xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision d34048812292b714a0bf99967270d18fe3097c62)
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/smp.h>
63 #include <sys/socket.h>
64 #include <sys/sysctl.h>
65 #include <sys/md5.h>
66 #include <sys/ucred.h>
67 
68 #include <net/if.h>
69 #include <net/if_var.h>
70 #include <net/vnet.h>
71 #include <net/route.h>
72 #include <net/pfil.h>
73 #include <net/pfvar.h>
74 #include <net/if_pfsync.h>
75 #include <net/if_pflog.h>
76 
77 #include <netinet/in.h>
78 #include <netinet/ip.h>
79 #include <netinet/ip_var.h>
80 #include <netinet6/ip6_var.h>
81 #include <netinet/ip_icmp.h>
82 
83 #ifdef INET6
84 #include <netinet/ip6.h>
85 #endif /* INET6 */
86 
87 #ifdef ALTQ
88 #include <net/altq/altq.h>
89 #endif
90 
91 static struct pf_pool	*pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t,
92 			    u_int8_t, u_int8_t, u_int8_t);
93 
94 static void		 pf_mv_pool(struct pf_palist *, struct pf_palist *);
95 static void		 pf_empty_pool(struct pf_palist *);
96 static int		 pfioctl(struct cdev *, u_long, caddr_t, int,
97 			    struct thread *);
98 #ifdef ALTQ
99 static int		 pf_begin_altq(u_int32_t *);
100 static int		 pf_rollback_altq(u_int32_t);
101 static int		 pf_commit_altq(u_int32_t);
102 static int		 pf_enable_altq(struct pf_altq *);
103 static int		 pf_disable_altq(struct pf_altq *);
104 static u_int32_t	 pf_qname2qid(char *);
105 static void		 pf_qid_unref(u_int32_t);
106 #endif /* ALTQ */
107 static int		 pf_begin_rules(u_int32_t *, int, const char *);
108 static int		 pf_rollback_rules(u_int32_t, int, char *);
109 static int		 pf_setup_pfsync_matching(struct pf_ruleset *);
110 static void		 pf_hash_rule(MD5_CTX *, struct pf_rule *);
111 static void		 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
112 static int		 pf_commit_rules(u_int32_t, int, char *);
113 static int		 pf_addr_setup(struct pf_ruleset *,
114 			    struct pf_addr_wrap *, sa_family_t);
115 static void		 pf_addr_copyout(struct pf_addr_wrap *);
116 #ifdef ALTQ
117 static int		 pf_export_kaltq(struct pf_altq *,
118 			    struct pfioc_altq_v1 *, size_t);
119 static int		 pf_import_kaltq(struct pfioc_altq_v1 *,
120 			    struct pf_altq *, size_t);
121 #endif /* ALTQ */
122 
123 VNET_DEFINE(struct pf_rule,	pf_default_rule);
124 
125 #ifdef ALTQ
126 VNET_DEFINE_STATIC(int,		pf_altq_running);
127 #define	V_pf_altq_running	VNET(pf_altq_running)
128 #endif
129 
130 #define	TAGID_MAX	 50000
131 struct pf_tagname {
132 	TAILQ_ENTRY(pf_tagname)	entries;
133 	char			name[PF_TAG_NAME_SIZE];
134 	uint16_t		tag;
135 	int			ref;
136 };
137 
138 TAILQ_HEAD(pf_tags, pf_tagname);
139 #define	V_pf_tags		VNET(pf_tags)
140 VNET_DEFINE(struct pf_tags, pf_tags);
141 #define	V_pf_qids		VNET(pf_qids)
142 VNET_DEFINE(struct pf_tags, pf_qids);
143 static MALLOC_DEFINE(M_PFTAG, "pf_tag", "pf(4) tag names");
144 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
145 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
146 
147 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
148 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
149 #endif
150 
151 static u_int16_t	 tagname2tag(struct pf_tags *, char *);
152 static u_int16_t	 pf_tagname2tag(char *);
153 static void		 tag_unref(struct pf_tags *, u_int16_t);
154 
155 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
156 
157 struct cdev *pf_dev;
158 
159 /*
160  * XXX - These are new and need to be checked when moveing to a new version
161  */
162 static void		 pf_clear_states(void);
163 static int		 pf_clear_tables(void);
164 static void		 pf_clear_srcnodes(struct pf_src_node *);
165 static void		 pf_kill_srcnodes(struct pfioc_src_node_kill *);
166 static void		 pf_tbladdr_copyout(struct pf_addr_wrap *);
167 
168 /*
169  * Wrapper functions for pfil(9) hooks
170  */
171 #ifdef INET
172 static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp,
173     int dir, int flags, struct inpcb *inp);
174 static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp,
175     int dir, int flags, struct inpcb *inp);
176 #endif
177 #ifdef INET6
178 static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp,
179     int dir, int flags, struct inpcb *inp);
180 static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp,
181     int dir, int flags, struct inpcb *inp);
182 #endif
183 
184 static int		hook_pf(void);
185 static int		dehook_pf(void);
186 static int		shutdown_pf(void);
187 static int		pf_load(void);
188 static void		pf_unload(void);
189 
190 static struct cdevsw pf_cdevsw = {
191 	.d_ioctl =	pfioctl,
192 	.d_name =	PF_NAME,
193 	.d_version =	D_VERSION,
194 };
195 
196 volatile VNET_DEFINE_STATIC(int, pf_pfil_hooked);
197 #define V_pf_pfil_hooked	VNET(pf_pfil_hooked)
198 
199 /*
200  * We need a flag that is neither hooked nor running to know when
201  * the VNET is "valid".  We primarily need this to control (global)
202  * external event, e.g., eventhandlers.
203  */
204 VNET_DEFINE(int, pf_vnet_active);
205 #define V_pf_vnet_active	VNET(pf_vnet_active)
206 
207 int pf_end_threads;
208 struct proc *pf_purge_proc;
209 
210 struct rmlock			pf_rules_lock;
211 struct sx			pf_ioctl_lock;
212 struct sx			pf_end_lock;
213 
214 /* pfsync */
215 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
216 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
217 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
218 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
219 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
220 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
221 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
222 
223 /* pflog */
224 pflog_packet_t			*pflog_packet_ptr = NULL;
225 
226 extern u_long	pf_ioctl_maxcount;
227 
228 static void
229 pfattach_vnet(void)
230 {
231 	u_int32_t *my_timeout = V_pf_default_rule.timeout;
232 
233 	pf_initialize();
234 	pfr_initialize();
235 	pfi_initialize_vnet();
236 	pf_normalize_init();
237 
238 	V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
239 	V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
240 
241 	RB_INIT(&V_pf_anchors);
242 	pf_init_ruleset(&pf_main_ruleset);
243 
244 	/* default rule should never be garbage collected */
245 	V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
246 #ifdef PF_DEFAULT_TO_DROP
247 	V_pf_default_rule.action = PF_DROP;
248 #else
249 	V_pf_default_rule.action = PF_PASS;
250 #endif
251 	V_pf_default_rule.nr = -1;
252 	V_pf_default_rule.rtableid = -1;
253 
254 	V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
255 	V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
256 	V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
257 
258 	/* initialize default timeouts */
259 	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
260 	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
261 	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
262 	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
263 	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
264 	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
265 	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
266 	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
267 	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
268 	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
269 	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
270 	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
271 	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
272 	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
273 	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
274 	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
275 	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
276 	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
277 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
278 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
279 
280 	bzero(&V_pf_status, sizeof(V_pf_status));
281 	V_pf_status.debug = PF_DEBUG_URGENT;
282 
283 	V_pf_pfil_hooked = 0;
284 
285 	/* XXX do our best to avoid a conflict */
286 	V_pf_status.hostid = arc4random();
287 
288 	for (int i = 0; i < PFRES_MAX; i++)
289 		V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
290 	for (int i = 0; i < LCNT_MAX; i++)
291 		V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
292 	for (int i = 0; i < FCNT_MAX; i++)
293 		V_pf_status.fcounters[i] = counter_u64_alloc(M_WAITOK);
294 	for (int i = 0; i < SCNT_MAX; i++)
295 		V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
296 
297 	if (swi_add(NULL, "pf send", pf_intr, curvnet, SWI_NET,
298 	    INTR_MPSAFE, &V_pf_swi_cookie) != 0)
299 		/* XXXGL: leaked all above. */
300 		return;
301 }
302 
303 
304 static struct pf_pool *
305 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
306     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
307     u_int8_t check_ticket)
308 {
309 	struct pf_ruleset	*ruleset;
310 	struct pf_rule		*rule;
311 	int			 rs_num;
312 
313 	ruleset = pf_find_ruleset(anchor);
314 	if (ruleset == NULL)
315 		return (NULL);
316 	rs_num = pf_get_ruleset_number(rule_action);
317 	if (rs_num >= PF_RULESET_MAX)
318 		return (NULL);
319 	if (active) {
320 		if (check_ticket && ticket !=
321 		    ruleset->rules[rs_num].active.ticket)
322 			return (NULL);
323 		if (r_last)
324 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
325 			    pf_rulequeue);
326 		else
327 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
328 	} else {
329 		if (check_ticket && ticket !=
330 		    ruleset->rules[rs_num].inactive.ticket)
331 			return (NULL);
332 		if (r_last)
333 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
334 			    pf_rulequeue);
335 		else
336 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
337 	}
338 	if (!r_last) {
339 		while ((rule != NULL) && (rule->nr != rule_number))
340 			rule = TAILQ_NEXT(rule, entries);
341 	}
342 	if (rule == NULL)
343 		return (NULL);
344 
345 	return (&rule->rpool);
346 }
347 
348 static void
349 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
350 {
351 	struct pf_pooladdr	*mv_pool_pa;
352 
353 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
354 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
355 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
356 	}
357 }
358 
359 static void
360 pf_empty_pool(struct pf_palist *poola)
361 {
362 	struct pf_pooladdr *pa;
363 
364 	while ((pa = TAILQ_FIRST(poola)) != NULL) {
365 		switch (pa->addr.type) {
366 		case PF_ADDR_DYNIFTL:
367 			pfi_dynaddr_remove(pa->addr.p.dyn);
368 			break;
369 		case PF_ADDR_TABLE:
370 			/* XXX: this could be unfinished pooladdr on pabuf */
371 			if (pa->addr.p.tbl != NULL)
372 				pfr_detach_table(pa->addr.p.tbl);
373 			break;
374 		}
375 		if (pa->kif)
376 			pfi_kif_unref(pa->kif);
377 		TAILQ_REMOVE(poola, pa, entries);
378 		free(pa, M_PFRULE);
379 	}
380 }
381 
382 static void
383 pf_unlink_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
384 {
385 
386 	PF_RULES_WASSERT();
387 
388 	TAILQ_REMOVE(rulequeue, rule, entries);
389 
390 	PF_UNLNKDRULES_LOCK();
391 	rule->rule_flag |= PFRULE_REFS;
392 	TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
393 	PF_UNLNKDRULES_UNLOCK();
394 }
395 
396 void
397 pf_free_rule(struct pf_rule *rule)
398 {
399 
400 	PF_RULES_WASSERT();
401 
402 	if (rule->tag)
403 		tag_unref(&V_pf_tags, rule->tag);
404 	if (rule->match_tag)
405 		tag_unref(&V_pf_tags, rule->match_tag);
406 #ifdef ALTQ
407 	if (rule->pqid != rule->qid)
408 		pf_qid_unref(rule->pqid);
409 	pf_qid_unref(rule->qid);
410 #endif
411 	switch (rule->src.addr.type) {
412 	case PF_ADDR_DYNIFTL:
413 		pfi_dynaddr_remove(rule->src.addr.p.dyn);
414 		break;
415 	case PF_ADDR_TABLE:
416 		pfr_detach_table(rule->src.addr.p.tbl);
417 		break;
418 	}
419 	switch (rule->dst.addr.type) {
420 	case PF_ADDR_DYNIFTL:
421 		pfi_dynaddr_remove(rule->dst.addr.p.dyn);
422 		break;
423 	case PF_ADDR_TABLE:
424 		pfr_detach_table(rule->dst.addr.p.tbl);
425 		break;
426 	}
427 	if (rule->overload_tbl)
428 		pfr_detach_table(rule->overload_tbl);
429 	if (rule->kif)
430 		pfi_kif_unref(rule->kif);
431 	pf_anchor_remove(rule);
432 	pf_empty_pool(&rule->rpool.list);
433 	counter_u64_free(rule->states_cur);
434 	counter_u64_free(rule->states_tot);
435 	counter_u64_free(rule->src_nodes);
436 	free(rule, M_PFRULE);
437 }
438 
439 static u_int16_t
440 tagname2tag(struct pf_tags *head, char *tagname)
441 {
442 	struct pf_tagname	*tag, *p = NULL;
443 	u_int16_t		 new_tagid = 1;
444 
445 	PF_RULES_WASSERT();
446 
447 	TAILQ_FOREACH(tag, head, entries)
448 		if (strcmp(tagname, tag->name) == 0) {
449 			tag->ref++;
450 			return (tag->tag);
451 		}
452 
453 	/*
454 	 * to avoid fragmentation, we do a linear search from the beginning
455 	 * and take the first free slot we find. if there is none or the list
456 	 * is empty, append a new entry at the end.
457 	 */
458 
459 	/* new entry */
460 	if (!TAILQ_EMPTY(head))
461 		for (p = TAILQ_FIRST(head); p != NULL &&
462 		    p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
463 			new_tagid = p->tag + 1;
464 
465 	if (new_tagid > TAGID_MAX)
466 		return (0);
467 
468 	/* allocate and fill new struct pf_tagname */
469 	tag = malloc(sizeof(*tag), M_PFTAG, M_NOWAIT|M_ZERO);
470 	if (tag == NULL)
471 		return (0);
472 	strlcpy(tag->name, tagname, sizeof(tag->name));
473 	tag->tag = new_tagid;
474 	tag->ref++;
475 
476 	if (p != NULL)	/* insert new entry before p */
477 		TAILQ_INSERT_BEFORE(p, tag, entries);
478 	else	/* either list empty or no free slot in between */
479 		TAILQ_INSERT_TAIL(head, tag, entries);
480 
481 	return (tag->tag);
482 }
483 
484 static void
485 tag_unref(struct pf_tags *head, u_int16_t tag)
486 {
487 	struct pf_tagname	*p, *next;
488 
489 	PF_RULES_WASSERT();
490 
491 	for (p = TAILQ_FIRST(head); p != NULL; p = next) {
492 		next = TAILQ_NEXT(p, entries);
493 		if (tag == p->tag) {
494 			if (--p->ref == 0) {
495 				TAILQ_REMOVE(head, p, entries);
496 				free(p, M_PFTAG);
497 			}
498 			break;
499 		}
500 	}
501 }
502 
503 static u_int16_t
504 pf_tagname2tag(char *tagname)
505 {
506 	return (tagname2tag(&V_pf_tags, tagname));
507 }
508 
509 #ifdef ALTQ
510 static u_int32_t
511 pf_qname2qid(char *qname)
512 {
513 	return ((u_int32_t)tagname2tag(&V_pf_qids, qname));
514 }
515 
516 static void
517 pf_qid_unref(u_int32_t qid)
518 {
519 	tag_unref(&V_pf_qids, (u_int16_t)qid);
520 }
521 
522 static int
523 pf_begin_altq(u_int32_t *ticket)
524 {
525 	struct pf_altq	*altq;
526 	int		 error = 0;
527 
528 	PF_RULES_WASSERT();
529 
530 	/* Purge the old altq list */
531 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
532 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
533 		if (altq->qname[0] == 0 &&
534 		    (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
535 			/* detach and destroy the discipline */
536 			error = altq_remove(altq);
537 		} else
538 			pf_qid_unref(altq->qid);
539 		free(altq, M_PFALTQ);
540 	}
541 	if (error)
542 		return (error);
543 	*ticket = ++V_ticket_altqs_inactive;
544 	V_altqs_inactive_open = 1;
545 	return (0);
546 }
547 
548 static int
549 pf_rollback_altq(u_int32_t ticket)
550 {
551 	struct pf_altq	*altq;
552 	int		 error = 0;
553 
554 	PF_RULES_WASSERT();
555 
556 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
557 		return (0);
558 	/* Purge the old altq list */
559 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
560 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
561 		if (altq->qname[0] == 0 &&
562 		   (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
563 			/* detach and destroy the discipline */
564 			error = altq_remove(altq);
565 		} else
566 			pf_qid_unref(altq->qid);
567 		free(altq, M_PFALTQ);
568 	}
569 	V_altqs_inactive_open = 0;
570 	return (error);
571 }
572 
573 static int
574 pf_commit_altq(u_int32_t ticket)
575 {
576 	struct pf_altqqueue	*old_altqs;
577 	struct pf_altq		*altq;
578 	int			 err, error = 0;
579 
580 	PF_RULES_WASSERT();
581 
582 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
583 		return (EBUSY);
584 
585 	/* swap altqs, keep the old. */
586 	old_altqs = V_pf_altqs_active;
587 	V_pf_altqs_active = V_pf_altqs_inactive;
588 	V_pf_altqs_inactive = old_altqs;
589 	V_ticket_altqs_active = V_ticket_altqs_inactive;
590 
591 	/* Attach new disciplines */
592 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
593 	if (altq->qname[0] == 0 &&
594 	   (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
595 			/* attach the discipline */
596 			error = altq_pfattach(altq);
597 			if (error == 0 && V_pf_altq_running)
598 				error = pf_enable_altq(altq);
599 			if (error != 0)
600 				return (error);
601 		}
602 	}
603 
604 	/* Purge the old altq list */
605 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
606 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
607 		if (altq->qname[0] == 0 &&
608 		    (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
609 			/* detach and destroy the discipline */
610 			if (V_pf_altq_running)
611 				error = pf_disable_altq(altq);
612 			err = altq_pfdetach(altq);
613 			if (err != 0 && error == 0)
614 				error = err;
615 			err = altq_remove(altq);
616 			if (err != 0 && error == 0)
617 				error = err;
618 		} else
619 			pf_qid_unref(altq->qid);
620 		free(altq, M_PFALTQ);
621 	}
622 
623 	V_altqs_inactive_open = 0;
624 	return (error);
625 }
626 
627 static int
628 pf_enable_altq(struct pf_altq *altq)
629 {
630 	struct ifnet		*ifp;
631 	struct tb_profile	 tb;
632 	int			 error = 0;
633 
634 	if ((ifp = ifunit(altq->ifname)) == NULL)
635 		return (EINVAL);
636 
637 	if (ifp->if_snd.altq_type != ALTQT_NONE)
638 		error = altq_enable(&ifp->if_snd);
639 
640 	/* set tokenbucket regulator */
641 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
642 		tb.rate = altq->ifbandwidth;
643 		tb.depth = altq->tbrsize;
644 		error = tbr_set(&ifp->if_snd, &tb);
645 	}
646 
647 	return (error);
648 }
649 
650 static int
651 pf_disable_altq(struct pf_altq *altq)
652 {
653 	struct ifnet		*ifp;
654 	struct tb_profile	 tb;
655 	int			 error;
656 
657 	if ((ifp = ifunit(altq->ifname)) == NULL)
658 		return (EINVAL);
659 
660 	/*
661 	 * when the discipline is no longer referenced, it was overridden
662 	 * by a new one.  if so, just return.
