xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision 2710751bc309af25c6dea1171781678258e83840)
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/mbuf.h>
57 #include <sys/module.h>
58 #include <sys/proc.h>
59 #include <sys/smp.h>
60 #include <sys/socket.h>
61 #include <sys/sysctl.h>
62 #include <sys/md5.h>
63 #include <sys/ucred.h>
64 
65 #include <net/if.h>
66 #include <net/route.h>
67 #include <net/pfil.h>
68 #include <net/pfvar.h>
69 #include <net/if_pfsync.h>
70 #include <net/if_pflog.h>
71 
72 #include <netinet/in.h>
73 #include <netinet/ip.h>
74 #include <netinet/ip_var.h>
75 #include <netinet/ip_icmp.h>
76 
77 #ifdef INET6
78 #include <netinet/ip6.h>
79 #endif /* INET6 */
80 
81 #ifdef ALTQ
82 #include <altq/altq.h>
83 #endif
84 
85 static int		 pfattach(void);
86 static struct pf_pool	*pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t,
87 			    u_int8_t, u_int8_t, u_int8_t);
88 
89 static void		 pf_mv_pool(struct pf_palist *, struct pf_palist *);
90 static void		 pf_empty_pool(struct pf_palist *);
91 static int		 pfioctl(struct cdev *, u_long, caddr_t, int,
92 			    struct thread *);
93 #ifdef ALTQ
94 static int		 pf_begin_altq(u_int32_t *);
95 static int		 pf_rollback_altq(u_int32_t);
96 static int		 pf_commit_altq(u_int32_t);
97 static int		 pf_enable_altq(struct pf_altq *);
98 static int		 pf_disable_altq(struct pf_altq *);
99 static u_int32_t	 pf_qname2qid(char *);
100 static void		 pf_qid_unref(u_int32_t);
101 #endif /* ALTQ */
102 static int		 pf_begin_rules(u_int32_t *, int, const char *);
103 static int		 pf_rollback_rules(u_int32_t, int, char *);
104 static int		 pf_setup_pfsync_matching(struct pf_ruleset *);
105 static void		 pf_hash_rule(MD5_CTX *, struct pf_rule *);
106 static void		 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
107 static int		 pf_commit_rules(u_int32_t, int, char *);
108 static int		 pf_addr_setup(struct pf_ruleset *,
109 			    struct pf_addr_wrap *, sa_family_t);
110 static void		 pf_addr_copyout(struct pf_addr_wrap *);
111 
112 VNET_DEFINE(struct pf_rule,	pf_default_rule);
113 
114 #ifdef ALTQ
115 static VNET_DEFINE(int,		pf_altq_running);
116 #define	V_pf_altq_running	VNET(pf_altq_running)
117 #endif
118 
119 #define	TAGID_MAX	 50000
120 struct pf_tagname {
121 	TAILQ_ENTRY(pf_tagname)	entries;
122 	char			name[PF_TAG_NAME_SIZE];
123 	uint16_t		tag;
124 	int			ref;
125 };
126 
127 TAILQ_HEAD(pf_tags, pf_tagname);
128 #define	V_pf_tags		VNET(pf_tags)
129 VNET_DEFINE(struct pf_tags, pf_tags);
130 #define	V_pf_qids		VNET(pf_qids)
131 VNET_DEFINE(struct pf_tags, pf_qids);
132 static MALLOC_DEFINE(M_PFTAG, "pf_tag", "pf(4) tag names");
133 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
134 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
135 
136 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
137 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
138 #endif
139 
140 static u_int16_t	 tagname2tag(struct pf_tags *, char *);
141 static u_int16_t	 pf_tagname2tag(char *);
142 static void		 tag_unref(struct pf_tags *, u_int16_t);
143 
144 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
145 
146 struct cdev *pf_dev;
147 
148 /*
149  * XXX - These are new and need to be checked when moveing to a new version
150  */
151 static void		 pf_clear_states(void);
152 static int		 pf_clear_tables(void);
153 static void		 pf_clear_srcnodes(struct pf_src_node *);
154 static void		 pf_tbladdr_copyout(struct pf_addr_wrap *);
155 
156 /*
157  * Wrapper functions for pfil(9) hooks
158  */
159 #ifdef INET
160 static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp,
161     int dir, struct inpcb *inp);
162 static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp,
163     int dir, struct inpcb *inp);
164 #endif
165 #ifdef INET6
166 static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp,
167     int dir, struct inpcb *inp);
168 static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp,
169     int dir, struct inpcb *inp);
170 #endif
171 
172 static int		hook_pf(void);
173 static int		dehook_pf(void);
174 static int		shutdown_pf(void);
175 static int		pf_load(void);
176 static int		pf_unload(void);
177 
178 static struct cdevsw pf_cdevsw = {
179 	.d_ioctl =	pfioctl,
180 	.d_name =	PF_NAME,
181 	.d_version =	D_VERSION,
182 };
183 
184 static volatile VNET_DEFINE(int, pf_pfil_hooked);
185 #define V_pf_pfil_hooked	VNET(pf_pfil_hooked)
186 VNET_DEFINE(int,		pf_end_threads);
187 
188 struct rwlock			pf_rules_lock;
189 
190 /* pfsync */
191 pfsync_state_import_t 		*pfsync_state_import_ptr = NULL;
192 pfsync_insert_state_t		*pfsync_insert_state_ptr = NULL;
193 pfsync_update_state_t		*pfsync_update_state_ptr = NULL;
194 pfsync_delete_state_t		*pfsync_delete_state_ptr = NULL;
195 pfsync_clear_states_t		*pfsync_clear_states_ptr = NULL;
196 pfsync_defer_t			*pfsync_defer_ptr = NULL;
197 /* pflog */
198 pflog_packet_t			*pflog_packet_ptr = NULL;
199 
200 static int
201 pfattach(void)
202 {
203 	u_int32_t *my_timeout = V_pf_default_rule.timeout;
204 	int error;
205 
206 	pf_initialize();
207 	pfr_initialize();
208 	pfi_initialize();
209 	pf_normalize_init();
210 
211 	V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
212 	V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
213 
214 	RB_INIT(&V_pf_anchors);
215 	pf_init_ruleset(&pf_main_ruleset);
216 
217 	/* default rule should never be garbage collected */
218 	V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
219 #ifdef PF_DEFAULT_TO_DROP
220 	V_pf_default_rule.action = PF_DROP;
221 #else
222 	V_pf_default_rule.action = PF_PASS;
223 #endif
224 	V_pf_default_rule.nr = -1;
225 	V_pf_default_rule.rtableid = -1;
226 
227 	/* initialize default timeouts */
228 	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
229 	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
230 	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
231 	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
232 	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
233 	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
234 	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
235 	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
236 	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
237 	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
238 	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
239 	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
240 	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
241 	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
242 	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
243 	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
244 	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
245 	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
246 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
247 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
248 
249 	bzero(&V_pf_status, sizeof(V_pf_status));
250 	V_pf_status.debug = PF_DEBUG_URGENT;
251 
252 	V_pf_pfil_hooked = 0;
253 
254 	/* XXX do our best to avoid a conflict */
255 	V_pf_status.hostid = arc4random();
256 
257 	if ((error = kproc_create(pf_purge_thread, curvnet, NULL, 0, 0,
258 	    "pf purge")) != 0)
259 		/* XXXGL: leaked all above. */
260 		return (error);
261 	if ((error = swi_add(NULL, "pf send", pf_intr, curvnet, SWI_NET,
262 	    INTR_MPSAFE, &V_pf_swi_cookie)) != 0)
263 		/* XXXGL: leaked all above. */
264 		return (error);
265 
266 	return (0);
267 }
268 
269 static struct pf_pool *
270 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
271     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
272     u_int8_t check_ticket)
273 {
274 	struct pf_ruleset	*ruleset;
275 	struct pf_rule		*rule;
276 	int			 rs_num;
277 
278 	ruleset = pf_find_ruleset(anchor);
279 	if (ruleset == NULL)
280 		return (NULL);
281 	rs_num = pf_get_ruleset_number(rule_action);
282 	if (rs_num >= PF_RULESET_MAX)
283 		return (NULL);
284 	if (active) {
285 		if (check_ticket && ticket !=
286 		    ruleset->rules[rs_num].active.ticket)
287 			return (NULL);
288 		if (r_last)
289 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
290 			    pf_rulequeue);
291 		else
292 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
293 	} else {
294 		if (check_ticket && ticket !=
295 		    ruleset->rules[rs_num].inactive.ticket)
296 			return (NULL);
297 		if (r_last)
298 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
299 			    pf_rulequeue);
300 		else
301 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
302 	}
303 	if (!r_last) {
304 		while ((rule != NULL) && (rule->nr != rule_number))
305 			rule = TAILQ_NEXT(rule, entries);
306 	}
307 	if (rule == NULL)
308 		return (NULL);
309 
310 	return (&rule->rpool);
311 }
312 
313 static void
314 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
315 {
316 	struct pf_pooladdr	*mv_pool_pa;
317 
318 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
319 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
320 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
321 	}
322 }
323 
324 static void
325 pf_empty_pool(struct pf_palist *poola)
326 {
327 	struct pf_pooladdr *pa;
328 
329 	while ((pa = TAILQ_FIRST(poola)) != NULL) {
330 		switch (pa->addr.type) {
331 		case PF_ADDR_DYNIFTL:
332 			pfi_dynaddr_remove(pa->addr.p.dyn);
333 			break;
334 		case PF_ADDR_TABLE:
335 			pfr_detach_table(pa->addr.p.tbl);
336 			break;
337 		}
338 		if (pa->kif)
339 			pfi_kif_unref(pa->kif);
340 		TAILQ_REMOVE(poola, pa, entries);
341 		free(pa, M_PFRULE);
342 	}
343 }
344 
345 static void
346 pf_unlink_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
347 {
348 
349 	PF_RULES_WASSERT();
350 
351 	TAILQ_REMOVE(rulequeue, rule, entries);
352 
353 	PF_UNLNKDRULES_LOCK();
354 	rule->rule_flag |= PFRULE_REFS;
355 	TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
356 	PF_UNLNKDRULES_UNLOCK();
357 }
358 
359 void
360 pf_free_rule(struct pf_rule *rule)
361 {
362 
363 	PF_RULES_WASSERT();
364 
365 	if (rule->tag)
366 		tag_unref(&V_pf_tags, rule->tag);
367 	if (rule->match_tag)
368 		tag_unref(&V_pf_tags, rule->match_tag);
369 #ifdef ALTQ
370 	if (rule->pqid != rule->qid)
371 		pf_qid_unref(rule->pqid);
372 	pf_qid_unref(rule->qid);
373 #endif
374 	switch (rule->src.addr.type) {
375 	case PF_ADDR_DYNIFTL:
376 		pfi_dynaddr_remove(rule->src.addr.p.dyn);
377 		break;
378 	case PF_ADDR_TABLE:
379 		pfr_detach_table(rule->src.addr.p.tbl);
380 		break;
381 	}
382 	switch (rule->dst.addr.type) {
383 	case PF_ADDR_DYNIFTL:
384 		pfi_dynaddr_remove(rule->dst.addr.p.dyn);
385 		break;
386 	case PF_ADDR_TABLE:
387 		pfr_detach_table(rule->dst.addr.p.tbl);
388 		break;
389 	}
390 	if (rule->overload_tbl)
391 		pfr_detach_table(rule->overload_tbl);
392 	if (rule->kif)
393 		pfi_kif_unref(rule->kif);
394 	pf_anchor_remove(rule);
395 	pf_empty_pool(&rule->rpool.list);
396 	free(rule, M_PFRULE);
397 }
398 
399 static u_int16_t
400 tagname2tag(struct pf_tags *head, char *tagname)
401 {
402 	struct pf_tagname	*tag, *p = NULL;
403 	u_int16_t		 new_tagid = 1;
404 
405 	PF_RULES_WASSERT();
406 
407 	TAILQ_FOREACH(tag, head, entries)
408 		if (strcmp(tagname, tag->name) == 0) {
409 			tag->ref++;
410 			return (tag->tag);
411 		}
412 
413 	/*
414 	 * to avoid fragmentation, we do a linear search from the beginning
415 	 * and take the first free slot we find. if there is none or the list
416 	 * is empty, append a new entry at the end.
