xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision c5fda9bac0325eb8c5b447717862d279006f318f)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2001 Daniel Hartmeier
5  * Copyright (c) 2002,2003 Henning Brauer
6  * Copyright (c) 2012 Gleb Smirnoff <glebius@FreeBSD.org>
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  *
13  *    - Redistributions of source code must retain the above copyright
14  *      notice, this list of conditions and the following disclaimer.
15  *    - Redistributions in binary form must reproduce the above
16  *      copyright notice, this list of conditions and the following
17  *      disclaimer in the documentation and/or other materials provided
18  *      with the distribution.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
26  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
28  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
30  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
31  * POSSIBILITY OF SUCH DAMAGE.
32  *
33  * Effort sponsored in part by the Defense Advanced Research Projects
34  * Agency (DARPA) and Air Force Research Laboratory, Air Force
35  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
36  *
37  *	$OpenBSD: pf_ioctl.c,v 1.213 2009/02/15 21:46:12 mbalmer Exp $
38  */
39 
40 #include <sys/cdefs.h>
41 __FBSDID("$FreeBSD$");
42 
43 #include "opt_inet.h"
44 #include "opt_inet6.h"
45 #include "opt_bpf.h"
46 #include "opt_pf.h"
47 
48 #include <sys/param.h>
49 #include <sys/bus.h>
50 #include <sys/conf.h>
51 #include <sys/endian.h>
52 #include <sys/fcntl.h>
53 #include <sys/filio.h>
54 #include <sys/interrupt.h>
55 #include <sys/jail.h>
56 #include <sys/kernel.h>
57 #include <sys/kthread.h>
58 #include <sys/lock.h>
59 #include <sys/mbuf.h>
60 #include <sys/module.h>
61 #include <sys/proc.h>
62 #include <sys/smp.h>
63 #include <sys/socket.h>
64 #include <sys/sysctl.h>
65 #include <sys/md5.h>
66 #include <sys/ucred.h>
67 
68 #include <net/if.h>
69 #include <net/if_var.h>
70 #include <net/vnet.h>
71 #include <net/route.h>
72 #include <net/pfil.h>
73 #include <net/pfvar.h>
74 #include <net/if_pfsync.h>
75 #include <net/if_pflog.h>
76 
77 #include <netinet/in.h>
78 #include <netinet/ip.h>
79 #include <netinet/ip_var.h>
80 #include <netinet6/ip6_var.h>
81 #include <netinet/ip_icmp.h>
82 
83 #ifdef INET6
84 #include <netinet/ip6.h>
85 #endif /* INET6 */
86 
87 #ifdef ALTQ
88 #include <net/altq/altq.h>
89 #endif
90 
91 static struct pf_pool	*pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t,
92 			    u_int8_t, u_int8_t, u_int8_t);
93 
94 static void		 pf_mv_pool(struct pf_palist *, struct pf_palist *);
95 static void		 pf_empty_pool(struct pf_palist *);
96 static int		 pfioctl(struct cdev *, u_long, caddr_t, int,
97 			    struct thread *);
98 #ifdef ALTQ
99 static int		 pf_begin_altq(u_int32_t *);
100 static int		 pf_rollback_altq(u_int32_t);
101 static int		 pf_commit_altq(u_int32_t);
102 static int		 pf_enable_altq(struct pf_altq *);
103 static int		 pf_disable_altq(struct pf_altq *);
104 static u_int32_t	 pf_qname2qid(char *);
105 static void		 pf_qid_unref(u_int32_t);
106 #endif /* ALTQ */
107 static int		 pf_begin_rules(u_int32_t *, int, const char *);
108 static int		 pf_rollback_rules(u_int32_t, int, char *);
109 static int		 pf_setup_pfsync_matching(struct pf_ruleset *);
110 static void		 pf_hash_rule(MD5_CTX *, struct pf_rule *);
111 static void		 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
112 static int		 pf_commit_rules(u_int32_t, int, char *);
113 static int		 pf_addr_setup(struct pf_ruleset *,
114 			    struct pf_addr_wrap *, sa_family_t);
115 static void		 pf_addr_copyout(struct pf_addr_wrap *);
116 
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, int flags, struct inpcb *inp);
168 static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp,
169     int dir, int flags, 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, int flags, struct inpcb *inp);
174 static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp,
175     int dir, int flags, 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 void		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 /*
194  * We need a flag that is neither hooked nor running to know when
195  * the VNET is "valid".  We primarily need this to control (global)
196  * external event, e.g., eventhandlers.
197  */
198 VNET_DEFINE(int, pf_vnet_active);
199 #define V_pf_vnet_active	VNET(pf_vnet_active)
200 
201 int pf_end_threads;
202 struct proc *pf_purge_proc;
203 
204 struct rmlock			pf_rules_lock;
205 struct sx			pf_ioctl_lock;
206 struct sx			pf_end_lock;
207 
208 /* pfsync */
209 pfsync_state_import_t 		*pfsync_state_import_ptr = NULL;
210 pfsync_insert_state_t		*pfsync_insert_state_ptr = NULL;
211 pfsync_update_state_t		*pfsync_update_state_ptr = NULL;
212 pfsync_delete_state_t		*pfsync_delete_state_ptr = NULL;
213 pfsync_clear_states_t		*pfsync_clear_states_ptr = NULL;
214 pfsync_defer_t			*pfsync_defer_ptr = NULL;
215 /* pflog */
216 pflog_packet_t			*pflog_packet_ptr = NULL;
217 
218 extern u_long	pf_ioctl_maxcount;
219 
220 static void
221 pfattach_vnet(void)
222 {
223 	u_int32_t *my_timeout = V_pf_default_rule.timeout;
224 
225 	pf_initialize();
226 	pfr_initialize();
227 	pfi_initialize_vnet();
228 	pf_normalize_init();
229 
230 	V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
231 	V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
232 
233 	RB_INIT(&V_pf_anchors);
234 	pf_init_ruleset(&pf_main_ruleset);
235 
236 	/* default rule should never be garbage collected */
237 	V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
238 #ifdef PF_DEFAULT_TO_DROP
239 	V_pf_default_rule.action = PF_DROP;
240 #else
241 	V_pf_default_rule.action = PF_PASS;
242 #endif
243 	V_pf_default_rule.nr = -1;
244 	V_pf_default_rule.rtableid = -1;
245 
246 	V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
247 	V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
248 	V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
249 
250 	/* initialize default timeouts */
251 	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
252 	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
253 	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
254 	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
255 	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
256 	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
257 	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
258 	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
259 	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
260 	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
261 	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
262 	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
263 	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
264 	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
265 	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
266 	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
267 	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
268 	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
269 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
270 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
271 
272 	bzero(&V_pf_status, sizeof(V_pf_status));
273 	V_pf_status.debug = PF_DEBUG_URGENT;
274 
275 	V_pf_pfil_hooked = 0;
276 
277 	/* XXX do our best to avoid a conflict */
278 	V_pf_status.hostid = arc4random();
279 
280 	for (int i = 0; i < PFRES_MAX; i++)
281 		V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
282 	for (int i = 0; i < LCNT_MAX; i++)
283 		V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
284 	for (int i = 0; i < FCNT_MAX; i++)
285 		V_pf_status.fcounters[i] = counter_u64_alloc(M_WAITOK);
286 	for (int i = 0; i < SCNT_MAX; i++)
287 		V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
288 
289 	if (swi_add(NULL, "pf send", pf_intr, curvnet, SWI_NET,
290 	    INTR_MPSAFE, &V_pf_swi_cookie) != 0)
291 		/* XXXGL: leaked all above. */
292 		return;
293 }
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 	PF_RULES_RLOCK_TRACKER;
998 
999 	/* XXX keep in sync with switch() below */
1000 	if (securelevel_gt(td->td_ucred, 2))
1001 		switch (cmd) {
1002 		case DIOCGETRULES:
1003 		case DIOCGETRULE:
1004 		case DIOCGETADDRS:
1005 		case DIOCGETADDR:
1006 		case DIOCGETSTATE:
1007 		case DIOCSETSTATUSIF:
1008 		case DIOCGETSTATUS:
1009 		case DIOCCLRSTATUS:
1010 		case DIOCNATLOOK:
1011 		case DIOCSETDEBUG:
1012 		case DIOCGETSTATES:
1013 		case DIOCGETTIMEOUT:
1014 		case DIOCCLRRULECTRS:
1015 		case DIOCGETLIMIT:
1016 		case DIOCGETALTQS:
1017 		case DIOCGETALTQ:
1018 		case DIOCGETQSTATS:
1019 		case DIOCGETRULESETS:
1020 		case DIOCGETRULESET:
1021 		case DIOCRGETTABLES:
1022 		case DIOCRGETTSTATS:
1023 		case DIOCRCLRTSTATS:
1024 		case DIOCRCLRADDRS:
1025 		case DIOCRADDADDRS:
1026 		case DIOCRDELADDRS:
1027 		case DIOCRSETADDRS:
1028 		case DIOCRGETADDRS:
1029 		case DIOCRGETASTATS:
1030 		case DIOCRCLRASTATS:
1031 		case DIOCRTSTADDRS:
1032 		case DIOCOSFPGET:
1033 		case DIOCGETSRCNODES:
1034 		case DIOCCLRSRCNODES:
1035 		case DIOCIGETIFACES:
1036 		case DIOCGIFSPEED:
1037 		case DIOCSETIFFLAG:
1038 		case DIOCCLRIFFLAG:
1039 			break;
1040 		case DIOCRCLRTABLES:
1041 		case DIOCRADDTABLES:
1042 		case DIOCRDELTABLES:
1043 		case DIOCRSETTFLAGS:
1044 			if (((struct pfioc_table *)addr)->pfrio_flags &
1045 			    PFR_FLAG_DUMMY)
1046 				break; /* dummy operation ok */
1047 			return (EPERM);
1048 		default:
1049 			return (EPERM);
1050 		}
1051 
1052 	if (!(flags & FWRITE))
1053 		switch (cmd) {
1054 		case DIOCGETRULES:
1055 		case DIOCGETADDRS:
1056 		case DIOCGETADDR:
1057 		case DIOCGETSTATE:
1058 		case DIOCGETSTATUS:
1059 		case DIOCGETSTATES:
1060 		case DIOCGETTIMEOUT:
1061 		case DIOCGETLIMIT:
1062 		case DIOCGETALTQS:
1063 		case DIOCGETALTQ:
1064 		case DIOCGETQSTATS:
1065 		case DIOCGETRULESETS:
1066 		case DIOCGETRULESET:
1067 		case DIOCNATLOOK:
1068 		case DIOCRGETTABLES:
1069 		case DIOCRGETTSTATS:
1070 		case DIOCRGETADDRS:
1071 		case DIOCRGETASTATS:
1072 		case DIOCRTSTADDRS:
1073 		case DIOCOSFPGET:
1074 		case DIOCGETSRCNODES:
1075 		case DIOCIGETIFACES:
1076 		case DIOCGIFSPEED:
1077 			break;
1078 		case DIOCRCLRTABLES:
1079 		case DIOCRADDTABLES:
1080 		case DIOCRDELTABLES:
1081 		case DIOCRCLRTSTATS:
1082 		case DIOCRCLRADDRS:
1083 		case DIOCRADDADDRS:
1084 		case DIOCRDELADDRS:
1085 		case DIOCRSETADDRS:
1086 		case DIOCRSETTFLAGS:
1087 			if (((struct pfioc_table *)addr)->pfrio_flags &
1088 			    PFR_FLAG_DUMMY) {
1089 				flags |= FWRITE; /* need write lock for dummy */
1090 				break; /* dummy operation ok */
1091 			}
1092 			return (EACCES);
1093 		case DIOCGETRULE:
1094 			if (((struct pfioc_rule *)addr)->action ==
1095 			    PF_GET_CLR_CNTR)
1096 				return (EACCES);
1097 			break;
1098 		default:
1099 			return (EACCES);
1100 		}
1101 
1102 	CURVNET_SET(TD_TO_VNET(td));
1103 
1104 	switch (cmd) {
1105 	case DIOCSTART:
1106 		sx_xlock(&pf_ioctl_lock);
1107 		if (V_pf_status.running)
1108 			error = EEXIST;
1109 		else {
1110 			int cpu;
1111 
1112 			error = hook_pf();
1113 			if (error) {
1114 				DPFPRINTF(PF_DEBUG_MISC,
1115 				    ("pf: pfil registration failed\n"));
1116 				break;
1117 			}
1118 			V_pf_status.running = 1;
1119 			V_pf_status.since = time_second;
1120 
1121 			CPU_FOREACH(cpu)
1122 				V_pf_stateid[cpu] = time_second;
1123 
1124 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1125 		}
1126 		break;
1127 
1128 	case DIOCSTOP:
1129 		sx_xlock(&pf_ioctl_lock);
1130 		if (!V_pf_status.running)
1131 			error = ENOENT;
1132 		else {
1133 			V_pf_status.running = 0;
1134 			error = dehook_pf();
1135 			if (error) {
1136 				V_pf_status.running = 1;
1137 				DPFPRINTF(PF_DEBUG_MISC,
1138 				    ("pf: pfil unregistration failed\n"));
1139 			}
1140 			V_pf_status.since = time_second;
1141 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1142 		}
1143 		break;
1144 
1145 	case DIOCADDRULE: {
1146 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1147 		struct pf_ruleset	*ruleset;
1148 		struct pf_rule		*rule, *tail;
1149 		struct pf_pooladdr	*pa;
1150 		struct pfi_kif		*kif = NULL;
1151 		int			 rs_num;
1152 
1153 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1154 			error = EINVAL;
1155 			break;
1156 		}
1157 #ifndef INET
1158 		if (pr->rule.af == AF_INET) {
1159 			error = EAFNOSUPPORT;
1160 			break;
1161 		}
1162 #endif /* INET */
1163 #ifndef INET6
1164 		if (pr->rule.af == AF_INET6) {
1165 			error = EAFNOSUPPORT;
1166 			break;
1167 		}
1168 #endif /* INET6 */
1169 
1170 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
1171 		bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1172 		if (rule->ifname[0])
1173 			kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1174 		rule->states_cur = counter_u64_alloc(M_WAITOK);
1175 		rule->states_tot = counter_u64_alloc(M_WAITOK);
1176 		rule->src_nodes = counter_u64_alloc(M_WAITOK);
1177 		rule->cuid = td->td_ucred->cr_ruid;
1178 		rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1179 		TAILQ_INIT(&rule->rpool.list);
1180 
1181 #define	ERROUT(x)	{ error = (x); goto DIOCADDRULE_error; }
1182 
1183 		PF_RULES_WLOCK();
1184 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1185 		ruleset = pf_find_ruleset(pr->anchor);
1186 		if (ruleset == NULL)
1187 			ERROUT(EINVAL);
1188 		rs_num = pf_get_ruleset_number(pr->rule.action);
1189 		if (rs_num >= PF_RULESET_MAX)
1190 			ERROUT(EINVAL);
1191 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1192 			DPFPRINTF(PF_DEBUG_MISC,
1193 			    ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
1194 			    ruleset->rules[rs_num].inactive.ticket));
1195 			ERROUT(EBUSY);
1196 		}
1197 		if (pr->pool_ticket != V_ticket_pabuf) {
1198 			DPFPRINTF(PF_DEBUG_MISC,
1199 			    ("pool_ticket: %d != %d\n", pr->pool_ticket,
