xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision dda5b39711dab90ae1c5624bdd6ff7453177df31)
1 /*-
2  * Copyright (c) 2001 Daniel Hartmeier
3  * Copyright (c) 2002,2003 Henning Brauer
4  * Copyright (c) 2012 Gleb Smirnoff <glebius@FreeBSD.org>
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  *
11  *    - Redistributions of source code must retain the above copyright
12  *      notice, this list of conditions and the following disclaimer.
13  *    - Redistributions in binary form must reproduce the above
14  *      copyright notice, this list of conditions and the following
15  *      disclaimer in the documentation and/or other materials provided
16  *      with the distribution.
17  *
18  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
19  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
20  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
21  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
22  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
23  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
24  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
25  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
26  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
28  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29  * POSSIBILITY OF SUCH DAMAGE.
30  *
31  * Effort sponsored in part by the Defense Advanced Research Projects
32  * Agency (DARPA) and Air Force Research Laboratory, Air Force
33  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
34  *
35  *	$OpenBSD: pf_ioctl.c,v 1.213 2009/02/15 21:46:12 mbalmer Exp $
36  */
37 
38 #include <sys/cdefs.h>
39 __FBSDID("$FreeBSD$");
40 
41 #include "opt_inet.h"
42 #include "opt_inet6.h"
43 #include "opt_bpf.h"
44 #include "opt_pf.h"
45 
46 #include <sys/param.h>
47 #include <sys/bus.h>
48 #include <sys/conf.h>
49 #include <sys/endian.h>
50 #include <sys/fcntl.h>
51 #include <sys/filio.h>
52 #include <sys/interrupt.h>
53 #include <sys/jail.h>
54 #include <sys/kernel.h>
55 #include <sys/kthread.h>
56 #include <sys/lock.h>
57 #include <sys/mbuf.h>
58 #include <sys/module.h>
59 #include <sys/proc.h>
60 #include <sys/rwlock.h>
61 #include <sys/smp.h>
62 #include <sys/socket.h>
63 #include <sys/sysctl.h>
64 #include <sys/md5.h>
65 #include <sys/ucred.h>
66 
67 #include <net/if.h>
68 #include <net/if_var.h>
69 #include <net/vnet.h>
70 #include <net/route.h>
71 #include <net/pfil.h>
72 #include <net/pfvar.h>
73 #include <net/if_pfsync.h>
74 #include <net/if_pflog.h>
75 
76 #include <netinet/in.h>
77 #include <netinet/ip.h>
78 #include <netinet/ip_var.h>
79 #include <netinet/ip_icmp.h>
80 
81 #ifdef INET6
82 #include <netinet/ip6.h>
83 #endif /* INET6 */
84 
85 #ifdef ALTQ
86 #include <altq/altq.h>
87 #endif
88 
89 static int		 pfattach(void);
90 static struct pf_pool	*pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t,
91 			    u_int8_t, u_int8_t, u_int8_t);
92 
93 static void		 pf_mv_pool(struct pf_palist *, struct pf_palist *);
94 static void		 pf_empty_pool(struct pf_palist *);
95 static int		 pfioctl(struct cdev *, u_long, caddr_t, int,
96 			    struct thread *);
97 #ifdef ALTQ
98 static int		 pf_begin_altq(u_int32_t *);
99 static int		 pf_rollback_altq(u_int32_t);
100 static int		 pf_commit_altq(u_int32_t);
101 static int		 pf_enable_altq(struct pf_altq *);
102 static int		 pf_disable_altq(struct pf_altq *);
103 static u_int32_t	 pf_qname2qid(char *);
104 static void		 pf_qid_unref(u_int32_t);
105 #endif /* ALTQ */
106 static int		 pf_begin_rules(u_int32_t *, int, const char *);
107 static int		 pf_rollback_rules(u_int32_t, int, char *);
108 static int		 pf_setup_pfsync_matching(struct pf_ruleset *);
109 static void		 pf_hash_rule(MD5_CTX *, struct pf_rule *);
110 static void		 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
111 static int		 pf_commit_rules(u_int32_t, int, char *);
112 static int		 pf_addr_setup(struct pf_ruleset *,
113 			    struct pf_addr_wrap *, sa_family_t);
114 static void		 pf_addr_copyout(struct pf_addr_wrap *);
115 
116 VNET_DEFINE(struct pf_rule,	pf_default_rule);
117 
118 #ifdef ALTQ
119 static VNET_DEFINE(int,		pf_altq_running);
120 #define	V_pf_altq_running	VNET(pf_altq_running)
121 #endif
122 
123 #define	TAGID_MAX	 50000
124 struct pf_tagname {
125 	TAILQ_ENTRY(pf_tagname)	entries;
126 	char			name[PF_TAG_NAME_SIZE];
127 	uint16_t		tag;
128 	int			ref;
129 };
130 
131 TAILQ_HEAD(pf_tags, pf_tagname);
132 #define	V_pf_tags		VNET(pf_tags)
133 VNET_DEFINE(struct pf_tags, pf_tags);
134 #define	V_pf_qids		VNET(pf_qids)
135 VNET_DEFINE(struct pf_tags, pf_qids);
136 static MALLOC_DEFINE(M_PFTAG, "pf_tag", "pf(4) tag names");
137 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
138 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
139 
140 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
141 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
142 #endif
143 
144 static u_int16_t	 tagname2tag(struct pf_tags *, char *);
145 static u_int16_t	 pf_tagname2tag(char *);
146 static void		 tag_unref(struct pf_tags *, u_int16_t);
147 
148 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
149 
150 struct cdev *pf_dev;
151 
152 /*
153  * XXX - These are new and need to be checked when moveing to a new version
154  */
155 static void		 pf_clear_states(void);
156 static int		 pf_clear_tables(void);
157 static void		 pf_clear_srcnodes(struct pf_src_node *);
158 static void		 pf_kill_srcnodes(struct pfioc_src_node_kill *);
159 static void		 pf_tbladdr_copyout(struct pf_addr_wrap *);
160 
161 /*
162  * Wrapper functions for pfil(9) hooks
163  */
164 #ifdef INET
165 static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp,
166     int dir, struct inpcb *inp);
167 static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp,
168     int dir, struct inpcb *inp);
169 #endif
170 #ifdef INET6
171 static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp,
172     int dir, struct inpcb *inp);
173 static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp,
174     int dir, struct inpcb *inp);
175 #endif
176 
177 static int		hook_pf(void);
178 static int		dehook_pf(void);
179 static int		shutdown_pf(void);
180 static int		pf_load(void);
181 static int		pf_unload(void);
182 
183 static struct cdevsw pf_cdevsw = {
184 	.d_ioctl =	pfioctl,
185 	.d_name =	PF_NAME,
186 	.d_version =	D_VERSION,
187 };
188 
189 static volatile VNET_DEFINE(int, pf_pfil_hooked);
190 #define V_pf_pfil_hooked	VNET(pf_pfil_hooked)
191 VNET_DEFINE(int,		pf_end_threads);
192 
193 struct rwlock			pf_rules_lock;
194 
195 /* pfsync */
196 pfsync_state_import_t 		*pfsync_state_import_ptr = NULL;
197 pfsync_insert_state_t		*pfsync_insert_state_ptr = NULL;
198 pfsync_update_state_t		*pfsync_update_state_ptr = NULL;
199 pfsync_delete_state_t		*pfsync_delete_state_ptr = NULL;
200 pfsync_clear_states_t		*pfsync_clear_states_ptr = NULL;
201 pfsync_defer_t			*pfsync_defer_ptr = NULL;
202 /* pflog */
203 pflog_packet_t			*pflog_packet_ptr = NULL;
204 
205 static int
206 pfattach(void)
207 {
208 	u_int32_t *my_timeout = V_pf_default_rule.timeout;
209 	int error;
210 
211 	pf_initialize();
212 	pfr_initialize();
213 	pfi_initialize();
214 	pf_normalize_init();
215 
216 	V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
217 	V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
218 
219 	RB_INIT(&V_pf_anchors);
220 	pf_init_ruleset(&pf_main_ruleset);
221 
222 	/* default rule should never be garbage collected */
223 	V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
224 #ifdef PF_DEFAULT_TO_DROP
225 	V_pf_default_rule.action = PF_DROP;
226 #else
227 	V_pf_default_rule.action = PF_PASS;
228 #endif
229 	V_pf_default_rule.nr = -1;
230 	V_pf_default_rule.rtableid = -1;
231 
232 	V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
233 	V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
234 	V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
235 
236 	/* initialize default timeouts */
237 	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
238 	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
239 	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
240 	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
241 	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
242 	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
243 	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
244 	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
245 	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
246 	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
247 	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
248 	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
249 	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
250 	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
251 	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
252 	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
253 	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
254 	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
255 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
256 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
257 
258 	bzero(&V_pf_status, sizeof(V_pf_status));
259 	V_pf_status.debug = PF_DEBUG_URGENT;
260 
261 	V_pf_pfil_hooked = 0;
262 
263 	/* XXX do our best to avoid a conflict */
264 	V_pf_status.hostid = arc4random();
265 
266 	if ((error = kproc_create(pf_purge_thread, curvnet, NULL, 0, 0,
267 	    "pf purge")) != 0)
268 		/* XXXGL: leaked all above. */
269 		return (error);
270 	if ((error = swi_add(NULL, "pf send", pf_intr, curvnet, SWI_NET,
271 	    INTR_MPSAFE, &V_pf_swi_cookie)) != 0)
272 		/* XXXGL: leaked all above. */
273 		return (error);
274 
275 	return (0);
276 }
277 
278 static struct pf_pool *
279 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
280     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
281     u_int8_t check_ticket)
282 {
283 	struct pf_ruleset	*ruleset;
284 	struct pf_rule		*rule;
285 	int			 rs_num;
286 
287 	ruleset = pf_find_ruleset(anchor);
288 	if (ruleset == NULL)
289 		return (NULL);
290 	rs_num = pf_get_ruleset_number(rule_action);
291 	if (rs_num >= PF_RULESET_MAX)
292 		return (NULL);
293 	if (active) {
294 		if (check_ticket && ticket !=
295 		    ruleset->rules[rs_num].active.ticket)
296 			return (NULL);
297 		if (r_last)
298 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
299 			    pf_rulequeue);
300 		else
301 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
302 	} else {
303 		if (check_ticket && ticket !=
304 		    ruleset->rules[rs_num].inactive.ticket)
305 			return (NULL);
306 		if (r_last)
307 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
308 			    pf_rulequeue);
309 		else
310 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
311 	}
312 	if (!r_last) {
313 		while ((rule != NULL) && (rule->nr != rule_number))
314 			rule = TAILQ_NEXT(rule, entries);
315 	}
316 	if (rule == NULL)
317 		return (NULL);
318 
319 	return (&rule->rpool);
320 }
321 
322 static void
323 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
324 {
325 	struct pf_pooladdr	*mv_pool_pa;
326 
327 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
328 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
329 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
330 	}
331 }
332 
333 static void
334 pf_empty_pool(struct pf_palist *poola)
335 {
336 	struct pf_pooladdr *pa;
337 
338 	while ((pa = TAILQ_FIRST(poola)) != NULL) {
339 		switch (pa->addr.type) {
340 		case PF_ADDR_DYNIFTL:
341 			pfi_dynaddr_remove(pa->addr.p.dyn);
342 			break;
343 		case PF_ADDR_TABLE:
344 			pfr_detach_table(pa->addr.p.tbl);
345 			break;
346 		}
347 		if (pa->kif)
348 			pfi_kif_unref(pa->kif);
349 		TAILQ_REMOVE(poola, pa, entries);
350 		free(pa, M_PFRULE);
351 	}
352 }
353 
354 static void
355 pf_unlink_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
356 {
357 
358 	PF_RULES_WASSERT();
359 
360 	TAILQ_REMOVE(rulequeue, rule, entries);
361 
362 	PF_UNLNKDRULES_LOCK();
363 	rule->rule_flag |= PFRULE_REFS;
364 	TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
365 	PF_UNLNKDRULES_UNLOCK();
366 }
367 
368 void
369 pf_free_rule(struct pf_rule *rule)
370 {
371 
372 	PF_RULES_WASSERT();
373 
374 	if (rule->tag)
375 		tag_unref(&V_pf_tags, rule->tag);
376 	if (rule->match_tag)
377 		tag_unref(&V_pf_tags, rule->match_tag);
378 #ifdef ALTQ
379 	if (rule->pqid != rule->qid)
380 		pf_qid_unref(rule->pqid);
381 	pf_qid_unref(rule->qid);
382 #endif
383 	switch (rule->src.addr.type) {
384 	case PF_ADDR_DYNIFTL:
385 		pfi_dynaddr_remove(rule->src.addr.p.dyn);
386 		break;
387 	case PF_ADDR_TABLE:
388 		pfr_detach_table(rule->src.addr.p.tbl);
389 		break;
390 	}
391 	switch (rule->dst.addr.type) {
392 	case PF_ADDR_DYNIFTL:
393 		pfi_dynaddr_remove(rule->dst.addr.p.dyn);
394 		break;
395 	case PF_ADDR_TABLE:
396 		pfr_detach_table(rule->dst.addr.p.tbl);
397 		break;
398 	}
399 	if (rule->overload_tbl)
400 		pfr_detach_table(rule->overload_tbl);
401 	if (rule->kif)
402 		pfi_kif_unref(rule->kif);
403 	pf_anchor_remove(rule);
404 	pf_empty_pool(&rule->rpool.list);
405 	counter_u64_free(rule->states_cur);
406 	counter_u64_free(rule->states_tot);
407 	counter_u64_free(rule->src_nodes);
408 	free(rule, M_PFRULE);
409 }
410 
411 static u_int16_t
412 tagname2tag(struct pf_tags *head, char *tagname)
413 {
414 	struct pf_tagname	*tag, *p = NULL;
415 	u_int16_t		 new_tagid = 1;
416 
417 	PF_RULES_WASSERT();
418 
419 	TAILQ_FOREACH(tag, head, entries)
420 		if (strcmp(tagname, tag->name) == 0) {
421 			tag->ref++;
422 			return (tag->tag);
423 		}
424 
425 	/*
426 	 * to avoid fragmentation, we do a linear search from the beginning
427 	 * and take the first free slot we find. if there is none or the list
428 	 * is empty, append a new entry at the end.
