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;
2494 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
2496 if (direction == PF_OUT) {
2497 bport = nport = th->th_sport;
2498 /* check outgoing packet for BINAT/NAT */
2499 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
2500 saddr, th->th_sport, daddr, th->th_dport,
2501 &pd->naddr, &nport)) != NULL) {
2502 PF_ACPY(&pd->baddr, saddr, af);
2503 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
2504 &th->th_sum, &pd->naddr, nport, 0, af);
2511 bport = nport = th->th_dport;
2512 /* check incoming packet for BINAT/RDR */
2513 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
2514 saddr, th->th_sport, daddr, th->th_dport,
2515 &pd->naddr, &nport)) != NULL) {
2516 PF_ACPY(&pd->baddr, daddr, af);
2517 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
2518 &th->th_sum, &pd->naddr, nport, 0, af);
2528 if (r->kif != NULL &&
2529 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
2530 r = r->skip[PF_SKIP_IFP].ptr;
2531 else if (r->direction && r->direction != direction)
2532 r = r->skip[PF_SKIP_DIR].ptr;
2533 else if (r->af && r->af != af)
2534 r = r->skip[PF_SKIP_AF].ptr;
2535 else if (r->proto && r->proto != IPPROTO_TCP)
2536 r = r->skip[PF_SKIP_PROTO].ptr;
2537 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
2538 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
2539 else if (r->src.port_op && !pf_match_port(r->src.port_op,
2540 r->src.port[0], r->src.port[1], th->th_sport))
2541 r = r->skip[PF_SKIP_SRC_PORT].ptr;
2542 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
2543 r = r->skip[PF_SKIP_DST_ADDR].ptr;
2544 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
2545 r->dst.port[0], r->dst.port[1], th->th_dport))
2546 r = r->skip[PF_SKIP_DST_PORT].ptr;
2547 else if (r->tos && !(r->tos & pd->tos))
2548 r = TAILQ_NEXT(r, entries);
2549 else if (r->rule_flag & PFRULE_FRAGMENT)
2550 r = TAILQ_NEXT(r, entries);
2551 else if ((r->flagset & th->th_flags) != r->flags)
2552 r = TAILQ_NEXT(r, entries);
2553 else if (r->uid.op && (lookup != -1 || (lookup =
2554 pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2555 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
2557 r = TAILQ_NEXT(r, entries);
2558 else if (r->gid.op && (lookup != -1 || (lookup =
2559 pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2560 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
2562 r = TAILQ_NEXT(r, entries);
2563 else if (r->prob && r->prob <= karc4random())
2564 r = TAILQ_NEXT(r, entries);
2565 else if (r->match_tag && !pf_match_tag(m, r, nr, &tag))
2566 r = TAILQ_NEXT(r, entries);
2567 else if (r->anchorname[0] && r->anchor == NULL)
2568 r = TAILQ_NEXT(r, entries);
2569 else if (r->os_fingerprint != PF_OSFP_ANY && !pf_osfp_match(
2570 pf_osfp_fingerprint(pd, m, off, th), r->os_fingerprint))
2571 r = TAILQ_NEXT(r, entries);
2575 if (r->anchor == NULL) {
2581 r = TAILQ_NEXT(r, entries);
2583 PF_STEP_INTO_ANCHOR(r, a, ruleset,
2586 if (r == NULL && a != NULL)
2587 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
2594 REASON_SET(&reason, PFRES_MATCH);
2598 m_copyback(m, off, sizeof(*th), (caddr_t)th);
2599 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
2602 if ((r->action == PF_DROP) &&
2603 ((r->rule_flag & PFRULE_RETURNRST) ||
2604 (r->rule_flag & PFRULE_RETURNICMP) ||
2605 (r->rule_flag & PFRULE_RETURN))) {
2606 /* undo NAT changes, if they have taken place */
2608 if (direction == PF_OUT) {
2609 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
2610 &th->th_sum, &pd->baddr, bport, 0, af);
2613 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
2614 &th->th_sum, &pd->baddr, bport, 0, af);
2618 if (((r->rule_flag & PFRULE_RETURNRST) ||
2619 (r->rule_flag & PFRULE_RETURN)) &&
2620 !(th->th_flags & TH_RST)) {
2621 u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
2623 if (th->th_flags & TH_SYN)
2625 if (th->th_flags & TH_FIN)
2627 pf_send_tcp(r, af, pd->dst,
2628 pd->src, th->th_dport, th->th_sport,
2629 ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0,
2631 } else if ((af == AF_INET) && r->return_icmp)
2632 pf_send_icmp(m, r->return_icmp >> 8,
2633 r->return_icmp & 255, af, r);
2634 else if ((af == AF_INET6) && r->return_icmp6)
2635 pf_send_icmp(m, r->return_icmp6 >> 8,
2636 r->return_icmp6 & 255, af, r);
2639 if (r->action == PF_DROP)
2642 pf_tag_packet(m, tag);
2644 if (r->keep_state || nr != NULL ||
2645 (pd->flags & PFDESC_TCP_NORM)) {
2646 /* create new state */
2648 struct pf_state *s = NULL;
2649 struct pf_src_node *sn = NULL;
2651 len = pd->tot_len - off - (th->th_off << 2);
2653 /* check maximums */
2654 if (r->max_states && (r->states >= r->max_states))
2656 /* src node for flter rule */
2657 if ((r->rule_flag & PFRULE_SRCTRACK ||
2658 r->rpool.opts & PF_POOL_STICKYADDR) &&
2659 pf_insert_src_node(&sn, r, saddr, af) != 0)
2661 /* src node for translation rule */
2662 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
2663 ((direction == PF_OUT &&
2664 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
2665 (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
2667 s = pool_get(&pf_state_pl, PR_NOWAIT);
2670 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
2671 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
2672 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
2673 pf_status.src_nodes--;
2674 pool_put(&pf_src_tree_pl, sn);
2676 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
2678 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
2679 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
2680 pf_status.src_nodes--;
2681 pool_put(&pf_src_tree_pl, nsn);
2683 REASON_SET(&reason, PFRES_MEMORY);
2686 bzero(s, sizeof(*s));
2691 s->nat_rule.ptr = nr;
2692 if (s->nat_rule.ptr != NULL)
2693 s->nat_rule.ptr->states++;
2695 s->allow_opts = r->allow_opts;
2696 s->log = r->log & 2;
2697 s->proto = IPPROTO_TCP;
2698 s->direction = direction;
2700 if (direction == PF_OUT) {
2701 PF_ACPY(&s->gwy.addr, saddr, af);
2702 s->gwy.port = th->th_sport; /* sport */
2703 PF_ACPY(&s->ext.addr, daddr, af);
2704 s->ext.port = th->th_dport;
2706 PF_ACPY(&s->lan.addr, &pd->baddr, af);
2707 s->lan.port = bport;
2709 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
2710 s->lan.port = s->gwy.port;
2713 PF_ACPY(&s->lan.addr, daddr, af);
2714 s->lan.port = th->th_dport;
2715 PF_ACPY(&s->ext.addr, saddr, af);
2716 s->ext.port = th->th_sport;
2718 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
2719 s->gwy.port = bport;
2721 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
2722 s->gwy.port = s->lan.port;
2726 s->hash = pf_state_hash(s);
2727 s->src.seqlo = ntohl(th->th_seq);
2728 s->src.seqhi = s->src.seqlo + len + 1;
2729 s->pickup_mode = r->pickup_mode;
2731 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
2732 r->keep_state == PF_STATE_MODULATE) {
2733 /* Generate sequence number modulator */
2734 while ((s->src.seqdiff = karc4random()) == 0)
2736 pf_change_a(&th->th_seq, &th->th_sum,
2737 htonl(s->src.seqlo + s->src.seqdiff), 0);
2743 * WARNING! NetBSD patched this to not scale max_win up
2744 * on the initial SYN, but they failed to correct the code
2745 * in pf_test_state_tcp() that 'undid' the scaling, and they
2746 * failed to remove the scale factor on successful window
2747 * scale negotiation (and doing so would be difficult in the
2748 * face of retransmission, without adding more flags to the
2751 * After discussions with Daniel Hartmeier and Max Laier
2752 * I've decided not to apply the NetBSD patch.
2754 * The worst that happens is that the undo code on window
2755 * scale negotiation failures will produce a larger
2756 * max_win then actual.
