xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision 6be3386466ab79a84b48429ae66244f21526d3df)
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(&V_pf_swi_ie, "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 
470 	pf_krule_free(rule);
471 }
472 
473 static void
474 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
475     unsigned int default_size)
476 {
477 	unsigned int i;
478 	unsigned int hashsize;
479 
480 	if (*tunable_size == 0 || !powerof2(*tunable_size))
481 		*tunable_size = default_size;
482 
483 	hashsize = *tunable_size;
484 	ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
485 	    M_WAITOK);
486 	ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
487 	    M_WAITOK);
488 	ts->mask = hashsize - 1;
489 	ts->seed = arc4random();
490 	for (i = 0; i < hashsize; i++) {
491 		TAILQ_INIT(&ts->namehash[i]);
492 		TAILQ_INIT(&ts->taghash[i]);
493 	}
494 	BIT_FILL(TAGID_MAX, &ts->avail);
495 }
496 
497 static void
498 pf_cleanup_tagset(struct pf_tagset *ts)
499 {
500 	unsigned int i;
501 	unsigned int hashsize;
502 	struct pf_tagname *t, *tmp;
503 
504 	/*
505 	 * Only need to clean up one of the hashes as each tag is hashed
506 	 * into each table.
507 	 */
508 	hashsize = ts->mask + 1;
509 	for (i = 0; i < hashsize; i++)
510 		TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
511 			uma_zfree(V_pf_tag_z, t);
512 
513 	free(ts->namehash, M_PFHASH);
514 	free(ts->taghash, M_PFHASH);
515 }
516 
517 static uint16_t
518 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
519 {
520 	size_t len;
521 
522 	len = strnlen(tagname, PF_TAG_NAME_SIZE - 1);
523 	return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask);
524 }
525 
526 static uint16_t
527 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
528 {
529 
530 	return (tag & ts->mask);
531 }
532 
533 static u_int16_t
534 tagname2tag(struct pf_tagset *ts, char *tagname)
535 {
536 	struct pf_tagname	*tag;
537 	u_int32_t		 index;
538 	u_int16_t		 new_tagid;
539 
540 	PF_RULES_WASSERT();
541 
542 	index = tagname2hashindex(ts, tagname);
543 	TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
544 		if (strcmp(tagname, tag->name) == 0) {
545 			tag->ref++;
546 			return (tag->tag);
547 		}
548 
549 	/*
550 	 * new entry
551 	 *
552 	 * to avoid fragmentation, we do a linear search from the beginning
553 	 * and take the first free slot we find.
554 	 */
555 	new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
556 	/*
557 	 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
558 	 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
559 	 * set.  It may also return a bit number greater than TAGID_MAX due
560 	 * to rounding of the number of bits in the vector up to a multiple
561 	 * of the vector word size at declaration/allocation time.
562 	 */
563 	if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
564 		return (0);
565 
566 	/* Mark the tag as in use.  Bits are 0-based for BIT_CLR() */
567 	BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
568 
569 	/* allocate and fill new struct pf_tagname */
570 	tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
571 	if (tag == NULL)
572 		return (0);
573 	strlcpy(tag->name, tagname, sizeof(tag->name));
574 	tag->tag = new_tagid;
575 	tag->ref = 1;
576 
577 	/* Insert into namehash */
578 	TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
579 
580 	/* Insert into taghash */
581 	index = tag2hashindex(ts, new_tagid);
582 	TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
583 
584 	return (tag->tag);
585 }
586 
587 static void
588 tag_unref(struct pf_tagset *ts, u_int16_t tag)
589 {
590 	struct pf_tagname	*t;
591 	uint16_t		 index;
592 
593 	PF_RULES_WASSERT();
594 
595 	index = tag2hashindex(ts, tag);
596 	TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
597 		if (tag == t->tag) {
598 			if (--t->ref == 0) {
599 				TAILQ_REMOVE(&ts->taghash[index], t,
600 				    taghash_entries);
601 				index = tagname2hashindex(ts, t->name);
602 				TAILQ_REMOVE(&ts->namehash[index], t,
603 				    namehash_entries);
604 				/* Bits are 0-based for BIT_SET() */
605 				BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
606 				uma_zfree(V_pf_tag_z, t);
607 			}
608 			break;
609 		}
610 }
611 
612 static u_int16_t
613 pf_tagname2tag(char *tagname)
614 {
615 	return (tagname2tag(&V_pf_tags, tagname));
616 }
617 
618 #ifdef ALTQ
619 static u_int32_t
620 pf_qname2qid(char *qname)
621 {
622 	return ((u_int32_t)tagname2tag(&V_pf_qids, qname));
623 }
624 
625 static void
626 pf_qid_unref(u_int32_t qid)
627 {
628 	tag_unref(&V_pf_qids, (u_int16_t)qid);
629 }
630 
631 static int
632 pf_begin_altq(u_int32_t *ticket)
633 {
634 	struct pf_altq	*altq, *tmp;
635 	int		 error = 0;
636 
637 	PF_RULES_WASSERT();
638 
639 	/* Purge the old altq lists */
640 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
641 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
642 			/* detach and destroy the discipline */
643 			error = altq_remove(altq);
644 		}
645 		free(altq, M_PFALTQ);
646 	}
647 	TAILQ_INIT(V_pf_altq_ifs_inactive);
648 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
649 		pf_qid_unref(altq->qid);
650 		free(altq, M_PFALTQ);
651 	}
652 	TAILQ_INIT(V_pf_altqs_inactive);
653 	if (error)
654 		return (error);
655 	*ticket = ++V_ticket_altqs_inactive;
656 	V_altqs_inactive_open = 1;
657 	return (0);
658 }
659 
660 static int
661 pf_rollback_altq(u_int32_t ticket)
662 {
663 	struct pf_altq	*altq, *tmp;
664 	int		 error = 0;
665 
666 	PF_RULES_WASSERT();
667 
668 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
669 		return (0);
670 	/* Purge the old altq lists */
671 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
672 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
673 			/* detach and destroy the discipline */
674 			error = altq_remove(altq);
675 		}
676 		free(altq, M_PFALTQ);
677 	}
678 	TAILQ_INIT(V_pf_altq_ifs_inactive);
679 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
680 		pf_qid_unref(altq->qid);
681 		free(altq, M_PFALTQ);
682 	}
683 	TAILQ_INIT(V_pf_altqs_inactive);
684 	V_altqs_inactive_open = 0;
685 	return (error);
686 }
687 
688 static int
689 pf_commit_altq(u_int32_t ticket)
690 {
691 	struct pf_altqqueue	*old_altqs, *old_altq_ifs;
692 	struct pf_altq		*altq, *tmp;
693 	int			 err, error = 0;
694 
695 	PF_RULES_WASSERT();
696 
697 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
698 		return (EBUSY);
699 
700 	/* swap altqs, keep the old. */
701 	old_altqs = V_pf_altqs_active;
702 	old_altq_ifs = V_pf_altq_ifs_active;
703 	V_pf_altqs_active = V_pf_altqs_inactive;
704 	V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
705 	V_pf_altqs_inactive = old_altqs;
706 	V_pf_altq_ifs_inactive = old_altq_ifs;
707 	V_ticket_altqs_active = V_ticket_altqs_inactive;
708 
709 	/* Attach new disciplines */
710 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
711 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
712 			/* attach the discipline */
713 			error = altq_pfattach(altq);
714 			if (error == 0 && V_pf_altq_running)
715 				error = pf_enable_altq(altq);
716 			if (error != 0)
717 				return (error);
718 		}
719 	}
720 
721 	/* Purge the old altq lists */
722 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
723 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
724 			/* detach and destroy the discipline */
725 			if (V_pf_altq_running)
726 				error = pf_disable_altq(altq);
727 			err = altq_pfdetach(altq);
728 			if (err != 0 && error == 0)
729 				error = err;
730 			err = altq_remove(altq);
731 			if (err != 0 && error == 0)
732 				error = err;
733 		}
734 		free(altq, M_PFALTQ);
735 	}
736 	TAILQ_INIT(V_pf_altq_ifs_inactive);
737 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
738 		pf_qid_unref(altq->qid);
739 		free(altq, M_PFALTQ);
740 	}
741 	TAILQ_INIT(V_pf_altqs_inactive);
742 
743 	V_altqs_inactive_open = 0;
744 	return (error);
745 }
746 
747 static int
748 pf_enable_altq(struct pf_altq *altq)
749 {
750 	struct ifnet		*ifp;
751 	struct tb_profile	 tb;
752 	int			 error = 0;
753 
754 	if ((ifp = ifunit(altq->ifname)) == NULL)
755 		return (EINVAL);
756 
757 	if (ifp->if_snd.altq_type != ALTQT_NONE)
758 		error = altq_enable(&ifp->if_snd);
759 
760 	/* set tokenbucket regulator */
761 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
762 		tb.rate = altq->ifbandwidth;
763 		tb.depth = altq->tbrsize;
764 		error = tbr_set(&ifp->if_snd, &tb);
765 	}
766 
767 	return (error);
768 }
769 
770 static int
771 pf_disable_altq(struct pf_altq *altq)
772 {
773 	struct ifnet		*ifp;
774 	struct tb_profile	 tb;
775 	int			 error;
776 
777 	if ((ifp = ifunit(altq->ifname)) == NULL)
778 		return (EINVAL);
779 
780 	/*
781 	 * when the discipline is no longer referenced, it was overridden
782 	 * by a new one.  if so, just return.
783 	 */
784 	if (altq->altq_disc != ifp->if_snd.altq_disc)
785 		return (0);
786 
787 	error = altq_disable(&ifp->if_snd);
788 
789 	if (error == 0) {
790 		/* clear tokenbucket regulator */
791 		tb.rate = 0;
792 		error = tbr_set(&ifp->if_snd, &tb);
793 	}
794 
795 	return (error);
796 }
797 
798 static int
799 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
800     struct pf_altq *altq)
801 {
802 	struct ifnet	*ifp1;
803 	int		 error = 0;
804 
805 	/* Deactivate the interface in question */
806 	altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
807 	if ((ifp1 = ifunit(altq->ifname)) == NULL ||
808 	    (remove && ifp1 == ifp)) {
809 		altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
810 	} else {
811 		error = altq_add(ifp1, altq);
812 
813 		if (ticket != V_ticket_altqs_inactive)
814 			error = EBUSY;
815 
816 		if (error)
817 			free(altq, M_PFALTQ);
818 	}
819 
820 	return (error);
821 }
822 
823 void
824 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
825 {
826 	struct pf_altq	*a1, *a2, *a3;
827 	u_int32_t	 ticket;
828 	int		 error = 0;
829 
830 	/*
831 	 * No need to re-evaluate the configuration for events on interfaces
832 	 * that do not support ALTQ, as it's not possible for such
833 	 * interfaces to be part of the configuration.
834 	 */
835 	if (!ALTQ_IS_READY(&ifp->if_snd))
836 		return;
837 
838 	/* Interrupt userland queue modifications */
839 	if (V_altqs_inactive_open)
840 		pf_rollback_altq(V_ticket_altqs_inactive);
841 
842 	/* Start new altq ruleset */
843 	if (pf_begin_altq(&ticket))
844 		return;
845 
846 	/* Copy the current active set */
847 	TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
848 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
849 		if (a2 == NULL) {
850 			error = ENOMEM;
851 			break;
852 		}
853 		bcopy(a1, a2, sizeof(struct pf_altq));
854 
855 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
856 		if (error)
857 			break;
858 
859 		TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
860 	}
861 	if (error)
862 		goto out;
863 	TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
864 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
865 		if (a2 == NULL) {
866 			error = ENOMEM;
867 			break;
868 		}
869 		bcopy(a1, a2, sizeof(struct pf_altq));
870 
871 		if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
872 			error = EBUSY;
873 			free(a2, M_PFALTQ);
874 			break;
875 		}
876 		a2->altq_disc = NULL;
877 		TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
878 			if (strncmp(a3->ifname, a2->ifname,
879 				IFNAMSIZ) == 0) {
880 				a2->altq_disc = a3->altq_disc;
881 				break;
882 			}
883 		}
884 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
885 		if (error)
886 			break;
887 
888 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
889 	}
890 
891 out:
892 	if (error != 0)
893 		pf_rollback_altq(ticket);
894 	else
895 		pf_commit_altq(ticket);
896 }
897 #endif /* ALTQ */
898 
899 static int
900 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
901 {
902 	struct pf_kruleset	*rs;
903 	struct pf_krule		*rule;
904 
905 	PF_RULES_WASSERT();
906 
907 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
908 		return (EINVAL);
909 	rs = pf_find_or_create_kruleset(anchor);
910 	if (rs == NULL)
911 		return (EINVAL);
912 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
913 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
914 		rs->rules[rs_num].inactive.rcount--;
915 	}
916 	*ticket = ++rs->rules[rs_num].inactive.ticket;
917 	rs->rules[rs_num].inactive.open = 1;
918 	return (0);
919 }
920 
921 static int
922 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
923 {
924 	struct pf_kruleset	*rs;
925 	struct pf_krule		*rule;
926 
927 	PF_RULES_WASSERT();
928 
929 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
930 		return (EINVAL);
931 	rs = pf_find_kruleset(anchor);
932 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
933 	    rs->rules[rs_num].inactive.ticket != ticket)
934 		return (0);
935 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
936 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
937 		rs->rules[rs_num].inactive.rcount--;
938 	}
939 	rs->rules[rs_num].inactive.open = 0;
940 	return (0);
941 }
942 
943 #define PF_MD5_UPD(st, elm)						\
944 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
945 
946 #define PF_MD5_UPD_STR(st, elm)						\
947 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
948 
949 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
950 		(stor) = htonl((st)->elm);				\
951 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
952 } while (0)
953 
954 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
955 		(stor) = htons((st)->elm);				\
956 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
957 } while (0)
958 
959 static void
960 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
961 {
962 	PF_MD5_UPD(pfr, addr.type);
963 	switch (pfr->addr.type) {
964 		case PF_ADDR_DYNIFTL:
965 			PF_MD5_UPD(pfr, addr.v.ifname);
966 			PF_MD5_UPD(pfr, addr.iflags);
967 			break;
968 		case PF_ADDR_TABLE:
969 			PF_MD5_UPD(pfr, addr.v.tblname);
970 			break;
971 		case PF_ADDR_ADDRMASK:
972 			/* XXX ignore af? */
973 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
974 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
975 			break;
976 	}
977 
978 	PF_MD5_UPD(pfr, port[0]);
979 	PF_MD5_UPD(pfr, port[1]);
980 	PF_MD5_UPD(pfr, neg);
981 	PF_MD5_UPD(pfr, port_op);
982 }
983 
984 static void
985 pf_hash_rule(MD5_CTX *ctx, struct pf_krule *rule)
986 {
987 	u_int16_t x;
988 	u_int32_t y;
989 
990 	pf_hash_rule_addr(ctx, &rule->src);
991 	pf_hash_rule_addr(ctx, &rule->dst);
992 	PF_MD5_UPD_STR(rule, label);
993 	PF_MD5_UPD_STR(rule, ifname);
994 	PF_MD5_UPD_STR(rule, match_tagname);
995 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
996 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
997 	PF_MD5_UPD_HTONL(rule, prob, y);
998 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
999 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1000 	PF_MD5_UPD(rule, uid.op);
1001 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1002 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1003 	PF_MD5_UPD(rule, gid.op);
1004 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
1005 	PF_MD5_UPD(rule, action);
1006 	PF_MD5_UPD(rule, direction);
1007 	PF_MD5_UPD(rule, af);
1008 	PF_MD5_UPD(rule, quick);
1009 	PF_MD5_UPD(rule, ifnot);
1010 	PF_MD5_UPD(rule, match_tag_not);
1011 	PF_MD5_UPD(rule, natpass);
1012 	PF_MD5_UPD(rule, keep_state);
1013 	PF_MD5_UPD(rule, proto);
1014 	PF_MD5_UPD(rule, type);
1015 	PF_MD5_UPD(rule, code);
1016 	PF_MD5_UPD(rule, flags);
1017 	PF_MD5_UPD(rule, flagset);
1018 	PF_MD5_UPD(rule, allow_opts);
1019 	PF_MD5_UPD(rule, rt);
1020 	PF_MD5_UPD(rule, tos);
1021 }
1022 
1023 static int
1024 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1025 {
1026 	struct pf_kruleset	*rs;
1027 	struct pf_krule		*rule, **old_array;
1028 	struct pf_krulequeue	*old_rules;
1029 	int			 error;
1030 	u_int32_t		 old_rcount;
1031 
1032 	PF_RULES_WASSERT();
1033 
1034 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1035 		return (EINVAL);
1036 	rs = pf_find_kruleset(anchor);
1037 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1038 	    ticket != rs->rules[rs_num].inactive.ticket)
1039 		return (EBUSY);
1040 
1041 	/* Calculate checksum for the main ruleset */
1042 	if (rs == &pf_main_ruleset) {
1043 		error = pf_setup_pfsync_matching(rs);
1044 		if (error != 0)
1045 			return (error);
1046 	}
1047 
1048 	/* Swap rules, keep the old. */
1049 	old_rules = rs->rules[rs_num].active.ptr;
1050 	old_rcount = rs->rules[rs_num].active.rcount;
1051 	old_array = rs->rules[rs_num].active.ptr_array;
1052 
1053 	rs->rules[rs_num].active.ptr =
1054 	    rs->rules[rs_num].inactive.ptr;
1055 	rs->rules[rs_num].active.ptr_array =
1056 	    rs->rules[rs_num].inactive.ptr_array;
1057 	rs->rules[rs_num].active.rcount =
1058 	    rs->rules[rs_num].inactive.rcount;
1059 	rs->rules[rs_num].inactive.ptr = old_rules;
1060 	rs->rules[rs_num].inactive.ptr_array = old_array;
1061 	rs->rules[rs_num].inactive.rcount = old_rcount;
1062 
1063 	rs->rules[rs_num].active.ticket =
1064 	    rs->rules[rs_num].inactive.ticket;
1065 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1066 
1067 	/* Purge the old rule list. */
1068 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1069 		pf_unlink_rule(old_rules, rule);
1070 	if (rs->rules[rs_num].inactive.ptr_array)
1071 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1072 	rs->rules[rs_num].inactive.ptr_array = NULL;
1073 	rs->rules[rs_num].inactive.rcount = 0;
1074 	rs->rules[rs_num].inactive.open = 0;
1075 	pf_remove_if_empty_kruleset(rs);
1076 
1077 	return (0);
1078 }
1079 
1080 static int
1081 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1082 {
1083 	MD5_CTX			 ctx;
1084 	struct pf_krule		*rule;
1085 	int			 rs_cnt;
1086 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
1087 
1088 	MD5Init(&ctx);
1089 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1090 		/* XXX PF_RULESET_SCRUB as well? */
1091 		if (rs_cnt == PF_RULESET_SCRUB)
1092 			continue;
1093 
1094 		if (rs->rules[rs_cnt].inactive.ptr_array)
1095 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1096 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
1097 
1098 		if (rs->rules[rs_cnt].inactive.rcount) {
1099 			rs->rules[rs_cnt].inactive.ptr_array =
1100 			    malloc(sizeof(caddr_t) *
1101 			    rs->rules[rs_cnt].inactive.rcount,
1102 			    M_TEMP, M_NOWAIT);
1103 
1104 			if (!rs->rules[rs_cnt].inactive.ptr_array)
1105 				return (ENOMEM);
1106 		}
1107 
1108 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1109 		    entries) {
1110 			pf_hash_rule(&ctx, rule);
1111 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1112 		}
1113 	}
1114 
1115 	MD5Final(digest, &ctx);
1116 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1117 	return (0);
1118 }
1119 
1120 static int
1121 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1122     sa_family_t af)
1123 {
1124 	int error = 0;
1125 
1126 	switch (addr->type) {
1127 	case PF_ADDR_TABLE:
1128 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1129 		if (addr->p.tbl == NULL)
1130 			error = ENOMEM;
1131 		break;
1132 	case PF_ADDR_DYNIFTL:
1133 		error = pfi_dynaddr_setup(addr, af);
1134 		break;
1135 	}
1136 
1137 	return (error);
1138 }
1139 
1140 static void
1141 pf_addr_copyout(struct pf_addr_wrap *addr)
1142 {
1143 
1144 	switch (addr->type) {
1145 	case PF_ADDR_DYNIFTL:
1146 		pfi_dynaddr_copyout(addr);
1147 		break;
1148 	case PF_ADDR_TABLE:
1149 		pf_tbladdr_copyout(addr);
1150 		break;
1151 	}
1152 }
1153 
1154 static void
1155 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
1156 {
1157 	int	secs = time_uptime, diff;
1158 
1159 	bzero(out, sizeof(struct pf_src_node));
1160 
1161 	bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
1162 	bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
1163 
1164 	if (in->rule.ptr != NULL)
1165 		out->rule.nr = in->rule.ptr->nr;
1166 
1167 	for (int i = 0; i < 2; i++) {
1168 		out->bytes[i] = counter_u64_fetch(in->bytes[i]);
1169 		out->packets[i] = counter_u64_fetch(in->packets[i]);
1170 	}
1171 
1172 	out->states = in->states;
1173 	out->conn = in->conn;
1174 	out->af = in->af;
1175 	out->ruletype = in->ruletype;
1176 
1177 	out->creation = secs - in->creation;
1178 	if (out->expire > secs)
1179 		out->expire -= secs;
1180 	else
1181 		out->expire = 0;
1182 
1183 	/* Adjust the connection rate estimate. */
1184 	diff = secs - in->conn_rate.last;
1185 	if (diff >= in->conn_rate.seconds)
1186 		out->conn_rate.count = 0;
1187 	else
1188 		out->conn_rate.count -=
1189 		    in->conn_rate.count * diff /
1190 		    in->conn_rate.seconds;
1191 }
1192 
1193 #ifdef ALTQ
1194 /*
1195  * Handle export of struct pf_kaltq to user binaries that may be using any
1196  * version of struct pf_altq.
