xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision 6ef6ba9950260f42b47499d17874d00ca9290955)
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_tbladdr_copyout(struct pf_addr_wrap *);
159 
160 /*
161  * Wrapper functions for pfil(9) hooks
162  */
163 #ifdef INET
164 static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp,
165     int dir, struct inpcb *inp);
166 static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp,
167     int dir, struct inpcb *inp);
168 #endif
169 #ifdef INET6
170 static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp,
171     int dir, struct inpcb *inp);
172 static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp,
173     int dir, struct inpcb *inp);
174 #endif
175 
176 static int		hook_pf(void);
177 static int		dehook_pf(void);
178 static int		shutdown_pf(void);
179 static int		pf_load(void);
180 static int		pf_unload(void);
181 
182 static struct cdevsw pf_cdevsw = {
183 	.d_ioctl =	pfioctl,
184 	.d_name =	PF_NAME,
185 	.d_version =	D_VERSION,
186 };
187 
188 static volatile VNET_DEFINE(int, pf_pfil_hooked);
189 #define V_pf_pfil_hooked	VNET(pf_pfil_hooked)
190 VNET_DEFINE(int,		pf_end_threads);
191 
192 struct rwlock			pf_rules_lock;
193 
194 /* pfsync */
195 pfsync_state_import_t 		*pfsync_state_import_ptr = NULL;
196 pfsync_insert_state_t		*pfsync_insert_state_ptr = NULL;
197 pfsync_update_state_t		*pfsync_update_state_ptr = NULL;
198 pfsync_delete_state_t		*pfsync_delete_state_ptr = NULL;
199 pfsync_clear_states_t		*pfsync_clear_states_ptr = NULL;
200 pfsync_defer_t			*pfsync_defer_ptr = NULL;
201 /* pflog */
202 pflog_packet_t			*pflog_packet_ptr = NULL;
203 
204 static int
205 pfattach(void)
206 {
207 	u_int32_t *my_timeout = V_pf_default_rule.timeout;
208 	int error;
209 
210 	pf_initialize();
211 	pfr_initialize();
212 	pfi_initialize();
213 	pf_normalize_init();
214 
215 	V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
216 	V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
217 
218 	RB_INIT(&V_pf_anchors);
219 	pf_init_ruleset(&pf_main_ruleset);
220 
221 	/* default rule should never be garbage collected */
222 	V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
223 #ifdef PF_DEFAULT_TO_DROP
224 	V_pf_default_rule.action = PF_DROP;
225 #else
226 	V_pf_default_rule.action = PF_PASS;
227 #endif
228 	V_pf_default_rule.nr = -1;
229 	V_pf_default_rule.rtableid = -1;
230 
231 	/* initialize default timeouts */
232 	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
233 	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
234 	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
235 	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
236 	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
237 	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
238 	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
239 	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
240 	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
241 	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
242 	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
243 	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
244 	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
245 	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
246 	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
247 	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
248 	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
249 	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
250 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
251 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
252 
253 	bzero(&V_pf_status, sizeof(V_pf_status));
254 	V_pf_status.debug = PF_DEBUG_URGENT;
255 
256 	V_pf_pfil_hooked = 0;
257 
258 	/* XXX do our best to avoid a conflict */
259 	V_pf_status.hostid = arc4random();
260 
261 	if ((error = kproc_create(pf_purge_thread, curvnet, NULL, 0, 0,
262 	    "pf purge")) != 0)
263 		/* XXXGL: leaked all above. */
264 		return (error);
265 	if ((error = swi_add(NULL, "pf send", pf_intr, curvnet, SWI_NET,
266 	    INTR_MPSAFE, &V_pf_swi_cookie)) != 0)
267 		/* XXXGL: leaked all above. */
268 		return (error);
269 
270 	return (0);
271 }
272 
273 static struct pf_pool *
274 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
275     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
276     u_int8_t check_ticket)
277 {
278 	struct pf_ruleset	*ruleset;
279 	struct pf_rule		*rule;
280 	int			 rs_num;
281 
282 	ruleset = pf_find_ruleset(anchor);
283 	if (ruleset == NULL)
284 		return (NULL);
285 	rs_num = pf_get_ruleset_number(rule_action);
286 	if (rs_num >= PF_RULESET_MAX)
287 		return (NULL);
288 	if (active) {
289 		if (check_ticket && ticket !=
290 		    ruleset->rules[rs_num].active.ticket)
291 			return (NULL);
292 		if (r_last)
293 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
294 			    pf_rulequeue);
295 		else
296 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
297 	} else {
298 		if (check_ticket && ticket !=
299 		    ruleset->rules[rs_num].inactive.ticket)
300 			return (NULL);
301 		if (r_last)
302 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
303 			    pf_rulequeue);
304 		else
305 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
306 	}
307 	if (!r_last) {
308 		while ((rule != NULL) && (rule->nr != rule_number))
309 			rule = TAILQ_NEXT(rule, entries);
310 	}
311 	if (rule == NULL)
312 		return (NULL);
313 
314 	return (&rule->rpool);
315 }
316 
317 static void
318 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
319 {
320 	struct pf_pooladdr	*mv_pool_pa;
321 
322 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
323 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
324 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
325 	}
326 }
327 
328 static void
329 pf_empty_pool(struct pf_palist *poola)
330 {
331 	struct pf_pooladdr *pa;
332 
333 	while ((pa = TAILQ_FIRST(poola)) != NULL) {
334 		switch (pa->addr.type) {
335 		case PF_ADDR_DYNIFTL:
336 			pfi_dynaddr_remove(pa->addr.p.dyn);
337 			break;
338 		case PF_ADDR_TABLE:
339 			pfr_detach_table(pa->addr.p.tbl);
340 			break;
341 		}
342 		if (pa->kif)
343 			pfi_kif_unref(pa->kif);
344 		TAILQ_REMOVE(poola, pa, entries);
345 		free(pa, M_PFRULE);
346 	}
347 }
348 
349 static void
350 pf_unlink_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
351 {
352 
353 	PF_RULES_WASSERT();
354 
355 	TAILQ_REMOVE(rulequeue, rule, entries);
356 
357 	PF_UNLNKDRULES_LOCK();
358 	rule->rule_flag |= PFRULE_REFS;
359 	TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
360 	PF_UNLNKDRULES_UNLOCK();
361 }
362 
363 void
364 pf_free_rule(struct pf_rule *rule)
365 {
366 
367 	PF_RULES_WASSERT();
368 
369 	if (rule->tag)
370 		tag_unref(&V_pf_tags, rule->tag);
371 	if (rule->match_tag)
372 		tag_unref(&V_pf_tags, rule->match_tag);
373 #ifdef ALTQ
374 	if (rule->pqid != rule->qid)
375 		pf_qid_unref(rule->pqid);
376 	pf_qid_unref(rule->qid);
377 #endif
378 	switch (rule->src.addr.type) {
379 	case PF_ADDR_DYNIFTL:
380 		pfi_dynaddr_remove(rule->src.addr.p.dyn);
381 		break;
382 	case PF_ADDR_TABLE:
383 		pfr_detach_table(rule->src.addr.p.tbl);
384 		break;
385 	}
386 	switch (rule->dst.addr.type) {
387 	case PF_ADDR_DYNIFTL:
388 		pfi_dynaddr_remove(rule->dst.addr.p.dyn);
389 		break;
390 	case PF_ADDR_TABLE:
391 		pfr_detach_table(rule->dst.addr.p.tbl);
392 		break;
393 	}
394 	if (rule->overload_tbl)
395 		pfr_detach_table(rule->overload_tbl);
396 	if (rule->kif)
397 		pfi_kif_unref(rule->kif);
398 	pf_anchor_remove(rule);
399 	pf_empty_pool(&rule->rpool.list);
400 	free(rule, M_PFRULE);
401 }
402 
403 static u_int16_t
404 tagname2tag(struct pf_tags *head, char *tagname)
405 {
406 	struct pf_tagname	*tag, *p = NULL;
407 	u_int16_t		 new_tagid = 1;
408 
409 	PF_RULES_WASSERT();
410 
411 	TAILQ_FOREACH(tag, head, entries)
412 		if (strcmp(tagname, tag->name) == 0) {
413 			tag->ref++;
414 			return (tag->tag);
415 		}
416 
417 	/*
418 	 * to avoid fragmentation, we do a linear search from the beginning
419 	 * and take the first free slot we find. if there is none or the list
420 	 * is empty, append a new entry at the end.
421 	 */
422 
423 	/* new entry */
424 	if (!TAILQ_EMPTY(head))
425 		for (p = TAILQ_FIRST(head); p != NULL &&
426 		    p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
427 			new_tagid = p->tag + 1;
428 
429 	if (new_tagid > TAGID_MAX)
430 		return (0);
431 
432 	/* allocate and fill new struct pf_tagname */
433 	tag = malloc(sizeof(*tag), M_PFTAG, M_NOWAIT|M_ZERO);
434 	if (tag == NULL)
435 		return (0);
436 	strlcpy(tag->name, tagname, sizeof(tag->name));
437 	tag->tag = new_tagid;
438 	tag->ref++;
439 
440 	if (p != NULL)	/* insert new entry before p */
441 		TAILQ_INSERT_BEFORE(p, tag, entries);
442 	else	/* either list empty or no free slot in between */
443 		TAILQ_INSERT_TAIL(head, tag, entries);
444 
445 	return (tag->tag);
446 }
447 
448 static void
449 tag_unref(struct pf_tags *head, u_int16_t tag)
450 {
451 	struct pf_tagname	*p, *next;
452 
453 	PF_RULES_WASSERT();
454 
455 	for (p = TAILQ_FIRST(head); p != NULL; p = next) {
456 		next = TAILQ_NEXT(p, entries);
457 		if (tag == p->tag) {
458 			if (--p->ref == 0) {
459 				TAILQ_REMOVE(head, p, entries);
460 				free(p, M_PFTAG);
461 			}
462 			break;
463 		}
464 	}
465 }
466 
467 static u_int16_t
468 pf_tagname2tag(char *tagname)
469 {
470 	return (tagname2tag(&V_pf_tags, tagname));
471 }
472 
473 #ifdef ALTQ
474 static u_int32_t
475 pf_qname2qid(char *qname)
476 {
477 	return ((u_int32_t)tagname2tag(&V_pf_qids, qname));
478 }
479 
480 static void
481 pf_qid_unref(u_int32_t qid)
482 {
483 	tag_unref(&V_pf_qids, (u_int16_t)qid);
484 }
485 
486 static int
487 pf_begin_altq(u_int32_t *ticket)
488 {
489 	struct pf_altq	*altq;
490 	int		 error = 0;
491 
492 	PF_RULES_WASSERT();
493 
494 	/* Purge the old altq list */
495 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
496 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
497 		if (altq->qname[0] == 0 &&
498 		    (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
499 			/* detach and destroy the discipline */
500 			error = altq_remove(altq);
501 		} else
502 			pf_qid_unref(altq->qid);
503 		free(altq, M_PFALTQ);
504 	}
505 	if (error)
506 		return (error);
507 	*ticket = ++V_ticket_altqs_inactive;
508 	V_altqs_inactive_open = 1;
509 	return (0);
510 }
511 
512 static int
513 pf_rollback_altq(u_int32_t ticket)
514 {
515 	struct pf_altq	*altq;
516 	int		 error = 0;
517 
518 	PF_RULES_WASSERT();
519 
520 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
521 		return (0);
522 	/* Purge the old altq list */
523 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
524 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
525 		if (altq->qname[0] == 0 &&
526 		   (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
527 			/* detach and destroy the discipline */
528 			error = altq_remove(altq);
529 		} else
530 			pf_qid_unref(altq->qid);
531 		free(altq, M_PFALTQ);
532 	}
533 	V_altqs_inactive_open = 0;
534 	return (error);
535 }
536 
537 static int
538 pf_commit_altq(u_int32_t ticket)
539 {
540 	struct pf_altqqueue	*old_altqs;
541 	struct pf_altq		*altq;
542 	int			 err, error = 0;
543 
544 	PF_RULES_WASSERT();
545 
546 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
547 		return (EBUSY);
548 
549 	/* swap altqs, keep the old. */
550 	old_altqs = V_pf_altqs_active;
551 	V_pf_altqs_active = V_pf_altqs_inactive;
552 	V_pf_altqs_inactive = old_altqs;
553 	V_ticket_altqs_active = V_ticket_altqs_inactive;
554 
555 	/* Attach new disciplines */
556 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
557 	if (altq->qname[0] == 0 &&
558 	   (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
559 			/* attach the discipline */
560 			error = altq_pfattach(altq);
561 			if (error == 0 && V_pf_altq_running)
562 				error = pf_enable_altq(altq);
563 			if (error != 0)
564 				return (error);
565 		}
566 	}
567 
568 	/* Purge the old altq list */
569 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
570 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
571 		if (altq->qname[0] == 0 &&
572 		    (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
573 			/* detach and destroy the discipline */
574 			if (V_pf_altq_running)
575 				error = pf_disable_altq(altq);
576 			err = altq_pfdetach(altq);
577 			if (err != 0 && error == 0)
578 				error = err;
579 			err = altq_remove(altq);
580 			if (err != 0 && error == 0)
581 				error = err;
582 		} else
583 			pf_qid_unref(altq->qid);
584 		free(altq, M_PFALTQ);
585 	}
586 
587 	V_altqs_inactive_open = 0;
588 	return (error);
589 }
590 
591 static int
592 pf_enable_altq(struct pf_altq *altq)
593 {
594 	struct ifnet		*ifp;
595 	struct tb_profile	 tb;
596 	int			 error = 0;
597 
598 	if ((ifp = ifunit(altq->ifname)) == NULL)
599 		return (EINVAL);
600 
601 	if (ifp->if_snd.altq_type != ALTQT_NONE)
602 		error = altq_enable(&ifp->if_snd);
603 
604 	/* set tokenbucket regulator */
605 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
606 		tb.rate = altq->ifbandwidth;
607 		tb.depth = altq->tbrsize;
608 		error = tbr_set(&ifp->if_snd, &tb);
609 	}
610 
611 	return (error);
612 }
613 
614 static int
615 pf_disable_altq(struct pf_altq *altq)
616 {
617 	struct ifnet		*ifp;
618 	struct tb_profile	 tb;
619 	int			 error;
620 
621 	if ((ifp = ifunit(altq->ifname)) == NULL)
622 		return (EINVAL);
623 
624 	/*
625 	 * when the discipline is no longer referenced, it was overridden
626 	 * by a new one.  if so, just return.
