xref: /netbsd-src/sys/dist/pf/net/pf_ioctl.c (revision 6a493d6bc668897c91594964a732d38505b70cbb)
1 /*	$NetBSD: pf_ioctl.c,v 1.48 2013/07/01 08:32:48 skrll 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 1.48 2013/07/01 08:32:48 skrll 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 #endif /* __NetBSD__ */
70 
71 #include <net/if.h>
72 #include <net/if_types.h>
73 #include <net/route.h>
74 
75 #include <netinet/in.h>
76 #include <netinet/in_var.h>
77 #include <netinet/in_systm.h>
78 #include <netinet/ip.h>
79 #include <netinet/ip_var.h>
80 #include <netinet/ip_icmp.h>
81 
82 #ifndef __NetBSD__
83 #include <dev/rndvar.h>
84 #include <crypto/md5.h>
85 #else
86 #include <sys/md5.h>
87 #endif /* __NetBSD__ */
88 #include <net/pfvar.h>
89 
90 #if NPFSYNC > 0
91 #include <net/if_pfsync.h>
92 #endif /* NPFSYNC > 0 */
93 
94 #if NPFLOG > 0
95 #include <net/if_pflog.h>
96 #endif /* NPFLOG > 0 */
97 
98 #ifdef INET6
99 #include <netinet/ip6.h>
100 #include <netinet/in_pcb.h>
101 #endif /* INET6 */
102 
103 #ifdef ALTQ
104 #include <altq/altq.h>
105 #endif
106 
107 void			 pfattach(int);
108 #ifdef _MODULE
109 void			 pfdetach(void);
110 #endif /* _MODULE */
111 #ifndef __NetBSD__
112 void			 pf_thread_create(void *);
113 #endif /* !__NetBSD__ */
114 int			 pfopen(dev_t, int, int, struct lwp *);
115 int			 pfclose(dev_t, int, int, struct lwp *);
116 struct pf_pool		*pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t,
117 			    u_int8_t, u_int8_t, u_int8_t);
118 
119 void			 pf_mv_pool(struct pf_palist *, struct pf_palist *);
120 void			 pf_empty_pool(struct pf_palist *);
121 int			 pfioctl(dev_t, u_long, void *, int, struct lwp *);
122 #ifdef ALTQ
123 int			 pf_begin_altq(u_int32_t *);
124 int			 pf_rollback_altq(u_int32_t);
125 int			 pf_commit_altq(u_int32_t);
126 int			 pf_enable_altq(struct pf_altq *);
127 int			 pf_disable_altq(struct pf_altq *);
128 #endif /* ALTQ */
129 int			 pf_begin_rules(u_int32_t *, int, const char *);
130 int			 pf_rollback_rules(u_int32_t, int, char *);
131 int			 pf_setup_pfsync_matching(struct pf_ruleset *);
132 void			 pf_hash_rule(MD5_CTX *, struct pf_rule *);
133 void			 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
134 int			 pf_commit_rules(u_int32_t, int, char *);
135 void			 pf_state_export(struct pfsync_state *,
136 			    struct pf_state_key *, struct pf_state *);
137 void			 pf_state_import(struct pfsync_state *,
138 			    struct pf_state_key *, struct pf_state *);
139 
140 static	int		pf_state_add(struct pfsync_state*);
141 
142 struct pf_rule		 pf_default_rule;
143 #ifdef __NetBSD__
144 krwlock_t		 pf_consistency_lock;
145 #else
146 struct rwlock		 pf_consistency_lock = RWLOCK_INITIALIZER("pfcnslk");
147 #endif /* __NetBSD__ */
148 #ifdef ALTQ
149 static int		 pf_altq_running;
150 #endif
151 
152 int		pf_state_lock = 0;
153 
154 #define	TAGID_MAX	 50000
155 TAILQ_HEAD(pf_tags, pf_tagname)	pf_tags = TAILQ_HEAD_INITIALIZER(pf_tags),
156 				pf_qids = TAILQ_HEAD_INITIALIZER(pf_qids);
157 
158 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
159 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
160 #endif
161 u_int16_t		 tagname2tag(struct pf_tags *, char *);
162 void			 tag2tagname(struct pf_tags *, u_int16_t, char *);
163 void			 tag_unref(struct pf_tags *, u_int16_t);
164 int			 pf_rtlabel_add(struct pf_addr_wrap *);
165 void			 pf_rtlabel_remove(struct pf_addr_wrap *);
166 void			 pf_rtlabel_copyout(struct pf_addr_wrap *);
167 
168 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) printf x
169 
170 #ifdef __NetBSD__
171 const struct cdevsw pf_cdevsw = {
172 	pfopen, pfclose, noread, nowrite, pfioctl,
173 	nostop, notty, nopoll, nommap, nokqfilter, D_OTHER
174 };
175 
176 static int pfil4_wrapper(void *, struct mbuf **, struct ifnet *, int);
177 #ifdef INET6
178 static int pfil6_wrapper(void *, struct mbuf **, struct ifnet *, int);
179 #endif /* INET6 */
180 
181 static int pf_pfil_attach(void);
182 static int pf_pfil_detach(void);
183 
184 static int pf_pfil_attached;
185 
186 static kauth_listener_t pf_listener;
187 #endif /* __NetBSD__ */
188 
189 #ifdef __NetBSD__
190 static int
191 pf_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
192     void *arg0, void *arg1, void *arg2, void *arg3)
193 {
194 	int result;
195 	enum kauth_network_req req;
196 
197 	result = KAUTH_RESULT_DEFER;
198 	req = (enum kauth_network_req)arg0;
199 
200 	if (action != KAUTH_NETWORK_FIREWALL)
201 		return result;
202 
203 	/* These must have came from device context. */
204 	if ((req == KAUTH_REQ_NETWORK_FIREWALL_FW) ||
205 	    (req == KAUTH_REQ_NETWORK_FIREWALL_NAT))
206 		result = KAUTH_RESULT_ALLOW;
207 
208 	return result;
209 }
210 #endif /* __NetBSD__ */
211 
212 void
213 pfattach(int num)
214 {
215 	u_int32_t *timeout = pf_default_rule.timeout;
216 
217 #ifdef __NetBSD__
218 	pool_init(&pf_rule_pl, sizeof(struct pf_rule), 0, 0, 0, "pfrulepl",
219 	    &pool_allocator_nointr, IPL_NONE);
220 	pool_init(&pf_src_tree_pl, sizeof(struct pf_src_node), 0, 0, 0,
221 	    "pfsrctrpl", NULL, IPL_SOFTNET);
222 	pool_init(&pf_state_pl, sizeof(struct pf_state), 0, 0, 0, "pfstatepl",
223 	    NULL, IPL_SOFTNET);
224 	pool_init(&pf_state_key_pl, sizeof(struct pf_state_key), 0, 0, 0,
225 	    "pfstatekeypl", NULL, IPL_SOFTNET);
226 	pool_init(&pf_altq_pl, sizeof(struct pf_altq), 0, 0, 0, "pfaltqpl",
227 	    &pool_allocator_nointr, IPL_NONE);
228 	pool_init(&pf_pooladdr_pl, sizeof(struct pf_pooladdr), 0, 0, 0,
229 	    "pfpooladdrpl", &pool_allocator_nointr, IPL_NONE);
230 #else
231 	pool_init(&pf_rule_pl, sizeof(struct pf_rule), 0, 0, 0, "pfrulepl",
232 	    &pool_allocator_nointr);
233 	pool_init(&pf_src_tree_pl, sizeof(struct pf_src_node), 0, 0, 0,
234 	    "pfsrctrpl", NULL);
235 	pool_init(&pf_state_pl, sizeof(struct pf_state), 0, 0, 0, "pfstatepl",
236 	    NULL);
237 	pool_init(&pf_state_key_pl, sizeof(struct pf_state_key), 0, 0, 0,
238 	    "pfstatekeypl", NULL);
239 	pool_init(&pf_altq_pl, sizeof(struct pf_altq), 0, 0, 0, "pfaltqpl",
240 	    &pool_allocator_nointr);
241 	pool_init(&pf_pooladdr_pl, sizeof(struct pf_pooladdr), 0, 0, 0,
242 	    "pfpooladdrpl", &pool_allocator_nointr);
243 #endif /* !__NetBSD__ */
244 
245 	pfr_initialize();
246 	pfi_initialize();
247 	pf_osfp_initialize();
248 
249 	pool_sethardlimit(pf_pool_limits[PF_LIMIT_STATES].pp,
250 	    pf_pool_limits[PF_LIMIT_STATES].limit, NULL, 0);
251 
252 	if (ctob(physmem) <= 100*1024*1024)
253 		pf_pool_limits[PF_LIMIT_TABLE_ENTRIES].limit =
254 		    PFR_KENTRY_HIWAT_SMALL;
255 
256 	RB_INIT(&tree_src_tracking);
257 	RB_INIT(&pf_anchors);
258 	pf_init_ruleset(&pf_main_ruleset);
259 	TAILQ_INIT(&pf_altqs[0]);
260 	TAILQ_INIT(&pf_altqs[1]);
261 	TAILQ_INIT(&pf_pabuf);
262 	pf_altqs_active = &pf_altqs[0];
263 	pf_altqs_inactive = &pf_altqs[1];
264 	TAILQ_INIT(&state_list);
265 
266 #ifdef __NetBSD__
267 	rw_init(&pf_consistency_lock);
268 #endif /* __NetBSD__ */
269 
270 	/* default rule should never be garbage collected */
271 	pf_default_rule.entries.tqe_prev = &pf_default_rule.entries.tqe_next;
272 	pf_default_rule.action = PF_PASS;
273 	pf_default_rule.nr = -1;
274 	pf_default_rule.rtableid = -1;
275 
276 	/* initialize default timeouts */
277 	timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
278 	timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
279 	timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
280 	timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
281 	timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
282 	timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
283 	timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
284 	timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
285 	timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
286 	timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
287 	timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
288 	timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
289 	timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
290 	timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
291 	timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
292 	timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
293 	timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
294 	timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
295 	timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
296 	timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
297 
298 	pf_normalize_init();
299 	bzero(&pf_status, sizeof(pf_status));
300 	pf_status.debug = PF_DEBUG_URGENT;
301 
302 	/* XXX do our best to avoid a conflict */
303 	pf_status.hostid = cprng_fast32();
304 
305 	/* require process context to purge states, so perform in a thread */
306 #ifdef __NetBSD__
307 	if (kthread_create(PRI_NONE, 0, NULL, pf_purge_thread, NULL, NULL,
308 	    "pfpurge"))
309 		panic("pfpurge thread");
310 #else
311 	kthread_create_deferred(pf_thread_create, NULL);
312 #endif /* !__NetBSD__ */
313 
314 #ifdef __NetBSD__
315 	pf_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
316 	    pf_listener_cb, NULL);
317 #endif /* __NetBSD__ */
318 }
319 
320 #ifdef _MODULE
321 void
322 pfdetach(void)
323 {
324 	extern int		 pf_purge_thread_running;
325 	extern int		 pf_purge_thread_stop;
326 	struct pf_anchor	*anchor;
327 	struct pf_state		*state;
328 	struct pf_src_node	*node;
329 	struct pfioc_table	 pt;
330 	u_int32_t		 ticket;
331 	int			 i;
332 	char			 r = '\0';
333 
334 	pf_purge_thread_stop = 1;
335 	wakeup(pf_purge_thread);
336 
337 	/* wait until the kthread exits */
338 	while (pf_purge_thread_running)
339 		tsleep(&pf_purge_thread_running, PWAIT, "pfdown", 0);
340 
341 	(void)pf_pfil_detach();
342 
343 	pf_status.running = 0;
344 
345 	/* clear the rulesets */
346 	for (i = 0; i < PF_RULESET_MAX; i++)
347 		if (pf_begin_rules(&ticket, i, &r) == 0)
348 			pf_commit_rules(ticket, i, &r);
349 #ifdef ALTQ
350 	if (pf_begin_altq(&ticket) == 0)
351 		pf_commit_altq(ticket);
352 #endif /* ALTQ */
353 
354 	/* clear states */
355 	RB_FOREACH(state, pf_state_tree_id, &tree_id) {
356 		state->timeout = PFTM_PURGE;
357 #if NPFSYNC > 0
358 		state->sync_flags = PFSTATE_NOSYNC;
359 #endif /* NPFSYNC > 0 */
360 	}
361 	pf_purge_expired_states(pf_status.states);
362 #if NPFSYNC > 0
363 	pfsync_clear_states(pf_status.hostid, NULL);
364 #endif /* NPFSYNC > 0 */
365 
366 	/* clear source nodes */
367 	RB_FOREACH(state, pf_state_tree_id, &tree_id) {
368 		state->src_node = NULL;
369 		state->nat_src_node = NULL;
370 	}
371 	RB_FOREACH(node, pf_src_tree, &tree_src_tracking) {
372 		node->expire = 1;
373 		node->states = 0;
374 	}
375 	pf_purge_expired_src_nodes(0);
376 
377 	/* clear tables */
378 	memset(&pt, '\0', sizeof(pt));
379 	pfr_clr_tables(&pt.pfrio_table, &pt.pfrio_ndel, pt.pfrio_flags);
380 
381 	/* destroy anchors */
382 	while ((anchor = RB_MIN(pf_anchor_global, &pf_anchors)) != NULL) {
383 		for (i = 0; i < PF_RULESET_MAX; i++)
384 			if (pf_begin_rules(&ticket, i, anchor->name) == 0)
385 				pf_commit_rules(ticket, i, anchor->name);
386 	}
387 
388 	/* destroy main ruleset */
389 	pf_remove_if_empty_ruleset(&pf_main_ruleset);
390 
391 	/* destroy the pools */
392 	pool_destroy(&pf_pooladdr_pl);
393 	pool_destroy(&pf_altq_pl);
394 	pool_destroy(&pf_state_key_pl);
395 	pool_destroy(&pf_state_pl);
396 	pool_destroy(&pf_rule_pl);
397 	pool_destroy(&pf_src_tree_pl);
398 
399 	rw_destroy(&pf_consistency_lock);
400 
401 	/* destroy subsystems */
402 	pf_normalize_destroy();
403 	pf_osfp_destroy();
404 	pfr_destroy();
405 	pfi_destroy();
406 
407 	/* cleanup kauth listener */
408 	kauth_unlisten_scope(pf_listener);
409 }
410 #endif /* _MODULE */
411 
412 #ifndef __NetBSD__
413 void
414 pf_thread_create(void *v)
415 {
416 	if (kthread_create(pf_purge_thread, NULL, NULL, "pfpurge"))
417 		panic("pfpurge thread");
418 }
419 #endif /* !__NetBSD__ */
420 
421 int
422 pfopen(dev_t dev, int flags, int fmt, struct lwp *l)
423 {
424 	if (minor(dev) >= 1)
425 		return (ENXIO);
426 	return (0);
427 }
428 
429 int
430 pfclose(dev_t dev, int flags, int fmt, struct lwp *l)
431 {
432 	if (minor(dev) >= 1)
433 		return (ENXIO);
434 	return (0);
435 }
436 
437 struct pf_pool *
438 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
439     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
440     u_int8_t check_ticket)
441 {
442 	struct pf_ruleset	*ruleset;
443 	struct pf_rule		*rule;
444 	int			 rs_num;
445 
446 	ruleset = pf_find_ruleset(anchor);
447 	if (ruleset == NULL)
448 		return (NULL);
449 	rs_num = pf_get_ruleset_number(rule_action);
450 	if (rs_num >= PF_RULESET_MAX)
451 		return (NULL);
452 	if (active) {
453 		if (check_ticket && ticket !=
454 		    ruleset->rules[rs_num].active.ticket)
455 			return (NULL);
456 		if (r_last)
457 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
458 			    pf_rulequeue);
459 		else
460 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
461 	} else {
462 		if (check_ticket && ticket !=
463 		    ruleset->rules[rs_num].inactive.ticket)
464 			return (NULL);
465 		if (r_last)
466 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
467 			    pf_rulequeue);
468 		else
469 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
470 	}
471 	if (!r_last) {
472 		while ((rule != NULL) && (rule->nr != rule_number))
473 			rule = TAILQ_NEXT(rule, entries);
474 	}
475 	if (rule == NULL)
476 		return (NULL);
477 
478 	return (&rule->rpool);
479 }
480 
481 void
482 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
483 {
484 	struct pf_pooladdr	*mv_pool_pa;
485 
486 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
487 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
488 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
489 	}
490 }
491 
492 void
493 pf_empty_pool(struct pf_palist *poola)
494 {
495 	struct pf_pooladdr	*empty_pool_pa;
496 
497 	while ((empty_pool_pa = TAILQ_FIRST(poola)) != NULL) {
498 		pfi_dynaddr_remove(&empty_pool_pa->addr);
499 		pf_tbladdr_remove(&empty_pool_pa->addr);
500 		pfi_kif_unref(empty_pool_pa->kif, PFI_KIF_REF_RULE);
501 		TAILQ_REMOVE(poola, empty_pool_pa, entries);
502 		pool_put(&pf_pooladdr_pl, empty_pool_pa);
503 	}
504 }
505 
506 void
507 pf_rm_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
508 {
509 	if (rulequeue != NULL) {
510 		if (rule->states <= 0) {
511 			/*
512 			 * XXX - we need to remove the table *before* detaching
513 			 * the rule to make sure the table code does not delete
514 			 * the anchor under our feet.