1197  */
1198 static int
1199 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1200 {
1201 	u_int32_t version;
1202 
1203 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1204 		version = 0;
1205 	else
1206 		version = pa->version;
1207 
1208 	if (version > PFIOC_ALTQ_VERSION)
1209 		return (EINVAL);
1210 
1211 #define ASSIGN(x) exported_q->x = q->x
1212 #define COPY(x) \
1213 	bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1214 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1215 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1216 
1217 	switch (version) {
1218 	case 0: {
1219 		struct pf_altq_v0 *exported_q =
1220 		    &((struct pfioc_altq_v0 *)pa)->altq;
1221 
1222 		COPY(ifname);
1223 
1224 		ASSIGN(scheduler);
1225 		ASSIGN(tbrsize);
1226 		exported_q->tbrsize = SATU16(q->tbrsize);
1227 		exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1228 
1229 		COPY(qname);
1230 		COPY(parent);
1231 		ASSIGN(parent_qid);
1232 		exported_q->bandwidth = SATU32(q->bandwidth);
1233 		ASSIGN(priority);
1234 		ASSIGN(local_flags);
1235 
1236 		ASSIGN(qlimit);
1237 		ASSIGN(flags);
1238 
1239 		if (q->scheduler == ALTQT_HFSC) {
1240 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1241 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1242 			    SATU32(q->pq_u.hfsc_opts.x)
1243 
1244 			ASSIGN_OPT_SATU32(rtsc_m1);
1245 			ASSIGN_OPT(rtsc_d);
1246 			ASSIGN_OPT_SATU32(rtsc_m2);
1247 
1248 			ASSIGN_OPT_SATU32(lssc_m1);
1249 			ASSIGN_OPT(lssc_d);
1250 			ASSIGN_OPT_SATU32(lssc_m2);
1251 
1252 			ASSIGN_OPT_SATU32(ulsc_m1);
1253 			ASSIGN_OPT(ulsc_d);
1254 			ASSIGN_OPT_SATU32(ulsc_m2);
1255 
1256 			ASSIGN_OPT(flags);
1257 
1258 #undef ASSIGN_OPT
1259 #undef ASSIGN_OPT_SATU32
1260 		} else
1261 			COPY(pq_u);
1262 
1263 		ASSIGN(qid);
1264 		break;
1265 	}
1266 	case 1:	{
1267 		struct pf_altq_v1 *exported_q =
1268 		    &((struct pfioc_altq_v1 *)pa)->altq;
1269 
1270 		COPY(ifname);
1271 
1272 		ASSIGN(scheduler);
1273 		ASSIGN(tbrsize);
1274 		ASSIGN(ifbandwidth);
1275 
1276 		COPY(qname);
1277 		COPY(parent);
1278 		ASSIGN(parent_qid);
1279 		ASSIGN(bandwidth);
1280 		ASSIGN(priority);
1281 		ASSIGN(local_flags);
1282 
1283 		ASSIGN(qlimit);
1284 		ASSIGN(flags);
1285 		COPY(pq_u);
1286 
1287 		ASSIGN(qid);
1288 		break;
1289 	}
1290 	default:
1291 		panic("%s: unhandled struct pfioc_altq version", __func__);
1292 		break;
1293 	}
1294 
1295 #undef ASSIGN
1296 #undef COPY
1297 #undef SATU16
1298 #undef SATU32
1299 
1300 	return (0);
1301 }
1302 
1303 /*
1304  * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1305  * that may be using any version of it.
1306  */
1307 static int
1308 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1309 {
1310 	u_int32_t version;
1311 
1312 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1313 		version = 0;
1314 	else
1315 		version = pa->version;
1316 
1317 	if (version > PFIOC_ALTQ_VERSION)
1318 		return (EINVAL);
1319 
1320 #define ASSIGN(x) q->x = imported_q->x
1321 #define COPY(x) \
1322 	bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1323 
1324 	switch (version) {
1325 	case 0: {
1326 		struct pf_altq_v0 *imported_q =
1327 		    &((struct pfioc_altq_v0 *)pa)->altq;
1328 
1329 		COPY(ifname);
1330 
1331 		ASSIGN(scheduler);
1332 		ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1333 		ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1334 
1335 		COPY(qname);
1336 		COPY(parent);
1337 		ASSIGN(parent_qid);
1338 		ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1339 		ASSIGN(priority);
1340 		ASSIGN(local_flags);
1341 
1342 		ASSIGN(qlimit);
1343 		ASSIGN(flags);
1344 
1345 		if (imported_q->scheduler == ALTQT_HFSC) {
1346 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1347 
1348 			/*
1349 			 * The m1 and m2 parameters are being copied from
1350 			 * 32-bit to 64-bit.
1351 			 */
1352 			ASSIGN_OPT(rtsc_m1);
1353 			ASSIGN_OPT(rtsc_d);
1354 			ASSIGN_OPT(rtsc_m2);
1355 
1356 			ASSIGN_OPT(lssc_m1);
1357 			ASSIGN_OPT(lssc_d);
1358 			ASSIGN_OPT(lssc_m2);
1359 
1360 			ASSIGN_OPT(ulsc_m1);
1361 			ASSIGN_OPT(ulsc_d);
1362 			ASSIGN_OPT(ulsc_m2);
1363 
1364 			ASSIGN_OPT(flags);
1365 
1366 #undef ASSIGN_OPT
1367 		} else
1368 			COPY(pq_u);
1369 
1370 		ASSIGN(qid);
1371 		break;
1372 	}
1373 	case 1: {
1374 		struct pf_altq_v1 *imported_q =
1375 		    &((struct pfioc_altq_v1 *)pa)->altq;
1376 
1377 		COPY(ifname);
1378 
1379 		ASSIGN(scheduler);
1380 		ASSIGN(tbrsize);
1381 		ASSIGN(ifbandwidth);
1382 
1383 		COPY(qname);
1384 		COPY(parent);
1385 		ASSIGN(parent_qid);
1386 		ASSIGN(bandwidth);
1387 		ASSIGN(priority);
1388 		ASSIGN(local_flags);
1389 
1390 		ASSIGN(qlimit);
1391 		ASSIGN(flags);
1392 		COPY(pq_u);
1393 
1394 		ASSIGN(qid);
1395 		break;
1396 	}
1397 	default:
1398 		panic("%s: unhandled struct pfioc_altq version", __func__);
1399 		break;
1400 	}
1401 
1402 #undef ASSIGN
1403 #undef COPY
1404 
1405 	return (0);
1406 }
1407 
1408 static struct pf_altq *
1409 pf_altq_get_nth_active(u_int32_t n)
1410 {
1411 	struct pf_altq		*altq;
1412 	u_int32_t		 nr;
1413 
1414 	nr = 0;
1415 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1416 		if (nr == n)
1417 			return (altq);
1418 		nr++;
1419 	}
1420 
1421 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1422 		if (nr == n)
1423 			return (altq);
1424 		nr++;
1425 	}
1426 
1427 	return (NULL);
1428 }
1429 #endif /* ALTQ */
1430 
1431 void
1432 pf_krule_free(struct pf_krule *rule)
1433 {
1434 	if (rule == NULL)
1435 		return;
1436 
1437 	counter_u64_free(rule->evaluations);
1438 	for (int i = 0; i < 2; i++) {
1439 		counter_u64_free(rule->packets[i]);
1440 		counter_u64_free(rule->bytes[i]);
1441 	}
1442 	counter_u64_free(rule->states_cur);
1443 	counter_u64_free(rule->states_tot);
1444 	counter_u64_free(rule->src_nodes);
1445 	free(rule, M_PFRULE);
1446 }
1447 
1448 static void
1449 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool,
1450     struct pf_pooladdr *pool)
1451 {
1452 
1453 	bzero(pool, sizeof(*pool));
1454 	bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr));
1455 	strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname));
1456 }
1457 
1458 static void
1459 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool,
1460     struct pf_kpooladdr *kpool)
1461 {
1462 
1463 	bzero(kpool, sizeof(*kpool));
1464 	bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr));
1465 	strlcpy(kpool->ifname, pool->ifname, sizeof(kpool->ifname));
1466 }
1467 
1468 static void
1469 pf_kpool_to_pool(const struct pf_kpool *kpool, struct pf_pool *pool)
1470 {
1471 	bzero(pool, sizeof(*pool));
1472 
1473 	bcopy(&kpool->key, &pool->key, sizeof(pool->key));
1474 	bcopy(&kpool->counter, &pool->counter, sizeof(pool->counter));
1475 
1476 	pool->tblidx = kpool->tblidx;
1477 	pool->proxy_port[0] = kpool->proxy_port[0];
1478 	pool->proxy_port[1] = kpool->proxy_port[1];
1479 	pool->opts = kpool->opts;
1480 }
1481 
1482 static int
1483 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool)
1484 {
1485 	_Static_assert(sizeof(pool->key) == sizeof(kpool->key), "");
1486 	_Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), "");
1487 
1488 	bzero(kpool, sizeof(*kpool));
1489 
1490 	bcopy(&pool->key, &kpool->key, sizeof(kpool->key));
1491 	bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter));
1492 
1493 	kpool->tblidx = pool->tblidx;
1494 	kpool->proxy_port[0] = pool->proxy_port[0];
1495 	kpool->proxy_port[1] = pool->proxy_port[1];
1496 	kpool->opts = pool->opts;
1497 
1498 	return (0);
1499 }
1500 
1501 static void
1502 pf_krule_to_rule(const struct pf_krule *krule, struct pf_rule *rule)
1503 {
1504 
1505 	bzero(rule, sizeof(*rule));
1506 
1507 	bcopy(&krule->src, &rule->src, sizeof(rule->src));
1508 	bcopy(&krule->dst, &rule->dst, sizeof(rule->dst));
1509 
1510 	for (int i = 0; i < PF_SKIP_COUNT; ++i) {
1511 		if (rule->skip[i].ptr == NULL)
1512 			rule->skip[i].nr = -1;
1513 		else
1514 			rule->skip[i].nr = krule->skip[i].ptr->nr;
1515 	}
1516 
1517 	strlcpy(rule->label, krule->label, sizeof(rule->label));
1518 	strlcpy(rule->ifname, krule->ifname, sizeof(rule->ifname));
1519 	strlcpy(rule->qname, krule->qname, sizeof(rule->qname));
1520 	strlcpy(rule->pqname, krule->pqname, sizeof(rule->pqname));
1521 	strlcpy(rule->tagname, krule->tagname, sizeof(rule->tagname));
1522 	strlcpy(rule->match_tagname, krule->match_tagname,
1523 	    sizeof(rule->match_tagname));
1524 	strlcpy(rule->overload_tblname, krule->overload_tblname,
1525 	    sizeof(rule->overload_tblname));
1526 
1527 	pf_kpool_to_pool(&krule->rpool, &rule->rpool);
1528 
1529 	rule->evaluations = counter_u64_fetch(krule->evaluations);
1530 	for (int i = 0; i < 2; i++) {
1531 		rule->packets[i] = counter_u64_fetch(krule->packets[i]);
1532 		rule->bytes[i] = counter_u64_fetch(krule->bytes[i]);
1533 	}
1534 
1535 	/* kif, anchor, overload_tbl are not copied over. */
1536 
1537 	rule->os_fingerprint = krule->os_fingerprint;
1538 
1539 	rule->rtableid = krule->rtableid;
1540 	bcopy(krule->timeout, rule->timeout, sizeof(krule->timeout));
1541 	rule->max_states = krule->max_states;
1542 	rule->max_src_nodes = krule->max_src_nodes;
1543 	rule->max_src_states = krule->max_src_states;
1544 	rule->max_src_conn = krule->max_src_conn;
1545 	rule->max_src_conn_rate.limit = krule->max_src_conn_rate.limit;
1546 	rule->max_src_conn_rate.seconds = krule->max_src_conn_rate.seconds;
1547 	rule->qid = krule->qid;
1548 	rule->pqid = krule->pqid;
1549 	rule->rt_listid = krule->rt_listid;
1550 	rule->nr = krule->nr;
1551 	rule->prob = krule->prob;
1552 	rule->cuid = krule->cuid;
1553 	rule->cpid = krule->cpid;
1554 
1555 	rule->return_icmp = krule->return_icmp;
1556 	rule->return_icmp6 = krule->return_icmp6;
1557 	rule->max_mss = krule->max_mss;
1558 	rule->tag = krule->tag;
1559 	rule->match_tag = krule->match_tag;
1560 	rule->scrub_flags = krule->scrub_flags;
1561 
1562 	bcopy(&krule->uid, &rule->uid, sizeof(krule->uid));
1563 	bcopy(&krule->gid, &rule->gid, sizeof(krule->gid));
1564 
1565 	rule->rule_flag = krule->rule_flag;
1566 	rule->action = krule->action;
1567 	rule->direction = krule->direction;
1568 	rule->log = krule->log;
1569 	rule->logif = krule->logif;
1570 	rule->quick = krule->quick;
1571 	rule->ifnot = krule->ifnot;
1572 	rule->match_tag_not = krule->match_tag_not;
1573 	rule->natpass = krule->natpass;
1574 
1575 	rule->keep_state = krule->keep_state;
1576 	rule->af = krule->af;
1577 	rule->proto = krule->proto;
1578 	rule->type = krule->type;
1579 	rule->code = krule->code;
1580 	rule->flags = krule->flags;
1581 	rule->flagset = krule->flagset;
1582 	rule->min_ttl = krule->min_ttl;
1583 	rule->allow_opts = krule->allow_opts;
1584 	rule->rt = krule->rt;
1585 	rule->return_ttl = krule->return_ttl;
1586 	rule->tos = krule->tos;
1587 	rule->set_tos = krule->set_tos;
1588 	rule->anchor_relative = krule->anchor_relative;
1589 	rule->anchor_wildcard = krule->anchor_wildcard;
1590 
1591 	rule->flush = krule->flush;
1592 	rule->prio = krule->prio;
1593 	rule->set_prio[0] = krule->set_prio[0];
1594 	rule->set_prio[1] = krule->set_prio[1];
1595 
1596 	bcopy(&krule->divert, &rule->divert, sizeof(krule->divert));
1597 
1598 	rule->u_states_cur = counter_u64_fetch(krule->states_cur);
1599 	rule->u_states_tot = counter_u64_fetch(krule->states_tot);
1600 	rule->u_src_nodes = counter_u64_fetch(krule->src_nodes);
1601 }
1602 
1603 static int
1604 pf_check_rule_addr(const struct pf_rule_addr *addr)
1605 {
1606 
1607 	switch (addr->addr.type) {
1608 	case PF_ADDR_ADDRMASK:
1609 	case PF_ADDR_NOROUTE:
1610 	case PF_ADDR_DYNIFTL:
1611 	case PF_ADDR_TABLE:
1612 	case PF_ADDR_URPFFAILED:
1613 	case PF_ADDR_RANGE:
1614 		break;
1615 	default:
1616 		return (EINVAL);
1617 	}
1618 
1619 	if (addr->addr.p.dyn != NULL) {
1620 		return (EINVAL);
1621 	}
1622 
1623 	return (0);
1624 }
1625 
1626 static int
1627 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
1628 {
1629 	int ret;
1630 
1631 #ifndef INET
1632 	if (rule->af == AF_INET) {
1633 		return (EAFNOSUPPORT);
1634 	}
1635 #endif /* INET */
1636 #ifndef INET6
1637 	if (rule->af == AF_INET6) {
1638 		return (EAFNOSUPPORT);
1639 	}
1640 #endif /* INET6 */
1641 
1642 	ret = pf_check_rule_addr(&rule->src);
1643 	if (ret != 0)
1644 		return (ret);
1645 	ret = pf_check_rule_addr(&rule->dst);
1646 	if (ret != 0)
1647 		return (ret);
1648 
1649 	bzero(krule, sizeof(*krule));
1650 
1651 	bcopy(&rule->src, &krule->src, sizeof(rule->src));
1652 	bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
1653 
1654 	strlcpy(krule->label, rule->label, sizeof(rule->label));
1655 	strlcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
1656 	strlcpy(krule->qname, rule->qname, sizeof(rule->qname));
1657 	strlcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
1658 	strlcpy(krule->tagname, rule->tagname, sizeof(rule->tagname));
1659 	strlcpy(krule->match_tagname, rule->match_tagname,
1660 	    sizeof(rule->match_tagname));
1661 	strlcpy(krule->overload_tblname, rule->overload_tblname,
1662 	    sizeof(rule->overload_tblname));
1663 
1664 	ret = pf_pool_to_kpool(&rule->rpool, &krule->rpool);
1665 	if (ret != 0)
1666 		return (ret);
1667 
1668 	/* Don't allow userspace to set evaulations, packets or bytes. */
1669 	/* kif, anchor, overload_tbl are not copied over. */
1670 
1671 	krule->os_fingerprint = rule->os_fingerprint;
1672 
1673 	krule->rtableid = rule->rtableid;
1674 	bcopy(rule->timeout, krule->timeout, sizeof(krule->timeout));
1675 	krule->max_states = rule->max_states;
1676 	krule->max_src_nodes = rule->max_src_nodes;
1677 	krule->max_src_states = rule->max_src_states;
1678 	krule->max_src_conn = rule->max_src_conn;
1679 	krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
1680 	krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
1681 	krule->qid = rule->qid;
1682 	krule->pqid = rule->pqid;
1683 	krule->rt_listid = rule->rt_listid;
1684 	krule->nr = rule->nr;
1685 	krule->prob = rule->prob;
1686 	krule->cuid = rule->cuid;
1687 	krule->cpid = rule->cpid;
1688 
1689 	krule->return_icmp = rule->return_icmp;
1690 	krule->return_icmp6 = rule->return_icmp6;
1691 	krule->max_mss = rule->max_mss;
1692 	krule->tag = rule->tag;
1693 	krule->match_tag = rule->match_tag;
1694 	krule->scrub_flags = rule->scrub_flags;
1695 
1696 	bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
1697 	bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
1698 
1699 	krule->rule_flag = rule->rule_flag;
1700 	krule->action = rule->action;
1701 	krule->direction = rule->direction;
1702 	krule->log = rule->log;
1703 	krule->logif = rule->logif;
1704 	krule->quick = rule->quick;
1705 	krule->ifnot = rule->ifnot;
1706 	krule->match_tag_not = rule->match_tag_not;
1707 	krule->natpass = rule->natpass;
1708 
1709 	krule->keep_state = rule->keep_state;
1710 	krule->af = rule->af;
1711 	krule->proto = rule->proto;
1712 	krule->type = rule->type;
1713 	krule->code = rule->code;
1714 	krule->flags = rule->flags;
1715 	krule->flagset = rule->flagset;
1716 	krule->min_ttl = rule->min_ttl;
1717 	krule->allow_opts = rule->allow_opts;
1718 	krule->rt = rule->rt;
1719 	krule->return_ttl = rule->return_ttl;
