xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision 1a61beb0549e05b33df31380e427d90f6e46ff7e)
1 /*-
2  * Copyright (c) 2001 Daniel Hartmeier
3  * Copyright (c) 2002,2003 Henning Brauer
4  * Copyright (c) 2012 Gleb Smirnoff <glebius@FreeBSD.org>
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  *
11  *    - Redistributions of source code must retain the above copyright
12  *      notice, this list of conditions and the following disclaimer.
13  *    - Redistributions in binary form must reproduce the above
14  *      copyright notice, this list of conditions and the following
15  *      disclaimer in the documentation and/or other materials provided
16  *      with the distribution.
17  *
18  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
19  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
20  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
21  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
22  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
23  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
24  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
25  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
26  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
28  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29  * POSSIBILITY OF SUCH DAMAGE.
30  *
31  * Effort sponsored in part by the Defense Advanced Research Projects
32  * Agency (DARPA) and Air Force Research Laboratory, Air Force
33  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
34  *
35  *	$OpenBSD: pf_ioctl.c,v 1.213 2009/02/15 21:46:12 mbalmer Exp $
36  */
37 
38 #include <sys/cdefs.h>
39 __FBSDID("$FreeBSD$");
40 
41 #include "opt_inet.h"
42 #include "opt_inet6.h"
43 #include "opt_bpf.h"
44 #include "opt_pf.h"
45 
46 #include <sys/param.h>
47 #include <sys/bus.h>
48 #include <sys/conf.h>
49 #include <sys/endian.h>
50 #include <sys/fcntl.h>
51 #include <sys/filio.h>
52 #include <sys/interrupt.h>
53 #include <sys/jail.h>
54 #include <sys/kernel.h>
55 #include <sys/kthread.h>
56 #include <sys/lock.h>
57 #include <sys/mbuf.h>
58 #include <sys/module.h>
59 #include <sys/proc.h>
60 #include <sys/rwlock.h>
61 #include <sys/smp.h>
62 #include <sys/socket.h>
63 #include <sys/sysctl.h>
64 #include <sys/md5.h>
65 #include <sys/ucred.h>
66 
67 #include <net/if.h>
68 #include <net/if_var.h>
69 #include <net/vnet.h>
70 #include <net/route.h>
71 #include <net/pfil.h>
72 #include <net/pfvar.h>
73 #include <net/if_pfsync.h>
74 #include <net/if_pflog.h>
75 
76 #include <netinet/in.h>
77 #include <netinet/ip.h>
78 #include <netinet/ip_var.h>
79 #include <netinet/ip_icmp.h>
80 
81 #ifdef INET6
82 #include <netinet/ip6.h>
83 #endif /* INET6 */
84 
85 #ifdef ALTQ
86 #include <altq/altq.h>
87 #endif
88 
89 static int		 pfattach(void);
90 static struct pf_pool	*pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t,
91 			    u_int8_t, u_int8_t, u_int8_t);
92 
93 static void		 pf_mv_pool(struct pf_palist *, struct pf_palist *);
94 static void		 pf_empty_pool(struct pf_palist *);
95 static int		 pfioctl(struct cdev *, u_long, caddr_t, int,
96 			    struct thread *);
97 #ifdef ALTQ
98 static int		 pf_begin_altq(u_int32_t *);
99 static int		 pf_rollback_altq(u_int32_t);
100 static int		 pf_commit_altq(u_int32_t);
101 static int		 pf_enable_altq(struct pf_altq *);
102 static int		 pf_disable_altq(struct pf_altq *);
103 static u_int32_t	 pf_qname2qid(char *);
104 static void		 pf_qid_unref(u_int32_t);
105 #endif /* ALTQ */
106 static int		 pf_begin_rules(u_int32_t *, int, const char *);
107 static int		 pf_rollback_rules(u_int32_t, int, char *);
108 static int		 pf_setup_pfsync_matching(struct pf_ruleset *);
109 static void		 pf_hash_rule(MD5_CTX *, struct pf_rule *);
110 static void		 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
111 static int		 pf_commit_rules(u_int32_t, int, char *);
112 static int		 pf_addr_setup(struct pf_ruleset *,
113 			    struct pf_addr_wrap *, sa_family_t);
114 static void		 pf_addr_copyout(struct pf_addr_wrap *);
115 
116 VNET_DEFINE(struct pf_rule,	pf_default_rule);
117 
118 #ifdef ALTQ
119 static VNET_DEFINE(int,		pf_altq_running);
120 #define	V_pf_altq_running	VNET(pf_altq_running)
121 #endif
122 
123 #define	TAGID_MAX	 50000
124 struct pf_tagname {
125 	TAILQ_ENTRY(pf_tagname)	entries;
126 	char			name[PF_TAG_NAME_SIZE];
127 	uint16_t		tag;
128 	int			ref;
129 };
130 
131 TAILQ_HEAD(pf_tags, pf_tagname);
132 #define	V_pf_tags		VNET(pf_tags)
133 VNET_DEFINE(struct pf_tags, pf_tags);
134 #define	V_pf_qids		VNET(pf_qids)
135 VNET_DEFINE(struct pf_tags, pf_qids);
136 static MALLOC_DEFINE(M_PFTAG, "pf_tag", "pf(4) tag names");
137 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
138 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
139 
140 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
141 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
142 #endif
143 
144 static u_int16_t	 tagname2tag(struct pf_tags *, char *);
145 static u_int16_t	 pf_tagname2tag(char *);
146 static void		 tag_unref(struct pf_tags *, u_int16_t);
147 
148 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
149 
150 struct cdev *pf_dev;
151 
152 /*
153  * XXX - These are new and need to be checked when moveing to a new version
154  */
155 static void		 pf_clear_states(void);
156 static int		 pf_clear_tables(void);
157 static void		 pf_clear_srcnodes(struct pf_src_node *);
158 static void		 pf_kill_srcnodes(struct pfioc_src_node_kill *);
159 static void		 pf_tbladdr_copyout(struct pf_addr_wrap *);
160 
161 /*
162  * Wrapper functions for pfil(9) hooks
163  */
164 #ifdef INET
165 static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp,
166     int dir, struct inpcb *inp);
167 static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp,
168     int dir, struct inpcb *inp);
169 #endif
170 #ifdef INET6
171 static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp,
172     int dir, struct inpcb *inp);
173 static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp,
174     int dir, struct inpcb *inp);
175 #endif
176 
177 static int		hook_pf(void);
178 static int		dehook_pf(void);
179 static int		shutdown_pf(void);
180 static int		pf_load(void);
181 static int		pf_unload(void);
182 
183 static struct cdevsw pf_cdevsw = {
184 	.d_ioctl =	pfioctl,
185 	.d_name =	PF_NAME,
186 	.d_version =	D_VERSION,
187 };
188 
189 static volatile VNET_DEFINE(int, pf_pfil_hooked);
190 #define V_pf_pfil_hooked	VNET(pf_pfil_hooked)
191 VNET_DEFINE(int,		pf_end_threads);
192 
193 struct rwlock			pf_rules_lock;
194 
195 /* pfsync */
196 pfsync_state_import_t 		*pfsync_state_import_ptr = NULL;
197 pfsync_insert_state_t		*pfsync_insert_state_ptr = NULL;
198 pfsync_update_state_t		*pfsync_update_state_ptr = NULL;
199 pfsync_delete_state_t		*pfsync_delete_state_ptr = NULL;
200 pfsync_clear_states_t		*pfsync_clear_states_ptr = NULL;
201 pfsync_defer_t			*pfsync_defer_ptr = NULL;
202 /* pflog */
203 pflog_packet_t			*pflog_packet_ptr = NULL;
204 
205 static int
206 pfattach(void)
207 {
208 	u_int32_t *my_timeout = V_pf_default_rule.timeout;
209 	int error;
210 
211 	pf_initialize();
212 	pfr_initialize();
213 	pfi_initialize();
214 	pf_normalize_init();
215 
216 	V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
217 	V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
218 
219 	RB_INIT(&V_pf_anchors);
220 	pf_init_ruleset(&pf_main_ruleset);
221 
222 	/* default rule should never be garbage collected */
223 	V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
224 #ifdef PF_DEFAULT_TO_DROP
225 	V_pf_default_rule.action = PF_DROP;
226 #else
227 	V_pf_default_rule.action = PF_PASS;
228 #endif
229 	V_pf_default_rule.nr = -1;
230 	V_pf_default_rule.rtableid = -1;
231 
232 	V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
233 	V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
234 	V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
235 
236 	/* initialize default timeouts */
237 	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
238 	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
239 	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
240 	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
241 	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
242 	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
243 	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
244 	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
245 	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
246 	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
247 	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
248 	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
249 	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
250 	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
251 	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
252 	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
253 	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
254 	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
255 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
256 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
257 
258 	bzero(&V_pf_status, sizeof(V_pf_status));
259 	V_pf_status.debug = PF_DEBUG_URGENT;
260 
261 	V_pf_pfil_hooked = 0;
262 
263 	/* XXX do our best to avoid a conflict */
264 	V_pf_status.hostid = arc4random();
265 
266 	if ((error = kproc_create(pf_purge_thread, curvnet, NULL, 0, 0,
267 	    "pf purge")) != 0)
268 		/* XXXGL: leaked all above. */
269 		return (error);
270 	if ((error = swi_add(NULL, "pf send", pf_intr, curvnet, SWI_NET,
271 	    INTR_MPSAFE, &V_pf_swi_cookie)) != 0)
272 		/* XXXGL: leaked all above. */
273 		return (error);
274 
275 	return (0);
276 }
277 
278 static struct pf_pool *
279 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
280     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
281     u_int8_t check_ticket)
282 {
283 	struct pf_ruleset	*ruleset;
284 	struct pf_rule		*rule;
285 	int			 rs_num;
286 
287 	ruleset = pf_find_ruleset(anchor);
288 	if (ruleset == NULL)
289 		return (NULL);
290 	rs_num = pf_get_ruleset_number(rule_action);
291 	if (rs_num >= PF_RULESET_MAX)
292 		return (NULL);
293 	if (active) {
294 		if (check_ticket && ticket !=
295 		    ruleset->rules[rs_num].active.ticket)
296 			return (NULL);
297 		if (r_last)
298 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
299 			    pf_rulequeue);
300 		else
301 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
302 	} else {
303 		if (check_ticket && ticket !=
304 		    ruleset->rules[rs_num].inactive.ticket)
305 			return (NULL);
306 		if (r_last)
307 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
308 			    pf_rulequeue);
309 		else
310 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
311 	}
312 	if (!r_last) {
313 		while ((rule != NULL) && (rule->nr != rule_number))
314 			rule = TAILQ_NEXT(rule, entries);
315 	}
316 	if (rule == NULL)
317 		return (NULL);
318 
319 	return (&rule->rpool);
320 }
321 
322 static void
323 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
324 {
325 	struct pf_pooladdr	*mv_pool_pa;
326 
327 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
328 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
329 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
330 	}
331 }
332 
333 static void
334 pf_empty_pool(struct pf_palist *poola)
335 {
336 	struct pf_pooladdr *pa;
337 
338 	while ((pa = TAILQ_FIRST(poola)) != NULL) {
339 		switch (pa->addr.type) {
340 		case PF_ADDR_DYNIFTL:
341 			pfi_dynaddr_remove(pa->addr.p.dyn);
342 			break;
343 		case PF_ADDR_TABLE:
344 			pfr_detach_table(pa->addr.p.tbl);
345 			break;
346 		}
347 		if (pa->kif)
348 			pfi_kif_unref(pa->kif);
349 		TAILQ_REMOVE(poola, pa, entries);
350 		free(pa, M_PFRULE);
351 	}
352 }
353 
354 static void
355 pf_unlink_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
356 {
357 
358 	PF_RULES_WASSERT();
359 
360 	TAILQ_REMOVE(rulequeue, rule, entries);
361 
362 	PF_UNLNKDRULES_LOCK();
363 	rule->rule_flag |= PFRULE_REFS;
364 	TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
365 	PF_UNLNKDRULES_UNLOCK();
366 }
367 
368 void
369 pf_free_rule(struct pf_rule *rule)
370 {
371 
372 	PF_RULES_WASSERT();
373 
374 	if (rule->tag)
375 		tag_unref(&V_pf_tags, rule->tag);
376 	if (rule->match_tag)
377 		tag_unref(&V_pf_tags, rule->match_tag);
378 #ifdef ALTQ
379 	if (rule->pqid != rule->qid)
380 		pf_qid_unref(rule->pqid);
381 	pf_qid_unref(rule->qid);
382 #endif
383 	switch (rule->src.addr.type) {
384 	case PF_ADDR_DYNIFTL:
385 		pfi_dynaddr_remove(rule->src.addr.p.dyn);
386 		break;
387 	case PF_ADDR_TABLE:
388 		pfr_detach_table(rule->src.addr.p.tbl);
389 		break;
390 	}
391 	switch (rule->dst.addr.type) {
392 	case PF_ADDR_DYNIFTL:
393 		pfi_dynaddr_remove(rule->dst.addr.p.dyn);
394 		break;
395 	case PF_ADDR_TABLE:
396 		pfr_detach_table(rule->dst.addr.p.tbl);
397 		break;
398 	}
399 	if (rule->overload_tbl)
400 		pfr_detach_table(rule->overload_tbl);
401 	if (rule->kif)
402 		pfi_kif_unref(rule->kif);
403 	pf_anchor_remove(rule);
404 	pf_empty_pool(&rule->rpool.list);
405 	counter_u64_free(rule->states_cur);
406 	counter_u64_free(rule->states_tot);
407 	counter_u64_free(rule->src_nodes);
408 	free(rule, M_PFRULE);
409 }
410 
411 static u_int16_t
412 tagname2tag(struct pf_tags *head, char *tagname)
413 {
414 	struct pf_tagname	*tag, *p = NULL;
415 	u_int16_t		 new_tagid = 1;
416 
417 	PF_RULES_WASSERT();
418 
419 	TAILQ_FOREACH(tag, head, entries)
420 		if (strcmp(tagname, tag->name) == 0) {
421 			tag->ref++;
422 			return (tag->tag);
423 		}
424 
425 	/*
426 	 * to avoid fragmentation, we do a linear search from the beginning
427 	 * and take the first free slot we find. if there is none or the list
428 	 * is empty, append a new entry at the end.
