xref: /netbsd-src/sys/dist/pf/net/pf_ioctl.c (revision c38e7cc395b1472a774ff828e46123de44c628e9)
1 /*	$NetBSD: pf_ioctl.c,v 1.52 2017/10/15 04:30:05 pgoyette Exp $	*/
2 /*	$OpenBSD: pf_ioctl.c,v 1.182 2007/06/24 11:17:13 mcbride Exp $ */
3 
4 /*
5  * Copyright (c) 2001 Daniel Hartmeier
6  * Copyright (c) 2002,2003 Henning Brauer
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  *
13  *    - Redistributions of source code must retain the above copyright
14  *      notice, this list of conditions and the following disclaimer.
15  *    - Redistributions in binary form must reproduce the above
16  *      copyright notice, this list of conditions and the following
17  *      disclaimer in the documentation and/or other materials provided
18  *      with the distribution.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
26  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
28  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
30  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
31  * POSSIBILITY OF SUCH DAMAGE.
32  *
33  * Effort sponsored in part by the Defense Advanced Research Projects
34  * Agency (DARPA) and Air Force Research Laboratory, Air Force
35  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
36  *
37  */
38 
39 #include <sys/cdefs.h>
40 __KERNEL_RCSID(0, "$NetBSD: pf_ioctl.c,v 1.52 2017/10/15 04:30:05 pgoyette Exp $");
41 
42 #ifdef _KERNEL_OPT
43 #include "opt_inet.h"
44 #endif
45 
46 #include "pfsync.h"
47 
48 #include <sys/param.h>
49 #include <sys/systm.h>
50 #include <sys/mbuf.h>
51 #include <sys/filio.h>
52 #include <sys/fcntl.h>
53 #include <sys/socket.h>
54 #include <sys/socketvar.h>
55 #include <sys/kernel.h>
56 #include <sys/time.h>
57 #include <sys/pool.h>
58 #include <sys/proc.h>
59 #include <sys/malloc.h>
60 #include <sys/kthread.h>
61 #include <sys/rwlock.h>
62 #include <uvm/uvm_extern.h>
63 #ifdef __NetBSD__
64 #include <sys/conf.h>
65 #include <sys/lwp.h>
66 #include <sys/kauth.h>
67 #include <sys/module.h>
68 #include <sys/cprng.h>
69 #include <sys/device.h>
70 #endif /* __NetBSD__ */
71 
72 #include <net/if.h>
73 #include <net/if_types.h>
74 #include <net/route.h>
75 
76 #include <netinet/in.h>
77 #include <netinet/in_var.h>
78 #include <netinet/in_systm.h>
79 #include <netinet/ip.h>
80 #include <netinet/ip_var.h>
81 #include <netinet/ip_icmp.h>
82 
83 #ifndef __NetBSD__
84 #include <dev/rndvar.h>
85 #include <crypto/md5.h>
86 #else
87 #include <sys/md5.h>
88 #endif /* __NetBSD__ */
89 #include <net/pfvar.h>
90 
91 #if NPFSYNC > 0
92 #include <net/if_pfsync.h>
93 #endif /* NPFSYNC > 0 */
94 
95 #if NPFLOG > 0
96 #include <net/if_pflog.h>
97 #endif /* NPFLOG > 0 */
98 
99 #ifdef INET6
100 #include <netinet/ip6.h>
101 #include <netinet/in_pcb.h>
102 #endif /* INET6 */
103 
104 #ifdef ALTQ
105 #include <altq/altq.h>
106 #endif
107 
108 #include "ioconf.h"
109 
110 #ifdef _MODULE
111 void			 pfdetach(void);
112 #endif /* _MODULE */
113 #ifndef __NetBSD__
114 void			 pf_thread_create(void *);
115 #endif /* !__NetBSD__ */
116 int			 pfopen(dev_t, int, int, struct lwp *);
117 int			 pfclose(dev_t, int, int, struct lwp *);
118 struct pf_pool		*pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t,
119 			    u_int8_t, u_int8_t, u_int8_t);
120 
121 void			 pf_mv_pool(struct pf_palist *, struct pf_palist *);
122 void			 pf_empty_pool(struct pf_palist *);
123 int			 pfioctl(dev_t, u_long, void *, int, struct lwp *);
124 #ifdef ALTQ
125 int			 pf_begin_altq(u_int32_t *);
126 int			 pf_rollback_altq(u_int32_t);
127 int			 pf_commit_altq(u_int32_t);
128 int			 pf_enable_altq(struct pf_altq *);
129 int			 pf_disable_altq(struct pf_altq *);
130 #endif /* ALTQ */
131 int			 pf_begin_rules(u_int32_t *, int, const char *);
132 int			 pf_rollback_rules(u_int32_t, int, char *);
133 int			 pf_setup_pfsync_matching(struct pf_ruleset *);
134 void			 pf_hash_rule(MD5_CTX *, struct pf_rule *);
135 void			 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
136 int			 pf_commit_rules(u_int32_t, int, char *);
137 void			 pf_state_export(struct pfsync_state *,
138 			    struct pf_state_key *, struct pf_state *);
139 void			 pf_state_import(struct pfsync_state *,
140 			    struct pf_state_key *, struct pf_state *);
141 
142 static	int		pf_state_add(struct pfsync_state*);
143 
144 struct pf_rule		 pf_default_rule;
145 #ifdef __NetBSD__
146 krwlock_t		 pf_consistency_lock;
147 #else
148 struct rwlock		 pf_consistency_lock = RWLOCK_INITIALIZER("pfcnslk");
149 #endif /* __NetBSD__ */
150 #ifdef ALTQ
151 static int		 pf_altq_running;
152 #endif
153 
154 int		pf_state_lock = 0;
155 
156 #define	TAGID_MAX	 50000
157 TAILQ_HEAD(pf_tags, pf_tagname)	pf_tags = TAILQ_HEAD_INITIALIZER(pf_tags),
158 				pf_qids = TAILQ_HEAD_INITIALIZER(pf_qids);
159 
160 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
161 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
162 #endif
163 u_int16_t		 tagname2tag(struct pf_tags *, char *);
164 void			 tag2tagname(struct pf_tags *, u_int16_t, char *);
165 void			 tag_unref(struct pf_tags *, u_int16_t);
166 int			 pf_rtlabel_add(struct pf_addr_wrap *);
167 void			 pf_rtlabel_remove(struct pf_addr_wrap *);
168 void			 pf_rtlabel_copyout(struct pf_addr_wrap *);
169 
170 #ifdef __NetBSD__
171 void	pf_deferred_init(device_t);
172 #endif
173 
174 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) printf x
175 
176 #ifdef __NetBSD__
177 const struct cdevsw pf_cdevsw = {
178 	.d_open = pfopen,
179 	.d_close = pfclose,
180 	.d_read = noread,
181 	.d_write = nowrite,
182 	.d_ioctl = pfioctl,
183 	.d_stop = nostop,
184 	.d_tty = notty,
185 	.d_poll = nopoll,
186 	.d_mmap = nommap,
187 	.d_kqfilter = nokqfilter,
188 	.d_discard = nodiscard,
189 	.d_flag = D_OTHER
190 };
191 
192 static int pfil4_wrapper(void *, struct mbuf **, struct ifnet *, int);
193 #ifdef INET6
194 static int pfil6_wrapper(void *, struct mbuf **, struct ifnet *, int);
195 #endif /* INET6 */
196 
197 static int pf_pfil_attach(void);
198 static int pf_pfil_detach(void);
199 
200 static int pf_pfil_attached;
201 
202 static kauth_listener_t pf_listener;
203 #endif /* __NetBSD__ */
204 
205 #ifdef __NetBSD__
206 static int
207 pf_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
208     void *arg0, void *arg1, void *arg2, void *arg3)
209 {
210 	int result;
211 	enum kauth_network_req req;
212 
213 	result = KAUTH_RESULT_DEFER;
214 	req = (enum kauth_network_req)arg0;
215 
216 	if (action != KAUTH_NETWORK_FIREWALL)
217 		return result;
218 
219 	/* These must have came from device context. */
220 	if ((req == KAUTH_REQ_NETWORK_FIREWALL_FW) ||
221 	    (req == KAUTH_REQ_NETWORK_FIREWALL_NAT))
222 		result = KAUTH_RESULT_ALLOW;
223 
224 	return result;
225 }
226 #endif /* __NetBSD__ */
227 
228 void
229 pfattach(int num)
230 {
231 	u_int32_t *timeout = pf_default_rule.timeout;
232 
233 #ifdef __NetBSD__
234 	pool_init(&pf_rule_pl, sizeof(struct pf_rule), 0, 0, 0, "pfrulepl",
235 	    &pool_allocator_nointr, IPL_NONE);
236 	pool_init(&pf_src_tree_pl, sizeof(struct pf_src_node), 0, 0, 0,
237 	    "pfsrctrpl", NULL, IPL_SOFTNET);
238 	pool_init(&pf_state_pl, sizeof(struct pf_state), 0, 0, 0, "pfstatepl",
239 	    NULL, IPL_SOFTNET);
240 	pool_init(&pf_state_key_pl, sizeof(struct pf_state_key), 0, 0, 0,
241 	    "pfstatekeypl", NULL, IPL_SOFTNET);
242 	pool_init(&pf_altq_pl, sizeof(struct pf_altq), 0, 0, 0, "pfaltqpl",
243 	    &pool_allocator_nointr, IPL_NONE);
244 	pool_init(&pf_pooladdr_pl, sizeof(struct pf_pooladdr), 0, 0, 0,
245 	    "pfpooladdrpl", &pool_allocator_nointr, IPL_NONE);
246 #else
247 	pool_init(&pf_rule_pl, sizeof(struct pf_rule), 0, 0, 0, "pfrulepl",
248 	    &pool_allocator_nointr);
249 	pool_init(&pf_src_tree_pl, sizeof(struct pf_src_node), 0, 0, 0,
250 	    "pfsrctrpl", NULL);
251 	pool_init(&pf_state_pl, sizeof(struct pf_state), 0, 0, 0, "pfstatepl",
252 	    NULL);
253 	pool_init(&pf_state_key_pl, sizeof(struct pf_state_key), 0, 0, 0,
254 	    "pfstatekeypl", NULL);
255 	pool_init(&pf_altq_pl, sizeof(struct pf_altq), 0, 0, 0, "pfaltqpl",
256 	    &pool_allocator_nointr);
257 	pool_init(&pf_pooladdr_pl, sizeof(struct pf_pooladdr), 0, 0, 0,
258 	    "pfpooladdrpl", &pool_allocator_nointr);
259 #endif /* !__NetBSD__ */
260 
261 	pfr_initialize();
262 	pfi_initialize();
263 	pf_osfp_initialize();
264 
265 	pool_sethardlimit(pf_pool_limits[PF_LIMIT_STATES].pp,
266 	    pf_pool_limits[PF_LIMIT_STATES].limit, NULL, 0);
267 
268 	if (ctob(physmem) <= 100*1024*1024)
269 		pf_pool_limits[PF_LIMIT_TABLE_ENTRIES].limit =
270 		    PFR_KENTRY_HIWAT_SMALL;
271 
272 	RB_INIT(&tree_src_tracking);
273 	RB_INIT(&pf_anchors);
274 	pf_init_ruleset(&pf_main_ruleset);
275 	TAILQ_INIT(&pf_altqs[0]);
276 	TAILQ_INIT(&pf_altqs[1]);
277 	TAILQ_INIT(&pf_pabuf);
278 	pf_altqs_active = &pf_altqs[0];
279 	pf_altqs_inactive = &pf_altqs[1];
280 	TAILQ_INIT(&state_list);
281 
282 #ifdef __NetBSD__
283 	rw_init(&pf_consistency_lock);
284 #endif /* __NetBSD__ */
285 
286 	/* default rule should never be garbage collected */
287 	pf_default_rule.entries.tqe_prev = &pf_default_rule.entries.tqe_next;
288 	pf_default_rule.action = PF_PASS;
289 	pf_default_rule.nr = -1;
290 	pf_default_rule.rtableid = -1;
291 
292 	/* initialize default timeouts */
293 	timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
294 	timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
295 	timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
296 	timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
297 	timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
298 	timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
299 	timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
300 	timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
301 	timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
302 	timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
303 	timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
304 	timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
305 	timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
306 	timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
307 	timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
308 	timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
309 	timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
310 	timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
311 	timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
312 	timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
313 
314 	pf_normalize_init();
315 	bzero(&pf_status, sizeof(pf_status));
316 	pf_status.debug = PF_DEBUG_URGENT;
317 
318 #ifdef __NetBSD__
319 	/*
320 	 * Defer rest of initialization until we can use cprng_fast32()
321 	 * which requires per-CPU data to have been initialized which
322 	 * in turn requires that all CPUs have been discovered and
323 	 * attached!
