1/*
2 * Copyright (c) 2007-2018 Apple Inc. All rights reserved.
3 *
4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5 *
6 * This file contains Original Code and/or Modifications of Original Code
7 * as defined in and that are subject to the Apple Public Source License
8 * Version 2.0 (the 'License'). You may not use this file except in
9 * compliance with the License. The rights granted to you under the License
10 * may not be used to create, or enable the creation or redistribution of,
11 * unlawful or unlicensed copies of an Apple operating system, or to
12 * circumvent, violate, or enable the circumvention or violation of, any
13 * terms of an Apple operating system software license agreement.
14 *
15 * Please obtain a copy of the License at
16 * http://www.opensource.apple.com/apsl/ and read it before using this file.
17 *
18 * The Original Code and all software distributed under the License are
19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23 * Please see the License for the specific language governing rights and
24 * limitations under the License.
25 *
26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27 */
28
29/* $apfw: git commit 6602420f2f101b74305cd78f7cd9e0c8fdedae97 $ */
30/* $OpenBSD: pf.c,v 1.567 2008/02/20 23:40:13 henning Exp $ */
31
32/*
33 * Copyright (c) 2001 Daniel Hartmeier
34 * Copyright (c) 2002 - 2013 Henning Brauer
35 * NAT64 - Copyright (c) 2010 Viagenie Inc. (http://www.viagenie.ca)
36 * All rights reserved.
37 *
38 * Redistribution and use in source and binary forms, with or without
39 * modification, are permitted provided that the following conditions
40 * are met:
41 *
42 * - Redistributions of source code must retain the above copyright
43 * notice, this list of conditions and the following disclaimer.
44 * - Redistributions in binary form must reproduce the above
45 * copyright notice, this list of conditions and the following
46 * disclaimer in the documentation and/or other materials provided
47 * with the distribution.
48 *
49 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
50 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
51 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
52 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
53 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
54 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
55 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
56 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
57 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
59 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
60 * POSSIBILITY OF SUCH DAMAGE.
61 *
62 * Effort sponsored in part by the Defense Advanced Research Projects
63 * Agency (DARPA) and Air Force Research Laboratory, Air Force
64 * Materiel Command, USAF, under agreement number F30602-01-2-0537.
65 *
66 */
67
68#include <machine/endian.h>
69#include <sys/param.h>
70#include <sys/systm.h>
71#include <sys/filio.h>
72#include <sys/socket.h>
73#include <sys/socketvar.h>
74#include <sys/kernel.h>
75#include <sys/time.h>
76#include <sys/proc.h>
77#include <sys/random.h>
78#include <sys/mcache.h>
79#include <sys/protosw.h>
80
81#include <libkern/crypto/md5.h>
82#include <libkern/libkern.h>
83
84#include <mach/thread_act.h>
85
86#include <net/if.h>
87#include <net/if_types.h>
88#include <net/bpf.h>
89#include <net/route.h>
90#include <net/dlil.h>
91
92#include <netinet/in.h>
93#include <netinet/in_var.h>
94#include <netinet/in_systm.h>
95#include <netinet/ip.h>
96#include <netinet/ip_var.h>
97#include <netinet/tcp.h>
98#include <netinet/tcp_seq.h>
99#include <netinet/udp.h>
100#include <netinet/ip_icmp.h>
101#include <netinet/in_pcb.h>
102#include <netinet/tcp_timer.h>
103#include <netinet/tcp_var.h>
104#include <netinet/tcp_fsm.h>
105#include <netinet/udp_var.h>
106#include <netinet/icmp_var.h>
107#include <net/if_ether.h>
108#include <net/ethernet.h>
109#include <net/flowhash.h>
110#include <net/nat464_utils.h>
111#include <net/pfvar.h>
112#include <net/if_pflog.h>
113
114#if NPFSYNC
115#include <net/if_pfsync.h>
116#endif /* NPFSYNC */
117
118#if INET6
119#include <netinet/ip6.h>
120#include <netinet6/in6_pcb.h>
121#include <netinet6/ip6_var.h>
122#include <netinet/icmp6.h>
123#include <netinet6/nd6.h>
124#endif /* INET6 */
125
126#if DUMMYNET
127#include <netinet/ip_dummynet.h>
128#endif /* DUMMYNET */
129
130/*
131 * For RandomULong(), to get a 32 bits random value
132 * Note that random() returns a 31 bits value, see rdar://11159750
133 */
134#include <dev/random/randomdev.h>
135
136#define DPFPRINTF(n, x) (pf_status.debug >= (n) ? printf x : ((void)0))
137
138/*
139 * On Mac OS X, the rtableid value is treated as the interface scope
140 * value that is equivalent to the interface index used for scoped
141 * routing. A valid scope value is anything but IFSCOPE_NONE (0),
142 * as per definition of ifindex which is a positive, non-zero number.
143 * The other BSDs treat a negative rtableid value as invalid, hence
144 * the test against INT_MAX to handle userland apps which initialize
145 * the field with a negative number.
146 */
147#define PF_RTABLEID_IS_VALID(r) \
148 ((r) > IFSCOPE_NONE && (r) <= INT_MAX)
149
150/*
151 * Global variables
152 */
153decl_lck_mtx_data(,pf_lock_data);
154decl_lck_rw_data(,pf_perim_lock_data);
155lck_mtx_t *pf_lock = &pf_lock_data;
156lck_rw_t *pf_perim_lock = &pf_perim_lock_data;
157
158/* state tables */
159struct pf_state_tree_lan_ext pf_statetbl_lan_ext;
160struct pf_state_tree_ext_gwy pf_statetbl_ext_gwy;
161
162struct pf_palist pf_pabuf;
163struct pf_status pf_status;
164
165u_int32_t ticket_pabuf;
166
167static MD5_CTX pf_tcp_secret_ctx;
168static u_char pf_tcp_secret[16];
169static int pf_tcp_secret_init;
170static int pf_tcp_iss_off;
171
172static struct pf_anchor_stackframe {
173 struct pf_ruleset *rs;
174 struct pf_rule *r;
175 struct pf_anchor_node *parent;
176 struct pf_anchor *child;
177} pf_anchor_stack[64];
178
179struct pool pf_src_tree_pl, pf_rule_pl, pf_pooladdr_pl;
180struct pool pf_state_pl, pf_state_key_pl;
181
182typedef void (*hook_fn_t)(void *);
183
184struct hook_desc {
185 TAILQ_ENTRY(hook_desc) hd_list;
186 hook_fn_t hd_fn;
187 void *hd_arg;
188};
189
190#define HOOK_REMOVE 0x01
191#define HOOK_FREE 0x02
192#define HOOK_ABORT 0x04
193
194static void *hook_establish(struct hook_desc_head *, int,
195 hook_fn_t, void *);
196static void hook_runloop(struct hook_desc_head *, int flags);
197
198struct pool pf_app_state_pl;
199static void pf_print_addr(struct pf_addr *addr, sa_family_t af);
200static void pf_print_sk_host(struct pf_state_host *, u_int8_t, int,
201 u_int8_t);
202
203static void pf_print_host(struct pf_addr *, u_int16_t, u_int8_t);
204
205static void pf_init_threshold(struct pf_threshold *, u_int32_t,
206 u_int32_t);
207static void pf_add_threshold(struct pf_threshold *);
208static int pf_check_threshold(struct pf_threshold *);
209
210static void pf_change_ap(int, pbuf_t *, struct pf_addr *,
211 u_int16_t *, u_int16_t *, u_int16_t *,
212 struct pf_addr *, u_int16_t, u_int8_t, sa_family_t,
213 sa_family_t, int);
214static int pf_modulate_sack(pbuf_t *, int, struct pf_pdesc *,
215 struct tcphdr *, struct pf_state_peer *);
216#if INET6
217static void pf_change_a6(struct pf_addr *, u_int16_t *,
218 struct pf_addr *, u_int8_t);
219void pf_change_addr(struct pf_addr *a, u_int16_t *c,
220 struct pf_addr *an, u_int8_t u,
221 sa_family_t af, sa_family_t afn);
222#endif /* INET6 */
223static void pf_change_icmp(struct pf_addr *, u_int16_t *,
224 struct pf_addr *, struct pf_addr *, u_int16_t,
225 u_int16_t *, u_int16_t *, u_int16_t *,
226 u_int16_t *, u_int8_t, sa_family_t);
227static void pf_send_tcp(const struct pf_rule *, sa_family_t,
228 const struct pf_addr *, const struct pf_addr *,
229 u_int16_t, u_int16_t, u_int32_t, u_int32_t,
230 u_int8_t, u_int16_t, u_int16_t, u_int8_t, int,
231 u_int16_t, struct ether_header *, struct ifnet *);
232static void pf_send_icmp(pbuf_t *, u_int8_t, u_int8_t,
233 sa_family_t, struct pf_rule *);
234static struct pf_rule *pf_match_translation(struct pf_pdesc *, pbuf_t *,
235 int, int, struct pfi_kif *, struct pf_addr *,
236 union pf_state_xport *, struct pf_addr *,
237 union pf_state_xport *, int);
238static struct pf_rule *pf_get_translation_aux(struct pf_pdesc *,
239 pbuf_t *, int, int, struct pfi_kif *,
240 struct pf_src_node **, struct pf_addr *,
241 union pf_state_xport *, struct pf_addr *,
242 union pf_state_xport *, union pf_state_xport *
243 );
244static void pf_attach_state(struct pf_state_key *,
245 struct pf_state *, int);
246static void pf_detach_state(struct pf_state *, int);
247static u_int32_t pf_tcp_iss(struct pf_pdesc *);
248static int pf_test_rule(struct pf_rule **, struct pf_state **,
249 int, struct pfi_kif *, pbuf_t *, int,
250 void *, struct pf_pdesc *, struct pf_rule **,
251 struct pf_ruleset **, struct ifqueue *);
252#if DUMMYNET
253static int pf_test_dummynet(struct pf_rule **, int,
254 struct pfi_kif *, pbuf_t **,
255 struct pf_pdesc *, struct ip_fw_args *);
256#endif /* DUMMYNET */
257static int pf_test_fragment(struct pf_rule **, int,
258 struct pfi_kif *, pbuf_t *, void *,
259 struct pf_pdesc *, struct pf_rule **,
260 struct pf_ruleset **);
261static int pf_test_state_tcp(struct pf_state **, int,
262 struct pfi_kif *, pbuf_t *, int,
263 void *, struct pf_pdesc *, u_short *);
264static int pf_test_state_udp(struct pf_state **, int,
265 struct pfi_kif *, pbuf_t *, int,
266 void *, struct pf_pdesc *, u_short *);
267static int pf_test_state_icmp(struct pf_state **, int,
268 struct pfi_kif *, pbuf_t *, int,
269 void *, struct pf_pdesc *, u_short *);
270static int pf_test_state_other(struct pf_state **, int,
271 struct pfi_kif *, struct pf_pdesc *);
272static int pf_match_tag(struct pf_rule *,
273 struct pf_mtag *, int *);
274static void pf_hash(struct pf_addr *, struct pf_addr *,
275 struct pf_poolhashkey *, sa_family_t);
276static int pf_map_addr(u_int8_t, struct pf_rule *,
277 struct pf_addr *, struct pf_addr *,
278 struct pf_addr *, struct pf_src_node **);
279static int pf_get_sport(struct pf_pdesc *, struct pfi_kif *,
280 struct pf_rule *, struct pf_addr *,
281 union pf_state_xport *, struct pf_addr *,
282 union pf_state_xport *, struct pf_addr *,
283 union pf_state_xport *, struct pf_src_node **
284 );
285static void pf_route(pbuf_t **, struct pf_rule *, int,
286 struct ifnet *, struct pf_state *,
287 struct pf_pdesc *);
288#if INET6
289static void pf_route6(pbuf_t **, struct pf_rule *, int,
290 struct ifnet *, struct pf_state *,
291 struct pf_pdesc *);
292#endif /* INET6 */
293static u_int8_t pf_get_wscale(pbuf_t *, int, u_int16_t,
294 sa_family_t);
295static u_int16_t pf_get_mss(pbuf_t *, int, u_int16_t,
296 sa_family_t);
297static u_int16_t pf_calc_mss(struct pf_addr *, sa_family_t,
298 u_int16_t);
299static void pf_set_rt_ifp(struct pf_state *,
300 struct pf_addr *, sa_family_t af);
301static int pf_check_proto_cksum(pbuf_t *, int, int,
302 u_int8_t, sa_family_t);
303static int pf_addr_wrap_neq(struct pf_addr_wrap *,
304 struct pf_addr_wrap *);
305static struct pf_state *pf_find_state(struct pfi_kif *,
306 struct pf_state_key_cmp *, u_int);
307static int pf_src_connlimit(struct pf_state **);
308static void pf_stateins_err(const char *, struct pf_state *,
309 struct pfi_kif *);
310static int pf_check_congestion(struct ifqueue *);
311
312#if 0
313static const char *pf_pptp_ctrl_type_name(u_int16_t code);
314#endif
315static void pf_pptp_handler(struct pf_state *, int, int,
316 struct pf_pdesc *, struct pfi_kif *);
317static void pf_pptp_unlink(struct pf_state *);
318static void pf_grev1_unlink(struct pf_state *);
319static int pf_test_state_grev1(struct pf_state **, int,
320 struct pfi_kif *, int, struct pf_pdesc *);
321static int pf_ike_compare(struct pf_app_state *,
322 struct pf_app_state *);
323static int pf_test_state_esp(struct pf_state **, int,
324 struct pfi_kif *, int, struct pf_pdesc *);
325
326extern struct pool pfr_ktable_pl;
327extern struct pool pfr_kentry_pl;
328extern int path_mtu_discovery;
329
330struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] = {
331 { &pf_state_pl, PFSTATE_HIWAT },
332 { &pf_app_state_pl, PFAPPSTATE_HIWAT },
333 { &pf_src_tree_pl, PFSNODE_HIWAT },
334 { &pf_frent_pl, PFFRAG_FRENT_HIWAT },
335 { &pfr_ktable_pl, PFR_KTABLE_HIWAT },
336 { &pfr_kentry_pl, PFR_KENTRY_HIWAT },
337};
338
339void *
340pf_lazy_makewritable(struct pf_pdesc *pd, pbuf_t *pbuf, int len)
341{
342 void *p;
343
344 if (pd->lmw < 0)
345 return (NULL);
346
347 VERIFY(pbuf == pd->mp);
348
349 p = pbuf->pb_data;
350 if (len > pd->lmw) {
351 if ((p = pbuf_ensure_writable(pbuf, len)) == NULL)
352 len = -1;
353 pd->lmw = len;
354 if (len >= 0) {
355 pd->pf_mtag = pf_find_mtag_pbuf(pbuf);
356
357 switch (pd->af) {
358 case AF_INET: {
359 struct ip *h = p;
360 pd->src = (struct pf_addr *)(uintptr_t)&h->ip_src;
361 pd->dst = (struct pf_addr *)(uintptr_t)&h->ip_dst;
362 pd->ip_sum = &h->ip_sum;
363 break;
364 }
365#if INET6
366 case AF_INET6: {
367 struct ip6_hdr *h = p;
368 pd->src = (struct pf_addr *)(uintptr_t)&h->ip6_src;
369 pd->dst = (struct pf_addr *)(uintptr_t)&h->ip6_dst;
370 break;
371 }
372#endif /* INET6 */
373 }
374 }
375 }
376
377 return (len < 0 ? NULL : p);
378}
379
380static const int *
381pf_state_lookup_aux(struct pf_state **state, struct pfi_kif *kif,
382 int direction, int *action)
383{
384 if (*state == NULL || (*state)->timeout == PFTM_PURGE) {
385 *action = PF_DROP;
386 return (action);
387 }
388
389 if (direction == PF_OUT &&
390 (((*state)->rule.ptr->rt == PF_ROUTETO &&
391 (*state)->rule.ptr->direction == PF_OUT) ||
392 ((*state)->rule.ptr->rt == PF_REPLYTO &&
393 (*state)->rule.ptr->direction == PF_IN)) &&
394 (*state)->rt_kif != NULL && (*state)->rt_kif != kif) {
395 *action = PF_PASS;
396 return (action);
397 }
398
399 return (0);
400}
401
402#define STATE_LOOKUP() \
403 do { \
404 int action; \
405 *state = pf_find_state(kif, &key, direction); \
406 if (*state != NULL && pd != NULL && \
407 !(pd->pktflags & PKTF_FLOW_ID)) { \
408 pd->flowsrc = (*state)->state_key->flowsrc; \
409 pd->flowhash = (*state)->state_key->flowhash; \
410 if (pd->flowhash != 0) { \
411 pd->pktflags |= PKTF_FLOW_ID; \
412 pd->pktflags &= ~PKTF_FLOW_ADV; \
413 } \
414 } \
415 if (pf_state_lookup_aux(state, kif, direction, &action)) \
416 return (action); \
417 } while (0)
418
419#define STATE_ADDR_TRANSLATE(sk) \
420 (sk)->lan.addr.addr32[0] != (sk)->gwy.addr.addr32[0] || \
421 ((sk)->af_lan == AF_INET6 && \
422 ((sk)->lan.addr.addr32[1] != (sk)->gwy.addr.addr32[1] || \
423 (sk)->lan.addr.addr32[2] != (sk)->gwy.addr.addr32[2] || \
424 (sk)->lan.addr.addr32[3] != (sk)->gwy.addr.addr32[3]))
425
426#define STATE_TRANSLATE(sk) \
427 ((sk)->af_lan != (sk)->af_gwy || \
428 STATE_ADDR_TRANSLATE(sk) || \
429 (sk)->lan.xport.port != (sk)->gwy.xport.port)
430
431#define STATE_GRE_TRANSLATE(sk) \
432 (STATE_ADDR_TRANSLATE(sk) || \
433 (sk)->lan.xport.call_id != (sk)->gwy.xport.call_id)
434
435#define BOUND_IFACE(r, k) \
436 ((r)->rule_flag & PFRULE_IFBOUND) ? (k) : pfi_all
437
438#define STATE_INC_COUNTERS(s) \
439 do { \
440 s->rule.ptr->states++; \
441 VERIFY(s->rule.ptr->states != 0); \
442 if (s->anchor.ptr != NULL) { \
443 s->anchor.ptr->states++; \
444 VERIFY(s->anchor.ptr->states != 0); \
445 } \
446 if (s->nat_rule.ptr != NULL) { \
447 s->nat_rule.ptr->states++; \
448 VERIFY(s->nat_rule.ptr->states != 0); \
449 } \
450 } while (0)
451
452#define STATE_DEC_COUNTERS(s) \
453 do { \
454 if (s->nat_rule.ptr != NULL) { \
455 VERIFY(s->nat_rule.ptr->states > 0); \
456 s->nat_rule.ptr->states--; \
457 } \
458 if (s->anchor.ptr != NULL) { \
459 VERIFY(s->anchor.ptr->states > 0); \
460 s->anchor.ptr->states--; \
461 } \
462 VERIFY(s->rule.ptr->states > 0); \
463 s->rule.ptr->states--; \
464 } while (0)
465
466static __inline int pf_src_compare(struct pf_src_node *, struct pf_src_node *);
467static __inline int pf_state_compare_lan_ext(struct pf_state_key *,
468 struct pf_state_key *);
469static __inline int pf_state_compare_ext_gwy(struct pf_state_key *,
470 struct pf_state_key *);
471static __inline int pf_state_compare_id(struct pf_state *,
472 struct pf_state *);
473
474struct pf_src_tree tree_src_tracking;
475
476struct pf_state_tree_id tree_id;
477struct pf_state_queue state_list;
478
479RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare);
480RB_GENERATE(pf_state_tree_lan_ext, pf_state_key,
481 entry_lan_ext, pf_state_compare_lan_ext);
482RB_GENERATE(pf_state_tree_ext_gwy, pf_state_key,
483 entry_ext_gwy, pf_state_compare_ext_gwy);
484RB_GENERATE(pf_state_tree_id, pf_state,
485 entry_id, pf_state_compare_id);
486
487#define PF_DT_SKIP_LANEXT 0x01
488#define PF_DT_SKIP_EXTGWY 0x02
489
490static const u_int16_t PF_PPTP_PORT = 1723;
491static const u_int32_t PF_PPTP_MAGIC_NUMBER = 0x1A2B3C4D;
492
493struct pf_pptp_hdr {
494 u_int16_t length;
495 u_int16_t type;
496 u_int32_t magic;
497};
498
499struct pf_pptp_ctrl_hdr {
500 u_int16_t type;
501 u_int16_t reserved_0;
502};
503
504struct pf_pptp_ctrl_generic {
505 u_int16_t data[0];
506};
507
508#define PF_PPTP_CTRL_TYPE_START_REQ 1
509struct pf_pptp_ctrl_start_req {
510 u_int16_t protocol_version;
511 u_int16_t reserved_1;
512 u_int32_t framing_capabilities;
513 u_int32_t bearer_capabilities;
514 u_int16_t maximum_channels;
515 u_int16_t firmware_revision;
516 u_int8_t host_name[64];
517 u_int8_t vendor_string[64];
518};
519
520#define PF_PPTP_CTRL_TYPE_START_RPY 2
521struct pf_pptp_ctrl_start_rpy {
522 u_int16_t protocol_version;
523 u_int8_t result_code;
524 u_int8_t error_code;
525 u_int32_t framing_capabilities;
526 u_int32_t bearer_capabilities;
527 u_int16_t maximum_channels;
528 u_int16_t firmware_revision;
529 u_int8_t host_name[64];
530 u_int8_t vendor_string[64];
531};
532
533#define PF_PPTP_CTRL_TYPE_STOP_REQ 3
534struct pf_pptp_ctrl_stop_req {
535 u_int8_t reason;
536 u_int8_t reserved_1;
537 u_int16_t reserved_2;
538};
539
540#define PF_PPTP_CTRL_TYPE_STOP_RPY 4
541struct pf_pptp_ctrl_stop_rpy {
542 u_int8_t reason;
543 u_int8_t error_code;
544 u_int16_t reserved_1;
545};
546
547#define PF_PPTP_CTRL_TYPE_ECHO_REQ 5
548struct pf_pptp_ctrl_echo_req {
549 u_int32_t identifier;
550};
551
552#define PF_PPTP_CTRL_TYPE_ECHO_RPY 6
553struct pf_pptp_ctrl_echo_rpy {
554 u_int32_t identifier;
555 u_int8_t result_code;
556 u_int8_t error_code;
557 u_int16_t reserved_1;
558};
559
560#define PF_PPTP_CTRL_TYPE_CALL_OUT_REQ 7
561struct pf_pptp_ctrl_call_out_req {
562 u_int16_t call_id;
563 u_int16_t call_sernum;
564 u_int32_t min_bps;
565 u_int32_t bearer_type;
566 u_int32_t framing_type;
567 u_int16_t rxwindow_size;
568 u_int16_t proc_delay;
569 u_int8_t phone_num[64];
570 u_int8_t sub_addr[64];
571};
572
573#define PF_PPTP_CTRL_TYPE_CALL_OUT_RPY 8
574struct pf_pptp_ctrl_call_out_rpy {
575 u_int16_t call_id;
576 u_int16_t peer_call_id;
577 u_int8_t result_code;
578 u_int8_t error_code;
579 u_int16_t cause_code;
580 u_int32_t connect_speed;
581 u_int16_t rxwindow_size;
582 u_int16_t proc_delay;
583 u_int32_t phy_channel_id;
584};
585
586#define PF_PPTP_CTRL_TYPE_CALL_IN_1ST 9
587struct pf_pptp_ctrl_call_in_1st {
588 u_int16_t call_id;
589 u_int16_t call_sernum;
590 u_int32_t bearer_type;
591 u_int32_t phy_channel_id;
592 u_int16_t dialed_number_len;
593 u_int16_t dialing_number_len;
594 u_int8_t dialed_num[64];
595 u_int8_t dialing_num[64];
596 u_int8_t sub_addr[64];
597};
598
599#define PF_PPTP_CTRL_TYPE_CALL_IN_2ND 10
600struct pf_pptp_ctrl_call_in_2nd {
601 u_int16_t call_id;
602 u_int16_t peer_call_id;
603 u_int8_t result_code;
604 u_int8_t error_code;
605 u_int16_t rxwindow_size;
606 u_int16_t txdelay;
607 u_int16_t reserved_1;
608};
609
610#define PF_PPTP_CTRL_TYPE_CALL_IN_3RD 11
611struct pf_pptp_ctrl_call_in_3rd {
612 u_int16_t call_id;
613 u_int16_t reserved_1;
614 u_int32_t connect_speed;
615 u_int16_t rxwindow_size;
616 u_int16_t txdelay;
617 u_int32_t framing_type;
618};
619
620#define PF_PPTP_CTRL_TYPE_CALL_CLR 12
621struct pf_pptp_ctrl_call_clr {
622 u_int16_t call_id;
623 u_int16_t reserved_1;
624};
625
626#define PF_PPTP_CTRL_TYPE_CALL_DISC 13
627struct pf_pptp_ctrl_call_disc {
628 u_int16_t call_id;
629 u_int8_t result_code;
630 u_int8_t error_code;
631 u_int16_t cause_code;
632 u_int16_t reserved_1;
633 u_int8_t statistics[128];
634};
635
636#define PF_PPTP_CTRL_TYPE_ERROR 14
637struct pf_pptp_ctrl_error {
638 u_int16_t peer_call_id;
639 u_int16_t reserved_1;
640 u_int32_t crc_errors;
641 u_int32_t fr_errors;
642 u_int32_t hw_errors;
643 u_int32_t buf_errors;
644 u_int32_t tim_errors;
645 u_int32_t align_errors;
646};
647
648#define PF_PPTP_CTRL_TYPE_SET_LINKINFO 15
649struct pf_pptp_ctrl_set_linkinfo {
650 u_int16_t peer_call_id;
651 u_int16_t reserved_1;
652 u_int32_t tx_accm;
653 u_int32_t rx_accm;
654};
655
656#if 0
657static const char *pf_pptp_ctrl_type_name(u_int16_t code)
658{
659 code = ntohs(code);
660
661 if (code < PF_PPTP_CTRL_TYPE_START_REQ ||
662 code > PF_PPTP_CTRL_TYPE_SET_LINKINFO) {
663 static char reserved[] = "reserved-00";
664
665 sprintf(&reserved[9], "%02x", code);
666 return (reserved);
667 } else {
668 static const char *name[] = {
669 "start_req", "start_rpy", "stop_req", "stop_rpy",
670 "echo_req", "echo_rpy", "call_out_req", "call_out_rpy",
671 "call_in_1st", "call_in_2nd", "call_in_3rd",
672 "call_clr", "call_disc", "error", "set_linkinfo"
673 };
674
675 return (name[code - 1]);
676 }
677};
678#endif
679
680static const size_t PF_PPTP_CTRL_MSG_MINSIZE =
681 sizeof (struct pf_pptp_hdr) + sizeof (struct pf_pptp_ctrl_hdr);
682
683union pf_pptp_ctrl_msg_union {
684 struct pf_pptp_ctrl_start_req start_req;
685 struct pf_pptp_ctrl_start_rpy start_rpy;
686 struct pf_pptp_ctrl_stop_req stop_req;
687 struct pf_pptp_ctrl_stop_rpy stop_rpy;
688 struct pf_pptp_ctrl_echo_req echo_req;
689 struct pf_pptp_ctrl_echo_rpy echo_rpy;
690 struct pf_pptp_ctrl_call_out_req call_out_req;
691 struct pf_pptp_ctrl_call_out_rpy call_out_rpy;
692 struct pf_pptp_ctrl_call_in_1st call_in_1st;
693 struct pf_pptp_ctrl_call_in_2nd call_in_2nd;
694 struct pf_pptp_ctrl_call_in_3rd call_in_3rd;
695 struct pf_pptp_ctrl_call_clr call_clr;
696 struct pf_pptp_ctrl_call_disc call_disc;
697 struct pf_pptp_ctrl_error error;
698 struct pf_pptp_ctrl_set_linkinfo set_linkinfo;
699 u_int8_t data[0];
700};
701
702struct pf_pptp_ctrl_msg {
703 struct pf_pptp_hdr hdr;
704 struct pf_pptp_ctrl_hdr ctrl;
705 union pf_pptp_ctrl_msg_union msg;
706};
707
708#define PF_GRE_FLAG_CHECKSUM_PRESENT 0x8000
709#define PF_GRE_FLAG_VERSION_MASK 0x0007
710#define PF_GRE_PPP_ETHERTYPE 0x880B
711
712struct pf_grev1_hdr {
713 u_int16_t flags;
714 u_int16_t protocol_type;
715 u_int16_t payload_length;
716 u_int16_t call_id;
717 /*
718 u_int32_t seqno;
719 u_int32_t ackno;
720 */
721};
722
723static const u_int16_t PF_IKE_PORT = 500;
724
725struct pf_ike_hdr {
726 u_int64_t initiator_cookie, responder_cookie;
727 u_int8_t next_payload, version, exchange_type, flags;
728 u_int32_t message_id, length;
729};
730
731#define PF_IKE_PACKET_MINSIZE (sizeof (struct pf_ike_hdr))
732
733#define PF_IKEv1_EXCHTYPE_BASE 1
734#define PF_IKEv1_EXCHTYPE_ID_PROTECT 2
735#define PF_IKEv1_EXCHTYPE_AUTH_ONLY 3
736#define PF_IKEv1_EXCHTYPE_AGGRESSIVE 4
737#define PF_IKEv1_EXCHTYPE_INFORMATIONAL 5
738#define PF_IKEv2_EXCHTYPE_SA_INIT 34
739#define PF_IKEv2_EXCHTYPE_AUTH 35
740#define PF_IKEv2_EXCHTYPE_CREATE_CHILD_SA 36
741#define PF_IKEv2_EXCHTYPE_INFORMATIONAL 37
742
743#define PF_IKEv1_FLAG_E 0x01
744#define PF_IKEv1_FLAG_C 0x02
745#define PF_IKEv1_FLAG_A 0x04
746#define PF_IKEv2_FLAG_I 0x08
747#define PF_IKEv2_FLAG_V 0x10
748#define PF_IKEv2_FLAG_R 0x20
749
750struct pf_esp_hdr {
751 u_int32_t spi;
752 u_int32_t seqno;
753 u_int8_t payload[];
754};
755
756static __inline int
757pf_addr_compare(struct pf_addr *a, struct pf_addr *b, sa_family_t af)
758{
759 switch (af) {
760#ifdef INET
761 case AF_INET:
762 if (a->addr32[0] > b->addr32[0])
763 return (1);
764 if (a->addr32[0] < b->addr32[0])
765 return (-1);
766 break;
767#endif /* INET */
768#ifdef INET6
769 case AF_INET6:
770 if (a->addr32[3] > b->addr32[3])
771 return (1);
772 if (a->addr32[3] < b->addr32[3])
773 return (-1);
774 if (a->addr32[2] > b->addr32[2])
775 return (1);
776 if (a->addr32[2] < b->addr32[2])
777 return (-1);
778 if (a->addr32[1] > b->addr32[1])
779 return (1);
780 if (a->addr32[1] < b->addr32[1])
781 return (-1);
782 if (a->addr32[0] > b->addr32[0])
783 return (1);
784 if (a->addr32[0] < b->addr32[0])
785 return (-1);
786 break;
787#endif /* INET6 */
788 }
789 return (0);
790}
791
792static __inline int
793pf_src_compare(struct pf_src_node *a, struct pf_src_node *b)
794{
795 int diff;
796
797 if (a->rule.ptr > b->rule.ptr)
798 return (1);
799 if (a->rule.ptr < b->rule.ptr)
800 return (-1);
801 if ((diff = a->af - b->af) != 0)
802 return (diff);
803 if ((diff = pf_addr_compare(&a->addr, &b->addr, a->af)) != 0)
804 return (diff);
805 return (0);
806}
807
808static __inline int
809pf_state_compare_lan_ext(struct pf_state_key *a, struct pf_state_key *b)
810{
811 int diff;
812 int extfilter;
813
814 if ((diff = a->proto - b->proto) != 0)
815 return (diff);
816 if ((diff = a->af_lan - b->af_lan) != 0)
817 return (diff);
818
819 extfilter = PF_EXTFILTER_APD;
820
821 switch (a->proto) {
822 case IPPROTO_ICMP:
823 case IPPROTO_ICMPV6:
824 if ((diff = a->lan.xport.port - b->lan.xport.port) != 0)
825 return (diff);
826 break;
827
828 case IPPROTO_TCP:
829 if ((diff = a->lan.xport.port - b->lan.xport.port) != 0)
830 return (diff);
831 if ((diff = a->ext_lan.xport.port - b->ext_lan.xport.port) != 0)
832 return (diff);
833 break;
834
835 case IPPROTO_UDP:
836 if ((diff = a->proto_variant - b->proto_variant))
837 return (diff);
838 extfilter = a->proto_variant;
839 if ((diff = a->lan.xport.port - b->lan.xport.port) != 0)
840 return (diff);
841 if ((extfilter < PF_EXTFILTER_AD) &&
842 (diff = a->ext_lan.xport.port - b->ext_lan.xport.port) != 0)
843 return (diff);
844 break;
845
846 case IPPROTO_GRE:
847 if (a->proto_variant == PF_GRE_PPTP_VARIANT &&
848 a->proto_variant == b->proto_variant) {
849 if (!!(diff = a->ext_lan.xport.call_id -
850 b->ext_lan.xport.call_id))
851 return (diff);
852 }
853 break;
854
855 case IPPROTO_ESP:
856 if (!!(diff = a->ext_lan.xport.spi - b->ext_lan.xport.spi))
857 return (diff);
858 break;
859
860 default:
861 break;
862 }
863
864 switch (a->af_lan) {
865#if INET
866 case AF_INET:
867 if ((diff = pf_addr_compare(&a->lan.addr, &b->lan.addr,
868 a->af_lan)) != 0)
869 return (diff);
870
871 if (extfilter < PF_EXTFILTER_EI) {
872 if ((diff = pf_addr_compare(&a->ext_lan.addr,
873 &b->ext_lan.addr,
874 a->af_lan)) != 0)
875 return (diff);
876 }
877 break;
878#endif /* INET */
879#if INET6
880 case AF_INET6:
881 if ((diff = pf_addr_compare(&a->lan.addr, &b->lan.addr,
882 a->af_lan)) != 0)
883 return (diff);
884
885 if (extfilter < PF_EXTFILTER_EI ||
886 !PF_AZERO(&b->ext_lan.addr, AF_INET6)) {
887 if ((diff = pf_addr_compare(&a->ext_lan.addr,
888 &b->ext_lan.addr,
889 a->af_lan)) != 0)
890 return (diff);
891 }
892 break;
893#endif /* INET6 */
894 }
895
896 if (a->app_state && b->app_state) {
897 if (a->app_state->compare_lan_ext &&
898 b->app_state->compare_lan_ext) {
899 diff = (const char *)b->app_state->compare_lan_ext -
900 (const char *)a->app_state->compare_lan_ext;
901 if (diff != 0)
902 return (diff);
903 diff = a->app_state->compare_lan_ext(a->app_state,
904 b->app_state);
905 if (diff != 0)
906 return (diff);
907 }
908 }
909
910 return (0);
911}
912
913static __inline int
914pf_state_compare_ext_gwy(struct pf_state_key *a, struct pf_state_key *b)
915{
916 int diff;
917 int extfilter;
918
919 if ((diff = a->proto - b->proto) != 0)
920 return (diff);
921
922 if ((diff = a->af_gwy - b->af_gwy) != 0)
923 return (diff);
924
925 extfilter = PF_EXTFILTER_APD;
926
927 switch (a->proto) {
928 case IPPROTO_ICMP:
929 case IPPROTO_ICMPV6:
930 if ((diff = a->gwy.xport.port - b->gwy.xport.port) != 0)
931 return (diff);
932 break;
933
934 case IPPROTO_TCP:
935 if ((diff = a->ext_gwy.xport.port - b->ext_gwy.xport.port) != 0)
936 return (diff);
937 if ((diff = a->gwy.xport.port - b->gwy.xport.port) != 0)
938 return (diff);
939 break;
940
941 case IPPROTO_UDP:
942 if ((diff = a->proto_variant - b->proto_variant))
943 return (diff);
944 extfilter = a->proto_variant;
945 if ((diff = a->gwy.xport.port - b->gwy.xport.port) != 0)
946 return (diff);
947 if ((extfilter < PF_EXTFILTER_AD) &&
948 (diff = a->ext_gwy.xport.port - b->ext_gwy.xport.port) != 0)
949 return (diff);
950 break;
951
952 case IPPROTO_GRE:
953 if (a->proto_variant == PF_GRE_PPTP_VARIANT &&
954 a->proto_variant == b->proto_variant) {
955 if (!!(diff = a->gwy.xport.call_id -
956 b->gwy.xport.call_id))
957 return (diff);
958 }
959 break;
960
961 case IPPROTO_ESP:
962 if (!!(diff = a->gwy.xport.spi - b->gwy.xport.spi))
963 return (diff);
964 break;
965
966 default:
967 break;
968 }
969
970 switch (a->af_gwy) {
971#if INET
972 case AF_INET:
973 if ((diff = pf_addr_compare(&a->gwy.addr, &b->gwy.addr,
974 a->af_gwy)) != 0)
975 return (diff);
976
977 if (extfilter < PF_EXTFILTER_EI) {
978 if ((diff = pf_addr_compare(&a->ext_gwy.addr, &b->ext_gwy.addr,
979 a->af_gwy)) != 0)
980 return (diff);
981 }
982 break;
983#endif /* INET */
984#if INET6
985 case AF_INET6:
986 if ((diff = pf_addr_compare(&a->gwy.addr, &b->gwy.addr,
987 a->af_gwy)) != 0)
988 return (diff);
989
990 if (extfilter < PF_EXTFILTER_EI ||
991 !PF_AZERO(&b->ext_gwy.addr, AF_INET6)) {
992 if ((diff = pf_addr_compare(&a->ext_gwy.addr, &b->ext_gwy.addr,
993 a->af_gwy)) != 0)
994 return (diff);
995 }
996 break;
997#endif /* INET6 */
998 }
999
1000 if (a->app_state && b->app_state) {
1001 if (a->app_state->compare_ext_gwy &&
1002 b->app_state->compare_ext_gwy) {
1003 diff = (const char *)b->app_state->compare_ext_gwy -
1004 (const char *)a->app_state->compare_ext_gwy;
1005 if (diff != 0)
1006 return (diff);
1007 diff = a->app_state->compare_ext_gwy(a->app_state,
1008 b->app_state);
1009 if (diff != 0)
1010 return (diff);
1011 }
1012 }
1013
1014 return (0);
1015}
1016
1017static __inline int
1018pf_state_compare_id(struct pf_state *a, struct pf_state *b)
1019{
1020 if (a->id > b->id)
1021 return (1);
1022 if (a->id < b->id)
1023 return (-1);
1024 if (a->creatorid > b->creatorid)
1025 return (1);
1026 if (a->creatorid < b->creatorid)
1027 return (-1);
1028
1029 return (0);
1030}
1031
1032#if INET6
1033void
1034pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af)
1035{
1036 switch (af) {
1037#if INET
1038 case AF_INET:
1039 dst->addr32[0] = src->addr32[0];
1040 break;
1041#endif /* INET */
1042 case AF_INET6:
1043 dst->addr32[0] = src->addr32[0];
1044 dst->addr32[1] = src->addr32[1];
1045 dst->addr32[2] = src->addr32[2];
1046 dst->addr32[3] = src->addr32[3];
1047 break;
1048 }
1049}
1050#endif /* INET6 */
1051
1052struct pf_state *
1053pf_find_state_byid(struct pf_state_cmp *key)
1054{
1055 pf_status.fcounters[FCNT_STATE_SEARCH]++;
1056
1057 return (RB_FIND(pf_state_tree_id, &tree_id,
1058 (struct pf_state *)(void *)key));
1059}
1060
1061static struct pf_state *
1062pf_find_state(struct pfi_kif *kif, struct pf_state_key_cmp *key, u_int dir)
1063{
1064 struct pf_state_key *sk = NULL;
1065 struct pf_state *s;
1066
1067 pf_status.fcounters[FCNT_STATE_SEARCH]++;
1068
1069 switch (dir) {
1070 case PF_OUT:
1071 sk = RB_FIND(pf_state_tree_lan_ext, &pf_statetbl_lan_ext,
1072 (struct pf_state_key *)key);
1073 break;
1074 case PF_IN:
1075 sk = RB_FIND(pf_state_tree_ext_gwy, &pf_statetbl_ext_gwy,
1076 (struct pf_state_key *)key);
1077 /*
1078 * NAT64 is done only on input, for packets coming in from
1079 * from the LAN side, need to lookup the lan_ext tree.
1080 */
1081 if (sk == NULL) {
1082 sk = RB_FIND(pf_state_tree_lan_ext,
1083 &pf_statetbl_lan_ext,
1084 (struct pf_state_key *)key);
1085 if (sk && sk->af_lan == sk->af_gwy)
1086 sk = NULL;
1087 }
1088 break;
1089 default:
1090 panic("pf_find_state");
1091 }
1092
1093 /* list is sorted, if-bound states before floating ones */
1094 if (sk != NULL)
1095 TAILQ_FOREACH(s, &sk->states, next)
1096 if (s->kif == pfi_all || s->kif == kif)
1097 return (s);
1098
1099 return (NULL);
1100}
1101
1102struct pf_state *
1103pf_find_state_all(struct pf_state_key_cmp *key, u_int dir, int *more)
1104{
1105 struct pf_state_key *sk = NULL;
1106 struct pf_state *s, *ret = NULL;
1107
1108 pf_status.fcounters[FCNT_STATE_SEARCH]++;
1109
1110 switch (dir) {
1111 case PF_OUT:
1112 sk = RB_FIND(pf_state_tree_lan_ext,
1113 &pf_statetbl_lan_ext, (struct pf_state_key *)key);
1114 break;
1115 case PF_IN:
1116 sk = RB_FIND(pf_state_tree_ext_gwy,
1117 &pf_statetbl_ext_gwy, (struct pf_state_key *)key);
1118 /*
1119 * NAT64 is done only on input, for packets coming in from
1120 * from the LAN side, need to lookup the lan_ext tree.
1121 */
1122 if ((sk == NULL) && pf_nat64_configured) {
1123 sk = RB_FIND(pf_state_tree_lan_ext,
1124 &pf_statetbl_lan_ext,
1125 (struct pf_state_key *)key);
1126 if (sk && sk->af_lan == sk->af_gwy)
1127 sk = NULL;
1128 }
1129 break;
1130 default:
1131 panic("pf_find_state_all");
1132 }
1133
1134 if (sk != NULL) {
1135 ret = TAILQ_FIRST(&sk->states);
1136 if (more == NULL)
1137 return (ret);
1138
1139 TAILQ_FOREACH(s, &sk->states, next)
1140 (*more)++;
1141 }
1142
1143 return (ret);
1144}
1145
1146static void
1147pf_init_threshold(struct pf_threshold *threshold,
1148 u_int32_t limit, u_int32_t seconds)
1149{
1150 threshold->limit = limit * PF_THRESHOLD_MULT;
1151 threshold->seconds = seconds;
1152 threshold->count = 0;
1153 threshold->last = pf_time_second();
1154}
1155
1156static void
1157pf_add_threshold(struct pf_threshold *threshold)
1158{
1159 u_int32_t t = pf_time_second(), diff = t - threshold->last;
1160
1161 if (diff >= threshold->seconds)
1162 threshold->count = 0;
1163 else
1164 threshold->count -= threshold->count * diff /
1165 threshold->seconds;
1166 threshold->count += PF_THRESHOLD_MULT;
1167 threshold->last = t;
1168}
1169
1170static int
1171pf_check_threshold(struct pf_threshold *threshold)
1172{
1173 return (threshold->count > threshold->limit);
1174}
1175
1176static int
1177pf_src_connlimit(struct pf_state **state)
1178{
1179 int bad = 0;
1180 (*state)->src_node->conn++;
1181 VERIFY((*state)->src_node->conn != 0);
1182 (*state)->src.tcp_est = 1;
1183 pf_add_threshold(&(*state)->src_node->conn_rate);
1184
1185 if ((*state)->rule.ptr->max_src_conn &&
1186 (*state)->rule.ptr->max_src_conn <
1187 (*state)->src_node->conn) {
1188 pf_status.lcounters[LCNT_SRCCONN]++;
1189 bad++;
1190 }
1191
1192 if ((*state)->rule.ptr->max_src_conn_rate.limit &&
1193 pf_check_threshold(&(*state)->src_node->conn_rate)) {
1194 pf_status.lcounters[LCNT_SRCCONNRATE]++;
1195 bad++;
1196 }
1197
1198 if (!bad)
1199 return (0);
1200
1201 if ((*state)->rule.ptr->overload_tbl) {
1202 struct pfr_addr p;
1203 u_int32_t killed = 0;
1204
1205 pf_status.lcounters[LCNT_OVERLOAD_TABLE]++;
1206 if (pf_status.debug >= PF_DEBUG_MISC) {
1207 printf("pf_src_connlimit: blocking address ");
1208 pf_print_host(&(*state)->src_node->addr, 0,
1209 (*state)->state_key->af_lan);
1210 }
1211
1212 bzero(&p, sizeof (p));
1213 p.pfra_af = (*state)->state_key->af_lan;
1214 switch ((*state)->state_key->af_lan) {
1215#if INET
1216 case AF_INET:
1217 p.pfra_net = 32;
1218 p.pfra_ip4addr = (*state)->src_node->addr.v4addr;
1219 break;
1220#endif /* INET */
1221#if INET6
1222 case AF_INET6:
1223 p.pfra_net = 128;
1224 p.pfra_ip6addr = (*state)->src_node->addr.v6addr;
1225 break;
1226#endif /* INET6 */
1227 }
1228
1229 pfr_insert_kentry((*state)->rule.ptr->overload_tbl,
1230 &p, pf_calendar_time_second());
1231
1232 /* kill existing states if that's required. */
1233 if ((*state)->rule.ptr->flush) {
1234 struct pf_state_key *sk;
1235 struct pf_state *st;
1236
1237 pf_status.lcounters[LCNT_OVERLOAD_FLUSH]++;
1238 RB_FOREACH(st, pf_state_tree_id, &tree_id) {
1239 sk = st->state_key;
1240 /*
1241 * Kill states from this source. (Only those
1242 * from the same rule if PF_FLUSH_GLOBAL is not
1243 * set)
1244 */
1245 if (sk->af_lan ==
1246 (*state)->state_key->af_lan &&
1247 (((*state)->state_key->direction ==
1248 PF_OUT &&
1249 PF_AEQ(&(*state)->src_node->addr,
1250 &sk->lan.addr, sk->af_lan)) ||
1251 ((*state)->state_key->direction == PF_IN &&
1252 PF_AEQ(&(*state)->src_node->addr,
1253 &sk->ext_lan.addr, sk->af_lan))) &&
1254 ((*state)->rule.ptr->flush &
1255 PF_FLUSH_GLOBAL ||
1256 (*state)->rule.ptr == st->rule.ptr)) {
1257 st->timeout = PFTM_PURGE;
1258 st->src.state = st->dst.state =
1259 TCPS_CLOSED;
1260 killed++;
1261 }
1262 }
1263 if (pf_status.debug >= PF_DEBUG_MISC)
1264 printf(", %u states killed", killed);
1265 }
1266 if (pf_status.debug >= PF_DEBUG_MISC)
1267 printf("\n");
1268 }
1269
1270 /* kill this state */
1271 (*state)->timeout = PFTM_PURGE;
1272 (*state)->src.state = (*state)->dst.state = TCPS_CLOSED;
1273 return (1);
1274}
1275
1276int
1277pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule,
1278 struct pf_addr *src, sa_family_t af)
1279{
1280 struct pf_src_node k;
1281
1282 if (*sn == NULL) {
1283 k.af = af;
1284 PF_ACPY(&k.addr, src, af);
1285 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
1286 rule->rpool.opts & PF_POOL_STICKYADDR)
1287 k.rule.ptr = rule;
1288 else
1289 k.rule.ptr = NULL;
1290 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
1291 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
1292 }
1293 if (*sn == NULL) {
1294 if (!rule->max_src_nodes ||
1295 rule->src_nodes < rule->max_src_nodes)
1296 (*sn) = pool_get(&pf_src_tree_pl, PR_WAITOK);
1297 else
1298 pf_status.lcounters[LCNT_SRCNODES]++;
1299 if ((*sn) == NULL)
1300 return (-1);
1301 bzero(*sn, sizeof (struct pf_src_node));
1302
1303 pf_init_threshold(&(*sn)->conn_rate,
1304 rule->max_src_conn_rate.limit,
1305 rule->max_src_conn_rate.seconds);
1306
1307 (*sn)->af = af;
1308 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
1309 rule->rpool.opts & PF_POOL_STICKYADDR)
1310 (*sn)->rule.ptr = rule;
1311 else
1312 (*sn)->rule.ptr = NULL;
1313 PF_ACPY(&(*sn)->addr, src, af);
1314 if (RB_INSERT(pf_src_tree,
1315 &tree_src_tracking, *sn) != NULL) {
1316 if (pf_status.debug >= PF_DEBUG_MISC) {
1317 printf("pf: src_tree insert failed: ");
1318 pf_print_host(&(*sn)->addr, 0, af);
1319 printf("\n");
1320 }
1321 pool_put(&pf_src_tree_pl, *sn);
1322 return (-1);
1323 }
1324 (*sn)->creation = pf_time_second();
1325 (*sn)->ruletype = rule->action;
1326 if ((*sn)->rule.ptr != NULL)
1327 (*sn)->rule.ptr->src_nodes++;
1328 pf_status.scounters[SCNT_SRC_NODE_INSERT]++;
1329 pf_status.src_nodes++;
1330 } else {
1331 if (rule->max_src_states &&
1332 (*sn)->states >= rule->max_src_states) {
1333 pf_status.lcounters[LCNT_SRCSTATES]++;
1334 return (-1);
1335 }
1336 }
1337 return (0);
1338}
1339
1340static void
1341pf_stateins_err(const char *tree, struct pf_state *s, struct pfi_kif *kif)
1342{
1343 struct pf_state_key *sk = s->state_key;
1344
1345 if (pf_status.debug >= PF_DEBUG_MISC) {
1346 printf("pf: state insert failed: %s %s ", tree, kif->pfik_name);
1347 switch (sk->proto) {
1348 case IPPROTO_TCP:
1349 printf("TCP");
1350 break;
1351 case IPPROTO_UDP:
1352 printf("UDP");
1353 break;
1354 case IPPROTO_ICMP:
1355 printf("ICMP4");
1356 break;
1357 case IPPROTO_ICMPV6:
1358 printf("ICMP6");
1359 break;
1360 default:
1361 printf("PROTO=%u", sk->proto);
1362 break;
1363 }
1364 printf(" lan: ");
1365 pf_print_sk_host(&sk->lan, sk->af_lan, sk->proto,
1366 sk->proto_variant);
1367 printf(" gwy: ");
1368 pf_print_sk_host(&sk->gwy, sk->af_gwy, sk->proto,
1369 sk->proto_variant);
1370 printf(" ext_lan: ");
1371 pf_print_sk_host(&sk->ext_lan, sk->af_lan, sk->proto,
1372 sk->proto_variant);
1373 printf(" ext_gwy: ");
1374 pf_print_sk_host(&sk->ext_gwy, sk->af_gwy, sk->proto,
1375 sk->proto_variant);
1376 if (s->sync_flags & PFSTATE_FROMSYNC)
1377 printf(" (from sync)");
1378 printf("\n");
1379 }
1380}
1381
1382int
1383pf_insert_state(struct pfi_kif *kif, struct pf_state *s)
1384{
1385 struct pf_state_key *cur;
1386 struct pf_state *sp;
1387
1388 VERIFY(s->state_key != NULL);
1389 s->kif = kif;
1390
1391 if ((cur = RB_INSERT(pf_state_tree_lan_ext, &pf_statetbl_lan_ext,
1392 s->state_key)) != NULL) {
1393 /* key exists. check for same kif, if none, add to key */
1394 TAILQ_FOREACH(sp, &cur->states, next)
1395 if (sp->kif == kif) { /* collision! */
1396 pf_stateins_err("tree_lan_ext", s, kif);
1397 pf_detach_state(s,
1398 PF_DT_SKIP_LANEXT|PF_DT_SKIP_EXTGWY);
1399 return (-1);
1400 }
1401 pf_detach_state(s, PF_DT_SKIP_LANEXT|PF_DT_SKIP_EXTGWY);
1402 pf_attach_state(cur, s, kif == pfi_all ? 1 : 0);
1403 }
1404
1405 /* if cur != NULL, we already found a state key and attached to it */
1406 if (cur == NULL && (cur = RB_INSERT(pf_state_tree_ext_gwy,
1407 &pf_statetbl_ext_gwy, s->state_key)) != NULL) {
1408 /* must not happen. we must have found the sk above! */
1409 pf_stateins_err("tree_ext_gwy", s, kif);
1410 pf_detach_state(s, PF_DT_SKIP_EXTGWY);
1411 return (-1);
1412 }
1413
1414 if (s->id == 0 && s->creatorid == 0) {
1415 s->id = htobe64(pf_status.stateid++);
1416 s->creatorid = pf_status.hostid;
1417 }
1418 if (RB_INSERT(pf_state_tree_id, &tree_id, s) != NULL) {
1419 if (pf_status.debug >= PF_DEBUG_MISC) {
1420 printf("pf: state insert failed: "
1421 "id: %016llx creatorid: %08x",
1422 be64toh(s->id), ntohl(s->creatorid));
1423 if (s->sync_flags & PFSTATE_FROMSYNC)
1424 printf(" (from sync)");
1425 printf("\n");
1426 }
1427 pf_detach_state(s, 0);
1428 return (-1);
1429 }
1430 TAILQ_INSERT_TAIL(&state_list, s, entry_list);
1431 pf_status.fcounters[FCNT_STATE_INSERT]++;
1432 pf_status.states++;
1433 VERIFY(pf_status.states != 0);
1434 pfi_kif_ref(kif, PFI_KIF_REF_STATE);
1435#if NPFSYNC
1436 pfsync_insert_state(s);
1437#endif
1438 return (0);
1439}
1440
1441static int
1442pf_purge_thread_cont(int err)
1443{
1444#pragma unused(err)
1445 static u_int32_t nloops = 0;
1446 int t = 1; /* 1 second */
1447
1448 /*
1449 * Update coarse-grained networking timestamp (in sec.); the idea
1450 * is to piggy-back on the periodic timeout callout to update
1451 * the counter returnable via net_uptime().