663 	 */
664 	if (altq->altq_disc != ifp->if_snd.altq_disc)
665 		return (0);
666 
667 	error = altq_disable(&ifp->if_snd);
668 
669 	if (error == 0) {
670 		/* clear tokenbucket regulator */
671 		tb.rate = 0;
672 		error = tbr_set(&ifp->if_snd, &tb);
673 	}
674 
675 	return (error);
676 }
677 
678 void
679 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
680 {
681 	struct ifnet	*ifp1;
682 	struct pf_altq	*a1, *a2, *a3;
683 	u_int32_t	 ticket;
684 	int		 error = 0;
685 
686 	/* Interrupt userland queue modifications */
687 	if (V_altqs_inactive_open)
688 		pf_rollback_altq(V_ticket_altqs_inactive);
689 
690 	/* Start new altq ruleset */
691 	if (pf_begin_altq(&ticket))
692 		return;
693 
694 	/* Copy the current active set */
695 	TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
696 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
697 		if (a2 == NULL) {
698 			error = ENOMEM;
699 			break;
700 		}
701 		bcopy(a1, a2, sizeof(struct pf_altq));
702 
703 		if (a2->qname[0] != 0) {
704 			if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
705 				error = EBUSY;
706 				free(a2, M_PFALTQ);
707 				break;
708 			}
709 			a2->altq_disc = NULL;
710 			TAILQ_FOREACH(a3, V_pf_altqs_inactive, entries) {
711 				if (strncmp(a3->ifname, a2->ifname,
712 				    IFNAMSIZ) == 0 && a3->qname[0] == 0) {
713 					a2->altq_disc = a3->altq_disc;
714 					break;
715 				}
716 			}
717 		}
718 		/* Deactivate the interface in question */
719 		a2->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
720 		if ((ifp1 = ifunit(a2->ifname)) == NULL ||
721 		    (remove && ifp1 == ifp)) {
722 			a2->local_flags |= PFALTQ_FLAG_IF_REMOVED;
723 		} else {
724 			error = altq_add(a2);
725 
726 			if (ticket != V_ticket_altqs_inactive)
727 				error = EBUSY;
728 
729 			if (error) {
730 				free(a2, M_PFALTQ);
731 				break;
732 			}
733 		}
734 
735 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
736 	}
737 
738 	if (error != 0)
739 		pf_rollback_altq(ticket);
740 	else
741 		pf_commit_altq(ticket);
742 }
743 #endif /* ALTQ */
744 
745 static int
746 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
747 {
748 	struct pf_ruleset	*rs;
749 	struct pf_rule		*rule;
750 
751 	PF_RULES_WASSERT();
752 
753 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
754 		return (EINVAL);
755 	rs = pf_find_or_create_ruleset(anchor);
756 	if (rs == NULL)
757 		return (EINVAL);
758 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
759 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
760 		rs->rules[rs_num].inactive.rcount--;
761 	}
762 	*ticket = ++rs->rules[rs_num].inactive.ticket;
763 	rs->rules[rs_num].inactive.open = 1;
764 	return (0);
765 }
766 
767 static int
768 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
769 {
770 	struct pf_ruleset	*rs;
771 	struct pf_rule		*rule;
772 
773 	PF_RULES_WASSERT();
774 
775 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
776 		return (EINVAL);
777 	rs = pf_find_ruleset(anchor);
778 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
779 	    rs->rules[rs_num].inactive.ticket != ticket)
780 		return (0);
781 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
782 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
783 		rs->rules[rs_num].inactive.rcount--;
784 	}
785 	rs->rules[rs_num].inactive.open = 0;
786 	return (0);
787 }
788 
789 #define PF_MD5_UPD(st, elm)						\
790 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
791 
792 #define PF_MD5_UPD_STR(st, elm)						\
793 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
794 
795 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
796 		(stor) = htonl((st)->elm);				\
797 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
798 } while (0)
799 
800 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
801 		(stor) = htons((st)->elm);				\
802 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
803 } while (0)
804 
805 static void
806 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
807 {
808 	PF_MD5_UPD(pfr, addr.type);
809 	switch (pfr->addr.type) {
810 		case PF_ADDR_DYNIFTL:
811 			PF_MD5_UPD(pfr, addr.v.ifname);
812 			PF_MD5_UPD(pfr, addr.iflags);
813 			break;
814 		case PF_ADDR_TABLE:
815 			PF_MD5_UPD(pfr, addr.v.tblname);
816 			break;
817 		case PF_ADDR_ADDRMASK:
818 			/* XXX ignore af? */
819 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
820 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
821 			break;
822 	}
823 
824 	PF_MD5_UPD(pfr, port[0]);
825 	PF_MD5_UPD(pfr, port[1]);
826 	PF_MD5_UPD(pfr, neg);
827 	PF_MD5_UPD(pfr, port_op);
828 }
829 
830 static void
831 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
832 {
833 	u_int16_t x;
834 	u_int32_t y;
835 
836 	pf_hash_rule_addr(ctx, &rule->src);
837 	pf_hash_rule_addr(ctx, &rule->dst);
838 	PF_MD5_UPD_STR(rule, label);
839 	PF_MD5_UPD_STR(rule, ifname);
840 	PF_MD5_UPD_STR(rule, match_tagname);
841 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
842 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
843 	PF_MD5_UPD_HTONL(rule, prob, y);
844 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
845 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
846 	PF_MD5_UPD(rule, uid.op);
847 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
848 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
849 	PF_MD5_UPD(rule, gid.op);
850 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
851 	PF_MD5_UPD(rule, action);
852 	PF_MD5_UPD(rule, direction);
853 	PF_MD5_UPD(rule, af);
854 	PF_MD5_UPD(rule, quick);
855 	PF_MD5_UPD(rule, ifnot);
856 	PF_MD5_UPD(rule, match_tag_not);
857 	PF_MD5_UPD(rule, natpass);
858 	PF_MD5_UPD(rule, keep_state);
859 	PF_MD5_UPD(rule, proto);
860 	PF_MD5_UPD(rule, type);
861 	PF_MD5_UPD(rule, code);
862 	PF_MD5_UPD(rule, flags);
863 	PF_MD5_UPD(rule, flagset);
864 	PF_MD5_UPD(rule, allow_opts);
865 	PF_MD5_UPD(rule, rt);
866 	PF_MD5_UPD(rule, tos);
867 }
868 
869 static int
870 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
871 {
872 	struct pf_ruleset	*rs;
873 	struct pf_rule		*rule, **old_array;
874 	struct pf_rulequeue	*old_rules;
875 	int			 error;
876 	u_int32_t		 old_rcount;
877 
878 	PF_RULES_WASSERT();
879 
880 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
881 		return (EINVAL);
882 	rs = pf_find_ruleset(anchor);
883 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
884 	    ticket != rs->rules[rs_num].inactive.ticket)
885 		return (EBUSY);
886 
887 	/* Calculate checksum for the main ruleset */
888 	if (rs == &pf_main_ruleset) {
889 		error = pf_setup_pfsync_matching(rs);
890 		if (error != 0)
891 			return (error);
892 	}
893 
894 	/* Swap rules, keep the old. */
895 	old_rules = rs->rules[rs_num].active.ptr;
896 	old_rcount = rs->rules[rs_num].active.rcount;
897 	old_array = rs->rules[rs_num].active.ptr_array;
898 
899 	rs->rules[rs_num].active.ptr =
900 	    rs->rules[rs_num].inactive.ptr;
901 	rs->rules[rs_num].active.ptr_array =
902 	    rs->rules[rs_num].inactive.ptr_array;
903 	rs->rules[rs_num].active.rcount =
904 	    rs->rules[rs_num].inactive.rcount;
905 	rs->rules[rs_num].inactive.ptr = old_rules;
906 	rs->rules[rs_num].inactive.ptr_array = old_array;
907 	rs->rules[rs_num].inactive.rcount = old_rcount;
908 
909 	rs->rules[rs_num].active.ticket =
910 	    rs->rules[rs_num].inactive.ticket;
911 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
912 
913 
914 	/* Purge the old rule list. */
915 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
916 		pf_unlink_rule(old_rules, rule);
917 	if (rs->rules[rs_num].inactive.ptr_array)
918 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
919 	rs->rules[rs_num].inactive.ptr_array = NULL;
920 	rs->rules[rs_num].inactive.rcount = 0;
921 	rs->rules[rs_num].inactive.open = 0;
922 	pf_remove_if_empty_ruleset(rs);
923 
924 	return (0);
925 }
926 
927 static int
928 pf_setup_pfsync_matching(struct pf_ruleset *rs)
929 {
930 	MD5_CTX			 ctx;
931 	struct pf_rule		*rule;
932 	int			 rs_cnt;
933 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
934 
935 	MD5Init(&ctx);
936 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
937 		/* XXX PF_RULESET_SCRUB as well? */
938 		if (rs_cnt == PF_RULESET_SCRUB)
939 			continue;
940 
941 		if (rs->rules[rs_cnt].inactive.ptr_array)
942 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
943 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
944 
945 		if (rs->rules[rs_cnt].inactive.rcount) {
946 			rs->rules[rs_cnt].inactive.ptr_array =
947 			    malloc(sizeof(caddr_t) *
948 			    rs->rules[rs_cnt].inactive.rcount,
949 			    M_TEMP, M_NOWAIT);
950 
951 			if (!rs->rules[rs_cnt].inactive.ptr_array)
952 				return (ENOMEM);
953 		}
954 
955 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
956 		    entries) {
957 			pf_hash_rule(&ctx, rule);
958 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
959 		}
960 	}
961 
962 	MD5Final(digest, &ctx);
963 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
964 	return (0);
965 }
966 
967 static int
968 pf_addr_setup(struct pf_ruleset *ruleset, struct pf_addr_wrap *addr,
969     sa_family_t af)
970 {
971 	int error = 0;
972 
973 	switch (addr->type) {
974 	case PF_ADDR_TABLE:
975 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
976 		if (addr->p.tbl == NULL)
977 			error = ENOMEM;
978 		break;
979 	case PF_ADDR_DYNIFTL:
980 		error = pfi_dynaddr_setup(addr, af);
981 		break;
982 	}
983 
984 	return (error);
985 }
986 
987 static void
988 pf_addr_copyout(struct pf_addr_wrap *addr)
989 {
990 
991 	switch (addr->type) {
992 	case PF_ADDR_DYNIFTL:
993 		pfi_dynaddr_copyout(addr);
994 		break;
995 	case PF_ADDR_TABLE:
996 		pf_tbladdr_copyout(addr);
997 		break;
998 	}
999 }
1000 
1001 #ifdef ALTQ
1002 /*
1003  * Handle export of struct pf_kaltq to user binaries that may be using any
1004  * version of struct pf_altq.
1005  */
1006 static int
1007 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1008 {
1009 	u_int32_t version;
1010 
1011 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1012 		version = 0;
1013 	else
1014 		version = pa->version;
1015 
1016 	if (version > PFIOC_ALTQ_VERSION)
1017 		return (EINVAL);
1018 
1019 #define ASSIGN(x) exported_q->x = q->x
1020 #define COPY(x) \
1021 	bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1022 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1023 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1024 
1025 	switch (version) {
1026 	case 0: {
1027 		struct pf_altq_v0 *exported_q =
1028 		    &((struct pfioc_altq_v0 *)pa)->altq;
1029 
1030 		COPY(ifname);
1031 
1032 		ASSIGN(scheduler);
1033 		ASSIGN(tbrsize);
1034 		exported_q->tbrsize = SATU16(q->tbrsize);
1035 		exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1036 
1037 		COPY(qname);
1038 		COPY(parent);
1039 		ASSIGN(parent_qid);
1040 		exported_q->bandwidth = SATU32(q->bandwidth);
1041 		ASSIGN(priority);
1042 		ASSIGN(local_flags);
1043 
1044 		ASSIGN(qlimit);
1045 		ASSIGN(flags);
1046 
1047 		if (q->scheduler == ALTQT_HFSC) {
1048 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1049 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1050 			    SATU32(q->pq_u.hfsc_opts.x)
1051 
1052 			ASSIGN_OPT_SATU32(rtsc_m1);
1053 			ASSIGN_OPT(rtsc_d);
1054 			ASSIGN_OPT_SATU32(rtsc_m2);
1055 
1056 			ASSIGN_OPT_SATU32(lssc_m1);
1057 			ASSIGN_OPT(lssc_d);
1058 			ASSIGN_OPT_SATU32(lssc_m2);
1059 
1060 			ASSIGN_OPT_SATU32(ulsc_m1);
1061 			ASSIGN_OPT(ulsc_d);
1062 			ASSIGN_OPT_SATU32(ulsc_m2);
1063 
1064 			ASSIGN_OPT(flags);
1065 
1066 #undef ASSIGN_OPT
1067 #undef ASSIGN_OPT_SATU32
1068 		} else
1069 			COPY(pq_u);
1070 
1071 		ASSIGN(qid);
1072 		break;
1073 	}
1074 	case 1:	{
1075 		struct pf_altq_v1 *exported_q =
1076 		    &((struct pfioc_altq_v1 *)pa)->altq;
1077 
1078 		COPY(ifname);
1079 
1080 		ASSIGN(scheduler);
1081 		ASSIGN(tbrsize);
1082 		ASSIGN(ifbandwidth);
1083 
1084 		COPY(qname);
1085 		COPY(parent);
1086 		ASSIGN(parent_qid);
1087 		ASSIGN(bandwidth);
1088 		ASSIGN(priority);
1089 		ASSIGN(local_flags);
1090 
1091 		ASSIGN(qlimit);
1092 		ASSIGN(flags);
1093 		COPY(pq_u);
1094 
1095 		ASSIGN(qid);
1096 		break;
1097 	}
1098 	default:
1099 		panic("%s: unhandled struct pfioc_altq version", __func__);
1100 		break;
1101 	}
1102 
1103 #undef ASSIGN
1104 #undef COPY
1105 #undef SATU16
1106 #undef SATU32
1107 
1108 	return (0);
1109 }
1110 
1111 /*
1112  * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1113  * that may be using any version of it.
1114  */
1115 static int
1116 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1117 {
1118 	u_int32_t version;
1119 
1120 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1121 		version = 0;
1122 	else
1123 		version = pa->version;
1124 
1125 	if (version > PFIOC_ALTQ_VERSION)
1126 		return (EINVAL);
1127 
1128 #define ASSIGN(x) q->x = imported_q->x
1129 #define COPY(x) \
1130 	bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1131 
1132 	switch (version) {
1133 	case 0: {
1134 		struct pf_altq_v0 *imported_q =
1135 		    &((struct pfioc_altq_v0 *)pa)->altq;
1136 
1137 		COPY(ifname);
1138 
1139 		ASSIGN(scheduler);
1140 		ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1141 		ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1142 
1143 		COPY(qname);
1144 		COPY(parent);
1145 		ASSIGN(parent_qid);
1146 		ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1147 		ASSIGN(priority);
1148 		ASSIGN(local_flags);
1149 
1150 		ASSIGN(qlimit);
1151 		ASSIGN(flags);
1152 
1153 		if (imported_q->scheduler == ALTQT_HFSC) {
1154 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1155 
1156 			/*
1157 			 * The m1 and m2 parameters are being copied from
1158 			 * 32-bit to 64-bit.
