xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision 8a272653d9fbd9fc37691c9aad6a05089b4ecb4d)
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 void		hook_pf(void);
217 static void		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 int
1561 pf_check_rule_addr(const struct pf_rule_addr *addr)
1562 {
1563 
1564 	switch (addr->addr.type) {
1565 	case PF_ADDR_ADDRMASK:
1566 	case PF_ADDR_NOROUTE:
1567 	case PF_ADDR_DYNIFTL:
1568 	case PF_ADDR_TABLE:
1569 	case PF_ADDR_URPFFAILED:
1570 	case PF_ADDR_RANGE:
1571 		break;
1572 	default:
1573 		return (EINVAL);
1574 	}
1575 
1576 	if (addr->addr.p.dyn != NULL) {
1577 		return (EINVAL);
1578 	}
1579 
1580 	return (0);
1581 }
1582 
1583 static int
1584 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
1585 {
1586 	int ret;
1587 
1588 #ifndef INET
1589 	if (rule->af == AF_INET) {
1590 		return (EAFNOSUPPORT);
1591 	}
1592 #endif /* INET */
1593 #ifndef INET6
1594 	if (rule->af == AF_INET6) {
1595 		return (EAFNOSUPPORT);
1596 	}
1597 #endif /* INET6 */
1598 
1599 	ret = pf_check_rule_addr(&rule->src);
1600 	if (ret != 0)
1601 		return (ret);
1602 	ret = pf_check_rule_addr(&rule->dst);
1603 	if (ret != 0)
1604 		return (ret);
1605 
1606 	bzero(krule, sizeof(*krule));
1607 
1608 	bcopy(&rule->src, &krule->src, sizeof(rule->src));
1609 	bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
1610 
1611 	strlcpy(krule->label, rule->label, sizeof(rule->label));
1612 	strlcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
1613 	strlcpy(krule->qname, rule->qname, sizeof(rule->qname));
1614 	strlcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
1615 	strlcpy(krule->tagname, rule->tagname, sizeof(rule->tagname));
1616 	strlcpy(krule->match_tagname, rule->match_tagname,
1617 	    sizeof(rule->match_tagname));
1618 	strlcpy(krule->overload_tblname, rule->overload_tblname,
1619 	    sizeof(rule->overload_tblname));
1620 
1621 	bcopy(&rule->rpool, &krule->rpool, sizeof(krule->rpool));
1622 
1623 	/* Don't allow userspace to set evaulations, packets or bytes. */
1624 	/* kif, anchor, overload_tbl are not copied over. */
1625 
1626 	krule->os_fingerprint = rule->os_fingerprint;
1627 
1628 	krule->rtableid = rule->rtableid;
1629 	bcopy(rule->timeout, krule->timeout, sizeof(krule->timeout));
1630 	krule->max_states = rule->max_states;
1631 	krule->max_src_nodes = rule->max_src_nodes;
1632 	krule->max_src_states = rule->max_src_states;
1633 	krule->max_src_conn = rule->max_src_conn;
1634 	krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
1635 	krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
1636 	krule->qid = rule->qid;
1637 	krule->pqid = rule->pqid;
1638 	krule->rt_listid = rule->rt_listid;
1639 	krule->nr = rule->nr;
1640 	krule->prob = rule->prob;
1641 	krule->cuid = rule->cuid;
1642 	krule->cpid = rule->cpid;
1643 
1644 	krule->return_icmp = rule->return_icmp;
1645 	krule->return_icmp6 = rule->return_icmp6;
1646 	krule->max_mss = rule->max_mss;
1647 	krule->tag = rule->tag;
1648 	krule->match_tag = rule->match_tag;
1649 	krule->scrub_flags = rule->scrub_flags;
1650 
1651 	bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
1652 	bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
1653 
1654 	krule->rule_flag = rule->rule_flag;
1655 	krule->action = rule->action;
1656 	krule->direction = rule->direction;
1657 	krule->log = rule->log;
1658 	krule->logif = rule->logif;
1659 	krule->quick = rule->quick;
1660 	krule->ifnot = rule->ifnot;
1661 	krule->match_tag_not = rule->match_tag_not;
1662 	krule->natpass = rule->natpass;
1663 
1664 	krule->keep_state = rule->keep_state;
1665 	krule->af = rule->af;
1666 	krule->proto = rule->proto;
1667 	krule->type = rule->type;
1668 	krule->code = rule->code;
1669 	krule->flags = rule->flags;
1670 	krule->flagset = rule->flagset;
1671 	krule->min_ttl = rule->min_ttl;
1672 	krule->allow_opts = rule->allow_opts;
1673 	krule->rt = rule->rt;
1674 	krule->return_ttl = rule->return_ttl;
1675 	krule->tos = rule->tos;
1676 	krule->set_tos = rule->set_tos;
1677 	krule->anchor_relative = rule->anchor_relative;
1678 	krule->anchor_wildcard = rule->anchor_wildcard;
1679 
1680 	krule->flush = rule->flush;
1681 	krule->prio = rule->prio;
1682 	krule->set_prio[0] = rule->set_prio[0];
1683 	krule->set_prio[1] = rule->set_prio[1];
1684 
1685 	bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
1686 
1687 	return (0);
1688 }
1689 
1690 static int
1691 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
1692 {
1693 	int			 error = 0;
1694 	PF_RULES_RLOCK_TRACKER;
1695 
1696 	/* XXX keep in sync with switch() below */
1697 	if (securelevel_gt(td->td_ucred, 2))
1698 		switch (cmd) {
1699 		case DIOCGETRULES:
1700 		case DIOCGETRULE:
1701 		case DIOCGETADDRS:
1702 		case DIOCGETADDR:
1703 		case DIOCGETSTATE:
1704 		case DIOCSETSTATUSIF:
1705 		case DIOCGETSTATUS:
1706 		case DIOCCLRSTATUS:
1707 		case DIOCNATLOOK:
1708 		case DIOCSETDEBUG:
1709 		case DIOCGETSTATES:
1710 		case DIOCGETTIMEOUT:
1711 		case DIOCCLRRULECTRS:
1712 		case DIOCGETLIMIT:
1713 		case DIOCGETALTQSV0:
1714 		case DIOCGETALTQSV1:
1715 		case DIOCGETALTQV0:
1716 		case DIOCGETALTQV1:
1717 		case DIOCGETQSTATSV0:
1718 		case DIOCGETQSTATSV1:
1719 		case DIOCGETRULESETS:
1720 		case DIOCGETRULESET:
1721 		case DIOCRGETTABLES:
1722 		case DIOCRGETTSTATS:
1723 		case DIOCRCLRTSTATS:
1724 		case DIOCRCLRADDRS:
1725 		case DIOCRADDADDRS:
1726 		case DIOCRDELADDRS:
1727 		case DIOCRSETADDRS:
1728 		case DIOCRGETADDRS:
1729 		case DIOCRGETASTATS:
1730 		case DIOCRCLRASTATS:
1731 		case DIOCRTSTADDRS:
1732 		case DIOCOSFPGET:
1733 		case DIOCGETSRCNODES:
1734 		case DIOCCLRSRCNODES:
1735 		case DIOCIGETIFACES:
1736 		case DIOCGIFSPEEDV0:
1737 		case DIOCGIFSPEEDV1:
1738 		case DIOCSETIFFLAG:
1739 		case DIOCCLRIFFLAG:
1740 			break;
1741 		case DIOCRCLRTABLES:
1742 		case DIOCRADDTABLES:
1743 		case DIOCRDELTABLES:
1744 		case DIOCRSETTFLAGS:
1745 			if (((struct pfioc_table *)addr)->pfrio_flags &
1746 			    PFR_FLAG_DUMMY)
1747 				break; /* dummy operation ok */
1748 			return (EPERM);
1749 		default:
1750 			return (EPERM);
1751 		}
1752 
1753 	if (!(flags & FWRITE))
1754 		switch (cmd) {
1755 		case DIOCGETRULES:
1756 		case DIOCGETADDRS:
1757 		case DIOCGETADDR:
1758 		case DIOCGETSTATE:
1759 		case DIOCGETSTATUS:
1760 		case DIOCGETSTATES:
1761 		case DIOCGETTIMEOUT:
1762 		case DIOCGETLIMIT:
1763 		case DIOCGETALTQSV0:
1764 		case DIOCGETALTQSV1:
1765 		case DIOCGETALTQV0:
1766 		case DIOCGETALTQV1:
1767 		case DIOCGETQSTATSV0:
1768 		case DIOCGETQSTATSV1:
1769 		case DIOCGETRULESETS:
1770 		case DIOCGETRULESET:
1771 		case DIOCNATLOOK:
1772 		case DIOCRGETTABLES:
1773 		case DIOCRGETTSTATS:
1774 		case DIOCRGETADDRS:
1775 		case DIOCRGETASTATS:
1776 		case DIOCRTSTADDRS:
1777 		case DIOCOSFPGET:
1778 		case DIOCGETSRCNODES:
1779 		case DIOCIGETIFACES:
1780 		case DIOCGIFSPEEDV1:
1781 		case DIOCGIFSPEEDV0:
1782 			break;
1783 		case DIOCRCLRTABLES:
1784 		case DIOCRADDTABLES:
1785 		case DIOCRDELTABLES:
1786 		case DIOCRCLRTSTATS:
1787 		case DIOCRCLRADDRS:
1788 		case DIOCRADDADDRS:
1789 		case DIOCRDELADDRS:
1790 		case DIOCRSETADDRS:
1791 		case DIOCRSETTFLAGS:
1792 			if (((struct pfioc_table *)addr)->pfrio_flags &
1793 			    PFR_FLAG_DUMMY) {
1794 				flags |= FWRITE; /* need write lock for dummy */
1795 				break; /* dummy operation ok */
1796 			}
1797 			return (EACCES);
1798 		case DIOCGETRULE:
1799 			if (((struct pfioc_rule *)addr)->action ==
1800 			    PF_GET_CLR_CNTR)
1801 				return (EACCES);
1802 			break;
1803 		default:
1804 			return (EACCES);
1805 		}
1806 
1807 	CURVNET_SET(TD_TO_VNET(td));
1808 
1809 	switch (cmd) {
1810 	case DIOCSTART:
1811 		sx_xlock(&pf_ioctl_lock);
1812 		if (V_pf_status.running)
1813 			error = EEXIST;
1814 		else {
1815 			int cpu;
1816 
1817 			hook_pf();
1818 			V_pf_status.running = 1;
1819 			V_pf_status.since = time_second;
1820 
1821 			CPU_FOREACH(cpu)
1822 				V_pf_stateid[cpu] = time_second;
1823 
1824 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1825 		}
1826 		break;
1827 
1828 	case DIOCSTOP:
1829 		sx_xlock(&pf_ioctl_lock);
1830 		if (!V_pf_status.running)
1831 			error = ENOENT;
1832 		else {
1833 			V_pf_status.running = 0;
1834 			dehook_pf();
1835 			V_pf_status.since = time_second;
1836 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1837 		}
1838 		break;
1839 
1840 	case DIOCADDRULE: {
1841 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1842 		struct pf_kruleset	*ruleset;
1843 		struct pf_krule		*rule, *tail;
1844 		struct pf_kpooladdr	*pa;
1845 		struct pfi_kkif		*kif = NULL;
1846 		int			 rs_num;
1847 
1848 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1849 			error = EINVAL;
1850 			break;
1851 		}
1852 
1853 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
1854 		error = pf_rule_to_krule(&pr->rule, rule);
1855 		if (error != 0) {
1856 			free(rule, M_PFRULE);
1857 			break;
1858 		}
1859 
1860 		if (rule->ifname[0])
1861 			kif = pf_kkif_create(M_WAITOK);
1862 		rule->evaluations = counter_u64_alloc(M_WAITOK);
1863 		for (int i = 0; i < 2; i++) {
1864 			rule->packets[i] = counter_u64_alloc(M_WAITOK);
1865 			rule->bytes[i] = counter_u64_alloc(M_WAITOK);
1866 		}
1867 		rule->states_cur = counter_u64_alloc(M_WAITOK);
1868 		rule->states_tot = counter_u64_alloc(M_WAITOK);
1869 		rule->src_nodes = counter_u64_alloc(M_WAITOK);
1870 		rule->cuid = td->td_ucred->cr_ruid;
1871 		rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1872 		TAILQ_INIT(&rule->rpool.list);
1873 #define	ERROUT(x)	{ error = (x); goto DIOCADDRULE_error; }
1874 
1875 		PF_RULES_WLOCK();
1876 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1877 		ruleset = pf_find_kruleset(pr->anchor);
1878 		if (ruleset == NULL)
1879 			ERROUT(EINVAL);
1880 		rs_num = pf_get_ruleset_number(pr->rule.action);
1881 		if (rs_num >= PF_RULESET_MAX)
1882 			ERROUT(EINVAL);
1883 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1884 			DPFPRINTF(PF_DEBUG_MISC,
1885 			    ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
1886 			    ruleset->rules[rs_num].inactive.ticket));
1887 			ERROUT(EBUSY);
1888 		}
1889 		if (pr->pool_ticket != V_ticket_pabuf) {
1890 			DPFPRINTF(PF_DEBUG_MISC,
1891 			    ("pool_ticket: %d != %d\n", pr->pool_ticket,
1892 			    V_ticket_pabuf));
1893 			ERROUT(EBUSY);
1894 		}
1895 
1896 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1897 		    pf_krulequeue);
1898 		if (tail)
1899 			rule->nr = tail->nr + 1;
1900 		else
1901 			rule->nr = 0;
1902 		if (rule->ifname[0]) {
1903 			rule->kif = pfi_kkif_attach(kif, rule->ifname);
1904 			pfi_kkif_ref(rule->kif);
1905 		} else
1906 			rule->kif = NULL;
1907 
1908 		if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
1909 			error = EBUSY;
1910 
1911 #ifdef ALTQ
1912 		/* set queue IDs */
1913 		if (rule->qname[0] != 0) {
1914 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1915 				error = EBUSY;
1916 			else if (rule->pqname[0] != 0) {
1917 				if ((rule->pqid =
1918 				    pf_qname2qid(rule->pqname)) == 0)
1919 					error = EBUSY;
1920 			} else
1921 				rule->pqid = rule->qid;
1922 		}
1923 #endif
1924 		if (rule->tagname[0])
1925 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1926 				error = EBUSY;
1927 		if (rule->match_tagname[0])
1928 			if ((rule->match_tag =
1929 			    pf_tagname2tag(rule->match_tagname)) == 0)
1930 				error = EBUSY;
1931 		if (rule->rt && !rule->direction)
1932 			error = EINVAL;
1933 		if (!rule->log)
1934 			rule->logif = 0;
1935 		if (rule->logif >= PFLOGIFS_MAX)
1936 			error = EINVAL;
1937 		if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1938 			error = ENOMEM;
1939 		if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1940 			error = ENOMEM;
1941 		if (pf_kanchor_setup(rule, ruleset, pr->anchor_call))
1942 			error = EINVAL;
1943 		if (rule->scrub_flags & PFSTATE_SETPRIO &&
1944 		    (rule->set_prio[0] > PF_PRIO_MAX ||
1945 		    rule->set_prio[1] > PF_PRIO_MAX))
1946 			error = EINVAL;
1947 		TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1948 			if (pa->addr.type == PF_ADDR_TABLE) {