627 	 */
628 	if (altq->altq_disc != ifp->if_snd.altq_disc)
629 		return (0);
630 
631 	error = altq_disable(&ifp->if_snd);
632 
633 	if (error == 0) {
634 		/* clear tokenbucket regulator */
635 		tb.rate = 0;
636 		error = tbr_set(&ifp->if_snd, &tb);
637 	}
638 
639 	return (error);
640 }
641 
642 void
643 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
644 {
645 	struct ifnet	*ifp1;
646 	struct pf_altq	*a1, *a2, *a3;
647 	u_int32_t	 ticket;
648 	int		 error = 0;
649 
650 	/* Interrupt userland queue modifications */
651 	if (V_altqs_inactive_open)
652 		pf_rollback_altq(V_ticket_altqs_inactive);
653 
654 	/* Start new altq ruleset */
655 	if (pf_begin_altq(&ticket))
656 		return;
657 
658 	/* Copy the current active set */
659 	TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
660 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
661 		if (a2 == NULL) {
662 			error = ENOMEM;
663 			break;
664 		}
665 		bcopy(a1, a2, sizeof(struct pf_altq));
666 
667 		if (a2->qname[0] != 0) {
668 			if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
669 				error = EBUSY;
670 				free(a2, M_PFALTQ);
671 				break;
672 			}
673 			a2->altq_disc = NULL;
674 			TAILQ_FOREACH(a3, V_pf_altqs_inactive, entries) {
675 				if (strncmp(a3->ifname, a2->ifname,
676 				    IFNAMSIZ) == 0 && a3->qname[0] == 0) {
677 					a2->altq_disc = a3->altq_disc;
678 					break;
679 				}
680 			}
681 		}
682 		/* Deactivate the interface in question */
683 		a2->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
684 		if ((ifp1 = ifunit(a2->ifname)) == NULL ||
685 		    (remove && ifp1 == ifp)) {
686 			a2->local_flags |= PFALTQ_FLAG_IF_REMOVED;
687 		} else {
688 			error = altq_add(a2);
689 
690 			if (ticket != V_ticket_altqs_inactive)
691 				error = EBUSY;
692 
693 			if (error) {
694 				free(a2, M_PFALTQ);
695 				break;
696 			}
697 		}
698 
699 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
700 	}
701 
702 	if (error != 0)
703 		pf_rollback_altq(ticket);
704 	else
705 		pf_commit_altq(ticket);
706 }
707 #endif /* ALTQ */
708 
709 static int
710 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
711 {
712 	struct pf_ruleset	*rs;
713 	struct pf_rule		*rule;
714 
715 	PF_RULES_WASSERT();
716 
717 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
718 		return (EINVAL);
719 	rs = pf_find_or_create_ruleset(anchor);
720 	if (rs == NULL)
721 		return (EINVAL);
722 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
723 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
724 		rs->rules[rs_num].inactive.rcount--;
725 	}
726 	*ticket = ++rs->rules[rs_num].inactive.ticket;
727 	rs->rules[rs_num].inactive.open = 1;
728 	return (0);
729 }
730 
731 static int
732 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
733 {
734 	struct pf_ruleset	*rs;
735 	struct pf_rule		*rule;
736 
737 	PF_RULES_WASSERT();
738 
739 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
740 		return (EINVAL);
741 	rs = pf_find_ruleset(anchor);
742 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
743 	    rs->rules[rs_num].inactive.ticket != ticket)
744 		return (0);
745 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
746 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
747 		rs->rules[rs_num].inactive.rcount--;
748 	}
749 	rs->rules[rs_num].inactive.open = 0;
750 	return (0);
751 }
752 
753 #define PF_MD5_UPD(st, elm)						\
754 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
755 
756 #define PF_MD5_UPD_STR(st, elm)						\
757 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
758 
759 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
760 		(stor) = htonl((st)->elm);				\
761 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
762 } while (0)
763 
764 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
765 		(stor) = htons((st)->elm);				\
766 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
767 } while (0)
768 
769 static void
770 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
771 {
772 	PF_MD5_UPD(pfr, addr.type);
773 	switch (pfr->addr.type) {
774 		case PF_ADDR_DYNIFTL:
775 			PF_MD5_UPD(pfr, addr.v.ifname);
776 			PF_MD5_UPD(pfr, addr.iflags);
777 			break;
778 		case PF_ADDR_TABLE:
779 			PF_MD5_UPD(pfr, addr.v.tblname);
780 			break;
781 		case PF_ADDR_ADDRMASK:
782 			/* XXX ignore af? */
783 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
784 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
785 			break;
786 	}
787 
788 	PF_MD5_UPD(pfr, port[0]);
789 	PF_MD5_UPD(pfr, port[1]);
790 	PF_MD5_UPD(pfr, neg);
791 	PF_MD5_UPD(pfr, port_op);
792 }
793 
794 static void
795 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
796 {
797 	u_int16_t x;
798 	u_int32_t y;
799 
800 	pf_hash_rule_addr(ctx, &rule->src);
801 	pf_hash_rule_addr(ctx, &rule->dst);
802 	PF_MD5_UPD_STR(rule, label);
803 	PF_MD5_UPD_STR(rule, ifname);
804 	PF_MD5_UPD_STR(rule, match_tagname);
805 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
806 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
807 	PF_MD5_UPD_HTONL(rule, prob, y);
808 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
809 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
810 	PF_MD5_UPD(rule, uid.op);
811 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
812 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
813 	PF_MD5_UPD(rule, gid.op);
814 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
815 	PF_MD5_UPD(rule, action);
816 	PF_MD5_UPD(rule, direction);
817 	PF_MD5_UPD(rule, af);
818 	PF_MD5_UPD(rule, quick);
819 	PF_MD5_UPD(rule, ifnot);
820 	PF_MD5_UPD(rule, match_tag_not);
821 	PF_MD5_UPD(rule, natpass);
822 	PF_MD5_UPD(rule, keep_state);
823 	PF_MD5_UPD(rule, proto);
824 	PF_MD5_UPD(rule, type);
825 	PF_MD5_UPD(rule, code);
826 	PF_MD5_UPD(rule, flags);
827 	PF_MD5_UPD(rule, flagset);
828 	PF_MD5_UPD(rule, allow_opts);
829 	PF_MD5_UPD(rule, rt);
830 	PF_MD5_UPD(rule, tos);
831 }
832 
833 static int
834 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
835 {
836 	struct pf_ruleset	*rs;
837 	struct pf_rule		*rule, **old_array;
838 	struct pf_rulequeue	*old_rules;
839 	int			 error;
840 	u_int32_t		 old_rcount;
841 
842 	PF_RULES_WASSERT();
843 
844 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
845 		return (EINVAL);
846 	rs = pf_find_ruleset(anchor);
847 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
848 	    ticket != rs->rules[rs_num].inactive.ticket)
849 		return (EBUSY);
850 
851 	/* Calculate checksum for the main ruleset */
852 	if (rs == &pf_main_ruleset) {
853 		error = pf_setup_pfsync_matching(rs);
854 		if (error != 0)
855 			return (error);
856 	}
857 
858 	/* Swap rules, keep the old. */
859 	old_rules = rs->rules[rs_num].active.ptr;
860 	old_rcount = rs->rules[rs_num].active.rcount;
861 	old_array = rs->rules[rs_num].active.ptr_array;
862 
863 	rs->rules[rs_num].active.ptr =
864 	    rs->rules[rs_num].inactive.ptr;
865 	rs->rules[rs_num].active.ptr_array =
866 	    rs->rules[rs_num].inactive.ptr_array;
867 	rs->rules[rs_num].active.rcount =
868 	    rs->rules[rs_num].inactive.rcount;
869 	rs->rules[rs_num].inactive.ptr = old_rules;
870 	rs->rules[rs_num].inactive.ptr_array = old_array;
871 	rs->rules[rs_num].inactive.rcount = old_rcount;
872 
873 	rs->rules[rs_num].active.ticket =
874 	    rs->rules[rs_num].inactive.ticket;
875 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
876 
877 
878 	/* Purge the old rule list. */
879 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
880 		pf_unlink_rule(old_rules, rule);
881 	if (rs->rules[rs_num].inactive.ptr_array)
882 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
883 	rs->rules[rs_num].inactive.ptr_array = NULL;
884 	rs->rules[rs_num].inactive.rcount = 0;
885 	rs->rules[rs_num].inactive.open = 0;
886 	pf_remove_if_empty_ruleset(rs);
887 
888 	return (0);
889 }
890 
891 static int
892 pf_setup_pfsync_matching(struct pf_ruleset *rs)
893 {
894 	MD5_CTX			 ctx;
895 	struct pf_rule		*rule;
896 	int			 rs_cnt;
897 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
898 
899 	MD5Init(&ctx);
900 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
901 		/* XXX PF_RULESET_SCRUB as well? */
902 		if (rs_cnt == PF_RULESET_SCRUB)
903 			continue;
904 
905 		if (rs->rules[rs_cnt].inactive.ptr_array)
906 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
907 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
908 
909 		if (rs->rules[rs_cnt].inactive.rcount) {
910 			rs->rules[rs_cnt].inactive.ptr_array =
911 			    malloc(sizeof(caddr_t) *
912 			    rs->rules[rs_cnt].inactive.rcount,
913 			    M_TEMP, M_NOWAIT);
914 
915 			if (!rs->rules[rs_cnt].inactive.ptr_array)
916 				return (ENOMEM);
917 		}
918 
919 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
920 		    entries) {
921 			pf_hash_rule(&ctx, rule);
922 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
923 		}
924 	}
925 
926 	MD5Final(digest, &ctx);
927 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
928 	return (0);
929 }
930 
931 static int
932 pf_addr_setup(struct pf_ruleset *ruleset, struct pf_addr_wrap *addr,
933     sa_family_t af)
934 {
935 	int error = 0;
936 
937 	switch (addr->type) {
938 	case PF_ADDR_TABLE:
939 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
940 		if (addr->p.tbl == NULL)
941 			error = ENOMEM;
942 		break;
943 	case PF_ADDR_DYNIFTL:
944 		error = pfi_dynaddr_setup(addr, af);
945 		break;
946 	}
947 
948 	return (error);
949 }
950 
951 static void
952 pf_addr_copyout(struct pf_addr_wrap *addr)
953 {
954 
955 	switch (addr->type) {
956 	case PF_ADDR_DYNIFTL:
957 		pfi_dynaddr_copyout(addr);
958 		break;
959 	case PF_ADDR_TABLE:
960 		pf_tbladdr_copyout(addr);
961 		break;
962 	}
963 }
964 
965 static int
966 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
967 {
968 	int			 error = 0;
969 
970 	/* XXX keep in sync with switch() below */
971 	if (securelevel_gt(td->td_ucred, 2))
972 		switch (cmd) {
973 		case DIOCGETRULES:
974 		case DIOCGETRULE:
975 		case DIOCGETADDRS:
976 		case DIOCGETADDR:
977 		case DIOCGETSTATE:
978 		case DIOCSETSTATUSIF:
979 		case DIOCGETSTATUS:
980 		case DIOCCLRSTATUS:
981 		case DIOCNATLOOK:
982 		case DIOCSETDEBUG:
983 		case DIOCGETSTATES:
984 		case DIOCGETTIMEOUT:
985 		case DIOCCLRRULECTRS:
986 		case DIOCGETLIMIT:
987 		case DIOCGETALTQS:
988 		case DIOCGETALTQ:
989 		case DIOCGETQSTATS:
990 		case DIOCGETRULESETS:
991 		case DIOCGETRULESET:
992 		case DIOCRGETTABLES:
993 		case DIOCRGETTSTATS:
994 		case DIOCRCLRTSTATS:
995 		case DIOCRCLRADDRS:
996 		case DIOCRADDADDRS:
997 		case DIOCRDELADDRS:
998 		case DIOCRSETADDRS:
999 		case DIOCRGETADDRS:
1000 		case DIOCRGETASTATS:
1001 		case DIOCRCLRASTATS:
1002 		case DIOCRTSTADDRS:
1003 		case DIOCOSFPGET:
1004 		case DIOCGETSRCNODES:
1005 		case DIOCCLRSRCNODES:
1006 		case DIOCIGETIFACES:
1007 		case DIOCGIFSPEED:
1008 		case DIOCSETIFFLAG:
1009 		case DIOCCLRIFFLAG:
1010 			break;
1011 		case DIOCRCLRTABLES:
1012 		case DIOCRADDTABLES:
1013 		case DIOCRDELTABLES:
1014 		case DIOCRSETTFLAGS:
1015 			if (((struct pfioc_table *)addr)->pfrio_flags &
1016 			    PFR_FLAG_DUMMY)
1017 				break; /* dummy operation ok */
1018 			return (EPERM);
1019 		default:
1020 			return (EPERM);
1021 		}
1022 
1023 	if (!(flags & FWRITE))
1024 		switch (cmd) {
1025 		case DIOCGETRULES:
1026 		case DIOCGETADDRS:
1027 		case DIOCGETADDR:
1028 		case DIOCGETSTATE:
1029 		case DIOCGETSTATUS:
1030 		case DIOCGETSTATES:
1031 		case DIOCGETTIMEOUT:
1032 		case DIOCGETLIMIT:
1033 		case DIOCGETALTQS:
1034 		case DIOCGETALTQ:
1035 		case DIOCGETQSTATS:
1036 		case DIOCGETRULESETS:
1037 		case DIOCGETRULESET:
1038 		case DIOCNATLOOK:
1039 		case DIOCRGETTABLES:
1040 		case DIOCRGETTSTATS:
1041 		case DIOCRGETADDRS:
1042 		case DIOCRGETASTATS:
1043 		case DIOCRTSTADDRS:
1044 		case DIOCOSFPGET:
1045 		case DIOCGETSRCNODES:
1046 		case DIOCIGETIFACES:
1047 		case DIOCGIFSPEED:
1048 			break;
1049 		case DIOCRCLRTABLES:
1050 		case DIOCRADDTABLES:
1051 		case DIOCRDELTABLES:
1052 		case DIOCRCLRTSTATS:
1053 		case DIOCRCLRADDRS:
1054 		case DIOCRADDADDRS:
1055 		case DIOCRDELADDRS:
1056 		case DIOCRSETADDRS:
1057 		case DIOCRSETTFLAGS:
1058 			if (((struct pfioc_table *)addr)->pfrio_flags &
1059 			    PFR_FLAG_DUMMY) {
1060 				flags |= FWRITE; /* need write lock for dummy */
1061 				break; /* dummy operation ok */
1062 			}
1063 			return (EACCES);
1064 		case DIOCGETRULE:
1065 			if (((struct pfioc_rule *)addr)->action ==
1066 			    PF_GET_CLR_CNTR)
1067 				return (EACCES);
1068 			break;
1069 		default:
1070 			return (EACCES);
1071 		}
1072 
1073 	CURVNET_SET(TD_TO_VNET(td));
1074 
1075 	switch (cmd) {
1076 	case DIOCSTART:
1077 		PF_RULES_WLOCK();
1078 		if (V_pf_status.running)
1079 			error = EEXIST;
1080 		else {
1081 			int cpu;
1082 
1083 			PF_RULES_WUNLOCK();
1084 			error = hook_pf();
1085 			if (error) {
1086 				DPFPRINTF(PF_DEBUG_MISC,
1087 				    ("pf: pfil registration failed\n"));
1088 				break;
1089 			}
1090 			PF_RULES_WLOCK();
1091 			V_pf_status.running = 1;
1092 			V_pf_status.since = time_second;
1093 
1094 			CPU_FOREACH(cpu)
1095 				V_pf_stateid[cpu] = time_second;
1096 
1097 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1098 		}
1099 		PF_RULES_WUNLOCK();
1100 		break;
1101 
1102 	case DIOCSTOP:
1103 		PF_RULES_WLOCK();
1104 		if (!V_pf_status.running)
1105 			error = ENOENT;
1106 		else {
1107 			V_pf_status.running = 0;
1108 			PF_RULES_WUNLOCK();
1109 			error = dehook_pf();
1110 			if (error) {
1111 				V_pf_status.running = 1;
1112 				DPFPRINTF(PF_DEBUG_MISC,
1113 				    ("pf: pfil unregistration failed\n"));
1114 			}
1115 			PF_RULES_WLOCK();
1116 			V_pf_status.since = time_second;
1117 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1118 		}
1119 		PF_RULES_WUNLOCK();
1120 		break;
1121 
1122 	case DIOCADDRULE: {
1123 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1124 		struct pf_ruleset	*ruleset;
1125 		struct pf_rule		*rule, *tail;
1126 		struct pf_pooladdr	*pa;
1127 		struct pfi_kif		*kif = NULL;
1128 		int			 rs_num;
1129 
1130 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1131 			error = EINVAL;
1132 			break;
1133 		}
1134 #ifndef INET
1135 		if (pr->rule.af == AF_INET) {
1136 			error = EAFNOSUPPORT;
1137 			break;
1138 		}
1139 #endif /* INET */
1140 #ifndef INET6
