xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision 28f42739a547ffe0b5dfaaf9f49fb4c4813aa232)
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 <netinet6/ip6_var.h>
80 #include <netinet/ip_icmp.h>
81 
82 #ifdef INET6
83 #include <netinet/ip6.h>
84 #endif /* INET6 */
85 
86 #ifdef ALTQ
87 #include <altq/altq.h>
88 #endif
89 
90 static int		 pf_vnet_init(void);
91 static struct pf_pool	*pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t,
92 			    u_int8_t, u_int8_t, u_int8_t);
93 
94 static void		 pf_mv_pool(struct pf_palist *, struct pf_palist *);
95 static void		 pf_empty_pool(struct pf_palist *);
96 static int		 pfioctl(struct cdev *, u_long, caddr_t, int,
97 			    struct thread *);
98 #ifdef ALTQ
99 static int		 pf_begin_altq(u_int32_t *);
100 static int		 pf_rollback_altq(u_int32_t);
101 static int		 pf_commit_altq(u_int32_t);
102 static int		 pf_enable_altq(struct pf_altq *);
103 static int		 pf_disable_altq(struct pf_altq *);
104 static u_int32_t	 pf_qname2qid(char *);
105 static void		 pf_qid_unref(u_int32_t);
106 #endif /* ALTQ */
107 static int		 pf_begin_rules(u_int32_t *, int, const char *);
108 static int		 pf_rollback_rules(u_int32_t, int, char *);
109 static int		 pf_setup_pfsync_matching(struct pf_ruleset *);
110 static void		 pf_hash_rule(MD5_CTX *, struct pf_rule *);
111 static void		 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
112 static int		 pf_commit_rules(u_int32_t, int, char *);
113 static int		 pf_addr_setup(struct pf_ruleset *,
114 			    struct pf_addr_wrap *, sa_family_t);
115 static void		 pf_addr_copyout(struct pf_addr_wrap *);
116 
117 VNET_DEFINE(struct pf_rule,	pf_default_rule);
118 
119 #ifdef ALTQ
120 static VNET_DEFINE(int,		pf_altq_running);
121 #define	V_pf_altq_running	VNET(pf_altq_running)
122 #endif
123 
124 #define	TAGID_MAX	 50000
125 struct pf_tagname {
126 	TAILQ_ENTRY(pf_tagname)	entries;
127 	char			name[PF_TAG_NAME_SIZE];
128 	uint16_t		tag;
129 	int			ref;
130 };
131 
132 TAILQ_HEAD(pf_tags, pf_tagname);
133 #define	V_pf_tags		VNET(pf_tags)
134 VNET_DEFINE(struct pf_tags, pf_tags);
135 #define	V_pf_qids		VNET(pf_qids)
136 VNET_DEFINE(struct pf_tags, pf_qids);
137 static MALLOC_DEFINE(M_PFTAG, "pf_tag", "pf(4) tag names");
138 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
139 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
140 
141 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
142 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
143 #endif
144 
145 static u_int16_t	 tagname2tag(struct pf_tags *, char *);
146 static u_int16_t	 pf_tagname2tag(char *);
147 static void		 tag_unref(struct pf_tags *, u_int16_t);
148 
149 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
150 
151 struct cdev *pf_dev;
152 
153 /*
154  * XXX - These are new and need to be checked when moveing to a new version
155  */
156 static void		 pf_clear_states(void);
157 static int		 pf_clear_tables(void);
158 static void		 pf_clear_srcnodes(struct pf_src_node *);
159 static void		 pf_kill_srcnodes(struct pfioc_src_node_kill *);
160 static void		 pf_tbladdr_copyout(struct pf_addr_wrap *);
161 
162 /*
163  * Wrapper functions for pfil(9) hooks
164  */
165 #ifdef INET
166 static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp,
167     int dir, struct inpcb *inp);
168 static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp,
169     int dir, struct inpcb *inp);
170 #endif
171 #ifdef INET6
172 static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp,
173     int dir, struct inpcb *inp);
174 static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp,
175     int dir, struct inpcb *inp);
176 #endif
177 
178 static int		hook_pf(void);
179 static int		dehook_pf(void);
180 static int		shutdown_pf(void);
181 static int		pf_load(void);
182 static int		pf_unload(void);
183 
184 static struct cdevsw pf_cdevsw = {
185 	.d_ioctl =	pfioctl,
186 	.d_name =	PF_NAME,
187 	.d_version =	D_VERSION,
188 };
189 
190 static volatile VNET_DEFINE(int, pf_pfil_hooked);
191 #define V_pf_pfil_hooked	VNET(pf_pfil_hooked)
192 
193 struct rwlock			pf_rules_lock;
194 struct sx			pf_ioctl_lock;
195 
196 /* pfsync */
197 pfsync_state_import_t 		*pfsync_state_import_ptr = NULL;
198 pfsync_insert_state_t		*pfsync_insert_state_ptr = NULL;
199 pfsync_update_state_t		*pfsync_update_state_ptr = NULL;
200 pfsync_delete_state_t		*pfsync_delete_state_ptr = NULL;
201 pfsync_clear_states_t		*pfsync_clear_states_ptr = NULL;
202 pfsync_defer_t			*pfsync_defer_ptr = NULL;
203 /* pflog */
204 pflog_packet_t			*pflog_packet_ptr = NULL;
205 
206 static int
207 pf_vnet_init(void)
208 {
209 	u_int32_t *my_timeout = V_pf_default_rule.timeout;
210 	int error;
211 
212 	if (IS_DEFAULT_VNET(curvnet))
213 		pf_mtag_initialize();
214 	TAILQ_INIT(&V_pf_tags);
215 	TAILQ_INIT(&V_pf_qids);
216 
217 	pf_vnet_initialize();
218 	pfr_initialize();
219 	pfi_vnet_initialize();
220 	pf_vnet_normalize_init();
221 
222 	V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
223 	V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
224 
225 	RB_INIT(&V_pf_anchors);
226 	pf_init_ruleset(&pf_main_ruleset);
227 
228 	/* default rule should never be garbage collected */
229 	V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
230 #ifdef PF_DEFAULT_TO_DROP
231 	V_pf_default_rule.action = PF_DROP;
232 #else
233 	V_pf_default_rule.action = PF_PASS;
234 #endif
235 	V_pf_default_rule.nr = -1;
236 	V_pf_default_rule.rtableid = -1;
237 
238 	V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
239 	V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
240 	V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
241 
242 	/* initialize default timeouts */
243 	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
244 	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
245 	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
246 	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
247 	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
248 	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
249 	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
250 	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
251 	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
252 	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
253 	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
254 	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
255 	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
256 	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
257 	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
258 	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
259 	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
260 	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
261 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
262 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
263 
264 	bzero(&V_pf_status, sizeof(V_pf_status));
265 	V_pf_status.debug = PF_DEBUG_URGENT;
266 
267 	V_pf_pfil_hooked = 0;
268 
269 	/* XXX do our best to avoid a conflict */
270 	V_pf_status.hostid = arc4random();
271 
272 	for (int i = 0; i < PFRES_MAX; i++)
273 		V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
274 	for (int i = 0; i < LCNT_MAX; i++)
275 		V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
276 	for (int i = 0; i < FCNT_MAX; i++)
277 		V_pf_status.fcounters[i] = counter_u64_alloc(M_WAITOK);
278 	for (int i = 0; i < SCNT_MAX; i++)
279 		V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
280 
281 	if (IS_DEFAULT_VNET(curvnet)) {
282 	    if ((error = kproc_create(pf_purge_thread, curvnet, NULL, 0, 0,
283 		"pf purge")) != 0) {
284 		    /* XXXGL: leaked all above. */
285 		    return (error);
286 	    }
287 	}
288 	if ((error = swi_add(NULL, "pf send", pf_intr, curvnet, SWI_NET,
289 	    INTR_MPSAFE, &V_pf_swi_cookie)) != 0)
290 		/* XXXGL: leaked all above. */
291 		return (error);
292 
293 	return (0);
294 }
295 
296 static struct pf_pool *
297 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
298     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
299     u_int8_t check_ticket)
300 {
301 	struct pf_ruleset	*ruleset;
302 	struct pf_rule		*rule;
303 	int			 rs_num;
304 
305 	ruleset = pf_find_ruleset(anchor);
306 	if (ruleset == NULL)
307 		return (NULL);
308 	rs_num = pf_get_ruleset_number(rule_action);
309 	if (rs_num >= PF_RULESET_MAX)
310 		return (NULL);
311 	if (active) {
312 		if (check_ticket && ticket !=
313 		    ruleset->rules[rs_num].active.ticket)
314 			return (NULL);
315 		if (r_last)
316 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
317 			    pf_rulequeue);
318 		else
319 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
320 	} else {
321 		if (check_ticket && ticket !=
322 		    ruleset->rules[rs_num].inactive.ticket)
323 			return (NULL);
324 		if (r_last)
325 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
326 			    pf_rulequeue);
327 		else
328 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
329 	}
330 	if (!r_last) {
331 		while ((rule != NULL) && (rule->nr != rule_number))
332 			rule = TAILQ_NEXT(rule, entries);
333 	}
334 	if (rule == NULL)
335 		return (NULL);
336 
337 	return (&rule->rpool);
338 }
339 
340 static void
341 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
342 {
343 	struct pf_pooladdr	*mv_pool_pa;
344 
345 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
346 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
347 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
348 	}
349 }
350 
351 static void
352 pf_empty_pool(struct pf_palist *poola)
353 {
354 	struct pf_pooladdr *pa;
355 
356 	while ((pa = TAILQ_FIRST(poola)) != NULL) {
357 		switch (pa->addr.type) {
358 		case PF_ADDR_DYNIFTL:
359 			pfi_dynaddr_remove(pa->addr.p.dyn);
360 			break;
361 		case PF_ADDR_TABLE:
362 			/* XXX: this could be unfinished pooladdr on pabuf */
363 			if (pa->addr.p.tbl != NULL)
364 				pfr_detach_table(pa->addr.p.tbl);
365 			break;
366 		}
367 		if (pa->kif)
368 			pfi_kif_unref(pa->kif);
369 		TAILQ_REMOVE(poola, pa, entries);
370 		free(pa, M_PFRULE);
371 	}
372 }
373 
374 static void
375 pf_unlink_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
376 {
377 
378 	PF_RULES_WASSERT();
379 
380 	TAILQ_REMOVE(rulequeue, rule, entries);
381 
382 	PF_UNLNKDRULES_LOCK();
383 	rule->rule_flag |= PFRULE_REFS;
384 	TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
385 	PF_UNLNKDRULES_UNLOCK();
386 }
387 
388 void
389 pf_free_rule(struct pf_rule *rule)
390 {
391 
392 	PF_RULES_WASSERT();
393 
394 	if (rule->tag)
395 		tag_unref(&V_pf_tags, rule->tag);
396 	if (rule->match_tag)
397 		tag_unref(&V_pf_tags, rule->match_tag);
398 #ifdef ALTQ
399 	if (rule->pqid != rule->qid)
400 		pf_qid_unref(rule->pqid);
401 	pf_qid_unref(rule->qid);
402 #endif
403 	switch (rule->src.addr.type) {
404 	case PF_ADDR_DYNIFTL:
405 		pfi_dynaddr_remove(rule->src.addr.p.dyn);
406 		break;
407 	case PF_ADDR_TABLE:
408 		pfr_detach_table(rule->src.addr.p.tbl);
409 		break;
410 	}
411 	switch (rule->dst.addr.type) {
412 	case PF_ADDR_DYNIFTL:
413 		pfi_dynaddr_remove(rule->dst.addr.p.dyn);
414 		break;
415 	case PF_ADDR_TABLE:
416 		pfr_detach_table(rule->dst.addr.p.tbl);
417 		break;
418 	}
419 	if (rule->overload_tbl)
420 		pfr_detach_table(rule->overload_tbl);
421 	if (rule->kif)
422 		pfi_kif_unref(rule->kif);
423 	pf_anchor_remove(rule);
424 	pf_empty_pool(&rule->rpool.list);
425 	counter_u64_free(rule->states_cur);
426 	counter_u64_free(rule->states_tot);
427 	counter_u64_free(rule->src_nodes);
428 	free(rule, M_PFRULE);
429 }
430 
431 static u_int16_t
432 tagname2tag(struct pf_tags *head, char *tagname)
433 {
434 	struct pf_tagname	*tag, *p = NULL;
435 	u_int16_t		 new_tagid = 1;
436 
437 	PF_RULES_WASSERT();
438 
439 	TAILQ_FOREACH(tag, head, entries)
440 		if (strcmp(tagname, tag->name) == 0) {
441 			tag->ref++;
442 			return (tag->tag);
443 		}
444 
445 	/*
446 	 * to avoid fragmentation, we do a linear search from the beginning
447 	 * and take the first free slot we find. if there is none or the list
448 	 * is empty, append a new entry at the end.
