xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision 8d20be1e22095c27faf8fe8b2f0d089739cc742e)
1 /*-
2  * Copyright (c) 2001 Daniel Hartmeier
3  * Copyright (c) 2002,2003 Henning Brauer
4  * Copyright (c) 2012 Gleb Smirnoff <glebius@FreeBSD.org>
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  *
11  *    - Redistributions of source code must retain the above copyright
12  *      notice, this list of conditions and the following disclaimer.
13  *    - Redistributions in binary form must reproduce the above
14  *      copyright notice, this list of conditions and the following
15  *      disclaimer in the documentation and/or other materials provided
16  *      with the distribution.
17  *
18  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
19  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
20  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
21  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
22  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
23  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
24  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
25  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
26  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
28  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29  * POSSIBILITY OF SUCH DAMAGE.
30  *
31  * Effort sponsored in part by the Defense Advanced Research Projects
32  * Agency (DARPA) and Air Force Research Laboratory, Air Force
33  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
34  *
35  *	$OpenBSD: pf_ioctl.c,v 1.213 2009/02/15 21:46:12 mbalmer Exp $
36  */
37 
38 #include <sys/cdefs.h>
39 __FBSDID("$FreeBSD$");
40 
41 #include "opt_inet.h"
42 #include "opt_inet6.h"
43 #include "opt_bpf.h"
44 #include "opt_pf.h"
45 
46 #include <sys/param.h>
47 #include <sys/bus.h>
48 #include <sys/conf.h>
49 #include <sys/endian.h>
50 #include <sys/fcntl.h>
51 #include <sys/filio.h>
52 #include <sys/interrupt.h>
53 #include <sys/jail.h>
54 #include <sys/kernel.h>
55 #include <sys/kthread.h>
56 #include <sys/lock.h>
57 #include <sys/mbuf.h>
58 #include <sys/module.h>
59 #include <sys/proc.h>
60 #include <sys/rwlock.h>
61 #include <sys/smp.h>
62 #include <sys/socket.h>
63 #include <sys/sysctl.h>
64 #include <sys/md5.h>
65 #include <sys/ucred.h>
66 
67 #include <net/if.h>
68 #include <net/if_var.h>
69 #include <net/vnet.h>
70 #include <net/route.h>
71 #include <net/pfil.h>
72 #include <net/pfvar.h>
73 #include <net/if_pfsync.h>
74 #include <net/if_pflog.h>
75 
76 #include <netinet/in.h>
77 #include <netinet/ip.h>
78 #include <netinet/ip_var.h>
79 #include <netinet/ip_icmp.h>
80 
81 #ifdef INET6
82 #include <netinet/ip6.h>
83 #endif /* INET6 */
84 
85 #ifdef ALTQ
86 #include <altq/altq.h>
87 #endif
88 
89 static int		 pfattach(void);
90 static struct pf_pool	*pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t,
91 			    u_int8_t, u_int8_t, u_int8_t);
92 
93 static void		 pf_mv_pool(struct pf_palist *, struct pf_palist *);
94 static void		 pf_empty_pool(struct pf_palist *);
95 static int		 pfioctl(struct cdev *, u_long, caddr_t, int,
96 			    struct thread *);
97 #ifdef ALTQ
98 static int		 pf_begin_altq(u_int32_t *);
99 static int		 pf_rollback_altq(u_int32_t);
100 static int		 pf_commit_altq(u_int32_t);
101 static int		 pf_enable_altq(struct pf_altq *);
102 static int		 pf_disable_altq(struct pf_altq *);
103 static u_int32_t	 pf_qname2qid(char *);
104 static void		 pf_qid_unref(u_int32_t);
105 #endif /* ALTQ */
106 static int		 pf_begin_rules(u_int32_t *, int, const char *);
107 static int		 pf_rollback_rules(u_int32_t, int, char *);
108 static int		 pf_setup_pfsync_matching(struct pf_ruleset *);
109 static void		 pf_hash_rule(MD5_CTX *, struct pf_rule *);
110 static void		 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
111 static int		 pf_commit_rules(u_int32_t, int, char *);
112 static int		 pf_addr_setup(struct pf_ruleset *,
113 			    struct pf_addr_wrap *, sa_family_t);
114 static void		 pf_addr_copyout(struct pf_addr_wrap *);
115 
116 VNET_DEFINE(struct pf_rule,	pf_default_rule);
117 
118 #ifdef ALTQ
119 static VNET_DEFINE(int,		pf_altq_running);
120 #define	V_pf_altq_running	VNET(pf_altq_running)
121 #endif
122 
123 #define	TAGID_MAX	 50000
124 struct pf_tagname {
125 	TAILQ_ENTRY(pf_tagname)	entries;
126 	char			name[PF_TAG_NAME_SIZE];
127 	uint16_t		tag;
128 	int			ref;
129 };
130 
131 TAILQ_HEAD(pf_tags, pf_tagname);
132 #define	V_pf_tags		VNET(pf_tags)
133 VNET_DEFINE(struct pf_tags, pf_tags);
134 #define	V_pf_qids		VNET(pf_qids)
135 VNET_DEFINE(struct pf_tags, pf_qids);
136 static MALLOC_DEFINE(M_PFTAG, "pf_tag", "pf(4) tag names");
137 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
138 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
139 
140 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
141 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
142 #endif
143 
144 static u_int16_t	 tagname2tag(struct pf_tags *, char *);
145 static u_int16_t	 pf_tagname2tag(char *);
146 static void		 tag_unref(struct pf_tags *, u_int16_t);
147 
148 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
149 
150 struct cdev *pf_dev;
151 
152 /*
153  * XXX - These are new and need to be checked when moveing to a new version
154  */
155 static void		 pf_clear_states(void);
156 static int		 pf_clear_tables(void);
157 static void		 pf_clear_srcnodes(struct pf_src_node *);
158 static void		 pf_tbladdr_copyout(struct pf_addr_wrap *);
159 
160 /*
161  * Wrapper functions for pfil(9) hooks
162  */
163 #ifdef INET
164 static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp,
165     int dir, struct inpcb *inp);
166 static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp,
167     int dir, struct inpcb *inp);
168 #endif
169 #ifdef INET6
170 static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp,
171     int dir, struct inpcb *inp);
172 static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp,
173     int dir, struct inpcb *inp);
174 #endif
175 
176 static int		hook_pf(void);
177 static int		dehook_pf(void);
178 static int		shutdown_pf(void);
179 static int		pf_load(void);
180 static int		pf_unload(void);
181 
182 static struct cdevsw pf_cdevsw = {
183 	.d_ioctl =	pfioctl,
184 	.d_name =	PF_NAME,
185 	.d_version =	D_VERSION,
186 };
187 
188 static volatile VNET_DEFINE(int, pf_pfil_hooked);
189 #define V_pf_pfil_hooked	VNET(pf_pfil_hooked)
190 VNET_DEFINE(int,		pf_end_threads);
191 
192 struct rwlock			pf_rules_lock;
193 
194 /* pfsync */
195 pfsync_state_import_t 		*pfsync_state_import_ptr = NULL;
196 pfsync_insert_state_t		*pfsync_insert_state_ptr = NULL;
197 pfsync_update_state_t		*pfsync_update_state_ptr = NULL;
198 pfsync_delete_state_t		*pfsync_delete_state_ptr = NULL;
199 pfsync_clear_states_t		*pfsync_clear_states_ptr = NULL;
200 pfsync_defer_t			*pfsync_defer_ptr = NULL;
201 /* pflog */
202 pflog_packet_t			*pflog_packet_ptr = NULL;
203 
204 static int
205 pfattach(void)
206 {
207 	u_int32_t *my_timeout = V_pf_default_rule.timeout;
208 	int error;
209 
210 	pf_initialize();
211 	pfr_initialize();
212 	pfi_initialize();
213 	pf_normalize_init();
214 
215 	V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
216 	V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
217 
218 	RB_INIT(&V_pf_anchors);
219 	pf_init_ruleset(&pf_main_ruleset);
220 
221 	/* default rule should never be garbage collected */
222 	V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
223 #ifdef PF_DEFAULT_TO_DROP
224 	V_pf_default_rule.action = PF_DROP;
225 #else
226 	V_pf_default_rule.action = PF_PASS;
227 #endif
228 	V_pf_default_rule.nr = -1;
229 	V_pf_default_rule.rtableid = -1;
230 
231 	/* initialize default timeouts */
232 	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
233 	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
234 	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
235 	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
236 	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
237 	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
238 	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
239 	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
240 	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
241 	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
242 	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
243 	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
244 	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
245 	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
246 	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
247 	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
248 	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
249 	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
250 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
251 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
252 
253 	bzero(&V_pf_status, sizeof(V_pf_status));
254 	V_pf_status.debug = PF_DEBUG_URGENT;
255 
256 	V_pf_pfil_hooked = 0;
257 
258 	/* XXX do our best to avoid a conflict */
259 	V_pf_status.hostid = arc4random();
260 
261 	if ((error = kproc_create(pf_purge_thread, curvnet, NULL, 0, 0,
262 	    "pf purge")) != 0)
263 		/* XXXGL: leaked all above. */
264 		return (error);
265 	if ((error = swi_add(NULL, "pf send", pf_intr, curvnet, SWI_NET,
266 	    INTR_MPSAFE, &V_pf_swi_cookie)) != 0)
267 		/* XXXGL: leaked all above. */
268 		return (error);
269 
270 	return (0);
271 }
272 
273 static struct pf_pool *
274 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
275     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
276     u_int8_t check_ticket)
277 {
278 	struct pf_ruleset	*ruleset;
279 	struct pf_rule		*rule;
280 	int			 rs_num;
281 
282 	ruleset = pf_find_ruleset(anchor);
283 	if (ruleset == NULL)
284 		return (NULL);
285 	rs_num = pf_get_ruleset_number(rule_action);
286 	if (rs_num >= PF_RULESET_MAX)
287 		return (NULL);
288 	if (active) {
289 		if (check_ticket && ticket !=
290 		    ruleset->rules[rs_num].active.ticket)
291 			return (NULL);
292 		if (r_last)
293 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
294 			    pf_rulequeue);
295 		else
296 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
297 	} else {
298 		if (check_ticket && ticket !=
299 		    ruleset->rules[rs_num].inactive.ticket)
300 			return (NULL);
301 		if (r_last)
302 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
303 			    pf_rulequeue);
304 		else
305 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
306 	}
307 	if (!r_last) {
308 		while ((rule != NULL) && (rule->nr != rule_number))
309 			rule = TAILQ_NEXT(rule, entries);
310 	}
311 	if (rule == NULL)
312 		return (NULL);
313 
314 	return (&rule->rpool);
315 }
316 
317 static void
318 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
319 {
320 	struct pf_pooladdr	*mv_pool_pa;
321 
322 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
323 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
324 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
325 	}
326 }
327 
328 static void
329 pf_empty_pool(struct pf_palist *poola)
330 {
331 	struct pf_pooladdr *pa;
332 
333 	while ((pa = TAILQ_FIRST(poola)) != NULL) {
334 		switch (pa->addr.type) {
335 		case PF_ADDR_DYNIFTL:
336 			pfi_dynaddr_remove(pa->addr.p.dyn);
337 			break;
338 		case PF_ADDR_TABLE:
339 			pfr_detach_table(pa->addr.p.tbl);
340 			break;
341 		}
342 		if (pa->kif)
343 			pfi_kif_unref(pa->kif);
344 		TAILQ_REMOVE(poola, pa, entries);
345 		free(pa, M_PFRULE);
346 	}
347 }
348 
349 static void
350 pf_unlink_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
351 {
352 
353 	PF_RULES_WASSERT();
354 
355 	TAILQ_REMOVE(rulequeue, rule, entries);
356 
357 	PF_UNLNKDRULES_LOCK();
358 	rule->rule_flag |= PFRULE_REFS;
359 	TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
360 	PF_UNLNKDRULES_UNLOCK();
361 }
362 
363 void
364 pf_free_rule(struct pf_rule *rule)
365 {
366 
367 	PF_RULES_WASSERT();
368 
369 	if (rule->tag)
370 		tag_unref(&V_pf_tags, rule->tag);
371 	if (rule->match_tag)
372 		tag_unref(&V_pf_tags, rule->match_tag);
373 #ifdef ALTQ
374 	if (rule->pqid != rule->qid)
375 		pf_qid_unref(rule->pqid);
376 	pf_qid_unref(rule->qid);
377 #endif
378 	switch (rule->src.addr.type) {
379 	case PF_ADDR_DYNIFTL:
380 		pfi_dynaddr_remove(rule->src.addr.p.dyn);
381 		break;
382 	case PF_ADDR_TABLE:
383 		pfr_detach_table(rule->src.addr.p.tbl);
384 		break;
385 	}
386 	switch (rule->dst.addr.type) {
387 	case PF_ADDR_DYNIFTL:
388 		pfi_dynaddr_remove(rule->dst.addr.p.dyn);
389 		break;
390 	case PF_ADDR_TABLE:
391 		pfr_detach_table(rule->dst.addr.p.tbl);
392 		break;
393 	}
394 	if (rule->overload_tbl)
395 		pfr_detach_table(rule->overload_tbl);
396 	if (rule->kif)
397 		pfi_kif_unref(rule->kif);
398 	pf_anchor_remove(rule);
399 	pf_empty_pool(&rule->rpool.list);
400 	free(rule, M_PFRULE);
401 }
402 
403 static u_int16_t
404 tagname2tag(struct pf_tags *head, char *tagname)
405 {
406 	struct pf_tagname	*tag, *p = NULL;
407 	u_int16_t		 new_tagid = 1;
408 
409 	PF_RULES_WASSERT();
410 
411 	TAILQ_FOREACH(tag, head, entries)
412 		if (strcmp(tagname, tag->name) == 0) {
413 			tag->ref++;
414 			return (tag->tag);
415 		}
416 
417 	/*
418 	 * to avoid fragmentation, we do a linear search from the beginning
419 	 * and take the first free slot we find. if there is none or the list
420 	 * is empty, append a new entry at the end.