515 			 */
516 			pf_tbladdr_remove(&rule->src.addr);
517 			pf_tbladdr_remove(&rule->dst.addr);
518 			if (rule->overload_tbl)
519 				pfr_detach_table(rule->overload_tbl);
520 		}
521 		TAILQ_REMOVE(rulequeue, rule, entries);
522 		rule->entries.tqe_prev = NULL;
523 		rule->nr = -1;
524 	}
525 
526 	if (rule->states > 0 || rule->src_nodes > 0 ||
527 	    rule->entries.tqe_prev != NULL)
528 		return;
529 	pf_tag_unref(rule->tag);
530 	pf_tag_unref(rule->match_tag);
531 #ifdef ALTQ
532 	if (rule->pqid != rule->qid)
533 		pf_qid_unref(rule->pqid);
534 	pf_qid_unref(rule->qid);
535 #endif
536 	pf_rtlabel_remove(&rule->src.addr);
537 	pf_rtlabel_remove(&rule->dst.addr);
538 	pfi_dynaddr_remove(&rule->src.addr);
539 	pfi_dynaddr_remove(&rule->dst.addr);
540 	if (rulequeue == NULL) {
541 		pf_tbladdr_remove(&rule->src.addr);
542 		pf_tbladdr_remove(&rule->dst.addr);
543 		if (rule->overload_tbl)
544 			pfr_detach_table(rule->overload_tbl);
545 	}
546 	pfi_kif_unref(rule->kif, PFI_KIF_REF_RULE);
547 	pf_anchor_remove(rule);
548 	pf_empty_pool(&rule->rpool.list);
549 	pool_put(&pf_rule_pl, rule);
550 }
551 
552 u_int16_t
553 tagname2tag(struct pf_tags *head, char *tagname)
554 {
555 	struct pf_tagname	*tag, *p = NULL;
556 	u_int16_t		 new_tagid = 1;
557 
558 	TAILQ_FOREACH(tag, head, entries)
559 		if (strcmp(tagname, tag->name) == 0) {
560 			tag->ref++;
561 			return (tag->tag);
562 		}
563 
564 	/*
565 	 * to avoid fragmentation, we do a linear search from the beginning
566 	 * and take the first free slot we find. if there is none or the list
567 	 * is empty, append a new entry at the end.
568 	 */
569 
570 	/* new entry */
571 	if (!TAILQ_EMPTY(head))
572 		for (p = TAILQ_FIRST(head); p != NULL &&
573 		    p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
574 			new_tagid = p->tag + 1;
575 
576 	if (new_tagid > TAGID_MAX)
577 		return (0);
578 
579 	/* allocate and fill new struct pf_tagname */
580 	tag = (struct pf_tagname *)malloc(sizeof(struct pf_tagname),
581 	    M_TEMP, M_NOWAIT);
582 	if (tag == NULL)
583 		return (0);
584 	bzero(tag, sizeof(struct pf_tagname));
585 	strlcpy(tag->name, tagname, sizeof(tag->name));
586 	tag->tag = new_tagid;
587 	tag->ref++;
588 
589 	if (p != NULL)	/* insert new entry before p */
590 		TAILQ_INSERT_BEFORE(p, tag, entries);
591 	else	/* either list empty or no free slot in between */
592 		TAILQ_INSERT_TAIL(head, tag, entries);
593 
594 	return (tag->tag);
595 }
596 
597 void
598 tag2tagname(struct pf_tags *head, u_int16_t tagid, char *p)
599 {
600 	struct pf_tagname	*tag;
601 
602 	TAILQ_FOREACH(tag, head, entries)
603 		if (tag->tag == tagid) {
604 			strlcpy(p, tag->name, PF_TAG_NAME_SIZE);
605 			return;
606 		}
607 }
608 
609 void
610 tag_unref(struct pf_tags *head, u_int16_t tag)
611 {
612 	struct pf_tagname	*p, *next;
613 
614 	if (tag == 0)
615 		return;
616 
617 	for (p = TAILQ_FIRST(head); p != NULL; p = next) {
618 		next = TAILQ_NEXT(p, entries);
619 		if (tag == p->tag) {
620 			if (--p->ref == 0) {
621 				TAILQ_REMOVE(head, p, entries);
622 				free(p, M_TEMP);
623 			}
624 			break;
625 		}
626 	}
627 }
628 
629 u_int16_t
630 pf_tagname2tag(char *tagname)
631 {
632 	return (tagname2tag(&pf_tags, tagname));
633 }
634 
635 void
636 pf_tag2tagname(u_int16_t tagid, char *p)
637 {
638 	tag2tagname(&pf_tags, tagid, p);
639 }
640 
641 void
642 pf_tag_ref(u_int16_t tag)
643 {
644 	struct pf_tagname *t;
645 
646 	TAILQ_FOREACH(t, &pf_tags, entries)
647 		if (t->tag == tag)
648 			break;
649 	if (t != NULL)
650 		t->ref++;
651 }
652 
653 void
654 pf_tag_unref(u_int16_t tag)
655 {
656 	tag_unref(&pf_tags, tag);
657 }
658 
659 int
660 pf_rtlabel_add(struct pf_addr_wrap *a)
661 {
662 #ifndef __NetBSD__
663 	if (a->type == PF_ADDR_RTLABEL &&
664 	    (a->v.rtlabel = rtlabel_name2id(a->v.rtlabelname)) == 0)
665 		return (-1);
666 #endif /* !__NetBSD__ */
667 	return (0);
668 }
669 
670 void
671 pf_rtlabel_remove(struct pf_addr_wrap *a)
672 {
673 #ifndef __NetBSD__
674 	if (a->type == PF_ADDR_RTLABEL)
675 		rtlabel_unref(a->v.rtlabel);
676 #endif /* !__NetBSD__ */
677 }
678 
679 void
680 pf_rtlabel_copyout(struct pf_addr_wrap *a)
681 {
682 #ifndef __NetBSD__
683 	const char	*name;
684 
685 	if (a->type == PF_ADDR_RTLABEL && a->v.rtlabel) {
686 		if ((name = rtlabel_id2name(a->v.rtlabel)) == NULL)
687 			strlcpy(a->v.rtlabelname, "?",
688 			    sizeof(a->v.rtlabelname));
689 		else
690 			strlcpy(a->v.rtlabelname, name,
691 			    sizeof(a->v.rtlabelname));
692 	}
693 #endif /* !__NetBSD__ */
694 }
695 
696 #ifdef ALTQ
697 u_int32_t
698 pf_qname2qid(char *qname)
699 {
700 	return ((u_int32_t)tagname2tag(&pf_qids, qname));
701 }
702 
703 void
704 pf_qid2qname(u_int32_t qid, char *p)
705 {
706 	tag2tagname(&pf_qids, (u_int16_t)qid, p);
707 }
708 
709 void
710 pf_qid_unref(u_int32_t qid)
711 {
712 	tag_unref(&pf_qids, (u_int16_t)qid);
713 }
714 
715 int
716 pf_begin_altq(u_int32_t *ticket)
717 {
718 	struct pf_altq	*altq;
719 	int		 error = 0;
720 
721 	/* Purge the old altq list */
722 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
723 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
724 		if (altq->qname[0] == 0) {
725 			/* detach and destroy the discipline */
726 			error = altq_remove(altq);
727 		} else
728 			pf_qid_unref(altq->qid);
729 		pool_put(&pf_altq_pl, altq);
730 	}
731 	if (error)
732 		return (error);
733 	*ticket = ++ticket_altqs_inactive;
734 	altqs_inactive_open = 1;
735 	return (0);
736 }
737 
738 int
739 pf_rollback_altq(u_int32_t ticket)
740 {
741 	struct pf_altq	*altq;
742 	int		 error = 0;
743 
744 	if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
745 		return (0);
746 	/* Purge the old altq list */
747 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
748 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
749 		if (altq->qname[0] == 0) {
750 			/* detach and destroy the discipline */
751 			error = altq_remove(altq);
752 		} else
753 			pf_qid_unref(altq->qid);
754 		pool_put(&pf_altq_pl, altq);
755 	}
756 	altqs_inactive_open = 0;
757 	return (error);
758 }
759 
760 int
761 pf_commit_altq(u_int32_t ticket)
762 {
763 	struct pf_altqqueue	*old_altqs;
764 	struct pf_altq		*altq;
765 	int			 s, err, error = 0;
766 
767 	if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
768 		return (EBUSY);
769 
770 	/* swap altqs, keep the old. */
771 	s = splsoftnet();
772 	old_altqs = pf_altqs_active;
773 	pf_altqs_active = pf_altqs_inactive;
774 	pf_altqs_inactive = old_altqs;
775 	ticket_altqs_active = ticket_altqs_inactive;
776 
777 	/* Attach new disciplines */
778 	TAILQ_FOREACH(altq, pf_altqs_active, entries) {
779 		if (altq->qname[0] == 0) {
780 			/* attach the discipline */
781 			error = altq_pfattach(altq);
782 			if (error == 0 && pf_altq_running)
783 				error = pf_enable_altq(altq);
784 			if (error != 0) {
785 				splx(s);
786 				return (error);
787 			}
788 		}
789 	}
790 
791 	/* Purge the old altq list */
792 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
793 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
794 		if (altq->qname[0] == 0) {
795 			/* detach and destroy the discipline */
796 			if (pf_altq_running)
797 				error = pf_disable_altq(altq);
798 			err = altq_pfdetach(altq);
799 			if (err != 0 && error == 0)
800 				error = err;
801 			err = altq_remove(altq);
802 			if (err != 0 && error == 0)
803 				error = err;
804 		} else
805 			pf_qid_unref(altq->qid);
806 		pool_put(&pf_altq_pl, altq);
807 	}
808 	splx(s);
809 
810 	altqs_inactive_open = 0;
811 	return (error);
812 }
813 
814 int
815 pf_enable_altq(struct pf_altq *altq)
816 {
817 	struct ifnet		*ifp;
818 	struct tb_profile	 tb;
819 	int			 s, error = 0;
820 
821 	if ((ifp = ifunit(altq->ifname)) == NULL)
822 		return (EINVAL);
823 
824 	if (ifp->if_snd.altq_type != ALTQT_NONE)
825 		error = altq_enable(&ifp->if_snd);
826 
827 	/* set tokenbucket regulator */
828 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
829 		tb.rate = altq->ifbandwidth;
830 		tb.depth = altq->tbrsize;
831 		s = splnet();
832 		error = tbr_set(&ifp->if_snd, &tb);
833 		splx(s);
834 	}
835 
836 	return (error);
837 }
838 
839 int
840 pf_disable_altq(struct pf_altq *altq)
841 {
842 	struct ifnet		*ifp;
843 	struct tb_profile	 tb;
844 	int			 s, error;
845 
846 	if ((ifp = ifunit(altq->ifname)) == NULL)
847 		return (EINVAL);
848 
849 	/*
850 	 * when the discipline is no longer referenced, it was overridden
851 	 * by a new one.  if so, just return.
