xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision 0572ccaa4543b0abef8ef81e384c1d04de9f3da1)
1 /*-
2  * Copyright (c) 2001 Daniel Hartmeier
3  * Copyright (c) 2002,2003 Henning Brauer
4  * Copyright (c) 2012 Gleb Smirnoff <glebius@FreeBSD.org>
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  *
11  *    - Redistributions of source code must retain the above copyright
12  *      notice, this list of conditions and the following disclaimer.
13  *    - Redistributions in binary form must reproduce the above
14  *      copyright notice, this list of conditions and the following
15  *      disclaimer in the documentation and/or other materials provided
16  *      with the distribution.
17  *
18  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
19  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
20  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
21  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
22  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
23  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
24  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
25  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
26  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
28  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29  * POSSIBILITY OF SUCH DAMAGE.
30  *
31  * Effort sponsored in part by the Defense Advanced Research Projects
32  * Agency (DARPA) and Air Force Research Laboratory, Air Force
33  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
34  *
35  *	$OpenBSD: pf_ioctl.c,v 1.213 2009/02/15 21:46:12 mbalmer Exp $
36  */
37 
38 #include <sys/cdefs.h>
39 __FBSDID("$FreeBSD$");
40 
41 #include "opt_inet.h"
42 #include "opt_inet6.h"
43 #include "opt_bpf.h"
44 #include "opt_pf.h"
45 
46 #include <sys/param.h>
47 #include <sys/bus.h>
48 #include <sys/conf.h>
49 #include <sys/endian.h>
50 #include <sys/fcntl.h>
51 #include <sys/filio.h>
52 #include <sys/interrupt.h>
53 #include <sys/jail.h>
54 #include <sys/kernel.h>
55 #include <sys/kthread.h>
56 #include <sys/lock.h>
57 #include <sys/mbuf.h>
58 #include <sys/module.h>
59 #include <sys/proc.h>
60 #include <sys/rwlock.h>
61 #include <sys/smp.h>
62 #include <sys/socket.h>
63 #include <sys/sysctl.h>
64 #include <sys/md5.h>
65 #include <sys/ucred.h>
66 
67 #include <net/if.h>
68 #include <net/if_var.h>
69 #include <net/vnet.h>
70 #include <net/route.h>
71 #include <net/pfil.h>
72 #include <net/pfvar.h>
73 #include <net/if_pfsync.h>
74 #include <net/if_pflog.h>
75 
76 #include <netinet/in.h>
77 #include <netinet/ip.h>
78 #include <netinet/ip_var.h>
79 #include <netinet/ip_icmp.h>
80 
81 #ifdef INET6
82 #include <netinet/ip6.h>
83 #endif /* INET6 */
84 
85 #ifdef ALTQ
86 #include <altq/altq.h>
87 #endif
88 
89 static int		 pfattach(void);
90 static struct pf_pool	*pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t,
91 			    u_int8_t, u_int8_t, u_int8_t);
92 
93 static void		 pf_mv_pool(struct pf_palist *, struct pf_palist *);
94 static void		 pf_empty_pool(struct pf_palist *);
95 static int		 pfioctl(struct cdev *, u_long, caddr_t, int,
96 			    struct thread *);
97 #ifdef ALTQ
98 static int		 pf_begin_altq(u_int32_t *);
99 static int		 pf_rollback_altq(u_int32_t);
100 static int		 pf_commit_altq(u_int32_t);
101 static int		 pf_enable_altq(struct pf_altq *);
102 static int		 pf_disable_altq(struct pf_altq *);
103 static u_int32_t	 pf_qname2qid(char *);
104 static void		 pf_qid_unref(u_int32_t);
105 #endif /* ALTQ */
106 static int		 pf_begin_rules(u_int32_t *, int, const char *);
107 static int		 pf_rollback_rules(u_int32_t, int, char *);
108 static int		 pf_setup_pfsync_matching(struct pf_ruleset *);
109 static void		 pf_hash_rule(MD5_CTX *, struct pf_rule *);
110 static void		 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
111 static int		 pf_commit_rules(u_int32_t, int, char *);
112 static int		 pf_addr_setup(struct pf_ruleset *,
113 			    struct pf_addr_wrap *, sa_family_t);
114 static void		 pf_addr_copyout(struct pf_addr_wrap *);
115 
116 VNET_DEFINE(struct pf_rule,	pf_default_rule);
117 
118 #ifdef ALTQ
119 static VNET_DEFINE(int,		pf_altq_running);
120 #define	V_pf_altq_running	VNET(pf_altq_running)
121 #endif
122 
123 #define	TAGID_MAX	 50000
124 struct pf_tagname {
125 	TAILQ_ENTRY(pf_tagname)	entries;
126 	char			name[PF_TAG_NAME_SIZE];
127 	uint16_t		tag;
128 	int			ref;
129 };
130 
131 TAILQ_HEAD(pf_tags, pf_tagname);
132 #define	V_pf_tags		VNET(pf_tags)
133 VNET_DEFINE(struct pf_tags, pf_tags);
134 #define	V_pf_qids		VNET(pf_qids)
135 VNET_DEFINE(struct pf_tags, pf_qids);
136 static MALLOC_DEFINE(M_PFTAG, "pf_tag", "pf(4) tag names");
137 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
138 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
139 
140 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
141 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
142 #endif
143 
144 static u_int16_t	 tagname2tag(struct pf_tags *, char *);
145 static u_int16_t	 pf_tagname2tag(char *);
146 static void		 tag_unref(struct pf_tags *, u_int16_t);
147 
148 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
149 
150 struct cdev *pf_dev;
151 
152 /*
153  * XXX - These are new and need to be checked when moveing to a new version
154  */
155 static void		 pf_clear_states(void);
156 static int		 pf_clear_tables(void);
157 static void		 pf_clear_srcnodes(struct pf_src_node *);
158 static void		 pf_kill_srcnodes(struct pfioc_src_node_kill *);
159 static void		 pf_tbladdr_copyout(struct pf_addr_wrap *);
160 
161 /*
162  * Wrapper functions for pfil(9) hooks
163  */
164 #ifdef INET
165 static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp,
166     int dir, struct inpcb *inp);
167 static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp,
168     int dir, struct inpcb *inp);
169 #endif
170 #ifdef INET6
171 static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp,
172     int dir, struct inpcb *inp);
173 static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp,
174     int dir, struct inpcb *inp);
175 #endif
176 
177 static int		hook_pf(void);
178 static int		dehook_pf(void);
179 static int		shutdown_pf(void);
180 static int		pf_load(void);
181 static int		pf_unload(void);
182 
183 static struct cdevsw pf_cdevsw = {
184 	.d_ioctl =	pfioctl,
185 	.d_name =	PF_NAME,
186 	.d_version =	D_VERSION,
187 };
188 
189 static volatile VNET_DEFINE(int, pf_pfil_hooked);
190 #define V_pf_pfil_hooked	VNET(pf_pfil_hooked)
191 VNET_DEFINE(int,		pf_end_threads);
192 
193 struct rwlock			pf_rules_lock;
194 
195 /* pfsync */
196 pfsync_state_import_t 		*pfsync_state_import_ptr = NULL;
197 pfsync_insert_state_t		*pfsync_insert_state_ptr = NULL;
198 pfsync_update_state_t		*pfsync_update_state_ptr = NULL;
199 pfsync_delete_state_t		*pfsync_delete_state_ptr = NULL;
200 pfsync_clear_states_t		*pfsync_clear_states_ptr = NULL;
201 pfsync_defer_t			*pfsync_defer_ptr = NULL;
202 /* pflog */
203 pflog_packet_t			*pflog_packet_ptr = NULL;
204 
205 static int
206 pfattach(void)
207 {
208 	u_int32_t *my_timeout = V_pf_default_rule.timeout;
209 	int error;
210 
211 	if (IS_DEFAULT_VNET(curvnet))
212 		pf_mtag_initialize();
213 	pf_initialize();
214 	pfr_initialize();
215 	pfi_initialize();
216 	pf_normalize_init();
217 
218 	V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
219 	V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
220 
221 	RB_INIT(&V_pf_anchors);
222 	pf_init_ruleset(&pf_main_ruleset);
223 
224 	/* default rule should never be garbage collected */
225 	V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
226 #ifdef PF_DEFAULT_TO_DROP
227 	V_pf_default_rule.action = PF_DROP;
228 #else
229 	V_pf_default_rule.action = PF_PASS;
230 #endif
231 	V_pf_default_rule.nr = -1;
232 	V_pf_default_rule.rtableid = -1;
233 
234 	V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
235 	V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
236 	V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
237 
238 	/* initialize default timeouts */
239 	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
240 	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
241 	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
242 	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
243 	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
244 	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
245 	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
246 	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
247 	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
248 	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
249 	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
250 	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
251 	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
252 	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
253 	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
254 	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
255 	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
256 	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
257 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
258 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
259 
260 	bzero(&V_pf_status, sizeof(V_pf_status));
261 	V_pf_status.debug = PF_DEBUG_URGENT;
262 
263 	V_pf_pfil_hooked = 0;
264 
265 	/* XXX do our best to avoid a conflict */
266 	V_pf_status.hostid = arc4random();
267 
268 	if ((error = kproc_create(pf_purge_thread, curvnet, NULL, 0, 0,
269 	    "pf purge")) != 0)
270 		/* XXXGL: leaked all above. */
271 		return (error);
272 	if ((error = swi_add(NULL, "pf send", pf_intr, curvnet, SWI_NET,
273 	    INTR_MPSAFE, &V_pf_swi_cookie)) != 0)
274 		/* XXXGL: leaked all above. */
275 		return (error);
276 
277 	return (0);
278 }
279 
280 static struct pf_pool *
281 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
282     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
283     u_int8_t check_ticket)
284 {
285 	struct pf_ruleset	*ruleset;
286 	struct pf_rule		*rule;
287 	int			 rs_num;
288 
289 	ruleset = pf_find_ruleset(anchor);
290 	if (ruleset == NULL)
291 		return (NULL);
292 	rs_num = pf_get_ruleset_number(rule_action);
293 	if (rs_num >= PF_RULESET_MAX)
294 		return (NULL);
295 	if (active) {
296 		if (check_ticket && ticket !=
297 		    ruleset->rules[rs_num].active.ticket)
298 			return (NULL);
299 		if (r_last)
300 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
301 			    pf_rulequeue);
302 		else
303 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
304 	} else {
305 		if (check_ticket && ticket !=
306 		    ruleset->rules[rs_num].inactive.ticket)
307 			return (NULL);
308 		if (r_last)
309 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
310 			    pf_rulequeue);
311 		else
312 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
313 	}
314 	if (!r_last) {
315 		while ((rule != NULL) && (rule->nr != rule_number))
316 			rule = TAILQ_NEXT(rule, entries);
317 	}
318 	if (rule == NULL)
319 		return (NULL);
320 
321 	return (&rule->rpool);
322 }
323 
324 static void
325 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
326 {
327 	struct pf_pooladdr	*mv_pool_pa;
328 
329 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
330 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
331 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
332 	}
333 }
334 
335 static void
336 pf_empty_pool(struct pf_palist *poola)
337 {
338 	struct pf_pooladdr *pa;
339 
340 	while ((pa = TAILQ_FIRST(poola)) != NULL) {
341 		switch (pa->addr.type) {
342 		case PF_ADDR_DYNIFTL:
343 			pfi_dynaddr_remove(pa->addr.p.dyn);
344 			break;
345 		case PF_ADDR_TABLE:
346 			/* XXX: this could be unfinished pooladdr on pabuf */
347 			if (pa->addr.p.tbl != NULL)
348 				pfr_detach_table(pa->addr.p.tbl);
349 			break;
350 		}
351 		if (pa->kif)
352 			pfi_kif_unref(pa->kif);
353 		TAILQ_REMOVE(poola, pa, entries);
354 		free(pa, M_PFRULE);
355 	}
356 }
357 
358 static void
359 pf_unlink_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
360 {
361 
362 	PF_RULES_WASSERT();
363 
364 	TAILQ_REMOVE(rulequeue, rule, entries);
365 
366 	PF_UNLNKDRULES_LOCK();
367 	rule->rule_flag |= PFRULE_REFS;
368 	TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
369 	PF_UNLNKDRULES_UNLOCK();
370 }
371 
372 void
373 pf_free_rule(struct pf_rule *rule)
374 {
375 
376 	PF_RULES_WASSERT();
377 
378 	if (rule->tag)
379 		tag_unref(&V_pf_tags, rule->tag);
380 	if (rule->match_tag)
381 		tag_unref(&V_pf_tags, rule->match_tag);
382 #ifdef ALTQ
383 	if (rule->pqid != rule->qid)
384 		pf_qid_unref(rule->pqid);
385 	pf_qid_unref(rule->qid);
386 #endif
387 	switch (rule->src.addr.type) {
388 	case PF_ADDR_DYNIFTL:
389 		pfi_dynaddr_remove(rule->src.addr.p.dyn);
390 		break;
391 	case PF_ADDR_TABLE:
392 		pfr_detach_table(rule->src.addr.p.tbl);
393 		break;
394 	}
395 	switch (rule->dst.addr.type) {
396 	case PF_ADDR_DYNIFTL:
397 		pfi_dynaddr_remove(rule->dst.addr.p.dyn);
398 		break;
399 	case PF_ADDR_TABLE:
400 		pfr_detach_table(rule->dst.addr.p.tbl);
401 		break;
402 	}
403 	if (rule->overload_tbl)
404 		pfr_detach_table(rule->overload_tbl);
405 	if (rule->kif)
406 		pfi_kif_unref(rule->kif);
407 	pf_anchor_remove(rule);
408 	pf_empty_pool(&rule->rpool.list);
409 	counter_u64_free(rule->states_cur);
410 	counter_u64_free(rule->states_tot);
411 	counter_u64_free(rule->src_nodes);
412 	free(rule, M_PFRULE);
413 }
414 
415 static u_int16_t
416 tagname2tag(struct pf_tags *head, char *tagname)
417 {
418 	struct pf_tagname	*tag, *p = NULL;
419 	u_int16_t		 new_tagid = 1;
420 
421 	PF_RULES_WASSERT();
422 
423 	TAILQ_FOREACH(tag, head, entries)
424 		if (strcmp(tagname, tag->name) == 0) {
425 			tag->ref++;
426 			return (tag->tag);
427 		}
428 
429 	/*
430 	 * to avoid fragmentation, we do a linear search from the beginning
431 	 * and take the first free slot we find. if there is none or the list
432 	 * is empty, append a new entry at the end.