324 	 */
325 	config_interrupts(NULL, pf_deferred_init);
326 #else
327 	/* XXX do our best to avoid a conflict */
328 	pf_status.hostid = cprng_fast32();
329 
330 	/* require process context to purge states, so perform in a thread */
331 	kthread_create_deferred(pf_thread_create, NULL);
332 #endif /* !__NetBSD__ */
333 
334 #ifdef __NetBSD__
335 	pf_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
336 	    pf_listener_cb, NULL);
337 #endif /* __NetBSD__ */
338 }
339 
340 #ifdef __NetBSD__
341 /* ARGSUSED */
342 void
343 pf_deferred_init(device_t dev)
344 {
345 
346 	/* XXX do our best to avoid a conflict */
347 	pf_status.hostid = cprng_fast32();
348 
349 	/* require process context to purge states, so perform in a thread */
350 	if (kthread_create(PRI_NONE, 0, NULL, pf_purge_thread, NULL, NULL,
351 	    "pfpurge"))
352 		panic("pfpurge thread");
353 }
354 #endif /* __NetBSD__ */
355 
356 #ifdef _MODULE
357 void
358 pfdetach(void)
359 {
360 	extern int		 pf_purge_thread_running;
361 	extern int		 pf_purge_thread_stop;
362 	struct pf_anchor	*anchor;
363 	struct pf_state		*state;
364 	struct pf_src_node	*node;
365 	struct pfioc_table	 pt;
366 	u_int32_t		 ticket;
367 	int			 i;
368 	char			 r = '\0';
369 
370 	pf_purge_thread_stop = 1;
371 	wakeup(pf_purge_thread);
372 
373 	/* wait until the kthread exits */
374 	while (pf_purge_thread_running)
375 		tsleep(&pf_purge_thread_running, PWAIT, "pfdown", 0);
376 
377 	(void)pf_pfil_detach();
378 
379 	pf_status.running = 0;
380 
381 	/* clear the rulesets */
382 	for (i = 0; i < PF_RULESET_MAX; i++)
383 		if (pf_begin_rules(&ticket, i, &r) == 0)
384 			pf_commit_rules(ticket, i, &r);
385 #ifdef ALTQ
386 	if (pf_begin_altq(&ticket) == 0)
387 		pf_commit_altq(ticket);
388 #endif /* ALTQ */
389 
390 	/* clear states */
391 	RB_FOREACH(state, pf_state_tree_id, &tree_id) {
392 		state->timeout = PFTM_PURGE;
393 #if NPFSYNC > 0
394 		state->sync_flags = PFSTATE_NOSYNC;
395 #endif /* NPFSYNC > 0 */
396 	}
397 	pf_purge_expired_states(pf_status.states);
398 #if NPFSYNC > 0
399 	pfsync_clear_states(pf_status.hostid, NULL);
400 #endif /* NPFSYNC > 0 */
401 
402 	/* clear source nodes */
403 	RB_FOREACH(state, pf_state_tree_id, &tree_id) {
404 		state->src_node = NULL;
405 		state->nat_src_node = NULL;
406 	}
407 	RB_FOREACH(node, pf_src_tree, &tree_src_tracking) {
408 		node->expire = 1;
409 		node->states = 0;
410 	}
411 	pf_purge_expired_src_nodes(0);
412 
413 	/* clear tables */
414 	memset(&pt, '\0', sizeof(pt));
415 	pfr_clr_tables(&pt.pfrio_table, &pt.pfrio_ndel, pt.pfrio_flags);
416 
417 	/* destroy anchors */
418 	while ((anchor = RB_MIN(pf_anchor_global, &pf_anchors)) != NULL) {
419 		for (i = 0; i < PF_RULESET_MAX; i++)
420 			if (pf_begin_rules(&ticket, i, anchor->name) == 0)
421 				pf_commit_rules(ticket, i, anchor->name);
422 	}
423 
424 	/* destroy main ruleset */
425 	pf_remove_if_empty_ruleset(&pf_main_ruleset);
426 
427 	/* destroy the pools */
428 	pool_destroy(&pf_pooladdr_pl);
429 	pool_destroy(&pf_altq_pl);
430 	pool_destroy(&pf_state_key_pl);
431 	pool_destroy(&pf_state_pl);
432 	pool_destroy(&pf_rule_pl);
433 	pool_destroy(&pf_src_tree_pl);
434 
435 	rw_destroy(&pf_consistency_lock);
436 
437 	/* destroy subsystems */
438 	pf_normalize_destroy();
439 	pf_osfp_destroy();
440 	pfr_destroy();
441 	pfi_destroy();
442 
443 	/* cleanup kauth listener */
444 	kauth_unlisten_scope(pf_listener);
445 }
446 #endif /* _MODULE */
447 
448 #ifndef __NetBSD__
449 void
450 pf_thread_create(void *v)
451 {
452 	if (kthread_create(pf_purge_thread, NULL, NULL, "pfpurge"))
453 		panic("pfpurge thread");
454 }
455 #endif /* !__NetBSD__ */
456 
457 int
458 pfopen(dev_t dev, int flags, int fmt, struct lwp *l)
459 {
460 	if (minor(dev) >= 1)
461 		return (ENXIO);
462 	return (0);
463 }
464 
465 int
466 pfclose(dev_t dev, int flags, int fmt, struct lwp *l)
467 {
468 	if (minor(dev) >= 1)
469 		return (ENXIO);
470 	return (0);
471 }
472 
473 struct pf_pool *
474 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
475     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
476     u_int8_t check_ticket)
477 {
478 	struct pf_ruleset	*ruleset;
479 	struct pf_rule		*rule;
480 	int			 rs_num;
481 
482 	ruleset = pf_find_ruleset(anchor);
483 	if (ruleset == NULL)
484 		return (NULL);
485 	rs_num = pf_get_ruleset_number(rule_action);
486 	if (rs_num >= PF_RULESET_MAX)
487 		return (NULL);
488 	if (active) {
489 		if (check_ticket && ticket !=
490 		    ruleset->rules[rs_num].active.ticket)
491 			return (NULL);
492 		if (r_last)
493 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
494 			    pf_rulequeue);
495 		else
496 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
497 	} else {
498 		if (check_ticket && ticket !=
499 		    ruleset->rules[rs_num].inactive.ticket)
500 			return (NULL);
501 		if (r_last)
502 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
503 			    pf_rulequeue);
504 		else
505 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
506 	}
507 	if (!r_last) {
508 		while ((rule != NULL) && (rule->nr != rule_number))
509 			rule = TAILQ_NEXT(rule, entries);
510 	}
511 	if (rule == NULL)
512 		return (NULL);
513 
514 	return (&rule->rpool);
515 }
516 
517 void
518 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
519 {
520 	struct pf_pooladdr	*mv_pool_pa;
521 
522 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
523 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
524 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
525 	}
526 }
527 
528 void
529 pf_empty_pool(struct pf_palist *poola)
530 {
531 	struct pf_pooladdr	*empty_pool_pa;
532 
533 	while ((empty_pool_pa = TAILQ_FIRST(poola)) != NULL) {
534 		pfi_dynaddr_remove(&empty_pool_pa->addr);
535 		pf_tbladdr_remove(&empty_pool_pa->addr);
536 		pfi_kif_unref(empty_pool_pa->kif, PFI_KIF_REF_RULE);
537 		TAILQ_REMOVE(poola, empty_pool_pa, entries);
538 		pool_put(&pf_pooladdr_pl, empty_pool_pa);
539 	}
540 }
541 
542 void
543 pf_rm_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
544 {
545 	if (rulequeue != NULL) {
546 		if (rule->states <= 0) {
547 			/*
548 			 * XXX - we need to remove the table *before* detaching
549 			 * the rule to make sure the table code does not delete
550 			 * the anchor under our feet.
551 			 */
552 			pf_tbladdr_remove(&rule->src.addr);
553 			pf_tbladdr_remove(&rule->dst.addr);
554 			if (rule->overload_tbl)
555 				pfr_detach_table(rule->overload_tbl);
556 		}
557 		TAILQ_REMOVE(rulequeue, rule, entries);
558 		rule->entries.tqe_prev = NULL;
559 		rule->nr = -1;
560 	}
561 
562 	if (rule->states > 0 || rule->src_nodes > 0 ||
563 	    rule->entries.tqe_prev != NULL)
564 		return;
565 	pf_tag_unref(rule->tag);
566 	pf_tag_unref(rule->match_tag);
567 #ifdef ALTQ
568 	if (rule->pqid != rule->qid)
569 		pf_qid_unref(rule->pqid);
570 	pf_qid_unref(rule->qid);
571 #endif
572 	pf_rtlabel_remove(&rule->src.addr);
573 	pf_rtlabel_remove(&rule->dst.addr);
574 	pfi_dynaddr_remove(&rule->src.addr);
575 	pfi_dynaddr_remove(&rule->dst.addr);
576 	if (rulequeue == NULL) {
577 		pf_tbladdr_remove(&rule->src.addr);
578 		pf_tbladdr_remove(&rule->dst.addr);
579 		if (rule->overload_tbl)
580 			pfr_detach_table(rule->overload_tbl);
581 	}
582 	pfi_kif_unref(rule->kif, PFI_KIF_REF_RULE);
583 	pf_anchor_remove(rule);
584 	pf_empty_pool(&rule->rpool.list);
585 	pool_put(&pf_rule_pl, rule);
586 }
587 
588 u_int16_t
589 tagname2tag(struct pf_tags *head, char *tagname)
590 {
591 	struct pf_tagname	*tag, *p = NULL;
592 	u_int16_t		 new_tagid = 1;
593 
594 	TAILQ_FOREACH(tag, head, entries)
595 		if (strcmp(tagname, tag->name) == 0) {
596 			tag->ref++;
597 			return (tag->tag);
598 		}
599 
600 	/*
601 	 * to avoid fragmentation, we do a linear search from the beginning
602 	 * and take the first free slot we find. if there is none or the list
603 	 * is empty, append a new entry at the end.
604 	 */
605 
606 	/* new entry */
607 	if (!TAILQ_EMPTY(head))
608 		for (p = TAILQ_FIRST(head); p != NULL &&
609 		    p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
610 			new_tagid = p->tag + 1;
611 
612 	if (new_tagid > TAGID_MAX)
613 		return (0);
614 
615 	/* allocate and fill new struct pf_tagname */
616 	tag = (struct pf_tagname *)malloc(sizeof(struct pf_tagname),
617 	    M_TEMP, M_NOWAIT);
618 	if (tag == NULL)
619 		return (0);
620 	bzero(tag, sizeof(struct pf_tagname));
621 	strlcpy(tag->name, tagname, sizeof(tag->name));
622 	tag->tag = new_tagid;
623 	tag->ref++;
624 
625 	if (p != NULL)	/* insert new entry before p */
626 		TAILQ_INSERT_BEFORE(p, tag, entries);
627 	else	/* either list empty or no free slot in between */
628 		TAILQ_INSERT_TAIL(head, tag, entries);
629 
630 	return (tag->tag);
631 }
632 
633 void
634 tag2tagname(struct pf_tags *head, u_int16_t tagid, char *p)
635 {
636 	struct pf_tagname	*tag;
637 
638 	TAILQ_FOREACH(tag, head, entries)
639 		if (tag->tag == tagid) {
640 			strlcpy(p, tag->name, PF_TAG_NAME_SIZE);
641 			return;
642 		}
643 }
644 
645 void
646 tag_unref(struct pf_tags *head, u_int16_t tag)
647 {
648 	struct pf_tagname	*p, *next;
649 
650 	if (tag == 0)
651 		return;
652 
653 	for (p = TAILQ_FIRST(head); p != NULL; p = next) {
654 		next = TAILQ_NEXT(p, entries);
655 		if (tag == p->tag) {
656 			if (--p->ref == 0) {
657 				TAILQ_REMOVE(head, p, entries);
658 				free(p, M_TEMP);
659 			}
660 			break;
661 		}
662 	}
663 }
664 
665 u_int16_t
666 pf_tagname2tag(char *tagname)
667 {
668 	return (tagname2tag(&pf_tags, tagname));
669 }
670 
671 void
672 pf_tag2tagname(u_int16_t tagid, char *p)
673 {
674 	tag2tagname(&pf_tags, tagid, p);
675 }
676 
677 void
678 pf_tag_ref(u_int16_t tag)
679 {
680 	struct pf_tagname *t;
681 
682 	TAILQ_FOREACH(t, &pf_tags, entries)
683 		if (t->tag == tag)
684 			break;
685 	if (t != NULL)
686 		t->ref++;
687 }
688 
689 void
690 pf_tag_unref(u_int16_t tag)
691 {
692 	tag_unref(&pf_tags, tag);
693 }
694 
695 int
696 pf_rtlabel_add(struct pf_addr_wrap *a)
697 {
698 #ifndef __NetBSD__
699 	if (a->type == PF_ADDR_RTLABEL &&
700 	    (a->v.rtlabel = rtlabel_name2id(a->v.rtlabelname)) == 0)
701 		return (-1);
702 #endif /* !__NetBSD__ */
703 	return (0);
704 }
705 
706 void
707 pf_rtlabel_remove(struct pf_addr_wrap *a)
708 {
709 #ifndef __NetBSD__
710 	if (a->type == PF_ADDR_RTLABEL)
711 		rtlabel_unref(a->v.rtlabel);
712 #endif /* !__NetBSD__ */
713 }
714 
715 void
716 pf_rtlabel_copyout(struct pf_addr_wrap *a)
717 {
718 #ifndef __NetBSD__
719 	const char	*name;
720 
721 	if (a->type == PF_ADDR_RTLABEL && a->v.rtlabel) {
722 		if ((name = rtlabel_id2name(a->v.rtlabel)) == NULL)
723 			strlcpy(a->v.rtlabelname, "?",
724 			    sizeof(a->v.rtlabelname));
725 		else
726 			strlcpy(a->v.rtlabelname, name,
727 			    sizeof(a->v.rtlabelname));
728 	}
729 #endif /* !__NetBSD__ */
730 }
731 
732 #ifdef ALTQ
733 u_int32_t
734 pf_qname2qid(char *qname)
735 {
736 	return ((u_int32_t)tagname2tag(&pf_qids, qname));
737 }
738 
739 void
740 pf_qid2qname(u_int32_t qid, char *p)
741 {
742 	tag2tagname(&pf_qids, (u_int16_t)qid, p);
743 }
744 
745 void
746 pf_qid_unref(u_int32_t qid)
747 {
748 	tag_unref(&pf_qids, (u_int16_t)qid);
749 }
750 
751 int
752 pf_begin_altq(u_int32_t *ticket)
753 {
754 	struct pf_altq	*altq;
755 	int		 error = 0;
756 
757 	/* Purge the old altq list */
758 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
759 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
760 		if (altq->qname[0] == 0) {
761 			/* detach and destroy the discipline */
762 			error = altq_remove(altq);
763 		} else
764 			pf_qid_unref(altq->qid);
765 		pool_put(&pf_altq_pl, altq);
766 	}
767 	if (error)
768 		return (error);
769 	*ticket = ++ticket_altqs_inactive;
770 	altqs_inactive_open = 1;
771 	return (0);
772 }
773 
774 int
775 pf_rollback_altq(u_int32_t ticket)
776 {
777 	struct pf_altq	*altq;
778 	int		 error = 0;
779 
780 	if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
781 		return (0);
782 	/* Purge the old altq list */
783 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
784 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
785 		if (altq->qname[0] == 0) {
786 			/* detach and destroy the discipline */
787 			error = altq_remove(altq);
788 		} else
789 			pf_qid_unref(altq->qid);
790 		pool_put(&pf_altq_pl, altq);
791 	}
792 	altqs_inactive_open = 0;
793 	return (error);
794 }
795 
796 int
797 pf_commit_altq(u_int32_t ticket)
798 {
799 	struct pf_altqqueue	*old_altqs;
800 	struct pf_altq		*altq;
801 	int			 s, err, error = 0;
802 
803 	if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
804 		return (EBUSY);
805 
806 	/* swap altqs, keep the old. */
807 	s = splsoftnet();
808 	old_altqs = pf_altqs_active;
809 	pf_altqs_active = pf_altqs_inactive;
810 	pf_altqs_inactive = old_altqs;
811 	ticket_altqs_active = ticket_altqs_inactive;
812 
813 	/* Attach new disciplines */
814 	TAILQ_FOREACH(altq, pf_altqs_active, entries) {
815 		if (altq->qname[0] == 0) {
816 			/* attach the discipline */
817 			error = altq_pfattach(altq);
818 			if (error == 0 && pf_altq_running)
819 				error = pf_enable_altq(altq);
820 			if (error != 0) {
821 				splx(s);
822 				return (error);
823 			}
824 		}
825 	}
826 
827 	/* Purge the old altq list */
828 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
829 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
830 		if (altq->qname[0] == 0) {
831 			/* detach and destroy the discipline */
832 			if (pf_altq_running)
833 				error = pf_disable_altq(altq);
834 			err = altq_pfdetach(altq);
835 			if (err != 0 && error == 0)
836 				error = err;
837 			err = altq_remove(altq);
838 			if (err != 0 && error == 0)
839 				error = err;
840 		} else
841 			pf_qid_unref(altq->qid);
842 		pool_put(&pf_altq_pl, altq);
843 	}
844 	splx(s);
845 
846 	altqs_inactive_open = 0;
847 	return (error);
848 }
849 
850 int
851 pf_enable_altq(struct pf_altq *altq)
852 {
853 	struct ifnet		*ifp;
854 	struct tb_profile	 tb;
855 	int			 s, error = 0;
856 
857 	if ((ifp = ifunit(altq->ifname)) == NULL)
858 		return (EINVAL);
859 
860 	if (ifp->if_snd.altq_type != ALTQT_NONE)
861 		error = altq_enable(&ifp->if_snd);
862 
863 	/* set tokenbucket regulator */
864 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
865 		tb.rate = altq->ifbandwidth;
866 		tb.depth = altq->tbrsize;
867 		s = splnet();
868 		error = tbr_set(&ifp->if_snd, &tb);
869 		splx(s);
870 	}
871 
872 	return (error);
873 }
874 
875 int
876 pf_disable_altq(struct pf_altq *altq)
877 {
878 	struct ifnet		*ifp;
879 	struct tb_profile	 tb;
880 	int			 s, error;
881 
882 	if ((ifp = ifunit(altq->ifname)) == NULL)
883 		return (EINVAL);
884 
885 	/*
886 	 * when the discipline is no longer referenced, it was overridden
887 	 * by a new one.  if so, just return.
