xref: /netbsd-src/sys/dist/pf/net/pf_ioctl.c (revision bbde328be4e75ea9ad02e9715ea13ca54b797ada)
1 /*	$NetBSD: pf_ioctl.c,v 1.41 2010/04/13 13:08:16 ahoka 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 1.41 2010/04/13 13:08:16 ahoka Exp $");
41 
42 #ifdef _KERNEL_OPT
43 #include "opt_inet.h"
44 #include "opt_pfil_hooks.h"
45 #endif
46 
47 #include "pfsync.h"
48 
49 #include <sys/param.h>
50 #include <sys/systm.h>
51 #include <sys/mbuf.h>
52 #include <sys/filio.h>
53 #include <sys/fcntl.h>
54 #include <sys/socket.h>
55 #include <sys/socketvar.h>
56 #include <sys/kernel.h>
57 #include <sys/time.h>
58 #include <sys/pool.h>
59 #include <sys/proc.h>
60 #include <sys/malloc.h>
61 #include <sys/kthread.h>
62 #include <sys/rwlock.h>
63 #include <uvm/uvm_extern.h>
64 #ifdef __NetBSD__
65 #include <sys/conf.h>
66 #include <sys/lwp.h>
67 #include <sys/kauth.h>
68 #include <sys/module.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 struct pf_rule		 pf_default_rule;
141 #ifdef __NetBSD__
142 krwlock_t		 pf_consistency_lock;
143 #else
144 struct rwlock		 pf_consistency_lock = RWLOCK_INITIALIZER("pfcnslk");
145 #endif /* __NetBSD__ */
146 #ifdef ALTQ
147 static int		 pf_altq_running;
148 #endif
149 
150 #define	TAGID_MAX	 50000
151 TAILQ_HEAD(pf_tags, pf_tagname)	pf_tags = TAILQ_HEAD_INITIALIZER(pf_tags),
152 				pf_qids = TAILQ_HEAD_INITIALIZER(pf_qids);
153 
154 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
155 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
156 #endif
157 u_int16_t		 tagname2tag(struct pf_tags *, char *);
158 void			 tag2tagname(struct pf_tags *, u_int16_t, char *);
159 void			 tag_unref(struct pf_tags *, u_int16_t);
160 int			 pf_rtlabel_add(struct pf_addr_wrap *);
161 void			 pf_rtlabel_remove(struct pf_addr_wrap *);
162 void			 pf_rtlabel_copyout(struct pf_addr_wrap *);
163 
164 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) printf x
165 
166 #ifdef __NetBSD__
167 const struct cdevsw pf_cdevsw = {
168 	pfopen, pfclose, noread, nowrite, pfioctl,
169 	nostop, notty, nopoll, nommap, nokqfilter, D_OTHER
170 };
171 
172 static int pfil4_wrapper(void *, struct mbuf **, struct ifnet *, int);
173 #ifdef INET6
174 static int pfil6_wrapper(void *, struct mbuf **, struct ifnet *, int);
175 #endif /* INET6 */
176 
177 static int pf_pfil_attach(void);
178 static int pf_pfil_detach(void);
179 
180 static int pf_pfil_attached;
181 
182 static kauth_listener_t pf_listener;
183 #endif /* __NetBSD__ */
184 
185 #ifdef __NetBSD__
186 static int
187 pf_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
188     void *arg0, void *arg1, void *arg2, void *arg3)
189 {
190 	int result;
191 	enum kauth_network_req req;
192 
193 	result = KAUTH_RESULT_DEFER;
194 	req = (enum kauth_network_req)arg0;
195 
196 	if (action != KAUTH_NETWORK_FIREWALL)
197 		return result;
198 
199 	/* These must have came from device context. */
200 	if ((req == KAUTH_REQ_NETWORK_FIREWALL_FW) ||
201 	    (req == KAUTH_REQ_NETWORK_FIREWALL_NAT))
202 		result = KAUTH_RESULT_ALLOW;
203 
204 	return result;
205 }
206 #endif /* __NetBSD__ */
207 
208 void
209 pfattach(int num)
210 {
211 	u_int32_t *timeout = pf_default_rule.timeout;
212 
213 #ifdef __NetBSD__
214 	pool_init(&pf_rule_pl, sizeof(struct pf_rule), 0, 0, 0, "pfrulepl",
215 	    &pool_allocator_nointr, IPL_NONE);
216 	pool_init(&pf_src_tree_pl, sizeof(struct pf_src_node), 0, 0, 0,
217 	    "pfsrctrpl", NULL, IPL_SOFTNET);
218 	pool_init(&pf_state_pl, sizeof(struct pf_state), 0, 0, 0, "pfstatepl",
219 	    NULL, IPL_SOFTNET);
220 	pool_init(&pf_state_key_pl, sizeof(struct pf_state_key), 0, 0, 0,
221 	    "pfstatekeypl", NULL, IPL_SOFTNET);
222 	pool_init(&pf_altq_pl, sizeof(struct pf_altq), 0, 0, 0, "pfaltqpl",
223 	    &pool_allocator_nointr, IPL_NONE);
224 	pool_init(&pf_pooladdr_pl, sizeof(struct pf_pooladdr), 0, 0, 0,
225 	    "pfpooladdrpl", &pool_allocator_nointr, IPL_NONE);
226 #else
227 	pool_init(&pf_rule_pl, sizeof(struct pf_rule), 0, 0, 0, "pfrulepl",
228 	    &pool_allocator_nointr);
229 	pool_init(&pf_src_tree_pl, sizeof(struct pf_src_node), 0, 0, 0,
230 	    "pfsrctrpl", NULL);
231 	pool_init(&pf_state_pl, sizeof(struct pf_state), 0, 0, 0, "pfstatepl",
232 	    NULL);
233 	pool_init(&pf_state_key_pl, sizeof(struct pf_state_key), 0, 0, 0,
234 	    "pfstatekeypl", NULL);
235 	pool_init(&pf_altq_pl, sizeof(struct pf_altq), 0, 0, 0, "pfaltqpl",
236 	    &pool_allocator_nointr);
237 	pool_init(&pf_pooladdr_pl, sizeof(struct pf_pooladdr), 0, 0, 0,
238 	    "pfpooladdrpl", &pool_allocator_nointr);
239 #endif /* !__NetBSD__ */
240 
241 	pfr_initialize();
242 	pfi_initialize();
243 	pf_osfp_initialize();
244 
245 	pool_sethardlimit(pf_pool_limits[PF_LIMIT_STATES].pp,
246 	    pf_pool_limits[PF_LIMIT_STATES].limit, NULL, 0);
247 
248 	if (ctob(physmem) <= 100*1024*1024)
249 		pf_pool_limits[PF_LIMIT_TABLE_ENTRIES].limit =
250 		    PFR_KENTRY_HIWAT_SMALL;
251 
252 	RB_INIT(&tree_src_tracking);
253 	RB_INIT(&pf_anchors);
254 	pf_init_ruleset(&pf_main_ruleset);
255 	TAILQ_INIT(&pf_altqs[0]);
256 	TAILQ_INIT(&pf_altqs[1]);
257 	TAILQ_INIT(&pf_pabuf);
258 	pf_altqs_active = &pf_altqs[0];
259 	pf_altqs_inactive = &pf_altqs[1];
260 	TAILQ_INIT(&state_list);
261 
262 #ifdef __NetBSD__
263 	rw_init(&pf_consistency_lock);
264 #endif /* __NetBSD__ */
265 
266 	/* default rule should never be garbage collected */
267 	pf_default_rule.entries.tqe_prev = &pf_default_rule.entries.tqe_next;
268 	pf_default_rule.action = PF_PASS;
269 	pf_default_rule.nr = -1;
270 	pf_default_rule.rtableid = -1;
271 
272 	/* initialize default timeouts */
273 	timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
274 	timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
275 	timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
276 	timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
277 	timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
278 	timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
279 	timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
280 	timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
281 	timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
282 	timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
283 	timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
284 	timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
285 	timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
286 	timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
287 	timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
288 	timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
289 	timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
290 	timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
291 	timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
292 	timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
293 
294 	pf_normalize_init();
295 	bzero(&pf_status, sizeof(pf_status));
296 	pf_status.debug = PF_DEBUG_URGENT;
297 
298 	/* XXX do our best to avoid a conflict */
299 	pf_status.hostid = arc4random();
300 
301 	/* require process context to purge states, so perform in a thread */
302 #ifdef __NetBSD__
303 	if (kthread_create(PRI_NONE, 0, NULL, pf_purge_thread, NULL, NULL,
304 	    "pfpurge"))
305 		panic("pfpurge thread");
306 #else
307 	kthread_create_deferred(pf_thread_create, NULL);
308 #endif /* !__NetBSD__ */
309 
310 #ifdef __NetBSD__
311 	pf_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
312 	    pf_listener_cb, NULL);
313 #endif /* __NetBSD__ */
314 }
315 
316 #ifdef _MODULE
317 void
318 pfdetach(void)
319 {
320 	extern int		 pf_purge_thread_running;
321 	extern int		 pf_purge_thread_stop;
322 	struct pf_anchor	*anchor;
323 	struct pf_state		*state;
324 	struct pf_src_node	*node;
325 	struct pfioc_table	 pt;
326 	u_int32_t		 ticket;
327 	int			 i;
328 	char			 r = '\0';
329 
330 	pf_purge_thread_stop = 1;
331 	wakeup(pf_purge_thread);
332 
333 	/* wait until the kthread exits */
334 	while (pf_purge_thread_running)
335 		tsleep(&pf_purge_thread_running, PWAIT, "pfdown", 0);
336 
337 	(void)pf_pfil_detach();
338 
339 	pf_status.running = 0;
340 
341 	/* clear the rulesets */
342 	for (i = 0; i < PF_RULESET_MAX; i++)
343 		if (pf_begin_rules(&ticket, i, &r) == 0)
344 			pf_commit_rules(ticket, i, &r);
345 #ifdef ALTQ
346 	if (pf_begin_altq(&ticket) == 0)
347 		pf_commit_altq(ticket);
348 #endif /* ALTQ */
349 
350 	/* clear states */
351 	RB_FOREACH(state, pf_state_tree_id, &tree_id) {
352 		state->timeout = PFTM_PURGE;
353 #if NPFSYNC > 0
354 		state->sync_flags = PFSTATE_NOSYNC;
355 #endif /* NPFSYNC > 0 */
356 	}
357 	pf_purge_expired_states(pf_status.states);
358 #if NPFSYNC > 0
359 	pfsync_clear_states(pf_status.hostid, NULL);
360 #endif /* NPFSYNC > 0 */
361 
362 	/* clear source nodes */
363 	RB_FOREACH(state, pf_state_tree_id, &tree_id) {
364 		state->src_node = NULL;
365 		state->nat_src_node = NULL;
366 	}
367 	RB_FOREACH(node, pf_src_tree, &tree_src_tracking) {
368 		node->expire = 1;
369 		node->states = 0;
370 	}
371 	pf_purge_expired_src_nodes(0);
372 
373 	/* clear tables */
374 	memset(&pt, '\0', sizeof(pt));
375 	pfr_clr_tables(&pt.pfrio_table, &pt.pfrio_ndel, pt.pfrio_flags);
376 
377 	/* destroy anchors */
378 	while ((anchor = RB_MIN(pf_anchor_global, &pf_anchors)) != NULL) {
379 		for (i = 0; i < PF_RULESET_MAX; i++)
380 			if (pf_begin_rules(&ticket, i, anchor->name) == 0)
381 				pf_commit_rules(ticket, i, anchor->name);
382 	}
383 
384 	/* destroy main ruleset */
385 	pf_remove_if_empty_ruleset(&pf_main_ruleset);
386 
387 	/* destroy the pools */
388 	pool_destroy(&pf_pooladdr_pl);
389 	pool_destroy(&pf_altq_pl);
390 	pool_destroy(&pf_state_key_pl);
391 	pool_destroy(&pf_state_pl);
392 	pool_destroy(&pf_rule_pl);
393 	pool_destroy(&pf_src_tree_pl);
394 
395 	rw_destroy(&pf_consistency_lock);
396 
397 	/* destroy subsystems */
398 	pf_normalize_destroy();
399 	pf_osfp_destroy();
400 	pfr_destroy();
401 	pfi_destroy();
402 
403 	/* cleanup kauth listener */
404 	kauth_unlisten_scope(pf_listener);
405 }
406 #endif /* _MODULE */
407 
408 #ifndef __NetBSD__
409 void
410 pf_thread_create(void *v)
411 {
412 	if (kthread_create(pf_purge_thread, NULL, NULL, "pfpurge"))
413 		panic("pfpurge thread");
414 }
415 #endif /* !__NetBSD__ */
416 
417 int
418 pfopen(dev_t dev, int flags, int fmt, struct lwp *l)
419 {
420 	if (minor(dev) >= 1)
421 		return (ENXIO);
422 	return (0);
423 }
424 
425 int
426 pfclose(dev_t dev, int flags, int fmt, struct lwp *l)
427 {
428 	if (minor(dev) >= 1)
429 		return (ENXIO);
430 	return (0);
431 }
432 
433 struct pf_pool *
434 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
435     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
436     u_int8_t check_ticket)
437 {
438 	struct pf_ruleset	*ruleset;
439 	struct pf_rule		*rule;
440 	int			 rs_num;
441 
442 	ruleset = pf_find_ruleset(anchor);
443 	if (ruleset == NULL)
444 		return (NULL);
445 	rs_num = pf_get_ruleset_number(rule_action);
446 	if (rs_num >= PF_RULESET_MAX)
447 		return (NULL);
448 	if (active) {
449 		if (check_ticket && ticket !=
450 		    ruleset->rules[rs_num].active.ticket)
451 			return (NULL);
452 		if (r_last)
453 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
454 			    pf_rulequeue);
455 		else
456 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
457 	} else {
458 		if (check_ticket && ticket !=
459 		    ruleset->rules[rs_num].inactive.ticket)
460 			return (NULL);
461 		if (r_last)
462 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
463 			    pf_rulequeue);
464 		else
465 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
466 	}
467 	if (!r_last) {
468 		while ((rule != NULL) && (rule->nr != rule_number))
469 			rule = TAILQ_NEXT(rule, entries);
470 	}
471 	if (rule == NULL)
472 		return (NULL);
473 
474 	return (&rule->rpool);
475 }
476 
477 void
478 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
479 {
480 	struct pf_pooladdr	*mv_pool_pa;
481 
482 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
483 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
484 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
485 	}
486 }
487 
488 void
489 pf_empty_pool(struct pf_palist *poola)
490 {
491 	struct pf_pooladdr	*empty_pool_pa;
492 
493 	while ((empty_pool_pa = TAILQ_FIRST(poola)) != NULL) {
494 		pfi_dynaddr_remove(&empty_pool_pa->addr);
495 		pf_tbladdr_remove(&empty_pool_pa->addr);
496 		pfi_kif_unref(empty_pool_pa->kif, PFI_KIF_REF_RULE);
497 		TAILQ_REMOVE(poola, empty_pool_pa, entries);
498 		pool_put(&pf_pooladdr_pl, empty_pool_pa);
499 	}
500 }
501 
502 void
503 pf_rm_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
504 {
505 	if (rulequeue != NULL) {
506 		if (rule->states <= 0) {
507 			/*
508 			 * XXX - we need to remove the table *before* detaching
509 			 * the rule to make sure the table code does not delete
510 			 * the anchor under our feet.