1159 			 */
1160 			ASSIGN_OPT(rtsc_m1);
1161 			ASSIGN_OPT(rtsc_d);
1162 			ASSIGN_OPT(rtsc_m2);
1163 
1164 			ASSIGN_OPT(lssc_m1);
1165 			ASSIGN_OPT(lssc_d);
1166 			ASSIGN_OPT(lssc_m2);
1167 
1168 			ASSIGN_OPT(ulsc_m1);
1169 			ASSIGN_OPT(ulsc_d);
1170 			ASSIGN_OPT(ulsc_m2);
1171 
1172 			ASSIGN_OPT(flags);
1173 
1174 #undef ASSIGN_OPT
1175 		} else
1176 			COPY(pq_u);
1177 
1178 		ASSIGN(qid);
1179 		break;
1180 	}
1181 	case 1: {
1182 		struct pf_altq_v1 *imported_q =
1183 		    &((struct pfioc_altq_v1 *)pa)->altq;
1184 
1185 		COPY(ifname);
1186 
1187 		ASSIGN(scheduler);
1188 		ASSIGN(tbrsize);
1189 		ASSIGN(ifbandwidth);
1190 
1191 		COPY(qname);
1192 		COPY(parent);
1193 		ASSIGN(parent_qid);
1194 		ASSIGN(bandwidth);
1195 		ASSIGN(priority);
1196 		ASSIGN(local_flags);
1197 
1198 		ASSIGN(qlimit);
1199 		ASSIGN(flags);
1200 		COPY(pq_u);
1201 
1202 		ASSIGN(qid);
1203 		break;
1204 	}
1205 	default:
1206 		panic("%s: unhandled struct pfioc_altq version", __func__);
1207 		break;
1208 	}
1209 
1210 #undef ASSIGN
1211 #undef COPY
1212 
1213 	return (0);
1214 }
1215 #endif /* ALTQ */
1216 
1217 static int
1218 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
1219 {
1220 	int			 error = 0;
1221 	PF_RULES_RLOCK_TRACKER;
1222 
1223 	/* XXX keep in sync with switch() below */
1224 	if (securelevel_gt(td->td_ucred, 2))
1225 		switch (cmd) {
1226 		case DIOCGETRULES:
1227 		case DIOCGETRULE:
1228 		case DIOCGETADDRS:
1229 		case DIOCGETADDR:
1230 		case DIOCGETSTATE:
1231 		case DIOCSETSTATUSIF:
1232 		case DIOCGETSTATUS:
1233 		case DIOCCLRSTATUS:
1234 		case DIOCNATLOOK:
1235 		case DIOCSETDEBUG:
1236 		case DIOCGETSTATES:
1237 		case DIOCGETTIMEOUT:
1238 		case DIOCCLRRULECTRS:
1239 		case DIOCGETLIMIT:
1240 		case DIOCGETALTQSV0:
1241 		case DIOCGETALTQSV1:
1242 		case DIOCGETALTQV0:
1243 		case DIOCGETALTQV1:
1244 		case DIOCGETQSTATSV0:
1245 		case DIOCGETQSTATSV1:
1246 		case DIOCGETRULESETS:
1247 		case DIOCGETRULESET:
1248 		case DIOCRGETTABLES:
1249 		case DIOCRGETTSTATS:
1250 		case DIOCRCLRTSTATS:
1251 		case DIOCRCLRADDRS:
1252 		case DIOCRADDADDRS:
1253 		case DIOCRDELADDRS:
1254 		case DIOCRSETADDRS:
1255 		case DIOCRGETADDRS:
1256 		case DIOCRGETASTATS:
1257 		case DIOCRCLRASTATS:
1258 		case DIOCRTSTADDRS:
1259 		case DIOCOSFPGET:
1260 		case DIOCGETSRCNODES:
1261 		case DIOCCLRSRCNODES:
1262 		case DIOCIGETIFACES:
1263 		case DIOCGIFSPEEDV0:
1264 		case DIOCGIFSPEEDV1:
1265 		case DIOCSETIFFLAG:
1266 		case DIOCCLRIFFLAG:
1267 			break;
1268 		case DIOCRCLRTABLES:
1269 		case DIOCRADDTABLES:
1270 		case DIOCRDELTABLES:
1271 		case DIOCRSETTFLAGS:
1272 			if (((struct pfioc_table *)addr)->pfrio_flags &
1273 			    PFR_FLAG_DUMMY)
1274 				break; /* dummy operation ok */
1275 			return (EPERM);
1276 		default:
1277 			return (EPERM);
1278 		}
1279 
1280 	if (!(flags & FWRITE))
1281 		switch (cmd) {
1282 		case DIOCGETRULES:
1283 		case DIOCGETADDRS:
1284 		case DIOCGETADDR:
1285 		case DIOCGETSTATE:
1286 		case DIOCGETSTATUS:
1287 		case DIOCGETSTATES:
1288 		case DIOCGETTIMEOUT:
1289 		case DIOCGETLIMIT:
1290 		case DIOCGETALTQSV0:
1291 		case DIOCGETALTQSV1:
1292 		case DIOCGETALTQV0:
1293 		case DIOCGETALTQV1:
1294 		case DIOCGETQSTATSV0:
1295 		case DIOCGETQSTATSV1:
1296 		case DIOCGETRULESETS:
1297 		case DIOCGETRULESET:
1298 		case DIOCNATLOOK:
1299 		case DIOCRGETTABLES:
1300 		case DIOCRGETTSTATS:
1301 		case DIOCRGETADDRS:
1302 		case DIOCRGETASTATS:
1303 		case DIOCRTSTADDRS:
1304 		case DIOCOSFPGET:
1305 		case DIOCGETSRCNODES:
1306 		case DIOCIGETIFACES:
1307 		case DIOCGIFSPEEDV1:
1308 		case DIOCGIFSPEEDV0:
1309 			break;
1310 		case DIOCRCLRTABLES:
1311 		case DIOCRADDTABLES:
1312 		case DIOCRDELTABLES:
1313 		case DIOCRCLRTSTATS:
1314 		case DIOCRCLRADDRS:
1315 		case DIOCRADDADDRS:
1316 		case DIOCRDELADDRS:
1317 		case DIOCRSETADDRS:
1318 		case DIOCRSETTFLAGS:
1319 			if (((struct pfioc_table *)addr)->pfrio_flags &
1320 			    PFR_FLAG_DUMMY) {
1321 				flags |= FWRITE; /* need write lock for dummy */
1322 				break; /* dummy operation ok */
1323 			}
1324 			return (EACCES);
1325 		case DIOCGETRULE:
1326 			if (((struct pfioc_rule *)addr)->action ==
1327 			    PF_GET_CLR_CNTR)
1328 				return (EACCES);
1329 			break;
1330 		default:
1331 			return (EACCES);
1332 		}
1333 
1334 	CURVNET_SET(TD_TO_VNET(td));
1335 
1336 	switch (cmd) {
1337 	case DIOCSTART:
1338 		sx_xlock(&pf_ioctl_lock);
1339 		if (V_pf_status.running)
1340 			error = EEXIST;
1341 		else {
1342 			int cpu;
1343 
1344 			error = hook_pf();
1345 			if (error) {
1346 				DPFPRINTF(PF_DEBUG_MISC,
1347 				    ("pf: pfil registration failed\n"));
1348 				break;
1349 			}
1350 			V_pf_status.running = 1;
1351 			V_pf_status.since = time_second;
1352 
1353 			CPU_FOREACH(cpu)
1354 				V_pf_stateid[cpu] = time_second;
1355 
1356 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1357 		}
1358 		break;
1359 
1360 	case DIOCSTOP:
1361 		sx_xlock(&pf_ioctl_lock);
1362 		if (!V_pf_status.running)
1363 			error = ENOENT;
1364 		else {
1365 			V_pf_status.running = 0;
1366 			error = dehook_pf();
1367 			if (error) {
1368 				V_pf_status.running = 1;
1369 				DPFPRINTF(PF_DEBUG_MISC,
1370 				    ("pf: pfil unregistration failed\n"));
1371 			}
1372 			V_pf_status.since = time_second;
1373 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1374 		}
1375 		break;
1376 
1377 	case DIOCADDRULE: {
1378 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1379 		struct pf_ruleset	*ruleset;
1380 		struct pf_rule		*rule, *tail;
1381 		struct pf_pooladdr	*pa;
1382 		struct pfi_kif		*kif = NULL;
1383 		int			 rs_num;
1384 
1385 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1386 			error = EINVAL;
1387 			break;
1388 		}
1389 #ifndef INET
1390 		if (pr->rule.af == AF_INET) {
1391 			error = EAFNOSUPPORT;
1392 			break;
1393 		}
1394 #endif /* INET */
1395 #ifndef INET6
1396 		if (pr->rule.af == AF_INET6) {
1397 			error = EAFNOSUPPORT;
1398 			break;
1399 		}
1400 #endif /* INET6 */
1401 
1402 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
1403 		bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1404 		if (rule->ifname[0])
1405 			kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1406 		rule->states_cur = counter_u64_alloc(M_WAITOK);
1407 		rule->states_tot = counter_u64_alloc(M_WAITOK);
1408 		rule->src_nodes = counter_u64_alloc(M_WAITOK);
1409 		rule->cuid = td->td_ucred->cr_ruid;
1410 		rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1411 		TAILQ_INIT(&rule->rpool.list);
1412 
1413 #define	ERROUT(x)	{ error = (x); goto DIOCADDRULE_error; }
1414 
1415 		PF_RULES_WLOCK();
1416 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1417 		ruleset = pf_find_ruleset(pr->anchor);
1418 		if (ruleset == NULL)
1419 			ERROUT(EINVAL);
1420 		rs_num = pf_get_ruleset_number(pr->rule.action);
1421 		if (rs_num >= PF_RULESET_MAX)
1422 			ERROUT(EINVAL);
1423 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1424 			DPFPRINTF(PF_DEBUG_MISC,
1425 			    ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
1426 			    ruleset->rules[rs_num].inactive.ticket));
1427 			ERROUT(EBUSY);
1428 		}
1429 		if (pr->pool_ticket != V_ticket_pabuf) {
1430 			DPFPRINTF(PF_DEBUG_MISC,
1431 			    ("pool_ticket: %d != %d\n", pr->pool_ticket,
1432 			    V_ticket_pabuf));
1433 			ERROUT(EBUSY);
1434 		}
1435 
1436 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1437 		    pf_rulequeue);
1438 		if (tail)
1439 			rule->nr = tail->nr + 1;
1440 		else
1441 			rule->nr = 0;
1442 		if (rule->ifname[0]) {
1443 			rule->kif = pfi_kif_attach(kif, rule->ifname);
1444 			pfi_kif_ref(rule->kif);
1445 		} else
1446 			rule->kif = NULL;
1447 
1448 		if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
1449 			error = EBUSY;
1450 
1451 #ifdef ALTQ
1452 		/* set queue IDs */
1453 		if (rule->qname[0] != 0) {
1454 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1455 				error = EBUSY;
1456 			else if (rule->pqname[0] != 0) {
1457 				if ((rule->pqid =
1458 				    pf_qname2qid(rule->pqname)) == 0)
1459 					error = EBUSY;
1460 			} else
1461 				rule->pqid = rule->qid;
1462 		}
1463 #endif
1464 		if (rule->tagname[0])
1465 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1466 				error = EBUSY;
1467 		if (rule->match_tagname[0])
1468 			if ((rule->match_tag =
1469 			    pf_tagname2tag(rule->match_tagname)) == 0)
1470 				error = EBUSY;
1471 		if (rule->rt && !rule->direction)
1472 			error = EINVAL;
1473 		if (!rule->log)
1474 			rule->logif = 0;
1475 		if (rule->logif >= PFLOGIFS_MAX)
1476 			error = EINVAL;
1477 		if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1478 			error = ENOMEM;
1479 		if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1480 			error = ENOMEM;
1481 		if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1482 			error = EINVAL;
1483 		if (rule->scrub_flags & PFSTATE_SETPRIO &&
1484 		    (rule->set_prio[0] > PF_PRIO_MAX ||
1485 		    rule->set_prio[1] > PF_PRIO_MAX))
1486 			error = EINVAL;
1487 		TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1488 			if (pa->addr.type == PF_ADDR_TABLE) {
1489 				pa->addr.p.tbl = pfr_attach_table(ruleset,
1490 				    pa->addr.v.tblname);
1491 				if (pa->addr.p.tbl == NULL)
1492 					error = ENOMEM;
1493 			}
1494 
1495 		rule->overload_tbl = NULL;
1496 		if (rule->overload_tblname[0]) {
1497 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1498 			    rule->overload_tblname)) == NULL)
1499 				error = EINVAL;
1500 			else
1501 				rule->overload_tbl->pfrkt_flags |=
1502 				    PFR_TFLAG_ACTIVE;
1503 		}
1504 
1505 		pf_mv_pool(&V_pf_pabuf, &rule->rpool.list);
1506 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1507 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1508 		    (rule->rt > PF_NOPFROUTE)) &&
1509 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1510 			error = EINVAL;
1511 
1512 		if (error) {
1513 			pf_free_rule(rule);
1514 			PF_RULES_WUNLOCK();
1515 			break;
1516 		}
1517 
1518 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1519 		rule->evaluations = rule->packets[0] = rule->packets[1] =
1520 		    rule->bytes[0] = rule->bytes[1] = 0;
1521 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1522 		    rule, entries);
1523 		ruleset->rules[rs_num].inactive.rcount++;
1524 		PF_RULES_WUNLOCK();
1525 		break;
1526 
1527 #undef ERROUT
1528 DIOCADDRULE_error:
1529 		PF_RULES_WUNLOCK();
1530 		counter_u64_free(rule->states_cur);
1531 		counter_u64_free(rule->states_tot);
1532 		counter_u64_free(rule->src_nodes);
1533 		free(rule, M_PFRULE);
1534 		if (kif)
1535 			free(kif, PFI_MTYPE);
1536 		break;
1537 	}
1538 
1539 	case DIOCGETRULES: {
1540 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1541 		struct pf_ruleset	*ruleset;
1542 		struct pf_rule		*tail;
1543 		int			 rs_num;
1544 
1545 		PF_RULES_WLOCK();
1546 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1547 		ruleset = pf_find_ruleset(pr->anchor);
1548 		if (ruleset == NULL) {
1549 			PF_RULES_WUNLOCK();
1550 			error = EINVAL;
1551 			break;
1552 		}
1553 		rs_num = pf_get_ruleset_number(pr->rule.action);
1554 		if (rs_num >= PF_RULESET_MAX) {
1555 			PF_RULES_WUNLOCK();
1556 			error = EINVAL;
1557 			break;
1558 		}
1559 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1560 		    pf_rulequeue);
1561 		if (tail)
1562 			pr->nr = tail->nr + 1;
1563 		else
1564 			pr->nr = 0;
1565 		pr->ticket = ruleset->rules[rs_num].active.ticket;
1566 		PF_RULES_WUNLOCK();
1567 		break;
1568 	}
1569 
1570 	case DIOCGETRULE: {
1571 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1572 		struct pf_ruleset	*ruleset;
1573 		struct pf_rule		*rule;
1574 		int			 rs_num, i;
1575 
1576 		PF_RULES_WLOCK();
1577 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1578 		ruleset = pf_find_ruleset(pr->anchor);
1579 		if (ruleset == NULL) {
1580 			PF_RULES_WUNLOCK();
1581 			error = EINVAL;
1582 			break;
1583 		}
1584 		rs_num = pf_get_ruleset_number(pr->rule.action);
1585 		if (rs_num >= PF_RULESET_MAX) {
1586 			PF_RULES_WUNLOCK();
1587 			error = EINVAL;
1588 			break;
1589 		}
1590 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1591 			PF_RULES_WUNLOCK();
1592 			error = EBUSY;
1593 			break;
1594 		}
1595 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1596 		while ((rule != NULL) && (rule->nr != pr->nr))
1597 			rule = TAILQ_NEXT(rule, entries);
1598 		if (rule == NULL) {
1599 			PF_RULES_WUNLOCK();
1600 			error = EBUSY;
1601 			break;
1602 		}
1603 		bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1604 		pr->rule.u_states_cur = counter_u64_fetch(rule->states_cur);
1605 		pr->rule.u_states_tot = counter_u64_fetch(rule->states_tot);
1606 		pr->rule.u_src_nodes = counter_u64_fetch(rule->src_nodes);
1607 		if (pf_anchor_copyout(ruleset, rule, pr)) {
1608 			PF_RULES_WUNLOCK();
1609 			error = EBUSY;
1610 			break;
1611 		}
1612 		pf_addr_copyout(&pr->rule.src.addr);
1613 		pf_addr_copyout(&pr->rule.dst.addr);
1614 		for (i = 0; i < PF_SKIP_COUNT; ++i)
1615 			if (rule->skip[i].ptr == NULL)
1616 				pr->rule.skip[i].nr = -1;
1617 			else
1618 				pr->rule.skip[i].nr =
1619 				    rule->skip[i].ptr->nr;
1620 
1621 		if (pr->action == PF_GET_CLR_CNTR) {
1622 			rule->evaluations = 0;
1623 			rule->packets[0] = rule->packets[1] = 0;
1624 			rule->bytes[0] = rule->bytes[1] = 0;
1625 			counter_u64_zero(rule->states_tot);
1626 		}
1627 		PF_RULES_WUNLOCK();
1628 		break;
1629 	}
1630 
1631 	case DIOCCHANGERULE: {
1632 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1633 		struct pf_ruleset	*ruleset;
1634 		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1635 		struct pfi_kif		*kif = NULL;
1636 		struct pf_pooladdr	*pa;
1637 		u_int32_t		 nr = 0;
1638 		int			 rs_num;
1639 
1640 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1641 		    pcr->action > PF_CHANGE_GET_TICKET) {
1642 			error = EINVAL;
1643 			break;
1644 		}
1645 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1646 			error = EINVAL;
1647 			break;
1648 		}
1649 
1650 		if (pcr->action != PF_CHANGE_REMOVE) {
1651 #ifndef INET
1652 			if (pcr->rule.af == AF_INET) {
1653 				error = EAFNOSUPPORT;
1654 				break;
1655 			}
1656 #endif /* INET */
1657 #ifndef INET6
1658 			if (pcr->rule.af == AF_INET6) {
1659 				error = EAFNOSUPPORT;
1660 				break;
1661 			}
1662 #endif /* INET6 */
1663 			newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
1664 			bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1665 			if (newrule->ifname[0])
1666 				kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1667 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
1668 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
1669 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
1670 			newrule->cuid = td->td_ucred->cr_ruid;
1671 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1672 			TAILQ_INIT(&newrule->rpool.list);
1673 		}
1674 
1675 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGERULE_error; }
1676 
1677 		PF_RULES_WLOCK();
1678 		if (!(pcr->action == PF_CHANGE_REMOVE ||
1679 		    pcr->action == PF_CHANGE_GET_TICKET) &&
1680 		    pcr->pool_ticket != V_ticket_pabuf)
1681 			ERROUT(EBUSY);
1682 
1683 		ruleset = pf_find_ruleset(pcr->anchor);
1684 		if (ruleset == NULL)
1685 			ERROUT(EINVAL);
1686 
1687 		rs_num = pf_get_ruleset_number(pcr->rule.action);
1688 		if (rs_num >= PF_RULESET_MAX)
1689 			ERROUT(EINVAL);
1690 
1691 		if (pcr->action == PF_CHANGE_GET_TICKET) {