888 	 */
889 	if (altq->altq_disc != ifp->if_snd.altq_disc)
890 		return (0);
891 
892 	error = altq_disable(&ifp->if_snd);
893 
894 	if (error == 0) {
895 		/* clear tokenbucket regulator */
896 		tb.rate = 0;
897 		s = splnet();
898 		error = tbr_set(&ifp->if_snd, &tb);
899 		splx(s);
900 	}
901 
902 	return (error);
903 }
904 #endif /* ALTQ */
905 
906 int
907 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
908 {
909 	struct pf_ruleset	*rs;
910 	struct pf_rule		*rule;
911 
912 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
913 		return (EINVAL);
914 	rs = pf_find_or_create_ruleset(anchor);
915 	if (rs == NULL)
916 		return (EINVAL);
917 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
918 		pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
919 		rs->rules[rs_num].inactive.rcount--;
920 	}
921 	*ticket = ++rs->rules[rs_num].inactive.ticket;
922 	rs->rules[rs_num].inactive.open = 1;
923 	return (0);
924 }
925 
926 int
927 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
928 {
929 	struct pf_ruleset	*rs;
930 	struct pf_rule		*rule;
931 
932 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
933 		return (EINVAL);
934 	rs = pf_find_ruleset(anchor);
935 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
936 	    rs->rules[rs_num].inactive.ticket != ticket)
937 		return (0);
938 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
939 		pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
940 		rs->rules[rs_num].inactive.rcount--;
941 	}
942 	rs->rules[rs_num].inactive.open = 0;
943 	return (0);
944 }
945 
946 #define PF_MD5_UPD(st, elm)						\
947 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
948 
949 #define PF_MD5_UPD_STR(st, elm)						\
950 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
951 
952 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
953 		(stor) = htonl((st)->elm);				\
954 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
955 } while (0)
956 
957 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
958 		(stor) = htons((st)->elm);				\
959 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
960 } while (0)
961 
962 void
963 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
964 {
965 	PF_MD5_UPD(pfr, addr.type);
966 	switch (pfr->addr.type) {
967 		case PF_ADDR_DYNIFTL:
968 			PF_MD5_UPD(pfr, addr.v.ifname);
969 			PF_MD5_UPD(pfr, addr.iflags);
970 			break;
971 		case PF_ADDR_TABLE:
972 			PF_MD5_UPD(pfr, addr.v.tblname);
973 			break;
974 		case PF_ADDR_ADDRMASK:
975 			/* XXX ignore af? */
976 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
977 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
978 			break;
979 		case PF_ADDR_RTLABEL:
980 			PF_MD5_UPD(pfr, addr.v.rtlabelname);
981 			break;
982 	}
983 
984 	PF_MD5_UPD(pfr, port[0]);
985 	PF_MD5_UPD(pfr, port[1]);
986 	PF_MD5_UPD(pfr, neg);
987 	PF_MD5_UPD(pfr, port_op);
988 }
989 
990 void
991 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
992 {
993 	u_int16_t x;
994 	u_int32_t y;
995 
996 	pf_hash_rule_addr(ctx, &rule->src);
997 	pf_hash_rule_addr(ctx, &rule->dst);
998 	PF_MD5_UPD_STR(rule, label);
999 	PF_MD5_UPD_STR(rule, ifname);
1000 	PF_MD5_UPD_STR(rule, match_tagname);
1001 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
1002 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
1003 	PF_MD5_UPD_HTONL(rule, prob, y);
1004 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
1005 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1006 	PF_MD5_UPD(rule, uid.op);
1007 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1008 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1009 	PF_MD5_UPD(rule, gid.op);
1010 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
1011 	PF_MD5_UPD(rule, action);
1012 	PF_MD5_UPD(rule, direction);
1013 	PF_MD5_UPD(rule, af);
1014 	PF_MD5_UPD(rule, quick);
1015 	PF_MD5_UPD(rule, ifnot);
1016 	PF_MD5_UPD(rule, match_tag_not);
1017 	PF_MD5_UPD(rule, natpass);
1018 	PF_MD5_UPD(rule, keep_state);
1019 	PF_MD5_UPD(rule, proto);
1020 	PF_MD5_UPD(rule, type);
1021 	PF_MD5_UPD(rule, code);
1022 	PF_MD5_UPD(rule, flags);
1023 	PF_MD5_UPD(rule, flagset);
1024 	PF_MD5_UPD(rule, allow_opts);
1025 	PF_MD5_UPD(rule, rt);
1026 	PF_MD5_UPD(rule, tos);
1027 }
1028 
1029 int
1030 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1031 {
1032 	struct pf_ruleset	*rs;
1033 	struct pf_rule		*rule, **old_array;
1034 	struct pf_rulequeue	*old_rules;
1035 	int			 s, error;
1036 	u_int32_t		 old_rcount;
1037 
1038 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1039 		return (EINVAL);
1040 	rs = pf_find_ruleset(anchor);
1041 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1042 	    ticket != rs->rules[rs_num].inactive.ticket)
1043 		return (EBUSY);
1044 
1045 	/* Calculate checksum for the main ruleset */
1046 	if (rs == &pf_main_ruleset) {
1047 		error = pf_setup_pfsync_matching(rs);
1048 		if (error != 0)
1049 			return (error);
1050 	}
1051 
1052 	/* Swap rules, keep the old. */
1053 	s = splsoftnet();
1054 	old_rules = rs->rules[rs_num].active.ptr;
1055 	old_rcount = rs->rules[rs_num].active.rcount;
1056 	old_array = rs->rules[rs_num].active.ptr_array;
1057 
1058 	rs->rules[rs_num].active.ptr =
1059 	    rs->rules[rs_num].inactive.ptr;
1060 	rs->rules[rs_num].active.ptr_array =
1061 	    rs->rules[rs_num].inactive.ptr_array;
1062 	rs->rules[rs_num].active.rcount =
1063 	    rs->rules[rs_num].inactive.rcount;
1064 	rs->rules[rs_num].inactive.ptr = old_rules;
1065 	rs->rules[rs_num].inactive.ptr_array = old_array;
1066 	rs->rules[rs_num].inactive.rcount = old_rcount;
1067 
1068 	rs->rules[rs_num].active.ticket =
1069 	    rs->rules[rs_num].inactive.ticket;
1070 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1071 
1072 
1073 	/* Purge the old rule list. */
1074 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1075 		pf_rm_rule(old_rules, rule);
1076 	if (rs->rules[rs_num].inactive.ptr_array)
1077 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1078 	rs->rules[rs_num].inactive.ptr_array = NULL;
1079 	rs->rules[rs_num].inactive.rcount = 0;
1080 	rs->rules[rs_num].inactive.open = 0;
1081 	pf_remove_if_empty_ruleset(rs);
1082 	splx(s);
1083 	return (0);
1084 }
1085 
1086 void
1087 pf_state_export(struct pfsync_state *sp, struct pf_state_key *sk,
1088    struct pf_state *s)
1089 {
1090 	int secs = time_second;
1091 	bzero(sp, sizeof(struct pfsync_state));
1092 
1093 	/* copy from state key */
1094 	sp->lan.addr = sk->lan.addr;
1095 	sp->lan.port = sk->lan.port;
1096 	sp->gwy.addr = sk->gwy.addr;
1097 	sp->gwy.port = sk->gwy.port;
1098 	sp->ext.addr = sk->ext.addr;
1099 	sp->ext.port = sk->ext.port;
1100 	sp->proto = sk->proto;
1101 	sp->af = sk->af;
1102 	sp->direction = sk->direction;
1103 
1104 	/* copy from state */
1105 	memcpy(&sp->id, &s->id, sizeof(sp->id));
1106 	sp->creatorid = s->creatorid;
1107 	strlcpy(sp->ifname, s->kif->pfik_name, sizeof(sp->ifname));
1108 	pf_state_peer_to_pfsync(&s->src, &sp->src);
1109 	pf_state_peer_to_pfsync(&s->dst, &sp->dst);
1110 
1111 	sp->rule = s->rule.ptr->nr;
1112 	sp->nat_rule = (s->nat_rule.ptr == NULL) ?  -1 : s->nat_rule.ptr->nr;
1113 	sp->anchor = (s->anchor.ptr == NULL) ?  -1 : s->anchor.ptr->nr;
1114 
1115 	pf_state_counter_to_pfsync(s->bytes[0], sp->bytes[0]);
1116 	pf_state_counter_to_pfsync(s->bytes[1], sp->bytes[1]);
1117 	pf_state_counter_to_pfsync(s->packets[0], sp->packets[0]);
1118 	pf_state_counter_to_pfsync(s->packets[1], sp->packets[1]);
1119 	sp->creation = secs - s->creation;
1120 	sp->expire = pf_state_expires(s);
1121 	sp->log = s->log;
1122 	sp->allow_opts = s->allow_opts;
1123 	sp->timeout = s->timeout;
1124 
1125 	if (s->src_node)
1126 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
1127 	if (s->nat_src_node)
1128 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
1129 
1130 	if (sp->expire > secs)
1131 		sp->expire -= secs;
1132 	else
1133 		sp->expire = 0;
1134 
1135 }
1136 
1137 void
1138 pf_state_import(struct pfsync_state *sp, struct pf_state_key *sk,
1139    struct pf_state *s)
1140 {
1141 	/* copy to state key */
1142 	sk->lan.addr = sp->lan.addr;
1143 	sk->lan.port = sp->lan.port;
1144 	sk->gwy.addr = sp->gwy.addr;
1145 	sk->gwy.port = sp->gwy.port;
1146 	sk->ext.addr = sp->ext.addr;
1147 	sk->ext.port = sp->ext.port;
1148 	sk->proto = sp->proto;
1149 	sk->af = sp->af;
1150 	sk->direction = sp->direction;
1151 
1152 	/* copy to state */
1153 	memcpy(&s->id, &sp->id, sizeof(sp->id));
1154 	s->creatorid = sp->creatorid;
1155 	pf_state_peer_from_pfsync(&sp->src, &s->src);
1156 	pf_state_peer_from_pfsync(&sp->dst, &s->dst);
1157 
1158 	s->rule.ptr = &pf_default_rule;
1159 	s->rule.ptr->states++;
1160 	s->nat_rule.ptr = NULL;
1161 	s->anchor.ptr = NULL;
1162 	s->rt_kif = NULL;
1163 	s->creation = time_second;
1164 	s->expire = time_second;
1165 	s->timeout = sp->timeout;
1166 	if (sp->expire > 0)
1167 		s->expire -= pf_default_rule.timeout[sp->timeout] - sp->expire;
1168 	s->pfsync_time = 0;
1169 	s->packets[0] = s->packets[1] = 0;
1170 	s->bytes[0] = s->bytes[1] = 0;
1171 }
1172 
1173 int
1174 pf_state_add(struct pfsync_state* sp)
1175 {
1176 	struct pf_state		*s;
1177 	struct pf_state_key	*sk;
1178 	struct pfi_kif		*kif;
1179 
1180 	if (sp->timeout >= PFTM_MAX &&
1181 			sp->timeout != PFTM_UNTIL_PACKET) {
1182 		return EINVAL;
1183 	}
1184 	s = pool_get(&pf_state_pl, PR_NOWAIT);
1185 	if (s == NULL) {
1186 		return ENOMEM;
1187 	}
1188 	bzero(s, sizeof(struct pf_state));
1189 	if ((sk = pf_alloc_state_key(s)) == NULL) {
1190 		pool_put(&pf_state_pl, s);
1191 		return ENOMEM;
1192 	}
1193 	pf_state_import(sp, sk, s);
1194 	kif = pfi_kif_get(sp->ifname);
1195 	if (kif == NULL) {
1196 		pool_put(&pf_state_pl, s);
1197 		pool_put(&pf_state_key_pl, sk);
1198 		return ENOENT;
1199 	}
1200 	if (pf_insert_state(kif, s)) {
1201 		pfi_kif_unref(kif, PFI_KIF_REF_NONE);
1202 		pool_put(&pf_state_pl, s);
1203 		return ENOMEM;
1204 	}
1205 
1206 	return 0;
1207 }
1208 
1209 
1210 int
1211 pf_setup_pfsync_matching(struct pf_ruleset *rs)
1212 {
1213 	MD5_CTX			 ctx;
1214 	struct pf_rule		*rule;
1215 	int			 rs_cnt;
1216 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
1217 
1218 	MD5Init(&ctx);
1219 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1220 		/* XXX PF_RULESET_SCRUB as well? */
1221 		if (rs_cnt == PF_RULESET_SCRUB)
1222 			continue;
1223 
1224 		if (rs->rules[rs_cnt].inactive.ptr_array)
1225 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1226 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
1227 
1228 		if (rs->rules[rs_cnt].inactive.rcount) {
1229 			rs->rules[rs_cnt].inactive.ptr_array =
1230 			    malloc(sizeof(void *) *
1231 			    rs->rules[rs_cnt].inactive.rcount,
1232 			    M_TEMP, M_NOWAIT);
1233 
1234 			if (!rs->rules[rs_cnt].inactive.ptr_array)
1235 				return (ENOMEM);
1236 		}
1237 
1238 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1239 		    entries) {
1240 			pf_hash_rule(&ctx, rule);
1241 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1242 		}
1243 	}
1244 
1245 	MD5Final(digest, &ctx);
1246 	memcpy(pf_status.pf_chksum, digest, sizeof(pf_status.pf_chksum));
1247 	return (0);
1248 }
1249 
1250 int
1251 pfioctl(dev_t dev, u_long cmd, void *addr, int flags, struct lwp *l)
1252 {
1253 	struct pf_pooladdr	*pa = NULL;
1254 	struct pf_pool		*pool = NULL;
1255 	int			 s;
1256 	int			 error = 0;
1257 
1258 	/* XXX keep in sync with switch() below */
1259 	if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FIREWALL,
1260 	    KAUTH_REQ_NETWORK_FIREWALL_FW, NULL, NULL, NULL))
1261 		switch (cmd) {
1262 		case DIOCGETRULES:
1263 		case DIOCGETRULE:
1264 		case DIOCGETADDRS:
1265 		case DIOCGETADDR:
1266 		case DIOCGETSTATE:
1267 		case DIOCSETSTATUSIF:
1268 		case DIOCGETSTATUS:
1269 		case DIOCCLRSTATUS:
1270 		case DIOCNATLOOK:
1271 		case DIOCSETDEBUG:
1272 		case DIOCGETSTATES:
1273 		case DIOCGETTIMEOUT:
1274 		case DIOCCLRRULECTRS:
1275 		case DIOCGETLIMIT:
1276 		case DIOCGETALTQS:
1277 		case DIOCGETALTQ:
1278 		case DIOCGETQSTATS:
1279 		case DIOCGETRULESETS:
1280 		case DIOCGETRULESET:
1281 		case DIOCRGETTABLES:
1282 		case DIOCRGETTSTATS:
1283 		case DIOCRCLRTSTATS:
1284 		case DIOCRCLRADDRS:
1285 		case DIOCRADDADDRS:
1286 		case DIOCRDELADDRS:
1287 		case DIOCRSETADDRS:
1288 		case DIOCRGETADDRS:
1289 		case DIOCRGETASTATS:
1290 		case DIOCRCLRASTATS:
1291 		case DIOCRTSTADDRS:
1292 		case DIOCOSFPGET:
1293 		case DIOCGETSRCNODES:
1294 		case DIOCCLRSRCNODES:
1295 		case DIOCIGETIFACES:
1296 		case DIOCSETIFFLAG:
1297 		case DIOCCLRIFFLAG:
1298 		case DIOCSETLCK:
1299 		case DIOCADDSTATES:
1300 			break;
1301 		case DIOCRCLRTABLES:
1302 		case DIOCRADDTABLES:
1303 		case DIOCRDELTABLES:
1304 		case DIOCRSETTFLAGS:
1305 			if (((struct pfioc_table *)addr)->pfrio_flags &
1306 			    PFR_FLAG_DUMMY)
1307 				break; /* dummy operation ok */
1308 			return (EPERM);
1309 		default:
1310 			return (EPERM);
1311 		}
1312 
1313 	if (!(flags & FWRITE))
1314 		switch (cmd) {
1315 		case DIOCGETRULES:
1316 		case DIOCGETADDRS:
1317 		case DIOCGETADDR:
1318 		case DIOCGETSTATE:
1319 		case DIOCGETSTATUS:
1320 		case DIOCGETSTATES:
1321 		case DIOCGETTIMEOUT:
1322 		case DIOCGETLIMIT:
1323 		case DIOCGETALTQS:
1324 		case DIOCGETALTQ:
1325 		case DIOCGETQSTATS:
1326 		case DIOCGETRULESETS:
1327 		case DIOCGETRULESET:
1328 		case DIOCNATLOOK:
1329 		case DIOCRGETTABLES:
1330 		case DIOCRGETTSTATS:
1331 		case DIOCRGETADDRS:
1332 		case DIOCRGETASTATS:
1333 		case DIOCRTSTADDRS:
1334 		case DIOCOSFPGET:
1335 		case DIOCGETSRCNODES:
1336 		case DIOCIGETIFACES:
1337 		case DIOCSETLCK:
1338 			break;
1339 		case DIOCRCLRTABLES:
1340 		case DIOCRADDTABLES:
1341 		case DIOCRDELTABLES:
1342 		case DIOCRCLRTSTATS:
1343 		case DIOCRCLRADDRS:
1344 		case DIOCRADDADDRS:
1345 		case DIOCRDELADDRS:
1346 		case DIOCRSETADDRS:
1347 		case DIOCRSETTFLAGS:
1348 		case DIOCADDSTATES:
1349 			if (((struct pfioc_table *)addr)->pfrio_flags &
1350 			    PFR_FLAG_DUMMY) {
1351 				flags |= FWRITE; /* need write lock for dummy */
1352 				break; /* dummy operation ok */
1353 			}
1354 			return (EACCES);
1355 		case DIOCGETRULE:
1356 			if (((struct pfioc_rule *)addr)->action == PF_GET_CLR_CNTR)
1357 				return (EACCES);
1358 			break;
1359 		default:
1360 			return (EACCES);
1361 		}
1362 
1363 	if (flags & FWRITE)
1364 		rw_enter_write(&pf_consistency_lock);
1365 	else
1366 		rw_enter_read(&pf_consistency_lock);
1367 
1368 	s = splsoftnet();
1369 	switch (cmd) {
1370 
1371 	case DIOCSTART:
1372 		if (pf_status.running)
1373 			error = EEXIST;
1374 		else {
1375 #ifdef __NetBSD__
1376 			error = pf_pfil_attach();
1377 			if (error)
1378 				break;
1379 #endif /* __NetBSD__ */
1380 			pf_status.running = 1;
1381 			pf_status.since = time_second;
1382 			if (pf_status.stateid == 0) {
1383 				pf_status.stateid = time_second;
1384 				pf_status.stateid = pf_status.stateid << 32;
1385 			}
1386 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1387 		}
1388 		break;
1389 
1390 	case DIOCSTOP:
1391 		if (!pf_status.running)
1392 			error = ENOENT;
1393 		else {
1394 #ifdef __NetBSD__
1395 			error = pf_pfil_detach();
1396 			if (error)
1397 				break;
1398 #endif /* __NetBSD__ */
1399 			pf_status.running = 0;
1400 			pf_status.since = time_second;
1401 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1402 		}
1403 		break;
1404 
1405 	case DIOCADDRULE: {
1406 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1407 		struct pf_ruleset	*ruleset;
1408 		struct pf_rule		*rule, *tail;
1409 		int			 rs_num;
1410 
1411 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1412 		ruleset = pf_find_ruleset(pr->anchor);
1413 		if (ruleset == NULL) {
1414 			error = EINVAL;
1415 			break;
1416 		}
1417 		rs_num = pf_get_ruleset_number(pr->rule.action);
1418 		if (rs_num >= PF_RULESET_MAX) {
1419 			error = EINVAL;
1420 			break;
1421 		}
1422 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1423 			error = EINVAL;
1424 			break;
1425 		}
1426 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1427 			error = EBUSY;
1428 			break;
1429 		}
1430 		if (pr->pool_ticket != ticket_pabuf) {
1431 			error = EBUSY;
1432 			break;
1433 		}
1434 		rule = pool_get(&pf_rule_pl, PR_NOWAIT);
1435 		if (rule == NULL) {
1436 			error = ENOMEM;
1437 			break;
1438 		}
1439 		bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1440 #ifdef __NetBSD__
1441 		rule->cuid = kauth_cred_getuid(l->l_cred);
1442 		rule->cpid = l->l_proc->p_pid;
1443 #else
1444 		rule->cuid = p->p_cred->p_ruid;
1445 		rule->cpid = p->p_pid;
1446 #endif /* !__NetBSD__ */
1447 		rule->anchor = NULL;
1448 		rule->kif = NULL;
1449 		TAILQ_INIT(&rule->rpool.list);
1450 		/* initialize refcounting */
1451 		rule->states = 0;
1452 		rule->src_nodes = 0;
1453 		rule->entries.tqe_prev = NULL;
1454 #ifndef INET
1455 		if (rule->af == AF_INET) {
1456 			pool_put(&pf_rule_pl, rule);
1457 			error = EAFNOSUPPORT;
1458 			break;
1459 		}
1460 #endif /* INET */
1461 #ifndef INET6
1462 		if (rule->af == AF_INET6) {
1463 			pool_put(&pf_rule_pl, rule);
1464 			error = EAFNOSUPPORT;
1465 			break;
1466 		}
1467 #endif /* INET6 */
1468 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1469 		    pf_rulequeue);
1470 		if (tail)
1471 			rule->nr = tail->nr + 1;
1472 		else
1473 			rule->nr = 0;
1474 		if (rule->ifname[0]) {
1475 			rule->kif = pfi_kif_get(rule->ifname);
1476 			if (rule->kif == NULL) {
1477 				pool_put(&pf_rule_pl, rule);
1478 				error = EINVAL;
1479 				break;
1480 			}
1481 			pfi_kif_ref(rule->kif, PFI_KIF_REF_RULE);
1482 		}
1483 
1484 #ifndef __NetBSD__
1485 		if (rule->rtableid > 0 && !rtable_exists(rule->rtableid))
1486 			error = EBUSY;
1487 #endif /* !__NetBSD__ */
1488 
1489 #ifdef ALTQ
1490 		/* set queue IDs */
1491 		if (rule->qname[0] != 0) {
1492 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1493 				error = EBUSY;
1494 			else if (rule->pqname[0] != 0) {
1495 				if ((rule->pqid =
1496 				    pf_qname2qid(rule->pqname)) == 0)
1497 					error = EBUSY;
1498 			} else
1499 				rule->pqid = rule->qid;
1500 		}
1501 #endif
1502 		if (rule->tagname[0])
1503 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1504 				error = EBUSY;
1505 		if (rule->match_tagname[0])
1506 			if ((rule->match_tag =
1507 			    pf_tagname2tag(rule->match_tagname)) == 0)
1508 				error = EBUSY;
1509 		if (rule->rt && !rule->direction)
1510 			error = EINVAL;
1511 #if NPFLOG > 0
1512 		if (!rule->log)
1513 			rule->logif = 0;
1514 		if (rule->logif >= PFLOGIFS_MAX)
1515 			error = EINVAL;
1516 #endif
1517 		if (pf_rtlabel_add(&rule->src.addr) ||
1518 		    pf_rtlabel_add(&rule->dst.addr))
1519 			error = EBUSY;
1520 		if (pfi_dynaddr_setup(&rule->src.addr, rule->af))
1521 			error = EINVAL;
1522 		if (pfi_dynaddr_setup(&rule->dst.addr, rule->af))
1523 			error = EINVAL;
1524 		if (pf_tbladdr_setup(ruleset, &rule->src.addr))
1525 			error = EINVAL;
1526 		if (pf_tbladdr_setup(ruleset, &rule->dst.addr))
1527 			error = EINVAL;
1528 		if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1529 			error = EINVAL;
1530 		TAILQ_FOREACH(pa, &pf_pabuf, entries)
1531 			if (pf_tbladdr_setup(ruleset, &pa->addr))
1532 				error = EINVAL;
1533 
1534 		rule->overload_tbl = NULL;
1535 		if (rule->overload_tblname[0]) {
1536 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1537 			    rule->overload_tblname)) == NULL)
1538 				error = EINVAL;
1539 			else
1540 				rule->overload_tbl->pfrkt_flags |=
1541 				    PFR_TFLAG_ACTIVE;
1542 		}
1543 
1544 		pf_mv_pool(&pf_pabuf, &rule->rpool.list);
1545 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1546 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1547 		    (rule->rt > PF_FASTROUTE)) &&
1548 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1549 			error = EINVAL;
1550 
1551 		if (error) {
1552 			pf_rm_rule(NULL, rule);
1553 			break;
1554 		}
1555 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1556 		rule->evaluations = rule->packets[0] = rule->packets[1] =
1557 		    rule->bytes[0] = rule->bytes[1] = 0;
1558 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1559 		    rule, entries);
1560 		ruleset->rules[rs_num].inactive.rcount++;
1561 		break;
1562 	}
1563 
1564 	case DIOCGETRULES: {
1565 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1566 		struct pf_ruleset	*ruleset;
1567 		struct pf_rule		*tail;
1568 		int			 rs_num;
1569 
1570 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1571 		ruleset = pf_find_ruleset(pr->anchor);
1572 		if (ruleset == NULL) {
1573 			error = EINVAL;
1574 			break;
1575 		}
1576 		rs_num = pf_get_ruleset_number(pr->rule.action);
1577 		if (rs_num >= PF_RULESET_MAX) {
1578 			error = EINVAL;
1579 			break;
1580 		}
1581 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1582 		    pf_rulequeue);
1583 		if (tail)
1584 			pr->nr = tail->nr + 1;
1585 		else
1586 			pr->nr = 0;
1587 		pr->ticket = ruleset->rules[rs_num].active.ticket;
1588 		break;
1589 	}
1590 
1591 	case DIOCGETRULE: {
1592 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1593 		struct pf_ruleset	*ruleset;
1594 		struct pf_rule		*rule;
1595 		int			 rs_num, i;
1596 
1597 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1598 		ruleset = pf_find_ruleset(pr->anchor);
1599 		if (ruleset == NULL) {
1600 			error = EINVAL;
1601 			break;
1602 		}
1603 		rs_num = pf_get_ruleset_number(pr->rule.action);
1604 		if (rs_num >= PF_RULESET_MAX) {
1605 			error = EINVAL;
1606 			break;
1607 		}
1608 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1609 			error = EBUSY;
1610 			break;
1611 		}
1612 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1613 		while ((rule != NULL) && (rule->nr != pr->nr))
1614 			rule = TAILQ_NEXT(rule, entries);
1615 		if (rule == NULL) {
1616 			error = EBUSY;
1617 			break;
1618 		}
1619 		bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1620 		if (pf_anchor_copyout(ruleset, rule, pr)) {
1621 			error = EBUSY;
1622 			break;
1623 		}
1624 		pfi_dynaddr_copyout(&pr->rule.src.addr);
1625 		pfi_dynaddr_copyout(&pr->rule.dst.addr);
1626 		pf_tbladdr_copyout(&pr->rule.src.addr);
1627 		pf_tbladdr_copyout(&pr->rule.dst.addr);
1628 		pf_rtlabel_copyout(&pr->rule.src.addr);
1629 		pf_rtlabel_copyout(&pr->rule.dst.addr);
1630 		for (i = 0; i < PF_SKIP_COUNT; ++i)
1631 			if (rule->skip[i].ptr == NULL)
1632 				pr->rule.skip[i].nr = -1;
1633 			else
1634 				pr->rule.skip[i].nr =
1635 				    rule->skip[i].ptr->nr;
1636 
1637 		if (pr->action == PF_GET_CLR_CNTR) {
1638 			rule->evaluations = 0;
1639 			rule->packets[0] = rule->packets[1] = 0;
1640 			rule->bytes[0] = rule->bytes[1] = 0;
1641 		}
1642 		break;
1643 	}
1644 
1645 	case DIOCCHANGERULE: {
1646 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1647 		struct pf_ruleset	*ruleset;
1648 		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1649 		u_int32_t		 nr = 0;
1650 		int			 rs_num;
1651 
1652 		if (!(pcr->action == PF_CHANGE_REMOVE ||
1653 		    pcr->action == PF_CHANGE_GET_TICKET) &&
1654 		    pcr->pool_ticket != ticket_pabuf) {
1655 			error = EBUSY;
1656 			break;
1657 		}
1658 
1659 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1660 		    pcr->action > PF_CHANGE_GET_TICKET) {
1661 			error = EINVAL;
1662 			break;
1663 		}
1664 		ruleset = pf_find_ruleset(pcr->anchor);
1665 		if (ruleset == NULL) {
1666 			error = EINVAL;
1667 			break;
1668 		}
1669 		rs_num = pf_get_ruleset_number(pcr->rule.action);
1670 		if (rs_num >= PF_RULESET_MAX) {
1671 			error = EINVAL;
1672 			break;
1673 		}
1674 
1675 		if (pcr->action == PF_CHANGE_GET_TICKET) {
1676 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1677 			break;
1678 		} else {
1679 			if (pcr->ticket !=
1680 			    ruleset->rules[rs_num].active.ticket) {
1681 				error = EINVAL;
1682 				break;
1683 			}
1684 			if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1685 				error = EINVAL;
1686 				break;
1687 			}
1688 		}
1689 
1690 		if (pcr->action != PF_CHANGE_REMOVE) {
1691 			newrule = pool_get(&pf_rule_pl, PR_NOWAIT);
1692 			if (newrule == NULL) {
1693 				error = ENOMEM;
1694 				break;
1695 			}
1696 			bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1697 #ifdef __NetBSD__
1698 			newrule->cuid = kauth_cred_getuid(l->l_cred);