1452 */
1453 net_update_uptime();
1454
1455 lck_rw_lock_shared(pf_perim_lock);
1456 lck_mtx_lock(pf_lock);
1457
1458 /* purge everything if not running */
1459 if (!pf_status.running) {
1460 pf_purge_expired_states(pf_status.states);
1461 pf_purge_expired_fragments();
1462 pf_purge_expired_src_nodes();
1463
1464 /* terminate thread (we don't currently do this) */
1465 if (pf_purge_thread == NULL) {
1466 lck_mtx_unlock(pf_lock);
1467 lck_rw_done(pf_perim_lock);
1468
1469 thread_deallocate(current_thread());
1470 thread_terminate(current_thread());
1471 /* NOTREACHED */
1472 return (0);
1473 } else {
1474 /* if there's nothing left, sleep w/o timeout */
1475 if (pf_status.states == 0 &&
1476 pf_normalize_isempty() &&
1477 RB_EMPTY(&tree_src_tracking)) {
1478 nloops = 0;
1479 t = 0;
1480 }
1481 goto done;
1482 }
1483 }
1484
1485 /* process a fraction of the state table every second */
1486 pf_purge_expired_states(1 + (pf_status.states
1487 / pf_default_rule.timeout[PFTM_INTERVAL]));
1488
1489 /* purge other expired types every PFTM_INTERVAL seconds */
1490 if (++nloops >= pf_default_rule.timeout[PFTM_INTERVAL]) {
1491 pf_purge_expired_fragments();
1492 pf_purge_expired_src_nodes();
1493 nloops = 0;
1494 }
1495done:
1496 lck_mtx_unlock(pf_lock);
1497 lck_rw_done(pf_perim_lock);
1498
1499 (void) tsleep0(pf_purge_thread_fn, PWAIT, "pf_purge_cont",
1500 t * hz, pf_purge_thread_cont);
1501 /* NOTREACHED */
1502 VERIFY(0);
1503
1504 return (0);
1505}
1506
1507void
1508pf_purge_thread_fn(void *v, wait_result_t w)
1509{
1510#pragma unused(v, w)
1511 (void) tsleep0(pf_purge_thread_fn, PWAIT, "pf_purge", 0,
1512 pf_purge_thread_cont);
1513 /*
1514 * tsleep0() shouldn't have returned as PCATCH was not set;
1515 * therefore assert in this case.
1516 */
1517 VERIFY(0);
1518}
1519
1520u_int64_t
1521pf_state_expires(const struct pf_state *state)
1522{
1523 u_int32_t t;
1524 u_int32_t start;
1525 u_int32_t end;
1526 u_int32_t states;
1527
1528 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
1529
1530 /* handle all PFTM_* > PFTM_MAX here */
1531 if (state->timeout == PFTM_PURGE)
1532 return (pf_time_second());
1533
1534 VERIFY(state->timeout != PFTM_UNLINKED);
1535 VERIFY(state->timeout < PFTM_MAX);
1536 t = state->rule.ptr->timeout[state->timeout];
1537 if (!t)
1538 t = pf_default_rule.timeout[state->timeout];
1539 start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START];
1540 if (start) {
1541 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END];
1542 states = state->rule.ptr->states;
1543 } else {
1544 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START];
1545 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END];
1546 states = pf_status.states;
1547 }
1548 if (end && states > start && start < end) {
1549 if (states < end)
1550 return (state->expire + t * (end - states) /
1551 (end - start));
1552 else
1553 return (pf_time_second());
1554 }
1555 return (state->expire + t);
1556}
1557
1558void
1559pf_purge_expired_src_nodes(void)
1560{
1561 struct pf_src_node *cur, *next;
1562
1563 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
1564
1565 for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) {
1566 next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur);
1567
1568 if (cur->states <= 0 && cur->expire <= pf_time_second()) {
1569 if (cur->rule.ptr != NULL) {
1570 cur->rule.ptr->src_nodes--;
1571 if (cur->rule.ptr->states <= 0 &&
1572 cur->rule.ptr->max_src_nodes <= 0)
1573 pf_rm_rule(NULL, cur->rule.ptr);
1574 }
1575 RB_REMOVE(pf_src_tree, &tree_src_tracking, cur);
1576 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
1577 pf_status.src_nodes--;
1578 pool_put(&pf_src_tree_pl, cur);
1579 }
1580 }
1581}
1582
1583void
1584pf_src_tree_remove_state(struct pf_state *s)
1585{
1586 u_int32_t t;
1587
1588 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
1589
1590 if (s->src_node != NULL) {
1591 if (s->src.tcp_est) {
1592 VERIFY(s->src_node->conn > 0);
1593 --s->src_node->conn;
1594 }
1595 VERIFY(s->src_node->states > 0);
1596 if (--s->src_node->states <= 0) {
1597 t = s->rule.ptr->timeout[PFTM_SRC_NODE];
1598 if (!t)
1599 t = pf_default_rule.timeout[PFTM_SRC_NODE];
1600 s->src_node->expire = pf_time_second() + t;
1601 }
1602 }
1603 if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) {
1604 VERIFY(s->nat_src_node->states > 0);
1605 if (--s->nat_src_node->states <= 0) {
1606 t = s->rule.ptr->timeout[PFTM_SRC_NODE];
1607 if (!t)
1608 t = pf_default_rule.timeout[PFTM_SRC_NODE];
1609 s->nat_src_node->expire = pf_time_second() + t;
1610 }
1611 }
1612 s->src_node = s->nat_src_node = NULL;
1613}
1614
1615void
1616pf_unlink_state(struct pf_state *cur)
1617{
1618 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
1619
1620 if (cur->src.state == PF_TCPS_PROXY_DST) {
1621 pf_send_tcp(cur->rule.ptr, cur->state_key->af_lan,
1622 &cur->state_key->ext_lan.addr, &cur->state_key->lan.addr,
1623 cur->state_key->ext_lan.xport.port,
1624 cur->state_key->lan.xport.port,
1625 cur->src.seqhi, cur->src.seqlo + 1,
1626 TH_RST|TH_ACK, 0, 0, 0, 1, cur->tag, NULL, NULL);
1627 }
1628
1629 hook_runloop(&cur->unlink_hooks, HOOK_REMOVE|HOOK_FREE);
1630 RB_REMOVE(pf_state_tree_id, &tree_id, cur);
1631#if NPFSYNC
1632 if (cur->creatorid == pf_status.hostid)
1633 pfsync_delete_state(cur);
1634#endif
1635 cur->timeout = PFTM_UNLINKED;
1636 pf_src_tree_remove_state(cur);
1637 pf_detach_state(cur, 0);
1638}
1639
1640/* callers should be at splpf and hold the
1641 * write_lock on pf_consistency_lock */
1642void
1643pf_free_state(struct pf_state *cur)
1644{
1645 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
1646#if NPFSYNC
1647 if (pfsyncif != NULL &&
1648 (pfsyncif->sc_bulk_send_next == cur ||
1649 pfsyncif->sc_bulk_terminator == cur))
1650 return;
1651#endif
1652 VERIFY(cur->timeout == PFTM_UNLINKED);
1653 VERIFY(cur->rule.ptr->states > 0);
1654 if (--cur->rule.ptr->states <= 0 &&
1655 cur->rule.ptr->src_nodes <= 0)
1656 pf_rm_rule(NULL, cur->rule.ptr);
1657 if (cur->nat_rule.ptr != NULL) {
1658 VERIFY(cur->nat_rule.ptr->states > 0);
1659 if (--cur->nat_rule.ptr->states <= 0 &&
1660 cur->nat_rule.ptr->src_nodes <= 0)
1661 pf_rm_rule(NULL, cur->nat_rule.ptr);
1662 }
1663 if (cur->anchor.ptr != NULL) {
1664 VERIFY(cur->anchor.ptr->states > 0);
1665 if (--cur->anchor.ptr->states <= 0)
1666 pf_rm_rule(NULL, cur->anchor.ptr);
1667 }
1668 pf_normalize_tcp_cleanup(cur);
1669 pfi_kif_unref(cur->kif, PFI_KIF_REF_STATE);
1670 TAILQ_REMOVE(&state_list, cur, entry_list);
1671 if (cur->tag)
1672 pf_tag_unref(cur->tag);
1673 pool_put(&pf_state_pl, cur);
1674 pf_status.fcounters[FCNT_STATE_REMOVALS]++;
1675 VERIFY(pf_status.states > 0);
1676 pf_status.states--;
1677}
1678
1679void
1680pf_purge_expired_states(u_int32_t maxcheck)
1681{
1682 static struct pf_state *cur = NULL;
1683 struct pf_state *next;
1684
1685 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
1686
1687 while (maxcheck--) {
1688 /* wrap to start of list when we hit the end */
1689 if (cur == NULL) {
1690 cur = TAILQ_FIRST(&state_list);
1691 if (cur == NULL)
1692 break; /* list empty */
1693 }
1694
1695 /* get next state, as cur may get deleted */
1696 next = TAILQ_NEXT(cur, entry_list);
1697
1698 if (cur->timeout == PFTM_UNLINKED) {
1699 pf_free_state(cur);
1700 } else if (pf_state_expires(cur) <= pf_time_second()) {
1701 /* unlink and free expired state */
1702 pf_unlink_state(cur);
1703 pf_free_state(cur);
1704 }
1705 cur = next;
1706 }
1707}
1708
1709int
1710pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw)
1711{
1712 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
1713
1714 if (aw->type != PF_ADDR_TABLE)
1715 return (0);
1716 if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL)
1717 return (1);
1718 return (0);
1719}
1720
1721void
1722pf_tbladdr_remove(struct pf_addr_wrap *aw)
1723{
1724 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
1725
1726 if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL)
1727 return;
1728 pfr_detach_table(aw->p.tbl);
1729 aw->p.tbl = NULL;
1730}
1731
1732void
1733pf_tbladdr_copyout(struct pf_addr_wrap *aw)
1734{
1735 struct pfr_ktable *kt = aw->p.tbl;
1736
1737 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
1738
1739 if (aw->type != PF_ADDR_TABLE || kt == NULL)
1740 return;
1741 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
1742 kt = kt->pfrkt_root;
1743 aw->p.tbl = NULL;
1744 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
1745 kt->pfrkt_cnt : -1;
1746}
1747
1748static void
1749pf_print_addr(struct pf_addr *addr, sa_family_t af)
1750{
1751 switch (af) {
1752#if INET
1753 case AF_INET: {
1754 u_int32_t a = ntohl(addr->addr32[0]);
1755 printf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255,
1756 (a>>8)&255, a&255);
1757 break;
1758 }
1759#endif /* INET */
1760#if INET6
1761 case AF_INET6: {
1762 u_int16_t b;
1763 u_int8_t i, curstart = 255, curend = 0,
1764 maxstart = 0, maxend = 0;
1765 for (i = 0; i < 8; i++) {
1766 if (!addr->addr16[i]) {
1767 if (curstart == 255)
1768 curstart = i;
1769 else
1770 curend = i;
1771 } else {
1772 if (curstart) {
1773 if ((curend - curstart) >
1774 (maxend - maxstart)) {
1775 maxstart = curstart;
1776 maxend = curend;
1777 curstart = 255;
1778 }
1779 }
1780 }
1781 }
1782 for (i = 0; i < 8; i++) {
1783 if (i >= maxstart && i <= maxend) {
1784 if (maxend != 7) {
1785 if (i == maxstart)
1786 printf(":");
1787 } else {
1788 if (i == maxend)
1789 printf(":");
1790 }
1791 } else {
1792 b = ntohs(addr->addr16[i]);
1793 printf("%x", b);
1794 if (i < 7)
1795 printf(":");
1796 }
1797 }
1798 break;
1799 }
1800#endif /* INET6 */
1801 }
1802}
1803
1804static void
1805pf_print_sk_host(struct pf_state_host *sh, sa_family_t af, int proto,
1806 u_int8_t proto_variant)
1807{
1808 pf_print_addr(&sh->addr, af);
1809
1810 switch (proto) {
1811 case IPPROTO_ESP:
1812 if (sh->xport.spi)
1813 printf("[%08x]", ntohl(sh->xport.spi));
1814 break;
1815
1816 case IPPROTO_GRE:
1817 if (proto_variant == PF_GRE_PPTP_VARIANT)
1818 printf("[%u]", ntohs(sh->xport.call_id));
1819 break;
1820
1821 case IPPROTO_TCP:
1822 case IPPROTO_UDP:
1823 printf("[%u]", ntohs(sh->xport.port));
1824 break;
1825
1826 default:
1827 break;
1828 }
1829}
1830
1831static void
1832pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af)
1833{
1834 pf_print_addr(addr, af);
1835 if (p)
1836 printf("[%u]", ntohs(p));
1837}
1838
1839void
1840pf_print_state(struct pf_state *s)
1841{
1842 struct pf_state_key *sk = s->state_key;
1843 switch (sk->proto) {
1844 case IPPROTO_ESP:
1845 printf("ESP ");
1846 break;
1847 case IPPROTO_GRE:
1848 printf("GRE%u ", sk->proto_variant);
1849 break;
1850 case IPPROTO_TCP:
1851 printf("TCP ");
1852 break;
1853 case IPPROTO_UDP:
1854 printf("UDP ");
1855 break;
1856 case IPPROTO_ICMP:
1857 printf("ICMP ");
1858 break;
1859 case IPPROTO_ICMPV6:
1860 printf("ICMPV6 ");
1861 break;
1862 default:
1863 printf("%u ", sk->proto);
1864 break;
1865 }
1866 pf_print_sk_host(&sk->lan, sk->af_lan, sk->proto, sk->proto_variant);
1867 printf(" ");
1868 pf_print_sk_host(&sk->gwy, sk->af_gwy, sk->proto, sk->proto_variant);
1869 printf(" ");
1870 pf_print_sk_host(&sk->ext_lan, sk->af_lan, sk->proto,
1871 sk->proto_variant);
1872 printf(" ");
1873 pf_print_sk_host(&sk->ext_gwy, sk->af_gwy, sk->proto,
1874 sk->proto_variant);
1875 printf(" [lo=%u high=%u win=%u modulator=%u", s->src.seqlo,
1876 s->src.seqhi, s->src.max_win, s->src.seqdiff);
1877 if (s->src.wscale && s->dst.wscale)
1878 printf(" wscale=%u", s->src.wscale & PF_WSCALE_MASK);
1879 printf("]");
1880 printf(" [lo=%u high=%u win=%u modulator=%u", s->dst.seqlo,
1881 s->dst.seqhi, s->dst.max_win, s->dst.seqdiff);
1882 if (s->src.wscale && s->dst.wscale)
1883 printf(" wscale=%u", s->dst.wscale & PF_WSCALE_MASK);
1884 printf("]");
1885 printf(" %u:%u", s->src.state, s->dst.state);
1886}
1887
1888void
1889pf_print_flags(u_int8_t f)
1890{
1891 if (f)
1892 printf(" ");
1893 if (f & TH_FIN)
1894 printf("F");
1895 if (f & TH_SYN)
1896 printf("S");
1897 if (f & TH_RST)
1898 printf("R");
1899 if (f & TH_PUSH)
1900 printf("P");
1901 if (f & TH_ACK)
1902 printf("A");
1903 if (f & TH_URG)
1904 printf("U");
1905 if (f & TH_ECE)
1906 printf("E");
1907 if (f & TH_CWR)
1908 printf("W");
1909}
1910
1911#define PF_SET_SKIP_STEPS(i) \
1912 do { \
1913 while (head[i] != cur) { \
1914 head[i]->skip[i].ptr = cur; \
1915 head[i] = TAILQ_NEXT(head[i], entries); \
1916 } \
1917 } while (0)
1918
1919void
1920pf_calc_skip_steps(struct pf_rulequeue *rules)
1921{
1922 struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT];
1923 int i;
1924
1925 cur = TAILQ_FIRST(rules);
1926 prev = cur;
1927 for (i = 0; i < PF_SKIP_COUNT; ++i)
1928 head[i] = cur;
1929 while (cur != NULL) {
1930
1931 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
1932 PF_SET_SKIP_STEPS(PF_SKIP_IFP);
1933 if (cur->direction != prev->direction)
1934 PF_SET_SKIP_STEPS(PF_SKIP_DIR);
1935 if (cur->af != prev->af)
1936 PF_SET_SKIP_STEPS(PF_SKIP_AF);
1937 if (cur->proto != prev->proto)
1938 PF_SET_SKIP_STEPS(PF_SKIP_PROTO);
1939 if (cur->src.neg != prev->src.neg ||
1940 pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr))
1941 PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR);
1942 {
1943 union pf_rule_xport *cx = &cur->src.xport;
1944 union pf_rule_xport *px = &prev->src.xport;
1945
1946 switch (cur->proto) {
1947 case IPPROTO_GRE:
1948 case IPPROTO_ESP:
1949 PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
1950 break;
1951 default:
1952 if (prev->proto == IPPROTO_GRE ||
1953 prev->proto == IPPROTO_ESP ||
1954 cx->range.op != px->range.op ||
1955 cx->range.port[0] != px->range.port[0] ||
1956 cx->range.port[1] != px->range.port[1])
1957 PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
1958 break;
1959 }
1960 }
1961 if (cur->dst.neg != prev->dst.neg ||
1962 pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr))
1963 PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR);
1964 {
1965 union pf_rule_xport *cx = &cur->dst.xport;
1966 union pf_rule_xport *px = &prev->dst.xport;
1967
1968 switch (cur->proto) {
1969 case IPPROTO_GRE:
1970 if (cur->proto != prev->proto ||
1971 cx->call_id != px->call_id)
1972 PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
1973 break;
1974 case IPPROTO_ESP:
1975 if (cur->proto != prev->proto ||
1976 cx->spi != px->spi)
1977 PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
1978 break;
1979 default:
1980 if (prev->proto == IPPROTO_GRE ||
1981 prev->proto == IPPROTO_ESP ||
1982 cx->range.op != px->range.op ||
1983 cx->range.port[0] != px->range.port[0] ||
1984 cx->range.port[1] != px->range.port[1])
1985 PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
1986 break;
1987 }
1988 }
1989
1990 prev = cur;
1991 cur = TAILQ_NEXT(cur, entries);
1992 }
1993 for (i = 0; i < PF_SKIP_COUNT; ++i)
1994 PF_SET_SKIP_STEPS(i);
1995}
1996
1997u_int32_t
1998pf_calc_state_key_flowhash(struct pf_state_key *sk)
1999{
2000 struct pf_flowhash_key fh __attribute__((aligned(8)));
2001 uint32_t flowhash = 0;
2002
2003 bzero(&fh, sizeof (fh));
2004 if (PF_ALEQ(&sk->lan.addr, &sk->ext_lan.addr, sk->af_lan)) {
2005 bcopy(&sk->lan.addr, &fh.ap1.addr, sizeof (fh.ap1.addr));
2006 bcopy(&sk->ext_lan.addr, &fh.ap2.addr, sizeof (fh.ap2.addr));
2007 } else {
2008 bcopy(&sk->ext_lan.addr, &fh.ap1.addr, sizeof (fh.ap1.addr));
2009 bcopy(&sk->lan.addr, &fh.ap2.addr, sizeof (fh.ap2.addr));
2010 }
2011 if (sk->lan.xport.spi <= sk->ext_lan.xport.spi) {
2012 fh.ap1.xport.spi = sk->lan.xport.spi;
2013 fh.ap2.xport.spi = sk->ext_lan.xport.spi;
2014 } else {
2015 fh.ap1.xport.spi = sk->ext_lan.xport.spi;
2016 fh.ap2.xport.spi = sk->lan.xport.spi;
2017 }
2018 fh.af = sk->af_lan;
2019 fh.proto = sk->proto;
2020
2021try_again:
2022 flowhash = net_flowhash(&fh, sizeof (fh), pf_hash_seed);
2023 if (flowhash == 0) {
2024 /* try to get a non-zero flowhash */
2025 pf_hash_seed = RandomULong();
2026 goto try_again;
2027 }
2028
2029 return (flowhash);
2030}
2031
2032static int
2033pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2)
2034{
2035 if (aw1->type != aw2->type)
2036 return (1);
2037 switch (aw1->type) {
2038 case PF_ADDR_ADDRMASK:
2039 case PF_ADDR_RANGE:
2040 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0))
2041 return (1);
2042 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0))
2043 return (1);
2044 return (0);
2045 case PF_ADDR_DYNIFTL:
2046 return (aw1->p.dyn == NULL || aw2->p.dyn == NULL ||
2047 aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt);
2048 case PF_ADDR_NOROUTE:
2049 case PF_ADDR_URPFFAILED:
2050 return (0);
2051 case PF_ADDR_TABLE:
2052 return (aw1->p.tbl != aw2->p.tbl);
2053 case PF_ADDR_RTLABEL:
2054 return (aw1->v.rtlabel != aw2->v.rtlabel);
2055 default:
2056 printf("invalid address type: %d\n", aw1->type);
2057 return (1);
2058 }
2059}
2060
2061u_int16_t
2062pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp)
2063{
2064 return (nat464_cksum_fixup(cksum, old, new, udp));
2065}
2066
2067/*
2068 * change ip address & port
2069 * dir : packet direction
2070 * a : address to be changed
2071 * p : port to be changed
2072 * ic : ip header checksum
2073 * pc : protocol checksum
2074 * an : new ip address
2075 * pn : new port
2076 * u : should be 1 if UDP packet else 0
2077 * af : address family of the packet
2078 * afn : address family of the new address
2079 * ua : should be 1 if ip address needs to be updated in the packet else
2080 * only the checksum is recalculated & updated.
2081 */
2082static void
2083pf_change_ap(int dir, pbuf_t *pbuf, struct pf_addr *a, u_int16_t *p,
2084 u_int16_t *ic, u_int16_t *pc, struct pf_addr *an, u_int16_t pn,
2085 u_int8_t u, sa_family_t af, sa_family_t afn, int ua)
2086{
2087 struct pf_addr ao;
2088 u_int16_t po = *p;
2089
2090 PF_ACPY(&ao, a, af);
2091 if (ua)
2092 PF_ACPY(a, an, afn);
2093
2094 *p = pn;
2095
2096 switch (af) {
2097#if INET
2098 case AF_INET:
2099 switch (afn) {
2100 case AF_INET:
2101 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
2102 ao.addr16[0], an->addr16[0], 0),
2103 ao.addr16[1], an->addr16[1], 0);
2104 *p = pn;
2105 /*
2106 * If the packet is originated from an ALG on the NAT gateway
2107 * (source address is loopback or local), in which case the
2108 * TCP/UDP checksum field contains the pseudo header checksum
2109 * that's not yet complemented.
2110 * In that case we do not need to fixup the checksum for port
2111 * translation as the pseudo header checksum doesn't include ports.
2112 *
2113 * A packet generated locally will have UDP/TCP CSUM flag
2114 * set (gets set in protocol output).
2115 *
2116 * It should be noted that the fixup doesn't do anything if the
2117 * checksum is 0.
2118 */
2119 if (dir == PF_OUT && pbuf != NULL &&
2120 (*pbuf->pb_csum_flags & (CSUM_TCP | CSUM_UDP))) {
2121 /* Pseudo-header checksum does not include ports */
2122 *pc = ~pf_cksum_fixup(pf_cksum_fixup(~*pc,
2123 ao.addr16[0], an->addr16[0], u),
2124 ao.addr16[1], an->addr16[1], u);
2125 } else {
2126 *pc =
2127 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2128 *pc, ao.addr16[0], an->addr16[0], u),
2129 ao.addr16[1], an->addr16[1], u),
2130 po, pn, u);
2131 }
2132 break;
2133#ifdef INET6
2134 case AF_INET6:
2135 *p = pn;
2136 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2137 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2138
2139 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
2140 ao.addr16[0], an->addr16[0], u),
2141 ao.addr16[1], an->addr16[1], u),
2142 0, an->addr16[2], u),
2143 0, an->addr16[3], u),
2144 0, an->addr16[4], u),
2145 0, an->addr16[5], u),
2146 0, an->addr16[6], u),
2147 0, an->addr16[7], u),
2148 po, pn, u);
2149 break;
2150#endif /* INET6 */
2151 }
2152 break;
2153#endif /* INET */
2154#if INET6
2155 case AF_INET6:
2156 switch (afn) {
2157 case AF_INET6:
2158 /*
2159 * If the packet is originated from an ALG on the NAT gateway
2160 * (source address is loopback or local), in which case the
2161 * TCP/UDP checksum field contains the pseudo header checksum
2162 * that's not yet complemented.
2163 * A packet generated locally
2164 * will have UDP/TCP CSUM flag set (gets set in protocol
2165 * output).
2166 */
2167 if (dir == PF_OUT && pbuf != NULL &&
2168 (*pbuf->pb_csum_flags & (CSUM_TCPIPV6 |
2169 CSUM_UDPIPV6))) {
2170 /* Pseudo-header checksum does not include ports */
2171 *pc =
2172 ~pf_cksum_fixup(pf_cksum_fixup(
2173 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2174 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2175 ~*pc,
2176 ao.addr16[0], an->addr16[0], u),
2177 ao.addr16[1], an->addr16[1], u),
2178 ao.addr16[2], an->addr16[2], u),
2179 ao.addr16[3], an->addr16[3], u),
2180 ao.addr16[4], an->addr16[4], u),
2181 ao.addr16[5], an->addr16[5], u),
2182 ao.addr16[6], an->addr16[6], u),
2183 ao.addr16[7], an->addr16[7], u);
2184 } else {
2185 *pc =
2186 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2187 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2188 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2189 *pc,
2190 ao.addr16[0], an->addr16[0], u),
2191 ao.addr16[1], an->addr16[1], u),
2192 ao.addr16[2], an->addr16[2], u),
2193 ao.addr16[3], an->addr16[3], u),
2194 ao.addr16[4], an->addr16[4], u),
2195 ao.addr16[5], an->addr16[5], u),
2196 ao.addr16[6], an->addr16[6], u),
2197 ao.addr16[7], an->addr16[7], u),
2198 po, pn, u);
2199 }
2200 break;
2201#ifdef INET
2202 case AF_INET:
2203 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2204 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2205 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
2206 ao.addr16[0], an->addr16[0], u),
2207 ao.addr16[1], an->addr16[1], u),
2208 ao.addr16[2], 0, u),
2209 ao.addr16[3], 0, u),
2210 ao.addr16[4], 0, u),
2211 ao.addr16[5], 0, u),
2212 ao.addr16[6], 0, u),
2213 ao.addr16[7], 0, u),
2214 po, pn, u);
2215 break;
2216#endif /* INET */
2217 }
2218 break;
2219#endif /* INET6 */
2220 }
2221}
2222
2223
2224/* Changes a u_int32_t. Uses a void * so there are no align restrictions */
2225void
2226pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u)
2227{
2228 u_int32_t ao;
2229
2230 memcpy(&ao, a, sizeof (ao));
2231 memcpy(a, &an, sizeof (u_int32_t));
2232 *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u),
2233 ao % 65536, an % 65536, u);
2234}
2235
2236#if INET6
2237static void
2238pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u)
2239{
2240 struct pf_addr ao;
2241
2242 PF_ACPY(&ao, a, AF_INET6);
2243 PF_ACPY(a, an, AF_INET6);
2244
2245 *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2246 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2247 pf_cksum_fixup(pf_cksum_fixup(*c,
2248 ao.addr16[0], an->addr16[0], u),
2249 ao.addr16[1], an->addr16[1], u),
2250 ao.addr16[2], an->addr16[2], u),
2251 ao.addr16[3], an->addr16[3], u),
2252 ao.addr16[4], an->addr16[4], u),
2253 ao.addr16[5], an->addr16[5], u),
2254 ao.addr16[6], an->addr16[6], u),
2255 ao.addr16[7], an->addr16[7], u);
2256}
2257
2258void
2259pf_change_addr(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u,
2260 sa_family_t af, sa_family_t afn)
2261{
2262 struct pf_addr ao;
2263
2264 PF_ACPY(&ao, a, af);
2265 PF_ACPY(a, an, afn);
2266
2267 switch (af) {
2268 case AF_INET:
2269 switch (afn) {
2270 case AF_INET:
2271 pf_change_a(a, c, an->v4addr.s_addr, u);
2272 break;
2273 case AF_INET6:
2274 *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2275 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2276 pf_cksum_fixup(pf_cksum_fixup(*c,
2277 ao.addr16[0], an->addr16[0], u),
2278 ao.addr16[1], an->addr16[1], u),
2279 0, an->addr16[2], u),
2280 0, an->addr16[3], u),
2281 0, an->addr16[4], u),
2282 0, an->addr16[5], u),
2283 0, an->addr16[6], u),
2284 0, an->addr16[7], u);
2285 break;
2286 }
2287 break;
2288 case AF_INET6:
2289 switch (afn) {
2290 case AF_INET:
2291 *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2292 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2293 pf_cksum_fixup(pf_cksum_fixup(*c,
2294 ao.addr16[0], an->addr16[0], u),
2295 ao.addr16[1], an->addr16[1], u),
2296 ao.addr16[2], 0, u),
2297 ao.addr16[3], 0, u),
2298 ao.addr16[4], 0, u),
2299 ao.addr16[5], 0, u),
2300 ao.addr16[6], 0, u),
2301 ao.addr16[7], 0, u);
2302 break;
2303 case AF_INET6:
2304 pf_change_a6(a, c, an, u);
2305 break;
2306 }
2307 break;
2308 }
2309}
2310
2311#endif /* INET6 */
2312
2313static void
2314pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa,
2315 struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c,
2316 u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af)
2317{
2318 struct pf_addr oia, ooa;
2319
2320 PF_ACPY(&oia, ia, af);
2321 PF_ACPY(&ooa, oa, af);
2322
2323 /* Change inner protocol port, fix inner protocol checksum. */
2324 if (ip != NULL) {
2325 u_int16_t oip = *ip;
2326 u_int32_t opc = 0;
2327
2328 if (pc != NULL)
2329 opc = *pc;
2330 *ip = np;
2331 if (pc != NULL)
2332 *pc = pf_cksum_fixup(*pc, oip, *ip, u);
2333 *ic = pf_cksum_fixup(*ic, oip, *ip, 0);
2334 if (pc != NULL)
2335 *ic = pf_cksum_fixup(*ic, opc, *pc, 0);
2336 }
2337 /* Change inner ip address, fix inner ip and icmp checksums. */
2338 PF_ACPY(ia, na, af);
2339 switch (af) {
2340#if INET
2341 case AF_INET: {
2342 u_int32_t oh2c = *h2c;
2343
2344 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c,
2345 oia.addr16[0], ia->addr16[0], 0),
2346 oia.addr16[1], ia->addr16[1], 0);
2347 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
2348 oia.addr16[0], ia->addr16[0], 0),
2349 oia.addr16[1], ia->addr16[1], 0);
2350 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0);
2351 break;
2352 }
2353#endif /* INET */
2354#if INET6
2355 case AF_INET6:
2356 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2357 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2358 pf_cksum_fixup(pf_cksum_fixup(*ic,
2359 oia.addr16[0], ia->addr16[0], u),
2360 oia.addr16[1], ia->addr16[1], u),
2361 oia.addr16[2], ia->addr16[2], u),
2362 oia.addr16[3], ia->addr16[3], u),
2363 oia.addr16[4], ia->addr16[4], u),
2364 oia.addr16[5], ia->addr16[5], u),
2365 oia.addr16[6], ia->addr16[6], u),
2366 oia.addr16[7], ia->addr16[7], u);
2367 break;
2368#endif /* INET6 */
2369 }
2370 /* Change outer ip address, fix outer ip or icmpv6 checksum. */
2371 PF_ACPY(oa, na, af);
2372 switch (af) {
2373#if INET
2374 case AF_INET:
2375 *hc = pf_cksum_fixup(pf_cksum_fixup(*hc,
2376 ooa.addr16[0], oa->addr16[0], 0),
2377 ooa.addr16[1], oa->addr16[1], 0);
2378 break;
2379#endif /* INET */
2380#if INET6
2381 case AF_INET6:
2382 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2383 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2384 pf_cksum_fixup(pf_cksum_fixup(*ic,
2385 ooa.addr16[0], oa->addr16[0], u),
2386 ooa.addr16[1], oa->addr16[1], u),
2387 ooa.addr16[2], oa->addr16[2], u),
2388 ooa.addr16[3], oa->addr16[3], u),
2389 ooa.addr16[4], oa->addr16[4], u),
2390 ooa.addr16[5], oa->addr16[5], u),
2391 ooa.addr16[6], oa->addr16[6], u),
2392 ooa.addr16[7], oa->addr16[7], u);
2393 break;
2394#endif /* INET6 */
2395 }
2396}
2397
2398
2399/*
2400 * Need to modulate the sequence numbers in the TCP SACK option
2401 * (credits to Krzysztof Pfaff for report and patch)
2402 */
2403static int
2404pf_modulate_sack(pbuf_t *pbuf, int off, struct pf_pdesc *pd,
2405 struct tcphdr *th, struct pf_state_peer *dst)
2406{
2407 int hlen = (th->th_off << 2) - sizeof (*th), thoptlen = hlen;
2408 u_int8_t opts[MAX_TCPOPTLEN], *opt = opts;
2409 int copyback = 0, i, olen;
2410 struct sackblk sack;
2411
2412#define TCPOLEN_SACKLEN (TCPOLEN_SACK + 2)
2413 if (hlen < TCPOLEN_SACKLEN ||
2414 !pf_pull_hdr(pbuf, off + sizeof (*th), opts, hlen, NULL, NULL, pd->af))
2415 return (0);
2416
2417 while (hlen >= TCPOLEN_SACKLEN) {
2418 olen = opt[1];
2419 switch (*opt) {
2420 case TCPOPT_EOL: /* FALLTHROUGH */
2421 case TCPOPT_NOP:
2422 opt++;
2423 hlen--;
2424 break;
2425 case TCPOPT_SACK:
2426 if (olen > hlen)
2427 olen = hlen;
2428 if (olen >= TCPOLEN_SACKLEN) {
2429 for (i = 2; i + TCPOLEN_SACK <= olen;
2430 i += TCPOLEN_SACK) {
2431 memcpy(&sack, &opt[i], sizeof (sack));
2432 pf_change_a(&sack.start, &th->th_sum,
2433 htonl(ntohl(sack.start) -
2434 dst->seqdiff), 0);
2435 pf_change_a(&sack.end, &th->th_sum,
2436 htonl(ntohl(sack.end) -
2437 dst->seqdiff), 0);
2438 memcpy(&opt[i], &sack, sizeof (sack));
2439 }
2440 copyback = off + sizeof (*th) + thoptlen;
2441 }
2442 /* FALLTHROUGH */
2443 default:
2444 if (olen < 2)
2445 olen = 2;
2446 hlen -= olen;
2447 opt += olen;
2448 }
2449 }
2450
2451 if (copyback) {
2452 if (pf_lazy_makewritable(pd, pbuf, copyback) == NULL)
2453 return (-1);
2454 pbuf_copy_back(pbuf, off + sizeof (*th), thoptlen, opts);
2455 }
2456 return (copyback);
2457}
2458
2459/*
2460 * XXX
2461 *
2462 * The following functions (pf_send_tcp and pf_send_icmp) are somewhat
2463 * special in that they originate "spurious" packets rather than
2464 * filter/NAT existing packets. As such, they're not a great fit for
2465 * the 'pbuf' shim, which assumes the underlying packet buffers are
2466 * allocated elsewhere.
2467 *
2468 * Since these functions are rarely used, we'll carry on allocating mbufs
2469 * and passing them to the IP stack for eventual routing.
2470 */
2471static void
2472pf_send_tcp(const struct pf_rule *r, sa_family_t af,
2473 const struct pf_addr *saddr, const struct pf_addr *daddr,
2474 u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack,
2475 u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl, int tag,
2476 u_int16_t rtag, struct ether_header *eh, struct ifnet *ifp)
2477{
2478#pragma unused(eh, ifp)
2479 struct mbuf *m;
2480 int len, tlen;
2481#if INET
2482 struct ip *h = NULL;
2483#endif /* INET */
2484#if INET6
2485 struct ip6_hdr *h6 = NULL;
2486#endif /* INET6 */
2487 struct tcphdr *th = NULL;
2488 char *opt;
2489 struct pf_mtag *pf_mtag;
2490
2491 /* maximum segment size tcp option */
2492 tlen = sizeof (struct tcphdr);
2493 if (mss)
2494 tlen += 4;
2495
2496 switch (af) {
2497#if INET
2498 case AF_INET:
2499 len = sizeof (struct ip) + tlen;
2500 break;
2501#endif /* INET */
2502#if INET6
2503 case AF_INET6:
2504 len = sizeof (struct ip6_hdr) + tlen;
2505 break;
2506#endif /* INET6 */
2507 default:
2508 panic("pf_send_tcp: not AF_INET or AF_INET6!");
2509 return;
2510 }
2511
2512 /* create outgoing mbuf */
2513 m = m_gethdr(M_DONTWAIT, MT_HEADER);
2514 if (m == NULL)
2515 return;
2516
2517 if ((pf_mtag = pf_get_mtag(m)) == NULL)
2518 return;
2519
2520 if (tag)
2521 pf_mtag->pftag_flags |= PF_TAG_GENERATED;
2522 pf_mtag->pftag_tag = rtag;
2523
2524 if (r != NULL && PF_RTABLEID_IS_VALID(r->rtableid))
2525 pf_mtag->pftag_rtableid = r->rtableid;
2526
2527#if PF_ECN
2528 /* add hints for ecn */
2529 pf_mtag->pftag_hdr = mtod(m, struct ip *);
2530 /* record address family */
2531 pf_mtag->pftag_flags &= ~(PF_TAG_HDR_INET | PF_TAG_HDR_INET6);
2532 switch (af) {
2533#if INET
2534 case AF_INET:
2535 pf_mtag->pftag_flags |= PF_TAG_HDR_INET;
2536 break;
2537#endif /* INET */
2538#if INET6
2539 case AF_INET6:
2540 pf_mtag->pftag_flags |= PF_TAG_HDR_INET6;
2541 break;
2542#endif /* INET6 */
2543 }
2544#endif /* PF_ECN */
2545
2546 /* indicate this is TCP */
2547 m->m_pkthdr.pkt_proto = IPPROTO_TCP;
2548
2549 /* Make sure headers are 32-bit aligned */
2550 m->m_data += max_linkhdr;
2551 m->m_pkthdr.len = m->m_len = len;
2552 m->m_pkthdr.rcvif = NULL;
2553 bzero(m->m_data, len);
2554 switch (af) {
2555#if INET
2556 case AF_INET:
2557 h = mtod(m, struct ip *);
2558
2559 /* IP header fields included in the TCP checksum */
2560 h->ip_p = IPPROTO_TCP;
2561 h->ip_len = htons(tlen);
2562 h->ip_src.s_addr = saddr->v4addr.s_addr;
2563 h->ip_dst.s_addr = daddr->v4addr.s_addr;
2564
2565 th = (struct tcphdr *)(void *)((caddr_t)h + sizeof (struct ip));
2566 break;
2567#endif /* INET */
2568#if INET6
2569 case AF_INET6:
2570 h6 = mtod(m, struct ip6_hdr *);
2571
2572 /* IP header fields included in the TCP checksum */
2573 h6->ip6_nxt = IPPROTO_TCP;
2574 h6->ip6_plen = htons(tlen);
2575 memcpy(&h6->ip6_src, &saddr->v6addr, sizeof (struct in6_addr));
2576 memcpy(&h6->ip6_dst, &daddr->v6addr, sizeof (struct in6_addr));
2577
2578 th = (struct tcphdr *)(void *)
2579 ((caddr_t)h6 + sizeof (struct ip6_hdr));
2580 break;
2581#endif /* INET6 */
2582 }
2583
2584 /* TCP header */
2585 th->th_sport = sport;
2586 th->th_dport = dport;
2587 th->th_seq = htonl(seq);
2588 th->th_ack = htonl(ack);
2589 th->th_off = tlen >> 2;
2590 th->th_flags = flags;
2591 th->th_win = htons(win);
2592
2593 if (mss) {
2594 opt = (char *)(th + 1);
2595 opt[0] = TCPOPT_MAXSEG;
2596 opt[1] = 4;
2597#if BYTE_ORDER != BIG_ENDIAN
2598 HTONS(mss);
2599#endif
2600 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2);
2601 }
2602
2603 switch (af) {
2604#if INET
2605 case AF_INET: {
2606 struct route ro;
2607
2608 /* TCP checksum */
2609 th->th_sum = in_cksum(m, len);
2610
2611 /* Finish the IP header */
2612 h->ip_v = 4;
2613 h->ip_hl = sizeof (*h) >> 2;
2614 h->ip_tos = IPTOS_LOWDELAY;
2615 /*
2616 * ip_output() expects ip_len and ip_off to be in host order.
2617 */
2618 h->ip_len = len;
2619 h->ip_off = (path_mtu_discovery ? IP_DF : 0);
2620 h->ip_ttl = ttl ? ttl : ip_defttl;
2621 h->ip_sum = 0;
2622
2623 bzero(&ro, sizeof (ro));
2624 ip_output(m, NULL, &ro, 0, NULL, NULL);
2625 ROUTE_RELEASE(&ro);
2626 break;
2627 }
2628#endif /* INET */
2629#if INET6
2630 case AF_INET6: {
2631 struct route_in6 ro6;
2632
2633 /* TCP checksum */
2634 th->th_sum = in6_cksum(m, IPPROTO_TCP,
2635 sizeof (struct ip6_hdr), tlen);
2636
2637 h6->ip6_vfc |= IPV6_VERSION;
2638 h6->ip6_hlim = IPV6_DEFHLIM;
2639
2640 bzero(&ro6, sizeof (ro6));
2641 ip6_output(m, NULL, &ro6, 0, NULL, NULL, NULL);
2642 ROUTE_RELEASE(&ro6);
2643 break;
2644 }
2645#endif /* INET6 */
2646 }
2647}
2648
2649static void
2650pf_send_icmp(pbuf_t *pbuf, u_int8_t type, u_int8_t code, sa_family_t af,
2651 struct pf_rule *r)
2652{
2653 struct mbuf *m0;
2654 struct pf_mtag *pf_mtag;
2655
2656 m0 = pbuf_clone_to_mbuf(pbuf);
2657 if (m0 == NULL)
2658 return;
2659
2660 if ((pf_mtag = pf_get_mtag(m0)) == NULL)
2661 return;
2662
2663 pf_mtag->pftag_flags |= PF_TAG_GENERATED;
2664
2665 if (PF_RTABLEID_IS_VALID(r->rtableid))
2666 pf_mtag->pftag_rtableid = r->rtableid;
2667
2668#if PF_ECN
2669 /* add hints for ecn */
2670 pf_mtag->pftag_hdr = mtod(m0, struct ip *);
2671 /* record address family */
2672 pf_mtag->pftag_flags &= ~(PF_TAG_HDR_INET | PF_TAG_HDR_INET6);
2673 switch (af) {
2674#if INET
2675 case AF_INET:
2676 pf_mtag->pftag_flags |= PF_TAG_HDR_INET;
2677 m0->m_pkthdr.pkt_proto = IPPROTO_ICMP;
2678 break;
2679#endif /* INET */
2680#if INET6
2681 case AF_INET6:
2682 pf_mtag->pftag_flags |= PF_TAG_HDR_INET6;
2683 m0->m_pkthdr.pkt_proto = IPPROTO_ICMPV6;
2684 break;
2685#endif /* INET6 */
2686 }
2687#endif /* PF_ECN */
2688
2689 switch (af) {
2690#if INET
2691 case AF_INET:
2692 icmp_error(m0, type, code, 0, 0);
2693 break;
2694#endif /* INET */
2695#if INET6
2696 case AF_INET6:
2697 icmp6_error(m0, type, code, 0);
2698 break;
2699#endif /* INET6 */
2700 }
2701}
2702
2703/*
2704 * Return 1 if the addresses a and b match (with mask m), otherwise return 0.
2705 * If n is 0, they match if they are equal. If n is != 0, they match if they
2706 * are different.
2707 */
2708int
2709pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m,
2710 struct pf_addr *b, sa_family_t af)
2711{
2712 int match = 0;
2713
2714 switch (af) {
2715#if INET
2716 case AF_INET:
2717 if ((a->addr32[0] & m->addr32[0]) ==
2718 (b->addr32[0] & m->addr32[0]))
2719 match++;
2720 break;
2721#endif /* INET */
2722#if INET6
2723 case AF_INET6:
2724 if (((a->addr32[0] & m->addr32[0]) ==
2725 (b->addr32[0] & m->addr32[0])) &&
2726 ((a->addr32[1] & m->addr32[1]) ==
2727 (b->addr32[1] & m->addr32[1])) &&
2728 ((a->addr32[2] & m->addr32[2]) ==
2729 (b->addr32[2] & m->addr32[2])) &&
2730 ((a->addr32[3] & m->addr32[3]) ==
2731 (b->addr32[3] & m->addr32[3])))
2732 match++;
2733 break;
2734#endif /* INET6 */
2735 }
2736 if (match) {
2737 if (n)
2738 return (0);
2739 else
2740 return (1);
2741 } else {
2742 if (n)
2743 return (1);
2744 else
2745 return (0);
2746 }
2747}
2748
2749/*
2750 * Return 1 if b <= a <= e, otherwise return 0.