1720 	krule->tos = rule->tos;
1721 	krule->set_tos = rule->set_tos;
1722 	krule->anchor_relative = rule->anchor_relative;
1723 	krule->anchor_wildcard = rule->anchor_wildcard;
1724 
1725 	krule->flush = rule->flush;
1726 	krule->prio = rule->prio;
1727 	krule->set_prio[0] = rule->set_prio[0];
1728 	krule->set_prio[1] = rule->set_prio[1];
1729 
1730 	bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
1731 
1732 	return (0);
1733 }
1734 
1735 static int
1736 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
1737 {
1738 	int			 error = 0;
1739 	PF_RULES_RLOCK_TRACKER;
1740 
1741 	/* XXX keep in sync with switch() below */
1742 	if (securelevel_gt(td->td_ucred, 2))
1743 		switch (cmd) {
1744 		case DIOCGETRULES:
1745 		case DIOCGETRULE:
1746 		case DIOCGETADDRS:
1747 		case DIOCGETADDR:
1748 		case DIOCGETSTATE:
1749 		case DIOCSETSTATUSIF:
1750 		case DIOCGETSTATUS:
1751 		case DIOCCLRSTATUS:
1752 		case DIOCNATLOOK:
1753 		case DIOCSETDEBUG:
1754 		case DIOCGETSTATES:
1755 		case DIOCGETTIMEOUT:
1756 		case DIOCCLRRULECTRS:
1757 		case DIOCGETLIMIT:
1758 		case DIOCGETALTQSV0:
1759 		case DIOCGETALTQSV1:
1760 		case DIOCGETALTQV0:
1761 		case DIOCGETALTQV1:
1762 		case DIOCGETQSTATSV0:
1763 		case DIOCGETQSTATSV1:
1764 		case DIOCGETRULESETS:
1765 		case DIOCGETRULESET:
1766 		case DIOCRGETTABLES:
1767 		case DIOCRGETTSTATS:
1768 		case DIOCRCLRTSTATS:
1769 		case DIOCRCLRADDRS:
1770 		case DIOCRADDADDRS:
1771 		case DIOCRDELADDRS:
1772 		case DIOCRSETADDRS:
1773 		case DIOCRGETADDRS:
1774 		case DIOCRGETASTATS:
1775 		case DIOCRCLRASTATS:
1776 		case DIOCRTSTADDRS:
1777 		case DIOCOSFPGET:
1778 		case DIOCGETSRCNODES:
1779 		case DIOCCLRSRCNODES:
1780 		case DIOCIGETIFACES:
1781 		case DIOCGIFSPEEDV0:
1782 		case DIOCGIFSPEEDV1:
1783 		case DIOCSETIFFLAG:
1784 		case DIOCCLRIFFLAG:
1785 			break;
1786 		case DIOCRCLRTABLES:
1787 		case DIOCRADDTABLES:
1788 		case DIOCRDELTABLES:
1789 		case DIOCRSETTFLAGS:
1790 			if (((struct pfioc_table *)addr)->pfrio_flags &
1791 			    PFR_FLAG_DUMMY)
1792 				break; /* dummy operation ok */
1793 			return (EPERM);
1794 		default:
1795 			return (EPERM);
1796 		}
1797 
1798 	if (!(flags & FWRITE))
1799 		switch (cmd) {
1800 		case DIOCGETRULES:
1801 		case DIOCGETADDRS:
1802 		case DIOCGETADDR:
1803 		case DIOCGETSTATE:
1804 		case DIOCGETSTATUS:
1805 		case DIOCGETSTATES:
1806 		case DIOCGETTIMEOUT:
1807 		case DIOCGETLIMIT:
1808 		case DIOCGETALTQSV0:
1809 		case DIOCGETALTQSV1:
1810 		case DIOCGETALTQV0:
1811 		case DIOCGETALTQV1:
1812 		case DIOCGETQSTATSV0:
1813 		case DIOCGETQSTATSV1:
1814 		case DIOCGETRULESETS:
1815 		case DIOCGETRULESET:
1816 		case DIOCNATLOOK:
1817 		case DIOCRGETTABLES:
1818 		case DIOCRGETTSTATS:
1819 		case DIOCRGETADDRS:
1820 		case DIOCRGETASTATS:
1821 		case DIOCRTSTADDRS:
1822 		case DIOCOSFPGET:
1823 		case DIOCGETSRCNODES:
1824 		case DIOCIGETIFACES:
1825 		case DIOCGIFSPEEDV1:
1826 		case DIOCGIFSPEEDV0:
1827 			break;
1828 		case DIOCRCLRTABLES:
1829 		case DIOCRADDTABLES:
1830 		case DIOCRDELTABLES:
1831 		case DIOCRCLRTSTATS:
1832 		case DIOCRCLRADDRS:
1833 		case DIOCRADDADDRS:
1834 		case DIOCRDELADDRS:
1835 		case DIOCRSETADDRS:
1836 		case DIOCRSETTFLAGS:
1837 			if (((struct pfioc_table *)addr)->pfrio_flags &
1838 			    PFR_FLAG_DUMMY) {
1839 				flags |= FWRITE; /* need write lock for dummy */
1840 				break; /* dummy operation ok */
1841 			}
1842 			return (EACCES);
1843 		case DIOCGETRULE:
1844 			if (((struct pfioc_rule *)addr)->action ==
1845 			    PF_GET_CLR_CNTR)
1846 				return (EACCES);
1847 			break;
1848 		default:
1849 			return (EACCES);
1850 		}
1851 
1852 	CURVNET_SET(TD_TO_VNET(td));
1853 
1854 	switch (cmd) {
1855 	case DIOCSTART:
1856 		sx_xlock(&pf_ioctl_lock);
1857 		if (V_pf_status.running)
1858 			error = EEXIST;
1859 		else {
1860 			int cpu;
1861 
1862 			hook_pf();
1863 			V_pf_status.running = 1;
1864 			V_pf_status.since = time_second;
1865 
1866 			CPU_FOREACH(cpu)
1867 				V_pf_stateid[cpu] = time_second;
1868 
1869 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1870 		}
1871 		break;
1872 
1873 	case DIOCSTOP:
1874 		sx_xlock(&pf_ioctl_lock);
1875 		if (!V_pf_status.running)
1876 			error = ENOENT;
1877 		else {
1878 			V_pf_status.running = 0;
1879 			dehook_pf();
1880 			V_pf_status.since = time_second;
1881 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1882 		}
1883 		break;
1884 
1885 	case DIOCADDRULE: {
1886 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1887 		struct pf_kruleset	*ruleset;
1888 		struct pf_krule		*rule, *tail;
1889 		struct pf_kpooladdr	*pa;
1890 		struct pfi_kkif		*kif = NULL;
1891 		int			 rs_num;
1892 
1893 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1894 			error = EINVAL;
1895 			break;
1896 		}
1897 
1898 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
1899 		error = pf_rule_to_krule(&pr->rule, rule);
1900 		if (error != 0) {
1901 			free(rule, M_PFRULE);
1902 			break;
1903 		}
1904 
1905 		if (rule->ifname[0])
1906 			kif = pf_kkif_create(M_WAITOK);
1907 		rule->evaluations = counter_u64_alloc(M_WAITOK);
1908 		for (int i = 0; i < 2; i++) {
1909 			rule->packets[i] = counter_u64_alloc(M_WAITOK);
1910 			rule->bytes[i] = counter_u64_alloc(M_WAITOK);
1911 		}
1912 		rule->states_cur = counter_u64_alloc(M_WAITOK);
1913 		rule->states_tot = counter_u64_alloc(M_WAITOK);
1914 		rule->src_nodes = counter_u64_alloc(M_WAITOK);
1915 		rule->cuid = td->td_ucred->cr_ruid;
1916 		rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1917 		TAILQ_INIT(&rule->rpool.list);
1918 #define	ERROUT(x)	{ error = (x); goto DIOCADDRULE_error; }
1919 
1920 		PF_RULES_WLOCK();
1921 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1922 		ruleset = pf_find_kruleset(pr->anchor);
1923 		if (ruleset == NULL)
1924 			ERROUT(EINVAL);
1925 		rs_num = pf_get_ruleset_number(pr->rule.action);
1926 		if (rs_num >= PF_RULESET_MAX)
1927 			ERROUT(EINVAL);
1928 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1929 			DPFPRINTF(PF_DEBUG_MISC,
1930 			    ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
1931 			    ruleset->rules[rs_num].inactive.ticket));
1932 			ERROUT(EBUSY);
1933 		}
1934 		if (pr->pool_ticket != V_ticket_pabuf) {
1935 			DPFPRINTF(PF_DEBUG_MISC,
1936 			    ("pool_ticket: %d != %d\n", pr->pool_ticket,
1937 			    V_ticket_pabuf));
1938 			ERROUT(EBUSY);
1939 		}
1940 
1941 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1942 		    pf_krulequeue);
1943 		if (tail)
1944 			rule->nr = tail->nr + 1;
1945 		else
1946 			rule->nr = 0;
1947 		if (rule->ifname[0]) {
1948 			rule->kif = pfi_kkif_attach(kif, rule->ifname);
1949 			pfi_kkif_ref(rule->kif);
1950 		} else
1951 			rule->kif = NULL;
1952 
1953 		if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
1954 			error = EBUSY;
1955 
1956 #ifdef ALTQ
1957 		/* set queue IDs */
1958 		if (rule->qname[0] != 0) {
1959 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1960 				error = EBUSY;
1961 			else if (rule->pqname[0] != 0) {
1962 				if ((rule->pqid =
1963 				    pf_qname2qid(rule->pqname)) == 0)
1964 					error = EBUSY;
1965 			} else
1966 				rule->pqid = rule->qid;
1967 		}
1968 #endif
1969 		if (rule->tagname[0])
1970 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1971 				error = EBUSY;
1972 		if (rule->match_tagname[0])
1973 			if ((rule->match_tag =
1974 			    pf_tagname2tag(rule->match_tagname)) == 0)
1975 				error = EBUSY;
1976 		if (rule->rt && !rule->direction)
1977 			error = EINVAL;
1978 		if (!rule->log)
1979 			rule->logif = 0;
1980 		if (rule->logif >= PFLOGIFS_MAX)
1981 			error = EINVAL;
1982 		if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1983 			error = ENOMEM;
1984 		if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1985 			error = ENOMEM;
1986 		if (pf_kanchor_setup(rule, ruleset, pr->anchor_call))
1987 			error = EINVAL;
1988 		if (rule->scrub_flags & PFSTATE_SETPRIO &&
1989 		    (rule->set_prio[0] > PF_PRIO_MAX ||
1990 		    rule->set_prio[1] > PF_PRIO_MAX))
1991 			error = EINVAL;
1992 		TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1993 			if (pa->addr.type == PF_ADDR_TABLE) {
1994 				pa->addr.p.tbl = pfr_attach_table(ruleset,
1995 				    pa->addr.v.tblname);
1996 				if (pa->addr.p.tbl == NULL)
1997 					error = ENOMEM;
1998 			}
1999 
2000 		rule->overload_tbl = NULL;
2001 		if (rule->overload_tblname[0]) {
2002 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
2003 			    rule->overload_tblname)) == NULL)
2004 				error = EINVAL;
2005 			else
2006 				rule->overload_tbl->pfrkt_flags |=
2007 				    PFR_TFLAG_ACTIVE;
2008 		}
2009 
2010 		pf_mv_kpool(&V_pf_pabuf, &rule->rpool.list);
2011 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
2012 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
2013 		    (rule->rt > PF_NOPFROUTE)) &&
2014 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
2015 			error = EINVAL;
2016 
2017 		if (error) {
2018 			pf_free_rule(rule);
2019 			PF_RULES_WUNLOCK();
2020 			break;
2021 		}
2022 
2023 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
2024 		counter_u64_zero(rule->evaluations);
2025 		for (int i = 0; i < 2; i++) {
2026 			counter_u64_zero(rule->packets[i]);
2027 			counter_u64_zero(rule->bytes[i]);
2028 		}
2029 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
2030 		    rule, entries);
2031 		ruleset->rules[rs_num].inactive.rcount++;
2032 		PF_RULES_WUNLOCK();
2033 		break;
2034 
2035 #undef ERROUT
2036 DIOCADDRULE_error:
2037 		PF_RULES_WUNLOCK();
2038 		pf_krule_free(rule);
2039 		pf_kkif_free(kif);
2040 		break;
2041 	}
2042 
2043 	case DIOCGETRULES: {
2044 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
2045 		struct pf_kruleset	*ruleset;
2046 		struct pf_krule		*tail;
2047 		int			 rs_num;
2048 
2049 		PF_RULES_WLOCK();
2050 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
2051 		ruleset = pf_find_kruleset(pr->anchor);
2052 		if (ruleset == NULL) {
2053 			PF_RULES_WUNLOCK();
2054 			error = EINVAL;
2055 			break;
2056 		}
2057 		rs_num = pf_get_ruleset_number(pr->rule.action);
2058 		if (rs_num >= PF_RULESET_MAX) {
2059 			PF_RULES_WUNLOCK();
2060 			error = EINVAL;
2061 			break;
2062 		}
2063 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
2064 		    pf_krulequeue);
2065 		if (tail)
2066 			pr->nr = tail->nr + 1;
2067 		else
2068 			pr->nr = 0;
2069 		pr->ticket = ruleset->rules[rs_num].active.ticket;
2070 		PF_RULES_WUNLOCK();
2071 		break;
2072 	}
2073 
2074 	case DIOCGETRULE: {
2075 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
2076 		struct pf_kruleset	*ruleset;
2077 		struct pf_krule		*rule;
2078 		int			 rs_num;
2079 
2080 		PF_RULES_WLOCK();
2081 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
2082 		ruleset = pf_find_kruleset(pr->anchor);
2083 		if (ruleset == NULL) {
2084 			PF_RULES_WUNLOCK();
2085 			error = EINVAL;
2086 			break;
2087 		}
2088 		rs_num = pf_get_ruleset_number(pr->rule.action);
2089 		if (rs_num >= PF_RULESET_MAX) {
2090 			PF_RULES_WUNLOCK();
2091 			error = EINVAL;
2092 			break;
2093 		}
2094 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
2095 			PF_RULES_WUNLOCK();
2096 			error = EBUSY;
2097 			break;
2098 		}
2099 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
2100 		while ((rule != NULL) && (rule->nr != pr->nr))
2101 			rule = TAILQ_NEXT(rule, entries);
2102 		if (rule == NULL) {
2103 			PF_RULES_WUNLOCK();
2104 			error = EBUSY;
2105 			break;
2106 		}
2107 
2108 		pf_krule_to_rule(rule, &pr->rule);
2109 
2110 		if (pf_kanchor_copyout(ruleset, rule, pr)) {
2111 			PF_RULES_WUNLOCK();
2112 			error = EBUSY;
2113 			break;
2114 		}
2115 		pf_addr_copyout(&pr->rule.src.addr);
2116 		pf_addr_copyout(&pr->rule.dst.addr);
2117 
2118 		if (pr->action == PF_GET_CLR_CNTR) {
2119 			counter_u64_zero(rule->evaluations);
2120 			for (int i = 0; i < 2; i++) {
2121 				counter_u64_zero(rule->packets[i]);
2122 				counter_u64_zero(rule->bytes[i]);
2123 			}
2124 			counter_u64_zero(rule->states_tot);
2125 		}
2126 		PF_RULES_WUNLOCK();
2127 		break;
2128 	}
2129 
2130 	case DIOCCHANGERULE: {
2131 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
2132 		struct pf_kruleset	*ruleset;
2133 		struct pf_krule		*oldrule = NULL, *newrule = NULL;
2134 		struct pfi_kkif		*kif = NULL;
2135 		struct pf_kpooladdr	*pa;
2136 		u_int32_t		 nr = 0;
2137 		int			 rs_num;
2138 
2139 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
2140 		    pcr->action > PF_CHANGE_GET_TICKET) {
2141 			error = EINVAL;
2142 			break;
2143 		}
2144 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
2145 			error = EINVAL;
2146 			break;
2147 		}
2148 
2149 		if (pcr->action != PF_CHANGE_REMOVE) {
2150 			newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
2151 			error = pf_rule_to_krule(&pcr->rule, newrule);
2152 			if (error != 0) {
2153 				free(newrule, M_PFRULE);
2154 				break;
2155 			}
2156 
2157 			if (newrule->ifname[0])
2158 				kif = pf_kkif_create(M_WAITOK);
2159 			newrule->evaluations = counter_u64_alloc(M_WAITOK);
2160 			for (int i = 0; i < 2; i++) {
2161 				newrule->packets[i] =
2162 				    counter_u64_alloc(M_WAITOK);
2163 				newrule->bytes[i] =
2164 				    counter_u64_alloc(M_WAITOK);
2165 			}
2166 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
2167 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
2168 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
2169 			newrule->cuid = td->td_ucred->cr_ruid;
2170 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
2171 			TAILQ_INIT(&newrule->rpool.list);
2172 		}
2173 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGERULE_error; }
2174 
2175 		PF_RULES_WLOCK();
2176 		if (!(pcr->action == PF_CHANGE_REMOVE ||
2177 		    pcr->action == PF_CHANGE_GET_TICKET) &&
2178 		    pcr->pool_ticket != V_ticket_pabuf)
2179 			ERROUT(EBUSY);
2180 
2181 		ruleset = pf_find_kruleset(pcr->anchor);
2182 		if (ruleset == NULL)
2183 			ERROUT(EINVAL);
2184 
2185 		rs_num = pf_get_ruleset_number(pcr->rule.action);
2186 		if (rs_num >= PF_RULESET_MAX)
2187 			ERROUT(EINVAL);
2188 
2189 		if (pcr->action == PF_CHANGE_GET_TICKET) {
2190 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
2191 			ERROUT(0);
2192 		} else if (pcr->ticket !=
2193 			    ruleset->rules[rs_num].active.ticket)
2194 				ERROUT(EINVAL);
2195 
2196 		if (pcr->action != PF_CHANGE_REMOVE) {
2197 			if (newrule->ifname[0]) {
2198 				newrule->kif = pfi_kkif_attach(kif,
2199 				    newrule->ifname);
2200 				pfi_kkif_ref(newrule->kif);
2201 			} else
2202 				newrule->kif = NULL;
2203 
2204 			if (newrule->rtableid > 0 &&
2205 			    newrule->rtableid >= rt_numfibs)
2206 				error = EBUSY;
2207 
2208 #ifdef ALTQ
2209 			/* set queue IDs */
2210 			if (newrule->qname[0] != 0) {
2211 				if ((newrule->qid =
2212 				    pf_qname2qid(newrule->qname)) == 0)
2213 					error = EBUSY;
2214 				else if (newrule->pqname[0] != 0) {
2215 					if ((newrule->pqid =
2216 					    pf_qname2qid(newrule->pqname)) == 0)
2217 						error = EBUSY;
2218 				} else
2219 					newrule->pqid = newrule->qid;
2220 			}
2221 #endif /* ALTQ */
2222 			if (newrule->tagname[0])
2223 				if ((newrule->tag =