417 	 */
418 
419 	/* new entry */
420 	if (!TAILQ_EMPTY(head))
421 		for (p = TAILQ_FIRST(head); p != NULL &&
422 		    p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
423 			new_tagid = p->tag + 1;
424 
425 	if (new_tagid > TAGID_MAX)
426 		return (0);
427 
428 	/* allocate and fill new struct pf_tagname */
429 	tag = malloc(sizeof(*tag), M_PFTAG, M_NOWAIT|M_ZERO);
430 	if (tag == NULL)
431 		return (0);
432 	strlcpy(tag->name, tagname, sizeof(tag->name));
433 	tag->tag = new_tagid;
434 	tag->ref++;
435 
436 	if (p != NULL)	/* insert new entry before p */
437 		TAILQ_INSERT_BEFORE(p, tag, entries);
438 	else	/* either list empty or no free slot in between */
439 		TAILQ_INSERT_TAIL(head, tag, entries);
440 
441 	return (tag->tag);
442 }
443 
444 static void
445 tag_unref(struct pf_tags *head, u_int16_t tag)
446 {
447 	struct pf_tagname	*p, *next;
448 
449 	PF_RULES_WASSERT();
450 
451 	for (p = TAILQ_FIRST(head); p != NULL; p = next) {
452 		next = TAILQ_NEXT(p, entries);
453 		if (tag == p->tag) {
454 			if (--p->ref == 0) {
455 				TAILQ_REMOVE(head, p, entries);
456 				free(p, M_PFTAG);
457 			}
458 			break;
459 		}
460 	}
461 }
462 
463 static u_int16_t
464 pf_tagname2tag(char *tagname)
465 {
466 	return (tagname2tag(&V_pf_tags, tagname));
467 }
468 
469 #ifdef ALTQ
470 static u_int32_t
471 pf_qname2qid(char *qname)
472 {
473 	return ((u_int32_t)tagname2tag(&V_pf_qids, qname));
474 }
475 
476 static void
477 pf_qid_unref(u_int32_t qid)
478 {
479 	tag_unref(&V_pf_qids, (u_int16_t)qid);
480 }
481 
482 static int
483 pf_begin_altq(u_int32_t *ticket)
484 {
485 	struct pf_altq	*altq;
486 	int		 error = 0;
487 
488 	PF_RULES_WASSERT();
489 
490 	/* Purge the old altq list */
491 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
492 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
493 		if (altq->qname[0] == 0 &&
494 		    (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
495 			/* detach and destroy the discipline */
496 			error = altq_remove(altq);
497 		} else
498 			pf_qid_unref(altq->qid);
499 		free(altq, M_PFALTQ);
500 	}
501 	if (error)
502 		return (error);
503 	*ticket = ++V_ticket_altqs_inactive;
504 	V_altqs_inactive_open = 1;
505 	return (0);
506 }
507 
508 static int
509 pf_rollback_altq(u_int32_t ticket)
510 {
511 	struct pf_altq	*altq;
512 	int		 error = 0;
513 
514 	PF_RULES_WASSERT();
515 
516 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
517 		return (0);
518 	/* Purge the old altq list */
519 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
520 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
521 		if (altq->qname[0] == 0 &&
522 		   (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
523 			/* detach and destroy the discipline */
524 			error = altq_remove(altq);
525 		} else
526 			pf_qid_unref(altq->qid);
527 		free(altq, M_PFALTQ);
528 	}
529 	V_altqs_inactive_open = 0;
530 	return (error);
531 }
532 
533 static int
534 pf_commit_altq(u_int32_t ticket)
535 {
536 	struct pf_altqqueue	*old_altqs;
537 	struct pf_altq		*altq;
538 	int			 err, error = 0;
539 
540 	PF_RULES_WASSERT();
541 
542 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
543 		return (EBUSY);
544 
545 	/* swap altqs, keep the old. */
546 	old_altqs = V_pf_altqs_active;
547 	V_pf_altqs_active = V_pf_altqs_inactive;
548 	V_pf_altqs_inactive = old_altqs;
549 	V_ticket_altqs_active = V_ticket_altqs_inactive;
550 
551 	/* Attach new disciplines */
552 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
553 	if (altq->qname[0] == 0 &&
554 	   (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
555 			/* attach the discipline */
556 			error = altq_pfattach(altq);
557 			if (error == 0 && V_pf_altq_running)
558 				error = pf_enable_altq(altq);
559 			if (error != 0)
560 				return (error);
561 		}
562 	}
563 
564 	/* Purge the old altq list */
565 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
566 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
567 		if (altq->qname[0] == 0 &&
568 		    (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
569 			/* detach and destroy the discipline */
570 			if (V_pf_altq_running)
571 				error = pf_disable_altq(altq);
572 			err = altq_pfdetach(altq);
573 			if (err != 0 && error == 0)
574 				error = err;
575 			err = altq_remove(altq);
576 			if (err != 0 && error == 0)
577 				error = err;
578 		} else
579 			pf_qid_unref(altq->qid);
580 		free(altq, M_PFALTQ);
581 	}
582 
583 	V_altqs_inactive_open = 0;
584 	return (error);
585 }
586 
587 static int
588 pf_enable_altq(struct pf_altq *altq)
589 {
590 	struct ifnet		*ifp;
591 	struct tb_profile	 tb;
592 	int			 error = 0;
593 
594 	if ((ifp = ifunit(altq->ifname)) == NULL)
595 		return (EINVAL);
596 
597 	if (ifp->if_snd.altq_type != ALTQT_NONE)
598 		error = altq_enable(&ifp->if_snd);
599 
600 	/* set tokenbucket regulator */
601 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
602 		tb.rate = altq->ifbandwidth;
603 		tb.depth = altq->tbrsize;
604 		error = tbr_set(&ifp->if_snd, &tb);
605 	}
606 
607 	return (error);
608 }
609 
610 static int
611 pf_disable_altq(struct pf_altq *altq)
612 {
613 	struct ifnet		*ifp;
614 	struct tb_profile	 tb;
615 	int			 error;
616 
617 	if ((ifp = ifunit(altq->ifname)) == NULL)
618 		return (EINVAL);
619 
620 	/*
621 	 * when the discipline is no longer referenced, it was overridden
622 	 * by a new one.  if so, just return.
623 	 */
624 	if (altq->altq_disc != ifp->if_snd.altq_disc)
625 		return (0);
626 
627 	error = altq_disable(&ifp->if_snd);
628 
629 	if (error == 0) {
630 		/* clear tokenbucket regulator */
631 		tb.rate = 0;
632 		error = tbr_set(&ifp->if_snd, &tb);
633 	}
634 
635 	return (error);
636 }
637 
638 void
639 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
640 {
641 	struct ifnet	*ifp1;
642 	struct pf_altq	*a1, *a2, *a3;
643 	u_int32_t	 ticket;
644 	int		 error = 0;
645 
646 	/* Interrupt userland queue modifications */
647 	if (V_altqs_inactive_open)
648 		pf_rollback_altq(V_ticket_altqs_inactive);
649 
650 	/* Start new altq ruleset */
651 	if (pf_begin_altq(&ticket))
652 		return;
653 
654 	/* Copy the current active set */
655 	TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
656 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
657 		if (a2 == NULL) {
658 			error = ENOMEM;
659 			break;
660 		}
661 		bcopy(a1, a2, sizeof(struct pf_altq));
662 
663 		if (a2->qname[0] != 0) {
664 			if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
665 				error = EBUSY;
666 				free(a2, M_PFALTQ);
667 				break;
668 			}
669 			a2->altq_disc = NULL;
670 			TAILQ_FOREACH(a3, V_pf_altqs_inactive, entries) {
671 				if (strncmp(a3->ifname, a2->ifname,
672 				    IFNAMSIZ) == 0 && a3->qname[0] == 0) {
673 					a2->altq_disc = a3->altq_disc;
674 					break;
675 				}
676 			}
677 		}
678 		/* Deactivate the interface in question */
679 		a2->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
680 		if ((ifp1 = ifunit(a2->ifname)) == NULL ||
681 		    (remove && ifp1 == ifp)) {
682 			a2->local_flags |= PFALTQ_FLAG_IF_REMOVED;
683 		} else {
684 			error = altq_add(a2);
685 
686 			if (ticket != V_ticket_altqs_inactive)
687 				error = EBUSY;
688 
689 			if (error) {
690 				free(a2, M_PFALTQ);
691 				break;
692 			}
693 		}
694 
695 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
696 	}
697 
698 	if (error != 0)
699 		pf_rollback_altq(ticket);
700 	else
701 		pf_commit_altq(ticket);
702 }
703 #endif /* ALTQ */
704 
705 static int
706 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
707 {
708 	struct pf_ruleset	*rs;
709 	struct pf_rule		*rule;
710 
711 	PF_RULES_WASSERT();
712 
713 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
714 		return (EINVAL);
715 	rs = pf_find_or_create_ruleset(anchor);
716 	if (rs == NULL)
717 		return (EINVAL);
718 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
719 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
720 		rs->rules[rs_num].inactive.rcount--;
721 	}
722 	*ticket = ++rs->rules[rs_num].inactive.ticket;
723 	rs->rules[rs_num].inactive.open = 1;
724 	return (0);
725 }
726 
727 static int
728 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
729 {
730 	struct pf_ruleset	*rs;
731 	struct pf_rule		*rule;
732 
733 	PF_RULES_WASSERT();
734 
735 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
736 		return (EINVAL);
737 	rs = pf_find_ruleset(anchor);
738 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
739 	    rs->rules[rs_num].inactive.ticket != ticket)
740 		return (0);
741 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
742 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
743 		rs->rules[rs_num].inactive.rcount--;
744 	}
745 	rs->rules[rs_num].inactive.open = 0;
746 	return (0);
747 }
748 
749 #define PF_MD5_UPD(st, elm)						\
750 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
751 
752 #define PF_MD5_UPD_STR(st, elm)						\
753 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
754 
755 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
756 		(stor) = htonl((st)->elm);				\
757 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
758 } while (0)
759 
760 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
761 		(stor) = htons((st)->elm);				\
762 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
763 } while (0)
764 
765 static void
766 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
767 {
768 	PF_MD5_UPD(pfr, addr.type);
769 	switch (pfr->addr.type) {
770 		case PF_ADDR_DYNIFTL:
771 			PF_MD5_UPD(pfr, addr.v.ifname);
772 			PF_MD5_UPD(pfr, addr.iflags);
773 			break;
774 		case PF_ADDR_TABLE:
775 			PF_MD5_UPD(pfr, addr.v.tblname);
776 			break;
777 		case PF_ADDR_ADDRMASK:
778 			/* XXX ignore af? */
779 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
780 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
781 			break;
782 	}
783 
784 	PF_MD5_UPD(pfr, port[0]);
785 	PF_MD5_UPD(pfr, port[1]);
786 	PF_MD5_UPD(pfr, neg);
787 	PF_MD5_UPD(pfr, port_op);
788 }
789 
790 static void
791 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
792 {
793 	u_int16_t x;
794 	u_int32_t y;
795 
796 	pf_hash_rule_addr(ctx, &rule->src);
797 	pf_hash_rule_addr(ctx, &rule->dst);
798 	PF_MD5_UPD_STR(rule, label);
799 	PF_MD5_UPD_STR(rule, ifname);
800 	PF_MD5_UPD_STR(rule, match_tagname);
801 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
802 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
803 	PF_MD5_UPD_HTONL(rule, prob, y);
804 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
805 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
806 	PF_MD5_UPD(rule, uid.op);
807 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
808 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
809 	PF_MD5_UPD(rule, gid.op);
810 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
811 	PF_MD5_UPD(rule, action);
812 	PF_MD5_UPD(rule, direction);
813 	PF_MD5_UPD(rule, af);
814 	PF_MD5_UPD(rule, quick);
815 	PF_MD5_UPD(rule, ifnot);
816 	PF_MD5_UPD(rule, match_tag_not);
817 	PF_MD5_UPD(rule, natpass);
818 	PF_MD5_UPD(rule, keep_state);
819 	PF_MD5_UPD(rule, proto);
820 	PF_MD5_UPD(rule, type);
821 	PF_MD5_UPD(rule, code);
822 	PF_MD5_UPD(rule, flags);
823 	PF_MD5_UPD(rule, flagset);
824 	PF_MD5_UPD(rule, allow_opts);
825 	PF_MD5_UPD(rule, rt);
826 	PF_MD5_UPD(rule, tos);
827 }
828 
829 static int
830 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
831 {
832 	struct pf_ruleset	*rs;
833 	struct pf_rule		*rule, **old_array;
834 	struct pf_rulequeue	*old_rules;
835 	int			 error;
836 	u_int32_t		 old_rcount;
837 
838 	PF_RULES_WASSERT();
839 
840 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
841 		return (EINVAL);
842 	rs = pf_find_ruleset(anchor);
843 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
844 	    ticket != rs->rules[rs_num].inactive.ticket)
845 		return (EBUSY);
846 
847 	/* Calculate checksum for the main ruleset */
848 	if (rs == &pf_main_ruleset) {
849 		error = pf_setup_pfsync_matching(rs);
850 		if (error != 0)
851 			return (error);
852 	}
853 
854 	/* Swap rules, keep the old. */
855 	old_rules = rs->rules[rs_num].active.ptr;
856 	old_rcount = rs->rules[rs_num].active.rcount;
857 	old_array = rs->rules[rs_num].active.ptr_array;
858 
859 	rs->rules[rs_num].active.ptr =
860 	    rs->rules[rs_num].inactive.ptr;
861 	rs->rules[rs_num].active.ptr_array =
862 	    rs->rules[rs_num].inactive.ptr_array;
863 	rs->rules[rs_num].active.rcount =
864 	    rs->rules[rs_num].inactive.rcount;
865 	rs->rules[rs_num].inactive.ptr = old_rules;
866 	rs->rules[rs_num].inactive.ptr_array = old_array;
867 	rs->rules[rs_num].inactive.rcount = old_rcount;
868 
869 	rs->rules[rs_num].active.ticket =
870 	    rs->rules[rs_num].inactive.ticket;
871 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
872 
873 
874 	/* Purge the old rule list. */
875 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
876 		pf_unlink_rule(old_rules, rule);
877 	if (rs->rules[rs_num].inactive.ptr_array)
878 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
879 	rs->rules[rs_num].inactive.ptr_array = NULL;
880 	rs->rules[rs_num].inactive.rcount = 0;
881 	rs->rules[rs_num].inactive.open = 0;
882 	pf_remove_if_empty_ruleset(rs);
883 
884 	return (0);
885 }
886 
887 static int
