xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision cfd6422a5217410fbd66f7a7a8a64d9d85e61229)
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/_bitset.h>
50 #include <sys/bitset.h>
51 #include <sys/bus.h>
52 #include <sys/conf.h>
53 #include <sys/endian.h>
54 #include <sys/fcntl.h>
55 #include <sys/filio.h>
56 #include <sys/hash.h>
57 #include <sys/interrupt.h>
58 #include <sys/jail.h>
59 #include <sys/kernel.h>
60 #include <sys/kthread.h>
61 #include <sys/lock.h>
62 #include <sys/mbuf.h>
63 #include <sys/module.h>
64 #include <sys/proc.h>
65 #include <sys/smp.h>
66 #include <sys/socket.h>
67 #include <sys/sysctl.h>
68 #include <sys/md5.h>
69 #include <sys/ucred.h>
70 
71 #include <net/if.h>
72 #include <net/if_var.h>
73 #include <net/vnet.h>
74 #include <net/route.h>
75 #include <net/pfil.h>
76 #include <net/pfvar.h>
77 #include <net/if_pfsync.h>
78 #include <net/if_pflog.h>
79 
80 #include <netinet/in.h>
81 #include <netinet/ip.h>
82 #include <netinet/ip_var.h>
83 #include <netinet6/ip6_var.h>
84 #include <netinet/ip_icmp.h>
85 
86 #ifdef INET6
87 #include <netinet/ip6.h>
88 #endif /* INET6 */
89 
90 #ifdef ALTQ
91 #include <net/altq/altq.h>
92 #endif
93 
94 static struct pf_kpool	*pf_get_kpool(char *, u_int32_t, u_int8_t, u_int32_t,
95 			    u_int8_t, u_int8_t, u_int8_t);
96 
97 static void		 pf_mv_kpool(struct pf_kpalist *, struct pf_kpalist *);
98 static void		 pf_empty_kpool(struct pf_kpalist *);
99 static int		 pfioctl(struct cdev *, u_long, caddr_t, int,
100 			    struct thread *);
101 #ifdef ALTQ
102 static int		 pf_begin_altq(u_int32_t *);
103 static int		 pf_rollback_altq(u_int32_t);
104 static int		 pf_commit_altq(u_int32_t);
105 static int		 pf_enable_altq(struct pf_altq *);
106 static int		 pf_disable_altq(struct pf_altq *);
107 static u_int32_t	 pf_qname2qid(char *);
108 static void		 pf_qid_unref(u_int32_t);
109 #endif /* ALTQ */
110 static int		 pf_begin_rules(u_int32_t *, int, const char *);
111 static int		 pf_rollback_rules(u_int32_t, int, char *);
112 static int		 pf_setup_pfsync_matching(struct pf_kruleset *);
113 static void		 pf_hash_rule(MD5_CTX *, struct pf_krule *);
114 static void		 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
115 static int		 pf_commit_rules(u_int32_t, int, char *);
116 static int		 pf_addr_setup(struct pf_kruleset *,
117 			    struct pf_addr_wrap *, sa_family_t);
118 static void		 pf_addr_copyout(struct pf_addr_wrap *);
119 static void		 pf_src_node_copy(const struct pf_ksrc_node *,
120 			    struct pf_src_node *);
121 #ifdef ALTQ
122 static int		 pf_export_kaltq(struct pf_altq *,
123 			    struct pfioc_altq_v1 *, size_t);
124 static int		 pf_import_kaltq(struct pfioc_altq_v1 *,
125 			    struct pf_altq *, size_t);
126 #endif /* ALTQ */
127 
128 VNET_DEFINE(struct pf_krule,	pf_default_rule);
129 
130 #ifdef ALTQ
131 VNET_DEFINE_STATIC(int,		pf_altq_running);
132 #define	V_pf_altq_running	VNET(pf_altq_running)
133 #endif
134 
135 #define	TAGID_MAX	 50000
136 struct pf_tagname {
137 	TAILQ_ENTRY(pf_tagname)	namehash_entries;
138 	TAILQ_ENTRY(pf_tagname)	taghash_entries;
139 	char			name[PF_TAG_NAME_SIZE];
140 	uint16_t		tag;
141 	int			ref;
142 };
143 
144 struct pf_tagset {
145 	TAILQ_HEAD(, pf_tagname)	*namehash;
146 	TAILQ_HEAD(, pf_tagname)	*taghash;
147 	unsigned int			 mask;
148 	uint32_t			 seed;
149 	BITSET_DEFINE(, TAGID_MAX)	 avail;
150 };
151 
152 VNET_DEFINE(struct pf_tagset, pf_tags);
153 #define	V_pf_tags	VNET(pf_tags)
154 static unsigned int	pf_rule_tag_hashsize;
155 #define	PF_RULE_TAG_HASH_SIZE_DEFAULT	128
156 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN,
157     &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT,
158     "Size of pf(4) rule tag hashtable");
159 
160 #ifdef ALTQ
161 VNET_DEFINE(struct pf_tagset, pf_qids);
162 #define	V_pf_qids	VNET(pf_qids)
163 static unsigned int	pf_queue_tag_hashsize;
164 #define	PF_QUEUE_TAG_HASH_SIZE_DEFAULT	128
165 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN,
166     &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT,
167     "Size of pf(4) queue tag hashtable");
168 #endif
169 VNET_DEFINE(uma_zone_t,	 pf_tag_z);
170 #define	V_pf_tag_z		 VNET(pf_tag_z)
171 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
172 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
173 
174 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
175 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
176 #endif
177 
178 static void		 pf_init_tagset(struct pf_tagset *, unsigned int *,
179 			    unsigned int);
180 static void		 pf_cleanup_tagset(struct pf_tagset *);
181 static uint16_t		 tagname2hashindex(const struct pf_tagset *, const char *);
182 static uint16_t		 tag2hashindex(const struct pf_tagset *, uint16_t);
183 static u_int16_t	 tagname2tag(struct pf_tagset *, char *);
184 static u_int16_t	 pf_tagname2tag(char *);
185 static void		 tag_unref(struct pf_tagset *, u_int16_t);
186 
187 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
188 
189 struct cdev *pf_dev;
190 
191 /*
192  * XXX - These are new and need to be checked when moveing to a new version
193  */
194 static void		 pf_clear_states(void);
195 static int		 pf_clear_tables(void);
196 static void		 pf_clear_srcnodes(struct pf_ksrc_node *);
197 static void		 pf_kill_srcnodes(struct pfioc_src_node_kill *);
198 static void		 pf_tbladdr_copyout(struct pf_addr_wrap *);
199 
200 /*
201  * Wrapper functions for pfil(9) hooks
202  */
203 #ifdef INET
204 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp,
205     int flags, void *ruleset __unused, struct inpcb *inp);
206 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp,
207     int flags, void *ruleset __unused, struct inpcb *inp);
208 #endif
209 #ifdef INET6
210 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp,
211     int flags, void *ruleset __unused, struct inpcb *inp);
212 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp,
213     int flags, void *ruleset __unused, struct inpcb *inp);
214 #endif
215 
216 static int		hook_pf(void);
217 static int		dehook_pf(void);
218 static int		shutdown_pf(void);
219 static int		pf_load(void);
220 static void		pf_unload(void);
221 
222 static struct cdevsw pf_cdevsw = {
223 	.d_ioctl =	pfioctl,
224 	.d_name =	PF_NAME,
225 	.d_version =	D_VERSION,
226 };
227 
228 volatile VNET_DEFINE_STATIC(int, pf_pfil_hooked);
229 #define V_pf_pfil_hooked	VNET(pf_pfil_hooked)
230 
231 /*
232  * We need a flag that is neither hooked nor running to know when
233  * the VNET is "valid".  We primarily need this to control (global)
234  * external event, e.g., eventhandlers.
235  */
236 VNET_DEFINE(int, pf_vnet_active);
237 #define V_pf_vnet_active	VNET(pf_vnet_active)
238 
239 int pf_end_threads;
240 struct proc *pf_purge_proc;
241 
242 struct rmlock			pf_rules_lock;
243 struct sx			pf_ioctl_lock;
244 struct sx			pf_end_lock;
245 
246 /* pfsync */
247 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
248 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
249 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
250 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
251 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
252 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
253 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
254 
255 /* pflog */
256 pflog_packet_t			*pflog_packet_ptr = NULL;
257 
258 extern u_long	pf_ioctl_maxcount;
259 
260 static void
261 pfattach_vnet(void)
262 {
263 	u_int32_t *my_timeout = V_pf_default_rule.timeout;
264 
265 	pf_initialize();
266 	pfr_initialize();
267 	pfi_initialize_vnet();
268 	pf_normalize_init();
269 
270 	V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
271 	V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
272 
273 	RB_INIT(&V_pf_anchors);
274 	pf_init_kruleset(&pf_main_ruleset);
275 
276 	/* default rule should never be garbage collected */
277 	V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
278 #ifdef PF_DEFAULT_TO_DROP
279 	V_pf_default_rule.action = PF_DROP;
280 #else
281 	V_pf_default_rule.action = PF_PASS;
282 #endif
283 	V_pf_default_rule.nr = -1;
284 	V_pf_default_rule.rtableid = -1;
285 
286 	V_pf_default_rule.evaluations = counter_u64_alloc(M_WAITOK);
287 	for (int i = 0; i < 2; i++) {
288 		V_pf_default_rule.packets[i] = counter_u64_alloc(M_WAITOK);
289 		V_pf_default_rule.bytes[i] = counter_u64_alloc(M_WAITOK);
290 	}
291 	V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
292 	V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
293 	V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
294 
295 	/* initialize default timeouts */
296 	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
297 	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
298 	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
299 	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
300 	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
301 	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
302 	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
303 	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
304 	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
305 	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
306 	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
307 	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
308 	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
309 	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
310 	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
311 	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
312 	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
313 	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
314 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
315 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
316 
317 	bzero(&V_pf_status, sizeof(V_pf_status));
318 	V_pf_status.debug = PF_DEBUG_URGENT;
319 
320 	V_pf_pfil_hooked = 0;
321 
322 	/* XXX do our best to avoid a conflict */
323 	V_pf_status.hostid = arc4random();
324 
325 	for (int i = 0; i < PFRES_MAX; i++)
326 		V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
327 	for (int i = 0; i < LCNT_MAX; i++)
328 		V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
329 	for (int i = 0; i < FCNT_MAX; i++)
330 		V_pf_status.fcounters[i] = counter_u64_alloc(M_WAITOK);
331 	for (int i = 0; i < SCNT_MAX; i++)
332 		V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
333 
334 	if (swi_add(NULL, "pf send", pf_intr, curvnet, SWI_NET,
335 	    INTR_MPSAFE, &V_pf_swi_cookie) != 0)
336 		/* XXXGL: leaked all above. */
337 		return;
338 }
339 
340 static struct pf_kpool *
341 pf_get_kpool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
342     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
343     u_int8_t check_ticket)
344 {
345 	struct pf_kruleset	*ruleset;
346 	struct pf_krule		*rule;
347 	int			 rs_num;
348 
349 	ruleset = pf_find_kruleset(anchor);
350 	if (ruleset == NULL)
351 		return (NULL);
352 	rs_num = pf_get_ruleset_number(rule_action);
353 	if (rs_num >= PF_RULESET_MAX)
354 		return (NULL);
355 	if (active) {
356 		if (check_ticket && ticket !=
357 		    ruleset->rules[rs_num].active.ticket)
358 			return (NULL);
359 		if (r_last)
360 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
361 			    pf_krulequeue);
362 		else
363 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
364 	} else {
365 		if (check_ticket && ticket !=
366 		    ruleset->rules[rs_num].inactive.ticket)
367 			return (NULL);
368 		if (r_last)
369 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
370 			    pf_krulequeue);
371 		else
372 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
373 	}
374 	if (!r_last) {
375 		while ((rule != NULL) && (rule->nr != rule_number))
376 			rule = TAILQ_NEXT(rule, entries);
377 	}
378 	if (rule == NULL)
379 		return (NULL);
380 
381 	return (&rule->rpool);
382 }
383 
384 static void
385 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb)
386 {
387 	struct pf_kpooladdr	*mv_pool_pa;
388 
389 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
390 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
391 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
392 	}
393 }
394 
395 static void
396 pf_empty_kpool(struct pf_kpalist *poola)
397 {
398 	struct pf_kpooladdr *pa;
399 
400 	while ((pa = TAILQ_FIRST(poola)) != NULL) {
401 		switch (pa->addr.type) {
402 		case PF_ADDR_DYNIFTL:
403 			pfi_dynaddr_remove(pa->addr.p.dyn);
404 			break;
405 		case PF_ADDR_TABLE:
406 			/* XXX: this could be unfinished pooladdr on pabuf */
407 			if (pa->addr.p.tbl != NULL)
408 				pfr_detach_table(pa->addr.p.tbl);
409 			break;
410 		}
411 		if (pa->kif)
412 			pfi_kkif_unref(pa->kif);
413 		TAILQ_REMOVE(poola, pa, entries);
414 		free(pa, M_PFRULE);
415 	}
416 }
417 
418 static void
419 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
420 {
421 
422 	PF_RULES_WASSERT();
423 
424 	TAILQ_REMOVE(rulequeue, rule, entries);
425 
426 	PF_UNLNKDRULES_LOCK();
427 	rule->rule_flag |= PFRULE_REFS;
428 	TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
429 	PF_UNLNKDRULES_UNLOCK();
430 }
431 
432 void
433 pf_free_rule(struct pf_krule *rule)
434 {
435 
436 	PF_RULES_WASSERT();
437 
438 	if (rule->tag)
439 		tag_unref(&V_pf_tags, rule->tag);
440 	if (rule->match_tag)
441 		tag_unref(&V_pf_tags, rule->match_tag);
442 #ifdef ALTQ
443 	if (rule->pqid != rule->qid)
444 		pf_qid_unref(rule->pqid);
445 	pf_qid_unref(rule->qid);
446 #endif
447 	switch (rule->src.addr.type) {
448 	case PF_ADDR_DYNIFTL:
449 		pfi_dynaddr_remove(rule->src.addr.p.dyn);
450 		break;
451 	case PF_ADDR_TABLE:
452 		pfr_detach_table(rule->src.addr.p.tbl);
453 		break;
454 	}
455 	switch (rule->dst.addr.type) {
456 	case PF_ADDR_DYNIFTL:
457 		pfi_dynaddr_remove(rule->dst.addr.p.dyn);
458 		break;
459 	case PF_ADDR_TABLE:
460 		pfr_detach_table(rule->dst.addr.p.tbl);
461 		break;
462 	}
463 	if (rule->overload_tbl)
464 		pfr_detach_table(rule->overload_tbl);
465 	if (rule->kif)
466 		pfi_kkif_unref(rule->kif);
467 	pf_kanchor_remove(rule);
468 	pf_empty_kpool(&rule->rpool.list);
469 	counter_u64_free(rule->evaluations);
470 	for (int i = 0; i < 2; i++) {
471 		counter_u64_free(rule->packets[i]);
472 		counter_u64_free(rule->bytes[i]);
473 	}
474 	counter_u64_free(rule->states_cur);
475 	counter_u64_free(rule->states_tot);
476 	counter_u64_free(rule->src_nodes);
477 	free(rule, M_PFRULE);
478 }
479 
480 static void
481 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
482     unsigned int default_size)
483 {
484 	unsigned int i;
485 	unsigned int hashsize;
486 
487 	if (*tunable_size == 0 || !powerof2(*tunable_size))
488 		*tunable_size = default_size;
489 
490 	hashsize = *tunable_size;
491 	ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
492 	    M_WAITOK);
493 	ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
494 	    M_WAITOK);
495 	ts->mask = hashsize - 1;
496 	ts->seed = arc4random();
497 	for (i = 0; i < hashsize; i++) {
498 		TAILQ_INIT(&ts->namehash[i]);
499 		TAILQ_INIT(&ts->taghash[i]);
500 	}
501 	BIT_FILL(TAGID_MAX, &ts->avail);
502 }
503 
504 static void
505 pf_cleanup_tagset(struct pf_tagset *ts)
506 {
507 	unsigned int i;
508 	unsigned int hashsize;
509 	struct pf_tagname *t, *tmp;
510 
511 	/*
512 	 * Only need to clean up one of the hashes as each tag is hashed
513 	 * into each table.
514 	 */
515 	hashsize = ts->mask + 1;
516 	for (i = 0; i < hashsize; i++)
517 		TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
518 			uma_zfree(V_pf_tag_z, t);
519 
520 	free(ts->namehash, M_PFHASH);
521 	free(ts->taghash, M_PFHASH);
522 }
523 
524 static uint16_t
525 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
526 {
527 	size_t len;
528 
529 	len = strnlen(tagname, PF_TAG_NAME_SIZE - 1);
530 	return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask);
531 }
532 
533 static uint16_t
534 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
535 {
536 
537 	return (tag & ts->mask);
538 }
539 
540 static u_int16_t
541 tagname2tag(struct pf_tagset *ts, char *tagname)
542 {
543 	struct pf_tagname	*tag;
544 	u_int32_t		 index;
545 	u_int16_t		 new_tagid;
546 
547 	PF_RULES_WASSERT();
548 
549 	index = tagname2hashindex(ts, tagname);
550 	TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
551 		if (strcmp(tagname, tag->name) == 0) {
552 			tag->ref++;
553 			return (tag->tag);
554 		}
555 
556 	/*
557 	 * new entry
558 	 *
559 	 * to avoid fragmentation, we do a linear search from the beginning
560 	 * and take the first free slot we find.
561 	 */
562 	new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
563 	/*
564 	 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
565 	 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
566 	 * set.  It may also return a bit number greater than TAGID_MAX due
567 	 * to rounding of the number of bits in the vector up to a multiple
568 	 * of the vector word size at declaration/allocation time.