433 	 */
434 
435 	/* new entry */
436 	if (!TAILQ_EMPTY(head))
437 		for (p = TAILQ_FIRST(head); p != NULL &&
438 		    p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
439 			new_tagid = p->tag + 1;
440 
441 	if (new_tagid > TAGID_MAX)
442 		return (0);
443 
444 	/* allocate and fill new struct pf_tagname */
445 	tag = malloc(sizeof(*tag), M_PFTAG, M_NOWAIT|M_ZERO);
446 	if (tag == NULL)
447 		return (0);
448 	strlcpy(tag->name, tagname, sizeof(tag->name));
449 	tag->tag = new_tagid;
450 	tag->ref++;
451 
452 	if (p != NULL)	/* insert new entry before p */
453 		TAILQ_INSERT_BEFORE(p, tag, entries);
454 	else	/* either list empty or no free slot in between */
455 		TAILQ_INSERT_TAIL(head, tag, entries);
456 
457 	return (tag->tag);
458 }
459 
460 static void
461 tag_unref(struct pf_tags *head, u_int16_t tag)
462 {
463 	struct pf_tagname	*p, *next;
464 
465 	PF_RULES_WASSERT();
466 
467 	for (p = TAILQ_FIRST(head); p != NULL; p = next) {
468 		next = TAILQ_NEXT(p, entries);
469 		if (tag == p->tag) {
470 			if (--p->ref == 0) {
471 				TAILQ_REMOVE(head, p, entries);
472 				free(p, M_PFTAG);
473 			}
474 			break;
475 		}
476 	}
477 }
478 
479 static u_int16_t
480 pf_tagname2tag(char *tagname)
481 {
482 	return (tagname2tag(&V_pf_tags, tagname));
483 }
484 
485 #ifdef ALTQ
486 static u_int32_t
487 pf_qname2qid(char *qname)
488 {
489 	return ((u_int32_t)tagname2tag(&V_pf_qids, qname));
490 }
491 
492 static void
493 pf_qid_unref(u_int32_t qid)
494 {
495 	tag_unref(&V_pf_qids, (u_int16_t)qid);
496 }
497 
498 static int
499 pf_begin_altq(u_int32_t *ticket)
500 {
501 	struct pf_altq	*altq;
502 	int		 error = 0;
503 
504 	PF_RULES_WASSERT();
505 
506 	/* Purge the old altq list */
507 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
508 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
509 		if (altq->qname[0] == 0 &&
510 		    (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
511 			/* detach and destroy the discipline */
512 			error = altq_remove(altq);
513 		} else
514 			pf_qid_unref(altq->qid);
515 		free(altq, M_PFALTQ);
516 	}
517 	if (error)
518 		return (error);
519 	*ticket = ++V_ticket_altqs_inactive;
520 	V_altqs_inactive_open = 1;
521 	return (0);
522 }
523 
524 static int
525 pf_rollback_altq(u_int32_t ticket)
526 {
527 	struct pf_altq	*altq;
528 	int		 error = 0;
529 
530 	PF_RULES_WASSERT();
531 
532 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
533 		return (0);
534 	/* Purge the old altq list */
535 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
536 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
537 		if (altq->qname[0] == 0 &&
538 		   (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
539 			/* detach and destroy the discipline */
540 			error = altq_remove(altq);
541 		} else
542 			pf_qid_unref(altq->qid);
543 		free(altq, M_PFALTQ);
544 	}
545 	V_altqs_inactive_open = 0;
546 	return (error);
547 }
548 
549 static int
550 pf_commit_altq(u_int32_t ticket)
551 {
552 	struct pf_altqqueue	*old_altqs;
553 	struct pf_altq		*altq;
554 	int			 err, error = 0;
555 
556 	PF_RULES_WASSERT();
557 
558 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
559 		return (EBUSY);
560 
561 	/* swap altqs, keep the old. */
562 	old_altqs = V_pf_altqs_active;
563 	V_pf_altqs_active = V_pf_altqs_inactive;
564 	V_pf_altqs_inactive = old_altqs;
565 	V_ticket_altqs_active = V_ticket_altqs_inactive;
566 
567 	/* Attach new disciplines */
568 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
569 	if (altq->qname[0] == 0 &&
570 	   (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
571 			/* attach the discipline */
572 			error = altq_pfattach(altq);
573 			if (error == 0 && V_pf_altq_running)
574 				error = pf_enable_altq(altq);
575 			if (error != 0)
576 				return (error);
577 		}
578 	}
579 
580 	/* Purge the old altq list */
581 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
582 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
583 		if (altq->qname[0] == 0 &&
584 		    (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
585 			/* detach and destroy the discipline */
586 			if (V_pf_altq_running)
587 				error = pf_disable_altq(altq);
588 			err = altq_pfdetach(altq);
589 			if (err != 0 && error == 0)
590 				error = err;
591 			err = altq_remove(altq);
592 			if (err != 0 && error == 0)
593 				error = err;
594 		} else
595 			pf_qid_unref(altq->qid);
596 		free(altq, M_PFALTQ);
597 	}
598 
599 	V_altqs_inactive_open = 0;
600 	return (error);
601 }
602 
603 static int
604 pf_enable_altq(struct pf_altq *altq)
605 {
606 	struct ifnet		*ifp;
607 	struct tb_profile	 tb;
608 	int			 error = 0;
609 
610 	if ((ifp = ifunit(altq->ifname)) == NULL)
611 		return (EINVAL);
612 
613 	if (ifp->if_snd.altq_type != ALTQT_NONE)
614 		error = altq_enable(&ifp->if_snd);
615 
616 	/* set tokenbucket regulator */
617 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
618 		tb.rate = altq->ifbandwidth;
619 		tb.depth = altq->tbrsize;
620 		error = tbr_set(&ifp->if_snd, &tb);
621 	}
622 
623 	return (error);
624 }
625 
626 static int
627 pf_disable_altq(struct pf_altq *altq)
628 {
629 	struct ifnet		*ifp;
630 	struct tb_profile	 tb;
631 	int			 error;
632 
633 	if ((ifp = ifunit(altq->ifname)) == NULL)
634 		return (EINVAL);
635 
636 	/*
637 	 * when the discipline is no longer referenced, it was overridden
638 	 * by a new one.  if so, just return.
639 	 */
640 	if (altq->altq_disc != ifp->if_snd.altq_disc)
641 		return (0);
642 
643 	error = altq_disable(&ifp->if_snd);
644 
645 	if (error == 0) {
646 		/* clear tokenbucket regulator */
647 		tb.rate = 0;
648 		error = tbr_set(&ifp->if_snd, &tb);
649 	}
650 
651 	return (error);
652 }
653 
654 void
655 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
656 {
657 	struct ifnet	*ifp1;
658 	struct pf_altq	*a1, *a2, *a3;
659 	u_int32_t	 ticket;
660 	int		 error = 0;
661 
662 	/* Interrupt userland queue modifications */
663 	if (V_altqs_inactive_open)
664 		pf_rollback_altq(V_ticket_altqs_inactive);
665 
666 	/* Start new altq ruleset */
667 	if (pf_begin_altq(&ticket))
668 		return;
669 
670 	/* Copy the current active set */
671 	TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
672 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
673 		if (a2 == NULL) {
674 			error = ENOMEM;
675 			break;
676 		}
677 		bcopy(a1, a2, sizeof(struct pf_altq));
678 
679 		if (a2->qname[0] != 0) {
680 			if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
681 				error = EBUSY;
682 				free(a2, M_PFALTQ);
683 				break;
684 			}
685 			a2->altq_disc = NULL;
686 			TAILQ_FOREACH(a3, V_pf_altqs_inactive, entries) {
687 				if (strncmp(a3->ifname, a2->ifname,
688 				    IFNAMSIZ) == 0 && a3->qname[0] == 0) {
689 					a2->altq_disc = a3->altq_disc;
690 					break;
691 				}
692 			}
693 		}
694 		/* Deactivate the interface in question */
695 		a2->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
696 		if ((ifp1 = ifunit(a2->ifname)) == NULL ||
697 		    (remove && ifp1 == ifp)) {
698 			a2->local_flags |= PFALTQ_FLAG_IF_REMOVED;
699 		} else {
700 			error = altq_add(a2);
701 
702 			if (ticket != V_ticket_altqs_inactive)
703 				error = EBUSY;
704 
705 			if (error) {
706 				free(a2, M_PFALTQ);
707 				break;
708 			}
709 		}
710 
711 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
712 	}
713 
714 	if (error != 0)
715 		pf_rollback_altq(ticket);
716 	else
717 		pf_commit_altq(ticket);
718 }
719 #endif /* ALTQ */
720 
721 static int
722 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
723 {
724 	struct pf_ruleset	*rs;
725 	struct pf_rule		*rule;
726 
727 	PF_RULES_WASSERT();
728 
729 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
730 		return (EINVAL);
731 	rs = pf_find_or_create_ruleset(anchor);
732 	if (rs == NULL)
733 		return (EINVAL);
734 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
735 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
736 		rs->rules[rs_num].inactive.rcount--;
737 	}
738 	*ticket = ++rs->rules[rs_num].inactive.ticket;
739 	rs->rules[rs_num].inactive.open = 1;
740 	return (0);
741 }
742 
743 static int
744 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
745 {
746 	struct pf_ruleset	*rs;
747 	struct pf_rule		*rule;
748 
749 	PF_RULES_WASSERT();
750 
751 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
752 		return (EINVAL);
753 	rs = pf_find_ruleset(anchor);
754 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
755 	    rs->rules[rs_num].inactive.ticket != ticket)
756 		return (0);
757 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
758 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
759 		rs->rules[rs_num].inactive.rcount--;
760 	}
761 	rs->rules[rs_num].inactive.open = 0;
762 	return (0);
763 }
764 
765 #define PF_MD5_UPD(st, elm)						\
766 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
767 
768 #define PF_MD5_UPD_STR(st, elm)						\
769 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
770 
771 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
772 		(stor) = htonl((st)->elm);				\
773 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
774 } while (0)
775 
776 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
777 		(stor) = htons((st)->elm);				\
778 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
779 } while (0)
780 
781 static void
782 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
783 {
784 	PF_MD5_UPD(pfr, addr.type);
785 	switch (pfr->addr.type) {
786 		case PF_ADDR_DYNIFTL:
787 			PF_MD5_UPD(pfr, addr.v.ifname);
788 			PF_MD5_UPD(pfr, addr.iflags);
789 			break;
790 		case PF_ADDR_TABLE:
791 			PF_MD5_UPD(pfr, addr.v.tblname);
792 			break;
793 		case PF_ADDR_ADDRMASK:
794 			/* XXX ignore af? */
795 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
796 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
797 			break;
798 	}
799 
800 	PF_MD5_UPD(pfr, port[0]);
801 	PF_MD5_UPD(pfr, port[1]);
802 	PF_MD5_UPD(pfr, neg);
803 	PF_MD5_UPD(pfr, port_op);
804 }
805 
806 static void
807 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
808 {
809 	u_int16_t x;
810 	u_int32_t y;
811 
812 	pf_hash_rule_addr(ctx, &rule->src);
813 	pf_hash_rule_addr(ctx, &rule->dst);
814 	PF_MD5_UPD_STR(rule, label);
815 	PF_MD5_UPD_STR(rule, ifname);
816 	PF_MD5_UPD_STR(rule, match_tagname);
817 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
818 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
819 	PF_MD5_UPD_HTONL(rule, prob, y);
820 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
821 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
822 	PF_MD5_UPD(rule, uid.op);
823 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
824 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
825 	PF_MD5_UPD(rule, gid.op);
826 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
827 	PF_MD5_UPD(rule, action);
828 	PF_MD5_UPD(rule, direction);
829 	PF_MD5_UPD(rule, af);
830 	PF_MD5_UPD(rule, quick);
831 	PF_MD5_UPD(rule, ifnot);
832 	PF_MD5_UPD(rule, match_tag_not);
833 	PF_MD5_UPD(rule, natpass);
834 	PF_MD5_UPD(rule, keep_state);
835 	PF_MD5_UPD(rule, proto);
836 	PF_MD5_UPD(rule, type);
837 	PF_MD5_UPD(rule, code);
838 	PF_MD5_UPD(rule, flags);
839 	PF_MD5_UPD(rule, flagset);
840 	PF_MD5_UPD(rule, allow_opts);
841 	PF_MD5_UPD(rule, rt);
842 	PF_MD5_UPD(rule, tos);
843 }
844 
845 static int
846 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
847 {
848 	struct pf_ruleset	*rs;
849 	struct pf_rule		*rule, **old_array;
850 	struct pf_rulequeue	*old_rules;
851 	int			 error;
852 	u_int32_t		 old_rcount;
853 
854 	PF_RULES_WASSERT();
855 
856 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
857 		return (EINVAL);
858 	rs = pf_find_ruleset(anchor);
859 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
860 	    ticket != rs->rules[rs_num].inactive.ticket)
861 		return (EBUSY);
862 
863 	/* Calculate checksum for the main ruleset */
864 	if (rs == &pf_main_ruleset) {
865 		error = pf_setup_pfsync_matching(rs);
866 		if (error != 0)
867 			return (error);
868 	}
869 
870 	/* Swap rules, keep the old. */
871 	old_rules = rs->rules[rs_num].active.ptr;
872 	old_rcount = rs->rules[rs_num].active.rcount;
873 	old_array = rs->rules[rs_num].active.ptr_array;
874 
875 	rs->rules[rs_num].active.ptr =
876 	    rs->rules[rs_num].inactive.ptr;
877 	rs->rules[rs_num].active.ptr_array =
878 	    rs->rules[rs_num].inactive.ptr_array;
879 	rs->rules[rs_num].active.rcount =
880 	    rs->rules[rs_num].inactive.rcount;
881 	rs->rules[rs_num].inactive.ptr = old_rules;
882 	rs->rules[rs_num].inactive.ptr_array = old_array;
883 	rs->rules[rs_num].inactive.rcount = old_rcount;
884 
885 	rs->rules[rs_num].active.ticket =
886 	    rs->rules[rs_num].inactive.ticket;
887 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
888 
889 
890 	/* Purge the old rule list. */
891 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
892 		pf_unlink_rule(old_rules, rule);
893 	if (rs->rules[rs_num].inactive.ptr_array)
894 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
895 	rs->rules[rs_num].inactive.ptr_array = NULL;
896 	rs->rules[rs_num].inactive.rcount = 0;
897 	rs->rules[rs_num].inactive.open = 0;
898 	pf_remove_if_empty_ruleset(rs);
899 
900 	return (0);
901 }
902 
903 static int
904 pf_setup_pfsync_matching(struct pf_ruleset *rs)
905 {
906 	MD5_CTX			 ctx;
907 	struct pf_rule		*rule;
908 	int			 rs_cnt;
909 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
910 
911 	MD5Init(&ctx);
912 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
913 		/* XXX PF_RULESET_SCRUB as well? */
914 		if (rs_cnt == PF_RULESET_SCRUB)
915 			continue;
916 
917 		if (rs->rules[rs_cnt].inactive.ptr_array)
918 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
919 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
920 
921 		if (rs->rules[rs_cnt].inactive.rcount) {
922 			rs->rules[rs_cnt].inactive.ptr_array =
923 			    malloc(sizeof(caddr_t) *
924 			    rs->rules[rs_cnt].inactive.rcount,
925 			    M_TEMP, M_NOWAIT);
926 