1699 			newrule->cpid = l->l_proc->p_pid;
1700 #else
1701 			newrule->cuid = p->p_cred->p_ruid;
1702 			newrule->cpid = p->p_pid;
1703 #endif /* !__NetBSD__ */
1704 			TAILQ_INIT(&newrule->rpool.list);
1705 			/* initialize refcounting */
1706 			newrule->states = 0;
1707 			newrule->entries.tqe_prev = NULL;
1708 #ifndef INET
1709 			if (newrule->af == AF_INET) {
1710 				pool_put(&pf_rule_pl, newrule);
1711 				error = EAFNOSUPPORT;
1712 				break;
1713 			}
1714 #endif /* INET */
1715 #ifndef INET6
1716 			if (newrule->af == AF_INET6) {
1717 				pool_put(&pf_rule_pl, newrule);
1718 				error = EAFNOSUPPORT;
1719 				break;
1720 			}
1721 #endif /* INET6 */
1722 			if (newrule->ifname[0]) {
1723 				newrule->kif = pfi_kif_get(newrule->ifname);
1724 				if (newrule->kif == NULL) {
1725 					pool_put(&pf_rule_pl, newrule);
1726 					error = EINVAL;
1727 					break;
1728 				}
1729 				pfi_kif_ref(newrule->kif, PFI_KIF_REF_RULE);
1730 			} else
1731 				newrule->kif = NULL;
1732 
1733 #ifndef __NetBSD__
1734 			if (newrule->rtableid > 0 &&
1735 			    !rtable_exists(newrule->rtableid))
1736 				error = EBUSY;
1737 #endif /* !__NetBSD__ */
1738 
1739 #ifdef ALTQ
1740 			/* set queue IDs */
1741 			if (newrule->qname[0] != 0) {
1742 				if ((newrule->qid =
1743 				    pf_qname2qid(newrule->qname)) == 0)
1744 					error = EBUSY;
1745 				else if (newrule->pqname[0] != 0) {
1746 					if ((newrule->pqid =
1747 					    pf_qname2qid(newrule->pqname)) == 0)
1748 						error = EBUSY;
1749 				} else
1750 					newrule->pqid = newrule->qid;
1751 			}
1752 #endif /* ALTQ */
1753 			if (newrule->tagname[0])
1754 				if ((newrule->tag =
1755 				    pf_tagname2tag(newrule->tagname)) == 0)
1756 					error = EBUSY;
1757 			if (newrule->match_tagname[0])
1758 				if ((newrule->match_tag = pf_tagname2tag(
1759 				    newrule->match_tagname)) == 0)
1760 					error = EBUSY;
1761 			if (newrule->rt && !newrule->direction)
1762 				error = EINVAL;
1763 #if NPFLOG > 0
1764 			if (!newrule->log)
1765 				newrule->logif = 0;
1766 			if (newrule->logif >= PFLOGIFS_MAX)
1767 				error = EINVAL;
1768 #endif
1769 			if (pf_rtlabel_add(&newrule->src.addr) ||
1770 			    pf_rtlabel_add(&newrule->dst.addr))
1771 				error = EBUSY;
1772 			if (pfi_dynaddr_setup(&newrule->src.addr, newrule->af))
1773 				error = EINVAL;
1774 			if (pfi_dynaddr_setup(&newrule->dst.addr, newrule->af))
1775 				error = EINVAL;
1776 			if (pf_tbladdr_setup(ruleset, &newrule->src.addr))
1777 				error = EINVAL;
1778 			if (pf_tbladdr_setup(ruleset, &newrule->dst.addr))
1779 				error = EINVAL;
1780 			if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1781 				error = EINVAL;
1782 			TAILQ_FOREACH(pa, &pf_pabuf, entries)
1783 				if (pf_tbladdr_setup(ruleset, &pa->addr))
1784 					error = EINVAL;
1785 
1786 			newrule->overload_tbl = NULL;
1787 			if (newrule->overload_tblname[0]) {
1788 				if ((newrule->overload_tbl = pfr_attach_table(
1789 				    ruleset, newrule->overload_tblname)) ==
1790 				    NULL)
1791 					error = EINVAL;
1792 				else
1793 					newrule->overload_tbl->pfrkt_flags |=
1794 					    PFR_TFLAG_ACTIVE;
1795 			}
1796 
1797 			pf_mv_pool(&pf_pabuf, &newrule->rpool.list);
1798 			if (((((newrule->action == PF_NAT) ||
1799 			    (newrule->action == PF_RDR) ||
1800 			    (newrule->action == PF_BINAT) ||
1801 			    (newrule->rt > PF_FASTROUTE)) &&
1802 			    !newrule->anchor)) &&
1803 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1804 				error = EINVAL;
1805 
1806 			if (error) {
1807 				pf_rm_rule(NULL, newrule);
1808 				break;
1809 			}
1810 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1811 			newrule->evaluations = 0;
1812 			newrule->packets[0] = newrule->packets[1] = 0;
1813 			newrule->bytes[0] = newrule->bytes[1] = 0;
1814 		}
1815 		pf_empty_pool(&pf_pabuf);
1816 
1817 		if (pcr->action == PF_CHANGE_ADD_HEAD)
1818 			oldrule = TAILQ_FIRST(
1819 			    ruleset->rules[rs_num].active.ptr);
1820 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1821 			oldrule = TAILQ_LAST(
1822 			    ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1823 		else {
1824 			oldrule = TAILQ_FIRST(
1825 			    ruleset->rules[rs_num].active.ptr);
1826 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1827 				oldrule = TAILQ_NEXT(oldrule, entries);
1828 			if (oldrule == NULL) {
1829 				if (newrule != NULL)
1830 					pf_rm_rule(NULL, newrule);
1831 				error = EINVAL;
1832 				break;
1833 			}
1834 		}
1835 
1836 		if (pcr->action == PF_CHANGE_REMOVE) {
1837 			pf_rm_rule(ruleset->rules[rs_num].active.ptr, oldrule);
1838 			ruleset->rules[rs_num].active.rcount--;
1839 		} else {
1840 			if (oldrule == NULL)
1841 				TAILQ_INSERT_TAIL(
1842 				    ruleset->rules[rs_num].active.ptr,
1843 				    newrule, entries);
1844 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1845 			    pcr->action == PF_CHANGE_ADD_BEFORE)
1846 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1847 			else
1848 				TAILQ_INSERT_AFTER(
1849 				    ruleset->rules[rs_num].active.ptr,
1850 				    oldrule, newrule, entries);
1851 			ruleset->rules[rs_num].active.rcount++;
1852 		}
1853 
1854 		nr = 0;
1855 		TAILQ_FOREACH(oldrule,
1856 		    ruleset->rules[rs_num].active.ptr, entries)
1857 			oldrule->nr = nr++;
1858 
1859 		ruleset->rules[rs_num].active.ticket++;
1860 
1861 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1862 		pf_remove_if_empty_ruleset(ruleset);
1863 
1864 		break;
1865 	}
1866 
1867 	case DIOCCLRSTATES: {
1868 		struct pf_state		*ps, *nexts;
1869 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1870 		int			 killed = 0;
1871 
1872 		for (ps = RB_MIN(pf_state_tree_id, &tree_id); ps; ps = nexts) {
1873 			nexts = RB_NEXT(pf_state_tree_id, &tree_id, ps);
1874 
1875 			if (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1876 			    ps->kif->pfik_name)) {
1877 #if NPFSYNC
1878 				/* don't send out individual delete messages */
1879 				ps->sync_flags = PFSTATE_NOSYNC;
1880 #endif
1881 				pf_unlink_state(ps);
1882 				killed++;
1883 			}
1884 		}
1885 		psk->psk_af = killed;
1886 #if NPFSYNC
1887 		pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
1888 #endif
1889 		break;
1890 	}
1891 
1892 	case DIOCKILLSTATES: {
1893 		struct pf_state		*ps, *nexts;
1894 		struct pf_state_key	*sk;
1895 		struct pf_state_host	*src, *dst;
1896 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
1897 		int			 killed = 0;
1898 
1899 		for (ps = RB_MIN(pf_state_tree_id, &tree_id); ps;
1900 		    ps = nexts) {
1901 			nexts = RB_NEXT(pf_state_tree_id, &tree_id, ps);
1902 			sk = ps->state_key;
1903 
1904 			if (sk->direction == PF_OUT) {
1905 				src = &sk->lan;
1906 				dst = &sk->ext;
1907 			} else {
1908 				src = &sk->ext;
1909 				dst = &sk->lan;
1910 			}
1911 			if ((!psk->psk_af || sk->af == psk->psk_af)
1912 			    && (!psk->psk_proto || psk->psk_proto ==
1913 			    sk->proto) &&
1914 			    PF_MATCHA(psk->psk_src.neg,
1915 			    &psk->psk_src.addr.v.a.addr,
1916 			    &psk->psk_src.addr.v.a.mask,
1917 			    &src->addr, sk->af) &&
1918 			    PF_MATCHA(psk->psk_dst.neg,
1919 			    &psk->psk_dst.addr.v.a.addr,
1920 			    &psk->psk_dst.addr.v.a.mask,
1921 			    &dst->addr, sk->af) &&
1922 			    (psk->psk_src.port_op == 0 ||
1923 			    pf_match_port(psk->psk_src.port_op,
1924 			    psk->psk_src.port[0], psk->psk_src.port[1],
1925 			    src->port)) &&
1926 			    (psk->psk_dst.port_op == 0 ||
1927 			    pf_match_port(psk->psk_dst.port_op,
1928 			    psk->psk_dst.port[0], psk->psk_dst.port[1],
1929 			    dst->port)) &&
1930 			    (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1931 			    ps->kif->pfik_name))) {
1932 #if NPFSYNC > 0
1933 				/* send immediate delete of state */
1934 				pfsync_delete_state(ps);
1935 				ps->sync_flags |= PFSTATE_NOSYNC;
1936 #endif
1937 				pf_unlink_state(ps);
1938 				killed++;
1939 			}
1940 		}
1941 		psk->psk_af = killed;
1942 		break;
1943 	}
1944 
1945 	case DIOCADDSTATE: {
1946 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1947 		struct pfsync_state 	*sp = (struct pfsync_state *)ps->state;
1948 
1949 		error = pf_state_add(sp);
1950 		break;
1951 	}
1952 
1953 	case DIOCADDSTATES: {
1954 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
1955 		struct pfsync_state	*p = (struct pfsync_state *) ps->ps_states;
1956 		struct pfsync_state *pk;
1957 		int size = ps->ps_len;
1958 		int i = 0;
1959 		error = 0;
1960 
1961 		pk = malloc(sizeof(*pk), M_TEMP,M_WAITOK);
1962 
1963 		while (error == 0 && i < size)
1964 		{
1965 			if (copyin(p, pk, sizeof(struct pfsync_state)))
1966 			{
1967 				error = EFAULT;
1968 				free(pk, M_TEMP);
1969 			} else {
1970 				error = pf_state_add(pk);
1971 				i += sizeof(*p);
1972 				p++;
1973 			}
1974 		}
1975 
1976 		free(pk, M_TEMP);
1977 		break;
1978 	}
1979 
1980 
1981 	case DIOCGETSTATE: {
1982 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1983 		struct pf_state		*pfs;
1984 		u_int32_t		 nr;
1985 
1986 		nr = 0;
1987 		RB_FOREACH(pfs, pf_state_tree_id, &tree_id) {
1988 			if (nr >= ps->nr)
1989 				break;
1990 			nr++;
1991 		}
1992 		if (pfs == NULL) {
1993 			error = EBUSY;
1994 			break;
1995 		}
1996 
1997 		pf_state_export((struct pfsync_state *)&ps->state,
1998 		    pfs->state_key, pfs);
1999 		break;
2000 	}
2001 
2002 	case DIOCGETSTATES: {
2003 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
2004 		struct pf_state		*state;
2005 		struct pfsync_state	*p, *pstore;
2006 		u_int32_t		 nr = 0;
2007 
2008 		if (ps->ps_len == 0) {
2009 			nr = pf_status.states;
2010 			ps->ps_len = sizeof(struct pfsync_state) * nr;
2011 			break;
2012 		}
2013 
2014 		pstore = malloc(sizeof(*pstore), M_TEMP, M_WAITOK);
2015 
2016 		p = ps->ps_states;
2017 
2018 		state = TAILQ_FIRST(&state_list);
2019 		while (state) {
2020 			if (state->timeout != PFTM_UNLINKED) {
2021 				if ((nr+1) * sizeof(*p) > (unsigned)ps->ps_len)
2022 					break;
2023 
2024 				pf_state_export(pstore,
2025 				    state->state_key, state);
2026 				error = copyout(pstore, p, sizeof(*p));
2027 				if (error) {
2028 					free(pstore, M_TEMP);
2029 					goto fail;
2030 				}
2031 				p++;
2032 				nr++;
2033 			}
2034 			state = TAILQ_NEXT(state, entry_list);
2035 		}
2036 
2037 		ps->ps_len = sizeof(struct pfsync_state) * nr;
2038 
2039 		free(pstore, M_TEMP);
2040 		break;
2041 	}
2042 
2043 	case DIOCGETSTATUS: {
2044 		struct pf_status *ps = (struct pf_status *)addr;
2045 		bcopy(&pf_status, ps, sizeof(struct pf_status));
2046 		pfi_fill_oldstatus(ps);
2047 		break;
2048 	}
2049 
2050 	case DIOCSETSTATUSIF: {
2051 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
2052 
2053 		if (pi->ifname[0] == 0) {
2054 			bzero(pf_status.ifname, IFNAMSIZ);
2055 			break;
2056 		}
2057 		if (ifunit(pi->ifname) == NULL) {
2058 			error = EINVAL;
2059 			break;
2060 		}
2061 		strlcpy(pf_status.ifname, pi->ifname, IFNAMSIZ);
2062 		break;
2063 	}
2064 
2065 	case DIOCCLRSTATUS: {
2066 		bzero(pf_status.counters, sizeof(pf_status.counters));
2067 		bzero(pf_status.fcounters, sizeof(pf_status.fcounters));
2068 		bzero(pf_status.scounters, sizeof(pf_status.scounters));
2069 		pf_status.since = time_second;
2070 		if (*pf_status.ifname)
2071 			pfi_clr_istats(pf_status.ifname);
2072 		break;
2073 	}
2074 
2075 	case DIOCNATLOOK: {
2076 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
2077 		struct pf_state_key	*sk;
2078 		struct pf_state		*state;
2079 		struct pf_state_key_cmp	 key;
2080 		int			 m = 0, direction = pnl->direction;
2081 
2082 		key.af = pnl->af;
2083 		key.proto = pnl->proto;
2084 
2085 		if (!pnl->proto ||
2086 		    PF_AZERO(&pnl->saddr, pnl->af) ||
2087 		    PF_AZERO(&pnl->daddr, pnl->af) ||
2088 		    ((pnl->proto == IPPROTO_TCP ||
2089 		    pnl->proto == IPPROTO_UDP) &&
2090 		    (!pnl->dport || !pnl->sport)))
2091 			error = EINVAL;
2092 		else {
2093 			/*
2094 			 * userland gives us source and dest of connection,
2095 			 * reverse the lookup so we ask for what happens with
2096 			 * the return traffic, enabling us to find it in the
2097 			 * state tree.
