]> git.saurik.com Git - apple/xnu.git/blob - bsd/net/pf.c
projects / apple / xnu.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
xnu-7195.101.1.tar.gz
[apple/xnu.git] / bsd / net / pf.c
1 /*
2 * Copyright (c) 2007-2020 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 #include <netinet/ip6.h>
119 #include <netinet6/in6_pcb.h>
120 #include <netinet6/ip6_var.h>
121 #include <netinet/icmp6.h>
122 #include <netinet6/nd6.h>
123
124 #if DUMMYNET
125 #include <netinet/ip_dummynet.h>
126 #endif /* DUMMYNET */
127
128 /*
129 * For RandomULong(), to get a 32 bits random value
130 * Note that random() returns a 31 bits value, see rdar://11159750
131 */
132 #include <dev/random/randomdev.h>
133
134 #define DPFPRINTF(n, x) (pf_status.debug >= (n) ? printf x : ((void)0))
135
136 /*
137 * On Mac OS X, the rtableid value is treated as the interface scope
138 * value that is equivalent to the interface index used for scoped
139 * routing. A valid scope value is anything but IFSCOPE_NONE (0),
140 * as per definition of ifindex which is a positive, non-zero number.
141 * The other BSDs treat a negative rtableid value as invalid, hence
142 * the test against INT_MAX to handle userland apps which initialize
143 * the field with a negative number.
144 */
145 #define PF_RTABLEID_IS_VALID(r) \
146 ((r) > IFSCOPE_NONE && (r) <= INT_MAX)
147
148 /*
149 * Global variables
150 */
151 decl_lck_mtx_data(, pf_lock_data);
152 decl_lck_rw_data(, pf_perim_lock_data);
153 lck_mtx_t *pf_lock = &pf_lock_data;
154 lck_rw_t *pf_perim_lock = &pf_perim_lock_data;
155
156 /* state tables */
157 struct pf_state_tree_lan_ext pf_statetbl_lan_ext;
158 struct pf_state_tree_ext_gwy pf_statetbl_ext_gwy;
159
160 struct pf_palist pf_pabuf;
161 struct pf_status pf_status;
162
163 u_int32_t ticket_pabuf;
164
165 static MD5_CTX pf_tcp_secret_ctx;
166 static u_char pf_tcp_secret[16];
167 static int pf_tcp_secret_init;
168 static int pf_tcp_iss_off;
169
170 static struct pf_anchor_stackframe {
171 struct pf_ruleset *rs;
172 struct pf_rule *r;
173 struct pf_anchor_node *parent;
174 struct pf_anchor *child;
175 } pf_anchor_stack[64];
176
177 struct pool pf_src_tree_pl, pf_rule_pl, pf_pooladdr_pl;
178 struct pool pf_state_pl, pf_state_key_pl;
179
180 typedef void (*hook_fn_t)(void *);
181
182 struct hook_desc {
183 TAILQ_ENTRY(hook_desc) hd_list;
184 hook_fn_t hd_fn;
185 void *hd_arg;
186 };
187
188 #define HOOK_REMOVE 0x01
189 #define HOOK_FREE 0x02
190 #define HOOK_ABORT 0x04
191
192 static void *hook_establish(struct hook_desc_head *, int,
193 hook_fn_t, void *);
194 static void hook_runloop(struct hook_desc_head *, int flags);
195
196 struct pool pf_app_state_pl;
197 static void pf_print_addr(struct pf_addr *addr, sa_family_t af);
198 static void pf_print_sk_host(struct pf_state_host *, u_int8_t, int,
199 u_int8_t);
200
201 static void pf_print_host(struct pf_addr *, u_int16_t, u_int8_t);
202
203 static void pf_init_threshold(struct pf_threshold *, u_int32_t,
204 u_int32_t);
205 static void pf_add_threshold(struct pf_threshold *);
206 static int pf_check_threshold(struct pf_threshold *);
207
208 static void pf_change_ap(int, pbuf_t *, struct pf_addr *,
209 u_int16_t *, u_int16_t *, u_int16_t *,
210 struct pf_addr *, u_int16_t, u_int8_t, sa_family_t,
211 sa_family_t, int);
212 static int pf_modulate_sack(pbuf_t *, int, struct pf_pdesc *,
213 struct tcphdr *, struct pf_state_peer *);
214 static void pf_change_a6(struct pf_addr *, u_int16_t *,
215 struct pf_addr *, u_int8_t);
216 void pf_change_addr(struct pf_addr *a, u_int16_t *c,
217 struct pf_addr *an, u_int8_t u,
218 sa_family_t af, sa_family_t afn);
219 static void pf_change_icmp(struct pf_addr *, u_int16_t *,
220 struct pf_addr *, struct pf_addr *, u_int16_t,
221 u_int16_t *, u_int16_t *, u_int16_t *,
222 u_int16_t *, u_int8_t, sa_family_t);
223 static void pf_send_tcp(const struct pf_rule *, sa_family_t,
224 const struct pf_addr *, const struct pf_addr *,
225 u_int16_t, u_int16_t, u_int32_t, u_int32_t,
226 u_int8_t, u_int16_t, u_int16_t, u_int8_t, int,
227 u_int16_t, struct ether_header *, struct ifnet *);
228 static void pf_send_icmp(pbuf_t *, u_int8_t, u_int8_t,
229 sa_family_t, struct pf_rule *);
230 static struct pf_rule *pf_match_translation(struct pf_pdesc *, pbuf_t *,
231 int, int, struct pfi_kif *, struct pf_addr *,
232 union pf_state_xport *, struct pf_addr *,
233 union pf_state_xport *, int);
234 static struct pf_rule *pf_get_translation_aux(struct pf_pdesc *,
235 pbuf_t *, int, int, struct pfi_kif *,
236 struct pf_src_node **, struct pf_addr *,
237 union pf_state_xport *, struct pf_addr *,
238 union pf_state_xport *, union pf_state_xport *
239 );
240 static void pf_attach_state(struct pf_state_key *,
241 struct pf_state *, int);
242 static void pf_detach_state(struct pf_state *, int);
243 static u_int32_t pf_tcp_iss(struct pf_pdesc *);
244 static int pf_test_rule(struct pf_rule **, struct pf_state **,
245 int, struct pfi_kif *, pbuf_t *, int,
246 void *, struct pf_pdesc *, struct pf_rule **,
247 struct pf_ruleset **, struct ifqueue *);
248 #if DUMMYNET
249 static int pf_test_dummynet(struct pf_rule **, int,
250 struct pfi_kif *, pbuf_t **,
251 struct pf_pdesc *, struct ip_fw_args *);
252 #endif /* DUMMYNET */
253 static int pf_test_fragment(struct pf_rule **, int,
254 struct pfi_kif *, pbuf_t *, void *,
255 struct pf_pdesc *, struct pf_rule **,
256 struct pf_ruleset **);
257 static int pf_test_state_tcp(struct pf_state **, int,
258 struct pfi_kif *, pbuf_t *, int,
259 void *, struct pf_pdesc *, u_short *);
260 static int pf_test_state_udp(struct pf_state **, int,
261 struct pfi_kif *, pbuf_t *, int,
262 void *, struct pf_pdesc *, u_short *);
263 static int pf_test_state_icmp(struct pf_state **, int,
264 struct pfi_kif *, pbuf_t *, int,
265 void *, struct pf_pdesc *, u_short *);
266 static int pf_test_state_other(struct pf_state **, int,
267 struct pfi_kif *, struct pf_pdesc *);
268 static int pf_match_tag(struct pf_rule *,
269 struct pf_mtag *, int *);
270 static void pf_hash(struct pf_addr *, struct pf_addr *,
271 struct pf_poolhashkey *, sa_family_t);
272 static int pf_map_addr(u_int8_t, struct pf_rule *,
273 struct pf_addr *, struct pf_addr *,
274 struct pf_addr *, struct pf_src_node **);
275 static int pf_get_sport(struct pf_pdesc *, struct pfi_kif *,
276 struct pf_rule *, struct pf_addr *,
277 union pf_state_xport *, struct pf_addr *,
278 union pf_state_xport *, struct pf_addr *,
279 union pf_state_xport *, struct pf_src_node **
280 );
281 static void pf_route(pbuf_t **, struct pf_rule *, int,
282 struct ifnet *, struct pf_state *,
283 struct pf_pdesc *);
284 static void pf_route6(pbuf_t **, struct pf_rule *, int,
285 struct ifnet *, struct pf_state *,
286 struct pf_pdesc *);
287 static u_int8_t pf_get_wscale(pbuf_t *, int, u_int16_t,
288 sa_family_t);
289 static u_int16_t pf_get_mss(pbuf_t *, int, u_int16_t,
290 sa_family_t);
291 static u_int16_t pf_calc_mss(struct pf_addr *, sa_family_t,
292 u_int16_t);
293 static void pf_set_rt_ifp(struct pf_state *,
294 struct pf_addr *, sa_family_t af);
295 static int pf_check_proto_cksum(pbuf_t *, int, int,
296 u_int8_t, sa_family_t);
297 static int pf_addr_wrap_neq(struct pf_addr_wrap *,
298 struct pf_addr_wrap *);
299 static struct pf_state *pf_find_state(struct pfi_kif *,
300 struct pf_state_key_cmp *, u_int);
301 static int pf_src_connlimit(struct pf_state **);
302 static void pf_stateins_err(const char *, struct pf_state *,
303 struct pfi_kif *);
304 static int pf_check_congestion(struct ifqueue *);
305
306 #if 0
307 static const char *pf_pptp_ctrl_type_name(u_int16_t code);
308 #endif
309 static void pf_pptp_handler(struct pf_state *, int, int,
310 struct pf_pdesc *, struct pfi_kif *);
311 static void pf_pptp_unlink(struct pf_state *);
312 static void pf_grev1_unlink(struct pf_state *);
313 static int pf_test_state_grev1(struct pf_state **, int,
314 struct pfi_kif *, int, struct pf_pdesc *);
315 static int pf_ike_compare(struct pf_app_state *,
316 struct pf_app_state *);
317 static int pf_test_state_esp(struct pf_state **, int,
318 struct pfi_kif *, int, struct pf_pdesc *);
319
320 extern struct pool pfr_ktable_pl;
321 extern struct pool pfr_kentry_pl;
322 extern int path_mtu_discovery;
323
324 struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX] = {
325 { .pp = &pf_state_pl, .limit = PFSTATE_HIWAT },
326 { .pp = &pf_app_state_pl, .limit = PFAPPSTATE_HIWAT },
327 { .pp = &pf_src_tree_pl, .limit = PFSNODE_HIWAT },
328 { .pp = &pf_frent_pl, .limit = PFFRAG_FRENT_HIWAT },
329 { .pp = &pfr_ktable_pl, .limit = PFR_KTABLE_HIWAT },
330 { .pp = &pfr_kentry_pl, .limit = PFR_KENTRY_HIWAT },
331 };
332
333 void *
334 pf_lazy_makewritable(struct pf_pdesc *pd, pbuf_t *pbuf, int len)
335 {
336 void *p;
337
338 if (pd->lmw < 0) {
339 return NULL;
340 }
341
342 VERIFY(pbuf == pd->mp);
343
344 p = pbuf->pb_data;
345 if (len > pd->lmw) {
346 if ((p = pbuf_ensure_writable(pbuf, len)) == NULL) {
347 len = -1;
348 }
349 pd->lmw = len;
350 if (len >= 0) {
351 pd->pf_mtag = pf_find_mtag_pbuf(pbuf);
352
353 switch (pd->af) {
354 case AF_INET: {
355 struct ip *h = p;
356 pd->src = (struct pf_addr *)(uintptr_t)&h->ip_src;
357 pd->dst = (struct pf_addr *)(uintptr_t)&h->ip_dst;
358 pd->ip_sum = &h->ip_sum;
359 break;
360 }
361 case AF_INET6: {
362 struct ip6_hdr *h = p;
363 pd->src = (struct pf_addr *)(uintptr_t)&h->ip6_src;
364 pd->dst = (struct pf_addr *)(uintptr_t)&h->ip6_dst;
365 break;
366 }
367 }
368 }
369 }
370
371 return len < 0 ? NULL : p;
372 }
373
374 static const int *
375 pf_state_lookup_aux(struct pf_state **state, struct pfi_kif *kif,
376 int direction, int *action)
377 {
378 if (*state == NULL || (*state)->timeout == PFTM_PURGE) {
379 *action = PF_DROP;
380 return action;
381 }
382
383 if (direction == PF_OUT &&
384 (((*state)->rule.ptr->rt == PF_ROUTETO &&
385 (*state)->rule.ptr->direction == PF_OUT) ||
386 ((*state)->rule.ptr->rt == PF_REPLYTO &&
387 (*state)->rule.ptr->direction == PF_IN)) &&
388 (*state)->rt_kif != NULL && (*state)->rt_kif != kif) {
389 *action = PF_PASS;
390 return action;
391 }
392
393 return 0;
394 }
395
396 #define STATE_LOOKUP() \
397 do { \
398 int action; \
399 *state = pf_find_state(kif, &key, direction); \
400 if (*state != NULL && pd != NULL && \
401 !(pd->pktflags & PKTF_FLOW_ID)) { \
402 pd->flowsrc = (*state)->state_key->flowsrc; \
403 pd->flowhash = (*state)->state_key->flowhash; \
404 if (pd->flowhash != 0) { \
405 pd->pktflags |= PKTF_FLOW_ID; \
406 pd->pktflags &= ~PKTF_FLOW_ADV; \
407 } \
408 } \
409 if (pf_state_lookup_aux(state, kif, direction, &action)) \
410 return (action); \
411 } while (0)
412
413 #define STATE_ADDR_TRANSLATE(sk) \
414 (sk)->lan.addr.addr32[0] != (sk)->gwy.addr.addr32[0] || \
415 ((sk)->af_lan == AF_INET6 && \
416 ((sk)->lan.addr.addr32[1] != (sk)->gwy.addr.addr32[1] || \
417 (sk)->lan.addr.addr32[2] != (sk)->gwy.addr.addr32[2] || \
418 (sk)->lan.addr.addr32[3] != (sk)->gwy.addr.addr32[3]))
419
420 #define STATE_TRANSLATE(sk) \
421 ((sk)->af_lan != (sk)->af_gwy || \
422 STATE_ADDR_TRANSLATE(sk) || \
423 (sk)->lan.xport.port != (sk)->gwy.xport.port)
424
425 #define STATE_GRE_TRANSLATE(sk) \
426 (STATE_ADDR_TRANSLATE(sk) || \
427 (sk)->lan.xport.call_id != (sk)->gwy.xport.call_id)
428
429 #define BOUND_IFACE(r, k) \
430 ((r)->rule_flag & PFRULE_IFBOUND) ? (k) : pfi_all
431
432 #define STATE_INC_COUNTERS(s) \
433 do { \
434 s->rule.ptr->states++; \
435 VERIFY(s->rule.ptr->states != 0); \
436 if (s->anchor.ptr != NULL) { \
437 s->anchor.ptr->states++; \
438 VERIFY(s->anchor.ptr->states != 0); \
439 } \
440 if (s->nat_rule.ptr != NULL) { \
441 s->nat_rule.ptr->states++; \
442 VERIFY(s->nat_rule.ptr->states != 0); \
443 } \
444 } while (0)
445
446 #define STATE_DEC_COUNTERS(s) \
447 do { \
448 if (s->nat_rule.ptr != NULL) { \
449 VERIFY(s->nat_rule.ptr->states > 0); \
450 s->nat_rule.ptr->states--; \
451 } \
452 if (s->anchor.ptr != NULL) { \
453 VERIFY(s->anchor.ptr->states > 0); \
454 s->anchor.ptr->states--; \
455 } \
456 VERIFY(s->rule.ptr->states > 0); \
457 s->rule.ptr->states--; \
458 } while (0)
459
460 static __inline int pf_src_compare(struct pf_src_node *, struct pf_src_node *);
461 static __inline int pf_state_compare_lan_ext(struct pf_state_key *,
462 struct pf_state_key *);
463 static __inline int pf_state_compare_ext_gwy(struct pf_state_key *,
464 struct pf_state_key *);
465 static __inline int pf_state_compare_id(struct pf_state *,
466 struct pf_state *);
467
468 struct pf_src_tree tree_src_tracking;
469
470 struct pf_state_tree_id tree_id;
471 struct pf_state_queue state_list;
472
473 RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare);
474 RB_GENERATE(pf_state_tree_lan_ext, pf_state_key,
475 entry_lan_ext, pf_state_compare_lan_ext);
476 RB_GENERATE(pf_state_tree_ext_gwy, pf_state_key,
477 entry_ext_gwy, pf_state_compare_ext_gwy);
478 RB_GENERATE(pf_state_tree_id, pf_state,
479 entry_id, pf_state_compare_id);
480
481 #define PF_DT_SKIP_LANEXT 0x01
482 #define PF_DT_SKIP_EXTGWY 0x02
483
484 static const u_int16_t PF_PPTP_PORT = 1723;
485 static const u_int32_t PF_PPTP_MAGIC_NUMBER = 0x1A2B3C4D;
486
487 struct pf_pptp_hdr {
488 u_int16_t length;
489 u_int16_t type;
490 u_int32_t magic;
491 };
492
493 struct pf_pptp_ctrl_hdr {
494 u_int16_t type;
495 u_int16_t reserved_0;
496 };
497
498 struct pf_pptp_ctrl_generic {
499 u_int16_t data[0];
500 };
501
502 #define PF_PPTP_CTRL_TYPE_START_REQ 1
503 struct pf_pptp_ctrl_start_req {
504 u_int16_t protocol_version;
505 u_int16_t reserved_1;
506 u_int32_t framing_capabilities;
507 u_int32_t bearer_capabilities;
508 u_int16_t maximum_channels;
509 u_int16_t firmware_revision;
510 u_int8_t host_name[64];
511 u_int8_t vendor_string[64];
512 };
513
514 #define PF_PPTP_CTRL_TYPE_START_RPY 2
515 struct pf_pptp_ctrl_start_rpy {
516 u_int16_t protocol_version;
517 u_int8_t result_code;
518 u_int8_t error_code;
519 u_int32_t framing_capabilities;
520 u_int32_t bearer_capabilities;
521 u_int16_t maximum_channels;
522 u_int16_t firmware_revision;
523 u_int8_t host_name[64];
524 u_int8_t vendor_string[64];
525 };
526
527 #define PF_PPTP_CTRL_TYPE_STOP_REQ 3
528 struct pf_pptp_ctrl_stop_req {
529 u_int8_t reason;
530 u_int8_t reserved_1;
531 u_int16_t reserved_2;
532 };
533
534 #define PF_PPTP_CTRL_TYPE_STOP_RPY 4
535 struct pf_pptp_ctrl_stop_rpy {
536 u_int8_t reason;
537 u_int8_t error_code;
538 u_int16_t reserved_1;
539 };
540
541 #define PF_PPTP_CTRL_TYPE_ECHO_REQ 5
542 struct pf_pptp_ctrl_echo_req {
543 u_int32_t identifier;
544 };
545
546 #define PF_PPTP_CTRL_TYPE_ECHO_RPY 6
547 struct pf_pptp_ctrl_echo_rpy {
548 u_int32_t identifier;
549 u_int8_t result_code;
550 u_int8_t error_code;
551 u_int16_t reserved_1;
552 };
553
554 #define PF_PPTP_CTRL_TYPE_CALL_OUT_REQ 7
555 struct pf_pptp_ctrl_call_out_req {
556 u_int16_t call_id;
557 u_int16_t call_sernum;
558 u_int32_t min_bps;
559 u_int32_t bearer_type;
560 u_int32_t framing_type;
561 u_int16_t rxwindow_size;
562 u_int16_t proc_delay;
563 u_int8_t phone_num[64];
564 u_int8_t sub_addr[64];
565 };
566
567 #define PF_PPTP_CTRL_TYPE_CALL_OUT_RPY 8
568 struct pf_pptp_ctrl_call_out_rpy {
569 u_int16_t call_id;
570 u_int16_t peer_call_id;
571 u_int8_t result_code;
572 u_int8_t error_code;
573 u_int16_t cause_code;
574 u_int32_t connect_speed;
575 u_int16_t rxwindow_size;
576 u_int16_t proc_delay;
577 u_int32_t phy_channel_id;
578 };
579
580 #define PF_PPTP_CTRL_TYPE_CALL_IN_1ST 9
581 struct pf_pptp_ctrl_call_in_1st {
582 u_int16_t call_id;
583 u_int16_t call_sernum;
584 u_int32_t bearer_type;
585 u_int32_t phy_channel_id;
586 u_int16_t dialed_number_len;
587 u_int16_t dialing_number_len;
588 u_int8_t dialed_num[64];
589 u_int8_t dialing_num[64];
590 u_int8_t sub_addr[64];
591 };
592
593 #define PF_PPTP_CTRL_TYPE_CALL_IN_2ND 10
594 struct pf_pptp_ctrl_call_in_2nd {
595 u_int16_t call_id;
596 u_int16_t peer_call_id;
597 u_int8_t result_code;
598 u_int8_t error_code;
599 u_int16_t rxwindow_size;
600 u_int16_t txdelay;
601 u_int16_t reserved_1;
602 };
603
604 #define PF_PPTP_CTRL_TYPE_CALL_IN_3RD 11
605 struct pf_pptp_ctrl_call_in_3rd {
606 u_int16_t call_id;
607 u_int16_t reserved_1;
608 u_int32_t connect_speed;
609 u_int16_t rxwindow_size;
610 u_int16_t txdelay;
611 u_int32_t framing_type;
612 };
613
614 #define PF_PPTP_CTRL_TYPE_CALL_CLR 12
615 struct pf_pptp_ctrl_call_clr {
616 u_int16_t call_id;
617 u_int16_t reserved_1;
618 };
619
620 #define PF_PPTP_CTRL_TYPE_CALL_DISC 13
621 struct pf_pptp_ctrl_call_disc {
622 u_int16_t call_id;
623 u_int8_t result_code;
624 u_int8_t error_code;
625 u_int16_t cause_code;
626 u_int16_t reserved_1;
627 u_int8_t statistics[128];
628 };
629
630 #define PF_PPTP_CTRL_TYPE_ERROR 14
631 struct pf_pptp_ctrl_error {
632 u_int16_t peer_call_id;
633 u_int16_t reserved_1;
634 u_int32_t crc_errors;
635 u_int32_t fr_errors;
636 u_int32_t hw_errors;
637 u_int32_t buf_errors;
638 u_int32_t tim_errors;
639 u_int32_t align_errors;
640 };
641
642 #define PF_PPTP_CTRL_TYPE_SET_LINKINFO 15
643 struct pf_pptp_ctrl_set_linkinfo {
644 u_int16_t peer_call_id;
645 u_int16_t reserved_1;
646 u_int32_t tx_accm;
647 u_int32_t rx_accm;
648 };
649
650 static const size_t PF_PPTP_CTRL_MSG_MINSIZE =
651 sizeof(struct pf_pptp_hdr) + sizeof(struct pf_pptp_ctrl_hdr);
652
653 union pf_pptp_ctrl_msg_union {
654 struct pf_pptp_ctrl_start_req start_req;
655 struct pf_pptp_ctrl_start_rpy start_rpy;
656 struct pf_pptp_ctrl_stop_req stop_req;
657 struct pf_pptp_ctrl_stop_rpy stop_rpy;
658 struct pf_pptp_ctrl_echo_req echo_req;
659 struct pf_pptp_ctrl_echo_rpy echo_rpy;
660 struct pf_pptp_ctrl_call_out_req call_out_req;
661 struct pf_pptp_ctrl_call_out_rpy call_out_rpy;
662 struct pf_pptp_ctrl_call_in_1st call_in_1st;
663 struct pf_pptp_ctrl_call_in_2nd call_in_2nd;
664 struct pf_pptp_ctrl_call_in_3rd call_in_3rd;
665 struct pf_pptp_ctrl_call_clr call_clr;
666 struct pf_pptp_ctrl_call_disc call_disc;
667 struct pf_pptp_ctrl_error error;
668 struct pf_pptp_ctrl_set_linkinfo set_linkinfo;
669 u_int8_t data[0];
670 };
671
672 struct pf_pptp_ctrl_msg {
673 struct pf_pptp_hdr hdr;
674 struct pf_pptp_ctrl_hdr ctrl;
675 union pf_pptp_ctrl_msg_union msg;
676 };
677
678 #define PF_GRE_FLAG_CHECKSUM_PRESENT 0x8000
679 #define PF_GRE_FLAG_VERSION_MASK 0x0007
680 #define PF_GRE_PPP_ETHERTYPE 0x880B
681
682 struct pf_grev1_hdr {
683 u_int16_t flags;
684 u_int16_t protocol_type;
685 u_int16_t payload_length;
686 u_int16_t call_id;
687 /*
688 * u_int32_t seqno;
689 * u_int32_t ackno;
690 */
691 };
692
693 static const u_int16_t PF_IKE_PORT = 500;
694
695 struct pf_ike_hdr {
696 u_int64_t initiator_cookie, responder_cookie;
697 u_int8_t next_payload, version, exchange_type, flags;
698 u_int32_t message_id, length;
699 };
700
701 #define PF_IKE_PACKET_MINSIZE (sizeof (struct pf_ike_hdr))
702
703 #define PF_IKEv1_EXCHTYPE_BASE 1
704 #define PF_IKEv1_EXCHTYPE_ID_PROTECT 2
705 #define PF_IKEv1_EXCHTYPE_AUTH_ONLY 3
706 #define PF_IKEv1_EXCHTYPE_AGGRESSIVE 4
707 #define PF_IKEv1_EXCHTYPE_INFORMATIONAL 5
708 #define PF_IKEv2_EXCHTYPE_SA_INIT 34
709 #define PF_IKEv2_EXCHTYPE_AUTH 35
710 #define PF_IKEv2_EXCHTYPE_CREATE_CHILD_SA 36
711 #define PF_IKEv2_EXCHTYPE_INFORMATIONAL 37
712
713 #define PF_IKEv1_FLAG_E 0x01
714 #define PF_IKEv1_FLAG_C 0x02
715 #define PF_IKEv1_FLAG_A 0x04
716 #define PF_IKEv2_FLAG_I 0x08
717 #define PF_IKEv2_FLAG_V 0x10
718 #define PF_IKEv2_FLAG_R 0x20
719
720 struct pf_esp_hdr {
721 u_int32_t spi;
722 u_int32_t seqno;
723 u_int8_t payload[];
724 };
725
726 static __inline int
727 pf_addr_compare(struct pf_addr *a, struct pf_addr *b, sa_family_t af)
728 {
729 switch (af) {
730 #ifdef INET
731 case AF_INET:
732 if (a->addr32[0] > b->addr32[0]) {
733 return 1;
734 }
735 if (a->addr32[0] < b->addr32[0]) {
736 return -1;
737 }
738 break;
739 #endif /* INET */
740 case AF_INET6:
741 if (a->addr32[3] > b->addr32[3]) {
742 return 1;
743 }
744 if (a->addr32[3] < b->addr32[3]) {
745 return -1;
746 }
747 if (a->addr32[2] > b->addr32[2]) {
748 return 1;
749 }
750 if (a->addr32[2] < b->addr32[2]) {
751 return -1;
752 }
753 if (a->addr32[1] > b->addr32[1]) {
754 return 1;
755 }
756 if (a->addr32[1] < b->addr32[1]) {
757 return -1;
758 }
759 if (a->addr32[0] > b->addr32[0]) {
760 return 1;
761 }
762 if (a->addr32[0] < b->addr32[0]) {
763 return -1;
764 }
765 break;
766 }
767 return 0;
768 }
769
770 static __inline int
771 pf_src_compare(struct pf_src_node *a, struct pf_src_node *b)
772 {
773 int diff;
774
775 if (a->rule.ptr > b->rule.ptr) {
776 return 1;
777 }
778 if (a->rule.ptr < b->rule.ptr) {
779 return -1;
780 }
781 if ((diff = a->af - b->af) != 0) {
782 return diff;
783 }
784 if ((diff = pf_addr_compare(&a->addr, &b->addr, a->af)) != 0) {
785 return diff;
786 }
787 return 0;
788 }
789
790 static __inline int
791 pf_state_compare_lan_ext(struct pf_state_key *a, struct pf_state_key *b)
792 {
793 int diff;
794 int extfilter;
795
796 if ((diff = a->proto - b->proto) != 0) {
797 return diff;
798 }
799 if ((diff = a->af_lan - b->af_lan) != 0) {
800 return diff;
801 }
802
803 extfilter = PF_EXTFILTER_APD;
804
805 switch (a->proto) {
806 case IPPROTO_ICMP:
807 case IPPROTO_ICMPV6:
808 if ((diff = a->lan.xport.port - b->lan.xport.port) != 0) {
809 return diff;
810 }
811 break;
812
813 case IPPROTO_TCP:
814 if ((diff = a->lan.xport.port - b->lan.xport.port) != 0) {
815 return diff;
816 }
817 if ((diff = a->ext_lan.xport.port - b->ext_lan.xport.port) != 0) {
818 return diff;
819 }
820 break;
821
822 case IPPROTO_UDP:
823 if ((diff = a->proto_variant - b->proto_variant)) {
824 return diff;
825 }
826 extfilter = a->proto_variant;
827 if ((diff = a->lan.xport.port - b->lan.xport.port) != 0) {
828 return diff;
829 }
830 if ((extfilter < PF_EXTFILTER_AD) &&
831 (diff = a->ext_lan.xport.port - b->ext_lan.xport.port) != 0) {
832 return diff;
833 }
834 break;
835
836 case IPPROTO_GRE:
837 if (a->proto_variant == PF_GRE_PPTP_VARIANT &&
838 a->proto_variant == b->proto_variant) {
839 if (!!(diff = a->ext_lan.xport.call_id -
840 b->ext_lan.xport.call_id)) {
841 return diff;
842 }
843 }
844 break;
845
846 case IPPROTO_ESP:
847 if (!!(diff = a->ext_lan.xport.spi - b->ext_lan.xport.spi)) {
848 return diff;
849 }
850 break;
851
852 default:
853 break;
854 }
855
856 switch (a->af_lan) {
857 #if INET
858 case AF_INET:
859 if ((diff = pf_addr_compare(&a->lan.addr, &b->lan.addr,
860 a->af_lan)) != 0) {
861 return diff;
862 }
863
864 if (extfilter < PF_EXTFILTER_EI) {
865 if ((diff = pf_addr_compare(&a->ext_lan.addr,
866 &b->ext_lan.addr,
867 a->af_lan)) != 0) {
868 return diff;
869 }
870 }
871 break;
872 #endif /* INET */
873 case AF_INET6:
874 if ((diff = pf_addr_compare(&a->lan.addr, &b->lan.addr,
875 a->af_lan)) != 0) {
876 return diff;
877 }
878
879 if (extfilter < PF_EXTFILTER_EI ||
880 !PF_AZERO(&b->ext_lan.addr, AF_INET6)) {
881 if ((diff = pf_addr_compare(&a->ext_lan.addr,
882 &b->ext_lan.addr,
883 a->af_lan)) != 0) {
884 return diff;
885 }
886 }
887 break;
888 }
889
890 if (a->app_state && b->app_state) {
891 if (a->app_state->compare_lan_ext &&
892 b->app_state->compare_lan_ext) {
893 diff = (const char *)b->app_state->compare_lan_ext -
894 (const char *)a->app_state->compare_lan_ext;
895 if (diff != 0) {
896 return diff;
897 }
898 diff = a->app_state->compare_lan_ext(a->app_state,
899 b->app_state);
900 if (diff != 0) {
901 return diff;
902 }
903 }
904 }
905
906 return 0;
907 }
908
909 static __inline int
910 pf_state_compare_ext_gwy(struct pf_state_key *a, struct pf_state_key *b)
911 {
912 int diff;
913 int extfilter;
914
915 if ((diff = a->proto - b->proto) != 0) {
916 return diff;
917 }
918
919 if ((diff = a->af_gwy - b->af_gwy) != 0) {
920 return diff;
921 }
922
923 extfilter = PF_EXTFILTER_APD;
924
925 switch (a->proto) {
926 case IPPROTO_ICMP:
927 case IPPROTO_ICMPV6:
928 if ((diff = a->gwy.xport.port - b->gwy.xport.port) != 0) {
929 return diff;
930 }
931 break;
932
933 case IPPROTO_TCP:
934 if ((diff = a->ext_gwy.xport.port - b->ext_gwy.xport.port) != 0) {
935 return diff;
936 }
937 if ((diff = a->gwy.xport.port - b->gwy.xport.port) != 0) {
938 return diff;
939 }
940 break;
941
942 case IPPROTO_UDP:
943 if ((diff = a->proto_variant - b->proto_variant)) {
944 return diff;
945 }
946 extfilter = a->proto_variant;
947 if ((diff = a->gwy.xport.port - b->gwy.xport.port) != 0) {
948 return diff;
949 }
950 if ((extfilter < PF_EXTFILTER_AD) &&
951 (diff = a->ext_gwy.xport.port - b->ext_gwy.xport.port) != 0) {
952 return diff;
953 }
954 break;
955
956 case IPPROTO_GRE:
957 if (a->proto_variant == PF_GRE_PPTP_VARIANT &&
958 a->proto_variant == b->proto_variant) {
959 if (!!(diff = a->gwy.xport.call_id -
960 b->gwy.xport.call_id)) {
961 return diff;
962 }
963 }
964 break;
965
966 case IPPROTO_ESP:
967 if (!!(diff = a->gwy.xport.spi - b->gwy.xport.spi)) {
968 return diff;
969 }
970 break;
971
972 default:
973 break;
974 }
975
976 switch (a->af_gwy) {
977 #if INET
978 case AF_INET:
979 if ((diff = pf_addr_compare(&a->gwy.addr, &b->gwy.addr,
980 a->af_gwy)) != 0) {
981 return diff;
982 }
983
984 if (extfilter < PF_EXTFILTER_EI) {
985 if ((diff = pf_addr_compare(&a->ext_gwy.addr, &b->ext_gwy.addr,
986 a->af_gwy)) != 0) {
987 return diff;
988 }
989 }
990 break;
991 #endif /* INET */
992 case AF_INET6:
993 if ((diff = pf_addr_compare(&a->gwy.addr, &b->gwy.addr,
994 a->af_gwy)) != 0) {
995 return diff;
996 }
997
998 if (extfilter < PF_EXTFILTER_EI ||
999 !PF_AZERO(&b->ext_gwy.addr, AF_INET6)) {
1000 if ((diff = pf_addr_compare(&a->ext_gwy.addr, &b->ext_gwy.addr,
1001 a->af_gwy)) != 0) {
1002 return diff;
1003 }
1004 }
1005 break;
1006 }
1007
1008 if (a->app_state && b->app_state) {
1009 if (a->app_state->compare_ext_gwy &&
1010 b->app_state->compare_ext_gwy) {
1011 diff = (const char *)b->app_state->compare_ext_gwy -
1012 (const char *)a->app_state->compare_ext_gwy;
1013 if (diff != 0) {
1014 return diff;
1015 }
1016 diff = a->app_state->compare_ext_gwy(a->app_state,
1017 b->app_state);
1018 if (diff != 0) {
1019 return diff;
1020 }
1021 }
1022 }
1023
1024 return 0;
1025 }
1026
1027 static __inline int
1028 pf_state_compare_id(struct pf_state *a, struct pf_state *b)
1029 {
1030 if (a->id > b->id) {
1031 return 1;
1032 }
1033 if (a->id < b->id) {
1034 return -1;
1035 }
1036 if (a->creatorid > b->creatorid) {
1037 return 1;
1038 }
1039 if (a->creatorid < b->creatorid) {
1040 return -1;
1041 }
1042
1043 return 0;
1044 }
1045
1046 void
1047 pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af)
1048 {
1049 switch (af) {
1050 #if INET
1051 case AF_INET:
1052 dst->addr32[0] = src->addr32[0];
1053 break;
1054 #endif /* INET */
1055 case AF_INET6:
1056 dst->addr32[0] = src->addr32[0];
1057 dst->addr32[1] = src->addr32[1];
1058 dst->addr32[2] = src->addr32[2];
1059 dst->addr32[3] = src->addr32[3];
1060 break;
1061 }
1062 }
1063
1064 struct pf_state *
1065 pf_find_state_byid(struct pf_state_cmp *key)
1066 {
1067 pf_status.fcounters[FCNT_STATE_SEARCH]++;
1068
1069 return RB_FIND(pf_state_tree_id, &tree_id,
1070 (struct pf_state *)(void *)key);
1071 }
1072
1073 static struct pf_state *
1074 pf_find_state(struct pfi_kif *kif, struct pf_state_key_cmp *key, u_int dir)
1075 {
1076 struct pf_state_key *sk = NULL;
1077 struct pf_state *s;
1078
1079 pf_status.fcounters[FCNT_STATE_SEARCH]++;
1080
1081 switch (dir) {
1082 case PF_OUT:
1083 sk = RB_FIND(pf_state_tree_lan_ext, &pf_statetbl_lan_ext,
1084 (struct pf_state_key *)key);
1085 break;
1086 case PF_IN:
1087 sk = RB_FIND(pf_state_tree_ext_gwy, &pf_statetbl_ext_gwy,
1088 (struct pf_state_key *)key);
1089 /*
1090 * NAT64 is done only on input, for packets coming in from
1091 * from the LAN side, need to lookup the lan_ext tree.
1092 */
1093 if (sk == NULL) {
1094 sk = RB_FIND(pf_state_tree_lan_ext,
1095 &pf_statetbl_lan_ext,
1096 (struct pf_state_key *)key);
1097 if (sk && sk->af_lan == sk->af_gwy) {
1098 sk = NULL;
1099 }
1100 }
1101 break;
1102 default:
1103 panic("pf_find_state");
1104 }
1105
1106 /* list is sorted, if-bound states before floating ones */
1107 if (sk != NULL) {
1108 TAILQ_FOREACH(s, &sk->states, next)
1109 if (s->kif == pfi_all || s->kif == kif) {
1110 return s;
1111 }
1112 }
1113
1114 return NULL;
1115 }
1116
1117 struct pf_state *
1118 pf_find_state_all(struct pf_state_key_cmp *key, u_int dir, int *more)
1119 {
1120 struct pf_state_key *sk = NULL;
1121 struct pf_state *s, *ret = NULL;
1122
1123 pf_status.fcounters[FCNT_STATE_SEARCH]++;
1124
1125 switch (dir) {
1126 case PF_OUT:
1127 sk = RB_FIND(pf_state_tree_lan_ext,
1128 &pf_statetbl_lan_ext, (struct pf_state_key *)key);
1129 break;
1130 case PF_IN:
1131 sk = RB_FIND(pf_state_tree_ext_gwy,
1132 &pf_statetbl_ext_gwy, (struct pf_state_key *)key);
1133 /*
1134 * NAT64 is done only on input, for packets coming in from
1135 * from the LAN side, need to lookup the lan_ext tree.
1136 */
1137 if ((sk == NULL) && pf_nat64_configured) {
1138 sk = RB_FIND(pf_state_tree_lan_ext,
1139 &pf_statetbl_lan_ext,
1140 (struct pf_state_key *)key);
1141 if (sk && sk->af_lan == sk->af_gwy) {
1142 sk = NULL;
1143 }
1144 }
1145 break;
1146 default:
1147 panic("pf_find_state_all");
1148 }
1149
1150 if (sk != NULL) {
1151 ret = TAILQ_FIRST(&sk->states);
1152 if (more == NULL) {
1153 return ret;
1154 }
1155
1156 TAILQ_FOREACH(s, &sk->states, next)
1157 (*more)++;
1158 }
1159
1160 return ret;
1161 }
1162
1163 static void
1164 pf_init_threshold(struct pf_threshold *threshold,
1165 u_int32_t limit, u_int32_t seconds)
1166 {
1167 threshold->limit = limit * PF_THRESHOLD_MULT;
1168 threshold->seconds = seconds;
1169 threshold->count = 0;
1170 threshold->last = pf_time_second();
1171 }
1172
1173 static void
1174 pf_add_threshold(struct pf_threshold *threshold)
1175 {
1176 u_int32_t t = pf_time_second(), diff = t - threshold->last;
1177
1178 if (diff >= threshold->seconds) {
1179 threshold->count = 0;
1180 } else {
1181 threshold->count -= threshold->count * diff /
1182 threshold->seconds;
1183 }
1184 threshold->count += PF_THRESHOLD_MULT;
1185 threshold->last = t;
1186 }
1187
1188 static int
1189 pf_check_threshold(struct pf_threshold *threshold)
1190 {
1191 return threshold->count > threshold->limit;
1192 }
1193
1194 static int
1195 pf_src_connlimit(struct pf_state **state)
1196 {
1197 int bad = 0;
1198 (*state)->src_node->conn++;
1199 VERIFY((*state)->src_node->conn != 0);
1200 (*state)->src.tcp_est = 1;
1201 pf_add_threshold(&(*state)->src_node->conn_rate);
1202
1203 if ((*state)->rule.ptr->max_src_conn &&
1204 (*state)->rule.ptr->max_src_conn <
1205 (*state)->src_node->conn) {
1206 pf_status.lcounters[LCNT_SRCCONN]++;
1207 bad++;
1208 }
1209
1210 if ((*state)->rule.ptr->max_src_conn_rate.limit &&
1211 pf_check_threshold(&(*state)->src_node->conn_rate)) {
1212 pf_status.lcounters[LCNT_SRCCONNRATE]++;
1213 bad++;
1214 }
1215
1216 if (!bad) {
1217 return 0;
1218 }
1219
1220 if ((*state)->rule.ptr->overload_tbl) {
1221 struct pfr_addr p;
1222 u_int32_t killed = 0;
1223
1224 pf_status.lcounters[LCNT_OVERLOAD_TABLE]++;
1225 if (pf_status.debug >= PF_DEBUG_MISC) {
1226 printf("pf_src_connlimit: blocking address ");
1227 pf_print_host(&(*state)->src_node->addr, 0,
1228 (*state)->state_key->af_lan);
1229 }
1230
1231 bzero(&p, sizeof(p));
1232 p.pfra_af = (*state)->state_key->af_lan;
1233 switch ((*state)->state_key->af_lan) {
1234 #if INET
1235 case AF_INET:
1236 p.pfra_net = 32;
1237 p.pfra_ip4addr = (*state)->src_node->addr.v4addr;
1238 break;
1239 #endif /* INET */
1240 case AF_INET6:
1241 p.pfra_net = 128;
1242 p.pfra_ip6addr = (*state)->src_node->addr.v6addr;
1243 break;
1244 }
1245
1246 pfr_insert_kentry((*state)->rule.ptr->overload_tbl,
1247 &p, pf_calendar_time_second());
1248
1249 /* kill existing states if that's required. */
1250 if ((*state)->rule.ptr->flush) {
1251 struct pf_state_key *sk;
1252 struct pf_state *st;
1253
1254 pf_status.lcounters[LCNT_OVERLOAD_FLUSH]++;
1255 RB_FOREACH(st, pf_state_tree_id, &tree_id) {
1256 sk = st->state_key;
1257 /*
1258 * Kill states from this source. (Only those
1259 * from the same rule if PF_FLUSH_GLOBAL is not
1260 * set)
1261 */
1262 if (sk->af_lan ==
1263 (*state)->state_key->af_lan &&
1264 (((*state)->state_key->direction ==
1265 PF_OUT &&
1266 PF_AEQ(&(*state)->src_node->addr,
1267 &sk->lan.addr, sk->af_lan)) ||
1268 ((*state)->state_key->direction == PF_IN &&
1269 PF_AEQ(&(*state)->src_node->addr,
1270 &sk->ext_lan.addr, sk->af_lan))) &&
1271 ((*state)->rule.ptr->flush &
1272 PF_FLUSH_GLOBAL ||
1273 (*state)->rule.ptr == st->rule.ptr)) {
1274 st->timeout = PFTM_PURGE;
1275 st->src.state = st->dst.state =
1276 TCPS_CLOSED;
1277 killed++;
1278 }
1279 }
1280 if (pf_status.debug >= PF_DEBUG_MISC) {
1281 printf(", %u states killed", killed);
1282 }
1283 }
1284 if (pf_status.debug >= PF_DEBUG_MISC) {
1285 printf("\n");
1286 }
1287 }
1288
1289 /* kill this state */
1290 (*state)->timeout = PFTM_PURGE;
1291 (*state)->src.state = (*state)->dst.state = TCPS_CLOSED;
1292 return 1;
1293 }
1294
1295 int
1296 pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule,
1297 struct pf_addr *src, sa_family_t af)
1298 {
1299 struct pf_src_node k;
1300
1301 if (*sn == NULL) {
1302 k.af = af;
1303 PF_ACPY(&k.addr, src, af);
1304 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
1305 rule->rpool.opts & PF_POOL_STICKYADDR) {
1306 k.rule.ptr = rule;
1307 } else {
1308 k.rule.ptr = NULL;
1309 }
1310 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
1311 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
1312 }
1313 if (*sn == NULL) {
1314 if (!rule->max_src_nodes ||
1315 rule->src_nodes < rule->max_src_nodes) {
1316 (*sn) = pool_get(&pf_src_tree_pl, PR_WAITOK);
1317 } else {
1318 pf_status.lcounters[LCNT_SRCNODES]++;
1319 }
1320 if ((*sn) == NULL) {
1321 return -1;
1322 }
1323 bzero(*sn, sizeof(struct pf_src_node));
1324
1325 pf_init_threshold(&(*sn)->conn_rate,
1326 rule->max_src_conn_rate.limit,
1327 rule->max_src_conn_rate.seconds);
1328
1329 (*sn)->af = af;
1330 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
1331 rule->rpool.opts & PF_POOL_STICKYADDR) {
1332 (*sn)->rule.ptr = rule;
1333 } else {
1334 (*sn)->rule.ptr = NULL;
1335 }
1336 PF_ACPY(&(*sn)->addr, src, af);
1337 if (RB_INSERT(pf_src_tree,
1338 &tree_src_tracking, *sn) != NULL) {
1339 if (pf_status.debug >= PF_DEBUG_MISC) {
1340 printf("pf: src_tree insert failed: ");
1341 pf_print_host(&(*sn)->addr, 0, af);
1342 printf("\n");
1343 }
1344 pool_put(&pf_src_tree_pl, *sn);
1345 return -1;
1346 }
1347 (*sn)->creation = pf_time_second();
1348 (*sn)->ruletype = rule->action;
1349 if ((*sn)->rule.ptr != NULL) {
1350 (*sn)->rule.ptr->src_nodes++;
1351 }
1352 pf_status.scounters[SCNT_SRC_NODE_INSERT]++;
1353 pf_status.src_nodes++;
1354 } else {
1355 if (rule->max_src_states &&
1356 (*sn)->states >= rule->max_src_states) {
1357 pf_status.lcounters[LCNT_SRCSTATES]++;
1358 return -1;
1359 }
1360 }
1361 return 0;
1362 }
1363
1364 static void
1365 pf_stateins_err(const char *tree, struct pf_state *s, struct pfi_kif *kif)
1366 {
1367 struct pf_state_key *sk = s->state_key;
1368
1369 if (pf_status.debug >= PF_DEBUG_MISC) {
1370 printf("pf: state insert failed: %s %s ", tree, kif->pfik_name);
1371 switch (sk->proto) {
1372 case IPPROTO_TCP:
1373 printf("TCP");
1374 break;
1375 case IPPROTO_UDP:
1376 printf("UDP");
1377 break;
1378 case IPPROTO_ICMP:
1379 printf("ICMP4");
1380 break;
1381 case IPPROTO_ICMPV6:
1382 printf("ICMP6");
1383 break;
1384 default:
1385 printf("PROTO=%u", sk->proto);
1386 break;
1387 }
1388 printf(" lan: ");
1389 pf_print_sk_host(&sk->lan, sk->af_lan, sk->proto,
1390 sk->proto_variant);
1391 printf(" gwy: ");
1392 pf_print_sk_host(&sk->gwy, sk->af_gwy, sk->proto,
1393 sk->proto_variant);
1394 printf(" ext_lan: ");
1395 pf_print_sk_host(&sk->ext_lan, sk->af_lan, sk->proto,
1396 sk->proto_variant);
1397 printf(" ext_gwy: ");
1398 pf_print_sk_host(&sk->ext_gwy, sk->af_gwy, sk->proto,
1399 sk->proto_variant);
1400 if (s->sync_flags & PFSTATE_FROMSYNC) {
1401 printf(" (from sync)");
1402 }
1403 printf("\n");
1404 }
1405 }
1406
1407 int
1408 pf_insert_state(struct pfi_kif *kif, struct pf_state *s)
1409 {
1410 struct pf_state_key *cur;
1411 struct pf_state *sp;
1412
1413 VERIFY(s->state_key != NULL);
1414 s->kif = kif;
1415
1416 if ((cur = RB_INSERT(pf_state_tree_lan_ext, &pf_statetbl_lan_ext,
1417 s->state_key)) != NULL) {
1418 /* key exists. check for same kif, if none, add to key */
1419 TAILQ_FOREACH(sp, &cur->states, next)
1420 if (sp->kif == kif) { /* collision! */
1421 pf_stateins_err("tree_lan_ext", s, kif);
1422 pf_detach_state(s,
1423 PF_DT_SKIP_LANEXT | PF_DT_SKIP_EXTGWY);
1424 return -1;
1425 }
1426 pf_detach_state(s, PF_DT_SKIP_LANEXT | PF_DT_SKIP_EXTGWY);
1427 pf_attach_state(cur, s, kif == pfi_all ? 1 : 0);
1428 }
1429
1430 /* if cur != NULL, we already found a state key and attached to it */
1431 if (cur == NULL && (cur = RB_INSERT(pf_state_tree_ext_gwy,
1432 &pf_statetbl_ext_gwy, s->state_key)) != NULL) {
1433 /* must not happen. we must have found the sk above! */
1434 pf_stateins_err("tree_ext_gwy", s, kif);
1435 pf_detach_state(s, PF_DT_SKIP_EXTGWY);
1436 return -1;
1437 }
1438
1439 if (s->id == 0 && s->creatorid == 0) {
1440 s->id = htobe64(pf_status.stateid++);
1441 s->creatorid = pf_status.hostid;
1442 }
1443 if (RB_INSERT(pf_state_tree_id, &tree_id, s) != NULL) {
1444 if (pf_status.debug >= PF_DEBUG_MISC) {
1445 printf("pf: state insert failed: "
1446 "id: %016llx creatorid: %08x",
1447 be64toh(s->id), ntohl(s->creatorid));
1448 if (s->sync_flags & PFSTATE_FROMSYNC) {
1449 printf(" (from sync)");
1450 }
1451 printf("\n");
1452 }
1453 pf_detach_state(s, 0);
1454 return -1;
1455 }
1456 TAILQ_INSERT_TAIL(&state_list, s, entry_list);
1457 pf_status.fcounters[FCNT_STATE_INSERT]++;
1458 pf_status.states++;
1459 VERIFY(pf_status.states != 0);
1460 pfi_kif_ref(kif, PFI_KIF_REF_STATE);
1461 #if NPFSYNC
1462 pfsync_insert_state(s);
1463 #endif
1464 return 0;
1465 }
1466
1467 static int
1468 pf_purge_thread_cont(int err)
1469 {
1470 #pragma unused(err)
1471 static u_int32_t nloops = 0;
1472 int t = 1; /* 1 second */
1473
1474 /*
1475 * Update coarse-grained networking timestamp (in sec.); the idea
1476 * is to piggy-back on the periodic timeout callout to update
1477 * the counter returnable via net_uptime().
