624 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) |
1625 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]);
1631 pf_addr_inc(struct pf_addr *addr, sa_family_t af)
1636 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1);
1640 if (addr->addr32[3] == 0xffffffff) {
1641 addr->addr32[3] = 0;
1642 if (addr->addr32[2] == 0xffffffff) {
1643 addr->addr32[2] = 0;
1644 if (addr->addr32[1] == 0xffffffff) {
1645 addr->addr32[1] = 0;
1647 htonl(ntohl(addr->addr32[0]) + 1);
1650 htonl(ntohl(addr->addr32[1]) + 1);
1653 htonl(ntohl(addr->addr32[2]) + 1);
1656 htonl(ntohl(addr->addr32[3]) + 1);
1662 #define mix(a,b,c) \
1664 a -= b; a -= c; a ^= (c >> 13); \
1665 b -= c; b -= a; b ^= (a << 8); \
1666 c -= a; c -= b; c ^= (b >> 13); \
1667 a -= b; a -= c; a ^= (c >> 12); \
1668 b -= c; b -= a; b ^= (a << 16); \
1669 c -= a; c -= b; c ^= (b >> 5); \
1670 a -= b; a -= c; a ^= (c >> 3); \
1671 b -= c; b -= a; b ^= (a << 10); \
1672 c -= a; c -= b; c ^= (b >> 15); \
1676 * hash function based on bridge_hash in if_bridge.c
1679 pf_hash(struct pf_addr *inaddr, struct pf_addr *hash,
1680 struct pf_poolhashkey *key, sa_family_t af)
1682 u_int32_t a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0];
1687 a += inaddr->addr32[0];
1690 hash->addr32[0] = c + key->key32[2];
1695 a += inaddr->addr32[0];
1696 b += inaddr->addr32[2];
1698 hash->addr32[0] = c;
1699 a += inaddr->addr32[1];
1700 b += inaddr->addr32[3];
1703 hash->addr32[1] = c;
1704 a += inaddr->addr32[2];
1705 b += inaddr->addr32[1];
1708 hash->addr32[2] = c;
1709 a += inaddr->addr32[3];
1710 b += inaddr->addr32[0];
1713 hash->addr32[3] = c;
1720 pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr,
1721 struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn)
1723 unsigned char hash[16];
1724 struct pf_pool *rpool = &r->rpool;
1725 struct pf_addr *raddr = &rpool->cur->addr.v.a.addr;
1726 struct pf_addr *rmask = &rpool->cur->addr.v.a.mask;
1727 struct pf_pooladdr *acur = rpool->cur;
1728 struct pf_src_node k;
1730 if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR &&
1731 (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
1733 PF_ACPY(&k.addr, saddr, af);
1734 if (r->rule_flag & PFRULE_RULESRCTRACK ||
1735 r->rpool.opts & PF_POOL_STICKYADDR)
1739 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
1740 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
1741 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) {
1742 PF_ACPY(naddr, &(*sn)->raddr, af);
1743 if (pf_status.debug >= PF_DEBUG_MISC) {
1744 kprintf("pf_map_addr: src tracking maps ");
1745 pf_print_host(&k.addr, 0, af);
1747 pf_print_host(naddr, 0, af);
1754 if (rpool->cur->addr.type == PF_ADDR_NOROUTE)
1756 if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
1757 if (af == AF_INET) {
1758 if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 &&
1759 (rpool->opts & PF_POOL_TYPEMASK) !=
1762 raddr = &rpool->cur->addr.p.dyn->pfid_addr4;
1763 rmask = &rpool->cur->addr.p.dyn->pfid_mask4;
1765 if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 &&
1766 (rpool->opts & PF_POOL_TYPEMASK) !=
1769 raddr = &rpool->cur->addr.p.dyn->pfid_addr6;
1770 rmask = &rpool->cur->addr.p.dyn->pfid_mask6;
1772 } else if (rpool->cur->addr.type == PF_ADDR_TABLE) {
1773 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN)
1774 return (1); /* unsupported */
1776 raddr = &rpool->cur->addr.v.a.addr;
1777 rmask = &rpool->cur->addr.v.a.mask;
1780 switch (rpool->opts & PF_POOL_TYPEMASK) {
1782 PF_ACPY(naddr, raddr, af);
1784 case PF_POOL_BITMASK:
1785 PF_POOLMASK(naddr, raddr, rmask, saddr, af);
1787 case PF_POOL_RANDOM:
1788 if (init_addr != NULL && PF_AZERO(init_addr, af)) {
1792 rpool->counter.addr32[0] = karc4random();
1797 if (rmask->addr32[3] != 0xffffffff)
1798 rpool->counter.addr32[3] = karc4random();
1801 if (rmask->addr32[2] != 0xffffffff)
1802 rpool->counter.addr32[2] = karc4random();
1805 if (rmask->addr32[1] != 0xffffffff)
1806 rpool->counter.addr32[1] = karc4random();
1809 if (rmask->addr32[0] != 0xffffffff)
1810 rpool->counter.addr32[0] = karc4random();
1814 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
1815 PF_ACPY(init_addr, naddr, af);
1818 PF_AINC(&rpool->counter, af);
1819 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
1822 case PF_POOL_SRCHASH:
1823 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af);
1824 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af);
1826 case PF_POOL_ROUNDROBIN:
1827 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
1828 if (!pfr_pool_get(rpool->cur->addr.p.tbl,
1829 &rpool->tblidx, &rpool->counter,
1830 &raddr, &rmask, af))
1832 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
1833 if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
1834 &rpool->tblidx, &rpool->counter,
1835 &raddr, &rmask, af))
1837 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af))
1841 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL)
1842 rpool->cur = TAILQ_FIRST(&rpool->list);
1843 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
1845 if (pfr_pool_get(rpool->cur->addr.p.tbl,
1846 &rpool->tblidx, &rpool->counter,
1847 &raddr, &rmask, af)) {
1848 /* table contains no address of type 'af' */
1849 if (rpool->cur != acur)
1853 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
1855 if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
1856 &rpool->tblidx, &rpool->counter,
1857 &raddr, &rmask, af)) {
1858 /* table contains no address of type 'af' */
1859 if (rpool->cur != acur)
1864 raddr = &rpool->cur->addr.v.a.addr;
1865 rmask = &rpool->cur->addr.v.a.mask;
1866 PF_ACPY(&rpool->counter, raddr, af);
1870 PF_ACPY(naddr, &rpool->counter, af);
1871 PF_AINC(&rpool->counter, af);
1875 PF_ACPY(&(*sn)->raddr, naddr, af);
1877 if (pf_status.debug >= PF_DEBUG_MISC &&
1878 (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
1879 kprintf("pf_map_addr: selected address ");
1880 pf_print_host(naddr, 0, af);
1888 pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r,
1889 struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport,
1890 struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high,
1891 struct pf_src_node **sn)
1893 struct pf_state key;
1894 struct pf_addr init_addr;
1897 bzero(&init_addr, sizeof(init_addr));
1898 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
1904 PF_ACPY(&key.ext.addr, daddr, key.af);
1905 PF_ACPY(&key.gwy.addr, naddr, key.af);
1906 key.ext.port = dport;
1909 * port search; start random, step;
1910 * similar 2 portloop in in_pcbbind
1912 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP)) {
1914 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
1916 } else if (low == 0 && high == 0) {
1917 key.gwy.port = *nport;
1918 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
1920 } else if (low == high) {
1921 key.gwy.port = htons(low);
1922 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) {
1923 *nport = htons(low);
1935 cut = karc4random() % (1 + high - low) + low;
1936 /* low <= cut <= high */
1937 for (tmp = cut; tmp <= high; ++(tmp)) {
1938 key.gwy.port = htons(tmp);
1939 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
1941 *nport = htons(tmp);
1945 for (tmp = cut - 1; tmp >= low; --(tmp)) {
1946 key.gwy.port = htons(tmp);
1947 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
1949 *nport = htons(tmp);
1955 switch (r->rpool.opts & PF_POOL_TYPEMASK) {
1956 case PF_POOL_RANDOM:
1957 case PF_POOL_ROUNDROBIN:
1958 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
1962 case PF_POOL_SRCHASH:
1963 case PF_POOL_BITMASK:
1967 } while (! PF_AEQ(&init_addr, naddr, af) );
1969 return (1); /* none available */
1973 pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off,
1974 int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport,
1975 struct pf_addr *daddr, u_int16_t dport, int rs_num)
1977 struct pf_rule *r, *rm = NULL, *anchorrule = NULL;
1978 struct pf_ruleset *ruleset = NULL;
1980 r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr);
1981 while (r && rm == NULL) {
1982 struct pf_rule_addr *src = NULL, *dst = NULL;
1983 struct pf_addr_wrap *xdst = NULL;
1985 if (r->action == PF_BINAT && direction == PF_IN) {
1987 if (r->rpool.cur != NULL)
1988 xdst = &r->rpool.cur->addr;
1995 if (r->kif != NULL &&
1996 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
1997 r = r->skip[PF_SKIP_IFP].ptr;
1998 else if (r->direction && r->direction != direction)
1999 r = r->skip[PF_SKIP_DIR].ptr;
2000 else if (r->af && r->af != pd->af)
2001 r = r->skip[PF_SKIP_AF].ptr;
2002 else if (r->proto && r->proto != pd->proto)
2003 r = r->skip[PF_SKIP_PROTO].ptr;
2004 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af, src->not))
2005 r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR :
2006 PF_SKIP_DST_ADDR].ptr;
2007 else if (src->port_op && !pf_match_port(src->port_op,
2008 src->port[0], src->port[1], sport))
2009 r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT :
2010 PF_SKIP_DST_PORT].ptr;
2011 else if (dst != NULL &&
2012 PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->not))
2013 r = r->skip[PF_SKIP_DST_ADDR].ptr;
2014 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af, 0))
2015 r = TAILQ_NEXT(r, entries);
2016 else if (dst != NULL && dst->port_op &&
2017 !pf_match_port(dst->port_op, dst->port[0],
2018 dst->port[1], dport))
2019 r = r->skip[PF_SKIP_DST_PORT].ptr;
2020 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto !=
2021 IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m,
2022 off, pd->hdr.tcp), r->os_fingerprint)))
2023 r = TAILQ_NEXT(r, entries);
2024 else if (r->anchorname[0] && r->anchor == NULL)
2025 r = TAILQ_NEXT(r, entries);
2026 else if (r->anchor == NULL)
2029 PF_STEP_INTO_ANCHOR(r, anchorrule, ruleset, rs_num);
2030 if (r == NULL && anchorrule != NULL)
2031 PF_STEP_OUT_OF_ANCHOR(r, anchorrule, ruleset,
2034 if (rm != NULL && (rm->action == PF_NONAT ||
2035 rm->action == PF_NORDR || rm->action == PF_NOBINAT))
2041 pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction,
2042 struct pfi_kif *kif, struct pf_src_node **sn,
2043 struct pf_addr *saddr, u_int16_t sport,
2044 struct pf_addr *daddr, u_int16_t dport,
2045 struct pf_addr *naddr, u_int16_t *nport)
2047 struct pf_rule *r = NULL;
2049 if (direction == PF_OUT) {
2050 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2051 sport, daddr, dport, PF_RULESET_BINAT);
2053 r = pf_match_translation(pd, m, off, direction, kif,
2054 saddr, sport, daddr, dport, PF_RULESET_NAT);
2056 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2057 sport, daddr, dport, PF_RULESET_RDR);
2059 r = pf_match_translation(pd, m, off, direction, kif,
2060 saddr, sport, daddr, dport, PF_RULESET_BINAT);
2064 switch (r->action) {
2070 if (pf_get_sport(pd->af, pd->proto, r, saddr,
2071 daddr, dport, naddr, nport, r->rpool.proxy_port[0],
2072 r->rpool.proxy_port[1], sn)) {
2073 DPFPRINTF(PF_DEBUG_MISC,
2074 ("pf: NAT proxy port allocation "
2076 r->rpool.proxy_port[0],
2077 r->rpool.proxy_port[1]));
2082 switch (direction) {
2084 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){
2085 if (pd->af == AF_INET) {
2086 if (r->rpool.cur->addr.p.dyn->
2090 &r->rpool.cur->addr.p.dyn->
2092 &r->rpool.cur->addr.p.dyn->
2096 if (r->rpool.cur->addr.p.dyn->
2100 &r->rpool.cur->addr.p.dyn->
2102 &r->rpool.cur->addr.p.dyn->
2108 &r->rpool.cur->addr.v.a.addr,
2109 &r->rpool.cur->addr.v.a.mask,
2113 if (r->src.addr.type == PF_ADDR_DYNIFTL){
2114 if (pd->af == AF_INET) {
2115 if (r->src.addr.p.dyn->
2119 &r->src.addr.p.dyn->
2121 &r->src.addr.p.dyn->
2125 if (r->src.addr.p.dyn->
2129 &r->src.addr.p.dyn->
2131 &r->src.addr.p.dyn->
2137 &r->src.addr.v.a.addr,
2138 &r->src.addr.v.a.mask, daddr,
2144 if (pf_map_addr(r->af, r, saddr, naddr, NULL, sn))
2147 if (r->rpool.proxy_port[1]) {
2148 u_int32_t tmp_nport;
2150 tmp_nport = ((ntohs(dport) -
2151 ntohs(r->dst.port[0])) %
2152 (r->rpool.proxy_port[1] -
2153 r->rpool.proxy_port[0] + 1)) +
2154 r->rpool.proxy_port[0];
2156 /* wrap around if necessary */
2157 if (tmp_nport > 65535)
2159 *nport = htons((u_int16_t)tmp_nport);
2160 } else if (r->rpool.proxy_port[0])
2161 *nport = htons(r->rpool.proxy_port[0]);
2173 struct netmsg_hashlookup {
2174 struct netmsg nm_netmsg;
2175 struct inpcb **nm_pinp;
2176 struct inpcbinfo *nm_pcbinfo;
2177 struct pf_addr *nm_saddr;
2178 struct pf_addr *nm_daddr;
2185 in_pcblookup_hash_handler(struct netmsg *msg0)
2187 struct netmsg_hashlookup *msg = (struct netmsg_hashlookup *)msg0;
2189 if (msg->nm_af == AF_INET)
2190 *msg->nm_pinp = in_pcblookup_hash(msg->nm_pcbinfo,
2191 msg->nm_saddr->v4, msg->nm_sport, msg->nm_daddr->v4,
2192 msg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2195 *msg->nm_pinp = in6_pcblookup_hash(msg->nm_pcbinfo,
2196 &msg->nm_saddr->v6, msg->nm_sport, &msg->nm_daddr->v6,
2197 msg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2199 lwkt_replymsg(&msg->nm_netmsg.nm_lmsg, 0);
2204 pf_socket_lookup(uid_t *uid, gid_t *gid, int direction, struct pf_pdesc *pd)
2206 struct pf_addr *saddr, *daddr;
2207 u_int16_t sport, dport;
2208 struct inpcbinfo *pi;
2211 struct netmsg_hashlookup *msg = NULL;
2217 if (direction == PF_IN) {
2224 switch (pd->proto) {
2226 sport = pd->hdr.tcp->th_sport;
2227 dport = pd->hdr.tcp->th_dport;
2229 pi_cpu = tcp_addrcpu(saddr->v4.s_addr, sport, daddr->v4.s_addr, dport);
2230 pi = &tcbinfo[pi_cpu];
2233 * Our netstack runs lockless on MP systems
2234 * (only for TCP connections at the moment).