569 	 */
570 	if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
571 		return (0);
572 
573 	/* Mark the tag as in use.  Bits are 0-based for BIT_CLR() */
574 	BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
575 
576 	/* allocate and fill new struct pf_tagname */
577 	tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
578 	if (tag == NULL)
579 		return (0);
580 	strlcpy(tag->name, tagname, sizeof(tag->name));
581 	tag->tag = new_tagid;
582 	tag->ref = 1;
583 
584 	/* Insert into namehash */
585 	TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
586 
587 	/* Insert into taghash */
588 	index = tag2hashindex(ts, new_tagid);
589 	TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
590 
591 	return (tag->tag);
592 }
593 
594 static void
595 tag_unref(struct pf_tagset *ts, u_int16_t tag)
596 {
597 	struct pf_tagname	*t;
598 	uint16_t		 index;
599 
600 	PF_RULES_WASSERT();
601 
602 	index = tag2hashindex(ts, tag);
603 	TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
604 		if (tag == t->tag) {
605 			if (--t->ref == 0) {
606 				TAILQ_REMOVE(&ts->taghash[index], t,
607 				    taghash_entries);
608 				index = tagname2hashindex(ts, t->name);
609 				TAILQ_REMOVE(&ts->namehash[index], t,
610 				    namehash_entries);
611 				/* Bits are 0-based for BIT_SET() */
612 				BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
613 				uma_zfree(V_pf_tag_z, t);
614 			}
615 			break;
616 		}
617 }
618 
619 static u_int16_t
620 pf_tagname2tag(char *tagname)
621 {
622 	return (tagname2tag(&V_pf_tags, tagname));
623 }
624 
625 #ifdef ALTQ
626 static u_int32_t
627 pf_qname2qid(char *qname)
628 {
629 	return ((u_int32_t)tagname2tag(&V_pf_qids, qname));
630 }
631 
632 static void
633 pf_qid_unref(u_int32_t qid)
634 {
635 	tag_unref(&V_pf_qids, (u_int16_t)qid);
636 }
637 
638 static int
639 pf_begin_altq(u_int32_t *ticket)
640 {
641 	struct pf_altq	*altq, *tmp;
642 	int		 error = 0;
643 
644 	PF_RULES_WASSERT();
645 
646 	/* Purge the old altq lists */
647 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
648 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
649 			/* detach and destroy the discipline */
650 			error = altq_remove(altq);
651 		}
652 		free(altq, M_PFALTQ);
653 	}
654 	TAILQ_INIT(V_pf_altq_ifs_inactive);
655 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
656 		pf_qid_unref(altq->qid);
657 		free(altq, M_PFALTQ);
658 	}
659 	TAILQ_INIT(V_pf_altqs_inactive);
660 	if (error)
661 		return (error);
662 	*ticket = ++V_ticket_altqs_inactive;
663 	V_altqs_inactive_open = 1;
664 	return (0);
665 }
666 
667 static int
668 pf_rollback_altq(u_int32_t ticket)
669 {
670 	struct pf_altq	*altq, *tmp;
671 	int		 error = 0;
672 
673 	PF_RULES_WASSERT();
674 
675 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
676 		return (0);
677 	/* Purge the old altq lists */
678 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
679 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
680 			/* detach and destroy the discipline */
681 			error = altq_remove(altq);
682 		}
683 		free(altq, M_PFALTQ);
684 	}
685 	TAILQ_INIT(V_pf_altq_ifs_inactive);
686 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
687 		pf_qid_unref(altq->qid);
688 		free(altq, M_PFALTQ);
689 	}
690 	TAILQ_INIT(V_pf_altqs_inactive);
691 	V_altqs_inactive_open = 0;
692 	return (error);
693 }
694 
695 static int
696 pf_commit_altq(u_int32_t ticket)
697 {
698 	struct pf_altqqueue	*old_altqs, *old_altq_ifs;
699 	struct pf_altq		*altq, *tmp;
700 	int			 err, error = 0;
701 
702 	PF_RULES_WASSERT();
703 
704 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
705 		return (EBUSY);
706 
707 	/* swap altqs, keep the old. */
708 	old_altqs = V_pf_altqs_active;
709 	old_altq_ifs = V_pf_altq_ifs_active;
710 	V_pf_altqs_active = V_pf_altqs_inactive;
711 	V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
712 	V_pf_altqs_inactive = old_altqs;
713 	V_pf_altq_ifs_inactive = old_altq_ifs;
714 	V_ticket_altqs_active = V_ticket_altqs_inactive;
715 
716 	/* Attach new disciplines */
717 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
718 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
719 			/* attach the discipline */
720 			error = altq_pfattach(altq);
721 			if (error == 0 && V_pf_altq_running)
722 				error = pf_enable_altq(altq);
723 			if (error != 0)
724 				return (error);
725 		}
726 	}
727 
728 	/* Purge the old altq lists */
729 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
730 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
731 			/* detach and destroy the discipline */
732 			if (V_pf_altq_running)
733 				error = pf_disable_altq(altq);
734 			err = altq_pfdetach(altq);
735 			if (err != 0 && error == 0)
736 				error = err;
737 			err = altq_remove(altq);
738 			if (err != 0 && error == 0)
739 				error = err;
740 		}
741 		free(altq, M_PFALTQ);
742 	}
743 	TAILQ_INIT(V_pf_altq_ifs_inactive);
744 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
745 		pf_qid_unref(altq->qid);
746 		free(altq, M_PFALTQ);
747 	}
748 	TAILQ_INIT(V_pf_altqs_inactive);
749 
750 	V_altqs_inactive_open = 0;
751 	return (error);
752 }
753 
754 static int
755 pf_enable_altq(struct pf_altq *altq)
756 {
757 	struct ifnet		*ifp;
758 	struct tb_profile	 tb;
759 	int			 error = 0;
760 
761 	if ((ifp = ifunit(altq->ifname)) == NULL)
762 		return (EINVAL);
763 
764 	if (ifp->if_snd.altq_type != ALTQT_NONE)
765 		error = altq_enable(&ifp->if_snd);
766 
767 	/* set tokenbucket regulator */
768 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
769 		tb.rate = altq->ifbandwidth;
770 		tb.depth = altq->tbrsize;
771 		error = tbr_set(&ifp->if_snd, &tb);
772 	}
773 
774 	return (error);
775 }
776 
777 static int
778 pf_disable_altq(struct pf_altq *altq)
779 {
780 	struct ifnet		*ifp;
781 	struct tb_profile	 tb;
782 	int			 error;
783 
784 	if ((ifp = ifunit(altq->ifname)) == NULL)
785 		return (EINVAL);
786 
787 	/*
788 	 * when the discipline is no longer referenced, it was overridden
789 	 * by a new one.  if so, just return.
790 	 */
791 	if (altq->altq_disc != ifp->if_snd.altq_disc)
792 		return (0);
793 
794 	error = altq_disable(&ifp->if_snd);
795 
796 	if (error == 0) {
797 		/* clear tokenbucket regulator */
798 		tb.rate = 0;
799 		error = tbr_set(&ifp->if_snd, &tb);
800 	}
801 
802 	return (error);
803 }
804 
805 static int
806 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
807     struct pf_altq *altq)
808 {
809 	struct ifnet	*ifp1;
810 	int		 error = 0;
811 
812 	/* Deactivate the interface in question */
813 	altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
814 	if ((ifp1 = ifunit(altq->ifname)) == NULL ||
815 	    (remove && ifp1 == ifp)) {
816 		altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
817 	} else {
818 		error = altq_add(ifp1, altq);
819 
820 		if (ticket != V_ticket_altqs_inactive)
821 			error = EBUSY;
822 
823 		if (error)
824 			free(altq, M_PFALTQ);
825 	}
826 
827 	return (error);
828 }
829 
830 void
831 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
832 {
833 	struct pf_altq	*a1, *a2, *a3;
834 	u_int32_t	 ticket;
835 	int		 error = 0;
836 
837 	/*
838 	 * No need to re-evaluate the configuration for events on interfaces
839 	 * that do not support ALTQ, as it's not possible for such
840 	 * interfaces to be part of the configuration.
841 	 */
842 	if (!ALTQ_IS_READY(&ifp->if_snd))
843 		return;
844 
845 	/* Interrupt userland queue modifications */
846 	if (V_altqs_inactive_open)
847 		pf_rollback_altq(V_ticket_altqs_inactive);
848 
849 	/* Start new altq ruleset */
850 	if (pf_begin_altq(&ticket))
851 		return;
852 
853 	/* Copy the current active set */
854 	TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
855 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
856 		if (a2 == NULL) {
857 			error = ENOMEM;
858 			break;
859 		}
860 		bcopy(a1, a2, sizeof(struct pf_altq));
861 
862 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
863 		if (error)
864 			break;
865 
866 		TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
867 	}
868 	if (error)
869 		goto out;
870 	TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
871 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
872 		if (a2 == NULL) {
873 			error = ENOMEM;
874 			break;
875 		}
876 		bcopy(a1, a2, sizeof(struct pf_altq));
877 
878 		if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
879 			error = EBUSY;
880 			free(a2, M_PFALTQ);
881 			break;
882 		}
883 		a2->altq_disc = NULL;
884 		TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
885 			if (strncmp(a3->ifname, a2->ifname,
886 				IFNAMSIZ) == 0) {
887 				a2->altq_disc = a3->altq_disc;
888 				break;
889 			}
890 		}
891 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
892 		if (error)
893 			break;
894 
895 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
896 	}
897 
898 out:
899 	if (error != 0)
900 		pf_rollback_altq(ticket);
901 	else
902 		pf_commit_altq(ticket);
903 }
904 #endif /* ALTQ */
905 
906 static int
907 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
908 {
909 	struct pf_kruleset	*rs;
910 	struct pf_krule		*rule;
911 
912 	PF_RULES_WASSERT();
913 
914 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
915 		return (EINVAL);
916 	rs = pf_find_or_create_kruleset(anchor);
917 	if (rs == NULL)
918 		return (EINVAL);
919 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
920 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
921 		rs->rules[rs_num].inactive.rcount--;
922 	}
923 	*ticket = ++rs->rules[rs_num].inactive.ticket;
924 	rs->rules[rs_num].inactive.open = 1;
925 	return (0);
926 }
927 
928 static int
929 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
930 {
931 	struct pf_kruleset	*rs;
932 	struct pf_krule		*rule;
933 
934 	PF_RULES_WASSERT();
935 
936 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
937 		return (EINVAL);
938 	rs = pf_find_kruleset(anchor);
939 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
940 	    rs->rules[rs_num].inactive.ticket != ticket)
941 		return (0);
942 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
943 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
944 		rs->rules[rs_num].inactive.rcount--;
945 	}
946 	rs->rules[rs_num].inactive.open = 0;
947 	return (0);
948 }
949 
950 #define PF_MD5_UPD(st, elm)						\
951 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
952 
953 #define PF_MD5_UPD_STR(st, elm)						\
954 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
955 
956 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
957 		(stor) = htonl((st)->elm);				\
958 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
959 } while (0)
960 
961 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
962 		(stor) = htons((st)->elm);				\
963 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
964 } while (0)
965 
966 static void
967 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
968 {
969 	PF_MD5_UPD(pfr, addr.type);
970 	switch (pfr->addr.type) {
971 		case PF_ADDR_DYNIFTL:
972 			PF_MD5_UPD(pfr, addr.v.ifname);
973 			PF_MD5_UPD(pfr, addr.iflags);
974 			break;
975 		case PF_ADDR_TABLE:
976 			PF_MD5_UPD(pfr, addr.v.tblname);
977 			break;
978 		case PF_ADDR_ADDRMASK:
979 			/* XXX ignore af? */
980 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
981 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
982 			break;
983 	}
984 
985 	PF_MD5_UPD(pfr, port[0]);
986 	PF_MD5_UPD(pfr, port[1]);
987 	PF_MD5_UPD(pfr, neg);
988 	PF_MD5_UPD(pfr, port_op);
989 }
990 
991 static void
992 pf_hash_rule(MD5_CTX *ctx, struct pf_krule *rule)
993 {
994 	u_int16_t x;
995 	u_int32_t y;
996 
997 	pf_hash_rule_addr(ctx, &rule->src);
998 	pf_hash_rule_addr(ctx, &rule->dst);
999 	PF_MD5_UPD_STR(rule, label);
1000 	PF_MD5_UPD_STR(rule, ifname);
1001 	PF_MD5_UPD_STR(rule, match_tagname);
1002 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
1003 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
1004 	PF_MD5_UPD_HTONL(rule, prob, y);
1005 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
1006 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1007 	PF_MD5_UPD(rule, uid.op);
1008 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1009 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1010 	PF_MD5_UPD(rule, gid.op);
1011 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
1012 	PF_MD5_UPD(rule, action);
1013 	PF_MD5_UPD(rule, direction);
1014 	PF_MD5_UPD(rule, af);
1015 	PF_MD5_UPD(rule, quick);
1016 	PF_MD5_UPD(rule, ifnot);
1017 	PF_MD5_UPD(rule, match_tag_not);
1018 	PF_MD5_UPD(rule, natpass);
1019 	PF_MD5_UPD(rule, keep_state);
1020 	PF_MD5_UPD(rule, proto);
1021 	PF_MD5_UPD(rule, type);
1022 	PF_MD5_UPD(rule, code);
1023 	PF_MD5_UPD(rule, flags);
1024 	PF_MD5_UPD(rule, flagset);
1025 	PF_MD5_UPD(rule, allow_opts);
1026 	PF_MD5_UPD(rule, rt);
1027 	PF_MD5_UPD(rule, tos);
1028 }
1029 
1030 static int
1031 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1032 {
1033 	struct pf_kruleset	*rs;
1034 	struct pf_krule		*rule, **old_array;
1035 	struct pf_krulequeue	*old_rules;
1036 	int			 error;
1037 	u_int32_t		 old_rcount;
1038 
1039 	PF_RULES_WASSERT();
1040 
1041 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1042 		return (EINVAL);
1043 	rs = pf_find_kruleset(anchor);
1044 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1045 	    ticket != rs->rules[rs_num].inactive.ticket)
1046 		return (EBUSY);
1047 
1048 	/* Calculate checksum for the main ruleset */
1049 	if (rs == &pf_main_ruleset) {
1050 		error = pf_setup_pfsync_matching(rs);
1051 		if (error != 0)
1052 			return (error);
1053 	}
1054 
1055 	/* Swap rules, keep the old. */
1056 	old_rules = rs->rules[rs_num].active.ptr;
1057 	old_rcount = rs->rules[rs_num].active.rcount;
1058 	old_array = rs->rules[rs_num].active.ptr_array;
1059 
1060 	rs->rules[rs_num].active.ptr =
1061 	    rs->rules[rs_num].inactive.ptr;
1062 	rs->rules[rs_num].active.ptr_array =
1063 	    rs->rules[rs_num].inactive.ptr_array;
1064 	rs->rules[rs_num].active.rcount =
1065 	    rs->rules[rs_num].inactive.rcount;
1066 	rs->rules[rs_num].inactive.ptr = old_rules;
1067 	rs->rules[rs_num].inactive.ptr_array = old_array;
1068 	rs->rules[rs_num].inactive.rcount = old_rcount;
1069 
1070 	rs->rules[rs_num].active.ticket =
1071 	    rs->rules[rs_num].inactive.ticket;
1072 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1073 
1074 	/* Purge the old rule list. */
1075 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1076 		pf_unlink_rule(old_rules, rule);
1077 	if (rs->rules[rs_num].inactive.ptr_array)
1078 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1079 	rs->rules[rs_num].inactive.ptr_array = NULL;
1080 	rs->rules[rs_num].inactive.rcount = 0;
1081 	rs->rules[rs_num].inactive.open = 0;
1082 	pf_remove_if_empty_kruleset(rs);
1083 
1084 	return (0);
1085 }
1086 
1087 static int
1088 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1089 {
1090 	MD5_CTX			 ctx;
1091 	struct pf_krule		*rule;
1092 	int			 rs_cnt;
1093 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
1094 
1095 	MD5Init(&ctx);
1096 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1097 		/* XXX PF_RULESET_SCRUB as well? */
1098 		if (rs_cnt == PF_RULESET_SCRUB)
1099 			continue;
1100 
1101 		if (rs->rules[rs_cnt].inactive.ptr_array)
1102 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1103 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
1104 
1105 		if (rs->rules[rs_cnt].inactive.rcount) {
1106 			rs->rules[rs_cnt].inactive.ptr_array =
1107 			    malloc(sizeof(caddr_t) *
1108 			    rs->rules[rs_cnt].inactive.rcount,
1109 			    M_TEMP, M_NOWAIT);
1110 
1111 			if (!rs->rules[rs_cnt].inactive.ptr_array)
1112 				return (ENOMEM);
1113 		}
1114 
1115 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1116 		    entries) {
1117 			pf_hash_rule(&ctx, rule);
1118 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1119 		}
1120 	}
1121 
1122 	MD5Final(digest, &ctx);
1123 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1124 	return (0);
1125 }
1126 
1127 static int
1128 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1129     sa_family_t af)
1130 {
1131 	int error = 0;
1132 
1133 	switch (addr->type) {
1134 	case PF_ADDR_TABLE:
1135 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1136 		if (addr->p.tbl == NULL)
1137 			error = ENOMEM;
1138 		break;
1139 	case PF_ADDR_DYNIFTL:
1140 		error = pfi_dynaddr_setup(addr, af);
1141 		break;
1142 	}
1143 
1144 	return (error);
1145 }
1146 
1147 static void
1148 pf_addr_copyout(struct pf_addr_wrap *addr)
1149 {
1150 
1151 	switch (addr->type) {
1152 	case PF_ADDR_DYNIFTL:
1153 		pfi_dynaddr_copyout(addr);
1154 		break;
1155 	case PF_ADDR_TABLE:
1156 		pf_tbladdr_copyout(addr);
1157 		break;
1158 	}
1159 }
1160 
1161 static void
1162 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
1163 {
1164 	int	secs = time_uptime, diff;
1165 
1166 	bzero(out, sizeof(struct pf_src_node));
1167 
1168 	bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
1169 	bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
1170 
1171 	if (in->rule.ptr != NULL)
1172 		out->rule.nr = in->rule.ptr->nr;
1173 
1174 	for (int i = 0; i < 2; i++) {
1175 		out->bytes[i] = counter_u64_fetch(in->bytes[i]);
1176 		out->packets[i] = counter_u64_fetch(in->packets[i]);
1177 	}
1178 
1179 	out->states = in->states;
1180 	out->conn = in->conn;
1181 	out->af = in->af;
1182 	out->ruletype = in->ruletype;
1183 
1184 	out->creation = secs - in->creation;
1185 	if (out->expire > secs)
1186 		out->expire -= secs;
1187 	else
1188 		out->expire = 0;
1189 
1190 	/* Adjust the connection rate estimate. */
1191 	diff = secs - in->conn_rate.last;
1192 	if (diff >= in->conn_rate.seconds)
1193 		out->conn_rate.count = 0;
1194 	else
1195 		out->conn_rate.count -=
1196 		    in->conn_rate.count * diff /
1197 		    in->conn_rate.seconds;
1198 }
1199 
1200 #ifdef ALTQ
1201 /*
1202  * Handle export of struct pf_kaltq to user binaries that may be using any
1203  * version of struct pf_altq.
1204  */
1205 static int
1206 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1207 {
1208 	u_int32_t version;
1209 
1210 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1211 		version = 0;
1212 	else
1213 		version = pa->version;
1214 
1215 	if (version > PFIOC_ALTQ_VERSION)
1216 		return (EINVAL);
1217 
1218 #define ASSIGN(x) exported_q->x = q->x
1219 #define COPY(x) \
1220 	bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1221 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1222 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1223 
1224 	switch (version) {
1225 	case 0: {
1226 		struct pf_altq_v0 *exported_q =
1227 		    &((struct pfioc_altq_v0 *)pa)->altq;
1228 
1229 		COPY(ifname);
1230 
1231 		ASSIGN(scheduler);
1232 		ASSIGN(tbrsize);
1233 		exported_q->tbrsize = SATU16(q->tbrsize);
1234 		exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1235 
1236 		COPY(qname);
1237 		COPY(parent);
1238 		ASSIGN(parent_qid);
1239 		exported_q->bandwidth = SATU32(q->bandwidth);
1240 		ASSIGN(priority);
1241 		ASSIGN(local_flags);
1242 
1243 		ASSIGN(qlimit);
1244 		ASSIGN(flags);
1245 
1246 		if (q->scheduler == ALTQT_HFSC) {
1247 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1248 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1249 			    SATU32(q->pq_u.hfsc_opts.x)
1250 
1251 			ASSIGN_OPT_SATU32(rtsc_m1);
1252 			ASSIGN_OPT(rtsc_d);
1253 			ASSIGN_OPT_SATU32(rtsc_m2);
1254 
1255 			ASSIGN_OPT_SATU32(lssc_m1);
1256 			ASSIGN_OPT(lssc_d);
1257 			ASSIGN_OPT_SATU32(lssc_m2);
1258 
1259 			ASSIGN_OPT_SATU32(ulsc_m1);
1260 			ASSIGN_OPT(ulsc_d);
1261 			ASSIGN_OPT_SATU32(ulsc_m2);
1262 
1263 			ASSIGN_OPT(flags);
1264 
1265 #undef ASSIGN_OPT
1266 #undef ASSIGN_OPT_SATU32
1267 		} else
1268 			COPY(pq_u);
1269 
1270 		ASSIGN(qid);
1271 		break;
1272 	}
1273 	case 1:	{
1274 		struct pf_altq_v1 *exported_q =
1275 		    &((struct pfioc_altq_v1 *)pa)->altq;
1276 
1277 		COPY(ifname);
1278 
1279 		ASSIGN(scheduler);
1280 		ASSIGN(tbrsize);
1281 		ASSIGN(ifbandwidth);
1282 
1283 		COPY(qname);
1284 		COPY(parent);
1285 		ASSIGN(parent_qid);
1286 		ASSIGN(bandwidth);
1287 		ASSIGN(priority);
1288 		ASSIGN(local_flags);
1289 
1290 		ASSIGN(qlimit);
1291 		ASSIGN(flags);
1292 		COPY(pq_u);
1293 
1294 		ASSIGN(qid);
1295 		break;
1296 	}
1297 	default:
1298 		panic("%s: unhandled struct pfioc_altq version", __func__);
1299 		break;
1300 	}
1301 
1302 #undef ASSIGN
1303 #undef COPY
1304 #undef SATU16
1305 #undef SATU32
1306 
1307 	return (0);
1308 }
1309 
1310 /*
1311  * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1312  * that may be using any version of it.