1478 */
1479 net_update_uptime();
1480
1481 lck_rw_lock_shared(pf_perim_lock);
1482 lck_mtx_lock(pf_lock);
1483
1484 /* purge everything if not running */
1485 if (!pf_status.running) {
1486 pf_purge_expired_states(pf_status.states);
1487 pf_purge_expired_fragments();
1488 pf_purge_expired_src_nodes();
1489
1490 /* terminate thread (we don't currently do this) */
1491 if (pf_purge_thread == NULL) {
1492 lck_mtx_unlock(pf_lock);
1493 lck_rw_done(pf_perim_lock);
1494
1495 thread_deallocate(current_thread());
1496 thread_terminate(current_thread());
1497 /* NOTREACHED */
1498 return 0;
1499 } else {
1500 /* if there's nothing left, sleep w/o timeout */
1501 if (pf_status.states == 0 &&
1502 pf_normalize_isempty() &&
1503 RB_EMPTY(&tree_src_tracking)) {
1504 nloops = 0;
1505 t = 0;
1506 }
1507 goto done;
1508 }
1509 }
1510
1511 /* process a fraction of the state table every second */
1512 pf_purge_expired_states(1 + (pf_status.states
1513 / pf_default_rule.timeout[PFTM_INTERVAL]));
1514
1515 /* purge other expired types every PFTM_INTERVAL seconds */
1516 if (++nloops >= pf_default_rule.timeout[PFTM_INTERVAL]) {
1517 pf_purge_expired_fragments();
1518 pf_purge_expired_src_nodes();
1519 nloops = 0;
1520 }
1521 done:
1522 lck_mtx_unlock(pf_lock);
1523 lck_rw_done(pf_perim_lock);
1524
1525 (void) tsleep0(pf_purge_thread_fn, PWAIT, "pf_purge_cont",
1526 t * hz, pf_purge_thread_cont);
1527 /* NOTREACHED */
1528 VERIFY(0);
1529
1530 return 0;
1531 }
1532
1533 void
1534 pf_purge_thread_fn(void *v, wait_result_t w)
1535 {
1536 #pragma unused(v, w)
1537 (void) tsleep0(pf_purge_thread_fn, PWAIT, "pf_purge", 0,
1538 pf_purge_thread_cont);
1539 /*
1540 * tsleep0() shouldn't have returned as PCATCH was not set;
1541 * therefore assert in this case.
1542 */
1543 VERIFY(0);
1544 }
1545
1546 u_int64_t
1547 pf_state_expires(const struct pf_state *state)
1548 {
1549 u_int32_t t;
1550 u_int32_t start;
1551 u_int32_t end;
1552 u_int32_t states;
1553
1554 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
1555
1556 /* handle all PFTM_* > PFTM_MAX here */
1557 if (state->timeout == PFTM_PURGE) {
1558 return pf_time_second();
1559 }
1560
1561 VERIFY(state->timeout != PFTM_UNLINKED);
1562 VERIFY(state->timeout < PFTM_MAX);
1563 t = state->rule.ptr->timeout[state->timeout];
1564 if (!t) {
1565 t = pf_default_rule.timeout[state->timeout];
1566 }
1567 start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START];
1568 if (start) {
1569 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END];
1570 states = state->rule.ptr->states;
1571 } else {
1572 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START];
1573 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END];
1574 states = pf_status.states;
1575 }
1576 if (end && states > start && start < end) {
1577 if (states < end) {
1578 return state->expire + t * (end - states) /
1579 (end - start);
1580 } else {
1581 return pf_time_second();
1582 }
1583 }
1584 return state->expire + t;
1585 }
1586
1587 void
1588 pf_purge_expired_src_nodes(void)
1589 {
1590 struct pf_src_node *cur, *next;
1591
1592 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
1593
1594 for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) {
1595 next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur);
1596
1597 if (cur->states <= 0 && cur->expire <= pf_time_second()) {
1598 if (cur->rule.ptr != NULL) {
1599 cur->rule.ptr->src_nodes--;
1600 if (cur->rule.ptr->states <= 0 &&
1601 cur->rule.ptr->max_src_nodes <= 0) {
1602 pf_rm_rule(NULL, cur->rule.ptr);
1603 }
1604 }
1605 RB_REMOVE(pf_src_tree, &tree_src_tracking, cur);
1606 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
1607 pf_status.src_nodes--;
1608 pool_put(&pf_src_tree_pl, cur);
1609 }
1610 }
1611 }
1612
1613 void
1614 pf_src_tree_remove_state(struct pf_state *s)
1615 {
1616 u_int32_t t;
1617
1618 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
1619
1620 if (s->src_node != NULL) {
1621 if (s->src.tcp_est) {
1622 VERIFY(s->src_node->conn > 0);
1623 --s->src_node->conn;
1624 }
1625 VERIFY(s->src_node->states > 0);
1626 if (--s->src_node->states <= 0) {
1627 t = s->rule.ptr->timeout[PFTM_SRC_NODE];
1628 if (!t) {
1629 t = pf_default_rule.timeout[PFTM_SRC_NODE];
1630 }
1631 s->src_node->expire = pf_time_second() + t;
1632 }
1633 }
1634 if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) {
1635 VERIFY(s->nat_src_node->states > 0);
1636 if (--s->nat_src_node->states <= 0) {
1637 t = s->rule.ptr->timeout[PFTM_SRC_NODE];
1638 if (!t) {
1639 t = pf_default_rule.timeout[PFTM_SRC_NODE];
1640 }
1641 s->nat_src_node->expire = pf_time_second() + t;
1642 }
1643 }
1644 s->src_node = s->nat_src_node = NULL;
1645 }
1646
1647 void
1648 pf_unlink_state(struct pf_state *cur)
1649 {
1650 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
1651
1652 if (cur->src.state == PF_TCPS_PROXY_DST) {
1653 pf_send_tcp(cur->rule.ptr, cur->state_key->af_lan,
1654 &cur->state_key->ext_lan.addr, &cur->state_key->lan.addr,
1655 cur->state_key->ext_lan.xport.port,
1656 cur->state_key->lan.xport.port,
1657 cur->src.seqhi, cur->src.seqlo + 1,
1658 TH_RST | TH_ACK, 0, 0, 0, 1, cur->tag, NULL, NULL);
1659 }
1660
1661 hook_runloop(&cur->unlink_hooks, HOOK_REMOVE | HOOK_FREE);
1662 RB_REMOVE(pf_state_tree_id, &tree_id, cur);
1663 #if NPFSYNC
1664 if (cur->creatorid == pf_status.hostid) {
1665 pfsync_delete_state(cur);
1666 }
1667 #endif
1668 cur->timeout = PFTM_UNLINKED;
1669 pf_src_tree_remove_state(cur);
1670 pf_detach_state(cur, 0);
1671 }
1672
1673 /* callers should be at splpf and hold the
1674 * write_lock on pf_consistency_lock */
1675 void
1676 pf_free_state(struct pf_state *cur)
1677 {
1678 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
1679 #if NPFSYNC
1680 if (pfsyncif != NULL &&
1681 (pfsyncif->sc_bulk_send_next == cur ||
1682 pfsyncif->sc_bulk_terminator == cur)) {
1683 return;
1684 }
1685 #endif
1686 VERIFY(cur->timeout == PFTM_UNLINKED);
1687 VERIFY(cur->rule.ptr->states > 0);
1688 if (--cur->rule.ptr->states <= 0 &&
1689 cur->rule.ptr->src_nodes <= 0) {
1690 pf_rm_rule(NULL, cur->rule.ptr);
1691 }
1692 if (cur->nat_rule.ptr != NULL) {
1693 VERIFY(cur->nat_rule.ptr->states > 0);
1694 if (--cur->nat_rule.ptr->states <= 0 &&
1695 cur->nat_rule.ptr->src_nodes <= 0) {
1696 pf_rm_rule(NULL, cur->nat_rule.ptr);
1697 }
1698 }
1699 if (cur->anchor.ptr != NULL) {
1700 VERIFY(cur->anchor.ptr->states > 0);
1701 if (--cur->anchor.ptr->states <= 0) {
1702 pf_rm_rule(NULL, cur->anchor.ptr);
1703 }
1704 }
1705 pf_normalize_tcp_cleanup(cur);
1706 pfi_kif_unref(cur->kif, PFI_KIF_REF_STATE);
1707 TAILQ_REMOVE(&state_list, cur, entry_list);
1708 if (cur->tag) {
1709 pf_tag_unref(cur->tag);
1710 }
1711 pool_put(&pf_state_pl, cur);
1712 pf_status.fcounters[FCNT_STATE_REMOVALS]++;
1713 VERIFY(pf_status.states > 0);
1714 pf_status.states--;
1715 }
1716
1717 void
1718 pf_purge_expired_states(u_int32_t maxcheck)
1719 {
1720 static struct pf_state *cur = NULL;
1721 struct pf_state *next;
1722
1723 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
1724
1725 while (maxcheck--) {
1726 /* wrap to start of list when we hit the end */
1727 if (cur == NULL) {
1728 cur = TAILQ_FIRST(&state_list);
1729 if (cur == NULL) {
1730 break; /* list empty */
1731 }
1732 }
1733
1734 /* get next state, as cur may get deleted */
1735 next = TAILQ_NEXT(cur, entry_list);
1736
1737 if (cur->timeout == PFTM_UNLINKED) {
1738 pf_free_state(cur);
1739 } else if (pf_state_expires(cur) <= pf_time_second()) {
1740 /* unlink and free expired state */
1741 pf_unlink_state(cur);
1742 pf_free_state(cur);
1743 }
1744 cur = next;
1745 }
1746 }
1747
1748 int
1749 pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw)
1750 {
1751 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
1752
1753 if (aw->type != PF_ADDR_TABLE) {
1754 return 0;
1755 }
1756 if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL) {
1757 return 1;
1758 }
1759 return 0;
1760 }
1761
1762 void
1763 pf_tbladdr_remove(struct pf_addr_wrap *aw)
1764 {
1765 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
1766
1767 if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL) {
1768 return;
1769 }
1770 pfr_detach_table(aw->p.tbl);
1771 aw->p.tbl = NULL;
1772 }
1773
1774 void
1775 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
1776 {
1777 struct pfr_ktable *kt = aw->p.tbl;
1778
1779 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
1780
1781 if (aw->type != PF_ADDR_TABLE || kt == NULL) {
1782 return;
1783 }
1784 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL) {
1785 kt = kt->pfrkt_root;
1786 }
1787 aw->p.tbl = NULL;
1788 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
1789 kt->pfrkt_cnt : -1;
1790 }
1791
1792 static void
1793 pf_print_addr(struct pf_addr *addr, sa_family_t af)
1794 {
1795 switch (af) {
1796 #if INET
1797 case AF_INET: {
1798 u_int32_t a = ntohl(addr->addr32[0]);
1799 printf("%u.%u.%u.%u", (a >> 24) & 255, (a >> 16) & 255,
1800 (a >> 8) & 255, a & 255);
1801 break;
1802 }
1803 #endif /* INET */
1804 case AF_INET6: {
1805 u_int16_t b;
1806 u_int8_t i, curstart = 255, curend = 0,
1807 maxstart = 0, maxend = 0;
1808 for (i = 0; i < 8; i++) {
1809 if (!addr->addr16[i]) {
1810 if (curstart == 255) {
1811 curstart = i;
1812 } else {
1813 curend = i;
1814 }
1815 } else {
1816 if (curstart) {
1817 if ((curend - curstart) >
1818 (maxend - maxstart)) {
1819 maxstart = curstart;
1820 maxend = curend;
1821 curstart = 255;
1822 }
1823 }
1824 }
1825 }
1826 for (i = 0; i < 8; i++) {
1827 if (i >= maxstart && i <= maxend) {
1828 if (maxend != 7) {
1829 if (i == maxstart) {
1830 printf(":");
1831 }
1832 } else {
1833 if (i == maxend) {
1834 printf(":");
1835 }
1836 }
1837 } else {
1838 b = ntohs(addr->addr16[i]);
1839 printf("%x", b);
1840 if (i < 7) {
1841 printf(":");
1842 }
1843 }
1844 }
1845 break;
1846 }
1847 }
1848 }
1849
1850 static void
1851 pf_print_sk_host(struct pf_state_host *sh, sa_family_t af, int proto,
1852 u_int8_t proto_variant)
1853 {
1854 pf_print_addr(&sh->addr, af);
1855
1856 switch (proto) {
1857 case IPPROTO_ESP:
1858 if (sh->xport.spi) {
1859 printf("[%08x]", ntohl(sh->xport.spi));
1860 }
1861 break;
1862
1863 case IPPROTO_GRE:
1864 if (proto_variant == PF_GRE_PPTP_VARIANT) {
1865 printf("[%u]", ntohs(sh->xport.call_id));
1866 }
1867 break;
1868
1869 case IPPROTO_TCP:
1870 case IPPROTO_UDP:
1871 printf("[%u]", ntohs(sh->xport.port));
1872 break;
1873
1874 default:
1875 break;
1876 }
1877 }
1878
1879 static void
1880 pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af)
1881 {
1882 pf_print_addr(addr, af);
1883 if (p) {
1884 printf("[%u]", ntohs(p));
1885 }
1886 }
1887
1888 void
1889 pf_print_state(struct pf_state *s)
1890 {
1891 struct pf_state_key *sk = s->state_key;
1892 switch (sk->proto) {
1893 case IPPROTO_ESP:
1894 printf("ESP ");
1895 break;
1896 case IPPROTO_GRE:
1897 printf("GRE%u ", sk->proto_variant);
1898 break;
1899 case IPPROTO_TCP:
1900 printf("TCP ");
1901 break;
1902 case IPPROTO_UDP:
1903 printf("UDP ");
1904 break;
1905 case IPPROTO_ICMP:
1906 printf("ICMP ");
1907 break;
1908 case IPPROTO_ICMPV6:
1909 printf("ICMPV6 ");
1910 break;
1911 default:
1912 printf("%u ", sk->proto);
1913 break;
1914 }
1915 pf_print_sk_host(&sk->lan, sk->af_lan, sk->proto, sk->proto_variant);
1916 printf(" ");
1917 pf_print_sk_host(&sk->gwy, sk->af_gwy, sk->proto, sk->proto_variant);
1918 printf(" ");
1919 pf_print_sk_host(&sk->ext_lan, sk->af_lan, sk->proto,
1920 sk->proto_variant);
1921 printf(" ");
1922 pf_print_sk_host(&sk->ext_gwy, sk->af_gwy, sk->proto,
1923 sk->proto_variant);
1924 printf(" [lo=%u high=%u win=%u modulator=%u", s->src.seqlo,
1925 s->src.seqhi, s->src.max_win, s->src.seqdiff);
1926 if (s->src.wscale && s->dst.wscale) {
1927 printf(" wscale=%u", s->src.wscale & PF_WSCALE_MASK);
1928 }
1929 printf("]");
1930 printf(" [lo=%u high=%u win=%u modulator=%u", s->dst.seqlo,
1931 s->dst.seqhi, s->dst.max_win, s->dst.seqdiff);
1932 if (s->src.wscale && s->dst.wscale) {
1933 printf(" wscale=%u", s->dst.wscale & PF_WSCALE_MASK);
1934 }
1935 printf("]");
1936 printf(" %u:%u", s->src.state, s->dst.state);
1937 }
1938
1939 void
1940 pf_print_flags(u_int8_t f)
1941 {
1942 if (f) {
1943 printf(" ");
1944 }
1945 if (f & TH_FIN) {
1946 printf("F");
1947 }
1948 if (f & TH_SYN) {
1949 printf("S");
1950 }
1951 if (f & TH_RST) {
1952 printf("R");
1953 }
1954 if (f & TH_PUSH) {
1955 printf("P");
1956 }
1957 if (f & TH_ACK) {
1958 printf("A");
1959 }
1960 if (f & TH_URG) {
1961 printf("U");
1962 }
1963 if (f & TH_ECE) {
1964 printf("E");
1965 }
1966 if (f & TH_CWR) {
1967 printf("W");
1968 }
1969 }
1970
1971 #define PF_SET_SKIP_STEPS(i) \
1972 do { \
1973 while (head[i] != cur) { \
1974 head[i]->skip[i].ptr = cur; \
1975 head[i] = TAILQ_NEXT(head[i], entries); \
1976 } \
1977 } while (0)
1978
1979 void
1980 pf_calc_skip_steps(struct pf_rulequeue *rules)
1981 {
1982 struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT];
1983 int i;
1984
1985 cur = TAILQ_FIRST(rules);
1986 prev = cur;
1987 for (i = 0; i < PF_SKIP_COUNT; ++i) {
1988 head[i] = cur;
1989 }
1990 while (cur != NULL) {
1991 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot) {
1992 PF_SET_SKIP_STEPS(PF_SKIP_IFP);
1993 }
1994 if (cur->direction != prev->direction) {
1995 PF_SET_SKIP_STEPS(PF_SKIP_DIR);
1996 }
1997 if (cur->af != prev->af) {
1998 PF_SET_SKIP_STEPS(PF_SKIP_AF);
1999 }
2000 if (cur->proto != prev->proto) {
2001 PF_SET_SKIP_STEPS(PF_SKIP_PROTO);
2002 }
2003 if (cur->src.neg != prev->src.neg ||
2004 pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr)) {
2005 PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR);
2006 }
2007 {
2008 union pf_rule_xport *cx = &cur->src.xport;
2009 union pf_rule_xport *px = &prev->src.xport;
2010
2011 switch (cur->proto) {
2012 case IPPROTO_GRE:
2013 case IPPROTO_ESP:
2014 PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
2015 break;
2016 default:
2017 if (prev->proto == IPPROTO_GRE ||
2018 prev->proto == IPPROTO_ESP ||
2019 cx->range.op != px->range.op ||
2020 cx->range.port[0] != px->range.port[0] ||
2021 cx->range.port[1] != px->range.port[1]) {
2022 PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
2023 }
2024 break;
2025 }
2026 }
2027 if (cur->dst.neg != prev->dst.neg ||
2028 pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr)) {
2029 PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR);
2030 }
2031 {
2032 union pf_rule_xport *cx = &cur->dst.xport;
2033 union pf_rule_xport *px = &prev->dst.xport;
2034
2035 switch (cur->proto) {
2036 case IPPROTO_GRE:
2037 if (cur->proto != prev->proto ||
2038 cx->call_id != px->call_id) {
2039 PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
2040 }
2041 break;
2042 case IPPROTO_ESP:
2043 if (cur->proto != prev->proto ||
2044 cx->spi != px->spi) {
2045 PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
2046 }
2047 break;
2048 default:
2049 if (prev->proto == IPPROTO_GRE ||
2050 prev->proto == IPPROTO_ESP ||
2051 cx->range.op != px->range.op ||
2052 cx->range.port[0] != px->range.port[0] ||
2053 cx->range.port[1] != px->range.port[1]) {
2054 PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
2055 }
2056 break;
2057 }
2058 }
2059
2060 prev = cur;
2061 cur = TAILQ_NEXT(cur, entries);
2062 }
2063 for (i = 0; i < PF_SKIP_COUNT; ++i) {
2064 PF_SET_SKIP_STEPS(i);
2065 }
2066 }
2067
2068 u_int32_t
2069 pf_calc_state_key_flowhash(struct pf_state_key *sk)
2070 {
2071 struct pf_flowhash_key fh __attribute__((aligned(8)));
2072 uint32_t flowhash = 0;
2073
2074 bzero(&fh, sizeof(fh));
2075 if (PF_ALEQ(&sk->lan.addr, &sk->ext_lan.addr, sk->af_lan)) {
2076 bcopy(&sk->lan.addr, &fh.ap1.addr, sizeof(fh.ap1.addr));
2077 bcopy(&sk->ext_lan.addr, &fh.ap2.addr, sizeof(fh.ap2.addr));
2078 } else {
2079 bcopy(&sk->ext_lan.addr, &fh.ap1.addr, sizeof(fh.ap1.addr));
2080 bcopy(&sk->lan.addr, &fh.ap2.addr, sizeof(fh.ap2.addr));
2081 }
2082 if (sk->lan.xport.spi <= sk->ext_lan.xport.spi) {
2083 fh.ap1.xport.spi = sk->lan.xport.spi;
2084 fh.ap2.xport.spi = sk->ext_lan.xport.spi;
2085 } else {
2086 fh.ap1.xport.spi = sk->ext_lan.xport.spi;
2087 fh.ap2.xport.spi = sk->lan.xport.spi;
2088 }
2089 fh.af = sk->af_lan;
2090 fh.proto = sk->proto;
2091
2092 try_again:
2093 flowhash = net_flowhash(&fh, sizeof(fh), pf_hash_seed);
2094 if (flowhash == 0) {
2095 /* try to get a non-zero flowhash */
2096 pf_hash_seed = RandomULong();
2097 goto try_again;
2098 }
2099
2100 return flowhash;
2101 }
2102
2103 static int
2104 pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2)
2105 {
2106 if (aw1->type != aw2->type) {
2107 return 1;
2108 }
2109 switch (aw1->type) {
2110 case PF_ADDR_ADDRMASK:
2111 case PF_ADDR_RANGE:
2112 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0)) {
2113 return 1;
2114 }
2115 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0)) {
2116 return 1;
2117 }
2118 return 0;
2119 case PF_ADDR_DYNIFTL:
2120 return aw1->p.dyn == NULL || aw2->p.dyn == NULL ||
2121 aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt;
2122 case PF_ADDR_NOROUTE:
2123 case PF_ADDR_URPFFAILED:
2124 return 0;
2125 case PF_ADDR_TABLE:
2126 return aw1->p.tbl != aw2->p.tbl;
2127 case PF_ADDR_RTLABEL:
2128 return aw1->v.rtlabel != aw2->v.rtlabel;
2129 default:
2130 printf("invalid address type: %d\n", aw1->type);
2131 return 1;
2132 }
2133 }
2134
2135 u_int16_t
2136 pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp)
2137 {
2138 return nat464_cksum_fixup(cksum, old, new, udp);
2139 }
2140
2141 /*
2142 * change ip address & port
2143 * dir : packet direction
2144 * a : address to be changed
2145 * p : port to be changed
2146 * ic : ip header checksum
2147 * pc : protocol checksum
2148 * an : new ip address
2149 * pn : new port
2150 * u : should be 1 if UDP packet else 0
2151 * af : address family of the packet
2152 * afn : address family of the new address
2153 * ua : should be 1 if ip address needs to be updated in the packet else
2154 * only the checksum is recalculated & updated.
2155 */
2156 static void
2157 pf_change_ap(int dir, pbuf_t *pbuf, struct pf_addr *a, u_int16_t *p,
2158 u_int16_t *ic, u_int16_t *pc, struct pf_addr *an, u_int16_t pn,
2159 u_int8_t u, sa_family_t af, sa_family_t afn, int ua)
2160 {
2161 struct pf_addr ao;
2162 u_int16_t po = *p;
2163
2164 PF_ACPY(&ao, a, af);
2165 if (ua) {
2166 PF_ACPY(a, an, afn);
2167 }
2168
2169 *p = pn;
2170
2171 switch (af) {
2172 #if INET
2173 case AF_INET:
2174 switch (afn) {
2175 case AF_INET:
2176 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
2177 ao.addr16[0], an->addr16[0], 0),
2178 ao.addr16[1], an->addr16[1], 0);
2179 *p = pn;
2180 /*
2181 * If the packet is originated from an ALG on the NAT gateway
2182 * (source address is loopback or local), in which case the
2183 * TCP/UDP checksum field contains the pseudo header checksum
2184 * that's not yet complemented.
2185 * In that case we do not need to fixup the checksum for port
2186 * translation as the pseudo header checksum doesn't include ports.
2187 *
2188 * A packet generated locally will have UDP/TCP CSUM flag
2189 * set (gets set in protocol output).
2190 *
2191 * It should be noted that the fixup doesn't do anything if the
2192 * checksum is 0.
2193 */
2194 if (dir == PF_OUT && pbuf != NULL &&
2195 (*pbuf->pb_csum_flags & (CSUM_TCP | CSUM_UDP))) {
2196 /* Pseudo-header checksum does not include ports */
2197 *pc = ~pf_cksum_fixup(pf_cksum_fixup(~*pc,
2198 ao.addr16[0], an->addr16[0], u),
2199 ao.addr16[1], an->addr16[1], u);
2200 } else {
2201 *pc =
2202 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2203 *pc, ao.addr16[0], an->addr16[0], u),
2204 ao.addr16[1], an->addr16[1], u),
2205 po, pn, u);
2206 }
2207 break;
2208 case AF_INET6:
2209 *p = pn;
2210 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2211 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2212
2213 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
2214 ao.addr16[0], an->addr16[0], u),
2215 ao.addr16[1], an->addr16[1], u),
2216 0, an->addr16[2], u),
2217 0, an->addr16[3], u),
2218 0, an->addr16[4], u),
2219 0, an->addr16[5], u),
2220 0, an->addr16[6], u),
2221 0, an->addr16[7], u),
2222 po, pn, u);
2223 break;
2224 }
2225 break;
2226 #endif /* INET */
2227 case AF_INET6:
2228 switch (afn) {
2229 case AF_INET6:
2230 /*
2231 * If the packet is originated from an ALG on the NAT gateway
2232 * (source address is loopback or local), in which case the
2233 * TCP/UDP checksum field contains the pseudo header checksum
2234 * that's not yet complemented.
2235 * A packet generated locally
2236 * will have UDP/TCP CSUM flag set (gets set in protocol
2237 * output).
2238 */
2239 if (dir == PF_OUT && pbuf != NULL &&
2240 (*pbuf->pb_csum_flags & (CSUM_TCPIPV6 |
2241 CSUM_UDPIPV6))) {
2242 /* Pseudo-header checksum does not include ports */
2243 *pc =
2244 ~pf_cksum_fixup(pf_cksum_fixup(
2245 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2246 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2247 ~*pc,
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 } else {
2257 *pc =
2258 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2259 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2260 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2261 *pc,
2262 ao.addr16[0], an->addr16[0], u),
2263 ao.addr16[1], an->addr16[1], u),
2264 ao.addr16[2], an->addr16[2], u),
2265 ao.addr16[3], an->addr16[3], u),
2266 ao.addr16[4], an->addr16[4], u),
2267 ao.addr16[5], an->addr16[5], u),
2268 ao.addr16[6], an->addr16[6], u),
2269 ao.addr16[7], an->addr16[7], u),
2270 po, pn, u);
2271 }
2272 break;
2273 #ifdef INET
2274 case AF_INET:
2275 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2276 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2277 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
2278 ao.addr16[0], an->addr16[0], u),
2279 ao.addr16[1], an->addr16[1], u),
2280 ao.addr16[2], 0, u),
2281 ao.addr16[3], 0, u),
2282 ao.addr16[4], 0, u),
2283 ao.addr16[5], 0, u),
2284 ao.addr16[6], 0, u),
2285 ao.addr16[7], 0, u),
2286 po, pn, u);
2287 break;
2288 #endif /* INET */
2289 }
2290 break;
2291 }
2292 }
2293
2294
2295 /* Changes a u_int32_t. Uses a void * so there are no align restrictions */
2296 void
2297 pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u)
2298 {
2299 u_int32_t ao;
2300
2301 memcpy(&ao, a, sizeof(ao));
2302 memcpy(a, &an, sizeof(u_int32_t));
2303 *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u),
2304 ao % 65536, an % 65536, u);
2305 }
2306
2307 static void
2308 pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u)
2309 {
2310 struct pf_addr ao;
2311
2312 PF_ACPY(&ao, a, AF_INET6);
2313 PF_ACPY(a, an, AF_INET6);
2314
2315 *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2316 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2317 pf_cksum_fixup(pf_cksum_fixup(*c,
2318 ao.addr16[0], an->addr16[0], u),
2319 ao.addr16[1], an->addr16[1], u),
2320 ao.addr16[2], an->addr16[2], u),
2321 ao.addr16[3], an->addr16[3], u),
2322 ao.addr16[4], an->addr16[4], u),
2323 ao.addr16[5], an->addr16[5], u),
2324 ao.addr16[6], an->addr16[6], u),
2325 ao.addr16[7], an->addr16[7], u);
2326 }
2327
2328 void
2329 pf_change_addr(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u,
2330 sa_family_t af, sa_family_t afn)
2331 {
2332 struct pf_addr ao;
2333
2334 if (af != afn) {
2335 PF_ACPY(&ao, a, af);
2336 PF_ACPY(a, an, afn);
2337 }
2338
2339 switch (af) {
2340 case AF_INET:
2341 switch (afn) {
2342 case AF_INET:
2343 pf_change_a(a, c, an->v4addr.s_addr, u);
2344 break;
2345 case AF_INET6:
2346 *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2347 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2348 pf_cksum_fixup(pf_cksum_fixup(*c,
2349 ao.addr16[0], an->addr16[0], u),
2350 ao.addr16[1], an->addr16[1], u),
2351 0, an->addr16[2], u),
2352 0, an->addr16[3], u),
2353 0, an->addr16[4], u),
2354 0, an->addr16[5], u),
2355 0, an->addr16[6], u),
2356 0, an->addr16[7], u);
2357 break;
2358 }
2359 break;
2360 case AF_INET6:
2361 switch (afn) {
2362 case AF_INET:
2363 *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2364 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2365 pf_cksum_fixup(pf_cksum_fixup(*c,
2366 ao.addr16[0], an->addr16[0], u),
2367 ao.addr16[1], an->addr16[1], u),
2368 ao.addr16[2], 0, u),
2369 ao.addr16[3], 0, u),
2370 ao.addr16[4], 0, u),
2371 ao.addr16[5], 0, u),
2372 ao.addr16[6], 0, u),
2373 ao.addr16[7], 0, u);
2374 break;
2375 case AF_INET6:
2376 pf_change_a6(a, c, an, u);
2377 break;
2378 }
2379 break;
2380 }
2381 }
2382
2383 static void
2384 pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa,
2385 struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c,
2386 u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af)
2387 {
2388 struct pf_addr oia, ooa;
2389
2390 PF_ACPY(&oia, ia, af);
2391 PF_ACPY(&ooa, oa, af);
2392
2393 /* Change inner protocol port, fix inner protocol checksum. */
2394 if (ip != NULL) {
2395 u_int16_t oip = *ip;
2396 u_int32_t opc = 0;
2397
2398 if (pc != NULL) {
2399 opc = *pc;
2400 }
2401 *ip = np;
2402 if (pc != NULL) {
2403 *pc = pf_cksum_fixup(*pc, oip, *ip, u);
2404 }
2405 *ic = pf_cksum_fixup(*ic, oip, *ip, 0);
2406 if (pc != NULL) {
2407 *ic = pf_cksum_fixup(*ic, opc, *pc, 0);
2408 }
2409 }
2410 /* Change inner ip address, fix inner ip and icmp checksums. */
2411 PF_ACPY(ia, na, af);
2412 switch (af) {
2413 #if INET
2414 case AF_INET: {
2415 u_int32_t oh2c = *h2c;
2416
2417 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c,
2418 oia.addr16[0], ia->addr16[0], 0),
2419 oia.addr16[1], ia->addr16[1], 0);
2420 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
2421 oia.addr16[0], ia->addr16[0], 0),
2422 oia.addr16[1], ia->addr16[1], 0);
2423 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0);
2424 break;
2425 }
2426 #endif /* INET */
2427 case AF_INET6:
2428 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2429 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2430 pf_cksum_fixup(pf_cksum_fixup(*ic,
2431 oia.addr16[0], ia->addr16[0], u),
2432 oia.addr16[1], ia->addr16[1], u),
2433 oia.addr16[2], ia->addr16[2], u),
2434 oia.addr16[3], ia->addr16[3], u),
2435 oia.addr16[4], ia->addr16[4], u),
2436 oia.addr16[5], ia->addr16[5], u),
2437 oia.addr16[6], ia->addr16[6], u),
2438 oia.addr16[7], ia->addr16[7], u);
2439 break;
2440 }
2441 /* Change outer ip address, fix outer ip or icmpv6 checksum. */
2442 PF_ACPY(oa, na, af);
2443 switch (af) {
2444 #if INET
2445 case AF_INET:
2446 *hc = pf_cksum_fixup(pf_cksum_fixup(*hc,
2447 ooa.addr16[0], oa->addr16[0], 0),
2448 ooa.addr16[1], oa->addr16[1], 0);
2449 break;
2450 #endif /* INET */
2451 case AF_INET6:
2452 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2453 pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
2454 pf_cksum_fixup(pf_cksum_fixup(*ic,
2455 ooa.addr16[0], oa->addr16[0], u),
2456 ooa.addr16[1], oa->addr16[1], u),
2457 ooa.addr16[2], oa->addr16[2], u),
2458 ooa.addr16[3], oa->addr16[3], u),
2459 ooa.addr16[4], oa->addr16[4], u),
2460 ooa.addr16[5], oa->addr16[5], u),
2461 ooa.addr16[6], oa->addr16[6], u),
2462 ooa.addr16[7], oa->addr16[7], u);
2463 break;
2464 }
2465 }
2466
2467
2468 /*
2469 * Need to modulate the sequence numbers in the TCP SACK option
2470 * (credits to Krzysztof Pfaff for report and patch)
2471 */
2472 static int
2473 pf_modulate_sack(pbuf_t *pbuf, int off, struct pf_pdesc *pd,
2474 struct tcphdr *th, struct pf_state_peer *dst)
2475 {
2476 int hlen = (th->th_off << 2) - sizeof(*th), thoptlen = hlen;
2477 u_int8_t opts[MAX_TCPOPTLEN], *opt = opts;
2478 int copyback = 0, i, olen;
2479 struct sackblk sack;
2480
2481 #define TCPOLEN_SACKLEN (TCPOLEN_SACK + 2)
2482 if (hlen < TCPOLEN_SACKLEN ||
2483 !pf_pull_hdr(pbuf, off + sizeof(*th), opts, hlen, NULL, NULL, pd->af)) {
2484 return 0;
2485 }
2486
2487 while (hlen >= TCPOLEN_SACKLEN) {
2488 olen = opt[1];
2489 switch (*opt) {
2490 case TCPOPT_EOL: /* FALLTHROUGH */
2491 case TCPOPT_NOP:
2492 opt++;
2493 hlen--;
2494 break;
2495 case TCPOPT_SACK:
2496 if (olen > hlen) {
2497 olen = hlen;
2498 }
2499 if (olen >= TCPOLEN_SACKLEN) {
2500 for (i = 2; i + TCPOLEN_SACK <= olen;
2501 i += TCPOLEN_SACK) {
2502 memcpy(&sack, &opt[i], sizeof(sack));
2503 pf_change_a(&sack.start, &th->th_sum,
2504 htonl(ntohl(sack.start) -
2505 dst->seqdiff), 0);
2506 pf_change_a(&sack.end, &th->th_sum,
2507 htonl(ntohl(sack.end) -
2508 dst->seqdiff), 0);
2509 memcpy(&opt[i], &sack, sizeof(sack));
2510 }
2511 copyback = off + sizeof(*th) + thoptlen;
2512 }
2513 OS_FALLTHROUGH;
2514 default:
2515 if (olen < 2) {
2516 olen = 2;
2517 }
2518 hlen -= olen;
2519 opt += olen;
2520 }
2521 }
2522
2523 if (copyback) {
2524 if (pf_lazy_makewritable(pd, pbuf, copyback) == NULL) {
2525 return -1;
2526 }
2527 pbuf_copy_back(pbuf, off + sizeof(*th), thoptlen, opts);
2528 }
2529 return copyback;
2530 }
2531
2532 /*
2533 * XXX
2534 *
2535 * The following functions (pf_send_tcp and pf_send_icmp) are somewhat
2536 * special in that they originate "spurious" packets rather than
2537 * filter/NAT existing packets. As such, they're not a great fit for
2538 * the 'pbuf' shim, which assumes the underlying packet buffers are
2539 * allocated elsewhere.
2540 *
2541 * Since these functions are rarely used, we'll carry on allocating mbufs
2542 * and passing them to the IP stack for eventual routing.
2543 */
2544 static void
2545 pf_send_tcp(const struct pf_rule *r, sa_family_t af,
2546 const struct pf_addr *saddr, const struct pf_addr *daddr,
2547 u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack,
2548 u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl, int tag,
2549 u_int16_t rtag, struct ether_header *eh, struct ifnet *ifp)
2550 {
2551 #pragma unused(eh, ifp)
2552 struct mbuf *m;
2553 int len, tlen;
2554 #if INET
2555 struct ip *h = NULL;
2556 #endif /* INET */
2557 struct ip6_hdr *h6 = NULL;
2558 struct tcphdr *th = NULL;
2559 char *opt;
2560 struct pf_mtag *pf_mtag;
2561
2562 /* maximum segment size tcp option */
2563 tlen = sizeof(struct tcphdr);
2564 if (mss) {
2565 tlen += 4;
2566 }
2567
2568 switch (af) {
2569 #if INET
2570 case AF_INET:
2571 len = sizeof(struct ip) + tlen;
2572 break;
2573 #endif /* INET */
2574 case AF_INET6:
2575 len = sizeof(struct ip6_hdr) + tlen;
2576 break;
2577 default:
2578 panic("pf_send_tcp: not AF_INET or AF_INET6!");
2579 return;
2580 }
2581
2582 /* create outgoing mbuf */
2583 m = m_gethdr(M_DONTWAIT, MT_HEADER);
2584 if (m == NULL) {
2585 return;
2586 }
2587
2588 if ((pf_mtag = pf_get_mtag(m)) == NULL) {
2589 return;
2590 }
2591
2592 if (tag) {
2593 pf_mtag->pftag_flags |= PF_TAG_GENERATED;
2594 }
2595 pf_mtag->pftag_tag = rtag;
2596
2597 if (r != NULL && PF_RTABLEID_IS_VALID(r->rtableid)) {
2598 pf_mtag->pftag_rtableid = r->rtableid;
2599 }
2600
2601 #if PF_ECN
2602 /* add hints for ecn */
2603 pf_mtag->pftag_hdr = mtod(m, struct ip *);
2604 /* record address family */
2605 pf_mtag->pftag_flags &= ~(PF_TAG_HDR_INET | PF_TAG_HDR_INET6);
2606 switch (af) {
2607 #if INET
2608 case AF_INET:
2609 pf_mtag->pftag_flags |= PF_TAG_HDR_INET;
2610 break;
2611 #endif /* INET */
2612 case AF_INET6:
2613 pf_mtag->pftag_flags |= PF_TAG_HDR_INET6;
2614 break;
2615 }
2616 #endif /* PF_ECN */
2617
2618 /* indicate this is TCP */
2619 m->m_pkthdr.pkt_proto = IPPROTO_TCP;
2620
2621 /* Make sure headers are 32-bit aligned */
2622 m->m_data += max_linkhdr;
2623 m->m_pkthdr.len = m->m_len = len;
2624 m->m_pkthdr.rcvif = NULL;
2625 bzero(m->m_data, len);
2626 switch (af) {
2627 #if INET
2628 case AF_INET:
2629 h = mtod(m, struct ip *);
2630
2631 /* IP header fields included in the TCP checksum */
2632 h->ip_p = IPPROTO_TCP;
2633 h->ip_len = htons(tlen);
2634 h->ip_src.s_addr = saddr->v4addr.s_addr;
2635 h->ip_dst.s_addr = daddr->v4addr.s_addr;
2636
2637 th = (struct tcphdr *)(void *)((caddr_t)h + sizeof(struct ip));
2638 break;
2639 #endif /* INET */
2640 case AF_INET6:
2641 h6 = mtod(m, struct ip6_hdr *);
2642
2643 /* IP header fields included in the TCP checksum */
2644 h6->ip6_nxt = IPPROTO_TCP;
2645 h6->ip6_plen = htons(tlen);
2646 memcpy(&h6->ip6_src, &saddr->v6addr, sizeof(struct in6_addr));
2647 memcpy(&h6->ip6_dst, &daddr->v6addr, sizeof(struct in6_addr));
2648
2649 th = (struct tcphdr *)(void *)
2650 ((caddr_t)h6 + sizeof(struct ip6_hdr));
2651 break;
2652 }
2653
2654 /* TCP header */
2655 th->th_sport = sport;
2656 th->th_dport = dport;
2657 th->th_seq = htonl(seq);
2658 th->th_ack = htonl(ack);
2659 th->th_off = tlen >> 2;
2660 th->th_flags = flags;
2661 th->th_win = htons(win);
2662
2663 if (mss) {
2664 opt = (char *)(th + 1);
2665 opt[0] = TCPOPT_MAXSEG;
2666 opt[1] = 4;
2667 #if BYTE_ORDER != BIG_ENDIAN
2668 HTONS(mss);
2669 #endif
2670 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2);
2671 }
2672
2673 switch (af) {
2674 #if INET
2675 case AF_INET: {
2676 struct route ro;
2677
2678 /* TCP checksum */
2679 th->th_sum = in_cksum(m, len);
2680
2681 /* Finish the IP header */
2682 h->ip_v = 4;
2683 h->ip_hl = sizeof(*h) >> 2;
2684 h->ip_tos = IPTOS_LOWDELAY;
2685 /*
2686 * ip_output() expects ip_len and ip_off to be in host order.
2687 */
2688 h->ip_len = len;
2689 h->ip_off = (path_mtu_discovery ? IP_DF : 0);
2690 h->ip_ttl = ttl ? ttl : ip_defttl;
2691 h->ip_sum = 0;
2692
2693 bzero(&ro, sizeof(ro));
2694 ip_output(m, NULL, &ro, 0, NULL, NULL);
2695 ROUTE_RELEASE(&ro);
2696 break;
2697 }
2698 #endif /* INET */
2699 case AF_INET6: {
2700 struct route_in6 ro6;
2701
2702 /* TCP checksum */
2703 th->th_sum = in6_cksum(m, IPPROTO_TCP,
2704 sizeof(struct ip6_hdr), tlen);
2705
2706 h6->ip6_vfc |= IPV6_VERSION;
2707 h6->ip6_hlim = IPV6_DEFHLIM;
2708
2709 bzero(&ro6, sizeof(ro6));
2710 ip6_output(m, NULL, &ro6, 0, NULL, NULL, NULL);
2711 ROUTE_RELEASE(&ro6);
2712 break;
2713 }
2714 }
2715 }
2716
2717 static void
2718 pf_send_icmp(pbuf_t *pbuf, u_int8_t type, u_int8_t code, sa_family_t af,
2719 struct pf_rule *r)
2720 {
2721 struct mbuf *m0;
2722 struct pf_mtag *pf_mtag;
2723
2724 m0 = pbuf_clone_to_mbuf(pbuf);
2725 if (m0 == NULL) {
2726 return;
2727 }
2728
2729 if ((pf_mtag = pf_get_mtag(m0)) == NULL) {
2730 return;
2731 }
2732
2733 pf_mtag->pftag_flags |= PF_TAG_GENERATED;
2734
2735 if (PF_RTABLEID_IS_VALID(r->rtableid)) {
2736 pf_mtag->pftag_rtableid = r->rtableid;
2737 }
2738
2739 #if PF_ECN
2740 /* add hints for ecn */
2741 pf_mtag->pftag_hdr = mtod(m0, struct ip *);
2742 /* record address family */
2743 pf_mtag->pftag_flags &= ~(PF_TAG_HDR_INET | PF_TAG_HDR_INET6);
2744 switch (af) {
2745 #if INET
2746 case AF_INET:
2747 pf_mtag->pftag_flags |= PF_TAG_HDR_INET;
2748 m0->m_pkthdr.pkt_proto = IPPROTO_ICMP;
2749 break;
2750 #endif /* INET */
2751 case AF_INET6:
2752 pf_mtag->pftag_flags |= PF_TAG_HDR_INET6;
2753 m0->m_pkthdr.pkt_proto = IPPROTO_ICMPV6;
2754 break;
2755 }
2756 #endif /* PF_ECN */
2757
2758 switch (af) {
2759 #if INET
2760 case AF_INET:
2761 icmp_error(m0, type, code, 0, 0);
2762 break;
2763 #endif /* INET */
2764 case AF_INET6:
2765 icmp6_error(m0, type, code, 0);
2766 break;
2767 }
2768 }
2769
2770 /*
2771 * Return 1 if the addresses a and b match (with mask m), otherwise return 0.
2772 * If n is 0, they match if they are equal. If n is != 0, they match if they
2773 * are different.
2774 */
2775 int
2776 pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m,
2777 struct pf_addr *b, sa_family_t af)
2778 {
2779 int match = 0;
2780
2781 switch (af) {
2782 #if INET
2783 case AF_INET:
2784 if ((a->addr32[0] & m->addr32[0]) ==
2785 (b->addr32[0] & m->addr32[0])) {
2786 match++;
2787 }
2788 break;
2789 #endif /* INET */
2790 case AF_INET6:
2791 if (((a->addr32[0] & m->addr32[0]) ==
2792 (b->addr32[0] & m->addr32[0])) &&
2793 ((a->addr32[1] & m->addr32[1]) ==
2794 (b->addr32[1] & m->addr32[1])) &&
2795 ((a->addr32[2] & m->addr32[2]) ==
2796 (b->addr32[2] & m->addr32[2])) &&
2797 ((a->addr32[3] & m->addr32[3]) ==
2798 (b->addr32[3] & m->addr32[3]))) {
2799 match++;
2800 }
2801 break;
2802 }
2803 if (match) {
2804 if (n) {
2805 return 0;
2806 } else {
2807 return 1;
2808 }
2809 } else {
2810 if (n) {
2811 return 1;
2812 } else {
2813 return 0;
2814 }
2815 }
2816 }
2817
2818 /*
2819 * Return 1 if b <= a <= e, otherwise return 0.