1200 			    V_ticket_pabuf));
1201 			ERROUT(EBUSY);
1202 		}
1203 
1204 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1205 		    pf_rulequeue);
1206 		if (tail)
1207 			rule->nr = tail->nr + 1;
1208 		else
1209 			rule->nr = 0;
1210 		if (rule->ifname[0]) {
1211 			rule->kif = pfi_kif_attach(kif, rule->ifname);
1212 			pfi_kif_ref(rule->kif);
1213 		} else
1214 			rule->kif = NULL;
1215 
1216 		if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
1217 			error = EBUSY;
1218 
1219 #ifdef ALTQ
1220 		/* set queue IDs */
1221 		if (rule->qname[0] != 0) {
1222 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1223 				error = EBUSY;
1224 			else if (rule->pqname[0] != 0) {
1225 				if ((rule->pqid =
1226 				    pf_qname2qid(rule->pqname)) == 0)
1227 					error = EBUSY;
1228 			} else
1229 				rule->pqid = rule->qid;
1230 		}
1231 #endif
1232 		if (rule->tagname[0])
1233 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1234 				error = EBUSY;
1235 		if (rule->match_tagname[0])
1236 			if ((rule->match_tag =
1237 			    pf_tagname2tag(rule->match_tagname)) == 0)
1238 				error = EBUSY;
1239 		if (rule->rt && !rule->direction)
1240 			error = EINVAL;
1241 		if (!rule->log)
1242 			rule->logif = 0;
1243 		if (rule->logif >= PFLOGIFS_MAX)
1244 			error = EINVAL;
1245 		if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1246 			error = ENOMEM;
1247 		if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1248 			error = ENOMEM;
1249 		if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1250 			error = EINVAL;
1251 		if (rule->scrub_flags & PFSTATE_SETPRIO &&
1252 		    (rule->set_prio[0] > PF_PRIO_MAX ||
1253 		    rule->set_prio[1] > PF_PRIO_MAX))
1254 			error = EINVAL;
1255 		TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1256 			if (pa->addr.type == PF_ADDR_TABLE) {
1257 				pa->addr.p.tbl = pfr_attach_table(ruleset,
1258 				    pa->addr.v.tblname);
1259 				if (pa->addr.p.tbl == NULL)
1260 					error = ENOMEM;
1261 			}
1262 
1263 		rule->overload_tbl = NULL;
1264 		if (rule->overload_tblname[0]) {
1265 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1266 			    rule->overload_tblname)) == NULL)
1267 				error = EINVAL;
1268 			else
1269 				rule->overload_tbl->pfrkt_flags |=
1270 				    PFR_TFLAG_ACTIVE;
1271 		}
1272 
1273 		pf_mv_pool(&V_pf_pabuf, &rule->rpool.list);
1274 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1275 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1276 		    (rule->rt > PF_NOPFROUTE)) &&
1277 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1278 			error = EINVAL;
1279 
1280 		if (error) {
1281 			pf_free_rule(rule);
1282 			PF_RULES_WUNLOCK();
1283 			break;
1284 		}
1285 
1286 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1287 		rule->evaluations = rule->packets[0] = rule->packets[1] =
1288 		    rule->bytes[0] = rule->bytes[1] = 0;
1289 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1290 		    rule, entries);
1291 		ruleset->rules[rs_num].inactive.rcount++;
1292 		PF_RULES_WUNLOCK();
1293 		break;
1294 
1295 #undef ERROUT
1296 DIOCADDRULE_error:
1297 		PF_RULES_WUNLOCK();
1298 		counter_u64_free(rule->states_cur);
1299 		counter_u64_free(rule->states_tot);
1300 		counter_u64_free(rule->src_nodes);
1301 		free(rule, M_PFRULE);
1302 		if (kif)
1303 			free(kif, PFI_MTYPE);
1304 		break;
1305 	}
1306 
1307 	case DIOCGETRULES: {
1308 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1309 		struct pf_ruleset	*ruleset;
1310 		struct pf_rule		*tail;
1311 		int			 rs_num;
1312 
1313 		PF_RULES_WLOCK();
1314 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1315 		ruleset = pf_find_ruleset(pr->anchor);
1316 		if (ruleset == NULL) {
1317 			PF_RULES_WUNLOCK();
1318 			error = EINVAL;
1319 			break;
1320 		}
1321 		rs_num = pf_get_ruleset_number(pr->rule.action);
1322 		if (rs_num >= PF_RULESET_MAX) {
1323 			PF_RULES_WUNLOCK();
1324 			error = EINVAL;
1325 			break;
1326 		}
1327 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1328 		    pf_rulequeue);
1329 		if (tail)
1330 			pr->nr = tail->nr + 1;
1331 		else
1332 			pr->nr = 0;
1333 		pr->ticket = ruleset->rules[rs_num].active.ticket;
1334 		PF_RULES_WUNLOCK();
1335 		break;
1336 	}
1337 
1338 	case DIOCGETRULE: {
1339 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1340 		struct pf_ruleset	*ruleset;
1341 		struct pf_rule		*rule;
1342 		int			 rs_num, i;
1343 
1344 		PF_RULES_WLOCK();
1345 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1346 		ruleset = pf_find_ruleset(pr->anchor);
1347 		if (ruleset == NULL) {
1348 			PF_RULES_WUNLOCK();
1349 			error = EINVAL;
1350 			break;
1351 		}
1352 		rs_num = pf_get_ruleset_number(pr->rule.action);
1353 		if (rs_num >= PF_RULESET_MAX) {
1354 			PF_RULES_WUNLOCK();
1355 			error = EINVAL;
1356 			break;
1357 		}
1358 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1359 			PF_RULES_WUNLOCK();
1360 			error = EBUSY;
1361 			break;
1362 		}
1363 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1364 		while ((rule != NULL) && (rule->nr != pr->nr))
1365 			rule = TAILQ_NEXT(rule, entries);
1366 		if (rule == NULL) {
1367 			PF_RULES_WUNLOCK();
1368 			error = EBUSY;
1369 			break;
1370 		}
1371 		bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1372 		pr->rule.u_states_cur = counter_u64_fetch(rule->states_cur);
1373 		pr->rule.u_states_tot = counter_u64_fetch(rule->states_tot);
1374 		pr->rule.u_src_nodes = counter_u64_fetch(rule->src_nodes);
1375 		if (pf_anchor_copyout(ruleset, rule, pr)) {
1376 			PF_RULES_WUNLOCK();
1377 			error = EBUSY;
1378 			break;
1379 		}
1380 		pf_addr_copyout(&pr->rule.src.addr);
1381 		pf_addr_copyout(&pr->rule.dst.addr);
1382 		for (i = 0; i < PF_SKIP_COUNT; ++i)
1383 			if (rule->skip[i].ptr == NULL)
1384 				pr->rule.skip[i].nr = -1;
1385 			else
1386 				pr->rule.skip[i].nr =
1387 				    rule->skip[i].ptr->nr;
1388 
1389 		if (pr->action == PF_GET_CLR_CNTR) {
1390 			rule->evaluations = 0;
1391 			rule->packets[0] = rule->packets[1] = 0;
1392 			rule->bytes[0] = rule->bytes[1] = 0;
1393 			counter_u64_zero(rule->states_tot);
1394 		}
1395 		PF_RULES_WUNLOCK();
1396 		break;
1397 	}
1398 
1399 	case DIOCCHANGERULE: {
1400 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1401 		struct pf_ruleset	*ruleset;
1402 		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1403 		struct pfi_kif		*kif = NULL;
1404 		struct pf_pooladdr	*pa;
1405 		u_int32_t		 nr = 0;
1406 		int			 rs_num;
1407 
1408 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1409 		    pcr->action > PF_CHANGE_GET_TICKET) {
1410 			error = EINVAL;
1411 			break;
1412 		}
1413 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1414 			error = EINVAL;
1415 			break;
1416 		}
1417 
1418 		if (pcr->action != PF_CHANGE_REMOVE) {
1419 #ifndef INET
1420 			if (pcr->rule.af == AF_INET) {
1421 				error = EAFNOSUPPORT;
1422 				break;
1423 			}
1424 #endif /* INET */
1425 #ifndef INET6
1426 			if (pcr->rule.af == AF_INET6) {
1427 				error = EAFNOSUPPORT;
1428 				break;
1429 			}
1430 #endif /* INET6 */
1431 			newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
1432 			bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1433 			if (newrule->ifname[0])
1434 				kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1435 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
1436 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
1437 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
1438 			newrule->cuid = td->td_ucred->cr_ruid;
1439 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1440 			TAILQ_INIT(&newrule->rpool.list);
1441 		}
1442 
1443 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGERULE_error; }
1444 
1445 		PF_RULES_WLOCK();
1446 		if (!(pcr->action == PF_CHANGE_REMOVE ||
1447 		    pcr->action == PF_CHANGE_GET_TICKET) &&
1448 		    pcr->pool_ticket != V_ticket_pabuf)
1449 			ERROUT(EBUSY);
1450 
1451 		ruleset = pf_find_ruleset(pcr->anchor);
1452 		if (ruleset == NULL)
1453 			ERROUT(EINVAL);
1454 
1455 		rs_num = pf_get_ruleset_number(pcr->rule.action);
1456 		if (rs_num >= PF_RULESET_MAX)
1457 			ERROUT(EINVAL);
1458 
1459 		if (pcr->action == PF_CHANGE_GET_TICKET) {
1460 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1461 			ERROUT(0);
1462 		} else if (pcr->ticket !=
1463 			    ruleset->rules[rs_num].active.ticket)
1464 				ERROUT(EINVAL);
1465 
1466 		if (pcr->action != PF_CHANGE_REMOVE) {
1467 			if (newrule->ifname[0]) {
1468 				newrule->kif = pfi_kif_attach(kif,
1469 				    newrule->ifname);
1470 				pfi_kif_ref(newrule->kif);
1471 			} else
1472 				newrule->kif = NULL;
1473 
1474 			if (newrule->rtableid > 0 &&
1475 			    newrule->rtableid >= rt_numfibs)
1476 				error = EBUSY;
1477 
1478 #ifdef ALTQ
1479 			/* set queue IDs */
1480 			if (newrule->qname[0] != 0) {
1481 				if ((newrule->qid =
1482 				    pf_qname2qid(newrule->qname)) == 0)
1483 					error = EBUSY;
1484 				else if (newrule->pqname[0] != 0) {
1485 					if ((newrule->pqid =
1486 					    pf_qname2qid(newrule->pqname)) == 0)
1487 						error = EBUSY;
1488 				} else
1489 					newrule->pqid = newrule->qid;
1490 			}
1491 #endif /* ALTQ */
1492 			if (newrule->tagname[0])
1493 				if ((newrule->tag =
1494 				    pf_tagname2tag(newrule->tagname)) == 0)
1495 					error = EBUSY;
1496 			if (newrule->match_tagname[0])
1497 				if ((newrule->match_tag = pf_tagname2tag(
1498 				    newrule->match_tagname)) == 0)
1499 					error = EBUSY;
1500 			if (newrule->rt && !newrule->direction)
1501 				error = EINVAL;
1502 			if (!newrule->log)
1503 				newrule->logif = 0;
1504 			if (newrule->logif >= PFLOGIFS_MAX)
1505 				error = EINVAL;
1506 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
1507 				error = ENOMEM;
1508 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
1509 				error = ENOMEM;
1510 			if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1511 				error = EINVAL;
1512 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1513 				if (pa->addr.type == PF_ADDR_TABLE) {
1514 					pa->addr.p.tbl =
1515 					    pfr_attach_table(ruleset,
1516 					    pa->addr.v.tblname);
1517 					if (pa->addr.p.tbl == NULL)
1518 						error = ENOMEM;
1519 				}
1520 
1521 			newrule->overload_tbl = NULL;
1522 			if (newrule->overload_tblname[0]) {
1523 				if ((newrule->overload_tbl = pfr_attach_table(
1524 				    ruleset, newrule->overload_tblname)) ==
1525 				    NULL)
1526 					error = EINVAL;
1527 				else
1528 					newrule->overload_tbl->pfrkt_flags |=
1529 					    PFR_TFLAG_ACTIVE;
1530 			}
1531 
1532 			pf_mv_pool(&V_pf_pabuf, &newrule->rpool.list);
1533 			if (((((newrule->action == PF_NAT) ||
1534 			    (newrule->action == PF_RDR) ||
1535 			    (newrule->action == PF_BINAT) ||
1536 			    (newrule->rt > PF_NOPFROUTE)) &&
1537 			    !newrule->anchor)) &&
1538 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1539 				error = EINVAL;
1540 
1541 			if (error) {
1542 				pf_free_rule(newrule);
1543 				PF_RULES_WUNLOCK();
1544 				break;
1545 			}
1546 
1547 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1548 			newrule->evaluations = 0;
1549 			newrule->packets[0] = newrule->packets[1] = 0;
1550 			newrule->bytes[0] = newrule->bytes[1] = 0;
1551 		}
1552 		pf_empty_pool(&V_pf_pabuf);
1553 
1554 		if (pcr->action == PF_CHANGE_ADD_HEAD)
1555 			oldrule = TAILQ_FIRST(
1556 			    ruleset->rules[rs_num].active.ptr);
1557 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1558 			oldrule = TAILQ_LAST(
1559 			    ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1560 		else {
1561 			oldrule = TAILQ_FIRST(
1562 			    ruleset->rules[rs_num].active.ptr);
1563 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1564 				oldrule = TAILQ_NEXT(oldrule, entries);
1565 			if (oldrule == NULL) {
1566 				if (newrule != NULL)
1567 					pf_free_rule(newrule);
1568 				PF_RULES_WUNLOCK();
1569 				error = EINVAL;
1570 				break;
1571 			}
1572 		}
1573 
1574 		if (pcr->action == PF_CHANGE_REMOVE) {
1575 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
1576 			    oldrule);
1577 			ruleset->rules[rs_num].active.rcount--;
1578 		} else {
1579 			if (oldrule == NULL)
1580 				TAILQ_INSERT_TAIL(
1581 				    ruleset->rules[rs_num].active.ptr,
1582 				    newrule, entries);
1583 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1584 			    pcr->action == PF_CHANGE_ADD_BEFORE)
1585 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1586 			else
1587 				TAILQ_INSERT_AFTER(
1588 				    ruleset->rules[rs_num].active.ptr,
1589 				    oldrule, newrule, entries);
1590 			ruleset->rules[rs_num].active.rcount++;
1591 		}
1592 
1593 		nr = 0;
1594 		TAILQ_FOREACH(oldrule,
1595 		    ruleset->rules[rs_num].active.ptr, entries)
1596 			oldrule->nr = nr++;
1597 
1598 		ruleset->rules[rs_num].active.ticket++;
1599 
1600 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1601 		pf_remove_if_empty_ruleset(ruleset);
1602 
1603 		PF_RULES_WUNLOCK();
1604 		break;
1605 
1606 #undef ERROUT
1607 DIOCCHANGERULE_error:
1608 		PF_RULES_WUNLOCK();
1609 		if (newrule != NULL) {
1610 			counter_u64_free(newrule->states_cur);
1611 			counter_u64_free(newrule->states_tot);
1612 			counter_u64_free(newrule->src_nodes);
1613 			free(newrule, M_PFRULE);
1614 		}
1615 		if (kif != NULL)
1616 			free(kif, PFI_MTYPE);
1617 		break;
1618 	}
1619 
1620 	case DIOCCLRSTATES: {
1621 		struct pf_state		*s;
1622 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1623 		u_int			 i, killed = 0;
1624 
1625 		for (i = 0; i <= pf_hashmask; i++) {
1626 			struct pf_idhash *ih = &V_pf_idhash[i];
1627 
1628 relock_DIOCCLRSTATES:
1629 			PF_HASHROW_LOCK(ih);
1630 			LIST_FOREACH(s, &ih->states, entry)
1631 				if (!psk->psk_ifname[0] ||
1632 				    !strcmp(psk->psk_ifname,
1633 				    s->kif->pfik_name)) {
1634 					/*
1635 					 * Don't send out individual
1636 					 * delete messages.