888 pf_setup_pfsync_matching(struct pf_ruleset *rs)
889 {
890 	MD5_CTX			 ctx;
891 	struct pf_rule		*rule;
892 	int			 rs_cnt;
893 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
894 
895 	MD5Init(&ctx);
896 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
897 		/* XXX PF_RULESET_SCRUB as well? */
898 		if (rs_cnt == PF_RULESET_SCRUB)
899 			continue;
900 
901 		if (rs->rules[rs_cnt].inactive.ptr_array)
902 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
903 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
904 
905 		if (rs->rules[rs_cnt].inactive.rcount) {
906 			rs->rules[rs_cnt].inactive.ptr_array =
907 			    malloc(sizeof(caddr_t) *
908 			    rs->rules[rs_cnt].inactive.rcount,
909 			    M_TEMP, M_NOWAIT);
910 
911 			if (!rs->rules[rs_cnt].inactive.ptr_array)
912 				return (ENOMEM);
913 		}
914 
915 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
916 		    entries) {
917 			pf_hash_rule(&ctx, rule);
918 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
919 		}
920 	}
921 
922 	MD5Final(digest, &ctx);
923 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
924 	return (0);
925 }
926 
927 static int
928 pf_addr_setup(struct pf_ruleset *ruleset, struct pf_addr_wrap *addr,
929     sa_family_t af)
930 {
931 	int error = 0;
932 
933 	switch (addr->type) {
934 	case PF_ADDR_TABLE:
935 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
936 		if (addr->p.tbl == NULL)
937 			error = ENOMEM;
938 		break;
939 	case PF_ADDR_DYNIFTL:
940 		error = pfi_dynaddr_setup(addr, af);
941 		break;
942 	}
943 
944 	return (error);
945 }
946 
947 static void
948 pf_addr_copyout(struct pf_addr_wrap *addr)
949 {
950 
951 	switch (addr->type) {
952 	case PF_ADDR_DYNIFTL:
953 		pfi_dynaddr_copyout(addr);
954 		break;
955 	case PF_ADDR_TABLE:
956 		pf_tbladdr_copyout(addr);
957 		break;
958 	}
959 }
960 
961 static int
962 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
963 {
964 	int			 error = 0;
965 
966 	/* XXX keep in sync with switch() below */
967 	if (securelevel_gt(td->td_ucred, 2))
968 		switch (cmd) {
969 		case DIOCGETRULES:
970 		case DIOCGETRULE:
971 		case DIOCGETADDRS:
972 		case DIOCGETADDR:
973 		case DIOCGETSTATE:
974 		case DIOCSETSTATUSIF:
975 		case DIOCGETSTATUS:
976 		case DIOCCLRSTATUS:
977 		case DIOCNATLOOK:
978 		case DIOCSETDEBUG:
979 		case DIOCGETSTATES:
980 		case DIOCGETTIMEOUT:
981 		case DIOCCLRRULECTRS:
982 		case DIOCGETLIMIT:
983 		case DIOCGETALTQS:
984 		case DIOCGETALTQ:
985 		case DIOCGETQSTATS:
986 		case DIOCGETRULESETS:
987 		case DIOCGETRULESET:
988 		case DIOCRGETTABLES:
989 		case DIOCRGETTSTATS:
990 		case DIOCRCLRTSTATS:
991 		case DIOCRCLRADDRS:
992 		case DIOCRADDADDRS:
993 		case DIOCRDELADDRS:
994 		case DIOCRSETADDRS:
995 		case DIOCRGETADDRS:
996 		case DIOCRGETASTATS:
997 		case DIOCRCLRASTATS:
998 		case DIOCRTSTADDRS:
999 		case DIOCOSFPGET:
1000 		case DIOCGETSRCNODES:
1001 		case DIOCCLRSRCNODES:
1002 		case DIOCIGETIFACES:
1003 		case DIOCGIFSPEED:
1004 		case DIOCSETIFFLAG:
1005 		case DIOCCLRIFFLAG:
1006 			break;
1007 		case DIOCRCLRTABLES:
1008 		case DIOCRADDTABLES:
1009 		case DIOCRDELTABLES:
1010 		case DIOCRSETTFLAGS:
1011 			if (((struct pfioc_table *)addr)->pfrio_flags &
1012 			    PFR_FLAG_DUMMY)
1013 				break; /* dummy operation ok */
1014 			return (EPERM);
1015 		default:
1016 			return (EPERM);
1017 		}
1018 
1019 	if (!(flags & FWRITE))
1020 		switch (cmd) {
1021 		case DIOCGETRULES:
1022 		case DIOCGETADDRS:
1023 		case DIOCGETADDR:
1024 		case DIOCGETSTATE:
1025 		case DIOCGETSTATUS:
1026 		case DIOCGETSTATES:
1027 		case DIOCGETTIMEOUT:
1028 		case DIOCGETLIMIT:
1029 		case DIOCGETALTQS:
1030 		case DIOCGETALTQ:
1031 		case DIOCGETQSTATS:
1032 		case DIOCGETRULESETS:
1033 		case DIOCGETRULESET:
1034 		case DIOCNATLOOK:
1035 		case DIOCRGETTABLES:
1036 		case DIOCRGETTSTATS:
1037 		case DIOCRGETADDRS:
1038 		case DIOCRGETASTATS:
1039 		case DIOCRTSTADDRS:
1040 		case DIOCOSFPGET:
1041 		case DIOCGETSRCNODES:
1042 		case DIOCIGETIFACES:
1043 		case DIOCGIFSPEED:
1044 			break;
1045 		case DIOCRCLRTABLES:
1046 		case DIOCRADDTABLES:
1047 		case DIOCRDELTABLES:
1048 		case DIOCRCLRTSTATS:
1049 		case DIOCRCLRADDRS:
1050 		case DIOCRADDADDRS:
1051 		case DIOCRDELADDRS:
1052 		case DIOCRSETADDRS:
1053 		case DIOCRSETTFLAGS:
1054 			if (((struct pfioc_table *)addr)->pfrio_flags &
1055 			    PFR_FLAG_DUMMY) {
1056 				flags |= FWRITE; /* need write lock for dummy */
1057 				break; /* dummy operation ok */
1058 			}
1059 			return (EACCES);
1060 		case DIOCGETRULE:
1061 			if (((struct pfioc_rule *)addr)->action ==
1062 			    PF_GET_CLR_CNTR)
1063 				return (EACCES);
1064 			break;
1065 		default:
1066 			return (EACCES);
1067 		}
1068 
1069 	CURVNET_SET(TD_TO_VNET(td));
1070 
1071 	switch (cmd) {
1072 	case DIOCSTART:
1073 		PF_RULES_WLOCK();
1074 		if (V_pf_status.running)
1075 			error = EEXIST;
1076 		else {
1077 			int cpu;
1078 
1079 			PF_RULES_WUNLOCK();
1080 			error = hook_pf();
1081 			if (error) {
1082 				DPFPRINTF(PF_DEBUG_MISC,
1083 				    ("pf: pfil registration failed\n"));
1084 				break;
1085 			}
1086 			PF_RULES_WLOCK();
1087 			V_pf_status.running = 1;
1088 			V_pf_status.since = time_second;
1089 
1090 			CPU_FOREACH(cpu)
1091 				V_pf_stateid[cpu] = time_second;
1092 
1093 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1094 		}
1095 		PF_RULES_WUNLOCK();
1096 		break;
1097 
1098 	case DIOCSTOP:
1099 		PF_RULES_WLOCK();
1100 		if (!V_pf_status.running)
1101 			error = ENOENT;
1102 		else {
1103 			V_pf_status.running = 0;
1104 			PF_RULES_WUNLOCK();
1105 			error = dehook_pf();
1106 			if (error) {
1107 				V_pf_status.running = 1;
1108 				DPFPRINTF(PF_DEBUG_MISC,
1109 				    ("pf: pfil unregistration failed\n"));
1110 			}
1111 			PF_RULES_WLOCK();
1112 			V_pf_status.since = time_second;
1113 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1114 		}
1115 		PF_RULES_WUNLOCK();
1116 		break;
1117 
1118 	case DIOCADDRULE: {
1119 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1120 		struct pf_ruleset	*ruleset;
1121 		struct pf_rule		*rule, *tail;
1122 		struct pf_pooladdr	*pa;
1123 		struct pfi_kif		*kif = NULL;
1124 		int			 rs_num;
1125 
1126 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1127 			error = EINVAL;
1128 			break;
1129 		}
1130 #ifndef INET
1131 		if (pr->rule.af == AF_INET) {
1132 			error = EAFNOSUPPORT;
1133 			break;
1134 		}
1135 #endif /* INET */
1136 #ifndef INET6
1137 		if (pr->rule.af == AF_INET6) {
1138 			error = EAFNOSUPPORT;
1139 			break;
1140 		}
1141 #endif /* INET6 */
1142 
1143 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
1144 		bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1145 		if (rule->ifname[0])
1146 			kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1147 		rule->cuid = td->td_ucred->cr_ruid;
1148 		rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1149 		TAILQ_INIT(&rule->rpool.list);
1150 
1151 #define	ERROUT(x)	{ error = (x); goto DIOCADDRULE_error; }
1152 
1153 		PF_RULES_WLOCK();
1154 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1155 		ruleset = pf_find_ruleset(pr->anchor);
1156 		if (ruleset == NULL)
1157 			ERROUT(EINVAL);
1158 		rs_num = pf_get_ruleset_number(pr->rule.action);
1159 		if (rs_num >= PF_RULESET_MAX)
1160 			ERROUT(EINVAL);
1161 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1162 			DPFPRINTF(PF_DEBUG_MISC,
1163 			    ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
1164 			    ruleset->rules[rs_num].inactive.ticket));
1165 			ERROUT(EBUSY);
1166 		}
1167 		if (pr->pool_ticket != V_ticket_pabuf) {
1168 			DPFPRINTF(PF_DEBUG_MISC,
1169 			    ("pool_ticket: %d != %d\n", pr->pool_ticket,
1170 			    V_ticket_pabuf));
1171 			ERROUT(EBUSY);
1172 		}
1173 
1174 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1175 		    pf_rulequeue);
1176 		if (tail)
1177 			rule->nr = tail->nr + 1;
1178 		else
1179 			rule->nr = 0;
1180 		if (rule->ifname[0]) {
1181 			rule->kif = pfi_kif_attach(kif, rule->ifname);
1182 			pfi_kif_ref(rule->kif);
1183 		} else
1184 			rule->kif = NULL;
1185 
1186 		if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
1187 			error = EBUSY;
1188 
1189 #ifdef ALTQ
1190 		/* set queue IDs */
1191 		if (rule->qname[0] != 0) {
1192 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1193 				error = EBUSY;
1194 			else if (rule->pqname[0] != 0) {
1195 				if ((rule->pqid =
1196 				    pf_qname2qid(rule->pqname)) == 0)
1197 					error = EBUSY;
1198 			} else
1199 				rule->pqid = rule->qid;
1200 		}
1201 #endif
1202 		if (rule->tagname[0])
1203 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1204 				error = EBUSY;
1205 		if (rule->match_tagname[0])
1206 			if ((rule->match_tag =
1207 			    pf_tagname2tag(rule->match_tagname)) == 0)
1208 				error = EBUSY;
1209 		if (rule->rt && !rule->direction)
1210 			error = EINVAL;
1211 		if (!rule->log)
1212 			rule->logif = 0;
1213 		if (rule->logif >= PFLOGIFS_MAX)
1214 			error = EINVAL;
1215 		if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1216 			error = ENOMEM;
1217 		if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1218 			error = ENOMEM;
1219 		if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1220 			error = EINVAL;
1221 		TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1222 			if (pa->addr.type == PF_ADDR_TABLE) {
1223 				pa->addr.p.tbl = pfr_attach_table(ruleset,
1224 				    pa->addr.v.tblname);
1225 				if (pa->addr.p.tbl == NULL)
1226 					error = ENOMEM;
1227 			}
1228 
1229 		if (rule->overload_tblname[0]) {
1230 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1231 			    rule->overload_tblname)) == NULL)
1232 				error = EINVAL;
1233 			else
1234 				rule->overload_tbl->pfrkt_flags |=
1235 				    PFR_TFLAG_ACTIVE;
1236 		}
1237 
1238 		pf_mv_pool(&V_pf_pabuf, &rule->rpool.list);
1239 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1240 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1241 		    (rule->rt > PF_FASTROUTE)) &&
1242 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1243 			error = EINVAL;
1244 
1245 		if (error) {
1246 			pf_free_rule(rule);
1247 			PF_RULES_WUNLOCK();
1248 			break;
1249 		}
1250 
1251 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1252 		rule->evaluations = rule->packets[0] = rule->packets[1] =
1253 		    rule->bytes[0] = rule->bytes[1] = 0;
1254 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1255 		    rule, entries);
1256 		ruleset->rules[rs_num].inactive.rcount++;
1257 		PF_RULES_WUNLOCK();
1258 		break;
1259 
1260 #undef ERROUT
1261 DIOCADDRULE_error:
1262 		PF_RULES_WUNLOCK();
1263 		free(rule, M_PFRULE);
1264 		if (kif)
1265 			free(kif, PFI_MTYPE);
1266 		break;
1267 	}
1268 
1269 	case DIOCGETRULES: {
1270 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1271 		struct pf_ruleset	*ruleset;
1272 		struct pf_rule		*tail;
1273 		int			 rs_num;
1274 
1275 		PF_RULES_WLOCK();
1276 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1277 		ruleset = pf_find_ruleset(pr->anchor);
1278 		if (ruleset == NULL) {
1279 			PF_RULES_WUNLOCK();
1280 			error = EINVAL;
1281 			break;
1282 		}
1283 		rs_num = pf_get_ruleset_number(pr->rule.action);
1284 		if (rs_num >= PF_RULESET_MAX) {
1285 			PF_RULES_WUNLOCK();
1286 			error = EINVAL;
1287 			break;
1288 		}
1289 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1290 		    pf_rulequeue);
1291 		if (tail)
1292 			pr->nr = tail->nr + 1;
1293 		else
1294 			pr->nr = 0;
1295 		pr->ticket = ruleset->rules[rs_num].active.ticket;
1296 		PF_RULES_WUNLOCK();
1297 		break;
1298 	}
1299 
1300 	case DIOCGETRULE: {
1301 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1302 		struct pf_ruleset	*ruleset;
1303 		struct pf_rule		*rule;
1304 		int			 rs_num, i;
1305 
1306 		PF_RULES_WLOCK();
1307 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1308 		ruleset = pf_find_ruleset(pr->anchor);
1309 		if (ruleset == NULL) {
1310 			PF_RULES_WUNLOCK();
1311 			error = EINVAL;
1312 			break;
1313 		}
1314 		rs_num = pf_get_ruleset_number(pr->rule.action);
1315 		if (rs_num >= PF_RULESET_MAX) {
1316 			PF_RULES_WUNLOCK();
1317 			error = EINVAL;
1318 			break;
1319 		}
1320 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1321 			PF_RULES_WUNLOCK();
1322 			error = EBUSY;
1323 			break;
1324 		}
1325 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1326 		while ((rule != NULL) && (rule->nr != pr->nr))
1327 			rule = TAILQ_NEXT(rule, entries);
1328 		if (rule == NULL) {
1329 			PF_RULES_WUNLOCK();
1330 			error = EBUSY;
1331 			break;
1332 		}
1333 		bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1334 		if (pf_anchor_copyout(ruleset, rule, pr)) {
1335 			PF_RULES_WUNLOCK();
1336 			error = EBUSY;
1337 			break;
1338 		}
1339 		pf_addr_copyout(&pr->rule.src.addr);
1340 		pf_addr_copyout(&pr->rule.dst.addr);
1341 		for (i = 0; i < PF_SKIP_COUNT; ++i)
1342 			if (rule->skip[i].ptr == NULL)
1343 				pr->rule.skip[i].nr = -1;
1344 			else
1345 				pr->rule.skip[i].nr =
1346 				    rule->skip[i].ptr->nr;
1347 
1348 		if (pr->action == PF_GET_CLR_CNTR) {
1349 			rule->evaluations = 0;
1350 			rule->packets[0] = rule->packets[1] = 0;
1351 			rule->bytes[0] = rule->bytes[1] = 0;
1352 			rule->states_tot = 0;
1353 		}
1354 		PF_RULES_WUNLOCK();