2751 */
2752int
2753pf_match_addr_range(struct pf_addr *b, struct pf_addr *e,
2754 struct pf_addr *a, sa_family_t af)
2755{
2756 switch (af) {
2757#if INET
2758 case AF_INET:
2759 if ((a->addr32[0] < b->addr32[0]) ||
2760 (a->addr32[0] > e->addr32[0]))
2761 return (0);
2762 break;
2763#endif /* INET */
2764#if INET6
2765 case AF_INET6: {
2766 int i;
2767
2768 /* check a >= b */
2769 for (i = 0; i < 4; ++i)
2770 if (a->addr32[i] > b->addr32[i])
2771 break;
2772 else if (a->addr32[i] < b->addr32[i])
2773 return (0);
2774 /* check a <= e */
2775 for (i = 0; i < 4; ++i)
2776 if (a->addr32[i] < e->addr32[i])
2777 break;
2778 else if (a->addr32[i] > e->addr32[i])
2779 return (0);
2780 break;
2781 }
2782#endif /* INET6 */
2783 }
2784 return (1);
2785}
2786
2787int
2788pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p)
2789{
2790 switch (op) {
2791 case PF_OP_IRG:
2792 return ((p > a1) && (p < a2));
2793 case PF_OP_XRG:
2794 return ((p < a1) || (p > a2));
2795 case PF_OP_RRG:
2796 return ((p >= a1) && (p <= a2));
2797 case PF_OP_EQ:
2798 return (p == a1);
2799 case PF_OP_NE:
2800 return (p != a1);
2801 case PF_OP_LT:
2802 return (p < a1);
2803 case PF_OP_LE:
2804 return (p <= a1);
2805 case PF_OP_GT:
2806 return (p > a1);
2807 case PF_OP_GE:
2808 return (p >= a1);
2809 }
2810 return (0); /* never reached */
2811}
2812
2813int
2814pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p)
2815{
2816#if BYTE_ORDER != BIG_ENDIAN
2817 NTOHS(a1);
2818 NTOHS(a2);
2819 NTOHS(p);
2820#endif
2821 return (pf_match(op, a1, a2, p));
2822}
2823
2824int
2825pf_match_xport(u_int8_t proto, u_int8_t proto_variant, union pf_rule_xport *rx,
2826 union pf_state_xport *sx)
2827{
2828 int d = !0;
2829
2830 if (sx) {
2831 switch (proto) {
2832 case IPPROTO_GRE:
2833 if (proto_variant == PF_GRE_PPTP_VARIANT)
2834 d = (rx->call_id == sx->call_id);
2835 break;
2836
2837 case IPPROTO_ESP:
2838 d = (rx->spi == sx->spi);
2839 break;
2840
2841 case IPPROTO_TCP:
2842 case IPPROTO_UDP:
2843 case IPPROTO_ICMP:
2844 case IPPROTO_ICMPV6:
2845 if (rx->range.op)
2846 d = pf_match_port(rx->range.op,
2847 rx->range.port[0], rx->range.port[1],
2848 sx->port);
2849 break;
2850
2851 default:
2852 break;
2853 }
2854 }
2855
2856 return (d);
2857}
2858
2859int
2860pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u)
2861{
2862 if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
2863 return (0);
2864 return (pf_match(op, a1, a2, u));
2865}
2866
2867int
2868pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g)
2869{
2870 if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
2871 return (0);
2872 return (pf_match(op, a1, a2, g));
2873}
2874
2875static int
2876pf_match_tag(struct pf_rule *r, struct pf_mtag *pf_mtag,
2877 int *tag)
2878{
2879 if (*tag == -1)
2880 *tag = pf_mtag->pftag_tag;
2881
2882 return ((!r->match_tag_not && r->match_tag == *tag) ||
2883 (r->match_tag_not && r->match_tag != *tag));
2884}
2885
2886int
2887pf_tag_packet(pbuf_t *pbuf, struct pf_mtag *pf_mtag, int tag,
2888 unsigned int rtableid, struct pf_pdesc *pd)
2889{
2890 if (tag <= 0 && !PF_RTABLEID_IS_VALID(rtableid) &&
2891 (pd == NULL || !(pd->pktflags & PKTF_FLOW_ID)))
2892 return (0);
2893
2894 if (pf_mtag == NULL && (pf_mtag = pf_get_mtag_pbuf(pbuf)) == NULL)
2895 return (1);
2896
2897 if (tag > 0)
2898 pf_mtag->pftag_tag = tag;
2899 if (PF_RTABLEID_IS_VALID(rtableid))
2900 pf_mtag->pftag_rtableid = rtableid;
2901 if (pd != NULL && (pd->pktflags & PKTF_FLOW_ID)) {
2902 *pbuf->pb_flowsrc = pd->flowsrc;
2903 *pbuf->pb_flowid = pd->flowhash;
2904 *pbuf->pb_flags |= pd->pktflags;
2905 *pbuf->pb_proto = pd->proto;
2906 }
2907
2908 return (0);
2909}
2910
2911void
2912pf_step_into_anchor(int *depth, struct pf_ruleset **rs, int n,
2913 struct pf_rule **r, struct pf_rule **a, int *match)
2914{
2915 struct pf_anchor_stackframe *f;
2916
2917 (*r)->anchor->match = 0;
2918 if (match)
2919 *match = 0;
2920 if (*depth >= (int)sizeof (pf_anchor_stack) /
2921 (int)sizeof (pf_anchor_stack[0])) {
2922 printf("pf_step_into_anchor: stack overflow\n");
2923 *r = TAILQ_NEXT(*r, entries);
2924 return;
2925 } else if (*depth == 0 && a != NULL)
2926 *a = *r;
2927 f = pf_anchor_stack + (*depth)++;
2928 f->rs = *rs;
2929 f->r = *r;
2930 if ((*r)->anchor_wildcard) {
2931 f->parent = &(*r)->anchor->children;
2932 if ((f->child = RB_MIN(pf_anchor_node, f->parent)) ==
2933 NULL) {
2934 *r = NULL;
2935 return;
2936 }
2937 *rs = &f->child->ruleset;
2938 } else {
2939 f->parent = NULL;
2940 f->child = NULL;
2941 *rs = &(*r)->anchor->ruleset;
2942 }
2943 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
2944}
2945
2946int
2947pf_step_out_of_anchor(int *depth, struct pf_ruleset **rs, int n,
2948 struct pf_rule **r, struct pf_rule **a, int *match)
2949{
2950 struct pf_anchor_stackframe *f;
2951 int quick = 0;
2952
2953 do {
2954 if (*depth <= 0)
2955 break;
2956 f = pf_anchor_stack + *depth - 1;
2957 if (f->parent != NULL && f->child != NULL) {
2958 if (f->child->match ||
2959 (match != NULL && *match)) {
2960 f->r->anchor->match = 1;
2961 if (match)
2962 *match = 0;
2963 }
2964 f->child = RB_NEXT(pf_anchor_node, f->parent, f->child);
2965 if (f->child != NULL) {
2966 *rs = &f->child->ruleset;
2967 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
2968 if (*r == NULL)
2969 continue;
2970 else
2971 break;
2972 }
2973 }
2974 (*depth)--;
2975 if (*depth == 0 && a != NULL)
2976 *a = NULL;
2977 *rs = f->rs;
2978 if (f->r->anchor->match || (match != NULL && *match))
2979 quick = f->r->quick;
2980 *r = TAILQ_NEXT(f->r, entries);
2981 } while (*r == NULL);
2982
2983 return (quick);
2984}
2985
2986#if INET6
2987void
2988pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr,
2989 struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af)
2990{
2991 switch (af) {
2992#if INET
2993 case AF_INET:
2994 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
2995 ((rmask->addr32[0] ^ 0xffffffff) & saddr->addr32[0]);
2996 break;
2997#endif /* INET */
2998 case AF_INET6:
2999 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
3000 ((rmask->addr32[0] ^ 0xffffffff) & saddr->addr32[0]);
3001 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) |
3002 ((rmask->addr32[1] ^ 0xffffffff) & saddr->addr32[1]);
3003 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) |
3004 ((rmask->addr32[2] ^ 0xffffffff) & saddr->addr32[2]);
3005 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) |
3006 ((rmask->addr32[3] ^ 0xffffffff) & saddr->addr32[3]);
3007 break;
3008 }
3009}
3010
3011void
3012pf_addr_inc(struct pf_addr *addr, sa_family_t af)
3013{
3014 switch (af) {
3015#if INET
3016 case AF_INET:
3017 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1);
3018 break;
3019#endif /* INET */
3020 case AF_INET6:
3021 if (addr->addr32[3] == 0xffffffff) {
3022 addr->addr32[3] = 0;
3023 if (addr->addr32[2] == 0xffffffff) {
3024 addr->addr32[2] = 0;
3025 if (addr->addr32[1] == 0xffffffff) {
3026 addr->addr32[1] = 0;
3027 addr->addr32[0] =
3028 htonl(ntohl(addr->addr32[0]) + 1);
3029 } else
3030 addr->addr32[1] =
3031 htonl(ntohl(addr->addr32[1]) + 1);
3032 } else
3033 addr->addr32[2] =
3034 htonl(ntohl(addr->addr32[2]) + 1);
3035 } else
3036 addr->addr32[3] =
3037 htonl(ntohl(addr->addr32[3]) + 1);
3038 break;
3039 }
3040}
3041#endif /* INET6 */
3042
3043#define mix(a, b, c) \
3044 do { \
3045 a -= b; a -= c; a ^= (c >> 13); \
3046 b -= c; b -= a; b ^= (a << 8); \
3047 c -= a; c -= b; c ^= (b >> 13); \
3048 a -= b; a -= c; a ^= (c >> 12); \
3049 b -= c; b -= a; b ^= (a << 16); \
3050 c -= a; c -= b; c ^= (b >> 5); \
3051 a -= b; a -= c; a ^= (c >> 3); \
3052 b -= c; b -= a; b ^= (a << 10); \
3053 c -= a; c -= b; c ^= (b >> 15); \
3054 } while (0)
3055
3056/*
3057 * hash function based on bridge_hash in if_bridge.c
3058 */
3059static void
3060pf_hash(struct pf_addr *inaddr, struct pf_addr *hash,
3061 struct pf_poolhashkey *key, sa_family_t af)
3062{
3063 u_int32_t a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0];
3064
3065 switch (af) {
3066#if INET
3067 case AF_INET:
3068 a += inaddr->addr32[0];
3069 b += key->key32[1];
3070 mix(a, b, c);
3071 hash->addr32[0] = c + key->key32[2];
3072 break;
3073#endif /* INET */
3074#if INET6
3075 case AF_INET6:
3076 a += inaddr->addr32[0];
3077 b += inaddr->addr32[2];
3078 mix(a, b, c);
3079 hash->addr32[0] = c;
3080 a += inaddr->addr32[1];
3081 b += inaddr->addr32[3];
3082 c += key->key32[1];
3083 mix(a, b, c);
3084 hash->addr32[1] = c;
3085 a += inaddr->addr32[2];
3086 b += inaddr->addr32[1];
3087 c += key->key32[2];
3088 mix(a, b, c);
3089 hash->addr32[2] = c;
3090 a += inaddr->addr32[3];
3091 b += inaddr->addr32[0];
3092 c += key->key32[3];
3093 mix(a, b, c);
3094 hash->addr32[3] = c;
3095 break;
3096#endif /* INET6 */
3097 }
3098}
3099
3100static int
3101pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr,
3102 struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn)
3103{
3104 unsigned char hash[16];
3105 struct pf_pool *rpool = &r->rpool;
3106 struct pf_addr *raddr = &rpool->cur->addr.v.a.addr;
3107 struct pf_addr *rmask = &rpool->cur->addr.v.a.mask;
3108 struct pf_pooladdr *acur = rpool->cur;
3109 struct pf_src_node k;
3110
3111 if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR &&
3112 (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
3113 k.af = af;
3114 PF_ACPY(&k.addr, saddr, af);
3115 if (r->rule_flag & PFRULE_RULESRCTRACK ||
3116 r->rpool.opts & PF_POOL_STICKYADDR)
3117 k.rule.ptr = r;
3118 else
3119 k.rule.ptr = NULL;
3120 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
3121 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
3122 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, rpool->af)) {
3123 PF_ACPY(naddr, &(*sn)->raddr, rpool->af);
3124 if (pf_status.debug >= PF_DEBUG_MISC) {
3125 printf("pf_map_addr: src tracking maps ");
3126 pf_print_host(&k.addr, 0, af);
3127 printf(" to ");
3128 pf_print_host(naddr, 0, rpool->af);
3129 printf("\n");
3130 }
3131 return (0);
3132 }
3133 }
3134
3135 if (rpool->cur->addr.type == PF_ADDR_NOROUTE)
3136 return (1);
3137 if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
3138 if (rpool->cur->addr.p.dyn == NULL)
3139 return (1);
3140 switch (rpool->af) {
3141#if INET
3142 case AF_INET:
3143 if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 &&
3144 (rpool->opts & PF_POOL_TYPEMASK) !=
3145 PF_POOL_ROUNDROBIN)
3146 return (1);
3147 raddr = &rpool->cur->addr.p.dyn->pfid_addr4;
3148 rmask = &rpool->cur->addr.p.dyn->pfid_mask4;
3149 break;
3150#endif /* INET */
3151#if INET6
3152 case AF_INET6:
3153 if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 &&
3154 (rpool->opts & PF_POOL_TYPEMASK) !=
3155 PF_POOL_ROUNDROBIN)
3156 return (1);
3157 raddr = &rpool->cur->addr.p.dyn->pfid_addr6;
3158 rmask = &rpool->cur->addr.p.dyn->pfid_mask6;
3159 break;
3160#endif /* INET6 */
3161 }
3162 } else if (rpool->cur->addr.type == PF_ADDR_TABLE) {
3163 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN)
3164 return (1); /* unsupported */
3165 } else {
3166 raddr = &rpool->cur->addr.v.a.addr;
3167 rmask = &rpool->cur->addr.v.a.mask;
3168 }
3169
3170 switch (rpool->opts & PF_POOL_TYPEMASK) {
3171 case PF_POOL_NONE:
3172 PF_ACPY(naddr, raddr, rpool->af);
3173 break;
3174 case PF_POOL_BITMASK:
3175 ASSERT(af == rpool->af);
3176 PF_POOLMASK(naddr, raddr, rmask, saddr, af);
3177 break;
3178 case PF_POOL_RANDOM:
3179 if (init_addr != NULL && PF_AZERO(init_addr, rpool->af)) {
3180 switch (af) {
3181#if INET
3182 case AF_INET:
3183 rpool->counter.addr32[0] = htonl(random());
3184 break;
3185#endif /* INET */
3186#if INET6
3187 case AF_INET6:
3188 if (rmask->addr32[3] != 0xffffffff)
3189 rpool->counter.addr32[3] =
3190 RandomULong();
3191 else
3192 break;
3193 if (rmask->addr32[2] != 0xffffffff)
3194 rpool->counter.addr32[2] =
3195 RandomULong();
3196 else
3197 break;
3198 if (rmask->addr32[1] != 0xffffffff)
3199 rpool->counter.addr32[1] =
3200 RandomULong();
3201 else
3202 break;
3203 if (rmask->addr32[0] != 0xffffffff)
3204 rpool->counter.addr32[0] =
3205 RandomULong();
3206 break;
3207#endif /* INET6 */
3208 }
3209 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter,
3210 rpool->af);
3211 PF_ACPY(init_addr, naddr, rpool->af);
3212
3213 } else {
3214 PF_AINC(&rpool->counter, rpool->af);
3215 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter,
3216 rpool->af);
3217 }
3218 break;
3219 case PF_POOL_SRCHASH:
3220 ASSERT(af == rpool->af);
3221 PF_POOLMASK(naddr, raddr, rmask, saddr, af);
3222 pf_hash(saddr, (struct pf_addr *)(void *)&hash,
3223 &rpool->key, af);
3224 PF_POOLMASK(naddr, raddr, rmask,
3225 (struct pf_addr *)(void *)&hash, af);
3226 break;
3227 case PF_POOL_ROUNDROBIN:
3228 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
3229 if (!pfr_pool_get(rpool->cur->addr.p.tbl,
3230 &rpool->tblidx, &rpool->counter,
3231 &raddr, &rmask, rpool->af))
3232 goto get_addr;
3233 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
3234 if (rpool->cur->addr.p.dyn != NULL &&
3235 !pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
3236 &rpool->tblidx, &rpool->counter,
3237 &raddr, &rmask, af))
3238 goto get_addr;
3239 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter,
3240 rpool->af))
3241 goto get_addr;
3242
3243 try_next:
3244 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL)
3245 rpool->cur = TAILQ_FIRST(&rpool->list);
3246 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
3247 rpool->tblidx = -1;
3248 if (pfr_pool_get(rpool->cur->addr.p.tbl,
3249 &rpool->tblidx, &rpool->counter,
3250 &raddr, &rmask, rpool->af)) {
3251 /* table contains no address of type
3252 * 'rpool->af' */
3253 if (rpool->cur != acur)
3254 goto try_next;
3255 return (1);
3256 }
3257 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
3258 rpool->tblidx = -1;
3259 if (rpool->cur->addr.p.dyn == NULL)
3260 return (1);
3261 if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
3262 &rpool->tblidx, &rpool->counter,
3263 &raddr, &rmask, rpool->af)) {
3264 /* table contains no address of type
3265 * 'rpool->af' */
3266 if (rpool->cur != acur)
3267 goto try_next;
3268 return (1);
3269 }
3270 } else {
3271 raddr = &rpool->cur->addr.v.a.addr;
3272 rmask = &rpool->cur->addr.v.a.mask;
3273 PF_ACPY(&rpool->counter, raddr, rpool->af);
3274 }
3275
3276 get_addr:
3277 PF_ACPY(naddr, &rpool->counter, rpool->af);
3278 if (init_addr != NULL && PF_AZERO(init_addr, rpool->af))
3279 PF_ACPY(init_addr, naddr, rpool->af);
3280 PF_AINC(&rpool->counter, rpool->af);
3281 break;
3282 }
3283 if (*sn != NULL)
3284 PF_ACPY(&(*sn)->raddr, naddr, rpool->af);
3285
3286 if (pf_status.debug >= PF_DEBUG_MISC &&
3287 (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
3288 printf("pf_map_addr: selected address ");
3289 pf_print_host(naddr, 0, rpool->af);
3290 printf("\n");
3291 }
3292
3293 return (0);
3294}
3295
3296static int
3297pf_get_sport(struct pf_pdesc *pd, struct pfi_kif *kif, struct pf_rule *r,
3298 struct pf_addr *saddr, union pf_state_xport *sxport, struct pf_addr *daddr,
3299 union pf_state_xport *dxport, struct pf_addr *naddr,
3300 union pf_state_xport *nxport, struct pf_src_node **sn
3301 )
3302{
3303#pragma unused(kif)
3304 struct pf_state_key_cmp key;
3305 struct pf_addr init_addr;
3306 unsigned int cut;
3307 sa_family_t af = pd->af;
3308 u_int8_t proto = pd->proto;
3309 unsigned int low = r->rpool.proxy_port[0];
3310 unsigned int high = r->rpool.proxy_port[1];
3311
3312 bzero(&init_addr, sizeof (init_addr));
3313 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
3314 return (1);
3315
3316 if (proto == IPPROTO_ICMP) {
3317 low = 1;
3318 high = 65535;
3319 }
3320
3321 if (!nxport)
3322 return (0); /* No output necessary. */
3323
3324 /*--- Special mapping rules for UDP ---*/
3325 if (proto == IPPROTO_UDP) {
3326
3327 /*--- Never float IKE source port ---*/
3328 if (ntohs(sxport->port) == PF_IKE_PORT) {
3329 nxport->port = sxport->port;
3330 return (0);
3331 }
3332
3333 /*--- Apply exterior mapping options ---*/
3334 if (r->extmap > PF_EXTMAP_APD) {
3335 struct pf_state *s;
3336
3337 TAILQ_FOREACH(s, &state_list, entry_list) {
3338 struct pf_state_key *sk = s->state_key;
3339 if (!sk)
3340 continue;
3341 if (s->nat_rule.ptr != r)
3342 continue;
3343 if (sk->proto != IPPROTO_UDP ||
3344 sk->af_lan != af)
3345 continue;
3346 if (sk->lan.xport.port != sxport->port)
3347 continue;
3348 if (PF_ANEQ(&sk->lan.addr, saddr, af))
3349 continue;
3350 if (r->extmap < PF_EXTMAP_EI &&
3351 PF_ANEQ(&sk->ext_lan.addr, daddr, af))
3352 continue;
3353
3354 nxport->port = sk->gwy.xport.port;
3355 return (0);
3356 }
3357 }
3358 } else if (proto == IPPROTO_TCP) {
3359 struct pf_state* s;
3360 /*
3361 * APPLE MODIFICATION: <rdar://problem/6546358>
3362 * Fix allows....NAT to use a single binding for TCP session
3363 * with same source IP and source port
3364 */
3365 TAILQ_FOREACH(s, &state_list, entry_list) {
3366 struct pf_state_key* sk = s->state_key;
3367 if (!sk)
3368 continue;
3369 if (s->nat_rule.ptr != r)
3370 continue;
3371 if (sk->proto != IPPROTO_TCP || sk->af_lan != af)
3372 continue;
3373 if (sk->lan.xport.port != sxport->port)
3374 continue;
3375 if (!(PF_AEQ(&sk->lan.addr, saddr, af)))
3376 continue;
3377 nxport->port = sk->gwy.xport.port;
3378 return (0);
3379 }
3380 }
3381 do {
3382 key.af_gwy = af;
3383 key.proto = proto;
3384 PF_ACPY(&key.ext_gwy.addr, daddr, key.af_gwy);
3385 PF_ACPY(&key.gwy.addr, naddr, key.af_gwy);
3386 switch (proto) {
3387 case IPPROTO_UDP:
3388 key.proto_variant = r->extfilter;
3389 break;
3390 default:
3391 key.proto_variant = 0;
3392 break;
3393 }
3394 if (dxport)
3395 key.ext_gwy.xport = *dxport;
3396 else
3397 memset(&key.ext_gwy.xport, 0,
3398 sizeof (key.ext_gwy.xport));
3399 /*
3400 * port search; start random, step;
3401 * similar 2 portloop in in_pcbbind
3402 */
3403 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP ||
3404 proto == IPPROTO_ICMP)) {
3405 if (dxport)
3406 key.gwy.xport = *dxport;
3407 else
3408 memset(&key.gwy.xport, 0,
3409 sizeof (key.gwy.xport));
3410 if (pf_find_state_all(&key, PF_IN, NULL) == NULL)
3411 return (0);
3412 } else if (low == 0 && high == 0) {
3413 key.gwy.xport = *nxport;
3414 if (pf_find_state_all(&key, PF_IN, NULL) == NULL
3415 ) {
3416 return (0);
3417 }
3418 } else if (low == high) {
3419 key.gwy.xport.port = htons(low);
3420 if (pf_find_state_all(&key, PF_IN, NULL) == NULL
3421 ) {
3422 nxport->port = htons(low);
3423 return (0);
3424 }
3425 } else {
3426 unsigned int tmp;
3427 if (low > high) {
3428 tmp = low;
3429 low = high;
3430 high = tmp;
3431 }
3432 /* low < high */
3433 cut = htonl(random()) % (1 + high - low) + low;
3434 /* low <= cut <= high */
3435 for (tmp = cut; tmp <= high; ++(tmp)) {
3436 key.gwy.xport.port = htons(tmp);
3437 if (pf_find_state_all(&key, PF_IN, NULL) == NULL
3438 ) {
3439 nxport->port = htons(tmp);
3440 return (0);
3441 }
3442 }
3443 for (tmp = cut - 1; tmp >= low; --(tmp)) {
3444 key.gwy.xport.port = htons(tmp);
3445 if (pf_find_state_all(&key, PF_IN, NULL) == NULL
3446 ) {
3447 nxport->port = htons(tmp);
3448 return (0);
3449 }
3450 }
3451 }
3452
3453 switch (r->rpool.opts & PF_POOL_TYPEMASK) {
3454 case PF_POOL_RANDOM:
3455 case PF_POOL_ROUNDROBIN:
3456 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
3457 return (1);
3458 break;
3459 case PF_POOL_NONE:
3460 case PF_POOL_SRCHASH:
3461 case PF_POOL_BITMASK:
3462 default:
3463 return (1);
3464 }
3465 } while (!PF_AEQ(&init_addr, naddr, af));
3466
3467 return (1); /* none available */
3468}
3469
3470static struct pf_rule *
3471pf_match_translation(struct pf_pdesc *pd, pbuf_t *pbuf, int off,
3472 int direction, struct pfi_kif *kif, struct pf_addr *saddr,
3473 union pf_state_xport *sxport, struct pf_addr *daddr,
3474 union pf_state_xport *dxport, int rs_num)
3475{
3476 struct pf_rule *r, *rm = NULL;
3477 struct pf_ruleset *ruleset = NULL;
3478 int tag = -1;
3479 unsigned int rtableid = IFSCOPE_NONE;
3480 int asd = 0;
3481
3482 r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr);
3483 while (r && rm == NULL) {
3484 struct pf_rule_addr *src = NULL, *dst = NULL;
3485 struct pf_addr_wrap *xdst = NULL;
3486 struct pf_addr_wrap *xsrc = NULL;
3487 union pf_rule_xport rdrxport;
3488
3489 if (r->action == PF_BINAT && direction == PF_IN) {
3490 src = &r->dst;
3491 if (r->rpool.cur != NULL)
3492 xdst = &r->rpool.cur->addr;
3493 } else if (r->action == PF_RDR && direction == PF_OUT) {
3494 dst = &r->src;
3495 src = &r->dst;
3496 if (r->rpool.cur != NULL) {
3497 rdrxport.range.op = PF_OP_EQ;
3498 rdrxport.range.port[0] =
3499 htons(r->rpool.proxy_port[0]);
3500 xsrc = &r->rpool.cur->addr;
3501 }
3502 } else {
3503 src = &r->src;
3504 dst = &r->dst;
3505 }
3506
3507 r->evaluations++;
3508 if (pfi_kif_match(r->kif, kif) == r->ifnot)
3509 r = r->skip[PF_SKIP_IFP].ptr;
3510 else if (r->direction && r->direction != direction)
3511 r = r->skip[PF_SKIP_DIR].ptr;
3512 else if (r->af && r->af != pd->af)
3513 r = r->skip[PF_SKIP_AF].ptr;
3514 else if (r->proto && r->proto != pd->proto)
3515 r = r->skip[PF_SKIP_PROTO].ptr;
3516 else if (xsrc && PF_MISMATCHAW(xsrc, saddr, pd->af, 0, NULL))
3517 r = TAILQ_NEXT(r, entries);
3518 else if (!xsrc && PF_MISMATCHAW(&src->addr, saddr, pd->af,
3519 src->neg, kif))
3520 r = TAILQ_NEXT(r, entries);
3521 else if (xsrc && (!rdrxport.range.port[0] ||
3522 !pf_match_xport(r->proto, r->proto_variant, &rdrxport,
3523 sxport)))
3524 r = TAILQ_NEXT(r, entries);
3525 else if (!xsrc && !pf_match_xport(r->proto,
3526 r->proto_variant, &src->xport, sxport))
3527 r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT :
3528 PF_SKIP_DST_PORT].ptr;
3529 else if (dst != NULL &&
3530 PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->neg, NULL))
3531 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3532 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af,
3533 0, NULL))
3534 r = TAILQ_NEXT(r, entries);
3535 else if (dst && !pf_match_xport(r->proto, r->proto_variant,
3536 &dst->xport, dxport))
3537 r = r->skip[PF_SKIP_DST_PORT].ptr;
3538 else if (r->match_tag && !pf_match_tag(r, pd->pf_mtag, &tag))
3539 r = TAILQ_NEXT(r, entries);
3540 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto !=
3541 IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, pbuf,
3542 off, pd->hdr.tcp), r->os_fingerprint)))
3543 r = TAILQ_NEXT(r, entries);
3544 else {
3545 if (r->tag)
3546 tag = r->tag;
3547 if (PF_RTABLEID_IS_VALID(r->rtableid))
3548 rtableid = r->rtableid;
3549 if (r->anchor == NULL) {
3550 rm = r;
3551 } else
3552 pf_step_into_anchor(&asd, &ruleset, rs_num,
3553 &r, NULL, NULL);
3554 }
3555 if (r == NULL)
3556 pf_step_out_of_anchor(&asd, &ruleset, rs_num, &r,
3557 NULL, NULL);
3558 }
3559 if (pf_tag_packet(pbuf, pd->pf_mtag, tag, rtableid, NULL))
3560 return (NULL);
3561 if (rm != NULL && (rm->action == PF_NONAT ||
3562 rm->action == PF_NORDR || rm->action == PF_NOBINAT ||
3563 rm->action == PF_NONAT64))
3564 return (NULL);
3565 return (rm);
3566}
3567
3568/*
3569 * Get address translation information for NAT/BINAT/RDR
3570 * pd : pf packet descriptor
3571 * pbuf : pbuf holding the packet
3572 * off : offset to protocol header
3573 * direction : direction of packet
3574 * kif : pf interface info obtained from the packet's recv interface
3575 * sn : source node pointer (output)
3576 * saddr : packet source address
3577 * sxport : packet source port
3578 * daddr : packet destination address
3579 * dxport : packet destination port
3580 * nsxport : translated source port (output)
3581 *
3582 * Translated source & destination address are updated in pd->nsaddr &
3583 * pd->ndaddr
3584 */
3585static struct pf_rule *
3586pf_get_translation_aux(struct pf_pdesc *pd, pbuf_t *pbuf, int off,
3587 int direction, struct pfi_kif *kif, struct pf_src_node **sn,
3588 struct pf_addr *saddr, union pf_state_xport *sxport, struct pf_addr *daddr,
3589 union pf_state_xport *dxport, union pf_state_xport *nsxport
3590 )
3591{
3592 struct pf_rule *r = NULL;
3593 pd->naf = pd->af;
3594
3595 if (direction == PF_OUT) {
3596 r = pf_match_translation(pd, pbuf, off, direction, kif, saddr,
3597 sxport, daddr, dxport, PF_RULESET_BINAT);
3598 if (r == NULL)
3599 r = pf_match_translation(pd, pbuf, off, direction, kif,
3600 saddr, sxport, daddr, dxport, PF_RULESET_RDR);
3601 if (r == NULL)
3602 r = pf_match_translation(pd, pbuf, off, direction, kif,
3603 saddr, sxport, daddr, dxport, PF_RULESET_NAT);
3604 } else {
3605 r = pf_match_translation(pd, pbuf, off, direction, kif, saddr,
3606 sxport, daddr, dxport, PF_RULESET_RDR);
3607 if (r == NULL)
3608 r = pf_match_translation(pd, pbuf, off, direction, kif,
3609 saddr, sxport, daddr, dxport, PF_RULESET_BINAT);
3610 }
3611
3612 if (r != NULL) {
3613 struct pf_addr *nsaddr = &pd->naddr;
3614 struct pf_addr *ndaddr = &pd->ndaddr;
3615
3616 *nsaddr = *saddr;
3617 *ndaddr = *daddr;
3618
3619 switch (r->action) {
3620 case PF_NONAT:
3621 case PF_NONAT64:
3622 case PF_NOBINAT:
3623 case PF_NORDR:
3624 return (NULL);
3625 case PF_NAT:
3626 case PF_NAT64:
3627 /*
3628 * we do NAT64 on incoming path and we call ip_input
3629 * which asserts receive interface to be not NULL.
3630 * The below check is to prevent NAT64 action on any
3631 * packet generated by local entity using synthesized
3632 * IPv6 address.
3633 */
3634 if ((r->action == PF_NAT64) && (direction == PF_OUT))
3635 return (NULL);
3636
3637 if (pf_get_sport(pd, kif, r, saddr, sxport, daddr,
3638 dxport, nsaddr, nsxport, sn
3639 ))
3640 {
3641 DPFPRINTF(PF_DEBUG_MISC,
3642 ("pf: NAT proxy port allocation "
3643 "(%u-%u) failed\n",
3644 r->rpool.proxy_port[0],
3645 r->rpool.proxy_port[1]));
3646 return (NULL);
3647 }
3648 /*
3649 * For NAT64 the destination IPv4 address is derived
3650 * from the last 32 bits of synthesized IPv6 address
3651 */
3652 if (r->action == PF_NAT64) {
3653 ndaddr->v4addr.s_addr = daddr->addr32[3];
3654 pd->naf = AF_INET;
3655 }
3656 break;
3657 case PF_BINAT:
3658 switch (direction) {
3659 case PF_OUT:
3660 if (r->rpool.cur->addr.type ==
3661 PF_ADDR_DYNIFTL) {
3662 if (r->rpool.cur->addr.p.dyn == NULL)
3663 return (NULL);
3664 switch (pd->af) {
3665#if INET
3666 case AF_INET:
3667 if (r->rpool.cur->addr.p.dyn->
3668 pfid_acnt4 < 1)
3669 return (NULL);
3670 PF_POOLMASK(nsaddr,
3671 &r->rpool.cur->addr.p.dyn->
3672 pfid_addr4,
3673 &r->rpool.cur->addr.p.dyn->
3674 pfid_mask4,
3675 saddr, AF_INET);
3676 break;
3677#endif /* INET */
3678#if INET6
3679 case AF_INET6:
3680 if (r->rpool.cur->addr.p.dyn->
3681 pfid_acnt6 < 1)
3682 return (NULL);
3683 PF_POOLMASK(nsaddr,
3684 &r->rpool.cur->addr.p.dyn->
3685 pfid_addr6,
3686 &r->rpool.cur->addr.p.dyn->
3687 pfid_mask6,
3688 saddr, AF_INET6);
3689 break;
3690#endif /* INET6 */
3691 }
3692 } else {
3693 PF_POOLMASK(nsaddr,
3694 &r->rpool.cur->addr.v.a.addr,
3695 &r->rpool.cur->addr.v.a.mask,
3696 saddr, pd->af);
3697 }
3698 break;
3699 case PF_IN:
3700 if (r->src.addr.type == PF_ADDR_DYNIFTL) {
3701 if (r->src.addr.p.dyn == NULL)
3702 return (NULL);
3703 switch (pd->af) {
3704#if INET
3705 case AF_INET:
3706 if (r->src.addr.p.dyn->
3707 pfid_acnt4 < 1)
3708 return (NULL);
3709 PF_POOLMASK(ndaddr,
3710 &r->src.addr.p.dyn->
3711 pfid_addr4,
3712 &r->src.addr.p.dyn->
3713 pfid_mask4,
3714 daddr, AF_INET);
3715 break;
3716#endif /* INET */
3717#if INET6
3718 case AF_INET6:
3719 if (r->src.addr.p.dyn->
3720 pfid_acnt6 < 1)
3721 return (NULL);
3722 PF_POOLMASK(ndaddr,
3723 &r->src.addr.p.dyn->
3724 pfid_addr6,
3725 &r->src.addr.p.dyn->
3726 pfid_mask6,
3727 daddr, AF_INET6);
3728 break;
3729#endif /* INET6 */
3730 }
3731 } else
3732 PF_POOLMASK(ndaddr,
3733 &r->src.addr.v.a.addr,
3734 &r->src.addr.v.a.mask, daddr,
3735 pd->af);
3736 break;
3737 }
3738 break;
3739 case PF_RDR: {
3740 switch (direction) {
3741 case PF_OUT:
3742 if (r->dst.addr.type == PF_ADDR_DYNIFTL) {
3743 if (r->dst.addr.p.dyn == NULL)
3744 return (NULL);
3745 switch (pd->af) {
3746#if INET
3747 case AF_INET:
3748 if (r->dst.addr.p.dyn->
3749 pfid_acnt4 < 1)
3750 return (NULL);
3751 PF_POOLMASK(nsaddr,
3752 &r->dst.addr.p.dyn->
3753 pfid_addr4,
3754 &r->dst.addr.p.dyn->
3755 pfid_mask4,
3756 daddr, AF_INET);
3757 break;
3758#endif /* INET */
3759#if INET6
3760 case AF_INET6:
3761 if (r->dst.addr.p.dyn->
3762 pfid_acnt6 < 1)
3763 return (NULL);
3764 PF_POOLMASK(nsaddr,
3765 &r->dst.addr.p.dyn->
3766 pfid_addr6,
3767 &r->dst.addr.p.dyn->
3768 pfid_mask6,
3769 daddr, AF_INET6);
3770 break;
3771#endif /* INET6 */
3772 }
3773 } else {
3774 PF_POOLMASK(nsaddr,
3775 &r->dst.addr.v.a.addr,
3776 &r->dst.addr.v.a.mask,
3777 daddr, pd->af);
3778 }
3779 if (nsxport && r->dst.xport.range.port[0])
3780 nsxport->port =
3781 r->dst.xport.range.port[0];
3782 break;
3783 case PF_IN:
3784 if (pf_map_addr(pd->af, r, saddr,
3785 ndaddr, NULL, sn))
3786 return (NULL);
3787 if ((r->rpool.opts & PF_POOL_TYPEMASK) ==
3788 PF_POOL_BITMASK)
3789 PF_POOLMASK(ndaddr, ndaddr,
3790 &r->rpool.cur->addr.v.a.mask, daddr,
3791 pd->af);
3792
3793 if (nsxport && dxport) {
3794 if (r->rpool.proxy_port[1]) {
3795 u_int32_t tmp_nport;
3796
3797 tmp_nport =
3798 ((ntohs(dxport->port) -
3799 ntohs(r->dst.xport.range.
3800 port[0])) %
3801 (r->rpool.proxy_port[1] -
3802 r->rpool.proxy_port[0] +
3803 1)) + r->rpool.proxy_port[0];
3804
3805 /* wrap around if necessary */
3806 if (tmp_nport > 65535)
3807 tmp_nport -= 65535;
3808 nsxport->port =
3809 htons((u_int16_t)tmp_nport);
3810 } else if (r->rpool.proxy_port[0]) {
3811 nsxport->port = htons(r->rpool.
3812 proxy_port[0]);
3813 }
3814 }
3815 break;
3816 }
3817 break;
3818 }
3819 default:
3820 return (NULL);
3821 }
3822 }
3823
3824 return (r);
3825}
3826
3827int
3828pf_socket_lookup(int direction, struct pf_pdesc *pd)
3829{
3830 struct pf_addr *saddr, *daddr;
3831 u_int16_t sport, dport;
3832 struct inpcbinfo *pi;
3833 int inp = 0;
3834
3835 if (pd == NULL)
3836 return (-1);
3837 pd->lookup.uid = UID_MAX;
3838 pd->lookup.gid = GID_MAX;
3839 pd->lookup.pid = NO_PID;
3840
3841 switch (pd->proto) {
3842 case IPPROTO_TCP:
3843 if (pd->hdr.tcp == NULL)
3844 return (-1);
3845 sport = pd->hdr.tcp->th_sport;
3846 dport = pd->hdr.tcp->th_dport;
3847 pi = &tcbinfo;
3848 break;
3849 case IPPROTO_UDP:
3850 if (pd->hdr.udp == NULL)
3851 return (-1);
3852 sport = pd->hdr.udp->uh_sport;
3853 dport = pd->hdr.udp->uh_dport;
3854 pi = &udbinfo;
3855 break;
3856 default:
3857 return (-1);
3858 }
3859 if (direction == PF_IN) {
3860 saddr = pd->src;
3861 daddr = pd->dst;
3862 } else {
3863 u_int16_t p;
3864
3865 p = sport;
3866 sport = dport;
3867 dport = p;
3868 saddr = pd->dst;
3869 daddr = pd->src;
3870 }
3871 switch (pd->af) {
3872#if INET
3873 case AF_INET:
3874 inp = in_pcblookup_hash_exists(pi, saddr->v4addr, sport, daddr->v4addr, dport,
3875 0, &pd->lookup.uid, &pd->lookup.gid, NULL);
3876#if INET6
3877 if (inp == 0) {
3878 struct in6_addr s6, d6;
3879
3880 memset(&s6, 0, sizeof (s6));
3881 s6.s6_addr16[5] = htons(0xffff);
3882 memcpy(&s6.s6_addr32[3], &saddr->v4addr,
3883 sizeof (saddr->v4addr));
3884
3885 memset(&d6, 0, sizeof (d6));
3886 d6.s6_addr16[5] = htons(0xffff);
3887 memcpy(&d6.s6_addr32[3], &daddr->v4addr,
3888 sizeof (daddr->v4addr));
3889
3890 inp = in6_pcblookup_hash_exists(pi, &s6, sport,
3891 &d6, dport, 0, &pd->lookup.uid, &pd->lookup.gid, NULL);
3892 if (inp == 0) {
3893 inp = in_pcblookup_hash_exists(pi, saddr->v4addr, sport,
3894 daddr->v4addr, dport, INPLOOKUP_WILDCARD, &pd->lookup.uid, &pd->lookup.gid, NULL);
3895 if (inp == 0) {
3896 inp = in6_pcblookup_hash_exists(pi, &s6, sport,
3897 &d6, dport, INPLOOKUP_WILDCARD,
3898 &pd->lookup.uid, &pd->lookup.gid, NULL);
3899 if (inp == 0)
3900 return (-1);
3901 }
3902 }
3903 }
3904#else
3905 if (inp == 0) {
3906 inp = in_pcblookup_hash_exists(pi, saddr->v4addr, sport,
3907 daddr->v4addr, dport, INPLOOKUP_WILDCARD,
3908 &pd->lookup.uid, &pd->lookup.gid, NULL);
3909 if (inp == 0)
3910 return (-1);
3911 }
3912#endif /* !INET6 */
3913 break;
3914#endif /* INET */
3915#if INET6
3916 case AF_INET6:
3917 inp = in6_pcblookup_hash_exists(pi, &saddr->v6addr, sport, &daddr->v6addr,
3918 dport, 0, &pd->lookup.uid, &pd->lookup.gid, NULL);
3919 if (inp == 0) {
3920 inp = in6_pcblookup_hash_exists(pi, &saddr->v6addr, sport,
3921 &daddr->v6addr, dport, INPLOOKUP_WILDCARD,
3922 &pd->lookup.uid, &pd->lookup.gid, NULL);
3923 if (inp == 0)
3924 return (-1);
3925 }
3926 break;
3927#endif /* INET6 */
3928
3929 default:
3930 return (-1);
3931 }
3932
3933 return (1);
3934}
3935
3936static u_int8_t
3937pf_get_wscale(pbuf_t *pbuf, int off, u_int16_t th_off, sa_family_t af)
3938{
3939 int hlen;
3940 u_int8_t hdr[60];
3941 u_int8_t *opt, optlen;
3942 u_int8_t wscale = 0;
3943
3944 hlen = th_off << 2; /* hlen <= sizeof (hdr) */
3945 if (hlen <= (int)sizeof (struct tcphdr))
3946 return (0);
3947 if (!pf_pull_hdr(pbuf, off, hdr, hlen, NULL, NULL, af))
3948 return (0);
3949 opt = hdr + sizeof (struct tcphdr);
3950 hlen -= sizeof (struct tcphdr);
3951 while (hlen >= 3) {
3952 switch (*opt) {
3953 case TCPOPT_EOL:
3954 case TCPOPT_NOP:
3955 ++opt;
3956 --hlen;
3957 break;
3958 case TCPOPT_WINDOW:
3959 wscale = opt[2];
3960 if (wscale > TCP_MAX_WINSHIFT)
3961 wscale = TCP_MAX_WINSHIFT;
3962 wscale |= PF_WSCALE_FLAG;
3963 /* FALLTHROUGH */
3964 default:
3965 optlen = opt[1];
3966 if (optlen < 2)
3967 optlen = 2;
3968 hlen -= optlen;
3969 opt += optlen;
3970 break;
3971 }
3972 }
3973 return (wscale);
3974}
3975
3976static u_int16_t
3977pf_get_mss(pbuf_t *pbuf, int off, u_int16_t th_off, sa_family_t af)
3978{
3979 int hlen;
3980 u_int8_t hdr[60];
3981 u_int8_t *opt, optlen;
3982 u_int16_t mss = tcp_mssdflt;
3983
3984 hlen = th_off << 2; /* hlen <= sizeof (hdr) */
3985 if (hlen <= (int)sizeof (struct tcphdr))
3986 return (0);
3987 if (!pf_pull_hdr(pbuf, off, hdr, hlen, NULL, NULL, af))
3988 return (0);
3989 opt = hdr + sizeof (struct tcphdr);
3990 hlen -= sizeof (struct tcphdr);
3991 while (hlen >= TCPOLEN_MAXSEG) {
3992 switch (*opt) {
3993 case TCPOPT_EOL:
3994 case TCPOPT_NOP:
3995 ++opt;
3996 --hlen;
3997 break;
3998 case TCPOPT_MAXSEG:
3999 bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2);
4000#if BYTE_ORDER != BIG_ENDIAN
4001 NTOHS(mss);
4002#endif
4003 /* FALLTHROUGH */
4004 default:
4005 optlen = opt[1];
4006 if (optlen < 2)
4007 optlen = 2;
4008 hlen -= optlen;
4009 opt += optlen;
4010 break;
4011 }
4012 }
4013 return (mss);
4014}
4015
4016static u_int16_t
4017pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer)
4018{
4019#if INET
4020 struct sockaddr_in *dst;
4021 struct route ro;
4022#endif /* INET */
4023#if INET6
4024 struct sockaddr_in6 *dst6;
4025 struct route_in6 ro6;
4026#endif /* INET6 */
4027 struct rtentry *rt = NULL;
4028 int hlen;
4029 u_int16_t mss = tcp_mssdflt;
4030
4031 switch (af) {
4032#if INET
4033 case AF_INET:
4034 hlen = sizeof (struct ip);
4035 bzero(&ro, sizeof (ro));
4036 dst = (struct sockaddr_in *)(void *)&ro.ro_dst;
4037 dst->sin_family = AF_INET;
4038 dst->sin_len = sizeof (*dst);
4039 dst->sin_addr = addr->v4addr;
4040 rtalloc(&ro);
4041 rt = ro.ro_rt;
4042 break;
4043#endif /* INET */
4044#if INET6
4045 case AF_INET6:
4046 hlen = sizeof (struct ip6_hdr);
4047 bzero(&ro6, sizeof (ro6));
4048 dst6 = (struct sockaddr_in6 *)(void *)&ro6.ro_dst;
4049 dst6->sin6_family = AF_INET6;
4050 dst6->sin6_len = sizeof (*dst6);
4051 dst6->sin6_addr = addr->v6addr;
4052 rtalloc((struct route *)&ro);
4053 rt = ro6.ro_rt;
4054 break;
4055#endif /* INET6 */
4056 default:
4057 panic("pf_calc_mss: not AF_INET or AF_INET6!");
4058 return (0);
4059 }
4060
4061 if (rt && rt->rt_ifp) {
4062 /* This is relevant only for PF SYN Proxy */
4063 int interface_mtu = rt->rt_ifp->if_mtu;
4064
4065 if (af == AF_INET &&
4066 INTF_ADJUST_MTU_FOR_CLAT46(rt->rt_ifp)) {
4067 interface_mtu = IN6_LINKMTU(rt->rt_ifp);
4068 /* Further adjust the size for CLAT46 expansion */
4069 interface_mtu -= CLAT46_HDR_EXPANSION_OVERHD;
4070 }
4071 mss = interface_mtu - hlen - sizeof (struct tcphdr);
4072 mss = max(tcp_mssdflt, mss);
4073 rtfree(rt);
4074 }
4075 mss = min(mss, offer);
4076 mss = max(mss, 64); /* sanity - at least max opt space */
4077 return (mss);
4078}
4079
4080static void
4081pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr, sa_family_t af)
4082{
4083 struct pf_rule *r = s->rule.ptr;
4084
4085 s->rt_kif = NULL;
4086
4087 if (!r->rt || r->rt == PF_FASTROUTE)
4088 return;
4089 if ((af == AF_INET) || (af == AF_INET6)) {
4090 pf_map_addr(af, r, saddr, &s->rt_addr, NULL,
4091 &s->nat_src_node);
4092 s->rt_kif = r->rpool.cur->kif;
4093 }
4094
4095 return;
4096}
4097
4098static void
4099pf_attach_state(struct pf_state_key *sk, struct pf_state *s, int tail)
4100{
4101 s->state_key = sk;
4102 sk->refcnt++;
4103
4104 /* list is sorted, if-bound states before floating */
4105 if (tail)
4106 TAILQ_INSERT_TAIL(&sk->states, s, next);
4107 else
4108 TAILQ_INSERT_HEAD(&sk->states, s, next);
4109}
4110
4111static void
4112pf_detach_state(struct pf_state *s, int flags)
4113{
4114 struct pf_state_key *sk = s->state_key;
4115
4116 if (sk == NULL)
4117 return;
4118
4119 s->state_key = NULL;
4120 TAILQ_REMOVE(&sk->states, s, next);
4121 if (--sk->refcnt == 0) {
4122 if (!(flags & PF_DT_SKIP_EXTGWY))
4123 RB_REMOVE(pf_state_tree_ext_gwy,
4124 &pf_statetbl_ext_gwy, sk);
4125 if (!(flags & PF_DT_SKIP_LANEXT))
4126 RB_REMOVE(pf_state_tree_lan_ext,
4127 &pf_statetbl_lan_ext, sk);
4128 if (sk->app_state)
4129 pool_put(&pf_app_state_pl, sk->app_state);
4130 pool_put(&pf_state_key_pl, sk);
4131 }
4132}
4133
4134struct pf_state_key *
4135pf_alloc_state_key(struct pf_state *s, struct pf_state_key *psk)
4136{
4137 struct pf_state_key *sk;
4138
4139 if ((sk = pool_get(&pf_state_key_pl, PR_WAITOK)) == NULL)
4140 return (NULL);
4141 bzero(sk, sizeof (*sk));
4142 TAILQ_INIT(&sk->states);
4143 pf_attach_state(sk, s, 0);
4144
4145 /* initialize state key from psk, if provided */
4146 if (psk != NULL) {
4147 bcopy(&psk->lan, &sk->lan, sizeof (sk->lan));
4148 bcopy(&psk->gwy, &sk->gwy, sizeof (sk->gwy));
4149 bcopy(&psk->ext_lan, &sk->ext_lan, sizeof (sk->ext_lan));
4150 bcopy(&psk->ext_gwy, &sk->ext_gwy, sizeof (sk->ext_gwy));
4151 sk->af_lan = psk->af_lan;
4152 sk->af_gwy = psk->af_gwy;
4153 sk->proto = psk->proto;
4154 sk->direction = psk->direction;
4155 sk->proto_variant = psk->proto_variant;
4156 VERIFY(psk->app_state == NULL);
4157 sk->flowsrc = psk->flowsrc;
4158 sk->flowhash = psk->flowhash;
4159 /* don't touch tree entries, states and refcnt on sk */
4160 }
4161
4162 return (sk);
4163}
4164
4165static u_int32_t
4166pf_tcp_iss(struct pf_pdesc *pd)
4167{
4168 MD5_CTX ctx;
4169 u_int32_t digest[4];
4170
4171 if (pf_tcp_secret_init == 0) {
4172 read_frandom(pf_tcp_secret, sizeof (pf_tcp_secret));
4173 MD5Init(&pf_tcp_secret_ctx);
4174 MD5Update(&pf_tcp_secret_ctx, pf_tcp_secret,
4175 sizeof (pf_tcp_secret));
4176 pf_tcp_secret_init = 1;
4177 }
4178 ctx = pf_tcp_secret_ctx;
4179
4180 MD5Update(&ctx, (char *)&pd->hdr.tcp->th_sport, sizeof (u_short));
4181 MD5Update(&ctx, (char *)&pd->hdr.tcp->th_dport, sizeof (u_short));
4182 if (pd->af == AF_INET6) {
4183 MD5Update(&ctx, (char *)&pd->src->v6addr, sizeof (struct in6_addr));
4184 MD5Update(&ctx, (char *)&pd->dst->v6addr, sizeof (struct in6_addr));
4185 } else {
4186 MD5Update(&ctx, (char *)&pd->src->v4addr, sizeof (struct in_addr));
4187 MD5Update(&ctx, (char *)&pd->dst->v4addr, sizeof (struct in_addr));
4188 }
4189 MD5Final((u_char *)digest, &ctx);
4190 pf_tcp_iss_off += 4096;
4191 return (digest[0] + random() + pf_tcp_iss_off);
4192}
4193
4194/*
4195 * This routine is called to perform address family translation on the
4196 * inner IP header (that may come as payload) of an ICMP(v4addr/6) error
4197 * response.