1637 					 */
1638 					s->state_flags |= PFSTATE_NOSYNC;
1639 					pf_unlink_state(s, PF_ENTER_LOCKED);
1640 					killed++;
1641 					goto relock_DIOCCLRSTATES;
1642 				}
1643 			PF_HASHROW_UNLOCK(ih);
1644 		}
1645 		psk->psk_killed = killed;
1646 		if (pfsync_clear_states_ptr != NULL)
1647 			pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
1648 		break;
1649 	}
1650 
1651 	case DIOCKILLSTATES: {
1652 		struct pf_state		*s;
1653 		struct pf_state_key	*sk;
1654 		struct pf_addr		*srcaddr, *dstaddr;
1655 		u_int16_t		 srcport, dstport;
1656 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
1657 		u_int			 i, killed = 0;
1658 
1659 		if (psk->psk_pfcmp.id) {
1660 			if (psk->psk_pfcmp.creatorid == 0)
1661 				psk->psk_pfcmp.creatorid = V_pf_status.hostid;
1662 			if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
1663 			    psk->psk_pfcmp.creatorid))) {
1664 				pf_unlink_state(s, PF_ENTER_LOCKED);
1665 				psk->psk_killed = 1;
1666 			}
1667 			break;
1668 		}
1669 
1670 		for (i = 0; i <= pf_hashmask; i++) {
1671 			struct pf_idhash *ih = &V_pf_idhash[i];
1672 
1673 relock_DIOCKILLSTATES:
1674 			PF_HASHROW_LOCK(ih);
1675 			LIST_FOREACH(s, &ih->states, entry) {
1676 				sk = s->key[PF_SK_WIRE];
1677 				if (s->direction == PF_OUT) {
1678 					srcaddr = &sk->addr[1];
1679 					dstaddr = &sk->addr[0];
1680 					srcport = sk->port[1];
1681 					dstport = sk->port[0];
1682 				} else {
1683 					srcaddr = &sk->addr[0];
1684 					dstaddr = &sk->addr[1];
1685 					srcport = sk->port[0];
1686 					dstport = sk->port[1];
1687 				}
1688 
1689 				if ((!psk->psk_af || sk->af == psk->psk_af)
1690 				    && (!psk->psk_proto || psk->psk_proto ==
1691 				    sk->proto) &&
1692 				    PF_MATCHA(psk->psk_src.neg,
1693 				    &psk->psk_src.addr.v.a.addr,
1694 				    &psk->psk_src.addr.v.a.mask,
1695 				    srcaddr, sk->af) &&
1696 				    PF_MATCHA(psk->psk_dst.neg,
1697 				    &psk->psk_dst.addr.v.a.addr,
1698 				    &psk->psk_dst.addr.v.a.mask,
1699 				    dstaddr, sk->af) &&
1700 				    (psk->psk_src.port_op == 0 ||
1701 				    pf_match_port(psk->psk_src.port_op,
1702 				    psk->psk_src.port[0], psk->psk_src.port[1],
1703 				    srcport)) &&
1704 				    (psk->psk_dst.port_op == 0 ||
1705 				    pf_match_port(psk->psk_dst.port_op,
1706 				    psk->psk_dst.port[0], psk->psk_dst.port[1],
1707 				    dstport)) &&
1708 				    (!psk->psk_label[0] ||
1709 				    (s->rule.ptr->label[0] &&
1710 				    !strcmp(psk->psk_label,
1711 				    s->rule.ptr->label))) &&
1712 				    (!psk->psk_ifname[0] ||
1713 				    !strcmp(psk->psk_ifname,
1714 				    s->kif->pfik_name))) {
1715 					pf_unlink_state(s, PF_ENTER_LOCKED);
1716 					killed++;
1717 					goto relock_DIOCKILLSTATES;
1718 				}
1719 			}
1720 			PF_HASHROW_UNLOCK(ih);
1721 		}
1722 		psk->psk_killed = killed;
1723 		break;
1724 	}
1725 
1726 	case DIOCADDSTATE: {
1727 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1728 		struct pfsync_state	*sp = &ps->state;
1729 
1730 		if (sp->timeout >= PFTM_MAX) {
1731 			error = EINVAL;
1732 			break;
1733 		}
1734 		if (pfsync_state_import_ptr != NULL) {
1735 			PF_RULES_RLOCK();
1736 			error = pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
1737 			PF_RULES_RUNLOCK();
1738 		} else
1739 			error = EOPNOTSUPP;
1740 		break;
1741 	}
1742 
1743 	case DIOCGETSTATE: {
1744 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1745 		struct pf_state		*s;
1746 
1747 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
1748 		if (s == NULL) {
1749 			error = ENOENT;
1750 			break;
1751 		}
1752 
1753 		pfsync_state_export(&ps->state, s);
1754 		PF_STATE_UNLOCK(s);
1755 		break;
1756 	}
1757 
1758 	case DIOCGETSTATES: {
1759 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
1760 		struct pf_state		*s;
1761 		struct pfsync_state	*pstore, *p;
1762 		int i, nr;
1763 
1764 		if (ps->ps_len == 0) {
1765 			nr = uma_zone_get_cur(V_pf_state_z);
1766 			ps->ps_len = sizeof(struct pfsync_state) * nr;
1767 			break;
1768 		}
1769 
1770 		p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK);
1771 		nr = 0;
1772 
1773 		for (i = 0; i <= pf_hashmask; i++) {
1774 			struct pf_idhash *ih = &V_pf_idhash[i];
1775 
1776 			PF_HASHROW_LOCK(ih);
1777 			LIST_FOREACH(s, &ih->states, entry) {
1778 
1779 				if (s->timeout == PFTM_UNLINKED)
1780 					continue;
1781 
1782 				if ((nr+1) * sizeof(*p) > ps->ps_len) {
1783 					PF_HASHROW_UNLOCK(ih);
1784 					goto DIOCGETSTATES_full;
1785 				}
1786 				pfsync_state_export(p, s);
1787 				p++;
1788 				nr++;
1789 			}
1790 			PF_HASHROW_UNLOCK(ih);
1791 		}
1792 DIOCGETSTATES_full:
1793 		error = copyout(pstore, ps->ps_states,
1794 		    sizeof(struct pfsync_state) * nr);
1795 		if (error) {
1796 			free(pstore, M_TEMP);
1797 			break;
1798 		}
1799 		ps->ps_len = sizeof(struct pfsync_state) * nr;
1800 		free(pstore, M_TEMP);
1801 
1802 		break;
1803 	}
1804 
1805 	case DIOCGETSTATUS: {
1806 		struct pf_status *s = (struct pf_status *)addr;
1807 
1808 		PF_RULES_RLOCK();
1809 		s->running = V_pf_status.running;
1810 		s->since   = V_pf_status.since;
1811 		s->debug   = V_pf_status.debug;
1812 		s->hostid  = V_pf_status.hostid;
1813 		s->states  = V_pf_status.states;
1814 		s->src_nodes = V_pf_status.src_nodes;
1815 
1816 		for (int i = 0; i < PFRES_MAX; i++)
1817 			s->counters[i] =
1818 			    counter_u64_fetch(V_pf_status.counters[i]);
1819 		for (int i = 0; i < LCNT_MAX; i++)
1820 			s->lcounters[i] =
1821 			    counter_u64_fetch(V_pf_status.lcounters[i]);
1822 		for (int i = 0; i < FCNT_MAX; i++)
1823 			s->fcounters[i] =
1824 			    counter_u64_fetch(V_pf_status.fcounters[i]);
1825 		for (int i = 0; i < SCNT_MAX; i++)
1826 			s->scounters[i] =
1827 			    counter_u64_fetch(V_pf_status.scounters[i]);
1828 
1829 		bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ);
1830 		bcopy(V_pf_status.pf_chksum, s->pf_chksum,
1831 		    PF_MD5_DIGEST_LENGTH);
1832 
1833 		pfi_update_status(s->ifname, s);
1834 		PF_RULES_RUNLOCK();
1835 		break;
1836 	}
1837 
1838 	case DIOCSETSTATUSIF: {
1839 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
1840 
1841 		if (pi->ifname[0] == 0) {
1842 			bzero(V_pf_status.ifname, IFNAMSIZ);
1843 			break;
1844 		}
1845 		PF_RULES_WLOCK();
1846 		strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
1847 		PF_RULES_WUNLOCK();
1848 		break;
1849 	}
1850 
1851 	case DIOCCLRSTATUS: {
1852 		PF_RULES_WLOCK();
1853 		for (int i = 0; i < PFRES_MAX; i++)
1854 			counter_u64_zero(V_pf_status.counters[i]);
1855 		for (int i = 0; i < FCNT_MAX; i++)
1856 			counter_u64_zero(V_pf_status.fcounters[i]);
1857 		for (int i = 0; i < SCNT_MAX; i++)
1858 			counter_u64_zero(V_pf_status.scounters[i]);
1859 		for (int i = 0; i < LCNT_MAX; i++)
1860 			counter_u64_zero(V_pf_status.lcounters[i]);
1861 		V_pf_status.since = time_second;
1862 		if (*V_pf_status.ifname)
1863 			pfi_update_status(V_pf_status.ifname, NULL);
1864 		PF_RULES_WUNLOCK();
1865 		break;
1866 	}
1867 
1868 	case DIOCNATLOOK: {
1869 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
1870 		struct pf_state_key	*sk;
1871 		struct pf_state		*state;
1872 		struct pf_state_key_cmp	 key;
1873 		int			 m = 0, direction = pnl->direction;
1874 		int			 sidx, didx;
1875 
1876 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
1877 		sidx = (direction == PF_IN) ? 1 : 0;
1878 		didx = (direction == PF_IN) ? 0 : 1;
1879 
1880 		if (!pnl->proto ||
1881 		    PF_AZERO(&pnl->saddr, pnl->af) ||
1882 		    PF_AZERO(&pnl->daddr, pnl->af) ||
1883 		    ((pnl->proto == IPPROTO_TCP ||
1884 		    pnl->proto == IPPROTO_UDP) &&
1885 		    (!pnl->dport || !pnl->sport)))
1886 			error = EINVAL;
1887 		else {
1888 			bzero(&key, sizeof(key));
1889 			key.af = pnl->af;
1890 			key.proto = pnl->proto;
1891 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
1892 			key.port[sidx] = pnl->sport;
1893 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
1894 			key.port[didx] = pnl->dport;
1895 
1896 			state = pf_find_state_all(&key, direction, &m);
1897 
1898 			if (m > 1)
1899 				error = E2BIG;	/* more than one state */
1900 			else if (state != NULL) {
1901 				/* XXXGL: not locked read */
1902 				sk = state->key[sidx];
1903 				PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
1904 				pnl->rsport = sk->port[sidx];
1905 				PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
1906 				pnl->rdport = sk->port[didx];
1907 			} else
1908 				error = ENOENT;
1909 		}
1910 		break;
1911 	}
1912 
1913 	case DIOCSETTIMEOUT: {
1914 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1915 		int		 old;
1916 
1917 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
1918 		    pt->seconds < 0) {
1919 			error = EINVAL;
1920 			break;
1921 		}
1922 		PF_RULES_WLOCK();
1923 		old = V_pf_default_rule.timeout[pt->timeout];
1924 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
1925 			pt->seconds = 1;
1926 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
1927 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
1928 			wakeup(pf_purge_thread);
1929 		pt->seconds = old;
1930 		PF_RULES_WUNLOCK();
1931 		break;
1932 	}
1933 
1934 	case DIOCGETTIMEOUT: {
1935 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1936 
1937 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
1938 			error = EINVAL;
1939 			break;
1940 		}
1941 		PF_RULES_RLOCK();
1942 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
1943 		PF_RULES_RUNLOCK();
1944 		break;
1945 	}
1946 
1947 	case DIOCGETLIMIT: {
1948 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1949 
1950 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
1951 			error = EINVAL;
1952 			break;
1953 		}
1954 		PF_RULES_RLOCK();
1955 		pl->limit = V_pf_limits[pl->index].limit;
1956 		PF_RULES_RUNLOCK();
1957 		break;
1958 	}
1959 
1960 	case DIOCSETLIMIT: {
1961 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1962 		int			 old_limit;
1963 
1964 		PF_RULES_WLOCK();
1965 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
1966 		    V_pf_limits[pl->index].zone == NULL) {
1967 			PF_RULES_WUNLOCK();
1968 			error = EINVAL;
1969 			break;
1970 		}
1971 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
1972 		old_limit = V_pf_limits[pl->index].limit;
1973 		V_pf_limits[pl->index].limit = pl->limit;
1974 		pl->limit = old_limit;
1975 		PF_RULES_WUNLOCK();
1976 		break;
1977 	}
1978 
1979 	case DIOCSETDEBUG: {
1980 		u_int32_t	*level = (u_int32_t *)addr;
1981 
1982 		PF_RULES_WLOCK();
1983 		V_pf_status.debug = *level;
1984 		PF_RULES_WUNLOCK();
1985 		break;
1986 	}
1987 
1988 	case DIOCCLRRULECTRS: {
1989 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
1990 		struct pf_ruleset	*ruleset = &pf_main_ruleset;
1991 		struct pf_rule		*rule;
1992 
1993 		PF_RULES_WLOCK();
1994 		TAILQ_FOREACH(rule,
1995 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
1996 			rule->evaluations = 0;
1997 			rule->packets[0] = rule->packets[1] = 0;
1998 			rule->bytes[0] = rule->bytes[1] = 0;
1999 		}
2000 		PF_RULES_WUNLOCK();
2001 		break;
2002 	}
2003 
2004 	case DIOCGIFSPEED: {
2005 		struct pf_ifspeed	*psp = (struct pf_ifspeed *)addr;