1949 				pa->addr.p.tbl = pfr_attach_table(ruleset,
1950 				    pa->addr.v.tblname);
1951 				if (pa->addr.p.tbl == NULL)
1952 					error = ENOMEM;
1953 			}
1954 
1955 		rule->overload_tbl = NULL;
1956 		if (rule->overload_tblname[0]) {
1957 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1958 			    rule->overload_tblname)) == NULL)
1959 				error = EINVAL;
1960 			else
1961 				rule->overload_tbl->pfrkt_flags |=
1962 				    PFR_TFLAG_ACTIVE;
1963 		}
1964 
1965 		pf_mv_kpool(&V_pf_pabuf, &rule->rpool.list);
1966 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1967 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1968 		    (rule->rt > PF_NOPFROUTE)) &&
1969 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1970 			error = EINVAL;
1971 
1972 		if (error) {
1973 			pf_free_rule(rule);
1974 			PF_RULES_WUNLOCK();
1975 			break;
1976 		}
1977 
1978 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1979 		counter_u64_zero(rule->evaluations);
1980 		for (int i = 0; i < 2; i++) {
1981 			counter_u64_zero(rule->packets[i]);
1982 			counter_u64_zero(rule->bytes[i]);
1983 		}
1984 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1985 		    rule, entries);
1986 		ruleset->rules[rs_num].inactive.rcount++;
1987 		PF_RULES_WUNLOCK();
1988 		break;
1989 
1990 #undef ERROUT
1991 DIOCADDRULE_error:
1992 		PF_RULES_WUNLOCK();
1993 		counter_u64_free(rule->evaluations);
1994 		for (int i = 0; i < 2; i++) {
1995 			counter_u64_free(rule->packets[i]);
1996 			counter_u64_free(rule->bytes[i]);
1997 		}
1998 		counter_u64_free(rule->states_cur);
1999 		counter_u64_free(rule->states_tot);
2000 		counter_u64_free(rule->src_nodes);
2001 		free(rule, M_PFRULE);
2002 		if (kif)
2003 			pf_kkif_free(kif);
2004 		break;
2005 	}
2006 
2007 	case DIOCGETRULES: {
2008 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
2009 		struct pf_kruleset	*ruleset;
2010 		struct pf_krule		*tail;
2011 		int			 rs_num;
2012 
2013 		PF_RULES_WLOCK();
2014 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
2015 		ruleset = pf_find_kruleset(pr->anchor);
2016 		if (ruleset == NULL) {
2017 			PF_RULES_WUNLOCK();
2018 			error = EINVAL;
2019 			break;
2020 		}
2021 		rs_num = pf_get_ruleset_number(pr->rule.action);
2022 		if (rs_num >= PF_RULESET_MAX) {
2023 			PF_RULES_WUNLOCK();
2024 			error = EINVAL;
2025 			break;
2026 		}
2027 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
2028 		    pf_krulequeue);
2029 		if (tail)
2030 			pr->nr = tail->nr + 1;
2031 		else
2032 			pr->nr = 0;
2033 		pr->ticket = ruleset->rules[rs_num].active.ticket;
2034 		PF_RULES_WUNLOCK();
2035 		break;
2036 	}
2037 
2038 	case DIOCGETRULE: {
2039 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
2040 		struct pf_kruleset	*ruleset;
2041 		struct pf_krule		*rule;
2042 		int			 rs_num;
2043 
2044 		PF_RULES_WLOCK();
2045 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
2046 		ruleset = pf_find_kruleset(pr->anchor);
2047 		if (ruleset == NULL) {
2048 			PF_RULES_WUNLOCK();
2049 			error = EINVAL;
2050 			break;
2051 		}
2052 		rs_num = pf_get_ruleset_number(pr->rule.action);
2053 		if (rs_num >= PF_RULESET_MAX) {
2054 			PF_RULES_WUNLOCK();
2055 			error = EINVAL;
2056 			break;
2057 		}
2058 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
2059 			PF_RULES_WUNLOCK();
2060 			error = EBUSY;
2061 			break;
2062 		}
2063 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
2064 		while ((rule != NULL) && (rule->nr != pr->nr))
2065 			rule = TAILQ_NEXT(rule, entries);
2066 		if (rule == NULL) {
2067 			PF_RULES_WUNLOCK();
2068 			error = EBUSY;
2069 			break;
2070 		}
2071 
2072 		pf_krule_to_rule(rule, &pr->rule);
2073 
2074 		if (pf_kanchor_copyout(ruleset, rule, pr)) {
2075 			PF_RULES_WUNLOCK();
2076 			error = EBUSY;
2077 			break;
2078 		}
2079 		pf_addr_copyout(&pr->rule.src.addr);
2080 		pf_addr_copyout(&pr->rule.dst.addr);
2081 
2082 		if (pr->action == PF_GET_CLR_CNTR) {
2083 			counter_u64_zero(rule->evaluations);
2084 			for (int i = 0; i < 2; i++) {
2085 				counter_u64_zero(rule->packets[i]);
2086 				counter_u64_zero(rule->bytes[i]);
2087 			}
2088 			counter_u64_zero(rule->states_tot);
2089 		}
2090 		PF_RULES_WUNLOCK();
2091 		break;
2092 	}
2093 
2094 	case DIOCCHANGERULE: {
2095 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
2096 		struct pf_kruleset	*ruleset;
2097 		struct pf_krule		*oldrule = NULL, *newrule = NULL;
2098 		struct pfi_kkif		*kif = NULL;
2099 		struct pf_kpooladdr	*pa;
2100 		u_int32_t		 nr = 0;
2101 		int			 rs_num;
2102 
2103 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
2104 		    pcr->action > PF_CHANGE_GET_TICKET) {
2105 			error = EINVAL;
2106 			break;
2107 		}
2108 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
2109 			error = EINVAL;
2110 			break;
2111 		}
2112 
2113 		if (pcr->action != PF_CHANGE_REMOVE) {
2114 			newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
2115 			error = pf_rule_to_krule(&pcr->rule, newrule);
2116 			if (error != 0) {
2117 				free(newrule, M_PFRULE);
2118 				break;
2119 			}
2120 
2121 			if (newrule->ifname[0])
2122 				kif = pf_kkif_create(M_WAITOK);
2123 			newrule->evaluations = counter_u64_alloc(M_WAITOK);
2124 			for (int i = 0; i < 2; i++) {
2125 				newrule->packets[i] =
2126 				    counter_u64_alloc(M_WAITOK);
2127 				newrule->bytes[i] =
2128 				    counter_u64_alloc(M_WAITOK);
2129 			}
2130 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
2131 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
2132 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
2133 			newrule->cuid = td->td_ucred->cr_ruid;
2134 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
2135 			TAILQ_INIT(&newrule->rpool.list);
2136 		}
2137 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGERULE_error; }
2138 
2139 		PF_RULES_WLOCK();
2140 		if (!(pcr->action == PF_CHANGE_REMOVE ||
2141 		    pcr->action == PF_CHANGE_GET_TICKET) &&
2142 		    pcr->pool_ticket != V_ticket_pabuf)
2143 			ERROUT(EBUSY);
2144 
2145 		ruleset = pf_find_kruleset(pcr->anchor);
2146 		if (ruleset == NULL)
2147 			ERROUT(EINVAL);
2148 
2149 		rs_num = pf_get_ruleset_number(pcr->rule.action);
2150 		if (rs_num >= PF_RULESET_MAX)
2151 			ERROUT(EINVAL);
2152 
2153 		if (pcr->action == PF_CHANGE_GET_TICKET) {
2154 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
2155 			ERROUT(0);
2156 		} else if (pcr->ticket !=
2157 			    ruleset->rules[rs_num].active.ticket)
2158 				ERROUT(EINVAL);
2159 
2160 		if (pcr->action != PF_CHANGE_REMOVE) {
2161 			if (newrule->ifname[0]) {
2162 				newrule->kif = pfi_kkif_attach(kif,
2163 				    newrule->ifname);
2164 				pfi_kkif_ref(newrule->kif);
2165 			} else
2166 				newrule->kif = NULL;
2167 
2168 			if (newrule->rtableid > 0 &&
2169 			    newrule->rtableid >= rt_numfibs)
2170 				error = EBUSY;
2171 
2172 #ifdef ALTQ
2173 			/* set queue IDs */
2174 			if (newrule->qname[0] != 0) {
2175 				if ((newrule->qid =
2176 				    pf_qname2qid(newrule->qname)) == 0)
2177 					error = EBUSY;
2178 				else if (newrule->pqname[0] != 0) {
2179 					if ((newrule->pqid =
2180 					    pf_qname2qid(newrule->pqname)) == 0)
2181 						error = EBUSY;
2182 				} else
2183 					newrule->pqid = newrule->qid;
2184 			}
2185 #endif /* ALTQ */
2186 			if (newrule->tagname[0])
2187 				if ((newrule->tag =
2188 				    pf_tagname2tag(newrule->tagname)) == 0)
2189 					error = EBUSY;
2190 			if (newrule->match_tagname[0])
2191 				if ((newrule->match_tag = pf_tagname2tag(
2192 				    newrule->match_tagname)) == 0)
2193 					error = EBUSY;
2194 			if (newrule->rt && !newrule->direction)
2195 				error = EINVAL;
2196 			if (!newrule->log)
2197 				newrule->logif = 0;
2198 			if (newrule->logif >= PFLOGIFS_MAX)
2199 				error = EINVAL;
2200 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
2201 				error = ENOMEM;
2202 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
2203 				error = ENOMEM;
2204 			if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
2205 				error = EINVAL;
2206 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
2207 				if (pa->addr.type == PF_ADDR_TABLE) {
2208 					pa->addr.p.tbl =
2209 					    pfr_attach_table(ruleset,
2210 					    pa->addr.v.tblname);
2211 					if (pa->addr.p.tbl == NULL)
2212 						error = ENOMEM;
2213 				}
2214 
2215 			newrule->overload_tbl = NULL;
2216 			if (newrule->overload_tblname[0]) {
2217 				if ((newrule->overload_tbl = pfr_attach_table(
2218 				    ruleset, newrule->overload_tblname)) ==
2219 				    NULL)
2220 					error = EINVAL;
2221 				else
2222 					newrule->overload_tbl->pfrkt_flags |=
2223 					    PFR_TFLAG_ACTIVE;
2224 			}
2225 
2226 			pf_mv_kpool(&V_pf_pabuf, &newrule->rpool.list);
2227 			if (((((newrule->action == PF_NAT) ||
2228 			    (newrule->action == PF_RDR) ||
2229 			    (newrule->action == PF_BINAT) ||
2230 			    (newrule->rt > PF_NOPFROUTE)) &&
2231 			    !newrule->anchor)) &&
2232 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
2233 				error = EINVAL;
2234 
2235 			if (error) {
2236 				pf_free_rule(newrule);
2237 				PF_RULES_WUNLOCK();
2238 				break;
2239 			}
2240 
2241 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
2242 		}
2243 		pf_empty_kpool(&V_pf_pabuf);
2244 
2245 		if (pcr->action == PF_CHANGE_ADD_HEAD)
2246 			oldrule = TAILQ_FIRST(
2247 			    ruleset->rules[rs_num].active.ptr);
2248 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
2249 			oldrule = TAILQ_LAST(
2250 			    ruleset->rules[rs_num].active.ptr, pf_krulequeue);
2251 		else {
2252 			oldrule = TAILQ_FIRST(
2253 			    ruleset->rules[rs_num].active.ptr);
2254 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
2255 				oldrule = TAILQ_NEXT(oldrule, entries);
2256 			if (oldrule == NULL) {
2257 				if (newrule != NULL)
2258 					pf_free_rule(newrule);
2259 				PF_RULES_WUNLOCK();
2260 				error = EINVAL;
2261 				break;
2262 			}
2263 		}
2264 
2265 		if (pcr->action == PF_CHANGE_REMOVE) {
2266 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
2267 			    oldrule);
2268 			ruleset->rules[rs_num].active.rcount--;
2269 		} else {
2270 			if (oldrule == NULL)
2271 				TAILQ_INSERT_TAIL(
2272 				    ruleset->rules[rs_num].active.ptr,
2273 				    newrule, entries);
2274 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
2275 			    pcr->action == PF_CHANGE_ADD_BEFORE)
2276 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
2277 			else
2278 				TAILQ_INSERT_AFTER(
2279 				    ruleset->rules[rs_num].active.ptr,
2280 				    oldrule, newrule, entries);
2281 			ruleset->rules[rs_num].active.rcount++;
2282 		}
2283 
2284 		nr = 0;
2285 		TAILQ_FOREACH(oldrule,
2286 		    ruleset->rules[rs_num].active.ptr, entries)
2287 			oldrule->nr = nr++;
2288 
2289 		ruleset->rules[rs_num].active.ticket++;
2290 
2291 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
2292 		pf_remove_if_empty_kruleset(ruleset);
2293 
2294 		PF_RULES_WUNLOCK();
2295 		break;
2296 
2297 #undef ERROUT
2298 DIOCCHANGERULE_error:
2299 		PF_RULES_WUNLOCK();
2300 		if (newrule != NULL) {
2301 			counter_u64_free(newrule->evaluations);
2302 			for (int i = 0; i < 2; i++) {
2303 				counter_u64_free(newrule->packets[i]);
2304 				counter_u64_free(newrule->bytes[i]);
2305 			}
2306 			counter_u64_free(newrule->states_cur);
2307 			counter_u64_free(newrule->states_tot);
2308 			counter_u64_free(newrule->src_nodes);
2309 			free(newrule, M_PFRULE);
2310 		}
2311 		if (kif != NULL)
2312 			pf_kkif_free(kif);
2313 		break;
2314 	}
2315 
2316 	case DIOCCLRSTATES: {
2317 		struct pf_state		*s;
2318 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
2319 		u_int			 i, killed = 0;
2320 
2321 		for (i = 0; i <= pf_hashmask; i++) {
2322 			struct pf_idhash *ih = &V_pf_idhash[i];
2323 
2324 relock_DIOCCLRSTATES:
2325 			PF_HASHROW_LOCK(ih);
2326 			LIST_FOREACH(s, &ih->states, entry)
2327 				if (!psk->psk_ifname[0] ||
2328 				    !strcmp(psk->psk_ifname,
2329 				    s->kif->pfik_name)) {
2330 					/*
2331 					 * Don't send out individual
2332 					 * delete messages.