421 	 */
422 
423 	/* new entry */
424 	if (!TAILQ_EMPTY(head))
425 		for (p = TAILQ_FIRST(head); p != NULL &&
426 		    p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
427 			new_tagid = p->tag + 1;
428 
429 	if (new_tagid > TAGID_MAX)
430 		return (0);
431 
432 	/* allocate and fill new struct pf_tagname */
433 	tag = malloc(sizeof(*tag), M_PFTAG, M_NOWAIT|M_ZERO);
434 	if (tag == NULL)
435 		return (0);
436 	strlcpy(tag->name, tagname, sizeof(tag->name));
437 	tag->tag = new_tagid;
438 	tag->ref++;
439 
440 	if (p != NULL)	/* insert new entry before p */
441 		TAILQ_INSERT_BEFORE(p, tag, entries);
442 	else	/* either list empty or no free slot in between */
443 		TAILQ_INSERT_TAIL(head, tag, entries);
444 
445 	return (tag->tag);
446 }
447 
448 static void
449 tag_unref(struct pf_tags *head, u_int16_t tag)
450 {
451 	struct pf_tagname	*p, *next;
452 
453 	PF_RULES_WASSERT();
454 
455 	for (p = TAILQ_FIRST(head); p != NULL; p = next) {
456 		next = TAILQ_NEXT(p, entries);
457 		if (tag == p->tag) {
458 			if (--p->ref == 0) {
459 				TAILQ_REMOVE(head, p, entries);
460 				free(p, M_PFTAG);
461 			}
462 			break;
463 		}
464 	}
465 }
466 
467 static u_int16_t
468 pf_tagname2tag(char *tagname)
469 {
470 	return (tagname2tag(&V_pf_tags, tagname));
471 }
472 
473 #ifdef ALTQ
474 static u_int32_t
475 pf_qname2qid(char *qname)
476 {
477 	return ((u_int32_t)tagname2tag(&V_pf_qids, qname));
478 }
479 
480 static void
481 pf_qid_unref(u_int32_t qid)
482 {
483 	tag_unref(&V_pf_qids, (u_int16_t)qid);
484 }
485 
486 static int
487 pf_begin_altq(u_int32_t *ticket)
488 {
489 	struct pf_altq	*altq;
490 	int		 error = 0;
491 
492 	PF_RULES_WASSERT();
493 
494 	/* Purge the old altq list */
495 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
496 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
497 		if (altq->qname[0] == 0 &&
498 		    (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
499 			/* detach and destroy the discipline */
500 			error = altq_remove(altq);
501 		} else
502 			pf_qid_unref(altq->qid);
503 		free(altq, M_PFALTQ);
504 	}
505 	if (error)
506 		return (error);
507 	*ticket = ++V_ticket_altqs_inactive;
508 	V_altqs_inactive_open = 1;
509 	return (0);
510 }
511 
512 static int
513 pf_rollback_altq(u_int32_t ticket)
514 {
515 	struct pf_altq	*altq;
516 	int		 error = 0;
517 
518 	PF_RULES_WASSERT();
519 
520 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
521 		return (0);
522 	/* Purge the old altq list */
523 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
524 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
525 		if (altq->qname[0] == 0 &&
526 		   (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
527 			/* detach and destroy the discipline */
528 			error = altq_remove(altq);
529 		} else
530 			pf_qid_unref(altq->qid);
531 		free(altq, M_PFALTQ);
532 	}
533 	V_altqs_inactive_open = 0;
534 	return (error);
535 }
536 
537 static int
538 pf_commit_altq(u_int32_t ticket)
539 {
540 	struct pf_altqqueue	*old_altqs;
541 	struct pf_altq		*altq;
542 	int			 err, error = 0;
543 
544 	PF_RULES_WASSERT();
545 
546 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
547 		return (EBUSY);
548 
549 	/* swap altqs, keep the old. */
550 	old_altqs = V_pf_altqs_active;
551 	V_pf_altqs_active = V_pf_altqs_inactive;
552 	V_pf_altqs_inactive = old_altqs;
553 	V_ticket_altqs_active = V_ticket_altqs_inactive;
554 
555 	/* Attach new disciplines */
556 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
557 	if (altq->qname[0] == 0 &&
558 	   (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
559 			/* attach the discipline */
560 			error = altq_pfattach(altq);
561 			if (error == 0 && V_pf_altq_running)
562 				error = pf_enable_altq(altq);
563 			if (error != 0)
564 				return (error);
565 		}
566 	}
567 
568 	/* Purge the old altq list */
569 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
570 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
571 		if (altq->qname[0] == 0 &&
572 		    (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
573 			/* detach and destroy the discipline */
574 			if (V_pf_altq_running)
575 				error = pf_disable_altq(altq);
576 			err = altq_pfdetach(altq);
577 			if (err != 0 && error == 0)
578 				error = err;
579 			err = altq_remove(altq);
580 			if (err != 0 && error == 0)
581 				error = err;
582 		} else
583 			pf_qid_unref(altq->qid);
584 		free(altq, M_PFALTQ);
585 	}
586 
587 	V_altqs_inactive_open = 0;
588 	return (error);
589 }
590 
591 static int
592 pf_enable_altq(struct pf_altq *altq)
593 {
594 	struct ifnet		*ifp;
595 	struct tb_profile	 tb;
596 	int			 error = 0;
597 
598 	if ((ifp = ifunit(altq->ifname)) == NULL)
599 		return (EINVAL);
600 
601 	if (ifp->if_snd.altq_type != ALTQT_NONE)
602 		error = altq_enable(&ifp->if_snd);
603 
604 	/* set tokenbucket regulator */
605 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
606 		tb.rate = altq->ifbandwidth;
607 		tb.depth = altq->tbrsize;
608 		error = tbr_set(&ifp->if_snd, &tb);
609 	}
610 
611 	return (error);
612 }
613 
614 static int
615 pf_disable_altq(struct pf_altq *altq)
616 {
617 	struct ifnet		*ifp;
618 	struct tb_profile	 tb;
619 	int			 error;
620 
621 	if ((ifp = ifunit(altq->ifname)) == NULL)
622 		return (EINVAL);
623 
624 	/*
625 	 * when the discipline is no longer referenced, it was overridden
626 	 * by a new one.  if so, just return.
627 	 */
628 	if (altq->altq_disc != ifp->if_snd.altq_disc)
629 		return (0);
630 
631 	error = altq_disable(&ifp->if_snd);
632 
633 	if (error == 0) {
634 		/* clear tokenbucket regulator */
635 		tb.rate = 0;
636 		error = tbr_set(&ifp->if_snd, &tb);
637 	}
638 
639 	return (error);
640 }
641 
642 void
643 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
644 {
645 	struct ifnet	*ifp1;
646 	struct pf_altq	*a1, *a2, *a3;
647 	u_int32_t	 ticket;
648 	int		 error = 0;
649 
650 	/* Interrupt userland queue modifications */
651 	if (V_altqs_inactive_open)
652 		pf_rollback_altq(V_ticket_altqs_inactive);
653 
654 	/* Start new altq ruleset */
655 	if (pf_begin_altq(&ticket))
656 		return;
657 
658 	/* Copy the current active set */
659 	TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
660 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
661 		if (a2 == NULL) {
662 			error = ENOMEM;
663 			break;
664 		}
665 		bcopy(a1, a2, sizeof(struct pf_altq));
666 
667 		if (a2->qname[0] != 0) {
668 			if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
669 				error = EBUSY;
670 				free(a2, M_PFALTQ);
671 				break;
672 			}
673 			a2->altq_disc = NULL;
674 			TAILQ_FOREACH(a3, V_pf_altqs_inactive, entries) {
675 				if (strncmp(a3->ifname, a2->ifname,
676 				    IFNAMSIZ) == 0 && a3->qname[0] == 0) {
677 					a2->altq_disc = a3->altq_disc;
678 					break;
679 				}
680 			}
681 		}
682 		/* Deactivate the interface in question */
683 		a2->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
684 		if ((ifp1 = ifunit(a2->ifname)) == NULL ||
685 		    (remove && ifp1 == ifp)) {
686 			a2->local_flags |= PFALTQ_FLAG_IF_REMOVED;
687 		} else {
688 			error = altq_add(a2);
689 
690 			if (ticket != V_ticket_altqs_inactive)
691 				error = EBUSY;
692 
693 			if (error) {
694 				free(a2, M_PFALTQ);
695 				break;
696 			}
697 		}
698 
699 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
700 	}
701 
702 	if (error != 0)
703 		pf_rollback_altq(ticket);
704 	else
705 		pf_commit_altq(ticket);
706 }
707 #endif /* ALTQ */
708 
709 static int
710 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
711 {
712 	struct pf_ruleset	*rs;
713 	struct pf_rule		*rule;
714 
715 	PF_RULES_WASSERT();
716 
717 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
718 		return (EINVAL);
719 	rs = pf_find_or_create_ruleset(anchor);
720 	if (rs == NULL)
721 		return (EINVAL);
722 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
723 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
724 		rs->rules[rs_num].inactive.rcount--;
725 	}
726 	*ticket = ++rs->rules[rs_num].inactive.ticket;
727 	rs->rules[rs_num].inactive.open = 1;
728 	return (0);
729 }
730 
731 static int
732 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
733 {
734 	struct pf_ruleset	*rs;
735 	struct pf_rule		*rule;
736 
737 	PF_RULES_WASSERT();
738 
739 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
740 		return (EINVAL);
741 	rs = pf_find_ruleset(anchor);
742 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
743 	    rs->rules[rs_num].inactive.ticket != ticket)
744 		return (0);
745 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
746 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
747 		rs->rules[rs_num].inactive.rcount--;
748 	}
749 	rs->rules[rs_num].inactive.open = 0;
750 	return (0);
751 }
752 
753 #define PF_MD5_UPD(st, elm)						\
754 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
755 
756 #define PF_MD5_UPD_STR(st, elm)						\
757 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
758 
759 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
760 		(stor) = htonl((st)->elm);				\
761 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
762 } while (0)
763 
764 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
765 		(stor) = htons((st)->elm);				\
766 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
767 } while (0)
768 
769 static void
770 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
771 {
772 	PF_MD5_UPD(pfr, addr.type);
773 	switch (pfr->addr.type) {
774 		case PF_ADDR_DYNIFTL:
775 			PF_MD5_UPD(pfr, addr.v.ifname);
776 			PF_MD5_UPD(pfr, addr.iflags);
777 			break;
778 		case PF_ADDR_TABLE:
779 			PF_MD5_UPD(pfr, addr.v.tblname);
780 			break;
781 		case PF_ADDR_ADDRMASK:
782 			/* XXX ignore af? */
783 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
784 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
785 			break;
786 	}
787 
788 	PF_MD5_UPD(pfr, port[0]);
789 	PF_MD5_UPD(pfr, port[1]);
790 	PF_MD5_UPD(pfr, neg);
791 	PF_MD5_UPD(pfr, port_op);
792 }
793 
794 static void
795 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
796 {
797 	u_int16_t x;
798 	u_int32_t y;
799 
800 	pf_hash_rule_addr(ctx, &rule->src);
801 	pf_hash_rule_addr(ctx, &rule->dst);
802 	PF_MD5_UPD_STR(rule, label);
803 	PF_MD5_UPD_STR(rule, ifname);
804 	PF_MD5_UPD_STR(rule, match_tagname);
805 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
806 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
807 	PF_MD5_UPD_HTONL(rule, prob, y);
808 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
809 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
810 	PF_MD5_UPD(rule, uid.op);
811 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
812 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
813 	PF_MD5_UPD(rule, gid.op);
814 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
815 	PF_MD5_UPD(rule, action);
816 	PF_MD5_UPD(rule, direction);
817 	PF_MD5_UPD(rule, af);
818 	PF_MD5_UPD(rule, quick);
819 	PF_MD5_UPD(rule, ifnot);
820 	PF_MD5_UPD(rule, match_tag_not);
821 	PF_MD5_UPD(rule, natpass);
822 	PF_MD5_UPD(rule, keep_state);
823 	PF_MD5_UPD(rule, proto);
824 	PF_MD5_UPD(rule, type);
825 	PF_MD5_UPD(rule, code);
826 	PF_MD5_UPD(rule, flags);
827 	PF_MD5_UPD(rule, flagset);
828 	PF_MD5_UPD(rule, allow_opts);
829 	PF_MD5_UPD(rule, rt);
830 	PF_MD5_UPD(rule, tos);
831 }
832 
833 static int
834 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
835 {
836 	struct pf_ruleset	*rs;
837 	struct pf_rule		*rule, **old_array;
838 	struct pf_rulequeue	*old_rules;
839 	int			 error;
840 	u_int32_t		 old_rcount;
841 
842 	PF_RULES_WASSERT();
843 
844 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
845 		return (EINVAL);
846 	rs = pf_find_ruleset(anchor);
847 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
848 	    ticket != rs->rules[rs_num].inactive.ticket)
849 		return (EBUSY);
850 
851 	/* Calculate checksum for the main ruleset */
852 	if (rs == &pf_main_ruleset) {
853 		error = pf_setup_pfsync_matching(rs);
854 		if (error != 0)
855 			return (error);
856 	}
857 
858 	/* Swap rules, keep the old. */
859 	old_rules = rs->rules[rs_num].active.ptr;
860 	old_rcount = rs->rules[rs_num].active.rcount;
861 	old_array = rs->rules[rs_num].active.ptr_array;
862 
863 	rs->rules[rs_num].active.ptr =
864 	    rs->rules[rs_num].inactive.ptr;
865 	rs->rules[rs_num].active.ptr_array =
866 	    rs->rules[rs_num].inactive.ptr_array;
867 	rs->rules[rs_num].active.rcount =
868 	    rs->rules[rs_num].inactive.rcount;
869 	rs->rules[rs_num].inactive.ptr = old_rules;
870 	rs->rules[rs_num].inactive.ptr_array = old_array;
871 	rs->rules[rs_num].inactive.rcount = old_rcount;
872 
873 	rs->rules[rs_num].active.ticket =
874 	    rs->rules[rs_num].inactive.ticket;
875 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
876 
877 
878 	/* Purge the old rule list. */
879 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
880 		pf_unlink_rule(old_rules, rule);
881 	if (rs->rules[rs_num].inactive.ptr_array)
882 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
883 	rs->rules[rs_num].inactive.ptr_array = NULL;
884 	rs->rules[rs_num].inactive.rcount = 0;
885 	rs->rules[rs_num].inactive.open = 0;
886 	pf_remove_if_empty_ruleset(rs);
887 
888 	return (0);
889 }
890 
891 static int