429 	 */
430 
431 	/* new entry */
432 	if (!TAILQ_EMPTY(head))
433 		for (p = TAILQ_FIRST(head); p != NULL &&
434 		    p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
435 			new_tagid = p->tag + 1;
436 
437 	if (new_tagid > TAGID_MAX)
438 		return (0);
439 
440 	/* allocate and fill new struct pf_tagname */
441 	tag = malloc(sizeof(*tag), M_PFTAG, M_NOWAIT|M_ZERO);
442 	if (tag == NULL)
443 		return (0);
444 	strlcpy(tag->name, tagname, sizeof(tag->name));
445 	tag->tag = new_tagid;
446 	tag->ref++;
447 
448 	if (p != NULL)	/* insert new entry before p */
449 		TAILQ_INSERT_BEFORE(p, tag, entries);
450 	else	/* either list empty or no free slot in between */
451 		TAILQ_INSERT_TAIL(head, tag, entries);
452 
453 	return (tag->tag);
454 }
455 
456 static void
457 tag_unref(struct pf_tags *head, u_int16_t tag)
458 {
459 	struct pf_tagname	*p, *next;
460 
461 	PF_RULES_WASSERT();
462 
463 	for (p = TAILQ_FIRST(head); p != NULL; p = next) {
464 		next = TAILQ_NEXT(p, entries);
465 		if (tag == p->tag) {
466 			if (--p->ref == 0) {
467 				TAILQ_REMOVE(head, p, entries);
468 				free(p, M_PFTAG);
469 			}
470 			break;
471 		}
472 	}
473 }
474 
475 static u_int16_t
476 pf_tagname2tag(char *tagname)
477 {
478 	return (tagname2tag(&V_pf_tags, tagname));
479 }
480 
481 #ifdef ALTQ
482 static u_int32_t
483 pf_qname2qid(char *qname)
484 {
485 	return ((u_int32_t)tagname2tag(&V_pf_qids, qname));
486 }
487 
488 static void
489 pf_qid_unref(u_int32_t qid)
490 {
491 	tag_unref(&V_pf_qids, (u_int16_t)qid);
492 }
493 
494 static int
495 pf_begin_altq(u_int32_t *ticket)
496 {
497 	struct pf_altq	*altq;
498 	int		 error = 0;
499 
500 	PF_RULES_WASSERT();
501 
502 	/* Purge the old altq list */
503 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
504 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
505 		if (altq->qname[0] == 0 &&
506 		    (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
507 			/* detach and destroy the discipline */
508 			error = altq_remove(altq);
509 		} else
510 			pf_qid_unref(altq->qid);
511 		free(altq, M_PFALTQ);
512 	}
513 	if (error)
514 		return (error);
515 	*ticket = ++V_ticket_altqs_inactive;
516 	V_altqs_inactive_open = 1;
517 	return (0);
518 }
519 
520 static int
521 pf_rollback_altq(u_int32_t ticket)
522 {
523 	struct pf_altq	*altq;
524 	int		 error = 0;
525 
526 	PF_RULES_WASSERT();
527 
528 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
529 		return (0);
530 	/* Purge the old altq list */
531 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
532 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
533 		if (altq->qname[0] == 0 &&
534 		   (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
535 			/* detach and destroy the discipline */
536 			error = altq_remove(altq);
537 		} else
538 			pf_qid_unref(altq->qid);
539 		free(altq, M_PFALTQ);
540 	}
541 	V_altqs_inactive_open = 0;
542 	return (error);
543 }
544 
545 static int
546 pf_commit_altq(u_int32_t ticket)
547 {
548 	struct pf_altqqueue	*old_altqs;
549 	struct pf_altq		*altq;
550 	int			 err, error = 0;
551 
552 	PF_RULES_WASSERT();
553 
554 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
555 		return (EBUSY);
556 
557 	/* swap altqs, keep the old. */
558 	old_altqs = V_pf_altqs_active;
559 	V_pf_altqs_active = V_pf_altqs_inactive;
560 	V_pf_altqs_inactive = old_altqs;
561 	V_ticket_altqs_active = V_ticket_altqs_inactive;
562 
563 	/* Attach new disciplines */
564 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
565 	if (altq->qname[0] == 0 &&
566 	   (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
567 			/* attach the discipline */
568 			error = altq_pfattach(altq);
569 			if (error == 0 && V_pf_altq_running)
570 				error = pf_enable_altq(altq);
571 			if (error != 0)
572 				return (error);
573 		}
574 	}
575 
576 	/* Purge the old altq list */
577 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
578 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
579 		if (altq->qname[0] == 0 &&
580 		    (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
581 			/* detach and destroy the discipline */
582 			if (V_pf_altq_running)
583 				error = pf_disable_altq(altq);
584 			err = altq_pfdetach(altq);
585 			if (err != 0 && error == 0)
586 				error = err;
587 			err = altq_remove(altq);
588 			if (err != 0 && error == 0)
589 				error = err;
590 		} else
591 			pf_qid_unref(altq->qid);
592 		free(altq, M_PFALTQ);
593 	}
594 
595 	V_altqs_inactive_open = 0;
596 	return (error);
597 }
598 
599 static int
600 pf_enable_altq(struct pf_altq *altq)
601 {
602 	struct ifnet		*ifp;
603 	struct tb_profile	 tb;
604 	int			 error = 0;
605 
606 	if ((ifp = ifunit(altq->ifname)) == NULL)
607 		return (EINVAL);
608 
609 	if (ifp->if_snd.altq_type != ALTQT_NONE)
610 		error = altq_enable(&ifp->if_snd);
611 
612 	/* set tokenbucket regulator */
613 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
614 		tb.rate = altq->ifbandwidth;
615 		tb.depth = altq->tbrsize;
616 		error = tbr_set(&ifp->if_snd, &tb);
617 	}
618 
619 	return (error);
620 }
621 
622 static int
623 pf_disable_altq(struct pf_altq *altq)
624 {
625 	struct ifnet		*ifp;
626 	struct tb_profile	 tb;
627 	int			 error;
628 
629 	if ((ifp = ifunit(altq->ifname)) == NULL)
630 		return (EINVAL);
631 
632 	/*
633 	 * when the discipline is no longer referenced, it was overridden
634 	 * by a new one.  if so, just return.
635 	 */
636 	if (altq->altq_disc != ifp->if_snd.altq_disc)
637 		return (0);
638 
639 	error = altq_disable(&ifp->if_snd);
640 
641 	if (error == 0) {
642 		/* clear tokenbucket regulator */
643 		tb.rate = 0;
644 		error = tbr_set(&ifp->if_snd, &tb);
645 	}
646 
647 	return (error);
648 }
649 
650 void
651 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
652 {
653 	struct ifnet	*ifp1;
654 	struct pf_altq	*a1, *a2, *a3;
655 	u_int32_t	 ticket;
656 	int		 error = 0;
657 
658 	/* Interrupt userland queue modifications */
659 	if (V_altqs_inactive_open)
660 		pf_rollback_altq(V_ticket_altqs_inactive);
661 
662 	/* Start new altq ruleset */
663 	if (pf_begin_altq(&ticket))
664 		return;
665 
666 	/* Copy the current active set */
667 	TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
668 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
669 		if (a2 == NULL) {
670 			error = ENOMEM;
671 			break;
672 		}
673 		bcopy(a1, a2, sizeof(struct pf_altq));
674 
675 		if (a2->qname[0] != 0) {
676 			if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
677 				error = EBUSY;
678 				free(a2, M_PFALTQ);
679 				break;
680 			}
681 			a2->altq_disc = NULL;
682 			TAILQ_FOREACH(a3, V_pf_altqs_inactive, entries) {
683 				if (strncmp(a3->ifname, a2->ifname,
684 				    IFNAMSIZ) == 0 && a3->qname[0] == 0) {
685 					a2->altq_disc = a3->altq_disc;
686 					break;
687 				}
688 			}
689 		}
690 		/* Deactivate the interface in question */
691 		a2->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
692 		if ((ifp1 = ifunit(a2->ifname)) == NULL ||
693 		    (remove && ifp1 == ifp)) {
694 			a2->local_flags |= PFALTQ_FLAG_IF_REMOVED;
695 		} else {
696 			error = altq_add(a2);
697 
698 			if (ticket != V_ticket_altqs_inactive)
699 				error = EBUSY;
700 
701 			if (error) {
702 				free(a2, M_PFALTQ);
703 				break;
704 			}
705 		}
706 
707 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
708 	}
709 
710 	if (error != 0)
711 		pf_rollback_altq(ticket);
712 	else
713 		pf_commit_altq(ticket);
714 }
715 #endif /* ALTQ */
716 
717 static int
718 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
719 {
720 	struct pf_ruleset	*rs;
721 	struct pf_rule		*rule;
722 
723 	PF_RULES_WASSERT();
724 
725 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
726 		return (EINVAL);
727 	rs = pf_find_or_create_ruleset(anchor);
728 	if (rs == NULL)
729 		return (EINVAL);
730 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
731 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
732 		rs->rules[rs_num].inactive.rcount--;
733 	}
734 	*ticket = ++rs->rules[rs_num].inactive.ticket;
735 	rs->rules[rs_num].inactive.open = 1;
736 	return (0);
737 }
738 
739 static int
740 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
741 {
742 	struct pf_ruleset	*rs;
743 	struct pf_rule		*rule;
744 
745 	PF_RULES_WASSERT();
746 
747 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
748 		return (EINVAL);
749 	rs = pf_find_ruleset(anchor);
750 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
751 	    rs->rules[rs_num].inactive.ticket != ticket)
752 		return (0);
753 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
754 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
755 		rs->rules[rs_num].inactive.rcount--;
756 	}
757 	rs->rules[rs_num].inactive.open = 0;
758 	return (0);
759 }
760 
761 #define PF_MD5_UPD(st, elm)						\
762 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
763 
764 #define PF_MD5_UPD_STR(st, elm)						\
765 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
766 
767 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
768 		(stor) = htonl((st)->elm);				\
769 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
770 } while (0)
771 
772 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
773 		(stor) = htons((st)->elm);				\
774 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
775 } while (0)
776 
777 static void
778 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
779 {
780 	PF_MD5_UPD(pfr, addr.type);
781 	switch (pfr->addr.type) {
782 		case PF_ADDR_DYNIFTL:
783 			PF_MD5_UPD(pfr, addr.v.ifname);
784 			PF_MD5_UPD(pfr, addr.iflags);
785 			break;
786 		case PF_ADDR_TABLE:
787 			PF_MD5_UPD(pfr, addr.v.tblname);
788 			break;
789 		case PF_ADDR_ADDRMASK:
790 			/* XXX ignore af? */
791 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
792 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
793 			break;
794 	}
795 
796 	PF_MD5_UPD(pfr, port[0]);
797 	PF_MD5_UPD(pfr, port[1]);
798 	PF_MD5_UPD(pfr, neg);
799 	PF_MD5_UPD(pfr, port_op);
800 }
801 
802 static void
803 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
804 {
805 	u_int16_t x;
806 	u_int32_t y;
807 
808 	pf_hash_rule_addr(ctx, &rule->src);
809 	pf_hash_rule_addr(ctx, &rule->dst);
810 	PF_MD5_UPD_STR(rule, label);
811 	PF_MD5_UPD_STR(rule, ifname);
812 	PF_MD5_UPD_STR(rule, match_tagname);
813 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
814 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
815 	PF_MD5_UPD_HTONL(rule, prob, y);
816 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
817 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
818 	PF_MD5_UPD(rule, uid.op);
819 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
820 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
821 	PF_MD5_UPD(rule, gid.op);
822 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
823 	PF_MD5_UPD(rule, action);
824 	PF_MD5_UPD(rule, direction);
825 	PF_MD5_UPD(rule, af);
826 	PF_MD5_UPD(rule, quick);
827 	PF_MD5_UPD(rule, ifnot);
828 	PF_MD5_UPD(rule, match_tag_not);
829 	PF_MD5_UPD(rule, natpass);
830 	PF_MD5_UPD(rule, keep_state);
831 	PF_MD5_UPD(rule, proto);
832 	PF_MD5_UPD(rule, type);
833 	PF_MD5_UPD(rule, code);
834 	PF_MD5_UPD(rule, flags);
835 	PF_MD5_UPD(rule, flagset);
836 	PF_MD5_UPD(rule, allow_opts);
837 	PF_MD5_UPD(rule, rt);
838 	PF_MD5_UPD(rule, tos);
839 }
840 
841 static int
842 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
843 {
844 	struct pf_ruleset	*rs;
845 	struct pf_rule		*rule, **old_array;
846 	struct pf_rulequeue	*old_rules;
847 	int			 error;
848 	u_int32_t		 old_rcount;
849 
850 	PF_RULES_WASSERT();
851 
852 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
853 		return (EINVAL);
854 	rs = pf_find_ruleset(anchor);
855 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
856 	    ticket != rs->rules[rs_num].inactive.ticket)
857 		return (EBUSY);
858 
859 	/* Calculate checksum for the main ruleset */
860 	if (rs == &pf_main_ruleset) {
861 		error = pf_setup_pfsync_matching(rs);
862 		if (error != 0)
863 			return (error);
864 	}
865 
866 	/* Swap rules, keep the old. */
867 	old_rules = rs->rules[rs_num].active.ptr;
868 	old_rcount = rs->rules[rs_num].active.rcount;
869 	old_array = rs->rules[rs_num].active.ptr_array;
870 
871 	rs->rules[rs_num].active.ptr =
872 	    rs->rules[rs_num].inactive.ptr;
873 	rs->rules[rs_num].active.ptr_array =
874 	    rs->rules[rs_num].inactive.ptr_array;
875 	rs->rules[rs_num].active.rcount =
876 	    rs->rules[rs_num].inactive.rcount;
877 	rs->rules[rs_num].inactive.ptr = old_rules;
878 	rs->rules[rs_num].inactive.ptr_array = old_array;
879 	rs->rules[rs_num].inactive.rcount = old_rcount;
880 
881 	rs->rules[rs_num].active.ticket =
882 	    rs->rules[rs_num].inactive.ticket;
883 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
884 
885 
886 	/* Purge the old rule list. */
887 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
888 		pf_unlink_rule(old_rules, rule);
889 	if (rs->rules[rs_num].inactive.ptr_array)
890 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
891 	rs->rules[rs_num].inactive.ptr_array = NULL;
892 	rs->rules[rs_num].inactive.rcount = 0;
893 	rs->rules[rs_num].inactive.open = 0;
894 	pf_remove_if_empty_ruleset(rs);
895 
896 	return (0);
897 }
898 
899 static int
900 pf_setup_pfsync_matching(struct pf_ruleset *rs)
901 {
902 	MD5_CTX			 ctx;
903 	struct pf_rule		*rule;
904 	int			 rs_cnt;
905 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
906 
907 	MD5Init(&ctx);
908 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
909 		/* XXX PF_RULESET_SCRUB as well? */
910 		if (rs_cnt == PF_RULESET_SCRUB)
911 			continue;
912 
913 		if (rs->rules[rs_cnt].inactive.ptr_array)
914 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
915 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
916 
917 		if (rs->rules[rs_cnt].inactive.rcount) {
918 			rs->rules[rs_cnt].inactive.ptr_array =
919 			    malloc(sizeof(caddr_t) *
920 			    rs->rules[rs_cnt].inactive.rcount,