2333 					 */
2334 					s->state_flags |= PFSTATE_NOSYNC;
2335 					pf_unlink_state(s, PF_ENTER_LOCKED);
2336 					killed++;
2337 					goto relock_DIOCCLRSTATES;
2338 				}
2339 			PF_HASHROW_UNLOCK(ih);
2340 		}
2341 		psk->psk_killed = killed;
2342 		if (V_pfsync_clear_states_ptr != NULL)
2343 			V_pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
2344 		break;
2345 	}
2346 
2347 	case DIOCKILLSTATES: {
2348 		struct pf_state		*s;
2349 		struct pf_state_key	*sk;
2350 		struct pf_addr		*srcaddr, *dstaddr;
2351 		u_int16_t		 srcport, dstport;
2352 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
2353 		u_int			 i, killed = 0;
2354 
2355 		if (psk->psk_pfcmp.id) {
2356 			if (psk->psk_pfcmp.creatorid == 0)
2357 				psk->psk_pfcmp.creatorid = V_pf_status.hostid;
2358 			if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
2359 			    psk->psk_pfcmp.creatorid))) {
2360 				pf_unlink_state(s, PF_ENTER_LOCKED);
2361 				psk->psk_killed = 1;
2362 			}
2363 			break;
2364 		}
2365 
2366 		for (i = 0; i <= pf_hashmask; i++) {
2367 			struct pf_idhash *ih = &V_pf_idhash[i];
2368 
2369 relock_DIOCKILLSTATES:
2370 			PF_HASHROW_LOCK(ih);
2371 			LIST_FOREACH(s, &ih->states, entry) {
2372 				sk = s->key[PF_SK_WIRE];
2373 				if (s->direction == PF_OUT) {
2374 					srcaddr = &sk->addr[1];
2375 					dstaddr = &sk->addr[0];
2376 					srcport = sk->port[1];
2377 					dstport = sk->port[0];
2378 				} else {
2379 					srcaddr = &sk->addr[0];
2380 					dstaddr = &sk->addr[1];
2381 					srcport = sk->port[0];
2382 					dstport = sk->port[1];
2383 				}
2384 
2385 				if ((!psk->psk_af || sk->af == psk->psk_af)
2386 				    && (!psk->psk_proto || psk->psk_proto ==
2387 				    sk->proto) &&
2388 				    PF_MATCHA(psk->psk_src.neg,
2389 				    &psk->psk_src.addr.v.a.addr,
2390 				    &psk->psk_src.addr.v.a.mask,
2391 				    srcaddr, sk->af) &&
2392 				    PF_MATCHA(psk->psk_dst.neg,
2393 				    &psk->psk_dst.addr.v.a.addr,
2394 				    &psk->psk_dst.addr.v.a.mask,
2395 				    dstaddr, sk->af) &&
2396 				    (psk->psk_src.port_op == 0 ||
2397 				    pf_match_port(psk->psk_src.port_op,
2398 				    psk->psk_src.port[0], psk->psk_src.port[1],
2399 				    srcport)) &&
2400 				    (psk->psk_dst.port_op == 0 ||
2401 				    pf_match_port(psk->psk_dst.port_op,
2402 				    psk->psk_dst.port[0], psk->psk_dst.port[1],
2403 				    dstport)) &&
2404 				    (!psk->psk_label[0] ||
2405 				    (s->rule.ptr->label[0] &&
2406 				    !strcmp(psk->psk_label,
2407 				    s->rule.ptr->label))) &&
2408 				    (!psk->psk_ifname[0] ||
2409 				    !strcmp(psk->psk_ifname,
2410 				    s->kif->pfik_name))) {
2411 					pf_unlink_state(s, PF_ENTER_LOCKED);
2412 					killed++;
2413 					goto relock_DIOCKILLSTATES;
2414 				}
2415 			}
2416 			PF_HASHROW_UNLOCK(ih);
2417 		}
2418 		psk->psk_killed = killed;
2419 		break;
2420 	}
2421 
2422 	case DIOCADDSTATE: {
2423 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
2424 		struct pfsync_state	*sp = &ps->state;
2425 
2426 		if (sp->timeout >= PFTM_MAX) {
2427 			error = EINVAL;
2428 			break;
2429 		}
2430 		if (V_pfsync_state_import_ptr != NULL) {
2431 			PF_RULES_RLOCK();
2432 			error = V_pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
2433 			PF_RULES_RUNLOCK();
2434 		} else
2435 			error = EOPNOTSUPP;
2436 		break;
2437 	}
2438 
2439 	case DIOCGETSTATE: {
2440 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
2441 		struct pf_state		*s;
2442 
2443 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
2444 		if (s == NULL) {
2445 			error = ENOENT;
2446 			break;
2447 		}
2448 
2449 		pfsync_state_export(&ps->state, s);
2450 		PF_STATE_UNLOCK(s);
2451 		break;
2452 	}
2453 
2454 	case DIOCGETSTATES: {
2455 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
2456 		struct pf_state		*s;
2457 		struct pfsync_state	*pstore, *p;
2458 		int i, nr;
2459 
2460 		if (ps->ps_len <= 0) {
2461 			nr = uma_zone_get_cur(V_pf_state_z);
2462 			ps->ps_len = sizeof(struct pfsync_state) * nr;
2463 			break;
2464 		}
2465 
2466 		p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK | M_ZERO);
2467 		nr = 0;
2468 
2469 		for (i = 0; i <= pf_hashmask; i++) {
2470 			struct pf_idhash *ih = &V_pf_idhash[i];
2471 
2472 			PF_HASHROW_LOCK(ih);
2473 			LIST_FOREACH(s, &ih->states, entry) {
2474 				if (s->timeout == PFTM_UNLINKED)
2475 					continue;
2476 
2477 				if ((nr+1) * sizeof(*p) > ps->ps_len) {
2478 					PF_HASHROW_UNLOCK(ih);
2479 					goto DIOCGETSTATES_full;
2480 				}
2481 				pfsync_state_export(p, s);
2482 				p++;
2483 				nr++;
2484 			}
2485 			PF_HASHROW_UNLOCK(ih);
2486 		}
2487 DIOCGETSTATES_full:
2488 		error = copyout(pstore, ps->ps_states,
2489 		    sizeof(struct pfsync_state) * nr);
2490 		if (error) {
2491 			free(pstore, M_TEMP);
2492 			break;
2493 		}
2494 		ps->ps_len = sizeof(struct pfsync_state) * nr;
2495 		free(pstore, M_TEMP);
2496 
2497 		break;
2498 	}
2499 
2500 	case DIOCGETSTATUS: {
2501 		struct pf_status *s = (struct pf_status *)addr;
2502 
2503 		PF_RULES_RLOCK();
2504 		s->running = V_pf_status.running;
2505 		s->since   = V_pf_status.since;
2506 		s->debug   = V_pf_status.debug;
2507 		s->hostid  = V_pf_status.hostid;
2508 		s->states  = V_pf_status.states;
2509 		s->src_nodes = V_pf_status.src_nodes;
2510 
2511 		for (int i = 0; i < PFRES_MAX; i++)
2512 			s->counters[i] =
2513 			    counter_u64_fetch(V_pf_status.counters[i]);
2514 		for (int i = 0; i < LCNT_MAX; i++)
2515 			s->lcounters[i] =
2516 			    counter_u64_fetch(V_pf_status.lcounters[i]);
2517 		for (int i = 0; i < FCNT_MAX; i++)
2518 			s->fcounters[i] =
2519 			    counter_u64_fetch(V_pf_status.fcounters[i]);
2520 		for (int i = 0; i < SCNT_MAX; i++)
2521 			s->scounters[i] =
2522 			    counter_u64_fetch(V_pf_status.scounters[i]);
2523 
2524 		bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ);
2525 		bcopy(V_pf_status.pf_chksum, s->pf_chksum,
2526 		    PF_MD5_DIGEST_LENGTH);
2527 
2528 		pfi_update_status(s->ifname, s);
2529 		PF_RULES_RUNLOCK();
2530 		break;
2531 	}
2532 
2533 	case DIOCSETSTATUSIF: {
2534 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
2535 
2536 		if (pi->ifname[0] == 0) {
2537 			bzero(V_pf_status.ifname, IFNAMSIZ);
2538 			break;
2539 		}
2540 		PF_RULES_WLOCK();
2541 		strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
2542 		PF_RULES_WUNLOCK();
2543 		break;
2544 	}
2545 
2546 	case DIOCCLRSTATUS: {
2547 		PF_RULES_WLOCK();
2548 		for (int i = 0; i < PFRES_MAX; i++)
2549 			counter_u64_zero(V_pf_status.counters[i]);
2550 		for (int i = 0; i < FCNT_MAX; i++)
2551 			counter_u64_zero(V_pf_status.fcounters[i]);
2552 		for (int i = 0; i < SCNT_MAX; i++)
2553 			counter_u64_zero(V_pf_status.scounters[i]);
2554 		for (int i = 0; i < LCNT_MAX; i++)
2555 			counter_u64_zero(V_pf_status.lcounters[i]);
2556 		V_pf_status.since = time_second;
2557 		if (*V_pf_status.ifname)
2558 			pfi_update_status(V_pf_status.ifname, NULL);
2559 		PF_RULES_WUNLOCK();
2560 		break;
2561 	}
2562 
2563 	case DIOCNATLOOK: {
2564 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
2565 		struct pf_state_key	*sk;
2566 		struct pf_state		*state;
2567 		struct pf_state_key_cmp	 key;
2568 		int			 m = 0, direction = pnl->direction;
2569 		int			 sidx, didx;
2570 
2571 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
2572 		sidx = (direction == PF_IN) ? 1 : 0;
2573 		didx = (direction == PF_IN) ? 0 : 1;
2574 
2575 		if (!pnl->proto ||
2576 		    PF_AZERO(&pnl->saddr, pnl->af) ||
2577 		    PF_AZERO(&pnl->daddr, pnl->af) ||
2578 		    ((pnl->proto == IPPROTO_TCP ||
2579 		    pnl->proto == IPPROTO_UDP) &&
2580 		    (!pnl->dport || !pnl->sport)))
2581 			error = EINVAL;
2582 		else {
2583 			bzero(&key, sizeof(key));
2584 			key.af = pnl->af;
2585 			key.proto = pnl->proto;
2586 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
2587 			key.port[sidx] = pnl->sport;
2588 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
2589 			key.port[didx] = pnl->dport;
2590 
2591 			state = pf_find_state_all(&key, direction, &m);
2592 
2593 			if (m > 1)
2594 				error = E2BIG;	/* more than one state */
2595 			else if (state != NULL) {
2596 				/* XXXGL: not locked read */
2597 				sk = state->key[sidx];
2598 				PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
2599 				pnl->rsport = sk->port[sidx];
2600 				PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
2601 				pnl->rdport = sk->port[didx];
2602 			} else
2603 				error = ENOENT;
2604 		}
2605 		break;
2606 	}
2607 
2608 	case DIOCSETTIMEOUT: {
2609 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2610 		int		 old;
2611 
2612 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
2613 		    pt->seconds < 0) {
2614 			error = EINVAL;
2615 			break;
2616 		}
2617 		PF_RULES_WLOCK();
2618 		old = V_pf_default_rule.timeout[pt->timeout];
2619 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
2620 			pt->seconds = 1;
2621 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
2622 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
2623 			wakeup(pf_purge_thread);
2624 		pt->seconds = old;
2625 		PF_RULES_WUNLOCK();
2626 		break;
2627 	}
2628 
2629 	case DIOCGETTIMEOUT: {
2630 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2631 
2632 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
2633 			error = EINVAL;
2634 			break;
2635 		}
2636 		PF_RULES_RLOCK();
2637 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
2638 		PF_RULES_RUNLOCK();
2639 		break;
2640 	}
2641 
2642 	case DIOCGETLIMIT: {
2643 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2644 
2645 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
2646 			error = EINVAL;
2647 			break;
2648 		}
2649 		PF_RULES_RLOCK();
2650 		pl->limit = V_pf_limits[pl->index].limit;
2651 		PF_RULES_RUNLOCK();
2652 		break;
2653 	}
2654 
2655 	case DIOCSETLIMIT: {
2656 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2657 		int			 old_limit;
2658 
2659 		PF_RULES_WLOCK();
2660 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
2661 		    V_pf_limits[pl->index].zone == NULL) {
2662 			PF_RULES_WUNLOCK();
2663 			error = EINVAL;
2664 			break;
2665 		}
2666 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
2667 		old_limit = V_pf_limits[pl->index].limit;