511 			 */
512 			pf_tbladdr_remove(&rule->src.addr);
513 			pf_tbladdr_remove(&rule->dst.addr);
514 			if (rule->overload_tbl)
515 				pfr_detach_table(rule->overload_tbl);
516 		}
517 		TAILQ_REMOVE(rulequeue, rule, entries);
518 		rule->entries.tqe_prev = NULL;
519 		rule->nr = -1;
520 	}
521 
522 	if (rule->states > 0 || rule->src_nodes > 0 ||
523 	    rule->entries.tqe_prev != NULL)
524 		return;
525 	pf_tag_unref(rule->tag);
526 	pf_tag_unref(rule->match_tag);
527 #ifdef ALTQ
528 	if (rule->pqid != rule->qid)
529 		pf_qid_unref(rule->pqid);
530 	pf_qid_unref(rule->qid);
531 #endif
532 	pf_rtlabel_remove(&rule->src.addr);
533 	pf_rtlabel_remove(&rule->dst.addr);
534 	pfi_dynaddr_remove(&rule->src.addr);
535 	pfi_dynaddr_remove(&rule->dst.addr);
536 	if (rulequeue == NULL) {
537 		pf_tbladdr_remove(&rule->src.addr);
538 		pf_tbladdr_remove(&rule->dst.addr);
539 		if (rule->overload_tbl)
540 			pfr_detach_table(rule->overload_tbl);
541 	}
542 	pfi_kif_unref(rule->kif, PFI_KIF_REF_RULE);
543 	pf_anchor_remove(rule);
544 	pf_empty_pool(&rule->rpool.list);
545 	pool_put(&pf_rule_pl, rule);
546 }
547 
548 u_int16_t
549 tagname2tag(struct pf_tags *head, char *tagname)
550 {
551 	struct pf_tagname	*tag, *p = NULL;
552 	u_int16_t		 new_tagid = 1;
553 
554 	TAILQ_FOREACH(tag, head, entries)
555 		if (strcmp(tagname, tag->name) == 0) {
556 			tag->ref++;
557 			return (tag->tag);
558 		}
559 
560 	/*
561 	 * to avoid fragmentation, we do a linear search from the beginning
562 	 * and take the first free slot we find. if there is none or the list
563 	 * is empty, append a new entry at the end.
564 	 */
565 
566 	/* new entry */
567 	if (!TAILQ_EMPTY(head))
568 		for (p = TAILQ_FIRST(head); p != NULL &&
569 		    p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
570 			new_tagid = p->tag + 1;
571 
572 	if (new_tagid > TAGID_MAX)
573 		return (0);
574 
575 	/* allocate and fill new struct pf_tagname */
576 	tag = (struct pf_tagname *)malloc(sizeof(struct pf_tagname),
577 	    M_TEMP, M_NOWAIT);
578 	if (tag == NULL)
579 		return (0);
580 	bzero(tag, sizeof(struct pf_tagname));
581 	strlcpy(tag->name, tagname, sizeof(tag->name));
582 	tag->tag = new_tagid;
583 	tag->ref++;
584 
585 	if (p != NULL)	/* insert new entry before p */
586 		TAILQ_INSERT_BEFORE(p, tag, entries);
587 	else	/* either list empty or no free slot in between */
588 		TAILQ_INSERT_TAIL(head, tag, entries);
589 
590 	return (tag->tag);
591 }
592 
593 void
594 tag2tagname(struct pf_tags *head, u_int16_t tagid, char *p)
595 {
596 	struct pf_tagname	*tag;
597 
598 	TAILQ_FOREACH(tag, head, entries)
599 		if (tag->tag == tagid) {
600 			strlcpy(p, tag->name, PF_TAG_NAME_SIZE);
601 			return;
602 		}
603 }
604 
605 void
606 tag_unref(struct pf_tags *head, u_int16_t tag)
607 {
608 	struct pf_tagname	*p, *next;
609 
610 	if (tag == 0)
611 		return;
612 
613 	for (p = TAILQ_FIRST(head); p != NULL; p = next) {
614 		next = TAILQ_NEXT(p, entries);
615 		if (tag == p->tag) {
616 			if (--p->ref == 0) {
617 				TAILQ_REMOVE(head, p, entries);
618 				free(p, M_TEMP);
619 			}
620 			break;
621 		}
622 	}
623 }
624 
625 u_int16_t
626 pf_tagname2tag(char *tagname)
627 {
628 	return (tagname2tag(&pf_tags, tagname));
629 }
630 
631 void
632 pf_tag2tagname(u_int16_t tagid, char *p)
633 {
634 	tag2tagname(&pf_tags, tagid, p);
635 }
636 
637 void
638 pf_tag_ref(u_int16_t tag)
639 {
640 	struct pf_tagname *t;
641 
642 	TAILQ_FOREACH(t, &pf_tags, entries)
643 		if (t->tag == tag)
644 			break;
645 	if (t != NULL)
646 		t->ref++;
647 }
648 
649 void
650 pf_tag_unref(u_int16_t tag)
651 {
652 	tag_unref(&pf_tags, tag);
653 }
654 
655 int
656 pf_rtlabel_add(struct pf_addr_wrap *a)
657 {
658 #ifndef __NetBSD__
659 	if (a->type == PF_ADDR_RTLABEL &&
660 	    (a->v.rtlabel = rtlabel_name2id(a->v.rtlabelname)) == 0)
661 		return (-1);
662 #endif /* !__NetBSD__ */
663 	return (0);
664 }
665 
666 void
667 pf_rtlabel_remove(struct pf_addr_wrap *a)
668 {
669 #ifndef __NetBSD__
670 	if (a->type == PF_ADDR_RTLABEL)
671 		rtlabel_unref(a->v.rtlabel);
672 #endif /* !__NetBSD__ */
673 }
674 
675 void
676 pf_rtlabel_copyout(struct pf_addr_wrap *a)
677 {
678 #ifndef __NetBSD__
679 	const char	*name;
680 
681 	if (a->type == PF_ADDR_RTLABEL && a->v.rtlabel) {
682 		if ((name = rtlabel_id2name(a->v.rtlabel)) == NULL)
683 			strlcpy(a->v.rtlabelname, "?",
684 			    sizeof(a->v.rtlabelname));
685 		else
686 			strlcpy(a->v.rtlabelname, name,
687 			    sizeof(a->v.rtlabelname));
688 	}
689 #endif /* !__NetBSD__ */
690 }
691 
692 #ifdef ALTQ
693 u_int32_t
694 pf_qname2qid(char *qname)
695 {
696 	return ((u_int32_t)tagname2tag(&pf_qids, qname));
697 }
698 
699 void
700 pf_qid2qname(u_int32_t qid, char *p)
701 {
702 	tag2tagname(&pf_qids, (u_int16_t)qid, p);
703 }
704 
705 void
706 pf_qid_unref(u_int32_t qid)
707 {
708 	tag_unref(&pf_qids, (u_int16_t)qid);
709 }
710 
711 int
712 pf_begin_altq(u_int32_t *ticket)
713 {
714 	struct pf_altq	*altq;
715 	int		 error = 0;
716 
717 	/* Purge the old altq list */
718 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
719 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
720 		if (altq->qname[0] == 0) {
721 			/* detach and destroy the discipline */
722 			error = altq_remove(altq);
723 		} else
724 			pf_qid_unref(altq->qid);
725 		pool_put(&pf_altq_pl, altq);
726 	}
727 	if (error)
728 		return (error);
729 	*ticket = ++ticket_altqs_inactive;
730 	altqs_inactive_open = 1;
731 	return (0);
732 }
733 
734 int
735 pf_rollback_altq(u_int32_t ticket)
736 {
737 	struct pf_altq	*altq;
738 	int		 error = 0;
739 
740 	if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
741 		return (0);
742 	/* Purge the old altq list */
743 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
744 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
745 		if (altq->qname[0] == 0) {
746 			/* detach and destroy the discipline */
747 			error = altq_remove(altq);
748 		} else
749 			pf_qid_unref(altq->qid);
750 		pool_put(&pf_altq_pl, altq);
751 	}
752 	altqs_inactive_open = 0;
753 	return (error);
754 }
755 
756 int
757 pf_commit_altq(u_int32_t ticket)
758 {
759 	struct pf_altqqueue	*old_altqs;
760 	struct pf_altq		*altq;
761 	int			 s, err, error = 0;
762 
763 	if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
764 		return (EBUSY);
765 
766 	/* swap altqs, keep the old. */
767 	s = splsoftnet();
768 	old_altqs = pf_altqs_active;
769 	pf_altqs_active = pf_altqs_inactive;
770 	pf_altqs_inactive = old_altqs;
771 	ticket_altqs_active = ticket_altqs_inactive;
772 
773 	/* Attach new disciplines */
774 	TAILQ_FOREACH(altq, pf_altqs_active, entries) {
775 		if (altq->qname[0] == 0) {
776 			/* attach the discipline */
777 			error = altq_pfattach(altq);
778 			if (error == 0 && pf_altq_running)
779 				error = pf_enable_altq(altq);
780 			if (error != 0) {
781 				splx(s);
782 				return (error);
783 			}
784 		}
785 	}
786 
787 	/* Purge the old altq list */
788 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
789 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
790 		if (altq->qname[0] == 0) {
791 			/* detach and destroy the discipline */
792 			if (pf_altq_running)
793 				error = pf_disable_altq(altq);
794 			err = altq_pfdetach(altq);
795 			if (err != 0 && error == 0)
796 				error = err;
797 			err = altq_remove(altq);
798 			if (err != 0 && error == 0)
799 				error = err;
800 		} else
801 			pf_qid_unref(altq->qid);
802 		pool_put(&pf_altq_pl, altq);
803 	}
804 	splx(s);
805 
806 	altqs_inactive_open = 0;
807 	return (error);
808 }
809 
810 int
811 pf_enable_altq(struct pf_altq *altq)
812 {
813 	struct ifnet		*ifp;
814 	struct tb_profile	 tb;
815 	int			 s, error = 0;
816 
817 	if ((ifp = ifunit(altq->ifname)) == NULL)
818 		return (EINVAL);
819 
820 	if (ifp->if_snd.altq_type != ALTQT_NONE)
821 		error = altq_enable(&ifp->if_snd);
822 
823 	/* set tokenbucket regulator */
824 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
825 		tb.rate = altq->ifbandwidth;
826 		tb.depth = altq->tbrsize;
827 		s = splnet();
828 		error = tbr_set(&ifp->if_snd, &tb);
829 		splx(s);
830 	}
831 
832 	return (error);
833 }
834 
835 int
836 pf_disable_altq(struct pf_altq *altq)
837 {
838 	struct ifnet		*ifp;
839 	struct tb_profile	 tb;
840 	int			 s, error;
841 
842 	if ((ifp = ifunit(altq->ifname)) == NULL)
843 		return (EINVAL);
844 
845 	/*
846 	 * when the discipline is no longer referenced, it was overridden
847 	 * by a new one.  if so, just return.