2224 				    pf_tagname2tag(newrule->tagname)) == 0)
2225 					error = EBUSY;
2226 			if (newrule->match_tagname[0])
2227 				if ((newrule->match_tag = pf_tagname2tag(
2228 				    newrule->match_tagname)) == 0)
2229 					error = EBUSY;
2230 			if (newrule->rt && !newrule->direction)
2231 				error = EINVAL;
2232 			if (!newrule->log)
2233 				newrule->logif = 0;
2234 			if (newrule->logif >= PFLOGIFS_MAX)
2235 				error = EINVAL;
2236 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
2237 				error = ENOMEM;
2238 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
2239 				error = ENOMEM;
2240 			if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
2241 				error = EINVAL;
2242 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
2243 				if (pa->addr.type == PF_ADDR_TABLE) {
2244 					pa->addr.p.tbl =
2245 					    pfr_attach_table(ruleset,
2246 					    pa->addr.v.tblname);
2247 					if (pa->addr.p.tbl == NULL)
2248 						error = ENOMEM;
2249 				}
2250 
2251 			newrule->overload_tbl = NULL;
2252 			if (newrule->overload_tblname[0]) {
2253 				if ((newrule->overload_tbl = pfr_attach_table(
2254 				    ruleset, newrule->overload_tblname)) ==
2255 				    NULL)
2256 					error = EINVAL;
2257 				else
2258 					newrule->overload_tbl->pfrkt_flags |=
2259 					    PFR_TFLAG_ACTIVE;
2260 			}
2261 
2262 			pf_mv_kpool(&V_pf_pabuf, &newrule->rpool.list);
2263 			if (((((newrule->action == PF_NAT) ||
2264 			    (newrule->action == PF_RDR) ||
2265 			    (newrule->action == PF_BINAT) ||
2266 			    (newrule->rt > PF_NOPFROUTE)) &&
2267 			    !newrule->anchor)) &&
2268 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
2269 				error = EINVAL;
2270 
2271 			if (error) {
2272 				pf_free_rule(newrule);
2273 				PF_RULES_WUNLOCK();
2274 				break;
2275 			}
2276 
2277 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
2278 		}
2279 		pf_empty_kpool(&V_pf_pabuf);
2280 
2281 		if (pcr->action == PF_CHANGE_ADD_HEAD)
2282 			oldrule = TAILQ_FIRST(
2283 			    ruleset->rules[rs_num].active.ptr);
2284 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
2285 			oldrule = TAILQ_LAST(
2286 			    ruleset->rules[rs_num].active.ptr, pf_krulequeue);
2287 		else {
2288 			oldrule = TAILQ_FIRST(
2289 			    ruleset->rules[rs_num].active.ptr);
2290 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
2291 				oldrule = TAILQ_NEXT(oldrule, entries);
2292 			if (oldrule == NULL) {
2293 				if (newrule != NULL)
2294 					pf_free_rule(newrule);
2295 				PF_RULES_WUNLOCK();
2296 				error = EINVAL;
2297 				break;
2298 			}
2299 		}
2300 
2301 		if (pcr->action == PF_CHANGE_REMOVE) {
2302 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
2303 			    oldrule);
2304 			ruleset->rules[rs_num].active.rcount--;
2305 		} else {
2306 			if (oldrule == NULL)
2307 				TAILQ_INSERT_TAIL(
2308 				    ruleset->rules[rs_num].active.ptr,
2309 				    newrule, entries);
2310 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
2311 			    pcr->action == PF_CHANGE_ADD_BEFORE)
2312 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
2313 			else
2314 				TAILQ_INSERT_AFTER(
2315 				    ruleset->rules[rs_num].active.ptr,
2316 				    oldrule, newrule, entries);
2317 			ruleset->rules[rs_num].active.rcount++;
2318 		}
2319 
2320 		nr = 0;
2321 		TAILQ_FOREACH(oldrule,
2322 		    ruleset->rules[rs_num].active.ptr, entries)
2323 			oldrule->nr = nr++;
2324 
2325 		ruleset->rules[rs_num].active.ticket++;
2326 
2327 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
2328 		pf_remove_if_empty_kruleset(ruleset);
2329 
2330 		PF_RULES_WUNLOCK();
2331 		break;
2332 
2333 #undef ERROUT
2334 DIOCCHANGERULE_error:
2335 		PF_RULES_WUNLOCK();
2336 		pf_krule_free(newrule);
2337 		pf_kkif_free(kif);
2338 		break;
2339 	}
2340 
2341 	case DIOCCLRSTATES: {
2342 		struct pf_state		*s;
2343 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
2344 		u_int			 i, killed = 0;
2345 
2346 		for (i = 0; i <= pf_hashmask; i++) {
2347 			struct pf_idhash *ih = &V_pf_idhash[i];
2348 
2349 relock_DIOCCLRSTATES:
2350 			PF_HASHROW_LOCK(ih);
2351 			LIST_FOREACH(s, &ih->states, entry)
2352 				if (!psk->psk_ifname[0] ||
2353 				    !strcmp(psk->psk_ifname,
2354 				    s->kif->pfik_name)) {
2355 					/*
2356 					 * Don't send out individual
2357 					 * delete messages.
2358 					 */
2359 					s->state_flags |= PFSTATE_NOSYNC;
2360 					pf_unlink_state(s, PF_ENTER_LOCKED);
2361 					killed++;
2362 					goto relock_DIOCCLRSTATES;
2363 				}
2364 			PF_HASHROW_UNLOCK(ih);
2365 		}
2366 		psk->psk_killed = killed;
2367 		if (V_pfsync_clear_states_ptr != NULL)
2368 			V_pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
2369 		break;
2370 	}
2371 
2372 	case DIOCKILLSTATES: {
2373 		struct pf_state		*s;
2374 		struct pf_state_key	*sk;
2375 		struct pf_addr		*srcaddr, *dstaddr;
2376 		u_int16_t		 srcport, dstport;
2377 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
2378 		u_int			 i, killed = 0;
2379 
2380 		if (psk->psk_pfcmp.id) {
2381 			if (psk->psk_pfcmp.creatorid == 0)
2382 				psk->psk_pfcmp.creatorid = V_pf_status.hostid;
2383 			if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
2384 			    psk->psk_pfcmp.creatorid))) {
2385 				pf_unlink_state(s, PF_ENTER_LOCKED);
2386 				psk->psk_killed = 1;
2387 			}
2388 			break;
2389 		}
2390 
2391 		for (i = 0; i <= pf_hashmask; i++) {
2392 			struct pf_idhash *ih = &V_pf_idhash[i];
2393 
2394 relock_DIOCKILLSTATES:
2395 			PF_HASHROW_LOCK(ih);
2396 			LIST_FOREACH(s, &ih->states, entry) {
2397 				sk = s->key[PF_SK_WIRE];
2398 				if (s->direction == PF_OUT) {
2399 					srcaddr = &sk->addr[1];
2400 					dstaddr = &sk->addr[0];
2401 					srcport = sk->port[1];
2402 					dstport = sk->port[0];
2403 				} else {
2404 					srcaddr = &sk->addr[0];
2405 					dstaddr = &sk->addr[1];
2406 					srcport = sk->port[0];
2407 					dstport = sk->port[1];
2408 				}
2409 
2410 				if ((!psk->psk_af || sk->af == psk->psk_af)
2411 				    && (!psk->psk_proto || psk->psk_proto ==
2412 				    sk->proto) &&
2413 				    PF_MATCHA(psk->psk_src.neg,
2414 				    &psk->psk_src.addr.v.a.addr,
2415 				    &psk->psk_src.addr.v.a.mask,
2416 				    srcaddr, sk->af) &&
2417 				    PF_MATCHA(psk->psk_dst.neg,
2418 				    &psk->psk_dst.addr.v.a.addr,
2419 				    &psk->psk_dst.addr.v.a.mask,
2420 				    dstaddr, sk->af) &&
2421 				    (psk->psk_src.port_op == 0 ||
2422 				    pf_match_port(psk->psk_src.port_op,
2423 				    psk->psk_src.port[0], psk->psk_src.port[1],
2424 				    srcport)) &&
2425 				    (psk->psk_dst.port_op == 0 ||
2426 				    pf_match_port(psk->psk_dst.port_op,
2427 				    psk->psk_dst.port[0], psk->psk_dst.port[1],
2428 				    dstport)) &&
2429 				    (!psk->psk_label[0] ||
2430 				    (s->rule.ptr->label[0] &&
2431 				    !strcmp(psk->psk_label,
2432 				    s->rule.ptr->label))) &&
2433 				    (!psk->psk_ifname[0] ||
2434 				    !strcmp(psk->psk_ifname,
2435 				    s->kif->pfik_name))) {
2436 					pf_unlink_state(s, PF_ENTER_LOCKED);
2437 					killed++;
2438 					goto relock_DIOCKILLSTATES;
2439 				}
2440 			}
2441 			PF_HASHROW_UNLOCK(ih);
2442 		}
2443 		psk->psk_killed = killed;
2444 		break;
2445 	}
2446 
2447 	case DIOCADDSTATE: {
2448 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
2449 		struct pfsync_state	*sp = &ps->state;
2450 
2451 		if (sp->timeout >= PFTM_MAX) {
2452 			error = EINVAL;
2453 			break;
2454 		}
2455 		if (V_pfsync_state_import_ptr != NULL) {
2456 			PF_RULES_RLOCK();
2457 			error = V_pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
2458 			PF_RULES_RUNLOCK();
2459 		} else
2460 			error = EOPNOTSUPP;
2461 		break;
2462 	}
2463 
2464 	case DIOCGETSTATE: {
2465 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
2466 		struct pf_state		*s;
2467 
2468 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
2469 		if (s == NULL) {
2470 			error = ENOENT;
2471 			break;
2472 		}
2473 
2474 		pfsync_state_export(&ps->state, s);
2475 		PF_STATE_UNLOCK(s);
2476 		break;
2477 	}
2478 
2479 	case DIOCGETSTATES: {
2480 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
2481 		struct pf_state		*s;
2482 		struct pfsync_state	*pstore, *p;
2483 		int i, nr;
2484 
2485 		if (ps->ps_len <= 0) {
2486 			nr = uma_zone_get_cur(V_pf_state_z);
2487 			ps->ps_len = sizeof(struct pfsync_state) * nr;
2488 			break;
2489 		}
2490 
2491 		p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK | M_ZERO);
2492 		nr = 0;
2493 
2494 		for (i = 0; i <= pf_hashmask; i++) {
2495 			struct pf_idhash *ih = &V_pf_idhash[i];
2496 
2497 			PF_HASHROW_LOCK(ih);
2498 			LIST_FOREACH(s, &ih->states, entry) {
2499 				if (s->timeout == PFTM_UNLINKED)
2500 					continue;
2501 
2502 				if ((nr+1) * sizeof(*p) > ps->ps_len) {
2503 					PF_HASHROW_UNLOCK(ih);
2504 					goto DIOCGETSTATES_full;
2505 				}
2506 				pfsync_state_export(p, s);
2507 				p++;
2508 				nr++;
2509 			}
2510 			PF_HASHROW_UNLOCK(ih);
2511 		}
2512 DIOCGETSTATES_full:
2513 		error = copyout(pstore, ps->ps_states,
2514 		    sizeof(struct pfsync_state) * nr);
2515 		if (error) {
2516 			free(pstore, M_TEMP);
2517 			break;
2518 		}
2519 		ps->ps_len = sizeof(struct pfsync_state) * nr;
2520 		free(pstore, M_TEMP);
2521 
2522 		break;
2523 	}
2524 
2525 	case DIOCGETSTATUS: {
2526 		struct pf_status *s = (struct pf_status *)addr;
2527 
2528 		PF_RULES_RLOCK();
2529 		s->running = V_pf_status.running;
2530 		s->since   = V_pf_status.since;
2531 		s->debug   = V_pf_status.debug;
2532 		s->hostid  = V_pf_status.hostid;
2533 		s->states  = V_pf_status.states;
2534 		s->src_nodes = V_pf_status.src_nodes;
2535 
2536 		for (int i = 0; i < PFRES_MAX; i++)
2537 			s->counters[i] =
2538 			    counter_u64_fetch(V_pf_status.counters[i]);
2539 		for (int i = 0; i < LCNT_MAX; i++)
2540 			s->lcounters[i] =
2541 			    counter_u64_fetch(V_pf_status.lcounters[i]);
2542 		for (int i = 0; i < FCNT_MAX; i++)
2543 			s->fcounters[i] =
2544 			    counter_u64_fetch(V_pf_status.fcounters[i]);
2545 		for (int i = 0; i < SCNT_MAX; i++)
2546 			s->scounters[i] =
2547 			    counter_u64_fetch(V_pf_status.scounters[i]);
2548 
2549 		bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ);
2550 		bcopy(V_pf_status.pf_chksum, s->pf_chksum,
2551 		    PF_MD5_DIGEST_LENGTH);
2552 
2553 		pfi_update_status(s->ifname, s);
2554 		PF_RULES_RUNLOCK();
2555 		break;
2556 	}
2557 
2558 	case DIOCSETSTATUSIF: {
2559 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
2560 
2561 		if (pi->ifname[0] == 0) {
2562 			bzero(V_pf_status.ifname, IFNAMSIZ);
2563 			break;
2564 		}
2565 		PF_RULES_WLOCK();
2566 		strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
2567 		PF_RULES_WUNLOCK();
2568 		break;
2569 	}
2570 
2571 	case DIOCCLRSTATUS: {
2572 		PF_RULES_WLOCK();
2573 		for (int i = 0; i < PFRES_MAX; i++)
2574 			counter_u64_zero(V_pf_status.counters[i]);
2575 		for (int i = 0; i < FCNT_MAX; i++)
2576 			counter_u64_zero(V_pf_status.fcounters[i]);
2577 		for (int i = 0; i < SCNT_MAX; i++)
2578 			counter_u64_zero(V_pf_status.scounters[i]);
2579 		for (int i = 0; i < LCNT_MAX; i++)
2580 			counter_u64_zero(V_pf_status.lcounters[i]);
2581 		V_pf_status.since = time_second;
2582 		if (*V_pf_status.ifname)
2583 			pfi_update_status(V_pf_status.ifname, NULL);
2584 		PF_RULES_WUNLOCK();
2585 		break;
2586 	}
2587 
2588 	case DIOCNATLOOK: {
2589 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
2590 		struct pf_state_key	*sk;
2591 		struct pf_state		*state;
2592 		struct pf_state_key_cmp	 key;
2593 		int			 m = 0, direction = pnl->direction;
2594 		int			 sidx, didx;
2595 
2596 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
2597 		sidx = (direction == PF_IN) ? 1 : 0;
2598 		didx = (direction == PF_IN) ? 0 : 1;
2599 
2600 		if (!pnl->proto ||
2601 		    PF_AZERO(&pnl->saddr, pnl->af) ||
2602 		    PF_AZERO(&pnl->daddr, pnl->af) ||
2603 		    ((pnl->proto == IPPROTO_TCP ||
2604 		    pnl->proto == IPPROTO_UDP) &&
2605 		    (!pnl->dport || !pnl->sport)))
2606 			error = EINVAL;
2607 		else {
2608 			bzero(&key, sizeof(key));
2609 			key.af = pnl->af;
2610 			key.proto = pnl->proto;
2611 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
2612 			key.port[sidx] = pnl->sport;
2613 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
2614 			key.port[didx] = pnl->dport;
2615 
2616 			state = pf_find_state_all(&key, direction, &m);
2617 
2618 			if (m > 1)
2619 				error = E2BIG;	/* more than one state */
2620 			else if (state != NULL) {
2621 				/* XXXGL: not locked read */
2622 				sk = state->key[sidx];
2623 				PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
2624 				pnl->rsport = sk->port[sidx];
2625 				PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
2626 				pnl->rdport = sk->port[didx];
2627 			} else
2628 				error = ENOENT;
2629 		}
2630 		break;
2631 	}
2632 
2633 	case DIOCSETTIMEOUT: {
2634 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2635 		int		 old;
2636 
2637 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
2638 		    pt->seconds < 0) {
2639 			error = EINVAL;
2640 			break;
2641 		}
2642 		PF_RULES_WLOCK();
2643 		old = V_pf_default_rule.timeout[pt->timeout];
2644 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
2645 			pt->seconds = 1;
2646 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
2647 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
2648 			wakeup(pf_purge_thread);
2649 		pt->seconds = old;
2650 		PF_RULES_WUNLOCK();
2651 		break;
2652 	}
2653 
2654 	case DIOCGETTIMEOUT: {
2655 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2656 
2657 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
2658 			error = EINVAL;
2659 			break;
2660 		}
2661 		PF_RULES_RLOCK();
2662 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
2663 		PF_RULES_RUNLOCK();
2664 		break;
2665 	}
2666 
2667 	case DIOCGETLIMIT: {
2668 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2669 
2670 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
2671 			error = EINVAL;
2672 			break;
2673 		}
2674 		PF_RULES_RLOCK();
2675 		pl->limit = V_pf_limits[pl->index].limit;
2676 		PF_RULES_RUNLOCK();
2677 		break;
2678 	}
2679 
2680 	case DIOCSETLIMIT: {
2681 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2682 		int			 old_limit;
2683 
2684 		PF_RULES_WLOCK();
2685 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
2686 		    V_pf_limits[pl->index].zone == NULL) {
2687 			PF_RULES_WUNLOCK();
2688 			error = EINVAL;
2689 			break;
2690 		}
2691 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
2692 		old_limit = V_pf_limits[pl->index].limit;
2693 		V_pf_limits[pl->index].limit = pl->limit;
2694 		pl->limit = old_limit;
2695 		PF_RULES_WUNLOCK();
2696 		break;
2697 	}
2698 
2699 	case DIOCSETDEBUG: {
2700 		u_int32_t	*level = (u_int32_t *)addr;
2701 
2702 		PF_RULES_WLOCK();
2703 		V_pf_status.debug = *level;
2704 		PF_RULES_WUNLOCK();
2705 		break;
2706 	}
2707 
2708 	case DIOCCLRRULECTRS: {
2709 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
2710 		struct pf_kruleset	*ruleset = &pf_main_ruleset;
2711 		struct pf_krule		*rule;
2712 
2713 		PF_RULES_WLOCK();
2714 		TAILQ_FOREACH(rule,