921 			    M_TEMP, M_NOWAIT);
922 
923 			if (!rs->rules[rs_cnt].inactive.ptr_array)
924 				return (ENOMEM);
925 		}
926 
927 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
928 		    entries) {
929 			pf_hash_rule(&ctx, rule);
930 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
931 		}
932 	}
933 
934 	MD5Final(digest, &ctx);
935 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
936 	return (0);
937 }
938 
939 static int
940 pf_addr_setup(struct pf_ruleset *ruleset, struct pf_addr_wrap *addr,
941     sa_family_t af)
942 {
943 	int error = 0;
944 
945 	switch (addr->type) {
946 	case PF_ADDR_TABLE:
947 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
948 		if (addr->p.tbl == NULL)
949 			error = ENOMEM;
950 		break;
951 	case PF_ADDR_DYNIFTL:
952 		error = pfi_dynaddr_setup(addr, af);
953 		break;
954 	}
955 
956 	return (error);
957 }
958 
959 static void
960 pf_addr_copyout(struct pf_addr_wrap *addr)
961 {
962 
963 	switch (addr->type) {
964 	case PF_ADDR_DYNIFTL:
965 		pfi_dynaddr_copyout(addr);
966 		break;
967 	case PF_ADDR_TABLE:
968 		pf_tbladdr_copyout(addr);
969 		break;
970 	}
971 }
972 
973 static int
974 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
975 {
976 	int			 error = 0;
977 
978 	/* XXX keep in sync with switch() below */
979 	if (securelevel_gt(td->td_ucred, 2))
980 		switch (cmd) {
981 		case DIOCGETRULES:
982 		case DIOCGETRULE:
983 		case DIOCGETADDRS:
984 		case DIOCGETADDR:
985 		case DIOCGETSTATE:
986 		case DIOCSETSTATUSIF:
987 		case DIOCGETSTATUS:
988 		case DIOCCLRSTATUS:
989 		case DIOCNATLOOK:
990 		case DIOCSETDEBUG:
991 		case DIOCGETSTATES:
992 		case DIOCGETTIMEOUT:
993 		case DIOCCLRRULECTRS:
994 		case DIOCGETLIMIT:
995 		case DIOCGETALTQS:
996 		case DIOCGETALTQ:
997 		case DIOCGETQSTATS:
998 		case DIOCGETRULESETS:
999 		case DIOCGETRULESET:
1000 		case DIOCRGETTABLES:
1001 		case DIOCRGETTSTATS:
1002 		case DIOCRCLRTSTATS:
1003 		case DIOCRCLRADDRS:
1004 		case DIOCRADDADDRS:
1005 		case DIOCRDELADDRS:
1006 		case DIOCRSETADDRS:
1007 		case DIOCRGETADDRS:
1008 		case DIOCRGETASTATS:
1009 		case DIOCRCLRASTATS:
1010 		case DIOCRTSTADDRS:
1011 		case DIOCOSFPGET:
1012 		case DIOCGETSRCNODES:
1013 		case DIOCCLRSRCNODES:
1014 		case DIOCIGETIFACES:
1015 		case DIOCGIFSPEED:
1016 		case DIOCSETIFFLAG:
1017 		case DIOCCLRIFFLAG:
1018 			break;
1019 		case DIOCRCLRTABLES:
1020 		case DIOCRADDTABLES:
1021 		case DIOCRDELTABLES:
1022 		case DIOCRSETTFLAGS:
1023 			if (((struct pfioc_table *)addr)->pfrio_flags &
1024 			    PFR_FLAG_DUMMY)
1025 				break; /* dummy operation ok */
1026 			return (EPERM);
1027 		default:
1028 			return (EPERM);
1029 		}
1030 
1031 	if (!(flags & FWRITE))
1032 		switch (cmd) {
1033 		case DIOCGETRULES:
1034 		case DIOCGETADDRS:
1035 		case DIOCGETADDR:
1036 		case DIOCGETSTATE:
1037 		case DIOCGETSTATUS:
1038 		case DIOCGETSTATES:
1039 		case DIOCGETTIMEOUT:
1040 		case DIOCGETLIMIT:
1041 		case DIOCGETALTQS:
1042 		case DIOCGETALTQ:
1043 		case DIOCGETQSTATS:
1044 		case DIOCGETRULESETS:
1045 		case DIOCGETRULESET:
1046 		case DIOCNATLOOK:
1047 		case DIOCRGETTABLES:
1048 		case DIOCRGETTSTATS:
1049 		case DIOCRGETADDRS:
1050 		case DIOCRGETASTATS:
1051 		case DIOCRTSTADDRS:
1052 		case DIOCOSFPGET:
1053 		case DIOCGETSRCNODES:
1054 		case DIOCIGETIFACES:
1055 		case DIOCGIFSPEED:
1056 			break;
1057 		case DIOCRCLRTABLES:
1058 		case DIOCRADDTABLES:
1059 		case DIOCRDELTABLES:
1060 		case DIOCRCLRTSTATS:
1061 		case DIOCRCLRADDRS:
1062 		case DIOCRADDADDRS:
1063 		case DIOCRDELADDRS:
1064 		case DIOCRSETADDRS:
1065 		case DIOCRSETTFLAGS:
1066 			if (((struct pfioc_table *)addr)->pfrio_flags &
1067 			    PFR_FLAG_DUMMY) {
1068 				flags |= FWRITE; /* need write lock for dummy */
1069 				break; /* dummy operation ok */
1070 			}
1071 			return (EACCES);
1072 		case DIOCGETRULE:
1073 			if (((struct pfioc_rule *)addr)->action ==
1074 			    PF_GET_CLR_CNTR)
1075 				return (EACCES);
1076 			break;
1077 		default:
1078 			return (EACCES);
1079 		}
1080 
1081 	CURVNET_SET(TD_TO_VNET(td));
1082 
1083 	switch (cmd) {
1084 	case DIOCSTART:
1085 		PF_RULES_WLOCK();
1086 		if (V_pf_status.running)
1087 			error = EEXIST;
1088 		else {
1089 			int cpu;
1090 
1091 			PF_RULES_WUNLOCK();
1092 			error = hook_pf();
1093 			if (error) {
1094 				DPFPRINTF(PF_DEBUG_MISC,
1095 				    ("pf: pfil registration failed\n"));
1096 				break;
1097 			}
1098 			PF_RULES_WLOCK();
1099 			V_pf_status.running = 1;
1100 			V_pf_status.since = time_second;
1101 
1102 			CPU_FOREACH(cpu)
1103 				V_pf_stateid[cpu] = time_second;
1104 
1105 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1106 		}
1107 		PF_RULES_WUNLOCK();
1108 		break;
1109 
1110 	case DIOCSTOP:
1111 		PF_RULES_WLOCK();
1112 		if (!V_pf_status.running)
1113 			error = ENOENT;
1114 		else {
1115 			V_pf_status.running = 0;
1116 			PF_RULES_WUNLOCK();
1117 			error = dehook_pf();
1118 			if (error) {
1119 				V_pf_status.running = 1;
1120 				DPFPRINTF(PF_DEBUG_MISC,
1121 				    ("pf: pfil unregistration failed\n"));
1122 			}
1123 			PF_RULES_WLOCK();
1124 			V_pf_status.since = time_second;
1125 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1126 		}
1127 		PF_RULES_WUNLOCK();
1128 		break;
1129 
1130 	case DIOCADDRULE: {
1131 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1132 		struct pf_ruleset	*ruleset;
1133 		struct pf_rule		*rule, *tail;
1134 		struct pf_pooladdr	*pa;
1135 		struct pfi_kif		*kif = NULL;
1136 		int			 rs_num;
1137 
1138 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1139 			error = EINVAL;
1140 			break;
1141 		}
1142 #ifndef INET
1143 		if (pr->rule.af == AF_INET) {
1144 			error = EAFNOSUPPORT;
1145 			break;
1146 		}
1147 #endif /* INET */
1148 #ifndef INET6
1149 		if (pr->rule.af == AF_INET6) {
1150 			error = EAFNOSUPPORT;
1151 			break;
1152 		}
1153 #endif /* INET6 */
1154 
1155 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
1156 		bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1157 		if (rule->ifname[0])
1158 			kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1159 		rule->states_cur = counter_u64_alloc(M_WAITOK);
1160 		rule->states_tot = counter_u64_alloc(M_WAITOK);
1161 		rule->src_nodes = counter_u64_alloc(M_WAITOK);
1162 		rule->cuid = td->td_ucred->cr_ruid;
1163 		rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1164 		TAILQ_INIT(&rule->rpool.list);
1165 
1166 #define	ERROUT(x)	{ error = (x); goto DIOCADDRULE_error; }
1167 
1168 		PF_RULES_WLOCK();
1169 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1170 		ruleset = pf_find_ruleset(pr->anchor);
1171 		if (ruleset == NULL)
1172 			ERROUT(EINVAL);
1173 		rs_num = pf_get_ruleset_number(pr->rule.action);
1174 		if (rs_num >= PF_RULESET_MAX)
1175 			ERROUT(EINVAL);
1176 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1177 			DPFPRINTF(PF_DEBUG_MISC,
1178 			    ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
1179 			    ruleset->rules[rs_num].inactive.ticket));
1180 			ERROUT(EBUSY);
1181 		}
1182 		if (pr->pool_ticket != V_ticket_pabuf) {
1183 			DPFPRINTF(PF_DEBUG_MISC,
1184 			    ("pool_ticket: %d != %d\n", pr->pool_ticket,
1185 			    V_ticket_pabuf));
1186 			ERROUT(EBUSY);
1187 		}
1188 
1189 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1190 		    pf_rulequeue);
1191 		if (tail)
1192 			rule->nr = tail->nr + 1;
1193 		else
1194 			rule->nr = 0;
1195 		if (rule->ifname[0]) {
1196 			rule->kif = pfi_kif_attach(kif, rule->ifname);
1197 			pfi_kif_ref(rule->kif);
1198 		} else
1199 			rule->kif = NULL;
1200 
1201 		if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
1202 			error = EBUSY;
1203 
1204 #ifdef ALTQ
1205 		/* set queue IDs */
1206 		if (rule->qname[0] != 0) {
1207 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1208 				error = EBUSY;
1209 			else if (rule->pqname[0] != 0) {
1210 				if ((rule->pqid =
1211 				    pf_qname2qid(rule->pqname)) == 0)
1212 					error = EBUSY;
1213 			} else
1214 				rule->pqid = rule->qid;
1215 		}
1216 #endif
1217 		if (rule->tagname[0])
1218 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1219 				error = EBUSY;
1220 		if (rule->match_tagname[0])
1221 			if ((rule->match_tag =
1222 			    pf_tagname2tag(rule->match_tagname)) == 0)
1223 				error = EBUSY;
1224 		if (rule->rt && !rule->direction)
1225 			error = EINVAL;
1226 		if (!rule->log)
1227 			rule->logif = 0;
1228 		if (rule->logif >= PFLOGIFS_MAX)
1229 			error = EINVAL;
1230 		if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1231 			error = ENOMEM;
1232 		if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1233 			error = ENOMEM;
1234 		if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1235 			error = EINVAL;
1236 		TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1237 			if (pa->addr.type == PF_ADDR_TABLE) {
1238 				pa->addr.p.tbl = pfr_attach_table(ruleset,
1239 				    pa->addr.v.tblname);
1240 				if (pa->addr.p.tbl == NULL)
1241 					error = ENOMEM;
1242 			}
1243 
1244 		if (rule->overload_tblname[0]) {
1245 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1246 			    rule->overload_tblname)) == NULL)
1247 				error = EINVAL;
1248 			else
1249 				rule->overload_tbl->pfrkt_flags |=
1250 				    PFR_TFLAG_ACTIVE;
1251 		}
1252 
1253 		pf_mv_pool(&V_pf_pabuf, &rule->rpool.list);
1254 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1255 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1256 		    (rule->rt > PF_FASTROUTE)) &&
1257 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1258 			error = EINVAL;
1259 
1260 		if (error) {
1261 			pf_free_rule(rule);
1262 			PF_RULES_WUNLOCK();
1263 			break;
1264 		}
1265 
1266 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1267 		rule->evaluations = rule->packets[0] = rule->packets[1] =
1268 		    rule->bytes[0] = rule->bytes[1] = 0;
1269 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1270 		    rule, entries);
1271 		ruleset->rules[rs_num].inactive.rcount++;
1272 		PF_RULES_WUNLOCK();
1273 		break;
1274 
1275 #undef ERROUT
1276 DIOCADDRULE_error:
1277 		PF_RULES_WUNLOCK();
1278 		counter_u64_free(rule->states_cur);
1279 		counter_u64_free(rule->states_tot);
1280 		counter_u64_free(rule->src_nodes);
1281 		free(rule, M_PFRULE);
1282 		if (kif)
1283 			free(kif, PFI_MTYPE);
1284 		break;
1285 	}
1286 
1287 	case DIOCGETRULES: {
1288 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1289 		struct pf_ruleset	*ruleset;
1290 		struct pf_rule		*tail;
1291 		int			 rs_num;
1292 
1293 		PF_RULES_WLOCK();
1294 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1295 		ruleset = pf_find_ruleset(pr->anchor);
1296 		if (ruleset == NULL) {
1297 			PF_RULES_WUNLOCK();
1298 			error = EINVAL;
1299 			break;
1300 		}
1301 		rs_num = pf_get_ruleset_number(pr->rule.action);
1302 		if (rs_num >= PF_RULESET_MAX) {
1303 			PF_RULES_WUNLOCK();
1304 			error = EINVAL;
1305 			break;
1306 		}
1307 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1308 		    pf_rulequeue);
1309 		if (tail)
1310 			pr->nr = tail->nr + 1;
1311 		else
1312 			pr->nr = 0;
1313 		pr->ticket = ruleset->rules[rs_num].active.ticket;
1314 		PF_RULES_WUNLOCK();
1315 		break;
1316 	}
1317 
1318 	case DIOCGETRULE: {
1319 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1320 		struct pf_ruleset	*ruleset;
1321 		struct pf_rule		*rule;
1322 		int			 rs_num, i;
1323 
1324 		PF_RULES_WLOCK();
1325 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1326 		ruleset = pf_find_ruleset(pr->anchor);
1327 		if (ruleset == NULL) {
1328 			PF_RULES_WUNLOCK();
1329 			error = EINVAL;
1330 			break;
1331 		}
1332 		rs_num = pf_get_ruleset_number(pr->rule.action);
1333 		if (rs_num >= PF_RULESET_MAX) {
1334 			PF_RULES_WUNLOCK();
1335 			error = EINVAL;
1336 			break;
1337 		}
1338 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1339 			PF_RULES_WUNLOCK();
1340 			error = EBUSY;
1341 			break;
1342 		}
1343 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1344 		while ((rule != NULL) && (rule->nr != pr->nr))
1345 			rule = TAILQ_NEXT(rule, entries);
1346 		if (rule == NULL) {
1347 			PF_RULES_WUNLOCK();
1348 			error = EBUSY;
1349 			break;
1350 		}
1351 		bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1352 		pr->rule.u_states_cur = counter_u64_fetch(rule->states_cur);
1353 		pr->rule.u_states_tot = counter_u64_fetch(rule->states_tot);
1354 		pr->rule.u_src_nodes = counter_u64_fetch(rule->src_nodes);
1355 		if (pf_anchor_copyout(ruleset, rule, pr)) {
1356 			PF_RULES_WUNLOCK();
1357 			error = EBUSY;
1358 			break;
1359 		}
1360 		pf_addr_copyout(&pr->rule.src.addr);
1361 		pf_addr_copyout(&pr->rule.dst.addr);
1362 		for (i = 0; i < PF_SKIP_COUNT; ++i)
1363 			if (rule->skip[i].ptr == NULL)
1364 				pr->rule.skip[i].nr = -1;
1365 			else
1366 				pr->rule.skip[i].nr =
1367 				    rule->skip[i].ptr->nr;
1368 
1369 		if (pr->action == PF_GET_CLR_CNTR) {
1370 			rule->evaluations = 0;
1371 			rule->packets[0] = rule->packets[1] = 0;
1372 			rule->bytes[0] = rule->bytes[1] = 0;
1373 			counter_u64_zero(rule->states_tot);
1374 		}
1375 		PF_RULES_WUNLOCK();
1376 		break;
1377 	}
1378 
1379 	case DIOCCHANGERULE: {
1380 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1381 		struct pf_ruleset	*ruleset;
1382 		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1383 		struct pfi_kif		*kif = NULL;
1384 		struct pf_pooladdr	*pa;
1385 		u_int32_t		 nr = 0;
1386 		int			 rs_num;
1387 
1388 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1389 		    pcr->action > PF_CHANGE_GET_TICKET) {
1390 			error = EINVAL;
1391 			break;
1392 		}
1393 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1394 			error = EINVAL;
1395 			break;
1396 		}
1397 
1398 		if (pcr->action != PF_CHANGE_REMOVE) {
1399 #ifndef INET
1400 			if (pcr->rule.af == AF_INET) {
1401 				error = EAFNOSUPPORT;
1402 				break;
1403 			}
1404 #endif /* INET */
1405 #ifndef INET6
1406 			if (pcr->rule.af == AF_INET6) {
1407 				error = EAFNOSUPPORT;
1408 				break;
1409 			}
1410 #endif /* INET6 */