848 	 */
849 	if (altq->altq_disc != ifp->if_snd.altq_disc)
850 		return (0);
851 
852 	error = altq_disable(&ifp->if_snd);
853 
854 	if (error == 0) {
855 		/* clear tokenbucket regulator */
856 		tb.rate = 0;
857 		s = splnet();
858 		error = tbr_set(&ifp->if_snd, &tb);
859 		splx(s);
860 	}
861 
862 	return (error);
863 }
864 #endif /* ALTQ */
865 
866 int
867 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
868 {
869 	struct pf_ruleset	*rs;
870 	struct pf_rule		*rule;
871 
872 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
873 		return (EINVAL);
874 	rs = pf_find_or_create_ruleset(anchor);
875 	if (rs == NULL)
876 		return (EINVAL);
877 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
878 		pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
879 		rs->rules[rs_num].inactive.rcount--;
880 	}
881 	*ticket = ++rs->rules[rs_num].inactive.ticket;
882 	rs->rules[rs_num].inactive.open = 1;
883 	return (0);
884 }
885 
886 int
887 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
888 {
889 	struct pf_ruleset	*rs;
890 	struct pf_rule		*rule;
891 
892 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
893 		return (EINVAL);
894 	rs = pf_find_ruleset(anchor);
895 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
896 	    rs->rules[rs_num].inactive.ticket != ticket)
897 		return (0);
898 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
899 		pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
900 		rs->rules[rs_num].inactive.rcount--;
901 	}
902 	rs->rules[rs_num].inactive.open = 0;
903 	return (0);
904 }
905 
906 #define PF_MD5_UPD(st, elm)						\
907 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
908 
909 #define PF_MD5_UPD_STR(st, elm)						\
910 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
911 
912 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
913 		(stor) = htonl((st)->elm);				\
914 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
915 } while (0)
916 
917 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
918 		(stor) = htons((st)->elm);				\
919 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
920 } while (0)
921 
922 void
923 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
924 {
925 	PF_MD5_UPD(pfr, addr.type);
926 	switch (pfr->addr.type) {
927 		case PF_ADDR_DYNIFTL:
928 			PF_MD5_UPD(pfr, addr.v.ifname);
929 			PF_MD5_UPD(pfr, addr.iflags);
930 			break;
931 		case PF_ADDR_TABLE:
932 			PF_MD5_UPD(pfr, addr.v.tblname);
933 			break;
934 		case PF_ADDR_ADDRMASK:
935 			/* XXX ignore af? */
936 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
937 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
938 			break;
939 		case PF_ADDR_RTLABEL:
940 			PF_MD5_UPD(pfr, addr.v.rtlabelname);
941 			break;
942 	}
943 
944 	PF_MD5_UPD(pfr, port[0]);
945 	PF_MD5_UPD(pfr, port[1]);
946 	PF_MD5_UPD(pfr, neg);
947 	PF_MD5_UPD(pfr, port_op);
948 }
949 
950 void
951 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
952 {
953 	u_int16_t x;
954 	u_int32_t y;
955 
956 	pf_hash_rule_addr(ctx, &rule->src);
957 	pf_hash_rule_addr(ctx, &rule->dst);
958 	PF_MD5_UPD_STR(rule, label);
959 	PF_MD5_UPD_STR(rule, ifname);
960 	PF_MD5_UPD_STR(rule, match_tagname);
961 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
962 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
963 	PF_MD5_UPD_HTONL(rule, prob, y);
964 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
965 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
966 	PF_MD5_UPD(rule, uid.op);
967 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
968 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
969 	PF_MD5_UPD(rule, gid.op);
970 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
971 	PF_MD5_UPD(rule, action);
972 	PF_MD5_UPD(rule, direction);
973 	PF_MD5_UPD(rule, af);
974 	PF_MD5_UPD(rule, quick);
975 	PF_MD5_UPD(rule, ifnot);
976 	PF_MD5_UPD(rule, match_tag_not);
977 	PF_MD5_UPD(rule, natpass);
978 	PF_MD5_UPD(rule, keep_state);
979 	PF_MD5_UPD(rule, proto);
980 	PF_MD5_UPD(rule, type);
981 	PF_MD5_UPD(rule, code);
982 	PF_MD5_UPD(rule, flags);
983 	PF_MD5_UPD(rule, flagset);
984 	PF_MD5_UPD(rule, allow_opts);
985 	PF_MD5_UPD(rule, rt);
986 	PF_MD5_UPD(rule, tos);
987 }
988 
989 int
990 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
991 {
992 	struct pf_ruleset	*rs;
993 	struct pf_rule		*rule, **old_array;
994 	struct pf_rulequeue	*old_rules;
995 	int			 s, error;
996 	u_int32_t		 old_rcount;
997 
998 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
999 		return (EINVAL);
1000 	rs = pf_find_ruleset(anchor);
1001 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1002 	    ticket != rs->rules[rs_num].inactive.ticket)
1003 		return (EBUSY);
1004 
1005 	/* Calculate checksum for the main ruleset */
1006 	if (rs == &pf_main_ruleset) {
1007 		error = pf_setup_pfsync_matching(rs);
1008 		if (error != 0)
1009 			return (error);
1010 	}
1011 
1012 	/* Swap rules, keep the old. */
1013 	s = splsoftnet();
1014 	old_rules = rs->rules[rs_num].active.ptr;
1015 	old_rcount = rs->rules[rs_num].active.rcount;
1016 	old_array = rs->rules[rs_num].active.ptr_array;
1017 
1018 	rs->rules[rs_num].active.ptr =
1019 	    rs->rules[rs_num].inactive.ptr;
1020 	rs->rules[rs_num].active.ptr_array =
1021 	    rs->rules[rs_num].inactive.ptr_array;
1022 	rs->rules[rs_num].active.rcount =
1023 	    rs->rules[rs_num].inactive.rcount;
1024 	rs->rules[rs_num].inactive.ptr = old_rules;
1025 	rs->rules[rs_num].inactive.ptr_array = old_array;
1026 	rs->rules[rs_num].inactive.rcount = old_rcount;
1027 
1028 	rs->rules[rs_num].active.ticket =
1029 	    rs->rules[rs_num].inactive.ticket;
1030 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1031 
1032 
1033 	/* Purge the old rule list. */
1034 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1035 		pf_rm_rule(old_rules, rule);
1036 	if (rs->rules[rs_num].inactive.ptr_array)
1037 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1038 	rs->rules[rs_num].inactive.ptr_array = NULL;
1039 	rs->rules[rs_num].inactive.rcount = 0;
1040 	rs->rules[rs_num].inactive.open = 0;
1041 	pf_remove_if_empty_ruleset(rs);
1042 	splx(s);
1043 	return (0);
1044 }
1045 
1046 void
1047 pf_state_export(struct pfsync_state *sp, struct pf_state_key *sk,
1048    struct pf_state *s)
1049 {
1050 	int secs = time_second;
1051 	bzero(sp, sizeof(struct pfsync_state));
1052 
1053 	/* copy from state key */
1054 	sp->lan.addr = sk->lan.addr;
1055 	sp->lan.port = sk->lan.port;
1056 	sp->gwy.addr = sk->gwy.addr;
1057 	sp->gwy.port = sk->gwy.port;
1058 	sp->ext.addr = sk->ext.addr;
1059 	sp->ext.port = sk->ext.port;
1060 	sp->proto = sk->proto;
1061 	sp->af = sk->af;
1062 	sp->direction = sk->direction;
1063 
1064 	/* copy from state */
1065 	memcpy(&sp->id, &s->id, sizeof(sp->id));
1066 	sp->creatorid = s->creatorid;
1067 	strlcpy(sp->ifname, s->kif->pfik_name, sizeof(sp->ifname));
1068 	pf_state_peer_to_pfsync(&s->src, &sp->src);
1069 	pf_state_peer_to_pfsync(&s->dst, &sp->dst);
1070 
1071 	sp->rule = s->rule.ptr->nr;
1072 	sp->nat_rule = (s->nat_rule.ptr == NULL) ?  -1 : s->nat_rule.ptr->nr;
1073 	sp->anchor = (s->anchor.ptr == NULL) ?  -1 : s->anchor.ptr->nr;
1074 
1075 	pf_state_counter_to_pfsync(s->bytes[0], sp->bytes[0]);
1076 	pf_state_counter_to_pfsync(s->bytes[1], sp->bytes[1]);
1077 	pf_state_counter_to_pfsync(s->packets[0], sp->packets[0]);
1078 	pf_state_counter_to_pfsync(s->packets[1], sp->packets[1]);
1079 	sp->creation = secs - s->creation;
1080 	sp->expire = pf_state_expires(s);
1081 	sp->log = s->log;
1082 	sp->allow_opts = s->allow_opts;
1083 	sp->timeout = s->timeout;
1084 
1085 	if (s->src_node)
1086 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
1087 	if (s->nat_src_node)
1088 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
1089 
1090 	if (sp->expire > secs)
1091 		sp->expire -= secs;
1092 	else
1093 		sp->expire = 0;
1094 
1095 }
1096 
1097 void
1098 pf_state_import(struct pfsync_state *sp, struct pf_state_key *sk,
1099    struct pf_state *s)
1100 {
1101 	/* copy to state key */
1102 	sk->lan.addr = sp->lan.addr;
1103 	sk->lan.port = sp->lan.port;
1104 	sk->gwy.addr = sp->gwy.addr;
1105 	sk->gwy.port = sp->gwy.port;
1106 	sk->ext.addr = sp->ext.addr;
1107 	sk->ext.port = sp->ext.port;
1108 	sk->proto = sp->proto;
1109 	sk->af = sp->af;
1110 	sk->direction = sp->direction;
1111 
1112 	/* copy to state */
1113 	memcpy(&s->id, &sp->id, sizeof(sp->id));
1114 	s->creatorid = sp->creatorid;
1115 	strlcpy(sp->ifname, s->kif->pfik_name, sizeof(sp->ifname));
1116 	pf_state_peer_from_pfsync(&sp->src, &s->src);
1117 	pf_state_peer_from_pfsync(&sp->dst, &s->dst);
1118 
1119 	s->rule.ptr = &pf_default_rule;
1120 	s->nat_rule.ptr = NULL;
1121 	s->anchor.ptr = NULL;
1122 	s->rt_kif = NULL;
1123 	s->creation = time_second;
1124 	s->pfsync_time = 0;
1125 	s->packets[0] = s->packets[1] = 0;
1126 	s->bytes[0] = s->bytes[1] = 0;
1127 }
1128 
1129 int
1130 pf_setup_pfsync_matching(struct pf_ruleset *rs)
1131 {
1132 	MD5_CTX			 ctx;
1133 	struct pf_rule		*rule;
1134 	int			 rs_cnt;
1135 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
1136 
1137 	MD5Init(&ctx);
1138 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1139 		/* XXX PF_RULESET_SCRUB as well? */
1140 		if (rs_cnt == PF_RULESET_SCRUB)
1141 			continue;
1142 
1143 		if (rs->rules[rs_cnt].inactive.ptr_array)
1144 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1145 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
1146 
1147 		if (rs->rules[rs_cnt].inactive.rcount) {
1148 			rs->rules[rs_cnt].inactive.ptr_array =
1149 			    malloc(sizeof(void *) *
1150 			    rs->rules[rs_cnt].inactive.rcount,
1151 			    M_TEMP, M_NOWAIT);
1152 
1153 			if (!rs->rules[rs_cnt].inactive.ptr_array)
1154 				return (ENOMEM);
1155 		}
1156 
1157 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1158 		    entries) {
1159 			pf_hash_rule(&ctx, rule);
1160 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1161 		}
1162 	}
1163 
1164 	MD5Final(digest, &ctx);
1165 	memcpy(pf_status.pf_chksum, digest, sizeof(pf_status.pf_chksum));
1166 	return (0);
1167 }
1168 
1169 int
1170 pfioctl(dev_t dev, u_long cmd, void *addr, int flags, struct lwp *l)
1171 {
1172 	struct pf_pooladdr	*pa = NULL;
1173 	struct pf_pool		*pool = NULL;
1174 	int			 s;
1175 	int			 error = 0;
1176 
1177 	/* XXX keep in sync with switch() below */
1178 	if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FIREWALL,
1179 	    KAUTH_REQ_NETWORK_FIREWALL_FW, NULL, NULL, NULL))
1180 		switch (cmd) {
1181 		case DIOCGETRULES:
1182 		case DIOCGETRULE:
1183 		case DIOCGETADDRS:
1184 		case DIOCGETADDR:
1185 		case DIOCGETSTATE:
1186 		case DIOCSETSTATUSIF:
1187 		case DIOCGETSTATUS:
1188 		case DIOCCLRSTATUS:
1189 		case DIOCNATLOOK:
1190 		case DIOCSETDEBUG:
1191 		case DIOCGETSTATES:
1192 		case DIOCGETTIMEOUT:
1193 		case DIOCCLRRULECTRS:
1194 		case DIOCGETLIMIT:
1195 		case DIOCGETALTQS:
1196 		case DIOCGETALTQ:
1197 		case DIOCGETQSTATS:
1198 		case DIOCGETRULESETS:
1199 		case DIOCGETRULESET:
1200 		case DIOCRGETTABLES:
1201 		case DIOCRGETTSTATS:
1202 		case DIOCRCLRTSTATS:
1203 		case DIOCRCLRADDRS:
1204 		case DIOCRADDADDRS:
1205 		case DIOCRDELADDRS:
1206 		case DIOCRSETADDRS:
1207 		case DIOCRGETADDRS:
1208 		case DIOCRGETASTATS:
1209 		case DIOCRCLRASTATS:
1210 		case DIOCRTSTADDRS:
1211 		case DIOCOSFPGET:
1212 		case DIOCGETSRCNODES:
1213 		case DIOCCLRSRCNODES:
1214 		case DIOCIGETIFACES:
1215 		case DIOCSETIFFLAG:
1216 		case DIOCCLRIFFLAG:
1217 			break;
1218 		case DIOCRCLRTABLES:
1219 		case DIOCRADDTABLES:
1220 		case DIOCRDELTABLES:
1221 		case DIOCRSETTFLAGS:
1222 			if (((struct pfioc_table *)addr)->pfrio_flags &
1223 			    PFR_FLAG_DUMMY)
1224 				break; /* dummy operation ok */
1225 			return (EPERM);
1226 		default:
1227 			return (EPERM);
1228 		}
1229 
1230 	if (!(flags & FWRITE))
1231 		switch (cmd) {
1232 		case DIOCGETRULES:
1233 		case DIOCGETADDRS:
1234 		case DIOCGETADDR:
1235 		case DIOCGETSTATE:
1236 		case DIOCGETSTATUS:
1237 		case DIOCGETSTATES:
1238 		case DIOCGETTIMEOUT:
1239 		case DIOCGETLIMIT:
1240 		case DIOCGETALTQS:
1241 		case DIOCGETALTQ:
1242 		case DIOCGETQSTATS:
1243 		case DIOCGETRULESETS:
1244 		case DIOCGETRULESET:
1245 		case DIOCNATLOOK:
1246 		case DIOCRGETTABLES:
1247 		case DIOCRGETTSTATS:
1248 		case DIOCRGETADDRS:
1249 		case DIOCRGETASTATS:
1250 		case DIOCRTSTADDRS:
1251 		case DIOCOSFPGET:
1252 		case DIOCGETSRCNODES:
1253 		case DIOCIGETIFACES:
1254 			break;
1255 		case DIOCRCLRTABLES:
1256 		case DIOCRADDTABLES:
1257 		case DIOCRDELTABLES:
1258 		case DIOCRCLRTSTATS:
1259 		case DIOCRCLRADDRS:
1260 		case DIOCRADDADDRS:
1261 		case DIOCRDELADDRS:
1262 		case DIOCRSETADDRS:
1263 		case DIOCRSETTFLAGS:
1264 			if (((struct pfioc_table *)addr)->pfrio_flags &
1265 			    PFR_FLAG_DUMMY) {
1266 				flags |= FWRITE; /* need write lock for dummy */
1267 				break; /* dummy operation ok */
1268 			}
1269 			return (EACCES);
1270 		case DIOCGETRULE:
1271 			if (((struct pfioc_rule *)addr)->action == PF_GET_CLR_CNTR)
1272 				return (EACCES);
1273 			break;
1274 		default:
1275 			return (EACCES);
1276 		}
1277 
1278 	if (flags & FWRITE)
1279 		rw_enter_write(&pf_consistency_lock);
1280 	else
1281 		rw_enter_read(&pf_consistency_lock);
1282 
1283 	s = splsoftnet();
1284 	switch (cmd) {
1285 
1286 	case DIOCSTART:
1287 		if (pf_status.running)
1288 			error = EEXIST;
1289 		else {
1290 #ifdef __NetBSD__
1291 			error = pf_pfil_attach();
1292 			if (error)
1293 				break;
1294 #endif /* __NetBSD__ */
1295 			pf_status.running = 1;
1296 			pf_status.since = time_second;
1297 			if (pf_status.stateid == 0) {
1298 				pf_status.stateid = time_second;
1299 				pf_status.stateid = pf_status.stateid << 32;
1300 			}
1301 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1302 		}
1303 		break;
1304 
1305 	case DIOCSTOP:
1306 		if (!pf_status.running)
1307 			error = ENOENT;
1308 		else {
1309 #ifdef __NetBSD__
1310 			error = pf_pfil_detach();
1311 			if (error)
1312 				break;
1313 #endif /* __NetBSD__ */
1314 			pf_status.running = 0;
1315 			pf_status.since = time_second;
1316 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1317 		}
1318 		break;
1319 
1320 	case DIOCADDRULE: {
1321 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1322 		struct pf_ruleset	*ruleset;
1323 		struct pf_rule		*rule, *tail;
1324 		struct pf_pooladdr	*pa;
1325 		int			 rs_num;
1326 
1327 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1328 		ruleset = pf_find_ruleset(pr->anchor);
1329 		if (ruleset == NULL) {
1330 			error = EINVAL;
1331 			break;
1332 		}
1333 		rs_num = pf_get_ruleset_number(pr->rule.action);
1334 		if (rs_num >= PF_RULESET_MAX) {
1335 			error = EINVAL;
1336 			break;
1337 		}
1338 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1339 			error = EINVAL;
1340 			break;
1341 		}
1342 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1343 			error = EBUSY;
1344 			break;
1345 		}
1346 		if (pr->pool_ticket != ticket_pabuf) {
1347 			error = EBUSY;
1348 			break;
1349 		}
1350 		rule = pool_get(&pf_rule_pl, PR_NOWAIT);
1351 		if (rule == NULL) {
1352 			error = ENOMEM;
1353 			break;
1354 		}
1355 		bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1356 #ifdef __NetBSD__
1357 		rule->cuid = kauth_cred_getuid(l->l_cred);
1358 		rule->cpid = l->l_proc->p_pid;
1359 #else
1360 		rule->cuid = p->p_cred->p_ruid;
1361 		rule->cpid = p->p_pid;
1362 #endif /* !__NetBSD__ */
1363 		rule->anchor = NULL;
1364 		rule->kif = NULL;
1365 		TAILQ_INIT(&rule->rpool.list);
1366 		/* initialize refcounting */
1367 		rule->states = 0;
1368 		rule->src_nodes = 0;
1369 		rule->entries.tqe_prev = NULL;
1370 #ifndef INET
1371 		if (rule->af == AF_INET) {
1372 			pool_put(&pf_rule_pl, rule);
1373 			error = EAFNOSUPPORT;
1374 			break;
1375 		}
1376 #endif /* INET */
1377 #ifndef INET6
1378 		if (rule->af == AF_INET6) {
1379 			pool_put(&pf_rule_pl, rule);
1380 			error = EAFNOSUPPORT;
1381 			break;
1382 		}
1383 #endif /* INET6 */
1384 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1385 		    pf_rulequeue);
1386 		if (tail)
1387 			rule->nr = tail->nr + 1;
1388 		else
1389 			rule->nr = 0;
1390 		if (rule->ifname[0]) {
1391 			rule->kif = pfi_kif_get(rule->ifname);
1392 			if (rule->kif == NULL) {
1393 				pool_put(&pf_rule_pl, rule);
1394 				error = EINVAL;
1395 				break;
1396 			}
1397 			pfi_kif_ref(rule->kif, PFI_KIF_REF_RULE);
1398 		}
1399 
1400 #ifndef __NetBSD__
1401 		if (rule->rtableid > 0 && !rtable_exists(rule->rtableid))
1402 			error = EBUSY;
1403 #endif /* !__NetBSD__ */
1404 
1405 #ifdef ALTQ
1406 		/* set queue IDs */
1407 		if (rule->qname[0] != 0) {
1408 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1409 				error = EBUSY;
1410 			else if (rule->pqname[0] != 0) {
1411 				if ((rule->pqid =
1412 				    pf_qname2qid(rule->pqname)) == 0)
1413 					error = EBUSY;
1414 			} else
1415 				rule->pqid = rule->qid;
1416 		}
1417 #endif
1418 		if (rule->tagname[0])
1419 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1420 				error = EBUSY;
1421 		if (rule->match_tagname[0])
1422 			if ((rule->match_tag =
1423 			    pf_tagname2tag(rule->match_tagname)) == 0)
1424 				error = EBUSY;
1425 		if (rule->rt && !rule->direction)
1426 			error = EINVAL;
1427 #if NPFLOG > 0
1428 		if (!rule->log)
1429 			rule->logif = 0;
1430 		if (rule->logif >= PFLOGIFS_MAX)
1431 			error = EINVAL;
1432 #endif
1433 		if (pf_rtlabel_add(&rule->src.addr) ||
1434 		    pf_rtlabel_add(&rule->dst.addr))
1435 			error = EBUSY;
1436 		if (pfi_dynaddr_setup(&rule->src.addr, rule->af))
1437 			error = EINVAL;
1438 		if (pfi_dynaddr_setup(&rule->dst.addr, rule->af))
1439 			error = EINVAL;
1440 		if (pf_tbladdr_setup(ruleset, &rule->src.addr))
1441 			error = EINVAL;
1442 		if (pf_tbladdr_setup(ruleset, &rule->dst.addr))
1443 			error = EINVAL;
1444 		if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1445 			error = EINVAL;
1446 		TAILQ_FOREACH(pa, &pf_pabuf, entries)
1447 			if (pf_tbladdr_setup(ruleset, &pa->addr))
1448 				error = EINVAL;
1449 
1450 		if (rule->overload_tblname[0]) {
1451 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1452 			    rule->overload_tblname)) == NULL)
1453 				error = EINVAL;
1454 			else
1455 				rule->overload_tbl->pfrkt_flags |=
1456 				    PFR_TFLAG_ACTIVE;
1457 		}
1458 
1459 		pf_mv_pool(&pf_pabuf, &rule->rpool.list);
1460 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1461 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1462 		    (rule->rt > PF_FASTROUTE)) &&
1463 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1464 			error = EINVAL;
1465 
1466 		if (error) {
1467 			pf_rm_rule(NULL, rule);
1468 			break;
1469 		}
1470 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1471 		rule->evaluations = rule->packets[0] = rule->packets[1] =
1472 		    rule->bytes[0] = rule->bytes[1] = 0;
1473 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1474 		    rule, entries);
1475 		ruleset->rules[rs_num].inactive.rcount++;
1476 		break;
1477 	}
1478 
1479 	case DIOCGETRULES: {
1480 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1481 		struct pf_ruleset	*ruleset;
1482 		struct pf_rule		*tail;
1483 		int			 rs_num;
1484 
1485 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1486 		ruleset = pf_find_ruleset(pr->anchor);
1487 		if (ruleset == NULL) {
1488 			error = EINVAL;
1489 			break;
1490 		}
1491 		rs_num = pf_get_ruleset_number(pr->rule.action);
1492 		if (rs_num >= PF_RULESET_MAX) {
1493 			error = EINVAL;
1494 			break;
1495 		}
1496 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1497 		    pf_rulequeue);
1498 		if (tail)
1499 			pr->nr = tail->nr + 1;
1500 		else
1501 			pr->nr = 0;
1502 		pr->ticket = ruleset->rules[rs_num].active.ticket;
1503 		break;
1504 	}
1505 
1506 	case DIOCGETRULE: {
1507 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1508 		struct pf_ruleset	*ruleset;
1509 		struct pf_rule		*rule;
1510 		int			 rs_num, i;
1511 
1512 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1513 		ruleset = pf_find_ruleset(pr->anchor);
1514 		if (ruleset == NULL) {
1515 			error = EINVAL;
1516 			break;
1517 		}
1518 		rs_num = pf_get_ruleset_number(pr->rule.action);
1519 		if (rs_num >= PF_RULESET_MAX) {
1520 			error = EINVAL;
1521 			break;
1522 		}
1523 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1524 			error = EBUSY;
1525 			break;
1526 		}
1527 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1528 		while ((rule != NULL) && (rule->nr != pr->nr))
1529 			rule = TAILQ_NEXT(rule, entries);
1530 		if (rule == NULL) {
1531 			error = EBUSY;
1532 			break;
1533 		}
1534 		bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1535 		if (pf_anchor_copyout(ruleset, rule, pr)) {
1536 			error = EBUSY;
1537 			break;
1538 		}
1539 		pfi_dynaddr_copyout(&pr->rule.src.addr);
1540 		pfi_dynaddr_copyout(&pr->rule.dst.addr);
1541 		pf_tbladdr_copyout(&pr->rule.src.addr);
1542 		pf_tbladdr_copyout(&pr->rule.dst.addr);
1543 		pf_rtlabel_copyout(&pr->rule.src.addr);
1544 		pf_rtlabel_copyout(&pr->rule.dst.addr);
1545 		for (i = 0; i < PF_SKIP_COUNT; ++i)
1546 			if (rule->skip[i].ptr == NULL)
1547 				pr->rule.skip[i].nr = -1;
1548 			else
1549 				pr->rule.skip[i].nr =
1550 				    rule->skip[i].ptr->nr;
1551 
1552 		if (pr->action == PF_GET_CLR_CNTR) {
1553 			rule->evaluations = 0;
1554 			rule->packets[0] = rule->packets[1] = 0;
1555 			rule->bytes[0] = rule->bytes[1] = 0;
1556 		}
1557 		break;
1558 	}
1559 
1560 	case DIOCCHANGERULE: {
1561 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1562 		struct pf_ruleset	*ruleset;
1563 		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1564 		u_int32_t		 nr = 0;
1565 		int			 rs_num;
1566 
1567 		if (!(pcr->action == PF_CHANGE_REMOVE ||
1568 		    pcr->action == PF_CHANGE_GET_TICKET) &&
1569 		    pcr->pool_ticket != ticket_pabuf) {
1570 			error = EBUSY;
1571 			break;
1572 		}
1573 
1574 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1575 		    pcr->action > PF_CHANGE_GET_TICKET) {
1576 			error = EINVAL;
1577 			break;
1578 		}
1579 		ruleset = pf_find_ruleset(pcr->anchor);
1580 		if (ruleset == NULL) {
1581 			error = EINVAL;
1582 			break;
1583 		}
1584 		rs_num = pf_get_ruleset_number(pcr->rule.action);
1585 		if (rs_num >= PF_RULESET_MAX) {
1586 			error = EINVAL;
1587 			break;
1588 		}
1589 
1590 		if (pcr->action == PF_CHANGE_GET_TICKET) {
1591 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1592 			break;
1593 		} else {
1594 			if (pcr->ticket !=
1595 			    ruleset->rules[rs_num].active.ticket) {
1596 				error = EINVAL;
1597 				break;
1598 			}
1599 			if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1600 				error = EINVAL;
1601 				break;
1602 			}
1603 		}
1604 
1605 		if (pcr->action != PF_CHANGE_REMOVE) {
1606 			newrule = pool_get(&pf_rule_pl, PR_NOWAIT);
1607 			if (newrule == NULL) {
1608 				error = ENOMEM;
1609 				break;
1610 			}
1611 			bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1612 #ifdef __NetBSD__
1613 			newrule->cuid = kauth_cred_getuid(l->l_cred);
1614 			newrule->cpid = l->l_proc->p_pid;
1615 #else
1616 			newrule->cuid = p->p_cred->p_ruid;
1617 			newrule->cpid = p->p_pid;
1618 #endif /* !__NetBSD__ */
1619 			TAILQ_INIT(&newrule->rpool.list);
1620 			/* initialize refcounting */
1621 			newrule->states = 0;
1622 			newrule->entries.tqe_prev = NULL;
1623 #ifndef INET
1624 			if (newrule->af == AF_INET) {
1625 				pool_put(&pf_rule_pl, newrule);
1626 				error = EAFNOSUPPORT;
1627 				break;
1628 			}