1141 		if (pr->rule.af == AF_INET6) {
1142 			error = EAFNOSUPPORT;
1143 			break;
1144 		}
1145 #endif /* INET6 */
1146 
1147 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
1148 		bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1149 		if (rule->ifname[0])
1150 			kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1151 		rule->cuid = td->td_ucred->cr_ruid;
1152 		rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1153 		TAILQ_INIT(&rule->rpool.list);
1154 
1155 #define	ERROUT(x)	{ error = (x); goto DIOCADDRULE_error; }
1156 
1157 		PF_RULES_WLOCK();
1158 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1159 		ruleset = pf_find_ruleset(pr->anchor);
1160 		if (ruleset == NULL)
1161 			ERROUT(EINVAL);
1162 		rs_num = pf_get_ruleset_number(pr->rule.action);
1163 		if (rs_num >= PF_RULESET_MAX)
1164 			ERROUT(EINVAL);
1165 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1166 			DPFPRINTF(PF_DEBUG_MISC,
1167 			    ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
1168 			    ruleset->rules[rs_num].inactive.ticket));
1169 			ERROUT(EBUSY);
1170 		}
1171 		if (pr->pool_ticket != V_ticket_pabuf) {
1172 			DPFPRINTF(PF_DEBUG_MISC,
1173 			    ("pool_ticket: %d != %d\n", pr->pool_ticket,
1174 			    V_ticket_pabuf));
1175 			ERROUT(EBUSY);
1176 		}
1177 
1178 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1179 		    pf_rulequeue);
1180 		if (tail)
1181 			rule->nr = tail->nr + 1;
1182 		else
1183 			rule->nr = 0;
1184 		if (rule->ifname[0]) {
1185 			rule->kif = pfi_kif_attach(kif, rule->ifname);
1186 			pfi_kif_ref(rule->kif);
1187 		} else
1188 			rule->kif = NULL;
1189 
1190 		if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
1191 			error = EBUSY;
1192 
1193 #ifdef ALTQ
1194 		/* set queue IDs */
1195 		if (rule->qname[0] != 0) {
1196 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1197 				error = EBUSY;
1198 			else if (rule->pqname[0] != 0) {
1199 				if ((rule->pqid =
1200 				    pf_qname2qid(rule->pqname)) == 0)
1201 					error = EBUSY;
1202 			} else
1203 				rule->pqid = rule->qid;
1204 		}
1205 #endif
1206 		if (rule->tagname[0])
1207 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1208 				error = EBUSY;
1209 		if (rule->match_tagname[0])
1210 			if ((rule->match_tag =
1211 			    pf_tagname2tag(rule->match_tagname)) == 0)
1212 				error = EBUSY;
1213 		if (rule->rt && !rule->direction)
1214 			error = EINVAL;
1215 		if (!rule->log)
1216 			rule->logif = 0;
1217 		if (rule->logif >= PFLOGIFS_MAX)
1218 			error = EINVAL;
1219 		if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1220 			error = ENOMEM;
1221 		if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1222 			error = ENOMEM;
1223 		if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1224 			error = EINVAL;
1225 		TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1226 			if (pa->addr.type == PF_ADDR_TABLE) {
1227 				pa->addr.p.tbl = pfr_attach_table(ruleset,
1228 				    pa->addr.v.tblname);
1229 				if (pa->addr.p.tbl == NULL)
1230 					error = ENOMEM;
1231 			}
1232 
1233 		if (rule->overload_tblname[0]) {
1234 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1235 			    rule->overload_tblname)) == NULL)
1236 				error = EINVAL;
1237 			else
1238 				rule->overload_tbl->pfrkt_flags |=
1239 				    PFR_TFLAG_ACTIVE;
1240 		}
1241 
1242 		pf_mv_pool(&V_pf_pabuf, &rule->rpool.list);
1243 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1244 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1245 		    (rule->rt > PF_FASTROUTE)) &&
1246 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1247 			error = EINVAL;
1248 
1249 		if (error) {
1250 			pf_free_rule(rule);
1251 			PF_RULES_WUNLOCK();
1252 			break;
1253 		}
1254 
1255 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1256 		rule->evaluations = rule->packets[0] = rule->packets[1] =
1257 		    rule->bytes[0] = rule->bytes[1] = 0;
1258 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1259 		    rule, entries);
1260 		ruleset->rules[rs_num].inactive.rcount++;
1261 		PF_RULES_WUNLOCK();
1262 		break;
1263 
1264 #undef ERROUT
1265 DIOCADDRULE_error:
1266 		PF_RULES_WUNLOCK();
1267 		free(rule, M_PFRULE);
1268 		if (kif)
1269 			free(kif, PFI_MTYPE);
1270 		break;
1271 	}
1272 
1273 	case DIOCGETRULES: {
1274 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1275 		struct pf_ruleset	*ruleset;
1276 		struct pf_rule		*tail;
1277 		int			 rs_num;
1278 
1279 		PF_RULES_WLOCK();
1280 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1281 		ruleset = pf_find_ruleset(pr->anchor);
1282 		if (ruleset == NULL) {
1283 			PF_RULES_WUNLOCK();
1284 			error = EINVAL;
1285 			break;
1286 		}
1287 		rs_num = pf_get_ruleset_number(pr->rule.action);
1288 		if (rs_num >= PF_RULESET_MAX) {
1289 			PF_RULES_WUNLOCK();
1290 			error = EINVAL;
1291 			break;
1292 		}
1293 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1294 		    pf_rulequeue);
1295 		if (tail)
1296 			pr->nr = tail->nr + 1;
1297 		else
1298 			pr->nr = 0;
1299 		pr->ticket = ruleset->rules[rs_num].active.ticket;
1300 		PF_RULES_WUNLOCK();
1301 		break;
1302 	}
1303 
1304 	case DIOCGETRULE: {
1305 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1306 		struct pf_ruleset	*ruleset;
1307 		struct pf_rule		*rule;
1308 		int			 rs_num, i;
1309 
1310 		PF_RULES_WLOCK();
1311 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1312 		ruleset = pf_find_ruleset(pr->anchor);
1313 		if (ruleset == NULL) {
1314 			PF_RULES_WUNLOCK();
1315 			error = EINVAL;
1316 			break;
1317 		}
1318 		rs_num = pf_get_ruleset_number(pr->rule.action);
1319 		if (rs_num >= PF_RULESET_MAX) {
1320 			PF_RULES_WUNLOCK();
1321 			error = EINVAL;
1322 			break;
1323 		}
1324 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1325 			PF_RULES_WUNLOCK();
1326 			error = EBUSY;
1327 			break;
1328 		}
1329 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1330 		while ((rule != NULL) && (rule->nr != pr->nr))
1331 			rule = TAILQ_NEXT(rule, entries);
1332 		if (rule == NULL) {
1333 			PF_RULES_WUNLOCK();
1334 			error = EBUSY;
1335 			break;
1336 		}
1337 		bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1338 		if (pf_anchor_copyout(ruleset, rule, pr)) {
1339 			PF_RULES_WUNLOCK();
1340 			error = EBUSY;
1341 			break;
1342 		}
1343 		pf_addr_copyout(&pr->rule.src.addr);
1344 		pf_addr_copyout(&pr->rule.dst.addr);
1345 		for (i = 0; i < PF_SKIP_COUNT; ++i)
1346 			if (rule->skip[i].ptr == NULL)
1347 				pr->rule.skip[i].nr = -1;
1348 			else
1349 				pr->rule.skip[i].nr =
1350 				    rule->skip[i].ptr->nr;
1351 
1352 		if (pr->action == PF_GET_CLR_CNTR) {
1353 			rule->evaluations = 0;
1354 			rule->packets[0] = rule->packets[1] = 0;
1355 			rule->bytes[0] = rule->bytes[1] = 0;
1356 			rule->states_tot = 0;
1357 		}
1358 		PF_RULES_WUNLOCK();
1359 		break;
1360 	}
1361 
1362 	case DIOCCHANGERULE: {
1363 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1364 		struct pf_ruleset	*ruleset;
1365 		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1366 		struct pfi_kif		*kif = NULL;
1367 		struct pf_pooladdr	*pa;
1368 		u_int32_t		 nr = 0;
1369 		int			 rs_num;
1370 
1371 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1372 		    pcr->action > PF_CHANGE_GET_TICKET) {
1373 			error = EINVAL;
1374 			break;
1375 		}
1376 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1377 			error = EINVAL;
1378 			break;
1379 		}
1380 
1381 		if (pcr->action != PF_CHANGE_REMOVE) {
1382 #ifndef INET
1383 			if (pcr->rule.af == AF_INET) {
1384 				error = EAFNOSUPPORT;
1385 				break;
1386 			}
1387 #endif /* INET */
1388 #ifndef INET6
1389 			if (pcr->rule.af == AF_INET6) {
1390 				error = EAFNOSUPPORT;
1391 				break;
1392 			}
1393 #endif /* INET6 */
1394 			newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
1395 			bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1396 			newrule->cuid = td->td_ucred->cr_ruid;
1397 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1398 			TAILQ_INIT(&newrule->rpool.list);
1399 			/* Initialize refcounting. */
1400 			newrule->states_cur = 0;
1401 			newrule->entries.tqe_prev = NULL;
1402 
1403 			if (newrule->ifname[0])
1404 				kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1405 		}
1406 
1407 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGERULE_error; }
1408 
1409 		PF_RULES_WLOCK();
1410 		if (!(pcr->action == PF_CHANGE_REMOVE ||
1411 		    pcr->action == PF_CHANGE_GET_TICKET) &&
1412 		    pcr->pool_ticket != V_ticket_pabuf)
1413 			ERROUT(EBUSY);
1414 
1415 		ruleset = pf_find_ruleset(pcr->anchor);
1416 		if (ruleset == NULL)
1417 			ERROUT(EINVAL);
1418 
1419 		rs_num = pf_get_ruleset_number(pcr->rule.action);
1420 		if (rs_num >= PF_RULESET_MAX)
1421 			ERROUT(EINVAL);
1422 
1423 		if (pcr->action == PF_CHANGE_GET_TICKET) {
1424 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1425 			ERROUT(0);
1426 		} else if (pcr->ticket !=
1427 			    ruleset->rules[rs_num].active.ticket)
1428 				ERROUT(EINVAL);
1429 
1430 		if (pcr->action != PF_CHANGE_REMOVE) {
1431 			if (newrule->ifname[0]) {
1432 				newrule->kif = pfi_kif_attach(kif,
1433 				    newrule->ifname);
1434 				pfi_kif_ref(newrule->kif);
1435 			} else
1436 				newrule->kif = NULL;
1437 
1438 			if (newrule->rtableid > 0 &&
1439 			    newrule->rtableid >= rt_numfibs)
1440 				error = EBUSY;
1441 
1442 #ifdef ALTQ
1443 			/* set queue IDs */
1444 			if (newrule->qname[0] != 0) {
1445 				if ((newrule->qid =
1446 				    pf_qname2qid(newrule->qname)) == 0)
1447 					error = EBUSY;
1448 				else if (newrule->pqname[0] != 0) {
1449 					if ((newrule->pqid =
1450 					    pf_qname2qid(newrule->pqname)) == 0)
1451 						error = EBUSY;
1452 				} else
1453 					newrule->pqid = newrule->qid;
1454 			}
1455 #endif /* ALTQ */
1456 			if (newrule->tagname[0])
1457 				if ((newrule->tag =
1458 				    pf_tagname2tag(newrule->tagname)) == 0)
1459 					error = EBUSY;
1460 			if (newrule->match_tagname[0])
1461 				if ((newrule->match_tag = pf_tagname2tag(
1462 				    newrule->match_tagname)) == 0)
1463 					error = EBUSY;
1464 			if (newrule->rt && !newrule->direction)
1465 				error = EINVAL;
1466 			if (!newrule->log)
1467 				newrule->logif = 0;
1468 			if (newrule->logif >= PFLOGIFS_MAX)
1469 				error = EINVAL;
1470 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
1471 				error = ENOMEM;
1472 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
1473 				error = ENOMEM;
1474 			if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1475 				error = EINVAL;
1476 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1477 				if (pa->addr.type == PF_ADDR_TABLE) {
1478 					pa->addr.p.tbl =
1479 					    pfr_attach_table(ruleset,
1480 					    pa->addr.v.tblname);
1481 					if (pa->addr.p.tbl == NULL)
1482 						error = ENOMEM;
1483 				}
1484 
1485 			if (newrule->overload_tblname[0]) {
1486 				if ((newrule->overload_tbl = pfr_attach_table(
1487 				    ruleset, newrule->overload_tblname)) ==
1488 				    NULL)
1489 					error = EINVAL;
1490 				else
1491 					newrule->overload_tbl->pfrkt_flags |=
1492 					    PFR_TFLAG_ACTIVE;
1493 			}
1494 
1495 			pf_mv_pool(&V_pf_pabuf, &newrule->rpool.list);
1496 			if (((((newrule->action == PF_NAT) ||
1497 			    (newrule->action == PF_RDR) ||
1498 			    (newrule->action == PF_BINAT) ||
1499 			    (newrule->rt > PF_FASTROUTE)) &&
1500 			    !newrule->anchor)) &&
1501 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1502 				error = EINVAL;
1503 
1504 			if (error) {
1505 				pf_free_rule(newrule);
1506 				PF_RULES_WUNLOCK();
1507 				break;
1508 			}
1509 
1510 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1511 			newrule->evaluations = 0;
1512 			newrule->packets[0] = newrule->packets[1] = 0;
1513 			newrule->bytes[0] = newrule->bytes[1] = 0;
1514 		}
1515 		pf_empty_pool(&V_pf_pabuf);
1516 
1517 		if (pcr->action == PF_CHANGE_ADD_HEAD)
1518 			oldrule = TAILQ_FIRST(
1519 			    ruleset->rules[rs_num].active.ptr);
1520 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1521 			oldrule = TAILQ_LAST(
1522 			    ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1523 		else {
1524 			oldrule = TAILQ_FIRST(
1525 			    ruleset->rules[rs_num].active.ptr);
1526 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1527 				oldrule = TAILQ_NEXT(oldrule, entries);
1528 			if (oldrule == NULL) {
1529 				if (newrule != NULL)
1530 					pf_free_rule(newrule);
1531 				PF_RULES_WUNLOCK();
1532 				error = EINVAL;
1533 				break;
1534 			}
1535 		}
1536 
1537 		if (pcr->action == PF_CHANGE_REMOVE) {
1538 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
1539 			    oldrule);
1540 			ruleset->rules[rs_num].active.rcount--;
1541 		} else {
1542 			if (oldrule == NULL)
1543 				TAILQ_INSERT_TAIL(
1544 				    ruleset->rules[rs_num].active.ptr,
1545 				    newrule, entries);
1546 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1547 			    pcr->action == PF_CHANGE_ADD_BEFORE)
1548 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1549 			else
1550 				TAILQ_INSERT_AFTER(
1551 				    ruleset->rules[rs_num].active.ptr,
1552 				    oldrule, newrule, entries);
1553 			ruleset->rules[rs_num].active.rcount++;
1554 		}
1555 
1556 		nr = 0;
1557 		TAILQ_FOREACH(oldrule,
1558 		    ruleset->rules[rs_num].active.ptr, entries)
1559 			oldrule->nr = nr++;
1560 
1561 		ruleset->rules[rs_num].active.ticket++;
1562 
1563 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1564 		pf_remove_if_empty_ruleset(ruleset);
1565 
1566 		PF_RULES_WUNLOCK();
1567 		break;
1568 
1569 #undef ERROUT
1570 DIOCCHANGERULE_error:
1571 		PF_RULES_WUNLOCK();
1572 		if (newrule != NULL)
1573 			free(newrule, M_PFRULE);
1574 		if (kif != NULL)
1575 			free(kif, PFI_MTYPE);
1576 		break;
1577 	}
1578 
1579 	case DIOCCLRSTATES: {
1580 		struct pf_state		*s;
1581 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1582 		u_int			 i, killed = 0;
1583 
1584 		for (i = 0; i <= V_pf_hashmask; i++) {
1585 			struct pf_idhash *ih = &V_pf_idhash[i];
1586 
1587 relock_DIOCCLRSTATES:
1588 			PF_HASHROW_LOCK(ih);
1589 			LIST_FOREACH(s, &ih->states, entry)
1590 				if (!psk->psk_ifname[0] ||
1591 				    !strcmp(psk->psk_ifname,
1592 				    s->kif->pfik_name)) {
1593 					/*
1594 					 * Don't send out individual
1595 					 * delete messages.
