xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision f0483545503a78e16e256d46d458a2faae2f07ea)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2001 Daniel Hartmeier
5  * Copyright (c) 2002,2003 Henning Brauer
6  * Copyright (c) 2012 Gleb Smirnoff <glebius@FreeBSD.org>
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  *
13  *    - Redistributions of source code must retain the above copyright
14  *      notice, this list of conditions and the following disclaimer.
15  *    - Redistributions in binary form must reproduce the above
16  *      copyright notice, this list of conditions and the following
17  *      disclaimer in the documentation and/or other materials provided
18  *      with the distribution.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
26  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
28  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
30  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
31  * POSSIBILITY OF SUCH DAMAGE.
32  *
33  * Effort sponsored in part by the Defense Advanced Research Projects
34  * Agency (DARPA) and Air Force Research Laboratory, Air Force
35  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
36  *
37  *	$OpenBSD: pf_ioctl.c,v 1.213 2009/02/15 21:46:12 mbalmer Exp $
38  */
39 
40 #include <sys/cdefs.h>
41 __FBSDID("$FreeBSD$");
42 
43 #include "opt_inet.h"
44 #include "opt_inet6.h"
45 #include "opt_bpf.h"
46 #include "opt_pf.h"
47 
48 #include <sys/param.h>
49 #include <sys/bus.h>
50 #include <sys/conf.h>
51 #include <sys/endian.h>
52 #include <sys/fcntl.h>
53 #include <sys/filio.h>
54 #include <sys/interrupt.h>
55 #include <sys/jail.h>
56 #include <sys/kernel.h>
57 #include <sys/kthread.h>
58 #include <sys/lock.h>
59 #include <sys/mbuf.h>
60 #include <sys/module.h>
61 #include <sys/proc.h>
62 #include <sys/smp.h>
63 #include <sys/socket.h>
64 #include <sys/sysctl.h>
65 #include <sys/md5.h>
66 #include <sys/ucred.h>
67 
68 #include <net/if.h>
69 #include <net/if_var.h>
70 #include <net/vnet.h>
71 #include <net/route.h>
72 #include <net/pfil.h>
73 #include <net/pfvar.h>
74 #include <net/if_pfsync.h>
75 #include <net/if_pflog.h>
76 
77 #include <netinet/in.h>
78 #include <netinet/ip.h>
79 #include <netinet/ip_var.h>
80 #include <netinet6/ip6_var.h>
81 #include <netinet/ip_icmp.h>
82 
83 #ifdef INET6
84 #include <netinet/ip6.h>
85 #endif /* INET6 */
86 
87 #ifdef ALTQ
88 #include <net/altq/altq.h>
89 #endif
90 
91 static struct pf_pool	*pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t,
92 			    u_int8_t, u_int8_t, u_int8_t);
93 
94 static void		 pf_mv_pool(struct pf_palist *, struct pf_palist *);
95 static void		 pf_empty_pool(struct pf_palist *);
96 static int		 pfioctl(struct cdev *, u_long, caddr_t, int,
97 			    struct thread *);
98 #ifdef ALTQ
99 static int		 pf_begin_altq(u_int32_t *);
100 static int		 pf_rollback_altq(u_int32_t);
101 static int		 pf_commit_altq(u_int32_t);
102 static int		 pf_enable_altq(struct pf_altq *);
103 static int		 pf_disable_altq(struct pf_altq *);
104 static u_int32_t	 pf_qname2qid(char *);
105 static void		 pf_qid_unref(u_int32_t);
106 #endif /* ALTQ */
107 static int		 pf_begin_rules(u_int32_t *, int, const char *);
108 static int		 pf_rollback_rules(u_int32_t, int, char *);
109 static int		 pf_setup_pfsync_matching(struct pf_ruleset *);
110 static void		 pf_hash_rule(MD5_CTX *, struct pf_rule *);
111 static void		 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
112 static int		 pf_commit_rules(u_int32_t, int, char *);
113 static int		 pf_addr_setup(struct pf_ruleset *,
114 			    struct pf_addr_wrap *, sa_family_t);
115 static void		 pf_addr_copyout(struct pf_addr_wrap *);
116 #ifdef ALTQ
117 static int		 pf_export_kaltq(struct pf_altq *,
118 			    struct pfioc_altq_v1 *, size_t);
119 static int		 pf_import_kaltq(struct pfioc_altq_v1 *,
120 			    struct pf_altq *, size_t);
121 #endif /* ALTQ */
122 
123 VNET_DEFINE(struct pf_rule,	pf_default_rule);
124 
125 #ifdef ALTQ
126 VNET_DEFINE_STATIC(int,		pf_altq_running);
127 #define	V_pf_altq_running	VNET(pf_altq_running)
128 #endif
129 
130 #define	TAGID_MAX	 50000
131 struct pf_tagname {
132 	TAILQ_ENTRY(pf_tagname)	entries;
133 	char			name[PF_TAG_NAME_SIZE];
134 	uint16_t		tag;
135 	int			ref;
136 };
137 
138 TAILQ_HEAD(pf_tags, pf_tagname);
139 #define	V_pf_tags		VNET(pf_tags)
140 VNET_DEFINE(struct pf_tags, pf_tags);
141 #define	V_pf_qids		VNET(pf_qids)
142 VNET_DEFINE(struct pf_tags, pf_qids);
143 static MALLOC_DEFINE(M_PFTAG, "pf_tag", "pf(4) tag names");
144 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
145 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
146 
147 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
148 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
149 #endif
150 
151 static u_int16_t	 tagname2tag(struct pf_tags *, char *);
152 static u_int16_t	 pf_tagname2tag(char *);
153 static void		 tag_unref(struct pf_tags *, u_int16_t);
154 
155 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
156 
157 struct cdev *pf_dev;
158 
159 /*
160  * XXX - These are new and need to be checked when moveing to a new version
161  */
162 static void		 pf_clear_states(void);
163 static int		 pf_clear_tables(void);
164 static void		 pf_clear_srcnodes(struct pf_src_node *);
165 static void		 pf_kill_srcnodes(struct pfioc_src_node_kill *);
166 static void		 pf_tbladdr_copyout(struct pf_addr_wrap *);
167 
168 /*
169  * Wrapper functions for pfil(9) hooks
170  */
171 #ifdef INET
172 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp,
173     int flags, void *ruleset __unused, struct inpcb *inp);
174 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp,
175     int flags, void *ruleset __unused, struct inpcb *inp);
176 #endif
177 #ifdef INET6
178 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp,
179     int flags, void *ruleset __unused, struct inpcb *inp);
180 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp,
181     int flags, void *ruleset __unused, struct inpcb *inp);
182 #endif
183 
184 static int		hook_pf(void);
185 static int		dehook_pf(void);
186 static int		shutdown_pf(void);
187 static int		pf_load(void);
188 static void		pf_unload(void);
189 
190 static struct cdevsw pf_cdevsw = {
191 	.d_ioctl =	pfioctl,
192 	.d_name =	PF_NAME,
193 	.d_version =	D_VERSION,
194 };
195 
196 volatile VNET_DEFINE_STATIC(int, pf_pfil_hooked);
197 #define V_pf_pfil_hooked	VNET(pf_pfil_hooked)
198 
199 /*
200  * We need a flag that is neither hooked nor running to know when
201  * the VNET is "valid".  We primarily need this to control (global)
202  * external event, e.g., eventhandlers.
203  */
204 VNET_DEFINE(int, pf_vnet_active);
205 #define V_pf_vnet_active	VNET(pf_vnet_active)
206 
207 int pf_end_threads;
208 struct proc *pf_purge_proc;
209 
210 struct rmlock			pf_rules_lock;
211 struct sx			pf_ioctl_lock;
212 struct sx			pf_end_lock;
213 
214 /* pfsync */
215 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
216 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
217 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
218 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
219 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
220 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
221 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
222 
223 /* pflog */
224 pflog_packet_t			*pflog_packet_ptr = NULL;
225 
226 extern u_long	pf_ioctl_maxcount;
227 
228 static void
229 pfattach_vnet(void)
230 {
231 	u_int32_t *my_timeout = V_pf_default_rule.timeout;
232 
233 	pf_initialize();
234 	pfr_initialize();
235 	pfi_initialize_vnet();
236 	pf_normalize_init();
237 
238 	V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
239 	V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
240 
241 	RB_INIT(&V_pf_anchors);
242 	pf_init_ruleset(&pf_main_ruleset);
243 
244 	/* default rule should never be garbage collected */
245 	V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
246 #ifdef PF_DEFAULT_TO_DROP
247 	V_pf_default_rule.action = PF_DROP;
248 #else
249 	V_pf_default_rule.action = PF_PASS;
250 #endif
251 	V_pf_default_rule.nr = -1;
252 	V_pf_default_rule.rtableid = -1;
253 
254 	V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
255 	V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
256 	V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
257 
258 	/* initialize default timeouts */
259 	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
260 	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
261 	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
262 	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
263 	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
264 	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
265 	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
266 	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
267 	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
268 	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
269 	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
270 	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
271 	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
272 	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
273 	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
274 	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
275 	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
276 	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
277 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
278 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
279 
280 	bzero(&V_pf_status, sizeof(V_pf_status));
281 	V_pf_status.debug = PF_DEBUG_URGENT;
282 
283 	V_pf_pfil_hooked = 0;
284 
285 	/* XXX do our best to avoid a conflict */
286 	V_pf_status.hostid = arc4random();
287 
288 	for (int i = 0; i < PFRES_MAX; i++)
289 		V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
290 	for (int i = 0; i < LCNT_MAX; i++)
291 		V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
292 	for (int i = 0; i < FCNT_MAX; i++)
293 		V_pf_status.fcounters[i] = counter_u64_alloc(M_WAITOK);
294 	for (int i = 0; i < SCNT_MAX; i++)
295 		V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
296 
297 	if (swi_add(NULL, "pf send", pf_intr, curvnet, SWI_NET,
298 	    INTR_MPSAFE, &V_pf_swi_cookie) != 0)
299 		/* XXXGL: leaked all above. */
300 		return;
301 }
302 
303 
304 static struct pf_pool *
305 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
306     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
307     u_int8_t check_ticket)
308 {
309 	struct pf_ruleset	*ruleset;
310 	struct pf_rule		*rule;
311 	int			 rs_num;
312 
313 	ruleset = pf_find_ruleset(anchor);
314 	if (ruleset == NULL)
315 		return (NULL);
316 	rs_num = pf_get_ruleset_number(rule_action);
317 	if (rs_num >= PF_RULESET_MAX)
318 		return (NULL);
319 	if (active) {
320 		if (check_ticket && ticket !=
321 		    ruleset->rules[rs_num].active.ticket)
322 			return (NULL);
323 		if (r_last)
324 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
325 			    pf_rulequeue);
326 		else
327 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
328 	} else {
329 		if (check_ticket && ticket !=
330 		    ruleset->rules[rs_num].inactive.ticket)
331 			return (NULL);
332 		if (r_last)
333 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
334 			    pf_rulequeue);
335 		else
336 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
337 	}
338 	if (!r_last) {
339 		while ((rule != NULL) && (rule->nr != rule_number))
340 			rule = TAILQ_NEXT(rule, entries);
341 	}
342 	if (rule == NULL)
343 		return (NULL);
344 
345 	return (&rule->rpool);
346 }
347 
348 static void
349 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
350 {
351 	struct pf_pooladdr	*mv_pool_pa;
352 
353 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
354 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
355 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
356 	}
357 }
358 
359 static void
360 pf_empty_pool(struct pf_palist *poola)
361 {
362 	struct pf_pooladdr *pa;
363 
364 	while ((pa = TAILQ_FIRST(poola)) != NULL) {
365 		switch (pa->addr.type) {
366 		case PF_ADDR_DYNIFTL:
367 			pfi_dynaddr_remove(pa->addr.p.dyn);
368 			break;
369 		case PF_ADDR_TABLE:
370 			/* XXX: this could be unfinished pooladdr on pabuf */
371 			if (pa->addr.p.tbl != NULL)
372 				pfr_detach_table(pa->addr.p.tbl);
373 			break;
374 		}
375 		if (pa->kif)
376 			pfi_kif_unref(pa->kif);
377 		TAILQ_REMOVE(poola, pa, entries);
378 		free(pa, M_PFRULE);
379 	}
380 }
381 
382 static void
383 pf_unlink_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
384 {
385 
386 	PF_RULES_WASSERT();
387 
388 	TAILQ_REMOVE(rulequeue, rule, entries);
389 
390 	PF_UNLNKDRULES_LOCK();
391 	rule->rule_flag |= PFRULE_REFS;
392 	TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
393 	PF_UNLNKDRULES_UNLOCK();
394 }
395 
396 void
397 pf_free_rule(struct pf_rule *rule)
398 {
399 
400 	PF_RULES_WASSERT();
401 
402 	if (rule->tag)
403 		tag_unref(&V_pf_tags, rule->tag);
404 	if (rule->match_tag)
405 		tag_unref(&V_pf_tags, rule->match_tag);
406 #ifdef ALTQ
407 	if (rule->pqid != rule->qid)
408 		pf_qid_unref(rule->pqid);
409 	pf_qid_unref(rule->qid);
410 #endif
411 	switch (rule->src.addr.type) {
412 	case PF_ADDR_DYNIFTL:
413 		pfi_dynaddr_remove(rule->src.addr.p.dyn);
414 		break;
415 	case PF_ADDR_TABLE:
416 		pfr_detach_table(rule->src.addr.p.tbl);
417 		break;
418 	}
419 	switch (rule->dst.addr.type) {
420 	case PF_ADDR_DYNIFTL:
421 		pfi_dynaddr_remove(rule->dst.addr.p.dyn);
422 		break;
423 	case PF_ADDR_TABLE:
424 		pfr_detach_table(rule->dst.addr.p.tbl);
425 		break;
426 	}
427 	if (rule->overload_tbl)
428 		pfr_detach_table(rule->overload_tbl);
429 	if (rule->kif)
430 		pfi_kif_unref(rule->kif);
431 	pf_anchor_remove(rule);
432 	pf_empty_pool(&rule->rpool.list);
433 	counter_u64_free(rule->states_cur);
434 	counter_u64_free(rule->states_tot);
435 	counter_u64_free(rule->src_nodes);
436 	free(rule, M_PFRULE);
437 }
438 
439 static u_int16_t
440 tagname2tag(struct pf_tags *head, char *tagname)
441 {
442 	struct pf_tagname	*tag, *p = NULL;
443 	u_int16_t		 new_tagid = 1;
444 
445 	PF_RULES_WASSERT();
446 
447 	TAILQ_FOREACH(tag, head, entries)
448 		if (strcmp(tagname, tag->name) == 0) {
449 			tag->ref++;
450 			return (tag->tag);
451 		}
452 
453 	/*
454 	 * to avoid fragmentation, we do a linear search from the beginning
455 	 * and take the first free slot we find. if there is none or the list
456 	 * is empty, append a new entry at the end.