892 pf_setup_pfsync_matching(struct pf_ruleset *rs)
893 {
894 	MD5_CTX			 ctx;
895 	struct pf_rule		*rule;
896 	int			 rs_cnt;
897 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
898 
899 	MD5Init(&ctx);
900 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
901 		/* XXX PF_RULESET_SCRUB as well? */
902 		if (rs_cnt == PF_RULESET_SCRUB)
903 			continue;
904 
905 		if (rs->rules[rs_cnt].inactive.ptr_array)
906 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
907 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
908 
909 		if (rs->rules[rs_cnt].inactive.rcount) {
910 			rs->rules[rs_cnt].inactive.ptr_array =
911 			    malloc(sizeof(caddr_t) *
912 			    rs->rules[rs_cnt].inactive.rcount,
913 			    M_TEMP, M_NOWAIT);
914 
915 			if (!rs->rules[rs_cnt].inactive.ptr_array)
916 				return (ENOMEM);
917 		}
918 
919 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
920 		    entries) {
921 			pf_hash_rule(&ctx, rule);
922 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
923 		}
924 	}
925 
926 	MD5Final(digest, &ctx);
927 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
928 	return (0);
929 }
930 
931 static int
932 pf_addr_setup(struct pf_ruleset *ruleset, struct pf_addr_wrap *addr,
933     sa_family_t af)
934 {
935 	int error = 0;
936 
937 	switch (addr->type) {
938 	case PF_ADDR_TABLE:
939 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
940 		if (addr->p.tbl == NULL)
941 			error = ENOMEM;
942 		break;
943 	case PF_ADDR_DYNIFTL:
944 		error = pfi_dynaddr_setup(addr, af);
945 		break;
946 	}
947 
948 	return (error);
949 }
950 
951 static void
952 pf_addr_copyout(struct pf_addr_wrap *addr)
953 {
954 
955 	switch (addr->type) {
956 	case PF_ADDR_DYNIFTL:
957 		pfi_dynaddr_copyout(addr);
958 		break;
959 	case PF_ADDR_TABLE:
960 		pf_tbladdr_copyout(addr);
961 		break;
962 	}
963 }
964 
965 static int
966 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
967 {
968 	int			 error = 0;
969 
970 	/* XXX keep in sync with switch() below */
971 	if (securelevel_gt(td->td_ucred, 2))
972 		switch (cmd) {
973 		case DIOCGETRULES:
974 		case DIOCGETRULE:
975 		case DIOCGETADDRS:
976 		case DIOCGETADDR:
977 		case DIOCGETSTATE:
978 		case DIOCSETSTATUSIF:
979 		case DIOCGETSTATUS:
980 		case DIOCCLRSTATUS:
981 		case DIOCNATLOOK:
982 		case DIOCSETDEBUG:
983 		case DIOCGETSTATES:
984 		case DIOCGETTIMEOUT:
985 		case DIOCCLRRULECTRS:
986 		case DIOCGETLIMIT:
987 		case DIOCGETALTQS:
988 		case DIOCGETALTQ:
989 		case DIOCGETQSTATS:
990 		case DIOCGETRULESETS:
991 		case DIOCGETRULESET:
992 		case DIOCRGETTABLES:
993 		case DIOCRGETTSTATS:
994 		case DIOCRCLRTSTATS:
995 		case DIOCRCLRADDRS:
996 		case DIOCRADDADDRS:
997 		case DIOCRDELADDRS:
998 		case DIOCRSETADDRS:
999 		case DIOCRGETADDRS:
1000 		case DIOCRGETASTATS:
1001 		case DIOCRCLRASTATS:
1002 		case DIOCRTSTADDRS:
1003 		case DIOCOSFPGET:
1004 		case DIOCGETSRCNODES:
1005 		case DIOCCLRSRCNODES:
1006 		case DIOCIGETIFACES:
1007 		case DIOCGIFSPEED:
1008 		case DIOCSETIFFLAG:
1009 		case DIOCCLRIFFLAG:
1010 			break;
1011 		case DIOCRCLRTABLES:
1012 		case DIOCRADDTABLES:
1013 		case DIOCRDELTABLES:
1014 		case DIOCRSETTFLAGS:
1015 			if (((struct pfioc_table *)addr)->pfrio_flags &
1016 			    PFR_FLAG_DUMMY)
1017 				break; /* dummy operation ok */
1018 			return (EPERM);
1019 		default:
1020 			return (EPERM);
1021 		}
1022 
1023 	if (!(flags & FWRITE))
1024 		switch (cmd) {
1025 		case DIOCGETRULES:
1026 		case DIOCGETADDRS:
1027 		case DIOCGETADDR:
1028 		case DIOCGETSTATE:
1029 		case DIOCGETSTATUS:
1030 		case DIOCGETSTATES:
1031 		case DIOCGETTIMEOUT:
1032 		case DIOCGETLIMIT:
1033 		case DIOCGETALTQS:
1034 		case DIOCGETALTQ:
1035 		case DIOCGETQSTATS:
1036 		case DIOCGETRULESETS:
1037 		case DIOCGETRULESET:
1038 		case DIOCNATLOOK:
1039 		case DIOCRGETTABLES:
1040 		case DIOCRGETTSTATS:
1041 		case DIOCRGETADDRS:
1042 		case DIOCRGETASTATS:
1043 		case DIOCRTSTADDRS:
1044 		case DIOCOSFPGET:
1045 		case DIOCGETSRCNODES:
1046 		case DIOCIGETIFACES:
1047 		case DIOCGIFSPEED:
1048 			break;
1049 		case DIOCRCLRTABLES:
1050 		case DIOCRADDTABLES:
1051 		case DIOCRDELTABLES:
1052 		case DIOCRCLRTSTATS:
1053 		case DIOCRCLRADDRS:
1054 		case DIOCRADDADDRS:
1055 		case DIOCRDELADDRS:
1056 		case DIOCRSETADDRS:
1057 		case DIOCRSETTFLAGS:
1058 			if (((struct pfioc_table *)addr)->pfrio_flags &
1059 			    PFR_FLAG_DUMMY) {
1060 				flags |= FWRITE; /* need write lock for dummy */
1061 				break; /* dummy operation ok */
1062 			}
1063 			return (EACCES);
1064 		case DIOCGETRULE:
1065 			if (((struct pfioc_rule *)addr)->action ==
1066 			    PF_GET_CLR_CNTR)
1067 				return (EACCES);
1068 			break;
1069 		default:
1070 			return (EACCES);
1071 		}
1072 
1073 	CURVNET_SET(TD_TO_VNET(td));
1074 
1075 	switch (cmd) {
1076 	case DIOCSTART:
1077 		PF_RULES_WLOCK();
1078 		if (V_pf_status.running)
1079 			error = EEXIST;
1080 		else {
1081 			int cpu;
1082 
1083 			PF_RULES_WUNLOCK();
1084 			error = hook_pf();
1085 			if (error) {
1086 				DPFPRINTF(PF_DEBUG_MISC,
1087 				    ("pf: pfil registration failed\n"));
1088 				break;
1089 			}
1090 			PF_RULES_WLOCK();
1091 			V_pf_status.running = 1;
1092 			V_pf_status.since = time_second;
1093 
1094 			CPU_FOREACH(cpu)
1095 				V_pf_stateid[cpu] = time_second;
1096 
1097 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1098 		}
1099 		PF_RULES_WUNLOCK();
1100 		break;
1101 
1102 	case DIOCSTOP:
1103 		PF_RULES_WLOCK();
1104 		if (!V_pf_status.running)
1105 			error = ENOENT;
1106 		else {
1107 			V_pf_status.running = 0;
1108 			PF_RULES_WUNLOCK();
1109 			error = dehook_pf();
1110 			if (error) {
1111 				V_pf_status.running = 1;
1112 				DPFPRINTF(PF_DEBUG_MISC,
1113 				    ("pf: pfil unregistration failed\n"));
1114 			}
1115 			PF_RULES_WLOCK();
1116 			V_pf_status.since = time_second;
1117 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1118 		}
1119 		PF_RULES_WUNLOCK();
1120 		break;
1121 
1122 	case DIOCADDRULE: {
1123 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1124 		struct pf_ruleset	*ruleset;
1125 		struct pf_rule		*rule, *tail;
1126 		struct pf_pooladdr	*pa;
1127 		struct pfi_kif		*kif = NULL;
1128 		int			 rs_num;
1129 
1130 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1131 			error = EINVAL;
1132 			break;
1133 		}
1134 #ifndef INET
1135 		if (pr->rule.af == AF_INET) {
1136 			error = EAFNOSUPPORT;
1137 			break;
1138 		}
1139 #endif /* INET */
1140 #ifndef INET6
1141 		if (pr->rule.af == AF_INET6) {
1142 			error = EAFNOSUPPORT;
1143 			break;
1144 		}
1145 #endif /* INET6 */
1146 
1147 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
1148 		bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1149 		if (rule->ifname[0])
1150 			kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1151 		rule->cuid = td->td_ucred->cr_ruid;
1152 		rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1153 		TAILQ_INIT(&rule->rpool.list);
1154 
1155 #define	ERROUT(x)	{ error = (x); goto DIOCADDRULE_error; }
1156 
1157 		PF_RULES_WLOCK();
1158 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1159 		ruleset = pf_find_ruleset(pr->anchor);
1160 		if (ruleset == NULL)
1161 			ERROUT(EINVAL);
1162 		rs_num = pf_get_ruleset_number(pr->rule.action);
1163 		if (rs_num >= PF_RULESET_MAX)
1164 			ERROUT(EINVAL);
1165 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1166 			DPFPRINTF(PF_DEBUG_MISC,
1167 			    ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
1168 			    ruleset->rules[rs_num].inactive.ticket));
1169 			ERROUT(EBUSY);
1170 		}
1171 		if (pr->pool_ticket != V_ticket_pabuf) {
1172 			DPFPRINTF(PF_DEBUG_MISC,
1173 			    ("pool_ticket: %d != %d\n", pr->pool_ticket,
1174 			    V_ticket_pabuf));
1175 			ERROUT(EBUSY);
1176 		}
1177 
1178 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1179 		    pf_rulequeue);
1180 		if (tail)
1181 			rule->nr = tail->nr + 1;
1182 		else
1183 			rule->nr = 0;
1184 		if (rule->ifname[0]) {
1185 			rule->kif = pfi_kif_attach(kif, rule->ifname);
1186 			pfi_kif_ref(rule->kif);
1187 		} else
1188 			rule->kif = NULL;
1189 
1190 		if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
1191 			error = EBUSY;
1192 
1193 #ifdef ALTQ
1194 		/* set queue IDs */
1195 		if (rule->qname[0] != 0) {
1196 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1197 				error = EBUSY;
1198 			else if (rule->pqname[0] != 0) {
1199 				if ((rule->pqid =
1200 				    pf_qname2qid(rule->pqname)) == 0)
1201 					error = EBUSY;
1202 			} else
1203 				rule->pqid = rule->qid;
1204 		}
1205 #endif
1206 		if (rule->tagname[0])
1207 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1208 				error = EBUSY;
1209 		if (rule->match_tagname[0])
1210 			if ((rule->match_tag =
1211 			    pf_tagname2tag(rule->match_tagname)) == 0)
1212 				error = EBUSY;
1213 		if (rule->rt && !rule->direction)
1214 			error = EINVAL;
1215 		if (!rule->log)
1216 			rule->logif = 0;
1217 		if (rule->logif >= PFLOGIFS_MAX)
1218 			error = EINVAL;
1219 		if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1220 			error = ENOMEM;
1221 		if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1222 			error = ENOMEM;
1223 		if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1224 			error = EINVAL;
1225 		TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1226 			if (pa->addr.type == PF_ADDR_TABLE) {
1227 				pa->addr.p.tbl = pfr_attach_table(ruleset,
1228 				    pa->addr.v.tblname);
1229 				if (pa->addr.p.tbl == NULL)
1230 					error = ENOMEM;
1231 			}
1232 
1233 		if (rule->overload_tblname[0]) {
1234 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1235 			    rule->overload_tblname)) == NULL)
1236 				error = EINVAL;
1237 			else
1238 				rule->overload_tbl->pfrkt_flags |=
1239 				    PFR_TFLAG_ACTIVE;
1240 		}
1241 
1242 		pf_mv_pool(&V_pf_pabuf, &rule->rpool.list);
1243 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1244 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1245 		    (rule->rt > PF_FASTROUTE)) &&
1246 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1247 			error = EINVAL;
1248 
1249 		if (error) {
1250 			pf_free_rule(rule);
1251 			PF_RULES_WUNLOCK();
1252 			break;
1253 		}
1254 
1255 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1256 		rule->evaluations = rule->packets[0] = rule->packets[1] =
1257 		    rule->bytes[0] = rule->bytes[1] = 0;
1258 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1259 		    rule, entries);
1260 		ruleset->rules[rs_num].inactive.rcount++;
1261 		PF_RULES_WUNLOCK();
1262 		break;
1263 
1264 #undef ERROUT
1265 DIOCADDRULE_error:
1266 		PF_RULES_WUNLOCK();
1267 		free(rule, M_PFRULE);
1268 		if (kif)
1269 			free(kif, PFI_MTYPE);
1270 		break;
1271 	}
1272 
1273 	case DIOCGETRULES: {
1274 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1275 		struct pf_ruleset	*ruleset;
1276 		struct pf_rule		*tail;
1277 		int			 rs_num;
1278 
1279 		PF_RULES_WLOCK();
1280 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1281 		ruleset = pf_find_ruleset(pr->anchor);
1282 		if (ruleset == NULL) {
1283 			PF_RULES_WUNLOCK();
1284 			error = EINVAL;
1285 			break;
1286 		}
1287 		rs_num = pf_get_ruleset_number(pr->rule.action);
1288 		if (rs_num >= PF_RULESET_MAX) {
1289 			PF_RULES_WUNLOCK();
1290 			error = EINVAL;
1291 			break;
1292 		}
1293 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1294 		    pf_rulequeue);
1295 		if (tail)
1296 			pr->nr = tail->nr + 1;
1297 		else
1298 			pr->nr = 0;
1299 		pr->ticket = ruleset->rules[rs_num].active.ticket;
1300 		PF_RULES_WUNLOCK();
1301 		break;
1302 	}
1303 
1304 	case DIOCGETRULE: {
1305 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1306 		struct pf_ruleset	*ruleset;
1307 		struct pf_rule		*rule;
1308 		int			 rs_num, i;
1309 
1310 		PF_RULES_WLOCK();
1311 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1312 		ruleset = pf_find_ruleset(pr->anchor);
1313 		if (ruleset == NULL) {
1314 			PF_RULES_WUNLOCK();
1315 			error = EINVAL;
1316 			break;
1317 		}
1318 		rs_num = pf_get_ruleset_number(pr->rule.action);
1319 		if (rs_num >= PF_RULESET_MAX) {
1320 			PF_RULES_WUNLOCK();
1321 			error = EINVAL;
1322 			break;
1323 		}
1324 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1325 			PF_RULES_WUNLOCK();
1326 			error = EBUSY;
1327 			break;
1328 		}
1329 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1330 		while ((rule != NULL) && (rule->nr != pr->nr))
1331 			rule = TAILQ_NEXT(rule, entries);
1332 		if (rule == NULL) {
1333 			PF_RULES_WUNLOCK();
1334 			error = EBUSY;
1335 			break;
1336 		}
1337 		bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1338 		if (pf_anchor_copyout(ruleset, rule, pr)) {
1339 			PF_RULES_WUNLOCK();
1340 			error = EBUSY;
1341 			break;
1342 		}
1343 		pf_addr_copyout(&pr->rule.src.addr);
1344 		pf_addr_copyout(&pr->rule.dst.addr);
1345 		for (i = 0; i < PF_SKIP_COUNT; ++i)
1346 			if (rule->skip[i].ptr == NULL)
1347 				pr->rule.skip[i].nr = -1;
1348 			else
1349 				pr->rule.skip[i].nr =
1350 				    rule->skip[i].ptr->nr;
1351 
1352 		if (pr->action == PF_GET_CLR_CNTR) {
1353 			rule->evaluations = 0;
1354 			rule->packets[0] = rule->packets[1] = 0;
1355 			rule->bytes[0] = rule->bytes[1] = 0;
1356 			rule->states_tot = 0;
1357 		}
1358 		PF_RULES_WUNLOCK();