2820 */
2821 int
2822 pf_match_addr_range(struct pf_addr *b, struct pf_addr *e,
2823 struct pf_addr *a, sa_family_t af)
2824 {
2825 switch (af) {
2826 #if INET
2827 case AF_INET:
2828 if ((a->addr32[0] < b->addr32[0]) ||
2829 (a->addr32[0] > e->addr32[0])) {
2830 return 0;
2831 }
2832 break;
2833 #endif /* INET */
2834 case AF_INET6: {
2835 int i;
2836
2837 /* check a >= b */
2838 for (i = 0; i < 4; ++i) {
2839 if (a->addr32[i] > b->addr32[i]) {
2840 break;
2841 } else if (a->addr32[i] < b->addr32[i]) {
2842 return 0;
2843 }
2844 }
2845 /* check a <= e */
2846 for (i = 0; i < 4; ++i) {
2847 if (a->addr32[i] < e->addr32[i]) {
2848 break;
2849 } else if (a->addr32[i] > e->addr32[i]) {
2850 return 0;
2851 }
2852 }
2853 break;
2854 }
2855 }
2856 return 1;
2857 }
2858
2859 int
2860 pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p)
2861 {
2862 switch (op) {
2863 case PF_OP_IRG:
2864 return (p > a1) && (p < a2);
2865 case PF_OP_XRG:
2866 return (p < a1) || (p > a2);
2867 case PF_OP_RRG:
2868 return (p >= a1) && (p <= a2);
2869 case PF_OP_EQ:
2870 return p == a1;
2871 case PF_OP_NE:
2872 return p != a1;
2873 case PF_OP_LT:
2874 return p < a1;
2875 case PF_OP_LE:
2876 return p <= a1;
2877 case PF_OP_GT:
2878 return p > a1;
2879 case PF_OP_GE:
2880 return p >= a1;
2881 }
2882 return 0; /* never reached */
2883 }
2884
2885 int
2886 pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p)
2887 {
2888 #if BYTE_ORDER != BIG_ENDIAN
2889 NTOHS(a1);
2890 NTOHS(a2);
2891 NTOHS(p);
2892 #endif
2893 return pf_match(op, a1, a2, p);
2894 }
2895
2896 int
2897 pf_match_xport(u_int8_t proto, u_int8_t proto_variant, union pf_rule_xport *rx,
2898 union pf_state_xport *sx)
2899 {
2900 int d = !0;
2901
2902 if (sx) {
2903 switch (proto) {
2904 case IPPROTO_GRE:
2905 if (proto_variant == PF_GRE_PPTP_VARIANT) {
2906 d = (rx->call_id == sx->call_id);
2907 }
2908 break;
2909
2910 case IPPROTO_ESP:
2911 d = (rx->spi == sx->spi);
2912 break;
2913
2914 case IPPROTO_TCP:
2915 case IPPROTO_UDP:
2916 case IPPROTO_ICMP:
2917 case IPPROTO_ICMPV6:
2918 if (rx->range.op) {
2919 d = pf_match_port(rx->range.op,
2920 rx->range.port[0], rx->range.port[1],
2921 sx->port);
2922 }
2923 break;
2924
2925 default:
2926 break;
2927 }
2928 }
2929
2930 return d;
2931 }
2932
2933 int
2934 pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u)
2935 {
2936 if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE) {
2937 return 0;
2938 }
2939 return pf_match(op, a1, a2, u);
2940 }
2941
2942 int
2943 pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g)
2944 {
2945 if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE) {
2946 return 0;
2947 }
2948 return pf_match(op, a1, a2, g);
2949 }
2950
2951 static int
2952 pf_match_tag(struct pf_rule *r, struct pf_mtag *pf_mtag,
2953 int *tag)
2954 {
2955 if (*tag == -1) {
2956 *tag = pf_mtag->pftag_tag;
2957 }
2958
2959 return (!r->match_tag_not && r->match_tag == *tag) ||
2960 (r->match_tag_not && r->match_tag != *tag);
2961 }
2962
2963 int
2964 pf_tag_packet(pbuf_t *pbuf, struct pf_mtag *pf_mtag, int tag,
2965 unsigned int rtableid, struct pf_pdesc *pd)
2966 {
2967 if (tag <= 0 && !PF_RTABLEID_IS_VALID(rtableid) &&
2968 (pd == NULL || !(pd->pktflags & PKTF_FLOW_ID))) {
2969 return 0;
2970 }
2971
2972 if (pf_mtag == NULL && (pf_mtag = pf_get_mtag_pbuf(pbuf)) == NULL) {
2973 return 1;
2974 }
2975
2976 if (tag > 0) {
2977 pf_mtag->pftag_tag = tag;
2978 }
2979 if (PF_RTABLEID_IS_VALID(rtableid)) {
2980 pf_mtag->pftag_rtableid = rtableid;
2981 }
2982 if (pd != NULL && (pd->pktflags & PKTF_FLOW_ID)) {
2983 *pbuf->pb_flowsrc = pd->flowsrc;
2984 *pbuf->pb_flowid = pd->flowhash;
2985 *pbuf->pb_flags |= pd->pktflags;
2986 *pbuf->pb_proto = pd->proto;
2987 }
2988
2989 return 0;
2990 }
2991
2992 void
2993 pf_step_into_anchor(int *depth, struct pf_ruleset **rs, int n,
2994 struct pf_rule **r, struct pf_rule **a, int *match)
2995 {
2996 struct pf_anchor_stackframe *f;
2997
2998 (*r)->anchor->match = 0;
2999 if (match) {
3000 *match = 0;
3001 }
3002 if (*depth >= (int)sizeof(pf_anchor_stack) /
3003 (int)sizeof(pf_anchor_stack[0])) {
3004 printf("pf_step_into_anchor: stack overflow\n");
3005 *r = TAILQ_NEXT(*r, entries);
3006 return;
3007 } else if (*depth == 0 && a != NULL) {
3008 *a = *r;
3009 }
3010 f = pf_anchor_stack + (*depth)++;
3011 f->rs = *rs;
3012 f->r = *r;
3013 if ((*r)->anchor_wildcard) {
3014 f->parent = &(*r)->anchor->children;
3015 if ((f->child = RB_MIN(pf_anchor_node, f->parent)) ==
3016 NULL) {
3017 *r = NULL;
3018 return;
3019 }
3020 *rs = &f->child->ruleset;
3021 } else {
3022 f->parent = NULL;
3023 f->child = NULL;
3024 *rs = &(*r)->anchor->ruleset;
3025 }
3026 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
3027 }
3028
3029 int
3030 pf_step_out_of_anchor(int *depth, struct pf_ruleset **rs, int n,
3031 struct pf_rule **r, struct pf_rule **a, int *match)
3032 {
3033 struct pf_anchor_stackframe *f;
3034 int quick = 0;
3035
3036 do {
3037 if (*depth <= 0) {
3038 break;
3039 }
3040 f = pf_anchor_stack + *depth - 1;
3041 if (f->parent != NULL && f->child != NULL) {
3042 if (f->child->match ||
3043 (match != NULL && *match)) {
3044 f->r->anchor->match = 1;
3045 if (match) {
3046 *match = 0;
3047 }
3048 }
3049 f->child = RB_NEXT(pf_anchor_node, f->parent, f->child);
3050 if (f->child != NULL) {
3051 *rs = &f->child->ruleset;
3052 *r = TAILQ_FIRST((*rs)->rules[n].active.ptr);
3053 if (*r == NULL) {
3054 continue;
3055 } else {
3056 break;
3057 }
3058 }
3059 }
3060 (*depth)--;
3061 if (*depth == 0 && a != NULL) {
3062 *a = NULL;
3063 }
3064 *rs = f->rs;
3065 if (f->r->anchor->match || (match != NULL && *match)) {
3066 quick = f->r->quick;
3067 }
3068 *r = TAILQ_NEXT(f->r, entries);
3069 } while (*r == NULL);
3070
3071 return quick;
3072 }
3073
3074 void
3075 pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr,
3076 struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af)
3077 {
3078 switch (af) {
3079 #if INET
3080 case AF_INET:
3081 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
3082 ((rmask->addr32[0] ^ 0xffffffff) & saddr->addr32[0]);
3083 break;
3084 #endif /* INET */
3085 case AF_INET6:
3086 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
3087 ((rmask->addr32[0] ^ 0xffffffff) & saddr->addr32[0]);
3088 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) |
3089 ((rmask->addr32[1] ^ 0xffffffff) & saddr->addr32[1]);
3090 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) |
3091 ((rmask->addr32[2] ^ 0xffffffff) & saddr->addr32[2]);
3092 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) |
3093 ((rmask->addr32[3] ^ 0xffffffff) & saddr->addr32[3]);
3094 break;
3095 }
3096 }
3097
3098 void
3099 pf_addr_inc(struct pf_addr *addr, sa_family_t af)
3100 {
3101 switch (af) {
3102 #if INET
3103 case AF_INET:
3104 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1);
3105 break;
3106 #endif /* INET */
3107 case AF_INET6:
3108 if (addr->addr32[3] == 0xffffffff) {
3109 addr->addr32[3] = 0;
3110 if (addr->addr32[2] == 0xffffffff) {
3111 addr->addr32[2] = 0;
3112 if (addr->addr32[1] == 0xffffffff) {
3113 addr->addr32[1] = 0;
3114 addr->addr32[0] =
3115 htonl(ntohl(addr->addr32[0]) + 1);
3116 } else {
3117 addr->addr32[1] =
3118 htonl(ntohl(addr->addr32[1]) + 1);
3119 }
3120 } else {
3121 addr->addr32[2] =
3122 htonl(ntohl(addr->addr32[2]) + 1);
3123 }
3124 } else {
3125 addr->addr32[3] =
3126 htonl(ntohl(addr->addr32[3]) + 1);
3127 }
3128 break;
3129 }
3130 }
3131
3132 #define mix(a, b, c) \
3133 do { \
3134 a -= b; a -= c; a ^= (c >> 13); \
3135 b -= c; b -= a; b ^= (a << 8); \
3136 c -= a; c -= b; c ^= (b >> 13); \
3137 a -= b; a -= c; a ^= (c >> 12); \
3138 b -= c; b -= a; b ^= (a << 16); \
3139 c -= a; c -= b; c ^= (b >> 5); \
3140 a -= b; a -= c; a ^= (c >> 3); \
3141 b -= c; b -= a; b ^= (a << 10); \
3142 c -= a; c -= b; c ^= (b >> 15); \
3143 } while (0)
3144
3145 /*
3146 * hash function based on bridge_hash in if_bridge.c
3147 */
3148 static void
3149 pf_hash(struct pf_addr *inaddr, struct pf_addr *hash,
3150 struct pf_poolhashkey *key, sa_family_t af)
3151 {
3152 u_int32_t a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0];
3153
3154 switch (af) {
3155 #if INET
3156 case AF_INET:
3157 a += inaddr->addr32[0];
3158 b += key->key32[1];
3159 mix(a, b, c);
3160 hash->addr32[0] = c + key->key32[2];
3161 break;
3162 #endif /* INET */
3163 case AF_INET6:
3164 a += inaddr->addr32[0];
3165 b += inaddr->addr32[2];
3166 mix(a, b, c);
3167 hash->addr32[0] = c;
3168 a += inaddr->addr32[1];
3169 b += inaddr->addr32[3];
3170 c += key->key32[1];
3171 mix(a, b, c);
3172 hash->addr32[1] = c;
3173 a += inaddr->addr32[2];
3174 b += inaddr->addr32[1];
3175 c += key->key32[2];
3176 mix(a, b, c);
3177 hash->addr32[2] = c;
3178 a += inaddr->addr32[3];
3179 b += inaddr->addr32[0];
3180 c += key->key32[3];
3181 mix(a, b, c);
3182 hash->addr32[3] = c;
3183 break;
3184 }
3185 }
3186
3187 static int
3188 pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr,
3189 struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn)
3190 {
3191 unsigned char hash[16];
3192 struct pf_pool *rpool = &r->rpool;
3193 struct pf_addr *raddr = &rpool->cur->addr.v.a.addr;
3194 struct pf_addr *rmask = &rpool->cur->addr.v.a.mask;
3195 struct pf_pooladdr *acur = rpool->cur;
3196 struct pf_src_node k;
3197
3198 if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR &&
3199 (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
3200 k.af = af;
3201 PF_ACPY(&k.addr, saddr, af);
3202 if (r->rule_flag & PFRULE_RULESRCTRACK ||
3203 r->rpool.opts & PF_POOL_STICKYADDR) {
3204 k.rule.ptr = r;
3205 } else {
3206 k.rule.ptr = NULL;
3207 }
3208 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
3209 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
3210 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, rpool->af)) {
3211 PF_ACPY(naddr, &(*sn)->raddr, rpool->af);
3212 if (pf_status.debug >= PF_DEBUG_MISC) {
3213 printf("pf_map_addr: src tracking maps ");
3214 pf_print_host(&k.addr, 0, af);
3215 printf(" to ");
3216 pf_print_host(naddr, 0, rpool->af);
3217 printf("\n");
3218 }
3219 return 0;
3220 }
3221 }
3222
3223 if (rpool->cur->addr.type == PF_ADDR_NOROUTE) {
3224 return 1;
3225 }
3226 if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
3227 if (rpool->cur->addr.p.dyn == NULL) {
3228 return 1;
3229 }
3230 switch (rpool->af) {
3231 #if INET
3232 case AF_INET:
3233 if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 &&
3234 (rpool->opts & PF_POOL_TYPEMASK) !=
3235 PF_POOL_ROUNDROBIN) {
3236 return 1;
3237 }
3238 raddr = &rpool->cur->addr.p.dyn->pfid_addr4;
3239 rmask = &rpool->cur->addr.p.dyn->pfid_mask4;
3240 break;
3241 #endif /* INET */
3242 case AF_INET6:
3243 if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 &&
3244 (rpool->opts & PF_POOL_TYPEMASK) !=
3245 PF_POOL_ROUNDROBIN) {
3246 return 1;
3247 }
3248 raddr = &rpool->cur->addr.p.dyn->pfid_addr6;
3249 rmask = &rpool->cur->addr.p.dyn->pfid_mask6;
3250 break;
3251 }
3252 } else if (rpool->cur->addr.type == PF_ADDR_TABLE) {
3253 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN) {
3254 return 1; /* unsupported */
3255 }
3256 } else {
3257 raddr = &rpool->cur->addr.v.a.addr;
3258 rmask = &rpool->cur->addr.v.a.mask;
3259 }
3260
3261 switch (rpool->opts & PF_POOL_TYPEMASK) {
3262 case PF_POOL_NONE:
3263 PF_ACPY(naddr, raddr, rpool->af);
3264 break;
3265 case PF_POOL_BITMASK:
3266 ASSERT(af == rpool->af);
3267 PF_POOLMASK(naddr, raddr, rmask, saddr, af);
3268 break;
3269 case PF_POOL_RANDOM:
3270 if (init_addr != NULL && PF_AZERO(init_addr, rpool->af)) {
3271 switch (af) {
3272 #if INET
3273 case AF_INET:
3274 rpool->counter.addr32[0] = htonl(random());
3275 break;
3276 #endif /* INET */
3277 case AF_INET6:
3278 if (rmask->addr32[3] != 0xffffffff) {
3279 rpool->counter.addr32[3] =
3280 RandomULong();
3281 } else {
3282 break;
3283 }
3284 if (rmask->addr32[2] != 0xffffffff) {
3285 rpool->counter.addr32[2] =
3286 RandomULong();
3287 } else {
3288 break;
3289 }
3290 if (rmask->addr32[1] != 0xffffffff) {
3291 rpool->counter.addr32[1] =
3292 RandomULong();
3293 } else {
3294 break;
3295 }
3296 if (rmask->addr32[0] != 0xffffffff) {
3297 rpool->counter.addr32[0] =
3298 RandomULong();
3299 }
3300 break;
3301 }
3302 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter,
3303 rpool->af);
3304 PF_ACPY(init_addr, naddr, rpool->af);
3305 } else {
3306 PF_AINC(&rpool->counter, rpool->af);
3307 PF_POOLMASK(naddr, raddr, rmask, &rpool->counter,
3308 rpool->af);
3309 }
3310 break;
3311 case PF_POOL_SRCHASH:
3312 ASSERT(af == rpool->af);
3313 PF_POOLMASK(naddr, raddr, rmask, saddr, af);
3314 pf_hash(saddr, (struct pf_addr *)(void *)&hash,
3315 &rpool->key, af);
3316 PF_POOLMASK(naddr, raddr, rmask,
3317 (struct pf_addr *)(void *)&hash, af);
3318 break;
3319 case PF_POOL_ROUNDROBIN:
3320 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
3321 if (!pfr_pool_get(rpool->cur->addr.p.tbl,
3322 &rpool->tblidx, &rpool->counter,
3323 &raddr, &rmask, rpool->af)) {
3324 goto get_addr;
3325 }
3326 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
3327 if (rpool->cur->addr.p.dyn != NULL &&
3328 !pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
3329 &rpool->tblidx, &rpool->counter,
3330 &raddr, &rmask, af)) {
3331 goto get_addr;
3332 }
3333 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter,
3334 rpool->af)) {
3335 goto get_addr;
3336 }
3337
3338 try_next:
3339 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL) {
3340 rpool->cur = TAILQ_FIRST(&rpool->list);
3341 }
3342 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
3343 rpool->tblidx = -1;
3344 if (pfr_pool_get(rpool->cur->addr.p.tbl,
3345 &rpool->tblidx, &rpool->counter,
3346 &raddr, &rmask, rpool->af)) {
3347 /* table contains no address of type
3348 * 'rpool->af' */
3349 if (rpool->cur != acur) {
3350 goto try_next;
3351 }
3352 return 1;
3353 }
3354 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
3355 rpool->tblidx = -1;
3356 if (rpool->cur->addr.p.dyn == NULL) {
3357 return 1;
3358 }
3359 if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
3360 &rpool->tblidx, &rpool->counter,
3361 &raddr, &rmask, rpool->af)) {
3362 /* table contains no address of type
3363 * 'rpool->af' */
3364 if (rpool->cur != acur) {
3365 goto try_next;
3366 }
3367 return 1;
3368 }
3369 } else {
3370 raddr = &rpool->cur->addr.v.a.addr;
3371 rmask = &rpool->cur->addr.v.a.mask;
3372 PF_ACPY(&rpool->counter, raddr, rpool->af);
3373 }
3374
3375 get_addr:
3376 PF_ACPY(naddr, &rpool->counter, rpool->af);
3377 if (init_addr != NULL && PF_AZERO(init_addr, rpool->af)) {
3378 PF_ACPY(init_addr, naddr, rpool->af);
3379 }
3380 PF_AINC(&rpool->counter, rpool->af);
3381 break;
3382 }
3383 if (*sn != NULL) {
3384 PF_ACPY(&(*sn)->raddr, naddr, rpool->af);
3385 }
3386
3387 if (pf_status.debug >= PF_DEBUG_MISC &&
3388 (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
3389 printf("pf_map_addr: selected address ");
3390 pf_print_host(naddr, 0, rpool->af);
3391 printf("\n");
3392 }
3393
3394 return 0;
3395 }
3396
3397 static int
3398 pf_get_sport(struct pf_pdesc *pd, struct pfi_kif *kif, struct pf_rule *r,
3399 struct pf_addr *saddr, union pf_state_xport *sxport, struct pf_addr *daddr,
3400 union pf_state_xport *dxport, struct pf_addr *naddr,
3401 union pf_state_xport *nxport, struct pf_src_node **sn
3402 )
3403 {
3404 #pragma unused(kif)
3405 struct pf_state_key_cmp key;
3406 struct pf_addr init_addr;
3407 unsigned int cut;
3408 sa_family_t af = pd->af;
3409 u_int8_t proto = pd->proto;
3410 unsigned int low = r->rpool.proxy_port[0];
3411 unsigned int high = r->rpool.proxy_port[1];
3412
3413 bzero(&init_addr, sizeof(init_addr));
3414 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn)) {
3415 return 1;
3416 }
3417
3418 if (proto == IPPROTO_ICMP) {
3419 low = 1;
3420 high = 65535;
3421 }
3422
3423 if (!nxport) {
3424 return 0; /* No output necessary. */
3425 }
3426 /*--- Special mapping rules for UDP ---*/
3427 if (proto == IPPROTO_UDP) {
3428 /*--- Never float IKE source port ---*/
3429 if (ntohs(sxport->port) == PF_IKE_PORT) {
3430 nxport->port = sxport->port;
3431 return 0;
3432 }
3433
3434 /*--- Apply exterior mapping options ---*/
3435 if (r->extmap > PF_EXTMAP_APD) {
3436 struct pf_state *s;
3437
3438 TAILQ_FOREACH(s, &state_list, entry_list) {
3439 struct pf_state_key *sk = s->state_key;
3440 if (!sk) {
3441 continue;
3442 }
3443 if (s->nat_rule.ptr != r) {
3444 continue;
3445 }
3446 if (sk->proto != IPPROTO_UDP ||
3447 sk->af_lan != af) {
3448 continue;
3449 }
3450 if (sk->lan.xport.port != sxport->port) {
3451 continue;
3452 }
3453 if (PF_ANEQ(&sk->lan.addr, saddr, af)) {
3454 continue;
3455 }
3456 if (r->extmap < PF_EXTMAP_EI &&
3457 PF_ANEQ(&sk->ext_lan.addr, daddr, af)) {
3458 continue;
3459 }
3460
3461 nxport->port = sk->gwy.xport.port;
3462 return 0;
3463 }
3464 }
3465 } else if (proto == IPPROTO_TCP) {
3466 struct pf_state* s;
3467 /*
3468 * APPLE MODIFICATION: <rdar://problem/6546358>
3469 * Fix allows....NAT to use a single binding for TCP session
3470 * with same source IP and source port
3471 */
3472 TAILQ_FOREACH(s, &state_list, entry_list) {
3473 struct pf_state_key* sk = s->state_key;
3474 if (!sk) {
3475 continue;
3476 }
3477 if (s->nat_rule.ptr != r) {
3478 continue;
3479 }
3480 if (sk->proto != IPPROTO_TCP || sk->af_lan != af) {
3481 continue;
3482 }
3483 if (sk->lan.xport.port != sxport->port) {
3484 continue;
3485 }
3486 if (!(PF_AEQ(&sk->lan.addr, saddr, af))) {
3487 continue;
3488 }
3489 nxport->port = sk->gwy.xport.port;
3490 return 0;
3491 }
3492 }
3493 do {
3494 key.af_gwy = af;
3495 key.proto = proto;
3496 PF_ACPY(&key.ext_gwy.addr, daddr, key.af_gwy);
3497 PF_ACPY(&key.gwy.addr, naddr, key.af_gwy);
3498 switch (proto) {
3499 case IPPROTO_UDP:
3500 key.proto_variant = r->extfilter;
3501 break;
3502 default:
3503 key.proto_variant = 0;
3504 break;
3505 }
3506 if (dxport) {
3507 key.ext_gwy.xport = *dxport;
3508 } else {
3509 memset(&key.ext_gwy.xport, 0,
3510 sizeof(key.ext_gwy.xport));
3511 }
3512 /*
3513 * port search; start random, step;
3514 * similar 2 portloop in in_pcbbind
3515 */
3516 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP ||
3517 proto == IPPROTO_ICMP)) {
3518 if (dxport) {
3519 key.gwy.xport = *dxport;
3520 } else {
3521 memset(&key.gwy.xport, 0,
3522 sizeof(key.gwy.xport));
3523 }
3524 if (pf_find_state_all(&key, PF_IN, NULL) == NULL) {
3525 return 0;
3526 }
3527 } else if (low == 0 && high == 0) {
3528 key.gwy.xport = *nxport;
3529 if (pf_find_state_all(&key, PF_IN, NULL) == NULL
3530 ) {
3531 return 0;
3532 }
3533 } else if (low == high) {
3534 key.gwy.xport.port = htons(low);
3535 if (pf_find_state_all(&key, PF_IN, NULL) == NULL
3536 ) {
3537 nxport->port = htons(low);
3538 return 0;
3539 }
3540 } else {
3541 unsigned int tmp;
3542 if (low > high) {
3543 tmp = low;
3544 low = high;
3545 high = tmp;
3546 }
3547 /* low < high */
3548 cut = htonl(random()) % (1 + high - low) + low;
3549 /* low <= cut <= high */
3550 for (tmp = cut; tmp <= high; ++(tmp)) {
3551 key.gwy.xport.port = htons(tmp);
3552 if (pf_find_state_all(&key, PF_IN, NULL) == NULL
3553 ) {
3554 nxport->port = htons(tmp);
3555 return 0;
3556 }
3557 }
3558 for (tmp = cut - 1; tmp >= low; --(tmp)) {
3559 key.gwy.xport.port = htons(tmp);
3560 if (pf_find_state_all(&key, PF_IN, NULL) == NULL
3561 ) {
3562 nxport->port = htons(tmp);
3563 return 0;
3564 }
3565 }
3566 }
3567
3568 switch (r->rpool.opts & PF_POOL_TYPEMASK) {
3569 case PF_POOL_RANDOM:
3570 case PF_POOL_ROUNDROBIN:
3571 if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn)) {
3572 return 1;
3573 }
3574 break;
3575 case PF_POOL_NONE:
3576 case PF_POOL_SRCHASH:
3577 case PF_POOL_BITMASK:
3578 default:
3579 return 1;
3580 }
3581 } while (!PF_AEQ(&init_addr, naddr, af));
3582
3583 return 1; /* none available */
3584 }
3585
3586 static struct pf_rule *
3587 pf_match_translation(struct pf_pdesc *pd, pbuf_t *pbuf, int off,
3588 int direction, struct pfi_kif *kif, struct pf_addr *saddr,
3589 union pf_state_xport *sxport, struct pf_addr *daddr,
3590 union pf_state_xport *dxport, int rs_num)
3591 {
3592 struct pf_rule *r, *rm = NULL;
3593 struct pf_ruleset *ruleset = NULL;
3594 int tag = -1;
3595 unsigned int rtableid = IFSCOPE_NONE;
3596 int asd = 0;
3597
3598 r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr);
3599 while (r && rm == NULL) {
3600 struct pf_rule_addr *src = NULL, *dst = NULL;
3601 struct pf_addr_wrap *xdst = NULL;
3602 struct pf_addr_wrap *xsrc = NULL;
3603 union pf_rule_xport rdrxport;
3604
3605 if (r->action == PF_BINAT && direction == PF_IN) {
3606 src = &r->dst;
3607 if (r->rpool.cur != NULL) {
3608 xdst = &r->rpool.cur->addr;
3609 }
3610 } else if (r->action == PF_RDR && direction == PF_OUT) {
3611 dst = &r->src;
3612 src = &r->dst;
3613 if (r->rpool.cur != NULL) {
3614 rdrxport.range.op = PF_OP_EQ;
3615 rdrxport.range.port[0] =
3616 htons(r->rpool.proxy_port[0]);
3617 xsrc = &r->rpool.cur->addr;
3618 }
3619 } else {
3620 src = &r->src;
3621 dst = &r->dst;
3622 }
3623
3624 r->evaluations++;
3625 if (pfi_kif_match(r->kif, kif) == r->ifnot) {
3626 r = r->skip[PF_SKIP_IFP].ptr;
3627 } else if (r->direction && r->direction != direction) {
3628 r = r->skip[PF_SKIP_DIR].ptr;
3629 } else if (r->af && r->af != pd->af) {
3630 r = r->skip[PF_SKIP_AF].ptr;
3631 } else if (r->proto && r->proto != pd->proto) {
3632 r = r->skip[PF_SKIP_PROTO].ptr;
3633 } else if (xsrc && PF_MISMATCHAW(xsrc, saddr, pd->af, 0, NULL)) {
3634 r = TAILQ_NEXT(r, entries);
3635 } else if (!xsrc && PF_MISMATCHAW(&src->addr, saddr, pd->af,
3636 src->neg, kif)) {
3637 r = TAILQ_NEXT(r, entries);
3638 } else if (xsrc && (!rdrxport.range.port[0] ||
3639 !pf_match_xport(r->proto, r->proto_variant, &rdrxport,
3640 sxport))) {
3641 r = TAILQ_NEXT(r, entries);
3642 } else if (!xsrc && !pf_match_xport(r->proto,
3643 r->proto_variant, &src->xport, sxport)) {
3644 r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT :
3645 PF_SKIP_DST_PORT].ptr;
3646 } else if (dst != NULL &&
3647 PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->neg, NULL)) {
3648 r = r->skip[PF_SKIP_DST_ADDR].ptr;
3649 } else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af,
3650 0, NULL)) {
3651 r = TAILQ_NEXT(r, entries);
3652 } else if (dst && !pf_match_xport(r->proto, r->proto_variant,
3653 &dst->xport, dxport)) {
3654 r = r->skip[PF_SKIP_DST_PORT].ptr;
3655 } else if (r->match_tag && !pf_match_tag(r, pd->pf_mtag, &tag)) {
3656 r = TAILQ_NEXT(r, entries);
3657 } else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto !=
3658 IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, pbuf,
3659 off, pd->hdr.tcp), r->os_fingerprint))) {
3660 r = TAILQ_NEXT(r, entries);
3661 } else {
3662 if (r->tag) {
3663 tag = r->tag;
3664 }
3665 if (PF_RTABLEID_IS_VALID(r->rtableid)) {
3666 rtableid = r->rtableid;
3667 }
3668 if (r->anchor == NULL) {
3669 rm = r;
3670 } else {
3671 pf_step_into_anchor(&asd, &ruleset, rs_num,
3672 &r, NULL, NULL);
3673 }
3674 }
3675 if (r == NULL) {
3676 pf_step_out_of_anchor(&asd, &ruleset, rs_num, &r,
3677 NULL, NULL);
3678 }
3679 }
3680 if (pf_tag_packet(pbuf, pd->pf_mtag, tag, rtableid, NULL)) {
3681 return NULL;
3682 }
3683 if (rm != NULL && (rm->action == PF_NONAT ||
3684 rm->action == PF_NORDR || rm->action == PF_NOBINAT ||
3685 rm->action == PF_NONAT64)) {
3686 return NULL;
3687 }
3688 return rm;
3689 }
3690
3691 /*
3692 * Get address translation information for NAT/BINAT/RDR
3693 * pd : pf packet descriptor
3694 * pbuf : pbuf holding the packet
3695 * off : offset to protocol header
3696 * direction : direction of packet
3697 * kif : pf interface info obtained from the packet's recv interface
3698 * sn : source node pointer (output)
3699 * saddr : packet source address
3700 * sxport : packet source port
3701 * daddr : packet destination address
3702 * dxport : packet destination port
3703 * nsxport : translated source port (output)
3704 *
3705 * Translated source & destination address are updated in pd->nsaddr &
3706 * pd->ndaddr
3707 */
3708 static struct pf_rule *
3709 pf_get_translation_aux(struct pf_pdesc *pd, pbuf_t *pbuf, int off,
3710 int direction, struct pfi_kif *kif, struct pf_src_node **sn,
3711 struct pf_addr *saddr, union pf_state_xport *sxport, struct pf_addr *daddr,
3712 union pf_state_xport *dxport, union pf_state_xport *nsxport
3713 )
3714 {
3715 struct pf_rule *r = NULL;
3716 pd->naf = pd->af;
3717
3718 if (direction == PF_OUT) {
3719 r = pf_match_translation(pd, pbuf, off, direction, kif, saddr,
3720 sxport, daddr, dxport, PF_RULESET_BINAT);
3721 if (r == NULL) {
3722 r = pf_match_translation(pd, pbuf, off, direction, kif,
3723 saddr, sxport, daddr, dxport, PF_RULESET_RDR);
3724 }
3725 if (r == NULL) {
3726 r = pf_match_translation(pd, pbuf, off, direction, kif,
3727 saddr, sxport, daddr, dxport, PF_RULESET_NAT);
3728 }
3729 } else {
3730 r = pf_match_translation(pd, pbuf, off, direction, kif, saddr,
3731 sxport, daddr, dxport, PF_RULESET_RDR);
3732 if (r == NULL) {
3733 r = pf_match_translation(pd, pbuf, off, direction, kif,
3734 saddr, sxport, daddr, dxport, PF_RULESET_BINAT);
3735 }
3736 }
3737
3738 if (r != NULL) {
3739 struct pf_addr *nsaddr = &pd->naddr;
3740 struct pf_addr *ndaddr = &pd->ndaddr;
3741
3742 *nsaddr = *saddr;
3743 *ndaddr = *daddr;
3744
3745 switch (r->action) {
3746 case PF_NONAT:
3747 case PF_NONAT64:
3748 case PF_NOBINAT:
3749 case PF_NORDR:
3750 return NULL;
3751 case PF_NAT:
3752 case PF_NAT64:
3753 /*
3754 * we do NAT64 on incoming path and we call ip_input
3755 * which asserts receive interface to be not NULL.
3756 * The below check is to prevent NAT64 action on any
3757 * packet generated by local entity using synthesized
3758 * IPv6 address.
3759 */
3760 if ((r->action == PF_NAT64) && (direction == PF_OUT)) {
3761 return NULL;
3762 }
3763
3764 if (pf_get_sport(pd, kif, r, saddr, sxport, daddr,
3765 dxport, nsaddr, nsxport, sn
3766 )) {
3767 DPFPRINTF(PF_DEBUG_MISC,
3768 ("pf: NAT proxy port allocation "
3769 "(%u-%u) failed\n",
3770 r->rpool.proxy_port[0],
3771 r->rpool.proxy_port[1]));
3772 return NULL;
3773 }
3774 /*
3775 * For NAT64 the destination IPv4 address is derived
3776 * from the last 32 bits of synthesized IPv6 address
3777 */
3778 if (r->action == PF_NAT64) {
3779 ndaddr->v4addr.s_addr = daddr->addr32[3];
3780 pd->naf = AF_INET;
3781 }
3782 break;
3783 case PF_BINAT:
3784 switch (direction) {
3785 case PF_OUT:
3786 if (r->rpool.cur->addr.type ==
3787 PF_ADDR_DYNIFTL) {
3788 if (r->rpool.cur->addr.p.dyn == NULL) {
3789 return NULL;
3790 }
3791 switch (pd->af) {
3792 #if INET
3793 case AF_INET:
3794 if (r->rpool.cur->addr.p.dyn->
3795 pfid_acnt4 < 1) {
3796 return NULL;
3797 }
3798 PF_POOLMASK(nsaddr,
3799 &r->rpool.cur->addr.p.dyn->
3800 pfid_addr4,
3801 &r->rpool.cur->addr.p.dyn->
3802 pfid_mask4,
3803 saddr, AF_INET);
3804 break;
3805 #endif /* INET */
3806 case AF_INET6:
3807 if (r->rpool.cur->addr.p.dyn->
3808 pfid_acnt6 < 1) {
3809 return NULL;
3810 }
3811 PF_POOLMASK(nsaddr,
3812 &r->rpool.cur->addr.p.dyn->
3813 pfid_addr6,
3814 &r->rpool.cur->addr.p.dyn->
3815 pfid_mask6,
3816 saddr, AF_INET6);
3817 break;
3818 }
3819 } else {
3820 PF_POOLMASK(nsaddr,
3821 &r->rpool.cur->addr.v.a.addr,
3822 &r->rpool.cur->addr.v.a.mask,
3823 saddr, pd->af);
3824 }
3825 break;
3826 case PF_IN:
3827 if (r->src.addr.type == PF_ADDR_DYNIFTL) {
3828 if (r->src.addr.p.dyn == NULL) {
3829 return NULL;
3830 }
3831 switch (pd->af) {
3832 #if INET
3833 case AF_INET:
3834 if (r->src.addr.p.dyn->
3835 pfid_acnt4 < 1) {
3836 return NULL;
3837 }
3838 PF_POOLMASK(ndaddr,
3839 &r->src.addr.p.dyn->
3840 pfid_addr4,
3841 &r->src.addr.p.dyn->
3842 pfid_mask4,
3843 daddr, AF_INET);
3844 break;
3845 #endif /* INET */
3846 case AF_INET6:
3847 if (r->src.addr.p.dyn->
3848 pfid_acnt6 < 1) {
3849 return NULL;
3850 }
3851 PF_POOLMASK(ndaddr,
3852 &r->src.addr.p.dyn->
3853 pfid_addr6,
3854 &r->src.addr.p.dyn->
3855 pfid_mask6,
3856 daddr, AF_INET6);
3857 break;
3858 }
3859 } else {
3860 PF_POOLMASK(ndaddr,
3861 &r->src.addr.v.a.addr,
3862 &r->src.addr.v.a.mask, daddr,
3863 pd->af);
3864 }
3865 break;
3866 }
3867 break;
3868 case PF_RDR: {
3869 switch (direction) {
3870 case PF_OUT:
3871 if (r->dst.addr.type == PF_ADDR_DYNIFTL) {
3872 if (r->dst.addr.p.dyn == NULL) {
3873 return NULL;
3874 }
3875 switch (pd->af) {
3876 #if INET
3877 case AF_INET:
3878 if (r->dst.addr.p.dyn->
3879 pfid_acnt4 < 1) {
3880 return NULL;
3881 }
3882 PF_POOLMASK(nsaddr,
3883 &r->dst.addr.p.dyn->
3884 pfid_addr4,
3885 &r->dst.addr.p.dyn->
3886 pfid_mask4,
3887 daddr, AF_INET);
3888 break;
3889 #endif /* INET */
3890 case AF_INET6:
3891 if (r->dst.addr.p.dyn->
3892 pfid_acnt6 < 1) {
3893 return NULL;
3894 }
3895 PF_POOLMASK(nsaddr,
3896 &r->dst.addr.p.dyn->
3897 pfid_addr6,
3898 &r->dst.addr.p.dyn->
3899 pfid_mask6,
3900 daddr, AF_INET6);
3901 break;
3902 }
3903 } else {
3904 PF_POOLMASK(nsaddr,
3905 &r->dst.addr.v.a.addr,
3906 &r->dst.addr.v.a.mask,
3907 daddr, pd->af);
3908 }
3909 if (nsxport && r->dst.xport.range.port[0]) {
3910 nsxport->port =
3911 r->dst.xport.range.port[0];
3912 }
3913 break;
3914 case PF_IN:
3915 if (pf_map_addr(pd->af, r, saddr,
3916 ndaddr, NULL, sn)) {
3917 return NULL;
3918 }
3919 if ((r->rpool.opts & PF_POOL_TYPEMASK) ==
3920 PF_POOL_BITMASK) {
3921 PF_POOLMASK(ndaddr, ndaddr,
3922 &r->rpool.cur->addr.v.a.mask, daddr,
3923 pd->af);
3924 }
3925
3926 if (nsxport && dxport) {
3927 if (r->rpool.proxy_port[1]) {
3928 u_int32_t tmp_nport;
3929
3930 tmp_nport =
3931 ((ntohs(dxport->port) -
3932 ntohs(r->dst.xport.range.
3933 port[0])) %
3934 (r->rpool.proxy_port[1] -
3935 r->rpool.proxy_port[0] +
3936 1)) + r->rpool.proxy_port[0];
3937
3938 /* wrap around if necessary */
3939 if (tmp_nport > 65535) {
3940 tmp_nport -= 65535;
3941 }
3942 nsxport->port =
3943 htons((u_int16_t)tmp_nport);
3944 } else if (r->rpool.proxy_port[0]) {
3945 nsxport->port = htons(r->rpool.
3946 proxy_port[0]);
3947 }
3948 }
3949 break;
3950 }
3951 break;
3952 }
3953 default:
3954 return NULL;
3955 }
3956 }
3957
3958 return r;
3959 }
3960
3961 int
3962 pf_socket_lookup(int direction, struct pf_pdesc *pd)
3963 {
3964 struct pf_addr *saddr, *daddr;
3965 u_int16_t sport, dport;
3966 struct inpcbinfo *pi;
3967 int inp = 0;
3968
3969 if (pd == NULL) {
3970 return -1;
3971 }
3972 pd->lookup.uid = UID_MAX;
3973 pd->lookup.gid = GID_MAX;
3974 pd->lookup.pid = NO_PID;
3975
3976 switch (pd->proto) {
3977 case IPPROTO_TCP:
3978 if (pd->hdr.tcp == NULL) {
3979 return -1;
3980 }
3981 sport = pd->hdr.tcp->th_sport;
3982 dport = pd->hdr.tcp->th_dport;
3983 pi = &tcbinfo;
3984 break;
3985 case IPPROTO_UDP:
3986 if (pd->hdr.udp == NULL) {
3987 return -1;
3988 }
3989 sport = pd->hdr.udp->uh_sport;
3990 dport = pd->hdr.udp->uh_dport;
3991 pi = &udbinfo;
3992 break;
3993 default:
3994 return -1;
3995 }
3996 if (direction == PF_IN) {
3997 saddr = pd->src;
3998 daddr = pd->dst;
3999 } else {
4000 u_int16_t p;
4001
4002 p = sport;
4003 sport = dport;
4004 dport = p;
4005 saddr = pd->dst;
4006 daddr = pd->src;
4007 }
4008 switch (pd->af) {
4009 #if INET
4010 case AF_INET:
4011 inp = in_pcblookup_hash_exists(pi, saddr->v4addr, sport, daddr->v4addr, dport,
4012 0, &pd->lookup.uid, &pd->lookup.gid, NULL);
4013 if (inp == 0) {
4014 struct in6_addr s6, d6;
4015
4016 memset(&s6, 0, sizeof(s6));
4017 s6.s6_addr16[5] = htons(0xffff);
4018 memcpy(&s6.s6_addr32[3], &saddr->v4addr,
4019 sizeof(saddr->v4addr));
4020
4021 memset(&d6, 0, sizeof(d6));
4022 d6.s6_addr16[5] = htons(0xffff);
4023 memcpy(&d6.s6_addr32[3], &daddr->v4addr,
4024 sizeof(daddr->v4addr));
4025
4026 inp = in6_pcblookup_hash_exists(pi, &s6, sport,
4027 &d6, dport, 0, &pd->lookup.uid, &pd->lookup.gid, NULL);
4028 if (inp == 0) {
4029 inp = in_pcblookup_hash_exists(pi, saddr->v4addr, sport,
4030 daddr->v4addr, dport, INPLOOKUP_WILDCARD, &pd->lookup.uid, &pd->lookup.gid, NULL);
4031 if (inp == 0) {
4032 inp = in6_pcblookup_hash_exists(pi, &s6, sport,
4033 &d6, dport, INPLOOKUP_WILDCARD,
4034 &pd->lookup.uid, &pd->lookup.gid, NULL);
4035 if (inp == 0) {
4036 return -1;
4037 }
4038 }
4039 }
4040 }
4041 break;
4042 #endif /* INET */
4043 case AF_INET6:
4044 inp = in6_pcblookup_hash_exists(pi, &saddr->v6addr, sport, &daddr->v6addr,
4045 dport, 0, &pd->lookup.uid, &pd->lookup.gid, NULL);
4046 if (inp == 0) {
4047 inp = in6_pcblookup_hash_exists(pi, &saddr->v6addr, sport,
4048 &daddr->v6addr, dport, INPLOOKUP_WILDCARD,
4049 &pd->lookup.uid, &pd->lookup.gid, NULL);
4050 if (inp == 0) {
4051 return -1;
4052 }
4053 }
4054 break;
4055
4056 default:
4057 return -1;
4058 }
4059
4060 return 1;
4061 }
4062
4063 static u_int8_t
4064 pf_get_wscale(pbuf_t *pbuf, int off, u_int16_t th_off, sa_family_t af)
4065 {
4066 int hlen;
4067 u_int8_t hdr[60];
4068 u_int8_t *opt, optlen;
4069 u_int8_t wscale = 0;
4070
4071 hlen = th_off << 2; /* hlen <= sizeof (hdr) */
4072 if (hlen <= (int)sizeof(struct tcphdr)) {
4073 return 0;
4074 }
4075 if (!pf_pull_hdr(pbuf, off, hdr, hlen, NULL, NULL, af)) {
4076 return 0;
4077 }
4078 opt = hdr + sizeof(struct tcphdr);
4079 hlen -= sizeof(struct tcphdr);
4080 while (hlen >= 3) {
4081 switch (*opt) {
4082 case TCPOPT_EOL:
4083 case TCPOPT_NOP:
4084 ++opt;
4085 --hlen;
4086 break;
4087 case TCPOPT_WINDOW:
4088 wscale = opt[2];
4089 if (wscale > TCP_MAX_WINSHIFT) {
4090 wscale = TCP_MAX_WINSHIFT;
4091 }
4092 wscale |= PF_WSCALE_FLAG;
4093 OS_FALLTHROUGH;
4094 default:
4095 optlen = opt[1];
4096 if (optlen < 2) {
4097 optlen = 2;
4098 }
4099 hlen -= optlen;
4100 opt += optlen;
4101 break;
4102 }
4103 }
4104 return wscale;
4105 }
4106
4107 static u_int16_t
4108 pf_get_mss(pbuf_t *pbuf, int off, u_int16_t th_off, sa_family_t af)
4109 {
4110 int hlen;
4111 u_int8_t hdr[60];
4112 u_int8_t *opt, optlen;
4113 u_int16_t mss = tcp_mssdflt;
4114
4115 hlen = th_off << 2; /* hlen <= sizeof (hdr) */
4116 if (hlen <= (int)sizeof(struct tcphdr)) {
4117 return 0;
4118 }
4119 if (!pf_pull_hdr(pbuf, off, hdr, hlen, NULL, NULL, af)) {
4120 return 0;
4121 }
4122 opt = hdr + sizeof(struct tcphdr);
4123 hlen -= sizeof(struct tcphdr);
4124 while (hlen >= TCPOLEN_MAXSEG) {
4125 switch (*opt) {
4126 case TCPOPT_EOL:
4127 case TCPOPT_NOP:
4128 ++opt;
4129 --hlen;
4130 break;
4131 case TCPOPT_MAXSEG:
4132 bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2);
4133 #if BYTE_ORDER != BIG_ENDIAN
4134 NTOHS(mss);
4135 #endif
4136 OS_FALLTHROUGH;
4137 default:
4138 optlen = opt[1];
4139 if (optlen < 2) {
4140 optlen = 2;
4141 }
4142 hlen -= optlen;
4143 opt += optlen;
4144 break;
4145 }
4146 }
4147 return mss;
4148 }
4149
4150 static u_int16_t
4151 pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer)
4152 {
4153 #if INET
4154 struct sockaddr_in *dst;
4155 struct route ro;
4156 #endif /* INET */
4157 struct sockaddr_in6 *dst6;
4158 struct route_in6 ro6;
4159 struct rtentry *rt = NULL;
4160 int hlen;
4161 u_int16_t mss = tcp_mssdflt;
4162
4163 switch (af) {
4164 #if INET
4165 case AF_INET:
4166 hlen = sizeof(struct ip);
4167 bzero(&ro, sizeof(ro));
4168 dst = (struct sockaddr_in *)(void *)&ro.ro_dst;
4169 dst->sin_family = AF_INET;
4170 dst->sin_len = sizeof(*dst);
4171 dst->sin_addr = addr->v4addr;
4172 rtalloc(&ro);
4173 rt = ro.ro_rt;
4174 break;
4175 #endif /* INET */
4176 case AF_INET6:
4177 hlen = sizeof(struct ip6_hdr);
4178 bzero(&ro6, sizeof(ro6));
4179 dst6 = (struct sockaddr_in6 *)(void *)&ro6.ro_dst;
4180 dst6->sin6_family = AF_INET6;
4181 dst6->sin6_len = sizeof(*dst6);
4182 dst6->sin6_addr = addr->v6addr;
4183 rtalloc((struct route *)&ro);
4184 rt = ro6.ro_rt;
4185 break;
4186 default:
4187 panic("pf_calc_mss: not AF_INET or AF_INET6!");
4188 return 0;
4189 }
4190
4191 if (rt && rt->rt_ifp) {
4192 /* This is relevant only for PF SYN Proxy */
4193 int interface_mtu = rt->rt_ifp->if_mtu;
4194
4195 if (af == AF_INET &&
4196 INTF_ADJUST_MTU_FOR_CLAT46(rt->rt_ifp)) {
4197 interface_mtu = IN6_LINKMTU(rt->rt_ifp);
4198 /* Further adjust the size for CLAT46 expansion */
4199 interface_mtu -= CLAT46_HDR_EXPANSION_OVERHD;
4200 }
4201 mss = interface_mtu - hlen - sizeof(struct tcphdr);
4202 mss = max(tcp_mssdflt, mss);
4203 rtfree(rt);
4204 }
4205 mss = min(mss, offer);
4206 mss = max(mss, 64); /* sanity - at least max opt space */
4207 return mss;
4208 }
4209
4210 static void
4211 pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr, sa_family_t af)
4212 {
4213 struct pf_rule *r = s->rule.ptr;
4214
4215 s->rt_kif = NULL;
4216
4217 if (!r->rt || r->rt == PF_FASTROUTE) {
4218 return;
4219 }
4220 if ((af == AF_INET) || (af == AF_INET6)) {
4221 pf_map_addr(af, r, saddr, &s->rt_addr, NULL,
4222 &s->nat_src_node);
4223 s->rt_kif = r->rpool.cur->kif;
4224 }
4225
4226 return;
4227 }
4228
4229 static void
4230 pf_attach_state(struct pf_state_key *sk, struct pf_state *s, int tail)
4231 {
4232 s->state_key = sk;
4233 sk->refcnt++;
4234
4235 /* list is sorted, if-bound states before floating */
4236 if (tail) {
4237 TAILQ_INSERT_TAIL(&sk->states, s, next);
4238 } else {
4239 TAILQ_INSERT_HEAD(&sk->states, s, next);
4240 }
4241 }
4242
4243 static void
4244 pf_detach_state(struct pf_state *s, int flags)
4245 {
4246 struct pf_state_key *sk = s->state_key;
4247
4248 if (sk == NULL) {
4249 return;
4250 }
4251
4252 s->state_key = NULL;
4253 TAILQ_REMOVE(&sk->states, s, next);
4254 if (--sk->refcnt == 0) {
4255 if (!(flags & PF_DT_SKIP_EXTGWY)) {
4256 RB_REMOVE(pf_state_tree_ext_gwy,
4257 &pf_statetbl_ext_gwy, sk);
4258 }
4259 if (!(flags & PF_DT_SKIP_LANEXT)) {
4260 RB_REMOVE(pf_state_tree_lan_ext,
4261 &pf_statetbl_lan_ext, sk);
4262 }
4263 if (sk->app_state) {
4264 pool_put(&pf_app_state_pl, sk->app_state);
4265 }
4266 pool_put(&pf_state_key_pl, sk);
4267 }
4268 }
4269
4270 struct pf_state_key *
4271 pf_alloc_state_key(struct pf_state *s, struct pf_state_key *psk)
4272 {
4273 struct pf_state_key *sk;
4274
4275 if ((sk = pool_get(&pf_state_key_pl, PR_WAITOK)) == NULL) {
4276 return NULL;
4277 }
4278 bzero(sk, sizeof(*sk));
4279 TAILQ_INIT(&sk->states);
4280 pf_attach_state(sk, s, 0);
4281
4282 /* initialize state key from psk, if provided */
4283 if (psk != NULL) {
4284 bcopy(&psk->lan, &sk->lan, sizeof(sk->lan));
4285 bcopy(&psk->gwy, &sk->gwy, sizeof(sk->gwy));
4286 bcopy(&psk->ext_lan, &sk->ext_lan, sizeof(sk->ext_lan));
4287 bcopy(&psk->ext_gwy, &sk->ext_gwy, sizeof(sk->ext_gwy));
4288 sk->af_lan = psk->af_lan;
4289 sk->af_gwy = psk->af_gwy;
4290 sk->proto = psk->proto;
4291 sk->direction = psk->direction;
4292 sk->proto_variant = psk->proto_variant;
4293 VERIFY(psk->app_state == NULL);
4294 sk->flowsrc = psk->flowsrc;
4295 sk->flowhash = psk->flowhash;
4296 /* don't touch tree entries, states and refcnt on sk */
4297 }
4298
4299 return sk;
4300 }
4301
4302 static u_int32_t
4303 pf_tcp_iss(struct pf_pdesc *pd)
4304 {
4305 MD5_CTX ctx;
4306 u_int32_t digest[4];
4307
4308 if (pf_tcp_secret_init == 0) {
4309 read_frandom(pf_tcp_secret, sizeof(pf_tcp_secret));
4310 MD5Init(&pf_tcp_secret_ctx);
4311 MD5Update(&pf_tcp_secret_ctx, pf_tcp_secret,
4312 sizeof(pf_tcp_secret));
4313 pf_tcp_secret_init = 1;
4314 }
4315 ctx = pf_tcp_secret_ctx;
4316
4317 MD5Update(&ctx, (char *)&pd->hdr.tcp->th_sport, sizeof(u_short));
4318 MD5Update(&ctx, (char *)&pd->hdr.tcp->th_dport, sizeof(u_short));
4319 if (pd->af == AF_INET6) {
4320 MD5Update(&ctx, (char *)&pd->src->v6addr, sizeof(struct in6_addr));
4321 MD5Update(&ctx, (char *)&pd->dst->v6addr, sizeof(struct in6_addr));
4322 } else {
4323 MD5Update(&ctx, (char *)&pd->src->v4addr, sizeof(struct in_addr));
4324 MD5Update(&ctx, (char *)&pd->dst->v4addr, sizeof(struct in_addr));
4325 }
4326 MD5Final((u_char *)digest, &ctx);
4327 pf_tcp_iss_off += 4096;
4328 return digest[0] + random() + pf_tcp_iss_off;
4329 }
4330
4331 /*
4332 * This routine is called to perform address family translation on the
4333 * inner IP header (that may come as payload) of an ICMP(v4addr/6) error
4334 * response.