457 	 */
458 
459 	/* new entry */
460 	if (!TAILQ_EMPTY(head))
461 		for (p = TAILQ_FIRST(head); p != NULL &&
462 		    p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
463 			new_tagid = p->tag + 1;
464 
465 	if (new_tagid > TAGID_MAX)
466 		return (0);
467 
468 	/* allocate and fill new struct pf_tagname */
469 	tag = malloc(sizeof(*tag), M_PFTAG, M_NOWAIT|M_ZERO);
470 	if (tag == NULL)
471 		return (0);
472 	strlcpy(tag->name, tagname, sizeof(tag->name));
473 	tag->tag = new_tagid;
474 	tag->ref++;
475 
476 	if (p != NULL)	/* insert new entry before p */
477 		TAILQ_INSERT_BEFORE(p, tag, entries);
478 	else	/* either list empty or no free slot in between */
479 		TAILQ_INSERT_TAIL(head, tag, entries);
480 
481 	return (tag->tag);
482 }
483 
484 static void
485 tag_unref(struct pf_tags *head, u_int16_t tag)
486 {
487 	struct pf_tagname	*p, *next;
488 
489 	PF_RULES_WASSERT();
490 
491 	for (p = TAILQ_FIRST(head); p != NULL; p = next) {
492 		next = TAILQ_NEXT(p, entries);
493 		if (tag == p->tag) {
494 			if (--p->ref == 0) {
495 				TAILQ_REMOVE(head, p, entries);
496 				free(p, M_PFTAG);
497 			}
498 			break;
499 		}
500 	}
501 }
502 
503 static u_int16_t
504 pf_tagname2tag(char *tagname)
505 {
506 	return (tagname2tag(&V_pf_tags, tagname));
507 }
508 
509 #ifdef ALTQ
510 static u_int32_t
511 pf_qname2qid(char *qname)
512 {
513 	return ((u_int32_t)tagname2tag(&V_pf_qids, qname));
514 }
515 
516 static void
517 pf_qid_unref(u_int32_t qid)
518 {
519 	tag_unref(&V_pf_qids, (u_int16_t)qid);
520 }
521 
522 static int
523 pf_begin_altq(u_int32_t *ticket)
524 {
525 	struct pf_altq	*altq;
526 	int		 error = 0;
527 
528 	PF_RULES_WASSERT();
529 
530 	/* Purge the old altq list */
531 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
532 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
533 		if (altq->qname[0] == 0 &&
534 		    (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
535 			/* detach and destroy the discipline */
536 			error = altq_remove(altq);
537 		} else
538 			pf_qid_unref(altq->qid);
539 		free(altq, M_PFALTQ);
540 	}
541 	if (error)
542 		return (error);
543 	*ticket = ++V_ticket_altqs_inactive;
544 	V_altqs_inactive_open = 1;
545 	return (0);
546 }
547 
548 static int
549 pf_rollback_altq(u_int32_t ticket)
550 {
551 	struct pf_altq	*altq;
552 	int		 error = 0;
553 
554 	PF_RULES_WASSERT();
555 
556 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
557 		return (0);
558 	/* Purge the old altq list */
559 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
560 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
561 		if (altq->qname[0] == 0 &&
562 		   (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
563 			/* detach and destroy the discipline */
564 			error = altq_remove(altq);
565 		} else
566 			pf_qid_unref(altq->qid);
567 		free(altq, M_PFALTQ);
568 	}
569 	V_altqs_inactive_open = 0;
570 	return (error);
571 }
572 
573 static int
574 pf_commit_altq(u_int32_t ticket)
575 {
576 	struct pf_altqqueue	*old_altqs;
577 	struct pf_altq		*altq;
578 	int			 err, error = 0;
579 
580 	PF_RULES_WASSERT();
581 
582 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
583 		return (EBUSY);
584 
585 	/* swap altqs, keep the old. */
586 	old_altqs = V_pf_altqs_active;
587 	V_pf_altqs_active = V_pf_altqs_inactive;
588 	V_pf_altqs_inactive = old_altqs;
589 	V_ticket_altqs_active = V_ticket_altqs_inactive;
590 
591 	/* Attach new disciplines */
592 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
593 	if (altq->qname[0] == 0 &&
594 	   (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
595 			/* attach the discipline */
596 			error = altq_pfattach(altq);
597 			if (error == 0 && V_pf_altq_running)
598 				error = pf_enable_altq(altq);
599 			if (error != 0)
600 				return (error);
601 		}
602 	}
603 
604 	/* Purge the old altq list */
605 	while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) {
606 		TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries);
607 		if (altq->qname[0] == 0 &&
608 		    (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
609 			/* detach and destroy the discipline */
610 			if (V_pf_altq_running)
611 				error = pf_disable_altq(altq);
612 			err = altq_pfdetach(altq);
613 			if (err != 0 && error == 0)
614 				error = err;
615 			err = altq_remove(altq);
616 			if (err != 0 && error == 0)
617 				error = err;
618 		} else
619 			pf_qid_unref(altq->qid);
620 		free(altq, M_PFALTQ);
621 	}
622 
623 	V_altqs_inactive_open = 0;
624 	return (error);
625 }
626 
627 static int
628 pf_enable_altq(struct pf_altq *altq)
629 {
630 	struct ifnet		*ifp;
631 	struct tb_profile	 tb;
632 	int			 error = 0;
633 
634 	if ((ifp = ifunit(altq->ifname)) == NULL)
635 		return (EINVAL);
636 
637 	if (ifp->if_snd.altq_type != ALTQT_NONE)
638 		error = altq_enable(&ifp->if_snd);
639 
640 	/* set tokenbucket regulator */
641 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
642 		tb.rate = altq->ifbandwidth;
643 		tb.depth = altq->tbrsize;
644 		error = tbr_set(&ifp->if_snd, &tb);
645 	}
646 
647 	return (error);
648 }
649 
650 static int
651 pf_disable_altq(struct pf_altq *altq)
652 {
653 	struct ifnet		*ifp;
654 	struct tb_profile	 tb;
655 	int			 error;
656 
657 	if ((ifp = ifunit(altq->ifname)) == NULL)
658 		return (EINVAL);
659 
660 	/*
661 	 * when the discipline is no longer referenced, it was overridden
662 	 * by a new one.  if so, just return.
663 	 */
664 	if (altq->altq_disc != ifp->if_snd.altq_disc)
665 		return (0);
666 
667 	error = altq_disable(&ifp->if_snd);
668 
669 	if (error == 0) {
670 		/* clear tokenbucket regulator */
671 		tb.rate = 0;
672 		error = tbr_set(&ifp->if_snd, &tb);
673 	}
674 
675 	return (error);
676 }
677 
678 void
679 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
680 {
681 	struct ifnet	*ifp1;
682 	struct pf_altq	*a1, *a2, *a3;
683 	u_int32_t	 ticket;
684 	int		 error = 0;
685 
686 	/*
687 	 * No need to re-evaluate the configuration for events on interfaces
688 	 * that do not support ALTQ, as it's not possible for such
689 	 * interfaces to be part of the configuration.
690 	 */
691 	if (!ALTQ_IS_READY(&ifp->if_snd))
692 		return;
693 
694 	/* Interrupt userland queue modifications */
695 	if (V_altqs_inactive_open)
696 		pf_rollback_altq(V_ticket_altqs_inactive);
697 
698 	/* Start new altq ruleset */
699 	if (pf_begin_altq(&ticket))
700 		return;
701 
702 	/* Copy the current active set */
703 	TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
704 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
705 		if (a2 == NULL) {
706 			error = ENOMEM;
707 			break;
708 		}
709 		bcopy(a1, a2, sizeof(struct pf_altq));
710 
711 		if (a2->qname[0] != 0) {
712 			if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
713 				error = EBUSY;
714 				free(a2, M_PFALTQ);
715 				break;
716 			}
717 			a2->altq_disc = NULL;
718 			TAILQ_FOREACH(a3, V_pf_altqs_inactive, entries) {
719 				if (strncmp(a3->ifname, a2->ifname,
720 				    IFNAMSIZ) == 0 && a3->qname[0] == 0) {
721 					a2->altq_disc = a3->altq_disc;
722 					break;
723 				}
724 			}
725 		}
726 		/* Deactivate the interface in question */
727 		a2->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
728 		if ((ifp1 = ifunit(a2->ifname)) == NULL ||
729 		    (remove && ifp1 == ifp)) {
730 			a2->local_flags |= PFALTQ_FLAG_IF_REMOVED;
731 		} else {
732 			error = altq_add(a2);
733 
734 			if (ticket != V_ticket_altqs_inactive)
735 				error = EBUSY;
736 
737 			if (error) {
738 				free(a2, M_PFALTQ);
739 				break;
740 			}
741 		}
742 
743 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
744 	}
745 
746 	if (error != 0)
747 		pf_rollback_altq(ticket);
748 	else
749 		pf_commit_altq(ticket);
750 }
751 #endif /* ALTQ */
752 
753 static int
754 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
755 {
756 	struct pf_ruleset	*rs;
757 	struct pf_rule		*rule;
758 
759 	PF_RULES_WASSERT();
760 
761 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
762 		return (EINVAL);
763 	rs = pf_find_or_create_ruleset(anchor);
764 	if (rs == NULL)
765 		return (EINVAL);
766 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
767 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
768 		rs->rules[rs_num].inactive.rcount--;
769 	}
770 	*ticket = ++rs->rules[rs_num].inactive.ticket;
771 	rs->rules[rs_num].inactive.open = 1;
772 	return (0);
773 }
774 
775 static int
776 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
777 {
778 	struct pf_ruleset	*rs;
779 	struct pf_rule		*rule;
780 
781 	PF_RULES_WASSERT();
782 
783 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
784 		return (EINVAL);
785 	rs = pf_find_ruleset(anchor);
786 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
787 	    rs->rules[rs_num].inactive.ticket != ticket)
788 		return (0);
789 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
790 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
791 		rs->rules[rs_num].inactive.rcount--;
792 	}
793 	rs->rules[rs_num].inactive.open = 0;
794 	return (0);
795 }
796 
797 #define PF_MD5_UPD(st, elm)						\
798 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
799 
800 #define PF_MD5_UPD_STR(st, elm)						\
801 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
802 
803 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
804 		(stor) = htonl((st)->elm);				\
805 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
806 } while (0)
807 
808 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
809 		(stor) = htons((st)->elm);				\
810 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
811 } while (0)
812 
813 static void
814 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
815 {
816 	PF_MD5_UPD(pfr, addr.type);
817 	switch (pfr->addr.type) {
818 		case PF_ADDR_DYNIFTL:
819 			PF_MD5_UPD(pfr, addr.v.ifname);
820 			PF_MD5_UPD(pfr, addr.iflags);
821 			break;
822 		case PF_ADDR_TABLE:
823 			PF_MD5_UPD(pfr, addr.v.tblname);
824 			break;
825 		case PF_ADDR_ADDRMASK:
826 			/* XXX ignore af? */
827 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
828 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
829 			break;
830 	}
831 
832 	PF_MD5_UPD(pfr, port[0]);
833 	PF_MD5_UPD(pfr, port[1]);
834 	PF_MD5_UPD(pfr, neg);
835 	PF_MD5_UPD(pfr, port_op);
836 }
837 
838 static void
839 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
840 {
841 	u_int16_t x;
842 	u_int32_t y;
843 
844 	pf_hash_rule_addr(ctx, &rule->src);
845 	pf_hash_rule_addr(ctx, &rule->dst);
846 	PF_MD5_UPD_STR(rule, label);
847 	PF_MD5_UPD_STR(rule, ifname);
848 	PF_MD5_UPD_STR(rule, match_tagname);
849 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
850 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
851 	PF_MD5_UPD_HTONL(rule, prob, y);
852 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
853 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
854 	PF_MD5_UPD(rule, uid.op);
855 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
856 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
857 	PF_MD5_UPD(rule, gid.op);
858 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
859 	PF_MD5_UPD(rule, action);
860 	PF_MD5_UPD(rule, direction);
861 	PF_MD5_UPD(rule, af);
862 	PF_MD5_UPD(rule, quick);
863 	PF_MD5_UPD(rule, ifnot);
864 	PF_MD5_UPD(rule, match_tag_not);
865 	PF_MD5_UPD(rule, natpass);
866 	PF_MD5_UPD(rule, keep_state);
867 	PF_MD5_UPD(rule, proto);
868 	PF_MD5_UPD(rule, type);
869 	PF_MD5_UPD(rule, code);
870 	PF_MD5_UPD(rule, flags);
871 	PF_MD5_UPD(rule, flagset);
872 	PF_MD5_UPD(rule, allow_opts);
873 	PF_MD5_UPD(rule, rt);
874 	PF_MD5_UPD(rule, tos);
875 }
876 
877 static int
878 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
879 {
880 	struct pf_ruleset	*rs;
881 	struct pf_rule		*rule, **old_array;
882 	struct pf_rulequeue	*old_rules;
883 	int			 error;
884 	u_int32_t		 old_rcount;
885 
886 	PF_RULES_WASSERT();
887 
888 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
889 		return (EINVAL);
890 	rs = pf_find_ruleset(anchor);
891 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
892 	    ticket != rs->rules[rs_num].inactive.ticket)
893 		return (EBUSY);
894 
895 	/* Calculate checksum for the main ruleset */
896 	if (rs == &pf_main_ruleset) {
897 		error = pf_setup_pfsync_matching(rs);
898 		if (error != 0)
899 			return (error);
900 	}
901 
902 	/* Swap rules, keep the old. */
903 	old_rules = rs->rules[rs_num].active.ptr;
904 	old_rcount = rs->rules[rs_num].active.rcount;
905 	old_array = rs->rules[rs_num].active.ptr_array;
906 
907 	rs->rules[rs_num].active.ptr =
908 	    rs->rules[rs_num].inactive.ptr;
909 	rs->rules[rs_num].active.ptr_array =
910 	    rs->rules[rs_num].inactive.ptr_array;
911 	rs->rules[rs_num].active.rcount =
912 	    rs->rules[rs_num].inactive.rcount;
913 	rs->rules[rs_num].inactive.ptr = old_rules;
914 	rs->rules[rs_num].inactive.ptr_array = old_array;
915 	rs->rules[rs_num].inactive.rcount = old_rcount;
916 
917 	rs->rules[rs_num].active.ticket =
918 	    rs->rules[rs_num].inactive.ticket;
919 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
920 
921 
922 	/* Purge the old rule list. */
923 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
924 		pf_unlink_rule(old_rules, rule);
925 	if (rs->rules[rs_num].inactive.ptr_array)
926 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
927 	rs->rules[rs_num].inactive.ptr_array = NULL;
928 	rs->rules[rs_num].inactive.rcount = 0;
929 	rs->rules[rs_num].inactive.open = 0;
930 	pf_remove_if_empty_ruleset(rs);
931 
932 	return (0);
933 }
934 
935 static int
936 pf_setup_pfsync_matching(struct pf_ruleset *rs)
937 {
938 	MD5_CTX			 ctx;
939 	struct pf_rule		*rule;
940 	int			 rs_cnt;
941 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
942 
943 	MD5Init(&ctx);
944 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
945 		/* XXX PF_RULESET_SCRUB as well? */
946 		if (rs_cnt == PF_RULESET_SCRUB)
947 			continue;
948 
949 		if (rs->rules[rs_cnt].inactive.ptr_array)
950 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
951 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
952 
953 		if (rs->rules[rs_cnt].inactive.rcount) {
954 			rs->rules[rs_cnt].inactive.ptr_array =
955 			    malloc(sizeof(caddr_t) *
956 			    rs->rules[rs_cnt].inactive.rcount,
957 			    M_TEMP, M_NOWAIT);
958 
959 			if (!rs->rules[rs_cnt].inactive.ptr_array)
960 				return (ENOMEM);
961 		}
962 
963 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
964 		    entries) {
965 			pf_hash_rule(&ctx, rule);
966 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
967 		}
968 	}
969 
970 	MD5Final(digest, &ctx);
971 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
972 	return (0);
973 }
974 
975 static int
976 pf_addr_setup(struct pf_ruleset *ruleset, struct pf_addr_wrap *addr,
977     sa_family_t af)
978 {
979 	int error = 0;
980 
981 	switch (addr->type) {
982 	case PF_ADDR_TABLE:
983 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
984 		if (addr->p.tbl == NULL)
985 			error = ENOMEM;
986 		break;
987 	case PF_ADDR_DYNIFTL:
988 		error = pfi_dynaddr_setup(addr, af);
989 		break;
990 	}
991 
992 	return (error);
993 }
994 
995 static void
996 pf_addr_copyout(struct pf_addr_wrap *addr)
997 {
998 
999 	switch (addr->type) {
1000 	case PF_ADDR_DYNIFTL:
1001 		pfi_dynaddr_copyout(addr);
1002 		break;
1003 	case PF_ADDR_TABLE:
1004 		pf_tbladdr_copyout(addr);
1005 		break;
1006 	}
1007 }
1008 
1009 #ifdef ALTQ
1010 /*
1011  * Handle export of struct pf_kaltq to user binaries that may be using any
1012  * version of struct pf_altq.
1013  */
1014 static int
1015 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1016 {
1017 	u_int32_t version;
1018 
1019 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1020 		version = 0;
1021 	else
1022 		version = pa->version;
1023 
1024 	if (version > PFIOC_ALTQ_VERSION)
1025 		return (EINVAL);
1026 
1027 #define ASSIGN(x) exported_q->x = q->x
1028 #define COPY(x) \
1029 	bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1030 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1031 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1032 
1033 	switch (version) {
1034 	case 0: {
1035 		struct pf_altq_v0 *exported_q =
1036 		    &((struct pfioc_altq_v0 *)pa)->altq;
1037 
1038 		COPY(ifname);
1039 
1040 		ASSIGN(scheduler);
1041 		ASSIGN(tbrsize);
1042 		exported_q->tbrsize = SATU16(q->tbrsize);
1043 		exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1044 
1045 		COPY(qname);
1046 		COPY(parent);
1047 		ASSIGN(parent_qid);
1048 		exported_q->bandwidth = SATU32(q->bandwidth);
1049 		ASSIGN(priority);
1050 		ASSIGN(local_flags);
1051 
1052 		ASSIGN(qlimit);
1053 		ASSIGN(flags);
1054 
1055 		if (q->scheduler == ALTQT_HFSC) {
1056 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1057 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1058 			    SATU32(q->pq_u.hfsc_opts.x)
1059 
1060 			ASSIGN_OPT_SATU32(rtsc_m1);
1061 			ASSIGN_OPT(rtsc_d);
1062 			ASSIGN_OPT_SATU32(rtsc_m2);
1063 
1064 			ASSIGN_OPT_SATU32(lssc_m1);
1065 			ASSIGN_OPT(lssc_d);
1066 			ASSIGN_OPT_SATU32(lssc_m2);
1067 
1068 			ASSIGN_OPT_SATU32(ulsc_m1);
1069 			ASSIGN_OPT(ulsc_d);
1070 			ASSIGN_OPT_SATU32(ulsc_m2);
1071 
1072 			ASSIGN_OPT(flags);
1073 
1074 #undef ASSIGN_OPT
1075 #undef ASSIGN_OPT_SATU32
1076 		} else
1077 			COPY(pq_u);
1078 
1079 		ASSIGN(qid);
1080 		break;
1081 	}
1082 	case 1:	{
1083 		struct pf_altq_v1 *exported_q =
1084 		    &((struct pfioc_altq_v1 *)pa)->altq;
1085 
1086 		COPY(ifname);
1087 
1088 		ASSIGN(scheduler);
1089 		ASSIGN(tbrsize);
1090 		ASSIGN(ifbandwidth);
1091 
1092 		COPY(qname);
1093 		COPY(parent);
1094 		ASSIGN(parent_qid);
1095 		ASSIGN(bandwidth);
1096 		ASSIGN(priority);
1097 		ASSIGN(local_flags);
1098 
1099 		ASSIGN(qlimit);
1100 		ASSIGN(flags);
1101 		COPY(pq_u);
1102 
1103 		ASSIGN(qid);
1104 		break;
1105 	}
1106 	default:
1107 		panic("%s: unhandled struct pfioc_altq version", __func__);
1108 		break;
1109 	}
1110 
1111 #undef ASSIGN
1112 #undef COPY
1113 #undef SATU16
1114 #undef SATU32
1115 
1116 	return (0);
1117 }
1118 
1119 /*
1120  * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1121  * that may be using any version of it.