4198 */
4199static int
4200pf_change_icmp_af(pbuf_t *pbuf, int off,
4201 struct pf_pdesc *pd, struct pf_pdesc *pd2, struct pf_addr *src,
4202 struct pf_addr *dst, sa_family_t af, sa_family_t naf)
4203{
4204 struct ip *ip4 = NULL;
4205 struct ip6_hdr *ip6 = NULL;
4206 void *hdr;
4207 int hlen, olen;
4208
4209 if (af == naf || (af != AF_INET && af != AF_INET6) ||
4210 (naf != AF_INET && naf != AF_INET6))
4211 return (-1);
4212
4213 /* old header */
4214 olen = pd2->off - off;
4215 /* new header */
4216 hlen = naf == AF_INET ? sizeof(*ip4) : sizeof(*ip6);
4217
4218 /* Modify the pbuf to accommodate the new header */
4219 hdr = pbuf_resize_segment(pbuf, off, olen, hlen);
4220 if (hdr == NULL)
4221 return (-1);
4222
4223 /* translate inner ip/ip6 header */
4224 switch (naf) {
4225 case AF_INET:
4226 ip4 = hdr;
4227 bzero(ip4, sizeof(*ip4));
4228 ip4->ip_v = IPVERSION;
4229 ip4->ip_hl = sizeof(*ip4) >> 2;
4230 ip4->ip_len = htons(sizeof(*ip4) + pd2->tot_len - olen);
4231 ip4->ip_id = rfc6864 ? 0 : htons(ip_randomid());
4232 ip4->ip_off = htons(IP_DF);
4233 ip4->ip_ttl = pd2->ttl;
4234 if (pd2->proto == IPPROTO_ICMPV6)
4235 ip4->ip_p = IPPROTO_ICMP;
4236 else
4237 ip4->ip_p = pd2->proto;
4238 ip4->ip_src = src->v4addr;
4239 ip4->ip_dst = dst->v4addr;
4240 ip4->ip_sum = pbuf_inet_cksum(pbuf, 0, 0, ip4->ip_hl << 2);
4241 break;
4242 case AF_INET6:
4243 ip6 = hdr;
4244 bzero(ip6, sizeof(*ip6));
4245 ip6->ip6_vfc = IPV6_VERSION;
4246 ip6->ip6_plen = htons(pd2->tot_len - olen);
4247 if (pd2->proto == IPPROTO_ICMP)
4248 ip6->ip6_nxt = IPPROTO_ICMPV6;
4249 else
4250 ip6->ip6_nxt = pd2->proto;
4251 if (!pd2->ttl || pd2->ttl > IPV6_DEFHLIM)
4252 ip6->ip6_hlim = IPV6_DEFHLIM;
4253 else
4254 ip6->ip6_hlim = pd2->ttl;
4255 ip6->ip6_src = src->v6addr;
4256 ip6->ip6_dst = dst->v6addr;
4257 break;
4258 }
4259
4260 /* adjust payload offset and total packet length */
4261 pd2->off += hlen - olen;
4262 pd->tot_len += hlen - olen;
4263
4264 return (0);
4265}
4266
4267#define PTR_IP(field) ((int32_t)offsetof(struct ip, field))
4268#define PTR_IP6(field) ((int32_t)offsetof(struct ip6_hdr, field))
4269
4270static int
4271pf_translate_icmp_af(int af, void *arg)
4272{
4273 struct icmp *icmp4;
4274 struct icmp6_hdr *icmp6;
4275 u_int32_t mtu;
4276 int32_t ptr = -1;
4277 u_int8_t type;
4278 u_int8_t code;
4279
4280 switch (af) {
4281 case AF_INET:
4282 icmp6 = arg;
4283 type = icmp6->icmp6_type;
4284 code = icmp6->icmp6_code;
4285 mtu = ntohl(icmp6->icmp6_mtu);
4286
4287 switch (type) {
4288 case ICMP6_ECHO_REQUEST:
4289 type = ICMP_ECHO;
4290 break;
4291 case ICMP6_ECHO_REPLY:
4292 type = ICMP_ECHOREPLY;
4293 break;
4294 case ICMP6_DST_UNREACH:
4295 type = ICMP_UNREACH;
4296 switch (code) {
4297 case ICMP6_DST_UNREACH_NOROUTE:
4298 case ICMP6_DST_UNREACH_BEYONDSCOPE:
4299 case ICMP6_DST_UNREACH_ADDR:
4300 code = ICMP_UNREACH_HOST;
4301 break;
4302 case ICMP6_DST_UNREACH_ADMIN:
4303 code = ICMP_UNREACH_HOST_PROHIB;
4304 break;
4305 case ICMP6_DST_UNREACH_NOPORT:
4306 code = ICMP_UNREACH_PORT;
4307 break;
4308 default:
4309 return (-1);
4310 }
4311 break;
4312 case ICMP6_PACKET_TOO_BIG:
4313 type = ICMP_UNREACH;
4314 code = ICMP_UNREACH_NEEDFRAG;
4315 mtu -= 20;
4316 break;
4317 case ICMP6_TIME_EXCEEDED:
4318 type = ICMP_TIMXCEED;
4319 break;
4320 case ICMP6_PARAM_PROB:
4321 switch (code) {
4322 case ICMP6_PARAMPROB_HEADER:
4323 type = ICMP_PARAMPROB;
4324 code = ICMP_PARAMPROB_ERRATPTR;
4325 ptr = ntohl(icmp6->icmp6_pptr);
4326
4327 if (ptr == PTR_IP6(ip6_vfc))
4328 ; /* preserve */
4329 else if (ptr == PTR_IP6(ip6_vfc) + 1)
4330 ptr = PTR_IP(ip_tos);
4331 else if (ptr == PTR_IP6(ip6_plen) ||
4332 ptr == PTR_IP6(ip6_plen) + 1)
4333 ptr = PTR_IP(ip_len);
4334 else if (ptr == PTR_IP6(ip6_nxt))
4335 ptr = PTR_IP(ip_p);
4336 else if (ptr == PTR_IP6(ip6_hlim))
4337 ptr = PTR_IP(ip_ttl);
4338 else if (ptr >= PTR_IP6(ip6_src) &&
4339 ptr < PTR_IP6(ip6_dst))
4340 ptr = PTR_IP(ip_src);
4341 else if (ptr >= PTR_IP6(ip6_dst) &&
4342 ptr < (int32_t)sizeof(struct ip6_hdr))
4343 ptr = PTR_IP(ip_dst);
4344 else {
4345 return (-1);
4346 }
4347 break;
4348 case ICMP6_PARAMPROB_NEXTHEADER:
4349 type = ICMP_UNREACH;
4350 code = ICMP_UNREACH_PROTOCOL;
4351 break;
4352 default:
4353 return (-1);
4354 }
4355 break;
4356 default:
4357 return (-1);
4358 }
4359 icmp6->icmp6_type = type;
4360 icmp6->icmp6_code = code;
4361 /* aligns well with a icmpv4 nextmtu */
4362 icmp6->icmp6_mtu = htonl(mtu);
4363 /* icmpv4 pptr is a one most significant byte */
4364 if (ptr >= 0)
4365 icmp6->icmp6_pptr = htonl(ptr << 24);
4366 break;
4367
4368 case AF_INET6:
4369 icmp4 = arg;
4370 type = icmp4->icmp_type;
4371 code = icmp4->icmp_code;
4372 mtu = ntohs(icmp4->icmp_nextmtu);
4373
4374 switch (type) {
4375 case ICMP_ECHO:
4376 type = ICMP6_ECHO_REQUEST;
4377 break;
4378 case ICMP_ECHOREPLY:
4379 type = ICMP6_ECHO_REPLY;
4380 break;
4381 case ICMP_UNREACH:
4382 type = ICMP6_DST_UNREACH;
4383 switch (code) {
4384 case ICMP_UNREACH_NET:
4385 case ICMP_UNREACH_HOST:
4386 case ICMP_UNREACH_NET_UNKNOWN:
4387 case ICMP_UNREACH_HOST_UNKNOWN:
4388 case ICMP_UNREACH_ISOLATED:
4389 case ICMP_UNREACH_TOSNET:
4390 case ICMP_UNREACH_TOSHOST:
4391 code = ICMP6_DST_UNREACH_NOROUTE;
4392 break;
4393 case ICMP_UNREACH_PORT:
4394 code = ICMP6_DST_UNREACH_NOPORT;
4395 break;
4396 case ICMP_UNREACH_NET_PROHIB:
4397 case ICMP_UNREACH_HOST_PROHIB:
4398 case ICMP_UNREACH_FILTER_PROHIB:
4399 case ICMP_UNREACH_PRECEDENCE_CUTOFF:
4400 code = ICMP6_DST_UNREACH_ADMIN;
4401 break;
4402 case ICMP_UNREACH_PROTOCOL:
4403 type = ICMP6_PARAM_PROB;
4404 code = ICMP6_PARAMPROB_NEXTHEADER;
4405 ptr = offsetof(struct ip6_hdr, ip6_nxt);
4406 break;
4407 case ICMP_UNREACH_NEEDFRAG:
4408 type = ICMP6_PACKET_TOO_BIG;
4409 code = 0;
4410 mtu += 20;
4411 break;
4412 default:
4413 return (-1);
4414 }
4415 break;
4416 case ICMP_TIMXCEED:
4417 type = ICMP6_TIME_EXCEEDED;
4418 break;
4419 case ICMP_PARAMPROB:
4420 type = ICMP6_PARAM_PROB;
4421 switch (code) {
4422 case ICMP_PARAMPROB_ERRATPTR:
4423 code = ICMP6_PARAMPROB_HEADER;
4424 break;
4425 case ICMP_PARAMPROB_LENGTH:
4426 code = ICMP6_PARAMPROB_HEADER;
4427 break;
4428 default:
4429 return (-1);
4430 }
4431
4432 ptr = icmp4->icmp_pptr;
4433 if (ptr == 0 || ptr == PTR_IP(ip_tos))
4434 ; /* preserve */
4435 else if (ptr == PTR_IP(ip_len) ||
4436 ptr == PTR_IP(ip_len) + 1)
4437 ptr = PTR_IP6(ip6_plen);
4438 else if (ptr == PTR_IP(ip_ttl))
4439 ptr = PTR_IP6(ip6_hlim);
4440 else if (ptr == PTR_IP(ip_p))
4441 ptr = PTR_IP6(ip6_nxt);
4442 else if (ptr >= PTR_IP(ip_src) &&
4443 ptr < PTR_IP(ip_dst))
4444 ptr = PTR_IP6(ip6_src);
4445 else if (ptr >= PTR_IP(ip_dst) &&
4446 ptr < (int32_t)sizeof(struct ip))
4447 ptr = PTR_IP6(ip6_dst);
4448 else {
4449 return (-1);
4450 }
4451 break;
4452 default:
4453 return (-1);
4454 }
4455 icmp4->icmp_type = type;
4456 icmp4->icmp_code = code;
4457 icmp4->icmp_nextmtu = htons(mtu);
4458 if (ptr >= 0)
4459 icmp4->icmp_void = htonl(ptr);
4460 break;
4461 }
4462
4463 return (0);
4464}
4465
4466/* Note: frees pbuf if PF_NAT64 is returned */
4467static int
4468pf_nat64_ipv6(pbuf_t *pbuf, int off, struct pf_pdesc *pd)
4469{
4470 struct ip *ip4;
4471 struct mbuf *m;
4472
4473 /*
4474 * ip_input asserts for rcvif to be not NULL
4475 * That may not be true for two corner cases
4476 * 1. If for some reason a local app sends DNS
4477 * AAAA query to local host
4478 * 2. If IPv6 stack in kernel internally generates a
4479 * message destined for a synthesized IPv6 end-point.
4480 */
4481 if (pbuf->pb_ifp == NULL)
4482 return (PF_DROP);
4483
4484 ip4 = (struct ip *)pbuf_resize_segment(pbuf, 0, off, sizeof(*ip4));
4485 if (ip4 == NULL)
4486 return (PF_DROP);
4487
4488 ip4->ip_v = 4;
4489 ip4->ip_hl = 5;
4490 ip4->ip_tos = pd->tos & htonl(0x0ff00000);
4491 ip4->ip_len = htons(sizeof(*ip4) + (pd->tot_len - off));
4492 ip4->ip_id = 0;
4493 ip4->ip_off = htons(IP_DF);
4494 ip4->ip_ttl = pd->ttl;
4495 ip4->ip_p = pd->proto;
4496 ip4->ip_sum = 0;
4497 ip4->ip_src = pd->naddr.v4addr;
4498 ip4->ip_dst = pd->ndaddr.v4addr;
4499 ip4->ip_sum = pbuf_inet_cksum(pbuf, 0, 0, ip4->ip_hl << 2);
4500
4501 /* recalculate icmp checksums */
4502 if (pd->proto == IPPROTO_ICMP) {
4503 struct icmp *icmp;
4504 int hlen = sizeof(*ip4);
4505
4506 icmp = (struct icmp *)pbuf_contig_segment(pbuf, hlen,
4507 ICMP_MINLEN);
4508 if (icmp == NULL)
4509 return (PF_DROP);
4510
4511 icmp->icmp_cksum = 0;
4512 icmp->icmp_cksum = pbuf_inet_cksum(pbuf, 0, hlen,
4513 ntohs(ip4->ip_len) - hlen);
4514 }
4515
4516 if ((m = pbuf_to_mbuf(pbuf, TRUE)) != NULL)
4517 ip_input(m);
4518
4519 return (PF_NAT64);
4520}
4521
4522static int
4523pf_nat64_ipv4(pbuf_t *pbuf, int off, struct pf_pdesc *pd)
4524{
4525 struct ip6_hdr *ip6;
4526 struct mbuf *m;
4527
4528 if (pbuf->pb_ifp == NULL)
4529 return (PF_DROP);
4530
4531 ip6 = (struct ip6_hdr *)pbuf_resize_segment(pbuf, 0, off, sizeof(*ip6));
4532 if (ip6 == NULL)
4533 return (PF_DROP);
4534
4535 ip6->ip6_vfc = htonl((6 << 28) | (pd->tos << 20));
4536 ip6->ip6_plen = htons(pd->tot_len - off);
4537 ip6->ip6_nxt = pd->proto;
4538 ip6->ip6_hlim = pd->ttl;
4539 ip6->ip6_src = pd->naddr.v6addr;
4540 ip6->ip6_dst = pd->ndaddr.v6addr;
4541
4542 /* recalculate icmp6 checksums */
4543 if (pd->proto == IPPROTO_ICMPV6) {
4544 struct icmp6_hdr *icmp6;
4545 int hlen = sizeof(*ip6);
4546
4547 icmp6 = (struct icmp6_hdr *)pbuf_contig_segment(pbuf, hlen,
4548 sizeof(*icmp6));
4549 if (icmp6 == NULL)
4550 return (PF_DROP);
4551
4552 icmp6->icmp6_cksum = 0;
4553 icmp6->icmp6_cksum = pbuf_inet6_cksum(pbuf,
4554 IPPROTO_ICMPV6, hlen,
4555 ntohs(ip6->ip6_plen));
4556 } else if (pd->proto == IPPROTO_UDP) {
4557 struct udphdr *uh;
4558 int hlen = sizeof(*ip6);
4559
4560 uh = (struct udphdr *)pbuf_contig_segment(pbuf, hlen,
4561 sizeof(*uh));
4562 if (uh == NULL)
4563 return (PF_DROP);
4564
4565 if (uh->uh_sum == 0)
4566 uh->uh_sum = pbuf_inet6_cksum(pbuf, IPPROTO_UDP,
4567 hlen, ntohs(ip6->ip6_plen));
4568 }
4569
4570 if ((m = pbuf_to_mbuf(pbuf, TRUE)) != NULL)
4571 ip6_input(m);
4572
4573 return (PF_NAT64);
4574}
4575
4576static int
4577pf_test_rule(struct pf_rule **rm, struct pf_state **sm, int direction,
4578 struct pfi_kif *kif, pbuf_t *pbuf, int off, void *h,
4579 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
4580 struct ifqueue *ifq)
4581{
4582#pragma unused(h)
4583 struct pf_rule *nr = NULL;
4584 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
4585 sa_family_t af = pd->af;
4586 struct pf_rule *r, *a = NULL;
4587 struct pf_ruleset *ruleset = NULL;
4588 struct pf_src_node *nsn = NULL;
4589 struct tcphdr *th = pd->hdr.tcp;
4590 struct udphdr *uh = pd->hdr.udp;
4591 u_short reason;
4592 int rewrite = 0, hdrlen = 0;
4593 int tag = -1;
4594 unsigned int rtableid = IFSCOPE_NONE;
4595 int asd = 0;
4596 int match = 0;
4597 int state_icmp = 0;
4598 u_int16_t mss = tcp_mssdflt;
4599 u_int8_t icmptype = 0, icmpcode = 0;
4600
4601 struct pf_grev1_hdr *grev1 = pd->hdr.grev1;
4602 union pf_state_xport bxport, bdxport, nxport, sxport, dxport;
4603 struct pf_state_key psk;
4604
4605 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
4606
4607 if (direction == PF_IN && pf_check_congestion(ifq)) {
4608 REASON_SET(&reason, PFRES_CONGEST);
4609 return (PF_DROP);
4610 }
4611
4612 hdrlen = 0;
4613 sxport.spi = 0;
4614 dxport.spi = 0;
4615 nxport.spi = 0;
4616
4617 switch (pd->proto) {
4618 case IPPROTO_TCP:
4619 sxport.port = th->th_sport;
4620 dxport.port = th->th_dport;
4621 hdrlen = sizeof (*th);
4622 break;
4623 case IPPROTO_UDP:
4624 sxport.port = uh->uh_sport;
4625 dxport.port = uh->uh_dport;
4626 hdrlen = sizeof (*uh);
4627 break;
4628#if INET
4629 case IPPROTO_ICMP:
4630 if (pd->af != AF_INET)
4631 break;
4632 sxport.port = dxport.port = pd->hdr.icmp->icmp_id;
4633 hdrlen = ICMP_MINLEN;
4634 icmptype = pd->hdr.icmp->icmp_type;
4635 icmpcode = pd->hdr.icmp->icmp_code;
4636
4637 if (ICMP_ERRORTYPE(icmptype))
4638 state_icmp++;
4639 break;
4640#endif /* INET */
4641#if INET6
4642 case IPPROTO_ICMPV6:
4643 if (pd->af != AF_INET6)
4644 break;
4645 sxport.port = dxport.port = pd->hdr.icmp6->icmp6_id;
4646 hdrlen = sizeof (*pd->hdr.icmp6);
4647 icmptype = pd->hdr.icmp6->icmp6_type;
4648 icmpcode = pd->hdr.icmp6->icmp6_code;
4649
4650 if (ICMP6_ERRORTYPE(icmptype))
4651 state_icmp++;
4652 break;
4653#endif /* INET6 */
4654 case IPPROTO_GRE:
4655 if (pd->proto_variant == PF_GRE_PPTP_VARIANT) {
4656 sxport.call_id = dxport.call_id =
4657 pd->hdr.grev1->call_id;
4658 hdrlen = sizeof (*pd->hdr.grev1);
4659 }
4660 break;
4661 case IPPROTO_ESP:
4662 sxport.spi = 0;
4663 dxport.spi = pd->hdr.esp->spi;
4664 hdrlen = sizeof (*pd->hdr.esp);
4665 break;
4666 }
4667
4668 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
4669
4670 bxport = sxport;
4671 bdxport = dxport;
4672
4673 if (direction == PF_OUT)
4674 nxport = sxport;
4675 else
4676 nxport = dxport;
4677
4678 /* check packet for BINAT/NAT/RDR */
4679 if ((nr = pf_get_translation_aux(pd, pbuf, off, direction, kif, &nsn,
4680 saddr, &sxport, daddr, &dxport, &nxport
4681 )) != NULL) {
4682 int ua;
4683 u_int16_t dport;
4684
4685 if (pd->af != pd->naf)
4686 ua = 0;
4687 else
4688 ua = 1;
4689
4690 PF_ACPY(&pd->baddr, saddr, af);
4691 PF_ACPY(&pd->bdaddr, daddr, af);
4692
4693 switch (pd->proto) {
4694 case IPPROTO_TCP:
4695 if (pd->af != pd->naf ||
4696 PF_ANEQ(saddr, &pd->naddr, pd->af)) {
4697 pf_change_ap(direction, pd->mp, saddr,
4698 &th->th_sport, pd->ip_sum, &th->th_sum,
4699 &pd->naddr, nxport.port, 0, af,
4700 pd->naf, ua);
4701 sxport.port = th->th_sport;
4702 }
4703
4704 if (pd->af != pd->naf ||
4705 PF_ANEQ(daddr, &pd->ndaddr, pd->af) ||
4706 (nr && (nr->action == PF_RDR) &&
4707 (th->th_dport != nxport.port))) {
4708 if (nr && nr->action == PF_RDR)
4709 dport = nxport.port;
4710 else
4711 dport = th->th_dport;
4712 pf_change_ap(direction, pd->mp, daddr,
4713 &th->th_dport, pd->ip_sum,
4714 &th->th_sum, &pd->ndaddr,
4715 dport, 0, af, pd->naf, ua);
4716 dxport.port = th->th_dport;
4717 }
4718 rewrite++;
4719 break;
4720
4721 case IPPROTO_UDP:
4722 if (pd->af != pd->naf ||
4723 PF_ANEQ(saddr, &pd->naddr, pd->af)) {
4724 pf_change_ap(direction, pd->mp, saddr,
4725 &uh->uh_sport, pd->ip_sum,
4726 &uh->uh_sum, &pd->naddr,
4727 nxport.port, 1, af, pd->naf, ua);
4728 sxport.port = uh->uh_sport;
4729 }
4730
4731 if (pd->af != pd->naf ||
4732 PF_ANEQ(daddr, &pd->ndaddr, pd->af) ||
4733 (nr && (nr->action == PF_RDR) &&
4734 (uh->uh_dport != nxport.port))) {
4735 if (nr && nr->action == PF_RDR)
4736 dport = nxport.port;
4737 else
4738 dport = uh->uh_dport;
4739 pf_change_ap(direction, pd->mp, daddr,
4740 &uh->uh_dport, pd->ip_sum,
4741 &uh->uh_sum, &pd->ndaddr,
4742 dport, 0, af, pd->naf, ua);
4743 dxport.port = uh->uh_dport;
4744 }
4745 rewrite++;
4746 break;
4747#if INET
4748 case IPPROTO_ICMP:
4749 if (pd->af != AF_INET)
4750 break;
4751 /*
4752 * TODO:
4753 * pd->af != pd->naf not handled yet here and would be
4754 * needed for NAT46 needed to support XLAT.
4755 * Will cross the bridge when it comes.
4756 */
4757 if (PF_ANEQ(saddr, &pd->naddr, pd->af)) {
4758 pf_change_a(&saddr->v4addr.s_addr, pd->ip_sum,
4759 pd->naddr.v4addr.s_addr, 0);
4760 pd->hdr.icmp->icmp_cksum = pf_cksum_fixup(
4761 pd->hdr.icmp->icmp_cksum, sxport.port,
4762 nxport.port, 0);
4763 pd->hdr.icmp->icmp_id = nxport.port;
4764 }
4765
4766 if (PF_ANEQ(daddr, &pd->ndaddr, pd->af)) {
4767 pf_change_a(&daddr->v4addr.s_addr, pd->ip_sum,
4768 pd->ndaddr.v4addr.s_addr, 0);
4769 }
4770 ++rewrite;
4771 break;
4772#endif /* INET */
4773#if INET6
4774 case IPPROTO_ICMPV6:
4775 if (pd->af != AF_INET6)
4776 break;
4777
4778 if (pd->af != pd->naf ||
4779 PF_ANEQ(saddr, &pd->naddr, pd->af)) {
4780 pf_change_addr(saddr,
4781 &pd->hdr.icmp6->icmp6_cksum,
4782 &pd->naddr, 0, pd->af, pd->naf);
4783 }
4784
4785 if (pd->af != pd->naf ||
4786 PF_ANEQ(daddr, &pd->ndaddr, pd->af)) {
4787 pf_change_addr(daddr,
4788 &pd->hdr.icmp6->icmp6_cksum,
4789 &pd->ndaddr, 0, pd->af, pd->naf);
4790 }
4791
4792 if (pd->af != pd->naf) {
4793 if (pf_translate_icmp_af(AF_INET,
4794 pd->hdr.icmp6))
4795 return (PF_DROP);
4796 pd->proto = IPPROTO_ICMP;
4797 }
4798 rewrite++;
4799 break;
4800#endif /* INET */
4801 case IPPROTO_GRE:
4802 if ((direction == PF_IN) &&
4803 (pd->proto_variant == PF_GRE_PPTP_VARIANT))
4804 grev1->call_id = nxport.call_id;
4805
4806 switch (pd->af) {
4807#if INET
4808 case AF_INET:
4809 if (PF_ANEQ(saddr, &pd->naddr, pd->af)) {
4810 pf_change_a(&saddr->v4addr.s_addr,
4811 pd->ip_sum,
4812 pd->naddr.v4addr.s_addr, 0);
4813 }
4814 if (PF_ANEQ(daddr, &pd->ndaddr, pd->af)) {
4815 pf_change_a(&daddr->v4addr.s_addr,
4816 pd->ip_sum,
4817 pd->ndaddr.v4addr.s_addr, 0);
4818 }
4819 break;
4820#endif /* INET */
4821#if INET6
4822 case AF_INET6:
4823 if (PF_ANEQ(saddr, &pd->naddr, pd->af))
4824 PF_ACPY(saddr, &pd->naddr, AF_INET6);
4825 if (PF_ANEQ(daddr, &pd->ndaddr, pd->af))
4826 PF_ACPY(daddr, &pd->ndaddr, AF_INET6);
4827 break;
4828#endif /* INET6 */
4829 }
4830 ++rewrite;
4831 break;
4832 case IPPROTO_ESP:
4833 if (direction == PF_OUT)
4834 bxport.spi = 0;
4835
4836 switch (pd->af) {
4837#if INET
4838 case AF_INET:
4839 if (PF_ANEQ(saddr, &pd->naddr, pd->af)) {
4840 pf_change_a(&saddr->v4addr.s_addr,
4841 pd->ip_sum, pd->naddr.v4addr.s_addr, 0);
4842 }
4843 if (PF_ANEQ(daddr, &pd->ndaddr, pd->af)) {
4844 pf_change_a(&daddr->v4addr.s_addr,
4845 pd->ip_sum,
4846 pd->ndaddr.v4addr.s_addr, 0);
4847 }
4848 break;
4849#endif /* INET */
4850#if INET6
4851 case AF_INET6:
4852 if (PF_ANEQ(saddr, &pd->naddr, pd->af))
4853 PF_ACPY(saddr, &pd->naddr, AF_INET6);
4854 if (PF_ANEQ(daddr, &pd->ndaddr, pd->af))
4855 PF_ACPY(daddr, &pd->ndaddr, AF_INET6);
4856 break;
4857#endif /* INET6 */
4858 }
4859 break;
4860 default:
4861 switch (pd->af) {
4862#if INET
4863 case AF_INET:
4864 if ((pd->naf != AF_INET) ||
4865 (PF_ANEQ(saddr, &pd->naddr, pd->af))) {
4866 pf_change_addr(saddr, pd->ip_sum,
4867 &pd->naddr, 0, af, pd->naf);
4868 }
4869
4870 if ((pd->naf != AF_INET) ||
4871 (PF_ANEQ(daddr, &pd->ndaddr, pd->af))) {
4872 pf_change_addr(daddr, pd->ip_sum,
4873 &pd->ndaddr, 0, af, pd->naf);
4874 }
4875 break;
4876#endif /* INET */
4877#if INET6
4878 case AF_INET6:
4879 if (PF_ANEQ(saddr, &pd->naddr, pd->af))
4880 PF_ACPY(saddr, &pd->naddr, af);
4881 if (PF_ANEQ(daddr, &pd->ndaddr, pd->af))
4882 PF_ACPY(daddr, &pd->ndaddr, af);
4883 break;
4884#endif /* INET */
4885 }
4886 break;
4887 }
4888
4889 if (nr->natpass)
4890 r = NULL;
4891 pd->nat_rule = nr;
4892 pd->af = pd->naf;
4893 } else {
4894 }
4895
4896 if (nr && nr->tag > 0)
4897 tag = nr->tag;
4898
4899 while (r != NULL) {
4900 r->evaluations++;
4901 if (pfi_kif_match(r->kif, kif) == r->ifnot)
4902 r = r->skip[PF_SKIP_IFP].ptr;
4903 else if (r->direction && r->direction != direction)
4904 r = r->skip[PF_SKIP_DIR].ptr;
4905 else if (r->af && r->af != pd->af)
4906 r = r->skip[PF_SKIP_AF].ptr;
4907 else if (r->proto && r->proto != pd->proto)
4908 r = r->skip[PF_SKIP_PROTO].ptr;
4909 else if (PF_MISMATCHAW(&r->src.addr, saddr, pd->af,
4910 r->src.neg, kif))
4911 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
4912 /* tcp/udp only. port_op always 0 in other cases */
4913 else if (r->proto == pd->proto &&
4914 (r->proto == IPPROTO_TCP || r->proto == IPPROTO_UDP) &&
4915 r->src.xport.range.op &&
4916 !pf_match_port(r->src.xport.range.op,
4917 r->src.xport.range.port[0], r->src.xport.range.port[1],
4918 th->th_sport))
4919 r = r->skip[PF_SKIP_SRC_PORT].ptr;
4920 else if (PF_MISMATCHAW(&r->dst.addr, daddr, pd->af,
4921 r->dst.neg, NULL))
4922 r = r->skip[PF_SKIP_DST_ADDR].ptr;
4923 /* tcp/udp only. port_op always 0 in other cases */
4924 else if (r->proto == pd->proto &&
4925 (r->proto == IPPROTO_TCP || r->proto == IPPROTO_UDP) &&
4926 r->dst.xport.range.op &&
4927 !pf_match_port(r->dst.xport.range.op,
4928 r->dst.xport.range.port[0], r->dst.xport.range.port[1],
4929 th->th_dport))
4930 r = r->skip[PF_SKIP_DST_PORT].ptr;
4931 /* icmp only. type always 0 in other cases */
4932 else if (r->type && r->type != icmptype + 1)
4933 r = TAILQ_NEXT(r, entries);
4934 /* icmp only. type always 0 in other cases */
4935 else if (r->code && r->code != icmpcode + 1)
4936 r = TAILQ_NEXT(r, entries);
4937 else if ((r->rule_flag & PFRULE_TOS) && r->tos &&
4938 !(r->tos & pd->tos))
4939 r = TAILQ_NEXT(r, entries);
4940 else if ((r->rule_flag & PFRULE_DSCP) && r->tos &&
4941 !(r->tos & (pd->tos & DSCP_MASK)))
4942 r = TAILQ_NEXT(r, entries);
4943 else if ((r->rule_flag & PFRULE_SC) && r->tos &&
4944 ((r->tos & SCIDX_MASK) != pd->sc))
4945 r = TAILQ_NEXT(r, entries);
4946 else if (r->rule_flag & PFRULE_FRAGMENT)
4947 r = TAILQ_NEXT(r, entries);
4948 else if (pd->proto == IPPROTO_TCP &&
4949 (r->flagset & th->th_flags) != r->flags)
4950 r = TAILQ_NEXT(r, entries);
4951 /* tcp/udp only. uid.op always 0 in other cases */
4952 else if (r->uid.op && (pd->lookup.done || ((void)(pd->lookup.done =
4953 pf_socket_lookup(direction, pd)), 1)) &&
4954 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
4955 pd->lookup.uid))
4956 r = TAILQ_NEXT(r, entries);
4957 /* tcp/udp only. gid.op always 0 in other cases */
4958 else if (r->gid.op && (pd->lookup.done || ((void)(pd->lookup.done =
4959 pf_socket_lookup(direction, pd)), 1)) &&
4960 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
4961 pd->lookup.gid))
4962 r = TAILQ_NEXT(r, entries);
4963 else if (r->prob && r->prob <= (RandomULong() % (UINT_MAX - 1) + 1))
4964 r = TAILQ_NEXT(r, entries);
4965 else if (r->match_tag && !pf_match_tag(r, pd->pf_mtag, &tag))
4966 r = TAILQ_NEXT(r, entries);
4967 else if (r->os_fingerprint != PF_OSFP_ANY &&
4968 (pd->proto != IPPROTO_TCP || !pf_osfp_match(
4969 pf_osfp_fingerprint(pd, pbuf, off, th),
4970 r->os_fingerprint)))
4971 r = TAILQ_NEXT(r, entries);
4972 else {
4973 if (r->tag)
4974 tag = r->tag;
4975 if (PF_RTABLEID_IS_VALID(r->rtableid))
4976 rtableid = r->rtableid;
4977 if (r->anchor == NULL) {
4978 match = 1;
4979 *rm = r;
4980 *am = a;
4981 *rsm = ruleset;
4982 if ((*rm)->quick)
4983 break;
4984 r = TAILQ_NEXT(r, entries);
4985 } else
4986 pf_step_into_anchor(&asd, &ruleset,
4987 PF_RULESET_FILTER, &r, &a, &match);
4988 }
4989 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
4990 PF_RULESET_FILTER, &r, &a, &match))
4991 break;
4992 }
4993 r = *rm;
4994 a = *am;
4995 ruleset = *rsm;
4996
4997 REASON_SET(&reason, PFRES_MATCH);
4998
4999 if (r->log || (nr != NULL && nr->log)) {
5000 if (rewrite > 0) {
5001 if (rewrite < off + hdrlen)
5002 rewrite = off + hdrlen;
5003
5004 if (pf_lazy_makewritable(pd, pbuf, rewrite) == NULL) {
5005 REASON_SET(&reason, PFRES_MEMORY);
5006 return (PF_DROP);
5007 }
5008
5009 pbuf_copy_back(pbuf, off, hdrlen, pd->hdr.any);
5010 }
5011 PFLOG_PACKET(kif, h, pbuf, pd->af, direction, reason,
5012 r->log ? r : nr, a, ruleset, pd);
5013 }
5014
5015 if ((r->action == PF_DROP) &&
5016 ((r->rule_flag & PFRULE_RETURNRST) ||
5017 (r->rule_flag & PFRULE_RETURNICMP) ||
5018 (r->rule_flag & PFRULE_RETURN))) {
5019 /* undo NAT changes, if they have taken place */
5020 /* XXX For NAT64 we are not reverting the changes */
5021 if (nr != NULL && nr->action != PF_NAT64) {
5022 if (direction == PF_OUT) {
5023 pd->af = af;
5024 switch (pd->proto) {
5025 case IPPROTO_TCP:
5026 pf_change_ap(direction, pd->mp, saddr,
5027 &th->th_sport, pd->ip_sum,
5028 &th->th_sum, &pd->baddr,
5029 bxport.port, 0, af, pd->af, 1);
5030 sxport.port = th->th_sport;
5031 rewrite++;
5032 break;
5033 case IPPROTO_UDP:
5034 pf_change_ap(direction, pd->mp, saddr,
5035 &pd->hdr.udp->uh_sport, pd->ip_sum,
5036 &pd->hdr.udp->uh_sum, &pd->baddr,
5037 bxport.port, 1, af, pd->af, 1);
5038 sxport.port = pd->hdr.udp->uh_sport;
5039 rewrite++;
5040 break;
5041 case IPPROTO_ICMP:
5042#if INET6
5043 case IPPROTO_ICMPV6:
5044#endif
5045 /* nothing! */
5046 break;
5047 case IPPROTO_GRE:
5048 PF_ACPY(&pd->baddr, saddr, af);
5049 ++rewrite;
5050 switch (af) {
5051#if INET
5052 case AF_INET:
5053 pf_change_a(&saddr->v4addr.s_addr,
5054 pd->ip_sum,
5055 pd->baddr.v4addr.s_addr, 0);
5056 break;
5057#endif /* INET */
5058#if INET6
5059 case AF_INET6:
5060 PF_ACPY(saddr, &pd->baddr,
5061 AF_INET6);
5062 break;
5063#endif /* INET6 */
5064 }
5065 break;
5066 case IPPROTO_ESP:
5067 PF_ACPY(&pd->baddr, saddr, af);
5068 switch (af) {
5069#if INET
5070 case AF_INET:
5071 pf_change_a(&saddr->v4addr.s_addr,
5072 pd->ip_sum,
5073 pd->baddr.v4addr.s_addr, 0);
5074 break;
5075#endif /* INET */
5076#if INET6
5077 case AF_INET6:
5078 PF_ACPY(saddr, &pd->baddr,
5079 AF_INET6);
5080 break;
5081#endif /* INET6 */
5082 }
5083 break;
5084 default:
5085 switch (af) {
5086 case AF_INET:
5087 pf_change_a(&saddr->v4addr.s_addr,
5088 pd->ip_sum,
5089 pd->baddr.v4addr.s_addr, 0);
5090 break;
5091 case AF_INET6:
5092 PF_ACPY(saddr, &pd->baddr, af);
5093 break;
5094 }
5095 }
5096 } else {
5097 switch (pd->proto) {
5098 case IPPROTO_TCP:
5099 pf_change_ap(direction, pd->mp, daddr,
5100 &th->th_dport, pd->ip_sum,
5101 &th->th_sum, &pd->bdaddr,
5102 bdxport.port, 0, af, pd->af, 1);
5103 dxport.port = th->th_dport;
5104 rewrite++;
5105 break;
5106 case IPPROTO_UDP:
5107 pf_change_ap(direction, pd->mp, daddr,
5108 &pd->hdr.udp->uh_dport, pd->ip_sum,
5109 &pd->hdr.udp->uh_sum, &pd->bdaddr,
5110 bdxport.port, 1, af, pd->af, 1);
5111 dxport.port = pd->hdr.udp->uh_dport;
5112 rewrite++;
5113 break;
5114 case IPPROTO_ICMP:
5115#if INET6
5116 case IPPROTO_ICMPV6:
5117#endif
5118 /* nothing! */
5119 break;
5120 case IPPROTO_GRE:
5121 if (pd->proto_variant ==
5122 PF_GRE_PPTP_VARIANT)
5123 grev1->call_id =
5124 bdxport.call_id;
5125 ++rewrite;
5126 switch (af) {
5127#if INET
5128 case AF_INET:
5129 pf_change_a(&daddr->v4addr.s_addr,
5130 pd->ip_sum,
5131 pd->bdaddr.v4addr.s_addr, 0);
5132 break;
5133#endif /* INET */
5134#if INET6
5135 case AF_INET6:
5136 PF_ACPY(daddr, &pd->bdaddr,
5137 AF_INET6);
5138 break;
5139#endif /* INET6 */
5140 }
5141 break;
5142 case IPPROTO_ESP:
5143 switch (af) {
5144#if INET
5145 case AF_INET:
5146 pf_change_a(&daddr->v4addr.s_addr,
5147 pd->ip_sum,
5148 pd->bdaddr.v4addr.s_addr, 0);
5149 break;
5150#endif /* INET */
5151#if INET6
5152 case AF_INET6:
5153 PF_ACPY(daddr, &pd->bdaddr,
5154 AF_INET6);
5155 break;
5156#endif /* INET6 */
5157 }
5158 break;
5159 default:
5160 switch (af) {
5161 case AF_INET:
5162 pf_change_a(&daddr->v4addr.s_addr,
5163 pd->ip_sum,
5164 pd->bdaddr.v4addr.s_addr, 0);
5165 break;
5166#if INET6
5167 case AF_INET6:
5168 PF_ACPY(daddr, &pd->bdaddr, af);
5169 break;
5170#endif /* INET6 */
5171 }
5172 }
5173 }
5174 }
5175 if (pd->proto == IPPROTO_TCP &&
5176 ((r->rule_flag & PFRULE_RETURNRST) ||
5177 (r->rule_flag & PFRULE_RETURN)) &&
5178 !(th->th_flags & TH_RST)) {
5179 u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
5180 int len = 0;
5181 struct ip *h4;
5182#if INET6
5183 struct ip6_hdr *h6;
5184#endif /* INET6 */
5185
5186 switch (pd->af) {
5187 case AF_INET:
5188 h4 = pbuf->pb_data;
5189 len = ntohs(h4->ip_len) - off;
5190 break;
5191#if INET6
5192 case AF_INET6:
5193 h6 = pbuf->pb_data;
5194 len = ntohs(h6->ip6_plen) -
5195 (off - sizeof (*h6));
5196 break;
5197#endif /* INET6 */
5198 }
5199
5200 if (pf_check_proto_cksum(pbuf, off, len, IPPROTO_TCP,
5201 pd->af))
5202 REASON_SET(&reason, PFRES_PROTCKSUM);
5203 else {
5204 if (th->th_flags & TH_SYN)
5205 ack++;
5206 if (th->th_flags & TH_FIN)
5207 ack++;
5208 pf_send_tcp(r, pd->af, pd->dst,
5209 pd->src, th->th_dport, th->th_sport,
5210 ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0,
5211 r->return_ttl, 1, 0, pd->eh, kif->pfik_ifp);
5212 }
5213 } else if (pd->proto != IPPROTO_ICMP && pd->af == AF_INET &&
5214 pd->proto != IPPROTO_ESP && pd->proto != IPPROTO_AH &&
5215 r->return_icmp)
5216 pf_send_icmp(pbuf, r->return_icmp >> 8,
5217 r->return_icmp & 255, pd->af, r);
5218 else if (pd->proto != IPPROTO_ICMPV6 && af == AF_INET6 &&
5219 pd->proto != IPPROTO_ESP && pd->proto != IPPROTO_AH &&
5220 r->return_icmp6)
5221 pf_send_icmp(pbuf, r->return_icmp6 >> 8,
5222 r->return_icmp6 & 255, pd->af, r);
5223 }
5224
5225 if (r->action == PF_DROP) {
5226 return (PF_DROP);
5227 }
5228
5229 /* prepare state key, for flowhash and/or the state (if created) */
5230 bzero(&psk, sizeof (psk));
5231 psk.proto = pd->proto;
5232 psk.direction = direction;
5233 if (pd->proto == IPPROTO_UDP) {
5234 if (ntohs(pd->hdr.udp->uh_sport) == PF_IKE_PORT &&
5235 ntohs(pd->hdr.udp->uh_dport) == PF_IKE_PORT) {
5236 psk.proto_variant = PF_EXTFILTER_APD;
5237 } else {
5238 psk.proto_variant = nr ? nr->extfilter : r->extfilter;
5239 if (psk.proto_variant < PF_EXTFILTER_APD)
5240 psk.proto_variant = PF_EXTFILTER_APD;
5241 }
5242 } else if (pd->proto == IPPROTO_GRE) {
5243 psk.proto_variant = pd->proto_variant;
5244 }
5245 if (direction == PF_OUT) {
5246 psk.af_gwy = af;
5247 PF_ACPY(&psk.gwy.addr, saddr, af);
5248 PF_ACPY(&psk.ext_gwy.addr, daddr, af);
5249 switch (pd->proto) {
5250 case IPPROTO_ESP:
5251 psk.gwy.xport.spi = 0;
5252 psk.ext_gwy.xport.spi = pd->hdr.esp->spi;
5253 break;
5254 case IPPROTO_ICMP:
5255#if INET6
5256 case IPPROTO_ICMPV6:
5257#endif
5258 /*
5259 * NAT64 requires protocol translation between ICMPv4
5260 * and ICMPv6. TCP and UDP do not require protocol
5261 * translation. To avoid adding complexity just to
5262 * handle ICMP(v4addr/v6addr), we always lookup for
5263 * proto = IPPROTO_ICMP on both LAN and WAN side
5264 */
5265 psk.proto = IPPROTO_ICMP;
5266 psk.gwy.xport.port = nxport.port;
5267 psk.ext_gwy.xport.spi = 0;
5268 break;
5269 default:
5270 psk.gwy.xport = sxport;
5271 psk.ext_gwy.xport = dxport;
5272 break;
5273 }
5274 psk.af_lan = af;
5275 if (nr != NULL) {
5276 PF_ACPY(&psk.lan.addr, &pd->baddr, af);
5277 psk.lan.xport = bxport;
5278 PF_ACPY(&psk.ext_lan.addr, &pd->bdaddr, af);
5279 psk.ext_lan.xport = bdxport;
5280 } else {
5281 PF_ACPY(&psk.lan.addr, &psk.gwy.addr, af);
5282 psk.lan.xport = psk.gwy.xport;
5283 PF_ACPY(&psk.ext_lan.addr, &psk.ext_gwy.addr, af);
5284 psk.ext_lan.xport = psk.ext_gwy.xport;
5285 }
5286 } else {
5287 psk.af_lan = af;
5288 if (nr && nr->action == PF_NAT64) {
5289 PF_ACPY(&psk.lan.addr, &pd->baddr, af);
5290 PF_ACPY(&psk.ext_lan.addr, &pd->bdaddr, af);
5291 } else {
5292 PF_ACPY(&psk.lan.addr, daddr, af);
5293 PF_ACPY(&psk.ext_lan.addr, saddr, af);
5294 }
5295 switch (pd->proto) {
5296 case IPPROTO_ICMP:
5297#if INET6
5298 case IPPROTO_ICMPV6:
5299#endif
5300 /*
5301 * NAT64 requires protocol translation between ICMPv4
5302 * and ICMPv6. TCP and UDP do not require protocol
5303 * translation. To avoid adding complexity just to
5304 * handle ICMP(v4addr/v6addr), we always lookup for
5305 * proto = IPPROTO_ICMP on both LAN and WAN side
5306 */
5307 psk.proto = IPPROTO_ICMP;
5308 if (nr && nr->action == PF_NAT64) {
5309 psk.lan.xport = bxport;
5310 psk.ext_lan.xport = bxport;
5311 } else {
5312 psk.lan.xport = nxport;
5313 psk.ext_lan.xport.spi = 0;
5314 }
5315 break;
5316 case IPPROTO_ESP:
5317 psk.ext_lan.xport.spi = 0;
5318 psk.lan.xport.spi = pd->hdr.esp->spi;
5319 break;
5320 default:
5321 if (nr != NULL) {
5322 if (nr->action == PF_NAT64) {
5323 psk.lan.xport = bxport;
5324 psk.ext_lan.xport = bdxport;
5325 } else {
5326 psk.lan.xport = dxport;
5327 psk.ext_lan.xport = sxport;
5328 }
5329 } else {
5330 psk.lan.xport = dxport;
5331 psk.ext_lan.xport = sxport;
5332 }
5333 break;
5334 }
5335 psk.af_gwy = pd->naf;
5336 if (nr != NULL) {
5337 if (nr->action == PF_NAT64) {
5338 PF_ACPY(&psk.gwy.addr, &pd->naddr, pd->naf);
5339 PF_ACPY(&psk.ext_gwy.addr, &pd->ndaddr,
5340 pd->naf);
5341 if ((pd->proto == IPPROTO_ICMPV6) ||
5342 (pd->proto == IPPROTO_ICMP)) {
5343 psk.gwy.xport = nxport;
5344 psk.ext_gwy.xport = nxport;
5345 } else {
5346 psk.gwy.xport = sxport;
5347 psk.ext_gwy.xport = dxport;
5348 }
5349 } else {
5350 PF_ACPY(&psk.gwy.addr, &pd->bdaddr, af);
5351 psk.gwy.xport = bdxport;
5352 PF_ACPY(&psk.ext_gwy.addr, saddr, af);
5353 psk.ext_gwy.xport = sxport;
5354 }
5355 } else {
5356 PF_ACPY(&psk.gwy.addr, &psk.lan.addr, af);
5357 psk.gwy.xport = psk.lan.xport;
5358 PF_ACPY(&psk.ext_gwy.addr, &psk.ext_lan.addr, af);
5359 psk.ext_gwy.xport = psk.ext_lan.xport;
5360 }
5361 }
5362 if (pd->pktflags & PKTF_FLOW_ID) {
5363 /* flow hash was already computed outside of PF */
5364 psk.flowsrc = pd->flowsrc;
5365 psk.flowhash = pd->flowhash;
5366 } else {
5367 /* compute flow hash and store it in state key */
5368 psk.flowsrc = FLOWSRC_PF;
5369 psk.flowhash = pf_calc_state_key_flowhash(&psk);
5370 pd->flowsrc = psk.flowsrc;
5371 pd->flowhash = psk.flowhash;
5372 pd->pktflags |= PKTF_FLOW_ID;
5373 pd->pktflags &= ~PKTF_FLOW_ADV;
5374 }
5375
5376 if (pf_tag_packet(pbuf, pd->pf_mtag, tag, rtableid, pd)) {
5377 REASON_SET(&reason, PFRES_MEMORY);
5378 return (PF_DROP);
5379 }
5380
5381 if (!state_icmp && (r->keep_state || nr != NULL ||
5382 (pd->flags & PFDESC_TCP_NORM))) {
5383 /* create new state */
5384 struct pf_state *s = NULL;
5385 struct pf_state_key *sk = NULL;
5386 struct pf_src_node *sn = NULL;
5387 struct pf_ike_hdr ike;
5388
5389 if (pd->proto == IPPROTO_UDP) {
5390 size_t plen = pbuf->pb_packet_len - off - sizeof(*uh);
5391
5392 if (ntohs(uh->uh_sport) == PF_IKE_PORT &&
5393 ntohs(uh->uh_dport) == PF_IKE_PORT &&
5394 plen >= PF_IKE_PACKET_MINSIZE) {
5395 if (plen > PF_IKE_PACKET_MINSIZE)
5396 plen = PF_IKE_PACKET_MINSIZE;
5397 pbuf_copy_data(pbuf, off + sizeof (*uh), plen,
5398 &ike);
5399 }
5400 }
5401
5402 if (nr != NULL && pd->proto == IPPROTO_ESP &&
5403 direction == PF_OUT) {
5404 struct pf_state_key_cmp sk0;
5405 struct pf_state *s0;
5406
5407 /*
5408 * <jhw@apple.com>
5409 * This squelches state creation if the external
5410 * address matches an existing incomplete state with a
5411 * different internal address. Only one 'blocking'
5412 * partial state is allowed for each external address.
