xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision 97cb52fa9aefd90fad38790fded50905aeeb9b9e)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2001 Daniel Hartmeier
5  * Copyright (c) 2002,2003 Henning Brauer
6  * Copyright (c) 2012 Gleb Smirnoff <glebius@FreeBSD.org>
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  *
13  *    - Redistributions of source code must retain the above copyright
14  *      notice, this list of conditions and the following disclaimer.
15  *    - Redistributions in binary form must reproduce the above
16  *      copyright notice, this list of conditions and the following
17  *      disclaimer in the documentation and/or other materials provided
18  *      with the distribution.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
26  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
28  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
30  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
31  * POSSIBILITY OF SUCH DAMAGE.
32  *
33  * Effort sponsored in part by the Defense Advanced Research Projects
34  * Agency (DARPA) and Air Force Research Laboratory, Air Force
35  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
36  *
37  *	$OpenBSD: pf_ioctl.c,v 1.213 2009/02/15 21:46:12 mbalmer Exp $
38  */
39 
40 #include <sys/cdefs.h>
41 __FBSDID("$FreeBSD$");
42 
43 #include "opt_inet.h"
44 #include "opt_inet6.h"
45 #include "opt_bpf.h"
46 #include "opt_pf.h"
47 
48 #include <sys/param.h>
49 #include <sys/bus.h>
50 #include <sys/conf.h>
51 #include <sys/endian.h>
52 #include <sys/fcntl.h>
53 #include <sys/filio.h>
54 #include <sys/interrupt.h>
55 #include <sys/jail.h>
56 #include <sys/kernel.h>
57 #include <sys/kthread.h>
58 #include <sys/lock.h>
59 #include <sys/mbuf.h>
60 #include <sys/module.h>
61 #include <sys/proc.h>
62 #include <sys/rwlock.h>
63 #include <sys/smp.h>
64 #include <sys/socket.h>
65 #include <sys/sysctl.h>
66 #include <sys/md5.h>
67 #include <sys/ucred.h>
68 
69 #include <net/if.h>
70 #include <net/if_var.h>
71 #include <net/vnet.h>
72 #include <net/route.h>
73 #include <net/pfil.h>
74 #include <net/pfvar.h>
75 #include <net/if_pfsync.h>
76 #include <net/if_pflog.h>
77 
78 #include <netinet/in.h>
79 #include <netinet/ip.h>
80 #include <netinet/ip_var.h>
81 #include <netinet6/ip6_var.h>
82 #include <netinet/ip_icmp.h>
83 
84 #ifdef INET6
85 #include <netinet/ip6.h>
86 #endif /* INET6 */
87 
88 #ifdef ALTQ
89 #include <net/altq/altq.h>
90 #endif
91 
92 static struct pf_pool	*pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t,
93 			    u_int8_t, u_int8_t, u_int8_t);
94 
95 static void		 pf_mv_pool(struct pf_palist *, struct pf_palist *);
96 static void		 pf_empty_pool(struct pf_palist *);
97 static int		 pfioctl(struct cdev *, u_long, caddr_t, int,
98 			    struct thread *);
99 #ifdef ALTQ
100 static int		 pf_begin_altq(u_int32_t *);
101 static int		 pf_rollback_altq(u_int32_t);
102 static int		 pf_commit_altq(u_int32_t);
103 static int		 pf_enable_altq(struct pf_altq *);
104 static int		 pf_disable_altq(struct pf_altq *);
105 static u_int32_t	 pf_qname2qid(char *);
106 static void		 pf_qid_unref(u_int32_t);
107 #endif /* ALTQ */
108 static int		 pf_begin_rules(u_int32_t *, int, const char *);
109 static int		 pf_rollback_rules(u_int32_t, int, char *);
110 static int		 pf_setup_pfsync_matching(struct pf_ruleset *);
111 static void		 pf_hash_rule(MD5_CTX *, struct pf_rule *);
112 static void		 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
113 static int		 pf_commit_rules(u_int32_t, int, char *);
114 static int		 pf_addr_setup(struct pf_ruleset *,
115 			    struct pf_addr_wrap *, sa_family_t);
116 static void		 pf_addr_copyout(struct pf_addr_wrap *);
117 
118 VNET_DEFINE(struct pf_rule,	pf_default_rule);
119 
120 #ifdef ALTQ
121 static VNET_DEFINE(int,		pf_altq_running);
122 #define	V_pf_altq_running	VNET(pf_altq_running)
123 #endif
124 
125 #define	TAGID_MAX	 50000
126 struct pf_tagname {
127 	TAILQ_ENTRY(pf_tagname)	entries;
128 	char			name[PF_TAG_NAME_SIZE];
129 	uint16_t		tag;
130 	int			ref;
131 };
132 
133 TAILQ_HEAD(pf_tags, pf_tagname);
134 #define	V_pf_tags		VNET(pf_tags)
135 VNET_DEFINE(struct pf_tags, pf_tags);
136 #define	V_pf_qids		VNET(pf_qids)
137 VNET_DEFINE(struct pf_tags, pf_qids);
138 static MALLOC_DEFINE(M_PFTAG, "pf_tag", "pf(4) tag names");
139 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
140 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
141 
142 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
143 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
144 #endif
145 
146 static u_int16_t	 tagname2tag(struct pf_tags *, char *);
147 static u_int16_t	 pf_tagname2tag(char *);
148 static void		 tag_unref(struct pf_tags *, u_int16_t);
149 
150 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
151 
152 struct cdev *pf_dev;
153 
154 /*
155  * XXX - These are new and need to be checked when moveing to a new version
156  */
157 static void		 pf_clear_states(void);
158 static int		 pf_clear_tables(void);
159 static void		 pf_clear_srcnodes(struct pf_src_node *);
160 static void		 pf_kill_srcnodes(struct pfioc_src_node_kill *);
161 static void		 pf_tbladdr_copyout(struct pf_addr_wrap *);
162 
163 /*
164  * Wrapper functions for pfil(9) hooks
165  */
166 #ifdef INET
167 static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp,
168     int dir, struct inpcb *inp);
169 static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp,
170     int dir, struct inpcb *inp);
171 #endif
172 #ifdef INET6
173 static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp,
174     int dir, struct inpcb *inp);
175 static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp,
176     int dir, struct inpcb *inp);
177 #endif
178 
179 static int		hook_pf(void);
180 static int		dehook_pf(void);
181 static int		shutdown_pf(void);
182 static int		pf_load(void);
183 static void		pf_unload(void);
184 
185 static struct cdevsw pf_cdevsw = {
186 	.d_ioctl =	pfioctl,
187 	.d_name =	PF_NAME,
188 	.d_version =	D_VERSION,
189 };
190 
191 static volatile VNET_DEFINE(int, pf_pfil_hooked);
192 #define V_pf_pfil_hooked	VNET(pf_pfil_hooked)
193 
194 /*
195  * We need a flag that is neither hooked nor running to know when
196  * the VNET is "valid".  We primarily need this to control (global)
197  * external event, e.g., eventhandlers.
198  */
199 VNET_DEFINE(int, pf_vnet_active);
200 #define V_pf_vnet_active	VNET(pf_vnet_active)
201 
202 int pf_end_threads;
203 struct proc *pf_purge_proc;
204 
205 struct rwlock			pf_rules_lock;
206 struct sx			pf_ioctl_lock;
207 struct sx			pf_end_lock;
208 
209 /* pfsync */
210 pfsync_state_import_t 		*pfsync_state_import_ptr = NULL;
211 pfsync_insert_state_t		*pfsync_insert_state_ptr = NULL;
212 pfsync_update_state_t		*pfsync_update_state_ptr = NULL;
213 pfsync_delete_state_t		*pfsync_delete_state_ptr = NULL;
214 pfsync_clear_states_t		*pfsync_clear_states_ptr = NULL;
215 pfsync_defer_t			*pfsync_defer_ptr = NULL;
216 /* pflog */
217 pflog_packet_t			*pflog_packet_ptr = NULL;
218 
219 static void
220 pfattach_vnet(void)
221 {
222 	u_int32_t *my_timeout = V_pf_default_rule.timeout;
223 
224 	pf_initialize();
225 	pfr_initialize();
226 	pfi_initialize_vnet();
227 	pf_normalize_init();
228 
229 	V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
230 	V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
231 
232 	RB_INIT(&V_pf_anchors);
233 	pf_init_ruleset(&pf_main_ruleset);
234 
235 	/* default rule should never be garbage collected */
236 	V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
237 #ifdef PF_DEFAULT_TO_DROP
238 	V_pf_default_rule.action = PF_DROP;
239 #else
240 	V_pf_default_rule.action = PF_PASS;
241 #endif
242 	V_pf_default_rule.nr = -1;
243 	V_pf_default_rule.rtableid = -1;
244 
245 	V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
246 	V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
247 	V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
248 
249 	/* initialize default timeouts */
250 	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
251 	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
252 	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
253 	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
254 	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
255 	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
256 	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
257 	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
258 	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
259 	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
260 	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
261 	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
262 	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
263 	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
264 	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
265 	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
266 	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
267 	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
268 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
269 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
270 
271 	bzero(&V_pf_status, sizeof(V_pf_status));
272 	V_pf_status.debug = PF_DEBUG_URGENT;
273 
274 	V_pf_pfil_hooked = 0;
275 
276 	/* XXX do our best to avoid a conflict */
277 	V_pf_status.hostid = arc4random();
278 
279 	for (int i = 0; i < PFRES_MAX; i++)
280 		V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
281 	for (int i = 0; i < LCNT_MAX; i++)
282 		V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
283 	for (int i = 0; i < FCNT_MAX; i++)
284 		V_pf_status.fcounters[i] = counter_u64_alloc(M_WAITOK);
285 	for (int i = 0; i < SCNT_MAX; i++)
286 		V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
287 
288 	if (swi_add(NULL, "pf send", pf_intr, curvnet, SWI_NET,
289 	    INTR_MPSAFE, &V_pf_swi_cookie) != 0)
290 		/* XXXGL: leaked all above. */
291 		return;
292 }
293 
294 
295 static struct pf_pool *
296 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
297     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
298     u_int8_t check_ticket)
299 {
300 	struct pf_ruleset	*ruleset;
301 	struct pf_rule		*rule;
302 	int			 rs_num;
303 
304 	ruleset = pf_find_ruleset(anchor);
305 	if (ruleset == NULL)
306 		return (NULL);
307 	rs_num = pf_get_ruleset_number(rule_action);
308 	if (rs_num >= PF_RULESET_MAX)
309 		return (NULL);
310 	if (active) {
311 		if (check_ticket && ticket !=
312 		    ruleset->rules[rs_num].active.ticket)
313 			return (NULL);
314 		if (r_last)
315 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
316 			    pf_rulequeue);
317 		else
318 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
319 	} else {
320 		if (check_ticket && ticket !=
321 		    ruleset->rules[rs_num].inactive.ticket)
322 			return (NULL);
323 		if (r_last)
324 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
325 			    pf_rulequeue);
326 		else
327 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
328 	}
329 	if (!r_last) {
330 		while ((rule != NULL) && (rule->nr != rule_number))
331 			rule = TAILQ_NEXT(rule, entries);
332 	}
333 	if (rule == NULL)
334 		return (NULL);
335 
336 	return (&rule->rpool);
337 }
338 
339 static void
340 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
341 {
342 	struct pf_pooladdr	*mv_pool_pa;
343 
344 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
345 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
346 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
347 	}
348 }
349 
350 static void
351 pf_empty_pool(struct pf_palist *poola)
352 {
353 	struct pf_pooladdr *pa;
354 
355 	while ((pa = TAILQ_FIRST(poola)) != NULL) {
356 		switch (pa->addr.type) {
357 		case PF_ADDR_DYNIFTL:
358 			pfi_dynaddr_remove(pa->addr.p.dyn);
359 			break;
360 		case PF_ADDR_TABLE:
361 			/* XXX: this could be unfinished pooladdr on pabuf */
362 			if (pa->addr.p.tbl != NULL)
363 				pfr_detach_table(pa->addr.p.tbl);
364 			break;
365 		}
366 		if (pa->kif)
367 			pfi_kif_unref(pa->kif);
368 		TAILQ_REMOVE(poola, pa, entries);
369 		free(pa, M_PFRULE);
370 	}
371 }
372 
373 static void
374 pf_unlink_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
375 {
376 
377 	PF_RULES_WASSERT();
378 
379 	TAILQ_REMOVE(rulequeue, rule, entries);
380 
381 	PF_UNLNKDRULES_LOCK();
382 	rule->rule_flag |= PFRULE_REFS;
383 	TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
384 	PF_UNLNKDRULES_UNLOCK();
385 }
386 
387 void
388 pf_free_rule(struct pf_rule *rule)
389 {
390 
391 	PF_RULES_WASSERT();
392 
393 	if (rule->tag)
394 		tag_unref(&V_pf_tags, rule->tag);
395 	if (rule->match_tag)
396 		tag_unref(&V_pf_tags, rule->match_tag);
397 #ifdef ALTQ
398 	if (rule->pqid != rule->qid)
399 		pf_qid_unref(rule->pqid);
400 	pf_qid_unref(rule->qid);
401 #endif
402 	switch (rule->src.addr.type) {
403 	case PF_ADDR_DYNIFTL:
404 		pfi_dynaddr_remove(rule->src.addr.p.dyn);
405 		break;
406 	case PF_ADDR_TABLE:
407 		pfr_detach_table(rule->src.addr.p.tbl);
408 		break;
409 	}
410 	switch (rule->dst.addr.type) {
411 	case PF_ADDR_DYNIFTL:
412 		pfi_dynaddr_remove(rule->dst.addr.p.dyn);
413 		break;
414 	case PF_ADDR_TABLE:
415 		pfr_detach_table(rule->dst.addr.p.tbl);
416 		break;
417 	}
418 	if (rule->overload_tbl)
419 		pfr_detach_table(rule->overload_tbl);
420 	if (rule->kif)
421 		pfi_kif_unref(rule->kif);
422 	pf_anchor_remove(rule);
423 	pf_empty_pool(&rule->rpool.list);
424 	counter_u64_free(rule->states_cur);
425 	counter_u64_free(rule->states_tot);
426 	counter_u64_free(rule->src_nodes);
427 	free(rule, M_PFRULE);
428 }
429 
430 static u_int16_t
431 tagname2tag(struct pf_tags *head, char *tagname)
432 {
433 	struct pf_tagname	*tag, *p = NULL;
434 	u_int16_t		 new_tagid = 1;
435 
436 	PF_RULES_WASSERT();
437 
438 	TAILQ_FOREACH(tag, head, entries)
439 		if (strcmp(tagname, tag->name) == 0) {
440 			tag->ref++;
441 			return (tag->tag);
442 		}
443 
444 	/*
445 	 * to avoid fragmentation, we do a linear search from the beginning
446 	 * and take the first free slot we find. if there is none or the list
447 	 * is empty, append a new entry at the end.