1629 #endif /* INET */
1630 #ifndef INET6
1631 			if (newrule->af == AF_INET6) {
1632 				pool_put(&pf_rule_pl, newrule);
1633 				error = EAFNOSUPPORT;
1634 				break;
1635 			}
1636 #endif /* INET6 */
1637 			if (newrule->ifname[0]) {
1638 				newrule->kif = pfi_kif_get(newrule->ifname);
1639 				if (newrule->kif == NULL) {
1640 					pool_put(&pf_rule_pl, newrule);
1641 					error = EINVAL;
1642 					break;
1643 				}
1644 				pfi_kif_ref(newrule->kif, PFI_KIF_REF_RULE);
1645 			} else
1646 				newrule->kif = NULL;
1647 
1648 #ifndef __NetBSD__
1649 			if (newrule->rtableid > 0 &&
1650 			    !rtable_exists(newrule->rtableid))
1651 				error = EBUSY;
1652 #endif /* !__NetBSD__ */
1653 
1654 #ifdef ALTQ
1655 			/* set queue IDs */
1656 			if (newrule->qname[0] != 0) {
1657 				if ((newrule->qid =
1658 				    pf_qname2qid(newrule->qname)) == 0)
1659 					error = EBUSY;
1660 				else if (newrule->pqname[0] != 0) {
1661 					if ((newrule->pqid =
1662 					    pf_qname2qid(newrule->pqname)) == 0)
1663 						error = EBUSY;
1664 				} else
1665 					newrule->pqid = newrule->qid;
1666 			}
1667 #endif /* ALTQ */
1668 			if (newrule->tagname[0])
1669 				if ((newrule->tag =
1670 				    pf_tagname2tag(newrule->tagname)) == 0)
1671 					error = EBUSY;
1672 			if (newrule->match_tagname[0])
1673 				if ((newrule->match_tag = pf_tagname2tag(
1674 				    newrule->match_tagname)) == 0)
1675 					error = EBUSY;
1676 			if (newrule->rt && !newrule->direction)
1677 				error = EINVAL;
1678 #if NPFLOG > 0
1679 			if (!newrule->log)
1680 				newrule->logif = 0;
1681 			if (newrule->logif >= PFLOGIFS_MAX)
1682 				error = EINVAL;
1683 #endif
1684 			if (pf_rtlabel_add(&newrule->src.addr) ||
1685 			    pf_rtlabel_add(&newrule->dst.addr))
1686 				error = EBUSY;
1687 			if (pfi_dynaddr_setup(&newrule->src.addr, newrule->af))
1688 				error = EINVAL;
1689 			if (pfi_dynaddr_setup(&newrule->dst.addr, newrule->af))
1690 				error = EINVAL;
1691 			if (pf_tbladdr_setup(ruleset, &newrule->src.addr))
1692 				error = EINVAL;
1693 			if (pf_tbladdr_setup(ruleset, &newrule->dst.addr))
1694 				error = EINVAL;
1695 			if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1696 				error = EINVAL;
1697 			TAILQ_FOREACH(pa, &pf_pabuf, entries)
1698 				if (pf_tbladdr_setup(ruleset, &pa->addr))
1699 					error = EINVAL;
1700 
1701 			if (newrule->overload_tblname[0]) {
1702 				if ((newrule->overload_tbl = pfr_attach_table(
1703 				    ruleset, newrule->overload_tblname)) ==
1704 				    NULL)
1705 					error = EINVAL;
1706 				else
1707 					newrule->overload_tbl->pfrkt_flags |=
1708 					    PFR_TFLAG_ACTIVE;
1709 			}
1710 
1711 			pf_mv_pool(&pf_pabuf, &newrule->rpool.list);
1712 			if (((((newrule->action == PF_NAT) ||
1713 			    (newrule->action == PF_RDR) ||
1714 			    (newrule->action == PF_BINAT) ||
1715 			    (newrule->rt > PF_FASTROUTE)) &&
1716 			    !newrule->anchor)) &&
1717 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1718 				error = EINVAL;
1719 
1720 			if (error) {
1721 				pf_rm_rule(NULL, newrule);
1722 				break;
1723 			}
1724 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1725 			newrule->evaluations = 0;
1726 			newrule->packets[0] = newrule->packets[1] = 0;
1727 			newrule->bytes[0] = newrule->bytes[1] = 0;
1728 		}
1729 		pf_empty_pool(&pf_pabuf);
1730 
1731 		if (pcr->action == PF_CHANGE_ADD_HEAD)
1732 			oldrule = TAILQ_FIRST(
1733 			    ruleset->rules[rs_num].active.ptr);
1734 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1735 			oldrule = TAILQ_LAST(
1736 			    ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1737 		else {
1738 			oldrule = TAILQ_FIRST(
1739 			    ruleset->rules[rs_num].active.ptr);
1740 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1741 				oldrule = TAILQ_NEXT(oldrule, entries);
1742 			if (oldrule == NULL) {
1743 				if (newrule != NULL)
1744 					pf_rm_rule(NULL, newrule);
1745 				error = EINVAL;
1746 				break;
1747 			}
1748 		}
1749 
1750 		if (pcr->action == PF_CHANGE_REMOVE) {
1751 			pf_rm_rule(ruleset->rules[rs_num].active.ptr, oldrule);
1752 			ruleset->rules[rs_num].active.rcount--;
1753 		} else {
1754 			if (oldrule == NULL)
1755 				TAILQ_INSERT_TAIL(
1756 				    ruleset->rules[rs_num].active.ptr,
1757 				    newrule, entries);
1758 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1759 			    pcr->action == PF_CHANGE_ADD_BEFORE)
1760 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1761 			else
1762 				TAILQ_INSERT_AFTER(
1763 				    ruleset->rules[rs_num].active.ptr,
1764 				    oldrule, newrule, entries);
1765 			ruleset->rules[rs_num].active.rcount++;
1766 		}
1767 
1768 		nr = 0;
1769 		TAILQ_FOREACH(oldrule,
1770 		    ruleset->rules[rs_num].active.ptr, entries)
1771 			oldrule->nr = nr++;
1772 
1773 		ruleset->rules[rs_num].active.ticket++;
1774 
1775 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1776 		pf_remove_if_empty_ruleset(ruleset);
1777 
1778 		break;
1779 	}
1780 
1781 	case DIOCCLRSTATES: {
1782 		struct pf_state		*s, *nexts;
1783 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1784 		int			 killed = 0;
1785 
1786 		for (s = RB_MIN(pf_state_tree_id, &tree_id); s; s = nexts) {
1787 			nexts = RB_NEXT(pf_state_tree_id, &tree_id, s);
1788 
1789 			if (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1790 			    s->kif->pfik_name)) {
1791 #if NPFSYNC
1792 				/* don't send out individual delete messages */
1793 				s->sync_flags = PFSTATE_NOSYNC;
1794 #endif
1795 				pf_unlink_state(s);
1796 				killed++;
1797 			}
1798 		}
1799 		psk->psk_af = killed;
1800 #if NPFSYNC
1801 		pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
1802 #endif
1803 		break;
1804 	}
1805 
1806 	case DIOCKILLSTATES: {
1807 		struct pf_state		*s, *nexts;
1808 		struct pf_state_key	*sk;
1809 		struct pf_state_host	*src, *dst;
1810 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
1811 		int			 killed = 0;
1812 
1813 		for (s = RB_MIN(pf_state_tree_id, &tree_id); s;
1814 		    s = nexts) {
1815 			nexts = RB_NEXT(pf_state_tree_id, &tree_id, s);
1816 			sk = s->state_key;
1817 
1818 			if (sk->direction == PF_OUT) {
1819 				src = &sk->lan;
1820 				dst = &sk->ext;
1821 			} else {
1822 				src = &sk->ext;
1823 				dst = &sk->lan;
1824 			}
1825 			if ((!psk->psk_af || sk->af == psk->psk_af)
1826 			    && (!psk->psk_proto || psk->psk_proto ==
1827 			    sk->proto) &&
1828 			    PF_MATCHA(psk->psk_src.neg,
1829 			    &psk->psk_src.addr.v.a.addr,
1830 			    &psk->psk_src.addr.v.a.mask,
1831 			    &src->addr, sk->af) &&
1832 			    PF_MATCHA(psk->psk_dst.neg,
1833 			    &psk->psk_dst.addr.v.a.addr,
1834 			    &psk->psk_dst.addr.v.a.mask,
1835 			    &dst->addr, sk->af) &&
1836 			    (psk->psk_src.port_op == 0 ||
1837 			    pf_match_port(psk->psk_src.port_op,
1838 			    psk->psk_src.port[0], psk->psk_src.port[1],
1839 			    src->port)) &&
1840 			    (psk->psk_dst.port_op == 0 ||
1841 			    pf_match_port(psk->psk_dst.port_op,
1842 			    psk->psk_dst.port[0], psk->psk_dst.port[1],
1843 			    dst->port)) &&
1844 			    (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1845 			    s->kif->pfik_name))) {
1846 #if NPFSYNC > 0
1847 				/* send immediate delete of state */
1848 				pfsync_delete_state(s);
1849 				s->sync_flags |= PFSTATE_NOSYNC;
1850 #endif
1851 				pf_unlink_state(s);
1852 				killed++;
1853 			}
1854 		}
1855 		psk->psk_af = killed;
1856 		break;
1857 	}
1858 
1859 	case DIOCADDSTATE: {
1860 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1861 		struct pfsync_state 	*sp = (struct pfsync_state *)ps->state;
1862 		struct pf_state		*s;
1863 		struct pf_state_key	*sk;
1864 		struct pfi_kif		*kif;
1865 
1866 		if (sp->timeout >= PFTM_MAX &&
1867 		    sp->timeout != PFTM_UNTIL_PACKET) {
1868 			error = EINVAL;
1869 			break;
1870 		}
1871 		s = pool_get(&pf_state_pl, PR_NOWAIT);
1872 		if (s == NULL) {
1873 			error = ENOMEM;
1874 			break;
1875 		}
1876 		bzero(s, sizeof(struct pf_state));
1877 		if ((sk = pf_alloc_state_key(s)) == NULL) {
1878 			error = ENOMEM;
1879 			break;
1880 		}
1881 		pf_state_import(sp, sk, s);
1882 		kif = pfi_kif_get(sp->ifname);
1883 		if (kif == NULL) {
1884 			pool_put(&pf_state_pl, s);
1885 			pool_put(&pf_state_key_pl, sk);
1886 			error = ENOENT;
1887 			break;
1888 		}
1889 		if (pf_insert_state(kif, s)) {
1890 			pfi_kif_unref(kif, PFI_KIF_REF_NONE);
1891 			pool_put(&pf_state_pl, s);
1892 			pool_put(&pf_state_key_pl, sk);
1893 			error = ENOMEM;
1894 		}
1895 		break;
1896 	}
1897 
1898 	case DIOCGETSTATE: {
1899 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1900 		struct pf_state		*s;
1901 		u_int32_t		 nr;
1902 
1903 		nr = 0;
1904 		RB_FOREACH(s, pf_state_tree_id, &tree_id) {
1905 			if (nr >= ps->nr)
1906 				break;
1907 			nr++;
1908 		}
1909 		if (s == NULL) {
1910 			error = EBUSY;
1911 			break;
1912 		}
1913 
1914 		pf_state_export((struct pfsync_state *)&ps->state,
1915 		    s->state_key, s);
1916 		break;
1917 	}
1918 
1919 	case DIOCGETSTATES: {
1920 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
1921 		struct pf_state		*state;
1922 		struct pfsync_state	*p, *pstore;
1923 		u_int32_t		 nr = 0;
1924 
1925 		if (ps->ps_len == 0) {
1926 			nr = pf_status.states;
1927 			ps->ps_len = sizeof(struct pfsync_state) * nr;
1928 			break;
1929 		}
1930 
1931 		pstore = malloc(sizeof(*pstore), M_TEMP, M_WAITOK);
1932 
1933 		p = ps->ps_states;
1934 
1935 		state = TAILQ_FIRST(&state_list);
1936 		while (state) {
1937 			if (state->timeout != PFTM_UNLINKED) {
1938 				if ((nr+1) * sizeof(*p) > (unsigned)ps->ps_len)
1939 					break;
1940 
1941 				pf_state_export(pstore,
1942 				    state->state_key, state);
1943 				error = copyout(pstore, p, sizeof(*p));
1944 				if (error) {
1945 					free(pstore, M_TEMP);
1946 					goto fail;
1947 				}
1948 				p++;
1949 				nr++;
1950 			}
1951 			state = TAILQ_NEXT(state, entry_list);
1952 		}
1953 
1954 		ps->ps_len = sizeof(struct pfsync_state) * nr;
1955 
1956 		free(pstore, M_TEMP);
1957 		break;
1958 	}
1959 
1960 	case DIOCGETSTATUS: {
1961 		struct pf_status *s = (struct pf_status *)addr;
1962 		bcopy(&pf_status, s, sizeof(struct pf_status));
1963 		pfi_fill_oldstatus(s);
1964 		break;
1965 	}
1966 
1967 	case DIOCSETSTATUSIF: {
1968 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
1969 
1970 		if (pi->ifname[0] == 0) {
1971 			bzero(pf_status.ifname, IFNAMSIZ);
1972 			break;
1973 		}
1974 		if (ifunit(pi->ifname) == NULL) {
1975 			error = EINVAL;
1976 			break;
1977 		}
1978 		strlcpy(pf_status.ifname, pi->ifname, IFNAMSIZ);
1979 		break;
1980 	}
1981 
1982 	case DIOCCLRSTATUS: {
1983 		bzero(pf_status.counters, sizeof(pf_status.counters));
1984 		bzero(pf_status.fcounters, sizeof(pf_status.fcounters));
1985 		bzero(pf_status.scounters, sizeof(pf_status.scounters));
1986 		pf_status.since = time_second;
1987 		if (*pf_status.ifname)
1988 			pfi_clr_istats(pf_status.ifname);
1989 		break;
1990 	}
1991 
1992 	case DIOCNATLOOK: {
1993 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
1994 		struct pf_state_key	*sk;
1995 		struct pf_state		*state;
1996 		struct pf_state_key_cmp	 key;
1997 		int			 m = 0, direction = pnl->direction;
1998 
1999 		key.af = pnl->af;
2000 		key.proto = pnl->proto;
2001 
2002 		if (!pnl->proto ||
2003 		    PF_AZERO(&pnl->saddr, pnl->af) ||
2004 		    PF_AZERO(&pnl->daddr, pnl->af) ||
2005 		    ((pnl->proto == IPPROTO_TCP ||
2006 		    pnl->proto == IPPROTO_UDP) &&
2007 		    (!pnl->dport || !pnl->sport)))
2008 			error = EINVAL;
2009 		else {
2010 			/*
2011 			 * userland gives us source and dest of connection,
2012 			 * reverse the lookup so we ask for what happens with
2013 			 * the return traffic, enabling us to find it in the
2014 			 * state tree.