927 			if (!rs->rules[rs_cnt].inactive.ptr_array)
928 				return (ENOMEM);
929 		}
930 
931 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
932 		    entries) {
933 			pf_hash_rule(&ctx, rule);
934 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
935 		}
936 	}
937 
938 	MD5Final(digest, &ctx);
939 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
940 	return (0);
941 }
942 
943 static int
944 pf_addr_setup(struct pf_ruleset *ruleset, struct pf_addr_wrap *addr,
945     sa_family_t af)
946 {
947 	int error = 0;
948 
949 	switch (addr->type) {
950 	case PF_ADDR_TABLE:
951 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
952 		if (addr->p.tbl == NULL)
953 			error = ENOMEM;
954 		break;
955 	case PF_ADDR_DYNIFTL:
956 		error = pfi_dynaddr_setup(addr, af);
957 		break;
958 	}
959 
960 	return (error);
961 }
962 
963 static void
964 pf_addr_copyout(struct pf_addr_wrap *addr)
965 {
966 
967 	switch (addr->type) {
968 	case PF_ADDR_DYNIFTL:
969 		pfi_dynaddr_copyout(addr);
970 		break;
971 	case PF_ADDR_TABLE:
972 		pf_tbladdr_copyout(addr);
973 		break;
974 	}
975 }
976 
977 static int
978 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
979 {
980 	int			 error = 0;
981 
982 	/* XXX keep in sync with switch() below */
983 	if (securelevel_gt(td->td_ucred, 2))
984 		switch (cmd) {
985 		case DIOCGETRULES:
986 		case DIOCGETRULE:
987 		case DIOCGETADDRS:
988 		case DIOCGETADDR:
989 		case DIOCGETSTATE:
990 		case DIOCSETSTATUSIF:
991 		case DIOCGETSTATUS:
992 		case DIOCCLRSTATUS:
993 		case DIOCNATLOOK:
994 		case DIOCSETDEBUG:
995 		case DIOCGETSTATES:
996 		case DIOCGETTIMEOUT:
997 		case DIOCCLRRULECTRS:
998 		case DIOCGETLIMIT:
999 		case DIOCGETALTQS:
1000 		case DIOCGETALTQ:
1001 		case DIOCGETQSTATS:
1002 		case DIOCGETRULESETS:
1003 		case DIOCGETRULESET:
1004 		case DIOCRGETTABLES:
1005 		case DIOCRGETTSTATS:
1006 		case DIOCRCLRTSTATS:
1007 		case DIOCRCLRADDRS:
1008 		case DIOCRADDADDRS:
1009 		case DIOCRDELADDRS:
1010 		case DIOCRSETADDRS:
1011 		case DIOCRGETADDRS:
1012 		case DIOCRGETASTATS:
1013 		case DIOCRCLRASTATS:
1014 		case DIOCRTSTADDRS:
1015 		case DIOCOSFPGET:
1016 		case DIOCGETSRCNODES:
1017 		case DIOCCLRSRCNODES:
1018 		case DIOCIGETIFACES:
1019 		case DIOCGIFSPEED:
1020 		case DIOCSETIFFLAG:
1021 		case DIOCCLRIFFLAG:
1022 			break;
1023 		case DIOCRCLRTABLES:
1024 		case DIOCRADDTABLES:
1025 		case DIOCRDELTABLES:
1026 		case DIOCRSETTFLAGS:
1027 			if (((struct pfioc_table *)addr)->pfrio_flags &
1028 			    PFR_FLAG_DUMMY)
1029 				break; /* dummy operation ok */
1030 			return (EPERM);
1031 		default:
1032 			return (EPERM);
1033 		}
1034 
1035 	if (!(flags & FWRITE))
1036 		switch (cmd) {
1037 		case DIOCGETRULES:
1038 		case DIOCGETADDRS:
1039 		case DIOCGETADDR:
1040 		case DIOCGETSTATE:
1041 		case DIOCGETSTATUS:
1042 		case DIOCGETSTATES:
1043 		case DIOCGETTIMEOUT:
1044 		case DIOCGETLIMIT:
1045 		case DIOCGETALTQS:
1046 		case DIOCGETALTQ:
1047 		case DIOCGETQSTATS:
1048 		case DIOCGETRULESETS:
1049 		case DIOCGETRULESET:
1050 		case DIOCNATLOOK:
1051 		case DIOCRGETTABLES:
1052 		case DIOCRGETTSTATS:
1053 		case DIOCRGETADDRS:
1054 		case DIOCRGETASTATS:
1055 		case DIOCRTSTADDRS:
1056 		case DIOCOSFPGET:
1057 		case DIOCGETSRCNODES:
1058 		case DIOCIGETIFACES:
1059 		case DIOCGIFSPEED:
1060 			break;
1061 		case DIOCRCLRTABLES:
1062 		case DIOCRADDTABLES:
1063 		case DIOCRDELTABLES:
1064 		case DIOCRCLRTSTATS:
1065 		case DIOCRCLRADDRS:
1066 		case DIOCRADDADDRS:
1067 		case DIOCRDELADDRS:
1068 		case DIOCRSETADDRS:
1069 		case DIOCRSETTFLAGS:
1070 			if (((struct pfioc_table *)addr)->pfrio_flags &
1071 			    PFR_FLAG_DUMMY) {
1072 				flags |= FWRITE; /* need write lock for dummy */
1073 				break; /* dummy operation ok */
1074 			}
1075 			return (EACCES);
1076 		case DIOCGETRULE:
1077 			if (((struct pfioc_rule *)addr)->action ==
1078 			    PF_GET_CLR_CNTR)
1079 				return (EACCES);
1080 			break;
1081 		default:
1082 			return (EACCES);
1083 		}
1084 
1085 	CURVNET_SET(TD_TO_VNET(td));
1086 
1087 	switch (cmd) {
1088 	case DIOCSTART:
1089 		PF_RULES_WLOCK();
1090 		if (V_pf_status.running)
1091 			error = EEXIST;
1092 		else {
1093 			int cpu;
1094 
1095 			PF_RULES_WUNLOCK();
1096 			error = hook_pf();
1097 			if (error) {
1098 				DPFPRINTF(PF_DEBUG_MISC,
1099 				    ("pf: pfil registration failed\n"));
1100 				break;
1101 			}
1102 			PF_RULES_WLOCK();
1103 			V_pf_status.running = 1;
1104 			V_pf_status.since = time_second;
1105 
1106 			CPU_FOREACH(cpu)
1107 				V_pf_stateid[cpu] = time_second;
1108 
1109 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1110 		}
1111 		PF_RULES_WUNLOCK();
1112 		break;
1113 
1114 	case DIOCSTOP:
1115 		PF_RULES_WLOCK();
1116 		if (!V_pf_status.running)
1117 			error = ENOENT;
1118 		else {
1119 			V_pf_status.running = 0;
1120 			PF_RULES_WUNLOCK();
1121 			error = dehook_pf();
1122 			if (error) {
1123 				V_pf_status.running = 1;
1124 				DPFPRINTF(PF_DEBUG_MISC,
1125 				    ("pf: pfil unregistration failed\n"));
1126 			}
1127 			PF_RULES_WLOCK();
1128 			V_pf_status.since = time_second;
1129 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1130 		}
1131 		PF_RULES_WUNLOCK();
1132 		break;
1133 
1134 	case DIOCADDRULE: {
1135 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1136 		struct pf_ruleset	*ruleset;
1137 		struct pf_rule		*rule, *tail;
1138 		struct pf_pooladdr	*pa;
1139 		struct pfi_kif		*kif = NULL;
1140 		int			 rs_num;
1141 
1142 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1143 			error = EINVAL;
1144 			break;
1145 		}
1146 #ifndef INET
1147 		if (pr->rule.af == AF_INET) {
1148 			error = EAFNOSUPPORT;
1149 			break;
1150 		}
1151 #endif /* INET */
1152 #ifndef INET6
1153 		if (pr->rule.af == AF_INET6) {
1154 			error = EAFNOSUPPORT;
1155 			break;
1156 		}
1157 #endif /* INET6 */
1158 
1159 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
1160 		bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1161 		if (rule->ifname[0])
1162 			kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1163 		rule->states_cur = counter_u64_alloc(M_WAITOK);
1164 		rule->states_tot = counter_u64_alloc(M_WAITOK);
1165 		rule->src_nodes = counter_u64_alloc(M_WAITOK);
1166 		rule->cuid = td->td_ucred->cr_ruid;
1167 		rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1168 		TAILQ_INIT(&rule->rpool.list);
1169 
1170 #define	ERROUT(x)	{ error = (x); goto DIOCADDRULE_error; }
1171 
1172 		PF_RULES_WLOCK();
1173 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1174 		ruleset = pf_find_ruleset(pr->anchor);
1175 		if (ruleset == NULL)
1176 			ERROUT(EINVAL);
1177 		rs_num = pf_get_ruleset_number(pr->rule.action);
1178 		if (rs_num >= PF_RULESET_MAX)
1179 			ERROUT(EINVAL);
1180 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1181 			DPFPRINTF(PF_DEBUG_MISC,
1182 			    ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
1183 			    ruleset->rules[rs_num].inactive.ticket));
1184 			ERROUT(EBUSY);
1185 		}
1186 		if (pr->pool_ticket != V_ticket_pabuf) {
1187 			DPFPRINTF(PF_DEBUG_MISC,
1188 			    ("pool_ticket: %d != %d\n", pr->pool_ticket,
1189 			    V_ticket_pabuf));
1190 			ERROUT(EBUSY);
1191 		}
1192 
1193 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1194 		    pf_rulequeue);
1195 		if (tail)
1196 			rule->nr = tail->nr + 1;
1197 		else
1198 			rule->nr = 0;
1199 		if (rule->ifname[0]) {
1200 			rule->kif = pfi_kif_attach(kif, rule->ifname);
1201 			pfi_kif_ref(rule->kif);
1202 		} else
1203 			rule->kif = NULL;
1204 
1205 		if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
1206 			error = EBUSY;
1207 
1208 #ifdef ALTQ
1209 		/* set queue IDs */
1210 		if (rule->qname[0] != 0) {
1211 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1212 				error = EBUSY;
1213 			else if (rule->pqname[0] != 0) {
1214 				if ((rule->pqid =
1215 				    pf_qname2qid(rule->pqname)) == 0)
1216 					error = EBUSY;
1217 			} else
1218 				rule->pqid = rule->qid;
1219 		}
1220 #endif
1221 		if (rule->tagname[0])
1222 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1223 				error = EBUSY;
1224 		if (rule->match_tagname[0])
1225 			if ((rule->match_tag =
1226 			    pf_tagname2tag(rule->match_tagname)) == 0)
1227 				error = EBUSY;
1228 		if (rule->rt && !rule->direction)
1229 			error = EINVAL;
1230 		if (!rule->log)
1231 			rule->logif = 0;
1232 		if (rule->logif >= PFLOGIFS_MAX)
1233 			error = EINVAL;
1234 		if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1235 			error = ENOMEM;
1236 		if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1237 			error = ENOMEM;
1238 		if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1239 			error = EINVAL;
1240 		TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1241 			if (pa->addr.type == PF_ADDR_TABLE) {
1242 				pa->addr.p.tbl = pfr_attach_table(ruleset,
1243 				    pa->addr.v.tblname);
1244 				if (pa->addr.p.tbl == NULL)
1245 					error = ENOMEM;
1246 			}
1247 
1248 		if (rule->overload_tblname[0]) {
1249 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1250 			    rule->overload_tblname)) == NULL)
1251 				error = EINVAL;
1252 			else
1253 				rule->overload_tbl->pfrkt_flags |=
1254 				    PFR_TFLAG_ACTIVE;
1255 		}
1256 
1257 		pf_mv_pool(&V_pf_pabuf, &rule->rpool.list);
1258 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1259 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1260 		    (rule->rt > PF_FASTROUTE)) &&
1261 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1262 			error = EINVAL;
1263 
1264 		if (error) {
1265 			pf_free_rule(rule);
1266 			PF_RULES_WUNLOCK();
1267 			break;
1268 		}
1269 
1270 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1271 		rule->evaluations = rule->packets[0] = rule->packets[1] =
1272 		    rule->bytes[0] = rule->bytes[1] = 0;
1273 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1274 		    rule, entries);
1275 		ruleset->rules[rs_num].inactive.rcount++;
1276 		PF_RULES_WUNLOCK();
1277 		break;
1278 
1279 #undef ERROUT
1280 DIOCADDRULE_error:
1281 		PF_RULES_WUNLOCK();
1282 		counter_u64_free(rule->states_cur);
1283 		counter_u64_free(rule->states_tot);
1284 		counter_u64_free(rule->src_nodes);
1285 		free(rule, M_PFRULE);
1286 		if (kif)
1287 			free(kif, PFI_MTYPE);
1288 		break;
1289 	}
1290 
1291 	case DIOCGETRULES: {
1292 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1293 		struct pf_ruleset	*ruleset;
1294 		struct pf_rule		*tail;
1295 		int			 rs_num;
1296 
1297 		PF_RULES_WLOCK();
1298 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1299 		ruleset = pf_find_ruleset(pr->anchor);
1300 		if (ruleset == NULL) {
1301 			PF_RULES_WUNLOCK();
1302 			error = EINVAL;
1303 			break;
1304 		}
1305 		rs_num = pf_get_ruleset_number(pr->rule.action);
1306 		if (rs_num >= PF_RULESET_MAX) {
1307 			PF_RULES_WUNLOCK();
1308 			error = EINVAL;
1309 			break;
1310 		}
1311 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1312 		    pf_rulequeue);
1313 		if (tail)
1314 			pr->nr = tail->nr + 1;
1315 		else
1316 			pr->nr = 0;
1317 		pr->ticket = ruleset->rules[rs_num].active.ticket;
1318 		PF_RULES_WUNLOCK();
1319 		break;
1320 	}
1321 
1322 	case DIOCGETRULE: {
1323 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1324 		struct pf_ruleset	*ruleset;
1325 		struct pf_rule		*rule;
1326 		int			 rs_num, i;
1327 
1328 		PF_RULES_WLOCK();
1329 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1330 		ruleset = pf_find_ruleset(pr->anchor);
1331 		if (ruleset == NULL) {
1332 			PF_RULES_WUNLOCK();
1333 			error = EINVAL;
1334 			break;
1335 		}
1336 		rs_num = pf_get_ruleset_number(pr->rule.action);
1337 		if (rs_num >= PF_RULESET_MAX) {
1338 			PF_RULES_WUNLOCK();
1339 			error = EINVAL;
1340 			break;
1341 		}
1342 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1343 			PF_RULES_WUNLOCK();
1344 			error = EBUSY;
1345 			break;
1346 		}
1347 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1348 		while ((rule != NULL) && (rule->nr != pr->nr))
1349 			rule = TAILQ_NEXT(rule, entries);
1350 		if (rule == NULL) {
1351 			PF_RULES_WUNLOCK();
1352 			error = EBUSY;
1353 			break;
1354 		}
1355 		bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1356 		pr->rule.u_states_cur = counter_u64_fetch(rule->states_cur);
1357 		pr->rule.u_states_tot = counter_u64_fetch(rule->states_tot);
1358 		pr->rule.u_src_nodes = counter_u64_fetch(rule->src_nodes);
1359 		if (pf_anchor_copyout(ruleset, rule, pr)) {
1360 			PF_RULES_WUNLOCK();
1361 			error = EBUSY;
1362 			break;
1363 		}
1364 		pf_addr_copyout(&pr->rule.src.addr);
1365 		pf_addr_copyout(&pr->rule.dst.addr);
1366 		for (i = 0; i < PF_SKIP_COUNT; ++i)
1367 			if (rule->skip[i].ptr == NULL)
1368 				pr->rule.skip[i].nr = -1;
1369 			else
1370 				pr->rule.skip[i].nr =
1371 				    rule->skip[i].ptr->nr;
1372 
1373 		if (pr->action == PF_GET_CLR_CNTR) {
1374 			rule->evaluations = 0;
1375 			rule->packets[0] = rule->packets[1] = 0;
1376 			rule->bytes[0] = rule->bytes[1] = 0;
1377 			counter_u64_zero(rule->states_tot);
1378 		}
1379 		PF_RULES_WUNLOCK();
1380 		break;
1381 	}
1382 
1383 	case DIOCCHANGERULE: {
1384 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1385 		struct pf_ruleset	*ruleset;
1386 		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1387 		struct pfi_kif		*kif = NULL;
1388 		struct pf_pooladdr	*pa;
1389 		u_int32_t		 nr = 0;
1390 		int			 rs_num;
1391 
1392 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1393 		    pcr->action > PF_CHANGE_GET_TICKET) {
1394 			error = EINVAL;
1395 			break;
1396 		}
1397 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1398 			error = EINVAL;
1399 			break;
1400 		}
1401 
1402 		if (pcr->action != PF_CHANGE_REMOVE) {
1403 #ifndef INET
1404 			if (pcr->rule.af == AF_INET) {
1405 				error = EAFNOSUPPORT;
1406 				break;
1407 			}
1408 #endif /* INET */
1409 #ifndef INET6
1410 			if (pcr->rule.af == AF_INET6) {
1411 				error = EAFNOSUPPORT;
1412 				break;
1413 			}
1414 #endif /* INET6 */
1415 			newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