448 	 */
449 
450 	/* new entry */
451 	if (!TAILQ_EMPTY(head))
452 		for (p = TAILQ_FIRST(head); p != NULL &&
453 		    p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
454 			new_tagid = p->tag + 1;
455 
456 	if (new_tagid > TAGID_MAX)
457 		return (0);
458 
459 	/* allocate and fill new struct pf_tagname */
460 	tag = malloc(sizeof(*tag), M_PFTAG, M_NOWAIT|M_ZERO);
461 	if (tag == NULL)
462 		return (0);
463 	strlcpy(tag->name, tagname, sizeof(tag->name));
464 	tag->tag = new_tagid;
465 	tag->ref++;
466 
467 	if (p != NULL)	/* insert new entry before p */
468 		TAILQ_INSERT_BEFORE(p, tag, entries);
469 	else	/* either list empty or no free slot in between */
470 		TAILQ_INSERT_TAIL(head, tag, entries);
471 
472 	return (tag->tag);
473 }
474 
475 static void
476 tag_unref(struct pf_tags *head, u_int16_t tag)
477 {
478 	struct pf_tagname	*p, *next;
479 
480 	PF_RULES_WASSERT();
481 
482 	for (p = TAILQ_FIRST(head); p != NULL; p = next) {
483 		next = TAILQ_NEXT(p, entries);
484 		if (tag == p->tag) {
485 			if (--p->ref == 0) {
486 				TAILQ_REMOVE(head, p, entries);
487 				free(p, M_PFTAG);
488 			}
489 			break;
490 		}
491 	}
492 }
493 
494 static u_int16_t
495 pf_tagname2tag(char *tagname)
496 {
497 	return (tagname2tag(&V_pf_tags, tagname));
498 }
499 
500 #ifdef ALTQ
501 static u_int32_t
502 pf_qname2qid(char *qname)
503 {
504 	return ((u_int32_t)tagname2tag(&V_pf_qids, qname));
505 }
506 
507 static void
508 pf_qid_unref(u_int32_t qid)
509 {
510 	tag_unref(&V_pf_qids, (u_int16_t)qid);
511 }
512 
513 static int
514 pf_begin_altq(u_int32_t *ticket)
515 {
516 	struct pf_altq	*altq;
517 	int		 error = 0;
518 
519 	PF_RULES_WASSERT();
520 
521 	/* Purge the old altq list */
522 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
523 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
524 		if (altq->qname[0] == 0 &&
525 		    (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
526 			/* detach and destroy the discipline */
527 			error = altq_remove(altq);
528 		} else
529 			pf_qid_unref(altq->qid);
530 		free(altq, M_PFALTQ);
531 	}
532 	if (error)
533 		return (error);
534 	*ticket = ++V_ticket_altqs_inactive;
535 	V_altqs_inactive_open = 1;
536 	return (0);
537 }
538 
539 static int
540 pf_rollback_altq(u_int32_t ticket)
541 {
542 	struct pf_altq	*altq;
543 	int		 error = 0;
544 
545 	PF_RULES_WASSERT();
546 
547 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
548 		return (0);
549 	/* Purge the old altq list */
550 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
551 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
552 		if (altq->qname[0] == 0 &&
553 		   (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
554 			/* detach and destroy the discipline */
555 			error = altq_remove(altq);
556 		} else
557 			pf_qid_unref(altq->qid);
558 		free(altq, M_PFALTQ);
559 	}
560 	V_altqs_inactive_open = 0;
561 	return (error);
562 }
563 
564 static int
565 pf_commit_altq(u_int32_t ticket)
566 {
567 	struct pf_altqqueue	*old_altqs;
568 	struct pf_altq		*altq;
569 	int			 err, error = 0;
570 
571 	PF_RULES_WASSERT();
572 
573 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
574 		return (EBUSY);
575 
576 	/* swap altqs, keep the old. */
577 	old_altqs = V_pf_altqs_active;
578 	V_pf_altqs_active = V_pf_altqs_inactive;
579 	V_pf_altqs_inactive = old_altqs;
580 	V_ticket_altqs_active = V_ticket_altqs_inactive;
581 
582 	/* Attach new disciplines */
583 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
584 	if (altq->qname[0] == 0 &&
585 	   (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
586 			/* attach the discipline */
587 			error = altq_pfattach(altq);
588 			if (error == 0 && V_pf_altq_running)
589 				error = pf_enable_altq(altq);
590 			if (error != 0)
591 				return (error);
592 		}
593 	}
594 
595 	/* Purge the old altq list */
596 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
597 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
598 		if (altq->qname[0] == 0 &&
599 		    (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
600 			/* detach and destroy the discipline */
601 			if (V_pf_altq_running)
602 				error = pf_disable_altq(altq);
603 			err = altq_pfdetach(altq);
604 			if (err != 0 && error == 0)
605 				error = err;
606 			err = altq_remove(altq);
607 			if (err != 0 && error == 0)
608 				error = err;
609 		} else
610 			pf_qid_unref(altq->qid);
611 		free(altq, M_PFALTQ);
612 	}
613 
614 	V_altqs_inactive_open = 0;
615 	return (error);
616 }
617 
618 static int
619 pf_enable_altq(struct pf_altq *altq)
620 {
621 	struct ifnet		*ifp;
622 	struct tb_profile	 tb;
623 	int			 error = 0;
624 
625 	if ((ifp = ifunit(altq->ifname)) == NULL)
626 		return (EINVAL);
627 
628 	if (ifp->if_snd.altq_type != ALTQT_NONE)
629 		error = altq_enable(&ifp->if_snd);
630 
631 	/* set tokenbucket regulator */
632 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
633 		tb.rate = altq->ifbandwidth;
634 		tb.depth = altq->tbrsize;
635 		error = tbr_set(&ifp->if_snd, &tb);
636 	}
637 
638 	return (error);
639 }
640 
641 static int
642 pf_disable_altq(struct pf_altq *altq)
643 {
644 	struct ifnet		*ifp;
645 	struct tb_profile	 tb;
646 	int			 error;
647 
648 	if ((ifp = ifunit(altq->ifname)) == NULL)
649 		return (EINVAL);
650 
651 	/*
652 	 * when the discipline is no longer referenced, it was overridden
653 	 * by a new one.  if so, just return.
654 	 */
655 	if (altq->altq_disc != ifp->if_snd.altq_disc)
656 		return (0);
657 
658 	error = altq_disable(&ifp->if_snd);
659 
660 	if (error == 0) {
661 		/* clear tokenbucket regulator */
662 		tb.rate = 0;
663 		error = tbr_set(&ifp->if_snd, &tb);
664 	}
665 
666 	return (error);
667 }
668 
669 void
670 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
671 {
672 	struct ifnet	*ifp1;
673 	struct pf_altq	*a1, *a2, *a3;
674 	u_int32_t	 ticket;
675 	int		 error = 0;
676 
677 	/* Interrupt userland queue modifications */
678 	if (V_altqs_inactive_open)
679 		pf_rollback_altq(V_ticket_altqs_inactive);
680 
681 	/* Start new altq ruleset */
682 	if (pf_begin_altq(&ticket))
683 		return;
684 
685 	/* Copy the current active set */
686 	TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
687 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
688 		if (a2 == NULL) {
689 			error = ENOMEM;
690 			break;
691 		}
692 		bcopy(a1, a2, sizeof(struct pf_altq));
693 
694 		if (a2->qname[0] != 0) {
695 			if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
696 				error = EBUSY;
697 				free(a2, M_PFALTQ);
698 				break;
699 			}
700 			a2->altq_disc = NULL;
701 			TAILQ_FOREACH(a3, V_pf_altqs_inactive, entries) {
702 				if (strncmp(a3->ifname, a2->ifname,
703 				    IFNAMSIZ) == 0 && a3->qname[0] == 0) {
704 					a2->altq_disc = a3->altq_disc;
705 					break;
706 				}
707 			}
708 		}
709 		/* Deactivate the interface in question */
710 		a2->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
711 		if ((ifp1 = ifunit(a2->ifname)) == NULL ||
712 		    (remove && ifp1 == ifp)) {
713 			a2->local_flags |= PFALTQ_FLAG_IF_REMOVED;
714 		} else {
715 			error = altq_add(a2);
716 
717 			if (ticket != V_ticket_altqs_inactive)
718 				error = EBUSY;
719 
720 			if (error) {
721 				free(a2, M_PFALTQ);
722 				break;
723 			}
724 		}
725 
726 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
727 	}
728 
729 	if (error != 0)
730 		pf_rollback_altq(ticket);
731 	else
732 		pf_commit_altq(ticket);
733 }
734 #endif /* ALTQ */
735 
736 static int
737 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
738 {
739 	struct pf_ruleset	*rs;
740 	struct pf_rule		*rule;
741 
742 	PF_RULES_WASSERT();
743 
744 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
745 		return (EINVAL);
746 	rs = pf_find_or_create_ruleset(anchor);
747 	if (rs == NULL)
748 		return (EINVAL);
749 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
750 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
751 		rs->rules[rs_num].inactive.rcount--;
752 	}
753 	*ticket = ++rs->rules[rs_num].inactive.ticket;
754 	rs->rules[rs_num].inactive.open = 1;
755 	return (0);
756 }
757 
758 static int
759 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
760 {
761 	struct pf_ruleset	*rs;
762 	struct pf_rule		*rule;
763 
764 	PF_RULES_WASSERT();
765 
766 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
767 		return (EINVAL);
768 	rs = pf_find_ruleset(anchor);
769 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
770 	    rs->rules[rs_num].inactive.ticket != ticket)
771 		return (0);
772 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
773 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
774 		rs->rules[rs_num].inactive.rcount--;
775 	}
776 	rs->rules[rs_num].inactive.open = 0;
777 	return (0);
778 }
779 
780 #define PF_MD5_UPD(st, elm)						\
781 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
782 
783 #define PF_MD5_UPD_STR(st, elm)						\
784 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
785 
786 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
787 		(stor) = htonl((st)->elm);				\
788 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
789 } while (0)
790 
791 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
792 		(stor) = htons((st)->elm);				\
793 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
794 } while (0)
795 
796 static void
797 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
798 {
799 	PF_MD5_UPD(pfr, addr.type);
800 	switch (pfr->addr.type) {
801 		case PF_ADDR_DYNIFTL:
802 			PF_MD5_UPD(pfr, addr.v.ifname);
803 			PF_MD5_UPD(pfr, addr.iflags);
804 			break;
805 		case PF_ADDR_TABLE:
806 			PF_MD5_UPD(pfr, addr.v.tblname);
807 			break;
808 		case PF_ADDR_ADDRMASK:
809 			/* XXX ignore af? */
810 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
811 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
812 			break;
813 	}
814 
815 	PF_MD5_UPD(pfr, port[0]);
816 	PF_MD5_UPD(pfr, port[1]);
817 	PF_MD5_UPD(pfr, neg);
818 	PF_MD5_UPD(pfr, port_op);
819 }
820 
821 static void
822 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
823 {
824 	u_int16_t x;
825 	u_int32_t y;
826 
827 	pf_hash_rule_addr(ctx, &rule->src);
828 	pf_hash_rule_addr(ctx, &rule->dst);
829 	PF_MD5_UPD_STR(rule, label);
830 	PF_MD5_UPD_STR(rule, ifname);
831 	PF_MD5_UPD_STR(rule, match_tagname);
832 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
833 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
834 	PF_MD5_UPD_HTONL(rule, prob, y);
835 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
836 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
837 	PF_MD5_UPD(rule, uid.op);
838 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
839 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
840 	PF_MD5_UPD(rule, gid.op);
841 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
842 	PF_MD5_UPD(rule, action);
843 	PF_MD5_UPD(rule, direction);
844 	PF_MD5_UPD(rule, af);
845 	PF_MD5_UPD(rule, quick);
846 	PF_MD5_UPD(rule, ifnot);
847 	PF_MD5_UPD(rule, match_tag_not);
848 	PF_MD5_UPD(rule, natpass);
849 	PF_MD5_UPD(rule, keep_state);
850 	PF_MD5_UPD(rule, proto);
851 	PF_MD5_UPD(rule, type);
852 	PF_MD5_UPD(rule, code);
853 	PF_MD5_UPD(rule, flags);
854 	PF_MD5_UPD(rule, flagset);
855 	PF_MD5_UPD(rule, allow_opts);
856 	PF_MD5_UPD(rule, rt);
857 	PF_MD5_UPD(rule, tos);
858 }
859 
860 static int
861 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
862 {
863 	struct pf_ruleset	*rs;
864 	struct pf_rule		*rule, **old_array;
865 	struct pf_rulequeue	*old_rules;
866 	int			 error;
867 	u_int32_t		 old_rcount;
868 
869 	PF_RULES_WASSERT();
870 
871 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
872 		return (EINVAL);
873 	rs = pf_find_ruleset(anchor);
874 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
875 	    ticket != rs->rules[rs_num].inactive.ticket)
876 		return (EBUSY);
877 
878 	/* Calculate checksum for the main ruleset */
879 	if (rs == &pf_main_ruleset) {
880 		error = pf_setup_pfsync_matching(rs);
881 		if (error != 0)
882 			return (error);
883 	}
884 
885 	/* Swap rules, keep the old. */
886 	old_rules = rs->rules[rs_num].active.ptr;
887 	old_rcount = rs->rules[rs_num].active.rcount;
888 	old_array = rs->rules[rs_num].active.ptr_array;
889 
890 	rs->rules[rs_num].active.ptr =
891 	    rs->rules[rs_num].inactive.ptr;
892 	rs->rules[rs_num].active.ptr_array =
893 	    rs->rules[rs_num].inactive.ptr_array;
894 	rs->rules[rs_num].active.rcount =
895 	    rs->rules[rs_num].inactive.rcount;
896 	rs->rules[rs_num].inactive.ptr = old_rules;
897 	rs->rules[rs_num].inactive.ptr_array = old_array;
898 	rs->rules[rs_num].inactive.rcount = old_rcount;
899 
900 	rs->rules[rs_num].active.ticket =
901 	    rs->rules[rs_num].inactive.ticket;
902 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
903 
904 
905 	/* Purge the old rule list. */
906 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
907 		pf_unlink_rule(old_rules, rule);
908 	if (rs->rules[rs_num].inactive.ptr_array)
909 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
910 	rs->rules[rs_num].inactive.ptr_array = NULL;
911 	rs->rules[rs_num].inactive.rcount = 0;
912 	rs->rules[rs_num].inactive.open = 0;
913 	pf_remove_if_empty_ruleset(rs);
914 
915 	return (0);
916 }
917 
918 static int
919 pf_setup_pfsync_matching(struct pf_ruleset *rs)
920 {
921 	MD5_CTX			 ctx;
922 	struct pf_rule		*rule;
923 	int			 rs_cnt;
924 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
925 
926 	MD5Init(&ctx);
927 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
928 		/* XXX PF_RULESET_SCRUB as well? */
929 		if (rs_cnt == PF_RULESET_SCRUB)
930 			continue;
931 
932 		if (rs->rules[rs_cnt].inactive.ptr_array)
933 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
934 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
935 
936 		if (rs->rules[rs_cnt].inactive.rcount) {
937 			rs->rules[rs_cnt].inactive.ptr_array =