5413 */
5414 memset(&sk0, 0, sizeof (sk0));
5415 sk0.af_gwy = pd->af;
5416 sk0.proto = IPPROTO_ESP;
5417 PF_ACPY(&sk0.gwy.addr, saddr, sk0.af_gwy);
5418 PF_ACPY(&sk0.ext_gwy.addr, daddr, sk0.af_gwy);
5419 s0 = pf_find_state(kif, &sk0, PF_IN);
5420
5421 if (s0 && PF_ANEQ(&s0->state_key->lan.addr,
5422 pd->src, pd->af)) {
5423 nsn = 0;
5424 goto cleanup;
5425 }
5426 }
5427
5428 /* check maximums */
5429 if (r->max_states && (r->states >= r->max_states)) {
5430 pf_status.lcounters[LCNT_STATES]++;
5431 REASON_SET(&reason, PFRES_MAXSTATES);
5432 goto cleanup;
5433 }
5434 /* src node for filter rule */
5435 if ((r->rule_flag & PFRULE_SRCTRACK ||
5436 r->rpool.opts & PF_POOL_STICKYADDR) &&
5437 pf_insert_src_node(&sn, r, saddr, af) != 0) {
5438 REASON_SET(&reason, PFRES_SRCLIMIT);
5439 goto cleanup;
5440 }
5441 /* src node for translation rule */
5442 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
5443 ((direction == PF_OUT &&
5444 nr->action != PF_RDR &&
5445 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
5446 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) {
5447 REASON_SET(&reason, PFRES_SRCLIMIT);
5448 goto cleanup;
5449 }
5450 s = pool_get(&pf_state_pl, PR_WAITOK);
5451 if (s == NULL) {
5452 REASON_SET(&reason, PFRES_MEMORY);
5453cleanup:
5454 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
5455 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
5456 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
5457 pf_status.src_nodes--;
5458 pool_put(&pf_src_tree_pl, sn);
5459 }
5460 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
5461 nsn->expire == 0) {
5462 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
5463 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
5464 pf_status.src_nodes--;
5465 pool_put(&pf_src_tree_pl, nsn);
5466 }
5467 if (sk != NULL) {
5468 if (sk->app_state)
5469 pool_put(&pf_app_state_pl,
5470 sk->app_state);
5471 pool_put(&pf_state_key_pl, sk);
5472 }
5473 return (PF_DROP);
5474 }
5475 bzero(s, sizeof (*s));
5476 TAILQ_INIT(&s->unlink_hooks);
5477 s->rule.ptr = r;
5478 s->nat_rule.ptr = nr;
5479 s->anchor.ptr = a;
5480 STATE_INC_COUNTERS(s);
5481 s->allow_opts = r->allow_opts;
5482 s->log = r->log & PF_LOG_ALL;
5483 if (nr != NULL)
5484 s->log |= nr->log & PF_LOG_ALL;
5485 switch (pd->proto) {
5486 case IPPROTO_TCP:
5487 s->src.seqlo = ntohl(th->th_seq);
5488 s->src.seqhi = s->src.seqlo + pd->p_len + 1;
5489 if ((th->th_flags & (TH_SYN|TH_ACK)) ==
5490 TH_SYN && r->keep_state == PF_STATE_MODULATE) {
5491 /* Generate sequence number modulator */
5492 if ((s->src.seqdiff = pf_tcp_iss(pd) -
5493 s->src.seqlo) == 0)
5494 s->src.seqdiff = 1;
5495 pf_change_a(&th->th_seq, &th->th_sum,
5496 htonl(s->src.seqlo + s->src.seqdiff), 0);
5497 rewrite = off + sizeof (*th);
5498 } else
5499 s->src.seqdiff = 0;
5500 if (th->th_flags & TH_SYN) {
5501 s->src.seqhi++;
5502 s->src.wscale = pf_get_wscale(pbuf, off,
5503 th->th_off, af);
5504 }
5505 s->src.max_win = MAX(ntohs(th->th_win), 1);
5506 if (s->src.wscale & PF_WSCALE_MASK) {
5507 /* Remove scale factor from initial window */
5508 int win = s->src.max_win;
5509 win += 1 << (s->src.wscale & PF_WSCALE_MASK);
5510 s->src.max_win = (win - 1) >>
5511 (s->src.wscale & PF_WSCALE_MASK);
5512 }
5513 if (th->th_flags & TH_FIN)
5514 s->src.seqhi++;
5515 s->dst.seqhi = 1;
5516 s->dst.max_win = 1;
5517 s->src.state = TCPS_SYN_SENT;
5518 s->dst.state = TCPS_CLOSED;
5519 s->timeout = PFTM_TCP_FIRST_PACKET;
5520 break;
5521 case IPPROTO_UDP:
5522 s->src.state = PFUDPS_SINGLE;
5523 s->dst.state = PFUDPS_NO_TRAFFIC;
5524 s->timeout = PFTM_UDP_FIRST_PACKET;
5525 break;
5526 case IPPROTO_ICMP:
5527#if INET6
5528 case IPPROTO_ICMPV6:
5529#endif
5530 s->timeout = PFTM_ICMP_FIRST_PACKET;
5531 break;
5532 case IPPROTO_GRE:
5533 s->src.state = PFGRE1S_INITIATING;
5534 s->dst.state = PFGRE1S_NO_TRAFFIC;
5535 s->timeout = PFTM_GREv1_INITIATING;
5536 break;
5537 case IPPROTO_ESP:
5538 s->src.state = PFESPS_INITIATING;
5539 s->dst.state = PFESPS_NO_TRAFFIC;
5540 s->timeout = PFTM_ESP_FIRST_PACKET;
5541 break;
5542 default:
5543 s->src.state = PFOTHERS_SINGLE;
5544 s->dst.state = PFOTHERS_NO_TRAFFIC;
5545 s->timeout = PFTM_OTHER_FIRST_PACKET;
5546 }
5547
5548 s->creation = pf_time_second();
5549 s->expire = pf_time_second();
5550
5551 if (sn != NULL) {
5552 s->src_node = sn;
5553 s->src_node->states++;
5554 VERIFY(s->src_node->states != 0);
5555 }
5556 if (nsn != NULL) {
5557 PF_ACPY(&nsn->raddr, &pd->naddr, af);
5558 s->nat_src_node = nsn;
5559 s->nat_src_node->states++;
5560 VERIFY(s->nat_src_node->states != 0);
5561 }
5562 if (pd->proto == IPPROTO_TCP) {
5563 if ((pd->flags & PFDESC_TCP_NORM) &&
5564 pf_normalize_tcp_init(pbuf, off, pd, th, &s->src,
5565 &s->dst)) {
5566 REASON_SET(&reason, PFRES_MEMORY);
5567 pf_src_tree_remove_state(s);
5568 STATE_DEC_COUNTERS(s);
5569 pool_put(&pf_state_pl, s);
5570 return (PF_DROP);
5571 }
5572 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
5573 pf_normalize_tcp_stateful(pbuf, off, pd, &reason,
5574 th, s, &s->src, &s->dst, &rewrite)) {
5575 /* This really shouldn't happen!!! */
5576 DPFPRINTF(PF_DEBUG_URGENT,
5577 ("pf_normalize_tcp_stateful failed on "
5578 "first pkt"));
5579 pf_normalize_tcp_cleanup(s);
5580 pf_src_tree_remove_state(s);
5581 STATE_DEC_COUNTERS(s);
5582 pool_put(&pf_state_pl, s);
5583 return (PF_DROP);
5584 }
5585 }
5586
5587 /* allocate state key and import values from psk */
5588 if ((sk = pf_alloc_state_key(s, &psk)) == NULL) {
5589 REASON_SET(&reason, PFRES_MEMORY);
5590 /*
5591 * XXXSCW: This will leak the freshly-allocated
5592 * state structure 's'. Although it should
5593 * eventually be aged-out and removed.
5594 */
5595 goto cleanup;
5596 }
5597
5598 pf_set_rt_ifp(s, saddr, af); /* needs s->state_key set */
5599
5600 pbuf = pd->mp; // XXXSCW: Why?
5601
5602 if (sk->app_state == 0) {
5603 switch (pd->proto) {
5604 case IPPROTO_TCP: {
5605 u_int16_t dport = (direction == PF_OUT) ?
5606 sk->ext_gwy.xport.port : sk->gwy.xport.port;
5607
5608 if (nr != NULL &&
5609 ntohs(dport) == PF_PPTP_PORT) {
5610 struct pf_app_state *as;
5611
5612 as = pool_get(&pf_app_state_pl,
5613 PR_WAITOK);
5614 if (!as) {
5615 REASON_SET(&reason,
5616 PFRES_MEMORY);
5617 goto cleanup;
5618 }
5619
5620 bzero(as, sizeof (*as));
5621 as->handler = pf_pptp_handler;
5622 as->compare_lan_ext = 0;
5623 as->compare_ext_gwy = 0;
5624 as->u.pptp.grev1_state = 0;
5625 sk->app_state = as;
5626 (void) hook_establish(&s->unlink_hooks,
5627 0, (hook_fn_t) pf_pptp_unlink, s);
5628 }
5629 break;
5630 }
5631
5632 case IPPROTO_UDP: {
5633 if (nr != NULL &&
5634 ntohs(uh->uh_sport) == PF_IKE_PORT &&
5635 ntohs(uh->uh_dport) == PF_IKE_PORT) {
5636 struct pf_app_state *as;
5637
5638 as = pool_get(&pf_app_state_pl,
5639 PR_WAITOK);
5640 if (!as) {
5641 REASON_SET(&reason,
5642 PFRES_MEMORY);
5643 goto cleanup;
5644 }
5645
5646 bzero(as, sizeof (*as));
5647 as->compare_lan_ext = pf_ike_compare;
5648 as->compare_ext_gwy = pf_ike_compare;
5649 as->u.ike.cookie = ike.initiator_cookie;
5650 sk->app_state = as;
5651 }
5652 break;
5653 }
5654
5655 default:
5656 break;
5657 }
5658 }
5659
5660 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
5661 if (pd->proto == IPPROTO_TCP)
5662 pf_normalize_tcp_cleanup(s);
5663 REASON_SET(&reason, PFRES_STATEINS);
5664 pf_src_tree_remove_state(s);
5665 STATE_DEC_COUNTERS(s);
5666 pool_put(&pf_state_pl, s);
5667 return (PF_DROP);
5668 } else {
5669 *sm = s;
5670 }
5671 if (tag > 0) {
5672 pf_tag_ref(tag);
5673 s->tag = tag;
5674 }
5675 if (pd->proto == IPPROTO_TCP &&
5676 (th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
5677 r->keep_state == PF_STATE_SYNPROXY) {
5678 int ua = (sk->af_lan == sk->af_gwy) ? 1 : 0;
5679 s->src.state = PF_TCPS_PROXY_SRC;
5680 if (nr != NULL) {
5681 if (direction == PF_OUT) {
5682 pf_change_ap(direction, pd->mp, saddr,
5683 &th->th_sport, pd->ip_sum,
5684 &th->th_sum, &pd->baddr,
5685 bxport.port, 0, af, pd->af, ua);
5686 sxport.port = th->th_sport;
5687 } else {
5688 pf_change_ap(direction, pd->mp, daddr,
5689 &th->th_dport, pd->ip_sum,
5690 &th->th_sum, &pd->baddr,
5691 bxport.port, 0, af, pd->af, ua);
5692 sxport.port = th->th_dport;
5693 }
5694 }
5695 s->src.seqhi = htonl(random());
5696 /* Find mss option */
5697 mss = pf_get_mss(pbuf, off, th->th_off, af);
5698 mss = pf_calc_mss(saddr, af, mss);
5699 mss = pf_calc_mss(daddr, af, mss);
5700 s->src.mss = mss;
5701 pf_send_tcp(r, af, daddr, saddr, th->th_dport,
5702 th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1,
5703 TH_SYN|TH_ACK, 0, s->src.mss, 0, 1, 0, NULL, NULL);
5704 REASON_SET(&reason, PFRES_SYNPROXY);
5705 return (PF_SYNPROXY_DROP);
5706 }
5707
5708 if (sk->app_state && sk->app_state->handler) {
5709 int offx = off;
5710
5711 switch (pd->proto) {
5712 case IPPROTO_TCP:
5713 offx += th->th_off << 2;
5714 break;
5715 case IPPROTO_UDP:
5716 offx += pd->hdr.udp->uh_ulen << 2;
5717 break;
5718 default:
5719 /* ALG handlers only apply to TCP and UDP rules */
5720 break;
5721 }
5722
5723 if (offx > off) {
5724 sk->app_state->handler(s, direction, offx,
5725 pd, kif);
5726 if (pd->lmw < 0) {
5727 REASON_SET(&reason, PFRES_MEMORY);
5728 return (PF_DROP);
5729 }
5730 pbuf = pd->mp; // XXXSCW: Why?
5731 }
5732 }
5733 }
5734
5735 /* copy back packet headers if we performed NAT operations */
5736 if (rewrite) {
5737 if (rewrite < off + hdrlen)
5738 rewrite = off + hdrlen;
5739
5740 if (pf_lazy_makewritable(pd, pd->mp, rewrite) == NULL) {
5741 REASON_SET(&reason, PFRES_MEMORY);
5742 return (PF_DROP);
5743 }
5744
5745 pbuf_copy_back(pbuf, off, hdrlen, pd->hdr.any);
5746 if (af == AF_INET6 && pd->naf == AF_INET)
5747 return pf_nat64_ipv6(pbuf, off, pd);
5748 else if (af == AF_INET && pd->naf == AF_INET6)
5749 return pf_nat64_ipv4(pbuf, off, pd);
5750
5751 }
5752
5753 return (PF_PASS);
5754}
5755
5756boolean_t is_nlc_enabled_glb = FALSE;
5757
5758static inline boolean_t
5759pf_is_dummynet_enabled(void)
5760{
5761#if DUMMYNET
5762 if (__probable(!PF_IS_ENABLED))
5763 return (FALSE);
5764
5765 if (__probable(!DUMMYNET_LOADED))
5766 return (FALSE);
5767
5768 if (__probable(TAILQ_EMPTY(pf_main_ruleset.
5769 rules[PF_RULESET_DUMMYNET].active.ptr)))
5770 return (FALSE);
5771
5772 return (TRUE);
5773#else
5774 return (FALSE);
5775#endif /* DUMMYNET */
5776}
5777
5778boolean_t
5779pf_is_nlc_enabled(void)
5780{
5781#if DUMMYNET
5782 if (__probable(!pf_is_dummynet_enabled()))
5783 return (FALSE);
5784
5785 if (__probable(!is_nlc_enabled_glb))
5786 return (FALSE);
5787
5788 return (TRUE);
5789#else
5790 return (FALSE);
5791#endif /* DUMMYNET */
5792}
5793
5794#if DUMMYNET
5795/*
5796 * When pf_test_dummynet() returns PF_PASS, the rule matching parameter "rm"
5797 * remains unchanged, meaning the packet did not match a dummynet rule.
5798 * when the packet does match a dummynet rule, pf_test_dummynet() returns
5799 * PF_PASS and zero out the mbuf rule as the packet is effectively siphoned
5800 * out by dummynet.
5801 */
5802static int
5803pf_test_dummynet(struct pf_rule **rm, int direction, struct pfi_kif *kif,
5804 pbuf_t **pbuf0, struct pf_pdesc *pd, struct ip_fw_args *fwa)
5805{
5806 pbuf_t *pbuf = *pbuf0;
5807 struct pf_rule *am = NULL;
5808 struct pf_ruleset *rsm = NULL;
5809 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
5810 sa_family_t af = pd->af;
5811 struct pf_rule *r, *a = NULL;
5812 struct pf_ruleset *ruleset = NULL;
5813 struct tcphdr *th = pd->hdr.tcp;
5814 u_short reason;
5815 int hdrlen = 0;
5816 int tag = -1;
5817 unsigned int rtableid = IFSCOPE_NONE;
5818 int asd = 0;
5819 int match = 0;
5820 u_int8_t icmptype = 0, icmpcode = 0;
5821 struct ip_fw_args dnflow;
5822 struct pf_rule *prev_matching_rule = fwa ? fwa->fwa_pf_rule : NULL;
5823 int found_prev_rule = (prev_matching_rule) ? 0 : 1;
5824
5825 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
5826
5827 if (!pf_is_dummynet_enabled())
5828 return (PF_PASS);
5829
5830 bzero(&dnflow, sizeof(dnflow));
5831
5832 hdrlen = 0;
5833
5834 /* Fragments don't gave protocol headers */
5835 if (!(pd->flags & PFDESC_IP_FRAG))
5836 switch (pd->proto) {
5837 case IPPROTO_TCP:
5838 dnflow.fwa_id.flags = pd->hdr.tcp->th_flags;
5839 dnflow.fwa_id.dst_port = ntohs(pd->hdr.tcp->th_dport);
5840 dnflow.fwa_id.src_port = ntohs(pd->hdr.tcp->th_sport);
5841 hdrlen = sizeof (*th);
5842 break;
5843 case IPPROTO_UDP:
5844 dnflow.fwa_id.dst_port = ntohs(pd->hdr.udp->uh_dport);
5845 dnflow.fwa_id.src_port = ntohs(pd->hdr.udp->uh_sport);
5846 hdrlen = sizeof (*pd->hdr.udp);
5847 break;
5848#if INET
5849 case IPPROTO_ICMP:
5850 if (af != AF_INET)
5851 break;
5852 hdrlen = ICMP_MINLEN;
5853 icmptype = pd->hdr.icmp->icmp_type;
5854 icmpcode = pd->hdr.icmp->icmp_code;
5855 break;
5856#endif /* INET */
5857#if INET6
5858 case IPPROTO_ICMPV6:
5859 if (af != AF_INET6)
5860 break;
5861 hdrlen = sizeof (*pd->hdr.icmp6);
5862 icmptype = pd->hdr.icmp6->icmp6_type;
5863 icmpcode = pd->hdr.icmp6->icmp6_code;
5864 break;
5865#endif /* INET6 */
5866 case IPPROTO_GRE:
5867 if (pd->proto_variant == PF_GRE_PPTP_VARIANT)
5868 hdrlen = sizeof (*pd->hdr.grev1);
5869 break;
5870 case IPPROTO_ESP:
5871 hdrlen = sizeof (*pd->hdr.esp);
5872 break;
5873 }
5874
5875 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_DUMMYNET].active.ptr);
5876
5877 while (r != NULL) {
5878 r->evaluations++;
5879 if (pfi_kif_match(r->kif, kif) == r->ifnot)
5880 r = r->skip[PF_SKIP_IFP].ptr;
5881 else if (r->direction && r->direction != direction)
5882 r = r->skip[PF_SKIP_DIR].ptr;
5883 else if (r->af && r->af != af)
5884 r = r->skip[PF_SKIP_AF].ptr;
5885 else if (r->proto && r->proto != pd->proto)
5886 r = r->skip[PF_SKIP_PROTO].ptr;
5887 else if (PF_MISMATCHAW(&r->src.addr, saddr, af,
5888 r->src.neg, kif))
5889 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
5890 /* tcp/udp only. port_op always 0 in other cases */
5891 else if (r->proto == pd->proto &&
5892 (r->proto == IPPROTO_TCP || r->proto == IPPROTO_UDP) &&
5893 ((pd->flags & PFDESC_IP_FRAG) ||
5894 ((r->src.xport.range.op &&
5895 !pf_match_port(r->src.xport.range.op,
5896 r->src.xport.range.port[0], r->src.xport.range.port[1],
5897 th->th_sport)))))
5898 r = r->skip[PF_SKIP_SRC_PORT].ptr;
5899 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af,
5900 r->dst.neg, NULL))
5901 r = r->skip[PF_SKIP_DST_ADDR].ptr;
5902 /* tcp/udp only. port_op always 0 in other cases */
5903 else if (r->proto == pd->proto &&
5904 (r->proto == IPPROTO_TCP || r->proto == IPPROTO_UDP) &&
5905 r->dst.xport.range.op &&
5906 ((pd->flags & PFDESC_IP_FRAG) ||
5907 !pf_match_port(r->dst.xport.range.op,
5908 r->dst.xport.range.port[0], r->dst.xport.range.port[1],
5909 th->th_dport)))
5910 r = r->skip[PF_SKIP_DST_PORT].ptr;
5911 /* icmp only. type always 0 in other cases */
5912 else if (r->type &&
5913 ((pd->flags & PFDESC_IP_FRAG) ||
5914 r->type != icmptype + 1))
5915 r = TAILQ_NEXT(r, entries);
5916 /* icmp only. type always 0 in other cases */
5917 else if (r->code &&
5918 ((pd->flags & PFDESC_IP_FRAG) ||
5919 r->code != icmpcode + 1))
5920 r = TAILQ_NEXT(r, entries);
5921 else if (r->tos && !(r->tos == pd->tos))
5922 r = TAILQ_NEXT(r, entries);
5923 else if (r->rule_flag & PFRULE_FRAGMENT)
5924 r = TAILQ_NEXT(r, entries);
5925 else if (pd->proto == IPPROTO_TCP &&
5926 ((pd->flags & PFDESC_IP_FRAG) ||
5927 (r->flagset & th->th_flags) != r->flags))
5928 r = TAILQ_NEXT(r, entries);
5929 else if (r->prob && r->prob <= (RandomULong() % (UINT_MAX - 1) + 1))
5930 r = TAILQ_NEXT(r, entries);
5931 else if (r->match_tag && !pf_match_tag(r, pd->pf_mtag, &tag))
5932 r = TAILQ_NEXT(r, entries);
5933 else {
5934 /*
5935 * Need to go past the previous dummynet matching rule
5936 */
5937 if (r->anchor == NULL) {
5938 if (found_prev_rule) {
5939 if (r->tag)
5940 tag = r->tag;
5941 if (PF_RTABLEID_IS_VALID(r->rtableid))
5942 rtableid = r->rtableid;
5943 match = 1;
5944 *rm = r;
5945 am = a;
5946 rsm = ruleset;
5947 if ((*rm)->quick)
5948 break;
5949 } else if (r == prev_matching_rule) {
5950 found_prev_rule = 1;
5951 }
5952 r = TAILQ_NEXT(r, entries);
5953 } else {
5954 pf_step_into_anchor(&asd, &ruleset,
5955 PF_RULESET_DUMMYNET, &r, &a, &match);
5956 }
5957 }
5958 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
5959 PF_RULESET_DUMMYNET, &r, &a, &match))
5960 break;
5961 }
5962 r = *rm;
5963 a = am;
5964 ruleset = rsm;
5965
5966 if (!match)
5967 return (PF_PASS);
5968
5969 REASON_SET(&reason, PFRES_DUMMYNET);
5970
5971 if (r->log) {
5972 PFLOG_PACKET(kif, h, pbuf, af, direction, reason, r,
5973 a, ruleset, pd);
5974 }
5975
5976 if (r->action == PF_NODUMMYNET) {
5977 int dirndx = (direction == PF_OUT);
5978
5979 r->packets[dirndx]++;
5980 r->bytes[dirndx] += pd->tot_len;
5981
5982 return (PF_PASS);
5983 }
5984 if (pf_tag_packet(pbuf, pd->pf_mtag, tag, rtableid, pd)) {
5985 REASON_SET(&reason, PFRES_MEMORY);
5986
5987 return (PF_DROP);
5988 }
5989
5990 if (r->dnpipe && ip_dn_io_ptr != NULL) {
5991 struct mbuf *m;
5992 int dirndx = (direction == PF_OUT);
5993
5994 r->packets[dirndx]++;
5995 r->bytes[dirndx] += pd->tot_len;
5996
5997 dnflow.fwa_cookie = r->dnpipe;
5998 dnflow.fwa_pf_rule = r;
5999 dnflow.fwa_id.proto = pd->proto;
6000 dnflow.fwa_flags = r->dntype;
6001 switch (af) {
6002 case AF_INET:
6003 dnflow.fwa_id.addr_type = 4;
6004 dnflow.fwa_id.src_ip = ntohl(saddr->v4addr.s_addr);
6005 dnflow.fwa_id.dst_ip = ntohl(daddr->v4addr.s_addr);
6006 break;
6007 case AF_INET6:
6008 dnflow.fwa_id.addr_type = 6;
6009 dnflow.fwa_id.src_ip6 = saddr->v6addr;
6010 dnflow.fwa_id.dst_ip6 = saddr->v6addr;
6011 break;
6012 }
6013
6014 if (fwa != NULL) {
6015 dnflow.fwa_oif = fwa->fwa_oif;
6016 dnflow.fwa_oflags = fwa->fwa_oflags;
6017 /*
6018 * Note that fwa_ro, fwa_dst and fwa_ipoa are
6019 * actually in a union so the following does work
6020 * for both IPv4 and IPv6
6021 */
6022 dnflow.fwa_ro = fwa->fwa_ro;
6023 dnflow.fwa_dst = fwa->fwa_dst;
6024 dnflow.fwa_ipoa = fwa->fwa_ipoa;
6025 dnflow.fwa_ro6_pmtu = fwa->fwa_ro6_pmtu;
6026 dnflow.fwa_origifp = fwa->fwa_origifp;
6027 dnflow.fwa_mtu = fwa->fwa_mtu;
6028 dnflow.fwa_alwaysfrag = fwa->fwa_alwaysfrag;
6029 dnflow.fwa_unfragpartlen = fwa->fwa_unfragpartlen;
6030 dnflow.fwa_exthdrs = fwa->fwa_exthdrs;
6031 }
6032
6033 if (af == AF_INET) {
6034 struct ip *iphdr = pbuf->pb_data;
6035 NTOHS(iphdr->ip_len);
6036 NTOHS(iphdr->ip_off);
6037 }
6038 /*
6039 * Don't need to unlock pf_lock as NET_THREAD_HELD_PF
6040 * allows for recursive behavior
6041 */
6042 m = pbuf_to_mbuf(pbuf, TRUE);
6043 if (m != NULL) {
6044 ip_dn_io_ptr(m,
6045 dnflow.fwa_cookie, (af == AF_INET) ?
6046 ((direction==PF_IN) ? DN_TO_IP_IN : DN_TO_IP_OUT) :
6047 ((direction==PF_IN) ? DN_TO_IP6_IN : DN_TO_IP6_OUT),
6048 &dnflow, DN_CLIENT_PF);
6049 }
6050
6051 /*
6052 * The packet is siphoned out by dummynet so return a NULL
6053 * pbuf so the caller can still return success.
6054 */
6055 *pbuf0 = NULL;
6056
6057 return (PF_PASS);
6058 }
6059
6060 return (PF_PASS);
6061}
6062#endif /* DUMMYNET */
6063
6064static int
6065pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif,
6066 pbuf_t *pbuf, void *h, struct pf_pdesc *pd, struct pf_rule **am,
6067 struct pf_ruleset **rsm)
6068{
6069#pragma unused(h)
6070 struct pf_rule *r, *a = NULL;
6071 struct pf_ruleset *ruleset = NULL;
6072 sa_family_t af = pd->af;
6073 u_short reason;
6074 int tag = -1;
6075 int asd = 0;
6076 int match = 0;
6077
6078 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
6079 while (r != NULL) {
6080 r->evaluations++;
6081 if (pfi_kif_match(r->kif, kif) == r->ifnot)
6082 r = r->skip[PF_SKIP_IFP].ptr;
6083 else if (r->direction && r->direction != direction)
6084 r = r->skip[PF_SKIP_DIR].ptr;
6085 else if (r->af && r->af != af)
6086 r = r->skip[PF_SKIP_AF].ptr;
6087 else if (r->proto && r->proto != pd->proto)
6088 r = r->skip[PF_SKIP_PROTO].ptr;
6089 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af,
6090 r->src.neg, kif))
6091 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
6092 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af,
6093 r->dst.neg, NULL))
6094 r = r->skip[PF_SKIP_DST_ADDR].ptr;
6095 else if ((r->rule_flag & PFRULE_TOS) && r->tos &&
6096 !(r->tos & pd->tos))
6097 r = TAILQ_NEXT(r, entries);
6098 else if ((r->rule_flag & PFRULE_DSCP) && r->tos &&
6099 !(r->tos & (pd->tos & DSCP_MASK)))
6100 r = TAILQ_NEXT(r, entries);
6101 else if ((r->rule_flag & PFRULE_SC) && r->tos &&
6102 ((r->tos & SCIDX_MASK) != pd->sc))
6103 r = TAILQ_NEXT(r, entries);
6104 else if (r->os_fingerprint != PF_OSFP_ANY)
6105 r = TAILQ_NEXT(r, entries);
6106 else if (pd->proto == IPPROTO_UDP &&
6107 (r->src.xport.range.op || r->dst.xport.range.op))
6108 r = TAILQ_NEXT(r, entries);
6109 else if (pd->proto == IPPROTO_TCP &&
6110 (r->src.xport.range.op || r->dst.xport.range.op ||
6111 r->flagset))
6112 r = TAILQ_NEXT(r, entries);
6113 else if ((pd->proto == IPPROTO_ICMP ||
6114 pd->proto == IPPROTO_ICMPV6) &&
6115 (r->type || r->code))
6116 r = TAILQ_NEXT(r, entries);
6117 else if (r->prob && r->prob <= (RandomULong() % (UINT_MAX - 1) + 1))
6118 r = TAILQ_NEXT(r, entries);
6119 else if (r->match_tag && !pf_match_tag(r, pd->pf_mtag, &tag))
6120 r = TAILQ_NEXT(r, entries);
6121 else {
6122 if (r->anchor == NULL) {
6123 match = 1;
6124 *rm = r;
6125 *am = a;
6126 *rsm = ruleset;
6127 if ((*rm)->quick)
6128 break;
6129 r = TAILQ_NEXT(r, entries);
6130 } else
6131 pf_step_into_anchor(&asd, &ruleset,
6132 PF_RULESET_FILTER, &r, &a, &match);
6133 }
6134 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
6135 PF_RULESET_FILTER, &r, &a, &match))
6136 break;
6137 }
6138 r = *rm;
6139 a = *am;
6140 ruleset = *rsm;
6141
6142 REASON_SET(&reason, PFRES_MATCH);
6143
6144 if (r->log)
6145 PFLOG_PACKET(kif, h, pbuf, af, direction, reason, r, a, ruleset,
6146 pd);
6147
6148 if (r->action != PF_PASS)
6149 return (PF_DROP);
6150
6151 if (pf_tag_packet(pbuf, pd->pf_mtag, tag, -1, NULL)) {
6152 REASON_SET(&reason, PFRES_MEMORY);
6153 return (PF_DROP);
6154 }
6155
6156 return (PF_PASS);
6157}
6158
6159static void
6160pf_pptp_handler(struct pf_state *s, int direction, int off,
6161 struct pf_pdesc *pd, struct pfi_kif *kif)
6162{
6163#pragma unused(direction)
6164 struct tcphdr *th;
6165 struct pf_pptp_state *pptps;
6166 struct pf_pptp_ctrl_msg cm;
6167 size_t plen, tlen;
6168 struct pf_state *gs;
6169 u_int16_t ct;
6170 u_int16_t *pac_call_id;
6171 u_int16_t *pns_call_id;
6172 u_int16_t *spoof_call_id;
6173 u_int8_t *pac_state;
6174 u_int8_t *pns_state;
6175 enum { PF_PPTP_PASS, PF_PPTP_INSERT_GRE, PF_PPTP_REMOVE_GRE } op;
6176 pbuf_t *pbuf;
6177 struct pf_state_key *sk;
6178 struct pf_state_key *gsk;
6179 struct pf_app_state *gas;
6180
6181 sk = s->state_key;
6182 pptps = &sk->app_state->u.pptp;
6183 gs = pptps->grev1_state;
6184
6185 if (gs)
6186 gs->expire = pf_time_second();
6187
6188 pbuf = pd->mp;
6189 plen = min(sizeof (cm), pbuf->pb_packet_len - off);
6190 if (plen < PF_PPTP_CTRL_MSG_MINSIZE)
6191 return;
6192 tlen = plen - PF_PPTP_CTRL_MSG_MINSIZE;
6193 pbuf_copy_data(pbuf, off, plen, &cm);
6194
6195 if (ntohl(cm.hdr.magic) != PF_PPTP_MAGIC_NUMBER)
6196 return;
6197 if (ntohs(cm.hdr.type) != 1)
6198 return;
6199
6200#define TYPE_LEN_CHECK(_type, _name) \
6201 case PF_PPTP_CTRL_TYPE_##_type: \
6202 if (tlen < sizeof(struct pf_pptp_ctrl_##_name)) \
6203 return; \
6204 break;
6205
6206 switch (cm.ctrl.type) {
6207 TYPE_LEN_CHECK(START_REQ, start_req);
6208 TYPE_LEN_CHECK(START_RPY, start_rpy);
6209 TYPE_LEN_CHECK(STOP_REQ, stop_req);
6210 TYPE_LEN_CHECK(STOP_RPY, stop_rpy);
6211 TYPE_LEN_CHECK(ECHO_REQ, echo_req);
6212 TYPE_LEN_CHECK(ECHO_RPY, echo_rpy);
6213 TYPE_LEN_CHECK(CALL_OUT_REQ, call_out_req);
6214 TYPE_LEN_CHECK(CALL_OUT_RPY, call_out_rpy);
6215 TYPE_LEN_CHECK(CALL_IN_1ST, call_in_1st);
6216 TYPE_LEN_CHECK(CALL_IN_2ND, call_in_2nd);
6217 TYPE_LEN_CHECK(CALL_IN_3RD, call_in_3rd);
6218 TYPE_LEN_CHECK(CALL_CLR, call_clr);
6219 TYPE_LEN_CHECK(CALL_DISC, call_disc);
6220 TYPE_LEN_CHECK(ERROR, error);
6221 TYPE_LEN_CHECK(SET_LINKINFO, set_linkinfo);
6222 default:
6223 return;
6224 }
6225#undef TYPE_LEN_CHECK
6226
6227 if (!gs) {
6228 gs = pool_get(&pf_state_pl, PR_WAITOK);
6229 if (!gs)
6230 return;
6231
6232 memcpy(gs, s, sizeof (*gs));
6233
6234 memset(&gs->entry_id, 0, sizeof (gs->entry_id));
6235 memset(&gs->entry_list, 0, sizeof (gs->entry_list));
6236
6237 TAILQ_INIT(&gs->unlink_hooks);
6238 gs->rt_kif = NULL;
6239 gs->creation = 0;
6240 gs->pfsync_time = 0;
6241 gs->packets[0] = gs->packets[1] = 0;
6242 gs->bytes[0] = gs->bytes[1] = 0;
6243 gs->timeout = PFTM_UNLINKED;
6244 gs->id = gs->creatorid = 0;
6245 gs->src.state = gs->dst.state = PFGRE1S_NO_TRAFFIC;
6246 gs->src.scrub = gs->dst.scrub = 0;
6247
6248 gas = pool_get(&pf_app_state_pl, PR_NOWAIT);
6249 if (!gas) {
6250 pool_put(&pf_state_pl, gs);
6251 return;
6252 }
6253
6254 gsk = pf_alloc_state_key(gs, NULL);
6255 if (!gsk) {
6256 pool_put(&pf_app_state_pl, gas);
6257 pool_put(&pf_state_pl, gs);
6258 return;
6259 }
6260
6261 memcpy(&gsk->lan, &sk->lan, sizeof (gsk->lan));
6262 memcpy(&gsk->gwy, &sk->gwy, sizeof (gsk->gwy));
6263 memcpy(&gsk->ext_lan, &sk->ext_lan, sizeof (gsk->ext_lan));
6264 memcpy(&gsk->ext_gwy, &sk->ext_gwy, sizeof (gsk->ext_gwy));
6265 gsk->af_lan = sk->af_lan;
6266 gsk->af_gwy = sk->af_gwy;
6267 gsk->proto = IPPROTO_GRE;
6268 gsk->proto_variant = PF_GRE_PPTP_VARIANT;
6269 gsk->app_state = gas;
6270 gsk->lan.xport.call_id = 0;
6271 gsk->gwy.xport.call_id = 0;
6272 gsk->ext_lan.xport.call_id = 0;
6273 gsk->ext_gwy.xport.call_id = 0;
6274 gsk->flowsrc = FLOWSRC_PF;
6275 gsk->flowhash = pf_calc_state_key_flowhash(gsk);
6276 memset(gas, 0, sizeof (*gas));
6277 gas->u.grev1.pptp_state = s;
6278 STATE_INC_COUNTERS(gs);
6279 pptps->grev1_state = gs;
6280 (void) hook_establish(&gs->unlink_hooks, 0,
6281 (hook_fn_t) pf_grev1_unlink, gs);
6282 } else {
6283 gsk = gs->state_key;
6284 }
6285
6286 switch (sk->direction) {
6287 case PF_IN:
6288 pns_call_id = &gsk->ext_lan.xport.call_id;
6289 pns_state = &gs->dst.state;
6290 pac_call_id = &gsk->lan.xport.call_id;
6291 pac_state = &gs->src.state;
6292 break;
6293
6294 case PF_OUT:
6295 pns_call_id = &gsk->lan.xport.call_id;
6296 pns_state = &gs->src.state;
6297 pac_call_id = &gsk->ext_lan.xport.call_id;
6298 pac_state = &gs->dst.state;
6299 break;
6300
6301 default:
6302 DPFPRINTF(PF_DEBUG_URGENT,
6303 ("pf_pptp_handler: bad directional!\n"));
6304 return;
6305 }
6306
6307 spoof_call_id = 0;
6308 op = PF_PPTP_PASS;
6309
6310 ct = ntohs(cm.ctrl.type);
6311
6312 switch (ct) {
6313 case PF_PPTP_CTRL_TYPE_CALL_OUT_REQ:
6314 *pns_call_id = cm.msg.call_out_req.call_id;
6315 *pns_state = PFGRE1S_INITIATING;
6316 if (s->nat_rule.ptr && pns_call_id == &gsk->lan.xport.call_id)
6317 spoof_call_id = &cm.msg.call_out_req.call_id;
6318 break;
6319
6320 case PF_PPTP_CTRL_TYPE_CALL_OUT_RPY:
6321 *pac_call_id = cm.msg.call_out_rpy.call_id;
6322 if (s->nat_rule.ptr)
6323 spoof_call_id =
6324 (pac_call_id == &gsk->lan.xport.call_id) ?
6325 &cm.msg.call_out_rpy.call_id :
6326 &cm.msg.call_out_rpy.peer_call_id;
6327 if (gs->timeout == PFTM_UNLINKED) {
6328 *pac_state = PFGRE1S_INITIATING;
6329 op = PF_PPTP_INSERT_GRE;
6330 }
6331 break;
6332
6333 case PF_PPTP_CTRL_TYPE_CALL_IN_1ST:
6334 *pns_call_id = cm.msg.call_in_1st.call_id;
6335 *pns_state = PFGRE1S_INITIATING;
6336 if (s->nat_rule.ptr && pns_call_id == &gsk->lan.xport.call_id)
6337 spoof_call_id = &cm.msg.call_in_1st.call_id;
6338 break;
6339
6340 case PF_PPTP_CTRL_TYPE_CALL_IN_2ND:
6341 *pac_call_id = cm.msg.call_in_2nd.call_id;
6342 *pac_state = PFGRE1S_INITIATING;
6343 if (s->nat_rule.ptr)
6344 spoof_call_id =
6345 (pac_call_id == &gsk->lan.xport.call_id) ?
6346 &cm.msg.call_in_2nd.call_id :
6347 &cm.msg.call_in_2nd.peer_call_id;
6348 break;
6349
6350 case PF_PPTP_CTRL_TYPE_CALL_IN_3RD:
6351 if (s->nat_rule.ptr && pns_call_id == &gsk->lan.xport.call_id)
6352 spoof_call_id = &cm.msg.call_in_3rd.call_id;
6353 if (cm.msg.call_in_3rd.call_id != *pns_call_id) {
6354 break;
6355 }
6356 if (gs->timeout == PFTM_UNLINKED)
6357 op = PF_PPTP_INSERT_GRE;
6358 break;
6359
6360 case PF_PPTP_CTRL_TYPE_CALL_CLR:
6361 if (cm.msg.call_clr.call_id != *pns_call_id)
6362 op = PF_PPTP_REMOVE_GRE;
6363 break;
6364
6365 case PF_PPTP_CTRL_TYPE_CALL_DISC:
6366 if (cm.msg.call_clr.call_id != *pac_call_id)
6367 op = PF_PPTP_REMOVE_GRE;
6368 break;
6369
6370 case PF_PPTP_CTRL_TYPE_ERROR:
6371 if (s->nat_rule.ptr && pns_call_id == &gsk->lan.xport.call_id)
6372 spoof_call_id = &cm.msg.error.peer_call_id;
6373 break;
6374
6375 case PF_PPTP_CTRL_TYPE_SET_LINKINFO:
6376 if (s->nat_rule.ptr && pac_call_id == &gsk->lan.xport.call_id)
6377 spoof_call_id = &cm.msg.set_linkinfo.peer_call_id;
6378 break;
6379
6380 default:
6381 op = PF_PPTP_PASS;
6382 break;
6383 }
6384
6385 if (!gsk->gwy.xport.call_id && gsk->lan.xport.call_id) {
6386 gsk->gwy.xport.call_id = gsk->lan.xport.call_id;
6387 if (spoof_call_id) {
6388 u_int16_t call_id = 0;
6389 int n = 0;
6390 struct pf_state_key_cmp key;
6391
6392 key.af_gwy = gsk->af_gwy;
6393 key.proto = IPPROTO_GRE;
6394 key.proto_variant = PF_GRE_PPTP_VARIANT;
6395 PF_ACPY(&key.gwy.addr, &gsk->gwy.addr, key.af_gwy);
6396 PF_ACPY(&key.ext_gwy.addr, &gsk->ext_gwy.addr, key.af_gwy);
6397 key.gwy.xport.call_id = gsk->gwy.xport.call_id;
6398 key.ext_gwy.xport.call_id = gsk->ext_gwy.xport.call_id;
6399 do {
6400 call_id = htonl(random());
6401 } while (!call_id);
6402
6403 while (pf_find_state_all(&key, PF_IN, 0)) {
6404 call_id = ntohs(call_id);
6405 --call_id;
6406 if (--call_id == 0) call_id = 0xffff;
6407 call_id = htons(call_id);
6408
6409 key.gwy.xport.call_id = call_id;
6410
6411 if (++n > 65535) {
6412 DPFPRINTF(PF_DEBUG_URGENT,
6413 ("pf_pptp_handler: failed to spoof "
6414 "call id\n"));
6415 key.gwy.xport.call_id = 0;
6416 break;
6417 }
6418 }
6419
6420 gsk->gwy.xport.call_id = call_id;
6421 }
6422 }
6423
6424 th = pd->hdr.tcp;
6425
6426 if (spoof_call_id && gsk->lan.xport.call_id != gsk->gwy.xport.call_id) {
6427 if (*spoof_call_id == gsk->gwy.xport.call_id) {
6428 *spoof_call_id = gsk->lan.xport.call_id;
6429 th->th_sum = pf_cksum_fixup(th->th_sum,
6430 gsk->gwy.xport.call_id, gsk->lan.xport.call_id, 0);
6431 } else {
6432 *spoof_call_id = gsk->gwy.xport.call_id;
6433 th->th_sum = pf_cksum_fixup(th->th_sum,
6434 gsk->lan.xport.call_id, gsk->gwy.xport.call_id, 0);
6435 }
6436
6437 if (pf_lazy_makewritable(pd, pbuf, off + plen) == NULL) {
6438 pptps->grev1_state = NULL;
6439 STATE_DEC_COUNTERS(gs);
6440 pool_put(&pf_state_pl, gs);
6441 return;
6442 }
6443 pbuf_copy_back(pbuf, off, plen, &cm);
6444 }
6445
6446 switch (op) {
6447 case PF_PPTP_REMOVE_GRE:
6448 gs->timeout = PFTM_PURGE;
6449 gs->src.state = gs->dst.state = PFGRE1S_NO_TRAFFIC;
6450 gsk->lan.xport.call_id = 0;
6451 gsk->gwy.xport.call_id = 0;
6452 gsk->ext_lan.xport.call_id = 0;
6453 gsk->ext_gwy.xport.call_id = 0;
6454 gs->id = gs->creatorid = 0;
6455 break;
6456
6457 case PF_PPTP_INSERT_GRE:
6458 gs->creation = pf_time_second();
6459 gs->expire = pf_time_second();
6460 gs->timeout = PFTM_TCP_ESTABLISHED;
6461 if (gs->src_node != NULL) {
6462 ++gs->src_node->states;
6463 VERIFY(gs->src_node->states != 0);
6464 }
6465 if (gs->nat_src_node != NULL) {
6466 ++gs->nat_src_node->states;
6467 VERIFY(gs->nat_src_node->states != 0);
6468 }
6469 pf_set_rt_ifp(gs, &sk->lan.addr, sk->af_lan);
6470 if (pf_insert_state(BOUND_IFACE(s->rule.ptr, kif), gs)) {
6471
6472 /*
6473 * <jhw@apple.com>
6474 * FIX ME: insertion can fail when multiple PNS
6475 * behind the same NAT open calls to the same PAC
6476 * simultaneously because spoofed call ID numbers
6477 * are chosen before states are inserted. This is
6478 * hard to fix and happens infrequently enough that
6479 * users will normally try again and this ALG will
6480 * succeed. Failures are expected to be rare enough
6481 * that fixing this is a low priority.
6482 */
6483 pptps->grev1_state = NULL;
6484 pd->lmw = -1; /* Force PF_DROP on PFRES_MEMORY */
6485 pf_src_tree_remove_state(gs);
6486 STATE_DEC_COUNTERS(gs);
6487 pool_put(&pf_state_pl, gs);
6488 DPFPRINTF(PF_DEBUG_URGENT, ("pf_pptp_handler: error "
6489 "inserting GREv1 state.\n"));
6490 }
6491 break;
6492
6493 default:
6494 break;
6495 }
6496}
6497
6498static void
6499pf_pptp_unlink(struct pf_state *s)
6500{
6501 struct pf_app_state *as = s->state_key->app_state;
6502 struct pf_state *grev1s = as->u.pptp.grev1_state;
6503
6504 if (grev1s) {
6505 struct pf_app_state *gas = grev1s->state_key->app_state;
6506
6507 if (grev1s->timeout < PFTM_MAX)
6508 grev1s->timeout = PFTM_PURGE;
6509 gas->u.grev1.pptp_state = NULL;
6510 as->u.pptp.grev1_state = NULL;
6511 }
6512}
6513
6514static void
6515pf_grev1_unlink(struct pf_state *s)
6516{
6517 struct pf_app_state *as = s->state_key->app_state;
6518 struct pf_state *pptps = as->u.grev1.pptp_state;
6519
6520 if (pptps) {
6521 struct pf_app_state *pas = pptps->state_key->app_state;
6522
6523 pas->u.pptp.grev1_state = NULL;
6524 as->u.grev1.pptp_state = NULL;
6525 }
6526}
6527
6528static int
6529pf_ike_compare(struct pf_app_state *a, struct pf_app_state *b)
6530{
6531 int64_t d = a->u.ike.cookie - b->u.ike.cookie;
6532 return ((d > 0) ? 1 : ((d < 0) ? -1 : 0));
6533}
6534
6535static int
6536pf_do_nat64(struct pf_state_key *sk, struct pf_pdesc *pd, pbuf_t *pbuf,
6537 int off)
6538{
6539 if (pd->af == AF_INET) {
6540 if (pd->af != sk->af_lan) {
6541 pd->ndaddr = sk->lan.addr;
6542 pd->naddr = sk->ext_lan.addr;
6543 } else {
6544 pd->naddr = sk->gwy.addr;
6545 pd->ndaddr = sk->ext_gwy.addr;
6546 }
6547 return (pf_nat64_ipv4(pbuf, off, pd));
6548 }
6549 else if (pd->af == AF_INET6) {
6550 if (pd->af != sk->af_lan) {
6551 pd->ndaddr = sk->lan.addr;
6552 pd->naddr = sk->ext_lan.addr;
6553 } else {
6554 pd->naddr = sk->gwy.addr;
6555 pd->ndaddr = sk->ext_gwy.addr;
6556 }
6557 return (pf_nat64_ipv6(pbuf, off, pd));
6558 }
6559 return (PF_DROP);
6560}
6561
6562static int
6563pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
6564 pbuf_t *pbuf, int off, void *h, struct pf_pdesc *pd,
6565 u_short *reason)
6566{
6567#pragma unused(h)
6568 struct pf_state_key_cmp key;
6569 struct tcphdr *th = pd->hdr.tcp;
6570 u_int16_t win = ntohs(th->th_win);
6571 u_int32_t ack, end, seq, orig_seq;
6572 u_int8_t sws, dws;
6573 int ackskew;
6574 int copyback = 0;
6575 struct pf_state_peer *src, *dst;
6576 struct pf_state_key *sk;
6577
6578 key.app_state = 0;
6579 key.proto = IPPROTO_TCP;
6580 key.af_lan = key.af_gwy = pd->af;
6581
6582 /*
6583 * For NAT64 the first time rule search and state creation
6584 * is done on the incoming side only.