429 	 */
430 
431 	/* new entry */
432 	if (!TAILQ_EMPTY(head))
433 		for (p = TAILQ_FIRST(head); p != NULL &&
434 		    p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
435 			new_tagid = p->tag + 1;
436 
437 	if (new_tagid > TAGID_MAX)
438 		return (0);
439 
440 	/* allocate and fill new struct pf_tagname */
441 	tag = malloc(sizeof(*tag), M_PFTAG, M_NOWAIT|M_ZERO);
442 	if (tag == NULL)
443 		return (0);
444 	strlcpy(tag->name, tagname, sizeof(tag->name));
445 	tag->tag = new_tagid;
446 	tag->ref++;
447 
448 	if (p != NULL)	/* insert new entry before p */
449 		TAILQ_INSERT_BEFORE(p, tag, entries);
450 	else	/* either list empty or no free slot in between */
451 		TAILQ_INSERT_TAIL(head, tag, entries);
452 
453 	return (tag->tag);
454 }
455 
456 static void
457 tag_unref(struct pf_tags *head, u_int16_t tag)
458 {
459 	struct pf_tagname	*p, *next;
460 
461 	PF_RULES_WASSERT();
462 
463 	for (p = TAILQ_FIRST(head); p != NULL; p = next) {
464 		next = TAILQ_NEXT(p, entries);
465 		if (tag == p->tag) {
466 			if (--p->ref == 0) {
467 				TAILQ_REMOVE(head, p, entries);
468 				free(p, M_PFTAG);
469 			}
470 			break;
471 		}
472 	}
473 }
474 
475 static u_int16_t
476 pf_tagname2tag(char *tagname)
477 {
478 	return (tagname2tag(&V_pf_tags, tagname));
479 }
480 
481 #ifdef ALTQ
482 static u_int32_t
483 pf_qname2qid(char *qname)
484 {
485 	return ((u_int32_t)tagname2tag(&V_pf_qids, qname));
486 }
487 
488 static void
489 pf_qid_unref(u_int32_t qid)
490 {
491 	tag_unref(&V_pf_qids, (u_int16_t)qid);
492 }
493 
494 static int
495 pf_begin_altq(u_int32_t *ticket)
496 {
497 	struct pf_altq	*altq;
498 	int		 error = 0;
499 
500 	PF_RULES_WASSERT();
501 
502 	/* Purge the old altq list */
503 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
504 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
505 		if (altq->qname[0] == 0 &&
506 		    (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
507 			/* detach and destroy the discipline */
508 			error = altq_remove(altq);
509 		} else
510 			pf_qid_unref(altq->qid);
511 		free(altq, M_PFALTQ);
512 	}
513 	if (error)
514 		return (error);
515 	*ticket = ++V_ticket_altqs_inactive;
516 	V_altqs_inactive_open = 1;
517 	return (0);
518 }
519 
520 static int
521 pf_rollback_altq(u_int32_t ticket)
522 {
523 	struct pf_altq	*altq;
524 	int		 error = 0;
525 
526 	PF_RULES_WASSERT();
527 
528 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
529 		return (0);
530 	/* Purge the old altq list */
531 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
532 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
533 		if (altq->qname[0] == 0 &&
534 		   (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
535 			/* detach and destroy the discipline */
536 			error = altq_remove(altq);
537 		} else
538 			pf_qid_unref(altq->qid);
539 		free(altq, M_PFALTQ);
540 	}
541 	V_altqs_inactive_open = 0;
542 	return (error);
543 }
544 
545 static int
546 pf_commit_altq(u_int32_t ticket)
547 {
548 	struct pf_altqqueue	*old_altqs;
549 	struct pf_altq		*altq;
550 	int			 err, error = 0;
551 
552 	PF_RULES_WASSERT();
553 
554 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
555 		return (EBUSY);
556 
557 	/* swap altqs, keep the old. */
558 	old_altqs = V_pf_altqs_active;
559 	V_pf_altqs_active = V_pf_altqs_inactive;
560 	V_pf_altqs_inactive = old_altqs;
561 	V_ticket_altqs_active = V_ticket_altqs_inactive;
562 
563 	/* Attach new disciplines */
564 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
565 	if (altq->qname[0] == 0 &&
566 	   (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
567 			/* attach the discipline */
568 			error = altq_pfattach(altq);
569 			if (error == 0 && V_pf_altq_running)
570 				error = pf_enable_altq(altq);
571 			if (error != 0)
572 				return (error);
573 		}
574 	}
575 
576 	/* Purge the old altq list */
577 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
578 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
579 		if (altq->qname[0] == 0 &&
580 		    (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
581 			/* detach and destroy the discipline */
582 			if (V_pf_altq_running)
583 				error = pf_disable_altq(altq);
584 			err = altq_pfdetach(altq);
585 			if (err != 0 && error == 0)
586 				error = err;
587 			err = altq_remove(altq);
588 			if (err != 0 && error == 0)
589 				error = err;
590 		} else
591 			pf_qid_unref(altq->qid);
592 		free(altq, M_PFALTQ);
593 	}
594 
595 	V_altqs_inactive_open = 0;
596 	return (error);
597 }
598 
599 static int
600 pf_enable_altq(struct pf_altq *altq)
601 {
602 	struct ifnet		*ifp;
603 	struct tb_profile	 tb;
604 	int			 error = 0;
605 
606 	if ((ifp = ifunit(altq->ifname)) == NULL)
607 		return (EINVAL);
608 
609 	if (ifp->if_snd.altq_type != ALTQT_NONE)
610 		error = altq_enable(&ifp->if_snd);
611 
612 	/* set tokenbucket regulator */
613 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
614 		tb.rate = altq->ifbandwidth;
615 		tb.depth = altq->tbrsize;
616 		error = tbr_set(&ifp->if_snd, &tb);
617 	}
618 
619 	return (error);
620 }
621 
622 static int
623 pf_disable_altq(struct pf_altq *altq)
624 {
625 	struct ifnet		*ifp;
626 	struct tb_profile	 tb;
627 	int			 error;
628 
629 	if ((ifp = ifunit(altq->ifname)) == NULL)
630 		return (EINVAL);
631 
632 	/*
633 	 * when the discipline is no longer referenced, it was overridden
634 	 * by a new one.  if so, just return.
635 	 */
636 	if (altq->altq_disc != ifp->if_snd.altq_disc)
637 		return (0);
638 
639 	error = altq_disable(&ifp->if_snd);
640 
641 	if (error == 0) {
642 		/* clear tokenbucket regulator */
643 		tb.rate = 0;
644 		error = tbr_set(&ifp->if_snd, &tb);
645 	}
646 
647 	return (error);
648 }
649 
650 void
651 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
652 {
653 	struct ifnet	*ifp1;
654 	struct pf_altq	*a1, *a2, *a3;
655 	u_int32_t	 ticket;
656 	int		 error = 0;
657 
658 	/* Interrupt userland queue modifications */
659 	if (V_altqs_inactive_open)
660 		pf_rollback_altq(V_ticket_altqs_inactive);
661 
662 	/* Start new altq ruleset */
663 	if (pf_begin_altq(&ticket))
664 		return;
665 
666 	/* Copy the current active set */
667 	TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
668 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
669 		if (a2 == NULL) {
670 			error = ENOMEM;
671 			break;
672 		}
673 		bcopy(a1, a2, sizeof(struct pf_altq));
674 
675 		if (a2->qname[0] != 0) {
676 			if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
677 				error = EBUSY;
678 				free(a2, M_PFALTQ);
679 				break;
680 			}
681 			a2->altq_disc = NULL;
682 			TAILQ_FOREACH(a3, V_pf_altqs_inactive, entries) {
683 				if (strncmp(a3->ifname, a2->ifname,
684 				    IFNAMSIZ) == 0 && a3->qname[0] == 0) {
685 					a2->altq_disc = a3->altq_disc;
686 					break;
687 				}
688 			}
689 		}
690 		/* Deactivate the interface in question */
691 		a2->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
692 		if ((ifp1 = ifunit(a2->ifname)) == NULL ||
693 		    (remove && ifp1 == ifp)) {
694 			a2->local_flags |= PFALTQ_FLAG_IF_REMOVED;
695 		} else {
696 			error = altq_add(a2);
697 
698 			if (ticket != V_ticket_altqs_inactive)
699 				error = EBUSY;
700 
701 			if (error) {
702 				free(a2, M_PFALTQ);
703 				break;
704 			}
705 		}
706 
707 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
708 	}
709 
710 	if (error != 0)
711 		pf_rollback_altq(ticket);
712 	else
713 		pf_commit_altq(ticket);
714 }
715 #endif /* ALTQ */
716 
717 static int
718 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
719 {
720 	struct pf_ruleset	*rs;
721 	struct pf_rule		*rule;
722 
723 	PF_RULES_WASSERT();
724 
725 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
726 		return (EINVAL);
727 	rs = pf_find_or_create_ruleset(anchor);
728 	if (rs == NULL)
729 		return (EINVAL);
730 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
731 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
732 		rs->rules[rs_num].inactive.rcount--;
733 	}
734 	*ticket = ++rs->rules[rs_num].inactive.ticket;
735 	rs->rules[rs_num].inactive.open = 1;
736 	return (0);
737 }
738 
739 static int
740 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
741 {
742 	struct pf_ruleset	*rs;
743 	struct pf_rule		*rule;
744 
745 	PF_RULES_WASSERT();
746 
747 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
748 		return (EINVAL);
749 	rs = pf_find_ruleset(anchor);
750 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
751 	    rs->rules[rs_num].inactive.ticket != ticket)
752 		return (0);
753 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
754 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
755 		rs->rules[rs_num].inactive.rcount--;
756 	}
757 	rs->rules[rs_num].inactive.open = 0;
758 	return (0);
759 }
760 
761 #define PF_MD5_UPD(st, elm)						\
762 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
763 
764 #define PF_MD5_UPD_STR(st, elm)						\
765 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
766 
767 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
768 		(stor) = htonl((st)->elm);				\
769 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
770 } while (0)
771 
772 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
773 		(stor) = htons((st)->elm);				\
774 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
775 } while (0)
776 
777 static void
778 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
779 {
780 	PF_MD5_UPD(pfr, addr.type);
781 	switch (pfr->addr.type) {
782 		case PF_ADDR_DYNIFTL:
783 			PF_MD5_UPD(pfr, addr.v.ifname);
784 			PF_MD5_UPD(pfr, addr.iflags);
785 			break;
786 		case PF_ADDR_TABLE:
787 			PF_MD5_UPD(pfr, addr.v.tblname);
788 			break;
789 		case PF_ADDR_ADDRMASK:
790 			/* XXX ignore af? */
791 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
792 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
793 			break;
794 	}
795 
796 	PF_MD5_UPD(pfr, port[0]);
797 	PF_MD5_UPD(pfr, port[1]);
798 	PF_MD5_UPD(pfr, neg);
799 	PF_MD5_UPD(pfr, port_op);
800 }
801 
802 static void
803 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
804 {
805 	u_int16_t x;
806 	u_int32_t y;
807 
808 	pf_hash_rule_addr(ctx, &rule->src);
809 	pf_hash_rule_addr(ctx, &rule->dst);
810 	PF_MD5_UPD_STR(rule, label);
811 	PF_MD5_UPD_STR(rule, ifname);
812 	PF_MD5_UPD_STR(rule, match_tagname);
813 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
814 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
815 	PF_MD5_UPD_HTONL(rule, prob, y);
816 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
817 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
818 	PF_MD5_UPD(rule, uid.op);
819 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
820 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
821 	PF_MD5_UPD(rule, gid.op);
822 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
823 	PF_MD5_UPD(rule, action);
824 	PF_MD5_UPD(rule, direction);
825 	PF_MD5_UPD(rule, af);
826 	PF_MD5_UPD(rule, quick);
827 	PF_MD5_UPD(rule, ifnot);
828 	PF_MD5_UPD(rule, match_tag_not);
829 	PF_MD5_UPD(rule, natpass);
830 	PF_MD5_UPD(rule, keep_state);
831 	PF_MD5_UPD(rule, proto);
832 	PF_MD5_UPD(rule, type);
833 	PF_MD5_UPD(rule, code);
834 	PF_MD5_UPD(rule, flags);
835 	PF_MD5_UPD(rule, flagset);
836 	PF_MD5_UPD(rule, allow_opts);
837 	PF_MD5_UPD(rule, rt);
838 	PF_MD5_UPD(rule, tos);
839 }
840 
841 static int
842 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
843 {
844 	struct pf_ruleset	*rs;
845 	struct pf_rule		*rule, **old_array;
846 	struct pf_rulequeue	*old_rules;
847 	int			 error;
848 	u_int32_t		 old_rcount;
849 
850 	PF_RULES_WASSERT();
851 
852 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
853 		return (EINVAL);
854 	rs = pf_find_ruleset(anchor);
855 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
856 	    ticket != rs->rules[rs_num].inactive.ticket)
857 		return (EBUSY);
858 
859 	/* Calculate checksum for the main ruleset */
860 	if (rs == &pf_main_ruleset) {
861 		error = pf_setup_pfsync_matching(rs);
862 		if (error != 0)
863 			return (error);
864 	}
865 
866 	/* Swap rules, keep the old. */
867 	old_rules = rs->rules[rs_num].active.ptr;
868 	old_rcount = rs->rules[rs_num].active.rcount;
869 	old_array = rs->rules[rs_num].active.ptr_array;
870 
871 	rs->rules[rs_num].active.ptr =
872 	    rs->rules[rs_num].inactive.ptr;
873 	rs->rules[rs_num].active.ptr_array =
874 	    rs->rules[rs_num].inactive.ptr_array;
875 	rs->rules[rs_num].active.rcount =
876 	    rs->rules[rs_num].inactive.rcount;
877 	rs->rules[rs_num].inactive.ptr = old_rules;
878 	rs->rules[rs_num].inactive.ptr_array = old_array;
879 	rs->rules[rs_num].inactive.rcount = old_rcount;
880 
881 	rs->rules[rs_num].active.ticket =
882 	    rs->rules[rs_num].inactive.ticket;
883 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
884 
885 
886 	/* Purge the old rule list. */
887 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
888 		pf_unlink_rule(old_rules, rule);
889 	if (rs->rules[rs_num].inactive.ptr_array)
890 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
891 	rs->rules[rs_num].inactive.ptr_array = NULL;
892 	rs->rules[rs_num].inactive.rcount = 0;
893 	rs->rules[rs_num].inactive.open = 0;