938 			    malloc(sizeof(caddr_t) *
939 			    rs->rules[rs_cnt].inactive.rcount,
940 			    M_TEMP, M_NOWAIT);
941 
942 			if (!rs->rules[rs_cnt].inactive.ptr_array)
943 				return (ENOMEM);
944 		}
945 
946 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
947 		    entries) {
948 			pf_hash_rule(&ctx, rule);
949 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
950 		}
951 	}
952 
953 	MD5Final(digest, &ctx);
954 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
955 	return (0);
956 }
957 
958 static int
959 pf_addr_setup(struct pf_ruleset *ruleset, struct pf_addr_wrap *addr,
960     sa_family_t af)
961 {
962 	int error = 0;
963 
964 	switch (addr->type) {
965 	case PF_ADDR_TABLE:
966 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
967 		if (addr->p.tbl == NULL)
968 			error = ENOMEM;
969 		break;
970 	case PF_ADDR_DYNIFTL:
971 		error = pfi_dynaddr_setup(addr, af);
972 		break;
973 	}
974 
975 	return (error);
976 }
977 
978 static void
979 pf_addr_copyout(struct pf_addr_wrap *addr)
980 {
981 
982 	switch (addr->type) {
983 	case PF_ADDR_DYNIFTL:
984 		pfi_dynaddr_copyout(addr);
985 		break;
986 	case PF_ADDR_TABLE:
987 		pf_tbladdr_copyout(addr);
988 		break;
989 	}
990 }
991 
992 static int
993 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
994 {
995 	int			 error = 0;
996 
997 	/* XXX keep in sync with switch() below */
998 	if (securelevel_gt(td->td_ucred, 2))
999 		switch (cmd) {
1000 		case DIOCGETRULES:
1001 		case DIOCGETRULE:
1002 		case DIOCGETADDRS:
1003 		case DIOCGETADDR:
1004 		case DIOCGETSTATE:
1005 		case DIOCSETSTATUSIF:
1006 		case DIOCGETSTATUS:
1007 		case DIOCCLRSTATUS:
1008 		case DIOCNATLOOK:
1009 		case DIOCSETDEBUG:
1010 		case DIOCGETSTATES:
1011 		case DIOCGETTIMEOUT:
1012 		case DIOCCLRRULECTRS:
1013 		case DIOCGETLIMIT:
1014 		case DIOCGETALTQS:
1015 		case DIOCGETALTQ:
1016 		case DIOCGETQSTATS:
1017 		case DIOCGETRULESETS:
1018 		case DIOCGETRULESET:
1019 		case DIOCRGETTABLES:
1020 		case DIOCRGETTSTATS:
1021 		case DIOCRCLRTSTATS:
1022 		case DIOCRCLRADDRS:
1023 		case DIOCRADDADDRS:
1024 		case DIOCRDELADDRS:
1025 		case DIOCRSETADDRS:
1026 		case DIOCRGETADDRS:
1027 		case DIOCRGETASTATS:
1028 		case DIOCRCLRASTATS:
1029 		case DIOCRTSTADDRS:
1030 		case DIOCOSFPGET:
1031 		case DIOCGETSRCNODES:
1032 		case DIOCCLRSRCNODES:
1033 		case DIOCIGETIFACES:
1034 		case DIOCGIFSPEED:
1035 		case DIOCSETIFFLAG:
1036 		case DIOCCLRIFFLAG:
1037 			break;
1038 		case DIOCRCLRTABLES:
1039 		case DIOCRADDTABLES:
1040 		case DIOCRDELTABLES:
1041 		case DIOCRSETTFLAGS:
1042 			if (((struct pfioc_table *)addr)->pfrio_flags &
1043 			    PFR_FLAG_DUMMY)
1044 				break; /* dummy operation ok */
1045 			return (EPERM);
1046 		default:
1047 			return (EPERM);
1048 		}
1049 
1050 	if (!(flags & FWRITE))
1051 		switch (cmd) {
1052 		case DIOCGETRULES:
1053 		case DIOCGETADDRS:
1054 		case DIOCGETADDR:
1055 		case DIOCGETSTATE:
1056 		case DIOCGETSTATUS:
1057 		case DIOCGETSTATES:
1058 		case DIOCGETTIMEOUT:
1059 		case DIOCGETLIMIT:
1060 		case DIOCGETALTQS:
1061 		case DIOCGETALTQ:
1062 		case DIOCGETQSTATS:
1063 		case DIOCGETRULESETS:
1064 		case DIOCGETRULESET:
1065 		case DIOCNATLOOK:
1066 		case DIOCRGETTABLES:
1067 		case DIOCRGETTSTATS:
1068 		case DIOCRGETADDRS:
1069 		case DIOCRGETASTATS:
1070 		case DIOCRTSTADDRS:
1071 		case DIOCOSFPGET:
1072 		case DIOCGETSRCNODES:
1073 		case DIOCIGETIFACES:
1074 		case DIOCGIFSPEED:
1075 			break;
1076 		case DIOCRCLRTABLES:
1077 		case DIOCRADDTABLES:
1078 		case DIOCRDELTABLES:
1079 		case DIOCRCLRTSTATS:
1080 		case DIOCRCLRADDRS:
1081 		case DIOCRADDADDRS:
1082 		case DIOCRDELADDRS:
1083 		case DIOCRSETADDRS:
1084 		case DIOCRSETTFLAGS:
1085 			if (((struct pfioc_table *)addr)->pfrio_flags &
1086 			    PFR_FLAG_DUMMY) {
1087 				flags |= FWRITE; /* need write lock for dummy */
1088 				break; /* dummy operation ok */
1089 			}
1090 			return (EACCES);
1091 		case DIOCGETRULE:
1092 			if (((struct pfioc_rule *)addr)->action ==
1093 			    PF_GET_CLR_CNTR)
1094 				return (EACCES);
1095 			break;
1096 		default:
1097 			return (EACCES);
1098 		}
1099 
1100 	CURVNET_SET(TD_TO_VNET(td));
1101 
1102 	switch (cmd) {
1103 	case DIOCSTART:
1104 		sx_xlock(&pf_ioctl_lock);
1105 		if (V_pf_status.running)
1106 			error = EEXIST;
1107 		else {
1108 			int cpu;
1109 
1110 			error = hook_pf();
1111 			if (error) {
1112 				DPFPRINTF(PF_DEBUG_MISC,
1113 				    ("pf: pfil registration failed\n"));
1114 				break;
1115 			}
1116 			V_pf_status.running = 1;
1117 			V_pf_status.since = time_second;
1118 
1119 			CPU_FOREACH(cpu)
1120 				V_pf_stateid[cpu] = time_second;
1121 
1122 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1123 		}
1124 		break;
1125 
1126 	case DIOCSTOP:
1127 		sx_xlock(&pf_ioctl_lock);
1128 		if (!V_pf_status.running)
1129 			error = ENOENT;
1130 		else {
1131 			V_pf_status.running = 0;
1132 			error = dehook_pf();
1133 			if (error) {
1134 				V_pf_status.running = 1;
1135 				DPFPRINTF(PF_DEBUG_MISC,
1136 				    ("pf: pfil unregistration failed\n"));
1137 			}
1138 			V_pf_status.since = time_second;
1139 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1140 		}
1141 		break;
1142 
1143 	case DIOCADDRULE: {
1144 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1145 		struct pf_ruleset	*ruleset;
1146 		struct pf_rule		*rule, *tail;
1147 		struct pf_pooladdr	*pa;
1148 		struct pfi_kif		*kif = NULL;
1149 		int			 rs_num;
1150 
1151 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1152 			error = EINVAL;
1153 			break;
1154 		}
1155 #ifndef INET
1156 		if (pr->rule.af == AF_INET) {
1157 			error = EAFNOSUPPORT;
1158 			break;
1159 		}
1160 #endif /* INET */
1161 #ifndef INET6
1162 		if (pr->rule.af == AF_INET6) {
1163 			error = EAFNOSUPPORT;
1164 			break;
1165 		}
1166 #endif /* INET6 */
1167 
1168 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
1169 		bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1170 		if (rule->ifname[0])
1171 			kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1172 		rule->states_cur = counter_u64_alloc(M_WAITOK);
1173 		rule->states_tot = counter_u64_alloc(M_WAITOK);
1174 		rule->src_nodes = counter_u64_alloc(M_WAITOK);
1175 		rule->cuid = td->td_ucred->cr_ruid;
1176 		rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1177 		TAILQ_INIT(&rule->rpool.list);
1178 
1179 #define	ERROUT(x)	{ error = (x); goto DIOCADDRULE_error; }
1180 
1181 		PF_RULES_WLOCK();
1182 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1183 		ruleset = pf_find_ruleset(pr->anchor);
1184 		if (ruleset == NULL)
1185 			ERROUT(EINVAL);
1186 		rs_num = pf_get_ruleset_number(pr->rule.action);
1187 		if (rs_num >= PF_RULESET_MAX)
1188 			ERROUT(EINVAL);
1189 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1190 			DPFPRINTF(PF_DEBUG_MISC,
1191 			    ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
1192 			    ruleset->rules[rs_num].inactive.ticket));
1193 			ERROUT(EBUSY);
1194 		}
1195 		if (pr->pool_ticket != V_ticket_pabuf) {
1196 			DPFPRINTF(PF_DEBUG_MISC,
1197 			    ("pool_ticket: %d != %d\n", pr->pool_ticket,
1198 			    V_ticket_pabuf));
1199 			ERROUT(EBUSY);
1200 		}
1201 
1202 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1203 		    pf_rulequeue);
1204 		if (tail)
1205 			rule->nr = tail->nr + 1;
1206 		else
1207 			rule->nr = 0;
1208 		if (rule->ifname[0]) {
1209 			rule->kif = pfi_kif_attach(kif, rule->ifname);
1210 			pfi_kif_ref(rule->kif);
1211 		} else
1212 			rule->kif = NULL;
1213 
1214 		if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
1215 			error = EBUSY;
1216 
1217 #ifdef ALTQ
1218 		/* set queue IDs */
1219 		if (rule->qname[0] != 0) {
1220 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1221 				error = EBUSY;
1222 			else if (rule->pqname[0] != 0) {
1223 				if ((rule->pqid =
1224 				    pf_qname2qid(rule->pqname)) == 0)
1225 					error = EBUSY;
1226 			} else
1227 				rule->pqid = rule->qid;
1228 		}
1229 #endif
1230 		if (rule->tagname[0])
1231 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1232 				error = EBUSY;
1233 		if (rule->match_tagname[0])
1234 			if ((rule->match_tag =
1235 			    pf_tagname2tag(rule->match_tagname)) == 0)
1236 				error = EBUSY;
1237 		if (rule->rt && !rule->direction)
1238 			error = EINVAL;
1239 		if (!rule->log)
1240 			rule->logif = 0;
1241 		if (rule->logif >= PFLOGIFS_MAX)
1242 			error = EINVAL;
1243 		if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1244 			error = ENOMEM;
1245 		if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1246 			error = ENOMEM;
1247 		if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1248 			error = EINVAL;
1249 		if (rule->scrub_flags & PFSTATE_SETPRIO &&
1250 		    (rule->set_prio[0] > PF_PRIO_MAX ||
1251 		    rule->set_prio[1] > PF_PRIO_MAX))
1252 			error = EINVAL;
1253 		TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1254 			if (pa->addr.type == PF_ADDR_TABLE) {
1255 				pa->addr.p.tbl = pfr_attach_table(ruleset,
1256 				    pa->addr.v.tblname);
1257 				if (pa->addr.p.tbl == NULL)
1258 					error = ENOMEM;
1259 			}
1260 
1261 		rule->overload_tbl = NULL;
1262 		if (rule->overload_tblname[0]) {
1263 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1264 			    rule->overload_tblname)) == NULL)
1265 				error = EINVAL;
1266 			else
1267 				rule->overload_tbl->pfrkt_flags |=
1268 				    PFR_TFLAG_ACTIVE;
1269 		}
1270 
1271 		pf_mv_pool(&V_pf_pabuf, &rule->rpool.list);
1272 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1273 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1274 		    (rule->rt > PF_NOPFROUTE)) &&
1275 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1276 			error = EINVAL;
1277 
1278 		if (error) {
1279 			pf_free_rule(rule);
1280 			PF_RULES_WUNLOCK();
1281 			break;
1282 		}
1283 
1284 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1285 		rule->evaluations = rule->packets[0] = rule->packets[1] =
1286 		    rule->bytes[0] = rule->bytes[1] = 0;
1287 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1288 		    rule, entries);
1289 		ruleset->rules[rs_num].inactive.rcount++;
1290 		PF_RULES_WUNLOCK();
1291 		break;
1292 
1293 #undef ERROUT
1294 DIOCADDRULE_error:
1295 		PF_RULES_WUNLOCK();
1296 		counter_u64_free(rule->states_cur);
1297 		counter_u64_free(rule->states_tot);
1298 		counter_u64_free(rule->src_nodes);
1299 		free(rule, M_PFRULE);
1300 		if (kif)
1301 			free(kif, PFI_MTYPE);
1302 		break;
1303 	}
1304 
1305 	case DIOCGETRULES: {
1306 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1307 		struct pf_ruleset	*ruleset;
1308 		struct pf_rule		*tail;
1309 		int			 rs_num;
1310 
1311 		PF_RULES_WLOCK();
1312 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1313 		ruleset = pf_find_ruleset(pr->anchor);
1314 		if (ruleset == NULL) {
1315 			PF_RULES_WUNLOCK();
1316 			error = EINVAL;
1317 			break;
1318 		}
1319 		rs_num = pf_get_ruleset_number(pr->rule.action);
1320 		if (rs_num >= PF_RULESET_MAX) {
1321 			PF_RULES_WUNLOCK();
1322 			error = EINVAL;
1323 			break;
1324 		}
1325 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1326 		    pf_rulequeue);
1327 		if (tail)
1328 			pr->nr = tail->nr + 1;
1329 		else
1330 			pr->nr = 0;
1331 		pr->ticket = ruleset->rules[rs_num].active.ticket;
1332 		PF_RULES_WUNLOCK();
1333 		break;
1334 	}
1335 
1336 	case DIOCGETRULE: {
1337 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1338 		struct pf_ruleset	*ruleset;
1339 		struct pf_rule		*rule;
1340 		int			 rs_num, i;
1341 
1342 		PF_RULES_WLOCK();
1343 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1344 		ruleset = pf_find_ruleset(pr->anchor);
1345 		if (ruleset == NULL) {
1346 			PF_RULES_WUNLOCK();
1347 			error = EINVAL;
1348 			break;
1349 		}
1350 		rs_num = pf_get_ruleset_number(pr->rule.action);
1351 		if (rs_num >= PF_RULESET_MAX) {
1352 			PF_RULES_WUNLOCK();
1353 			error = EINVAL;
1354 			break;
1355 		}
1356 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1357 			PF_RULES_WUNLOCK();
1358 			error = EBUSY;
1359 			break;
1360 		}
1361 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1362 		while ((rule != NULL) && (rule->nr != pr->nr))
1363 			rule = TAILQ_NEXT(rule, entries);
1364 		if (rule == NULL) {
1365 			PF_RULES_WUNLOCK();
1366 			error = EBUSY;
1367 			break;
1368 		}
1369 		bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1370 		pr->rule.u_states_cur = counter_u64_fetch(rule->states_cur);
1371 		pr->rule.u_states_tot = counter_u64_fetch(rule->states_tot);
1372 		pr->rule.u_src_nodes = counter_u64_fetch(rule->src_nodes);
1373 		if (pf_anchor_copyout(ruleset, rule, pr)) {
1374 			PF_RULES_WUNLOCK();
1375 			error = EBUSY;
1376 			break;
1377 		}
1378 		pf_addr_copyout(&pr->rule.src.addr);
1379 		pf_addr_copyout(&pr->rule.dst.addr);
1380 		for (i = 0; i < PF_SKIP_COUNT; ++i)
1381 			if (rule->skip[i].ptr == NULL)
1382 				pr->rule.skip[i].nr = -1;
1383 			else
1384 				pr->rule.skip[i].nr =
1385 				    rule->skip[i].ptr->nr;
1386 
1387 		if (pr->action == PF_GET_CLR_CNTR) {
1388 			rule->evaluations = 0;
1389 			rule->packets[0] = rule->packets[1] = 0;
1390 			rule->bytes[0] = rule->bytes[1] = 0;
1391 			counter_u64_zero(rule->states_tot);
1392 		}
1393 		PF_RULES_WUNLOCK();
1394 		break;
1395 	}
1396 
1397 	case DIOCCHANGERULE: {
1398 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1399 		struct pf_ruleset	*ruleset;
1400 		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1401 		struct pfi_kif		*kif = NULL;
1402 		struct pf_pooladdr	*pa;
1403 		u_int32_t		 nr = 0;
1404 		int			 rs_num;
1405 
1406 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1407 		    pcr->action > PF_CHANGE_GET_TICKET) {
1408 			error = EINVAL;
1409 			break;
1410 		}
1411 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1412 			error = EINVAL;
1413 			break;
1414 		}
1415 
1416 		if (pcr->action != PF_CHANGE_REMOVE) {