852 	 */
853 	if (altq->altq_disc != ifp->if_snd.altq_disc)
854 		return (0);
855 
856 	error = altq_disable(&ifp->if_snd);
857 
858 	if (error == 0) {
859 		/* clear tokenbucket regulator */
860 		tb.rate = 0;
861 		s = splnet();
862 		error = tbr_set(&ifp->if_snd, &tb);
863 		splx(s);
864 	}
865 
866 	return (error);
867 }
868 #endif /* ALTQ */
869 
870 int
871 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
872 {
873 	struct pf_ruleset	*rs;
874 	struct pf_rule		*rule;
875 
876 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
877 		return (EINVAL);
878 	rs = pf_find_or_create_ruleset(anchor);
879 	if (rs == NULL)
880 		return (EINVAL);
881 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
882 		pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
883 		rs->rules[rs_num].inactive.rcount--;
884 	}
885 	*ticket = ++rs->rules[rs_num].inactive.ticket;
886 	rs->rules[rs_num].inactive.open = 1;
887 	return (0);
888 }
889 
890 int
891 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
892 {
893 	struct pf_ruleset	*rs;
894 	struct pf_rule		*rule;
895 
896 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
897 		return (EINVAL);
898 	rs = pf_find_ruleset(anchor);
899 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
900 	    rs->rules[rs_num].inactive.ticket != ticket)
901 		return (0);
902 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
903 		pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
904 		rs->rules[rs_num].inactive.rcount--;
905 	}
906 	rs->rules[rs_num].inactive.open = 0;
907 	return (0);
908 }
909 
910 #define PF_MD5_UPD(st, elm)						\
911 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
912 
913 #define PF_MD5_UPD_STR(st, elm)						\
914 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
915 
916 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
917 		(stor) = htonl((st)->elm);				\
918 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
919 } while (0)
920 
921 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
922 		(stor) = htons((st)->elm);				\
923 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
924 } while (0)
925 
926 void
927 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
928 {
929 	PF_MD5_UPD(pfr, addr.type);
930 	switch (pfr->addr.type) {
931 		case PF_ADDR_DYNIFTL:
932 			PF_MD5_UPD(pfr, addr.v.ifname);
933 			PF_MD5_UPD(pfr, addr.iflags);
934 			break;
935 		case PF_ADDR_TABLE:
936 			PF_MD5_UPD(pfr, addr.v.tblname);
937 			break;
938 		case PF_ADDR_ADDRMASK:
939 			/* XXX ignore af? */
940 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
941 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
942 			break;
943 		case PF_ADDR_RTLABEL:
944 			PF_MD5_UPD(pfr, addr.v.rtlabelname);
945 			break;
946 	}
947 
948 	PF_MD5_UPD(pfr, port[0]);
949 	PF_MD5_UPD(pfr, port[1]);
950 	PF_MD5_UPD(pfr, neg);
951 	PF_MD5_UPD(pfr, port_op);
952 }
953 
954 void
955 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
956 {
957 	u_int16_t x;
958 	u_int32_t y;
959 
960 	pf_hash_rule_addr(ctx, &rule->src);
961 	pf_hash_rule_addr(ctx, &rule->dst);
962 	PF_MD5_UPD_STR(rule, label);
963 	PF_MD5_UPD_STR(rule, ifname);
964 	PF_MD5_UPD_STR(rule, match_tagname);
965 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
966 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
967 	PF_MD5_UPD_HTONL(rule, prob, y);
968 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
969 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
970 	PF_MD5_UPD(rule, uid.op);
971 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
972 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
973 	PF_MD5_UPD(rule, gid.op);
974 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
975 	PF_MD5_UPD(rule, action);
976 	PF_MD5_UPD(rule, direction);
977 	PF_MD5_UPD(rule, af);
978 	PF_MD5_UPD(rule, quick);
979 	PF_MD5_UPD(rule, ifnot);
980 	PF_MD5_UPD(rule, match_tag_not);
981 	PF_MD5_UPD(rule, natpass);
982 	PF_MD5_UPD(rule, keep_state);
983 	PF_MD5_UPD(rule, proto);
984 	PF_MD5_UPD(rule, type);
985 	PF_MD5_UPD(rule, code);
986 	PF_MD5_UPD(rule, flags);
987 	PF_MD5_UPD(rule, flagset);
988 	PF_MD5_UPD(rule, allow_opts);
989 	PF_MD5_UPD(rule, rt);
990 	PF_MD5_UPD(rule, tos);
991 }
992 
993 int
994 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
995 {
996 	struct pf_ruleset	*rs;
997 	struct pf_rule		*rule, **old_array;
998 	struct pf_rulequeue	*old_rules;
999 	int			 s, error;
1000 	u_int32_t		 old_rcount;
1001 
1002 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1003 		return (EINVAL);
1004 	rs = pf_find_ruleset(anchor);
1005 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1006 	    ticket != rs->rules[rs_num].inactive.ticket)
1007 		return (EBUSY);
1008 
1009 	/* Calculate checksum for the main ruleset */
1010 	if (rs == &pf_main_ruleset) {
1011 		error = pf_setup_pfsync_matching(rs);
1012 		if (error != 0)
1013 			return (error);
1014 	}
1015 
1016 	/* Swap rules, keep the old. */
1017 	s = splsoftnet();
1018 	old_rules = rs->rules[rs_num].active.ptr;
1019 	old_rcount = rs->rules[rs_num].active.rcount;
1020 	old_array = rs->rules[rs_num].active.ptr_array;
1021 
1022 	rs->rules[rs_num].active.ptr =
1023 	    rs->rules[rs_num].inactive.ptr;
1024 	rs->rules[rs_num].active.ptr_array =
1025 	    rs->rules[rs_num].inactive.ptr_array;
1026 	rs->rules[rs_num].active.rcount =
1027 	    rs->rules[rs_num].inactive.rcount;
1028 	rs->rules[rs_num].inactive.ptr = old_rules;
1029 	rs->rules[rs_num].inactive.ptr_array = old_array;
1030 	rs->rules[rs_num].inactive.rcount = old_rcount;
1031 
1032 	rs->rules[rs_num].active.ticket =
1033 	    rs->rules[rs_num].inactive.ticket;
1034 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1035 
1036 
1037 	/* Purge the old rule list. */
1038 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1039 		pf_rm_rule(old_rules, rule);
1040 	if (rs->rules[rs_num].inactive.ptr_array)
1041 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1042 	rs->rules[rs_num].inactive.ptr_array = NULL;
1043 	rs->rules[rs_num].inactive.rcount = 0;
1044 	rs->rules[rs_num].inactive.open = 0;
1045 	pf_remove_if_empty_ruleset(rs);
1046 	splx(s);
1047 	return (0);
1048 }
1049 
1050 void
1051 pf_state_export(struct pfsync_state *sp, struct pf_state_key *sk,
1052    struct pf_state *s)
1053 {
1054 	int secs = time_second;
1055 	bzero(sp, sizeof(struct pfsync_state));
1056 
1057 	/* copy from state key */
1058 	sp->lan.addr = sk->lan.addr;
1059 	sp->lan.port = sk->lan.port;
1060 	sp->gwy.addr = sk->gwy.addr;
1061 	sp->gwy.port = sk->gwy.port;
1062 	sp->ext.addr = sk->ext.addr;
1063 	sp->ext.port = sk->ext.port;
1064 	sp->proto = sk->proto;
1065 	sp->af = sk->af;
1066 	sp->direction = sk->direction;
1067 
1068 	/* copy from state */
1069 	memcpy(&sp->id, &s->id, sizeof(sp->id));
1070 	sp->creatorid = s->creatorid;
1071 	strlcpy(sp->ifname, s->kif->pfik_name, sizeof(sp->ifname));
1072 	pf_state_peer_to_pfsync(&s->src, &sp->src);
1073 	pf_state_peer_to_pfsync(&s->dst, &sp->dst);
1074 
1075 	sp->rule = s->rule.ptr->nr;
1076 	sp->nat_rule = (s->nat_rule.ptr == NULL) ?  -1 : s->nat_rule.ptr->nr;
1077 	sp->anchor = (s->anchor.ptr == NULL) ?  -1 : s->anchor.ptr->nr;
1078 
1079 	pf_state_counter_to_pfsync(s->bytes[0], sp->bytes[0]);
1080 	pf_state_counter_to_pfsync(s->bytes[1], sp->bytes[1]);
1081 	pf_state_counter_to_pfsync(s->packets[0], sp->packets[0]);
1082 	pf_state_counter_to_pfsync(s->packets[1], sp->packets[1]);
1083 	sp->creation = secs - s->creation;
1084 	sp->expire = pf_state_expires(s);
1085 	sp->log = s->log;
1086 	sp->allow_opts = s->allow_opts;
1087 	sp->timeout = s->timeout;
1088 
1089 	if (s->src_node)
1090 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
1091 	if (s->nat_src_node)
1092 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
1093 
1094 	if (sp->expire > secs)
1095 		sp->expire -= secs;
1096 	else
1097 		sp->expire = 0;
1098 
1099 }
1100 
1101 void
1102 pf_state_import(struct pfsync_state *sp, struct pf_state_key *sk,
1103    struct pf_state *s)
1104 {
1105 	/* copy to state key */
1106 	sk->lan.addr = sp->lan.addr;
1107 	sk->lan.port = sp->lan.port;
1108 	sk->gwy.addr = sp->gwy.addr;
1109 	sk->gwy.port = sp->gwy.port;
1110 	sk->ext.addr = sp->ext.addr;
1111 	sk->ext.port = sp->ext.port;
1112 	sk->proto = sp->proto;
1113 	sk->af = sp->af;
1114 	sk->direction = sp->direction;
1115 
1116 	/* copy to state */
1117 	memcpy(&s->id, &sp->id, sizeof(sp->id));
1118 	s->creatorid = sp->creatorid;
1119 	pf_state_peer_from_pfsync(&sp->src, &s->src);
1120 	pf_state_peer_from_pfsync(&sp->dst, &s->dst);
1121 
1122 	s->rule.ptr = &pf_default_rule;
1123 	s->rule.ptr->states++;
1124 	s->nat_rule.ptr = NULL;
1125 	s->anchor.ptr = NULL;
1126 	s->rt_kif = NULL;
1127 	s->creation = time_second;
1128 	s->expire = time_second;
1129 	s->timeout = sp->timeout;
1130 	if (sp->expire > 0)
1131 		s->expire -= pf_default_rule.timeout[sp->timeout] - sp->expire;
1132 	s->pfsync_time = 0;
1133 	s->packets[0] = s->packets[1] = 0;
1134 	s->bytes[0] = s->bytes[1] = 0;
1135 }
1136 
1137 int
1138 pf_state_add(struct pfsync_state* sp)
1139 {
1140 	struct pf_state		*s;
1141 	struct pf_state_key	*sk;
1142 	struct pfi_kif		*kif;
1143 
1144 	if (sp->timeout >= PFTM_MAX &&
1145 			sp->timeout != PFTM_UNTIL_PACKET) {
1146 		return EINVAL;
1147 	}
1148 	s = pool_get(&pf_state_pl, PR_NOWAIT);
1149 	if (s == NULL) {
1150 		return ENOMEM;
1151 	}
1152 	bzero(s, sizeof(struct pf_state));
1153 	if ((sk = pf_alloc_state_key(s)) == NULL) {
1154 		pool_put(&pf_state_pl, s);
1155 		return ENOMEM;
1156 	}
1157 	pf_state_import(sp, sk, s);
1158 	kif = pfi_kif_get(sp->ifname);
1159 	if (kif == NULL) {
1160 		pool_put(&pf_state_pl, s);
1161 		pool_put(&pf_state_key_pl, sk);
1162 		return ENOENT;
1163 	}
1164 	if (pf_insert_state(kif, s)) {
1165 		pfi_kif_unref(kif, PFI_KIF_REF_NONE);
1166 		pool_put(&pf_state_pl, s);
1167 		return ENOMEM;
1168 	}
1169 
1170 	return 0;
1171 }
1172 
1173 
1174 int
1175 pf_setup_pfsync_matching(struct pf_ruleset *rs)
1176 {
1177 	MD5_CTX			 ctx;
1178 	struct pf_rule		*rule;
1179 	int			 rs_cnt;
1180 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
1181 
1182 	MD5Init(&ctx);
1183 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1184 		/* XXX PF_RULESET_SCRUB as well? */
1185 		if (rs_cnt == PF_RULESET_SCRUB)
1186 			continue;
1187 
1188 		if (rs->rules[rs_cnt].inactive.ptr_array)
1189 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1190 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
1191 
1192 		if (rs->rules[rs_cnt].inactive.rcount) {
1193 			rs->rules[rs_cnt].inactive.ptr_array =
1194 			    malloc(sizeof(void *) *
1195 			    rs->rules[rs_cnt].inactive.rcount,
1196 			    M_TEMP, M_NOWAIT);
1197 
1198 			if (!rs->rules[rs_cnt].inactive.ptr_array)
1199 				return (ENOMEM);
1200 		}
1201 
1202 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1203 		    entries) {
1204 			pf_hash_rule(&ctx, rule);
1205 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1206 		}
1207 	}
1208 
1209 	MD5Final(digest, &ctx);
1210 	memcpy(pf_status.pf_chksum, digest, sizeof(pf_status.pf_chksum));
1211 	return (0);
1212 }
1213 
1214 int
1215 pfioctl(dev_t dev, u_long cmd, void *addr, int flags, struct lwp *l)
1216 {
1217 	struct pf_pooladdr	*pa = NULL;
1218 	struct pf_pool		*pool = NULL;
1219 	int			 s;
1220 	int			 error = 0;
1221 
1222 	/* XXX keep in sync with switch() below */
1223 	if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FIREWALL,
1224 	    KAUTH_REQ_NETWORK_FIREWALL_FW, NULL, NULL, NULL))
1225 		switch (cmd) {
1226 		case DIOCGETRULES:
1227 		case DIOCGETRULE:
1228 		case DIOCGETADDRS:
1229 		case DIOCGETADDR:
1230 		case DIOCGETSTATE:
1231 		case DIOCSETSTATUSIF:
1232 		case DIOCGETSTATUS:
1233 		case DIOCCLRSTATUS:
1234 		case DIOCNATLOOK:
1235 		case DIOCSETDEBUG:
1236 		case DIOCGETSTATES:
1237 		case DIOCGETTIMEOUT:
1238 		case DIOCCLRRULECTRS:
1239 		case DIOCGETLIMIT:
1240 		case DIOCGETALTQS:
1241 		case DIOCGETALTQ:
1242 		case DIOCGETQSTATS:
1243 		case DIOCGETRULESETS:
1244 		case DIOCGETRULESET:
1245 		case DIOCRGETTABLES:
1246 		case DIOCRGETTSTATS:
1247 		case DIOCRCLRTSTATS:
1248 		case DIOCRCLRADDRS:
1249 		case DIOCRADDADDRS:
1250 		case DIOCRDELADDRS:
1251 		case DIOCRSETADDRS:
1252 		case DIOCRGETADDRS:
1253 		case DIOCRGETASTATS:
1254 		case DIOCRCLRASTATS:
1255 		case DIOCRTSTADDRS:
1256 		case DIOCOSFPGET:
1257 		case DIOCGETSRCNODES:
1258 		case DIOCCLRSRCNODES:
1259 		case DIOCIGETIFACES:
1260 		case DIOCSETIFFLAG:
1261 		case DIOCCLRIFFLAG:
1262 		case DIOCSETLCK:
1263 		case DIOCADDSTATES:
1264 			break;
1265 		case DIOCRCLRTABLES:
1266 		case DIOCRADDTABLES:
1267 		case DIOCRDELTABLES:
1268 		case DIOCRSETTFLAGS:
1269 			if (((struct pfioc_table *)addr)->pfrio_flags &
1270 			    PFR_FLAG_DUMMY)
1271 				break; /* dummy operation ok */
1272 			return (EPERM);
1273 		default:
1274 			return (EPERM);
1275 		}
1276 
1277 	if (!(flags & FWRITE))
1278 		switch (cmd) {
1279 		case DIOCGETRULES:
1280 		case DIOCGETADDRS:
1281 		case DIOCGETADDR:
1282 		case DIOCGETSTATE:
1283 		case DIOCGETSTATUS:
1284 		case DIOCGETSTATES:
1285 		case DIOCGETTIMEOUT:
1286 		case DIOCGETLIMIT:
1287 		case DIOCGETALTQS:
1288 		case DIOCGETALTQ:
1289 		case DIOCGETQSTATS:
1290 		case DIOCGETRULESETS:
1291 		case DIOCGETRULESET:
1292 		case DIOCNATLOOK:
1293 		case DIOCRGETTABLES:
1294 		case DIOCRGETTSTATS:
1295 		case DIOCRGETADDRS:
1296 		case DIOCRGETASTATS:
1297 		case DIOCRTSTADDRS:
1298 		case DIOCOSFPGET:
1299 		case DIOCGETSRCNODES:
1300 		case DIOCIGETIFACES:
1301 		case DIOCSETLCK:
1302 			break;
1303 		case DIOCRCLRTABLES:
1304 		case DIOCRADDTABLES:
1305 		case DIOCRDELTABLES:
1306 		case DIOCRCLRTSTATS:
1307 		case DIOCRCLRADDRS:
1308 		case DIOCRADDADDRS:
1309 		case DIOCRDELADDRS:
1310 		case DIOCRSETADDRS:
1311 		case DIOCRSETTFLAGS:
1312 		case DIOCADDSTATES:
1313 			if (((struct pfioc_table *)addr)->pfrio_flags &
1314 			    PFR_FLAG_DUMMY) {
1315 				flags |= FWRITE; /* need write lock for dummy */
1316 				break; /* dummy operation ok */
1317 			}
1318 			return (EACCES);
1319 		case DIOCGETRULE:
1320 			if (((struct pfioc_rule *)addr)->action == PF_GET_CLR_CNTR)
1321 				return (EACCES);
1322 			break;
1323 		default:
1324 			return (EACCES);
1325 		}
1326 
1327 	if (flags & FWRITE)
1328 		rw_enter_write(&pf_consistency_lock);
1329 	else
1330 		rw_enter_read(&pf_consistency_lock);
1331 
1332 	s = splsoftnet();
1333 	switch (cmd) {
1334 
1335 	case DIOCSTART:
1336 		if (pf_status.running)
1337 			error = EEXIST;
1338 		else {
1339 #ifdef __NetBSD__
1340 			error = pf_pfil_attach();
1341 			if (error)
1342 				break;
1343 #endif /* __NetBSD__ */
1344 			pf_status.running = 1;
1345 			pf_status.since = time_second;
1346 			if (pf_status.stateid == 0) {
1347 				pf_status.stateid = time_second;
1348 				pf_status.stateid = pf_status.stateid << 32;
1349 			}
1350 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1351 		}
1352 		break;
1353 
1354 	case DIOCSTOP:
1355 		if (!pf_status.running)
1356 			error = ENOENT;
1357 		else {
1358 #ifdef __NetBSD__
1359 			error = pf_pfil_detach();
1360 			if (error)
1361 				break;
1362 #endif /* __NetBSD__ */
1363 			pf_status.running = 0;
1364 			pf_status.since = time_second;
1365 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1366 		}
1367 		break;
1368 
1369 	case DIOCADDRULE: {
1370 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1371 		struct pf_ruleset	*ruleset;
1372 		struct pf_rule		*rule, *tail;
1373 		int			 rs_num;
1374 
1375 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1376 		ruleset = pf_find_ruleset(pr->anchor);
1377 		if (ruleset == NULL) {
1378 			error = EINVAL;
1379 			break;
1380 		}
1381 		rs_num = pf_get_ruleset_number(pr->rule.action);
1382 		if (rs_num >= PF_RULESET_MAX) {
1383 			error = EINVAL;
1384 			break;
1385 		}
1386 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1387 			error = EINVAL;
1388 			break;
1389 		}
1390 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1391 			error = EBUSY;
1392 			break;
1393 		}
1394 		if (pr->pool_ticket != ticket_pabuf) {
1395 			error = EBUSY;
1396 			break;
1397 		}
1398 		rule = pool_get(&pf_rule_pl, PR_NOWAIT);
1399 		if (rule == NULL) {
1400 			error = ENOMEM;
1401 			break;
1402 		}
1403 		bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1404 #ifdef __NetBSD__
1405 		rule->cuid = kauth_cred_getuid(l->l_cred);
1406 		rule->cpid = l->l_proc->p_pid;
1407 #else
1408 		rule->cuid = p->p_cred->p_ruid;
1409 		rule->cpid = p->p_pid;
1410 #endif /* !__NetBSD__ */
1411 		rule->anchor = NULL;
1412 		rule->kif = NULL;
1413 		TAILQ_INIT(&rule->rpool.list);
1414 		/* initialize refcounting */
1415 		rule->states = 0;
1416 		rule->src_nodes = 0;
1417 		rule->entries.tqe_prev = NULL;
1418 #ifndef INET
1419 		if (rule->af == AF_INET) {
1420 			pool_put(&pf_rule_pl, rule);
1421 			error = EAFNOSUPPORT;
1422 			break;
1423 		}
1424 #endif /* INET */
1425 #ifndef INET6
1426 		if (rule->af == AF_INET6) {
1427 			pool_put(&pf_rule_pl, rule);
1428 			error = EAFNOSUPPORT;
1429 			break;
1430 		}
1431 #endif /* INET6 */
1432 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1433 		    pf_rulequeue);
1434 		if (tail)
1435 			rule->nr = tail->nr + 1;
1436 		else
1437 			rule->nr = 0;
1438 		if (rule->ifname[0]) {
1439 			rule->kif = pfi_kif_get(rule->ifname);
1440 			if (rule->kif == NULL) {
1441 				pool_put(&pf_rule_pl, rule);
1442 				error = EINVAL;
1443 				break;
1444 			}
1445 			pfi_kif_ref(rule->kif, PFI_KIF_REF_RULE);
1446 		}
1447 
1448 #ifndef __NetBSD__
1449 		if (rule->rtableid > 0 && !rtable_exists(rule->rtableid))
1450 			error = EBUSY;
1451 #endif /* !__NetBSD__ */
1452 
1453 #ifdef ALTQ
1454 		/* set queue IDs */
1455 		if (rule->qname[0] != 0) {
1456 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1457 				error = EBUSY;
1458 			else if (rule->pqname[0] != 0) {
1459 				if ((rule->pqid =
1460 				    pf_qname2qid(rule->pqname)) == 0)
1461 					error = EBUSY;
1462 			} else
1463 				rule->pqid = rule->qid;
1464 		}
1465 #endif
1466 		if (rule->tagname[0])
1467 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1468 				error = EBUSY;
1469 		if (rule->match_tagname[0])
1470 			if ((rule->match_tag =
1471 			    pf_tagname2tag(rule->match_tagname)) == 0)
1472 				error = EBUSY;
1473 		if (rule->rt && !rule->direction)
1474 			error = EINVAL;
1475 #if NPFLOG > 0
1476 		if (!rule->log)
1477 			rule->logif = 0;
1478 		if (rule->logif >= PFLOGIFS_MAX)
1479 			error = EINVAL;
1480 #endif
1481 		if (pf_rtlabel_add(&rule->src.addr) ||
1482 		    pf_rtlabel_add(&rule->dst.addr))
1483 			error = EBUSY;
1484 		if (pfi_dynaddr_setup(&rule->src.addr, rule->af))
1485 			error = EINVAL;
1486 		if (pfi_dynaddr_setup(&rule->dst.addr, rule->af))
1487 			error = EINVAL;
1488 		if (pf_tbladdr_setup(ruleset, &rule->src.addr))
1489 			error = EINVAL;
1490 		if (pf_tbladdr_setup(ruleset, &rule->dst.addr))
1491 			error = EINVAL;
1492 		if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1493 			error = EINVAL;
1494 		TAILQ_FOREACH(pa, &pf_pabuf, entries)
1495 			if (pf_tbladdr_setup(ruleset, &pa->addr))
1496 				error = EINVAL;
1497 
1498 		rule->overload_tbl = NULL;
1499 		if (rule->overload_tblname[0]) {
1500 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1501 			    rule->overload_tblname)) == NULL)
1502 				error = EINVAL;
1503 			else
1504 				rule->overload_tbl->pfrkt_flags |=
1505 				    PFR_TFLAG_ACTIVE;
1506 		}
1507 
1508 		pf_mv_pool(&pf_pabuf, &rule->rpool.list);
1509 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1510 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1511 		    (rule->rt > PF_FASTROUTE)) &&
1512 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1513 			error = EINVAL;
1514 
1515 		if (error) {
1516 			pf_rm_rule(NULL, rule);
1517 			break;
1518 		}
1519 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1520 		rule->evaluations = rule->packets[0] = rule->packets[1] =
1521 		    rule->bytes[0] = rule->bytes[1] = 0;
1522 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1523 		    rule, entries);
1524 		ruleset->rules[rs_num].inactive.rcount++;
1525 		break;
1526 	}
1527 
1528 	case DIOCGETRULES: {
1529 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1530 		struct pf_ruleset	*ruleset;
1531 		struct pf_rule		*tail;
1532 		int			 rs_num;
1533 
1534 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1535 		ruleset = pf_find_ruleset(pr->anchor);
1536 		if (ruleset == NULL) {
1537 			error = EINVAL;
1538 			break;
1539 		}
1540 		rs_num = pf_get_ruleset_number(pr->rule.action);
1541 		if (rs_num >= PF_RULESET_MAX) {
1542 			error = EINVAL;
1543 			break;
1544 		}
1545 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1546 		    pf_rulequeue);
1547 		if (tail)
1548 			pr->nr = tail->nr + 1;
1549 		else
1550 			pr->nr = 0;
1551 		pr->ticket = ruleset->rules[rs_num].active.ticket;
1552 		break;
1553 	}
1554 
1555 	case DIOCGETRULE: {
1556 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1557 		struct pf_ruleset	*ruleset;
1558 		struct pf_rule		*rule;
1559 		int			 rs_num, i;
1560 
1561 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1562 		ruleset = pf_find_ruleset(pr->anchor);
1563 		if (ruleset == NULL) {
1564 			error = EINVAL;
1565 			break;
1566 		}
1567 		rs_num = pf_get_ruleset_number(pr->rule.action);
1568 		if (rs_num >= PF_RULESET_MAX) {
1569 			error = EINVAL;
1570 			break;
1571 		}
1572 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1573 			error = EBUSY;
1574 			break;
1575 		}
1576 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1577 		while ((rule != NULL) && (rule->nr != pr->nr))
1578 			rule = TAILQ_NEXT(rule, entries);
1579 		if (rule == NULL) {
1580 			error = EBUSY;
1581 			break;
1582 		}
1583 		bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1584 		if (pf_anchor_copyout(ruleset, rule, pr)) {
1585 			error = EBUSY;
1586 			break;
1587 		}
1588 		pfi_dynaddr_copyout(&pr->rule.src.addr);
1589 		pfi_dynaddr_copyout(&pr->rule.dst.addr);
1590 		pf_tbladdr_copyout(&pr->rule.src.addr);
1591 		pf_tbladdr_copyout(&pr->rule.dst.addr);
1592 		pf_rtlabel_copyout(&pr->rule.src.addr);
1593 		pf_rtlabel_copyout(&pr->rule.dst.addr);
1594 		for (i = 0; i < PF_SKIP_COUNT; ++i)
1595 			if (rule->skip[i].ptr == NULL)
1596 				pr->rule.skip[i].nr = -1;
1597 			else
1598 				pr->rule.skip[i].nr =
1599 				    rule->skip[i].ptr->nr;
1600 
1601 		if (pr->action == PF_GET_CLR_CNTR) {
1602 			rule->evaluations = 0;
1603 			rule->packets[0] = rule->packets[1] = 0;
1604 			rule->bytes[0] = rule->bytes[1] = 0;
1605 		}
1606 		break;
1607 	}
1608 
1609 	case DIOCCHANGERULE: {
1610 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1611 		struct pf_ruleset	*ruleset;
1612 		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1613 		u_int32_t		 nr = 0;
1614 		int			 rs_num;
1615 
1616 		if (!(pcr->action == PF_CHANGE_REMOVE ||
1617 		    pcr->action == PF_CHANGE_GET_TICKET) &&
1618 		    pcr->pool_ticket != ticket_pabuf) {
1619 			error = EBUSY;
1620 			break;
1621 		}
1622 
1623 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1624 		    pcr->action > PF_CHANGE_GET_TICKET) {
1625 			error = EINVAL;
1626 			break;
1627 		}
1628 		ruleset = pf_find_ruleset(pcr->anchor);
1629 		if (ruleset == NULL) {
1630 			error = EINVAL;
1631 			break;
1632 		}
1633 		rs_num = pf_get_ruleset_number(pcr->rule.action);
1634 		if (rs_num >= PF_RULESET_MAX) {
1635 			error = EINVAL;
1636 			break;
1637 		}
1638 
1639 		if (pcr->action == PF_CHANGE_GET_TICKET) {
1640 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1641 			break;
1642 		} else {
1643 			if (pcr->ticket !=
1644 			    ruleset->rules[rs_num].active.ticket) {
1645 				error = EINVAL;
1646 				break;
1647 			}
1648 			if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1649 				error = EINVAL;
1650 				break;
1651 			}
1652 		}
1653 
1654 		if (pcr->action != PF_CHANGE_REMOVE) {
1655 			newrule = pool_get(&pf_rule_pl, PR_NOWAIT);
1656 			if (newrule == NULL) {
1657 				error = ENOMEM;
1658 				break;
1659 			}
1660 			bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1661 #ifdef __NetBSD__
1662 			newrule->cuid = kauth_cred_getuid(l->l_cred);