2098 			 */
2099 			if (direction == PF_IN) {
2100 				PF_ACPY(&key.ext.addr, &pnl->daddr, pnl->af);
2101 				key.ext.port = pnl->dport;
2102 				PF_ACPY(&key.gwy.addr, &pnl->saddr, pnl->af);
2103 				key.gwy.port = pnl->sport;
2104 				state = pf_find_state_all(&key, PF_EXT_GWY, &m);
2105 			} else {
2106 				PF_ACPY(&key.lan.addr, &pnl->daddr, pnl->af);
2107 				key.lan.port = pnl->dport;
2108 				PF_ACPY(&key.ext.addr, &pnl->saddr, pnl->af);
2109 				key.ext.port = pnl->sport;
2110 				state = pf_find_state_all(&key, PF_LAN_EXT, &m);
2111 			}
2112 			if (m > 1)
2113 				error = E2BIG;	/* more than one state */
2114 			else if (state != NULL) {
2115 				sk = state->state_key;
2116 				if (direction == PF_IN) {
2117 					PF_ACPY(&pnl->rsaddr, &sk->lan.addr,
2118 					    sk->af);
2119 					pnl->rsport = sk->lan.port;
2120 					PF_ACPY(&pnl->rdaddr, &pnl->daddr,
2121 					    pnl->af);
2122 					pnl->rdport = pnl->dport;
2123 				} else {
2124 					PF_ACPY(&pnl->rdaddr, &sk->gwy.addr,
2125 					    sk->af);
2126 					pnl->rdport = sk->gwy.port;
2127 					PF_ACPY(&pnl->rsaddr, &pnl->saddr,
2128 					    pnl->af);
2129 					pnl->rsport = pnl->sport;
2130 				}
2131 			} else
2132 				error = ENOENT;
2133 		}
2134 		break;
2135 	}
2136 
2137 	case DIOCSETTIMEOUT: {
2138 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2139 		int		 old;
2140 
2141 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
2142 		    pt->seconds < 0) {
2143 			error = EINVAL;
2144 			goto fail;
2145 		}
2146 		old = pf_default_rule.timeout[pt->timeout];
2147 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
2148 			pt->seconds = 1;
2149 		pf_default_rule.timeout[pt->timeout] = pt->seconds;
2150 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
2151 			wakeup(pf_purge_thread);
2152 		pt->seconds = old;
2153 		break;
2154 	}
2155 
2156 	case DIOCGETTIMEOUT: {
2157 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2158 
2159 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
2160 			error = EINVAL;
2161 			goto fail;
2162 		}
2163 		pt->seconds = pf_default_rule.timeout[pt->timeout];
2164 		break;
2165 	}
2166 
2167 	case DIOCGETLIMIT: {
2168 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2169 
2170 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
2171 			error = EINVAL;
2172 			goto fail;
2173 		}
2174 		pl->limit = pf_pool_limits[pl->index].limit;
2175 		break;
2176 	}
2177 
2178 	case DIOCSETLIMIT: {
2179 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2180 		int			 old_limit;
2181 
2182 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
2183 		    pf_pool_limits[pl->index].pp == NULL) {
2184 			error = EINVAL;
2185 			goto fail;
2186 		}
2187 #ifdef __NetBSD__
2188 		pool_sethardlimit(pf_pool_limits[pl->index].pp,
2189 		    pl->limit, NULL, 0);
2190 #else
2191 		if (pool_sethardlimit(pf_pool_limits[pl->index].pp,
2192 		    pl->limit, NULL, 0) != 0) {
2193 			error = EBUSY;
2194 			goto fail;
2195 		}
2196 #endif /* !__NetBSD__ */
2197 		old_limit = pf_pool_limits[pl->index].limit;
2198 		pf_pool_limits[pl->index].limit = pl->limit;
2199 		pl->limit = old_limit;
2200 		break;
2201 	}
2202 
2203 	case DIOCSETDEBUG: {
2204 		u_int32_t	*level = (u_int32_t *)addr;
2205 
2206 		pf_status.debug = *level;
2207 		break;
2208 	}
2209 
2210 	case DIOCCLRRULECTRS: {
2211 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
2212 		struct pf_ruleset	*ruleset = &pf_main_ruleset;
2213 		struct pf_rule		*rule;
2214 
2215 		TAILQ_FOREACH(rule,
2216 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
2217 			rule->evaluations = 0;
2218 			rule->packets[0] = rule->packets[1] = 0;
2219 			rule->bytes[0] = rule->bytes[1] = 0;
2220 		}
2221 		break;
2222 	}
2223 
2224 #ifdef ALTQ
2225 	case DIOCSTARTALTQ: {
2226 		struct pf_altq		*altq;
2227 
2228 		/* enable all altq interfaces on active list */
2229 		TAILQ_FOREACH(altq, pf_altqs_active, entries) {
2230 			if (altq->qname[0] == 0) {
2231 				error = pf_enable_altq(altq);
2232 				if (error != 0)
2233 					break;
2234 			}
2235 		}
2236 		if (error == 0)
2237 			pf_altq_running = 1;
2238 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
2239 		break;
2240 	}
2241 
2242 	case DIOCSTOPALTQ: {
2243 		struct pf_altq		*altq;
2244 
2245 		/* disable all altq interfaces on active list */
2246 		TAILQ_FOREACH(altq, pf_altqs_active, entries) {
2247 			if (altq->qname[0] == 0) {
2248 				error = pf_disable_altq(altq);
2249 				if (error != 0)
2250 					break;
2251 			}
2252 		}
2253 		if (error == 0)
2254 			pf_altq_running = 0;
2255 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2256 		break;
2257 	}
2258 
2259 	case DIOCADDALTQ: {
2260 		struct pfioc_altq	*paa = (struct pfioc_altq *)addr;
2261 		struct pf_altq		*altq, *a;
2262 
2263 		if (paa->ticket != ticket_altqs_inactive) {
2264 			error = EBUSY;
2265 			break;
2266 		}
2267 		altq = pool_get(&pf_altq_pl, PR_NOWAIT);
2268 		if (altq == NULL) {
2269 			error = ENOMEM;
2270 			break;
2271 		}
2272 		bcopy(&paa->altq, altq, sizeof(struct pf_altq));
2273 
2274 		/*
2275 		 * if this is for a queue, find the discipline and
2276 		 * copy the necessary fields
2277 		 */
2278 		if (altq->qname[0] != 0) {
2279 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2280 				error = EBUSY;
2281 				pool_put(&pf_altq_pl, altq);
2282 				break;
2283 			}
2284 			TAILQ_FOREACH(a, pf_altqs_inactive, entries) {
2285 				if (strncmp(a->ifname, altq->ifname,
2286 				    IFNAMSIZ) == 0 && a->qname[0] == 0) {
2287 					altq->altq_disc = a->altq_disc;
2288 					break;
2289 				}
2290 			}
2291 		}
2292 
2293 		error = altq_add(altq);
2294 		if (error) {
2295 			pool_put(&pf_altq_pl, altq);
2296 			break;
2297 		}
2298 
2299 		TAILQ_INSERT_TAIL(pf_altqs_inactive, altq, entries);
2300 		bcopy(altq, &paa->altq, sizeof(struct pf_altq));
2301 		break;
2302 	}
2303 
2304 	case DIOCGETALTQS: {
2305 		struct pfioc_altq	*paa = (struct pfioc_altq *)addr;
2306 		struct pf_altq		*altq;
2307 
2308 		paa->nr = 0;
2309 		TAILQ_FOREACH(altq, pf_altqs_active, entries)
2310 			paa->nr++;
2311 		paa->ticket = ticket_altqs_active;
2312 		break;
2313 	}
2314 
2315 	case DIOCGETALTQ: {
2316 		struct pfioc_altq	*paa = (struct pfioc_altq *)addr;
2317 		struct pf_altq		*altq;
2318 		u_int32_t		 nr;
2319 
2320 		if (paa->ticket != ticket_altqs_active) {
2321 			error = EBUSY;
2322 			break;
2323 		}
2324 		nr = 0;
2325 		altq = TAILQ_FIRST(pf_altqs_active);
2326 		while ((altq != NULL) && (nr < paa->nr)) {
2327 			altq = TAILQ_NEXT(altq, entries);
2328 			nr++;
2329 		}
2330 		if (altq == NULL) {
2331 			error = EBUSY;
2332 			break;
2333 		}
2334 		bcopy(altq, &paa->altq, sizeof(struct pf_altq));
2335 		break;
2336 	}
2337 
2338 	case DIOCCHANGEALTQ:
2339 		/* CHANGEALTQ not supported yet! */
2340 		error = ENODEV;
2341 		break;
2342 
2343 	case DIOCGETQSTATS: {
2344 		struct pfioc_qstats	*pq = (struct pfioc_qstats *)addr;
2345 		struct pf_altq		*altq;
2346 		u_int32_t		 nr;
2347 		int			 nbytes;
2348 
2349 		if (pq->ticket != ticket_altqs_active) {
2350 			error = EBUSY;
2351 			break;
2352 		}
2353 		nbytes = pq->nbytes;
2354 		nr = 0;
2355 		altq = TAILQ_FIRST(pf_altqs_active);
2356 		while ((altq != NULL) && (nr < pq->nr)) {
2357 			altq = TAILQ_NEXT(altq, entries);
2358 			nr++;
2359 		}
2360 		if (altq == NULL) {
2361 			error = EBUSY;
2362 			break;
2363 		}
2364 		error = altq_getqstats(altq, pq->buf, &nbytes);
2365 		if (error == 0) {
2366 			pq->scheduler = altq->scheduler;
2367 			pq->nbytes = nbytes;
2368 		}
2369 		break;
2370 	}
2371 #endif /* ALTQ */
2372 
2373 	case DIOCBEGINADDRS: {
2374 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2375 
2376 		pf_empty_pool(&pf_pabuf);
2377 		pp->ticket = ++ticket_pabuf;
2378 		break;
2379 	}
2380 
2381 	case DIOCADDADDR: {
2382 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2383 
2384 		if (pp->ticket != ticket_pabuf) {
2385 			error = EBUSY;
2386 			break;
2387 		}
2388 #ifndef INET
2389 		if (pp->af == AF_INET) {
2390 			error = EAFNOSUPPORT;
2391 			break;
2392 		}
2393 #endif /* INET */
2394 #ifndef INET6
2395 		if (pp->af == AF_INET6) {
2396 			error = EAFNOSUPPORT;
2397 			break;
2398 		}
2399 #endif /* INET6 */
2400 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2401 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2402 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2403 			error = EINVAL;
2404 			break;
2405 		}
2406 		pa = pool_get(&pf_pooladdr_pl, PR_NOWAIT);
2407 		if (pa == NULL) {
2408 			error = ENOMEM;
2409 			break;
2410 		}
2411 		bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2412 		if (pa->ifname[0]) {
2413 			pa->kif = pfi_kif_get(pa->ifname);
2414 			if (pa->kif == NULL) {
2415 				pool_put(&pf_pooladdr_pl, pa);
2416 				error = EINVAL;
2417 				break;
2418 			}
2419 			pfi_kif_ref(pa->kif, PFI_KIF_REF_RULE);
2420 		}
2421 		if (pfi_dynaddr_setup(&pa->addr, pp->af)) {
2422 			pfi_dynaddr_remove(&pa->addr);
2423 			pfi_kif_unref(pa->kif, PFI_KIF_REF_RULE);
2424 			pool_put(&pf_pooladdr_pl, pa);
2425 			error = EINVAL;
2426 			break;
2427 		}
2428 		TAILQ_INSERT_TAIL(&pf_pabuf, pa, entries);
2429 		break;
2430 	}
2431 
2432 	case DIOCGETADDRS: {
2433 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2434 
2435 		pp->nr = 0;
2436 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2437 		    pp->r_num, 0, 1, 0);
2438 		if (pool == NULL) {
2439 			error = EBUSY;
2440 			break;
2441 		}
2442 		TAILQ_FOREACH(pa, &pool->list, entries)
2443 			pp->nr++;
2444 		break;
2445 	}
2446 
2447 	case DIOCGETADDR: {
2448 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2449 		u_int32_t		 nr = 0;
2450 
2451 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2452 		    pp->r_num, 0, 1, 1);
2453 		if (pool == NULL) {
2454 			error = EBUSY;
2455 			break;
2456 		}
2457 		pa = TAILQ_FIRST(&pool->list);
2458 		while ((pa != NULL) && (nr < pp->nr)) {
2459 			pa = TAILQ_NEXT(pa, entries);
2460 			nr++;
2461 		}
2462 		if (pa == NULL) {
2463 			error = EBUSY;
2464 			break;
2465 		}
2466 		bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2467 		pfi_dynaddr_copyout(&pp->addr.addr);
2468 		pf_tbladdr_copyout(&pp->addr.addr);
2469 		pf_rtlabel_copyout(&pp->addr.addr);
2470 		break;
2471 	}
2472 
2473 	case DIOCCHANGEADDR: {
2474 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
2475 		struct pf_pooladdr	*oldpa = NULL, *newpa = NULL;
2476 		struct pf_ruleset	*ruleset;
2477 
2478 		if (pca->action < PF_CHANGE_ADD_HEAD ||
2479 		    pca->action > PF_CHANGE_REMOVE) {
2480 			error = EINVAL;
2481 			break;
2482 		}
2483 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2484 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2485 		    pca->addr.addr.type != PF_ADDR_TABLE) {
2486 			error = EINVAL;
2487 			break;
2488 		}
2489 
2490 		ruleset = pf_find_ruleset(pca->anchor);
2491 		if (ruleset == NULL) {
2492 			error = EBUSY;
2493 			break;
2494 		}
2495 		pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2496 		    pca->r_num, pca->r_last, 1, 1);
2497 		if (pool == NULL) {
2498 			error = EBUSY;
2499 			break;
2500 		}
2501 		if (pca->action != PF_CHANGE_REMOVE) {
2502 			newpa = pool_get(&pf_pooladdr_pl, PR_NOWAIT);
2503 			if (newpa == NULL) {
2504 				error = ENOMEM;
2505 				break;
2506 			}
2507 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2508 #ifndef INET
2509 			if (pca->af == AF_INET) {
2510 				pool_put(&pf_pooladdr_pl, newpa);
2511 				error = EAFNOSUPPORT;
2512 				break;
2513 			}
2514 #endif /* INET */
2515 #ifndef INET6
2516 			if (pca->af == AF_INET6) {
2517 				pool_put(&pf_pooladdr_pl, newpa);
2518 				error = EAFNOSUPPORT;
2519 				break;
2520 			}
2521 #endif /* INET6 */
2522 			if (newpa->ifname[0]) {
2523 				newpa->kif = pfi_kif_get(newpa->ifname);
2524 				if (newpa->kif == NULL) {
2525 					pool_put(&pf_pooladdr_pl, newpa);
2526 					error = EINVAL;
2527 					break;
2528 				}
2529 				pfi_kif_ref(newpa->kif, PFI_KIF_REF_RULE);
2530 			} else
2531 				newpa->kif = NULL;
2532 			if (pfi_dynaddr_setup(&newpa->addr, pca->af) ||
2533 			    pf_tbladdr_setup(ruleset, &newpa->addr)) {
2534 				pfi_dynaddr_remove(&newpa->addr);
2535 				pfi_kif_unref(newpa->kif, PFI_KIF_REF_RULE);
2536 				pool_put(&pf_pooladdr_pl, newpa);
2537 				error = EINVAL;
2538 				break;
2539 			}
2540 		}
2541 
2542 		if (pca->action == PF_CHANGE_ADD_HEAD)
2543 			oldpa = TAILQ_FIRST(&pool->list);
2544 		else if (pca->action == PF_CHANGE_ADD_TAIL)
2545 			oldpa = TAILQ_LAST(&pool->list, pf_palist);
2546 		else {
2547 			int	i = 0;
2548 
2549 			oldpa = TAILQ_FIRST(&pool->list);
2550 			while ((oldpa != NULL) && (i < pca->nr)) {