1596 					 */
1597 					s->state_flags |= PFSTATE_NOSYNC;
1598 					pf_unlink_state(s, PF_ENTER_LOCKED);
1599 					killed++;
1600 					goto relock_DIOCCLRSTATES;
1601 				}
1602 			PF_HASHROW_UNLOCK(ih);
1603 		}
1604 		psk->psk_killed = killed;
1605 		if (pfsync_clear_states_ptr != NULL)
1606 			pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
1607 		break;
1608 	}
1609 
1610 	case DIOCKILLSTATES: {
1611 		struct pf_state		*s;
1612 		struct pf_state_key	*sk;
1613 		struct pf_addr		*srcaddr, *dstaddr;
1614 		u_int16_t		 srcport, dstport;
1615 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
1616 		u_int			 i, killed = 0;
1617 
1618 		if (psk->psk_pfcmp.id) {
1619 			if (psk->psk_pfcmp.creatorid == 0)
1620 				psk->psk_pfcmp.creatorid = V_pf_status.hostid;
1621 			if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
1622 			    psk->psk_pfcmp.creatorid))) {
1623 				pf_unlink_state(s, PF_ENTER_LOCKED);
1624 				psk->psk_killed = 1;
1625 			}
1626 			break;
1627 		}
1628 
1629 		for (i = 0; i <= V_pf_hashmask; i++) {
1630 			struct pf_idhash *ih = &V_pf_idhash[i];
1631 
1632 relock_DIOCKILLSTATES:
1633 			PF_HASHROW_LOCK(ih);
1634 			LIST_FOREACH(s, &ih->states, entry) {
1635 				sk = s->key[PF_SK_WIRE];
1636 				if (s->direction == PF_OUT) {
1637 					srcaddr = &sk->addr[1];
1638 					dstaddr = &sk->addr[0];
1639 					srcport = sk->port[0];
1640 					dstport = sk->port[0];
1641 				} else {
1642 					srcaddr = &sk->addr[0];
1643 					dstaddr = &sk->addr[1];
1644 					srcport = sk->port[0];
1645 					dstport = sk->port[0];
1646 				}
1647 
1648 				if ((!psk->psk_af || sk->af == psk->psk_af)
1649 				    && (!psk->psk_proto || psk->psk_proto ==
1650 				    sk->proto) &&
1651 				    PF_MATCHA(psk->psk_src.neg,
1652 				    &psk->psk_src.addr.v.a.addr,
1653 				    &psk->psk_src.addr.v.a.mask,
1654 				    srcaddr, sk->af) &&
1655 				    PF_MATCHA(psk->psk_dst.neg,
1656 				    &psk->psk_dst.addr.v.a.addr,
1657 				    &psk->psk_dst.addr.v.a.mask,
1658 				    dstaddr, sk->af) &&
1659 				    (psk->psk_src.port_op == 0 ||
1660 				    pf_match_port(psk->psk_src.port_op,
1661 				    psk->psk_src.port[0], psk->psk_src.port[1],
1662 				    srcport)) &&
1663 				    (psk->psk_dst.port_op == 0 ||
1664 				    pf_match_port(psk->psk_dst.port_op,
1665 				    psk->psk_dst.port[0], psk->psk_dst.port[1],
1666 				    dstport)) &&
1667 				    (!psk->psk_label[0] ||
1668 				    (s->rule.ptr->label[0] &&
1669 				    !strcmp(psk->psk_label,
1670 				    s->rule.ptr->label))) &&
1671 				    (!psk->psk_ifname[0] ||
1672 				    !strcmp(psk->psk_ifname,
1673 				    s->kif->pfik_name))) {
1674 					pf_unlink_state(s, PF_ENTER_LOCKED);
1675 					killed++;
1676 					goto relock_DIOCKILLSTATES;
1677 				}
1678 			}
1679 			PF_HASHROW_UNLOCK(ih);
1680 		}
1681 		psk->psk_killed = killed;
1682 		break;
1683 	}
1684 
1685 	case DIOCADDSTATE: {
1686 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1687 		struct pfsync_state	*sp = &ps->state;
1688 
1689 		if (sp->timeout >= PFTM_MAX) {
1690 			error = EINVAL;
1691 			break;
1692 		}
1693 		if (pfsync_state_import_ptr != NULL) {
1694 			PF_RULES_RLOCK();
1695 			error = pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
1696 			PF_RULES_RUNLOCK();
1697 		} else
1698 			error = EOPNOTSUPP;
1699 		break;
1700 	}
1701 
1702 	case DIOCGETSTATE: {
1703 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1704 		struct pf_state		*s;
1705 
1706 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
1707 		if (s == NULL) {
1708 			error = ENOENT;
1709 			break;
1710 		}
1711 
1712 		pfsync_state_export(&ps->state, s);
1713 		PF_STATE_UNLOCK(s);
1714 		break;
1715 	}
1716 
1717 	case DIOCGETSTATES: {
1718 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
1719 		struct pf_state		*s;
1720 		struct pfsync_state	*pstore, *p;
1721 		int i, nr;
1722 
1723 		if (ps->ps_len == 0) {
1724 			nr = uma_zone_get_cur(V_pf_state_z);
1725 			ps->ps_len = sizeof(struct pfsync_state) * nr;
1726 			break;
1727 		}
1728 
1729 		p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK);
1730 		nr = 0;
1731 
1732 		for (i = 0; i <= V_pf_hashmask; i++) {
1733 			struct pf_idhash *ih = &V_pf_idhash[i];
1734 
1735 			PF_HASHROW_LOCK(ih);
1736 			LIST_FOREACH(s, &ih->states, entry) {
1737 
1738 				if (s->timeout == PFTM_UNLINKED)
1739 					continue;
1740 
1741 				if ((nr+1) * sizeof(*p) > ps->ps_len) {
1742 					PF_HASHROW_UNLOCK(ih);
1743 					goto DIOCGETSTATES_full;
1744 				}
1745 				pfsync_state_export(p, s);
1746 				p++;
1747 				nr++;
1748 			}
1749 			PF_HASHROW_UNLOCK(ih);
1750 		}
1751 DIOCGETSTATES_full:
1752 		error = copyout(pstore, ps->ps_states,
1753 		    sizeof(struct pfsync_state) * nr);
1754 		if (error) {
1755 			free(pstore, M_TEMP);
1756 			break;
1757 		}
1758 		ps->ps_len = sizeof(struct pfsync_state) * nr;
1759 		free(pstore, M_TEMP);
1760 
1761 		break;
1762 	}
1763 
1764 	case DIOCGETSTATUS: {
1765 		struct pf_status *s = (struct pf_status *)addr;
1766 		PF_RULES_RLOCK();
1767 		bcopy(&V_pf_status, s, sizeof(struct pf_status));
1768 		pfi_update_status(s->ifname, s);
1769 		PF_RULES_RUNLOCK();
1770 		break;
1771 	}
1772 
1773 	case DIOCSETSTATUSIF: {
1774 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
1775 
1776 		if (pi->ifname[0] == 0) {
1777 			bzero(V_pf_status.ifname, IFNAMSIZ);
1778 			break;
1779 		}
1780 		PF_RULES_WLOCK();
1781 		strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
1782 		PF_RULES_WUNLOCK();
1783 		break;
1784 	}
1785 
1786 	case DIOCCLRSTATUS: {
1787 		PF_RULES_WLOCK();
1788 		bzero(V_pf_status.counters, sizeof(V_pf_status.counters));
1789 		bzero(V_pf_status.fcounters, sizeof(V_pf_status.fcounters));
1790 		bzero(V_pf_status.scounters, sizeof(V_pf_status.scounters));
1791 		V_pf_status.since = time_second;
1792 		if (*V_pf_status.ifname)
1793 			pfi_update_status(V_pf_status.ifname, NULL);
1794 		PF_RULES_WUNLOCK();
1795 		break;
1796 	}
1797 
1798 	case DIOCNATLOOK: {
1799 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
1800 		struct pf_state_key	*sk;
1801 		struct pf_state		*state;
1802 		struct pf_state_key_cmp	 key;
1803 		int			 m = 0, direction = pnl->direction;
1804 		int			 sidx, didx;
1805 
1806 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
1807 		sidx = (direction == PF_IN) ? 1 : 0;
1808 		didx = (direction == PF_IN) ? 0 : 1;
1809 
1810 		if (!pnl->proto ||
1811 		    PF_AZERO(&pnl->saddr, pnl->af) ||
1812 		    PF_AZERO(&pnl->daddr, pnl->af) ||
1813 		    ((pnl->proto == IPPROTO_TCP ||
1814 		    pnl->proto == IPPROTO_UDP) &&
1815 		    (!pnl->dport || !pnl->sport)))
1816 			error = EINVAL;
1817 		else {
1818 			bzero(&key, sizeof(key));
1819 			key.af = pnl->af;
1820 			key.proto = pnl->proto;
1821 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
1822 			key.port[sidx] = pnl->sport;
1823 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
1824 			key.port[didx] = pnl->dport;
1825 
1826 			state = pf_find_state_all(&key, direction, &m);
1827 
1828 			if (m > 1)
1829 				error = E2BIG;	/* more than one state */
1830 			else if (state != NULL) {
1831 				/* XXXGL: not locked read */
1832 				sk = state->key[sidx];
1833 				PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
1834 				pnl->rsport = sk->port[sidx];
1835 				PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
1836 				pnl->rdport = sk->port[didx];
1837 			} else
1838 				error = ENOENT;
1839 		}
1840 		break;
1841 	}
1842 
1843 	case DIOCSETTIMEOUT: {
1844 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1845 		int		 old;
1846 
1847 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
1848 		    pt->seconds < 0) {
1849 			error = EINVAL;
1850 			break;
1851 		}
1852 		PF_RULES_WLOCK();
1853 		old = V_pf_default_rule.timeout[pt->timeout];
1854 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
1855 			pt->seconds = 1;
1856 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
1857 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
1858 			wakeup(pf_purge_thread);
1859 		pt->seconds = old;
1860 		PF_RULES_WUNLOCK();
1861 		break;
1862 	}
1863 
1864 	case DIOCGETTIMEOUT: {
1865 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1866 
1867 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
1868 			error = EINVAL;
1869 			break;
1870 		}
1871 		PF_RULES_RLOCK();
1872 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
1873 		PF_RULES_RUNLOCK();
1874 		break;
1875 	}
1876 
1877 	case DIOCGETLIMIT: {
1878 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1879 
1880 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
1881 			error = EINVAL;
1882 			break;
1883 		}
1884 		PF_RULES_RLOCK();
1885 		pl->limit = V_pf_limits[pl->index].limit;
1886 		PF_RULES_RUNLOCK();
1887 		break;
1888 	}
1889 
1890 	case DIOCSETLIMIT: {
1891 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1892 		int			 old_limit;
1893 
1894 		PF_RULES_WLOCK();
1895 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
1896 		    V_pf_limits[pl->index].zone == NULL) {
1897 			PF_RULES_WUNLOCK();
1898 			error = EINVAL;
1899 			break;
1900 		}
1901 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
1902 		old_limit = V_pf_limits[pl->index].limit;
1903 		V_pf_limits[pl->index].limit = pl->limit;
1904 		pl->limit = old_limit;
1905 		PF_RULES_WUNLOCK();
1906 		break;
1907 	}
1908 
1909 	case DIOCSETDEBUG: {
1910 		u_int32_t	*level = (u_int32_t *)addr;
1911 
1912 		PF_RULES_WLOCK();
1913 		V_pf_status.debug = *level;
1914 		PF_RULES_WUNLOCK();
1915 		break;
1916 	}
1917 
1918 	case DIOCCLRRULECTRS: {
1919 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
1920 		struct pf_ruleset	*ruleset = &pf_main_ruleset;
1921 		struct pf_rule		*rule;
1922 
1923 		PF_RULES_WLOCK();
1924 		TAILQ_FOREACH(rule,
1925 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
1926 			rule->evaluations = 0;
1927 			rule->packets[0] = rule->packets[1] = 0;
1928 			rule->bytes[0] = rule->bytes[1] = 0;
1929 		}
1930 		PF_RULES_WUNLOCK();
1931 		break;
1932 	}
1933 
1934 	case DIOCGIFSPEED: {
1935 		struct pf_ifspeed	*psp = (struct pf_ifspeed *)addr;
1936 		struct pf_ifspeed	ps;
1937 		struct ifnet		*ifp;
1938 
1939 		if (psp->ifname[0] != 0) {
1940 			/* Can we completely trust user-land? */
1941 			strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
1942 			ifp = ifunit(ps.ifname);
1943 			if (ifp != NULL)
1944 				psp->baudrate = ifp->if_baudrate;
1945 			else
1946 				error = EINVAL;
1947 		} else
1948 			error = EINVAL;
1949 		break;
1950 	}
1951 
1952 #ifdef ALTQ
1953 	case DIOCSTARTALTQ: {
1954 		struct pf_altq		*altq;
1955 
1956 		PF_RULES_WLOCK();
1957 		/* enable all altq interfaces on active list */
1958 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1959 			if (altq->qname[0] == 0 && (altq->local_flags &
1960 			    PFALTQ_FLAG_IF_REMOVED) == 0) {
1961 				error = pf_enable_altq(altq);
1962 				if (error != 0)
1963 					break;
1964 			}
1965 		}
1966 		if (error == 0)
1967 			V_pf_altq_running = 1;
1968 		PF_RULES_WUNLOCK();
1969 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
1970 		break;
1971 	}
1972 
1973 	case DIOCSTOPALTQ: {
1974 		struct pf_altq		*altq;
1975 
1976 		PF_RULES_WLOCK();
1977 		/* disable all altq interfaces on active list */
1978 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1979 			if (altq->qname[0] == 0 && (altq->local_flags &