449 	 */
450 
451 	/* new entry */
452 	if (!TAILQ_EMPTY(head))
453 		for (p = TAILQ_FIRST(head); p != NULL &&
454 		    p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
455 			new_tagid = p->tag + 1;
456 
457 	if (new_tagid > TAGID_MAX)
458 		return (0);
459 
460 	/* allocate and fill new struct pf_tagname */
461 	tag = malloc(sizeof(*tag), M_PFTAG, M_NOWAIT|M_ZERO);
462 	if (tag == NULL)
463 		return (0);
464 	strlcpy(tag->name, tagname, sizeof(tag->name));
465 	tag->tag = new_tagid;
466 	tag->ref++;
467 
468 	if (p != NULL)	/* insert new entry before p */
469 		TAILQ_INSERT_BEFORE(p, tag, entries);
470 	else	/* either list empty or no free slot in between */
471 		TAILQ_INSERT_TAIL(head, tag, entries);
472 
473 	return (tag->tag);
474 }
475 
476 static void
477 tag_unref(struct pf_tags *head, u_int16_t tag)
478 {
479 	struct pf_tagname	*p, *next;
480 
481 	PF_RULES_WASSERT();
482 
483 	for (p = TAILQ_FIRST(head); p != NULL; p = next) {
484 		next = TAILQ_NEXT(p, entries);
485 		if (tag == p->tag) {
486 			if (--p->ref == 0) {
487 				TAILQ_REMOVE(head, p, entries);
488 				free(p, M_PFTAG);
489 			}
490 			break;
491 		}
492 	}
493 }
494 
495 static u_int16_t
496 pf_tagname2tag(char *tagname)
497 {
498 	return (tagname2tag(&V_pf_tags, tagname));
499 }
500 
501 #ifdef ALTQ
502 static u_int32_t
503 pf_qname2qid(char *qname)
504 {
505 	return ((u_int32_t)tagname2tag(&V_pf_qids, qname));
506 }
507 
508 static void
509 pf_qid_unref(u_int32_t qid)
510 {
511 	tag_unref(&V_pf_qids, (u_int16_t)qid);
512 }
513 
514 static int
515 pf_begin_altq(u_int32_t *ticket)
516 {
517 	struct pf_altq	*altq;
518 	int		 error = 0;
519 
520 	PF_RULES_WASSERT();
521 
522 	/* Purge the old altq list */
523 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
524 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
525 		if (altq->qname[0] == 0 &&
526 		    (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
527 			/* detach and destroy the discipline */
528 			error = altq_remove(altq);
529 		} else
530 			pf_qid_unref(altq->qid);
531 		free(altq, M_PFALTQ);
532 	}
533 	if (error)
534 		return (error);
535 	*ticket = ++V_ticket_altqs_inactive;
536 	V_altqs_inactive_open = 1;
537 	return (0);
538 }
539 
540 static int
541 pf_rollback_altq(u_int32_t ticket)
542 {
543 	struct pf_altq	*altq;
544 	int		 error = 0;
545 
546 	PF_RULES_WASSERT();
547 
548 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
549 		return (0);
550 	/* Purge the old altq list */
551 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
552 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
553 		if (altq->qname[0] == 0 &&
554 		   (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
555 			/* detach and destroy the discipline */
556 			error = altq_remove(altq);
557 		} else
558 			pf_qid_unref(altq->qid);
559 		free(altq, M_PFALTQ);
560 	}
561 	V_altqs_inactive_open = 0;
562 	return (error);
563 }
564 
565 static int
566 pf_commit_altq(u_int32_t ticket)
567 {
568 	struct pf_altqqueue	*old_altqs;
569 	struct pf_altq		*altq;
570 	int			 err, error = 0;
571 
572 	PF_RULES_WASSERT();
573 
574 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
575 		return (EBUSY);
576 
577 	/* swap altqs, keep the old. */
578 	old_altqs = V_pf_altqs_active;
579 	V_pf_altqs_active = V_pf_altqs_inactive;
580 	V_pf_altqs_inactive = old_altqs;
581 	V_ticket_altqs_active = V_ticket_altqs_inactive;
582 
583 	/* Attach new disciplines */
584 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
585 	if (altq->qname[0] == 0 &&
586 	   (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
587 			/* attach the discipline */
588 			error = altq_pfattach(altq);
589 			if (error == 0 && V_pf_altq_running)
590 				error = pf_enable_altq(altq);
591 			if (error != 0)
592 				return (error);
593 		}
594 	}
595 
596 	/* Purge the old altq list */
597 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
598 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
599 		if (altq->qname[0] == 0 &&
600 		    (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
601 			/* detach and destroy the discipline */
602 			if (V_pf_altq_running)
603 				error = pf_disable_altq(altq);
604 			err = altq_pfdetach(altq);
605 			if (err != 0 && error == 0)
606 				error = err;
607 			err = altq_remove(altq);
608 			if (err != 0 && error == 0)
609 				error = err;
610 		} else
611 			pf_qid_unref(altq->qid);
612 		free(altq, M_PFALTQ);
613 	}
614 
615 	V_altqs_inactive_open = 0;
616 	return (error);
617 }
618 
619 static int
620 pf_enable_altq(struct pf_altq *altq)
621 {
622 	struct ifnet		*ifp;
623 	struct tb_profile	 tb;
624 	int			 error = 0;
625 
626 	if ((ifp = ifunit(altq->ifname)) == NULL)
627 		return (EINVAL);
628 
629 	if (ifp->if_snd.altq_type != ALTQT_NONE)
630 		error = altq_enable(&ifp->if_snd);
631 
632 	/* set tokenbucket regulator */
633 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
634 		tb.rate = altq->ifbandwidth;
635 		tb.depth = altq->tbrsize;
636 		error = tbr_set(&ifp->if_snd, &tb);
637 	}
638 
639 	return (error);
640 }
641 
642 static int
643 pf_disable_altq(struct pf_altq *altq)
644 {
645 	struct ifnet		*ifp;
646 	struct tb_profile	 tb;
647 	int			 error;
648 
649 	if ((ifp = ifunit(altq->ifname)) == NULL)
650 		return (EINVAL);
651 
652 	/*
653 	 * when the discipline is no longer referenced, it was overridden
654 	 * by a new one.  if so, just return.
655 	 */
656 	if (altq->altq_disc != ifp->if_snd.altq_disc)
657 		return (0);
658 
659 	error = altq_disable(&ifp->if_snd);
660 
661 	if (error == 0) {
662 		/* clear tokenbucket regulator */
663 		tb.rate = 0;
664 		error = tbr_set(&ifp->if_snd, &tb);
665 	}
666 
667 	return (error);
668 }
669 
670 void
671 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
672 {
673 	struct ifnet	*ifp1;
674 	struct pf_altq	*a1, *a2, *a3;
675 	u_int32_t	 ticket;
676 	int		 error = 0;
677 
678 	/* Interrupt userland queue modifications */
679 	if (V_altqs_inactive_open)
680 		pf_rollback_altq(V_ticket_altqs_inactive);
681 
682 	/* Start new altq ruleset */
683 	if (pf_begin_altq(&ticket))
684 		return;
685 
686 	/* Copy the current active set */
687 	TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
688 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
689 		if (a2 == NULL) {
690 			error = ENOMEM;
691 			break;
692 		}
693 		bcopy(a1, a2, sizeof(struct pf_altq));
694 
695 		if (a2->qname[0] != 0) {
696 			if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
697 				error = EBUSY;
698 				free(a2, M_PFALTQ);
699 				break;
700 			}
701 			a2->altq_disc = NULL;
702 			TAILQ_FOREACH(a3, V_pf_altqs_inactive, entries) {
703 				if (strncmp(a3->ifname, a2->ifname,
704 				    IFNAMSIZ) == 0 && a3->qname[0] == 0) {
705 					a2->altq_disc = a3->altq_disc;
706 					break;
707 				}
708 			}
709 		}
710 		/* Deactivate the interface in question */
711 		a2->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
712 		if ((ifp1 = ifunit(a2->ifname)) == NULL ||
713 		    (remove && ifp1 == ifp)) {
714 			a2->local_flags |= PFALTQ_FLAG_IF_REMOVED;
715 		} else {
716 			error = altq_add(a2);
717 
718 			if (ticket != V_ticket_altqs_inactive)
719 				error = EBUSY;
720 
721 			if (error) {
722 				free(a2, M_PFALTQ);
723 				break;
724 			}
725 		}
726 
727 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
728 	}
729 
730 	if (error != 0)
731 		pf_rollback_altq(ticket);
732 	else
733 		pf_commit_altq(ticket);
734 }
735 #endif /* ALTQ */
736 
737 static int
738 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
739 {
740 	struct pf_ruleset	*rs;
741 	struct pf_rule		*rule;
742 
743 	PF_RULES_WASSERT();
744 
745 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
746 		return (EINVAL);
747 	rs = pf_find_or_create_ruleset(anchor);
748 	if (rs == NULL)
749 		return (EINVAL);
750 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
751 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
752 		rs->rules[rs_num].inactive.rcount--;
753 	}
754 	*ticket = ++rs->rules[rs_num].inactive.ticket;
755 	rs->rules[rs_num].inactive.open = 1;
756 	return (0);
757 }
758 
759 static int
760 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
761 {
762 	struct pf_ruleset	*rs;
763 	struct pf_rule		*rule;
764 
765 	PF_RULES_WASSERT();
766 
767 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
768 		return (EINVAL);
769 	rs = pf_find_ruleset(anchor);
770 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
771 	    rs->rules[rs_num].inactive.ticket != ticket)
772 		return (0);
773 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
774 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
775 		rs->rules[rs_num].inactive.rcount--;
776 	}
777 	rs->rules[rs_num].inactive.open = 0;
778 	return (0);
779 }
780 
781 #define PF_MD5_UPD(st, elm)						\
782 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
783 
784 #define PF_MD5_UPD_STR(st, elm)						\
785 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
786 
787 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
788 		(stor) = htonl((st)->elm);				\
789 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
790 } while (0)
791 
792 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
793 		(stor) = htons((st)->elm);				\
794 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
795 } while (0)
796 
797 static void
798 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
799 {
800 	PF_MD5_UPD(pfr, addr.type);
801 	switch (pfr->addr.type) {
802 		case PF_ADDR_DYNIFTL:
803 			PF_MD5_UPD(pfr, addr.v.ifname);
804 			PF_MD5_UPD(pfr, addr.iflags);
805 			break;
806 		case PF_ADDR_TABLE:
807 			PF_MD5_UPD(pfr, addr.v.tblname);
808 			break;
809 		case PF_ADDR_ADDRMASK:
810 			/* XXX ignore af? */
811 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
812 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
813 			break;
814 	}
815 
816 	PF_MD5_UPD(pfr, port[0]);
817 	PF_MD5_UPD(pfr, port[1]);
818 	PF_MD5_UPD(pfr, neg);
819 	PF_MD5_UPD(pfr, port_op);
820 }
821 
822 static void
823 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
824 {
825 	u_int16_t x;
826 	u_int32_t y;
827 
828 	pf_hash_rule_addr(ctx, &rule->src);
829 	pf_hash_rule_addr(ctx, &rule->dst);
830 	PF_MD5_UPD_STR(rule, label);
831 	PF_MD5_UPD_STR(rule, ifname);
832 	PF_MD5_UPD_STR(rule, match_tagname);
833 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
834 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
835 	PF_MD5_UPD_HTONL(rule, prob, y);
836 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
837 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
838 	PF_MD5_UPD(rule, uid.op);
839 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
840 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
841 	PF_MD5_UPD(rule, gid.op);
842 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
843 	PF_MD5_UPD(rule, action);
844 	PF_MD5_UPD(rule, direction);
845 	PF_MD5_UPD(rule, af);
846 	PF_MD5_UPD(rule, quick);
847 	PF_MD5_UPD(rule, ifnot);
848 	PF_MD5_UPD(rule, match_tag_not);
849 	PF_MD5_UPD(rule, natpass);
850 	PF_MD5_UPD(rule, keep_state);
851 	PF_MD5_UPD(rule, proto);
852 	PF_MD5_UPD(rule, type);
853 	PF_MD5_UPD(rule, code);
854 	PF_MD5_UPD(rule, flags);
855 	PF_MD5_UPD(rule, flagset);
856 	PF_MD5_UPD(rule, allow_opts);
857 	PF_MD5_UPD(rule, rt);
858 	PF_MD5_UPD(rule, tos);
859 }
860 
861 static int
862 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
863 {
864 	struct pf_ruleset	*rs;
865 	struct pf_rule		*rule, **old_array;
866 	struct pf_rulequeue	*old_rules;
867 	int			 error;
868 	u_int32_t		 old_rcount;
869 
870 	PF_RULES_WASSERT();
871 
872 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
873 		return (EINVAL);
874 	rs = pf_find_ruleset(anchor);
875 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
876 	    ticket != rs->rules[rs_num].inactive.ticket)
877 		return (EBUSY);
878 
879 	/* Calculate checksum for the main ruleset */
880 	if (rs == &pf_main_ruleset) {
881 		error = pf_setup_pfsync_matching(rs);
882 		if (error != 0)
883 			return (error);
884 	}
885 
886 	/* Swap rules, keep the old. */
887 	old_rules = rs->rules[rs_num].active.ptr;
888 	old_rcount = rs->rules[rs_num].active.rcount;
889 	old_array = rs->rules[rs_num].active.ptr_array;
890 
891 	rs->rules[rs_num].active.ptr =
892 	    rs->rules[rs_num].inactive.ptr;
893 	rs->rules[rs_num].active.ptr_array =
894 	    rs->rules[rs_num].inactive.ptr_array;
895 	rs->rules[rs_num].active.rcount =
896 	    rs->rules[rs_num].inactive.rcount;
897 	rs->rules[rs_num].inactive.ptr = old_rules;
898 	rs->rules[rs_num].inactive.ptr_array = old_array;
899 	rs->rules[rs_num].inactive.rcount = old_rcount;
900 
901 	rs->rules[rs_num].active.ticket =
902 	    rs->rules[rs_num].inactive.ticket;
903 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
904 
905 
906 	/* Purge the old rule list. */
907 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
908 		pf_unlink_rule(old_rules, rule);
909 	if (rs->rules[rs_num].inactive.ptr_array)
910 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
911 	rs->rules[rs_num].inactive.ptr_array = NULL;
912 	rs->rules[rs_num].inactive.rcount = 0;
913 	rs->rules[rs_num].inactive.open = 0;
914 	pf_remove_if_empty_ruleset(rs);
915 
916 	return (0);
917 }
918 
919 static int
920 pf_setup_pfsync_matching(struct pf_ruleset *rs)
921 {
922 	MD5_CTX			 ctx;
923 	struct pf_rule		*rule;
924 	int			 rs_cnt;
925 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
926 
927 	MD5Init(&ctx);
928 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
929 		/* XXX PF_RULESET_SCRUB as well? */
930 		if (rs_cnt == PF_RULESET_SCRUB)
931 			continue;
932 
933 		if (rs->rules[rs_cnt].inactive.ptr_array)
934 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
935 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
936 
937 		if (rs->rules[rs_cnt].inactive.rcount) {
938 			rs->rules[rs_cnt].inactive.ptr_array =
939 			    malloc(sizeof(caddr_t) *
940 			    rs->rules[rs_cnt].inactive.rcount,
941 			    M_TEMP, M_NOWAIT);
942 
943 			if (!rs->rules[rs_cnt].inactive.ptr_array)
944 				return (ENOMEM);
945 		}
946 
947 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