2006 		struct pf_ifspeed	ps;
2007 		struct ifnet		*ifp;
2008 
2009 		if (psp->ifname[0] != 0) {
2010 			/* Can we completely trust user-land? */
2011 			strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
2012 			ifp = ifunit(ps.ifname);
2013 			if (ifp != NULL)
2014 				psp->baudrate = ifp->if_baudrate;
2015 			else
2016 				error = EINVAL;
2017 		} else
2018 			error = EINVAL;
2019 		break;
2020 	}
2021 
2022 #ifdef ALTQ
2023 	case DIOCSTARTALTQ: {
2024 		struct pf_altq		*altq;
2025 
2026 		PF_RULES_WLOCK();
2027 		/* enable all altq interfaces on active list */
2028 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
2029 			if (altq->qname[0] == 0 && (altq->local_flags &
2030 			    PFALTQ_FLAG_IF_REMOVED) == 0) {
2031 				error = pf_enable_altq(altq);
2032 				if (error != 0)
2033 					break;
2034 			}
2035 		}
2036 		if (error == 0)
2037 			V_pf_altq_running = 1;
2038 		PF_RULES_WUNLOCK();
2039 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
2040 		break;
2041 	}
2042 
2043 	case DIOCSTOPALTQ: {
2044 		struct pf_altq		*altq;
2045 
2046 		PF_RULES_WLOCK();
2047 		/* disable all altq interfaces on active list */
2048 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
2049 			if (altq->qname[0] == 0 && (altq->local_flags &
2050 			    PFALTQ_FLAG_IF_REMOVED) == 0) {
2051 				error = pf_disable_altq(altq);
2052 				if (error != 0)
2053 					break;
2054 			}
2055 		}
2056 		if (error == 0)
2057 			V_pf_altq_running = 0;
2058 		PF_RULES_WUNLOCK();
2059 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2060 		break;
2061 	}
2062 
2063 	case DIOCADDALTQ: {
2064 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2065 		struct pf_altq		*altq, *a;
2066 		struct ifnet		*ifp;
2067 
2068 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK);
2069 		bcopy(&pa->altq, altq, sizeof(struct pf_altq));
2070 		altq->local_flags = 0;
2071 
2072 		PF_RULES_WLOCK();
2073 		if (pa->ticket != V_ticket_altqs_inactive) {
2074 			PF_RULES_WUNLOCK();
2075 			free(altq, M_PFALTQ);
2076 			error = EBUSY;
2077 			break;
2078 		}
2079 
2080 		/*
2081 		 * if this is for a queue, find the discipline and
2082 		 * copy the necessary fields
2083 		 */
2084 		if (altq->qname[0] != 0) {
2085 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2086 				PF_RULES_WUNLOCK();
2087 				error = EBUSY;
2088 				free(altq, M_PFALTQ);
2089 				break;
2090 			}
2091 			altq->altq_disc = NULL;
2092 			TAILQ_FOREACH(a, V_pf_altqs_inactive, entries) {
2093 				if (strncmp(a->ifname, altq->ifname,
2094 				    IFNAMSIZ) == 0 && a->qname[0] == 0) {
2095 					altq->altq_disc = a->altq_disc;
2096 					break;
2097 				}
2098 			}
2099 		}
2100 
2101 		if ((ifp = ifunit(altq->ifname)) == NULL)
2102 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
2103 		else
2104 			error = altq_add(altq);
2105 
2106 		if (error) {
2107 			PF_RULES_WUNLOCK();
2108 			free(altq, M_PFALTQ);
2109 			break;
2110 		}
2111 
2112 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
2113 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2114 		PF_RULES_WUNLOCK();
2115 		break;
2116 	}
2117 
2118 	case DIOCGETALTQS: {
2119 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2120 		struct pf_altq		*altq;
2121 
2122 		PF_RULES_RLOCK();
2123 		pa->nr = 0;
2124 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
2125 			pa->nr++;
2126 		pa->ticket = V_ticket_altqs_active;
2127 		PF_RULES_RUNLOCK();
2128 		break;
2129 	}
2130 
2131 	case DIOCGETALTQ: {
2132 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2133 		struct pf_altq		*altq;
2134 		u_int32_t		 nr;
2135 
2136 		PF_RULES_RLOCK();
2137 		if (pa->ticket != V_ticket_altqs_active) {
2138 			PF_RULES_RUNLOCK();
2139 			error = EBUSY;
2140 			break;
2141 		}
2142 		nr = 0;
2143 		altq = TAILQ_FIRST(V_pf_altqs_active);
2144 		while ((altq != NULL) && (nr < pa->nr)) {
2145 			altq = TAILQ_NEXT(altq, entries);
2146 			nr++;
2147 		}
2148 		if (altq == NULL) {
2149 			PF_RULES_RUNLOCK();
2150 			error = EBUSY;
2151 			break;
2152 		}
2153 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2154 		PF_RULES_RUNLOCK();
2155 		break;
2156 	}
2157 
2158 	case DIOCCHANGEALTQ:
2159 		/* CHANGEALTQ not supported yet! */
2160 		error = ENODEV;
2161 		break;
2162 
2163 	case DIOCGETQSTATS: {
2164 		struct pfioc_qstats	*pq = (struct pfioc_qstats *)addr;
2165 		struct pf_altq		*altq;
2166 		u_int32_t		 nr;
2167 		int			 nbytes;
2168 
2169 		PF_RULES_RLOCK();
2170 		if (pq->ticket != V_ticket_altqs_active) {
2171 			PF_RULES_RUNLOCK();
2172 			error = EBUSY;
2173 			break;
2174 		}
2175 		nbytes = pq->nbytes;
2176 		nr = 0;
2177 		altq = TAILQ_FIRST(V_pf_altqs_active);
2178 		while ((altq != NULL) && (nr < pq->nr)) {
2179 			altq = TAILQ_NEXT(altq, entries);
2180 			nr++;
2181 		}
2182 		if (altq == NULL) {
2183 			PF_RULES_RUNLOCK();
2184 			error = EBUSY;
2185 			break;
2186 		}
2187 
2188 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
2189 			PF_RULES_RUNLOCK();
2190 			error = ENXIO;
2191 			break;
2192 		}
2193 		PF_RULES_RUNLOCK();
2194 		error = altq_getqstats(altq, pq->buf, &nbytes);
2195 		if (error == 0) {
2196 			pq->scheduler = altq->scheduler;
2197 			pq->nbytes = nbytes;
2198 		}
2199 		break;
2200 	}
2201 #endif /* ALTQ */
2202 
2203 	case DIOCBEGINADDRS: {
2204 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2205 
2206 		PF_RULES_WLOCK();
2207 		pf_empty_pool(&V_pf_pabuf);
2208 		pp->ticket = ++V_ticket_pabuf;
2209 		PF_RULES_WUNLOCK();
2210 		break;
2211 	}
2212 
2213 	case DIOCADDADDR: {
2214 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2215 		struct pf_pooladdr	*pa;
2216 		struct pfi_kif		*kif = NULL;
2217 
2218 #ifndef INET
2219 		if (pp->af == AF_INET) {
2220 			error = EAFNOSUPPORT;
2221 			break;
2222 		}
2223 #endif /* INET */
2224 #ifndef INET6
2225 		if (pp->af == AF_INET6) {
2226 			error = EAFNOSUPPORT;
2227 			break;
2228 		}
2229 #endif /* INET6 */
2230 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2231 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2232 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2233 			error = EINVAL;
2234 			break;
2235 		}
2236 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2237 		bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2238 		if (pa->ifname[0])
2239 			kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2240 		PF_RULES_WLOCK();
2241 		if (pp->ticket != V_ticket_pabuf) {
2242 			PF_RULES_WUNLOCK();
2243 			if (pa->ifname[0])
2244 				free(kif, PFI_MTYPE);
2245 			free(pa, M_PFRULE);
2246 			error = EBUSY;
2247 			break;
2248 		}
2249 		if (pa->ifname[0]) {
2250 			pa->kif = pfi_kif_attach(kif, pa->ifname);
2251 			pfi_kif_ref(pa->kif);
2252 		} else
2253 			pa->kif = NULL;
2254 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2255 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2256 			if (pa->ifname[0])
2257 				pfi_kif_unref(pa->kif);
2258 			PF_RULES_WUNLOCK();
2259 			free(pa, M_PFRULE);
2260 			break;
2261 		}
2262 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2263 		PF_RULES_WUNLOCK();
2264 		break;
2265 	}
2266 
2267 	case DIOCGETADDRS: {
2268 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2269 		struct pf_pool		*pool;
2270 		struct pf_pooladdr	*pa;
2271 
2272 		PF_RULES_RLOCK();
2273 		pp->nr = 0;
2274 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2275 		    pp->r_num, 0, 1, 0);
2276 		if (pool == NULL) {
2277 			PF_RULES_RUNLOCK();
2278 			error = EBUSY;
2279 			break;
2280 		}
2281 		TAILQ_FOREACH(pa, &pool->list, entries)
2282 			pp->nr++;
2283 		PF_RULES_RUNLOCK();
2284 		break;
2285 	}
2286 
2287 	case DIOCGETADDR: {
2288 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2289 		struct pf_pool		*pool;
2290 		struct pf_pooladdr	*pa;
2291 		u_int32_t		 nr = 0;
2292 
2293 		PF_RULES_RLOCK();
2294 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2295 		    pp->r_num, 0, 1, 1);
2296 		if (pool == NULL) {
2297 			PF_RULES_RUNLOCK();
2298 			error = EBUSY;
2299 			break;
2300 		}
2301 		pa = TAILQ_FIRST(&pool->list);
2302 		while ((pa != NULL) && (nr < pp->nr)) {
2303 			pa = TAILQ_NEXT(pa, entries);
2304 			nr++;
2305 		}
2306 		if (pa == NULL) {
2307 			PF_RULES_RUNLOCK();
2308 			error = EBUSY;
2309 			break;
2310 		}
2311 		bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2312 		pf_addr_copyout(&pp->addr.addr);
2313 		PF_RULES_RUNLOCK();
2314 		break;
2315 	}
2316 
2317 	case DIOCCHANGEADDR: {
2318 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
2319 		struct pf_pool		*pool;
2320 		struct pf_pooladdr	*oldpa = NULL, *newpa = NULL;
2321 		struct pf_ruleset	*ruleset;
2322 		struct pfi_kif		*kif = NULL;
2323 
2324 		if (pca->action < PF_CHANGE_ADD_HEAD ||
2325 		    pca->action > PF_CHANGE_REMOVE) {
2326 			error = EINVAL;
2327 			break;
2328 		}
2329 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2330 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2331 		    pca->addr.addr.type != PF_ADDR_TABLE) {
2332 			error = EINVAL;
2333 			break;
2334 		}
2335 
2336 		if (pca->action != PF_CHANGE_REMOVE) {
2337 #ifndef INET
2338 			if (pca->af == AF_INET) {
2339 				error = EAFNOSUPPORT;
2340 				break;
2341 			}
2342 #endif /* INET */
2343 #ifndef INET6
2344 			if (pca->af == AF_INET6) {
2345 				error = EAFNOSUPPORT;
2346 				break;
2347 			}
2348 #endif /* INET6 */
2349 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
2350 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2351 			if (newpa->ifname[0])
2352 				kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2353 			newpa->kif = NULL;
2354 		}
2355 
2356 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGEADDR_error; }
2357 		PF_RULES_WLOCK();
2358 		ruleset = pf_find_ruleset(pca->anchor);
2359 		if (ruleset == NULL)
2360 			ERROUT(EBUSY);
2361 
2362 		pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2363 		    pca->r_num, pca->r_last, 1, 1);
2364 		if (pool == NULL)
2365 			ERROUT(EBUSY);
2366 
2367 		if (pca->action != PF_CHANGE_REMOVE) {
2368 			if (newpa->ifname[0]) {
2369 				newpa->kif = pfi_kif_attach(kif, newpa->ifname);
2370 				pfi_kif_ref(newpa->kif);
2371 				kif = NULL;
2372 			}
2373 
2374 			switch (newpa->addr.type) {
2375 			case PF_ADDR_DYNIFTL:
2376 				error = pfi_dynaddr_setup(&newpa->addr,
2377 				    pca->af);
2378 				break;
2379 			case PF_ADDR_TABLE:
2380 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
2381 				    newpa->addr.v.tblname);
2382 				if (newpa->addr.p.tbl == NULL)
2383 					error = ENOMEM;
2384 				break;
2385 			}
2386 			if (error)
2387 				goto DIOCCHANGEADDR_error;
2388 		}
2389 
2390 		switch (pca->action) {
2391 		case PF_CHANGE_ADD_HEAD:
2392 			oldpa = TAILQ_FIRST(&pool->list);
2393 			break;
2394 		case PF_CHANGE_ADD_TAIL:
2395 			oldpa = TAILQ_LAST(&pool->list, pf_palist);
2396 			break;
2397 		default:
2398 			oldpa = TAILQ_FIRST(&pool->list);
2399 			for (int i = 0; oldpa && i < pca->nr; i++)
2400 				oldpa = TAILQ_NEXT(oldpa, entries);
2401 
2402 			if (oldpa == NULL)
2403 				ERROUT(EINVAL);
2404 		}
2405 
2406 		if (pca->action == PF_CHANGE_REMOVE) {