1313  */
1314 static int
1315 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1316 {
1317 	u_int32_t version;
1318 
1319 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1320 		version = 0;
1321 	else
1322 		version = pa->version;
1323 
1324 	if (version > PFIOC_ALTQ_VERSION)
1325 		return (EINVAL);
1326 
1327 #define ASSIGN(x) q->x = imported_q->x
1328 #define COPY(x) \
1329 	bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1330 
1331 	switch (version) {
1332 	case 0: {
1333 		struct pf_altq_v0 *imported_q =
1334 		    &((struct pfioc_altq_v0 *)pa)->altq;
1335 
1336 		COPY(ifname);
1337 
1338 		ASSIGN(scheduler);
1339 		ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1340 		ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1341 
1342 		COPY(qname);
1343 		COPY(parent);
1344 		ASSIGN(parent_qid);
1345 		ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1346 		ASSIGN(priority);
1347 		ASSIGN(local_flags);
1348 
1349 		ASSIGN(qlimit);
1350 		ASSIGN(flags);
1351 
1352 		if (imported_q->scheduler == ALTQT_HFSC) {
1353 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1354 
1355 			/*
1356 			 * The m1 and m2 parameters are being copied from
1357 			 * 32-bit to 64-bit.
1358 			 */
1359 			ASSIGN_OPT(rtsc_m1);
1360 			ASSIGN_OPT(rtsc_d);
1361 			ASSIGN_OPT(rtsc_m2);
1362 
1363 			ASSIGN_OPT(lssc_m1);
1364 			ASSIGN_OPT(lssc_d);
1365 			ASSIGN_OPT(lssc_m2);
1366 
1367 			ASSIGN_OPT(ulsc_m1);
1368 			ASSIGN_OPT(ulsc_d);
1369 			ASSIGN_OPT(ulsc_m2);
1370 
1371 			ASSIGN_OPT(flags);
1372 
1373 #undef ASSIGN_OPT
1374 		} else
1375 			COPY(pq_u);
1376 
1377 		ASSIGN(qid);
1378 		break;
1379 	}
1380 	case 1: {
1381 		struct pf_altq_v1 *imported_q =
1382 		    &((struct pfioc_altq_v1 *)pa)->altq;
1383 
1384 		COPY(ifname);
1385 
1386 		ASSIGN(scheduler);
1387 		ASSIGN(tbrsize);
1388 		ASSIGN(ifbandwidth);
1389 
1390 		COPY(qname);
1391 		COPY(parent);
1392 		ASSIGN(parent_qid);
1393 		ASSIGN(bandwidth);
1394 		ASSIGN(priority);
1395 		ASSIGN(local_flags);
1396 
1397 		ASSIGN(qlimit);
1398 		ASSIGN(flags);
1399 		COPY(pq_u);
1400 
1401 		ASSIGN(qid);
1402 		break;
1403 	}
1404 	default:
1405 		panic("%s: unhandled struct pfioc_altq version", __func__);
1406 		break;
1407 	}
1408 
1409 #undef ASSIGN
1410 #undef COPY
1411 
1412 	return (0);
1413 }
1414 
1415 static struct pf_altq *
1416 pf_altq_get_nth_active(u_int32_t n)
1417 {
1418 	struct pf_altq		*altq;
1419 	u_int32_t		 nr;
1420 
1421 	nr = 0;
1422 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1423 		if (nr == n)
1424 			return (altq);
1425 		nr++;
1426 	}
1427 
1428 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1429 		if (nr == n)
1430 			return (altq);
1431 		nr++;
1432 	}
1433 
1434 	return (NULL);
1435 }
1436 #endif /* ALTQ */
1437 
1438 static void
1439 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool,
1440     struct pf_pooladdr *pool)
1441 {
1442 
1443 	bzero(pool, sizeof(*pool));
1444 	bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr));
1445 	strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname));
1446 }
1447 
1448 static void
1449 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool,
1450     struct pf_kpooladdr *kpool)
1451 {
1452 
1453 	bzero(kpool, sizeof(*kpool));
1454 	bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr));
1455 	strlcpy(kpool->ifname, pool->ifname, sizeof(kpool->ifname));
1456 }
1457 
1458 static void
1459 pf_krule_to_rule(const struct pf_krule *krule, struct pf_rule *rule)
1460 {
1461 
1462 	bzero(rule, sizeof(*rule));
1463 
1464 	bcopy(&krule->src, &rule->src, sizeof(rule->src));
1465 	bcopy(&krule->dst, &rule->dst, sizeof(rule->dst));
1466 
1467 	for (int i = 0; i < PF_SKIP_COUNT; ++i) {
1468 		if (rule->skip[i].ptr == NULL)
1469 			rule->skip[i].nr = -1;
1470 		else
1471 			rule->skip[i].nr = krule->skip[i].ptr->nr;
1472 	}
1473 
1474 	strlcpy(rule->label, krule->label, sizeof(rule->label));
1475 	strlcpy(rule->ifname, krule->ifname, sizeof(rule->ifname));
1476 	strlcpy(rule->qname, krule->qname, sizeof(rule->qname));
1477 	strlcpy(rule->pqname, krule->pqname, sizeof(rule->pqname));
1478 	strlcpy(rule->tagname, krule->tagname, sizeof(rule->tagname));
1479 	strlcpy(rule->match_tagname, krule->match_tagname,
1480 	    sizeof(rule->match_tagname));
1481 	strlcpy(rule->overload_tblname, krule->overload_tblname,
1482 	    sizeof(rule->overload_tblname));
1483 
1484 	bcopy(&krule->rpool, &rule->rpool, sizeof(krule->rpool));
1485 
1486 	rule->evaluations = counter_u64_fetch(krule->evaluations);
1487 	for (int i = 0; i < 2; i++) {
1488 		rule->packets[i] = counter_u64_fetch(krule->packets[i]);
1489 		rule->bytes[i] = counter_u64_fetch(krule->bytes[i]);
1490 	}
1491 
1492 	/* kif, anchor, overload_tbl are not copied over. */
1493 
1494 	rule->os_fingerprint = krule->os_fingerprint;
1495 
1496 	rule->rtableid = krule->rtableid;
1497 	bcopy(krule->timeout, rule->timeout, sizeof(krule->timeout));
1498 	rule->max_states = krule->max_states;
1499 	rule->max_src_nodes = krule->max_src_nodes;
1500 	rule->max_src_states = krule->max_src_states;
1501 	rule->max_src_conn = krule->max_src_conn;
1502 	rule->max_src_conn_rate.limit = krule->max_src_conn_rate.limit;
1503 	rule->max_src_conn_rate.seconds = krule->max_src_conn_rate.seconds;
1504 	rule->qid = krule->qid;
1505 	rule->pqid = krule->pqid;
1506 	rule->rt_listid = krule->rt_listid;
1507 	rule->nr = krule->nr;
1508 	rule->prob = krule->prob;
1509 	rule->cuid = krule->cuid;
1510 	rule->cpid = krule->cpid;
1511 
1512 	rule->return_icmp = krule->return_icmp;
1513 	rule->return_icmp6 = krule->return_icmp6;
1514 	rule->max_mss = krule->max_mss;
1515 	rule->tag = krule->tag;
1516 	rule->match_tag = krule->match_tag;
1517 	rule->scrub_flags = krule->scrub_flags;
1518 
1519 	bcopy(&krule->uid, &rule->uid, sizeof(krule->uid));
1520 	bcopy(&krule->gid, &rule->gid, sizeof(krule->gid));
1521 
1522 	rule->rule_flag = krule->rule_flag;
1523 	rule->action = krule->action;
1524 	rule->direction = krule->direction;
1525 	rule->log = krule->log;
1526 	rule->logif = krule->logif;
1527 	rule->quick = krule->quick;
1528 	rule->ifnot = krule->ifnot;
1529 	rule->match_tag_not = krule->match_tag_not;
1530 	rule->natpass = krule->natpass;
1531 
1532 	rule->keep_state = krule->keep_state;
1533 	rule->af = krule->af;
1534 	rule->proto = krule->proto;
1535 	rule->type = krule->type;
1536 	rule->code = krule->code;
1537 	rule->flags = krule->flags;
1538 	rule->flagset = krule->flagset;
1539 	rule->min_ttl = krule->min_ttl;
1540 	rule->allow_opts = krule->allow_opts;
1541 	rule->rt = krule->rt;
1542 	rule->return_ttl = krule->return_ttl;
1543 	rule->tos = krule->tos;
1544 	rule->set_tos = krule->set_tos;
1545 	rule->anchor_relative = krule->anchor_relative;
1546 	rule->anchor_wildcard = krule->anchor_wildcard;
1547 
1548 	rule->flush = krule->flush;
1549 	rule->prio = krule->prio;
1550 	rule->set_prio[0] = krule->set_prio[0];
1551 	rule->set_prio[1] = krule->set_prio[1];
1552 
1553 	bcopy(&krule->divert, &rule->divert, sizeof(krule->divert));
1554 
1555 	rule->u_states_cur = counter_u64_fetch(krule->states_cur);
1556 	rule->u_states_tot = counter_u64_fetch(krule->states_tot);
1557 	rule->u_src_nodes = counter_u64_fetch(krule->src_nodes);
1558 }
1559 
1560 static void
1561 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
1562 {
1563 
1564 	bzero(krule, sizeof(*krule));
1565 
1566 	bcopy(&rule->src, &krule->src, sizeof(rule->src));
1567 	bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
1568 
1569 	strlcpy(krule->label, rule->label, sizeof(rule->label));
1570 	strlcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
1571 	strlcpy(krule->qname, rule->qname, sizeof(rule->qname));
1572 	strlcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
1573 	strlcpy(krule->tagname, rule->tagname, sizeof(rule->tagname));
1574 	strlcpy(krule->match_tagname, rule->match_tagname,
1575 	    sizeof(rule->match_tagname));
1576 	strlcpy(krule->overload_tblname, rule->overload_tblname,
1577 	    sizeof(rule->overload_tblname));
1578 
1579 	bcopy(&rule->rpool, &krule->rpool, sizeof(krule->rpool));
1580 
1581 	/* Don't allow userspace to set evaulations, packets or bytes. */
1582 	/* kif, anchor, overload_tbl are not copied over. */
1583 
1584 	krule->os_fingerprint = krule->os_fingerprint;
1585 
1586 	krule->rtableid = rule->rtableid;
1587 	bcopy(rule->timeout, krule->timeout, sizeof(krule->timeout));
1588 	krule->max_states = rule->max_states;
1589 	krule->max_src_nodes = rule->max_src_nodes;
1590 	krule->max_src_states = rule->max_src_states;
1591 	krule->max_src_conn = rule->max_src_conn;
1592 	krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
1593 	krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
1594 	krule->qid = rule->qid;
1595 	krule->pqid = rule->pqid;
1596 	krule->rt_listid = rule->rt_listid;
1597 	krule->nr = rule->nr;
1598 	krule->prob = rule->prob;
1599 	krule->cuid = rule->cuid;
1600 	krule->cpid = rule->cpid;
1601 
1602 	krule->return_icmp = rule->return_icmp;
1603 	krule->return_icmp6 = rule->return_icmp6;
1604 	krule->max_mss = rule->max_mss;
1605 	krule->tag = rule->tag;
1606 	krule->match_tag = rule->match_tag;
1607 	krule->scrub_flags = rule->scrub_flags;
1608 
1609 	bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
1610 	bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
1611 
1612 	krule->rule_flag = rule->rule_flag;
1613 	krule->action = rule->action;
1614 	krule->direction = rule->direction;
1615 	krule->log = rule->log;
1616 	krule->logif = rule->logif;
1617 	krule->quick = rule->quick;
1618 	krule->ifnot = rule->ifnot;
1619 	krule->match_tag_not = rule->match_tag_not;
1620 	krule->natpass = rule->natpass;
1621 
1622 	krule->keep_state = rule->keep_state;
1623 	krule->af = rule->af;
1624 	krule->proto = rule->proto;
1625 	krule->type = rule->type;
1626 	krule->code = rule->code;
1627 	krule->flags = rule->flags;
1628 	krule->flagset = rule->flagset;
1629 	krule->min_ttl = rule->min_ttl;
1630 	krule->allow_opts = rule->allow_opts;
1631 	krule->rt = rule->rt;
1632 	krule->return_ttl = rule->return_ttl;
1633 	krule->tos = rule->tos;
1634 	krule->set_tos = rule->set_tos;
1635 	krule->anchor_relative = rule->anchor_relative;
1636 	krule->anchor_wildcard = rule->anchor_wildcard;
1637 
1638 	krule->flush = rule->flush;
1639 	krule->prio = rule->prio;
1640 	krule->set_prio[0] = rule->set_prio[0];
1641 	krule->set_prio[1] = rule->set_prio[1];
1642 
1643 	bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
1644 }
1645 
1646 static int
1647 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
1648 {
1649 	int			 error = 0;
1650 	PF_RULES_RLOCK_TRACKER;
1651 
1652 	/* XXX keep in sync with switch() below */
1653 	if (securelevel_gt(td->td_ucred, 2))
1654 		switch (cmd) {
1655 		case DIOCGETRULES:
1656 		case DIOCGETRULE:
1657 		case DIOCGETADDRS:
1658 		case DIOCGETADDR:
1659 		case DIOCGETSTATE:
1660 		case DIOCSETSTATUSIF:
1661 		case DIOCGETSTATUS:
1662 		case DIOCCLRSTATUS:
1663 		case DIOCNATLOOK:
1664 		case DIOCSETDEBUG:
1665 		case DIOCGETSTATES:
1666 		case DIOCGETTIMEOUT:
1667 		case DIOCCLRRULECTRS:
1668 		case DIOCGETLIMIT:
1669 		case DIOCGETALTQSV0:
1670 		case DIOCGETALTQSV1:
1671 		case DIOCGETALTQV0:
1672 		case DIOCGETALTQV1:
1673 		case DIOCGETQSTATSV0:
1674 		case DIOCGETQSTATSV1:
1675 		case DIOCGETRULESETS:
1676 		case DIOCGETRULESET:
1677 		case DIOCRGETTABLES:
1678 		case DIOCRGETTSTATS:
1679 		case DIOCRCLRTSTATS:
1680 		case DIOCRCLRADDRS:
1681 		case DIOCRADDADDRS:
1682 		case DIOCRDELADDRS:
1683 		case DIOCRSETADDRS:
1684 		case DIOCRGETADDRS:
1685 		case DIOCRGETASTATS:
1686 		case DIOCRCLRASTATS:
1687 		case DIOCRTSTADDRS:
1688 		case DIOCOSFPGET:
1689 		case DIOCGETSRCNODES:
1690 		case DIOCCLRSRCNODES:
1691 		case DIOCIGETIFACES:
1692 		case DIOCGIFSPEEDV0:
1693 		case DIOCGIFSPEEDV1:
1694 		case DIOCSETIFFLAG:
1695 		case DIOCCLRIFFLAG:
1696 			break;
1697 		case DIOCRCLRTABLES:
1698 		case DIOCRADDTABLES:
1699 		case DIOCRDELTABLES:
1700 		case DIOCRSETTFLAGS:
1701 			if (((struct pfioc_table *)addr)->pfrio_flags &
1702 			    PFR_FLAG_DUMMY)
1703 				break; /* dummy operation ok */
1704 			return (EPERM);
1705 		default:
1706 			return (EPERM);
1707 		}
1708 
1709 	if (!(flags & FWRITE))
1710 		switch (cmd) {
1711 		case DIOCGETRULES:
1712 		case DIOCGETADDRS:
1713 		case DIOCGETADDR:
1714 		case DIOCGETSTATE:
1715 		case DIOCGETSTATUS:
1716 		case DIOCGETSTATES:
1717 		case DIOCGETTIMEOUT:
1718 		case DIOCGETLIMIT:
1719 		case DIOCGETALTQSV0:
1720 		case DIOCGETALTQSV1:
1721 		case DIOCGETALTQV0:
1722 		case DIOCGETALTQV1:
1723 		case DIOCGETQSTATSV0:
1724 		case DIOCGETQSTATSV1:
1725 		case DIOCGETRULESETS:
1726 		case DIOCGETRULESET:
1727 		case DIOCNATLOOK:
1728 		case DIOCRGETTABLES:
1729 		case DIOCRGETTSTATS:
1730 		case DIOCRGETADDRS:
1731 		case DIOCRGETASTATS:
1732 		case DIOCRTSTADDRS:
1733 		case DIOCOSFPGET:
1734 		case DIOCGETSRCNODES:
1735 		case DIOCIGETIFACES:
1736 		case DIOCGIFSPEEDV1:
1737 		case DIOCGIFSPEEDV0:
1738 			break;
1739 		case DIOCRCLRTABLES:
1740 		case DIOCRADDTABLES:
1741 		case DIOCRDELTABLES:
1742 		case DIOCRCLRTSTATS:
1743 		case DIOCRCLRADDRS:
1744 		case DIOCRADDADDRS:
1745 		case DIOCRDELADDRS:
1746 		case DIOCRSETADDRS:
1747 		case DIOCRSETTFLAGS:
1748 			if (((struct pfioc_table *)addr)->pfrio_flags &
1749 			    PFR_FLAG_DUMMY) {
1750 				flags |= FWRITE; /* need write lock for dummy */
1751 				break; /* dummy operation ok */
1752 			}
1753 			return (EACCES);
1754 		case DIOCGETRULE:
1755 			if (((struct pfioc_rule *)addr)->action ==
1756 			    PF_GET_CLR_CNTR)
1757 				return (EACCES);
1758 			break;
1759 		default:
1760 			return (EACCES);
1761 		}
1762 
1763 	CURVNET_SET(TD_TO_VNET(td));
1764 
1765 	switch (cmd) {
1766 	case DIOCSTART:
1767 		sx_xlock(&pf_ioctl_lock);
1768 		if (V_pf_status.running)
1769 			error = EEXIST;
1770 		else {
1771 			int cpu;
1772 
1773 			error = hook_pf();
1774 			if (error) {
1775 				DPFPRINTF(PF_DEBUG_MISC,
1776 				    ("pf: pfil registration failed\n"));
1777 				break;
1778 			}
1779 			V_pf_status.running = 1;
1780 			V_pf_status.since = time_second;
1781 
1782 			CPU_FOREACH(cpu)
1783 				V_pf_stateid[cpu] = time_second;
1784 
1785 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1786 		}
1787 		break;
1788 
1789 	case DIOCSTOP:
1790 		sx_xlock(&pf_ioctl_lock);
1791 		if (!V_pf_status.running)
1792 			error = ENOENT;
1793 		else {
1794 			V_pf_status.running = 0;
1795 			error = dehook_pf();
1796 			if (error) {
1797 				V_pf_status.running = 1;
1798 				DPFPRINTF(PF_DEBUG_MISC,
1799 				    ("pf: pfil unregistration failed\n"));
1800 			}
1801 			V_pf_status.since = time_second;
1802 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1803 		}
1804 		break;
1805 
1806 	case DIOCADDRULE: {
1807 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1808 		struct pf_kruleset	*ruleset;
1809 		struct pf_krule		*rule, *tail;
1810 		struct pf_kpooladdr	*pa;
1811 		struct pfi_kkif		*kif = NULL;
1812 		int			 rs_num;
1813 
1814 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1815 			error = EINVAL;
1816 			break;
1817 		}
1818 		if (pr->rule.src.addr.p.dyn != NULL ||
1819 		    pr->rule.dst.addr.p.dyn != NULL) {
1820 			error = EINVAL;
1821 			break;
1822 		}
1823 #ifndef INET
1824 		if (pr->rule.af == AF_INET) {
1825 			error = EAFNOSUPPORT;
1826 			break;
1827 		}
1828 #endif /* INET */
1829 #ifndef INET6
1830 		if (pr->rule.af == AF_INET6) {
1831 			error = EAFNOSUPPORT;
1832 			break;
1833 		}