1980 			    PFALTQ_FLAG_IF_REMOVED) == 0) {
1981 				error = pf_disable_altq(altq);
1982 				if (error != 0)
1983 					break;
1984 			}
1985 		}
1986 		if (error == 0)
1987 			V_pf_altq_running = 0;
1988 		PF_RULES_WUNLOCK();
1989 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
1990 		break;
1991 	}
1992 
1993 	case DIOCADDALTQ: {
1994 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
1995 		struct pf_altq		*altq, *a;
1996 		struct ifnet		*ifp;
1997 
1998 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK);
1999 		bcopy(&pa->altq, altq, sizeof(struct pf_altq));
2000 		altq->local_flags = 0;
2001 
2002 		PF_RULES_WLOCK();
2003 		if (pa->ticket != V_ticket_altqs_inactive) {
2004 			PF_RULES_WUNLOCK();
2005 			free(altq, M_PFALTQ);
2006 			error = EBUSY;
2007 			break;
2008 		}
2009 
2010 		/*
2011 		 * if this is for a queue, find the discipline and
2012 		 * copy the necessary fields
2013 		 */
2014 		if (altq->qname[0] != 0) {
2015 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2016 				PF_RULES_WUNLOCK();
2017 				error = EBUSY;
2018 				free(altq, M_PFALTQ);
2019 				break;
2020 			}
2021 			altq->altq_disc = NULL;
2022 			TAILQ_FOREACH(a, V_pf_altqs_inactive, entries) {
2023 				if (strncmp(a->ifname, altq->ifname,
2024 				    IFNAMSIZ) == 0 && a->qname[0] == 0) {
2025 					altq->altq_disc = a->altq_disc;
2026 					break;
2027 				}
2028 			}
2029 		}
2030 
2031 		if ((ifp = ifunit(altq->ifname)) == NULL)
2032 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
2033 		else
2034 			error = altq_add(altq);
2035 
2036 		if (error) {
2037 			PF_RULES_WUNLOCK();
2038 			free(altq, M_PFALTQ);
2039 			break;
2040 		}
2041 
2042 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
2043 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2044 		PF_RULES_WUNLOCK();
2045 		break;
2046 	}
2047 
2048 	case DIOCGETALTQS: {
2049 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2050 		struct pf_altq		*altq;
2051 
2052 		PF_RULES_RLOCK();
2053 		pa->nr = 0;
2054 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
2055 			pa->nr++;
2056 		pa->ticket = V_ticket_altqs_active;
2057 		PF_RULES_RUNLOCK();
2058 		break;
2059 	}
2060 
2061 	case DIOCGETALTQ: {
2062 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2063 		struct pf_altq		*altq;
2064 		u_int32_t		 nr;
2065 
2066 		PF_RULES_RLOCK();
2067 		if (pa->ticket != V_ticket_altqs_active) {
2068 			PF_RULES_RUNLOCK();
2069 			error = EBUSY;
2070 			break;
2071 		}
2072 		nr = 0;
2073 		altq = TAILQ_FIRST(V_pf_altqs_active);
2074 		while ((altq != NULL) && (nr < pa->nr)) {
2075 			altq = TAILQ_NEXT(altq, entries);
2076 			nr++;
2077 		}
2078 		if (altq == NULL) {
2079 			PF_RULES_RUNLOCK();
2080 			error = EBUSY;
2081 			break;
2082 		}
2083 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2084 		PF_RULES_RUNLOCK();
2085 		break;
2086 	}
2087 
2088 	case DIOCCHANGEALTQ:
2089 		/* CHANGEALTQ not supported yet! */
2090 		error = ENODEV;
2091 		break;
2092 
2093 	case DIOCGETQSTATS: {
2094 		struct pfioc_qstats	*pq = (struct pfioc_qstats *)addr;
2095 		struct pf_altq		*altq;
2096 		u_int32_t		 nr;
2097 		int			 nbytes;
2098 
2099 		PF_RULES_RLOCK();
2100 		if (pq->ticket != V_ticket_altqs_active) {
2101 			PF_RULES_RUNLOCK();
2102 			error = EBUSY;
2103 			break;
2104 		}
2105 		nbytes = pq->nbytes;
2106 		nr = 0;
2107 		altq = TAILQ_FIRST(V_pf_altqs_active);
2108 		while ((altq != NULL) && (nr < pq->nr)) {
2109 			altq = TAILQ_NEXT(altq, entries);
2110 			nr++;
2111 		}
2112 		if (altq == NULL) {
2113 			PF_RULES_RUNLOCK();
2114 			error = EBUSY;
2115 			break;
2116 		}
2117 
2118 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
2119 			PF_RULES_RUNLOCK();
2120 			error = ENXIO;
2121 			break;
2122 		}
2123 		PF_RULES_RUNLOCK();
2124 		error = altq_getqstats(altq, pq->buf, &nbytes);
2125 		if (error == 0) {
2126 			pq->scheduler = altq->scheduler;
2127 			pq->nbytes = nbytes;
2128 		}
2129 		break;
2130 	}
2131 #endif /* ALTQ */
2132 
2133 	case DIOCBEGINADDRS: {
2134 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2135 
2136 		PF_RULES_WLOCK();
2137 		pf_empty_pool(&V_pf_pabuf);
2138 		pp->ticket = ++V_ticket_pabuf;
2139 		PF_RULES_WUNLOCK();
2140 		break;
2141 	}
2142 
2143 	case DIOCADDADDR: {
2144 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2145 		struct pf_pooladdr	*pa;
2146 		struct pfi_kif		*kif = NULL;
2147 
2148 #ifndef INET
2149 		if (pp->af == AF_INET) {
2150 			error = EAFNOSUPPORT;
2151 			break;
2152 		}
2153 #endif /* INET */
2154 #ifndef INET6
2155 		if (pp->af == AF_INET6) {
2156 			error = EAFNOSUPPORT;
2157 			break;
2158 		}
2159 #endif /* INET6 */
2160 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2161 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2162 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2163 			error = EINVAL;
2164 			break;
2165 		}
2166 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2167 		bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2168 		if (pa->ifname[0])
2169 			kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2170 		PF_RULES_WLOCK();
2171 		if (pp->ticket != V_ticket_pabuf) {
2172 			PF_RULES_WUNLOCK();
2173 			if (pa->ifname[0])
2174 				free(kif, PFI_MTYPE);
2175 			free(pa, M_PFRULE);
2176 			error = EBUSY;
2177 			break;
2178 		}
2179 		if (pa->ifname[0]) {
2180 			pa->kif = pfi_kif_attach(kif, pa->ifname);
2181 			pfi_kif_ref(pa->kif);
2182 		} else
2183 			pa->kif = NULL;
2184 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2185 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2186 			if (pa->ifname[0])
2187 				pfi_kif_unref(pa->kif);
2188 			PF_RULES_WUNLOCK();
2189 			free(pa, M_PFRULE);
2190 			break;
2191 		}
2192 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2193 		PF_RULES_WUNLOCK();
2194 		break;
2195 	}
2196 
2197 	case DIOCGETADDRS: {
2198 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2199 		struct pf_pool		*pool;
2200 		struct pf_pooladdr	*pa;
2201 
2202 		PF_RULES_RLOCK();
2203 		pp->nr = 0;
2204 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2205 		    pp->r_num, 0, 1, 0);
2206 		if (pool == NULL) {
2207 			PF_RULES_RUNLOCK();
2208 			error = EBUSY;
2209 			break;
2210 		}
2211 		TAILQ_FOREACH(pa, &pool->list, entries)
2212 			pp->nr++;
2213 		PF_RULES_RUNLOCK();
2214 		break;
2215 	}
2216 
2217 	case DIOCGETADDR: {
2218 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2219 		struct pf_pool		*pool;
2220 		struct pf_pooladdr	*pa;
2221 		u_int32_t		 nr = 0;
2222 
2223 		PF_RULES_RLOCK();
2224 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2225 		    pp->r_num, 0, 1, 1);
2226 		if (pool == NULL) {
2227 			PF_RULES_RUNLOCK();
2228 			error = EBUSY;
2229 			break;
2230 		}
2231 		pa = TAILQ_FIRST(&pool->list);
2232 		while ((pa != NULL) && (nr < pp->nr)) {
2233 			pa = TAILQ_NEXT(pa, entries);
2234 			nr++;
2235 		}
2236 		if (pa == NULL) {
2237 			PF_RULES_RUNLOCK();
2238 			error = EBUSY;
2239 			break;
2240 		}
2241 		bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2242 		pf_addr_copyout(&pp->addr.addr);
2243 		PF_RULES_RUNLOCK();
2244 		break;
2245 	}
2246 
2247 	case DIOCCHANGEADDR: {
2248 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
2249 		struct pf_pool		*pool;
2250 		struct pf_pooladdr	*oldpa = NULL, *newpa = NULL;
2251 		struct pf_ruleset	*ruleset;
2252 		struct pfi_kif		*kif = NULL;
2253 
2254 		if (pca->action < PF_CHANGE_ADD_HEAD ||
2255 		    pca->action > PF_CHANGE_REMOVE) {
2256 			error = EINVAL;
2257 			break;
2258 		}
2259 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2260 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2261 		    pca->addr.addr.type != PF_ADDR_TABLE) {
2262 			error = EINVAL;
2263 			break;
2264 		}
2265 
2266 		if (pca->action != PF_CHANGE_REMOVE) {
2267 #ifndef INET
2268 			if (pca->af == AF_INET) {
2269 				error = EAFNOSUPPORT;
2270 				break;
2271 			}
2272 #endif /* INET */
2273 #ifndef INET6
2274 			if (pca->af == AF_INET6) {
2275 				error = EAFNOSUPPORT;
2276 				break;
2277 			}
2278 #endif /* INET6 */
2279 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
2280 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2281 			if (newpa->ifname[0])
2282 				kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2283 		}
2284 
2285 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGEADDR_error; }
2286 		PF_RULES_WLOCK();
2287 		ruleset = pf_find_ruleset(pca->anchor);
2288 		if (ruleset == NULL)
2289 			ERROUT(EBUSY);
2290 
2291 		pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2292 		    pca->r_num, pca->r_last, 1, 1);
2293 		if (pool == NULL)
2294 			ERROUT(EBUSY);
2295 
2296 		if (pca->action != PF_CHANGE_REMOVE) {
2297 			if (newpa->ifname[0]) {
2298 				newpa->kif = pfi_kif_attach(kif, newpa->ifname);
2299 				pfi_kif_ref(newpa->kif);
2300 			} else
2301 				newpa->kif = NULL;
2302 
2303 			switch (newpa->addr.type) {
2304 			case PF_ADDR_DYNIFTL:
2305 				error = pfi_dynaddr_setup(&newpa->addr,
2306 				    pca->af);
2307 				break;
2308 			case PF_ADDR_TABLE:
2309 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
2310 				    newpa->addr.v.tblname);
2311 				if (newpa->addr.p.tbl == NULL)
2312 					error = ENOMEM;
2313 				break;
2314 			}
2315 			if (error) {
2316 				if (newpa->kif)
2317 					pfi_kif_unref(newpa->kif);
2318 				PF_RULES_WUNLOCK();
2319 				free(newpa, M_PFRULE);
2320 				break;
2321 			}
2322 		}
2323 
2324 		if (pca->action == PF_CHANGE_ADD_HEAD)
2325 			oldpa = TAILQ_FIRST(&pool->list);
2326 		else if (pca->action == PF_CHANGE_ADD_TAIL)
2327 			oldpa = TAILQ_LAST(&pool->list, pf_palist);
2328 		else {
2329 			int	i = 0;
2330 
2331 			oldpa = TAILQ_FIRST(&pool->list);
2332 			while ((oldpa != NULL) && (i < pca->nr)) {
2333 				oldpa = TAILQ_NEXT(oldpa, entries);
2334 				i++;
2335 			}
2336 			if (oldpa == NULL) {
2337 				PF_RULES_WUNLOCK();
2338 				error = EINVAL;
2339 				break;
2340 			}
2341 		}
2342 
2343 		if (pca->action == PF_CHANGE_REMOVE) {
2344 			TAILQ_REMOVE(&pool->list, oldpa, entries);
2345 			switch (oldpa->addr.type) {
2346 			case PF_ADDR_DYNIFTL:
2347 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
2348 				break;
2349 			case PF_ADDR_TABLE:
2350 				pfr_detach_table(oldpa->addr.p.tbl);
2351 				break;
2352 			}
2353 			if (oldpa->kif)
2354 				pfi_kif_unref(oldpa->kif);
2355 			free(oldpa, M_PFRULE);
2356 		} else {
2357 			if (oldpa == NULL)
2358 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2359 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
2360 			    pca->action == PF_CHANGE_ADD_BEFORE)
2361 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2362 			else
2363 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
2364 				    newpa, entries);
2365 		}
2366 
2367 		pool->cur = TAILQ_FIRST(&pool->list);
2368 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr,
2369 		    pca->af);
2370 		PF_RULES_WUNLOCK();
2371 		break;
2372 
2373 #undef ERROUT
2374 DIOCCHANGEADDR_error:
2375 		PF_RULES_WUNLOCK();
2376 		if (newpa != NULL)
2377 			free(newpa, M_PFRULE);
2378 		if (kif != NULL)
2379 			free(kif, PFI_MTYPE);
2380 		break;
2381 	}
2382 
2383 	case DIOCGETRULESETS: {
2384 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2385 		struct pf_ruleset	*ruleset;
2386 		struct pf_anchor	*anchor;
2387 
2388 		PF_RULES_RLOCK();
2389 		pr->path[sizeof(pr->path) - 1] = 0;