4335 */
4336 static int
4337 pf_change_icmp_af(pbuf_t *pbuf, int off,
4338 struct pf_pdesc *pd, struct pf_pdesc *pd2, struct pf_addr *src,
4339 struct pf_addr *dst, sa_family_t af, sa_family_t naf)
4340 {
4341 struct ip *ip4 = NULL;
4342 struct ip6_hdr *ip6 = NULL;
4343 void *hdr;
4344 int hlen, olen;
4345
4346 if (af == naf || (af != AF_INET && af != AF_INET6) ||
4347 (naf != AF_INET && naf != AF_INET6)) {
4348 return -1;
4349 }
4350
4351 /* old header */
4352 olen = pd2->off - off;
4353 /* new header */
4354 hlen = naf == AF_INET ? sizeof(*ip4) : sizeof(*ip6);
4355
4356 /* Modify the pbuf to accommodate the new header */
4357 hdr = pbuf_resize_segment(pbuf, off, olen, hlen);
4358 if (hdr == NULL) {
4359 return -1;
4360 }
4361
4362 /* translate inner ip/ip6 header */
4363 switch (naf) {
4364 case AF_INET:
4365 ip4 = hdr;
4366 bzero(ip4, sizeof(*ip4));
4367 ip4->ip_v = IPVERSION;
4368 ip4->ip_hl = sizeof(*ip4) >> 2;
4369 ip4->ip_len = htons(sizeof(*ip4) + pd2->tot_len - olen);
4370 ip4->ip_id = rfc6864 ? 0 : htons(ip_randomid());
4371 ip4->ip_off = htons(IP_DF);
4372 ip4->ip_ttl = pd2->ttl;
4373 if (pd2->proto == IPPROTO_ICMPV6) {
4374 ip4->ip_p = IPPROTO_ICMP;
4375 } else {
4376 ip4->ip_p = pd2->proto;
4377 }
4378 ip4->ip_src = src->v4addr;
4379 ip4->ip_dst = dst->v4addr;
4380 ip4->ip_sum = pbuf_inet_cksum(pbuf, 0, 0, ip4->ip_hl << 2);
4381 break;
4382 case AF_INET6:
4383 ip6 = hdr;
4384 bzero(ip6, sizeof(*ip6));
4385 ip6->ip6_vfc = IPV6_VERSION;
4386 ip6->ip6_plen = htons(pd2->tot_len - olen);
4387 if (pd2->proto == IPPROTO_ICMP) {
4388 ip6->ip6_nxt = IPPROTO_ICMPV6;
4389 } else {
4390 ip6->ip6_nxt = pd2->proto;
4391 }
4392 if (!pd2->ttl || pd2->ttl > IPV6_DEFHLIM) {
4393 ip6->ip6_hlim = IPV6_DEFHLIM;
4394 } else {
4395 ip6->ip6_hlim = pd2->ttl;
4396 }
4397 ip6->ip6_src = src->v6addr;
4398 ip6->ip6_dst = dst->v6addr;
4399 break;
4400 }
4401
4402 /* adjust payload offset and total packet length */
4403 pd2->off += hlen - olen;
4404 pd->tot_len += hlen - olen;
4405
4406 return 0;
4407 }
4408
4409 #define PTR_IP(field) ((int32_t)offsetof(struct ip, field))
4410 #define PTR_IP6(field) ((int32_t)offsetof(struct ip6_hdr, field))
4411
4412 static int
4413 pf_translate_icmp_af(int af, void *arg)
4414 {
4415 struct icmp *icmp4;
4416 struct icmp6_hdr *icmp6;
4417 u_int32_t mtu;
4418 int32_t ptr = -1;
4419 u_int8_t type;
4420 u_int8_t code;
4421
4422 switch (af) {
4423 case AF_INET:
4424 icmp6 = arg;
4425 type = icmp6->icmp6_type;
4426 code = icmp6->icmp6_code;
4427 mtu = ntohl(icmp6->icmp6_mtu);
4428
4429 switch (type) {
4430 case ICMP6_ECHO_REQUEST:
4431 type = ICMP_ECHO;
4432 break;
4433 case ICMP6_ECHO_REPLY:
4434 type = ICMP_ECHOREPLY;
4435 break;
4436 case ICMP6_DST_UNREACH:
4437 type = ICMP_UNREACH;
4438 switch (code) {
4439 case ICMP6_DST_UNREACH_NOROUTE:
4440 case ICMP6_DST_UNREACH_BEYONDSCOPE:
4441 case ICMP6_DST_UNREACH_ADDR:
4442 code = ICMP_UNREACH_HOST;
4443 break;
4444 case ICMP6_DST_UNREACH_ADMIN:
4445 code = ICMP_UNREACH_HOST_PROHIB;
4446 break;
4447 case ICMP6_DST_UNREACH_NOPORT:
4448 code = ICMP_UNREACH_PORT;
4449 break;
4450 default:
4451 return -1;
4452 }
4453 break;
4454 case ICMP6_PACKET_TOO_BIG:
4455 type = ICMP_UNREACH;
4456 code = ICMP_UNREACH_NEEDFRAG;
4457 mtu -= 20;
4458 break;
4459 case ICMP6_TIME_EXCEEDED:
4460 type = ICMP_TIMXCEED;
4461 break;
4462 case ICMP6_PARAM_PROB:
4463 switch (code) {
4464 case ICMP6_PARAMPROB_HEADER:
4465 type = ICMP_PARAMPROB;
4466 code = ICMP_PARAMPROB_ERRATPTR;
4467 ptr = ntohl(icmp6->icmp6_pptr);
4468
4469 if (ptr == PTR_IP6(ip6_vfc)) {
4470 ; /* preserve */
4471 } else if (ptr == PTR_IP6(ip6_vfc) + 1) {
4472 ptr = PTR_IP(ip_tos);
4473 } else if (ptr == PTR_IP6(ip6_plen) ||
4474 ptr == PTR_IP6(ip6_plen) + 1) {
4475 ptr = PTR_IP(ip_len);
4476 } else if (ptr == PTR_IP6(ip6_nxt)) {
4477 ptr = PTR_IP(ip_p);
4478 } else if (ptr == PTR_IP6(ip6_hlim)) {
4479 ptr = PTR_IP(ip_ttl);
4480 } else if (ptr >= PTR_IP6(ip6_src) &&
4481 ptr < PTR_IP6(ip6_dst)) {
4482 ptr = PTR_IP(ip_src);
4483 } else if (ptr >= PTR_IP6(ip6_dst) &&
4484 ptr < (int32_t)sizeof(struct ip6_hdr)) {
4485 ptr = PTR_IP(ip_dst);
4486 } else {
4487 return -1;
4488 }
4489 break;
4490 case ICMP6_PARAMPROB_NEXTHEADER:
4491 type = ICMP_UNREACH;
4492 code = ICMP_UNREACH_PROTOCOL;
4493 break;
4494 default:
4495 return -1;
4496 }
4497 break;
4498 default:
4499 return -1;
4500 }
4501 icmp6->icmp6_type = type;
4502 icmp6->icmp6_code = code;
4503 /* aligns well with a icmpv4 nextmtu */
4504 icmp6->icmp6_mtu = htonl(mtu);
4505 /* icmpv4 pptr is a one most significant byte */
4506 if (ptr >= 0) {
4507 icmp6->icmp6_pptr = htonl(ptr << 24);
4508 }
4509 break;
4510
4511 case AF_INET6:
4512 icmp4 = arg;
4513 type = icmp4->icmp_type;
4514 code = icmp4->icmp_code;
4515 mtu = ntohs(icmp4->icmp_nextmtu);
4516
4517 switch (type) {
4518 case ICMP_ECHO:
4519 type = ICMP6_ECHO_REQUEST;
4520 break;
4521 case ICMP_ECHOREPLY:
4522 type = ICMP6_ECHO_REPLY;
4523 break;
4524 case ICMP_UNREACH:
4525 type = ICMP6_DST_UNREACH;
4526 switch (code) {
4527 case ICMP_UNREACH_NET:
4528 case ICMP_UNREACH_HOST:
4529 case ICMP_UNREACH_NET_UNKNOWN:
4530 case ICMP_UNREACH_HOST_UNKNOWN:
4531 case ICMP_UNREACH_ISOLATED:
4532 case ICMP_UNREACH_TOSNET:
4533 case ICMP_UNREACH_TOSHOST:
4534 code = ICMP6_DST_UNREACH_NOROUTE;
4535 break;
4536 case ICMP_UNREACH_PORT:
4537 code = ICMP6_DST_UNREACH_NOPORT;
4538 break;
4539 case ICMP_UNREACH_NET_PROHIB:
4540 case ICMP_UNREACH_HOST_PROHIB:
4541 case ICMP_UNREACH_FILTER_PROHIB:
4542 case ICMP_UNREACH_PRECEDENCE_CUTOFF:
4543 code = ICMP6_DST_UNREACH_ADMIN;
4544 break;
4545 case ICMP_UNREACH_PROTOCOL:
4546 type = ICMP6_PARAM_PROB;
4547 code = ICMP6_PARAMPROB_NEXTHEADER;
4548 ptr = offsetof(struct ip6_hdr, ip6_nxt);
4549 break;
4550 case ICMP_UNREACH_NEEDFRAG:
4551 type = ICMP6_PACKET_TOO_BIG;
4552 code = 0;
4553 mtu += 20;
4554 break;
4555 default:
4556 return -1;
4557 }
4558 break;
4559 case ICMP_TIMXCEED:
4560 type = ICMP6_TIME_EXCEEDED;
4561 break;
4562 case ICMP_PARAMPROB:
4563 type = ICMP6_PARAM_PROB;
4564 switch (code) {
4565 case ICMP_PARAMPROB_ERRATPTR:
4566 code = ICMP6_PARAMPROB_HEADER;
4567 break;
4568 case ICMP_PARAMPROB_LENGTH:
4569 code = ICMP6_PARAMPROB_HEADER;
4570 break;
4571 default:
4572 return -1;
4573 }
4574
4575 ptr = icmp4->icmp_pptr;
4576 if (ptr == 0 || ptr == PTR_IP(ip_tos)) {
4577 ; /* preserve */
4578 } else if (ptr == PTR_IP(ip_len) ||
4579 ptr == PTR_IP(ip_len) + 1) {
4580 ptr = PTR_IP6(ip6_plen);
4581 } else if (ptr == PTR_IP(ip_ttl)) {
4582 ptr = PTR_IP6(ip6_hlim);
4583 } else if (ptr == PTR_IP(ip_p)) {
4584 ptr = PTR_IP6(ip6_nxt);
4585 } else if (ptr >= PTR_IP(ip_src) &&
4586 ptr < PTR_IP(ip_dst)) {
4587 ptr = PTR_IP6(ip6_src);
4588 } else if (ptr >= PTR_IP(ip_dst) &&
4589 ptr < (int32_t)sizeof(struct ip)) {
4590 ptr = PTR_IP6(ip6_dst);
4591 } else {
4592 return -1;
4593 }
4594 break;
4595 default:
4596 return -1;
4597 }
4598 icmp4->icmp_type = type;
4599 icmp4->icmp_code = code;
4600 icmp4->icmp_nextmtu = htons(mtu);
4601 if (ptr >= 0) {
4602 icmp4->icmp_void = htonl(ptr);
4603 }
4604 break;
4605 }
4606
4607 return 0;
4608 }
4609
4610 /* Note: frees pbuf if PF_NAT64 is returned */
4611 static int
4612 pf_nat64_ipv6(pbuf_t *pbuf, int off, struct pf_pdesc *pd)
4613 {
4614 struct ip *ip4;
4615 struct mbuf *m;
4616
4617 /*
4618 * ip_input asserts for rcvif to be not NULL
4619 * That may not be true for two corner cases
4620 * 1. If for some reason a local app sends DNS
4621 * AAAA query to local host
4622 * 2. If IPv6 stack in kernel internally generates a
4623 * message destined for a synthesized IPv6 end-point.
4624 */
4625 if (pbuf->pb_ifp == NULL) {
4626 return PF_DROP;
4627 }
4628
4629 ip4 = (struct ip *)pbuf_resize_segment(pbuf, 0, off, sizeof(*ip4));
4630 if (ip4 == NULL) {
4631 return PF_DROP;
4632 }
4633
4634 ip4->ip_v = 4;
4635 ip4->ip_hl = 5;
4636 ip4->ip_tos = pd->tos & htonl(0x0ff00000);
4637 ip4->ip_len = htons(sizeof(*ip4) + (pd->tot_len - off));
4638 ip4->ip_id = 0;
4639 ip4->ip_off = htons(IP_DF);
4640 ip4->ip_ttl = pd->ttl;
4641 ip4->ip_p = pd->proto;
4642 ip4->ip_sum = 0;
4643 ip4->ip_src = pd->naddr.v4addr;
4644 ip4->ip_dst = pd->ndaddr.v4addr;
4645 ip4->ip_sum = pbuf_inet_cksum(pbuf, 0, 0, ip4->ip_hl << 2);
4646
4647 /* recalculate icmp checksums */
4648 if (pd->proto == IPPROTO_ICMP) {
4649 struct icmp *icmp;
4650 int hlen = sizeof(*ip4);
4651
4652 icmp = (struct icmp *)pbuf_contig_segment(pbuf, hlen,
4653 ICMP_MINLEN);
4654 if (icmp == NULL) {
4655 return PF_DROP;
4656 }
4657
4658 icmp->icmp_cksum = 0;
4659 icmp->icmp_cksum = pbuf_inet_cksum(pbuf, 0, hlen,
4660 ntohs(ip4->ip_len) - hlen);
4661 }
4662
4663 if ((m = pbuf_to_mbuf(pbuf, TRUE)) != NULL) {
4664 ip_input(m);
4665 }
4666
4667 return PF_NAT64;
4668 }
4669
4670 static int
4671 pf_nat64_ipv4(pbuf_t *pbuf, int off, struct pf_pdesc *pd)
4672 {
4673 struct ip6_hdr *ip6;
4674 struct mbuf *m;
4675
4676 if (pbuf->pb_ifp == NULL) {
4677 return PF_DROP;
4678 }
4679
4680 ip6 = (struct ip6_hdr *)pbuf_resize_segment(pbuf, 0, off, sizeof(*ip6));
4681 if (ip6 == NULL) {
4682 return PF_DROP;
4683 }
4684
4685 ip6->ip6_vfc = htonl((6 << 28) | (pd->tos << 20));
4686 ip6->ip6_plen = htons(pd->tot_len - off);
4687 ip6->ip6_nxt = pd->proto;
4688 ip6->ip6_hlim = pd->ttl;
4689 ip6->ip6_src = pd->naddr.v6addr;
4690 ip6->ip6_dst = pd->ndaddr.v6addr;
4691
4692 /* recalculate icmp6 checksums */
4693 if (pd->proto == IPPROTO_ICMPV6) {
4694 struct icmp6_hdr *icmp6;
4695 int hlen = sizeof(*ip6);
4696
4697 icmp6 = (struct icmp6_hdr *)pbuf_contig_segment(pbuf, hlen,
4698 sizeof(*icmp6));
4699 if (icmp6 == NULL) {
4700 return PF_DROP;
4701 }
4702
4703 icmp6->icmp6_cksum = 0;
4704 icmp6->icmp6_cksum = pbuf_inet6_cksum(pbuf,
4705 IPPROTO_ICMPV6, hlen,
4706 ntohs(ip6->ip6_plen));
4707 } else if (pd->proto == IPPROTO_UDP) {
4708 struct udphdr *uh;
4709 int hlen = sizeof(*ip6);
4710
4711 uh = (struct udphdr *)pbuf_contig_segment(pbuf, hlen,
4712 sizeof(*uh));
4713 if (uh == NULL) {
4714 return PF_DROP;
4715 }
4716
4717 if (uh->uh_sum == 0) {
4718 uh->uh_sum = pbuf_inet6_cksum(pbuf, IPPROTO_UDP,
4719 hlen, ntohs(ip6->ip6_plen));
4720 }
4721 }
4722
4723 if ((m = pbuf_to_mbuf(pbuf, TRUE)) != NULL) {
4724 ip6_input(m);
4725 }
4726
4727 return PF_NAT64;
4728 }
4729
4730 static int
4731 pf_test_rule(struct pf_rule **rm, struct pf_state **sm, int direction,
4732 struct pfi_kif *kif, pbuf_t *pbuf, int off, void *h,
4733 struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm,
4734 struct ifqueue *ifq)
4735 {
4736 #pragma unused(h)
4737 struct pf_rule *nr = NULL;
4738 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
4739 sa_family_t af = pd->af;
4740 struct pf_rule *r, *a = NULL;
4741 struct pf_ruleset *ruleset = NULL;
4742 struct pf_src_node *nsn = NULL;
4743 struct tcphdr *th = pd->hdr.tcp;
4744 struct udphdr *uh = pd->hdr.udp;
4745 u_short reason;
4746 int rewrite = 0, hdrlen = 0;
4747 int tag = -1;
4748 unsigned int rtableid = IFSCOPE_NONE;
4749 int asd = 0;
4750 int match = 0;
4751 int state_icmp = 0;
4752 u_int16_t mss = tcp_mssdflt;
4753 u_int8_t icmptype = 0, icmpcode = 0;
4754
4755 struct pf_grev1_hdr *grev1 = pd->hdr.grev1;
4756 union pf_state_xport bxport, bdxport, nxport, sxport, dxport;
4757 struct pf_state_key psk;
4758
4759 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
4760
4761 if (direction == PF_IN && pf_check_congestion(ifq)) {
4762 REASON_SET(&reason, PFRES_CONGEST);
4763 return PF_DROP;
4764 }
4765
4766 hdrlen = 0;
4767 sxport.spi = 0;
4768 dxport.spi = 0;
4769 nxport.spi = 0;
4770
4771 switch (pd->proto) {
4772 case IPPROTO_TCP:
4773 sxport.port = th->th_sport;
4774 dxport.port = th->th_dport;
4775 hdrlen = sizeof(*th);
4776 break;
4777 case IPPROTO_UDP:
4778 sxport.port = uh->uh_sport;
4779 dxport.port = uh->uh_dport;
4780 hdrlen = sizeof(*uh);
4781 break;
4782 #if INET
4783 case IPPROTO_ICMP:
4784 if (pd->af != AF_INET) {
4785 break;
4786 }
4787 sxport.port = dxport.port = pd->hdr.icmp->icmp_id;
4788 hdrlen = ICMP_MINLEN;
4789 icmptype = pd->hdr.icmp->icmp_type;
4790 icmpcode = pd->hdr.icmp->icmp_code;
4791
4792 if (ICMP_ERRORTYPE(icmptype)) {
4793 state_icmp++;
4794 }
4795 break;
4796 #endif /* INET */
4797 case IPPROTO_ICMPV6:
4798 if (pd->af != AF_INET6) {
4799 break;
4800 }
4801 sxport.port = dxport.port = pd->hdr.icmp6->icmp6_id;
4802 hdrlen = sizeof(*pd->hdr.icmp6);
4803 icmptype = pd->hdr.icmp6->icmp6_type;
4804 icmpcode = pd->hdr.icmp6->icmp6_code;
4805
4806 if (ICMP6_ERRORTYPE(icmptype)) {
4807 state_icmp++;
4808 }
4809 break;
4810 case IPPROTO_GRE:
4811 if (pd->proto_variant == PF_GRE_PPTP_VARIANT) {
4812 sxport.call_id = dxport.call_id =
4813 pd->hdr.grev1->call_id;
4814 hdrlen = sizeof(*pd->hdr.grev1);
4815 }
4816 break;
4817 case IPPROTO_ESP:
4818 sxport.spi = 0;
4819 dxport.spi = pd->hdr.esp->spi;
4820 hdrlen = sizeof(*pd->hdr.esp);
4821 break;
4822 }
4823
4824 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
4825
4826 bxport = sxport;
4827 bdxport = dxport;
4828
4829 if (direction == PF_OUT) {
4830 nxport = sxport;
4831 } else {
4832 nxport = dxport;
4833 }
4834
4835 /* check packet for BINAT/NAT/RDR */
4836 if ((nr = pf_get_translation_aux(pd, pbuf, off, direction, kif, &nsn,
4837 saddr, &sxport, daddr, &dxport, &nxport
4838 )) != NULL) {
4839 int ua;
4840 u_int16_t dport;
4841
4842 if (pd->af != pd->naf) {
4843 ua = 0;
4844 } else {
4845 ua = 1;
4846 }
4847
4848 PF_ACPY(&pd->baddr, saddr, af);
4849 PF_ACPY(&pd->bdaddr, daddr, af);
4850
4851 switch (pd->proto) {
4852 case IPPROTO_TCP:
4853 if (pd->af != pd->naf ||
4854 PF_ANEQ(saddr, &pd->naddr, pd->af)) {
4855 pf_change_ap(direction, pd->mp, saddr,
4856 &th->th_sport, pd->ip_sum, &th->th_sum,
4857 &pd->naddr, nxport.port, 0, af,
4858 pd->naf, ua);
4859 sxport.port = th->th_sport;
4860 }
4861
4862 if (pd->af != pd->naf ||
4863 PF_ANEQ(daddr, &pd->ndaddr, pd->af) ||
4864 (nr && (nr->action == PF_RDR) &&
4865 (th->th_dport != nxport.port))) {
4866 if (nr && nr->action == PF_RDR) {
4867 dport = nxport.port;
4868 } else {
4869 dport = th->th_dport;
4870 }
4871 pf_change_ap(direction, pd->mp, daddr,
4872 &th->th_dport, pd->ip_sum,
4873 &th->th_sum, &pd->ndaddr,
4874 dport, 0, af, pd->naf, ua);
4875 dxport.port = th->th_dport;
4876 }
4877 rewrite++;
4878 break;
4879
4880 case IPPROTO_UDP:
4881 if (pd->af != pd->naf ||
4882 PF_ANEQ(saddr, &pd->naddr, pd->af)) {
4883 pf_change_ap(direction, pd->mp, saddr,
4884 &uh->uh_sport, pd->ip_sum,
4885 &uh->uh_sum, &pd->naddr,
4886 nxport.port, 1, af, pd->naf, ua);
4887 sxport.port = uh->uh_sport;
4888 }
4889
4890 if (pd->af != pd->naf ||
4891 PF_ANEQ(daddr, &pd->ndaddr, pd->af) ||
4892 (nr && (nr->action == PF_RDR) &&
4893 (uh->uh_dport != nxport.port))) {
4894 if (nr && nr->action == PF_RDR) {
4895 dport = nxport.port;
4896 } else {
4897 dport = uh->uh_dport;
4898 }
4899 pf_change_ap(direction, pd->mp, daddr,
4900 &uh->uh_dport, pd->ip_sum,
4901 &uh->uh_sum, &pd->ndaddr,
4902 dport, 0, af, pd->naf, ua);
4903 dxport.port = uh->uh_dport;
4904 }
4905 rewrite++;
4906 break;
4907 #if INET
4908 case IPPROTO_ICMP:
4909 if (pd->af != AF_INET) {
4910 break;
4911 }
4912 /*
4913 * TODO:
4914 * pd->af != pd->naf not handled yet here and would be
4915 * needed for NAT46 needed to support XLAT.
4916 * Will cross the bridge when it comes.
4917 */
4918 if (PF_ANEQ(saddr, &pd->naddr, pd->af)) {
4919 pf_change_a(&saddr->v4addr.s_addr, pd->ip_sum,
4920 pd->naddr.v4addr.s_addr, 0);
4921 pd->hdr.icmp->icmp_cksum = pf_cksum_fixup(
4922 pd->hdr.icmp->icmp_cksum, sxport.port,
4923 nxport.port, 0);
4924 pd->hdr.icmp->icmp_id = nxport.port;
4925 }
4926
4927 if (PF_ANEQ(daddr, &pd->ndaddr, pd->af)) {
4928 pf_change_a(&daddr->v4addr.s_addr, pd->ip_sum,
4929 pd->ndaddr.v4addr.s_addr, 0);
4930 }
4931 ++rewrite;
4932 break;
4933 #endif /* INET */
4934 case IPPROTO_ICMPV6:
4935 if (pd->af != AF_INET6) {
4936 break;
4937 }
4938
4939 if (pd->af != pd->naf ||
4940 PF_ANEQ(saddr, &pd->naddr, pd->af)) {
4941 pf_change_addr(saddr,
4942 &pd->hdr.icmp6->icmp6_cksum,
4943 &pd->naddr, 0, pd->af, pd->naf);
4944 }
4945
4946 if (pd->af != pd->naf ||
4947 PF_ANEQ(daddr, &pd->ndaddr, pd->af)) {
4948 pf_change_addr(daddr,
4949 &pd->hdr.icmp6->icmp6_cksum,
4950 &pd->ndaddr, 0, pd->af, pd->naf);
4951 }
4952
4953 if (pd->af != pd->naf) {
4954 if (pf_translate_icmp_af(AF_INET,
4955 pd->hdr.icmp6)) {
4956 return PF_DROP;
4957 }
4958 pd->proto = IPPROTO_ICMP;
4959 }
4960 rewrite++;
4961 break;
4962 case IPPROTO_GRE:
4963 if ((direction == PF_IN) &&
4964 (pd->proto_variant == PF_GRE_PPTP_VARIANT)) {
4965 grev1->call_id = nxport.call_id;
4966 }
4967
4968 switch (pd->af) {
4969 #if INET
4970 case AF_INET:
4971 if (PF_ANEQ(saddr, &pd->naddr, pd->af)) {
4972 pf_change_a(&saddr->v4addr.s_addr,
4973 pd->ip_sum,
4974 pd->naddr.v4addr.s_addr, 0);
4975 }
4976 if (PF_ANEQ(daddr, &pd->ndaddr, pd->af)) {
4977 pf_change_a(&daddr->v4addr.s_addr,
4978 pd->ip_sum,
4979 pd->ndaddr.v4addr.s_addr, 0);
4980 }
4981 break;
4982 #endif /* INET */
4983 case AF_INET6:
4984 if (PF_ANEQ(saddr, &pd->naddr, pd->af)) {
4985 PF_ACPY(saddr, &pd->naddr, AF_INET6);
4986 }
4987 if (PF_ANEQ(daddr, &pd->ndaddr, pd->af)) {
4988 PF_ACPY(daddr, &pd->ndaddr, AF_INET6);
4989 }
4990 break;
4991 }
4992 ++rewrite;
4993 break;
4994 case IPPROTO_ESP:
4995 if (direction == PF_OUT) {
4996 bxport.spi = 0;
4997 }
4998
4999 switch (pd->af) {
5000 #if INET
5001 case AF_INET:
5002 if (PF_ANEQ(saddr, &pd->naddr, pd->af)) {
5003 pf_change_a(&saddr->v4addr.s_addr,
5004 pd->ip_sum, pd->naddr.v4addr.s_addr, 0);
5005 }
5006 if (PF_ANEQ(daddr, &pd->ndaddr, pd->af)) {
5007 pf_change_a(&daddr->v4addr.s_addr,
5008 pd->ip_sum,
5009 pd->ndaddr.v4addr.s_addr, 0);
5010 }
5011 break;
5012 #endif /* INET */
5013 case AF_INET6:
5014 if (PF_ANEQ(saddr, &pd->naddr, pd->af)) {
5015 PF_ACPY(saddr, &pd->naddr, AF_INET6);
5016 }
5017 if (PF_ANEQ(daddr, &pd->ndaddr, pd->af)) {
5018 PF_ACPY(daddr, &pd->ndaddr, AF_INET6);
5019 }
5020 break;
5021 }
5022 break;
5023 default:
5024 switch (pd->af) {
5025 #if INET
5026 case AF_INET:
5027 if ((pd->naf != AF_INET) ||
5028 (PF_ANEQ(saddr, &pd->naddr, pd->af))) {
5029 pf_change_addr(saddr, pd->ip_sum,
5030 &pd->naddr, 0, af, pd->naf);
5031 }
5032
5033 if ((pd->naf != AF_INET) ||
5034 (PF_ANEQ(daddr, &pd->ndaddr, pd->af))) {
5035 pf_change_addr(daddr, pd->ip_sum,
5036 &pd->ndaddr, 0, af, pd->naf);
5037 }
5038 break;
5039 #endif /* INET */
5040 case AF_INET6:
5041 if (PF_ANEQ(saddr, &pd->naddr, pd->af)) {
5042 PF_ACPY(saddr, &pd->naddr, af);
5043 }
5044 if (PF_ANEQ(daddr, &pd->ndaddr, pd->af)) {
5045 PF_ACPY(daddr, &pd->ndaddr, af);
5046 }
5047 break;
5048 }
5049 break;
5050 }
5051
5052 if (nr->natpass) {
5053 r = NULL;
5054 }
5055 pd->nat_rule = nr;
5056 pd->af = pd->naf;
5057 } else {
5058 }
5059
5060 if (nr && nr->tag > 0) {
5061 tag = nr->tag;
5062 }
5063
5064 while (r != NULL) {
5065 r->evaluations++;
5066 if (pfi_kif_match(r->kif, kif) == r->ifnot) {
5067 r = r->skip[PF_SKIP_IFP].ptr;
5068 } else if (r->direction && r->direction != direction) {
5069 r = r->skip[PF_SKIP_DIR].ptr;
5070 } else if (r->af && r->af != pd->af) {
5071 r = r->skip[PF_SKIP_AF].ptr;
5072 } else if (r->proto && r->proto != pd->proto) {
5073 r = r->skip[PF_SKIP_PROTO].ptr;
5074 } else if (PF_MISMATCHAW(&r->src.addr, saddr, pd->af,
5075 r->src.neg, kif)) {
5076 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
5077 }
5078 /* tcp/udp only. port_op always 0 in other cases */
5079 else if (r->proto == pd->proto &&
5080 (r->proto == IPPROTO_TCP || r->proto == IPPROTO_UDP) &&
5081 r->src.xport.range.op &&
5082 !pf_match_port(r->src.xport.range.op,
5083 r->src.xport.range.port[0], r->src.xport.range.port[1],
5084 th->th_sport)) {
5085 r = r->skip[PF_SKIP_SRC_PORT].ptr;
5086 } else if (PF_MISMATCHAW(&r->dst.addr, daddr, pd->af,
5087 r->dst.neg, NULL)) {
5088 r = r->skip[PF_SKIP_DST_ADDR].ptr;
5089 }
5090 /* tcp/udp only. port_op always 0 in other cases */
5091 else if (r->proto == pd->proto &&
5092 (r->proto == IPPROTO_TCP || r->proto == IPPROTO_UDP) &&
5093 r->dst.xport.range.op &&
5094 !pf_match_port(r->dst.xport.range.op,
5095 r->dst.xport.range.port[0], r->dst.xport.range.port[1],
5096 th->th_dport)) {
5097 r = r->skip[PF_SKIP_DST_PORT].ptr;
5098 }
5099 /* icmp only. type always 0 in other cases */
5100 else if (r->type && r->type != icmptype + 1) {
5101 r = TAILQ_NEXT(r, entries);
5102 }
5103 /* icmp only. type always 0 in other cases */
5104 else if (r->code && r->code != icmpcode + 1) {
5105 r = TAILQ_NEXT(r, entries);
5106 } else if ((r->rule_flag & PFRULE_TOS) && r->tos &&
5107 !(r->tos & pd->tos)) {
5108 r = TAILQ_NEXT(r, entries);
5109 } else if ((r->rule_flag & PFRULE_DSCP) && r->tos &&
5110 !(r->tos & (pd->tos & DSCP_MASK))) {
5111 r = TAILQ_NEXT(r, entries);
5112 } else if ((r->rule_flag & PFRULE_SC) && r->tos &&
5113 ((r->tos & SCIDX_MASK) != pd->sc)) {
5114 r = TAILQ_NEXT(r, entries);
5115 } else if (r->rule_flag & PFRULE_FRAGMENT) {
5116 r = TAILQ_NEXT(r, entries);
5117 } else if (pd->proto == IPPROTO_TCP &&
5118 (r->flagset & th->th_flags) != r->flags) {
5119 r = TAILQ_NEXT(r, entries);
5120 }
5121 /* tcp/udp only. uid.op always 0 in other cases */
5122 else if (r->uid.op && (pd->lookup.done || ((void)(pd->lookup.done =
5123 pf_socket_lookup(direction, pd)), 1)) &&
5124 !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
5125 pd->lookup.uid)) {
5126 r = TAILQ_NEXT(r, entries);
5127 }
5128 /* tcp/udp only. gid.op always 0 in other cases */
5129 else if (r->gid.op && (pd->lookup.done || ((void)(pd->lookup.done =
5130 pf_socket_lookup(direction, pd)), 1)) &&
5131 !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
5132 pd->lookup.gid)) {
5133 r = TAILQ_NEXT(r, entries);
5134 } else if (r->prob && r->prob <= (RandomULong() % (UINT_MAX - 1) + 1)) {
5135 r = TAILQ_NEXT(r, entries);
5136 } else if (r->match_tag && !pf_match_tag(r, pd->pf_mtag, &tag)) {
5137 r = TAILQ_NEXT(r, entries);
5138 } else if (r->os_fingerprint != PF_OSFP_ANY &&
5139 (pd->proto != IPPROTO_TCP || !pf_osfp_match(
5140 pf_osfp_fingerprint(pd, pbuf, off, th),
5141 r->os_fingerprint))) {
5142 r = TAILQ_NEXT(r, entries);
5143 } else {
5144 if (r->tag) {
5145 tag = r->tag;
5146 }
5147 if (PF_RTABLEID_IS_VALID(r->rtableid)) {
5148 rtableid = r->rtableid;
5149 }
5150 if (r->anchor == NULL) {
5151 match = 1;
5152 *rm = r;
5153 *am = a;
5154 *rsm = ruleset;
5155 if ((*rm)->quick) {
5156 break;
5157 }
5158 r = TAILQ_NEXT(r, entries);
5159 } else {
5160 pf_step_into_anchor(&asd, &ruleset,
5161 PF_RULESET_FILTER, &r, &a, &match);
5162 }
5163 }
5164 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
5165 PF_RULESET_FILTER, &r, &a, &match)) {
5166 break;
5167 }
5168 }
5169 r = *rm;
5170 a = *am;
5171 ruleset = *rsm;
5172
5173 REASON_SET(&reason, PFRES_MATCH);
5174
5175 if (r->log || (nr != NULL && nr->log)) {
5176 if (rewrite > 0) {
5177 if (rewrite < off + hdrlen) {
5178 rewrite = off + hdrlen;
5179 }
5180
5181 if (pf_lazy_makewritable(pd, pbuf, rewrite) == NULL) {
5182 REASON_SET(&reason, PFRES_MEMORY);
5183 return PF_DROP;
5184 }
5185
5186 pbuf_copy_back(pbuf, off, hdrlen, pd->hdr.any);
5187 }
5188 PFLOG_PACKET(kif, h, pbuf, pd->af, direction, reason,
5189 r->log ? r : nr, a, ruleset, pd);
5190 }
5191
5192 if ((r->action == PF_DROP) &&
5193 ((r->rule_flag & PFRULE_RETURNRST) ||
5194 (r->rule_flag & PFRULE_RETURNICMP) ||
5195 (r->rule_flag & PFRULE_RETURN))) {
5196 /* undo NAT changes, if they have taken place */
5197 /* XXX For NAT64 we are not reverting the changes */
5198 if (nr != NULL && nr->action != PF_NAT64) {
5199 if (direction == PF_OUT) {
5200 pd->af = af;
5201 switch (pd->proto) {
5202 case IPPROTO_TCP:
5203 pf_change_ap(direction, pd->mp, saddr,
5204 &th->th_sport, pd->ip_sum,
5205 &th->th_sum, &pd->baddr,
5206 bxport.port, 0, af, pd->af, 1);
5207 sxport.port = th->th_sport;
5208 rewrite++;
5209 break;
5210 case IPPROTO_UDP:
5211 pf_change_ap(direction, pd->mp, saddr,
5212 &pd->hdr.udp->uh_sport, pd->ip_sum,
5213 &pd->hdr.udp->uh_sum, &pd->baddr,
5214 bxport.port, 1, af, pd->af, 1);
5215 sxport.port = pd->hdr.udp->uh_sport;
5216 rewrite++;
5217 break;
5218 case IPPROTO_ICMP:
5219 case IPPROTO_ICMPV6:
5220 /* nothing! */
5221 break;
5222 case IPPROTO_GRE:
5223 PF_ACPY(&pd->baddr, saddr, af);
5224 ++rewrite;
5225 switch (af) {
5226 #if INET
5227 case AF_INET:
5228 pf_change_a(&saddr->v4addr.s_addr,
5229 pd->ip_sum,
5230 pd->baddr.v4addr.s_addr, 0);
5231 break;
5232 #endif /* INET */
5233 case AF_INET6:
5234 PF_ACPY(saddr, &pd->baddr,
5235 AF_INET6);
5236 break;
5237 }
5238 break;
5239 case IPPROTO_ESP:
5240 PF_ACPY(&pd->baddr, saddr, af);
5241 switch (af) {
5242 #if INET
5243 case AF_INET:
5244 pf_change_a(&saddr->v4addr.s_addr,
5245 pd->ip_sum,
5246 pd->baddr.v4addr.s_addr, 0);
5247 break;
5248 #endif /* INET */
5249 case AF_INET6:
5250 PF_ACPY(saddr, &pd->baddr,
5251 AF_INET6);
5252 break;
5253 }
5254 break;
5255 default:
5256 switch (af) {
5257 case AF_INET:
5258 pf_change_a(&saddr->v4addr.s_addr,
5259 pd->ip_sum,
5260 pd->baddr.v4addr.s_addr, 0);
5261 break;
5262 case AF_INET6:
5263 PF_ACPY(saddr, &pd->baddr, af);
5264 break;
5265 }
5266 }
5267 } else {
5268 switch (pd->proto) {
5269 case IPPROTO_TCP:
5270 pf_change_ap(direction, pd->mp, daddr,
5271 &th->th_dport, pd->ip_sum,
5272 &th->th_sum, &pd->bdaddr,
5273 bdxport.port, 0, af, pd->af, 1);
5274 dxport.port = th->th_dport;
5275 rewrite++;
5276 break;
5277 case IPPROTO_UDP:
5278 pf_change_ap(direction, pd->mp, daddr,
5279 &pd->hdr.udp->uh_dport, pd->ip_sum,
5280 &pd->hdr.udp->uh_sum, &pd->bdaddr,
5281 bdxport.port, 1, af, pd->af, 1);
5282 dxport.port = pd->hdr.udp->uh_dport;
5283 rewrite++;
5284 break;
5285 case IPPROTO_ICMP:
5286 case IPPROTO_ICMPV6:
5287 /* nothing! */
5288 break;
5289 case IPPROTO_GRE:
5290 if (pd->proto_variant ==
5291 PF_GRE_PPTP_VARIANT) {
5292 grev1->call_id =
5293 bdxport.call_id;
5294 }
5295 ++rewrite;
5296 switch (af) {
5297 #if INET
5298 case AF_INET:
5299 pf_change_a(&daddr->v4addr.s_addr,
5300 pd->ip_sum,
5301 pd->bdaddr.v4addr.s_addr, 0);
5302 break;
5303 #endif /* INET */
5304 case AF_INET6:
5305 PF_ACPY(daddr, &pd->bdaddr,
5306 AF_INET6);
5307 break;
5308 }
5309 break;
5310 case IPPROTO_ESP:
5311 switch (af) {
5312 #if INET
5313 case AF_INET:
5314 pf_change_a(&daddr->v4addr.s_addr,
5315 pd->ip_sum,
5316 pd->bdaddr.v4addr.s_addr, 0);
5317 break;
5318 #endif /* INET */
5319 case AF_INET6:
5320 PF_ACPY(daddr, &pd->bdaddr,
5321 AF_INET6);
5322 break;
5323 }
5324 break;
5325 default:
5326 switch (af) {
5327 case AF_INET:
5328 pf_change_a(&daddr->v4addr.s_addr,
5329 pd->ip_sum,
5330 pd->bdaddr.v4addr.s_addr, 0);
5331 break;
5332 case AF_INET6:
5333 PF_ACPY(daddr, &pd->bdaddr, af);
5334 break;
5335 }
5336 }
5337 }
5338 }
5339 if (pd->proto == IPPROTO_TCP &&
5340 ((r->rule_flag & PFRULE_RETURNRST) ||
5341 (r->rule_flag & PFRULE_RETURN)) &&
5342 !(th->th_flags & TH_RST)) {
5343 u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
5344 int len = 0;
5345 struct ip *h4;
5346 struct ip6_hdr *h6;
5347
5348 switch (pd->af) {
5349 case AF_INET:
5350 h4 = pbuf->pb_data;
5351 len = ntohs(h4->ip_len) - off;
5352 break;
5353 case AF_INET6:
5354 h6 = pbuf->pb_data;
5355 len = ntohs(h6->ip6_plen) -
5356 (off - sizeof(*h6));
5357 break;
5358 }
5359
5360 if (pf_check_proto_cksum(pbuf, off, len, IPPROTO_TCP,
5361 pd->af)) {
5362 REASON_SET(&reason, PFRES_PROTCKSUM);
5363 } else {
5364 if (th->th_flags & TH_SYN) {
5365 ack++;
5366 }
5367 if (th->th_flags & TH_FIN) {
5368 ack++;
5369 }
5370 pf_send_tcp(r, pd->af, pd->dst,
5371 pd->src, th->th_dport, th->th_sport,
5372 ntohl(th->th_ack), ack, TH_RST | TH_ACK, 0, 0,
5373 r->return_ttl, 1, 0, pd->eh, kif->pfik_ifp);
5374 }
5375 } else if (pd->proto != IPPROTO_ICMP && pd->af == AF_INET &&
5376 pd->proto != IPPROTO_ESP && pd->proto != IPPROTO_AH &&
5377 r->return_icmp) {
5378 pf_send_icmp(pbuf, r->return_icmp >> 8,
5379 r->return_icmp & 255, pd->af, r);
5380 } else if (pd->proto != IPPROTO_ICMPV6 && af == AF_INET6 &&
5381 pd->proto != IPPROTO_ESP && pd->proto != IPPROTO_AH &&
5382 r->return_icmp6) {
5383 pf_send_icmp(pbuf, r->return_icmp6 >> 8,
5384 r->return_icmp6 & 255, pd->af, r);
5385 }
5386 }
5387
5388 if (r->action == PF_DROP) {
5389 return PF_DROP;
5390 }
5391
5392 /* prepare state key, for flowhash and/or the state (if created) */
5393 bzero(&psk, sizeof(psk));
5394 psk.proto = pd->proto;
5395 psk.direction = direction;
5396 if (pd->proto == IPPROTO_UDP) {
5397 if (ntohs(pd->hdr.udp->uh_sport) == PF_IKE_PORT &&
5398 ntohs(pd->hdr.udp->uh_dport) == PF_IKE_PORT) {
5399 psk.proto_variant = PF_EXTFILTER_APD;
5400 } else {
5401 psk.proto_variant = nr ? nr->extfilter : r->extfilter;
5402 if (psk.proto_variant < PF_EXTFILTER_APD) {
5403 psk.proto_variant = PF_EXTFILTER_APD;
5404 }
5405 }
5406 } else if (pd->proto == IPPROTO_GRE) {
5407 psk.proto_variant = pd->proto_variant;
5408 }
5409 if (direction == PF_OUT) {
5410 psk.af_gwy = af;
5411 PF_ACPY(&psk.gwy.addr, saddr, af);
5412 PF_ACPY(&psk.ext_gwy.addr, daddr, af);
5413 switch (pd->proto) {
5414 case IPPROTO_ESP:
5415 psk.gwy.xport.spi = 0;
5416 psk.ext_gwy.xport.spi = pd->hdr.esp->spi;
5417 break;
5418 case IPPROTO_ICMP:
5419 case IPPROTO_ICMPV6:
5420 /*
5421 * NAT64 requires protocol translation between ICMPv4
5422 * and ICMPv6. TCP and UDP do not require protocol
5423 * translation. To avoid adding complexity just to
5424 * handle ICMP(v4addr/v6addr), we always lookup for
5425 * proto = IPPROTO_ICMP on both LAN and WAN side
5426 */
5427 psk.proto = IPPROTO_ICMP;
5428 psk.gwy.xport.port = nxport.port;
5429 psk.ext_gwy.xport.spi = 0;
5430 break;
5431 default:
5432 psk.gwy.xport = sxport;
5433 psk.ext_gwy.xport = dxport;
5434 break;
5435 }
5436 psk.af_lan = af;
5437 if (nr != NULL) {
5438 PF_ACPY(&psk.lan.addr, &pd->baddr, af);
5439 psk.lan.xport = bxport;
5440 PF_ACPY(&psk.ext_lan.addr, &pd->bdaddr, af);
5441 psk.ext_lan.xport = bdxport;
5442 } else {
5443 PF_ACPY(&psk.lan.addr, &psk.gwy.addr, af);
5444 psk.lan.xport = psk.gwy.xport;
5445 PF_ACPY(&psk.ext_lan.addr, &psk.ext_gwy.addr, af);
5446 psk.ext_lan.xport = psk.ext_gwy.xport;
5447 }
5448 } else {
5449 psk.af_lan = af;
5450 if (nr && nr->action == PF_NAT64) {
5451 PF_ACPY(&psk.lan.addr, &pd->baddr, af);
5452 PF_ACPY(&psk.ext_lan.addr, &pd->bdaddr, af);
5453 } else {
5454 PF_ACPY(&psk.lan.addr, daddr, af);
5455 PF_ACPY(&psk.ext_lan.addr, saddr, af);
5456 }
5457 switch (pd->proto) {
5458 case IPPROTO_ICMP:
5459 case IPPROTO_ICMPV6:
5460 /*
5461 * NAT64 requires protocol translation between ICMPv4
5462 * and ICMPv6. TCP and UDP do not require protocol
5463 * translation. To avoid adding complexity just to
5464 * handle ICMP(v4addr/v6addr), we always lookup for
5465 * proto = IPPROTO_ICMP on both LAN and WAN side
5466 */
5467 psk.proto = IPPROTO_ICMP;
5468 if (nr && nr->action == PF_NAT64) {
5469 psk.lan.xport = bxport;
5470 psk.ext_lan.xport = bxport;
5471 } else {
5472 psk.lan.xport = nxport;
5473 psk.ext_lan.xport.spi = 0;
5474 }
5475 break;
5476 case IPPROTO_ESP:
5477 psk.ext_lan.xport.spi = 0;
5478 psk.lan.xport.spi = pd->hdr.esp->spi;
5479 break;
5480 default:
5481 if (nr != NULL) {
5482 if (nr->action == PF_NAT64) {
5483 psk.lan.xport = bxport;
5484 psk.ext_lan.xport = bdxport;
5485 } else {
5486 psk.lan.xport = dxport;
5487 psk.ext_lan.xport = sxport;
5488 }
5489 } else {
5490 psk.lan.xport = dxport;
5491 psk.ext_lan.xport = sxport;
5492 }
5493 break;
5494 }
5495 psk.af_gwy = pd->naf;
5496 if (nr != NULL) {
5497 if (nr->action == PF_NAT64) {
5498 PF_ACPY(&psk.gwy.addr, &pd->naddr, pd->naf);
5499 PF_ACPY(&psk.ext_gwy.addr, &pd->ndaddr,
5500 pd->naf);
5501 if ((pd->proto == IPPROTO_ICMPV6) ||
5502 (pd->proto == IPPROTO_ICMP)) {
5503 psk.gwy.xport = nxport;
5504 psk.ext_gwy.xport = nxport;
5505 } else {
5506 psk.gwy.xport = sxport;
5507 psk.ext_gwy.xport = dxport;
5508 }
5509 } else {
5510 PF_ACPY(&psk.gwy.addr, &pd->bdaddr, af);
5511 psk.gwy.xport = bdxport;
5512 PF_ACPY(&psk.ext_gwy.addr, saddr, af);
5513 psk.ext_gwy.xport = sxport;
5514 }
5515 } else {
5516 PF_ACPY(&psk.gwy.addr, &psk.lan.addr, af);
5517 psk.gwy.xport = psk.lan.xport;
5518 PF_ACPY(&psk.ext_gwy.addr, &psk.ext_lan.addr, af);
5519 psk.ext_gwy.xport = psk.ext_lan.xport;
5520 }
5521 }
5522 if (pd->pktflags & PKTF_FLOW_ID) {
5523 /* flow hash was already computed outside of PF */
5524 psk.flowsrc = pd->flowsrc;
5525 psk.flowhash = pd->flowhash;
5526 } else {
5527 /* compute flow hash and store it in state key */
5528 psk.flowsrc = FLOWSRC_PF;
5529 psk.flowhash = pf_calc_state_key_flowhash(&psk);
5530 pd->flowsrc = psk.flowsrc;
5531 pd->flowhash = psk.flowhash;
5532 pd->pktflags |= PKTF_FLOW_ID;
5533 pd->pktflags &= ~PKTF_FLOW_ADV;
5534 }
5535
5536 if (pf_tag_packet(pbuf, pd->pf_mtag, tag, rtableid, pd)) {
5537 REASON_SET(&reason, PFRES_MEMORY);
5538 return PF_DROP;
5539 }
5540
5541 if (!state_icmp && (r->keep_state || nr != NULL ||
5542 (pd->flags & PFDESC_TCP_NORM))) {
5543 /* create new state */
5544 struct pf_state *s = NULL;
5545 struct pf_state_key *sk = NULL;
5546 struct pf_src_node *sn = NULL;
5547 struct pf_ike_hdr ike;
5548
5549 if (pd->proto == IPPROTO_UDP) {
5550 size_t plen = pbuf->pb_packet_len - off - sizeof(*uh);
5551
5552 if (ntohs(uh->uh_sport) == PF_IKE_PORT &&
5553 ntohs(uh->uh_dport) == PF_IKE_PORT &&
5554 plen >= PF_IKE_PACKET_MINSIZE) {
5555 if (plen > PF_IKE_PACKET_MINSIZE) {
5556 plen = PF_IKE_PACKET_MINSIZE;
5557 }
5558 pbuf_copy_data(pbuf, off + sizeof(*uh), plen,
5559 &ike);
5560 }
5561 }
5562
5563 if (nr != NULL && pd->proto == IPPROTO_ESP &&
5564 direction == PF_OUT) {
5565 struct pf_state_key_cmp sk0;
5566 struct pf_state *s0;
5567
5568 /*
5569 * <jhw@apple.com>
5570 * This squelches state creation if the external
5571 * address matches an existing incomplete state with a
5572 * different internal address. Only one 'blocking'
5573 * partial state is allowed for each external address.