1411 			newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
1412 			bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1413 			if (newrule->ifname[0])
1414 				kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1415 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
1416 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
1417 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
1418 			newrule->cuid = td->td_ucred->cr_ruid;
1419 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1420 			TAILQ_INIT(&newrule->rpool.list);
1421 		}
1422 
1423 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGERULE_error; }
1424 
1425 		PF_RULES_WLOCK();
1426 		if (!(pcr->action == PF_CHANGE_REMOVE ||
1427 		    pcr->action == PF_CHANGE_GET_TICKET) &&
1428 		    pcr->pool_ticket != V_ticket_pabuf)
1429 			ERROUT(EBUSY);
1430 
1431 		ruleset = pf_find_ruleset(pcr->anchor);
1432 		if (ruleset == NULL)
1433 			ERROUT(EINVAL);
1434 
1435 		rs_num = pf_get_ruleset_number(pcr->rule.action);
1436 		if (rs_num >= PF_RULESET_MAX)
1437 			ERROUT(EINVAL);
1438 
1439 		if (pcr->action == PF_CHANGE_GET_TICKET) {
1440 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1441 			ERROUT(0);
1442 		} else if (pcr->ticket !=
1443 			    ruleset->rules[rs_num].active.ticket)
1444 				ERROUT(EINVAL);
1445 
1446 		if (pcr->action != PF_CHANGE_REMOVE) {
1447 			if (newrule->ifname[0]) {
1448 				newrule->kif = pfi_kif_attach(kif,
1449 				    newrule->ifname);
1450 				pfi_kif_ref(newrule->kif);
1451 			} else
1452 				newrule->kif = NULL;
1453 
1454 			if (newrule->rtableid > 0 &&
1455 			    newrule->rtableid >= rt_numfibs)
1456 				error = EBUSY;
1457 
1458 #ifdef ALTQ
1459 			/* set queue IDs */
1460 			if (newrule->qname[0] != 0) {
1461 				if ((newrule->qid =
1462 				    pf_qname2qid(newrule->qname)) == 0)
1463 					error = EBUSY;
1464 				else if (newrule->pqname[0] != 0) {
1465 					if ((newrule->pqid =
1466 					    pf_qname2qid(newrule->pqname)) == 0)
1467 						error = EBUSY;
1468 				} else
1469 					newrule->pqid = newrule->qid;
1470 			}
1471 #endif /* ALTQ */
1472 			if (newrule->tagname[0])
1473 				if ((newrule->tag =
1474 				    pf_tagname2tag(newrule->tagname)) == 0)
1475 					error = EBUSY;
1476 			if (newrule->match_tagname[0])
1477 				if ((newrule->match_tag = pf_tagname2tag(
1478 				    newrule->match_tagname)) == 0)
1479 					error = EBUSY;
1480 			if (newrule->rt && !newrule->direction)
1481 				error = EINVAL;
1482 			if (!newrule->log)
1483 				newrule->logif = 0;
1484 			if (newrule->logif >= PFLOGIFS_MAX)
1485 				error = EINVAL;
1486 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
1487 				error = ENOMEM;
1488 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
1489 				error = ENOMEM;
1490 			if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1491 				error = EINVAL;
1492 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1493 				if (pa->addr.type == PF_ADDR_TABLE) {
1494 					pa->addr.p.tbl =
1495 					    pfr_attach_table(ruleset,
1496 					    pa->addr.v.tblname);
1497 					if (pa->addr.p.tbl == NULL)
1498 						error = ENOMEM;
1499 				}
1500 
1501 			if (newrule->overload_tblname[0]) {
1502 				if ((newrule->overload_tbl = pfr_attach_table(
1503 				    ruleset, newrule->overload_tblname)) ==
1504 				    NULL)
1505 					error = EINVAL;
1506 				else
1507 					newrule->overload_tbl->pfrkt_flags |=
1508 					    PFR_TFLAG_ACTIVE;
1509 			}
1510 
1511 			pf_mv_pool(&V_pf_pabuf, &newrule->rpool.list);
1512 			if (((((newrule->action == PF_NAT) ||
1513 			    (newrule->action == PF_RDR) ||
1514 			    (newrule->action == PF_BINAT) ||
1515 			    (newrule->rt > PF_FASTROUTE)) &&
1516 			    !newrule->anchor)) &&
1517 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1518 				error = EINVAL;
1519 
1520 			if (error) {
1521 				pf_free_rule(newrule);
1522 				PF_RULES_WUNLOCK();
1523 				break;
1524 			}
1525 
1526 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1527 			newrule->evaluations = 0;
1528 			newrule->packets[0] = newrule->packets[1] = 0;
1529 			newrule->bytes[0] = newrule->bytes[1] = 0;
1530 		}
1531 		pf_empty_pool(&V_pf_pabuf);
1532 
1533 		if (pcr->action == PF_CHANGE_ADD_HEAD)
1534 			oldrule = TAILQ_FIRST(
1535 			    ruleset->rules[rs_num].active.ptr);
1536 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1537 			oldrule = TAILQ_LAST(
1538 			    ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1539 		else {
1540 			oldrule = TAILQ_FIRST(
1541 			    ruleset->rules[rs_num].active.ptr);
1542 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1543 				oldrule = TAILQ_NEXT(oldrule, entries);
1544 			if (oldrule == NULL) {
1545 				if (newrule != NULL)
1546 					pf_free_rule(newrule);
1547 				PF_RULES_WUNLOCK();
1548 				error = EINVAL;
1549 				break;
1550 			}
1551 		}
1552 
1553 		if (pcr->action == PF_CHANGE_REMOVE) {
1554 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
1555 			    oldrule);
1556 			ruleset->rules[rs_num].active.rcount--;
1557 		} else {
1558 			if (oldrule == NULL)
1559 				TAILQ_INSERT_TAIL(
1560 				    ruleset->rules[rs_num].active.ptr,
1561 				    newrule, entries);
1562 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1563 			    pcr->action == PF_CHANGE_ADD_BEFORE)
1564 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1565 			else
1566 				TAILQ_INSERT_AFTER(
1567 				    ruleset->rules[rs_num].active.ptr,
1568 				    oldrule, newrule, entries);
1569 			ruleset->rules[rs_num].active.rcount++;
1570 		}
1571 
1572 		nr = 0;
1573 		TAILQ_FOREACH(oldrule,
1574 		    ruleset->rules[rs_num].active.ptr, entries)
1575 			oldrule->nr = nr++;
1576 
1577 		ruleset->rules[rs_num].active.ticket++;
1578 
1579 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1580 		pf_remove_if_empty_ruleset(ruleset);
1581 
1582 		PF_RULES_WUNLOCK();
1583 		break;
1584 
1585 #undef ERROUT
1586 DIOCCHANGERULE_error:
1587 		PF_RULES_WUNLOCK();
1588 		if (newrule != NULL) {
1589 			counter_u64_free(newrule->states_cur);
1590 			counter_u64_free(newrule->states_tot);
1591 			counter_u64_free(newrule->src_nodes);
1592 			free(newrule, M_PFRULE);
1593 		}
1594 		if (kif != NULL)
1595 			free(kif, PFI_MTYPE);
1596 		break;
1597 	}
1598 
1599 	case DIOCCLRSTATES: {
1600 		struct pf_state		*s;
1601 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1602 		u_int			 i, killed = 0;
1603 
1604 		for (i = 0; i <= V_pf_hashmask; i++) {
1605 			struct pf_idhash *ih = &V_pf_idhash[i];
1606 
1607 relock_DIOCCLRSTATES:
1608 			PF_HASHROW_LOCK(ih);
1609 			LIST_FOREACH(s, &ih->states, entry)
1610 				if (!psk->psk_ifname[0] ||
1611 				    !strcmp(psk->psk_ifname,
1612 				    s->kif->pfik_name)) {
1613 					/*
1614 					 * Don't send out individual
1615 					 * delete messages.
1616 					 */
1617 					s->state_flags |= PFSTATE_NOSYNC;
1618 					pf_unlink_state(s, PF_ENTER_LOCKED);
1619 					killed++;
1620 					goto relock_DIOCCLRSTATES;
1621 				}
1622 			PF_HASHROW_UNLOCK(ih);
1623 		}
1624 		psk->psk_killed = killed;
1625 		if (pfsync_clear_states_ptr != NULL)
1626 			pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
1627 		break;
1628 	}
1629 
1630 	case DIOCKILLSTATES: {
1631 		struct pf_state		*s;
1632 		struct pf_state_key	*sk;
1633 		struct pf_addr		*srcaddr, *dstaddr;
1634 		u_int16_t		 srcport, dstport;
1635 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
1636 		u_int			 i, killed = 0;
1637 
1638 		if (psk->psk_pfcmp.id) {
1639 			if (psk->psk_pfcmp.creatorid == 0)
1640 				psk->psk_pfcmp.creatorid = V_pf_status.hostid;
1641 			if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
1642 			    psk->psk_pfcmp.creatorid))) {
1643 				pf_unlink_state(s, PF_ENTER_LOCKED);
1644 				psk->psk_killed = 1;
1645 			}
1646 			break;
1647 		}
1648 
1649 		for (i = 0; i <= V_pf_hashmask; i++) {
1650 			struct pf_idhash *ih = &V_pf_idhash[i];
1651 
1652 relock_DIOCKILLSTATES:
1653 			PF_HASHROW_LOCK(ih);
1654 			LIST_FOREACH(s, &ih->states, entry) {
1655 				sk = s->key[PF_SK_WIRE];
1656 				if (s->direction == PF_OUT) {
1657 					srcaddr = &sk->addr[1];
1658 					dstaddr = &sk->addr[0];
1659 					srcport = sk->port[0];
1660 					dstport = sk->port[0];
1661 				} else {
1662 					srcaddr = &sk->addr[0];
1663 					dstaddr = &sk->addr[1];
1664 					srcport = sk->port[0];
1665 					dstport = sk->port[0];
1666 				}
1667 
1668 				if ((!psk->psk_af || sk->af == psk->psk_af)
1669 				    && (!psk->psk_proto || psk->psk_proto ==
1670 				    sk->proto) &&
1671 				    PF_MATCHA(psk->psk_src.neg,
1672 				    &psk->psk_src.addr.v.a.addr,
1673 				    &psk->psk_src.addr.v.a.mask,
1674 				    srcaddr, sk->af) &&
1675 				    PF_MATCHA(psk->psk_dst.neg,
1676 				    &psk->psk_dst.addr.v.a.addr,
1677 				    &psk->psk_dst.addr.v.a.mask,
1678 				    dstaddr, sk->af) &&
1679 				    (psk->psk_src.port_op == 0 ||
1680 				    pf_match_port(psk->psk_src.port_op,
1681 				    psk->psk_src.port[0], psk->psk_src.port[1],
1682 				    srcport)) &&
1683 				    (psk->psk_dst.port_op == 0 ||
1684 				    pf_match_port(psk->psk_dst.port_op,
1685 				    psk->psk_dst.port[0], psk->psk_dst.port[1],
1686 				    dstport)) &&
1687 				    (!psk->psk_label[0] ||
1688 				    (s->rule.ptr->label[0] &&
1689 				    !strcmp(psk->psk_label,
1690 				    s->rule.ptr->label))) &&
1691 				    (!psk->psk_ifname[0] ||
1692 				    !strcmp(psk->psk_ifname,
1693 				    s->kif->pfik_name))) {
1694 					pf_unlink_state(s, PF_ENTER_LOCKED);
1695 					killed++;
1696 					goto relock_DIOCKILLSTATES;
1697 				}
1698 			}
1699 			PF_HASHROW_UNLOCK(ih);
1700 		}
1701 		psk->psk_killed = killed;
1702 		break;
1703 	}
1704 
1705 	case DIOCADDSTATE: {
1706 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1707 		struct pfsync_state	*sp = &ps->state;
1708 
1709 		if (sp->timeout >= PFTM_MAX) {
1710 			error = EINVAL;
1711 			break;
1712 		}
1713 		if (pfsync_state_import_ptr != NULL) {
1714 			PF_RULES_RLOCK();
1715 			error = pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
1716 			PF_RULES_RUNLOCK();
1717 		} else
1718 			error = EOPNOTSUPP;
1719 		break;
1720 	}
1721 
1722 	case DIOCGETSTATE: {
1723 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1724 		struct pf_state		*s;
1725 
1726 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
1727 		if (s == NULL) {
1728 			error = ENOENT;
1729 			break;
1730 		}
1731 
1732 		pfsync_state_export(&ps->state, s);
1733 		PF_STATE_UNLOCK(s);
1734 		break;
1735 	}
1736 
1737 	case DIOCGETSTATES: {
1738 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
1739 		struct pf_state		*s;
1740 		struct pfsync_state	*pstore, *p;
1741 		int i, nr;
1742 
1743 		if (ps->ps_len == 0) {
1744 			nr = uma_zone_get_cur(V_pf_state_z);
1745 			ps->ps_len = sizeof(struct pfsync_state) * nr;
1746 			break;
1747 		}
1748 
1749 		p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK);
1750 		nr = 0;
1751 
1752 		for (i = 0; i <= V_pf_hashmask; i++) {
1753 			struct pf_idhash *ih = &V_pf_idhash[i];
1754 
1755 			PF_HASHROW_LOCK(ih);
1756 			LIST_FOREACH(s, &ih->states, entry) {
1757 
1758 				if (s->timeout == PFTM_UNLINKED)
1759 					continue;
1760 
1761 				if ((nr+1) * sizeof(*p) > ps->ps_len) {
1762 					PF_HASHROW_UNLOCK(ih);
1763 					goto DIOCGETSTATES_full;
1764 				}
1765 				pfsync_state_export(p, s);
1766 				p++;
1767 				nr++;
1768 			}
1769 			PF_HASHROW_UNLOCK(ih);
1770 		}
1771 DIOCGETSTATES_full:
1772 		error = copyout(pstore, ps->ps_states,
1773 		    sizeof(struct pfsync_state) * nr);
1774 		if (error) {
1775 			free(pstore, M_TEMP);
1776 			break;
1777 		}
1778 		ps->ps_len = sizeof(struct pfsync_state) * nr;
1779 		free(pstore, M_TEMP);
1780 
1781 		break;
1782 	}
1783 
1784 	case DIOCGETSTATUS: {
1785 		struct pf_status *s = (struct pf_status *)addr;
1786 		PF_RULES_RLOCK();
1787 		bcopy(&V_pf_status, s, sizeof(struct pf_status));
1788 		pfi_update_status(s->ifname, s);
1789 		PF_RULES_RUNLOCK();
1790 		break;
1791 	}
1792 
1793 	case DIOCSETSTATUSIF: {
1794 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
1795 
1796 		if (pi->ifname[0] == 0) {
1797 			bzero(V_pf_status.ifname, IFNAMSIZ);
1798 			break;
1799 		}
1800 		PF_RULES_WLOCK();
1801 		strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
1802 		PF_RULES_WUNLOCK();
1803 		break;
1804 	}
1805 
1806 	case DIOCCLRSTATUS: {
1807 		PF_RULES_WLOCK();
1808 		bzero(V_pf_status.counters, sizeof(V_pf_status.counters));
1809 		bzero(V_pf_status.fcounters, sizeof(V_pf_status.fcounters));
1810 		bzero(V_pf_status.scounters, sizeof(V_pf_status.scounters));
1811 		V_pf_status.since = time_second;
1812 		if (*V_pf_status.ifname)
1813 			pfi_update_status(V_pf_status.ifname, NULL);
1814 		PF_RULES_WUNLOCK();
1815 		break;
1816 	}
1817 
1818 	case DIOCNATLOOK: {
1819 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
1820 		struct pf_state_key	*sk;
1821 		struct pf_state		*state;
1822 		struct pf_state_key_cmp	 key;
1823 		int			 m = 0, direction = pnl->direction;
1824 		int			 sidx, didx;
1825 
1826 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
1827 		sidx = (direction == PF_IN) ? 1 : 0;
1828 		didx = (direction == PF_IN) ? 0 : 1;
1829 
1830 		if (!pnl->proto ||
1831 		    PF_AZERO(&pnl->saddr, pnl->af) ||
1832 		    PF_AZERO(&pnl->daddr, pnl->af) ||
1833 		    ((pnl->proto == IPPROTO_TCP ||
1834 		    pnl->proto == IPPROTO_UDP) &&
1835 		    (!pnl->dport || !pnl->sport)))
1836 			error = EINVAL;
1837 		else {
1838 			bzero(&key, sizeof(key));
1839 			key.af = pnl->af;
1840 			key.proto = pnl->proto;
1841 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
1842 			key.port[sidx] = pnl->sport;
1843 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
1844 			key.port[didx] = pnl->dport;
1845 
1846 			state = pf_find_state_all(&key, direction, &m);
1847 
1848 			if (m > 1)
1849 				error = E2BIG;	/* more than one state */
1850 			else if (state != NULL) {
1851 				/* XXXGL: not locked read */
1852 				sk = state->key[sidx];