1663 			newrule->cpid = l->l_proc->p_pid;
1664 #else
1665 			newrule->cuid = p->p_cred->p_ruid;
1666 			newrule->cpid = p->p_pid;
1667 #endif /* !__NetBSD__ */
1668 			TAILQ_INIT(&newrule->rpool.list);
1669 			/* initialize refcounting */
1670 			newrule->states = 0;
1671 			newrule->entries.tqe_prev = NULL;
1672 #ifndef INET
1673 			if (newrule->af == AF_INET) {
1674 				pool_put(&pf_rule_pl, newrule);
1675 				error = EAFNOSUPPORT;
1676 				break;
1677 			}
1678 #endif /* INET */
1679 #ifndef INET6
1680 			if (newrule->af == AF_INET6) {
1681 				pool_put(&pf_rule_pl, newrule);
1682 				error = EAFNOSUPPORT;
1683 				break;
1684 			}
1685 #endif /* INET6 */
1686 			if (newrule->ifname[0]) {
1687 				newrule->kif = pfi_kif_get(newrule->ifname);
1688 				if (newrule->kif == NULL) {
1689 					pool_put(&pf_rule_pl, newrule);
1690 					error = EINVAL;
1691 					break;
1692 				}
1693 				pfi_kif_ref(newrule->kif, PFI_KIF_REF_RULE);
1694 			} else
1695 				newrule->kif = NULL;
1696 
1697 #ifndef __NetBSD__
1698 			if (newrule->rtableid > 0 &&
1699 			    !rtable_exists(newrule->rtableid))
1700 				error = EBUSY;
1701 #endif /* !__NetBSD__ */
1702 
1703 #ifdef ALTQ
1704 			/* set queue IDs */
1705 			if (newrule->qname[0] != 0) {
1706 				if ((newrule->qid =
1707 				    pf_qname2qid(newrule->qname)) == 0)
1708 					error = EBUSY;
1709 				else if (newrule->pqname[0] != 0) {
1710 					if ((newrule->pqid =
1711 					    pf_qname2qid(newrule->pqname)) == 0)
1712 						error = EBUSY;
1713 				} else
1714 					newrule->pqid = newrule->qid;
1715 			}
1716 #endif /* ALTQ */
1717 			if (newrule->tagname[0])
1718 				if ((newrule->tag =
1719 				    pf_tagname2tag(newrule->tagname)) == 0)
1720 					error = EBUSY;
1721 			if (newrule->match_tagname[0])
1722 				if ((newrule->match_tag = pf_tagname2tag(
1723 				    newrule->match_tagname)) == 0)
1724 					error = EBUSY;
1725 			if (newrule->rt && !newrule->direction)
1726 				error = EINVAL;
1727 #if NPFLOG > 0
1728 			if (!newrule->log)
1729 				newrule->logif = 0;
1730 			if (newrule->logif >= PFLOGIFS_MAX)
1731 				error = EINVAL;
1732 #endif
1733 			if (pf_rtlabel_add(&newrule->src.addr) ||
1734 			    pf_rtlabel_add(&newrule->dst.addr))
1735 				error = EBUSY;
1736 			if (pfi_dynaddr_setup(&newrule->src.addr, newrule->af))
1737 				error = EINVAL;
1738 			if (pfi_dynaddr_setup(&newrule->dst.addr, newrule->af))
1739 				error = EINVAL;
1740 			if (pf_tbladdr_setup(ruleset, &newrule->src.addr))
1741 				error = EINVAL;
1742 			if (pf_tbladdr_setup(ruleset, &newrule->dst.addr))
1743 				error = EINVAL;
1744 			if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1745 				error = EINVAL;
1746 			TAILQ_FOREACH(pa, &pf_pabuf, entries)
1747 				if (pf_tbladdr_setup(ruleset, &pa->addr))
1748 					error = EINVAL;
1749 
1750 			newrule->overload_tbl = NULL;
1751 			if (newrule->overload_tblname[0]) {
1752 				if ((newrule->overload_tbl = pfr_attach_table(
1753 				    ruleset, newrule->overload_tblname)) ==
1754 				    NULL)
1755 					error = EINVAL;
1756 				else
1757 					newrule->overload_tbl->pfrkt_flags |=
1758 					    PFR_TFLAG_ACTIVE;
1759 			}
1760 
1761 			pf_mv_pool(&pf_pabuf, &newrule->rpool.list);
1762 			if (((((newrule->action == PF_NAT) ||
1763 			    (newrule->action == PF_RDR) ||
1764 			    (newrule->action == PF_BINAT) ||
1765 			    (newrule->rt > PF_FASTROUTE)) &&
1766 			    !newrule->anchor)) &&
1767 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1768 				error = EINVAL;
1769 
1770 			if (error) {
1771 				pf_rm_rule(NULL, newrule);
1772 				break;
1773 			}
1774 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1775 			newrule->evaluations = 0;
1776 			newrule->packets[0] = newrule->packets[1] = 0;
1777 			newrule->bytes[0] = newrule->bytes[1] = 0;
1778 		}
1779 		pf_empty_pool(&pf_pabuf);
1780 
1781 		if (pcr->action == PF_CHANGE_ADD_HEAD)
1782 			oldrule = TAILQ_FIRST(
1783 			    ruleset->rules[rs_num].active.ptr);
1784 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1785 			oldrule = TAILQ_LAST(
1786 			    ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1787 		else {
1788 			oldrule = TAILQ_FIRST(
1789 			    ruleset->rules[rs_num].active.ptr);
1790 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1791 				oldrule = TAILQ_NEXT(oldrule, entries);
1792 			if (oldrule == NULL) {
1793 				if (newrule != NULL)
1794 					pf_rm_rule(NULL, newrule);
1795 				error = EINVAL;
1796 				break;
1797 			}
1798 		}
1799 
1800 		if (pcr->action == PF_CHANGE_REMOVE) {
1801 			pf_rm_rule(ruleset->rules[rs_num].active.ptr, oldrule);
1802 			ruleset->rules[rs_num].active.rcount--;
1803 		} else {
1804 			if (oldrule == NULL)
1805 				TAILQ_INSERT_TAIL(
1806 				    ruleset->rules[rs_num].active.ptr,
1807 				    newrule, entries);
1808 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1809 			    pcr->action == PF_CHANGE_ADD_BEFORE)
1810 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1811 			else
1812 				TAILQ_INSERT_AFTER(
1813 				    ruleset->rules[rs_num].active.ptr,
1814 				    oldrule, newrule, entries);
1815 			ruleset->rules[rs_num].active.rcount++;
1816 		}
1817 
1818 		nr = 0;
1819 		TAILQ_FOREACH(oldrule,
1820 		    ruleset->rules[rs_num].active.ptr, entries)
1821 			oldrule->nr = nr++;
1822 
1823 		ruleset->rules[rs_num].active.ticket++;
1824 
1825 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1826 		pf_remove_if_empty_ruleset(ruleset);
1827 
1828 		break;
1829 	}
1830 
1831 	case DIOCCLRSTATES: {
1832 		struct pf_state		*ps, *nexts;
1833 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1834 		int			 killed = 0;
1835 
1836 		for (ps = RB_MIN(pf_state_tree_id, &tree_id); ps; ps = nexts) {
1837 			nexts = RB_NEXT(pf_state_tree_id, &tree_id, ps);
1838 
1839 			if (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1840 			    ps->kif->pfik_name)) {
1841 #if NPFSYNC
1842 				/* don't send out individual delete messages */
1843 				ps->sync_flags = PFSTATE_NOSYNC;
1844 #endif
1845 				pf_unlink_state(ps);
1846 				killed++;
1847 			}
1848 		}
1849 		psk->psk_af = killed;
1850 #if NPFSYNC
1851 		pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
1852 #endif
1853 		break;
1854 	}
1855 
1856 	case DIOCKILLSTATES: {
1857 		struct pf_state		*ps, *nexts;
1858 		struct pf_state_key	*sk;
1859 		struct pf_state_host	*src, *dst;
1860 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
1861 		int			 killed = 0;
1862 
1863 		for (ps = RB_MIN(pf_state_tree_id, &tree_id); ps;
1864 		    ps = nexts) {
1865 			nexts = RB_NEXT(pf_state_tree_id, &tree_id, ps);
1866 			sk = ps->state_key;
1867 
1868 			if (sk->direction == PF_OUT) {
1869 				src = &sk->lan;
1870 				dst = &sk->ext;
1871 			} else {
1872 				src = &sk->ext;
1873 				dst = &sk->lan;
1874 			}
1875 			if ((!psk->psk_af || sk->af == psk->psk_af)
1876 			    && (!psk->psk_proto || psk->psk_proto ==
1877 			    sk->proto) &&
1878 			    PF_MATCHA(psk->psk_src.neg,
1879 			    &psk->psk_src.addr.v.a.addr,
1880 			    &psk->psk_src.addr.v.a.mask,
1881 			    &src->addr, sk->af) &&
1882 			    PF_MATCHA(psk->psk_dst.neg,
1883 			    &psk->psk_dst.addr.v.a.addr,
1884 			    &psk->psk_dst.addr.v.a.mask,
1885 			    &dst->addr, sk->af) &&
1886 			    (psk->psk_src.port_op == 0 ||
1887 			    pf_match_port(psk->psk_src.port_op,
1888 			    psk->psk_src.port[0], psk->psk_src.port[1],
1889 			    src->port)) &&
1890 			    (psk->psk_dst.port_op == 0 ||
1891 			    pf_match_port(psk->psk_dst.port_op,
1892 			    psk->psk_dst.port[0], psk->psk_dst.port[1],
1893 			    dst->port)) &&
1894 			    (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1895 			    ps->kif->pfik_name))) {
1896 #if NPFSYNC > 0
1897 				/* send immediate delete of state */
1898 				pfsync_delete_state(ps);
1899 				ps->sync_flags |= PFSTATE_NOSYNC;
1900 #endif
1901 				pf_unlink_state(ps);
1902 				killed++;
1903 			}
1904 		}
1905 		psk->psk_af = killed;
1906 		break;
1907 	}
1908 
1909 	case DIOCADDSTATE: {
1910 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1911 		struct pfsync_state 	*sp = (struct pfsync_state *)ps->state;
1912 
1913 		error = pf_state_add(sp);
1914 		break;
1915 	}
1916 
1917 	case DIOCADDSTATES: {
1918 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
1919 		struct pfsync_state	*p = (struct pfsync_state *) ps->ps_states;
1920 		struct pfsync_state *pk;
1921 		int size = ps->ps_len;
1922 		int i = 0;
1923 		error = 0;
1924 
1925 		pk = malloc(sizeof(*pk), M_TEMP,M_WAITOK);
1926 
1927 		while (error == 0 && i < size)
1928 		{
1929 			if (copyin(p, pk, sizeof(struct pfsync_state)))
1930 			{
1931 				error = EFAULT;
1932 				free(pk, M_TEMP);
1933 			} else {
1934 				error = pf_state_add(pk);
1935 				i += sizeof(*p);
1936 				p++;
1937 			}
1938 		}
1939 
1940 		free(pk, M_TEMP);
1941 		break;
1942 	}
1943 
1944 
1945 	case DIOCGETSTATE: {
1946 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1947 		struct pf_state		*pfs;
1948 		u_int32_t		 nr;
1949 
1950 		nr = 0;
1951 		RB_FOREACH(pfs, pf_state_tree_id, &tree_id) {
1952 			if (nr >= ps->nr)
1953 				break;
1954 			nr++;
1955 		}
1956 		if (pfs == NULL) {
1957 			error = EBUSY;
1958 			break;
1959 		}
1960 
1961 		pf_state_export((struct pfsync_state *)&ps->state,
1962 		    pfs->state_key, pfs);
1963 		break;
1964 	}
1965 
1966 	case DIOCGETSTATES: {
1967 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
1968 		struct pf_state		*state;
1969 		struct pfsync_state	*p, *pstore;
1970 		u_int32_t		 nr = 0;
1971 
1972 		if (ps->ps_len == 0) {
1973 			nr = pf_status.states;
1974 			ps->ps_len = sizeof(struct pfsync_state) * nr;
1975 			break;
1976 		}
1977 
1978 		pstore = malloc(sizeof(*pstore), M_TEMP, M_WAITOK);
1979 
1980 		p = ps->ps_states;
1981 
1982 		state = TAILQ_FIRST(&state_list);
1983 		while (state) {
1984 			if (state->timeout != PFTM_UNLINKED) {
1985 				if ((nr+1) * sizeof(*p) > (unsigned)ps->ps_len)
1986 					break;
1987 
1988 				pf_state_export(pstore,
1989 				    state->state_key, state);
1990 				error = copyout(pstore, p, sizeof(*p));
1991 				if (error) {
1992 					free(pstore, M_TEMP);
1993 					goto fail;
1994 				}
1995 				p++;
1996 				nr++;
1997 			}
1998 			state = TAILQ_NEXT(state, entry_list);
1999 		}
2000 
2001 		ps->ps_len = sizeof(struct pfsync_state) * nr;
2002 
2003 		free(pstore, M_TEMP);
2004 		break;
2005 	}
2006 
2007 	case DIOCGETSTATUS: {
2008 		struct pf_status *ps = (struct pf_status *)addr;
2009 		bcopy(&pf_status, ps, sizeof(struct pf_status));
2010 		pfi_fill_oldstatus(ps);
2011 		break;
2012 	}
2013 
2014 	case DIOCSETSTATUSIF: {
2015 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
2016 
2017 		if (pi->ifname[0] == 0) {
2018 			bzero(pf_status.ifname, IFNAMSIZ);
2019 			break;
2020 		}
2021 		if (ifunit(pi->ifname) == NULL) {
2022 			error = EINVAL;
2023 			break;
2024 		}
2025 		strlcpy(pf_status.ifname, pi->ifname, IFNAMSIZ);
2026 		break;
2027 	}
2028 
2029 	case DIOCCLRSTATUS: {
2030 		bzero(pf_status.counters, sizeof(pf_status.counters));
2031 		bzero(pf_status.fcounters, sizeof(pf_status.fcounters));
2032 		bzero(pf_status.scounters, sizeof(pf_status.scounters));
2033 		pf_status.since = time_second;
2034 		if (*pf_status.ifname)
2035 			pfi_clr_istats(pf_status.ifname);
2036 		break;
2037 	}
2038 
2039 	case DIOCNATLOOK: {
2040 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
2041 		struct pf_state_key	*sk;
2042 		struct pf_state		*state;
2043 		struct pf_state_key_cmp	 key;
2044 		int			 m = 0, direction = pnl->direction;
2045 
2046 		key.af = pnl->af;
2047 		key.proto = pnl->proto;
2048 
2049 		if (!pnl->proto ||
2050 		    PF_AZERO(&pnl->saddr, pnl->af) ||
2051 		    PF_AZERO(&pnl->daddr, pnl->af) ||
2052 		    ((pnl->proto == IPPROTO_TCP ||
2053 		    pnl->proto == IPPROTO_UDP) &&
2054 		    (!pnl->dport || !pnl->sport)))
2055 			error = EINVAL;
2056 		else {
2057 			/*
2058 			 * userland gives us source and dest of connection,
2059 			 * reverse the lookup so we ask for what happens with
2060 			 * the return traffic, enabling us to find it in the
2061 			 * state tree.