1417 #ifndef INET
1418 			if (pcr->rule.af == AF_INET) {
1419 				error = EAFNOSUPPORT;
1420 				break;
1421 			}
1422 #endif /* INET */
1423 #ifndef INET6
1424 			if (pcr->rule.af == AF_INET6) {
1425 				error = EAFNOSUPPORT;
1426 				break;
1427 			}
1428 #endif /* INET6 */
1429 			newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
1430 			bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1431 			if (newrule->ifname[0])
1432 				kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1433 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
1434 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
1435 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
1436 			newrule->cuid = td->td_ucred->cr_ruid;
1437 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1438 			TAILQ_INIT(&newrule->rpool.list);
1439 		}
1440 
1441 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGERULE_error; }
1442 
1443 		PF_RULES_WLOCK();
1444 		if (!(pcr->action == PF_CHANGE_REMOVE ||
1445 		    pcr->action == PF_CHANGE_GET_TICKET) &&
1446 		    pcr->pool_ticket != V_ticket_pabuf)
1447 			ERROUT(EBUSY);
1448 
1449 		ruleset = pf_find_ruleset(pcr->anchor);
1450 		if (ruleset == NULL)
1451 			ERROUT(EINVAL);
1452 
1453 		rs_num = pf_get_ruleset_number(pcr->rule.action);
1454 		if (rs_num >= PF_RULESET_MAX)
1455 			ERROUT(EINVAL);
1456 
1457 		if (pcr->action == PF_CHANGE_GET_TICKET) {
1458 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1459 			ERROUT(0);
1460 		} else if (pcr->ticket !=
1461 			    ruleset->rules[rs_num].active.ticket)
1462 				ERROUT(EINVAL);
1463 
1464 		if (pcr->action != PF_CHANGE_REMOVE) {
1465 			if (newrule->ifname[0]) {
1466 				newrule->kif = pfi_kif_attach(kif,
1467 				    newrule->ifname);
1468 				pfi_kif_ref(newrule->kif);
1469 			} else
1470 				newrule->kif = NULL;
1471 
1472 			if (newrule->rtableid > 0 &&
1473 			    newrule->rtableid >= rt_numfibs)
1474 				error = EBUSY;
1475 
1476 #ifdef ALTQ
1477 			/* set queue IDs */
1478 			if (newrule->qname[0] != 0) {
1479 				if ((newrule->qid =
1480 				    pf_qname2qid(newrule->qname)) == 0)
1481 					error = EBUSY;
1482 				else if (newrule->pqname[0] != 0) {
1483 					if ((newrule->pqid =
1484 					    pf_qname2qid(newrule->pqname)) == 0)
1485 						error = EBUSY;
1486 				} else
1487 					newrule->pqid = newrule->qid;
1488 			}
1489 #endif /* ALTQ */
1490 			if (newrule->tagname[0])
1491 				if ((newrule->tag =
1492 				    pf_tagname2tag(newrule->tagname)) == 0)
1493 					error = EBUSY;
1494 			if (newrule->match_tagname[0])
1495 				if ((newrule->match_tag = pf_tagname2tag(
1496 				    newrule->match_tagname)) == 0)
1497 					error = EBUSY;
1498 			if (newrule->rt && !newrule->direction)
1499 				error = EINVAL;
1500 			if (!newrule->log)
1501 				newrule->logif = 0;
1502 			if (newrule->logif >= PFLOGIFS_MAX)
1503 				error = EINVAL;
1504 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
1505 				error = ENOMEM;
1506 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
1507 				error = ENOMEM;
1508 			if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1509 				error = EINVAL;
1510 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1511 				if (pa->addr.type == PF_ADDR_TABLE) {
1512 					pa->addr.p.tbl =
1513 					    pfr_attach_table(ruleset,
1514 					    pa->addr.v.tblname);
1515 					if (pa->addr.p.tbl == NULL)
1516 						error = ENOMEM;
1517 				}
1518 
1519 			newrule->overload_tbl = NULL;
1520 			if (newrule->overload_tblname[0]) {
1521 				if ((newrule->overload_tbl = pfr_attach_table(
1522 				    ruleset, newrule->overload_tblname)) ==
1523 				    NULL)
1524 					error = EINVAL;
1525 				else
1526 					newrule->overload_tbl->pfrkt_flags |=
1527 					    PFR_TFLAG_ACTIVE;
1528 			}
1529 
1530 			pf_mv_pool(&V_pf_pabuf, &newrule->rpool.list);
1531 			if (((((newrule->action == PF_NAT) ||
1532 			    (newrule->action == PF_RDR) ||
1533 			    (newrule->action == PF_BINAT) ||
1534 			    (newrule->rt > PF_NOPFROUTE)) &&
1535 			    !newrule->anchor)) &&
1536 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1537 				error = EINVAL;
1538 
1539 			if (error) {
1540 				pf_free_rule(newrule);
1541 				PF_RULES_WUNLOCK();
1542 				break;
1543 			}
1544 
1545 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1546 			newrule->evaluations = 0;
1547 			newrule->packets[0] = newrule->packets[1] = 0;
1548 			newrule->bytes[0] = newrule->bytes[1] = 0;
1549 		}
1550 		pf_empty_pool(&V_pf_pabuf);
1551 
1552 		if (pcr->action == PF_CHANGE_ADD_HEAD)
1553 			oldrule = TAILQ_FIRST(
1554 			    ruleset->rules[rs_num].active.ptr);
1555 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1556 			oldrule = TAILQ_LAST(
1557 			    ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1558 		else {
1559 			oldrule = TAILQ_FIRST(
1560 			    ruleset->rules[rs_num].active.ptr);
1561 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1562 				oldrule = TAILQ_NEXT(oldrule, entries);
1563 			if (oldrule == NULL) {
1564 				if (newrule != NULL)
1565 					pf_free_rule(newrule);
1566 				PF_RULES_WUNLOCK();
1567 				error = EINVAL;
1568 				break;
1569 			}
1570 		}
1571 
1572 		if (pcr->action == PF_CHANGE_REMOVE) {
1573 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
1574 			    oldrule);
1575 			ruleset->rules[rs_num].active.rcount--;
1576 		} else {
1577 			if (oldrule == NULL)
1578 				TAILQ_INSERT_TAIL(
1579 				    ruleset->rules[rs_num].active.ptr,
1580 				    newrule, entries);
1581 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1582 			    pcr->action == PF_CHANGE_ADD_BEFORE)
1583 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1584 			else
1585 				TAILQ_INSERT_AFTER(
1586 				    ruleset->rules[rs_num].active.ptr,
1587 				    oldrule, newrule, entries);
1588 			ruleset->rules[rs_num].active.rcount++;
1589 		}
1590 
1591 		nr = 0;
1592 		TAILQ_FOREACH(oldrule,
1593 		    ruleset->rules[rs_num].active.ptr, entries)
1594 			oldrule->nr = nr++;
1595 
1596 		ruleset->rules[rs_num].active.ticket++;
1597 
1598 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1599 		pf_remove_if_empty_ruleset(ruleset);
1600 
1601 		PF_RULES_WUNLOCK();
1602 		break;
1603 
1604 #undef ERROUT
1605 DIOCCHANGERULE_error:
1606 		PF_RULES_WUNLOCK();
1607 		if (newrule != NULL) {
1608 			counter_u64_free(newrule->states_cur);
1609 			counter_u64_free(newrule->states_tot);
1610 			counter_u64_free(newrule->src_nodes);
1611 			free(newrule, M_PFRULE);
1612 		}
1613 		if (kif != NULL)
1614 			free(kif, PFI_MTYPE);
1615 		break;
1616 	}
1617 
1618 	case DIOCCLRSTATES: {
1619 		struct pf_state		*s;
1620 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1621 		u_int			 i, killed = 0;
1622 
1623 		for (i = 0; i <= pf_hashmask; i++) {
1624 			struct pf_idhash *ih = &V_pf_idhash[i];
1625 
1626 relock_DIOCCLRSTATES:
1627 			PF_HASHROW_LOCK(ih);
1628 			LIST_FOREACH(s, &ih->states, entry)
1629 				if (!psk->psk_ifname[0] ||
1630 				    !strcmp(psk->psk_ifname,
1631 				    s->kif->pfik_name)) {
1632 					/*
1633 					 * Don't send out individual
1634 					 * delete messages.
1635 					 */
1636 					s->state_flags |= PFSTATE_NOSYNC;
1637 					pf_unlink_state(s, PF_ENTER_LOCKED);
1638 					killed++;
1639 					goto relock_DIOCCLRSTATES;
1640 				}
1641 			PF_HASHROW_UNLOCK(ih);
1642 		}
1643 		psk->psk_killed = killed;
1644 		if (pfsync_clear_states_ptr != NULL)
1645 			pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
1646 		break;
1647 	}
1648 
1649 	case DIOCKILLSTATES: {
1650 		struct pf_state		*s;
1651 		struct pf_state_key	*sk;
1652 		struct pf_addr		*srcaddr, *dstaddr;
1653 		u_int16_t		 srcport, dstport;
1654 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
1655 		u_int			 i, killed = 0;
1656 
1657 		if (psk->psk_pfcmp.id) {
1658 			if (psk->psk_pfcmp.creatorid == 0)
1659 				psk->psk_pfcmp.creatorid = V_pf_status.hostid;
1660 			if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
1661 			    psk->psk_pfcmp.creatorid))) {
1662 				pf_unlink_state(s, PF_ENTER_LOCKED);
1663 				psk->psk_killed = 1;
1664 			}
1665 			break;
1666 		}
1667 
1668 		for (i = 0; i <= pf_hashmask; i++) {
1669 			struct pf_idhash *ih = &V_pf_idhash[i];
1670 
1671 relock_DIOCKILLSTATES:
1672 			PF_HASHROW_LOCK(ih);
1673 			LIST_FOREACH(s, &ih->states, entry) {
1674 				sk = s->key[PF_SK_WIRE];
1675 				if (s->direction == PF_OUT) {
1676 					srcaddr = &sk->addr[1];
1677 					dstaddr = &sk->addr[0];
1678 					srcport = sk->port[1];
1679 					dstport = sk->port[0];
1680 				} else {
1681 					srcaddr = &sk->addr[0];
1682 					dstaddr = &sk->addr[1];
1683 					srcport = sk->port[0];
1684 					dstport = sk->port[1];
1685 				}
1686 
1687 				if ((!psk->psk_af || sk->af == psk->psk_af)
1688 				    && (!psk->psk_proto || psk->psk_proto ==
1689 				    sk->proto) &&
1690 				    PF_MATCHA(psk->psk_src.neg,
1691 				    &psk->psk_src.addr.v.a.addr,
1692 				    &psk->psk_src.addr.v.a.mask,
1693 				    srcaddr, sk->af) &&
1694 				    PF_MATCHA(psk->psk_dst.neg,
1695 				    &psk->psk_dst.addr.v.a.addr,
1696 				    &psk->psk_dst.addr.v.a.mask,
1697 				    dstaddr, sk->af) &&
1698 				    (psk->psk_src.port_op == 0 ||
1699 				    pf_match_port(psk->psk_src.port_op,
1700 				    psk->psk_src.port[0], psk->psk_src.port[1],
1701 				    srcport)) &&
1702 				    (psk->psk_dst.port_op == 0 ||
1703 				    pf_match_port(psk->psk_dst.port_op,
1704 				    psk->psk_dst.port[0], psk->psk_dst.port[1],
1705 				    dstport)) &&
1706 				    (!psk->psk_label[0] ||
1707 				    (s->rule.ptr->label[0] &&
1708 				    !strcmp(psk->psk_label,
1709 				    s->rule.ptr->label))) &&
1710 				    (!psk->psk_ifname[0] ||
1711 				    !strcmp(psk->psk_ifname,
1712 				    s->kif->pfik_name))) {
1713 					pf_unlink_state(s, PF_ENTER_LOCKED);
1714 					killed++;
1715 					goto relock_DIOCKILLSTATES;
1716 				}
1717 			}
1718 			PF_HASHROW_UNLOCK(ih);
1719 		}
1720 		psk->psk_killed = killed;
1721 		break;
1722 	}
1723 
1724 	case DIOCADDSTATE: {
1725 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1726 		struct pfsync_state	*sp = &ps->state;
1727 
1728 		if (sp->timeout >= PFTM_MAX) {
1729 			error = EINVAL;
1730 			break;
1731 		}
1732 		if (pfsync_state_import_ptr != NULL) {
1733 			PF_RULES_RLOCK();
1734 			error = pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
1735 			PF_RULES_RUNLOCK();
1736 		} else
1737 			error = EOPNOTSUPP;
1738 		break;
1739 	}
1740 
1741 	case DIOCGETSTATE: {
1742 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1743 		struct pf_state		*s;
1744 
1745 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
1746 		if (s == NULL) {
1747 			error = ENOENT;
1748 			break;
1749 		}
1750 
1751 		pfsync_state_export(&ps->state, s);
1752 		PF_STATE_UNLOCK(s);
1753 		break;
1754 	}
1755 
1756 	case DIOCGETSTATES: {
1757 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
1758 		struct pf_state		*s;
1759 		struct pfsync_state	*pstore, *p;
1760 		int i, nr;
1761 
1762 		if (ps->ps_len == 0) {
1763 			nr = uma_zone_get_cur(V_pf_state_z);
1764 			ps->ps_len = sizeof(struct pfsync_state) * nr;
1765 			break;
1766 		}
1767 
1768 		p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK);
1769 		nr = 0;
1770 
1771 		for (i = 0; i <= pf_hashmask; i++) {
1772 			struct pf_idhash *ih = &V_pf_idhash[i];
1773 
1774 			PF_HASHROW_LOCK(ih);
1775 			LIST_FOREACH(s, &ih->states, entry) {
1776 
1777 				if (s->timeout == PFTM_UNLINKED)
1778 					continue;
1779 
1780 				if ((nr+1) * sizeof(*p) > ps->ps_len) {
1781 					PF_HASHROW_UNLOCK(ih);
1782 					goto DIOCGETSTATES_full;
1783 				}
1784 				pfsync_state_export(p, s);
1785 				p++;
1786 				nr++;
1787 			}
1788 			PF_HASHROW_UNLOCK(ih);
1789 		}
1790 DIOCGETSTATES_full:
1791 		error = copyout(pstore, ps->ps_states,
1792 		    sizeof(struct pfsync_state) * nr);
1793 		if (error) {
1794 			free(pstore, M_TEMP);
1795 			break;
1796 		}
1797 		ps->ps_len = sizeof(struct pfsync_state) * nr;
1798 		free(pstore, M_TEMP);
1799 
1800 		break;
1801 	}
1802 
1803 	case DIOCGETSTATUS: {
1804 		struct pf_status *s = (struct pf_status *)addr;
1805 
1806 		PF_RULES_RLOCK();
1807 		s->running = V_pf_status.running;
1808 		s->since   = V_pf_status.since;
1809 		s->debug   = V_pf_status.debug;
1810 		s->hostid  = V_pf_status.hostid;
1811 		s->states  = V_pf_status.states;
1812 		s->src_nodes = V_pf_status.src_nodes;
1813 
1814 		for (int i = 0; i < PFRES_MAX; i++)
1815 			s->counters[i] =
1816 			    counter_u64_fetch(V_pf_status.counters[i]);
1817 		for (int i = 0; i < LCNT_MAX; i++)
1818 			s->lcounters[i] =
1819 			    counter_u64_fetch(V_pf_status.lcounters[i]);
1820 		for (int i = 0; i < FCNT_MAX; i++)
1821 			s->fcounters[i] =
1822 			    counter_u64_fetch(V_pf_status.fcounters[i]);
1823 		for (int i = 0; i < SCNT_MAX; i++)
1824 			s->scounters[i] =
1825 			    counter_u64_fetch(V_pf_status.scounters[i]);
1826 
1827 		bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ);
1828 		bcopy(V_pf_status.pf_chksum, s->pf_chksum,
1829 		    PF_MD5_DIGEST_LENGTH);
1830 
1831 		pfi_update_status(s->ifname, s);
1832 		PF_RULES_RUNLOCK();
1833 		break;
1834 	}
1835 
1836 	case DIOCSETSTATUSIF: {
1837 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
1838 
1839 		if (pi->ifname[0] == 0) {
1840 			bzero(V_pf_status.ifname, IFNAMSIZ);
1841 			break;
1842 		}
1843 		PF_RULES_WLOCK();
1844 		strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
1845 		PF_RULES_WUNLOCK();
1846 		break;
1847 	}
1848 
1849 	case DIOCCLRSTATUS: {
1850 		PF_RULES_WLOCK();
1851 		for (int i = 0; i < PFRES_MAX; i++)
1852 			counter_u64_zero(V_pf_status.counters[i]);
1853 		for (int i = 0; i < FCNT_MAX; i++)
1854 			counter_u64_zero(V_pf_status.fcounters[i]);
1855 		for (int i = 0; i < SCNT_MAX; i++)
1856 			counter_u64_zero(V_pf_status.scounters[i]);
1857 		for (int i = 0; i < LCNT_MAX; i++)
1858 			counter_u64_zero(V_pf_status.lcounters[i]);