948 		    entries) {
949 			pf_hash_rule(&ctx, rule);
950 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
951 		}
952 	}
953 
954 	MD5Final(digest, &ctx);
955 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
956 	return (0);
957 }
958 
959 static int
960 pf_addr_setup(struct pf_ruleset *ruleset, struct pf_addr_wrap *addr,
961     sa_family_t af)
962 {
963 	int error = 0;
964 
965 	switch (addr->type) {
966 	case PF_ADDR_TABLE:
967 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
968 		if (addr->p.tbl == NULL)
969 			error = ENOMEM;
970 		break;
971 	case PF_ADDR_DYNIFTL:
972 		error = pfi_dynaddr_setup(addr, af);
973 		break;
974 	}
975 
976 	return (error);
977 }
978 
979 static void
980 pf_addr_copyout(struct pf_addr_wrap *addr)
981 {
982 
983 	switch (addr->type) {
984 	case PF_ADDR_DYNIFTL:
985 		pfi_dynaddr_copyout(addr);
986 		break;
987 	case PF_ADDR_TABLE:
988 		pf_tbladdr_copyout(addr);
989 		break;
990 	}
991 }
992 
993 static int
994 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
995 {
996 	int			 error = 0;
997 
998 	/* XXX keep in sync with switch() below */
999 	if (securelevel_gt(td->td_ucred, 2))
1000 		switch (cmd) {
1001 		case DIOCGETRULES:
1002 		case DIOCGETRULE:
1003 		case DIOCGETADDRS:
1004 		case DIOCGETADDR:
1005 		case DIOCGETSTATE:
1006 		case DIOCSETSTATUSIF:
1007 		case DIOCGETSTATUS:
1008 		case DIOCCLRSTATUS:
1009 		case DIOCNATLOOK:
1010 		case DIOCSETDEBUG:
1011 		case DIOCGETSTATES:
1012 		case DIOCGETTIMEOUT:
1013 		case DIOCCLRRULECTRS:
1014 		case DIOCGETLIMIT:
1015 		case DIOCGETALTQS:
1016 		case DIOCGETALTQ:
1017 		case DIOCGETQSTATS:
1018 		case DIOCGETRULESETS:
1019 		case DIOCGETRULESET:
1020 		case DIOCRGETTABLES:
1021 		case DIOCRGETTSTATS:
1022 		case DIOCRCLRTSTATS:
1023 		case DIOCRCLRADDRS:
1024 		case DIOCRADDADDRS:
1025 		case DIOCRDELADDRS:
1026 		case DIOCRSETADDRS:
1027 		case DIOCRGETADDRS:
1028 		case DIOCRGETASTATS:
1029 		case DIOCRCLRASTATS:
1030 		case DIOCRTSTADDRS:
1031 		case DIOCOSFPGET:
1032 		case DIOCGETSRCNODES:
1033 		case DIOCCLRSRCNODES:
1034 		case DIOCIGETIFACES:
1035 		case DIOCGIFSPEED:
1036 		case DIOCSETIFFLAG:
1037 		case DIOCCLRIFFLAG:
1038 			break;
1039 		case DIOCRCLRTABLES:
1040 		case DIOCRADDTABLES:
1041 		case DIOCRDELTABLES:
1042 		case DIOCRSETTFLAGS:
1043 			if (((struct pfioc_table *)addr)->pfrio_flags &
1044 			    PFR_FLAG_DUMMY)
1045 				break; /* dummy operation ok */
1046 			return (EPERM);
1047 		default:
1048 			return (EPERM);
1049 		}
1050 
1051 	if (!(flags & FWRITE))
1052 		switch (cmd) {
1053 		case DIOCGETRULES:
1054 		case DIOCGETADDRS:
1055 		case DIOCGETADDR:
1056 		case DIOCGETSTATE:
1057 		case DIOCGETSTATUS:
1058 		case DIOCGETSTATES:
1059 		case DIOCGETTIMEOUT:
1060 		case DIOCGETLIMIT:
1061 		case DIOCGETALTQS:
1062 		case DIOCGETALTQ:
1063 		case DIOCGETQSTATS:
1064 		case DIOCGETRULESETS:
1065 		case DIOCGETRULESET:
1066 		case DIOCNATLOOK:
1067 		case DIOCRGETTABLES:
1068 		case DIOCRGETTSTATS:
1069 		case DIOCRGETADDRS:
1070 		case DIOCRGETASTATS:
1071 		case DIOCRTSTADDRS:
1072 		case DIOCOSFPGET:
1073 		case DIOCGETSRCNODES:
1074 		case DIOCIGETIFACES:
1075 		case DIOCGIFSPEED:
1076 			break;
1077 		case DIOCRCLRTABLES:
1078 		case DIOCRADDTABLES:
1079 		case DIOCRDELTABLES:
1080 		case DIOCRCLRTSTATS:
1081 		case DIOCRCLRADDRS:
1082 		case DIOCRADDADDRS:
1083 		case DIOCRDELADDRS:
1084 		case DIOCRSETADDRS:
1085 		case DIOCRSETTFLAGS:
1086 			if (((struct pfioc_table *)addr)->pfrio_flags &
1087 			    PFR_FLAG_DUMMY) {
1088 				flags |= FWRITE; /* need write lock for dummy */
1089 				break; /* dummy operation ok */
1090 			}
1091 			return (EACCES);
1092 		case DIOCGETRULE:
1093 			if (((struct pfioc_rule *)addr)->action ==
1094 			    PF_GET_CLR_CNTR)
1095 				return (EACCES);
1096 			break;
1097 		default:
1098 			return (EACCES);
1099 		}
1100 
1101 	CURVNET_SET(TD_TO_VNET(td));
1102 
1103 	switch (cmd) {
1104 	case DIOCSTART:
1105 		sx_xlock(&pf_ioctl_lock);
1106 		if (V_pf_status.running)
1107 			error = EEXIST;
1108 		else {
1109 			int cpu;
1110 
1111 			error = hook_pf();
1112 			if (error) {
1113 				DPFPRINTF(PF_DEBUG_MISC,
1114 				    ("pf: pfil registration failed\n"));
1115 				break;
1116 			}
1117 			V_pf_status.running = 1;
1118 			V_pf_status.since = time_second;
1119 
1120 			CPU_FOREACH(cpu)
1121 				V_pf_stateid[cpu] = time_second;
1122 
1123 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1124 		}
1125 		break;
1126 
1127 	case DIOCSTOP:
1128 		sx_xlock(&pf_ioctl_lock);
1129 		if (!V_pf_status.running)
1130 			error = ENOENT;
1131 		else {
1132 			V_pf_status.running = 0;
1133 			error = dehook_pf();
1134 			if (error) {
1135 				V_pf_status.running = 1;
1136 				DPFPRINTF(PF_DEBUG_MISC,
1137 				    ("pf: pfil unregistration failed\n"));
1138 			}
1139 			V_pf_status.since = time_second;
1140 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1141 		}
1142 		break;
1143 
1144 	case DIOCADDRULE: {
1145 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1146 		struct pf_ruleset	*ruleset;
1147 		struct pf_rule		*rule, *tail;
1148 		struct pf_pooladdr	*pa;
1149 		struct pfi_kif		*kif = NULL;
1150 		int			 rs_num;
1151 
1152 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1153 			error = EINVAL;
1154 			break;
1155 		}
1156 #ifndef INET
1157 		if (pr->rule.af == AF_INET) {
1158 			error = EAFNOSUPPORT;
1159 			break;
1160 		}
1161 #endif /* INET */
1162 #ifndef INET6
1163 		if (pr->rule.af == AF_INET6) {
1164 			error = EAFNOSUPPORT;
1165 			break;
1166 		}
1167 #endif /* INET6 */
1168 
1169 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
1170 		bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1171 		if (rule->ifname[0])
1172 			kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1173 		rule->states_cur = counter_u64_alloc(M_WAITOK);
1174 		rule->states_tot = counter_u64_alloc(M_WAITOK);
1175 		rule->src_nodes = counter_u64_alloc(M_WAITOK);
1176 		rule->cuid = td->td_ucred->cr_ruid;
1177 		rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1178 		TAILQ_INIT(&rule->rpool.list);
1179 
1180 #define	ERROUT(x)	{ error = (x); goto DIOCADDRULE_error; }
1181 
1182 		PF_RULES_WLOCK();
1183 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1184 		ruleset = pf_find_ruleset(pr->anchor);
1185 		if (ruleset == NULL)
1186 			ERROUT(EINVAL);
1187 		rs_num = pf_get_ruleset_number(pr->rule.action);
1188 		if (rs_num >= PF_RULESET_MAX)
1189 			ERROUT(EINVAL);
1190 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1191 			DPFPRINTF(PF_DEBUG_MISC,
1192 			    ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
1193 			    ruleset->rules[rs_num].inactive.ticket));
1194 			ERROUT(EBUSY);
1195 		}
1196 		if (pr->pool_ticket != V_ticket_pabuf) {
1197 			DPFPRINTF(PF_DEBUG_MISC,
1198 			    ("pool_ticket: %d != %d\n", pr->pool_ticket,
1199 			    V_ticket_pabuf));
1200 			ERROUT(EBUSY);
1201 		}
1202 
1203 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1204 		    pf_rulequeue);
1205 		if (tail)
1206 			rule->nr = tail->nr + 1;
1207 		else
1208 			rule->nr = 0;
1209 		if (rule->ifname[0]) {
1210 			rule->kif = pfi_kif_attach(kif, rule->ifname);
1211 			pfi_kif_ref(rule->kif);
1212 		} else
1213 			rule->kif = NULL;
1214 
1215 		if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
1216 			error = EBUSY;
1217 
1218 #ifdef ALTQ
1219 		/* set queue IDs */
1220 		if (rule->qname[0] != 0) {
1221 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1222 				error = EBUSY;
1223 			else if (rule->pqname[0] != 0) {
1224 				if ((rule->pqid =
1225 				    pf_qname2qid(rule->pqname)) == 0)
1226 					error = EBUSY;
1227 			} else
1228 				rule->pqid = rule->qid;
1229 		}
1230 #endif
1231 		if (rule->tagname[0])
1232 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1233 				error = EBUSY;
1234 		if (rule->match_tagname[0])
1235 			if ((rule->match_tag =
1236 			    pf_tagname2tag(rule->match_tagname)) == 0)
1237 				error = EBUSY;
1238 		if (rule->rt && !rule->direction)
1239 			error = EINVAL;
1240 		if (!rule->log)
1241 			rule->logif = 0;
1242 		if (rule->logif >= PFLOGIFS_MAX)
1243 			error = EINVAL;
1244 		if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1245 			error = ENOMEM;
1246 		if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1247 			error = ENOMEM;
1248 		if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1249 			error = EINVAL;
1250 		TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1251 			if (pa->addr.type == PF_ADDR_TABLE) {
1252 				pa->addr.p.tbl = pfr_attach_table(ruleset,
1253 				    pa->addr.v.tblname);
1254 				if (pa->addr.p.tbl == NULL)
1255 					error = ENOMEM;
1256 			}
1257 
1258 		if (rule->overload_tblname[0]) {
1259 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1260 			    rule->overload_tblname)) == NULL)
1261 				error = EINVAL;
1262 			else
1263 				rule->overload_tbl->pfrkt_flags |=
1264 				    PFR_TFLAG_ACTIVE;
1265 		}
1266 
1267 		pf_mv_pool(&V_pf_pabuf, &rule->rpool.list);
1268 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1269 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1270 		    (rule->rt > PF_FASTROUTE)) &&
1271 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1272 			error = EINVAL;
1273 
1274 		if (error) {
1275 			pf_free_rule(rule);
1276 			PF_RULES_WUNLOCK();
1277 			break;
1278 		}
1279 
1280 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1281 		rule->evaluations = rule->packets[0] = rule->packets[1] =
1282 		    rule->bytes[0] = rule->bytes[1] = 0;
1283 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1284 		    rule, entries);
1285 		ruleset->rules[rs_num].inactive.rcount++;
1286 		PF_RULES_WUNLOCK();
1287 		break;
1288 
1289 #undef ERROUT
1290 DIOCADDRULE_error:
1291 		PF_RULES_WUNLOCK();
1292 		counter_u64_free(rule->states_cur);
1293 		counter_u64_free(rule->states_tot);
1294 		counter_u64_free(rule->src_nodes);
1295 		free(rule, M_PFRULE);
1296 		if (kif)
1297 			free(kif, PFI_MTYPE);
1298 		break;
1299 	}
1300 
1301 	case DIOCGETRULES: {
1302 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1303 		struct pf_ruleset	*ruleset;
1304 		struct pf_rule		*tail;
1305 		int			 rs_num;
1306 
1307 		PF_RULES_WLOCK();
1308 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1309 		ruleset = pf_find_ruleset(pr->anchor);
1310 		if (ruleset == NULL) {
1311 			PF_RULES_WUNLOCK();
1312 			error = EINVAL;
1313 			break;
1314 		}
1315 		rs_num = pf_get_ruleset_number(pr->rule.action);
1316 		if (rs_num >= PF_RULESET_MAX) {
1317 			PF_RULES_WUNLOCK();
1318 			error = EINVAL;
1319 			break;
1320 		}
1321 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1322 		    pf_rulequeue);
1323 		if (tail)
1324 			pr->nr = tail->nr + 1;
1325 		else
1326 			pr->nr = 0;
1327 		pr->ticket = ruleset->rules[rs_num].active.ticket;
1328 		PF_RULES_WUNLOCK();
1329 		break;
1330 	}
1331 
1332 	case DIOCGETRULE: {
1333 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1334 		struct pf_ruleset	*ruleset;
1335 		struct pf_rule		*rule;
1336 		int			 rs_num, i;
1337 
1338 		PF_RULES_WLOCK();
1339 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1340 		ruleset = pf_find_ruleset(pr->anchor);
1341 		if (ruleset == NULL) {
1342 			PF_RULES_WUNLOCK();
1343 			error = EINVAL;
1344 			break;
1345 		}
1346 		rs_num = pf_get_ruleset_number(pr->rule.action);
1347 		if (rs_num >= PF_RULESET_MAX) {
1348 			PF_RULES_WUNLOCK();
1349 			error = EINVAL;
1350 			break;
1351 		}
1352 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1353 			PF_RULES_WUNLOCK();
1354 			error = EBUSY;
1355 			break;
1356 		}
1357 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1358 		while ((rule != NULL) && (rule->nr != pr->nr))
1359 			rule = TAILQ_NEXT(rule, entries);
1360 		if (rule == NULL) {
1361 			PF_RULES_WUNLOCK();
1362 			error = EBUSY;
1363 			break;
1364 		}
1365 		bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1366 		pr->rule.u_states_cur = counter_u64_fetch(rule->states_cur);
1367 		pr->rule.u_states_tot = counter_u64_fetch(rule->states_tot);
1368 		pr->rule.u_src_nodes = counter_u64_fetch(rule->src_nodes);
1369 		if (pf_anchor_copyout(ruleset, rule, pr)) {
1370 			PF_RULES_WUNLOCK();
1371 			error = EBUSY;
1372 			break;
1373 		}
1374 		pf_addr_copyout(&pr->rule.src.addr);
1375 		pf_addr_copyout(&pr->rule.dst.addr);
1376 		for (i = 0; i < PF_SKIP_COUNT; ++i)
1377 			if (rule->skip[i].ptr == NULL)
1378 				pr->rule.skip[i].nr = -1;
1379 			else
1380 				pr->rule.skip[i].nr =
1381 				    rule->skip[i].ptr->nr;
1382 
1383 		if (pr->action == PF_GET_CLR_CNTR) {
1384 			rule->evaluations = 0;
1385 			rule->packets[0] = rule->packets[1] = 0;
1386 			rule->bytes[0] = rule->bytes[1] = 0;
1387 			counter_u64_zero(rule->states_tot);
1388 		}
1389 		PF_RULES_WUNLOCK();
1390 		break;
1391 	}
1392 
1393 	case DIOCCHANGERULE: {
1394 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1395 		struct pf_ruleset	*ruleset;
1396 		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1397 		struct pfi_kif		*kif = NULL;
1398 		struct pf_pooladdr	*pa;
1399 		u_int32_t		 nr = 0;
1400 		int			 rs_num;
1401 
1402 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1403 		    pcr->action > PF_CHANGE_GET_TICKET) {
1404 			error = EINVAL;
1405 			break;
1406 		}
1407 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1408 			error = EINVAL;
1409 			break;
1410 		}
1411 
1412 		if (pcr->action != PF_CHANGE_REMOVE) {
1413 #ifndef INET
1414 			if (pcr->rule.af == AF_INET) {
1415 				error = EAFNOSUPPORT;
1416 				break;
1417 			}
1418 #endif /* INET */
1419 #ifndef INET6
1420 			if (pcr->rule.af == AF_INET6) {
1421 				error = EAFNOSUPPORT;
1422 				break;
1423 			}
1424 #endif /* INET6 */
1425 			newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
1426 			bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1427 			if (newrule->ifname[0])
1428 				kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1429 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
1430 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
1431 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
1432 			newrule->cuid = td->td_ucred->cr_ruid;