894 	pf_remove_if_empty_ruleset(rs);
895 
896 	return (0);
897 }
898 
899 static int
900 pf_setup_pfsync_matching(struct pf_ruleset *rs)
901 {
902 	MD5_CTX			 ctx;
903 	struct pf_rule		*rule;
904 	int			 rs_cnt;
905 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
906 
907 	MD5Init(&ctx);
908 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
909 		/* XXX PF_RULESET_SCRUB as well? */
910 		if (rs_cnt == PF_RULESET_SCRUB)
911 			continue;
912 
913 		if (rs->rules[rs_cnt].inactive.ptr_array)
914 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
915 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
916 
917 		if (rs->rules[rs_cnt].inactive.rcount) {
918 			rs->rules[rs_cnt].inactive.ptr_array =
919 			    malloc(sizeof(caddr_t) *
920 			    rs->rules[rs_cnt].inactive.rcount,
921 			    M_TEMP, M_NOWAIT);
922 
923 			if (!rs->rules[rs_cnt].inactive.ptr_array)
924 				return (ENOMEM);
925 		}
926 
927 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
928 		    entries) {
929 			pf_hash_rule(&ctx, rule);
930 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
931 		}
932 	}
933 
934 	MD5Final(digest, &ctx);
935 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
936 	return (0);
937 }
938 
939 static int
940 pf_addr_setup(struct pf_ruleset *ruleset, struct pf_addr_wrap *addr,
941     sa_family_t af)
942 {
943 	int error = 0;
944 
945 	switch (addr->type) {
946 	case PF_ADDR_TABLE:
947 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
948 		if (addr->p.tbl == NULL)
949 			error = ENOMEM;
950 		break;
951 	case PF_ADDR_DYNIFTL:
952 		error = pfi_dynaddr_setup(addr, af);
953 		break;
954 	}
955 
956 	return (error);
957 }
958 
959 static void
960 pf_addr_copyout(struct pf_addr_wrap *addr)
961 {
962 
963 	switch (addr->type) {
964 	case PF_ADDR_DYNIFTL:
965 		pfi_dynaddr_copyout(addr);
966 		break;
967 	case PF_ADDR_TABLE:
968 		pf_tbladdr_copyout(addr);
969 		break;
970 	}
971 }
972 
973 static int
974 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
975 {
976 	int			 error = 0;
977 
978 	/* XXX keep in sync with switch() below */
979 	if (securelevel_gt(td->td_ucred, 2))
980 		switch (cmd) {
981 		case DIOCGETRULES:
982 		case DIOCGETRULE:
983 		case DIOCGETADDRS:
984 		case DIOCGETADDR:
985 		case DIOCGETSTATE:
986 		case DIOCSETSTATUSIF:
987 		case DIOCGETSTATUS:
988 		case DIOCCLRSTATUS:
989 		case DIOCNATLOOK:
990 		case DIOCSETDEBUG:
991 		case DIOCGETSTATES:
992 		case DIOCGETTIMEOUT:
993 		case DIOCCLRRULECTRS:
994 		case DIOCGETLIMIT:
995 		case DIOCGETALTQS:
996 		case DIOCGETALTQ:
997 		case DIOCGETQSTATS:
998 		case DIOCGETRULESETS:
999 		case DIOCGETRULESET:
1000 		case DIOCRGETTABLES:
1001 		case DIOCRGETTSTATS:
1002 		case DIOCRCLRTSTATS:
1003 		case DIOCRCLRADDRS:
1004 		case DIOCRADDADDRS:
1005 		case DIOCRDELADDRS:
1006 		case DIOCRSETADDRS:
1007 		case DIOCRGETADDRS:
1008 		case DIOCRGETASTATS:
1009 		case DIOCRCLRASTATS:
1010 		case DIOCRTSTADDRS:
1011 		case DIOCOSFPGET:
1012 		case DIOCGETSRCNODES:
1013 		case DIOCCLRSRCNODES:
1014 		case DIOCIGETIFACES:
1015 		case DIOCGIFSPEED:
1016 		case DIOCSETIFFLAG:
1017 		case DIOCCLRIFFLAG:
1018 			break;
1019 		case DIOCRCLRTABLES:
1020 		case DIOCRADDTABLES:
1021 		case DIOCRDELTABLES:
1022 		case DIOCRSETTFLAGS:
1023 			if (((struct pfioc_table *)addr)->pfrio_flags &
1024 			    PFR_FLAG_DUMMY)
1025 				break; /* dummy operation ok */
1026 			return (EPERM);
1027 		default:
1028 			return (EPERM);
1029 		}
1030 
1031 	if (!(flags & FWRITE))
1032 		switch (cmd) {
1033 		case DIOCGETRULES:
1034 		case DIOCGETADDRS:
1035 		case DIOCGETADDR:
1036 		case DIOCGETSTATE:
1037 		case DIOCGETSTATUS:
1038 		case DIOCGETSTATES:
1039 		case DIOCGETTIMEOUT:
1040 		case DIOCGETLIMIT:
1041 		case DIOCGETALTQS:
1042 		case DIOCGETALTQ:
1043 		case DIOCGETQSTATS:
1044 		case DIOCGETRULESETS:
1045 		case DIOCGETRULESET:
1046 		case DIOCNATLOOK:
1047 		case DIOCRGETTABLES:
1048 		case DIOCRGETTSTATS:
1049 		case DIOCRGETADDRS:
1050 		case DIOCRGETASTATS:
1051 		case DIOCRTSTADDRS:
1052 		case DIOCOSFPGET:
1053 		case DIOCGETSRCNODES:
1054 		case DIOCIGETIFACES:
1055 		case DIOCGIFSPEED:
1056 			break;
1057 		case DIOCRCLRTABLES:
1058 		case DIOCRADDTABLES:
1059 		case DIOCRDELTABLES:
1060 		case DIOCRCLRTSTATS:
1061 		case DIOCRCLRADDRS:
1062 		case DIOCRADDADDRS:
1063 		case DIOCRDELADDRS:
1064 		case DIOCRSETADDRS:
1065 		case DIOCRSETTFLAGS:
1066 			if (((struct pfioc_table *)addr)->pfrio_flags &
1067 			    PFR_FLAG_DUMMY) {
1068 				flags |= FWRITE; /* need write lock for dummy */
1069 				break; /* dummy operation ok */
1070 			}
1071 			return (EACCES);
1072 		case DIOCGETRULE:
1073 			if (((struct pfioc_rule *)addr)->action ==
1074 			    PF_GET_CLR_CNTR)
1075 				return (EACCES);
1076 			break;
1077 		default:
1078 			return (EACCES);
1079 		}
1080 
1081 	CURVNET_SET(TD_TO_VNET(td));
1082 
1083 	switch (cmd) {
1084 	case DIOCSTART:
1085 		PF_RULES_WLOCK();
1086 		if (V_pf_status.running)
1087 			error = EEXIST;
1088 		else {
1089 			int cpu;
1090 
1091 			PF_RULES_WUNLOCK();
1092 			error = hook_pf();
1093 			if (error) {
1094 				DPFPRINTF(PF_DEBUG_MISC,
1095 				    ("pf: pfil registration failed\n"));
1096 				break;
1097 			}
1098 			PF_RULES_WLOCK();
1099 			V_pf_status.running = 1;
1100 			V_pf_status.since = time_second;
1101 
1102 			CPU_FOREACH(cpu)
1103 				V_pf_stateid[cpu] = time_second;
1104 
1105 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1106 		}
1107 		PF_RULES_WUNLOCK();
1108 		break;
1109 
1110 	case DIOCSTOP:
1111 		PF_RULES_WLOCK();
1112 		if (!V_pf_status.running)
1113 			error = ENOENT;
1114 		else {
1115 			V_pf_status.running = 0;
1116 			PF_RULES_WUNLOCK();
1117 			error = dehook_pf();
1118 			if (error) {
1119 				V_pf_status.running = 1;
1120 				DPFPRINTF(PF_DEBUG_MISC,
1121 				    ("pf: pfil unregistration failed\n"));
1122 			}
1123 			PF_RULES_WLOCK();
1124 			V_pf_status.since = time_second;
1125 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1126 		}
1127 		PF_RULES_WUNLOCK();
1128 		break;
1129 
1130 	case DIOCADDRULE: {
1131 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1132 		struct pf_ruleset	*ruleset;
1133 		struct pf_rule		*rule, *tail;
1134 		struct pf_pooladdr	*pa;
1135 		struct pfi_kif		*kif = NULL;
1136 		int			 rs_num;
1137 
1138 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1139 			error = EINVAL;
1140 			break;
1141 		}
1142 #ifndef INET
1143 		if (pr->rule.af == AF_INET) {
1144 			error = EAFNOSUPPORT;
1145 			break;
1146 		}
1147 #endif /* INET */
1148 #ifndef INET6
1149 		if (pr->rule.af == AF_INET6) {
1150 			error = EAFNOSUPPORT;
1151 			break;
1152 		}
1153 #endif /* INET6 */
1154 
1155 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
1156 		bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1157 		if (rule->ifname[0])
1158 			kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1159 		rule->states_cur = counter_u64_alloc(M_WAITOK);
1160 		rule->states_tot = counter_u64_alloc(M_WAITOK);
1161 		rule->src_nodes = counter_u64_alloc(M_WAITOK);
1162 		rule->cuid = td->td_ucred->cr_ruid;
1163 		rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1164 		TAILQ_INIT(&rule->rpool.list);
1165 
1166 #define	ERROUT(x)	{ error = (x); goto DIOCADDRULE_error; }
1167 
1168 		PF_RULES_WLOCK();
1169 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1170 		ruleset = pf_find_ruleset(pr->anchor);
1171 		if (ruleset == NULL)
1172 			ERROUT(EINVAL);
1173 		rs_num = pf_get_ruleset_number(pr->rule.action);
1174 		if (rs_num >= PF_RULESET_MAX)
1175 			ERROUT(EINVAL);
1176 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1177 			DPFPRINTF(PF_DEBUG_MISC,
1178 			    ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
1179 			    ruleset->rules[rs_num].inactive.ticket));
1180 			ERROUT(EBUSY);
1181 		}
1182 		if (pr->pool_ticket != V_ticket_pabuf) {
1183 			DPFPRINTF(PF_DEBUG_MISC,
1184 			    ("pool_ticket: %d != %d\n", pr->pool_ticket,
1185 			    V_ticket_pabuf));
1186 			ERROUT(EBUSY);
1187 		}
1188 
1189 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1190 		    pf_rulequeue);
1191 		if (tail)
1192 			rule->nr = tail->nr + 1;
1193 		else
1194 			rule->nr = 0;
1195 		if (rule->ifname[0]) {
1196 			rule->kif = pfi_kif_attach(kif, rule->ifname);
1197 			pfi_kif_ref(rule->kif);
1198 		} else
1199 			rule->kif = NULL;
1200 
1201 		if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
1202 			error = EBUSY;
1203 
1204 #ifdef ALTQ
1205 		/* set queue IDs */
1206 		if (rule->qname[0] != 0) {
1207 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1208 				error = EBUSY;
1209 			else if (rule->pqname[0] != 0) {
1210 				if ((rule->pqid =
1211 				    pf_qname2qid(rule->pqname)) == 0)
1212 					error = EBUSY;
1213 			} else
1214 				rule->pqid = rule->qid;
1215 		}
1216 #endif
1217 		if (rule->tagname[0])
1218 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1219 				error = EBUSY;
1220 		if (rule->match_tagname[0])
1221 			if ((rule->match_tag =
1222 			    pf_tagname2tag(rule->match_tagname)) == 0)
1223 				error = EBUSY;
1224 		if (rule->rt && !rule->direction)
1225 			error = EINVAL;
1226 		if (!rule->log)
1227 			rule->logif = 0;
1228 		if (rule->logif >= PFLOGIFS_MAX)
1229 			error = EINVAL;
1230 		if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1231 			error = ENOMEM;
1232 		if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1233 			error = ENOMEM;
1234 		if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1235 			error = EINVAL;
1236 		TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1237 			if (pa->addr.type == PF_ADDR_TABLE) {
1238 				pa->addr.p.tbl = pfr_attach_table(ruleset,
1239 				    pa->addr.v.tblname);
1240 				if (pa->addr.p.tbl == NULL)
1241 					error = ENOMEM;
1242 			}
1243 
1244 		if (rule->overload_tblname[0]) {
1245 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1246 			    rule->overload_tblname)) == NULL)
1247 				error = EINVAL;
1248 			else
1249 				rule->overload_tbl->pfrkt_flags |=
1250 				    PFR_TFLAG_ACTIVE;
1251 		}
1252 
1253 		pf_mv_pool(&V_pf_pabuf, &rule->rpool.list);
1254 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1255 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1256 		    (rule->rt > PF_FASTROUTE)) &&
1257 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1258 			error = EINVAL;
1259 
1260 		if (error) {
1261 			pf_free_rule(rule);
1262 			PF_RULES_WUNLOCK();
1263 			break;
1264 		}
1265 
1266 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1267 		rule->evaluations = rule->packets[0] = rule->packets[1] =
1268 		    rule->bytes[0] = rule->bytes[1] = 0;
1269 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1270 		    rule, entries);
1271 		ruleset->rules[rs_num].inactive.rcount++;
1272 		PF_RULES_WUNLOCK();
1273 		break;
1274 
1275 #undef ERROUT
1276 DIOCADDRULE_error:
1277 		PF_RULES_WUNLOCK();
1278 		counter_u64_free(rule->states_cur);
1279 		counter_u64_free(rule->states_tot);
1280 		counter_u64_free(rule->src_nodes);
1281 		free(rule, M_PFRULE);
1282 		if (kif)
1283 			free(kif, PFI_MTYPE);
1284 		break;
1285 	}
1286 
1287 	case DIOCGETRULES: {
1288 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1289 		struct pf_ruleset	*ruleset;
1290 		struct pf_rule		*tail;
1291 		int			 rs_num;
1292 
1293 		PF_RULES_WLOCK();
1294 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1295 		ruleset = pf_find_ruleset(pr->anchor);
1296 		if (ruleset == NULL) {
1297 			PF_RULES_WUNLOCK();
1298 			error = EINVAL;
1299 			break;
1300 		}
1301 		rs_num = pf_get_ruleset_number(pr->rule.action);
1302 		if (rs_num >= PF_RULESET_MAX) {
1303 			PF_RULES_WUNLOCK();
1304 			error = EINVAL;
1305 			break;
1306 		}
1307 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1308 		    pf_rulequeue);
1309 		if (tail)
1310 			pr->nr = tail->nr + 1;
1311 		else
1312 			pr->nr = 0;
1313 		pr->ticket = ruleset->rules[rs_num].active.ticket;
1314 		PF_RULES_WUNLOCK();
1315 		break;
1316 	}
1317 
1318 	case DIOCGETRULE: {
1319 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1320 		struct pf_ruleset	*ruleset;
1321 		struct pf_rule		*rule;
1322 		int			 rs_num, i;
1323 
1324 		PF_RULES_WLOCK();
1325 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1326 		ruleset = pf_find_ruleset(pr->anchor);
1327 		if (ruleset == NULL) {
1328 			PF_RULES_WUNLOCK();
1329 			error = EINVAL;
1330 			break;
1331 		}
1332 		rs_num = pf_get_ruleset_number(pr->rule.action);
1333 		if (rs_num >= PF_RULESET_MAX) {
1334 			PF_RULES_WUNLOCK();
1335 			error = EINVAL;
1336 			break;
1337 		}
1338 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1339 			PF_RULES_WUNLOCK();
1340 			error = EBUSY;
1341 			break;
1342 		}
1343 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1344 		while ((rule != NULL) && (rule->nr != pr->nr))
1345 			rule = TAILQ_NEXT(rule, entries);
1346 		if (rule == NULL) {
1347 			PF_RULES_WUNLOCK();
1348 			error = EBUSY;
1349 			break;
1350 		}
1351 		bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1352 		/*
1353 		 * XXXGL: this is what happens when internal kernel
1354 		 * structures are used as ioctl API structures.