2715 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
2716 			counter_u64_zero(rule->evaluations);
2717 			for (int i = 0; i < 2; i++) {
2718 				counter_u64_zero(rule->packets[i]);
2719 				counter_u64_zero(rule->bytes[i]);
2720 			}
2721 		}
2722 		PF_RULES_WUNLOCK();
2723 		break;
2724 	}
2725 
2726 	case DIOCGIFSPEEDV0:
2727 	case DIOCGIFSPEEDV1: {
2728 		struct pf_ifspeed_v1	*psp = (struct pf_ifspeed_v1 *)addr;
2729 		struct pf_ifspeed_v1	ps;
2730 		struct ifnet		*ifp;
2731 
2732 		if (psp->ifname[0] != 0) {
2733 			/* Can we completely trust user-land? */
2734 			strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
2735 			ifp = ifunit(ps.ifname);
2736 			if (ifp != NULL) {
2737 				psp->baudrate32 =
2738 				    (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
2739 				if (cmd == DIOCGIFSPEEDV1)
2740 					psp->baudrate = ifp->if_baudrate;
2741 			} else
2742 				error = EINVAL;
2743 		} else
2744 			error = EINVAL;
2745 		break;
2746 	}
2747 
2748 #ifdef ALTQ
2749 	case DIOCSTARTALTQ: {
2750 		struct pf_altq		*altq;
2751 
2752 		PF_RULES_WLOCK();
2753 		/* enable all altq interfaces on active list */
2754 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
2755 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
2756 				error = pf_enable_altq(altq);
2757 				if (error != 0)
2758 					break;
2759 			}
2760 		}
2761 		if (error == 0)
2762 			V_pf_altq_running = 1;
2763 		PF_RULES_WUNLOCK();
2764 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
2765 		break;
2766 	}
2767 
2768 	case DIOCSTOPALTQ: {
2769 		struct pf_altq		*altq;
2770 
2771 		PF_RULES_WLOCK();
2772 		/* disable all altq interfaces on active list */
2773 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
2774 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
2775 				error = pf_disable_altq(altq);
2776 				if (error != 0)
2777 					break;
2778 			}
2779 		}
2780 		if (error == 0)
2781 			V_pf_altq_running = 0;
2782 		PF_RULES_WUNLOCK();
2783 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2784 		break;
2785 	}
2786 
2787 	case DIOCADDALTQV0:
2788 	case DIOCADDALTQV1: {
2789 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2790 		struct pf_altq		*altq, *a;
2791 		struct ifnet		*ifp;
2792 
2793 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
2794 		error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
2795 		if (error)
2796 			break;
2797 		altq->local_flags = 0;
2798 
2799 		PF_RULES_WLOCK();
2800 		if (pa->ticket != V_ticket_altqs_inactive) {
2801 			PF_RULES_WUNLOCK();
2802 			free(altq, M_PFALTQ);
2803 			error = EBUSY;
2804 			break;
2805 		}
2806 
2807 		/*
2808 		 * if this is for a queue, find the discipline and
2809 		 * copy the necessary fields
2810 		 */
2811 		if (altq->qname[0] != 0) {
2812 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2813 				PF_RULES_WUNLOCK();
2814 				error = EBUSY;
2815 				free(altq, M_PFALTQ);
2816 				break;
2817 			}
2818 			altq->altq_disc = NULL;
2819 			TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
2820 				if (strncmp(a->ifname, altq->ifname,
2821 				    IFNAMSIZ) == 0) {
2822 					altq->altq_disc = a->altq_disc;
2823 					break;
2824 				}
2825 			}
2826 		}
2827 
2828 		if ((ifp = ifunit(altq->ifname)) == NULL)
2829 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
2830 		else
2831 			error = altq_add(ifp, altq);
2832 
2833 		if (error) {
2834 			PF_RULES_WUNLOCK();
2835 			free(altq, M_PFALTQ);
2836 			break;
2837 		}
2838 
2839 		if (altq->qname[0] != 0)
2840 			TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
2841 		else
2842 			TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
2843 		/* version error check done on import above */
2844 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
2845 		PF_RULES_WUNLOCK();
2846 		break;
2847 	}
2848 
2849 	case DIOCGETALTQSV0:
2850 	case DIOCGETALTQSV1: {
2851 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2852 		struct pf_altq		*altq;
2853 
2854 		PF_RULES_RLOCK();
2855 		pa->nr = 0;
2856 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
2857 			pa->nr++;
2858 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
2859 			pa->nr++;
2860 		pa->ticket = V_ticket_altqs_active;
2861 		PF_RULES_RUNLOCK();
2862 		break;
2863 	}
2864 
2865 	case DIOCGETALTQV0:
2866 	case DIOCGETALTQV1: {
2867 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2868 		struct pf_altq		*altq;
2869 
2870 		PF_RULES_RLOCK();
2871 		if (pa->ticket != V_ticket_altqs_active) {
2872 			PF_RULES_RUNLOCK();
2873 			error = EBUSY;
2874 			break;
2875 		}
2876 		altq = pf_altq_get_nth_active(pa->nr);
2877 		if (altq == NULL) {
2878 			PF_RULES_RUNLOCK();
2879 			error = EBUSY;
2880 			break;
2881 		}
2882 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
2883 		PF_RULES_RUNLOCK();
2884 		break;
2885 	}
2886 
2887 	case DIOCCHANGEALTQV0:
2888 	case DIOCCHANGEALTQV1:
2889 		/* CHANGEALTQ not supported yet! */
2890 		error = ENODEV;
2891 		break;
2892 
2893 	case DIOCGETQSTATSV0:
2894 	case DIOCGETQSTATSV1: {
2895 		struct pfioc_qstats_v1	*pq = (struct pfioc_qstats_v1 *)addr;
2896 		struct pf_altq		*altq;
2897 		int			 nbytes;
2898 		u_int32_t		 version;
2899 
2900 		PF_RULES_RLOCK();
2901 		if (pq->ticket != V_ticket_altqs_active) {
2902 			PF_RULES_RUNLOCK();
2903 			error = EBUSY;
2904 			break;
2905 		}
2906 		nbytes = pq->nbytes;
2907 		altq = pf_altq_get_nth_active(pq->nr);
2908 		if (altq == NULL) {
2909 			PF_RULES_RUNLOCK();
2910 			error = EBUSY;
2911 			break;
2912 		}
2913 
2914 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
2915 			PF_RULES_RUNLOCK();
2916 			error = ENXIO;
2917 			break;
2918 		}
2919 		PF_RULES_RUNLOCK();
2920 		if (cmd == DIOCGETQSTATSV0)
2921 			version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
2922 		else
2923 			version = pq->version;
2924 		error = altq_getqstats(altq, pq->buf, &nbytes, version);
2925 		if (error == 0) {
2926 			pq->scheduler = altq->scheduler;
2927 			pq->nbytes = nbytes;
2928 		}
2929 		break;
2930 	}
2931 #endif /* ALTQ */
2932 
2933 	case DIOCBEGINADDRS: {
2934 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2935 
2936 		PF_RULES_WLOCK();
2937 		pf_empty_kpool(&V_pf_pabuf);
2938 		pp->ticket = ++V_ticket_pabuf;
2939 		PF_RULES_WUNLOCK();
2940 		break;
2941 	}
2942 
2943 	case DIOCADDADDR: {
2944 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2945 		struct pf_kpooladdr	*pa;
2946 		struct pfi_kkif		*kif = NULL;
2947 
2948 #ifndef INET
2949 		if (pp->af == AF_INET) {
2950 			error = EAFNOSUPPORT;
2951 			break;
2952 		}
2953 #endif /* INET */
2954 #ifndef INET6
2955 		if (pp->af == AF_INET6) {
2956 			error = EAFNOSUPPORT;
2957 			break;
2958 		}
2959 #endif /* INET6 */
2960 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2961 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2962 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2963 			error = EINVAL;
2964 			break;
2965 		}
2966 		if (pp->addr.addr.p.dyn != NULL) {
2967 			error = EINVAL;
2968 			break;
2969 		}
2970 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2971 		pf_pooladdr_to_kpooladdr(&pp->addr, pa);
2972 		if (pa->ifname[0])
2973 			kif = pf_kkif_create(M_WAITOK);
2974 		PF_RULES_WLOCK();
2975 		if (pp->ticket != V_ticket_pabuf) {
2976 			PF_RULES_WUNLOCK();
2977 			if (pa->ifname[0])
2978 				pf_kkif_free(kif);
2979 			free(pa, M_PFRULE);
2980 			error = EBUSY;
2981 			break;
2982 		}
2983 		if (pa->ifname[0]) {
2984 			pa->kif = pfi_kkif_attach(kif, pa->ifname);
2985 			pfi_kkif_ref(pa->kif);
2986 		} else
2987 			pa->kif = NULL;
2988 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2989 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2990 			if (pa->ifname[0])
2991 				pfi_kkif_unref(pa->kif);
2992 			PF_RULES_WUNLOCK();
2993 			free(pa, M_PFRULE);
2994 			break;
2995 		}
2996 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2997 		PF_RULES_WUNLOCK();
2998 		break;
2999 	}
3000 
3001 	case DIOCGETADDRS: {
3002 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
3003 		struct pf_kpool		*pool;
3004 		struct pf_kpooladdr	*pa;
3005 
3006 		PF_RULES_RLOCK();
3007 		pp->nr = 0;
3008 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
3009 		    pp->r_num, 0, 1, 0);
3010 		if (pool == NULL) {
3011 			PF_RULES_RUNLOCK();
3012 			error = EBUSY;
3013 			break;
3014 		}
3015 		TAILQ_FOREACH(pa, &pool->list, entries)
3016 			pp->nr++;
3017 		PF_RULES_RUNLOCK();
3018 		break;
3019 	}
3020 
3021 	case DIOCGETADDR: {
3022 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
3023 		struct pf_kpool		*pool;
3024 		struct pf_kpooladdr	*pa;
3025 		u_int32_t		 nr = 0;
3026 
3027 		PF_RULES_RLOCK();
3028 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
3029 		    pp->r_num, 0, 1, 1);
3030 		if (pool == NULL) {
3031 			PF_RULES_RUNLOCK();
3032 			error = EBUSY;
3033 			break;
3034 		}
3035 		pa = TAILQ_FIRST(&pool->list);
3036 		while ((pa != NULL) && (nr < pp->nr)) {
3037 			pa = TAILQ_NEXT(pa, entries);
3038 			nr++;
3039 		}
3040 		if (pa == NULL) {
3041 			PF_RULES_RUNLOCK();
3042 			error = EBUSY;
3043 			break;
3044 		}
3045 		pf_kpooladdr_to_pooladdr(pa, &pp->addr);
3046 		pf_addr_copyout(&pp->addr.addr);
3047 		PF_RULES_RUNLOCK();
3048 		break;
3049 	}
3050 
3051 	case DIOCCHANGEADDR: {
3052 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
3053 		struct pf_kpool		*pool;
3054 		struct pf_kpooladdr	*oldpa = NULL, *newpa = NULL;
3055 		struct pf_kruleset	*ruleset;
3056 		struct pfi_kkif		*kif = NULL;
3057 
3058 		if (pca->action < PF_CHANGE_ADD_HEAD ||
3059 		    pca->action > PF_CHANGE_REMOVE) {
3060 			error = EINVAL;
3061 			break;
3062 		}
3063 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
3064 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
3065 		    pca->addr.addr.type != PF_ADDR_TABLE) {
3066 			error = EINVAL;
3067 			break;
3068 		}
3069 		if (pca->addr.addr.p.dyn != NULL) {
3070 			error = EINVAL;
3071 			break;
3072 		}
3073 
3074 		if (pca->action != PF_CHANGE_REMOVE) {
3075 #ifndef INET
3076 			if (pca->af == AF_INET) {
3077 				error = EAFNOSUPPORT;
3078 				break;
3079 			}
3080 #endif /* INET */
3081 #ifndef INET6
3082 			if (pca->af == AF_INET6) {
3083 				error = EAFNOSUPPORT;
3084 				break;
3085 			}
3086 #endif /* INET6 */
3087 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
3088 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
3089 			if (newpa->ifname[0])
3090 				kif = pf_kkif_create(M_WAITOK);
3091 			newpa->kif = NULL;
3092 		}
3093 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGEADDR_error; }
3094 		PF_RULES_WLOCK();
3095 		ruleset = pf_find_kruleset(pca->anchor);
3096 		if (ruleset == NULL)
3097 			ERROUT(EBUSY);
3098 
3099 		pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
3100 		    pca->r_num, pca->r_last, 1, 1);
3101 		if (pool == NULL)
3102 			ERROUT(EBUSY);
3103 
3104 		if (pca->action != PF_CHANGE_REMOVE) {
3105 			if (newpa->ifname[0]) {
3106 				newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
3107 				pfi_kkif_ref(newpa->kif);
3108 				kif = NULL;
3109 			}
3110 
3111 			switch (newpa->addr.type) {
3112 			case PF_ADDR_DYNIFTL:
3113 				error = pfi_dynaddr_setup(&newpa->addr,
3114 				    pca->af);
3115 				break;
3116 			case PF_ADDR_TABLE:
3117 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
3118 				    newpa->addr.v.tblname);
3119 				if (newpa->addr.p.tbl == NULL)
3120 					error = ENOMEM;
3121 				break;
3122 			}
3123 			if (error)
3124 				goto DIOCCHANGEADDR_error;
3125 		}
3126 
3127 		switch (pca->action) {
3128 		case PF_CHANGE_ADD_HEAD:
3129 			oldpa = TAILQ_FIRST(&pool->list);
3130 			break;
3131 		case PF_CHANGE_ADD_TAIL:
3132 			oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
3133 			break;
3134 		default:
3135 			oldpa = TAILQ_FIRST(&pool->list);
3136 			for (int i = 0; oldpa && i < pca->nr; i++)
3137 				oldpa = TAILQ_NEXT(oldpa, entries);
3138 
3139 			if (oldpa == NULL)
3140 				ERROUT(EINVAL);
3141 		}
3142 
3143 		if (pca->action == PF_CHANGE_REMOVE) {
3144 			TAILQ_REMOVE(&pool->list, oldpa, entries);
3145 			switch (oldpa->addr.type) {
3146 			case PF_ADDR_DYNIFTL:
3147 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
3148 				break;
3149 			case PF_ADDR_TABLE:
3150 				pfr_detach_table(oldpa->addr.p.tbl);
3151 				break;
3152 			}
3153 			if (oldpa->kif)
3154 				pfi_kkif_unref(oldpa->kif);
3155 			free(oldpa, M_PFRULE);
3156 		} else {
3157 			if (oldpa == NULL)
3158 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
3159 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
3160 			    pca->action == PF_CHANGE_ADD_BEFORE)
3161 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
3162 			else
3163 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
3164 				    newpa, entries);
3165 		}
3166 
3167 		pool->cur = TAILQ_FIRST(&pool->list);
3168 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
3169 		PF_RULES_WUNLOCK();
3170 		break;
3171 
3172 #undef ERROUT
3173 DIOCCHANGEADDR_error:
3174 		if (newpa != NULL) {
3175 			if (newpa->kif)
3176 				pfi_kkif_unref(newpa->kif);
3177 			free(newpa, M_PFRULE);
3178 		}
3179 		PF_RULES_WUNLOCK();
3180 		pf_kkif_free(kif);
3181 		break;
3182 	}
3183 
3184 	case DIOCGETRULESETS: {
3185 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
3186 		struct pf_kruleset	*ruleset;
3187 		struct pf_kanchor	*anchor;
3188 
3189 		PF_RULES_RLOCK();
3190 		pr->path[sizeof(pr->path) - 1] = 0;
3191 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
3192 			PF_RULES_RUNLOCK();
3193 			error = ENOENT;
3194 			break;
3195 		}
3196 		pr->nr = 0;
3197 		if (ruleset->anchor == NULL) {
3198 			/* XXX kludge for pf_main_ruleset */
3199 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
3200 				if (anchor->parent == NULL)
3201 					pr->nr++;
3202 		} else {
3203 			RB_FOREACH(anchor, pf_kanchor_node,
3204 			    &ruleset->anchor->children)
3205 				pr->nr++;
3206 		}
3207 		PF_RULES_RUNLOCK();
3208 		break;
3209 	}
3210 
3211 	case DIOCGETRULESET: {
3212 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
3213 		struct pf_kruleset	*ruleset;
3214 		struct pf_kanchor	*anchor;
3215 		u_int32_t		 nr = 0;
3216 
3217 		PF_RULES_RLOCK();
3218 		pr->path[sizeof(pr->path) - 1] = 0;
3219 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
3220 			PF_RULES_RUNLOCK();
3221 			error = ENOENT;
3222 			break;
3223 		}
3224 		pr->name[0] = 0;
3225 		if (ruleset->anchor == NULL) {
3226 			/* XXX kludge for pf_main_ruleset */
3227 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
3228 				if (anchor->parent == NULL && nr++ == pr->nr) {
3229 					strlcpy(pr->name, anchor->name,
3230 					    sizeof(pr->name));
3231 					break;
3232 				}
3233 		} else {
3234 			RB_FOREACH(anchor, pf_kanchor_node,
3235 			    &ruleset->anchor->children)
3236 				if (nr++ == pr->nr) {
3237 					strlcpy(pr->name, anchor->name,
3238 					    sizeof(pr->name));
3239 					break;
3240 				}
3241 		}
3242 		if (!pr->name[0])
3243 			error = EBUSY;
3244 		PF_RULES_RUNLOCK();
3245 		break;
3246 	}
3247 
3248 	case DIOCRCLRTABLES: {
3249 		struct pfioc_table *io = (struct pfioc_table *)addr;
3250 
3251 		if (io->pfrio_esize != 0) {