1122  */
1123 static int
1124 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1125 {
1126 	u_int32_t version;
1127 
1128 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1129 		version = 0;
1130 	else
1131 		version = pa->version;
1132 
1133 	if (version > PFIOC_ALTQ_VERSION)
1134 		return (EINVAL);
1135 
1136 #define ASSIGN(x) q->x = imported_q->x
1137 #define COPY(x) \
1138 	bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1139 
1140 	switch (version) {
1141 	case 0: {
1142 		struct pf_altq_v0 *imported_q =
1143 		    &((struct pfioc_altq_v0 *)pa)->altq;
1144 
1145 		COPY(ifname);
1146 
1147 		ASSIGN(scheduler);
1148 		ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1149 		ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1150 
1151 		COPY(qname);
1152 		COPY(parent);
1153 		ASSIGN(parent_qid);
1154 		ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1155 		ASSIGN(priority);
1156 		ASSIGN(local_flags);
1157 
1158 		ASSIGN(qlimit);
1159 		ASSIGN(flags);
1160 
1161 		if (imported_q->scheduler == ALTQT_HFSC) {
1162 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1163 
1164 			/*
1165 			 * The m1 and m2 parameters are being copied from
1166 			 * 32-bit to 64-bit.
1167 			 */
1168 			ASSIGN_OPT(rtsc_m1);
1169 			ASSIGN_OPT(rtsc_d);
1170 			ASSIGN_OPT(rtsc_m2);
1171 
1172 			ASSIGN_OPT(lssc_m1);
1173 			ASSIGN_OPT(lssc_d);
1174 			ASSIGN_OPT(lssc_m2);
1175 
1176 			ASSIGN_OPT(ulsc_m1);
1177 			ASSIGN_OPT(ulsc_d);
1178 			ASSIGN_OPT(ulsc_m2);
1179 
1180 			ASSIGN_OPT(flags);
1181 
1182 #undef ASSIGN_OPT
1183 		} else
1184 			COPY(pq_u);
1185 
1186 		ASSIGN(qid);
1187 		break;
1188 	}
1189 	case 1: {
1190 		struct pf_altq_v1 *imported_q =
1191 		    &((struct pfioc_altq_v1 *)pa)->altq;
1192 
1193 		COPY(ifname);
1194 
1195 		ASSIGN(scheduler);
1196 		ASSIGN(tbrsize);
1197 		ASSIGN(ifbandwidth);
1198 
1199 		COPY(qname);
1200 		COPY(parent);
1201 		ASSIGN(parent_qid);
1202 		ASSIGN(bandwidth);
1203 		ASSIGN(priority);
1204 		ASSIGN(local_flags);
1205 
1206 		ASSIGN(qlimit);
1207 		ASSIGN(flags);
1208 		COPY(pq_u);
1209 
1210 		ASSIGN(qid);
1211 		break;
1212 	}
1213 	default:
1214 		panic("%s: unhandled struct pfioc_altq version", __func__);
1215 		break;
1216 	}
1217 
1218 #undef ASSIGN
1219 #undef COPY
1220 
1221 	return (0);
1222 }
1223 #endif /* ALTQ */
1224 
1225 static int
1226 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
1227 {
1228 	int			 error = 0;
1229 	PF_RULES_RLOCK_TRACKER;
1230 
1231 	/* XXX keep in sync with switch() below */
1232 	if (securelevel_gt(td->td_ucred, 2))
1233 		switch (cmd) {
1234 		case DIOCGETRULES:
1235 		case DIOCGETRULE:
1236 		case DIOCGETADDRS:
1237 		case DIOCGETADDR:
1238 		case DIOCGETSTATE:
1239 		case DIOCSETSTATUSIF:
1240 		case DIOCGETSTATUS:
1241 		case DIOCCLRSTATUS:
1242 		case DIOCNATLOOK:
1243 		case DIOCSETDEBUG:
1244 		case DIOCGETSTATES:
1245 		case DIOCGETTIMEOUT:
1246 		case DIOCCLRRULECTRS:
1247 		case DIOCGETLIMIT:
1248 		case DIOCGETALTQSV0:
1249 		case DIOCGETALTQSV1:
1250 		case DIOCGETALTQV0:
1251 		case DIOCGETALTQV1:
1252 		case DIOCGETQSTATSV0:
1253 		case DIOCGETQSTATSV1:
1254 		case DIOCGETRULESETS:
1255 		case DIOCGETRULESET:
1256 		case DIOCRGETTABLES:
1257 		case DIOCRGETTSTATS:
1258 		case DIOCRCLRTSTATS:
1259 		case DIOCRCLRADDRS:
1260 		case DIOCRADDADDRS:
1261 		case DIOCRDELADDRS:
1262 		case DIOCRSETADDRS:
1263 		case DIOCRGETADDRS:
1264 		case DIOCRGETASTATS:
1265 		case DIOCRCLRASTATS:
1266 		case DIOCRTSTADDRS:
1267 		case DIOCOSFPGET:
1268 		case DIOCGETSRCNODES:
1269 		case DIOCCLRSRCNODES:
1270 		case DIOCIGETIFACES:
1271 		case DIOCGIFSPEEDV0:
1272 		case DIOCGIFSPEEDV1:
1273 		case DIOCSETIFFLAG:
1274 		case DIOCCLRIFFLAG:
1275 			break;
1276 		case DIOCRCLRTABLES:
1277 		case DIOCRADDTABLES:
1278 		case DIOCRDELTABLES:
1279 		case DIOCRSETTFLAGS:
1280 			if (((struct pfioc_table *)addr)->pfrio_flags &
1281 			    PFR_FLAG_DUMMY)
1282 				break; /* dummy operation ok */
1283 			return (EPERM);
1284 		default:
1285 			return (EPERM);
1286 		}
1287 
1288 	if (!(flags & FWRITE))
1289 		switch (cmd) {
1290 		case DIOCGETRULES:
1291 		case DIOCGETADDRS:
1292 		case DIOCGETADDR:
1293 		case DIOCGETSTATE:
1294 		case DIOCGETSTATUS:
1295 		case DIOCGETSTATES:
1296 		case DIOCGETTIMEOUT:
1297 		case DIOCGETLIMIT:
1298 		case DIOCGETALTQSV0:
1299 		case DIOCGETALTQSV1:
1300 		case DIOCGETALTQV0:
1301 		case DIOCGETALTQV1:
1302 		case DIOCGETQSTATSV0:
1303 		case DIOCGETQSTATSV1:
1304 		case DIOCGETRULESETS:
1305 		case DIOCGETRULESET:
1306 		case DIOCNATLOOK:
1307 		case DIOCRGETTABLES:
1308 		case DIOCRGETTSTATS:
1309 		case DIOCRGETADDRS:
1310 		case DIOCRGETASTATS:
1311 		case DIOCRTSTADDRS:
1312 		case DIOCOSFPGET:
1313 		case DIOCGETSRCNODES:
1314 		case DIOCIGETIFACES:
1315 		case DIOCGIFSPEEDV1:
1316 		case DIOCGIFSPEEDV0:
1317 			break;
1318 		case DIOCRCLRTABLES:
1319 		case DIOCRADDTABLES:
1320 		case DIOCRDELTABLES:
1321 		case DIOCRCLRTSTATS:
1322 		case DIOCRCLRADDRS:
1323 		case DIOCRADDADDRS:
1324 		case DIOCRDELADDRS:
1325 		case DIOCRSETADDRS:
1326 		case DIOCRSETTFLAGS:
1327 			if (((struct pfioc_table *)addr)->pfrio_flags &
1328 			    PFR_FLAG_DUMMY) {
1329 				flags |= FWRITE; /* need write lock for dummy */
1330 				break; /* dummy operation ok */
1331 			}
1332 			return (EACCES);
1333 		case DIOCGETRULE:
1334 			if (((struct pfioc_rule *)addr)->action ==
1335 			    PF_GET_CLR_CNTR)
1336 				return (EACCES);
1337 			break;
1338 		default:
1339 			return (EACCES);
1340 		}
1341 
1342 	CURVNET_SET(TD_TO_VNET(td));
1343 
1344 	switch (cmd) {
1345 	case DIOCSTART:
1346 		sx_xlock(&pf_ioctl_lock);
1347 		if (V_pf_status.running)
1348 			error = EEXIST;
1349 		else {
1350 			int cpu;
1351 
1352 			error = hook_pf();
1353 			if (error) {
1354 				DPFPRINTF(PF_DEBUG_MISC,
1355 				    ("pf: pfil registration failed\n"));
1356 				break;
1357 			}
1358 			V_pf_status.running = 1;
1359 			V_pf_status.since = time_second;
1360 
1361 			CPU_FOREACH(cpu)
1362 				V_pf_stateid[cpu] = time_second;
1363 
1364 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1365 		}
1366 		break;
1367 
1368 	case DIOCSTOP:
1369 		sx_xlock(&pf_ioctl_lock);
1370 		if (!V_pf_status.running)
1371 			error = ENOENT;
1372 		else {
1373 			V_pf_status.running = 0;
1374 			error = dehook_pf();
1375 			if (error) {
1376 				V_pf_status.running = 1;
1377 				DPFPRINTF(PF_DEBUG_MISC,
1378 				    ("pf: pfil unregistration failed\n"));
1379 			}
1380 			V_pf_status.since = time_second;
1381 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1382 		}
1383 		break;
1384 
1385 	case DIOCADDRULE: {
1386 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1387 		struct pf_ruleset	*ruleset;
1388 		struct pf_rule		*rule, *tail;
1389 		struct pf_pooladdr	*pa;
1390 		struct pfi_kif		*kif = NULL;
1391 		int			 rs_num;
1392 
1393 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1394 			error = EINVAL;
1395 			break;
1396 		}
1397 #ifndef INET
1398 		if (pr->rule.af == AF_INET) {
1399 			error = EAFNOSUPPORT;
1400 			break;
1401 		}
1402 #endif /* INET */
1403 #ifndef INET6
1404 		if (pr->rule.af == AF_INET6) {
1405 			error = EAFNOSUPPORT;
1406 			break;
1407 		}
1408 #endif /* INET6 */
1409 
1410 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
1411 		bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1412 		if (rule->ifname[0])
1413 			kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1414 		rule->states_cur = counter_u64_alloc(M_WAITOK);
1415 		rule->states_tot = counter_u64_alloc(M_WAITOK);
1416 		rule->src_nodes = counter_u64_alloc(M_WAITOK);
1417 		rule->cuid = td->td_ucred->cr_ruid;
1418 		rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1419 		TAILQ_INIT(&rule->rpool.list);
1420 
1421 #define	ERROUT(x)	{ error = (x); goto DIOCADDRULE_error; }
1422 
1423 		PF_RULES_WLOCK();
1424 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1425 		ruleset = pf_find_ruleset(pr->anchor);
1426 		if (ruleset == NULL)
1427 			ERROUT(EINVAL);
1428 		rs_num = pf_get_ruleset_number(pr->rule.action);
1429 		if (rs_num >= PF_RULESET_MAX)
1430 			ERROUT(EINVAL);
1431 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1432 			DPFPRINTF(PF_DEBUG_MISC,
1433 			    ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
1434 			    ruleset->rules[rs_num].inactive.ticket));
1435 			ERROUT(EBUSY);
1436 		}
1437 		if (pr->pool_ticket != V_ticket_pabuf) {
1438 			DPFPRINTF(PF_DEBUG_MISC,
1439 			    ("pool_ticket: %d != %d\n", pr->pool_ticket,
1440 			    V_ticket_pabuf));
1441 			ERROUT(EBUSY);
1442 		}
1443 
1444 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1445 		    pf_rulequeue);
1446 		if (tail)
1447 			rule->nr = tail->nr + 1;
1448 		else
1449 			rule->nr = 0;
1450 		if (rule->ifname[0]) {
1451 			rule->kif = pfi_kif_attach(kif, rule->ifname);
1452 			pfi_kif_ref(rule->kif);
1453 		} else
1454 			rule->kif = NULL;
1455 
1456 		if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
1457 			error = EBUSY;
1458 
1459 #ifdef ALTQ
1460 		/* set queue IDs */
1461 		if (rule->qname[0] != 0) {
1462 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1463 				error = EBUSY;
1464 			else if (rule->pqname[0] != 0) {
1465 				if ((rule->pqid =
1466 				    pf_qname2qid(rule->pqname)) == 0)
1467 					error = EBUSY;
1468 			} else
1469 				rule->pqid = rule->qid;
1470 		}
1471 #endif
1472 		if (rule->tagname[0])
1473 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1474 				error = EBUSY;
1475 		if (rule->match_tagname[0])
1476 			if ((rule->match_tag =
1477 			    pf_tagname2tag(rule->match_tagname)) == 0)
1478 				error = EBUSY;
1479 		if (rule->rt && !rule->direction)
1480 			error = EINVAL;
1481 		if (!rule->log)
1482 			rule->logif = 0;
1483 		if (rule->logif >= PFLOGIFS_MAX)
1484 			error = EINVAL;
1485 		if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1486 			error = ENOMEM;
1487 		if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1488 			error = ENOMEM;
1489 		if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1490 			error = EINVAL;
1491 		if (rule->scrub_flags & PFSTATE_SETPRIO &&
1492 		    (rule->set_prio[0] > PF_PRIO_MAX ||
1493 		    rule->set_prio[1] > PF_PRIO_MAX))
1494 			error = EINVAL;
1495 		TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1496 			if (pa->addr.type == PF_ADDR_TABLE) {
1497 				pa->addr.p.tbl = pfr_attach_table(ruleset,
1498 				    pa->addr.v.tblname);
1499 				if (pa->addr.p.tbl == NULL)
1500 					error = ENOMEM;
1501 			}
1502 
1503 		rule->overload_tbl = NULL;
1504 		if (rule->overload_tblname[0]) {
1505 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1506 			    rule->overload_tblname)) == NULL)
1507 				error = EINVAL;
1508 			else
1509 				rule->overload_tbl->pfrkt_flags |=
1510 				    PFR_TFLAG_ACTIVE;
1511 		}
1512 
1513 		pf_mv_pool(&V_pf_pabuf, &rule->rpool.list);
1514 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1515 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1516 		    (rule->rt > PF_NOPFROUTE)) &&
1517 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1518 			error = EINVAL;
1519 
1520 		if (error) {
1521 			pf_free_rule(rule);
1522 			PF_RULES_WUNLOCK();
1523 			break;
1524 		}
1525 
1526 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1527 		rule->evaluations = rule->packets[0] = rule->packets[1] =
1528 		    rule->bytes[0] = rule->bytes[1] = 0;
1529 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1530 		    rule, entries);
1531 		ruleset->rules[rs_num].inactive.rcount++;
1532 		PF_RULES_WUNLOCK();
1533 		break;
1534 
1535 #undef ERROUT
1536 DIOCADDRULE_error:
1537 		PF_RULES_WUNLOCK();
1538 		counter_u64_free(rule->states_cur);
1539 		counter_u64_free(rule->states_tot);
1540 		counter_u64_free(rule->src_nodes);
1541 		free(rule, M_PFRULE);
1542 		if (kif)
1543 			free(kif, PFI_MTYPE);
1544 		break;
1545 	}
1546 
1547 	case DIOCGETRULES: {
1548 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1549 		struct pf_ruleset	*ruleset;
1550 		struct pf_rule		*tail;
1551 		int			 rs_num;
1552 
1553 		PF_RULES_WLOCK();
1554 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1555 		ruleset = pf_find_ruleset(pr->anchor);
1556 		if (ruleset == NULL) {
1557 			PF_RULES_WUNLOCK();
1558 			error = EINVAL;
1559 			break;
1560 		}
1561 		rs_num = pf_get_ruleset_number(pr->rule.action);
1562 		if (rs_num >= PF_RULESET_MAX) {
1563 			PF_RULES_WUNLOCK();
1564 			error = EINVAL;
1565 			break;
1566 		}
1567 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1568 		    pf_rulequeue);
1569 		if (tail)
1570 			pr->nr = tail->nr + 1;
1571 		else
1572 			pr->nr = 0;
1573 		pr->ticket = ruleset->rules[rs_num].active.ticket;
1574 		PF_RULES_WUNLOCK();
1575 		break;
1576 	}
1577 
1578 	case DIOCGETRULE: {
1579 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1580 		struct pf_ruleset	*ruleset;
1581 		struct pf_rule		*rule;
1582 		int			 rs_num, i;
1583 
1584 		PF_RULES_WLOCK();
1585 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1586 		ruleset = pf_find_ruleset(pr->anchor);
1587 		if (ruleset == NULL) {
1588 			PF_RULES_WUNLOCK();
1589 			error = EINVAL;
1590 			break;
1591 		}
1592 		rs_num = pf_get_ruleset_number(pr->rule.action);
1593 		if (rs_num >= PF_RULESET_MAX) {
1594 			PF_RULES_WUNLOCK();
1595 			error = EINVAL;
1596 			break;
1597 		}
1598 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1599 			PF_RULES_WUNLOCK();
1600 			error = EBUSY;
1601 			break;
1602 		}
1603 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1604 		while ((rule != NULL) && (rule->nr != pr->nr))
1605 			rule = TAILQ_NEXT(rule, entries);
1606 		if (rule == NULL) {
1607 			PF_RULES_WUNLOCK();
1608 			error = EBUSY;
1609 			break;
1610 		}
1611 		bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1612 		pr->rule.u_states_cur = counter_u64_fetch(rule->states_cur);
1613 		pr->rule.u_states_tot = counter_u64_fetch(rule->states_tot);
1614 		pr->rule.u_src_nodes = counter_u64_fetch(rule->src_nodes);
1615 		if (pf_anchor_copyout(ruleset, rule, pr)) {
1616 			PF_RULES_WUNLOCK();
1617 			error = EBUSY;
1618 			break;
1619 		}
1620 		pf_addr_copyout(&pr->rule.src.addr);
1621 		pf_addr_copyout(&pr->rule.dst.addr);
1622 		for (i = 0; i < PF_SKIP_COUNT; ++i)
1623 			if (rule->skip[i].ptr == NULL)
1624 				pr->rule.skip[i].nr = -1;
1625 			else
1626 				pr->rule.skip[i].nr =
1627 				    rule->skip[i].ptr->nr;
1628 
1629 		if (pr->action == PF_GET_CLR_CNTR) {
1630 			rule->evaluations = 0;
1631 			rule->packets[0] = rule->packets[1] = 0;
1632 			rule->bytes[0] = rule->bytes[1] = 0;
1633 			counter_u64_zero(rule->states_tot);
1634 		}
1635 		PF_RULES_WUNLOCK();
1636 		break;
1637 	}
1638 
1639 	case DIOCCHANGERULE: {
1640 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1641 		struct pf_ruleset	*ruleset;
1642 		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1643 		struct pfi_kif		*kif = NULL;
1644 		struct pf_pooladdr	*pa;
1645 		u_int32_t		 nr = 0;
1646 		int			 rs_num;
1647 
1648 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1649 		    pcr->action > PF_CHANGE_GET_TICKET) {
1650 			error = EINVAL;
1651 			break;
1652 		}
1653 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1654 			error = EINVAL;
1655 			break;
1656 		}
1657 
1658 		if (pcr->action != PF_CHANGE_REMOVE) {
1659 #ifndef INET
1660 			if (pcr->rule.af == AF_INET) {
1661 				error = EAFNOSUPPORT;
1662 				break;
1663 			}
1664 #endif /* INET */
1665 #ifndef INET6
1666 			if (pcr->rule.af == AF_INET6) {
1667 				error = EAFNOSUPPORT;
1668 				break;
1669 			}
1670 #endif /* INET6 */
1671 			newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
1672 			bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1673 			if (newrule->ifname[0])