1359 		break;
1360 	}
1361 
1362 	case DIOCCHANGERULE: {
1363 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1364 		struct pf_ruleset	*ruleset;
1365 		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1366 		struct pfi_kif		*kif = NULL;
1367 		struct pf_pooladdr	*pa;
1368 		u_int32_t		 nr = 0;
1369 		int			 rs_num;
1370 
1371 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1372 		    pcr->action > PF_CHANGE_GET_TICKET) {
1373 			error = EINVAL;
1374 			break;
1375 		}
1376 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1377 			error = EINVAL;
1378 			break;
1379 		}
1380 
1381 		if (pcr->action != PF_CHANGE_REMOVE) {
1382 #ifndef INET
1383 			if (pcr->rule.af == AF_INET) {
1384 				error = EAFNOSUPPORT;
1385 				break;
1386 			}
1387 #endif /* INET */
1388 #ifndef INET6
1389 			if (pcr->rule.af == AF_INET6) {
1390 				error = EAFNOSUPPORT;
1391 				break;
1392 			}
1393 #endif /* INET6 */
1394 			newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
1395 			bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1396 			newrule->cuid = td->td_ucred->cr_ruid;
1397 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1398 			TAILQ_INIT(&newrule->rpool.list);
1399 			/* Initialize refcounting. */
1400 			newrule->states_cur = 0;
1401 			newrule->entries.tqe_prev = NULL;
1402 
1403 			if (newrule->ifname[0])
1404 				kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1405 		}
1406 
1407 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGERULE_error; }
1408 
1409 		PF_RULES_WLOCK();
1410 		if (!(pcr->action == PF_CHANGE_REMOVE ||
1411 		    pcr->action == PF_CHANGE_GET_TICKET) &&
1412 		    pcr->pool_ticket != V_ticket_pabuf)
1413 			ERROUT(EBUSY);
1414 
1415 		ruleset = pf_find_ruleset(pcr->anchor);
1416 		if (ruleset == NULL)
1417 			ERROUT(EINVAL);
1418 
1419 		rs_num = pf_get_ruleset_number(pcr->rule.action);
1420 		if (rs_num >= PF_RULESET_MAX)
1421 			ERROUT(EINVAL);
1422 
1423 		if (pcr->action == PF_CHANGE_GET_TICKET) {
1424 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1425 			ERROUT(0);
1426 		} else if (pcr->ticket !=
1427 			    ruleset->rules[rs_num].active.ticket)
1428 				ERROUT(EINVAL);
1429 
1430 		if (pcr->action != PF_CHANGE_REMOVE) {
1431 			if (newrule->ifname[0]) {
1432 				newrule->kif = pfi_kif_attach(kif,
1433 				    newrule->ifname);
1434 				pfi_kif_ref(newrule->kif);
1435 			} else
1436 				newrule->kif = NULL;
1437 
1438 			if (newrule->rtableid > 0 &&
1439 			    newrule->rtableid >= rt_numfibs)
1440 				error = EBUSY;
1441 
1442 #ifdef ALTQ
1443 			/* set queue IDs */
1444 			if (newrule->qname[0] != 0) {
1445 				if ((newrule->qid =
1446 				    pf_qname2qid(newrule->qname)) == 0)
1447 					error = EBUSY;
1448 				else if (newrule->pqname[0] != 0) {
1449 					if ((newrule->pqid =
1450 					    pf_qname2qid(newrule->pqname)) == 0)
1451 						error = EBUSY;
1452 				} else
1453 					newrule->pqid = newrule->qid;
1454 			}
1455 #endif /* ALTQ */
1456 			if (newrule->tagname[0])
1457 				if ((newrule->tag =
1458 				    pf_tagname2tag(newrule->tagname)) == 0)
1459 					error = EBUSY;
1460 			if (newrule->match_tagname[0])
1461 				if ((newrule->match_tag = pf_tagname2tag(
1462 				    newrule->match_tagname)) == 0)
1463 					error = EBUSY;
1464 			if (newrule->rt && !newrule->direction)
1465 				error = EINVAL;
1466 			if (!newrule->log)
1467 				newrule->logif = 0;
1468 			if (newrule->logif >= PFLOGIFS_MAX)
1469 				error = EINVAL;
1470 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
1471 				error = ENOMEM;
1472 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
1473 				error = ENOMEM;
1474 			if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1475 				error = EINVAL;
1476 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1477 				if (pa->addr.type == PF_ADDR_TABLE) {
1478 					pa->addr.p.tbl =
1479 					    pfr_attach_table(ruleset,
1480 					    pa->addr.v.tblname);
1481 					if (pa->addr.p.tbl == NULL)
1482 						error = ENOMEM;
1483 				}
1484 
1485 			if (newrule->overload_tblname[0]) {
1486 				if ((newrule->overload_tbl = pfr_attach_table(
1487 				    ruleset, newrule->overload_tblname)) ==
1488 				    NULL)
1489 					error = EINVAL;
1490 				else
1491 					newrule->overload_tbl->pfrkt_flags |=
1492 					    PFR_TFLAG_ACTIVE;
1493 			}
1494 
1495 			pf_mv_pool(&V_pf_pabuf, &newrule->rpool.list);
1496 			if (((((newrule->action == PF_NAT) ||
1497 			    (newrule->action == PF_RDR) ||
1498 			    (newrule->action == PF_BINAT) ||
1499 			    (newrule->rt > PF_FASTROUTE)) &&
1500 			    !newrule->anchor)) &&
1501 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1502 				error = EINVAL;
1503 
1504 			if (error) {
1505 				pf_free_rule(newrule);
1506 				PF_RULES_WUNLOCK();
1507 				break;
1508 			}
1509 
1510 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1511 			newrule->evaluations = 0;
1512 			newrule->packets[0] = newrule->packets[1] = 0;
1513 			newrule->bytes[0] = newrule->bytes[1] = 0;
1514 		}
1515 		pf_empty_pool(&V_pf_pabuf);
1516 
1517 		if (pcr->action == PF_CHANGE_ADD_HEAD)
1518 			oldrule = TAILQ_FIRST(
1519 			    ruleset->rules[rs_num].active.ptr);
1520 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1521 			oldrule = TAILQ_LAST(
1522 			    ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1523 		else {
1524 			oldrule = TAILQ_FIRST(
1525 			    ruleset->rules[rs_num].active.ptr);
1526 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1527 				oldrule = TAILQ_NEXT(oldrule, entries);
1528 			if (oldrule == NULL) {
1529 				if (newrule != NULL)
1530 					pf_free_rule(newrule);
1531 				PF_RULES_WUNLOCK();
1532 				error = EINVAL;
1533 				break;
1534 			}
1535 		}
1536 
1537 		if (pcr->action == PF_CHANGE_REMOVE) {
1538 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
1539 			    oldrule);
1540 			ruleset->rules[rs_num].active.rcount--;
1541 		} else {
1542 			if (oldrule == NULL)
1543 				TAILQ_INSERT_TAIL(
1544 				    ruleset->rules[rs_num].active.ptr,
1545 				    newrule, entries);
1546 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1547 			    pcr->action == PF_CHANGE_ADD_BEFORE)
1548 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1549 			else
1550 				TAILQ_INSERT_AFTER(
1551 				    ruleset->rules[rs_num].active.ptr,
1552 				    oldrule, newrule, entries);
1553 			ruleset->rules[rs_num].active.rcount++;
1554 		}
1555 
1556 		nr = 0;
1557 		TAILQ_FOREACH(oldrule,
1558 		    ruleset->rules[rs_num].active.ptr, entries)
1559 			oldrule->nr = nr++;
1560 
1561 		ruleset->rules[rs_num].active.ticket++;
1562 
1563 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1564 		pf_remove_if_empty_ruleset(ruleset);
1565 
1566 		PF_RULES_WUNLOCK();
1567 		break;
1568 
1569 #undef ERROUT
1570 DIOCCHANGERULE_error:
1571 		PF_RULES_WUNLOCK();
1572 		if (newrule != NULL)
1573 			free(newrule, M_PFRULE);
1574 		if (kif != NULL)
1575 			free(kif, PFI_MTYPE);
1576 		break;
1577 	}
1578 
1579 	case DIOCCLRSTATES: {
1580 		struct pf_state		*s;
1581 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1582 		u_int			 i, killed = 0;
1583 
1584 		for (i = 0; i <= V_pf_hashmask; i++) {
1585 			struct pf_idhash *ih = &V_pf_idhash[i];
1586 
1587 relock_DIOCCLRSTATES:
1588 			PF_HASHROW_LOCK(ih);
1589 			LIST_FOREACH(s, &ih->states, entry)
1590 				if (!psk->psk_ifname[0] ||
1591 				    !strcmp(psk->psk_ifname,
1592 				    s->kif->pfik_name)) {
1593 					/*
1594 					 * Don't send out individual
1595 					 * delete messages.
1596 					 */
1597 					s->state_flags |= PFSTATE_NOSYNC;
1598 					pf_unlink_state(s, PF_ENTER_LOCKED);
1599 					killed++;
1600 					goto relock_DIOCCLRSTATES;
1601 				}
1602 			PF_HASHROW_UNLOCK(ih);
1603 		}
1604 		psk->psk_killed = killed;
1605 		if (pfsync_clear_states_ptr != NULL)
1606 			pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
1607 		break;
1608 	}
1609 
1610 	case DIOCKILLSTATES: {
1611 		struct pf_state		*s;
1612 		struct pf_state_key	*sk;
1613 		struct pf_addr		*srcaddr, *dstaddr;
1614 		u_int16_t		 srcport, dstport;
1615 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
1616 		u_int			 i, killed = 0;
1617 
1618 		if (psk->psk_pfcmp.id) {
1619 			if (psk->psk_pfcmp.creatorid == 0)
1620 				psk->psk_pfcmp.creatorid = V_pf_status.hostid;
1621 			if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
1622 			    psk->psk_pfcmp.creatorid))) {
1623 				pf_unlink_state(s, PF_ENTER_LOCKED);
1624 				psk->psk_killed = 1;
1625 			}
1626 			break;
1627 		}
1628 
1629 		for (i = 0; i <= V_pf_hashmask; i++) {
1630 			struct pf_idhash *ih = &V_pf_idhash[i];
1631 
1632 relock_DIOCKILLSTATES:
1633 			PF_HASHROW_LOCK(ih);
1634 			LIST_FOREACH(s, &ih->states, entry) {
1635 				sk = s->key[PF_SK_WIRE];
1636 				if (s->direction == PF_OUT) {
1637 					srcaddr = &sk->addr[1];
1638 					dstaddr = &sk->addr[0];
1639 					srcport = sk->port[0];
1640 					dstport = sk->port[0];
1641 				} else {
1642 					srcaddr = &sk->addr[0];
1643 					dstaddr = &sk->addr[1];
1644 					srcport = sk->port[0];
1645 					dstport = sk->port[0];
1646 				}
1647 
1648 				if ((!psk->psk_af || sk->af == psk->psk_af)
1649 				    && (!psk->psk_proto || psk->psk_proto ==
1650 				    sk->proto) &&
1651 				    PF_MATCHA(psk->psk_src.neg,
1652 				    &psk->psk_src.addr.v.a.addr,
1653 				    &psk->psk_src.addr.v.a.mask,
1654 				    srcaddr, sk->af) &&
1655 				    PF_MATCHA(psk->psk_dst.neg,
1656 				    &psk->psk_dst.addr.v.a.addr,
1657 				    &psk->psk_dst.addr.v.a.mask,
1658 				    dstaddr, sk->af) &&
1659 				    (psk->psk_src.port_op == 0 ||
1660 				    pf_match_port(psk->psk_src.port_op,
1661 				    psk->psk_src.port[0], psk->psk_src.port[1],
1662 				    srcport)) &&
1663 				    (psk->psk_dst.port_op == 0 ||
1664 				    pf_match_port(psk->psk_dst.port_op,
1665 				    psk->psk_dst.port[0], psk->psk_dst.port[1],
1666 				    dstport)) &&
1667 				    (!psk->psk_label[0] ||
1668 				    (s->rule.ptr->label[0] &&
1669 				    !strcmp(psk->psk_label,
1670 				    s->rule.ptr->label))) &&
1671 				    (!psk->psk_ifname[0] ||
1672 				    !strcmp(psk->psk_ifname,
1673 				    s->kif->pfik_name))) {
1674 					pf_unlink_state(s, PF_ENTER_LOCKED);
1675 					killed++;
1676 					goto relock_DIOCKILLSTATES;
1677 				}
1678 			}
1679 			PF_HASHROW_UNLOCK(ih);
1680 		}
1681 		psk->psk_killed = killed;
1682 		break;
1683 	}
1684 
1685 	case DIOCADDSTATE: {
1686 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1687 		struct pfsync_state	*sp = &ps->state;
1688 
1689 		if (sp->timeout >= PFTM_MAX &&
1690 		    sp->timeout != PFTM_UNTIL_PACKET) {
1691 			error = EINVAL;
1692 			break;
1693 		}
1694 		if (pfsync_state_import_ptr != NULL) {
1695 			PF_RULES_RLOCK();
1696 			error = pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
1697 			PF_RULES_RUNLOCK();
1698 		} else
1699 			error = EOPNOTSUPP;
1700 		break;
1701 	}
1702 
1703 	case DIOCGETSTATE: {
1704 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1705 		struct pf_state		*s;
1706 
1707 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
1708 		if (s == NULL) {
1709 			error = ENOENT;
1710 			break;
1711 		}
1712 
1713 		pfsync_state_export(&ps->state, s);
1714 		PF_STATE_UNLOCK(s);
1715 		break;
1716 	}
1717 
1718 	case DIOCGETSTATES: {
1719 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
1720 		struct pf_state		*s;
1721 		struct pfsync_state	*pstore, *p;
1722 		int i, nr;
1723 
1724 		if (ps->ps_len == 0) {
1725 			nr = uma_zone_get_cur(V_pf_state_z);
1726 			ps->ps_len = sizeof(struct pfsync_state) * nr;
1727 			break;
1728 		}
1729 
1730 		p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK);
1731 		nr = 0;
1732 
1733 		for (i = 0; i <= V_pf_hashmask; i++) {
1734 			struct pf_idhash *ih = &V_pf_idhash[i];
1735 
1736 			PF_HASHROW_LOCK(ih);
1737 			LIST_FOREACH(s, &ih->states, entry) {
1738 
1739 				if (s->timeout == PFTM_UNLINKED)
1740 					continue;
1741 
1742 				if ((nr+1) * sizeof(*p) > ps->ps_len) {
1743 					PF_HASHROW_UNLOCK(ih);
1744 					goto DIOCGETSTATES_full;
1745 				}
1746 				pfsync_state_export(p, s);
1747 				p++;
1748 				nr++;
1749 			}
1750 			PF_HASHROW_UNLOCK(ih);
1751 		}
1752 DIOCGETSTATES_full:
1753 		error = copyout(pstore, ps->ps_states,
1754 		    sizeof(struct pfsync_state) * nr);
1755 		if (error) {
1756 			free(pstore, M_TEMP);
1757 			break;
1758 		}
1759 		ps->ps_len = sizeof(struct pfsync_state) * nr;
1760 		free(pstore, M_TEMP);
1761 
1762 		break;
1763 	}
1764 
1765 	case DIOCGETSTATUS: {
1766 		struct pf_status *s = (struct pf_status *)addr;
1767 		PF_RULES_RLOCK();
1768 		bcopy(&V_pf_status, s, sizeof(struct pf_status));
1769 		pfi_update_status(s->ifname, s);
1770 		PF_RULES_RUNLOCK();
1771 		break;
1772 	}
1773 
1774 	case DIOCSETSTATUSIF: {
1775 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
1776 
1777 		if (pi->ifname[0] == 0) {
1778 			bzero(V_pf_status.ifname, IFNAMSIZ);
1779 			break;
1780 		}
1781 		PF_RULES_WLOCK();
1782 		strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
1783 		PF_RULES_WUNLOCK();
1784 		break;
1785 	}
1786 
1787 	case DIOCCLRSTATUS: {
1788 		PF_RULES_WLOCK();
1789 		bzero(V_pf_status.counters, sizeof(V_pf_status.counters));
1790 		bzero(V_pf_status.fcounters, sizeof(V_pf_status.fcounters));