2668 		V_pf_limits[pl->index].limit = pl->limit;
2669 		pl->limit = old_limit;
2670 		PF_RULES_WUNLOCK();
2671 		break;
2672 	}
2673 
2674 	case DIOCSETDEBUG: {
2675 		u_int32_t	*level = (u_int32_t *)addr;
2676 
2677 		PF_RULES_WLOCK();
2678 		V_pf_status.debug = *level;
2679 		PF_RULES_WUNLOCK();
2680 		break;
2681 	}
2682 
2683 	case DIOCCLRRULECTRS: {
2684 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
2685 		struct pf_kruleset	*ruleset = &pf_main_ruleset;
2686 		struct pf_krule		*rule;
2687 
2688 		PF_RULES_WLOCK();
2689 		TAILQ_FOREACH(rule,
2690 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
2691 			counter_u64_zero(rule->evaluations);
2692 			for (int i = 0; i < 2; i++) {
2693 				counter_u64_zero(rule->packets[i]);
2694 				counter_u64_zero(rule->bytes[i]);
2695 			}
2696 		}
2697 		PF_RULES_WUNLOCK();
2698 		break;
2699 	}
2700 
2701 	case DIOCGIFSPEEDV0:
2702 	case DIOCGIFSPEEDV1: {
2703 		struct pf_ifspeed_v1	*psp = (struct pf_ifspeed_v1 *)addr;
2704 		struct pf_ifspeed_v1	ps;
2705 		struct ifnet		*ifp;
2706 
2707 		if (psp->ifname[0] != 0) {
2708 			/* Can we completely trust user-land? */
2709 			strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
2710 			ifp = ifunit(ps.ifname);
2711 			if (ifp != NULL) {
2712 				psp->baudrate32 =
2713 				    (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
2714 				if (cmd == DIOCGIFSPEEDV1)
2715 					psp->baudrate = ifp->if_baudrate;
2716 			} else
2717 				error = EINVAL;
2718 		} else
2719 			error = EINVAL;
2720 		break;
2721 	}
2722 
2723 #ifdef ALTQ
2724 	case DIOCSTARTALTQ: {
2725 		struct pf_altq		*altq;
2726 
2727 		PF_RULES_WLOCK();
2728 		/* enable all altq interfaces on active list */
2729 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
2730 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
2731 				error = pf_enable_altq(altq);
2732 				if (error != 0)
2733 					break;
2734 			}
2735 		}
2736 		if (error == 0)
2737 			V_pf_altq_running = 1;
2738 		PF_RULES_WUNLOCK();
2739 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
2740 		break;
2741 	}
2742 
2743 	case DIOCSTOPALTQ: {
2744 		struct pf_altq		*altq;
2745 
2746 		PF_RULES_WLOCK();
2747 		/* disable all altq interfaces on active list */
2748 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
2749 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
2750 				error = pf_disable_altq(altq);
2751 				if (error != 0)
2752 					break;
2753 			}
2754 		}
2755 		if (error == 0)
2756 			V_pf_altq_running = 0;
2757 		PF_RULES_WUNLOCK();
2758 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2759 		break;
2760 	}
2761 
2762 	case DIOCADDALTQV0:
2763 	case DIOCADDALTQV1: {
2764 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2765 		struct pf_altq		*altq, *a;
2766 		struct ifnet		*ifp;
2767 
2768 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
2769 		error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
2770 		if (error)
2771 			break;
2772 		altq->local_flags = 0;
2773 
2774 		PF_RULES_WLOCK();
2775 		if (pa->ticket != V_ticket_altqs_inactive) {
2776 			PF_RULES_WUNLOCK();
2777 			free(altq, M_PFALTQ);
2778 			error = EBUSY;
2779 			break;
2780 		}
2781 
2782 		/*
2783 		 * if this is for a queue, find the discipline and
2784 		 * copy the necessary fields
2785 		 */
2786 		if (altq->qname[0] != 0) {
2787 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2788 				PF_RULES_WUNLOCK();
2789 				error = EBUSY;
2790 				free(altq, M_PFALTQ);
2791 				break;
2792 			}
2793 			altq->altq_disc = NULL;
2794 			TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
2795 				if (strncmp(a->ifname, altq->ifname,
2796 				    IFNAMSIZ) == 0) {
2797 					altq->altq_disc = a->altq_disc;
2798 					break;
2799 				}
2800 			}
2801 		}
2802 
2803 		if ((ifp = ifunit(altq->ifname)) == NULL)
2804 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
2805 		else
2806 			error = altq_add(ifp, altq);
2807 
2808 		if (error) {
2809 			PF_RULES_WUNLOCK();
2810 			free(altq, M_PFALTQ);
2811 			break;
2812 		}
2813 
2814 		if (altq->qname[0] != 0)
2815 			TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
2816 		else
2817 			TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
2818 		/* version error check done on import above */
2819 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
2820 		PF_RULES_WUNLOCK();
2821 		break;
2822 	}
2823 
2824 	case DIOCGETALTQSV0:
2825 	case DIOCGETALTQSV1: {
2826 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2827 		struct pf_altq		*altq;
2828 
2829 		PF_RULES_RLOCK();
2830 		pa->nr = 0;
2831 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
2832 			pa->nr++;
2833 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
2834 			pa->nr++;
2835 		pa->ticket = V_ticket_altqs_active;
2836 		PF_RULES_RUNLOCK();
2837 		break;
2838 	}
2839 
2840 	case DIOCGETALTQV0:
2841 	case DIOCGETALTQV1: {
2842 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2843 		struct pf_altq		*altq;
2844 
2845 		PF_RULES_RLOCK();
2846 		if (pa->ticket != V_ticket_altqs_active) {
2847 			PF_RULES_RUNLOCK();
2848 			error = EBUSY;
2849 			break;
2850 		}
2851 		altq = pf_altq_get_nth_active(pa->nr);
2852 		if (altq == NULL) {
2853 			PF_RULES_RUNLOCK();
2854 			error = EBUSY;
2855 			break;
2856 		}
2857 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
2858 		PF_RULES_RUNLOCK();
2859 		break;
2860 	}
2861 
2862 	case DIOCCHANGEALTQV0:
2863 	case DIOCCHANGEALTQV1:
2864 		/* CHANGEALTQ not supported yet! */
2865 		error = ENODEV;
2866 		break;
2867 
2868 	case DIOCGETQSTATSV0:
2869 	case DIOCGETQSTATSV1: {
2870 		struct pfioc_qstats_v1	*pq = (struct pfioc_qstats_v1 *)addr;
2871 		struct pf_altq		*altq;
2872 		int			 nbytes;
2873 		u_int32_t		 version;
2874 
2875 		PF_RULES_RLOCK();
2876 		if (pq->ticket != V_ticket_altqs_active) {
2877 			PF_RULES_RUNLOCK();
2878 			error = EBUSY;
2879 			break;
2880 		}
2881 		nbytes = pq->nbytes;
2882 		altq = pf_altq_get_nth_active(pq->nr);
2883 		if (altq == NULL) {
2884 			PF_RULES_RUNLOCK();
2885 			error = EBUSY;
2886 			break;
2887 		}
2888 
2889 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
2890 			PF_RULES_RUNLOCK();
2891 			error = ENXIO;
2892 			break;
2893 		}
2894 		PF_RULES_RUNLOCK();
2895 		if (cmd == DIOCGETQSTATSV0)
2896 			version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
2897 		else
2898 			version = pq->version;
2899 		error = altq_getqstats(altq, pq->buf, &nbytes, version);
2900 		if (error == 0) {
2901 			pq->scheduler = altq->scheduler;
2902 			pq->nbytes = nbytes;
2903 		}
2904 		break;
2905 	}
2906 #endif /* ALTQ */
2907 
2908 	case DIOCBEGINADDRS: {
2909 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2910 
2911 		PF_RULES_WLOCK();
2912 		pf_empty_kpool(&V_pf_pabuf);
2913 		pp->ticket = ++V_ticket_pabuf;
2914 		PF_RULES_WUNLOCK();
2915 		break;
2916 	}
2917 
2918 	case DIOCADDADDR: {
2919 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2920 		struct pf_kpooladdr	*pa;
2921 		struct pfi_kkif		*kif = NULL;
2922 
2923 #ifndef INET
2924 		if (pp->af == AF_INET) {
2925 			error = EAFNOSUPPORT;
2926 			break;
2927 		}
2928 #endif /* INET */
2929 #ifndef INET6
2930 		if (pp->af == AF_INET6) {
2931 			error = EAFNOSUPPORT;
2932 			break;
2933 		}
2934 #endif /* INET6 */
2935 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2936 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2937 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2938 			error = EINVAL;
2939 			break;
2940 		}
2941 		if (pp->addr.addr.p.dyn != NULL) {
2942 			error = EINVAL;
2943 			break;
2944 		}
2945 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2946 		pf_pooladdr_to_kpooladdr(&pp->addr, pa);
2947 		if (pa->ifname[0])
2948 			kif = pf_kkif_create(M_WAITOK);
2949 		PF_RULES_WLOCK();
2950 		if (pp->ticket != V_ticket_pabuf) {
2951 			PF_RULES_WUNLOCK();
2952 			if (pa->ifname[0])
2953 				pf_kkif_free(kif);
2954 			free(pa, M_PFRULE);
2955 			error = EBUSY;
2956 			break;
2957 		}
2958 		if (pa->ifname[0]) {
2959 			pa->kif = pfi_kkif_attach(kif, pa->ifname);
2960 			pfi_kkif_ref(pa->kif);
2961 		} else
2962 			pa->kif = NULL;
2963 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2964 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2965 			if (pa->ifname[0])
2966 				pfi_kkif_unref(pa->kif);
2967 			PF_RULES_WUNLOCK();
2968 			free(pa, M_PFRULE);
2969 			break;
2970 		}
2971 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2972 		PF_RULES_WUNLOCK();
2973 		break;
2974 	}
2975 
2976 	case DIOCGETADDRS: {
2977 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2978 		struct pf_kpool		*pool;
2979 		struct pf_kpooladdr	*pa;
2980 
2981 		PF_RULES_RLOCK();
2982 		pp->nr = 0;
2983 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2984 		    pp->r_num, 0, 1, 0);
2985 		if (pool == NULL) {
2986 			PF_RULES_RUNLOCK();
2987 			error = EBUSY;
2988 			break;
2989 		}
2990 		TAILQ_FOREACH(pa, &pool->list, entries)
2991 			pp->nr++;
2992 		PF_RULES_RUNLOCK();
2993 		break;
2994 	}
2995 
2996 	case DIOCGETADDR: {
2997 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2998 		struct pf_kpool		*pool;
2999 		struct pf_kpooladdr	*pa;
3000 		u_int32_t		 nr = 0;
3001 
3002 		PF_RULES_RLOCK();
3003 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
3004 		    pp->r_num, 0, 1, 1);
3005 		if (pool == NULL) {
3006 			PF_RULES_RUNLOCK();
3007 			error = EBUSY;
3008 			break;
3009 		}
3010 		pa = TAILQ_FIRST(&pool->list);
3011 		while ((pa != NULL) && (nr < pp->nr)) {
3012 			pa = TAILQ_NEXT(pa, entries);
3013 			nr++;
3014 		}
3015 		if (pa == NULL) {
3016 			PF_RULES_RUNLOCK();
3017 			error = EBUSY;
3018 			break;
3019 		}
3020 		pf_kpooladdr_to_pooladdr(pa, &pp->addr);
3021 		pf_addr_copyout(&pp->addr.addr);
3022 		PF_RULES_RUNLOCK();
3023 		break;
3024 	}
3025 
3026 	case DIOCCHANGEADDR: {
3027 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
3028 		struct pf_kpool		*pool;
3029 		struct pf_kpooladdr	*oldpa = NULL, *newpa = NULL;
3030 		struct pf_kruleset	*ruleset;
3031 		struct pfi_kkif		*kif = NULL;
3032 
3033 		if (pca->action < PF_CHANGE_ADD_HEAD ||
3034 		    pca->action > PF_CHANGE_REMOVE) {
3035 			error = EINVAL;
3036 			break;
3037 		}
3038 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