1674 				kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1675 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
1676 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
1677 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
1678 			newrule->cuid = td->td_ucred->cr_ruid;
1679 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1680 			TAILQ_INIT(&newrule->rpool.list);
1681 		}
1682 
1683 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGERULE_error; }
1684 
1685 		PF_RULES_WLOCK();
1686 		if (!(pcr->action == PF_CHANGE_REMOVE ||
1687 		    pcr->action == PF_CHANGE_GET_TICKET) &&
1688 		    pcr->pool_ticket != V_ticket_pabuf)
1689 			ERROUT(EBUSY);
1690 
1691 		ruleset = pf_find_ruleset(pcr->anchor);
1692 		if (ruleset == NULL)
1693 			ERROUT(EINVAL);
1694 
1695 		rs_num = pf_get_ruleset_number(pcr->rule.action);
1696 		if (rs_num >= PF_RULESET_MAX)
1697 			ERROUT(EINVAL);
1698 
1699 		if (pcr->action == PF_CHANGE_GET_TICKET) {
1700 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1701 			ERROUT(0);
1702 		} else if (pcr->ticket !=
1703 			    ruleset->rules[rs_num].active.ticket)
1704 				ERROUT(EINVAL);
1705 
1706 		if (pcr->action != PF_CHANGE_REMOVE) {
1707 			if (newrule->ifname[0]) {
1708 				newrule->kif = pfi_kif_attach(kif,
1709 				    newrule->ifname);
1710 				pfi_kif_ref(newrule->kif);
1711 			} else
1712 				newrule->kif = NULL;
1713 
1714 			if (newrule->rtableid > 0 &&
1715 			    newrule->rtableid >= rt_numfibs)
1716 				error = EBUSY;
1717 
1718 #ifdef ALTQ
1719 			/* set queue IDs */
1720 			if (newrule->qname[0] != 0) {
1721 				if ((newrule->qid =
1722 				    pf_qname2qid(newrule->qname)) == 0)
1723 					error = EBUSY;
1724 				else if (newrule->pqname[0] != 0) {
1725 					if ((newrule->pqid =
1726 					    pf_qname2qid(newrule->pqname)) == 0)
1727 						error = EBUSY;
1728 				} else
1729 					newrule->pqid = newrule->qid;
1730 			}
1731 #endif /* ALTQ */
1732 			if (newrule->tagname[0])
1733 				if ((newrule->tag =
1734 				    pf_tagname2tag(newrule->tagname)) == 0)
1735 					error = EBUSY;
1736 			if (newrule->match_tagname[0])
1737 				if ((newrule->match_tag = pf_tagname2tag(
1738 				    newrule->match_tagname)) == 0)
1739 					error = EBUSY;
1740 			if (newrule->rt && !newrule->direction)
1741 				error = EINVAL;
1742 			if (!newrule->log)
1743 				newrule->logif = 0;
1744 			if (newrule->logif >= PFLOGIFS_MAX)
1745 				error = EINVAL;
1746 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
1747 				error = ENOMEM;
1748 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
1749 				error = ENOMEM;
1750 			if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1751 				error = EINVAL;
1752 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1753 				if (pa->addr.type == PF_ADDR_TABLE) {
1754 					pa->addr.p.tbl =
1755 					    pfr_attach_table(ruleset,
1756 					    pa->addr.v.tblname);
1757 					if (pa->addr.p.tbl == NULL)
1758 						error = ENOMEM;
1759 				}
1760 
1761 			newrule->overload_tbl = NULL;
1762 			if (newrule->overload_tblname[0]) {
1763 				if ((newrule->overload_tbl = pfr_attach_table(
1764 				    ruleset, newrule->overload_tblname)) ==
1765 				    NULL)
1766 					error = EINVAL;
1767 				else
1768 					newrule->overload_tbl->pfrkt_flags |=
1769 					    PFR_TFLAG_ACTIVE;
1770 			}
1771 
1772 			pf_mv_pool(&V_pf_pabuf, &newrule->rpool.list);
1773 			if (((((newrule->action == PF_NAT) ||
1774 			    (newrule->action == PF_RDR) ||
1775 			    (newrule->action == PF_BINAT) ||
1776 			    (newrule->rt > PF_NOPFROUTE)) &&
1777 			    !newrule->anchor)) &&
1778 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1779 				error = EINVAL;
1780 
1781 			if (error) {
1782 				pf_free_rule(newrule);
1783 				PF_RULES_WUNLOCK();
1784 				break;
1785 			}
1786 
1787 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1788 			newrule->evaluations = 0;
1789 			newrule->packets[0] = newrule->packets[1] = 0;
1790 			newrule->bytes[0] = newrule->bytes[1] = 0;
1791 		}
1792 		pf_empty_pool(&V_pf_pabuf);
1793 
1794 		if (pcr->action == PF_CHANGE_ADD_HEAD)
1795 			oldrule = TAILQ_FIRST(
1796 			    ruleset->rules[rs_num].active.ptr);
1797 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1798 			oldrule = TAILQ_LAST(
1799 			    ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1800 		else {
1801 			oldrule = TAILQ_FIRST(
1802 			    ruleset->rules[rs_num].active.ptr);
1803 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1804 				oldrule = TAILQ_NEXT(oldrule, entries);
1805 			if (oldrule == NULL) {
1806 				if (newrule != NULL)
1807 					pf_free_rule(newrule);
1808 				PF_RULES_WUNLOCK();
1809 				error = EINVAL;
1810 				break;
1811 			}
1812 		}
1813 
1814 		if (pcr->action == PF_CHANGE_REMOVE) {
1815 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
1816 			    oldrule);
1817 			ruleset->rules[rs_num].active.rcount--;
1818 		} else {
1819 			if (oldrule == NULL)
1820 				TAILQ_INSERT_TAIL(
1821 				    ruleset->rules[rs_num].active.ptr,
1822 				    newrule, entries);
1823 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1824 			    pcr->action == PF_CHANGE_ADD_BEFORE)
1825 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1826 			else
1827 				TAILQ_INSERT_AFTER(
1828 				    ruleset->rules[rs_num].active.ptr,
1829 				    oldrule, newrule, entries);
1830 			ruleset->rules[rs_num].active.rcount++;
1831 		}
1832 
1833 		nr = 0;
1834 		TAILQ_FOREACH(oldrule,
1835 		    ruleset->rules[rs_num].active.ptr, entries)
1836 			oldrule->nr = nr++;
1837 
1838 		ruleset->rules[rs_num].active.ticket++;
1839 
1840 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1841 		pf_remove_if_empty_ruleset(ruleset);
1842 
1843 		PF_RULES_WUNLOCK();
1844 		break;
1845 
1846 #undef ERROUT
1847 DIOCCHANGERULE_error:
1848 		PF_RULES_WUNLOCK();
1849 		if (newrule != NULL) {
1850 			counter_u64_free(newrule->states_cur);
1851 			counter_u64_free(newrule->states_tot);
1852 			counter_u64_free(newrule->src_nodes);
1853 			free(newrule, M_PFRULE);
1854 		}
1855 		if (kif != NULL)
1856 			free(kif, PFI_MTYPE);
1857 		break;
1858 	}
1859 
1860 	case DIOCCLRSTATES: {
1861 		struct pf_state		*s;
1862 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1863 		u_int			 i, killed = 0;
1864 
1865 		for (i = 0; i <= pf_hashmask; i++) {
1866 			struct pf_idhash *ih = &V_pf_idhash[i];
1867 
1868 relock_DIOCCLRSTATES:
1869 			PF_HASHROW_LOCK(ih);
1870 			LIST_FOREACH(s, &ih->states, entry)
1871 				if (!psk->psk_ifname[0] ||
1872 				    !strcmp(psk->psk_ifname,
1873 				    s->kif->pfik_name)) {
1874 					/*
1875 					 * Don't send out individual
1876 					 * delete messages.
1877 					 */
1878 					s->state_flags |= PFSTATE_NOSYNC;
1879 					pf_unlink_state(s, PF_ENTER_LOCKED);
1880 					killed++;
1881 					goto relock_DIOCCLRSTATES;
1882 				}
1883 			PF_HASHROW_UNLOCK(ih);
1884 		}
1885 		psk->psk_killed = killed;
1886 		if (V_pfsync_clear_states_ptr != NULL)
1887 			V_pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
1888 		break;
1889 	}
1890 
1891 	case DIOCKILLSTATES: {
1892 		struct pf_state		*s;
1893 		struct pf_state_key	*sk;
1894 		struct pf_addr		*srcaddr, *dstaddr;
1895 		u_int16_t		 srcport, dstport;
1896 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
1897 		u_int			 i, killed = 0;
1898 
1899 		if (psk->psk_pfcmp.id) {
1900 			if (psk->psk_pfcmp.creatorid == 0)
1901 				psk->psk_pfcmp.creatorid = V_pf_status.hostid;
1902 			if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
1903 			    psk->psk_pfcmp.creatorid))) {
1904 				pf_unlink_state(s, PF_ENTER_LOCKED);
1905 				psk->psk_killed = 1;
1906 			}
1907 			break;
1908 		}
1909 
1910 		for (i = 0; i <= pf_hashmask; i++) {
1911 			struct pf_idhash *ih = &V_pf_idhash[i];
1912 
1913 relock_DIOCKILLSTATES:
1914 			PF_HASHROW_LOCK(ih);
1915 			LIST_FOREACH(s, &ih->states, entry) {
1916 				sk = s->key[PF_SK_WIRE];
1917 				if (s->direction == PF_OUT) {
1918 					srcaddr = &sk->addr[1];
1919 					dstaddr = &sk->addr[0];
1920 					srcport = sk->port[1];
1921 					dstport = sk->port[0];
1922 				} else {
1923 					srcaddr = &sk->addr[0];
1924 					dstaddr = &sk->addr[1];
1925 					srcport = sk->port[0];
1926 					dstport = sk->port[1];
1927 				}
1928 
1929 				if ((!psk->psk_af || sk->af == psk->psk_af)
1930 				    && (!psk->psk_proto || psk->psk_proto ==
1931 				    sk->proto) &&
1932 				    PF_MATCHA(psk->psk_src.neg,
1933 				    &psk->psk_src.addr.v.a.addr,
1934 				    &psk->psk_src.addr.v.a.mask,
1935 				    srcaddr, sk->af) &&
1936 				    PF_MATCHA(psk->psk_dst.neg,
1937 				    &psk->psk_dst.addr.v.a.addr,
1938 				    &psk->psk_dst.addr.v.a.mask,
1939 				    dstaddr, sk->af) &&
1940 				    (psk->psk_src.port_op == 0 ||
1941 				    pf_match_port(psk->psk_src.port_op,
1942 				    psk->psk_src.port[0], psk->psk_src.port[1],
1943 				    srcport)) &&
1944 				    (psk->psk_dst.port_op == 0 ||
1945 				    pf_match_port(psk->psk_dst.port_op,
1946 				    psk->psk_dst.port[0], psk->psk_dst.port[1],
1947 				    dstport)) &&
1948 				    (!psk->psk_label[0] ||
1949 				    (s->rule.ptr->label[0] &&
1950 				    !strcmp(psk->psk_label,
1951 				    s->rule.ptr->label))) &&
1952 				    (!psk->psk_ifname[0] ||
1953 				    !strcmp(psk->psk_ifname,
1954 				    s->kif->pfik_name))) {
1955 					pf_unlink_state(s, PF_ENTER_LOCKED);
1956 					killed++;
1957 					goto relock_DIOCKILLSTATES;
1958 				}
1959 			}
1960 			PF_HASHROW_UNLOCK(ih);
1961 		}
1962 		psk->psk_killed = killed;
1963 		break;
1964 	}
1965 
1966 	case DIOCADDSTATE: {
1967 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1968 		struct pfsync_state	*sp = &ps->state;
1969 
1970 		if (sp->timeout >= PFTM_MAX) {
1971 			error = EINVAL;
1972 			break;
1973 		}
1974 		if (V_pfsync_state_import_ptr != NULL) {
1975 			PF_RULES_RLOCK();
1976 			error = V_pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
1977 			PF_RULES_RUNLOCK();
1978 		} else
1979 			error = EOPNOTSUPP;
1980 		break;
1981 	}
1982 
1983 	case DIOCGETSTATE: {
1984 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1985 		struct pf_state		*s;
1986 
1987 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
1988 		if (s == NULL) {
1989 			error = ENOENT;
1990 			break;
1991 		}
1992 
1993 		pfsync_state_export(&ps->state, s);
1994 		PF_STATE_UNLOCK(s);
1995 		break;
1996 	}
1997 
1998 	case DIOCGETSTATES: {
1999 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
2000 		struct pf_state		*s;
2001 		struct pfsync_state	*pstore, *p;
2002 		int i, nr;
2003 
2004 		if (ps->ps_len == 0) {
2005 			nr = uma_zone_get_cur(V_pf_state_z);
2006 			ps->ps_len = sizeof(struct pfsync_state) * nr;
2007 			break;
2008 		}
2009 
2010 		p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK);
2011 		nr = 0;
2012 
2013 		for (i = 0; i <= pf_hashmask; i++) {
2014 			struct pf_idhash *ih = &V_pf_idhash[i];
2015 
2016 			PF_HASHROW_LOCK(ih);
2017 			LIST_FOREACH(s, &ih->states, entry) {
2018 
2019 				if (s->timeout == PFTM_UNLINKED)
2020 					continue;
2021 
2022 				if ((nr+1) * sizeof(*p) > ps->ps_len) {
2023 					PF_HASHROW_UNLOCK(ih);
2024 					goto DIOCGETSTATES_full;
2025 				}
2026 				pfsync_state_export(p, s);
2027 				p++;
2028 				nr++;
2029 			}
2030 			PF_HASHROW_UNLOCK(ih);
2031 		}
2032 DIOCGETSTATES_full:
2033 		error = copyout(pstore, ps->ps_states,
2034 		    sizeof(struct pfsync_state) * nr);
2035 		if (error) {
2036 			free(pstore, M_TEMP);
2037 			break;
2038 		}
2039 		ps->ps_len = sizeof(struct pfsync_state) * nr;
2040 		free(pstore, M_TEMP);
2041 
2042 		break;
2043 	}
2044 
2045 	case DIOCGETSTATUS: {
2046 		struct pf_status *s = (struct pf_status *)addr;
2047 
2048 		PF_RULES_RLOCK();
2049 		s->running = V_pf_status.running;
2050 		s->since   = V_pf_status.since;
2051 		s->debug   = V_pf_status.debug;
2052 		s->hostid  = V_pf_status.hostid;
2053 		s->states  = V_pf_status.states;
2054 		s->src_nodes = V_pf_status.src_nodes;
2055 
2056 		for (int i = 0; i < PFRES_MAX; i++)
2057 			s->counters[i] =
2058 			    counter_u64_fetch(V_pf_status.counters[i]);
2059 		for (int i = 0; i < LCNT_MAX; i++)
2060 			s->lcounters[i] =
2061 			    counter_u64_fetch(V_pf_status.lcounters[i]);
2062 		for (int i = 0; i < FCNT_MAX; i++)
2063 			s->fcounters[i] =
2064 			    counter_u64_fetch(V_pf_status.fcounters[i]);
2065 		for (int i = 0; i < SCNT_MAX; i++)
2066 			s->scounters[i] =
2067 			    counter_u64_fetch(V_pf_status.scounters[i]);
2068 
2069 		bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ);
2070 		bcopy(V_pf_status.pf_chksum, s->pf_chksum,
2071 		    PF_MD5_DIGEST_LENGTH);
2072 
2073 		pfi_update_status(s->ifname, s);
2074 		PF_RULES_RUNLOCK();
2075 		break;
2076 	}
2077 
2078 	case DIOCSETSTATUSIF: {
2079 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
2080 
2081 		if (pi->ifname[0] == 0) {
2082 			bzero(V_pf_status.ifname, IFNAMSIZ);
2083 			break;
2084 		}
2085 		PF_RULES_WLOCK();
2086 		strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
2087 		PF_RULES_WUNLOCK();
2088 		break;
2089 	}
2090 
2091 	case DIOCCLRSTATUS: {
2092 		PF_RULES_WLOCK();
2093 		for (int i = 0; i < PFRES_MAX; i++)
2094 			counter_u64_zero(V_pf_status.counters[i]);
2095 		for (int i = 0; i < FCNT_MAX; i++)
2096 			counter_u64_zero(V_pf_status.fcounters[i]);
2097 		for (int i = 0; i < SCNT_MAX; i++)
2098 			counter_u64_zero(V_pf_status.scounters[i]);
2099 		for (int i = 0; i < LCNT_MAX; i++)
2100 			counter_u64_zero(V_pf_status.lcounters[i]);
2101 		V_pf_status.since = time_second;
2102 		if (*V_pf_status.ifname)
2103 			pfi_update_status(V_pf_status.ifname, NULL);
2104 		PF_RULES_WUNLOCK();
2105 		break;
2106 	}
2107 
2108 	case DIOCNATLOOK: {
2109 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
2110 		struct pf_state_key	*sk;
2111 		struct pf_state		*state;
2112 		struct pf_state_key_cmp	 key;
2113 		int			 m = 0, direction = pnl->direction;
2114 		int			 sidx, didx;
2115 
2116 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
2117 		sidx = (direction == PF_IN) ? 1 : 0;
2118 		didx = (direction == PF_IN) ? 0 : 1;
2119 
2120 		if (!pnl->proto ||
2121 		    PF_AZERO(&pnl->saddr, pnl->af) ||
2122 		    PF_AZERO(&pnl->daddr, pnl->af) ||
2123 		    ((pnl->proto == IPPROTO_TCP ||
2124 		    pnl->proto == IPPROTO_UDP) &&
2125 		    (!pnl->dport || !pnl->sport)))
2126 			error = EINVAL;
2127 		else {
2128 			bzero(&key, sizeof(key));
2129 			key.af = pnl->af;
2130 			key.proto = pnl->proto;
2131 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
2132 			key.port[sidx] = pnl->sport;
2133 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
2134 			key.port[didx] = pnl->dport;
2135 
2136 			state = pf_find_state_all(&key, direction, &m);
2137 
2138 			if (m > 1)
2139 				error = E2BIG;	/* more than one state */
2140 			else if (state != NULL) {
2141 				/* XXXGL: not locked read */
2142 				sk = state->key[sidx];
2143 				PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
2144 				pnl->rsport = sk->port[sidx];
2145 				PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
2146 				pnl->rdport = sk->port[didx];
2147 			} else
2148 				error = ENOENT;
2149 		}
2150 		break;
2151 	}
2152 
2153 	case DIOCSETTIMEOUT: {
2154 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2155 		int		 old;
2156 
2157 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
2158 		    pt->seconds < 0) {
2159 			error = EINVAL;
2160 			break;
2161 		}
2162 		PF_RULES_WLOCK();
2163 		old = V_pf_default_rule.timeout[pt->timeout];
2164 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
2165 			pt->seconds = 1;
2166 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
2167 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
2168 			wakeup(pf_purge_thread);
2169 		pt->seconds = old;