1355 		break;
1356 	}
1357 
1358 	case DIOCCHANGERULE: {
1359 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1360 		struct pf_ruleset	*ruleset;
1361 		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1362 		struct pfi_kif		*kif = NULL;
1363 		struct pf_pooladdr	*pa;
1364 		u_int32_t		 nr = 0;
1365 		int			 rs_num;
1366 
1367 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1368 		    pcr->action > PF_CHANGE_GET_TICKET) {
1369 			error = EINVAL;
1370 			break;
1371 		}
1372 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1373 			error = EINVAL;
1374 			break;
1375 		}
1376 
1377 		if (pcr->action != PF_CHANGE_REMOVE) {
1378 #ifndef INET
1379 			if (pcr->rule.af == AF_INET) {
1380 				error = EAFNOSUPPORT;
1381 				break;
1382 			}
1383 #endif /* INET */
1384 #ifndef INET6
1385 			if (pcr->rule.af == AF_INET6) {
1386 				error = EAFNOSUPPORT;
1387 				break;
1388 			}
1389 #endif /* INET6 */
1390 			newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
1391 			bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1392 			newrule->cuid = td->td_ucred->cr_ruid;
1393 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1394 			TAILQ_INIT(&newrule->rpool.list);
1395 			/* Initialize refcounting. */
1396 			newrule->states_cur = 0;
1397 			newrule->entries.tqe_prev = NULL;
1398 
1399 			if (newrule->ifname[0])
1400 				kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1401 		}
1402 
1403 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGERULE_error; }
1404 
1405 		PF_RULES_WLOCK();
1406 		if (!(pcr->action == PF_CHANGE_REMOVE ||
1407 		    pcr->action == PF_CHANGE_GET_TICKET) &&
1408 		    pcr->pool_ticket != V_ticket_pabuf)
1409 			ERROUT(EBUSY);
1410 
1411 		ruleset = pf_find_ruleset(pcr->anchor);
1412 		if (ruleset == NULL)
1413 			ERROUT(EINVAL);
1414 
1415 		rs_num = pf_get_ruleset_number(pcr->rule.action);
1416 		if (rs_num >= PF_RULESET_MAX)
1417 			ERROUT(EINVAL);
1418 
1419 		if (pcr->action == PF_CHANGE_GET_TICKET) {
1420 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1421 			ERROUT(0);
1422 		} else if (pcr->ticket !=
1423 			    ruleset->rules[rs_num].active.ticket)
1424 				ERROUT(EINVAL);
1425 
1426 		if (pcr->action != PF_CHANGE_REMOVE) {
1427 			if (newrule->ifname[0]) {
1428 				newrule->kif = pfi_kif_attach(kif,
1429 				    newrule->ifname);
1430 				pfi_kif_ref(newrule->kif);
1431 			} else
1432 				newrule->kif = NULL;
1433 
1434 			if (newrule->rtableid > 0 &&
1435 			    newrule->rtableid >= rt_numfibs)
1436 				error = EBUSY;
1437 
1438 #ifdef ALTQ
1439 			/* set queue IDs */
1440 			if (newrule->qname[0] != 0) {
1441 				if ((newrule->qid =
1442 				    pf_qname2qid(newrule->qname)) == 0)
1443 					error = EBUSY;
1444 				else if (newrule->pqname[0] != 0) {
1445 					if ((newrule->pqid =
1446 					    pf_qname2qid(newrule->pqname)) == 0)
1447 						error = EBUSY;
1448 				} else
1449 					newrule->pqid = newrule->qid;
1450 			}
1451 #endif /* ALTQ */
1452 			if (newrule->tagname[0])
1453 				if ((newrule->tag =
1454 				    pf_tagname2tag(newrule->tagname)) == 0)
1455 					error = EBUSY;
1456 			if (newrule->match_tagname[0])
1457 				if ((newrule->match_tag = pf_tagname2tag(
1458 				    newrule->match_tagname)) == 0)
1459 					error = EBUSY;
1460 			if (newrule->rt && !newrule->direction)
1461 				error = EINVAL;
1462 			if (!newrule->log)
1463 				newrule->logif = 0;
1464 			if (newrule->logif >= PFLOGIFS_MAX)
1465 				error = EINVAL;
1466 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
1467 				error = ENOMEM;
1468 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
1469 				error = ENOMEM;
1470 			if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1471 				error = EINVAL;
1472 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1473 				if (pa->addr.type == PF_ADDR_TABLE) {
1474 					pa->addr.p.tbl =
1475 					    pfr_attach_table(ruleset,
1476 					    pa->addr.v.tblname);
1477 					if (pa->addr.p.tbl == NULL)
1478 						error = ENOMEM;
1479 				}
1480 
1481 			if (newrule->overload_tblname[0]) {
1482 				if ((newrule->overload_tbl = pfr_attach_table(
1483 				    ruleset, newrule->overload_tblname)) ==
1484 				    NULL)
1485 					error = EINVAL;
1486 				else
1487 					newrule->overload_tbl->pfrkt_flags |=
1488 					    PFR_TFLAG_ACTIVE;
1489 			}
1490 
1491 			pf_mv_pool(&V_pf_pabuf, &newrule->rpool.list);
1492 			if (((((newrule->action == PF_NAT) ||
1493 			    (newrule->action == PF_RDR) ||
1494 			    (newrule->action == PF_BINAT) ||
1495 			    (newrule->rt > PF_FASTROUTE)) &&
1496 			    !newrule->anchor)) &&
1497 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1498 				error = EINVAL;
1499 
1500 			if (error) {
1501 				pf_free_rule(newrule);
1502 				PF_RULES_WUNLOCK();
1503 				break;
1504 			}
1505 
1506 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1507 			newrule->evaluations = 0;
1508 			newrule->packets[0] = newrule->packets[1] = 0;
1509 			newrule->bytes[0] = newrule->bytes[1] = 0;
1510 		}
1511 		pf_empty_pool(&V_pf_pabuf);
1512 
1513 		if (pcr->action == PF_CHANGE_ADD_HEAD)
1514 			oldrule = TAILQ_FIRST(
1515 			    ruleset->rules[rs_num].active.ptr);
1516 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1517 			oldrule = TAILQ_LAST(
1518 			    ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1519 		else {
1520 			oldrule = TAILQ_FIRST(
1521 			    ruleset->rules[rs_num].active.ptr);
1522 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1523 				oldrule = TAILQ_NEXT(oldrule, entries);
1524 			if (oldrule == NULL) {
1525 				if (newrule != NULL)
1526 					pf_free_rule(newrule);
1527 				PF_RULES_WUNLOCK();
1528 				error = EINVAL;
1529 				break;
1530 			}
1531 		}
1532 
1533 		if (pcr->action == PF_CHANGE_REMOVE) {
1534 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
1535 			    oldrule);
1536 			ruleset->rules[rs_num].active.rcount--;
1537 		} else {
1538 			if (oldrule == NULL)
1539 				TAILQ_INSERT_TAIL(
1540 				    ruleset->rules[rs_num].active.ptr,
1541 				    newrule, entries);
1542 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1543 			    pcr->action == PF_CHANGE_ADD_BEFORE)
1544 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1545 			else
1546 				TAILQ_INSERT_AFTER(
1547 				    ruleset->rules[rs_num].active.ptr,
1548 				    oldrule, newrule, entries);
1549 			ruleset->rules[rs_num].active.rcount++;
1550 		}
1551 
1552 		nr = 0;
1553 		TAILQ_FOREACH(oldrule,
1554 		    ruleset->rules[rs_num].active.ptr, entries)
1555 			oldrule->nr = nr++;
1556 
1557 		ruleset->rules[rs_num].active.ticket++;
1558 
1559 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1560 		pf_remove_if_empty_ruleset(ruleset);
1561 
1562 		PF_RULES_WUNLOCK();
1563 		break;
1564 
1565 #undef ERROUT
1566 DIOCCHANGERULE_error:
1567 		PF_RULES_WUNLOCK();
1568 		if (newrule != NULL)
1569 			free(newrule, M_PFRULE);
1570 		if (kif != NULL)
1571 			free(kif, PFI_MTYPE);
1572 		break;
1573 	}
1574 
1575 	case DIOCCLRSTATES: {
1576 		struct pf_state		*s;
1577 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1578 		u_int			 i, killed = 0;
1579 
1580 		for (i = 0; i <= V_pf_hashmask; i++) {
1581 			struct pf_idhash *ih = &V_pf_idhash[i];
1582 
1583 relock_DIOCCLRSTATES:
1584 			PF_HASHROW_LOCK(ih);
1585 			LIST_FOREACH(s, &ih->states, entry)
1586 				if (!psk->psk_ifname[0] ||
1587 				    !strcmp(psk->psk_ifname,
1588 				    s->kif->pfik_name)) {
1589 					/*
1590 					 * Don't send out individual
1591 					 * delete messages.
1592 					 */
1593 					s->state_flags |= PFSTATE_NOSYNC;
1594 					pf_unlink_state(s, PF_ENTER_LOCKED);
1595 					killed++;
1596 					goto relock_DIOCCLRSTATES;
1597 				}
1598 			PF_HASHROW_UNLOCK(ih);
1599 		}
1600 		psk->psk_killed = killed;
1601 		if (pfsync_clear_states_ptr != NULL)
1602 			pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
1603 		break;
1604 	}
1605 
1606 	case DIOCKILLSTATES: {
1607 		struct pf_state		*s;
1608 		struct pf_state_key	*sk;
1609 		struct pf_addr		*srcaddr, *dstaddr;
1610 		u_int16_t		 srcport, dstport;
1611 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
1612 		u_int			 i, killed = 0;
1613 
1614 		if (psk->psk_pfcmp.id) {
1615 			if (psk->psk_pfcmp.creatorid == 0)
1616 				psk->psk_pfcmp.creatorid = V_pf_status.hostid;
1617 			if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
1618 			    psk->psk_pfcmp.creatorid))) {
1619 				pf_unlink_state(s, PF_ENTER_LOCKED);
1620 				psk->psk_killed = 1;
1621 			}
1622 			break;
1623 		}
1624 
1625 		for (i = 0; i <= V_pf_hashmask; i++) {
1626 			struct pf_idhash *ih = &V_pf_idhash[i];
1627 
1628 relock_DIOCKILLSTATES:
1629 			PF_HASHROW_LOCK(ih);
1630 			LIST_FOREACH(s, &ih->states, entry) {
1631 				sk = s->key[PF_SK_WIRE];
1632 				if (s->direction == PF_OUT) {
1633 					srcaddr = &sk->addr[1];
1634 					dstaddr = &sk->addr[0];
1635 					srcport = sk->port[0];
1636 					dstport = sk->port[0];
1637 				} else {
1638 					srcaddr = &sk->addr[0];
1639 					dstaddr = &sk->addr[1];
1640 					srcport = sk->port[0];
1641 					dstport = sk->port[0];
1642 				}
1643 
1644 				if ((!psk->psk_af || sk->af == psk->psk_af)
1645 				    && (!psk->psk_proto || psk->psk_proto ==
1646 				    sk->proto) &&
1647 				    PF_MATCHA(psk->psk_src.neg,
1648 				    &psk->psk_src.addr.v.a.addr,
1649 				    &psk->psk_src.addr.v.a.mask,
1650 				    srcaddr, sk->af) &&
1651 				    PF_MATCHA(psk->psk_dst.neg,
1652 				    &psk->psk_dst.addr.v.a.addr,
1653 				    &psk->psk_dst.addr.v.a.mask,
1654 				    dstaddr, sk->af) &&
1655 				    (psk->psk_src.port_op == 0 ||
1656 				    pf_match_port(psk->psk_src.port_op,
1657 				    psk->psk_src.port[0], psk->psk_src.port[1],
1658 				    srcport)) &&
1659 				    (psk->psk_dst.port_op == 0 ||
1660 				    pf_match_port(psk->psk_dst.port_op,
1661 				    psk->psk_dst.port[0], psk->psk_dst.port[1],
1662 				    dstport)) &&
1663 				    (!psk->psk_label[0] ||
1664 				    (s->rule.ptr->label[0] &&
1665 				    !strcmp(psk->psk_label,
1666 				    s->rule.ptr->label))) &&
1667 				    (!psk->psk_ifname[0] ||
1668 				    !strcmp(psk->psk_ifname,
1669 				    s->kif->pfik_name))) {
1670 					pf_unlink_state(s, PF_ENTER_LOCKED);
1671 					killed++;
1672 					goto relock_DIOCKILLSTATES;
1673 				}
1674 			}
1675 			PF_HASHROW_UNLOCK(ih);
1676 		}
1677 		psk->psk_killed = killed;
1678 		break;
1679 	}
1680 
1681 	case DIOCADDSTATE: {
1682 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1683 		struct pfsync_state	*sp = &ps->state;
1684 
1685 		if (sp->timeout >= PFTM_MAX &&
1686 		    sp->timeout != PFTM_UNTIL_PACKET) {
1687 			error = EINVAL;
1688 			break;
1689 		}
1690 		if (pfsync_state_import_ptr != NULL) {
1691 			PF_RULES_RLOCK();
1692 			error = pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
1693 			PF_RULES_RUNLOCK();
1694 		} else
1695 			error = EOPNOTSUPP;
1696 		break;
1697 	}
1698 
1699 	case DIOCGETSTATE: {
1700 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1701 		struct pf_state		*s;
1702 
1703 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
1704 		if (s == NULL) {
1705 			error = ENOENT;
1706 			break;
1707 		}
1708 
1709 		pfsync_state_export(&ps->state, s);
1710 		PF_STATE_UNLOCK(s);
1711 		break;
1712 	}
1713 
1714 	case DIOCGETSTATES: {
1715 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
1716 		struct pf_state		*s;
1717 		struct pfsync_state	*pstore, *p;
1718 		int i, nr;
1719 
1720 		if (ps->ps_len == 0) {
1721 			nr = uma_zone_get_cur(V_pf_state_z);
1722 			ps->ps_len = sizeof(struct pfsync_state) * nr;
1723 			break;
1724 		}
1725 
1726 		p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK);
1727 		nr = 0;
1728 
1729 		for (i = 0; i <= V_pf_hashmask; i++) {
1730 			struct pf_idhash *ih = &V_pf_idhash[i];
1731 
1732 			PF_HASHROW_LOCK(ih);
1733 			LIST_FOREACH(s, &ih->states, entry) {
1734 
1735 				if (s->timeout == PFTM_UNLINKED)
1736 					continue;
1737 
1738 				if ((nr+1) * sizeof(*p) > ps->ps_len) {
1739 					PF_HASHROW_UNLOCK(ih);
1740 					goto DIOCGETSTATES_full;
1741 				}
1742 				pfsync_state_export(p, s);
1743 				p++;
1744 				nr++;
1745 			}
1746 			PF_HASHROW_UNLOCK(ih);
1747 		}
1748 DIOCGETSTATES_full:
1749 		error = copyout(pstore, ps->ps_states,
1750 		    sizeof(struct pfsync_state) * nr);
1751 		if (error) {
1752 			free(pstore, M_TEMP);
1753 			break;
1754 		}
1755 		ps->ps_len = sizeof(struct pfsync_state) * nr;
1756 		free(pstore, M_TEMP);
1757 
1758 		break;
1759 	}
1760 
1761 	case DIOCGETSTATUS: {
1762 		struct pf_status *s = (struct pf_status *)addr;
1763 		PF_RULES_RLOCK();
1764 		bcopy(&V_pf_status, s, sizeof(struct pf_status));
1765 		pfi_update_status(s->ifname, s);
1766 		PF_RULES_RUNLOCK();
1767 		break;
1768 	}
1769 
1770 	case DIOCSETSTATUSIF: {
1771 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
1772 
1773 		if (pi->ifname[0] == 0) {
1774 			bzero(V_pf_status.ifname, IFNAMSIZ);
1775 			break;
1776 		}
1777 		PF_RULES_WLOCK();
1778 		strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
1779 		PF_RULES_WUNLOCK();
1780 		break;
1781 	}
1782 
1783 	case DIOCCLRSTATUS: {
1784 		PF_RULES_WLOCK();
1785 		bzero(V_pf_status.counters, sizeof(V_pf_status.counters));
1786 		bzero(V_pf_status.fcounters, sizeof(V_pf_status.fcounters));