3252 			error = ENODEV;
3253 			break;
3254 		}
3255 		PF_RULES_WLOCK();
3256 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
3257 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
3258 		PF_RULES_WUNLOCK();
3259 		break;
3260 	}
3261 
3262 	case DIOCRADDTABLES: {
3263 		struct pfioc_table *io = (struct pfioc_table *)addr;
3264 		struct pfr_table *pfrts;
3265 		size_t totlen;
3266 
3267 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3268 			error = ENODEV;
3269 			break;
3270 		}
3271 
3272 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
3273 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
3274 			error = ENOMEM;
3275 			break;
3276 		}
3277 
3278 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3279 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3280 		    M_TEMP, M_WAITOK);
3281 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3282 		if (error) {
3283 			free(pfrts, M_TEMP);
3284 			break;
3285 		}
3286 		PF_RULES_WLOCK();
3287 		error = pfr_add_tables(pfrts, io->pfrio_size,
3288 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3289 		PF_RULES_WUNLOCK();
3290 		free(pfrts, M_TEMP);
3291 		break;
3292 	}
3293 
3294 	case DIOCRDELTABLES: {
3295 		struct pfioc_table *io = (struct pfioc_table *)addr;
3296 		struct pfr_table *pfrts;
3297 		size_t totlen;
3298 
3299 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3300 			error = ENODEV;
3301 			break;
3302 		}
3303 
3304 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
3305 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
3306 			error = ENOMEM;
3307 			break;
3308 		}
3309 
3310 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3311 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3312 		    M_TEMP, M_WAITOK);
3313 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3314 		if (error) {
3315 			free(pfrts, M_TEMP);
3316 			break;
3317 		}
3318 		PF_RULES_WLOCK();
3319 		error = pfr_del_tables(pfrts, io->pfrio_size,
3320 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3321 		PF_RULES_WUNLOCK();
3322 		free(pfrts, M_TEMP);
3323 		break;
3324 	}
3325 
3326 	case DIOCRGETTABLES: {
3327 		struct pfioc_table *io = (struct pfioc_table *)addr;
3328 		struct pfr_table *pfrts;
3329 		size_t totlen;
3330 		int n;
3331 
3332 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3333 			error = ENODEV;
3334 			break;
3335 		}
3336 		PF_RULES_RLOCK();
3337 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3338 		if (n < 0) {
3339 			PF_RULES_RUNLOCK();
3340 			error = EINVAL;
3341 			break;
3342 		}
3343 		io->pfrio_size = min(io->pfrio_size, n);
3344 
3345 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3346 
3347 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3348 		    M_TEMP, M_NOWAIT);
3349 		if (pfrts == NULL) {
3350 			error = ENOMEM;
3351 			PF_RULES_RUNLOCK();
3352 			break;
3353 		}
3354 		error = pfr_get_tables(&io->pfrio_table, pfrts,
3355 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3356 		PF_RULES_RUNLOCK();
3357 		if (error == 0)
3358 			error = copyout(pfrts, io->pfrio_buffer, totlen);
3359 		free(pfrts, M_TEMP);
3360 		break;
3361 	}
3362 
3363 	case DIOCRGETTSTATS: {
3364 		struct pfioc_table *io = (struct pfioc_table *)addr;
3365 		struct pfr_tstats *pfrtstats;
3366 		size_t totlen;
3367 		int n;
3368 
3369 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
3370 			error = ENODEV;
3371 			break;
3372 		}
3373 		PF_RULES_WLOCK();
3374 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3375 		if (n < 0) {
3376 			PF_RULES_WUNLOCK();
3377 			error = EINVAL;
3378 			break;
3379 		}
3380 		io->pfrio_size = min(io->pfrio_size, n);
3381 
3382 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
3383 		pfrtstats = mallocarray(io->pfrio_size,
3384 		    sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT);
3385 		if (pfrtstats == NULL) {
3386 			error = ENOMEM;
3387 			PF_RULES_WUNLOCK();
3388 			break;
3389 		}
3390 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
3391 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3392 		PF_RULES_WUNLOCK();
3393 		if (error == 0)
3394 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
3395 		free(pfrtstats, M_TEMP);
3396 		break;
3397 	}
3398 
3399 	case DIOCRCLRTSTATS: {
3400 		struct pfioc_table *io = (struct pfioc_table *)addr;
3401 		struct pfr_table *pfrts;
3402 		size_t totlen;
3403 
3404 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3405 			error = ENODEV;
3406 			break;
3407 		}
3408 
3409 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
3410 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
3411 			/* We used to count tables and use the minimum required
3412 			 * size, so we didn't fail on overly large requests.
3413 			 * Keep doing so. */
3414 			io->pfrio_size = pf_ioctl_maxcount;
3415 			break;
3416 		}
3417 
3418 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3419 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3420 		    M_TEMP, M_NOWAIT);
3421 		if (pfrts == NULL) {
3422 			error = ENOMEM;
3423 			break;
3424 		}
3425 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3426 		if (error) {
3427 			free(pfrts, M_TEMP);
3428 			break;
3429 		}
3430 
3431 		PF_RULES_WLOCK();
3432 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
3433 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3434 		PF_RULES_WUNLOCK();
3435 		free(pfrts, M_TEMP);
3436 		break;
3437 	}
3438 
3439 	case DIOCRSETTFLAGS: {
3440 		struct pfioc_table *io = (struct pfioc_table *)addr;
3441 		struct pfr_table *pfrts;
3442 		size_t totlen;
3443 		int n;
3444 
3445 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3446 			error = ENODEV;
3447 			break;
3448 		}
3449 
3450 		PF_RULES_RLOCK();
3451 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3452 		if (n < 0) {
3453 			PF_RULES_RUNLOCK();
3454 			error = EINVAL;
3455 			break;
3456 		}
3457 
3458 		io->pfrio_size = min(io->pfrio_size, n);
3459 		PF_RULES_RUNLOCK();
3460 
3461 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3462 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3463 		    M_TEMP, M_WAITOK);
3464 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3465 		if (error) {
3466 			free(pfrts, M_TEMP);
3467 			break;
3468 		}
3469 		PF_RULES_WLOCK();
3470 		error = pfr_set_tflags(pfrts, io->pfrio_size,
3471 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
3472 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3473 		PF_RULES_WUNLOCK();
3474 		free(pfrts, M_TEMP);
3475 		break;
3476 	}
3477 
3478 	case DIOCRCLRADDRS: {
3479 		struct pfioc_table *io = (struct pfioc_table *)addr;
3480 
3481 		if (io->pfrio_esize != 0) {
3482 			error = ENODEV;
3483 			break;
3484 		}
3485 		PF_RULES_WLOCK();
3486 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
3487 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
3488 		PF_RULES_WUNLOCK();
3489 		break;
3490 	}
3491 
3492 	case DIOCRADDADDRS: {
3493 		struct pfioc_table *io = (struct pfioc_table *)addr;
3494 		struct pfr_addr *pfras;
3495 		size_t totlen;
3496 
3497 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3498 			error = ENODEV;
3499 			break;
3500 		}
3501 		if (io->pfrio_size < 0 ||
3502 		    io->pfrio_size > pf_ioctl_maxcount ||
3503 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3504 			error = EINVAL;
3505 			break;
3506 		}
3507 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3508 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3509 		    M_TEMP, M_NOWAIT);
3510 		if (! pfras) {
3511 			error = ENOMEM;
3512 			break;
3513 		}
3514 		error = copyin(io->pfrio_buffer, pfras, totlen);
3515 		if (error) {
3516 			free(pfras, M_TEMP);
3517 			break;
3518 		}
3519 		PF_RULES_WLOCK();
3520 		error = pfr_add_addrs(&io->pfrio_table, pfras,
3521 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
3522 		    PFR_FLAG_USERIOCTL);
3523 		PF_RULES_WUNLOCK();
3524 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3525 			error = copyout(pfras, io->pfrio_buffer, totlen);
3526 		free(pfras, M_TEMP);
3527 		break;
3528 	}
3529 
3530 	case DIOCRDELADDRS: {
3531 		struct pfioc_table *io = (struct pfioc_table *)addr;
3532 		struct pfr_addr *pfras;
3533 		size_t totlen;
3534 
3535 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3536 			error = ENODEV;
3537 			break;
3538 		}
3539 		if (io->pfrio_size < 0 ||
3540 		    io->pfrio_size > pf_ioctl_maxcount ||
3541 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3542 			error = EINVAL;
3543 			break;
3544 		}
3545 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3546 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3547 		    M_TEMP, M_NOWAIT);
3548 		if (! pfras) {
3549 			error = ENOMEM;
3550 			break;
3551 		}
3552 		error = copyin(io->pfrio_buffer, pfras, totlen);
3553 		if (error) {
3554 			free(pfras, M_TEMP);
3555 			break;
3556 		}
3557 		PF_RULES_WLOCK();
3558 		error = pfr_del_addrs(&io->pfrio_table, pfras,
3559 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
3560 		    PFR_FLAG_USERIOCTL);
3561 		PF_RULES_WUNLOCK();
3562 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3563 			error = copyout(pfras, io->pfrio_buffer, totlen);
3564 		free(pfras, M_TEMP);
3565 		break;
3566 	}
3567 
3568 	case DIOCRSETADDRS: {
3569 		struct pfioc_table *io = (struct pfioc_table *)addr;
3570 		struct pfr_addr *pfras;
3571 		size_t totlen, count;
3572 
3573 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3574 			error = ENODEV;
3575 			break;
3576 		}
3577 		if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
3578 			error = EINVAL;
3579 			break;
3580 		}
3581 		count = max(io->pfrio_size, io->pfrio_size2);
3582 		if (count > pf_ioctl_maxcount ||
3583 		    WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
3584 			error = EINVAL;
3585 			break;
3586 		}
3587 		totlen = count * sizeof(struct pfr_addr);
3588 		pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
3589 		    M_NOWAIT);
3590 		if (! pfras) {
3591 			error = ENOMEM;
3592 			break;
3593 		}
3594 		error = copyin(io->pfrio_buffer, pfras, totlen);
3595 		if (error) {
3596 			free(pfras, M_TEMP);
3597 			break;
3598 		}
3599 		PF_RULES_WLOCK();
3600 		error = pfr_set_addrs(&io->pfrio_table, pfras,
3601 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
3602 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
3603 		    PFR_FLAG_USERIOCTL, 0);
3604 		PF_RULES_WUNLOCK();
3605 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3606 			error = copyout(pfras, io->pfrio_buffer, totlen);
3607 		free(pfras, M_TEMP);
3608 		break;
3609 	}
3610 
3611 	case DIOCRGETADDRS: {
3612 		struct pfioc_table *io = (struct pfioc_table *)addr;
3613 		struct pfr_addr *pfras;
3614 		size_t totlen;
3615 
3616 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3617 			error = ENODEV;
3618 			break;
3619 		}
3620 		if (io->pfrio_size < 0 ||
3621 		    io->pfrio_size > pf_ioctl_maxcount ||
3622 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3623 			error = EINVAL;
3624 			break;
3625 		}
3626 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3627 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3628 		    M_TEMP, M_NOWAIT);
3629 		if (! pfras) {
3630 			error = ENOMEM;
3631 			break;
3632 		}
3633 		PF_RULES_RLOCK();
3634 		error = pfr_get_addrs(&io->pfrio_table, pfras,
3635 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3636 		PF_RULES_RUNLOCK();
3637 		if (error == 0)
3638 			error = copyout(pfras, io->pfrio_buffer, totlen);
3639 		free(pfras, M_TEMP);
3640 		break;
3641 	}
3642 
3643 	case DIOCRGETASTATS: {
3644 		struct pfioc_table *io = (struct pfioc_table *)addr;
3645 		struct pfr_astats *pfrastats;
3646 		size_t totlen;
3647 
3648 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
3649 			error = ENODEV;
3650 			break;
3651 		}
3652 		if (io->pfrio_size < 0 ||
3653 		    io->pfrio_size > pf_ioctl_maxcount ||
3654 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
3655 			error = EINVAL;
3656 			break;
3657 		}
3658 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
3659 		pfrastats = mallocarray(io->pfrio_size,
3660 		    sizeof(struct pfr_astats), M_TEMP, M_NOWAIT);
3661 		if (! pfrastats) {
3662 			error = ENOMEM;
3663 			break;
3664 		}
3665 		PF_RULES_RLOCK();
3666 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
3667 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3668 		PF_RULES_RUNLOCK();
3669 		if (error == 0)
3670 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
3671 		free(pfrastats, M_TEMP);
3672 		break;
3673 	}
3674 
3675 	case DIOCRCLRASTATS: {
3676 		struct pfioc_table *io = (struct pfioc_table *)addr;
3677 		struct pfr_addr *pfras;
3678 		size_t totlen;
3679 
3680 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3681 			error = ENODEV;
3682 			break;
3683 		}
3684 		if (io->pfrio_size < 0 ||
3685 		    io->pfrio_size > pf_ioctl_maxcount ||
3686 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3687 			error = EINVAL;
3688 			break;
3689 		}
3690 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3691 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3692 		    M_TEMP, M_NOWAIT);
3693 		if (! pfras) {
3694 			error = ENOMEM;
3695 			break;
3696 		}
3697 		error = copyin(io->pfrio_buffer, pfras, totlen);
3698 		if (error) {
3699 			free(pfras, M_TEMP);
3700 			break;
3701 		}
3702 		PF_RULES_WLOCK();
3703 		error = pfr_clr_astats(&io->pfrio_table, pfras,
3704 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
3705 		    PFR_FLAG_USERIOCTL);
3706 		PF_RULES_WUNLOCK();
3707 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3708 			error = copyout(pfras, io->pfrio_buffer, totlen);
3709 		free(pfras, M_TEMP);
3710 		break;
3711 	}
3712 
3713 	case DIOCRTSTADDRS: {
3714 		struct pfioc_table *io = (struct pfioc_table *)addr;
3715 		struct pfr_addr *pfras;
3716 		size_t totlen;
3717 
3718 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3719 			error = ENODEV;
3720 			break;
3721 		}
3722 		if (io->pfrio_size < 0 ||
3723 		    io->pfrio_size > pf_ioctl_maxcount ||
3724 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3725 			error = EINVAL;
3726 			break;
3727 		}
3728 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3729 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3730 		    M_TEMP, M_NOWAIT);
3731 		if (! pfras) {
3732 			error = ENOMEM;
3733 			break;
3734 		}
3735 		error = copyin(io->pfrio_buffer, pfras, totlen);
3736 		if (error) {
3737 			free(pfras, M_TEMP);
3738 			break;
3739 		}
3740 		PF_RULES_RLOCK();
3741 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
3742 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
3743 		    PFR_FLAG_USERIOCTL);
3744 		PF_RULES_RUNLOCK();
3745 		if (error == 0)
3746 			error = copyout(pfras, io->pfrio_buffer, totlen);
3747 		free(pfras, M_TEMP);
3748 		break;
3749 	}
3750 
3751 	case DIOCRINADEFINE: {
3752 		struct pfioc_table *io = (struct pfioc_table *)addr;
3753 		struct pfr_addr *pfras;
3754 		size_t totlen;
3755 
3756 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3757 			error = ENODEV;
3758 			break;
3759 		}
3760 		if (io->pfrio_size < 0 ||
3761 		    io->pfrio_size > pf_ioctl_maxcount ||