2170 		PF_RULES_WUNLOCK();
2171 		break;
2172 	}
2173 
2174 	case DIOCGETTIMEOUT: {
2175 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2176 
2177 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
2178 			error = EINVAL;
2179 			break;
2180 		}
2181 		PF_RULES_RLOCK();
2182 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
2183 		PF_RULES_RUNLOCK();
2184 		break;
2185 	}
2186 
2187 	case DIOCGETLIMIT: {
2188 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2189 
2190 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
2191 			error = EINVAL;
2192 			break;
2193 		}
2194 		PF_RULES_RLOCK();
2195 		pl->limit = V_pf_limits[pl->index].limit;
2196 		PF_RULES_RUNLOCK();
2197 		break;
2198 	}
2199 
2200 	case DIOCSETLIMIT: {
2201 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2202 		int			 old_limit;
2203 
2204 		PF_RULES_WLOCK();
2205 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
2206 		    V_pf_limits[pl->index].zone == NULL) {
2207 			PF_RULES_WUNLOCK();
2208 			error = EINVAL;
2209 			break;
2210 		}
2211 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
2212 		old_limit = V_pf_limits[pl->index].limit;
2213 		V_pf_limits[pl->index].limit = pl->limit;
2214 		pl->limit = old_limit;
2215 		PF_RULES_WUNLOCK();
2216 		break;
2217 	}
2218 
2219 	case DIOCSETDEBUG: {
2220 		u_int32_t	*level = (u_int32_t *)addr;
2221 
2222 		PF_RULES_WLOCK();
2223 		V_pf_status.debug = *level;
2224 		PF_RULES_WUNLOCK();
2225 		break;
2226 	}
2227 
2228 	case DIOCCLRRULECTRS: {
2229 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
2230 		struct pf_ruleset	*ruleset = &pf_main_ruleset;
2231 		struct pf_rule		*rule;
2232 
2233 		PF_RULES_WLOCK();
2234 		TAILQ_FOREACH(rule,
2235 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
2236 			rule->evaluations = 0;
2237 			rule->packets[0] = rule->packets[1] = 0;
2238 			rule->bytes[0] = rule->bytes[1] = 0;
2239 		}
2240 		PF_RULES_WUNLOCK();
2241 		break;
2242 	}
2243 
2244 	case DIOCGIFSPEEDV0:
2245 	case DIOCGIFSPEEDV1: {
2246 		struct pf_ifspeed_v1	*psp = (struct pf_ifspeed_v1 *)addr;
2247 		struct pf_ifspeed_v1	ps;
2248 		struct ifnet		*ifp;
2249 
2250 		if (psp->ifname[0] != 0) {
2251 			/* Can we completely trust user-land? */
2252 			strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
2253 			ifp = ifunit(ps.ifname);
2254 			if (ifp != NULL) {
2255 				psp->baudrate32 =
2256 				    (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
2257 				if (cmd == DIOCGIFSPEEDV1)
2258 					psp->baudrate = ifp->if_baudrate;
2259 			} else
2260 				error = EINVAL;
2261 		} else
2262 			error = EINVAL;
2263 		break;
2264 	}
2265 
2266 #ifdef ALTQ
2267 	case DIOCSTARTALTQ: {
2268 		struct pf_altq		*altq;
2269 
2270 		PF_RULES_WLOCK();
2271 		/* enable all altq interfaces on active list */
2272 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
2273 			if (altq->qname[0] == 0 && (altq->local_flags &
2274 			    PFALTQ_FLAG_IF_REMOVED) == 0) {
2275 				error = pf_enable_altq(altq);
2276 				if (error != 0)
2277 					break;
2278 			}
2279 		}
2280 		if (error == 0)
2281 			V_pf_altq_running = 1;
2282 		PF_RULES_WUNLOCK();
2283 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
2284 		break;
2285 	}
2286 
2287 	case DIOCSTOPALTQ: {
2288 		struct pf_altq		*altq;
2289 
2290 		PF_RULES_WLOCK();
2291 		/* disable all altq interfaces on active list */
2292 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
2293 			if (altq->qname[0] == 0 && (altq->local_flags &
2294 			    PFALTQ_FLAG_IF_REMOVED) == 0) {
2295 				error = pf_disable_altq(altq);
2296 				if (error != 0)
2297 					break;
2298 			}
2299 		}
2300 		if (error == 0)
2301 			V_pf_altq_running = 0;
2302 		PF_RULES_WUNLOCK();
2303 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2304 		break;
2305 	}
2306 
2307 	case DIOCADDALTQV0:
2308 	case DIOCADDALTQV1: {
2309 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2310 		struct pf_altq		*altq, *a;
2311 		struct ifnet		*ifp;
2312 
2313 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
2314 		error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
2315 		if (error)
2316 			break;
2317 		altq->local_flags = 0;
2318 
2319 		PF_RULES_WLOCK();
2320 		if (pa->ticket != V_ticket_altqs_inactive) {
2321 			PF_RULES_WUNLOCK();
2322 			free(altq, M_PFALTQ);
2323 			error = EBUSY;
2324 			break;
2325 		}
2326 
2327 		/*
2328 		 * if this is for a queue, find the discipline and
2329 		 * copy the necessary fields
2330 		 */
2331 		if (altq->qname[0] != 0) {
2332 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2333 				PF_RULES_WUNLOCK();
2334 				error = EBUSY;
2335 				free(altq, M_PFALTQ);
2336 				break;
2337 			}
2338 			altq->altq_disc = NULL;
2339 			TAILQ_FOREACH(a, V_pf_altqs_inactive, entries) {
2340 				if (strncmp(a->ifname, altq->ifname,
2341 				    IFNAMSIZ) == 0 && a->qname[0] == 0) {
2342 					altq->altq_disc = a->altq_disc;
2343 					break;
2344 				}
2345 			}
2346 		}
2347 
2348 		if ((ifp = ifunit(altq->ifname)) == NULL)
2349 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
2350 		else
2351 			error = altq_add(altq);
2352 
2353 		if (error) {
2354 			PF_RULES_WUNLOCK();
2355 			free(altq, M_PFALTQ);
2356 			break;
2357 		}
2358 
2359 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
2360 		/* version error check done on import above */
2361 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
2362 		PF_RULES_WUNLOCK();
2363 		break;
2364 	}
2365 
2366 	case DIOCGETALTQSV0:
2367 	case DIOCGETALTQSV1: {
2368 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2369 		struct pf_altq		*altq;
2370 
2371 		PF_RULES_RLOCK();
2372 		pa->nr = 0;
2373 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
2374 			pa->nr++;
2375 		pa->ticket = V_ticket_altqs_active;
2376 		PF_RULES_RUNLOCK();
2377 		break;
2378 	}
2379 
2380 	case DIOCGETALTQV0:
2381 	case DIOCGETALTQV1: {
2382 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2383 		struct pf_altq		*altq;
2384 		u_int32_t		 nr;
2385 
2386 		PF_RULES_RLOCK();
2387 		if (pa->ticket != V_ticket_altqs_active) {
2388 			PF_RULES_RUNLOCK();
2389 			error = EBUSY;
2390 			break;
2391 		}
2392 		nr = 0;
2393 		altq = TAILQ_FIRST(V_pf_altqs_active);
2394 		while ((altq != NULL) && (nr < pa->nr)) {
2395 			altq = TAILQ_NEXT(altq, entries);
2396 			nr++;
2397 		}
2398 		if (altq == NULL) {
2399 			PF_RULES_RUNLOCK();
2400 			error = EBUSY;
2401 			break;
2402 		}
2403 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
2404 		PF_RULES_RUNLOCK();
2405 		break;
2406 	}
2407 
2408 	case DIOCCHANGEALTQV0:
2409 	case DIOCCHANGEALTQV1:
2410 		/* CHANGEALTQ not supported yet! */
2411 		error = ENODEV;
2412 		break;
2413 
2414 	case DIOCGETQSTATSV0:
2415 	case DIOCGETQSTATSV1: {
2416 		struct pfioc_qstats_v1	*pq = (struct pfioc_qstats_v1 *)addr;
2417 		struct pf_altq		*altq;
2418 		u_int32_t		 nr;
2419 		int			 nbytes;
2420 		u_int32_t		 version;
2421 
2422 		PF_RULES_RLOCK();
2423 		if (pq->ticket != V_ticket_altqs_active) {
2424 			PF_RULES_RUNLOCK();
2425 			error = EBUSY;
2426 			break;
2427 		}
2428 		nbytes = pq->nbytes;
2429 		nr = 0;
2430 		altq = TAILQ_FIRST(V_pf_altqs_active);
2431 		while ((altq != NULL) && (nr < pq->nr)) {
2432 			altq = TAILQ_NEXT(altq, entries);
2433 			nr++;
2434 		}
2435 		if (altq == NULL) {
2436 			PF_RULES_RUNLOCK();
2437 			error = EBUSY;
2438 			break;
2439 		}
2440 
2441 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
2442 			PF_RULES_RUNLOCK();
2443 			error = ENXIO;
2444 			break;
2445 		}
2446 		PF_RULES_RUNLOCK();
2447 		if (cmd == DIOCGETQSTATSV0)
2448 			version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
2449 		else
2450 			version = pq->version;
2451 		error = altq_getqstats(altq, pq->buf, &nbytes, version);
2452 		if (error == 0) {
2453 			pq->scheduler = altq->scheduler;
2454 			pq->nbytes = nbytes;
2455 		}
2456 		break;
2457 	}
2458 #endif /* ALTQ */
2459 
2460 	case DIOCBEGINADDRS: {
2461 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2462 
2463 		PF_RULES_WLOCK();
2464 		pf_empty_pool(&V_pf_pabuf);
2465 		pp->ticket = ++V_ticket_pabuf;
2466 		PF_RULES_WUNLOCK();
2467 		break;
2468 	}
2469 
2470 	case DIOCADDADDR: {
2471 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2472 		struct pf_pooladdr	*pa;
2473 		struct pfi_kif		*kif = NULL;
2474 
2475 #ifndef INET
2476 		if (pp->af == AF_INET) {
2477 			error = EAFNOSUPPORT;
2478 			break;
2479 		}
2480 #endif /* INET */
2481 #ifndef INET6
2482 		if (pp->af == AF_INET6) {
2483 			error = EAFNOSUPPORT;
2484 			break;
2485 		}
2486 #endif /* INET6 */
2487 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2488 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2489 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2490 			error = EINVAL;
2491 			break;
2492 		}
2493 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2494 		bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2495 		if (pa->ifname[0])
2496 			kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2497 		PF_RULES_WLOCK();
2498 		if (pp->ticket != V_ticket_pabuf) {
2499 			PF_RULES_WUNLOCK();
2500 			if (pa->ifname[0])
2501 				free(kif, PFI_MTYPE);
2502 			free(pa, M_PFRULE);
2503 			error = EBUSY;
2504 			break;
2505 		}
2506 		if (pa->ifname[0]) {
2507 			pa->kif = pfi_kif_attach(kif, pa->ifname);
2508 			pfi_kif_ref(pa->kif);
2509 		} else
2510 			pa->kif = NULL;
2511 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2512 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2513 			if (pa->ifname[0])
2514 				pfi_kif_unref(pa->kif);
2515 			PF_RULES_WUNLOCK();
2516 			free(pa, M_PFRULE);
2517 			break;
2518 		}
2519 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2520 		PF_RULES_WUNLOCK();
2521 		break;
2522 	}
2523 
2524 	case DIOCGETADDRS: {
2525 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2526 		struct pf_pool		*pool;
2527 		struct pf_pooladdr	*pa;
2528 
2529 		PF_RULES_RLOCK();
2530 		pp->nr = 0;
2531 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2532 		    pp->r_num, 0, 1, 0);
2533 		if (pool == NULL) {
2534 			PF_RULES_RUNLOCK();
2535 			error = EBUSY;
2536 			break;
2537 		}
2538 		TAILQ_FOREACH(pa, &pool->list, entries)
2539 			pp->nr++;
2540 		PF_RULES_RUNLOCK();
2541 		break;
2542 	}
2543 
2544 	case DIOCGETADDR: {
2545 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2546 		struct pf_pool		*pool;
2547 		struct pf_pooladdr	*pa;
2548 		u_int32_t		 nr = 0;
2549 
2550 		PF_RULES_RLOCK();
2551 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2552 		    pp->r_num, 0, 1, 1);
2553 		if (pool == NULL) {
2554 			PF_RULES_RUNLOCK();
2555 			error = EBUSY;
2556 			break;
2557 		}
2558 		pa = TAILQ_FIRST(&pool->list);
2559 		while ((pa != NULL) && (nr < pp->nr)) {
2560 			pa = TAILQ_NEXT(pa, entries);
2561 			nr++;
2562 		}
2563 		if (pa == NULL) {
2564 			PF_RULES_RUNLOCK();
2565 			error = EBUSY;
2566 			break;
2567 		}
2568 		bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2569 		pf_addr_copyout(&pp->addr.addr);
2570 		PF_RULES_RUNLOCK();
2571 		break;
2572 	}
2573 
2574 	case DIOCCHANGEADDR: {
2575 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
2576 		struct pf_pool		*pool;
2577 		struct pf_pooladdr	*oldpa = NULL, *newpa = NULL;
2578 		struct pf_ruleset	*ruleset;
2579 		struct pfi_kif		*kif = NULL;
2580 
2581 		if (pca->action < PF_CHANGE_ADD_HEAD ||
2582 		    pca->action > PF_CHANGE_REMOVE) {
2583 			error = EINVAL;
2584 			break;
2585 		}
2586 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2587 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2588 		    pca->addr.addr.type != PF_ADDR_TABLE) {
2589 			error = EINVAL;
2590 			break;
2591 		}
2592 
2593 		if (pca->action != PF_CHANGE_REMOVE) {
2594 #ifndef INET
2595 			if (pca->af == AF_INET) {
2596 				error = EAFNOSUPPORT;
2597 				break;
2598 			}
2599 #endif /* INET */
2600 #ifndef INET6
2601 			if (pca->af == AF_INET6) {
2602 				error = EAFNOSUPPORT;
2603 				break;
2604 			}
2605 #endif /* INET6 */
2606 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
2607 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2608 			if (newpa->ifname[0])
2609 				kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2610 			newpa->kif = NULL;
2611 		}
2612 
2613 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGEADDR_error; }
2614 		PF_RULES_WLOCK();
2615 		ruleset = pf_find_ruleset(pca->anchor);
2616 		if (ruleset == NULL)
2617 			ERROUT(EBUSY);
2618 
2619 		pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2620 		    pca->r_num, pca->r_last, 1, 1);
2621 		if (pool == NULL)
2622 			ERROUT(EBUSY);
2623 
2624 		if (pca->action != PF_CHANGE_REMOVE) {
2625 			if (newpa->ifname[0]) {
2626 				newpa->kif = pfi_kif_attach(kif, newpa->ifname);
2627 				pfi_kif_ref(newpa->kif);
2628 				kif = NULL;
2629 			}
2630 
2631 			switch (newpa->addr.type) {
2632 			case PF_ADDR_DYNIFTL:
2633 				error = pfi_dynaddr_setup(&newpa->addr,
2634 				    pca->af);
2635 				break;
2636 			case PF_ADDR_TABLE:
2637 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
2638 				    newpa->addr.v.tblname);
2639 				if (newpa->addr.p.tbl == NULL)
2640 					error = ENOMEM;
2641 				break;
2642 			}
2643 			if (error)
2644 				goto DIOCCHANGEADDR_error;
2645 		}
2646 
2647 		switch (pca->action) {
2648 		case PF_CHANGE_ADD_HEAD:
2649 			oldpa = TAILQ_FIRST(&pool->list);
2650 			break;
2651 		case PF_CHANGE_ADD_TAIL:
2652 			oldpa = TAILQ_LAST(&pool->list, pf_palist);
2653 			break;
2654 		default:
2655 			oldpa = TAILQ_FIRST(&pool->list);
2656 			for (int i = 0; oldpa && i < pca->nr; i++)
2657 				oldpa = TAILQ_NEXT(oldpa, entries);
2658 
2659 			if (oldpa == NULL)
2660 				ERROUT(EINVAL);
2661 		}
2662 
2663 		if (pca->action == PF_CHANGE_REMOVE) {
2664 			TAILQ_REMOVE(&pool->list, oldpa, entries);
2665 			switch (oldpa->addr.type) {
2666 			case PF_ADDR_DYNIFTL:
2667 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
2668 				break;
2669 			case PF_ADDR_TABLE:
2670 				pfr_detach_table(oldpa->addr.p.tbl);
2671 				break;
2672 			}
2673 			if (oldpa->kif)
2674 				pfi_kif_unref(oldpa->kif);
2675 			free(oldpa, M_PFRULE);
2676 		} else {
2677 			if (oldpa == NULL)
2678 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2679 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
2680 			    pca->action == PF_CHANGE_ADD_BEFORE)
2681 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2682 			else
2683 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
2684 				    newpa, entries);
2685 		}
2686 
2687 		pool->cur = TAILQ_FIRST(&pool->list);
2688 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
2689 		PF_RULES_WUNLOCK();
2690 		break;
2691 
2692 #undef ERROUT
2693 DIOCCHANGEADDR_error:
2694 		if (newpa != NULL) {
2695 			if (newpa->kif)
2696 				pfi_kif_unref(newpa->kif);
2697 			free(newpa, M_PFRULE);
2698 		}
2699 		PF_RULES_WUNLOCK();
2700 		if (kif != NULL)
2701 			free(kif, PFI_MTYPE);
2702 		break;
2703 	}
2704 
2705 	case DIOCGETRULESETS: {
2706 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2707 		struct pf_ruleset	*ruleset;
2708 		struct pf_anchor	*anchor;
2709 
2710 		PF_RULES_RLOCK();
2711 		pr->path[sizeof(pr->path) - 1] = 0;
2712 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2713 			PF_RULES_RUNLOCK();
2714 			error = ENOENT;
2715 			break;
2716 		}
2717 		pr->nr = 0;
2718 		if (ruleset->anchor == NULL) {
2719 			/* XXX kludge for pf_main_ruleset */
2720 			RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2721 				if (anchor->parent == NULL)
2722 					pr->nr++;
2723 		} else {
2724 			RB_FOREACH(anchor, pf_anchor_node,
2725 			    &ruleset->anchor->children)
2726 				pr->nr++;
2727 		}
2728 		PF_RULES_RUNLOCK();
2729 		break;
2730 	}
2731 