2407 			TAILQ_REMOVE(&pool->list, oldpa, entries);
2408 			switch (oldpa->addr.type) {
2409 			case PF_ADDR_DYNIFTL:
2410 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
2411 				break;
2412 			case PF_ADDR_TABLE:
2413 				pfr_detach_table(oldpa->addr.p.tbl);
2414 				break;
2415 			}
2416 			if (oldpa->kif)
2417 				pfi_kif_unref(oldpa->kif);
2418 			free(oldpa, M_PFRULE);
2419 		} else {
2420 			if (oldpa == NULL)
2421 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2422 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
2423 			    pca->action == PF_CHANGE_ADD_BEFORE)
2424 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2425 			else
2426 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
2427 				    newpa, entries);
2428 		}
2429 
2430 		pool->cur = TAILQ_FIRST(&pool->list);
2431 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
2432 		PF_RULES_WUNLOCK();
2433 		break;
2434 
2435 #undef ERROUT
2436 DIOCCHANGEADDR_error:
2437 		if (newpa != NULL) {
2438 			if (newpa->kif)
2439 				pfi_kif_unref(newpa->kif);
2440 			free(newpa, M_PFRULE);
2441 		}
2442 		PF_RULES_WUNLOCK();
2443 		if (kif != NULL)
2444 			free(kif, PFI_MTYPE);
2445 		break;
2446 	}
2447 
2448 	case DIOCGETRULESETS: {
2449 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2450 		struct pf_ruleset	*ruleset;
2451 		struct pf_anchor	*anchor;
2452 
2453 		PF_RULES_RLOCK();
2454 		pr->path[sizeof(pr->path) - 1] = 0;
2455 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2456 			PF_RULES_RUNLOCK();
2457 			error = ENOENT;
2458 			break;
2459 		}
2460 		pr->nr = 0;
2461 		if (ruleset->anchor == NULL) {
2462 			/* XXX kludge for pf_main_ruleset */
2463 			RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2464 				if (anchor->parent == NULL)
2465 					pr->nr++;
2466 		} else {
2467 			RB_FOREACH(anchor, pf_anchor_node,
2468 			    &ruleset->anchor->children)
2469 				pr->nr++;
2470 		}
2471 		PF_RULES_RUNLOCK();
2472 		break;
2473 	}
2474 
2475 	case DIOCGETRULESET: {
2476 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2477 		struct pf_ruleset	*ruleset;
2478 		struct pf_anchor	*anchor;
2479 		u_int32_t		 nr = 0;
2480 
2481 		PF_RULES_RLOCK();
2482 		pr->path[sizeof(pr->path) - 1] = 0;
2483 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2484 			PF_RULES_RUNLOCK();
2485 			error = ENOENT;
2486 			break;
2487 		}
2488 		pr->name[0] = 0;
2489 		if (ruleset->anchor == NULL) {
2490 			/* XXX kludge for pf_main_ruleset */
2491 			RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2492 				if (anchor->parent == NULL && nr++ == pr->nr) {
2493 					strlcpy(pr->name, anchor->name,
2494 					    sizeof(pr->name));
2495 					break;
2496 				}
2497 		} else {
2498 			RB_FOREACH(anchor, pf_anchor_node,
2499 			    &ruleset->anchor->children)
2500 				if (nr++ == pr->nr) {
2501 					strlcpy(pr->name, anchor->name,
2502 					    sizeof(pr->name));
2503 					break;
2504 				}
2505 		}
2506 		if (!pr->name[0])
2507 			error = EBUSY;
2508 		PF_RULES_RUNLOCK();
2509 		break;
2510 	}
2511 
2512 	case DIOCRCLRTABLES: {
2513 		struct pfioc_table *io = (struct pfioc_table *)addr;
2514 
2515 		if (io->pfrio_esize != 0) {
2516 			error = ENODEV;
2517 			break;
2518 		}
2519 		PF_RULES_WLOCK();
2520 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2521 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2522 		PF_RULES_WUNLOCK();
2523 		break;
2524 	}
2525 
2526 	case DIOCRADDTABLES: {
2527 		struct pfioc_table *io = (struct pfioc_table *)addr;
2528 		struct pfr_table *pfrts;
2529 		size_t totlen;
2530 
2531 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2532 			error = ENODEV;
2533 			break;
2534 		}
2535 
2536 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
2537 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
2538 			error = ENOMEM;
2539 			break;
2540 		}
2541 
2542 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2543 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
2544 		    M_TEMP, M_WAITOK);
2545 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2546 		if (error) {
2547 			free(pfrts, M_TEMP);
2548 			break;
2549 		}
2550 		PF_RULES_WLOCK();
2551 		error = pfr_add_tables(pfrts, io->pfrio_size,
2552 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2553 		PF_RULES_WUNLOCK();
2554 		free(pfrts, M_TEMP);
2555 		break;
2556 	}
2557 
2558 	case DIOCRDELTABLES: {
2559 		struct pfioc_table *io = (struct pfioc_table *)addr;
2560 		struct pfr_table *pfrts;
2561 		size_t totlen;
2562 
2563 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2564 			error = ENODEV;
2565 			break;
2566 		}
2567 
2568 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
2569 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
2570 			error = ENOMEM;
2571 			break;
2572 		}
2573 
2574 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2575 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
2576 		    M_TEMP, M_WAITOK);
2577 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2578 		if (error) {
2579 			free(pfrts, M_TEMP);
2580 			break;
2581 		}
2582 		PF_RULES_WLOCK();
2583 		error = pfr_del_tables(pfrts, io->pfrio_size,
2584 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2585 		PF_RULES_WUNLOCK();
2586 		free(pfrts, M_TEMP);
2587 		break;
2588 	}
2589 
2590 	case DIOCRGETTABLES: {
2591 		struct pfioc_table *io = (struct pfioc_table *)addr;
2592 		struct pfr_table *pfrts;
2593 		size_t totlen, n;
2594 
2595 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2596 			error = ENODEV;
2597 			break;
2598 		}
2599 		PF_RULES_RLOCK();
2600 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
2601 		io->pfrio_size = min(io->pfrio_size, n);
2602 
2603 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2604 
2605 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
2606 		    M_TEMP, M_NOWAIT);
2607 		if (pfrts == NULL) {
2608 			error = ENOMEM;
2609 			PF_RULES_RUNLOCK();
2610 			break;
2611 		}
2612 		error = pfr_get_tables(&io->pfrio_table, pfrts,
2613 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2614 		PF_RULES_RUNLOCK();
2615 		if (error == 0)
2616 			error = copyout(pfrts, io->pfrio_buffer, totlen);
2617 		free(pfrts, M_TEMP);
2618 		break;
2619 	}
2620 
2621 	case DIOCRGETTSTATS: {
2622 		struct pfioc_table *io = (struct pfioc_table *)addr;
2623 		struct pfr_tstats *pfrtstats;
2624 		size_t totlen, n;
2625 
2626 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2627 			error = ENODEV;
2628 			break;
2629 		}
2630 		PF_RULES_WLOCK();
2631 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
2632 		io->pfrio_size = min(io->pfrio_size, n);
2633 
2634 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
2635 		pfrtstats = mallocarray(io->pfrio_size,
2636 		    sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT);
2637 		if (pfrtstats == NULL) {
2638 			error = ENOMEM;
2639 			PF_RULES_WUNLOCK();
2640 			break;
2641 		}
2642 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
2643 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2644 		PF_RULES_WUNLOCK();
2645 		if (error == 0)
2646 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
2647 		free(pfrtstats, M_TEMP);
2648 		break;
2649 	}
2650 
2651 	case DIOCRCLRTSTATS: {
2652 		struct pfioc_table *io = (struct pfioc_table *)addr;
2653 		struct pfr_table *pfrts;
2654 		size_t totlen, n;
2655 
2656 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2657 			error = ENODEV;
2658 			break;
2659 		}
2660 
2661 		PF_RULES_WLOCK();
2662 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
2663 		io->pfrio_size = min(io->pfrio_size, n);
2664 
2665 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2666 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
2667 		    M_TEMP, M_NOWAIT);
2668 		if (pfrts == NULL) {
2669 			error = ENOMEM;
2670 			PF_RULES_WUNLOCK();
2671 			break;
2672 		}
2673 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2674 		if (error) {
2675 			free(pfrts, M_TEMP);
2676 			PF_RULES_WUNLOCK();
2677 			break;
2678 		}
2679 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
2680 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2681 		PF_RULES_WUNLOCK();
2682 		free(pfrts, M_TEMP);
2683 		break;
2684 	}
2685 
2686 	case DIOCRSETTFLAGS: {
2687 		struct pfioc_table *io = (struct pfioc_table *)addr;
2688 		struct pfr_table *pfrts;
2689 		size_t totlen, n;
2690 
2691 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2692 			error = ENODEV;
2693 			break;
2694 		}
2695 
2696 		PF_RULES_WLOCK();
2697 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
2698 		io->pfrio_size = min(io->pfrio_size, n);
2699 
2700 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2701 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
2702 		    M_TEMP, M_NOWAIT);
2703 		if (pfrts == NULL) {
2704 			error = ENOMEM;
2705 			PF_RULES_WUNLOCK();
2706 			break;
2707 		}
2708 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2709 		if (error) {
2710 			free(pfrts, M_TEMP);
2711 			PF_RULES_WUNLOCK();
2712 			break;
2713 		}
2714 		error = pfr_set_tflags(pfrts, io->pfrio_size,
2715 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2716 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2717 		PF_RULES_WUNLOCK();
2718 		free(pfrts, M_TEMP);
2719 		break;
2720 	}
2721 
2722 	case DIOCRCLRADDRS: {
2723 		struct pfioc_table *io = (struct pfioc_table *)addr;
2724 
2725 		if (io->pfrio_esize != 0) {
2726 			error = ENODEV;
2727 			break;
2728 		}
2729 		PF_RULES_WLOCK();
2730 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2731 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2732 		PF_RULES_WUNLOCK();
2733 		break;
2734 	}
2735 
2736 	case DIOCRADDADDRS: {
2737 		struct pfioc_table *io = (struct pfioc_table *)addr;
2738 		struct pfr_addr *pfras;
2739 		size_t totlen;
2740 
2741 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2742 			error = ENODEV;
2743 			break;
2744 		}
2745 		if (io->pfrio_size < 0 ||
2746 		    io->pfrio_size > pf_ioctl_maxcount ||
2747 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
2748 			error = EINVAL;
2749 			break;
2750 		}
2751 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2752 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
2753 		    M_TEMP, M_NOWAIT);
2754 		if (! pfras) {
2755 			error = ENOMEM;
2756 			break;
2757 		}
2758 		error = copyin(io->pfrio_buffer, pfras, totlen);
2759 		if (error) {
2760 			free(pfras, M_TEMP);
2761 			break;
2762 		}
2763 		PF_RULES_WLOCK();
2764 		error = pfr_add_addrs(&io->pfrio_table, pfras,
2765 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2766 		    PFR_FLAG_USERIOCTL);
2767 		PF_RULES_WUNLOCK();
2768 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2769 			error = copyout(pfras, io->pfrio_buffer, totlen);
2770 		free(pfras, M_TEMP);
2771 		break;
2772 	}
2773 
2774 	case DIOCRDELADDRS: {
2775 		struct pfioc_table *io = (struct pfioc_table *)addr;
2776 		struct pfr_addr *pfras;
2777 		size_t totlen;
2778 
2779 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2780 			error = ENODEV;
2781 			break;
2782 		}
2783 		if (io->pfrio_size < 0 ||
2784 		    io->pfrio_size > pf_ioctl_maxcount ||