1859 		V_pf_status.since = time_second;
1860 		if (*V_pf_status.ifname)
1861 			pfi_update_status(V_pf_status.ifname, NULL);
1862 		PF_RULES_WUNLOCK();
1863 		break;
1864 	}
1865 
1866 	case DIOCNATLOOK: {
1867 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
1868 		struct pf_state_key	*sk;
1869 		struct pf_state		*state;
1870 		struct pf_state_key_cmp	 key;
1871 		int			 m = 0, direction = pnl->direction;
1872 		int			 sidx, didx;
1873 
1874 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
1875 		sidx = (direction == PF_IN) ? 1 : 0;
1876 		didx = (direction == PF_IN) ? 0 : 1;
1877 
1878 		if (!pnl->proto ||
1879 		    PF_AZERO(&pnl->saddr, pnl->af) ||
1880 		    PF_AZERO(&pnl->daddr, pnl->af) ||
1881 		    ((pnl->proto == IPPROTO_TCP ||
1882 		    pnl->proto == IPPROTO_UDP) &&
1883 		    (!pnl->dport || !pnl->sport)))
1884 			error = EINVAL;
1885 		else {
1886 			bzero(&key, sizeof(key));
1887 			key.af = pnl->af;
1888 			key.proto = pnl->proto;
1889 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
1890 			key.port[sidx] = pnl->sport;
1891 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
1892 			key.port[didx] = pnl->dport;
1893 
1894 			state = pf_find_state_all(&key, direction, &m);
1895 
1896 			if (m > 1)
1897 				error = E2BIG;	/* more than one state */
1898 			else if (state != NULL) {
1899 				/* XXXGL: not locked read */
1900 				sk = state->key[sidx];
1901 				PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
1902 				pnl->rsport = sk->port[sidx];
1903 				PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
1904 				pnl->rdport = sk->port[didx];
1905 			} else
1906 				error = ENOENT;
1907 		}
1908 		break;
1909 	}
1910 
1911 	case DIOCSETTIMEOUT: {
1912 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1913 		int		 old;
1914 
1915 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
1916 		    pt->seconds < 0) {
1917 			error = EINVAL;
1918 			break;
1919 		}
1920 		PF_RULES_WLOCK();
1921 		old = V_pf_default_rule.timeout[pt->timeout];
1922 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
1923 			pt->seconds = 1;
1924 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
1925 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
1926 			wakeup(pf_purge_thread);
1927 		pt->seconds = old;
1928 		PF_RULES_WUNLOCK();
1929 		break;
1930 	}
1931 
1932 	case DIOCGETTIMEOUT: {
1933 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1934 
1935 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
1936 			error = EINVAL;
1937 			break;
1938 		}
1939 		PF_RULES_RLOCK();
1940 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
1941 		PF_RULES_RUNLOCK();
1942 		break;
1943 	}
1944 
1945 	case DIOCGETLIMIT: {
1946 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1947 
1948 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
1949 			error = EINVAL;
1950 			break;
1951 		}
1952 		PF_RULES_RLOCK();
1953 		pl->limit = V_pf_limits[pl->index].limit;
1954 		PF_RULES_RUNLOCK();
1955 		break;
1956 	}
1957 
1958 	case DIOCSETLIMIT: {
1959 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1960 		int			 old_limit;
1961 
1962 		PF_RULES_WLOCK();
1963 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
1964 		    V_pf_limits[pl->index].zone == NULL) {
1965 			PF_RULES_WUNLOCK();
1966 			error = EINVAL;
1967 			break;
1968 		}
1969 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
1970 		old_limit = V_pf_limits[pl->index].limit;
1971 		V_pf_limits[pl->index].limit = pl->limit;
1972 		pl->limit = old_limit;
1973 		PF_RULES_WUNLOCK();
1974 		break;
1975 	}
1976 
1977 	case DIOCSETDEBUG: {
1978 		u_int32_t	*level = (u_int32_t *)addr;
1979 
1980 		PF_RULES_WLOCK();
1981 		V_pf_status.debug = *level;
1982 		PF_RULES_WUNLOCK();
1983 		break;
1984 	}
1985 
1986 	case DIOCCLRRULECTRS: {
1987 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
1988 		struct pf_ruleset	*ruleset = &pf_main_ruleset;
1989 		struct pf_rule		*rule;
1990 
1991 		PF_RULES_WLOCK();
1992 		TAILQ_FOREACH(rule,
1993 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
1994 			rule->evaluations = 0;
1995 			rule->packets[0] = rule->packets[1] = 0;
1996 			rule->bytes[0] = rule->bytes[1] = 0;
1997 		}
1998 		PF_RULES_WUNLOCK();
1999 		break;
2000 	}
2001 
2002 	case DIOCGIFSPEED: {
2003 		struct pf_ifspeed	*psp = (struct pf_ifspeed *)addr;
2004 		struct pf_ifspeed	ps;
2005 		struct ifnet		*ifp;
2006 
2007 		if (psp->ifname[0] != 0) {
2008 			/* Can we completely trust user-land? */
2009 			strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
2010 			ifp = ifunit(ps.ifname);
2011 			if (ifp != NULL)
2012 				psp->baudrate = ifp->if_baudrate;
2013 			else
2014 				error = EINVAL;
2015 		} else
2016 			error = EINVAL;
2017 		break;
2018 	}
2019 
2020 #ifdef ALTQ
2021 	case DIOCSTARTALTQ: {
2022 		struct pf_altq		*altq;
2023 
2024 		PF_RULES_WLOCK();
2025 		/* enable all altq interfaces on active list */
2026 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
2027 			if (altq->qname[0] == 0 && (altq->local_flags &
2028 			    PFALTQ_FLAG_IF_REMOVED) == 0) {
2029 				error = pf_enable_altq(altq);
2030 				if (error != 0)
2031 					break;
2032 			}
2033 		}
2034 		if (error == 0)
2035 			V_pf_altq_running = 1;
2036 		PF_RULES_WUNLOCK();
2037 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
2038 		break;
2039 	}
2040 
2041 	case DIOCSTOPALTQ: {
2042 		struct pf_altq		*altq;
2043 
2044 		PF_RULES_WLOCK();
2045 		/* disable all altq interfaces on active list */
2046 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
2047 			if (altq->qname[0] == 0 && (altq->local_flags &
2048 			    PFALTQ_FLAG_IF_REMOVED) == 0) {
2049 				error = pf_disable_altq(altq);
2050 				if (error != 0)
2051 					break;
2052 			}
2053 		}
2054 		if (error == 0)
2055 			V_pf_altq_running = 0;
2056 		PF_RULES_WUNLOCK();
2057 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2058 		break;
2059 	}
2060 
2061 	case DIOCADDALTQ: {
2062 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2063 		struct pf_altq		*altq, *a;
2064 		struct ifnet		*ifp;
2065 
2066 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK);
2067 		bcopy(&pa->altq, altq, sizeof(struct pf_altq));
2068 		altq->local_flags = 0;
2069 
2070 		PF_RULES_WLOCK();
2071 		if (pa->ticket != V_ticket_altqs_inactive) {
2072 			PF_RULES_WUNLOCK();
2073 			free(altq, M_PFALTQ);
2074 			error = EBUSY;
2075 			break;
2076 		}
2077 
2078 		/*
2079 		 * if this is for a queue, find the discipline and
2080 		 * copy the necessary fields
2081 		 */
2082 		if (altq->qname[0] != 0) {
2083 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2084 				PF_RULES_WUNLOCK();
2085 				error = EBUSY;
2086 				free(altq, M_PFALTQ);
2087 				break;
2088 			}
2089 			altq->altq_disc = NULL;
2090 			TAILQ_FOREACH(a, V_pf_altqs_inactive, entries) {
2091 				if (strncmp(a->ifname, altq->ifname,
2092 				    IFNAMSIZ) == 0 && a->qname[0] == 0) {
2093 					altq->altq_disc = a->altq_disc;
2094 					break;
2095 				}
2096 			}
2097 		}
2098 
2099 		if ((ifp = ifunit(altq->ifname)) == NULL)
2100 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
2101 		else
2102 			error = altq_add(altq);
2103 
2104 		if (error) {
2105 			PF_RULES_WUNLOCK();
2106 			free(altq, M_PFALTQ);
2107 			break;
2108 		}
2109 
2110 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
2111 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2112 		PF_RULES_WUNLOCK();
2113 		break;
2114 	}
2115 
2116 	case DIOCGETALTQS: {
2117 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2118 		struct pf_altq		*altq;
2119 
2120 		PF_RULES_RLOCK();
2121 		pa->nr = 0;
2122 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
2123 			pa->nr++;
2124 		pa->ticket = V_ticket_altqs_active;
2125 		PF_RULES_RUNLOCK();
2126 		break;
2127 	}
2128 
2129 	case DIOCGETALTQ: {
2130 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2131 		struct pf_altq		*altq;
2132 		u_int32_t		 nr;
2133 
2134 		PF_RULES_RLOCK();
2135 		if (pa->ticket != V_ticket_altqs_active) {
2136 			PF_RULES_RUNLOCK();
2137 			error = EBUSY;
2138 			break;
2139 		}
2140 		nr = 0;
2141 		altq = TAILQ_FIRST(V_pf_altqs_active);
2142 		while ((altq != NULL) && (nr < pa->nr)) {
2143 			altq = TAILQ_NEXT(altq, entries);
2144 			nr++;
2145 		}
2146 		if (altq == NULL) {
2147 			PF_RULES_RUNLOCK();
2148 			error = EBUSY;
2149 			break;
2150 		}
2151 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2152 		PF_RULES_RUNLOCK();
2153 		break;
2154 	}
2155 
2156 	case DIOCCHANGEALTQ:
2157 		/* CHANGEALTQ not supported yet! */
2158 		error = ENODEV;
2159 		break;
2160 
2161 	case DIOCGETQSTATS: {
2162 		struct pfioc_qstats	*pq = (struct pfioc_qstats *)addr;
2163 		struct pf_altq		*altq;
2164 		u_int32_t		 nr;
2165 		int			 nbytes;
2166 
2167 		PF_RULES_RLOCK();
2168 		if (pq->ticket != V_ticket_altqs_active) {
2169 			PF_RULES_RUNLOCK();
2170 			error = EBUSY;
2171 			break;
2172 		}
2173 		nbytes = pq->nbytes;
2174 		nr = 0;
2175 		altq = TAILQ_FIRST(V_pf_altqs_active);
2176 		while ((altq != NULL) && (nr < pq->nr)) {
2177 			altq = TAILQ_NEXT(altq, entries);
2178 			nr++;
2179 		}
2180 		if (altq == NULL) {
2181 			PF_RULES_RUNLOCK();
2182 			error = EBUSY;
2183 			break;
2184 		}
2185 
2186 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
2187 			PF_RULES_RUNLOCK();
2188 			error = ENXIO;
2189 			break;
2190 		}
2191 		PF_RULES_RUNLOCK();
2192 		error = altq_getqstats(altq, pq->buf, &nbytes);
2193 		if (error == 0) {
2194 			pq->scheduler = altq->scheduler;
2195 			pq->nbytes = nbytes;
2196 		}
2197 		break;
2198 	}
2199 #endif /* ALTQ */
2200 
2201 	case DIOCBEGINADDRS: {
2202 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2203 
2204 		PF_RULES_WLOCK();
2205 		pf_empty_pool(&V_pf_pabuf);
2206 		pp->ticket = ++V_ticket_pabuf;
2207 		PF_RULES_WUNLOCK();
2208 		break;
2209 	}
2210 
2211 	case DIOCADDADDR: {
2212 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2213 		struct pf_pooladdr	*pa;
2214 		struct pfi_kif		*kif = NULL;
2215 
2216 #ifndef INET
2217 		if (pp->af == AF_INET) {
2218 			error = EAFNOSUPPORT;
2219 			break;
2220 		}
2221 #endif /* INET */
2222 #ifndef INET6
2223 		if (pp->af == AF_INET6) {
2224 			error = EAFNOSUPPORT;
2225 			break;
2226 		}
2227 #endif /* INET6 */
2228 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2229 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2230 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2231 			error = EINVAL;
2232 			break;
2233 		}
2234 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2235 		bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2236 		if (pa->ifname[0])
2237 			kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2238 		PF_RULES_WLOCK();
2239 		if (pp->ticket != V_ticket_pabuf) {
2240 			PF_RULES_WUNLOCK();
2241 			if (pa->ifname[0])
2242 				free(kif, PFI_MTYPE);
2243 			free(pa, M_PFRULE);
2244 			error = EBUSY;
2245 			break;
2246 		}
2247 		if (pa->ifname[0]) {
2248 			pa->kif = pfi_kif_attach(kif, pa->ifname);
2249 			pfi_kif_ref(pa->kif);
2250 		} else
2251 			pa->kif = NULL;
2252 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2253 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2254 			if (pa->ifname[0])
2255 				pfi_kif_unref(pa->kif);
2256 			PF_RULES_WUNLOCK();
2257 			free(pa, M_PFRULE);
2258 			break;
2259 		}
2260 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2261 		PF_RULES_WUNLOCK();
2262 		break;
2263 	}
2264 
2265 	case DIOCGETADDRS: {
2266 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2267 		struct pf_pool		*pool;
2268 		struct pf_pooladdr	*pa;
2269 
2270 		PF_RULES_RLOCK();
2271 		pp->nr = 0;
2272 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2273 		    pp->r_num, 0, 1, 0);
2274 		if (pool == NULL) {
2275 			PF_RULES_RUNLOCK();
2276 			error = EBUSY;
2277 			break;
2278 		}
2279 		TAILQ_FOREACH(pa, &pool->list, entries)
2280 			pp->nr++;
2281 		PF_RULES_RUNLOCK();
2282 		break;
2283 	}
2284 
2285 	case DIOCGETADDR: {
2286 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2287 		struct pf_pool		*pool;
2288 		struct pf_pooladdr	*pa;
2289 		u_int32_t		 nr = 0;
2290 
2291 		PF_RULES_RLOCK();
2292 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2293 		    pp->r_num, 0, 1, 1);
2294 		if (pool == NULL) {
2295 			PF_RULES_RUNLOCK();
2296 			error = EBUSY;
2297 			break;
2298 		}
2299 		pa = TAILQ_FIRST(&pool->list);
2300 		while ((pa != NULL) && (nr < pp->nr)) {
2301 			pa = TAILQ_NEXT(pa, entries);
2302 			nr++;
2303 		}
2304 		if (pa == NULL) {
2305 			PF_RULES_RUNLOCK();
2306 			error = EBUSY;
2307 			break;
2308 		}
2309 		bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2310 		pf_addr_copyout(&pp->addr.addr);
2311 		PF_RULES_RUNLOCK();
2312 		break;
2313 	}
2314 
2315 	case DIOCCHANGEADDR: {
2316 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
2317 		struct pf_pool		*pool;
2318 		struct pf_pooladdr	*oldpa = NULL, *newpa = NULL;
2319 		struct pf_ruleset	*ruleset;
2320 		struct pfi_kif		*kif = NULL;
2321 
2322 		if (pca->action < PF_CHANGE_ADD_HEAD ||
2323 		    pca->action > PF_CHANGE_REMOVE) {
2324 			error = EINVAL;
2325 			break;
2326 		}
2327 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2328 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2329 		    pca->addr.addr.type != PF_ADDR_TABLE) {
2330 			error = EINVAL;
2331 			break;
2332 		}
2333 
2334 		if (pca->action != PF_CHANGE_REMOVE) {
2335 #ifndef INET
2336 			if (pca->af == AF_INET) {
2337 				error = EAFNOSUPPORT;
2338 				break;
2339 			}
2340 #endif /* INET */
2341 #ifndef INET6
2342 			if (pca->af == AF_INET6) {
2343 				error = EAFNOSUPPORT;
2344 				break;
2345 			}
2346 #endif /* INET6 */
2347 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
2348 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2349 			if (newpa->ifname[0])
2350 				kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2351 			newpa->kif = NULL;