1355 		 */
1356 		pr->rule.states_cur =
1357 		    (counter_u64_t )counter_u64_fetch(rule->states_cur);
1358 		pr->rule.states_tot =
1359 		    (counter_u64_t )counter_u64_fetch(rule->states_tot);
1360 		pr->rule.src_nodes =
1361 		    (counter_u64_t )counter_u64_fetch(rule->src_nodes);
1362 		if (pf_anchor_copyout(ruleset, rule, pr)) {
1363 			PF_RULES_WUNLOCK();
1364 			error = EBUSY;
1365 			break;
1366 		}
1367 		pf_addr_copyout(&pr->rule.src.addr);
1368 		pf_addr_copyout(&pr->rule.dst.addr);
1369 		for (i = 0; i < PF_SKIP_COUNT; ++i)
1370 			if (rule->skip[i].ptr == NULL)
1371 				pr->rule.skip[i].nr = -1;
1372 			else
1373 				pr->rule.skip[i].nr =
1374 				    rule->skip[i].ptr->nr;
1375 
1376 		if (pr->action == PF_GET_CLR_CNTR) {
1377 			rule->evaluations = 0;
1378 			rule->packets[0] = rule->packets[1] = 0;
1379 			rule->bytes[0] = rule->bytes[1] = 0;
1380 			counter_u64_zero(rule->states_tot);
1381 		}
1382 		PF_RULES_WUNLOCK();
1383 		break;
1384 	}
1385 
1386 	case DIOCCHANGERULE: {
1387 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1388 		struct pf_ruleset	*ruleset;
1389 		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1390 		struct pfi_kif		*kif = NULL;
1391 		struct pf_pooladdr	*pa;
1392 		u_int32_t		 nr = 0;
1393 		int			 rs_num;
1394 
1395 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1396 		    pcr->action > PF_CHANGE_GET_TICKET) {
1397 			error = EINVAL;
1398 			break;
1399 		}
1400 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1401 			error = EINVAL;
1402 			break;
1403 		}
1404 
1405 		if (pcr->action != PF_CHANGE_REMOVE) {
1406 #ifndef INET
1407 			if (pcr->rule.af == AF_INET) {
1408 				error = EAFNOSUPPORT;
1409 				break;
1410 			}
1411 #endif /* INET */
1412 #ifndef INET6
1413 			if (pcr->rule.af == AF_INET6) {
1414 				error = EAFNOSUPPORT;
1415 				break;
1416 			}
1417 #endif /* INET6 */
1418 			newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
1419 			bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1420 			if (newrule->ifname[0])
1421 				kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1422 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
1423 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
1424 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
1425 			newrule->cuid = td->td_ucred->cr_ruid;
1426 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1427 			TAILQ_INIT(&newrule->rpool.list);
1428 		}
1429 
1430 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGERULE_error; }
1431 
1432 		PF_RULES_WLOCK();
1433 		if (!(pcr->action == PF_CHANGE_REMOVE ||
1434 		    pcr->action == PF_CHANGE_GET_TICKET) &&
1435 		    pcr->pool_ticket != V_ticket_pabuf)
1436 			ERROUT(EBUSY);
1437 
1438 		ruleset = pf_find_ruleset(pcr->anchor);
1439 		if (ruleset == NULL)
1440 			ERROUT(EINVAL);
1441 
1442 		rs_num = pf_get_ruleset_number(pcr->rule.action);
1443 		if (rs_num >= PF_RULESET_MAX)
1444 			ERROUT(EINVAL);
1445 
1446 		if (pcr->action == PF_CHANGE_GET_TICKET) {
1447 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1448 			ERROUT(0);
1449 		} else if (pcr->ticket !=
1450 			    ruleset->rules[rs_num].active.ticket)
1451 				ERROUT(EINVAL);
1452 
1453 		if (pcr->action != PF_CHANGE_REMOVE) {
1454 			if (newrule->ifname[0]) {
1455 				newrule->kif = pfi_kif_attach(kif,
1456 				    newrule->ifname);
1457 				pfi_kif_ref(newrule->kif);
1458 			} else
1459 				newrule->kif = NULL;
1460 
1461 			if (newrule->rtableid > 0 &&
1462 			    newrule->rtableid >= rt_numfibs)
1463 				error = EBUSY;
1464 
1465 #ifdef ALTQ
1466 			/* set queue IDs */
1467 			if (newrule->qname[0] != 0) {
1468 				if ((newrule->qid =
1469 				    pf_qname2qid(newrule->qname)) == 0)
1470 					error = EBUSY;
1471 				else if (newrule->pqname[0] != 0) {
1472 					if ((newrule->pqid =
1473 					    pf_qname2qid(newrule->pqname)) == 0)
1474 						error = EBUSY;
1475 				} else
1476 					newrule->pqid = newrule->qid;
1477 			}
1478 #endif /* ALTQ */
1479 			if (newrule->tagname[0])
1480 				if ((newrule->tag =
1481 				    pf_tagname2tag(newrule->tagname)) == 0)
1482 					error = EBUSY;
1483 			if (newrule->match_tagname[0])
1484 				if ((newrule->match_tag = pf_tagname2tag(
1485 				    newrule->match_tagname)) == 0)
1486 					error = EBUSY;
1487 			if (newrule->rt && !newrule->direction)
1488 				error = EINVAL;
1489 			if (!newrule->log)
1490 				newrule->logif = 0;
1491 			if (newrule->logif >= PFLOGIFS_MAX)
1492 				error = EINVAL;
1493 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
1494 				error = ENOMEM;
1495 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
1496 				error = ENOMEM;
1497 			if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1498 				error = EINVAL;
1499 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1500 				if (pa->addr.type == PF_ADDR_TABLE) {
1501 					pa->addr.p.tbl =
1502 					    pfr_attach_table(ruleset,
1503 					    pa->addr.v.tblname);
1504 					if (pa->addr.p.tbl == NULL)
1505 						error = ENOMEM;
1506 				}
1507 
1508 			if (newrule->overload_tblname[0]) {
1509 				if ((newrule->overload_tbl = pfr_attach_table(
1510 				    ruleset, newrule->overload_tblname)) ==
1511 				    NULL)
1512 					error = EINVAL;
1513 				else
1514 					newrule->overload_tbl->pfrkt_flags |=
1515 					    PFR_TFLAG_ACTIVE;
1516 			}
1517 
1518 			pf_mv_pool(&V_pf_pabuf, &newrule->rpool.list);
1519 			if (((((newrule->action == PF_NAT) ||
1520 			    (newrule->action == PF_RDR) ||
1521 			    (newrule->action == PF_BINAT) ||
1522 			    (newrule->rt > PF_FASTROUTE)) &&
1523 			    !newrule->anchor)) &&
1524 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1525 				error = EINVAL;
1526 
1527 			if (error) {
1528 				pf_free_rule(newrule);
1529 				PF_RULES_WUNLOCK();
1530 				break;
1531 			}
1532 
1533 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1534 			newrule->evaluations = 0;
1535 			newrule->packets[0] = newrule->packets[1] = 0;
1536 			newrule->bytes[0] = newrule->bytes[1] = 0;
1537 		}
1538 		pf_empty_pool(&V_pf_pabuf);
1539 
1540 		if (pcr->action == PF_CHANGE_ADD_HEAD)
1541 			oldrule = TAILQ_FIRST(
1542 			    ruleset->rules[rs_num].active.ptr);
1543 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1544 			oldrule = TAILQ_LAST(
1545 			    ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1546 		else {
1547 			oldrule = TAILQ_FIRST(
1548 			    ruleset->rules[rs_num].active.ptr);
1549 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1550 				oldrule = TAILQ_NEXT(oldrule, entries);
1551 			if (oldrule == NULL) {
1552 				if (newrule != NULL)
1553 					pf_free_rule(newrule);
1554 				PF_RULES_WUNLOCK();
1555 				error = EINVAL;
1556 				break;
1557 			}
1558 		}
1559 
1560 		if (pcr->action == PF_CHANGE_REMOVE) {
1561 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
1562 			    oldrule);
1563 			ruleset->rules[rs_num].active.rcount--;
1564 		} else {
1565 			if (oldrule == NULL)
1566 				TAILQ_INSERT_TAIL(
1567 				    ruleset->rules[rs_num].active.ptr,
1568 				    newrule, entries);
1569 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1570 			    pcr->action == PF_CHANGE_ADD_BEFORE)
1571 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1572 			else
1573 				TAILQ_INSERT_AFTER(
1574 				    ruleset->rules[rs_num].active.ptr,
1575 				    oldrule, newrule, entries);
1576 			ruleset->rules[rs_num].active.rcount++;
1577 		}
1578 
1579 		nr = 0;
1580 		TAILQ_FOREACH(oldrule,
1581 		    ruleset->rules[rs_num].active.ptr, entries)
1582 			oldrule->nr = nr++;
1583 
1584 		ruleset->rules[rs_num].active.ticket++;
1585 
1586 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1587 		pf_remove_if_empty_ruleset(ruleset);
1588 
1589 		PF_RULES_WUNLOCK();
1590 		break;
1591 
1592 #undef ERROUT
1593 DIOCCHANGERULE_error:
1594 		PF_RULES_WUNLOCK();
1595 		if (newrule != NULL) {
1596 			counter_u64_free(newrule->states_cur);
1597 			counter_u64_free(newrule->states_tot);
1598 			counter_u64_free(newrule->src_nodes);
1599 			free(newrule, M_PFRULE);
1600 		}
1601 		if (kif != NULL)
1602 			free(kif, PFI_MTYPE);
1603 		break;
1604 	}
1605 
1606 	case DIOCCLRSTATES: {
1607 		struct pf_state		*s;
1608 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1609 		u_int			 i, killed = 0;
1610 
1611 		for (i = 0; i <= V_pf_hashmask; i++) {
1612 			struct pf_idhash *ih = &V_pf_idhash[i];
1613 
1614 relock_DIOCCLRSTATES:
1615 			PF_HASHROW_LOCK(ih);
1616 			LIST_FOREACH(s, &ih->states, entry)
1617 				if (!psk->psk_ifname[0] ||
1618 				    !strcmp(psk->psk_ifname,
1619 				    s->kif->pfik_name)) {
1620 					/*
1621 					 * Don't send out individual
1622 					 * delete messages.
1623 					 */
1624 					s->state_flags |= PFSTATE_NOSYNC;
1625 					pf_unlink_state(s, PF_ENTER_LOCKED);
1626 					killed++;
1627 					goto relock_DIOCCLRSTATES;
1628 				}
1629 			PF_HASHROW_UNLOCK(ih);
1630 		}
1631 		psk->psk_killed = killed;
1632 		if (pfsync_clear_states_ptr != NULL)
1633 			pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
1634 		break;
1635 	}
1636 
1637 	case DIOCKILLSTATES: {
1638 		struct pf_state		*s;
1639 		struct pf_state_key	*sk;
1640 		struct pf_addr		*srcaddr, *dstaddr;
1641 		u_int16_t		 srcport, dstport;
1642 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
1643 		u_int			 i, killed = 0;
1644 
1645 		if (psk->psk_pfcmp.id) {
1646 			if (psk->psk_pfcmp.creatorid == 0)
1647 				psk->psk_pfcmp.creatorid = V_pf_status.hostid;
1648 			if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
1649 			    psk->psk_pfcmp.creatorid))) {
1650 				pf_unlink_state(s, PF_ENTER_LOCKED);
1651 				psk->psk_killed = 1;
1652 			}
1653 			break;
1654 		}
1655 
1656 		for (i = 0; i <= V_pf_hashmask; i++) {
1657 			struct pf_idhash *ih = &V_pf_idhash[i];
1658 
1659 relock_DIOCKILLSTATES:
1660 			PF_HASHROW_LOCK(ih);
1661 			LIST_FOREACH(s, &ih->states, entry) {
1662 				sk = s->key[PF_SK_WIRE];
1663 				if (s->direction == PF_OUT) {
1664 					srcaddr = &sk->addr[1];
1665 					dstaddr = &sk->addr[0];
1666 					srcport = sk->port[0];
1667 					dstport = sk->port[0];
1668 				} else {
1669 					srcaddr = &sk->addr[0];
1670 					dstaddr = &sk->addr[1];
1671 					srcport = sk->port[0];
1672 					dstport = sk->port[0];
1673 				}
1674 
1675 				if ((!psk->psk_af || sk->af == psk->psk_af)
1676 				    && (!psk->psk_proto || psk->psk_proto ==
1677 				    sk->proto) &&
1678 				    PF_MATCHA(psk->psk_src.neg,
1679 				    &psk->psk_src.addr.v.a.addr,
1680 				    &psk->psk_src.addr.v.a.mask,
1681 				    srcaddr, sk->af) &&
1682 				    PF_MATCHA(psk->psk_dst.neg,
1683 				    &psk->psk_dst.addr.v.a.addr,
1684 				    &psk->psk_dst.addr.v.a.mask,
1685 				    dstaddr, sk->af) &&
1686 				    (psk->psk_src.port_op == 0 ||
1687 				    pf_match_port(psk->psk_src.port_op,
1688 				    psk->psk_src.port[0], psk->psk_src.port[1],
1689 				    srcport)) &&
1690 				    (psk->psk_dst.port_op == 0 ||
1691 				    pf_match_port(psk->psk_dst.port_op,
1692 				    psk->psk_dst.port[0], psk->psk_dst.port[1],
1693 				    dstport)) &&
1694 				    (!psk->psk_label[0] ||
1695 				    (s->rule.ptr->label[0] &&
1696 				    !strcmp(psk->psk_label,
1697 				    s->rule.ptr->label))) &&
1698 				    (!psk->psk_ifname[0] ||
1699 				    !strcmp(psk->psk_ifname,
1700 				    s->kif->pfik_name))) {
1701 					pf_unlink_state(s, PF_ENTER_LOCKED);
1702 					killed++;
1703 					goto relock_DIOCKILLSTATES;
1704 				}
1705 			}
1706 			PF_HASHROW_UNLOCK(ih);
1707 		}
1708 		psk->psk_killed = killed;
1709 		break;
1710 	}
1711 
1712 	case DIOCADDSTATE: {
1713 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1714 		struct pfsync_state	*sp = &ps->state;
1715 
1716 		if (sp->timeout >= PFTM_MAX) {
1717 			error = EINVAL;
1718 			break;
1719 		}
1720 		if (pfsync_state_import_ptr != NULL) {
1721 			PF_RULES_RLOCK();
1722 			error = pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
1723 			PF_RULES_RUNLOCK();
1724 		} else
1725 			error = EOPNOTSUPP;
1726 		break;
1727 	}
1728 
1729 	case DIOCGETSTATE: {
1730 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1731 		struct pf_state		*s;
1732 
1733 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
1734 		if (s == NULL) {
1735 			error = ENOENT;
1736 			break;
1737 		}
1738 
1739 		pfsync_state_export(&ps->state, s);
1740 		PF_STATE_UNLOCK(s);
1741 		break;
1742 	}
1743 
1744 	case DIOCGETSTATES: {
1745 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
1746 		struct pf_state		*s;
1747 		struct pfsync_state	*pstore, *p;
1748 		int i, nr;
1749 
1750 		if (ps->ps_len == 0) {
1751 			nr = uma_zone_get_cur(V_pf_state_z);
1752 			ps->ps_len = sizeof(struct pfsync_state) * nr;
1753 			break;
1754 		}
1755 
1756 		p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK);
1757 		nr = 0;
1758 
1759 		for (i = 0; i <= V_pf_hashmask; i++) {
1760 			struct pf_idhash *ih = &V_pf_idhash[i];
1761 
1762 			PF_HASHROW_LOCK(ih);
1763 			LIST_FOREACH(s, &ih->states, entry) {
1764 
1765 				if (s->timeout == PFTM_UNLINKED)
1766 					continue;
1767 
1768 				if ((nr+1) * sizeof(*p) > ps->ps_len) {
1769 					PF_HASHROW_UNLOCK(ih);
1770 					goto DIOCGETSTATES_full;
1771 				}
1772 				pfsync_state_export(p, s);
1773 				p++;
1774 				nr++;
1775 			}
1776 			PF_HASHROW_UNLOCK(ih);
1777 		}
1778 DIOCGETSTATES_full:
1779 		error = copyout(pstore, ps->ps_states,
1780 		    sizeof(struct pfsync_state) * nr);
1781 		if (error) {
1782 			free(pstore, M_TEMP);
1783 			break;
1784 		}
1785 		ps->ps_len = sizeof(struct pfsync_state) * nr;
1786 		free(pstore, M_TEMP);
1787 
1788 		break;
1789 	}
1790 
1791 	case DIOCGETSTATUS: {
1792 		struct pf_status *s = (struct pf_status *)addr;
1793 		PF_RULES_RLOCK();
1794 		bcopy(&V_pf_status, s, sizeof(struct pf_status));
1795 		pfi_update_status(s->ifname, s);
1796 		PF_RULES_RUNLOCK();
1797 		break;
1798 	}
1799 
1800 	case DIOCSETSTATUSIF: {
1801 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
1802 
1803 		if (pi->ifname[0] == 0) {
1804 			bzero(V_pf_status.ifname, IFNAMSIZ);
1805 			break;
1806 		}
1807 		PF_RULES_WLOCK();
1808 		strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
1809 		PF_RULES_WUNLOCK();
1810 		break;
1811 	}
1812 
1813 	case DIOCCLRSTATUS: {
1814 		PF_RULES_WLOCK();
1815 		bzero(V_pf_status.counters, sizeof(V_pf_status.counters));
1816 		bzero(V_pf_status.fcounters, sizeof(V_pf_status.fcounters));
1817 		bzero(V_pf_status.scounters, sizeof(V_pf_status.scounters));
1818 		V_pf_status.since = time_second;
1819 		if (*V_pf_status.ifname)