2062 			 */
2063 			if (direction == PF_IN) {
2064 				PF_ACPY(&key.ext.addr, &pnl->daddr, pnl->af);
2065 				key.ext.port = pnl->dport;
2066 				PF_ACPY(&key.gwy.addr, &pnl->saddr, pnl->af);
2067 				key.gwy.port = pnl->sport;
2068 				state = pf_find_state_all(&key, PF_EXT_GWY, &m);
2069 			} else {
2070 				PF_ACPY(&key.lan.addr, &pnl->daddr, pnl->af);
2071 				key.lan.port = pnl->dport;
2072 				PF_ACPY(&key.ext.addr, &pnl->saddr, pnl->af);
2073 				key.ext.port = pnl->sport;
2074 				state = pf_find_state_all(&key, PF_LAN_EXT, &m);
2075 			}
2076 			if (m > 1)
2077 				error = E2BIG;	/* more than one state */
2078 			else if (state != NULL) {
2079 				sk = state->state_key;
2080 				if (direction == PF_IN) {
2081 					PF_ACPY(&pnl->rsaddr, &sk->lan.addr,
2082 					    sk->af);
2083 					pnl->rsport = sk->lan.port;
2084 					PF_ACPY(&pnl->rdaddr, &pnl->daddr,
2085 					    pnl->af);
2086 					pnl->rdport = pnl->dport;
2087 				} else {
2088 					PF_ACPY(&pnl->rdaddr, &sk->gwy.addr,
2089 					    sk->af);
2090 					pnl->rdport = sk->gwy.port;
2091 					PF_ACPY(&pnl->rsaddr, &pnl->saddr,
2092 					    pnl->af);
2093 					pnl->rsport = pnl->sport;
2094 				}
2095 			} else
2096 				error = ENOENT;
2097 		}
2098 		break;
2099 	}
2100 
2101 	case DIOCSETTIMEOUT: {
2102 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2103 		int		 old;
2104 
2105 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
2106 		    pt->seconds < 0) {
2107 			error = EINVAL;
2108 			goto fail;
2109 		}
2110 		old = pf_default_rule.timeout[pt->timeout];
2111 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
2112 			pt->seconds = 1;
2113 		pf_default_rule.timeout[pt->timeout] = pt->seconds;
2114 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
2115 			wakeup(pf_purge_thread);
2116 		pt->seconds = old;
2117 		break;
2118 	}
2119 
2120 	case DIOCGETTIMEOUT: {
2121 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2122 
2123 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
2124 			error = EINVAL;
2125 			goto fail;
2126 		}
2127 		pt->seconds = pf_default_rule.timeout[pt->timeout];
2128 		break;
2129 	}
2130 
2131 	case DIOCGETLIMIT: {
2132 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2133 
2134 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
2135 			error = EINVAL;
2136 			goto fail;
2137 		}
2138 		pl->limit = pf_pool_limits[pl->index].limit;
2139 		break;
2140 	}
2141 
2142 	case DIOCSETLIMIT: {
2143 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2144 		int			 old_limit;
2145 
2146 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
2147 		    pf_pool_limits[pl->index].pp == NULL) {
2148 			error = EINVAL;
2149 			goto fail;
2150 		}
2151 #ifdef __NetBSD__
2152 		pool_sethardlimit(pf_pool_limits[pl->index].pp,
2153 		    pl->limit, NULL, 0);
2154 #else
2155 		if (pool_sethardlimit(pf_pool_limits[pl->index].pp,
2156 		    pl->limit, NULL, 0) != 0) {
2157 			error = EBUSY;
2158 			goto fail;
2159 		}
2160 #endif /* !__NetBSD__ */
2161 		old_limit = pf_pool_limits[pl->index].limit;
2162 		pf_pool_limits[pl->index].limit = pl->limit;
2163 		pl->limit = old_limit;
2164 		break;
2165 	}
2166 
2167 	case DIOCSETDEBUG: {
2168 		u_int32_t	*level = (u_int32_t *)addr;
2169 
2170 		pf_status.debug = *level;
2171 		break;
2172 	}
2173 
2174 	case DIOCCLRRULECTRS: {
2175 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
2176 		struct pf_ruleset	*ruleset = &pf_main_ruleset;
2177 		struct pf_rule		*rule;
2178 
2179 		TAILQ_FOREACH(rule,
2180 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
2181 			rule->evaluations = 0;
2182 			rule->packets[0] = rule->packets[1] = 0;
2183 			rule->bytes[0] = rule->bytes[1] = 0;
2184 		}
2185 		break;
2186 	}
2187 
2188 #ifdef ALTQ
2189 	case DIOCSTARTALTQ: {
2190 		struct pf_altq		*altq;
2191 
2192 		/* enable all altq interfaces on active list */
2193 		TAILQ_FOREACH(altq, pf_altqs_active, entries) {
2194 			if (altq->qname[0] == 0) {
2195 				error = pf_enable_altq(altq);
2196 				if (error != 0)
2197 					break;
2198 			}
2199 		}
2200 		if (error == 0)
2201 			pf_altq_running = 1;
2202 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
2203 		break;
2204 	}
2205 
2206 	case DIOCSTOPALTQ: {
2207 		struct pf_altq		*altq;
2208 
2209 		/* disable all altq interfaces on active list */
2210 		TAILQ_FOREACH(altq, pf_altqs_active, entries) {
2211 			if (altq->qname[0] == 0) {
2212 				error = pf_disable_altq(altq);
2213 				if (error != 0)
2214 					break;
2215 			}
2216 		}
2217 		if (error == 0)
2218 			pf_altq_running = 0;
2219 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2220 		break;
2221 	}
2222 
2223 	case DIOCADDALTQ: {
2224 		struct pfioc_altq	*paa = (struct pfioc_altq *)addr;
2225 		struct pf_altq		*altq, *a;
2226 
2227 		if (paa->ticket != ticket_altqs_inactive) {
2228 			error = EBUSY;
2229 			break;
2230 		}
2231 		altq = pool_get(&pf_altq_pl, PR_NOWAIT);
2232 		if (altq == NULL) {
2233 			error = ENOMEM;
2234 			break;
2235 		}
2236 		bcopy(&paa->altq, altq, sizeof(struct pf_altq));
2237 
2238 		/*
2239 		 * if this is for a queue, find the discipline and
2240 		 * copy the necessary fields
2241 		 */
2242 		if (altq->qname[0] != 0) {
2243 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2244 				error = EBUSY;
2245 				pool_put(&pf_altq_pl, altq);
2246 				break;
2247 			}
2248 			TAILQ_FOREACH(a, pf_altqs_inactive, entries) {
2249 				if (strncmp(a->ifname, altq->ifname,
2250 				    IFNAMSIZ) == 0 && a->qname[0] == 0) {
2251 					altq->altq_disc = a->altq_disc;
2252 					break;
2253 				}
2254 			}
2255 		}
2256 
2257 		error = altq_add(altq);
2258 		if (error) {
2259 			pool_put(&pf_altq_pl, altq);
2260 			break;
2261 		}
2262 
2263 		TAILQ_INSERT_TAIL(pf_altqs_inactive, altq, entries);
2264 		bcopy(altq, &paa->altq, sizeof(struct pf_altq));
2265 		break;
2266 	}
2267 
2268 	case DIOCGETALTQS: {
2269 		struct pfioc_altq	*paa = (struct pfioc_altq *)addr;
2270 		struct pf_altq		*altq;
2271 
2272 		paa->nr = 0;
2273 		TAILQ_FOREACH(altq, pf_altqs_active, entries)
2274 			paa->nr++;
2275 		paa->ticket = ticket_altqs_active;
2276 		break;
2277 	}
2278 
2279 	case DIOCGETALTQ: {
2280 		struct pfioc_altq	*paa = (struct pfioc_altq *)addr;
2281 		struct pf_altq		*altq;
2282 		u_int32_t		 nr;
2283 
2284 		if (paa->ticket != ticket_altqs_active) {
2285 			error = EBUSY;
2286 			break;
2287 		}
2288 		nr = 0;
2289 		altq = TAILQ_FIRST(pf_altqs_active);
2290 		while ((altq != NULL) && (nr < paa->nr)) {
2291 			altq = TAILQ_NEXT(altq, entries);
2292 			nr++;
2293 		}
2294 		if (altq == NULL) {
2295 			error = EBUSY;
2296 			break;
2297 		}
2298 		bcopy(altq, &paa->altq, sizeof(struct pf_altq));
2299 		break;
2300 	}
2301 
2302 	case DIOCCHANGEALTQ:
2303 		/* CHANGEALTQ not supported yet! */
2304 		error = ENODEV;
2305 		break;
2306 
2307 	case DIOCGETQSTATS: {
2308 		struct pfioc_qstats	*pq = (struct pfioc_qstats *)addr;
2309 		struct pf_altq		*altq;
2310 		u_int32_t		 nr;
2311 		int			 nbytes;
2312 
2313 		if (pq->ticket != ticket_altqs_active) {
2314 			error = EBUSY;
2315 			break;
2316 		}
2317 		nbytes = pq->nbytes;
2318 		nr = 0;
2319 		altq = TAILQ_FIRST(pf_altqs_active);
2320 		while ((altq != NULL) && (nr < pq->nr)) {
2321 			altq = TAILQ_NEXT(altq, entries);
2322 			nr++;
2323 		}
2324 		if (altq == NULL) {
2325 			error = EBUSY;
2326 			break;
2327 		}
2328 		error = altq_getqstats(altq, pq->buf, &nbytes);
2329 		if (error == 0) {
2330 			pq->scheduler = altq->scheduler;
2331 			pq->nbytes = nbytes;
2332 		}
2333 		break;
2334 	}
2335 #endif /* ALTQ */
2336 
2337 	case DIOCBEGINADDRS: {
2338 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2339 
2340 		pf_empty_pool(&pf_pabuf);
2341 		pp->ticket = ++ticket_pabuf;
2342 		break;
2343 	}
2344 
2345 	case DIOCADDADDR: {
2346 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2347 
2348 		if (pp->ticket != ticket_pabuf) {
2349 			error = EBUSY;
2350 			break;
2351 		}
2352 #ifndef INET
2353 		if (pp->af == AF_INET) {
2354 			error = EAFNOSUPPORT;
2355 			break;
2356 		}
2357 #endif /* INET */
2358 #ifndef INET6
2359 		if (pp->af == AF_INET6) {
2360 			error = EAFNOSUPPORT;
2361 			break;
2362 		}
2363 #endif /* INET6 */
2364 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2365 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2366 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2367 			error = EINVAL;
2368 			break;
2369 		}
2370 		pa = pool_get(&pf_pooladdr_pl, PR_NOWAIT);
2371 		if (pa == NULL) {
2372 			error = ENOMEM;
2373 			break;
2374 		}
2375 		bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2376 		if (pa->ifname[0]) {
2377 			pa->kif = pfi_kif_get(pa->ifname);
2378 			if (pa->kif == NULL) {
2379 				pool_put(&pf_pooladdr_pl, pa);
2380 				error = EINVAL;
2381 				break;
2382 			}
2383 			pfi_kif_ref(pa->kif, PFI_KIF_REF_RULE);
2384 		}
2385 		if (pfi_dynaddr_setup(&pa->addr, pp->af)) {
2386 			pfi_dynaddr_remove(&pa->addr);
2387 			pfi_kif_unref(pa->kif, PFI_KIF_REF_RULE);
2388 			pool_put(&pf_pooladdr_pl, pa);
2389 			error = EINVAL;
2390 			break;
2391 		}
2392 		TAILQ_INSERT_TAIL(&pf_pabuf, pa, entries);
2393 		break;
2394 	}
2395 
2396 	case DIOCGETADDRS: {
2397 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2398 
2399 		pp->nr = 0;
2400 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2401 		    pp->r_num, 0, 1, 0);
2402 		if (pool == NULL) {
2403 			error = EBUSY;
2404 			break;
2405 		}
2406 		TAILQ_FOREACH(pa, &pool->list, entries)
2407 			pp->nr++;
2408 		break;
2409 	}
2410 
2411 	case DIOCGETADDR: {
2412 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2413 		u_int32_t		 nr = 0;
2414 
2415 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2416 		    pp->r_num, 0, 1, 1);
2417 		if (pool == NULL) {
2418 			error = EBUSY;
2419 			break;
2420 		}
2421 		pa = TAILQ_FIRST(&pool->list);
2422 		while ((pa != NULL) && (nr < pp->nr)) {
2423 			pa = TAILQ_NEXT(pa, entries);
2424 			nr++;
2425 		}
2426 		if (pa == NULL) {
2427 			error = EBUSY;
2428 			break;
2429 		}
2430 		bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2431 		pfi_dynaddr_copyout(&pp->addr.addr);
2432 		pf_tbladdr_copyout(&pp->addr.addr);
2433 		pf_rtlabel_copyout(&pp->addr.addr);
2434 		break;
2435 	}
2436 
2437 	case DIOCCHANGEADDR: {
2438 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
2439 		struct pf_pooladdr	*oldpa = NULL, *newpa = NULL;
2440 		struct pf_ruleset	*ruleset;
2441 
2442 		if (pca->action < PF_CHANGE_ADD_HEAD ||
2443 		    pca->action > PF_CHANGE_REMOVE) {
2444 			error = EINVAL;
2445 			break;
2446 		}
2447 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2448 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2449 		    pca->addr.addr.type != PF_ADDR_TABLE) {
2450 			error = EINVAL;
2451 			break;
2452 		}
2453 
2454 		ruleset = pf_find_ruleset(pca->anchor);
2455 		if (ruleset == NULL) {
2456 			error = EBUSY;
2457 			break;
2458 		}
2459 		pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2460 		    pca->r_num, pca->r_last, 1, 1);
2461 		if (pool == NULL) {
2462 			error = EBUSY;
2463 			break;
2464 		}
2465 		if (pca->action != PF_CHANGE_REMOVE) {
2466 			newpa = pool_get(&pf_pooladdr_pl, PR_NOWAIT);
2467 			if (newpa == NULL) {
2468 				error = ENOMEM;
2469 				break;
2470 			}
2471 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2472 #ifndef INET
2473 			if (pca->af == AF_INET) {
2474 				pool_put(&pf_pooladdr_pl, newpa);
2475 				error = EAFNOSUPPORT;
2476 				break;
2477 			}
2478 #endif /* INET */
2479 #ifndef INET6
2480 			if (pca->af == AF_INET6) {
2481 				pool_put(&pf_pooladdr_pl, newpa);
2482 				error = EAFNOSUPPORT;
2483 				break;
2484 			}
2485 #endif /* INET6 */
2486 			if (newpa->ifname[0]) {
2487 				newpa->kif = pfi_kif_get(newpa->ifname);
2488 				if (newpa->kif == NULL) {
2489 					pool_put(&pf_pooladdr_pl, newpa);
2490 					error = EINVAL;
2491 					break;
2492 				}
2493 				pfi_kif_ref(newpa->kif, PFI_KIF_REF_RULE);
2494 			} else
2495 				newpa->kif = NULL;
2496 			if (pfi_dynaddr_setup(&newpa->addr, pca->af) ||
2497 			    pf_tbladdr_setup(ruleset, &newpa->addr)) {
2498 				pfi_dynaddr_remove(&newpa->addr);
2499 				pfi_kif_unref(newpa->kif, PFI_KIF_REF_RULE);
2500 				pool_put(&pf_pooladdr_pl, newpa);
2501 				error = EINVAL;
2502 				break;
2503 			}
2504 		}
2505 
2506 		if (pca->action == PF_CHANGE_ADD_HEAD)
2507 			oldpa = TAILQ_FIRST(&pool->list);
2508 		else if (pca->action == PF_CHANGE_ADD_TAIL)
2509 			oldpa = TAILQ_LAST(&pool->list, pf_palist);
2510 		else {
2511 			int	i = 0;
2512 
2513 			oldpa = TAILQ_FIRST(&pool->list);