1433 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1434 			TAILQ_INIT(&newrule->rpool.list);
1435 		}
1436 
1437 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGERULE_error; }
1438 
1439 		PF_RULES_WLOCK();
1440 		if (!(pcr->action == PF_CHANGE_REMOVE ||
1441 		    pcr->action == PF_CHANGE_GET_TICKET) &&
1442 		    pcr->pool_ticket != V_ticket_pabuf)
1443 			ERROUT(EBUSY);
1444 
1445 		ruleset = pf_find_ruleset(pcr->anchor);
1446 		if (ruleset == NULL)
1447 			ERROUT(EINVAL);
1448 
1449 		rs_num = pf_get_ruleset_number(pcr->rule.action);
1450 		if (rs_num >= PF_RULESET_MAX)
1451 			ERROUT(EINVAL);
1452 
1453 		if (pcr->action == PF_CHANGE_GET_TICKET) {
1454 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1455 			ERROUT(0);
1456 		} else if (pcr->ticket !=
1457 			    ruleset->rules[rs_num].active.ticket)
1458 				ERROUT(EINVAL);
1459 
1460 		if (pcr->action != PF_CHANGE_REMOVE) {
1461 			if (newrule->ifname[0]) {
1462 				newrule->kif = pfi_kif_attach(kif,
1463 				    newrule->ifname);
1464 				pfi_kif_ref(newrule->kif);
1465 			} else
1466 				newrule->kif = NULL;
1467 
1468 			if (newrule->rtableid > 0 &&
1469 			    newrule->rtableid >= rt_numfibs)
1470 				error = EBUSY;
1471 
1472 #ifdef ALTQ
1473 			/* set queue IDs */
1474 			if (newrule->qname[0] != 0) {
1475 				if ((newrule->qid =
1476 				    pf_qname2qid(newrule->qname)) == 0)
1477 					error = EBUSY;
1478 				else if (newrule->pqname[0] != 0) {
1479 					if ((newrule->pqid =
1480 					    pf_qname2qid(newrule->pqname)) == 0)
1481 						error = EBUSY;
1482 				} else
1483 					newrule->pqid = newrule->qid;
1484 			}
1485 #endif /* ALTQ */
1486 			if (newrule->tagname[0])
1487 				if ((newrule->tag =
1488 				    pf_tagname2tag(newrule->tagname)) == 0)
1489 					error = EBUSY;
1490 			if (newrule->match_tagname[0])
1491 				if ((newrule->match_tag = pf_tagname2tag(
1492 				    newrule->match_tagname)) == 0)
1493 					error = EBUSY;
1494 			if (newrule->rt && !newrule->direction)
1495 				error = EINVAL;
1496 			if (!newrule->log)
1497 				newrule->logif = 0;
1498 			if (newrule->logif >= PFLOGIFS_MAX)
1499 				error = EINVAL;
1500 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
1501 				error = ENOMEM;
1502 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
1503 				error = ENOMEM;
1504 			if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1505 				error = EINVAL;
1506 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1507 				if (pa->addr.type == PF_ADDR_TABLE) {
1508 					pa->addr.p.tbl =
1509 					    pfr_attach_table(ruleset,
1510 					    pa->addr.v.tblname);
1511 					if (pa->addr.p.tbl == NULL)
1512 						error = ENOMEM;
1513 				}
1514 
1515 			if (newrule->overload_tblname[0]) {
1516 				if ((newrule->overload_tbl = pfr_attach_table(
1517 				    ruleset, newrule->overload_tblname)) ==
1518 				    NULL)
1519 					error = EINVAL;
1520 				else
1521 					newrule->overload_tbl->pfrkt_flags |=
1522 					    PFR_TFLAG_ACTIVE;
1523 			}
1524 
1525 			pf_mv_pool(&V_pf_pabuf, &newrule->rpool.list);
1526 			if (((((newrule->action == PF_NAT) ||
1527 			    (newrule->action == PF_RDR) ||
1528 			    (newrule->action == PF_BINAT) ||
1529 			    (newrule->rt > PF_FASTROUTE)) &&
1530 			    !newrule->anchor)) &&
1531 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1532 				error = EINVAL;
1533 
1534 			if (error) {
1535 				pf_free_rule(newrule);
1536 				PF_RULES_WUNLOCK();
1537 				break;
1538 			}
1539 
1540 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1541 			newrule->evaluations = 0;
1542 			newrule->packets[0] = newrule->packets[1] = 0;
1543 			newrule->bytes[0] = newrule->bytes[1] = 0;
1544 		}
1545 		pf_empty_pool(&V_pf_pabuf);
1546 
1547 		if (pcr->action == PF_CHANGE_ADD_HEAD)
1548 			oldrule = TAILQ_FIRST(
1549 			    ruleset->rules[rs_num].active.ptr);
1550 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1551 			oldrule = TAILQ_LAST(
1552 			    ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1553 		else {
1554 			oldrule = TAILQ_FIRST(
1555 			    ruleset->rules[rs_num].active.ptr);
1556 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1557 				oldrule = TAILQ_NEXT(oldrule, entries);
1558 			if (oldrule == NULL) {
1559 				if (newrule != NULL)
1560 					pf_free_rule(newrule);
1561 				PF_RULES_WUNLOCK();
1562 				error = EINVAL;
1563 				break;
1564 			}
1565 		}
1566 
1567 		if (pcr->action == PF_CHANGE_REMOVE) {
1568 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
1569 			    oldrule);
1570 			ruleset->rules[rs_num].active.rcount--;
1571 		} else {
1572 			if (oldrule == NULL)
1573 				TAILQ_INSERT_TAIL(
1574 				    ruleset->rules[rs_num].active.ptr,
1575 				    newrule, entries);
1576 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1577 			    pcr->action == PF_CHANGE_ADD_BEFORE)
1578 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1579 			else
1580 				TAILQ_INSERT_AFTER(
1581 				    ruleset->rules[rs_num].active.ptr,
1582 				    oldrule, newrule, entries);
1583 			ruleset->rules[rs_num].active.rcount++;
1584 		}
1585 
1586 		nr = 0;
1587 		TAILQ_FOREACH(oldrule,
1588 		    ruleset->rules[rs_num].active.ptr, entries)
1589 			oldrule->nr = nr++;
1590 
1591 		ruleset->rules[rs_num].active.ticket++;
1592 
1593 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1594 		pf_remove_if_empty_ruleset(ruleset);
1595 
1596 		PF_RULES_WUNLOCK();
1597 		break;
1598 
1599 #undef ERROUT
1600 DIOCCHANGERULE_error:
1601 		PF_RULES_WUNLOCK();
1602 		if (newrule != NULL) {
1603 			counter_u64_free(newrule->states_cur);
1604 			counter_u64_free(newrule->states_tot);
1605 			counter_u64_free(newrule->src_nodes);
1606 			free(newrule, M_PFRULE);
1607 		}
1608 		if (kif != NULL)
1609 			free(kif, PFI_MTYPE);
1610 		break;
1611 	}
1612 
1613 	case DIOCCLRSTATES: {
1614 		struct pf_state		*s;
1615 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1616 		u_int			 i, killed = 0;
1617 
1618 		for (i = 0; i <= pf_hashmask; i++) {
1619 			struct pf_idhash *ih = &V_pf_idhash[i];
1620 
1621 relock_DIOCCLRSTATES:
1622 			PF_HASHROW_LOCK(ih);
1623 			LIST_FOREACH(s, &ih->states, entry)
1624 				if (!psk->psk_ifname[0] ||
1625 				    !strcmp(psk->psk_ifname,
1626 				    s->kif->pfik_name)) {
1627 					/*
1628 					 * Don't send out individual
1629 					 * delete messages.
1630 					 */
1631 					s->state_flags |= PFSTATE_NOSYNC;
1632 					pf_unlink_state(s, PF_ENTER_LOCKED);
1633 					killed++;
1634 					goto relock_DIOCCLRSTATES;
1635 				}
1636 			PF_HASHROW_UNLOCK(ih);
1637 		}
1638 		psk->psk_killed = killed;
1639 		if (pfsync_clear_states_ptr != NULL)
1640 			pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
1641 		break;
1642 	}
1643 
1644 	case DIOCKILLSTATES: {
1645 		struct pf_state		*s;
1646 		struct pf_state_key	*sk;
1647 		struct pf_addr		*srcaddr, *dstaddr;
1648 		u_int16_t		 srcport, dstport;
1649 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
1650 		u_int			 i, killed = 0;
1651 
1652 		if (psk->psk_pfcmp.id) {
1653 			if (psk->psk_pfcmp.creatorid == 0)
1654 				psk->psk_pfcmp.creatorid = V_pf_status.hostid;
1655 			if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
1656 			    psk->psk_pfcmp.creatorid))) {
1657 				pf_unlink_state(s, PF_ENTER_LOCKED);
1658 				psk->psk_killed = 1;
1659 			}
1660 			break;
1661 		}
1662 
1663 		for (i = 0; i <= pf_hashmask; i++) {
1664 			struct pf_idhash *ih = &V_pf_idhash[i];
1665 
1666 relock_DIOCKILLSTATES:
1667 			PF_HASHROW_LOCK(ih);
1668 			LIST_FOREACH(s, &ih->states, entry) {
1669 				sk = s->key[PF_SK_WIRE];
1670 				if (s->direction == PF_OUT) {
1671 					srcaddr = &sk->addr[1];
1672 					dstaddr = &sk->addr[0];
1673 					srcport = sk->port[0];
1674 					dstport = sk->port[0];
1675 				} else {
1676 					srcaddr = &sk->addr[0];
1677 					dstaddr = &sk->addr[1];
1678 					srcport = sk->port[0];
1679 					dstport = sk->port[0];
1680 				}
1681 
1682 				if ((!psk->psk_af || sk->af == psk->psk_af)
1683 				    && (!psk->psk_proto || psk->psk_proto ==
1684 				    sk->proto) &&
1685 				    PF_MATCHA(psk->psk_src.neg,
1686 				    &psk->psk_src.addr.v.a.addr,
1687 				    &psk->psk_src.addr.v.a.mask,
1688 				    srcaddr, sk->af) &&
1689 				    PF_MATCHA(psk->psk_dst.neg,
1690 				    &psk->psk_dst.addr.v.a.addr,
1691 				    &psk->psk_dst.addr.v.a.mask,
1692 				    dstaddr, sk->af) &&
1693 				    (psk->psk_src.port_op == 0 ||
1694 				    pf_match_port(psk->psk_src.port_op,
1695 				    psk->psk_src.port[0], psk->psk_src.port[1],
1696 				    srcport)) &&
1697 				    (psk->psk_dst.port_op == 0 ||
1698 				    pf_match_port(psk->psk_dst.port_op,
1699 				    psk->psk_dst.port[0], psk->psk_dst.port[1],
1700 				    dstport)) &&
1701 				    (!psk->psk_label[0] ||
1702 				    (s->rule.ptr->label[0] &&
1703 				    !strcmp(psk->psk_label,
1704 				    s->rule.ptr->label))) &&
1705 				    (!psk->psk_ifname[0] ||
1706 				    !strcmp(psk->psk_ifname,
1707 				    s->kif->pfik_name))) {
1708 					pf_unlink_state(s, PF_ENTER_LOCKED);
1709 					killed++;
1710 					goto relock_DIOCKILLSTATES;
1711 				}
1712 			}
1713 			PF_HASHROW_UNLOCK(ih);
1714 		}
1715 		psk->psk_killed = killed;
1716 		break;
1717 	}
1718 
1719 	case DIOCADDSTATE: {
1720 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1721 		struct pfsync_state	*sp = &ps->state;
1722 
1723 		if (sp->timeout >= PFTM_MAX) {
1724 			error = EINVAL;
1725 			break;
1726 		}
1727 		if (pfsync_state_import_ptr != NULL) {
1728 			PF_RULES_RLOCK();
1729 			error = pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
1730 			PF_RULES_RUNLOCK();
1731 		} else
1732 			error = EOPNOTSUPP;
1733 		break;
1734 	}
1735 
1736 	case DIOCGETSTATE: {
1737 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1738 		struct pf_state		*s;
1739 
1740 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
1741 		if (s == NULL) {
1742 			error = ENOENT;
1743 			break;
1744 		}
1745 
1746 		pfsync_state_export(&ps->state, s);
1747 		PF_STATE_UNLOCK(s);
1748 		break;
1749 	}
1750 
1751 	case DIOCGETSTATES: {
1752 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
1753 		struct pf_state		*s;
1754 		struct pfsync_state	*pstore, *p;
1755 		int i, nr;
1756 
1757 		if (ps->ps_len == 0) {
1758 			nr = uma_zone_get_cur(V_pf_state_z);
1759 			ps->ps_len = sizeof(struct pfsync_state) * nr;
1760 			break;
1761 		}
1762 
1763 		p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK);
1764 		nr = 0;
1765 
1766 		for (i = 0; i <= pf_hashmask; i++) {
1767 			struct pf_idhash *ih = &V_pf_idhash[i];
1768 
1769 			PF_HASHROW_LOCK(ih);
1770 			LIST_FOREACH(s, &ih->states, entry) {
1771 
1772 				if (s->timeout == PFTM_UNLINKED)
1773 					continue;
1774 
1775 				if ((nr+1) * sizeof(*p) > ps->ps_len) {
1776 					PF_HASHROW_UNLOCK(ih);
1777 					goto DIOCGETSTATES_full;
1778 				}
1779 				pfsync_state_export(p, s);
1780 				p++;
1781 				nr++;
1782 			}
1783 			PF_HASHROW_UNLOCK(ih);
1784 		}
1785 DIOCGETSTATES_full:
1786 		error = copyout(pstore, ps->ps_states,
1787 		    sizeof(struct pfsync_state) * nr);
1788 		if (error) {
1789 			free(pstore, M_TEMP);
1790 			break;
1791 		}
1792 		ps->ps_len = sizeof(struct pfsync_state) * nr;
1793 		free(pstore, M_TEMP);
1794 
1795 		break;
1796 	}
1797 
1798 	case DIOCGETSTATUS: {
1799 		struct pf_status *s = (struct pf_status *)addr;
1800 
1801 		PF_RULES_RLOCK();
1802 		s->running = V_pf_status.running;
1803 		s->since   = V_pf_status.since;
1804 		s->debug   = V_pf_status.debug;
1805 		s->hostid  = V_pf_status.hostid;
1806 		s->states  = V_pf_status.states;
1807 		s->src_nodes = V_pf_status.src_nodes;
1808 
1809 		for (int i = 0; i < PFRES_MAX; i++)
1810 			s->counters[i] =
1811 			    counter_u64_fetch(V_pf_status.counters[i]);
1812 		for (int i = 0; i < LCNT_MAX; i++)
1813 			s->lcounters[i] =
1814 			    counter_u64_fetch(V_pf_status.lcounters[i]);
1815 		for (int i = 0; i < FCNT_MAX; i++)
1816 			s->fcounters[i] =
1817 			    counter_u64_fetch(V_pf_status.fcounters[i]);
1818 		for (int i = 0; i < SCNT_MAX; i++)
1819 			s->scounters[i] =
1820 			    counter_u64_fetch(V_pf_status.scounters[i]);
1821 
1822 		bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ);
1823 		bcopy(V_pf_status.pf_chksum, s->pf_chksum,
1824 		    PF_MD5_DIGEST_LENGTH);
1825 
1826 		pfi_update_status(s->ifname, s);
1827 		PF_RULES_RUNLOCK();
1828 		break;
1829 	}
1830 
1831 	case DIOCSETSTATUSIF: {
1832 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
1833 
1834 		if (pi->ifname[0] == 0) {
1835 			bzero(V_pf_status.ifname, IFNAMSIZ);
1836 			break;
1837 		}
1838 		PF_RULES_WLOCK();
1839 		strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
1840 		PF_RULES_WUNLOCK();
1841 		break;
1842 	}
1843 
1844 	case DIOCCLRSTATUS: {
1845 		PF_RULES_WLOCK();
1846 		for (int i = 0; i < PFRES_MAX; i++)
1847 			counter_u64_zero(V_pf_status.counters[i]);
1848 		for (int i = 0; i < FCNT_MAX; i++)
1849 			counter_u64_zero(V_pf_status.fcounters[i]);
1850 		for (int i = 0; i < SCNT_MAX; i++)
1851 			counter_u64_zero(V_pf_status.scounters[i]);
1852 		V_pf_status.since = time_second;
1853 		if (*V_pf_status.ifname)
1854 			pfi_update_status(V_pf_status.ifname, NULL);
1855 		PF_RULES_WUNLOCK();
1856 		break;
1857 	}
1858 
1859 	case DIOCNATLOOK: {
1860 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
1861 		struct pf_state_key	*sk;
1862 		struct pf_state		*state;
1863 		struct pf_state_key_cmp	 key;
1864 		int			 m = 0, direction = pnl->direction;
1865 		int			 sidx, didx;
1866 
1867 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
1868 		sidx = (direction == PF_IN) ? 1 : 0;
1869 		didx = (direction == PF_IN) ? 0 : 1;
1870 
1871 		if (!pnl->proto ||
1872 		    PF_AZERO(&pnl->saddr, pnl->af) ||
1873 		    PF_AZERO(&pnl->daddr, pnl->af) ||
1874 		    ((pnl->proto == IPPROTO_TCP ||
1875 		    pnl->proto == IPPROTO_UDP) &&
1876 		    (!pnl->dport || !pnl->sport)))
1877 			error = EINVAL;
1878 		else {
1879 			bzero(&key, sizeof(key));
1880 			key.af = pnl->af;
1881 			key.proto = pnl->proto;
1882 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
1883 			key.port[sidx] = pnl->sport;
1884 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