2758 if (th->th_flags & TH_SYN) {
2760 s->src.wscale = pf_get_wscale(m, off, th->th_off, af);
2761 s->sync_flags |= PFSTATE_GOT_SYN1;
2763 s->src.max_win = MAX(ntohs(th->th_win), 1);
2764 if (s->src.wscale & PF_WSCALE_MASK) {
2765 /* Remove scale factor from initial window */
2766 u_int win = s->src.max_win;
2767 win += 1 << (s->src.wscale & PF_WSCALE_MASK);
2768 s->src.max_win = (win - 1) >>
2769 (s->src.wscale & PF_WSCALE_MASK);
2771 if (th->th_flags & TH_FIN)
2775 s->src.state = TCPS_SYN_SENT;
2776 s->dst.state = TCPS_CLOSED;
2777 s->creation = time_second;
2778 s->expire = time_second;
2779 s->timeout = PFTM_TCP_FIRST_PACKET;
2780 pf_set_rt_ifp(s, saddr);
2783 s->src_node->states++;
2786 PF_ACPY(&nsn->raddr, &pd->naddr, af);
2787 s->nat_src_node = nsn;
2788 s->nat_src_node->states++;
2790 if ((pd->flags & PFDESC_TCP_NORM) && pf_normalize_tcp_init(m,
2791 off, pd, th, &s->src, &s->dst)) {
2792 REASON_SET(&reason, PFRES_MEMORY);
2793 pf_src_tree_remove_state(s);
2794 pool_put(&pf_state_pl, s);
2797 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
2798 pf_normalize_tcp_stateful(m, off, pd, &reason, th, &s->src,
2799 &s->dst, &rewrite)) {
2800 pf_normalize_tcp_cleanup(s);
2801 pf_src_tree_remove_state(s);
2802 pool_put(&pf_state_pl, s);
2805 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
2806 pf_normalize_tcp_cleanup(s);
2807 REASON_SET(&reason, PFRES_MEMORY);
2808 pf_src_tree_remove_state(s);
2809 pool_put(&pf_state_pl, s);
2813 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
2814 r->keep_state == PF_STATE_SYNPROXY) {
2817 s->src.state = PF_TCPS_PROXY_SRC;
2819 if (direction == PF_OUT) {
2820 pf_change_ap(saddr, &th->th_sport,
2821 pd->ip_sum, &th->th_sum, &pd->baddr,
2824 pf_change_ap(daddr, &th->th_dport,
2825 pd->ip_sum, &th->th_sum, &pd->baddr,
2829 s->src.seqhi = karc4random();
2830 /* Find mss option */
2831 mss = pf_get_mss(m, off, th->th_off, af);
2832 mss = pf_calc_mss(saddr, af, mss);
2833 mss = pf_calc_mss(daddr, af, mss);
2835 pf_send_tcp(r, af, daddr, saddr, th->th_dport,
2836 th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1,
2837 TH_SYN|TH_ACK, 0, s->src.mss, 0);
2838 return (PF_SYNPROXY_DROP);
2842 /* copy back packet headers if we performed NAT operations */
2844 m_copyback(m, off, sizeof(*th), (caddr_t)th);
2850 pf_test_udp(struct pf_rule **rm, struct pf_state **sm, int direction,
2851 struct pfi_kif *kif, struct mbuf *m, int off, void *h,
2852 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
2854 struct pf_rule *nr = NULL;
2855 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
2856 struct udphdr *uh = pd->hdr.udp;
2857 u_int16_t bport, nport = 0;
2858 sa_family_t af = pd->af;
2862 struct pf_rule *r, *a = NULL;
2863 struct pf_ruleset *ruleset = NULL;
2864 struct pf_src_node *nsn = NULL;
2869 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
2871 if (direction == PF_OUT) {
2872 bport = nport = uh->uh_sport;
2873 /* check outgoing packet for BINAT/NAT */
2874 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
2875 saddr, uh->uh_sport, daddr, uh->uh_dport,
2876 &pd->naddr, &nport)) != NULL) {
2877 PF_ACPY(&pd->baddr, saddr, af);
2878 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum,
2879 &uh->uh_sum, &pd->naddr, nport, 1, af);
2886 bport = nport = uh->uh_dport;
2887 /* check incoming packet for BINAT/RDR */
2888 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
2889 saddr, uh->uh_sport, daddr, uh->uh_dport, &pd->naddr,
2891 PF_ACPY(&pd->baddr, daddr, af);
2892 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum,
2893 &uh->uh_sum, &pd->naddr, nport, 1, af);
2903 if (r->kif != NULL &&
2904 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
2905 r = r->skip[PF_SKIP_IFP].ptr;
2906 else if (r->direction && r->direction != direction)
2907 r = r->skip[PF_SKIP_DIR].ptr;
2908 else if (r->af && r->af != af)
2909 r = r->skip[PF_SKIP_AF].ptr;
2910 else if (r->proto && r->proto != IPPROTO_UDP)
2911 r = r->skip[PF_SKIP_PROTO].ptr;
2912 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
2913 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
2914 else if (r->src.port_op && !pf_match_port(r->src.port_op,
2915 r->src.port[0], r->src.port[1], uh->uh_sport))
2916 r = r->skip[PF_SKIP_SRC_PORT].ptr;
2917 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
2918 r = r->skip[PF_SKIP_DST_ADDR].ptr;
2919 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
2920 r->dst.port[0], r->dst.port[1], uh->uh_dport))
2921 r = r->skip[PF_SKIP_DST_PORT].ptr;
2922 else if (r->tos && !(r->tos & pd->tos))
2923 r = TAILQ_NEXT(r, entries);
2924 else if (r->rule_flag & PFRULE_FRAGMENT)
2925 r = TAILQ_NEXT(r, entries);
2926 else if (r->uid.op && (lookup != -1 || (lookup =
2927 pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2928 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
2930 r = TAILQ_NEXT(r, entries);
2931 else if (r->gid.op && (lookup != -1 || (lookup =
2932 pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2933 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
2935 r = TAILQ_NEXT(r, entries);
2936 else if (r->prob && r->prob <= karc4random())
2937 r = TAILQ_NEXT(r, entries);
2938 else if (r->match_tag && !pf_match_tag(m, r, nr, &tag))
2939 r = TAILQ_NEXT(r, entries);
2940 else if (r->anchorname[0] && r->anchor == NULL)
2941 r = TAILQ_NEXT(r, entries);
2942 else if (r->os_fingerprint != PF_OSFP_ANY)
2943 r = TAILQ_NEXT(r, entries);
2947 if (r->anchor == NULL) {
2953 r = TAILQ_NEXT(r, entries);
2955 PF_STEP_INTO_ANCHOR(r, a, ruleset,
2958 if (r == NULL && a != NULL)
2959 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
2966 REASON_SET(&reason, PFRES_MATCH);
2970 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
2971 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
2974 if ((r->action == PF_DROP) &&
2975 ((r->rule_flag & PFRULE_RETURNICMP) ||
2976 (r->rule_flag & PFRULE_RETURN))) {
2977 /* undo NAT changes, if they have taken place */
2979 if (direction == PF_OUT) {
2980 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum,
2981 &uh->uh_sum, &pd->baddr, bport, 1, af);
2984 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum,
2985 &uh->uh_sum, &pd->baddr, bport, 1, af);
2989 if ((af == AF_INET) && r->return_icmp)
2990 pf_send_icmp(m, r->return_icmp >> 8,
2991 r->return_icmp & 255, af, r);
2992 else if ((af == AF_INET6) && r->return_icmp6)
2993 pf_send_icmp(m, r->return_icmp6 >> 8,
2994 r->return_icmp6 & 255, af, r);
2997 if (r->action == PF_DROP)
3000 pf_tag_packet(m, tag);
3002 if (r->keep_state || nr != NULL) {
3003 /* create new state */
3004 struct pf_state *s = NULL;
3005 struct pf_src_node *sn = NULL;
3007 /* check maximums */
3008 if (r->max_states && (r->states >= r->max_states))
3010 /* src node for flter rule */
3011 if ((r->rule_flag & PFRULE_SRCTRACK ||
3012 r->rpool.opts & PF_POOL_STICKYADDR) &&
3013 pf_insert_src_node(&sn, r, saddr, af) != 0)
3015 /* src node for translation rule */
3016 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3017 ((direction == PF_OUT &&
3018 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3019 (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
3021 s = pool_get(&pf_state_pl, PR_NOWAIT);
3024 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3025 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3026 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3027 pf_status.src_nodes--;
3028 pool_put(&pf_src_tree_pl, sn);
3030 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3032 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3033 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3034 pf_status.src_nodes--;
3035 pool_put(&pf_src_tree_pl, nsn);
3037 REASON_SET(&reason, PFRES_MEMORY);
3040 bzero(s, sizeof(*s));
3045 s->nat_rule.ptr = nr;
3046 if (s->nat_rule.ptr != NULL)
3047 s->nat_rule.ptr->states++;
3049 s->allow_opts = r->allow_opts;
3050 s->log = r->log & 2;
3051 s->proto = IPPROTO_UDP;
3052 s->direction = direction;
3054 if (direction == PF_OUT) {
3055 PF_ACPY(&s->gwy.addr, saddr, af);
3056 s->gwy.port = uh->uh_sport;
3057 PF_ACPY(&s->ext.addr, daddr, af);
3058 s->ext.port = uh->uh_dport;
3060 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3061 s->lan.port = bport;
3063 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3064 s->lan.port = s->gwy.port;
3067 PF_ACPY(&s->lan.addr, daddr, af);
3068 s->lan.port = uh->uh_dport;
3069 PF_ACPY(&s->ext.addr, saddr, af);
3070 s->ext.port = uh->uh_sport;
3072 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3073 s->gwy.port = bport;
3075 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3076 s->gwy.port = s->lan.port;
3079 s->hash = pf_state_hash(s);
3080 s->src.state = PFUDPS_SINGLE;
3081 s->dst.state = PFUDPS_NO_TRAFFIC;
3082 s->creation = time_second;
3083 s->expire = time_second;
3084 s->timeout = PFTM_UDP_FIRST_PACKET;
3085 pf_set_rt_ifp(s, saddr);
3088 s->src_node->states++;
3091 PF_ACPY(&nsn->raddr, &pd->naddr, af);
3092 s->nat_src_node = nsn;
3093 s->nat_src_node->states++;
3095 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3096 REASON_SET(&reason, PFRES_MEMORY);
3097 pf_src_tree_remove_state(s);
3098 pool_put(&pf_state_pl, s);
3104 /* copy back packet headers if we performed NAT operations */
3106 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
3112 pf_test_icmp(struct pf_rule **rm, struct pf_state **sm, int direction,
3113 struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3114 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
3116 struct pf_rule *nr = NULL;
3117 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
3118 struct pf_rule *r, *a = NULL;
3119 struct pf_ruleset *ruleset = NULL;
3120 struct pf_src_node *nsn = NULL;
3122 u_int16_t icmpid = 0;
3123 sa_family_t af = pd->af;
3124 u_int8_t icmptype = 0, icmpcode = 0;
3131 switch (pd->proto) {
3134 icmptype = pd->hdr.icmp->icmp_type;
3135 icmpcode = pd->hdr.icmp->icmp_code;
3136 icmpid = pd->hdr.icmp->icmp_id;
3138 if (icmptype == ICMP_UNREACH ||
3139 icmptype == ICMP_SOURCEQUENCH ||
3140 icmptype == ICMP_REDIRECT ||
3141 icmptype == ICMP_TIMXCEED ||
3142 icmptype == ICMP_PARAMPROB)
3147 case IPPROTO_ICMPV6:
3148 icmptype = pd->hdr.icmp6->icmp6_type;
3149 icmpcode = pd->hdr.icmp6->icmp6_code;
3150 icmpid = pd->hdr.icmp6->icmp6_id;
3152 if (icmptype == ICMP6_DST_UNREACH ||
3153 icmptype == ICMP6_PACKET_TOO_BIG ||
3154 icmptype == ICMP6_TIME_EXCEEDED ||
3155 icmptype == ICMP6_PARAM_PROB)
3161 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3163 if (direction == PF_OUT) {
3164 /* check outgoing packet for BINAT/NAT */
3165 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3166 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3167 PF_ACPY(&pd->baddr, saddr, af);
3171 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3172 pd->naddr.v4.s_addr, 0);
3177 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum,
3188 /* check incoming packet for BINAT/RDR */
3189 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3190 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3191 PF_ACPY(&pd->baddr, daddr, af);
3195 pf_change_a(&daddr->v4.s_addr,
3196 pd->ip_sum, pd->naddr.v4.s_addr, 0);
3201 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum,
3215 if (r->kif != NULL &&
3216 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3217 r = r->skip[PF_SKIP_IFP].ptr;