2236 * As we are not allowed to read another CPU's tcbinfo,
2237 * we have to ask that CPU via remote call to search the
2240 * Prepare a msg iff data belongs to another CPU.
2242 if (pi_cpu != mycpu->gd_cpuid) {
2243 msg = kmalloc(sizeof(*msg), M_LWKTMSG, M_INTWAIT);
2244 netmsg_init(&msg->nm_netmsg, &netisr_afree_rport, 0,
2245 in_pcblookup_hash_handler);
2246 msg->nm_pinp = &inp;
2247 msg->nm_pcbinfo = pi;
2248 msg->nm_saddr = saddr;
2249 msg->nm_sport = sport;
2250 msg->nm_daddr = daddr;
2251 msg->nm_dport = dport;
2252 msg->nm_af = pd->af;
2257 sport = pd->hdr.udp->uh_sport;
2258 dport = pd->hdr.udp->uh_dport;
2264 if (direction != PF_IN) {
2276 * Query other CPU, second part
2278 * msg only gets initialized when:
2280 * 2) the info belongs to another CPU
2282 * Use some switch/case magic to avoid code duplication.
2287 inp = in6_pcblookup_hash(pi, &saddr->v6, sport,
2288 &daddr->v6, dport, INPLOOKUP_WILDCARD, NULL);
2294 /* FALLTHROUGH if SMP and on other CPU */
2299 lwkt_sendmsg(tcp_cport(pi_cpu),
2300 &msg->nm_netmsg.nm_lmsg);
2304 inp = in_pcblookup_hash(pi, saddr->v4, sport, daddr->v4,
2305 dport, INPLOOKUP_WILDCARD, NULL);
2314 *uid = inp->inp_socket->so_cred->cr_uid;
2315 *gid = inp->inp_socket->so_cred->cr_groups[0];
2320 pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2324 u_int8_t *opt, optlen;
2325 u_int8_t wscale = 0;
2327 hlen = th_off << 2; /* hlen <= sizeof(hdr) */
2328 if (hlen <= sizeof(struct tcphdr))
2330 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2332 opt = hdr + sizeof(struct tcphdr);
2333 hlen -= sizeof(struct tcphdr);
2343 if (wscale > TCP_MAX_WINSHIFT)
2344 wscale = TCP_MAX_WINSHIFT;
2345 wscale |= PF_WSCALE_FLAG;
2360 pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2364 u_int8_t *opt, optlen;
2365 u_int16_t mss = tcp_mssdflt;
2367 hlen = th_off << 2; /* hlen <= sizeof(hdr) */
2368 if (hlen <= sizeof(struct tcphdr))
2370 if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2372 opt = hdr + sizeof(struct tcphdr);
2373 hlen -= sizeof(struct tcphdr);
2374 while (hlen >= TCPOLEN_MAXSEG) {
2382 bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2);
2397 pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer)
2400 struct sockaddr_in *dst;
2404 struct sockaddr_in6 *dst6;
2405 struct route_in6 ro6;
2407 struct rtentry *rt = NULL;
2409 u_int16_t mss = tcp_mssdflt;
2414 hlen = sizeof(struct ip);
2415 bzero(&ro, sizeof(ro));
2416 dst = (struct sockaddr_in *)&ro.ro_dst;
2417 dst->sin_family = AF_INET;
2418 dst->sin_len = sizeof(*dst);
2419 dst->sin_addr = addr->v4;
2420 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
2426 hlen = sizeof(struct ip6_hdr);
2427 bzero(&ro6, sizeof(ro6));
2428 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst;
2429 dst6->sin6_family = AF_INET6;
2430 dst6->sin6_len = sizeof(*dst6);
2431 dst6->sin6_addr = addr->v6;
2432 rtalloc_ign((struct route *)&ro6, (RTF_CLONING | RTF_PRCLONING));
2438 if (rt && rt->rt_ifp) {
2439 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr);
2440 mss = max(tcp_mssdflt, mss);
2443 mss = min(mss, offer);
2444 mss = max(mss, 64); /* sanity - at least max opt space */
2449 pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr)
2451 struct pf_rule *r = s->rule.ptr;
2454 if (!r->rt || r->rt == PF_FASTROUTE)
2459 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL,
2461 s->rt_kif = r->rpool.cur->kif;
2466 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL,
2468 s->rt_kif = r->rpool.cur->kif;
2475 pf_test_tcp(struct pf_rule **rm, struct pf_state **sm, int direction,
2476 struct pfi_kif *kif, struct mbuf *m, int off, void *h,
2477 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
2479 struct pf_rule *nr = NULL;
2480 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
2481 struct tcphdr *th = pd->hdr.tcp;
2482 u_int16_t bport, nport = 0;
2483 sa_family_t af = pd->af;
2487 struct pf_rule *r, *a = NULL;
2488 struct pf_ruleset *ruleset = NULL;
2489 struct pf_src_node *nsn = NULL;
2493 u_int16_t mss = tcp_mssdflt;
2495 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
2497 if (direction == PF_OUT) {
2498 bport = nport = th->th_sport;
2499 /* check outgoing packet for BINAT/NAT */
2500 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
2501 saddr, th->th_sport, daddr, th->th_dport,
2502 &pd->naddr, &nport)) != NULL) {
2503 PF_ACPY(&pd->baddr, saddr, af);
2504 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
2505 &th->th_sum, &pd->naddr, nport, 0, af);
2512 bport = nport = th->th_dport;
2513 /* check incoming packet for BINAT/RDR */
2514 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
2515 saddr, th->th_sport, daddr, th->th_dport,
2516 &pd->naddr, &nport)) != NULL) {
2517 PF_ACPY(&pd->baddr, daddr, af);
2518 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
2519 &th->th_sum, &pd->naddr, nport, 0, af);
2529 if (r->kif != NULL &&
2530 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
2531 r = r->skip[PF_SKIP_IFP].ptr;
2532 else if (r->direction && r->direction != direction)
2533 r = r->skip[PF_SKIP_DIR].ptr;
2534 else if (r->af && r->af != af)
2535 r = r->skip[PF_SKIP_AF].ptr;
2536 else if (r->proto && r->proto != IPPROTO_TCP)
2537 r = r->skip[PF_SKIP_PROTO].ptr;
2538 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
2539 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
2540 else if (r->src.port_op && !pf_match_port(r->src.port_op,
2541 r->src.port[0], r->src.port[1], th->th_sport))
2542 r = r->skip[PF_SKIP_SRC_PORT].ptr;
2543 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
2544 r = r->skip[PF_SKIP_DST_ADDR].ptr;
2545 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
2546 r->dst.port[0], r->dst.port[1], th->th_dport))
2547 r = r->skip[PF_SKIP_DST_PORT].ptr;
2548 else if (r->tos && !(r->tos & pd->tos))
2549 r = TAILQ_NEXT(r, entries);
2550 else if (r->rule_flag & PFRULE_FRAGMENT)
2551 r = TAILQ_NEXT(r, entries);
2552 else if ((r->flagset & th->th_flags) != r->flags)
2553 r = TAILQ_NEXT(r, entries);
2554 else if (r->uid.op && (lookup != -1 || (lookup =
2555 pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2556 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
2558 r = TAILQ_NEXT(r, entries);
2559 else if (r->gid.op && (lookup != -1 || (lookup =
2560 pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2561 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
2563 r = TAILQ_NEXT(r, entries);
2564 else if (r->prob && r->prob <= karc4random())
2565 r = TAILQ_NEXT(r, entries);
2566 else if (r->match_tag && !pf_match_tag(m, r, nr, &tag))
2567 r = TAILQ_NEXT(r, entries);
2568 else if (r->anchorname[0] && r->anchor == NULL)
2569 r = TAILQ_NEXT(r, entries);
2570 else if (r->os_fingerprint != PF_OSFP_ANY && !pf_osfp_match(
2571 pf_osfp_fingerprint(pd, m, off, th), r->os_fingerprint))
2572 r = TAILQ_NEXT(r, entries);
2576 if (r->anchor == NULL) {
2582 r = TAILQ_NEXT(r, entries);
2584 PF_STEP_INTO_ANCHOR(r, a, ruleset,
2587 if (r == NULL && a != NULL)
2588 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
2595 REASON_SET(&reason, PFRES_MATCH);
2599 m_copyback(m, off, sizeof(*th), (caddr_t)th);
2600 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
2603 if ((r->action == PF_DROP) &&
2604 ((r->rule_flag & PFRULE_RETURNRST) ||
2605 (r->rule_flag & PFRULE_RETURNICMP) ||
2606 (r->rule_flag & PFRULE_RETURN))) {
2607 /* undo NAT changes, if they have taken place */
2609 if (direction == PF_OUT) {
2610 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
2611 &th->th_sum, &pd->baddr, bport, 0, af);
2614 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
2615 &th->th_sum, &pd->baddr, bport, 0, af);
2619 if (((r->rule_flag & PFRULE_RETURNRST) ||
2620 (r->rule_flag & PFRULE_RETURN)) &&
2621 !(th->th_flags & TH_RST)) {
2622 u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
2624 if (th->th_flags & TH_SYN)
2626 if (th->th_flags & TH_FIN)
2628 pf_send_tcp(r, af, pd->dst,
2629 pd->src, th->th_dport, th->th_sport,
2630 ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0,
2632 } else if ((af == AF_INET) && r->return_icmp)
2633 pf_send_icmp(m, r->return_icmp >> 8,
2634 r->return_icmp & 255, af, r);
2635 else if ((af == AF_INET6) && r->return_icmp6)
2636 pf_send_icmp(m, r->return_icmp6 >> 8,
2637 r->return_icmp6 & 255, af, r);
2640 if (r->action == PF_DROP)
2643 pf_tag_packet(m, tag);
2645 if (r->keep_state || nr != NULL ||
2646 (pd->flags & PFDESC_TCP_NORM)) {
2647 /* create new state */
2649 struct pf_state *s = NULL;
2650 struct pf_src_node *sn = NULL;
2652 len = pd->tot_len - off - (th->th_off << 2);
2654 /* check maximums */
2655 if (r->max_states && (r->states >= r->max_states))
2657 /* src node for flter rule */
2658 if ((r->rule_flag & PFRULE_SRCTRACK ||
2659 r->rpool.opts & PF_POOL_STICKYADDR) &&
2660 pf_insert_src_node(&sn, r, saddr, af) != 0)
2662 /* src node for translation rule */
2663 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
2664 ((direction == PF_OUT &&
2665 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
2666 (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
2668 s = pool_get(&pf_state_pl, PR_NOWAIT);
2671 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
2672 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
2673 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
2674 pf_status.src_nodes--;
2675 pool_put(&pf_src_tree_pl, sn);
2677 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
2679 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
2680 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
2681 pf_status.src_nodes--;
2682 pool_put(&pf_src_tree_pl, nsn);
2684 REASON_SET(&reason, PFRES_MEMORY);
2687 bzero(s, sizeof(*s));
2692 s->nat_rule.ptr = nr;
2693 if (s->nat_rule.ptr != NULL)
2694 s->nat_rule.ptr->states++;
2696 s->allow_opts = r->allow_opts;
2697 s->log = r->log & 2;
2698 s->proto = IPPROTO_TCP;
2699 s->direction = direction;
2701 if (direction == PF_OUT) {
2702 PF_ACPY(&s->gwy.addr, saddr, af);
2703 s->gwy.port = th->th_sport; /* sport */
2704 PF_ACPY(&s->ext.addr, daddr, af);
2705 s->ext.port = th->th_dport;
2707 PF_ACPY(&s->lan.addr, &pd->baddr, af);
2708 s->lan.port = bport;
2710 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
2711 s->lan.port = s->gwy.port;
2714 PF_ACPY(&s->lan.addr, daddr, af);
2715 s->lan.port = th->th_dport;
2716 PF_ACPY(&s->ext.addr, saddr, af);
2717 s->ext.port = th->th_sport;
2719 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
2720 s->gwy.port = bport;
2722 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
2723 s->gwy.port = s->lan.port;
2727 s->hash = pf_state_hash(s);
2728 s->src.seqlo = ntohl(th->th_seq);
2729 s->src.seqhi = s->src.seqlo + len + 1;
2730 s->pickup_mode = r->pickup_mode;
2732 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
2733 r->keep_state == PF_STATE_MODULATE) {
2734 /* Generate sequence number modulator */
2735 while ((s->src.seqdiff = karc4random()) == 0)
2737 pf_change_a(&th->th_seq, &th->th_sum,
2738 htonl(s->src.seqlo + s->src.seqdiff), 0);
2744 * WARNING! NetBSD patched this to not scale max_win up
2745 * on the initial SYN, but they failed to correct the code
2746 * in pf_test_state_tcp() that 'undid' the scaling, and they
2747 * failed to remove the scale factor on successful window
2748 * scale negotiation (and doing so would be difficult in the
2749 * face of retransmission, without adding more flags to the
2752 * After discussions with Daniel Hartmeier and Max Laier
2753 * I've decided not to apply the NetBSD patch.