1787 		bzero(V_pf_status.scounters, sizeof(V_pf_status.scounters));
1788 		V_pf_status.since = time_second;
1789 		if (*V_pf_status.ifname)
1790 			pfi_update_status(V_pf_status.ifname, NULL);
1791 		PF_RULES_WUNLOCK();
1792 		break;
1793 	}
1794 
1795 	case DIOCNATLOOK: {
1796 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
1797 		struct pf_state_key	*sk;
1798 		struct pf_state		*state;
1799 		struct pf_state_key_cmp	 key;
1800 		int			 m = 0, direction = pnl->direction;
1801 		int			 sidx, didx;
1802 
1803 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
1804 		sidx = (direction == PF_IN) ? 1 : 0;
1805 		didx = (direction == PF_IN) ? 0 : 1;
1806 
1807 		if (!pnl->proto ||
1808 		    PF_AZERO(&pnl->saddr, pnl->af) ||
1809 		    PF_AZERO(&pnl->daddr, pnl->af) ||
1810 		    ((pnl->proto == IPPROTO_TCP ||
1811 		    pnl->proto == IPPROTO_UDP) &&
1812 		    (!pnl->dport || !pnl->sport)))
1813 			error = EINVAL;
1814 		else {
1815 			bzero(&key, sizeof(key));
1816 			key.af = pnl->af;
1817 			key.proto = pnl->proto;
1818 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
1819 			key.port[sidx] = pnl->sport;
1820 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
1821 			key.port[didx] = pnl->dport;
1822 
1823 			state = pf_find_state_all(&key, direction, &m);
1824 
1825 			if (m > 1)
1826 				error = E2BIG;	/* more than one state */
1827 			else if (state != NULL) {
1828 				/* XXXGL: not locked read */
1829 				sk = state->key[sidx];
1830 				PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
1831 				pnl->rsport = sk->port[sidx];
1832 				PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
1833 				pnl->rdport = sk->port[didx];
1834 			} else
1835 				error = ENOENT;
1836 		}
1837 		break;
1838 	}
1839 
1840 	case DIOCSETTIMEOUT: {
1841 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1842 		int		 old;
1843 
1844 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
1845 		    pt->seconds < 0) {
1846 			error = EINVAL;
1847 			break;
1848 		}
1849 		PF_RULES_WLOCK();
1850 		old = V_pf_default_rule.timeout[pt->timeout];
1851 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
1852 			pt->seconds = 1;
1853 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
1854 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
1855 			wakeup(pf_purge_thread);
1856 		pt->seconds = old;
1857 		PF_RULES_WUNLOCK();
1858 		break;
1859 	}
1860 
1861 	case DIOCGETTIMEOUT: {
1862 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1863 
1864 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
1865 			error = EINVAL;
1866 			break;
1867 		}
1868 		PF_RULES_RLOCK();
1869 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
1870 		PF_RULES_RUNLOCK();
1871 		break;
1872 	}
1873 
1874 	case DIOCGETLIMIT: {
1875 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1876 
1877 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
1878 			error = EINVAL;
1879 			break;
1880 		}
1881 		PF_RULES_RLOCK();
1882 		pl->limit = V_pf_limits[pl->index].limit;
1883 		PF_RULES_RUNLOCK();
1884 		break;
1885 	}
1886 
1887 	case DIOCSETLIMIT: {
1888 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1889 		int			 old_limit;
1890 
1891 		PF_RULES_WLOCK();
1892 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
1893 		    V_pf_limits[pl->index].zone == NULL) {
1894 			PF_RULES_WUNLOCK();
1895 			error = EINVAL;
1896 			break;
1897 		}
1898 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
1899 		old_limit = V_pf_limits[pl->index].limit;
1900 		V_pf_limits[pl->index].limit = pl->limit;
1901 		pl->limit = old_limit;
1902 		PF_RULES_WUNLOCK();
1903 		break;
1904 	}
1905 
1906 	case DIOCSETDEBUG: {
1907 		u_int32_t	*level = (u_int32_t *)addr;
1908 
1909 		PF_RULES_WLOCK();
1910 		V_pf_status.debug = *level;
1911 		PF_RULES_WUNLOCK();
1912 		break;
1913 	}
1914 
1915 	case DIOCCLRRULECTRS: {
1916 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
1917 		struct pf_ruleset	*ruleset = &pf_main_ruleset;
1918 		struct pf_rule		*rule;
1919 
1920 		PF_RULES_WLOCK();
1921 		TAILQ_FOREACH(rule,
1922 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
1923 			rule->evaluations = 0;
1924 			rule->packets[0] = rule->packets[1] = 0;
1925 			rule->bytes[0] = rule->bytes[1] = 0;
1926 		}
1927 		PF_RULES_WUNLOCK();
1928 		break;
1929 	}
1930 
1931 	case DIOCGIFSPEED: {
1932 		struct pf_ifspeed	*psp = (struct pf_ifspeed *)addr;
1933 		struct pf_ifspeed	ps;
1934 		struct ifnet		*ifp;
1935 
1936 		if (psp->ifname[0] != 0) {
1937 			/* Can we completely trust user-land? */
1938 			strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
1939 			ifp = ifunit(ps.ifname);
1940 			if (ifp != NULL)
1941 				psp->baudrate = ifp->if_baudrate;
1942 			else
1943 				error = EINVAL;
1944 		} else
1945 			error = EINVAL;
1946 		break;
1947 	}
1948 
1949 #ifdef ALTQ
1950 	case DIOCSTARTALTQ: {
1951 		struct pf_altq		*altq;
1952 
1953 		PF_RULES_WLOCK();
1954 		/* enable all altq interfaces on active list */
1955 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1956 			if (altq->qname[0] == 0 && (altq->local_flags &
1957 			    PFALTQ_FLAG_IF_REMOVED) == 0) {
1958 				error = pf_enable_altq(altq);
1959 				if (error != 0)
1960 					break;
1961 			}
1962 		}
1963 		if (error == 0)
1964 			V_pf_altq_running = 1;
1965 		PF_RULES_WUNLOCK();
1966 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
1967 		break;
1968 	}
1969 
1970 	case DIOCSTOPALTQ: {
1971 		struct pf_altq		*altq;
1972 
1973 		PF_RULES_WLOCK();
1974 		/* disable all altq interfaces on active list */
1975 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1976 			if (altq->qname[0] == 0 && (altq->local_flags &
1977 			    PFALTQ_FLAG_IF_REMOVED) == 0) {
1978 				error = pf_disable_altq(altq);
1979 				if (error != 0)
1980 					break;
1981 			}
1982 		}
1983 		if (error == 0)
1984 			V_pf_altq_running = 0;
1985 		PF_RULES_WUNLOCK();
1986 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
1987 		break;
1988 	}
1989 
1990 	case DIOCADDALTQ: {
1991 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
1992 		struct pf_altq		*altq, *a;
1993 		struct ifnet		*ifp;
1994 
1995 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK);
1996 		bcopy(&pa->altq, altq, sizeof(struct pf_altq));
1997 		altq->local_flags = 0;
1998 
1999 		PF_RULES_WLOCK();
2000 		if (pa->ticket != V_ticket_altqs_inactive) {
2001 			PF_RULES_WUNLOCK();
2002 			free(altq, M_PFALTQ);
2003 			error = EBUSY;
2004 			break;
2005 		}
2006 
2007 		/*
2008 		 * if this is for a queue, find the discipline and
2009 		 * copy the necessary fields
2010 		 */
2011 		if (altq->qname[0] != 0) {
2012 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2013 				PF_RULES_WUNLOCK();
2014 				error = EBUSY;
2015 				free(altq, M_PFALTQ);
2016 				break;
2017 			}
2018 			altq->altq_disc = NULL;
2019 			TAILQ_FOREACH(a, V_pf_altqs_inactive, entries) {
2020 				if (strncmp(a->ifname, altq->ifname,
2021 				    IFNAMSIZ) == 0 && a->qname[0] == 0) {
2022 					altq->altq_disc = a->altq_disc;
2023 					break;
2024 				}
2025 			}
2026 		}
2027 
2028 		if ((ifp = ifunit(altq->ifname)) == NULL)
2029 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
2030 		else
2031 			error = altq_add(altq);
2032 
2033 		if (error) {
2034 			PF_RULES_WUNLOCK();
2035 			free(altq, M_PFALTQ);
2036 			break;
2037 		}
2038 
2039 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
2040 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2041 		PF_RULES_WUNLOCK();
2042 		break;
2043 	}
2044 
2045 	case DIOCGETALTQS: {
2046 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2047 		struct pf_altq		*altq;
2048 
2049 		PF_RULES_RLOCK();
2050 		pa->nr = 0;
2051 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
2052 			pa->nr++;
2053 		pa->ticket = V_ticket_altqs_active;
2054 		PF_RULES_RUNLOCK();
2055 		break;
2056 	}
2057 
2058 	case DIOCGETALTQ: {
2059 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2060 		struct pf_altq		*altq;
2061 		u_int32_t		 nr;
2062 
2063 		PF_RULES_RLOCK();
2064 		if (pa->ticket != V_ticket_altqs_active) {
2065 			PF_RULES_RUNLOCK();
2066 			error = EBUSY;
2067 			break;
2068 		}
2069 		nr = 0;
2070 		altq = TAILQ_FIRST(V_pf_altqs_active);
2071 		while ((altq != NULL) && (nr < pa->nr)) {
2072 			altq = TAILQ_NEXT(altq, entries);
2073 			nr++;
2074 		}
2075 		if (altq == NULL) {
2076 			PF_RULES_RUNLOCK();
2077 			error = EBUSY;
2078 			break;
2079 		}
2080 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2081 		PF_RULES_RUNLOCK();
2082 		break;
2083 	}
2084 
2085 	case DIOCCHANGEALTQ:
2086 		/* CHANGEALTQ not supported yet! */
2087 		error = ENODEV;
2088 		break;
2089 
2090 	case DIOCGETQSTATS: {
2091 		struct pfioc_qstats	*pq = (struct pfioc_qstats *)addr;
2092 		struct pf_altq		*altq;
2093 		u_int32_t		 nr;
2094 		int			 nbytes;
2095 
2096 		PF_RULES_RLOCK();
2097 		if (pq->ticket != V_ticket_altqs_active) {
2098 			PF_RULES_RUNLOCK();
2099 			error = EBUSY;
2100 			break;
2101 		}
2102 		nbytes = pq->nbytes;
2103 		nr = 0;
2104 		altq = TAILQ_FIRST(V_pf_altqs_active);
2105 		while ((altq != NULL) && (nr < pq->nr)) {
2106 			altq = TAILQ_NEXT(altq, entries);
2107 			nr++;
2108 		}
2109 		if (altq == NULL) {
2110 			PF_RULES_RUNLOCK();
2111 			error = EBUSY;
2112 			break;
2113 		}
2114 
2115 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
2116 			PF_RULES_RUNLOCK();
2117 			error = ENXIO;
2118 			break;
2119 		}
2120 		PF_RULES_RUNLOCK();
2121 		error = altq_getqstats(altq, pq->buf, &nbytes);
2122 		if (error == 0) {
2123 			pq->scheduler = altq->scheduler;
2124 			pq->nbytes = nbytes;
2125 		}
2126 		break;
2127 	}
2128 #endif /* ALTQ */
2129 
2130 	case DIOCBEGINADDRS: {
2131 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2132 
2133 		PF_RULES_WLOCK();
2134 		pf_empty_pool(&V_pf_pabuf);
2135 		pp->ticket = ++V_ticket_pabuf;
2136 		PF_RULES_WUNLOCK();
2137 		break;
2138 	}
2139 
2140 	case DIOCADDADDR: {
2141 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2142 		struct pf_pooladdr	*pa;
2143 		struct pfi_kif		*kif = NULL;
2144 
2145 #ifndef INET
2146 		if (pp->af == AF_INET) {
2147 			error = EAFNOSUPPORT;
2148 			break;
2149 		}
2150 #endif /* INET */
2151 #ifndef INET6
2152 		if (pp->af == AF_INET6) {
2153 			error = EAFNOSUPPORT;
2154 			break;
2155 		}
2156 #endif /* INET6 */
2157 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2158 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2159 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2160 			error = EINVAL;
2161 			break;
2162 		}
2163 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2164 		bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2165 		if (pa->ifname[0])
2166 			kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2167 		PF_RULES_WLOCK();
2168 		if (pp->ticket != V_ticket_pabuf) {
2169 			PF_RULES_WUNLOCK();
2170 			if (pa->ifname[0])
2171 				free(kif, PFI_MTYPE);
2172 			free(pa, M_PFRULE);
2173 			error = EBUSY;
2174 			break;
2175 		}
2176 		if (pa->ifname[0]) {
2177 			pa->kif = pfi_kif_attach(kif, pa->ifname);
2178 			pfi_kif_ref(pa->kif);
2179 		} else
2180 			pa->kif = NULL;
2181 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2182 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2183 			if (pa->ifname[0])
2184 				pfi_kif_unref(pa->kif);
2185 			PF_RULES_WUNLOCK();
2186 			free(pa, M_PFRULE);
2187 			break;
2188 		}
2189 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2190 		PF_RULES_WUNLOCK();
2191 		break;
2192 	}
2193 
2194 	case DIOCGETADDRS: {
2195 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2196 		struct pf_pool		*pool;
2197 		struct pf_pooladdr	*pa;
2198 
2199 		PF_RULES_RLOCK();
2200 		pp->nr = 0;
2201 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2202 		    pp->r_num, 0, 1, 0);
2203 		if (pool == NULL) {
2204 			PF_RULES_RUNLOCK();
2205 			error = EBUSY;
2206 			break;
2207 		}
2208 		TAILQ_FOREACH(pa, &pool->list, entries)
2209 			pp->nr++;
2210 		PF_RULES_RUNLOCK();
2211 		break;
2212 	}
2213 
2214 	case DIOCGETADDR: {
2215 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2216 		struct pf_pool		*pool;
2217 		struct pf_pooladdr	*pa;
2218 		u_int32_t		 nr = 0;
2219 
2220 		PF_RULES_RLOCK();
2221 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2222 		    pp->r_num, 0, 1, 1);
2223 		if (pool == NULL) {
2224 			PF_RULES_RUNLOCK();
2225 			error = EBUSY;
2226 			break;
2227 		}
2228 		pa = TAILQ_FIRST(&pool->list);
2229 		while ((pa != NULL) && (nr < pp->nr)) {
2230 			pa = TAILQ_NEXT(pa, entries);
2231 			nr++;
2232 		}
2233 		if (pa == NULL) {
2234 			PF_RULES_RUNLOCK();
2235 			error = EBUSY;
2236 			break;
2237 		}
2238 		bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2239 		pf_addr_copyout(&pp->addr.addr);
2240 		PF_RULES_RUNLOCK();
2241 		break;
2242 	}
2243 
2244 	case DIOCCHANGEADDR: {
2245 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
2246 		struct pf_pool		*pool;
2247 		struct pf_pooladdr	*oldpa = NULL, *newpa = NULL;
2248 		struct pf_ruleset	*ruleset;
2249 		struct pfi_kif		*kif = NULL;
2250 
2251 		if (pca->action < PF_CHANGE_ADD_HEAD ||
2252 		    pca->action > PF_CHANGE_REMOVE) {
2253 			error = EINVAL;
2254 			break;
2255 		}
2256 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2257 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2258 		    pca->addr.addr.type != PF_ADDR_TABLE) {