3218 else if (r->direction && r->direction != direction)
3219 r = r->skip[PF_SKIP_DIR].ptr;
3220 else if (r->af && r->af != af)
3221 r = r->skip[PF_SKIP_AF].ptr;
3222 else if (r->proto && r->proto != pd->proto)
3223 r = r->skip[PF_SKIP_PROTO].ptr;
3224 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
3225 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3226 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
3227 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3228 else if (r->type && r->type != icmptype + 1)
3229 r = TAILQ_NEXT(r, entries);
3230 else if (r->code && r->code != icmpcode + 1)
3231 r = TAILQ_NEXT(r, entries);
3232 else if (r->tos && !(r->tos & pd->tos))
3233 r = TAILQ_NEXT(r, entries);
3234 else if (r->rule_flag & PFRULE_FRAGMENT)
3235 r = TAILQ_NEXT(r, entries);
3236 else if (r->prob && r->prob <= karc4random())
3237 r = TAILQ_NEXT(r, entries);
3238 else if (r->match_tag && !pf_match_tag(m, r, nr, &tag))
3239 r = TAILQ_NEXT(r, entries);
3240 else if (r->anchorname[0] && r->anchor == NULL)
3241 r = TAILQ_NEXT(r, entries);
3242 else if (r->os_fingerprint != PF_OSFP_ANY)
3243 r = TAILQ_NEXT(r, entries);
3247 if (r->anchor == NULL) {
3253 r = TAILQ_NEXT(r, entries);
3255 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3258 if (r == NULL && a != NULL)
3259 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3266 REASON_SET(&reason, PFRES_MATCH);
3271 m_copyback(m, off, sizeof(struct icmp6_hdr),
3272 (caddr_t)pd->hdr.icmp6);
3274 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3277 if (r->action != PF_PASS)
3280 pf_tag_packet(m, tag);
3282 if (!state_icmp && (r->keep_state || nr != NULL)) {
3283 /* create new state */
3284 struct pf_state *s = NULL;
3285 struct pf_src_node *sn = NULL;
3287 /* check maximums */
3288 if (r->max_states && (r->states >= r->max_states))
3290 /* src node for flter rule */
3291 if ((r->rule_flag & PFRULE_SRCTRACK ||
3292 r->rpool.opts & PF_POOL_STICKYADDR) &&
3293 pf_insert_src_node(&sn, r, saddr, af) != 0)
3295 /* src node for translation rule */
3296 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3297 ((direction == PF_OUT &&
3298 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3299 (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
3301 s = pool_get(&pf_state_pl, PR_NOWAIT);
3304 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3305 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3306 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3307 pf_status.src_nodes--;
3308 pool_put(&pf_src_tree_pl, sn);
3310 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3312 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3313 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3314 pf_status.src_nodes--;
3315 pool_put(&pf_src_tree_pl, nsn);
3317 REASON_SET(&reason, PFRES_MEMORY);
3320 bzero(s, sizeof(*s));
3325 s->nat_rule.ptr = nr;
3326 if (s->nat_rule.ptr != NULL)
3327 s->nat_rule.ptr->states++;
3329 s->allow_opts = r->allow_opts;
3330 s->log = r->log & 2;
3331 s->proto = pd->proto;
3332 s->direction = direction;
3334 if (direction == PF_OUT) {
3335 PF_ACPY(&s->gwy.addr, saddr, af);
3336 s->gwy.port = icmpid;
3337 PF_ACPY(&s->ext.addr, daddr, af);
3338 s->ext.port = icmpid;
3340 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3342 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3343 s->lan.port = icmpid;
3345 PF_ACPY(&s->lan.addr, daddr, af);
3346 s->lan.port = icmpid;
3347 PF_ACPY(&s->ext.addr, saddr, af);
3348 s->ext.port = icmpid;
3350 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3352 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3353 s->gwy.port = icmpid;
3355 s->hash = pf_state_hash(s);
3356 s->creation = time_second;
3357 s->expire = time_second;
3358 s->timeout = PFTM_ICMP_FIRST_PACKET;
3359 pf_set_rt_ifp(s, saddr);
3362 s->src_node->states++;
3365 PF_ACPY(&nsn->raddr, &pd->naddr, af);
3366 s->nat_src_node = nsn;
3367 s->nat_src_node->states++;
3369 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3370 REASON_SET(&reason, PFRES_MEMORY);
3371 pf_src_tree_remove_state(s);
3372 pool_put(&pf_state_pl, s);
3379 /* copy back packet headers if we performed IPv6 NAT operations */
3381 m_copyback(m, off, sizeof(struct icmp6_hdr),
3382 (caddr_t)pd->hdr.icmp6);
3389 pf_test_other(struct pf_rule **rm, struct pf_state **sm, int direction,
3390 struct pfi_kif *kif, struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
3391 struct pf_rule **am, struct pf_ruleset **rsm)
3393 struct pf_rule *nr = NULL;
3394 struct pf_rule *r, *a = NULL;
3395 struct pf_ruleset *ruleset = NULL;
3396 struct pf_src_node *nsn = NULL;
3397 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
3398 sa_family_t af = pd->af;
3402 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3404 if (direction == PF_OUT) {
3405 /* check outgoing packet for BINAT/NAT */
3406 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3407 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3408 PF_ACPY(&pd->baddr, saddr, af);
3412 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3413 pd->naddr.v4.s_addr, 0);
3418 PF_ACPY(saddr, &pd->naddr, af);
3427 /* check incoming packet for BINAT/RDR */
3428 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3429 saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3430 PF_ACPY(&pd->baddr, daddr, af);
3434 pf_change_a(&daddr->v4.s_addr,
3435 pd->ip_sum, pd->naddr.v4.s_addr, 0);
3440 PF_ACPY(daddr, &pd->naddr, af);
3452 if (r->kif != NULL &&
3453 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3454 r = r->skip[PF_SKIP_IFP].ptr;
3455 else if (r->direction && r->direction != direction)
3456 r = r->skip[PF_SKIP_DIR].ptr;
3457 else if (r->af && r->af != af)
3458 r = r->skip[PF_SKIP_AF].ptr;
3459 else if (r->proto && r->proto != pd->proto)
3460 r = r->skip[PF_SKIP_PROTO].ptr;
3461 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, r->src.not))
3462 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3463 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, r->dst.not))
3464 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3465 else if (r->tos && !(r->tos & pd->tos))
3466 r = TAILQ_NEXT(r, entries);
3467 else if (r->rule_flag & PFRULE_FRAGMENT)
3468 r = TAILQ_NEXT(r, entries);
3469 else if (r->prob && r->prob <= karc4random())
3470 r = TAILQ_NEXT(r, entries);
3471 else if (r->match_tag && !pf_match_tag(m, r, nr, &tag))
3472 r = TAILQ_NEXT(r, entries);
3473 else if (r->anchorname[0] && r->anchor == NULL)
3474 r = TAILQ_NEXT(r, entries);
3475 else if (r->os_fingerprint != PF_OSFP_ANY)
3476 r = TAILQ_NEXT(r, entries);
3480 if (r->anchor == NULL) {
3486 r = TAILQ_NEXT(r, entries);
3488 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3491 if (r == NULL && a != NULL)
3492 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3499 REASON_SET(&reason, PFRES_MATCH);
3502 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3504 if ((r->action == PF_DROP) &&
3505 ((r->rule_flag & PFRULE_RETURNICMP) ||
3506 (r->rule_flag & PFRULE_RETURN))) {
3507 struct pf_addr *a = NULL;
3510 if (direction == PF_OUT)
3519 pf_change_a(&a->v4.s_addr, pd->ip_sum,
3520 pd->baddr.v4.s_addr, 0);
3525 PF_ACPY(a, &pd->baddr, af);
3530 if ((af == AF_INET) && r->return_icmp)
3531 pf_send_icmp(m, r->return_icmp >> 8,
3532 r->return_icmp & 255, af, r);
3533 else if ((af == AF_INET6) && r->return_icmp6)
3534 pf_send_icmp(m, r->return_icmp6 >> 8,
3535 r->return_icmp6 & 255, af, r);
3538 if (r->action != PF_PASS)
3541 pf_tag_packet(m, tag);
3543 if (r->keep_state || nr != NULL) {
3544 /* create new state */
3545 struct pf_state *s = NULL;
3546 struct pf_src_node *sn = NULL;
3548 /* check maximums */
3549 if (r->max_states && (r->states >= r->max_states))
3551 /* src node for flter rule */
3552 if ((r->rule_flag & PFRULE_SRCTRACK ||
3553 r->rpool.opts & PF_POOL_STICKYADDR) &&
3554 pf_insert_src_node(&sn, r, saddr, af) != 0)
3556 /* src node for translation rule */
3557 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3558 ((direction == PF_OUT &&
3559 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3560 (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
3562 s = pool_get(&pf_state_pl, PR_NOWAIT);
3565 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3566 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3567 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3568 pf_status.src_nodes--;
3569 pool_put(&pf_src_tree_pl, sn);
3571 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3573 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3574 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3575 pf_status.src_nodes--;
3576 pool_put(&pf_src_tree_pl, nsn);
3578 REASON_SET(&reason, PFRES_MEMORY);
3581 bzero(s, sizeof(*s));
3586 s->nat_rule.ptr = nr;
3587 if (s->nat_rule.ptr != NULL)
3588 s->nat_rule.ptr->states++;
3590 s->allow_opts = r->allow_opts;
3591 s->log = r->log & 2;
3592 s->proto = pd->proto;
3593 s->direction = direction;
3595 if (direction == PF_OUT) {
3596 PF_ACPY(&s->gwy.addr, saddr, af);
3597 PF_ACPY(&s->ext.addr, daddr, af);
3599 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3601 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3603 PF_ACPY(&s->lan.addr, daddr, af);
3604 PF_ACPY(&s->ext.addr, saddr, af);
3606 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3608 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3610 s->hash = pf_state_hash(s);
3611 s->src.state = PFOTHERS_SINGLE;
3612 s->dst.state = PFOTHERS_NO_TRAFFIC;
3613 s->creation = time_second;
3614 s->expire = time_second;
3615 s->timeout = PFTM_OTHER_FIRST_PACKET;
3616 pf_set_rt_ifp(s, saddr);
3619 s->src_node->states++;
3622 PF_ACPY(&nsn->raddr, &pd->naddr, af);
3623 s->nat_src_node = nsn;
3624 s->nat_src_node->states++;
3626 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3627 REASON_SET(&reason, PFRES_MEMORY);
3628 pf_src_tree_remove_state(s);
3629 pool_put(&pf_state_pl, s);
3639 pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif,
3640 struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am,
3641 struct pf_ruleset **rsm)
3643 struct pf_rule *r, *a = NULL;
3644 struct pf_ruleset *ruleset = NULL;
3645 sa_family_t af = pd->af;
3649 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3652 if (r->kif != NULL &&
3653 (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3654 r = r->skip[PF_SKIP_IFP].ptr;
3655 else if (r->direction && r->direction != direction)
3656 r = r->skip[PF_SKIP_DIR].ptr;
3657 else if (r->af && r->af != af)
3658 r = r->skip[PF_SKIP_AF].ptr;
3659 else if (r->proto && r->proto != pd->proto)
3660 r = r->skip[PF_SKIP_PROTO].ptr;
3661 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, r->src.not))
3662 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3663 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, r->dst.not))
3664 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3665 else if (r->tos && !(r->tos & pd->tos))
3666 r = TAILQ_NEXT(r, entries);
3667 else if (r->os_fingerprint != PF_OSFP_ANY)
3668 r = TAILQ_NEXT(r, entries);
3669 else if (pd->proto == IPPROTO_UDP &&
3670 (r->src.port_op || r->dst.port_op))
3671 r = TAILQ_NEXT(r, entries);
3672 else if (pd->proto == IPPROTO_TCP &&
3673 (r->src.port_op || r->dst.port_op || r->flagset))
3674 r = TAILQ_NEXT(r, entries);