2390 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2391 			PF_RULES_RUNLOCK();
2392 			error = ENOENT;
2393 			break;
2394 		}
2395 		pr->nr = 0;
2396 		if (ruleset->anchor == NULL) {
2397 			/* XXX kludge for pf_main_ruleset */
2398 			RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2399 				if (anchor->parent == NULL)
2400 					pr->nr++;
2401 		} else {
2402 			RB_FOREACH(anchor, pf_anchor_node,
2403 			    &ruleset->anchor->children)
2404 				pr->nr++;
2405 		}
2406 		PF_RULES_RUNLOCK();
2407 		break;
2408 	}
2409 
2410 	case DIOCGETRULESET: {
2411 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2412 		struct pf_ruleset	*ruleset;
2413 		struct pf_anchor	*anchor;
2414 		u_int32_t		 nr = 0;
2415 
2416 		PF_RULES_RLOCK();
2417 		pr->path[sizeof(pr->path) - 1] = 0;
2418 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2419 			PF_RULES_RUNLOCK();
2420 			error = ENOENT;
2421 			break;
2422 		}
2423 		pr->name[0] = 0;
2424 		if (ruleset->anchor == NULL) {
2425 			/* XXX kludge for pf_main_ruleset */
2426 			RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2427 				if (anchor->parent == NULL && nr++ == pr->nr) {
2428 					strlcpy(pr->name, anchor->name,
2429 					    sizeof(pr->name));
2430 					break;
2431 				}
2432 		} else {
2433 			RB_FOREACH(anchor, pf_anchor_node,
2434 			    &ruleset->anchor->children)
2435 				if (nr++ == pr->nr) {
2436 					strlcpy(pr->name, anchor->name,
2437 					    sizeof(pr->name));
2438 					break;
2439 				}
2440 		}
2441 		if (!pr->name[0])
2442 			error = EBUSY;
2443 		PF_RULES_RUNLOCK();
2444 		break;
2445 	}
2446 
2447 	case DIOCRCLRTABLES: {
2448 		struct pfioc_table *io = (struct pfioc_table *)addr;
2449 
2450 		if (io->pfrio_esize != 0) {
2451 			error = ENODEV;
2452 			break;
2453 		}
2454 		PF_RULES_WLOCK();
2455 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2456 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2457 		PF_RULES_WUNLOCK();
2458 		break;
2459 	}
2460 
2461 	case DIOCRADDTABLES: {
2462 		struct pfioc_table *io = (struct pfioc_table *)addr;
2463 		struct pfr_table *pfrts;
2464 		size_t totlen;
2465 
2466 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2467 			error = ENODEV;
2468 			break;
2469 		}
2470 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2471 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2472 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2473 		if (error) {
2474 			free(pfrts, M_TEMP);
2475 			break;
2476 		}
2477 		PF_RULES_WLOCK();
2478 		error = pfr_add_tables(pfrts, io->pfrio_size,
2479 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2480 		PF_RULES_WUNLOCK();
2481 		free(pfrts, M_TEMP);
2482 		break;
2483 	}
2484 
2485 	case DIOCRDELTABLES: {
2486 		struct pfioc_table *io = (struct pfioc_table *)addr;
2487 		struct pfr_table *pfrts;
2488 		size_t totlen;
2489 
2490 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2491 			error = ENODEV;
2492 			break;
2493 		}
2494 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2495 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2496 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2497 		if (error) {
2498 			free(pfrts, M_TEMP);
2499 			break;
2500 		}
2501 		PF_RULES_WLOCK();
2502 		error = pfr_del_tables(pfrts, io->pfrio_size,
2503 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2504 		PF_RULES_WUNLOCK();
2505 		free(pfrts, M_TEMP);
2506 		break;
2507 	}
2508 
2509 	case DIOCRGETTABLES: {
2510 		struct pfioc_table *io = (struct pfioc_table *)addr;
2511 		struct pfr_table *pfrts;
2512 		size_t totlen;
2513 
2514 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2515 			error = ENODEV;
2516 			break;
2517 		}
2518 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2519 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2520 		PF_RULES_RLOCK();
2521 		error = pfr_get_tables(&io->pfrio_table, pfrts,
2522 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2523 		PF_RULES_RUNLOCK();
2524 		if (error == 0)
2525 			error = copyout(pfrts, io->pfrio_buffer, totlen);
2526 		free(pfrts, M_TEMP);
2527 		break;
2528 	}
2529 
2530 	case DIOCRGETTSTATS: {
2531 		struct pfioc_table *io = (struct pfioc_table *)addr;
2532 		struct pfr_tstats *pfrtstats;
2533 		size_t totlen;
2534 
2535 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2536 			error = ENODEV;
2537 			break;
2538 		}
2539 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
2540 		pfrtstats = malloc(totlen, M_TEMP, M_WAITOK);
2541 		PF_RULES_WLOCK();
2542 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
2543 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2544 		PF_RULES_WUNLOCK();
2545 		if (error == 0)
2546 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
2547 		free(pfrtstats, M_TEMP);
2548 		break;
2549 	}
2550 
2551 	case DIOCRCLRTSTATS: {
2552 		struct pfioc_table *io = (struct pfioc_table *)addr;
2553 		struct pfr_table *pfrts;
2554 		size_t totlen;
2555 
2556 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2557 			error = ENODEV;
2558 			break;
2559 		}
2560 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2561 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2562 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2563 		if (error) {
2564 			free(pfrts, M_TEMP);
2565 			break;
2566 		}
2567 		PF_RULES_WLOCK();
2568 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
2569 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2570 		PF_RULES_WUNLOCK();
2571 		free(pfrts, M_TEMP);
2572 		break;
2573 	}
2574 
2575 	case DIOCRSETTFLAGS: {
2576 		struct pfioc_table *io = (struct pfioc_table *)addr;
2577 		struct pfr_table *pfrts;
2578 		size_t totlen;
2579 
2580 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2581 			error = ENODEV;
2582 			break;
2583 		}
2584 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2585 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2586 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2587 		if (error) {
2588 			free(pfrts, M_TEMP);
2589 			break;
2590 		}
2591 		PF_RULES_WLOCK();
2592 		error = pfr_set_tflags(pfrts, io->pfrio_size,
2593 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2594 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2595 		PF_RULES_WUNLOCK();
2596 		free(pfrts, M_TEMP);
2597 		break;
2598 	}
2599 
2600 	case DIOCRCLRADDRS: {
2601 		struct pfioc_table *io = (struct pfioc_table *)addr;
2602 
2603 		if (io->pfrio_esize != 0) {
2604 			error = ENODEV;
2605 			break;
2606 		}
2607 		PF_RULES_WLOCK();
2608 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2609 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2610 		PF_RULES_WUNLOCK();
2611 		break;
2612 	}
2613 
2614 	case DIOCRADDADDRS: {
2615 		struct pfioc_table *io = (struct pfioc_table *)addr;
2616 		struct pfr_addr *pfras;
2617 		size_t totlen;
2618 
2619 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2620 			error = ENODEV;
2621 			break;
2622 		}
2623 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2624 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2625 		error = copyin(io->pfrio_buffer, pfras, totlen);
2626 		if (error) {
2627 			free(pfras, M_TEMP);
2628 			break;
2629 		}
2630 		PF_RULES_WLOCK();
2631 		error = pfr_add_addrs(&io->pfrio_table, pfras,
2632 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2633 		    PFR_FLAG_USERIOCTL);
2634 		PF_RULES_WUNLOCK();
2635 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2636 			error = copyout(pfras, io->pfrio_buffer, totlen);
2637 		free(pfras, M_TEMP);
2638 		break;
2639 	}
2640 
2641 	case DIOCRDELADDRS: {
2642 		struct pfioc_table *io = (struct pfioc_table *)addr;
2643 		struct pfr_addr *pfras;
2644 		size_t totlen;
2645 
2646 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2647 			error = ENODEV;
2648 			break;
2649 		}
2650 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2651 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2652 		error = copyin(io->pfrio_buffer, pfras, totlen);
2653 		if (error) {
2654 			free(pfras, M_TEMP);
2655 			break;
2656 		}
2657 		PF_RULES_WLOCK();
2658 		error = pfr_del_addrs(&io->pfrio_table, pfras,
2659 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2660 		    PFR_FLAG_USERIOCTL);
2661 		PF_RULES_WUNLOCK();
2662 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2663 			error = copyout(pfras, io->pfrio_buffer, totlen);
2664 		free(pfras, M_TEMP);
2665 		break;
2666 	}
2667 
2668 	case DIOCRSETADDRS: {
2669 		struct pfioc_table *io = (struct pfioc_table *)addr;
2670 		struct pfr_addr *pfras;
2671 		size_t totlen;
2672 
2673 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2674 			error = ENODEV;
2675 			break;
2676 		}
2677 		totlen = (io->pfrio_size + io->pfrio_size2) *
2678 		    sizeof(struct pfr_addr);
2679 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2680 		error = copyin(io->pfrio_buffer, pfras, totlen);
2681 		if (error) {
2682 			free(pfras, M_TEMP);
2683 			break;
2684 		}
2685 		PF_RULES_WLOCK();
2686 		error = pfr_set_addrs(&io->pfrio_table, pfras,
2687 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2688 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2689 		    PFR_FLAG_USERIOCTL, 0);
2690 		PF_RULES_WUNLOCK();
2691 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2692 			error = copyout(pfras, io->pfrio_buffer, totlen);
2693 		free(pfras, M_TEMP);
2694 		break;
2695 	}
2696 
2697 	case DIOCRGETADDRS: {
2698 		struct pfioc_table *io = (struct pfioc_table *)addr;
2699 		struct pfr_addr *pfras;
2700 		size_t totlen;
2701 
2702 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2703 			error = ENODEV;
2704 			break;
2705 		}
2706 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2707 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2708 		PF_RULES_RLOCK();
2709 		error = pfr_get_addrs(&io->pfrio_table, pfras,
2710 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2711 		PF_RULES_RUNLOCK();
2712 		if (error == 0)
2713 			error = copyout(pfras, io->pfrio_buffer, totlen);
2714 		free(pfras, M_TEMP);
2715 		break;
2716 	}
2717 
2718 	case DIOCRGETASTATS: {
2719 		struct pfioc_table *io = (struct pfioc_table *)addr;
2720 		struct pfr_astats *pfrastats;
2721 		size_t totlen;
2722 
2723 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2724 			error = ENODEV;
2725 			break;
2726 		}
2727 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
2728 		pfrastats = malloc(totlen, M_TEMP, M_WAITOK);
2729 		PF_RULES_RLOCK();
2730 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
2731 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2732 		PF_RULES_RUNLOCK();
2733 		if (error == 0)
2734 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
2735 		free(pfrastats, M_TEMP);
2736 		break;
2737 	}
2738 
2739 	case DIOCRCLRASTATS: {
2740 		struct pfioc_table *io = (struct pfioc_table *)addr;
2741 		struct pfr_addr *pfras;
2742 		size_t totlen;
2743 
2744 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2745 			error = ENODEV;
2746 			break;
2747 		}
2748 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2749 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2750 		error = copyin(io->pfrio_buffer, pfras, totlen);
2751 		if (error) {
2752 			free(pfras, M_TEMP);
2753 			break;
2754 		}
2755 		PF_RULES_WLOCK();
2756 		error = pfr_clr_astats(&io->pfrio_table, pfras,
2757 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2758 		    PFR_FLAG_USERIOCTL);
2759 		PF_RULES_WUNLOCK();