5574 */
5575 memset(&sk0, 0, sizeof(sk0));
5576 sk0.af_gwy = pd->af;
5577 sk0.proto = IPPROTO_ESP;
5578 PF_ACPY(&sk0.gwy.addr, saddr, sk0.af_gwy);
5579 PF_ACPY(&sk0.ext_gwy.addr, daddr, sk0.af_gwy);
5580 s0 = pf_find_state(kif, &sk0, PF_IN);
5581
5582 if (s0 && PF_ANEQ(&s0->state_key->lan.addr,
5583 pd->src, pd->af)) {
5584 nsn = 0;
5585 goto cleanup;
5586 }
5587 }
5588
5589 /* check maximums */
5590 if (r->max_states && (r->states >= r->max_states)) {
5591 pf_status.lcounters[LCNT_STATES]++;
5592 REASON_SET(&reason, PFRES_MAXSTATES);
5593 goto cleanup;
5594 }
5595 /* src node for filter rule */
5596 if ((r->rule_flag & PFRULE_SRCTRACK ||
5597 r->rpool.opts & PF_POOL_STICKYADDR) &&
5598 pf_insert_src_node(&sn, r, saddr, af) != 0) {
5599 REASON_SET(&reason, PFRES_SRCLIMIT);
5600 goto cleanup;
5601 }
5602 /* src node for translation rule */
5603 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
5604 ((direction == PF_OUT &&
5605 nr->action != PF_RDR &&
5606 pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
5607 (pf_insert_src_node(&nsn, nr, saddr, af) != 0))) {
5608 REASON_SET(&reason, PFRES_SRCLIMIT);
5609 goto cleanup;
5610 }
5611 s = pool_get(&pf_state_pl, PR_WAITOK);
5612 if (s == NULL) {
5613 REASON_SET(&reason, PFRES_MEMORY);
5614 cleanup:
5615 if (sn != NULL && sn->states == 0 && sn->expire == 0) {
5616 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
5617 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
5618 pf_status.src_nodes--;
5619 pool_put(&pf_src_tree_pl, sn);
5620 }
5621 if (nsn != sn && nsn != NULL && nsn->states == 0 &&
5622 nsn->expire == 0) {
5623 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
5624 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
5625 pf_status.src_nodes--;
5626 pool_put(&pf_src_tree_pl, nsn);
5627 }
5628 if (sk != NULL) {
5629 if (sk->app_state) {
5630 pool_put(&pf_app_state_pl,
5631 sk->app_state);
5632 }
5633 pool_put(&pf_state_key_pl, sk);
5634 }
5635 return PF_DROP;
5636 }
5637 bzero(s, sizeof(*s));
5638 TAILQ_INIT(&s->unlink_hooks);
5639 s->rule.ptr = r;
5640 s->nat_rule.ptr = nr;
5641 s->anchor.ptr = a;
5642 STATE_INC_COUNTERS(s);
5643 s->allow_opts = r->allow_opts;
5644 s->log = r->log & PF_LOG_ALL;
5645 if (nr != NULL) {
5646 s->log |= nr->log & PF_LOG_ALL;
5647 }
5648 switch (pd->proto) {
5649 case IPPROTO_TCP:
5650 s->src.seqlo = ntohl(th->th_seq);
5651 s->src.seqhi = s->src.seqlo + pd->p_len + 1;
5652 if ((th->th_flags & (TH_SYN | TH_ACK)) ==
5653 TH_SYN && r->keep_state == PF_STATE_MODULATE) {
5654 /* Generate sequence number modulator */
5655 if ((s->src.seqdiff = pf_tcp_iss(pd) -
5656 s->src.seqlo) == 0) {
5657 s->src.seqdiff = 1;
5658 }
5659 pf_change_a(&th->th_seq, &th->th_sum,
5660 htonl(s->src.seqlo + s->src.seqdiff), 0);
5661 rewrite = off + sizeof(*th);
5662 } else {
5663 s->src.seqdiff = 0;
5664 }
5665 if (th->th_flags & TH_SYN) {
5666 s->src.seqhi++;
5667 s->src.wscale = pf_get_wscale(pbuf, off,
5668 th->th_off, af);
5669 }
5670 s->src.max_win = MAX(ntohs(th->th_win), 1);
5671 if (s->src.wscale & PF_WSCALE_MASK) {
5672 /* Remove scale factor from initial window */
5673 int win = s->src.max_win;
5674 win += 1 << (s->src.wscale & PF_WSCALE_MASK);
5675 s->src.max_win = (win - 1) >>
5676 (s->src.wscale & PF_WSCALE_MASK);
5677 }
5678 if (th->th_flags & TH_FIN) {
5679 s->src.seqhi++;
5680 }
5681 s->dst.seqhi = 1;
5682 s->dst.max_win = 1;
5683 s->src.state = TCPS_SYN_SENT;
5684 s->dst.state = TCPS_CLOSED;
5685 s->timeout = PFTM_TCP_FIRST_PACKET;
5686 break;
5687 case IPPROTO_UDP:
5688 s->src.state = PFUDPS_SINGLE;
5689 s->dst.state = PFUDPS_NO_TRAFFIC;
5690 s->timeout = PFTM_UDP_FIRST_PACKET;
5691 break;
5692 case IPPROTO_ICMP:
5693 case IPPROTO_ICMPV6:
5694 s->timeout = PFTM_ICMP_FIRST_PACKET;
5695 break;
5696 case IPPROTO_GRE:
5697 s->src.state = PFGRE1S_INITIATING;
5698 s->dst.state = PFGRE1S_NO_TRAFFIC;
5699 s->timeout = PFTM_GREv1_INITIATING;
5700 break;
5701 case IPPROTO_ESP:
5702 s->src.state = PFESPS_INITIATING;
5703 s->dst.state = PFESPS_NO_TRAFFIC;
5704 s->timeout = PFTM_ESP_FIRST_PACKET;
5705 break;
5706 default:
5707 s->src.state = PFOTHERS_SINGLE;
5708 s->dst.state = PFOTHERS_NO_TRAFFIC;
5709 s->timeout = PFTM_OTHER_FIRST_PACKET;
5710 }
5711
5712 s->creation = pf_time_second();
5713 s->expire = pf_time_second();
5714
5715 if (sn != NULL) {
5716 s->src_node = sn;
5717 s->src_node->states++;
5718 VERIFY(s->src_node->states != 0);
5719 }
5720 if (nsn != NULL) {
5721 PF_ACPY(&nsn->raddr, &pd->naddr, af);
5722 s->nat_src_node = nsn;
5723 s->nat_src_node->states++;
5724 VERIFY(s->nat_src_node->states != 0);
5725 }
5726 if (pd->proto == IPPROTO_TCP) {
5727 if ((pd->flags & PFDESC_TCP_NORM) &&
5728 pf_normalize_tcp_init(pbuf, off, pd, th, &s->src,
5729 &s->dst)) {
5730 REASON_SET(&reason, PFRES_MEMORY);
5731 pf_src_tree_remove_state(s);
5732 STATE_DEC_COUNTERS(s);
5733 pool_put(&pf_state_pl, s);
5734 return PF_DROP;
5735 }
5736 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
5737 pf_normalize_tcp_stateful(pbuf, off, pd, &reason,
5738 th, s, &s->src, &s->dst, &rewrite)) {
5739 /* This really shouldn't happen!!! */
5740 DPFPRINTF(PF_DEBUG_URGENT,
5741 ("pf_normalize_tcp_stateful failed on "
5742 "first pkt"));
5743 pf_normalize_tcp_cleanup(s);
5744 pf_src_tree_remove_state(s);
5745 STATE_DEC_COUNTERS(s);
5746 pool_put(&pf_state_pl, s);
5747 return PF_DROP;
5748 }
5749 }
5750
5751 /* allocate state key and import values from psk */
5752 if ((sk = pf_alloc_state_key(s, &psk)) == NULL) {
5753 REASON_SET(&reason, PFRES_MEMORY);
5754 /*
5755 * XXXSCW: This will leak the freshly-allocated
5756 * state structure 's'. Although it should
5757 * eventually be aged-out and removed.
5758 */
5759 goto cleanup;
5760 }
5761
5762 pf_set_rt_ifp(s, saddr, af); /* needs s->state_key set */
5763
5764 pbuf = pd->mp; // XXXSCW: Why?
5765
5766 if (sk->app_state == 0) {
5767 switch (pd->proto) {
5768 case IPPROTO_TCP: {
5769 u_int16_t dport = (direction == PF_OUT) ?
5770 sk->ext_gwy.xport.port : sk->gwy.xport.port;
5771
5772 if (nr != NULL &&
5773 ntohs(dport) == PF_PPTP_PORT) {
5774 struct pf_app_state *as;
5775
5776 as = pool_get(&pf_app_state_pl,
5777 PR_WAITOK);
5778 if (!as) {
5779 REASON_SET(&reason,
5780 PFRES_MEMORY);
5781 goto cleanup;
5782 }
5783
5784 bzero(as, sizeof(*as));
5785 as->handler = pf_pptp_handler;
5786 as->compare_lan_ext = 0;
5787 as->compare_ext_gwy = 0;
5788 as->u.pptp.grev1_state = 0;
5789 sk->app_state = as;
5790 (void) hook_establish(&s->unlink_hooks,
5791 0, (hook_fn_t) pf_pptp_unlink, s);
5792 }
5793 break;
5794 }
5795
5796 case IPPROTO_UDP: {
5797 if (nr != NULL &&
5798 ntohs(uh->uh_sport) == PF_IKE_PORT &&
5799 ntohs(uh->uh_dport) == PF_IKE_PORT) {
5800 struct pf_app_state *as;
5801
5802 as = pool_get(&pf_app_state_pl,
5803 PR_WAITOK);
5804 if (!as) {
5805 REASON_SET(&reason,
5806 PFRES_MEMORY);
5807 goto cleanup;
5808 }
5809
5810 bzero(as, sizeof(*as));
5811 as->compare_lan_ext = pf_ike_compare;
5812 as->compare_ext_gwy = pf_ike_compare;
5813 as->u.ike.cookie = ike.initiator_cookie;
5814 sk->app_state = as;
5815 }
5816 break;
5817 }
5818
5819 default:
5820 break;
5821 }
5822 }
5823
5824 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
5825 if (pd->proto == IPPROTO_TCP) {
5826 pf_normalize_tcp_cleanup(s);
5827 }
5828 REASON_SET(&reason, PFRES_STATEINS);
5829 pf_src_tree_remove_state(s);
5830 STATE_DEC_COUNTERS(s);
5831 pool_put(&pf_state_pl, s);
5832 return PF_DROP;
5833 } else {
5834 *sm = s;
5835 }
5836 if (tag > 0) {
5837 pf_tag_ref(tag);
5838 s->tag = tag;
5839 }
5840 if (pd->proto == IPPROTO_TCP &&
5841 (th->th_flags & (TH_SYN | TH_ACK)) == TH_SYN &&
5842 r->keep_state == PF_STATE_SYNPROXY) {
5843 int ua = (sk->af_lan == sk->af_gwy) ? 1 : 0;
5844 s->src.state = PF_TCPS_PROXY_SRC;
5845 if (nr != NULL) {
5846 if (direction == PF_OUT) {
5847 pf_change_ap(direction, pd->mp, saddr,
5848 &th->th_sport, pd->ip_sum,
5849 &th->th_sum, &pd->baddr,
5850 bxport.port, 0, af, pd->af, ua);
5851 sxport.port = th->th_sport;
5852 } else {
5853 pf_change_ap(direction, pd->mp, daddr,
5854 &th->th_dport, pd->ip_sum,
5855 &th->th_sum, &pd->baddr,
5856 bxport.port, 0, af, pd->af, ua);
5857 sxport.port = th->th_dport;
5858 }
5859 }
5860 s->src.seqhi = htonl(random());
5861 /* Find mss option */
5862 mss = pf_get_mss(pbuf, off, th->th_off, af);
5863 mss = pf_calc_mss(saddr, af, mss);
5864 mss = pf_calc_mss(daddr, af, mss);
5865 s->src.mss = mss;
5866 pf_send_tcp(r, af, daddr, saddr, th->th_dport,
5867 th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1,
5868 TH_SYN | TH_ACK, 0, s->src.mss, 0, 1, 0, NULL, NULL);
5869 REASON_SET(&reason, PFRES_SYNPROXY);
5870 return PF_SYNPROXY_DROP;
5871 }
5872
5873 if (sk->app_state && sk->app_state->handler) {
5874 int offx = off;
5875
5876 switch (pd->proto) {
5877 case IPPROTO_TCP:
5878 offx += th->th_off << 2;
5879 break;
5880 case IPPROTO_UDP:
5881 offx += pd->hdr.udp->uh_ulen << 2;
5882 break;
5883 default:
5884 /* ALG handlers only apply to TCP and UDP rules */
5885 break;
5886 }
5887
5888 if (offx > off) {
5889 sk->app_state->handler(s, direction, offx,
5890 pd, kif);
5891 if (pd->lmw < 0) {
5892 REASON_SET(&reason, PFRES_MEMORY);
5893 return PF_DROP;
5894 }
5895 pbuf = pd->mp; // XXXSCW: Why?
5896 }
5897 }
5898 }
5899
5900 /* copy back packet headers if we performed NAT operations */
5901 if (rewrite) {
5902 if (rewrite < off + hdrlen) {
5903 rewrite = off + hdrlen;
5904 }
5905
5906 if (pf_lazy_makewritable(pd, pd->mp, rewrite) == NULL) {
5907 REASON_SET(&reason, PFRES_MEMORY);
5908 return PF_DROP;
5909 }
5910
5911 pbuf_copy_back(pbuf, off, hdrlen, pd->hdr.any);
5912 if (af == AF_INET6 && pd->naf == AF_INET) {
5913 return pf_nat64_ipv6(pbuf, off, pd);
5914 } else if (af == AF_INET && pd->naf == AF_INET6) {
5915 return pf_nat64_ipv4(pbuf, off, pd);
5916 }
5917 }
5918
5919 return PF_PASS;
5920 }
5921
5922 boolean_t is_nlc_enabled_glb = FALSE;
5923
5924 static inline boolean_t
5925 pf_is_dummynet_enabled(void)
5926 {
5927 #if DUMMYNET
5928 if (__probable(!PF_IS_ENABLED)) {
5929 return FALSE;
5930 }
5931
5932 if (__probable(!DUMMYNET_LOADED)) {
5933 return FALSE;
5934 }
5935
5936 if (__probable(TAILQ_EMPTY(pf_main_ruleset.
5937 rules[PF_RULESET_DUMMYNET].active.ptr))) {
5938 return FALSE;
5939 }
5940
5941 return TRUE;
5942 #else
5943 return FALSE;
5944 #endif /* DUMMYNET */
5945 }
5946
5947 #if DUMMYNET
5948 /*
5949 * When pf_test_dummynet() returns PF_PASS, the rule matching parameter "rm"
5950 * remains unchanged, meaning the packet did not match a dummynet rule.
5951 * when the packet does match a dummynet rule, pf_test_dummynet() returns
5952 * PF_PASS and zero out the mbuf rule as the packet is effectively siphoned
5953 * out by dummynet.
5954 */
5955 static int
5956 pf_test_dummynet(struct pf_rule **rm, int direction, struct pfi_kif *kif,
5957 pbuf_t **pbuf0, struct pf_pdesc *pd, struct ip_fw_args *fwa)
5958 {
5959 pbuf_t *pbuf = *pbuf0;
5960 struct pf_rule *am = NULL;
5961 struct pf_ruleset *rsm = NULL;
5962 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
5963 sa_family_t af = pd->af;
5964 struct pf_rule *r, *a = NULL;
5965 struct pf_ruleset *ruleset = NULL;
5966 struct tcphdr *th = pd->hdr.tcp;
5967 u_short reason;
5968 int hdrlen = 0;
5969 int tag = -1;
5970 unsigned int rtableid = IFSCOPE_NONE;
5971 int asd = 0;
5972 int match = 0;
5973 u_int8_t icmptype = 0, icmpcode = 0;
5974 struct ip_fw_args dnflow;
5975 struct pf_rule *prev_matching_rule = fwa ? fwa->fwa_pf_rule : NULL;
5976 int found_prev_rule = (prev_matching_rule) ? 0 : 1;
5977
5978 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
5979
5980 if (!pf_is_dummynet_enabled()) {
5981 return PF_PASS;
5982 }
5983
5984 bzero(&dnflow, sizeof(dnflow));
5985
5986 hdrlen = 0;
5987
5988 /* Fragments don't gave protocol headers */
5989 if (!(pd->flags & PFDESC_IP_FRAG)) {
5990 switch (pd->proto) {
5991 case IPPROTO_TCP:
5992 dnflow.fwa_id.flags = pd->hdr.tcp->th_flags;
5993 dnflow.fwa_id.dst_port = ntohs(pd->hdr.tcp->th_dport);
5994 dnflow.fwa_id.src_port = ntohs(pd->hdr.tcp->th_sport);
5995 hdrlen = sizeof(*th);
5996 break;
5997 case IPPROTO_UDP:
5998 dnflow.fwa_id.dst_port = ntohs(pd->hdr.udp->uh_dport);
5999 dnflow.fwa_id.src_port = ntohs(pd->hdr.udp->uh_sport);
6000 hdrlen = sizeof(*pd->hdr.udp);
6001 break;
6002 #if INET
6003 case IPPROTO_ICMP:
6004 if (af != AF_INET) {
6005 break;
6006 }
6007 hdrlen = ICMP_MINLEN;
6008 icmptype = pd->hdr.icmp->icmp_type;
6009 icmpcode = pd->hdr.icmp->icmp_code;
6010 break;
6011 #endif /* INET */
6012 case IPPROTO_ICMPV6:
6013 if (af != AF_INET6) {
6014 break;
6015 }
6016 hdrlen = sizeof(*pd->hdr.icmp6);
6017 icmptype = pd->hdr.icmp6->icmp6_type;
6018 icmpcode = pd->hdr.icmp6->icmp6_code;
6019 break;
6020 case IPPROTO_GRE:
6021 if (pd->proto_variant == PF_GRE_PPTP_VARIANT) {
6022 hdrlen = sizeof(*pd->hdr.grev1);
6023 }
6024 break;
6025 case IPPROTO_ESP:
6026 hdrlen = sizeof(*pd->hdr.esp);
6027 break;
6028 }
6029 }
6030
6031 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_DUMMYNET].active.ptr);
6032
6033 while (r != NULL) {
6034 r->evaluations++;
6035 if (pfi_kif_match(r->kif, kif) == r->ifnot) {
6036 r = r->skip[PF_SKIP_IFP].ptr;
6037 } else if (r->direction && r->direction != direction) {
6038 r = r->skip[PF_SKIP_DIR].ptr;
6039 } else if (r->af && r->af != af) {
6040 r = r->skip[PF_SKIP_AF].ptr;
6041 } else if (r->proto && r->proto != pd->proto) {
6042 r = r->skip[PF_SKIP_PROTO].ptr;
6043 } else if (PF_MISMATCHAW(&r->src.addr, saddr, af,
6044 r->src.neg, kif)) {
6045 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
6046 }
6047 /* tcp/udp only. port_op always 0 in other cases */
6048 else if (r->proto == pd->proto &&
6049 (r->proto == IPPROTO_TCP || r->proto == IPPROTO_UDP) &&
6050 ((pd->flags & PFDESC_IP_FRAG) ||
6051 ((r->src.xport.range.op &&
6052 !pf_match_port(r->src.xport.range.op,
6053 r->src.xport.range.port[0], r->src.xport.range.port[1],
6054 th->th_sport))))) {
6055 r = r->skip[PF_SKIP_SRC_PORT].ptr;
6056 } else if (PF_MISMATCHAW(&r->dst.addr, daddr, af,
6057 r->dst.neg, NULL)) {
6058 r = r->skip[PF_SKIP_DST_ADDR].ptr;
6059 }
6060 /* tcp/udp only. port_op always 0 in other cases */
6061 else if (r->proto == pd->proto &&
6062 (r->proto == IPPROTO_TCP || r->proto == IPPROTO_UDP) &&
6063 r->dst.xport.range.op &&
6064 ((pd->flags & PFDESC_IP_FRAG) ||
6065 !pf_match_port(r->dst.xport.range.op,
6066 r->dst.xport.range.port[0], r->dst.xport.range.port[1],
6067 th->th_dport))) {
6068 r = r->skip[PF_SKIP_DST_PORT].ptr;
6069 }
6070 /* icmp only. type always 0 in other cases */
6071 else if (r->type &&
6072 ((pd->flags & PFDESC_IP_FRAG) ||
6073 r->type != icmptype + 1)) {
6074 r = TAILQ_NEXT(r, entries);
6075 }
6076 /* icmp only. type always 0 in other cases */
6077 else if (r->code &&
6078 ((pd->flags & PFDESC_IP_FRAG) ||
6079 r->code != icmpcode + 1)) {
6080 r = TAILQ_NEXT(r, entries);
6081 } else if (r->tos && !(r->tos == pd->tos)) {
6082 r = TAILQ_NEXT(r, entries);
6083 } else if (r->rule_flag & PFRULE_FRAGMENT) {
6084 r = TAILQ_NEXT(r, entries);
6085 } else if (pd->proto == IPPROTO_TCP &&
6086 ((pd->flags & PFDESC_IP_FRAG) ||
6087 (r->flagset & th->th_flags) != r->flags)) {
6088 r = TAILQ_NEXT(r, entries);
6089 } else if (r->prob && r->prob <= (RandomULong() % (UINT_MAX - 1) + 1)) {
6090 r = TAILQ_NEXT(r, entries);
6091 } else if (r->match_tag && !pf_match_tag(r, pd->pf_mtag, &tag)) {
6092 r = TAILQ_NEXT(r, entries);
6093 } else {
6094 /*
6095 * Need to go past the previous dummynet matching rule
6096 */
6097 if (r->anchor == NULL) {
6098 if (found_prev_rule) {
6099 if (r->tag) {
6100 tag = r->tag;
6101 }
6102 if (PF_RTABLEID_IS_VALID(r->rtableid)) {
6103 rtableid = r->rtableid;
6104 }
6105 match = 1;
6106 *rm = r;
6107 am = a;
6108 rsm = ruleset;
6109 if ((*rm)->quick) {
6110 break;
6111 }
6112 } else if (r == prev_matching_rule) {
6113 found_prev_rule = 1;
6114 }
6115 r = TAILQ_NEXT(r, entries);
6116 } else {
6117 pf_step_into_anchor(&asd, &ruleset,
6118 PF_RULESET_DUMMYNET, &r, &a, &match);
6119 }
6120 }
6121 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
6122 PF_RULESET_DUMMYNET, &r, &a, &match)) {
6123 break;
6124 }
6125 }
6126 r = *rm;
6127 a = am;
6128 ruleset = rsm;
6129
6130 if (!match) {
6131 return PF_PASS;
6132 }
6133
6134 REASON_SET(&reason, PFRES_DUMMYNET);
6135
6136 if (r->log) {
6137 PFLOG_PACKET(kif, h, pbuf, af, direction, reason, r,
6138 a, ruleset, pd);
6139 }
6140
6141 if (r->action == PF_NODUMMYNET) {
6142 int dirndx = (direction == PF_OUT);
6143
6144 r->packets[dirndx]++;
6145 r->bytes[dirndx] += pd->tot_len;
6146
6147 return PF_PASS;
6148 }
6149 if (pf_tag_packet(pbuf, pd->pf_mtag, tag, rtableid, pd)) {
6150 REASON_SET(&reason, PFRES_MEMORY);
6151
6152 return PF_DROP;
6153 }
6154
6155 if (r->dnpipe && ip_dn_io_ptr != NULL) {
6156 struct mbuf *m;
6157 int dirndx = (direction == PF_OUT);
6158
6159 r->packets[dirndx]++;
6160 r->bytes[dirndx] += pd->tot_len;
6161
6162 dnflow.fwa_cookie = r->dnpipe;
6163 dnflow.fwa_pf_rule = r;
6164 dnflow.fwa_id.proto = pd->proto;
6165 dnflow.fwa_flags = r->dntype;
6166 switch (af) {
6167 case AF_INET:
6168 dnflow.fwa_id.addr_type = 4;
6169 dnflow.fwa_id.src_ip = ntohl(saddr->v4addr.s_addr);
6170 dnflow.fwa_id.dst_ip = ntohl(daddr->v4addr.s_addr);
6171 break;
6172 case AF_INET6:
6173 dnflow.fwa_id.addr_type = 6;
6174 dnflow.fwa_id.src_ip6 = saddr->v6addr;
6175 dnflow.fwa_id.dst_ip6 = saddr->v6addr;
6176 break;
6177 }
6178
6179 if (fwa != NULL) {
6180 dnflow.fwa_oif = fwa->fwa_oif;
6181 dnflow.fwa_oflags = fwa->fwa_oflags;
6182 /*
6183 * Note that fwa_ro, fwa_dst and fwa_ipoa are
6184 * actually in a union so the following does work
6185 * for both IPv4 and IPv6
6186 */
6187 dnflow.fwa_ro = fwa->fwa_ro;
6188 dnflow.fwa_dst = fwa->fwa_dst;
6189 dnflow.fwa_ipoa = fwa->fwa_ipoa;
6190 dnflow.fwa_ro6_pmtu = fwa->fwa_ro6_pmtu;
6191 dnflow.fwa_origifp = fwa->fwa_origifp;
6192 dnflow.fwa_mtu = fwa->fwa_mtu;
6193 dnflow.fwa_unfragpartlen = fwa->fwa_unfragpartlen;
6194 dnflow.fwa_exthdrs = fwa->fwa_exthdrs;
6195 }
6196
6197 if (af == AF_INET) {
6198 struct ip *iphdr = pbuf->pb_data;
6199 NTOHS(iphdr->ip_len);
6200 NTOHS(iphdr->ip_off);
6201 }
6202 /*
6203 * Don't need to unlock pf_lock as NET_THREAD_HELD_PF
6204 * allows for recursive behavior
6205 */
6206 m = pbuf_to_mbuf(pbuf, TRUE);
6207 if (m != NULL) {
6208 ip_dn_io_ptr(m,
6209 dnflow.fwa_cookie, (af == AF_INET) ?
6210 ((direction == PF_IN) ? DN_TO_IP_IN : DN_TO_IP_OUT) :
6211 ((direction == PF_IN) ? DN_TO_IP6_IN : DN_TO_IP6_OUT),
6212 &dnflow);
6213 }
6214
6215 /*
6216 * The packet is siphoned out by dummynet so return a NULL
6217 * pbuf so the caller can still return success.
6218 */
6219 *pbuf0 = NULL;
6220
6221 return PF_PASS;
6222 }
6223
6224 return PF_PASS;
6225 }
6226 #endif /* DUMMYNET */
6227
6228 static int
6229 pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif,
6230 pbuf_t *pbuf, void *h, struct pf_pdesc *pd, struct pf_rule **am,
6231 struct pf_ruleset **rsm)
6232 {
6233 #pragma unused(h)
6234 struct pf_rule *r, *a = NULL;
6235 struct pf_ruleset *ruleset = NULL;
6236 sa_family_t af = pd->af;
6237 u_short reason;
6238 int tag = -1;
6239 int asd = 0;
6240 int match = 0;
6241
6242 r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
6243 while (r != NULL) {
6244 r->evaluations++;
6245 if (pfi_kif_match(r->kif, kif) == r->ifnot) {
6246 r = r->skip[PF_SKIP_IFP].ptr;
6247 } else if (r->direction && r->direction != direction) {
6248 r = r->skip[PF_SKIP_DIR].ptr;
6249 } else if (r->af && r->af != af) {
6250 r = r->skip[PF_SKIP_AF].ptr;
6251 } else if (r->proto && r->proto != pd->proto) {
6252 r = r->skip[PF_SKIP_PROTO].ptr;
6253 } else if (PF_MISMATCHAW(&r->src.addr, pd->src, af,
6254 r->src.neg, kif)) {
6255 r = r->skip[PF_SKIP_SRC_ADDR].ptr;
6256 } else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af,
6257 r->dst.neg, NULL)) {
6258 r = r->skip[PF_SKIP_DST_ADDR].ptr;
6259 } else if ((r->rule_flag & PFRULE_TOS) && r->tos &&
6260 !(r->tos & pd->tos)) {
6261 r = TAILQ_NEXT(r, entries);
6262 } else if ((r->rule_flag & PFRULE_DSCP) && r->tos &&
6263 !(r->tos & (pd->tos & DSCP_MASK))) {
6264 r = TAILQ_NEXT(r, entries);
6265 } else if ((r->rule_flag & PFRULE_SC) && r->tos &&
6266 ((r->tos & SCIDX_MASK) != pd->sc)) {
6267 r = TAILQ_NEXT(r, entries);
6268 } else if (r->os_fingerprint != PF_OSFP_ANY) {
6269 r = TAILQ_NEXT(r, entries);
6270 } else if (pd->proto == IPPROTO_UDP &&
6271 (r->src.xport.range.op || r->dst.xport.range.op)) {
6272 r = TAILQ_NEXT(r, entries);
6273 } else if (pd->proto == IPPROTO_TCP &&
6274 (r->src.xport.range.op || r->dst.xport.range.op ||
6275 r->flagset)) {
6276 r = TAILQ_NEXT(r, entries);
6277 } else if ((pd->proto == IPPROTO_ICMP ||
6278 pd->proto == IPPROTO_ICMPV6) &&
6279 (r->type || r->code)) {
6280 r = TAILQ_NEXT(r, entries);
6281 } else if (r->prob && r->prob <= (RandomULong() % (UINT_MAX - 1) + 1)) {
6282 r = TAILQ_NEXT(r, entries);
6283 } else if (r->match_tag && !pf_match_tag(r, pd->pf_mtag, &tag)) {
6284 r = TAILQ_NEXT(r, entries);
6285 } else {
6286 if (r->anchor == NULL) {
6287 match = 1;
6288 *rm = r;
6289 *am = a;
6290 *rsm = ruleset;
6291 if ((*rm)->quick) {
6292 break;
6293 }
6294 r = TAILQ_NEXT(r, entries);
6295 } else {
6296 pf_step_into_anchor(&asd, &ruleset,
6297 PF_RULESET_FILTER, &r, &a, &match);
6298 }
6299 }
6300 if (r == NULL && pf_step_out_of_anchor(&asd, &ruleset,
6301 PF_RULESET_FILTER, &r, &a, &match)) {
6302 break;
6303 }
6304 }
6305 r = *rm;
6306 a = *am;
6307 ruleset = *rsm;
6308
6309 REASON_SET(&reason, PFRES_MATCH);
6310
6311 if (r->log) {
6312 PFLOG_PACKET(kif, h, pbuf, af, direction, reason, r, a, ruleset,
6313 pd);
6314 }
6315
6316 if (r->action != PF_PASS) {
6317 return PF_DROP;
6318 }
6319
6320 if (pf_tag_packet(pbuf, pd->pf_mtag, tag, -1, NULL)) {
6321 REASON_SET(&reason, PFRES_MEMORY);
6322 return PF_DROP;
6323 }
6324
6325 return PF_PASS;
6326 }
6327
6328 static void
6329 pf_pptp_handler(struct pf_state *s, int direction, int off,
6330 struct pf_pdesc *pd, struct pfi_kif *kif)
6331 {
6332 #pragma unused(direction)
6333 struct tcphdr *th;
6334 struct pf_pptp_state *pptps;
6335 struct pf_pptp_ctrl_msg cm;
6336 size_t plen, tlen;
6337 struct pf_state *gs;
6338 u_int16_t ct;
6339 u_int16_t *pac_call_id;
6340 u_int16_t *pns_call_id;
6341 u_int16_t *spoof_call_id;
6342 u_int8_t *pac_state;
6343 u_int8_t *pns_state;
6344 enum { PF_PPTP_PASS, PF_PPTP_INSERT_GRE, PF_PPTP_REMOVE_GRE } op;
6345 pbuf_t *pbuf;
6346 struct pf_state_key *sk;
6347 struct pf_state_key *gsk;
6348 struct pf_app_state *gas;
6349
6350 sk = s->state_key;
6351 pptps = &sk->app_state->u.pptp;
6352 gs = pptps->grev1_state;
6353
6354 if (gs) {
6355 gs->expire = pf_time_second();
6356 }
6357
6358 pbuf = pd->mp;
6359 plen = min(sizeof(cm), pbuf->pb_packet_len - off);
6360 if (plen < PF_PPTP_CTRL_MSG_MINSIZE) {
6361 return;
6362 }
6363 tlen = plen - PF_PPTP_CTRL_MSG_MINSIZE;
6364 pbuf_copy_data(pbuf, off, plen, &cm);
6365
6366 if (ntohl(cm.hdr.magic) != PF_PPTP_MAGIC_NUMBER) {
6367 return;
6368 }
6369 if (ntohs(cm.hdr.type) != 1) {
6370 return;
6371 }
6372
6373 #define TYPE_LEN_CHECK(_type, _name) \
6374 case PF_PPTP_CTRL_TYPE_##_type: \
6375 if (tlen < sizeof(struct pf_pptp_ctrl_##_name)) \
6376 return; \
6377 break;
6378
6379 switch (cm.ctrl.type) {
6380 TYPE_LEN_CHECK(START_REQ, start_req);
6381 TYPE_LEN_CHECK(START_RPY, start_rpy);
6382 TYPE_LEN_CHECK(STOP_REQ, stop_req);
6383 TYPE_LEN_CHECK(STOP_RPY, stop_rpy);
6384 TYPE_LEN_CHECK(ECHO_REQ, echo_req);
6385 TYPE_LEN_CHECK(ECHO_RPY, echo_rpy);
6386 TYPE_LEN_CHECK(CALL_OUT_REQ, call_out_req);
6387 TYPE_LEN_CHECK(CALL_OUT_RPY, call_out_rpy);
6388 TYPE_LEN_CHECK(CALL_IN_1ST, call_in_1st);
6389 TYPE_LEN_CHECK(CALL_IN_2ND, call_in_2nd);
6390 TYPE_LEN_CHECK(CALL_IN_3RD, call_in_3rd);
6391 TYPE_LEN_CHECK(CALL_CLR, call_clr);
6392 TYPE_LEN_CHECK(CALL_DISC, call_disc);
6393 TYPE_LEN_CHECK(ERROR, error);
6394 TYPE_LEN_CHECK(SET_LINKINFO, set_linkinfo);
6395 default:
6396 return;
6397 }
6398 #undef TYPE_LEN_CHECK
6399
6400 if (!gs) {
6401 gs = pool_get(&pf_state_pl, PR_WAITOK);
6402 if (!gs) {
6403 return;
6404 }
6405
6406 memcpy(gs, s, sizeof(*gs));
6407
6408 memset(&gs->entry_id, 0, sizeof(gs->entry_id));
6409 memset(&gs->entry_list, 0, sizeof(gs->entry_list));
6410
6411 TAILQ_INIT(&gs->unlink_hooks);
6412 gs->rt_kif = NULL;
6413 gs->creation = 0;
6414 gs->pfsync_time = 0;
6415 gs->packets[0] = gs->packets[1] = 0;
6416 gs->bytes[0] = gs->bytes[1] = 0;
6417 gs->timeout = PFTM_UNLINKED;
6418 gs->id = gs->creatorid = 0;
6419 gs->src.state = gs->dst.state = PFGRE1S_NO_TRAFFIC;
6420 gs->src.scrub = gs->dst.scrub = 0;
6421
6422 gas = pool_get(&pf_app_state_pl, PR_NOWAIT);
6423 if (!gas) {
6424 pool_put(&pf_state_pl, gs);
6425 return;
6426 }
6427
6428 gsk = pf_alloc_state_key(gs, NULL);
6429 if (!gsk) {
6430 pool_put(&pf_app_state_pl, gas);
6431 pool_put(&pf_state_pl, gs);
6432 return;
6433 }
6434
6435 memcpy(&gsk->lan, &sk->lan, sizeof(gsk->lan));
6436 memcpy(&gsk->gwy, &sk->gwy, sizeof(gsk->gwy));
6437 memcpy(&gsk->ext_lan, &sk->ext_lan, sizeof(gsk->ext_lan));
6438 memcpy(&gsk->ext_gwy, &sk->ext_gwy, sizeof(gsk->ext_gwy));
6439 gsk->af_lan = sk->af_lan;
6440 gsk->af_gwy = sk->af_gwy;
6441 gsk->proto = IPPROTO_GRE;
6442 gsk->proto_variant = PF_GRE_PPTP_VARIANT;
6443 gsk->app_state = gas;
6444 gsk->lan.xport.call_id = 0;
6445 gsk->gwy.xport.call_id = 0;
6446 gsk->ext_lan.xport.call_id = 0;
6447 gsk->ext_gwy.xport.call_id = 0;
6448 gsk->flowsrc = FLOWSRC_PF;
6449 gsk->flowhash = pf_calc_state_key_flowhash(gsk);
6450 memset(gas, 0, sizeof(*gas));
6451 gas->u.grev1.pptp_state = s;
6452 STATE_INC_COUNTERS(gs);
6453 pptps->grev1_state = gs;
6454 (void) hook_establish(&gs->unlink_hooks, 0,
6455 (hook_fn_t) pf_grev1_unlink, gs);
6456 } else {
6457 gsk = gs->state_key;
6458 }
6459
6460 switch (sk->direction) {
6461 case PF_IN:
6462 pns_call_id = &gsk->ext_lan.xport.call_id;
6463 pns_state = &gs->dst.state;
6464 pac_call_id = &gsk->lan.xport.call_id;
6465 pac_state = &gs->src.state;
6466 break;
6467
6468 case PF_OUT:
6469 pns_call_id = &gsk->lan.xport.call_id;
6470 pns_state = &gs->src.state;
6471 pac_call_id = &gsk->ext_lan.xport.call_id;
6472 pac_state = &gs->dst.state;
6473 break;
6474
6475 default:
6476 DPFPRINTF(PF_DEBUG_URGENT,
6477 ("pf_pptp_handler: bad directional!\n"));
6478 return;
6479 }
6480
6481 spoof_call_id = 0;
6482 op = PF_PPTP_PASS;
6483
6484 ct = ntohs(cm.ctrl.type);
6485
6486 switch (ct) {
6487 case PF_PPTP_CTRL_TYPE_CALL_OUT_REQ:
6488 *pns_call_id = cm.msg.call_out_req.call_id;
6489 *pns_state = PFGRE1S_INITIATING;
6490 if (s->nat_rule.ptr && pns_call_id == &gsk->lan.xport.call_id) {
6491 spoof_call_id = &cm.msg.call_out_req.call_id;
6492 }
6493 break;
6494
6495 case PF_PPTP_CTRL_TYPE_CALL_OUT_RPY:
6496 *pac_call_id = cm.msg.call_out_rpy.call_id;
6497 if (s->nat_rule.ptr) {
6498 spoof_call_id =
6499 (pac_call_id == &gsk->lan.xport.call_id) ?
6500 &cm.msg.call_out_rpy.call_id :
6501 &cm.msg.call_out_rpy.peer_call_id;
6502 }
6503 if (gs->timeout == PFTM_UNLINKED) {
6504 *pac_state = PFGRE1S_INITIATING;
6505 op = PF_PPTP_INSERT_GRE;
6506 }
6507 break;
6508
6509 case PF_PPTP_CTRL_TYPE_CALL_IN_1ST:
6510 *pns_call_id = cm.msg.call_in_1st.call_id;
6511 *pns_state = PFGRE1S_INITIATING;
6512 if (s->nat_rule.ptr && pns_call_id == &gsk->lan.xport.call_id) {
6513 spoof_call_id = &cm.msg.call_in_1st.call_id;
6514 }
6515 break;
6516
6517 case PF_PPTP_CTRL_TYPE_CALL_IN_2ND:
6518 *pac_call_id = cm.msg.call_in_2nd.call_id;
6519 *pac_state = PFGRE1S_INITIATING;
6520 if (s->nat_rule.ptr) {
6521 spoof_call_id =
6522 (pac_call_id == &gsk->lan.xport.call_id) ?
6523 &cm.msg.call_in_2nd.call_id :
6524 &cm.msg.call_in_2nd.peer_call_id;
6525 }
6526 break;
6527
6528 case PF_PPTP_CTRL_TYPE_CALL_IN_3RD:
6529 if (s->nat_rule.ptr && pns_call_id == &gsk->lan.xport.call_id) {
6530 spoof_call_id = &cm.msg.call_in_3rd.call_id;
6531 }
6532 if (cm.msg.call_in_3rd.call_id != *pns_call_id) {
6533 break;
6534 }
6535 if (gs->timeout == PFTM_UNLINKED) {
6536 op = PF_PPTP_INSERT_GRE;
6537 }
6538 break;
6539
6540 case PF_PPTP_CTRL_TYPE_CALL_CLR:
6541 if (cm.msg.call_clr.call_id != *pns_call_id) {
6542 op = PF_PPTP_REMOVE_GRE;
6543 }
6544 break;
6545
6546 case PF_PPTP_CTRL_TYPE_CALL_DISC:
6547 if (cm.msg.call_clr.call_id != *pac_call_id) {
6548 op = PF_PPTP_REMOVE_GRE;
6549 }
6550 break;
6551
6552 case PF_PPTP_CTRL_TYPE_ERROR:
6553 if (s->nat_rule.ptr && pns_call_id == &gsk->lan.xport.call_id) {
6554 spoof_call_id = &cm.msg.error.peer_call_id;
6555 }
6556 break;
6557
6558 case PF_PPTP_CTRL_TYPE_SET_LINKINFO:
6559 if (s->nat_rule.ptr && pac_call_id == &gsk->lan.xport.call_id) {
6560 spoof_call_id = &cm.msg.set_linkinfo.peer_call_id;
6561 }
6562 break;
6563
6564 default:
6565 op = PF_PPTP_PASS;
6566 break;
6567 }
6568
6569 if (!gsk->gwy.xport.call_id && gsk->lan.xport.call_id) {
6570 gsk->gwy.xport.call_id = gsk->lan.xport.call_id;
6571 if (spoof_call_id) {
6572 u_int16_t call_id = 0;
6573 int n = 0;
6574 struct pf_state_key_cmp key;
6575
6576 key.af_gwy = gsk->af_gwy;
6577 key.proto = IPPROTO_GRE;
6578 key.proto_variant = PF_GRE_PPTP_VARIANT;
6579 PF_ACPY(&key.gwy.addr, &gsk->gwy.addr, key.af_gwy);
6580 PF_ACPY(&key.ext_gwy.addr, &gsk->ext_gwy.addr, key.af_gwy);
6581 key.gwy.xport.call_id = gsk->gwy.xport.call_id;
6582 key.ext_gwy.xport.call_id = gsk->ext_gwy.xport.call_id;
6583 do {
6584 call_id = htonl(random());
6585 } while (!call_id);
6586
6587 while (pf_find_state_all(&key, PF_IN, 0)) {
6588 call_id = ntohs(call_id);
6589 --call_id;
6590 if (--call_id == 0) {
6591 call_id = 0xffff;
6592 }
6593 call_id = htons(call_id);
6594
6595 key.gwy.xport.call_id = call_id;
6596
6597 if (++n > 65535) {
6598 DPFPRINTF(PF_DEBUG_URGENT,
6599 ("pf_pptp_handler: failed to spoof "
6600 "call id\n"));
6601 key.gwy.xport.call_id = 0;
6602 break;
6603 }
6604 }
6605
6606 gsk->gwy.xport.call_id = call_id;
6607 }
6608 }
6609
6610 th = pd->hdr.tcp;
6611
6612 if (spoof_call_id && gsk->lan.xport.call_id != gsk->gwy.xport.call_id) {
6613 if (*spoof_call_id == gsk->gwy.xport.call_id) {
6614 *spoof_call_id = gsk->lan.xport.call_id;
6615 th->th_sum = pf_cksum_fixup(th->th_sum,
6616 gsk->gwy.xport.call_id, gsk->lan.xport.call_id, 0);
6617 } else {
6618 *spoof_call_id = gsk->gwy.xport.call_id;
6619 th->th_sum = pf_cksum_fixup(th->th_sum,
6620 gsk->lan.xport.call_id, gsk->gwy.xport.call_id, 0);
6621 }
6622
6623 if (pf_lazy_makewritable(pd, pbuf, off + plen) == NULL) {
6624 pptps->grev1_state = NULL;
6625 STATE_DEC_COUNTERS(gs);
6626 pool_put(&pf_state_pl, gs);
6627 return;
6628 }
6629 pbuf_copy_back(pbuf, off, plen, &cm);
6630 }
6631
6632 switch (op) {
6633 case PF_PPTP_REMOVE_GRE:
6634 gs->timeout = PFTM_PURGE;
6635 gs->src.state = gs->dst.state = PFGRE1S_NO_TRAFFIC;
6636 gsk->lan.xport.call_id = 0;
6637 gsk->gwy.xport.call_id = 0;
6638 gsk->ext_lan.xport.call_id = 0;
6639 gsk->ext_gwy.xport.call_id = 0;
6640 gs->id = gs->creatorid = 0;
6641 break;
6642
6643 case PF_PPTP_INSERT_GRE:
6644 gs->creation = pf_time_second();
6645 gs->expire = pf_time_second();
6646 gs->timeout = PFTM_TCP_ESTABLISHED;
6647 if (gs->src_node != NULL) {
6648 ++gs->src_node->states;
6649 VERIFY(gs->src_node->states != 0);
6650 }
6651 if (gs->nat_src_node != NULL) {
6652 ++gs->nat_src_node->states;
6653 VERIFY(gs->nat_src_node->states != 0);
6654 }
6655 pf_set_rt_ifp(gs, &sk->lan.addr, sk->af_lan);
6656 if (pf_insert_state(BOUND_IFACE(s->rule.ptr, kif), gs)) {
6657 /*
6658 * <jhw@apple.com>
6659 * FIX ME: insertion can fail when multiple PNS
6660 * behind the same NAT open calls to the same PAC
6661 * simultaneously because spoofed call ID numbers
6662 * are chosen before states are inserted. This is
6663 * hard to fix and happens infrequently enough that
6664 * users will normally try again and this ALG will
6665 * succeed. Failures are expected to be rare enough
6666 * that fixing this is a low priority.
6667 */
6668 pptps->grev1_state = NULL;
6669 pd->lmw = -1; /* Force PF_DROP on PFRES_MEMORY */
6670 pf_src_tree_remove_state(gs);
6671 STATE_DEC_COUNTERS(gs);
6672 pool_put(&pf_state_pl, gs);
6673 DPFPRINTF(PF_DEBUG_URGENT, ("pf_pptp_handler: error "
6674 "inserting GREv1 state.\n"));
6675 }
6676 break;
6677
6678 default:
6679 break;
6680 }
6681 }
6682
6683 static void
6684 pf_pptp_unlink(struct pf_state *s)
6685 {
6686 struct pf_app_state *as = s->state_key->app_state;
6687 struct pf_state *grev1s = as->u.pptp.grev1_state;
6688
6689 if (grev1s) {
6690 struct pf_app_state *gas = grev1s->state_key->app_state;
6691
6692 if (grev1s->timeout < PFTM_MAX) {
6693 grev1s->timeout = PFTM_PURGE;
6694 }
6695 gas->u.grev1.pptp_state = NULL;
6696 as->u.pptp.grev1_state = NULL;
6697 }
6698 }
6699
6700 static void
6701 pf_grev1_unlink(struct pf_state *s)
6702 {
6703 struct pf_app_state *as = s->state_key->app_state;
6704 struct pf_state *pptps = as->u.grev1.pptp_state;
6705
6706 if (pptps) {
6707 struct pf_app_state *pas = pptps->state_key->app_state;
6708
6709 pas->u.pptp.grev1_state = NULL;
6710 as->u.grev1.pptp_state = NULL;
6711 }
6712 }
6713
6714 static int
6715 pf_ike_compare(struct pf_app_state *a, struct pf_app_state *b)
6716 {
6717 int64_t d = a->u.ike.cookie - b->u.ike.cookie;
6718 return (d > 0) ? 1 : ((d < 0) ? -1 : 0);
6719 }
6720
6721 static int
6722 pf_do_nat64(struct pf_state_key *sk, struct pf_pdesc *pd, pbuf_t *pbuf,
6723 int off)
6724 {
6725 if (pd->af == AF_INET) {
6726 if (pd->af != sk->af_lan) {
6727 pd->ndaddr = sk->lan.addr;
6728 pd->naddr = sk->ext_lan.addr;
6729 } else {
6730 pd->naddr = sk->gwy.addr;
6731 pd->ndaddr = sk->ext_gwy.addr;
6732 }
6733 return pf_nat64_ipv4(pbuf, off, pd);
6734 } else if (pd->af == AF_INET6) {
6735 if (pd->af != sk->af_lan) {
6736 pd->ndaddr = sk->lan.addr;
6737 pd->naddr = sk->ext_lan.addr;
6738 } else {
6739 pd->naddr = sk->gwy.addr;
6740 pd->ndaddr = sk->ext_gwy.addr;
6741 }
6742 return pf_nat64_ipv6(pbuf, off, pd);
6743 }
6744 return PF_DROP;
6745 }
6746
6747 static int
6748 pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
6749 pbuf_t *pbuf, int off, void *h, struct pf_pdesc *pd,
6750 u_short *reason)
6751 {
6752 #pragma unused(h)
6753 struct pf_state_key_cmp key;
6754 struct tcphdr *th = pd->hdr.tcp;
6755 u_int16_t win = ntohs(th->th_win);
6756 u_int32_t ack, end, seq, orig_seq;
6757 u_int8_t sws, dws;
6758 int ackskew;
6759 int copyback = 0;
6760 struct pf_state_peer *src, *dst;
6761 struct pf_state_key *sk;
6762
6763 key.app_state = 0;
6764 key.proto = IPPROTO_TCP;
6765 key.af_lan = key.af_gwy = pd->af;
6766
6767 /*
6768 * For NAT64 the first time rule search and state creation
6769 * is done on the incoming side only.
6770 * Once the state gets created, NAT64's LAN side (ipv6) will
6771 * not be able to find the state in ext-gwy tree as that normally
6772 * is intended to be looked up for incoming traffic from the
6773 * WAN side.
6774 * Therefore to handle NAT64 case we init keys here for both
6775 * lan-ext as well as ext-gwy trees.
6776 * In the state lookup we attempt a lookup on both trees if
6777 * first one does not return any result and return a match if
6778 * the match state's was created by NAT64 rule.
6779 */
6780 PF_ACPY(&key.ext_gwy.addr, pd->src, key.af_gwy);
6781 PF_ACPY(&key.gwy.addr, pd->dst, key.af_gwy);
6782 key.ext_gwy.xport.port = th->th_sport;
6783 key.gwy.xport.port = th->th_dport;
6784
6785 PF_ACPY(&key.lan.addr, pd->src, key.af_lan);
6786 PF_ACPY(&key.ext_lan.addr, pd->dst, key.af_lan);
6787 key.lan.xport.port = th->th_sport;
6788 key.ext_lan.xport.port = th->th_dport;
6789
6790 STATE_LOOKUP();
6791
6792 sk = (*state)->state_key;
6793 /*
6794 * In case of NAT64 the translation is first applied on the LAN
6795 * side. Therefore for stack's address family comparison
6796 * we use sk->af_lan.