3675 else if ((pd->proto == IPPROTO_ICMP ||
3676 pd->proto == IPPROTO_ICMPV6) &&
3677 (r->type || r->code))
3678 r = TAILQ_NEXT(r, entries);
3679 else if (r->prob && r->prob <= karc4random())
3680 r = TAILQ_NEXT(r, entries);
3681 else if (r->match_tag && !pf_match_tag(m, r, NULL, &tag))
3682 r = TAILQ_NEXT(r, entries);
3683 else if (r->anchorname[0] && r->anchor == NULL)
3684 r = TAILQ_NEXT(r, entries);
3686 if (r->anchor == NULL) {
3692 r = TAILQ_NEXT(r, entries);
3694 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3697 if (r == NULL && a != NULL)
3698 PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3705 REASON_SET(&reason, PFRES_MATCH);
3708 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3710 if (r->action != PF_PASS)
3713 pf_tag_packet(m, tag);
3719 pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
3720 struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
3723 struct pf_state key;
3724 struct tcphdr *th = pd->hdr.tcp;
3725 u_int16_t win = ntohs(th->th_win);
3726 u_int32_t ack, end, seq;
3730 struct pf_state_peer *src, *dst;
3733 key.proto = IPPROTO_TCP;
3734 if (direction == PF_IN) {
3735 PF_ACPY(&key.ext.addr, pd->src, key.af);
3736 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
3737 key.ext.port = th->th_sport;
3738 key.gwy.port = th->th_dport;
3740 PF_ACPY(&key.lan.addr, pd->src, key.af);
3741 PF_ACPY(&key.ext.addr, pd->dst, key.af);
3742 key.lan.port = th->th_sport;
3743 key.ext.port = th->th_dport;
3748 if (direction == (*state)->direction) {
3749 src = &(*state)->src;
3750 dst = &(*state)->dst;
3752 src = &(*state)->dst;
3753 dst = &(*state)->src;
3756 if ((*state)->src.state == PF_TCPS_PROXY_SRC) {
3757 if (direction != (*state)->direction)
3758 return (PF_SYNPROXY_DROP);
3759 if (th->th_flags & TH_SYN) {
3760 if (ntohl(th->th_seq) != (*state)->src.seqlo)
3762 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
3763 pd->src, th->th_dport, th->th_sport,
3764 (*state)->src.seqhi, ntohl(th->th_seq) + 1,
3765 TH_SYN|TH_ACK, 0, (*state)->src.mss, 0);
3766 return (PF_SYNPROXY_DROP);
3767 } else if (!(th->th_flags & TH_ACK) ||
3768 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
3769 (ntohl(th->th_seq) != (*state)->src.seqlo + 1))
3772 (*state)->src.state = PF_TCPS_PROXY_DST;
3774 if ((*state)->src.state == PF_TCPS_PROXY_DST) {
3775 struct pf_state_host *src, *dst;
3777 if (direction == PF_OUT) {
3778 src = &(*state)->gwy;
3779 dst = &(*state)->ext;
3781 src = &(*state)->ext;
3782 dst = &(*state)->lan;
3784 if (direction == (*state)->direction) {
3785 if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) ||
3786 (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
3787 (ntohl(th->th_seq) != (*state)->src.seqlo + 1))
3789 (*state)->src.max_win = MAX(ntohs(th->th_win), 1);
3790 if ((*state)->dst.seqhi == 1)
3791 (*state)->dst.seqhi = karc4random();
3792 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
3793 &dst->addr, src->port, dst->port,
3794 (*state)->dst.seqhi, 0, TH_SYN, 0,
3795 (*state)->src.mss, 0);
3796 return (PF_SYNPROXY_DROP);
3797 } else if (((th->th_flags & (TH_SYN|TH_ACK)) !=
3799 (ntohl(th->th_ack) != (*state)->dst.seqhi + 1))
3802 (*state)->dst.max_win = MAX(ntohs(th->th_win), 1);
3803 (*state)->dst.seqlo = ntohl(th->th_seq);
3804 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
3805 pd->src, th->th_dport, th->th_sport,
3806 ntohl(th->th_ack), ntohl(th->th_seq) + 1,
3807 TH_ACK, (*state)->src.max_win, 0, 0);
3808 pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
3809 &dst->addr, src->port, dst->port,
3810 (*state)->src.seqhi + 1, (*state)->src.seqlo + 1,
3811 TH_ACK, (*state)->dst.max_win, 0, 0);
3812 (*state)->src.seqdiff = (*state)->dst.seqhi -
3813 (*state)->src.seqlo;
3814 (*state)->dst.seqdiff = (*state)->src.seqhi -
3815 (*state)->dst.seqlo;
3816 (*state)->src.seqhi = (*state)->src.seqlo +
3817 (*state)->dst.max_win;
3818 (*state)->dst.seqhi = (*state)->dst.seqlo +
3819 (*state)->src.max_win;
3820 (*state)->src.wscale = (*state)->dst.wscale = 0;
3821 (*state)->src.state = (*state)->dst.state =
3823 return (PF_SYNPROXY_DROP);
3827 if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) {
3828 sws = src->wscale & PF_WSCALE_MASK;
3829 dws = dst->wscale & PF_WSCALE_MASK;
3834 * Sequence tracking algorithm from Guido van Rooij's paper:
3835 * http://www.madison-gurkha.com/publications/tcp_filtering/
3839 seq = ntohl(th->th_seq);
3840 if (src->seqlo == 0) {
3842 * First packet from this end. The other end has already set
3843 * the seqlo field. Set its state.
3845 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) &&
3846 src->scrub == NULL) {
3847 if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) {
3848 REASON_SET(reason, PFRES_MEMORY);
3853 /* Deferred generation of sequence number modulator */
3854 if (dst->seqdiff && !src->seqdiff) {
3855 while ((src->seqdiff = karc4random()) == 0)
3857 ack = ntohl(th->th_ack) - dst->seqdiff;
3858 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3860 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3863 ack = ntohl(th->th_ack);
3866 end = seq + pd->p_len;
3867 if (th->th_flags & TH_SYN) {
3869 (*state)->sync_flags |= PFSTATE_GOT_SYN2;
3870 if (dst->wscale & PF_WSCALE_FLAG) {
3871 src->wscale = pf_get_wscale(m, off, th->th_off,
3873 if (src->wscale & PF_WSCALE_FLAG) {
3874 /* Remove scale factor from initial
3876 sws = src->wscale & PF_WSCALE_MASK;
3877 win = ((u_int32_t)win + (1 << sws) - 1)
3879 dws = dst->wscale & PF_WSCALE_MASK;
3882 * Fixup other window. Undo the
3883 * normalization that was done on
3884 * the initial SYN. This can result
3885 * in max_win being larger then
3886 * actual but we don't really have
3889 dst->max_win <<= dst->wscale &
3891 /* in case of a retrans SYN|ACK */
3896 if (th->th_flags & TH_FIN)
3900 if (src->state < TCPS_SYN_SENT)
3901 src->state = TCPS_SYN_SENT;
3904 * May need to slide the window (seqhi may have been set by
3905 * the crappy stack check or if we picked up the connection
3906 * after establishment)
3908 if (src->seqhi == 1 ||
3909 SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi))
3910 src->seqhi = end + MAX(1, dst->max_win << dws);
3911 if (win > src->max_win)
3915 ack = ntohl(th->th_ack) - dst->seqdiff;
3917 /* Modulate sequence numbers */
3918 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3920 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3923 end = seq + pd->p_len;
3924 if (th->th_flags & TH_SYN)
3926 if (th->th_flags & TH_FIN)
3930 if ((th->th_flags & TH_ACK) == 0) {
3931 /* Let it pass through the ack skew check */
3933 } else if ((ack == 0 &&
3934 (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) ||
3935 /* broken tcp stacks do not set ack */
3936 (dst->state < TCPS_SYN_SENT)) {
3938 * Many stacks (ours included) will set the ACK number in an
3939 * FIN|ACK if the SYN times out -- no sequence to ACK.
3945 /* Ease sequencing restrictions on no data packets */
3950 ackskew = dst->seqlo - ack;
3952 #define MAXACKWINDOW (0xffff + 1500) /* 1500 is an arbitrary fudge factor */
3954 if (SEQ_GEQ(src->seqhi, end) &&
3955 /* Last octet inside other's window space */
3956 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) &&
3957 /* Retrans: not more than one window back */
3958 (ackskew >= -MAXACKWINDOW) &&
3959 /* Acking not more than one reassembled fragment backwards */
3960 (ackskew <= (MAXACKWINDOW << sws))) {
3961 /* Acking not more than one window forward */
3963 /* update max window */
3964 if (src->max_win < win)
3966 /* synchronize sequencing */
3967 if (SEQ_GT(end, src->seqlo))
3969 /* slide the window of what the other end can send */
3970 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
3971 dst->seqhi = ack + MAX((win << sws), 1);
3975 if (th->th_flags & TH_SYN)
3976 if (src->state < TCPS_SYN_SENT)
3977 src->state = TCPS_SYN_SENT;
3978 if (th->th_flags & TH_FIN)
3979 if (src->state < TCPS_CLOSING)
3980 src->state = TCPS_CLOSING;
3981 if (th->th_flags & TH_ACK) {
3982 if (dst->state == TCPS_SYN_SENT)
3983 dst->state = TCPS_ESTABLISHED;
3984 else if (dst->state == TCPS_CLOSING)
3985 dst->state = TCPS_FIN_WAIT_2;
3987 if (th->th_flags & TH_RST)
3988 src->state = dst->state = TCPS_TIME_WAIT;
3990 /* update expire time */
3991 (*state)->expire = time_second;
3992 if (src->state >= TCPS_FIN_WAIT_2 &&
3993 dst->state >= TCPS_FIN_WAIT_2)
3994 (*state)->timeout = PFTM_TCP_CLOSED;
3995 else if (src->state >= TCPS_FIN_WAIT_2 ||
3996 dst->state >= TCPS_FIN_WAIT_2)
3997 (*state)->timeout = PFTM_TCP_FIN_WAIT;
3998 else if (src->state < TCPS_ESTABLISHED ||
3999 dst->state < TCPS_ESTABLISHED)
4000 (*state)->timeout = PFTM_TCP_OPENING;
4001 else if (src->state >= TCPS_CLOSING ||
4002 dst->state >= TCPS_CLOSING)
4003 (*state)->timeout = PFTM_TCP_CLOSING;
4005 (*state)->timeout = PFTM_TCP_ESTABLISHED;
4007 /* Fall through to PASS packet */
4009 } else if ((dst->state < TCPS_SYN_SENT ||
4010 dst->state >= TCPS_FIN_WAIT_2 ||
4011 src->state >= TCPS_FIN_WAIT_2) &&
4012 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) &&
4013 /* Within a window forward of the originating packet */
4014 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) {
4015 /* Within a window backward of the originating packet */
4018 * This currently handles three situations:
4019 * 1) Stupid stacks will shotgun SYNs before their peer
4021 * 2) When PF catches an already established stream (the
4022 * firewall rebooted, the state table was flushed, routes
4024 * 3) Packets get funky immediately after the connection
4025 * closes (this should catch Solaris spurious ACK|FINs
4026 * that web servers like to spew after a close)
4028 * This must be a little more careful than the above code
4029 * since packet floods will also be caught here. We don't
4030 * update the TTL here to mitigate the damage of a packet
4031 * flood and so the same code can handle awkward establishment
4032 * and a loosened connection close.
4033 * In the establishment case, a correct peer response will
4034 * validate the connection, go through the normal state code
4035 * and keep updating the state TTL.
4038 if (pf_status.debug >= PF_DEBUG_MISC) {
4039 kprintf("pf: loose state match: ");
4040 pf_print_state(*state);
4041 pf_print_flags(th->th_flags);
4042 kprintf(" seq=%u ack=%u len=%u ackskew=%d pkts=%d:%d\n",
4043 seq, ack, pd->p_len, ackskew,
4044 (*state)->packets[0], (*state)->packets[1]);
4047 /* update max window */
4048 if (src->max_win < win)
4050 /* synchronize sequencing */
4051 if (SEQ_GT(end, src->seqlo))
4053 /* slide the window of what the other end can send */
4054 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4055 dst->seqhi = ack + MAX((win << sws), 1);
4058 * Cannot set dst->seqhi here since this could be a shotgunned
4059 * SYN and not an already established connection.
4062 if (th->th_flags & TH_FIN)
4063 if (src->state < TCPS_CLOSING)
4064 src->state = TCPS_CLOSING;
4065 if (th->th_flags & TH_RST)
4066 src->state = dst->state = TCPS_TIME_WAIT;
4068 /* Fall through to PASS packet */
4070 } else if ((*state)->pickup_mode == PF_PICKUPS_HASHONLY ||
4071 ((*state)->pickup_mode == PF_PICKUPS_ENABLED &&
4072 ((*state)->sync_flags & PFSTATE_GOT_SYN_MASK) !=
4073 PFSTATE_GOT_SYN_MASK)) {
4075 * If pickup mode is hash only, do not fail on sequence checks.