1692 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1693 			ERROUT(0);
1694 		} else if (pcr->ticket !=
1695 			    ruleset->rules[rs_num].active.ticket)
1696 				ERROUT(EINVAL);
1697 
1698 		if (pcr->action != PF_CHANGE_REMOVE) {
1699 			if (newrule->ifname[0]) {
1700 				newrule->kif = pfi_kif_attach(kif,
1701 				    newrule->ifname);
1702 				pfi_kif_ref(newrule->kif);
1703 			} else
1704 				newrule->kif = NULL;
1705 
1706 			if (newrule->rtableid > 0 &&
1707 			    newrule->rtableid >= rt_numfibs)
1708 				error = EBUSY;
1709 
1710 #ifdef ALTQ
1711 			/* set queue IDs */
1712 			if (newrule->qname[0] != 0) {
1713 				if ((newrule->qid =
1714 				    pf_qname2qid(newrule->qname)) == 0)
1715 					error = EBUSY;
1716 				else if (newrule->pqname[0] != 0) {
1717 					if ((newrule->pqid =
1718 					    pf_qname2qid(newrule->pqname)) == 0)
1719 						error = EBUSY;
1720 				} else
1721 					newrule->pqid = newrule->qid;
1722 			}
1723 #endif /* ALTQ */
1724 			if (newrule->tagname[0])
1725 				if ((newrule->tag =
1726 				    pf_tagname2tag(newrule->tagname)) == 0)
1727 					error = EBUSY;
1728 			if (newrule->match_tagname[0])
1729 				if ((newrule->match_tag = pf_tagname2tag(
1730 				    newrule->match_tagname)) == 0)
1731 					error = EBUSY;
1732 			if (newrule->rt && !newrule->direction)
1733 				error = EINVAL;
1734 			if (!newrule->log)
1735 				newrule->logif = 0;
1736 			if (newrule->logif >= PFLOGIFS_MAX)
1737 				error = EINVAL;
1738 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
1739 				error = ENOMEM;
1740 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
1741 				error = ENOMEM;
1742 			if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1743 				error = EINVAL;
1744 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1745 				if (pa->addr.type == PF_ADDR_TABLE) {
1746 					pa->addr.p.tbl =
1747 					    pfr_attach_table(ruleset,
1748 					    pa->addr.v.tblname);
1749 					if (pa->addr.p.tbl == NULL)
1750 						error = ENOMEM;
1751 				}
1752 
1753 			newrule->overload_tbl = NULL;
1754 			if (newrule->overload_tblname[0]) {
1755 				if ((newrule->overload_tbl = pfr_attach_table(
1756 				    ruleset, newrule->overload_tblname)) ==
1757 				    NULL)
1758 					error = EINVAL;
1759 				else
1760 					newrule->overload_tbl->pfrkt_flags |=
1761 					    PFR_TFLAG_ACTIVE;
1762 			}
1763 
1764 			pf_mv_pool(&V_pf_pabuf, &newrule->rpool.list);
1765 			if (((((newrule->action == PF_NAT) ||
1766 			    (newrule->action == PF_RDR) ||
1767 			    (newrule->action == PF_BINAT) ||
1768 			    (newrule->rt > PF_NOPFROUTE)) &&
1769 			    !newrule->anchor)) &&
1770 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1771 				error = EINVAL;
1772 
1773 			if (error) {
1774 				pf_free_rule(newrule);
1775 				PF_RULES_WUNLOCK();
1776 				break;
1777 			}
1778 
1779 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1780 			newrule->evaluations = 0;
1781 			newrule->packets[0] = newrule->packets[1] = 0;
1782 			newrule->bytes[0] = newrule->bytes[1] = 0;
1783 		}
1784 		pf_empty_pool(&V_pf_pabuf);
1785 
1786 		if (pcr->action == PF_CHANGE_ADD_HEAD)
1787 			oldrule = TAILQ_FIRST(
1788 			    ruleset->rules[rs_num].active.ptr);
1789 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1790 			oldrule = TAILQ_LAST(
1791 			    ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1792 		else {
1793 			oldrule = TAILQ_FIRST(
1794 			    ruleset->rules[rs_num].active.ptr);
1795 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1796 				oldrule = TAILQ_NEXT(oldrule, entries);
1797 			if (oldrule == NULL) {
1798 				if (newrule != NULL)
1799 					pf_free_rule(newrule);
1800 				PF_RULES_WUNLOCK();
1801 				error = EINVAL;
1802 				break;
1803 			}
1804 		}
1805 
1806 		if (pcr->action == PF_CHANGE_REMOVE) {
1807 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
1808 			    oldrule);
1809 			ruleset->rules[rs_num].active.rcount--;
1810 		} else {
1811 			if (oldrule == NULL)
1812 				TAILQ_INSERT_TAIL(
1813 				    ruleset->rules[rs_num].active.ptr,
1814 				    newrule, entries);
1815 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1816 			    pcr->action == PF_CHANGE_ADD_BEFORE)
1817 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1818 			else
1819 				TAILQ_INSERT_AFTER(
1820 				    ruleset->rules[rs_num].active.ptr,
1821 				    oldrule, newrule, entries);
1822 			ruleset->rules[rs_num].active.rcount++;
1823 		}
1824 
1825 		nr = 0;
1826 		TAILQ_FOREACH(oldrule,
1827 		    ruleset->rules[rs_num].active.ptr, entries)
1828 			oldrule->nr = nr++;
1829 
1830 		ruleset->rules[rs_num].active.ticket++;
1831 
1832 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1833 		pf_remove_if_empty_ruleset(ruleset);
1834 
1835 		PF_RULES_WUNLOCK();
1836 		break;
1837 
1838 #undef ERROUT
1839 DIOCCHANGERULE_error:
1840 		PF_RULES_WUNLOCK();
1841 		if (newrule != NULL) {
1842 			counter_u64_free(newrule->states_cur);
1843 			counter_u64_free(newrule->states_tot);
1844 			counter_u64_free(newrule->src_nodes);
1845 			free(newrule, M_PFRULE);
1846 		}
1847 		if (kif != NULL)
1848 			free(kif, PFI_MTYPE);
1849 		break;
1850 	}
1851 
1852 	case DIOCCLRSTATES: {
1853 		struct pf_state		*s;
1854 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1855 		u_int			 i, killed = 0;
1856 
1857 		for (i = 0; i <= pf_hashmask; i++) {
1858 			struct pf_idhash *ih = &V_pf_idhash[i];
1859 
1860 relock_DIOCCLRSTATES:
1861 			PF_HASHROW_LOCK(ih);
1862 			LIST_FOREACH(s, &ih->states, entry)
1863 				if (!psk->psk_ifname[0] ||
1864 				    !strcmp(psk->psk_ifname,
1865 				    s->kif->pfik_name)) {
1866 					/*
1867 					 * Don't send out individual
1868 					 * delete messages.
1869 					 */
1870 					s->state_flags |= PFSTATE_NOSYNC;
1871 					pf_unlink_state(s, PF_ENTER_LOCKED);
1872 					killed++;
1873 					goto relock_DIOCCLRSTATES;
1874 				}
1875 			PF_HASHROW_UNLOCK(ih);
1876 		}
1877 		psk->psk_killed = killed;
1878 		if (V_pfsync_clear_states_ptr != NULL)
1879 			V_pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
1880 		break;
1881 	}
1882 
1883 	case DIOCKILLSTATES: {
1884 		struct pf_state		*s;
1885 		struct pf_state_key	*sk;
1886 		struct pf_addr		*srcaddr, *dstaddr;
1887 		u_int16_t		 srcport, dstport;
1888 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
1889 		u_int			 i, killed = 0;
1890 
1891 		if (psk->psk_pfcmp.id) {
1892 			if (psk->psk_pfcmp.creatorid == 0)
1893 				psk->psk_pfcmp.creatorid = V_pf_status.hostid;
1894 			if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
1895 			    psk->psk_pfcmp.creatorid))) {
1896 				pf_unlink_state(s, PF_ENTER_LOCKED);
1897 				psk->psk_killed = 1;
1898 			}
1899 			break;
1900 		}
1901 
1902 		for (i = 0; i <= pf_hashmask; i++) {
1903 			struct pf_idhash *ih = &V_pf_idhash[i];
1904 
1905 relock_DIOCKILLSTATES:
1906 			PF_HASHROW_LOCK(ih);
1907 			LIST_FOREACH(s, &ih->states, entry) {
1908 				sk = s->key[PF_SK_WIRE];
1909 				if (s->direction == PF_OUT) {
1910 					srcaddr = &sk->addr[1];
1911 					dstaddr = &sk->addr[0];
1912 					srcport = sk->port[1];
1913 					dstport = sk->port[0];
1914 				} else {
1915 					srcaddr = &sk->addr[0];
1916 					dstaddr = &sk->addr[1];
1917 					srcport = sk->port[0];
1918 					dstport = sk->port[1];
1919 				}
1920 
1921 				if ((!psk->psk_af || sk->af == psk->psk_af)
1922 				    && (!psk->psk_proto || psk->psk_proto ==
1923 				    sk->proto) &&
1924 				    PF_MATCHA(psk->psk_src.neg,
1925 				    &psk->psk_src.addr.v.a.addr,
1926 				    &psk->psk_src.addr.v.a.mask,
1927 				    srcaddr, sk->af) &&
1928 				    PF_MATCHA(psk->psk_dst.neg,
1929 				    &psk->psk_dst.addr.v.a.addr,
1930 				    &psk->psk_dst.addr.v.a.mask,
1931 				    dstaddr, sk->af) &&
1932 				    (psk->psk_src.port_op == 0 ||
1933 				    pf_match_port(psk->psk_src.port_op,
1934 				    psk->psk_src.port[0], psk->psk_src.port[1],
1935 				    srcport)) &&
1936 				    (psk->psk_dst.port_op == 0 ||
1937 				    pf_match_port(psk->psk_dst.port_op,
1938 				    psk->psk_dst.port[0], psk->psk_dst.port[1],
1939 				    dstport)) &&
1940 				    (!psk->psk_label[0] ||
1941 				    (s->rule.ptr->label[0] &&
1942 				    !strcmp(psk->psk_label,
1943 				    s->rule.ptr->label))) &&
1944 				    (!psk->psk_ifname[0] ||
1945 				    !strcmp(psk->psk_ifname,
1946 				    s->kif->pfik_name))) {
1947 					pf_unlink_state(s, PF_ENTER_LOCKED);
1948 					killed++;
1949 					goto relock_DIOCKILLSTATES;
1950 				}
1951 			}
1952 			PF_HASHROW_UNLOCK(ih);
1953 		}
1954 		psk->psk_killed = killed;
1955 		break;
1956 	}
1957 
1958 	case DIOCADDSTATE: {
1959 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1960 		struct pfsync_state	*sp = &ps->state;
1961 
1962 		if (sp->timeout >= PFTM_MAX) {
1963 			error = EINVAL;
1964 			break;
1965 		}
1966 		if (V_pfsync_state_import_ptr != NULL) {
1967 			PF_RULES_RLOCK();
1968 			error = V_pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
1969 			PF_RULES_RUNLOCK();
1970 		} else
1971 			error = EOPNOTSUPP;
1972 		break;
1973 	}
1974 
1975 	case DIOCGETSTATE: {
1976 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1977 		struct pf_state		*s;
1978 
1979 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
1980 		if (s == NULL) {
1981 			error = ENOENT;
1982 			break;
1983 		}
1984 
1985 		pfsync_state_export(&ps->state, s);
1986 		PF_STATE_UNLOCK(s);
1987 		break;
1988 	}
1989 
1990 	case DIOCGETSTATES: {
1991 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
1992 		struct pf_state		*s;
1993 		struct pfsync_state	*pstore, *p;
1994 		int i, nr;
1995 
1996 		if (ps->ps_len == 0) {
1997 			nr = uma_zone_get_cur(V_pf_state_z);
1998 			ps->ps_len = sizeof(struct pfsync_state) * nr;
1999 			break;
2000 		}
2001 
2002 		p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK);
2003 		nr = 0;
2004 
2005 		for (i = 0; i <= pf_hashmask; i++) {
2006 			struct pf_idhash *ih = &V_pf_idhash[i];
2007 
2008 			PF_HASHROW_LOCK(ih);
2009 			LIST_FOREACH(s, &ih->states, entry) {
2010 
2011 				if (s->timeout == PFTM_UNLINKED)
2012 					continue;
2013 
2014 				if ((nr+1) * sizeof(*p) > ps->ps_len) {
2015 					PF_HASHROW_UNLOCK(ih);
2016 					goto DIOCGETSTATES_full;
2017 				}
2018 				pfsync_state_export(p, s);
2019 				p++;
2020 				nr++;
2021 			}
2022 			PF_HASHROW_UNLOCK(ih);
2023 		}
2024 DIOCGETSTATES_full:
2025 		error = copyout(pstore, ps->ps_states,
2026 		    sizeof(struct pfsync_state) * nr);
2027 		if (error) {
2028 			free(pstore, M_TEMP);
2029 			break;
2030 		}
2031 		ps->ps_len = sizeof(struct pfsync_state) * nr;
2032 		free(pstore, M_TEMP);
2033 
2034 		break;
2035 	}
2036 
2037 	case DIOCGETSTATUS: {
2038 		struct pf_status *s = (struct pf_status *)addr;
2039 
2040 		PF_RULES_RLOCK();
2041 		s->running = V_pf_status.running;
2042 		s->since   = V_pf_status.since;
2043 		s->debug   = V_pf_status.debug;
2044 		s->hostid  = V_pf_status.hostid;
2045 		s->states  = V_pf_status.states;
2046 		s->src_nodes = V_pf_status.src_nodes;
2047 
2048 		for (int i = 0; i < PFRES_MAX; i++)
2049 			s->counters[i] =
2050 			    counter_u64_fetch(V_pf_status.counters[i]);
2051 		for (int i = 0; i < LCNT_MAX; i++)
2052 			s->lcounters[i] =
2053 			    counter_u64_fetch(V_pf_status.lcounters[i]);
2054 		for (int i = 0; i < FCNT_MAX; i++)
2055 			s->fcounters[i] =
2056 			    counter_u64_fetch(V_pf_status.fcounters[i]);
2057 		for (int i = 0; i < SCNT_MAX; i++)
2058 			s->scounters[i] =
2059 			    counter_u64_fetch(V_pf_status.scounters[i]);
2060 
2061 		bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ);
2062 		bcopy(V_pf_status.pf_chksum, s->pf_chksum,
2063 		    PF_MD5_DIGEST_LENGTH);
2064 
2065 		pfi_update_status(s->ifname, s);
2066 		PF_RULES_RUNLOCK();
2067 		break;
2068 	}
2069 
2070 	case DIOCSETSTATUSIF: {
2071 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
2072 
2073 		if (pi->ifname[0] == 0) {
2074 			bzero(V_pf_status.ifname, IFNAMSIZ);
2075 			break;
2076 		}
2077 		PF_RULES_WLOCK();
2078 		strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
2079 		PF_RULES_WUNLOCK();
2080 		break;
2081 	}
2082 
2083 	case DIOCCLRSTATUS: {
2084 		PF_RULES_WLOCK();
2085 		for (int i = 0; i < PFRES_MAX; i++)
2086 			counter_u64_zero(V_pf_status.counters[i]);
2087 		for (int i = 0; i < FCNT_MAX; i++)
2088 			counter_u64_zero(V_pf_status.fcounters[i]);
2089 		for (int i = 0; i < SCNT_MAX; i++)
2090 			counter_u64_zero(V_pf_status.scounters[i]);
2091 		for (int i = 0; i < LCNT_MAX; i++)
2092 			counter_u64_zero(V_pf_status.lcounters[i]);
2093 		V_pf_status.since = time_second;
2094 		if (*V_pf_status.ifname)
2095 			pfi_update_status(V_pf_status.ifname, NULL);
2096 		PF_RULES_WUNLOCK();
2097 		break;
2098 	}
2099 
2100 	case DIOCNATLOOK: {
2101 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
2102 		struct pf_state_key	*sk;
2103 		struct pf_state		*state;
2104 		struct pf_state_key_cmp	 key;
2105 		int			 m = 0, direction = pnl->direction;
2106 		int			 sidx, didx;
2107 
2108 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
2109 		sidx = (direction == PF_IN) ? 1 : 0;
2110 		didx = (direction == PF_IN) ? 0 : 1;
2111 
2112 		if (!pnl->proto ||
2113 		    PF_AZERO(&pnl->saddr, pnl->af) ||
2114 		    PF_AZERO(&pnl->daddr, pnl->af) ||
2115 		    ((pnl->proto == IPPROTO_TCP ||
2116 		    pnl->proto == IPPROTO_UDP) &&
2117 		    (!pnl->dport || !pnl->sport)))
2118 			error = EINVAL;
2119 		else {
2120 			bzero(&key, sizeof(key));
2121 			key.af = pnl->af;
2122 			key.proto = pnl->proto;
2123 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
2124 			key.port[sidx] = pnl->sport;
2125 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
2126 			key.port[didx] = pnl->dport;
2127 
2128 			state = pf_find_state_all(&key, direction, &m);
2129 
2130 			if (m > 1)
2131 				error = E2BIG;	/* more than one state */
2132 			else if (state != NULL) {
2133 				/* XXXGL: not locked read */
2134 				sk = state->key[sidx];
2135 				PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
2136 				pnl->rsport = sk->port[sidx];
2137 				PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
2138 				pnl->rdport = sk->port[didx];
2139 			} else
2140 				error = ENOENT;
2141 		}
2142 		break;
2143 	}
2144 
2145 	case DIOCSETTIMEOUT: {
2146 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2147 		int		 old;
2148 
2149 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
2150 		    pt->seconds < 0) {
2151 			error = EINVAL;
2152 			break;
2153 		}
2154 		PF_RULES_WLOCK();
2155 		old = V_pf_default_rule.timeout[pt->timeout];
2156 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
2157 			pt->seconds = 1;
2158 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
2159 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
2160 			wakeup(pf_purge_thread);
2161 		pt->seconds = old;
2162 		PF_RULES_WUNLOCK();
2163 		break;
2164 	}
2165 
2166 	case DIOCGETTIMEOUT: {
2167 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2168 
2169 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
2170 			error = EINVAL;
2171 			break;
2172 		}
2173 		PF_RULES_RLOCK();
2174 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
2175 		PF_RULES_RUNLOCK();
2176 		break;
2177 	}
2178 
2179 	case DIOCGETLIMIT: {
2180 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2181 
2182 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
2183 			error = EINVAL;
2184 			break;
2185 		}
2186 		PF_RULES_RLOCK();
2187 		pl->limit = V_pf_limits[pl->index].limit;
2188 		PF_RULES_RUNLOCK();