2732 	case DIOCGETRULESET: {
2733 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2734 		struct pf_ruleset	*ruleset;
2735 		struct pf_anchor	*anchor;
2736 		u_int32_t		 nr = 0;
2737 
2738 		PF_RULES_RLOCK();
2739 		pr->path[sizeof(pr->path) - 1] = 0;
2740 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2741 			PF_RULES_RUNLOCK();
2742 			error = ENOENT;
2743 			break;
2744 		}
2745 		pr->name[0] = 0;
2746 		if (ruleset->anchor == NULL) {
2747 			/* XXX kludge for pf_main_ruleset */
2748 			RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2749 				if (anchor->parent == NULL && nr++ == pr->nr) {
2750 					strlcpy(pr->name, anchor->name,
2751 					    sizeof(pr->name));
2752 					break;
2753 				}
2754 		} else {
2755 			RB_FOREACH(anchor, pf_anchor_node,
2756 			    &ruleset->anchor->children)
2757 				if (nr++ == pr->nr) {
2758 					strlcpy(pr->name, anchor->name,
2759 					    sizeof(pr->name));
2760 					break;
2761 				}
2762 		}
2763 		if (!pr->name[0])
2764 			error = EBUSY;
2765 		PF_RULES_RUNLOCK();
2766 		break;
2767 	}
2768 
2769 	case DIOCRCLRTABLES: {
2770 		struct pfioc_table *io = (struct pfioc_table *)addr;
2771 
2772 		if (io->pfrio_esize != 0) {
2773 			error = ENODEV;
2774 			break;
2775 		}
2776 		PF_RULES_WLOCK();
2777 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2778 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2779 		PF_RULES_WUNLOCK();
2780 		break;
2781 	}
2782 
2783 	case DIOCRADDTABLES: {
2784 		struct pfioc_table *io = (struct pfioc_table *)addr;
2785 		struct pfr_table *pfrts;
2786 		size_t totlen;
2787 
2788 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2789 			error = ENODEV;
2790 			break;
2791 		}
2792 
2793 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
2794 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
2795 			error = ENOMEM;
2796 			break;
2797 		}
2798 
2799 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2800 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
2801 		    M_TEMP, M_WAITOK);
2802 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2803 		if (error) {
2804 			free(pfrts, M_TEMP);
2805 			break;
2806 		}
2807 		PF_RULES_WLOCK();
2808 		error = pfr_add_tables(pfrts, io->pfrio_size,
2809 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2810 		PF_RULES_WUNLOCK();
2811 		free(pfrts, M_TEMP);
2812 		break;
2813 	}
2814 
2815 	case DIOCRDELTABLES: {
2816 		struct pfioc_table *io = (struct pfioc_table *)addr;
2817 		struct pfr_table *pfrts;
2818 		size_t totlen;
2819 
2820 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2821 			error = ENODEV;
2822 			break;
2823 		}
2824 
2825 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
2826 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
2827 			error = ENOMEM;
2828 			break;
2829 		}
2830 
2831 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2832 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
2833 		    M_TEMP, M_WAITOK);
2834 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2835 		if (error) {
2836 			free(pfrts, M_TEMP);
2837 			break;
2838 		}
2839 		PF_RULES_WLOCK();
2840 		error = pfr_del_tables(pfrts, io->pfrio_size,
2841 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2842 		PF_RULES_WUNLOCK();
2843 		free(pfrts, M_TEMP);
2844 		break;
2845 	}
2846 
2847 	case DIOCRGETTABLES: {
2848 		struct pfioc_table *io = (struct pfioc_table *)addr;
2849 		struct pfr_table *pfrts;
2850 		size_t totlen, n;
2851 
2852 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2853 			error = ENODEV;
2854 			break;
2855 		}
2856 		PF_RULES_RLOCK();
2857 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
2858 		io->pfrio_size = min(io->pfrio_size, n);
2859 
2860 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2861 
2862 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
2863 		    M_TEMP, M_NOWAIT);
2864 		if (pfrts == NULL) {
2865 			error = ENOMEM;
2866 			PF_RULES_RUNLOCK();
2867 			break;
2868 		}
2869 		error = pfr_get_tables(&io->pfrio_table, pfrts,
2870 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2871 		PF_RULES_RUNLOCK();
2872 		if (error == 0)
2873 			error = copyout(pfrts, io->pfrio_buffer, totlen);
2874 		free(pfrts, M_TEMP);
2875 		break;
2876 	}
2877 
2878 	case DIOCRGETTSTATS: {
2879 		struct pfioc_table *io = (struct pfioc_table *)addr;
2880 		struct pfr_tstats *pfrtstats;
2881 		size_t totlen, n;
2882 
2883 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2884 			error = ENODEV;
2885 			break;
2886 		}
2887 		PF_RULES_WLOCK();
2888 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
2889 		io->pfrio_size = min(io->pfrio_size, n);
2890 
2891 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
2892 		pfrtstats = mallocarray(io->pfrio_size,
2893 		    sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT);
2894 		if (pfrtstats == NULL) {
2895 			error = ENOMEM;
2896 			PF_RULES_WUNLOCK();
2897 			break;
2898 		}
2899 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
2900 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2901 		PF_RULES_WUNLOCK();
2902 		if (error == 0)
2903 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
2904 		free(pfrtstats, M_TEMP);
2905 		break;
2906 	}
2907 
2908 	case DIOCRCLRTSTATS: {
2909 		struct pfioc_table *io = (struct pfioc_table *)addr;
2910 		struct pfr_table *pfrts;
2911 		size_t totlen, n;
2912 
2913 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2914 			error = ENODEV;
2915 			break;
2916 		}
2917 
2918 		PF_RULES_WLOCK();
2919 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
2920 		io->pfrio_size = min(io->pfrio_size, n);
2921 
2922 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2923 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
2924 		    M_TEMP, M_NOWAIT);
2925 		if (pfrts == NULL) {
2926 			error = ENOMEM;
2927 			PF_RULES_WUNLOCK();
2928 			break;
2929 		}
2930 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2931 		if (error) {
2932 			free(pfrts, M_TEMP);
2933 			PF_RULES_WUNLOCK();
2934 			break;
2935 		}
2936 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
2937 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2938 		PF_RULES_WUNLOCK();
2939 		free(pfrts, M_TEMP);
2940 		break;
2941 	}
2942 
2943 	case DIOCRSETTFLAGS: {
2944 		struct pfioc_table *io = (struct pfioc_table *)addr;
2945 		struct pfr_table *pfrts;
2946 		size_t totlen, n;
2947 
2948 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2949 			error = ENODEV;
2950 			break;
2951 		}
2952 
2953 		PF_RULES_WLOCK();
2954 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
2955 		io->pfrio_size = min(io->pfrio_size, n);
2956 
2957 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2958 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
2959 		    M_TEMP, M_NOWAIT);
2960 		if (pfrts == NULL) {
2961 			error = ENOMEM;
2962 			PF_RULES_WUNLOCK();
2963 			break;
2964 		}
2965 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2966 		if (error) {
2967 			free(pfrts, M_TEMP);
2968 			PF_RULES_WUNLOCK();
2969 			break;
2970 		}
2971 		error = pfr_set_tflags(pfrts, io->pfrio_size,
2972 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2973 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2974 		PF_RULES_WUNLOCK();
2975 		free(pfrts, M_TEMP);
2976 		break;
2977 	}
2978 
2979 	case DIOCRCLRADDRS: {
2980 		struct pfioc_table *io = (struct pfioc_table *)addr;
2981 
2982 		if (io->pfrio_esize != 0) {
2983 			error = ENODEV;
2984 			break;
2985 		}
2986 		PF_RULES_WLOCK();
2987 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2988 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2989 		PF_RULES_WUNLOCK();
2990 		break;
2991 	}
2992 
2993 	case DIOCRADDADDRS: {
2994 		struct pfioc_table *io = (struct pfioc_table *)addr;
2995 		struct pfr_addr *pfras;
2996 		size_t totlen;
2997 
2998 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2999 			error = ENODEV;
3000 			break;
3001 		}
3002 		if (io->pfrio_size < 0 ||
3003 		    io->pfrio_size > pf_ioctl_maxcount ||
3004 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3005 			error = EINVAL;
3006 			break;
3007 		}
3008 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3009 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3010 		    M_TEMP, M_NOWAIT);
3011 		if (! pfras) {
3012 			error = ENOMEM;
3013 			break;
3014 		}
3015 		error = copyin(io->pfrio_buffer, pfras, totlen);
3016 		if (error) {
3017 			free(pfras, M_TEMP);
3018 			break;
3019 		}
3020 		PF_RULES_WLOCK();
3021 		error = pfr_add_addrs(&io->pfrio_table, pfras,
3022 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
3023 		    PFR_FLAG_USERIOCTL);
3024 		PF_RULES_WUNLOCK();
3025 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3026 			error = copyout(pfras, io->pfrio_buffer, totlen);
3027 		free(pfras, M_TEMP);
3028 		break;
3029 	}
3030 
3031 	case DIOCRDELADDRS: {
3032 		struct pfioc_table *io = (struct pfioc_table *)addr;
3033 		struct pfr_addr *pfras;
3034 		size_t totlen;
3035 
3036 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3037 			error = ENODEV;
3038 			break;
3039 		}
3040 		if (io->pfrio_size < 0 ||
3041 		    io->pfrio_size > pf_ioctl_maxcount ||
3042 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3043 			error = EINVAL;
3044 			break;
3045 		}
3046 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3047 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3048 		    M_TEMP, M_NOWAIT);
3049 		if (! pfras) {
3050 			error = ENOMEM;
3051 			break;
3052 		}
3053 		error = copyin(io->pfrio_buffer, pfras, totlen);
3054 		if (error) {
3055 			free(pfras, M_TEMP);
3056 			break;
3057 		}
3058 		PF_RULES_WLOCK();
3059 		error = pfr_del_addrs(&io->pfrio_table, pfras,
3060 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
3061 		    PFR_FLAG_USERIOCTL);
3062 		PF_RULES_WUNLOCK();
3063 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3064 			error = copyout(pfras, io->pfrio_buffer, totlen);
3065 		free(pfras, M_TEMP);
3066 		break;
3067 	}
3068 
3069 	case DIOCRSETADDRS: {
3070 		struct pfioc_table *io = (struct pfioc_table *)addr;
3071 		struct pfr_addr *pfras;
3072 		size_t totlen, count;
3073 
3074 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3075 			error = ENODEV;
3076 			break;
3077 		}
3078 		if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
3079 			error = EINVAL;
3080 			break;
3081 		}
3082 		count = max(io->pfrio_size, io->pfrio_size2);
3083 		if (count > pf_ioctl_maxcount ||
3084 		    WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
3085 			error = EINVAL;
3086 			break;
3087 		}
3088 		totlen = count * sizeof(struct pfr_addr);
3089 		pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
3090 		    M_NOWAIT);
3091 		if (! pfras) {
3092 			error = ENOMEM;
3093 			break;
3094 		}
3095 		error = copyin(io->pfrio_buffer, pfras, totlen);
3096 		if (error) {
3097 			free(pfras, M_TEMP);
3098 			break;
3099 		}
3100 		PF_RULES_WLOCK();
3101 		error = pfr_set_addrs(&io->pfrio_table, pfras,
3102 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
3103 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
3104 		    PFR_FLAG_USERIOCTL, 0);
3105 		PF_RULES_WUNLOCK();
3106 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3107 			error = copyout(pfras, io->pfrio_buffer, totlen);
3108 		free(pfras, M_TEMP);
3109 		break;
3110 	}
3111 
3112 	case DIOCRGETADDRS: {
3113 		struct pfioc_table *io = (struct pfioc_table *)addr;
3114 		struct pfr_addr *pfras;
3115 		size_t totlen;
3116 
3117 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3118 			error = ENODEV;
3119 			break;
3120 		}
3121 		if (io->pfrio_size < 0 ||
3122 		    io->pfrio_size > pf_ioctl_maxcount ||
3123 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3124 			error = EINVAL;
3125 			break;
3126 		}
3127 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3128 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3129 		    M_TEMP, M_NOWAIT);
3130 		if (! pfras) {
3131 			error = ENOMEM;
3132 			break;
3133 		}
3134 		PF_RULES_RLOCK();
3135 		error = pfr_get_addrs(&io->pfrio_table, pfras,
3136 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3137 		PF_RULES_RUNLOCK();
3138 		if (error == 0)
3139 			error = copyout(pfras, io->pfrio_buffer, totlen);
3140 		free(pfras, M_TEMP);
3141 		break;
3142 	}
3143 
3144 	case DIOCRGETASTATS: {
3145 		struct pfioc_table *io = (struct pfioc_table *)addr;
3146 		struct pfr_astats *pfrastats;
3147 		size_t totlen;
3148 
3149 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
3150 			error = ENODEV;
3151 			break;
3152 		}
3153 		if (io->pfrio_size < 0 ||
3154 		    io->pfrio_size > pf_ioctl_maxcount ||
3155 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
3156 			error = EINVAL;
3157 			break;
3158 		}
3159 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
3160 		pfrastats = mallocarray(io->pfrio_size,
3161 		    sizeof(struct pfr_astats), M_TEMP, M_NOWAIT);
3162 		if (! pfrastats) {
3163 			error = ENOMEM;
3164 			break;
3165 		}
3166 		PF_RULES_RLOCK();
3167 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
3168 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3169 		PF_RULES_RUNLOCK();
3170 		if (error == 0)
3171 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
3172 		free(pfrastats, M_TEMP);
3173 		break;
3174 	}
3175 
3176 	case DIOCRCLRASTATS: {
3177 		struct pfioc_table *io = (struct pfioc_table *)addr;
3178 		struct pfr_addr *pfras;
3179 		size_t totlen;
3180 
3181 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3182 			error = ENODEV;
3183 			break;
3184 		}
3185 		if (io->pfrio_size < 0 ||
3186 		    io->pfrio_size > pf_ioctl_maxcount ||
3187 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3188 			error = EINVAL;
3189 			break;
3190 		}
3191 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3192 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3193 		    M_TEMP, M_NOWAIT);
3194 		if (! pfras) {
3195 			error = ENOMEM;
3196 			break;
3197 		}
3198 		error = copyin(io->pfrio_buffer, pfras, totlen);
3199 		if (error) {
3200 			free(pfras, M_TEMP);
3201 			break;
3202 		}
3203 		PF_RULES_WLOCK();
3204 		error = pfr_clr_astats(&io->pfrio_table, pfras,
3205 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
3206 		    PFR_FLAG_USERIOCTL);
3207 		PF_RULES_WUNLOCK();
3208 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3209 			error = copyout(pfras, io->pfrio_buffer, totlen);
3210 		free(pfras, M_TEMP);
3211 		break;
3212 	}
3213 
3214 	case DIOCRTSTADDRS: {
3215 		struct pfioc_table *io = (struct pfioc_table *)addr;
3216 		struct pfr_addr *pfras;
3217 		size_t totlen;
3218 
3219 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3220 			error = ENODEV;
3221 			break;
3222 		}
3223 		if (io->pfrio_size < 0 ||
3224 		    io->pfrio_size > pf_ioctl_maxcount ||
3225 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3226 			error = EINVAL;
3227 			break;
3228 		}
3229 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3230 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3231 		    M_TEMP, M_NOWAIT);
3232 		if (! pfras) {
3233 			error = ENOMEM;
3234 			break;
3235 		}
3236 		error = copyin(io->pfrio_buffer, pfras, totlen);
3237 		if (error) {
3238 			free(pfras, M_TEMP);
3239 			break;
3240 		}
3241 		PF_RULES_RLOCK();
3242 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
3243 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
3244 		    PFR_FLAG_USERIOCTL);
3245 		PF_RULES_RUNLOCK();
3246 		if (error == 0)
3247 			error = copyout(pfras, io->pfrio_buffer, totlen);
3248 		free(pfras, M_TEMP);
3249 		break;
3250 	}
3251 
3252 	case DIOCRINADEFINE: {
3253 		struct pfioc_table *io = (struct pfioc_table *)addr;
3254 		struct pfr_addr *pfras;
3255 		size_t totlen;
3256 