4077 * If pickup mode is enabled and we did not see the SYN in
4078 * both direction, do not fail on sequence checks because
4079 * we do not have complete information on window scale.
4081 * Adjust expiration and fall through to PASS packet.
4082 * XXX Add a FIN check to reduce timeout?
4084 (*state)->expire = time_second;
4087 * Failure processing
4089 if ((*state)->dst.state == TCPS_SYN_SENT &&
4090 (*state)->src.state == TCPS_SYN_SENT) {
4091 /* Send RST for state mismatches during handshake */
4092 if (!(th->th_flags & TH_RST)) {
4093 u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
4095 if (th->th_flags & TH_SYN)
4097 if (th->th_flags & TH_FIN)
4099 pf_send_tcp((*state)->rule.ptr, pd->af,
4100 pd->dst, pd->src, th->th_dport,
4101 th->th_sport, ntohl(th->th_ack), ack,
4102 TH_RST|TH_ACK, 0, 0,
4103 (*state)->rule.ptr->return_ttl);
4108 } else if (pf_status.debug >= PF_DEBUG_MISC) {
4109 kprintf("pf: BAD state: ");
4110 pf_print_state(*state);
4111 pf_print_flags(th->th_flags);
4112 kprintf(" seq=%u ack=%u len=%u ackskew=%d pkts=%d:%d "
4113 "dir=%s,%s\n", seq, ack, pd->p_len, ackskew,
4114 (*state)->packets[0], (*state)->packets[1],
4115 direction == PF_IN ? "in" : "out",
4116 direction == (*state)->direction ? "fwd" : "rev");
4117 kprintf("pf: State failure on: %c %c %c %c | %c %c\n",
4118 SEQ_GEQ(src->seqhi, end) ? ' ' : '1',
4119 SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ?
4121 (ackskew >= -MAXACKWINDOW) ? ' ' : '3',
4122 (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4',
4123 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5',
4124 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6');
4129 if (dst->scrub || src->scrub) {
4130 if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4131 src, dst, ©back))
4135 /* Any packets which have gotten here are to be passed */
4137 /* translate source/destination address, if necessary */
4138 if (STATE_TRANSLATE(*state)) {
4139 if (direction == PF_OUT)
4140 pf_change_ap(pd->src, &th->th_sport, pd->ip_sum,
4141 &th->th_sum, &(*state)->gwy.addr,
4142 (*state)->gwy.port, 0, pd->af);
4144 pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum,
4145 &th->th_sum, &(*state)->lan.addr,
4146 (*state)->lan.port, 0, pd->af);
4147 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4148 } else if (copyback) {
4149 /* Copyback sequence modulation or stateful scrub changes */
4150 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4157 pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif,
4158 struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4160 struct pf_state_peer *src, *dst;
4161 struct pf_state key;
4162 struct udphdr *uh = pd->hdr.udp;
4165 key.proto = IPPROTO_UDP;
4166 if (direction == PF_IN) {
4167 PF_ACPY(&key.ext.addr, pd->src, key.af);
4168 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4169 key.ext.port = uh->uh_sport;
4170 key.gwy.port = uh->uh_dport;
4172 PF_ACPY(&key.lan.addr, pd->src, key.af);
4173 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4174 key.lan.port = uh->uh_sport;
4175 key.ext.port = uh->uh_dport;
4180 if (direction == (*state)->direction) {
4181 src = &(*state)->src;
4182 dst = &(*state)->dst;
4184 src = &(*state)->dst;
4185 dst = &(*state)->src;
4189 if (src->state < PFUDPS_SINGLE)
4190 src->state = PFUDPS_SINGLE;
4191 if (dst->state == PFUDPS_SINGLE)
4192 dst->state = PFUDPS_MULTIPLE;
4194 /* update expire time */
4195 (*state)->expire = time_second;
4196 if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE)
4197 (*state)->timeout = PFTM_UDP_MULTIPLE;
4199 (*state)->timeout = PFTM_UDP_SINGLE;
4201 /* translate source/destination address, if necessary */
4202 if (STATE_TRANSLATE(*state)) {
4203 if (direction == PF_OUT)
4204 pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum,
4205 &uh->uh_sum, &(*state)->gwy.addr,
4206 (*state)->gwy.port, 1, pd->af);
4208 pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum,
4209 &uh->uh_sum, &(*state)->lan.addr,
4210 (*state)->lan.port, 1, pd->af);
4211 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
4218 pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif,
4219 struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4221 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
4222 u_int16_t icmpid = 0;
4223 u_int16_t *icmpsum = NULL;
4224 u_int8_t icmptype = 0;
4227 switch (pd->proto) {
4230 icmptype = pd->hdr.icmp->icmp_type;
4231 icmpid = pd->hdr.icmp->icmp_id;
4232 icmpsum = &pd->hdr.icmp->icmp_cksum;
4234 if (icmptype == ICMP_UNREACH ||
4235 icmptype == ICMP_SOURCEQUENCH ||
4236 icmptype == ICMP_REDIRECT ||
4237 icmptype == ICMP_TIMXCEED ||
4238 icmptype == ICMP_PARAMPROB)
4243 case IPPROTO_ICMPV6:
4244 icmptype = pd->hdr.icmp6->icmp6_type;
4245 icmpid = pd->hdr.icmp6->icmp6_id;
4246 icmpsum = &pd->hdr.icmp6->icmp6_cksum;
4248 if (icmptype == ICMP6_DST_UNREACH ||
4249 icmptype == ICMP6_PACKET_TOO_BIG ||
4250 icmptype == ICMP6_TIME_EXCEEDED ||
4251 icmptype == ICMP6_PARAM_PROB)
4260 * ICMP query/reply message not related to a TCP/UDP packet.
4261 * Search for an ICMP state.
4263 struct pf_state key;
4266 key.proto = pd->proto;
4267 if (direction == PF_IN) {
4268 PF_ACPY(&key.ext.addr, pd->src, key.af);
4269 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4270 key.ext.port = icmpid;
4271 key.gwy.port = icmpid;
4273 PF_ACPY(&key.lan.addr, pd->src, key.af);
4274 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4275 key.lan.port = icmpid;
4276 key.ext.port = icmpid;
4281 (*state)->expire = time_second;
4282 (*state)->timeout = PFTM_ICMP_ERROR_REPLY;
4284 /* translate source/destination address, if necessary */
4285 if (PF_ANEQ(&(*state)->lan.addr, &(*state)->gwy.addr, pd->af)) {
4286 if (direction == PF_OUT) {
4290 pf_change_a(&saddr->v4.s_addr,
4292 (*state)->gwy.addr.v4.s_addr, 0);
4298 &pd->hdr.icmp6->icmp6_cksum,
4299 &(*state)->gwy.addr, 0);
4301 sizeof(struct icmp6_hdr),
4302 (caddr_t)pd->hdr.icmp6);
4310 pf_change_a(&daddr->v4.s_addr,
4312 (*state)->lan.addr.v4.s_addr, 0);
4318 &pd->hdr.icmp6->icmp6_cksum,
4319 &(*state)->lan.addr, 0);
4321 sizeof(struct icmp6_hdr),
4322 (caddr_t)pd->hdr.icmp6);
4333 * ICMP error message in response to a TCP/UDP packet.
4334 * Extract the inner TCP/UDP header and search for that state.
4337 struct pf_pdesc pd2;
4342 struct ip6_hdr h2_6;
4352 /* offset of h2 in mbuf chain */
4353 ipoff2 = off + ICMP_MINLEN;
4355 if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2),
4356 NULL, NULL, pd2.af)) {
4357 DPFPRINTF(PF_DEBUG_MISC,
4358 ("pf: ICMP error message too short "
4363 * ICMP error messages don't refer to non-first
4367 * Note: We are dealing with an encapsulated
4368 * header. This means ip_off/ip_len are not
4369 * in host byte order!
4371 if (h2.ip_off & htons(IP_OFFMASK))
4374 /* offset of protocol header that follows h2 */
4375 off2 = ipoff2 + (h2.ip_hl << 2);
4377 pd2.proto = h2.ip_p;
4378 pd2.src = (struct pf_addr *)&h2.ip_src;
4379 pd2.dst = (struct pf_addr *)&h2.ip_dst;
4380 pd2.ip_sum = &h2.ip_sum;
4385 ipoff2 = off + sizeof(struct icmp6_hdr);
4387 if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6),
4388 NULL, NULL, pd2.af)) {
4389 DPFPRINTF(PF_DEBUG_MISC,
4390 ("pf: ICMP error message too short "
4394 pd2.proto = h2_6.ip6_nxt;
4395 pd2.src = (struct pf_addr *)&h2_6.ip6_src;
4396 pd2.dst = (struct pf_addr *)&h2_6.ip6_dst;
4398 off2 = ipoff2 + sizeof(h2_6);
4400 switch (pd2.proto) {
4401 case IPPROTO_FRAGMENT:
4403 * ICMPv6 error messages for
4404 * non-first fragments
4408 case IPPROTO_HOPOPTS:
4409 case IPPROTO_ROUTING:
4410 case IPPROTO_DSTOPTS: {
4411 /* get next header and header length */
4412 struct ip6_ext opt6;
4414 if (!pf_pull_hdr(m, off2, &opt6,
4415 sizeof(opt6), NULL, NULL, pd2.af)) {
4416 DPFPRINTF(PF_DEBUG_MISC,
4417 ("pf: ICMPv6 short opt\n"));
4420 if (pd2.proto == IPPROTO_AH)
4421 off2 += (opt6.ip6e_len + 2) * 4;
4423 off2 += (opt6.ip6e_len + 1) * 8;
4424 pd2.proto = opt6.ip6e_nxt;
4425 /* goto the next header */
4432 } while (!terminal);
4436 DPFPRINTF(PF_DEBUG_MISC,
4437 ("pf: ICMP AF %d unknown (ip6)\n", pd->af));
4442 switch (pd2.proto) {
4446 struct pf_state key;
4447 struct pf_state_peer *src, *dst;
4452 * Only the first 8 bytes of the TCP header can be
4453 * expected. Don't access any TCP header fields after
4454 * th_seq, an ackskew test is not possible.