2551 				oldpa = TAILQ_NEXT(oldpa, entries);
2552 				i++;
2553 			}
2554 			if (oldpa == NULL) {
2555 				error = EINVAL;
2556 				break;
2557 			}
2558 		}
2559 
2560 		if (pca->action == PF_CHANGE_REMOVE) {
2561 			TAILQ_REMOVE(&pool->list, oldpa, entries);
2562 			pfi_dynaddr_remove(&oldpa->addr);
2563 			pf_tbladdr_remove(&oldpa->addr);
2564 			pfi_kif_unref(oldpa->kif, PFI_KIF_REF_RULE);
2565 			pool_put(&pf_pooladdr_pl, oldpa);
2566 		} else {
2567 			if (oldpa == NULL)
2568 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2569 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
2570 			    pca->action == PF_CHANGE_ADD_BEFORE)
2571 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2572 			else
2573 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
2574 				    newpa, entries);
2575 		}
2576 
2577 		pool->cur = TAILQ_FIRST(&pool->list);
2578 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr,
2579 		    pca->af);
2580 		break;
2581 	}
2582 
2583 	case DIOCGETRULESETS: {
2584 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2585 		struct pf_ruleset	*ruleset;
2586 		struct pf_anchor	*anchor;
2587 
2588 		pr->path[sizeof(pr->path) - 1] = 0;
2589 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2590 			error = EINVAL;
2591 			break;
2592 		}
2593 		pr->nr = 0;
2594 		if (ruleset->anchor == NULL) {
2595 			/* XXX kludge for pf_main_ruleset */
2596 			RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2597 				if (anchor->parent == NULL)
2598 					pr->nr++;
2599 		} else {
2600 			RB_FOREACH(anchor, pf_anchor_node,
2601 			    &ruleset->anchor->children)
2602 				pr->nr++;
2603 		}
2604 		break;
2605 	}
2606 
2607 	case DIOCGETRULESET: {
2608 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2609 		struct pf_ruleset	*ruleset;
2610 		struct pf_anchor	*anchor;
2611 		u_int32_t		 nr = 0;
2612 
2613 		pr->path[sizeof(pr->path) - 1] = 0;
2614 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2615 			error = EINVAL;
2616 			break;
2617 		}
2618 		pr->name[0] = 0;
2619 		if (ruleset->anchor == NULL) {
2620 			/* XXX kludge for pf_main_ruleset */
2621 			RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2622 				if (anchor->parent == NULL && nr++ == pr->nr) {
2623 					strlcpy(pr->name, anchor->name,
2624 					    sizeof(pr->name));
2625 					break;
2626 				}
2627 		} else {
2628 			RB_FOREACH(anchor, pf_anchor_node,
2629 			    &ruleset->anchor->children)
2630 				if (nr++ == pr->nr) {
2631 					strlcpy(pr->name, anchor->name,
2632 					    sizeof(pr->name));
2633 					break;
2634 				}
2635 		}
2636 		if (!pr->name[0])
2637 			error = EBUSY;
2638 		break;
2639 	}
2640 
2641 	case DIOCRCLRTABLES: {
2642 		struct pfioc_table *io = (struct pfioc_table *)addr;
2643 
2644 		if (io->pfrio_esize != 0) {
2645 			error = ENODEV;
2646 			break;
2647 		}
2648 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2649 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2650 		break;
2651 	}
2652 
2653 	case DIOCRADDTABLES: {
2654 		struct pfioc_table *io = (struct pfioc_table *)addr;
2655 
2656 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2657 			error = ENODEV;
2658 			break;
2659 		}
2660 		error = pfr_add_tables(io->pfrio_buffer, io->pfrio_size,
2661 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2662 		break;
2663 	}
2664 
2665 	case DIOCRDELTABLES: {
2666 		struct pfioc_table *io = (struct pfioc_table *)addr;
2667 
2668 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2669 			error = ENODEV;
2670 			break;
2671 		}
2672 		error = pfr_del_tables(io->pfrio_buffer, io->pfrio_size,
2673 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2674 		break;
2675 	}
2676 
2677 	case DIOCRGETTABLES: {
2678 		struct pfioc_table *io = (struct pfioc_table *)addr;
2679 
2680 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2681 			error = ENODEV;
2682 			break;
2683 		}
2684 		error = pfr_get_tables(&io->pfrio_table, io->pfrio_buffer,
2685 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2686 		break;
2687 	}
2688 
2689 	case DIOCRGETTSTATS: {
2690 		struct pfioc_table *io = (struct pfioc_table *)addr;
2691 
2692 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2693 			error = ENODEV;
2694 			break;
2695 		}
2696 		error = pfr_get_tstats(&io->pfrio_table, io->pfrio_buffer,
2697 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2698 		break;
2699 	}
2700 
2701 	case DIOCRCLRTSTATS: {
2702 		struct pfioc_table *io = (struct pfioc_table *)addr;
2703 
2704 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2705 			error = ENODEV;
2706 			break;
2707 		}
2708 		error = pfr_clr_tstats(io->pfrio_buffer, io->pfrio_size,
2709 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2710 		break;
2711 	}
2712 
2713 	case DIOCRSETTFLAGS: {
2714 		struct pfioc_table *io = (struct pfioc_table *)addr;
2715 
2716 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2717 			error = ENODEV;
2718 			break;
2719 		}
2720 		error = pfr_set_tflags(io->pfrio_buffer, io->pfrio_size,
2721 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2722 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2723 		break;
2724 	}
2725 
2726 	case DIOCRCLRADDRS: {
2727 		struct pfioc_table *io = (struct pfioc_table *)addr;
2728 
2729 		if (io->pfrio_esize != 0) {
2730 			error = ENODEV;
2731 			break;
2732 		}
2733 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2734 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2735 		break;
2736 	}
2737 
2738 	case DIOCRADDADDRS: {
2739 		struct pfioc_table *io = (struct pfioc_table *)addr;
2740 
2741 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2742 			error = ENODEV;
2743 			break;
2744 		}
2745 		error = pfr_add_addrs(&io->pfrio_table, io->pfrio_buffer,
2746 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2747 		    PFR_FLAG_USERIOCTL);
2748 		break;
2749 	}
2750 
2751 	case DIOCRDELADDRS: {
2752 		struct pfioc_table *io = (struct pfioc_table *)addr;
2753 
2754 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2755 			error = ENODEV;
2756 			break;
2757 		}
2758 		error = pfr_del_addrs(&io->pfrio_table, io->pfrio_buffer,
2759 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2760 		    PFR_FLAG_USERIOCTL);
2761 		break;
2762 	}
2763 
2764 	case DIOCRSETADDRS: {
2765 		struct pfioc_table *io = (struct pfioc_table *)addr;
2766 
2767 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2768 			error = ENODEV;
2769 			break;
2770 		}
2771 		error = pfr_set_addrs(&io->pfrio_table, io->pfrio_buffer,
2772 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2773 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2774 		    PFR_FLAG_USERIOCTL, 0);
2775 		break;
2776 	}
2777 
2778 	case DIOCRGETADDRS: {
2779 		struct pfioc_table *io = (struct pfioc_table *)addr;
2780 
2781 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2782 			error = ENODEV;
2783 			break;
2784 		}
2785 		error = pfr_get_addrs(&io->pfrio_table, io->pfrio_buffer,
2786 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2787 		break;
2788 	}
2789 
2790 	case DIOCRGETASTATS: {
2791 		struct pfioc_table *io = (struct pfioc_table *)addr;
2792 
2793 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2794 			error = ENODEV;
2795 			break;
2796 		}
2797 		error = pfr_get_astats(&io->pfrio_table, io->pfrio_buffer,
2798 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2799 		break;
2800 	}
2801 
2802 	case DIOCRCLRASTATS: {
2803 		struct pfioc_table *io = (struct pfioc_table *)addr;
2804 
2805 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2806 			error = ENODEV;
2807 			break;
2808 		}
2809 		error = pfr_clr_astats(&io->pfrio_table, io->pfrio_buffer,
2810 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2811 		    PFR_FLAG_USERIOCTL);
2812 		break;
2813 	}
2814 
2815 	case DIOCRTSTADDRS: {
2816 		struct pfioc_table *io = (struct pfioc_table *)addr;
2817 
2818 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2819 			error = ENODEV;
2820 			break;
2821 		}
2822 		error = pfr_tst_addrs(&io->pfrio_table, io->pfrio_buffer,
2823 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2824 		    PFR_FLAG_USERIOCTL);
2825 		break;
2826 	}
2827 
2828 	case DIOCRINADEFINE: {
2829 		struct pfioc_table *io = (struct pfioc_table *)addr;
2830 
2831 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2832 			error = ENODEV;
2833 			break;
2834 		}
2835 		error = pfr_ina_define(&io->pfrio_table, io->pfrio_buffer,
2836 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2837 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2838 		break;
2839 	}
2840 
2841 	case DIOCOSFPADD: {
2842 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2843 		error = pf_osfp_add(io);
2844 		break;
2845 	}
2846 
2847 	case DIOCOSFPGET: {
2848 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2849 		error = pf_osfp_get(io);
2850 		break;
2851 	}
2852 
2853 	case DIOCXBEGIN: {
2854 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2855 		struct pfioc_trans_e	*ioe;
2856 		struct pfr_table	*table;
2857 		int			 i;
2858 
2859 		if (io->esize != sizeof(*ioe)) {
2860 			error = ENODEV;
2861 			goto fail;
2862 		}
2863 		ioe = (struct pfioc_trans_e *)malloc(sizeof(*ioe),
2864 		    M_TEMP, M_WAITOK);
2865 		table = (struct pfr_table *)malloc(sizeof(*table),
2866 		    M_TEMP, M_WAITOK);
2867 		for (i = 0; i < io->size; i++) {
2868 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2869 				free(table, M_TEMP);
2870 				free(ioe, M_TEMP);
2871 				error = EFAULT;
2872 				goto fail;
2873 			}
2874 			switch (ioe->rs_num) {
2875 #ifdef ALTQ
2876 			case PF_RULESET_ALTQ:
2877 				if (ioe->anchor[0]) {
2878 					free(table, M_TEMP);
2879 					free(ioe, M_TEMP);
2880 					error = EINVAL;
2881 					goto fail;
2882 				}
2883 				if ((error = pf_begin_altq(&ioe->ticket))) {
2884 					free(table, M_TEMP);
2885 					free(ioe, M_TEMP);
2886 					goto fail;
2887 				}
2888 				break;
2889 #endif /* ALTQ */
2890 			case PF_RULESET_TABLE:
2891 				bzero(table, sizeof(*table));
2892 				strlcpy(table->pfrt_anchor, ioe->anchor,
2893 				    sizeof(table->pfrt_anchor));
2894 				if ((error = pfr_ina_begin(table,
2895 				    &ioe->ticket, NULL, 0))) {
2896 					free(table, M_TEMP);
2897 					free(ioe, M_TEMP);
2898 					goto fail;
2899 				}
2900 				break;
2901 			default:
2902 				if ((error = pf_begin_rules(&ioe->ticket,
2903 				    ioe->rs_num, ioe->anchor))) {
2904 					free(table, M_TEMP);
2905 					free(ioe, M_TEMP);
2906 					goto fail;
2907 				}
2908 				break;
2909 			}
2910 			if (copyout(ioe, io->array+i, sizeof(io->array[i]))) {
2911 				free(table, M_TEMP);
2912 				free(ioe, M_TEMP);
2913 				error = EFAULT;
2914 				goto fail;
2915 			}
2916 		}
2917 		free(table, M_TEMP);
2918 		free(ioe, M_TEMP);
2919 		break;
2920 	}
2921 
2922 	case DIOCXROLLBACK: {
2923 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2924 		struct pfioc_trans_e	*ioe;
2925 		struct pfr_table	*table;
2926 		int			 i;
2927 
2928 		if (io->esize != sizeof(*ioe)) {
2929 			error = ENODEV;
2930 			goto fail;
2931 		}
2932 		ioe = (struct pfioc_trans_e *)malloc(sizeof(*ioe),
2933 		    M_TEMP, M_WAITOK);
2934 		table = (struct pfr_table *)malloc(sizeof(*table),
2935 		    M_TEMP, M_WAITOK);
2936 		for (i = 0; i < io->size; i++) {
2937 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2938 				free(table, M_TEMP);
2939 				free(ioe, M_TEMP);
2940 				error = EFAULT;
2941 				goto fail;
2942 			}
2943 			switch (ioe->rs_num) {
2944 #ifdef ALTQ
2945 			case PF_RULESET_ALTQ:
2946 				if (ioe->anchor[0]) {
2947 					free(table, M_TEMP);
2948 					free(ioe, M_TEMP);
2949 					error = EINVAL;
2950 					goto fail;
2951 				}
2952 				if ((error = pf_rollback_altq(ioe->ticket))) {
2953 					free(table, M_TEMP);
2954 					free(ioe, M_TEMP);
2955 					goto fail; /* really bad */
2956 				}
2957 				break;
2958 #endif /* ALTQ */
2959 			case PF_RULESET_TABLE:
2960 				bzero(table, sizeof(*table));
2961 				strlcpy(table->pfrt_anchor, ioe->anchor,
2962 				    sizeof(table->pfrt_anchor));
2963 				if ((error = pfr_ina_rollback(table,
2964 				    ioe->ticket, NULL, 0))) {
2965 					free(table, M_TEMP);
2966 					free(ioe, M_TEMP);
2967 					goto fail; /* really bad */
2968 				}
2969 				break;
2970 			default:
2971 				if ((error = pf_rollback_rules(ioe->ticket,
2972 				    ioe->rs_num, ioe->anchor))) {
2973 					free(table, M_TEMP);
2974 					free(ioe, M_TEMP);
2975 					goto fail; /* really bad */
2976 				}
2977 				break;
2978 			}
2979 		}
2980 		free(table, M_TEMP);
2981 		free(ioe, M_TEMP);
2982 		break;
2983 	}
2984 
2985 	case DIOCXCOMMIT: {
2986 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2987 		struct pfioc_trans_e	*ioe;