2015 			 */
2016 			if (direction == PF_IN) {
2017 				PF_ACPY(&key.ext.addr, &pnl->daddr, pnl->af);
2018 				key.ext.port = pnl->dport;
2019 				PF_ACPY(&key.gwy.addr, &pnl->saddr, pnl->af);
2020 				key.gwy.port = pnl->sport;
2021 				state = pf_find_state_all(&key, PF_EXT_GWY, &m);
2022 			} else {
2023 				PF_ACPY(&key.lan.addr, &pnl->daddr, pnl->af);
2024 				key.lan.port = pnl->dport;
2025 				PF_ACPY(&key.ext.addr, &pnl->saddr, pnl->af);
2026 				key.ext.port = pnl->sport;
2027 				state = pf_find_state_all(&key, PF_LAN_EXT, &m);
2028 			}
2029 			if (m > 1)
2030 				error = E2BIG;	/* more than one state */
2031 			else if (state != NULL) {
2032 				sk = state->state_key;
2033 				if (direction == PF_IN) {
2034 					PF_ACPY(&pnl->rsaddr, &sk->lan.addr,
2035 					    sk->af);
2036 					pnl->rsport = sk->lan.port;
2037 					PF_ACPY(&pnl->rdaddr, &pnl->daddr,
2038 					    pnl->af);
2039 					pnl->rdport = pnl->dport;
2040 				} else {
2041 					PF_ACPY(&pnl->rdaddr, &sk->gwy.addr,
2042 					    sk->af);
2043 					pnl->rdport = sk->gwy.port;
2044 					PF_ACPY(&pnl->rsaddr, &pnl->saddr,
2045 					    pnl->af);
2046 					pnl->rsport = pnl->sport;
2047 				}
2048 			} else
2049 				error = ENOENT;
2050 		}
2051 		break;
2052 	}
2053 
2054 	case DIOCSETTIMEOUT: {
2055 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2056 		int		 old;
2057 
2058 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
2059 		    pt->seconds < 0) {
2060 			error = EINVAL;
2061 			goto fail;
2062 		}
2063 		old = pf_default_rule.timeout[pt->timeout];
2064 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
2065 			pt->seconds = 1;
2066 		pf_default_rule.timeout[pt->timeout] = pt->seconds;
2067 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
2068 			wakeup(pf_purge_thread);
2069 		pt->seconds = old;
2070 		break;
2071 	}
2072 
2073 	case DIOCGETTIMEOUT: {
2074 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2075 
2076 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
2077 			error = EINVAL;
2078 			goto fail;
2079 		}
2080 		pt->seconds = pf_default_rule.timeout[pt->timeout];
2081 		break;
2082 	}
2083 
2084 	case DIOCGETLIMIT: {
2085 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2086 
2087 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
2088 			error = EINVAL;
2089 			goto fail;
2090 		}
2091 		pl->limit = pf_pool_limits[pl->index].limit;
2092 		break;
2093 	}
2094 
2095 	case DIOCSETLIMIT: {
2096 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2097 		int			 old_limit;
2098 
2099 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
2100 		    pf_pool_limits[pl->index].pp == NULL) {
2101 			error = EINVAL;
2102 			goto fail;
2103 		}
2104 #ifdef __NetBSD__
2105 		pool_sethardlimit(pf_pool_limits[pl->index].pp,
2106 		    pl->limit, NULL, 0);
2107 #else
2108 		if (pool_sethardlimit(pf_pool_limits[pl->index].pp,
2109 		    pl->limit, NULL, 0) != 0) {
2110 			error = EBUSY;
2111 			goto fail;
2112 		}
2113 #endif /* !__NetBSD__ */
2114 		old_limit = pf_pool_limits[pl->index].limit;
2115 		pf_pool_limits[pl->index].limit = pl->limit;
2116 		pl->limit = old_limit;
2117 		break;
2118 	}
2119 
2120 	case DIOCSETDEBUG: {
2121 		u_int32_t	*level = (u_int32_t *)addr;
2122 
2123 		pf_status.debug = *level;
2124 		break;
2125 	}
2126 
2127 	case DIOCCLRRULECTRS: {
2128 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
2129 		struct pf_ruleset	*ruleset = &pf_main_ruleset;
2130 		struct pf_rule		*rule;
2131 
2132 		TAILQ_FOREACH(rule,
2133 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
2134 			rule->evaluations = 0;
2135 			rule->packets[0] = rule->packets[1] = 0;
2136 			rule->bytes[0] = rule->bytes[1] = 0;
2137 		}
2138 		break;
2139 	}
2140 
2141 #ifdef ALTQ
2142 	case DIOCSTARTALTQ: {
2143 		struct pf_altq		*altq;
2144 
2145 		/* enable all altq interfaces on active list */
2146 		TAILQ_FOREACH(altq, pf_altqs_active, entries) {
2147 			if (altq->qname[0] == 0) {
2148 				error = pf_enable_altq(altq);
2149 				if (error != 0)
2150 					break;
2151 			}
2152 		}
2153 		if (error == 0)
2154 			pf_altq_running = 1;
2155 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
2156 		break;
2157 	}
2158 
2159 	case DIOCSTOPALTQ: {
2160 		struct pf_altq		*altq;
2161 
2162 		/* disable all altq interfaces on active list */
2163 		TAILQ_FOREACH(altq, pf_altqs_active, entries) {
2164 			if (altq->qname[0] == 0) {
2165 				error = pf_disable_altq(altq);
2166 				if (error != 0)
2167 					break;
2168 			}
2169 		}
2170 		if (error == 0)
2171 			pf_altq_running = 0;
2172 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2173 		break;
2174 	}
2175 
2176 	case DIOCADDALTQ: {
2177 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2178 		struct pf_altq		*altq, *a;
2179 
2180 		if (pa->ticket != ticket_altqs_inactive) {
2181 			error = EBUSY;
2182 			break;
2183 		}
2184 		altq = pool_get(&pf_altq_pl, PR_NOWAIT);
2185 		if (altq == NULL) {
2186 			error = ENOMEM;
2187 			break;
2188 		}
2189 		bcopy(&pa->altq, altq, sizeof(struct pf_altq));
2190 
2191 		/*
2192 		 * if this is for a queue, find the discipline and
2193 		 * copy the necessary fields
2194 		 */
2195 		if (altq->qname[0] != 0) {
2196 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2197 				error = EBUSY;
2198 				pool_put(&pf_altq_pl, altq);
2199 				break;
2200 			}
2201 			TAILQ_FOREACH(a, pf_altqs_inactive, entries) {
2202 				if (strncmp(a->ifname, altq->ifname,
2203 				    IFNAMSIZ) == 0 && a->qname[0] == 0) {
2204 					altq->altq_disc = a->altq_disc;
2205 					break;
2206 				}
2207 			}
2208 		}
2209 
2210 		error = altq_add(altq);
2211 		if (error) {
2212 			pool_put(&pf_altq_pl, altq);
2213 			break;
2214 		}
2215 
2216 		TAILQ_INSERT_TAIL(pf_altqs_inactive, altq, entries);
2217 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2218 		break;
2219 	}
2220 
2221 	case DIOCGETALTQS: {
2222 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2223 		struct pf_altq		*altq;
2224 
2225 		pa->nr = 0;
2226 		TAILQ_FOREACH(altq, pf_altqs_active, entries)
2227 			pa->nr++;
2228 		pa->ticket = ticket_altqs_active;
2229 		break;
2230 	}
2231 
2232 	case DIOCGETALTQ: {
2233 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2234 		struct pf_altq		*altq;
2235 		u_int32_t		 nr;
2236 
2237 		if (pa->ticket != ticket_altqs_active) {
2238 			error = EBUSY;
2239 			break;
2240 		}
2241 		nr = 0;
2242 		altq = TAILQ_FIRST(pf_altqs_active);
2243 		while ((altq != NULL) && (nr < pa->nr)) {
2244 			altq = TAILQ_NEXT(altq, entries);
2245 			nr++;
2246 		}
2247 		if (altq == NULL) {
2248 			error = EBUSY;
2249 			break;
2250 		}
2251 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2252 		break;
2253 	}
2254 
2255 	case DIOCCHANGEALTQ:
2256 		/* CHANGEALTQ not supported yet! */
2257 		error = ENODEV;
2258 		break;
2259 
2260 	case DIOCGETQSTATS: {
2261 		struct pfioc_qstats	*pq = (struct pfioc_qstats *)addr;
2262 		struct pf_altq		*altq;
2263 		u_int32_t		 nr;
2264 		int			 nbytes;
2265 
2266 		if (pq->ticket != ticket_altqs_active) {
2267 			error = EBUSY;
2268 			break;
2269 		}
2270 		nbytes = pq->nbytes;
2271 		nr = 0;
2272 		altq = TAILQ_FIRST(pf_altqs_active);
2273 		while ((altq != NULL) && (nr < pq->nr)) {
2274 			altq = TAILQ_NEXT(altq, entries);
2275 			nr++;
2276 		}
2277 		if (altq == NULL) {
2278 			error = EBUSY;
2279 			break;
2280 		}
2281 		error = altq_getqstats(altq, pq->buf, &nbytes);
2282 		if (error == 0) {
2283 			pq->scheduler = altq->scheduler;
2284 			pq->nbytes = nbytes;
2285 		}
2286 		break;
2287 	}
2288 #endif /* ALTQ */
2289 
2290 	case DIOCBEGINADDRS: {
2291 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2292 
2293 		pf_empty_pool(&pf_pabuf);
2294 		pp->ticket = ++ticket_pabuf;
2295 		break;
2296 	}
2297 
2298 	case DIOCADDADDR: {
2299 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2300 
2301 		if (pp->ticket != ticket_pabuf) {
2302 			error = EBUSY;
2303 			break;
2304 		}
2305 #ifndef INET
2306 		if (pp->af == AF_INET) {
2307 			error = EAFNOSUPPORT;
2308 			break;
2309 		}
2310 #endif /* INET */
2311 #ifndef INET6
2312 		if (pp->af == AF_INET6) {
2313 			error = EAFNOSUPPORT;
2314 			break;
2315 		}
2316 #endif /* INET6 */
2317 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2318 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2319 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2320 			error = EINVAL;
2321 			break;
2322 		}
2323 		pa = pool_get(&pf_pooladdr_pl, PR_NOWAIT);
2324 		if (pa == NULL) {
2325 			error = ENOMEM;
2326 			break;
2327 		}
2328 		bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2329 		if (pa->ifname[0]) {
2330 			pa->kif = pfi_kif_get(pa->ifname);
2331 			if (pa->kif == NULL) {
2332 				pool_put(&pf_pooladdr_pl, pa);
2333 				error = EINVAL;
2334 				break;
2335 			}
2336 			pfi_kif_ref(pa->kif, PFI_KIF_REF_RULE);
2337 		}
2338 		if (pfi_dynaddr_setup(&pa->addr, pp->af)) {
2339 			pfi_dynaddr_remove(&pa->addr);
2340 			pfi_kif_unref(pa->kif, PFI_KIF_REF_RULE);
2341 			pool_put(&pf_pooladdr_pl, pa);
2342 			error = EINVAL;
2343 			break;
2344 		}
2345 		TAILQ_INSERT_TAIL(&pf_pabuf, pa, entries);
2346 		break;
2347 	}
2348 
2349 	case DIOCGETADDRS: {
2350 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2351 
2352 		pp->nr = 0;
2353 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2354 		    pp->r_num, 0, 1, 0);
2355 		if (pool == NULL) {
2356 			error = EBUSY;
2357 			break;
2358 		}
2359 		TAILQ_FOREACH(pa, &pool->list, entries)
2360 			pp->nr++;
2361 		break;
2362 	}
2363 
2364 	case DIOCGETADDR: {
2365 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2366 		u_int32_t		 nr = 0;
2367 
2368 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2369 		    pp->r_num, 0, 1, 1);
2370 		if (pool == NULL) {
2371 			error = EBUSY;
2372 			break;
2373 		}
2374 		pa = TAILQ_FIRST(&pool->list);
2375 		while ((pa != NULL) && (nr < pp->nr)) {
2376 			pa = TAILQ_NEXT(pa, entries);
2377 			nr++;
2378 		}
2379 		if (pa == NULL) {
2380 			error = EBUSY;
2381 			break;
2382 		}
2383 		bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2384 		pfi_dynaddr_copyout(&pp->addr.addr);
2385 		pf_tbladdr_copyout(&pp->addr.addr);
2386 		pf_rtlabel_copyout(&pp->addr.addr);
2387 		break;
2388 	}
2389 
2390 	case DIOCCHANGEADDR: {
2391 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
2392 		struct pf_pooladdr	*oldpa = NULL, *newpa = NULL;
2393 		struct pf_ruleset	*ruleset;
2394 
2395 		if (pca->action < PF_CHANGE_ADD_HEAD ||
2396 		    pca->action > PF_CHANGE_REMOVE) {
2397 			error = EINVAL;
2398 			break;
2399 		}
2400 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2401 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2402 		    pca->addr.addr.type != PF_ADDR_TABLE) {
2403 			error = EINVAL;
2404 			break;
2405 		}
2406 
2407 		ruleset = pf_find_ruleset(pca->anchor);
2408 		if (ruleset == NULL) {
2409 			error = EBUSY;
2410 			break;
2411 		}
2412 		pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2413 		    pca->r_num, pca->r_last, 1, 1);
2414 		if (pool == NULL) {
2415 			error = EBUSY;
2416 			break;
2417 		}
2418 		if (pca->action != PF_CHANGE_REMOVE) {
2419 			newpa = pool_get(&pf_pooladdr_pl, PR_NOWAIT);
2420 			if (newpa == NULL) {
2421 				error = ENOMEM;
2422 				break;
2423 			}
2424 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2425 #ifndef INET
2426 			if (pca->af == AF_INET) {
2427 				pool_put(&pf_pooladdr_pl, newpa);
2428 				error = EAFNOSUPPORT;
2429 				break;
2430 			}
2431 #endif /* INET */
2432 #ifndef INET6
2433 			if (pca->af == AF_INET6) {
2434 				pool_put(&pf_pooladdr_pl, newpa);
2435 				error = EAFNOSUPPORT;
2436 				break;
2437 			}
2438 #endif /* INET6 */
2439 			if (newpa->ifname[0]) {
2440 				newpa->kif = pfi_kif_get(newpa->ifname);
2441 				if (newpa->kif == NULL) {
2442 					pool_put(&pf_pooladdr_pl, newpa);
2443 					error = EINVAL;
2444 					break;
2445 				}
2446 				pfi_kif_ref(newpa->kif, PFI_KIF_REF_RULE);
2447 			} else
2448 				newpa->kif = NULL;
2449 			if (pfi_dynaddr_setup(&newpa->addr, pca->af) ||
2450 			    pf_tbladdr_setup(ruleset, &newpa->addr)) {
2451 				pfi_dynaddr_remove(&newpa->addr);
2452 				pfi_kif_unref(newpa->kif, PFI_KIF_REF_RULE);
2453 				pool_put(&pf_pooladdr_pl, newpa);
2454 				error = EINVAL;
2455 				break;
2456 			}
2457 		}
2458 
2459 		if (pca->action == PF_CHANGE_ADD_HEAD)
2460 			oldpa = TAILQ_FIRST(&pool->list);
2461 		else if (pca->action == PF_CHANGE_ADD_TAIL)
2462 			oldpa = TAILQ_LAST(&pool->list, pf_palist);
2463 		else {
2464 			int	i = 0;
2465 
2466 			oldpa = TAILQ_FIRST(&pool->list);
2467 			while ((oldpa != NULL) && (i < pca->nr)) {