3762 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3763 			error = EINVAL;
3764 			break;
3765 		}
3766 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3767 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3768 		    M_TEMP, M_NOWAIT);
3769 		if (! pfras) {
3770 			error = ENOMEM;
3771 			break;
3772 		}
3773 		error = copyin(io->pfrio_buffer, pfras, totlen);
3774 		if (error) {
3775 			free(pfras, M_TEMP);
3776 			break;
3777 		}
3778 		PF_RULES_WLOCK();
3779 		error = pfr_ina_define(&io->pfrio_table, pfras,
3780 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
3781 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3782 		PF_RULES_WUNLOCK();
3783 		free(pfras, M_TEMP);
3784 		break;
3785 	}
3786 
3787 	case DIOCOSFPADD: {
3788 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
3789 		PF_RULES_WLOCK();
3790 		error = pf_osfp_add(io);
3791 		PF_RULES_WUNLOCK();
3792 		break;
3793 	}
3794 
3795 	case DIOCOSFPGET: {
3796 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
3797 		PF_RULES_RLOCK();
3798 		error = pf_osfp_get(io);
3799 		PF_RULES_RUNLOCK();
3800 		break;
3801 	}
3802 
3803 	case DIOCXBEGIN: {
3804 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3805 		struct pfioc_trans_e	*ioes, *ioe;
3806 		size_t			 totlen;
3807 		int			 i;
3808 
3809 		if (io->esize != sizeof(*ioe)) {
3810 			error = ENODEV;
3811 			break;
3812 		}
3813 		if (io->size < 0 ||
3814 		    io->size > pf_ioctl_maxcount ||
3815 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3816 			error = EINVAL;
3817 			break;
3818 		}
3819 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3820 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3821 		    M_TEMP, M_NOWAIT);
3822 		if (! ioes) {
3823 			error = ENOMEM;
3824 			break;
3825 		}
3826 		error = copyin(io->array, ioes, totlen);
3827 		if (error) {
3828 			free(ioes, M_TEMP);
3829 			break;
3830 		}
3831 		PF_RULES_WLOCK();
3832 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3833 			switch (ioe->rs_num) {
3834 #ifdef ALTQ
3835 			case PF_RULESET_ALTQ:
3836 				if (ioe->anchor[0]) {
3837 					PF_RULES_WUNLOCK();
3838 					free(ioes, M_TEMP);
3839 					error = EINVAL;
3840 					goto fail;
3841 				}
3842 				if ((error = pf_begin_altq(&ioe->ticket))) {
3843 					PF_RULES_WUNLOCK();
3844 					free(ioes, M_TEMP);
3845 					goto fail;
3846 				}
3847 				break;
3848 #endif /* ALTQ */
3849 			case PF_RULESET_TABLE:
3850 			    {
3851 				struct pfr_table table;
3852 
3853 				bzero(&table, sizeof(table));
3854 				strlcpy(table.pfrt_anchor, ioe->anchor,
3855 				    sizeof(table.pfrt_anchor));
3856 				if ((error = pfr_ina_begin(&table,
3857 				    &ioe->ticket, NULL, 0))) {
3858 					PF_RULES_WUNLOCK();
3859 					free(ioes, M_TEMP);
3860 					goto fail;
3861 				}
3862 				break;
3863 			    }
3864 			default:
3865 				if ((error = pf_begin_rules(&ioe->ticket,
3866 				    ioe->rs_num, ioe->anchor))) {
3867 					PF_RULES_WUNLOCK();
3868 					free(ioes, M_TEMP);
3869 					goto fail;
3870 				}
3871 				break;
3872 			}
3873 		}
3874 		PF_RULES_WUNLOCK();
3875 		error = copyout(ioes, io->array, totlen);
3876 		free(ioes, M_TEMP);
3877 		break;
3878 	}
3879 
3880 	case DIOCXROLLBACK: {
3881 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3882 		struct pfioc_trans_e	*ioe, *ioes;
3883 		size_t			 totlen;
3884 		int			 i;
3885 
3886 		if (io->esize != sizeof(*ioe)) {
3887 			error = ENODEV;
3888 			break;
3889 		}
3890 		if (io->size < 0 ||
3891 		    io->size > pf_ioctl_maxcount ||
3892 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3893 			error = EINVAL;
3894 			break;
3895 		}
3896 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3897 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3898 		    M_TEMP, M_NOWAIT);
3899 		if (! ioes) {
3900 			error = ENOMEM;
3901 			break;
3902 		}
3903 		error = copyin(io->array, ioes, totlen);
3904 		if (error) {
3905 			free(ioes, M_TEMP);
3906 			break;
3907 		}
3908 		PF_RULES_WLOCK();
3909 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3910 			switch (ioe->rs_num) {
3911 #ifdef ALTQ
3912 			case PF_RULESET_ALTQ:
3913 				if (ioe->anchor[0]) {
3914 					PF_RULES_WUNLOCK();
3915 					free(ioes, M_TEMP);
3916 					error = EINVAL;
3917 					goto fail;
3918 				}
3919 				if ((error = pf_rollback_altq(ioe->ticket))) {
3920 					PF_RULES_WUNLOCK();
3921 					free(ioes, M_TEMP);
3922 					goto fail; /* really bad */
3923 				}
3924 				break;
3925 #endif /* ALTQ */
3926 			case PF_RULESET_TABLE:
3927 			    {
3928 				struct pfr_table table;
3929 
3930 				bzero(&table, sizeof(table));
3931 				strlcpy(table.pfrt_anchor, ioe->anchor,
3932 				    sizeof(table.pfrt_anchor));
3933 				if ((error = pfr_ina_rollback(&table,
3934 				    ioe->ticket, NULL, 0))) {
3935 					PF_RULES_WUNLOCK();
3936 					free(ioes, M_TEMP);
3937 					goto fail; /* really bad */
3938 				}
3939 				break;
3940 			    }
3941 			default:
3942 				if ((error = pf_rollback_rules(ioe->ticket,
3943 				    ioe->rs_num, ioe->anchor))) {
3944 					PF_RULES_WUNLOCK();
3945 					free(ioes, M_TEMP);
3946 					goto fail; /* really bad */
3947 				}
3948 				break;
3949 			}
3950 		}
3951 		PF_RULES_WUNLOCK();
3952 		free(ioes, M_TEMP);
3953 		break;
3954 	}
3955 
3956 	case DIOCXCOMMIT: {
3957 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3958 		struct pfioc_trans_e	*ioe, *ioes;
3959 		struct pf_kruleset	*rs;
3960 		size_t			 totlen;
3961 		int			 i;
3962 
3963 		if (io->esize != sizeof(*ioe)) {
3964 			error = ENODEV;
3965 			break;
3966 		}
3967 
3968 		if (io->size < 0 ||
3969 		    io->size > pf_ioctl_maxcount ||
3970 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3971 			error = EINVAL;
3972 			break;
3973 		}
3974 
3975 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3976 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3977 		    M_TEMP, M_NOWAIT);
3978 		if (ioes == NULL) {
3979 			error = ENOMEM;
3980 			break;
3981 		}
3982 		error = copyin(io->array, ioes, totlen);
3983 		if (error) {
3984 			free(ioes, M_TEMP);
3985 			break;
3986 		}
3987 		PF_RULES_WLOCK();
3988 		/* First makes sure everything will succeed. */
3989 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3990 			switch (ioe->rs_num) {
3991 #ifdef ALTQ
3992 			case PF_RULESET_ALTQ:
3993 				if (ioe->anchor[0]) {
3994 					PF_RULES_WUNLOCK();
3995 					free(ioes, M_TEMP);
3996 					error = EINVAL;
3997 					goto fail;
3998 				}
3999 				if (!V_altqs_inactive_open || ioe->ticket !=
4000 				    V_ticket_altqs_inactive) {
4001 					PF_RULES_WUNLOCK();
4002 					free(ioes, M_TEMP);
4003 					error = EBUSY;
4004 					goto fail;
4005 				}
4006 				break;
4007 #endif /* ALTQ */
4008 			case PF_RULESET_TABLE:
4009 				rs = pf_find_kruleset(ioe->anchor);
4010 				if (rs == NULL || !rs->topen || ioe->ticket !=
4011 				    rs->tticket) {
4012 					PF_RULES_WUNLOCK();
4013 					free(ioes, M_TEMP);
4014 					error = EBUSY;
4015 					goto fail;
4016 				}
4017 				break;
4018 			default:
4019 				if (ioe->rs_num < 0 || ioe->rs_num >=
4020 				    PF_RULESET_MAX) {
4021 					PF_RULES_WUNLOCK();
4022 					free(ioes, M_TEMP);
4023 					error = EINVAL;
4024 					goto fail;
4025 				}
4026 				rs = pf_find_kruleset(ioe->anchor);
4027 				if (rs == NULL ||
4028 				    !rs->rules[ioe->rs_num].inactive.open ||
4029 				    rs->rules[ioe->rs_num].inactive.ticket !=
4030 				    ioe->ticket) {
4031 					PF_RULES_WUNLOCK();
4032 					free(ioes, M_TEMP);
4033 					error = EBUSY;
4034 					goto fail;
4035 				}
4036 				break;
4037 			}
4038 		}
4039 		/* Now do the commit - no errors should happen here. */
4040 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
4041 			switch (ioe->rs_num) {
4042 #ifdef ALTQ
4043 			case PF_RULESET_ALTQ:
4044 				if ((error = pf_commit_altq(ioe->ticket))) {
4045 					PF_RULES_WUNLOCK();
4046 					free(ioes, M_TEMP);
4047 					goto fail; /* really bad */
4048 				}
4049 				break;
4050 #endif /* ALTQ */
4051 			case PF_RULESET_TABLE:
4052 			    {
4053 				struct pfr_table table;
4054 
4055 				bzero(&table, sizeof(table));
4056 				strlcpy(table.pfrt_anchor, ioe->anchor,
4057 				    sizeof(table.pfrt_anchor));
4058 				if ((error = pfr_ina_commit(&table,
4059 				    ioe->ticket, NULL, NULL, 0))) {
4060 					PF_RULES_WUNLOCK();
4061 					free(ioes, M_TEMP);
4062 					goto fail; /* really bad */
4063 				}
4064 				break;
4065 			    }
4066 			default:
4067 				if ((error = pf_commit_rules(ioe->ticket,
4068 				    ioe->rs_num, ioe->anchor))) {
4069 					PF_RULES_WUNLOCK();
4070 					free(ioes, M_TEMP);
4071 					goto fail; /* really bad */
4072 				}
4073 				break;
4074 			}
4075 		}
4076 		PF_RULES_WUNLOCK();
4077 		free(ioes, M_TEMP);
4078 		break;
4079 	}
4080 
4081 	case DIOCGETSRCNODES: {
4082 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
4083 		struct pf_srchash	*sh;
4084 		struct pf_ksrc_node	*n;
4085 		struct pf_src_node	*p, *pstore;
4086 		uint32_t		 i, nr = 0;
4087 
4088 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4089 				i++, sh++) {
4090 			PF_HASHROW_LOCK(sh);
4091 			LIST_FOREACH(n, &sh->nodes, entry)
4092 				nr++;
4093 			PF_HASHROW_UNLOCK(sh);
4094 		}
4095 
4096 		psn->psn_len = min(psn->psn_len,
4097 		    sizeof(struct pf_src_node) * nr);
4098 
4099 		if (psn->psn_len == 0) {
4100 			psn->psn_len = sizeof(struct pf_src_node) * nr;
4101 			break;
4102 		}
4103 
4104 		nr = 0;
4105 
4106 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
4107 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4108 		    i++, sh++) {
4109 		    PF_HASHROW_LOCK(sh);
4110 		    LIST_FOREACH(n, &sh->nodes, entry) {
4111 
4112 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
4113 				break;
4114 
4115 			pf_src_node_copy(n, p);
4116 
4117 			p++;
4118 			nr++;
4119 		    }
4120 		    PF_HASHROW_UNLOCK(sh);
4121 		}
4122 		error = copyout(pstore, psn->psn_src_nodes,
4123 		    sizeof(struct pf_src_node) * nr);
4124 		if (error) {
4125 			free(pstore, M_TEMP);
4126 			break;
4127 		}
4128 		psn->psn_len = sizeof(struct pf_src_node) * nr;
4129 		free(pstore, M_TEMP);
4130 		break;
4131 	}
4132 
4133 	case DIOCCLRSRCNODES: {
4134 		pf_clear_srcnodes(NULL);
4135 		pf_purge_expired_src_nodes();
4136 		break;
4137 	}
4138 
4139 	case DIOCKILLSRCNODES:
4140 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
4141 		break;
4142 
4143 	case DIOCSETHOSTID: {
4144 		u_int32_t	*hostid = (u_int32_t *)addr;
4145 
4146 		PF_RULES_WLOCK();
4147 		if (*hostid == 0)
4148 			V_pf_status.hostid = arc4random();
4149 		else
4150 			V_pf_status.hostid = *hostid;
4151 		PF_RULES_WUNLOCK();
4152 		break;
4153 	}
4154 
4155 	case DIOCOSFPFLUSH:
4156 		PF_RULES_WLOCK();
4157 		pf_osfp_flush();
4158 		PF_RULES_WUNLOCK();
4159 		break;
4160 
4161 	case DIOCIGETIFACES: {
4162 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
4163 		struct pfi_kif *ifstore;
4164 		size_t bufsiz;
4165 
4166 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
4167 			error = ENODEV;
4168 			break;
4169 		}
4170 
4171 		if (io->pfiio_size < 0 ||
4172 		    io->pfiio_size > pf_ioctl_maxcount ||
4173 		    WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
4174 			error = EINVAL;
4175 			break;
4176 		}
4177 
4178 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
4179 		ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
4180 		    M_TEMP, M_NOWAIT);
4181 		if (ifstore == NULL) {
4182 			error = ENOMEM;
4183 			break;
4184 		}
4185 
4186 		PF_RULES_RLOCK();
4187 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
4188 		PF_RULES_RUNLOCK();
4189 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
4190 		free(ifstore, M_TEMP);
4191 		break;
4192 	}
4193 
4194 	case DIOCSETIFFLAG: {
4195 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
4196 
4197 		PF_RULES_WLOCK();
4198 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
4199 		PF_RULES_WUNLOCK();
4200 		break;
4201 	}
4202 
4203 	case DIOCCLRIFFLAG: {
4204 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
4205 
4206 		PF_RULES_WLOCK();
4207 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
4208 		PF_RULES_WUNLOCK();
4209 		break;
4210 	}
4211 
4212 	default:
4213 		error = ENODEV;
4214 		break;
4215 	}
4216 fail:
4217 	if (sx_xlocked(&pf_ioctl_lock))
4218 		sx_xunlock(&pf_ioctl_lock);
4219 	CURVNET_RESTORE();
4220 
4221 	return (error);
4222 }
4223 
4224 void
4225 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
4226 {
4227 	bzero(sp, sizeof(struct pfsync_state));
4228 
4229 	/* copy from state key */
4230 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
4231 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
4232 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
4233 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
4234 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
4235 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
4236 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
4237 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
4238 	sp->proto = st->key[PF_SK_WIRE]->proto;
4239 	sp->af = st->key[PF_SK_WIRE]->af;
4240 
4241 	/* copy from state */
4242 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
4243 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
4244 	sp->creation = htonl(time_uptime - st->creation);
4245 	sp->expire = pf_state_expires(st);
4246 	if (sp->expire <= time_uptime)
4247 		sp->expire = htonl(0);
4248 	else
4249 		sp->expire = htonl(sp->expire - time_uptime);
4250 
4251 	sp->direction = st->direction;
4252 	sp->log = st->log;
4253 	sp->timeout = st->timeout;
4254 	sp->state_flags = st->state_flags;
4255 	if (st->src_node)
4256 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
4257 	if (st->nat_src_node)
4258 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
4259 
4260 	sp->id = st->id;
4261 	sp->creatorid = st->creatorid;
4262 	pf_state_peer_hton(&st->src, &sp->src);
4263 	pf_state_peer_hton(&st->dst, &sp->dst);
4264 
4265 	if (st->rule.ptr == NULL)
4266 		sp->rule = htonl(-1);
4267 	else
4268 		sp->rule = htonl(st->rule.ptr->nr);
4269 	if (st->anchor.ptr == NULL)
4270 		sp->anchor = htonl(-1);
4271 	else
4272 		sp->anchor = htonl(st->anchor.ptr->nr);
4273 	if (st->nat_rule.ptr == NULL)
4274 		sp->nat_rule = htonl(-1);
4275 	else
4276 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
4277 
4278 	pf_state_counter_hton(counter_u64_fetch(st->packets[0]),
4279 	    sp->packets[0]);
4280 	pf_state_counter_hton(counter_u64_fetch(st->packets[1]),
4281 	    sp->packets[1]);
4282 	pf_state_counter_hton(counter_u64_fetch(st->bytes[0]), sp->bytes[0]);
4283 	pf_state_counter_hton(counter_u64_fetch(st->bytes[1]), sp->bytes[1]);
4284 
4285 }
4286 
4287 static void
4288 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
4289 {
4290 	struct pfr_ktable *kt;
4291 
4292 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
4293 
4294 	kt = aw->p.tbl;
4295 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
4296 		kt = kt->pfrkt_root;
4297 	aw->p.tbl = NULL;
4298 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
4299 		kt->pfrkt_cnt : -1;
4300 }
4301 
4302 /*
4303  * XXX - Check for version missmatch!!!