3039 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
3040 		    pca->addr.addr.type != PF_ADDR_TABLE) {
3041 			error = EINVAL;
3042 			break;
3043 		}
3044 		if (pca->addr.addr.p.dyn != NULL) {
3045 			error = EINVAL;
3046 			break;
3047 		}
3048 
3049 		if (pca->action != PF_CHANGE_REMOVE) {
3050 #ifndef INET
3051 			if (pca->af == AF_INET) {
3052 				error = EAFNOSUPPORT;
3053 				break;
3054 			}
3055 #endif /* INET */
3056 #ifndef INET6
3057 			if (pca->af == AF_INET6) {
3058 				error = EAFNOSUPPORT;
3059 				break;
3060 			}
3061 #endif /* INET6 */
3062 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
3063 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
3064 			if (newpa->ifname[0])
3065 				kif = pf_kkif_create(M_WAITOK);
3066 			newpa->kif = NULL;
3067 		}
3068 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGEADDR_error; }
3069 		PF_RULES_WLOCK();
3070 		ruleset = pf_find_kruleset(pca->anchor);
3071 		if (ruleset == NULL)
3072 			ERROUT(EBUSY);
3073 
3074 		pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
3075 		    pca->r_num, pca->r_last, 1, 1);
3076 		if (pool == NULL)
3077 			ERROUT(EBUSY);
3078 
3079 		if (pca->action != PF_CHANGE_REMOVE) {
3080 			if (newpa->ifname[0]) {
3081 				newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
3082 				pfi_kkif_ref(newpa->kif);
3083 				kif = NULL;
3084 			}
3085 
3086 			switch (newpa->addr.type) {
3087 			case PF_ADDR_DYNIFTL:
3088 				error = pfi_dynaddr_setup(&newpa->addr,
3089 				    pca->af);
3090 				break;
3091 			case PF_ADDR_TABLE:
3092 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
3093 				    newpa->addr.v.tblname);
3094 				if (newpa->addr.p.tbl == NULL)
3095 					error = ENOMEM;
3096 				break;
3097 			}
3098 			if (error)
3099 				goto DIOCCHANGEADDR_error;
3100 		}
3101 
3102 		switch (pca->action) {
3103 		case PF_CHANGE_ADD_HEAD:
3104 			oldpa = TAILQ_FIRST(&pool->list);
3105 			break;
3106 		case PF_CHANGE_ADD_TAIL:
3107 			oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
3108 			break;
3109 		default:
3110 			oldpa = TAILQ_FIRST(&pool->list);
3111 			for (int i = 0; oldpa && i < pca->nr; i++)
3112 				oldpa = TAILQ_NEXT(oldpa, entries);
3113 
3114 			if (oldpa == NULL)
3115 				ERROUT(EINVAL);
3116 		}
3117 
3118 		if (pca->action == PF_CHANGE_REMOVE) {
3119 			TAILQ_REMOVE(&pool->list, oldpa, entries);
3120 			switch (oldpa->addr.type) {
3121 			case PF_ADDR_DYNIFTL:
3122 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
3123 				break;
3124 			case PF_ADDR_TABLE:
3125 				pfr_detach_table(oldpa->addr.p.tbl);
3126 				break;
3127 			}
3128 			if (oldpa->kif)
3129 				pfi_kkif_unref(oldpa->kif);
3130 			free(oldpa, M_PFRULE);
3131 		} else {
3132 			if (oldpa == NULL)
3133 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
3134 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
3135 			    pca->action == PF_CHANGE_ADD_BEFORE)
3136 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
3137 			else
3138 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
3139 				    newpa, entries);
3140 		}
3141 
3142 		pool->cur = TAILQ_FIRST(&pool->list);
3143 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
3144 		PF_RULES_WUNLOCK();
3145 		break;
3146 
3147 #undef ERROUT
3148 DIOCCHANGEADDR_error:
3149 		if (newpa != NULL) {
3150 			if (newpa->kif)
3151 				pfi_kkif_unref(newpa->kif);
3152 			free(newpa, M_PFRULE);
3153 		}
3154 		PF_RULES_WUNLOCK();
3155 		if (kif != NULL)
3156 			pf_kkif_free(kif);
3157 		break;
3158 	}
3159 
3160 	case DIOCGETRULESETS: {
3161 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
3162 		struct pf_kruleset	*ruleset;
3163 		struct pf_kanchor	*anchor;
3164 
3165 		PF_RULES_RLOCK();
3166 		pr->path[sizeof(pr->path) - 1] = 0;
3167 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
3168 			PF_RULES_RUNLOCK();
3169 			error = ENOENT;
3170 			break;
3171 		}
3172 		pr->nr = 0;
3173 		if (ruleset->anchor == NULL) {
3174 			/* XXX kludge for pf_main_ruleset */
3175 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
3176 				if (anchor->parent == NULL)
3177 					pr->nr++;
3178 		} else {
3179 			RB_FOREACH(anchor, pf_kanchor_node,
3180 			    &ruleset->anchor->children)
3181 				pr->nr++;
3182 		}
3183 		PF_RULES_RUNLOCK();
3184 		break;
3185 	}
3186 
3187 	case DIOCGETRULESET: {
3188 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
3189 		struct pf_kruleset	*ruleset;
3190 		struct pf_kanchor	*anchor;
3191 		u_int32_t		 nr = 0;
3192 
3193 		PF_RULES_RLOCK();
3194 		pr->path[sizeof(pr->path) - 1] = 0;
3195 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
3196 			PF_RULES_RUNLOCK();
3197 			error = ENOENT;
3198 			break;
3199 		}
3200 		pr->name[0] = 0;
3201 		if (ruleset->anchor == NULL) {
3202 			/* XXX kludge for pf_main_ruleset */
3203 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
3204 				if (anchor->parent == NULL && nr++ == pr->nr) {
3205 					strlcpy(pr->name, anchor->name,
3206 					    sizeof(pr->name));
3207 					break;
3208 				}
3209 		} else {
3210 			RB_FOREACH(anchor, pf_kanchor_node,
3211 			    &ruleset->anchor->children)
3212 				if (nr++ == pr->nr) {
3213 					strlcpy(pr->name, anchor->name,
3214 					    sizeof(pr->name));
3215 					break;
3216 				}
3217 		}
3218 		if (!pr->name[0])
3219 			error = EBUSY;
3220 		PF_RULES_RUNLOCK();
3221 		break;
3222 	}
3223 
3224 	case DIOCRCLRTABLES: {
3225 		struct pfioc_table *io = (struct pfioc_table *)addr;
3226 
3227 		if (io->pfrio_esize != 0) {
3228 			error = ENODEV;
3229 			break;
3230 		}
3231 		PF_RULES_WLOCK();
3232 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
3233 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
3234 		PF_RULES_WUNLOCK();
3235 		break;
3236 	}
3237 
3238 	case DIOCRADDTABLES: {
3239 		struct pfioc_table *io = (struct pfioc_table *)addr;
3240 		struct pfr_table *pfrts;
3241 		size_t totlen;
3242 
3243 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3244 			error = ENODEV;
3245 			break;
3246 		}
3247 
3248 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
3249 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
3250 			error = ENOMEM;
3251 			break;
3252 		}
3253 
3254 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3255 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3256 		    M_TEMP, M_WAITOK);
3257 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3258 		if (error) {
3259 			free(pfrts, M_TEMP);
3260 			break;
3261 		}
3262 		PF_RULES_WLOCK();
3263 		error = pfr_add_tables(pfrts, io->pfrio_size,
3264 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3265 		PF_RULES_WUNLOCK();
3266 		free(pfrts, M_TEMP);
3267 		break;
3268 	}
3269 
3270 	case DIOCRDELTABLES: {
3271 		struct pfioc_table *io = (struct pfioc_table *)addr;
3272 		struct pfr_table *pfrts;
3273 		size_t totlen;
3274 
3275 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3276 			error = ENODEV;
3277 			break;
3278 		}
3279 
3280 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
3281 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
3282 			error = ENOMEM;
3283 			break;
3284 		}
3285 
3286 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3287 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3288 		    M_TEMP, M_WAITOK);
3289 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3290 		if (error) {
3291 			free(pfrts, M_TEMP);
3292 			break;
3293 		}
3294 		PF_RULES_WLOCK();
3295 		error = pfr_del_tables(pfrts, io->pfrio_size,
3296 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3297 		PF_RULES_WUNLOCK();
3298 		free(pfrts, M_TEMP);
3299 		break;
3300 	}
3301 
3302 	case DIOCRGETTABLES: {
3303 		struct pfioc_table *io = (struct pfioc_table *)addr;
3304 		struct pfr_table *pfrts;
3305 		size_t totlen;
3306 		int n;
3307 
3308 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3309 			error = ENODEV;
3310 			break;
3311 		}
3312 		PF_RULES_RLOCK();
3313 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3314 		if (n < 0) {
3315 			PF_RULES_RUNLOCK();
3316 			error = EINVAL;
3317 			break;
3318 		}
3319 		io->pfrio_size = min(io->pfrio_size, n);
3320 
3321 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3322 
3323 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3324 		    M_TEMP, M_NOWAIT);
3325 		if (pfrts == NULL) {
3326 			error = ENOMEM;
3327 			PF_RULES_RUNLOCK();
3328 			break;
3329 		}
3330 		error = pfr_get_tables(&io->pfrio_table, pfrts,
3331 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3332 		PF_RULES_RUNLOCK();
3333 		if (error == 0)
3334 			error = copyout(pfrts, io->pfrio_buffer, totlen);
3335 		free(pfrts, M_TEMP);
3336 		break;
3337 	}
3338 
3339 	case DIOCRGETTSTATS: {
3340 		struct pfioc_table *io = (struct pfioc_table *)addr;
3341 		struct pfr_tstats *pfrtstats;
3342 		size_t totlen;
3343 		int n;
3344 
3345 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
3346 			error = ENODEV;
3347 			break;
3348 		}
3349 		PF_RULES_WLOCK();
3350 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3351 		if (n < 0) {
3352 			PF_RULES_WUNLOCK();
3353 			error = EINVAL;
3354 			break;
3355 		}
3356 		io->pfrio_size = min(io->pfrio_size, n);
3357 
3358 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
3359 		pfrtstats = mallocarray(io->pfrio_size,
3360 		    sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT);
3361 		if (pfrtstats == NULL) {
3362 			error = ENOMEM;
3363 			PF_RULES_WUNLOCK();
3364 			break;
3365 		}
3366 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
3367 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3368 		PF_RULES_WUNLOCK();
3369 		if (error == 0)
3370 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
3371 		free(pfrtstats, M_TEMP);
3372 		break;
3373 	}
3374 
3375 	case DIOCRCLRTSTATS: {
3376 		struct pfioc_table *io = (struct pfioc_table *)addr;
3377 		struct pfr_table *pfrts;
3378 		size_t totlen;
3379 
3380 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3381 			error = ENODEV;
3382 			break;
3383 		}
3384 
3385 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
3386 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
3387 			/* We used to count tables and use the minimum required
3388 			 * size, so we didn't fail on overly large requests.