2189 		break;
2190 	}
2191 
2192 	case DIOCSETLIMIT: {
2193 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2194 		int			 old_limit;
2195 
2196 		PF_RULES_WLOCK();
2197 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
2198 		    V_pf_limits[pl->index].zone == NULL) {
2199 			PF_RULES_WUNLOCK();
2200 			error = EINVAL;
2201 			break;
2202 		}
2203 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
2204 		old_limit = V_pf_limits[pl->index].limit;
2205 		V_pf_limits[pl->index].limit = pl->limit;
2206 		pl->limit = old_limit;
2207 		PF_RULES_WUNLOCK();
2208 		break;
2209 	}
2210 
2211 	case DIOCSETDEBUG: {
2212 		u_int32_t	*level = (u_int32_t *)addr;
2213 
2214 		PF_RULES_WLOCK();
2215 		V_pf_status.debug = *level;
2216 		PF_RULES_WUNLOCK();
2217 		break;
2218 	}
2219 
2220 	case DIOCCLRRULECTRS: {
2221 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
2222 		struct pf_ruleset	*ruleset = &pf_main_ruleset;
2223 		struct pf_rule		*rule;
2224 
2225 		PF_RULES_WLOCK();
2226 		TAILQ_FOREACH(rule,
2227 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
2228 			rule->evaluations = 0;
2229 			rule->packets[0] = rule->packets[1] = 0;
2230 			rule->bytes[0] = rule->bytes[1] = 0;
2231 		}
2232 		PF_RULES_WUNLOCK();
2233 		break;
2234 	}
2235 
2236 	case DIOCGIFSPEEDV0:
2237 	case DIOCGIFSPEEDV1: {
2238 		struct pf_ifspeed_v1	*psp = (struct pf_ifspeed_v1 *)addr;
2239 		struct pf_ifspeed_v1	ps;
2240 		struct ifnet		*ifp;
2241 
2242 		if (psp->ifname[0] != 0) {
2243 			/* Can we completely trust user-land? */
2244 			strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
2245 			ifp = ifunit(ps.ifname);
2246 			if (ifp != NULL) {
2247 				psp->baudrate32 =
2248 				    (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
2249 				if (cmd == DIOCGIFSPEEDV1)
2250 					psp->baudrate = ifp->if_baudrate;
2251 			} else
2252 				error = EINVAL;
2253 		} else
2254 			error = EINVAL;
2255 		break;
2256 	}
2257 
2258 #ifdef ALTQ
2259 	case DIOCSTARTALTQ: {
2260 		struct pf_altq		*altq;
2261 
2262 		PF_RULES_WLOCK();
2263 		/* enable all altq interfaces on active list */
2264 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
2265 			if (altq->qname[0] == 0 && (altq->local_flags &
2266 			    PFALTQ_FLAG_IF_REMOVED) == 0) {
2267 				error = pf_enable_altq(altq);
2268 				if (error != 0)
2269 					break;
2270 			}
2271 		}
2272 		if (error == 0)
2273 			V_pf_altq_running = 1;
2274 		PF_RULES_WUNLOCK();
2275 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
2276 		break;
2277 	}
2278 
2279 	case DIOCSTOPALTQ: {
2280 		struct pf_altq		*altq;
2281 
2282 		PF_RULES_WLOCK();
2283 		/* disable all altq interfaces on active list */
2284 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
2285 			if (altq->qname[0] == 0 && (altq->local_flags &
2286 			    PFALTQ_FLAG_IF_REMOVED) == 0) {
2287 				error = pf_disable_altq(altq);
2288 				if (error != 0)
2289 					break;
2290 			}
2291 		}
2292 		if (error == 0)
2293 			V_pf_altq_running = 0;
2294 		PF_RULES_WUNLOCK();
2295 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2296 		break;
2297 	}
2298 
2299 	case DIOCADDALTQV0:
2300 	case DIOCADDALTQV1: {
2301 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2302 		struct pf_altq		*altq, *a;
2303 		struct ifnet		*ifp;
2304 
2305 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
2306 		error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
2307 		if (error)
2308 			break;
2309 		altq->local_flags = 0;
2310 
2311 		PF_RULES_WLOCK();
2312 		if (pa->ticket != V_ticket_altqs_inactive) {
2313 			PF_RULES_WUNLOCK();
2314 			free(altq, M_PFALTQ);
2315 			error = EBUSY;
2316 			break;
2317 		}
2318 
2319 		/*
2320 		 * if this is for a queue, find the discipline and
2321 		 * copy the necessary fields
2322 		 */
2323 		if (altq->qname[0] != 0) {
2324 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2325 				PF_RULES_WUNLOCK();
2326 				error = EBUSY;
2327 				free(altq, M_PFALTQ);
2328 				break;
2329 			}
2330 			altq->altq_disc = NULL;
2331 			TAILQ_FOREACH(a, V_pf_altqs_inactive, entries) {
2332 				if (strncmp(a->ifname, altq->ifname,
2333 				    IFNAMSIZ) == 0 && a->qname[0] == 0) {
2334 					altq->altq_disc = a->altq_disc;
2335 					break;
2336 				}
2337 			}
2338 		}
2339 
2340 		if ((ifp = ifunit(altq->ifname)) == NULL)
2341 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
2342 		else
2343 			error = altq_add(altq);
2344 
2345 		if (error) {
2346 			PF_RULES_WUNLOCK();
2347 			free(altq, M_PFALTQ);
2348 			break;
2349 		}
2350 
2351 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
2352 		/* version error check done on import above */
2353 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
2354 		PF_RULES_WUNLOCK();
2355 		break;
2356 	}
2357 
2358 	case DIOCGETALTQSV0:
2359 	case DIOCGETALTQSV1: {
2360 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2361 		struct pf_altq		*altq;
2362 
2363 		PF_RULES_RLOCK();
2364 		pa->nr = 0;
2365 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
2366 			pa->nr++;
2367 		pa->ticket = V_ticket_altqs_active;
2368 		PF_RULES_RUNLOCK();
2369 		break;
2370 	}
2371 
2372 	case DIOCGETALTQV0:
2373 	case DIOCGETALTQV1: {
2374 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2375 		struct pf_altq		*altq;
2376 		u_int32_t		 nr;
2377 
2378 		PF_RULES_RLOCK();
2379 		if (pa->ticket != V_ticket_altqs_active) {
2380 			PF_RULES_RUNLOCK();
2381 			error = EBUSY;
2382 			break;
2383 		}
2384 		nr = 0;
2385 		altq = TAILQ_FIRST(V_pf_altqs_active);
2386 		while ((altq != NULL) && (nr < pa->nr)) {
2387 			altq = TAILQ_NEXT(altq, entries);
2388 			nr++;
2389 		}
2390 		if (altq == NULL) {
2391 			PF_RULES_RUNLOCK();
2392 			error = EBUSY;
2393 			break;
2394 		}
2395 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
2396 		PF_RULES_RUNLOCK();
2397 		break;
2398 	}
2399 
2400 	case DIOCCHANGEALTQV0:
2401 	case DIOCCHANGEALTQV1:
2402 		/* CHANGEALTQ not supported yet! */
2403 		error = ENODEV;
2404 		break;
2405 
2406 	case DIOCGETQSTATSV0:
2407 	case DIOCGETQSTATSV1: {
2408 		struct pfioc_qstats_v1	*pq = (struct pfioc_qstats_v1 *)addr;
2409 		struct pf_altq		*altq;
2410 		u_int32_t		 nr;
2411 		int			 nbytes;
2412 		u_int32_t		 version;
2413 
2414 		PF_RULES_RLOCK();
2415 		if (pq->ticket != V_ticket_altqs_active) {
2416 			PF_RULES_RUNLOCK();
2417 			error = EBUSY;
2418 			break;
2419 		}
2420 		nbytes = pq->nbytes;
2421 		nr = 0;
2422 		altq = TAILQ_FIRST(V_pf_altqs_active);
2423 		while ((altq != NULL) && (nr < pq->nr)) {
2424 			altq = TAILQ_NEXT(altq, entries);
2425 			nr++;
2426 		}
2427 		if (altq == NULL) {
2428 			PF_RULES_RUNLOCK();
2429 			error = EBUSY;
2430 			break;
2431 		}
2432 
2433 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
2434 			PF_RULES_RUNLOCK();
2435 			error = ENXIO;
2436 			break;
2437 		}
2438 		PF_RULES_RUNLOCK();
2439 		if (cmd == DIOCGETQSTATSV0)
2440 			version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
2441 		else
2442 			version = pq->version;
2443 		error = altq_getqstats(altq, pq->buf, &nbytes, version);
2444 		if (error == 0) {
2445 			pq->scheduler = altq->scheduler;
2446 			pq->nbytes = nbytes;
2447 		}
2448 		break;
2449 	}
2450 #endif /* ALTQ */
2451 
2452 	case DIOCBEGINADDRS: {
2453 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2454 
2455 		PF_RULES_WLOCK();
2456 		pf_empty_pool(&V_pf_pabuf);
2457 		pp->ticket = ++V_ticket_pabuf;
2458 		PF_RULES_WUNLOCK();
2459 		break;
2460 	}
2461 
2462 	case DIOCADDADDR: {
2463 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2464 		struct pf_pooladdr	*pa;
2465 		struct pfi_kif		*kif = NULL;
2466 
2467 #ifndef INET
2468 		if (pp->af == AF_INET) {
2469 			error = EAFNOSUPPORT;
2470 			break;
2471 		}
2472 #endif /* INET */
2473 #ifndef INET6
2474 		if (pp->af == AF_INET6) {
2475 			error = EAFNOSUPPORT;
2476 			break;
2477 		}
2478 #endif /* INET6 */
2479 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2480 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2481 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2482 			error = EINVAL;
2483 			break;
2484 		}
2485 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2486 		bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2487 		if (pa->ifname[0])
2488 			kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2489 		PF_RULES_WLOCK();
2490 		if (pp->ticket != V_ticket_pabuf) {
2491 			PF_RULES_WUNLOCK();
2492 			if (pa->ifname[0])
2493 				free(kif, PFI_MTYPE);
2494 			free(pa, M_PFRULE);
2495 			error = EBUSY;
2496 			break;
2497 		}
2498 		if (pa->ifname[0]) {
2499 			pa->kif = pfi_kif_attach(kif, pa->ifname);
2500 			pfi_kif_ref(pa->kif);
2501 		} else
2502 			pa->kif = NULL;
2503 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2504 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2505 			if (pa->ifname[0])
2506 				pfi_kif_unref(pa->kif);
2507 			PF_RULES_WUNLOCK();
2508 			free(pa, M_PFRULE);
2509 			break;
2510 		}
2511 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2512 		PF_RULES_WUNLOCK();
2513 		break;
2514 	}
2515 
2516 	case DIOCGETADDRS: {
2517 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2518 		struct pf_pool		*pool;
2519 		struct pf_pooladdr	*pa;
2520 
2521 		PF_RULES_RLOCK();
2522 		pp->nr = 0;
2523 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2524 		    pp->r_num, 0, 1, 0);
2525 		if (pool == NULL) {
2526 			PF_RULES_RUNLOCK();
2527 			error = EBUSY;
2528 			break;
2529 		}
2530 		TAILQ_FOREACH(pa, &pool->list, entries)
2531 			pp->nr++;
2532 		PF_RULES_RUNLOCK();
2533 		break;
2534 	}
2535 
2536 	case DIOCGETADDR: {
2537 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2538 		struct pf_pool		*pool;
2539 		struct pf_pooladdr	*pa;
2540 		u_int32_t		 nr = 0;
2541 
2542 		PF_RULES_RLOCK();
2543 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2544 		    pp->r_num, 0, 1, 1);
2545 		if (pool == NULL) {
2546 			PF_RULES_RUNLOCK();
2547 			error = EBUSY;
2548 			break;
2549 		}
2550 		pa = TAILQ_FIRST(&pool->list);
2551 		while ((pa != NULL) && (nr < pp->nr)) {
2552 			pa = TAILQ_NEXT(pa, entries);
2553 			nr++;
2554 		}
2555 		if (pa == NULL) {
2556 			PF_RULES_RUNLOCK();
2557 			error = EBUSY;
2558 			break;
2559 		}
2560 		bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2561 		pf_addr_copyout(&pp->addr.addr);
2562 		PF_RULES_RUNLOCK();
2563 		break;
2564 	}
2565 
2566 	case DIOCCHANGEADDR: {
2567 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
2568 		struct pf_pool		*pool;
2569 		struct pf_pooladdr	*oldpa = NULL, *newpa = NULL;
2570 		struct pf_ruleset	*ruleset;
2571 		struct pfi_kif		*kif = NULL;
2572 
2573 		if (pca->action < PF_CHANGE_ADD_HEAD ||
2574 		    pca->action > PF_CHANGE_REMOVE) {
2575 			error = EINVAL;
2576 			break;
2577 		}
2578 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2579 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2580 		    pca->addr.addr.type != PF_ADDR_TABLE) {
2581 			error = EINVAL;
2582 			break;
2583 		}
2584 
2585 		if (pca->action != PF_CHANGE_REMOVE) {
2586 #ifndef INET
2587 			if (pca->af == AF_INET) {
2588 				error = EAFNOSUPPORT;
2589 				break;
2590 			}
2591 #endif /* INET */
2592 #ifndef INET6
2593 			if (pca->af == AF_INET6) {
2594 				error = EAFNOSUPPORT;
2595 				break;
2596 			}
2597 #endif /* INET6 */
2598 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
2599 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2600 			if (newpa->ifname[0])
2601 				kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2602 			newpa->kif = NULL;
2603 		}
2604 
2605 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGEADDR_error; }
2606 		PF_RULES_WLOCK();
2607 		ruleset = pf_find_ruleset(pca->anchor);
2608 		if (ruleset == NULL)
2609 			ERROUT(EBUSY);
2610 
2611 		pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2612 		    pca->r_num, pca->r_last, 1, 1);
2613 		if (pool == NULL)
2614 			ERROUT(EBUSY);
2615 
2616 		if (pca->action != PF_CHANGE_REMOVE) {
2617 			if (newpa->ifname[0]) {
2618 				newpa->kif = pfi_kif_attach(kif, newpa->ifname);
2619 				pfi_kif_ref(newpa->kif);
2620 				kif = NULL;
2621 			}
2622 
2623 			switch (newpa->addr.type) {
2624 			case PF_ADDR_DYNIFTL:
2625 				error = pfi_dynaddr_setup(&newpa->addr,
2626 				    pca->af);
2627 				break;
2628 			case PF_ADDR_TABLE:
2629 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
2630 				    newpa->addr.v.tblname);
2631 				if (newpa->addr.p.tbl == NULL)
2632 					error = ENOMEM;
2633 				break;
2634 			}
2635 			if (error)
2636 				goto DIOCCHANGEADDR_error;
2637 		}
2638 
2639 		switch (pca->action) {
2640 		case PF_CHANGE_ADD_HEAD:
2641 			oldpa = TAILQ_FIRST(&pool->list);
2642 			break;
2643 		case PF_CHANGE_ADD_TAIL:
2644 			oldpa = TAILQ_LAST(&pool->list, pf_palist);
2645 			break;
2646 		default:
2647 			oldpa = TAILQ_FIRST(&pool->list);
2648 			for (int i = 0; oldpa && i < pca->nr; i++)
2649 				oldpa = TAILQ_NEXT(oldpa, entries);
2650 
2651 			if (oldpa == NULL)
2652 				ERROUT(EINVAL);
2653 		}
2654 
2655 		if (pca->action == PF_CHANGE_REMOVE) {
2656 			TAILQ_REMOVE(&pool->list, oldpa, entries);
2657 			switch (oldpa->addr.type) {
2658 			case PF_ADDR_DYNIFTL:
2659 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
2660 				break;
2661 			case PF_ADDR_TABLE:
2662 				pfr_detach_table(oldpa->addr.p.tbl);
2663 				break;
2664 			}
2665 			if (oldpa->kif)
2666 				pfi_kif_unref(oldpa->kif);
2667 			free(oldpa, M_PFRULE);
2668 		} else {
2669 			if (oldpa == NULL)
2670 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2671 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
2672 			    pca->action == PF_CHANGE_ADD_BEFORE)
2673 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2674 			else
2675 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
2676 				    newpa, entries);
2677 		}
2678 
2679 		pool->cur = TAILQ_FIRST(&pool->list);
2680 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
2681 		PF_RULES_WUNLOCK();
2682 		break;
2683 
2684 #undef ERROUT
2685 DIOCCHANGEADDR_error:
2686 		if (newpa != NULL) {
2687 			if (newpa->kif)
2688 				pfi_kif_unref(newpa->kif);
2689 			free(newpa, M_PFRULE);
2690 		}
2691 		PF_RULES_WUNLOCK();
2692 		if (kif != NULL)
2693 			free(kif, PFI_MTYPE);
2694 		break;
2695 	}
2696 
2697 	case DIOCGETRULESETS: {
2698 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2699 		struct pf_ruleset	*ruleset;
2700 		struct pf_anchor	*anchor;
2701 
2702 		PF_RULES_RLOCK();
2703 		pr->path[sizeof(pr->path) - 1] = 0;
2704 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2705 			PF_RULES_RUNLOCK();
2706 			error = ENOENT;
2707 			break;
2708 		}
2709 		pr->nr = 0;
2710 		if (ruleset->anchor == NULL) {
2711 			/* XXX kludge for pf_main_ruleset */
2712 			RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2713 				if (anchor->parent == NULL)
2714 					pr->nr++;
2715 		} else {
2716 			RB_FOREACH(anchor, pf_anchor_node,
2717 			    &ruleset->anchor->children)
2718 				pr->nr++;
2719 		}
2720 		PF_RULES_RUNLOCK();
2721 		break;
2722 	}
2723 
2724 	case DIOCGETRULESET: {
2725 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2726 		struct pf_ruleset	*ruleset;
2727 		struct pf_anchor	*anchor;
2728 		u_int32_t		 nr = 0;
2729 
2730 		PF_RULES_RLOCK();
2731 		pr->path[sizeof(pr->path) - 1] = 0;
2732 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2733 			PF_RULES_RUNLOCK();
2734 			error = ENOENT;
2735 			break;
2736 		}
2737 		pr->name[0] = 0;
2738 		if (ruleset->anchor == NULL) {