1820 			pfi_update_status(V_pf_status.ifname, NULL);
1821 		PF_RULES_WUNLOCK();
1822 		break;
1823 	}
1824 
1825 	case DIOCNATLOOK: {
1826 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
1827 		struct pf_state_key	*sk;
1828 		struct pf_state		*state;
1829 		struct pf_state_key_cmp	 key;
1830 		int			 m = 0, direction = pnl->direction;
1831 		int			 sidx, didx;
1832 
1833 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
1834 		sidx = (direction == PF_IN) ? 1 : 0;
1835 		didx = (direction == PF_IN) ? 0 : 1;
1836 
1837 		if (!pnl->proto ||
1838 		    PF_AZERO(&pnl->saddr, pnl->af) ||
1839 		    PF_AZERO(&pnl->daddr, pnl->af) ||
1840 		    ((pnl->proto == IPPROTO_TCP ||
1841 		    pnl->proto == IPPROTO_UDP) &&
1842 		    (!pnl->dport || !pnl->sport)))
1843 			error = EINVAL;
1844 		else {
1845 			bzero(&key, sizeof(key));
1846 			key.af = pnl->af;
1847 			key.proto = pnl->proto;
1848 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
1849 			key.port[sidx] = pnl->sport;
1850 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
1851 			key.port[didx] = pnl->dport;
1852 
1853 			state = pf_find_state_all(&key, direction, &m);
1854 
1855 			if (m > 1)
1856 				error = E2BIG;	/* more than one state */
1857 			else if (state != NULL) {
1858 				/* XXXGL: not locked read */
1859 				sk = state->key[sidx];
1860 				PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
1861 				pnl->rsport = sk->port[sidx];
1862 				PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
1863 				pnl->rdport = sk->port[didx];
1864 			} else
1865 				error = ENOENT;
1866 		}
1867 		break;
1868 	}
1869 
1870 	case DIOCSETTIMEOUT: {
1871 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1872 		int		 old;
1873 
1874 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
1875 		    pt->seconds < 0) {
1876 			error = EINVAL;
1877 			break;
1878 		}
1879 		PF_RULES_WLOCK();
1880 		old = V_pf_default_rule.timeout[pt->timeout];
1881 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
1882 			pt->seconds = 1;
1883 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
1884 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
1885 			wakeup(pf_purge_thread);
1886 		pt->seconds = old;
1887 		PF_RULES_WUNLOCK();
1888 		break;
1889 	}
1890 
1891 	case DIOCGETTIMEOUT: {
1892 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1893 
1894 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
1895 			error = EINVAL;
1896 			break;
1897 		}
1898 		PF_RULES_RLOCK();
1899 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
1900 		PF_RULES_RUNLOCK();
1901 		break;
1902 	}
1903 
1904 	case DIOCGETLIMIT: {
1905 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1906 
1907 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
1908 			error = EINVAL;
1909 			break;
1910 		}
1911 		PF_RULES_RLOCK();
1912 		pl->limit = V_pf_limits[pl->index].limit;
1913 		PF_RULES_RUNLOCK();
1914 		break;
1915 	}
1916 
1917 	case DIOCSETLIMIT: {
1918 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1919 		int			 old_limit;
1920 
1921 		PF_RULES_WLOCK();
1922 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
1923 		    V_pf_limits[pl->index].zone == NULL) {
1924 			PF_RULES_WUNLOCK();
1925 			error = EINVAL;
1926 			break;
1927 		}
1928 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
1929 		old_limit = V_pf_limits[pl->index].limit;
1930 		V_pf_limits[pl->index].limit = pl->limit;
1931 		pl->limit = old_limit;
1932 		PF_RULES_WUNLOCK();
1933 		break;
1934 	}
1935 
1936 	case DIOCSETDEBUG: {
1937 		u_int32_t	*level = (u_int32_t *)addr;
1938 
1939 		PF_RULES_WLOCK();
1940 		V_pf_status.debug = *level;
1941 		PF_RULES_WUNLOCK();
1942 		break;
1943 	}
1944 
1945 	case DIOCCLRRULECTRS: {
1946 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
1947 		struct pf_ruleset	*ruleset = &pf_main_ruleset;
1948 		struct pf_rule		*rule;
1949 
1950 		PF_RULES_WLOCK();
1951 		TAILQ_FOREACH(rule,
1952 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
1953 			rule->evaluations = 0;
1954 			rule->packets[0] = rule->packets[1] = 0;
1955 			rule->bytes[0] = rule->bytes[1] = 0;
1956 		}
1957 		PF_RULES_WUNLOCK();
1958 		break;
1959 	}
1960 
1961 	case DIOCGIFSPEED: {
1962 		struct pf_ifspeed	*psp = (struct pf_ifspeed *)addr;
1963 		struct pf_ifspeed	ps;
1964 		struct ifnet		*ifp;
1965 
1966 		if (psp->ifname[0] != 0) {
1967 			/* Can we completely trust user-land? */
1968 			strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
1969 			ifp = ifunit(ps.ifname);
1970 			if (ifp != NULL)
1971 				psp->baudrate = ifp->if_baudrate;
1972 			else
1973 				error = EINVAL;
1974 		} else
1975 			error = EINVAL;
1976 		break;
1977 	}
1978 
1979 #ifdef ALTQ
1980 	case DIOCSTARTALTQ: {
1981 		struct pf_altq		*altq;
1982 
1983 		PF_RULES_WLOCK();
1984 		/* enable all altq interfaces on active list */
1985 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1986 			if (altq->qname[0] == 0 && (altq->local_flags &
1987 			    PFALTQ_FLAG_IF_REMOVED) == 0) {
1988 				error = pf_enable_altq(altq);
1989 				if (error != 0)
1990 					break;
1991 			}
1992 		}
1993 		if (error == 0)
1994 			V_pf_altq_running = 1;
1995 		PF_RULES_WUNLOCK();
1996 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
1997 		break;
1998 	}
1999 
2000 	case DIOCSTOPALTQ: {
2001 		struct pf_altq		*altq;
2002 
2003 		PF_RULES_WLOCK();
2004 		/* disable all altq interfaces on active list */
2005 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
2006 			if (altq->qname[0] == 0 && (altq->local_flags &
2007 			    PFALTQ_FLAG_IF_REMOVED) == 0) {
2008 				error = pf_disable_altq(altq);
2009 				if (error != 0)
2010 					break;
2011 			}
2012 		}
2013 		if (error == 0)
2014 			V_pf_altq_running = 0;
2015 		PF_RULES_WUNLOCK();
2016 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2017 		break;
2018 	}
2019 
2020 	case DIOCADDALTQ: {
2021 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2022 		struct pf_altq		*altq, *a;
2023 		struct ifnet		*ifp;
2024 
2025 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK);
2026 		bcopy(&pa->altq, altq, sizeof(struct pf_altq));
2027 		altq->local_flags = 0;
2028 
2029 		PF_RULES_WLOCK();
2030 		if (pa->ticket != V_ticket_altqs_inactive) {
2031 			PF_RULES_WUNLOCK();
2032 			free(altq, M_PFALTQ);
2033 			error = EBUSY;
2034 			break;
2035 		}
2036 
2037 		/*
2038 		 * if this is for a queue, find the discipline and
2039 		 * copy the necessary fields
2040 		 */
2041 		if (altq->qname[0] != 0) {
2042 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2043 				PF_RULES_WUNLOCK();
2044 				error = EBUSY;
2045 				free(altq, M_PFALTQ);
2046 				break;
2047 			}
2048 			altq->altq_disc = NULL;
2049 			TAILQ_FOREACH(a, V_pf_altqs_inactive, entries) {
2050 				if (strncmp(a->ifname, altq->ifname,
2051 				    IFNAMSIZ) == 0 && a->qname[0] == 0) {
2052 					altq->altq_disc = a->altq_disc;
2053 					break;
2054 				}
2055 			}
2056 		}
2057 
2058 		if ((ifp = ifunit(altq->ifname)) == NULL)
2059 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
2060 		else
2061 			error = altq_add(altq);
2062 
2063 		if (error) {
2064 			PF_RULES_WUNLOCK();
2065 			free(altq, M_PFALTQ);
2066 			break;
2067 		}
2068 
2069 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
2070 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2071 		PF_RULES_WUNLOCK();
2072 		break;
2073 	}
2074 
2075 	case DIOCGETALTQS: {
2076 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2077 		struct pf_altq		*altq;
2078 
2079 		PF_RULES_RLOCK();
2080 		pa->nr = 0;
2081 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
2082 			pa->nr++;
2083 		pa->ticket = V_ticket_altqs_active;
2084 		PF_RULES_RUNLOCK();
2085 		break;
2086 	}
2087 
2088 	case DIOCGETALTQ: {
2089 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2090 		struct pf_altq		*altq;
2091 		u_int32_t		 nr;
2092 
2093 		PF_RULES_RLOCK();
2094 		if (pa->ticket != V_ticket_altqs_active) {
2095 			PF_RULES_RUNLOCK();
2096 			error = EBUSY;
2097 			break;
2098 		}
2099 		nr = 0;
2100 		altq = TAILQ_FIRST(V_pf_altqs_active);
2101 		while ((altq != NULL) && (nr < pa->nr)) {
2102 			altq = TAILQ_NEXT(altq, entries);
2103 			nr++;
2104 		}
2105 		if (altq == NULL) {
2106 			PF_RULES_RUNLOCK();
2107 			error = EBUSY;
2108 			break;
2109 		}
2110 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2111 		PF_RULES_RUNLOCK();
2112 		break;
2113 	}
2114 
2115 	case DIOCCHANGEALTQ:
2116 		/* CHANGEALTQ not supported yet! */
2117 		error = ENODEV;
2118 		break;
2119 
2120 	case DIOCGETQSTATS: {
2121 		struct pfioc_qstats	*pq = (struct pfioc_qstats *)addr;
2122 		struct pf_altq		*altq;
2123 		u_int32_t		 nr;
2124 		int			 nbytes;
2125 
2126 		PF_RULES_RLOCK();
2127 		if (pq->ticket != V_ticket_altqs_active) {
2128 			PF_RULES_RUNLOCK();
2129 			error = EBUSY;
2130 			break;
2131 		}
2132 		nbytes = pq->nbytes;
2133 		nr = 0;
2134 		altq = TAILQ_FIRST(V_pf_altqs_active);
2135 		while ((altq != NULL) && (nr < pq->nr)) {
2136 			altq = TAILQ_NEXT(altq, entries);
2137 			nr++;
2138 		}
2139 		if (altq == NULL) {
2140 			PF_RULES_RUNLOCK();
2141 			error = EBUSY;
2142 			break;
2143 		}
2144 
2145 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
2146 			PF_RULES_RUNLOCK();
2147 			error = ENXIO;
2148 			break;
2149 		}
2150 		PF_RULES_RUNLOCK();
2151 		error = altq_getqstats(altq, pq->buf, &nbytes);
2152 		if (error == 0) {
2153 			pq->scheduler = altq->scheduler;
2154 			pq->nbytes = nbytes;
2155 		}
2156 		break;
2157 	}
2158 #endif /* ALTQ */
2159 
2160 	case DIOCBEGINADDRS: {
2161 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2162 
2163 		PF_RULES_WLOCK();
2164 		pf_empty_pool(&V_pf_pabuf);
2165 		pp->ticket = ++V_ticket_pabuf;
2166 		PF_RULES_WUNLOCK();
2167 		break;
2168 	}
2169 
2170 	case DIOCADDADDR: {
2171 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2172 		struct pf_pooladdr	*pa;
2173 		struct pfi_kif		*kif = NULL;
2174 
2175 #ifndef INET
2176 		if (pp->af == AF_INET) {
2177 			error = EAFNOSUPPORT;
2178 			break;
2179 		}
2180 #endif /* INET */
2181 #ifndef INET6
2182 		if (pp->af == AF_INET6) {
2183 			error = EAFNOSUPPORT;
2184 			break;
2185 		}
2186 #endif /* INET6 */
2187 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2188 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2189 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2190 			error = EINVAL;
2191 			break;
2192 		}
2193 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2194 		bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2195 		if (pa->ifname[0])
2196 			kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2197 		PF_RULES_WLOCK();
2198 		if (pp->ticket != V_ticket_pabuf) {
2199 			PF_RULES_WUNLOCK();
2200 			if (pa->ifname[0])
2201 				free(kif, PFI_MTYPE);
2202 			free(pa, M_PFRULE);
2203 			error = EBUSY;
2204 			break;
2205 		}
2206 		if (pa->ifname[0]) {
2207 			pa->kif = pfi_kif_attach(kif, pa->ifname);
2208 			pfi_kif_ref(pa->kif);
2209 		} else
2210 			pa->kif = NULL;
2211 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2212 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2213 			if (pa->ifname[0])
2214 				pfi_kif_unref(pa->kif);
2215 			PF_RULES_WUNLOCK();
2216 			free(pa, M_PFRULE);
2217 			break;
2218 		}
2219 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2220 		PF_RULES_WUNLOCK();
2221 		break;
2222 	}
2223 
2224 	case DIOCGETADDRS: {
2225 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2226 		struct pf_pool		*pool;
2227 		struct pf_pooladdr	*pa;
2228 
2229 		PF_RULES_RLOCK();
2230 		pp->nr = 0;
2231 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2232 		    pp->r_num, 0, 1, 0);
2233 		if (pool == NULL) {
2234 			PF_RULES_RUNLOCK();
2235 			error = EBUSY;
2236 			break;
2237 		}
2238 		TAILQ_FOREACH(pa, &pool->list, entries)
2239 			pp->nr++;
2240 		PF_RULES_RUNLOCK();
2241 		break;
2242 	}
2243 
2244 	case DIOCGETADDR: {
2245 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2246 		struct pf_pool		*pool;
2247 		struct pf_pooladdr	*pa;
2248 		u_int32_t		 nr = 0;
2249 
2250 		PF_RULES_RLOCK();
2251 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2252 		    pp->r_num, 0, 1, 1);
2253 		if (pool == NULL) {
2254 			PF_RULES_RUNLOCK();
2255 			error = EBUSY;
2256 			break;
2257 		}
2258 		pa = TAILQ_FIRST(&pool->list);
2259 		while ((pa != NULL) && (nr < pp->nr)) {
2260 			pa = TAILQ_NEXT(pa, entries);
2261 			nr++;
2262 		}
2263 		if (pa == NULL) {
2264 			PF_RULES_RUNLOCK();
2265 			error = EBUSY;
2266 			break;
2267 		}
2268 		bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2269 		pf_addr_copyout(&pp->addr.addr);
2270 		PF_RULES_RUNLOCK();
2271 		break;
2272 	}
2273 
2274 	case DIOCCHANGEADDR: {
2275 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
2276 		struct pf_pool		*pool;
2277 		struct pf_pooladdr	*oldpa = NULL, *newpa = NULL;
2278 		struct pf_ruleset	*ruleset;
2279 		struct pfi_kif		*kif = NULL;
2280 
2281 		if (pca->action < PF_CHANGE_ADD_HEAD ||
2282 		    pca->action > PF_CHANGE_REMOVE) {
2283 			error = EINVAL;
2284 			break;
2285 		}
2286 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2287 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2288 		    pca->addr.addr.type != PF_ADDR_TABLE) {
2289 			error = EINVAL;
2290 			break;
2291 		}
2292 
2293 		if (pca->action != PF_CHANGE_REMOVE) {
2294 #ifndef INET
2295 			if (pca->af == AF_INET) {
2296 				error = EAFNOSUPPORT;
2297 				break;
2298 			}
2299 #endif /* INET */
2300 #ifndef INET6
2301 			if (pca->af == AF_INET6) {
2302 				error = EAFNOSUPPORT;
2303 				break;
2304 			}
2305 #endif /* INET6 */
2306 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
2307 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2308 			if (newpa->ifname[0])
2309 				kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2310 			newpa->kif = NULL;
2311 		}
2312 
2313 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGEADDR_error; }
2314 		PF_RULES_WLOCK();
2315 		ruleset = pf_find_ruleset(pca->anchor);
2316 		if (ruleset == NULL)
2317 			ERROUT(EBUSY);
2318 
2319 		pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2320 		    pca->r_num, pca->r_last, 1, 1);
2321 		if (pool == NULL)
2322 			ERROUT(EBUSY);
2323 
2324 		if (pca->action != PF_CHANGE_REMOVE) {
2325 			if (newpa->ifname[0]) {
2326 				newpa->kif = pfi_kif_attach(kif, newpa->ifname);
2327 				pfi_kif_ref(newpa->kif);
2328 				kif = NULL;
2329 			}
2330 
2331 			switch (newpa->addr.type) {
2332 			case PF_ADDR_DYNIFTL:
2333 				error = pfi_dynaddr_setup(&newpa->addr,
2334 				    pca->af);
2335 				break;
2336 			case PF_ADDR_TABLE:
2337 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
2338 				    newpa->addr.v.tblname);
2339 				if (newpa->addr.p.tbl == NULL)
2340 					error = ENOMEM;
2341 				break;
2342 			}
2343 			if (error)
2344 				goto DIOCCHANGEADDR_error;
2345 		}
2346 
2347 		switch (pca->action) {
2348 		case PF_CHANGE_ADD_HEAD:
2349 			oldpa = TAILQ_FIRST(&pool->list);
2350 			break;
2351 		case PF_CHANGE_ADD_TAIL:
2352 			oldpa = TAILQ_LAST(&pool->list, pf_palist);
2353 			break;
2354 		default:
2355 			oldpa = TAILQ_FIRST(&pool->list);
2356 			for (int i = 0; oldpa && i < pca->nr; i++)
2357 				oldpa = TAILQ_NEXT(oldpa, entries);
2358 
2359 			if (oldpa == NULL)
2360 				ERROUT(EINVAL);
2361 		}
2362 
2363 		if (pca->action == PF_CHANGE_REMOVE) {
2364 			TAILQ_REMOVE(&pool->list, oldpa, entries);
2365 			switch (oldpa->addr.type) {
2366 			case PF_ADDR_DYNIFTL:
2367 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
2368 				break;
2369 			case PF_ADDR_TABLE:
2370 				pfr_detach_table(oldpa->addr.p.tbl);
2371 				break;
2372 			}
2373 			if (oldpa->kif)
2374 				pfi_kif_unref(oldpa->kif);
2375 			free(oldpa, M_PFRULE);
2376 		} else {
2377 			if (oldpa == NULL)
2378 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2379 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
2380 			    pca->action == PF_CHANGE_ADD_BEFORE)
2381 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2382 			else
2383 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
2384 				    newpa, entries);
2385 		}
2386 
2387 		pool->cur = TAILQ_FIRST(&pool->list);
2388 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
2389 		PF_RULES_WUNLOCK();
2390 		break;
2391 
2392 #undef ERROUT
2393 DIOCCHANGEADDR_error:
2394 		if (newpa->kif)
2395 			pfi_kif_unref(newpa->kif);
2396 		PF_RULES_WUNLOCK();
2397 		if (newpa != NULL)
2398 			free(newpa, M_PFRULE);
2399 		if (kif != NULL)
2400 			free(kif, PFI_MTYPE);
2401 		break;
2402 	}
2403 
2404 	case DIOCGETRULESETS: {
2405 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2406 		struct pf_ruleset	*ruleset;
2407 		struct pf_anchor	*anchor;
2408 
2409 		PF_RULES_RLOCK();
2410 		pr->path[sizeof(pr->path) - 1] = 0;
2411 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2412 			PF_RULES_RUNLOCK();
2413 			error = ENOENT;
2414 			break;
2415 		}
2416 		pr->nr = 0;
2417 		if (ruleset->anchor == NULL) {
2418 			/* XXX kludge for pf_main_ruleset */
2419 			RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2420 				if (anchor->parent == NULL)
2421 					pr->nr++;
2422 		} else {
2423 			RB_FOREACH(anchor, pf_anchor_node,
2424 			    &ruleset->anchor->children)
2425 				pr->nr++;
2426 		}
2427 		PF_RULES_RUNLOCK();
2428 		break;
2429 	}
2430 
2431 	case DIOCGETRULESET: {
2432 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2433 		struct pf_ruleset	*ruleset;
2434 		struct pf_anchor	*anchor;
2435 		u_int32_t		 nr = 0;
2436 
2437 		PF_RULES_RLOCK();
2438 		pr->path[sizeof(pr->path) - 1] = 0;
2439 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2440 			PF_RULES_RUNLOCK();
2441 			error = ENOENT;
2442 			break;
2443 		}
2444 		pr->name[0] = 0;
2445 		if (ruleset->anchor == NULL) {
2446 			/* XXX kludge for pf_main_ruleset */
2447 			RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2448 				if (anchor->parent == NULL && nr++ == pr->nr) {
2449 					strlcpy(pr->name, anchor->name,
2450 					    sizeof(pr->name));
2451 					break;
2452 				}
2453 		} else {
2454 			RB_FOREACH(anchor, pf_anchor_node,
2455 			    &ruleset->anchor->children)
2456 				if (nr++ == pr->nr) {
2457 					strlcpy(pr->name, anchor->name,
2458 					    sizeof(pr->name));
2459 					break;
2460 				}
2461 		}
2462 		if (!pr->name[0])
2463 			error = EBUSY;
2464 		PF_RULES_RUNLOCK();
2465 		break;
2466 	}
2467 
2468 	case DIOCRCLRTABLES: {
2469 		struct pfioc_table *io = (struct pfioc_table *)addr;
2470 
2471 		if (io->pfrio_esize != 0) {
2472 			error = ENODEV;
2473 			break;
2474 		}
2475 		PF_RULES_WLOCK();
2476 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2477 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2478 		PF_RULES_WUNLOCK();
2479 		break;
2480 	}
2481 
2482 	case DIOCRADDTABLES: {
2483 		struct pfioc_table *io = (struct pfioc_table *)addr;
2484 		struct pfr_table *pfrts;
2485 		size_t totlen;
2486 
2487 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2488 			error = ENODEV;
2489 			break;
2490 		}
2491 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2492 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2493 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2494 		if (error) {
2495 			free(pfrts, M_TEMP);
2496 			break;
2497 		}
2498 		PF_RULES_WLOCK();
2499 		error = pfr_add_tables(pfrts, io->pfrio_size,
2500 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2501 		PF_RULES_WUNLOCK();
2502 		free(pfrts, M_TEMP);
2503 		break;
2504 	}
2505 
2506 	case DIOCRDELTABLES: {
2507 		struct pfioc_table *io = (struct pfioc_table *)addr;
2508 		struct pfr_table *pfrts;
2509 		size_t totlen;
2510 
2511 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2512 			error = ENODEV;
2513 			break;
2514 		}
2515 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2516 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2517 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2518 		if (error) {
2519 			free(pfrts, M_TEMP);
2520 			break;
2521 		}
2522 		PF_RULES_WLOCK();
2523 		error = pfr_del_tables(pfrts, io->pfrio_size,
2524 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2525 		PF_RULES_WUNLOCK();
2526 		free(pfrts, M_TEMP);
2527 		break;
2528 	}
2529 
2530 	case DIOCRGETTABLES: {
2531 		struct pfioc_table *io = (struct pfioc_table *)addr;
2532 		struct pfr_table *pfrts;
2533 		size_t totlen;
2534 
2535 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2536 			error = ENODEV;
2537 			break;
2538 		}
2539 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2540 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2541 		PF_RULES_RLOCK();
2542 		error = pfr_get_tables(&io->pfrio_table, pfrts,
2543 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2544 		PF_RULES_RUNLOCK();
2545 		if (error == 0)
2546 			error = copyout(pfrts, io->pfrio_buffer, totlen);
2547 		free(pfrts, M_TEMP);
2548 		break;
2549 	}
2550 
2551 	case DIOCRGETTSTATS: {
2552 		struct pfioc_table *io = (struct pfioc_table *)addr;
2553 		struct pfr_tstats *pfrtstats;
2554 		size_t totlen;
2555 
2556 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2557 			error = ENODEV;
2558 			break;
2559 		}
2560 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
2561 		pfrtstats = malloc(totlen, M_TEMP, M_WAITOK);
2562 		PF_RULES_WLOCK();
2563 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
2564 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2565 		PF_RULES_WUNLOCK();
2566 		if (error == 0)
2567 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
2568 		free(pfrtstats, M_TEMP);
2569 		break;
2570 	}
2571 
2572 	case DIOCRCLRTSTATS: {
2573 		struct pfioc_table *io = (struct pfioc_table *)addr;
2574 		struct pfr_table *pfrts;
2575 		size_t totlen;
2576 
2577 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2578 			error = ENODEV;
2579 			break;
2580 		}
2581 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2582 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2583 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2584 		if (error) {
2585 			free(pfrts, M_TEMP);
2586 			break;
2587 		}
2588 		PF_RULES_WLOCK();
2589 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
2590 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2591 		PF_RULES_WUNLOCK();
2592 		free(pfrts, M_TEMP);
2593 		break;
2594 	}
2595 
2596 	case DIOCRSETTFLAGS: {
2597 		struct pfioc_table *io = (struct pfioc_table *)addr;
2598 		struct pfr_table *pfrts;
2599 		size_t totlen;
2600 
2601 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2602 			error = ENODEV;
2603 			break;
2604 		}
2605 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2606 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2607 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2608 		if (error) {
2609 			free(pfrts, M_TEMP);
2610 			break;
2611 		}
2612 		PF_RULES_WLOCK();
2613 		error = pfr_set_tflags(pfrts, io->pfrio_size,
2614 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2615 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2616 		PF_RULES_WUNLOCK();
2617 		free(pfrts, M_TEMP);
2618 		break;
2619 	}
2620 
2621 	case DIOCRCLRADDRS: {
2622 		struct pfioc_table *io = (struct pfioc_table *)addr;
2623 
2624 		if (io->pfrio_esize != 0) {
2625 			error = ENODEV;
2626 			break;
2627 		}
2628 		PF_RULES_WLOCK();
2629 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2630 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2631 		PF_RULES_WUNLOCK();
2632 		break;
2633 	}
2634 
2635 	case DIOCRADDADDRS: {
2636 		struct pfioc_table *io = (struct pfioc_table *)addr;
2637 		struct pfr_addr *pfras;
2638 		size_t totlen;
2639 
2640 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2641 			error = ENODEV;
2642 			break;
2643 		}
2644 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2645 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2646 		error = copyin(io->pfrio_buffer, pfras, totlen);
2647 		if (error) {
2648 			free(pfras, M_TEMP);
2649 			break;
2650 		}
2651 		PF_RULES_WLOCK();
2652 		error = pfr_add_addrs(&io->pfrio_table, pfras,
2653 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2654 		    PFR_FLAG_USERIOCTL);
2655 		PF_RULES_WUNLOCK();
2656 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2657 			error = copyout(pfras, io->pfrio_buffer, totlen);
2658 		free(pfras, M_TEMP);
2659 		break;
2660 	}
2661 
2662 	case DIOCRDELADDRS: {
2663 		struct pfioc_table *io = (struct pfioc_table *)addr;
2664 		struct pfr_addr *pfras;
2665 		size_t totlen;
2666 
2667 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2668 			error = ENODEV;
2669 			break;
2670 		}
2671 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2672 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2673 		error = copyin(io->pfrio_buffer, pfras, totlen);
2674 		if (error) {
2675 			free(pfras, M_TEMP);
2676 			break;
2677 		}
2678 		PF_RULES_WLOCK();
2679 		error = pfr_del_addrs(&io->pfrio_table, pfras,
2680 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2681 		    PFR_FLAG_USERIOCTL);
2682 		PF_RULES_WUNLOCK();
2683 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2684 			error = copyout(pfras, io->pfrio_buffer, totlen);
2685 		free(pfras, M_TEMP);
2686 		break;
2687 	}
2688 
2689 	case DIOCRSETADDRS: {
2690 		struct pfioc_table *io = (struct pfioc_table *)addr;
2691 		struct pfr_addr *pfras;
2692 		size_t totlen;
2693 
2694 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2695 			error = ENODEV;
2696 			break;
2697 		}
2698 		totlen = (io->pfrio_size + io->pfrio_size2) *
2699 		    sizeof(struct pfr_addr);
2700 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2701 		error = copyin(io->pfrio_buffer, pfras, totlen);
2702 		if (error) {
2703 			free(pfras, M_TEMP);
2704 			break;
2705 		}
2706 		PF_RULES_WLOCK();
2707 		error = pfr_set_addrs(&io->pfrio_table, pfras,
2708 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2709 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2710 		    PFR_FLAG_USERIOCTL, 0);
2711 		PF_RULES_WUNLOCK();
2712 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2713 			error = copyout(pfras, io->pfrio_buffer, totlen);
2714 		free(pfras, M_TEMP);
2715 		break;
2716 	}
2717 
2718 	case DIOCRGETADDRS: {
2719 		struct pfioc_table *io = (struct pfioc_table *)addr;
2720 		struct pfr_addr *pfras;
2721 		size_t totlen;
2722 
2723 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2724 			error = ENODEV;
2725 			break;
2726 		}
2727 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2728 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2729 		PF_RULES_RLOCK();
2730 		error = pfr_get_addrs(&io->pfrio_table, pfras,
2731 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2732 		PF_RULES_RUNLOCK();
2733 		if (error == 0)
2734 			error = copyout(pfras, io->pfrio_buffer, totlen);
2735 		free(pfras, M_TEMP);
2736 		break;
2737 	}
2738 
2739 	case DIOCRGETASTATS: {
2740 		struct pfioc_table *io = (struct pfioc_table *)addr;
2741 		struct pfr_astats *pfrastats;
2742 		size_t totlen;
2743 
2744 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2745 			error = ENODEV;
2746 			break;
2747 		}
2748 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
2749 		pfrastats = malloc(totlen, M_TEMP, M_WAITOK);
2750 		PF_RULES_RLOCK();
2751 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
2752 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2753 		PF_RULES_RUNLOCK();
2754 		if (error == 0)
2755 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
2756 		free(pfrastats, M_TEMP);
2757 		break;
2758 	}
2759 
2760 	case DIOCRCLRASTATS: {
2761 		struct pfioc_table *io = (struct pfioc_table *)addr;
2762 		struct pfr_addr *pfras;
2763 		size_t totlen;
2764 
2765 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2766 			error = ENODEV;
2767 			break;
2768 		}
2769 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2770 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2771 		error = copyin(io->pfrio_buffer, pfras, totlen);
2772 		if (error) {
2773 			free(pfras, M_TEMP);
2774 			break;
2775 		}
2776 		PF_RULES_WLOCK();
2777 		error = pfr_clr_astats(&io->pfrio_table, pfras,
2778 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2779 		    PFR_FLAG_USERIOCTL);
2780 		PF_RULES_WUNLOCK();
2781 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2782 			error = copyout(pfras, io->pfrio_buffer, totlen);
2783 		free(pfras, M_TEMP);
2784 		break;
2785 	}
2786 
2787 	case DIOCRTSTADDRS: {
2788 		struct pfioc_table *io = (struct pfioc_table *)addr;
2789 		struct pfr_addr *pfras;
2790 		size_t totlen;
2791 
2792 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2793 			error = ENODEV;
2794 			break;
2795 		}
2796 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2797 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2798 		error = copyin(io->pfrio_buffer, pfras, totlen);
2799 		if (error) {
2800 			free(pfras, M_TEMP);
2801 			break;
2802 		}
2803 		PF_RULES_RLOCK();
2804 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
2805 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2806 		    PFR_FLAG_USERIOCTL);