6585 * Once the state gets created, NAT64's LAN side (ipv6) will
6586 * not be able to find the state in ext-gwy tree as that normally
6587 * is intended to be looked up for incoming traffic from the
6588 * WAN side.
6589 * Therefore to handle NAT64 case we init keys here for both
6590 * lan-ext as well as ext-gwy trees.
6591 * In the state lookup we attempt a lookup on both trees if
6592 * first one does not return any result and return a match if
6593 * the match state's was created by NAT64 rule.
6594 */
6595 PF_ACPY(&key.ext_gwy.addr, pd->src, key.af_gwy);
6596 PF_ACPY(&key.gwy.addr, pd->dst, key.af_gwy);
6597 key.ext_gwy.xport.port = th->th_sport;
6598 key.gwy.xport.port = th->th_dport;
6599
6600 PF_ACPY(&key.lan.addr, pd->src, key.af_lan);
6601 PF_ACPY(&key.ext_lan.addr, pd->dst, key.af_lan);
6602 key.lan.xport.port = th->th_sport;
6603 key.ext_lan.xport.port = th->th_dport;
6604
6605 STATE_LOOKUP();
6606
6607 sk = (*state)->state_key;
6608 /*
6609 * In case of NAT64 the translation is first applied on the LAN
6610 * side. Therefore for stack's address family comparison
6611 * we use sk->af_lan.
6612 */
6613 if ((direction == sk->direction) && (pd->af == sk->af_lan)) {
6614 src = &(*state)->src;
6615 dst = &(*state)->dst;
6616 } else {
6617 src = &(*state)->dst;
6618 dst = &(*state)->src;
6619 }
6620
6621 if (src->state == PF_TCPS_PROXY_SRC) {
6622 if (direction != sk->direction) {
6623 REASON_SET(reason, PFRES_SYNPROXY);
6624 return (PF_SYNPROXY_DROP);
6625 }
6626 if (th->th_flags & TH_SYN) {
6627 if (ntohl(th->th_seq) != src->seqlo) {
6628 REASON_SET(reason, PFRES_SYNPROXY);
6629 return (PF_DROP);
6630 }
6631 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
6632 pd->src, th->th_dport, th->th_sport,
6633 src->seqhi, ntohl(th->th_seq) + 1,
6634 TH_SYN|TH_ACK, 0, src->mss, 0, 1,
6635 0, NULL, NULL);
6636 REASON_SET(reason, PFRES_SYNPROXY);
6637 return (PF_SYNPROXY_DROP);
6638 } else if (!(th->th_flags & TH_ACK) ||
6639 (ntohl(th->th_ack) != src->seqhi + 1) ||
6640 (ntohl(th->th_seq) != src->seqlo + 1)) {
6641 REASON_SET(reason, PFRES_SYNPROXY);
6642 return (PF_DROP);
6643 } else if ((*state)->src_node != NULL &&
6644 pf_src_connlimit(state)) {
6645 REASON_SET(reason, PFRES_SRCLIMIT);
6646 return (PF_DROP);
6647 } else
6648 src->state = PF_TCPS_PROXY_DST;
6649 }
6650 if (src->state == PF_TCPS_PROXY_DST) {
6651 struct pf_state_host *psrc, *pdst;
6652
6653 if (direction == PF_OUT) {
6654 psrc = &sk->gwy;
6655 pdst = &sk->ext_gwy;
6656 } else {
6657 psrc = &sk->ext_lan;
6658 pdst = &sk->lan;
6659 }
6660 if (direction == sk->direction) {
6661 if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) ||
6662 (ntohl(th->th_ack) != src->seqhi + 1) ||
6663 (ntohl(th->th_seq) != src->seqlo + 1)) {
6664 REASON_SET(reason, PFRES_SYNPROXY);
6665 return (PF_DROP);
6666 }
6667 src->max_win = MAX(ntohs(th->th_win), 1);
6668 if (dst->seqhi == 1)
6669 dst->seqhi = htonl(random());
6670 pf_send_tcp((*state)->rule.ptr, pd->af, &psrc->addr,
6671 &pdst->addr, psrc->xport.port, pdst->xport.port,
6672 dst->seqhi, 0, TH_SYN, 0,
6673 src->mss, 0, 0, (*state)->tag, NULL, NULL);
6674 REASON_SET(reason, PFRES_SYNPROXY);
6675 return (PF_SYNPROXY_DROP);
6676 } else if (((th->th_flags & (TH_SYN|TH_ACK)) !=
6677 (TH_SYN|TH_ACK)) ||
6678 (ntohl(th->th_ack) != dst->seqhi + 1)) {
6679 REASON_SET(reason, PFRES_SYNPROXY);
6680 return (PF_DROP);
6681 } else {
6682 dst->max_win = MAX(ntohs(th->th_win), 1);
6683 dst->seqlo = ntohl(th->th_seq);
6684 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
6685 pd->src, th->th_dport, th->th_sport,
6686 ntohl(th->th_ack), ntohl(th->th_seq) + 1,
6687 TH_ACK, src->max_win, 0, 0, 0,
6688 (*state)->tag, NULL, NULL);
6689 pf_send_tcp((*state)->rule.ptr, pd->af, &psrc->addr,
6690 &pdst->addr, psrc->xport.port, pdst->xport.port,
6691 src->seqhi + 1, src->seqlo + 1,
6692 TH_ACK, dst->max_win, 0, 0, 1,
6693 0, NULL, NULL);
6694 src->seqdiff = dst->seqhi -
6695 src->seqlo;
6696 dst->seqdiff = src->seqhi -
6697 dst->seqlo;
6698 src->seqhi = src->seqlo +
6699 dst->max_win;
6700 dst->seqhi = dst->seqlo +
6701 src->max_win;
6702 src->wscale = dst->wscale = 0;
6703 src->state = dst->state =
6704 TCPS_ESTABLISHED;
6705 REASON_SET(reason, PFRES_SYNPROXY);
6706 return (PF_SYNPROXY_DROP);
6707 }
6708 }
6709
6710 if (((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN) &&
6711 dst->state >= TCPS_FIN_WAIT_2 &&
6712 src->state >= TCPS_FIN_WAIT_2) {
6713 if (pf_status.debug >= PF_DEBUG_MISC) {
6714 printf("pf: state reuse ");
6715 pf_print_state(*state);
6716 pf_print_flags(th->th_flags);
6717 printf("\n");
6718 }
6719 /* XXX make sure it's the same direction ?? */
6720 src->state = dst->state = TCPS_CLOSED;
6721 pf_unlink_state(*state);
6722 *state = NULL;
6723 return (PF_DROP);
6724 }
6725
6726 if ((th->th_flags & TH_SYN) == 0) {
6727 sws = (src->wscale & PF_WSCALE_FLAG) ?
6728 (src->wscale & PF_WSCALE_MASK) : TCP_MAX_WINSHIFT;
6729 dws = (dst->wscale & PF_WSCALE_FLAG) ?
6730 (dst->wscale & PF_WSCALE_MASK) : TCP_MAX_WINSHIFT;
6731 }
6732 else
6733 sws = dws = 0;
6734
6735 /*
6736 * Sequence tracking algorithm from Guido van Rooij's paper:
6737 * http://www.madison-gurkha.com/publications/tcp_filtering/
6738 * tcp_filtering.ps
6739 */
6740
6741 orig_seq = seq = ntohl(th->th_seq);
6742 if (src->seqlo == 0) {
6743 /* First packet from this end. Set its state */
6744
6745 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) &&
6746 src->scrub == NULL) {
6747 if (pf_normalize_tcp_init(pbuf, off, pd, th, src, dst)) {
6748 REASON_SET(reason, PFRES_MEMORY);
6749 return (PF_DROP);
6750 }
6751 }
6752
6753 /* Deferred generation of sequence number modulator */
6754 if (dst->seqdiff && !src->seqdiff) {
6755 /* use random iss for the TCP server */
6756 while ((src->seqdiff = random() - seq) == 0)
6757 ;
6758 ack = ntohl(th->th_ack) - dst->seqdiff;
6759 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
6760 src->seqdiff), 0);
6761 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
6762 copyback = off + sizeof (*th);
6763 } else {
6764 ack = ntohl(th->th_ack);
6765 }
6766
6767 end = seq + pd->p_len;
6768 if (th->th_flags & TH_SYN) {
6769 end++;
6770 if (dst->wscale & PF_WSCALE_FLAG) {
6771 src->wscale = pf_get_wscale(pbuf, off,
6772 th->th_off, pd->af);
6773 if (src->wscale & PF_WSCALE_FLAG) {
6774 /*
6775 * Remove scale factor from initial
6776 * window
6777 */
6778 sws = src->wscale & PF_WSCALE_MASK;
6779 win = ((u_int32_t)win + (1 << sws) - 1)
6780 >> sws;
6781 dws = dst->wscale & PF_WSCALE_MASK;
6782 } else {
6783 /*
6784 * Window scale negotiation has failed,
6785 * therefore we must restore the window
6786 * scale in the state record that we
6787 * optimistically removed in
6788 * pf_test_rule(). Care is required to
6789 * prevent arithmetic overflow from
6790 * zeroing the window when it's
6791 * truncated down to 16-bits.
6792 */
6793 u_int32_t max_win = dst->max_win;
6794 max_win <<=
6795 dst->wscale & PF_WSCALE_MASK;
6796 dst->max_win = MIN(0xffff, max_win);
6797 /* in case of a retrans SYN|ACK */
6798 dst->wscale = 0;
6799 }
6800 }
6801 }
6802 if (th->th_flags & TH_FIN)
6803 end++;
6804
6805 src->seqlo = seq;
6806 if (src->state < TCPS_SYN_SENT)
6807 src->state = TCPS_SYN_SENT;
6808
6809 /*
6810 * May need to slide the window (seqhi may have been set by
6811 * the crappy stack check or if we picked up the connection
6812 * after establishment)
6813 */
6814 if (src->seqhi == 1 ||
6815 SEQ_GEQ(end + MAX(1, (u_int32_t)dst->max_win << dws),
6816 src->seqhi))
6817 src->seqhi = end + MAX(1, (u_int32_t)dst->max_win << dws);
6818 if (win > src->max_win)
6819 src->max_win = win;
6820
6821 } else {
6822 ack = ntohl(th->th_ack) - dst->seqdiff;
6823 if (src->seqdiff) {
6824 /* Modulate sequence numbers */
6825 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
6826 src->seqdiff), 0);
6827 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
6828 copyback = off+ sizeof (*th);
6829 }
6830 end = seq + pd->p_len;
6831 if (th->th_flags & TH_SYN)
6832 end++;
6833 if (th->th_flags & TH_FIN)
6834 end++;
6835 }
6836
6837 if ((th->th_flags & TH_ACK) == 0) {
6838 /* Let it pass through the ack skew check */
6839 ack = dst->seqlo;
6840 } else if ((ack == 0 &&
6841 (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) ||
6842 /* broken tcp stacks do not set ack */
6843 (dst->state < TCPS_SYN_SENT)) {
6844 /*
6845 * Many stacks (ours included) will set the ACK number in an
6846 * FIN|ACK if the SYN times out -- no sequence to ACK.
6847 */
6848 ack = dst->seqlo;
6849 }
6850
6851 if (seq == end) {
6852 /* Ease sequencing restrictions on no data packets */
6853 seq = src->seqlo;
6854 end = seq;
6855 }
6856
6857 ackskew = dst->seqlo - ack;
6858
6859
6860 /*
6861 * Need to demodulate the sequence numbers in any TCP SACK options
6862 * (Selective ACK). We could optionally validate the SACK values
6863 * against the current ACK window, either forwards or backwards, but
6864 * I'm not confident that SACK has been implemented properly
6865 * everywhere. It wouldn't surprise me if several stacks accidently
6866 * SACK too far backwards of previously ACKed data. There really aren't
6867 * any security implications of bad SACKing unless the target stack
6868 * doesn't validate the option length correctly. Someone trying to
6869 * spoof into a TCP connection won't bother blindly sending SACK
6870 * options anyway.
6871 */
6872 if (dst->seqdiff && (th->th_off << 2) > (int)sizeof (struct tcphdr)) {
6873 copyback = pf_modulate_sack(pbuf, off, pd, th, dst);
6874 if (copyback == -1) {
6875 REASON_SET(reason, PFRES_MEMORY);
6876 return (PF_DROP);
6877 }
6878
6879 pbuf = pd->mp; // XXXSCW: Why?
6880 }
6881
6882
6883#define MAXACKWINDOW (0xffff + 1500) /* 1500 is an arbitrary fudge factor */
6884 if (SEQ_GEQ(src->seqhi, end) &&
6885 /* Last octet inside other's window space */
6886 SEQ_GEQ(seq, src->seqlo - ((u_int32_t)dst->max_win << dws)) &&
6887 /* Retrans: not more than one window back */
6888 (ackskew >= -MAXACKWINDOW) &&
6889 /* Acking not more than one reassembled fragment backwards */
6890 (ackskew <= (MAXACKWINDOW << sws)) &&
6891 /* Acking not more than one window forward */
6892 ((th->th_flags & TH_RST) == 0 || orig_seq == src->seqlo ||
6893 (orig_seq == src->seqlo + 1) || (orig_seq + 1 == src->seqlo) ||
6894 (pd->flags & PFDESC_IP_REAS) == 0)) {
6895 /* Require an exact/+1 sequence match on resets when possible */
6896
6897 if (dst->scrub || src->scrub) {
6898 if (pf_normalize_tcp_stateful(pbuf, off, pd, reason, th,
6899 *state, src, dst, &copyback))
6900 return (PF_DROP);
6901
6902 pbuf = pd->mp; // XXXSCW: Why?
6903 }
6904
6905 /* update max window */
6906 if (src->max_win < win)
6907 src->max_win = win;
6908 /* synchronize sequencing */
6909 if (SEQ_GT(end, src->seqlo))
6910 src->seqlo = end;
6911 /* slide the window of what the other end can send */
6912 if (SEQ_GEQ(ack + ((u_int32_t)win << sws), dst->seqhi))
6913 dst->seqhi = ack + MAX(((u_int32_t)win << sws), 1);
6914
6915 /* update states */
6916 if (th->th_flags & TH_SYN)
6917 if (src->state < TCPS_SYN_SENT)
6918 src->state = TCPS_SYN_SENT;
6919 if (th->th_flags & TH_FIN)
6920 if (src->state < TCPS_CLOSING)
6921 src->state = TCPS_CLOSING;
6922 if (th->th_flags & TH_ACK) {
6923 if (dst->state == TCPS_SYN_SENT) {
6924 dst->state = TCPS_ESTABLISHED;
6925 if (src->state == TCPS_ESTABLISHED &&
6926 (*state)->src_node != NULL &&
6927 pf_src_connlimit(state)) {
6928 REASON_SET(reason, PFRES_SRCLIMIT);
6929 return (PF_DROP);
6930 }
6931 } else if (dst->state == TCPS_CLOSING)
6932 dst->state = TCPS_FIN_WAIT_2;
6933 }
6934 if (th->th_flags & TH_RST)
6935 src->state = dst->state = TCPS_TIME_WAIT;
6936
6937 /* update expire time */
6938 (*state)->expire = pf_time_second();
6939 if (src->state >= TCPS_FIN_WAIT_2 &&
6940 dst->state >= TCPS_FIN_WAIT_2)
6941 (*state)->timeout = PFTM_TCP_CLOSED;
6942 else if (src->state >= TCPS_CLOSING &&
6943 dst->state >= TCPS_CLOSING)
6944 (*state)->timeout = PFTM_TCP_FIN_WAIT;
6945 else if (src->state < TCPS_ESTABLISHED ||
6946 dst->state < TCPS_ESTABLISHED)
6947 (*state)->timeout = PFTM_TCP_OPENING;
6948 else if (src->state >= TCPS_CLOSING ||
6949 dst->state >= TCPS_CLOSING)
6950 (*state)->timeout = PFTM_TCP_CLOSING;
6951 else
6952 (*state)->timeout = PFTM_TCP_ESTABLISHED;
6953
6954 /* Fall through to PASS packet */
6955
6956 } else if ((dst->state < TCPS_SYN_SENT ||
6957 dst->state >= TCPS_FIN_WAIT_2 || src->state >= TCPS_FIN_WAIT_2) &&
6958 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) &&
6959 /* Within a window forward of the originating packet */
6960 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) {
6961 /* Within a window backward of the originating packet */
6962
6963 /*
6964 * This currently handles three situations:
6965 * 1) Stupid stacks will shotgun SYNs before their peer
6966 * replies.
6967 * 2) When PF catches an already established stream (the
6968 * firewall rebooted, the state table was flushed, routes
6969 * changed...)
6970 * 3) Packets get funky immediately after the connection
6971 * closes (this should catch Solaris spurious ACK|FINs
6972 * that web servers like to spew after a close)
6973 *
6974 * This must be a little more careful than the above code
6975 * since packet floods will also be caught here. We don't
6976 * update the TTL here to mitigate the damage of a packet
6977 * flood and so the same code can handle awkward establishment
6978 * and a loosened connection close.
6979 * In the establishment case, a correct peer response will
6980 * validate the connection, go through the normal state code
6981 * and keep updating the state TTL.
6982 */
6983
6984 if (pf_status.debug >= PF_DEBUG_MISC) {
6985 printf("pf: loose state match: ");
6986 pf_print_state(*state);
6987 pf_print_flags(th->th_flags);
6988 printf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
6989 "pkts=%llu:%llu dir=%s,%s\n", seq, orig_seq, ack,
6990 pd->p_len, ackskew, (*state)->packets[0],
6991 (*state)->packets[1],
6992 direction == PF_IN ? "in" : "out",
6993 direction == sk->direction ?
6994 "fwd" : "rev");
6995 }
6996
6997 if (dst->scrub || src->scrub) {
6998 if (pf_normalize_tcp_stateful(pbuf, off, pd, reason, th,
6999 *state, src, dst, &copyback))
7000 return (PF_DROP);
7001 pbuf = pd->mp; // XXXSCW: Why?
7002 }
7003
7004 /* update max window */
7005 if (src->max_win < win)
7006 src->max_win = win;
7007 /* synchronize sequencing */
7008 if (SEQ_GT(end, src->seqlo))
7009 src->seqlo = end;
7010 /* slide the window of what the other end can send */
7011 if (SEQ_GEQ(ack + ((u_int32_t)win << sws), dst->seqhi))
7012 dst->seqhi = ack + MAX(((u_int32_t)win << sws), 1);
7013
7014 /*
7015 * Cannot set dst->seqhi here since this could be a shotgunned
7016 * SYN and not an already established connection.
7017 */
7018
7019 if (th->th_flags & TH_FIN)
7020 if (src->state < TCPS_CLOSING)
7021 src->state = TCPS_CLOSING;
7022 if (th->th_flags & TH_RST)
7023 src->state = dst->state = TCPS_TIME_WAIT;
7024
7025 /* Fall through to PASS packet */
7026
7027 } else {
7028 if (dst->state == TCPS_SYN_SENT &&
7029 src->state == TCPS_SYN_SENT) {
7030 /* Send RST for state mismatches during handshake */
7031 if (!(th->th_flags & TH_RST))
7032 pf_send_tcp((*state)->rule.ptr, pd->af,
7033 pd->dst, pd->src, th->th_dport,
7034 th->th_sport, ntohl(th->th_ack), 0,
7035 TH_RST, 0, 0,
7036 (*state)->rule.ptr->return_ttl, 1, 0,
7037 pd->eh, kif->pfik_ifp);
7038 src->seqlo = 0;
7039 src->seqhi = 1;
7040 src->max_win = 1;
7041 } else if (pf_status.debug >= PF_DEBUG_MISC) {
7042 printf("pf: BAD state: ");
7043 pf_print_state(*state);
7044 pf_print_flags(th->th_flags);
7045 printf("\n seq=%u (%u) ack=%u len=%u ackskew=%d "
7046 "sws=%u dws=%u pkts=%llu:%llu dir=%s,%s\n",
7047 seq, orig_seq, ack, pd->p_len, ackskew,
7048 (unsigned int)sws, (unsigned int)dws,
7049 (*state)->packets[0], (*state)->packets[1],
7050 direction == PF_IN ? "in" : "out",
7051 direction == sk->direction ?
7052 "fwd" : "rev");
7053 printf("pf: State failure on: %c %c %c %c | %c %c\n",
7054 SEQ_GEQ(src->seqhi, end) ? ' ' : '1',
7055 SEQ_GEQ(seq,
7056 src->seqlo - ((u_int32_t)dst->max_win << dws)) ?
7057 ' ': '2',
7058 (ackskew >= -MAXACKWINDOW) ? ' ' : '3',
7059 (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4',
7060 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5',
7061 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6');
7062 }
7063 REASON_SET(reason, PFRES_BADSTATE);
7064 return (PF_DROP);
7065 }
7066
7067 /* Any packets which have gotten here are to be passed */
7068
7069 if (sk->app_state &&
7070 sk->app_state->handler) {
7071 sk->app_state->handler(*state, direction,
7072 off + (th->th_off << 2), pd, kif);
7073 if (pd->lmw < 0) {
7074 REASON_SET(reason, PFRES_MEMORY);
7075 return (PF_DROP);
7076 }
7077 pbuf = pd->mp; // XXXSCW: Why?
7078 }
7079
7080 /* translate source/destination address, if necessary */
7081 if (STATE_TRANSLATE(sk)) {
7082 pd->naf = (pd->af == sk->af_lan) ? sk->af_gwy : sk->af_lan;
7083
7084 if (direction == PF_OUT) {
7085 pf_change_ap(direction, pd->mp, pd->src, &th->th_sport,
7086 pd->ip_sum, &th->th_sum, &sk->gwy.addr,
7087 sk->gwy.xport.port, 0, pd->af, pd->naf, 1);
7088 } else {
7089 if (pd->af != pd->naf) {
7090 if (pd->af == sk->af_gwy) {
7091 pf_change_ap(direction, pd->mp, pd->dst,
7092 &th->th_dport, pd->ip_sum,
7093 &th->th_sum, &sk->lan.addr,
7094 sk->lan.xport.port, 0,
7095 pd->af, pd->naf, 0);
7096
7097 pf_change_ap(direction, pd->mp, pd->src,
7098 &th->th_sport, pd->ip_sum,
7099 &th->th_sum, &sk->ext_lan.addr,
7100 th->th_sport, 0, pd->af,
7101 pd->naf, 0);
7102
7103 } else {
7104 pf_change_ap(direction, pd->mp, pd->dst,
7105 &th->th_dport, pd->ip_sum,
7106 &th->th_sum, &sk->ext_gwy.addr,
7107 th->th_dport, 0, pd->af,
7108 pd->naf, 0);
7109
7110 pf_change_ap(direction, pd->mp, pd->src,
7111 &th->th_sport, pd->ip_sum,
7112 &th->th_sum, &sk->gwy.addr,
7113 sk->gwy.xport.port, 0, pd->af,
7114 pd->naf, 0);
7115 }
7116 } else {
7117 pf_change_ap(direction, pd->mp, pd->dst,
7118 &th->th_dport, pd->ip_sum,
7119 &th->th_sum, &sk->lan.addr,
7120 sk->lan.xport.port, 0, pd->af,
7121 pd->naf, 1);
7122 }
7123 }
7124
7125 copyback = off + sizeof (*th);
7126 }
7127
7128 if (copyback) {
7129 if (pf_lazy_makewritable(pd, pbuf, copyback) == NULL) {
7130 REASON_SET(reason, PFRES_MEMORY);
7131 return (PF_DROP);
7132 }
7133
7134 /* Copyback sequence modulation or stateful scrub changes */
7135 pbuf_copy_back(pbuf, off, sizeof (*th), th);
7136
7137 if (sk->af_lan != sk->af_gwy)
7138 return (pf_do_nat64(sk, pd, pbuf, off));
7139 }
7140 return (PF_PASS);
7141}
7142
7143static int
7144pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif,
7145 pbuf_t *pbuf, int off, void *h, struct pf_pdesc *pd, u_short *reason)
7146{
7147#pragma unused(h)
7148 struct pf_state_peer *src, *dst;
7149 struct pf_state_key_cmp key;
7150 struct pf_state_key *sk;
7151 struct udphdr *uh = pd->hdr.udp;
7152 struct pf_app_state as;
7153 int action, extfilter;
7154 key.app_state = 0;
7155 key.proto_variant = PF_EXTFILTER_APD;
7156
7157 key.proto = IPPROTO_UDP;
7158 key.af_lan = key.af_gwy = pd->af;
7159
7160 /*
7161 * For NAT64 the first time rule search and state creation
7162 * is done on the incoming side only.
7163 * Once the state gets created, NAT64's LAN side (ipv6) will
7164 * not be able to find the state in ext-gwy tree as that normally
7165 * is intended to be looked up for incoming traffic from the
7166 * WAN side.
7167 * Therefore to handle NAT64 case we init keys here for both
7168 * lan-ext as well as ext-gwy trees.
7169 * In the state lookup we attempt a lookup on both trees if
7170 * first one does not return any result and return a match if
7171 * the match state's was created by NAT64 rule.
7172 */
7173 PF_ACPY(&key.ext_gwy.addr, pd->src, key.af_gwy);
7174 PF_ACPY(&key.gwy.addr, pd->dst, key.af_gwy);
7175 key.ext_gwy.xport.port = uh->uh_sport;
7176 key.gwy.xport.port = uh->uh_dport;
7177
7178 PF_ACPY(&key.lan.addr, pd->src, key.af_lan);
7179 PF_ACPY(&key.ext_lan.addr, pd->dst, key.af_lan);
7180 key.lan.xport.port = uh->uh_sport;
7181 key.ext_lan.xport.port = uh->uh_dport;
7182
7183 if (ntohs(uh->uh_sport) == PF_IKE_PORT &&
7184 ntohs(uh->uh_dport) == PF_IKE_PORT) {
7185 struct pf_ike_hdr ike;
7186 size_t plen = pbuf->pb_packet_len - off - sizeof (*uh);
7187 if (plen < PF_IKE_PACKET_MINSIZE) {
7188 DPFPRINTF(PF_DEBUG_MISC,
7189 ("pf: IKE message too small.\n"));
7190 return (PF_DROP);
7191 }
7192
7193 if (plen > sizeof (ike))
7194 plen = sizeof (ike);
7195 pbuf_copy_data(pbuf, off + sizeof (*uh), plen, &ike);
7196
7197 if (ike.initiator_cookie) {
7198 key.app_state = &as;
7199 as.compare_lan_ext = pf_ike_compare;
7200 as.compare_ext_gwy = pf_ike_compare;
7201 as.u.ike.cookie = ike.initiator_cookie;
7202 } else {
7203 /*
7204 * <http://tools.ietf.org/html/\
7205 * draft-ietf-ipsec-nat-t-ike-01>
7206 * Support non-standard NAT-T implementations that
7207 * push the ESP packet over the top of the IKE packet.
7208 * Do not drop packet.
7209 */
7210 DPFPRINTF(PF_DEBUG_MISC,
7211 ("pf: IKE initiator cookie = 0.\n"));
7212 }
7213 }
7214
7215 *state = pf_find_state(kif, &key, direction);
7216
7217 if (!key.app_state && *state == 0) {
7218 key.proto_variant = PF_EXTFILTER_AD;
7219 *state = pf_find_state(kif, &key, direction);
7220 }
7221
7222 if (!key.app_state && *state == 0) {
7223 key.proto_variant = PF_EXTFILTER_EI;
7224 *state = pf_find_state(kif, &key, direction);
7225 }
7226
7227 /* similar to STATE_LOOKUP() */
7228 if (*state != NULL && pd != NULL && !(pd->pktflags & PKTF_FLOW_ID)) {
7229 pd->flowsrc = (*state)->state_key->flowsrc;
7230 pd->flowhash = (*state)->state_key->flowhash;
7231 if (pd->flowhash != 0) {
7232 pd->pktflags |= PKTF_FLOW_ID;
7233 pd->pktflags &= ~PKTF_FLOW_ADV;
7234 }
7235 }
7236
7237 if (pf_state_lookup_aux(state, kif, direction, &action))
7238 return (action);
7239
7240 sk = (*state)->state_key;
7241
7242 /*
7243 * In case of NAT64 the translation is first applied on the LAN
7244 * side. Therefore for stack's address family comparison
7245 * we use sk->af_lan.
7246 */
7247 if ((direction == sk->direction) && (pd->af == sk->af_lan)) {
7248 src = &(*state)->src;
7249 dst = &(*state)->dst;
7250 } else {
7251 src = &(*state)->dst;
7252 dst = &(*state)->src;
7253 }
7254
7255 /* update states */
7256 if (src->state < PFUDPS_SINGLE)
7257 src->state = PFUDPS_SINGLE;
7258 if (dst->state == PFUDPS_SINGLE)
7259 dst->state = PFUDPS_MULTIPLE;
7260
7261 /* update expire time */
7262 (*state)->expire = pf_time_second();
7263 if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE)
7264 (*state)->timeout = PFTM_UDP_MULTIPLE;
7265 else
7266 (*state)->timeout = PFTM_UDP_SINGLE;
7267
7268 extfilter = sk->proto_variant;
7269 if (extfilter > PF_EXTFILTER_APD) {
7270 if (direction == PF_OUT) {
7271 sk->ext_lan.xport.port = key.ext_lan.xport.port;
7272 if (extfilter > PF_EXTFILTER_AD)
7273 PF_ACPY(&sk->ext_lan.addr, &key.ext_lan.addr,
7274 key.af_lan);
7275 } else {
7276 sk->ext_gwy.xport.port = key.ext_gwy.xport.port;
7277 if (extfilter > PF_EXTFILTER_AD)
7278 PF_ACPY(&sk->ext_gwy.addr, &key.ext_gwy.addr,
7279 key.af_gwy);
7280 }
7281 }
7282
7283 if (sk->app_state && sk->app_state->handler) {
7284 sk->app_state->handler(*state, direction, off + uh->uh_ulen,
7285 pd, kif);
7286 if (pd->lmw < 0) {
7287 REASON_SET(reason, PFRES_MEMORY);
7288 return (PF_DROP);
7289 }
7290 pbuf = pd->mp; // XXXSCW: Why?
7291 }
7292
7293 /* translate source/destination address, if necessary */
7294 if (STATE_TRANSLATE(sk)) {
7295 if (pf_lazy_makewritable(pd, pbuf, off + sizeof (*uh)) == NULL) {
7296 REASON_SET(reason, PFRES_MEMORY);
7297 return (PF_DROP);
7298 }
7299
7300 pd->naf = (pd->af == sk->af_lan) ? sk->af_gwy : sk->af_lan;
7301
7302 if (direction == PF_OUT) {
7303 pf_change_ap(direction, pd->mp, pd->src, &uh->uh_sport,
7304 pd->ip_sum, &uh->uh_sum, &sk->gwy.addr,
7305 sk->gwy.xport.port, 1, pd->af, pd->naf, 1);
7306 } else {
7307 if (pd->af != pd->naf) {
7308
7309 if (pd->af == sk->af_gwy) {
7310 pf_change_ap(direction, pd->mp, pd->dst,
7311 &uh->uh_dport, pd->ip_sum,
7312 &uh->uh_sum, &sk->lan.addr,
7313 sk->lan.xport.port, 1,
7314 pd->af, pd->naf, 0);
7315
7316 pf_change_ap(direction, pd->mp, pd->src,
7317 &uh->uh_sport, pd->ip_sum,
7318 &uh->uh_sum, &sk->ext_lan.addr,
7319 uh->uh_sport, 1, pd->af,
7320 pd->naf, 0);
7321
7322 } else {
7323 pf_change_ap(direction, pd->mp, pd->dst,
7324 &uh->uh_dport, pd->ip_sum,
7325 &uh->uh_sum, &sk->ext_gwy.addr,
7326 uh->uh_dport, 1, pd->af,
7327 pd->naf, 0);
7328
7329 pf_change_ap(direction, pd->mp, pd->src,
7330 &uh->uh_sport, pd->ip_sum,
7331 &uh->uh_sum, &sk->gwy.addr,
7332 sk->gwy.xport.port, 1, pd->af,
7333 pd->naf, 0);
7334 }
7335 } else {
7336 pf_change_ap(direction, pd->mp, pd->dst,
7337 &uh->uh_dport, pd->ip_sum,
7338 &uh->uh_sum, &sk->lan.addr,
7339 sk->lan.xport.port, 1,
7340 pd->af, pd->naf, 1);
7341 }
7342 }
7343
7344 pbuf_copy_back(pbuf, off, sizeof (*uh), uh);
7345 if (sk->af_lan != sk->af_gwy)
7346 return (pf_do_nat64(sk, pd, pbuf, off));
7347 }
7348 return (PF_PASS);
7349}
7350
7351static int
7352pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif,
7353 pbuf_t *pbuf, int off, void *h, struct pf_pdesc *pd, u_short *reason)
7354{
7355#pragma unused(h)
7356 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
7357 struct in_addr srcv4_inaddr = saddr->v4addr;
7358 u_int16_t icmpid = 0, *icmpsum = NULL;
7359 u_int8_t icmptype = 0;
7360 int state_icmp = 0;
7361 struct pf_state_key_cmp key;
7362 struct pf_state_key *sk;
7363
7364 struct pf_app_state as;
7365 key.app_state = 0;
7366
7367 pd->off = off;
7368
7369 switch (pd->proto) {
7370#if INET
7371 case IPPROTO_ICMP:
7372 icmptype = pd->hdr.icmp->icmp_type;
7373 icmpid = pd->hdr.icmp->icmp_id;
7374 icmpsum = &pd->hdr.icmp->icmp_cksum;
7375
7376 if (ICMP_ERRORTYPE(icmptype))
7377 state_icmp++;
7378 break;
7379#endif /* INET */
7380#if INET6
7381 case IPPROTO_ICMPV6:
7382 icmptype = pd->hdr.icmp6->icmp6_type;
7383 icmpid = pd->hdr.icmp6->icmp6_id;
7384 icmpsum = &pd->hdr.icmp6->icmp6_cksum;
7385
7386 if (ICMP6_ERRORTYPE(icmptype))
7387 state_icmp++;
7388 break;
7389#endif /* INET6 */
7390 }
7391
7392 if (!state_icmp) {
7393
7394 /*
7395 * ICMP query/reply message not related to a TCP/UDP packet.
7396 * Search for an ICMP state.
7397 */
7398 /*
7399 * NAT64 requires protocol translation between ICMPv4
7400 * and ICMPv6. TCP and UDP do not require protocol
7401 * translation. To avoid adding complexity just to
7402 * handle ICMP(v4addr/v6addr), we always lookup for
7403 * proto = IPPROTO_ICMP on both LAN and WAN side
7404 */
7405 key.proto = IPPROTO_ICMP;
7406 key.af_lan = key.af_gwy = pd->af;
7407
7408 PF_ACPY(&key.ext_gwy.addr, pd->src, key.af_gwy);
7409 PF_ACPY(&key.gwy.addr, pd->dst, key.af_gwy);
7410 key.ext_gwy.xport.port = 0;
7411 key.gwy.xport.port = icmpid;
7412
7413 PF_ACPY(&key.lan.addr, pd->src, key.af_lan);
7414 PF_ACPY(&key.ext_lan.addr, pd->dst, key.af_lan);
7415 key.lan.xport.port = icmpid;
7416 key.ext_lan.xport.port = 0;
7417
7418 STATE_LOOKUP();
7419
7420 sk = (*state)->state_key;
7421 (*state)->expire = pf_time_second();
7422 (*state)->timeout = PFTM_ICMP_ERROR_REPLY;
7423
7424 /* translate source/destination address, if necessary */
7425 if (STATE_TRANSLATE(sk)) {
7426 pd->naf = (pd->af == sk->af_lan) ?
7427 sk->af_gwy : sk->af_lan;
7428 if (direction == PF_OUT) {
7429 switch (pd->af) {
7430#if INET
7431 case AF_INET:
7432 pf_change_a(&saddr->v4addr.s_addr,
7433 pd->ip_sum,
7434 sk->gwy.addr.v4addr.s_addr, 0);
7435 pd->hdr.icmp->icmp_cksum =
7436 pf_cksum_fixup(
7437 pd->hdr.icmp->icmp_cksum, icmpid,
7438 sk->gwy.xport.port, 0);
7439 pd->hdr.icmp->icmp_id =
7440 sk->gwy.xport.port;
7441 if (pf_lazy_makewritable(pd, pbuf,
7442 off + ICMP_MINLEN) == NULL)
7443 return (PF_DROP);
7444 pbuf_copy_back(pbuf, off, ICMP_MINLEN,
7445 pd->hdr.icmp);
7446 break;
7447#endif /* INET */
7448#if INET6
7449 case AF_INET6:
7450 pf_change_a6(saddr,
7451 &pd->hdr.icmp6->icmp6_cksum,
7452 &sk->gwy.addr, 0);
7453 if (pf_lazy_makewritable(pd, NULL,
7454 off + sizeof (struct icmp6_hdr)) ==
7455 NULL)
7456 return (PF_DROP);
7457 pbuf_copy_back(pbuf, off,
7458 sizeof (struct icmp6_hdr),
7459 pd->hdr.icmp6);
7460 break;
7461#endif /* INET6 */
7462 }
7463 } else {
7464 switch (pd->af) {
7465#if INET
7466 case AF_INET:
7467 if (pd->naf != AF_INET) {
7468 if (pf_translate_icmp_af(
7469 AF_INET6, pd->hdr.icmp))
7470 return (PF_DROP);
7471
7472 pd->proto = IPPROTO_ICMPV6;
7473
7474 } else {
7475
7476 pf_change_a(&daddr->v4addr.s_addr,
7477 pd->ip_sum,
7478 sk->lan.addr.v4addr.s_addr, 0);
7479
7480 pd->hdr.icmp->icmp_cksum =
7481 pf_cksum_fixup(
7482 pd->hdr.icmp->icmp_cksum,
7483 icmpid, sk->lan.xport.port, 0);
7484
7485 pd->hdr.icmp->icmp_id =
7486 sk->lan.xport.port;
7487 }
7488
7489 if (pf_lazy_makewritable(pd, pbuf,
7490 off + ICMP_MINLEN) == NULL)
7491 return (PF_DROP);
7492 pbuf_copy_back(pbuf, off, ICMP_MINLEN,
7493 pd->hdr.icmp);
7494 if (sk->af_lan != sk->af_gwy)
7495 return (pf_do_nat64(sk, pd,
7496 pbuf, off));
7497 break;
7498#endif /* INET */
7499#if INET6
7500 case AF_INET6:
7501 if (pd->naf != AF_INET6) {
7502 if (pf_translate_icmp_af(
7503 AF_INET, pd->hdr.icmp6))
7504 return (PF_DROP);
7505
7506 pd->proto = IPPROTO_ICMP;
7507 } else {
7508 pf_change_a6(daddr,
7509 &pd->hdr.icmp6->icmp6_cksum,
7510 &sk->lan.addr, 0);
7511 }
7512 if (pf_lazy_makewritable(pd, pbuf,
7513 off + sizeof (struct icmp6_hdr)) ==
7514 NULL)
7515 return (PF_DROP);
7516 pbuf_copy_back(pbuf, off,
7517 sizeof (struct icmp6_hdr),
7518 pd->hdr.icmp6);
7519 if (sk->af_lan != sk->af_gwy)
7520 return (pf_do_nat64(sk, pd,
7521 pbuf, off));
7522 break;
7523#endif /* INET6 */
7524 }
7525 }
7526 }
7527
7528 return (PF_PASS);
7529
7530 } else {
7531 /*
7532 * ICMP error message in response to a TCP/UDP packet.
7533 * Extract the inner TCP/UDP header and search for that state.
7534 */
7535 struct pf_pdesc pd2; /* For inner (original) header */
7536#if INET
7537 struct ip h2;
7538#endif /* INET */
7539#if INET6
7540 struct ip6_hdr h2_6;
7541 int terminal = 0;
7542#endif /* INET6 */
7543 int ipoff2 = 0;
7544 int off2 = 0;
7545
7546 memset(&pd2, 0, sizeof (pd2));
7547
7548 pd2.af = pd->af;
7549 switch (pd->af) {
7550#if INET
7551 case AF_INET:
7552 /* offset of h2 in mbuf chain */
7553 ipoff2 = off + ICMP_MINLEN;
7554
7555 if (!pf_pull_hdr(pbuf, ipoff2, &h2, sizeof (h2),
7556 NULL, reason, pd2.af)) {
7557 DPFPRINTF(PF_DEBUG_MISC,
7558 ("pf: ICMP error message too short "
7559 "(ip)\n"));
7560 return (PF_DROP);
7561 }
7562 /*
7563 * ICMP error messages don't refer to non-first
7564 * fragments
7565 */
7566 if (h2.ip_off & htons(IP_OFFMASK)) {
7567 REASON_SET(reason, PFRES_FRAG);
7568 return (PF_DROP);
7569 }
7570
7571 /* offset of protocol header that follows h2 */
7572 off2 = ipoff2 + (h2.ip_hl << 2);
7573 /* TODO */
7574 pd2.off = ipoff2 + (h2.ip_hl << 2);
7575
7576 pd2.proto = h2.ip_p;
7577 pd2.src = (struct pf_addr *)&h2.ip_src;
7578 pd2.dst = (struct pf_addr *)&h2.ip_dst;
7579 pd2.ip_sum = &h2.ip_sum;
7580 break;
7581#endif /* INET */
7582#if INET6
7583 case AF_INET6:
7584 ipoff2 = off + sizeof (struct icmp6_hdr);
7585
7586 if (!pf_pull_hdr(pbuf, ipoff2, &h2_6, sizeof (h2_6),
7587 NULL, reason, pd2.af)) {
7588 DPFPRINTF(PF_DEBUG_MISC,
7589 ("pf: ICMP error message too short "
7590 "(ip6)\n"));
7591 return (PF_DROP);
7592 }
7593 pd2.proto = h2_6.ip6_nxt;
7594 pd2.src = (struct pf_addr *)(uintptr_t)&h2_6.ip6_src;
7595 pd2.dst = (struct pf_addr *)(uintptr_t)&h2_6.ip6_dst;
7596 pd2.ip_sum = NULL;
7597 off2 = ipoff2 + sizeof (h2_6);
7598 do {
7599 switch (pd2.proto) {
7600 case IPPROTO_FRAGMENT:
7601 /*
7602 * ICMPv6 error messages for
7603 * non-first fragments
7604 */
7605 REASON_SET(reason, PFRES_FRAG);
7606 return (PF_DROP);
7607 case IPPROTO_AH:
7608 case IPPROTO_HOPOPTS:
7609 case IPPROTO_ROUTING:
7610 case IPPROTO_DSTOPTS: {
7611 /* get next header and header length */
7612 struct ip6_ext opt6;
7613
7614 if (!pf_pull_hdr(pbuf, off2, &opt6,
7615 sizeof (opt6), NULL, reason,
7616 pd2.af)) {
7617 DPFPRINTF(PF_DEBUG_MISC,
7618 ("pf: ICMPv6 short opt\n"));
7619 return (PF_DROP);
7620 }
7621 if (pd2.proto == IPPROTO_AH)
7622 off2 += (opt6.ip6e_len + 2) * 4;
7623 else
7624 off2 += (opt6.ip6e_len + 1) * 8;
7625 pd2.proto = opt6.ip6e_nxt;
7626 /* goto the next header */
7627 break;
7628 }
7629 default:
7630 terminal++;
7631 break;
7632 }
7633 } while (!terminal);
7634 /* TODO */
7635 pd2.off = ipoff2;
7636 break;
7637#endif /* INET6 */
7638 }
7639
7640 switch (pd2.proto) {
7641 case IPPROTO_TCP: {
7642 struct tcphdr th;
7643 u_int32_t seq;
7644 struct pf_state_peer *src, *dst;
7645 u_int8_t dws;
7646 int copyback = 0;
7647
7648 /*
7649 * Only the first 8 bytes of the TCP header can be
7650 * expected. Don't access any TCP header fields after
7651 * th_seq, an ackskew test is not possible.
7652 */
7653 if (!pf_pull_hdr(pbuf, off2, &th, 8, NULL, reason,
7654 pd2.af)) {
7655 DPFPRINTF(PF_DEBUG_MISC,
7656 ("pf: ICMP error message too short "
7657 "(tcp)\n"));
7658 return (PF_DROP);
7659 }
7660
7661 key.proto = IPPROTO_TCP;
7662 key.af_gwy = pd2.af;
7663 PF_ACPY(&key.ext_gwy.addr, pd2.dst, key.af_gwy);
7664 PF_ACPY(&key.gwy.addr, pd2.src, key.af_gwy);
7665 key.ext_gwy.xport.port = th.th_dport;
7666 key.gwy.xport.port = th.th_sport;
7667
7668 key.af_lan = pd2.af;
7669 PF_ACPY(&key.lan.addr, pd2.dst, key.af_lan);
7670 PF_ACPY(&key.ext_lan.addr, pd2.src, key.af_lan);
7671 key.lan.xport.port = th.th_dport;
7672 key.ext_lan.xport.port = th.th_sport;
7673
7674 STATE_LOOKUP();
7675
7676 sk = (*state)->state_key;
7677 if ((direction == sk->direction) &&
7678 ((sk->af_lan == sk->af_gwy) ||
7679 (pd2.af == sk->af_lan))) {
7680 src = &(*state)->dst;
7681 dst = &(*state)->src;
7682 } else {
7683 src = &(*state)->src;
7684 dst = &(*state)->dst;
7685 }
7686
7687 if (src->wscale && (dst->wscale & PF_WSCALE_FLAG))
7688 dws = dst->wscale & PF_WSCALE_MASK;
7689 else
7690 dws = TCP_MAX_WINSHIFT;
7691
7692 /* Demodulate sequence number */
7693 seq = ntohl(th.th_seq) - src->seqdiff;
7694 if (src->seqdiff) {
7695 pf_change_a(&th.th_seq, icmpsum,
7696 htonl(seq), 0);
7697 copyback = 1;
7698 }
7699
7700 if (!SEQ_GEQ(src->seqhi, seq) ||
7701 !SEQ_GEQ(seq,
7702 src->seqlo - ((u_int32_t)dst->max_win << dws))) {
7703 if (pf_status.debug >= PF_DEBUG_MISC) {
7704 printf("pf: BAD ICMP %d:%d ",
7705 icmptype, pd->hdr.icmp->icmp_code);
7706 pf_print_host(pd->src, 0, pd->af);
7707 printf(" -> ");
7708 pf_print_host(pd->dst, 0, pd->af);
7709 printf(" state: ");
7710 pf_print_state(*state);
7711 printf(" seq=%u\n", seq);
7712 }
7713 REASON_SET(reason, PFRES_BADSTATE);
7714 return (PF_DROP);
7715 }
7716
7717 pd->naf = pd2.naf = (pd2.af == sk->af_lan) ?