1853 				PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
1854 				pnl->rsport = sk->port[sidx];
1855 				PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
1856 				pnl->rdport = sk->port[didx];
1857 			} else
1858 				error = ENOENT;
1859 		}
1860 		break;
1861 	}
1862 
1863 	case DIOCSETTIMEOUT: {
1864 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1865 		int		 old;
1866 
1867 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
1868 		    pt->seconds < 0) {
1869 			error = EINVAL;
1870 			break;
1871 		}
1872 		PF_RULES_WLOCK();
1873 		old = V_pf_default_rule.timeout[pt->timeout];
1874 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
1875 			pt->seconds = 1;
1876 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
1877 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
1878 			wakeup(pf_purge_thread);
1879 		pt->seconds = old;
1880 		PF_RULES_WUNLOCK();
1881 		break;
1882 	}
1883 
1884 	case DIOCGETTIMEOUT: {
1885 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1886 
1887 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
1888 			error = EINVAL;
1889 			break;
1890 		}
1891 		PF_RULES_RLOCK();
1892 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
1893 		PF_RULES_RUNLOCK();
1894 		break;
1895 	}
1896 
1897 	case DIOCGETLIMIT: {
1898 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1899 
1900 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
1901 			error = EINVAL;
1902 			break;
1903 		}
1904 		PF_RULES_RLOCK();
1905 		pl->limit = V_pf_limits[pl->index].limit;
1906 		PF_RULES_RUNLOCK();
1907 		break;
1908 	}
1909 
1910 	case DIOCSETLIMIT: {
1911 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1912 		int			 old_limit;
1913 
1914 		PF_RULES_WLOCK();
1915 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
1916 		    V_pf_limits[pl->index].zone == NULL) {
1917 			PF_RULES_WUNLOCK();
1918 			error = EINVAL;
1919 			break;
1920 		}
1921 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
1922 		old_limit = V_pf_limits[pl->index].limit;
1923 		V_pf_limits[pl->index].limit = pl->limit;
1924 		pl->limit = old_limit;
1925 		PF_RULES_WUNLOCK();
1926 		break;
1927 	}
1928 
1929 	case DIOCSETDEBUG: {
1930 		u_int32_t	*level = (u_int32_t *)addr;
1931 
1932 		PF_RULES_WLOCK();
1933 		V_pf_status.debug = *level;
1934 		PF_RULES_WUNLOCK();
1935 		break;
1936 	}
1937 
1938 	case DIOCCLRRULECTRS: {
1939 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
1940 		struct pf_ruleset	*ruleset = &pf_main_ruleset;
1941 		struct pf_rule		*rule;
1942 
1943 		PF_RULES_WLOCK();
1944 		TAILQ_FOREACH(rule,
1945 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
1946 			rule->evaluations = 0;
1947 			rule->packets[0] = rule->packets[1] = 0;
1948 			rule->bytes[0] = rule->bytes[1] = 0;
1949 		}
1950 		PF_RULES_WUNLOCK();
1951 		break;
1952 	}
1953 
1954 	case DIOCGIFSPEED: {
1955 		struct pf_ifspeed	*psp = (struct pf_ifspeed *)addr;
1956 		struct pf_ifspeed	ps;
1957 		struct ifnet		*ifp;
1958 
1959 		if (psp->ifname[0] != 0) {
1960 			/* Can we completely trust user-land? */
1961 			strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
1962 			ifp = ifunit(ps.ifname);
1963 			if (ifp != NULL)
1964 				psp->baudrate = ifp->if_baudrate;
1965 			else
1966 				error = EINVAL;
1967 		} else
1968 			error = EINVAL;
1969 		break;
1970 	}
1971 
1972 #ifdef ALTQ
1973 	case DIOCSTARTALTQ: {
1974 		struct pf_altq		*altq;
1975 
1976 		PF_RULES_WLOCK();
1977 		/* enable 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_enable_altq(altq);
1982 				if (error != 0)
1983 					break;
1984 			}
1985 		}
1986 		if (error == 0)
1987 			V_pf_altq_running = 1;
1988 		PF_RULES_WUNLOCK();
1989 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
1990 		break;
1991 	}
1992 
1993 	case DIOCSTOPALTQ: {
1994 		struct pf_altq		*altq;
1995 
1996 		PF_RULES_WLOCK();
1997 		/* disable all altq interfaces on active list */
1998 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1999 			if (altq->qname[0] == 0 && (altq->local_flags &
2000 			    PFALTQ_FLAG_IF_REMOVED) == 0) {
2001 				error = pf_disable_altq(altq);
2002 				if (error != 0)
2003 					break;
2004 			}
2005 		}
2006 		if (error == 0)
2007 			V_pf_altq_running = 0;
2008 		PF_RULES_WUNLOCK();
2009 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2010 		break;
2011 	}
2012 
2013 	case DIOCADDALTQ: {
2014 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2015 		struct pf_altq		*altq, *a;
2016 		struct ifnet		*ifp;
2017 
2018 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK);
2019 		bcopy(&pa->altq, altq, sizeof(struct pf_altq));
2020 		altq->local_flags = 0;
2021 
2022 		PF_RULES_WLOCK();
2023 		if (pa->ticket != V_ticket_altqs_inactive) {
2024 			PF_RULES_WUNLOCK();
2025 			free(altq, M_PFALTQ);
2026 			error = EBUSY;
2027 			break;
2028 		}
2029 
2030 		/*
2031 		 * if this is for a queue, find the discipline and
2032 		 * copy the necessary fields
2033 		 */
2034 		if (altq->qname[0] != 0) {
2035 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2036 				PF_RULES_WUNLOCK();
2037 				error = EBUSY;
2038 				free(altq, M_PFALTQ);
2039 				break;
2040 			}
2041 			altq->altq_disc = NULL;
2042 			TAILQ_FOREACH(a, V_pf_altqs_inactive, entries) {
2043 				if (strncmp(a->ifname, altq->ifname,
2044 				    IFNAMSIZ) == 0 && a->qname[0] == 0) {
2045 					altq->altq_disc = a->altq_disc;
2046 					break;
2047 				}
2048 			}
2049 		}
2050 
2051 		if ((ifp = ifunit(altq->ifname)) == NULL)
2052 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
2053 		else
2054 			error = altq_add(altq);
2055 
2056 		if (error) {
2057 			PF_RULES_WUNLOCK();
2058 			free(altq, M_PFALTQ);
2059 			break;
2060 		}
2061 
2062 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
2063 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2064 		PF_RULES_WUNLOCK();
2065 		break;
2066 	}
2067 
2068 	case DIOCGETALTQS: {
2069 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2070 		struct pf_altq		*altq;
2071 
2072 		PF_RULES_RLOCK();
2073 		pa->nr = 0;
2074 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
2075 			pa->nr++;
2076 		pa->ticket = V_ticket_altqs_active;
2077 		PF_RULES_RUNLOCK();
2078 		break;
2079 	}
2080 
2081 	case DIOCGETALTQ: {
2082 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2083 		struct pf_altq		*altq;
2084 		u_int32_t		 nr;
2085 
2086 		PF_RULES_RLOCK();
2087 		if (pa->ticket != V_ticket_altqs_active) {
2088 			PF_RULES_RUNLOCK();
2089 			error = EBUSY;
2090 			break;
2091 		}
2092 		nr = 0;
2093 		altq = TAILQ_FIRST(V_pf_altqs_active);
2094 		while ((altq != NULL) && (nr < pa->nr)) {
2095 			altq = TAILQ_NEXT(altq, entries);
2096 			nr++;
2097 		}
2098 		if (altq == NULL) {
2099 			PF_RULES_RUNLOCK();
2100 			error = EBUSY;
2101 			break;
2102 		}
2103 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2104 		PF_RULES_RUNLOCK();
2105 		break;
2106 	}
2107 
2108 	case DIOCCHANGEALTQ:
2109 		/* CHANGEALTQ not supported yet! */
2110 		error = ENODEV;
2111 		break;
2112 
2113 	case DIOCGETQSTATS: {
2114 		struct pfioc_qstats	*pq = (struct pfioc_qstats *)addr;
2115 		struct pf_altq		*altq;
2116 		u_int32_t		 nr;
2117 		int			 nbytes;
2118 
2119 		PF_RULES_RLOCK();
2120 		if (pq->ticket != V_ticket_altqs_active) {
2121 			PF_RULES_RUNLOCK();
2122 			error = EBUSY;
2123 			break;
2124 		}
2125 		nbytes = pq->nbytes;
2126 		nr = 0;
2127 		altq = TAILQ_FIRST(V_pf_altqs_active);
2128 		while ((altq != NULL) && (nr < pq->nr)) {
2129 			altq = TAILQ_NEXT(altq, entries);
2130 			nr++;
2131 		}
2132 		if (altq == NULL) {
2133 			PF_RULES_RUNLOCK();
2134 			error = EBUSY;
2135 			break;
2136 		}
2137 
2138 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
2139 			PF_RULES_RUNLOCK();
2140 			error = ENXIO;
2141 			break;
2142 		}
2143 		PF_RULES_RUNLOCK();
2144 		error = altq_getqstats(altq, pq->buf, &nbytes);
2145 		if (error == 0) {
2146 			pq->scheduler = altq->scheduler;
2147 			pq->nbytes = nbytes;
2148 		}
2149 		break;
2150 	}
2151 #endif /* ALTQ */
2152 
2153 	case DIOCBEGINADDRS: {
2154 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2155 
2156 		PF_RULES_WLOCK();
2157 		pf_empty_pool(&V_pf_pabuf);
2158 		pp->ticket = ++V_ticket_pabuf;
2159 		PF_RULES_WUNLOCK();
2160 		break;
2161 	}
2162 
2163 	case DIOCADDADDR: {
2164 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2165 		struct pf_pooladdr	*pa;
2166 		struct pfi_kif		*kif = NULL;
2167 
2168 #ifndef INET
2169 		if (pp->af == AF_INET) {
2170 			error = EAFNOSUPPORT;
2171 			break;
2172 		}
2173 #endif /* INET */
2174 #ifndef INET6
2175 		if (pp->af == AF_INET6) {
2176 			error = EAFNOSUPPORT;
2177 			break;
2178 		}
2179 #endif /* INET6 */
2180 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2181 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2182 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2183 			error = EINVAL;
2184 			break;
2185 		}
2186 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2187 		bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2188 		if (pa->ifname[0])
2189 			kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2190 		PF_RULES_WLOCK();
2191 		if (pp->ticket != V_ticket_pabuf) {
2192 			PF_RULES_WUNLOCK();
2193 			if (pa->ifname[0])
2194 				free(kif, PFI_MTYPE);
2195 			free(pa, M_PFRULE);
2196 			error = EBUSY;
2197 			break;
2198 		}
2199 		if (pa->ifname[0]) {
2200 			pa->kif = pfi_kif_attach(kif, pa->ifname);
2201 			pfi_kif_ref(pa->kif);
2202 		} else
2203 			pa->kif = NULL;
2204 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2205 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2206 			if (pa->ifname[0])
2207 				pfi_kif_unref(pa->kif);
2208 			PF_RULES_WUNLOCK();
2209 			free(pa, M_PFRULE);
2210 			break;
2211 		}
2212 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2213 		PF_RULES_WUNLOCK();
2214 		break;
2215 	}
2216 
2217 	case DIOCGETADDRS: {
2218 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2219 		struct pf_pool		*pool;
2220 		struct pf_pooladdr	*pa;
2221 
2222 		PF_RULES_RLOCK();
2223 		pp->nr = 0;
2224 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2225 		    pp->r_num, 0, 1, 0);
2226 		if (pool == NULL) {
2227 			PF_RULES_RUNLOCK();
2228 			error = EBUSY;
2229 			break;
2230 		}
2231 		TAILQ_FOREACH(pa, &pool->list, entries)
2232 			pp->nr++;
2233 		PF_RULES_RUNLOCK();
2234 		break;
2235 	}
2236 
2237 	case DIOCGETADDR: {
2238 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2239 		struct pf_pool		*pool;
2240 		struct pf_pooladdr	*pa;
2241 		u_int32_t		 nr = 0;
2242 
2243 		PF_RULES_RLOCK();
2244 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2245 		    pp->r_num, 0, 1, 1);
2246 		if (pool == NULL) {
2247 			PF_RULES_RUNLOCK();
2248 			error = EBUSY;
2249 			break;
2250 		}
2251 		pa = TAILQ_FIRST(&pool->list);
2252 		while ((pa != NULL) && (nr < pp->nr)) {
2253 			pa = TAILQ_NEXT(pa, entries);
2254 			nr++;
2255 		}
2256 		if (pa == NULL) {
2257 			PF_RULES_RUNLOCK();
2258 			error = EBUSY;
2259 			break;
2260 		}
2261 		bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2262 		pf_addr_copyout(&pp->addr.addr);
2263 		PF_RULES_RUNLOCK();
2264 		break;
2265 	}
2266 
2267 	case DIOCCHANGEADDR: {
2268 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
2269 		struct pf_pool		*pool;
2270 		struct pf_pooladdr	*oldpa = NULL, *newpa = NULL;
2271 		struct pf_ruleset	*ruleset;
2272 		struct pfi_kif		*kif = NULL;
2273 
2274 		if (pca->action < PF_CHANGE_ADD_HEAD ||
2275 		    pca->action > PF_CHANGE_REMOVE) {
2276 			error = EINVAL;
2277 			break;
2278 		}
2279 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2280 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2281 		    pca->addr.addr.type != PF_ADDR_TABLE) {
2282 			error = EINVAL;
2283 			break;
2284 		}
2285 
2286 		if (pca->action != PF_CHANGE_REMOVE) {
2287 #ifndef INET
2288 			if (pca->af == AF_INET) {
2289 				error = EAFNOSUPPORT;
2290 				break;
2291 			}
2292 #endif /* INET */
2293 #ifndef INET6
2294 			if (pca->af == AF_INET6) {
2295 				error = EAFNOSUPPORT;
2296 				break;
2297 			}
2298 #endif /* INET6 */
2299 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
2300 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2301 			if (newpa->ifname[0])
2302 				kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2303 			newpa->kif = NULL;
2304 		}
2305 
2306 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGEADDR_error; }
2307 		PF_RULES_WLOCK();
2308 		ruleset = pf_find_ruleset(pca->anchor);
2309 		if (ruleset == NULL)
2310 			ERROUT(EBUSY);
2311 
2312 		pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2313 		    pca->r_num, pca->r_last, 1, 1);
2314 		if (pool == NULL)
2315 			ERROUT(EBUSY);
2316 
2317 		if (pca->action != PF_CHANGE_REMOVE) {
2318 			if (newpa->ifname[0]) {
2319 				newpa->kif = pfi_kif_attach(kif, newpa->ifname);
2320 				pfi_kif_ref(newpa->kif);
2321 				kif = NULL;
2322 			}
2323 
2324 			switch (newpa->addr.type) {
2325 			case PF_ADDR_DYNIFTL:
2326 				error = pfi_dynaddr_setup(&newpa->addr,
2327 				    pca->af);
2328 				break;
2329 			case PF_ADDR_TABLE:
2330 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
2331 				    newpa->addr.v.tblname);
2332 				if (newpa->addr.p.tbl == NULL)
2333 					error = ENOMEM;
2334 				break;
2335 			}
2336 			if (error)
2337 				goto DIOCCHANGEADDR_error;
2338 		}
2339 
2340 		switch (pca->action) {
2341 		case PF_CHANGE_ADD_HEAD:
2342 			oldpa = TAILQ_FIRST(&pool->list);
2343 			break;
2344 		case PF_CHANGE_ADD_TAIL:
2345 			oldpa = TAILQ_LAST(&pool->list, pf_palist);
2346 			break;
2347 		default:
2348 			oldpa = TAILQ_FIRST(&pool->list);
2349 			for (int i = 0; oldpa && i < pca->nr; i++)
2350 				oldpa = TAILQ_NEXT(oldpa, entries);
2351 
2352 			if (oldpa == NULL)
2353 				ERROUT(EINVAL);
2354 		}
2355 
2356 		if (pca->action == PF_CHANGE_REMOVE) {
2357 			TAILQ_REMOVE(&pool->list, oldpa, entries);
2358 			switch (oldpa->addr.type) {
2359 			case PF_ADDR_DYNIFTL:
2360 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