2807 		PF_RULES_RUNLOCK();
2808 		if (error == 0)
2809 			error = copyout(pfras, io->pfrio_buffer, totlen);
2810 		free(pfras, M_TEMP);
2811 		break;
2812 	}
2813 
2814 	case DIOCRINADEFINE: {
2815 		struct pfioc_table *io = (struct pfioc_table *)addr;
2816 		struct pfr_addr *pfras;
2817 		size_t totlen;
2818 
2819 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2820 			error = ENODEV;
2821 			break;
2822 		}
2823 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2824 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2825 		error = copyin(io->pfrio_buffer, pfras, totlen);
2826 		if (error) {
2827 			free(pfras, M_TEMP);
2828 			break;
2829 		}
2830 		PF_RULES_WLOCK();
2831 		error = pfr_ina_define(&io->pfrio_table, pfras,
2832 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2833 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2834 		PF_RULES_WUNLOCK();
2835 		free(pfras, M_TEMP);
2836 		break;
2837 	}
2838 
2839 	case DIOCOSFPADD: {
2840 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2841 		PF_RULES_WLOCK();
2842 		error = pf_osfp_add(io);
2843 		PF_RULES_WUNLOCK();
2844 		break;
2845 	}
2846 
2847 	case DIOCOSFPGET: {
2848 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2849 		PF_RULES_RLOCK();
2850 		error = pf_osfp_get(io);
2851 		PF_RULES_RUNLOCK();
2852 		break;
2853 	}
2854 
2855 	case DIOCXBEGIN: {
2856 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2857 		struct pfioc_trans_e	*ioes, *ioe;
2858 		size_t			 totlen;
2859 		int			 i;
2860 
2861 		if (io->esize != sizeof(*ioe)) {
2862 			error = ENODEV;
2863 			break;
2864 		}
2865 		totlen = sizeof(struct pfioc_trans_e) * io->size;
2866 		ioes = malloc(totlen, M_TEMP, M_WAITOK);
2867 		error = copyin(io->array, ioes, totlen);
2868 		if (error) {
2869 			free(ioes, M_TEMP);
2870 			break;
2871 		}
2872 		PF_RULES_WLOCK();
2873 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2874 			switch (ioe->rs_num) {
2875 #ifdef ALTQ
2876 			case PF_RULESET_ALTQ:
2877 				if (ioe->anchor[0]) {
2878 					PF_RULES_WUNLOCK();
2879 					free(ioes, M_TEMP);
2880 					error = EINVAL;
2881 					goto fail;
2882 				}
2883 				if ((error = pf_begin_altq(&ioe->ticket))) {
2884 					PF_RULES_WUNLOCK();
2885 					free(ioes, M_TEMP);
2886 					goto fail;
2887 				}
2888 				break;
2889 #endif /* ALTQ */
2890 			case PF_RULESET_TABLE:
2891 			    {
2892 				struct pfr_table table;
2893 
2894 				bzero(&table, sizeof(table));
2895 				strlcpy(table.pfrt_anchor, ioe->anchor,
2896 				    sizeof(table.pfrt_anchor));
2897 				if ((error = pfr_ina_begin(&table,
2898 				    &ioe->ticket, NULL, 0))) {
2899 					PF_RULES_WUNLOCK();
2900 					free(ioes, M_TEMP);
2901 					goto fail;
2902 				}
2903 				break;
2904 			    }
2905 			default:
2906 				if ((error = pf_begin_rules(&ioe->ticket,
2907 				    ioe->rs_num, ioe->anchor))) {
2908 					PF_RULES_WUNLOCK();
2909 					free(ioes, M_TEMP);
2910 					goto fail;
2911 				}
2912 				break;
2913 			}
2914 		}
2915 		PF_RULES_WUNLOCK();
2916 		error = copyout(ioes, io->array, totlen);
2917 		free(ioes, M_TEMP);
2918 		break;
2919 	}
2920 
2921 	case DIOCXROLLBACK: {
2922 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2923 		struct pfioc_trans_e	*ioe, *ioes;
2924 		size_t			 totlen;
2925 		int			 i;
2926 
2927 		if (io->esize != sizeof(*ioe)) {
2928 			error = ENODEV;
2929 			break;
2930 		}
2931 		totlen = sizeof(struct pfioc_trans_e) * io->size;
2932 		ioes = malloc(totlen, M_TEMP, M_WAITOK);
2933 		error = copyin(io->array, ioes, totlen);
2934 		if (error) {
2935 			free(ioes, M_TEMP);
2936 			break;
2937 		}
2938 		PF_RULES_WLOCK();
2939 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2940 			switch (ioe->rs_num) {
2941 #ifdef ALTQ
2942 			case PF_RULESET_ALTQ:
2943 				if (ioe->anchor[0]) {
2944 					PF_RULES_WUNLOCK();
2945 					free(ioes, M_TEMP);
2946 					error = EINVAL;
2947 					goto fail;
2948 				}
2949 				if ((error = pf_rollback_altq(ioe->ticket))) {
2950 					PF_RULES_WUNLOCK();
2951 					free(ioes, M_TEMP);
2952 					goto fail; /* really bad */
2953 				}
2954 				break;
2955 #endif /* ALTQ */
2956 			case PF_RULESET_TABLE:
2957 			    {
2958 				struct pfr_table table;
2959 
2960 				bzero(&table, sizeof(table));
2961 				strlcpy(table.pfrt_anchor, ioe->anchor,
2962 				    sizeof(table.pfrt_anchor));
2963 				if ((error = pfr_ina_rollback(&table,
2964 				    ioe->ticket, NULL, 0))) {
2965 					PF_RULES_WUNLOCK();
2966 					free(ioes, M_TEMP);
2967 					goto fail; /* really bad */
2968 				}
2969 				break;
2970 			    }
2971 			default:
2972 				if ((error = pf_rollback_rules(ioe->ticket,
2973 				    ioe->rs_num, ioe->anchor))) {
2974 					PF_RULES_WUNLOCK();
2975 					free(ioes, M_TEMP);
2976 					goto fail; /* really bad */
2977 				}
2978 				break;
2979 			}
2980 		}
2981 		PF_RULES_WUNLOCK();
2982 		free(ioes, M_TEMP);
2983 		break;
2984 	}
2985 
2986 	case DIOCXCOMMIT: {
2987 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2988 		struct pfioc_trans_e	*ioe, *ioes;
2989 		struct pf_ruleset	*rs;
2990 		size_t			 totlen;
2991 		int			 i;
2992 
2993 		if (io->esize != sizeof(*ioe)) {
2994 			error = ENODEV;
2995 			break;
2996 		}
2997 		totlen = sizeof(struct pfioc_trans_e) * io->size;
2998 		ioes = malloc(totlen, M_TEMP, M_WAITOK);
2999 		error = copyin(io->array, ioes, totlen);
3000 		if (error) {
3001 			free(ioes, M_TEMP);
3002 			break;
3003 		}
3004 		PF_RULES_WLOCK();
3005 		/* First makes sure everything will succeed. */
3006 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3007 			switch (ioe->rs_num) {
3008 #ifdef ALTQ
3009 			case PF_RULESET_ALTQ:
3010 				if (ioe->anchor[0]) {
3011 					PF_RULES_WUNLOCK();
3012 					free(ioes, M_TEMP);
3013 					error = EINVAL;
3014 					goto fail;
3015 				}
3016 				if (!V_altqs_inactive_open || ioe->ticket !=
3017 				    V_ticket_altqs_inactive) {
3018 					PF_RULES_WUNLOCK();
3019 					free(ioes, M_TEMP);
3020 					error = EBUSY;
3021 					goto fail;
3022 				}
3023 				break;
3024 #endif /* ALTQ */
3025 			case PF_RULESET_TABLE:
3026 				rs = pf_find_ruleset(ioe->anchor);
3027 				if (rs == NULL || !rs->topen || ioe->ticket !=
3028 				    rs->tticket) {
3029 					PF_RULES_WUNLOCK();
3030 					free(ioes, M_TEMP);
3031 					error = EBUSY;
3032 					goto fail;
3033 				}
3034 				break;
3035 			default:
3036 				if (ioe->rs_num < 0 || ioe->rs_num >=
3037 				    PF_RULESET_MAX) {
3038 					PF_RULES_WUNLOCK();
3039 					free(ioes, M_TEMP);
3040 					error = EINVAL;
3041 					goto fail;
3042 				}
3043 				rs = pf_find_ruleset(ioe->anchor);
3044 				if (rs == NULL ||
3045 				    !rs->rules[ioe->rs_num].inactive.open ||
3046 				    rs->rules[ioe->rs_num].inactive.ticket !=
3047 				    ioe->ticket) {
3048 					PF_RULES_WUNLOCK();
3049 					free(ioes, M_TEMP);
3050 					error = EBUSY;
3051 					goto fail;
3052 				}
3053 				break;
3054 			}
3055 		}
3056 		/* Now do the commit - no errors should happen here. */
3057 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3058 			switch (ioe->rs_num) {
3059 #ifdef ALTQ
3060 			case PF_RULESET_ALTQ:
3061 				if ((error = pf_commit_altq(ioe->ticket))) {
3062 					PF_RULES_WUNLOCK();
3063 					free(ioes, M_TEMP);
3064 					goto fail; /* really bad */
3065 				}
3066 				break;
3067 #endif /* ALTQ */
3068 			case PF_RULESET_TABLE:
3069 			    {
3070 				struct pfr_table table;
3071 
3072 				bzero(&table, sizeof(table));
3073 				strlcpy(table.pfrt_anchor, ioe->anchor,
3074 				    sizeof(table.pfrt_anchor));
3075 				if ((error = pfr_ina_commit(&table,
3076 				    ioe->ticket, NULL, NULL, 0))) {
3077 					PF_RULES_WUNLOCK();
3078 					free(ioes, M_TEMP);
3079 					goto fail; /* really bad */
3080 				}
3081 				break;
3082 			    }
3083 			default:
3084 				if ((error = pf_commit_rules(ioe->ticket,
3085 				    ioe->rs_num, ioe->anchor))) {
3086 					PF_RULES_WUNLOCK();
3087 					free(ioes, M_TEMP);
3088 					goto fail; /* really bad */
3089 				}
3090 				break;
3091 			}
3092 		}
3093 		PF_RULES_WUNLOCK();
3094 		free(ioes, M_TEMP);
3095 		break;
3096 	}
3097 
3098 	case DIOCGETSRCNODES: {
3099 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
3100 		struct pf_srchash	*sh;
3101 		struct pf_src_node	*n, *p, *pstore;
3102 		uint32_t		 i, nr = 0;
3103 
3104 		if (psn->psn_len == 0) {
3105 			for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
3106 			    i++, sh++) {
3107 				PF_HASHROW_LOCK(sh);
3108 				LIST_FOREACH(n, &sh->nodes, entry)
3109 					nr++;
3110 				PF_HASHROW_UNLOCK(sh);
3111 			}
3112 			psn->psn_len = sizeof(struct pf_src_node) * nr;
3113 			break;
3114 		}
3115 
3116 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK);
3117 		for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
3118 		    i++, sh++) {
3119 		    PF_HASHROW_LOCK(sh);
3120 		    LIST_FOREACH(n, &sh->nodes, entry) {
3121 			int	secs = time_uptime, diff;
3122 
3123 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
3124 				break;
3125 
3126 			bcopy(n, p, sizeof(struct pf_src_node));
3127 			if (n->rule.ptr != NULL)
3128 				p->rule.nr = n->rule.ptr->nr;
3129 			p->creation = secs - p->creation;
3130 			if (p->expire > secs)
3131 				p->expire -= secs;
3132 			else
3133 				p->expire = 0;
3134 
3135 			/* Adjust the connection rate estimate. */
3136 			diff = secs - n->conn_rate.last;
3137 			if (diff >= n->conn_rate.seconds)
3138 				p->conn_rate.count = 0;
3139 			else
3140 				p->conn_rate.count -=
3141 				    n->conn_rate.count * diff /
3142 				    n->conn_rate.seconds;
3143 			p++;
3144 			nr++;
3145 		    }
3146 		    PF_HASHROW_UNLOCK(sh);
3147 		}
3148 		error = copyout(pstore, psn->psn_src_nodes,
3149 		    sizeof(struct pf_src_node) * nr);
3150 		if (error) {
3151 			free(pstore, M_TEMP);
3152 			break;
3153 		}
3154 		psn->psn_len = sizeof(struct pf_src_node) * nr;
3155 		free(pstore, M_TEMP);
3156 		break;
3157 	}
3158 
3159 	case DIOCCLRSRCNODES: {
3160 
3161 		pf_clear_srcnodes(NULL);
3162 		pf_purge_expired_src_nodes();
3163 		V_pf_status.src_nodes = 0;
3164 		break;
3165 	}
3166 
3167 	case DIOCKILLSRCNODES:
3168 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
3169 		break;
3170 
3171 	case DIOCSETHOSTID: {
3172 		u_int32_t	*hostid = (u_int32_t *)addr;
3173 
3174 		PF_RULES_WLOCK();
3175 		if (*hostid == 0)
3176 			V_pf_status.hostid = arc4random();
3177 		else
3178 			V_pf_status.hostid = *hostid;
3179 		PF_RULES_WUNLOCK();
3180 		break;
3181 	}
3182 
3183 	case DIOCOSFPFLUSH:
3184 		PF_RULES_WLOCK();
3185 		pf_osfp_flush();
3186 		PF_RULES_WUNLOCK();
3187 		break;
3188 
3189 	case DIOCIGETIFACES: {
3190 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3191 		struct pfi_kif *ifstore;
3192 		size_t bufsiz;
3193 
3194 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
3195 			error = ENODEV;
3196 			break;
3197 		}
3198 
3199 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
3200 		ifstore = malloc(bufsiz, M_TEMP, M_WAITOK);
3201 		PF_RULES_RLOCK();
3202 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
3203 		PF_RULES_RUNLOCK();
3204 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
3205 		free(ifstore, M_TEMP);
3206 		break;
3207 	}
3208 
3209 	case DIOCSETIFFLAG: {
3210 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3211 
3212 		PF_RULES_WLOCK();
3213 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3214 		PF_RULES_WUNLOCK();
3215 		break;
3216 	}
3217 
3218 	case DIOCCLRIFFLAG: {
3219 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3220 
3221 		PF_RULES_WLOCK();
3222 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3223 		PF_RULES_WUNLOCK();
3224 		break;
3225 	}
3226 
3227 	default:
3228 		error = ENODEV;
3229 		break;
3230 	}
3231 fail:
3232 	CURVNET_RESTORE();
3233 
3234 	return (error);
3235 }
3236 
3237 void
3238 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
3239 {
3240 	bzero(sp, sizeof(struct pfsync_state));
3241 
3242 	/* copy from state key */
3243 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
3244 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
3245 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
3246 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
3247 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
3248 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
3249 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
3250 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
3251 	sp->proto = st->key[PF_SK_WIRE]->proto;
3252 	sp->af = st->key[PF_SK_WIRE]->af;
3253 
3254 	/* copy from state */
3255 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
3256 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
3257 	sp->creation = htonl(time_uptime - st->creation);
3258 	sp->expire = pf_state_expires(st);
3259 	if (sp->expire <= time_uptime)
3260 		sp->expire = htonl(0);
3261 	else
3262 		sp->expire = htonl(sp->expire - time_uptime);
3263 
3264 	sp->direction = st->direction;
3265 	sp->log = st->log;
3266 	sp->timeout = st->timeout;
3267 	sp->state_flags = st->state_flags;
3268 	if (st->src_node)
3269 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
3270 	if (st->nat_src_node)
3271 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
3272 
3273 	sp->id = st->id;
3274 	sp->creatorid = st->creatorid;
3275 	pf_state_peer_hton(&st->src, &sp->src);
3276 	pf_state_peer_hton(&st->dst, &sp->dst);
3277 
3278 	if (st->rule.ptr == NULL)
3279 		sp->rule = htonl(-1);
3280 	else
3281 		sp->rule = htonl(st->rule.ptr->nr);
3282 	if (st->anchor.ptr == NULL)
3283 		sp->anchor = htonl(-1);
3284 	else
3285 		sp->anchor = htonl(st->anchor.ptr->nr);
3286 	if (st->nat_rule.ptr == NULL)
3287 		sp->nat_rule = htonl(-1);
3288 	else
3289 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
3290 
3291 	pf_state_counter_hton(st->packets[0], sp->packets[0]);
3292 	pf_state_counter_hton(st->packets[1], sp->packets[1]);
3293 	pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
3294 	pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
3295 
3296 }
3297 
3298 static void
3299 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
3300 {
3301 	struct pfr_ktable *kt;
3302 
3303 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
3304 
3305 	kt = aw->p.tbl;
3306 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
3307 		kt = kt->pfrkt_root;
3308 	aw->p.tbl = NULL;
3309 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
3310 		kt->pfrkt_cnt : -1;
3311 }
3312 
3313 /*
3314  * XXX - Check for version missmatch!!!