2755 * The worst that happens is that the undo code on window
2756 * scale negotiation failures will produce a larger
2757 * max_win then actual.
2759 if (th->th_flags & TH_SYN) {
2761 s->src.wscale = pf_get_wscale(m, off, th->th_off, af);
2762 s->sync_flags |= PFSTATE_GOT_SYN1;
2764 s->src.max_win = MAX(ntohs(th->th_win), 1);
2765 if (s->src.wscale & PF_WSCALE_MASK) {
2766 /* Remove scale factor from initial window */
2767 u_int win = s->src.max_win;
2768 win += 1 << (s->src.wscale & PF_WSCALE_MASK);
2769 s->src.max_win = (win - 1) >>
2770 (s->src.wscale & PF_WSCALE_MASK);
2772 if (th->th_flags & TH_FIN)
2776 s->src.state = TCPS_SYN_SENT;
2777 s->dst.state = TCPS_CLOSED;
2778 s->creation = time_second;
2779 s->expire = time_second;
2780 s->timeout = PFTM_TCP_FIRST_PACKET;
2781 pf_set_rt_ifp(s, saddr);
2784 s->src_node->states++;
2787 PF_ACPY(&nsn->raddr, &pd->naddr, af);
2788 s->nat_src_node = nsn;
2789 s->nat_src_node->states++;
2791 if ((pd->flags & PFDESC_TCP_NORM) && pf_normalize_tcp_init(m,
2792 off, pd, th, &s->src, &s->dst)) {
2793 REASON_SET(&reason, PFRES_MEMORY);
2794 pf_src_tree_remove_state(s);
2795 pool_put(&pf_state_pl, s);
2798 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
2799 pf_normalize_tcp_stateful(m, off, pd, &reason, th, &s->src,
2800 &s->dst, &rewrite)) {
2801 pf_normalize_tcp_cleanup(s);
2802 pf_src_tree_remove_state(s);
2803 pool_put(&pf_state_pl, s);
2806 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
2807 pf_normalize_tcp_cleanup(s);
2808 REASON_SET(&reason, PFRES_MEMORY);
2809 pf_src_tree_remove_state(s);
2810 pool_put(&pf_state_pl, s);
2814 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
2815 r->keep_state == PF_STATE_SYNPROXY) {
2816 s->src.state = PF_TCPS_PROXY_SRC;
2818 if (direction == PF_OUT) {
2819 pf_change_ap(saddr, &th->th_sport,
2820 pd->ip_sum, &th->th_sum, &pd->baddr,
2823 pf_change_ap(daddr, &th->th_dport,
2824 pd->ip_sum, &th->th_sum, &pd->baddr,
2828 s->src.seqhi = karc4random();
2829 /* Find mss option */
2830 mss = pf_get_mss(m, off, th->th_off, af);
2831 mss = pf_calc_mss(saddr, af, mss);
2832 mss = pf_calc_mss(daddr, af, mss);
2834 pf_send_tcp(r, af, daddr, saddr, th->th_dport,
2835 th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1,
2836 TH_SYN|TH_ACK, 0, s->src.mss, 0);
2837 return (PF_SYNPROXY_DROP);
2841 /* copy back packet headers if we performed NAT operations */
2843 m_copyback(m, off, sizeof(*th), (caddr_t)th);
2849 pf_test_udp(struct pf_rule **rm, struct pf_state **sm, int direction,
2850 struct pfi_kif *kif, struct mbuf *m, int off, void *h,
2851 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
2853 struct pf_rule *nr = NULL;
2854 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
2855 struct udphdr *uh = pd->hdr.udp;
2856 u_int16_t bport, nport = 0;
2857 sa_family_t af = pd->af;
2861 struct pf_rule *r, *a = NULL;
2862 struct pf_ruleset *ruleset = NULL;
2863 struct pf_src_node *nsn = NULL;
2868 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
2870 if (direction == PF_OUT) {
2871 bport = nport = uh->uh_sport;
2872 /* check outgoing packet for BINAT/NAT */
2873 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
2874 saddr, uh->uh_sport, daddr, uh->uh_dport,
2875 &pd->naddr, &nport)) != NULL) {
2876 PF_ACPY(&pd->baddr, saddr, af);
2877 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum,
2878 &uh->uh_sum, &pd->naddr, nport, 1, af);
2885 bport = nport = uh->uh_dport;
2886 /* check incoming packet for BINAT/RDR */
2887 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
2888 saddr, uh->uh_sport, daddr, uh->uh_dport, &pd->naddr,
2890 PF_ACPY(&pd->baddr, daddr, af);
2891 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum,
2892 &uh->uh_sum, &pd->naddr, nport, 1, af);
2902 if (r->kif != NULL &&
2903 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
2904 r = r->skip[PF_SKIP_IFP].ptr;
2905 else if (r->direction && r->direction != direction)
2906 r = r->skip[PF_SKIP_DIR].ptr;
2907 else if (r->af && r->af != af)
2908 r = r->skip[PF_SKIP_AF].ptr;
2909 else if (r->proto && r->proto != IPPROTO_UDP)
2910 r = r->skip[PF_SKIP_PROTO].ptr;
2911 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
2912 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
2913 else if (r->src.port_op && !pf_match_port(r->src.port_op,
2914 r->src.port[0], r->src.port[1], uh->uh_sport))
2915 r = r->skip[PF_SKIP_SRC_PORT].ptr;
2916 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
2917 r = r->skip[PF_SKIP_DST_ADDR].ptr;
2918 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
2919 r->dst.port[0], r->dst.port[1], uh->uh_dport))
2920 r = r->skip[PF_SKIP_DST_PORT].ptr;
2921 else if (r->tos && !(r->tos & pd->tos))
2922 r = TAILQ_NEXT(r, entries);
2923 else if (r->rule_flag & PFRULE_FRAGMENT)
2924 r = TAILQ_NEXT(r, entries);
2925 else if (r->uid.op && (lookup != -1 || (lookup =
2926 pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2927 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
2929 r = TAILQ_NEXT(r, entries);
2930 else if (r->gid.op && (lookup != -1 || (lookup =
2931 pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2932 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
2934 r = TAILQ_NEXT(r, entries);
2935 else if (r->prob && r->prob <= karc4random())
2936 r = TAILQ_NEXT(r, entries);
2937 else if (r->match_tag && !pf_match_tag(m, r, nr, &tag))
2938 r = TAILQ_NEXT(r, entries);
2939 else if (r->anchorname[0] && r->anchor == NULL)
2940 r = TAILQ_NEXT(r, entries);
2941 else if (r->os_fingerprint != PF_OSFP_ANY)
2942 r = TAILQ_NEXT(r, entries);
2946 if (r->anchor == NULL) {
2952 r = TAILQ_NEXT(r, entries);
2954 PF_STEP_INTO_ANCHOR(r, a, ruleset,
2957 if (r == NULL && a != NULL)
2958 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
2965 REASON_SET(&reason, PFRES_MATCH);
2969 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
2970 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
2973 if ((r->action == PF_DROP) &&
2974 ((r->rule_flag & PFRULE_RETURNICMP) ||
2975 (r->rule_flag & PFRULE_RETURN))) {
2976 /* undo NAT changes, if they have taken place */
2978 if (direction == PF_OUT) {
2979 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum,
2980 &uh->uh_sum, &pd->baddr, bport, 1, af);
2983 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum,
2984 &uh->uh_sum, &pd->baddr, bport, 1, af);
2988 if ((af == AF_INET) && r->return_icmp)
2989 pf_send_icmp(m, r->return_icmp >> 8,
2990 r->return_icmp & 255, af, r);
2991 else if ((af == AF_INET6) && r->return_icmp6)
2992 pf_send_icmp(m, r->return_icmp6 >> 8,
2993 r->return_icmp6 & 255, af, r);
2996 if (r->action == PF_DROP)
2999 pf_tag_packet(m, tag);
3001 if (r->keep_state || nr != NULL) {
3002 /* create new state */
3003 struct pf_state *s = NULL;
3004 struct pf_src_node *sn = NULL;
3006 /* check maximums */
3007 if (r->max_states && (r->states >= r->max_states))
3009 /* src node for flter rule */
3010 if ((r->rule_flag & PFRULE_SRCTRACK ||
3011 r->rpool.opts & PF_POOL_STICKYADDR) &&
3012 pf_insert_src_node(&sn, r, saddr, af) != 0)
3014 /* src node for translation rule */
3015 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3016 ((direction == PF_OUT &&
3017 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3018 (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
3020 s = pool_get(&pf_state_pl, PR_NOWAIT);
3023 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3024 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3025 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3026 pf_status.src_nodes--;
3027 pool_put(&pf_src_tree_pl, sn);
3029 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3031 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3032 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3033 pf_status.src_nodes--;
3034 pool_put(&pf_src_tree_pl, nsn);
3036 REASON_SET(&reason, PFRES_MEMORY);
3039 bzero(s, sizeof(*s));
3044 s->nat_rule.ptr = nr;
3045 if (s->nat_rule.ptr != NULL)
3046 s->nat_rule.ptr->states++;
3048 s->allow_opts = r->allow_opts;
3049 s->log = r->log & 2;
3050 s->proto = IPPROTO_UDP;
3051 s->direction = direction;
3053 if (direction == PF_OUT) {
3054 PF_ACPY(&s->gwy.addr, saddr, af);
3055 s->gwy.port = uh->uh_sport;
3056 PF_ACPY(&s->ext.addr, daddr, af);
3057 s->ext.port = uh->uh_dport;
3059 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3060 s->lan.port = bport;
3062 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3063 s->lan.port = s->gwy.port;
3066 PF_ACPY(&s->lan.addr, daddr, af);
3067 s->lan.port = uh->uh_dport;
3068 PF_ACPY(&s->ext.addr, saddr, af);
3069 s->ext.port = uh->uh_sport;
3071 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3072 s->gwy.port = bport;
3074 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3075 s->gwy.port = s->lan.port;
3078 s->hash = pf_state_hash(s);
3079 s->src.state = PFUDPS_SINGLE;
3080 s->dst.state = PFUDPS_NO_TRAFFIC;
3081 s->creation = time_second;
3082 s->expire = time_second;
3083 s->timeout = PFTM_UDP_FIRST_PACKET;
3084 pf_set_rt_ifp(s, saddr);
3087 s->src_node->states++;
3090 PF_ACPY(&nsn->raddr, &pd->naddr, af);
3091 s->nat_src_node = nsn;
3092 s->nat_src_node->states++;
3094 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3095 REASON_SET(&reason, PFRES_MEMORY);
3096 pf_src_tree_remove_state(s);
3097 pool_put(&pf_state_pl, s);
3103 /* copy back packet headers if we performed NAT operations */
3105 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
3111 pf_test_icmp(struct pf_rule **rm, struct pf_state **sm, int direction,
3112 struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3113 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
3115 struct pf_rule *nr = NULL;
3116 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
3117 struct pf_rule *r, *a = NULL;
3118 struct pf_ruleset *ruleset = NULL;
3119 struct pf_src_node *nsn = NULL;
3121 u_int16_t icmpid = 0;
3122 sa_family_t af = pd->af;
3123 u_int8_t icmptype = 0, icmpcode = 0;
3130 switch (pd->proto) {
3133 icmptype = pd->hdr.icmp->icmp_type;
3134 icmpcode = pd->hdr.icmp->icmp_code;
3135 icmpid = pd->hdr.icmp->icmp_id;
3137 if (icmptype == ICMP_UNREACH ||
3138 icmptype == ICMP_SOURCEQUENCH ||
3139 icmptype == ICMP_REDIRECT ||
3140 icmptype == ICMP_TIMXCEED ||
3141 icmptype == ICMP_PARAMPROB)
3146 case IPPROTO_ICMPV6:
3147 icmptype = pd->hdr.icmp6->icmp6_type;
3148 icmpcode = pd->hdr.icmp6->icmp6_code;
3149 icmpid = pd->hdr.icmp6->icmp6_id;
3151 if (icmptype == ICMP6_DST_UNREACH ||
3152 icmptype == ICMP6_PACKET_TOO_BIG ||
3153 icmptype == ICMP6_TIME_EXCEEDED ||
3154 icmptype == ICMP6_PARAM_PROB)
3160 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3162 if (direction == PF_OUT) {
3163 /* check outgoing packet for BINAT/NAT */
3164 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3165 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3166 PF_ACPY(&pd->baddr, saddr, af);
3170 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3171 pd->naddr.v4.s_addr, 0);
3176 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum,
3187 /* check incoming packet for BINAT/RDR */
3188 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3189 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3190 PF_ACPY(&pd->baddr, daddr, af);
3194 pf_change_a(&daddr->v4.s_addr,
3195 pd->ip_sum, pd->naddr.v4.s_addr, 0);
3200 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum,
3214 if (r->kif != NULL &&
3215 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3216 r = r->skip[PF_SKIP_IFP].ptr;
3217 else if (r->direction && r->direction != direction)
3218 r = r->skip[PF_SKIP_DIR].ptr;
3219 else if (r->af && r->af != af)
3220 r = r->skip[PF_SKIP_AF].ptr;
3221 else if (r->proto && r->proto != pd->proto)
3222 r = r->skip[PF_SKIP_PROTO].ptr;
3223 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
3224 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3225 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
3226 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3227 else if (r->type && r->type != icmptype + 1)
3228 r = TAILQ_NEXT(r, entries);
3229 else if (r->code && r->code != icmpcode + 1)
3230 r = TAILQ_NEXT(r, entries);
3231 else if (r->tos && !(r->tos & pd->tos))
3232 r = TAILQ_NEXT(r, entries);
3233 else if (r->rule_flag & PFRULE_FRAGMENT)
3234 r = TAILQ_NEXT(r, entries);
3235 else if (r->prob && r->prob <= karc4random())
3236 r = TAILQ_NEXT(r, entries);
3237 else if (r->match_tag && !pf_match_tag(m, r, nr, &tag))
3238 r = TAILQ_NEXT(r, entries);
3239 else if (r->anchorname[0] && r->anchor == NULL)
3240 r = TAILQ_NEXT(r, entries);
3241 else if (r->os_fingerprint != PF_OSFP_ANY)
3242 r = TAILQ_NEXT(r, entries);
3246 if (r->anchor == NULL) {
3252 r = TAILQ_NEXT(r, entries);
3254 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3257 if (r == NULL && a != NULL)
3258 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3265 REASON_SET(&reason, PFRES_MATCH);
3270 m_copyback(m, off, sizeof(struct icmp6_hdr),
3271 (caddr_t)pd->hdr.icmp6);
3273 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3276 if (r->action != PF_PASS)
3279 pf_tag_packet(m, tag);
3281 if (!state_icmp && (r->keep_state || nr != NULL)) {
3282 /* create new state */
3283 struct pf_state *s = NULL;
3284 struct pf_src_node *sn = NULL;
3286 /* check maximums */
3287 if (r->max_states && (r->states >= r->max_states))
3289 /* src node for flter rule */
3290 if ((r->rule_flag & PFRULE_SRCTRACK ||
3291 r->rpool.opts & PF_POOL_STICKYADDR) &&
3292 pf_insert_src_node(&sn, r, saddr, af) != 0)
3294 /* src node for translation rule */
3295 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3296 ((direction == PF_OUT &&