7718 sk->af_gwy : sk->af_lan;
7719
7720 if (STATE_TRANSLATE(sk)) {
7721 /* NAT64 case */
7722 if (sk->af_lan != sk->af_gwy) {
7723 struct pf_state_host *saddr2, *daddr2;
7724
7725 if (pd2.naf == sk->af_lan) {
7726 saddr2 = &sk->lan;
7727 daddr2 = &sk->ext_lan;
7728 } else {
7729 saddr2 = &sk->ext_gwy;
7730 daddr2 = &sk->gwy;
7731 }
7732
7733 /* translate ICMP message types and codes */
7734 if (pf_translate_icmp_af(pd->naf,
7735 pd->hdr.icmp))
7736 return (PF_DROP);
7737
7738 if (pf_lazy_makewritable(pd, pbuf,
7739 off2 + 8) == NULL)
7740 return (PF_DROP);
7741
7742 pbuf_copy_back(pbuf, pd->off,
7743 sizeof(struct icmp6_hdr),
7744 pd->hdr.icmp6);
7745
7746 /*
7747 * translate inner ip header within the
7748 * ICMP message
7749 */
7750 if (pf_change_icmp_af(pbuf, ipoff2, pd,
7751 &pd2, &saddr2->addr, &daddr2->addr,
7752 pd->af, pd->naf))
7753 return (PF_DROP);
7754
7755 if (pd->naf == AF_INET)
7756 pd->proto = IPPROTO_ICMP;
7757 else
7758 pd->proto = IPPROTO_ICMPV6;
7759
7760 /*
7761 * translate inner tcp header within
7762 * the ICMP message
7763 */
7764 pf_change_ap(direction, NULL, pd2.src,
7765 &th.th_sport, pd2.ip_sum,
7766 &th.th_sum, &daddr2->addr,
7767 saddr2->xport.port, 0, pd2.af,
7768 pd2.naf, 0);
7769
7770 pf_change_ap(direction, NULL, pd2.dst,
7771 &th.th_dport, pd2.ip_sum,
7772 &th.th_sum, &saddr2->addr,
7773 daddr2->xport.port, 0, pd2.af,
7774 pd2.naf, 0);
7775
7776 pbuf_copy_back(pbuf, pd2.off, 8, &th);
7777
7778 /* translate outer ip header */
7779 PF_ACPY(&pd->naddr, &daddr2->addr,
7780 pd->naf);
7781 PF_ACPY(&pd->ndaddr, &saddr2->addr,
7782 pd->naf);
7783 if (pd->af == AF_INET) {
7784 memcpy(&pd->naddr.addr32[3],
7785 &srcv4_inaddr,
7786 sizeof(pd->naddr.addr32[3]));
7787 return (pf_nat64_ipv4(pbuf, off,
7788 pd));
7789 } else {
7790 return (pf_nat64_ipv6(pbuf, off,
7791 pd));
7792 }
7793 }
7794 if (direction == PF_IN) {
7795 pf_change_icmp(pd2.src, &th.th_sport,
7796 daddr, &sk->lan.addr,
7797 sk->lan.xport.port, NULL,
7798 pd2.ip_sum, icmpsum,
7799 pd->ip_sum, 0, pd2.af);
7800 } else {
7801 pf_change_icmp(pd2.dst, &th.th_dport,
7802 saddr, &sk->gwy.addr,
7803 sk->gwy.xport.port, NULL,
7804 pd2.ip_sum, icmpsum,
7805 pd->ip_sum, 0, pd2.af);
7806 }
7807 copyback = 1;
7808 }
7809
7810 if (copyback) {
7811 if (pf_lazy_makewritable(pd, pbuf, off2 + 8) ==
7812 NULL)
7813 return (PF_DROP);
7814 switch (pd2.af) {
7815#if INET
7816 case AF_INET:
7817 pbuf_copy_back(pbuf, off, ICMP_MINLEN,
7818 pd->hdr.icmp);
7819 pbuf_copy_back(pbuf, ipoff2, sizeof(h2),
7820 &h2);
7821 break;
7822#endif /* INET */
7823#if INET6
7824 case AF_INET6:
7825 pbuf_copy_back(pbuf, off,
7826 sizeof (struct icmp6_hdr),
7827 pd->hdr.icmp6);
7828 pbuf_copy_back(pbuf, ipoff2,
7829 sizeof (h2_6), &h2_6);
7830 break;
7831#endif /* INET6 */
7832 }
7833 pbuf_copy_back(pbuf, off2, 8, &th);
7834 }
7835
7836 return (PF_PASS);
7837 }
7838 case IPPROTO_UDP: {
7839 struct udphdr uh;
7840 int dx, action;
7841 if (!pf_pull_hdr(pbuf, off2, &uh, sizeof (uh),
7842 NULL, reason, pd2.af)) {
7843 DPFPRINTF(PF_DEBUG_MISC,
7844 ("pf: ICMP error message too short "
7845 "(udp)\n"));
7846 return (PF_DROP);
7847 }
7848
7849 key.af_gwy = pd2.af;
7850 PF_ACPY(&key.ext_gwy.addr, pd2.dst, key.af_gwy);
7851 PF_ACPY(&key.gwy.addr, pd2.src, key.af_gwy);
7852 key.ext_gwy.xport.port = uh.uh_dport;
7853 key.gwy.xport.port = uh.uh_sport;
7854
7855 key.af_lan = pd2.af;
7856 PF_ACPY(&key.lan.addr, pd2.dst, key.af_lan);
7857 PF_ACPY(&key.ext_lan.addr, pd2.src, key.af_lan);
7858 key.lan.xport.port = uh.uh_dport;
7859 key.ext_lan.xport.port = uh.uh_sport;
7860
7861 key.proto = IPPROTO_UDP;
7862 key.proto_variant = PF_EXTFILTER_APD;
7863 dx = direction;
7864
7865 if (ntohs(uh.uh_sport) == PF_IKE_PORT &&
7866 ntohs(uh.uh_dport) == PF_IKE_PORT) {
7867 struct pf_ike_hdr ike;
7868 size_t plen = pbuf->pb_packet_len - off2 -
7869 sizeof (uh);
7870 if (direction == PF_IN &&
7871 plen < 8 /* PF_IKE_PACKET_MINSIZE */) {
7872 DPFPRINTF(PF_DEBUG_MISC, ("pf: "
7873 "ICMP error, embedded IKE message "
7874 "too small.\n"));
7875 return (PF_DROP);
7876 }
7877
7878 if (plen > sizeof (ike))
7879 plen = sizeof (ike);
7880 pbuf_copy_data(pbuf, off + sizeof (uh), plen,
7881 &ike);
7882
7883 key.app_state = &as;
7884 as.compare_lan_ext = pf_ike_compare;
7885 as.compare_ext_gwy = pf_ike_compare;
7886 as.u.ike.cookie = ike.initiator_cookie;
7887 }
7888
7889 *state = pf_find_state(kif, &key, dx);
7890
7891 if (key.app_state && *state == 0) {
7892 key.app_state = 0;
7893 *state = pf_find_state(kif, &key, dx);
7894 }
7895
7896 if (*state == 0) {
7897 key.proto_variant = PF_EXTFILTER_AD;
7898 *state = pf_find_state(kif, &key, dx);
7899 }
7900
7901 if (*state == 0) {
7902 key.proto_variant = PF_EXTFILTER_EI;
7903 *state = pf_find_state(kif, &key, dx);
7904 }
7905
7906 /* similar to STATE_LOOKUP() */
7907 if (*state != NULL && pd != NULL &&
7908 !(pd->pktflags & PKTF_FLOW_ID)) {
7909 pd->flowsrc = (*state)->state_key->flowsrc;
7910 pd->flowhash = (*state)->state_key->flowhash;
7911 if (pd->flowhash != 0) {
7912 pd->pktflags |= PKTF_FLOW_ID;
7913 pd->pktflags &= ~PKTF_FLOW_ADV;
7914 }
7915 }
7916
7917 if (pf_state_lookup_aux(state, kif, direction, &action))
7918 return (action);
7919
7920 sk = (*state)->state_key;
7921 pd->naf = pd2.naf = (pd2.af == sk->af_lan) ?
7922 sk->af_gwy : sk->af_lan;
7923
7924 if (STATE_TRANSLATE(sk)) {
7925 /* NAT64 case */
7926 if (sk->af_lan != sk->af_gwy) {
7927 struct pf_state_host *saddr2, *daddr2;
7928
7929 if (pd2.naf == sk->af_lan) {
7930 saddr2 = &sk->lan;
7931 daddr2 = &sk->ext_lan;
7932 } else {
7933 saddr2 = &sk->ext_gwy;
7934 daddr2 = &sk->gwy;
7935 }
7936
7937 /* translate ICMP message */
7938 if (pf_translate_icmp_af(pd->naf,
7939 pd->hdr.icmp))
7940 return (PF_DROP);
7941 if (pf_lazy_makewritable(pd, pbuf,
7942 off2 + 8) == NULL)
7943 return (PF_DROP);
7944
7945 pbuf_copy_back(pbuf, pd->off,
7946 sizeof(struct icmp6_hdr),
7947 pd->hdr.icmp6);
7948
7949 /*
7950 * translate inner ip header within the
7951 * ICMP message
7952 */
7953 if (pf_change_icmp_af(pbuf, ipoff2, pd,
7954 &pd2, &saddr2->addr, &daddr2->addr,
7955 pd->af, pd->naf))
7956 return (PF_DROP);
7957
7958 if (pd->naf == AF_INET)
7959 pd->proto = IPPROTO_ICMP;
7960 else
7961 pd->proto = IPPROTO_ICMPV6;
7962
7963 /*
7964 * translate inner udp header within
7965 * the ICMP message
7966 */
7967 pf_change_ap(direction, NULL, pd2.src,
7968 &uh.uh_sport, pd2.ip_sum,
7969 &uh.uh_sum, &daddr2->addr,
7970 saddr2->xport.port, 0, pd2.af,
7971 pd2.naf, 0);
7972
7973 pf_change_ap(direction, NULL, pd2.dst,
7974 &uh.uh_dport, pd2.ip_sum,
7975 &uh.uh_sum, &saddr2->addr,
7976 daddr2->xport.port, 0, pd2.af,
7977 pd2.naf, 0);
7978
7979 pbuf_copy_back(pbuf, pd2.off,
7980 sizeof(uh), &uh);
7981
7982 /* translate outer ip header */
7983 PF_ACPY(&pd->naddr, &daddr2->addr,
7984 pd->naf);
7985 PF_ACPY(&pd->ndaddr, &saddr2->addr,
7986 pd->naf);
7987 if (pd->af == AF_INET) {
7988 memcpy(&pd->naddr.addr32[3],
7989 &srcv4_inaddr,
7990 sizeof(pd->naddr.addr32[3]));
7991 return (pf_nat64_ipv4(pbuf, off,
7992 pd));
7993 } else {
7994 return (pf_nat64_ipv6(pbuf, off,
7995 pd));
7996 }
7997 }
7998 if (direction == PF_IN) {
7999 pf_change_icmp(pd2.src, &uh.uh_sport,
8000 daddr, &sk->lan.addr,
8001 sk->lan.xport.port, &uh.uh_sum,
8002 pd2.ip_sum, icmpsum,
8003 pd->ip_sum, 1, pd2.af);
8004 } else {
8005 pf_change_icmp(pd2.dst, &uh.uh_dport,
8006 saddr, &sk->gwy.addr,
8007 sk->gwy.xport.port, &uh.uh_sum,
8008 pd2.ip_sum, icmpsum,
8009 pd->ip_sum, 1, pd2.af);
8010 }
8011 if (pf_lazy_makewritable(pd, pbuf,
8012 off2 + sizeof (uh)) == NULL)
8013 return (PF_DROP);
8014 switch (pd2.af) {
8015#if INET
8016 case AF_INET:
8017 pbuf_copy_back(pbuf, off, ICMP_MINLEN,
8018 pd->hdr.icmp);
8019 pbuf_copy_back(pbuf, ipoff2,
8020 sizeof (h2), &h2);
8021 break;
8022#endif /* INET */
8023#if INET6
8024 case AF_INET6:
8025 pbuf_copy_back(pbuf, off,
8026 sizeof (struct icmp6_hdr),
8027 pd->hdr.icmp6);
8028 pbuf_copy_back(pbuf, ipoff2,
8029 sizeof (h2_6), &h2_6);
8030 break;
8031#endif /* INET6 */
8032 }
8033 pbuf_copy_back(pbuf, off2, sizeof (uh), &uh);
8034 }
8035
8036 return (PF_PASS);
8037 }
8038#if INET
8039 case IPPROTO_ICMP: {
8040 struct icmp iih;
8041
8042 if (!pf_pull_hdr(pbuf, off2, &iih, ICMP_MINLEN,
8043 NULL, reason, pd2.af)) {
8044 DPFPRINTF(PF_DEBUG_MISC,
8045 ("pf: ICMP error message too short i"
8046 "(icmp)\n"));
8047 return (PF_DROP);
8048 }
8049
8050 key.proto = IPPROTO_ICMP;
8051 if (direction == PF_IN) {
8052 key.af_gwy = pd2.af;
8053 PF_ACPY(&key.ext_gwy.addr, pd2.dst, key.af_gwy);
8054 PF_ACPY(&key.gwy.addr, pd2.src, key.af_gwy);
8055 key.ext_gwy.xport.port = 0;
8056 key.gwy.xport.port = iih.icmp_id;
8057 } else {
8058 key.af_lan = pd2.af;
8059 PF_ACPY(&key.lan.addr, pd2.dst, key.af_lan);
8060 PF_ACPY(&key.ext_lan.addr, pd2.src, key.af_lan);
8061 key.lan.xport.port = iih.icmp_id;
8062 key.ext_lan.xport.port = 0;
8063 }
8064
8065 STATE_LOOKUP();
8066
8067 sk = (*state)->state_key;
8068 if (STATE_TRANSLATE(sk)) {
8069 if (direction == PF_IN) {
8070 pf_change_icmp(pd2.src, &iih.icmp_id,
8071 daddr, &sk->lan.addr,
8072 sk->lan.xport.port, NULL,
8073 pd2.ip_sum, icmpsum,
8074 pd->ip_sum, 0, AF_INET);
8075 } else {
8076 pf_change_icmp(pd2.dst, &iih.icmp_id,
8077 saddr, &sk->gwy.addr,
8078 sk->gwy.xport.port, NULL,
8079 pd2.ip_sum, icmpsum,
8080 pd->ip_sum, 0, AF_INET);
8081 }
8082 if (pf_lazy_makewritable(pd, pbuf,
8083 off2 + ICMP_MINLEN) == NULL)
8084 return (PF_DROP);
8085 pbuf_copy_back(pbuf, off, ICMP_MINLEN,
8086 pd->hdr.icmp);
8087 pbuf_copy_back(pbuf, ipoff2, sizeof (h2), &h2);
8088 pbuf_copy_back(pbuf, off2, ICMP_MINLEN, &iih);
8089 }
8090
8091 return (PF_PASS);
8092 }
8093#endif /* INET */
8094#if INET6
8095 case IPPROTO_ICMPV6: {
8096 struct icmp6_hdr iih;
8097
8098 if (!pf_pull_hdr(pbuf, off2, &iih,
8099 sizeof (struct icmp6_hdr), NULL, reason, pd2.af)) {
8100 DPFPRINTF(PF_DEBUG_MISC,
8101 ("pf: ICMP error message too short "
8102 "(icmp6)\n"));
8103 return (PF_DROP);
8104 }
8105
8106 key.proto = IPPROTO_ICMPV6;
8107 if (direction == PF_IN) {
8108 key.af_gwy = pd2.af;
8109 PF_ACPY(&key.ext_gwy.addr, pd2.dst, key.af_gwy);
8110 PF_ACPY(&key.gwy.addr, pd2.src, key.af_gwy);
8111 key.ext_gwy.xport.port = 0;
8112 key.gwy.xport.port = iih.icmp6_id;
8113 } else {
8114 key.af_lan = pd2.af;
8115 PF_ACPY(&key.lan.addr, pd2.dst, key.af_lan);
8116 PF_ACPY(&key.ext_lan.addr, pd2.src, key.af_lan);
8117 key.lan.xport.port = iih.icmp6_id;
8118 key.ext_lan.xport.port = 0;
8119 }
8120
8121 STATE_LOOKUP();
8122
8123 sk = (*state)->state_key;
8124 if (STATE_TRANSLATE(sk)) {
8125 if (direction == PF_IN) {
8126 pf_change_icmp(pd2.src, &iih.icmp6_id,
8127 daddr, &sk->lan.addr,
8128 sk->lan.xport.port, NULL,
8129 pd2.ip_sum, icmpsum,
8130 pd->ip_sum, 0, AF_INET6);
8131 } else {
8132 pf_change_icmp(pd2.dst, &iih.icmp6_id,
8133 saddr, &sk->gwy.addr,
8134 sk->gwy.xport.port, NULL,
8135 pd2.ip_sum, icmpsum,
8136 pd->ip_sum, 0, AF_INET6);
8137 }
8138 if (pf_lazy_makewritable(pd, pbuf, off2 +
8139 sizeof (struct icmp6_hdr)) == NULL)
8140 return (PF_DROP);
8141 pbuf_copy_back(pbuf, off,
8142 sizeof (struct icmp6_hdr), pd->hdr.icmp6);
8143 pbuf_copy_back(pbuf, ipoff2, sizeof (h2_6),
8144 &h2_6);
8145 pbuf_copy_back(pbuf, off2,
8146 sizeof (struct icmp6_hdr), &iih);
8147 }
8148
8149 return (PF_PASS);
8150 }
8151#endif /* INET6 */
8152 default: {
8153 key.proto = pd2.proto;
8154 if (direction == PF_IN) {
8155 key.af_gwy = pd2.af;
8156 PF_ACPY(&key.ext_gwy.addr, pd2.dst, key.af_gwy);
8157 PF_ACPY(&key.gwy.addr, pd2.src, key.af_gwy);
8158 key.ext_gwy.xport.port = 0;
8159 key.gwy.xport.port = 0;
8160 } else {
8161 key.af_lan = pd2.af;
8162 PF_ACPY(&key.lan.addr, pd2.dst, key.af_lan);
8163 PF_ACPY(&key.ext_lan.addr, pd2.src, key.af_lan);
8164 key.lan.xport.port = 0;
8165 key.ext_lan.xport.port = 0;
8166 }
8167
8168 STATE_LOOKUP();
8169
8170 sk = (*state)->state_key;
8171 if (STATE_TRANSLATE(sk)) {
8172 if (direction == PF_IN) {
8173 pf_change_icmp(pd2.src, NULL, daddr,
8174 &sk->lan.addr, 0, NULL,
8175 pd2.ip_sum, icmpsum,
8176 pd->ip_sum, 0, pd2.af);
8177 } else {
8178 pf_change_icmp(pd2.dst, NULL, saddr,
8179 &sk->gwy.addr, 0, NULL,
8180 pd2.ip_sum, icmpsum,
8181 pd->ip_sum, 0, pd2.af);
8182 }
8183 switch (pd2.af) {
8184#if INET
8185 case AF_INET:
8186 if (pf_lazy_makewritable(pd, pbuf,
8187 ipoff2 + sizeof (h2)) == NULL)
8188 return (PF_DROP);
8189 /*
8190 * <XXXSCW>
8191 * Xnu was missing the following...
8192 */
8193 pbuf_copy_back(pbuf, off, ICMP_MINLEN,
8194 pd->hdr.icmp);
8195 pbuf_copy_back(pbuf, ipoff2,
8196 sizeof(h2), &h2);
8197 break;
8198 /*
8199 * </XXXSCW>
8200 */
8201#endif /* INET */
8202#if INET6
8203 case AF_INET6:
8204 if (pf_lazy_makewritable(pd, pbuf,
8205 ipoff2 + sizeof (h2_6)) == NULL)
8206 return (PF_DROP);
8207 pbuf_copy_back(pbuf, off,
8208 sizeof (struct icmp6_hdr),
8209 pd->hdr.icmp6);
8210 pbuf_copy_back(pbuf, ipoff2,
8211 sizeof (h2_6), &h2_6);
8212 break;
8213#endif /* INET6 */
8214 }
8215 }
8216
8217 return (PF_PASS);
8218 }
8219 }
8220 }
8221}
8222
8223static int
8224pf_test_state_grev1(struct pf_state **state, int direction,
8225 struct pfi_kif *kif, int off, struct pf_pdesc *pd)
8226{
8227 struct pf_state_peer *src;
8228 struct pf_state_peer *dst;
8229 struct pf_state_key_cmp key;
8230 struct pf_grev1_hdr *grev1 = pd->hdr.grev1;
8231
8232 key.app_state = 0;
8233 key.proto = IPPROTO_GRE;
8234 key.proto_variant = PF_GRE_PPTP_VARIANT;
8235 if (direction == PF_IN) {
8236 key.af_gwy = pd->af;
8237 PF_ACPY(&key.ext_gwy.addr, pd->src, key.af_gwy);
8238 PF_ACPY(&key.gwy.addr, pd->dst, key.af_gwy);
8239 key.gwy.xport.call_id = grev1->call_id;
8240 } else {
8241 key.af_lan = pd->af;
8242 PF_ACPY(&key.lan.addr, pd->src, key.af_lan);
8243 PF_ACPY(&key.ext_lan.addr, pd->dst, key.af_lan);
8244 key.ext_lan.xport.call_id = grev1->call_id;
8245 }
8246
8247 STATE_LOOKUP();
8248
8249 if (direction == (*state)->state_key->direction) {
8250 src = &(*state)->src;
8251 dst = &(*state)->dst;
8252 } else {
8253 src = &(*state)->dst;
8254 dst = &(*state)->src;
8255 }
8256
8257 /* update states */
8258 if (src->state < PFGRE1S_INITIATING)
8259 src->state = PFGRE1S_INITIATING;
8260
8261 /* update expire time */
8262 (*state)->expire = pf_time_second();
8263 if (src->state >= PFGRE1S_INITIATING &&
8264 dst->state >= PFGRE1S_INITIATING) {
8265 if ((*state)->timeout != PFTM_TCP_ESTABLISHED)
8266 (*state)->timeout = PFTM_GREv1_ESTABLISHED;
8267 src->state = PFGRE1S_ESTABLISHED;
8268 dst->state = PFGRE1S_ESTABLISHED;
8269 } else {
8270 (*state)->timeout = PFTM_GREv1_INITIATING;
8271 }
8272
8273 if ((*state)->state_key->app_state)
8274 (*state)->state_key->app_state->u.grev1.pptp_state->expire =
8275 pf_time_second();
8276
8277 /* translate source/destination address, if necessary */
8278 if (STATE_GRE_TRANSLATE((*state)->state_key)) {
8279 if (direction == PF_OUT) {
8280 switch (pd->af) {
8281#if INET
8282 case AF_INET:
8283 pf_change_a(&pd->src->v4addr.s_addr,
8284 pd->ip_sum,
8285 (*state)->state_key->gwy.addr.v4addr.s_addr, 0);
8286 break;
8287#endif /* INET */
8288#if INET6
8289 case AF_INET6:
8290 PF_ACPY(pd->src, &(*state)->state_key->gwy.addr,
8291 pd->af);
8292 break;
8293#endif /* INET6 */
8294 }
8295 } else {
8296 grev1->call_id = (*state)->state_key->lan.xport.call_id;
8297
8298 switch (pd->af) {
8299#if INET
8300 case AF_INET:
8301 pf_change_a(&pd->dst->v4addr.s_addr,
8302 pd->ip_sum,
8303 (*state)->state_key->lan.addr.v4addr.s_addr, 0);
8304 break;
8305#endif /* INET */
8306#if INET6
8307 case AF_INET6:
8308 PF_ACPY(pd->dst, &(*state)->state_key->lan.addr,
8309 pd->af);
8310 break;
8311#endif /* INET6 */
8312 }
8313 }
8314
8315 if (pf_lazy_makewritable(pd, pd->mp, off + sizeof (*grev1)) ==
8316 NULL)
8317 return (PF_DROP);
8318 pbuf_copy_back(pd->mp, off, sizeof (*grev1), grev1);
8319 }
8320
8321 return (PF_PASS);
8322}
8323
8324static int
8325pf_test_state_esp(struct pf_state **state, int direction, struct pfi_kif *kif,
8326 int off, struct pf_pdesc *pd)
8327{
8328#pragma unused(off)
8329 struct pf_state_peer *src;
8330 struct pf_state_peer *dst;
8331 struct pf_state_key_cmp key;
8332 struct pf_esp_hdr *esp = pd->hdr.esp;
8333 int action;
8334
8335 memset(&key, 0, sizeof (key));
8336 key.proto = IPPROTO_ESP;
8337 if (direction == PF_IN) {
8338 key.af_gwy = pd->af;
8339 PF_ACPY(&key.ext_gwy.addr, pd->src, key.af_gwy);
8340 PF_ACPY(&key.gwy.addr, pd->dst, key.af_gwy);
8341 key.gwy.xport.spi = esp->spi;
8342 } else {
8343 key.af_lan = pd->af;
8344 PF_ACPY(&key.lan.addr, pd->src, key.af_lan);
8345 PF_ACPY(&key.ext_lan.addr, pd->dst, key.af_lan);
8346 key.ext_lan.xport.spi = esp->spi;
8347 }
8348
8349 *state = pf_find_state(kif, &key, direction);
8350
8351 if (*state == 0) {
8352 struct pf_state *s;
8353
8354 /*
8355 * <jhw@apple.com>
8356 * No matching state. Look for a blocking state. If we find
8357 * one, then use that state and move it so that it's keyed to
8358 * the SPI in the current packet.
8359 */
8360 if (direction == PF_IN) {
8361 key.gwy.xport.spi = 0;
8362
8363 s = pf_find_state(kif, &key, direction);
8364 if (s) {
8365 struct pf_state_key *sk = s->state_key;
8366
8367 RB_REMOVE(pf_state_tree_ext_gwy,
8368 &pf_statetbl_ext_gwy, sk);
8369 sk->lan.xport.spi = sk->gwy.xport.spi =
8370 esp->spi;
8371
8372 if (RB_INSERT(pf_state_tree_ext_gwy,
8373 &pf_statetbl_ext_gwy, sk))
8374 pf_detach_state(s, PF_DT_SKIP_EXTGWY);
8375 else
8376 *state = s;
8377 }
8378 } else {
8379 key.ext_lan.xport.spi = 0;
8380
8381 s = pf_find_state(kif, &key, direction);
8382 if (s) {
8383 struct pf_state_key *sk = s->state_key;
8384
8385 RB_REMOVE(pf_state_tree_lan_ext,
8386 &pf_statetbl_lan_ext, sk);
8387 sk->ext_lan.xport.spi = esp->spi;
8388
8389 if (RB_INSERT(pf_state_tree_lan_ext,
8390 &pf_statetbl_lan_ext, sk))
8391 pf_detach_state(s, PF_DT_SKIP_LANEXT);
8392 else
8393 *state = s;
8394 }
8395 }
8396
8397 if (s) {
8398 if (*state == 0) {
8399#if NPFSYNC
8400 if (s->creatorid == pf_status.hostid)
8401 pfsync_delete_state(s);
8402#endif
8403 s->timeout = PFTM_UNLINKED;
8404 hook_runloop(&s->unlink_hooks,
8405 HOOK_REMOVE|HOOK_FREE);
8406 pf_src_tree_remove_state(s);
8407 pf_free_state(s);
8408 return (PF_DROP);
8409 }
8410 }
8411 }
8412
8413 /* similar to STATE_LOOKUP() */
8414 if (*state != NULL && pd != NULL && !(pd->pktflags & PKTF_FLOW_ID)) {
8415 pd->flowsrc = (*state)->state_key->flowsrc;
8416 pd->flowhash = (*state)->state_key->flowhash;
8417 if (pd->flowhash != 0) {
8418 pd->pktflags |= PKTF_FLOW_ID;
8419 pd->pktflags &= ~PKTF_FLOW_ADV;
8420 }
8421 }
8422
8423 if (pf_state_lookup_aux(state, kif, direction, &action))
8424 return (action);
8425
8426 if (direction == (*state)->state_key->direction) {
8427 src = &(*state)->src;
8428 dst = &(*state)->dst;
8429 } else {
8430 src = &(*state)->dst;
8431 dst = &(*state)->src;
8432 }
8433
8434 /* update states */
8435 if (src->state < PFESPS_INITIATING)
8436 src->state = PFESPS_INITIATING;
8437
8438 /* update expire time */
8439 (*state)->expire = pf_time_second();
8440 if (src->state >= PFESPS_INITIATING &&
8441 dst->state >= PFESPS_INITIATING) {
8442 (*state)->timeout = PFTM_ESP_ESTABLISHED;
8443 src->state = PFESPS_ESTABLISHED;
8444 dst->state = PFESPS_ESTABLISHED;
8445 } else {
8446 (*state)->timeout = PFTM_ESP_INITIATING;
8447 }
8448 /* translate source/destination address, if necessary */
8449 if (STATE_ADDR_TRANSLATE((*state)->state_key)) {
8450 if (direction == PF_OUT) {
8451 switch (pd->af) {
8452#if INET
8453 case AF_INET:
8454 pf_change_a(&pd->src->v4addr.s_addr,
8455 pd->ip_sum,
8456 (*state)->state_key->gwy.addr.v4addr.s_addr, 0);
8457 break;
8458#endif /* INET */
8459#if INET6
8460 case AF_INET6:
8461 PF_ACPY(pd->src, &(*state)->state_key->gwy.addr,
8462 pd->af);
8463 break;
8464#endif /* INET6 */
8465 }
8466 } else {
8467 switch (pd->af) {
8468#if INET
8469 case AF_INET:
8470 pf_change_a(&pd->dst->v4addr.s_addr,
8471 pd->ip_sum,
8472 (*state)->state_key->lan.addr.v4addr.s_addr, 0);
8473 break;
8474#endif /* INET */
8475#if INET6
8476 case AF_INET6:
8477 PF_ACPY(pd->dst, &(*state)->state_key->lan.addr,
8478 pd->af);
8479 break;
8480#endif /* INET6 */
8481 }
8482 }
8483 }
8484
8485 return (PF_PASS);
8486}
8487
8488static int
8489pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif,
8490 struct pf_pdesc *pd)
8491{
8492 struct pf_state_peer *src, *dst;
8493 struct pf_state_key_cmp key;
8494
8495 key.app_state = 0;
8496 key.proto = pd->proto;
8497 if (direction == PF_IN) {
8498 key.af_gwy = pd->af;
8499 PF_ACPY(&key.ext_gwy.addr, pd->src, key.af_gwy);
8500 PF_ACPY(&key.gwy.addr, pd->dst, key.af_gwy);
8501 key.ext_gwy.xport.port = 0;
8502 key.gwy.xport.port = 0;
8503 } else {
8504 key.af_lan = pd->af;
8505 PF_ACPY(&key.lan.addr, pd->src, key.af_lan);
8506 PF_ACPY(&key.ext_lan.addr, pd->dst, key.af_lan);
8507 key.lan.xport.port = 0;
8508 key.ext_lan.xport.port = 0;
8509 }
8510
8511 STATE_LOOKUP();
8512
8513 if (direction == (*state)->state_key->direction) {
8514 src = &(*state)->src;
8515 dst = &(*state)->dst;
8516 } else {
8517 src = &(*state)->dst;
8518 dst = &(*state)->src;
8519 }
8520
8521 /* update states */
8522 if (src->state < PFOTHERS_SINGLE)
8523 src->state = PFOTHERS_SINGLE;
8524 if (dst->state == PFOTHERS_SINGLE)
8525 dst->state = PFOTHERS_MULTIPLE;
8526
8527 /* update expire time */
8528 (*state)->expire = pf_time_second();
8529 if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE)
8530 (*state)->timeout = PFTM_OTHER_MULTIPLE;
8531 else
8532 (*state)->timeout = PFTM_OTHER_SINGLE;
8533
8534 /* translate source/destination address, if necessary */
8535 if (STATE_ADDR_TRANSLATE((*state)->state_key)) {
8536 if (direction == PF_OUT) {
8537 switch (pd->af) {
8538#if INET
8539 case AF_INET:
8540 pf_change_a(&pd->src->v4addr.s_addr,
8541 pd->ip_sum,
8542 (*state)->state_key->gwy.addr.v4addr.s_addr,
8543 0);
8544 break;
8545#endif /* INET */
8546#if INET6
8547 case AF_INET6:
8548 PF_ACPY(pd->src,
8549 &(*state)->state_key->gwy.addr, pd->af);
8550 break;
8551#endif /* INET6 */
8552 }
8553 } else {
8554 switch (pd->af) {
8555#if INET
8556 case AF_INET:
8557 pf_change_a(&pd->dst->v4addr.s_addr,
8558 pd->ip_sum,
8559 (*state)->state_key->lan.addr.v4addr.s_addr,
8560 0);
8561 break;
8562#endif /* INET */
8563#if INET6
8564 case AF_INET6:
8565 PF_ACPY(pd->dst,
8566 &(*state)->state_key->lan.addr, pd->af);
8567 break;
8568#endif /* INET6 */
8569 }
8570 }
8571 }
8572
8573 return (PF_PASS);
8574}
8575
8576/*
8577 * ipoff and off are measured from the start of the mbuf chain.
8578 * h must be at "ipoff" on the mbuf chain.
8579 */
8580void *
8581pf_pull_hdr(pbuf_t *pbuf, int off, void *p, int len,
8582 u_short *actionp, u_short *reasonp, sa_family_t af)
8583{
8584 switch (af) {
8585#if INET
8586 case AF_INET: {
8587 struct ip *h = pbuf->pb_data;
8588 u_int16_t fragoff = (ntohs(h->ip_off) & IP_OFFMASK) << 3;
8589
8590 if (fragoff) {
8591 if (fragoff >= len) {
8592 ACTION_SET(actionp, PF_PASS);
8593 } else {
8594 ACTION_SET(actionp, PF_DROP);
8595 REASON_SET(reasonp, PFRES_FRAG);
8596 }
8597 return (NULL);
8598 }
8599 if (pbuf->pb_packet_len < (unsigned)(off + len) ||
8600 ntohs(h->ip_len) < off + len) {
8601 ACTION_SET(actionp, PF_DROP);
8602 REASON_SET(reasonp, PFRES_SHORT);
8603 return (NULL);
8604 }
8605 break;
8606 }
8607#endif /* INET */
8608#if INET6
8609 case AF_INET6: {
8610 struct ip6_hdr *h = pbuf->pb_data;
8611
8612 if (pbuf->pb_packet_len < (unsigned)(off + len) ||
8613 (ntohs(h->ip6_plen) + sizeof (struct ip6_hdr)) <
8614 (unsigned)(off + len)) {
8615 ACTION_SET(actionp, PF_DROP);
8616 REASON_SET(reasonp, PFRES_SHORT);
8617 return (NULL);
8618 }
8619 break;
8620 }
8621#endif /* INET6 */
8622 }
8623 pbuf_copy_data(pbuf, off, len, p);
8624 return (p);
8625}
8626
8627int
8628pf_routable(struct pf_addr *addr, sa_family_t af, struct pfi_kif *kif)
8629{
8630#pragma unused(kif)
8631 struct sockaddr_in *dst;
8632 int ret = 1;
8633#if INET6
8634 struct sockaddr_in6 *dst6;
8635 struct route_in6 ro;
8636#else
8637 struct route ro;
8638#endif
8639
8640 bzero(&ro, sizeof (ro));
8641 switch (af) {
8642 case AF_INET:
8643 dst = satosin(&ro.ro_dst);
8644 dst->sin_family = AF_INET;
8645 dst->sin_len = sizeof (*dst);
8646 dst->sin_addr = addr->v4addr;
8647 break;
8648#if INET6
8649 case AF_INET6:
8650 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
8651 dst6->sin6_family = AF_INET6;
8652 dst6->sin6_len = sizeof (*dst6);
8653 dst6->sin6_addr = addr->v6addr;
8654 break;
8655#endif /* INET6 */
8656 default:
8657 return (0);
8658 }
8659
8660 /* XXX: IFT_ENC is not currently used by anything*/
8661 /* Skip checks for ipsec interfaces */
8662 if (kif != NULL && kif->pfik_ifp->if_type == IFT_ENC)
8663 goto out;
8664
8665 /* XXX: what is the point of this? */
8666 rtalloc((struct route *)&ro);
8667
8668out:
8669 ROUTE_RELEASE(&ro);
8670 return (ret);
8671}
8672
8673int
8674pf_rtlabel_match(struct pf_addr *addr, sa_family_t af, struct pf_addr_wrap *aw)
8675{
8676#pragma unused(aw)
8677 struct sockaddr_in *dst;
8678#if INET6
8679 struct sockaddr_in6 *dst6;
8680 struct route_in6 ro;
8681#else
8682 struct route ro;
8683#endif
8684 int ret = 0;
8685
8686 bzero(&ro, sizeof (ro));
8687 switch (af) {
8688 case AF_INET:
8689 dst = satosin(&ro.ro_dst);
8690 dst->sin_family = AF_INET;
8691 dst->sin_len = sizeof (*dst);
8692 dst->sin_addr = addr->v4addr;
8693 break;
8694#if INET6
8695 case AF_INET6:
8696 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
8697 dst6->sin6_family = AF_INET6;
8698 dst6->sin6_len = sizeof (*dst6);
8699 dst6->sin6_addr = addr->v6addr;
8700 break;
8701#endif /* INET6 */
8702 default:
8703 return (0);
8704 }
8705
8706 /* XXX: what is the point of this? */
8707 rtalloc((struct route *)&ro);
8708
8709 ROUTE_RELEASE(&ro);
8710
8711 return (ret);
8712}
8713
8714#if INET
8715static void
8716pf_route(pbuf_t **pbufp, struct pf_rule *r, int dir, struct ifnet *oifp,
8717 struct pf_state *s, struct pf_pdesc *pd)
8718{
8719#pragma unused(pd)
8720 struct mbuf *m0, *m1;
8721 struct route iproute;
8722 struct route *ro = &iproute;
8723 struct sockaddr_in *dst;
8724 struct ip *ip;
8725 struct ifnet *ifp = NULL;
8726 struct pf_addr naddr;
8727 struct pf_src_node *sn = NULL;
8728 int error = 0;
8729 uint32_t sw_csum;
8730 int interface_mtu = 0;
8731 bzero(&iproute, sizeof (iproute));
8732
8733 if (pbufp == NULL || !pbuf_is_valid(*pbufp) || r == NULL ||
8734 (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
8735 panic("pf_route: invalid parameters");
8736
8737 if (pd->pf_mtag->pftag_routed++ > 3) {
8738 pbuf_destroy(*pbufp);
8739 *pbufp = NULL;
8740 m0 = NULL;
8741 goto bad;
8742 }
8743
8744 /*
8745 * Since this is something of an edge case and may involve the
8746 * host stack (for routing, at least for now), we convert the
8747 * incoming pbuf into an mbuf.
8748 */
8749 if (r->rt == PF_DUPTO)
8750 m0 = pbuf_clone_to_mbuf(*pbufp);
8751 else
8752 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
8753 return;
8754 else {
8755 /* We're going to consume this packet */
8756 m0 = pbuf_to_mbuf(*pbufp, TRUE);
8757 *pbufp = NULL;
8758 }
8759
8760 if (m0 == NULL)
8761 goto bad;
8762
8763 /* We now have the packet in an mbuf (m0) */
8764
8765 if (m0->m_len < (int)sizeof (struct ip)) {
8766 DPFPRINTF(PF_DEBUG_URGENT,
8767 ("pf_route: packet length < sizeof (struct ip)\n"));
8768 goto bad;
8769 }
8770
8771 ip = mtod(m0, struct ip *);
8772
8773 dst = satosin((void *)&ro->ro_dst);
8774 dst->sin_family = AF_INET;
8775 dst->sin_len = sizeof (*dst);
8776 dst->sin_addr = ip->ip_dst;
8777
8778 if (r->rt == PF_FASTROUTE) {
8779 rtalloc(ro);
8780 if (ro->ro_rt == NULL) {
8781 ipstat.ips_noroute++;
8782 goto bad;
8783 }
8784
8785 ifp = ro->ro_rt->rt_ifp;
8786 RT_LOCK(ro->ro_rt);
8787 ro->ro_rt->rt_use++;
8788
8789 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
8790 dst = satosin((void *)ro->ro_rt->rt_gateway);
8791 RT_UNLOCK(ro->ro_rt);
8792 } else {
8793 if (TAILQ_EMPTY(&r->rpool.list)) {
8794 DPFPRINTF(PF_DEBUG_URGENT,
8795 ("pf_route: TAILQ_EMPTY(&r->rpool.list)\n"));
8796 goto bad;
8797 }
8798 if (s == NULL) {
8799 pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src,
8800 &naddr, NULL, &sn);
8801 if (!PF_AZERO(&naddr, AF_INET))
8802 dst->sin_addr.s_addr = naddr.v4addr.s_addr;
8803 ifp = r->rpool.cur->kif ?
8804 r->rpool.cur->kif->pfik_ifp : NULL;
8805 } else {
8806 if (!PF_AZERO(&s->rt_addr, AF_INET))
8807 dst->sin_addr.s_addr =
8808 s->rt_addr.v4addr.s_addr;
8809 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
8810 }
8811 }
8812 if (ifp == NULL)
8813 goto bad;
8814
8815 if (oifp != ifp) {
8816 if (pf_test_mbuf(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS)
8817 goto bad;
8818 else if (m0 == NULL)
8819 goto done;
8820 if (m0->m_len < (int)sizeof (struct ip)) {
8821 DPFPRINTF(PF_DEBUG_URGENT,
8822 ("pf_route: packet length < sizeof (struct ip)\n"));
8823 goto bad;
8824 }
8825 ip = mtod(m0, struct ip *);
8826 }
8827
8828 /* Catch routing changes wrt. hardware checksumming for TCP or UDP. */
8829 ip_output_checksum(ifp, m0, ((ip->ip_hl) << 2), ntohs(ip->ip_len),
8830 &sw_csum);
8831
8832 interface_mtu = ifp->if_mtu;
8833
8834 if (INTF_ADJUST_MTU_FOR_CLAT46(ifp)) {
8835 interface_mtu = IN6_LINKMTU(ifp);
8836 /* Further adjust the size for CLAT46 expansion */
8837 interface_mtu -= CLAT46_HDR_EXPANSION_OVERHD;
8838 }
8839
8840 if (ntohs(ip->ip_len) <= interface_mtu || TSO_IPV4_OK(ifp, m0) ||
8841 (!(ip->ip_off & htons(IP_DF)) &&
8842 (ifp->if_hwassist & CSUM_FRAGMENT))) {
8843 ip->ip_sum = 0;
8844 if (sw_csum & CSUM_DELAY_IP) {
8845 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
8846 sw_csum &= ~CSUM_DELAY_IP;
8847 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_IP;
8848 }
8849 error = ifnet_output(ifp, PF_INET, m0, ro->ro_rt, sintosa(dst));
8850 goto done;
8851 }
8852
8853 /*
8854 * Too large for interface; fragment if possible.
8855 * Must be able to put at least 8 bytes per fragment.
8856 * Balk when DF bit is set or the interface didn't support TSO.
8857 */
8858 if ((ip->ip_off & htons(IP_DF)) ||
8859 (m0->m_pkthdr.csum_flags & CSUM_TSO_IPV4)) {
8860 ipstat.ips_cantfrag++;
8861 if (r->rt != PF_DUPTO) {
8862 icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0,
8863 interface_mtu);
8864 goto done;
8865 } else
8866 goto bad;
8867 }
8868
8869 m1 = m0;
8870
8871 /* PR-8933605: send ip_len,ip_off to ip_fragment in host byte order */
8872#if BYTE_ORDER != BIG_ENDIAN
8873 NTOHS(ip->ip_off);
8874 NTOHS(ip->ip_len);
8875#endif
8876 error = ip_fragment(m0, ifp, interface_mtu, sw_csum);
8877
8878 if (error) {
8879 m0 = NULL;
8880 goto bad;
8881 }
8882
8883 for (m0 = m1; m0; m0 = m1) {
8884 m1 = m0->m_nextpkt;
8885 m0->m_nextpkt = 0;
8886 if (error == 0)
8887 error = ifnet_output(ifp, PF_INET, m0, ro->ro_rt,
8888 sintosa(dst));
8889 else
8890 m_freem(m0);
8891 }
8892
8893 if (error == 0)
8894 ipstat.ips_fragmented++;
8895
8896done:
8897 ROUTE_RELEASE(&iproute);
8898 return;
8899
8900bad:
8901 if (m0)
8902 m_freem(m0);
8903 goto done;
8904}
8905#endif /* INET */
8906
8907#if INET6
8908static void
8909pf_route6(pbuf_t **pbufp, struct pf_rule *r, int dir, struct ifnet *oifp,
8910 struct pf_state *s, struct pf_pdesc *pd)
8911{
8912#pragma unused(pd)
8913 struct mbuf *m0;
8914 struct route_in6 ip6route;
8915 struct route_in6 *ro;
8916 struct sockaddr_in6 *dst;
8917 struct ip6_hdr *ip6;
8918 struct ifnet *ifp = NULL;
8919 struct pf_addr naddr;
8920 struct pf_src_node *sn = NULL;
8921 int error = 0;
8922
8923 if (pbufp == NULL || !pbuf_is_valid(*pbufp) || r == NULL ||
8924 (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
8925 panic("pf_route6: invalid parameters");
8926
8927 if (pd->pf_mtag->pftag_routed++ > 3) {
8928 pbuf_destroy(*pbufp);
8929 *pbufp = NULL;
8930 m0 = NULL;
8931 goto bad;
8932 }
8933
8934 /*
8935 * Since this is something of an edge case and may involve the
8936 * host stack (for routing, at least for now), we convert the
8937 * incoming pbuf into an mbuf.
8938 */
8939 if (r->rt == PF_DUPTO) {
8940 m0 = pbuf_clone_to_mbuf(*pbufp);
8941 } else
8942 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
8943 return;
8944 else {
8945 /* We're about to consume this packet */
8946 m0 = pbuf_to_mbuf(*pbufp, TRUE);
8947 *pbufp = NULL;
8948 }
8949
8950 if (m0 == NULL)
8951 goto bad;
8952
8953 if (m0->m_len < (int)sizeof (struct ip6_hdr)) {
8954 DPFPRINTF(PF_DEBUG_URGENT,
8955 ("pf_route6: m0->m_len < sizeof (struct ip6_hdr)\n"));
8956 goto bad;
8957 }
8958 ip6 = mtod(m0, struct ip6_hdr *);
8959
8960 ro = &ip6route;
8961 bzero((caddr_t)ro, sizeof (*ro));
8962 dst = (struct sockaddr_in6 *)&ro->ro_dst;
8963 dst->sin6_family = AF_INET6;
8964 dst->sin6_len = sizeof (*dst);
8965 dst->sin6_addr = ip6->ip6_dst;
8966
8967 /* Cheat. XXX why only in the v6addr case??? */
8968 if (r->rt == PF_FASTROUTE) {
8969 struct pf_mtag *pf_mtag;
8970
8971 if ((pf_mtag = pf_get_mtag(m0)) == NULL)
8972 goto bad;
8973 pf_mtag->pftag_flags |= PF_TAG_GENERATED;
8974 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
8975 return;
8976 }
8977
8978 if (TAILQ_EMPTY(&r->rpool.list)) {
8979 DPFPRINTF(PF_DEBUG_URGENT,
8980 ("pf_route6: TAILQ_EMPTY(&r->rpool.list)\n"));
8981 goto bad;
8982 }
8983 if (s == NULL) {
8984 pf_map_addr(AF_INET6, r, (struct pf_addr *)(uintptr_t)&ip6->ip6_src,
8985 &naddr, NULL, &sn);
8986 if (!PF_AZERO(&naddr, AF_INET6))
8987 PF_ACPY((struct pf_addr *)&dst->sin6_addr,
8988 &naddr, AF_INET6);
8989 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL;
8990 } else {
8991 if (!PF_AZERO(&s->rt_addr, AF_INET6))
8992 PF_ACPY((struct pf_addr *)&dst->sin6_addr,
8993 &s->rt_addr, AF_INET6);
8994 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
8995 }
8996 if (ifp == NULL)
8997 goto bad;
8998
8999 if (oifp != ifp) {
9000 if (pf_test6_mbuf(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS)
9001 goto bad;
9002 else if (m0 == NULL)
9003 goto done;
9004 if (m0->m_len < (int)sizeof (struct ip6_hdr)) {
9005 DPFPRINTF(PF_DEBUG_URGENT, ("pf_route6: m0->m_len "
9006 "< sizeof (struct ip6_hdr)\n"));
9007 goto bad;
9008 }
9009 ip6 = mtod(m0, struct ip6_hdr *);
9010 }
9011
9012 /*
9013 * If the packet is too large for the outgoing interface,
9014 * send back an icmp6 error.
9015 */
9016 if (IN6_IS_SCOPE_EMBED(&dst->sin6_addr))
9017 dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
9018 if ((unsigned)m0->m_pkthdr.len <= ifp->if_mtu) {
9019 error = nd6_output(ifp, ifp, m0, dst, NULL, NULL);
9020 } else {
9021 in6_ifstat_inc(ifp, ifs6_in_toobig);
9022 if (r->rt != PF_DUPTO)
9023 icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
9024 else
9025 goto bad;
9026 }
9027
9028done:
9029 return;
9030
9031bad:
9032 if (m0)
9033 m_freem(m0);
9034 goto done;
9035}
9036#endif /* INET6 */
9037
9038
9039/*
9040 * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag
9041 * off is the offset where the protocol header starts
9042 * len is the total length of protocol header plus payload
9043 * returns 0 when the checksum is valid, otherwise returns 1.
9044 */
9045static int
9046pf_check_proto_cksum(pbuf_t *pbuf, int off, int len, u_int8_t p,
9047 sa_family_t af)
9048{
9049 u_int16_t sum;
9050
9051 switch (p) {
9052 case IPPROTO_TCP:
9053 case IPPROTO_UDP:
9054 /*
9055 * Optimize for the common case; if the hardware calculated
9056 * value doesn't include pseudo-header checksum, or if it
9057 * is partially-computed (only 16-bit summation), do it in
9058 * software below.