2739 			/* XXX kludge for pf_main_ruleset */
2740 			RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2741 				if (anchor->parent == NULL && nr++ == pr->nr) {
2742 					strlcpy(pr->name, anchor->name,
2743 					    sizeof(pr->name));
2744 					break;
2745 				}
2746 		} else {
2747 			RB_FOREACH(anchor, pf_anchor_node,
2748 			    &ruleset->anchor->children)
2749 				if (nr++ == pr->nr) {
2750 					strlcpy(pr->name, anchor->name,
2751 					    sizeof(pr->name));
2752 					break;
2753 				}
2754 		}
2755 		if (!pr->name[0])
2756 			error = EBUSY;
2757 		PF_RULES_RUNLOCK();
2758 		break;
2759 	}
2760 
2761 	case DIOCRCLRTABLES: {
2762 		struct pfioc_table *io = (struct pfioc_table *)addr;
2763 
2764 		if (io->pfrio_esize != 0) {
2765 			error = ENODEV;
2766 			break;
2767 		}
2768 		PF_RULES_WLOCK();
2769 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2770 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2771 		PF_RULES_WUNLOCK();
2772 		break;
2773 	}
2774 
2775 	case DIOCRADDTABLES: {
2776 		struct pfioc_table *io = (struct pfioc_table *)addr;
2777 		struct pfr_table *pfrts;
2778 		size_t totlen;
2779 
2780 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2781 			error = ENODEV;
2782 			break;
2783 		}
2784 
2785 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
2786 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
2787 			error = ENOMEM;
2788 			break;
2789 		}
2790 
2791 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2792 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
2793 		    M_TEMP, M_WAITOK);
2794 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2795 		if (error) {
2796 			free(pfrts, M_TEMP);
2797 			break;
2798 		}
2799 		PF_RULES_WLOCK();
2800 		error = pfr_add_tables(pfrts, io->pfrio_size,
2801 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2802 		PF_RULES_WUNLOCK();
2803 		free(pfrts, M_TEMP);
2804 		break;
2805 	}
2806 
2807 	case DIOCRDELTABLES: {
2808 		struct pfioc_table *io = (struct pfioc_table *)addr;
2809 		struct pfr_table *pfrts;
2810 		size_t totlen;
2811 
2812 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2813 			error = ENODEV;
2814 			break;
2815 		}
2816 
2817 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
2818 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
2819 			error = ENOMEM;
2820 			break;
2821 		}
2822 
2823 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2824 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
2825 		    M_TEMP, M_WAITOK);
2826 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2827 		if (error) {
2828 			free(pfrts, M_TEMP);
2829 			break;
2830 		}
2831 		PF_RULES_WLOCK();
2832 		error = pfr_del_tables(pfrts, io->pfrio_size,
2833 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2834 		PF_RULES_WUNLOCK();
2835 		free(pfrts, M_TEMP);
2836 		break;
2837 	}
2838 
2839 	case DIOCRGETTABLES: {
2840 		struct pfioc_table *io = (struct pfioc_table *)addr;
2841 		struct pfr_table *pfrts;
2842 		size_t totlen, n;
2843 
2844 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2845 			error = ENODEV;
2846 			break;
2847 		}
2848 		PF_RULES_RLOCK();
2849 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
2850 		io->pfrio_size = min(io->pfrio_size, n);
2851 
2852 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2853 
2854 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
2855 		    M_TEMP, M_NOWAIT);
2856 		if (pfrts == NULL) {
2857 			error = ENOMEM;
2858 			PF_RULES_RUNLOCK();
2859 			break;
2860 		}
2861 		error = pfr_get_tables(&io->pfrio_table, pfrts,
2862 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2863 		PF_RULES_RUNLOCK();
2864 		if (error == 0)
2865 			error = copyout(pfrts, io->pfrio_buffer, totlen);
2866 		free(pfrts, M_TEMP);
2867 		break;
2868 	}
2869 
2870 	case DIOCRGETTSTATS: {
2871 		struct pfioc_table *io = (struct pfioc_table *)addr;
2872 		struct pfr_tstats *pfrtstats;
2873 		size_t totlen, n;
2874 
2875 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2876 			error = ENODEV;
2877 			break;
2878 		}
2879 		PF_RULES_WLOCK();
2880 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
2881 		io->pfrio_size = min(io->pfrio_size, n);
2882 
2883 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
2884 		pfrtstats = mallocarray(io->pfrio_size,
2885 		    sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT);
2886 		if (pfrtstats == NULL) {
2887 			error = ENOMEM;
2888 			PF_RULES_WUNLOCK();
2889 			break;
2890 		}
2891 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
2892 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2893 		PF_RULES_WUNLOCK();
2894 		if (error == 0)
2895 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
2896 		free(pfrtstats, M_TEMP);
2897 		break;
2898 	}
2899 
2900 	case DIOCRCLRTSTATS: {
2901 		struct pfioc_table *io = (struct pfioc_table *)addr;
2902 		struct pfr_table *pfrts;
2903 		size_t totlen, n;
2904 
2905 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2906 			error = ENODEV;
2907 			break;
2908 		}
2909 
2910 		PF_RULES_WLOCK();
2911 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
2912 		io->pfrio_size = min(io->pfrio_size, n);
2913 
2914 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2915 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
2916 		    M_TEMP, M_NOWAIT);
2917 		if (pfrts == NULL) {
2918 			error = ENOMEM;
2919 			PF_RULES_WUNLOCK();
2920 			break;
2921 		}
2922 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2923 		if (error) {
2924 			free(pfrts, M_TEMP);
2925 			PF_RULES_WUNLOCK();
2926 			break;
2927 		}
2928 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
2929 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2930 		PF_RULES_WUNLOCK();
2931 		free(pfrts, M_TEMP);
2932 		break;
2933 	}
2934 
2935 	case DIOCRSETTFLAGS: {
2936 		struct pfioc_table *io = (struct pfioc_table *)addr;
2937 		struct pfr_table *pfrts;
2938 		size_t totlen, n;
2939 
2940 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2941 			error = ENODEV;
2942 			break;
2943 		}
2944 
2945 		PF_RULES_WLOCK();
2946 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
2947 		io->pfrio_size = min(io->pfrio_size, n);
2948 
2949 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2950 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
2951 		    M_TEMP, M_NOWAIT);
2952 		if (pfrts == NULL) {
2953 			error = ENOMEM;
2954 			PF_RULES_WUNLOCK();
2955 			break;
2956 		}
2957 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2958 		if (error) {
2959 			free(pfrts, M_TEMP);
2960 			PF_RULES_WUNLOCK();
2961 			break;
2962 		}
2963 		error = pfr_set_tflags(pfrts, io->pfrio_size,
2964 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2965 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2966 		PF_RULES_WUNLOCK();
2967 		free(pfrts, M_TEMP);
2968 		break;
2969 	}
2970 
2971 	case DIOCRCLRADDRS: {
2972 		struct pfioc_table *io = (struct pfioc_table *)addr;
2973 
2974 		if (io->pfrio_esize != 0) {
2975 			error = ENODEV;
2976 			break;
2977 		}
2978 		PF_RULES_WLOCK();
2979 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2980 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2981 		PF_RULES_WUNLOCK();
2982 		break;
2983 	}
2984 
2985 	case DIOCRADDADDRS: {
2986 		struct pfioc_table *io = (struct pfioc_table *)addr;
2987 		struct pfr_addr *pfras;
2988 		size_t totlen;
2989 
2990 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2991 			error = ENODEV;
2992 			break;
2993 		}
2994 		if (io->pfrio_size < 0 ||
2995 		    io->pfrio_size > pf_ioctl_maxcount ||
2996 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
2997 			error = EINVAL;
2998 			break;
2999 		}
3000 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3001 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3002 		    M_TEMP, M_NOWAIT);
3003 		if (! pfras) {
3004 			error = ENOMEM;
3005 			break;
3006 		}
3007 		error = copyin(io->pfrio_buffer, pfras, totlen);
3008 		if (error) {
3009 			free(pfras, M_TEMP);
3010 			break;
3011 		}
3012 		PF_RULES_WLOCK();
3013 		error = pfr_add_addrs(&io->pfrio_table, pfras,
3014 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
3015 		    PFR_FLAG_USERIOCTL);
3016 		PF_RULES_WUNLOCK();
3017 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3018 			error = copyout(pfras, io->pfrio_buffer, totlen);
3019 		free(pfras, M_TEMP);
3020 		break;
3021 	}
3022 
3023 	case DIOCRDELADDRS: {
3024 		struct pfioc_table *io = (struct pfioc_table *)addr;
3025 		struct pfr_addr *pfras;
3026 		size_t totlen;
3027 
3028 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3029 			error = ENODEV;
3030 			break;
3031 		}
3032 		if (io->pfrio_size < 0 ||
3033 		    io->pfrio_size > pf_ioctl_maxcount ||
3034 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3035 			error = EINVAL;
3036 			break;
3037 		}
3038 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3039 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3040 		    M_TEMP, M_NOWAIT);
3041 		if (! pfras) {
3042 			error = ENOMEM;
3043 			break;
3044 		}
3045 		error = copyin(io->pfrio_buffer, pfras, totlen);
3046 		if (error) {
3047 			free(pfras, M_TEMP);
3048 			break;
3049 		}
3050 		PF_RULES_WLOCK();
3051 		error = pfr_del_addrs(&io->pfrio_table, pfras,
3052 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
3053 		    PFR_FLAG_USERIOCTL);
3054 		PF_RULES_WUNLOCK();
3055 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3056 			error = copyout(pfras, io->pfrio_buffer, totlen);
3057 		free(pfras, M_TEMP);
3058 		break;
3059 	}
3060 
3061 	case DIOCRSETADDRS: {
3062 		struct pfioc_table *io = (struct pfioc_table *)addr;
3063 		struct pfr_addr *pfras;
3064 		size_t totlen, count;
3065 
3066 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3067 			error = ENODEV;
3068 			break;
3069 		}
3070 		if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
3071 			error = EINVAL;
3072 			break;
3073 		}
3074 		count = max(io->pfrio_size, io->pfrio_size2);
3075 		if (count > pf_ioctl_maxcount ||
3076 		    WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
3077 			error = EINVAL;
3078 			break;
3079 		}
3080 		totlen = count * sizeof(struct pfr_addr);
3081 		pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
3082 		    M_NOWAIT);
3083 		if (! pfras) {
3084 			error = ENOMEM;
3085 			break;
3086 		}
3087 		error = copyin(io->pfrio_buffer, pfras, totlen);
3088 		if (error) {
3089 			free(pfras, M_TEMP);
3090 			break;
3091 		}
3092 		PF_RULES_WLOCK();
3093 		error = pfr_set_addrs(&io->pfrio_table, pfras,
3094 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
3095 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
3096 		    PFR_FLAG_USERIOCTL, 0);
3097 		PF_RULES_WUNLOCK();
3098 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3099 			error = copyout(pfras, io->pfrio_buffer, totlen);
3100 		free(pfras, M_TEMP);
3101 		break;
3102 	}
3103 
3104 	case DIOCRGETADDRS: {
3105 		struct pfioc_table *io = (struct pfioc_table *)addr;
3106 		struct pfr_addr *pfras;
3107 		size_t totlen;
3108 
3109 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3110 			error = ENODEV;
3111 			break;
3112 		}
3113 		if (io->pfrio_size < 0 ||
3114 		    io->pfrio_size > pf_ioctl_maxcount ||
3115 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3116 			error = EINVAL;
3117 			break;
3118 		}
3119 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3120 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3121 		    M_TEMP, M_NOWAIT);
3122 		if (! pfras) {
3123 			error = ENOMEM;
3124 			break;
3125 		}
3126 		PF_RULES_RLOCK();
3127 		error = pfr_get_addrs(&io->pfrio_table, pfras,
3128 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3129 		PF_RULES_RUNLOCK();
3130 		if (error == 0)
3131 			error = copyout(pfras, io->pfrio_buffer, totlen);
3132 		free(pfras, M_TEMP);
3133 		break;
3134 	}
3135 
3136 	case DIOCRGETASTATS: {
3137 		struct pfioc_table *io = (struct pfioc_table *)addr;
3138 		struct pfr_astats *pfrastats;
3139 		size_t totlen;
3140 
3141 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
3142 			error = ENODEV;
3143 			break;
3144 		}
3145 		if (io->pfrio_size < 0 ||
3146 		    io->pfrio_size > pf_ioctl_maxcount ||
3147 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
3148 			error = EINVAL;
3149 			break;
3150 		}
3151 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
3152 		pfrastats = mallocarray(io->pfrio_size,
3153 		    sizeof(struct pfr_astats), M_TEMP, M_NOWAIT);
3154 		if (! pfrastats) {
3155 			error = ENOMEM;
3156 			break;
3157 		}
3158 		PF_RULES_RLOCK();
3159 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
3160 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3161 		PF_RULES_RUNLOCK();
3162 		if (error == 0)
3163 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
3164 		free(pfrastats, M_TEMP);
3165 		break;
3166 	}
3167 
3168 	case DIOCRCLRASTATS: {
3169 		struct pfioc_table *io = (struct pfioc_table *)addr;
3170 		struct pfr_addr *pfras;
3171 		size_t totlen;
3172 
3173 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3174 			error = ENODEV;
3175 			break;
3176 		}
3177 		if (io->pfrio_size < 0 ||
3178 		    io->pfrio_size > pf_ioctl_maxcount ||
3179 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3180 			error = EINVAL;
3181 			break;
3182 		}
3183 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3184 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3185 		    M_TEMP, M_NOWAIT);
3186 		if (! pfras) {
3187 			error = ENOMEM;
3188 			break;
3189 		}
3190 		error = copyin(io->pfrio_buffer, pfras, totlen);
3191 		if (error) {
3192 			free(pfras, M_TEMP);
3193 			break;
3194 		}
3195 		PF_RULES_WLOCK();
3196 		error = pfr_clr_astats(&io->pfrio_table, pfras,
3197 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
3198 		    PFR_FLAG_USERIOCTL);
3199 		PF_RULES_WUNLOCK();
3200 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3201 			error = copyout(pfras, io->pfrio_buffer, totlen);
3202 		free(pfras, M_TEMP);
3203 		break;
3204 	}
3205 
3206 	case DIOCRTSTADDRS: {
3207 		struct pfioc_table *io = (struct pfioc_table *)addr;
3208 		struct pfr_addr *pfras;
3209 		size_t totlen;
3210 
3211 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3212 			error = ENODEV;
3213 			break;
3214 		}
3215 		if (io->pfrio_size < 0 ||
3216 		    io->pfrio_size > pf_ioctl_maxcount ||
3217 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3218 			error = EINVAL;
3219 			break;
3220 		}
3221 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3222 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3223 		    M_TEMP, M_NOWAIT);
3224 		if (! pfras) {
3225 			error = ENOMEM;
3226 			break;
3227 		}
3228 		error = copyin(io->pfrio_buffer, pfras, totlen);
3229 		if (error) {
3230 			free(pfras, M_TEMP);
3231 			break;
3232 		}
3233 		PF_RULES_RLOCK();
3234 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
3235 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
3236 		    PFR_FLAG_USERIOCTL);
3237 		PF_RULES_RUNLOCK();
3238 		if (error == 0)
3239 			error = copyout(pfras, io->pfrio_buffer, totlen);
3240 		free(pfras, M_TEMP);
3241 		break;
3242 	}
3243 
3244 	case DIOCRINADEFINE: {
3245 		struct pfioc_table *io = (struct pfioc_table *)addr;
3246 		struct pfr_addr *pfras;
3247 		size_t totlen;
3248 
3249 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3250 			error = ENODEV;
3251 			break;
3252 		}
3253 		if (io->pfrio_size < 0 ||
3254 		    io->pfrio_size > pf_ioctl_maxcount ||