3297 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3298 (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
3300 s = pool_get(&pf_state_pl, PR_NOWAIT);
3303 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3304 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3305 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3306 pf_status.src_nodes--;
3307 pool_put(&pf_src_tree_pl, sn);
3309 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3311 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3312 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3313 pf_status.src_nodes--;
3314 pool_put(&pf_src_tree_pl, nsn);
3316 REASON_SET(&reason, PFRES_MEMORY);
3319 bzero(s, sizeof(*s));
3324 s->nat_rule.ptr = nr;
3325 if (s->nat_rule.ptr != NULL)
3326 s->nat_rule.ptr->states++;
3328 s->allow_opts = r->allow_opts;
3329 s->log = r->log & 2;
3330 s->proto = pd->proto;
3331 s->direction = direction;
3333 if (direction == PF_OUT) {
3334 PF_ACPY(&s->gwy.addr, saddr, af);
3335 s->gwy.port = icmpid;
3336 PF_ACPY(&s->ext.addr, daddr, af);
3337 s->ext.port = icmpid;
3339 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3341 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3342 s->lan.port = icmpid;
3344 PF_ACPY(&s->lan.addr, daddr, af);
3345 s->lan.port = icmpid;
3346 PF_ACPY(&s->ext.addr, saddr, af);
3347 s->ext.port = icmpid;
3349 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3351 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3352 s->gwy.port = icmpid;
3354 s->hash = pf_state_hash(s);
3355 s->creation = time_second;
3356 s->expire = time_second;
3357 s->timeout = PFTM_ICMP_FIRST_PACKET;
3358 pf_set_rt_ifp(s, saddr);
3361 s->src_node->states++;
3364 PF_ACPY(&nsn->raddr, &pd->naddr, af);
3365 s->nat_src_node = nsn;
3366 s->nat_src_node->states++;
3368 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3369 REASON_SET(&reason, PFRES_MEMORY);
3370 pf_src_tree_remove_state(s);
3371 pool_put(&pf_state_pl, s);
3378 /* copy back packet headers if we performed IPv6 NAT operations */
3380 m_copyback(m, off, sizeof(struct icmp6_hdr),
3381 (caddr_t)pd->hdr.icmp6);
3388 pf_test_other(struct pf_rule **rm, struct pf_state **sm, int direction,
3389 struct pfi_kif *kif, struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
3390 struct pf_rule **am, struct pf_ruleset **rsm)
3392 struct pf_rule *nr = NULL;
3393 struct pf_rule *r, *a = NULL;
3394 struct pf_ruleset *ruleset = NULL;
3395 struct pf_src_node *nsn = NULL;
3396 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
3397 sa_family_t af = pd->af;
3401 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3403 if (direction == PF_OUT) {
3404 /* check outgoing packet for BINAT/NAT */
3405 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3406 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3407 PF_ACPY(&pd->baddr, saddr, af);
3411 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3412 pd->naddr.v4.s_addr, 0);
3417 PF_ACPY(saddr, &pd->naddr, af);
3426 /* check incoming packet for BINAT/RDR */
3427 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3428 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3429 PF_ACPY(&pd->baddr, daddr, af);
3433 pf_change_a(&daddr->v4.s_addr,
3434 pd->ip_sum, pd->naddr.v4.s_addr, 0);
3439 PF_ACPY(daddr, &pd->naddr, af);
3451 if (r->kif != NULL &&
3452 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3453 r = r->skip[PF_SKIP_IFP].ptr;
3454 else if (r->direction && r->direction != direction)
3455 r = r->skip[PF_SKIP_DIR].ptr;
3456 else if (r->af && r->af != af)
3457 r = r->skip[PF_SKIP_AF].ptr;
3458 else if (r->proto && r->proto != pd->proto)
3459 r = r->skip[PF_SKIP_PROTO].ptr;
3460 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, r->src.not))
3461 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3462 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, r->dst.not))
3463 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3464 else if (r->tos && !(r->tos & pd->tos))
3465 r = TAILQ_NEXT(r, entries);
3466 else if (r->rule_flag & PFRULE_FRAGMENT)
3467 r = TAILQ_NEXT(r, entries);
3468 else if (r->prob && r->prob <= karc4random())
3469 r = TAILQ_NEXT(r, entries);
3470 else if (r->match_tag && !pf_match_tag(m, r, nr, &tag))
3471 r = TAILQ_NEXT(r, entries);
3472 else if (r->anchorname[0] && r->anchor == NULL)
3473 r = TAILQ_NEXT(r, entries);
3474 else if (r->os_fingerprint != PF_OSFP_ANY)
3475 r = TAILQ_NEXT(r, entries);
3479 if (r->anchor == NULL) {
3485 r = TAILQ_NEXT(r, entries);
3487 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3490 if (r == NULL && a != NULL)
3491 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3498 REASON_SET(&reason, PFRES_MATCH);
3501 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3503 if ((r->action == PF_DROP) &&
3504 ((r->rule_flag & PFRULE_RETURNICMP) ||
3505 (r->rule_flag & PFRULE_RETURN))) {
3506 struct pf_addr *a = NULL;
3509 if (direction == PF_OUT)
3518 pf_change_a(&a->v4.s_addr, pd->ip_sum,
3519 pd->baddr.v4.s_addr, 0);
3524 PF_ACPY(a, &pd->baddr, af);
3529 if ((af == AF_INET) && r->return_icmp)
3530 pf_send_icmp(m, r->return_icmp >> 8,
3531 r->return_icmp & 255, af, r);
3532 else if ((af == AF_INET6) && r->return_icmp6)
3533 pf_send_icmp(m, r->return_icmp6 >> 8,
3534 r->return_icmp6 & 255, af, r);
3537 if (r->action != PF_PASS)
3540 pf_tag_packet(m, tag);
3542 if (r->keep_state || nr != NULL) {
3543 /* create new state */
3544 struct pf_state *s = NULL;
3545 struct pf_src_node *sn = NULL;
3547 /* check maximums */
3548 if (r->max_states && (r->states >= r->max_states))
3550 /* src node for flter rule */
3551 if ((r->rule_flag & PFRULE_SRCTRACK ||
3552 r->rpool.opts & PF_POOL_STICKYADDR) &&
3553 pf_insert_src_node(&sn, r, saddr, af) != 0)
3555 /* src node for translation rule */
3556 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3557 ((direction == PF_OUT &&
3558 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3559 (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
3561 s = pool_get(&pf_state_pl, PR_NOWAIT);
3564 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3565 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3566 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3567 pf_status.src_nodes--;
3568 pool_put(&pf_src_tree_pl, sn);
3570 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3572 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3573 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3574 pf_status.src_nodes--;
3575 pool_put(&pf_src_tree_pl, nsn);
3577 REASON_SET(&reason, PFRES_MEMORY);
3580 bzero(s, sizeof(*s));
3585 s->nat_rule.ptr = nr;
3586 if (s->nat_rule.ptr != NULL)
3587 s->nat_rule.ptr->states++;
3589 s->allow_opts = r->allow_opts;
3590 s->log = r->log & 2;
3591 s->proto = pd->proto;
3592 s->direction = direction;
3594 if (direction == PF_OUT) {
3595 PF_ACPY(&s->gwy.addr, saddr, af);
3596 PF_ACPY(&s->ext.addr, daddr, af);
3598 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3600 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3602 PF_ACPY(&s->lan.addr, daddr, af);
3603 PF_ACPY(&s->ext.addr, saddr, af);
3605 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3607 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3609 s->hash = pf_state_hash(s);
3610 s->src.state = PFOTHERS_SINGLE;
3611 s->dst.state = PFOTHERS_NO_TRAFFIC;
3612 s->creation = time_second;
3613 s->expire = time_second;
3614 s->timeout = PFTM_OTHER_FIRST_PACKET;
3615 pf_set_rt_ifp(s, saddr);
3618 s->src_node->states++;
3621 PF_ACPY(&nsn->raddr, &pd->naddr, af);
3622 s->nat_src_node = nsn;
3623 s->nat_src_node->states++;
3625 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3626 REASON_SET(&reason, PFRES_MEMORY);
3627 pf_src_tree_remove_state(s);
3628 pool_put(&pf_state_pl, s);
3638 pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif,
3639 struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am,
3640 struct pf_ruleset **rsm)
3642 struct pf_rule *r, *a = NULL;
3643 struct pf_ruleset *ruleset = NULL;
3644 sa_family_t af = pd->af;
3648 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3651 if (r->kif != NULL &&
3652 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3653 r = r->skip[PF_SKIP_IFP].ptr;
3654 else if (r->direction && r->direction != direction)
3655 r = r->skip[PF_SKIP_DIR].ptr;
3656 else if (r->af && r->af != af)
3657 r = r->skip[PF_SKIP_AF].ptr;
3658 else if (r->proto && r->proto != pd->proto)
3659 r = r->skip[PF_SKIP_PROTO].ptr;
3660 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, r->src.not))
3661 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3662 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, r->dst.not))
3663 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3664 else if (r->tos && !(r->tos & pd->tos))
3665 r = TAILQ_NEXT(r, entries);
3666 else if (r->os_fingerprint != PF_OSFP_ANY)
3667 r = TAILQ_NEXT(r, entries);
3668 else if (pd->proto == IPPROTO_UDP &&
3669 (r->src.port_op || r->dst.port_op))
3670 r = TAILQ_NEXT(r, entries);
3671 else if (pd->proto == IPPROTO_TCP &&
3672 (r->src.port_op || r->dst.port_op || r->flagset))
3673 r = TAILQ_NEXT(r, entries);
3674 else if ((pd->proto == IPPROTO_ICMP ||
3675 pd->proto == IPPROTO_ICMPV6) &&
3676 (r->type || r->code))
3677 r = TAILQ_NEXT(r, entries);
3678 else if (r->prob && r->prob <= karc4random())
3679 r = TAILQ_NEXT(r, entries);
3680 else if (r->match_tag && !pf_match_tag(m, r, NULL, &tag))
3681 r = TAILQ_NEXT(r, entries);
3682 else if (r->anchorname[0] && r->anchor == NULL)
3683 r = TAILQ_NEXT(r, entries);
3685 if (r->anchor == NULL) {
3691 r = TAILQ_NEXT(r, entries);
3693 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3696 if (r == NULL && a != NULL)
3697 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3704 REASON_SET(&reason, PFRES_MATCH);
3707 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3709 if (r->action != PF_PASS)
3712 pf_tag_packet(m, tag);
3718 pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
3719 struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
3722 struct pf_state key;
3723 struct tcphdr *th = pd->hdr.tcp;
3724 u_int16_t win = ntohs(th->th_win);
3725 u_int32_t ack, end, seq;
3729 struct pf_state_peer *src, *dst;
3732 key.proto = IPPROTO_TCP;
3733 if (direction == PF_IN) {
3734 PF_ACPY(&key.ext.addr, pd->src, key.af);
3735 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
3736 key.ext.port = th->th_sport;
3737 key.gwy.port = th->th_dport;
3739 PF_ACPY(&key.lan.addr, pd->src, key.af);
3740 PF_ACPY(&key.ext.addr, pd->dst, key.af);
3741 key.lan.port = th->th_sport;
3742 key.ext.port = th->th_dport;
3747 if (direction == (*state)->direction) {
3748 src = &(*state)->src;
3749 dst = &(*state)->dst;
3751 src = &(*state)->dst;
3752 dst = &(*state)->src;
3755 if ((*state)->src.state == PF_TCPS_PROXY_SRC) {
3756 if (direction != (*state)->direction)
3757 return (PF_SYNPROXY_DROP);
3758 if (th->th_flags & TH_SYN) {
3759 if (ntohl(th->th_seq) != (*state)->src.seqlo)
3761 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
3762 pd->src, th->th_dport, th->th_sport,
3763 (*state)->src.seqhi, ntohl(th->th_seq) + 1,
3764 TH_SYN|TH_ACK, 0, (*state)->src.mss, 0);
3765 return (PF_SYNPROXY_DROP);
3766 } else if (!(th->th_flags & TH_ACK) ||
3767 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
3768 (ntohl(th->th_seq) != (*state)->src.seqlo + 1))
3771 (*state)->src.state = PF_TCPS_PROXY_DST;
3773 if ((*state)->src.state == PF_TCPS_PROXY_DST) {
3774 struct pf_state_host *src, *dst;
3776 if (direction == PF_OUT) {
3777 src = &(*state)->gwy;
3778 dst = &(*state)->ext;
3780 src = &(*state)->ext;
3781 dst = &(*state)->lan;
3783 if (direction == (*state)->direction) {
3784 if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) ||
3785 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
3786 (ntohl(th->th_seq) != (*state)->src.seqlo + 1))
3788 (*state)->src.max_win = MAX(ntohs(th->th_win), 1);
3789 if ((*state)->dst.seqhi == 1)
3790 (*state)->dst.seqhi = karc4random();
3791 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
3792 &dst->addr, src->port, dst->port,
3793 (*state)->dst.seqhi, 0, TH_SYN, 0,
3794 (*state)->src.mss, 0);
3795 return (PF_SYNPROXY_DROP);
3796 } else if (((th->th_flags & (TH_SYN|TH_ACK)) !=
3798 (ntohl(th->th_ack) != (*state)->dst.seqhi + 1))
3801 (*state)->dst.max_win = MAX(ntohs(th->th_win), 1);
3802 (*state)->dst.seqlo = ntohl(th->th_seq);
3803 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
3804 pd->src, th->th_dport, th->th_sport,
3805 ntohl(th->th_ack), ntohl(th->th_seq) + 1,
3806 TH_ACK, (*state)->src.max_win, 0, 0);
3807 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
3808 &dst->addr, src->port, dst->port,
3809 (*state)->src.seqhi + 1, (*state)->src.seqlo + 1,
3810 TH_ACK, (*state)->dst.max_win, 0, 0);
3811 (*state)->src.seqdiff = (*state)->dst.seqhi -
3812 (*state)->src.seqlo;
3813 (*state)->dst.seqdiff = (*state)->src.seqhi -
3814 (*state)->dst.seqlo;
3815 (*state)->src.seqhi = (*state)->src.seqlo +
3816 (*state)->dst.max_win;
3817 (*state)->dst.seqhi = (*state)->dst.seqlo +
3818 (*state)->src.max_win;
3819 (*state)->src.wscale = (*state)->dst.wscale = 0;
3820 (*state)->src.state = (*state)->dst.state =
3822 return (PF_SYNPROXY_DROP);
3826 if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) {
3827 sws = src->wscale & PF_WSCALE_MASK;
3828 dws = dst->wscale & PF_WSCALE_MASK;
3833 * Sequence tracking algorithm from Guido van Rooij's paper:
3834 * http://www.madison-gurkha.com/publications/tcp_filtering/
3838 seq = ntohl(th->th_seq);
3839 if (src->seqlo == 0) {
3841 * First packet from this end. The other end has already set
3842 * the seqlo field. Set its state.