2259 			error = EINVAL;
2260 			break;
2261 		}
2262 
2263 		if (pca->action != PF_CHANGE_REMOVE) {
2264 #ifndef INET
2265 			if (pca->af == AF_INET) {
2266 				error = EAFNOSUPPORT;
2267 				break;
2268 			}
2269 #endif /* INET */
2270 #ifndef INET6
2271 			if (pca->af == AF_INET6) {
2272 				error = EAFNOSUPPORT;
2273 				break;
2274 			}
2275 #endif /* INET6 */
2276 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
2277 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2278 			if (newpa->ifname[0])
2279 				kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2280 		}
2281 
2282 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGEADDR_error; }
2283 		PF_RULES_WLOCK();
2284 		ruleset = pf_find_ruleset(pca->anchor);
2285 		if (ruleset == NULL)
2286 			ERROUT(EBUSY);
2287 
2288 		pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2289 		    pca->r_num, pca->r_last, 1, 1);
2290 		if (pool == NULL)
2291 			ERROUT(EBUSY);
2292 
2293 		if (pca->action != PF_CHANGE_REMOVE) {
2294 			if (newpa->ifname[0]) {
2295 				newpa->kif = pfi_kif_attach(kif, newpa->ifname);
2296 				pfi_kif_ref(newpa->kif);
2297 			} else
2298 				newpa->kif = NULL;
2299 
2300 			switch (newpa->addr.type) {
2301 			case PF_ADDR_DYNIFTL:
2302 				error = pfi_dynaddr_setup(&newpa->addr,
2303 				    pca->af);
2304 				break;
2305 			case PF_ADDR_TABLE:
2306 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
2307 				    newpa->addr.v.tblname);
2308 				if (newpa->addr.p.tbl == NULL)
2309 					error = ENOMEM;
2310 				break;
2311 			}
2312 			if (error) {
2313 				if (newpa->kif)
2314 					pfi_kif_unref(newpa->kif);
2315 				PF_RULES_WUNLOCK();
2316 				free(newpa, M_PFRULE);
2317 				break;
2318 			}
2319 		}
2320 
2321 		if (pca->action == PF_CHANGE_ADD_HEAD)
2322 			oldpa = TAILQ_FIRST(&pool->list);
2323 		else if (pca->action == PF_CHANGE_ADD_TAIL)
2324 			oldpa = TAILQ_LAST(&pool->list, pf_palist);
2325 		else {
2326 			int	i = 0;
2327 
2328 			oldpa = TAILQ_FIRST(&pool->list);
2329 			while ((oldpa != NULL) && (i < pca->nr)) {
2330 				oldpa = TAILQ_NEXT(oldpa, entries);
2331 				i++;
2332 			}
2333 			if (oldpa == NULL) {
2334 				PF_RULES_WUNLOCK();
2335 				error = EINVAL;
2336 				break;
2337 			}
2338 		}
2339 
2340 		if (pca->action == PF_CHANGE_REMOVE) {
2341 			TAILQ_REMOVE(&pool->list, oldpa, entries);
2342 			switch (oldpa->addr.type) {
2343 			case PF_ADDR_DYNIFTL:
2344 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
2345 				break;
2346 			case PF_ADDR_TABLE:
2347 				pfr_detach_table(oldpa->addr.p.tbl);
2348 				break;
2349 			}
2350 			if (oldpa->kif)
2351 				pfi_kif_unref(oldpa->kif);
2352 			free(oldpa, M_PFRULE);
2353 		} else {
2354 			if (oldpa == NULL)
2355 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2356 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
2357 			    pca->action == PF_CHANGE_ADD_BEFORE)
2358 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2359 			else
2360 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
2361 				    newpa, entries);
2362 		}
2363 
2364 		pool->cur = TAILQ_FIRST(&pool->list);
2365 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr,
2366 		    pca->af);
2367 		PF_RULES_WUNLOCK();
2368 		break;
2369 
2370 #undef ERROUT
2371 DIOCCHANGEADDR_error:
2372 		PF_RULES_WUNLOCK();
2373 		if (newpa != NULL)
2374 			free(newpa, M_PFRULE);
2375 		if (kif != NULL)
2376 			free(kif, PFI_MTYPE);
2377 		break;
2378 	}
2379 
2380 	case DIOCGETRULESETS: {
2381 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2382 		struct pf_ruleset	*ruleset;
2383 		struct pf_anchor	*anchor;
2384 
2385 		PF_RULES_RLOCK();
2386 		pr->path[sizeof(pr->path) - 1] = 0;
2387 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2388 			PF_RULES_RUNLOCK();
2389 			error = ENOENT;
2390 			break;
2391 		}
2392 		pr->nr = 0;
2393 		if (ruleset->anchor == NULL) {
2394 			/* XXX kludge for pf_main_ruleset */
2395 			RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2396 				if (anchor->parent == NULL)
2397 					pr->nr++;
2398 		} else {
2399 			RB_FOREACH(anchor, pf_anchor_node,
2400 			    &ruleset->anchor->children)
2401 				pr->nr++;
2402 		}
2403 		PF_RULES_RUNLOCK();
2404 		break;
2405 	}
2406 
2407 	case DIOCGETRULESET: {
2408 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2409 		struct pf_ruleset	*ruleset;
2410 		struct pf_anchor	*anchor;
2411 		u_int32_t		 nr = 0;
2412 
2413 		PF_RULES_RLOCK();
2414 		pr->path[sizeof(pr->path) - 1] = 0;
2415 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2416 			PF_RULES_RUNLOCK();
2417 			error = ENOENT;
2418 			break;
2419 		}
2420 		pr->name[0] = 0;
2421 		if (ruleset->anchor == NULL) {
2422 			/* XXX kludge for pf_main_ruleset */
2423 			RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2424 				if (anchor->parent == NULL && nr++ == pr->nr) {
2425 					strlcpy(pr->name, anchor->name,
2426 					    sizeof(pr->name));
2427 					break;
2428 				}
2429 		} else {
2430 			RB_FOREACH(anchor, pf_anchor_node,
2431 			    &ruleset->anchor->children)
2432 				if (nr++ == pr->nr) {
2433 					strlcpy(pr->name, anchor->name,
2434 					    sizeof(pr->name));
2435 					break;
2436 				}
2437 		}
2438 		if (!pr->name[0])
2439 			error = EBUSY;
2440 		PF_RULES_RUNLOCK();
2441 		break;
2442 	}
2443 
2444 	case DIOCRCLRTABLES: {
2445 		struct pfioc_table *io = (struct pfioc_table *)addr;
2446 
2447 		if (io->pfrio_esize != 0) {
2448 			error = ENODEV;
2449 			break;
2450 		}
2451 		PF_RULES_WLOCK();
2452 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2453 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2454 		PF_RULES_WUNLOCK();
2455 		break;
2456 	}
2457 
2458 	case DIOCRADDTABLES: {
2459 		struct pfioc_table *io = (struct pfioc_table *)addr;
2460 		struct pfr_table *pfrts;
2461 		size_t totlen;
2462 
2463 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2464 			error = ENODEV;
2465 			break;
2466 		}
2467 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2468 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2469 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2470 		if (error) {
2471 			free(pfrts, M_TEMP);
2472 			break;
2473 		}
2474 		PF_RULES_WLOCK();
2475 		error = pfr_add_tables(pfrts, io->pfrio_size,
2476 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2477 		PF_RULES_WUNLOCK();
2478 		free(pfrts, M_TEMP);
2479 		break;
2480 	}
2481 
2482 	case DIOCRDELTABLES: {
2483 		struct pfioc_table *io = (struct pfioc_table *)addr;
2484 		struct pfr_table *pfrts;
2485 		size_t totlen;
2486 
2487 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2488 			error = ENODEV;
2489 			break;
2490 		}
2491 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2492 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2493 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2494 		if (error) {
2495 			free(pfrts, M_TEMP);
2496 			break;
2497 		}
2498 		PF_RULES_WLOCK();
2499 		error = pfr_del_tables(pfrts, io->pfrio_size,
2500 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2501 		PF_RULES_WUNLOCK();
2502 		free(pfrts, M_TEMP);
2503 		break;
2504 	}
2505 
2506 	case DIOCRGETTABLES: {
2507 		struct pfioc_table *io = (struct pfioc_table *)addr;
2508 		struct pfr_table *pfrts;
2509 		size_t totlen;
2510 
2511 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2512 			error = ENODEV;
2513 			break;
2514 		}
2515 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2516 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2517 		PF_RULES_RLOCK();
2518 		error = pfr_get_tables(&io->pfrio_table, pfrts,
2519 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2520 		PF_RULES_RUNLOCK();
2521 		if (error == 0)
2522 			error = copyout(pfrts, io->pfrio_buffer, totlen);
2523 		free(pfrts, M_TEMP);
2524 		break;
2525 	}
2526 
2527 	case DIOCRGETTSTATS: {
2528 		struct pfioc_table *io = (struct pfioc_table *)addr;
2529 		struct pfr_tstats *pfrtstats;
2530 		size_t totlen;
2531 
2532 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2533 			error = ENODEV;
2534 			break;
2535 		}
2536 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
2537 		pfrtstats = malloc(totlen, M_TEMP, M_WAITOK);
2538 		PF_RULES_WLOCK();
2539 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
2540 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2541 		PF_RULES_WUNLOCK();
2542 		if (error == 0)
2543 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
2544 		free(pfrtstats, M_TEMP);
2545 		break;
2546 	}
2547 
2548 	case DIOCRCLRTSTATS: {
2549 		struct pfioc_table *io = (struct pfioc_table *)addr;
2550 		struct pfr_table *pfrts;
2551 		size_t totlen;
2552 
2553 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2554 			error = ENODEV;
2555 			break;
2556 		}
2557 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2558 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2559 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2560 		if (error) {
2561 			free(pfrts, M_TEMP);
2562 			break;
2563 		}
2564 		PF_RULES_WLOCK();
2565 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
2566 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2567 		PF_RULES_WUNLOCK();
2568 		free(pfrts, M_TEMP);
2569 		break;
2570 	}
2571 
2572 	case DIOCRSETTFLAGS: {
2573 		struct pfioc_table *io = (struct pfioc_table *)addr;
2574 		struct pfr_table *pfrts;
2575 		size_t totlen;
2576 
2577 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2578 			error = ENODEV;
2579 			break;
2580 		}
2581 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2582 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2583 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2584 		if (error) {
2585 			free(pfrts, M_TEMP);
2586 			break;
2587 		}
2588 		PF_RULES_WLOCK();
2589 		error = pfr_set_tflags(pfrts, io->pfrio_size,
2590 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2591 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2592 		PF_RULES_WUNLOCK();
2593 		free(pfrts, M_TEMP);
2594 		break;
2595 	}
2596 
2597 	case DIOCRCLRADDRS: {
2598 		struct pfioc_table *io = (struct pfioc_table *)addr;
2599 
2600 		if (io->pfrio_esize != 0) {
2601 			error = ENODEV;
2602 			break;
2603 		}
2604 		PF_RULES_WLOCK();
2605 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2606 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2607 		PF_RULES_WUNLOCK();
2608 		break;
2609 	}
2610 
2611 	case DIOCRADDADDRS: {
2612 		struct pfioc_table *io = (struct pfioc_table *)addr;
2613 		struct pfr_addr *pfras;
2614 		size_t totlen;
2615 
2616 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2617 			error = ENODEV;
2618 			break;
2619 		}
2620 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2621 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2622 		error = copyin(io->pfrio_buffer, pfras, totlen);
2623 		if (error) {
2624 			free(pfras, M_TEMP);
2625 			break;
2626 		}
2627 		PF_RULES_WLOCK();
2628 		error = pfr_add_addrs(&io->pfrio_table, pfras,
2629 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2630 		    PFR_FLAG_USERIOCTL);
2631 		PF_RULES_WUNLOCK();
2632 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2633 			error = copyout(pfras, io->pfrio_buffer, totlen);
2634 		free(pfras, M_TEMP);
2635 		break;
2636 	}
2637 
2638 	case DIOCRDELADDRS: {
2639 		struct pfioc_table *io = (struct pfioc_table *)addr;
2640 		struct pfr_addr *pfras;
2641 		size_t totlen;
2642 
2643 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2644 			error = ENODEV;
2645 			break;
2646 		}
2647 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2648 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2649 		error = copyin(io->pfrio_buffer, pfras, totlen);
2650 		if (error) {
2651 			free(pfras, M_TEMP);
2652 			break;
2653 		}
2654 		PF_RULES_WLOCK();
2655 		error = pfr_del_addrs(&io->pfrio_table, pfras,
2656 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2657 		    PFR_FLAG_USERIOCTL);
2658 		PF_RULES_WUNLOCK();
2659 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2660 			error = copyout(pfras, io->pfrio_buffer, totlen);
2661 		free(pfras, M_TEMP);
2662 		break;
2663 	}
2664 
2665 	case DIOCRSETADDRS: {
2666 		struct pfioc_table *io = (struct pfioc_table *)addr;
2667 		struct pfr_addr *pfras;
2668 		size_t totlen;
2669 
2670 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2671 			error = ENODEV;
2672 			break;
2673 		}
2674 		totlen = (io->pfrio_size + io->pfrio_size2) *
2675 		    sizeof(struct pfr_addr);
2676 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2677 		error = copyin(io->pfrio_buffer, pfras, totlen);
2678 		if (error) {
2679 			free(pfras, M_TEMP);
2680 			break;
2681 		}
2682 		PF_RULES_WLOCK();
2683 		error = pfr_set_addrs(&io->pfrio_table, pfras,
2684 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2685 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2686 		    PFR_FLAG_USERIOCTL, 0);
2687 		PF_RULES_WUNLOCK();
2688 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2689 			error = copyout(pfras, io->pfrio_buffer, totlen);
2690 		free(pfras, M_TEMP);
2691 		break;
2692 	}
2693 
2694 	case DIOCRGETADDRS: {
2695 		struct pfioc_table *io = (struct pfioc_table *)addr;
2696 		struct pfr_addr *pfras;
2697 		size_t totlen;
2698 