1834 #endif /* INET6 */
1835 
1836 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
1837 		pf_rule_to_krule(&pr->rule, rule);
1838 
1839 		if (rule->ifname[0])
1840 			kif = pf_kkif_create(M_WAITOK);
1841 		rule->evaluations = counter_u64_alloc(M_WAITOK);
1842 		for (int i = 0; i < 2; i++) {
1843 			rule->packets[i] = counter_u64_alloc(M_WAITOK);
1844 			rule->bytes[i] = counter_u64_alloc(M_WAITOK);
1845 		}
1846 		rule->states_cur = counter_u64_alloc(M_WAITOK);
1847 		rule->states_tot = counter_u64_alloc(M_WAITOK);
1848 		rule->src_nodes = counter_u64_alloc(M_WAITOK);
1849 		rule->cuid = td->td_ucred->cr_ruid;
1850 		rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1851 		TAILQ_INIT(&rule->rpool.list);
1852 #define	ERROUT(x)	{ error = (x); goto DIOCADDRULE_error; }
1853 
1854 		PF_RULES_WLOCK();
1855 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1856 		ruleset = pf_find_kruleset(pr->anchor);
1857 		if (ruleset == NULL)
1858 			ERROUT(EINVAL);
1859 		rs_num = pf_get_ruleset_number(pr->rule.action);
1860 		if (rs_num >= PF_RULESET_MAX)
1861 			ERROUT(EINVAL);
1862 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1863 			DPFPRINTF(PF_DEBUG_MISC,
1864 			    ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
1865 			    ruleset->rules[rs_num].inactive.ticket));
1866 			ERROUT(EBUSY);
1867 		}
1868 		if (pr->pool_ticket != V_ticket_pabuf) {
1869 			DPFPRINTF(PF_DEBUG_MISC,
1870 			    ("pool_ticket: %d != %d\n", pr->pool_ticket,
1871 			    V_ticket_pabuf));
1872 			ERROUT(EBUSY);
1873 		}
1874 
1875 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1876 		    pf_krulequeue);
1877 		if (tail)
1878 			rule->nr = tail->nr + 1;
1879 		else
1880 			rule->nr = 0;
1881 		if (rule->ifname[0]) {
1882 			rule->kif = pfi_kkif_attach(kif, rule->ifname);
1883 			pfi_kkif_ref(rule->kif);
1884 		} else
1885 			rule->kif = NULL;
1886 
1887 		if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
1888 			error = EBUSY;
1889 
1890 #ifdef ALTQ
1891 		/* set queue IDs */
1892 		if (rule->qname[0] != 0) {
1893 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1894 				error = EBUSY;
1895 			else if (rule->pqname[0] != 0) {
1896 				if ((rule->pqid =
1897 				    pf_qname2qid(rule->pqname)) == 0)
1898 					error = EBUSY;
1899 			} else
1900 				rule->pqid = rule->qid;
1901 		}
1902 #endif
1903 		if (rule->tagname[0])
1904 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1905 				error = EBUSY;
1906 		if (rule->match_tagname[0])
1907 			if ((rule->match_tag =
1908 			    pf_tagname2tag(rule->match_tagname)) == 0)
1909 				error = EBUSY;
1910 		if (rule->rt && !rule->direction)
1911 			error = EINVAL;
1912 		if (!rule->log)
1913 			rule->logif = 0;
1914 		if (rule->logif >= PFLOGIFS_MAX)
1915 			error = EINVAL;
1916 		if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1917 			error = ENOMEM;
1918 		if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1919 			error = ENOMEM;
1920 		if (pf_kanchor_setup(rule, ruleset, pr->anchor_call))
1921 			error = EINVAL;
1922 		if (rule->scrub_flags & PFSTATE_SETPRIO &&
1923 		    (rule->set_prio[0] > PF_PRIO_MAX ||
1924 		    rule->set_prio[1] > PF_PRIO_MAX))
1925 			error = EINVAL;
1926 		TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1927 			if (pa->addr.type == PF_ADDR_TABLE) {
1928 				pa->addr.p.tbl = pfr_attach_table(ruleset,
1929 				    pa->addr.v.tblname);
1930 				if (pa->addr.p.tbl == NULL)
1931 					error = ENOMEM;
1932 			}
1933 
1934 		rule->overload_tbl = NULL;
1935 		if (rule->overload_tblname[0]) {
1936 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1937 			    rule->overload_tblname)) == NULL)
1938 				error = EINVAL;
1939 			else
1940 				rule->overload_tbl->pfrkt_flags |=
1941 				    PFR_TFLAG_ACTIVE;
1942 		}
1943 
1944 		pf_mv_kpool(&V_pf_pabuf, &rule->rpool.list);
1945 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1946 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1947 		    (rule->rt > PF_NOPFROUTE)) &&
1948 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1949 			error = EINVAL;
1950 
1951 		if (error) {
1952 			pf_free_rule(rule);
1953 			PF_RULES_WUNLOCK();
1954 			break;
1955 		}
1956 
1957 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1958 		counter_u64_zero(rule->evaluations);
1959 		for (int i = 0; i < 2; i++) {
1960 			counter_u64_zero(rule->packets[i]);
1961 			counter_u64_zero(rule->bytes[i]);
1962 		}
1963 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1964 		    rule, entries);
1965 		ruleset->rules[rs_num].inactive.rcount++;
1966 		PF_RULES_WUNLOCK();
1967 		break;
1968 
1969 #undef ERROUT
1970 DIOCADDRULE_error:
1971 		PF_RULES_WUNLOCK();
1972 		counter_u64_free(rule->evaluations);
1973 		for (int i = 0; i < 2; i++) {
1974 			counter_u64_free(rule->packets[i]);
1975 			counter_u64_free(rule->bytes[i]);
1976 		}
1977 		counter_u64_free(rule->states_cur);
1978 		counter_u64_free(rule->states_tot);
1979 		counter_u64_free(rule->src_nodes);
1980 		free(rule, M_PFRULE);
1981 		if (kif)
1982 			pf_kkif_free(kif);
1983 		break;
1984 	}
1985 
1986 	case DIOCGETRULES: {
1987 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1988 		struct pf_kruleset	*ruleset;
1989 		struct pf_krule		*tail;
1990 		int			 rs_num;
1991 
1992 		PF_RULES_WLOCK();
1993 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1994 		ruleset = pf_find_kruleset(pr->anchor);
1995 		if (ruleset == NULL) {
1996 			PF_RULES_WUNLOCK();
1997 			error = EINVAL;
1998 			break;
1999 		}
2000 		rs_num = pf_get_ruleset_number(pr->rule.action);
2001 		if (rs_num >= PF_RULESET_MAX) {
2002 			PF_RULES_WUNLOCK();
2003 			error = EINVAL;
2004 			break;
2005 		}
2006 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
2007 		    pf_krulequeue);
2008 		if (tail)
2009 			pr->nr = tail->nr + 1;
2010 		else
2011 			pr->nr = 0;
2012 		pr->ticket = ruleset->rules[rs_num].active.ticket;
2013 		PF_RULES_WUNLOCK();
2014 		break;
2015 	}
2016 
2017 	case DIOCGETRULE: {
2018 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
2019 		struct pf_kruleset	*ruleset;
2020 		struct pf_krule		*rule;
2021 		int			 rs_num;
2022 
2023 		PF_RULES_WLOCK();
2024 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
2025 		ruleset = pf_find_kruleset(pr->anchor);
2026 		if (ruleset == NULL) {
2027 			PF_RULES_WUNLOCK();
2028 			error = EINVAL;
2029 			break;
2030 		}
2031 		rs_num = pf_get_ruleset_number(pr->rule.action);
2032 		if (rs_num >= PF_RULESET_MAX) {
2033 			PF_RULES_WUNLOCK();
2034 			error = EINVAL;
2035 			break;
2036 		}
2037 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
2038 			PF_RULES_WUNLOCK();
2039 			error = EBUSY;
2040 			break;
2041 		}
2042 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
2043 		while ((rule != NULL) && (rule->nr != pr->nr))
2044 			rule = TAILQ_NEXT(rule, entries);
2045 		if (rule == NULL) {
2046 			PF_RULES_WUNLOCK();
2047 			error = EBUSY;
2048 			break;
2049 		}
2050 
2051 		pf_krule_to_rule(rule, &pr->rule);
2052 
2053 		if (pf_kanchor_copyout(ruleset, rule, pr)) {
2054 			PF_RULES_WUNLOCK();
2055 			error = EBUSY;
2056 			break;
2057 		}
2058 		pf_addr_copyout(&pr->rule.src.addr);
2059 		pf_addr_copyout(&pr->rule.dst.addr);
2060 
2061 		if (pr->action == PF_GET_CLR_CNTR) {
2062 			counter_u64_zero(rule->evaluations);
2063 			for (int i = 0; i < 2; i++) {
2064 				counter_u64_zero(rule->packets[i]);
2065 				counter_u64_zero(rule->bytes[i]);
2066 			}
2067 			counter_u64_zero(rule->states_tot);
2068 		}
2069 		PF_RULES_WUNLOCK();
2070 		break;
2071 	}
2072 
2073 	case DIOCCHANGERULE: {
2074 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
2075 		struct pf_kruleset	*ruleset;
2076 		struct pf_krule		*oldrule = NULL, *newrule = NULL;
2077 		struct pfi_kkif		*kif = NULL;
2078 		struct pf_kpooladdr	*pa;
2079 		u_int32_t		 nr = 0;
2080 		int			 rs_num;
2081 
2082 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
2083 		    pcr->action > PF_CHANGE_GET_TICKET) {
2084 			error = EINVAL;
2085 			break;
2086 		}
2087 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
2088 			error = EINVAL;
2089 			break;
2090 		}
2091 
2092 		if (pcr->action != PF_CHANGE_REMOVE) {
2093 #ifndef INET
2094 			if (pcr->rule.af == AF_INET) {
2095 				error = EAFNOSUPPORT;
2096 				break;
2097 			}
2098 #endif /* INET */
2099 #ifndef INET6
2100 			if (pcr->rule.af == AF_INET6) {
2101 				error = EAFNOSUPPORT;
2102 				break;
2103 			}
2104 #endif /* INET6 */
2105 			newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
2106 			pf_rule_to_krule(&pcr->rule, newrule);
2107 
2108 			if (newrule->ifname[0])
2109 				kif = pf_kkif_create(M_WAITOK);
2110 			newrule->evaluations = counter_u64_alloc(M_WAITOK);
2111 			for (int i = 0; i < 2; i++) {
2112 				newrule->packets[i] =
2113 				    counter_u64_alloc(M_WAITOK);
2114 				newrule->bytes[i] =
2115 				    counter_u64_alloc(M_WAITOK);
2116 			}
2117 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
2118 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
2119 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
2120 			newrule->cuid = td->td_ucred->cr_ruid;
2121 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
2122 			TAILQ_INIT(&newrule->rpool.list);
2123 		}
2124 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGERULE_error; }
2125 
2126 		PF_RULES_WLOCK();
2127 		if (!(pcr->action == PF_CHANGE_REMOVE ||
2128 		    pcr->action == PF_CHANGE_GET_TICKET) &&
2129 		    pcr->pool_ticket != V_ticket_pabuf)
2130 			ERROUT(EBUSY);
2131 
2132 		ruleset = pf_find_kruleset(pcr->anchor);
2133 		if (ruleset == NULL)
2134 			ERROUT(EINVAL);
2135 
2136 		rs_num = pf_get_ruleset_number(pcr->rule.action);
2137 		if (rs_num >= PF_RULESET_MAX)
2138 			ERROUT(EINVAL);
2139 
2140 		if (pcr->action == PF_CHANGE_GET_TICKET) {
2141 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
2142 			ERROUT(0);
2143 		} else if (pcr->ticket !=
2144 			    ruleset->rules[rs_num].active.ticket)
2145 				ERROUT(EINVAL);
2146 
2147 		if (pcr->action != PF_CHANGE_REMOVE) {
2148 			if (newrule->ifname[0]) {
2149 				newrule->kif = pfi_kkif_attach(kif,
2150 				    newrule->ifname);
2151 				pfi_kkif_ref(newrule->kif);
2152 			} else
2153 				newrule->kif = NULL;
2154 
2155 			if (newrule->rtableid > 0 &&
2156 			    newrule->rtableid >= rt_numfibs)
2157 				error = EBUSY;
2158 
2159 #ifdef ALTQ
2160 			/* set queue IDs */
2161 			if (newrule->qname[0] != 0) {
2162 				if ((newrule->qid =
2163 				    pf_qname2qid(newrule->qname)) == 0)
2164 					error = EBUSY;
2165 				else if (newrule->pqname[0] != 0) {
2166 					if ((newrule->pqid =
2167 					    pf_qname2qid(newrule->pqname)) == 0)
2168 						error = EBUSY;
2169 				} else
2170 					newrule->pqid = newrule->qid;
2171 			}
2172 #endif /* ALTQ */
2173 			if (newrule->tagname[0])
2174 				if ((newrule->tag =
2175 				    pf_tagname2tag(newrule->tagname)) == 0)
2176 					error = EBUSY;
2177 			if (newrule->match_tagname[0])
2178 				if ((newrule->match_tag = pf_tagname2tag(
2179 				    newrule->match_tagname)) == 0)
2180 					error = EBUSY;
2181 			if (newrule->rt && !newrule->direction)
2182 				error = EINVAL;
2183 			if (!newrule->log)
2184 				newrule->logif = 0;
2185 			if (newrule->logif >= PFLOGIFS_MAX)
2186 				error = EINVAL;
2187 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
2188 				error = ENOMEM;
2189 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
2190 				error = ENOMEM;
2191 			if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
2192 				error = EINVAL;
2193 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
2194 				if (pa->addr.type == PF_ADDR_TABLE) {
2195 					pa->addr.p.tbl =
2196 					    pfr_attach_table(ruleset,
2197 					    pa->addr.v.tblname);
2198 					if (pa->addr.p.tbl == NULL)
2199 						error = ENOMEM;
2200 				}
2201 
2202 			newrule->overload_tbl = NULL;
2203 			if (newrule->overload_tblname[0]) {
2204 				if ((newrule->overload_tbl = pfr_attach_table(
2205 				    ruleset, newrule->overload_tblname)) ==
2206 				    NULL)
2207 					error = EINVAL;
2208 				else
2209 					newrule->overload_tbl->pfrkt_flags |=
2210 					    PFR_TFLAG_ACTIVE;
2211 			}
2212 
2213 			pf_mv_kpool(&V_pf_pabuf, &newrule->rpool.list);
2214 			if (((((newrule->action == PF_NAT) ||
2215 			    (newrule->action == PF_RDR) ||
2216 			    (newrule->action == PF_BINAT) ||
2217 			    (newrule->rt > PF_NOPFROUTE)) &&
2218 			    !newrule->anchor)) &&
2219 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
2220 				error = EINVAL;
2221 
2222 			if (error) {
2223 				pf_free_rule(newrule);
2224 				PF_RULES_WUNLOCK();
2225 				break;
2226 			}
2227 
2228 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
2229 		}
2230 		pf_empty_kpool(&V_pf_pabuf);
2231 
2232 		if (pcr->action == PF_CHANGE_ADD_HEAD)
2233 			oldrule = TAILQ_FIRST(
2234 			    ruleset->rules[rs_num].active.ptr);
2235 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
2236 			oldrule = TAILQ_LAST(
2237 			    ruleset->rules[rs_num].active.ptr, pf_krulequeue);
2238 		else {
2239 			oldrule = TAILQ_FIRST(
2240 			    ruleset->rules[rs_num].active.ptr);
2241 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
2242 				oldrule = TAILQ_NEXT(oldrule, entries);
2243 			if (oldrule == NULL) {
2244 				if (newrule != NULL)
2245 					pf_free_rule(newrule);
2246 				PF_RULES_WUNLOCK();
2247 				error = EINVAL;
2248 				break;
2249 			}
2250 		}
2251 
2252 		if (pcr->action == PF_CHANGE_REMOVE) {
2253 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
2254 			    oldrule);
2255 			ruleset->rules[rs_num].active.rcount--;
2256 		} else {
2257 			if (oldrule == NULL)
2258 				TAILQ_INSERT_TAIL(
2259 				    ruleset->rules[rs_num].active.ptr,
2260 				    newrule, entries);
2261 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
2262 			    pcr->action == PF_CHANGE_ADD_BEFORE)
2263 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
2264 			else
2265 				TAILQ_INSERT_AFTER(
2266 				    ruleset->rules[rs_num].active.ptr,
2267 				    oldrule, newrule, entries);
2268 			ruleset->rules[rs_num].active.rcount++;
2269 		}
2270 
2271 		nr = 0;
2272 		TAILQ_FOREACH(oldrule,
2273 		    ruleset->rules[rs_num].active.ptr, entries)
2274 			oldrule->nr = nr++;
2275 
2276 		ruleset->rules[rs_num].active.ticket++;
2277 
2278 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
2279 		pf_remove_if_empty_kruleset(ruleset);
2280 
2281 		PF_RULES_WUNLOCK();
2282 		break;
2283 
2284 #undef ERROUT
2285 DIOCCHANGERULE_error:
2286 		PF_RULES_WUNLOCK();
2287 		if (newrule != NULL) {
2288 			counter_u64_free(newrule->evaluations);
2289 			for (int i = 0; i < 2; i++) {
2290 				counter_u64_free(newrule->packets[i]);
2291 				counter_u64_free(newrule->bytes[i]);
2292 			}
2293 			counter_u64_free(newrule->states_cur);
2294 			counter_u64_free(newrule->states_tot);
2295 			counter_u64_free(newrule->src_nodes);
2296 			free(newrule, M_PFRULE);
2297 		}
2298 		if (kif != NULL)
2299 			pf_kkif_free(kif);
2300 		break;
2301 	}
2302 
2303 	case DIOCCLRSTATES: {
2304 		struct pf_state		*s;
2305 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
2306 		u_int			 i, killed = 0;
2307 
2308 		for (i = 0; i <= pf_hashmask; i++) {
2309 			struct pf_idhash *ih = &V_pf_idhash[i];
2310 
2311 relock_DIOCCLRSTATES:
2312 			PF_HASHROW_LOCK(ih);
2313 			LIST_FOREACH(s, &ih->states, entry)
2314 				if (!psk->psk_ifname[0] ||
2315 				    !strcmp(psk->psk_ifname,
2316 				    s->kif->pfik_name)) {
2317 					/*
2318 					 * Don't send out individual
2319 					 * delete messages.