3844 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) &&
3845 src->scrub == NULL) {
3846 if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) {
3847 REASON_SET(reason, PFRES_MEMORY);
3852 /* Deferred generation of sequence number modulator */
3853 if (dst->seqdiff && !src->seqdiff) {
3854 while ((src->seqdiff = karc4random()) == 0)
3856 ack = ntohl(th->th_ack) - dst->seqdiff;
3857 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3859 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3862 ack = ntohl(th->th_ack);
3865 end = seq + pd->p_len;
3866 if (th->th_flags & TH_SYN) {
3868 (*state)->sync_flags |= PFSTATE_GOT_SYN2;
3869 if (dst->wscale & PF_WSCALE_FLAG) {
3870 src->wscale = pf_get_wscale(m, off, th->th_off,
3872 if (src->wscale & PF_WSCALE_FLAG) {
3873 /* Remove scale factor from initial
3875 sws = src->wscale & PF_WSCALE_MASK;
3876 win = ((u_int32_t)win + (1 << sws) - 1)
3878 dws = dst->wscale & PF_WSCALE_MASK;
3881 * Fixup other window. Undo the
3882 * normalization that was done on
3883 * the initial SYN. This can result
3884 * in max_win being larger then
3885 * actual but we don't really have
3888 dst->max_win <<= dst->wscale &
3890 /* in case of a retrans SYN|ACK */
3895 if (th->th_flags & TH_FIN)
3899 if (src->state < TCPS_SYN_SENT)
3900 src->state = TCPS_SYN_SENT;
3903 * May need to slide the window (seqhi may have been set by
3904 * the crappy stack check or if we picked up the connection
3905 * after establishment)
3907 if (src->seqhi == 1 ||
3908 SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi))
3909 src->seqhi = end + MAX(1, dst->max_win << dws);
3910 if (win > src->max_win)
3914 ack = ntohl(th->th_ack) - dst->seqdiff;
3916 /* Modulate sequence numbers */
3917 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3919 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3922 end = seq + pd->p_len;
3923 if (th->th_flags & TH_SYN)
3925 if (th->th_flags & TH_FIN)
3929 if ((th->th_flags & TH_ACK) == 0) {
3930 /* Let it pass through the ack skew check */
3932 } else if ((ack == 0 &&
3933 (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) ||
3934 /* broken tcp stacks do not set ack */
3935 (dst->state < TCPS_SYN_SENT)) {
3937 * Many stacks (ours included) will set the ACK number in an
3938 * FIN|ACK if the SYN times out -- no sequence to ACK.
3944 /* Ease sequencing restrictions on no data packets */
3949 ackskew = dst->seqlo - ack;
3951 #define MAXACKWINDOW (0xffff + 1500) /* 1500 is an arbitrary fudge factor */
3953 if (SEQ_GEQ(src->seqhi, end) &&
3954 /* Last octet inside other's window space */
3955 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) &&
3956 /* Retrans: not more than one window back */
3957 (ackskew >= -MAXACKWINDOW) &&
3958 /* Acking not more than one reassembled fragment backwards */
3959 (ackskew <= (MAXACKWINDOW << sws))) {
3960 /* Acking not more than one window forward */
3962 /* update max window */
3963 if (src->max_win < win)
3965 /* synchronize sequencing */
3966 if (SEQ_GT(end, src->seqlo))
3968 /* slide the window of what the other end can send */
3969 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
3970 dst->seqhi = ack + MAX((win << sws), 1);
3974 if (th->th_flags & TH_SYN)
3975 if (src->state < TCPS_SYN_SENT)
3976 src->state = TCPS_SYN_SENT;
3977 if (th->th_flags & TH_FIN)
3978 if (src->state < TCPS_CLOSING)
3979 src->state = TCPS_CLOSING;
3980 if (th->th_flags & TH_ACK) {
3981 if (dst->state == TCPS_SYN_SENT)
3982 dst->state = TCPS_ESTABLISHED;
3983 else if (dst->state == TCPS_CLOSING)
3984 dst->state = TCPS_FIN_WAIT_2;
3986 if (th->th_flags & TH_RST)
3987 src->state = dst->state = TCPS_TIME_WAIT;
3989 /* update expire time */
3990 (*state)->expire = time_second;
3991 if (src->state >= TCPS_FIN_WAIT_2 &&
3992 dst->state >= TCPS_FIN_WAIT_2)
3993 (*state)->timeout = PFTM_TCP_CLOSED;
3994 else if (src->state >= TCPS_FIN_WAIT_2 ||
3995 dst->state >= TCPS_FIN_WAIT_2)
3996 (*state)->timeout = PFTM_TCP_FIN_WAIT;
3997 else if (src->state < TCPS_ESTABLISHED ||
3998 dst->state < TCPS_ESTABLISHED)
3999 (*state)->timeout = PFTM_TCP_OPENING;
4000 else if (src->state >= TCPS_CLOSING ||
4001 dst->state >= TCPS_CLOSING)
4002 (*state)->timeout = PFTM_TCP_CLOSING;
4004 (*state)->timeout = PFTM_TCP_ESTABLISHED;
4006 /* Fall through to PASS packet */
4008 } else if ((dst->state < TCPS_SYN_SENT ||
4009 dst->state >= TCPS_FIN_WAIT_2 ||
4010 src->state >= TCPS_FIN_WAIT_2) &&
4011 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) &&
4012 /* Within a window forward of the originating packet */
4013 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) {
4014 /* Within a window backward of the originating packet */
4017 * This currently handles three situations:
4018 * 1) Stupid stacks will shotgun SYNs before their peer
4020 * 2) When PF catches an already established stream (the
4021 * firewall rebooted, the state table was flushed, routes
4023 * 3) Packets get funky immediately after the connection
4024 * closes (this should catch Solaris spurious ACK|FINs
4025 * that web servers like to spew after a close)
4027 * This must be a little more careful than the above code
4028 * since packet floods will also be caught here. We don't
4029 * update the TTL here to mitigate the damage of a packet
4030 * flood and so the same code can handle awkward establishment
4031 * and a loosened connection close.
4032 * In the establishment case, a correct peer response will
4033 * validate the connection, go through the normal state code
4034 * and keep updating the state TTL.
4037 if (pf_status.debug >= PF_DEBUG_MISC) {
4038 kprintf("pf: loose state match: ");
4039 pf_print_state(*state);
4040 pf_print_flags(th->th_flags);
4041 kprintf(" seq=%u ack=%u len=%u ackskew=%d pkts=%d:%d\n",
4042 seq, ack, pd->p_len, ackskew,
4043 (*state)->packets[0], (*state)->packets[1]);
4046 /* update max window */
4047 if (src->max_win < win)
4049 /* synchronize sequencing */
4050 if (SEQ_GT(end, src->seqlo))
4052 /* slide the window of what the other end can send */
4053 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4054 dst->seqhi = ack + MAX((win << sws), 1);
4057 * Cannot set dst->seqhi here since this could be a shotgunned
4058 * SYN and not an already established connection.
4061 if (th->th_flags & TH_FIN)
4062 if (src->state < TCPS_CLOSING)
4063 src->state = TCPS_CLOSING;
4064 if (th->th_flags & TH_RST)
4065 src->state = dst->state = TCPS_TIME_WAIT;
4067 /* Fall through to PASS packet */
4069 } else if ((*state)->pickup_mode == PF_PICKUPS_HASHONLY ||
4070 ((*state)->pickup_mode == PF_PICKUPS_ENABLED &&
4071 ((*state)->sync_flags & PFSTATE_GOT_SYN_MASK) !=
4072 PFSTATE_GOT_SYN_MASK)) {
4074 * If pickup mode is hash only, do not fail on sequence checks.
4076 * If pickup mode is enabled and we did not see the SYN in
4077 * both direction, do not fail on sequence checks because
4078 * we do not have complete information on window scale.
4080 * Adjust expiration and fall through to PASS packet.
4081 * XXX Add a FIN check to reduce timeout?
4083 (*state)->expire = time_second;
4086 * Failure processing
4088 if ((*state)->dst.state == TCPS_SYN_SENT &&
4089 (*state)->src.state == TCPS_SYN_SENT) {
4090 /* Send RST for state mismatches during handshake */
4091 if (!(th->th_flags & TH_RST)) {
4092 u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
4094 if (th->th_flags & TH_SYN)
4096 if (th->th_flags & TH_FIN)
4098 pf_send_tcp((*state)->rule.ptr, pd->af,
4099 pd->dst, pd->src, th->th_dport,
4100 th->th_sport, ntohl(th->th_ack), ack,
4101 TH_RST|TH_ACK, 0, 0,
4102 (*state)->rule.ptr->return_ttl);
4107 } else if (pf_status.debug >= PF_DEBUG_MISC) {
4108 kprintf("pf: BAD state: ");
4109 pf_print_state(*state);
4110 pf_print_flags(th->th_flags);
4111 kprintf(" seq=%u ack=%u len=%u ackskew=%d pkts=%d:%d "
4112 "dir=%s,%s\n", seq, ack, pd->p_len, ackskew,
4113 (*state)->packets[0], (*state)->packets[1],
4114 direction == PF_IN ? "in" : "out",
4115 direction == (*state)->direction ? "fwd" : "rev");
4116 kprintf("pf: State failure on: %c %c %c %c | %c %c\n",
4117 SEQ_GEQ(src->seqhi, end) ? ' ' : '1',
4118 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ?
4120 (ackskew >= -MAXACKWINDOW) ? ' ' : '3',
4121 (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4',
4122 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5',
4123 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6');
4128 if (dst->scrub || src->scrub) {
4129 if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4130 src, dst, ©back))
4134 /* Any packets which have gotten here are to be passed */
4136 /* translate source/destination address, if necessary */
4137 if (STATE_TRANSLATE(*state)) {
4138 if (direction == PF_OUT)
4139 pf_change_ap(pd->src, &th->th_sport, pd->ip_sum,
4140 &th->th_sum, &(*state)->gwy.addr,
4141 (*state)->gwy.port, 0, pd->af);
4143 pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum,
4144 &th->th_sum, &(*state)->lan.addr,
4145 (*state)->lan.port, 0, pd->af);
4146 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4147 } else if (copyback) {
4148 /* Copyback sequence modulation or stateful scrub changes */
4149 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4156 pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif,
4157 struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4159 struct pf_state_peer *src, *dst;
4160 struct pf_state key;
4161 struct udphdr *uh = pd->hdr.udp;
4164 key.proto = IPPROTO_UDP;
4165 if (direction == PF_IN) {
4166 PF_ACPY(&key.ext.addr, pd->src, key.af);
4167 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4168 key.ext.port = uh->uh_sport;
4169 key.gwy.port = uh->uh_dport;
4171 PF_ACPY(&key.lan.addr, pd->src, key.af);
4172 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4173 key.lan.port = uh->uh_sport;
4174 key.ext.port = uh->uh_dport;
4179 if (direction == (*state)->direction) {
4180 src = &(*state)->src;
4181 dst = &(*state)->dst;
4183 src = &(*state)->dst;
4184 dst = &(*state)->src;
4188 if (src->state < PFUDPS_SINGLE)
4189 src->state = PFUDPS_SINGLE;
4190 if (dst->state == PFUDPS_SINGLE)
4191 dst->state = PFUDPS_MULTIPLE;
4193 /* update expire time */
4194 (*state)->expire = time_second;
4195 if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE)
4196 (*state)->timeout = PFTM_UDP_MULTIPLE;
4198 (*state)->timeout = PFTM_UDP_SINGLE;
4200 /* translate source/destination address, if necessary */
4201 if (STATE_TRANSLATE(*state)) {
4202 if (direction == PF_OUT)
4203 pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum,
4204 &uh->uh_sum, &(*state)->gwy.addr,
4205 (*state)->gwy.port, 1, pd->af);
4207 pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum,
4208 &uh->uh_sum, &(*state)->lan.addr,
4209 (*state)->lan.port, 1, pd->af);
4210 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
4217 pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif,
4218 struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4220 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
4221 u_int16_t icmpid = 0;
4222 u_int16_t *icmpsum = NULL;
4223 u_int8_t icmptype = 0;
4226 switch (pd->proto) {
4229 icmptype = pd->hdr.icmp->icmp_type;
4230 icmpid = pd->hdr.icmp->icmp_id;
4231 icmpsum = &pd->hdr.icmp->icmp_cksum;
4233 if (icmptype == ICMP_UNREACH ||
4234 icmptype == ICMP_SOURCEQUENCH ||
4235 icmptype == ICMP_REDIRECT ||
4236 icmptype == ICMP_TIMXCEED ||
4237 icmptype == ICMP_PARAMPROB)
4242 case IPPROTO_ICMPV6:
4243 icmptype = pd->hdr.icmp6->icmp6_type;
4244 icmpid = pd->hdr.icmp6->icmp6_id;
4245 icmpsum = &pd->hdr.icmp6->icmp6_cksum;
4247 if (icmptype == ICMP6_DST_UNREACH ||
4248 icmptype == ICMP6_PACKET_TOO_BIG ||
4249 icmptype == ICMP6_TIME_EXCEEDED ||
4250 icmptype == ICMP6_PARAM_PROB)
4259 * ICMP query/reply message not related to a TCP/UDP packet.