3257 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3258 			error = ENODEV;
3259 			break;
3260 		}
3261 		if (io->pfrio_size < 0 ||
3262 		    io->pfrio_size > pf_ioctl_maxcount ||
3263 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3264 			error = EINVAL;
3265 			break;
3266 		}
3267 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3268 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3269 		    M_TEMP, M_NOWAIT);
3270 		if (! pfras) {
3271 			error = ENOMEM;
3272 			break;
3273 		}
3274 		error = copyin(io->pfrio_buffer, pfras, totlen);
3275 		if (error) {
3276 			free(pfras, M_TEMP);
3277 			break;
3278 		}
3279 		PF_RULES_WLOCK();
3280 		error = pfr_ina_define(&io->pfrio_table, pfras,
3281 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
3282 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3283 		PF_RULES_WUNLOCK();
3284 		free(pfras, M_TEMP);
3285 		break;
3286 	}
3287 
3288 	case DIOCOSFPADD: {
3289 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
3290 		PF_RULES_WLOCK();
3291 		error = pf_osfp_add(io);
3292 		PF_RULES_WUNLOCK();
3293 		break;
3294 	}
3295 
3296 	case DIOCOSFPGET: {
3297 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
3298 		PF_RULES_RLOCK();
3299 		error = pf_osfp_get(io);
3300 		PF_RULES_RUNLOCK();
3301 		break;
3302 	}
3303 
3304 	case DIOCXBEGIN: {
3305 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3306 		struct pfioc_trans_e	*ioes, *ioe;
3307 		size_t			 totlen;
3308 		int			 i;
3309 
3310 		if (io->esize != sizeof(*ioe)) {
3311 			error = ENODEV;
3312 			break;
3313 		}
3314 		if (io->size < 0 ||
3315 		    io->size > pf_ioctl_maxcount ||
3316 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3317 			error = EINVAL;
3318 			break;
3319 		}
3320 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3321 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3322 		    M_TEMP, M_NOWAIT);
3323 		if (! ioes) {
3324 			error = ENOMEM;
3325 			break;
3326 		}
3327 		error = copyin(io->array, ioes, totlen);
3328 		if (error) {
3329 			free(ioes, M_TEMP);
3330 			break;
3331 		}
3332 		PF_RULES_WLOCK();
3333 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3334 			switch (ioe->rs_num) {
3335 #ifdef ALTQ
3336 			case PF_RULESET_ALTQ:
3337 				if (ioe->anchor[0]) {
3338 					PF_RULES_WUNLOCK();
3339 					free(ioes, M_TEMP);
3340 					error = EINVAL;
3341 					goto fail;
3342 				}
3343 				if ((error = pf_begin_altq(&ioe->ticket))) {
3344 					PF_RULES_WUNLOCK();
3345 					free(ioes, M_TEMP);
3346 					goto fail;
3347 				}
3348 				break;
3349 #endif /* ALTQ */
3350 			case PF_RULESET_TABLE:
3351 			    {
3352 				struct pfr_table table;
3353 
3354 				bzero(&table, sizeof(table));
3355 				strlcpy(table.pfrt_anchor, ioe->anchor,
3356 				    sizeof(table.pfrt_anchor));
3357 				if ((error = pfr_ina_begin(&table,
3358 				    &ioe->ticket, NULL, 0))) {
3359 					PF_RULES_WUNLOCK();
3360 					free(ioes, M_TEMP);
3361 					goto fail;
3362 				}
3363 				break;
3364 			    }
3365 			default:
3366 				if ((error = pf_begin_rules(&ioe->ticket,
3367 				    ioe->rs_num, ioe->anchor))) {
3368 					PF_RULES_WUNLOCK();
3369 					free(ioes, M_TEMP);
3370 					goto fail;
3371 				}
3372 				break;
3373 			}
3374 		}
3375 		PF_RULES_WUNLOCK();
3376 		error = copyout(ioes, io->array, totlen);
3377 		free(ioes, M_TEMP);
3378 		break;
3379 	}
3380 
3381 	case DIOCXROLLBACK: {
3382 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3383 		struct pfioc_trans_e	*ioe, *ioes;
3384 		size_t			 totlen;
3385 		int			 i;
3386 
3387 		if (io->esize != sizeof(*ioe)) {
3388 			error = ENODEV;
3389 			break;
3390 		}
3391 		if (io->size < 0 ||
3392 		    io->size > pf_ioctl_maxcount ||
3393 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3394 			error = EINVAL;
3395 			break;
3396 		}
3397 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3398 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3399 		    M_TEMP, M_NOWAIT);
3400 		if (! ioes) {
3401 			error = ENOMEM;
3402 			break;
3403 		}
3404 		error = copyin(io->array, ioes, totlen);
3405 		if (error) {
3406 			free(ioes, M_TEMP);
3407 			break;
3408 		}
3409 		PF_RULES_WLOCK();
3410 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3411 			switch (ioe->rs_num) {
3412 #ifdef ALTQ
3413 			case PF_RULESET_ALTQ:
3414 				if (ioe->anchor[0]) {
3415 					PF_RULES_WUNLOCK();
3416 					free(ioes, M_TEMP);
3417 					error = EINVAL;
3418 					goto fail;
3419 				}
3420 				if ((error = pf_rollback_altq(ioe->ticket))) {
3421 					PF_RULES_WUNLOCK();
3422 					free(ioes, M_TEMP);
3423 					goto fail; /* really bad */
3424 				}
3425 				break;
3426 #endif /* ALTQ */
3427 			case PF_RULESET_TABLE:
3428 			    {
3429 				struct pfr_table table;
3430 
3431 				bzero(&table, sizeof(table));
3432 				strlcpy(table.pfrt_anchor, ioe->anchor,
3433 				    sizeof(table.pfrt_anchor));
3434 				if ((error = pfr_ina_rollback(&table,
3435 				    ioe->ticket, NULL, 0))) {
3436 					PF_RULES_WUNLOCK();
3437 					free(ioes, M_TEMP);
3438 					goto fail; /* really bad */
3439 				}
3440 				break;
3441 			    }
3442 			default:
3443 				if ((error = pf_rollback_rules(ioe->ticket,
3444 				    ioe->rs_num, ioe->anchor))) {
3445 					PF_RULES_WUNLOCK();
3446 					free(ioes, M_TEMP);
3447 					goto fail; /* really bad */
3448 				}
3449 				break;
3450 			}
3451 		}
3452 		PF_RULES_WUNLOCK();
3453 		free(ioes, M_TEMP);
3454 		break;
3455 	}
3456 
3457 	case DIOCXCOMMIT: {
3458 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3459 		struct pfioc_trans_e	*ioe, *ioes;
3460 		struct pf_ruleset	*rs;
3461 		size_t			 totlen;
3462 		int			 i;
3463 
3464 		if (io->esize != sizeof(*ioe)) {
3465 			error = ENODEV;
3466 			break;
3467 		}
3468 
3469 		if (io->size < 0 ||
3470 		    io->size > pf_ioctl_maxcount ||
3471 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3472 			error = EINVAL;
3473 			break;
3474 		}
3475 
3476 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3477 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3478 		    M_TEMP, M_NOWAIT);
3479 		if (ioes == NULL) {
3480 			error = ENOMEM;
3481 			break;
3482 		}
3483 		error = copyin(io->array, ioes, totlen);
3484 		if (error) {
3485 			free(ioes, M_TEMP);
3486 			break;
3487 		}
3488 		PF_RULES_WLOCK();
3489 		/* First makes sure everything will succeed. */
3490 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3491 			switch (ioe->rs_num) {
3492 #ifdef ALTQ
3493 			case PF_RULESET_ALTQ:
3494 				if (ioe->anchor[0]) {
3495 					PF_RULES_WUNLOCK();
3496 					free(ioes, M_TEMP);
3497 					error = EINVAL;
3498 					goto fail;
3499 				}
3500 				if (!V_altqs_inactive_open || ioe->ticket !=
3501 				    V_ticket_altqs_inactive) {
3502 					PF_RULES_WUNLOCK();
3503 					free(ioes, M_TEMP);
3504 					error = EBUSY;
3505 					goto fail;
3506 				}
3507 				break;
3508 #endif /* ALTQ */
3509 			case PF_RULESET_TABLE:
3510 				rs = pf_find_ruleset(ioe->anchor);
3511 				if (rs == NULL || !rs->topen || ioe->ticket !=
3512 				    rs->tticket) {
3513 					PF_RULES_WUNLOCK();
3514 					free(ioes, M_TEMP);
3515 					error = EBUSY;
3516 					goto fail;
3517 				}
3518 				break;
3519 			default:
3520 				if (ioe->rs_num < 0 || ioe->rs_num >=
3521 				    PF_RULESET_MAX) {
3522 					PF_RULES_WUNLOCK();
3523 					free(ioes, M_TEMP);
3524 					error = EINVAL;
3525 					goto fail;
3526 				}
3527 				rs = pf_find_ruleset(ioe->anchor);
3528 				if (rs == NULL ||
3529 				    !rs->rules[ioe->rs_num].inactive.open ||
3530 				    rs->rules[ioe->rs_num].inactive.ticket !=
3531 				    ioe->ticket) {
3532 					PF_RULES_WUNLOCK();
3533 					free(ioes, M_TEMP);
3534 					error = EBUSY;
3535 					goto fail;
3536 				}
3537 				break;
3538 			}
3539 		}
3540 		/* Now do the commit - no errors should happen here. */
3541 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3542 			switch (ioe->rs_num) {
3543 #ifdef ALTQ
3544 			case PF_RULESET_ALTQ:
3545 				if ((error = pf_commit_altq(ioe->ticket))) {
3546 					PF_RULES_WUNLOCK();
3547 					free(ioes, M_TEMP);
3548 					goto fail; /* really bad */
3549 				}
3550 				break;
3551 #endif /* ALTQ */
3552 			case PF_RULESET_TABLE:
3553 			    {
3554 				struct pfr_table table;
3555 
3556 				bzero(&table, sizeof(table));
3557 				strlcpy(table.pfrt_anchor, ioe->anchor,
3558 				    sizeof(table.pfrt_anchor));
3559 				if ((error = pfr_ina_commit(&table,
3560 				    ioe->ticket, NULL, NULL, 0))) {
3561 					PF_RULES_WUNLOCK();
3562 					free(ioes, M_TEMP);
3563 					goto fail; /* really bad */
3564 				}
3565 				break;
3566 			    }
3567 			default:
3568 				if ((error = pf_commit_rules(ioe->ticket,
3569 				    ioe->rs_num, ioe->anchor))) {
3570 					PF_RULES_WUNLOCK();
3571 					free(ioes, M_TEMP);
3572 					goto fail; /* really bad */
3573 				}
3574 				break;
3575 			}
3576 		}
3577 		PF_RULES_WUNLOCK();
3578 		free(ioes, M_TEMP);
3579 		break;
3580 	}
3581 
3582 	case DIOCGETSRCNODES: {
3583 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
3584 		struct pf_srchash	*sh;
3585 		struct pf_src_node	*n, *p, *pstore;
3586 		uint32_t		 i, nr = 0;
3587 
3588 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3589 				i++, sh++) {
3590 			PF_HASHROW_LOCK(sh);
3591 			LIST_FOREACH(n, &sh->nodes, entry)
3592 				nr++;
3593 			PF_HASHROW_UNLOCK(sh);
3594 		}
3595 
3596 		psn->psn_len = min(psn->psn_len,
3597 		    sizeof(struct pf_src_node) * nr);
3598 
3599 		if (psn->psn_len == 0) {
3600 			psn->psn_len = sizeof(struct pf_src_node) * nr;
3601 			break;
3602 		}
3603 
3604 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK);
3605 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3606 		    i++, sh++) {
3607 		    PF_HASHROW_LOCK(sh);
3608 		    LIST_FOREACH(n, &sh->nodes, entry) {
3609 			int	secs = time_uptime, diff;
3610 
3611 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
3612 				break;
3613 
3614 			bcopy(n, p, sizeof(struct pf_src_node));
3615 			if (n->rule.ptr != NULL)
3616 				p->rule.nr = n->rule.ptr->nr;
3617 			p->creation = secs - p->creation;
3618 			if (p->expire > secs)
3619 				p->expire -= secs;
3620 			else
3621 				p->expire = 0;
3622 
3623 			/* Adjust the connection rate estimate. */
3624 			diff = secs - n->conn_rate.last;
3625 			if (diff >= n->conn_rate.seconds)
3626 				p->conn_rate.count = 0;
3627 			else
3628 				p->conn_rate.count -=
3629 				    n->conn_rate.count * diff /
3630 				    n->conn_rate.seconds;
3631 			p++;
3632 			nr++;
3633 		    }
3634 		    PF_HASHROW_UNLOCK(sh);
3635 		}
3636 		error = copyout(pstore, psn->psn_src_nodes,
3637 		    sizeof(struct pf_src_node) * nr);
3638 		if (error) {
3639 			free(pstore, M_TEMP);
3640 			break;
3641 		}
3642 		psn->psn_len = sizeof(struct pf_src_node) * nr;
3643 		free(pstore, M_TEMP);
3644 		break;
3645 	}
3646 
3647 	case DIOCCLRSRCNODES: {
3648 
3649 		pf_clear_srcnodes(NULL);
3650 		pf_purge_expired_src_nodes();
3651 		break;
3652 	}
3653 
3654 	case DIOCKILLSRCNODES:
3655 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
3656 		break;
3657 
3658 	case DIOCSETHOSTID: {
3659 		u_int32_t	*hostid = (u_int32_t *)addr;
3660 
3661 		PF_RULES_WLOCK();
3662 		if (*hostid == 0)
3663 			V_pf_status.hostid = arc4random();
3664 		else
3665 			V_pf_status.hostid = *hostid;
3666 		PF_RULES_WUNLOCK();
3667 		break;
3668 	}
3669 
3670 	case DIOCOSFPFLUSH:
3671 		PF_RULES_WLOCK();
3672 		pf_osfp_flush();
3673 		PF_RULES_WUNLOCK();
3674 		break;
3675 
3676 	case DIOCIGETIFACES: {
3677 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3678 		struct pfi_kif *ifstore;
3679 		size_t bufsiz;
3680 
3681 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
3682 			error = ENODEV;
3683 			break;
3684 		}
3685 
3686 		if (io->pfiio_size < 0 ||
3687 		    io->pfiio_size > pf_ioctl_maxcount ||
3688 		    WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
3689 			error = EINVAL;
3690 			break;
3691 		}
3692 
3693 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
3694 		ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
3695 		    M_TEMP, M_NOWAIT);
3696 		if (ifstore == NULL) {
3697 			error = ENOMEM;
3698 			break;
3699 		}
3700 
3701 		PF_RULES_RLOCK();
3702 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
3703 		PF_RULES_RUNLOCK();
3704 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
3705 		free(ifstore, M_TEMP);
3706 		break;
3707 	}
3708 
3709 	case DIOCSETIFFLAG: {
3710 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3711 
3712 		PF_RULES_WLOCK();
3713 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3714 		PF_RULES_WUNLOCK();
3715 		break;
3716 	}
3717 
3718 	case DIOCCLRIFFLAG: {
3719 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3720 
3721 		PF_RULES_WLOCK();
3722 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3723 		PF_RULES_WUNLOCK();
3724 		break;
3725 	}
3726 
3727 	default:
3728 		error = ENODEV;
3729 		break;
3730 	}
3731 fail:
3732 	if (sx_xlocked(&pf_ioctl_lock))
3733 		sx_xunlock(&pf_ioctl_lock);
3734 	CURVNET_RESTORE();
3735 
3736 	return (error);
3737 }
3738 
3739 void
3740 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
3741 {
3742 	bzero(sp, sizeof(struct pfsync_state));
3743 
3744 	/* copy from state key */
3745 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
3746 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
3747 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
3748 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
3749 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
3750 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
3751 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
3752 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
3753 	sp->proto = st->key[PF_SK_WIRE]->proto;
3754 	sp->af = st->key[PF_SK_WIRE]->af;
3755 
3756 	/* copy from state */
3757 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
3758 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
3759 	sp->creation = htonl(time_uptime - st->creation);
3760 	sp->expire = pf_state_expires(st);
3761 	if (sp->expire <= time_uptime)
3762 		sp->expire = htonl(0);
3763 	else
3764 		sp->expire = htonl(sp->expire - time_uptime);
3765 
3766 	sp->direction = st->direction;
3767 	sp->log = st->log;
3768 	sp->timeout = st->timeout;
3769 	sp->state_flags = st->state_flags;
3770 	if (st->src_node)
3771 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
3772 	if (st->nat_src_node)
3773 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
3774 
3775 	sp->id = st->id;
3776 	sp->creatorid = st->creatorid;
3777 	pf_state_peer_hton(&st->src, &sp->src);
3778 	pf_state_peer_hton(&st->dst, &sp->dst);
3779 
3780 	if (st->rule.ptr == NULL)
3781 		sp->rule = htonl(-1);
3782 	else
3783 		sp->rule = htonl(st->rule.ptr->nr);
3784 	if (st->anchor.ptr == NULL)
3785 		sp->anchor = htonl(-1);
3786 	else
3787 		sp->anchor = htonl(st->anchor.ptr->nr);
3788 	if (st->nat_rule.ptr == NULL)
3789 		sp->nat_rule = htonl(-1);
3790 	else
3791 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
3792 
3793 	pf_state_counter_hton(st->packets[0], sp->packets[0]);
3794 	pf_state_counter_hton(st->packets[1], sp->packets[1]);
3795 	pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
3796 	pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
3797 
3798 }
3799 
3800 static void
3801 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
3802 {
3803 	struct pfr_ktable *kt;
3804 
3805 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
3806 
3807 	kt = aw->p.tbl;
3808 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
3809 		kt = kt->pfrkt_root;
3810 	aw->p.tbl = NULL;
3811 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
3812 		kt->pfrkt_cnt : -1;
3813 }
3814 
3815 /*
3816  * XXX - Check for version missmatch!!!