2352 		}
2353 
2354 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGEADDR_error; }
2355 		PF_RULES_WLOCK();
2356 		ruleset = pf_find_ruleset(pca->anchor);
2357 		if (ruleset == NULL)
2358 			ERROUT(EBUSY);
2359 
2360 		pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2361 		    pca->r_num, pca->r_last, 1, 1);
2362 		if (pool == NULL)
2363 			ERROUT(EBUSY);
2364 
2365 		if (pca->action != PF_CHANGE_REMOVE) {
2366 			if (newpa->ifname[0]) {
2367 				newpa->kif = pfi_kif_attach(kif, newpa->ifname);
2368 				pfi_kif_ref(newpa->kif);
2369 				kif = NULL;
2370 			}
2371 
2372 			switch (newpa->addr.type) {
2373 			case PF_ADDR_DYNIFTL:
2374 				error = pfi_dynaddr_setup(&newpa->addr,
2375 				    pca->af);
2376 				break;
2377 			case PF_ADDR_TABLE:
2378 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
2379 				    newpa->addr.v.tblname);
2380 				if (newpa->addr.p.tbl == NULL)
2381 					error = ENOMEM;
2382 				break;
2383 			}
2384 			if (error)
2385 				goto DIOCCHANGEADDR_error;
2386 		}
2387 
2388 		switch (pca->action) {
2389 		case PF_CHANGE_ADD_HEAD:
2390 			oldpa = TAILQ_FIRST(&pool->list);
2391 			break;
2392 		case PF_CHANGE_ADD_TAIL:
2393 			oldpa = TAILQ_LAST(&pool->list, pf_palist);
2394 			break;
2395 		default:
2396 			oldpa = TAILQ_FIRST(&pool->list);
2397 			for (int i = 0; oldpa && i < pca->nr; i++)
2398 				oldpa = TAILQ_NEXT(oldpa, entries);
2399 
2400 			if (oldpa == NULL)
2401 				ERROUT(EINVAL);
2402 		}
2403 
2404 		if (pca->action == PF_CHANGE_REMOVE) {
2405 			TAILQ_REMOVE(&pool->list, oldpa, entries);
2406 			switch (oldpa->addr.type) {
2407 			case PF_ADDR_DYNIFTL:
2408 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
2409 				break;
2410 			case PF_ADDR_TABLE:
2411 				pfr_detach_table(oldpa->addr.p.tbl);
2412 				break;
2413 			}
2414 			if (oldpa->kif)
2415 				pfi_kif_unref(oldpa->kif);
2416 			free(oldpa, M_PFRULE);
2417 		} else {
2418 			if (oldpa == NULL)
2419 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2420 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
2421 			    pca->action == PF_CHANGE_ADD_BEFORE)
2422 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2423 			else
2424 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
2425 				    newpa, entries);
2426 		}
2427 
2428 		pool->cur = TAILQ_FIRST(&pool->list);
2429 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
2430 		PF_RULES_WUNLOCK();
2431 		break;
2432 
2433 #undef ERROUT
2434 DIOCCHANGEADDR_error:
2435 		if (newpa != NULL) {
2436 			if (newpa->kif)
2437 				pfi_kif_unref(newpa->kif);
2438 			free(newpa, M_PFRULE);
2439 		}
2440 		PF_RULES_WUNLOCK();
2441 		if (kif != NULL)
2442 			free(kif, PFI_MTYPE);
2443 		break;
2444 	}
2445 
2446 	case DIOCGETRULESETS: {
2447 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2448 		struct pf_ruleset	*ruleset;
2449 		struct pf_anchor	*anchor;
2450 
2451 		PF_RULES_RLOCK();
2452 		pr->path[sizeof(pr->path) - 1] = 0;
2453 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2454 			PF_RULES_RUNLOCK();
2455 			error = ENOENT;
2456 			break;
2457 		}
2458 		pr->nr = 0;
2459 		if (ruleset->anchor == NULL) {
2460 			/* XXX kludge for pf_main_ruleset */
2461 			RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2462 				if (anchor->parent == NULL)
2463 					pr->nr++;
2464 		} else {
2465 			RB_FOREACH(anchor, pf_anchor_node,
2466 			    &ruleset->anchor->children)
2467 				pr->nr++;
2468 		}
2469 		PF_RULES_RUNLOCK();
2470 		break;
2471 	}
2472 
2473 	case DIOCGETRULESET: {
2474 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2475 		struct pf_ruleset	*ruleset;
2476 		struct pf_anchor	*anchor;
2477 		u_int32_t		 nr = 0;
2478 
2479 		PF_RULES_RLOCK();
2480 		pr->path[sizeof(pr->path) - 1] = 0;
2481 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2482 			PF_RULES_RUNLOCK();
2483 			error = ENOENT;
2484 			break;
2485 		}
2486 		pr->name[0] = 0;
2487 		if (ruleset->anchor == NULL) {
2488 			/* XXX kludge for pf_main_ruleset */
2489 			RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2490 				if (anchor->parent == NULL && nr++ == pr->nr) {
2491 					strlcpy(pr->name, anchor->name,
2492 					    sizeof(pr->name));
2493 					break;
2494 				}
2495 		} else {
2496 			RB_FOREACH(anchor, pf_anchor_node,
2497 			    &ruleset->anchor->children)
2498 				if (nr++ == pr->nr) {
2499 					strlcpy(pr->name, anchor->name,
2500 					    sizeof(pr->name));
2501 					break;
2502 				}
2503 		}
2504 		if (!pr->name[0])
2505 			error = EBUSY;
2506 		PF_RULES_RUNLOCK();
2507 		break;
2508 	}
2509 
2510 	case DIOCRCLRTABLES: {
2511 		struct pfioc_table *io = (struct pfioc_table *)addr;
2512 
2513 		if (io->pfrio_esize != 0) {
2514 			error = ENODEV;
2515 			break;
2516 		}
2517 		PF_RULES_WLOCK();
2518 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2519 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2520 		PF_RULES_WUNLOCK();
2521 		break;
2522 	}
2523 
2524 	case DIOCRADDTABLES: {
2525 		struct pfioc_table *io = (struct pfioc_table *)addr;
2526 		struct pfr_table *pfrts;
2527 		size_t totlen;
2528 
2529 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2530 			error = ENODEV;
2531 			break;
2532 		}
2533 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2534 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2535 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2536 		if (error) {
2537 			free(pfrts, M_TEMP);
2538 			break;
2539 		}
2540 		PF_RULES_WLOCK();
2541 		error = pfr_add_tables(pfrts, io->pfrio_size,
2542 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2543 		PF_RULES_WUNLOCK();
2544 		free(pfrts, M_TEMP);
2545 		break;
2546 	}
2547 
2548 	case DIOCRDELTABLES: {
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_del_tables(pfrts, io->pfrio_size,
2566 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2567 		PF_RULES_WUNLOCK();
2568 		free(pfrts, M_TEMP);
2569 		break;
2570 	}
2571 
2572 	case DIOCRGETTABLES: {
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 		PF_RULES_RLOCK();
2584 		error = pfr_get_tables(&io->pfrio_table, pfrts,
2585 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2586 		PF_RULES_RUNLOCK();
2587 		if (error == 0)
2588 			error = copyout(pfrts, io->pfrio_buffer, totlen);
2589 		free(pfrts, M_TEMP);
2590 		break;
2591 	}
2592 
2593 	case DIOCRGETTSTATS: {
2594 		struct pfioc_table *io = (struct pfioc_table *)addr;
2595 		struct pfr_tstats *pfrtstats;
2596 		size_t totlen;
2597 
2598 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2599 			error = ENODEV;
2600 			break;
2601 		}
2602 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
2603 		pfrtstats = malloc(totlen, M_TEMP, M_WAITOK);
2604 		PF_RULES_WLOCK();
2605 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
2606 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2607 		PF_RULES_WUNLOCK();
2608 		if (error == 0)
2609 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
2610 		free(pfrtstats, M_TEMP);
2611 		break;
2612 	}
2613 
2614 	case DIOCRCLRTSTATS: {
2615 		struct pfioc_table *io = (struct pfioc_table *)addr;
2616 		struct pfr_table *pfrts;
2617 		size_t totlen;
2618 
2619 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2620 			error = ENODEV;
2621 			break;
2622 		}
2623 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2624 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2625 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2626 		if (error) {
2627 			free(pfrts, M_TEMP);
2628 			break;
2629 		}
2630 		PF_RULES_WLOCK();
2631 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
2632 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2633 		PF_RULES_WUNLOCK();
2634 		free(pfrts, M_TEMP);
2635 		break;
2636 	}
2637 
2638 	case DIOCRSETTFLAGS: {
2639 		struct pfioc_table *io = (struct pfioc_table *)addr;
2640 		struct pfr_table *pfrts;
2641 		size_t totlen;
2642 
2643 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2644 			error = ENODEV;
2645 			break;
2646 		}
2647 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2648 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2649 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2650 		if (error) {
2651 			free(pfrts, M_TEMP);
2652 			break;
2653 		}
2654 		PF_RULES_WLOCK();
2655 		error = pfr_set_tflags(pfrts, io->pfrio_size,
2656 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2657 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2658 		PF_RULES_WUNLOCK();
2659 		free(pfrts, M_TEMP);
2660 		break;
2661 	}
2662 
2663 	case DIOCRCLRADDRS: {
2664 		struct pfioc_table *io = (struct pfioc_table *)addr;
2665 
2666 		if (io->pfrio_esize != 0) {
2667 			error = ENODEV;
2668 			break;
2669 		}
2670 		PF_RULES_WLOCK();
2671 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2672 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2673 		PF_RULES_WUNLOCK();
2674 		break;
2675 	}
2676 
2677 	case DIOCRADDADDRS: {
2678 		struct pfioc_table *io = (struct pfioc_table *)addr;
2679 		struct pfr_addr *pfras;
2680 		size_t totlen;
2681 
2682 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2683 			error = ENODEV;
2684 			break;
2685 		}
2686 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2687 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2688 		error = copyin(io->pfrio_buffer, pfras, totlen);
2689 		if (error) {
2690 			free(pfras, M_TEMP);
2691 			break;
2692 		}
2693 		PF_RULES_WLOCK();
2694 		error = pfr_add_addrs(&io->pfrio_table, pfras,
2695 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2696 		    PFR_FLAG_USERIOCTL);
2697 		PF_RULES_WUNLOCK();
2698 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2699 			error = copyout(pfras, io->pfrio_buffer, totlen);
2700 		free(pfras, M_TEMP);
2701 		break;
2702 	}
2703 
2704 	case DIOCRDELADDRS: {
2705 		struct pfioc_table *io = (struct pfioc_table *)addr;
2706 		struct pfr_addr *pfras;
2707 		size_t totlen;
2708 
2709 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2710 			error = ENODEV;
2711 			break;
2712 		}
2713 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2714 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2715 		error = copyin(io->pfrio_buffer, pfras, totlen);
2716 		if (error) {
2717 			free(pfras, M_TEMP);
2718 			break;
2719 		}
2720 		PF_RULES_WLOCK();
2721 		error = pfr_del_addrs(&io->pfrio_table, pfras,
2722 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2723 		    PFR_FLAG_USERIOCTL);
2724 		PF_RULES_WUNLOCK();
2725 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2726 			error = copyout(pfras, io->pfrio_buffer, totlen);
2727 		free(pfras, M_TEMP);
2728 		break;
2729 	}
2730 
2731 	case DIOCRSETADDRS: {
2732 		struct pfioc_table *io = (struct pfioc_table *)addr;
2733 		struct pfr_addr *pfras;
2734 		size_t totlen, count;
2735 
2736 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2737 			error = ENODEV;
2738 			break;
2739 		}
2740 		count = max(io->pfrio_size, io->pfrio_size2);
2741 		totlen = count * sizeof(struct pfr_addr);
2742 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2743 		error = copyin(io->pfrio_buffer, pfras, totlen);
2744 		if (error) {
2745 			free(pfras, M_TEMP);
2746 			break;
2747 		}
2748 		PF_RULES_WLOCK();
2749 		error = pfr_set_addrs(&io->pfrio_table, pfras,
2750 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2751 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2752 		    PFR_FLAG_USERIOCTL, 0);
2753 		PF_RULES_WUNLOCK();
2754 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2755 			error = copyout(pfras, io->pfrio_buffer, totlen);
2756 		free(pfras, M_TEMP);
2757 		break;
2758 	}
2759 
2760 	case DIOCRGETADDRS: {
2761 		struct pfioc_table *io = (struct pfioc_table *)addr;
2762 		struct pfr_addr *pfras;
2763 		size_t totlen;
2764 
2765 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2766 			error = ENODEV;
2767 			break;
2768 		}
2769 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2770 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2771 		PF_RULES_RLOCK();
2772 		error = pfr_get_addrs(&io->pfrio_table, pfras,
2773 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2774 		PF_RULES_RUNLOCK();
2775 		if (error == 0)
2776 			error = copyout(pfras, io->pfrio_buffer, totlen);
2777 		free(pfras, M_TEMP);
2778 		break;
2779 	}
2780 
2781 	case DIOCRGETASTATS: {
2782 		struct pfioc_table *io = (struct pfioc_table *)addr;
2783 		struct pfr_astats *pfrastats;
2784 		size_t totlen;
2785 
2786 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2787 			error = ENODEV;
2788 			break;
2789 		}
2790 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
2791 		pfrastats = malloc(totlen, M_TEMP, M_WAITOK);
2792 		PF_RULES_RLOCK();
2793 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
2794 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2795 		PF_RULES_RUNLOCK();
2796 		if (error == 0)
2797 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
2798 		free(pfrastats, M_TEMP);
2799 		break;
2800 	}
2801 
2802 	case DIOCRCLRASTATS: {
2803 		struct pfioc_table *io = (struct pfioc_table *)addr;
2804 		struct pfr_addr *pfras;
2805 		size_t totlen;
2806 
2807 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2808 			error = ENODEV;
2809 			break;
2810 		}
2811 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2812 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2813 		error = copyin(io->pfrio_buffer, pfras, totlen);
2814 		if (error) {
2815 			free(pfras, M_TEMP);
2816 			break;
2817 		}
2818 		PF_RULES_WLOCK();
2819 		error = pfr_clr_astats(&io->pfrio_table, pfras,
2820 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2821 		    PFR_FLAG_USERIOCTL);
2822 		PF_RULES_WUNLOCK();
2823 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2824 			error = copyout(pfras, io->pfrio_buffer, totlen);
2825 		free(pfras, M_TEMP);
2826 		break;
2827 	}
2828 
2829 	case DIOCRTSTADDRS: {