4260 * Search for an ICMP state.
4262 struct pf_state key;
4265 key.proto = pd->proto;
4266 if (direction == PF_IN) {
4267 PF_ACPY(&key.ext.addr, pd->src, key.af);
4268 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4269 key.ext.port = icmpid;
4270 key.gwy.port = icmpid;
4272 PF_ACPY(&key.lan.addr, pd->src, key.af);
4273 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4274 key.lan.port = icmpid;
4275 key.ext.port = icmpid;
4280 (*state)->expire = time_second;
4281 (*state)->timeout = PFTM_ICMP_ERROR_REPLY;
4283 /* translate source/destination address, if necessary */
4284 if (PF_ANEQ(&(*state)->lan.addr, &(*state)->gwy.addr, pd->af)) {
4285 if (direction == PF_OUT) {
4289 pf_change_a(&saddr->v4.s_addr,
4291 (*state)->gwy.addr.v4.s_addr, 0);
4297 &pd->hdr.icmp6->icmp6_cksum,
4298 &(*state)->gwy.addr, 0);
4300 sizeof(struct icmp6_hdr),
4301 (caddr_t)pd->hdr.icmp6);
4309 pf_change_a(&daddr->v4.s_addr,
4311 (*state)->lan.addr.v4.s_addr, 0);
4317 &pd->hdr.icmp6->icmp6_cksum,
4318 &(*state)->lan.addr, 0);
4320 sizeof(struct icmp6_hdr),
4321 (caddr_t)pd->hdr.icmp6);
4332 * ICMP error message in response to a TCP/UDP packet.
4333 * Extract the inner TCP/UDP header and search for that state.
4336 struct pf_pdesc pd2;
4341 struct ip6_hdr h2_6;
4351 /* offset of h2 in mbuf chain */
4352 ipoff2 = off + ICMP_MINLEN;
4354 if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2),
4355 NULL, NULL, pd2.af)) {
4356 DPFPRINTF(PF_DEBUG_MISC,
4357 ("pf: ICMP error message too short "
4362 * ICMP error messages don't refer to non-first
4366 * Note: We are dealing with an encapsulated
4367 * header. This means ip_off/ip_len are not
4368 * in host byte order!
4370 if (h2.ip_off & htons(IP_OFFMASK))
4373 /* offset of protocol header that follows h2 */
4374 off2 = ipoff2 + (h2.ip_hl << 2);
4376 pd2.proto = h2.ip_p;
4377 pd2.src = (struct pf_addr *)&h2.ip_src;
4378 pd2.dst = (struct pf_addr *)&h2.ip_dst;
4379 pd2.ip_sum = &h2.ip_sum;
4384 ipoff2 = off + sizeof(struct icmp6_hdr);
4386 if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6),
4387 NULL, NULL, pd2.af)) {
4388 DPFPRINTF(PF_DEBUG_MISC,
4389 ("pf: ICMP error message too short "
4393 pd2.proto = h2_6.ip6_nxt;
4394 pd2.src = (struct pf_addr *)&h2_6.ip6_src;
4395 pd2.dst = (struct pf_addr *)&h2_6.ip6_dst;
4397 off2 = ipoff2 + sizeof(h2_6);
4399 switch (pd2.proto) {
4400 case IPPROTO_FRAGMENT:
4402 * ICMPv6 error messages for
4403 * non-first fragments
4407 case IPPROTO_HOPOPTS:
4408 case IPPROTO_ROUTING:
4409 case IPPROTO_DSTOPTS: {
4410 /* get next header and header length */
4411 struct ip6_ext opt6;
4413 if (!pf_pull_hdr(m, off2, &opt6,
4414 sizeof(opt6), NULL, NULL, pd2.af)) {
4415 DPFPRINTF(PF_DEBUG_MISC,
4416 ("pf: ICMPv6 short opt\n"));
4419 if (pd2.proto == IPPROTO_AH)
4420 off2 += (opt6.ip6e_len + 2) * 4;
4422 off2 += (opt6.ip6e_len + 1) * 8;
4423 pd2.proto = opt6.ip6e_nxt;
4424 /* goto the next header */
4431 } while (!terminal);
4436 switch (pd2.proto) {
4440 struct pf_state key;
4441 struct pf_state_peer *src, *dst;
4446 * Only the first 8 bytes of the TCP header can be
4447 * expected. Don't access any TCP header fields after
4448 * th_seq, an ackskew test is not possible.
4450 if (!pf_pull_hdr(m, off2, &th, 8, NULL, NULL, pd2.af)) {
4451 DPFPRINTF(PF_DEBUG_MISC,
4452 ("pf: ICMP error message too short "
4458 key.proto = IPPROTO_TCP;
4459 if (direction == PF_IN) {
4460 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4461 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4462 key.ext.port = th.th_dport;
4463 key.gwy.port = th.th_sport;
4465 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4466 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4467 key.lan.port = th.th_dport;
4468 key.ext.port = th.th_sport;
4473 if (direction == (*state)->direction) {
4474 src = &(*state)->dst;
4475 dst = &(*state)->src;
4477 src = &(*state)->src;
4478 dst = &(*state)->dst;
4481 if (src->wscale && dst->wscale &&
4482 !(th.th_flags & TH_SYN))
4483 dws = dst->wscale & PF_WSCALE_MASK;
4487 /* Demodulate sequence number */
4488 seq = ntohl(th.th_seq) - src->seqdiff;
4490 pf_change_a(&th.th_seq, icmpsum,
4495 if (!SEQ_GEQ(src->seqhi, seq) ||
4496 !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws))) {
4497 if (pf_status.debug >= PF_DEBUG_MISC) {
4498 kprintf("pf: BAD ICMP %d:%d ",
4499 icmptype, pd->hdr.icmp->icmp_code);
4500 pf_print_host(pd->src, 0, pd->af);
4502 pf_print_host(pd->dst, 0, pd->af);
4503 kprintf(" state: ");
4504 pf_print_state(*state);
4505 kprintf(" seq=%u\n", seq);
4510 if (STATE_TRANSLATE(*state)) {
4511 if (direction == PF_IN) {
4512 pf_change_icmp(pd2.src, &th.th_sport,
4513 daddr, &(*state)->lan.addr,
4514 (*state)->lan.port, NULL,
4515 pd2.ip_sum, icmpsum,
4516 pd->ip_sum, 0, pd2.af);
4518 pf_change_icmp(pd2.dst, &th.th_dport,
4519 saddr, &(*state)->gwy.addr,
4520 (*state)->gwy.port, NULL,
4521 pd2.ip_sum, icmpsum,
4522 pd->ip_sum, 0, pd2.af);
4531 m_copyback(m, off, ICMP_MINLEN,
4532 (caddr_t)pd->hdr.icmp);
4533 m_copyback(m, ipoff2, sizeof(h2),
4540 sizeof(struct icmp6_hdr),
4541 (caddr_t)pd->hdr.icmp6);
4542 m_copyback(m, ipoff2, sizeof(h2_6),
4547 m_copyback(m, off2, 8, (caddr_t)&th);
4555 struct pf_state key;
4557 if (!pf_pull_hdr(m, off2, &uh, sizeof(uh),
4558 NULL, NULL, pd2.af)) {
4559 DPFPRINTF(PF_DEBUG_MISC,
4560 ("pf: ICMP error message too short "
4566 key.proto = IPPROTO_UDP;
4567 if (direction == PF_IN) {
4568 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4569 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4570 key.ext.port = uh.uh_dport;
4571 key.gwy.port = uh.uh_sport;
4573 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4574 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4575 key.lan.port = uh.uh_dport;
4576 key.ext.port = uh.uh_sport;
4581 if (STATE_TRANSLATE(*state)) {
4582 if (direction == PF_IN) {
4583 pf_change_icmp(pd2.src, &uh.uh_sport,
4584 daddr, &(*state)->lan.addr,
4585 (*state)->lan.port, &uh.uh_sum,
4586 pd2.ip_sum, icmpsum,
4587 pd->ip_sum, 1, pd2.af);
4589 pf_change_icmp(pd2.dst, &uh.uh_dport,
4590 saddr, &(*state)->gwy.addr,
4591 (*state)->gwy.port, &uh.uh_sum,
4592 pd2.ip_sum, icmpsum,
4593 pd->ip_sum, 1, pd2.af);
4598 m_copyback(m, off, ICMP_MINLEN,
4599 (caddr_t)pd->hdr.icmp);
4600 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4606 sizeof(struct icmp6_hdr),
4607 (caddr_t)pd->hdr.icmp6);
4608 m_copyback(m, ipoff2, sizeof(h2_6),
4613 m_copyback(m, off2, sizeof(uh), (caddr_t)&uh);
4620 case IPPROTO_ICMP: {
4622 struct pf_state key;
4624 if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN,
4625 NULL, NULL, pd2.af)) {
4626 DPFPRINTF(PF_DEBUG_MISC,
4627 ("pf: ICMP error message too short i"
4633 key.proto = IPPROTO_ICMP;
4634 if (direction == PF_IN) {
4635 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4636 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4637 key.ext.port = iih.icmp_id;
4638 key.gwy.port = iih.icmp_id;
4640 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4641 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4642 key.lan.port = iih.icmp_id;
4643 key.ext.port = iih.icmp_id;
4648 if (STATE_TRANSLATE(*state)) {
4649 if (direction == PF_IN) {
4650 pf_change_icmp(pd2.src, &iih.icmp_id,
4651 daddr, &(*state)->lan.addr,
4652 (*state)->lan.port, NULL,
4653 pd2.ip_sum, icmpsum,
4654 pd->ip_sum, 0, AF_INET);
4656 pf_change_icmp(pd2.dst, &iih.icmp_id,
4657 saddr, &(*state)->gwy.addr,
4658 (*state)->gwy.port, NULL,
4659 pd2.ip_sum, icmpsum,
4660 pd->ip_sum, 0, AF_INET);
4662 m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
4663 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4664 m_copyback(m, off2, ICMP_MINLEN, (caddr_t)&iih);
4672 case IPPROTO_ICMPV6: {
4673 struct icmp6_hdr iih;
4674 struct pf_state key;
4676 if (!pf_pull_hdr(m, off2, &iih,
4677 sizeof(struct icmp6_hdr), NULL, NULL, pd2.af)) {
4678 DPFPRINTF(PF_DEBUG_MISC,
4679 ("pf: ICMP error message too short "
4685 key.proto = IPPROTO_ICMPV6;
4686 if (direction == PF_IN) {
4687 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4688 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4689 key.ext.port = iih.icmp6_id;
4690 key.gwy.port = iih.icmp6_id;
4692 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4693 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4694 key.lan.port = iih.icmp6_id;
4695 key.ext.port = iih.icmp6_id;
4700 if (STATE_TRANSLATE(*state)) {
4701 if (direction == PF_IN) {
4702 pf_change_icmp(pd2.src, &iih.icmp6_id,
4703 daddr, &(*state)->lan.addr,
4704 (*state)->lan.port, NULL,
4705 pd2.ip_sum, icmpsum,
4706 pd->ip_sum, 0, AF_INET6);
4708 pf_change_icmp(pd2.dst, &iih.icmp6_id,
4709 saddr, &(*state)->gwy.addr,
4710 (*state)->gwy.port, NULL,
4711 pd2.ip_sum, icmpsum,
4712 pd->ip_sum, 0, AF_INET6);
4714 m_copyback(m, off, sizeof(struct icmp6_hdr),
4715 (caddr_t)pd->hdr.icmp6);
4716 m_copyback(m, ipoff2, sizeof(h2_6), (caddr_t)&h2_6);
4717 m_copyback(m, off2, sizeof(struct icmp6_hdr),
4726 struct pf_state key;
4729 key.proto = pd2.proto;
4730 if (direction == PF_IN) {
4731 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4732 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4736 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4737 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4744 if (STATE_TRANSLATE(*state)) {
4745 if (direction == PF_IN) {
4746 pf_change_icmp(pd2.src, NULL,
4747 daddr, &(*state)->lan.addr,
4749 pd2.ip_sum, icmpsum,
4750 pd->ip_sum, 0, pd2.af);
4752 pf_change_icmp(pd2.dst, NULL,
4753 saddr, &(*state)->gwy.addr,
4755 pd2.ip_sum, icmpsum,
4756 pd->ip_sum, 0, pd2.af);
4761 m_copyback(m, off, ICMP_MINLEN,
4762 (caddr_t)pd->hdr.icmp);
4763 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4769 sizeof(struct icmp6_hdr),
4770 (caddr_t)pd->hdr.icmp6);
4771 m_copyback(m, ipoff2, sizeof(h2_6),
4786 pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif,
4787 struct pf_pdesc *pd)
4789 struct pf_state_peer *src, *dst;
4790 struct pf_state key;
4793 key.proto = pd->proto;
4794 if (direction == PF_IN) {
4795 PF_ACPY(&key.ext.addr, pd->src, key.af);
4796 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4800 PF_ACPY(&key.lan.addr, pd->src, key.af);
4801 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4808 if (direction == (*state)->direction) {
4809 src = &(*state)->src;
4810 dst = &(*state)->dst;
4812 src = &(*state)->dst;
4813 dst = &(*state)->src;
4817 if (src->state < PFOTHERS_SINGLE)
4818 src->state = PFOTHERS_SINGLE;
4819 if (dst->state == PFOTHERS_SINGLE)
4820 dst->state = PFOTHERS_MULTIPLE;
4822 /* update expire time */
4823 (*state)->expire = time_second;
4824 if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE)
4825 (*state)->timeout = PFTM_OTHER_MULTIPLE;
4827 (*state)->timeout = PFTM_OTHER_SINGLE;
4829 /* translate source/destination address, if necessary */
4830 if (STATE_TRANSLATE(*state)) {
4831 if (direction == PF_OUT)
4835 pf_change_a(&pd->src->v4.s_addr,
4836 pd->ip_sum, (*state)->gwy.addr.v4.s_addr,
4842 PF_ACPY(pd->src, &(*state)->gwy.addr, pd->af);
4850 pf_change_a(&pd->dst->v4.s_addr,
4851 pd->ip_sum, (*state)->lan.addr.v4.s_addr,
4857 PF_ACPY(pd->dst, &(*state)->lan.addr, pd->af);
4867 * ipoff and off are measured from the start of the mbuf chain.