1416 			bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1417 			if (newrule->ifname[0])
1418 				kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1419 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
1420 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
1421 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
1422 			newrule->cuid = td->td_ucred->cr_ruid;
1423 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1424 			TAILQ_INIT(&newrule->rpool.list);
1425 		}
1426 
1427 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGERULE_error; }
1428 
1429 		PF_RULES_WLOCK();
1430 		if (!(pcr->action == PF_CHANGE_REMOVE ||
1431 		    pcr->action == PF_CHANGE_GET_TICKET) &&
1432 		    pcr->pool_ticket != V_ticket_pabuf)
1433 			ERROUT(EBUSY);
1434 
1435 		ruleset = pf_find_ruleset(pcr->anchor);
1436 		if (ruleset == NULL)
1437 			ERROUT(EINVAL);
1438 
1439 		rs_num = pf_get_ruleset_number(pcr->rule.action);
1440 		if (rs_num >= PF_RULESET_MAX)
1441 			ERROUT(EINVAL);
1442 
1443 		if (pcr->action == PF_CHANGE_GET_TICKET) {
1444 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1445 			ERROUT(0);
1446 		} else if (pcr->ticket !=
1447 			    ruleset->rules[rs_num].active.ticket)
1448 				ERROUT(EINVAL);
1449 
1450 		if (pcr->action != PF_CHANGE_REMOVE) {
1451 			if (newrule->ifname[0]) {
1452 				newrule->kif = pfi_kif_attach(kif,
1453 				    newrule->ifname);
1454 				pfi_kif_ref(newrule->kif);
1455 			} else
1456 				newrule->kif = NULL;
1457 
1458 			if (newrule->rtableid > 0 &&
1459 			    newrule->rtableid >= rt_numfibs)
1460 				error = EBUSY;
1461 
1462 #ifdef ALTQ
1463 			/* set queue IDs */
1464 			if (newrule->qname[0] != 0) {
1465 				if ((newrule->qid =
1466 				    pf_qname2qid(newrule->qname)) == 0)
1467 					error = EBUSY;
1468 				else if (newrule->pqname[0] != 0) {
1469 					if ((newrule->pqid =
1470 					    pf_qname2qid(newrule->pqname)) == 0)
1471 						error = EBUSY;
1472 				} else
1473 					newrule->pqid = newrule->qid;
1474 			}
1475 #endif /* ALTQ */
1476 			if (newrule->tagname[0])
1477 				if ((newrule->tag =
1478 				    pf_tagname2tag(newrule->tagname)) == 0)
1479 					error = EBUSY;
1480 			if (newrule->match_tagname[0])
1481 				if ((newrule->match_tag = pf_tagname2tag(
1482 				    newrule->match_tagname)) == 0)
1483 					error = EBUSY;
1484 			if (newrule->rt && !newrule->direction)
1485 				error = EINVAL;
1486 			if (!newrule->log)
1487 				newrule->logif = 0;
1488 			if (newrule->logif >= PFLOGIFS_MAX)
1489 				error = EINVAL;
1490 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
1491 				error = ENOMEM;
1492 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
1493 				error = ENOMEM;
1494 			if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1495 				error = EINVAL;
1496 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1497 				if (pa->addr.type == PF_ADDR_TABLE) {
1498 					pa->addr.p.tbl =
1499 					    pfr_attach_table(ruleset,
1500 					    pa->addr.v.tblname);
1501 					if (pa->addr.p.tbl == NULL)
1502 						error = ENOMEM;
1503 				}
1504 
1505 			if (newrule->overload_tblname[0]) {
1506 				if ((newrule->overload_tbl = pfr_attach_table(
1507 				    ruleset, newrule->overload_tblname)) ==
1508 				    NULL)
1509 					error = EINVAL;
1510 				else
1511 					newrule->overload_tbl->pfrkt_flags |=
1512 					    PFR_TFLAG_ACTIVE;
1513 			}
1514 
1515 			pf_mv_pool(&V_pf_pabuf, &newrule->rpool.list);
1516 			if (((((newrule->action == PF_NAT) ||
1517 			    (newrule->action == PF_RDR) ||
1518 			    (newrule->action == PF_BINAT) ||
1519 			    (newrule->rt > PF_FASTROUTE)) &&
1520 			    !newrule->anchor)) &&
1521 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1522 				error = EINVAL;
1523 
1524 			if (error) {
1525 				pf_free_rule(newrule);
1526 				PF_RULES_WUNLOCK();
1527 				break;
1528 			}
1529 
1530 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1531 			newrule->evaluations = 0;
1532 			newrule->packets[0] = newrule->packets[1] = 0;
1533 			newrule->bytes[0] = newrule->bytes[1] = 0;
1534 		}
1535 		pf_empty_pool(&V_pf_pabuf);
1536 
1537 		if (pcr->action == PF_CHANGE_ADD_HEAD)
1538 			oldrule = TAILQ_FIRST(
1539 			    ruleset->rules[rs_num].active.ptr);
1540 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1541 			oldrule = TAILQ_LAST(
1542 			    ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1543 		else {
1544 			oldrule = TAILQ_FIRST(
1545 			    ruleset->rules[rs_num].active.ptr);
1546 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1547 				oldrule = TAILQ_NEXT(oldrule, entries);
1548 			if (oldrule == NULL) {
1549 				if (newrule != NULL)
1550 					pf_free_rule(newrule);
1551 				PF_RULES_WUNLOCK();
1552 				error = EINVAL;
1553 				break;
1554 			}
1555 		}
1556 
1557 		if (pcr->action == PF_CHANGE_REMOVE) {
1558 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
1559 			    oldrule);
1560 			ruleset->rules[rs_num].active.rcount--;
1561 		} else {
1562 			if (oldrule == NULL)
1563 				TAILQ_INSERT_TAIL(
1564 				    ruleset->rules[rs_num].active.ptr,
1565 				    newrule, entries);
1566 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1567 			    pcr->action == PF_CHANGE_ADD_BEFORE)
1568 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1569 			else
1570 				TAILQ_INSERT_AFTER(
1571 				    ruleset->rules[rs_num].active.ptr,
1572 				    oldrule, newrule, entries);
1573 			ruleset->rules[rs_num].active.rcount++;
1574 		}
1575 
1576 		nr = 0;
1577 		TAILQ_FOREACH(oldrule,
1578 		    ruleset->rules[rs_num].active.ptr, entries)
1579 			oldrule->nr = nr++;
1580 
1581 		ruleset->rules[rs_num].active.ticket++;
1582 
1583 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1584 		pf_remove_if_empty_ruleset(ruleset);
1585 
1586 		PF_RULES_WUNLOCK();
1587 		break;
1588 
1589 #undef ERROUT
1590 DIOCCHANGERULE_error:
1591 		PF_RULES_WUNLOCK();
1592 		if (newrule != NULL) {
1593 			counter_u64_free(newrule->states_cur);
1594 			counter_u64_free(newrule->states_tot);
1595 			counter_u64_free(newrule->src_nodes);
1596 			free(newrule, M_PFRULE);
1597 		}
1598 		if (kif != NULL)
1599 			free(kif, PFI_MTYPE);
1600 		break;
1601 	}
1602 
1603 	case DIOCCLRSTATES: {
1604 		struct pf_state		*s;
1605 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1606 		u_int			 i, killed = 0;
1607 
1608 		for (i = 0; i <= pf_hashmask; i++) {
1609 			struct pf_idhash *ih = &V_pf_idhash[i];
1610 
1611 relock_DIOCCLRSTATES:
1612 			PF_HASHROW_LOCK(ih);
1613 			LIST_FOREACH(s, &ih->states, entry)
1614 				if (!psk->psk_ifname[0] ||
1615 				    !strcmp(psk->psk_ifname,
1616 				    s->kif->pfik_name)) {
1617 					/*
1618 					 * Don't send out individual
1619 					 * delete messages.
1620 					 */
1621 					s->state_flags |= PFSTATE_NOSYNC;
1622 					pf_unlink_state(s, PF_ENTER_LOCKED);
1623 					killed++;
1624 					goto relock_DIOCCLRSTATES;
1625 				}
1626 			PF_HASHROW_UNLOCK(ih);
1627 		}
1628 		psk->psk_killed = killed;
1629 		if (pfsync_clear_states_ptr != NULL)
1630 			pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
1631 		break;
1632 	}
1633 
1634 	case DIOCKILLSTATES: {
1635 		struct pf_state		*s;
1636 		struct pf_state_key	*sk;
1637 		struct pf_addr		*srcaddr, *dstaddr;
1638 		u_int16_t		 srcport, dstport;
1639 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
1640 		u_int			 i, killed = 0;
1641 
1642 		if (psk->psk_pfcmp.id) {
1643 			if (psk->psk_pfcmp.creatorid == 0)
1644 				psk->psk_pfcmp.creatorid = V_pf_status.hostid;
1645 			if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
1646 			    psk->psk_pfcmp.creatorid))) {
1647 				pf_unlink_state(s, PF_ENTER_LOCKED);
1648 				psk->psk_killed = 1;
1649 			}
1650 			break;
1651 		}
1652 
1653 		for (i = 0; i <= pf_hashmask; i++) {
1654 			struct pf_idhash *ih = &V_pf_idhash[i];
1655 
1656 relock_DIOCKILLSTATES:
1657 			PF_HASHROW_LOCK(ih);
1658 			LIST_FOREACH(s, &ih->states, entry) {
1659 				sk = s->key[PF_SK_WIRE];
1660 				if (s->direction == PF_OUT) {
1661 					srcaddr = &sk->addr[1];
1662 					dstaddr = &sk->addr[0];
1663 					srcport = sk->port[0];
1664 					dstport = sk->port[0];
1665 				} else {
1666 					srcaddr = &sk->addr[0];
1667 					dstaddr = &sk->addr[1];
1668 					srcport = sk->port[0];
1669 					dstport = sk->port[0];
1670 				}
1671 
1672 				if ((!psk->psk_af || sk->af == psk->psk_af)
1673 				    && (!psk->psk_proto || psk->psk_proto ==
1674 				    sk->proto) &&
1675 				    PF_MATCHA(psk->psk_src.neg,
1676 				    &psk->psk_src.addr.v.a.addr,
1677 				    &psk->psk_src.addr.v.a.mask,
1678 				    srcaddr, sk->af) &&
1679 				    PF_MATCHA(psk->psk_dst.neg,
1680 				    &psk->psk_dst.addr.v.a.addr,
1681 				    &psk->psk_dst.addr.v.a.mask,
1682 				    dstaddr, sk->af) &&
1683 				    (psk->psk_src.port_op == 0 ||
1684 				    pf_match_port(psk->psk_src.port_op,
1685 				    psk->psk_src.port[0], psk->psk_src.port[1],
1686 				    srcport)) &&
1687 				    (psk->psk_dst.port_op == 0 ||
1688 				    pf_match_port(psk->psk_dst.port_op,
1689 				    psk->psk_dst.port[0], psk->psk_dst.port[1],
1690 				    dstport)) &&
1691 				    (!psk->psk_label[0] ||
1692 				    (s->rule.ptr->label[0] &&
1693 				    !strcmp(psk->psk_label,
1694 				    s->rule.ptr->label))) &&
1695 				    (!psk->psk_ifname[0] ||
1696 				    !strcmp(psk->psk_ifname,
1697 				    s->kif->pfik_name))) {
1698 					pf_unlink_state(s, PF_ENTER_LOCKED);
1699 					killed++;
1700 					goto relock_DIOCKILLSTATES;
1701 				}
1702 			}
1703 			PF_HASHROW_UNLOCK(ih);
1704 		}
1705 		psk->psk_killed = killed;
1706 		break;
1707 	}
1708 
1709 	case DIOCADDSTATE: {
1710 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1711 		struct pfsync_state	*sp = &ps->state;
1712 
1713 		if (sp->timeout >= PFTM_MAX) {
1714 			error = EINVAL;
1715 			break;
1716 		}
1717 		if (pfsync_state_import_ptr != NULL) {
1718 			PF_RULES_RLOCK();
1719 			error = pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
1720 			PF_RULES_RUNLOCK();
1721 		} else
1722 			error = EOPNOTSUPP;
1723 		break;
1724 	}
1725 
1726 	case DIOCGETSTATE: {
1727 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1728 		struct pf_state		*s;
1729 
1730 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
1731 		if (s == NULL) {
1732 			error = ENOENT;
1733 			break;
1734 		}
1735 
1736 		pfsync_state_export(&ps->state, s);
1737 		PF_STATE_UNLOCK(s);
1738 		break;
1739 	}
1740 
1741 	case DIOCGETSTATES: {
1742 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
1743 		struct pf_state		*s;
1744 		struct pfsync_state	*pstore, *p;
1745 		int i, nr;
1746 
1747 		if (ps->ps_len == 0) {
1748 			nr = uma_zone_get_cur(V_pf_state_z);
1749 			ps->ps_len = sizeof(struct pfsync_state) * nr;
1750 			break;
1751 		}
1752 
1753 		p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK);
1754 		nr = 0;
1755 
1756 		for (i = 0; i <= pf_hashmask; i++) {
1757 			struct pf_idhash *ih = &V_pf_idhash[i];
1758 
1759 			PF_HASHROW_LOCK(ih);
1760 			LIST_FOREACH(s, &ih->states, entry) {
1761 
1762 				if (s->timeout == PFTM_UNLINKED)
1763 					continue;
1764 
1765 				if ((nr+1) * sizeof(*p) > ps->ps_len) {
1766 					PF_HASHROW_UNLOCK(ih);
1767 					goto DIOCGETSTATES_full;
1768 				}
1769 				pfsync_state_export(p, s);
1770 				p++;
1771 				nr++;
1772 			}
1773 			PF_HASHROW_UNLOCK(ih);
1774 		}
1775 DIOCGETSTATES_full:
1776 		error = copyout(pstore, ps->ps_states,
1777 		    sizeof(struct pfsync_state) * nr);
1778 		if (error) {
1779 			free(pstore, M_TEMP);
1780 			break;
1781 		}
1782 		ps->ps_len = sizeof(struct pfsync_state) * nr;
1783 		free(pstore, M_TEMP);
1784 
1785 		break;
1786 	}
1787 
1788 	case DIOCGETSTATUS: {
1789 		struct pf_status *s = (struct pf_status *)addr;
1790 		PF_RULES_RLOCK();
1791 		bcopy(&V_pf_status, s, sizeof(struct pf_status));
1792 		pfi_update_status(s->ifname, s);
1793 		PF_RULES_RUNLOCK();
1794 		break;
1795 	}
1796 
1797 	case DIOCSETSTATUSIF: {
1798 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
1799 
1800 		if (pi->ifname[0] == 0) {
1801 			bzero(V_pf_status.ifname, IFNAMSIZ);
1802 			break;
1803 		}
1804 		PF_RULES_WLOCK();
1805 		strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
1806 		PF_RULES_WUNLOCK();
1807 		break;
1808 	}
1809 
1810 	case DIOCCLRSTATUS: {
1811 		PF_RULES_WLOCK();
1812 		bzero(V_pf_status.counters, sizeof(V_pf_status.counters));
1813 		bzero(V_pf_status.fcounters, sizeof(V_pf_status.fcounters));
1814 		bzero(V_pf_status.scounters, sizeof(V_pf_status.scounters));
1815 		V_pf_status.since = time_second;
1816 		if (*V_pf_status.ifname)
1817 			pfi_update_status(V_pf_status.ifname, NULL);
1818 		PF_RULES_WUNLOCK();
1819 		break;
1820 	}
1821 
1822 	case DIOCNATLOOK: {
1823 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
1824 		struct pf_state_key	*sk;
1825 		struct pf_state		*state;
1826 		struct pf_state_key_cmp	 key;
1827 		int			 m = 0, direction = pnl->direction;
1828 		int			 sidx, didx;
1829 
1830 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
1831 		sidx = (direction == PF_IN) ? 1 : 0;
1832 		didx = (direction == PF_IN) ? 0 : 1;
1833 
1834 		if (!pnl->proto ||
1835 		    PF_AZERO(&pnl->saddr, pnl->af) ||
1836 		    PF_AZERO(&pnl->daddr, pnl->af) ||
1837 		    ((pnl->proto == IPPROTO_TCP ||
1838 		    pnl->proto == IPPROTO_UDP) &&
1839 		    (!pnl->dport || !pnl->sport)))
1840 			error = EINVAL;
1841 		else {
1842 			bzero(&key, sizeof(key));
1843 			key.af = pnl->af;
1844 			key.proto = pnl->proto;
1845 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
1846 			key.port[sidx] = pnl->sport;
1847 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
1848 			key.port[didx] = pnl->dport;
1849 
1850 			state = pf_find_state_all(&key, direction, &m);
1851 
1852 			if (m > 1)
1853 				error = E2BIG;	/* more than one state */
1854 			else if (state != NULL) {
1855 				/* XXXGL: not locked read */
1856 				sk = state->key[sidx];
1857 				PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
1858 				pnl->rsport = sk->port[sidx];