2830 		struct pfioc_table *io = (struct pfioc_table *)addr;
2831 		struct pfr_addr *pfras;
2832 		size_t totlen;
2833 
2834 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2835 			error = ENODEV;
2836 			break;
2837 		}
2838 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2839 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2840 		error = copyin(io->pfrio_buffer, pfras, totlen);
2841 		if (error) {
2842 			free(pfras, M_TEMP);
2843 			break;
2844 		}
2845 		PF_RULES_RLOCK();
2846 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
2847 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2848 		    PFR_FLAG_USERIOCTL);
2849 		PF_RULES_RUNLOCK();
2850 		if (error == 0)
2851 			error = copyout(pfras, io->pfrio_buffer, totlen);
2852 		free(pfras, M_TEMP);
2853 		break;
2854 	}
2855 
2856 	case DIOCRINADEFINE: {
2857 		struct pfioc_table *io = (struct pfioc_table *)addr;
2858 		struct pfr_addr *pfras;
2859 		size_t totlen;
2860 
2861 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2862 			error = ENODEV;
2863 			break;
2864 		}
2865 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2866 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2867 		error = copyin(io->pfrio_buffer, pfras, totlen);
2868 		if (error) {
2869 			free(pfras, M_TEMP);
2870 			break;
2871 		}
2872 		PF_RULES_WLOCK();
2873 		error = pfr_ina_define(&io->pfrio_table, pfras,
2874 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2875 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2876 		PF_RULES_WUNLOCK();
2877 		free(pfras, M_TEMP);
2878 		break;
2879 	}
2880 
2881 	case DIOCOSFPADD: {
2882 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2883 		PF_RULES_WLOCK();
2884 		error = pf_osfp_add(io);
2885 		PF_RULES_WUNLOCK();
2886 		break;
2887 	}
2888 
2889 	case DIOCOSFPGET: {
2890 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2891 		PF_RULES_RLOCK();
2892 		error = pf_osfp_get(io);
2893 		PF_RULES_RUNLOCK();
2894 		break;
2895 	}
2896 
2897 	case DIOCXBEGIN: {
2898 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2899 		struct pfioc_trans_e	*ioes, *ioe;
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_begin_altq(&ioe->ticket))) {
2926 					PF_RULES_WUNLOCK();
2927 					free(ioes, M_TEMP);
2928 					goto fail;
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_begin(&table,
2940 				    &ioe->ticket, NULL, 0))) {
2941 					PF_RULES_WUNLOCK();
2942 					free(ioes, M_TEMP);
2943 					goto fail;
2944 				}
2945 				break;
2946 			    }
2947 			default:
2948 				if ((error = pf_begin_rules(&ioe->ticket,
2949 				    ioe->rs_num, ioe->anchor))) {
2950 					PF_RULES_WUNLOCK();
2951 					free(ioes, M_TEMP);
2952 					goto fail;
2953 				}
2954 				break;
2955 			}
2956 		}
2957 		PF_RULES_WUNLOCK();
2958 		error = copyout(ioes, io->array, totlen);
2959 		free(ioes, M_TEMP);
2960 		break;
2961 	}
2962 
2963 	case DIOCXROLLBACK: {
2964 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2965 		struct pfioc_trans_e	*ioe, *ioes;
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 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2982 			switch (ioe->rs_num) {
2983 #ifdef ALTQ
2984 			case PF_RULESET_ALTQ:
2985 				if (ioe->anchor[0]) {
2986 					PF_RULES_WUNLOCK();
2987 					free(ioes, M_TEMP);
2988 					error = EINVAL;
2989 					goto fail;
2990 				}
2991 				if ((error = pf_rollback_altq(ioe->ticket))) {
2992 					PF_RULES_WUNLOCK();
2993 					free(ioes, M_TEMP);
2994 					goto fail; /* really bad */
2995 				}
2996 				break;
2997 #endif /* ALTQ */
2998 			case PF_RULESET_TABLE:
2999 			    {
3000 				struct pfr_table table;
3001 
3002 				bzero(&table, sizeof(table));
3003 				strlcpy(table.pfrt_anchor, ioe->anchor,
3004 				    sizeof(table.pfrt_anchor));
3005 				if ((error = pfr_ina_rollback(&table,
3006 				    ioe->ticket, NULL, 0))) {
3007 					PF_RULES_WUNLOCK();
3008 					free(ioes, M_TEMP);
3009 					goto fail; /* really bad */
3010 				}
3011 				break;
3012 			    }
3013 			default:
3014 				if ((error = pf_rollback_rules(ioe->ticket,
3015 				    ioe->rs_num, ioe->anchor))) {
3016 					PF_RULES_WUNLOCK();
3017 					free(ioes, M_TEMP);
3018 					goto fail; /* really bad */
3019 				}
3020 				break;
3021 			}
3022 		}
3023 		PF_RULES_WUNLOCK();
3024 		free(ioes, M_TEMP);
3025 		break;
3026 	}
3027 
3028 	case DIOCXCOMMIT: {
3029 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3030 		struct pfioc_trans_e	*ioe, *ioes;
3031 		struct pf_ruleset	*rs;
3032 		size_t			 totlen;
3033 		int			 i;
3034 
3035 		if (io->esize != sizeof(*ioe)) {
3036 			error = ENODEV;
3037 			break;
3038 		}
3039 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3040 		ioes = malloc(totlen, M_TEMP, M_WAITOK);
3041 		error = copyin(io->array, ioes, totlen);
3042 		if (error) {
3043 			free(ioes, M_TEMP);
3044 			break;
3045 		}
3046 		PF_RULES_WLOCK();
3047 		/* First makes sure everything will succeed. */
3048 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3049 			switch (ioe->rs_num) {
3050 #ifdef ALTQ
3051 			case PF_RULESET_ALTQ:
3052 				if (ioe->anchor[0]) {
3053 					PF_RULES_WUNLOCK();
3054 					free(ioes, M_TEMP);
3055 					error = EINVAL;
3056 					goto fail;
3057 				}
3058 				if (!V_altqs_inactive_open || ioe->ticket !=
3059 				    V_ticket_altqs_inactive) {
3060 					PF_RULES_WUNLOCK();
3061 					free(ioes, M_TEMP);
3062 					error = EBUSY;
3063 					goto fail;
3064 				}
3065 				break;
3066 #endif /* ALTQ */
3067 			case PF_RULESET_TABLE:
3068 				rs = pf_find_ruleset(ioe->anchor);
3069 				if (rs == NULL || !rs->topen || ioe->ticket !=
3070 				    rs->tticket) {
3071 					PF_RULES_WUNLOCK();
3072 					free(ioes, M_TEMP);
3073 					error = EBUSY;
3074 					goto fail;
3075 				}
3076 				break;
3077 			default:
3078 				if (ioe->rs_num < 0 || ioe->rs_num >=
3079 				    PF_RULESET_MAX) {
3080 					PF_RULES_WUNLOCK();
3081 					free(ioes, M_TEMP);
3082 					error = EINVAL;
3083 					goto fail;
3084 				}
3085 				rs = pf_find_ruleset(ioe->anchor);
3086 				if (rs == NULL ||
3087 				    !rs->rules[ioe->rs_num].inactive.open ||
3088 				    rs->rules[ioe->rs_num].inactive.ticket !=
3089 				    ioe->ticket) {
3090 					PF_RULES_WUNLOCK();
3091 					free(ioes, M_TEMP);
3092 					error = EBUSY;
3093 					goto fail;
3094 				}
3095 				break;
3096 			}
3097 		}
3098 		/* Now do the commit - no errors should happen here. */
3099 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3100 			switch (ioe->rs_num) {
3101 #ifdef ALTQ
3102 			case PF_RULESET_ALTQ:
3103 				if ((error = pf_commit_altq(ioe->ticket))) {
3104 					PF_RULES_WUNLOCK();
3105 					free(ioes, M_TEMP);
3106 					goto fail; /* really bad */
3107 				}
3108 				break;
3109 #endif /* ALTQ */
3110 			case PF_RULESET_TABLE:
3111 			    {
3112 				struct pfr_table table;
3113 
3114 				bzero(&table, sizeof(table));
3115 				strlcpy(table.pfrt_anchor, ioe->anchor,
3116 				    sizeof(table.pfrt_anchor));
3117 				if ((error = pfr_ina_commit(&table,
3118 				    ioe->ticket, NULL, NULL, 0))) {
3119 					PF_RULES_WUNLOCK();
3120 					free(ioes, M_TEMP);
3121 					goto fail; /* really bad */
3122 				}
3123 				break;
3124 			    }
3125 			default:
3126 				if ((error = pf_commit_rules(ioe->ticket,
3127 				    ioe->rs_num, ioe->anchor))) {
3128 					PF_RULES_WUNLOCK();
3129 					free(ioes, M_TEMP);
3130 					goto fail; /* really bad */
3131 				}
3132 				break;
3133 			}
3134 		}
3135 		PF_RULES_WUNLOCK();
3136 		free(ioes, M_TEMP);
3137 		break;
3138 	}
3139 
3140 	case DIOCGETSRCNODES: {
3141 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
3142 		struct pf_srchash	*sh;
3143 		struct pf_src_node	*n, *p, *pstore;
3144 		uint32_t		 i, nr = 0;
3145 
3146 		if (psn->psn_len == 0) {
3147 			for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3148 			    i++, sh++) {
3149 				PF_HASHROW_LOCK(sh);
3150 				LIST_FOREACH(n, &sh->nodes, entry)
3151 					nr++;
3152 				PF_HASHROW_UNLOCK(sh);
3153 			}
3154 			psn->psn_len = sizeof(struct pf_src_node) * nr;
3155 			break;
3156 		}
3157 
3158 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK);
3159 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3160 		    i++, sh++) {
3161 		    PF_HASHROW_LOCK(sh);
3162 		    LIST_FOREACH(n, &sh->nodes, entry) {
3163 			int	secs = time_uptime, diff;
3164 
3165 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
3166 				break;
3167 
3168 			bcopy(n, p, sizeof(struct pf_src_node));
3169 			if (n->rule.ptr != NULL)
3170 				p->rule.nr = n->rule.ptr->nr;
3171 			p->creation = secs - p->creation;
3172 			if (p->expire > secs)
3173 				p->expire -= secs;
3174 			else
3175 				p->expire = 0;
3176 
3177 			/* Adjust the connection rate estimate. */
3178 			diff = secs - n->conn_rate.last;
3179 			if (diff >= n->conn_rate.seconds)
3180 				p->conn_rate.count = 0;
3181 			else
3182 				p->conn_rate.count -=
3183 				    n->conn_rate.count * diff /
3184 				    n->conn_rate.seconds;
3185 			p++;
3186 			nr++;
3187 		    }
3188 		    PF_HASHROW_UNLOCK(sh);
3189 		}
3190 		error = copyout(pstore, psn->psn_src_nodes,
3191 		    sizeof(struct pf_src_node) * nr);
3192 		if (error) {
3193 			free(pstore, M_TEMP);
3194 			break;
3195 		}
3196 		psn->psn_len = sizeof(struct pf_src_node) * nr;
3197 		free(pstore, M_TEMP);
3198 		break;
3199 	}
3200 
3201 	case DIOCCLRSRCNODES: {
3202 
3203 		pf_clear_srcnodes(NULL);
3204 		pf_purge_expired_src_nodes();
3205 		break;
3206 	}
3207 
3208 	case DIOCKILLSRCNODES:
3209 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
3210 		break;
3211 
3212 	case DIOCSETHOSTID: {
3213 		u_int32_t	*hostid = (u_int32_t *)addr;
3214 
3215 		PF_RULES_WLOCK();
3216 		if (*hostid == 0)
3217 			V_pf_status.hostid = arc4random();
3218 		else
3219 			V_pf_status.hostid = *hostid;
3220 		PF_RULES_WUNLOCK();
3221 		break;
3222 	}
3223 
3224 	case DIOCOSFPFLUSH:
3225 		PF_RULES_WLOCK();
3226 		pf_osfp_flush();
3227 		PF_RULES_WUNLOCK();
3228 		break;
3229 
3230 	case DIOCIGETIFACES: {
3231 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3232 		struct pfi_kif *ifstore;
3233 		size_t bufsiz;
3234 
3235 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
3236 			error = ENODEV;
3237 			break;
3238 		}
3239 
3240 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
3241 		ifstore = malloc(bufsiz, M_TEMP, M_WAITOK);
3242 		PF_RULES_RLOCK();
3243 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
3244 		PF_RULES_RUNLOCK();
3245 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
3246 		free(ifstore, M_TEMP);
3247 		break;
3248 	}
3249 
3250 	case DIOCSETIFFLAG: {
3251 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3252 
3253 		PF_RULES_WLOCK();
3254 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3255 		PF_RULES_WUNLOCK();
3256 		break;
3257 	}
3258 
3259 	case DIOCCLRIFFLAG: {
3260 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3261 
3262 		PF_RULES_WLOCK();
3263 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3264 		PF_RULES_WUNLOCK();
3265 		break;
3266 	}
3267 
3268 	default:
3269 		error = ENODEV;
3270 		break;
3271 	}
3272 fail:
3273 	if (sx_xlocked(&pf_ioctl_lock))
3274 		sx_xunlock(&pf_ioctl_lock);
3275 	CURVNET_RESTORE();
3276 
3277 	return (error);
3278 }
3279 
3280 void
3281 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
3282 {
3283 	bzero(sp, sizeof(struct pfsync_state));
3284 
3285 	/* copy from state key */
3286 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
3287 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
3288 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
3289 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
3290 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
3291 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
3292 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
3293 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
3294 	sp->proto = st->key[PF_SK_WIRE]->proto;
3295 	sp->af = st->key[PF_SK_WIRE]->af;
3296 
3297 	/* copy from state */
3298 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
3299 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
3300 	sp->creation = htonl(time_uptime - st->creation);
3301 	sp->expire = pf_state_expires(st);
3302 	if (sp->expire <= time_uptime)
3303 		sp->expire = htonl(0);
3304 	else
3305 		sp->expire = htonl(sp->expire - time_uptime);
3306 
3307 	sp->direction = st->direction;
3308 	sp->log = st->log;
3309 	sp->timeout = st->timeout;
3310 	sp->state_flags = st->state_flags;
3311 	if (st->src_node)
3312 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
3313 	if (st->nat_src_node)
3314 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
3315 
3316 	sp->id = st->id;
3317 	sp->creatorid = st->creatorid;
3318 	pf_state_peer_hton(&st->src, &sp->src);
3319 	pf_state_peer_hton(&st->dst, &sp->dst);
3320 
3321 	if (st->rule.ptr == NULL)
3322 		sp->rule = htonl(-1);
3323 	else
3324 		sp->rule = htonl(st->rule.ptr->nr);
3325 	if (st->anchor.ptr == NULL)
3326 		sp->anchor = htonl(-1);
3327 	else
3328 		sp->anchor = htonl(st->anchor.ptr->nr);
3329 	if (st->nat_rule.ptr == NULL)
3330 		sp->nat_rule = htonl(-1);
3331 	else
3332 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
3333 
3334 	pf_state_counter_hton(st->packets[0], sp->packets[0]);
3335 	pf_state_counter_hton(st->packets[1], sp->packets[1]);
3336 	pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
3337 	pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
3338 
3339 }
3340 
3341 static void
3342 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
3343 {
3344 	struct pfr_ktable *kt;
3345 
3346 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
3347 
3348 	kt = aw->p.tbl;
3349 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
3350 		kt = kt->pfrkt_root;
3351 	aw->p.tbl = NULL;
3352 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
3353 		kt->pfrkt_cnt : -1;
3354 }
3355 
3356 /*
3357  * XXX - Check for version missmatch!!!