4868 * h must be at "ipoff" on the mbuf chain.
4871 pf_pull_hdr(struct mbuf *m, int off, void *p, int len,
4872 u_short *actionp, u_short *reasonp, sa_family_t af)
4877 struct ip *h = mtod(m, struct ip *);
4878 u_int16_t fragoff = (h->ip_off & IP_OFFMASK) << 3;
4882 ACTION_SET(actionp, PF_PASS);
4884 ACTION_SET(actionp, PF_DROP);
4885 REASON_SET(reasonp, PFRES_FRAG);
4889 if (m->m_pkthdr.len < off + len ||
4890 h->ip_len < off + len) {
4891 ACTION_SET(actionp, PF_DROP);
4892 REASON_SET(reasonp, PFRES_SHORT);
4900 struct ip6_hdr *h = mtod(m, struct ip6_hdr *);
4902 if (m->m_pkthdr.len < off + len ||
4903 (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) <
4904 (unsigned)(off + len)) {
4905 ACTION_SET(actionp, PF_DROP);
4906 REASON_SET(reasonp, PFRES_SHORT);
4913 m_copydata(m, off, len, p);
4918 pf_routable(struct pf_addr *addr, sa_family_t af)
4920 struct sockaddr_in *dst;
4924 bzero(&ro, sizeof(ro));
4925 dst = satosin(&ro.ro_dst);
4926 dst->sin_family = af;
4927 dst->sin_len = sizeof(*dst);
4928 dst->sin_addr = addr->v4;
4929 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
4931 if (ro.ro_rt != NULL) {
4941 pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
4944 struct mbuf *m0, *m1;
4945 struct route iproute;
4946 struct route *ro = NULL;
4947 struct sockaddr_in *dst;
4949 struct ifnet *ifp = NULL;
4950 struct pf_addr naddr;
4951 struct pf_src_node *sn = NULL;
4955 if (m == NULL || *m == NULL || r == NULL ||
4956 (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
4957 panic("pf_route: invalid parameters");
4959 if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
4960 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
4961 (*m)->m_pkthdr.pf_routed = 1;
4963 if ((*m)->m_pkthdr.pf_routed > 3) {
4968 (*m)->m_pkthdr.pf_routed++;
4971 if (r->rt == PF_DUPTO) {
4972 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL)
4975 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
4980 if (m0->m_len < sizeof(struct ip))
4981 panic("pf_route: m0->m_len < sizeof(struct ip)");
4982 ip = mtod(m0, struct ip *);
4985 bzero((caddr_t)ro, sizeof(*ro));
4986 dst = satosin(&ro->ro_dst);
4987 dst->sin_family = AF_INET;
4988 dst->sin_len = sizeof(*dst);
4989 dst->sin_addr = ip->ip_dst;
4991 if (r->rt == PF_FASTROUTE) {
4993 if (ro->ro_rt == 0) {
4994 ipstat.ips_noroute++;
4998 ifp = ro->ro_rt->rt_ifp;
4999 ro->ro_rt->rt_use++;
5001 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
5002 dst = satosin(ro->ro_rt->rt_gateway);
5004 if (TAILQ_EMPTY(&r->rpool.list))
5005 panic("pf_route: TAILQ_EMPTY(&r->rpool.list)");
5007 pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src,
5009 if (!PF_AZERO(&naddr, AF_INET))
5010 dst->sin_addr.s_addr = naddr.v4.s_addr;
5011 ifp = r->rpool.cur->kif ?
5012 r->rpool.cur->kif->pfik_ifp : NULL;
5014 if (!PF_AZERO(&s->rt_addr, AF_INET))
5015 dst->sin_addr.s_addr =
5016 s->rt_addr.v4.s_addr;
5017 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5024 if (pf_test(PF_OUT, ifp, &m0) != PF_PASS)
5026 else if (m0 == NULL)
5028 if (m0->m_len < sizeof(struct ip))
5029 panic("pf_route: m0->m_len < sizeof(struct ip)");
5030 ip = mtod(m0, struct ip *);
5033 /* Copied from ip_output. */
5034 m0->m_pkthdr.csum_flags |= CSUM_IP;
5035 sw_csum = m0->m_pkthdr.csum_flags & ~ifp->if_hwassist;
5036 if (sw_csum & CSUM_DELAY_DATA) {
5037 in_delayed_cksum(m0);
5038 sw_csum &= ~CSUM_DELAY_DATA;
5040 m0->m_pkthdr.csum_flags &= ifp->if_hwassist;
5043 * If small enough for interface, or the interface will take
5044 * care of the fragmentation for us, can just send directly.
5046 if (ip->ip_len <= ifp->if_mtu || ((ifp->if_hwassist & CSUM_FRAGMENT) &&
5047 (ip->ip_off & IP_DF) == 0)) {
5048 ip->ip_len = htons(ip->ip_len);
5049 ip->ip_off = htons(ip->ip_off);
5051 if (sw_csum & CSUM_DELAY_IP) {
5053 if (ip->ip_v == IPVERSION &&
5054 (ip->ip_hl << 2) == sizeof(*ip)) {
5055 ip->ip_sum = in_cksum_hdr(ip);
5057 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
5061 error = ifp->if_output(ifp, m0, sintosa(dst), ro->ro_rt);
5066 * Too large for interface; fragment if possible.
5067 * Must be able to put at least 8 bytes per fragment.
5069 if (ip->ip_off & IP_DF) {
5070 ipstat.ips_cantfrag++;
5071 if (r->rt != PF_DUPTO) {
5072 icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0,
5080 error = ip_fragment(ip, &m0, ifp->if_mtu, ifp->if_hwassist, sw_csum);
5084 for (m0 = m1; m0; m0 = m1) {
5088 error = ifp->if_output(ifp, m0, sintosa(dst), NULL);
5095 ipstat.ips_fragmented++;
5098 if (r->rt != PF_DUPTO)
5100 if (ro == &iproute && ro->ro_rt)
5112 pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5116 struct route_in6 ip6route;
5117 struct route_in6 *ro;
5118 struct sockaddr_in6 *dst;
5119 struct ip6_hdr *ip6;
5120 struct ifnet *ifp = NULL;
5121 struct pf_addr naddr;
5122 struct pf_src_node *sn = NULL;
5125 if (m == NULL || *m == NULL || r == NULL ||
5126 (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5127 panic("pf_route6: invalid parameters");
5129 if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
5130 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
5131 (*m)->m_pkthdr.pf_routed = 1;
5133 if ((*m)->m_pkthdr.pf_routed > 3) {
5138 (*m)->m_pkthdr.pf_routed++;
5141 if (r->rt == PF_DUPTO) {
5142 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL)
5145 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5150 if (m0->m_len < sizeof(struct ip6_hdr))
5151 panic("pf_route6: m0->m_len < sizeof(struct ip6_hdr)");
5152 ip6 = mtod(m0, struct ip6_hdr *);
5155 bzero((caddr_t)ro, sizeof(*ro));
5156 dst = (struct sockaddr_in6 *)&ro->ro_dst;
5157 dst->sin6_family = AF_INET6;
5158 dst->sin6_len = sizeof(*dst);
5159 dst->sin6_addr = ip6->ip6_dst;
5162 if (r->rt == PF_FASTROUTE) {
5163 m0->m_pkthdr.fw_flags |= PF_MBUF_GENERATED;
5164 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
5168 if (TAILQ_EMPTY(&r->rpool.list))
5169 panic("pf_route6: TAILQ_EMPTY(&r->rpool.list)");
5171 pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src,
5173 if (!PF_AZERO(&naddr, AF_INET6))
5174 PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5176 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL;
5178 if (!PF_AZERO(&s->rt_addr, AF_INET6))
5179 PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5180 &s->rt_addr, AF_INET6);
5181 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5187 if (pf_test6(PF_OUT, ifp, &m0) != PF_PASS)
5189 else if (m0 == NULL)
5191 if (m0->m_len < sizeof(struct ip6_hdr))
5192 panic("pf_route6: m0->m_len < sizeof(struct ip6_hdr)");
5193 ip6 = mtod(m0, struct ip6_hdr *);
5197 * If the packet is too large for the outgoing interface,
5198 * send back an icmp6 error.
5200 if (IN6_IS_ADDR_LINKLOCAL(&dst->sin6_addr))
5201 dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
5202 if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) {
5203 error = nd6_output(ifp, ifp, m0, dst, NULL);
5205 in6_ifstat_inc(ifp, ifs6_in_toobig);
5206 if (r->rt != PF_DUPTO)
5207 icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
5213 if (r->rt != PF_DUPTO)
5225 * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag
5226 * off is the offset where the protocol header starts
5227 * len is the total length of protocol header plus payload
5228 * returns 0 when the checksum is valid, otherwise returns 1.
5232 * FreeBSD supports cksum offload for the following drivers.
5233 * em(4), gx(4), lge(4), nge(4), ti(4), xl(4)
5234 * If we can make full use of it we would outperform ipfw/ipfilter in
5235 * very heavy traffic.
5236 * I have not tested 'cause I don't have NICs that supports cksum offload.
5237 * (There might be problems. Typical phenomena would be
5238 * 1. No route message for UDP packet.
5239 * 2. No connection acceptance from external hosts regardless of rule set.)
5242 pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p,
5249 if (off < sizeof(struct ip) || len < sizeof(struct udphdr))
5251 if (m->m_pkthdr.len < off + len)
5257 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
5258 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) {
5259 sum = m->m_pkthdr.csum_data;
5261 ip = mtod(m, struct ip *);
5262 sum = in_pseudo(ip->ip_src.s_addr,
5263 ip->ip_dst.s_addr, htonl((u_short)len +
5264 m->m_pkthdr.csum_data + p));
5272 case IPPROTO_ICMPV6:
5282 if (p == IPPROTO_ICMP) {
5287 sum = in_cksum(m, len);
5291 if (m->m_len < sizeof(struct ip))
5293 sum = in_cksum_range(m, p, off, len);
5295 m->m_pkthdr.csum_flags |=
5298 m->m_pkthdr.csum_data = 0xffff;
5304 if (m->m_len < sizeof(struct ip6_hdr))
5306 sum = in6_cksum(m, p, off, len);
5309 * IPv6 H/W cksum off-load not supported yet!