3389 			 * Keep doing so. */
3390 			io->pfrio_size = pf_ioctl_maxcount;
3391 			break;
3392 		}
3393 
3394 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3395 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3396 		    M_TEMP, M_NOWAIT);
3397 		if (pfrts == NULL) {
3398 			error = ENOMEM;
3399 			break;
3400 		}
3401 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3402 		if (error) {
3403 			free(pfrts, M_TEMP);
3404 			break;
3405 		}
3406 
3407 		PF_RULES_WLOCK();
3408 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
3409 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3410 		PF_RULES_WUNLOCK();
3411 		free(pfrts, M_TEMP);
3412 		break;
3413 	}
3414 
3415 	case DIOCRSETTFLAGS: {
3416 		struct pfioc_table *io = (struct pfioc_table *)addr;
3417 		struct pfr_table *pfrts;
3418 		size_t totlen;
3419 		int n;
3420 
3421 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3422 			error = ENODEV;
3423 			break;
3424 		}
3425 
3426 		PF_RULES_RLOCK();
3427 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3428 		if (n < 0) {
3429 			PF_RULES_RUNLOCK();
3430 			error = EINVAL;
3431 			break;
3432 		}
3433 
3434 		io->pfrio_size = min(io->pfrio_size, n);
3435 		PF_RULES_RUNLOCK();
3436 
3437 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3438 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3439 		    M_TEMP, M_WAITOK);
3440 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3441 		if (error) {
3442 			free(pfrts, M_TEMP);
3443 			break;
3444 		}
3445 		PF_RULES_WLOCK();
3446 		error = pfr_set_tflags(pfrts, io->pfrio_size,
3447 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
3448 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3449 		PF_RULES_WUNLOCK();
3450 		free(pfrts, M_TEMP);
3451 		break;
3452 	}
3453 
3454 	case DIOCRCLRADDRS: {
3455 		struct pfioc_table *io = (struct pfioc_table *)addr;
3456 
3457 		if (io->pfrio_esize != 0) {
3458 			error = ENODEV;
3459 			break;
3460 		}
3461 		PF_RULES_WLOCK();
3462 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
3463 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
3464 		PF_RULES_WUNLOCK();
3465 		break;
3466 	}
3467 
3468 	case DIOCRADDADDRS: {
3469 		struct pfioc_table *io = (struct pfioc_table *)addr;
3470 		struct pfr_addr *pfras;
3471 		size_t totlen;
3472 
3473 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3474 			error = ENODEV;
3475 			break;
3476 		}
3477 		if (io->pfrio_size < 0 ||
3478 		    io->pfrio_size > pf_ioctl_maxcount ||
3479 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3480 			error = EINVAL;
3481 			break;
3482 		}
3483 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3484 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3485 		    M_TEMP, M_NOWAIT);
3486 		if (! pfras) {
3487 			error = ENOMEM;
3488 			break;
3489 		}
3490 		error = copyin(io->pfrio_buffer, pfras, totlen);
3491 		if (error) {
3492 			free(pfras, M_TEMP);
3493 			break;
3494 		}
3495 		PF_RULES_WLOCK();
3496 		error = pfr_add_addrs(&io->pfrio_table, pfras,
3497 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
3498 		    PFR_FLAG_USERIOCTL);
3499 		PF_RULES_WUNLOCK();
3500 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3501 			error = copyout(pfras, io->pfrio_buffer, totlen);
3502 		free(pfras, M_TEMP);
3503 		break;
3504 	}
3505 
3506 	case DIOCRDELADDRS: {
3507 		struct pfioc_table *io = (struct pfioc_table *)addr;
3508 		struct pfr_addr *pfras;
3509 		size_t totlen;
3510 
3511 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3512 			error = ENODEV;
3513 			break;
3514 		}
3515 		if (io->pfrio_size < 0 ||
3516 		    io->pfrio_size > pf_ioctl_maxcount ||
3517 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3518 			error = EINVAL;
3519 			break;
3520 		}
3521 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3522 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3523 		    M_TEMP, M_NOWAIT);
3524 		if (! pfras) {
3525 			error = ENOMEM;
3526 			break;
3527 		}
3528 		error = copyin(io->pfrio_buffer, pfras, totlen);
3529 		if (error) {
3530 			free(pfras, M_TEMP);
3531 			break;
3532 		}
3533 		PF_RULES_WLOCK();
3534 		error = pfr_del_addrs(&io->pfrio_table, pfras,
3535 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
3536 		    PFR_FLAG_USERIOCTL);
3537 		PF_RULES_WUNLOCK();
3538 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3539 			error = copyout(pfras, io->pfrio_buffer, totlen);
3540 		free(pfras, M_TEMP);
3541 		break;
3542 	}
3543 
3544 	case DIOCRSETADDRS: {
3545 		struct pfioc_table *io = (struct pfioc_table *)addr;
3546 		struct pfr_addr *pfras;
3547 		size_t totlen, count;
3548 
3549 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3550 			error = ENODEV;
3551 			break;
3552 		}
3553 		if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
3554 			error = EINVAL;
3555 			break;
3556 		}
3557 		count = max(io->pfrio_size, io->pfrio_size2);
3558 		if (count > pf_ioctl_maxcount ||
3559 		    WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
3560 			error = EINVAL;
3561 			break;
3562 		}
3563 		totlen = count * sizeof(struct pfr_addr);
3564 		pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
3565 		    M_NOWAIT);
3566 		if (! pfras) {
3567 			error = ENOMEM;
3568 			break;
3569 		}
3570 		error = copyin(io->pfrio_buffer, pfras, totlen);
3571 		if (error) {
3572 			free(pfras, M_TEMP);
3573 			break;
3574 		}
3575 		PF_RULES_WLOCK();
3576 		error = pfr_set_addrs(&io->pfrio_table, pfras,
3577 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
3578 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
3579 		    PFR_FLAG_USERIOCTL, 0);
3580 		PF_RULES_WUNLOCK();
3581 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3582 			error = copyout(pfras, io->pfrio_buffer, totlen);
3583 		free(pfras, M_TEMP);
3584 		break;
3585 	}
3586 
3587 	case DIOCRGETADDRS: {
3588 		struct pfioc_table *io = (struct pfioc_table *)addr;
3589 		struct pfr_addr *pfras;
3590 		size_t totlen;
3591 
3592 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3593 			error = ENODEV;
3594 			break;
3595 		}
3596 		if (io->pfrio_size < 0 ||
3597 		    io->pfrio_size > pf_ioctl_maxcount ||
3598 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3599 			error = EINVAL;
3600 			break;
3601 		}
3602 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3603 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3604 		    M_TEMP, M_NOWAIT);
3605 		if (! pfras) {
3606 			error = ENOMEM;
3607 			break;
3608 		}
3609 		PF_RULES_RLOCK();
3610 		error = pfr_get_addrs(&io->pfrio_table, pfras,
3611 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3612 		PF_RULES_RUNLOCK();
3613 		if (error == 0)
3614 			error = copyout(pfras, io->pfrio_buffer, totlen);
3615 		free(pfras, M_TEMP);
3616 		break;
3617 	}
3618 
3619 	case DIOCRGETASTATS: {
3620 		struct pfioc_table *io = (struct pfioc_table *)addr;
3621 		struct pfr_astats *pfrastats;
3622 		size_t totlen;
3623 
3624 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
3625 			error = ENODEV;
3626 			break;
3627 		}
3628 		if (io->pfrio_size < 0 ||
3629 		    io->pfrio_size > pf_ioctl_maxcount ||
3630 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
3631 			error = EINVAL;
3632 			break;
3633 		}
3634 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
3635 		pfrastats = mallocarray(io->pfrio_size,
3636 		    sizeof(struct pfr_astats), M_TEMP, M_NOWAIT);
3637 		if (! pfrastats) {
3638 			error = ENOMEM;
3639 			break;
3640 		}
3641 		PF_RULES_RLOCK();
3642 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
3643 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3644 		PF_RULES_RUNLOCK();
3645 		if (error == 0)
3646 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
3647 		free(pfrastats, M_TEMP);
3648 		break;
3649 	}
3650 
3651 	case DIOCRCLRASTATS: {
3652 		struct pfioc_table *io = (struct pfioc_table *)addr;
3653 		struct pfr_addr *pfras;
3654 		size_t totlen;
3655 
3656 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3657 			error = ENODEV;
3658 			break;
3659 		}
3660 		if (io->pfrio_size < 0 ||
3661 		    io->pfrio_size > pf_ioctl_maxcount ||
3662 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3663 			error = EINVAL;
3664 			break;
3665 		}
3666 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3667 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3668 		    M_TEMP, M_NOWAIT);
3669 		if (! pfras) {
3670 			error = ENOMEM;
3671 			break;
3672 		}
3673 		error = copyin(io->pfrio_buffer, pfras, totlen);
3674 		if (error) {
3675 			free(pfras, M_TEMP);
3676 			break;
3677 		}
3678 		PF_RULES_WLOCK();
3679 		error = pfr_clr_astats(&io->pfrio_table, pfras,
3680 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
3681 		    PFR_FLAG_USERIOCTL);
3682 		PF_RULES_WUNLOCK();
3683 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3684 			error = copyout(pfras, io->pfrio_buffer, totlen);
3685 		free(pfras, M_TEMP);
3686 		break;
3687 	}
3688 
3689 	case DIOCRTSTADDRS: {
3690 		struct pfioc_table *io = (struct pfioc_table *)addr;
3691 		struct pfr_addr *pfras;
3692 		size_t totlen;
3693 
3694 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3695 			error = ENODEV;
3696 			break;
3697 		}
3698 		if (io->pfrio_size < 0 ||
3699 		    io->pfrio_size > pf_ioctl_maxcount ||
3700 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3701 			error = EINVAL;
3702 			break;
3703 		}
3704 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3705 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3706 		    M_TEMP, M_NOWAIT);
3707 		if (! pfras) {
3708 			error = ENOMEM;
3709 			break;
3710 		}
3711 		error = copyin(io->pfrio_buffer, pfras, totlen);
3712 		if (error) {
3713 			free(pfras, M_TEMP);
3714 			break;
3715 		}
3716 		PF_RULES_RLOCK();
3717 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
3718 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
3719 		    PFR_FLAG_USERIOCTL);
3720 		PF_RULES_RUNLOCK();
3721 		if (error == 0)
3722 			error = copyout(pfras, io->pfrio_buffer, totlen);
3723 		free(pfras, M_TEMP);
3724 		break;
3725 	}
3726 
3727 	case DIOCRINADEFINE: {
3728 		struct pfioc_table *io = (struct pfioc_table *)addr;
3729 		struct pfr_addr *pfras;
3730 		size_t totlen;
3731 
3732 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3733 			error = ENODEV;
3734 			break;
3735 		}
3736 		if (io->pfrio_size < 0 ||
3737 		    io->pfrio_size > pf_ioctl_maxcount ||
3738 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3739 			error = EINVAL;
3740 			break;
3741 		}
3742 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3743 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3744 		    M_TEMP, M_NOWAIT);
3745 		if (! pfras) {
3746 			error = ENOMEM;
3747 			break;
3748 		}
3749 		error = copyin(io->pfrio_buffer, pfras, totlen);
3750 		if (error) {
3751 			free(pfras, M_TEMP);
3752 			break;
3753 		}
3754 		PF_RULES_WLOCK();
3755 		error = pfr_ina_define(&io->pfrio_table, pfras,
3756 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
3757 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3758 		PF_RULES_WUNLOCK();
3759 		free(pfras, M_TEMP);
3760 		break;
3761 	}
3762 
3763 	case DIOCOSFPADD: {
3764 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
3765 		PF_RULES_WLOCK();
3766 		error = pf_osfp_add(io);
3767 		PF_RULES_WUNLOCK();
3768 		break;
3769 	}
3770 
3771 	case DIOCOSFPGET: {
3772 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
3773 		PF_RULES_RLOCK();
3774 		error = pf_osfp_get(io);
3775 		PF_RULES_RUNLOCK();
3776 		break;
3777 	}
3778 
3779 	case DIOCXBEGIN: {
3780 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3781 		struct pfioc_trans_e	*ioes, *ioe;
3782 		size_t			 totlen;
3783 		int			 i;
3784 
3785 		if (io->esize != sizeof(*ioe)) {
3786 			error = ENODEV;
3787 			break;
3788 		}
3789 		if (io->size < 0 ||
3790 		    io->size > pf_ioctl_maxcount ||
3791 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3792 			error = EINVAL;
3793 			break;
3794 		}
3795 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3796 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3797 		    M_TEMP, M_NOWAIT);
3798 		if (! ioes) {
3799 			error = ENOMEM;
3800 			break;
3801 		}
3802 		error = copyin(io->array, ioes, totlen);
3803 		if (error) {
3804 			free(ioes, M_TEMP);
3805 			break;
3806 		}
3807 		PF_RULES_WLOCK();
3808 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3809 			switch (ioe->rs_num) {
3810 #ifdef ALTQ
3811 			case PF_RULESET_ALTQ:
3812 				if (ioe->anchor[0]) {
3813 					PF_RULES_WUNLOCK();
3814 					free(ioes, M_TEMP);
3815 					error = EINVAL;
3816 					goto fail;
3817 				}
3818 				if ((error = pf_begin_altq(&ioe->ticket))) {
3819 					PF_RULES_WUNLOCK();
3820 					free(ioes, M_TEMP);
3821 					goto fail;
3822 				}
3823 				break;
3824 #endif /* ALTQ */
3825 			case PF_RULESET_TABLE:
3826 			    {
3827 				struct pfr_table table;
3828 
3829 				bzero(&table, sizeof(table));
3830 				strlcpy(table.pfrt_anchor, ioe->anchor,
3831 				    sizeof(table.pfrt_anchor));
3832 				if ((error = pfr_ina_begin(&table,
3833 				    &ioe->ticket, NULL, 0))) {
3834 					PF_RULES_WUNLOCK();
3835 					free(ioes, M_TEMP);
3836 					goto fail;
3837 				}
3838 				break;
3839 			    }
3840 			default:
3841 				if ((error = pf_begin_rules(&ioe->ticket,
3842 				    ioe->rs_num, ioe->anchor))) {
3843 					PF_RULES_WUNLOCK();
3844 					free(ioes, M_TEMP);
3845 					goto fail;
3846 				}
3847 				break;
3848 			}
3849 		}
3850 		PF_RULES_WUNLOCK();
3851 		error = copyout(ioes, io->array, totlen);
3852 		free(ioes, M_TEMP);
3853 		break;
3854 	}
3855 
3856 	case DIOCXROLLBACK: {
3857 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3858 		struct pfioc_trans_e	*ioe, *ioes;
3859 		size_t			 totlen;
3860 		int			 i;
3861 
3862 		if (io->esize != sizeof(*ioe)) {
3863 			error = ENODEV;
3864 			break;
3865 		}
3866 		if (io->size < 0 ||
3867 		    io->size > pf_ioctl_maxcount ||
3868 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3869 			error = EINVAL;
3870 			break;
3871 		}
3872 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3873 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3874 		    M_TEMP, M_NOWAIT);
3875 		if (! ioes) {
3876 			error = ENOMEM;
3877 			break;
3878 		}
3879 		error = copyin(io->array, ioes, totlen);
3880 		if (error) {
3881 			free(ioes, M_TEMP);
3882 			break;
3883 		}