3315  */
3316 static void
3317 pf_clear_states(void)
3318 {
3319 	struct pf_state	*s;
3320 	u_int i;
3321 
3322 	for (i = 0; i <= V_pf_hashmask; i++) {
3323 		struct pf_idhash *ih = &V_pf_idhash[i];
3324 relock:
3325 		PF_HASHROW_LOCK(ih);
3326 		LIST_FOREACH(s, &ih->states, entry) {
3327 			s->timeout = PFTM_PURGE;
3328 			/* Don't send out individual delete messages. */
3329 			s->sync_state = PFSTATE_NOSYNC;
3330 			pf_unlink_state(s, PF_ENTER_LOCKED);
3331 			goto relock;
3332 		}
3333 		PF_HASHROW_UNLOCK(ih);
3334 	}
3335 }
3336 
3337 static int
3338 pf_clear_tables(void)
3339 {
3340 	struct pfioc_table io;
3341 	int error;
3342 
3343 	bzero(&io, sizeof(io));
3344 
3345 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
3346 	    io.pfrio_flags);
3347 
3348 	return (error);
3349 }
3350 
3351 static void
3352 pf_clear_srcnodes(struct pf_src_node *n)
3353 {
3354 	struct pf_state *s;
3355 	int i;
3356 
3357 	for (i = 0; i <= V_pf_hashmask; i++) {
3358 		struct pf_idhash *ih = &V_pf_idhash[i];
3359 
3360 		PF_HASHROW_LOCK(ih);
3361 		LIST_FOREACH(s, &ih->states, entry) {
3362 			if (n == NULL || n == s->src_node)
3363 				s->src_node = NULL;
3364 			if (n == NULL || n == s->nat_src_node)
3365 				s->nat_src_node = NULL;
3366 		}
3367 		PF_HASHROW_UNLOCK(ih);
3368 	}
3369 
3370 	if (n == NULL) {
3371 		struct pf_srchash *sh;
3372 
3373 		for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
3374 		    i++, sh++) {
3375 			PF_HASHROW_LOCK(sh);
3376 			LIST_FOREACH(n, &sh->nodes, entry) {
3377 				n->expire = 1;
3378 				n->states = 0;
3379 			}
3380 			PF_HASHROW_UNLOCK(sh);
3381 		}
3382 	} else {
3383 		/* XXX: hash slot should already be locked here. */
3384 		n->expire = 1;
3385 		n->states = 0;
3386 	}
3387 }
3388 
3389 static void
3390 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
3391 {
3392 	struct pf_src_node_list	 kill;
3393 
3394 	LIST_INIT(&kill);
3395 	for (int i = 0; i <= V_pf_srchashmask; i++) {
3396 		struct pf_srchash *sh = &V_pf_srchash[i];
3397 		struct pf_src_node *sn, *tmp;
3398 
3399 		PF_HASHROW_LOCK(sh);
3400 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
3401 			if (PF_MATCHA(psnk->psnk_src.neg,
3402 			      &psnk->psnk_src.addr.v.a.addr,
3403 			      &psnk->psnk_src.addr.v.a.mask,
3404 			      &sn->addr, sn->af) &&
3405 			    PF_MATCHA(psnk->psnk_dst.neg,
3406 			      &psnk->psnk_dst.addr.v.a.addr,
3407 			      &psnk->psnk_dst.addr.v.a.mask,
3408 			      &sn->raddr, sn->af)) {
3409 				pf_unlink_src_node_locked(sn);
3410 				LIST_INSERT_HEAD(&kill, sn, entry);
3411 				sn->expire = 1;
3412 			}
3413 		PF_HASHROW_UNLOCK(sh);
3414 	}
3415 
3416 	for (int i = 0; i <= V_pf_hashmask; i++) {
3417 		struct pf_idhash *ih = &V_pf_idhash[i];
3418 		struct pf_state *s;
3419 
3420 		PF_HASHROW_LOCK(ih);
3421 		LIST_FOREACH(s, &ih->states, entry) {
3422 			if (s->src_node && s->src_node->expire == 1) {
3423 #ifdef INVARIANTS
3424 				s->src_node->states--;
3425 #endif
3426 				s->src_node = NULL;
3427 			}
3428 			if (s->nat_src_node && s->nat_src_node->expire == 1) {
3429 #ifdef INVARIANTS
3430 				s->nat_src_node->states--;
3431 #endif
3432 				s->nat_src_node = NULL;
3433 			}
3434 		}
3435 		PF_HASHROW_UNLOCK(ih);
3436 	}
3437 
3438 	psnk->psnk_killed = pf_free_src_nodes(&kill);
3439 }
3440 
3441 /*
3442  * XXX - Check for version missmatch!!!
3443  */
3444 
3445 /*
3446  * Duplicate pfctl -Fa operation to get rid of as much as we can.
3447  */
3448 static int
3449 shutdown_pf(void)
3450 {
3451 	int error = 0;
3452 	u_int32_t t[5];
3453 	char nn = '\0';
3454 
3455 	V_pf_status.running = 0;
3456 
3457 	counter_u64_free(V_pf_default_rule.states_cur);
3458 	counter_u64_free(V_pf_default_rule.states_tot);
3459 	counter_u64_free(V_pf_default_rule.src_nodes);
3460 
3461 	do {
3462 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
3463 		    != 0) {
3464 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
3465 			break;
3466 		}
3467 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
3468 		    != 0) {
3469 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
3470 			break;		/* XXX: rollback? */
3471 		}
3472 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
3473 		    != 0) {
3474 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
3475 			break;		/* XXX: rollback? */
3476 		}
3477 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
3478 		    != 0) {
3479 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
3480 			break;		/* XXX: rollback? */
3481 		}
3482 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
3483 		    != 0) {
3484 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
3485 			break;		/* XXX: rollback? */
3486 		}
3487 
3488 		/* XXX: these should always succeed here */
3489 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
3490 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
3491 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
3492 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
3493 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
3494 
3495 		if ((error = pf_clear_tables()) != 0)
3496 			break;
3497 
3498 #ifdef ALTQ
3499 		if ((error = pf_begin_altq(&t[0])) != 0) {
3500 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
3501 			break;
3502 		}
3503 		pf_commit_altq(t[0]);
3504 #endif
3505 
3506 		pf_clear_states();
3507 
3508 		pf_clear_srcnodes(NULL);
3509 
3510 		/* status does not use malloced mem so no need to cleanup */
3511 		/* fingerprints and interfaces have thier own cleanup code */
3512 	} while(0);
3513 
3514 	return (error);
3515 }
3516 
3517 #ifdef INET
3518 static int
3519 pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3520     struct inpcb *inp)
3521 {
3522 	int chk;
3523 
3524 	chk = pf_test(PF_IN, ifp, m, inp);
3525 	if (chk && *m) {
3526 		m_freem(*m);
3527 		*m = NULL;
3528 	}
3529 
3530 	return (chk);
3531 }
3532 
3533 static int
3534 pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3535     struct inpcb *inp)
3536 {
3537 	int chk;
3538 
3539 	/* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3540 	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3541 		in_delayed_cksum(*m);
3542 		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3543 	}
3544 
3545 	chk = pf_test(PF_OUT, ifp, m, inp);
3546 	if (chk && *m) {
3547 		m_freem(*m);
3548 		*m = NULL;
3549 	}
3550 
3551 	return (chk);
3552 }
3553 #endif
3554 
3555 #ifdef INET6
3556 static int
3557 pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3558     struct inpcb *inp)
3559 {
3560 	int chk;
3561 
3562 	/*
3563 	 * In case of loopback traffic IPv6 uses the real interface in
3564 	 * order to support scoped addresses. In order to support stateful
3565 	 * filtering we have change this to lo0 as it is the case in IPv4.
3566 	 */
3567 	CURVNET_SET(ifp->if_vnet);
3568 	chk = pf_test6(PF_IN, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
3569 	CURVNET_RESTORE();
3570 	if (chk && *m) {
3571 		m_freem(*m);
3572 		*m = NULL;
3573 	}
3574 	return chk;
3575 }
3576 
3577 static int
3578 pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3579     struct inpcb *inp)
3580 {
3581 	int chk;
3582 
3583 	/* We need a proper CSUM before we start (s. OpenBSD ip_output) */
3584 	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3585 #ifdef INET
3586 		/* XXX-BZ copy&paste error from r126261? */
3587 		in_delayed_cksum(*m);
3588 #endif
3589 		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3590 	}
3591 	CURVNET_SET(ifp->if_vnet);
3592 	chk = pf_test6(PF_OUT, ifp, m, inp);
3593 	CURVNET_RESTORE();
3594 	if (chk && *m) {
3595 		m_freem(*m);
3596 		*m = NULL;
3597 	}
3598 	return chk;
3599 }
3600 #endif /* INET6 */
3601 
3602 static int
3603 hook_pf(void)
3604 {
3605 #ifdef INET
3606 	struct pfil_head *pfh_inet;
3607 #endif
3608 #ifdef INET6
3609 	struct pfil_head *pfh_inet6;
3610 #endif
3611 
3612 	if (V_pf_pfil_hooked)
3613 		return (0);
3614 
3615 #ifdef INET
3616 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3617 	if (pfh_inet == NULL)
3618 		return (ESRCH); /* XXX */
3619 	pfil_add_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet);
3620 	pfil_add_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet);
3621 #endif
3622 #ifdef INET6
3623 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3624 	if (pfh_inet6 == NULL) {
3625 #ifdef INET
3626 		pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3627 		    pfh_inet);
3628 		pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3629 		    pfh_inet);
3630 #endif
3631 		return (ESRCH); /* XXX */
3632 	}
3633 	pfil_add_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet6);
3634 	pfil_add_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet6);
3635 #endif
3636 
3637 	V_pf_pfil_hooked = 1;
3638 	return (0);
3639 }
3640 
3641 static int
3642 dehook_pf(void)
3643 {
3644 #ifdef INET
3645 	struct pfil_head *pfh_inet;
3646 #endif
3647 #ifdef INET6
3648 	struct pfil_head *pfh_inet6;
3649 #endif
3650 
3651 	if (V_pf_pfil_hooked == 0)
3652 		return (0);
3653 
3654 #ifdef INET
3655 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3656 	if (pfh_inet == NULL)
3657 		return (ESRCH); /* XXX */
3658 	pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3659 	    pfh_inet);
3660 	pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3661 	    pfh_inet);
3662 #endif
3663 #ifdef INET6
3664 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3665 	if (pfh_inet6 == NULL)
3666 		return (ESRCH); /* XXX */
3667 	pfil_remove_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK,
3668 	    pfh_inet6);
3669 	pfil_remove_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK,
3670 	    pfh_inet6);
3671 #endif
3672 
3673 	V_pf_pfil_hooked = 0;
3674 	return (0);
3675 }
3676 
3677 static int
3678 pf_load(void)
3679 {
3680 	int error;
3681 
3682 	VNET_ITERATOR_DECL(vnet_iter);
3683 
3684 	VNET_LIST_RLOCK();
3685 	VNET_FOREACH(vnet_iter) {
3686 		CURVNET_SET(vnet_iter);
3687 		V_pf_pfil_hooked = 0;
3688 		V_pf_end_threads = 0;
3689 		TAILQ_INIT(&V_pf_tags);
3690 		TAILQ_INIT(&V_pf_qids);
3691 		CURVNET_RESTORE();
3692 	}
3693 	VNET_LIST_RUNLOCK();
3694 
3695 	rw_init(&pf_rules_lock, "pf rulesets");
3696 
3697 	pf_dev = make_dev(&pf_cdevsw, 0, 0, 0, 0600, PF_NAME);
3698 	if ((error = pfattach()) != 0)
3699 		return (error);
3700 
3701 	return (0);
3702 }
3703 
3704 static int
3705 pf_unload(void)
3706 {
3707 	int error = 0;
3708 
3709 	PF_RULES_WLOCK();
3710 	V_pf_status.running = 0;
3711 	PF_RULES_WUNLOCK();
3712 	swi_remove(V_pf_swi_cookie);
3713 	error = dehook_pf();
3714 	if (error) {
3715 		/*
3716 		 * Should not happen!
3717 		 * XXX Due to error code ESRCH, kldunload will show
3718 		 * a message like 'No such process'.
3719 		 */
3720 		printf("%s : pfil unregisteration fail\n", __FUNCTION__);
3721 		return error;
3722 	}
3723 	PF_RULES_WLOCK();
3724 	shutdown_pf();
3725 	V_pf_end_threads = 1;
3726 	while (V_pf_end_threads < 2) {
3727 		wakeup_one(pf_purge_thread);
3728 		rw_sleep(pf_purge_thread, &pf_rules_lock, 0, "pftmo", 0);
3729 	}
3730 	pf_normalize_cleanup();
3731 	pfi_cleanup();
3732 	pfr_cleanup();
3733 	pf_osfp_flush();
3734 	pf_cleanup();
3735 	PF_RULES_WUNLOCK();
3736 	destroy_dev(pf_dev);
3737 	rw_destroy(&pf_rules_lock);
3738 
3739 	return (error);
3740 }
3741 
3742 static int
3743 pf_modevent(module_t mod, int type, void *data)
3744 {
3745 	int error = 0;
3746 
3747 	switch(type) {
3748 	case MOD_LOAD:
3749 		error = pf_load();
3750 		break;
3751 	case MOD_QUIESCE:
3752 		/*
3753 		 * Module should not be unloaded due to race conditions.
3754 		 */
3755 		error = EBUSY;
3756 		break;
3757 	case MOD_UNLOAD:
3758 		error = pf_unload();
3759 		break;
3760 	default:
3761 		error = EINVAL;
3762 		break;
3763 	}
3764 
3765 	return (error);
3766 }
3767 
3768 static moduledata_t pf_mod = {
3769 	"pf",
3770 	pf_modevent,
3771 	0
3772 };
3773 
3774 DECLARE_MODULE(pf, pf_mod, SI_SUB_PSEUDO, SI_ORDER_FIRST);
3775 MODULE_VERSION(pf, PF_MODVER);
3776