4456 if (!pf_pull_hdr(m, off2, &th, 8, NULL, NULL, pd2.af)) {
4457 DPFPRINTF(PF_DEBUG_MISC,
4458 ("pf: ICMP error message too short "
4464 key.proto = IPPROTO_TCP;
4465 if (direction == PF_IN) {
4466 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4467 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4468 key.ext.port = th.th_dport;
4469 key.gwy.port = th.th_sport;
4471 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4472 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4473 key.lan.port = th.th_dport;
4474 key.ext.port = th.th_sport;
4479 if (direction == (*state)->direction) {
4480 src = &(*state)->dst;
4481 dst = &(*state)->src;
4483 src = &(*state)->src;
4484 dst = &(*state)->dst;
4487 if (src->wscale && dst->wscale &&
4488 !(th.th_flags & TH_SYN))
4489 dws = dst->wscale & PF_WSCALE_MASK;
4493 /* Demodulate sequence number */
4494 seq = ntohl(th.th_seq) - src->seqdiff;
4496 pf_change_a(&th.th_seq, icmpsum,
4501 if (!SEQ_GEQ(src->seqhi, seq) ||
4502 !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws))) {
4503 if (pf_status.debug >= PF_DEBUG_MISC) {
4504 kprintf("pf: BAD ICMP %d:%d ",
4505 icmptype, pd->hdr.icmp->icmp_code);
4506 pf_print_host(pd->src, 0, pd->af);
4508 pf_print_host(pd->dst, 0, pd->af);
4509 kprintf(" state: ");
4510 pf_print_state(*state);
4511 kprintf(" seq=%u\n", seq);
4516 if (STATE_TRANSLATE(*state)) {
4517 if (direction == PF_IN) {
4518 pf_change_icmp(pd2.src, &th.th_sport,
4519 daddr, &(*state)->lan.addr,
4520 (*state)->lan.port, NULL,
4521 pd2.ip_sum, icmpsum,
4522 pd->ip_sum, 0, pd2.af);
4524 pf_change_icmp(pd2.dst, &th.th_dport,
4525 saddr, &(*state)->gwy.addr,
4526 (*state)->gwy.port, NULL,
4527 pd2.ip_sum, icmpsum,
4528 pd->ip_sum, 0, pd2.af);
4537 m_copyback(m, off, ICMP_MINLEN,
4538 (caddr_t)pd->hdr.icmp);
4539 m_copyback(m, ipoff2, sizeof(h2),
4546 sizeof(struct icmp6_hdr),
4547 (caddr_t)pd->hdr.icmp6);
4548 m_copyback(m, ipoff2, sizeof(h2_6),
4553 m_copyback(m, off2, 8, (caddr_t)&th);
4561 struct pf_state key;
4563 if (!pf_pull_hdr(m, off2, &uh, sizeof(uh),
4564 NULL, NULL, pd2.af)) {
4565 DPFPRINTF(PF_DEBUG_MISC,
4566 ("pf: ICMP error message too short "
4572 key.proto = IPPROTO_UDP;
4573 if (direction == PF_IN) {
4574 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4575 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4576 key.ext.port = uh.uh_dport;
4577 key.gwy.port = uh.uh_sport;
4579 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4580 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4581 key.lan.port = uh.uh_dport;
4582 key.ext.port = uh.uh_sport;
4587 if (STATE_TRANSLATE(*state)) {
4588 if (direction == PF_IN) {
4589 pf_change_icmp(pd2.src, &uh.uh_sport,
4590 daddr, &(*state)->lan.addr,
4591 (*state)->lan.port, &uh.uh_sum,
4592 pd2.ip_sum, icmpsum,
4593 pd->ip_sum, 1, pd2.af);
4595 pf_change_icmp(pd2.dst, &uh.uh_dport,
4596 saddr, &(*state)->gwy.addr,
4597 (*state)->gwy.port, &uh.uh_sum,
4598 pd2.ip_sum, icmpsum,
4599 pd->ip_sum, 1, pd2.af);
4604 m_copyback(m, off, ICMP_MINLEN,
4605 (caddr_t)pd->hdr.icmp);
4606 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4612 sizeof(struct icmp6_hdr),
4613 (caddr_t)pd->hdr.icmp6);
4614 m_copyback(m, ipoff2, sizeof(h2_6),
4619 m_copyback(m, off2, sizeof(uh), (caddr_t)&uh);
4626 case IPPROTO_ICMP: {
4628 struct pf_state key;
4630 if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN,
4631 NULL, NULL, pd2.af)) {
4632 DPFPRINTF(PF_DEBUG_MISC,
4633 ("pf: ICMP error message too short i"
4639 key.proto = IPPROTO_ICMP;
4640 if (direction == PF_IN) {
4641 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4642 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4643 key.ext.port = iih.icmp_id;
4644 key.gwy.port = iih.icmp_id;
4646 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4647 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4648 key.lan.port = iih.icmp_id;
4649 key.ext.port = iih.icmp_id;
4654 if (STATE_TRANSLATE(*state)) {
4655 if (direction == PF_IN) {
4656 pf_change_icmp(pd2.src, &iih.icmp_id,
4657 daddr, &(*state)->lan.addr,
4658 (*state)->lan.port, NULL,
4659 pd2.ip_sum, icmpsum,
4660 pd->ip_sum, 0, AF_INET);
4662 pf_change_icmp(pd2.dst, &iih.icmp_id,
4663 saddr, &(*state)->gwy.addr,
4664 (*state)->gwy.port, NULL,
4665 pd2.ip_sum, icmpsum,
4666 pd->ip_sum, 0, AF_INET);
4668 m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
4669 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4670 m_copyback(m, off2, ICMP_MINLEN, (caddr_t)&iih);
4678 case IPPROTO_ICMPV6: {
4679 struct icmp6_hdr iih;
4680 struct pf_state key;
4682 if (!pf_pull_hdr(m, off2, &iih,
4683 sizeof(struct icmp6_hdr), NULL, NULL, pd2.af)) {
4684 DPFPRINTF(PF_DEBUG_MISC,
4685 ("pf: ICMP error message too short "
4691 key.proto = IPPROTO_ICMPV6;
4692 if (direction == PF_IN) {
4693 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4694 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4695 key.ext.port = iih.icmp6_id;
4696 key.gwy.port = iih.icmp6_id;
4698 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4699 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4700 key.lan.port = iih.icmp6_id;
4701 key.ext.port = iih.icmp6_id;
4706 if (STATE_TRANSLATE(*state)) {
4707 if (direction == PF_IN) {
4708 pf_change_icmp(pd2.src, &iih.icmp6_id,
4709 daddr, &(*state)->lan.addr,
4710 (*state)->lan.port, NULL,
4711 pd2.ip_sum, icmpsum,
4712 pd->ip_sum, 0, AF_INET6);
4714 pf_change_icmp(pd2.dst, &iih.icmp6_id,
4715 saddr, &(*state)->gwy.addr,
4716 (*state)->gwy.port, NULL,
4717 pd2.ip_sum, icmpsum,
4718 pd->ip_sum, 0, AF_INET6);
4720 m_copyback(m, off, sizeof(struct icmp6_hdr),
4721 (caddr_t)pd->hdr.icmp6);
4722 m_copyback(m, ipoff2, sizeof(h2_6), (caddr_t)&h2_6);
4723 m_copyback(m, off2, sizeof(struct icmp6_hdr),
4732 struct pf_state key;
4735 key.proto = pd2.proto;
4736 if (direction == PF_IN) {
4737 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4738 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4742 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4743 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4750 if (STATE_TRANSLATE(*state)) {
4751 if (direction == PF_IN) {
4752 pf_change_icmp(pd2.src, NULL,
4753 daddr, &(*state)->lan.addr,
4755 pd2.ip_sum, icmpsum,
4756 pd->ip_sum, 0, pd2.af);
4758 pf_change_icmp(pd2.dst, NULL,
4759 saddr, &(*state)->gwy.addr,
4761 pd2.ip_sum, icmpsum,
4762 pd->ip_sum, 0, pd2.af);
4767 m_copyback(m, off, ICMP_MINLEN,
4768 (caddr_t)pd->hdr.icmp);
4769 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4775 sizeof(struct icmp6_hdr),
4776 (caddr_t)pd->hdr.icmp6);
4777 m_copyback(m, ipoff2, sizeof(h2_6),
4792 pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif,
4793 struct pf_pdesc *pd)
4795 struct pf_state_peer *src, *dst;
4796 struct pf_state key;
4799 key.proto = pd->proto;
4800 if (direction == PF_IN) {
4801 PF_ACPY(&key.ext.addr, pd->src, key.af);
4802 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4806 PF_ACPY(&key.lan.addr, pd->src, key.af);
4807 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4814 if (direction == (*state)->direction) {
4815 src = &(*state)->src;
4816 dst = &(*state)->dst;
4818 src = &(*state)->dst;
4819 dst = &(*state)->src;
4823 if (src->state < PFOTHERS_SINGLE)
4824 src->state = PFOTHERS_SINGLE;
4825 if (dst->state == PFOTHERS_SINGLE)
4826 dst->state = PFOTHERS_MULTIPLE;
4828 /* update expire time */
4829 (*state)->expire = time_second;
4830 if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE)
4831 (*state)->timeout = PFTM_OTHER_MULTIPLE;
4833 (*state)->timeout = PFTM_OTHER_SINGLE;
4835 /* translate source/destination address, if necessary */
4836 if (STATE_TRANSLATE(*state)) {
4837 if (direction == PF_OUT)
4841 pf_change_a(&pd->src->v4.s_addr,
4842 pd->ip_sum, (*state)->gwy.addr.v4.s_addr,
4848 PF_ACPY(pd->src, &(*state)->gwy.addr, pd->af);
4856 pf_change_a(&pd->dst->v4.s_addr,
4857 pd->ip_sum, (*state)->lan.addr.v4.s_addr,
4863 PF_ACPY(pd->dst, &(*state)->lan.addr, pd->af);
4873 * ipoff and off are measured from the start of the mbuf chain.
4874 * h must be at "ipoff" on the mbuf chain.
4877 pf_pull_hdr(struct mbuf *m, int off, void *p, int len,
4878 u_short *actionp, u_short *reasonp, sa_family_t af)
4883 struct ip *h = mtod(m, struct ip *);
4884 u_int16_t fragoff = (h->ip_off & IP_OFFMASK) << 3;
4888 ACTION_SET(actionp, PF_PASS);
4890 ACTION_SET(actionp, PF_DROP);
4891 REASON_SET(reasonp, PFRES_FRAG);
4895 if (m->m_pkthdr.len < off + len ||
4896 h->ip_len < off + len) {
4897 ACTION_SET(actionp, PF_DROP);
4898 REASON_SET(reasonp, PFRES_SHORT);
4906 struct ip6_hdr *h = mtod(m, struct ip6_hdr *);
4908 if (m->m_pkthdr.len < off + len ||
4909 (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) <
4910 (unsigned)(off + len)) {
4911 ACTION_SET(actionp, PF_DROP);
4912 REASON_SET(reasonp, PFRES_SHORT);
4919 m_copydata(m, off, len, p);
4924 pf_routable(struct pf_addr *addr, sa_family_t af)
4926 struct sockaddr_in *dst;
4930 bzero(&ro, sizeof(ro));
4931 dst = satosin(&ro.ro_dst);
4932 dst->sin_family = af;
4933 dst->sin_len = sizeof(*dst);
4934 dst->sin_addr = addr->v4;
4935 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
4937 if (ro.ro_rt != NULL) {
4947 pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
4950 struct mbuf *m0, *m1;
4951 struct route iproute;
4952 struct route *ro = NULL;
4953 struct sockaddr_in *dst;
4955 struct ifnet *ifp = NULL;
4956 struct pf_addr naddr;
4957 struct pf_src_node *sn = NULL;
4961 if (m == NULL || *m == NULL || r == NULL ||
4962 (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
4963 panic("pf_route: invalid parameters");
4965 if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
4966 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
4967 (*m)->m_pkthdr.pf_routed = 1;
4969 if ((*m)->m_pkthdr.pf_routed > 3) {
4974 (*m)->m_pkthdr.pf_routed++;
4977 if (r->rt == PF_DUPTO) {
4978 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL)
4981 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
4986 if (m0->m_len < sizeof(struct ip))
4987 panic("pf_route: m0->m_len < sizeof(struct ip)");
4988 ip = mtod(m0, struct ip *);
4991 bzero((caddr_t)ro, sizeof(*ro));
4992 dst = satosin(&ro->ro_dst);
4993 dst->sin_family = AF_INET;
4994 dst->sin_len = sizeof(*dst);
4995 dst->sin_addr = ip->ip_dst;
4997 if (r->rt == PF_FASTROUTE) {
4999 if (ro->ro_rt == 0) {
5000 ipstat.ips_noroute++;
5004 ifp = ro->ro_rt->rt_ifp;
5005 ro->ro_rt->rt_use++;
5007 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
5008 dst = satosin(ro->ro_rt->rt_gateway);
5010 if (TAILQ_EMPTY(&r->rpool.list))
5011 panic("pf_route: TAILQ_EMPTY(&r->rpool.list)");
5013 pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src,
5015 if (!PF_AZERO(&naddr, AF_INET))
5016 dst->sin_addr.s_addr = naddr.v4.s_addr;
5017 ifp = r->rpool.cur->kif ?