5312 * m->m_pkthdr.csum_flags |=
5313 * (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
5314 * m->m_pkthdr.csum_data = 0xffff;
5326 tcpstat.tcps_rcvbadsum++;
5329 udpstat.udps_badsum++;
5332 icmpstat.icps_checksum++;
5335 case IPPROTO_ICMPV6:
5336 icmp6stat.icp6s_checksum++;
5347 pf_test(int dir, struct ifnet *ifp, struct mbuf **m0)
5349 struct pfi_kif *kif;
5350 u_short action, reason = 0, log = 0;
5351 struct mbuf *m = *m0;
5352 struct ip *h = NULL;
5353 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr;
5354 struct pf_state *s = NULL;
5355 struct pf_ruleset *ruleset = NULL;
5357 int off, dirndx, pqid = 0;
5359 if (!pf_status.running || (m->m_pkthdr.fw_flags & PF_MBUF_GENERATED))
5362 kif = pfi_index2kif[ifp->if_index];
5367 if ((m->m_flags & M_PKTHDR) == 0)
5368 panic("non-M_PKTHDR is passed to pf_test");
5371 memset(&pd, 0, sizeof(pd));
5372 if (m->m_pkthdr.len < (int)sizeof(*h)) {
5374 REASON_SET(&reason, PFRES_SHORT);
5379 /* We do IP header normalization and packet reassembly here */
5380 if (pf_normalize_ip(m0, dir, kif, &reason) != PF_PASS) {
5385 h = mtod(m, struct ip *);
5387 off = h->ip_hl << 2;
5388 if (off < (int)sizeof(*h)) {
5390 REASON_SET(&reason, PFRES_SHORT);
5395 pd.src = (struct pf_addr *)&h->ip_src;
5396 pd.dst = (struct pf_addr *)&h->ip_dst;
5397 PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET);
5398 pd.ip_sum = &h->ip_sum;
5402 pd.tot_len = h->ip_len;
5404 /* handle fragments that didn't get reassembled by normalization */
5405 if (h->ip_off & (IP_MF | IP_OFFMASK)) {
5406 action = pf_test_fragment(&r, dir, kif, m, h,
5417 if (!pf_pull_hdr(m, off, &th, sizeof(th),
5418 &action, &reason, AF_INET)) {
5419 log = action != PF_PASS;
5422 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5423 h->ip_len - off, IPPROTO_TCP, AF_INET)) {
5427 pd.p_len = pd.tot_len - off - (th.th_off << 2);
5428 if ((th.th_flags & TH_ACK) && pd.p_len == 0)
5430 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
5431 if (action == PF_DROP)
5433 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
5435 if (action == PF_PASS) {
5437 pfsync_update_state(s);
5442 } else if (s == NULL)
5443 action = pf_test_tcp(&r, &s, dir, kif,
5444 m, off, h, &pd, &a, &ruleset);
5452 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
5453 &action, &reason, AF_INET)) {
5454 log = action != PF_PASS;
5457 if (dir == PF_IN && uh.uh_sum && pf_check_proto_cksum(m,
5458 off, h->ip_len - off, IPPROTO_UDP, AF_INET)) {
5462 if (uh.uh_dport == 0 ||
5463 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
5464 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
5468 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
5469 if (action == PF_PASS) {
5471 pfsync_update_state(s);
5476 } else if (s == NULL)
5477 action = pf_test_udp(&r, &s, dir, kif,
5478 m, off, h, &pd, &a, &ruleset);
5482 case IPPROTO_ICMP: {
5486 if (!pf_pull_hdr(m, off, &ih, ICMP_MINLEN,
5487 &action, &reason, AF_INET)) {
5488 log = action != PF_PASS;
5491 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5492 h->ip_len - off, IPPROTO_ICMP, AF_INET)) {
5496 action = pf_test_state_icmp(&s, dir, kif, m, off, h, &pd);
5497 if (action == PF_PASS) {
5499 pfsync_update_state(s);
5504 } else if (s == NULL)
5505 action = pf_test_icmp(&r, &s, dir, kif,
5506 m, off, h, &pd, &a, &ruleset);
5511 action = pf_test_state_other(&s, dir, kif, &pd);
5512 if (action == PF_PASS) {
5514 pfsync_update_state(s);
5519 } else if (s == NULL)
5520 action = pf_test_other(&r, &s, dir, kif, m, off, h,
5526 if (action == PF_PASS && h->ip_hl > 5 &&
5527 !((s && s->allow_opts) || r->allow_opts)) {
5529 REASON_SET(&reason, PFRES_SHORT);
5531 DPFPRINTF(PF_DEBUG_MISC,
5532 ("pf: dropping packet with ip options\n"));
5536 if (action == PF_PASS && r->qid) {
5537 m->m_pkthdr.fw_flags |= ALTQ_MBUF_TAGGED;
5538 if (pd.tos == IPTOS_LOWDELAY)
5539 m->m_pkthdr.altq_qid = r->pqid;
5541 m->m_pkthdr.altq_qid = r->qid;
5543 KKASSERT(s->hash != 0);
5544 m->m_pkthdr.fw_flags |= ALTQ_MBUF_STATE_HASHED;
5545 m->m_pkthdr.altq_state_hash = s->hash;
5547 m->m_pkthdr.ecn_af = AF_INET;
5548 m->m_pkthdr.header = h;
5553 * connections redirected to loopback should not match sockets
5554 * bound specifically to loopback due to security implications,
5555 * see tcp_input() and in_pcblookup_listen().
5557 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
5558 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
5559 (s->nat_rule.ptr->action == PF_RDR ||
5560 s->nat_rule.ptr->action == PF_BINAT) &&
5561 (ntohl(pd.dst->v4.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
5563 REASON_SET(&reason, PFRES_MEMORY);
5566 m->m_pkthdr.fw_flags |= PF_MBUF_TRANSLATE_LOCALHOST;
5569 PFLOG_PACKET(kif, h, m, AF_INET, dir, reason, r, a, ruleset);
5571 kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
5572 kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++;
5574 if (action == PF_PASS || r->action == PF_DROP) {
5576 r->bytes += pd.tot_len;
5579 a->bytes += pd.tot_len;
5582 dirndx = (dir == s->direction) ? 0 : 1;
5583 s->packets[dirndx]++;
5584 s->bytes[dirndx] += pd.tot_len;
5585 if (s->nat_rule.ptr != NULL) {
5586 s->nat_rule.ptr->packets++;
5587 s->nat_rule.ptr->bytes += pd.tot_len;
5589 if (s->src_node != NULL) {
5590 s->src_node->packets++;
5591 s->src_node->bytes += pd.tot_len;
5593 if (s->nat_src_node != NULL) {
5594 s->nat_src_node->packets++;
5595 s->nat_src_node->bytes += pd.tot_len;
5599 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
5603 * XXX: we need to make sure that the addresses
5604 * passed to pfr_update_stats() are the same than
5605 * the addresses used during matching (pfr_match)
5607 if (r == &pf_default_rule) {
5609 x = (s == NULL || s->direction == dir) ?
5610 &pd.baddr : &pd.naddr;
5612 x = (s == NULL || s->direction == dir) ?
5613 &pd.naddr : &pd.baddr;
5614 if (x == &pd.baddr || s == NULL) {
5615 /* we need to change the address */
5622 if (tr->src.addr.type == PF_ADDR_TABLE)
5623 pfr_update_stats(tr->src.addr.p.tbl, (s == NULL ||
5624 s->direction == dir) ? pd.src : pd.dst, pd.af,
5625 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5627 if (tr->dst.addr.type == PF_ADDR_TABLE)
5628 pfr_update_stats(tr->dst.addr.p.tbl, (s == NULL ||
5629 s->direction == dir) ? pd.dst : pd.src, pd.af,
5630 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5635 if (action == PF_SYNPROXY_DROP) {
5640 /* pf_route can free the mbuf causing *m0 to become NULL */
5641 pf_route(m0, r, dir, ifp, s);
5649 pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0)
5651 struct pfi_kif *kif;
5652 u_short action, reason = 0, log = 0;
5653 struct mbuf *m = *m0;
5654 struct ip6_hdr *h = NULL;
5655 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr;
5656 struct pf_state *s = NULL;
5657 struct pf_ruleset *ruleset = NULL;
5659 int off, terminal = 0, dirndx;
5661 if (!pf_status.running || (m->m_pkthdr.fw_flags & PF_MBUF_GENERATED))
5664 kif = pfi_index2kif[ifp->if_index];
5669 if ((m->m_flags & M_PKTHDR) == 0)
5670 panic("non-M_PKTHDR is passed to pf_test");
5673 memset(&pd, 0, sizeof(pd));
5674 if (m->m_pkthdr.len < (int)sizeof(*h)) {
5676 REASON_SET(&reason, PFRES_SHORT);
5681 /* We do IP header normalization and packet reassembly here */
5682 if (pf_normalize_ip6(m0, dir, kif, &reason) != PF_PASS) {
5687 h = mtod(m, struct ip6_hdr *);
5689 pd.src = (struct pf_addr *)&h->ip6_src;
5690 pd.dst = (struct pf_addr *)&h->ip6_dst;
5691 PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET6);
5695 pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr);
5697 off = ((caddr_t)h - m->m_data) + sizeof(struct ip6_hdr);
5698 pd.proto = h->ip6_nxt;
5701 case IPPROTO_FRAGMENT:
5702 action = pf_test_fragment(&r, dir, kif, m, h,
5704 if (action == PF_DROP)
5705 REASON_SET(&reason, PFRES_FRAG);
5708 case IPPROTO_HOPOPTS:
5709 case IPPROTO_ROUTING:
5710 case IPPROTO_DSTOPTS: {
5711 /* get next header and header length */
5712 struct ip6_ext opt6;
5714 if (!pf_pull_hdr(m, off, &opt6, sizeof(opt6),
5715 NULL, NULL, pd.af)) {
5716 DPFPRINTF(PF_DEBUG_MISC,
5717 ("pf: IPv6 short opt\n"));
5719 REASON_SET(&reason, PFRES_SHORT);
5723 if (pd.proto == IPPROTO_AH)
5724 off += (opt6.ip6e_len + 2) * 4;
5726 off += (opt6.ip6e_len + 1) * 8;
5727 pd.proto = opt6.ip6e_nxt;
5728 /* goto the next header */
5735 } while (!terminal);
5743 if (!pf_pull_hdr(m, off, &th, sizeof(th),
5744 &action, &reason, AF_INET6)) {
5745 log = action != PF_PASS;
5748 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5749 ntohs(h->ip6_plen), IPPROTO_TCP, AF_INET6)) {
5753 pd.p_len = pd.tot_len - off - (th.th_off << 2);
5754 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
5755 if (action == PF_DROP)
5757 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
5759 if (action == PF_PASS) {
5761 pfsync_update_state(s);
5766 } else if (s == NULL)
5767 action = pf_test_tcp(&r, &s, dir, kif,
5768 m, off, h, &pd, &a, &ruleset);
5776 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
5777 &action, &reason, AF_INET6)) {
5778 log = action != PF_PASS;
5781 if (dir == PF_IN && uh.uh_sum && pf_check_proto_cksum(m,
5782 off, ntohs(h->ip6_plen), IPPROTO_UDP, AF_INET6)) {
5786 if (uh.uh_dport == 0 ||
5787 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
5788 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
5792 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
5793 if (action == PF_PASS) {
5795 pfsync_update_state(s);
5800 } else if (s == NULL)
5801 action = pf_test_udp(&r, &s, dir, kif,
5802 m, off, h, &pd, &a, &ruleset);
5806 case IPPROTO_ICMPV6: {
5807 struct icmp6_hdr ih;
5810 if (!pf_pull_hdr(m, off, &ih, sizeof(ih),
5811 &action, &reason, AF_INET6)) {
5812 log = action != PF_PASS;
5815 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5816 ntohs(h->ip6_plen), IPPROTO_ICMPV6, AF_INET6)) {
5820 action = pf_test_state_icmp(&s, dir, kif,
5822 if (action == PF_PASS) {
5824 pfsync_update_state(s);
5829 } else if (s == NULL)
5830 action = pf_test_icmp(&r, &s, dir, kif,
5831 m, off, h, &pd, &a, &ruleset);
5836 action = pf_test_state_other(&s, dir, kif, &pd);
5837 if (action == PF_PASS) {
5841 } else if (s == NULL)
5842 action = pf_test_other(&r, &s, dir, kif, m, off, h,
5848 /* XXX handle IPv6 options, if not allowed. not implemented. */
5851 if (action == PF_PASS && r->qid) {
5852 m->m_pkthdr.fw_flags |= ALTQ_MBUF_TAGGED;
5853 if (pd.tos == IPTOS_LOWDELAY)
5854 m->m_pkthdr.altq_qid = r->pqid;
5856 m->m_pkthdr.altq_qid = r->qid;
5858 KKASSERT(s->hash != 0);
5859 m->m_pkthdr.fw_flags |= ALTQ_MBUF_STATE_HASHED;
5860 m->m_pkthdr.altq_state_hash = s->hash;
5862 m->m_pkthdr.ecn_af = AF_INET6;
5863 m->m_pkthdr.header = h;
5867 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
5868 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
5869 (s->nat_rule.ptr->action == PF_RDR ||
5870 s->nat_rule.ptr->action == PF_BINAT) &&
5871 IN6_IS_ADDR_LOOPBACK(&pd.dst->v6)) {
5873 REASON_SET(&reason, PFRES_MEMORY);
5876 m->m_pkthdr.fw_flags |= PF_MBUF_TRANSLATE_LOCALHOST;
5879 PFLOG_PACKET(kif, h, m, AF_INET6, dir, reason, r, a, ruleset);
5881 kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
5882 kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++;
5884 if (action == PF_PASS || r->action == PF_DROP) {
5886 r->bytes += pd.tot_len;
5889 a->bytes += pd.tot_len;
5892 dirndx = (dir == s->direction) ? 0 : 1;
5893 s->packets[dirndx]++;
5894 s->bytes[dirndx] += pd.tot_len;
5895 if (s->nat_rule.ptr != NULL) {
5896 s->nat_rule.ptr->packets++;
5897 s->nat_rule.ptr->bytes += pd.tot_len;
5899 if (s->src_node != NULL) {
5900 s->src_node->packets++;
5901 s->src_node->bytes += pd.tot_len;
5903 if (s->nat_src_node != NULL) {
5904 s->nat_src_node->packets++;
5905 s->nat_src_node->bytes += pd.tot_len;
5909 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
5913 * XXX: we need to make sure that the addresses
5914 * passed to pfr_update_stats() are the same than
5915 * the addresses used during matching (pfr_match)
5917 if (r == &pf_default_rule) {
5919 x = (s == NULL || s->direction == dir) ?
5920 &pd.baddr : &pd.naddr;
5922 x = (s == NULL || s->direction == dir) ?
5923 &pd.naddr : &pd.baddr;
5925 if (x == &pd.baddr || s == NULL) {
5932 if (tr->src.addr.type == PF_ADDR_TABLE)
5933 pfr_update_stats(tr->src.addr.p.tbl, (s == NULL ||
5934 s->direction == dir) ? pd.src : pd.dst, pd.af,
5935 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5937 if (tr->dst.addr.type == PF_ADDR_TABLE)
5938 pfr_update_stats(tr->dst.addr.p.tbl, (s == NULL ||
5939 s->direction == dir) ? pd.dst : pd.src, pd.af,
5940 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5945 if (action == PF_SYNPROXY_DROP) {
5950 /* pf_route6 can free the mbuf causing *m0 to become NULL */
5951 pf_route6(m0, r, dir, ifp, s);