3817  */
3818 static void
3819 pf_clear_states(void)
3820 {
3821 	struct pf_state	*s;
3822 	u_int i;
3823 
3824 	for (i = 0; i <= pf_hashmask; i++) {
3825 		struct pf_idhash *ih = &V_pf_idhash[i];
3826 relock:
3827 		PF_HASHROW_LOCK(ih);
3828 		LIST_FOREACH(s, &ih->states, entry) {
3829 			s->timeout = PFTM_PURGE;
3830 			/* Don't send out individual delete messages. */
3831 			s->state_flags |= PFSTATE_NOSYNC;
3832 			pf_unlink_state(s, PF_ENTER_LOCKED);
3833 			goto relock;
3834 		}
3835 		PF_HASHROW_UNLOCK(ih);
3836 	}
3837 }
3838 
3839 static int
3840 pf_clear_tables(void)
3841 {
3842 	struct pfioc_table io;
3843 	int error;
3844 
3845 	bzero(&io, sizeof(io));
3846 
3847 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
3848 	    io.pfrio_flags);
3849 
3850 	return (error);
3851 }
3852 
3853 static void
3854 pf_clear_srcnodes(struct pf_src_node *n)
3855 {
3856 	struct pf_state *s;
3857 	int i;
3858 
3859 	for (i = 0; i <= pf_hashmask; i++) {
3860 		struct pf_idhash *ih = &V_pf_idhash[i];
3861 
3862 		PF_HASHROW_LOCK(ih);
3863 		LIST_FOREACH(s, &ih->states, entry) {
3864 			if (n == NULL || n == s->src_node)
3865 				s->src_node = NULL;
3866 			if (n == NULL || n == s->nat_src_node)
3867 				s->nat_src_node = NULL;
3868 		}
3869 		PF_HASHROW_UNLOCK(ih);
3870 	}
3871 
3872 	if (n == NULL) {
3873 		struct pf_srchash *sh;
3874 
3875 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3876 		    i++, sh++) {
3877 			PF_HASHROW_LOCK(sh);
3878 			LIST_FOREACH(n, &sh->nodes, entry) {
3879 				n->expire = 1;
3880 				n->states = 0;
3881 			}
3882 			PF_HASHROW_UNLOCK(sh);
3883 		}
3884 	} else {
3885 		/* XXX: hash slot should already be locked here. */
3886 		n->expire = 1;
3887 		n->states = 0;
3888 	}
3889 }
3890 
3891 static void
3892 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
3893 {
3894 	struct pf_src_node_list	 kill;
3895 
3896 	LIST_INIT(&kill);
3897 	for (int i = 0; i <= pf_srchashmask; i++) {
3898 		struct pf_srchash *sh = &V_pf_srchash[i];
3899 		struct pf_src_node *sn, *tmp;
3900 
3901 		PF_HASHROW_LOCK(sh);
3902 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
3903 			if (PF_MATCHA(psnk->psnk_src.neg,
3904 			      &psnk->psnk_src.addr.v.a.addr,
3905 			      &psnk->psnk_src.addr.v.a.mask,
3906 			      &sn->addr, sn->af) &&
3907 			    PF_MATCHA(psnk->psnk_dst.neg,
3908 			      &psnk->psnk_dst.addr.v.a.addr,
3909 			      &psnk->psnk_dst.addr.v.a.mask,
3910 			      &sn->raddr, sn->af)) {
3911 				pf_unlink_src_node(sn);
3912 				LIST_INSERT_HEAD(&kill, sn, entry);
3913 				sn->expire = 1;
3914 			}
3915 		PF_HASHROW_UNLOCK(sh);
3916 	}
3917 
3918 	for (int i = 0; i <= pf_hashmask; i++) {
3919 		struct pf_idhash *ih = &V_pf_idhash[i];
3920 		struct pf_state *s;
3921 
3922 		PF_HASHROW_LOCK(ih);
3923 		LIST_FOREACH(s, &ih->states, entry) {
3924 			if (s->src_node && s->src_node->expire == 1)
3925 				s->src_node = NULL;
3926 			if (s->nat_src_node && s->nat_src_node->expire == 1)
3927 				s->nat_src_node = NULL;
3928 		}
3929 		PF_HASHROW_UNLOCK(ih);
3930 	}
3931 
3932 	psnk->psnk_killed = pf_free_src_nodes(&kill);
3933 }
3934 
3935 /*
3936  * XXX - Check for version missmatch!!!
3937  */
3938 
3939 /*
3940  * Duplicate pfctl -Fa operation to get rid of as much as we can.
3941  */
3942 static int
3943 shutdown_pf(void)
3944 {
3945 	int error = 0;
3946 	u_int32_t t[5];
3947 	char nn = '\0';
3948 
3949 	do {
3950 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
3951 		    != 0) {
3952 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
3953 			break;
3954 		}
3955 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
3956 		    != 0) {
3957 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
3958 			break;		/* XXX: rollback? */
3959 		}
3960 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
3961 		    != 0) {
3962 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
3963 			break;		/* XXX: rollback? */
3964 		}
3965 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
3966 		    != 0) {
3967 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
3968 			break;		/* XXX: rollback? */
3969 		}
3970 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
3971 		    != 0) {
3972 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
3973 			break;		/* XXX: rollback? */
3974 		}
3975 
3976 		/* XXX: these should always succeed here */
3977 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
3978 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
3979 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
3980 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
3981 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
3982 
3983 		if ((error = pf_clear_tables()) != 0)
3984 			break;
3985 
3986 #ifdef ALTQ
3987 		if ((error = pf_begin_altq(&t[0])) != 0) {
3988 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
3989 			break;
3990 		}
3991 		pf_commit_altq(t[0]);
3992 #endif
3993 
3994 		pf_clear_states();
3995 
3996 		pf_clear_srcnodes(NULL);
3997 
3998 		/* status does not use malloced mem so no need to cleanup */
3999 		/* fingerprints and interfaces have their own cleanup code */
4000 	} while(0);
4001 
4002 	return (error);
4003 }
4004 
4005 static pfil_return_t
4006 pf_check_return(int chk, struct mbuf **m)
4007 {
4008 
4009 	switch (chk) {
4010 	case PF_PASS:
4011 		if (*m == NULL)
4012 			return (PFIL_CONSUMED);
4013 		else
4014 			return (PFIL_PASS);
4015 		break;
4016 	default:
4017 		if (*m != NULL) {
4018 			m_freem(*m);
4019 			*m = NULL;
4020 		}
4021 		return (PFIL_DROPPED);
4022 	}
4023 }
4024 
4025 #ifdef INET
4026 static pfil_return_t
4027 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
4028     void *ruleset __unused, struct inpcb *inp)
4029 {
4030 	int chk;
4031 
4032 	chk = pf_test(PF_IN, flags, ifp, m, inp);
4033 
4034 	return (pf_check_return(chk, m));
4035 }
4036 
4037 static pfil_return_t
4038 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
4039     void *ruleset __unused,  struct inpcb *inp)
4040 {
4041 	int chk;
4042 
4043 	chk = pf_test(PF_OUT, flags, ifp, m, inp);
4044 
4045 	return (pf_check_return(chk, m));
4046 }
4047 #endif
4048 
4049 #ifdef INET6
4050 static pfil_return_t
4051 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
4052     void *ruleset __unused,  struct inpcb *inp)
4053 {
4054 	int chk;
4055 
4056 	/*
4057 	 * In case of loopback traffic IPv6 uses the real interface in
4058 	 * order to support scoped addresses. In order to support stateful
4059 	 * filtering we have change this to lo0 as it is the case in IPv4.
4060 	 */
4061 	CURVNET_SET(ifp->if_vnet);
4062 	chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
4063 	CURVNET_RESTORE();
4064 
4065 	return (pf_check_return(chk, m));
4066 }
4067 
4068 static pfil_return_t
4069 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
4070     void *ruleset __unused,  struct inpcb *inp)
4071 {
4072 	int chk;
4073 
4074 	CURVNET_SET(ifp->if_vnet);
4075 	chk = pf_test6(PF_OUT, flags, ifp, m, inp);
4076 	CURVNET_RESTORE();
4077 
4078 	return (pf_check_return(chk, m));
4079 }
4080 #endif /* INET6 */
4081 
4082 #ifdef INET
4083 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
4084 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
4085 #define	V_pf_ip4_in_hook	VNET(pf_ip4_in_hook)
4086 #define	V_pf_ip4_out_hook	VNET(pf_ip4_out_hook)
4087 #endif
4088 #ifdef INET6
4089 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
4090 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
4091 #define	V_pf_ip6_in_hook	VNET(pf_ip6_in_hook)
4092 #define	V_pf_ip6_out_hook	VNET(pf_ip6_out_hook)
4093 #endif
4094 
4095 static int
4096 hook_pf(void)
4097 {
4098 	struct pfil_hook_args pha;
4099 	struct pfil_link_args pla;
4100 
4101 	if (V_pf_pfil_hooked)
4102 		return (0);
4103 
4104 	pha.pa_version = PFIL_VERSION;
4105 	pha.pa_modname = "pf";
4106 	pha.pa_ruleset = NULL;
4107 
4108 	pla.pa_version = PFIL_VERSION;
4109 
4110 #ifdef INET
4111 	pha.pa_type = PFIL_TYPE_IP4;
4112 	pha.pa_func = pf_check_in;
4113 	pha.pa_flags = PFIL_IN;
4114 	pha.pa_rulname = "default-in";
4115 	V_pf_ip4_in_hook = pfil_add_hook(&pha);
4116 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
4117 	pla.pa_head = V_inet_pfil_head;
4118 	pla.pa_hook = V_pf_ip4_in_hook;
4119 	(void)pfil_link(&pla);
4120 	pha.pa_func = pf_check_out;
4121 	pha.pa_flags = PFIL_OUT;
4122 	pha.pa_rulname = "default-out";
4123 	V_pf_ip4_out_hook = pfil_add_hook(&pha);
4124 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
4125 	pla.pa_head = V_inet_pfil_head;
4126 	pla.pa_hook = V_pf_ip4_out_hook;
4127 	(void)pfil_link(&pla);
4128 #endif
4129 #ifdef INET6
4130 	pha.pa_type = PFIL_TYPE_IP6;
4131 	pha.pa_func = pf_check6_in;
4132 	pha.pa_flags = PFIL_IN;
4133 	pha.pa_rulname = "default-in6";
4134 	V_pf_ip6_in_hook = pfil_add_hook(&pha);
4135 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
4136 	pla.pa_head = V_inet6_pfil_head;
4137 	pla.pa_hook = V_pf_ip6_in_hook;
4138 	(void)pfil_link(&pla);
4139 	pha.pa_func = pf_check6_out;
4140 	pha.pa_rulname = "default-out6";
4141 	pha.pa_flags = PFIL_OUT;
4142 	V_pf_ip6_out_hook = pfil_add_hook(&pha);
4143 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
4144 	pla.pa_head = V_inet6_pfil_head;
4145 	pla.pa_hook = V_pf_ip6_out_hook;
4146 	(void)pfil_link(&pla);
4147 #endif
4148 
4149 	V_pf_pfil_hooked = 1;
4150 	return (0);
4151 }
4152 
4153 static int
4154 dehook_pf(void)
4155 {
4156 
4157 	if (V_pf_pfil_hooked == 0)
4158 		return (0);
4159 
4160 #ifdef INET
4161 	pfil_remove_hook(V_pf_ip4_in_hook);
4162 	pfil_remove_hook(V_pf_ip4_out_hook);
4163 #endif
4164 #ifdef INET6
4165 	pfil_remove_hook(V_pf_ip6_in_hook);
4166 	pfil_remove_hook(V_pf_ip6_out_hook);
4167 #endif
4168 
4169 	V_pf_pfil_hooked = 0;
4170 	return (0);
4171 }
4172 
4173 static void
4174 pf_load_vnet(void)
4175 {
4176 	TAILQ_INIT(&V_pf_tags);
4177 	TAILQ_INIT(&V_pf_qids);
4178 
4179 	pfattach_vnet();
4180 	V_pf_vnet_active = 1;
4181 }
4182 
4183 static int
4184 pf_load(void)
4185 {
4186 	int error;
4187 
4188 	rm_init(&pf_rules_lock, "pf rulesets");
4189 	sx_init(&pf_ioctl_lock, "pf ioctl");
4190 	sx_init(&pf_end_lock, "pf end thread");
4191 
4192 	pf_mtag_initialize();
4193 
4194 	pf_dev = make_dev(&pf_cdevsw, 0, 0, 0, 0600, PF_NAME);
4195 	if (pf_dev == NULL)
4196 		return (ENOMEM);
4197 
4198 	pf_end_threads = 0;
4199 	error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
4200 	if (error != 0)
4201 		return (error);
4202 
4203 	pfi_initialize();
4204 
4205 	return (0);
4206 }
4207 
4208 static void
4209 pf_unload_vnet(void)
4210 {
4211 	int error;
4212 
4213 	V_pf_vnet_active = 0;
4214 	V_pf_status.running = 0;
4215 	error = dehook_pf();
4216 	if (error) {
4217 		/*
4218 		 * Should not happen!
4219 		 * XXX Due to error code ESRCH, kldunload will show
4220 		 * a message like 'No such process'.
4221 		 */
4222 		printf("%s : pfil unregisteration fail\n", __FUNCTION__);
4223 		return;
4224 	}
4225 
4226 	PF_RULES_WLOCK();
4227 	shutdown_pf();
4228 	PF_RULES_WUNLOCK();
4229 
4230 	swi_remove(V_pf_swi_cookie);
4231 
4232 	pf_unload_vnet_purge();
4233 
4234 	pf_normalize_cleanup();
4235 	PF_RULES_WLOCK();
4236 	pfi_cleanup_vnet();
4237 	PF_RULES_WUNLOCK();
4238 	pfr_cleanup();
4239 	pf_osfp_flush();
4240 	pf_cleanup();
4241 	if (IS_DEFAULT_VNET(curvnet))
4242 		pf_mtag_cleanup();
4243 
4244 	/* Free counters last as we updated them during shutdown. */
4245 	counter_u64_free(V_pf_default_rule.states_cur);
4246 	counter_u64_free(V_pf_default_rule.states_tot);
4247 	counter_u64_free(V_pf_default_rule.src_nodes);
4248 
4249 	for (int i = 0; i < PFRES_MAX; i++)
4250 		counter_u64_free(V_pf_status.counters[i]);
4251 	for (int i = 0; i < LCNT_MAX; i++)
4252 		counter_u64_free(V_pf_status.lcounters[i]);
4253 	for (int i = 0; i < FCNT_MAX; i++)
4254 		counter_u64_free(V_pf_status.fcounters[i]);
4255 	for (int i = 0; i < SCNT_MAX; i++)
4256 		counter_u64_free(V_pf_status.scounters[i]);
4257 }
4258 
4259 static void
4260 pf_unload(void)
4261 {
4262 
4263 	sx_xlock(&pf_end_lock);
4264 	pf_end_threads = 1;
4265 	while (pf_end_threads < 2) {
4266 		wakeup_one(pf_purge_thread);
4267 		sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
4268 	}
4269 	sx_xunlock(&pf_end_lock);
4270 
4271 	if (pf_dev != NULL)
4272 		destroy_dev(pf_dev);
4273 
4274 	pfi_cleanup();
4275 
4276 	rm_destroy(&pf_rules_lock);
4277 	sx_destroy(&pf_ioctl_lock);
4278 	sx_destroy(&pf_end_lock);
4279 }
4280 
4281 static void
4282 vnet_pf_init(void *unused __unused)
4283 {
4284 
4285 	pf_load_vnet();
4286 }
4287 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
4288     vnet_pf_init, NULL);
4289 
4290 static void
4291 vnet_pf_uninit(const void *unused __unused)
4292 {
4293 
4294 	pf_unload_vnet();
4295 }
4296 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
4297 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
4298     vnet_pf_uninit, NULL);
4299 
4300 
4301 static int
4302 pf_modevent(module_t mod, int type, void *data)
4303 {
4304 	int error = 0;
4305 
4306 	switch(type) {
4307 	case MOD_LOAD:
4308 		error = pf_load();
4309 		break;
4310 	case MOD_UNLOAD:
4311 		/* Handled in SYSUNINIT(pf_unload) to ensure it's done after
4312 		 * the vnet_pf_uninit()s */
4313 		break;
4314 	default:
4315 		error = EINVAL;
4316 		break;
4317 	}
4318 
4319 	return (error);
4320 }
4321 
4322 static moduledata_t pf_mod = {
4323 	"pf",
4324 	pf_modevent,
4325 	0
4326 };
4327 
4328 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
4329 MODULE_VERSION(pf, PF_MODVER);
4330