1791 		bzero(V_pf_status.scounters, sizeof(V_pf_status.scounters));
1792 		V_pf_status.since = time_second;
1793 		if (*V_pf_status.ifname)
1794 			pfi_update_status(V_pf_status.ifname, NULL);
1795 		PF_RULES_WUNLOCK();
1796 		break;
1797 	}
1798 
1799 	case DIOCNATLOOK: {
1800 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
1801 		struct pf_state_key	*sk;
1802 		struct pf_state		*state;
1803 		struct pf_state_key_cmp	 key;
1804 		int			 m = 0, direction = pnl->direction;
1805 		int			 sidx, didx;
1806 
1807 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
1808 		sidx = (direction == PF_IN) ? 1 : 0;
1809 		didx = (direction == PF_IN) ? 0 : 1;
1810 
1811 		if (!pnl->proto ||
1812 		    PF_AZERO(&pnl->saddr, pnl->af) ||
1813 		    PF_AZERO(&pnl->daddr, pnl->af) ||
1814 		    ((pnl->proto == IPPROTO_TCP ||
1815 		    pnl->proto == IPPROTO_UDP) &&
1816 		    (!pnl->dport || !pnl->sport)))
1817 			error = EINVAL;
1818 		else {
1819 			bzero(&key, sizeof(key));
1820 			key.af = pnl->af;
1821 			key.proto = pnl->proto;
1822 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
1823 			key.port[sidx] = pnl->sport;
1824 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
1825 			key.port[didx] = pnl->dport;
1826 
1827 			state = pf_find_state_all(&key, direction, &m);
1828 
1829 			if (m > 1)
1830 				error = E2BIG;	/* more than one state */
1831 			else if (state != NULL) {
1832 				/* XXXGL: not locked read */
1833 				sk = state->key[sidx];
1834 				PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
1835 				pnl->rsport = sk->port[sidx];
1836 				PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
1837 				pnl->rdport = sk->port[didx];
1838 			} else
1839 				error = ENOENT;
1840 		}
1841 		break;
1842 	}
1843 
1844 	case DIOCSETTIMEOUT: {
1845 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1846 		int		 old;
1847 
1848 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
1849 		    pt->seconds < 0) {
1850 			error = EINVAL;
1851 			break;
1852 		}
1853 		PF_RULES_WLOCK();
1854 		old = V_pf_default_rule.timeout[pt->timeout];
1855 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
1856 			pt->seconds = 1;
1857 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
1858 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
1859 			wakeup(pf_purge_thread);
1860 		pt->seconds = old;
1861 		PF_RULES_WUNLOCK();
1862 		break;
1863 	}
1864 
1865 	case DIOCGETTIMEOUT: {
1866 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1867 
1868 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
1869 			error = EINVAL;
1870 			break;
1871 		}
1872 		PF_RULES_RLOCK();
1873 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
1874 		PF_RULES_RUNLOCK();
1875 		break;
1876 	}
1877 
1878 	case DIOCGETLIMIT: {
1879 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1880 
1881 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
1882 			error = EINVAL;
1883 			break;
1884 		}
1885 		PF_RULES_RLOCK();
1886 		pl->limit = V_pf_limits[pl->index].limit;
1887 		PF_RULES_RUNLOCK();
1888 		break;
1889 	}
1890 
1891 	case DIOCSETLIMIT: {
1892 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1893 		int			 old_limit;
1894 
1895 		PF_RULES_WLOCK();
1896 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
1897 		    V_pf_limits[pl->index].zone == NULL) {
1898 			PF_RULES_WUNLOCK();
1899 			error = EINVAL;
1900 			break;
1901 		}
1902 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
1903 		old_limit = V_pf_limits[pl->index].limit;
1904 		V_pf_limits[pl->index].limit = pl->limit;
1905 		pl->limit = old_limit;
1906 		PF_RULES_WUNLOCK();
1907 		break;
1908 	}
1909 
1910 	case DIOCSETDEBUG: {
1911 		u_int32_t	*level = (u_int32_t *)addr;
1912 
1913 		PF_RULES_WLOCK();
1914 		V_pf_status.debug = *level;
1915 		PF_RULES_WUNLOCK();
1916 		break;
1917 	}
1918 
1919 	case DIOCCLRRULECTRS: {
1920 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
1921 		struct pf_ruleset	*ruleset = &pf_main_ruleset;
1922 		struct pf_rule		*rule;
1923 
1924 		PF_RULES_WLOCK();
1925 		TAILQ_FOREACH(rule,
1926 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
1927 			rule->evaluations = 0;
1928 			rule->packets[0] = rule->packets[1] = 0;
1929 			rule->bytes[0] = rule->bytes[1] = 0;
1930 		}
1931 		PF_RULES_WUNLOCK();
1932 		break;
1933 	}
1934 
1935 	case DIOCGIFSPEED: {
1936 		struct pf_ifspeed	*psp = (struct pf_ifspeed *)addr;
1937 		struct pf_ifspeed	ps;
1938 		struct ifnet		*ifp;
1939 
1940 		if (psp->ifname[0] != 0) {
1941 			/* Can we completely trust user-land? */
1942 			strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
1943 			ifp = ifunit(ps.ifname);
1944 			if (ifp != NULL)
1945 				psp->baudrate = ifp->if_baudrate;
1946 			else
1947 				error = EINVAL;
1948 		} else
1949 			error = EINVAL;
1950 		break;
1951 	}
1952 
1953 #ifdef ALTQ
1954 	case DIOCSTARTALTQ: {
1955 		struct pf_altq		*altq;
1956 
1957 		PF_RULES_WLOCK();
1958 		/* enable all altq interfaces on active list */
1959 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1960 			if (altq->qname[0] == 0 && (altq->local_flags &
1961 			    PFALTQ_FLAG_IF_REMOVED) == 0) {
1962 				error = pf_enable_altq(altq);
1963 				if (error != 0)
1964 					break;
1965 			}
1966 		}
1967 		if (error == 0)
1968 			V_pf_altq_running = 1;
1969 		PF_RULES_WUNLOCK();
1970 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
1971 		break;
1972 	}
1973 
1974 	case DIOCSTOPALTQ: {
1975 		struct pf_altq		*altq;
1976 
1977 		PF_RULES_WLOCK();
1978 		/* disable all altq interfaces on active list */
1979 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1980 			if (altq->qname[0] == 0 && (altq->local_flags &
1981 			    PFALTQ_FLAG_IF_REMOVED) == 0) {
1982 				error = pf_disable_altq(altq);
1983 				if (error != 0)
1984 					break;
1985 			}
1986 		}
1987 		if (error == 0)
1988 			V_pf_altq_running = 0;
1989 		PF_RULES_WUNLOCK();
1990 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
1991 		break;
1992 	}
1993 
1994 	case DIOCADDALTQ: {
1995 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
1996 		struct pf_altq		*altq, *a;
1997 		struct ifnet		*ifp;
1998 
1999 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK);
2000 		bcopy(&pa->altq, altq, sizeof(struct pf_altq));
2001 		altq->local_flags = 0;
2002 
2003 		PF_RULES_WLOCK();
2004 		if (pa->ticket != V_ticket_altqs_inactive) {
2005 			PF_RULES_WUNLOCK();
2006 			free(altq, M_PFALTQ);
2007 			error = EBUSY;
2008 			break;
2009 		}
2010 
2011 		/*
2012 		 * if this is for a queue, find the discipline and
2013 		 * copy the necessary fields
2014 		 */
2015 		if (altq->qname[0] != 0) {
2016 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2017 				PF_RULES_WUNLOCK();
2018 				error = EBUSY;
2019 				free(altq, M_PFALTQ);
2020 				break;
2021 			}
2022 			altq->altq_disc = NULL;
2023 			TAILQ_FOREACH(a, V_pf_altqs_inactive, entries) {
2024 				if (strncmp(a->ifname, altq->ifname,
2025 				    IFNAMSIZ) == 0 && a->qname[0] == 0) {
2026 					altq->altq_disc = a->altq_disc;
2027 					break;
2028 				}
2029 			}
2030 		}
2031 
2032 		if ((ifp = ifunit(altq->ifname)) == NULL)
2033 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
2034 		else
2035 			error = altq_add(altq);
2036 
2037 		if (error) {
2038 			PF_RULES_WUNLOCK();
2039 			free(altq, M_PFALTQ);
2040 			break;
2041 		}
2042 
2043 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
2044 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2045 		PF_RULES_WUNLOCK();
2046 		break;
2047 	}
2048 
2049 	case DIOCGETALTQS: {
2050 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2051 		struct pf_altq		*altq;
2052 
2053 		PF_RULES_RLOCK();
2054 		pa->nr = 0;
2055 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
2056 			pa->nr++;
2057 		pa->ticket = V_ticket_altqs_active;
2058 		PF_RULES_RUNLOCK();
2059 		break;
2060 	}
2061 
2062 	case DIOCGETALTQ: {
2063 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2064 		struct pf_altq		*altq;
2065 		u_int32_t		 nr;
2066 
2067 		PF_RULES_RLOCK();
2068 		if (pa->ticket != V_ticket_altqs_active) {
2069 			PF_RULES_RUNLOCK();
2070 			error = EBUSY;
2071 			break;
2072 		}
2073 		nr = 0;
2074 		altq = TAILQ_FIRST(V_pf_altqs_active);
2075 		while ((altq != NULL) && (nr < pa->nr)) {
2076 			altq = TAILQ_NEXT(altq, entries);
2077 			nr++;
2078 		}
2079 		if (altq == NULL) {
2080 			PF_RULES_RUNLOCK();
2081 			error = EBUSY;
2082 			break;
2083 		}
2084 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2085 		PF_RULES_RUNLOCK();
2086 		break;
2087 	}
2088 
2089 	case DIOCCHANGEALTQ:
2090 		/* CHANGEALTQ not supported yet! */
2091 		error = ENODEV;
2092 		break;
2093 
2094 	case DIOCGETQSTATS: {
2095 		struct pfioc_qstats	*pq = (struct pfioc_qstats *)addr;
2096 		struct pf_altq		*altq;
2097 		u_int32_t		 nr;
2098 		int			 nbytes;
2099 
2100 		PF_RULES_RLOCK();
2101 		if (pq->ticket != V_ticket_altqs_active) {
2102 			PF_RULES_RUNLOCK();
2103 			error = EBUSY;
2104 			break;
2105 		}
2106 		nbytes = pq->nbytes;
2107 		nr = 0;
2108 		altq = TAILQ_FIRST(V_pf_altqs_active);
2109 		while ((altq != NULL) && (nr < pq->nr)) {
2110 			altq = TAILQ_NEXT(altq, entries);
2111 			nr++;
2112 		}
2113 		if (altq == NULL) {
2114 			PF_RULES_RUNLOCK();
2115 			error = EBUSY;
2116 			break;
2117 		}
2118 
2119 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
2120 			PF_RULES_RUNLOCK();
2121 			error = ENXIO;
2122 			break;
2123 		}
2124 		PF_RULES_RUNLOCK();
2125 		error = altq_getqstats(altq, pq->buf, &nbytes);
2126 		if (error == 0) {
2127 			pq->scheduler = altq->scheduler;
2128 			pq->nbytes = nbytes;
2129 		}
2130 		break;
2131 	}
2132 #endif /* ALTQ */
2133 
2134 	case DIOCBEGINADDRS: {
2135 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2136 
2137 		PF_RULES_WLOCK();
2138 		pf_empty_pool(&V_pf_pabuf);
2139 		pp->ticket = ++V_ticket_pabuf;
2140 		PF_RULES_WUNLOCK();
2141 		break;
2142 	}
2143 
2144 	case DIOCADDADDR: {
2145 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2146 		struct pf_pooladdr	*pa;
2147 		struct pfi_kif		*kif = NULL;
2148 
2149 #ifndef INET
2150 		if (pp->af == AF_INET) {
2151 			error = EAFNOSUPPORT;
2152 			break;
2153 		}
2154 #endif /* INET */
2155 #ifndef INET6
2156 		if (pp->af == AF_INET6) {
2157 			error = EAFNOSUPPORT;
2158 			break;
2159 		}
2160 #endif /* INET6 */
2161 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2162 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2163 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2164 			error = EINVAL;
2165 			break;
2166 		}
2167 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2168 		bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2169 		if (pa->ifname[0])
2170 			kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2171 		PF_RULES_WLOCK();
2172 		if (pp->ticket != V_ticket_pabuf) {
2173 			PF_RULES_WUNLOCK();
2174 			if (pa->ifname[0])
2175 				free(kif, PFI_MTYPE);
2176 			free(pa, M_PFRULE);
2177 			error = EBUSY;
2178 			break;
2179 		}
2180 		if (pa->ifname[0]) {
2181 			pa->kif = pfi_kif_attach(kif, pa->ifname);
2182 			pfi_kif_ref(pa->kif);
2183 		} else
2184 			pa->kif = NULL;
2185 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2186 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2187 			if (pa->ifname[0])
2188 				pfi_kif_unref(pa->kif);
2189 			PF_RULES_WUNLOCK();
2190 			free(pa, M_PFRULE);
2191 			break;
2192 		}
2193 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2194 		PF_RULES_WUNLOCK();
2195 		break;
2196 	}
2197 
2198 	case DIOCGETADDRS: {
2199 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2200 		struct pf_pool		*pool;
2201 		struct pf_pooladdr	*pa;
2202 
2203 		PF_RULES_RLOCK();
2204 		pp->nr = 0;
2205 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2206 		    pp->r_num, 0, 1, 0);
2207 		if (pool == NULL) {
2208 			PF_RULES_RUNLOCK();
2209 			error = EBUSY;
2210 			break;
2211 		}
2212 		TAILQ_FOREACH(pa, &pool->list, entries)
2213 			pp->nr++;
2214 		PF_RULES_RUNLOCK();
2215 		break;
2216 	}
2217 
2218 	case DIOCGETADDR: {
2219 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2220 		struct pf_pool		*pool;
2221 		struct pf_pooladdr	*pa;
2222 		u_int32_t		 nr = 0;
2223 
2224 		PF_RULES_RLOCK();
2225 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2226 		    pp->r_num, 0, 1, 1);
2227 		if (pool == NULL) {
2228 			PF_RULES_RUNLOCK();
2229 			error = EBUSY;
2230 			break;
2231 		}
2232 		pa = TAILQ_FIRST(&pool->list);
2233 		while ((pa != NULL) && (nr < pp->nr)) {
2234 			pa = TAILQ_NEXT(pa, entries);
2235 			nr++;
2236 		}
2237 		if (pa == NULL) {
2238 			PF_RULES_RUNLOCK();
2239 			error = EBUSY;
2240 			break;
2241 		}
2242 		bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2243 		pf_addr_copyout(&pp->addr.addr);
2244 		PF_RULES_RUNLOCK();
2245 		break;
2246 	}
2247 
2248 	case DIOCCHANGEADDR: {
2249 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
2250 		struct pf_pool		*pool;
2251 		struct pf_pooladdr	*oldpa = NULL, *newpa = NULL;
2252 		struct pf_ruleset	*ruleset;
2253 		struct pfi_kif		*kif = NULL;
2254 
2255 		if (pca->action < PF_CHANGE_ADD_HEAD ||
2256 		    pca->action > PF_CHANGE_REMOVE) {
2257 			error = EINVAL;
2258 			break;
2259 		}
2260 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2261 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2262 		    pca->addr.addr.type != PF_ADDR_TABLE) {