1859 				PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
1860 				pnl->rdport = sk->port[didx];
1861 			} else
1862 				error = ENOENT;
1863 		}
1864 		break;
1865 	}
1866 
1867 	case DIOCSETTIMEOUT: {
1868 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1869 		int		 old;
1870 
1871 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
1872 		    pt->seconds < 0) {
1873 			error = EINVAL;
1874 			break;
1875 		}
1876 		PF_RULES_WLOCK();
1877 		old = V_pf_default_rule.timeout[pt->timeout];
1878 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
1879 			pt->seconds = 1;
1880 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
1881 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
1882 			wakeup(pf_purge_thread);
1883 		pt->seconds = old;
1884 		PF_RULES_WUNLOCK();
1885 		break;
1886 	}
1887 
1888 	case DIOCGETTIMEOUT: {
1889 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1890 
1891 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
1892 			error = EINVAL;
1893 			break;
1894 		}
1895 		PF_RULES_RLOCK();
1896 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
1897 		PF_RULES_RUNLOCK();
1898 		break;
1899 	}
1900 
1901 	case DIOCGETLIMIT: {
1902 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1903 
1904 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
1905 			error = EINVAL;
1906 			break;
1907 		}
1908 		PF_RULES_RLOCK();
1909 		pl->limit = V_pf_limits[pl->index].limit;
1910 		PF_RULES_RUNLOCK();
1911 		break;
1912 	}
1913 
1914 	case DIOCSETLIMIT: {
1915 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1916 		int			 old_limit;
1917 
1918 		PF_RULES_WLOCK();
1919 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
1920 		    V_pf_limits[pl->index].zone == NULL) {
1921 			PF_RULES_WUNLOCK();
1922 			error = EINVAL;
1923 			break;
1924 		}
1925 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
1926 		old_limit = V_pf_limits[pl->index].limit;
1927 		V_pf_limits[pl->index].limit = pl->limit;
1928 		pl->limit = old_limit;
1929 		PF_RULES_WUNLOCK();
1930 		break;
1931 	}
1932 
1933 	case DIOCSETDEBUG: {
1934 		u_int32_t	*level = (u_int32_t *)addr;
1935 
1936 		PF_RULES_WLOCK();
1937 		V_pf_status.debug = *level;
1938 		PF_RULES_WUNLOCK();
1939 		break;
1940 	}
1941 
1942 	case DIOCCLRRULECTRS: {
1943 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
1944 		struct pf_ruleset	*ruleset = &pf_main_ruleset;
1945 		struct pf_rule		*rule;
1946 
1947 		PF_RULES_WLOCK();
1948 		TAILQ_FOREACH(rule,
1949 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
1950 			rule->evaluations = 0;
1951 			rule->packets[0] = rule->packets[1] = 0;
1952 			rule->bytes[0] = rule->bytes[1] = 0;
1953 		}
1954 		PF_RULES_WUNLOCK();
1955 		break;
1956 	}
1957 
1958 	case DIOCGIFSPEED: {
1959 		struct pf_ifspeed	*psp = (struct pf_ifspeed *)addr;
1960 		struct pf_ifspeed	ps;
1961 		struct ifnet		*ifp;
1962 
1963 		if (psp->ifname[0] != 0) {
1964 			/* Can we completely trust user-land? */
1965 			strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
1966 			ifp = ifunit(ps.ifname);
1967 			if (ifp != NULL)
1968 				psp->baudrate = ifp->if_baudrate;
1969 			else
1970 				error = EINVAL;
1971 		} else
1972 			error = EINVAL;
1973 		break;
1974 	}
1975 
1976 #ifdef ALTQ
1977 	case DIOCSTARTALTQ: {
1978 		struct pf_altq		*altq;
1979 
1980 		PF_RULES_WLOCK();
1981 		/* enable all altq interfaces on active list */
1982 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1983 			if (altq->qname[0] == 0 && (altq->local_flags &
1984 			    PFALTQ_FLAG_IF_REMOVED) == 0) {
1985 				error = pf_enable_altq(altq);
1986 				if (error != 0)
1987 					break;
1988 			}
1989 		}
1990 		if (error == 0)
1991 			V_pf_altq_running = 1;
1992 		PF_RULES_WUNLOCK();
1993 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
1994 		break;
1995 	}
1996 
1997 	case DIOCSTOPALTQ: {
1998 		struct pf_altq		*altq;
1999 
2000 		PF_RULES_WLOCK();
2001 		/* disable all altq interfaces on active list */
2002 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
2003 			if (altq->qname[0] == 0 && (altq->local_flags &
2004 			    PFALTQ_FLAG_IF_REMOVED) == 0) {
2005 				error = pf_disable_altq(altq);
2006 				if (error != 0)
2007 					break;
2008 			}
2009 		}
2010 		if (error == 0)
2011 			V_pf_altq_running = 0;
2012 		PF_RULES_WUNLOCK();
2013 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2014 		break;
2015 	}
2016 
2017 	case DIOCADDALTQ: {
2018 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2019 		struct pf_altq		*altq, *a;
2020 		struct ifnet		*ifp;
2021 
2022 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK);
2023 		bcopy(&pa->altq, altq, sizeof(struct pf_altq));
2024 		altq->local_flags = 0;
2025 
2026 		PF_RULES_WLOCK();
2027 		if (pa->ticket != V_ticket_altqs_inactive) {
2028 			PF_RULES_WUNLOCK();
2029 			free(altq, M_PFALTQ);
2030 			error = EBUSY;
2031 			break;
2032 		}
2033 
2034 		/*
2035 		 * if this is for a queue, find the discipline and
2036 		 * copy the necessary fields
2037 		 */
2038 		if (altq->qname[0] != 0) {
2039 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2040 				PF_RULES_WUNLOCK();
2041 				error = EBUSY;
2042 				free(altq, M_PFALTQ);
2043 				break;
2044 			}
2045 			altq->altq_disc = NULL;
2046 			TAILQ_FOREACH(a, V_pf_altqs_inactive, entries) {
2047 				if (strncmp(a->ifname, altq->ifname,
2048 				    IFNAMSIZ) == 0 && a->qname[0] == 0) {
2049 					altq->altq_disc = a->altq_disc;
2050 					break;
2051 				}
2052 			}
2053 		}
2054 
2055 		if ((ifp = ifunit(altq->ifname)) == NULL)
2056 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
2057 		else
2058 			error = altq_add(altq);
2059 
2060 		if (error) {
2061 			PF_RULES_WUNLOCK();
2062 			free(altq, M_PFALTQ);
2063 			break;
2064 		}
2065 
2066 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
2067 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2068 		PF_RULES_WUNLOCK();
2069 		break;
2070 	}
2071 
2072 	case DIOCGETALTQS: {
2073 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2074 		struct pf_altq		*altq;
2075 
2076 		PF_RULES_RLOCK();
2077 		pa->nr = 0;
2078 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
2079 			pa->nr++;
2080 		pa->ticket = V_ticket_altqs_active;
2081 		PF_RULES_RUNLOCK();
2082 		break;
2083 	}
2084 
2085 	case DIOCGETALTQ: {
2086 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2087 		struct pf_altq		*altq;
2088 		u_int32_t		 nr;
2089 
2090 		PF_RULES_RLOCK();
2091 		if (pa->ticket != V_ticket_altqs_active) {
2092 			PF_RULES_RUNLOCK();
2093 			error = EBUSY;
2094 			break;
2095 		}
2096 		nr = 0;
2097 		altq = TAILQ_FIRST(V_pf_altqs_active);
2098 		while ((altq != NULL) && (nr < pa->nr)) {
2099 			altq = TAILQ_NEXT(altq, entries);
2100 			nr++;
2101 		}
2102 		if (altq == NULL) {
2103 			PF_RULES_RUNLOCK();
2104 			error = EBUSY;
2105 			break;
2106 		}
2107 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2108 		PF_RULES_RUNLOCK();
2109 		break;
2110 	}
2111 
2112 	case DIOCCHANGEALTQ:
2113 		/* CHANGEALTQ not supported yet! */
2114 		error = ENODEV;
2115 		break;
2116 
2117 	case DIOCGETQSTATS: {
2118 		struct pfioc_qstats	*pq = (struct pfioc_qstats *)addr;
2119 		struct pf_altq		*altq;
2120 		u_int32_t		 nr;
2121 		int			 nbytes;
2122 
2123 		PF_RULES_RLOCK();
2124 		if (pq->ticket != V_ticket_altqs_active) {
2125 			PF_RULES_RUNLOCK();
2126 			error = EBUSY;
2127 			break;
2128 		}
2129 		nbytes = pq->nbytes;
2130 		nr = 0;
2131 		altq = TAILQ_FIRST(V_pf_altqs_active);
2132 		while ((altq != NULL) && (nr < pq->nr)) {
2133 			altq = TAILQ_NEXT(altq, entries);
2134 			nr++;
2135 		}
2136 		if (altq == NULL) {
2137 			PF_RULES_RUNLOCK();
2138 			error = EBUSY;
2139 			break;
2140 		}
2141 
2142 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
2143 			PF_RULES_RUNLOCK();
2144 			error = ENXIO;
2145 			break;
2146 		}
2147 		PF_RULES_RUNLOCK();
2148 		error = altq_getqstats(altq, pq->buf, &nbytes);
2149 		if (error == 0) {
2150 			pq->scheduler = altq->scheduler;
2151 			pq->nbytes = nbytes;
2152 		}
2153 		break;
2154 	}
2155 #endif /* ALTQ */
2156 
2157 	case DIOCBEGINADDRS: {
2158 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2159 
2160 		PF_RULES_WLOCK();
2161 		pf_empty_pool(&V_pf_pabuf);
2162 		pp->ticket = ++V_ticket_pabuf;
2163 		PF_RULES_WUNLOCK();
2164 		break;
2165 	}
2166 
2167 	case DIOCADDADDR: {
2168 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2169 		struct pf_pooladdr	*pa;
2170 		struct pfi_kif		*kif = NULL;
2171 
2172 #ifndef INET
2173 		if (pp->af == AF_INET) {
2174 			error = EAFNOSUPPORT;
2175 			break;
2176 		}
2177 #endif /* INET */
2178 #ifndef INET6
2179 		if (pp->af == AF_INET6) {
2180 			error = EAFNOSUPPORT;
2181 			break;
2182 		}
2183 #endif /* INET6 */
2184 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2185 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2186 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2187 			error = EINVAL;
2188 			break;
2189 		}
2190 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2191 		bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2192 		if (pa->ifname[0])
2193 			kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2194 		PF_RULES_WLOCK();
2195 		if (pp->ticket != V_ticket_pabuf) {
2196 			PF_RULES_WUNLOCK();
2197 			if (pa->ifname[0])
2198 				free(kif, PFI_MTYPE);
2199 			free(pa, M_PFRULE);
2200 			error = EBUSY;
2201 			break;
2202 		}
2203 		if (pa->ifname[0]) {
2204 			pa->kif = pfi_kif_attach(kif, pa->ifname);
2205 			pfi_kif_ref(pa->kif);
2206 		} else
2207 			pa->kif = NULL;
2208 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2209 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2210 			if (pa->ifname[0])
2211 				pfi_kif_unref(pa->kif);
2212 			PF_RULES_WUNLOCK();
2213 			free(pa, M_PFRULE);
2214 			break;
2215 		}
2216 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2217 		PF_RULES_WUNLOCK();
2218 		break;
2219 	}
2220 
2221 	case DIOCGETADDRS: {
2222 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2223 		struct pf_pool		*pool;
2224 		struct pf_pooladdr	*pa;
2225 
2226 		PF_RULES_RLOCK();
2227 		pp->nr = 0;
2228 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2229 		    pp->r_num, 0, 1, 0);
2230 		if (pool == NULL) {
2231 			PF_RULES_RUNLOCK();
2232 			error = EBUSY;
2233 			break;
2234 		}
2235 		TAILQ_FOREACH(pa, &pool->list, entries)
2236 			pp->nr++;
2237 		PF_RULES_RUNLOCK();
2238 		break;
2239 	}
2240 
2241 	case DIOCGETADDR: {
2242 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2243 		struct pf_pool		*pool;
2244 		struct pf_pooladdr	*pa;
2245 		u_int32_t		 nr = 0;
2246 
2247 		PF_RULES_RLOCK();
2248 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2249 		    pp->r_num, 0, 1, 1);
2250 		if (pool == NULL) {
2251 			PF_RULES_RUNLOCK();
2252 			error = EBUSY;
2253 			break;
2254 		}
2255 		pa = TAILQ_FIRST(&pool->list);
2256 		while ((pa != NULL) && (nr < pp->nr)) {
2257 			pa = TAILQ_NEXT(pa, entries);
2258 			nr++;
2259 		}
2260 		if (pa == NULL) {
2261 			PF_RULES_RUNLOCK();
2262 			error = EBUSY;
2263 			break;
2264 		}
2265 		bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2266 		pf_addr_copyout(&pp->addr.addr);
2267 		PF_RULES_RUNLOCK();
2268 		break;
2269 	}
2270 
2271 	case DIOCCHANGEADDR: {
2272 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
2273 		struct pf_pool		*pool;
2274 		struct pf_pooladdr	*oldpa = NULL, *newpa = NULL;
2275 		struct pf_ruleset	*ruleset;
2276 		struct pfi_kif		*kif = NULL;
2277 
2278 		if (pca->action < PF_CHANGE_ADD_HEAD ||
2279 		    pca->action > PF_CHANGE_REMOVE) {
2280 			error = EINVAL;
2281 			break;
2282 		}
2283 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2284 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2285 		    pca->addr.addr.type != PF_ADDR_TABLE) {
2286 			error = EINVAL;
2287 			break;
2288 		}
2289 
2290 		if (pca->action != PF_CHANGE_REMOVE) {
2291 #ifndef INET
2292 			if (pca->af == AF_INET) {
2293 				error = EAFNOSUPPORT;
2294 				break;
2295 			}
2296 #endif /* INET */
2297 #ifndef INET6
2298 			if (pca->af == AF_INET6) {
2299 				error = EAFNOSUPPORT;
2300 				break;
2301 			}
2302 #endif /* INET6 */
2303 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
2304 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2305 			if (newpa->ifname[0])
2306 				kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2307 			newpa->kif = NULL;
2308 		}
2309 
2310 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGEADDR_error; }
2311 		PF_RULES_WLOCK();
2312 		ruleset = pf_find_ruleset(pca->anchor);
2313 		if (ruleset == NULL)
2314 			ERROUT(EBUSY);
2315 
2316 		pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2317 		    pca->r_num, pca->r_last, 1, 1);
2318 		if (pool == NULL)
2319 			ERROUT(EBUSY);
2320 
2321 		if (pca->action != PF_CHANGE_REMOVE) {
2322 			if (newpa->ifname[0]) {
2323 				newpa->kif = pfi_kif_attach(kif, newpa->ifname);
2324 				pfi_kif_ref(newpa->kif);
2325 				kif = NULL;
2326 			}
2327 
2328 			switch (newpa->addr.type) {
2329 			case PF_ADDR_DYNIFTL:
2330 				error = pfi_dynaddr_setup(&newpa->addr,
2331 				    pca->af);
2332 				break;
2333 			case PF_ADDR_TABLE:
2334 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
2335 				    newpa->addr.v.tblname);
2336 				if (newpa->addr.p.tbl == NULL)
2337 					error = ENOMEM;
2338 				break;
2339 			}
2340 			if (error)
2341 				goto DIOCCHANGEADDR_error;
2342 		}
2343 
2344 		switch (pca->action) {
2345 		case PF_CHANGE_ADD_HEAD:
2346 			oldpa = TAILQ_FIRST(&pool->list);
2347 			break;
2348 		case PF_CHANGE_ADD_TAIL:
2349 			oldpa = TAILQ_LAST(&pool->list, pf_palist);
2350 			break;
2351 		default:
2352 			oldpa = TAILQ_FIRST(&pool->list);
2353 			for (int i = 0; oldpa && i < pca->nr; i++)
2354 				oldpa = TAILQ_NEXT(oldpa, entries);
2355 
2356 			if (oldpa == NULL)
2357 				ERROUT(EINVAL);
2358 		}
2359 
2360 		if (pca->action == PF_CHANGE_REMOVE) {
2361 			TAILQ_REMOVE(&pool->list, oldpa, entries);
2362 			switch (oldpa->addr.type) {
2363 			case PF_ADDR_DYNIFTL:
2364 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
2365 				break;
2366 			case PF_ADDR_TABLE:
2367 				pfr_detach_table(oldpa->addr.p.tbl);
2368 				break;
2369 			}
2370 			if (oldpa->kif)
2371 				pfi_kif_unref(oldpa->kif);
2372 			free(oldpa, M_PFRULE);
2373 		} else {
2374 			if (oldpa == NULL)
2375 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2376 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
2377 			    pca->action == PF_CHANGE_ADD_BEFORE)
2378 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2379 			else
2380 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
2381 				    newpa, entries);
2382 		}
2383 
2384 		pool->cur = TAILQ_FIRST(&pool->list);
2385 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
2386 		PF_RULES_WUNLOCK();
2387 		break;
2388 
2389 #undef ERROUT
2390 DIOCCHANGEADDR_error:
2391 		if (newpa->kif)
2392 			pfi_kif_unref(newpa->kif);
2393 		PF_RULES_WUNLOCK();
2394 		if (newpa != NULL)
2395 			free(newpa, M_PFRULE);
2396 		if (kif != NULL)
2397 			free(kif, PFI_MTYPE);
2398 		break;
2399 	}
2400 
2401 	case DIOCGETRULESETS: {
2402 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2403 		struct pf_ruleset	*ruleset;
2404 		struct pf_anchor	*anchor;
2405 
2406 		PF_RULES_RLOCK();
2407 		pr->path[sizeof(pr->path) - 1] = 0;
2408 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2409 			PF_RULES_RUNLOCK();
2410 			error = ENOENT;
2411 			break;
2412 		}
2413 		pr->nr = 0;
2414 		if (ruleset->anchor == NULL) {
2415 			/* XXX kludge for pf_main_ruleset */
2416 			RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2417 				if (anchor->parent == NULL)
2418 					pr->nr++;
2419 		} else {
2420 			RB_FOREACH(anchor, pf_anchor_node,
2421 			    &ruleset->anchor->children)
2422 				pr->nr++;
2423 		}
2424 		PF_RULES_RUNLOCK();
2425 		break;
2426 	}
2427 
2428 	case DIOCGETRULESET: {
2429 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2430 		struct pf_ruleset	*ruleset;
2431 		struct pf_anchor	*anchor;
2432 		u_int32_t		 nr = 0;
2433 
2434 		PF_RULES_RLOCK();
2435 		pr->path[sizeof(pr->path) - 1] = 0;
2436 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2437 			PF_RULES_RUNLOCK();
2438 			error = ENOENT;
2439 			break;
2440 		}
2441 		pr->name[0] = 0;
2442 		if (ruleset->anchor == NULL) {
2443 			/* XXX kludge for pf_main_ruleset */
2444 			RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2445 				if (anchor->parent == NULL && nr++ == pr->nr) {
2446 					strlcpy(pr->name, anchor->name,
2447 					    sizeof(pr->name));
2448 					break;
2449 				}
2450 		} else {
2451 			RB_FOREACH(anchor, pf_anchor_node,
2452 			    &ruleset->anchor->children)
2453 				if (nr++ == pr->nr) {
2454 					strlcpy(pr->name, anchor->name,
2455 					    sizeof(pr->name));
2456 					break;
2457 				}
2458 		}
2459 		if (!pr->name[0])
2460 			error = EBUSY;
2461 		PF_RULES_RUNLOCK();
2462 		break;
2463 	}
2464 
2465 	case DIOCRCLRTABLES: {
2466 		struct pfioc_table *io = (struct pfioc_table *)addr;
2467 
2468 		if (io->pfrio_esize != 0) {
2469 			error = ENODEV;
2470 			break;
2471 		}
2472 		PF_RULES_WLOCK();
2473 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2474 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2475 		PF_RULES_WUNLOCK();
2476 		break;
2477 	}
2478 
2479 	case DIOCRADDTABLES: {
2480 		struct pfioc_table *io = (struct pfioc_table *)addr;
2481 		struct pfr_table *pfrts;
2482 		size_t totlen;
2483 
2484 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2485 			error = ENODEV;
2486 			break;
2487 		}
2488 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2489 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2490 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2491 		if (error) {
2492 			free(pfrts, M_TEMP);
2493 			break;
2494 		}
2495 		PF_RULES_WLOCK();
2496 		error = pfr_add_tables(pfrts, io->pfrio_size,
2497 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2498 		PF_RULES_WUNLOCK();
2499 		free(pfrts, M_TEMP);
2500 		break;
2501 	}
2502 
2503 	case DIOCRDELTABLES: {
2504 		struct pfioc_table *io = (struct pfioc_table *)addr;
2505 		struct pfr_table *pfrts;
2506 		size_t totlen;
2507 
2508 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2509 			error = ENODEV;
2510 			break;
2511 		}
2512 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2513 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2514 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2515 		if (error) {
2516 			free(pfrts, M_TEMP);
2517 			break;
2518 		}
2519 		PF_RULES_WLOCK();
2520 		error = pfr_del_tables(pfrts, io->pfrio_size,
2521 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2522 		PF_RULES_WUNLOCK();
2523 		free(pfrts, M_TEMP);
2524 		break;
2525 	}
2526 
2527 	case DIOCRGETTABLES: {
2528 		struct pfioc_table *io = (struct pfioc_table *)addr;
2529 		struct pfr_table *pfrts;
2530 		size_t totlen;
2531 
2532 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2533 			error = ENODEV;
2534 			break;
2535 		}
2536 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2537 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2538 		PF_RULES_RLOCK();
2539 		error = pfr_get_tables(&io->pfrio_table, pfrts,
2540 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2541 		PF_RULES_RUNLOCK();
2542 		if (error == 0)
2543 			error = copyout(pfrts, io->pfrio_buffer, totlen);
2544 		free(pfrts, M_TEMP);
2545 		break;
2546 	}
2547 
2548 	case DIOCRGETTSTATS: {
2549 		struct pfioc_table *io = (struct pfioc_table *)addr;
2550 		struct pfr_tstats *pfrtstats;
2551 		size_t totlen;
2552 
2553 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2554 			error = ENODEV;
2555 			break;
2556 		}
2557 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
2558 		pfrtstats = malloc(totlen, M_TEMP, M_WAITOK);
2559 		PF_RULES_WLOCK();
2560 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
2561 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2562 		PF_RULES_WUNLOCK();
2563 		if (error == 0)
2564 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
2565 		free(pfrtstats, M_TEMP);
2566 		break;
2567 	}
2568 
2569 	case DIOCRCLRTSTATS: {
2570 		struct pfioc_table *io = (struct pfioc_table *)addr;
2571 		struct pfr_table *pfrts;
2572 		size_t totlen;
2573 
2574 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2575 			error = ENODEV;
2576 			break;
2577 		}
2578 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2579 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2580 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2581 		if (error) {
2582 			free(pfrts, M_TEMP);
2583 			break;
2584 		}
2585 		PF_RULES_WLOCK();
2586 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
2587 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2588 		PF_RULES_WUNLOCK();
2589 		free(pfrts, M_TEMP);
2590 		break;
2591 	}
2592 
2593 	case DIOCRSETTFLAGS: {
2594 		struct pfioc_table *io = (struct pfioc_table *)addr;
2595 		struct pfr_table *pfrts;
2596 		size_t totlen;
2597 
2598 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2599 			error = ENODEV;
2600 			break;
2601 		}
2602 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2603 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2604 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2605 		if (error) {
2606 			free(pfrts, M_TEMP);
2607 			break;
2608 		}
2609 		PF_RULES_WLOCK();
2610 		error = pfr_set_tflags(pfrts, io->pfrio_size,
2611 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2612 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2613 		PF_RULES_WUNLOCK();
2614 		free(pfrts, M_TEMP);
2615 		break;
2616 	}
2617 
2618 	case DIOCRCLRADDRS: {
2619 		struct pfioc_table *io = (struct pfioc_table *)addr;
2620 
2621 		if (io->pfrio_esize != 0) {
2622 			error = ENODEV;
2623 			break;
2624 		}
2625 		PF_RULES_WLOCK();
2626 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2627 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2628 		PF_RULES_WUNLOCK();
2629 		break;
2630 	}
2631 
2632 	case DIOCRADDADDRS: {
2633 		struct pfioc_table *io = (struct pfioc_table *)addr;
2634 		struct pfr_addr *pfras;
2635 		size_t totlen;
2636 
2637 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2638 			error = ENODEV;
2639 			break;
2640 		}
2641 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2642 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2643 		error = copyin(io->pfrio_buffer, pfras, totlen);
2644 		if (error) {
2645 			free(pfras, M_TEMP);
2646 			break;
2647 		}
2648 		PF_RULES_WLOCK();
2649 		error = pfr_add_addrs(&io->pfrio_table, pfras,
2650 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2651 		    PFR_FLAG_USERIOCTL);
2652 		PF_RULES_WUNLOCK();
2653 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2654 			error = copyout(pfras, io->pfrio_buffer, totlen);
2655 		free(pfras, M_TEMP);
2656 		break;
2657 	}
2658 
2659 	case DIOCRDELADDRS: {
2660 		struct pfioc_table *io = (struct pfioc_table *)addr;
2661 		struct pfr_addr *pfras;
2662 		size_t totlen;
2663 
2664 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2665 			error = ENODEV;
2666 			break;
2667 		}
2668 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2669 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2670 		error = copyin(io->pfrio_buffer, pfras, totlen);
2671 		if (error) {
2672 			free(pfras, M_TEMP);
2673 			break;
2674 		}
2675 		PF_RULES_WLOCK();
2676 		error = pfr_del_addrs(&io->pfrio_table, pfras,
2677 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2678 		    PFR_FLAG_USERIOCTL);
2679 		PF_RULES_WUNLOCK();
2680 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2681 			error = copyout(pfras, io->pfrio_buffer, totlen);
2682 		free(pfras, M_TEMP);
2683 		break;
2684 	}
2685 
2686 	case DIOCRSETADDRS: {
2687 		struct pfioc_table *io = (struct pfioc_table *)addr;
2688 		struct pfr_addr *pfras;
2689 		size_t totlen;
2690 
2691 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2692 			error = ENODEV;
2693 			break;
2694 		}
2695 		totlen = (io->pfrio_size + io->pfrio_size2) *
2696 		    sizeof(struct pfr_addr);
2697 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2698 		error = copyin(io->pfrio_buffer, pfras, totlen);
2699 		if (error) {
2700 			free(pfras, M_TEMP);
2701 			break;
2702 		}
2703 		PF_RULES_WLOCK();
2704 		error = pfr_set_addrs(&io->pfrio_table, pfras,
2705 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2706 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2707 		    PFR_FLAG_USERIOCTL, 0);
2708 		PF_RULES_WUNLOCK();
2709 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2710 			error = copyout(pfras, io->pfrio_buffer, totlen);
2711 		free(pfras, M_TEMP);
2712 		break;
2713 	}
2714 
2715 	case DIOCRGETADDRS: {
2716 		struct pfioc_table *io = (struct pfioc_table *)addr;
2717 		struct pfr_addr *pfras;
2718 		size_t totlen;
2719 
2720 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2721 			error = ENODEV;
2722 			break;
2723 		}
2724 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2725 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2726 		PF_RULES_RLOCK();
2727 		error = pfr_get_addrs(&io->pfrio_table, pfras,
2728 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2729 		PF_RULES_RUNLOCK();
2730 		if (error == 0)
2731 			error = copyout(pfras, io->pfrio_buffer, totlen);
2732 		free(pfras, M_TEMP);
2733 		break;
2734 	}
2735 
2736 	case DIOCRGETASTATS: {
2737 		struct pfioc_table *io = (struct pfioc_table *)addr;
2738 		struct pfr_astats *pfrastats;
2739 		size_t totlen;
2740 
2741 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2742 			error = ENODEV;
2743 			break;
2744 		}
2745 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
2746 		pfrastats = malloc(totlen, M_TEMP, M_WAITOK);
2747 		PF_RULES_RLOCK();
2748 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
2749 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2750 		PF_RULES_RUNLOCK();
2751 		if (error == 0)
2752 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
2753 		free(pfrastats, M_TEMP);
2754 		break;
2755 	}
2756 
2757 	case DIOCRCLRASTATS: {
2758 		struct pfioc_table *io = (struct pfioc_table *)addr;
2759 		struct pfr_addr *pfras;
2760 		size_t totlen;
2761 
2762 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2763 			error = ENODEV;
2764 			break;
2765 		}
2766 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2767 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2768 		error = copyin(io->pfrio_buffer, pfras, totlen);
2769 		if (error) {
2770 			free(pfras, M_TEMP);
2771 			break;
2772 		}
2773 		PF_RULES_WLOCK();
2774 		error = pfr_clr_astats(&io->pfrio_table, pfras,
2775 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2776 		    PFR_FLAG_USERIOCTL);
2777 		PF_RULES_WUNLOCK();
2778 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2779 			error = copyout(pfras, io->pfrio_buffer, totlen);
2780 		free(pfras, M_TEMP);
2781 		break;
2782 	}
2783 
2784 	case DIOCRTSTADDRS: {
2785 		struct pfioc_table *io = (struct pfioc_table *)addr;
2786 		struct pfr_addr *pfras;
2787 		size_t totlen;
2788 
2789 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2790 			error = ENODEV;
2791 			break;
2792 		}
2793 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2794 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2795 		error = copyin(io->pfrio_buffer, pfras, totlen);
2796 		if (error) {
2797 			free(pfras, M_TEMP);
2798 			break;
2799 		}
2800 		PF_RULES_RLOCK();
2801 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
2802 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2803 		    PFR_FLAG_USERIOCTL);
2804 		PF_RULES_RUNLOCK();
2805 		if (error == 0)
2806 			error = copyout(pfras, io->pfrio_buffer, totlen);
2807 		free(pfras, M_TEMP);
2808 		break;
2809 	}
2810 
2811 	case DIOCRINADEFINE: {
2812 		struct pfioc_table *io = (struct pfioc_table *)addr;
2813 		struct pfr_addr *pfras;
2814 		size_t totlen;
2815 
2816 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2817 			error = ENODEV;
2818 			break;
2819 		}