1885 			key.port[didx] = pnl->dport;
1886 
1887 			state = pf_find_state_all(&key, direction, &m);
1888 
1889 			if (m > 1)
1890 				error = E2BIG;	/* more than one state */
1891 			else if (state != NULL) {
1892 				/* XXXGL: not locked read */
1893 				sk = state->key[sidx];
1894 				PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
1895 				pnl->rsport = sk->port[sidx];
1896 				PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
1897 				pnl->rdport = sk->port[didx];
1898 			} else
1899 				error = ENOENT;
1900 		}
1901 		break;
1902 	}
1903 
1904 	case DIOCSETTIMEOUT: {
1905 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1906 		int		 old;
1907 
1908 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
1909 		    pt->seconds < 0) {
1910 			error = EINVAL;
1911 			break;
1912 		}
1913 		PF_RULES_WLOCK();
1914 		old = V_pf_default_rule.timeout[pt->timeout];
1915 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
1916 			pt->seconds = 1;
1917 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
1918 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
1919 			wakeup(pf_purge_thread);
1920 		pt->seconds = old;
1921 		PF_RULES_WUNLOCK();
1922 		break;
1923 	}
1924 
1925 	case DIOCGETTIMEOUT: {
1926 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1927 
1928 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
1929 			error = EINVAL;
1930 			break;
1931 		}
1932 		PF_RULES_RLOCK();
1933 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
1934 		PF_RULES_RUNLOCK();
1935 		break;
1936 	}
1937 
1938 	case DIOCGETLIMIT: {
1939 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1940 
1941 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
1942 			error = EINVAL;
1943 			break;
1944 		}
1945 		PF_RULES_RLOCK();
1946 		pl->limit = V_pf_limits[pl->index].limit;
1947 		PF_RULES_RUNLOCK();
1948 		break;
1949 	}
1950 
1951 	case DIOCSETLIMIT: {
1952 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1953 		int			 old_limit;
1954 
1955 		PF_RULES_WLOCK();
1956 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
1957 		    V_pf_limits[pl->index].zone == NULL) {
1958 			PF_RULES_WUNLOCK();
1959 			error = EINVAL;
1960 			break;
1961 		}
1962 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
1963 		old_limit = V_pf_limits[pl->index].limit;
1964 		V_pf_limits[pl->index].limit = pl->limit;
1965 		pl->limit = old_limit;
1966 		PF_RULES_WUNLOCK();
1967 		break;
1968 	}
1969 
1970 	case DIOCSETDEBUG: {
1971 		u_int32_t	*level = (u_int32_t *)addr;
1972 
1973 		PF_RULES_WLOCK();
1974 		V_pf_status.debug = *level;
1975 		PF_RULES_WUNLOCK();
1976 		break;
1977 	}
1978 
1979 	case DIOCCLRRULECTRS: {
1980 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
1981 		struct pf_ruleset	*ruleset = &pf_main_ruleset;
1982 		struct pf_rule		*rule;
1983 
1984 		PF_RULES_WLOCK();
1985 		TAILQ_FOREACH(rule,
1986 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
1987 			rule->evaluations = 0;
1988 			rule->packets[0] = rule->packets[1] = 0;
1989 			rule->bytes[0] = rule->bytes[1] = 0;
1990 		}
1991 		PF_RULES_WUNLOCK();
1992 		break;
1993 	}
1994 
1995 	case DIOCGIFSPEED: {
1996 		struct pf_ifspeed	*psp = (struct pf_ifspeed *)addr;
1997 		struct pf_ifspeed	ps;
1998 		struct ifnet		*ifp;
1999 
2000 		if (psp->ifname[0] != 0) {
2001 			/* Can we completely trust user-land? */
2002 			strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
2003 			ifp = ifunit(ps.ifname);
2004 			if (ifp != NULL)
2005 				psp->baudrate = ifp->if_baudrate;
2006 			else
2007 				error = EINVAL;
2008 		} else
2009 			error = EINVAL;
2010 		break;
2011 	}
2012 
2013 #ifdef ALTQ
2014 	case DIOCSTARTALTQ: {
2015 		struct pf_altq		*altq;
2016 
2017 		PF_RULES_WLOCK();
2018 		/* enable all altq interfaces on active list */
2019 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
2020 			if (altq->qname[0] == 0 && (altq->local_flags &
2021 			    PFALTQ_FLAG_IF_REMOVED) == 0) {
2022 				error = pf_enable_altq(altq);
2023 				if (error != 0)
2024 					break;
2025 			}
2026 		}
2027 		if (error == 0)
2028 			V_pf_altq_running = 1;
2029 		PF_RULES_WUNLOCK();
2030 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
2031 		break;
2032 	}
2033 
2034 	case DIOCSTOPALTQ: {
2035 		struct pf_altq		*altq;
2036 
2037 		PF_RULES_WLOCK();
2038 		/* disable all altq interfaces on active list */
2039 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
2040 			if (altq->qname[0] == 0 && (altq->local_flags &
2041 			    PFALTQ_FLAG_IF_REMOVED) == 0) {
2042 				error = pf_disable_altq(altq);
2043 				if (error != 0)
2044 					break;
2045 			}
2046 		}
2047 		if (error == 0)
2048 			V_pf_altq_running = 0;
2049 		PF_RULES_WUNLOCK();
2050 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2051 		break;
2052 	}
2053 
2054 	case DIOCADDALTQ: {
2055 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2056 		struct pf_altq		*altq, *a;
2057 		struct ifnet		*ifp;
2058 
2059 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK);
2060 		bcopy(&pa->altq, altq, sizeof(struct pf_altq));
2061 		altq->local_flags = 0;
2062 
2063 		PF_RULES_WLOCK();
2064 		if (pa->ticket != V_ticket_altqs_inactive) {
2065 			PF_RULES_WUNLOCK();
2066 			free(altq, M_PFALTQ);
2067 			error = EBUSY;
2068 			break;
2069 		}
2070 
2071 		/*
2072 		 * if this is for a queue, find the discipline and
2073 		 * copy the necessary fields
2074 		 */
2075 		if (altq->qname[0] != 0) {
2076 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2077 				PF_RULES_WUNLOCK();
2078 				error = EBUSY;
2079 				free(altq, M_PFALTQ);
2080 				break;
2081 			}
2082 			altq->altq_disc = NULL;
2083 			TAILQ_FOREACH(a, V_pf_altqs_inactive, entries) {
2084 				if (strncmp(a->ifname, altq->ifname,
2085 				    IFNAMSIZ) == 0 && a->qname[0] == 0) {
2086 					altq->altq_disc = a->altq_disc;
2087 					break;
2088 				}
2089 			}
2090 		}
2091 
2092 		if ((ifp = ifunit(altq->ifname)) == NULL)
2093 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
2094 		else
2095 			error = altq_add(altq);
2096 
2097 		if (error) {
2098 			PF_RULES_WUNLOCK();
2099 			free(altq, M_PFALTQ);
2100 			break;
2101 		}
2102 
2103 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
2104 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2105 		PF_RULES_WUNLOCK();
2106 		break;
2107 	}
2108 
2109 	case DIOCGETALTQS: {
2110 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2111 		struct pf_altq		*altq;
2112 
2113 		PF_RULES_RLOCK();
2114 		pa->nr = 0;
2115 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
2116 			pa->nr++;
2117 		pa->ticket = V_ticket_altqs_active;
2118 		PF_RULES_RUNLOCK();
2119 		break;
2120 	}
2121 
2122 	case DIOCGETALTQ: {
2123 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2124 		struct pf_altq		*altq;
2125 		u_int32_t		 nr;
2126 
2127 		PF_RULES_RLOCK();
2128 		if (pa->ticket != V_ticket_altqs_active) {
2129 			PF_RULES_RUNLOCK();
2130 			error = EBUSY;
2131 			break;
2132 		}
2133 		nr = 0;
2134 		altq = TAILQ_FIRST(V_pf_altqs_active);
2135 		while ((altq != NULL) && (nr < pa->nr)) {
2136 			altq = TAILQ_NEXT(altq, entries);
2137 			nr++;
2138 		}
2139 		if (altq == NULL) {
2140 			PF_RULES_RUNLOCK();
2141 			error = EBUSY;
2142 			break;
2143 		}
2144 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2145 		PF_RULES_RUNLOCK();
2146 		break;
2147 	}
2148 
2149 	case DIOCCHANGEALTQ:
2150 		/* CHANGEALTQ not supported yet! */
2151 		error = ENODEV;
2152 		break;
2153 
2154 	case DIOCGETQSTATS: {
2155 		struct pfioc_qstats	*pq = (struct pfioc_qstats *)addr;
2156 		struct pf_altq		*altq;
2157 		u_int32_t		 nr;
2158 		int			 nbytes;
2159 
2160 		PF_RULES_RLOCK();
2161 		if (pq->ticket != V_ticket_altqs_active) {
2162 			PF_RULES_RUNLOCK();
2163 			error = EBUSY;
2164 			break;
2165 		}
2166 		nbytes = pq->nbytes;
2167 		nr = 0;
2168 		altq = TAILQ_FIRST(V_pf_altqs_active);
2169 		while ((altq != NULL) && (nr < pq->nr)) {
2170 			altq = TAILQ_NEXT(altq, entries);
2171 			nr++;
2172 		}
2173 		if (altq == NULL) {
2174 			PF_RULES_RUNLOCK();
2175 			error = EBUSY;
2176 			break;
2177 		}
2178 
2179 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
2180 			PF_RULES_RUNLOCK();
2181 			error = ENXIO;
2182 			break;
2183 		}
2184 		PF_RULES_RUNLOCK();
2185 		error = altq_getqstats(altq, pq->buf, &nbytes);
2186 		if (error == 0) {
2187 			pq->scheduler = altq->scheduler;
2188 			pq->nbytes = nbytes;
2189 		}
2190 		break;
2191 	}
2192 #endif /* ALTQ */
2193 
2194 	case DIOCBEGINADDRS: {
2195 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2196 
2197 		PF_RULES_WLOCK();
2198 		pf_empty_pool(&V_pf_pabuf);
2199 		pp->ticket = ++V_ticket_pabuf;
2200 		PF_RULES_WUNLOCK();
2201 		break;
2202 	}
2203 
2204 	case DIOCADDADDR: {
2205 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2206 		struct pf_pooladdr	*pa;
2207 		struct pfi_kif		*kif = NULL;
2208 
2209 #ifndef INET
2210 		if (pp->af == AF_INET) {
2211 			error = EAFNOSUPPORT;
2212 			break;
2213 		}
2214 #endif /* INET */
2215 #ifndef INET6
2216 		if (pp->af == AF_INET6) {
2217 			error = EAFNOSUPPORT;
2218 			break;
2219 		}
2220 #endif /* INET6 */
2221 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2222 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2223 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2224 			error = EINVAL;
2225 			break;
2226 		}
2227 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2228 		bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2229 		if (pa->ifname[0])
2230 			kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2231 		PF_RULES_WLOCK();
2232 		if (pp->ticket != V_ticket_pabuf) {
2233 			PF_RULES_WUNLOCK();
2234 			if (pa->ifname[0])
2235 				free(kif, PFI_MTYPE);
2236 			free(pa, M_PFRULE);
2237 			error = EBUSY;
2238 			break;
2239 		}
2240 		if (pa->ifname[0]) {
2241 			pa->kif = pfi_kif_attach(kif, pa->ifname);
2242 			pfi_kif_ref(pa->kif);
2243 		} else
2244 			pa->kif = NULL;
2245 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2246 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2247 			if (pa->ifname[0])
2248 				pfi_kif_unref(pa->kif);
2249 			PF_RULES_WUNLOCK();
2250 			free(pa, M_PFRULE);
2251 			break;
2252 		}
2253 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2254 		PF_RULES_WUNLOCK();
2255 		break;
2256 	}
2257 
2258 	case DIOCGETADDRS: {
2259 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2260 		struct pf_pool		*pool;
2261 		struct pf_pooladdr	*pa;
2262 
2263 		PF_RULES_RLOCK();
2264 		pp->nr = 0;
2265 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2266 		    pp->r_num, 0, 1, 0);
2267 		if (pool == NULL) {
2268 			PF_RULES_RUNLOCK();
2269 			error = EBUSY;
2270 			break;
2271 		}
2272 		TAILQ_FOREACH(pa, &pool->list, entries)
2273 			pp->nr++;
2274 		PF_RULES_RUNLOCK();
2275 		break;
2276 	}
2277 
2278 	case DIOCGETADDR: {
2279 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2280 		struct pf_pool		*pool;
2281 		struct pf_pooladdr	*pa;
2282 		u_int32_t		 nr = 0;
2283 
2284 		PF_RULES_RLOCK();
2285 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2286 		    pp->r_num, 0, 1, 1);
2287 		if (pool == NULL) {
2288 			PF_RULES_RUNLOCK();
2289 			error = EBUSY;
2290 			break;
2291 		}
2292 		pa = TAILQ_FIRST(&pool->list);
2293 		while ((pa != NULL) && (nr < pp->nr)) {
2294 			pa = TAILQ_NEXT(pa, entries);
2295 			nr++;
2296 		}
2297 		if (pa == NULL) {
2298 			PF_RULES_RUNLOCK();
2299 			error = EBUSY;
2300 			break;
2301 		}
2302 		bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2303 		pf_addr_copyout(&pp->addr.addr);
2304 		PF_RULES_RUNLOCK();
2305 		break;
2306 	}
2307 
2308 	case DIOCCHANGEADDR: {
2309 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
2310 		struct pf_pool		*pool;
2311 		struct pf_pooladdr	*oldpa = NULL, *newpa = NULL;
2312 		struct pf_ruleset	*ruleset;
2313 		struct pfi_kif		*kif = NULL;
2314 
2315 		if (pca->action < PF_CHANGE_ADD_HEAD ||
2316 		    pca->action > PF_CHANGE_REMOVE) {
2317 			error = EINVAL;
2318 			break;
2319 		}
2320 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2321 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2322 		    pca->addr.addr.type != PF_ADDR_TABLE) {
2323 			error = EINVAL;
2324 			break;
2325 		}
2326 
2327 		if (pca->action != PF_CHANGE_REMOVE) {
2328 #ifndef INET
2329 			if (pca->af == AF_INET) {
2330 				error = EAFNOSUPPORT;
2331 				break;
2332 			}
2333 #endif /* INET */
2334 #ifndef INET6
2335 			if (pca->af == AF_INET6) {
2336 				error = EAFNOSUPPORT;
2337 				break;
2338 			}
2339 #endif /* INET6 */
2340 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
2341 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2342 			if (newpa->ifname[0])
2343 				kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2344 			newpa->kif = NULL;
2345 		}
2346 
2347 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGEADDR_error; }
2348 		PF_RULES_WLOCK();
2349 		ruleset = pf_find_ruleset(pca->anchor);
2350 		if (ruleset == NULL)
2351 			ERROUT(EBUSY);
2352 
2353 		pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2354 		    pca->r_num, pca->r_last, 1, 1);
2355 		if (pool == NULL)
2356 			ERROUT(EBUSY);
2357 
2358 		if (pca->action != PF_CHANGE_REMOVE) {
2359 			if (newpa->ifname[0]) {
2360 				newpa->kif = pfi_kif_attach(kif, newpa->ifname);
2361 				pfi_kif_ref(newpa->kif);
2362 				kif = NULL;
2363 			}
2364 
2365 			switch (newpa->addr.type) {
2366 			case PF_ADDR_DYNIFTL:
2367 				error = pfi_dynaddr_setup(&newpa->addr,
2368 				    pca->af);
2369 				break;
2370 			case PF_ADDR_TABLE:
2371 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
2372 				    newpa->addr.v.tblname);
2373 				if (newpa->addr.p.tbl == NULL)
2374 					error = ENOMEM;
2375 				break;
2376 			}
2377 			if (error)
2378 				goto DIOCCHANGEADDR_error;
2379 		}
2380 
2381 		switch (pca->action) {
2382 		case PF_CHANGE_ADD_HEAD:
2383 			oldpa = TAILQ_FIRST(&pool->list);
2384 			break;
2385 		case PF_CHANGE_ADD_TAIL:
2386 			oldpa = TAILQ_LAST(&pool->list, pf_palist);
2387 			break;
2388 		default:
2389 			oldpa = TAILQ_FIRST(&pool->list);
2390 			for (int i = 0; oldpa && i < pca->nr; i++)