2263 			error = EINVAL;
2264 			break;
2265 		}
2266 
2267 		if (pca->action != PF_CHANGE_REMOVE) {
2268 #ifndef INET
2269 			if (pca->af == AF_INET) {
2270 				error = EAFNOSUPPORT;
2271 				break;
2272 			}
2273 #endif /* INET */
2274 #ifndef INET6
2275 			if (pca->af == AF_INET6) {
2276 				error = EAFNOSUPPORT;
2277 				break;
2278 			}
2279 #endif /* INET6 */
2280 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
2281 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2282 			if (newpa->ifname[0])
2283 				kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2284 		}
2285 
2286 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGEADDR_error; }
2287 		PF_RULES_WLOCK();
2288 		ruleset = pf_find_ruleset(pca->anchor);
2289 		if (ruleset == NULL)
2290 			ERROUT(EBUSY);
2291 
2292 		pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2293 		    pca->r_num, pca->r_last, 1, 1);
2294 		if (pool == NULL)
2295 			ERROUT(EBUSY);
2296 
2297 		if (pca->action != PF_CHANGE_REMOVE) {
2298 			if (newpa->ifname[0]) {
2299 				newpa->kif = pfi_kif_attach(kif, newpa->ifname);
2300 				pfi_kif_ref(newpa->kif);
2301 			} else
2302 				newpa->kif = NULL;
2303 
2304 			switch (newpa->addr.type) {
2305 			case PF_ADDR_DYNIFTL:
2306 				error = pfi_dynaddr_setup(&newpa->addr,
2307 				    pca->af);
2308 				break;
2309 			case PF_ADDR_TABLE:
2310 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
2311 				    newpa->addr.v.tblname);
2312 				if (newpa->addr.p.tbl == NULL)
2313 					error = ENOMEM;
2314 				break;
2315 			}
2316 			if (error) {
2317 				if (newpa->kif)
2318 					pfi_kif_unref(newpa->kif);
2319 				PF_RULES_WUNLOCK();
2320 				free(newpa, M_PFRULE);
2321 				break;
2322 			}
2323 		}
2324 
2325 		if (pca->action == PF_CHANGE_ADD_HEAD)
2326 			oldpa = TAILQ_FIRST(&pool->list);
2327 		else if (pca->action == PF_CHANGE_ADD_TAIL)
2328 			oldpa = TAILQ_LAST(&pool->list, pf_palist);
2329 		else {
2330 			int	i = 0;
2331 
2332 			oldpa = TAILQ_FIRST(&pool->list);
2333 			while ((oldpa != NULL) && (i < pca->nr)) {
2334 				oldpa = TAILQ_NEXT(oldpa, entries);
2335 				i++;
2336 			}
2337 			if (oldpa == NULL) {
2338 				PF_RULES_WUNLOCK();
2339 				error = EINVAL;
2340 				break;
2341 			}
2342 		}
2343 
2344 		if (pca->action == PF_CHANGE_REMOVE) {
2345 			TAILQ_REMOVE(&pool->list, oldpa, entries);
2346 			switch (oldpa->addr.type) {
2347 			case PF_ADDR_DYNIFTL:
2348 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
2349 				break;
2350 			case PF_ADDR_TABLE:
2351 				pfr_detach_table(oldpa->addr.p.tbl);
2352 				break;
2353 			}
2354 			if (oldpa->kif)
2355 				pfi_kif_unref(oldpa->kif);
2356 			free(oldpa, M_PFRULE);
2357 		} else {
2358 			if (oldpa == NULL)
2359 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2360 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
2361 			    pca->action == PF_CHANGE_ADD_BEFORE)
2362 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2363 			else
2364 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
2365 				    newpa, entries);
2366 		}
2367 
2368 		pool->cur = TAILQ_FIRST(&pool->list);
2369 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr,
2370 		    pca->af);
2371 		PF_RULES_WUNLOCK();
2372 		break;
2373 
2374 #undef ERROUT
2375 DIOCCHANGEADDR_error:
2376 		PF_RULES_WUNLOCK();
2377 		if (newpa != NULL)
2378 			free(newpa, M_PFRULE);
2379 		if (kif != NULL)
2380 			free(kif, PFI_MTYPE);
2381 		break;
2382 	}
2383 
2384 	case DIOCGETRULESETS: {
2385 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2386 		struct pf_ruleset	*ruleset;
2387 		struct pf_anchor	*anchor;
2388 
2389 		PF_RULES_RLOCK();
2390 		pr->path[sizeof(pr->path) - 1] = 0;
2391 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2392 			PF_RULES_RUNLOCK();
2393 			error = ENOENT;
2394 			break;
2395 		}
2396 		pr->nr = 0;
2397 		if (ruleset->anchor == NULL) {
2398 			/* XXX kludge for pf_main_ruleset */
2399 			RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2400 				if (anchor->parent == NULL)
2401 					pr->nr++;
2402 		} else {
2403 			RB_FOREACH(anchor, pf_anchor_node,
2404 			    &ruleset->anchor->children)
2405 				pr->nr++;
2406 		}
2407 		PF_RULES_RUNLOCK();
2408 		break;
2409 	}
2410 
2411 	case DIOCGETRULESET: {
2412 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2413 		struct pf_ruleset	*ruleset;
2414 		struct pf_anchor	*anchor;
2415 		u_int32_t		 nr = 0;
2416 
2417 		PF_RULES_RLOCK();
2418 		pr->path[sizeof(pr->path) - 1] = 0;
2419 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2420 			PF_RULES_RUNLOCK();
2421 			error = ENOENT;
2422 			break;
2423 		}
2424 		pr->name[0] = 0;
2425 		if (ruleset->anchor == NULL) {
2426 			/* XXX kludge for pf_main_ruleset */
2427 			RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2428 				if (anchor->parent == NULL && nr++ == pr->nr) {
2429 					strlcpy(pr->name, anchor->name,
2430 					    sizeof(pr->name));
2431 					break;
2432 				}
2433 		} else {
2434 			RB_FOREACH(anchor, pf_anchor_node,
2435 			    &ruleset->anchor->children)
2436 				if (nr++ == pr->nr) {
2437 					strlcpy(pr->name, anchor->name,
2438 					    sizeof(pr->name));
2439 					break;
2440 				}
2441 		}
2442 		if (!pr->name[0])
2443 			error = EBUSY;
2444 		PF_RULES_RUNLOCK();
2445 		break;
2446 	}
2447 
2448 	case DIOCRCLRTABLES: {
2449 		struct pfioc_table *io = (struct pfioc_table *)addr;
2450 
2451 		if (io->pfrio_esize != 0) {
2452 			error = ENODEV;
2453 			break;
2454 		}
2455 		PF_RULES_WLOCK();
2456 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2457 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2458 		PF_RULES_WUNLOCK();
2459 		break;
2460 	}
2461 
2462 	case DIOCRADDTABLES: {
2463 		struct pfioc_table *io = (struct pfioc_table *)addr;
2464 		struct pfr_table *pfrts;
2465 		size_t totlen;
2466 
2467 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2468 			error = ENODEV;
2469 			break;
2470 		}
2471 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2472 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2473 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2474 		if (error) {
2475 			free(pfrts, M_TEMP);
2476 			break;
2477 		}
2478 		PF_RULES_WLOCK();
2479 		error = pfr_add_tables(pfrts, io->pfrio_size,
2480 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2481 		PF_RULES_WUNLOCK();
2482 		free(pfrts, M_TEMP);
2483 		break;
2484 	}
2485 
2486 	case DIOCRDELTABLES: {
2487 		struct pfioc_table *io = (struct pfioc_table *)addr;
2488 		struct pfr_table *pfrts;
2489 		size_t totlen;
2490 
2491 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2492 			error = ENODEV;
2493 			break;
2494 		}
2495 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2496 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2497 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2498 		if (error) {
2499 			free(pfrts, M_TEMP);
2500 			break;
2501 		}
2502 		PF_RULES_WLOCK();
2503 		error = pfr_del_tables(pfrts, io->pfrio_size,
2504 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2505 		PF_RULES_WUNLOCK();
2506 		free(pfrts, M_TEMP);
2507 		break;
2508 	}
2509 
2510 	case DIOCRGETTABLES: {
2511 		struct pfioc_table *io = (struct pfioc_table *)addr;
2512 		struct pfr_table *pfrts;
2513 		size_t totlen;
2514 
2515 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2516 			error = ENODEV;
2517 			break;
2518 		}
2519 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2520 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2521 		PF_RULES_RLOCK();
2522 		error = pfr_get_tables(&io->pfrio_table, pfrts,
2523 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2524 		PF_RULES_RUNLOCK();
2525 		if (error == 0)
2526 			error = copyout(pfrts, io->pfrio_buffer, totlen);
2527 		free(pfrts, M_TEMP);
2528 		break;
2529 	}
2530 
2531 	case DIOCRGETTSTATS: {
2532 		struct pfioc_table *io = (struct pfioc_table *)addr;
2533 		struct pfr_tstats *pfrtstats;
2534 		size_t totlen;
2535 
2536 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2537 			error = ENODEV;
2538 			break;
2539 		}
2540 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
2541 		pfrtstats = malloc(totlen, M_TEMP, M_WAITOK);
2542 		PF_RULES_WLOCK();
2543 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
2544 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2545 		PF_RULES_WUNLOCK();
2546 		if (error == 0)
2547 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
2548 		free(pfrtstats, M_TEMP);
2549 		break;
2550 	}
2551 
2552 	case DIOCRCLRTSTATS: {
2553 		struct pfioc_table *io = (struct pfioc_table *)addr;
2554 		struct pfr_table *pfrts;
2555 		size_t totlen;
2556 
2557 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2558 			error = ENODEV;
2559 			break;
2560 		}
2561 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2562 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2563 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2564 		if (error) {
2565 			free(pfrts, M_TEMP);
2566 			break;
2567 		}
2568 		PF_RULES_WLOCK();
2569 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
2570 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2571 		PF_RULES_WUNLOCK();
2572 		free(pfrts, M_TEMP);
2573 		break;
2574 	}
2575 
2576 	case DIOCRSETTFLAGS: {
2577 		struct pfioc_table *io = (struct pfioc_table *)addr;
2578 		struct pfr_table *pfrts;
2579 		size_t totlen;
2580 
2581 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2582 			error = ENODEV;
2583 			break;
2584 		}
2585 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2586 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2587 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2588 		if (error) {
2589 			free(pfrts, M_TEMP);
2590 			break;
2591 		}
2592 		PF_RULES_WLOCK();
2593 		error = pfr_set_tflags(pfrts, io->pfrio_size,
2594 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2595 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2596 		PF_RULES_WUNLOCK();
2597 		free(pfrts, M_TEMP);
2598 		break;
2599 	}
2600 
2601 	case DIOCRCLRADDRS: {
2602 		struct pfioc_table *io = (struct pfioc_table *)addr;
2603 
2604 		if (io->pfrio_esize != 0) {
2605 			error = ENODEV;
2606 			break;
2607 		}
2608 		PF_RULES_WLOCK();
2609 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2610 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2611 		PF_RULES_WUNLOCK();
2612 		break;
2613 	}
2614 
2615 	case DIOCRADDADDRS: {
2616 		struct pfioc_table *io = (struct pfioc_table *)addr;
2617 		struct pfr_addr *pfras;
2618 		size_t totlen;
2619 
2620 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2621 			error = ENODEV;
2622 			break;
2623 		}
2624 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2625 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2626 		error = copyin(io->pfrio_buffer, pfras, totlen);
2627 		if (error) {
2628 			free(pfras, M_TEMP);
2629 			break;
2630 		}
2631 		PF_RULES_WLOCK();
2632 		error = pfr_add_addrs(&io->pfrio_table, pfras,
2633 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2634 		    PFR_FLAG_USERIOCTL);
2635 		PF_RULES_WUNLOCK();
2636 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2637 			error = copyout(pfras, io->pfrio_buffer, totlen);
2638 		free(pfras, M_TEMP);
2639 		break;
2640 	}
2641 
2642 	case DIOCRDELADDRS: {
2643 		struct pfioc_table *io = (struct pfioc_table *)addr;
2644 		struct pfr_addr *pfras;
2645 		size_t totlen;
2646 
2647 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2648 			error = ENODEV;
2649 			break;
2650 		}
2651 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2652 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2653 		error = copyin(io->pfrio_buffer, pfras, totlen);
2654 		if (error) {
2655 			free(pfras, M_TEMP);
2656 			break;
2657 		}
2658 		PF_RULES_WLOCK();
2659 		error = pfr_del_addrs(&io->pfrio_table, pfras,
2660 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2661 		    PFR_FLAG_USERIOCTL);
2662 		PF_RULES_WUNLOCK();
2663 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2664 			error = copyout(pfras, io->pfrio_buffer, totlen);
2665 		free(pfras, M_TEMP);
2666 		break;
2667 	}
2668 
2669 	case DIOCRSETADDRS: {
2670 		struct pfioc_table *io = (struct pfioc_table *)addr;
2671 		struct pfr_addr *pfras;
2672 		size_t totlen;
2673 
2674 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2675 			error = ENODEV;
2676 			break;
2677 		}
2678 		totlen = (io->pfrio_size + io->pfrio_size2) *
2679 		    sizeof(struct pfr_addr);
2680 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2681 		error = copyin(io->pfrio_buffer, pfras, totlen);
2682 		if (error) {
2683 			free(pfras, M_TEMP);
2684 			break;
2685 		}
2686 		PF_RULES_WLOCK();
2687 		error = pfr_set_addrs(&io->pfrio_table, pfras,
2688 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2689 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2690 		    PFR_FLAG_USERIOCTL, 0);
2691 		PF_RULES_WUNLOCK();
2692 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2693 			error = copyout(pfras, io->pfrio_buffer, totlen);
2694 		free(pfras, M_TEMP);
2695 		break;
2696 	}
2697 
2698 	case DIOCRGETADDRS: {
2699 		struct pfioc_table *io = (struct pfioc_table *)addr;
2700 		struct pfr_addr *pfras;
2701 		size_t totlen;
2702 