2514 			while ((oldpa != NULL) && (i < pca->nr)) {
2515 				oldpa = TAILQ_NEXT(oldpa, entries);
2516 				i++;
2517 			}
2518 			if (oldpa == NULL) {
2519 				error = EINVAL;
2520 				break;
2521 			}
2522 		}
2523 
2524 		if (pca->action == PF_CHANGE_REMOVE) {
2525 			TAILQ_REMOVE(&pool->list, oldpa, entries);
2526 			pfi_dynaddr_remove(&oldpa->addr);
2527 			pf_tbladdr_remove(&oldpa->addr);
2528 			pfi_kif_unref(oldpa->kif, PFI_KIF_REF_RULE);
2529 			pool_put(&pf_pooladdr_pl, oldpa);
2530 		} else {
2531 			if (oldpa == NULL)
2532 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2533 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
2534 			    pca->action == PF_CHANGE_ADD_BEFORE)
2535 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2536 			else
2537 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
2538 				    newpa, entries);
2539 		}
2540 
2541 		pool->cur = TAILQ_FIRST(&pool->list);
2542 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr,
2543 		    pca->af);
2544 		break;
2545 	}
2546 
2547 	case DIOCGETRULESETS: {
2548 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2549 		struct pf_ruleset	*ruleset;
2550 		struct pf_anchor	*anchor;
2551 
2552 		pr->path[sizeof(pr->path) - 1] = 0;
2553 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2554 			error = EINVAL;
2555 			break;
2556 		}
2557 		pr->nr = 0;
2558 		if (ruleset->anchor == NULL) {
2559 			/* XXX kludge for pf_main_ruleset */
2560 			RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2561 				if (anchor->parent == NULL)
2562 					pr->nr++;
2563 		} else {
2564 			RB_FOREACH(anchor, pf_anchor_node,
2565 			    &ruleset->anchor->children)
2566 				pr->nr++;
2567 		}
2568 		break;
2569 	}
2570 
2571 	case DIOCGETRULESET: {
2572 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2573 		struct pf_ruleset	*ruleset;
2574 		struct pf_anchor	*anchor;
2575 		u_int32_t		 nr = 0;
2576 
2577 		pr->path[sizeof(pr->path) - 1] = 0;
2578 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2579 			error = EINVAL;
2580 			break;
2581 		}
2582 		pr->name[0] = 0;
2583 		if (ruleset->anchor == NULL) {
2584 			/* XXX kludge for pf_main_ruleset */
2585 			RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2586 				if (anchor->parent == NULL && nr++ == pr->nr) {
2587 					strlcpy(pr->name, anchor->name,
2588 					    sizeof(pr->name));
2589 					break;
2590 				}
2591 		} else {
2592 			RB_FOREACH(anchor, pf_anchor_node,
2593 			    &ruleset->anchor->children)
2594 				if (nr++ == pr->nr) {
2595 					strlcpy(pr->name, anchor->name,
2596 					    sizeof(pr->name));
2597 					break;
2598 				}
2599 		}
2600 		if (!pr->name[0])
2601 			error = EBUSY;
2602 		break;
2603 	}
2604 
2605 	case DIOCRCLRTABLES: {
2606 		struct pfioc_table *io = (struct pfioc_table *)addr;
2607 
2608 		if (io->pfrio_esize != 0) {
2609 			error = ENODEV;
2610 			break;
2611 		}
2612 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2613 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2614 		break;
2615 	}
2616 
2617 	case DIOCRADDTABLES: {
2618 		struct pfioc_table *io = (struct pfioc_table *)addr;
2619 
2620 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2621 			error = ENODEV;
2622 			break;
2623 		}
2624 		error = pfr_add_tables(io->pfrio_buffer, io->pfrio_size,
2625 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2626 		break;
2627 	}
2628 
2629 	case DIOCRDELTABLES: {
2630 		struct pfioc_table *io = (struct pfioc_table *)addr;
2631 
2632 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2633 			error = ENODEV;
2634 			break;
2635 		}
2636 		error = pfr_del_tables(io->pfrio_buffer, io->pfrio_size,
2637 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2638 		break;
2639 	}
2640 
2641 	case DIOCRGETTABLES: {
2642 		struct pfioc_table *io = (struct pfioc_table *)addr;
2643 
2644 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2645 			error = ENODEV;
2646 			break;
2647 		}
2648 		error = pfr_get_tables(&io->pfrio_table, io->pfrio_buffer,
2649 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2650 		break;
2651 	}
2652 
2653 	case DIOCRGETTSTATS: {
2654 		struct pfioc_table *io = (struct pfioc_table *)addr;
2655 
2656 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2657 			error = ENODEV;
2658 			break;
2659 		}
2660 		error = pfr_get_tstats(&io->pfrio_table, io->pfrio_buffer,
2661 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2662 		break;
2663 	}
2664 
2665 	case DIOCRCLRTSTATS: {
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_clr_tstats(io->pfrio_buffer, io->pfrio_size,
2673 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2674 		break;
2675 	}
2676 
2677 	case DIOCRSETTFLAGS: {
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_set_tflags(io->pfrio_buffer, io->pfrio_size,
2685 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2686 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2687 		break;
2688 	}
2689 
2690 	case DIOCRCLRADDRS: {
2691 		struct pfioc_table *io = (struct pfioc_table *)addr;
2692 
2693 		if (io->pfrio_esize != 0) {
2694 			error = ENODEV;
2695 			break;
2696 		}
2697 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2698 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2699 		break;
2700 	}
2701 
2702 	case DIOCRADDADDRS: {
2703 		struct pfioc_table *io = (struct pfioc_table *)addr;
2704 
2705 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2706 			error = ENODEV;
2707 			break;
2708 		}
2709 		error = pfr_add_addrs(&io->pfrio_table, io->pfrio_buffer,
2710 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2711 		    PFR_FLAG_USERIOCTL);
2712 		break;
2713 	}
2714 
2715 	case DIOCRDELADDRS: {
2716 		struct pfioc_table *io = (struct pfioc_table *)addr;
2717 
2718 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2719 			error = ENODEV;
2720 			break;
2721 		}
2722 		error = pfr_del_addrs(&io->pfrio_table, io->pfrio_buffer,
2723 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2724 		    PFR_FLAG_USERIOCTL);
2725 		break;
2726 	}
2727 
2728 	case DIOCRSETADDRS: {
2729 		struct pfioc_table *io = (struct pfioc_table *)addr;
2730 
2731 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2732 			error = ENODEV;
2733 			break;
2734 		}
2735 		error = pfr_set_addrs(&io->pfrio_table, io->pfrio_buffer,
2736 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2737 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2738 		    PFR_FLAG_USERIOCTL, 0);
2739 		break;
2740 	}
2741 
2742 	case DIOCRGETADDRS: {
2743 		struct pfioc_table *io = (struct pfioc_table *)addr;
2744 
2745 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2746 			error = ENODEV;
2747 			break;
2748 		}
2749 		error = pfr_get_addrs(&io->pfrio_table, io->pfrio_buffer,
2750 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2751 		break;
2752 	}
2753 
2754 	case DIOCRGETASTATS: {
2755 		struct pfioc_table *io = (struct pfioc_table *)addr;
2756 
2757 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2758 			error = ENODEV;
2759 			break;
2760 		}
2761 		error = pfr_get_astats(&io->pfrio_table, io->pfrio_buffer,
2762 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2763 		break;
2764 	}
2765 
2766 	case DIOCRCLRASTATS: {
2767 		struct pfioc_table *io = (struct pfioc_table *)addr;
2768 
2769 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2770 			error = ENODEV;
2771 			break;
2772 		}
2773 		error = pfr_clr_astats(&io->pfrio_table, io->pfrio_buffer,
2774 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2775 		    PFR_FLAG_USERIOCTL);
2776 		break;
2777 	}
2778 
2779 	case DIOCRTSTADDRS: {
2780 		struct pfioc_table *io = (struct pfioc_table *)addr;
2781 
2782 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2783 			error = ENODEV;
2784 			break;
2785 		}
2786 		error = pfr_tst_addrs(&io->pfrio_table, io->pfrio_buffer,
2787 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2788 		    PFR_FLAG_USERIOCTL);
2789 		break;
2790 	}
2791 
2792 	case DIOCRINADEFINE: {
2793 		struct pfioc_table *io = (struct pfioc_table *)addr;
2794 
2795 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2796 			error = ENODEV;
2797 			break;
2798 		}
2799 		error = pfr_ina_define(&io->pfrio_table, io->pfrio_buffer,
2800 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2801 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2802 		break;
2803 	}
2804 
2805 	case DIOCOSFPADD: {
2806 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2807 		error = pf_osfp_add(io);
2808 		break;
2809 	}
2810 
2811 	case DIOCOSFPGET: {
2812 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2813 		error = pf_osfp_get(io);
2814 		break;
2815 	}
2816 
2817 	case DIOCXBEGIN: {
2818 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2819 		struct pfioc_trans_e	*ioe;
2820 		struct pfr_table	*table;
2821 		int			 i;
2822 
2823 		if (io->esize != sizeof(*ioe)) {
2824 			error = ENODEV;
2825 			goto fail;
2826 		}
2827 		ioe = (struct pfioc_trans_e *)malloc(sizeof(*ioe),
2828 		    M_TEMP, M_WAITOK);
2829 		table = (struct pfr_table *)malloc(sizeof(*table),
2830 		    M_TEMP, M_WAITOK);
2831 		for (i = 0; i < io->size; i++) {
2832 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2833 				free(table, M_TEMP);
2834 				free(ioe, M_TEMP);
2835 				error = EFAULT;
2836 				goto fail;
2837 			}
2838 			switch (ioe->rs_num) {
2839 #ifdef ALTQ
2840 			case PF_RULESET_ALTQ:
2841 				if (ioe->anchor[0]) {
2842 					free(table, M_TEMP);
2843 					free(ioe, M_TEMP);
2844 					error = EINVAL;
2845 					goto fail;
2846 				}
2847 				if ((error = pf_begin_altq(&ioe->ticket))) {
2848 					free(table, M_TEMP);
2849 					free(ioe, M_TEMP);
2850 					goto fail;
2851 				}
2852 				break;
2853 #endif /* ALTQ */
2854 			case PF_RULESET_TABLE:
2855 				bzero(table, sizeof(*table));
2856 				strlcpy(table->pfrt_anchor, ioe->anchor,
2857 				    sizeof(table->pfrt_anchor));
2858 				if ((error = pfr_ina_begin(table,
2859 				    &ioe->ticket, NULL, 0))) {
2860 					free(table, M_TEMP);
2861 					free(ioe, M_TEMP);
2862 					goto fail;
2863 				}
2864 				break;
2865 			default:
2866 				if ((error = pf_begin_rules(&ioe->ticket,
2867 				    ioe->rs_num, ioe->anchor))) {
2868 					free(table, M_TEMP);
2869 					free(ioe, M_TEMP);
2870 					goto fail;
2871 				}
2872 				break;
2873 			}
2874 			if (copyout(ioe, io->array+i, sizeof(io->array[i]))) {
2875 				free(table, M_TEMP);
2876 				free(ioe, M_TEMP);
2877 				error = EFAULT;
2878 				goto fail;
2879 			}
2880 		}
2881 		free(table, M_TEMP);
2882 		free(ioe, M_TEMP);
2883 		break;
2884 	}
2885 
2886 	case DIOCXROLLBACK: {
2887 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2888 		struct pfioc_trans_e	*ioe;
2889 		struct pfr_table	*table;
2890 		int			 i;
2891 
2892 		if (io->esize != sizeof(*ioe)) {
2893 			error = ENODEV;
2894 			goto fail;
2895 		}
2896 		ioe = (struct pfioc_trans_e *)malloc(sizeof(*ioe),
2897 		    M_TEMP, M_WAITOK);
2898 		table = (struct pfr_table *)malloc(sizeof(*table),
2899 		    M_TEMP, M_WAITOK);
2900 		for (i = 0; i < io->size; i++) {
2901 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2902 				free(table, M_TEMP);
2903 				free(ioe, M_TEMP);
2904 				error = EFAULT;
2905 				goto fail;
2906 			}
2907 			switch (ioe->rs_num) {
2908 #ifdef ALTQ
2909 			case PF_RULESET_ALTQ:
2910 				if (ioe->anchor[0]) {
2911 					free(table, M_TEMP);
2912 					free(ioe, M_TEMP);
2913 					error = EINVAL;
2914 					goto fail;
2915 				}
2916 				if ((error = pf_rollback_altq(ioe->ticket))) {
2917 					free(table, M_TEMP);
2918 					free(ioe, M_TEMP);
2919 					goto fail; /* really bad */
2920 				}
2921 				break;
2922 #endif /* ALTQ */
2923 			case PF_RULESET_TABLE:
2924 				bzero(table, sizeof(*table));
2925 				strlcpy(table->pfrt_anchor, ioe->anchor,
2926 				    sizeof(table->pfrt_anchor));
2927 				if ((error = pfr_ina_rollback(table,
2928 				    ioe->ticket, NULL, 0))) {
2929 					free(table, M_TEMP);
2930 					free(ioe, M_TEMP);
2931 					goto fail; /* really bad */
2932 				}
2933 				break;
2934 			default:
2935 				if ((error = pf_rollback_rules(ioe->ticket,
2936 				    ioe->rs_num, ioe->anchor))) {
2937 					free(table, M_TEMP);
2938 					free(ioe, M_TEMP);
2939 					goto fail; /* really bad */
2940 				}
2941 				break;
2942 			}
2943 		}
2944 		free(table, M_TEMP);
2945 		free(ioe, M_TEMP);
2946 		break;
2947 	}
2948 
2949 	case DIOCXCOMMIT: {
2950 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2951 		struct pfioc_trans_e	*ioe;
2952 		struct pfr_table	*table;
2953 		struct pf_ruleset	*rs;
2954 		int			 i;
2955 