2785 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
2786 			error = EINVAL;
2787 			break;
2788 		}
2789 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2790 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
2791 		    M_TEMP, M_NOWAIT);
2792 		if (! pfras) {
2793 			error = ENOMEM;
2794 			break;
2795 		}
2796 		error = copyin(io->pfrio_buffer, pfras, totlen);
2797 		if (error) {
2798 			free(pfras, M_TEMP);
2799 			break;
2800 		}
2801 		PF_RULES_WLOCK();
2802 		error = pfr_del_addrs(&io->pfrio_table, pfras,
2803 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2804 		    PFR_FLAG_USERIOCTL);
2805 		PF_RULES_WUNLOCK();
2806 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2807 			error = copyout(pfras, io->pfrio_buffer, totlen);
2808 		free(pfras, M_TEMP);
2809 		break;
2810 	}
2811 
2812 	case DIOCRSETADDRS: {
2813 		struct pfioc_table *io = (struct pfioc_table *)addr;
2814 		struct pfr_addr *pfras;
2815 		size_t totlen, count;
2816 
2817 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2818 			error = ENODEV;
2819 			break;
2820 		}
2821 		if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
2822 			error = EINVAL;
2823 			break;
2824 		}
2825 		count = max(io->pfrio_size, io->pfrio_size2);
2826 		if (count > pf_ioctl_maxcount ||
2827 		    WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
2828 			error = EINVAL;
2829 			break;
2830 		}
2831 		totlen = count * sizeof(struct pfr_addr);
2832 		pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
2833 		    M_NOWAIT);
2834 		if (! pfras) {
2835 			error = ENOMEM;
2836 			break;
2837 		}
2838 		error = copyin(io->pfrio_buffer, pfras, totlen);
2839 		if (error) {
2840 			free(pfras, M_TEMP);
2841 			break;
2842 		}
2843 		PF_RULES_WLOCK();
2844 		error = pfr_set_addrs(&io->pfrio_table, pfras,
2845 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2846 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2847 		    PFR_FLAG_USERIOCTL, 0);
2848 		PF_RULES_WUNLOCK();
2849 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2850 			error = copyout(pfras, io->pfrio_buffer, totlen);
2851 		free(pfras, M_TEMP);
2852 		break;
2853 	}
2854 
2855 	case DIOCRGETADDRS: {
2856 		struct pfioc_table *io = (struct pfioc_table *)addr;
2857 		struct pfr_addr *pfras;
2858 		size_t totlen;
2859 
2860 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2861 			error = ENODEV;
2862 			break;
2863 		}
2864 		if (io->pfrio_size < 0 ||
2865 		    io->pfrio_size > pf_ioctl_maxcount ||
2866 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
2867 			error = EINVAL;
2868 			break;
2869 		}
2870 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2871 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
2872 		    M_TEMP, M_NOWAIT);
2873 		if (! pfras) {
2874 			error = ENOMEM;
2875 			break;
2876 		}
2877 		PF_RULES_RLOCK();
2878 		error = pfr_get_addrs(&io->pfrio_table, pfras,
2879 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2880 		PF_RULES_RUNLOCK();
2881 		if (error == 0)
2882 			error = copyout(pfras, io->pfrio_buffer, totlen);
2883 		free(pfras, M_TEMP);
2884 		break;
2885 	}
2886 
2887 	case DIOCRGETASTATS: {
2888 		struct pfioc_table *io = (struct pfioc_table *)addr;
2889 		struct pfr_astats *pfrastats;
2890 		size_t totlen;
2891 
2892 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2893 			error = ENODEV;
2894 			break;
2895 		}
2896 		if (io->pfrio_size < 0 ||
2897 		    io->pfrio_size > pf_ioctl_maxcount ||
2898 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
2899 			error = EINVAL;
2900 			break;
2901 		}
2902 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
2903 		pfrastats = mallocarray(io->pfrio_size,
2904 		    sizeof(struct pfr_astats), M_TEMP, M_NOWAIT);
2905 		if (! pfrastats) {
2906 			error = ENOMEM;
2907 			break;
2908 		}
2909 		PF_RULES_RLOCK();
2910 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
2911 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2912 		PF_RULES_RUNLOCK();
2913 		if (error == 0)
2914 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
2915 		free(pfrastats, M_TEMP);
2916 		break;
2917 	}
2918 
2919 	case DIOCRCLRASTATS: {
2920 		struct pfioc_table *io = (struct pfioc_table *)addr;
2921 		struct pfr_addr *pfras;
2922 		size_t totlen;
2923 
2924 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2925 			error = ENODEV;
2926 			break;
2927 		}
2928 		if (io->pfrio_size < 0 ||
2929 		    io->pfrio_size > pf_ioctl_maxcount ||
2930 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
2931 			error = EINVAL;
2932 			break;
2933 		}
2934 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2935 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
2936 		    M_TEMP, M_NOWAIT);
2937 		if (! pfras) {
2938 			error = ENOMEM;
2939 			break;
2940 		}
2941 		error = copyin(io->pfrio_buffer, pfras, totlen);
2942 		if (error) {
2943 			free(pfras, M_TEMP);
2944 			break;
2945 		}
2946 		PF_RULES_WLOCK();
2947 		error = pfr_clr_astats(&io->pfrio_table, pfras,
2948 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2949 		    PFR_FLAG_USERIOCTL);
2950 		PF_RULES_WUNLOCK();
2951 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
2952 			error = copyout(pfras, io->pfrio_buffer, totlen);
2953 		free(pfras, M_TEMP);
2954 		break;
2955 	}
2956 
2957 	case DIOCRTSTADDRS: {
2958 		struct pfioc_table *io = (struct pfioc_table *)addr;
2959 		struct pfr_addr *pfras;
2960 		size_t totlen;
2961 
2962 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2963 			error = ENODEV;
2964 			break;
2965 		}
2966 		if (io->pfrio_size < 0 ||
2967 		    io->pfrio_size > pf_ioctl_maxcount ||
2968 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
2969 			error = EINVAL;
2970 			break;
2971 		}
2972 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
2973 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
2974 		    M_TEMP, M_NOWAIT);
2975 		if (! pfras) {
2976 			error = ENOMEM;
2977 			break;
2978 		}
2979 		error = copyin(io->pfrio_buffer, pfras, totlen);
2980 		if (error) {
2981 			free(pfras, M_TEMP);
2982 			break;
2983 		}
2984 		PF_RULES_RLOCK();
2985 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
2986 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2987 		    PFR_FLAG_USERIOCTL);
2988 		PF_RULES_RUNLOCK();
2989 		if (error == 0)
2990 			error = copyout(pfras, io->pfrio_buffer, totlen);
2991 		free(pfras, M_TEMP);
2992 		break;
2993 	}
2994 
2995 	case DIOCRINADEFINE: {
2996 		struct pfioc_table *io = (struct pfioc_table *)addr;
2997 		struct pfr_addr *pfras;
2998 		size_t totlen;
2999 
3000 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3001 			error = ENODEV;
3002 			break;
3003 		}
3004 		if (io->pfrio_size < 0 ||
3005 		    io->pfrio_size > pf_ioctl_maxcount ||
3006 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3007 			error = EINVAL;
3008 			break;
3009 		}
3010 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3011 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3012 		    M_TEMP, M_NOWAIT);
3013 		if (! pfras) {
3014 			error = ENOMEM;
3015 			break;
3016 		}
3017 		error = copyin(io->pfrio_buffer, pfras, totlen);
3018 		if (error) {
3019 			free(pfras, M_TEMP);
3020 			break;
3021 		}
3022 		PF_RULES_WLOCK();
3023 		error = pfr_ina_define(&io->pfrio_table, pfras,
3024 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
3025 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3026 		PF_RULES_WUNLOCK();
3027 		free(pfras, M_TEMP);
3028 		break;
3029 	}
3030 
3031 	case DIOCOSFPADD: {
3032 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
3033 		PF_RULES_WLOCK();
3034 		error = pf_osfp_add(io);
3035 		PF_RULES_WUNLOCK();
3036 		break;
3037 	}
3038 
3039 	case DIOCOSFPGET: {
3040 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
3041 		PF_RULES_RLOCK();
3042 		error = pf_osfp_get(io);
3043 		PF_RULES_RUNLOCK();
3044 		break;
3045 	}
3046 
3047 	case DIOCXBEGIN: {
3048 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3049 		struct pfioc_trans_e	*ioes, *ioe;
3050 		size_t			 totlen;
3051 		int			 i;
3052 
3053 		if (io->esize != sizeof(*ioe)) {
3054 			error = ENODEV;
3055 			break;
3056 		}
3057 		if (io->size < 0 ||
3058 		    io->size > pf_ioctl_maxcount ||
3059 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3060 			error = EINVAL;
3061 			break;
3062 		}
3063 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3064 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3065 		    M_TEMP, M_NOWAIT);
3066 		if (! ioes) {
3067 			error = ENOMEM;
3068 			break;
3069 		}
3070 		error = copyin(io->array, ioes, totlen);
3071 		if (error) {
3072 			free(ioes, M_TEMP);
3073 			break;
3074 		}
3075 		PF_RULES_WLOCK();
3076 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3077 			switch (ioe->rs_num) {
3078 #ifdef ALTQ
3079 			case PF_RULESET_ALTQ:
3080 				if (ioe->anchor[0]) {
3081 					PF_RULES_WUNLOCK();
3082 					free(ioes, M_TEMP);
3083 					error = EINVAL;
3084 					goto fail;
3085 				}
3086 				if ((error = pf_begin_altq(&ioe->ticket))) {
3087 					PF_RULES_WUNLOCK();
3088 					free(ioes, M_TEMP);
3089 					goto fail;
3090 				}
3091 				break;
3092 #endif /* ALTQ */
3093 			case PF_RULESET_TABLE:
3094 			    {
3095 				struct pfr_table table;
3096 
3097 				bzero(&table, sizeof(table));
3098 				strlcpy(table.pfrt_anchor, ioe->anchor,
3099 				    sizeof(table.pfrt_anchor));
3100 				if ((error = pfr_ina_begin(&table,
3101 				    &ioe->ticket, NULL, 0))) {
3102 					PF_RULES_WUNLOCK();
3103 					free(ioes, M_TEMP);
3104 					goto fail;
3105 				}
3106 				break;
3107 			    }
3108 			default:
3109 				if ((error = pf_begin_rules(&ioe->ticket,
3110 				    ioe->rs_num, ioe->anchor))) {
3111 					PF_RULES_WUNLOCK();
3112 					free(ioes, M_TEMP);
3113 					goto fail;
3114 				}
3115 				break;
3116 			}
3117 		}
3118 		PF_RULES_WUNLOCK();
3119 		error = copyout(ioes, io->array, totlen);
3120 		free(ioes, M_TEMP);
3121 		break;
3122 	}
3123 
3124 	case DIOCXROLLBACK: {
3125 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3126 		struct pfioc_trans_e	*ioe, *ioes;
3127 		size_t			 totlen;
3128 		int			 i;
3129 
3130 		if (io->esize != sizeof(*ioe)) {
3131 			error = ENODEV;
3132 			break;
3133 		}
3134 		if (io->size < 0 ||
3135 		    io->size > pf_ioctl_maxcount ||
3136 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3137 			error = EINVAL;
3138 			break;
3139 		}
3140 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3141 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3142 		    M_TEMP, M_NOWAIT);
3143 		if (! ioes) {
3144 			error = ENOMEM;
3145 			break;
3146 		}
3147 		error = copyin(io->array, ioes, totlen);
3148 		if (error) {
3149 			free(ioes, M_TEMP);
3150 			break;
3151 		}
3152 		PF_RULES_WLOCK();
3153 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3154 			switch (ioe->rs_num) {
3155 #ifdef ALTQ
3156 			case PF_RULESET_ALTQ:
3157 				if (ioe->anchor[0]) {
3158 					PF_RULES_WUNLOCK();
3159 					free(ioes, M_TEMP);
3160 					error = EINVAL;
3161 					goto fail;
3162 				}
3163 				if ((error = pf_rollback_altq(ioe->ticket))) {
3164 					PF_RULES_WUNLOCK();