3884 		PF_RULES_WLOCK();
3885 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3886 			switch (ioe->rs_num) {
3887 #ifdef ALTQ
3888 			case PF_RULESET_ALTQ:
3889 				if (ioe->anchor[0]) {
3890 					PF_RULES_WUNLOCK();
3891 					free(ioes, M_TEMP);
3892 					error = EINVAL;
3893 					goto fail;
3894 				}
3895 				if ((error = pf_rollback_altq(ioe->ticket))) {
3896 					PF_RULES_WUNLOCK();
3897 					free(ioes, M_TEMP);
3898 					goto fail; /* really bad */
3899 				}
3900 				break;
3901 #endif /* ALTQ */
3902 			case PF_RULESET_TABLE:
3903 			    {
3904 				struct pfr_table table;
3905 
3906 				bzero(&table, sizeof(table));
3907 				strlcpy(table.pfrt_anchor, ioe->anchor,
3908 				    sizeof(table.pfrt_anchor));
3909 				if ((error = pfr_ina_rollback(&table,
3910 				    ioe->ticket, NULL, 0))) {
3911 					PF_RULES_WUNLOCK();
3912 					free(ioes, M_TEMP);
3913 					goto fail; /* really bad */
3914 				}
3915 				break;
3916 			    }
3917 			default:
3918 				if ((error = pf_rollback_rules(ioe->ticket,
3919 				    ioe->rs_num, ioe->anchor))) {
3920 					PF_RULES_WUNLOCK();
3921 					free(ioes, M_TEMP);
3922 					goto fail; /* really bad */
3923 				}
3924 				break;
3925 			}
3926 		}
3927 		PF_RULES_WUNLOCK();
3928 		free(ioes, M_TEMP);
3929 		break;
3930 	}
3931 
3932 	case DIOCXCOMMIT: {
3933 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3934 		struct pfioc_trans_e	*ioe, *ioes;
3935 		struct pf_kruleset	*rs;
3936 		size_t			 totlen;
3937 		int			 i;
3938 
3939 		if (io->esize != sizeof(*ioe)) {
3940 			error = ENODEV;
3941 			break;
3942 		}
3943 
3944 		if (io->size < 0 ||
3945 		    io->size > pf_ioctl_maxcount ||
3946 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3947 			error = EINVAL;
3948 			break;
3949 		}
3950 
3951 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3952 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3953 		    M_TEMP, M_NOWAIT);
3954 		if (ioes == NULL) {
3955 			error = ENOMEM;
3956 			break;
3957 		}
3958 		error = copyin(io->array, ioes, totlen);
3959 		if (error) {
3960 			free(ioes, M_TEMP);
3961 			break;
3962 		}
3963 		PF_RULES_WLOCK();
3964 		/* First makes sure everything will succeed. */
3965 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3966 			switch (ioe->rs_num) {
3967 #ifdef ALTQ
3968 			case PF_RULESET_ALTQ:
3969 				if (ioe->anchor[0]) {
3970 					PF_RULES_WUNLOCK();
3971 					free(ioes, M_TEMP);
3972 					error = EINVAL;
3973 					goto fail;
3974 				}
3975 				if (!V_altqs_inactive_open || ioe->ticket !=
3976 				    V_ticket_altqs_inactive) {
3977 					PF_RULES_WUNLOCK();
3978 					free(ioes, M_TEMP);
3979 					error = EBUSY;
3980 					goto fail;
3981 				}
3982 				break;
3983 #endif /* ALTQ */
3984 			case PF_RULESET_TABLE:
3985 				rs = pf_find_kruleset(ioe->anchor);
3986 				if (rs == NULL || !rs->topen || ioe->ticket !=
3987 				    rs->tticket) {
3988 					PF_RULES_WUNLOCK();
3989 					free(ioes, M_TEMP);
3990 					error = EBUSY;
3991 					goto fail;
3992 				}
3993 				break;
3994 			default:
3995 				if (ioe->rs_num < 0 || ioe->rs_num >=
3996 				    PF_RULESET_MAX) {
3997 					PF_RULES_WUNLOCK();
3998 					free(ioes, M_TEMP);
3999 					error = EINVAL;
4000 					goto fail;
4001 				}
4002 				rs = pf_find_kruleset(ioe->anchor);
4003 				if (rs == NULL ||
4004 				    !rs->rules[ioe->rs_num].inactive.open ||
4005 				    rs->rules[ioe->rs_num].inactive.ticket !=
4006 				    ioe->ticket) {
4007 					PF_RULES_WUNLOCK();
4008 					free(ioes, M_TEMP);
4009 					error = EBUSY;
4010 					goto fail;
4011 				}
4012 				break;
4013 			}
4014 		}
4015 		/* Now do the commit - no errors should happen here. */
4016 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
4017 			switch (ioe->rs_num) {
4018 #ifdef ALTQ
4019 			case PF_RULESET_ALTQ:
4020 				if ((error = pf_commit_altq(ioe->ticket))) {
4021 					PF_RULES_WUNLOCK();
4022 					free(ioes, M_TEMP);
4023 					goto fail; /* really bad */
4024 				}
4025 				break;
4026 #endif /* ALTQ */
4027 			case PF_RULESET_TABLE:
4028 			    {
4029 				struct pfr_table table;
4030 
4031 				bzero(&table, sizeof(table));
4032 				strlcpy(table.pfrt_anchor, ioe->anchor,
4033 				    sizeof(table.pfrt_anchor));
4034 				if ((error = pfr_ina_commit(&table,
4035 				    ioe->ticket, NULL, NULL, 0))) {
4036 					PF_RULES_WUNLOCK();
4037 					free(ioes, M_TEMP);
4038 					goto fail; /* really bad */
4039 				}
4040 				break;
4041 			    }
4042 			default:
4043 				if ((error = pf_commit_rules(ioe->ticket,
4044 				    ioe->rs_num, ioe->anchor))) {
4045 					PF_RULES_WUNLOCK();
4046 					free(ioes, M_TEMP);
4047 					goto fail; /* really bad */
4048 				}
4049 				break;
4050 			}
4051 		}
4052 		PF_RULES_WUNLOCK();
4053 		free(ioes, M_TEMP);
4054 		break;
4055 	}
4056 
4057 	case DIOCGETSRCNODES: {
4058 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
4059 		struct pf_srchash	*sh;
4060 		struct pf_ksrc_node	*n;
4061 		struct pf_src_node	*p, *pstore;
4062 		uint32_t		 i, nr = 0;
4063 
4064 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4065 				i++, sh++) {
4066 			PF_HASHROW_LOCK(sh);
4067 			LIST_FOREACH(n, &sh->nodes, entry)
4068 				nr++;
4069 			PF_HASHROW_UNLOCK(sh);
4070 		}
4071 
4072 		psn->psn_len = min(psn->psn_len,
4073 		    sizeof(struct pf_src_node) * nr);
4074 
4075 		if (psn->psn_len == 0) {
4076 			psn->psn_len = sizeof(struct pf_src_node) * nr;
4077 			break;
4078 		}
4079 
4080 		nr = 0;
4081 
4082 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
4083 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4084 		    i++, sh++) {
4085 		    PF_HASHROW_LOCK(sh);
4086 		    LIST_FOREACH(n, &sh->nodes, entry) {
4087 
4088 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
4089 				break;
4090 
4091 			pf_src_node_copy(n, p);
4092 
4093 			p++;
4094 			nr++;
4095 		    }
4096 		    PF_HASHROW_UNLOCK(sh);
4097 		}
4098 		error = copyout(pstore, psn->psn_src_nodes,
4099 		    sizeof(struct pf_src_node) * nr);
4100 		if (error) {
4101 			free(pstore, M_TEMP);
4102 			break;
4103 		}
4104 		psn->psn_len = sizeof(struct pf_src_node) * nr;
4105 		free(pstore, M_TEMP);
4106 		break;
4107 	}
4108 
4109 	case DIOCCLRSRCNODES: {
4110 		pf_clear_srcnodes(NULL);
4111 		pf_purge_expired_src_nodes();
4112 		break;
4113 	}
4114 
4115 	case DIOCKILLSRCNODES:
4116 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
4117 		break;
4118 
4119 	case DIOCSETHOSTID: {
4120 		u_int32_t	*hostid = (u_int32_t *)addr;
4121 
4122 		PF_RULES_WLOCK();
4123 		if (*hostid == 0)
4124 			V_pf_status.hostid = arc4random();
4125 		else
4126 			V_pf_status.hostid = *hostid;
4127 		PF_RULES_WUNLOCK();
4128 		break;
4129 	}
4130 
4131 	case DIOCOSFPFLUSH:
4132 		PF_RULES_WLOCK();
4133 		pf_osfp_flush();
4134 		PF_RULES_WUNLOCK();
4135 		break;
4136 
4137 	case DIOCIGETIFACES: {
4138 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
4139 		struct pfi_kif *ifstore;
4140 		size_t bufsiz;
4141 
4142 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
4143 			error = ENODEV;
4144 			break;
4145 		}
4146 
4147 		if (io->pfiio_size < 0 ||
4148 		    io->pfiio_size > pf_ioctl_maxcount ||
4149 		    WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
4150 			error = EINVAL;
4151 			break;
4152 		}
4153 
4154 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
4155 		ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
4156 		    M_TEMP, M_NOWAIT);
4157 		if (ifstore == NULL) {
4158 			error = ENOMEM;
4159 			break;
4160 		}
4161 
4162 		PF_RULES_RLOCK();
4163 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
4164 		PF_RULES_RUNLOCK();
4165 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
4166 		free(ifstore, M_TEMP);
4167 		break;
4168 	}
4169 
4170 	case DIOCSETIFFLAG: {
4171 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
4172 
4173 		PF_RULES_WLOCK();
4174 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
4175 		PF_RULES_WUNLOCK();
4176 		break;
4177 	}
4178 
4179 	case DIOCCLRIFFLAG: {
4180 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
4181 
4182 		PF_RULES_WLOCK();
4183 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
4184 		PF_RULES_WUNLOCK();
4185 		break;
4186 	}
4187 
4188 	default:
4189 		error = ENODEV;
4190 		break;
4191 	}
4192 fail:
4193 	if (sx_xlocked(&pf_ioctl_lock))
4194 		sx_xunlock(&pf_ioctl_lock);
4195 	CURVNET_RESTORE();
4196 
4197 	return (error);
4198 }
4199 
4200 void
4201 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
4202 {
4203 	bzero(sp, sizeof(struct pfsync_state));
4204 
4205 	/* copy from state key */
4206 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
4207 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
4208 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
4209 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
4210 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
4211 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
4212 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
4213 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
4214 	sp->proto = st->key[PF_SK_WIRE]->proto;
4215 	sp->af = st->key[PF_SK_WIRE]->af;
4216 
4217 	/* copy from state */
4218 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
4219 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
4220 	sp->creation = htonl(time_uptime - st->creation);
4221 	sp->expire = pf_state_expires(st);
4222 	if (sp->expire <= time_uptime)
4223 		sp->expire = htonl(0);
4224 	else
4225 		sp->expire = htonl(sp->expire - time_uptime);
4226 
4227 	sp->direction = st->direction;
4228 	sp->log = st->log;
4229 	sp->timeout = st->timeout;
4230 	sp->state_flags = st->state_flags;
4231 	if (st->src_node)
4232 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
4233 	if (st->nat_src_node)
4234 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
4235 
4236 	sp->id = st->id;
4237 	sp->creatorid = st->creatorid;
4238 	pf_state_peer_hton(&st->src, &sp->src);
4239 	pf_state_peer_hton(&st->dst, &sp->dst);
4240 
4241 	if (st->rule.ptr == NULL)
4242 		sp->rule = htonl(-1);
4243 	else
4244 		sp->rule = htonl(st->rule.ptr->nr);
4245 	if (st->anchor.ptr == NULL)
4246 		sp->anchor = htonl(-1);
4247 	else
4248 		sp->anchor = htonl(st->anchor.ptr->nr);
4249 	if (st->nat_rule.ptr == NULL)
4250 		sp->nat_rule = htonl(-1);
4251 	else
4252 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
4253 
4254 	pf_state_counter_hton(counter_u64_fetch(st->packets[0]),
4255 	    sp->packets[0]);
4256 	pf_state_counter_hton(counter_u64_fetch(st->packets[1]),
4257 	    sp->packets[1]);
4258 	pf_state_counter_hton(counter_u64_fetch(st->bytes[0]), sp->bytes[0]);
4259 	pf_state_counter_hton(counter_u64_fetch(st->bytes[1]), sp->bytes[1]);
4260 
4261 }
4262 
4263 static void
4264 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
4265 {
4266 	struct pfr_ktable *kt;
4267 
4268 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
4269 
4270 	kt = aw->p.tbl;
4271 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
4272 		kt = kt->pfrkt_root;
4273 	aw->p.tbl = NULL;
4274 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
4275 		kt->pfrkt_cnt : -1;
4276 }
4277 
4278 /*
4279  * XXX - Check for version missmatch!!!
4280  */
4281 static void
4282 pf_clear_states(void)
4283 {
4284 	struct pf_state	*s;
4285 	u_int i;
4286 
4287 	for (i = 0; i <= pf_hashmask; i++) {
4288 		struct pf_idhash *ih = &V_pf_idhash[i];
4289 relock:
4290 		PF_HASHROW_LOCK(ih);
4291 		LIST_FOREACH(s, &ih->states, entry) {
4292 			s->timeout = PFTM_PURGE;
4293 			/* Don't send out individual delete messages. */
4294 			s->state_flags |= PFSTATE_NOSYNC;
4295 			pf_unlink_state(s, PF_ENTER_LOCKED);
4296 			goto relock;
4297 		}
4298 		PF_HASHROW_UNLOCK(ih);
4299 	}
4300 }
4301 
4302 static int
4303 pf_clear_tables(void)
4304 {
4305 	struct pfioc_table io;
4306 	int error;
4307 
4308 	bzero(&io, sizeof(io));
4309 
4310 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
4311 	    io.pfrio_flags);
4312 
4313 	return (error);
4314 }
4315 
4316 static void
4317 pf_clear_srcnodes(struct pf_ksrc_node *n)
4318 {
4319 	struct pf_state *s;
4320 	int i;
4321 
4322 	for (i = 0; i <= pf_hashmask; i++) {
4323 		struct pf_idhash *ih = &V_pf_idhash[i];
4324 
4325 		PF_HASHROW_LOCK(ih);
4326 		LIST_FOREACH(s, &ih->states, entry) {
4327 			if (n == NULL || n == s->src_node)
4328 				s->src_node = NULL;
4329 			if (n == NULL || n == s->nat_src_node)
4330 				s->nat_src_node = NULL;
4331 		}
4332 		PF_HASHROW_UNLOCK(ih);
4333 	}
4334 
4335 	if (n == NULL) {
4336 		struct pf_srchash *sh;
4337 
4338 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4339 		    i++, sh++) {
4340 			PF_HASHROW_LOCK(sh);
4341 			LIST_FOREACH(n, &sh->nodes, entry) {
4342 				n->expire = 1;
4343 				n->states = 0;
4344 			}
4345 			PF_HASHROW_UNLOCK(sh);
4346 		}
4347 	} else {
4348 		/* XXX: hash slot should already be locked here. */
4349 		n->expire = 1;
4350 		n->states = 0;
4351 	}
4352 }
4353 
4354 static void
4355 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
4356 {
4357 	struct pf_ksrc_node_list	 kill;
4358 
4359 	LIST_INIT(&kill);
4360 	for (int i = 0; i <= pf_srchashmask; i++) {
4361 		struct pf_srchash *sh = &V_pf_srchash[i];
4362 		struct pf_ksrc_node *sn, *tmp;
4363 
4364 		PF_HASHROW_LOCK(sh);
4365 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
4366 			if (PF_MATCHA(psnk->psnk_src.neg,
4367 			      &psnk->psnk_src.addr.v.a.addr,
4368 			      &psnk->psnk_src.addr.v.a.mask,
4369 			      &sn->addr, sn->af) &&
4370 			    PF_MATCHA(psnk->psnk_dst.neg,
4371 			      &psnk->psnk_dst.addr.v.a.addr,
4372 			      &psnk->psnk_dst.addr.v.a.mask,
4373 			      &sn->raddr, sn->af)) {
4374 				pf_unlink_src_node(sn);
4375 				LIST_INSERT_HEAD(&kill, sn, entry);
4376 				sn->expire = 1;
4377 			}
4378 		PF_HASHROW_UNLOCK(sh);
4379 	}
4380 
4381 	for (int i = 0; i <= pf_hashmask; i++) {
4382 		struct pf_idhash *ih = &V_pf_idhash[i];
4383 		struct pf_state *s;
4384 
4385 		PF_HASHROW_LOCK(ih);
4386 		LIST_FOREACH(s, &ih->states, entry) {
4387 			if (s->src_node && s->src_node->expire == 1)
4388 				s->src_node = NULL;
4389 			if (s->nat_src_node && s->nat_src_node->expire == 1)
4390 				s->nat_src_node = NULL;
4391 		}
4392 		PF_HASHROW_UNLOCK(ih);
4393 	}
4394 
4395 	psnk->psnk_killed = pf_free_src_nodes(&kill);
4396 }
4397 
4398 /*
4399  * XXX - Check for version missmatch!!!