2820 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2821 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2822 		error = copyin(io->pfrio_buffer, pfras, totlen);
2823 		if (error) {
2824 			free(pfras, M_TEMP);
2825 			break;
2826 		}
2827 		PF_RULES_WLOCK();
2828 		error = pfr_ina_define(&io->pfrio_table, pfras,
2829 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2830 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2831 		PF_RULES_WUNLOCK();
2832 		free(pfras, M_TEMP);
2833 		break;
2834 	}
2835 
2836 	case DIOCOSFPADD: {
2837 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2838 		PF_RULES_WLOCK();
2839 		error = pf_osfp_add(io);
2840 		PF_RULES_WUNLOCK();
2841 		break;
2842 	}
2843 
2844 	case DIOCOSFPGET: {
2845 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2846 		PF_RULES_RLOCK();
2847 		error = pf_osfp_get(io);
2848 		PF_RULES_RUNLOCK();
2849 		break;
2850 	}
2851 
2852 	case DIOCXBEGIN: {
2853 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2854 		struct pfioc_trans_e	*ioes, *ioe;
2855 		size_t			 totlen;
2856 		int			 i;
2857 
2858 		if (io->esize != sizeof(*ioe)) {
2859 			error = ENODEV;
2860 			break;
2861 		}
2862 		totlen = sizeof(struct pfioc_trans_e) * io->size;
2863 		ioes = malloc(totlen, M_TEMP, M_WAITOK);
2864 		error = copyin(io->array, ioes, totlen);
2865 		if (error) {
2866 			free(ioes, M_TEMP);
2867 			break;
2868 		}
2869 		PF_RULES_WLOCK();
2870 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2871 			switch (ioe->rs_num) {
2872 #ifdef ALTQ
2873 			case PF_RULESET_ALTQ:
2874 				if (ioe->anchor[0]) {
2875 					PF_RULES_WUNLOCK();
2876 					free(ioes, M_TEMP);
2877 					error = EINVAL;
2878 					goto fail;
2879 				}
2880 				if ((error = pf_begin_altq(&ioe->ticket))) {
2881 					PF_RULES_WUNLOCK();
2882 					free(ioes, M_TEMP);
2883 					goto fail;
2884 				}
2885 				break;
2886 #endif /* ALTQ */
2887 			case PF_RULESET_TABLE:
2888 			    {
2889 				struct pfr_table table;
2890 
2891 				bzero(&table, sizeof(table));
2892 				strlcpy(table.pfrt_anchor, ioe->anchor,
2893 				    sizeof(table.pfrt_anchor));
2894 				if ((error = pfr_ina_begin(&table,
2895 				    &ioe->ticket, NULL, 0))) {
2896 					PF_RULES_WUNLOCK();
2897 					free(ioes, M_TEMP);
2898 					goto fail;
2899 				}
2900 				break;
2901 			    }
2902 			default:
2903 				if ((error = pf_begin_rules(&ioe->ticket,
2904 				    ioe->rs_num, ioe->anchor))) {
2905 					PF_RULES_WUNLOCK();
2906 					free(ioes, M_TEMP);
2907 					goto fail;
2908 				}
2909 				break;
2910 			}
2911 		}
2912 		PF_RULES_WUNLOCK();
2913 		error = copyout(ioes, io->array, totlen);
2914 		free(ioes, M_TEMP);
2915 		break;
2916 	}
2917 
2918 	case DIOCXROLLBACK: {
2919 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2920 		struct pfioc_trans_e	*ioe, *ioes;
2921 		size_t			 totlen;
2922 		int			 i;
2923 
2924 		if (io->esize != sizeof(*ioe)) {
2925 			error = ENODEV;
2926 			break;
2927 		}
2928 		totlen = sizeof(struct pfioc_trans_e) * io->size;
2929 		ioes = malloc(totlen, M_TEMP, M_WAITOK);
2930 		error = copyin(io->array, ioes, totlen);
2931 		if (error) {
2932 			free(ioes, M_TEMP);
2933 			break;
2934 		}
2935 		PF_RULES_WLOCK();
2936 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2937 			switch (ioe->rs_num) {
2938 #ifdef ALTQ
2939 			case PF_RULESET_ALTQ:
2940 				if (ioe->anchor[0]) {
2941 					PF_RULES_WUNLOCK();
2942 					free(ioes, M_TEMP);
2943 					error = EINVAL;
2944 					goto fail;
2945 				}
2946 				if ((error = pf_rollback_altq(ioe->ticket))) {
2947 					PF_RULES_WUNLOCK();
2948 					free(ioes, M_TEMP);
2949 					goto fail; /* really bad */
2950 				}
2951 				break;
2952 #endif /* ALTQ */
2953 			case PF_RULESET_TABLE:
2954 			    {
2955 				struct pfr_table table;
2956 
2957 				bzero(&table, sizeof(table));
2958 				strlcpy(table.pfrt_anchor, ioe->anchor,
2959 				    sizeof(table.pfrt_anchor));
2960 				if ((error = pfr_ina_rollback(&table,
2961 				    ioe->ticket, NULL, 0))) {
2962 					PF_RULES_WUNLOCK();
2963 					free(ioes, M_TEMP);
2964 					goto fail; /* really bad */
2965 				}
2966 				break;
2967 			    }
2968 			default:
2969 				if ((error = pf_rollback_rules(ioe->ticket,
2970 				    ioe->rs_num, ioe->anchor))) {
2971 					PF_RULES_WUNLOCK();
2972 					free(ioes, M_TEMP);
2973 					goto fail; /* really bad */
2974 				}
2975 				break;
2976 			}
2977 		}
2978 		PF_RULES_WUNLOCK();
2979 		free(ioes, M_TEMP);
2980 		break;
2981 	}
2982 
2983 	case DIOCXCOMMIT: {
2984 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2985 		struct pfioc_trans_e	*ioe, *ioes;
2986 		struct pf_ruleset	*rs;
2987 		size_t			 totlen;
2988 		int			 i;
2989 
2990 		if (io->esize != sizeof(*ioe)) {
2991 			error = ENODEV;
2992 			break;
2993 		}
2994 		totlen = sizeof(struct pfioc_trans_e) * io->size;
2995 		ioes = malloc(totlen, M_TEMP, M_WAITOK);
2996 		error = copyin(io->array, ioes, totlen);
2997 		if (error) {
2998 			free(ioes, M_TEMP);
2999 			break;
3000 		}
3001 		PF_RULES_WLOCK();
3002 		/* First makes sure everything will succeed. */
3003 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3004 			switch (ioe->rs_num) {
3005 #ifdef ALTQ
3006 			case PF_RULESET_ALTQ:
3007 				if (ioe->anchor[0]) {
3008 					PF_RULES_WUNLOCK();
3009 					free(ioes, M_TEMP);
3010 					error = EINVAL;
3011 					goto fail;
3012 				}
3013 				if (!V_altqs_inactive_open || ioe->ticket !=
3014 				    V_ticket_altqs_inactive) {
3015 					PF_RULES_WUNLOCK();
3016 					free(ioes, M_TEMP);
3017 					error = EBUSY;
3018 					goto fail;
3019 				}
3020 				break;
3021 #endif /* ALTQ */
3022 			case PF_RULESET_TABLE:
3023 				rs = pf_find_ruleset(ioe->anchor);
3024 				if (rs == NULL || !rs->topen || ioe->ticket !=
3025 				    rs->tticket) {
3026 					PF_RULES_WUNLOCK();
3027 					free(ioes, M_TEMP);
3028 					error = EBUSY;
3029 					goto fail;
3030 				}
3031 				break;
3032 			default:
3033 				if (ioe->rs_num < 0 || ioe->rs_num >=
3034 				    PF_RULESET_MAX) {
3035 					PF_RULES_WUNLOCK();
3036 					free(ioes, M_TEMP);
3037 					error = EINVAL;
3038 					goto fail;
3039 				}
3040 				rs = pf_find_ruleset(ioe->anchor);
3041 				if (rs == NULL ||
3042 				    !rs->rules[ioe->rs_num].inactive.open ||
3043 				    rs->rules[ioe->rs_num].inactive.ticket !=
3044 				    ioe->ticket) {
3045 					PF_RULES_WUNLOCK();
3046 					free(ioes, M_TEMP);
3047 					error = EBUSY;
3048 					goto fail;
3049 				}
3050 				break;
3051 			}
3052 		}
3053 		/* Now do the commit - no errors should happen here. */
3054 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3055 			switch (ioe->rs_num) {
3056 #ifdef ALTQ
3057 			case PF_RULESET_ALTQ:
3058 				if ((error = pf_commit_altq(ioe->ticket))) {
3059 					PF_RULES_WUNLOCK();
3060 					free(ioes, M_TEMP);
3061 					goto fail; /* really bad */
3062 				}
3063 				break;
3064 #endif /* ALTQ */
3065 			case PF_RULESET_TABLE:
3066 			    {
3067 				struct pfr_table table;
3068 
3069 				bzero(&table, sizeof(table));
3070 				strlcpy(table.pfrt_anchor, ioe->anchor,
3071 				    sizeof(table.pfrt_anchor));
3072 				if ((error = pfr_ina_commit(&table,
3073 				    ioe->ticket, NULL, NULL, 0))) {
3074 					PF_RULES_WUNLOCK();
3075 					free(ioes, M_TEMP);
3076 					goto fail; /* really bad */
3077 				}
3078 				break;
3079 			    }
3080 			default:
3081 				if ((error = pf_commit_rules(ioe->ticket,
3082 				    ioe->rs_num, ioe->anchor))) {
3083 					PF_RULES_WUNLOCK();
3084 					free(ioes, M_TEMP);
3085 					goto fail; /* really bad */
3086 				}
3087 				break;
3088 			}
3089 		}
3090 		PF_RULES_WUNLOCK();
3091 		free(ioes, M_TEMP);
3092 		break;
3093 	}
3094 
3095 	case DIOCGETSRCNODES: {
3096 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
3097 		struct pf_srchash	*sh;
3098 		struct pf_src_node	*n, *p, *pstore;
3099 		uint32_t		 i, nr = 0;
3100 
3101 		if (psn->psn_len == 0) {
3102 			for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3103 			    i++, sh++) {
3104 				PF_HASHROW_LOCK(sh);
3105 				LIST_FOREACH(n, &sh->nodes, entry)
3106 					nr++;
3107 				PF_HASHROW_UNLOCK(sh);
3108 			}
3109 			psn->psn_len = sizeof(struct pf_src_node) * nr;
3110 			break;
3111 		}
3112 
3113 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK);
3114 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3115 		    i++, sh++) {
3116 		    PF_HASHROW_LOCK(sh);
3117 		    LIST_FOREACH(n, &sh->nodes, entry) {
3118 			int	secs = time_uptime, diff;
3119 
3120 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
3121 				break;
3122 
3123 			bcopy(n, p, sizeof(struct pf_src_node));
3124 			if (n->rule.ptr != NULL)
3125 				p->rule.nr = n->rule.ptr->nr;
3126 			p->creation = secs - p->creation;
3127 			if (p->expire > secs)
3128 				p->expire -= secs;
3129 			else
3130 				p->expire = 0;
3131 
3132 			/* Adjust the connection rate estimate. */
3133 			diff = secs - n->conn_rate.last;
3134 			if (diff >= n->conn_rate.seconds)
3135 				p->conn_rate.count = 0;
3136 			else
3137 				p->conn_rate.count -=
3138 				    n->conn_rate.count * diff /
3139 				    n->conn_rate.seconds;
3140 			p++;
3141 			nr++;
3142 		    }
3143 		    PF_HASHROW_UNLOCK(sh);
3144 		}
3145 		error = copyout(pstore, psn->psn_src_nodes,
3146 		    sizeof(struct pf_src_node) * nr);
3147 		if (error) {
3148 			free(pstore, M_TEMP);
3149 			break;
3150 		}
3151 		psn->psn_len = sizeof(struct pf_src_node) * nr;
3152 		free(pstore, M_TEMP);
3153 		break;
3154 	}
3155 
3156 	case DIOCCLRSRCNODES: {
3157 
3158 		pf_clear_srcnodes(NULL);
3159 		pf_purge_expired_src_nodes();
3160 		V_pf_status.src_nodes = 0;
3161 		break;
3162 	}
3163 
3164 	case DIOCKILLSRCNODES:
3165 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
3166 		break;
3167 
3168 	case DIOCSETHOSTID: {
3169 		u_int32_t	*hostid = (u_int32_t *)addr;
3170 
3171 		PF_RULES_WLOCK();
3172 		if (*hostid == 0)
3173 			V_pf_status.hostid = arc4random();
3174 		else
3175 			V_pf_status.hostid = *hostid;
3176 		PF_RULES_WUNLOCK();
3177 		break;
3178 	}
3179 
3180 	case DIOCOSFPFLUSH:
3181 		PF_RULES_WLOCK();
3182 		pf_osfp_flush();
3183 		PF_RULES_WUNLOCK();
3184 		break;
3185 
3186 	case DIOCIGETIFACES: {
3187 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3188 		struct pfi_kif *ifstore;
3189 		size_t bufsiz;
3190 
3191 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
3192 			error = ENODEV;
3193 			break;
3194 		}
3195 
3196 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
3197 		ifstore = malloc(bufsiz, M_TEMP, M_WAITOK);
3198 		PF_RULES_RLOCK();
3199 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
3200 		PF_RULES_RUNLOCK();
3201 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
3202 		free(ifstore, M_TEMP);
3203 		break;
3204 	}
3205 
3206 	case DIOCSETIFFLAG: {
3207 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3208 
3209 		PF_RULES_WLOCK();
3210 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3211 		PF_RULES_WUNLOCK();
3212 		break;
3213 	}
3214 
3215 	case DIOCCLRIFFLAG: {
3216 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3217 
3218 		PF_RULES_WLOCK();
3219 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3220 		PF_RULES_WUNLOCK();
3221 		break;
3222 	}
3223 
3224 	default:
3225 		error = ENODEV;
3226 		break;
3227 	}
3228 fail:
3229 	CURVNET_RESTORE();
3230 
3231 	return (error);
3232 }
3233 
3234 void
3235 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
3236 {
3237 	bzero(sp, sizeof(struct pfsync_state));
3238 
3239 	/* copy from state key */
3240 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
3241 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
3242 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
3243 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
3244 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
3245 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
3246 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
3247 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
3248 	sp->proto = st->key[PF_SK_WIRE]->proto;
3249 	sp->af = st->key[PF_SK_WIRE]->af;
3250 
3251 	/* copy from state */
3252 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
3253 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
3254 	sp->creation = htonl(time_uptime - st->creation);
3255 	sp->expire = pf_state_expires(st);
3256 	if (sp->expire <= time_uptime)
3257 		sp->expire = htonl(0);
3258 	else
3259 		sp->expire = htonl(sp->expire - time_uptime);
3260 
3261 	sp->direction = st->direction;
3262 	sp->log = st->log;
3263 	sp->timeout = st->timeout;
3264 	sp->state_flags = st->state_flags;
3265 	if (st->src_node)
3266 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
3267 	if (st->nat_src_node)
3268 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
3269 
3270 	sp->id = st->id;
3271 	sp->creatorid = st->creatorid;
3272 	pf_state_peer_hton(&st->src, &sp->src);
3273 	pf_state_peer_hton(&st->dst, &sp->dst);
3274 
3275 	if (st->rule.ptr == NULL)
3276 		sp->rule = htonl(-1);
3277 	else
3278 		sp->rule = htonl(st->rule.ptr->nr);
3279 	if (st->anchor.ptr == NULL)
3280 		sp->anchor = htonl(-1);
3281 	else
3282 		sp->anchor = htonl(st->anchor.ptr->nr);
3283 	if (st->nat_rule.ptr == NULL)
3284 		sp->nat_rule = htonl(-1);
3285 	else
3286 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
3287 
3288 	pf_state_counter_hton(st->packets[0], sp->packets[0]);
3289 	pf_state_counter_hton(st->packets[1], sp->packets[1]);
3290 	pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
3291 	pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
3292 
3293 }
3294 
3295 static void
3296 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
3297 {
3298 	struct pfr_ktable *kt;
3299 
3300 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
3301 
3302 	kt = aw->p.tbl;
3303 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
3304 		kt = kt->pfrkt_root;
3305 	aw->p.tbl = NULL;
3306 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
3307 		kt->pfrkt_cnt : -1;
3308 }
3309 
3310 /*
3311  * XXX - Check for version missmatch!!!