2320 					 */
2321 					s->state_flags |= PFSTATE_NOSYNC;
2322 					pf_unlink_state(s, PF_ENTER_LOCKED);
2323 					killed++;
2324 					goto relock_DIOCCLRSTATES;
2325 				}
2326 			PF_HASHROW_UNLOCK(ih);
2327 		}
2328 		psk->psk_killed = killed;
2329 		if (V_pfsync_clear_states_ptr != NULL)
2330 			V_pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
2331 		break;
2332 	}
2333 
2334 	case DIOCKILLSTATES: {
2335 		struct pf_state		*s;
2336 		struct pf_state_key	*sk;
2337 		struct pf_addr		*srcaddr, *dstaddr;
2338 		u_int16_t		 srcport, dstport;
2339 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
2340 		u_int			 i, killed = 0;
2341 
2342 		if (psk->psk_pfcmp.id) {
2343 			if (psk->psk_pfcmp.creatorid == 0)
2344 				psk->psk_pfcmp.creatorid = V_pf_status.hostid;
2345 			if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
2346 			    psk->psk_pfcmp.creatorid))) {
2347 				pf_unlink_state(s, PF_ENTER_LOCKED);
2348 				psk->psk_killed = 1;
2349 			}
2350 			break;
2351 		}
2352 
2353 		for (i = 0; i <= pf_hashmask; i++) {
2354 			struct pf_idhash *ih = &V_pf_idhash[i];
2355 
2356 relock_DIOCKILLSTATES:
2357 			PF_HASHROW_LOCK(ih);
2358 			LIST_FOREACH(s, &ih->states, entry) {
2359 				sk = s->key[PF_SK_WIRE];
2360 				if (s->direction == PF_OUT) {
2361 					srcaddr = &sk->addr[1];
2362 					dstaddr = &sk->addr[0];
2363 					srcport = sk->port[1];
2364 					dstport = sk->port[0];
2365 				} else {
2366 					srcaddr = &sk->addr[0];
2367 					dstaddr = &sk->addr[1];
2368 					srcport = sk->port[0];
2369 					dstport = sk->port[1];
2370 				}
2371 
2372 				if ((!psk->psk_af || sk->af == psk->psk_af)
2373 				    && (!psk->psk_proto || psk->psk_proto ==
2374 				    sk->proto) &&
2375 				    PF_MATCHA(psk->psk_src.neg,
2376 				    &psk->psk_src.addr.v.a.addr,
2377 				    &psk->psk_src.addr.v.a.mask,
2378 				    srcaddr, sk->af) &&
2379 				    PF_MATCHA(psk->psk_dst.neg,
2380 				    &psk->psk_dst.addr.v.a.addr,
2381 				    &psk->psk_dst.addr.v.a.mask,
2382 				    dstaddr, sk->af) &&
2383 				    (psk->psk_src.port_op == 0 ||
2384 				    pf_match_port(psk->psk_src.port_op,
2385 				    psk->psk_src.port[0], psk->psk_src.port[1],
2386 				    srcport)) &&
2387 				    (psk->psk_dst.port_op == 0 ||
2388 				    pf_match_port(psk->psk_dst.port_op,
2389 				    psk->psk_dst.port[0], psk->psk_dst.port[1],
2390 				    dstport)) &&
2391 				    (!psk->psk_label[0] ||
2392 				    (s->rule.ptr->label[0] &&
2393 				    !strcmp(psk->psk_label,
2394 				    s->rule.ptr->label))) &&
2395 				    (!psk->psk_ifname[0] ||
2396 				    !strcmp(psk->psk_ifname,
2397 				    s->kif->pfik_name))) {
2398 					pf_unlink_state(s, PF_ENTER_LOCKED);
2399 					killed++;
2400 					goto relock_DIOCKILLSTATES;
2401 				}
2402 			}
2403 			PF_HASHROW_UNLOCK(ih);
2404 		}
2405 		psk->psk_killed = killed;
2406 		break;
2407 	}
2408 
2409 	case DIOCADDSTATE: {
2410 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
2411 		struct pfsync_state	*sp = &ps->state;
2412 
2413 		if (sp->timeout >= PFTM_MAX) {
2414 			error = EINVAL;
2415 			break;
2416 		}
2417 		if (V_pfsync_state_import_ptr != NULL) {
2418 			PF_RULES_RLOCK();
2419 			error = V_pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
2420 			PF_RULES_RUNLOCK();
2421 		} else
2422 			error = EOPNOTSUPP;
2423 		break;
2424 	}
2425 
2426 	case DIOCGETSTATE: {
2427 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
2428 		struct pf_state		*s;
2429 
2430 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
2431 		if (s == NULL) {
2432 			error = ENOENT;
2433 			break;
2434 		}
2435 
2436 		pfsync_state_export(&ps->state, s);
2437 		PF_STATE_UNLOCK(s);
2438 		break;
2439 	}
2440 
2441 	case DIOCGETSTATES: {
2442 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
2443 		struct pf_state		*s;
2444 		struct pfsync_state	*pstore, *p;
2445 		int i, nr;
2446 
2447 		if (ps->ps_len <= 0) {
2448 			nr = uma_zone_get_cur(V_pf_state_z);
2449 			ps->ps_len = sizeof(struct pfsync_state) * nr;
2450 			break;
2451 		}
2452 
2453 		p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK | M_ZERO);
2454 		nr = 0;
2455 
2456 		for (i = 0; i <= pf_hashmask; i++) {
2457 			struct pf_idhash *ih = &V_pf_idhash[i];
2458 
2459 			PF_HASHROW_LOCK(ih);
2460 			LIST_FOREACH(s, &ih->states, entry) {
2461 				if (s->timeout == PFTM_UNLINKED)
2462 					continue;
2463 
2464 				if ((nr+1) * sizeof(*p) > ps->ps_len) {
2465 					PF_HASHROW_UNLOCK(ih);
2466 					goto DIOCGETSTATES_full;
2467 				}
2468 				pfsync_state_export(p, s);
2469 				p++;
2470 				nr++;
2471 			}
2472 			PF_HASHROW_UNLOCK(ih);
2473 		}
2474 DIOCGETSTATES_full:
2475 		error = copyout(pstore, ps->ps_states,
2476 		    sizeof(struct pfsync_state) * nr);
2477 		if (error) {
2478 			free(pstore, M_TEMP);
2479 			break;
2480 		}
2481 		ps->ps_len = sizeof(struct pfsync_state) * nr;
2482 		free(pstore, M_TEMP);
2483 
2484 		break;
2485 	}
2486 
2487 	case DIOCGETSTATUS: {
2488 		struct pf_status *s = (struct pf_status *)addr;
2489 
2490 		PF_RULES_RLOCK();
2491 		s->running = V_pf_status.running;
2492 		s->since   = V_pf_status.since;
2493 		s->debug   = V_pf_status.debug;
2494 		s->hostid  = V_pf_status.hostid;
2495 		s->states  = V_pf_status.states;
2496 		s->src_nodes = V_pf_status.src_nodes;
2497 
2498 		for (int i = 0; i < PFRES_MAX; i++)
2499 			s->counters[i] =
2500 			    counter_u64_fetch(V_pf_status.counters[i]);
2501 		for (int i = 0; i < LCNT_MAX; i++)
2502 			s->lcounters[i] =
2503 			    counter_u64_fetch(V_pf_status.lcounters[i]);
2504 		for (int i = 0; i < FCNT_MAX; i++)
2505 			s->fcounters[i] =
2506 			    counter_u64_fetch(V_pf_status.fcounters[i]);
2507 		for (int i = 0; i < SCNT_MAX; i++)
2508 			s->scounters[i] =
2509 			    counter_u64_fetch(V_pf_status.scounters[i]);
2510 
2511 		bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ);
2512 		bcopy(V_pf_status.pf_chksum, s->pf_chksum,
2513 		    PF_MD5_DIGEST_LENGTH);
2514 
2515 		pfi_update_status(s->ifname, s);
2516 		PF_RULES_RUNLOCK();
2517 		break;
2518 	}
2519 
2520 	case DIOCSETSTATUSIF: {
2521 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
2522 
2523 		if (pi->ifname[0] == 0) {
2524 			bzero(V_pf_status.ifname, IFNAMSIZ);
2525 			break;
2526 		}
2527 		PF_RULES_WLOCK();
2528 		strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
2529 		PF_RULES_WUNLOCK();
2530 		break;
2531 	}
2532 
2533 	case DIOCCLRSTATUS: {
2534 		PF_RULES_WLOCK();
2535 		for (int i = 0; i < PFRES_MAX; i++)
2536 			counter_u64_zero(V_pf_status.counters[i]);
2537 		for (int i = 0; i < FCNT_MAX; i++)
2538 			counter_u64_zero(V_pf_status.fcounters[i]);
2539 		for (int i = 0; i < SCNT_MAX; i++)
2540 			counter_u64_zero(V_pf_status.scounters[i]);
2541 		for (int i = 0; i < LCNT_MAX; i++)
2542 			counter_u64_zero(V_pf_status.lcounters[i]);
2543 		V_pf_status.since = time_second;
2544 		if (*V_pf_status.ifname)
2545 			pfi_update_status(V_pf_status.ifname, NULL);
2546 		PF_RULES_WUNLOCK();
2547 		break;
2548 	}
2549 
2550 	case DIOCNATLOOK: {
2551 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
2552 		struct pf_state_key	*sk;
2553 		struct pf_state		*state;
2554 		struct pf_state_key_cmp	 key;
2555 		int			 m = 0, direction = pnl->direction;
2556 		int			 sidx, didx;
2557 
2558 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
2559 		sidx = (direction == PF_IN) ? 1 : 0;
2560 		didx = (direction == PF_IN) ? 0 : 1;
2561 
2562 		if (!pnl->proto ||
2563 		    PF_AZERO(&pnl->saddr, pnl->af) ||
2564 		    PF_AZERO(&pnl->daddr, pnl->af) ||
2565 		    ((pnl->proto == IPPROTO_TCP ||
2566 		    pnl->proto == IPPROTO_UDP) &&
2567 		    (!pnl->dport || !pnl->sport)))
2568 			error = EINVAL;
2569 		else {
2570 			bzero(&key, sizeof(key));
2571 			key.af = pnl->af;
2572 			key.proto = pnl->proto;
2573 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
2574 			key.port[sidx] = pnl->sport;
2575 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
2576 			key.port[didx] = pnl->dport;
2577 
2578 			state = pf_find_state_all(&key, direction, &m);
2579 
2580 			if (m > 1)
2581 				error = E2BIG;	/* more than one state */
2582 			else if (state != NULL) {
2583 				/* XXXGL: not locked read */
2584 				sk = state->key[sidx];
2585 				PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
2586 				pnl->rsport = sk->port[sidx];
2587 				PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
2588 				pnl->rdport = sk->port[didx];
2589 			} else
2590 				error = ENOENT;
2591 		}
2592 		break;
2593 	}
2594 
2595 	case DIOCSETTIMEOUT: {
2596 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2597 		int		 old;
2598 
2599 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
2600 		    pt->seconds < 0) {
2601 			error = EINVAL;
2602 			break;
2603 		}
2604 		PF_RULES_WLOCK();
2605 		old = V_pf_default_rule.timeout[pt->timeout];
2606 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
2607 			pt->seconds = 1;
2608 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
2609 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
2610 			wakeup(pf_purge_thread);
2611 		pt->seconds = old;
2612 		PF_RULES_WUNLOCK();
2613 		break;
2614 	}
2615 
2616 	case DIOCGETTIMEOUT: {
2617 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2618 
2619 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
2620 			error = EINVAL;
2621 			break;
2622 		}
2623 		PF_RULES_RLOCK();
2624 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
2625 		PF_RULES_RUNLOCK();
2626 		break;
2627 	}
2628 
2629 	case DIOCGETLIMIT: {
2630 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2631 
2632 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
2633 			error = EINVAL;
2634 			break;
2635 		}
2636 		PF_RULES_RLOCK();
2637 		pl->limit = V_pf_limits[pl->index].limit;
2638 		PF_RULES_RUNLOCK();
2639 		break;
2640 	}
2641 
2642 	case DIOCSETLIMIT: {
2643 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2644 		int			 old_limit;
2645 
2646 		PF_RULES_WLOCK();
2647 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
2648 		    V_pf_limits[pl->index].zone == NULL) {
2649 			PF_RULES_WUNLOCK();
2650 			error = EINVAL;
2651 			break;
2652 		}
2653 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
2654 		old_limit = V_pf_limits[pl->index].limit;
2655 		V_pf_limits[pl->index].limit = pl->limit;
2656 		pl->limit = old_limit;
2657 		PF_RULES_WUNLOCK();
2658 		break;
2659 	}
2660 
2661 	case DIOCSETDEBUG: {
2662 		u_int32_t	*level = (u_int32_t *)addr;
2663 
2664 		PF_RULES_WLOCK();
2665 		V_pf_status.debug = *level;
2666 		PF_RULES_WUNLOCK();
2667 		break;
2668 	}
2669 
2670 	case DIOCCLRRULECTRS: {
2671 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
2672 		struct pf_kruleset	*ruleset = &pf_main_ruleset;
2673 		struct pf_krule		*rule;
2674 
2675 		PF_RULES_WLOCK();
2676 		TAILQ_FOREACH(rule,
2677 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
2678 			counter_u64_zero(rule->evaluations);
2679 			for (int i = 0; i < 2; i++) {
2680 				counter_u64_zero(rule->packets[i]);
2681 				counter_u64_zero(rule->bytes[i]);
2682 			}
2683 		}
2684 		PF_RULES_WUNLOCK();
2685 		break;
2686 	}
2687 
2688 	case DIOCGIFSPEEDV0:
2689 	case DIOCGIFSPEEDV1: {
2690 		struct pf_ifspeed_v1	*psp = (struct pf_ifspeed_v1 *)addr;
2691 		struct pf_ifspeed_v1	ps;
2692 		struct ifnet		*ifp;
2693 
2694 		if (psp->ifname[0] != 0) {
2695 			/* Can we completely trust user-land? */
2696 			strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
2697 			ifp = ifunit(ps.ifname);
2698 			if (ifp != NULL) {
2699 				psp->baudrate32 =
2700 				    (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
2701 				if (cmd == DIOCGIFSPEEDV1)
2702 					psp->baudrate = ifp->if_baudrate;
2703 			} else
2704 				error = EINVAL;
2705 		} else
2706 			error = EINVAL;
2707 		break;
2708 	}
2709 
2710 #ifdef ALTQ
2711 	case DIOCSTARTALTQ: {
2712 		struct pf_altq		*altq;
2713 
2714 		PF_RULES_WLOCK();
2715 		/* enable all altq interfaces on active list */
2716 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
2717 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
2718 				error = pf_enable_altq(altq);
2719 				if (error != 0)
2720 					break;
2721 			}
2722 		}
2723 		if (error == 0)
2724 			V_pf_altq_running = 1;
2725 		PF_RULES_WUNLOCK();
2726 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
2727 		break;
2728 	}
2729 
2730 	case DIOCSTOPALTQ: {
2731 		struct pf_altq		*altq;
2732 
2733 		PF_RULES_WLOCK();
2734 		/* disable all altq interfaces on active list */
2735 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
2736 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
2737 				error = pf_disable_altq(altq);
2738 				if (error != 0)
2739 					break;
2740 			}
2741 		}
2742 		if (error == 0)
2743 			V_pf_altq_running = 0;
2744 		PF_RULES_WUNLOCK();
2745 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2746 		break;
2747 	}
2748 
2749 	case DIOCADDALTQV0:
2750 	case DIOCADDALTQV1: {
2751 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2752 		struct pf_altq		*altq, *a;
2753 		struct ifnet		*ifp;
2754 
2755 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
2756 		error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
2757 		if (error)
2758 			break;
2759 		altq->local_flags = 0;
2760 
2761 		PF_RULES_WLOCK();
2762 		if (pa->ticket != V_ticket_altqs_inactive) {
2763 			PF_RULES_WUNLOCK();
2764 			free(altq, M_PFALTQ);
2765 			error = EBUSY;
2766 			break;
2767 		}
2768 
2769 		/*
2770 		 * if this is for a queue, find the discipline and
2771 		 * copy the necessary fields
2772 		 */
2773 		if (altq->qname[0] != 0) {
2774 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2775 				PF_RULES_WUNLOCK();
2776 				error = EBUSY;
2777 				free(altq, M_PFALTQ);
2778 				break;
2779 			}
2780 			altq->altq_disc = NULL;
2781 			TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
2782 				if (strncmp(a->ifname, altq->ifname,
2783 				    IFNAMSIZ) == 0) {
2784 					altq->altq_disc = a->altq_disc;
2785 					break;
2786 				}
2787 			}
2788 		}
2789 
2790 		if ((ifp = ifunit(altq->ifname)) == NULL)
2791 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
2792 		else
2793 			error = altq_add(ifp, altq);
2794 
2795 		if (error) {
2796 			PF_RULES_WUNLOCK();
2797 			free(altq, M_PFALTQ);
2798 			break;
2799 		}
2800 
2801 		if (altq->qname[0] != 0)
2802 			TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
2803 		else
2804 			TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
2805 		/* version error check done on import above */
2806 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
2807 		PF_RULES_WUNLOCK();
2808 		break;
2809 	}
2810 
2811 	case DIOCGETALTQSV0:
2812 	case DIOCGETALTQSV1: {
2813 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2814 		struct pf_altq		*altq;
2815 
2816 		PF_RULES_RLOCK();
2817 		pa->nr = 0;
2818 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
2819 			pa->nr++;
2820 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
2821 			pa->nr++;
2822 		pa->ticket = V_ticket_altqs_active;
2823 		PF_RULES_RUNLOCK();
2824 		break;
2825 	}
2826 
2827 	case DIOCGETALTQV0:
2828 	case DIOCGETALTQV1: {
2829 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2830 		struct pf_altq		*altq;
2831 
2832 		PF_RULES_RLOCK();
2833 		if (pa->ticket != V_ticket_altqs_active) {
2834 			PF_RULES_RUNLOCK();
2835 			error = EBUSY;
2836 			break;