2703 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2704 			error = ENODEV;
2705 			break;
2706 		}
2707 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2708 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2709 		PF_RULES_RLOCK();
2710 		error = pfr_get_addrs(&io->pfrio_table, pfras,
2711 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2712 		PF_RULES_RUNLOCK();
2713 		if (error == 0)
2714 			error = copyout(pfras, io->pfrio_buffer, totlen);
2715 		free(pfras, M_TEMP);
2716 		break;
2717 	}
2718 
2719 	case DIOCRGETASTATS: {
2720 		struct pfioc_table *io = (struct pfioc_table *)addr;
2721 		struct pfr_astats *pfrastats;
2722 		size_t totlen;
2723 
2724 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2725 			error = ENODEV;
2726 			break;
2727 		}
2728 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
2729 		pfrastats = malloc(totlen, M_TEMP, M_WAITOK);
2730 		PF_RULES_RLOCK();
2731 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
2732 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2733 		PF_RULES_RUNLOCK();
2734 		if (error == 0)
2735 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
2736 		free(pfrastats, M_TEMP);
2737 		break;
2738 	}
2739 
2740 	case DIOCRCLRASTATS: {
2741 		struct pfioc_table *io = (struct pfioc_table *)addr;
2742 		struct pfr_addr *pfras;
2743 		size_t totlen;
2744 
2745 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2746 			error = ENODEV;
2747 			break;
2748 		}
2749 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2750 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2751 		error = copyin(io->pfrio_buffer, pfras, totlen);
2752 		if (error) {
2753 			free(pfras, M_TEMP);
2754 			break;
2755 		}
2756 		PF_RULES_WLOCK();
2757 		error = pfr_clr_astats(&io->pfrio_table, pfras,
2758 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2759 		    PFR_FLAG_USERIOCTL);
2760 		PF_RULES_WUNLOCK();
2761 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2762 			error = copyout(pfras, io->pfrio_buffer, totlen);
2763 		free(pfras, M_TEMP);
2764 		break;
2765 	}
2766 
2767 	case DIOCRTSTADDRS: {
2768 		struct pfioc_table *io = (struct pfioc_table *)addr;
2769 		struct pfr_addr *pfras;
2770 		size_t totlen;
2771 
2772 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2773 			error = ENODEV;
2774 			break;
2775 		}
2776 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2777 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2778 		error = copyin(io->pfrio_buffer, pfras, totlen);
2779 		if (error) {
2780 			free(pfras, M_TEMP);
2781 			break;
2782 		}
2783 		PF_RULES_RLOCK();
2784 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
2785 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2786 		    PFR_FLAG_USERIOCTL);
2787 		PF_RULES_RUNLOCK();
2788 		if (error == 0)
2789 			error = copyout(pfras, io->pfrio_buffer, totlen);
2790 		free(pfras, M_TEMP);
2791 		break;
2792 	}
2793 
2794 	case DIOCRINADEFINE: {
2795 		struct pfioc_table *io = (struct pfioc_table *)addr;
2796 		struct pfr_addr *pfras;
2797 		size_t totlen;
2798 
2799 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2800 			error = ENODEV;
2801 			break;
2802 		}
2803 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2804 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2805 		error = copyin(io->pfrio_buffer, pfras, totlen);
2806 		if (error) {
2807 			free(pfras, M_TEMP);
2808 			break;
2809 		}
2810 		PF_RULES_WLOCK();
2811 		error = pfr_ina_define(&io->pfrio_table, pfras,
2812 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2813 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2814 		PF_RULES_WUNLOCK();
2815 		free(pfras, M_TEMP);
2816 		break;
2817 	}
2818 
2819 	case DIOCOSFPADD: {
2820 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2821 		PF_RULES_WLOCK();
2822 		error = pf_osfp_add(io);
2823 		PF_RULES_WUNLOCK();
2824 		break;
2825 	}
2826 
2827 	case DIOCOSFPGET: {
2828 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2829 		PF_RULES_RLOCK();
2830 		error = pf_osfp_get(io);
2831 		PF_RULES_RUNLOCK();
2832 		break;
2833 	}
2834 
2835 	case DIOCXBEGIN: {
2836 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2837 		struct pfioc_trans_e	*ioes, *ioe;
2838 		size_t			 totlen;
2839 		int			 i;
2840 
2841 		if (io->esize != sizeof(*ioe)) {
2842 			error = ENODEV;
2843 			break;
2844 		}
2845 		totlen = sizeof(struct pfioc_trans_e) * io->size;
2846 		ioes = malloc(totlen, M_TEMP, M_WAITOK);
2847 		error = copyin(io->array, ioes, totlen);
2848 		if (error) {
2849 			free(ioes, M_TEMP);
2850 			break;
2851 		}
2852 		PF_RULES_WLOCK();
2853 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2854 			switch (ioe->rs_num) {
2855 #ifdef ALTQ
2856 			case PF_RULESET_ALTQ:
2857 				if (ioe->anchor[0]) {
2858 					PF_RULES_WUNLOCK();
2859 					free(ioes, M_TEMP);
2860 					error = EINVAL;
2861 					goto fail;
2862 				}
2863 				if ((error = pf_begin_altq(&ioe->ticket))) {
2864 					PF_RULES_WUNLOCK();
2865 					free(ioes, M_TEMP);
2866 					goto fail;
2867 				}
2868 				break;
2869 #endif /* ALTQ */
2870 			case PF_RULESET_TABLE:
2871 			    {
2872 				struct pfr_table table;
2873 
2874 				bzero(&table, sizeof(table));
2875 				strlcpy(table.pfrt_anchor, ioe->anchor,
2876 				    sizeof(table.pfrt_anchor));
2877 				if ((error = pfr_ina_begin(&table,
2878 				    &ioe->ticket, NULL, 0))) {
2879 					PF_RULES_WUNLOCK();
2880 					free(ioes, M_TEMP);
2881 					goto fail;
2882 				}
2883 				break;
2884 			    }
2885 			default:
2886 				if ((error = pf_begin_rules(&ioe->ticket,
2887 				    ioe->rs_num, ioe->anchor))) {
2888 					PF_RULES_WUNLOCK();
2889 					free(ioes, M_TEMP);
2890 					goto fail;
2891 				}
2892 				break;
2893 			}
2894 		}
2895 		PF_RULES_WUNLOCK();
2896 		error = copyout(ioes, io->array, totlen);
2897 		free(ioes, M_TEMP);
2898 		break;
2899 	}
2900 
2901 	case DIOCXROLLBACK: {
2902 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2903 		struct pfioc_trans_e	*ioe, *ioes;
2904 		size_t			 totlen;
2905 		int			 i;
2906 
2907 		if (io->esize != sizeof(*ioe)) {
2908 			error = ENODEV;
2909 			break;
2910 		}
2911 		totlen = sizeof(struct pfioc_trans_e) * io->size;
2912 		ioes = malloc(totlen, M_TEMP, M_WAITOK);
2913 		error = copyin(io->array, ioes, totlen);
2914 		if (error) {
2915 			free(ioes, M_TEMP);
2916 			break;
2917 		}
2918 		PF_RULES_WLOCK();
2919 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2920 			switch (ioe->rs_num) {
2921 #ifdef ALTQ
2922 			case PF_RULESET_ALTQ:
2923 				if (ioe->anchor[0]) {
2924 					PF_RULES_WUNLOCK();
2925 					free(ioes, M_TEMP);
2926 					error = EINVAL;
2927 					goto fail;
2928 				}
2929 				if ((error = pf_rollback_altq(ioe->ticket))) {
2930 					PF_RULES_WUNLOCK();
2931 					free(ioes, M_TEMP);
2932 					goto fail; /* really bad */
2933 				}
2934 				break;
2935 #endif /* ALTQ */
2936 			case PF_RULESET_TABLE:
2937 			    {
2938 				struct pfr_table table;
2939 
2940 				bzero(&table, sizeof(table));
2941 				strlcpy(table.pfrt_anchor, ioe->anchor,
2942 				    sizeof(table.pfrt_anchor));
2943 				if ((error = pfr_ina_rollback(&table,
2944 				    ioe->ticket, NULL, 0))) {
2945 					PF_RULES_WUNLOCK();
2946 					free(ioes, M_TEMP);
2947 					goto fail; /* really bad */
2948 				}
2949 				break;
2950 			    }
2951 			default:
2952 				if ((error = pf_rollback_rules(ioe->ticket,
2953 				    ioe->rs_num, ioe->anchor))) {
2954 					PF_RULES_WUNLOCK();
2955 					free(ioes, M_TEMP);
2956 					goto fail; /* really bad */
2957 				}
2958 				break;
2959 			}
2960 		}
2961 		PF_RULES_WUNLOCK();
2962 		free(ioes, M_TEMP);
2963 		break;
2964 	}
2965 
2966 	case DIOCXCOMMIT: {
2967 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2968 		struct pfioc_trans_e	*ioe, *ioes;
2969 		struct pf_ruleset	*rs;
2970 		size_t			 totlen;
2971 		int			 i;
2972 
2973 		if (io->esize != sizeof(*ioe)) {
2974 			error = ENODEV;
2975 			break;
2976 		}
2977 		totlen = sizeof(struct pfioc_trans_e) * io->size;
2978 		ioes = malloc(totlen, M_TEMP, M_WAITOK);
2979 		error = copyin(io->array, ioes, totlen);
2980 		if (error) {
2981 			free(ioes, M_TEMP);
2982 			break;
2983 		}
2984 		PF_RULES_WLOCK();
2985 		/* First makes sure everything will succeed. */
2986 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2987 			switch (ioe->rs_num) {
2988 #ifdef ALTQ
2989 			case PF_RULESET_ALTQ:
2990 				if (ioe->anchor[0]) {
2991 					PF_RULES_WUNLOCK();
2992 					free(ioes, M_TEMP);
2993 					error = EINVAL;
2994 					goto fail;
2995 				}
2996 				if (!V_altqs_inactive_open || ioe->ticket !=
2997 				    V_ticket_altqs_inactive) {
2998 					PF_RULES_WUNLOCK();
2999 					free(ioes, M_TEMP);
3000 					error = EBUSY;
3001 					goto fail;
3002 				}
3003 				break;
3004 #endif /* ALTQ */
3005 			case PF_RULESET_TABLE:
3006 				rs = pf_find_ruleset(ioe->anchor);
3007 				if (rs == NULL || !rs->topen || ioe->ticket !=
3008 				    rs->tticket) {
3009 					PF_RULES_WUNLOCK();
3010 					free(ioes, M_TEMP);
3011 					error = EBUSY;
3012 					goto fail;
3013 				}
3014 				break;
3015 			default:
3016 				if (ioe->rs_num < 0 || ioe->rs_num >=
3017 				    PF_RULESET_MAX) {
3018 					PF_RULES_WUNLOCK();
3019 					free(ioes, M_TEMP);
3020 					error = EINVAL;
3021 					goto fail;
3022 				}
3023 				rs = pf_find_ruleset(ioe->anchor);
3024 				if (rs == NULL ||
3025 				    !rs->rules[ioe->rs_num].inactive.open ||
3026 				    rs->rules[ioe->rs_num].inactive.ticket !=
3027 				    ioe->ticket) {
3028 					PF_RULES_WUNLOCK();
3029 					free(ioes, M_TEMP);
3030 					error = EBUSY;
3031 					goto fail;
3032 				}
3033 				break;
3034 			}
3035 		}
3036 		/* Now do the commit - no errors should happen here. */
3037 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3038 			switch (ioe->rs_num) {
3039 #ifdef ALTQ
3040 			case PF_RULESET_ALTQ:
3041 				if ((error = pf_commit_altq(ioe->ticket))) {
3042 					PF_RULES_WUNLOCK();
3043 					free(ioes, M_TEMP);
3044 					goto fail; /* really bad */
3045 				}
3046 				break;
3047 #endif /* ALTQ */
3048 			case PF_RULESET_TABLE:
3049 			    {
3050 				struct pfr_table table;
3051 
3052 				bzero(&table, sizeof(table));
3053 				strlcpy(table.pfrt_anchor, ioe->anchor,
3054 				    sizeof(table.pfrt_anchor));
3055 				if ((error = pfr_ina_commit(&table,
3056 				    ioe->ticket, NULL, NULL, 0))) {
3057 					PF_RULES_WUNLOCK();
3058 					free(ioes, M_TEMP);
3059 					goto fail; /* really bad */
3060 				}
3061 				break;
3062 			    }
3063 			default:
3064 				if ((error = pf_commit_rules(ioe->ticket,
3065 				    ioe->rs_num, ioe->anchor))) {
3066 					PF_RULES_WUNLOCK();
3067 					free(ioes, M_TEMP);
3068 					goto fail; /* really bad */
3069 				}
3070 				break;
3071 			}
3072 		}
3073 		PF_RULES_WUNLOCK();
3074 		free(ioes, M_TEMP);
3075 		break;
3076 	}
3077 
3078 	case DIOCGETSRCNODES: {
3079 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
3080 		struct pf_srchash	*sh;
3081 		struct pf_src_node	*n, *p, *pstore;
3082 		uint32_t		 i, nr = 0;
3083 
3084 		if (psn->psn_len == 0) {
3085 			for (i = 0, sh = V_pf_srchash; i < V_pf_srchashmask;
3086 			    i++, sh++) {
3087 				PF_HASHROW_LOCK(sh);
3088 				LIST_FOREACH(n, &sh->nodes, entry)
3089 					nr++;
3090 				PF_HASHROW_UNLOCK(sh);
3091 			}
3092 			psn->psn_len = sizeof(struct pf_src_node) * nr;
3093 			break;
3094 		}
3095 
3096 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK);
3097 		for (i = 0, sh = V_pf_srchash; i < V_pf_srchashmask;
3098 		    i++, sh++) {
3099 		    PF_HASHROW_LOCK(sh);
3100 		    LIST_FOREACH(n, &sh->nodes, entry) {
3101 			int	secs = time_uptime, diff;
3102 
3103 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
3104 				break;
3105 
3106 			bcopy(n, p, sizeof(struct pf_src_node));
3107 			if (n->rule.ptr != NULL)
3108 				p->rule.nr = n->rule.ptr->nr;
3109 			p->creation = secs - p->creation;
3110 			if (p->expire > secs)
3111 				p->expire -= secs;
3112 			else
3113 				p->expire = 0;
3114 
3115 			/* Adjust the connection rate estimate. */
3116 			diff = secs - n->conn_rate.last;
3117 			if (diff >= n->conn_rate.seconds)
3118 				p->conn_rate.count = 0;
3119 			else
3120 				p->conn_rate.count -=
3121 				    n->conn_rate.count * diff /
3122 				    n->conn_rate.seconds;
3123 			p++;
3124 			nr++;
3125 		    }
3126 		    PF_HASHROW_UNLOCK(sh);
3127 		}
3128 		error = copyout(pstore, psn->psn_src_nodes,
3129 		    sizeof(struct pf_src_node) * nr);
3130 		if (error) {
3131 			free(pstore, M_TEMP);
3132 			break;
3133 		}
3134 		psn->psn_len = sizeof(struct pf_src_node) * nr;
3135 		free(pstore, M_TEMP);
3136 		break;
3137 	}
3138 
3139 	case DIOCCLRSRCNODES: {
3140 
3141 		pf_clear_srcnodes(NULL);
3142 		pf_purge_expired_src_nodes();
3143 		V_pf_status.src_nodes = 0;
3144 		break;
3145 	}
3146 
3147 	case DIOCKILLSRCNODES: {
3148 		struct pfioc_src_node_kill *psnk =
3149 		    (struct pfioc_src_node_kill *)addr;
3150 		struct pf_srchash	*sh;
3151 		struct pf_src_node	*sn;
3152 		u_int			i, killed = 0;
3153 
3154 		for (i = 0, sh = V_pf_srchash; i < V_pf_srchashmask;
3155 		    i++, sh++) {
3156 		    /*
3157 		     * XXXGL: we don't ever acquire sources hash lock
3158 		     * but if we ever do, the below call to pf_clear_srcnodes()
3159 		     * would lead to a LOR.