6797 */
6798 if ((direction == sk->direction) && (pd->af == sk->af_lan)) {
6799 src = &(*state)->src;
6800 dst = &(*state)->dst;
6801 } else {
6802 src = &(*state)->dst;
6803 dst = &(*state)->src;
6804 }
6805
6806 if (src->state == PF_TCPS_PROXY_SRC) {
6807 if (direction != sk->direction) {
6808 REASON_SET(reason, PFRES_SYNPROXY);
6809 return PF_SYNPROXY_DROP;
6810 }
6811 if (th->th_flags & TH_SYN) {
6812 if (ntohl(th->th_seq) != src->seqlo) {
6813 REASON_SET(reason, PFRES_SYNPROXY);
6814 return PF_DROP;
6815 }
6816 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
6817 pd->src, th->th_dport, th->th_sport,
6818 src->seqhi, ntohl(th->th_seq) + 1,
6819 TH_SYN | TH_ACK, 0, src->mss, 0, 1,
6820 0, NULL, NULL);
6821 REASON_SET(reason, PFRES_SYNPROXY);
6822 return PF_SYNPROXY_DROP;
6823 } else if (!(th->th_flags & TH_ACK) ||
6824 (ntohl(th->th_ack) != src->seqhi + 1) ||
6825 (ntohl(th->th_seq) != src->seqlo + 1)) {
6826 REASON_SET(reason, PFRES_SYNPROXY);
6827 return PF_DROP;
6828 } else if ((*state)->src_node != NULL &&
6829 pf_src_connlimit(state)) {
6830 REASON_SET(reason, PFRES_SRCLIMIT);
6831 return PF_DROP;
6832 } else {
6833 src->state = PF_TCPS_PROXY_DST;
6834 }
6835 }
6836 if (src->state == PF_TCPS_PROXY_DST) {
6837 struct pf_state_host *psrc, *pdst;
6838
6839 if (direction == PF_OUT) {
6840 psrc = &sk->gwy;
6841 pdst = &sk->ext_gwy;
6842 } else {
6843 psrc = &sk->ext_lan;
6844 pdst = &sk->lan;
6845 }
6846 if (direction == sk->direction) {
6847 if (((th->th_flags & (TH_SYN | TH_ACK)) != TH_ACK) ||
6848 (ntohl(th->th_ack) != src->seqhi + 1) ||
6849 (ntohl(th->th_seq) != src->seqlo + 1)) {
6850 REASON_SET(reason, PFRES_SYNPROXY);
6851 return PF_DROP;
6852 }
6853 src->max_win = MAX(ntohs(th->th_win), 1);
6854 if (dst->seqhi == 1) {
6855 dst->seqhi = htonl(random());
6856 }
6857 pf_send_tcp((*state)->rule.ptr, pd->af, &psrc->addr,
6858 &pdst->addr, psrc->xport.port, pdst->xport.port,
6859 dst->seqhi, 0, TH_SYN, 0,
6860 src->mss, 0, 0, (*state)->tag, NULL, NULL);
6861 REASON_SET(reason, PFRES_SYNPROXY);
6862 return PF_SYNPROXY_DROP;
6863 } else if (((th->th_flags & (TH_SYN | TH_ACK)) !=
6864 (TH_SYN | TH_ACK)) ||
6865 (ntohl(th->th_ack) != dst->seqhi + 1)) {
6866 REASON_SET(reason, PFRES_SYNPROXY);
6867 return PF_DROP;
6868 } else {
6869 dst->max_win = MAX(ntohs(th->th_win), 1);
6870 dst->seqlo = ntohl(th->th_seq);
6871 pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
6872 pd->src, th->th_dport, th->th_sport,
6873 ntohl(th->th_ack), ntohl(th->th_seq) + 1,
6874 TH_ACK, src->max_win, 0, 0, 0,
6875 (*state)->tag, NULL, NULL);
6876 pf_send_tcp((*state)->rule.ptr, pd->af, &psrc->addr,
6877 &pdst->addr, psrc->xport.port, pdst->xport.port,
6878 src->seqhi + 1, src->seqlo + 1,
6879 TH_ACK, dst->max_win, 0, 0, 1,
6880 0, NULL, NULL);
6881 src->seqdiff = dst->seqhi -
6882 src->seqlo;
6883 dst->seqdiff = src->seqhi -
6884 dst->seqlo;
6885 src->seqhi = src->seqlo +
6886 dst->max_win;
6887 dst->seqhi = dst->seqlo +
6888 src->max_win;
6889 src->wscale = dst->wscale = 0;
6890 src->state = dst->state =
6891 TCPS_ESTABLISHED;
6892 REASON_SET(reason, PFRES_SYNPROXY);
6893 return PF_SYNPROXY_DROP;
6894 }
6895 }
6896
6897 if (((th->th_flags & (TH_SYN | TH_ACK)) == TH_SYN) &&
6898 dst->state >= TCPS_FIN_WAIT_2 &&
6899 src->state >= TCPS_FIN_WAIT_2) {
6900 if (pf_status.debug >= PF_DEBUG_MISC) {
6901 printf("pf: state reuse ");
6902 pf_print_state(*state);
6903 pf_print_flags(th->th_flags);
6904 printf("\n");
6905 }
6906 /* XXX make sure it's the same direction ?? */
6907 src->state = dst->state = TCPS_CLOSED;
6908 pf_unlink_state(*state);
6909 *state = NULL;
6910 return PF_DROP;
6911 }
6912
6913 if ((th->th_flags & TH_SYN) == 0) {
6914 sws = (src->wscale & PF_WSCALE_FLAG) ?
6915 (src->wscale & PF_WSCALE_MASK) : TCP_MAX_WINSHIFT;
6916 dws = (dst->wscale & PF_WSCALE_FLAG) ?
6917 (dst->wscale & PF_WSCALE_MASK) : TCP_MAX_WINSHIFT;
6918 } else {
6919 sws = dws = 0;
6920 }
6921
6922 /*
6923 * Sequence tracking algorithm from Guido van Rooij's paper:
6924 * http://www.madison-gurkha.com/publications/tcp_filtering/
6925 * tcp_filtering.ps
6926 */
6927
6928 orig_seq = seq = ntohl(th->th_seq);
6929 if (src->seqlo == 0) {
6930 /* First packet from this end. Set its state */
6931
6932 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) &&
6933 src->scrub == NULL) {
6934 if (pf_normalize_tcp_init(pbuf, off, pd, th, src, dst)) {
6935 REASON_SET(reason, PFRES_MEMORY);
6936 return PF_DROP;
6937 }
6938 }
6939
6940 /* Deferred generation of sequence number modulator */
6941 if (dst->seqdiff && !src->seqdiff) {
6942 /* use random iss for the TCP server */
6943 while ((src->seqdiff = random() - seq) == 0) {
6944 ;
6945 }
6946 ack = ntohl(th->th_ack) - dst->seqdiff;
6947 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
6948 src->seqdiff), 0);
6949 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
6950 copyback = off + sizeof(*th);
6951 } else {
6952 ack = ntohl(th->th_ack);
6953 }
6954
6955 end = seq + pd->p_len;
6956 if (th->th_flags & TH_SYN) {
6957 end++;
6958 if (dst->wscale & PF_WSCALE_FLAG) {
6959 src->wscale = pf_get_wscale(pbuf, off,
6960 th->th_off, pd->af);
6961 if (src->wscale & PF_WSCALE_FLAG) {
6962 /*
6963 * Remove scale factor from initial
6964 * window
6965 */
6966 sws = src->wscale & PF_WSCALE_MASK;
6967 win = ((u_int32_t)win + (1 << sws) - 1)
6968 >> sws;
6969 dws = dst->wscale & PF_WSCALE_MASK;
6970 } else {
6971 /*
6972 * Window scale negotiation has failed,
6973 * therefore we must restore the window
6974 * scale in the state record that we
6975 * optimistically removed in
6976 * pf_test_rule(). Care is required to
6977 * prevent arithmetic overflow from
6978 * zeroing the window when it's
6979 * truncated down to 16-bits.
6980 */
6981 u_int32_t max_win = dst->max_win;
6982 max_win <<=
6983 dst->wscale & PF_WSCALE_MASK;
6984 dst->max_win = MIN(0xffff, max_win);
6985 /* in case of a retrans SYN|ACK */
6986 dst->wscale = 0;
6987 }
6988 }
6989 }
6990 if (th->th_flags & TH_FIN) {
6991 end++;
6992 }
6993
6994 src->seqlo = seq;
6995 if (src->state < TCPS_SYN_SENT) {
6996 src->state = TCPS_SYN_SENT;
6997 }
6998
6999 /*
7000 * May need to slide the window (seqhi may have been set by
7001 * the crappy stack check or if we picked up the connection
7002 * after establishment)
7003 */
7004 if (src->seqhi == 1 ||
7005 SEQ_GEQ(end + MAX(1, (u_int32_t)dst->max_win << dws),
7006 src->seqhi)) {
7007 src->seqhi = end + MAX(1, (u_int32_t)dst->max_win << dws);
7008 }
7009 if (win > src->max_win) {
7010 src->max_win = win;
7011 }
7012 } else {
7013 ack = ntohl(th->th_ack) - dst->seqdiff;
7014 if (src->seqdiff) {
7015 /* Modulate sequence numbers */
7016 pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
7017 src->seqdiff), 0);
7018 pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
7019 copyback = off + sizeof(*th);
7020 }
7021 end = seq + pd->p_len;
7022 if (th->th_flags & TH_SYN) {
7023 end++;
7024 }
7025 if (th->th_flags & TH_FIN) {
7026 end++;
7027 }
7028 }
7029
7030 if ((th->th_flags & TH_ACK) == 0) {
7031 /* Let it pass through the ack skew check */
7032 ack = dst->seqlo;
7033 } else if ((ack == 0 &&
7034 (th->th_flags & (TH_ACK | TH_RST)) == (TH_ACK | TH_RST)) ||
7035 /* broken tcp stacks do not set ack */
7036 (dst->state < TCPS_SYN_SENT)) {
7037 /*
7038 * Many stacks (ours included) will set the ACK number in an
7039 * FIN|ACK if the SYN times out -- no sequence to ACK.
7040 */
7041 ack = dst->seqlo;
7042 }
7043
7044 if (seq == end) {
7045 /* Ease sequencing restrictions on no data packets */
7046 seq = src->seqlo;
7047 end = seq;
7048 }
7049
7050 ackskew = dst->seqlo - ack;
7051
7052
7053 /*
7054 * Need to demodulate the sequence numbers in any TCP SACK options
7055 * (Selective ACK). We could optionally validate the SACK values
7056 * against the current ACK window, either forwards or backwards, but
7057 * I'm not confident that SACK has been implemented properly
7058 * everywhere. It wouldn't surprise me if several stacks accidently
7059 * SACK too far backwards of previously ACKed data. There really aren't
7060 * any security implications of bad SACKing unless the target stack
7061 * doesn't validate the option length correctly. Someone trying to
7062 * spoof into a TCP connection won't bother blindly sending SACK
7063 * options anyway.
7064 */
7065 if (dst->seqdiff && (th->th_off << 2) > (int)sizeof(struct tcphdr)) {
7066 copyback = pf_modulate_sack(pbuf, off, pd, th, dst);
7067 if (copyback == -1) {
7068 REASON_SET(reason, PFRES_MEMORY);
7069 return PF_DROP;
7070 }
7071
7072 pbuf = pd->mp; // XXXSCW: Why?
7073 }
7074
7075
7076 #define MAXACKWINDOW (0xffff + 1500) /* 1500 is an arbitrary fudge factor */
7077 if (SEQ_GEQ(src->seqhi, end) &&
7078 /* Last octet inside other's window space */
7079 SEQ_GEQ(seq, src->seqlo - ((u_int32_t)dst->max_win << dws)) &&
7080 /* Retrans: not more than one window back */
7081 (ackskew >= -MAXACKWINDOW) &&
7082 /* Acking not more than one reassembled fragment backwards */
7083 (ackskew <= (MAXACKWINDOW << sws)) &&
7084 /* Acking not more than one window forward */
7085 ((th->th_flags & TH_RST) == 0 || orig_seq == src->seqlo ||
7086 (orig_seq == src->seqlo + 1) || (orig_seq + 1 == src->seqlo) ||
7087 (pd->flags & PFDESC_IP_REAS) == 0)) {
7088 /* Require an exact/+1 sequence match on resets when possible */
7089
7090 if (dst->scrub || src->scrub) {
7091 if (pf_normalize_tcp_stateful(pbuf, off, pd, reason, th,
7092 *state, src, dst, &copyback)) {
7093 return PF_DROP;
7094 }
7095
7096 pbuf = pd->mp; // XXXSCW: Why?
7097 }
7098
7099 /* update max window */
7100 if (src->max_win < win) {
7101 src->max_win = win;
7102 }
7103 /* synchronize sequencing */
7104 if (SEQ_GT(end, src->seqlo)) {
7105 src->seqlo = end;
7106 }
7107 /* slide the window of what the other end can send */
7108 if (SEQ_GEQ(ack + ((u_int32_t)win << sws), dst->seqhi)) {
7109 dst->seqhi = ack + MAX(((u_int32_t)win << sws), 1);
7110 }
7111
7112 /* update states */
7113 if (th->th_flags & TH_SYN) {
7114 if (src->state < TCPS_SYN_SENT) {
7115 src->state = TCPS_SYN_SENT;
7116 }
7117 }
7118 if (th->th_flags & TH_FIN) {
7119 if (src->state < TCPS_CLOSING) {
7120 src->state = TCPS_CLOSING;
7121 }
7122 }
7123 if (th->th_flags & TH_ACK) {
7124 if (dst->state == TCPS_SYN_SENT) {
7125 dst->state = TCPS_ESTABLISHED;
7126 if (src->state == TCPS_ESTABLISHED &&
7127 (*state)->src_node != NULL &&
7128 pf_src_connlimit(state)) {
7129 REASON_SET(reason, PFRES_SRCLIMIT);
7130 return PF_DROP;
7131 }
7132 } else if (dst->state == TCPS_CLOSING) {
7133 dst->state = TCPS_FIN_WAIT_2;
7134 }
7135 }
7136 if (th->th_flags & TH_RST) {
7137 src->state = dst->state = TCPS_TIME_WAIT;
7138 }
7139
7140 /* update expire time */
7141 (*state)->expire = pf_time_second();
7142 if (src->state >= TCPS_FIN_WAIT_2 &&
7143 dst->state >= TCPS_FIN_WAIT_2) {
7144 (*state)->timeout = PFTM_TCP_CLOSED;
7145 } else if (src->state >= TCPS_CLOSING &&
7146 dst->state >= TCPS_CLOSING) {
7147 (*state)->timeout = PFTM_TCP_FIN_WAIT;
7148 } else if (src->state < TCPS_ESTABLISHED ||
7149 dst->state < TCPS_ESTABLISHED) {
7150 (*state)->timeout = PFTM_TCP_OPENING;
7151 } else if (src->state >= TCPS_CLOSING ||
7152 dst->state >= TCPS_CLOSING) {
7153 (*state)->timeout = PFTM_TCP_CLOSING;
7154 } else {
7155 (*state)->timeout = PFTM_TCP_ESTABLISHED;
7156 }
7157
7158 /* Fall through to PASS packet */
7159 } else if ((dst->state < TCPS_SYN_SENT ||
7160 dst->state >= TCPS_FIN_WAIT_2 || src->state >= TCPS_FIN_WAIT_2) &&
7161 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) &&
7162 /* Within a window forward of the originating packet */
7163 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) {
7164 /* Within a window backward of the originating packet */
7165
7166 /*
7167 * This currently handles three situations:
7168 * 1) Stupid stacks will shotgun SYNs before their peer
7169 * replies.
7170 * 2) When PF catches an already established stream (the
7171 * firewall rebooted, the state table was flushed, routes
7172 * changed...)
7173 * 3) Packets get funky immediately after the connection
7174 * closes (this should catch Solaris spurious ACK|FINs
7175 * that web servers like to spew after a close)
7176 *
7177 * This must be a little more careful than the above code
7178 * since packet floods will also be caught here. We don't
7179 * update the TTL here to mitigate the damage of a packet
7180 * flood and so the same code can handle awkward establishment
7181 * and a loosened connection close.
7182 * In the establishment case, a correct peer response will
7183 * validate the connection, go through the normal state code
7184 * and keep updating the state TTL.
7185 */
7186
7187 if (pf_status.debug >= PF_DEBUG_MISC) {
7188 printf("pf: loose state match: ");
7189 pf_print_state(*state);
7190 pf_print_flags(th->th_flags);
7191 printf(" seq=%u (%u) ack=%u len=%u ackskew=%d "
7192 "pkts=%llu:%llu dir=%s,%s\n", seq, orig_seq, ack,
7193 pd->p_len, ackskew, (*state)->packets[0],
7194 (*state)->packets[1],
7195 direction == PF_IN ? "in" : "out",
7196 direction == sk->direction ?
7197 "fwd" : "rev");
7198 }
7199
7200 if (dst->scrub || src->scrub) {
7201 if (pf_normalize_tcp_stateful(pbuf, off, pd, reason, th,
7202 *state, src, dst, &copyback)) {
7203 return PF_DROP;
7204 }
7205 pbuf = pd->mp; // XXXSCW: Why?
7206 }
7207
7208 /* update max window */
7209 if (src->max_win < win) {
7210 src->max_win = win;
7211 }
7212 /* synchronize sequencing */
7213 if (SEQ_GT(end, src->seqlo)) {
7214 src->seqlo = end;
7215 }
7216 /* slide the window of what the other end can send */
7217 if (SEQ_GEQ(ack + ((u_int32_t)win << sws), dst->seqhi)) {
7218 dst->seqhi = ack + MAX(((u_int32_t)win << sws), 1);
7219 }
7220
7221 /*
7222 * Cannot set dst->seqhi here since this could be a shotgunned
7223 * SYN and not an already established connection.
7224 */
7225
7226 if (th->th_flags & TH_FIN) {
7227 if (src->state < TCPS_CLOSING) {
7228 src->state = TCPS_CLOSING;
7229 }
7230 }
7231 if (th->th_flags & TH_RST) {
7232 src->state = dst->state = TCPS_TIME_WAIT;
7233 }
7234
7235 /* Fall through to PASS packet */
7236 } else {
7237 if (dst->state == TCPS_SYN_SENT &&
7238 src->state == TCPS_SYN_SENT) {
7239 /* Send RST for state mismatches during handshake */
7240 if (!(th->th_flags & TH_RST)) {
7241 pf_send_tcp((*state)->rule.ptr, pd->af,
7242 pd->dst, pd->src, th->th_dport,
7243 th->th_sport, ntohl(th->th_ack), 0,
7244 TH_RST, 0, 0,
7245 (*state)->rule.ptr->return_ttl, 1, 0,
7246 pd->eh, kif->pfik_ifp);
7247 }
7248 src->seqlo = 0;
7249 src->seqhi = 1;
7250 src->max_win = 1;
7251 } else if (pf_status.debug >= PF_DEBUG_MISC) {
7252 printf("pf: BAD state: ");
7253 pf_print_state(*state);
7254 pf_print_flags(th->th_flags);
7255 printf("\n seq=%u (%u) ack=%u len=%u ackskew=%d "
7256 "sws=%u dws=%u pkts=%llu:%llu dir=%s,%s\n",
7257 seq, orig_seq, ack, pd->p_len, ackskew,
7258 (unsigned int)sws, (unsigned int)dws,
7259 (*state)->packets[0], (*state)->packets[1],
7260 direction == PF_IN ? "in" : "out",
7261 direction == sk->direction ?
7262 "fwd" : "rev");
7263 printf("pf: State failure on: %c %c %c %c | %c %c\n",
7264 SEQ_GEQ(src->seqhi, end) ? ' ' : '1',
7265 SEQ_GEQ(seq,
7266 src->seqlo - ((u_int32_t)dst->max_win << dws)) ?
7267 ' ': '2',
7268 (ackskew >= -MAXACKWINDOW) ? ' ' : '3',
7269 (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4',
7270 SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5',
7271 SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6');
7272 }
7273 REASON_SET(reason, PFRES_BADSTATE);
7274 return PF_DROP;
7275 }
7276
7277 /* Any packets which have gotten here are to be passed */
7278
7279 if (sk->app_state &&
7280 sk->app_state->handler) {
7281 sk->app_state->handler(*state, direction,
7282 off + (th->th_off << 2), pd, kif);
7283 if (pd->lmw < 0) {
7284 REASON_SET(reason, PFRES_MEMORY);
7285 return PF_DROP;
7286 }
7287 pbuf = pd->mp; // XXXSCW: Why?
7288 }
7289
7290 /* translate source/destination address, if necessary */
7291 if (STATE_TRANSLATE(sk)) {
7292 pd->naf = (pd->af == sk->af_lan) ? sk->af_gwy : sk->af_lan;
7293
7294 if (direction == PF_OUT) {
7295 pf_change_ap(direction, pd->mp, pd->src, &th->th_sport,
7296 pd->ip_sum, &th->th_sum, &sk->gwy.addr,
7297 sk->gwy.xport.port, 0, pd->af, pd->naf, 1);
7298 } else {
7299 if (pd->af != pd->naf) {
7300 if (pd->af == sk->af_gwy) {
7301 pf_change_ap(direction, pd->mp, pd->dst,
7302 &th->th_dport, pd->ip_sum,
7303 &th->th_sum, &sk->lan.addr,
7304 sk->lan.xport.port, 0,
7305 pd->af, pd->naf, 0);
7306
7307 pf_change_ap(direction, pd->mp, pd->src,
7308 &th->th_sport, pd->ip_sum,
7309 &th->th_sum, &sk->ext_lan.addr,
7310 th->th_sport, 0, pd->af,
7311 pd->naf, 0);
7312 } else {
7313 pf_change_ap(direction, pd->mp, pd->dst,
7314 &th->th_dport, pd->ip_sum,
7315 &th->th_sum, &sk->ext_gwy.addr,
7316 th->th_dport, 0, pd->af,
7317 pd->naf, 0);
7318
7319 pf_change_ap(direction, pd->mp, pd->src,
7320 &th->th_sport, pd->ip_sum,
7321 &th->th_sum, &sk->gwy.addr,
7322 sk->gwy.xport.port, 0, pd->af,
7323 pd->naf, 0);
7324 }
7325 } else {
7326 pf_change_ap(direction, pd->mp, pd->dst,
7327 &th->th_dport, pd->ip_sum,
7328 &th->th_sum, &sk->lan.addr,
7329 sk->lan.xport.port, 0, pd->af,
7330 pd->naf, 1);
7331 }
7332 }
7333
7334 copyback = off + sizeof(*th);
7335 }
7336
7337 if (copyback) {
7338 if (pf_lazy_makewritable(pd, pbuf, copyback) == NULL) {
7339 REASON_SET(reason, PFRES_MEMORY);
7340 return PF_DROP;
7341 }
7342
7343 /* Copyback sequence modulation or stateful scrub changes */
7344 pbuf_copy_back(pbuf, off, sizeof(*th), th);
7345
7346 if (sk->af_lan != sk->af_gwy) {
7347 return pf_do_nat64(sk, pd, pbuf, off);
7348 }
7349 }
7350 return PF_PASS;
7351 }
7352
7353 static int
7354 pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif,
7355 pbuf_t *pbuf, int off, void *h, struct pf_pdesc *pd, u_short *reason)
7356 {
7357 #pragma unused(h)
7358 struct pf_state_peer *src, *dst;
7359 struct pf_state_key_cmp key;
7360 struct pf_state_key *sk;
7361 struct udphdr *uh = pd->hdr.udp;
7362 struct pf_app_state as;
7363 int action, extfilter;
7364 key.app_state = 0;
7365 key.proto_variant = PF_EXTFILTER_APD;
7366
7367 key.proto = IPPROTO_UDP;
7368 key.af_lan = key.af_gwy = pd->af;
7369
7370 /*
7371 * For NAT64 the first time rule search and state creation
7372 * is done on the incoming side only.
7373 * Once the state gets created, NAT64's LAN side (ipv6) will
7374 * not be able to find the state in ext-gwy tree as that normally
7375 * is intended to be looked up for incoming traffic from the
7376 * WAN side.
7377 * Therefore to handle NAT64 case we init keys here for both
7378 * lan-ext as well as ext-gwy trees.
7379 * In the state lookup we attempt a lookup on both trees if
7380 * first one does not return any result and return a match if
7381 * the match state's was created by NAT64 rule.
7382 */
7383 PF_ACPY(&key.ext_gwy.addr, pd->src, key.af_gwy);
7384 PF_ACPY(&key.gwy.addr, pd->dst, key.af_gwy);
7385 key.ext_gwy.xport.port = uh->uh_sport;
7386 key.gwy.xport.port = uh->uh_dport;
7387
7388 PF_ACPY(&key.lan.addr, pd->src, key.af_lan);
7389 PF_ACPY(&key.ext_lan.addr, pd->dst, key.af_lan);
7390 key.lan.xport.port = uh->uh_sport;
7391 key.ext_lan.xport.port = uh->uh_dport;
7392
7393 if (ntohs(uh->uh_sport) == PF_IKE_PORT &&
7394 ntohs(uh->uh_dport) == PF_IKE_PORT) {
7395 struct pf_ike_hdr ike;
7396 size_t plen = pbuf->pb_packet_len - off - sizeof(*uh);
7397 if (plen < PF_IKE_PACKET_MINSIZE) {
7398 DPFPRINTF(PF_DEBUG_MISC,
7399 ("pf: IKE message too small.\n"));
7400 return PF_DROP;
7401 }
7402
7403 if (plen > sizeof(ike)) {
7404 plen = sizeof(ike);
7405 }
7406 pbuf_copy_data(pbuf, off + sizeof(*uh), plen, &ike);
7407
7408 if (ike.initiator_cookie) {
7409 key.app_state = &as;
7410 as.compare_lan_ext = pf_ike_compare;
7411 as.compare_ext_gwy = pf_ike_compare;
7412 as.u.ike.cookie = ike.initiator_cookie;
7413 } else {
7414 /*
7415 * <http://tools.ietf.org/html/\
7416 * draft-ietf-ipsec-nat-t-ike-01>
7417 * Support non-standard NAT-T implementations that
7418 * push the ESP packet over the top of the IKE packet.
7419 * Do not drop packet.
7420 */
7421 DPFPRINTF(PF_DEBUG_MISC,
7422 ("pf: IKE initiator cookie = 0.\n"));
7423 }
7424 }
7425
7426 *state = pf_find_state(kif, &key, direction);
7427
7428 if (!key.app_state && *state == 0) {
7429 key.proto_variant = PF_EXTFILTER_AD;
7430 *state = pf_find_state(kif, &key, direction);
7431 }
7432
7433 if (!key.app_state && *state == 0) {
7434 key.proto_variant = PF_EXTFILTER_EI;
7435 *state = pf_find_state(kif, &key, direction);
7436 }
7437
7438 /* similar to STATE_LOOKUP() */
7439 if (*state != NULL && pd != NULL && !(pd->pktflags & PKTF_FLOW_ID)) {
7440 pd->flowsrc = (*state)->state_key->flowsrc;
7441 pd->flowhash = (*state)->state_key->flowhash;
7442 if (pd->flowhash != 0) {
7443 pd->pktflags |= PKTF_FLOW_ID;
7444 pd->pktflags &= ~PKTF_FLOW_ADV;
7445 }
7446 }
7447
7448 if (pf_state_lookup_aux(state, kif, direction, &action)) {
7449 return action;
7450 }
7451
7452 sk = (*state)->state_key;
7453
7454 /*
7455 * In case of NAT64 the translation is first applied on the LAN
7456 * side. Therefore for stack's address family comparison
7457 * we use sk->af_lan.
7458 */
7459 if ((direction == sk->direction) && (pd->af == sk->af_lan)) {
7460 src = &(*state)->src;
7461 dst = &(*state)->dst;
7462 } else {
7463 src = &(*state)->dst;
7464 dst = &(*state)->src;
7465 }
7466
7467 /* update states */
7468 if (src->state < PFUDPS_SINGLE) {
7469 src->state = PFUDPS_SINGLE;
7470 }
7471 if (dst->state == PFUDPS_SINGLE) {
7472 dst->state = PFUDPS_MULTIPLE;
7473 }
7474
7475 /* update expire time */
7476 (*state)->expire = pf_time_second();
7477 if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE) {
7478 (*state)->timeout = PFTM_UDP_MULTIPLE;
7479 } else {
7480 (*state)->timeout = PFTM_UDP_SINGLE;
7481 }
7482
7483 extfilter = sk->proto_variant;
7484 if (extfilter > PF_EXTFILTER_APD) {
7485 if (direction == PF_OUT) {
7486 sk->ext_lan.xport.port = key.ext_lan.xport.port;
7487 if (extfilter > PF_EXTFILTER_AD) {
7488 PF_ACPY(&sk->ext_lan.addr, &key.ext_lan.addr,
7489 key.af_lan);
7490 }
7491 } else {
7492 sk->ext_gwy.xport.port = key.ext_gwy.xport.port;
7493 if (extfilter > PF_EXTFILTER_AD) {
7494 PF_ACPY(&sk->ext_gwy.addr, &key.ext_gwy.addr,
7495 key.af_gwy);
7496 }
7497 }
7498 }
7499
7500 if (sk->app_state && sk->app_state->handler) {
7501 sk->app_state->handler(*state, direction, off + uh->uh_ulen,
7502 pd, kif);
7503 if (pd->lmw < 0) {
7504 REASON_SET(reason, PFRES_MEMORY);
7505 return PF_DROP;
7506 }
7507 pbuf = pd->mp; // XXXSCW: Why?
7508 }
7509
7510 /* translate source/destination address, if necessary */
7511 if (STATE_TRANSLATE(sk)) {
7512 if (pf_lazy_makewritable(pd, pbuf, off + sizeof(*uh)) == NULL) {
7513 REASON_SET(reason, PFRES_MEMORY);
7514 return PF_DROP;
7515 }
7516
7517 pd->naf = (pd->af == sk->af_lan) ? sk->af_gwy : sk->af_lan;
7518
7519 if (direction == PF_OUT) {
7520 pf_change_ap(direction, pd->mp, pd->src, &uh->uh_sport,
7521 pd->ip_sum, &uh->uh_sum, &sk->gwy.addr,
7522 sk->gwy.xport.port, 1, pd->af, pd->naf, 1);
7523 } else {
7524 if (pd->af != pd->naf) {
7525 if (pd->af == sk->af_gwy) {
7526 pf_change_ap(direction, pd->mp, pd->dst,
7527 &uh->uh_dport, pd->ip_sum,
7528 &uh->uh_sum, &sk->lan.addr,
7529 sk->lan.xport.port, 1,
7530 pd->af, pd->naf, 0);
7531
7532 pf_change_ap(direction, pd->mp, pd->src,
7533 &uh->uh_sport, pd->ip_sum,
7534 &uh->uh_sum, &sk->ext_lan.addr,
7535 uh->uh_sport, 1, pd->af,
7536 pd->naf, 0);
7537 } else {
7538 pf_change_ap(direction, pd->mp, pd->dst,
7539 &uh->uh_dport, pd->ip_sum,
7540 &uh->uh_sum, &sk->ext_gwy.addr,
7541 uh->uh_dport, 1, pd->af,
7542 pd->naf, 0);
7543
7544 pf_change_ap(direction, pd->mp, pd->src,
7545 &uh->uh_sport, pd->ip_sum,
7546 &uh->uh_sum, &sk->gwy.addr,
7547 sk->gwy.xport.port, 1, pd->af,
7548 pd->naf, 0);
7549 }
7550 } else {
7551 pf_change_ap(direction, pd->mp, pd->dst,
7552 &uh->uh_dport, pd->ip_sum,
7553 &uh->uh_sum, &sk->lan.addr,
7554 sk->lan.xport.port, 1,
7555 pd->af, pd->naf, 1);
7556 }
7557 }
7558
7559 pbuf_copy_back(pbuf, off, sizeof(*uh), uh);
7560 if (sk->af_lan != sk->af_gwy) {
7561 return pf_do_nat64(sk, pd, pbuf, off);
7562 }
7563 }
7564 return PF_PASS;
7565 }
7566
7567 static int
7568 pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif,
7569 pbuf_t *pbuf, int off, void *h, struct pf_pdesc *pd, u_short *reason)
7570 {
7571 #pragma unused(h)
7572 struct pf_addr *saddr = pd->src, *daddr = pd->dst;
7573 struct in_addr srcv4_inaddr = saddr->v4addr;
7574 u_int16_t icmpid = 0, *icmpsum = NULL;
7575 u_int8_t icmptype = 0;
7576 int state_icmp = 0;
7577 struct pf_state_key_cmp key;
7578 struct pf_state_key *sk;
7579
7580 struct pf_app_state as;
7581 key.app_state = 0;
7582
7583 pd->off = off;
7584
7585 switch (pd->proto) {
7586 #if INET
7587 case IPPROTO_ICMP:
7588 icmptype = pd->hdr.icmp->icmp_type;
7589 icmpid = pd->hdr.icmp->icmp_id;
7590 icmpsum = &pd->hdr.icmp->icmp_cksum;
7591
7592 if (ICMP_ERRORTYPE(icmptype)) {
7593 state_icmp++;
7594 }
7595 break;
7596 #endif /* INET */
7597 case IPPROTO_ICMPV6:
7598 icmptype = pd->hdr.icmp6->icmp6_type;
7599 icmpid = pd->hdr.icmp6->icmp6_id;
7600 icmpsum = &pd->hdr.icmp6->icmp6_cksum;
7601
7602 if (ICMP6_ERRORTYPE(icmptype)) {
7603 state_icmp++;
7604 }
7605 break;
7606 }
7607
7608 if (!state_icmp) {
7609 /*
7610 * ICMP query/reply message not related to a TCP/UDP packet.
7611 * Search for an ICMP state.
7612 */
7613 /*
7614 * NAT64 requires protocol translation between ICMPv4
7615 * and ICMPv6. TCP and UDP do not require protocol
7616 * translation. To avoid adding complexity just to
7617 * handle ICMP(v4addr/v6addr), we always lookup for
7618 * proto = IPPROTO_ICMP on both LAN and WAN side
7619 */
7620 key.proto = IPPROTO_ICMP;
7621 key.af_lan = key.af_gwy = pd->af;
7622
7623 PF_ACPY(&key.ext_gwy.addr, pd->src, key.af_gwy);
7624 PF_ACPY(&key.gwy.addr, pd->dst, key.af_gwy);
7625 key.ext_gwy.xport.port = 0;
7626 key.gwy.xport.port = icmpid;
7627
7628 PF_ACPY(&key.lan.addr, pd->src, key.af_lan);
7629 PF_ACPY(&key.ext_lan.addr, pd->dst, key.af_lan);
7630 key.lan.xport.port = icmpid;
7631 key.ext_lan.xport.port = 0;
7632
7633 STATE_LOOKUP();
7634
7635 sk = (*state)->state_key;
7636 (*state)->expire = pf_time_second();
7637 (*state)->timeout = PFTM_ICMP_ERROR_REPLY;
7638
7639 /* translate source/destination address, if necessary */
7640 if (STATE_TRANSLATE(sk)) {
7641 pd->naf = (pd->af == sk->af_lan) ?
7642 sk->af_gwy : sk->af_lan;
7643 if (direction == PF_OUT) {
7644 switch (pd->af) {
7645 #if INET
7646 case AF_INET:
7647 pf_change_a(&saddr->v4addr.s_addr,
7648 pd->ip_sum,
7649 sk->gwy.addr.v4addr.s_addr, 0);
7650 pd->hdr.icmp->icmp_cksum =
7651 pf_cksum_fixup(
7652 pd->hdr.icmp->icmp_cksum, icmpid,
7653 sk->gwy.xport.port, 0);
7654 pd->hdr.icmp->icmp_id =
7655 sk->gwy.xport.port;
7656 if (pf_lazy_makewritable(pd, pbuf,
7657 off + ICMP_MINLEN) == NULL) {
7658 return PF_DROP;
7659 }
7660 pbuf_copy_back(pbuf, off, ICMP_MINLEN,
7661 pd->hdr.icmp);
7662 break;
7663 #endif /* INET */
7664 case AF_INET6:
7665 pf_change_a6(saddr,
7666 &pd->hdr.icmp6->icmp6_cksum,
7667 &sk->gwy.addr, 0);
7668 if (pf_lazy_makewritable(pd, pbuf,
7669 off + sizeof(struct icmp6_hdr)) ==
7670 NULL) {
7671 return PF_DROP;
7672 }
7673 pbuf_copy_back(pbuf, off,
7674 sizeof(struct icmp6_hdr),
7675 pd->hdr.icmp6);
7676 break;
7677 }
7678 } else {
7679 switch (pd->af) {
7680 #if INET
7681 case AF_INET:
7682 if (pd->naf != AF_INET) {
7683 if (pf_translate_icmp_af(
7684 AF_INET6, pd->hdr.icmp)) {
7685 return PF_DROP;
7686 }
7687
7688 pd->proto = IPPROTO_ICMPV6;
7689 } else {
7690 pf_change_a(&daddr->v4addr.s_addr,
7691 pd->ip_sum,
7692 sk->lan.addr.v4addr.s_addr, 0);
7693
7694 pd->hdr.icmp->icmp_cksum =
7695 pf_cksum_fixup(
7696 pd->hdr.icmp->icmp_cksum,
7697 icmpid, sk->lan.xport.port, 0);
7698
7699 pd->hdr.icmp->icmp_id =
7700 sk->lan.xport.port;
7701 }
7702
7703 if (pf_lazy_makewritable(pd, pbuf,
7704 off + ICMP_MINLEN) == NULL) {
7705 return PF_DROP;
7706 }
7707 pbuf_copy_back(pbuf, off, ICMP_MINLEN,
7708 pd->hdr.icmp);
7709 if (sk->af_lan != sk->af_gwy) {
7710 return pf_do_nat64(sk, pd,
7711 pbuf, off);
7712 }
7713 break;
7714 #endif /* INET */
7715 case AF_INET6:
7716 if (pd->naf != AF_INET6) {
7717 if (pf_translate_icmp_af(
7718 AF_INET, pd->hdr.icmp6)) {
7719 return PF_DROP;
7720 }
7721
7722 pd->proto = IPPROTO_ICMP;
7723 } else {
7724 pf_change_a6(daddr,
7725 &pd->hdr.icmp6->icmp6_cksum,
7726 &sk->lan.addr, 0);
7727 }
7728 if (pf_lazy_makewritable(pd, pbuf,
7729 off + sizeof(struct icmp6_hdr)) ==
7730 NULL) {
7731 return PF_DROP;
7732 }
7733 pbuf_copy_back(pbuf, off,
7734 sizeof(struct icmp6_hdr),
7735 pd->hdr.icmp6);
7736 if (sk->af_lan != sk->af_gwy) {
7737 return pf_do_nat64(sk, pd,
7738 pbuf, off);
7739 }
7740 break;
7741 }
7742 }
7743 }
7744
7745 return PF_PASS;
7746 } else {
7747 /*
7748 * ICMP error message in response to a TCP/UDP packet.
7749 * Extract the inner TCP/UDP header and search for that state.
7750 */
7751 struct pf_pdesc pd2; /* For inner (original) header */
7752 #if INET
7753 struct ip h2;
7754 #endif /* INET */
7755 struct ip6_hdr h2_6;
7756 int terminal = 0;
7757 int ipoff2 = 0;
7758 int off2 = 0;
7759
7760 memset(&pd2, 0, sizeof(pd2));
7761
7762 pd2.af = pd->af;
7763 switch (pd->af) {
7764 #if INET
7765 case AF_INET:
7766 /* offset of h2 in mbuf chain */
7767 ipoff2 = off + ICMP_MINLEN;
7768
7769 if (!pf_pull_hdr(pbuf, ipoff2, &h2, sizeof(h2),
7770 NULL, reason, pd2.af)) {
7771 DPFPRINTF(PF_DEBUG_MISC,
7772 ("pf: ICMP error message too short "
7773 "(ip)\n"));
7774 return PF_DROP;
7775 }
7776 /*
7777 * ICMP error messages don't refer to non-first
7778 * fragments
7779 */
7780 if (h2.ip_off & htons(IP_OFFMASK)) {
7781 REASON_SET(reason, PFRES_FRAG);
7782 return PF_DROP;
7783 }
7784
7785 /* offset of protocol header that follows h2 */
7786 off2 = ipoff2 + (h2.ip_hl << 2);
7787 /* TODO */
7788 pd2.off = ipoff2 + (h2.ip_hl << 2);
7789
7790 pd2.proto = h2.ip_p;
7791 pd2.src = (struct pf_addr *)&h2.ip_src;
7792 pd2.dst = (struct pf_addr *)&h2.ip_dst;
7793 pd2.ip_sum = &h2.ip_sum;
7794 break;
7795 #endif /* INET */
7796 case AF_INET6:
7797 ipoff2 = off + sizeof(struct icmp6_hdr);
7798
7799 if (!pf_pull_hdr(pbuf, ipoff2, &h2_6, sizeof(h2_6),
7800 NULL, reason, pd2.af)) {
7801 DPFPRINTF(PF_DEBUG_MISC,
7802 ("pf: ICMP error message too short "
7803 "(ip6)\n"));
7804 return PF_DROP;
7805 }
7806 pd2.proto = h2_6.ip6_nxt;
7807 pd2.src = (struct pf_addr *)(uintptr_t)&h2_6.ip6_src;
7808 pd2.dst = (struct pf_addr *)(uintptr_t)&h2_6.ip6_dst;
7809 pd2.ip_sum = NULL;
7810 off2 = ipoff2 + sizeof(h2_6);
7811 do {
7812 switch (pd2.proto) {
7813 case IPPROTO_FRAGMENT:
7814 /*
7815 * ICMPv6 error messages for
7816 * non-first fragments
7817 */
7818 REASON_SET(reason, PFRES_FRAG);
7819 return PF_DROP;
7820 case IPPROTO_AH:
7821 case IPPROTO_HOPOPTS:
7822 case IPPROTO_ROUTING:
7823 case IPPROTO_DSTOPTS: {
7824 /* get next header and header length */
7825 struct ip6_ext opt6;
7826
7827 if (!pf_pull_hdr(pbuf, off2, &opt6,
7828 sizeof(opt6), NULL, reason,
7829 pd2.af)) {
7830 DPFPRINTF(PF_DEBUG_MISC,
7831 ("pf: ICMPv6 short opt\n"));
7832 return PF_DROP;
7833 }
7834 if (pd2.proto == IPPROTO_AH) {
7835 off2 += (opt6.ip6e_len + 2) * 4;
7836 } else {
7837 off2 += (opt6.ip6e_len + 1) * 8;
7838 }
7839 pd2.proto = opt6.ip6e_nxt;
7840 /* goto the next header */
7841 break;
7842 }
7843 default:
7844 terminal++;
7845 break;
7846 }
7847 } while (!terminal);
7848 /* TODO */
7849 pd2.off = ipoff2;
7850 break;
7851 }
7852
7853 switch (pd2.proto) {
7854 case IPPROTO_TCP: {
7855 struct tcphdr th;
7856 u_int32_t seq;
7857 struct pf_state_peer *src, *dst;
7858 u_int8_t dws;
7859 int copyback = 0;
7860
7861 /*
7862 * Only the first 8 bytes of the TCP header can be
7863 * expected. Don't access any TCP header fields after
7864 * th_seq, an ackskew test is not possible.
7865 */
7866 if (!pf_pull_hdr(pbuf, off2, &th, 8, NULL, reason,
7867 pd2.af)) {
7868 DPFPRINTF(PF_DEBUG_MISC,
7869 ("pf: ICMP error message too short "
7870 "(tcp)\n"));
7871 return PF_DROP;
7872 }
7873
7874 key.proto = IPPROTO_TCP;
7875 key.af_gwy = pd2.af;
7876 PF_ACPY(&key.ext_gwy.addr, pd2.dst, key.af_gwy);
7877 PF_ACPY(&key.gwy.addr, pd2.src, key.af_gwy);
7878 key.ext_gwy.xport.port = th.th_dport;
7879 key.gwy.xport.port = th.th_sport;
7880
7881 key.af_lan = pd2.af;
7882 PF_ACPY(&key.lan.addr, pd2.dst, key.af_lan);
7883 PF_ACPY(&key.ext_lan.addr, pd2.src, key.af_lan);
7884 key.lan.xport.port = th.th_dport;
7885 key.ext_lan.xport.port = th.th_sport;
7886
7887 STATE_LOOKUP();
7888
7889 sk = (*state)->state_key;
7890 if ((direction == sk->direction) &&
7891 ((sk->af_lan == sk->af_gwy) ||
7892 (pd2.af == sk->af_lan))) {
7893 src = &(*state)->dst;
7894 dst = &(*state)->src;
7895 } else {
7896 src = &(*state)->src;
7897 dst = &(*state)->dst;
7898 }
7899
7900 if (src->wscale && (dst->wscale & PF_WSCALE_FLAG)) {
7901 dws = dst->wscale & PF_WSCALE_MASK;
7902 } else {
7903 dws = TCP_MAX_WINSHIFT;
7904 }
7905
7906 /* Demodulate sequence number */
7907 seq = ntohl(th.th_seq) - src->seqdiff;
7908 if (src->seqdiff) {
7909 pf_change_a(&th.th_seq, icmpsum,
7910 htonl(seq), 0);
7911 copyback = 1;
7912 }
7913
7914 if (!SEQ_GEQ(src->seqhi, seq) ||
7915 !SEQ_GEQ(seq,
7916 src->seqlo - ((u_int32_t)dst->max_win << dws))) {
7917 if (pf_status.debug >= PF_DEBUG_MISC) {
7918 printf("pf: BAD ICMP %d:%d ",
7919 icmptype, pd->hdr.icmp->icmp_code);
7920 pf_print_host(pd->src, 0, pd->af);
7921 printf(" -> ");
7922 pf_print_host(pd->dst, 0, pd->af);
7923 printf(" state: ");
7924 pf_print_state(*state);
7925 printf(" seq=%u\n", seq);
7926 }
7927 REASON_SET(reason, PFRES_BADSTATE);
7928 return PF_DROP;
7929 }
7930
7931 pd->naf = pd2.naf = (pd2.af == sk->af_lan) ?
7932 sk->af_gwy : sk->af_lan;
7933
7934 if (STATE_TRANSLATE(sk)) {
7935 /* NAT64 case */
7936 if (sk->af_lan != sk->af_gwy) {
7937 struct pf_state_host *saddr2, *daddr2;
7938
7939 if (pd2.naf == sk->af_lan) {
7940 saddr2 = &sk->lan;
7941 daddr2 = &sk->ext_lan;
7942 } else {
7943 saddr2 = &sk->ext_gwy;
7944 daddr2 = &sk->gwy;
7945 }
7946
7947 /* translate ICMP message types and codes */
7948 if (pf_translate_icmp_af(pd->naf,
7949 pd->hdr.icmp)) {
7950 return PF_DROP;
7951 }
7952
7953 if (pf_lazy_makewritable(pd, pbuf,
7954 off2 + 8) == NULL) {
7955 return PF_DROP;
7956 }
7957
7958 pbuf_copy_back(pbuf, pd->off,
7959 sizeof(struct icmp6_hdr),
7960 pd->hdr.icmp6);
7961
7962 /*
7963 * translate inner ip header within the
7964 * ICMP message
7965 */
7966 if (pf_change_icmp_af(pbuf, ipoff2, pd,
7967 &pd2, &saddr2->addr, &daddr2->addr,
7968 pd->af, pd->naf)) {
7969 return PF_DROP;
7970 }
7971
7972 if (pd->naf == AF_INET) {
7973 pd->proto = IPPROTO_ICMP;
7974 } else {
7975 pd->proto = IPPROTO_ICMPV6;
7976 }
7977
7978 /*
7979 * translate inner tcp header within
7980 * the ICMP message
7981 */
7982 pf_change_ap(direction, NULL, pd2.src,
7983 &th.th_sport, pd2.ip_sum,
7984 &th.th_sum, &daddr2->addr,
7985 saddr2->xport.port, 0, pd2.af,
7986 pd2.naf, 0);
7987
7988 pf_change_ap(direction, NULL, pd2.dst,
7989 &th.th_dport, pd2.ip_sum,
7990 &th.th_sum, &saddr2->addr,
7991 daddr2->xport.port, 0, pd2.af,
7992 pd2.naf, 0);
7993
7994 pbuf_copy_back(pbuf, pd2.off, 8, &th);
7995
7996 /* translate outer ip header */
7997 PF_ACPY(&pd->naddr, &daddr2->addr,
7998 pd->naf);
7999 PF_ACPY(&pd->ndaddr, &saddr2->addr,
8000 pd->naf);
8001 if (pd->af == AF_INET) {
8002 memcpy(&pd->naddr.addr32[3],
8003 &srcv4_inaddr,
8004 sizeof(pd->naddr.addr32[3]));
8005 return pf_nat64_ipv4(pbuf, off,
8006 pd);
8007 } else {
8008 return pf_nat64_ipv6(pbuf, off,
8009 pd);
8010 }
8011 }
8012 if (direction == PF_IN) {
8013 pf_change_icmp(pd2.src, &th.th_sport,
8014 daddr, &sk->lan.addr,
8015 sk->lan.xport.port, NULL,
8016 pd2.ip_sum, icmpsum,
8017 pd->ip_sum, 0, pd2.af);
8018 } else {
8019 pf_change_icmp(pd2.dst, &th.th_dport,
8020 saddr, &sk->gwy.addr,
8021 sk->gwy.xport.port, NULL,
8022 pd2.ip_sum, icmpsum,
8023 pd->ip_sum, 0, pd2.af);
8024 }
8025 copyback = 1;
8026 }
8027
8028 if (copyback) {
8029 if (pf_lazy_makewritable(pd, pbuf, off2 + 8) ==
8030 NULL) {
8031 return PF_DROP;
8032 }
8033 switch (pd2.af) {
8034 #if INET
8035 case AF_INET:
8036 pbuf_copy_back(pbuf, off, ICMP_MINLEN,
8037 pd->hdr.icmp);
8038 pbuf_copy_back(pbuf, ipoff2, sizeof(h2),
8039 &h2);
8040 break;
8041 #endif /* INET */
8042 case AF_INET6:
8043 pbuf_copy_back(pbuf, off,
8044 sizeof(struct icmp6_hdr),
8045 pd->hdr.icmp6);
8046 pbuf_copy_back(pbuf, ipoff2,
8047 sizeof(h2_6), &h2_6);
8048 break;
8049 }
8050 pbuf_copy_back(pbuf, off2, 8, &th);
8051 }
8052
8053 return PF_PASS;
8054 }
8055 case IPPROTO_UDP: {
8056 struct udphdr uh;
8057 int dx, action;
8058 if (!pf_pull_hdr(pbuf, off2, &uh, sizeof(uh),
8059 NULL, reason, pd2.af)) {
8060 DPFPRINTF(PF_DEBUG_MISC,
8061 ("pf: ICMP error message too short "
8062 "(udp)\n"));
8063 return PF_DROP;
8064 }
8065
8066 key.af_gwy = pd2.af;
8067 PF_ACPY(&key.ext_gwy.addr, pd2.dst, key.af_gwy);
8068 PF_ACPY(&key.gwy.addr, pd2.src, key.af_gwy);
8069 key.ext_gwy.xport.port = uh.uh_dport;
8070 key.gwy.xport.port = uh.uh_sport;
8071
8072 key.af_lan = pd2.af;
8073 PF_ACPY(&key.lan.addr, pd2.dst, key.af_lan);
8074 PF_ACPY(&key.ext_lan.addr, pd2.src, key.af_lan);
8075 key.lan.xport.port = uh.uh_dport;
8076 key.ext_lan.xport.port = uh.uh_sport;
8077
8078 key.proto = IPPROTO_UDP;
8079 key.proto_variant = PF_EXTFILTER_APD;
8080 dx = direction;
8081
8082 if (ntohs(uh.uh_sport) == PF_IKE_PORT &&
8083 ntohs(uh.uh_dport) == PF_IKE_PORT) {
8084 struct pf_ike_hdr ike;
8085 size_t plen = pbuf->pb_packet_len - off2 -
8086 sizeof(uh);
8087 if (direction == PF_IN &&
8088 plen < 8 /* PF_IKE_PACKET_MINSIZE */) {
8089 DPFPRINTF(PF_DEBUG_MISC, ("pf: "
8090 "ICMP error, embedded IKE message "
8091 "too small.\n"));
8092 return PF_DROP;
8093 }
8094
8095 if (plen > sizeof(ike)) {
8096 plen = sizeof(ike);
8097 }
8098 pbuf_copy_data(pbuf, off + sizeof(uh), plen,
8099 &ike);
8100
8101 key.app_state = &as;
8102 as.compare_lan_ext = pf_ike_compare;
8103 as.compare_ext_gwy = pf_ike_compare;
8104 as.u.ike.cookie = ike.initiator_cookie;
8105 }
8106
8107 *state = pf_find_state(kif, &key, dx);
8108
8109 if (key.app_state && *state == 0) {
8110 key.app_state = 0;
8111 *state = pf_find_state(kif, &key, dx);
8112 }
8113
8114 if (*state == 0) {
8115 key.proto_variant = PF_EXTFILTER_AD;
8116 *state = pf_find_state(kif, &key, dx);
8117 }
8118
8119 if (*state == 0) {
8120 key.proto_variant = PF_EXTFILTER_EI;
8121 *state = pf_find_state(kif, &key, dx);
8122 }
8123
8124 /* similar to STATE_LOOKUP() */
8125 if (*state != NULL && pd != NULL &&
8126 !(pd->pktflags & PKTF_FLOW_ID)) {
8127 pd->flowsrc = (*state)->state_key->flowsrc;
8128 pd->flowhash = (*state)->state_key->flowhash;
8129 if (pd->flowhash != 0) {
8130 pd->pktflags |= PKTF_FLOW_ID;
8131 pd->pktflags &= ~PKTF_FLOW_ADV;
8132 }
8133 }
8134
8135 if (pf_state_lookup_aux(state, kif, direction, &action)) {
8136 return action;
8137 }
8138
8139 sk = (*state)->state_key;
8140 pd->naf = pd2.naf = (pd2.af == sk->af_lan) ?