2699 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2700 			error = ENODEV;
2701 			break;
2702 		}
2703 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2704 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2705 		PF_RULES_RLOCK();
2706 		error = pfr_get_addrs(&io->pfrio_table, pfras,
2707 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2708 		PF_RULES_RUNLOCK();
2709 		if (error == 0)
2710 			error = copyout(pfras, io->pfrio_buffer, totlen);
2711 		free(pfras, M_TEMP);
2712 		break;
2713 	}
2714 
2715 	case DIOCRGETASTATS: {
2716 		struct pfioc_table *io = (struct pfioc_table *)addr;
2717 		struct pfr_astats *pfrastats;
2718 		size_t totlen;
2719 
2720 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2721 			error = ENODEV;
2722 			break;
2723 		}
2724 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
2725 		pfrastats = malloc(totlen, M_TEMP, M_WAITOK);
2726 		PF_RULES_RLOCK();
2727 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
2728 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2729 		PF_RULES_RUNLOCK();
2730 		if (error == 0)
2731 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
2732 		free(pfrastats, M_TEMP);
2733 		break;
2734 	}
2735 
2736 	case DIOCRCLRASTATS: {
2737 		struct pfioc_table *io = (struct pfioc_table *)addr;
2738 		struct pfr_addr *pfras;
2739 		size_t totlen;
2740 
2741 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2742 			error = ENODEV;
2743 			break;
2744 		}
2745 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2746 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2747 		error = copyin(io->pfrio_buffer, pfras, totlen);
2748 		if (error) {
2749 			free(pfras, M_TEMP);
2750 			break;
2751 		}
2752 		PF_RULES_WLOCK();
2753 		error = pfr_clr_astats(&io->pfrio_table, pfras,
2754 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2755 		    PFR_FLAG_USERIOCTL);
2756 		PF_RULES_WUNLOCK();
2757 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2758 			error = copyout(pfras, io->pfrio_buffer, totlen);
2759 		free(pfras, M_TEMP);
2760 		break;
2761 	}
2762 
2763 	case DIOCRTSTADDRS: {
2764 		struct pfioc_table *io = (struct pfioc_table *)addr;
2765 		struct pfr_addr *pfras;
2766 		size_t totlen;
2767 
2768 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2769 			error = ENODEV;
2770 			break;
2771 		}
2772 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2773 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2774 		error = copyin(io->pfrio_buffer, pfras, totlen);
2775 		if (error) {
2776 			free(pfras, M_TEMP);
2777 			break;
2778 		}
2779 		PF_RULES_RLOCK();
2780 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
2781 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2782 		    PFR_FLAG_USERIOCTL);
2783 		PF_RULES_RUNLOCK();
2784 		if (error == 0)
2785 			error = copyout(pfras, io->pfrio_buffer, totlen);
2786 		free(pfras, M_TEMP);
2787 		break;
2788 	}
2789 
2790 	case DIOCRINADEFINE: {
2791 		struct pfioc_table *io = (struct pfioc_table *)addr;
2792 		struct pfr_addr *pfras;
2793 		size_t totlen;
2794 
2795 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2796 			error = ENODEV;
2797 			break;
2798 		}
2799 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2800 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2801 		error = copyin(io->pfrio_buffer, pfras, totlen);
2802 		if (error) {
2803 			free(pfras, M_TEMP);
2804 			break;
2805 		}
2806 		PF_RULES_WLOCK();
2807 		error = pfr_ina_define(&io->pfrio_table, pfras,
2808 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2809 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2810 		PF_RULES_WUNLOCK();
2811 		free(pfras, M_TEMP);
2812 		break;
2813 	}
2814 
2815 	case DIOCOSFPADD: {
2816 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2817 		PF_RULES_WLOCK();
2818 		error = pf_osfp_add(io);
2819 		PF_RULES_WUNLOCK();
2820 		break;
2821 	}
2822 
2823 	case DIOCOSFPGET: {
2824 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2825 		PF_RULES_RLOCK();
2826 		error = pf_osfp_get(io);
2827 		PF_RULES_RUNLOCK();
2828 		break;
2829 	}
2830 
2831 	case DIOCXBEGIN: {
2832 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2833 		struct pfioc_trans_e	*ioes, *ioe;
2834 		size_t			 totlen;
2835 		int			 i;
2836 
2837 		if (io->esize != sizeof(*ioe)) {
2838 			error = ENODEV;
2839 			break;
2840 		}
2841 		totlen = sizeof(struct pfioc_trans_e) * io->size;
2842 		ioes = malloc(totlen, M_TEMP, M_WAITOK);
2843 		error = copyin(io->array, ioes, totlen);
2844 		if (error) {
2845 			free(ioes, M_TEMP);
2846 			break;
2847 		}
2848 		PF_RULES_WLOCK();
2849 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2850 			switch (ioe->rs_num) {
2851 #ifdef ALTQ
2852 			case PF_RULESET_ALTQ:
2853 				if (ioe->anchor[0]) {
2854 					PF_RULES_WUNLOCK();
2855 					free(ioes, M_TEMP);
2856 					error = EINVAL;
2857 					goto fail;
2858 				}
2859 				if ((error = pf_begin_altq(&ioe->ticket))) {
2860 					PF_RULES_WUNLOCK();
2861 					free(ioes, M_TEMP);
2862 					goto fail;
2863 				}
2864 				break;
2865 #endif /* ALTQ */
2866 			case PF_RULESET_TABLE:
2867 			    {
2868 				struct pfr_table table;
2869 
2870 				bzero(&table, sizeof(table));
2871 				strlcpy(table.pfrt_anchor, ioe->anchor,
2872 				    sizeof(table.pfrt_anchor));
2873 				if ((error = pfr_ina_begin(&table,
2874 				    &ioe->ticket, NULL, 0))) {
2875 					PF_RULES_WUNLOCK();
2876 					free(ioes, M_TEMP);
2877 					goto fail;
2878 				}
2879 				break;
2880 			    }
2881 			default:
2882 				if ((error = pf_begin_rules(&ioe->ticket,
2883 				    ioe->rs_num, ioe->anchor))) {
2884 					PF_RULES_WUNLOCK();
2885 					free(ioes, M_TEMP);
2886 					goto fail;
2887 				}
2888 				break;
2889 			}
2890 		}
2891 		PF_RULES_WUNLOCK();
2892 		error = copyout(ioes, io->array, totlen);
2893 		free(ioes, M_TEMP);
2894 		break;
2895 	}
2896 
2897 	case DIOCXROLLBACK: {
2898 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2899 		struct pfioc_trans_e	*ioe, *ioes;
2900 		size_t			 totlen;
2901 		int			 i;
2902 
2903 		if (io->esize != sizeof(*ioe)) {
2904 			error = ENODEV;
2905 			break;
2906 		}
2907 		totlen = sizeof(struct pfioc_trans_e) * io->size;
2908 		ioes = malloc(totlen, M_TEMP, M_WAITOK);
2909 		error = copyin(io->array, ioes, totlen);
2910 		if (error) {
2911 			free(ioes, M_TEMP);
2912 			break;
2913 		}
2914 		PF_RULES_WLOCK();
2915 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2916 			switch (ioe->rs_num) {
2917 #ifdef ALTQ
2918 			case PF_RULESET_ALTQ:
2919 				if (ioe->anchor[0]) {
2920 					PF_RULES_WUNLOCK();
2921 					free(ioes, M_TEMP);
2922 					error = EINVAL;
2923 					goto fail;
2924 				}
2925 				if ((error = pf_rollback_altq(ioe->ticket))) {
2926 					PF_RULES_WUNLOCK();
2927 					free(ioes, M_TEMP);
2928 					goto fail; /* really bad */
2929 				}
2930 				break;
2931 #endif /* ALTQ */
2932 			case PF_RULESET_TABLE:
2933 			    {
2934 				struct pfr_table table;
2935 
2936 				bzero(&table, sizeof(table));
2937 				strlcpy(table.pfrt_anchor, ioe->anchor,
2938 				    sizeof(table.pfrt_anchor));
2939 				if ((error = pfr_ina_rollback(&table,
2940 				    ioe->ticket, NULL, 0))) {
2941 					PF_RULES_WUNLOCK();
2942 					free(ioes, M_TEMP);
2943 					goto fail; /* really bad */
2944 				}
2945 				break;
2946 			    }
2947 			default:
2948 				if ((error = pf_rollback_rules(ioe->ticket,
2949 				    ioe->rs_num, ioe->anchor))) {
2950 					PF_RULES_WUNLOCK();
2951 					free(ioes, M_TEMP);
2952 					goto fail; /* really bad */
2953 				}
2954 				break;
2955 			}
2956 		}
2957 		PF_RULES_WUNLOCK();
2958 		free(ioes, M_TEMP);
2959 		break;
2960 	}
2961 
2962 	case DIOCXCOMMIT: {
2963 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2964 		struct pfioc_trans_e	*ioe, *ioes;
2965 		struct pf_ruleset	*rs;
2966 		size_t			 totlen;
2967 		int			 i;
2968 
2969 		if (io->esize != sizeof(*ioe)) {
2970 			error = ENODEV;
2971 			break;
2972 		}
2973 		totlen = sizeof(struct pfioc_trans_e) * io->size;
2974 		ioes = malloc(totlen, M_TEMP, M_WAITOK);
2975 		error = copyin(io->array, ioes, totlen);
2976 		if (error) {
2977 			free(ioes, M_TEMP);
2978 			break;
2979 		}
2980 		PF_RULES_WLOCK();
2981 		/* First makes sure everything will succeed. */
2982 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2983 			switch (ioe->rs_num) {
2984 #ifdef ALTQ
2985 			case PF_RULESET_ALTQ:
2986 				if (ioe->anchor[0]) {
2987 					PF_RULES_WUNLOCK();
2988 					free(ioes, M_TEMP);
2989 					error = EINVAL;
2990 					goto fail;
2991 				}
2992 				if (!V_altqs_inactive_open || ioe->ticket !=
2993 				    V_ticket_altqs_inactive) {
2994 					PF_RULES_WUNLOCK();
2995 					free(ioes, M_TEMP);
2996 					error = EBUSY;
2997 					goto fail;
2998 				}
2999 				break;
3000 #endif /* ALTQ */
3001 			case PF_RULESET_TABLE:
3002 				rs = pf_find_ruleset(ioe->anchor);
3003 				if (rs == NULL || !rs->topen || ioe->ticket !=
3004 				    rs->tticket) {
3005 					PF_RULES_WUNLOCK();
3006 					free(ioes, M_TEMP);
3007 					error = EBUSY;
3008 					goto fail;
3009 				}
3010 				break;
3011 			default:
3012 				if (ioe->rs_num < 0 || ioe->rs_num >=
3013 				    PF_RULESET_MAX) {
3014 					PF_RULES_WUNLOCK();
3015 					free(ioes, M_TEMP);
3016 					error = EINVAL;
3017 					goto fail;
3018 				}
3019 				rs = pf_find_ruleset(ioe->anchor);
3020 				if (rs == NULL ||
3021 				    !rs->rules[ioe->rs_num].inactive.open ||
3022 				    rs->rules[ioe->rs_num].inactive.ticket !=
3023 				    ioe->ticket) {
3024 					PF_RULES_WUNLOCK();
3025 					free(ioes, M_TEMP);
3026 					error = EBUSY;
3027 					goto fail;
3028 				}
3029 				break;
3030 			}
3031 		}
3032 		/* Now do the commit - no errors should happen here. */
3033 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3034 			switch (ioe->rs_num) {
3035 #ifdef ALTQ
3036 			case PF_RULESET_ALTQ:
3037 				if ((error = pf_commit_altq(ioe->ticket))) {
3038 					PF_RULES_WUNLOCK();
3039 					free(ioes, M_TEMP);
3040 					goto fail; /* really bad */
3041 				}
3042 				break;
3043 #endif /* ALTQ */
3044 			case PF_RULESET_TABLE:
3045 			    {
3046 				struct pfr_table table;
3047 
3048 				bzero(&table, sizeof(table));
3049 				strlcpy(table.pfrt_anchor, ioe->anchor,
3050 				    sizeof(table.pfrt_anchor));
3051 				if ((error = pfr_ina_commit(&table,
3052 				    ioe->ticket, NULL, NULL, 0))) {
3053 					PF_RULES_WUNLOCK();
3054 					free(ioes, M_TEMP);
3055 					goto fail; /* really bad */
3056 				}
3057 				break;
3058 			    }
3059 			default:
3060 				if ((error = pf_commit_rules(ioe->ticket,
3061 				    ioe->rs_num, ioe->anchor))) {
3062 					PF_RULES_WUNLOCK();
3063 					free(ioes, M_TEMP);
3064 					goto fail; /* really bad */
3065 				}
3066 				break;
3067 			}
3068 		}
3069 		PF_RULES_WUNLOCK();
3070 		free(ioes, M_TEMP);
3071 		break;
3072 	}
3073 
3074 	case DIOCGETSRCNODES: {
3075 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
3076 		struct pf_srchash	*sh;
3077 		struct pf_src_node	*n, *p, *pstore;
3078 		uint32_t		 i, nr = 0;
3079 
3080 		if (psn->psn_len == 0) {
3081 			for (i = 0, sh = V_pf_srchash; i < V_pf_srchashmask;
3082 			    i++, sh++) {
3083 				PF_HASHROW_LOCK(sh);
3084 				LIST_FOREACH(n, &sh->nodes, entry)
3085 					nr++;
3086 				PF_HASHROW_UNLOCK(sh);
3087 			}
3088 			psn->psn_len = sizeof(struct pf_src_node) * nr;
3089 			break;
3090 		}
3091 
3092 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK);
3093 		for (i = 0, sh = V_pf_srchash; i < V_pf_srchashmask;
3094 		    i++, sh++) {
3095 		    PF_HASHROW_LOCK(sh);
3096 		    LIST_FOREACH(n, &sh->nodes, entry) {
3097 			int	secs = time_uptime, diff;
3098 
3099 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
3100 				break;
3101 
3102 			bcopy(n, p, sizeof(struct pf_src_node));
3103 			if (n->rule.ptr != NULL)
3104 				p->rule.nr = n->rule.ptr->nr;
3105 			p->creation = secs - p->creation;
3106 			if (p->expire > secs)
3107 				p->expire -= secs;
3108 			else
3109 				p->expire = 0;
3110 
3111 			/* Adjust the connection rate estimate. */
3112 			diff = secs - n->conn_rate.last;
3113 			if (diff >= n->conn_rate.seconds)
3114 				p->conn_rate.count = 0;
3115 			else
3116 				p->conn_rate.count -=
3117 				    n->conn_rate.count * diff /
3118 				    n->conn_rate.seconds;
3119 			p++;
3120 			nr++;
3121 		    }
3122 		    PF_HASHROW_UNLOCK(sh);
3123 		}
3124 		error = copyout(pstore, psn->psn_src_nodes,
3125 		    sizeof(struct pf_src_node) * nr);
3126 		if (error) {
3127 			free(pstore, M_TEMP);
3128 			break;
3129 		}
3130 		psn->psn_len = sizeof(struct pf_src_node) * nr;
3131 		free(pstore, M_TEMP);
3132 		break;
3133 	}
3134 
3135 	case DIOCCLRSRCNODES: {
3136 
3137 		pf_clear_srcnodes(NULL);
3138 		pf_purge_expired_src_nodes();
3139 		V_pf_status.src_nodes = 0;
3140 		break;
3141 	}
3142 
3143 	case DIOCKILLSRCNODES: {
3144 		struct pfioc_src_node_kill *psnk =
3145 		    (struct pfioc_src_node_kill *)addr;
3146 		struct pf_srchash	*sh;
3147 		struct pf_src_node	*sn;
3148 		u_int			i, killed = 0;
3149 
3150 		for (i = 0, sh = V_pf_srchash; i < V_pf_srchashmask;
3151 		    i++, sh++) {
3152 		    /*
3153 		     * XXXGL: we don't ever acquire sources hash lock
3154 		     * but if we ever do, the below call to pf_clear_srcnodes()
3155 		     * would lead to a LOR.