3312  */
3313 static void
3314 pf_clear_states(void)
3315 {
3316 	struct pf_state	*s;
3317 	u_int i;
3318 
3319 	for (i = 0; i <= pf_hashmask; i++) {
3320 		struct pf_idhash *ih = &V_pf_idhash[i];
3321 relock:
3322 		PF_HASHROW_LOCK(ih);
3323 		LIST_FOREACH(s, &ih->states, entry) {
3324 			s->timeout = PFTM_PURGE;
3325 			/* Don't send out individual delete messages. */
3326 			s->sync_state = PFSTATE_NOSYNC;
3327 			pf_unlink_state(s, PF_ENTER_LOCKED);
3328 			goto relock;
3329 		}
3330 		PF_HASHROW_UNLOCK(ih);
3331 	}
3332 }
3333 
3334 static int
3335 pf_clear_tables(void)
3336 {
3337 	struct pfioc_table io;
3338 	int error;
3339 
3340 	bzero(&io, sizeof(io));
3341 
3342 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
3343 	    io.pfrio_flags);
3344 
3345 	return (error);
3346 }
3347 
3348 static void
3349 pf_clear_srcnodes(struct pf_src_node *n)
3350 {
3351 	struct pf_state *s;
3352 	int i;
3353 
3354 	for (i = 0; i <= pf_hashmask; i++) {
3355 		struct pf_idhash *ih = &V_pf_idhash[i];
3356 
3357 		PF_HASHROW_LOCK(ih);
3358 		LIST_FOREACH(s, &ih->states, entry) {
3359 			if (n == NULL || n == s->src_node)
3360 				s->src_node = NULL;
3361 			if (n == NULL || n == s->nat_src_node)
3362 				s->nat_src_node = NULL;
3363 		}
3364 		PF_HASHROW_UNLOCK(ih);
3365 	}
3366 
3367 	if (n == NULL) {
3368 		struct pf_srchash *sh;
3369 
3370 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3371 		    i++, sh++) {
3372 			PF_HASHROW_LOCK(sh);
3373 			LIST_FOREACH(n, &sh->nodes, entry) {
3374 				n->expire = 1;
3375 				n->states = 0;
3376 			}
3377 			PF_HASHROW_UNLOCK(sh);
3378 		}
3379 	} else {
3380 		/* XXX: hash slot should already be locked here. */
3381 		n->expire = 1;
3382 		n->states = 0;
3383 	}
3384 }
3385 
3386 static void
3387 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
3388 {
3389 	struct pf_src_node_list	 kill;
3390 
3391 	LIST_INIT(&kill);
3392 	for (int i = 0; i <= pf_srchashmask; i++) {
3393 		struct pf_srchash *sh = &V_pf_srchash[i];
3394 		struct pf_src_node *sn, *tmp;
3395 
3396 		PF_HASHROW_LOCK(sh);
3397 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
3398 			if (PF_MATCHA(psnk->psnk_src.neg,
3399 			      &psnk->psnk_src.addr.v.a.addr,
3400 			      &psnk->psnk_src.addr.v.a.mask,
3401 			      &sn->addr, sn->af) &&
3402 			    PF_MATCHA(psnk->psnk_dst.neg,
3403 			      &psnk->psnk_dst.addr.v.a.addr,
3404 			      &psnk->psnk_dst.addr.v.a.mask,
3405 			      &sn->raddr, sn->af)) {
3406 				pf_unlink_src_node_locked(sn);
3407 				LIST_INSERT_HEAD(&kill, sn, entry);
3408 				sn->expire = 1;
3409 			}
3410 		PF_HASHROW_UNLOCK(sh);
3411 	}
3412 
3413 	for (int i = 0; i <= pf_hashmask; i++) {
3414 		struct pf_idhash *ih = &V_pf_idhash[i];
3415 		struct pf_state *s;
3416 
3417 		PF_HASHROW_LOCK(ih);
3418 		LIST_FOREACH(s, &ih->states, entry) {
3419 			if (s->src_node && s->src_node->expire == 1) {
3420 #ifdef INVARIANTS
3421 				s->src_node->states--;
3422 #endif
3423 				s->src_node = NULL;
3424 			}
3425 			if (s->nat_src_node && s->nat_src_node->expire == 1) {
3426 #ifdef INVARIANTS
3427 				s->nat_src_node->states--;
3428 #endif
3429 				s->nat_src_node = NULL;
3430 			}
3431 		}
3432 		PF_HASHROW_UNLOCK(ih);
3433 	}
3434 
3435 	psnk->psnk_killed = pf_free_src_nodes(&kill);
3436 }
3437 
3438 /*
3439  * XXX - Check for version missmatch!!!
3440  */
3441 
3442 /*
3443  * Duplicate pfctl -Fa operation to get rid of as much as we can.
3444  */
3445 static int
3446 shutdown_pf(void)
3447 {
3448 	int error = 0;
3449 	u_int32_t t[5];
3450 	char nn = '\0';
3451 
3452 	V_pf_status.running = 0;
3453 
3454 	counter_u64_free(V_pf_default_rule.states_cur);
3455 	counter_u64_free(V_pf_default_rule.states_tot);
3456 	counter_u64_free(V_pf_default_rule.src_nodes);
3457 
3458 	do {
3459 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
3460 		    != 0) {
3461 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
3462 			break;
3463 		}
3464 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
3465 		    != 0) {
3466 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
3467 			break;		/* XXX: rollback? */
3468 		}
3469 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
3470 		    != 0) {
3471 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
3472 			break;		/* XXX: rollback? */
3473 		}
3474 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
3475 		    != 0) {
3476 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
3477 			break;		/* XXX: rollback? */
3478 		}
3479 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
3480 		    != 0) {
3481 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
3482 			break;		/* XXX: rollback? */
3483 		}
3484 
3485 		/* XXX: these should always succeed here */
3486 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
3487 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
3488 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
3489 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
3490 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
3491 
3492 		if ((error = pf_clear_tables()) != 0)
3493 			break;
3494 
3495 #ifdef ALTQ
3496 		if ((error = pf_begin_altq(&t[0])) != 0) {
3497 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
3498 			break;
3499 		}
3500 		pf_commit_altq(t[0]);
3501 #endif
3502 
3503 		pf_clear_states();
3504 
3505 		pf_clear_srcnodes(NULL);
3506 
3507 		/* status does not use malloced mem so no need to cleanup */
3508 		/* fingerprints and interfaces have thier own cleanup code */
3509 	} while(0);
3510 
3511 	return (error);
3512 }
3513 
3514 #ifdef INET
3515 static int
3516 pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3517     struct inpcb *inp)
3518 {
3519 	int chk;
3520 
3521 	chk = pf_test(PF_IN, ifp, m, inp);
3522 	if (chk && *m) {
3523 		m_freem(*m);
3524 		*m = NULL;
3525 	}
3526 
3527 	return (chk);
3528 }
3529 
3530 static int
3531 pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3532     struct inpcb *inp)
3533 {
3534 	int chk;
3535 
3536 	/* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3537 	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3538 		in_delayed_cksum(*m);
3539 		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3540 	}
3541 
3542 	chk = pf_test(PF_OUT, ifp, m, inp);
3543 	if (chk && *m) {
3544 		m_freem(*m);
3545 		*m = NULL;
3546 	}
3547 
3548 	return (chk);
3549 }
3550 #endif
3551 
3552 #ifdef INET6
3553 static int
3554 pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3555     struct inpcb *inp)
3556 {
3557 	int chk;
3558 
3559 	/*
3560 	 * In case of loopback traffic IPv6 uses the real interface in
3561 	 * order to support scoped addresses. In order to support stateful
3562 	 * filtering we have change this to lo0 as it is the case in IPv4.
3563 	 */
3564 	CURVNET_SET(ifp->if_vnet);
3565 	chk = pf_test6(PF_IN, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
3566 	CURVNET_RESTORE();
3567 	if (chk && *m) {
3568 		m_freem(*m);
3569 		*m = NULL;
3570 	}
3571 	return chk;
3572 }
3573 
3574 static int
3575 pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3576     struct inpcb *inp)
3577 {
3578 	int chk;
3579 
3580 	/* We need a proper CSUM before we start (s. OpenBSD ip_output) */
3581 	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3582 #ifdef INET
3583 		/* XXX-BZ copy&paste error from r126261? */
3584 		in_delayed_cksum(*m);
3585 #endif
3586 		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3587 	}
3588 	CURVNET_SET(ifp->if_vnet);
3589 	chk = pf_test6(PF_OUT, ifp, m, inp);
3590 	CURVNET_RESTORE();
3591 	if (chk && *m) {
3592 		m_freem(*m);
3593 		*m = NULL;
3594 	}
3595 	return chk;
3596 }
3597 #endif /* INET6 */
3598 
3599 static int
3600 hook_pf(void)
3601 {
3602 #ifdef INET
3603 	struct pfil_head *pfh_inet;
3604 #endif
3605 #ifdef INET6
3606 	struct pfil_head *pfh_inet6;
3607 #endif
3608 
3609 	if (V_pf_pfil_hooked)
3610 		return (0);
3611 
3612 #ifdef INET
3613 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3614 	if (pfh_inet == NULL)
3615 		return (ESRCH); /* XXX */
3616 	pfil_add_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet);
3617 	pfil_add_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet);
3618 #endif
3619 #ifdef INET6
3620 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3621 	if (pfh_inet6 == NULL) {
3622 #ifdef INET
3623 		pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3624 		    pfh_inet);
3625 		pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3626 		    pfh_inet);
3627 #endif
3628 		return (ESRCH); /* XXX */
3629 	}
3630 	pfil_add_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet6);
3631 	pfil_add_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet6);
3632 #endif
3633 
3634 	V_pf_pfil_hooked = 1;
3635 	return (0);
3636 }
3637 
3638 static int
3639 dehook_pf(void)
3640 {
3641 #ifdef INET
3642 	struct pfil_head *pfh_inet;
3643 #endif
3644 #ifdef INET6
3645 	struct pfil_head *pfh_inet6;
3646 #endif
3647 
3648 	if (V_pf_pfil_hooked == 0)
3649 		return (0);
3650 
3651 #ifdef INET
3652 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3653 	if (pfh_inet == NULL)
3654 		return (ESRCH); /* XXX */
3655 	pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3656 	    pfh_inet);
3657 	pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3658 	    pfh_inet);
3659 #endif
3660 #ifdef INET6
3661 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3662 	if (pfh_inet6 == NULL)
3663 		return (ESRCH); /* XXX */
3664 	pfil_remove_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK,
3665 	    pfh_inet6);
3666 	pfil_remove_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK,
3667 	    pfh_inet6);
3668 #endif
3669 
3670 	V_pf_pfil_hooked = 0;
3671 	return (0);
3672 }
3673 
3674 static int
3675 pf_load(void)
3676 {
3677 	int error;
3678 
3679 	VNET_ITERATOR_DECL(vnet_iter);
3680 
3681 	VNET_LIST_RLOCK();
3682 	VNET_FOREACH(vnet_iter) {
3683 		CURVNET_SET(vnet_iter);
3684 		V_pf_pfil_hooked = 0;
3685 		V_pf_end_threads = 0;
3686 		TAILQ_INIT(&V_pf_tags);
3687 		TAILQ_INIT(&V_pf_qids);
3688 		CURVNET_RESTORE();
3689 	}
3690 	VNET_LIST_RUNLOCK();
3691 
3692 	rw_init(&pf_rules_lock, "pf rulesets");
3693 
3694 	pf_dev = make_dev(&pf_cdevsw, 0, 0, 0, 0600, PF_NAME);
3695 	if ((error = pfattach()) != 0)
3696 		return (error);
3697 
3698 	return (0);
3699 }
3700 
3701 static int
3702 pf_unload(void)
3703 {
3704 	int error = 0;
3705 
3706 	PF_RULES_WLOCK();
3707 	V_pf_status.running = 0;
3708 	PF_RULES_WUNLOCK();
3709 	swi_remove(V_pf_swi_cookie);
3710 	error = dehook_pf();
3711 	if (error) {
3712 		/*
3713 		 * Should not happen!
3714 		 * XXX Due to error code ESRCH, kldunload will show
3715 		 * a message like 'No such process'.
3716 		 */
3717 		printf("%s : pfil unregisteration fail\n", __FUNCTION__);
3718 		return error;
3719 	}
3720 	PF_RULES_WLOCK();
3721 	shutdown_pf();
3722 	V_pf_end_threads = 1;
3723 	while (V_pf_end_threads < 2) {
3724 		wakeup_one(pf_purge_thread);
3725 		rw_sleep(pf_purge_thread, &pf_rules_lock, 0, "pftmo", 0);
3726 	}
3727 	pf_normalize_cleanup();
3728 	pfi_cleanup();
3729 	pfr_cleanup();
3730 	pf_osfp_flush();
3731 	pf_cleanup();
3732 	if (IS_DEFAULT_VNET(curvnet))
3733 		pf_mtag_cleanup();
3734 	PF_RULES_WUNLOCK();
3735 	destroy_dev(pf_dev);
3736 	rw_destroy(&pf_rules_lock);
3737 
3738 	return (error);
3739 }
3740 
3741 static int
3742 pf_modevent(module_t mod, int type, void *data)
3743 {
3744 	int error = 0;
3745 
3746 	switch(type) {
3747 	case MOD_LOAD:
3748 		error = pf_load();
3749 		break;
3750 	case MOD_QUIESCE:
3751 		/*
3752 		 * Module should not be unloaded due to race conditions.
3753 		 */
3754 		error = EBUSY;
3755 		break;
3756 	case MOD_UNLOAD:
3757 		error = pf_unload();
3758 		break;
3759 	default:
3760 		error = EINVAL;
3761 		break;
3762 	}
3763 
3764 	return (error);
3765 }
3766 
3767 static moduledata_t pf_mod = {
3768 	"pf",
3769 	pf_modevent,
3770 	0
3771 };
3772 
3773 DECLARE_MODULE(pf, pf_mod, SI_SUB_PSEUDO, SI_ORDER_FIRST);
3774 MODULE_VERSION(pf, PF_MODVER);
3775