3165 					free(ioes, M_TEMP);
3166 					goto fail; /* really bad */
3167 				}
3168 				break;
3169 #endif /* ALTQ */
3170 			case PF_RULESET_TABLE:
3171 			    {
3172 				struct pfr_table table;
3173 
3174 				bzero(&table, sizeof(table));
3175 				strlcpy(table.pfrt_anchor, ioe->anchor,
3176 				    sizeof(table.pfrt_anchor));
3177 				if ((error = pfr_ina_rollback(&table,
3178 				    ioe->ticket, NULL, 0))) {
3179 					PF_RULES_WUNLOCK();
3180 					free(ioes, M_TEMP);
3181 					goto fail; /* really bad */
3182 				}
3183 				break;
3184 			    }
3185 			default:
3186 				if ((error = pf_rollback_rules(ioe->ticket,
3187 				    ioe->rs_num, ioe->anchor))) {
3188 					PF_RULES_WUNLOCK();
3189 					free(ioes, M_TEMP);
3190 					goto fail; /* really bad */
3191 				}
3192 				break;
3193 			}
3194 		}
3195 		PF_RULES_WUNLOCK();
3196 		free(ioes, M_TEMP);
3197 		break;
3198 	}
3199 
3200 	case DIOCXCOMMIT: {
3201 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3202 		struct pfioc_trans_e	*ioe, *ioes;
3203 		struct pf_ruleset	*rs;
3204 		size_t			 totlen;
3205 		int			 i;
3206 
3207 		if (io->esize != sizeof(*ioe)) {
3208 			error = ENODEV;
3209 			break;
3210 		}
3211 
3212 		if (io->size < 0 ||
3213 		    io->size > pf_ioctl_maxcount ||
3214 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3215 			error = EINVAL;
3216 			break;
3217 		}
3218 
3219 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3220 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3221 		    M_TEMP, M_NOWAIT);
3222 		if (ioes == NULL) {
3223 			error = ENOMEM;
3224 			break;
3225 		}
3226 		error = copyin(io->array, ioes, totlen);
3227 		if (error) {
3228 			free(ioes, M_TEMP);
3229 			break;
3230 		}
3231 		PF_RULES_WLOCK();
3232 		/* First makes sure everything will succeed. */
3233 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3234 			switch (ioe->rs_num) {
3235 #ifdef ALTQ
3236 			case PF_RULESET_ALTQ:
3237 				if (ioe->anchor[0]) {
3238 					PF_RULES_WUNLOCK();
3239 					free(ioes, M_TEMP);
3240 					error = EINVAL;
3241 					goto fail;
3242 				}
3243 				if (!V_altqs_inactive_open || ioe->ticket !=
3244 				    V_ticket_altqs_inactive) {
3245 					PF_RULES_WUNLOCK();
3246 					free(ioes, M_TEMP);
3247 					error = EBUSY;
3248 					goto fail;
3249 				}
3250 				break;
3251 #endif /* ALTQ */
3252 			case PF_RULESET_TABLE:
3253 				rs = pf_find_ruleset(ioe->anchor);
3254 				if (rs == NULL || !rs->topen || ioe->ticket !=
3255 				    rs->tticket) {
3256 					PF_RULES_WUNLOCK();
3257 					free(ioes, M_TEMP);
3258 					error = EBUSY;
3259 					goto fail;
3260 				}
3261 				break;
3262 			default:
3263 				if (ioe->rs_num < 0 || ioe->rs_num >=
3264 				    PF_RULESET_MAX) {
3265 					PF_RULES_WUNLOCK();
3266 					free(ioes, M_TEMP);
3267 					error = EINVAL;
3268 					goto fail;
3269 				}
3270 				rs = pf_find_ruleset(ioe->anchor);
3271 				if (rs == NULL ||
3272 				    !rs->rules[ioe->rs_num].inactive.open ||
3273 				    rs->rules[ioe->rs_num].inactive.ticket !=
3274 				    ioe->ticket) {
3275 					PF_RULES_WUNLOCK();
3276 					free(ioes, M_TEMP);
3277 					error = EBUSY;
3278 					goto fail;
3279 				}
3280 				break;
3281 			}
3282 		}
3283 		/* Now do the commit - no errors should happen here. */
3284 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3285 			switch (ioe->rs_num) {
3286 #ifdef ALTQ
3287 			case PF_RULESET_ALTQ:
3288 				if ((error = pf_commit_altq(ioe->ticket))) {
3289 					PF_RULES_WUNLOCK();
3290 					free(ioes, M_TEMP);
3291 					goto fail; /* really bad */
3292 				}
3293 				break;
3294 #endif /* ALTQ */
3295 			case PF_RULESET_TABLE:
3296 			    {
3297 				struct pfr_table table;
3298 
3299 				bzero(&table, sizeof(table));
3300 				strlcpy(table.pfrt_anchor, ioe->anchor,
3301 				    sizeof(table.pfrt_anchor));
3302 				if ((error = pfr_ina_commit(&table,
3303 				    ioe->ticket, NULL, NULL, 0))) {
3304 					PF_RULES_WUNLOCK();
3305 					free(ioes, M_TEMP);
3306 					goto fail; /* really bad */
3307 				}
3308 				break;
3309 			    }
3310 			default:
3311 				if ((error = pf_commit_rules(ioe->ticket,
3312 				    ioe->rs_num, ioe->anchor))) {
3313 					PF_RULES_WUNLOCK();
3314 					free(ioes, M_TEMP);
3315 					goto fail; /* really bad */
3316 				}
3317 				break;
3318 			}
3319 		}
3320 		PF_RULES_WUNLOCK();
3321 		free(ioes, M_TEMP);
3322 		break;
3323 	}
3324 
3325 	case DIOCGETSRCNODES: {
3326 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
3327 		struct pf_srchash	*sh;
3328 		struct pf_src_node	*n, *p, *pstore;
3329 		uint32_t		 i, nr = 0;
3330 
3331 		if (psn->psn_len == 0) {
3332 			for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3333 			    i++, sh++) {
3334 				PF_HASHROW_LOCK(sh);
3335 				LIST_FOREACH(n, &sh->nodes, entry)
3336 					nr++;
3337 				PF_HASHROW_UNLOCK(sh);
3338 			}
3339 			psn->psn_len = sizeof(struct pf_src_node) * nr;
3340 			break;
3341 		}
3342 
3343 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK);
3344 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3345 		    i++, sh++) {
3346 		    PF_HASHROW_LOCK(sh);
3347 		    LIST_FOREACH(n, &sh->nodes, entry) {
3348 			int	secs = time_uptime, diff;
3349 
3350 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
3351 				break;
3352 
3353 			bcopy(n, p, sizeof(struct pf_src_node));
3354 			if (n->rule.ptr != NULL)
3355 				p->rule.nr = n->rule.ptr->nr;
3356 			p->creation = secs - p->creation;
3357 			if (p->expire > secs)
3358 				p->expire -= secs;
3359 			else
3360 				p->expire = 0;
3361 
3362 			/* Adjust the connection rate estimate. */
3363 			diff = secs - n->conn_rate.last;
3364 			if (diff >= n->conn_rate.seconds)
3365 				p->conn_rate.count = 0;
3366 			else
3367 				p->conn_rate.count -=
3368 				    n->conn_rate.count * diff /
3369 				    n->conn_rate.seconds;
3370 			p++;
3371 			nr++;
3372 		    }
3373 		    PF_HASHROW_UNLOCK(sh);
3374 		}
3375 		error = copyout(pstore, psn->psn_src_nodes,
3376 		    sizeof(struct pf_src_node) * nr);
3377 		if (error) {
3378 			free(pstore, M_TEMP);
3379 			break;
3380 		}
3381 		psn->psn_len = sizeof(struct pf_src_node) * nr;
3382 		free(pstore, M_TEMP);
3383 		break;
3384 	}
3385 
3386 	case DIOCCLRSRCNODES: {
3387 
3388 		pf_clear_srcnodes(NULL);
3389 		pf_purge_expired_src_nodes();
3390 		break;
3391 	}
3392 
3393 	case DIOCKILLSRCNODES:
3394 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
3395 		break;
3396 
3397 	case DIOCSETHOSTID: {
3398 		u_int32_t	*hostid = (u_int32_t *)addr;
3399 
3400 		PF_RULES_WLOCK();
3401 		if (*hostid == 0)
3402 			V_pf_status.hostid = arc4random();
3403 		else
3404 			V_pf_status.hostid = *hostid;
3405 		PF_RULES_WUNLOCK();
3406 		break;
3407 	}
3408 
3409 	case DIOCOSFPFLUSH:
3410 		PF_RULES_WLOCK();
3411 		pf_osfp_flush();
3412 		PF_RULES_WUNLOCK();
3413 		break;
3414 
3415 	case DIOCIGETIFACES: {
3416 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3417 		struct pfi_kif *ifstore;
3418 		size_t bufsiz;
3419 
3420 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
3421 			error = ENODEV;
3422 			break;
3423 		}
3424 
3425 		if (io->pfiio_size < 0 ||
3426 		    io->pfiio_size > pf_ioctl_maxcount ||
3427 		    WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
3428 			error = EINVAL;
3429 			break;
3430 		}
3431 
3432 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
3433 		ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
3434 		    M_TEMP, M_NOWAIT);
3435 		if (ifstore == NULL) {
3436 			error = ENOMEM;
3437 			break;
3438 		}
3439 
3440 		PF_RULES_RLOCK();
3441 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
3442 		PF_RULES_RUNLOCK();
3443 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
3444 		free(ifstore, M_TEMP);
3445 		break;
3446 	}
3447 
3448 	case DIOCSETIFFLAG: {
3449 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3450 
3451 		PF_RULES_WLOCK();
3452 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3453 		PF_RULES_WUNLOCK();
3454 		break;
3455 	}
3456 
3457 	case DIOCCLRIFFLAG: {
3458 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3459 
3460 		PF_RULES_WLOCK();
3461 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3462 		PF_RULES_WUNLOCK();
3463 		break;
3464 	}
3465 
3466 	default:
3467 		error = ENODEV;
3468 		break;
3469 	}
3470 fail:
3471 	if (sx_xlocked(&pf_ioctl_lock))
3472 		sx_xunlock(&pf_ioctl_lock);
3473 	CURVNET_RESTORE();
3474 
3475 	return (error);
3476 }
3477 
3478 void
3479 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
3480 {
3481 	bzero(sp, sizeof(struct pfsync_state));
3482 
3483 	/* copy from state key */
3484 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
3485 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
3486 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
3487 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
3488 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
3489 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
3490 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
3491 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
3492 	sp->proto = st->key[PF_SK_WIRE]->proto;
3493 	sp->af = st->key[PF_SK_WIRE]->af;
3494 
3495 	/* copy from state */
3496 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
3497 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
3498 	sp->creation = htonl(time_uptime - st->creation);
3499 	sp->expire = pf_state_expires(st);
3500 	if (sp->expire <= time_uptime)
3501 		sp->expire = htonl(0);
3502 	else
3503 		sp->expire = htonl(sp->expire - time_uptime);
3504 
3505 	sp->direction = st->direction;
3506 	sp->log = st->log;
3507 	sp->timeout = st->timeout;
3508 	sp->state_flags = st->state_flags;
3509 	if (st->src_node)
3510 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
3511 	if (st->nat_src_node)
3512 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
3513 
3514 	sp->id = st->id;
3515 	sp->creatorid = st->creatorid;
3516 	pf_state_peer_hton(&st->src, &sp->src);
3517 	pf_state_peer_hton(&st->dst, &sp->dst);
3518 
3519 	if (st->rule.ptr == NULL)
3520 		sp->rule = htonl(-1);
3521 	else
3522 		sp->rule = htonl(st->rule.ptr->nr);
3523 	if (st->anchor.ptr == NULL)
3524 		sp->anchor = htonl(-1);
3525 	else
3526 		sp->anchor = htonl(st->anchor.ptr->nr);
3527 	if (st->nat_rule.ptr == NULL)
3528 		sp->nat_rule = htonl(-1);
3529 	else
3530 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
3531 
3532 	pf_state_counter_hton(st->packets[0], sp->packets[0]);
3533 	pf_state_counter_hton(st->packets[1], sp->packets[1]);
3534 	pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
3535 	pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
3536 
3537 }
3538 
3539 static void
3540 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
3541 {
3542 	struct pfr_ktable *kt;
3543 
3544 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
3545 
3546 	kt = aw->p.tbl;
3547 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
3548 		kt = kt->pfrkt_root;
3549 	aw->p.tbl = NULL;
3550 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
3551 		kt->pfrkt_cnt : -1;
3552 }
3553 
3554 /*
3555  * XXX - Check for version missmatch!!!