2988 		struct pfr_table	*table;
2989 		struct pf_ruleset	*rs;
2990 		int			 i;
2991 
2992 		if (io->esize != sizeof(*ioe)) {
2993 			error = ENODEV;
2994 			goto fail;
2995 		}
2996 		ioe = (struct pfioc_trans_e *)malloc(sizeof(*ioe),
2997 		    M_TEMP, M_WAITOK);
2998 		table = (struct pfr_table *)malloc(sizeof(*table),
2999 		    M_TEMP, M_WAITOK);
3000 		/* first makes sure everything will succeed */
3001 		for (i = 0; i < io->size; i++) {
3002 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
3003 				free(table, M_TEMP);
3004 				free(ioe, M_TEMP);
3005 				error = EFAULT;
3006 				goto fail;
3007 			}
3008 			switch (ioe->rs_num) {
3009 #ifdef ALTQ
3010 			case PF_RULESET_ALTQ:
3011 				if (ioe->anchor[0]) {
3012 					free(table, M_TEMP);
3013 					free(ioe, M_TEMP);
3014 					error = EINVAL;
3015 					goto fail;
3016 				}
3017 				if (!altqs_inactive_open || ioe->ticket !=
3018 				    ticket_altqs_inactive) {
3019 					free(table, M_TEMP);
3020 					free(ioe, M_TEMP);
3021 					error = EBUSY;
3022 					goto fail;
3023 				}
3024 				break;
3025 #endif /* ALTQ */
3026 			case PF_RULESET_TABLE:
3027 				rs = pf_find_ruleset(ioe->anchor);
3028 				if (rs == NULL || !rs->topen || ioe->ticket !=
3029 				     rs->tticket) {
3030 					free(table, M_TEMP);
3031 					free(ioe, M_TEMP);
3032 					error = EBUSY;
3033 					goto fail;
3034 				}
3035 				break;
3036 			default:
3037 				if (ioe->rs_num < 0 || ioe->rs_num >=
3038 				    PF_RULESET_MAX) {
3039 					free(table, M_TEMP);
3040 					free(ioe, M_TEMP);
3041 					error = EINVAL;
3042 					goto fail;
3043 				}
3044 				rs = pf_find_ruleset(ioe->anchor);
3045 				if (rs == NULL ||
3046 				    !rs->rules[ioe->rs_num].inactive.open ||
3047 				    rs->rules[ioe->rs_num].inactive.ticket !=
3048 				    ioe->ticket) {
3049 					free(table, M_TEMP);
3050 					free(ioe, M_TEMP);
3051 					error = EBUSY;
3052 					goto fail;
3053 				}
3054 				break;
3055 			}
3056 		}
3057 		/* now do the commit - no errors should happen here */
3058 		for (i = 0; i < io->size; i++) {
3059 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
3060 				free(table, M_TEMP);
3061 				free(ioe, M_TEMP);
3062 				error = EFAULT;
3063 				goto fail;
3064 			}
3065 			switch (ioe->rs_num) {
3066 #ifdef ALTQ
3067 			case PF_RULESET_ALTQ:
3068 				if ((error = pf_commit_altq(ioe->ticket))) {
3069 					free(table, M_TEMP);
3070 					free(ioe, M_TEMP);
3071 					goto fail; /* really bad */
3072 				}
3073 				break;
3074 #endif /* ALTQ */
3075 			case PF_RULESET_TABLE:
3076 				bzero(table, sizeof(*table));
3077 				strlcpy(table->pfrt_anchor, ioe->anchor,
3078 				    sizeof(table->pfrt_anchor));
3079 				if ((error = pfr_ina_commit(table, ioe->ticket,
3080 				    NULL, NULL, 0))) {
3081 					free(table, M_TEMP);
3082 					free(ioe, M_TEMP);
3083 					goto fail; /* really bad */
3084 				}
3085 				break;
3086 			default:
3087 				if ((error = pf_commit_rules(ioe->ticket,
3088 				    ioe->rs_num, ioe->anchor))) {
3089 					free(table, M_TEMP);
3090 					free(ioe, M_TEMP);
3091 					goto fail; /* really bad */
3092 				}
3093 				break;
3094 			}
3095 		}
3096 		free(table, M_TEMP);
3097 		free(ioe, M_TEMP);
3098 		break;
3099 	}
3100 
3101 	case DIOCGETSRCNODES: {
3102 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
3103 		struct pf_src_node	*n, *p, *pstore;
3104 		u_int32_t		 nr = 0;
3105 		int			 space = psn->psn_len;
3106 
3107 		if (space == 0) {
3108 			RB_FOREACH(n, pf_src_tree, &tree_src_tracking)
3109 				nr++;
3110 			psn->psn_len = sizeof(struct pf_src_node) * nr;
3111 			break;
3112 		}
3113 
3114 		pstore = malloc(sizeof(*pstore), M_TEMP, M_WAITOK);
3115 
3116 		p = psn->psn_src_nodes;
3117 		RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
3118 			int	secs = time_second, diff;
3119 
3120 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
3121 				break;
3122 
3123 			bcopy(n, pstore, sizeof(*pstore));
3124 			if (n->rule.ptr != NULL)
3125 				pstore->rule.nr = n->rule.ptr->nr;
3126 			pstore->creation = secs - pstore->creation;
3127 			if (pstore->expire > secs)
3128 				pstore->expire -= secs;
3129 			else
3130 				pstore->expire = 0;
3131 
3132 			/* adjust the connection rate estimate */
3133 			diff = secs - n->conn_rate.last;
3134 			if (diff >= n->conn_rate.seconds)
3135 				pstore->conn_rate.count = 0;
3136 			else
3137 				pstore->conn_rate.count -=
3138 				    n->conn_rate.count * diff /
3139 				    n->conn_rate.seconds;
3140 
3141 			error = copyout(pstore, p, sizeof(*p));
3142 			if (error) {
3143 				free(pstore, M_TEMP);
3144 				goto fail;
3145 			}
3146 			p++;
3147 			nr++;
3148 		}
3149 		psn->psn_len = sizeof(struct pf_src_node) * nr;
3150 
3151 		free(pstore, M_TEMP);
3152 		break;
3153 	}
3154 
3155 	case DIOCCLRSRCNODES: {
3156 		struct pf_src_node	*n;
3157 		struct pf_state		*state;
3158 
3159 		RB_FOREACH(state, pf_state_tree_id, &tree_id) {
3160 			state->src_node = NULL;
3161 			state->nat_src_node = NULL;
3162 		}
3163 		RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
3164 			n->expire = 1;
3165 			n->states = 0;
3166 		}
3167 		pf_purge_expired_src_nodes(1);
3168 		pf_status.src_nodes = 0;
3169 		break;
3170 	}
3171 
3172 	case DIOCKILLSRCNODES: {
3173 		struct pf_src_node	*sn;
3174 		struct pf_state		*ps;
3175 		struct pfioc_src_node_kill *psnk = \
3176 			(struct pfioc_src_node_kill *) addr;
3177 		int			killed = 0;
3178 
3179 		RB_FOREACH(sn, pf_src_tree, &tree_src_tracking) {
3180         		if (PF_MATCHA(psnk->psnk_src.neg, \
3181 				      &psnk->psnk_src.addr.v.a.addr, \
3182 				      &psnk->psnk_src.addr.v.a.mask, \
3183 				      &sn->addr, sn->af) &&
3184 			    PF_MATCHA(psnk->psnk_dst.neg, \
3185 				      &psnk->psnk_dst.addr.v.a.addr, \
3186 				      &psnk->psnk_dst.addr.v.a.mask, \
3187 				      &sn->raddr, sn->af)) {
3188 				/* Handle state to src_node linkage */
3189 				if (sn->states != 0) {
3190 					RB_FOREACH(ps, pf_state_tree_id,
3191 					    &tree_id) {
3192 						if (ps->src_node == sn)
3193 							ps->src_node = NULL;
3194 						if (ps->nat_src_node == sn)
3195 							ps->nat_src_node = NULL;
3196 					}
3197 					sn->states = 0;
3198 				}
3199 				sn->expire = 1;
3200 				killed++;
3201 			}
3202 		}
3203 
3204 		if (killed > 0)
3205 			pf_purge_expired_src_nodes(1);
3206 
3207 		psnk->psnk_af = killed;
3208 		break;
3209 	}
3210 
3211 	case DIOCSETHOSTID: {
3212 		u_int32_t	*hid = (u_int32_t *)addr;
3213 
3214 		if (*hid == 0)
3215 			pf_status.hostid = cprng_fast32();
3216 		else
3217 			pf_status.hostid = *hid;
3218 		break;
3219 	}
3220 
3221 	case DIOCOSFPFLUSH:
3222 		pf_osfp_flush();
3223 		break;
3224 
3225 	case DIOCIGETIFACES: {
3226 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3227 
3228 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
3229 			error = ENODEV;
3230 			break;
3231 		}
3232 		error = pfi_get_ifaces(io->pfiio_name, io->pfiio_buffer,
3233 		    &io->pfiio_size);
3234 		break;
3235 	}
3236 
3237 	case DIOCSETIFFLAG: {
3238 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3239 
3240 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3241 		break;
3242 	}
3243 
3244 	case DIOCCLRIFFLAG: {
3245 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3246 
3247 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3248 		break;
3249 	}
3250 
3251 	case DIOCSETLCK: {
3252 		pf_state_lock = *(uint32_t*)addr;
3253 		break;
3254 	}
3255 
3256 	default:
3257 		error = ENODEV;
3258 		break;
3259 	}
3260 fail:
3261 	splx(s);
3262 	if (flags & FWRITE)
3263 		rw_exit_write(&pf_consistency_lock);
3264 	else
3265 		rw_exit_read(&pf_consistency_lock);
3266 	return (error);
3267 }
3268 
3269 #ifdef __NetBSD__
3270 #ifdef INET
3271 static int
3272 pfil4_wrapper(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
3273 {
3274 	int error;
3275 
3276 	/*
3277 	 * ensure that mbufs are writable beforehand
3278 	 * as it's assumed by pf code.
3279 	 * ip hdr (60 bytes) + tcp hdr (60 bytes) should be enough.
3280 	 * XXX inefficient
3281 	 */
3282 	error = m_makewritable(mp, 0, 60 + 60, M_DONTWAIT);
3283 	if (error) {
3284 		m_freem(*mp);
3285 		*mp = NULL;
3286 		return error;
3287 	}
3288 
3289 	/*
3290 	 * If the packet is out-bound, we can't delay checksums
3291 	 * here.  For in-bound, the checksum has already been
3292 	 * validated.
3293 	 */
3294 	if (dir == PFIL_OUT) {
3295 		if ((*mp)->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
3296 			in_delayed_cksum(*mp);
3297 			(*mp)->m_pkthdr.csum_flags &=
3298 			    ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
3299 		}
3300 	}
3301 
3302 	if (pf_test(dir == PFIL_OUT ? PF_OUT : PF_IN, ifp, mp, NULL)
3303 	    != PF_PASS) {
3304 		m_freem(*mp);
3305 		*mp = NULL;
3306 		return EHOSTUNREACH;
3307 	}
3308 
3309 	/*
3310 	 * we're not compatible with fast-forward.
3311 	 */
3312 
3313 	if (dir == PFIL_IN && *mp) {
3314 		(*mp)->m_flags &= ~M_CANFASTFWD;
3315 	}
3316 
3317 	return (0);
3318 }
3319 #endif /* INET */
3320 
3321 #ifdef INET6
3322 static int
3323 pfil6_wrapper(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
3324 {
3325 	int error;
3326 
3327 	/*
3328 	 * ensure that mbufs are writable beforehand
3329 	 * as it's assumed by pf code.
3330 	 * XXX inefficient
3331 	 */
3332 	error = m_makewritable(mp, 0, M_COPYALL, M_DONTWAIT);
3333 	if (error) {
3334 		m_freem(*mp);
3335 		*mp = NULL;
3336 		return error;
3337 	}
3338 
3339 	/*
3340 	 * If the packet is out-bound, we can't delay checksums
3341 	 * here.  For in-bound, the checksum has already been
3342 	 * validated.
3343 	 */
3344 	if (dir == PFIL_OUT) {
3345 		if ((*mp)->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) {
3346 			in6_delayed_cksum(*mp);
3347 			(*mp)->m_pkthdr.csum_flags &=
3348 			    ~(M_CSUM_TCPv6|M_CSUM_UDPv6);
3349 		}
3350 	}
3351 
3352 	if (pf_test6(dir == PFIL_OUT ? PF_OUT : PF_IN, ifp, mp, NULL)
3353 	    != PF_PASS) {
3354 		m_freem(*mp);
3355 		*mp = NULL;
3356 		return EHOSTUNREACH;
3357 	} else
3358 		return (0);
3359 }
3360 #endif /* INET6 */
3361 
3362 static int
3363 pf_pfil_attach(void)
3364 {
3365 	pfil_head_t *ph_inet;
3366 #ifdef INET6
3367 	pfil_head_t *ph_inet6;
3368 #endif /* INET6 */
3369 	int error;
3370 
3371 	if (pf_pfil_attached)
3372 		return (EBUSY);
3373 
3374 	ph_inet = pfil_head_get(PFIL_TYPE_AF, (void *)AF_INET);
3375 	if (ph_inet)
3376 		error = pfil_add_hook((void *)pfil4_wrapper, NULL,
3377 		    PFIL_IN|PFIL_OUT, ph_inet);
3378 	else
3379 		error = ENOENT;
3380 	if (error)
3381 		return (error);
3382 
3383 #ifdef INET6
3384 	ph_inet6 = pfil_head_get(PFIL_TYPE_AF, (void *)AF_INET6);
3385 	if (ph_inet6)
3386 		error = pfil_add_hook((void *)pfil6_wrapper, NULL,
3387 		    PFIL_IN|PFIL_OUT, ph_inet6);
3388 	else
3389 		error = ENOENT;
3390 	if (error)
3391 		goto bad;
3392 #endif /* INET6 */
3393 
3394 	pf_pfil_attached = 1;
3395 
3396 	return (0);
3397 
3398 #ifdef INET6
3399 bad:
3400 	pfil_remove_hook(pfil4_wrapper, NULL, PFIL_IN|PFIL_OUT, ph_inet);
3401 #endif /* INET6 */
3402 
3403 	return (error);
3404 }
3405 
3406 static int
3407 pf_pfil_detach(void)
3408 {
3409 	pfil_head_t *ph_inet;
3410 #ifdef INET6
3411 	pfil_head_t *ph_inet6;
3412 #endif /* INET6 */
3413 
3414 	if (pf_pfil_attached == 0)
3415 		return (EBUSY);
3416 
3417 	ph_inet = pfil_head_get(PFIL_TYPE_AF, (void *)AF_INET);
3418 	if (ph_inet)
3419 		pfil_remove_hook((void *)pfil4_wrapper, NULL,
3420 		    PFIL_IN|PFIL_OUT, ph_inet);
3421 #ifdef INET6
3422 	ph_inet6 = pfil_head_get(PFIL_TYPE_AF, (void *)AF_INET6);
3423 	if (ph_inet6)
3424 		pfil_remove_hook((void *)pfil6_wrapper, NULL,
3425 		    PFIL_IN|PFIL_OUT, ph_inet6);
3426 #endif /* INET6 */
3427 	pf_pfil_attached = 0;
3428 
3429 	return (0);
3430 }
3431 #endif /* __NetBSD__ */
3432 
3433 #if defined(__NetBSD__)
3434 MODULE(MODULE_CLASS_DRIVER, pf, "bpf");
3435 
3436 static int
3437 pf_modcmd(modcmd_t cmd, void *opaque)
3438 {
3439 #ifdef _MODULE
3440 	extern void pflogattach(int);
3441 	extern void pflogdetach(void);
3442 
3443 	devmajor_t cmajor = NODEVMAJOR, bmajor = NODEVMAJOR;
3444 	int err;
3445 
3446 	switch (cmd) {
3447 	case MODULE_CMD_INIT:
3448 		err = devsw_attach("pf", NULL, &bmajor, &pf_cdevsw, &cmajor);
3449 		if (err)
3450 			return err;
3451 		pfattach(1);
3452 		pflogattach(1);
3453 		return 0;
3454 	case MODULE_CMD_FINI:
3455 		if (pf_status.running) {
3456 			return EBUSY;
3457 		} else {
3458 			pfdetach();
3459 			pflogdetach();
3460 			return devsw_detach(NULL, &pf_cdevsw);
3461 		}
3462 	default:
3463 		return ENOTTY;
3464 	}
3465 #else
3466 	if (cmd == MODULE_CMD_INIT)
3467 		return 0;
3468 	return ENOTTY;
3469 #endif
3470 }
3471 #endif
3472