8141 sk->af_gwy : sk->af_lan;
8142
8143 if (STATE_TRANSLATE(sk)) {
8144 /* NAT64 case */
8145 if (sk->af_lan != sk->af_gwy) {
8146 struct pf_state_host *saddr2, *daddr2;
8147
8148 if (pd2.naf == sk->af_lan) {
8149 saddr2 = &sk->lan;
8150 daddr2 = &sk->ext_lan;
8151 } else {
8152 saddr2 = &sk->ext_gwy;
8153 daddr2 = &sk->gwy;
8154 }
8155
8156 /* translate ICMP message */
8157 if (pf_translate_icmp_af(pd->naf,
8158 pd->hdr.icmp)) {
8159 return PF_DROP;
8160 }
8161 if (pf_lazy_makewritable(pd, pbuf,
8162 off2 + 8) == NULL) {
8163 return PF_DROP;
8164 }
8165
8166 pbuf_copy_back(pbuf, pd->off,
8167 sizeof(struct icmp6_hdr),
8168 pd->hdr.icmp6);
8169
8170 /*
8171 * translate inner ip header within the
8172 * ICMP message
8173 */
8174 if (pf_change_icmp_af(pbuf, ipoff2, pd,
8175 &pd2, &saddr2->addr, &daddr2->addr,
8176 pd->af, pd->naf)) {
8177 return PF_DROP;
8178 }
8179
8180 if (pd->naf == AF_INET) {
8181 pd->proto = IPPROTO_ICMP;
8182 } else {
8183 pd->proto = IPPROTO_ICMPV6;
8184 }
8185
8186 /*
8187 * translate inner udp header within
8188 * the ICMP message
8189 */
8190 pf_change_ap(direction, NULL, pd2.src,
8191 &uh.uh_sport, pd2.ip_sum,
8192 &uh.uh_sum, &daddr2->addr,
8193 saddr2->xport.port, 0, pd2.af,
8194 pd2.naf, 0);
8195
8196 pf_change_ap(direction, NULL, pd2.dst,
8197 &uh.uh_dport, pd2.ip_sum,
8198 &uh.uh_sum, &saddr2->addr,
8199 daddr2->xport.port, 0, pd2.af,
8200 pd2.naf, 0);
8201
8202 pbuf_copy_back(pbuf, pd2.off,
8203 sizeof(uh), &uh);
8204
8205 /* translate outer ip header */
8206 PF_ACPY(&pd->naddr, &daddr2->addr,
8207 pd->naf);
8208 PF_ACPY(&pd->ndaddr, &saddr2->addr,
8209 pd->naf);
8210 if (pd->af == AF_INET) {
8211 memcpy(&pd->naddr.addr32[3],
8212 &srcv4_inaddr,
8213 sizeof(pd->naddr.addr32[3]));
8214 return pf_nat64_ipv4(pbuf, off,
8215 pd);
8216 } else {
8217 return pf_nat64_ipv6(pbuf, off,
8218 pd);
8219 }
8220 }
8221 if (direction == PF_IN) {
8222 pf_change_icmp(pd2.src, &uh.uh_sport,
8223 daddr, &sk->lan.addr,
8224 sk->lan.xport.port, &uh.uh_sum,
8225 pd2.ip_sum, icmpsum,
8226 pd->ip_sum, 1, pd2.af);
8227 } else {
8228 pf_change_icmp(pd2.dst, &uh.uh_dport,
8229 saddr, &sk->gwy.addr,
8230 sk->gwy.xport.port, &uh.uh_sum,
8231 pd2.ip_sum, icmpsum,
8232 pd->ip_sum, 1, pd2.af);
8233 }
8234 if (pf_lazy_makewritable(pd, pbuf,
8235 off2 + sizeof(uh)) == NULL) {
8236 return PF_DROP;
8237 }
8238 switch (pd2.af) {
8239 #if INET
8240 case AF_INET:
8241 pbuf_copy_back(pbuf, off, ICMP_MINLEN,
8242 pd->hdr.icmp);
8243 pbuf_copy_back(pbuf, ipoff2,
8244 sizeof(h2), &h2);
8245 break;
8246 #endif /* INET */
8247 case AF_INET6:
8248 pbuf_copy_back(pbuf, off,
8249 sizeof(struct icmp6_hdr),
8250 pd->hdr.icmp6);
8251 pbuf_copy_back(pbuf, ipoff2,
8252 sizeof(h2_6), &h2_6);
8253 break;
8254 }
8255 pbuf_copy_back(pbuf, off2, sizeof(uh), &uh);
8256 }
8257
8258 return PF_PASS;
8259 }
8260 #if INET
8261 case IPPROTO_ICMP: {
8262 struct icmp iih;
8263
8264 if (!pf_pull_hdr(pbuf, off2, &iih, ICMP_MINLEN,
8265 NULL, reason, pd2.af)) {
8266 DPFPRINTF(PF_DEBUG_MISC,
8267 ("pf: ICMP error message too short i"
8268 "(icmp)\n"));
8269 return PF_DROP;
8270 }
8271
8272 key.proto = IPPROTO_ICMP;
8273 if (direction == PF_IN) {
8274 key.af_gwy = pd2.af;
8275 PF_ACPY(&key.ext_gwy.addr, pd2.dst, key.af_gwy);
8276 PF_ACPY(&key.gwy.addr, pd2.src, key.af_gwy);
8277 key.ext_gwy.xport.port = 0;
8278 key.gwy.xport.port = iih.icmp_id;
8279 } else {
8280 key.af_lan = pd2.af;
8281 PF_ACPY(&key.lan.addr, pd2.dst, key.af_lan);
8282 PF_ACPY(&key.ext_lan.addr, pd2.src, key.af_lan);
8283 key.lan.xport.port = iih.icmp_id;
8284 key.ext_lan.xport.port = 0;
8285 }
8286
8287 STATE_LOOKUP();
8288
8289 sk = (*state)->state_key;
8290 if (STATE_TRANSLATE(sk)) {
8291 if (direction == PF_IN) {
8292 pf_change_icmp(pd2.src, &iih.icmp_id,
8293 daddr, &sk->lan.addr,
8294 sk->lan.xport.port, NULL,
8295 pd2.ip_sum, icmpsum,
8296 pd->ip_sum, 0, AF_INET);
8297 } else {
8298 pf_change_icmp(pd2.dst, &iih.icmp_id,
8299 saddr, &sk->gwy.addr,
8300 sk->gwy.xport.port, NULL,
8301 pd2.ip_sum, icmpsum,
8302 pd->ip_sum, 0, AF_INET);
8303 }
8304 if (pf_lazy_makewritable(pd, pbuf,
8305 off2 + ICMP_MINLEN) == NULL) {
8306 return PF_DROP;
8307 }
8308 pbuf_copy_back(pbuf, off, ICMP_MINLEN,
8309 pd->hdr.icmp);
8310 pbuf_copy_back(pbuf, ipoff2, sizeof(h2), &h2);
8311 pbuf_copy_back(pbuf, off2, ICMP_MINLEN, &iih);
8312 }
8313
8314 return PF_PASS;
8315 }
8316 #endif /* INET */
8317 case IPPROTO_ICMPV6: {
8318 struct icmp6_hdr iih;
8319
8320 if (!pf_pull_hdr(pbuf, off2, &iih,
8321 sizeof(struct icmp6_hdr), NULL, reason, pd2.af)) {
8322 DPFPRINTF(PF_DEBUG_MISC,
8323 ("pf: ICMP error message too short "
8324 "(icmp6)\n"));
8325 return PF_DROP;
8326 }
8327
8328 key.proto = IPPROTO_ICMPV6;
8329 if (direction == PF_IN) {
8330 key.af_gwy = pd2.af;
8331 PF_ACPY(&key.ext_gwy.addr, pd2.dst, key.af_gwy);
8332 PF_ACPY(&key.gwy.addr, pd2.src, key.af_gwy);
8333 key.ext_gwy.xport.port = 0;
8334 key.gwy.xport.port = iih.icmp6_id;
8335 } else {
8336 key.af_lan = pd2.af;
8337 PF_ACPY(&key.lan.addr, pd2.dst, key.af_lan);
8338 PF_ACPY(&key.ext_lan.addr, pd2.src, key.af_lan);
8339 key.lan.xport.port = iih.icmp6_id;
8340 key.ext_lan.xport.port = 0;
8341 }
8342
8343 STATE_LOOKUP();
8344
8345 sk = (*state)->state_key;
8346 if (STATE_TRANSLATE(sk)) {
8347 if (direction == PF_IN) {
8348 pf_change_icmp(pd2.src, &iih.icmp6_id,
8349 daddr, &sk->lan.addr,
8350 sk->lan.xport.port, NULL,
8351 pd2.ip_sum, icmpsum,
8352 pd->ip_sum, 0, AF_INET6);
8353 } else {
8354 pf_change_icmp(pd2.dst, &iih.icmp6_id,
8355 saddr, &sk->gwy.addr,
8356 sk->gwy.xport.port, NULL,
8357 pd2.ip_sum, icmpsum,
8358 pd->ip_sum, 0, AF_INET6);
8359 }
8360 if (pf_lazy_makewritable(pd, pbuf, off2 +
8361 sizeof(struct icmp6_hdr)) == NULL) {
8362 return PF_DROP;
8363 }
8364 pbuf_copy_back(pbuf, off,
8365 sizeof(struct icmp6_hdr), pd->hdr.icmp6);
8366 pbuf_copy_back(pbuf, ipoff2, sizeof(h2_6),
8367 &h2_6);
8368 pbuf_copy_back(pbuf, off2,
8369 sizeof(struct icmp6_hdr), &iih);
8370 }
8371
8372 return PF_PASS;
8373 }
8374 default: {
8375 key.proto = pd2.proto;
8376 if (direction == PF_IN) {
8377 key.af_gwy = pd2.af;
8378 PF_ACPY(&key.ext_gwy.addr, pd2.dst, key.af_gwy);
8379 PF_ACPY(&key.gwy.addr, pd2.src, key.af_gwy);
8380 key.ext_gwy.xport.port = 0;
8381 key.gwy.xport.port = 0;
8382 } else {
8383 key.af_lan = pd2.af;
8384 PF_ACPY(&key.lan.addr, pd2.dst, key.af_lan);
8385 PF_ACPY(&key.ext_lan.addr, pd2.src, key.af_lan);
8386 key.lan.xport.port = 0;
8387 key.ext_lan.xport.port = 0;
8388 }
8389
8390 STATE_LOOKUP();
8391
8392 sk = (*state)->state_key;
8393 if (STATE_TRANSLATE(sk)) {
8394 if (direction == PF_IN) {
8395 pf_change_icmp(pd2.src, NULL, daddr,
8396 &sk->lan.addr, 0, NULL,
8397 pd2.ip_sum, icmpsum,
8398 pd->ip_sum, 0, pd2.af);
8399 } else {
8400 pf_change_icmp(pd2.dst, NULL, saddr,
8401 &sk->gwy.addr, 0, NULL,
8402 pd2.ip_sum, icmpsum,
8403 pd->ip_sum, 0, pd2.af);
8404 }
8405 switch (pd2.af) {
8406 #if INET
8407 case AF_INET:
8408 if (pf_lazy_makewritable(pd, pbuf,
8409 ipoff2 + sizeof(h2)) == NULL) {
8410 return PF_DROP;
8411 }
8412 /*
8413 * <XXXSCW>
8414 * Xnu was missing the following...
8415 */
8416 pbuf_copy_back(pbuf, off, ICMP_MINLEN,
8417 pd->hdr.icmp);
8418 pbuf_copy_back(pbuf, ipoff2,
8419 sizeof(h2), &h2);
8420 break;
8421 /*
8422 * </XXXSCW>
8423 */
8424 #endif /* INET */
8425 case AF_INET6:
8426 if (pf_lazy_makewritable(pd, pbuf,
8427 ipoff2 + sizeof(h2_6)) == NULL) {
8428 return PF_DROP;
8429 }
8430 pbuf_copy_back(pbuf, off,
8431 sizeof(struct icmp6_hdr),
8432 pd->hdr.icmp6);
8433 pbuf_copy_back(pbuf, ipoff2,
8434 sizeof(h2_6), &h2_6);
8435 break;
8436 }
8437 }
8438
8439 return PF_PASS;
8440 }
8441 }
8442 }
8443 }
8444
8445 static int
8446 pf_test_state_grev1(struct pf_state **state, int direction,
8447 struct pfi_kif *kif, int off, struct pf_pdesc *pd)
8448 {
8449 struct pf_state_peer *src;
8450 struct pf_state_peer *dst;
8451 struct pf_state_key_cmp key;
8452 struct pf_grev1_hdr *grev1 = pd->hdr.grev1;
8453
8454 key.app_state = 0;
8455 key.proto = IPPROTO_GRE;
8456 key.proto_variant = PF_GRE_PPTP_VARIANT;
8457 if (direction == PF_IN) {
8458 key.af_gwy = pd->af;
8459 PF_ACPY(&key.ext_gwy.addr, pd->src, key.af_gwy);
8460 PF_ACPY(&key.gwy.addr, pd->dst, key.af_gwy);
8461 key.gwy.xport.call_id = grev1->call_id;
8462 } else {
8463 key.af_lan = pd->af;
8464 PF_ACPY(&key.lan.addr, pd->src, key.af_lan);
8465 PF_ACPY(&key.ext_lan.addr, pd->dst, key.af_lan);
8466 key.ext_lan.xport.call_id = grev1->call_id;
8467 }
8468
8469 STATE_LOOKUP();
8470
8471 if (direction == (*state)->state_key->direction) {
8472 src = &(*state)->src;
8473 dst = &(*state)->dst;
8474 } else {
8475 src = &(*state)->dst;
8476 dst = &(*state)->src;
8477 }
8478
8479 /* update states */
8480 if (src->state < PFGRE1S_INITIATING) {
8481 src->state = PFGRE1S_INITIATING;
8482 }
8483
8484 /* update expire time */
8485 (*state)->expire = pf_time_second();
8486 if (src->state >= PFGRE1S_INITIATING &&
8487 dst->state >= PFGRE1S_INITIATING) {
8488 if ((*state)->timeout != PFTM_TCP_ESTABLISHED) {
8489 (*state)->timeout = PFTM_GREv1_ESTABLISHED;
8490 }
8491 src->state = PFGRE1S_ESTABLISHED;
8492 dst->state = PFGRE1S_ESTABLISHED;
8493 } else {
8494 (*state)->timeout = PFTM_GREv1_INITIATING;
8495 }
8496
8497 if ((*state)->state_key->app_state) {
8498 (*state)->state_key->app_state->u.grev1.pptp_state->expire =
8499 pf_time_second();
8500 }
8501
8502 /* translate source/destination address, if necessary */
8503 if (STATE_GRE_TRANSLATE((*state)->state_key)) {
8504 if (direction == PF_OUT) {
8505 switch (pd->af) {
8506 #if INET
8507 case AF_INET:
8508 pf_change_a(&pd->src->v4addr.s_addr,
8509 pd->ip_sum,
8510 (*state)->state_key->gwy.addr.v4addr.s_addr, 0);
8511 break;
8512 #endif /* INET */
8513 case AF_INET6:
8514 PF_ACPY(pd->src, &(*state)->state_key->gwy.addr,
8515 pd->af);
8516 break;
8517 }
8518 } else {
8519 grev1->call_id = (*state)->state_key->lan.xport.call_id;
8520
8521 switch (pd->af) {
8522 #if INET
8523 case AF_INET:
8524 pf_change_a(&pd->dst->v4addr.s_addr,
8525 pd->ip_sum,
8526 (*state)->state_key->lan.addr.v4addr.s_addr, 0);
8527 break;
8528 #endif /* INET */
8529 case AF_INET6:
8530 PF_ACPY(pd->dst, &(*state)->state_key->lan.addr,
8531 pd->af);
8532 break;
8533 }
8534 }
8535
8536 if (pf_lazy_makewritable(pd, pd->mp, off + sizeof(*grev1)) ==
8537 NULL) {
8538 return PF_DROP;
8539 }
8540 pbuf_copy_back(pd->mp, off, sizeof(*grev1), grev1);
8541 }
8542
8543 return PF_PASS;
8544 }
8545
8546 static int
8547 pf_test_state_esp(struct pf_state **state, int direction, struct pfi_kif *kif,
8548 int off, struct pf_pdesc *pd)
8549 {
8550 #pragma unused(off)
8551 struct pf_state_peer *src;
8552 struct pf_state_peer *dst;
8553 struct pf_state_key_cmp key;
8554 struct pf_esp_hdr *esp = pd->hdr.esp;
8555 int action;
8556
8557 memset(&key, 0, sizeof(key));
8558 key.proto = IPPROTO_ESP;
8559 if (direction == PF_IN) {
8560 key.af_gwy = pd->af;
8561 PF_ACPY(&key.ext_gwy.addr, pd->src, key.af_gwy);
8562 PF_ACPY(&key.gwy.addr, pd->dst, key.af_gwy);
8563 key.gwy.xport.spi = esp->spi;
8564 } else {
8565 key.af_lan = pd->af;
8566 PF_ACPY(&key.lan.addr, pd->src, key.af_lan);
8567 PF_ACPY(&key.ext_lan.addr, pd->dst, key.af_lan);
8568 key.ext_lan.xport.spi = esp->spi;
8569 }
8570
8571 *state = pf_find_state(kif, &key, direction);
8572
8573 if (*state == 0) {
8574 struct pf_state *s;
8575
8576 /*
8577 * <jhw@apple.com>
8578 * No matching state. Look for a blocking state. If we find
8579 * one, then use that state and move it so that it's keyed to
8580 * the SPI in the current packet.
8581 */
8582 if (direction == PF_IN) {
8583 key.gwy.xport.spi = 0;
8584
8585 s = pf_find_state(kif, &key, direction);
8586 if (s) {
8587 struct pf_state_key *sk = s->state_key;
8588
8589 RB_REMOVE(pf_state_tree_ext_gwy,
8590 &pf_statetbl_ext_gwy, sk);
8591 sk->lan.xport.spi = sk->gwy.xport.spi =
8592 esp->spi;
8593
8594 if (RB_INSERT(pf_state_tree_ext_gwy,
8595 &pf_statetbl_ext_gwy, sk)) {
8596 pf_detach_state(s, PF_DT_SKIP_EXTGWY);
8597 } else {
8598 *state = s;
8599 }
8600 }
8601 } else {
8602 key.ext_lan.xport.spi = 0;
8603
8604 s = pf_find_state(kif, &key, direction);
8605 if (s) {
8606 struct pf_state_key *sk = s->state_key;
8607
8608 RB_REMOVE(pf_state_tree_lan_ext,
8609 &pf_statetbl_lan_ext, sk);
8610 sk->ext_lan.xport.spi = esp->spi;
8611
8612 if (RB_INSERT(pf_state_tree_lan_ext,
8613 &pf_statetbl_lan_ext, sk)) {
8614 pf_detach_state(s, PF_DT_SKIP_LANEXT);
8615 } else {
8616 *state = s;
8617 }
8618 }
8619 }
8620
8621 if (s) {
8622 if (*state == 0) {
8623 #if NPFSYNC
8624 if (s->creatorid == pf_status.hostid) {
8625 pfsync_delete_state(s);
8626 }
8627 #endif
8628 s->timeout = PFTM_UNLINKED;
8629 hook_runloop(&s->unlink_hooks,
8630 HOOK_REMOVE | HOOK_FREE);
8631 pf_src_tree_remove_state(s);
8632 pf_free_state(s);
8633 return PF_DROP;
8634 }
8635 }
8636 }
8637
8638 /* similar to STATE_LOOKUP() */
8639 if (*state != NULL && pd != NULL && !(pd->pktflags & PKTF_FLOW_ID)) {
8640 pd->flowsrc = (*state)->state_key->flowsrc;
8641 pd->flowhash = (*state)->state_key->flowhash;
8642 if (pd->flowhash != 0) {
8643 pd->pktflags |= PKTF_FLOW_ID;
8644 pd->pktflags &= ~PKTF_FLOW_ADV;
8645 }
8646 }
8647
8648 if (pf_state_lookup_aux(state, kif, direction, &action)) {
8649 return action;
8650 }
8651
8652 if (direction == (*state)->state_key->direction) {
8653 src = &(*state)->src;
8654 dst = &(*state)->dst;
8655 } else {
8656 src = &(*state)->dst;
8657 dst = &(*state)->src;
8658 }
8659
8660 /* update states */
8661 if (src->state < PFESPS_INITIATING) {
8662 src->state = PFESPS_INITIATING;
8663 }
8664
8665 /* update expire time */
8666 (*state)->expire = pf_time_second();
8667 if (src->state >= PFESPS_INITIATING &&
8668 dst->state >= PFESPS_INITIATING) {
8669 (*state)->timeout = PFTM_ESP_ESTABLISHED;
8670 src->state = PFESPS_ESTABLISHED;
8671 dst->state = PFESPS_ESTABLISHED;
8672 } else {
8673 (*state)->timeout = PFTM_ESP_INITIATING;
8674 }
8675 /* translate source/destination address, if necessary */
8676 if (STATE_ADDR_TRANSLATE((*state)->state_key)) {
8677 if (direction == PF_OUT) {
8678 switch (pd->af) {
8679 #if INET
8680 case AF_INET:
8681 pf_change_a(&pd->src->v4addr.s_addr,
8682 pd->ip_sum,
8683 (*state)->state_key->gwy.addr.v4addr.s_addr, 0);
8684 break;
8685 #endif /* INET */
8686 case AF_INET6:
8687 PF_ACPY(pd->src, &(*state)->state_key->gwy.addr,
8688 pd->af);
8689 break;
8690 }
8691 } else {
8692 switch (pd->af) {
8693 #if INET
8694 case AF_INET:
8695 pf_change_a(&pd->dst->v4addr.s_addr,
8696 pd->ip_sum,
8697 (*state)->state_key->lan.addr.v4addr.s_addr, 0);
8698 break;
8699 #endif /* INET */
8700 case AF_INET6:
8701 PF_ACPY(pd->dst, &(*state)->state_key->lan.addr,
8702 pd->af);
8703 break;
8704 }
8705 }
8706 }
8707
8708 return PF_PASS;
8709 }
8710
8711 static int
8712 pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif,
8713 struct pf_pdesc *pd)
8714 {
8715 struct pf_state_peer *src, *dst;
8716 struct pf_state_key_cmp key;
8717
8718 key.app_state = 0;
8719 key.proto = pd->proto;
8720 if (direction == PF_IN) {
8721 key.af_gwy = pd->af;
8722 PF_ACPY(&key.ext_gwy.addr, pd->src, key.af_gwy);
8723 PF_ACPY(&key.gwy.addr, pd->dst, key.af_gwy);
8724 key.ext_gwy.xport.port = 0;
8725 key.gwy.xport.port = 0;
8726 } else {
8727 key.af_lan = pd->af;
8728 PF_ACPY(&key.lan.addr, pd->src, key.af_lan);
8729 PF_ACPY(&key.ext_lan.addr, pd->dst, key.af_lan);
8730 key.lan.xport.port = 0;
8731 key.ext_lan.xport.port = 0;
8732 }
8733
8734 STATE_LOOKUP();
8735
8736 if (direction == (*state)->state_key->direction) {
8737 src = &(*state)->src;
8738 dst = &(*state)->dst;
8739 } else {
8740 src = &(*state)->dst;
8741 dst = &(*state)->src;
8742 }
8743
8744 /* update states */
8745 if (src->state < PFOTHERS_SINGLE) {
8746 src->state = PFOTHERS_SINGLE;
8747 }
8748 if (dst->state == PFOTHERS_SINGLE) {
8749 dst->state = PFOTHERS_MULTIPLE;
8750 }
8751
8752 /* update expire time */
8753 (*state)->expire = pf_time_second();
8754 if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE) {
8755 (*state)->timeout = PFTM_OTHER_MULTIPLE;
8756 } else {
8757 (*state)->timeout = PFTM_OTHER_SINGLE;
8758 }
8759
8760 /* translate source/destination address, if necessary */
8761 if (STATE_ADDR_TRANSLATE((*state)->state_key)) {
8762 if (direction == PF_OUT) {
8763 switch (pd->af) {
8764 #if INET
8765 case AF_INET:
8766 pf_change_a(&pd->src->v4addr.s_addr,
8767 pd->ip_sum,
8768 (*state)->state_key->gwy.addr.v4addr.s_addr,
8769 0);
8770 break;
8771 #endif /* INET */
8772 case AF_INET6:
8773 PF_ACPY(pd->src,
8774 &(*state)->state_key->gwy.addr, pd->af);
8775 break;
8776 }
8777 } else {
8778 switch (pd->af) {
8779 #if INET
8780 case AF_INET:
8781 pf_change_a(&pd->dst->v4addr.s_addr,
8782 pd->ip_sum,
8783 (*state)->state_key->lan.addr.v4addr.s_addr,
8784 0);
8785 break;
8786 #endif /* INET */
8787 case AF_INET6:
8788 PF_ACPY(pd->dst,
8789 &(*state)->state_key->lan.addr, pd->af);
8790 break;
8791 }
8792 }
8793 }
8794
8795 return PF_PASS;
8796 }
8797
8798 /*
8799 * ipoff and off are measured from the start of the mbuf chain.
8800 * h must be at "ipoff" on the mbuf chain.
8801 */
8802 void *
8803 pf_pull_hdr(pbuf_t *pbuf, int off, void *p, int len,
8804 u_short *actionp, u_short *reasonp, sa_family_t af)
8805 {
8806 switch (af) {
8807 #if INET
8808 case AF_INET: {
8809 struct ip *h = pbuf->pb_data;
8810 u_int16_t fragoff = (ntohs(h->ip_off) & IP_OFFMASK) << 3;
8811
8812 if (fragoff) {
8813 if (fragoff >= len) {
8814 ACTION_SET(actionp, PF_PASS);
8815 } else {
8816 ACTION_SET(actionp, PF_DROP);
8817 REASON_SET(reasonp, PFRES_FRAG);
8818 }
8819 return NULL;
8820 }
8821 if (pbuf->pb_packet_len < (unsigned)(off + len) ||
8822 ntohs(h->ip_len) < off + len) {
8823 ACTION_SET(actionp, PF_DROP);
8824 REASON_SET(reasonp, PFRES_SHORT);
8825 return NULL;
8826 }
8827 break;
8828 }
8829 #endif /* INET */
8830 case AF_INET6: {
8831 struct ip6_hdr *h = pbuf->pb_data;
8832
8833 if (pbuf->pb_packet_len < (unsigned)(off + len) ||
8834 (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) <
8835 (unsigned)(off + len)) {
8836 ACTION_SET(actionp, PF_DROP);
8837 REASON_SET(reasonp, PFRES_SHORT);
8838 return NULL;
8839 }
8840 break;
8841 }
8842 }
8843 pbuf_copy_data(pbuf, off, len, p);
8844 return p;
8845 }
8846
8847 int
8848 pf_routable(struct pf_addr *addr, sa_family_t af, struct pfi_kif *kif)
8849 {
8850 #pragma unused(kif)
8851 struct sockaddr_in *dst;
8852 int ret = 1;
8853 struct sockaddr_in6 *dst6;
8854 struct route_in6 ro;
8855
8856 bzero(&ro, sizeof(ro));
8857 switch (af) {
8858 case AF_INET:
8859 dst = satosin(&ro.ro_dst);
8860 dst->sin_family = AF_INET;
8861 dst->sin_len = sizeof(*dst);
8862 dst->sin_addr = addr->v4addr;
8863 break;
8864 case AF_INET6:
8865 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
8866 dst6->sin6_family = AF_INET6;
8867 dst6->sin6_len = sizeof(*dst6);
8868 dst6->sin6_addr = addr->v6addr;
8869 break;
8870 default:
8871 return 0;
8872 }
8873
8874 /* XXX: IFT_ENC is not currently used by anything*/
8875 /* Skip checks for ipsec interfaces */
8876 if (kif != NULL && kif->pfik_ifp->if_type == IFT_ENC) {
8877 goto out;
8878 }
8879
8880 /* XXX: what is the point of this? */
8881 rtalloc((struct route *)&ro);
8882
8883 out:
8884 ROUTE_RELEASE(&ro);
8885 return ret;
8886 }
8887
8888 int
8889 pf_rtlabel_match(struct pf_addr *addr, sa_family_t af, struct pf_addr_wrap *aw)
8890 {
8891 #pragma unused(aw)
8892 struct sockaddr_in *dst;
8893 struct sockaddr_in6 *dst6;
8894 struct route_in6 ro;
8895 int ret = 0;
8896
8897 bzero(&ro, sizeof(ro));
8898 switch (af) {
8899 case AF_INET:
8900 dst = satosin(&ro.ro_dst);
8901 dst->sin_family = AF_INET;
8902 dst->sin_len = sizeof(*dst);
8903 dst->sin_addr = addr->v4addr;
8904 break;
8905 case AF_INET6:
8906 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
8907 dst6->sin6_family = AF_INET6;
8908 dst6->sin6_len = sizeof(*dst6);
8909 dst6->sin6_addr = addr->v6addr;
8910 break;
8911 default:
8912 return 0;
8913 }
8914
8915 /* XXX: what is the point of this? */
8916 rtalloc((struct route *)&ro);
8917
8918 ROUTE_RELEASE(&ro);
8919
8920 return ret;
8921 }
8922
8923 #if INET
8924 static void
8925 pf_route(pbuf_t **pbufp, struct pf_rule *r, int dir, struct ifnet *oifp,
8926 struct pf_state *s, struct pf_pdesc *pd)
8927 {
8928 #pragma unused(pd)
8929 struct mbuf *m0, *m1;
8930 struct route iproute;
8931 struct route *ro = &iproute;
8932 struct sockaddr_in *dst;
8933 struct ip *ip;
8934 struct ifnet *ifp = NULL;
8935 struct pf_addr naddr;
8936 struct pf_src_node *sn = NULL;
8937 int error = 0;
8938 uint32_t sw_csum;
8939 int interface_mtu = 0;
8940 bzero(&iproute, sizeof(iproute));
8941
8942 if (pbufp == NULL || !pbuf_is_valid(*pbufp) || r == NULL ||
8943 (dir != PF_IN && dir != PF_OUT) || oifp == NULL) {
8944 panic("pf_route: invalid parameters");
8945 }
8946
8947 if (pd->pf_mtag->pftag_routed++ > 3) {
8948 pbuf_destroy(*pbufp);
8949 *pbufp = NULL;
8950 m0 = NULL;
8951 goto bad;
8952 }
8953
8954 /*
8955 * Since this is something of an edge case and may involve the
8956 * host stack (for routing, at least for now), we convert the
8957 * incoming pbuf into an mbuf.
8958 */
8959 if (r->rt == PF_DUPTO) {
8960 m0 = pbuf_clone_to_mbuf(*pbufp);
8961 } else if ((r->rt == PF_REPLYTO) == (r->direction == dir)) {
8962 return;
8963 } else {
8964 /* We're going to consume this packet */
8965 m0 = pbuf_to_mbuf(*pbufp, TRUE);
8966 *pbufp = NULL;
8967 }
8968
8969 if (m0 == NULL) {
8970 goto bad;
8971 }
8972
8973 /* We now have the packet in an mbuf (m0) */
8974
8975 if (m0->m_len < (int)sizeof(struct ip)) {
8976 DPFPRINTF(PF_DEBUG_URGENT,
8977 ("pf_route: packet length < sizeof (struct ip)\n"));
8978 goto bad;
8979 }
8980
8981 ip = mtod(m0, struct ip *);
8982
8983 dst = satosin((void *)&ro->ro_dst);
8984 dst->sin_family = AF_INET;
8985 dst->sin_len = sizeof(*dst);
8986 dst->sin_addr = ip->ip_dst;
8987
8988 if (r->rt == PF_FASTROUTE) {
8989 rtalloc(ro);
8990 if (ro->ro_rt == NULL) {
8991 ipstat.ips_noroute++;
8992 goto bad;
8993 }
8994
8995 ifp = ro->ro_rt->rt_ifp;
8996 RT_LOCK(ro->ro_rt);
8997 ro->ro_rt->rt_use++;
8998
8999 if (ro->ro_rt->rt_flags & RTF_GATEWAY) {
9000 dst = satosin((void *)ro->ro_rt->rt_gateway);
9001 }
9002 RT_UNLOCK(ro->ro_rt);
9003 } else {
9004 if (TAILQ_EMPTY(&r->rpool.list)) {
9005 DPFPRINTF(PF_DEBUG_URGENT,
9006 ("pf_route: TAILQ_EMPTY(&r->rpool.list)\n"));
9007 goto bad;
9008 }
9009 if (s == NULL) {
9010 pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src,
9011 &naddr, NULL, &sn);
9012 if (!PF_AZERO(&naddr, AF_INET)) {
9013 dst->sin_addr.s_addr = naddr.v4addr.s_addr;
9014 }
9015 ifp = r->rpool.cur->kif ?
9016 r->rpool.cur->kif->pfik_ifp : NULL;
9017 } else {
9018 if (!PF_AZERO(&s->rt_addr, AF_INET)) {
9019 dst->sin_addr.s_addr =
9020 s->rt_addr.v4addr.s_addr;
9021 }
9022 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
9023 }
9024 }
9025 if (ifp == NULL) {
9026 goto bad;
9027 }
9028
9029 if (oifp != ifp) {
9030 if (pf_test_mbuf(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) {
9031 goto bad;
9032 } else if (m0 == NULL) {
9033 goto done;
9034 }
9035 if (m0->m_len < (int)sizeof(struct ip)) {
9036 DPFPRINTF(PF_DEBUG_URGENT,
9037 ("pf_route: packet length < sizeof (struct ip)\n"));
9038 goto bad;
9039 }
9040 ip = mtod(m0, struct ip *);
9041 }
9042
9043 /* Catch routing changes wrt. hardware checksumming for TCP or UDP. */
9044 ip_output_checksum(ifp, m0, ((ip->ip_hl) << 2), ntohs(ip->ip_len),
9045 &sw_csum);
9046
9047 interface_mtu = ifp->if_mtu;
9048
9049 if (INTF_ADJUST_MTU_FOR_CLAT46(ifp)) {
9050 interface_mtu = IN6_LINKMTU(ifp);
9051 /* Further adjust the size for CLAT46 expansion */
9052 interface_mtu -= CLAT46_HDR_EXPANSION_OVERHD;
9053 }
9054
9055 if (ntohs(ip->ip_len) <= interface_mtu || TSO_IPV4_OK(ifp, m0) ||
9056 (!(ip->ip_off & htons(IP_DF)) &&
9057 (ifp->if_hwassist & CSUM_FRAGMENT))) {
9058 ip->ip_sum = 0;
9059 if (sw_csum & CSUM_DELAY_IP) {
9060 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
9061 sw_csum &= ~CSUM_DELAY_IP;
9062 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_IP;
9063 }
9064 error = ifnet_output(ifp, PF_INET, m0, ro->ro_rt, sintosa(dst));
9065 goto done;
9066 }
9067
9068 /*
9069 * Too large for interface; fragment if possible.
9070 * Must be able to put at least 8 bytes per fragment.
9071 * Balk when DF bit is set or the interface didn't support TSO.
9072 */
9073 if ((ip->ip_off & htons(IP_DF)) ||
9074 (m0->m_pkthdr.csum_flags & CSUM_TSO_IPV4)) {
9075 ipstat.ips_cantfrag++;
9076 if (r->rt != PF_DUPTO) {
9077 icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0,
9078 interface_mtu);
9079 goto done;
9080 } else {
9081 goto bad;
9082 }
9083 }
9084
9085 m1 = m0;
9086
9087 /* PR-8933605: send ip_len,ip_off to ip_fragment in host byte order */
9088 #if BYTE_ORDER != BIG_ENDIAN
9089 NTOHS(ip->ip_off);
9090 NTOHS(ip->ip_len);
9091 #endif
9092 error = ip_fragment(m0, ifp, interface_mtu, sw_csum);
9093
9094 if (error) {
9095 m0 = NULL;
9096 goto bad;
9097 }
9098
9099 for (m0 = m1; m0; m0 = m1) {
9100 m1 = m0->m_nextpkt;
9101 m0->m_nextpkt = 0;
9102 if (error == 0) {
9103 error = ifnet_output(ifp, PF_INET, m0, ro->ro_rt,
9104 sintosa(dst));
9105 } else {
9106 m_freem(m0);
9107 }
9108 }
9109
9110 if (error == 0) {
9111 ipstat.ips_fragmented++;
9112 }
9113
9114 done:
9115 ROUTE_RELEASE(&iproute);
9116 return;
9117
9118 bad:
9119 if (m0) {
9120 m_freem(m0);
9121 }
9122 goto done;
9123 }
9124 #endif /* INET */
9125
9126 static void
9127 pf_route6(pbuf_t **pbufp, struct pf_rule *r, int dir, struct ifnet *oifp,
9128 struct pf_state *s, struct pf_pdesc *pd)
9129 {
9130 #pragma unused(pd)
9131 struct mbuf *m0;
9132 struct route_in6 ip6route;
9133 struct route_in6 *ro;
9134 struct sockaddr_in6 *dst;
9135 struct ip6_hdr *ip6;
9136 struct ifnet *ifp = NULL;
9137 struct pf_addr naddr;
9138 struct pf_src_node *sn = NULL;
9139 int error = 0;
9140 struct pf_mtag *pf_mtag;
9141
9142 if (pbufp == NULL || !pbuf_is_valid(*pbufp) || r == NULL ||
9143 (dir != PF_IN && dir != PF_OUT) || oifp == NULL) {
9144 panic("pf_route6: invalid parameters");
9145 }
9146
9147 if (pd->pf_mtag->pftag_routed++ > 3) {
9148 pbuf_destroy(*pbufp);
9149 *pbufp = NULL;
9150 m0 = NULL;
9151 goto bad;
9152 }
9153
9154 /*
9155 * Since this is something of an edge case and may involve the
9156 * host stack (for routing, at least for now), we convert the
9157 * incoming pbuf into an mbuf.
9158 */
9159 if (r->rt == PF_DUPTO) {
9160 m0 = pbuf_clone_to_mbuf(*pbufp);
9161 } else if ((r->rt == PF_REPLYTO) == (r->direction == dir)) {
9162 return;
9163 } else {
9164 /* We're about to consume this packet */
9165 m0 = pbuf_to_mbuf(*pbufp, TRUE);
9166 *pbufp = NULL;
9167 }
9168
9169 if (m0 == NULL) {
9170 goto bad;
9171 }
9172
9173 if (m0->m_len < (int)sizeof(struct ip6_hdr)) {
9174 DPFPRINTF(PF_DEBUG_URGENT,
9175 ("pf_route6: m0->m_len < sizeof (struct ip6_hdr)\n"));
9176 goto bad;
9177 }
9178 ip6 = mtod(m0, struct ip6_hdr *);
9179
9180 ro = &ip6route;
9181 bzero((caddr_t)ro, sizeof(*ro));
9182 dst = (struct sockaddr_in6 *)&ro->ro_dst;
9183 dst->sin6_family = AF_INET6;
9184 dst->sin6_len = sizeof(*dst);
9185 dst->sin6_addr = ip6->ip6_dst;
9186
9187 /* Cheat. XXX why only in the v6addr case??? */
9188 if (r->rt == PF_FASTROUTE) {
9189 pf_mtag = pf_get_mtag(m0);
9190 ASSERT(pf_mtag != NULL);
9191 pf_mtag->pftag_flags |= PF_TAG_GENERATED;
9192 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
9193 return;
9194 }
9195
9196 if (TAILQ_EMPTY(&r->rpool.list)) {
9197 DPFPRINTF(PF_DEBUG_URGENT,
9198 ("pf_route6: TAILQ_EMPTY(&r->rpool.list)\n"));
9199 goto bad;
9200 }
9201 if (s == NULL) {
9202 pf_map_addr(AF_INET6, r, (struct pf_addr *)(uintptr_t)&ip6->ip6_src,
9203 &naddr, NULL, &sn);
9204 if (!PF_AZERO(&naddr, AF_INET6)) {
9205 PF_ACPY((struct pf_addr *)&dst->sin6_addr,
9206 &naddr, AF_INET6);
9207 }
9208 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL;
9209 } else {
9210 if (!PF_AZERO(&s->rt_addr, AF_INET6)) {
9211 PF_ACPY((struct pf_addr *)&dst->sin6_addr,
9212 &s->rt_addr, AF_INET6);
9213 }
9214 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
9215 }
9216 if (ifp == NULL) {
9217 goto bad;
9218 }
9219
9220 if (oifp != ifp) {
9221 if (pf_test6_mbuf(PF_OUT, ifp, &m0, NULL, NULL) != PF_PASS) {
9222 goto bad;
9223 } else if (m0 == NULL) {
9224 goto done;
9225 }
9226 if (m0->m_len < (int)sizeof(struct ip6_hdr)) {
9227 DPFPRINTF(PF_DEBUG_URGENT, ("pf_route6: m0->m_len "
9228 "< sizeof (struct ip6_hdr)\n"));
9229 goto bad;
9230 }
9231 pf_mtag = pf_get_mtag(m0);
9232 /*
9233 * send refragmented packets.
9234 */
9235 if ((pf_mtag->pftag_flags & PF_TAG_REFRAGMENTED) != 0) {
9236 pf_mtag->pftag_flags &= ~PF_TAG_REFRAGMENTED;
9237 /*
9238 * nd6_output() frees packet chain in both success and
9239 * failure cases.
9240 */
9241 error = nd6_output(ifp, ifp, m0, dst, NULL, NULL);
9242 m0 = NULL;
9243 if (error) {
9244 DPFPRINTF(PF_DEBUG_URGENT, ("pf_route6:"
9245 "dropped refragmented packet\n"));
9246 }
9247 goto done;
9248 }
9249 ip6 = mtod(m0, struct ip6_hdr *);
9250 }
9251
9252 /*
9253 * If the packet is too large for the outgoing interface,
9254 * send back an icmp6 error.
9255 */
9256 if (IN6_IS_SCOPE_EMBED(&dst->sin6_addr)) {
9257 dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
9258 }
9259 if ((unsigned)m0->m_pkthdr.len <= ifp->if_mtu) {
9260 error = nd6_output(ifp, ifp, m0, dst, NULL, NULL);
9261 } else {
9262 in6_ifstat_inc(ifp, ifs6_in_toobig);
9263 if (r->rt != PF_DUPTO) {
9264 icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
9265 } else {
9266 goto bad;
9267 }
9268 }
9269
9270 done:
9271 return;
9272
9273 bad:
9274 if (m0) {
9275 m_freem(m0);
9276 m0 = NULL;
9277 }
9278 goto done;
9279 }
9280
9281
9282 /*
9283 * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag
9284 * off is the offset where the protocol header starts
9285 * len is the total length of protocol header plus payload
9286 * returns 0 when the checksum is valid, otherwise returns 1.
9287 */
9288 static int
9289 pf_check_proto_cksum(pbuf_t *pbuf, int off, int len, u_int8_t p,
9290 sa_family_t af)
9291 {
9292 u_int16_t sum;
9293
9294 switch (p) {
9295 case IPPROTO_TCP:
9296 case IPPROTO_UDP:
9297 /*
9298 * Optimize for the common case; if the hardware calculated
9299 * value doesn't include pseudo-header checksum, or if it
9300 * is partially-computed (only 16-bit summation), do it in
9301 * software below.