5018 r->rpool.cur->kif->pfik_ifp : NULL;
5020 if (!PF_AZERO(&s->rt_addr, AF_INET))
5021 dst->sin_addr.s_addr =
5022 s->rt_addr.v4.s_addr;
5023 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5030 if (pf_test(PF_OUT, ifp, &m0) != PF_PASS)
5032 else if (m0 == NULL)
5034 if (m0->m_len < sizeof(struct ip))
5035 panic("pf_route: m0->m_len < sizeof(struct ip)");
5036 ip = mtod(m0, struct ip *);
5039 /* Copied from ip_output. */
5040 m0->m_pkthdr.csum_flags |= CSUM_IP;
5041 sw_csum = m0->m_pkthdr.csum_flags & ~ifp->if_hwassist;
5042 if (sw_csum & CSUM_DELAY_DATA) {
5043 in_delayed_cksum(m0);
5044 sw_csum &= ~CSUM_DELAY_DATA;
5046 m0->m_pkthdr.csum_flags &= ifp->if_hwassist;
5049 * If small enough for interface, or the interface will take
5050 * care of the fragmentation for us, can just send directly.
5052 if (ip->ip_len <= ifp->if_mtu || ((ifp->if_hwassist & CSUM_FRAGMENT) &&
5053 (ip->ip_off & IP_DF) == 0)) {
5054 ip->ip_len = htons(ip->ip_len);
5055 ip->ip_off = htons(ip->ip_off);
5057 if (sw_csum & CSUM_DELAY_IP) {
5059 if (ip->ip_v == IPVERSION &&
5060 (ip->ip_hl << 2) == sizeof(*ip)) {
5061 ip->ip_sum = in_cksum_hdr(ip);
5063 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
5067 error = ifp->if_output(ifp, m0, sintosa(dst), ro->ro_rt);
5072 * Too large for interface; fragment if possible.
5073 * Must be able to put at least 8 bytes per fragment.
5075 if (ip->ip_off & IP_DF) {
5076 ipstat.ips_cantfrag++;
5077 if (r->rt != PF_DUPTO) {
5078 icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0,
5086 error = ip_fragment(ip, &m0, ifp->if_mtu, ifp->if_hwassist, sw_csum);
5090 for (m0 = m1; m0; m0 = m1) {
5094 error = ifp->if_output(ifp, m0, sintosa(dst), NULL);
5101 ipstat.ips_fragmented++;
5104 if (r->rt != PF_DUPTO)
5106 if (ro == &iproute && ro->ro_rt)
5118 pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5122 struct route_in6 ip6route;
5123 struct route_in6 *ro;
5124 struct sockaddr_in6 *dst;
5125 struct ip6_hdr *ip6;
5126 struct ifnet *ifp = NULL;
5127 struct pf_addr naddr;
5128 struct pf_src_node *sn = NULL;
5131 if (m == NULL || *m == NULL || r == NULL ||
5132 (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5133 panic("pf_route6: invalid parameters");
5135 if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
5136 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
5137 (*m)->m_pkthdr.pf_routed = 1;
5139 if ((*m)->m_pkthdr.pf_routed > 3) {
5144 (*m)->m_pkthdr.pf_routed++;
5147 if (r->rt == PF_DUPTO) {
5148 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL)
5151 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5156 if (m0->m_len < sizeof(struct ip6_hdr))
5157 panic("pf_route6: m0->m_len < sizeof(struct ip6_hdr)");
5158 ip6 = mtod(m0, struct ip6_hdr *);
5161 bzero((caddr_t)ro, sizeof(*ro));
5162 dst = (struct sockaddr_in6 *)&ro->ro_dst;
5163 dst->sin6_family = AF_INET6;
5164 dst->sin6_len = sizeof(*dst);
5165 dst->sin6_addr = ip6->ip6_dst;
5168 if (r->rt == PF_FASTROUTE) {
5169 m0->m_pkthdr.fw_flags |= PF_MBUF_GENERATED;
5170 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
5174 if (TAILQ_EMPTY(&r->rpool.list))
5175 panic("pf_route6: TAILQ_EMPTY(&r->rpool.list)");
5177 pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src,
5179 if (!PF_AZERO(&naddr, AF_INET6))
5180 PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5182 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL;
5184 if (!PF_AZERO(&s->rt_addr, AF_INET6))
5185 PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5186 &s->rt_addr, AF_INET6);
5187 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5193 if (pf_test6(PF_OUT, ifp, &m0) != PF_PASS)
5195 else if (m0 == NULL)
5197 if (m0->m_len < sizeof(struct ip6_hdr))
5198 panic("pf_route6: m0->m_len < sizeof(struct ip6_hdr)");
5199 ip6 = mtod(m0, struct ip6_hdr *);
5203 * If the packet is too large for the outgoing interface,
5204 * send back an icmp6 error.
5206 if (IN6_IS_ADDR_LINKLOCAL(&dst->sin6_addr))
5207 dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
5208 if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) {
5209 error = nd6_output(ifp, ifp, m0, dst, NULL);
5211 in6_ifstat_inc(ifp, ifs6_in_toobig);
5212 if (r->rt != PF_DUPTO)
5213 icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
5219 if (r->rt != PF_DUPTO)
5231 * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag
5232 * off is the offset where the protocol header starts
5233 * len is the total length of protocol header plus payload
5234 * returns 0 when the checksum is valid, otherwise returns 1.
5238 * FreeBSD supports cksum offload for the following drivers.
5239 * em(4), gx(4), lge(4), nge(4), ti(4), xl(4)
5240 * If we can make full use of it we would outperform ipfw/ipfilter in
5241 * very heavy traffic.
5242 * I have not tested 'cause I don't have NICs that supports cksum offload.
5243 * (There might be problems. Typical phenomena would be
5244 * 1. No route message for UDP packet.
5245 * 2. No connection acceptance from external hosts regardless of rule set.)
5248 pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p,
5255 if (off < sizeof(struct ip) || len < sizeof(struct udphdr))
5257 if (m->m_pkthdr.len < off + len)
5263 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
5264 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) {
5265 sum = m->m_pkthdr.csum_data;
5267 ip = mtod(m, struct ip *);
5268 sum = in_pseudo(ip->ip_src.s_addr,
5269 ip->ip_dst.s_addr, htonl((u_short)len +
5270 m->m_pkthdr.csum_data + p));
5278 case IPPROTO_ICMPV6:
5288 if (p == IPPROTO_ICMP) {
5293 sum = in_cksum(m, len);
5297 if (m->m_len < sizeof(struct ip))
5299 sum = in_cksum_range(m, p, off, len);
5301 m->m_pkthdr.csum_flags |=
5304 m->m_pkthdr.csum_data = 0xffff;
5310 if (m->m_len < sizeof(struct ip6_hdr))
5312 sum = in6_cksum(m, p, off, len);
5315 * IPv6 H/W cksum off-load not supported yet!
5318 * m->m_pkthdr.csum_flags |=
5319 * (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
5320 * m->m_pkthdr.csum_data = 0xffff;
5332 tcpstat.tcps_rcvbadsum++;
5335 udpstat.udps_badsum++;
5338 icmpstat.icps_checksum++;
5341 case IPPROTO_ICMPV6:
5342 icmp6stat.icp6s_checksum++;
5353 pf_test(int dir, struct ifnet *ifp, struct mbuf **m0)
5355 struct pfi_kif *kif;
5356 u_short action, reason = 0, log = 0;
5357 struct mbuf *m = *m0;
5358 struct ip *h = NULL;
5359 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr;
5360 struct pf_state *s = NULL;
5361 struct pf_ruleset *ruleset = NULL;
5363 int off, dirndx, pqid = 0;
5365 if (!pf_status.running || (m->m_pkthdr.fw_flags & PF_MBUF_GENERATED))
5368 kif = pfi_index2kif[ifp->if_index];
5373 if ((m->m_flags & M_PKTHDR) == 0)
5374 panic("non-M_PKTHDR is passed to pf_test");
5377 memset(&pd, 0, sizeof(pd));
5378 if (m->m_pkthdr.len < (int)sizeof(*h)) {
5380 REASON_SET(&reason, PFRES_SHORT);
5385 /* We do IP header normalization and packet reassembly here */
5386 if (pf_normalize_ip(m0, dir, kif, &reason) != PF_PASS) {
5391 h = mtod(m, struct ip *);
5393 off = h->ip_hl << 2;
5394 if (off < (int)sizeof(*h)) {
5396 REASON_SET(&reason, PFRES_SHORT);
5401 pd.src = (struct pf_addr *)&h->ip_src;
5402 pd.dst = (struct pf_addr *)&h->ip_dst;
5403 PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET);
5404 pd.ip_sum = &h->ip_sum;
5408 pd.tot_len = h->ip_len;
5410 /* handle fragments that didn't get reassembled by normalization */
5411 if (h->ip_off & (IP_MF | IP_OFFMASK)) {
5412 action = pf_test_fragment(&r, dir, kif, m, h,
5423 if (!pf_pull_hdr(m, off, &th, sizeof(th),
5424 &action, &reason, AF_INET)) {
5425 log = action != PF_PASS;
5428 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5429 h->ip_len - off, IPPROTO_TCP, AF_INET)) {
5433 pd.p_len = pd.tot_len - off - (th.th_off << 2);
5434 if ((th.th_flags & TH_ACK) && pd.p_len == 0)
5436 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
5437 if (action == PF_DROP)
5439 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
5441 if (action == PF_PASS) {
5443 pfsync_update_state(s);
5448 } else if (s == NULL)
5449 action = pf_test_tcp(&r, &s, dir, kif,
5450 m, off, h, &pd, &a, &ruleset);
5458 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
5459 &action, &reason, AF_INET)) {
5460 log = action != PF_PASS;
5463 if (dir == PF_IN && uh.uh_sum && pf_check_proto_cksum(m,
5464 off, h->ip_len - off, IPPROTO_UDP, AF_INET)) {
5468 if (uh.uh_dport == 0 ||
5469 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
5470 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
5474 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
5475 if (action == PF_PASS) {
5477 pfsync_update_state(s);
5482 } else if (s == NULL)
5483 action = pf_test_udp(&r, &s, dir, kif,
5484 m, off, h, &pd, &a, &ruleset);
5488 case IPPROTO_ICMP: {
5492 if (!pf_pull_hdr(m, off, &ih, ICMP_MINLEN,
5493 &action, &reason, AF_INET)) {
5494 log = action != PF_PASS;
5497 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5498 h->ip_len - off, IPPROTO_ICMP, AF_INET)) {
5502 action = pf_test_state_icmp(&s, dir, kif, m, off, h, &pd);
5503 if (action == PF_PASS) {
5505 pfsync_update_state(s);
5510 } else if (s == NULL)
5511 action = pf_test_icmp(&r, &s, dir, kif,
5512 m, off, h, &pd, &a, &ruleset);
5517 action = pf_test_state_other(&s, dir, kif, &pd);
5518 if (action == PF_PASS) {
5520 pfsync_update_state(s);
5525 } else if (s == NULL)
5526 action = pf_test_other(&r, &s, dir, kif, m, off, h,
5532 if (action == PF_PASS && h->ip_hl > 5 &&
5533 !((s && s->allow_opts) || r->allow_opts)) {
5535 REASON_SET(&reason, PFRES_SHORT);
5537 DPFPRINTF(PF_DEBUG_MISC,
5538 ("pf: dropping packet with ip options\n"));
5542 if (action == PF_PASS && r->qid) {
5543 m->m_pkthdr.fw_flags |= ALTQ_MBUF_TAGGED;
5544 if (pd.tos == IPTOS_LOWDELAY)
5545 m->m_pkthdr.altq_qid = r->pqid;
5547 m->m_pkthdr.altq_qid = r->qid;
5549 KKASSERT(s->hash != 0);
5550 m->m_pkthdr.fw_flags |= ALTQ_MBUF_STATE_HASHED;
5551 m->m_pkthdr.altq_state_hash = s->hash;
5553 m->m_pkthdr.ecn_af = AF_INET;
5554 m->m_pkthdr.header = h;
5559 * connections redirected to loopback should not match sockets
5560 * bound specifically to loopback due to security implications,
5561 * see tcp_input() and in_pcblookup_listen().