3358  */
3359 static void
3360 pf_clear_states(void)
3361 {
3362 	struct pf_state	*s;
3363 	u_int i;
3364 
3365 	for (i = 0; i <= pf_hashmask; i++) {
3366 		struct pf_idhash *ih = &V_pf_idhash[i];
3367 relock:
3368 		PF_HASHROW_LOCK(ih);
3369 		LIST_FOREACH(s, &ih->states, entry) {
3370 			s->timeout = PFTM_PURGE;
3371 			/* Don't send out individual delete messages. */
3372 			s->state_flags |= PFSTATE_NOSYNC;
3373 			pf_unlink_state(s, PF_ENTER_LOCKED);
3374 			goto relock;
3375 		}
3376 		PF_HASHROW_UNLOCK(ih);
3377 	}
3378 }
3379 
3380 static int
3381 pf_clear_tables(void)
3382 {
3383 	struct pfioc_table io;
3384 	int error;
3385 
3386 	bzero(&io, sizeof(io));
3387 
3388 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
3389 	    io.pfrio_flags);
3390 
3391 	return (error);
3392 }
3393 
3394 static void
3395 pf_clear_srcnodes(struct pf_src_node *n)
3396 {
3397 	struct pf_state *s;
3398 	int i;
3399 
3400 	for (i = 0; i <= pf_hashmask; i++) {
3401 		struct pf_idhash *ih = &V_pf_idhash[i];
3402 
3403 		PF_HASHROW_LOCK(ih);
3404 		LIST_FOREACH(s, &ih->states, entry) {
3405 			if (n == NULL || n == s->src_node)
3406 				s->src_node = NULL;
3407 			if (n == NULL || n == s->nat_src_node)
3408 				s->nat_src_node = NULL;
3409 		}
3410 		PF_HASHROW_UNLOCK(ih);
3411 	}
3412 
3413 	if (n == NULL) {
3414 		struct pf_srchash *sh;
3415 
3416 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3417 		    i++, sh++) {
3418 			PF_HASHROW_LOCK(sh);
3419 			LIST_FOREACH(n, &sh->nodes, entry) {
3420 				n->expire = 1;
3421 				n->states = 0;
3422 			}
3423 			PF_HASHROW_UNLOCK(sh);
3424 		}
3425 	} else {
3426 		/* XXX: hash slot should already be locked here. */
3427 		n->expire = 1;
3428 		n->states = 0;
3429 	}
3430 }
3431 
3432 static void
3433 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
3434 {
3435 	struct pf_src_node_list	 kill;
3436 
3437 	LIST_INIT(&kill);
3438 	for (int i = 0; i <= pf_srchashmask; i++) {
3439 		struct pf_srchash *sh = &V_pf_srchash[i];
3440 		struct pf_src_node *sn, *tmp;
3441 
3442 		PF_HASHROW_LOCK(sh);
3443 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
3444 			if (PF_MATCHA(psnk->psnk_src.neg,
3445 			      &psnk->psnk_src.addr.v.a.addr,
3446 			      &psnk->psnk_src.addr.v.a.mask,
3447 			      &sn->addr, sn->af) &&
3448 			    PF_MATCHA(psnk->psnk_dst.neg,
3449 			      &psnk->psnk_dst.addr.v.a.addr,
3450 			      &psnk->psnk_dst.addr.v.a.mask,
3451 			      &sn->raddr, sn->af)) {
3452 				pf_unlink_src_node(sn);
3453 				LIST_INSERT_HEAD(&kill, sn, entry);
3454 				sn->expire = 1;
3455 			}
3456 		PF_HASHROW_UNLOCK(sh);
3457 	}
3458 
3459 	for (int i = 0; i <= pf_hashmask; i++) {
3460 		struct pf_idhash *ih = &V_pf_idhash[i];
3461 		struct pf_state *s;
3462 
3463 		PF_HASHROW_LOCK(ih);
3464 		LIST_FOREACH(s, &ih->states, entry) {
3465 			if (s->src_node && s->src_node->expire == 1)
3466 				s->src_node = NULL;
3467 			if (s->nat_src_node && s->nat_src_node->expire == 1)
3468 				s->nat_src_node = NULL;
3469 		}
3470 		PF_HASHROW_UNLOCK(ih);
3471 	}
3472 
3473 	psnk->psnk_killed = pf_free_src_nodes(&kill);
3474 }
3475 
3476 /*
3477  * XXX - Check for version missmatch!!!
3478  */
3479 
3480 /*
3481  * Duplicate pfctl -Fa operation to get rid of as much as we can.
3482  */
3483 static int
3484 shutdown_pf(void)
3485 {
3486 	int error = 0;
3487 	u_int32_t t[5];
3488 	char nn = '\0';
3489 
3490 	do {
3491 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
3492 		    != 0) {
3493 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
3494 			break;
3495 		}
3496 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
3497 		    != 0) {
3498 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
3499 			break;		/* XXX: rollback? */
3500 		}
3501 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
3502 		    != 0) {
3503 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
3504 			break;		/* XXX: rollback? */
3505 		}
3506 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
3507 		    != 0) {
3508 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
3509 			break;		/* XXX: rollback? */
3510 		}
3511 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
3512 		    != 0) {
3513 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
3514 			break;		/* XXX: rollback? */
3515 		}
3516 
3517 		/* XXX: these should always succeed here */
3518 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
3519 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
3520 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
3521 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
3522 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
3523 
3524 		if ((error = pf_clear_tables()) != 0)
3525 			break;
3526 
3527 #ifdef ALTQ
3528 		if ((error = pf_begin_altq(&t[0])) != 0) {
3529 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
3530 			break;
3531 		}
3532 		pf_commit_altq(t[0]);
3533 #endif
3534 
3535 		pf_clear_states();
3536 
3537 		pf_clear_srcnodes(NULL);
3538 
3539 		/* status does not use malloced mem so no need to cleanup */
3540 		/* fingerprints and interfaces have their own cleanup code */
3541 
3542 		/* Free counters last as we updated them during shutdown. */
3543 		counter_u64_free(V_pf_default_rule.states_cur);
3544 		counter_u64_free(V_pf_default_rule.states_tot);
3545 		counter_u64_free(V_pf_default_rule.src_nodes);
3546 
3547 		for (int i = 0; i < PFRES_MAX; i++)
3548 			counter_u64_free(V_pf_status.counters[i]);
3549 		for (int i = 0; i < LCNT_MAX; i++)
3550 			counter_u64_free(V_pf_status.lcounters[i]);
3551 		for (int i = 0; i < FCNT_MAX; i++)
3552 			counter_u64_free(V_pf_status.fcounters[i]);
3553 		for (int i = 0; i < SCNT_MAX; i++)
3554 			counter_u64_free(V_pf_status.scounters[i]);
3555 	} while(0);
3556 
3557 	return (error);
3558 }
3559 
3560 #ifdef INET
3561 static int
3562 pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3563     struct inpcb *inp)
3564 {
3565 	int chk;
3566 
3567 	chk = pf_test(PF_IN, ifp, m, inp);
3568 	if (chk && *m) {
3569 		m_freem(*m);
3570 		*m = NULL;
3571 	}
3572 
3573 	if (chk != PF_PASS)
3574 		return (EACCES);
3575 	return (0);
3576 }
3577 
3578 static int
3579 pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3580     struct inpcb *inp)
3581 {
3582 	int chk;
3583 
3584 	chk = pf_test(PF_OUT, ifp, m, inp);
3585 	if (chk && *m) {
3586 		m_freem(*m);
3587 		*m = NULL;
3588 	}
3589 
3590 	if (chk != PF_PASS)
3591 		return (EACCES);
3592 	return (0);
3593 }
3594 #endif
3595 
3596 #ifdef INET6
3597 static int
3598 pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3599     struct inpcb *inp)
3600 {
3601 	int chk;
3602 
3603 	/*
3604 	 * In case of loopback traffic IPv6 uses the real interface in
3605 	 * order to support scoped addresses. In order to support stateful
3606 	 * filtering we have change this to lo0 as it is the case in IPv4.
3607 	 */
3608 	CURVNET_SET(ifp->if_vnet);
3609 	chk = pf_test6(PF_IN, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
3610 	CURVNET_RESTORE();
3611 	if (chk && *m) {
3612 		m_freem(*m);
3613 		*m = NULL;
3614 	}
3615 	if (chk != PF_PASS)
3616 		return (EACCES);
3617 	return (0);
3618 }
3619 
3620 static int
3621 pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3622     struct inpcb *inp)
3623 {
3624 	int chk;
3625 
3626 	CURVNET_SET(ifp->if_vnet);
3627 	chk = pf_test6(PF_OUT, ifp, m, inp);
3628 	CURVNET_RESTORE();
3629 	if (chk && *m) {
3630 		m_freem(*m);
3631 		*m = NULL;
3632 	}
3633 	if (chk != PF_PASS)
3634 		return (EACCES);
3635 	return (0);
3636 }
3637 #endif /* INET6 */
3638 
3639 static int
3640 hook_pf(void)
3641 {
3642 #ifdef INET
3643 	struct pfil_head *pfh_inet;
3644 #endif
3645 #ifdef INET6
3646 	struct pfil_head *pfh_inet6;
3647 #endif
3648 
3649 	if (V_pf_pfil_hooked)
3650 		return (0);
3651 
3652 #ifdef INET
3653 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3654 	if (pfh_inet == NULL)
3655 		return (ESRCH); /* XXX */
3656 	pfil_add_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet);
3657 	pfil_add_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet);
3658 #endif
3659 #ifdef INET6
3660 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3661 	if (pfh_inet6 == NULL) {
3662 #ifdef INET
3663 		pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3664 		    pfh_inet);
3665 		pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3666 		    pfh_inet);
3667 #endif
3668 		return (ESRCH); /* XXX */
3669 	}
3670 	pfil_add_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet6);
3671 	pfil_add_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet6);
3672 #endif
3673 
3674 	V_pf_pfil_hooked = 1;
3675 	return (0);
3676 }
3677 
3678 static int
3679 dehook_pf(void)
3680 {
3681 #ifdef INET
3682 	struct pfil_head *pfh_inet;
3683 #endif
3684 #ifdef INET6
3685 	struct pfil_head *pfh_inet6;
3686 #endif
3687 
3688 	if (V_pf_pfil_hooked == 0)
3689 		return (0);
3690 
3691 #ifdef INET
3692 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3693 	if (pfh_inet == NULL)
3694 		return (ESRCH); /* XXX */
3695 	pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3696 	    pfh_inet);
3697 	pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3698 	    pfh_inet);
3699 #endif
3700 #ifdef INET6
3701 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3702 	if (pfh_inet6 == NULL)
3703 		return (ESRCH); /* XXX */
3704 	pfil_remove_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK,
3705 	    pfh_inet6);
3706 	pfil_remove_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK,
3707 	    pfh_inet6);
3708 #endif
3709 
3710 	V_pf_pfil_hooked = 0;
3711 	return (0);
3712 }
3713 
3714 static void
3715 pf_load_vnet(void)
3716 {
3717 	TAILQ_INIT(&V_pf_tags);
3718 	TAILQ_INIT(&V_pf_qids);
3719 
3720 	pfattach_vnet();
3721 	V_pf_vnet_active = 1;
3722 }
3723 
3724 static int
3725 pf_load(void)
3726 {
3727 	int error;
3728 
3729 	rw_init(&pf_rules_lock, "pf rulesets");
3730 	sx_init(&pf_ioctl_lock, "pf ioctl");
3731 	sx_init(&pf_end_lock, "pf end thread");
3732 
3733 	pf_mtag_initialize();
3734 
3735 	pf_dev = make_dev(&pf_cdevsw, 0, 0, 0, 0600, PF_NAME);
3736 	if (pf_dev == NULL)
3737 		return (ENOMEM);
3738 
3739 	pf_end_threads = 0;
3740 	error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
3741 	if (error != 0)
3742 		return (error);
3743 
3744 	pfi_initialize();
3745 
3746 	return (0);
3747 }
3748 
3749 static void
3750 pf_unload_vnet(void)
3751 {
3752 	int error;
3753 
3754 	V_pf_vnet_active = 0;
3755 	V_pf_status.running = 0;
3756 	swi_remove(V_pf_swi_cookie);
3757 	error = dehook_pf();
3758 	if (error) {
3759 		/*
3760 		 * Should not happen!
3761 		 * XXX Due to error code ESRCH, kldunload will show
3762 		 * a message like 'No such process'.
3763 		 */
3764 		printf("%s : pfil unregisteration fail\n", __FUNCTION__);
3765 		return;
3766 	}
3767 
3768 	PF_RULES_WLOCK();
3769 	shutdown_pf();
3770 	PF_RULES_WUNLOCK();
3771 
3772 	pf_unload_vnet_purge();
3773 
3774 	pf_normalize_cleanup();
3775 	PF_RULES_WLOCK();
3776 	pfi_cleanup_vnet();
3777 	PF_RULES_WUNLOCK();
3778 	pfr_cleanup();
3779 	pf_osfp_flush();
3780 	pf_cleanup();
3781 	if (IS_DEFAULT_VNET(curvnet))
3782 		pf_mtag_cleanup();
3783 }
3784 
3785 static void
3786 pf_unload(void)
3787 {
3788 
3789 	sx_xlock(&pf_end_lock);
3790 	pf_end_threads = 1;
3791 	while (pf_end_threads < 2) {
3792 		wakeup_one(pf_purge_thread);
3793 		sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
3794 	}
3795 	sx_xunlock(&pf_end_lock);
3796 
3797 	if (pf_dev != NULL)
3798 		destroy_dev(pf_dev);
3799 
3800 	pfi_cleanup();
3801 
3802 	rw_destroy(&pf_rules_lock);
3803 	sx_destroy(&pf_ioctl_lock);
3804 	sx_destroy(&pf_end_lock);
3805 }
3806 
3807 static void
3808 vnet_pf_init(void *unused __unused)
3809 {
3810 
3811 	pf_load_vnet();
3812 }
3813 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
3814     vnet_pf_init, NULL);
3815 
3816 static void
3817 vnet_pf_uninit(const void *unused __unused)
3818 {
3819 
3820 	pf_unload_vnet();
3821 }
3822 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
3823 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
3824     vnet_pf_uninit, NULL);
3825 
3826 
3827 static int
3828 pf_modevent(module_t mod, int type, void *data)
3829 {
3830 	int error = 0;
3831 
3832 	switch(type) {
3833 	case MOD_LOAD:
3834 		error = pf_load();
3835 		break;
3836 	case MOD_QUIESCE:
3837 		/*
3838 		 * Module should not be unloaded due to race conditions.
3839 		 */
3840 		error = EBUSY;
3841 		break;
3842 	case MOD_UNLOAD:
3843 		/* Handled in SYSUNINIT(pf_unload) to ensure it's done after
3844 		 * the vnet_pf_uninit()s */
3845 		break;
3846 	default:
3847 		error = EINVAL;
3848 		break;
3849 	}
3850 
3851 	return (error);
3852 }
3853 
3854 static moduledata_t pf_mod = {
3855 	"pf",
3856 	pf_modevent,
3857 	0
3858 };
3859 
3860 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
3861 MODULE_VERSION(pf, PF_MODVER);
3862