9302 */
9303 if ((*pbuf->pb_csum_flags &
9304 (CSUM_DATA_VALID | CSUM_PSEUDO_HDR)) ==
9305 (CSUM_DATA_VALID | CSUM_PSEUDO_HDR) &&
9306 (*pbuf->pb_csum_data ^ 0xffff) == 0) {
9307 return 0;
9308 }
9309 break;
9310 case IPPROTO_ICMP:
9311 case IPPROTO_ICMPV6:
9312 break;
9313 default:
9314 return 1;
9315 }
9316 if (off < (int)sizeof(struct ip) || len < (int)sizeof(struct udphdr)) {
9317 return 1;
9318 }
9319 if (pbuf->pb_packet_len < (unsigned)(off + len)) {
9320 return 1;
9321 }
9322 switch (af) {
9323 #if INET
9324 case AF_INET:
9325 if (p == IPPROTO_ICMP) {
9326 if (pbuf->pb_contig_len < (unsigned)off) {
9327 return 1;
9328 }
9329 sum = pbuf_inet_cksum(pbuf, 0, off, len);
9330 } else {
9331 if (pbuf->pb_contig_len < (int)sizeof(struct ip)) {
9332 return 1;
9333 }
9334 sum = pbuf_inet_cksum(pbuf, p, off, len);
9335 }
9336 break;
9337 #endif /* INET */
9338 case AF_INET6:
9339 if (pbuf->pb_contig_len < (int)sizeof(struct ip6_hdr)) {
9340 return 1;
9341 }
9342 sum = pbuf_inet6_cksum(pbuf, p, off, len);
9343 break;
9344 default:
9345 return 1;
9346 }
9347 if (sum) {
9348 switch (p) {
9349 case IPPROTO_TCP:
9350 tcpstat.tcps_rcvbadsum++;
9351 break;
9352 case IPPROTO_UDP:
9353 udpstat.udps_badsum++;
9354 break;
9355 case IPPROTO_ICMP:
9356 icmpstat.icps_checksum++;
9357 break;
9358 case IPPROTO_ICMPV6:
9359 icmp6stat.icp6s_checksum++;
9360 break;
9361 }
9362 return 1;
9363 }
9364 return 0;
9365 }
9366
9367 #if INET
9368 #define PF_APPLE_UPDATE_PDESC_IPv4() \
9369 do { \
9370 if (pbuf && pd.mp && pbuf != pd.mp) { \
9371 pbuf = pd.mp; \
9372 h = pbuf->pb_data; \
9373 pd.pf_mtag = pf_get_mtag_pbuf(pbuf); \
9374 } \
9375 } while (0)
9376
9377 int
9378 pf_test_mbuf(int dir, struct ifnet *ifp, struct mbuf **m0,
9379 struct ether_header *eh, struct ip_fw_args *fwa)
9380 {
9381 pbuf_t pbuf_store, *pbuf;
9382 int rv;
9383
9384 pbuf_init_mbuf(&pbuf_store, *m0, (*m0)->m_pkthdr.rcvif);
9385 pbuf = &pbuf_store;
9386
9387 rv = pf_test(dir, ifp, &pbuf, eh, fwa);
9388
9389 if (pbuf_is_valid(pbuf)) {
9390 *m0 = pbuf->pb_mbuf;
9391 pbuf->pb_mbuf = NULL;
9392 pbuf_destroy(pbuf);
9393 } else {
9394 *m0 = NULL;
9395 }
9396
9397 return rv;
9398 }
9399
9400 int
9401 pf_test(int dir, struct ifnet *ifp, pbuf_t **pbufp,
9402 struct ether_header *eh, struct ip_fw_args *fwa)
9403 {
9404 #if !DUMMYNET
9405 #pragma unused(fwa)
9406 #endif
9407 struct pfi_kif *kif;
9408 u_short action = PF_PASS, reason = 0, log = 0;
9409 pbuf_t *pbuf = *pbufp;
9410 struct ip *h = 0;
9411 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr;
9412 struct pf_state *s = NULL;
9413 struct pf_state_key *sk = NULL;
9414 struct pf_ruleset *ruleset = NULL;
9415 struct pf_pdesc pd;
9416 int off, dirndx, pqid = 0;
9417
9418 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
9419
9420 if (!pf_status.running) {
9421 return PF_PASS;
9422 }
9423
9424 memset(&pd, 0, sizeof(pd));
9425
9426 if ((pd.pf_mtag = pf_get_mtag_pbuf(pbuf)) == NULL) {
9427 DPFPRINTF(PF_DEBUG_URGENT,
9428 ("pf_test: pf_get_mtag_pbuf returned NULL\n"));
9429 return PF_DROP;
9430 }
9431
9432 if (pd.pf_mtag->pftag_flags & PF_TAG_GENERATED) {
9433 return PF_PASS;
9434 }
9435
9436 kif = (struct pfi_kif *)ifp->if_pf_kif;
9437
9438 if (kif == NULL) {
9439 DPFPRINTF(PF_DEBUG_URGENT,
9440 ("pf_test: kif == NULL, if_name %s\n", ifp->if_name));
9441 return PF_DROP;
9442 }
9443 if (kif->pfik_flags & PFI_IFLAG_SKIP) {
9444 return PF_PASS;
9445 }
9446
9447 if (pbuf->pb_packet_len < (int)sizeof(*h)) {
9448 REASON_SET(&reason, PFRES_SHORT);
9449 return PF_DROP;
9450 }
9451
9452 /* initialize enough of pd for the done label */
9453 h = pbuf->pb_data;
9454 pd.mp = pbuf;
9455 pd.lmw = 0;
9456 pd.pf_mtag = pf_get_mtag_pbuf(pbuf);
9457 pd.src = (struct pf_addr *)&h->ip_src;
9458 pd.dst = (struct pf_addr *)&h->ip_dst;
9459 PF_ACPY(&pd.baddr, pd.src, AF_INET);
9460 PF_ACPY(&pd.bdaddr, pd.dst, AF_INET);
9461 pd.ip_sum = &h->ip_sum;
9462 pd.proto = h->ip_p;
9463 pd.proto_variant = 0;
9464 pd.af = AF_INET;
9465 pd.tos = h->ip_tos;
9466 pd.ttl = h->ip_ttl;
9467 pd.tot_len = ntohs(h->ip_len);
9468 pd.eh = eh;
9469
9470 #if DUMMYNET
9471 if (fwa != NULL && fwa->fwa_pf_rule != NULL) {
9472 goto nonormalize;
9473 }
9474 #endif /* DUMMYNET */
9475
9476 /* We do IP header normalization and packet reassembly here */
9477 action = pf_normalize_ip(pbuf, dir, kif, &reason, &pd);
9478 if (action != PF_PASS || pd.lmw < 0) {
9479 action = PF_DROP;
9480 goto done;
9481 }
9482
9483 #if DUMMYNET
9484 nonormalize:
9485 #endif /* DUMMYNET */
9486 /* pf_normalize can mess with pb_data */
9487 h = pbuf->pb_data;
9488
9489 off = h->ip_hl << 2;
9490 if (off < (int)sizeof(*h)) {
9491 action = PF_DROP;
9492 REASON_SET(&reason, PFRES_SHORT);
9493 log = 1;
9494 goto done;
9495 }
9496
9497 pd.src = (struct pf_addr *)&h->ip_src;
9498 pd.dst = (struct pf_addr *)&h->ip_dst;
9499 PF_ACPY(&pd.baddr, pd.src, AF_INET);
9500 PF_ACPY(&pd.bdaddr, pd.dst, AF_INET);
9501 pd.ip_sum = &h->ip_sum;
9502 pd.proto = h->ip_p;
9503 pd.proto_variant = 0;
9504 pd.mp = pbuf;
9505 pd.lmw = 0;
9506 pd.pf_mtag = pf_get_mtag_pbuf(pbuf);
9507 pd.af = AF_INET;
9508 pd.tos = h->ip_tos;
9509 pd.ttl = h->ip_ttl;
9510 pd.sc = MBUF_SCIDX(pbuf_get_service_class(pbuf));
9511 pd.tot_len = ntohs(h->ip_len);
9512 pd.eh = eh;
9513
9514 if (*pbuf->pb_flags & PKTF_FLOW_ID) {
9515 pd.flowsrc = *pbuf->pb_flowsrc;
9516 pd.flowhash = *pbuf->pb_flowid;
9517 pd.pktflags = *pbuf->pb_flags & PKTF_FLOW_MASK;
9518 }
9519
9520 /* handle fragments that didn't get reassembled by normalization */
9521 if (h->ip_off & htons(IP_MF | IP_OFFMASK)) {
9522 pd.flags |= PFDESC_IP_FRAG;
9523 #if DUMMYNET
9524 /* Traffic goes through dummynet first */
9525 action = pf_test_dummynet(&r, dir, kif, &pbuf, &pd, fwa);
9526 if (action == PF_DROP || pbuf == NULL) {
9527 *pbufp = NULL;
9528 return action;
9529 }
9530 #endif /* DUMMYNET */
9531 action = pf_test_fragment(&r, dir, kif, pbuf, h,
9532 &pd, &a, &ruleset);
9533 goto done;
9534 }
9535
9536 switch (h->ip_p) {
9537 case IPPROTO_TCP: {
9538 struct tcphdr th;
9539 pd.hdr.tcp = &th;
9540 if (!pf_pull_hdr(pbuf, off, &th, sizeof(th),
9541 &action, &reason, AF_INET)) {
9542 log = action != PF_PASS;
9543 goto done;
9544 }
9545 pd.p_len = pd.tot_len - off - (th.th_off << 2);
9546 if ((th.th_flags & TH_ACK) && pd.p_len == 0) {
9547 pqid = 1;
9548 }
9549 #if DUMMYNET
9550 /* Traffic goes through dummynet first */
9551 action = pf_test_dummynet(&r, dir, kif, &pbuf, &pd, fwa);
9552 if (action == PF_DROP || pbuf == NULL) {
9553 *pbufp = NULL;
9554 return action;
9555 }
9556 #endif /* DUMMYNET */
9557 action = pf_normalize_tcp(dir, kif, pbuf, 0, off, h, &pd);
9558 if (pd.lmw < 0) {
9559 goto done;
9560 }
9561 PF_APPLE_UPDATE_PDESC_IPv4();
9562 if (action == PF_DROP) {
9563 goto done;
9564 }
9565 action = pf_test_state_tcp(&s, dir, kif, pbuf, off, h, &pd,
9566 &reason);
9567 if (action == PF_NAT64) {
9568 goto done;
9569 }
9570 if (pd.lmw < 0) {
9571 goto done;
9572 }
9573 PF_APPLE_UPDATE_PDESC_IPv4();
9574 if (action == PF_PASS) {
9575 #if NPFSYNC
9576 pfsync_update_state(s);
9577 #endif /* NPFSYNC */
9578 r = s->rule.ptr;
9579 a = s->anchor.ptr;
9580 log = s->log;
9581 } else if (s == NULL) {
9582 action = pf_test_rule(&r, &s, dir, kif,
9583 pbuf, off, h, &pd, &a, &ruleset, NULL);
9584 }
9585 break;
9586 }
9587
9588 case IPPROTO_UDP: {
9589 struct udphdr uh;
9590
9591 pd.hdr.udp = &uh;
9592 if (!pf_pull_hdr(pbuf, off, &uh, sizeof(uh),
9593 &action, &reason, AF_INET)) {
9594 log = action != PF_PASS;
9595 goto done;
9596 }
9597 if (uh.uh_dport == 0 ||
9598 ntohs(uh.uh_ulen) > pbuf->pb_packet_len - off ||
9599 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
9600 action = PF_DROP;
9601 REASON_SET(&reason, PFRES_SHORT);
9602 goto done;
9603 }
9604 #if DUMMYNET
9605 /* Traffic goes through dummynet first */
9606 action = pf_test_dummynet(&r, dir, kif, &pbuf, &pd, fwa);
9607 if (action == PF_DROP || pbuf == NULL) {
9608 *pbufp = NULL;
9609 return action;
9610 }
9611 #endif /* DUMMYNET */
9612 action = pf_test_state_udp(&s, dir, kif, pbuf, off, h, &pd,
9613 &reason);
9614 if (action == PF_NAT64) {
9615 goto done;
9616 }
9617 if (pd.lmw < 0) {
9618 goto done;
9619 }
9620 PF_APPLE_UPDATE_PDESC_IPv4();
9621 if (action == PF_PASS) {
9622 #if NPFSYNC
9623 pfsync_update_state(s);
9624 #endif /* NPFSYNC */
9625 r = s->rule.ptr;
9626 a = s->anchor.ptr;
9627 log = s->log;
9628 } else if (s == NULL) {
9629 action = pf_test_rule(&r, &s, dir, kif,
9630 pbuf, off, h, &pd, &a, &ruleset, NULL);
9631 }
9632 break;
9633 }
9634
9635 case IPPROTO_ICMP: {
9636 struct icmp ih;
9637
9638 pd.hdr.icmp = &ih;
9639 if (!pf_pull_hdr(pbuf, off, &ih, ICMP_MINLEN,
9640 &action, &reason, AF_INET)) {
9641 log = action != PF_PASS;
9642 goto done;
9643 }
9644 #if DUMMYNET
9645 /* Traffic goes through dummynet first */
9646 action = pf_test_dummynet(&r, dir, kif, &pbuf, &pd, fwa);
9647 if (action == PF_DROP || pbuf == NULL) {
9648 *pbufp = NULL;
9649 return action;
9650 }
9651 #endif /* DUMMYNET */
9652 action = pf_test_state_icmp(&s, dir, kif, pbuf, off, h, &pd,
9653 &reason);
9654 if (action == PF_NAT64) {
9655 goto done;
9656 }
9657 if (pd.lmw < 0) {
9658 goto done;
9659 }
9660 PF_APPLE_UPDATE_PDESC_IPv4();
9661 if (action == PF_PASS) {
9662 #if NPFSYNC
9663 pfsync_update_state(s);
9664 #endif /* NPFSYNC */
9665 r = s->rule.ptr;
9666 a = s->anchor.ptr;
9667 log = s->log;
9668 } else if (s == NULL) {
9669 action = pf_test_rule(&r, &s, dir, kif,
9670 pbuf, off, h, &pd, &a, &ruleset, NULL);
9671 }
9672 break;
9673 }
9674
9675 case IPPROTO_ESP: {
9676 struct pf_esp_hdr esp;
9677
9678 pd.hdr.esp = &esp;
9679 if (!pf_pull_hdr(pbuf, off, &esp, sizeof(esp), &action, &reason,
9680 AF_INET)) {
9681 log = action != PF_PASS;
9682 goto done;
9683 }
9684 #if DUMMYNET
9685 /* Traffic goes through dummynet first */
9686 action = pf_test_dummynet(&r, dir, kif, &pbuf, &pd, fwa);
9687 if (action == PF_DROP || pbuf == NULL) {
9688 *pbufp = NULL;
9689 return action;
9690 }
9691 #endif /* DUMMYNET */
9692 action = pf_test_state_esp(&s, dir, kif, off, &pd);
9693 if (pd.lmw < 0) {
9694 goto done;
9695 }
9696 PF_APPLE_UPDATE_PDESC_IPv4();
9697 if (action == PF_PASS) {
9698 #if NPFSYNC
9699 pfsync_update_state(s);
9700 #endif /* NPFSYNC */
9701 r = s->rule.ptr;
9702 a = s->anchor.ptr;
9703 log = s->log;
9704 } else if (s == NULL) {
9705 action = pf_test_rule(&r, &s, dir, kif,
9706 pbuf, off, h, &pd, &a, &ruleset, NULL);
9707 }
9708 break;
9709 }
9710
9711 case IPPROTO_GRE: {
9712 struct pf_grev1_hdr grev1;
9713 pd.hdr.grev1 = &grev1;
9714 if (!pf_pull_hdr(pbuf, off, &grev1, sizeof(grev1), &action,
9715 &reason, AF_INET)) {
9716 log = (action != PF_PASS);
9717 goto done;
9718 }
9719 #if DUMMYNET
9720 /* Traffic goes through dummynet first */
9721 action = pf_test_dummynet(&r, dir, kif, &pbuf, &pd, fwa);
9722 if (action == PF_DROP || pbuf == NULL) {
9723 *pbufp = NULL;
9724 return action;
9725 }
9726 #endif /* DUMMYNET */
9727 if ((ntohs(grev1.flags) & PF_GRE_FLAG_VERSION_MASK) == 1 &&
9728 ntohs(grev1.protocol_type) == PF_GRE_PPP_ETHERTYPE) {
9729 if (ntohs(grev1.payload_length) >
9730 pbuf->pb_packet_len - off) {
9731 action = PF_DROP;
9732 REASON_SET(&reason, PFRES_SHORT);
9733 goto done;
9734 }
9735 pd.proto_variant = PF_GRE_PPTP_VARIANT;
9736 action = pf_test_state_grev1(&s, dir, kif, off, &pd);
9737 if (pd.lmw < 0) {
9738 goto done;
9739 }
9740 PF_APPLE_UPDATE_PDESC_IPv4();
9741 if (action == PF_PASS) {
9742 #if NPFSYNC
9743 pfsync_update_state(s);
9744 #endif /* NPFSYNC */
9745 r = s->rule.ptr;
9746 a = s->anchor.ptr;
9747 log = s->log;
9748 break;
9749 } else if (s == NULL) {
9750 action = pf_test_rule(&r, &s, dir, kif, pbuf,
9751 off, h, &pd, &a, &ruleset, NULL);
9752 if (action == PF_PASS) {
9753 break;
9754 }
9755 }
9756 }
9757
9758 /* not GREv1/PPTP, so treat as ordinary GRE... */
9759 OS_FALLTHROUGH;
9760 }
9761
9762 default:
9763 #if DUMMYNET
9764 /* Traffic goes through dummynet first */
9765 action = pf_test_dummynet(&r, dir, kif, &pbuf, &pd, fwa);
9766 if (action == PF_DROP || pbuf == NULL) {
9767 *pbufp = NULL;
9768 return action;
9769 }
9770 #endif /* DUMMYNET */
9771 action = pf_test_state_other(&s, dir, kif, &pd);
9772 if (pd.lmw < 0) {
9773 goto done;
9774 }
9775 PF_APPLE_UPDATE_PDESC_IPv4();
9776 if (action == PF_PASS) {
9777 #if NPFSYNC
9778 pfsync_update_state(s);
9779 #endif /* NPFSYNC */
9780 r = s->rule.ptr;
9781 a = s->anchor.ptr;
9782 log = s->log;
9783 } else if (s == NULL) {
9784 action = pf_test_rule(&r, &s, dir, kif, pbuf, off, h,
9785 &pd, &a, &ruleset, NULL);
9786 }
9787 break;
9788 }
9789
9790 done:
9791 if (action == PF_NAT64) {
9792 *pbufp = NULL;
9793 return action;
9794 }
9795
9796 *pbufp = pd.mp;
9797 PF_APPLE_UPDATE_PDESC_IPv4();
9798
9799 if (action != PF_DROP) {
9800 if (action == PF_PASS && h->ip_hl > 5 &&
9801 !((s && s->allow_opts) || r->allow_opts)) {
9802 action = PF_DROP;
9803 REASON_SET(&reason, PFRES_IPOPTIONS);
9804 log = 1;
9805 DPFPRINTF(PF_DEBUG_MISC,
9806 ("pf: dropping packet with ip options [hlen=%u]\n",
9807 (unsigned int) h->ip_hl));
9808 }
9809
9810 if ((s && s->tag) || PF_RTABLEID_IS_VALID(r->rtableid) ||
9811 (pd.pktflags & PKTF_FLOW_ID)) {
9812 (void) pf_tag_packet(pbuf, pd.pf_mtag, s ? s->tag : 0,
9813 r->rtableid, &pd);
9814 }
9815
9816 if (action == PF_PASS) {
9817 #if PF_ECN
9818 /* add hints for ecn */
9819 pd.pf_mtag->pftag_hdr = h;
9820 /* record address family */
9821 pd.pf_mtag->pftag_flags &= ~PF_TAG_HDR_INET6;
9822 pd.pf_mtag->pftag_flags |= PF_TAG_HDR_INET;
9823 #endif /* PF_ECN */
9824 /* record protocol */
9825 *pbuf->pb_proto = pd.proto;
9826
9827 /*
9828 * connections redirected to loopback should not match sockets
9829 * bound specifically to loopback due to security implications,
9830 * see tcp_input() and in_pcblookup_listen().
9831 */
9832 if (dir == PF_IN && (pd.proto == IPPROTO_TCP ||
9833 pd.proto == IPPROTO_UDP) && s != NULL &&
9834 s->nat_rule.ptr != NULL &&
9835 (s->nat_rule.ptr->action == PF_RDR ||
9836 s->nat_rule.ptr->action == PF_BINAT) &&
9837 (ntohl(pd.dst->v4addr.s_addr) >> IN_CLASSA_NSHIFT)
9838 == IN_LOOPBACKNET) {
9839 pd.pf_mtag->pftag_flags |= PF_TAG_TRANSLATE_LOCALHOST;
9840 }
9841 }
9842 }
9843
9844 if (log) {
9845 struct pf_rule *lr;
9846
9847 if (s != NULL && s->nat_rule.ptr != NULL &&
9848 s->nat_rule.ptr->log & PF_LOG_ALL) {
9849 lr = s->nat_rule.ptr;
9850 } else {
9851 lr = r;
9852 }
9853 PFLOG_PACKET(kif, h, pbuf, AF_INET, dir, reason, lr, a, ruleset,
9854 &pd);
9855 }
9856
9857 kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
9858 kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++;
9859
9860 if (action == PF_PASS || r->action == PF_DROP) {
9861 dirndx = (dir == PF_OUT);
9862 r->packets[dirndx]++;
9863 r->bytes[dirndx] += pd.tot_len;
9864 if (a != NULL) {
9865 a->packets[dirndx]++;
9866 a->bytes[dirndx] += pd.tot_len;
9867 }
9868 if (s != NULL) {
9869 sk = s->state_key;
9870 if (s->nat_rule.ptr != NULL) {
9871 s->nat_rule.ptr->packets[dirndx]++;
9872 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len;
9873 }
9874 if (s->src_node != NULL) {
9875 s->src_node->packets[dirndx]++;
9876 s->src_node->bytes[dirndx] += pd.tot_len;
9877 }
9878 if (s->nat_src_node != NULL) {
9879 s->nat_src_node->packets[dirndx]++;
9880 s->nat_src_node->bytes[dirndx] += pd.tot_len;
9881 }
9882 dirndx = (dir == sk->direction) ? 0 : 1;
9883 s->packets[dirndx]++;
9884 s->bytes[dirndx] += pd.tot_len;
9885 }
9886 tr = r;
9887 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
9888 if (nr != NULL) {
9889 struct pf_addr *x;
9890 /*
9891 * XXX: we need to make sure that the addresses
9892 * passed to pfr_update_stats() are the same than
9893 * the addresses used during matching (pfr_match)
9894 */
9895 if (r == &pf_default_rule) {
9896 tr = nr;
9897 x = (sk == NULL || sk->direction == dir) ?
9898 &pd.baddr : &pd.naddr;
9899 } else {
9900 x = (sk == NULL || sk->direction == dir) ?
9901 &pd.naddr : &pd.baddr;
9902 }
9903 if (x == &pd.baddr || s == NULL) {
9904 /* we need to change the address */
9905 if (dir == PF_OUT) {
9906 pd.src = x;
9907 } else {
9908 pd.dst = x;
9909 }
9910 }
9911 }
9912 if (tr->src.addr.type == PF_ADDR_TABLE) {
9913 pfr_update_stats(tr->src.addr.p.tbl, (sk == NULL ||
9914 sk->direction == dir) ?
9915 pd.src : pd.dst, pd.af,
9916 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
9917 tr->src.neg);
9918 }
9919 if (tr->dst.addr.type == PF_ADDR_TABLE) {
9920 pfr_update_stats(tr->dst.addr.p.tbl, (sk == NULL ||
9921 sk->direction == dir) ? pd.dst : pd.src, pd.af,
9922 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
9923 tr->dst.neg);
9924 }
9925 }
9926
9927 VERIFY(pbuf == NULL || pd.mp == NULL || pd.mp == pbuf);
9928
9929 if (*pbufp) {
9930 if (pd.lmw < 0) {
9931 REASON_SET(&reason, PFRES_MEMORY);
9932 action = PF_DROP;
9933 }
9934
9935 if (action == PF_DROP) {
9936 pbuf_destroy(*pbufp);
9937 *pbufp = NULL;
9938 return PF_DROP;
9939 }
9940
9941 *pbufp = pbuf;
9942 }
9943
9944 if (action == PF_SYNPROXY_DROP) {
9945 pbuf_destroy(*pbufp);
9946 *pbufp = NULL;
9947 action = PF_PASS;
9948 } else if (r->rt) {
9949 /* pf_route can free the pbuf causing *pbufp to become NULL */
9950 pf_route(pbufp, r, dir, kif->pfik_ifp, s, &pd);
9951 }
9952
9953 return action;
9954 }
9955 #endif /* INET */
9956
9957 #define PF_APPLE_UPDATE_PDESC_IPv6() \
9958 do { \
9959 if (pbuf && pd.mp && pbuf != pd.mp) { \
9960 pbuf = pd.mp; \
9961 } \
9962 h = pbuf->pb_data; \
9963 } while (0)
9964
9965 int
9966 pf_test6_mbuf(int dir, struct ifnet *ifp, struct mbuf **m0,
9967 struct ether_header *eh, struct ip_fw_args *fwa)
9968 {
9969 pbuf_t pbuf_store, *pbuf;
9970 int rv;
9971
9972 pbuf_init_mbuf(&pbuf_store, *m0, (*m0)->m_pkthdr.rcvif);
9973 pbuf = &pbuf_store;
9974
9975 rv = pf_test6(dir, ifp, &pbuf, eh, fwa);
9976
9977 if (pbuf_is_valid(pbuf)) {
9978 *m0 = pbuf->pb_mbuf;
9979 pbuf->pb_mbuf = NULL;
9980 pbuf_destroy(pbuf);
9981 } else {
9982 *m0 = NULL;
9983 }
9984
9985 return rv;
9986 }
9987
9988 int
9989 pf_test6(int dir, struct ifnet *ifp, pbuf_t **pbufp,
9990 struct ether_header *eh, struct ip_fw_args *fwa)
9991 {
9992 #if !DUMMYNET
9993 #pragma unused(fwa)
9994 #endif
9995 struct pfi_kif *kif;
9996 u_short action = PF_PASS, reason = 0, log = 0;
9997 pbuf_t *pbuf = *pbufp;
9998 struct ip6_hdr *h;
9999 struct pf_rule *a = NULL, *r = &pf_default_rule, *tr, *nr;
10000 struct pf_state *s = NULL;
10001 struct pf_state_key *sk = NULL;
10002 struct pf_ruleset *ruleset = NULL;
10003 struct pf_pdesc pd;
10004 int off, terminal = 0, dirndx, rh_cnt = 0;
10005 u_int8_t nxt;
10006 boolean_t fwd = FALSE;
10007
10008 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
10009
10010 ASSERT(ifp != NULL);
10011 if ((dir == PF_OUT) && (pbuf->pb_ifp) && (ifp != pbuf->pb_ifp)) {
10012 fwd = TRUE;
10013 }
10014
10015 if (!pf_status.running) {
10016 return PF_PASS;
10017 }
10018
10019 memset(&pd, 0, sizeof(pd));
10020
10021 if ((pd.pf_mtag = pf_get_mtag_pbuf(pbuf)) == NULL) {
10022 DPFPRINTF(PF_DEBUG_URGENT,
10023 ("pf_test6: pf_get_mtag_pbuf returned NULL\n"));
10024 return PF_DROP;
10025 }
10026
10027 if (pd.pf_mtag->pftag_flags & PF_TAG_GENERATED) {
10028 return PF_PASS;
10029 }
10030
10031 kif = (struct pfi_kif *)ifp->if_pf_kif;
10032
10033 if (kif == NULL) {
10034 DPFPRINTF(PF_DEBUG_URGENT,
10035 ("pf_test6: kif == NULL, if_name %s\n", ifp->if_name));
10036 return PF_DROP;
10037 }
10038 if (kif->pfik_flags & PFI_IFLAG_SKIP) {
10039 return PF_PASS;
10040 }
10041
10042 if (pbuf->pb_packet_len < (int)sizeof(*h)) {
10043 REASON_SET(&reason, PFRES_SHORT);
10044 return PF_DROP;
10045 }
10046
10047 h = pbuf->pb_data;
10048 nxt = h->ip6_nxt;
10049 off = ((caddr_t)h - (caddr_t)pbuf->pb_data) + sizeof(struct ip6_hdr);
10050 pd.mp = pbuf;
10051 pd.lmw = 0;
10052 pd.pf_mtag = pf_get_mtag_pbuf(pbuf);
10053 pd.src = (struct pf_addr *)(uintptr_t)&h->ip6_src;
10054 pd.dst = (struct pf_addr *)(uintptr_t)&h->ip6_dst;
10055 PF_ACPY(&pd.baddr, pd.src, AF_INET6);
10056 PF_ACPY(&pd.bdaddr, pd.dst, AF_INET6);
10057 pd.ip_sum = NULL;
10058 pd.af = AF_INET6;
10059 pd.proto = nxt;
10060 pd.proto_variant = 0;
10061 pd.tos = 0;
10062 pd.ttl = h->ip6_hlim;
10063 pd.sc = MBUF_SCIDX(pbuf_get_service_class(pbuf));
10064 pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr);
10065 pd.eh = eh;
10066
10067 if (*pbuf->pb_flags & PKTF_FLOW_ID) {
10068 pd.flowsrc = *pbuf->pb_flowsrc;
10069 pd.flowhash = *pbuf->pb_flowid;
10070 pd.pktflags = (*pbuf->pb_flags & PKTF_FLOW_MASK);
10071 }
10072
10073 #if DUMMYNET
10074 if (fwa != NULL && fwa->fwa_pf_rule != NULL) {
10075 goto nonormalize;
10076 }
10077 #endif /* DUMMYNET */
10078
10079 /* We do IP header normalization and packet reassembly here */
10080 action = pf_normalize_ip6(pbuf, dir, kif, &reason, &pd);
10081 if (action != PF_PASS || pd.lmw < 0) {
10082 action = PF_DROP;
10083 goto done;
10084 }
10085
10086 #if DUMMYNET
10087 nonormalize:
10088 #endif /* DUMMYNET */
10089 h = pbuf->pb_data;
10090
10091 /*
10092 * we do not support jumbogram yet. if we keep going, zero ip6_plen
10093 * will do something bad, so drop the packet for now.
10094 */
10095 if (htons(h->ip6_plen) == 0) {
10096 action = PF_DROP;
10097 REASON_SET(&reason, PFRES_NORM); /*XXX*/
10098 goto done;
10099 }
10100 pd.src = (struct pf_addr *)(uintptr_t)&h->ip6_src;
10101 pd.dst = (struct pf_addr *)(uintptr_t)&h->ip6_dst;
10102 PF_ACPY(&pd.baddr, pd.src, AF_INET6);
10103 PF_ACPY(&pd.bdaddr, pd.dst, AF_INET6);
10104 pd.ip_sum = NULL;
10105 pd.af = AF_INET6;
10106 pd.tos = 0;
10107 pd.ttl = h->ip6_hlim;
10108 pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr);
10109 pd.eh = eh;
10110
10111 off = ((caddr_t)h - (caddr_t)pbuf->pb_data) + sizeof(struct ip6_hdr);
10112 pd.proto = h->ip6_nxt;
10113 pd.proto_variant = 0;
10114 pd.mp = pbuf;
10115 pd.lmw = 0;
10116 pd.pf_mtag = pf_get_mtag_pbuf(pbuf);
10117
10118 do {
10119 switch (pd.proto) {
10120 case IPPROTO_FRAGMENT: {
10121 struct ip6_frag ip6f;
10122
10123 pd.flags |= PFDESC_IP_FRAG;
10124 if (!pf_pull_hdr(pbuf, off, &ip6f, sizeof ip6f, NULL,
10125 &reason, pd.af)) {
10126 DPFPRINTF(PF_DEBUG_MISC,
10127 ("pf: IPv6 short fragment header\n"));
10128 action = PF_DROP;
10129 REASON_SET(&reason, PFRES_SHORT);
10130 log = 1;
10131 goto done;
10132 }
10133 pd.proto = ip6f.ip6f_nxt;
10134 #if DUMMYNET
10135 /* Traffic goes through dummynet first */
10136 action = pf_test_dummynet(&r, dir, kif, &pbuf, &pd,
10137 fwa);
10138 if (action == PF_DROP || pbuf == NULL) {
10139 *pbufp = NULL;
10140 return action;
10141 }
10142 #endif /* DUMMYNET */
10143 action = pf_test_fragment(&r, dir, kif, pbuf, h, &pd,
10144 &a, &ruleset);
10145 if (action == PF_DROP) {
10146 REASON_SET(&reason, PFRES_FRAG);
10147 log = 1;
10148 }
10149 goto done;
10150 }
10151 case IPPROTO_ROUTING:
10152 ++rh_cnt;
10153 OS_FALLTHROUGH;
10154
10155 case IPPROTO_AH:
10156 case IPPROTO_HOPOPTS:
10157 case IPPROTO_DSTOPTS: {
10158 /* get next header and header length */
10159 struct ip6_ext opt6;
10160
10161 if (!pf_pull_hdr(pbuf, off, &opt6, sizeof(opt6),
10162 NULL, &reason, pd.af)) {
10163 DPFPRINTF(PF_DEBUG_MISC,
10164 ("pf: IPv6 short opt\n"));
10165 action = PF_DROP;
10166 log = 1;
10167 goto done;
10168 }
10169 if (pd.proto == IPPROTO_AH) {
10170 off += (opt6.ip6e_len + 2) * 4;
10171 } else {
10172 off += (opt6.ip6e_len + 1) * 8;
10173 }
10174 pd.proto = opt6.ip6e_nxt;
10175 /* goto the next header */
10176 break;
10177 }
10178 default:
10179 terminal++;
10180 break;
10181 }
10182 } while (!terminal);
10183
10184
10185 switch (pd.proto) {
10186 case IPPROTO_TCP: {
10187 struct tcphdr th;
10188
10189 pd.hdr.tcp = &th;
10190 if (!pf_pull_hdr(pbuf, off, &th, sizeof(th),
10191 &action, &reason, AF_INET6)) {
10192 log = action != PF_PASS;
10193 goto done;
10194 }
10195 pd.p_len = pd.tot_len - off - (th.th_off << 2);
10196 #if DUMMYNET
10197 /* Traffic goes through dummynet first */
10198 action = pf_test_dummynet(&r, dir, kif, &pbuf, &pd, fwa);
10199 if (action == PF_DROP || pbuf == NULL) {
10200 *pbufp = NULL;
10201 return action;
10202 }
10203 #endif /* DUMMYNET */
10204 action = pf_normalize_tcp(dir, kif, pbuf, 0, off, h, &pd);
10205 if (pd.lmw < 0) {
10206 goto done;
10207 }
10208 PF_APPLE_UPDATE_PDESC_IPv6();
10209 if (action == PF_DROP) {
10210 goto done;
10211 }
10212 action = pf_test_state_tcp(&s, dir, kif, pbuf, off, h, &pd,
10213 &reason);
10214 if (action == PF_NAT64) {
10215 goto done;
10216 }
10217 if (pd.lmw < 0) {
10218 goto done;
10219 }
10220 PF_APPLE_UPDATE_PDESC_IPv6();
10221 if (action == PF_PASS) {
10222 #if NPFSYNC
10223 pfsync_update_state(s);
10224 #endif /* NPFSYNC */
10225 r = s->rule.ptr;
10226 a = s->anchor.ptr;
10227 log = s->log;
10228 } else if (s == NULL) {
10229 action = pf_test_rule(&r, &s, dir, kif,
10230 pbuf, off, h, &pd, &a, &ruleset, NULL);
10231 }
10232 break;
10233 }
10234
10235 case IPPROTO_UDP: {
10236 struct udphdr uh;
10237
10238 pd.hdr.udp = &uh;
10239 if (!pf_pull_hdr(pbuf, off, &uh, sizeof(uh),
10240 &action, &reason, AF_INET6)) {
10241 log = action != PF_PASS;
10242 goto done;
10243 }
10244 if (uh.uh_dport == 0 ||
10245 ntohs(uh.uh_ulen) > pbuf->pb_packet_len - off ||
10246 ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
10247 action = PF_DROP;
10248 REASON_SET(&reason, PFRES_SHORT);
10249 goto done;
10250 }
10251 #if DUMMYNET
10252 /* Traffic goes through dummynet first */
10253 action = pf_test_dummynet(&r, dir, kif, &pbuf, &pd, fwa);
10254 if (action == PF_DROP || pbuf == NULL) {
10255 *pbufp = NULL;
10256 return action;
10257 }
10258 #endif /* DUMMYNET */
10259 action = pf_test_state_udp(&s, dir, kif, pbuf, off, h, &pd,
10260 &reason);
10261 if (action == PF_NAT64) {
10262 goto done;
10263 }
10264 if (pd.lmw < 0) {
10265 goto done;
10266 }
10267 PF_APPLE_UPDATE_PDESC_IPv6();
10268 if (action == PF_PASS) {
10269 #if NPFSYNC
10270 pfsync_update_state(s);
10271 #endif /* NPFSYNC */
10272 r = s->rule.ptr;
10273 a = s->anchor.ptr;
10274 log = s->log;
10275 } else if (s == NULL) {
10276 action = pf_test_rule(&r, &s, dir, kif,
10277 pbuf, off, h, &pd, &a, &ruleset, NULL);
10278 }
10279 break;
10280 }
10281
10282 case IPPROTO_ICMPV6: {
10283 struct icmp6_hdr ih;
10284
10285 pd.hdr.icmp6 = &ih;
10286 if (!pf_pull_hdr(pbuf, off, &ih, sizeof(ih),
10287 &action, &reason, AF_INET6)) {
10288 log = action != PF_PASS;
10289 goto done;
10290 }
10291 #if DUMMYNET
10292 /* Traffic goes through dummynet first */
10293 action = pf_test_dummynet(&r, dir, kif, &pbuf, &pd, fwa);
10294 if (action == PF_DROP || pbuf == NULL) {
10295 *pbufp = NULL;
10296 return action;
10297 }
10298 #endif /* DUMMYNET */
10299 action = pf_test_state_icmp(&s, dir, kif,
10300 pbuf, off, h, &pd, &reason);
10301 if (action == PF_NAT64) {
10302 goto done;
10303 }
10304 if (pd.lmw < 0) {
10305 goto done;
10306 }
10307 PF_APPLE_UPDATE_PDESC_IPv6();
10308 if (action == PF_PASS) {
10309 #if NPFSYNC
10310 pfsync_update_state(s);
10311 #endif /* NPFSYNC */
10312 r = s->rule.ptr;
10313 a = s->anchor.ptr;
10314 log = s->log;
10315 } else if (s == NULL) {
10316 action = pf_test_rule(&r, &s, dir, kif,
10317 pbuf, off, h, &pd, &a, &ruleset, NULL);
10318 }
10319 break;
10320 }
10321
10322 case IPPROTO_ESP: {
10323 struct pf_esp_hdr esp;
10324
10325 pd.hdr.esp = &esp;
10326 if (!pf_pull_hdr(pbuf, off, &esp, sizeof(esp), &action,
10327 &reason, AF_INET6)) {
10328 log = action != PF_PASS;
10329 goto done;
10330 }
10331 #if DUMMYNET
10332 /* Traffic goes through dummynet first */
10333 action = pf_test_dummynet(&r, dir, kif, &pbuf, &pd, fwa);
10334 if (action == PF_DROP || pbuf == NULL) {
10335 *pbufp = NULL;
10336 return action;
10337 }
10338 #endif /* DUMMYNET */
10339 action = pf_test_state_esp(&s, dir, kif, off, &pd);
10340 if (pd.lmw < 0) {
10341 goto done;
10342 }
10343 PF_APPLE_UPDATE_PDESC_IPv6();
10344 if (action == PF_PASS) {
10345 #if NPFSYNC
10346 pfsync_update_state(s);
10347 #endif /* NPFSYNC */
10348 r = s->rule.ptr;
10349 a = s->anchor.ptr;
10350 log = s->log;
10351 } else if (s == NULL) {
10352 action = pf_test_rule(&r, &s, dir, kif,
10353 pbuf, off, h, &pd, &a, &ruleset, NULL);
10354 }
10355 break;
10356 }
10357
10358 case IPPROTO_GRE: {
10359 struct pf_grev1_hdr grev1;
10360
10361 pd.hdr.grev1 = &grev1;
10362 if (!pf_pull_hdr(pbuf, off, &grev1, sizeof(grev1), &action,
10363 &reason, AF_INET6)) {
10364 log = (action != PF_PASS);
10365 goto done;
10366 }
10367 #if DUMMYNET
10368 /* Traffic goes through dummynet first */
10369 action = pf_test_dummynet(&r, dir, kif, &pbuf, &pd, fwa);
10370 if (action == PF_DROP || pbuf == NULL) {
10371 *pbufp = NULL;
10372 return action;
10373 }
10374 #endif /* DUMMYNET */
10375 if ((ntohs(grev1.flags) & PF_GRE_FLAG_VERSION_MASK) == 1 &&
10376 ntohs(grev1.protocol_type) == PF_GRE_PPP_ETHERTYPE) {
10377 if (ntohs(grev1.payload_length) >
10378 pbuf->pb_packet_len - off) {
10379 action = PF_DROP;
10380 REASON_SET(&reason, PFRES_SHORT);
10381 goto done;
10382 }
10383 action = pf_test_state_grev1(&s, dir, kif, off, &pd);
10384 if (pd.lmw < 0) {
10385 goto done;
10386 }
10387 PF_APPLE_UPDATE_PDESC_IPv6();
10388 if (action == PF_PASS) {
10389 #if NPFSYNC
10390 pfsync_update_state(s);
10391 #endif /* NPFSYNC */
10392 r = s->rule.ptr;
10393 a = s->anchor.ptr;
10394 log = s->log;
10395 break;
10396 } else if (s == NULL) {
10397 action = pf_test_rule(&r, &s, dir, kif, pbuf,
10398 off, h, &pd, &a, &ruleset, NULL);
10399 if (action == PF_PASS) {
10400 break;
10401 }
10402 }
10403 }
10404
10405 /* not GREv1/PPTP, so treat as ordinary GRE... */
10406 OS_FALLTHROUGH; /* XXX is this correct? */
10407 }
10408
10409 default:
10410 #if DUMMYNET
10411 /* Traffic goes through dummynet first */
10412 action = pf_test_dummynet(&r, dir, kif, &pbuf, &pd, fwa);
10413 if (action == PF_DROP || pbuf == NULL) {
10414 *pbufp = NULL;
10415 return action;
10416 }
10417 #endif /* DUMMYNET */
10418 action = pf_test_state_other(&s, dir, kif, &pd);
10419 if (pd.lmw < 0) {
10420 goto done;
10421 }
10422 PF_APPLE_UPDATE_PDESC_IPv6();
10423 if (action == PF_PASS) {
10424 #if NPFSYNC
10425 pfsync_update_state(s);
10426 #endif /* NPFSYNC */
10427 r = s->rule.ptr;
10428 a = s->anchor.ptr;
10429 log = s->log;
10430 } else if (s == NULL) {
10431 action = pf_test_rule(&r, &s, dir, kif, pbuf, off, h,
10432 &pd, &a, &ruleset, NULL);
10433 }
10434 break;
10435 }
10436
10437 done:
10438 if (action == PF_NAT64) {
10439 *pbufp = NULL;
10440 return action;
10441 }
10442
10443 *pbufp = pd.mp;
10444 PF_APPLE_UPDATE_PDESC_IPv6();
10445
10446 /* handle dangerous IPv6 extension headers. */
10447 if (action != PF_DROP) {
10448 if (action == PF_PASS && rh_cnt &&
10449 !((s && s->allow_opts) || r->allow_opts)) {
10450 action = PF_DROP;
10451 REASON_SET(&reason, PFRES_IPOPTIONS);
10452 log = 1;
10453 DPFPRINTF(PF_DEBUG_MISC,
10454 ("pf: dropping packet with dangerous v6addr headers\n"));
10455 }
10456
10457 if ((s && s->tag) || PF_RTABLEID_IS_VALID(r->rtableid) ||
10458 (pd.pktflags & PKTF_FLOW_ID)) {
10459 (void) pf_tag_packet(pbuf, pd.pf_mtag, s ? s->tag : 0,
10460 r->rtableid, &pd);
10461 }
10462
10463 if (action == PF_PASS) {
10464 #if PF_ECN
10465 /* add hints for ecn */
10466 pd.pf_mtag->pftag_hdr = h;
10467 /* record address family */
10468 pd.pf_mtag->pftag_flags &= ~PF_TAG_HDR_INET;
10469 pd.pf_mtag->pftag_flags |= PF_TAG_HDR_INET6;
10470 #endif /* PF_ECN */
10471 /* record protocol */
10472 *pbuf->pb_proto = pd.proto;
10473 if (dir == PF_IN && (pd.proto == IPPROTO_TCP ||
10474 pd.proto == IPPROTO_UDP) && s != NULL &&
10475 s->nat_rule.ptr != NULL &&
10476 (s->nat_rule.ptr->action == PF_RDR ||
10477 s->nat_rule.ptr->action == PF_BINAT) &&
10478 IN6_IS_ADDR_LOOPBACK(&pd.dst->v6addr)) {
10479 pd.pf_mtag->pftag_flags |= PF_TAG_TRANSLATE_LOCALHOST;
10480 }
10481 }
10482 }
10483
10484
10485 if (log) {
10486 struct pf_rule *lr;
10487
10488 if (s != NULL && s->nat_rule.ptr != NULL &&
10489 s->nat_rule.ptr->log & PF_LOG_ALL) {
10490 lr = s->nat_rule.ptr;
10491 } else {
10492 lr = r;
10493 }
10494 PFLOG_PACKET(kif, h, pbuf, AF_INET6, dir, reason, lr, a, ruleset,
10495 &pd);
10496 }
10497
10498 kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
10499 kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++;
10500
10501 if (action == PF_PASS || r->action == PF_DROP) {
10502 dirndx = (dir == PF_OUT);
10503 r->packets[dirndx]++;
10504 r->bytes[dirndx] += pd.tot_len;
10505 if (a != NULL) {
10506 a->packets[dirndx]++;
10507 a->bytes[dirndx] += pd.tot_len;
10508 }
10509 if (s != NULL) {
10510 sk = s->state_key;
10511 if (s->nat_rule.ptr != NULL) {
10512 s->nat_rule.ptr->packets[dirndx]++;
10513 s->nat_rule.ptr->bytes[dirndx] += pd.tot_len;
10514 }
10515 if (s->src_node != NULL) {
10516 s->src_node->packets[dirndx]++;
10517 s->src_node->bytes[dirndx] += pd.tot_len;
10518 }
10519 if (s->nat_src_node != NULL) {
10520 s->nat_src_node->packets[dirndx]++;
10521 s->nat_src_node->bytes[dirndx] += pd.tot_len;
10522 }
10523 dirndx = (dir == sk->direction) ? 0 : 1;
10524 s->packets[dirndx]++;
10525 s->bytes[dirndx] += pd.tot_len;
10526 }
10527 tr = r;
10528 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
10529 if (nr != NULL) {
10530 struct pf_addr *x;
10531 /*
10532 * XXX: we need to make sure that the addresses
10533 * passed to pfr_update_stats() are the same than
10534 * the addresses used during matching (pfr_match)
10535 */
10536 if (r == &pf_default_rule) {
10537 tr = nr;
10538 x = (s == NULL || sk->direction == dir) ?
10539 &pd.baddr : &pd.naddr;
10540 } else {
10541 x = (s == NULL || sk->direction == dir) ?
10542 &pd.naddr : &pd.baddr;
10543 }
10544 if (x == &pd.baddr || s == NULL) {
10545 if (dir == PF_OUT) {
10546 pd.src = x;
10547 } else {
10548 pd.dst = x;
10549 }
10550 }
10551 }
10552 if (tr->src.addr.type == PF_ADDR_TABLE) {
10553 pfr_update_stats(tr->src.addr.p.tbl, (sk == NULL ||
10554 sk->direction == dir) ? pd.src : pd.dst, pd.af,
10555 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
10556 tr->src.neg);
10557 }
10558 if (tr->dst.addr.type == PF_ADDR_TABLE) {
10559 pfr_update_stats(tr->dst.addr.p.tbl, (sk == NULL ||
10560 sk->direction == dir) ? pd.dst : pd.src, pd.af,
10561 pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
10562 tr->dst.neg);
10563 }
10564 }
10565
10566 VERIFY(pbuf == NULL || pd.mp == NULL || pd.mp == pbuf);
10567
10568 if (*pbufp) {
10569 if (pd.lmw < 0) {
10570 REASON_SET(&reason, PFRES_MEMORY);
10571 action = PF_DROP;
10572 }
10573
10574 if (action == PF_DROP) {
10575 pbuf_destroy(*pbufp);
10576 *pbufp = NULL;
10577 return PF_DROP;
10578 }
10579
10580 *pbufp = pbuf;
10581 }
10582
10583 if (action == PF_SYNPROXY_DROP) {
10584 pbuf_destroy(*pbufp);
10585 *pbufp = NULL;
10586 action = PF_PASS;
10587 } else if (r->rt) {
10588 /* pf_route6 can free the mbuf causing *pbufp to become NULL */
10589 pf_route6(pbufp, r, dir, kif->pfik_ifp, s, &pd);
10590 }
10591
10592 /* if reassembled packet passed, create new fragments */
10593 struct pf_fragment_tag *ftag = NULL;
10594 if ((action == PF_PASS) && (*pbufp != NULL) && (fwd) &&
10595 ((ftag = pf_find_fragment_tag_pbuf(*pbufp)) != NULL)) {
10596 action = pf_refragment6(ifp, pbufp, ftag);
10597 }
10598 return action;
10599 }
10600
10601 static int
10602 pf_check_congestion(struct ifqueue *ifq)
10603 {
10604 #pragma unused(ifq)
10605 return 0;
10606 }
10607
10608 void
10609 pool_init(struct pool *pp, size_t size, unsigned int align, unsigned int ioff,
10610 int flags, const char *wchan, void *palloc)
10611 {
10612 #pragma unused(align, ioff, flags, palloc)
10613 bzero(pp, sizeof(*pp));
10614 pp->pool_zone = zone_create(wchan, size, ZC_DESTRUCTIBLE);
10615 pp->pool_hiwat = pp->pool_limit = (unsigned int)-1;
10616 pp->pool_name = wchan;
10617 }
10618
10619 /* Zones cannot be currently destroyed */
10620 void
10621 pool_destroy(struct pool *pp)
10622 {
10623 #pragma unused(pp)
10624 }
10625
10626 void
10627 pool_sethiwat(struct pool *pp, int n)
10628 {
10629 pp->pool_hiwat = n; /* Currently unused */
10630 }
10631
10632 void
10633 pool_sethardlimit(struct pool *pp, int n, const char *warnmess, int ratecap)
10634 {
10635 #pragma unused(warnmess, ratecap)
10636 pp->pool_limit = n;
10637 }
10638
10639 void *
10640 pool_get(struct pool *pp, int flags)
10641 {
10642 void *buf;
10643
10644 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
10645
10646 if (pp->pool_count > pp->pool_limit) {
10647 DPFPRINTF(PF_DEBUG_NOISY,
10648 ("pf: pool %s hard limit reached (%d)\n",
10649 pp->pool_name != NULL ? pp->pool_name : "unknown",
10650 pp->pool_limit));
10651 pp->pool_fails++;
10652 return NULL;
10653 }
10654
10655 buf = zalloc_flags(pp->pool_zone,
10656 (flags & PR_WAITOK) ? Z_WAITOK : Z_NOWAIT);
10657 if (buf != NULL) {
10658 pp->pool_count++;
10659 VERIFY(pp->pool_count != 0);
10660 }
10661 return buf;
10662 }
10663
10664 void
10665 pool_put(struct pool *pp, void *v)
10666 {
10667 LCK_MTX_ASSERT(pf_lock, LCK_MTX_ASSERT_OWNED);
10668
10669 zfree(pp->pool_zone, v);
10670 VERIFY(pp->pool_count != 0);
10671 pp->pool_count--;
10672 }
10673
10674 struct pf_mtag *
10675 pf_find_mtag_pbuf(pbuf_t *pbuf)
10676 {
10677 return pbuf->pb_pftag;
10678 }
10679
10680 struct pf_mtag *
10681 pf_find_mtag(struct mbuf *m)
10682 {
10683 return m_pftag(m);
10684 }
10685
10686 struct pf_mtag *
10687 pf_get_mtag(struct mbuf *m)
10688 {
10689 return pf_find_mtag(m);
10690 }
10691
10692 struct pf_mtag *
10693 pf_get_mtag_pbuf(pbuf_t *pbuf)
10694 {
10695 return pf_find_mtag_pbuf(pbuf);
10696 }
10697
10698 struct pf_fragment_tag *
10699 pf_copy_fragment_tag(struct mbuf *m, struct pf_fragment_tag *ftag, int how)
10700 {
10701 struct m_tag *tag;
10702 struct pf_mtag *pftag = pf_find_mtag(m);
10703
10704 tag = m_tag_create(KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_PF_REASS,
10705 sizeof(*ftag), how, m);
10706 if (tag == NULL) {
10707 return NULL;
10708 } else {
10709 m_tag_prepend(m, tag);
10710 tag = tag + 1;
10711 }
10712 bcopy(ftag, tag, sizeof(*ftag));
10713 pftag->pftag_flags |= PF_TAG_REASSEMBLED;
10714 return (struct pf_fragment_tag *)tag;
10715 }
10716
10717 struct pf_fragment_tag *
10718 pf_find_fragment_tag(struct mbuf *m)
10719 {
10720 struct m_tag *tag;
10721 struct pf_fragment_tag *ftag;
10722 struct pf_mtag *pftag = pf_find_mtag(m);
10723
10724 tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_PF_REASS,
10725 NULL);
10726 VERIFY((tag == NULL) || (pftag->pftag_flags & PF_TAG_REASSEMBLED));
10727 if (tag != NULL) {
10728 tag = tag + 1;
10729 }
10730 ftag = (struct pf_fragment_tag *)tag;
10731 return ftag;
10732 }
10733
10734 struct pf_fragment_tag *
10735 pf_find_fragment_tag_pbuf(pbuf_t *pbuf)
10736 {
10737 struct pf_mtag *mtag = pf_find_mtag_pbuf(pbuf);
10738
10739 return (mtag->pftag_flags & PF_TAG_REASSEMBLED) ?
10740 pbuf->pb_pf_fragtag : NULL;
10741 }
10742
10743 uint64_t
10744 pf_time_second(void)
10745 {
10746 struct timeval t;
10747
10748 microuptime(&t);
10749 return t.tv_sec;
10750 }
10751
10752 uint64_t
10753 pf_calendar_time_second(void)
10754 {
10755 struct timeval t;
10756
10757 getmicrotime(&t);
10758 return t.tv_sec;
10759 }
10760
10761 static void *
10762 hook_establish(struct hook_desc_head *head, int tail, hook_fn_t fn, void *arg)
10763 {
10764 struct hook_desc *hd;
10765
10766 hd = _MALLOC(sizeof(*hd), M_DEVBUF, M_WAITOK);
10767 if (hd == NULL) {
10768 return NULL;
10769 }
10770
10771 hd->hd_fn = fn;
10772 hd->hd_arg = arg;
10773 if (tail) {
10774 TAILQ_INSERT_TAIL(head, hd, hd_list);
10775 } else {
10776 TAILQ_INSERT_HEAD(head, hd, hd_list);
10777 }
10778
10779 return hd;
10780 }
10781
10782 static void
10783 hook_runloop(struct hook_desc_head *head, int flags)
10784 {
10785 struct hook_desc *hd;
10786
10787 if (!(flags & HOOK_REMOVE)) {
10788 if (!(flags & HOOK_ABORT)) {
10789 TAILQ_FOREACH(hd, head, hd_list)
10790 hd->hd_fn(hd->hd_arg);
10791 }
10792 } else {
10793 while (!!(hd = TAILQ_FIRST(head))) {
10794 TAILQ_REMOVE(head, hd, hd_list);
10795 if (!(flags & HOOK_ABORT)) {
10796 hd->hd_fn(hd->hd_arg);
10797 }
10798 if (flags & HOOK_FREE) {
10799 _FREE(hd, M_DEVBUF);
10800 }
10801 }
10802 }
10803 }
mirror of XNU