5563 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
5564 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
5565 (s->nat_rule.ptr->action == PF_RDR ||
5566 s->nat_rule.ptr->action == PF_BINAT) &&
5567 (ntohl(pd.dst->v4.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
5569 REASON_SET(&reason, PFRES_MEMORY);
5572 m->m_pkthdr.fw_flags |= PF_MBUF_TRANSLATE_LOCALHOST;
5575 PFLOG_PACKET(kif, h, m, AF_INET, dir, reason, r, a, ruleset);
5577 kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
5578 kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++;
5580 if (action == PF_PASS || r->action == PF_DROP) {
5582 r->bytes += pd.tot_len;
5585 a->bytes += pd.tot_len;
5588 dirndx = (dir == s->direction) ? 0 : 1;
5589 s->packets[dirndx]++;
5590 s->bytes[dirndx] += pd.tot_len;
5591 if (s->nat_rule.ptr != NULL) {
5592 s->nat_rule.ptr->packets++;
5593 s->nat_rule.ptr->bytes += pd.tot_len;
5595 if (s->src_node != NULL) {
5596 s->src_node->packets++;
5597 s->src_node->bytes += pd.tot_len;
5599 if (s->nat_src_node != NULL) {
5600 s->nat_src_node->packets++;
5601 s->nat_src_node->bytes += pd.tot_len;
5605 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
5609 * XXX: we need to make sure that the addresses
5610 * passed to pfr_update_stats() are the same than
5611 * the addresses used during matching (pfr_match)
5613 if (r == &pf_default_rule) {
5615 x = (s == NULL || s->direction == dir) ?
5616 &pd.baddr : &pd.naddr;
5618 x = (s == NULL || s->direction == dir) ?
5619 &pd.naddr : &pd.baddr;
5620 if (x == &pd.baddr || s == NULL) {
5621 /* we need to change the address */
5628 if (tr->src.addr.type == PF_ADDR_TABLE)
5629 pfr_update_stats(tr->src.addr.p.tbl, (s == NULL ||
5630 s->direction == dir) ? pd.src : pd.dst, pd.af,
5631 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5633 if (tr->dst.addr.type == PF_ADDR_TABLE)
5634 pfr_update_stats(tr->dst.addr.p.tbl, (s == NULL ||
5635 s->direction == dir) ? pd.dst : pd.src, pd.af,
5636 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5641 if (action == PF_SYNPROXY_DROP) {
5646 /* pf_route can free the mbuf causing *m0 to become NULL */
5647 pf_route(m0, r, dir, ifp, s);
5655 pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0)
5657 struct pfi_kif *kif;
5658 u_short action, reason = 0, log = 0;
5659 struct mbuf *m = *m0;
5660 struct ip6_hdr *h = NULL;
5661 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr;
5662 struct pf_state *s = NULL;
5663 struct pf_ruleset *ruleset = NULL;
5665 int off, terminal = 0, dirndx;
5667 if (!pf_status.running || (m->m_pkthdr.fw_flags & PF_MBUF_GENERATED))
5670 kif = pfi_index2kif[ifp->if_index];
5675 if ((m->m_flags & M_PKTHDR) == 0)
5676 panic("non-M_PKTHDR is passed to pf_test");
5679 memset(&pd, 0, sizeof(pd));
5680 if (m->m_pkthdr.len < (int)sizeof(*h)) {
5682 REASON_SET(&reason, PFRES_SHORT);
5687 /* We do IP header normalization and packet reassembly here */
5688 if (pf_normalize_ip6(m0, dir, kif, &reason) != PF_PASS) {
5693 h = mtod(m, struct ip6_hdr *);
5695 pd.src = (struct pf_addr *)&h->ip6_src;
5696 pd.dst = (struct pf_addr *)&h->ip6_dst;
5697 PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET6);
5701 pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr);
5703 off = ((caddr_t)h - m->m_data) + sizeof(struct ip6_hdr);
5704 pd.proto = h->ip6_nxt;
5707 case IPPROTO_FRAGMENT:
5708 action = pf_test_fragment(&r, dir, kif, m, h,
5710 if (action == PF_DROP)
5711 REASON_SET(&reason, PFRES_FRAG);
5714 case IPPROTO_HOPOPTS:
5715 case IPPROTO_ROUTING:
5716 case IPPROTO_DSTOPTS: {
5717 /* get next header and header length */
5718 struct ip6_ext opt6;
5720 if (!pf_pull_hdr(m, off, &opt6, sizeof(opt6),
5721 NULL, NULL, pd.af)) {
5722 DPFPRINTF(PF_DEBUG_MISC,
5723 ("pf: IPv6 short opt\n"));
5725 REASON_SET(&reason, PFRES_SHORT);
5729 if (pd.proto == IPPROTO_AH)
5730 off += (opt6.ip6e_len + 2) * 4;
5732 off += (opt6.ip6e_len + 1) * 8;
5733 pd.proto = opt6.ip6e_nxt;
5734 /* goto the next header */
5741 } while (!terminal);
5749 if (!pf_pull_hdr(m, off, &th, sizeof(th),
5750 &action, &reason, AF_INET6)) {
5751 log = action != PF_PASS;
5754 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5755 ntohs(h->ip6_plen), IPPROTO_TCP, AF_INET6)) {
5759 pd.p_len = pd.tot_len - off - (th.th_off << 2);
5760 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
5761 if (action == PF_DROP)
5763 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
5765 if (action == PF_PASS) {
5767 pfsync_update_state(s);
5772 } else if (s == NULL)
5773 action = pf_test_tcp(&r, &s, dir, kif,
5774 m, off, h, &pd, &a, &ruleset);
5782 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
5783 &action, &reason, AF_INET6)) {
5784 log = action != PF_PASS;
5787 if (dir == PF_IN && uh.uh_sum && pf_check_proto_cksum(m,
5788 off, ntohs(h->ip6_plen), IPPROTO_UDP, AF_INET6)) {
5792 if (uh.uh_dport == 0 ||
5793 ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
5794 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
5798 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
5799 if (action == PF_PASS) {
5801 pfsync_update_state(s);
5806 } else if (s == NULL)
5807 action = pf_test_udp(&r, &s, dir, kif,
5808 m, off, h, &pd, &a, &ruleset);
5812 case IPPROTO_ICMPV6: {
5813 struct icmp6_hdr ih;
5816 if (!pf_pull_hdr(m, off, &ih, sizeof(ih),
5817 &action, &reason, AF_INET6)) {
5818 log = action != PF_PASS;
5821 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5822 ntohs(h->ip6_plen), IPPROTO_ICMPV6, AF_INET6)) {
5826 action = pf_test_state_icmp(&s, dir, kif,
5828 if (action == PF_PASS) {
5830 pfsync_update_state(s);
5835 } else if (s == NULL)
5836 action = pf_test_icmp(&r, &s, dir, kif,
5837 m, off, h, &pd, &a, &ruleset);
5842 action = pf_test_state_other(&s, dir, kif, &pd);
5843 if (action == PF_PASS) {
5847 } else if (s == NULL)
5848 action = pf_test_other(&r, &s, dir, kif, m, off, h,
5854 /* XXX handle IPv6 options, if not allowed. not implemented. */
5857 if (action == PF_PASS && r->qid) {
5858 m->m_pkthdr.fw_flags |= ALTQ_MBUF_TAGGED;
5859 if (pd.tos == IPTOS_LOWDELAY)
5860 m->m_pkthdr.altq_qid = r->pqid;
5862 m->m_pkthdr.altq_qid = r->qid;
5864 KKASSERT(s->hash != 0);
5865 m->m_pkthdr.fw_flags |= ALTQ_MBUF_STATE_HASHED;
5866 m->m_pkthdr.altq_state_hash = s->hash;
5868 m->m_pkthdr.ecn_af = AF_INET6;
5869 m->m_pkthdr.header = h;
5873 if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
5874 pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
5875 (s->nat_rule.ptr->action == PF_RDR ||
5876 s->nat_rule.ptr->action == PF_BINAT) &&
5877 IN6_IS_ADDR_LOOPBACK(&pd.dst->v6)) {
5879 REASON_SET(&reason, PFRES_MEMORY);
5882 m->m_pkthdr.fw_flags |= PF_MBUF_TRANSLATE_LOCALHOST;
5885 PFLOG_PACKET(kif, h, m, AF_INET6, dir, reason, r, a, ruleset);
5887 kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
5888 kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++;
5890 if (action == PF_PASS || r->action == PF_DROP) {
5892 r->bytes += pd.tot_len;
5895 a->bytes += pd.tot_len;
5898 dirndx = (dir == s->direction) ? 0 : 1;
5899 s->packets[dirndx]++;
5900 s->bytes[dirndx] += pd.tot_len;
5901 if (s->nat_rule.ptr != NULL) {
5902 s->nat_rule.ptr->packets++;
5903 s->nat_rule.ptr->bytes += pd.tot_len;
5905 if (s->src_node != NULL) {
5906 s->src_node->packets++;
5907 s->src_node->bytes += pd.tot_len;
5909 if (s->nat_src_node != NULL) {
5910 s->nat_src_node->packets++;
5911 s->nat_src_node->bytes += pd.tot_len;
5915 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
5919 * XXX: we need to make sure that the addresses
5920 * passed to pfr_update_stats() are the same than
5921 * the addresses used during matching (pfr_match)
5923 if (r == &pf_default_rule) {
5925 x = (s == NULL || s->direction == dir) ?
5926 &pd.baddr : &pd.naddr;
5928 x = (s == NULL || s->direction == dir) ?
5929 &pd.naddr : &pd.baddr;
5931 if (x == &pd.baddr || s == NULL) {
5938 if (tr->src.addr.type == PF_ADDR_TABLE)
5939 pfr_update_stats(tr->src.addr.p.tbl, (s == NULL ||
5940 s->direction == dir) ? pd.src : pd.dst, pd.af,
5941 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5943 if (tr->dst.addr.type == PF_ADDR_TABLE)
5944 pfr_update_stats(tr->dst.addr.p.tbl, (s == NULL ||
5945 s->direction == dir) ? pd.dst : pd.src, pd.af,
5946 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5951 if (action == PF_SYNPROXY_DROP) {
5956 /* pf_route6 can free the mbuf causing *m0 to become NULL */
5957 pf_route6(m0, r, dir, ifp, s);