2361 				break;
2362 			case PF_ADDR_TABLE:
2363 				pfr_detach_table(oldpa->addr.p.tbl);
2364 				break;
2365 			}
2366 			if (oldpa->kif)
2367 				pfi_kif_unref(oldpa->kif);
2368 			free(oldpa, M_PFRULE);
2369 		} else {
2370 			if (oldpa == NULL)
2371 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2372 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
2373 			    pca->action == PF_CHANGE_ADD_BEFORE)
2374 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2375 			else
2376 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
2377 				    newpa, entries);
2378 		}
2379 
2380 		pool->cur = TAILQ_FIRST(&pool->list);
2381 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
2382 		PF_RULES_WUNLOCK();
2383 		break;
2384 
2385 #undef ERROUT
2386 DIOCCHANGEADDR_error:
2387 		if (newpa->kif)
2388 			pfi_kif_unref(newpa->kif);
2389 		PF_RULES_WUNLOCK();
2390 		if (newpa != NULL)
2391 			free(newpa, M_PFRULE);
2392 		if (kif != NULL)
2393 			free(kif, PFI_MTYPE);
2394 		break;
2395 	}
2396 
2397 	case DIOCGETRULESETS: {
2398 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2399 		struct pf_ruleset	*ruleset;
2400 		struct pf_anchor	*anchor;
2401 
2402 		PF_RULES_RLOCK();
2403 		pr->path[sizeof(pr->path) - 1] = 0;
2404 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2405 			PF_RULES_RUNLOCK();
2406 			error = ENOENT;
2407 			break;
2408 		}
2409 		pr->nr = 0;
2410 		if (ruleset->anchor == NULL) {
2411 			/* XXX kludge for pf_main_ruleset */
2412 			RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2413 				if (anchor->parent == NULL)
2414 					pr->nr++;
2415 		} else {
2416 			RB_FOREACH(anchor, pf_anchor_node,
2417 			    &ruleset->anchor->children)
2418 				pr->nr++;
2419 		}
2420 		PF_RULES_RUNLOCK();
2421 		break;
2422 	}
2423 
2424 	case DIOCGETRULESET: {
2425 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2426 		struct pf_ruleset	*ruleset;
2427 		struct pf_anchor	*anchor;
2428 		u_int32_t		 nr = 0;
2429 
2430 		PF_RULES_RLOCK();
2431 		pr->path[sizeof(pr->path) - 1] = 0;
2432 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2433 			PF_RULES_RUNLOCK();
2434 			error = ENOENT;
2435 			break;
2436 		}
2437 		pr->name[0] = 0;
2438 		if (ruleset->anchor == NULL) {
2439 			/* XXX kludge for pf_main_ruleset */
2440 			RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2441 				if (anchor->parent == NULL && nr++ == pr->nr) {
2442 					strlcpy(pr->name, anchor->name,
2443 					    sizeof(pr->name));
2444 					break;
2445 				}
2446 		} else {
2447 			RB_FOREACH(anchor, pf_anchor_node,
2448 			    &ruleset->anchor->children)
2449 				if (nr++ == pr->nr) {
2450 					strlcpy(pr->name, anchor->name,
2451 					    sizeof(pr->name));
2452 					break;
2453 				}
2454 		}
2455 		if (!pr->name[0])
2456 			error = EBUSY;
2457 		PF_RULES_RUNLOCK();
2458 		break;
2459 	}
2460 
2461 	case DIOCRCLRTABLES: {
2462 		struct pfioc_table *io = (struct pfioc_table *)addr;
2463 
2464 		if (io->pfrio_esize != 0) {
2465 			error = ENODEV;
2466 			break;
2467 		}
2468 		PF_RULES_WLOCK();
2469 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2470 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2471 		PF_RULES_WUNLOCK();
2472 		break;
2473 	}
2474 
2475 	case DIOCRADDTABLES: {
2476 		struct pfioc_table *io = (struct pfioc_table *)addr;
2477 		struct pfr_table *pfrts;
2478 		size_t totlen;
2479 
2480 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2481 			error = ENODEV;
2482 			break;
2483 		}
2484 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2485 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2486 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2487 		if (error) {
2488 			free(pfrts, M_TEMP);
2489 			break;
2490 		}
2491 		PF_RULES_WLOCK();
2492 		error = pfr_add_tables(pfrts, io->pfrio_size,
2493 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2494 		PF_RULES_WUNLOCK();
2495 		free(pfrts, M_TEMP);
2496 		break;
2497 	}
2498 
2499 	case DIOCRDELTABLES: {
2500 		struct pfioc_table *io = (struct pfioc_table *)addr;
2501 		struct pfr_table *pfrts;
2502 		size_t totlen;
2503 
2504 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2505 			error = ENODEV;
2506 			break;
2507 		}
2508 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2509 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2510 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2511 		if (error) {
2512 			free(pfrts, M_TEMP);
2513 			break;
2514 		}
2515 		PF_RULES_WLOCK();
2516 		error = pfr_del_tables(pfrts, io->pfrio_size,
2517 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2518 		PF_RULES_WUNLOCK();
2519 		free(pfrts, M_TEMP);
2520 		break;
2521 	}
2522 
2523 	case DIOCRGETTABLES: {
2524 		struct pfioc_table *io = (struct pfioc_table *)addr;
2525 		struct pfr_table *pfrts;
2526 		size_t totlen;
2527 
2528 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2529 			error = ENODEV;
2530 			break;
2531 		}
2532 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2533 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2534 		PF_RULES_RLOCK();
2535 		error = pfr_get_tables(&io->pfrio_table, pfrts,
2536 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2537 		PF_RULES_RUNLOCK();
2538 		if (error == 0)
2539 			error = copyout(pfrts, io->pfrio_buffer, totlen);
2540 		free(pfrts, M_TEMP);
2541 		break;
2542 	}
2543 
2544 	case DIOCRGETTSTATS: {
2545 		struct pfioc_table *io = (struct pfioc_table *)addr;
2546 		struct pfr_tstats *pfrtstats;
2547 		size_t totlen;
2548 
2549 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2550 			error = ENODEV;
2551 			break;
2552 		}
2553 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
2554 		pfrtstats = malloc(totlen, M_TEMP, M_WAITOK);
2555 		PF_RULES_WLOCK();
2556 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
2557 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2558 		PF_RULES_WUNLOCK();
2559 		if (error == 0)
2560 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
2561 		free(pfrtstats, M_TEMP);
2562 		break;
2563 	}
2564 
2565 	case DIOCRCLRTSTATS: {
2566 		struct pfioc_table *io = (struct pfioc_table *)addr;
2567 		struct pfr_table *pfrts;
2568 		size_t totlen;
2569 
2570 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2571 			error = ENODEV;
2572 			break;
2573 		}
2574 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2575 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2576 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2577 		if (error) {
2578 			free(pfrts, M_TEMP);
2579 			break;
2580 		}
2581 		PF_RULES_WLOCK();
2582 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
2583 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2584 		PF_RULES_WUNLOCK();
2585 		free(pfrts, M_TEMP);
2586 		break;
2587 	}
2588 
2589 	case DIOCRSETTFLAGS: {
2590 		struct pfioc_table *io = (struct pfioc_table *)addr;
2591 		struct pfr_table *pfrts;
2592 		size_t totlen;
2593 
2594 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2595 			error = ENODEV;
2596 			break;
2597 		}
2598 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2599 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2600 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2601 		if (error) {
2602 			free(pfrts, M_TEMP);
2603 			break;
2604 		}
2605 		PF_RULES_WLOCK();
2606 		error = pfr_set_tflags(pfrts, io->pfrio_size,
2607 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2608 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2609 		PF_RULES_WUNLOCK();
2610 		free(pfrts, M_TEMP);
2611 		break;
2612 	}
2613 
2614 	case DIOCRCLRADDRS: {
2615 		struct pfioc_table *io = (struct pfioc_table *)addr;
2616 
2617 		if (io->pfrio_esize != 0) {
2618 			error = ENODEV;
2619 			break;
2620 		}
2621 		PF_RULES_WLOCK();
2622 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2623 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2624 		PF_RULES_WUNLOCK();
2625 		break;
2626 	}
2627 
2628 	case DIOCRADDADDRS: {
2629 		struct pfioc_table *io = (struct pfioc_table *)addr;
2630 		struct pfr_addr *pfras;
2631 		size_t totlen;
2632 
2633 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2634 			error = ENODEV;
2635 			break;
2636 		}
2637 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2638 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2639 		error = copyin(io->pfrio_buffer, pfras, totlen);
2640 		if (error) {
2641 			free(pfras, M_TEMP);
2642 			break;
2643 		}
2644 		PF_RULES_WLOCK();
2645 		error = pfr_add_addrs(&io->pfrio_table, pfras,
2646 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2647 		    PFR_FLAG_USERIOCTL);
2648 		PF_RULES_WUNLOCK();
2649 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2650 			error = copyout(pfras, io->pfrio_buffer, totlen);
2651 		free(pfras, M_TEMP);
2652 		break;
2653 	}
2654 
2655 	case DIOCRDELADDRS: {
2656 		struct pfioc_table *io = (struct pfioc_table *)addr;
2657 		struct pfr_addr *pfras;
2658 		size_t totlen;
2659 
2660 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2661 			error = ENODEV;
2662 			break;
2663 		}
2664 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2665 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2666 		error = copyin(io->pfrio_buffer, pfras, totlen);
2667 		if (error) {
2668 			free(pfras, M_TEMP);
2669 			break;
2670 		}
2671 		PF_RULES_WLOCK();
2672 		error = pfr_del_addrs(&io->pfrio_table, pfras,
2673 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2674 		    PFR_FLAG_USERIOCTL);
2675 		PF_RULES_WUNLOCK();
2676 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2677 			error = copyout(pfras, io->pfrio_buffer, totlen);
2678 		free(pfras, M_TEMP);
2679 		break;
2680 	}
2681 
2682 	case DIOCRSETADDRS: {
2683 		struct pfioc_table *io = (struct pfioc_table *)addr;
2684 		struct pfr_addr *pfras;
2685 		size_t totlen;
2686 
2687 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2688 			error = ENODEV;
2689 			break;
2690 		}
2691 		totlen = (io->pfrio_size + io->pfrio_size2) *
2692 		    sizeof(struct pfr_addr);
2693 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2694 		error = copyin(io->pfrio_buffer, pfras, totlen);
2695 		if (error) {
2696 			free(pfras, M_TEMP);
2697 			break;
2698 		}
2699 		PF_RULES_WLOCK();
2700 		error = pfr_set_addrs(&io->pfrio_table, pfras,
2701 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2702 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2703 		    PFR_FLAG_USERIOCTL, 0);
2704 		PF_RULES_WUNLOCK();
2705 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2706 			error = copyout(pfras, io->pfrio_buffer, totlen);
2707 		free(pfras, M_TEMP);
2708 		break;
2709 	}
2710 
2711 	case DIOCRGETADDRS: {
2712 		struct pfioc_table *io = (struct pfioc_table *)addr;
2713 		struct pfr_addr *pfras;
2714 		size_t totlen;
2715 
2716 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2717 			error = ENODEV;
2718 			break;
2719 		}
2720 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2721 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2722 		PF_RULES_RLOCK();
2723 		error = pfr_get_addrs(&io->pfrio_table, pfras,
2724 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2725 		PF_RULES_RUNLOCK();
2726 		if (error == 0)
2727 			error = copyout(pfras, io->pfrio_buffer, totlen);
2728 		free(pfras, M_TEMP);
2729 		break;
2730 	}
2731 
2732 	case DIOCRGETASTATS: {
2733 		struct pfioc_table *io = (struct pfioc_table *)addr;
2734 		struct pfr_astats *pfrastats;
2735 		size_t totlen;
2736 
2737 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2738 			error = ENODEV;
2739 			break;
2740 		}
2741 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
2742 		pfrastats = malloc(totlen, M_TEMP, M_WAITOK);
2743 		PF_RULES_RLOCK();
2744 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
2745 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2746 		PF_RULES_RUNLOCK();
2747 		if (error == 0)
2748 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
2749 		free(pfrastats, M_TEMP);
2750 		break;
2751 	}
2752 
2753 	case DIOCRCLRASTATS: {
2754 		struct pfioc_table *io = (struct pfioc_table *)addr;
2755 		struct pfr_addr *pfras;
2756 		size_t totlen;
2757 
2758 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2759 			error = ENODEV;
2760 			break;
2761 		}
2762 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2763 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2764 		error = copyin(io->pfrio_buffer, pfras, totlen);
2765 		if (error) {
2766 			free(pfras, M_TEMP);
2767 			break;
2768 		}
2769 		PF_RULES_WLOCK();
2770 		error = pfr_clr_astats(&io->pfrio_table, pfras,
2771 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2772 		    PFR_FLAG_USERIOCTL);
2773 		PF_RULES_WUNLOCK();
2774 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2775 			error = copyout(pfras, io->pfrio_buffer, totlen);
2776 		free(pfras, M_TEMP);
2777 		break;
2778 	}
2779 
2780 	case DIOCRTSTADDRS: {
2781 		struct pfioc_table *io = (struct pfioc_table *)addr;
2782 		struct pfr_addr *pfras;
2783 		size_t totlen;
2784 
2785 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2786 			error = ENODEV;
2787 			break;
2788 		}
2789 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2790 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2791 		error = copyin(io->pfrio_buffer, pfras, totlen);
2792 		if (error) {
2793 			free(pfras, M_TEMP);
2794 			break;
2795 		}
2796 		PF_RULES_RLOCK();
2797 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
2798 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2799 		    PFR_FLAG_USERIOCTL);
2800 		PF_RULES_RUNLOCK();
2801 		if (error == 0)
2802 			error = copyout(pfras, io->pfrio_buffer, totlen);
2803 		free(pfras, M_TEMP);
2804 		break;
2805 	}
2806 
2807 	case DIOCRINADEFINE: {
2808 		struct pfioc_table *io = (struct pfioc_table *)addr;
2809 		struct pfr_addr *pfras;
2810 		size_t totlen;
2811 
2812 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2813 			error = ENODEV;
2814 			break;
2815 		}