3255 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3256 			error = EINVAL;
3257 			break;
3258 		}
3259 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3260 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3261 		    M_TEMP, M_NOWAIT);
3262 		if (! pfras) {
3263 			error = ENOMEM;
3264 			break;
3265 		}
3266 		error = copyin(io->pfrio_buffer, pfras, totlen);
3267 		if (error) {
3268 			free(pfras, M_TEMP);
3269 			break;
3270 		}
3271 		PF_RULES_WLOCK();
3272 		error = pfr_ina_define(&io->pfrio_table, pfras,
3273 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
3274 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3275 		PF_RULES_WUNLOCK();
3276 		free(pfras, M_TEMP);
3277 		break;
3278 	}
3279 
3280 	case DIOCOSFPADD: {
3281 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
3282 		PF_RULES_WLOCK();
3283 		error = pf_osfp_add(io);
3284 		PF_RULES_WUNLOCK();
3285 		break;
3286 	}
3287 
3288 	case DIOCOSFPGET: {
3289 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
3290 		PF_RULES_RLOCK();
3291 		error = pf_osfp_get(io);
3292 		PF_RULES_RUNLOCK();
3293 		break;
3294 	}
3295 
3296 	case DIOCXBEGIN: {
3297 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3298 		struct pfioc_trans_e	*ioes, *ioe;
3299 		size_t			 totlen;
3300 		int			 i;
3301 
3302 		if (io->esize != sizeof(*ioe)) {
3303 			error = ENODEV;
3304 			break;
3305 		}
3306 		if (io->size < 0 ||
3307 		    io->size > pf_ioctl_maxcount ||
3308 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3309 			error = EINVAL;
3310 			break;
3311 		}
3312 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3313 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3314 		    M_TEMP, M_NOWAIT);
3315 		if (! ioes) {
3316 			error = ENOMEM;
3317 			break;
3318 		}
3319 		error = copyin(io->array, ioes, totlen);
3320 		if (error) {
3321 			free(ioes, M_TEMP);
3322 			break;
3323 		}
3324 		PF_RULES_WLOCK();
3325 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3326 			switch (ioe->rs_num) {
3327 #ifdef ALTQ
3328 			case PF_RULESET_ALTQ:
3329 				if (ioe->anchor[0]) {
3330 					PF_RULES_WUNLOCK();
3331 					free(ioes, M_TEMP);
3332 					error = EINVAL;
3333 					goto fail;
3334 				}
3335 				if ((error = pf_begin_altq(&ioe->ticket))) {
3336 					PF_RULES_WUNLOCK();
3337 					free(ioes, M_TEMP);
3338 					goto fail;
3339 				}
3340 				break;
3341 #endif /* ALTQ */
3342 			case PF_RULESET_TABLE:
3343 			    {
3344 				struct pfr_table table;
3345 
3346 				bzero(&table, sizeof(table));
3347 				strlcpy(table.pfrt_anchor, ioe->anchor,
3348 				    sizeof(table.pfrt_anchor));
3349 				if ((error = pfr_ina_begin(&table,
3350 				    &ioe->ticket, NULL, 0))) {
3351 					PF_RULES_WUNLOCK();
3352 					free(ioes, M_TEMP);
3353 					goto fail;
3354 				}
3355 				break;
3356 			    }
3357 			default:
3358 				if ((error = pf_begin_rules(&ioe->ticket,
3359 				    ioe->rs_num, ioe->anchor))) {
3360 					PF_RULES_WUNLOCK();
3361 					free(ioes, M_TEMP);
3362 					goto fail;
3363 				}
3364 				break;
3365 			}
3366 		}
3367 		PF_RULES_WUNLOCK();
3368 		error = copyout(ioes, io->array, totlen);
3369 		free(ioes, M_TEMP);
3370 		break;
3371 	}
3372 
3373 	case DIOCXROLLBACK: {
3374 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3375 		struct pfioc_trans_e	*ioe, *ioes;
3376 		size_t			 totlen;
3377 		int			 i;
3378 
3379 		if (io->esize != sizeof(*ioe)) {
3380 			error = ENODEV;
3381 			break;
3382 		}
3383 		if (io->size < 0 ||
3384 		    io->size > pf_ioctl_maxcount ||
3385 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3386 			error = EINVAL;
3387 			break;
3388 		}
3389 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3390 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3391 		    M_TEMP, M_NOWAIT);
3392 		if (! ioes) {
3393 			error = ENOMEM;
3394 			break;
3395 		}
3396 		error = copyin(io->array, ioes, totlen);
3397 		if (error) {
3398 			free(ioes, M_TEMP);
3399 			break;
3400 		}
3401 		PF_RULES_WLOCK();
3402 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3403 			switch (ioe->rs_num) {
3404 #ifdef ALTQ
3405 			case PF_RULESET_ALTQ:
3406 				if (ioe->anchor[0]) {
3407 					PF_RULES_WUNLOCK();
3408 					free(ioes, M_TEMP);
3409 					error = EINVAL;
3410 					goto fail;
3411 				}
3412 				if ((error = pf_rollback_altq(ioe->ticket))) {
3413 					PF_RULES_WUNLOCK();
3414 					free(ioes, M_TEMP);
3415 					goto fail; /* really bad */
3416 				}
3417 				break;
3418 #endif /* ALTQ */
3419 			case PF_RULESET_TABLE:
3420 			    {
3421 				struct pfr_table table;
3422 
3423 				bzero(&table, sizeof(table));
3424 				strlcpy(table.pfrt_anchor, ioe->anchor,
3425 				    sizeof(table.pfrt_anchor));
3426 				if ((error = pfr_ina_rollback(&table,
3427 				    ioe->ticket, NULL, 0))) {
3428 					PF_RULES_WUNLOCK();
3429 					free(ioes, M_TEMP);
3430 					goto fail; /* really bad */
3431 				}
3432 				break;
3433 			    }
3434 			default:
3435 				if ((error = pf_rollback_rules(ioe->ticket,
3436 				    ioe->rs_num, ioe->anchor))) {
3437 					PF_RULES_WUNLOCK();
3438 					free(ioes, M_TEMP);
3439 					goto fail; /* really bad */
3440 				}
3441 				break;
3442 			}
3443 		}
3444 		PF_RULES_WUNLOCK();
3445 		free(ioes, M_TEMP);
3446 		break;
3447 	}
3448 
3449 	case DIOCXCOMMIT: {
3450 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3451 		struct pfioc_trans_e	*ioe, *ioes;
3452 		struct pf_ruleset	*rs;
3453 		size_t			 totlen;
3454 		int			 i;
3455 
3456 		if (io->esize != sizeof(*ioe)) {
3457 			error = ENODEV;
3458 			break;
3459 		}
3460 
3461 		if (io->size < 0 ||
3462 		    io->size > pf_ioctl_maxcount ||
3463 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3464 			error = EINVAL;
3465 			break;
3466 		}
3467 
3468 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3469 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3470 		    M_TEMP, M_NOWAIT);
3471 		if (ioes == NULL) {
3472 			error = ENOMEM;
3473 			break;
3474 		}
3475 		error = copyin(io->array, ioes, totlen);
3476 		if (error) {
3477 			free(ioes, M_TEMP);
3478 			break;
3479 		}
3480 		PF_RULES_WLOCK();
3481 		/* First makes sure everything will succeed. */
3482 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3483 			switch (ioe->rs_num) {
3484 #ifdef ALTQ
3485 			case PF_RULESET_ALTQ:
3486 				if (ioe->anchor[0]) {
3487 					PF_RULES_WUNLOCK();
3488 					free(ioes, M_TEMP);
3489 					error = EINVAL;
3490 					goto fail;
3491 				}
3492 				if (!V_altqs_inactive_open || ioe->ticket !=
3493 				    V_ticket_altqs_inactive) {
3494 					PF_RULES_WUNLOCK();
3495 					free(ioes, M_TEMP);
3496 					error = EBUSY;
3497 					goto fail;
3498 				}
3499 				break;
3500 #endif /* ALTQ */
3501 			case PF_RULESET_TABLE:
3502 				rs = pf_find_ruleset(ioe->anchor);
3503 				if (rs == NULL || !rs->topen || ioe->ticket !=
3504 				    rs->tticket) {
3505 					PF_RULES_WUNLOCK();
3506 					free(ioes, M_TEMP);
3507 					error = EBUSY;
3508 					goto fail;
3509 				}
3510 				break;
3511 			default:
3512 				if (ioe->rs_num < 0 || ioe->rs_num >=
3513 				    PF_RULESET_MAX) {
3514 					PF_RULES_WUNLOCK();
3515 					free(ioes, M_TEMP);
3516 					error = EINVAL;
3517 					goto fail;
3518 				}
3519 				rs = pf_find_ruleset(ioe->anchor);
3520 				if (rs == NULL ||
3521 				    !rs->rules[ioe->rs_num].inactive.open ||
3522 				    rs->rules[ioe->rs_num].inactive.ticket !=
3523 				    ioe->ticket) {
3524 					PF_RULES_WUNLOCK();
3525 					free(ioes, M_TEMP);
3526 					error = EBUSY;
3527 					goto fail;
3528 				}
3529 				break;
3530 			}
3531 		}
3532 		/* Now do the commit - no errors should happen here. */
3533 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3534 			switch (ioe->rs_num) {
3535 #ifdef ALTQ
3536 			case PF_RULESET_ALTQ:
3537 				if ((error = pf_commit_altq(ioe->ticket))) {
3538 					PF_RULES_WUNLOCK();
3539 					free(ioes, M_TEMP);
3540 					goto fail; /* really bad */
3541 				}
3542 				break;
3543 #endif /* ALTQ */
3544 			case PF_RULESET_TABLE:
3545 			    {
3546 				struct pfr_table table;
3547 
3548 				bzero(&table, sizeof(table));
3549 				strlcpy(table.pfrt_anchor, ioe->anchor,
3550 				    sizeof(table.pfrt_anchor));
3551 				if ((error = pfr_ina_commit(&table,
3552 				    ioe->ticket, NULL, NULL, 0))) {
3553 					PF_RULES_WUNLOCK();
3554 					free(ioes, M_TEMP);
3555 					goto fail; /* really bad */
3556 				}
3557 				break;
3558 			    }
3559 			default:
3560 				if ((error = pf_commit_rules(ioe->ticket,
3561 				    ioe->rs_num, ioe->anchor))) {
3562 					PF_RULES_WUNLOCK();
3563 					free(ioes, M_TEMP);
3564 					goto fail; /* really bad */
3565 				}
3566 				break;
3567 			}
3568 		}
3569 		PF_RULES_WUNLOCK();
3570 		free(ioes, M_TEMP);
3571 		break;
3572 	}
3573 
3574 	case DIOCGETSRCNODES: {
3575 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
3576 		struct pf_srchash	*sh;
3577 		struct pf_src_node	*n, *p, *pstore;
3578 		uint32_t		 i, nr = 0;
3579 
3580 		if (psn->psn_len == 0) {
3581 			for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3582 			    i++, sh++) {
3583 				PF_HASHROW_LOCK(sh);
3584 				LIST_FOREACH(n, &sh->nodes, entry)
3585 					nr++;
3586 				PF_HASHROW_UNLOCK(sh);
3587 			}
3588 			psn->psn_len = sizeof(struct pf_src_node) * nr;
3589 			break;
3590 		}
3591 
3592 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK);
3593 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3594 		    i++, sh++) {
3595 		    PF_HASHROW_LOCK(sh);
3596 		    LIST_FOREACH(n, &sh->nodes, entry) {
3597 			int	secs = time_uptime, diff;
3598 
3599 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
3600 				break;
3601 
3602 			bcopy(n, p, sizeof(struct pf_src_node));
3603 			if (n->rule.ptr != NULL)
3604 				p->rule.nr = n->rule.ptr->nr;
3605 			p->creation = secs - p->creation;
3606 			if (p->expire > secs)
3607 				p->expire -= secs;
3608 			else
3609 				p->expire = 0;
3610 
3611 			/* Adjust the connection rate estimate. */
3612 			diff = secs - n->conn_rate.last;
3613 			if (diff >= n->conn_rate.seconds)
3614 				p->conn_rate.count = 0;
3615 			else
3616 				p->conn_rate.count -=
3617 				    n->conn_rate.count * diff /
3618 				    n->conn_rate.seconds;
3619 			p++;
3620 			nr++;
3621 		    }
3622 		    PF_HASHROW_UNLOCK(sh);
3623 		}
3624 		error = copyout(pstore, psn->psn_src_nodes,
3625 		    sizeof(struct pf_src_node) * nr);
3626 		if (error) {
3627 			free(pstore, M_TEMP);
3628 			break;
3629 		}
3630 		psn->psn_len = sizeof(struct pf_src_node) * nr;
3631 		free(pstore, M_TEMP);
3632 		break;
3633 	}
3634 
3635 	case DIOCCLRSRCNODES: {
3636 
3637 		pf_clear_srcnodes(NULL);
3638 		pf_purge_expired_src_nodes();
3639 		break;
3640 	}
3641 
3642 	case DIOCKILLSRCNODES:
3643 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
3644 		break;
3645 
3646 	case DIOCSETHOSTID: {
3647 		u_int32_t	*hostid = (u_int32_t *)addr;
3648 
3649 		PF_RULES_WLOCK();
3650 		if (*hostid == 0)
3651 			V_pf_status.hostid = arc4random();
3652 		else
3653 			V_pf_status.hostid = *hostid;
3654 		PF_RULES_WUNLOCK();
3655 		break;
3656 	}
3657 
3658 	case DIOCOSFPFLUSH:
3659 		PF_RULES_WLOCK();
3660 		pf_osfp_flush();
3661 		PF_RULES_WUNLOCK();
3662 		break;
3663 
3664 	case DIOCIGETIFACES: {
3665 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3666 		struct pfi_kif *ifstore;
3667 		size_t bufsiz;
3668 
3669 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
3670 			error = ENODEV;
3671 			break;
3672 		}
3673 
3674 		if (io->pfiio_size < 0 ||
3675 		    io->pfiio_size > pf_ioctl_maxcount ||
3676 		    WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
3677 			error = EINVAL;
3678 			break;
3679 		}
3680 
3681 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
3682 		ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
3683 		    M_TEMP, M_NOWAIT);
3684 		if (ifstore == NULL) {
3685 			error = ENOMEM;
3686 			break;
3687 		}
3688 
3689 		PF_RULES_RLOCK();
3690 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
3691 		PF_RULES_RUNLOCK();
3692 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
3693 		free(ifstore, M_TEMP);
3694 		break;
3695 	}
3696 
3697 	case DIOCSETIFFLAG: {
3698 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3699 
3700 		PF_RULES_WLOCK();
3701 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3702 		PF_RULES_WUNLOCK();
3703 		break;
3704 	}
3705 
3706 	case DIOCCLRIFFLAG: {
3707 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3708 
3709 		PF_RULES_WLOCK();
3710 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3711 		PF_RULES_WUNLOCK();
3712 		break;
3713 	}
3714 
3715 	default:
3716 		error = ENODEV;
3717 		break;
3718 	}
3719 fail:
3720 	if (sx_xlocked(&pf_ioctl_lock))
3721 		sx_xunlock(&pf_ioctl_lock);
3722 	CURVNET_RESTORE();
3723 
3724 	return (error);
3725 }
3726 
3727 void
3728 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
3729 {
3730 	bzero(sp, sizeof(struct pfsync_state));
3731 
3732 	/* copy from state key */
3733 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
3734 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
3735 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
3736 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
3737 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
3738 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
3739 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
3740 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
3741 	sp->proto = st->key[PF_SK_WIRE]->proto;
3742 	sp->af = st->key[PF_SK_WIRE]->af;
3743 
3744 	/* copy from state */
3745 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
3746 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
3747 	sp->creation = htonl(time_uptime - st->creation);
3748 	sp->expire = pf_state_expires(st);
3749 	if (sp->expire <= time_uptime)
3750 		sp->expire = htonl(0);
3751 	else
3752 		sp->expire = htonl(sp->expire - time_uptime);
3753 
3754 	sp->direction = st->direction;
3755 	sp->log = st->log;
3756 	sp->timeout = st->timeout;
3757 	sp->state_flags = st->state_flags;
3758 	if (st->src_node)
3759 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
3760 	if (st->nat_src_node)
3761 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
3762 
3763 	sp->id = st->id;
3764 	sp->creatorid = st->creatorid;
3765 	pf_state_peer_hton(&st->src, &sp->src);
3766 	pf_state_peer_hton(&st->dst, &sp->dst);
3767 
3768 	if (st->rule.ptr == NULL)
3769 		sp->rule = htonl(-1);
3770 	else
3771 		sp->rule = htonl(st->rule.ptr->nr);
3772 	if (st->anchor.ptr == NULL)
3773 		sp->anchor = htonl(-1);
3774 	else
3775 		sp->anchor = htonl(st->anchor.ptr->nr);
3776 	if (st->nat_rule.ptr == NULL)
3777 		sp->nat_rule = htonl(-1);
3778 	else
3779 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
3780 
3781 	pf_state_counter_hton(st->packets[0], sp->packets[0]);
3782 	pf_state_counter_hton(st->packets[1], sp->packets[1]);
3783 	pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
3784 	pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
3785 
3786 }
3787 
3788 static void
3789 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
3790 {
3791 	struct pfr_ktable *kt;
3792 
3793 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
3794 
3795 	kt = aw->p.tbl;
3796 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
3797 		kt = kt->pfrkt_root;
3798 	aw->p.tbl = NULL;
3799 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
3800 		kt->pfrkt_cnt : -1;
3801 }
3802 
3803 /*
3804  * XXX - Check for version missmatch!!!