3556  */
3557 static void
3558 pf_clear_states(void)
3559 {
3560 	struct pf_state	*s;
3561 	u_int i;
3562 
3563 	for (i = 0; i <= pf_hashmask; i++) {
3564 		struct pf_idhash *ih = &V_pf_idhash[i];
3565 relock:
3566 		PF_HASHROW_LOCK(ih);
3567 		LIST_FOREACH(s, &ih->states, entry) {
3568 			s->timeout = PFTM_PURGE;
3569 			/* Don't send out individual delete messages. */
3570 			s->state_flags |= PFSTATE_NOSYNC;
3571 			pf_unlink_state(s, PF_ENTER_LOCKED);
3572 			goto relock;
3573 		}
3574 		PF_HASHROW_UNLOCK(ih);
3575 	}
3576 }
3577 
3578 static int
3579 pf_clear_tables(void)
3580 {
3581 	struct pfioc_table io;
3582 	int error;
3583 
3584 	bzero(&io, sizeof(io));
3585 
3586 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
3587 	    io.pfrio_flags);
3588 
3589 	return (error);
3590 }
3591 
3592 static void
3593 pf_clear_srcnodes(struct pf_src_node *n)
3594 {
3595 	struct pf_state *s;
3596 	int i;
3597 
3598 	for (i = 0; i <= pf_hashmask; i++) {
3599 		struct pf_idhash *ih = &V_pf_idhash[i];
3600 
3601 		PF_HASHROW_LOCK(ih);
3602 		LIST_FOREACH(s, &ih->states, entry) {
3603 			if (n == NULL || n == s->src_node)
3604 				s->src_node = NULL;
3605 			if (n == NULL || n == s->nat_src_node)
3606 				s->nat_src_node = NULL;
3607 		}
3608 		PF_HASHROW_UNLOCK(ih);
3609 	}
3610 
3611 	if (n == NULL) {
3612 		struct pf_srchash *sh;
3613 
3614 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3615 		    i++, sh++) {
3616 			PF_HASHROW_LOCK(sh);
3617 			LIST_FOREACH(n, &sh->nodes, entry) {
3618 				n->expire = 1;
3619 				n->states = 0;
3620 			}
3621 			PF_HASHROW_UNLOCK(sh);
3622 		}
3623 	} else {
3624 		/* XXX: hash slot should already be locked here. */
3625 		n->expire = 1;
3626 		n->states = 0;
3627 	}
3628 }
3629 
3630 static void
3631 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
3632 {
3633 	struct pf_src_node_list	 kill;
3634 
3635 	LIST_INIT(&kill);
3636 	for (int i = 0; i <= pf_srchashmask; i++) {
3637 		struct pf_srchash *sh = &V_pf_srchash[i];
3638 		struct pf_src_node *sn, *tmp;
3639 
3640 		PF_HASHROW_LOCK(sh);
3641 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
3642 			if (PF_MATCHA(psnk->psnk_src.neg,
3643 			      &psnk->psnk_src.addr.v.a.addr,
3644 			      &psnk->psnk_src.addr.v.a.mask,
3645 			      &sn->addr, sn->af) &&
3646 			    PF_MATCHA(psnk->psnk_dst.neg,
3647 			      &psnk->psnk_dst.addr.v.a.addr,
3648 			      &psnk->psnk_dst.addr.v.a.mask,
3649 			      &sn->raddr, sn->af)) {
3650 				pf_unlink_src_node(sn);
3651 				LIST_INSERT_HEAD(&kill, sn, entry);
3652 				sn->expire = 1;
3653 			}
3654 		PF_HASHROW_UNLOCK(sh);
3655 	}
3656 
3657 	for (int i = 0; i <= pf_hashmask; i++) {
3658 		struct pf_idhash *ih = &V_pf_idhash[i];
3659 		struct pf_state *s;
3660 
3661 		PF_HASHROW_LOCK(ih);
3662 		LIST_FOREACH(s, &ih->states, entry) {
3663 			if (s->src_node && s->src_node->expire == 1)
3664 				s->src_node = NULL;
3665 			if (s->nat_src_node && s->nat_src_node->expire == 1)
3666 				s->nat_src_node = NULL;
3667 		}
3668 		PF_HASHROW_UNLOCK(ih);
3669 	}
3670 
3671 	psnk->psnk_killed = pf_free_src_nodes(&kill);
3672 }
3673 
3674 /*
3675  * XXX - Check for version missmatch!!!
3676  */
3677 
3678 /*
3679  * Duplicate pfctl -Fa operation to get rid of as much as we can.
3680  */
3681 static int
3682 shutdown_pf(void)
3683 {
3684 	int error = 0;
3685 	u_int32_t t[5];
3686 	char nn = '\0';
3687 
3688 	do {
3689 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
3690 		    != 0) {
3691 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
3692 			break;
3693 		}
3694 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
3695 		    != 0) {
3696 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
3697 			break;		/* XXX: rollback? */
3698 		}
3699 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
3700 		    != 0) {
3701 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
3702 			break;		/* XXX: rollback? */
3703 		}
3704 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
3705 		    != 0) {
3706 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
3707 			break;		/* XXX: rollback? */
3708 		}
3709 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
3710 		    != 0) {
3711 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
3712 			break;		/* XXX: rollback? */
3713 		}
3714 
3715 		/* XXX: these should always succeed here */
3716 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
3717 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
3718 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
3719 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
3720 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
3721 
3722 		if ((error = pf_clear_tables()) != 0)
3723 			break;
3724 
3725 #ifdef ALTQ
3726 		if ((error = pf_begin_altq(&t[0])) != 0) {
3727 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
3728 			break;
3729 		}
3730 		pf_commit_altq(t[0]);
3731 #endif
3732 
3733 		pf_clear_states();
3734 
3735 		pf_clear_srcnodes(NULL);
3736 
3737 		/* status does not use malloced mem so no need to cleanup */
3738 		/* fingerprints and interfaces have their own cleanup code */
3739 
3740 		/* Free counters last as we updated them during shutdown. */
3741 		counter_u64_free(V_pf_default_rule.states_cur);
3742 		counter_u64_free(V_pf_default_rule.states_tot);
3743 		counter_u64_free(V_pf_default_rule.src_nodes);
3744 
3745 		for (int i = 0; i < PFRES_MAX; i++)
3746 			counter_u64_free(V_pf_status.counters[i]);
3747 		for (int i = 0; i < LCNT_MAX; i++)
3748 			counter_u64_free(V_pf_status.lcounters[i]);
3749 		for (int i = 0; i < FCNT_MAX; i++)
3750 			counter_u64_free(V_pf_status.fcounters[i]);
3751 		for (int i = 0; i < SCNT_MAX; i++)
3752 			counter_u64_free(V_pf_status.scounters[i]);
3753 	} while(0);
3754 
3755 	return (error);
3756 }
3757 
3758 #ifdef INET
3759 static int
3760 pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir, int flags,
3761     struct inpcb *inp)
3762 {
3763 	int chk;
3764 
3765 	chk = pf_test(PF_IN, flags, ifp, m, inp);
3766 	if (chk && *m) {
3767 		m_freem(*m);
3768 		*m = NULL;
3769 	}
3770 
3771 	if (chk != PF_PASS)
3772 		return (EACCES);
3773 	return (0);
3774 }
3775 
3776 static int
3777 pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir, int flags,
3778     struct inpcb *inp)
3779 {
3780 	int chk;
3781 
3782 	chk = pf_test(PF_OUT, flags, ifp, m, inp);
3783 	if (chk && *m) {
3784 		m_freem(*m);
3785 		*m = NULL;
3786 	}
3787 
3788 	if (chk != PF_PASS)
3789 		return (EACCES);
3790 	return (0);
3791 }
3792 #endif
3793 
3794 #ifdef INET6
3795 static int
3796 pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir, int flags,
3797     struct inpcb *inp)
3798 {
3799 	int chk;
3800 
3801 	/*
3802 	 * In case of loopback traffic IPv6 uses the real interface in
3803 	 * order to support scoped addresses. In order to support stateful
3804 	 * filtering we have change this to lo0 as it is the case in IPv4.
3805 	 */
3806 	CURVNET_SET(ifp->if_vnet);
3807 	chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
3808 	CURVNET_RESTORE();
3809 	if (chk && *m) {
3810 		m_freem(*m);
3811 		*m = NULL;
3812 	}
3813 	if (chk != PF_PASS)
3814 		return (EACCES);
3815 	return (0);
3816 }
3817 
3818 static int
3819 pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir, int flags,
3820     struct inpcb *inp)
3821 {
3822 	int chk;
3823 
3824 	CURVNET_SET(ifp->if_vnet);
3825 	chk = pf_test6(PF_OUT, flags, ifp, m, inp);
3826 	CURVNET_RESTORE();
3827 	if (chk && *m) {
3828 		m_freem(*m);
3829 		*m = NULL;
3830 	}
3831 	if (chk != PF_PASS)
3832 		return (EACCES);
3833 	return (0);
3834 }
3835 #endif /* INET6 */
3836 
3837 static int
3838 hook_pf(void)
3839 {
3840 #ifdef INET
3841 	struct pfil_head *pfh_inet;
3842 #endif
3843 #ifdef INET6
3844 	struct pfil_head *pfh_inet6;
3845 #endif
3846 
3847 	if (V_pf_pfil_hooked)
3848 		return (0);
3849 
3850 #ifdef INET
3851 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3852 	if (pfh_inet == NULL)
3853 		return (ESRCH); /* XXX */
3854 	pfil_add_hook_flags(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet);
3855 	pfil_add_hook_flags(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet);
3856 #endif
3857 #ifdef INET6
3858 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3859 	if (pfh_inet6 == NULL) {
3860 #ifdef INET
3861 		pfil_remove_hook_flags(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3862 		    pfh_inet);
3863 		pfil_remove_hook_flags(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3864 		    pfh_inet);
3865 #endif
3866 		return (ESRCH); /* XXX */
3867 	}
3868 	pfil_add_hook_flags(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet6);
3869 	pfil_add_hook_flags(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet6);
3870 #endif
3871 
3872 	V_pf_pfil_hooked = 1;
3873 	return (0);
3874 }
3875 
3876 static int
3877 dehook_pf(void)
3878 {
3879 #ifdef INET
3880 	struct pfil_head *pfh_inet;
3881 #endif
3882 #ifdef INET6
3883 	struct pfil_head *pfh_inet6;
3884 #endif
3885 
3886 	if (V_pf_pfil_hooked == 0)
3887 		return (0);
3888 
3889 #ifdef INET
3890 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3891 	if (pfh_inet == NULL)
3892 		return (ESRCH); /* XXX */
3893 	pfil_remove_hook_flags(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3894 	    pfh_inet);
3895 	pfil_remove_hook_flags(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3896 	    pfh_inet);
3897 #endif
3898 #ifdef INET6
3899 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3900 	if (pfh_inet6 == NULL)
3901 		return (ESRCH); /* XXX */
3902 	pfil_remove_hook_flags(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK,
3903 	    pfh_inet6);
3904 	pfil_remove_hook_flags(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK,
3905 	    pfh_inet6);
3906 #endif
3907 
3908 	V_pf_pfil_hooked = 0;
3909 	return (0);
3910 }
3911 
3912 static void
3913 pf_load_vnet(void)
3914 {
3915 	TAILQ_INIT(&V_pf_tags);
3916 	TAILQ_INIT(&V_pf_qids);
3917 
3918 	pfattach_vnet();
3919 	V_pf_vnet_active = 1;
3920 }
3921 
3922 static int
3923 pf_load(void)
3924 {
3925 	int error;
3926 
3927 	rm_init(&pf_rules_lock, "pf rulesets");
3928 	sx_init(&pf_ioctl_lock, "pf ioctl");
3929 	sx_init(&pf_end_lock, "pf end thread");
3930 
3931 	pf_mtag_initialize();
3932 
3933 	pf_dev = make_dev(&pf_cdevsw, 0, 0, 0, 0600, PF_NAME);
3934 	if (pf_dev == NULL)
3935 		return (ENOMEM);
3936 
3937 	pf_end_threads = 0;
3938 	error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
3939 	if (error != 0)
3940 		return (error);
3941 
3942 	pfi_initialize();
3943 
3944 	return (0);
3945 }
3946 
3947 static void
3948 pf_unload_vnet(void)
3949 {
3950 	int error;
3951 
3952 	V_pf_vnet_active = 0;
3953 	V_pf_status.running = 0;
3954 	swi_remove(V_pf_swi_cookie);
3955 	error = dehook_pf();
3956 	if (error) {
3957 		/*
3958 		 * Should not happen!
3959 		 * XXX Due to error code ESRCH, kldunload will show
3960 		 * a message like 'No such process'.
3961 		 */
3962 		printf("%s : pfil unregisteration fail\n", __FUNCTION__);
3963 		return;
3964 	}
3965 
3966 	PF_RULES_WLOCK();
3967 	shutdown_pf();
3968 	PF_RULES_WUNLOCK();
3969 
3970 	pf_unload_vnet_purge();
3971 
3972 	pf_normalize_cleanup();
3973 	PF_RULES_WLOCK();
3974 	pfi_cleanup_vnet();
3975 	PF_RULES_WUNLOCK();
3976 	pfr_cleanup();
3977 	pf_osfp_flush();
3978 	pf_cleanup();
3979 	if (IS_DEFAULT_VNET(curvnet))
3980 		pf_mtag_cleanup();
3981 }
3982 
3983 static void
3984 pf_unload(void)
3985 {
3986 
3987 	sx_xlock(&pf_end_lock);
3988 	pf_end_threads = 1;
3989 	while (pf_end_threads < 2) {
3990 		wakeup_one(pf_purge_thread);
3991 		sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
3992 	}
3993 	sx_xunlock(&pf_end_lock);
3994 
3995 	if (pf_dev != NULL)
3996 		destroy_dev(pf_dev);
3997 
3998 	pfi_cleanup();
3999 
4000 	rm_destroy(&pf_rules_lock);
4001 	sx_destroy(&pf_ioctl_lock);
4002 	sx_destroy(&pf_end_lock);
4003 }
4004 
4005 static void
4006 vnet_pf_init(void *unused __unused)
4007 {
4008 
4009 	pf_load_vnet();
4010 }
4011 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
4012     vnet_pf_init, NULL);
4013 
4014 static void
4015 vnet_pf_uninit(const void *unused __unused)
4016 {
4017 
4018 	pf_unload_vnet();
4019 }
4020 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
4021 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
4022     vnet_pf_uninit, NULL);
4023 
4024 
4025 static int
4026 pf_modevent(module_t mod, int type, void *data)
4027 {
4028 	int error = 0;
4029 
4030 	switch(type) {
4031 	case MOD_LOAD:
4032 		error = pf_load();
4033 		break;
4034 	case MOD_UNLOAD:
4035 		/* Handled in SYSUNINIT(pf_unload) to ensure it's done after
4036 		 * the vnet_pf_uninit()s */
4037 		break;
4038 	default:
4039 		error = EINVAL;
4040 		break;
4041 	}
4042 
4043 	return (error);
4044 }
4045 
4046 static moduledata_t pf_mod = {
4047 	"pf",
4048 	pf_modevent,
4049 	0
4050 };
4051 
4052 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
4053 MODULE_VERSION(pf, PF_MODVER);
4054