2468 				oldpa = TAILQ_NEXT(oldpa, entries);
2469 				i++;
2470 			}
2471 			if (oldpa == NULL) {
2472 				error = EINVAL;
2473 				break;
2474 			}
2475 		}
2476 
2477 		if (pca->action == PF_CHANGE_REMOVE) {
2478 			TAILQ_REMOVE(&pool->list, oldpa, entries);
2479 			pfi_dynaddr_remove(&oldpa->addr);
2480 			pf_tbladdr_remove(&oldpa->addr);
2481 			pfi_kif_unref(oldpa->kif, PFI_KIF_REF_RULE);
2482 			pool_put(&pf_pooladdr_pl, oldpa);
2483 		} else {
2484 			if (oldpa == NULL)
2485 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2486 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
2487 			    pca->action == PF_CHANGE_ADD_BEFORE)
2488 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2489 			else
2490 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
2491 				    newpa, entries);
2492 		}
2493 
2494 		pool->cur = TAILQ_FIRST(&pool->list);
2495 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr,
2496 		    pca->af);
2497 		break;
2498 	}
2499 
2500 	case DIOCGETRULESETS: {
2501 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2502 		struct pf_ruleset	*ruleset;
2503 		struct pf_anchor	*anchor;
2504 
2505 		pr->path[sizeof(pr->path) - 1] = 0;
2506 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2507 			error = EINVAL;
2508 			break;
2509 		}
2510 		pr->nr = 0;
2511 		if (ruleset->anchor == NULL) {
2512 			/* XXX kludge for pf_main_ruleset */
2513 			RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2514 				if (anchor->parent == NULL)
2515 					pr->nr++;
2516 		} else {
2517 			RB_FOREACH(anchor, pf_anchor_node,
2518 			    &ruleset->anchor->children)
2519 				pr->nr++;
2520 		}
2521 		break;
2522 	}
2523 
2524 	case DIOCGETRULESET: {
2525 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2526 		struct pf_ruleset	*ruleset;
2527 		struct pf_anchor	*anchor;
2528 		u_int32_t		 nr = 0;
2529 
2530 		pr->path[sizeof(pr->path) - 1] = 0;
2531 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2532 			error = EINVAL;
2533 			break;
2534 		}
2535 		pr->name[0] = 0;
2536 		if (ruleset->anchor == NULL) {
2537 			/* XXX kludge for pf_main_ruleset */
2538 			RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2539 				if (anchor->parent == NULL && nr++ == pr->nr) {
2540 					strlcpy(pr->name, anchor->name,
2541 					    sizeof(pr->name));
2542 					break;
2543 				}
2544 		} else {
2545 			RB_FOREACH(anchor, pf_anchor_node,
2546 			    &ruleset->anchor->children)
2547 				if (nr++ == pr->nr) {
2548 					strlcpy(pr->name, anchor->name,
2549 					    sizeof(pr->name));
2550 					break;
2551 				}
2552 		}
2553 		if (!pr->name[0])
2554 			error = EBUSY;
2555 		break;
2556 	}
2557 
2558 	case DIOCRCLRTABLES: {
2559 		struct pfioc_table *io = (struct pfioc_table *)addr;
2560 
2561 		if (io->pfrio_esize != 0) {
2562 			error = ENODEV;
2563 			break;
2564 		}
2565 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2566 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2567 		break;
2568 	}
2569 
2570 	case DIOCRADDTABLES: {
2571 		struct pfioc_table *io = (struct pfioc_table *)addr;
2572 
2573 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2574 			error = ENODEV;
2575 			break;
2576 		}
2577 		error = pfr_add_tables(io->pfrio_buffer, io->pfrio_size,
2578 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2579 		break;
2580 	}
2581 
2582 	case DIOCRDELTABLES: {
2583 		struct pfioc_table *io = (struct pfioc_table *)addr;
2584 
2585 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2586 			error = ENODEV;
2587 			break;
2588 		}
2589 		error = pfr_del_tables(io->pfrio_buffer, io->pfrio_size,
2590 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2591 		break;
2592 	}
2593 
2594 	case DIOCRGETTABLES: {
2595 		struct pfioc_table *io = (struct pfioc_table *)addr;
2596 
2597 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2598 			error = ENODEV;
2599 			break;
2600 		}
2601 		error = pfr_get_tables(&io->pfrio_table, io->pfrio_buffer,
2602 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2603 		break;
2604 	}
2605 
2606 	case DIOCRGETTSTATS: {
2607 		struct pfioc_table *io = (struct pfioc_table *)addr;
2608 
2609 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2610 			error = ENODEV;
2611 			break;
2612 		}
2613 		error = pfr_get_tstats(&io->pfrio_table, io->pfrio_buffer,
2614 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2615 		break;
2616 	}
2617 
2618 	case DIOCRCLRTSTATS: {
2619 		struct pfioc_table *io = (struct pfioc_table *)addr;
2620 
2621 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2622 			error = ENODEV;
2623 			break;
2624 		}
2625 		error = pfr_clr_tstats(io->pfrio_buffer, io->pfrio_size,
2626 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2627 		break;
2628 	}
2629 
2630 	case DIOCRSETTFLAGS: {
2631 		struct pfioc_table *io = (struct pfioc_table *)addr;
2632 
2633 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2634 			error = ENODEV;
2635 			break;
2636 		}
2637 		error = pfr_set_tflags(io->pfrio_buffer, io->pfrio_size,
2638 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2639 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2640 		break;
2641 	}
2642 
2643 	case DIOCRCLRADDRS: {
2644 		struct pfioc_table *io = (struct pfioc_table *)addr;
2645 
2646 		if (io->pfrio_esize != 0) {
2647 			error = ENODEV;
2648 			break;
2649 		}
2650 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2651 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2652 		break;
2653 	}
2654 
2655 	case DIOCRADDADDRS: {
2656 		struct pfioc_table *io = (struct pfioc_table *)addr;
2657 
2658 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2659 			error = ENODEV;
2660 			break;
2661 		}
2662 		error = pfr_add_addrs(&io->pfrio_table, io->pfrio_buffer,
2663 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2664 		    PFR_FLAG_USERIOCTL);
2665 		break;
2666 	}
2667 
2668 	case DIOCRDELADDRS: {
2669 		struct pfioc_table *io = (struct pfioc_table *)addr;
2670 
2671 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2672 			error = ENODEV;
2673 			break;
2674 		}
2675 		error = pfr_del_addrs(&io->pfrio_table, io->pfrio_buffer,
2676 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2677 		    PFR_FLAG_USERIOCTL);
2678 		break;
2679 	}
2680 
2681 	case DIOCRSETADDRS: {
2682 		struct pfioc_table *io = (struct pfioc_table *)addr;
2683 
2684 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2685 			error = ENODEV;
2686 			break;
2687 		}
2688 		error = pfr_set_addrs(&io->pfrio_table, io->pfrio_buffer,
2689 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2690 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2691 		    PFR_FLAG_USERIOCTL, 0);
2692 		break;
2693 	}
2694 
2695 	case DIOCRGETADDRS: {
2696 		struct pfioc_table *io = (struct pfioc_table *)addr;
2697 
2698 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2699 			error = ENODEV;
2700 			break;
2701 		}
2702 		error = pfr_get_addrs(&io->pfrio_table, io->pfrio_buffer,
2703 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2704 		break;
2705 	}
2706 
2707 	case DIOCRGETASTATS: {
2708 		struct pfioc_table *io = (struct pfioc_table *)addr;
2709 
2710 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2711 			error = ENODEV;
2712 			break;
2713 		}
2714 		error = pfr_get_astats(&io->pfrio_table, io->pfrio_buffer,
2715 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2716 		break;
2717 	}
2718 
2719 	case DIOCRCLRASTATS: {
2720 		struct pfioc_table *io = (struct pfioc_table *)addr;
2721 
2722 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2723 			error = ENODEV;
2724 			break;
2725 		}
2726 		error = pfr_clr_astats(&io->pfrio_table, io->pfrio_buffer,
2727 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2728 		    PFR_FLAG_USERIOCTL);
2729 		break;
2730 	}
2731 
2732 	case DIOCRTSTADDRS: {
2733 		struct pfioc_table *io = (struct pfioc_table *)addr;
2734 
2735 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2736 			error = ENODEV;
2737 			break;
2738 		}
2739 		error = pfr_tst_addrs(&io->pfrio_table, io->pfrio_buffer,
2740 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2741 		    PFR_FLAG_USERIOCTL);
2742 		break;
2743 	}
2744 
2745 	case DIOCRINADEFINE: {
2746 		struct pfioc_table *io = (struct pfioc_table *)addr;
2747 
2748 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2749 			error = ENODEV;
2750 			break;
2751 		}
2752 		error = pfr_ina_define(&io->pfrio_table, io->pfrio_buffer,
2753 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2754 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2755 		break;
2756 	}
2757 
2758 	case DIOCOSFPADD: {
2759 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2760 		error = pf_osfp_add(io);
2761 		break;
2762 	}
2763 
2764 	case DIOCOSFPGET: {
2765 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2766 		error = pf_osfp_get(io);
2767 		break;
2768 	}
2769 
2770 	case DIOCXBEGIN: {
2771 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2772 		struct pfioc_trans_e	*ioe;
2773 		struct pfr_table	*table;
2774 		int			 i;
2775 
2776 		if (io->esize != sizeof(*ioe)) {
2777 			error = ENODEV;
2778 			goto fail;
2779 		}
2780 		ioe = (struct pfioc_trans_e *)malloc(sizeof(*ioe),
2781 		    M_TEMP, M_WAITOK);
2782 		table = (struct pfr_table *)malloc(sizeof(*table),
2783 		    M_TEMP, M_WAITOK);
2784 		for (i = 0; i < io->size; i++) {
2785 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2786 				free(table, M_TEMP);
2787 				free(ioe, M_TEMP);
2788 				error = EFAULT;
2789 				goto fail;
2790 			}
2791 			switch (ioe->rs_num) {
2792 #ifdef ALTQ
2793 			case PF_RULESET_ALTQ:
2794 				if (ioe->anchor[0]) {
2795 					free(table, M_TEMP);
2796 					free(ioe, M_TEMP);
2797 					error = EINVAL;
2798 					goto fail;
2799 				}
2800 				if ((error = pf_begin_altq(&ioe->ticket))) {
2801 					free(table, M_TEMP);
2802 					free(ioe, M_TEMP);
2803 					goto fail;
2804 				}
2805 				break;
2806 #endif /* ALTQ */
2807 			case PF_RULESET_TABLE:
2808 				bzero(table, sizeof(*table));
2809 				strlcpy(table->pfrt_anchor, ioe->anchor,
2810 				    sizeof(table->pfrt_anchor));
2811 				if ((error = pfr_ina_begin(table,
2812 				    &ioe->ticket, NULL, 0))) {
2813 					free(table, M_TEMP);
2814 					free(ioe, M_TEMP);
2815 					goto fail;
2816 				}
2817 				break;
2818 			default:
2819 				if ((error = pf_begin_rules(&ioe->ticket,
2820 				    ioe->rs_num, ioe->anchor))) {
2821 					free(table, M_TEMP);
2822 					free(ioe, M_TEMP);
2823 					goto fail;
2824 				}
2825 				break;
2826 			}
2827 			if (copyout(ioe, io->array+i, sizeof(io->array[i]))) {
2828 				free(table, M_TEMP);
2829 				free(ioe, M_TEMP);
2830 				error = EFAULT;
2831 				goto fail;
2832 			}
2833 		}
2834 		free(table, M_TEMP);
2835 		free(ioe, M_TEMP);
2836 		break;
2837 	}
2838 
2839 	case DIOCXROLLBACK: {
2840 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2841 		struct pfioc_trans_e	*ioe;
2842 		struct pfr_table	*table;
2843 		int			 i;
2844 
2845 		if (io->esize != sizeof(*ioe)) {
2846 			error = ENODEV;
2847 			goto fail;
2848 		}
2849 		ioe = (struct pfioc_trans_e *)malloc(sizeof(*ioe),
2850 		    M_TEMP, M_WAITOK);
2851 		table = (struct pfr_table *)malloc(sizeof(*table),
2852 		    M_TEMP, M_WAITOK);
2853 		for (i = 0; i < io->size; i++) {
2854 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2855 				free(table, M_TEMP);
2856 				free(ioe, M_TEMP);
2857 				error = EFAULT;
2858 				goto fail;
2859 			}
2860 			switch (ioe->rs_num) {
2861 #ifdef ALTQ
2862 			case PF_RULESET_ALTQ:
2863 				if (ioe->anchor[0]) {
2864 					free(table, M_TEMP);
2865 					free(ioe, M_TEMP);
2866 					error = EINVAL;
2867 					goto fail;
2868 				}
2869 				if ((error = pf_rollback_altq(ioe->ticket))) {
2870 					free(table, M_TEMP);
2871 					free(ioe, M_TEMP);
2872 					goto fail; /* really bad */
2873 				}
2874 				break;
2875 #endif /* ALTQ */
2876 			case PF_RULESET_TABLE:
2877 				bzero(table, sizeof(*table));
2878 				strlcpy(table->pfrt_anchor, ioe->anchor,
2879 				    sizeof(table->pfrt_anchor));
2880 				if ((error = pfr_ina_rollback(table,
2881 				    ioe->ticket, NULL, 0))) {
2882 					free(table, M_TEMP);
2883 					free(ioe, M_TEMP);
2884 					goto fail; /* really bad */
2885 				}
2886 				break;
2887 			default:
2888 				if ((error = pf_rollback_rules(ioe->ticket,
2889 				    ioe->rs_num, ioe->anchor))) {
2890 					free(table, M_TEMP);
2891 					free(ioe, M_TEMP);
2892 					goto fail; /* really bad */
2893 				}
2894 				break;
2895 			}
2896 		}
2897 		free(table, M_TEMP);
2898 		free(ioe, M_TEMP);
2899 		break;
2900 	}
2901 
2902 	case DIOCXCOMMIT: {
2903 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2904 		struct pfioc_trans_e	*ioe;