2956 		if (io->esize != sizeof(*ioe)) {
2957 			error = ENODEV;
2958 			goto fail;
2959 		}
2960 		ioe = (struct pfioc_trans_e *)malloc(sizeof(*ioe),
2961 		    M_TEMP, M_WAITOK);
2962 		table = (struct pfr_table *)malloc(sizeof(*table),
2963 		    M_TEMP, M_WAITOK);
2964 		/* first makes sure everything will succeed */
2965 		for (i = 0; i < io->size; i++) {
2966 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2967 				free(table, M_TEMP);
2968 				free(ioe, M_TEMP);
2969 				error = EFAULT;
2970 				goto fail;
2971 			}
2972 			switch (ioe->rs_num) {
2973 #ifdef ALTQ
2974 			case PF_RULESET_ALTQ:
2975 				if (ioe->anchor[0]) {
2976 					free(table, M_TEMP);
2977 					free(ioe, M_TEMP);
2978 					error = EINVAL;
2979 					goto fail;
2980 				}
2981 				if (!altqs_inactive_open || ioe->ticket !=
2982 				    ticket_altqs_inactive) {
2983 					free(table, M_TEMP);
2984 					free(ioe, M_TEMP);
2985 					error = EBUSY;
2986 					goto fail;
2987 				}
2988 				break;
2989 #endif /* ALTQ */
2990 			case PF_RULESET_TABLE:
2991 				rs = pf_find_ruleset(ioe->anchor);
2992 				if (rs == NULL || !rs->topen || ioe->ticket !=
2993 				     rs->tticket) {
2994 					free(table, M_TEMP);
2995 					free(ioe, M_TEMP);
2996 					error = EBUSY;
2997 					goto fail;
2998 				}
2999 				break;
3000 			default:
3001 				if (ioe->rs_num < 0 || ioe->rs_num >=
3002 				    PF_RULESET_MAX) {
3003 					free(table, M_TEMP);
3004 					free(ioe, M_TEMP);
3005 					error = EINVAL;
3006 					goto fail;
3007 				}
3008 				rs = pf_find_ruleset(ioe->anchor);
3009 				if (rs == NULL ||
3010 				    !rs->rules[ioe->rs_num].inactive.open ||
3011 				    rs->rules[ioe->rs_num].inactive.ticket !=
3012 				    ioe->ticket) {
3013 					free(table, M_TEMP);
3014 					free(ioe, M_TEMP);
3015 					error = EBUSY;
3016 					goto fail;
3017 				}
3018 				break;
3019 			}
3020 		}
3021 		/* now do the commit - no errors should happen here */
3022 		for (i = 0; i < io->size; i++) {
3023 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
3024 				free(table, M_TEMP);
3025 				free(ioe, M_TEMP);
3026 				error = EFAULT;
3027 				goto fail;
3028 			}
3029 			switch (ioe->rs_num) {
3030 #ifdef ALTQ
3031 			case PF_RULESET_ALTQ:
3032 				if ((error = pf_commit_altq(ioe->ticket))) {
3033 					free(table, M_TEMP);
3034 					free(ioe, M_TEMP);
3035 					goto fail; /* really bad */
3036 				}
3037 				break;
3038 #endif /* ALTQ */
3039 			case PF_RULESET_TABLE:
3040 				bzero(table, sizeof(*table));
3041 				strlcpy(table->pfrt_anchor, ioe->anchor,
3042 				    sizeof(table->pfrt_anchor));
3043 				if ((error = pfr_ina_commit(table, ioe->ticket,
3044 				    NULL, NULL, 0))) {
3045 					free(table, M_TEMP);
3046 					free(ioe, M_TEMP);
3047 					goto fail; /* really bad */
3048 				}
3049 				break;
3050 			default:
3051 				if ((error = pf_commit_rules(ioe->ticket,
3052 				    ioe->rs_num, ioe->anchor))) {
3053 					free(table, M_TEMP);
3054 					free(ioe, M_TEMP);
3055 					goto fail; /* really bad */
3056 				}
3057 				break;
3058 			}
3059 		}
3060 		free(table, M_TEMP);
3061 		free(ioe, M_TEMP);
3062 		break;
3063 	}
3064 
3065 	case DIOCGETSRCNODES: {
3066 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
3067 		struct pf_src_node	*n, *p, *pstore;
3068 		u_int32_t		 nr = 0;
3069 		int			 space = psn->psn_len;
3070 
3071 		if (space == 0) {
3072 			RB_FOREACH(n, pf_src_tree, &tree_src_tracking)
3073 				nr++;
3074 			psn->psn_len = sizeof(struct pf_src_node) * nr;
3075 			break;
3076 		}
3077 
3078 		pstore = malloc(sizeof(*pstore), M_TEMP, M_WAITOK);
3079 
3080 		p = psn->psn_src_nodes;
3081 		RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
3082 			int	secs = time_second, diff;
3083 
3084 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
3085 				break;
3086 
3087 			bcopy(n, pstore, sizeof(*pstore));
3088 			if (n->rule.ptr != NULL)
3089 				pstore->rule.nr = n->rule.ptr->nr;
3090 			pstore->creation = secs - pstore->creation;
3091 			if (pstore->expire > secs)
3092 				pstore->expire -= secs;
3093 			else
3094 				pstore->expire = 0;
3095 
3096 			/* adjust the connection rate estimate */
3097 			diff = secs - n->conn_rate.last;
3098 			if (diff >= n->conn_rate.seconds)
3099 				pstore->conn_rate.count = 0;
3100 			else
3101 				pstore->conn_rate.count -=
3102 				    n->conn_rate.count * diff /
3103 				    n->conn_rate.seconds;
3104 
3105 			error = copyout(pstore, p, sizeof(*p));
3106 			if (error) {
3107 				free(pstore, M_TEMP);
3108 				goto fail;
3109 			}
3110 			p++;
3111 			nr++;
3112 		}
3113 		psn->psn_len = sizeof(struct pf_src_node) * nr;
3114 
3115 		free(pstore, M_TEMP);
3116 		break;
3117 	}
3118 
3119 	case DIOCCLRSRCNODES: {
3120 		struct pf_src_node	*n;
3121 		struct pf_state		*state;
3122 
3123 		RB_FOREACH(state, pf_state_tree_id, &tree_id) {
3124 			state->src_node = NULL;
3125 			state->nat_src_node = NULL;
3126 		}
3127 		RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
3128 			n->expire = 1;
3129 			n->states = 0;
3130 		}
3131 		pf_purge_expired_src_nodes(1);
3132 		pf_status.src_nodes = 0;
3133 		break;
3134 	}
3135 
3136 	case DIOCKILLSRCNODES: {
3137 		struct pf_src_node	*sn;
3138 		struct pf_state		*ps;
3139 		struct pfioc_src_node_kill *psnk = \
3140 			(struct pfioc_src_node_kill *) addr;
3141 		int			killed = 0;
3142 
3143 		RB_FOREACH(sn, pf_src_tree, &tree_src_tracking) {
3144         		if (PF_MATCHA(psnk->psnk_src.neg, \
3145 				      &psnk->psnk_src.addr.v.a.addr, \
3146 				      &psnk->psnk_src.addr.v.a.mask, \
3147 				      &sn->addr, sn->af) &&
3148 			    PF_MATCHA(psnk->psnk_dst.neg, \
3149 				      &psnk->psnk_dst.addr.v.a.addr, \
3150 				      &psnk->psnk_dst.addr.v.a.mask, \
3151 				      &sn->raddr, sn->af)) {
3152 				/* Handle state to src_node linkage */
3153 				if (sn->states != 0) {
3154 					RB_FOREACH(ps, pf_state_tree_id,
3155 					    &tree_id) {
3156 						if (ps->src_node == sn)
3157 							ps->src_node = NULL;
3158 						if (ps->nat_src_node == sn)
3159 							ps->nat_src_node = NULL;
3160 					}
3161 					sn->states = 0;
3162 				}
3163 				sn->expire = 1;
3164 				killed++;
3165 			}
3166 		}
3167 
3168 		if (killed > 0)
3169 			pf_purge_expired_src_nodes(1);
3170 
3171 		psnk->psnk_af = killed;
3172 		break;
3173 	}
3174 
3175 	case DIOCSETHOSTID: {
3176 		u_int32_t	*hid = (u_int32_t *)addr;
3177 
3178 		if (*hid == 0)
3179 			pf_status.hostid = cprng_fast32();
3180 		else
3181 			pf_status.hostid = *hid;
3182 		break;
3183 	}
3184 
3185 	case DIOCOSFPFLUSH:
3186 		pf_osfp_flush();
3187 		break;
3188 
3189 	case DIOCIGETIFACES: {
3190 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3191 
3192 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
3193 			error = ENODEV;
3194 			break;
3195 		}
3196 		error = pfi_get_ifaces(io->pfiio_name, io->pfiio_buffer,
3197 		    &io->pfiio_size);
3198 		break;
3199 	}
3200 
3201 	case DIOCSETIFFLAG: {
3202 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3203 
3204 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3205 		break;
3206 	}
3207 
3208 	case DIOCCLRIFFLAG: {
3209 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3210 
3211 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3212 		break;
3213 	}
3214 
3215 	case DIOCSETLCK: {
3216 		pf_state_lock = *(uint32_t*)addr;
3217 		break;
3218 	}
3219 
3220 	default:
3221 		error = ENODEV;
3222 		break;
3223 	}
3224 fail:
3225 	splx(s);
3226 	if (flags & FWRITE)
3227 		rw_exit_write(&pf_consistency_lock);
3228 	else
3229 		rw_exit_read(&pf_consistency_lock);
3230 	return (error);
3231 }
3232 
3233 #ifdef __NetBSD__
3234 #ifdef INET
3235 static int
3236 pfil4_wrapper(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
3237 {
3238 	int error;
3239 
3240 	/*
3241 	 * ensure that mbufs are writable beforehand
3242 	 * as it's assumed by pf code.
3243 	 * ip hdr (60 bytes) + tcp hdr (60 bytes) should be enough.
3244 	 * XXX inefficient
3245 	 */
3246 	error = m_makewritable(mp, 0, 60 + 60, M_DONTWAIT);
3247 	if (error) {
3248 		m_freem(*mp);
3249 		*mp = NULL;
3250 		return error;
3251 	}
3252 
3253 	/*
3254 	 * If the packet is out-bound, we can't delay checksums
3255 	 * here.  For in-bound, the checksum has already been
3256 	 * validated.
3257 	 */
3258 	if (dir == PFIL_OUT) {
3259 		if ((*mp)->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
3260 			in_delayed_cksum(*mp);
3261 			(*mp)->m_pkthdr.csum_flags &=
3262 			    ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
3263 		}
3264 	}
3265 
3266 	if (pf_test(dir == PFIL_OUT ? PF_OUT : PF_IN, ifp, mp, NULL)
3267 	    != PF_PASS) {
3268 		m_freem(*mp);
3269 		*mp = NULL;
3270 		return EHOSTUNREACH;
3271 	}
3272 
3273 	/*
3274 	 * we're not compatible with fast-forward.
3275 	 */
3276 
3277 	if (dir == PFIL_IN && *mp) {
3278 		(*mp)->m_flags &= ~M_CANFASTFWD;
3279 	}
3280 
3281 	return (0);
3282 }
3283 #endif /* INET */
3284 
3285 #ifdef INET6
3286 static int
3287 pfil6_wrapper(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
3288 {
3289 	int error;
3290 
3291 	/*
3292 	 * ensure that mbufs are writable beforehand
3293 	 * as it's assumed by pf code.
3294 	 * XXX inefficient
3295 	 */
3296 	error = m_makewritable(mp, 0, M_COPYALL, M_DONTWAIT);
3297 	if (error) {
3298 		m_freem(*mp);
3299 		*mp = NULL;
3300 		return error;
3301 	}
3302 
3303 	/*
3304 	 * If the packet is out-bound, we can't delay checksums
3305 	 * here.  For in-bound, the checksum has already been
3306 	 * validated.
3307 	 */
3308 	if (dir == PFIL_OUT) {
3309 		if ((*mp)->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) {
3310 			in6_delayed_cksum(*mp);
3311 			(*mp)->m_pkthdr.csum_flags &=
3312 			    ~(M_CSUM_TCPv6|M_CSUM_UDPv6);
3313 		}
3314 	}
3315 
3316 	if (pf_test6(dir == PFIL_OUT ? PF_OUT : PF_IN, ifp, mp, NULL)
3317 	    != PF_PASS) {
3318 		m_freem(*mp);
3319 		*mp = NULL;
3320 		return EHOSTUNREACH;
3321 	} else
3322 		return (0);
3323 }
3324 #endif /* INET6 */
3325 
3326 static int
3327 pf_pfil_attach(void)
3328 {
3329 	pfil_head_t *ph_inet;
3330 #ifdef INET6
3331 	pfil_head_t *ph_inet6;
3332 #endif /* INET6 */
3333 	int error;
3334 
3335 	if (pf_pfil_attached)
3336 		return (EBUSY);
3337 
3338 	ph_inet = pfil_head_get(PFIL_TYPE_AF, (void *)AF_INET);
3339 	if (ph_inet)
3340 		error = pfil_add_hook((void *)pfil4_wrapper, NULL,
3341 		    PFIL_IN|PFIL_OUT, ph_inet);
3342 	else
3343 		error = ENOENT;
3344 	if (error)
3345 		return (error);
3346 
3347 #ifdef INET6
3348 	ph_inet6 = pfil_head_get(PFIL_TYPE_AF, (void *)AF_INET6);
3349 	if (ph_inet6)
3350 		error = pfil_add_hook((void *)pfil6_wrapper, NULL,
3351 		    PFIL_IN|PFIL_OUT, ph_inet6);
3352 	else
3353 		error = ENOENT;
3354 	if (error)
3355 		goto bad;
3356 #endif /* INET6 */
3357 
3358 	pf_pfil_attached = 1;
3359 
3360 	return (0);
3361 
3362 #ifdef INET6
3363 bad:
3364 	pfil_remove_hook(pfil4_wrapper, NULL, PFIL_IN|PFIL_OUT, ph_inet);
3365 #endif /* INET6 */
3366 
3367 	return (error);
3368 }
3369 
3370 static int
3371 pf_pfil_detach(void)
3372 {
3373 	pfil_head_t *ph_inet;
3374 #ifdef INET6
3375 	pfil_head_t *ph_inet6;
3376 #endif /* INET6 */
3377 
3378 	if (pf_pfil_attached == 0)
3379 		return (EBUSY);
3380 
3381 	ph_inet = pfil_head_get(PFIL_TYPE_AF, (void *)AF_INET);
3382 	if (ph_inet)
3383 		pfil_remove_hook((void *)pfil4_wrapper, NULL,
3384 		    PFIL_IN|PFIL_OUT, ph_inet);
3385 #ifdef INET6
3386 	ph_inet6 = pfil_head_get(PFIL_TYPE_AF, (void *)AF_INET6);
3387 	if (ph_inet6)
3388 		pfil_remove_hook((void *)pfil6_wrapper, NULL,
3389 		    PFIL_IN|PFIL_OUT, ph_inet6);
3390 #endif /* INET6 */
3391 	pf_pfil_attached = 0;
3392 
3393 	return (0);
3394 }
3395 #endif /* __NetBSD__ */
3396 
3397 #if defined(__NetBSD__)
3398 MODULE(MODULE_CLASS_DRIVER, pf, "bpf");
3399 
3400 static int
3401 pf_modcmd(modcmd_t cmd, void *opaque)
3402 {
3403 #ifdef _MODULE
3404 	extern void pflogattach(int);
3405 	extern void pflogdetach(void);
3406 
3407 	devmajor_t cmajor = NODEVMAJOR, bmajor = NODEVMAJOR;
3408 	int err;
3409 
3410 	switch (cmd) {
3411 	case MODULE_CMD_INIT:
3412 		err = devsw_attach("pf", NULL, &bmajor, &pf_cdevsw, &cmajor);
3413 		if (err)
3414 			return err;
3415 		pfattach(1);
3416 		pflogattach(1);
3417 		return 0;
3418 	case MODULE_CMD_FINI:
3419 		if (pf_status.running) {
3420 			return EBUSY;
3421 		} else {
3422 			pfdetach();
3423 			pflogdetach();
3424 			return devsw_detach(NULL, &pf_cdevsw);
3425 		}
3426 	default:
3427 		return ENOTTY;
3428 	}
3429 #else
3430 	if (cmd == MODULE_CMD_INIT)
3431 		return 0;
3432 	return ENOTTY;
3433 #endif
3434 }
3435 #endif
3436