2391 				oldpa = TAILQ_NEXT(oldpa, entries);
2392 
2393 			if (oldpa == NULL)
2394 				ERROUT(EINVAL);
2395 		}
2396 
2397 		if (pca->action == PF_CHANGE_REMOVE) {
2398 			TAILQ_REMOVE(&pool->list, oldpa, entries);
2399 			switch (oldpa->addr.type) {
2400 			case PF_ADDR_DYNIFTL:
2401 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
2402 				break;
2403 			case PF_ADDR_TABLE:
2404 				pfr_detach_table(oldpa->addr.p.tbl);
2405 				break;
2406 			}
2407 			if (oldpa->kif)
2408 				pfi_kif_unref(oldpa->kif);
2409 			free(oldpa, M_PFRULE);
2410 		} else {
2411 			if (oldpa == NULL)
2412 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2413 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
2414 			    pca->action == PF_CHANGE_ADD_BEFORE)
2415 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2416 			else
2417 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
2418 				    newpa, entries);
2419 		}
2420 
2421 		pool->cur = TAILQ_FIRST(&pool->list);
2422 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
2423 		PF_RULES_WUNLOCK();
2424 		break;
2425 
2426 #undef ERROUT
2427 DIOCCHANGEADDR_error:
2428 		if (newpa->kif)
2429 			pfi_kif_unref(newpa->kif);
2430 		PF_RULES_WUNLOCK();
2431 		if (newpa != NULL)
2432 			free(newpa, M_PFRULE);
2433 		if (kif != NULL)
2434 			free(kif, PFI_MTYPE);
2435 		break;
2436 	}
2437 
2438 	case DIOCGETRULESETS: {
2439 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2440 		struct pf_ruleset	*ruleset;
2441 		struct pf_anchor	*anchor;
2442 
2443 		PF_RULES_RLOCK();
2444 		pr->path[sizeof(pr->path) - 1] = 0;
2445 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2446 			PF_RULES_RUNLOCK();
2447 			error = ENOENT;
2448 			break;
2449 		}
2450 		pr->nr = 0;
2451 		if (ruleset->anchor == NULL) {
2452 			/* XXX kludge for pf_main_ruleset */
2453 			RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2454 				if (anchor->parent == NULL)
2455 					pr->nr++;
2456 		} else {
2457 			RB_FOREACH(anchor, pf_anchor_node,
2458 			    &ruleset->anchor->children)
2459 				pr->nr++;
2460 		}
2461 		PF_RULES_RUNLOCK();
2462 		break;
2463 	}
2464 
2465 	case DIOCGETRULESET: {
2466 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2467 		struct pf_ruleset	*ruleset;
2468 		struct pf_anchor	*anchor;
2469 		u_int32_t		 nr = 0;
2470 
2471 		PF_RULES_RLOCK();
2472 		pr->path[sizeof(pr->path) - 1] = 0;
2473 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2474 			PF_RULES_RUNLOCK();
2475 			error = ENOENT;
2476 			break;
2477 		}
2478 		pr->name[0] = 0;
2479 		if (ruleset->anchor == NULL) {
2480 			/* XXX kludge for pf_main_ruleset */
2481 			RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2482 				if (anchor->parent == NULL && nr++ == pr->nr) {
2483 					strlcpy(pr->name, anchor->name,
2484 					    sizeof(pr->name));
2485 					break;
2486 				}
2487 		} else {
2488 			RB_FOREACH(anchor, pf_anchor_node,
2489 			    &ruleset->anchor->children)
2490 				if (nr++ == pr->nr) {
2491 					strlcpy(pr->name, anchor->name,
2492 					    sizeof(pr->name));
2493 					break;
2494 				}
2495 		}
2496 		if (!pr->name[0])
2497 			error = EBUSY;
2498 		PF_RULES_RUNLOCK();
2499 		break;
2500 	}
2501 
2502 	case DIOCRCLRTABLES: {
2503 		struct pfioc_table *io = (struct pfioc_table *)addr;
2504 
2505 		if (io->pfrio_esize != 0) {
2506 			error = ENODEV;
2507 			break;
2508 		}
2509 		PF_RULES_WLOCK();
2510 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2511 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2512 		PF_RULES_WUNLOCK();
2513 		break;
2514 	}
2515 
2516 	case DIOCRADDTABLES: {
2517 		struct pfioc_table *io = (struct pfioc_table *)addr;
2518 		struct pfr_table *pfrts;
2519 		size_t totlen;
2520 
2521 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2522 			error = ENODEV;
2523 			break;
2524 		}
2525 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2526 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2527 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2528 		if (error) {
2529 			free(pfrts, M_TEMP);
2530 			break;
2531 		}
2532 		PF_RULES_WLOCK();
2533 		error = pfr_add_tables(pfrts, io->pfrio_size,
2534 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2535 		PF_RULES_WUNLOCK();
2536 		free(pfrts, M_TEMP);
2537 		break;
2538 	}
2539 
2540 	case DIOCRDELTABLES: {
2541 		struct pfioc_table *io = (struct pfioc_table *)addr;
2542 		struct pfr_table *pfrts;
2543 		size_t totlen;
2544 
2545 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2546 			error = ENODEV;
2547 			break;
2548 		}
2549 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2550 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2551 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2552 		if (error) {
2553 			free(pfrts, M_TEMP);
2554 			break;
2555 		}
2556 		PF_RULES_WLOCK();
2557 		error = pfr_del_tables(pfrts, io->pfrio_size,
2558 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2559 		PF_RULES_WUNLOCK();
2560 		free(pfrts, M_TEMP);
2561 		break;
2562 	}
2563 
2564 	case DIOCRGETTABLES: {
2565 		struct pfioc_table *io = (struct pfioc_table *)addr;
2566 		struct pfr_table *pfrts;
2567 		size_t totlen;
2568 
2569 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2570 			error = ENODEV;
2571 			break;
2572 		}
2573 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2574 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2575 		PF_RULES_RLOCK();
2576 		error = pfr_get_tables(&io->pfrio_table, pfrts,
2577 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2578 		PF_RULES_RUNLOCK();
2579 		if (error == 0)
2580 			error = copyout(pfrts, io->pfrio_buffer, totlen);
2581 		free(pfrts, M_TEMP);
2582 		break;
2583 	}
2584 
2585 	case DIOCRGETTSTATS: {
2586 		struct pfioc_table *io = (struct pfioc_table *)addr;
2587 		struct pfr_tstats *pfrtstats;
2588 		size_t totlen;
2589 
2590 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2591 			error = ENODEV;
2592 			break;
2593 		}
2594 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
2595 		pfrtstats = malloc(totlen, M_TEMP, M_WAITOK);
2596 		PF_RULES_WLOCK();
2597 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
2598 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2599 		PF_RULES_WUNLOCK();
2600 		if (error == 0)
2601 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
2602 		free(pfrtstats, M_TEMP);
2603 		break;
2604 	}
2605 
2606 	case DIOCRCLRTSTATS: {
2607 		struct pfioc_table *io = (struct pfioc_table *)addr;
2608 		struct pfr_table *pfrts;
2609 		size_t totlen;
2610 
2611 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2612 			error = ENODEV;
2613 			break;
2614 		}
2615 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2616 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2617 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2618 		if (error) {
2619 			free(pfrts, M_TEMP);
2620 			break;
2621 		}
2622 		PF_RULES_WLOCK();
2623 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
2624 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2625 		PF_RULES_WUNLOCK();
2626 		free(pfrts, M_TEMP);
2627 		break;
2628 	}
2629 
2630 	case DIOCRSETTFLAGS: {
2631 		struct pfioc_table *io = (struct pfioc_table *)addr;
2632 		struct pfr_table *pfrts;
2633 		size_t totlen;
2634 
2635 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2636 			error = ENODEV;
2637 			break;
2638 		}
2639 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2640 		pfrts = malloc(totlen, M_TEMP, M_WAITOK);
2641 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2642 		if (error) {
2643 			free(pfrts, M_TEMP);
2644 			break;
2645 		}
2646 		PF_RULES_WLOCK();
2647 		error = pfr_set_tflags(pfrts, io->pfrio_size,
2648 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2649 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2650 		PF_RULES_WUNLOCK();
2651 		free(pfrts, M_TEMP);
2652 		break;
2653 	}
2654 
2655 	case DIOCRCLRADDRS: {
2656 		struct pfioc_table *io = (struct pfioc_table *)addr;
2657 
2658 		if (io->pfrio_esize != 0) {
2659 			error = ENODEV;
2660 			break;
2661 		}
2662 		PF_RULES_WLOCK();
2663 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2664 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2665 		PF_RULES_WUNLOCK();
2666 		break;
2667 	}
2668 
2669 	case DIOCRADDADDRS: {
2670 		struct pfioc_table *io = (struct pfioc_table *)addr;
2671 		struct pfr_addr *pfras;
2672 		size_t totlen;
2673 
2674 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2675 			error = ENODEV;
2676 			break;
2677 		}
2678 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2679 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2680 		error = copyin(io->pfrio_buffer, pfras, totlen);
2681 		if (error) {
2682 			free(pfras, M_TEMP);
2683 			break;
2684 		}
2685 		PF_RULES_WLOCK();
2686 		error = pfr_add_addrs(&io->pfrio_table, pfras,
2687 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2688 		    PFR_FLAG_USERIOCTL);
2689 		PF_RULES_WUNLOCK();
2690 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2691 			error = copyout(pfras, io->pfrio_buffer, totlen);
2692 		free(pfras, M_TEMP);
2693 		break;
2694 	}
2695 
2696 	case DIOCRDELADDRS: {
2697 		struct pfioc_table *io = (struct pfioc_table *)addr;
2698 		struct pfr_addr *pfras;
2699 		size_t totlen;
2700 
2701 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2702 			error = ENODEV;
2703 			break;
2704 		}
2705 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2706 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2707 		error = copyin(io->pfrio_buffer, pfras, totlen);
2708 		if (error) {
2709 			free(pfras, M_TEMP);
2710 			break;
2711 		}
2712 		PF_RULES_WLOCK();
2713 		error = pfr_del_addrs(&io->pfrio_table, pfras,
2714 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2715 		    PFR_FLAG_USERIOCTL);
2716 		PF_RULES_WUNLOCK();
2717 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2718 			error = copyout(pfras, io->pfrio_buffer, totlen);
2719 		free(pfras, M_TEMP);
2720 		break;
2721 	}
2722 
2723 	case DIOCRSETADDRS: {
2724 		struct pfioc_table *io = (struct pfioc_table *)addr;
2725 		struct pfr_addr *pfras;
2726 		size_t totlen;
2727 
2728 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2729 			error = ENODEV;
2730 			break;
2731 		}
2732 		totlen = (io->pfrio_size + io->pfrio_size2) *
2733 		    sizeof(struct pfr_addr);
2734 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2735 		error = copyin(io->pfrio_buffer, pfras, totlen);
2736 		if (error) {
2737 			free(pfras, M_TEMP);
2738 			break;
2739 		}
2740 		PF_RULES_WLOCK();
2741 		error = pfr_set_addrs(&io->pfrio_table, pfras,
2742 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2743 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2744 		    PFR_FLAG_USERIOCTL, 0);
2745 		PF_RULES_WUNLOCK();
2746 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2747 			error = copyout(pfras, io->pfrio_buffer, totlen);
2748 		free(pfras, M_TEMP);
2749 		break;
2750 	}
2751 
2752 	case DIOCRGETADDRS: {
2753 		struct pfioc_table *io = (struct pfioc_table *)addr;
2754 		struct pfr_addr *pfras;
2755 		size_t totlen;
2756 
2757 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2758 			error = ENODEV;
2759 			break;
2760 		}
2761 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2762 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2763 		PF_RULES_RLOCK();
2764 		error = pfr_get_addrs(&io->pfrio_table, pfras,
2765 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2766 		PF_RULES_RUNLOCK();
2767 		if (error == 0)
2768 			error = copyout(pfras, io->pfrio_buffer, totlen);
2769 		free(pfras, M_TEMP);
2770 		break;
2771 	}
2772 
2773 	case DIOCRGETASTATS: {
2774 		struct pfioc_table *io = (struct pfioc_table *)addr;
2775 		struct pfr_astats *pfrastats;
2776 		size_t totlen;
2777 
2778 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2779 			error = ENODEV;
2780 			break;
2781 		}
2782 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
2783 		pfrastats = malloc(totlen, M_TEMP, M_WAITOK);
2784 		PF_RULES_RLOCK();
2785 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
2786 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2787 		PF_RULES_RUNLOCK();
2788 		if (error == 0)
2789 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
2790 		free(pfrastats, M_TEMP);
2791 		break;
2792 	}
2793 
2794 	case DIOCRCLRASTATS: {
2795 		struct pfioc_table *io = (struct pfioc_table *)addr;
2796 		struct pfr_addr *pfras;
2797 		size_t totlen;
2798 
2799 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2800 			error = ENODEV;
2801 			break;
2802 		}
2803 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2804 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2805 		error = copyin(io->pfrio_buffer, pfras, totlen);
2806 		if (error) {
2807 			free(pfras, M_TEMP);
2808 			break;
2809 		}
2810 		PF_RULES_WLOCK();
2811 		error = pfr_clr_astats(&io->pfrio_table, pfras,
2812 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2813 		    PFR_FLAG_USERIOCTL);
2814 		PF_RULES_WUNLOCK();
2815 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2816 			error = copyout(pfras, io->pfrio_buffer, totlen);
2817 		free(pfras, M_TEMP);
2818 		break;
2819 	}
2820 
2821 	case DIOCRTSTADDRS: {
2822 		struct pfioc_table *io = (struct pfioc_table *)addr;
2823 		struct pfr_addr *pfras;
2824 		size_t totlen;
2825 
2826 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2827 			error = ENODEV;
2828 			break;
2829 		}
2830 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2831 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2832 		error = copyin(io->pfrio_buffer, pfras, totlen);
2833 		if (error) {
2834 			free(pfras, M_TEMP);
2835 			break;
2836 		}
2837 		PF_RULES_RLOCK();
2838 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
2839 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2840 		    PFR_FLAG_USERIOCTL);
2841 		PF_RULES_RUNLOCK();
2842 		if (error == 0)
2843 			error = copyout(pfras, io->pfrio_buffer, totlen);
2844 		free(pfras, M_TEMP);
2845 		break;
2846 	}
2847 
2848 	case DIOCRINADEFINE: {
2849 		struct pfioc_table *io = (struct pfioc_table *)addr;
2850 		struct pfr_addr *pfras;
2851 		size_t totlen;
2852 