2837 		}
2838 		altq = pf_altq_get_nth_active(pa->nr);
2839 		if (altq == NULL) {
2840 			PF_RULES_RUNLOCK();
2841 			error = EBUSY;
2842 			break;
2843 		}
2844 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
2845 		PF_RULES_RUNLOCK();
2846 		break;
2847 	}
2848 
2849 	case DIOCCHANGEALTQV0:
2850 	case DIOCCHANGEALTQV1:
2851 		/* CHANGEALTQ not supported yet! */
2852 		error = ENODEV;
2853 		break;
2854 
2855 	case DIOCGETQSTATSV0:
2856 	case DIOCGETQSTATSV1: {
2857 		struct pfioc_qstats_v1	*pq = (struct pfioc_qstats_v1 *)addr;
2858 		struct pf_altq		*altq;
2859 		int			 nbytes;
2860 		u_int32_t		 version;
2861 
2862 		PF_RULES_RLOCK();
2863 		if (pq->ticket != V_ticket_altqs_active) {
2864 			PF_RULES_RUNLOCK();
2865 			error = EBUSY;
2866 			break;
2867 		}
2868 		nbytes = pq->nbytes;
2869 		altq = pf_altq_get_nth_active(pq->nr);
2870 		if (altq == NULL) {
2871 			PF_RULES_RUNLOCK();
2872 			error = EBUSY;
2873 			break;
2874 		}
2875 
2876 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
2877 			PF_RULES_RUNLOCK();
2878 			error = ENXIO;
2879 			break;
2880 		}
2881 		PF_RULES_RUNLOCK();
2882 		if (cmd == DIOCGETQSTATSV0)
2883 			version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
2884 		else
2885 			version = pq->version;
2886 		error = altq_getqstats(altq, pq->buf, &nbytes, version);
2887 		if (error == 0) {
2888 			pq->scheduler = altq->scheduler;
2889 			pq->nbytes = nbytes;
2890 		}
2891 		break;
2892 	}
2893 #endif /* ALTQ */
2894 
2895 	case DIOCBEGINADDRS: {
2896 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2897 
2898 		PF_RULES_WLOCK();
2899 		pf_empty_kpool(&V_pf_pabuf);
2900 		pp->ticket = ++V_ticket_pabuf;
2901 		PF_RULES_WUNLOCK();
2902 		break;
2903 	}
2904 
2905 	case DIOCADDADDR: {
2906 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2907 		struct pf_kpooladdr	*pa;
2908 		struct pfi_kkif		*kif = NULL;
2909 
2910 #ifndef INET
2911 		if (pp->af == AF_INET) {
2912 			error = EAFNOSUPPORT;
2913 			break;
2914 		}
2915 #endif /* INET */
2916 #ifndef INET6
2917 		if (pp->af == AF_INET6) {
2918 			error = EAFNOSUPPORT;
2919 			break;
2920 		}
2921 #endif /* INET6 */
2922 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2923 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2924 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2925 			error = EINVAL;
2926 			break;
2927 		}
2928 		if (pp->addr.addr.p.dyn != NULL) {
2929 			error = EINVAL;
2930 			break;
2931 		}
2932 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2933 		pf_pooladdr_to_kpooladdr(&pp->addr, pa);
2934 		if (pa->ifname[0])
2935 			kif = pf_kkif_create(M_WAITOK);
2936 		PF_RULES_WLOCK();
2937 		if (pp->ticket != V_ticket_pabuf) {
2938 			PF_RULES_WUNLOCK();
2939 			if (pa->ifname[0])
2940 				pf_kkif_free(kif);
2941 			free(pa, M_PFRULE);
2942 			error = EBUSY;
2943 			break;
2944 		}
2945 		if (pa->ifname[0]) {
2946 			pa->kif = pfi_kkif_attach(kif, pa->ifname);
2947 			pfi_kkif_ref(pa->kif);
2948 		} else
2949 			pa->kif = NULL;
2950 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2951 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2952 			if (pa->ifname[0])
2953 				pfi_kkif_unref(pa->kif);
2954 			PF_RULES_WUNLOCK();
2955 			free(pa, M_PFRULE);
2956 			break;
2957 		}
2958 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2959 		PF_RULES_WUNLOCK();
2960 		break;
2961 	}
2962 
2963 	case DIOCGETADDRS: {
2964 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2965 		struct pf_kpool		*pool;
2966 		struct pf_kpooladdr	*pa;
2967 
2968 		PF_RULES_RLOCK();
2969 		pp->nr = 0;
2970 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2971 		    pp->r_num, 0, 1, 0);
2972 		if (pool == NULL) {
2973 			PF_RULES_RUNLOCK();
2974 			error = EBUSY;
2975 			break;
2976 		}
2977 		TAILQ_FOREACH(pa, &pool->list, entries)
2978 			pp->nr++;
2979 		PF_RULES_RUNLOCK();
2980 		break;
2981 	}
2982 
2983 	case DIOCGETADDR: {
2984 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2985 		struct pf_kpool		*pool;
2986 		struct pf_kpooladdr	*pa;
2987 		u_int32_t		 nr = 0;
2988 
2989 		PF_RULES_RLOCK();
2990 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2991 		    pp->r_num, 0, 1, 1);
2992 		if (pool == NULL) {
2993 			PF_RULES_RUNLOCK();
2994 			error = EBUSY;
2995 			break;
2996 		}
2997 		pa = TAILQ_FIRST(&pool->list);
2998 		while ((pa != NULL) && (nr < pp->nr)) {
2999 			pa = TAILQ_NEXT(pa, entries);
3000 			nr++;
3001 		}
3002 		if (pa == NULL) {
3003 			PF_RULES_RUNLOCK();
3004 			error = EBUSY;
3005 			break;
3006 		}
3007 		pf_kpooladdr_to_pooladdr(pa, &pp->addr);
3008 		pf_addr_copyout(&pp->addr.addr);
3009 		PF_RULES_RUNLOCK();
3010 		break;
3011 	}
3012 
3013 	case DIOCCHANGEADDR: {
3014 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
3015 		struct pf_kpool		*pool;
3016 		struct pf_kpooladdr	*oldpa = NULL, *newpa = NULL;
3017 		struct pf_kruleset	*ruleset;
3018 		struct pfi_kkif		*kif = NULL;
3019 
3020 		if (pca->action < PF_CHANGE_ADD_HEAD ||
3021 		    pca->action > PF_CHANGE_REMOVE) {
3022 			error = EINVAL;
3023 			break;
3024 		}
3025 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
3026 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
3027 		    pca->addr.addr.type != PF_ADDR_TABLE) {
3028 			error = EINVAL;
3029 			break;
3030 		}
3031 		if (pca->addr.addr.p.dyn != NULL) {
3032 			error = EINVAL;
3033 			break;
3034 		}
3035 
3036 		if (pca->action != PF_CHANGE_REMOVE) {
3037 #ifndef INET
3038 			if (pca->af == AF_INET) {
3039 				error = EAFNOSUPPORT;
3040 				break;
3041 			}
3042 #endif /* INET */
3043 #ifndef INET6
3044 			if (pca->af == AF_INET6) {
3045 				error = EAFNOSUPPORT;
3046 				break;
3047 			}
3048 #endif /* INET6 */
3049 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
3050 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
3051 			if (newpa->ifname[0])
3052 				kif = pf_kkif_create(M_WAITOK);
3053 			newpa->kif = NULL;
3054 		}
3055 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGEADDR_error; }
3056 		PF_RULES_WLOCK();
3057 		ruleset = pf_find_kruleset(pca->anchor);
3058 		if (ruleset == NULL)
3059 			ERROUT(EBUSY);
3060 
3061 		pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
3062 		    pca->r_num, pca->r_last, 1, 1);
3063 		if (pool == NULL)
3064 			ERROUT(EBUSY);
3065 
3066 		if (pca->action != PF_CHANGE_REMOVE) {
3067 			if (newpa->ifname[0]) {
3068 				newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
3069 				pfi_kkif_ref(newpa->kif);
3070 				kif = NULL;
3071 			}
3072 
3073 			switch (newpa->addr.type) {
3074 			case PF_ADDR_DYNIFTL:
3075 				error = pfi_dynaddr_setup(&newpa->addr,
3076 				    pca->af);
3077 				break;
3078 			case PF_ADDR_TABLE:
3079 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
3080 				    newpa->addr.v.tblname);
3081 				if (newpa->addr.p.tbl == NULL)
3082 					error = ENOMEM;
3083 				break;
3084 			}
3085 			if (error)
3086 				goto DIOCCHANGEADDR_error;
3087 		}
3088 
3089 		switch (pca->action) {
3090 		case PF_CHANGE_ADD_HEAD:
3091 			oldpa = TAILQ_FIRST(&pool->list);
3092 			break;
3093 		case PF_CHANGE_ADD_TAIL:
3094 			oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
3095 			break;
3096 		default:
3097 			oldpa = TAILQ_FIRST(&pool->list);
3098 			for (int i = 0; oldpa && i < pca->nr; i++)
3099 				oldpa = TAILQ_NEXT(oldpa, entries);
3100 
3101 			if (oldpa == NULL)
3102 				ERROUT(EINVAL);
3103 		}
3104 
3105 		if (pca->action == PF_CHANGE_REMOVE) {
3106 			TAILQ_REMOVE(&pool->list, oldpa, entries);
3107 			switch (oldpa->addr.type) {
3108 			case PF_ADDR_DYNIFTL:
3109 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
3110 				break;
3111 			case PF_ADDR_TABLE:
3112 				pfr_detach_table(oldpa->addr.p.tbl);
3113 				break;
3114 			}
3115 			if (oldpa->kif)
3116 				pfi_kkif_unref(oldpa->kif);
3117 			free(oldpa, M_PFRULE);
3118 		} else {
3119 			if (oldpa == NULL)
3120 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
3121 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
3122 			    pca->action == PF_CHANGE_ADD_BEFORE)
3123 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
3124 			else
3125 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
3126 				    newpa, entries);
3127 		}
3128 
3129 		pool->cur = TAILQ_FIRST(&pool->list);
3130 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
3131 		PF_RULES_WUNLOCK();
3132 		break;
3133 
3134 #undef ERROUT
3135 DIOCCHANGEADDR_error:
3136 		if (newpa != NULL) {
3137 			if (newpa->kif)
3138 				pfi_kkif_unref(newpa->kif);
3139 			free(newpa, M_PFRULE);
3140 		}
3141 		PF_RULES_WUNLOCK();
3142 		if (kif != NULL)
3143 			pf_kkif_free(kif);
3144 		break;
3145 	}
3146 
3147 	case DIOCGETRULESETS: {
3148 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
3149 		struct pf_kruleset	*ruleset;
3150 		struct pf_kanchor	*anchor;
3151 
3152 		PF_RULES_RLOCK();
3153 		pr->path[sizeof(pr->path) - 1] = 0;
3154 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
3155 			PF_RULES_RUNLOCK();
3156 			error = ENOENT;
3157 			break;
3158 		}
3159 		pr->nr = 0;
3160 		if (ruleset->anchor == NULL) {
3161 			/* XXX kludge for pf_main_ruleset */
3162 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
3163 				if (anchor->parent == NULL)
3164 					pr->nr++;
3165 		} else {
3166 			RB_FOREACH(anchor, pf_kanchor_node,
3167 			    &ruleset->anchor->children)
3168 				pr->nr++;
3169 		}
3170 		PF_RULES_RUNLOCK();
3171 		break;
3172 	}
3173 
3174 	case DIOCGETRULESET: {
3175 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
3176 		struct pf_kruleset	*ruleset;
3177 		struct pf_kanchor	*anchor;
3178 		u_int32_t		 nr = 0;
3179 
3180 		PF_RULES_RLOCK();
3181 		pr->path[sizeof(pr->path) - 1] = 0;
3182 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
3183 			PF_RULES_RUNLOCK();
3184 			error = ENOENT;
3185 			break;
3186 		}
3187 		pr->name[0] = 0;
3188 		if (ruleset->anchor == NULL) {
3189 			/* XXX kludge for pf_main_ruleset */
3190 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
3191 				if (anchor->parent == NULL && nr++ == pr->nr) {
3192 					strlcpy(pr->name, anchor->name,
3193 					    sizeof(pr->name));
3194 					break;
3195 				}
3196 		} else {
3197 			RB_FOREACH(anchor, pf_kanchor_node,
3198 			    &ruleset->anchor->children)
3199 				if (nr++ == pr->nr) {
3200 					strlcpy(pr->name, anchor->name,
3201 					    sizeof(pr->name));
3202 					break;
3203 				}
3204 		}
3205 		if (!pr->name[0])
3206 			error = EBUSY;
3207 		PF_RULES_RUNLOCK();
3208 		break;
3209 	}
3210 
3211 	case DIOCRCLRTABLES: {
3212 		struct pfioc_table *io = (struct pfioc_table *)addr;
3213 
3214 		if (io->pfrio_esize != 0) {
3215 			error = ENODEV;
3216 			break;
3217 		}
3218 		PF_RULES_WLOCK();
3219 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
3220 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
3221 		PF_RULES_WUNLOCK();
3222 		break;
3223 	}
3224 
3225 	case DIOCRADDTABLES: {
3226 		struct pfioc_table *io = (struct pfioc_table *)addr;
3227 		struct pfr_table *pfrts;
3228 		size_t totlen;
3229 
3230 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3231 			error = ENODEV;
3232 			break;
3233 		}
3234 
3235 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
3236 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
3237 			error = ENOMEM;
3238 			break;
3239 		}
3240 
3241 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3242 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3243 		    M_TEMP, M_WAITOK);
3244 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3245 		if (error) {
3246 			free(pfrts, M_TEMP);
3247 			break;
3248 		}
3249 		PF_RULES_WLOCK();
3250 		error = pfr_add_tables(pfrts, io->pfrio_size,
3251 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3252 		PF_RULES_WUNLOCK();
3253 		free(pfrts, M_TEMP);
3254 		break;
3255 	}
3256 
3257 	case DIOCRDELTABLES: {
3258 		struct pfioc_table *io = (struct pfioc_table *)addr;
3259 		struct pfr_table *pfrts;
3260 		size_t totlen;
3261 
3262 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3263 			error = ENODEV;
3264 			break;
3265 		}
3266 
3267 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
3268 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
3269 			error = ENOMEM;
3270 			break;
3271 		}
3272 
3273 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3274 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3275 		    M_TEMP, M_WAITOK);
3276 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3277 		if (error) {
3278 			free(pfrts, M_TEMP);
3279 			break;
3280 		}
3281 		PF_RULES_WLOCK();
3282 		error = pfr_del_tables(pfrts, io->pfrio_size,
3283 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3284 		PF_RULES_WUNLOCK();
3285 		free(pfrts, M_TEMP);
3286 		break;
3287 	}
3288 
3289 	case DIOCRGETTABLES: {
3290 		struct pfioc_table *io = (struct pfioc_table *)addr;
3291 		struct pfr_table *pfrts;
3292 		size_t totlen;
3293 		int n;
3294 
3295 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3296 			error = ENODEV;
3297 			break;
3298 		}
3299 		PF_RULES_RLOCK();
3300 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3301 		if (n < 0) {
3302 			PF_RULES_RUNLOCK();
3303 			error = EINVAL;
3304 			break;
3305 		}
3306 		io->pfrio_size = min(io->pfrio_size, n);
3307 
3308 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3309 
3310 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3311 		    M_TEMP, M_NOWAIT);
3312 		if (pfrts == NULL) {
3313 			error = ENOMEM;
3314 			PF_RULES_RUNLOCK();
3315 			break;
3316 		}
3317 		error = pfr_get_tables(&io->pfrio_table, pfrts,
3318 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3319 		PF_RULES_RUNLOCK();
3320 		if (error == 0)
3321 			error = copyout(pfrts, io->pfrio_buffer, totlen);
3322 		free(pfrts, M_TEMP);
3323 		break;
3324 	}
3325 
3326 	case DIOCRGETTSTATS: {
3327 		struct pfioc_table *io = (struct pfioc_table *)addr;
3328 		struct pfr_tstats *pfrtstats;
3329 		size_t totlen;
3330 		int n;
3331 
3332 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
3333 			error = ENODEV;
3334 			break;
3335 		}
3336 		PF_RULES_WLOCK();
3337 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3338 		if (n < 0) {
3339 			PF_RULES_WUNLOCK();
3340 			error = EINVAL;
3341 			break;
3342 		}
3343 		io->pfrio_size = min(io->pfrio_size, n);
3344 
3345 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
3346 		pfrtstats = mallocarray(io->pfrio_size,
3347 		    sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT);
3348 		if (pfrtstats == NULL) {
3349 			error = ENOMEM;
3350 			PF_RULES_WUNLOCK();
3351 			break;
3352 		}
3353 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
3354 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3355 		PF_RULES_WUNLOCK();
3356 		if (error == 0)
3357 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
3358 		free(pfrtstats, M_TEMP);
3359 		break;
3360 	}
3361 
3362 	case DIOCRCLRTSTATS: {
3363 		struct pfioc_table *io = (struct pfioc_table *)addr;
3364 		struct pfr_table *pfrts;
3365 		size_t totlen;
3366 
3367 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3368 			error = ENODEV;
3369 			break;
3370 		}
3371 
3372 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
3373 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
3374 			/* We used to count tables and use the minimum required
3375 			 * size, so we didn't fail on overly large requests.