2816 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2817 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2818 		error = copyin(io->pfrio_buffer, pfras, totlen);
2819 		if (error) {
2820 			free(pfras, M_TEMP);
2821 			break;
2822 		}
2823 		PF_RULES_WLOCK();
2824 		error = pfr_ina_define(&io->pfrio_table, pfras,
2825 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2826 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2827 		PF_RULES_WUNLOCK();
2828 		free(pfras, M_TEMP);
2829 		break;
2830 	}
2831 
2832 	case DIOCOSFPADD: {
2833 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2834 		PF_RULES_WLOCK();
2835 		error = pf_osfp_add(io);
2836 		PF_RULES_WUNLOCK();
2837 		break;
2838 	}
2839 
2840 	case DIOCOSFPGET: {
2841 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2842 		PF_RULES_RLOCK();
2843 		error = pf_osfp_get(io);
2844 		PF_RULES_RUNLOCK();
2845 		break;
2846 	}
2847 
2848 	case DIOCXBEGIN: {
2849 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2850 		struct pfioc_trans_e	*ioes, *ioe;
2851 		size_t			 totlen;
2852 		int			 i;
2853 
2854 		if (io->esize != sizeof(*ioe)) {
2855 			error = ENODEV;
2856 			break;
2857 		}
2858 		totlen = sizeof(struct pfioc_trans_e) * io->size;
2859 		ioes = malloc(totlen, M_TEMP, M_WAITOK);
2860 		error = copyin(io->array, ioes, totlen);
2861 		if (error) {
2862 			free(ioes, M_TEMP);
2863 			break;
2864 		}
2865 		PF_RULES_WLOCK();
2866 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2867 			switch (ioe->rs_num) {
2868 #ifdef ALTQ
2869 			case PF_RULESET_ALTQ:
2870 				if (ioe->anchor[0]) {
2871 					PF_RULES_WUNLOCK();
2872 					free(ioes, M_TEMP);
2873 					error = EINVAL;
2874 					goto fail;
2875 				}
2876 				if ((error = pf_begin_altq(&ioe->ticket))) {
2877 					PF_RULES_WUNLOCK();
2878 					free(ioes, M_TEMP);
2879 					goto fail;
2880 				}
2881 				break;
2882 #endif /* ALTQ */
2883 			case PF_RULESET_TABLE:
2884 			    {
2885 				struct pfr_table table;
2886 
2887 				bzero(&table, sizeof(table));
2888 				strlcpy(table.pfrt_anchor, ioe->anchor,
2889 				    sizeof(table.pfrt_anchor));
2890 				if ((error = pfr_ina_begin(&table,
2891 				    &ioe->ticket, NULL, 0))) {
2892 					PF_RULES_WUNLOCK();
2893 					free(ioes, M_TEMP);
2894 					goto fail;
2895 				}
2896 				break;
2897 			    }
2898 			default:
2899 				if ((error = pf_begin_rules(&ioe->ticket,
2900 				    ioe->rs_num, ioe->anchor))) {
2901 					PF_RULES_WUNLOCK();
2902 					free(ioes, M_TEMP);
2903 					goto fail;
2904 				}
2905 				break;
2906 			}
2907 		}
2908 		PF_RULES_WUNLOCK();
2909 		error = copyout(ioes, io->array, totlen);
2910 		free(ioes, M_TEMP);
2911 		break;
2912 	}
2913 
2914 	case DIOCXROLLBACK: {
2915 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2916 		struct pfioc_trans_e	*ioe, *ioes;
2917 		size_t			 totlen;
2918 		int			 i;
2919 
2920 		if (io->esize != sizeof(*ioe)) {
2921 			error = ENODEV;
2922 			break;
2923 		}
2924 		totlen = sizeof(struct pfioc_trans_e) * io->size;
2925 		ioes = malloc(totlen, M_TEMP, M_WAITOK);
2926 		error = copyin(io->array, ioes, totlen);
2927 		if (error) {
2928 			free(ioes, M_TEMP);
2929 			break;
2930 		}
2931 		PF_RULES_WLOCK();
2932 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2933 			switch (ioe->rs_num) {
2934 #ifdef ALTQ
2935 			case PF_RULESET_ALTQ:
2936 				if (ioe->anchor[0]) {
2937 					PF_RULES_WUNLOCK();
2938 					free(ioes, M_TEMP);
2939 					error = EINVAL;
2940 					goto fail;
2941 				}
2942 				if ((error = pf_rollback_altq(ioe->ticket))) {
2943 					PF_RULES_WUNLOCK();
2944 					free(ioes, M_TEMP);
2945 					goto fail; /* really bad */
2946 				}
2947 				break;
2948 #endif /* ALTQ */
2949 			case PF_RULESET_TABLE:
2950 			    {
2951 				struct pfr_table table;
2952 
2953 				bzero(&table, sizeof(table));
2954 				strlcpy(table.pfrt_anchor, ioe->anchor,
2955 				    sizeof(table.pfrt_anchor));
2956 				if ((error = pfr_ina_rollback(&table,
2957 				    ioe->ticket, NULL, 0))) {
2958 					PF_RULES_WUNLOCK();
2959 					free(ioes, M_TEMP);
2960 					goto fail; /* really bad */
2961 				}
2962 				break;
2963 			    }
2964 			default:
2965 				if ((error = pf_rollback_rules(ioe->ticket,
2966 				    ioe->rs_num, ioe->anchor))) {
2967 					PF_RULES_WUNLOCK();
2968 					free(ioes, M_TEMP);
2969 					goto fail; /* really bad */
2970 				}
2971 				break;
2972 			}
2973 		}
2974 		PF_RULES_WUNLOCK();
2975 		free(ioes, M_TEMP);
2976 		break;
2977 	}
2978 
2979 	case DIOCXCOMMIT: {
2980 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2981 		struct pfioc_trans_e	*ioe, *ioes;
2982 		struct pf_ruleset	*rs;
2983 		size_t			 totlen;
2984 		int			 i;
2985 
2986 		if (io->esize != sizeof(*ioe)) {
2987 			error = ENODEV;
2988 			break;
2989 		}
2990 		totlen = sizeof(struct pfioc_trans_e) * io->size;
2991 		ioes = malloc(totlen, M_TEMP, M_WAITOK);
2992 		error = copyin(io->array, ioes, totlen);
2993 		if (error) {
2994 			free(ioes, M_TEMP);
2995 			break;
2996 		}
2997 		PF_RULES_WLOCK();
2998 		/* First makes sure everything will succeed. */
2999 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3000 			switch (ioe->rs_num) {
3001 #ifdef ALTQ
3002 			case PF_RULESET_ALTQ:
3003 				if (ioe->anchor[0]) {
3004 					PF_RULES_WUNLOCK();
3005 					free(ioes, M_TEMP);
3006 					error = EINVAL;
3007 					goto fail;
3008 				}
3009 				if (!V_altqs_inactive_open || ioe->ticket !=
3010 				    V_ticket_altqs_inactive) {
3011 					PF_RULES_WUNLOCK();
3012 					free(ioes, M_TEMP);
3013 					error = EBUSY;
3014 					goto fail;
3015 				}
3016 				break;
3017 #endif /* ALTQ */
3018 			case PF_RULESET_TABLE:
3019 				rs = pf_find_ruleset(ioe->anchor);
3020 				if (rs == NULL || !rs->topen || ioe->ticket !=
3021 				    rs->tticket) {
3022 					PF_RULES_WUNLOCK();
3023 					free(ioes, M_TEMP);
3024 					error = EBUSY;
3025 					goto fail;
3026 				}
3027 				break;
3028 			default:
3029 				if (ioe->rs_num < 0 || ioe->rs_num >=
3030 				    PF_RULESET_MAX) {
3031 					PF_RULES_WUNLOCK();
3032 					free(ioes, M_TEMP);
3033 					error = EINVAL;
3034 					goto fail;
3035 				}
3036 				rs = pf_find_ruleset(ioe->anchor);
3037 				if (rs == NULL ||
3038 				    !rs->rules[ioe->rs_num].inactive.open ||
3039 				    rs->rules[ioe->rs_num].inactive.ticket !=
3040 				    ioe->ticket) {
3041 					PF_RULES_WUNLOCK();
3042 					free(ioes, M_TEMP);
3043 					error = EBUSY;
3044 					goto fail;
3045 				}
3046 				break;
3047 			}
3048 		}
3049 		/* Now do the commit - no errors should happen here. */
3050 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3051 			switch (ioe->rs_num) {
3052 #ifdef ALTQ
3053 			case PF_RULESET_ALTQ:
3054 				if ((error = pf_commit_altq(ioe->ticket))) {
3055 					PF_RULES_WUNLOCK();
3056 					free(ioes, M_TEMP);
3057 					goto fail; /* really bad */
3058 				}
3059 				break;
3060 #endif /* ALTQ */
3061 			case PF_RULESET_TABLE:
3062 			    {
3063 				struct pfr_table table;
3064 
3065 				bzero(&table, sizeof(table));
3066 				strlcpy(table.pfrt_anchor, ioe->anchor,
3067 				    sizeof(table.pfrt_anchor));
3068 				if ((error = pfr_ina_commit(&table,
3069 				    ioe->ticket, NULL, NULL, 0))) {
3070 					PF_RULES_WUNLOCK();
3071 					free(ioes, M_TEMP);
3072 					goto fail; /* really bad */
3073 				}
3074 				break;
3075 			    }
3076 			default:
3077 				if ((error = pf_commit_rules(ioe->ticket,
3078 				    ioe->rs_num, ioe->anchor))) {
3079 					PF_RULES_WUNLOCK();
3080 					free(ioes, M_TEMP);
3081 					goto fail; /* really bad */
3082 				}
3083 				break;
3084 			}
3085 		}
3086 		PF_RULES_WUNLOCK();
3087 		free(ioes, M_TEMP);
3088 		break;
3089 	}
3090 
3091 	case DIOCGETSRCNODES: {
3092 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
3093 		struct pf_srchash	*sh;
3094 		struct pf_src_node	*n, *p, *pstore;
3095 		uint32_t		 i, nr = 0;
3096 
3097 		if (psn->psn_len == 0) {
3098 			for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
3099 			    i++, sh++) {
3100 				PF_HASHROW_LOCK(sh);
3101 				LIST_FOREACH(n, &sh->nodes, entry)
3102 					nr++;
3103 				PF_HASHROW_UNLOCK(sh);
3104 			}
3105 			psn->psn_len = sizeof(struct pf_src_node) * nr;
3106 			break;
3107 		}
3108 
3109 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK);
3110 		for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
3111 		    i++, sh++) {
3112 		    PF_HASHROW_LOCK(sh);
3113 		    LIST_FOREACH(n, &sh->nodes, entry) {
3114 			int	secs = time_uptime, diff;
3115 
3116 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
3117 				break;
3118 
3119 			bcopy(n, p, sizeof(struct pf_src_node));
3120 			if (n->rule.ptr != NULL)
3121 				p->rule.nr = n->rule.ptr->nr;
3122 			p->creation = secs - p->creation;
3123 			if (p->expire > secs)
3124 				p->expire -= secs;
3125 			else
3126 				p->expire = 0;
3127 
3128 			/* Adjust the connection rate estimate. */
3129 			diff = secs - n->conn_rate.last;
3130 			if (diff >= n->conn_rate.seconds)
3131 				p->conn_rate.count = 0;
3132 			else
3133 				p->conn_rate.count -=
3134 				    n->conn_rate.count * diff /
3135 				    n->conn_rate.seconds;
3136 			p++;
3137 			nr++;
3138 		    }
3139 		    PF_HASHROW_UNLOCK(sh);
3140 		}
3141 		error = copyout(pstore, psn->psn_src_nodes,
3142 		    sizeof(struct pf_src_node) * nr);
3143 		if (error) {
3144 			free(pstore, M_TEMP);
3145 			break;
3146 		}
3147 		psn->psn_len = sizeof(struct pf_src_node) * nr;
3148 		free(pstore, M_TEMP);
3149 		break;
3150 	}
3151 
3152 	case DIOCCLRSRCNODES: {
3153 
3154 		pf_clear_srcnodes(NULL);
3155 		pf_purge_expired_src_nodes();
3156 		V_pf_status.src_nodes = 0;
3157 		break;
3158 	}
3159 
3160 	case DIOCKILLSRCNODES:
3161 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
3162 		break;
3163 
3164 	case DIOCSETHOSTID: {
3165 		u_int32_t	*hostid = (u_int32_t *)addr;
3166 
3167 		PF_RULES_WLOCK();
3168 		if (*hostid == 0)
3169 			V_pf_status.hostid = arc4random();
3170 		else
3171 			V_pf_status.hostid = *hostid;
3172 		PF_RULES_WUNLOCK();
3173 		break;
3174 	}
3175 
3176 	case DIOCOSFPFLUSH:
3177 		PF_RULES_WLOCK();
3178 		pf_osfp_flush();
3179 		PF_RULES_WUNLOCK();
3180 		break;
3181 
3182 	case DIOCIGETIFACES: {
3183 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3184 		struct pfi_kif *ifstore;
3185 		size_t bufsiz;
3186 
3187 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
3188 			error = ENODEV;
3189 			break;
3190 		}
3191 
3192 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
3193 		ifstore = malloc(bufsiz, M_TEMP, M_WAITOK);
3194 		PF_RULES_RLOCK();
3195 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
3196 		PF_RULES_RUNLOCK();
3197 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
3198 		free(ifstore, M_TEMP);
3199 		break;
3200 	}
3201 
3202 	case DIOCSETIFFLAG: {
3203 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3204 
3205 		PF_RULES_WLOCK();
3206 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3207 		PF_RULES_WUNLOCK();
3208 		break;
3209 	}
3210 
3211 	case DIOCCLRIFFLAG: {
3212 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3213 
3214 		PF_RULES_WLOCK();
3215 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3216 		PF_RULES_WUNLOCK();
3217 		break;
3218 	}
3219 
3220 	default:
3221 		error = ENODEV;
3222 		break;
3223 	}
3224 fail:
3225 	CURVNET_RESTORE();
3226 
3227 	return (error);
3228 }
3229 
3230 void
3231 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
3232 {
3233 	bzero(sp, sizeof(struct pfsync_state));
3234 
3235 	/* copy from state key */
3236 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
3237 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
3238 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
3239 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
3240 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
3241 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
3242 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
3243 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
3244 	sp->proto = st->key[PF_SK_WIRE]->proto;
3245 	sp->af = st->key[PF_SK_WIRE]->af;
3246 
3247 	/* copy from state */
3248 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
3249 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
3250 	sp->creation = htonl(time_uptime - st->creation);
3251 	sp->expire = pf_state_expires(st);
3252 	if (sp->expire <= time_uptime)
3253 		sp->expire = htonl(0);
3254 	else
3255 		sp->expire = htonl(sp->expire - time_uptime);
3256 
3257 	sp->direction = st->direction;
3258 	sp->log = st->log;
3259 	sp->timeout = st->timeout;
3260 	sp->state_flags = st->state_flags;
3261 	if (st->src_node)
3262 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
3263 	if (st->nat_src_node)
3264 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
3265 
3266 	sp->id = st->id;
3267 	sp->creatorid = st->creatorid;
3268 	pf_state_peer_hton(&st->src, &sp->src);
3269 	pf_state_peer_hton(&st->dst, &sp->dst);
3270 
3271 	if (st->rule.ptr == NULL)
3272 		sp->rule = htonl(-1);
3273 	else
3274 		sp->rule = htonl(st->rule.ptr->nr);
3275 	if (st->anchor.ptr == NULL)
3276 		sp->anchor = htonl(-1);
3277 	else
3278 		sp->anchor = htonl(st->anchor.ptr->nr);
3279 	if (st->nat_rule.ptr == NULL)
3280 		sp->nat_rule = htonl(-1);
3281 	else
3282 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
3283 
3284 	pf_state_counter_hton(st->packets[0], sp->packets[0]);
3285 	pf_state_counter_hton(st->packets[1], sp->packets[1]);
3286 	pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
3287 	pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
3288 
3289 }
3290 
3291 static void
3292 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
3293 {
3294 	struct pfr_ktable *kt;
3295 
3296 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
3297 
3298 	kt = aw->p.tbl;
3299 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
3300 		kt = kt->pfrkt_root;
3301 	aw->p.tbl = NULL;
3302 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
3303 		kt->pfrkt_cnt : -1;
3304 }
3305 
3306 /*
3307  * XXX - Check for version missmatch!!!