3805  */
3806 static void
3807 pf_clear_states(void)
3808 {
3809 	struct pf_state	*s;
3810 	u_int i;
3811 
3812 	for (i = 0; i <= pf_hashmask; i++) {
3813 		struct pf_idhash *ih = &V_pf_idhash[i];
3814 relock:
3815 		PF_HASHROW_LOCK(ih);
3816 		LIST_FOREACH(s, &ih->states, entry) {
3817 			s->timeout = PFTM_PURGE;
3818 			/* Don't send out individual delete messages. */
3819 			s->state_flags |= PFSTATE_NOSYNC;
3820 			pf_unlink_state(s, PF_ENTER_LOCKED);
3821 			goto relock;
3822 		}
3823 		PF_HASHROW_UNLOCK(ih);
3824 	}
3825 }
3826 
3827 static int
3828 pf_clear_tables(void)
3829 {
3830 	struct pfioc_table io;
3831 	int error;
3832 
3833 	bzero(&io, sizeof(io));
3834 
3835 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
3836 	    io.pfrio_flags);
3837 
3838 	return (error);
3839 }
3840 
3841 static void
3842 pf_clear_srcnodes(struct pf_src_node *n)
3843 {
3844 	struct pf_state *s;
3845 	int i;
3846 
3847 	for (i = 0; i <= pf_hashmask; i++) {
3848 		struct pf_idhash *ih = &V_pf_idhash[i];
3849 
3850 		PF_HASHROW_LOCK(ih);
3851 		LIST_FOREACH(s, &ih->states, entry) {
3852 			if (n == NULL || n == s->src_node)
3853 				s->src_node = NULL;
3854 			if (n == NULL || n == s->nat_src_node)
3855 				s->nat_src_node = NULL;
3856 		}
3857 		PF_HASHROW_UNLOCK(ih);
3858 	}
3859 
3860 	if (n == NULL) {
3861 		struct pf_srchash *sh;
3862 
3863 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3864 		    i++, sh++) {
3865 			PF_HASHROW_LOCK(sh);
3866 			LIST_FOREACH(n, &sh->nodes, entry) {
3867 				n->expire = 1;
3868 				n->states = 0;
3869 			}
3870 			PF_HASHROW_UNLOCK(sh);
3871 		}
3872 	} else {
3873 		/* XXX: hash slot should already be locked here. */
3874 		n->expire = 1;
3875 		n->states = 0;
3876 	}
3877 }
3878 
3879 static void
3880 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
3881 {
3882 	struct pf_src_node_list	 kill;
3883 
3884 	LIST_INIT(&kill);
3885 	for (int i = 0; i <= pf_srchashmask; i++) {
3886 		struct pf_srchash *sh = &V_pf_srchash[i];
3887 		struct pf_src_node *sn, *tmp;
3888 
3889 		PF_HASHROW_LOCK(sh);
3890 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
3891 			if (PF_MATCHA(psnk->psnk_src.neg,
3892 			      &psnk->psnk_src.addr.v.a.addr,
3893 			      &psnk->psnk_src.addr.v.a.mask,
3894 			      &sn->addr, sn->af) &&
3895 			    PF_MATCHA(psnk->psnk_dst.neg,
3896 			      &psnk->psnk_dst.addr.v.a.addr,
3897 			      &psnk->psnk_dst.addr.v.a.mask,
3898 			      &sn->raddr, sn->af)) {
3899 				pf_unlink_src_node(sn);
3900 				LIST_INSERT_HEAD(&kill, sn, entry);
3901 				sn->expire = 1;
3902 			}
3903 		PF_HASHROW_UNLOCK(sh);
3904 	}
3905 
3906 	for (int i = 0; i <= pf_hashmask; i++) {
3907 		struct pf_idhash *ih = &V_pf_idhash[i];
3908 		struct pf_state *s;
3909 
3910 		PF_HASHROW_LOCK(ih);
3911 		LIST_FOREACH(s, &ih->states, entry) {
3912 			if (s->src_node && s->src_node->expire == 1)
3913 				s->src_node = NULL;
3914 			if (s->nat_src_node && s->nat_src_node->expire == 1)
3915 				s->nat_src_node = NULL;
3916 		}
3917 		PF_HASHROW_UNLOCK(ih);
3918 	}
3919 
3920 	psnk->psnk_killed = pf_free_src_nodes(&kill);
3921 }
3922 
3923 /*
3924  * XXX - Check for version missmatch!!!
3925  */
3926 
3927 /*
3928  * Duplicate pfctl -Fa operation to get rid of as much as we can.
3929  */
3930 static int
3931 shutdown_pf(void)
3932 {
3933 	int error = 0;
3934 	u_int32_t t[5];
3935 	char nn = '\0';
3936 
3937 	do {
3938 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
3939 		    != 0) {
3940 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
3941 			break;
3942 		}
3943 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
3944 		    != 0) {
3945 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
3946 			break;		/* XXX: rollback? */
3947 		}
3948 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
3949 		    != 0) {
3950 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
3951 			break;		/* XXX: rollback? */
3952 		}
3953 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
3954 		    != 0) {
3955 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
3956 			break;		/* XXX: rollback? */
3957 		}
3958 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
3959 		    != 0) {
3960 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
3961 			break;		/* XXX: rollback? */
3962 		}
3963 
3964 		/* XXX: these should always succeed here */
3965 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
3966 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
3967 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
3968 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
3969 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
3970 
3971 		if ((error = pf_clear_tables()) != 0)
3972 			break;
3973 
3974 #ifdef ALTQ
3975 		if ((error = pf_begin_altq(&t[0])) != 0) {
3976 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
3977 			break;
3978 		}
3979 		pf_commit_altq(t[0]);
3980 #endif
3981 
3982 		pf_clear_states();
3983 
3984 		pf_clear_srcnodes(NULL);
3985 
3986 		/* status does not use malloced mem so no need to cleanup */
3987 		/* fingerprints and interfaces have their own cleanup code */
3988 
3989 		/* Free counters last as we updated them during shutdown. */
3990 		counter_u64_free(V_pf_default_rule.states_cur);
3991 		counter_u64_free(V_pf_default_rule.states_tot);
3992 		counter_u64_free(V_pf_default_rule.src_nodes);
3993 
3994 		for (int i = 0; i < PFRES_MAX; i++)
3995 			counter_u64_free(V_pf_status.counters[i]);
3996 		for (int i = 0; i < LCNT_MAX; i++)
3997 			counter_u64_free(V_pf_status.lcounters[i]);
3998 		for (int i = 0; i < FCNT_MAX; i++)
3999 			counter_u64_free(V_pf_status.fcounters[i]);
4000 		for (int i = 0; i < SCNT_MAX; i++)
4001 			counter_u64_free(V_pf_status.scounters[i]);
4002 	} while(0);
4003 
4004 	return (error);
4005 }
4006 
4007 #ifdef INET
4008 static int
4009 pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir, int flags,
4010     struct inpcb *inp)
4011 {
4012 	int chk;
4013 
4014 	chk = pf_test(PF_IN, flags, ifp, m, inp);
4015 	if (chk && *m) {
4016 		m_freem(*m);
4017 		*m = NULL;
4018 	}
4019 
4020 	if (chk != PF_PASS)
4021 		return (EACCES);
4022 	return (0);
4023 }
4024 
4025 static int
4026 pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir, int flags,
4027     struct inpcb *inp)
4028 {
4029 	int chk;
4030 
4031 	chk = pf_test(PF_OUT, flags, ifp, m, inp);
4032 	if (chk && *m) {
4033 		m_freem(*m);
4034 		*m = NULL;
4035 	}
4036 
4037 	if (chk != PF_PASS)
4038 		return (EACCES);
4039 	return (0);
4040 }
4041 #endif
4042 
4043 #ifdef INET6
4044 static int
4045 pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir, int flags,
4046     struct inpcb *inp)
4047 {
4048 	int chk;
4049 
4050 	/*
4051 	 * In case of loopback traffic IPv6 uses the real interface in
4052 	 * order to support scoped addresses. In order to support stateful
4053 	 * filtering we have change this to lo0 as it is the case in IPv4.
4054 	 */
4055 	CURVNET_SET(ifp->if_vnet);
4056 	chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
4057 	CURVNET_RESTORE();
4058 	if (chk && *m) {
4059 		m_freem(*m);
4060 		*m = NULL;
4061 	}
4062 	if (chk != PF_PASS)
4063 		return (EACCES);
4064 	return (0);
4065 }
4066 
4067 static int
4068 pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir, int flags,
4069     struct inpcb *inp)
4070 {
4071 	int chk;
4072 
4073 	CURVNET_SET(ifp->if_vnet);
4074 	chk = pf_test6(PF_OUT, flags, ifp, m, inp);
4075 	CURVNET_RESTORE();
4076 	if (chk && *m) {
4077 		m_freem(*m);
4078 		*m = NULL;
4079 	}
4080 	if (chk != PF_PASS)
4081 		return (EACCES);
4082 	return (0);
4083 }
4084 #endif /* INET6 */
4085 
4086 static int
4087 hook_pf(void)
4088 {
4089 #ifdef INET
4090 	struct pfil_head *pfh_inet;
4091 #endif
4092 #ifdef INET6
4093 	struct pfil_head *pfh_inet6;
4094 #endif
4095 
4096 	if (V_pf_pfil_hooked)
4097 		return (0);
4098 
4099 #ifdef INET
4100 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
4101 	if (pfh_inet == NULL)
4102 		return (ESRCH); /* XXX */
4103 	pfil_add_hook_flags(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet);
4104 	pfil_add_hook_flags(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet);
4105 #endif
4106 #ifdef INET6
4107 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
4108 	if (pfh_inet6 == NULL) {
4109 #ifdef INET
4110 		pfil_remove_hook_flags(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
4111 		    pfh_inet);
4112 		pfil_remove_hook_flags(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
4113 		    pfh_inet);
4114 #endif
4115 		return (ESRCH); /* XXX */
4116 	}
4117 	pfil_add_hook_flags(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet6);
4118 	pfil_add_hook_flags(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet6);
4119 #endif
4120 
4121 	V_pf_pfil_hooked = 1;
4122 	return (0);
4123 }
4124 
4125 static int
4126 dehook_pf(void)
4127 {
4128 #ifdef INET
4129 	struct pfil_head *pfh_inet;
4130 #endif
4131 #ifdef INET6
4132 	struct pfil_head *pfh_inet6;
4133 #endif
4134 
4135 	if (V_pf_pfil_hooked == 0)
4136 		return (0);
4137 
4138 #ifdef INET
4139 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
4140 	if (pfh_inet == NULL)
4141 		return (ESRCH); /* XXX */
4142 	pfil_remove_hook_flags(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
4143 	    pfh_inet);
4144 	pfil_remove_hook_flags(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
4145 	    pfh_inet);
4146 #endif
4147 #ifdef INET6
4148 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
4149 	if (pfh_inet6 == NULL)
4150 		return (ESRCH); /* XXX */
4151 	pfil_remove_hook_flags(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK,
4152 	    pfh_inet6);
4153 	pfil_remove_hook_flags(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK,
4154 	    pfh_inet6);
4155 #endif
4156 
4157 	V_pf_pfil_hooked = 0;
4158 	return (0);
4159 }
4160 
4161 static void
4162 pf_load_vnet(void)
4163 {
4164 	TAILQ_INIT(&V_pf_tags);
4165 	TAILQ_INIT(&V_pf_qids);
4166 
4167 	pfattach_vnet();
4168 	V_pf_vnet_active = 1;
4169 }
4170 
4171 static int
4172 pf_load(void)
4173 {
4174 	int error;
4175 
4176 	rm_init(&pf_rules_lock, "pf rulesets");
4177 	sx_init(&pf_ioctl_lock, "pf ioctl");
4178 	sx_init(&pf_end_lock, "pf end thread");
4179 
4180 	pf_mtag_initialize();
4181 
4182 	pf_dev = make_dev(&pf_cdevsw, 0, 0, 0, 0600, PF_NAME);
4183 	if (pf_dev == NULL)
4184 		return (ENOMEM);
4185 
4186 	pf_end_threads = 0;
4187 	error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
4188 	if (error != 0)
4189 		return (error);
4190 
4191 	pfi_initialize();
4192 
4193 	return (0);
4194 }
4195 
4196 static void
4197 pf_unload_vnet(void)
4198 {
4199 	int error;
4200 
4201 	V_pf_vnet_active = 0;
4202 	V_pf_status.running = 0;
4203 	error = dehook_pf();
4204 	if (error) {
4205 		/*
4206 		 * Should not happen!
4207 		 * XXX Due to error code ESRCH, kldunload will show
4208 		 * a message like 'No such process'.
4209 		 */
4210 		printf("%s : pfil unregisteration fail\n", __FUNCTION__);
4211 		return;
4212 	}
4213 
4214 	PF_RULES_WLOCK();
4215 	shutdown_pf();
4216 	PF_RULES_WUNLOCK();
4217 
4218 	swi_remove(V_pf_swi_cookie);
4219 
4220 	pf_unload_vnet_purge();
4221 
4222 	pf_normalize_cleanup();
4223 	PF_RULES_WLOCK();
4224 	pfi_cleanup_vnet();
4225 	PF_RULES_WUNLOCK();
4226 	pfr_cleanup();
4227 	pf_osfp_flush();
4228 	pf_cleanup();
4229 	if (IS_DEFAULT_VNET(curvnet))
4230 		pf_mtag_cleanup();
4231 }
4232 
4233 static void
4234 pf_unload(void)
4235 {
4236 
4237 	sx_xlock(&pf_end_lock);
4238 	pf_end_threads = 1;
4239 	while (pf_end_threads < 2) {
4240 		wakeup_one(pf_purge_thread);
4241 		sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
4242 	}
4243 	sx_xunlock(&pf_end_lock);
4244 
4245 	if (pf_dev != NULL)
4246 		destroy_dev(pf_dev);
4247 
4248 	pfi_cleanup();
4249 
4250 	rm_destroy(&pf_rules_lock);
4251 	sx_destroy(&pf_ioctl_lock);
4252 	sx_destroy(&pf_end_lock);
4253 }
4254 
4255 static void
4256 vnet_pf_init(void *unused __unused)
4257 {
4258 
4259 	pf_load_vnet();
4260 }
4261 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
4262     vnet_pf_init, NULL);
4263 
4264 static void
4265 vnet_pf_uninit(const void *unused __unused)
4266 {
4267 
4268 	pf_unload_vnet();
4269 }
4270 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
4271 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
4272     vnet_pf_uninit, NULL);
4273 
4274 
4275 static int
4276 pf_modevent(module_t mod, int type, void *data)
4277 {
4278 	int error = 0;
4279 
4280 	switch(type) {
4281 	case MOD_LOAD:
4282 		error = pf_load();
4283 		break;
4284 	case MOD_UNLOAD:
4285 		/* Handled in SYSUNINIT(pf_unload) to ensure it's done after
4286 		 * the vnet_pf_uninit()s */
4287 		break;
4288 	default:
4289 		error = EINVAL;
4290 		break;
4291 	}
4292 
4293 	return (error);
4294 }
4295 
4296 static moduledata_t pf_mod = {
4297 	"pf",
4298 	pf_modevent,
4299 	0
4300 };
4301 
4302 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
4303 MODULE_VERSION(pf, PF_MODVER);
4304