2853 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2854 			error = ENODEV;
2855 			break;
2856 		}
2857 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2858 		pfras = malloc(totlen, M_TEMP, M_WAITOK);
2859 		error = copyin(io->pfrio_buffer, pfras, totlen);
2860 		if (error) {
2861 			free(pfras, M_TEMP);
2862 			break;
2863 		}
2864 		PF_RULES_WLOCK();
2865 		error = pfr_ina_define(&io->pfrio_table, pfras,
2866 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2867 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2868 		PF_RULES_WUNLOCK();
2869 		free(pfras, M_TEMP);
2870 		break;
2871 	}
2872 
2873 	case DIOCOSFPADD: {
2874 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2875 		PF_RULES_WLOCK();
2876 		error = pf_osfp_add(io);
2877 		PF_RULES_WUNLOCK();
2878 		break;
2879 	}
2880 
2881 	case DIOCOSFPGET: {
2882 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2883 		PF_RULES_RLOCK();
2884 		error = pf_osfp_get(io);
2885 		PF_RULES_RUNLOCK();
2886 		break;
2887 	}
2888 
2889 	case DIOCXBEGIN: {
2890 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2891 		struct pfioc_trans_e	*ioes, *ioe;
2892 		size_t			 totlen;
2893 		int			 i;
2894 
2895 		if (io->esize != sizeof(*ioe)) {
2896 			error = ENODEV;
2897 			break;
2898 		}
2899 		totlen = sizeof(struct pfioc_trans_e) * io->size;
2900 		ioes = malloc(totlen, M_TEMP, M_WAITOK);
2901 		error = copyin(io->array, ioes, totlen);
2902 		if (error) {
2903 			free(ioes, M_TEMP);
2904 			break;
2905 		}
2906 		PF_RULES_WLOCK();
2907 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2908 			switch (ioe->rs_num) {
2909 #ifdef ALTQ
2910 			case PF_RULESET_ALTQ:
2911 				if (ioe->anchor[0]) {
2912 					PF_RULES_WUNLOCK();
2913 					free(ioes, M_TEMP);
2914 					error = EINVAL;
2915 					goto fail;
2916 				}
2917 				if ((error = pf_begin_altq(&ioe->ticket))) {
2918 					PF_RULES_WUNLOCK();
2919 					free(ioes, M_TEMP);
2920 					goto fail;
2921 				}
2922 				break;
2923 #endif /* ALTQ */
2924 			case PF_RULESET_TABLE:
2925 			    {
2926 				struct pfr_table table;
2927 
2928 				bzero(&table, sizeof(table));
2929 				strlcpy(table.pfrt_anchor, ioe->anchor,
2930 				    sizeof(table.pfrt_anchor));
2931 				if ((error = pfr_ina_begin(&table,
2932 				    &ioe->ticket, NULL, 0))) {
2933 					PF_RULES_WUNLOCK();
2934 					free(ioes, M_TEMP);
2935 					goto fail;
2936 				}
2937 				break;
2938 			    }
2939 			default:
2940 				if ((error = pf_begin_rules(&ioe->ticket,
2941 				    ioe->rs_num, ioe->anchor))) {
2942 					PF_RULES_WUNLOCK();
2943 					free(ioes, M_TEMP);
2944 					goto fail;
2945 				}
2946 				break;
2947 			}
2948 		}
2949 		PF_RULES_WUNLOCK();
2950 		error = copyout(ioes, io->array, totlen);
2951 		free(ioes, M_TEMP);
2952 		break;
2953 	}
2954 
2955 	case DIOCXROLLBACK: {
2956 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2957 		struct pfioc_trans_e	*ioe, *ioes;
2958 		size_t			 totlen;
2959 		int			 i;
2960 
2961 		if (io->esize != sizeof(*ioe)) {
2962 			error = ENODEV;
2963 			break;
2964 		}
2965 		totlen = sizeof(struct pfioc_trans_e) * io->size;
2966 		ioes = malloc(totlen, M_TEMP, M_WAITOK);
2967 		error = copyin(io->array, ioes, totlen);
2968 		if (error) {
2969 			free(ioes, M_TEMP);
2970 			break;
2971 		}
2972 		PF_RULES_WLOCK();
2973 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
2974 			switch (ioe->rs_num) {
2975 #ifdef ALTQ
2976 			case PF_RULESET_ALTQ:
2977 				if (ioe->anchor[0]) {
2978 					PF_RULES_WUNLOCK();
2979 					free(ioes, M_TEMP);
2980 					error = EINVAL;
2981 					goto fail;
2982 				}
2983 				if ((error = pf_rollback_altq(ioe->ticket))) {
2984 					PF_RULES_WUNLOCK();
2985 					free(ioes, M_TEMP);
2986 					goto fail; /* really bad */
2987 				}
2988 				break;
2989 #endif /* ALTQ */
2990 			case PF_RULESET_TABLE:
2991 			    {
2992 				struct pfr_table table;
2993 
2994 				bzero(&table, sizeof(table));
2995 				strlcpy(table.pfrt_anchor, ioe->anchor,
2996 				    sizeof(table.pfrt_anchor));
2997 				if ((error = pfr_ina_rollback(&table,
2998 				    ioe->ticket, NULL, 0))) {
2999 					PF_RULES_WUNLOCK();
3000 					free(ioes, M_TEMP);
3001 					goto fail; /* really bad */
3002 				}
3003 				break;
3004 			    }
3005 			default:
3006 				if ((error = pf_rollback_rules(ioe->ticket,
3007 				    ioe->rs_num, ioe->anchor))) {
3008 					PF_RULES_WUNLOCK();
3009 					free(ioes, M_TEMP);
3010 					goto fail; /* really bad */
3011 				}
3012 				break;
3013 			}
3014 		}
3015 		PF_RULES_WUNLOCK();
3016 		free(ioes, M_TEMP);
3017 		break;
3018 	}
3019 
3020 	case DIOCXCOMMIT: {
3021 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3022 		struct pfioc_trans_e	*ioe, *ioes;
3023 		struct pf_ruleset	*rs;
3024 		size_t			 totlen;
3025 		int			 i;
3026 
3027 		if (io->esize != sizeof(*ioe)) {
3028 			error = ENODEV;
3029 			break;
3030 		}
3031 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3032 		ioes = malloc(totlen, M_TEMP, M_WAITOK);
3033 		error = copyin(io->array, ioes, totlen);
3034 		if (error) {
3035 			free(ioes, M_TEMP);
3036 			break;
3037 		}
3038 		PF_RULES_WLOCK();
3039 		/* First makes sure everything will succeed. */
3040 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3041 			switch (ioe->rs_num) {
3042 #ifdef ALTQ
3043 			case PF_RULESET_ALTQ:
3044 				if (ioe->anchor[0]) {
3045 					PF_RULES_WUNLOCK();
3046 					free(ioes, M_TEMP);
3047 					error = EINVAL;
3048 					goto fail;
3049 				}
3050 				if (!V_altqs_inactive_open || ioe->ticket !=
3051 				    V_ticket_altqs_inactive) {
3052 					PF_RULES_WUNLOCK();
3053 					free(ioes, M_TEMP);
3054 					error = EBUSY;
3055 					goto fail;
3056 				}
3057 				break;
3058 #endif /* ALTQ */
3059 			case PF_RULESET_TABLE:
3060 				rs = pf_find_ruleset(ioe->anchor);
3061 				if (rs == NULL || !rs->topen || ioe->ticket !=
3062 				    rs->tticket) {
3063 					PF_RULES_WUNLOCK();
3064 					free(ioes, M_TEMP);
3065 					error = EBUSY;
3066 					goto fail;
3067 				}
3068 				break;
3069 			default:
3070 				if (ioe->rs_num < 0 || ioe->rs_num >=
3071 				    PF_RULESET_MAX) {
3072 					PF_RULES_WUNLOCK();
3073 					free(ioes, M_TEMP);
3074 					error = EINVAL;
3075 					goto fail;
3076 				}
3077 				rs = pf_find_ruleset(ioe->anchor);
3078 				if (rs == NULL ||
3079 				    !rs->rules[ioe->rs_num].inactive.open ||
3080 				    rs->rules[ioe->rs_num].inactive.ticket !=
3081 				    ioe->ticket) {
3082 					PF_RULES_WUNLOCK();
3083 					free(ioes, M_TEMP);
3084 					error = EBUSY;
3085 					goto fail;
3086 				}
3087 				break;
3088 			}
3089 		}
3090 		/* Now do the commit - no errors should happen here. */
3091 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3092 			switch (ioe->rs_num) {
3093 #ifdef ALTQ
3094 			case PF_RULESET_ALTQ:
3095 				if ((error = pf_commit_altq(ioe->ticket))) {
3096 					PF_RULES_WUNLOCK();
3097 					free(ioes, M_TEMP);
3098 					goto fail; /* really bad */
3099 				}
3100 				break;
3101 #endif /* ALTQ */
3102 			case PF_RULESET_TABLE:
3103 			    {
3104 				struct pfr_table table;
3105 
3106 				bzero(&table, sizeof(table));
3107 				strlcpy(table.pfrt_anchor, ioe->anchor,
3108 				    sizeof(table.pfrt_anchor));
3109 				if ((error = pfr_ina_commit(&table,
3110 				    ioe->ticket, NULL, NULL, 0))) {
3111 					PF_RULES_WUNLOCK();
3112 					free(ioes, M_TEMP);
3113 					goto fail; /* really bad */
3114 				}
3115 				break;
3116 			    }
3117 			default:
3118 				if ((error = pf_commit_rules(ioe->ticket,
3119 				    ioe->rs_num, ioe->anchor))) {
3120 					PF_RULES_WUNLOCK();
3121 					free(ioes, M_TEMP);
3122 					goto fail; /* really bad */
3123 				}
3124 				break;
3125 			}
3126 		}
3127 		PF_RULES_WUNLOCK();
3128 		free(ioes, M_TEMP);
3129 		break;
3130 	}
3131 
3132 	case DIOCGETSRCNODES: {
3133 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
3134 		struct pf_srchash	*sh;
3135 		struct pf_src_node	*n, *p, *pstore;
3136 		uint32_t		 i, nr = 0;
3137 
3138 		if (psn->psn_len == 0) {
3139 			for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3140 			    i++, sh++) {
3141 				PF_HASHROW_LOCK(sh);
3142 				LIST_FOREACH(n, &sh->nodes, entry)
3143 					nr++;
3144 				PF_HASHROW_UNLOCK(sh);
3145 			}
3146 			psn->psn_len = sizeof(struct pf_src_node) * nr;
3147 			break;
3148 		}
3149 
3150 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK);
3151 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3152 		    i++, sh++) {
3153 		    PF_HASHROW_LOCK(sh);
3154 		    LIST_FOREACH(n, &sh->nodes, entry) {
3155 			int	secs = time_uptime, diff;
3156 
3157 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
3158 				break;
3159 
3160 			bcopy(n, p, sizeof(struct pf_src_node));
3161 			if (n->rule.ptr != NULL)
3162 				p->rule.nr = n->rule.ptr->nr;
3163 			p->creation = secs - p->creation;
3164 			if (p->expire > secs)
3165 				p->expire -= secs;
3166 			else
3167 				p->expire = 0;
3168 
3169 			/* Adjust the connection rate estimate. */
3170 			diff = secs - n->conn_rate.last;
3171 			if (diff >= n->conn_rate.seconds)
3172 				p->conn_rate.count = 0;
3173 			else
3174 				p->conn_rate.count -=
3175 				    n->conn_rate.count * diff /
3176 				    n->conn_rate.seconds;
3177 			p++;
3178 			nr++;
3179 		    }
3180 		    PF_HASHROW_UNLOCK(sh);
3181 		}
3182 		error = copyout(pstore, psn->psn_src_nodes,
3183 		    sizeof(struct pf_src_node) * nr);
3184 		if (error) {
3185 			free(pstore, M_TEMP);
3186 			break;
3187 		}
3188 		psn->psn_len = sizeof(struct pf_src_node) * nr;
3189 		free(pstore, M_TEMP);
3190 		break;
3191 	}
3192 
3193 	case DIOCCLRSRCNODES: {
3194 
3195 		pf_clear_srcnodes(NULL);
3196 		pf_purge_expired_src_nodes();
3197 		break;
3198 	}
3199 
3200 	case DIOCKILLSRCNODES:
3201 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
3202 		break;
3203 
3204 	case DIOCSETHOSTID: {
3205 		u_int32_t	*hostid = (u_int32_t *)addr;
3206 
3207 		PF_RULES_WLOCK();
3208 		if (*hostid == 0)
3209 			V_pf_status.hostid = arc4random();
3210 		else
3211 			V_pf_status.hostid = *hostid;
3212 		PF_RULES_WUNLOCK();
3213 		break;
3214 	}
3215 
3216 	case DIOCOSFPFLUSH:
3217 		PF_RULES_WLOCK();
3218 		pf_osfp_flush();
3219 		PF_RULES_WUNLOCK();
3220 		break;
3221 
3222 	case DIOCIGETIFACES: {
3223 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3224 		struct pfi_kif *ifstore;
3225 		size_t bufsiz;
3226 
3227 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
3228 			error = ENODEV;
3229 			break;
3230 		}
3231 
3232 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
3233 		ifstore = malloc(bufsiz, M_TEMP, M_WAITOK);
3234 		PF_RULES_RLOCK();
3235 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
3236 		PF_RULES_RUNLOCK();
3237 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
3238 		free(ifstore, M_TEMP);
3239 		break;
3240 	}
3241 
3242 	case DIOCSETIFFLAG: {
3243 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3244 
3245 		PF_RULES_WLOCK();
3246 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3247 		PF_RULES_WUNLOCK();
3248 		break;
3249 	}
3250 
3251 	case DIOCCLRIFFLAG: {
3252 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3253 
3254 		PF_RULES_WLOCK();
3255 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3256 		PF_RULES_WUNLOCK();
3257 		break;
3258 	}
3259 
3260 	default:
3261 		error = ENODEV;
3262 		break;
3263 	}
3264 fail:
3265 	if (sx_xlocked(&pf_ioctl_lock))
3266 		sx_xunlock(&pf_ioctl_lock);
3267 	CURVNET_RESTORE();
3268 
3269 	return (error);
3270 }
3271 
3272 void
3273 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
3274 {
3275 	bzero(sp, sizeof(struct pfsync_state));
3276 
3277 	/* copy from state key */
3278 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
3279 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
3280 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
3281 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
3282 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
3283 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
3284 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
3285 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
3286 	sp->proto = st->key[PF_SK_WIRE]->proto;
3287 	sp->af = st->key[PF_SK_WIRE]->af;
3288 
3289 	/* copy from state */
3290 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
3291 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
3292 	sp->creation = htonl(time_uptime - st->creation);
3293 	sp->expire = pf_state_expires(st);
3294 	if (sp->expire <= time_uptime)
3295 		sp->expire = htonl(0);
3296 	else
3297 		sp->expire = htonl(sp->expire - time_uptime);
3298 
3299 	sp->direction = st->direction;
3300 	sp->log = st->log;
3301 	sp->timeout = st->timeout;
3302 	sp->state_flags = st->state_flags;
3303 	if (st->src_node)
3304 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
3305 	if (st->nat_src_node)
3306 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
3307 
3308 	sp->id = st->id;
3309 	sp->creatorid = st->creatorid;
3310 	pf_state_peer_hton(&st->src, &sp->src);
3311 	pf_state_peer_hton(&st->dst, &sp->dst);
3312 
3313 	if (st->rule.ptr == NULL)
3314 		sp->rule = htonl(-1);
3315 	else
3316 		sp->rule = htonl(st->rule.ptr->nr);
3317 	if (st->anchor.ptr == NULL)
3318 		sp->anchor = htonl(-1);
3319 	else
3320 		sp->anchor = htonl(st->anchor.ptr->nr);
3321 	if (st->nat_rule.ptr == NULL)
3322 		sp->nat_rule = htonl(-1);
3323 	else
3324 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
3325 
3326 	pf_state_counter_hton(st->packets[0], sp->packets[0]);
3327 	pf_state_counter_hton(st->packets[1], sp->packets[1]);
3328 	pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
3329 	pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
3330 
3331 }
3332 
3333 static void
3334 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
3335 {
3336 	struct pfr_ktable *kt;
3337 
3338 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
3339 
3340 	kt = aw->p.tbl;
3341 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
3342 		kt = kt->pfrkt_root;
3343 	aw->p.tbl = NULL;
3344 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
3345 		kt->pfrkt_cnt : -1;
3346 }
3347 
3348 /*
3349  * XXX - Check for version missmatch!!!