9059 */
9060 if ((*pbuf->pb_csum_flags &
9061 (CSUM_DATA_VALID | CSUM_PSEUDO_HDR)) ==
9062 (CSUM_DATA_VALID | CSUM_PSEUDO_HDR) &&
9063 (*pbuf->pb_csum_data ^ 0xffff) == 0) {
9064 return (0);
9065 }
9066 break;
9067 case IPPROTO_ICMP:
9068#if INET6
9069 case IPPROTO_ICMPV6:
9070#endif /* INET6 */
9071 break;
9072 default:
9073 return (1);
9074 }
9075 if (off < (int)sizeof (struct ip) || len < (int)sizeof (struct udphdr))
9076 return (1);
9077 if (pbuf->pb_packet_len < (unsigned)(off + len))
9078 return (1);
9079 switch (af) {
9080#if INET
9081 case AF_INET:
9082 if (p == IPPROTO_ICMP) {
9083#if 0
9084 if (m->m_len < off)
9085 return (1);
9086 m->m_data += off;
9087 m->m_len -= off;
9088 sum = in_cksum(m, len);
9089 m->m_data -= off;
9090 m->m_len += off;
9091#else
9092 if (pbuf->pb_contig_len < (unsigned)off)
9093 return (1);
9094 sum = pbuf_inet_cksum(pbuf, 0, off, len);
9095#endif
9096 } else {
9097 if (pbuf->pb_contig_len < (int)sizeof (struct ip))
9098 return (1);
9099 sum = pbuf_inet_cksum(pbuf, p, off, len);
9100 }
9101 break;
9102#endif /* INET */
9103#if INET6
9104 case AF_INET6:
9105 if (pbuf->pb_contig_len < (int)sizeof (struct ip6_hdr))
9106 return (1);
9107 sum = pbuf_inet6_cksum(pbuf, p, off, len);
9108 break;
9109#endif /* INET6 */
9110 default:
9111 return (1);
9112 }
9113 if (sum) {
9114 switch (p) {
9115 case IPPROTO_TCP:
9116 tcpstat.tcps_rcvbadsum++;
9117 break;
9118 case IPPROTO_UDP:
9119 udpstat.udps_badsum++;
9120 break;
9121 case IPPROTO_ICMP:
9122 icmpstat.icps_checksum++;
9123 break;
9124#if INET6
9125 case IPPROTO_ICMPV6:
9126 icmp6stat.icp6s_checksum++;
9127 break;
9128#endif /* INET6 */
9129 }
9130 return (1);
9131 }
9132 return (0);
9133}
9134
9135#if INET
9136#define PF_APPLE_UPDATE_PDESC_IPv4() \
9137 do { \
9138 if (pbuf && pd.mp && pbuf != pd.mp) { \
9139 pbuf = pd.mp; \
9140 h = pbuf->pb_data; \
9141 pd.pf_mtag = pf_get_mtag_pbuf(pbuf); \
9142 } \
9143 } while (0)
9144
9145int
9146pf_test_mbuf(int dir, struct ifnet *ifp, struct mbuf **m0,
9147 struct ether_header *eh, struct ip_fw_args *fwa)
9148{
9149 pbuf_t pbuf_store, *pbuf;
9150 int rv;
9151
9152 pbuf_init_mbuf(&pbuf_store, *m0, (*m0)->m_pkthdr.rcvif);
9153 pbuf = &pbuf_store;
9154
9155 rv = pf_test(dir, ifp, &pbuf, eh, fwa);
9156
9157 if (pbuf_is_valid(pbuf)) {
9158 *m0 = pbuf->pb_mbuf;
9159 pbuf->pb_mbuf = NULL;
9160 pbuf_destroy(pbuf);
9161 } else
9162 *m0 = NULL;
9163
9164 return (rv);
9165}
9166
9167int
9168pf_test(int dir, struct ifnet *ifp, pbuf_t **pbufp,
9169 struct ether_header *eh, struct ip_fw_args *fwa)
9170{
9171#if !DUMMYNET
9172#pragma unused(fwa)
9173#endif
9174 struct pfi_kif *kif;
9175 u_short action = PF_PASS, reason = 0, log = 0;
9176 pbuf_t *pbuf = *pbufp;
9177 struct ip *h = 0;
9178 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr;
9179 struct pf_state *s = NULL;
9180 struct pf_state_key *sk = NULL;
9181 struct pf_ruleset *ruleset = NULL;
9182 struct pf_pdesc pd;
9183 int off, dirndx, pqid = 0;
9184
9185 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
9186
9187 if (!pf_status.running)
9188 return (PF_PASS);
9189
9190 memset(&pd, 0, sizeof (pd));
9191
9192 if ((pd.pf_mtag = pf_get_mtag_pbuf(pbuf)) == NULL) {
9193 DPFPRINTF(PF_DEBUG_URGENT,
9194 ("pf_test: pf_get_mtag_pbuf returned NULL\n"));
9195 return (PF_DROP);
9196 }
9197
9198 if (pd.pf_mtag->pftag_flags & PF_TAG_GENERATED)
9199 return (PF_PASS);
9200
9201 kif = (struct pfi_kif *)ifp->if_pf_kif;
9202
9203 if (kif == NULL) {
9204 DPFPRINTF(PF_DEBUG_URGENT,
9205 ("pf_test: kif == NULL, if_name %s\n", ifp->if_name));
9206 return (PF_DROP);
9207 }
9208 if (kif->pfik_flags & PFI_IFLAG_SKIP)
9209 return (PF_PASS);
9210
9211 /* initialize enough of pd for the done label */
9212 h = pbuf->pb_data;
9213 pd.mp = pbuf;
9214 pd.lmw = 0;
9215 pd.pf_mtag = pf_get_mtag_pbuf(pbuf);
9216 pd.src = (struct pf_addr *)&h->ip_src;
9217 pd.dst = (struct pf_addr *)&h->ip_dst;
9218 PF_ACPY(&pd.baddr, pd.src, AF_INET);
9219 PF_ACPY(&pd.bdaddr, pd.dst, AF_INET);
9220 pd.ip_sum = &h->ip_sum;
9221 pd.proto = h->ip_p;
9222 pd.proto_variant = 0;
9223 pd.af = AF_INET;
9224 pd.tos = h->ip_tos;
9225 pd.ttl = h->ip_ttl;
9226 pd.tot_len = ntohs(h->ip_len);
9227 pd.eh = eh;
9228
9229 if (pbuf->pb_packet_len < (int)sizeof (*h)) {
9230 action = PF_DROP;
9231 REASON_SET(&reason, PFRES_SHORT);
9232 log = 1;
9233 goto done;
9234 }
9235
9236#if DUMMYNET
9237 if (fwa != NULL && fwa->fwa_pf_rule != NULL)
9238 goto nonormalize;
9239#endif /* DUMMYNET */
9240
9241 /* We do IP header normalization and packet reassembly here */
9242 action = pf_normalize_ip(pbuf, dir, kif, &reason, &pd);
9243 if (action != PF_PASS || pd.lmw < 0) {
9244 action = PF_DROP;
9245 goto done;
9246 }
9247
9248#if DUMMYNET
9249nonormalize:
9250#endif /* DUMMYNET */
9251 /* pf_normalize can mess with pb_data */
9252 h = pbuf->pb_data;
9253
9254 off = h->ip_hl << 2;
9255 if (off < (int)sizeof (*h)) {
9256 action = PF_DROP;
9257 REASON_SET(&reason, PFRES_SHORT);
9258 log = 1;
9259 goto done;
9260 }
9261
9262 pd.src = (struct pf_addr *)&h->ip_src;
9263 pd.dst = (struct pf_addr *)&h->ip_dst;
9264 PF_ACPY(&pd.baddr, pd.src, AF_INET);
9265 PF_ACPY(&pd.bdaddr, pd.dst, AF_INET);
9266 pd.ip_sum = &h->ip_sum;
9267 pd.proto = h->ip_p;
9268 pd.proto_variant = 0;
9269 pd.mp = pbuf;
9270 pd.lmw = 0;
9271 pd.pf_mtag = pf_get_mtag_pbuf(pbuf);
9272 pd.af = AF_INET;
9273 pd.tos = h->ip_tos;
9274 pd.ttl = h->ip_ttl;
9275 pd.sc = MBUF_SCIDX(pbuf_get_service_class(pbuf));
9276 pd.tot_len = ntohs(h->ip_len);
9277 pd.eh = eh;
9278
9279 if (*pbuf->pb_flags & PKTF_FLOW_ID) {
9280 pd.flowsrc = *pbuf->pb_flowsrc;
9281 pd.flowhash = *pbuf->pb_flowid;
9282 pd.pktflags = *pbuf->pb_flags & PKTF_FLOW_MASK;
9283 }
9284
9285 /* handle fragments that didn't get reassembled by normalization */
9286 if (h->ip_off & htons(IP_MF | IP_OFFMASK)) {
9287 pd.flags |= PFDESC_IP_FRAG;
9288#if DUMMYNET
9289 /* Traffic goes through dummynet first */
9290 action = pf_test_dummynet(&r, dir, kif, &pbuf, &pd, fwa);
9291 if (action == PF_DROP || pbuf == NULL) {
9292 *pbufp = NULL;
9293 return (action);
9294 }
9295#endif /* DUMMYNET */
9296 action = pf_test_fragment(&r, dir, kif, pbuf, h,
9297 &pd, &a, &ruleset);
9298 goto done;
9299 }
9300
9301 switch (h->ip_p) {
9302
9303 case IPPROTO_TCP: {
9304 struct tcphdr th;
9305 pd.hdr.tcp = &th;
9306 if (!pf_pull_hdr(pbuf, off, &th, sizeof (th),
9307 &action, &reason, AF_INET)) {
9308 log = action != PF_PASS;
9309 goto done;
9310 }
9311 pd.p_len = pd.tot_len - off - (th.th_off << 2);
9312 if ((th.th_flags & TH_ACK) && pd.p_len == 0)
9313 pqid = 1;
9314#if DUMMYNET
9315 /* Traffic goes through dummynet first */
9316 action = pf_test_dummynet(&r, dir, kif, &pbuf, &pd, fwa);
9317 if (action == PF_DROP || pbuf == NULL) {
9318 *pbufp = NULL;
9319 return (action);
9320 }
9321#endif /* DUMMYNET */
9322 action = pf_normalize_tcp(dir, kif, pbuf, 0, off, h, &pd);
9323 if (pd.lmw < 0)
9324 goto done;
9325 PF_APPLE_UPDATE_PDESC_IPv4();
9326 if (action == PF_DROP)
9327 goto done;
9328 action = pf_test_state_tcp(&s, dir, kif, pbuf, off, h, &pd,
9329 &reason);
9330 if (action == PF_NAT64)
9331 goto done;
9332 if (pd.lmw < 0)
9333 goto done;
9334 PF_APPLE_UPDATE_PDESC_IPv4();
9335 if (action == PF_PASS) {
9336#if NPFSYNC
9337 pfsync_update_state(s);
9338#endif /* NPFSYNC */
9339 r = s->rule.ptr;
9340 a = s->anchor.ptr;
9341 log = s->log;
9342 } else if (s == NULL)
9343 action = pf_test_rule(&r, &s, dir, kif,
9344 pbuf, off, h, &pd, &a, &ruleset, NULL);
9345 break;
9346 }
9347
9348 case IPPROTO_UDP: {
9349 struct udphdr uh;
9350
9351 pd.hdr.udp = &uh;
9352 if (!pf_pull_hdr(pbuf, off, &uh, sizeof (uh),
9353 &action, &reason, AF_INET)) {
9354 log = action != PF_PASS;
9355 goto done;
9356 }
9357 if (uh.uh_dport == 0 ||
9358 ntohs(uh.uh_ulen) > pbuf->pb_packet_len - off ||
9359 ntohs(uh.uh_ulen) < sizeof (struct udphdr)) {
9360 action = PF_DROP;
9361 REASON_SET(&reason, PFRES_SHORT);
9362 goto done;
9363 }
9364#if DUMMYNET
9365 /* Traffic goes through dummynet first */
9366 action = pf_test_dummynet(&r, dir, kif, &pbuf, &pd, fwa);
9367 if (action == PF_DROP || pbuf == NULL) {
9368 *pbufp = NULL;
9369 return (action);
9370 }
9371#endif /* DUMMYNET */
9372 action = pf_test_state_udp(&s, dir, kif, pbuf, off, h, &pd,
9373 &reason);
9374 if (action == PF_NAT64)
9375 goto done;
9376 if (pd.lmw < 0)
9377 goto done;
9378 PF_APPLE_UPDATE_PDESC_IPv4();
9379 if (action == PF_PASS) {
9380#if NPFSYNC
9381 pfsync_update_state(s);
9382#endif /* NPFSYNC */
9383 r = s->rule.ptr;
9384 a = s->anchor.ptr;
9385 log = s->log;
9386 } else if (s == NULL)
9387 action = pf_test_rule(&r, &s, dir, kif,
9388 pbuf, off, h, &pd, &a, &ruleset, NULL);
9389 break;
9390 }
9391
9392 case IPPROTO_ICMP: {
9393 struct icmp ih;
9394
9395 pd.hdr.icmp = &ih;
9396 if (!pf_pull_hdr(pbuf, off, &ih, ICMP_MINLEN,
9397 &action, &reason, AF_INET)) {
9398 log = action != PF_PASS;
9399 goto done;
9400 }
9401#if DUMMYNET
9402 /* Traffic goes through dummynet first */
9403 action = pf_test_dummynet(&r, dir, kif, &pbuf, &pd, fwa);
9404 if (action == PF_DROP || pbuf == NULL) {
9405 *pbufp = NULL;
9406 return (action);
9407 }
9408#endif /* DUMMYNET */
9409 action = pf_test_state_icmp(&s, dir, kif, pbuf, off, h, &pd,
9410 &reason);
9411 if (action == PF_NAT64)
9412 goto done;
9413 if (pd.lmw < 0)
9414 goto done;
9415 PF_APPLE_UPDATE_PDESC_IPv4();
9416 if (action == PF_PASS) {
9417#if NPFSYNC
9418 pfsync_update_state(s);
9419#endif /* NPFSYNC */
9420 r = s->rule.ptr;
9421 a = s->anchor.ptr;
9422 log = s->log;
9423 } else if (s == NULL)
9424 action = pf_test_rule(&r, &s, dir, kif,
9425 pbuf, off, h, &pd, &a, &ruleset, NULL);
9426 break;
9427 }
9428
9429 case IPPROTO_ESP: {
9430 struct pf_esp_hdr esp;
9431
9432 pd.hdr.esp = &esp;
9433 if (!pf_pull_hdr(pbuf, off, &esp, sizeof (esp), &action, &reason,
9434 AF_INET)) {
9435 log = action != PF_PASS;
9436 goto done;
9437 }
9438#if DUMMYNET
9439 /* Traffic goes through dummynet first */
9440 action = pf_test_dummynet(&r, dir, kif, &pbuf, &pd, fwa);
9441 if (action == PF_DROP || pbuf == NULL) {
9442 *pbufp = NULL;
9443 return (action);
9444 }
9445#endif /* DUMMYNET */
9446 action = pf_test_state_esp(&s, dir, kif, off, &pd);
9447 if (pd.lmw < 0)
9448 goto done;
9449 PF_APPLE_UPDATE_PDESC_IPv4();
9450 if (action == PF_PASS) {
9451#if NPFSYNC
9452 pfsync_update_state(s);
9453#endif /* NPFSYNC */
9454 r = s->rule.ptr;
9455 a = s->anchor.ptr;
9456 log = s->log;
9457 } else if (s == NULL)
9458 action = pf_test_rule(&r, &s, dir, kif,
9459 pbuf, off, h, &pd, &a, &ruleset, NULL);
9460 break;
9461 }
9462
9463 case IPPROTO_GRE: {
9464 struct pf_grev1_hdr grev1;
9465 pd.hdr.grev1 = &grev1;
9466 if (!pf_pull_hdr(pbuf, off, &grev1, sizeof (grev1), &action,
9467 &reason, AF_INET)) {
9468 log = (action != PF_PASS);
9469 goto done;
9470 }
9471#if DUMMYNET
9472 /* Traffic goes through dummynet first */
9473 action = pf_test_dummynet(&r, dir, kif, &pbuf, &pd, fwa);
9474 if (action == PF_DROP || pbuf == NULL) {
9475 *pbufp = NULL;
9476 return (action);
9477 }
9478#endif /* DUMMYNET */
9479 if ((ntohs(grev1.flags) & PF_GRE_FLAG_VERSION_MASK) == 1 &&
9480 ntohs(grev1.protocol_type) == PF_GRE_PPP_ETHERTYPE) {
9481 if (ntohs(grev1.payload_length) >
9482 pbuf->pb_packet_len - off) {
9483 action = PF_DROP;
9484 REASON_SET(&reason, PFRES_SHORT);
9485 goto done;
9486 }
9487 pd.proto_variant = PF_GRE_PPTP_VARIANT;
9488 action = pf_test_state_grev1(&s, dir, kif, off, &pd);
9489 if (pd.lmw < 0) goto done;
9490 PF_APPLE_UPDATE_PDESC_IPv4();
9491 if (action == PF_PASS) {
9492#if NPFSYNC
9493 pfsync_update_state(s);
9494#endif /* NPFSYNC */
9495 r = s->rule.ptr;
9496 a = s->anchor.ptr;
9497 log = s->log;
9498 break;
9499 } else if (s == NULL) {
9500 action = pf_test_rule(&r, &s, dir, kif, pbuf,
9501 off, h, &pd, &a, &ruleset, NULL);
9502 if (action == PF_PASS)
9503 break;
9504 }
9505 }
9506
9507 /* not GREv1/PPTP, so treat as ordinary GRE... */
9508 }
9509
9510 default:
9511#if DUMMYNET
9512 /* Traffic goes through dummynet first */
9513 action = pf_test_dummynet(&r, dir, kif, &pbuf, &pd, fwa);
9514 if (action == PF_DROP || pbuf == NULL) {
9515 *pbufp = NULL;
9516 return (action);
9517 }
9518#endif /* DUMMYNET */
9519 action = pf_test_state_other(&s, dir, kif, &pd);
9520 if (pd.lmw < 0)
9521 goto done;
9522 PF_APPLE_UPDATE_PDESC_IPv4();
9523 if (action == PF_PASS) {
9524#if NPFSYNC
9525 pfsync_update_state(s);
9526#endif /* NPFSYNC */
9527 r = s->rule.ptr;
9528 a = s->anchor.ptr;
9529 log = s->log;
9530 } else if (s == NULL)
9531 action = pf_test_rule(&r, &s, dir, kif, pbuf, off, h,
9532 &pd, &a, &ruleset, NULL);
9533 break;
9534 }
9535
9536done:
9537 if (action == PF_NAT64) {
9538 *pbufp = NULL;
9539 return (action);
9540 }
9541
9542 *pbufp = pd.mp;
9543 PF_APPLE_UPDATE_PDESC_IPv4();
9544
9545 if (action != PF_DROP) {
9546 if (action == PF_PASS && h->ip_hl > 5 &&
9547 !((s && s->allow_opts) || r->allow_opts)) {
9548 action = PF_DROP;
9549 REASON_SET(&reason, PFRES_IPOPTIONS);
9550 log = 1;
9551 DPFPRINTF(PF_DEBUG_MISC,
9552 ("pf: dropping packet with ip options [hlen=%u]\n",
9553 (unsigned int) h->ip_hl));
9554 }
9555
9556 if ((s && s->tag) || PF_RTABLEID_IS_VALID(r->rtableid) ||
9557 (pd.pktflags & PKTF_FLOW_ID))
9558 (void) pf_tag_packet(pbuf, pd.pf_mtag, s ? s->tag : 0,
9559 r->rtableid, &pd);
9560
9561 if (action == PF_PASS) {
9562#if PF_ECN
9563 /* add hints for ecn */
9564 pd.pf_mtag->pftag_hdr = h;
9565 /* record address family */
9566 pd.pf_mtag->pftag_flags &= ~PF_TAG_HDR_INET6;
9567 pd.pf_mtag->pftag_flags |= PF_TAG_HDR_INET;
9568#endif /* PF_ECN */
9569 /* record protocol */
9570 *pbuf->pb_proto = pd.proto;
9571
9572 /*
9573 * connections redirected to loopback should not match sockets
9574 * bound specifically to loopback due to security implications,
9575 * see tcp_input() and in_pcblookup_listen().
9576 */
9577 if (dir == PF_IN && (pd.proto == IPPROTO_TCP ||
9578 pd.proto == IPPROTO_UDP) && s != NULL &&
9579 s->nat_rule.ptr != NULL &&
9580 (s->nat_rule.ptr->action == PF_RDR ||
9581 s->nat_rule.ptr->action == PF_BINAT) &&
9582 (ntohl(pd.dst->v4addr.s_addr) >> IN_CLASSA_NSHIFT)
9583 == IN_LOOPBACKNET)
9584 pd.pf_mtag->pftag_flags |= PF_TAG_TRANSLATE_LOCALHOST;
9585 }
9586 }
9587
9588 if (log) {
9589 struct pf_rule *lr;
9590
9591 if (s != NULL && s->nat_rule.ptr != NULL &&
9592 s->nat_rule.ptr->log & PF_LOG_ALL)
9593 lr = s->nat_rule.ptr;
9594 else
9595 lr = r;
9596 PFLOG_PACKET(kif, h, pbuf, AF_INET, dir, reason, lr, a, ruleset,
9597 &pd);
9598 }
9599
9600 kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
9601 kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++;
9602
9603 if (action == PF_PASS || r->action == PF_DROP) {
9604 dirndx = (dir == PF_OUT);
9605 r->packets[dirndx]++;
9606 r->bytes[dirndx] += pd.tot_len;
9607 if (a != NULL) {
9608 a->packets[dirndx]++;
9609 a->bytes[dirndx] += pd.tot_len;
9610 }
9611 if (s != NULL) {
9612 sk = s->state_key;
9613 if (s->nat_rule.ptr != NULL) {
9614 s->nat_rule.ptr->packets[dirndx]++;
9615 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len;
9616 }
9617 if (s->src_node != NULL) {
9618 s->src_node->packets[dirndx]++;
9619 s->src_node->bytes[dirndx] += pd.tot_len;
9620 }
9621 if (s->nat_src_node != NULL) {
9622 s->nat_src_node->packets[dirndx]++;
9623 s->nat_src_node->bytes[dirndx] += pd.tot_len;
9624 }
9625 dirndx = (dir == sk->direction) ? 0 : 1;
9626 s->packets[dirndx]++;
9627 s->bytes[dirndx] += pd.tot_len;
9628 }
9629 tr = r;
9630 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
9631 if (nr != NULL) {
9632 struct pf_addr *x;
9633 /*
9634 * XXX: we need to make sure that the addresses
9635 * passed to pfr_update_stats() are the same than
9636 * the addresses used during matching (pfr_match)
9637 */
9638 if (r == &pf_default_rule) {
9639 tr = nr;
9640 x = (sk == NULL || sk->direction == dir) ?
9641 &pd.baddr : &pd.naddr;
9642 } else
9643 x = (sk == NULL || sk->direction == dir) ?
9644 &pd.naddr : &pd.baddr;
9645 if (x == &pd.baddr || s == NULL) {
9646 /* we need to change the address */
9647 if (dir == PF_OUT)
9648 pd.src = x;
9649 else
9650 pd.dst = x;
9651 }
9652 }
9653 if (tr->src.addr.type == PF_ADDR_TABLE)
9654 pfr_update_stats(tr->src.addr.p.tbl, (sk == NULL ||
9655 sk->direction == dir) ?
9656 pd.src : pd.dst, pd.af,
9657 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
9658 tr->src.neg);
9659 if (tr->dst.addr.type == PF_ADDR_TABLE)
9660 pfr_update_stats(tr->dst.addr.p.tbl, (sk == NULL ||
9661 sk->direction == dir) ? pd.dst : pd.src, pd.af,
9662 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
9663 tr->dst.neg);
9664 }
9665
9666 VERIFY(pbuf == NULL || pd.mp == NULL || pd.mp == pbuf);
9667
9668 if (*pbufp) {
9669 if (pd.lmw < 0) {
9670 REASON_SET(&reason, PFRES_MEMORY);
9671 action = PF_DROP;
9672 }
9673
9674 if (action == PF_DROP) {
9675 pbuf_destroy(*pbufp);
9676 *pbufp = NULL;
9677 return (PF_DROP);
9678 }
9679
9680 *pbufp = pbuf;
9681 }
9682
9683 if (action == PF_SYNPROXY_DROP) {
9684 pbuf_destroy(*pbufp);
9685 *pbufp = NULL;
9686 action = PF_PASS;
9687 } else if (r->rt)
9688 /* pf_route can free the pbuf causing *pbufp to become NULL */
9689 pf_route(pbufp, r, dir, kif->pfik_ifp, s, &pd);
9690
9691 return (action);
9692}
9693#endif /* INET */
9694
9695#if INET6
9696#define PF_APPLE_UPDATE_PDESC_IPv6() \
9697 do { \
9698 if (pbuf && pd.mp && pbuf != pd.mp) { \
9699 pbuf = pd.mp; \
9700 } \
9701 h = pbuf->pb_data; \
9702 } while (0)
9703
9704int
9705pf_test6_mbuf(int dir, struct ifnet *ifp, struct mbuf **m0,
9706 struct ether_header *eh, struct ip_fw_args *fwa)
9707{
9708 pbuf_t pbuf_store, *pbuf;
9709 int rv;
9710
9711 pbuf_init_mbuf(&pbuf_store, *m0, (*m0)->m_pkthdr.rcvif);
9712 pbuf = &pbuf_store;
9713
9714 rv = pf_test6(dir, ifp, &pbuf, eh, fwa);
9715
9716 if (pbuf_is_valid(pbuf)) {
9717 *m0 = pbuf->pb_mbuf;
9718 pbuf->pb_mbuf = NULL;
9719 pbuf_destroy(pbuf);
9720 } else
9721 *m0 = NULL;
9722
9723 return (rv);
9724}
9725
9726int
9727pf_test6(int dir, struct ifnet *ifp, pbuf_t **pbufp,
9728 struct ether_header *eh, struct ip_fw_args *fwa)
9729{
9730#if !DUMMYNET
9731#pragma unused(fwa)
9732#endif
9733 struct pfi_kif *kif;
9734 u_short action = PF_PASS, reason = 0, log = 0;
9735 pbuf_t *pbuf = *pbufp;
9736 struct ip6_hdr *h;
9737 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr;
9738 struct pf_state *s = NULL;
9739 struct pf_state_key *sk = NULL;
9740 struct pf_ruleset *ruleset = NULL;
9741 struct pf_pdesc pd;
9742 int off, terminal = 0, dirndx, rh_cnt = 0;
9743 u_int8_t nxt;
9744
9745 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
9746
9747 if (!pf_status.running)
9748 return (PF_PASS);
9749
9750 memset(&pd, 0, sizeof (pd));
9751
9752 if ((pd.pf_mtag = pf_get_mtag_pbuf(pbuf)) == NULL) {
9753 DPFPRINTF(PF_DEBUG_URGENT,
9754 ("pf_test6: pf_get_mtag_pbuf returned NULL\n"));
9755 return (PF_DROP);
9756 }
9757
9758 if (pd.pf_mtag->pftag_flags & PF_TAG_GENERATED)
9759 return (PF_PASS);
9760
9761 kif = (struct pfi_kif *)ifp->if_pf_kif;
9762
9763 if (kif == NULL) {
9764 DPFPRINTF(PF_DEBUG_URGENT,
9765 ("pf_test6: kif == NULL, if_name %s\n", ifp->if_name));
9766 return (PF_DROP);
9767 }
9768 if (kif->pfik_flags & PFI_IFLAG_SKIP)
9769 return (PF_PASS);
9770
9771 h = pbuf->pb_data;
9772
9773 nxt = h->ip6_nxt;
9774 off = ((caddr_t)h - (caddr_t)pbuf->pb_data) + sizeof(struct ip6_hdr);
9775 pd.mp = pbuf;
9776 pd.lmw = 0;
9777 pd.pf_mtag = pf_get_mtag_pbuf(pbuf);
9778 pd.src = (struct pf_addr *)(uintptr_t)&h->ip6_src;
9779 pd.dst = (struct pf_addr *)(uintptr_t)&h->ip6_dst;
9780 PF_ACPY(&pd.baddr, pd.src, AF_INET6);
9781 PF_ACPY(&pd.bdaddr, pd.dst, AF_INET6);
9782 pd.ip_sum = NULL;
9783 pd.af = AF_INET6;
9784 pd.proto = nxt;
9785 pd.proto_variant = 0;
9786 pd.tos = 0;
9787 pd.ttl = h->ip6_hlim;
9788 pd.sc = MBUF_SCIDX(pbuf_get_service_class(pbuf));
9789 pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr);
9790 pd.eh = eh;
9791
9792 if (*pbuf->pb_flags & PKTF_FLOW_ID) {
9793 pd.flowsrc = *pbuf->pb_flowsrc;
9794 pd.flowhash = *pbuf->pb_flowid;
9795 pd.pktflags = (*pbuf->pb_flags & PKTF_FLOW_MASK);
9796 }
9797
9798 if (pbuf->pb_packet_len < (int)sizeof (*h)) {
9799 action = PF_DROP;
9800 REASON_SET(&reason, PFRES_SHORT);
9801 log = 1;
9802 goto done;
9803 }
9804
9805#if DUMMYNET
9806 if (fwa != NULL && fwa->fwa_pf_rule != NULL)
9807 goto nonormalize;
9808#endif /* DUMMYNET */
9809
9810 /* We do IP header normalization and packet reassembly here */
9811 action = pf_normalize_ip6(pbuf, dir, kif, &reason, &pd);
9812 if (action != PF_PASS || pd.lmw < 0) {
9813 action = PF_DROP;
9814 goto done;
9815 }
9816
9817#if DUMMYNET
9818nonormalize:
9819#endif /* DUMMYNET */
9820 h = pbuf->pb_data;
9821
9822#if 1
9823 /*
9824 * we do not support jumbogram yet. if we keep going, zero ip6_plen
9825 * will do something bad, so drop the packet for now.
9826 */
9827 if (htons(h->ip6_plen) == 0) {
9828 action = PF_DROP;
9829 REASON_SET(&reason, PFRES_NORM); /*XXX*/
9830 goto done;
9831 }
9832#endif
9833
9834 pd.src = (struct pf_addr *)(uintptr_t)&h->ip6_src;
9835 pd.dst = (struct pf_addr *)(uintptr_t)&h->ip6_dst;
9836 PF_ACPY(&pd.baddr, pd.src, AF_INET6);
9837 PF_ACPY(&pd.bdaddr, pd.dst, AF_INET6);
9838 pd.ip_sum = NULL;
9839 pd.af = AF_INET6;
9840 pd.tos = 0;
9841 pd.ttl = h->ip6_hlim;
9842 pd.tot_len = ntohs(h->ip6_plen) + sizeof (struct ip6_hdr);
9843 pd.eh = eh;
9844
9845 off = ((caddr_t)h - (caddr_t)pbuf->pb_data) + sizeof (struct ip6_hdr);
9846 pd.proto = h->ip6_nxt;
9847 pd.proto_variant = 0;
9848 pd.mp = pbuf;
9849 pd.lmw = 0;
9850 pd.pf_mtag = pf_get_mtag_pbuf(pbuf);
9851
9852 do {
9853 switch (nxt) {
9854 case IPPROTO_FRAGMENT: {
9855 struct ip6_frag ip6f;
9856
9857 pd.flags |= PFDESC_IP_FRAG;
9858 if (!pf_pull_hdr(pbuf, off, &ip6f, sizeof ip6f, NULL,
9859 &reason, pd.af)) {
9860 DPFPRINTF(PF_DEBUG_MISC,
9861 ("pf: IPv6 short fragment header\n"));
9862 action = PF_DROP;
9863 REASON_SET(&reason, PFRES_SHORT);
9864 log = 1;
9865 goto done;
9866 }
9867 pd.proto = nxt = ip6f.ip6f_nxt;
9868#if DUMMYNET
9869 /* Traffic goes through dummynet first */
9870 action = pf_test_dummynet(&r, dir, kif, &pbuf, &pd,
9871 fwa);
9872 if (action == PF_DROP || pbuf == NULL) {
9873 *pbufp = NULL;
9874 return (action);
9875 }
9876#endif /* DUMMYNET */
9877 action = pf_test_fragment(&r, dir, kif, pbuf, h, &pd,
9878 &a, &ruleset);
9879 if (action == PF_DROP) {
9880 REASON_SET(&reason, PFRES_FRAG);
9881 log = 1;
9882 }
9883 goto done;
9884 }
9885 case IPPROTO_ROUTING:
9886 ++rh_cnt;
9887 /* FALL THROUGH */
9888
9889 case IPPROTO_AH:
9890 case IPPROTO_HOPOPTS:
9891 case IPPROTO_DSTOPTS: {
9892 /* get next header and header length */
9893 struct ip6_ext opt6;
9894
9895 if (!pf_pull_hdr(pbuf, off, &opt6, sizeof(opt6),
9896 NULL, &reason, pd.af)) {
9897 DPFPRINTF(PF_DEBUG_MISC,
9898 ("pf: IPv6 short opt\n"));
9899 action = PF_DROP;
9900 log = 1;
9901 goto done;
9902 }
9903 if (pd.proto == IPPROTO_AH)
9904 off += (opt6.ip6e_len + 2) * 4;
9905 else
9906 off += (opt6.ip6e_len + 1) * 8;
9907 nxt = opt6.ip6e_nxt;
9908 /* goto the next header */
9909 break;
9910 }
9911 default:
9912 terminal++;
9913 break;
9914 }
9915 } while (!terminal);
9916
9917
9918 switch (pd.proto) {
9919
9920 case IPPROTO_TCP: {
9921 struct tcphdr th;
9922
9923 pd.hdr.tcp = &th;
9924 if (!pf_pull_hdr(pbuf, off, &th, sizeof (th),
9925 &action, &reason, AF_INET6)) {
9926 log = action != PF_PASS;
9927 goto done;
9928 }
9929 pd.p_len = pd.tot_len - off - (th.th_off << 2);
9930#if DUMMYNET
9931 /* Traffic goes through dummynet first */
9932 action = pf_test_dummynet(&r, dir, kif, &pbuf, &pd, fwa);
9933 if (action == PF_DROP || pbuf == NULL) {
9934 *pbufp = NULL;
9935 return (action);
9936 }
9937#endif /* DUMMYNET */
9938 action = pf_normalize_tcp(dir, kif, pbuf, 0, off, h, &pd);
9939 if (pd.lmw < 0)
9940 goto done;
9941 PF_APPLE_UPDATE_PDESC_IPv6();
9942 if (action == PF_DROP)
9943 goto done;
9944 action = pf_test_state_tcp(&s, dir, kif, pbuf, off, h, &pd,
9945 &reason);
9946 if (action == PF_NAT64)
9947 goto done;
9948 if (pd.lmw < 0)
9949 goto done;
9950 PF_APPLE_UPDATE_PDESC_IPv6();
9951 if (action == PF_PASS) {
9952#if NPFSYNC
9953 pfsync_update_state(s);
9954#endif /* NPFSYNC */
9955 r = s->rule.ptr;
9956 a = s->anchor.ptr;
9957 log = s->log;
9958 } else if (s == NULL)
9959 action = pf_test_rule(&r, &s, dir, kif,
9960 pbuf, off, h, &pd, &a, &ruleset, NULL);
9961 break;
9962 }
9963
9964 case IPPROTO_UDP: {
9965 struct udphdr uh;
9966
9967 pd.hdr.udp = &uh;
9968 if (!pf_pull_hdr(pbuf, off, &uh, sizeof (uh),
9969 &action, &reason, AF_INET6)) {
9970 log = action != PF_PASS;
9971 goto done;
9972 }
9973 if (uh.uh_dport == 0 ||
9974 ntohs(uh.uh_ulen) > pbuf->pb_packet_len - off ||
9975 ntohs(uh.uh_ulen) < sizeof (struct udphdr)) {
9976 action = PF_DROP;
9977 REASON_SET(&reason, PFRES_SHORT);
9978 goto done;
9979 }
9980#if DUMMYNET
9981 /* Traffic goes through dummynet first */
9982 action = pf_test_dummynet(&r, dir, kif, &pbuf, &pd, fwa);
9983 if (action == PF_DROP || pbuf == NULL) {
9984 *pbufp = NULL;
9985 return (action);
9986 }
9987#endif /* DUMMYNET */
9988 action = pf_test_state_udp(&s, dir, kif, pbuf, off, h, &pd,
9989 &reason);
9990 if (action == PF_NAT64)
9991 goto done;
9992 if (pd.lmw < 0)
9993 goto done;
9994 PF_APPLE_UPDATE_PDESC_IPv6();
9995 if (action == PF_PASS) {
9996#if NPFSYNC
9997 pfsync_update_state(s);
9998#endif /* NPFSYNC */
9999 r = s->rule.ptr;
10000 a = s->anchor.ptr;
10001 log = s->log;
10002 } else if (s == NULL)
10003 action = pf_test_rule(&r, &s, dir, kif,
10004 pbuf, off, h, &pd, &a, &ruleset, NULL);
10005 break;
10006 }
10007
10008 case IPPROTO_ICMPV6: {
10009 struct icmp6_hdr ih;
10010
10011 pd.hdr.icmp6 = &ih;
10012 if (!pf_pull_hdr(pbuf, off, &ih, sizeof (ih),
10013 &action, &reason, AF_INET6)) {
10014 log = action != PF_PASS;
10015 goto done;
10016 }
10017#if DUMMYNET
10018 /* Traffic goes through dummynet first */
10019 action = pf_test_dummynet(&r, dir, kif, &pbuf, &pd, fwa);
10020 if (action == PF_DROP || pbuf == NULL) {
10021 *pbufp = NULL;
10022 return (action);
10023 }
10024#endif /* DUMMYNET */
10025 action = pf_test_state_icmp(&s, dir, kif,
10026 pbuf, off, h, &pd, &reason);
10027 if (action == PF_NAT64)
10028 goto done;
10029 if (pd.lmw < 0)
10030 goto done;
10031 PF_APPLE_UPDATE_PDESC_IPv6();
10032 if (action == PF_PASS) {
10033#if NPFSYNC
10034 pfsync_update_state(s);
10035#endif /* NPFSYNC */
10036 r = s->rule.ptr;
10037 a = s->anchor.ptr;
10038 log = s->log;
10039 } else if (s == NULL)
10040 action = pf_test_rule(&r, &s, dir, kif,
10041 pbuf, off, h, &pd, &a, &ruleset, NULL);
10042 break;
10043 }
10044
10045 case IPPROTO_ESP: {
10046 struct pf_esp_hdr esp;
10047
10048 pd.hdr.esp = &esp;
10049 if (!pf_pull_hdr(pbuf, off, &esp, sizeof (esp), &action,
10050 &reason, AF_INET6)) {
10051 log = action != PF_PASS;
10052 goto done;
10053 }
10054#if DUMMYNET
10055 /* Traffic goes through dummynet first */
10056 action = pf_test_dummynet(&r, dir, kif, &pbuf, &pd, fwa);
10057 if (action == PF_DROP || pbuf == NULL) {
10058 *pbufp = NULL;
10059 return (action);
10060 }
10061#endif /* DUMMYNET */
10062 action = pf_test_state_esp(&s, dir, kif, off, &pd);
10063 if (pd.lmw < 0)
10064 goto done;
10065 PF_APPLE_UPDATE_PDESC_IPv6();
10066 if (action == PF_PASS) {
10067#if NPFSYNC
10068 pfsync_update_state(s);
10069#endif /* NPFSYNC */
10070 r = s->rule.ptr;
10071 a = s->anchor.ptr;
10072 log = s->log;
10073 } else if (s == NULL)
10074 action = pf_test_rule(&r, &s, dir, kif,
10075 pbuf, off, h, &pd, &a, &ruleset, NULL);
10076 break;
10077 }
10078
10079 case IPPROTO_GRE: {
10080 struct pf_grev1_hdr grev1;
10081
10082 pd.hdr.grev1 = &grev1;
10083 if (!pf_pull_hdr(pbuf, off, &grev1, sizeof (grev1), &action,
10084 &reason, AF_INET6)) {
10085 log = (action != PF_PASS);
10086 goto done;
10087 }
10088#if DUMMYNET
10089 /* Traffic goes through dummynet first */
10090 action = pf_test_dummynet(&r, dir, kif, &pbuf, &pd, fwa);
10091 if (action == PF_DROP || pbuf == NULL) {
10092 *pbufp = NULL;
10093 return (action);
10094 }
10095#endif /* DUMMYNET */
10096 if ((ntohs(grev1.flags) & PF_GRE_FLAG_VERSION_MASK) == 1 &&
10097 ntohs(grev1.protocol_type) == PF_GRE_PPP_ETHERTYPE) {
10098 if (ntohs(grev1.payload_length) >
10099 pbuf->pb_packet_len - off) {
10100 action = PF_DROP;
10101 REASON_SET(&reason, PFRES_SHORT);
10102 goto done;
10103 }
10104 action = pf_test_state_grev1(&s, dir, kif, off, &pd);
10105 if (pd.lmw < 0)
10106 goto done;
10107 PF_APPLE_UPDATE_PDESC_IPv6();
10108 if (action == PF_PASS) {
10109#if NPFSYNC
10110 pfsync_update_state(s);
10111#endif /* NPFSYNC */
10112 r = s->rule.ptr;
10113 a = s->anchor.ptr;
10114 log = s->log;
10115 break;
10116 } else if (s == NULL) {
10117 action = pf_test_rule(&r, &s, dir, kif, pbuf,
10118 off, h, &pd, &a, &ruleset, NULL);
10119 if (action == PF_PASS)
10120 break;
10121 }
10122 }
10123
10124 /* not GREv1/PPTP, so treat as ordinary GRE... */
10125 }
10126
10127 default:
10128#if DUMMYNET
10129 /* Traffic goes through dummynet first */
10130 action = pf_test_dummynet(&r, dir, kif, &pbuf, &pd, fwa);
10131 if (action == PF_DROP || pbuf == NULL) {
10132 *pbufp = NULL;
10133 return (action);
10134 }
10135#endif /* DUMMYNET */
10136 action = pf_test_state_other(&s, dir, kif, &pd);
10137 if (pd.lmw < 0)
10138 goto done;
10139 PF_APPLE_UPDATE_PDESC_IPv6();
10140 if (action == PF_PASS) {
10141#if NPFSYNC
10142 pfsync_update_state(s);
10143#endif /* NPFSYNC */
10144 r = s->rule.ptr;
10145 a = s->anchor.ptr;
10146 log = s->log;
10147 } else if (s == NULL)
10148 action = pf_test_rule(&r, &s, dir, kif, pbuf, off, h,
10149 &pd, &a, &ruleset, NULL);
10150 break;
10151 }
10152
10153done:
10154 if (action == PF_NAT64) {
10155 *pbufp = NULL;
10156 return (action);
10157 }
10158
10159 *pbufp = pd.mp;
10160 PF_APPLE_UPDATE_PDESC_IPv6();
10161
10162 /* handle dangerous IPv6 extension headers. */
10163 if (action != PF_DROP) {
10164 if (action == PF_PASS && rh_cnt &&
10165 !((s && s->allow_opts) || r->allow_opts)) {
10166 action = PF_DROP;
10167 REASON_SET(&reason, PFRES_IPOPTIONS);
10168 log = 1;
10169 DPFPRINTF(PF_DEBUG_MISC,
10170 ("pf: dropping packet with dangerous v6addr headers\n"));
10171 }
10172
10173 if ((s && s->tag) || PF_RTABLEID_IS_VALID(r->rtableid) ||
10174 (pd.pktflags & PKTF_FLOW_ID))
10175 (void) pf_tag_packet(pbuf, pd.pf_mtag, s ? s->tag : 0,
10176 r->rtableid, &pd);
10177
10178 if (action == PF_PASS) {
10179#if PF_ECN
10180 /* add hints for ecn */
10181 pd.pf_mtag->pftag_hdr = h;
10182 /* record address family */
10183 pd.pf_mtag->pftag_flags &= ~PF_TAG_HDR_INET;
10184 pd.pf_mtag->pftag_flags |= PF_TAG_HDR_INET6;
10185#endif /* PF_ECN */
10186 /* record protocol */
10187 *pbuf->pb_proto = pd.proto;
10188 if (dir == PF_IN && (pd.proto == IPPROTO_TCP ||
10189 pd.proto == IPPROTO_UDP) && s != NULL &&
10190 s->nat_rule.ptr != NULL &&
10191 (s->nat_rule.ptr->action == PF_RDR ||
10192 s->nat_rule.ptr->action == PF_BINAT) &&
10193 IN6_IS_ADDR_LOOPBACK(&pd.dst->v6addr))
10194 pd.pf_mtag->pftag_flags |= PF_TAG_TRANSLATE_LOCALHOST;
10195 }
10196 }
10197
10198
10199 if (log) {
10200 struct pf_rule *lr;
10201
10202 if (s != NULL && s->nat_rule.ptr != NULL &&
10203 s->nat_rule.ptr->log & PF_LOG_ALL)
10204 lr = s->nat_rule.ptr;
10205 else
10206 lr = r;
10207 PFLOG_PACKET(kif, h, pbuf, AF_INET6, dir, reason, lr, a, ruleset,
10208 &pd);
10209 }
10210
10211 kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
10212 kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++;
10213
10214 if (action == PF_PASS || r->action == PF_DROP) {
10215 dirndx = (dir == PF_OUT);
10216 r->packets[dirndx]++;
10217 r->bytes[dirndx] += pd.tot_len;
10218 if (a != NULL) {
10219 a->packets[dirndx]++;
10220 a->bytes[dirndx] += pd.tot_len;
10221 }
10222 if (s != NULL) {
10223 sk = s->state_key;
10224 if (s->nat_rule.ptr != NULL) {
10225 s->nat_rule.ptr->packets[dirndx]++;
10226 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len;
10227 }
10228 if (s->src_node != NULL) {
10229 s->src_node->packets[dirndx]++;
10230 s->src_node->bytes[dirndx] += pd.tot_len;
10231 }
10232 if (s->nat_src_node != NULL) {
10233 s->nat_src_node->packets[dirndx]++;
10234 s->nat_src_node->bytes[dirndx] += pd.tot_len;
10235 }
10236 dirndx = (dir == sk->direction) ? 0 : 1;
10237 s->packets[dirndx]++;
10238 s->bytes[dirndx] += pd.tot_len;
10239 }
10240 tr = r;
10241 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
10242 if (nr != NULL) {
10243 struct pf_addr *x;
10244 /*
10245 * XXX: we need to make sure that the addresses
10246 * passed to pfr_update_stats() are the same than
10247 * the addresses used during matching (pfr_match)
10248 */
10249 if (r == &pf_default_rule) {
10250 tr = nr;
10251 x = (s == NULL || sk->direction == dir) ?
10252 &pd.baddr : &pd.naddr;
10253 } else {
10254 x = (s == NULL || sk->direction == dir) ?
10255 &pd.naddr : &pd.baddr;
10256 }
10257 if (x == &pd.baddr || s == NULL) {
10258 if (dir == PF_OUT)
10259 pd.src = x;
10260 else
10261 pd.dst = x;
10262 }
10263 }
10264 if (tr->src.addr.type == PF_ADDR_TABLE)
10265 pfr_update_stats(tr->src.addr.p.tbl, (sk == NULL ||
10266 sk->direction == dir) ? pd.src : pd.dst, pd.af,
10267 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
10268 tr->src.neg);
10269 if (tr->dst.addr.type == PF_ADDR_TABLE)
10270 pfr_update_stats(tr->dst.addr.p.tbl, (sk == NULL ||
10271 sk->direction == dir) ? pd.dst : pd.src, pd.af,
10272 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
10273 tr->dst.neg);
10274 }
10275
10276#if 0
10277 if (action == PF_SYNPROXY_DROP) {
10278 m_freem(*m0);
10279 *m0 = NULL;
10280 action = PF_PASS;
10281 } else if (r->rt)
10282 /* pf_route6 can free the mbuf causing *m0 to become NULL */
10283 pf_route6(m0, r, dir, kif->pfik_ifp, s, &pd);
10284#else
10285 VERIFY(pbuf == NULL || pd.mp == NULL || pd.mp == pbuf);
10286
10287 if (*pbufp) {
10288 if (pd.lmw < 0) {
10289 REASON_SET(&reason, PFRES_MEMORY);
10290 action = PF_DROP;
10291 }
10292
10293 if (action == PF_DROP) {
10294 pbuf_destroy(*pbufp);
10295 *pbufp = NULL;
10296 return (PF_DROP);
10297 }
10298
10299 *pbufp = pbuf;
10300 }
10301
10302 if (action == PF_SYNPROXY_DROP) {
10303 pbuf_destroy(*pbufp);
10304 *pbufp = NULL;
10305 action = PF_PASS;
10306 } else if (r->rt) {
10307 /* pf_route6 can free the mbuf causing *m0 to become NULL */
10308 pf_route6(pbufp, r, dir, kif->pfik_ifp, s, &pd);
10309 }
10310#endif /* 0 */
10311
10312 return (action);
10313}
10314#endif /* INET6 */
10315
10316static int
10317pf_check_congestion(struct ifqueue *ifq)
10318{
10319#pragma unused(ifq)
10320 return (0);
10321}
10322
10323void
10324pool_init(struct pool *pp, size_t size, unsigned int align, unsigned int ioff,
10325 int flags, const char *wchan, void *palloc)
10326{
10327#pragma unused(align, ioff, flags, palloc)
10328 bzero(pp, sizeof (*pp));
10329 pp->pool_zone = zinit(size, 1024 * size, PAGE_SIZE, wchan);
10330 if (pp->pool_zone != NULL) {
10331 zone_change(pp->pool_zone, Z_EXPAND, TRUE);
10332 zone_change(pp->pool_zone, Z_CALLERACCT, FALSE);
10333 pp->pool_hiwat = pp->pool_limit = (unsigned int)-1;
10334 pp->pool_name = wchan;
10335 }
10336}
10337
10338/* Zones cannot be currently destroyed */
10339void
10340pool_destroy(struct pool *pp)
10341{
10342#pragma unused(pp)
10343}
10344
10345void
10346pool_sethiwat(struct pool *pp, int n)
10347{
10348 pp->pool_hiwat = n; /* Currently unused */
10349}
10350
10351void
10352pool_sethardlimit(struct pool *pp, int n, const char *warnmess, int ratecap)
10353{
10354#pragma unused(warnmess, ratecap)
10355 pp->pool_limit = n;
10356}
10357
10358void *
10359pool_get(struct pool *pp, int flags)
10360{
10361 void *buf;
10362
10363 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
10364
10365 if (pp->pool_count > pp->pool_limit) {
10366 DPFPRINTF(PF_DEBUG_NOISY,
10367 ("pf: pool %s hard limit reached (%d)\n",
10368 pp->pool_name != NULL ? pp->pool_name : "unknown",
10369 pp->pool_limit));
10370 pp->pool_fails++;
10371 return (NULL);
10372 }
10373
10374 buf = zalloc_canblock(pp->pool_zone, (flags & (PR_NOWAIT | PR_WAITOK)));
10375 if (buf != NULL) {
10376 pp->pool_count++;
10377 VERIFY(pp->pool_count != 0);
10378 }
10379 return (buf);
10380}
10381
10382void
10383pool_put(struct pool *pp, void *v)
10384{
10385 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
10386
10387 zfree(pp->pool_zone, v);
10388 VERIFY(pp->pool_count != 0);
10389 pp->pool_count--;
10390}
10391
10392struct pf_mtag *
10393pf_find_mtag_pbuf(pbuf_t *pbuf)
10394{
10395
10396 return (pbuf->pb_pftag);
10397}
10398
10399struct pf_mtag *
10400pf_find_mtag(struct mbuf *m)
10401{
10402
10403 return (m_pftag(m));
10404}
10405
10406struct pf_mtag *
10407pf_get_mtag(struct mbuf *m)
10408{
10409 return (pf_find_mtag(m));
10410}
10411
10412struct pf_mtag *
10413pf_get_mtag_pbuf(pbuf_t *pbuf)
10414{
10415 return (pf_find_mtag_pbuf(pbuf));
10416}
10417
10418uint64_t
10419pf_time_second(void)
10420{
10421 struct timeval t;
10422
10423 microuptime(&t);
10424 return (t.tv_sec);
10425}
10426
10427uint64_t
10428pf_calendar_time_second(void)
10429{
10430 struct timeval t;
10431
10432 getmicrotime(&t);
10433 return (t.tv_sec);
10434}
10435
10436static void *
10437hook_establish(struct hook_desc_head *head, int tail, hook_fn_t fn, void *arg)
10438{
10439 struct hook_desc *hd;
10440
10441 hd = _MALLOC(sizeof(*hd), M_DEVBUF, M_WAITOK);
10442 if (hd == NULL)
10443 return (NULL);
10444
10445 hd->hd_fn = fn;
10446 hd->hd_arg = arg;
10447 if (tail)
10448 TAILQ_INSERT_TAIL(head, hd, hd_list);
10449 else
10450 TAILQ_INSERT_HEAD(head, hd, hd_list);
10451
10452 return (hd);
10453}
10454
10455static void
10456hook_runloop(struct hook_desc_head *head, int flags)
10457{
10458 struct hook_desc *hd;
10459
10460 if (!(flags & HOOK_REMOVE)) {
10461 if (!(flags & HOOK_ABORT))
10462 TAILQ_FOREACH(hd, head, hd_list)
10463 hd->hd_fn(hd->hd_arg);
10464 } else {
10465 while (!!(hd = TAILQ_FIRST(head))) {
10466 TAILQ_REMOVE(head, hd, hd_list);
10467 if (!(flags & HOOK_ABORT))
10468 hd->hd_fn(hd->hd_arg);
10469 if (flags & HOOK_FREE)
10470 _FREE(hd, M_DEVBUF);
10471 }
10472 }
10473}
10474