3376 			 * Keep doing so. */
3377 			io->pfrio_size = pf_ioctl_maxcount;
3378 			break;
3379 		}
3380 
3381 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3382 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3383 		    M_TEMP, M_NOWAIT);
3384 		if (pfrts == NULL) {
3385 			error = ENOMEM;
3386 			break;
3387 		}
3388 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3389 		if (error) {
3390 			free(pfrts, M_TEMP);
3391 			break;
3392 		}
3393 
3394 		PF_RULES_WLOCK();
3395 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
3396 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3397 		PF_RULES_WUNLOCK();
3398 		free(pfrts, M_TEMP);
3399 		break;
3400 	}
3401 
3402 	case DIOCRSETTFLAGS: {
3403 		struct pfioc_table *io = (struct pfioc_table *)addr;
3404 		struct pfr_table *pfrts;
3405 		size_t totlen;
3406 		int n;
3407 
3408 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3409 			error = ENODEV;
3410 			break;
3411 		}
3412 
3413 		PF_RULES_RLOCK();
3414 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3415 		if (n < 0) {
3416 			PF_RULES_RUNLOCK();
3417 			error = EINVAL;
3418 			break;
3419 		}
3420 
3421 		io->pfrio_size = min(io->pfrio_size, n);
3422 		PF_RULES_RUNLOCK();
3423 
3424 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3425 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3426 		    M_TEMP, M_WAITOK);
3427 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3428 		if (error) {
3429 			free(pfrts, M_TEMP);
3430 			break;
3431 		}
3432 		PF_RULES_WLOCK();
3433 		error = pfr_set_tflags(pfrts, io->pfrio_size,
3434 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
3435 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3436 		PF_RULES_WUNLOCK();
3437 		free(pfrts, M_TEMP);
3438 		break;
3439 	}
3440 
3441 	case DIOCRCLRADDRS: {
3442 		struct pfioc_table *io = (struct pfioc_table *)addr;
3443 
3444 		if (io->pfrio_esize != 0) {
3445 			error = ENODEV;
3446 			break;
3447 		}
3448 		PF_RULES_WLOCK();
3449 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
3450 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
3451 		PF_RULES_WUNLOCK();
3452 		break;
3453 	}
3454 
3455 	case DIOCRADDADDRS: {
3456 		struct pfioc_table *io = (struct pfioc_table *)addr;
3457 		struct pfr_addr *pfras;
3458 		size_t totlen;
3459 
3460 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3461 			error = ENODEV;
3462 			break;
3463 		}
3464 		if (io->pfrio_size < 0 ||
3465 		    io->pfrio_size > pf_ioctl_maxcount ||
3466 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3467 			error = EINVAL;
3468 			break;
3469 		}
3470 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3471 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3472 		    M_TEMP, M_NOWAIT);
3473 		if (! pfras) {
3474 			error = ENOMEM;
3475 			break;
3476 		}
3477 		error = copyin(io->pfrio_buffer, pfras, totlen);
3478 		if (error) {
3479 			free(pfras, M_TEMP);
3480 			break;
3481 		}
3482 		PF_RULES_WLOCK();
3483 		error = pfr_add_addrs(&io->pfrio_table, pfras,
3484 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
3485 		    PFR_FLAG_USERIOCTL);
3486 		PF_RULES_WUNLOCK();
3487 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3488 			error = copyout(pfras, io->pfrio_buffer, totlen);
3489 		free(pfras, M_TEMP);
3490 		break;
3491 	}
3492 
3493 	case DIOCRDELADDRS: {
3494 		struct pfioc_table *io = (struct pfioc_table *)addr;
3495 		struct pfr_addr *pfras;
3496 		size_t totlen;
3497 
3498 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3499 			error = ENODEV;
3500 			break;
3501 		}
3502 		if (io->pfrio_size < 0 ||
3503 		    io->pfrio_size > pf_ioctl_maxcount ||
3504 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3505 			error = EINVAL;
3506 			break;
3507 		}
3508 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3509 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3510 		    M_TEMP, M_NOWAIT);
3511 		if (! pfras) {
3512 			error = ENOMEM;
3513 			break;
3514 		}
3515 		error = copyin(io->pfrio_buffer, pfras, totlen);
3516 		if (error) {
3517 			free(pfras, M_TEMP);
3518 			break;
3519 		}
3520 		PF_RULES_WLOCK();
3521 		error = pfr_del_addrs(&io->pfrio_table, pfras,
3522 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
3523 		    PFR_FLAG_USERIOCTL);
3524 		PF_RULES_WUNLOCK();
3525 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3526 			error = copyout(pfras, io->pfrio_buffer, totlen);
3527 		free(pfras, M_TEMP);
3528 		break;
3529 	}
3530 
3531 	case DIOCRSETADDRS: {
3532 		struct pfioc_table *io = (struct pfioc_table *)addr;
3533 		struct pfr_addr *pfras;
3534 		size_t totlen, count;
3535 
3536 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3537 			error = ENODEV;
3538 			break;
3539 		}
3540 		if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
3541 			error = EINVAL;
3542 			break;
3543 		}
3544 		count = max(io->pfrio_size, io->pfrio_size2);
3545 		if (count > pf_ioctl_maxcount ||
3546 		    WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
3547 			error = EINVAL;
3548 			break;
3549 		}
3550 		totlen = count * sizeof(struct pfr_addr);
3551 		pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
3552 		    M_NOWAIT);
3553 		if (! pfras) {
3554 			error = ENOMEM;
3555 			break;
3556 		}
3557 		error = copyin(io->pfrio_buffer, pfras, totlen);
3558 		if (error) {
3559 			free(pfras, M_TEMP);
3560 			break;
3561 		}
3562 		PF_RULES_WLOCK();
3563 		error = pfr_set_addrs(&io->pfrio_table, pfras,
3564 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
3565 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
3566 		    PFR_FLAG_USERIOCTL, 0);
3567 		PF_RULES_WUNLOCK();
3568 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3569 			error = copyout(pfras, io->pfrio_buffer, totlen);
3570 		free(pfras, M_TEMP);
3571 		break;
3572 	}
3573 
3574 	case DIOCRGETADDRS: {
3575 		struct pfioc_table *io = (struct pfioc_table *)addr;
3576 		struct pfr_addr *pfras;
3577 		size_t totlen;
3578 
3579 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3580 			error = ENODEV;
3581 			break;
3582 		}
3583 		if (io->pfrio_size < 0 ||
3584 		    io->pfrio_size > pf_ioctl_maxcount ||
3585 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3586 			error = EINVAL;
3587 			break;
3588 		}
3589 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3590 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3591 		    M_TEMP, M_NOWAIT);
3592 		if (! pfras) {
3593 			error = ENOMEM;
3594 			break;
3595 		}
3596 		PF_RULES_RLOCK();
3597 		error = pfr_get_addrs(&io->pfrio_table, pfras,
3598 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3599 		PF_RULES_RUNLOCK();
3600 		if (error == 0)
3601 			error = copyout(pfras, io->pfrio_buffer, totlen);
3602 		free(pfras, M_TEMP);
3603 		break;
3604 	}
3605 
3606 	case DIOCRGETASTATS: {
3607 		struct pfioc_table *io = (struct pfioc_table *)addr;
3608 		struct pfr_astats *pfrastats;
3609 		size_t totlen;
3610 
3611 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
3612 			error = ENODEV;
3613 			break;
3614 		}
3615 		if (io->pfrio_size < 0 ||
3616 		    io->pfrio_size > pf_ioctl_maxcount ||
3617 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
3618 			error = EINVAL;
3619 			break;
3620 		}
3621 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
3622 		pfrastats = mallocarray(io->pfrio_size,
3623 		    sizeof(struct pfr_astats), M_TEMP, M_NOWAIT);
3624 		if (! pfrastats) {
3625 			error = ENOMEM;
3626 			break;
3627 		}
3628 		PF_RULES_RLOCK();
3629 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
3630 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3631 		PF_RULES_RUNLOCK();
3632 		if (error == 0)
3633 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
3634 		free(pfrastats, M_TEMP);
3635 		break;
3636 	}
3637 
3638 	case DIOCRCLRASTATS: {
3639 		struct pfioc_table *io = (struct pfioc_table *)addr;
3640 		struct pfr_addr *pfras;
3641 		size_t totlen;
3642 
3643 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3644 			error = ENODEV;
3645 			break;
3646 		}
3647 		if (io->pfrio_size < 0 ||
3648 		    io->pfrio_size > pf_ioctl_maxcount ||
3649 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3650 			error = EINVAL;
3651 			break;
3652 		}
3653 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3654 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3655 		    M_TEMP, M_NOWAIT);
3656 		if (! pfras) {
3657 			error = ENOMEM;
3658 			break;
3659 		}
3660 		error = copyin(io->pfrio_buffer, pfras, totlen);
3661 		if (error) {
3662 			free(pfras, M_TEMP);
3663 			break;
3664 		}
3665 		PF_RULES_WLOCK();
3666 		error = pfr_clr_astats(&io->pfrio_table, pfras,
3667 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
3668 		    PFR_FLAG_USERIOCTL);
3669 		PF_RULES_WUNLOCK();
3670 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3671 			error = copyout(pfras, io->pfrio_buffer, totlen);
3672 		free(pfras, M_TEMP);
3673 		break;
3674 	}
3675 
3676 	case DIOCRTSTADDRS: {
3677 		struct pfioc_table *io = (struct pfioc_table *)addr;
3678 		struct pfr_addr *pfras;
3679 		size_t totlen;
3680 
3681 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3682 			error = ENODEV;
3683 			break;
3684 		}
3685 		if (io->pfrio_size < 0 ||
3686 		    io->pfrio_size > pf_ioctl_maxcount ||
3687 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3688 			error = EINVAL;
3689 			break;
3690 		}
3691 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3692 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3693 		    M_TEMP, M_NOWAIT);
3694 		if (! pfras) {
3695 			error = ENOMEM;
3696 			break;
3697 		}
3698 		error = copyin(io->pfrio_buffer, pfras, totlen);
3699 		if (error) {
3700 			free(pfras, M_TEMP);
3701 			break;
3702 		}
3703 		PF_RULES_RLOCK();
3704 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
3705 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
3706 		    PFR_FLAG_USERIOCTL);
3707 		PF_RULES_RUNLOCK();
3708 		if (error == 0)
3709 			error = copyout(pfras, io->pfrio_buffer, totlen);
3710 		free(pfras, M_TEMP);
3711 		break;
3712 	}
3713 
3714 	case DIOCRINADEFINE: {
3715 		struct pfioc_table *io = (struct pfioc_table *)addr;
3716 		struct pfr_addr *pfras;
3717 		size_t totlen;
3718 
3719 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3720 			error = ENODEV;
3721 			break;
3722 		}
3723 		if (io->pfrio_size < 0 ||
3724 		    io->pfrio_size > pf_ioctl_maxcount ||
3725 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3726 			error = EINVAL;
3727 			break;
3728 		}
3729 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3730 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3731 		    M_TEMP, M_NOWAIT);
3732 		if (! pfras) {
3733 			error = ENOMEM;
3734 			break;
3735 		}
3736 		error = copyin(io->pfrio_buffer, pfras, totlen);
3737 		if (error) {
3738 			free(pfras, M_TEMP);
3739 			break;
3740 		}
3741 		PF_RULES_WLOCK();
3742 		error = pfr_ina_define(&io->pfrio_table, pfras,
3743 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
3744 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3745 		PF_RULES_WUNLOCK();
3746 		free(pfras, M_TEMP);
3747 		break;
3748 	}
3749 
3750 	case DIOCOSFPADD: {
3751 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
3752 		PF_RULES_WLOCK();
3753 		error = pf_osfp_add(io);
3754 		PF_RULES_WUNLOCK();
3755 		break;
3756 	}
3757 
3758 	case DIOCOSFPGET: {
3759 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
3760 		PF_RULES_RLOCK();
3761 		error = pf_osfp_get(io);
3762 		PF_RULES_RUNLOCK();
3763 		break;
3764 	}
3765 
3766 	case DIOCXBEGIN: {
3767 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3768 		struct pfioc_trans_e	*ioes, *ioe;
3769 		size_t			 totlen;
3770 		int			 i;
3771 
3772 		if (io->esize != sizeof(*ioe)) {
3773 			error = ENODEV;
3774 			break;
3775 		}
3776 		if (io->size < 0 ||
3777 		    io->size > pf_ioctl_maxcount ||
3778 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3779 			error = EINVAL;
3780 			break;
3781 		}
3782 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3783 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3784 		    M_TEMP, M_NOWAIT);
3785 		if (! ioes) {
3786 			error = ENOMEM;
3787 			break;
3788 		}
3789 		error = copyin(io->array, ioes, totlen);
3790 		if (error) {
3791 			free(ioes, M_TEMP);
3792 			break;
3793 		}
3794 		PF_RULES_WLOCK();
3795 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3796 			switch (ioe->rs_num) {
3797 #ifdef ALTQ
3798 			case PF_RULESET_ALTQ:
3799 				if (ioe->anchor[0]) {
3800 					PF_RULES_WUNLOCK();
3801 					free(ioes, M_TEMP);
3802 					error = EINVAL;
3803 					goto fail;
3804 				}
3805 				if ((error = pf_begin_altq(&ioe->ticket))) {
3806 					PF_RULES_WUNLOCK();
3807 					free(ioes, M_TEMP);
3808 					goto fail;
3809 				}
3810 				break;
3811 #endif /* ALTQ */
3812 			case PF_RULESET_TABLE:
3813 			    {
3814 				struct pfr_table table;
3815 
3816 				bzero(&table, sizeof(table));
3817 				strlcpy(table.pfrt_anchor, ioe->anchor,
3818 				    sizeof(table.pfrt_anchor));
3819 				if ((error = pfr_ina_begin(&table,
3820 				    &ioe->ticket, NULL, 0))) {
3821 					PF_RULES_WUNLOCK();
3822 					free(ioes, M_TEMP);
3823 					goto fail;
3824 				}
3825 				break;
3826 			    }
3827 			default:
3828 				if ((error = pf_begin_rules(&ioe->ticket,
3829 				    ioe->rs_num, ioe->anchor))) {
3830 					PF_RULES_WUNLOCK();
3831 					free(ioes, M_TEMP);
3832 					goto fail;
3833 				}
3834 				break;
3835 			}
3836 		}
3837 		PF_RULES_WUNLOCK();
3838 		error = copyout(ioes, io->array, totlen);
3839 		free(ioes, M_TEMP);
3840 		break;
3841 	}
3842 
3843 	case DIOCXROLLBACK: {
3844 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3845 		struct pfioc_trans_e	*ioe, *ioes;
3846 		size_t			 totlen;
3847 		int			 i;
3848 
3849 		if (io->esize != sizeof(*ioe)) {
3850 			error = ENODEV;
3851 			break;
3852 		}
3853 		if (io->size < 0 ||
3854 		    io->size > pf_ioctl_maxcount ||
3855 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3856 			error = EINVAL;
3857 			break;
3858 		}
3859 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3860 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3861 		    M_TEMP, M_NOWAIT);
3862 		if (! ioes) {
3863 			error = ENOMEM;
3864 			break;
3865 		}
3866 		error = copyin(io->array, ioes, totlen);
3867 		if (error) {
3868 			free(ioes, M_TEMP);
3869 			break;
3870 		}
3871 		PF_RULES_WLOCK();
3872 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3873 			switch (ioe->rs_num) {
3874 #ifdef ALTQ
3875 			case PF_RULESET_ALTQ:
3876 				if (ioe->anchor[0]) {
3877 					PF_RULES_WUNLOCK();
3878 					free(ioes, M_TEMP);
3879 					error = EINVAL;
3880 					goto fail;
3881 				}
3882 				if ((error = pf_rollback_altq(ioe->ticket))) {
3883 					PF_RULES_WUNLOCK();
3884 					free(ioes, M_TEMP);
3885 					goto fail; /* really bad */
3886 				}
3887 				break;
3888 #endif /* ALTQ */
3889 			case PF_RULESET_TABLE:
3890 			    {
3891 				struct pfr_table table;
3892 
3893 				bzero(&table, sizeof(table));
3894 				strlcpy(table.pfrt_anchor, ioe->anchor,
3895 				    sizeof(table.pfrt_anchor));
3896 				if ((error = pfr_ina_rollback(&table,
3897 				    ioe->ticket, NULL, 0))) {
3898 					PF_RULES_WUNLOCK();
3899 					free(ioes, M_TEMP);
3900 					goto fail; /* really bad */
3901 				}
3902 				break;
3903 			    }
3904 			default:
3905 				if ((error = pf_rollback_rules(ioe->ticket,
3906 				    ioe->rs_num, ioe->anchor))) {
3907 					PF_RULES_WUNLOCK();
3908 					free(ioes, M_TEMP);
3909 					goto fail; /* really bad */
3910 				}
3911 				break;
3912 			}
3913 		}
3914 		PF_RULES_WUNLOCK();
3915 		free(ioes, M_TEMP);
3916 		break;
3917 	}
3918 
3919 	case DIOCXCOMMIT: {
3920 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3921 		struct pfioc_trans_e	*ioe, *ioes;
3922 		struct pf_kruleset	*rs;
3923 		size_t			 totlen;
3924 		int			 i;
3925 
3926 		if (io->esize != sizeof(*ioe)) {
3927 			error = ENODEV;
3928 			break;
3929 		}
3930 
3931 		if (io->size < 0 ||
3932 		    io->size > pf_ioctl_maxcount ||
3933 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3934 			error = EINVAL;
3935 			break;
3936 		}
3937 
3938 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3939 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3940 		    M_TEMP, M_NOWAIT);
3941 		if (ioes == NULL) {
3942 			error = ENOMEM;
3943 			break;
3944 		}
3945 		error = copyin(io->array, ioes, totlen);
3946 		if (error) {
3947 			free(ioes, M_TEMP);
3948 			break;
3949 		}
3950 		PF_RULES_WLOCK();
3951 		/* First makes sure everything will succeed. */
3952 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3953 			switch (ioe->rs_num) {
3954 #ifdef ALTQ
3955 			case PF_RULESET_ALTQ:
3956 				if (ioe->anchor[0]) {
3957 					PF_RULES_WUNLOCK();
3958 					free(ioes, M_TEMP);
3959 					error = EINVAL;
3960 					goto fail;
3961 				}
3962 				if (!V_altqs_inactive_open || ioe->ticket !=
3963 				    V_ticket_altqs_inactive) {
3964 					PF_RULES_WUNLOCK();
3965 					free(ioes, M_TEMP);
3966 					error = EBUSY;
3967 					goto fail;
3968 				}
3969 				break;
3970 #endif /* ALTQ */
3971 			case PF_RULESET_TABLE:
3972 				rs = pf_find_kruleset(ioe->anchor);
3973 				if (rs == NULL || !rs->topen || ioe->ticket !=
3974 				    rs->tticket) {
3975 					PF_RULES_WUNLOCK();
3976 					free(ioes, M_TEMP);
3977 					error = EBUSY;
3978 					goto fail;
3979 				}
3980 				break;
3981 			default:
3982 				if (ioe->rs_num < 0 || ioe->rs_num >=
3983 				    PF_RULESET_MAX) {
3984 					PF_RULES_WUNLOCK();
3985 					free(ioes, M_TEMP);
3986 					error = EINVAL;
3987 					goto fail;
3988 				}
3989 				rs = pf_find_kruleset(ioe->anchor);
3990 				if (rs == NULL ||
3991 				    !rs->rules[ioe->rs_num].inactive.open ||
3992 				    rs->rules[ioe->rs_num].inactive.ticket !=
3993 				    ioe->ticket) {
3994 					PF_RULES_WUNLOCK();
3995 					free(ioes, M_TEMP);
3996 					error = EBUSY;
3997 					goto fail;
3998 				}
3999 				break;
4000 			}
4001 		}
4002 		/* Now do the commit - no errors should happen here. */
4003 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
4004 			switch (ioe->rs_num) {
4005 #ifdef ALTQ
4006 			case PF_RULESET_ALTQ:
4007 				if ((error = pf_commit_altq(ioe->ticket))) {
4008 					PF_RULES_WUNLOCK();
4009 					free(ioes, M_TEMP);
4010 					goto fail; /* really bad */
4011 				}
4012 				break;
4013 #endif /* ALTQ */
4014 			case PF_RULESET_TABLE:
4015 			    {
4016 				struct pfr_table table;
4017 
4018 				bzero(&table, sizeof(table));
4019 				strlcpy(table.pfrt_anchor, ioe->anchor,
4020 				    sizeof(table.pfrt_anchor));
4021 				if ((error = pfr_ina_commit(&table,
4022 				    ioe->ticket, NULL, NULL, 0))) {
4023 					PF_RULES_WUNLOCK();
4024 					free(ioes, M_TEMP);
4025 					goto fail; /* really bad */
4026 				}
4027 				break;
4028 			    }
4029 			default:
4030 				if ((error = pf_commit_rules(ioe->ticket,
4031 				    ioe->rs_num, ioe->anchor))) {
4032 					PF_RULES_WUNLOCK();
4033 					free(ioes, M_TEMP);
4034 					goto fail; /* really bad */
4035 				}
4036 				break;
4037 			}
4038 		}
4039 		PF_RULES_WUNLOCK();
4040 		free(ioes, M_TEMP);
4041 		break;
4042 	}
4043 
4044 	case DIOCGETSRCNODES: {
4045 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
4046 		struct pf_srchash	*sh;
4047 		struct pf_ksrc_node	*n;
4048 		struct pf_src_node	*p, *pstore;
4049 		uint32_t		 i, nr = 0;
4050 
4051 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4052 				i++, sh++) {
4053 			PF_HASHROW_LOCK(sh);
4054 			LIST_FOREACH(n, &sh->nodes, entry)
4055 				nr++;
4056 			PF_HASHROW_UNLOCK(sh);
4057 		}
4058 
4059 		psn->psn_len = min(psn->psn_len,
4060 		    sizeof(struct pf_src_node) * nr);
4061 
4062 		if (psn->psn_len == 0) {
4063 			psn->psn_len = sizeof(struct pf_src_node) * nr;
4064 			break;
4065 		}
4066 
4067 		nr = 0;
4068 
4069 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
4070 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4071 		    i++, sh++) {
4072 		    PF_HASHROW_LOCK(sh);
4073 		    LIST_FOREACH(n, &sh->nodes, entry) {
4074 
4075 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
4076 				break;
4077 
4078 			pf_src_node_copy(n, p);
4079 
4080 			p++;
4081 			nr++;
4082 		    }
4083 		    PF_HASHROW_UNLOCK(sh);
4084 		}
4085 		error = copyout(pstore, psn->psn_src_nodes,
4086 		    sizeof(struct pf_src_node) * nr);
4087 		if (error) {
4088 			free(pstore, M_TEMP);
4089 			break;
4090 		}
4091 		psn->psn_len = sizeof(struct pf_src_node) * nr;
4092 		free(pstore, M_TEMP);
4093 		break;
4094 	}
4095 
4096 	case DIOCCLRSRCNODES: {
4097 		pf_clear_srcnodes(NULL);
4098 		pf_purge_expired_src_nodes();
4099 		break;
4100 	}
4101 
4102 	case DIOCKILLSRCNODES:
4103 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
4104 		break;
4105 
4106 	case DIOCSETHOSTID: {
4107 		u_int32_t	*hostid = (u_int32_t *)addr;
4108 
4109 		PF_RULES_WLOCK();
4110 		if (*hostid == 0)
4111 			V_pf_status.hostid = arc4random();
4112 		else
4113 			V_pf_status.hostid = *hostid;
4114 		PF_RULES_WUNLOCK();
4115 		break;
4116 	}
4117 
4118 	case DIOCOSFPFLUSH:
4119 		PF_RULES_WLOCK();
4120 		pf_osfp_flush();
4121 		PF_RULES_WUNLOCK();
4122 		break;
4123 
4124 	case DIOCIGETIFACES: {
4125 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
4126 		struct pfi_kif *ifstore;
4127 		size_t bufsiz;
4128 
4129 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
4130 			error = ENODEV;
4131 			break;
4132 		}
4133 
4134 		if (io->pfiio_size < 0 ||
4135 		    io->pfiio_size > pf_ioctl_maxcount ||
4136 		    WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
4137 			error = EINVAL;
4138 			break;
4139 		}
4140 
4141 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
4142 		ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
4143 		    M_TEMP, M_NOWAIT);
4144 		if (ifstore == NULL) {
4145 			error = ENOMEM;
4146 			break;
4147 		}
4148 
4149 		PF_RULES_RLOCK();
4150 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
4151 		PF_RULES_RUNLOCK();
4152 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
4153 		free(ifstore, M_TEMP);
4154 		break;
4155 	}
4156 
4157 	case DIOCSETIFFLAG: {
4158 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
4159 
4160 		PF_RULES_WLOCK();
4161 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
4162 		PF_RULES_WUNLOCK();
4163 		break;
4164 	}
4165 
4166 	case DIOCCLRIFFLAG: {
4167 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
4168 
4169 		PF_RULES_WLOCK();
4170 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
4171 		PF_RULES_WUNLOCK();
4172 		break;
4173 	}
4174 
4175 	default:
4176 		error = ENODEV;
4177 		break;
4178 	}
4179 fail:
4180 	if (sx_xlocked(&pf_ioctl_lock))
4181 		sx_xunlock(&pf_ioctl_lock);
4182 	CURVNET_RESTORE();
4183 
4184 	return (error);
4185 }
4186 
4187 void
4188 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
4189 {
4190 	bzero(sp, sizeof(struct pfsync_state));
4191 
4192 	/* copy from state key */
4193 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
4194 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
4195 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
4196 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
4197 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
4198 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
4199 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
4200 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
4201 	sp->proto = st->key[PF_SK_WIRE]->proto;
4202 	sp->af = st->key[PF_SK_WIRE]->af;
4203 
4204 	/* copy from state */
4205 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
4206 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
4207 	sp->creation = htonl(time_uptime - st->creation);
4208 	sp->expire = pf_state_expires(st);
4209 	if (sp->expire <= time_uptime)
4210 		sp->expire = htonl(0);
4211 	else
4212 		sp->expire = htonl(sp->expire - time_uptime);
4213 
4214 	sp->direction = st->direction;
4215 	sp->log = st->log;
4216 	sp->timeout = st->timeout;
4217 	sp->state_flags = st->state_flags;
4218 	if (st->src_node)
4219 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
4220 	if (st->nat_src_node)
4221 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
4222 
4223 	sp->id = st->id;
4224 	sp->creatorid = st->creatorid;
4225 	pf_state_peer_hton(&st->src, &sp->src);
4226 	pf_state_peer_hton(&st->dst, &sp->dst);
4227 
4228 	if (st->rule.ptr == NULL)
4229 		sp->rule = htonl(-1);
4230 	else
4231 		sp->rule = htonl(st->rule.ptr->nr);
4232 	if (st->anchor.ptr == NULL)
4233 		sp->anchor = htonl(-1);
4234 	else
4235 		sp->anchor = htonl(st->anchor.ptr->nr);
4236 	if (st->nat_rule.ptr == NULL)
4237 		sp->nat_rule = htonl(-1);
4238 	else
4239 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
4240 
4241 	pf_state_counter_hton(counter_u64_fetch(st->packets[0]),
4242 	    sp->packets[0]);
4243 	pf_state_counter_hton(counter_u64_fetch(st->packets[1]),
4244 	    sp->packets[1]);
4245 	pf_state_counter_hton(counter_u64_fetch(st->bytes[0]), sp->bytes[0]);
4246 	pf_state_counter_hton(counter_u64_fetch(st->bytes[1]), sp->bytes[1]);
4247 
4248 }
4249 
4250 static void
4251 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
4252 {
4253 	struct pfr_ktable *kt;
4254 
4255 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
4256 
4257 	kt = aw->p.tbl;
4258 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
4259 		kt = kt->pfrkt_root;
4260 	aw->p.tbl = NULL;
4261 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
4262 		kt->pfrkt_cnt : -1;
4263 }
4264 
4265 /*
4266  * XXX - Check for version missmatch!!!