3350  */
3351 static void
3352 pf_clear_states(void)
3353 {
3354 	struct pf_state	*s;
3355 	u_int i;
3356 
3357 	for (i = 0; i <= pf_hashmask; i++) {
3358 		struct pf_idhash *ih = &V_pf_idhash[i];
3359 relock:
3360 		PF_HASHROW_LOCK(ih);
3361 		LIST_FOREACH(s, &ih->states, entry) {
3362 			s->timeout = PFTM_PURGE;
3363 			/* Don't send out individual delete messages. */
3364 			s->sync_state = PFSTATE_NOSYNC;
3365 			pf_unlink_state(s, PF_ENTER_LOCKED);
3366 			goto relock;
3367 		}
3368 		PF_HASHROW_UNLOCK(ih);
3369 	}
3370 }
3371 
3372 static int
3373 pf_clear_tables(void)
3374 {
3375 	struct pfioc_table io;
3376 	int error;
3377 
3378 	bzero(&io, sizeof(io));
3379 
3380 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
3381 	    io.pfrio_flags);
3382 
3383 	return (error);
3384 }
3385 
3386 static void
3387 pf_clear_srcnodes(struct pf_src_node *n)
3388 {
3389 	struct pf_state *s;
3390 	int i;
3391 
3392 	for (i = 0; i <= pf_hashmask; i++) {
3393 		struct pf_idhash *ih = &V_pf_idhash[i];
3394 
3395 		PF_HASHROW_LOCK(ih);
3396 		LIST_FOREACH(s, &ih->states, entry) {
3397 			if (n == NULL || n == s->src_node)
3398 				s->src_node = NULL;
3399 			if (n == NULL || n == s->nat_src_node)
3400 				s->nat_src_node = NULL;
3401 		}
3402 		PF_HASHROW_UNLOCK(ih);
3403 	}
3404 
3405 	if (n == NULL) {
3406 		struct pf_srchash *sh;
3407 
3408 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3409 		    i++, sh++) {
3410 			PF_HASHROW_LOCK(sh);
3411 			LIST_FOREACH(n, &sh->nodes, entry) {
3412 				n->expire = 1;
3413 				n->states = 0;
3414 			}
3415 			PF_HASHROW_UNLOCK(sh);
3416 		}
3417 	} else {
3418 		/* XXX: hash slot should already be locked here. */
3419 		n->expire = 1;
3420 		n->states = 0;
3421 	}
3422 }
3423 
3424 static void
3425 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
3426 {
3427 	struct pf_src_node_list	 kill;
3428 
3429 	LIST_INIT(&kill);
3430 	for (int i = 0; i <= pf_srchashmask; i++) {
3431 		struct pf_srchash *sh = &V_pf_srchash[i];
3432 		struct pf_src_node *sn, *tmp;
3433 
3434 		PF_HASHROW_LOCK(sh);
3435 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
3436 			if (PF_MATCHA(psnk->psnk_src.neg,
3437 			      &psnk->psnk_src.addr.v.a.addr,
3438 			      &psnk->psnk_src.addr.v.a.mask,
3439 			      &sn->addr, sn->af) &&
3440 			    PF_MATCHA(psnk->psnk_dst.neg,
3441 			      &psnk->psnk_dst.addr.v.a.addr,
3442 			      &psnk->psnk_dst.addr.v.a.mask,
3443 			      &sn->raddr, sn->af)) {
3444 				pf_unlink_src_node_locked(sn);
3445 				LIST_INSERT_HEAD(&kill, sn, entry);
3446 				sn->expire = 1;
3447 			}
3448 		PF_HASHROW_UNLOCK(sh);
3449 	}
3450 
3451 	for (int i = 0; i <= pf_hashmask; i++) {
3452 		struct pf_idhash *ih = &V_pf_idhash[i];
3453 		struct pf_state *s;
3454 
3455 		PF_HASHROW_LOCK(ih);
3456 		LIST_FOREACH(s, &ih->states, entry) {
3457 			if (s->src_node && s->src_node->expire == 1) {
3458 #ifdef INVARIANTS
3459 				s->src_node->states--;
3460 #endif
3461 				s->src_node = NULL;
3462 			}
3463 			if (s->nat_src_node && s->nat_src_node->expire == 1) {
3464 #ifdef INVARIANTS
3465 				s->nat_src_node->states--;
3466 #endif
3467 				s->nat_src_node = NULL;
3468 			}
3469 		}
3470 		PF_HASHROW_UNLOCK(ih);
3471 	}
3472 
3473 	psnk->psnk_killed = pf_free_src_nodes(&kill);
3474 }
3475 
3476 /*
3477  * XXX - Check for version missmatch!!!
3478  */
3479 
3480 /*
3481  * Duplicate pfctl -Fa operation to get rid of as much as we can.
3482  */
3483 static int
3484 shutdown_pf(void)
3485 {
3486 	int error = 0;
3487 	u_int32_t t[5];
3488 	char nn = '\0';
3489 
3490 	V_pf_status.running = 0;
3491 
3492 	counter_u64_free(V_pf_default_rule.states_cur);
3493 	counter_u64_free(V_pf_default_rule.states_tot);
3494 	counter_u64_free(V_pf_default_rule.src_nodes);
3495 
3496 	for (int i = 0; i < PFRES_MAX; i++)
3497 		counter_u64_free(V_pf_status.counters[i]);
3498 	for (int i = 0; i < LCNT_MAX; i++)
3499 		counter_u64_free(V_pf_status.lcounters[i]);
3500 	for (int i = 0; i < FCNT_MAX; i++)
3501 		counter_u64_free(V_pf_status.fcounters[i]);
3502 	for (int i = 0; i < SCNT_MAX; i++)
3503 		counter_u64_free(V_pf_status.scounters[i]);
3504 
3505 	do {
3506 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
3507 		    != 0) {
3508 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
3509 			break;
3510 		}
3511 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
3512 		    != 0) {
3513 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
3514 			break;		/* XXX: rollback? */
3515 		}
3516 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
3517 		    != 0) {
3518 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
3519 			break;		/* XXX: rollback? */
3520 		}
3521 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
3522 		    != 0) {
3523 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
3524 			break;		/* XXX: rollback? */
3525 		}
3526 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
3527 		    != 0) {
3528 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
3529 			break;		/* XXX: rollback? */
3530 		}
3531 
3532 		/* XXX: these should always succeed here */
3533 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
3534 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
3535 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
3536 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
3537 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
3538 
3539 		if ((error = pf_clear_tables()) != 0)
3540 			break;
3541 
3542 #ifdef ALTQ
3543 		if ((error = pf_begin_altq(&t[0])) != 0) {
3544 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
3545 			break;
3546 		}
3547 		pf_commit_altq(t[0]);
3548 #endif
3549 
3550 		pf_clear_states();
3551 
3552 		pf_clear_srcnodes(NULL);
3553 
3554 		/* status does not use malloced mem so no need to cleanup */
3555 		/* fingerprints and interfaces have thier own cleanup code */
3556 	} while(0);
3557 
3558 	return (error);
3559 }
3560 
3561 #ifdef INET
3562 static int
3563 pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3564     struct inpcb *inp)
3565 {
3566 	int chk;
3567 
3568 	chk = pf_test(PF_IN, ifp, m, inp);
3569 	if (chk && *m) {
3570 		m_freem(*m);
3571 		*m = NULL;
3572 	}
3573 
3574 	return (chk);
3575 }
3576 
3577 static int
3578 pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3579     struct inpcb *inp)
3580 {
3581 	int chk;
3582 
3583 	/* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3584 	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3585 		in_delayed_cksum(*m);
3586 		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3587 	}
3588 
3589 	chk = pf_test(PF_OUT, ifp, m, inp);
3590 	if (chk && *m) {
3591 		m_freem(*m);
3592 		*m = NULL;
3593 	}
3594 
3595 	return (chk);
3596 }
3597 #endif
3598 
3599 #ifdef INET6
3600 static int
3601 pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3602     struct inpcb *inp)
3603 {
3604 	int chk;
3605 
3606 	/*
3607 	 * In case of loopback traffic IPv6 uses the real interface in
3608 	 * order to support scoped addresses. In order to support stateful
3609 	 * filtering we have change this to lo0 as it is the case in IPv4.
3610 	 */
3611 	CURVNET_SET(ifp->if_vnet);
3612 	chk = pf_test6(PF_IN, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
3613 	CURVNET_RESTORE();
3614 	if (chk && *m) {
3615 		m_freem(*m);
3616 		*m = NULL;
3617 	}
3618 	return chk;
3619 }
3620 
3621 static int
3622 pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3623     struct inpcb *inp)
3624 {
3625 	int chk;
3626 
3627 	/* We need a proper CSUM before we start (s. OpenBSD ip_output) */
3628 	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3629 		in6_delayed_cksum(*m,
3630 		    (*m)->m_pkthdr.len - sizeof(struct ip6_hdr),
3631 		    sizeof(struct ip6_hdr));
3632 		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
3633 	}
3634 	CURVNET_SET(ifp->if_vnet);
3635 	chk = pf_test6(PF_OUT, ifp, m, inp);
3636 	CURVNET_RESTORE();
3637 	if (chk && *m) {
3638 		m_freem(*m);
3639 		*m = NULL;
3640 	}
3641 	return chk;
3642 }
3643 #endif /* INET6 */
3644 
3645 static int
3646 hook_pf(void)
3647 {
3648 #ifdef INET
3649 	struct pfil_head *pfh_inet;
3650 #endif
3651 #ifdef INET6
3652 	struct pfil_head *pfh_inet6;
3653 #endif
3654 
3655 	if (V_pf_pfil_hooked)
3656 		return (0);
3657 
3658 #ifdef INET
3659 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3660 	if (pfh_inet == NULL)
3661 		return (ESRCH); /* XXX */
3662 	pfil_add_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet);
3663 	pfil_add_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet);
3664 #endif
3665 #ifdef INET6
3666 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3667 	if (pfh_inet6 == NULL) {
3668 #ifdef INET
3669 		pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3670 		    pfh_inet);
3671 		pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3672 		    pfh_inet);
3673 #endif
3674 		return (ESRCH); /* XXX */
3675 	}
3676 	pfil_add_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet6);
3677 	pfil_add_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet6);
3678 #endif
3679 
3680 	V_pf_pfil_hooked = 1;
3681 	return (0);
3682 }
3683 
3684 static int
3685 dehook_pf(void)
3686 {
3687 #ifdef INET
3688 	struct pfil_head *pfh_inet;
3689 #endif
3690 #ifdef INET6
3691 	struct pfil_head *pfh_inet6;
3692 #endif
3693 
3694 	if (V_pf_pfil_hooked == 0)
3695 		return (0);
3696 
3697 #ifdef INET
3698 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3699 	if (pfh_inet == NULL)
3700 		return (ESRCH); /* XXX */
3701 	pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3702 	    pfh_inet);
3703 	pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3704 	    pfh_inet);
3705 #endif
3706 #ifdef INET6
3707 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3708 	if (pfh_inet6 == NULL)
3709 		return (ESRCH); /* XXX */
3710 	pfil_remove_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK,
3711 	    pfh_inet6);
3712 	pfil_remove_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK,
3713 	    pfh_inet6);
3714 #endif
3715 
3716 	V_pf_pfil_hooked = 0;
3717 	return (0);
3718 }
3719 
3720 static int
3721 pf_load(void)
3722 {
3723 
3724 	rw_init(&pf_rules_lock, "pf rulesets");
3725 	sx_init(&pf_ioctl_lock, "pf ioctl");
3726 
3727 	pf_dev = make_dev(&pf_cdevsw, 0, 0, 0, 0600, PF_NAME);
3728 
3729 	return (0);
3730 }
3731 
3732 static int
3733 pf_unload(void)
3734 {
3735 	int error = 0;
3736 
3737 	V_pf_status.running = 0;
3738 	swi_remove(V_pf_swi_cookie);
3739 	error = dehook_pf();
3740 	if (error) {
3741 		/*
3742 		 * Should not happen!
3743 		 * XXX Due to error code ESRCH, kldunload will show
3744 		 * a message like 'No such process'.
3745 		 */
3746 		printf("%s : pfil unregisteration fail\n", __FUNCTION__);
3747 		return error;
3748 	}
3749 	PF_RULES_WLOCK();
3750 	shutdown_pf();
3751 	pf_normalize_cleanup();
3752 	pfi_cleanup();
3753 	pfr_cleanup();
3754 	pf_osfp_flush();
3755 	pf_cleanup();
3756 	if (IS_DEFAULT_VNET(curvnet))
3757 		pf_mtag_cleanup();
3758 	PF_RULES_WUNLOCK();
3759 	destroy_dev(pf_dev);
3760 	rw_destroy(&pf_rules_lock);
3761 	sx_destroy(&pf_ioctl_lock);
3762 
3763 	return (error);
3764 }
3765 
3766 static int
3767 pf_modevent(module_t mod, int type, void *data)
3768 {
3769 	int error = 0;
3770 
3771 	switch(type) {
3772 	case MOD_LOAD:
3773 		error = pf_load();
3774 		break;
3775 	case MOD_QUIESCE:
3776 		/*
3777 		 * Module should not be unloaded due to race conditions.
3778 		 */
3779 		error = EBUSY;
3780 		break;
3781 	case MOD_UNLOAD:
3782 		error = pf_unload();
3783 		break;
3784 	default:
3785 		error = EINVAL;
3786 		break;
3787 	}
3788 
3789 	return (error);
3790 }
3791 
3792 static moduledata_t pf_mod = {
3793 	"pf",
3794 	pf_modevent,
3795 	0
3796 };
3797 
3798 DECLARE_MODULE(pf, pf_mod, SI_SUB_PSEUDO, SI_ORDER_FIRST);
3799 MODULE_VERSION(pf, PF_MODVER);
3800 VNET_SYSINIT(pf_vnet_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY - 255,
3801 	    pf_vnet_init, NULL);
3802