2905 		struct pfr_table	*table;
2906 		struct pf_ruleset	*rs;
2907 		int			 i;
2908 
2909 		if (io->esize != sizeof(*ioe)) {
2910 			error = ENODEV;
2911 			goto fail;
2912 		}
2913 		ioe = (struct pfioc_trans_e *)malloc(sizeof(*ioe),
2914 		    M_TEMP, M_WAITOK);
2915 		table = (struct pfr_table *)malloc(sizeof(*table),
2916 		    M_TEMP, M_WAITOK);
2917 		/* first makes sure everything will succeed */
2918 		for (i = 0; i < io->size; i++) {
2919 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2920 				free(table, M_TEMP);
2921 				free(ioe, M_TEMP);
2922 				error = EFAULT;
2923 				goto fail;
2924 			}
2925 			switch (ioe->rs_num) {
2926 #ifdef ALTQ
2927 			case PF_RULESET_ALTQ:
2928 				if (ioe->anchor[0]) {
2929 					free(table, M_TEMP);
2930 					free(ioe, M_TEMP);
2931 					error = EINVAL;
2932 					goto fail;
2933 				}
2934 				if (!altqs_inactive_open || ioe->ticket !=
2935 				    ticket_altqs_inactive) {
2936 					free(table, M_TEMP);
2937 					free(ioe, M_TEMP);
2938 					error = EBUSY;
2939 					goto fail;
2940 				}
2941 				break;
2942 #endif /* ALTQ */
2943 			case PF_RULESET_TABLE:
2944 				rs = pf_find_ruleset(ioe->anchor);
2945 				if (rs == NULL || !rs->topen || ioe->ticket !=
2946 				     rs->tticket) {
2947 					free(table, M_TEMP);
2948 					free(ioe, M_TEMP);
2949 					error = EBUSY;
2950 					goto fail;
2951 				}
2952 				break;
2953 			default:
2954 				if (ioe->rs_num < 0 || ioe->rs_num >=
2955 				    PF_RULESET_MAX) {
2956 					free(table, M_TEMP);
2957 					free(ioe, M_TEMP);
2958 					error = EINVAL;
2959 					goto fail;
2960 				}
2961 				rs = pf_find_ruleset(ioe->anchor);
2962 				if (rs == NULL ||
2963 				    !rs->rules[ioe->rs_num].inactive.open ||
2964 				    rs->rules[ioe->rs_num].inactive.ticket !=
2965 				    ioe->ticket) {
2966 					free(table, M_TEMP);
2967 					free(ioe, M_TEMP);
2968 					error = EBUSY;
2969 					goto fail;
2970 				}
2971 				break;
2972 			}
2973 		}
2974 		/* now do the commit - no errors should happen here */
2975 		for (i = 0; i < io->size; i++) {
2976 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2977 				free(table, M_TEMP);
2978 				free(ioe, M_TEMP);
2979 				error = EFAULT;
2980 				goto fail;
2981 			}
2982 			switch (ioe->rs_num) {
2983 #ifdef ALTQ
2984 			case PF_RULESET_ALTQ:
2985 				if ((error = pf_commit_altq(ioe->ticket))) {
2986 					free(table, M_TEMP);
2987 					free(ioe, M_TEMP);
2988 					goto fail; /* really bad */
2989 				}
2990 				break;
2991 #endif /* ALTQ */
2992 			case PF_RULESET_TABLE:
2993 				bzero(table, sizeof(*table));
2994 				strlcpy(table->pfrt_anchor, ioe->anchor,
2995 				    sizeof(table->pfrt_anchor));
2996 				if ((error = pfr_ina_commit(table, ioe->ticket,
2997 				    NULL, NULL, 0))) {
2998 					free(table, M_TEMP);
2999 					free(ioe, M_TEMP);
3000 					goto fail; /* really bad */
3001 				}
3002 				break;
3003 			default:
3004 				if ((error = pf_commit_rules(ioe->ticket,
3005 				    ioe->rs_num, ioe->anchor))) {
3006 					free(table, M_TEMP);
3007 					free(ioe, M_TEMP);
3008 					goto fail; /* really bad */
3009 				}
3010 				break;
3011 			}
3012 		}
3013 		free(table, M_TEMP);
3014 		free(ioe, M_TEMP);
3015 		break;
3016 	}
3017 
3018 	case DIOCGETSRCNODES: {
3019 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
3020 		struct pf_src_node	*n, *p, *pstore;
3021 		u_int32_t		 nr = 0;
3022 		int			 space = psn->psn_len;
3023 
3024 		if (space == 0) {
3025 			RB_FOREACH(n, pf_src_tree, &tree_src_tracking)
3026 				nr++;
3027 			psn->psn_len = sizeof(struct pf_src_node) * nr;
3028 			break;
3029 		}
3030 
3031 		pstore = malloc(sizeof(*pstore), M_TEMP, M_WAITOK);
3032 
3033 		p = psn->psn_src_nodes;
3034 		RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
3035 			int	secs = time_second, diff;
3036 
3037 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
3038 				break;
3039 
3040 			bcopy(n, pstore, sizeof(*pstore));
3041 			if (n->rule.ptr != NULL)
3042 				pstore->rule.nr = n->rule.ptr->nr;
3043 			pstore->creation = secs - pstore->creation;
3044 			if (pstore->expire > secs)
3045 				pstore->expire -= secs;
3046 			else
3047 				pstore->expire = 0;
3048 
3049 			/* adjust the connection rate estimate */
3050 			diff = secs - n->conn_rate.last;
3051 			if (diff >= n->conn_rate.seconds)
3052 				pstore->conn_rate.count = 0;
3053 			else
3054 				pstore->conn_rate.count -=
3055 				    n->conn_rate.count * diff /
3056 				    n->conn_rate.seconds;
3057 
3058 			error = copyout(pstore, p, sizeof(*p));
3059 			if (error) {
3060 				free(pstore, M_TEMP);
3061 				goto fail;
3062 			}
3063 			p++;
3064 			nr++;
3065 		}
3066 		psn->psn_len = sizeof(struct pf_src_node) * nr;
3067 
3068 		free(pstore, M_TEMP);
3069 		break;
3070 	}
3071 
3072 	case DIOCCLRSRCNODES: {
3073 		struct pf_src_node	*n;
3074 		struct pf_state		*state;
3075 
3076 		RB_FOREACH(state, pf_state_tree_id, &tree_id) {
3077 			state->src_node = NULL;
3078 			state->nat_src_node = NULL;
3079 		}
3080 		RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
3081 			n->expire = 1;
3082 			n->states = 0;
3083 		}
3084 		pf_purge_expired_src_nodes(1);
3085 		pf_status.src_nodes = 0;
3086 		break;
3087 	}
3088 
3089 	case DIOCKILLSRCNODES: {
3090 		struct pf_src_node	*sn;
3091 		struct pf_state		*s;
3092 		struct pfioc_src_node_kill *psnk = \
3093 			(struct pfioc_src_node_kill *) addr;
3094 		int			killed = 0;
3095 
3096 		RB_FOREACH(sn, pf_src_tree, &tree_src_tracking) {
3097         		if (PF_MATCHA(psnk->psnk_src.neg, \
3098 				      &psnk->psnk_src.addr.v.a.addr, \
3099 				      &psnk->psnk_src.addr.v.a.mask, \
3100 				      &sn->addr, sn->af) &&
3101 			    PF_MATCHA(psnk->psnk_dst.neg, \
3102 				      &psnk->psnk_dst.addr.v.a.addr, \
3103 				      &psnk->psnk_dst.addr.v.a.mask, \
3104 				      &sn->raddr, sn->af)) {
3105 				/* Handle state to src_node linkage */
3106 				if (sn->states != 0) {
3107 					RB_FOREACH(s, pf_state_tree_id,
3108 					    &tree_id) {
3109 						if (s->src_node == sn)
3110 							s->src_node = NULL;
3111 						if (s->nat_src_node == sn)
3112 							s->nat_src_node = NULL;
3113 					}
3114 					sn->states = 0;
3115 				}
3116 				sn->expire = 1;
3117 				killed++;
3118 			}
3119 		}
3120 
3121 		if (killed > 0)
3122 			pf_purge_expired_src_nodes(1);
3123 
3124 		psnk->psnk_af = killed;
3125 		break;
3126 	}
3127 
3128 	case DIOCSETHOSTID: {
3129 		u_int32_t	*hostid = (u_int32_t *)addr;
3130 
3131 		if (*hostid == 0)
3132 			pf_status.hostid = arc4random();
3133 		else
3134 			pf_status.hostid = *hostid;
3135 		break;
3136 	}
3137 
3138 	case DIOCOSFPFLUSH:
3139 		pf_osfp_flush();
3140 		break;
3141 
3142 	case DIOCIGETIFACES: {
3143 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3144 
3145 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
3146 			error = ENODEV;
3147 			break;
3148 		}
3149 		error = pfi_get_ifaces(io->pfiio_name, io->pfiio_buffer,
3150 		    &io->pfiio_size);
3151 		break;
3152 	}
3153 
3154 	case DIOCSETIFFLAG: {
3155 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3156 
3157 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3158 		break;
3159 	}
3160 
3161 	case DIOCCLRIFFLAG: {
3162 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3163 
3164 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3165 		break;
3166 	}
3167 
3168 	default:
3169 		error = ENODEV;
3170 		break;
3171 	}
3172 fail:
3173 	splx(s);
3174 	if (flags & FWRITE)
3175 		rw_exit_write(&pf_consistency_lock);
3176 	else
3177 		rw_exit_read(&pf_consistency_lock);
3178 	return (error);
3179 }
3180 
3181 #ifdef __NetBSD__
3182 #ifdef INET
3183 static int
3184 pfil4_wrapper(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
3185 {
3186 	int error;
3187 
3188 	/*
3189 	 * ensure that mbufs are writable beforehand
3190 	 * as it's assumed by pf code.
3191 	 * ip hdr (60 bytes) + tcp hdr (60 bytes) should be enough.
3192 	 * XXX inefficient
3193 	 */
3194 	error = m_makewritable(mp, 0, 60 + 60, M_DONTWAIT);
3195 	if (error) {
3196 		m_freem(*mp);
3197 		*mp = NULL;
3198 		return error;
3199 	}
3200 
3201 	/*
3202 	 * If the packet is out-bound, we can't delay checksums
3203 	 * here.  For in-bound, the checksum has already been
3204 	 * validated.
3205 	 */
3206 	if (dir == PFIL_OUT) {
3207 		if ((*mp)->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
3208 			in_delayed_cksum(*mp);
3209 			(*mp)->m_pkthdr.csum_flags &=
3210 			    ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
3211 		}
3212 	}
3213 
3214 	if (pf_test(dir == PFIL_OUT ? PF_OUT : PF_IN, ifp, mp, NULL)
3215 	    != PF_PASS) {
3216 		m_freem(*mp);
3217 		*mp = NULL;
3218 		return EHOSTUNREACH;
3219 	}
3220 
3221 	/*
3222 	 * we're not compatible with fast-forward.
3223 	 */
3224 
3225 	if (dir == PFIL_IN && *mp) {
3226 		(*mp)->m_flags &= ~M_CANFASTFWD;
3227 	}
3228 
3229 	return (0);
3230 }
3231 #endif /* INET */
3232 
3233 #ifdef INET6
3234 static int
3235 pfil6_wrapper(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
3236 {
3237 	int error;
3238 
3239 	/*
3240 	 * ensure that mbufs are writable beforehand
3241 	 * as it's assumed by pf code.
3242 	 * XXX inefficient
3243 	 */
3244 	error = m_makewritable(mp, 0, M_COPYALL, M_DONTWAIT);
3245 	if (error) {
3246 		m_freem(*mp);
3247 		*mp = NULL;
3248 		return error;
3249 	}
3250 
3251 	/*
3252 	 * If the packet is out-bound, we can't delay checksums
3253 	 * here.  For in-bound, the checksum has already been
3254 	 * validated.
3255 	 */
3256 	if (dir == PFIL_OUT) {
3257 		if ((*mp)->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) {
3258 			in6_delayed_cksum(*mp);
3259 			(*mp)->m_pkthdr.csum_flags &=
3260 			    ~(M_CSUM_TCPv6|M_CSUM_UDPv6);
3261 		}
3262 	}
3263 
3264 	if (pf_test6(dir == PFIL_OUT ? PF_OUT : PF_IN, ifp, mp, NULL)
3265 	    != PF_PASS) {
3266 		m_freem(*mp);
3267 		*mp = NULL;
3268 		return EHOSTUNREACH;
3269 	} else
3270 		return (0);
3271 }
3272 #endif /* INET6 */
3273 
3274 static int
3275 pf_pfil_attach(void)
3276 {
3277 	struct pfil_head *ph_inet;
3278 #ifdef INET6
3279 	struct pfil_head *ph_inet6;
3280 #endif /* INET6 */
3281 	int error;
3282 
3283 	if (pf_pfil_attached)
3284 		return (EBUSY);
3285 
3286 	ph_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3287 	if (ph_inet)
3288 		error = pfil_add_hook((void *)pfil4_wrapper, NULL,
3289 		    PFIL_IN|PFIL_OUT, ph_inet);
3290 	else
3291 		error = ENOENT;
3292 	if (error)
3293 		return (error);
3294 
3295 #ifdef INET6
3296 	ph_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3297 	if (ph_inet6)
3298 		error = pfil_add_hook((void *)pfil6_wrapper, NULL,
3299 		    PFIL_IN|PFIL_OUT, ph_inet6);
3300 	else
3301 		error = ENOENT;
3302 	if (error)
3303 		goto bad;
3304 #endif /* INET6 */
3305 
3306 	pf_pfil_attached = 1;
3307 
3308 	return (0);
3309 
3310 #ifdef INET6
3311 bad:
3312 	pfil_remove_hook(pfil4_wrapper, NULL, PFIL_IN|PFIL_OUT, ph_inet);
3313 #endif /* INET6 */
3314 
3315 	return (error);
3316 }
3317 
3318 static int
3319 pf_pfil_detach(void)
3320 {
3321 	struct pfil_head *ph_inet;
3322 #ifdef INET6
3323 	struct pfil_head *ph_inet6;
3324 #endif /* INET6 */
3325 
3326 	if (pf_pfil_attached == 0)
3327 		return (EBUSY);
3328 
3329 	ph_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3330 	if (ph_inet)
3331 		pfil_remove_hook((void *)pfil4_wrapper, NULL,
3332 		    PFIL_IN|PFIL_OUT, ph_inet);
3333 #ifdef INET6
3334 	ph_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3335 	if (ph_inet6)
3336 		pfil_remove_hook((void *)pfil6_wrapper, NULL,
3337 		    PFIL_IN|PFIL_OUT, ph_inet6);
3338 #endif /* INET6 */
3339 	pf_pfil_attached = 0;
3340 
3341 	return (0);
3342 }
3343 #endif /* __NetBSD__ */
3344 
3345 #if defined(__NetBSD__)
3346 MODULE(MODULE_CLASS_DRIVER, pf, "bpf");
3347 
3348 static int
3349 pf_modcmd(modcmd_t cmd, void *opaque)
3350 {
3351 #ifdef _MODULE
3352 	extern void pflogattach(int);
3353 	extern void pflogdetach(void);
3354 
3355 	devmajor_t cmajor = NODEVMAJOR, bmajor = NODEVMAJOR;
3356 	int err;
3357 
3358 	switch (cmd) {
3359 	case MODULE_CMD_INIT:
3360 		err = devsw_attach("pf", NULL, &bmajor, &pf_cdevsw, &cmajor);
3361 		if (err)
3362 			return err;
3363 		pfattach(1);
3364 		pflogattach(1);
3365 		return 0;
3366 	case MODULE_CMD_FINI:
3367 		if (pf_status.running) {
3368 			return EBUSY;
3369 		} else {
3370 			pfdetach();
3371 			pflogdetach();
3372 			return devsw_detach(NULL, &pf_cdevsw);
3373 		}
3374 	default:
3375 		return ENOTTY;
3376 	}
3377 #else
3378 	if (cmd == MODULE_CMD_INIT)
3379 		return 0;
3380 	return ENOTTY;
3381 #endif
3382 }
3383 #endif
3384