2760 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2761 			error = copyout(pfras, io->pfrio_buffer, totlen);
2762 		free(pfras, M_TEMP);
2763 		break;
2764 	}
2765 
2766 	case DIOCRTSTADDRS: {
2767 		struct pfioc_table *io = (struct pfioc_table *)addr;
2768 		struct pfr_addr *pfras;
2769 		size_t totlen;
2770 
2771 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2772 			error = ENODEV;
2773 			break;
2774 		}
2775 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2776 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2777 		error = copyin(io->pfrio_buffer, pfras, totlen);
2778 		if (error) {
2779 			free(pfras, M_TEMP);
2780 			break;
2781 		}
2782 		PF_RULES_RLOCK();
2783 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
2784 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2785 		    PFR_FLAG_USERIOCTL);
2786 		PF_RULES_RUNLOCK();
2787 		if (error == 0)
2788 			error = copyout(pfras, io->pfrio_buffer, totlen);
2789 		free(pfras, M_TEMP);
2790 		break;
2791 	}
2792 
2793 	case DIOCRINADEFINE: {
2794 		struct pfioc_table *io = (struct pfioc_table *)addr;
2795 		struct pfr_addr *pfras;
2796 		size_t totlen;
2797 
2798 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2799 			error = ENODEV;
2800 			break;
2801 		}
2802 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2803 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2804 		error = copyin(io->pfrio_buffer, pfras, totlen);
2805 		if (error) {
2806 			free(pfras, M_TEMP);
2807 			break;
2808 		}
2809 		PF_RULES_WLOCK();
2810 		error = pfr_ina_define(&io->pfrio_table, pfras,
2811 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2812 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2813 		PF_RULES_WUNLOCK();
2814 		free(pfras, M_TEMP);
2815 		break;
2816 	}
2817 
2818 	case DIOCOSFPADD: {
2819 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2820 		PF_RULES_WLOCK();
2821 		error = pf_osfp_add(io);
2822 		PF_RULES_WUNLOCK();
2823 		break;
2824 	}
2825 
2826 	case DIOCOSFPGET: {
2827 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2828 		PF_RULES_RLOCK();
2829 		error = pf_osfp_get(io);
2830 		PF_RULES_RUNLOCK();
2831 		break;
2832 	}
2833 
2834 	case DIOCXBEGIN: {
2835 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2836 		struct pfioc_trans_e	*ioes, *ioe;
2837 		size_t			 totlen;
2838 		int			 i;
2839 
2840 		if (io->esize != sizeof(*ioe)) {
2841 			error = ENODEV;
2842 			break;
2843 		}
2844 		totlen = sizeof(struct pfioc_trans_e) * io->size;
2845 		ioes = malloc(totlen, M_TEMP, M_WAITOK);
2846 		error = copyin(io->array, ioes, totlen);
2847 		if (error) {
2848 			free(ioes, M_TEMP);
2849 			break;
2850 		}
2851 		PF_RULES_WLOCK();
2852 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2853 			switch (ioe->rs_num) {
2854 #ifdef ALTQ
2855 			case PF_RULESET_ALTQ:
2856 				if (ioe->anchor[0]) {
2857 					PF_RULES_WUNLOCK();
2858 					free(ioes, M_TEMP);
2859 					error = EINVAL;
2860 					goto fail;
2861 				}
2862 				if ((error = pf_begin_altq(&ioe->ticket))) {
2863 					PF_RULES_WUNLOCK();
2864 					free(ioes, M_TEMP);
2865 					goto fail;
2866 				}
2867 				break;
2868 #endif /* ALTQ */
2869 			case PF_RULESET_TABLE:
2870 			    {
2871 				struct pfr_table table;
2872 
2873 				bzero(&table, sizeof(table));
2874 				strlcpy(table.pfrt_anchor, ioe->anchor,
2875 				    sizeof(table.pfrt_anchor));
2876 				if ((error = pfr_ina_begin(&table,
2877 				    &ioe->ticket, NULL, 0))) {
2878 					PF_RULES_WUNLOCK();
2879 					free(ioes, M_TEMP);
2880 					goto fail;
2881 				}
2882 				break;
2883 			    }
2884 			default:
2885 				if ((error = pf_begin_rules(&ioe->ticket,
2886 				    ioe->rs_num, ioe->anchor))) {
2887 					PF_RULES_WUNLOCK();
2888 					free(ioes, M_TEMP);
2889 					goto fail;
2890 				}
2891 				break;
2892 			}
2893 		}
2894 		PF_RULES_WUNLOCK();
2895 		error = copyout(ioes, io->array, totlen);
2896 		free(ioes, M_TEMP);
2897 		break;
2898 	}
2899 
2900 	case DIOCXROLLBACK: {
2901 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2902 		struct pfioc_trans_e	*ioe, *ioes;
2903 		size_t			 totlen;
2904 		int			 i;
2905 
2906 		if (io->esize != sizeof(*ioe)) {
2907 			error = ENODEV;
2908 			break;
2909 		}
2910 		totlen = sizeof(struct pfioc_trans_e) * io->size;
2911 		ioes = malloc(totlen, M_TEMP, M_WAITOK);
2912 		error = copyin(io->array, ioes, totlen);
2913 		if (error) {
2914 			free(ioes, M_TEMP);
2915 			break;
2916 		}
2917 		PF_RULES_WLOCK();
2918 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2919 			switch (ioe->rs_num) {
2920 #ifdef ALTQ
2921 			case PF_RULESET_ALTQ:
2922 				if (ioe->anchor[0]) {
2923 					PF_RULES_WUNLOCK();
2924 					free(ioes, M_TEMP);
2925 					error = EINVAL;
2926 					goto fail;
2927 				}
2928 				if ((error = pf_rollback_altq(ioe->ticket))) {
2929 					PF_RULES_WUNLOCK();
2930 					free(ioes, M_TEMP);
2931 					goto fail; /* really bad */
2932 				}
2933 				break;
2934 #endif /* ALTQ */
2935 			case PF_RULESET_TABLE:
2936 			    {
2937 				struct pfr_table table;
2938 
2939 				bzero(&table, sizeof(table));
2940 				strlcpy(table.pfrt_anchor, ioe->anchor,
2941 				    sizeof(table.pfrt_anchor));
2942 				if ((error = pfr_ina_rollback(&table,
2943 				    ioe->ticket, NULL, 0))) {
2944 					PF_RULES_WUNLOCK();
2945 					free(ioes, M_TEMP);
2946 					goto fail; /* really bad */
2947 				}
2948 				break;
2949 			    }
2950 			default:
2951 				if ((error = pf_rollback_rules(ioe->ticket,
2952 				    ioe->rs_num, ioe->anchor))) {
2953 					PF_RULES_WUNLOCK();
2954 					free(ioes, M_TEMP);
2955 					goto fail; /* really bad */
2956 				}
2957 				break;
2958 			}
2959 		}
2960 		PF_RULES_WUNLOCK();
2961 		free(ioes, M_TEMP);
2962 		break;
2963 	}
2964 
2965 	case DIOCXCOMMIT: {
2966 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2967 		struct pfioc_trans_e	*ioe, *ioes;
2968 		struct pf_ruleset	*rs;
2969 		size_t			 totlen;
2970 		int			 i;
2971 
2972 		if (io->esize != sizeof(*ioe)) {
2973 			error = ENODEV;
2974 			break;
2975 		}
2976 		totlen = sizeof(struct pfioc_trans_e) * io->size;
2977 		ioes = malloc(totlen, M_TEMP, M_WAITOK);
2978 		error = copyin(io->array, ioes, totlen);
2979 		if (error) {
2980 			free(ioes, M_TEMP);
2981 			break;
2982 		}
2983 		PF_RULES_WLOCK();
2984 		/* First makes sure everything will succeed. */
2985 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2986 			switch (ioe->rs_num) {
2987 #ifdef ALTQ
2988 			case PF_RULESET_ALTQ:
2989 				if (ioe->anchor[0]) {
2990 					PF_RULES_WUNLOCK();
2991 					free(ioes, M_TEMP);
2992 					error = EINVAL;
2993 					goto fail;
2994 				}
2995 				if (!V_altqs_inactive_open || ioe->ticket !=
2996 				    V_ticket_altqs_inactive) {
2997 					PF_RULES_WUNLOCK();
2998 					free(ioes, M_TEMP);
2999 					error = EBUSY;
3000 					goto fail;
3001 				}
3002 				break;
3003 #endif /* ALTQ */
3004 			case PF_RULESET_TABLE:
3005 				rs = pf_find_ruleset(ioe->anchor);
3006 				if (rs == NULL || !rs->topen || ioe->ticket !=
3007 				    rs->tticket) {
3008 					PF_RULES_WUNLOCK();
3009 					free(ioes, M_TEMP);
3010 					error = EBUSY;
3011 					goto fail;
3012 				}
3013 				break;
3014 			default:
3015 				if (ioe->rs_num < 0 || ioe->rs_num >=
3016 				    PF_RULESET_MAX) {
3017 					PF_RULES_WUNLOCK();
3018 					free(ioes, M_TEMP);
3019 					error = EINVAL;
3020 					goto fail;
3021 				}
3022 				rs = pf_find_ruleset(ioe->anchor);
3023 				if (rs == NULL ||
3024 				    !rs->rules[ioe->rs_num].inactive.open ||
3025 				    rs->rules[ioe->rs_num].inactive.ticket !=
3026 				    ioe->ticket) {
3027 					PF_RULES_WUNLOCK();
3028 					free(ioes, M_TEMP);
3029 					error = EBUSY;
3030 					goto fail;
3031 				}
3032 				break;
3033 			}
3034 		}
3035 		/* Now do the commit - no errors should happen here. */
3036 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3037 			switch (ioe->rs_num) {
3038 #ifdef ALTQ
3039 			case PF_RULESET_ALTQ:
3040 				if ((error = pf_commit_altq(ioe->ticket))) {
3041 					PF_RULES_WUNLOCK();
3042 					free(ioes, M_TEMP);
3043 					goto fail; /* really bad */
3044 				}
3045 				break;
3046 #endif /* ALTQ */
3047 			case PF_RULESET_TABLE:
3048 			    {
3049 				struct pfr_table table;
3050 
3051 				bzero(&table, sizeof(table));
3052 				strlcpy(table.pfrt_anchor, ioe->anchor,
3053 				    sizeof(table.pfrt_anchor));
3054 				if ((error = pfr_ina_commit(&table,
3055 				    ioe->ticket, NULL, NULL, 0))) {
3056 					PF_RULES_WUNLOCK();
3057 					free(ioes, M_TEMP);
3058 					goto fail; /* really bad */
3059 				}
3060 				break;
3061 			    }
3062 			default:
3063 				if ((error = pf_commit_rules(ioe->ticket,
3064 				    ioe->rs_num, ioe->anchor))) {
3065 					PF_RULES_WUNLOCK();
3066 					free(ioes, M_TEMP);
3067 					goto fail; /* really bad */
3068 				}
3069 				break;
3070 			}
3071 		}
3072 		PF_RULES_WUNLOCK();
3073 		free(ioes, M_TEMP);
3074 		break;
3075 	}
3076 
3077 	case DIOCGETSRCNODES: {
3078 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
3079 		struct pf_srchash	*sh;
3080 		struct pf_src_node	*n, *p, *pstore;
3081 		uint32_t		 i, nr = 0;
3082 
3083 		if (psn->psn_len == 0) {
3084 			for (i = 0, sh = V_pf_srchash; i < V_pf_srchashmask;
3085 			    i++, sh++) {
3086 				PF_HASHROW_LOCK(sh);
3087 				LIST_FOREACH(n, &sh->nodes, entry)
3088 					nr++;
3089 				PF_HASHROW_UNLOCK(sh);
3090 			}
3091 			psn->psn_len = sizeof(struct pf_src_node) * nr;
3092 			break;
3093 		}
3094 
3095 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK);
3096 		for (i = 0, sh = V_pf_srchash; i < V_pf_srchashmask;
3097 		    i++, sh++) {
3098 		    PF_HASHROW_LOCK(sh);
3099 		    LIST_FOREACH(n, &sh->nodes, entry) {
3100 			int	secs = time_uptime, diff;
3101 
3102 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
3103 				break;
3104 
3105 			bcopy(n, p, sizeof(struct pf_src_node));
3106 			if (n->rule.ptr != NULL)
3107 				p->rule.nr = n->rule.ptr->nr;
3108 			p->creation = secs - p->creation;
3109 			if (p->expire > secs)
3110 				p->expire -= secs;
3111 			else
3112 				p->expire = 0;
3113 
3114 			/* Adjust the connection rate estimate. */
3115 			diff = secs - n->conn_rate.last;
3116 			if (diff >= n->conn_rate.seconds)
3117 				p->conn_rate.count = 0;
3118 			else
3119 				p->conn_rate.count -=
3120 				    n->conn_rate.count * diff /
3121 				    n->conn_rate.seconds;
3122 			p++;
3123 			nr++;
3124 		    }
3125 		    PF_HASHROW_UNLOCK(sh);
3126 		}
3127 		error = copyout(pstore, psn->psn_src_nodes,
3128 		    sizeof(struct pf_src_node) * nr);
3129 		if (error) {
3130 			free(pstore, M_TEMP);
3131 			break;
3132 		}
3133 		psn->psn_len = sizeof(struct pf_src_node) * nr;
3134 		free(pstore, M_TEMP);
3135 		break;
3136 	}
3137 
3138 	case DIOCCLRSRCNODES: {
3139 
3140 		pf_clear_srcnodes(NULL);
3141 		pf_purge_expired_src_nodes();
3142 		V_pf_status.src_nodes = 0;
3143 		break;
3144 	}
3145 
3146 	case DIOCKILLSRCNODES: {
3147 		struct pfioc_src_node_kill *psnk =
3148 		    (struct pfioc_src_node_kill *)addr;
3149 		struct pf_srchash	*sh;
3150 		struct pf_src_node	*sn;
3151 		u_int			i, killed = 0;
3152 
3153 		for (i = 0, sh = V_pf_srchash; i < V_pf_srchashmask;
3154 		    i++, sh++) {
3155 		    /*
3156 		     * XXXGL: we don't ever acquire sources hash lock
3157 		     * but if we ever do, the below call to pf_clear_srcnodes()
3158 		     * would lead to a LOR.