4400  */
4401 
4402 /*
4403  * Duplicate pfctl -Fa operation to get rid of as much as we can.
4404  */
4405 static int
4406 shutdown_pf(void)
4407 {
4408 	int error = 0;
4409 	u_int32_t t[5];
4410 	char nn = '\0';
4411 
4412 	do {
4413 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
4414 		    != 0) {
4415 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
4416 			break;
4417 		}
4418 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
4419 		    != 0) {
4420 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
4421 			break;		/* XXX: rollback? */
4422 		}
4423 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
4424 		    != 0) {
4425 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
4426 			break;		/* XXX: rollback? */
4427 		}
4428 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
4429 		    != 0) {
4430 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
4431 			break;		/* XXX: rollback? */
4432 		}
4433 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
4434 		    != 0) {
4435 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
4436 			break;		/* XXX: rollback? */
4437 		}
4438 
4439 		/* XXX: these should always succeed here */
4440 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
4441 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
4442 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
4443 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
4444 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
4445 
4446 		if ((error = pf_clear_tables()) != 0)
4447 			break;
4448 
4449 #ifdef ALTQ
4450 		if ((error = pf_begin_altq(&t[0])) != 0) {
4451 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
4452 			break;
4453 		}
4454 		pf_commit_altq(t[0]);
4455 #endif
4456 
4457 		pf_clear_states();
4458 
4459 		pf_clear_srcnodes(NULL);
4460 
4461 		/* status does not use malloced mem so no need to cleanup */
4462 		/* fingerprints and interfaces have their own cleanup code */
4463 	} while(0);
4464 
4465 	return (error);
4466 }
4467 
4468 static pfil_return_t
4469 pf_check_return(int chk, struct mbuf **m)
4470 {
4471 
4472 	switch (chk) {
4473 	case PF_PASS:
4474 		if (*m == NULL)
4475 			return (PFIL_CONSUMED);
4476 		else
4477 			return (PFIL_PASS);
4478 		break;
4479 	default:
4480 		if (*m != NULL) {
4481 			m_freem(*m);
4482 			*m = NULL;
4483 		}
4484 		return (PFIL_DROPPED);
4485 	}
4486 }
4487 
4488 #ifdef INET
4489 static pfil_return_t
4490 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
4491     void *ruleset __unused, struct inpcb *inp)
4492 {
4493 	int chk;
4494 
4495 	chk = pf_test(PF_IN, flags, ifp, m, inp);
4496 
4497 	return (pf_check_return(chk, m));
4498 }
4499 
4500 static pfil_return_t
4501 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
4502     void *ruleset __unused,  struct inpcb *inp)
4503 {
4504 	int chk;
4505 
4506 	chk = pf_test(PF_OUT, flags, ifp, m, inp);
4507 
4508 	return (pf_check_return(chk, m));
4509 }
4510 #endif
4511 
4512 #ifdef INET6
4513 static pfil_return_t
4514 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
4515     void *ruleset __unused,  struct inpcb *inp)
4516 {
4517 	int chk;
4518 
4519 	/*
4520 	 * In case of loopback traffic IPv6 uses the real interface in
4521 	 * order to support scoped addresses. In order to support stateful
4522 	 * filtering we have change this to lo0 as it is the case in IPv4.
4523 	 */
4524 	CURVNET_SET(ifp->if_vnet);
4525 	chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
4526 	CURVNET_RESTORE();
4527 
4528 	return (pf_check_return(chk, m));
4529 }
4530 
4531 static pfil_return_t
4532 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
4533     void *ruleset __unused,  struct inpcb *inp)
4534 {
4535 	int chk;
4536 
4537 	CURVNET_SET(ifp->if_vnet);
4538 	chk = pf_test6(PF_OUT, flags, ifp, m, inp);
4539 	CURVNET_RESTORE();
4540 
4541 	return (pf_check_return(chk, m));
4542 }
4543 #endif /* INET6 */
4544 
4545 #ifdef INET
4546 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
4547 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
4548 #define	V_pf_ip4_in_hook	VNET(pf_ip4_in_hook)
4549 #define	V_pf_ip4_out_hook	VNET(pf_ip4_out_hook)
4550 #endif
4551 #ifdef INET6
4552 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
4553 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
4554 #define	V_pf_ip6_in_hook	VNET(pf_ip6_in_hook)
4555 #define	V_pf_ip6_out_hook	VNET(pf_ip6_out_hook)
4556 #endif
4557 
4558 static void
4559 hook_pf(void)
4560 {
4561 	struct pfil_hook_args pha;
4562 	struct pfil_link_args pla;
4563 	int ret;
4564 
4565 	if (V_pf_pfil_hooked)
4566 		return;
4567 
4568 	pha.pa_version = PFIL_VERSION;
4569 	pha.pa_modname = "pf";
4570 	pha.pa_ruleset = NULL;
4571 
4572 	pla.pa_version = PFIL_VERSION;
4573 
4574 #ifdef INET
4575 	pha.pa_type = PFIL_TYPE_IP4;
4576 	pha.pa_func = pf_check_in;
4577 	pha.pa_flags = PFIL_IN;
4578 	pha.pa_rulname = "default-in";
4579 	V_pf_ip4_in_hook = pfil_add_hook(&pha);
4580 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
4581 	pla.pa_head = V_inet_pfil_head;
4582 	pla.pa_hook = V_pf_ip4_in_hook;
4583 	ret = pfil_link(&pla);
4584 	MPASS(ret == 0);
4585 	pha.pa_func = pf_check_out;
4586 	pha.pa_flags = PFIL_OUT;
4587 	pha.pa_rulname = "default-out";
4588 	V_pf_ip4_out_hook = pfil_add_hook(&pha);
4589 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
4590 	pla.pa_head = V_inet_pfil_head;
4591 	pla.pa_hook = V_pf_ip4_out_hook;
4592 	ret = pfil_link(&pla);
4593 	MPASS(ret == 0);
4594 #endif
4595 #ifdef INET6
4596 	pha.pa_type = PFIL_TYPE_IP6;
4597 	pha.pa_func = pf_check6_in;
4598 	pha.pa_flags = PFIL_IN;
4599 	pha.pa_rulname = "default-in6";
4600 	V_pf_ip6_in_hook = pfil_add_hook(&pha);
4601 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
4602 	pla.pa_head = V_inet6_pfil_head;
4603 	pla.pa_hook = V_pf_ip6_in_hook;
4604 	ret = pfil_link(&pla);
4605 	MPASS(ret == 0);
4606 	pha.pa_func = pf_check6_out;
4607 	pha.pa_rulname = "default-out6";
4608 	pha.pa_flags = PFIL_OUT;
4609 	V_pf_ip6_out_hook = pfil_add_hook(&pha);
4610 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
4611 	pla.pa_head = V_inet6_pfil_head;
4612 	pla.pa_hook = V_pf_ip6_out_hook;
4613 	ret = pfil_link(&pla);
4614 	MPASS(ret == 0);
4615 #endif
4616 
4617 	V_pf_pfil_hooked = 1;
4618 }
4619 
4620 static void
4621 dehook_pf(void)
4622 {
4623 
4624 	if (V_pf_pfil_hooked == 0)
4625 		return;
4626 
4627 #ifdef INET
4628 	pfil_remove_hook(V_pf_ip4_in_hook);
4629 	pfil_remove_hook(V_pf_ip4_out_hook);
4630 #endif
4631 #ifdef INET6
4632 	pfil_remove_hook(V_pf_ip6_in_hook);
4633 	pfil_remove_hook(V_pf_ip6_out_hook);
4634 #endif
4635 
4636 	V_pf_pfil_hooked = 0;
4637 }
4638 
4639 static void
4640 pf_load_vnet(void)
4641 {
4642 	V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
4643 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
4644 
4645 	pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
4646 	    PF_RULE_TAG_HASH_SIZE_DEFAULT);
4647 #ifdef ALTQ
4648 	pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
4649 	    PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
4650 #endif
4651 
4652 	pfattach_vnet();
4653 	V_pf_vnet_active = 1;
4654 }
4655 
4656 static int
4657 pf_load(void)
4658 {
4659 	int error;
4660 
4661 	rm_init(&pf_rules_lock, "pf rulesets");
4662 	sx_init(&pf_ioctl_lock, "pf ioctl");
4663 	sx_init(&pf_end_lock, "pf end thread");
4664 
4665 	pf_mtag_initialize();
4666 
4667 	pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
4668 	if (pf_dev == NULL)
4669 		return (ENOMEM);
4670 
4671 	pf_end_threads = 0;
4672 	error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
4673 	if (error != 0)
4674 		return (error);
4675 
4676 	pfi_initialize();
4677 
4678 	return (0);
4679 }
4680 
4681 static void
4682 pf_unload_vnet(void)
4683 {
4684 
4685 	V_pf_vnet_active = 0;
4686 	V_pf_status.running = 0;
4687 	dehook_pf();
4688 
4689 	PF_RULES_WLOCK();
4690 	shutdown_pf();
4691 	PF_RULES_WUNLOCK();
4692 
4693 	swi_remove(V_pf_swi_cookie);
4694 
4695 	pf_unload_vnet_purge();
4696 
4697 	pf_normalize_cleanup();
4698 	PF_RULES_WLOCK();
4699 	pfi_cleanup_vnet();
4700 	PF_RULES_WUNLOCK();
4701 	pfr_cleanup();
4702 	pf_osfp_flush();
4703 	pf_cleanup();
4704 	if (IS_DEFAULT_VNET(curvnet))
4705 		pf_mtag_cleanup();
4706 
4707 	pf_cleanup_tagset(&V_pf_tags);
4708 #ifdef ALTQ
4709 	pf_cleanup_tagset(&V_pf_qids);
4710 #endif
4711 	uma_zdestroy(V_pf_tag_z);
4712 
4713 	/* Free counters last as we updated them during shutdown. */
4714 	counter_u64_free(V_pf_default_rule.evaluations);
4715 	for (int i = 0; i < 2; i++) {
4716 		counter_u64_free(V_pf_default_rule.packets[i]);
4717 		counter_u64_free(V_pf_default_rule.bytes[i]);
4718 	}
4719 	counter_u64_free(V_pf_default_rule.states_cur);
4720 	counter_u64_free(V_pf_default_rule.states_tot);
4721 	counter_u64_free(V_pf_default_rule.src_nodes);
4722 
4723 	for (int i = 0; i < PFRES_MAX; i++)
4724 		counter_u64_free(V_pf_status.counters[i]);
4725 	for (int i = 0; i < LCNT_MAX; i++)
4726 		counter_u64_free(V_pf_status.lcounters[i]);
4727 	for (int i = 0; i < FCNT_MAX; i++)
4728 		counter_u64_free(V_pf_status.fcounters[i]);
4729 	for (int i = 0; i < SCNT_MAX; i++)
4730 		counter_u64_free(V_pf_status.scounters[i]);
4731 }
4732 
4733 static void
4734 pf_unload(void)
4735 {
4736 
4737 	sx_xlock(&pf_end_lock);
4738 	pf_end_threads = 1;
4739 	while (pf_end_threads < 2) {
4740 		wakeup_one(pf_purge_thread);
4741 		sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
4742 	}
4743 	sx_xunlock(&pf_end_lock);
4744 
4745 	if (pf_dev != NULL)
4746 		destroy_dev(pf_dev);
4747 
4748 	pfi_cleanup();
4749 
4750 	rm_destroy(&pf_rules_lock);
4751 	sx_destroy(&pf_ioctl_lock);
4752 	sx_destroy(&pf_end_lock);
4753 }
4754 
4755 static void
4756 vnet_pf_init(void *unused __unused)
4757 {
4758 
4759 	pf_load_vnet();
4760 }
4761 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
4762     vnet_pf_init, NULL);
4763 
4764 static void
4765 vnet_pf_uninit(const void *unused __unused)
4766 {
4767 
4768 	pf_unload_vnet();
4769 }
4770 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
4771 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
4772     vnet_pf_uninit, NULL);
4773 
4774 static int
4775 pf_modevent(module_t mod, int type, void *data)
4776 {
4777 	int error = 0;
4778 
4779 	switch(type) {
4780 	case MOD_LOAD:
4781 		error = pf_load();
4782 		break;
4783 	case MOD_UNLOAD:
4784 		/* Handled in SYSUNINIT(pf_unload) to ensure it's done after
4785 		 * the vnet_pf_uninit()s */
4786 		break;
4787 	default:
4788 		error = EINVAL;
4789 		break;
4790 	}
4791 
4792 	return (error);
4793 }
4794 
4795 static moduledata_t pf_mod = {
4796 	"pf",
4797 	pf_modevent,
4798 	0
4799 };
4800 
4801 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
4802 MODULE_VERSION(pf, PF_MODVER);
4803