4267  */
4268 static void
4269 pf_clear_states(void)
4270 {
4271 	struct pf_state	*s;
4272 	u_int i;
4273 
4274 	for (i = 0; i <= pf_hashmask; i++) {
4275 		struct pf_idhash *ih = &V_pf_idhash[i];
4276 relock:
4277 		PF_HASHROW_LOCK(ih);
4278 		LIST_FOREACH(s, &ih->states, entry) {
4279 			s->timeout = PFTM_PURGE;
4280 			/* Don't send out individual delete messages. */
4281 			s->state_flags |= PFSTATE_NOSYNC;
4282 			pf_unlink_state(s, PF_ENTER_LOCKED);
4283 			goto relock;
4284 		}
4285 		PF_HASHROW_UNLOCK(ih);
4286 	}
4287 }
4288 
4289 static int
4290 pf_clear_tables(void)
4291 {
4292 	struct pfioc_table io;
4293 	int error;
4294 
4295 	bzero(&io, sizeof(io));
4296 
4297 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
4298 	    io.pfrio_flags);
4299 
4300 	return (error);
4301 }
4302 
4303 static void
4304 pf_clear_srcnodes(struct pf_ksrc_node *n)
4305 {
4306 	struct pf_state *s;
4307 	int i;
4308 
4309 	for (i = 0; i <= pf_hashmask; i++) {
4310 		struct pf_idhash *ih = &V_pf_idhash[i];
4311 
4312 		PF_HASHROW_LOCK(ih);
4313 		LIST_FOREACH(s, &ih->states, entry) {
4314 			if (n == NULL || n == s->src_node)
4315 				s->src_node = NULL;
4316 			if (n == NULL || n == s->nat_src_node)
4317 				s->nat_src_node = NULL;
4318 		}
4319 		PF_HASHROW_UNLOCK(ih);
4320 	}
4321 
4322 	if (n == NULL) {
4323 		struct pf_srchash *sh;
4324 
4325 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4326 		    i++, sh++) {
4327 			PF_HASHROW_LOCK(sh);
4328 			LIST_FOREACH(n, &sh->nodes, entry) {
4329 				n->expire = 1;
4330 				n->states = 0;
4331 			}
4332 			PF_HASHROW_UNLOCK(sh);
4333 		}
4334 	} else {
4335 		/* XXX: hash slot should already be locked here. */
4336 		n->expire = 1;
4337 		n->states = 0;
4338 	}
4339 }
4340 
4341 static void
4342 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
4343 {
4344 	struct pf_ksrc_node_list	 kill;
4345 
4346 	LIST_INIT(&kill);
4347 	for (int i = 0; i <= pf_srchashmask; i++) {
4348 		struct pf_srchash *sh = &V_pf_srchash[i];
4349 		struct pf_ksrc_node *sn, *tmp;
4350 
4351 		PF_HASHROW_LOCK(sh);
4352 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
4353 			if (PF_MATCHA(psnk->psnk_src.neg,
4354 			      &psnk->psnk_src.addr.v.a.addr,
4355 			      &psnk->psnk_src.addr.v.a.mask,
4356 			      &sn->addr, sn->af) &&
4357 			    PF_MATCHA(psnk->psnk_dst.neg,
4358 			      &psnk->psnk_dst.addr.v.a.addr,
4359 			      &psnk->psnk_dst.addr.v.a.mask,
4360 			      &sn->raddr, sn->af)) {
4361 				pf_unlink_src_node(sn);
4362 				LIST_INSERT_HEAD(&kill, sn, entry);
4363 				sn->expire = 1;
4364 			}
4365 		PF_HASHROW_UNLOCK(sh);
4366 	}
4367 
4368 	for (int i = 0; i <= pf_hashmask; i++) {
4369 		struct pf_idhash *ih = &V_pf_idhash[i];
4370 		struct pf_state *s;
4371 
4372 		PF_HASHROW_LOCK(ih);
4373 		LIST_FOREACH(s, &ih->states, entry) {
4374 			if (s->src_node && s->src_node->expire == 1)
4375 				s->src_node = NULL;
4376 			if (s->nat_src_node && s->nat_src_node->expire == 1)
4377 				s->nat_src_node = NULL;
4378 		}
4379 		PF_HASHROW_UNLOCK(ih);
4380 	}
4381 
4382 	psnk->psnk_killed = pf_free_src_nodes(&kill);
4383 }
4384 
4385 /*
4386  * XXX - Check for version missmatch!!!
4387  */
4388 
4389 /*
4390  * Duplicate pfctl -Fa operation to get rid of as much as we can.
4391  */
4392 static int
4393 shutdown_pf(void)
4394 {
4395 	int error = 0;
4396 	u_int32_t t[5];
4397 	char nn = '\0';
4398 
4399 	do {
4400 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
4401 		    != 0) {
4402 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
4403 			break;
4404 		}
4405 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
4406 		    != 0) {
4407 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
4408 			break;		/* XXX: rollback? */
4409 		}
4410 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
4411 		    != 0) {
4412 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
4413 			break;		/* XXX: rollback? */
4414 		}
4415 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
4416 		    != 0) {
4417 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
4418 			break;		/* XXX: rollback? */
4419 		}
4420 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
4421 		    != 0) {
4422 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
4423 			break;		/* XXX: rollback? */
4424 		}
4425 
4426 		/* XXX: these should always succeed here */
4427 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
4428 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
4429 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
4430 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
4431 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
4432 
4433 		if ((error = pf_clear_tables()) != 0)
4434 			break;
4435 
4436 #ifdef ALTQ
4437 		if ((error = pf_begin_altq(&t[0])) != 0) {
4438 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
4439 			break;
4440 		}
4441 		pf_commit_altq(t[0]);
4442 #endif
4443 
4444 		pf_clear_states();
4445 
4446 		pf_clear_srcnodes(NULL);
4447 
4448 		/* status does not use malloced mem so no need to cleanup */
4449 		/* fingerprints and interfaces have their own cleanup code */
4450 	} while(0);
4451 
4452 	return (error);
4453 }
4454 
4455 static pfil_return_t
4456 pf_check_return(int chk, struct mbuf **m)
4457 {
4458 
4459 	switch (chk) {
4460 	case PF_PASS:
4461 		if (*m == NULL)
4462 			return (PFIL_CONSUMED);
4463 		else
4464 			return (PFIL_PASS);
4465 		break;
4466 	default:
4467 		if (*m != NULL) {
4468 			m_freem(*m);
4469 			*m = NULL;
4470 		}
4471 		return (PFIL_DROPPED);
4472 	}
4473 }
4474 
4475 #ifdef INET
4476 static pfil_return_t
4477 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
4478     void *ruleset __unused, struct inpcb *inp)
4479 {
4480 	int chk;
4481 
4482 	chk = pf_test(PF_IN, flags, ifp, m, inp);
4483 
4484 	return (pf_check_return(chk, m));
4485 }
4486 
4487 static pfil_return_t
4488 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
4489     void *ruleset __unused,  struct inpcb *inp)
4490 {
4491 	int chk;
4492 
4493 	chk = pf_test(PF_OUT, flags, ifp, m, inp);
4494 
4495 	return (pf_check_return(chk, m));
4496 }
4497 #endif
4498 
4499 #ifdef INET6
4500 static pfil_return_t
4501 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
4502     void *ruleset __unused,  struct inpcb *inp)
4503 {
4504 	int chk;
4505 
4506 	/*
4507 	 * In case of loopback traffic IPv6 uses the real interface in
4508 	 * order to support scoped addresses. In order to support stateful
4509 	 * filtering we have change this to lo0 as it is the case in IPv4.
4510 	 */
4511 	CURVNET_SET(ifp->if_vnet);
4512 	chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
4513 	CURVNET_RESTORE();
4514 
4515 	return (pf_check_return(chk, m));
4516 }
4517 
4518 static pfil_return_t
4519 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
4520     void *ruleset __unused,  struct inpcb *inp)
4521 {
4522 	int chk;
4523 
4524 	CURVNET_SET(ifp->if_vnet);
4525 	chk = pf_test6(PF_OUT, flags, ifp, m, inp);
4526 	CURVNET_RESTORE();
4527 
4528 	return (pf_check_return(chk, m));
4529 }
4530 #endif /* INET6 */
4531 
4532 #ifdef INET
4533 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
4534 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
4535 #define	V_pf_ip4_in_hook	VNET(pf_ip4_in_hook)
4536 #define	V_pf_ip4_out_hook	VNET(pf_ip4_out_hook)
4537 #endif
4538 #ifdef INET6
4539 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
4540 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
4541 #define	V_pf_ip6_in_hook	VNET(pf_ip6_in_hook)
4542 #define	V_pf_ip6_out_hook	VNET(pf_ip6_out_hook)
4543 #endif
4544 
4545 static int
4546 hook_pf(void)
4547 {
4548 	struct pfil_hook_args pha;
4549 	struct pfil_link_args pla;
4550 
4551 	if (V_pf_pfil_hooked)
4552 		return (0);
4553 
4554 	pha.pa_version = PFIL_VERSION;
4555 	pha.pa_modname = "pf";
4556 	pha.pa_ruleset = NULL;
4557 
4558 	pla.pa_version = PFIL_VERSION;
4559 
4560 #ifdef INET
4561 	pha.pa_type = PFIL_TYPE_IP4;
4562 	pha.pa_func = pf_check_in;
4563 	pha.pa_flags = PFIL_IN;
4564 	pha.pa_rulname = "default-in";
4565 	V_pf_ip4_in_hook = pfil_add_hook(&pha);
4566 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
4567 	pla.pa_head = V_inet_pfil_head;
4568 	pla.pa_hook = V_pf_ip4_in_hook;
4569 	(void)pfil_link(&pla);
4570 	pha.pa_func = pf_check_out;
4571 	pha.pa_flags = PFIL_OUT;
4572 	pha.pa_rulname = "default-out";
4573 	V_pf_ip4_out_hook = pfil_add_hook(&pha);
4574 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
4575 	pla.pa_head = V_inet_pfil_head;
4576 	pla.pa_hook = V_pf_ip4_out_hook;
4577 	(void)pfil_link(&pla);
4578 #endif
4579 #ifdef INET6
4580 	pha.pa_type = PFIL_TYPE_IP6;
4581 	pha.pa_func = pf_check6_in;
4582 	pha.pa_flags = PFIL_IN;
4583 	pha.pa_rulname = "default-in6";
4584 	V_pf_ip6_in_hook = pfil_add_hook(&pha);
4585 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
4586 	pla.pa_head = V_inet6_pfil_head;
4587 	pla.pa_hook = V_pf_ip6_in_hook;
4588 	(void)pfil_link(&pla);
4589 	pha.pa_func = pf_check6_out;
4590 	pha.pa_rulname = "default-out6";
4591 	pha.pa_flags = PFIL_OUT;
4592 	V_pf_ip6_out_hook = pfil_add_hook(&pha);
4593 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
4594 	pla.pa_head = V_inet6_pfil_head;
4595 	pla.pa_hook = V_pf_ip6_out_hook;
4596 	(void)pfil_link(&pla);
4597 #endif
4598 
4599 	V_pf_pfil_hooked = 1;
4600 	return (0);
4601 }
4602 
4603 static int
4604 dehook_pf(void)
4605 {
4606 
4607 	if (V_pf_pfil_hooked == 0)
4608 		return (0);
4609 
4610 #ifdef INET
4611 	pfil_remove_hook(V_pf_ip4_in_hook);
4612 	pfil_remove_hook(V_pf_ip4_out_hook);
4613 #endif
4614 #ifdef INET6
4615 	pfil_remove_hook(V_pf_ip6_in_hook);
4616 	pfil_remove_hook(V_pf_ip6_out_hook);
4617 #endif
4618 
4619 	V_pf_pfil_hooked = 0;
4620 	return (0);
4621 }
4622 
4623 static void
4624 pf_load_vnet(void)
4625 {
4626 	V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
4627 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
4628 
4629 	pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
4630 	    PF_RULE_TAG_HASH_SIZE_DEFAULT);
4631 #ifdef ALTQ
4632 	pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
4633 	    PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
4634 #endif
4635 
4636 	pfattach_vnet();
4637 	V_pf_vnet_active = 1;
4638 }
4639 
4640 static int
4641 pf_load(void)
4642 {
4643 	int error;
4644 
4645 	rm_init(&pf_rules_lock, "pf rulesets");
4646 	sx_init(&pf_ioctl_lock, "pf ioctl");
4647 	sx_init(&pf_end_lock, "pf end thread");
4648 
4649 	pf_mtag_initialize();
4650 
4651 	pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
4652 	if (pf_dev == NULL)
4653 		return (ENOMEM);
4654 
4655 	pf_end_threads = 0;
4656 	error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
4657 	if (error != 0)
4658 		return (error);
4659 
4660 	pfi_initialize();
4661 
4662 	return (0);
4663 }
4664 
4665 static void
4666 pf_unload_vnet(void)
4667 {
4668 	int error;
4669 
4670 	V_pf_vnet_active = 0;
4671 	V_pf_status.running = 0;
4672 	error = dehook_pf();
4673 	if (error) {
4674 		/*
4675 		 * Should not happen!
4676 		 * XXX Due to error code ESRCH, kldunload will show
4677 		 * a message like 'No such process'.
4678 		 */
4679 		printf("%s : pfil unregisteration fail\n", __FUNCTION__);
4680 		return;
4681 	}
4682 
4683 	PF_RULES_WLOCK();
4684 	shutdown_pf();
4685 	PF_RULES_WUNLOCK();
4686 
4687 	swi_remove(V_pf_swi_cookie);
4688 
4689 	pf_unload_vnet_purge();
4690 
4691 	pf_normalize_cleanup();
4692 	PF_RULES_WLOCK();
4693 	pfi_cleanup_vnet();
4694 	PF_RULES_WUNLOCK();
4695 	pfr_cleanup();
4696 	pf_osfp_flush();
4697 	pf_cleanup();
4698 	if (IS_DEFAULT_VNET(curvnet))
4699 		pf_mtag_cleanup();
4700 
4701 	pf_cleanup_tagset(&V_pf_tags);
4702 #ifdef ALTQ
4703 	pf_cleanup_tagset(&V_pf_qids);
4704 #endif
4705 	uma_zdestroy(V_pf_tag_z);
4706 
4707 	/* Free counters last as we updated them during shutdown. */
4708 	counter_u64_free(V_pf_default_rule.evaluations);
4709 	for (int i = 0; i < 2; i++) {
4710 		counter_u64_free(V_pf_default_rule.packets[i]);
4711 		counter_u64_free(V_pf_default_rule.bytes[i]);
4712 	}
4713 	counter_u64_free(V_pf_default_rule.states_cur);
4714 	counter_u64_free(V_pf_default_rule.states_tot);
4715 	counter_u64_free(V_pf_default_rule.src_nodes);
4716 
4717 	for (int i = 0; i < PFRES_MAX; i++)
4718 		counter_u64_free(V_pf_status.counters[i]);
4719 	for (int i = 0; i < LCNT_MAX; i++)
4720 		counter_u64_free(V_pf_status.lcounters[i]);
4721 	for (int i = 0; i < FCNT_MAX; i++)
4722 		counter_u64_free(V_pf_status.fcounters[i]);
4723 	for (int i = 0; i < SCNT_MAX; i++)
4724 		counter_u64_free(V_pf_status.scounters[i]);
4725 }
4726 
4727 static void
4728 pf_unload(void)
4729 {
4730 
4731 	sx_xlock(&pf_end_lock);
4732 	pf_end_threads = 1;
4733 	while (pf_end_threads < 2) {
4734 		wakeup_one(pf_purge_thread);
4735 		sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
4736 	}
4737 	sx_xunlock(&pf_end_lock);
4738 
4739 	if (pf_dev != NULL)
4740 		destroy_dev(pf_dev);
4741 
4742 	pfi_cleanup();
4743 
4744 	rm_destroy(&pf_rules_lock);
4745 	sx_destroy(&pf_ioctl_lock);
4746 	sx_destroy(&pf_end_lock);
4747 }
4748 
4749 static void
4750 vnet_pf_init(void *unused __unused)
4751 {
4752 
4753 	pf_load_vnet();
4754 }
4755 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
4756     vnet_pf_init, NULL);
4757 
4758 static void
4759 vnet_pf_uninit(const void *unused __unused)
4760 {
4761 
4762 	pf_unload_vnet();
4763 }
4764 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
4765 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
4766     vnet_pf_uninit, NULL);
4767 
4768 static int
4769 pf_modevent(module_t mod, int type, void *data)
4770 {
4771 	int error = 0;
4772 
4773 	switch(type) {
4774 	case MOD_LOAD:
4775 		error = pf_load();
4776 		break;
4777 	case MOD_UNLOAD:
4778 		/* Handled in SYSUNINIT(pf_unload) to ensure it's done after
4779 		 * the vnet_pf_uninit()s */
4780 		break;
4781 	default:
4782 		error = EINVAL;
4783 		break;
4784 	}
4785 
4786 	return (error);
4787 }
4788 
4789 static moduledata_t pf_mod = {
4790 	"pf",
4791 	pf_modevent,
4792 	0
4793 };
4794 
4795 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
4796 MODULE_VERSION(pf, PF_MODVER);
4797