3156 		     */
3157 		    PF_HASHROW_LOCK(sh);
3158 		    LIST_FOREACH(sn, &sh->nodes, entry)
3159 			if (PF_MATCHA(psnk->psnk_src.neg,
3160 				&psnk->psnk_src.addr.v.a.addr,
3161 				&psnk->psnk_src.addr.v.a.mask,
3162 				&sn->addr, sn->af) &&
3163 			    PF_MATCHA(psnk->psnk_dst.neg,
3164 				&psnk->psnk_dst.addr.v.a.addr,
3165 				&psnk->psnk_dst.addr.v.a.mask,
3166 				&sn->raddr, sn->af)) {
3167 				/* Handle state to src_node linkage */
3168 				if (sn->states != 0)
3169 					pf_clear_srcnodes(sn);
3170 				sn->expire = 1;
3171 				killed++;
3172 			}
3173 		    PF_HASHROW_UNLOCK(sh);
3174 		}
3175 
3176 		if (killed > 0)
3177 			pf_purge_expired_src_nodes();
3178 
3179 		psnk->psnk_killed = killed;
3180 		break;
3181 	}
3182 
3183 	case DIOCSETHOSTID: {
3184 		u_int32_t	*hostid = (u_int32_t *)addr;
3185 
3186 		PF_RULES_WLOCK();
3187 		if (*hostid == 0)
3188 			V_pf_status.hostid = arc4random();
3189 		else
3190 			V_pf_status.hostid = *hostid;
3191 		PF_RULES_WUNLOCK();
3192 		break;
3193 	}
3194 
3195 	case DIOCOSFPFLUSH:
3196 		PF_RULES_WLOCK();
3197 		pf_osfp_flush();
3198 		PF_RULES_WUNLOCK();
3199 		break;
3200 
3201 	case DIOCIGETIFACES: {
3202 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3203 		struct pfi_kif *ifstore;
3204 		size_t bufsiz;
3205 
3206 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
3207 			error = ENODEV;
3208 			break;
3209 		}
3210 
3211 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
3212 		ifstore = malloc(bufsiz, M_TEMP, M_WAITOK);
3213 		PF_RULES_RLOCK();
3214 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
3215 		PF_RULES_RUNLOCK();
3216 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
3217 		free(ifstore, M_TEMP);
3218 		break;
3219 	}
3220 
3221 	case DIOCSETIFFLAG: {
3222 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3223 
3224 		PF_RULES_WLOCK();
3225 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3226 		PF_RULES_WUNLOCK();
3227 		break;
3228 	}
3229 
3230 	case DIOCCLRIFFLAG: {
3231 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3232 
3233 		PF_RULES_WLOCK();
3234 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3235 		PF_RULES_WUNLOCK();
3236 		break;
3237 	}
3238 
3239 	default:
3240 		error = ENODEV;
3241 		break;
3242 	}
3243 fail:
3244 	CURVNET_RESTORE();
3245 
3246 	return (error);
3247 }
3248 
3249 void
3250 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
3251 {
3252 	bzero(sp, sizeof(struct pfsync_state));
3253 
3254 	/* copy from state key */
3255 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
3256 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
3257 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
3258 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
3259 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
3260 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
3261 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
3262 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
3263 	sp->proto = st->key[PF_SK_WIRE]->proto;
3264 	sp->af = st->key[PF_SK_WIRE]->af;
3265 
3266 	/* copy from state */
3267 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
3268 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
3269 	sp->creation = htonl(time_uptime - st->creation);
3270 	sp->expire = pf_state_expires(st);
3271 	if (sp->expire <= time_uptime)
3272 		sp->expire = htonl(0);
3273 	else
3274 		sp->expire = htonl(sp->expire - time_uptime);
3275 
3276 	sp->direction = st->direction;
3277 	sp->log = st->log;
3278 	sp->timeout = st->timeout;
3279 	sp->state_flags = st->state_flags;
3280 	if (st->src_node)
3281 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
3282 	if (st->nat_src_node)
3283 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
3284 
3285 	sp->id = st->id;
3286 	sp->creatorid = st->creatorid;
3287 	pf_state_peer_hton(&st->src, &sp->src);
3288 	pf_state_peer_hton(&st->dst, &sp->dst);
3289 
3290 	if (st->rule.ptr == NULL)
3291 		sp->rule = htonl(-1);
3292 	else
3293 		sp->rule = htonl(st->rule.ptr->nr);
3294 	if (st->anchor.ptr == NULL)
3295 		sp->anchor = htonl(-1);
3296 	else
3297 		sp->anchor = htonl(st->anchor.ptr->nr);
3298 	if (st->nat_rule.ptr == NULL)
3299 		sp->nat_rule = htonl(-1);
3300 	else
3301 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
3302 
3303 	pf_state_counter_hton(st->packets[0], sp->packets[0]);
3304 	pf_state_counter_hton(st->packets[1], sp->packets[1]);
3305 	pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
3306 	pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
3307 
3308 }
3309 
3310 static void
3311 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
3312 {
3313 	struct pfr_ktable *kt;
3314 
3315 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
3316 
3317 	kt = aw->p.tbl;
3318 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
3319 		kt = kt->pfrkt_root;
3320 	aw->p.tbl = NULL;
3321 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
3322 		kt->pfrkt_cnt : -1;
3323 }
3324 
3325 /*
3326  * XXX - Check for version missmatch!!!
3327  */
3328 static void
3329 pf_clear_states(void)
3330 {
3331 	struct pf_state	*s;
3332 	u_int i;
3333 
3334 	for (i = 0; i <= V_pf_hashmask; i++) {
3335 		struct pf_idhash *ih = &V_pf_idhash[i];
3336 relock:
3337 		PF_HASHROW_LOCK(ih);
3338 		LIST_FOREACH(s, &ih->states, entry) {
3339 			s->timeout = PFTM_PURGE;
3340 			/* Don't send out individual delete messages. */
3341 			s->sync_state = PFSTATE_NOSYNC;
3342 			pf_unlink_state(s, PF_ENTER_LOCKED);
3343 			goto relock;
3344 		}
3345 		PF_HASHROW_UNLOCK(ih);
3346 	}
3347 }
3348 
3349 static int
3350 pf_clear_tables(void)
3351 {
3352 	struct pfioc_table io;
3353 	int error;
3354 
3355 	bzero(&io, sizeof(io));
3356 
3357 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
3358 	    io.pfrio_flags);
3359 
3360 	return (error);
3361 }
3362 
3363 static void
3364 pf_clear_srcnodes(struct pf_src_node *n)
3365 {
3366 	struct pf_state *s;
3367 	int i;
3368 
3369 	for (i = 0; i <= V_pf_hashmask; i++) {
3370 		struct pf_idhash *ih = &V_pf_idhash[i];
3371 
3372 		PF_HASHROW_LOCK(ih);
3373 		LIST_FOREACH(s, &ih->states, entry) {
3374 			if (n == NULL || n == s->src_node)
3375 				s->src_node = NULL;
3376 			if (n == NULL || n == s->nat_src_node)
3377 				s->nat_src_node = NULL;
3378 		}
3379 		PF_HASHROW_UNLOCK(ih);
3380 	}
3381 
3382 	if (n == NULL) {
3383 		struct pf_srchash *sh;
3384 
3385 		for (i = 0, sh = V_pf_srchash; i < V_pf_srchashmask;
3386 		    i++, sh++) {
3387 			PF_HASHROW_LOCK(sh);
3388 			LIST_FOREACH(n, &sh->nodes, entry) {
3389 				n->expire = 1;
3390 				n->states = 0;
3391 			}
3392 			PF_HASHROW_UNLOCK(sh);
3393 		}
3394 	} else {
3395 		/* XXX: hash slot should already be locked here. */
3396 		n->expire = 1;
3397 		n->states = 0;
3398 	}
3399 }
3400 /*
3401  * XXX - Check for version missmatch!!!
3402  */
3403 
3404 /*
3405  * Duplicate pfctl -Fa operation to get rid of as much as we can.
3406  */
3407 static int
3408 shutdown_pf(void)
3409 {
3410 	int error = 0;
3411 	u_int32_t t[5];
3412 	char nn = '\0';
3413 
3414 	V_pf_status.running = 0;
3415 	do {
3416 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
3417 		    != 0) {
3418 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
3419 			break;
3420 		}
3421 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
3422 		    != 0) {
3423 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
3424 			break;		/* XXX: rollback? */
3425 		}
3426 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
3427 		    != 0) {
3428 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
3429 			break;		/* XXX: rollback? */
3430 		}
3431 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
3432 		    != 0) {
3433 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
3434 			break;		/* XXX: rollback? */
3435 		}
3436 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
3437 		    != 0) {
3438 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
3439 			break;		/* XXX: rollback? */
3440 		}
3441 
3442 		/* XXX: these should always succeed here */
3443 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
3444 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
3445 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
3446 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
3447 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
3448 
3449 		if ((error = pf_clear_tables()) != 0)
3450 			break;
3451 
3452 #ifdef ALTQ
3453 		if ((error = pf_begin_altq(&t[0])) != 0) {
3454 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
3455 			break;
3456 		}
3457 		pf_commit_altq(t[0]);
3458 #endif
3459 
3460 		pf_clear_states();
3461 
3462 		pf_clear_srcnodes(NULL);
3463 
3464 		/* status does not use malloced mem so no need to cleanup */
3465 		/* fingerprints and interfaces have thier own cleanup code */
3466 	} while(0);
3467 
3468 	return (error);
3469 }
3470 
3471 #ifdef INET
3472 static int
3473 pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3474     struct inpcb *inp)
3475 {
3476 	int chk;
3477 
3478 	chk = pf_test(PF_IN, ifp, m, inp);
3479 	if (chk && *m) {
3480 		m_freem(*m);
3481 		*m = NULL;
3482 	}
3483 
3484 	return (chk);
3485 }
3486 
3487 static int
3488 pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3489     struct inpcb *inp)
3490 {
3491 	int chk;
3492 
3493 	/* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3494 	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3495 		in_delayed_cksum(*m);
3496 		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3497 	}
3498 
3499 	chk = pf_test(PF_OUT, ifp, m, inp);
3500 	if (chk && *m) {
3501 		m_freem(*m);
3502 		*m = NULL;
3503 	}
3504 
3505 	return (chk);
3506 }
3507 #endif
3508 
3509 #ifdef INET6
3510 static int
3511 pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3512     struct inpcb *inp)
3513 {
3514 	int chk;
3515 
3516 	/*
3517 	 * In case of loopback traffic IPv6 uses the real interface in
3518 	 * order to support scoped addresses. In order to support stateful
3519 	 * filtering we have change this to lo0 as it is the case in IPv4.
3520 	 */
3521 	CURVNET_SET(ifp->if_vnet);
3522 	chk = pf_test6(PF_IN, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
3523 	CURVNET_RESTORE();
3524 	if (chk && *m) {
3525 		m_freem(*m);
3526 		*m = NULL;
3527 	}
3528 	return chk;
3529 }
3530 
3531 static int
3532 pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3533     struct inpcb *inp)
3534 {
3535 	int chk;
3536 
3537 	/* We need a proper CSUM before we start (s. OpenBSD ip_output) */
3538 	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3539 #ifdef INET
3540 		/* XXX-BZ copy&paste error from r126261? */
3541 		in_delayed_cksum(*m);
3542 #endif
3543 		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3544 	}
3545 	CURVNET_SET(ifp->if_vnet);
3546 	chk = pf_test6(PF_OUT, ifp, m, inp);
3547 	CURVNET_RESTORE();
3548 	if (chk && *m) {
3549 		m_freem(*m);
3550 		*m = NULL;
3551 	}
3552 	return chk;
3553 }
3554 #endif /* INET6 */
3555 
3556 static int
3557 hook_pf(void)
3558 {
3559 #ifdef INET
3560 	struct pfil_head *pfh_inet;
3561 #endif
3562 #ifdef INET6
3563 	struct pfil_head *pfh_inet6;
3564 #endif
3565 
3566 	if (V_pf_pfil_hooked)
3567 		return (0);
3568 
3569 #ifdef INET
3570 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3571 	if (pfh_inet == NULL)
3572 		return (ESRCH); /* XXX */
3573 	pfil_add_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet);
3574 	pfil_add_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet);
3575 #endif
3576 #ifdef INET6
3577 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3578 	if (pfh_inet6 == NULL) {
3579 #ifdef INET
3580 		pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3581 		    pfh_inet);
3582 		pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3583 		    pfh_inet);
3584 #endif
3585 		return (ESRCH); /* XXX */
3586 	}
3587 	pfil_add_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet6);
3588 	pfil_add_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet6);
3589 #endif
3590 
3591 	V_pf_pfil_hooked = 1;
3592 	return (0);
3593 }
3594 
3595 static int
3596 dehook_pf(void)
3597 {
3598 #ifdef INET
3599 	struct pfil_head *pfh_inet;
3600 #endif
3601 #ifdef INET6
3602 	struct pfil_head *pfh_inet6;
3603 #endif
3604 
3605 	if (V_pf_pfil_hooked == 0)
3606 		return (0);
3607 
3608 #ifdef INET
3609 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3610 	if (pfh_inet == NULL)
3611 		return (ESRCH); /* XXX */
3612 	pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3613 	    pfh_inet);
3614 	pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3615 	    pfh_inet);
3616 #endif
3617 #ifdef INET6
3618 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3619 	if (pfh_inet6 == NULL)
3620 		return (ESRCH); /* XXX */
3621 	pfil_remove_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK,
3622 	    pfh_inet6);
3623 	pfil_remove_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK,
3624 	    pfh_inet6);
3625 #endif
3626 
3627 	V_pf_pfil_hooked = 0;
3628 	return (0);
3629 }
3630 
3631 static int
3632 pf_load(void)
3633 {
3634 	int error;
3635 
3636 	VNET_ITERATOR_DECL(vnet_iter);
3637 
3638 	VNET_LIST_RLOCK();
3639 	VNET_FOREACH(vnet_iter) {
3640 		CURVNET_SET(vnet_iter);
3641 		V_pf_pfil_hooked = 0;
3642 		V_pf_end_threads = 0;
3643 		TAILQ_INIT(&V_pf_tags);
3644 		TAILQ_INIT(&V_pf_qids);
3645 		CURVNET_RESTORE();
3646 	}
3647 	VNET_LIST_RUNLOCK();
3648 
3649 	rw_init(&pf_rules_lock, "pf rulesets");
3650 
3651 	pf_dev = make_dev(&pf_cdevsw, 0, 0, 0, 0600, PF_NAME);
3652 	if ((error = pfattach()) != 0)
3653 		return (error);
3654 
3655 	return (0);
3656 }
3657 
3658 static int
3659 pf_unload(void)
3660 {
3661 	int error = 0;
3662 
3663 	PF_RULES_WLOCK();
3664 	V_pf_status.running = 0;
3665 	PF_RULES_WUNLOCK();
3666 	swi_remove(V_pf_swi_cookie);
3667 	error = dehook_pf();
3668 	if (error) {
3669 		/*
3670 		 * Should not happen!
3671 		 * XXX Due to error code ESRCH, kldunload will show
3672 		 * a message like 'No such process'.
3673 		 */
3674 		printf("%s : pfil unregisteration fail\n", __FUNCTION__);
3675 		return error;
3676 	}
3677 	PF_RULES_WLOCK();
3678 	shutdown_pf();
3679 	V_pf_end_threads = 1;
3680 	while (V_pf_end_threads < 2) {
3681 		wakeup_one(pf_purge_thread);
3682 		rw_sleep(pf_purge_thread, &pf_rules_lock, 0, "pftmo", 0);
3683 	}
3684 	pf_normalize_cleanup();
3685 	pfi_cleanup();
3686 	pfr_cleanup();
3687 	pf_osfp_flush();
3688 	pf_cleanup();
3689 	PF_RULES_WUNLOCK();
3690 	destroy_dev(pf_dev);
3691 	rw_destroy(&pf_rules_lock);
3692 
3693 	return (error);
3694 }
3695 
3696 static int
3697 pf_modevent(module_t mod, int type, void *data)
3698 {
3699 	int error = 0;
3700 
3701 	switch(type) {
3702 	case MOD_LOAD:
3703 		error = pf_load();
3704 		break;
3705 	case MOD_QUIESCE:
3706 		/*
3707 		 * Module should not be unloaded due to race conditions.
3708 		 */
3709 		error = EBUSY;
3710 		break;
3711 	case MOD_UNLOAD:
3712 		error = pf_unload();
3713 		break;
3714 	default:
3715 		error = EINVAL;
3716 		break;
3717 	}
3718 
3719 	return (error);
3720 }
3721 
3722 static moduledata_t pf_mod = {
3723 	"pf",
3724 	pf_modevent,
3725 	0
3726 };
3727 
3728 DECLARE_MODULE(pf, pf_mod, SI_SUB_PSEUDO, SI_ORDER_FIRST);
3729 MODULE_VERSION(pf, PF_MODVER);
3730