3308  */
3309 static void
3310 pf_clear_states(void)
3311 {
3312 	struct pf_state	*s;
3313 	u_int i;
3314 
3315 	for (i = 0; i <= V_pf_hashmask; i++) {
3316 		struct pf_idhash *ih = &V_pf_idhash[i];
3317 relock:
3318 		PF_HASHROW_LOCK(ih);
3319 		LIST_FOREACH(s, &ih->states, entry) {
3320 			s->timeout = PFTM_PURGE;
3321 			/* Don't send out individual delete messages. */
3322 			s->sync_state = PFSTATE_NOSYNC;
3323 			pf_unlink_state(s, PF_ENTER_LOCKED);
3324 			goto relock;
3325 		}
3326 		PF_HASHROW_UNLOCK(ih);
3327 	}
3328 }
3329 
3330 static int
3331 pf_clear_tables(void)
3332 {
3333 	struct pfioc_table io;
3334 	int error;
3335 
3336 	bzero(&io, sizeof(io));
3337 
3338 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
3339 	    io.pfrio_flags);
3340 
3341 	return (error);
3342 }
3343 
3344 static void
3345 pf_clear_srcnodes(struct pf_src_node *n)
3346 {
3347 	struct pf_state *s;
3348 	int i;
3349 
3350 	for (i = 0; i <= V_pf_hashmask; i++) {
3351 		struct pf_idhash *ih = &V_pf_idhash[i];
3352 
3353 		PF_HASHROW_LOCK(ih);
3354 		LIST_FOREACH(s, &ih->states, entry) {
3355 			if (n == NULL || n == s->src_node)
3356 				s->src_node = NULL;
3357 			if (n == NULL || n == s->nat_src_node)
3358 				s->nat_src_node = NULL;
3359 		}
3360 		PF_HASHROW_UNLOCK(ih);
3361 	}
3362 
3363 	if (n == NULL) {
3364 		struct pf_srchash *sh;
3365 
3366 		for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
3367 		    i++, sh++) {
3368 			PF_HASHROW_LOCK(sh);
3369 			LIST_FOREACH(n, &sh->nodes, entry) {
3370 				n->expire = 1;
3371 				n->states = 0;
3372 			}
3373 			PF_HASHROW_UNLOCK(sh);
3374 		}
3375 	} else {
3376 		/* XXX: hash slot should already be locked here. */
3377 		n->expire = 1;
3378 		n->states = 0;
3379 	}
3380 }
3381 
3382 static void
3383 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
3384 {
3385 	struct pf_src_node_list	 kill;
3386 
3387 	LIST_INIT(&kill);
3388 	for (int i = 0; i <= V_pf_srchashmask; i++) {
3389 		struct pf_srchash *sh = &V_pf_srchash[i];
3390 		struct pf_src_node *sn, *tmp;
3391 
3392 		PF_HASHROW_LOCK(sh);
3393 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
3394 			if (PF_MATCHA(psnk->psnk_src.neg,
3395 			      &psnk->psnk_src.addr.v.a.addr,
3396 			      &psnk->psnk_src.addr.v.a.mask,
3397 			      &sn->addr, sn->af) &&
3398 			    PF_MATCHA(psnk->psnk_dst.neg,
3399 			      &psnk->psnk_dst.addr.v.a.addr,
3400 			      &psnk->psnk_dst.addr.v.a.mask,
3401 			      &sn->raddr, sn->af)) {
3402 				pf_unlink_src_node_locked(sn);
3403 				LIST_INSERT_HEAD(&kill, sn, entry);
3404 				sn->expire = 1;
3405 			}
3406 		PF_HASHROW_UNLOCK(sh);
3407 	}
3408 
3409 	for (int i = 0; i <= V_pf_hashmask; i++) {
3410 		struct pf_idhash *ih = &V_pf_idhash[i];
3411 		struct pf_state *s;
3412 
3413 		PF_HASHROW_LOCK(ih);
3414 		LIST_FOREACH(s, &ih->states, entry) {
3415 			if (s->src_node && s->src_node->expire == 1) {
3416 #ifdef INVARIANTS
3417 				s->src_node->states--;
3418 #endif
3419 				s->src_node = NULL;
3420 			}
3421 			if (s->nat_src_node && s->nat_src_node->expire == 1) {
3422 #ifdef INVARIANTS
3423 				s->nat_src_node->states--;
3424 #endif
3425 				s->nat_src_node = NULL;
3426 			}
3427 		}
3428 		PF_HASHROW_UNLOCK(ih);
3429 	}
3430 
3431 	psnk->psnk_killed = pf_free_src_nodes(&kill);
3432 }
3433 
3434 /*
3435  * XXX - Check for version missmatch!!!
3436  */
3437 
3438 /*
3439  * Duplicate pfctl -Fa operation to get rid of as much as we can.
3440  */
3441 static int
3442 shutdown_pf(void)
3443 {
3444 	int error = 0;
3445 	u_int32_t t[5];
3446 	char nn = '\0';
3447 
3448 	V_pf_status.running = 0;
3449 
3450 	counter_u64_free(V_pf_default_rule.states_cur);
3451 	counter_u64_free(V_pf_default_rule.states_tot);
3452 	counter_u64_free(V_pf_default_rule.src_nodes);
3453 
3454 	do {
3455 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
3456 		    != 0) {
3457 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
3458 			break;
3459 		}
3460 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
3461 		    != 0) {
3462 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
3463 			break;		/* XXX: rollback? */
3464 		}
3465 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
3466 		    != 0) {
3467 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
3468 			break;		/* XXX: rollback? */
3469 		}
3470 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
3471 		    != 0) {
3472 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
3473 			break;		/* XXX: rollback? */
3474 		}
3475 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
3476 		    != 0) {
3477 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
3478 			break;		/* XXX: rollback? */
3479 		}
3480 
3481 		/* XXX: these should always succeed here */
3482 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
3483 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
3484 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
3485 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
3486 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
3487 
3488 		if ((error = pf_clear_tables()) != 0)
3489 			break;
3490 
3491 #ifdef ALTQ
3492 		if ((error = pf_begin_altq(&t[0])) != 0) {
3493 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
3494 			break;
3495 		}
3496 		pf_commit_altq(t[0]);
3497 #endif
3498 
3499 		pf_clear_states();
3500 
3501 		pf_clear_srcnodes(NULL);
3502 
3503 		/* status does not use malloced mem so no need to cleanup */
3504 		/* fingerprints and interfaces have thier own cleanup code */
3505 	} while(0);
3506 
3507 	return (error);
3508 }
3509 
3510 #ifdef INET
3511 static int
3512 pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3513     struct inpcb *inp)
3514 {
3515 	int chk;
3516 
3517 	chk = pf_test(PF_IN, ifp, m, inp);
3518 	if (chk && *m) {
3519 		m_freem(*m);
3520 		*m = NULL;
3521 	}
3522 
3523 	return (chk);
3524 }
3525 
3526 static int
3527 pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3528     struct inpcb *inp)
3529 {
3530 	int chk;
3531 
3532 	/* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3533 	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3534 		in_delayed_cksum(*m);
3535 		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3536 	}
3537 
3538 	chk = pf_test(PF_OUT, ifp, m, inp);
3539 	if (chk && *m) {
3540 		m_freem(*m);
3541 		*m = NULL;
3542 	}
3543 
3544 	return (chk);
3545 }
3546 #endif
3547 
3548 #ifdef INET6
3549 static int
3550 pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3551     struct inpcb *inp)
3552 {
3553 	int chk;
3554 
3555 	/*
3556 	 * In case of loopback traffic IPv6 uses the real interface in
3557 	 * order to support scoped addresses. In order to support stateful
3558 	 * filtering we have change this to lo0 as it is the case in IPv4.
3559 	 */
3560 	CURVNET_SET(ifp->if_vnet);
3561 	chk = pf_test6(PF_IN, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
3562 	CURVNET_RESTORE();
3563 	if (chk && *m) {
3564 		m_freem(*m);
3565 		*m = NULL;
3566 	}
3567 	return chk;
3568 }
3569 
3570 static int
3571 pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3572     struct inpcb *inp)
3573 {
3574 	int chk;
3575 
3576 	/* We need a proper CSUM before we start (s. OpenBSD ip_output) */
3577 	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3578 #ifdef INET
3579 		/* XXX-BZ copy&paste error from r126261? */
3580 		in_delayed_cksum(*m);
3581 #endif
3582 		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3583 	}
3584 	CURVNET_SET(ifp->if_vnet);
3585 	chk = pf_test6(PF_OUT, ifp, m, inp);
3586 	CURVNET_RESTORE();
3587 	if (chk && *m) {
3588 		m_freem(*m);
3589 		*m = NULL;
3590 	}
3591 	return chk;
3592 }
3593 #endif /* INET6 */
3594 
3595 static int
3596 hook_pf(void)
3597 {
3598 #ifdef INET
3599 	struct pfil_head *pfh_inet;
3600 #endif
3601 #ifdef INET6
3602 	struct pfil_head *pfh_inet6;
3603 #endif
3604 
3605 	if (V_pf_pfil_hooked)
3606 		return (0);
3607 
3608 #ifdef INET
3609 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3610 	if (pfh_inet == NULL)
3611 		return (ESRCH); /* XXX */
3612 	pfil_add_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet);
3613 	pfil_add_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet);
3614 #endif
3615 #ifdef INET6
3616 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3617 	if (pfh_inet6 == NULL) {
3618 #ifdef INET
3619 		pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3620 		    pfh_inet);
3621 		pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3622 		    pfh_inet);
3623 #endif
3624 		return (ESRCH); /* XXX */
3625 	}
3626 	pfil_add_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet6);
3627 	pfil_add_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet6);
3628 #endif
3629 
3630 	V_pf_pfil_hooked = 1;
3631 	return (0);
3632 }
3633 
3634 static int
3635 dehook_pf(void)
3636 {
3637 #ifdef INET
3638 	struct pfil_head *pfh_inet;
3639 #endif
3640 #ifdef INET6
3641 	struct pfil_head *pfh_inet6;
3642 #endif
3643 
3644 	if (V_pf_pfil_hooked == 0)
3645 		return (0);
3646 
3647 #ifdef INET
3648 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3649 	if (pfh_inet == NULL)
3650 		return (ESRCH); /* XXX */
3651 	pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3652 	    pfh_inet);
3653 	pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3654 	    pfh_inet);
3655 #endif
3656 #ifdef INET6
3657 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3658 	if (pfh_inet6 == NULL)
3659 		return (ESRCH); /* XXX */
3660 	pfil_remove_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK,
3661 	    pfh_inet6);
3662 	pfil_remove_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK,
3663 	    pfh_inet6);
3664 #endif
3665 
3666 	V_pf_pfil_hooked = 0;
3667 	return (0);
3668 }
3669 
3670 static int
3671 pf_load(void)
3672 {
3673 	int error;
3674 
3675 	VNET_ITERATOR_DECL(vnet_iter);
3676 
3677 	VNET_LIST_RLOCK();
3678 	VNET_FOREACH(vnet_iter) {
3679 		CURVNET_SET(vnet_iter);
3680 		V_pf_pfil_hooked = 0;
3681 		V_pf_end_threads = 0;
3682 		TAILQ_INIT(&V_pf_tags);
3683 		TAILQ_INIT(&V_pf_qids);
3684 		CURVNET_RESTORE();
3685 	}
3686 	VNET_LIST_RUNLOCK();
3687 
3688 	rw_init(&pf_rules_lock, "pf rulesets");
3689 
3690 	pf_dev = make_dev(&pf_cdevsw, 0, 0, 0, 0600, PF_NAME);
3691 	if ((error = pfattach()) != 0)
3692 		return (error);
3693 
3694 	return (0);
3695 }
3696 
3697 static int
3698 pf_unload(void)
3699 {
3700 	int error = 0;
3701 
3702 	PF_RULES_WLOCK();
3703 	V_pf_status.running = 0;
3704 	PF_RULES_WUNLOCK();
3705 	swi_remove(V_pf_swi_cookie);
3706 	error = dehook_pf();
3707 	if (error) {
3708 		/*
3709 		 * Should not happen!
3710 		 * XXX Due to error code ESRCH, kldunload will show
3711 		 * a message like 'No such process'.
3712 		 */
3713 		printf("%s : pfil unregisteration fail\n", __FUNCTION__);
3714 		return error;
3715 	}
3716 	PF_RULES_WLOCK();
3717 	shutdown_pf();
3718 	V_pf_end_threads = 1;
3719 	while (V_pf_end_threads < 2) {
3720 		wakeup_one(pf_purge_thread);
3721 		rw_sleep(pf_purge_thread, &pf_rules_lock, 0, "pftmo", 0);
3722 	}
3723 	pf_normalize_cleanup();
3724 	pfi_cleanup();
3725 	pfr_cleanup();
3726 	pf_osfp_flush();
3727 	pf_cleanup();
3728 	PF_RULES_WUNLOCK();
3729 	destroy_dev(pf_dev);
3730 	rw_destroy(&pf_rules_lock);
3731 
3732 	return (error);
3733 }
3734 
3735 static int
3736 pf_modevent(module_t mod, int type, void *data)
3737 {
3738 	int error = 0;
3739 
3740 	switch(type) {
3741 	case MOD_LOAD:
3742 		error = pf_load();
3743 		break;
3744 	case MOD_QUIESCE:
3745 		/*
3746 		 * Module should not be unloaded due to race conditions.
3747 		 */
3748 		error = EBUSY;
3749 		break;
3750 	case MOD_UNLOAD:
3751 		error = pf_unload();
3752 		break;
3753 	default:
3754 		error = EINVAL;
3755 		break;
3756 	}
3757 
3758 	return (error);
3759 }
3760 
3761 static moduledata_t pf_mod = {
3762 	"pf",
3763 	pf_modevent,
3764 	0
3765 };
3766 
3767 DECLARE_MODULE(pf, pf_mod, SI_SUB_PSEUDO, SI_ORDER_FIRST);
3768 MODULE_VERSION(pf, PF_MODVER);
3769