3159 		     */
3160 		    PF_HASHROW_LOCK(sh);
3161 		    LIST_FOREACH(sn, &sh->nodes, entry)
3162 			if (PF_MATCHA(psnk->psnk_src.neg,
3163 				&psnk->psnk_src.addr.v.a.addr,
3164 				&psnk->psnk_src.addr.v.a.mask,
3165 				&sn->addr, sn->af) &&
3166 			    PF_MATCHA(psnk->psnk_dst.neg,
3167 				&psnk->psnk_dst.addr.v.a.addr,
3168 				&psnk->psnk_dst.addr.v.a.mask,
3169 				&sn->raddr, sn->af)) {
3170 				/* Handle state to src_node linkage */
3171 				if (sn->states != 0)
3172 					pf_clear_srcnodes(sn);
3173 				sn->expire = 1;
3174 				killed++;
3175 			}
3176 		    PF_HASHROW_UNLOCK(sh);
3177 		}
3178 
3179 		if (killed > 0)
3180 			pf_purge_expired_src_nodes();
3181 
3182 		psnk->psnk_killed = killed;
3183 		break;
3184 	}
3185 
3186 	case DIOCSETHOSTID: {
3187 		u_int32_t	*hostid = (u_int32_t *)addr;
3188 
3189 		PF_RULES_WLOCK();
3190 		if (*hostid == 0)
3191 			V_pf_status.hostid = arc4random();
3192 		else
3193 			V_pf_status.hostid = *hostid;
3194 		PF_RULES_WUNLOCK();
3195 		break;
3196 	}
3197 
3198 	case DIOCOSFPFLUSH:
3199 		PF_RULES_WLOCK();
3200 		pf_osfp_flush();
3201 		PF_RULES_WUNLOCK();
3202 		break;
3203 
3204 	case DIOCIGETIFACES: {
3205 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3206 		struct pfi_kif *ifstore;
3207 		size_t bufsiz;
3208 
3209 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
3210 			error = ENODEV;
3211 			break;
3212 		}
3213 
3214 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
3215 		ifstore = malloc(bufsiz, M_TEMP, M_WAITOK);
3216 		PF_RULES_RLOCK();
3217 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
3218 		PF_RULES_RUNLOCK();
3219 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
3220 		free(ifstore, M_TEMP);
3221 		break;
3222 	}
3223 
3224 	case DIOCSETIFFLAG: {
3225 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3226 
3227 		PF_RULES_WLOCK();
3228 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3229 		PF_RULES_WUNLOCK();
3230 		break;
3231 	}
3232 
3233 	case DIOCCLRIFFLAG: {
3234 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3235 
3236 		PF_RULES_WLOCK();
3237 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3238 		PF_RULES_WUNLOCK();
3239 		break;
3240 	}
3241 
3242 	default:
3243 		error = ENODEV;
3244 		break;
3245 	}
3246 fail:
3247 	CURVNET_RESTORE();
3248 
3249 	return (error);
3250 }
3251 
3252 void
3253 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
3254 {
3255 	bzero(sp, sizeof(struct pfsync_state));
3256 
3257 	/* copy from state key */
3258 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
3259 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
3260 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
3261 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
3262 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
3263 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
3264 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
3265 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
3266 	sp->proto = st->key[PF_SK_WIRE]->proto;
3267 	sp->af = st->key[PF_SK_WIRE]->af;
3268 
3269 	/* copy from state */
3270 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
3271 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
3272 	sp->creation = htonl(time_uptime - st->creation);
3273 	sp->expire = pf_state_expires(st);
3274 	if (sp->expire <= time_uptime)
3275 		sp->expire = htonl(0);
3276 	else
3277 		sp->expire = htonl(sp->expire - time_uptime);
3278 
3279 	sp->direction = st->direction;
3280 	sp->log = st->log;
3281 	sp->timeout = st->timeout;
3282 	sp->state_flags = st->state_flags;
3283 	if (st->src_node)
3284 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
3285 	if (st->nat_src_node)
3286 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
3287 
3288 	sp->id = st->id;
3289 	sp->creatorid = st->creatorid;
3290 	pf_state_peer_hton(&st->src, &sp->src);
3291 	pf_state_peer_hton(&st->dst, &sp->dst);
3292 
3293 	if (st->rule.ptr == NULL)
3294 		sp->rule = htonl(-1);
3295 	else
3296 		sp->rule = htonl(st->rule.ptr->nr);
3297 	if (st->anchor.ptr == NULL)
3298 		sp->anchor = htonl(-1);
3299 	else
3300 		sp->anchor = htonl(st->anchor.ptr->nr);
3301 	if (st->nat_rule.ptr == NULL)
3302 		sp->nat_rule = htonl(-1);
3303 	else
3304 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
3305 
3306 	pf_state_counter_hton(st->packets[0], sp->packets[0]);
3307 	pf_state_counter_hton(st->packets[1], sp->packets[1]);
3308 	pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
3309 	pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
3310 
3311 }
3312 
3313 static void
3314 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
3315 {
3316 	struct pfr_ktable *kt;
3317 
3318 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
3319 
3320 	kt = aw->p.tbl;
3321 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
3322 		kt = kt->pfrkt_root;
3323 	aw->p.tbl = NULL;
3324 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
3325 		kt->pfrkt_cnt : -1;
3326 }
3327 
3328 /*
3329  * XXX - Check for version missmatch!!!
3330  */
3331 static void
3332 pf_clear_states(void)
3333 {
3334 	struct pf_state	*s;
3335 	u_int i;
3336 
3337 	for (i = 0; i <= V_pf_hashmask; i++) {
3338 		struct pf_idhash *ih = &V_pf_idhash[i];
3339 relock:
3340 		PF_HASHROW_LOCK(ih);
3341 		LIST_FOREACH(s, &ih->states, entry) {
3342 			s->timeout = PFTM_PURGE;
3343 			/* Don't send out individual delete messages. */
3344 			s->sync_state = PFSTATE_NOSYNC;
3345 			pf_unlink_state(s, PF_ENTER_LOCKED);
3346 			goto relock;
3347 		}
3348 		PF_HASHROW_UNLOCK(ih);
3349 	}
3350 }
3351 
3352 static int
3353 pf_clear_tables(void)
3354 {
3355 	struct pfioc_table io;
3356 	int error;
3357 
3358 	bzero(&io, sizeof(io));
3359 
3360 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
3361 	    io.pfrio_flags);
3362 
3363 	return (error);
3364 }
3365 
3366 static void
3367 pf_clear_srcnodes(struct pf_src_node *n)
3368 {
3369 	struct pf_state *s;
3370 	int i;
3371 
3372 	for (i = 0; i <= V_pf_hashmask; i++) {
3373 		struct pf_idhash *ih = &V_pf_idhash[i];
3374 
3375 		PF_HASHROW_LOCK(ih);
3376 		LIST_FOREACH(s, &ih->states, entry) {
3377 			if (n == NULL || n == s->src_node)
3378 				s->src_node = NULL;
3379 			if (n == NULL || n == s->nat_src_node)
3380 				s->nat_src_node = NULL;
3381 		}
3382 		PF_HASHROW_UNLOCK(ih);
3383 	}
3384 
3385 	if (n == NULL) {
3386 		struct pf_srchash *sh;
3387 
3388 		for (i = 0, sh = V_pf_srchash; i < V_pf_srchashmask;
3389 		    i++, sh++) {
3390 			PF_HASHROW_LOCK(sh);
3391 			LIST_FOREACH(n, &sh->nodes, entry) {
3392 				n->expire = 1;
3393 				n->states = 0;
3394 			}
3395 			PF_HASHROW_UNLOCK(sh);
3396 		}
3397 	} else {
3398 		/* XXX: hash slot should already be locked here. */
3399 		n->expire = 1;
3400 		n->states = 0;
3401 	}
3402 }
3403 /*
3404  * XXX - Check for version missmatch!!!
3405  */
3406 
3407 /*
3408  * Duplicate pfctl -Fa operation to get rid of as much as we can.
3409  */
3410 static int
3411 shutdown_pf(void)
3412 {
3413 	int error = 0;
3414 	u_int32_t t[5];
3415 	char nn = '\0';
3416 
3417 	V_pf_status.running = 0;
3418 	do {
3419 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
3420 		    != 0) {
3421 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
3422 			break;
3423 		}
3424 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
3425 		    != 0) {
3426 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
3427 			break;		/* XXX: rollback? */
3428 		}
3429 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
3430 		    != 0) {
3431 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
3432 			break;		/* XXX: rollback? */
3433 		}
3434 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
3435 		    != 0) {
3436 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
3437 			break;		/* XXX: rollback? */
3438 		}
3439 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
3440 		    != 0) {
3441 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
3442 			break;		/* XXX: rollback? */
3443 		}
3444 
3445 		/* XXX: these should always succeed here */
3446 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
3447 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
3448 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
3449 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
3450 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
3451 
3452 		if ((error = pf_clear_tables()) != 0)
3453 			break;
3454 
3455 #ifdef ALTQ
3456 		if ((error = pf_begin_altq(&t[0])) != 0) {
3457 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
3458 			break;
3459 		}
3460 		pf_commit_altq(t[0]);
3461 #endif
3462 
3463 		pf_clear_states();
3464 
3465 		pf_clear_srcnodes(NULL);
3466 
3467 		/* status does not use malloced mem so no need to cleanup */
3468 		/* fingerprints and interfaces have thier own cleanup code */
3469 	} while(0);
3470 
3471 	return (error);
3472 }
3473 
3474 #ifdef INET
3475 static int
3476 pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3477     struct inpcb *inp)
3478 {
3479 	int chk;
3480 
3481 	chk = pf_test(PF_IN, ifp, m, inp);
3482 	if (chk && *m) {
3483 		m_freem(*m);
3484 		*m = NULL;
3485 	}
3486 
3487 	return (chk);
3488 }
3489 
3490 static int
3491 pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3492     struct inpcb *inp)
3493 {
3494 	int chk;
3495 
3496 	/* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3497 	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3498 		in_delayed_cksum(*m);
3499 		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3500 	}
3501 
3502 	chk = pf_test(PF_OUT, ifp, m, inp);
3503 	if (chk && *m) {
3504 		m_freem(*m);
3505 		*m = NULL;
3506 	}
3507 
3508 	return (chk);
3509 }
3510 #endif
3511 
3512 #ifdef INET6
3513 static int
3514 pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3515     struct inpcb *inp)
3516 {
3517 	int chk;
3518 
3519 	/*
3520 	 * In case of loopback traffic IPv6 uses the real interface in
3521 	 * order to support scoped addresses. In order to support stateful
3522 	 * filtering we have change this to lo0 as it is the case in IPv4.
3523 	 */
3524 	CURVNET_SET(ifp->if_vnet);
3525 	chk = pf_test6(PF_IN, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
3526 	CURVNET_RESTORE();
3527 	if (chk && *m) {
3528 		m_freem(*m);
3529 		*m = NULL;
3530 	}
3531 	return chk;
3532 }
3533 
3534 static int
3535 pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3536     struct inpcb *inp)
3537 {
3538 	int chk;
3539 
3540 	/* We need a proper CSUM before we start (s. OpenBSD ip_output) */
3541 	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3542 #ifdef INET
3543 		/* XXX-BZ copy&paste error from r126261? */
3544 		in_delayed_cksum(*m);
3545 #endif
3546 		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3547 	}
3548 	CURVNET_SET(ifp->if_vnet);
3549 	chk = pf_test6(PF_OUT, ifp, m, inp);
3550 	CURVNET_RESTORE();
3551 	if (chk && *m) {
3552 		m_freem(*m);
3553 		*m = NULL;
3554 	}
3555 	return chk;
3556 }
3557 #endif /* INET6 */
3558 
3559 static int
3560 hook_pf(void)
3561 {
3562 #ifdef INET
3563 	struct pfil_head *pfh_inet;
3564 #endif
3565 #ifdef INET6
3566 	struct pfil_head *pfh_inet6;
3567 #endif
3568 
3569 	if (V_pf_pfil_hooked)
3570 		return (0);
3571 
3572 #ifdef INET
3573 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3574 	if (pfh_inet == NULL)
3575 		return (ESRCH); /* XXX */
3576 	pfil_add_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet);
3577 	pfil_add_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet);
3578 #endif
3579 #ifdef INET6
3580 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3581 	if (pfh_inet6 == NULL) {
3582 #ifdef INET
3583 		pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3584 		    pfh_inet);
3585 		pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3586 		    pfh_inet);
3587 #endif
3588 		return (ESRCH); /* XXX */
3589 	}
3590 	pfil_add_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet6);
3591 	pfil_add_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet6);
3592 #endif
3593 
3594 	V_pf_pfil_hooked = 1;
3595 	return (0);
3596 }
3597 
3598 static int
3599 dehook_pf(void)
3600 {
3601 #ifdef INET
3602 	struct pfil_head *pfh_inet;
3603 #endif
3604 #ifdef INET6
3605 	struct pfil_head *pfh_inet6;
3606 #endif
3607 
3608 	if (V_pf_pfil_hooked == 0)
3609 		return (0);
3610 
3611 #ifdef INET
3612 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3613 	if (pfh_inet == NULL)
3614 		return (ESRCH); /* XXX */
3615 	pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3616 	    pfh_inet);
3617 	pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3618 	    pfh_inet);
3619 #endif
3620 #ifdef INET6
3621 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3622 	if (pfh_inet6 == NULL)
3623 		return (ESRCH); /* XXX */
3624 	pfil_remove_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK,
3625 	    pfh_inet6);
3626 	pfil_remove_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK,
3627 	    pfh_inet6);
3628 #endif
3629 
3630 	V_pf_pfil_hooked = 0;
3631 	return (0);
3632 }
3633 
3634 static int
3635 pf_load(void)
3636 {
3637 	int error;
3638 
3639 	VNET_ITERATOR_DECL(vnet_iter);
3640 
3641 	VNET_LIST_RLOCK();
3642 	VNET_FOREACH(vnet_iter) {
3643 		CURVNET_SET(vnet_iter);
3644 		V_pf_pfil_hooked = 0;
3645 		V_pf_end_threads = 0;
3646 		TAILQ_INIT(&V_pf_tags);
3647 		TAILQ_INIT(&V_pf_qids);
3648 		CURVNET_RESTORE();
3649 	}
3650 	VNET_LIST_RUNLOCK();
3651 
3652 	rw_init(&pf_rules_lock, "pf rulesets");
3653 
3654 	pf_dev = make_dev(&pf_cdevsw, 0, 0, 0, 0600, PF_NAME);
3655 	if ((error = pfattach()) != 0)
3656 		return (error);
3657 
3658 	return (0);
3659 }
3660 
3661 static int
3662 pf_unload(void)
3663 {
3664 	int error = 0;
3665 
3666 	PF_RULES_WLOCK();
3667 	V_pf_status.running = 0;
3668 	PF_RULES_WUNLOCK();
3669 	swi_remove(V_pf_swi_cookie);
3670 	error = dehook_pf();
3671 	if (error) {
3672 		/*
3673 		 * Should not happen!
3674 		 * XXX Due to error code ESRCH, kldunload will show
3675 		 * a message like 'No such process'.
3676 		 */
3677 		printf("%s : pfil unregisteration fail\n", __FUNCTION__);
3678 		return error;
3679 	}
3680 	PF_RULES_WLOCK();
3681 	shutdown_pf();
3682 	V_pf_end_threads = 1;
3683 	while (V_pf_end_threads < 2) {
3684 		wakeup_one(pf_purge_thread);
3685 		rw_sleep(pf_purge_thread, &pf_rules_lock, 0, "pftmo", 0);
3686 	}
3687 	pf_normalize_cleanup();
3688 	pfi_cleanup();
3689 	pfr_cleanup();
3690 	pf_osfp_flush();
3691 	pf_cleanup();
3692 	PF_RULES_WUNLOCK();
3693 	destroy_dev(pf_dev);
3694 	rw_destroy(&pf_rules_lock);
3695 
3696 	return (error);
3697 }
3698 
3699 static int
3700 pf_modevent(module_t mod, int type, void *data)
3701 {
3702 	int error = 0;
3703 
3704 	switch(type) {
3705 	case MOD_LOAD:
3706 		error = pf_load();
3707 		break;
3708 	case MOD_QUIESCE:
3709 		/*
3710 		 * Module should not be unloaded due to race conditions.
3711 		 */
3712 		error = EBUSY;
3713 		break;
3714 	case MOD_UNLOAD:
3715 		error = pf_unload();
3716 		break;
3717 	default:
3718 		error = EINVAL;
3719 		break;
3720 	}
3721 
3722 	return (error);
3723 }
3724 
3725 static moduledata_t pf_mod = {
3726 	"pf",
3727 	pf_modevent,
3728 	0
3729 };
3730 
3731 DECLARE_MODULE(pf, pf_mod, SI_SUB_PSEUDO, SI_ORDER_FIRST);
3732 MODULE_VERSION(pf, PF_MODVER);
3733