3160 		     */
3161 		    PF_HASHROW_LOCK(sh);
3162 		    LIST_FOREACH(sn, &sh->nodes, entry)
3163 			if (PF_MATCHA(psnk->psnk_src.neg,
3164 				&psnk->psnk_src.addr.v.a.addr,
3165 				&psnk->psnk_src.addr.v.a.mask,
3166 				&sn->addr, sn->af) &&
3167 			    PF_MATCHA(psnk->psnk_dst.neg,
3168 				&psnk->psnk_dst.addr.v.a.addr,
3169 				&psnk->psnk_dst.addr.v.a.mask,
3170 				&sn->raddr, sn->af)) {
3171 				/* Handle state to src_node linkage */
3172 				if (sn->states != 0)
3173 					pf_clear_srcnodes(sn);
3174 				sn->expire = 1;
3175 				killed++;
3176 			}
3177 		    PF_HASHROW_UNLOCK(sh);
3178 		}
3179 
3180 		if (killed > 0)
3181 			pf_purge_expired_src_nodes();
3182 
3183 		psnk->psnk_killed = killed;
3184 		break;
3185 	}
3186 
3187 	case DIOCSETHOSTID: {
3188 		u_int32_t	*hostid = (u_int32_t *)addr;
3189 
3190 		PF_RULES_WLOCK();
3191 		if (*hostid == 0)
3192 			V_pf_status.hostid = arc4random();
3193 		else
3194 			V_pf_status.hostid = *hostid;
3195 		PF_RULES_WUNLOCK();
3196 		break;
3197 	}
3198 
3199 	case DIOCOSFPFLUSH:
3200 		PF_RULES_WLOCK();
3201 		pf_osfp_flush();
3202 		PF_RULES_WUNLOCK();
3203 		break;
3204 
3205 	case DIOCIGETIFACES: {
3206 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3207 		struct pfi_kif *ifstore;
3208 		size_t bufsiz;
3209 
3210 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
3211 			error = ENODEV;
3212 			break;
3213 		}
3214 
3215 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
3216 		ifstore = malloc(bufsiz, M_TEMP, M_WAITOK);
3217 		PF_RULES_RLOCK();
3218 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
3219 		PF_RULES_RUNLOCK();
3220 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
3221 		free(ifstore, M_TEMP);
3222 		break;
3223 	}
3224 
3225 	case DIOCSETIFFLAG: {
3226 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3227 
3228 		PF_RULES_WLOCK();
3229 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3230 		PF_RULES_WUNLOCK();
3231 		break;
3232 	}
3233 
3234 	case DIOCCLRIFFLAG: {
3235 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3236 
3237 		PF_RULES_WLOCK();
3238 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3239 		PF_RULES_WUNLOCK();
3240 		break;
3241 	}
3242 
3243 	default:
3244 		error = ENODEV;
3245 		break;
3246 	}
3247 fail:
3248 	CURVNET_RESTORE();
3249 
3250 	return (error);
3251 }
3252 
3253 void
3254 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
3255 {
3256 	bzero(sp, sizeof(struct pfsync_state));
3257 
3258 	/* copy from state key */
3259 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
3260 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
3261 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
3262 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
3263 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
3264 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
3265 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
3266 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
3267 	sp->proto = st->key[PF_SK_WIRE]->proto;
3268 	sp->af = st->key[PF_SK_WIRE]->af;
3269 
3270 	/* copy from state */
3271 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
3272 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
3273 	sp->creation = htonl(time_uptime - st->creation);
3274 	sp->expire = pf_state_expires(st);
3275 	if (sp->expire <= time_uptime)
3276 		sp->expire = htonl(0);
3277 	else
3278 		sp->expire = htonl(sp->expire - time_uptime);
3279 
3280 	sp->direction = st->direction;
3281 	sp->log = st->log;
3282 	sp->timeout = st->timeout;
3283 	sp->state_flags = st->state_flags;
3284 	if (st->src_node)
3285 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
3286 	if (st->nat_src_node)
3287 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
3288 
3289 	sp->id = st->id;
3290 	sp->creatorid = st->creatorid;
3291 	pf_state_peer_hton(&st->src, &sp->src);
3292 	pf_state_peer_hton(&st->dst, &sp->dst);
3293 
3294 	if (st->rule.ptr == NULL)
3295 		sp->rule = htonl(-1);
3296 	else
3297 		sp->rule = htonl(st->rule.ptr->nr);
3298 	if (st->anchor.ptr == NULL)
3299 		sp->anchor = htonl(-1);
3300 	else
3301 		sp->anchor = htonl(st->anchor.ptr->nr);
3302 	if (st->nat_rule.ptr == NULL)
3303 		sp->nat_rule = htonl(-1);
3304 	else
3305 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
3306 
3307 	pf_state_counter_hton(st->packets[0], sp->packets[0]);
3308 	pf_state_counter_hton(st->packets[1], sp->packets[1]);
3309 	pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
3310 	pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
3311 
3312 }
3313 
3314 static void
3315 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
3316 {
3317 	struct pfr_ktable *kt;
3318 
3319 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
3320 
3321 	kt = aw->p.tbl;
3322 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
3323 		kt = kt->pfrkt_root;
3324 	aw->p.tbl = NULL;
3325 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
3326 		kt->pfrkt_cnt : -1;
3327 }
3328 
3329 /*
3330  * XXX - Check for version missmatch!!!
3331  */
3332 static void
3333 pf_clear_states(void)
3334 {
3335 	struct pf_state	*s;
3336 	u_int i;
3337 
3338 	for (i = 0; i <= V_pf_hashmask; i++) {
3339 		struct pf_idhash *ih = &V_pf_idhash[i];
3340 relock:
3341 		PF_HASHROW_LOCK(ih);
3342 		LIST_FOREACH(s, &ih->states, entry) {
3343 			s->timeout = PFTM_PURGE;
3344 			/* Don't send out individual delete messages. */
3345 			s->sync_state = PFSTATE_NOSYNC;
3346 			pf_unlink_state(s, PF_ENTER_LOCKED);
3347 			goto relock;
3348 		}
3349 		PF_HASHROW_UNLOCK(ih);
3350 	}
3351 }
3352 
3353 static int
3354 pf_clear_tables(void)
3355 {
3356 	struct pfioc_table io;
3357 	int error;
3358 
3359 	bzero(&io, sizeof(io));
3360 
3361 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
3362 	    io.pfrio_flags);
3363 
3364 	return (error);
3365 }
3366 
3367 static void
3368 pf_clear_srcnodes(struct pf_src_node *n)
3369 {
3370 	struct pf_state *s;
3371 	int i;
3372 
3373 	for (i = 0; i <= V_pf_hashmask; i++) {
3374 		struct pf_idhash *ih = &V_pf_idhash[i];
3375 
3376 		PF_HASHROW_LOCK(ih);
3377 		LIST_FOREACH(s, &ih->states, entry) {
3378 			if (n == NULL || n == s->src_node)
3379 				s->src_node = NULL;
3380 			if (n == NULL || n == s->nat_src_node)
3381 				s->nat_src_node = NULL;
3382 		}
3383 		PF_HASHROW_UNLOCK(ih);
3384 	}
3385 
3386 	if (n == NULL) {
3387 		struct pf_srchash *sh;
3388 
3389 		for (i = 0, sh = V_pf_srchash; i < V_pf_srchashmask;
3390 		    i++, sh++) {
3391 			PF_HASHROW_LOCK(sh);
3392 			LIST_FOREACH(n, &sh->nodes, entry) {
3393 				n->expire = 1;
3394 				n->states = 0;
3395 			}
3396 			PF_HASHROW_UNLOCK(sh);
3397 		}
3398 	} else {
3399 		/* XXX: hash slot should already be locked here. */
3400 		n->expire = 1;
3401 		n->states = 0;
3402 	}
3403 }
3404 /*
3405  * XXX - Check for version missmatch!!!
3406  */
3407 
3408 /*
3409  * Duplicate pfctl -Fa operation to get rid of as much as we can.
3410  */
3411 static int
3412 shutdown_pf(void)
3413 {
3414 	int error = 0;
3415 	u_int32_t t[5];
3416 	char nn = '\0';
3417 
3418 	V_pf_status.running = 0;
3419 	do {
3420 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
3421 		    != 0) {
3422 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
3423 			break;
3424 		}
3425 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
3426 		    != 0) {
3427 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
3428 			break;		/* XXX: rollback? */
3429 		}
3430 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
3431 		    != 0) {
3432 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
3433 			break;		/* XXX: rollback? */
3434 		}
3435 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
3436 		    != 0) {
3437 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
3438 			break;		/* XXX: rollback? */
3439 		}
3440 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
3441 		    != 0) {
3442 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
3443 			break;		/* XXX: rollback? */
3444 		}
3445 
3446 		/* XXX: these should always succeed here */
3447 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
3448 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
3449 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
3450 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
3451 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
3452 
3453 		if ((error = pf_clear_tables()) != 0)
3454 			break;
3455 
3456 #ifdef ALTQ
3457 		if ((error = pf_begin_altq(&t[0])) != 0) {
3458 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
3459 			break;
3460 		}
3461 		pf_commit_altq(t[0]);
3462 #endif
3463 
3464 		pf_clear_states();
3465 
3466 		pf_clear_srcnodes(NULL);
3467 
3468 		/* status does not use malloced mem so no need to cleanup */
3469 		/* fingerprints and interfaces have thier own cleanup code */
3470 	} while(0);
3471 
3472 	return (error);
3473 }
3474 
3475 #ifdef INET
3476 static int
3477 pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3478     struct inpcb *inp)
3479 {
3480 	int chk;
3481 
3482 	chk = pf_test(PF_IN, ifp, m, inp);
3483 	if (chk && *m) {
3484 		m_freem(*m);
3485 		*m = NULL;
3486 	}
3487 
3488 	return (chk);
3489 }
3490 
3491 static int
3492 pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3493     struct inpcb *inp)
3494 {
3495 	int chk;
3496 
3497 	/* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3498 	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3499 		in_delayed_cksum(*m);
3500 		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3501 	}
3502 
3503 	chk = pf_test(PF_OUT, ifp, m, inp);
3504 	if (chk && *m) {
3505 		m_freem(*m);
3506 		*m = NULL;
3507 	}
3508 
3509 	return (chk);
3510 }
3511 #endif
3512 
3513 #ifdef INET6
3514 static int
3515 pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3516     struct inpcb *inp)
3517 {
3518 	int chk;
3519 
3520 	/*
3521 	 * In case of loopback traffic IPv6 uses the real interface in
3522 	 * order to support scoped addresses. In order to support stateful
3523 	 * filtering we have change this to lo0 as it is the case in IPv4.
3524 	 */
3525 	CURVNET_SET(ifp->if_vnet);
3526 	chk = pf_test6(PF_IN, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
3527 	CURVNET_RESTORE();
3528 	if (chk && *m) {
3529 		m_freem(*m);
3530 		*m = NULL;
3531 	}
3532 	return chk;
3533 }
3534 
3535 static int
3536 pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3537     struct inpcb *inp)
3538 {
3539 	int chk;
3540 
3541 	/* We need a proper CSUM before we start (s. OpenBSD ip_output) */
3542 	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3543 #ifdef INET
3544 		/* XXX-BZ copy&paste error from r126261? */
3545 		in_delayed_cksum(*m);
3546 #endif
3547 		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3548 	}
3549 	CURVNET_SET(ifp->if_vnet);
3550 	chk = pf_test6(PF_OUT, ifp, m, inp);
3551 	CURVNET_RESTORE();
3552 	if (chk && *m) {
3553 		m_freem(*m);
3554 		*m = NULL;
3555 	}
3556 	return chk;
3557 }
3558 #endif /* INET6 */
3559 
3560 static int
3561 hook_pf(void)
3562 {
3563 #ifdef INET
3564 	struct pfil_head *pfh_inet;
3565 #endif
3566 #ifdef INET6
3567 	struct pfil_head *pfh_inet6;
3568 #endif
3569 
3570 	if (V_pf_pfil_hooked)
3571 		return (0);
3572 
3573 #ifdef INET
3574 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3575 	if (pfh_inet == NULL)
3576 		return (ESRCH); /* XXX */
3577 	pfil_add_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet);
3578 	pfil_add_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet);
3579 #endif
3580 #ifdef INET6
3581 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3582 	if (pfh_inet6 == NULL) {
3583 #ifdef INET
3584 		pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3585 		    pfh_inet);
3586 		pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3587 		    pfh_inet);
3588 #endif
3589 		return (ESRCH); /* XXX */
3590 	}
3591 	pfil_add_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet6);
3592 	pfil_add_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet6);
3593 #endif
3594 
3595 	V_pf_pfil_hooked = 1;
3596 	return (0);
3597 }
3598 
3599 static int
3600 dehook_pf(void)
3601 {
3602 #ifdef INET
3603 	struct pfil_head *pfh_inet;
3604 #endif
3605 #ifdef INET6
3606 	struct pfil_head *pfh_inet6;
3607 #endif
3608 
3609 	if (V_pf_pfil_hooked == 0)
3610 		return (0);
3611 
3612 #ifdef INET
3613 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3614 	if (pfh_inet == NULL)
3615 		return (ESRCH); /* XXX */
3616 	pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3617 	    pfh_inet);
3618 	pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3619 	    pfh_inet);
3620 #endif
3621 #ifdef INET6
3622 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3623 	if (pfh_inet6 == NULL)
3624 		return (ESRCH); /* XXX */
3625 	pfil_remove_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK,
3626 	    pfh_inet6);
3627 	pfil_remove_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK,
3628 	    pfh_inet6);
3629 #endif
3630 
3631 	V_pf_pfil_hooked = 0;
3632 	return (0);
3633 }
3634 
3635 static int
3636 pf_load(void)
3637 {
3638 	int error;
3639 
3640 	VNET_ITERATOR_DECL(vnet_iter);
3641 
3642 	VNET_LIST_RLOCK();
3643 	VNET_FOREACH(vnet_iter) {
3644 		CURVNET_SET(vnet_iter);
3645 		V_pf_pfil_hooked = 0;
3646 		V_pf_end_threads = 0;
3647 		TAILQ_INIT(&V_pf_tags);
3648 		TAILQ_INIT(&V_pf_qids);
3649 		CURVNET_RESTORE();
3650 	}
3651 	VNET_LIST_RUNLOCK();
3652 
3653 	rw_init(&pf_rules_lock, "pf rulesets");
3654 
3655 	pf_dev = make_dev(&pf_cdevsw, 0, 0, 0, 0600, PF_NAME);
3656 	if ((error = pfattach()) != 0)
3657 		return (error);
3658 
3659 	return (0);
3660 }
3661 
3662 static int
3663 pf_unload(void)
3664 {
3665 	int error = 0;
3666 
3667 	PF_RULES_WLOCK();
3668 	V_pf_status.running = 0;
3669 	PF_RULES_WUNLOCK();
3670 	swi_remove(V_pf_swi_cookie);
3671 	error = dehook_pf();
3672 	if (error) {
3673 		/*
3674 		 * Should not happen!
3675 		 * XXX Due to error code ESRCH, kldunload will show
3676 		 * a message like 'No such process'.
3677 		 */
3678 		printf("%s : pfil unregisteration fail\n", __FUNCTION__);
3679 		return error;
3680 	}
3681 	PF_RULES_WLOCK();
3682 	shutdown_pf();
3683 	V_pf_end_threads = 1;
3684 	while (V_pf_end_threads < 2) {
3685 		wakeup_one(pf_purge_thread);
3686 		rw_sleep(pf_purge_thread, &pf_rules_lock, 0, "pftmo", 0);
3687 	}
3688 	pf_normalize_cleanup();
3689 	pfi_cleanup();
3690 	pfr_cleanup();
3691 	pf_osfp_flush();
3692 	pf_cleanup();
3693 	PF_RULES_WUNLOCK();
3694 	destroy_dev(pf_dev);
3695 	rw_destroy(&pf_rules_lock);
3696 
3697 	return (error);
3698 }
3699 
3700 static int
3701 pf_modevent(module_t mod, int type, void *data)
3702 {
3703 	int error = 0;
3704 
3705 	switch(type) {
3706 	case MOD_LOAD:
3707 		error = pf_load();
3708 		break;
3709 	case MOD_QUIESCE:
3710 		/*
3711 		 * Module should not be unloaded due to race conditions.
3712 		 */
3713 		error = EBUSY;
3714 		break;
3715 	case MOD_UNLOAD:
3716 		error = pf_unload();
3717 		break;
3718 	default:
3719 		error = EINVAL;
3720 		break;
3721 	}
3722 
3723 	return (error);
3724 }
3725 
3726 static moduledata_t pf_mod = {
3727 	"pf",
3728 	pf_modevent,
3729 	0
3730 };
3731 
3732 DECLARE_MODULE(pf, pf_mod, SI_SUB_PSEUDO, SI_ORDER_FIRST);
3733 MODULE_VERSION(pf, PF_MODVER);
3734