4304  */
4305 static void
4306 pf_clear_states(void)
4307 {
4308 	struct pf_state	*s;
4309 	u_int i;
4310 
4311 	for (i = 0; i <= pf_hashmask; i++) {
4312 		struct pf_idhash *ih = &V_pf_idhash[i];
4313 relock:
4314 		PF_HASHROW_LOCK(ih);
4315 		LIST_FOREACH(s, &ih->states, entry) {
4316 			s->timeout = PFTM_PURGE;
4317 			/* Don't send out individual delete messages. */
4318 			s->state_flags |= PFSTATE_NOSYNC;
4319 			pf_unlink_state(s, PF_ENTER_LOCKED);
4320 			goto relock;
4321 		}
4322 		PF_HASHROW_UNLOCK(ih);
4323 	}
4324 }
4325 
4326 static int
4327 pf_clear_tables(void)
4328 {
4329 	struct pfioc_table io;
4330 	int error;
4331 
4332 	bzero(&io, sizeof(io));
4333 
4334 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
4335 	    io.pfrio_flags);
4336 
4337 	return (error);
4338 }
4339 
4340 static void
4341 pf_clear_srcnodes(struct pf_ksrc_node *n)
4342 {
4343 	struct pf_state *s;
4344 	int i;
4345 
4346 	for (i = 0; i <= pf_hashmask; i++) {
4347 		struct pf_idhash *ih = &V_pf_idhash[i];
4348 
4349 		PF_HASHROW_LOCK(ih);
4350 		LIST_FOREACH(s, &ih->states, entry) {
4351 			if (n == NULL || n == s->src_node)
4352 				s->src_node = NULL;
4353 			if (n == NULL || n == s->nat_src_node)
4354 				s->nat_src_node = NULL;
4355 		}
4356 		PF_HASHROW_UNLOCK(ih);
4357 	}
4358 
4359 	if (n == NULL) {
4360 		struct pf_srchash *sh;
4361 
4362 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4363 		    i++, sh++) {
4364 			PF_HASHROW_LOCK(sh);
4365 			LIST_FOREACH(n, &sh->nodes, entry) {
4366 				n->expire = 1;
4367 				n->states = 0;
4368 			}
4369 			PF_HASHROW_UNLOCK(sh);
4370 		}
4371 	} else {
4372 		/* XXX: hash slot should already be locked here. */
4373 		n->expire = 1;
4374 		n->states = 0;
4375 	}
4376 }
4377 
4378 static void
4379 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
4380 {
4381 	struct pf_ksrc_node_list	 kill;
4382 
4383 	LIST_INIT(&kill);
4384 	for (int i = 0; i <= pf_srchashmask; i++) {
4385 		struct pf_srchash *sh = &V_pf_srchash[i];
4386 		struct pf_ksrc_node *sn, *tmp;
4387 
4388 		PF_HASHROW_LOCK(sh);
4389 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
4390 			if (PF_MATCHA(psnk->psnk_src.neg,
4391 			      &psnk->psnk_src.addr.v.a.addr,
4392 			      &psnk->psnk_src.addr.v.a.mask,
4393 			      &sn->addr, sn->af) &&
4394 			    PF_MATCHA(psnk->psnk_dst.neg,
4395 			      &psnk->psnk_dst.addr.v.a.addr,
4396 			      &psnk->psnk_dst.addr.v.a.mask,
4397 			      &sn->raddr, sn->af)) {
4398 				pf_unlink_src_node(sn);
4399 				LIST_INSERT_HEAD(&kill, sn, entry);
4400 				sn->expire = 1;
4401 			}
4402 		PF_HASHROW_UNLOCK(sh);
4403 	}
4404 
4405 	for (int i = 0; i <= pf_hashmask; i++) {
4406 		struct pf_idhash *ih = &V_pf_idhash[i];
4407 		struct pf_state *s;
4408 
4409 		PF_HASHROW_LOCK(ih);
4410 		LIST_FOREACH(s, &ih->states, entry) {
4411 			if (s->src_node && s->src_node->expire == 1)
4412 				s->src_node = NULL;
4413 			if (s->nat_src_node && s->nat_src_node->expire == 1)
4414 				s->nat_src_node = NULL;
4415 		}
4416 		PF_HASHROW_UNLOCK(ih);
4417 	}
4418 
4419 	psnk->psnk_killed = pf_free_src_nodes(&kill);
4420 }
4421 
4422 /*
4423  * XXX - Check for version missmatch!!!
4424  */
4425 
4426 /*
4427  * Duplicate pfctl -Fa operation to get rid of as much as we can.
4428  */
4429 static int
4430 shutdown_pf(void)
4431 {
4432 	int error = 0;
4433 	u_int32_t t[5];
4434 	char nn = '\0';
4435 
4436 	do {
4437 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
4438 		    != 0) {
4439 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
4440 			break;
4441 		}
4442 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
4443 		    != 0) {
4444 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
4445 			break;		/* XXX: rollback? */
4446 		}
4447 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
4448 		    != 0) {
4449 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
4450 			break;		/* XXX: rollback? */
4451 		}
4452 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
4453 		    != 0) {
4454 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
4455 			break;		/* XXX: rollback? */
4456 		}
4457 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
4458 		    != 0) {
4459 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
4460 			break;		/* XXX: rollback? */
4461 		}
4462 
4463 		/* XXX: these should always succeed here */
4464 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
4465 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
4466 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
4467 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
4468 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
4469 
4470 		if ((error = pf_clear_tables()) != 0)
4471 			break;
4472 
4473 #ifdef ALTQ
4474 		if ((error = pf_begin_altq(&t[0])) != 0) {
4475 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
4476 			break;
4477 		}
4478 		pf_commit_altq(t[0]);
4479 #endif
4480 
4481 		pf_clear_states();
4482 
4483 		pf_clear_srcnodes(NULL);
4484 
4485 		/* status does not use malloced mem so no need to cleanup */
4486 		/* fingerprints and interfaces have their own cleanup code */
4487 	} while(0);
4488 
4489 	return (error);
4490 }
4491 
4492 static pfil_return_t
4493 pf_check_return(int chk, struct mbuf **m)
4494 {
4495 
4496 	switch (chk) {
4497 	case PF_PASS:
4498 		if (*m == NULL)
4499 			return (PFIL_CONSUMED);
4500 		else
4501 			return (PFIL_PASS);
4502 		break;
4503 	default:
4504 		if (*m != NULL) {
4505 			m_freem(*m);
4506 			*m = NULL;
4507 		}
4508 		return (PFIL_DROPPED);
4509 	}
4510 }
4511 
4512 #ifdef INET
4513 static pfil_return_t
4514 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
4515     void *ruleset __unused, struct inpcb *inp)
4516 {
4517 	int chk;
4518 
4519 	chk = pf_test(PF_IN, flags, ifp, m, inp);
4520 
4521 	return (pf_check_return(chk, m));
4522 }
4523 
4524 static pfil_return_t
4525 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
4526     void *ruleset __unused,  struct inpcb *inp)
4527 {
4528 	int chk;
4529 
4530 	chk = pf_test(PF_OUT, flags, ifp, m, inp);
4531 
4532 	return (pf_check_return(chk, m));
4533 }
4534 #endif
4535 
4536 #ifdef INET6
4537 static pfil_return_t
4538 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
4539     void *ruleset __unused,  struct inpcb *inp)
4540 {
4541 	int chk;
4542 
4543 	/*
4544 	 * In case of loopback traffic IPv6 uses the real interface in
4545 	 * order to support scoped addresses. In order to support stateful
4546 	 * filtering we have change this to lo0 as it is the case in IPv4.
4547 	 */
4548 	CURVNET_SET(ifp->if_vnet);
4549 	chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
4550 	CURVNET_RESTORE();
4551 
4552 	return (pf_check_return(chk, m));
4553 }
4554 
4555 static pfil_return_t
4556 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
4557     void *ruleset __unused,  struct inpcb *inp)
4558 {
4559 	int chk;
4560 
4561 	CURVNET_SET(ifp->if_vnet);
4562 	chk = pf_test6(PF_OUT, flags, ifp, m, inp);
4563 	CURVNET_RESTORE();
4564 
4565 	return (pf_check_return(chk, m));
4566 }
4567 #endif /* INET6 */
4568 
4569 #ifdef INET
4570 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
4571 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
4572 #define	V_pf_ip4_in_hook	VNET(pf_ip4_in_hook)
4573 #define	V_pf_ip4_out_hook	VNET(pf_ip4_out_hook)
4574 #endif
4575 #ifdef INET6
4576 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
4577 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
4578 #define	V_pf_ip6_in_hook	VNET(pf_ip6_in_hook)
4579 #define	V_pf_ip6_out_hook	VNET(pf_ip6_out_hook)
4580 #endif
4581 
4582 static void
4583 hook_pf(void)
4584 {
4585 	struct pfil_hook_args pha;
4586 	struct pfil_link_args pla;
4587 	int ret;
4588 
4589 	if (V_pf_pfil_hooked)
4590 		return;
4591 
4592 	pha.pa_version = PFIL_VERSION;
4593 	pha.pa_modname = "pf";
4594 	pha.pa_ruleset = NULL;
4595 
4596 	pla.pa_version = PFIL_VERSION;
4597 
4598 #ifdef INET
4599 	pha.pa_type = PFIL_TYPE_IP4;
4600 	pha.pa_func = pf_check_in;
4601 	pha.pa_flags = PFIL_IN;
4602 	pha.pa_rulname = "default-in";
4603 	V_pf_ip4_in_hook = pfil_add_hook(&pha);
4604 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
4605 	pla.pa_head = V_inet_pfil_head;
4606 	pla.pa_hook = V_pf_ip4_in_hook;
4607 	ret = pfil_link(&pla);
4608 	MPASS(ret == 0);
4609 	pha.pa_func = pf_check_out;
4610 	pha.pa_flags = PFIL_OUT;
4611 	pha.pa_rulname = "default-out";
4612 	V_pf_ip4_out_hook = pfil_add_hook(&pha);
4613 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
4614 	pla.pa_head = V_inet_pfil_head;
4615 	pla.pa_hook = V_pf_ip4_out_hook;
4616 	ret = pfil_link(&pla);
4617 	MPASS(ret == 0);
4618 #endif
4619 #ifdef INET6
4620 	pha.pa_type = PFIL_TYPE_IP6;
4621 	pha.pa_func = pf_check6_in;
4622 	pha.pa_flags = PFIL_IN;
4623 	pha.pa_rulname = "default-in6";
4624 	V_pf_ip6_in_hook = pfil_add_hook(&pha);
4625 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
4626 	pla.pa_head = V_inet6_pfil_head;
4627 	pla.pa_hook = V_pf_ip6_in_hook;
4628 	ret = pfil_link(&pla);
4629 	MPASS(ret == 0);
4630 	pha.pa_func = pf_check6_out;
4631 	pha.pa_rulname = "default-out6";
4632 	pha.pa_flags = PFIL_OUT;
4633 	V_pf_ip6_out_hook = pfil_add_hook(&pha);
4634 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
4635 	pla.pa_head = V_inet6_pfil_head;
4636 	pla.pa_hook = V_pf_ip6_out_hook;
4637 	ret = pfil_link(&pla);
4638 	MPASS(ret == 0);
4639 #endif
4640 
4641 	V_pf_pfil_hooked = 1;
4642 }
4643 
4644 static void
4645 dehook_pf(void)
4646 {
4647 
4648 	if (V_pf_pfil_hooked == 0)
4649 		return;
4650 
4651 #ifdef INET
4652 	pfil_remove_hook(V_pf_ip4_in_hook);
4653 	pfil_remove_hook(V_pf_ip4_out_hook);
4654 #endif
4655 #ifdef INET6
4656 	pfil_remove_hook(V_pf_ip6_in_hook);
4657 	pfil_remove_hook(V_pf_ip6_out_hook);
4658 #endif
4659 
4660 	V_pf_pfil_hooked = 0;
4661 }
4662 
4663 static void
4664 pf_load_vnet(void)
4665 {
4666 	V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
4667 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
4668 
4669 	pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
4670 	    PF_RULE_TAG_HASH_SIZE_DEFAULT);
4671 #ifdef ALTQ
4672 	pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
4673 	    PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
4674 #endif
4675 
4676 	pfattach_vnet();
4677 	V_pf_vnet_active = 1;
4678 }
4679 
4680 static int
4681 pf_load(void)
4682 {
4683 	int error;
4684 
4685 	rm_init(&pf_rules_lock, "pf rulesets");
4686 	sx_init(&pf_ioctl_lock, "pf ioctl");
4687 	sx_init(&pf_end_lock, "pf end thread");
4688 
4689 	pf_mtag_initialize();
4690 
4691 	pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
4692 	if (pf_dev == NULL)
4693 		return (ENOMEM);
4694 
4695 	pf_end_threads = 0;
4696 	error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
4697 	if (error != 0)
4698 		return (error);
4699 
4700 	pfi_initialize();
4701 
4702 	return (0);
4703 }
4704 
4705 static void
4706 pf_unload_vnet(void)
4707 {
4708 	int ret;
4709 
4710 	V_pf_vnet_active = 0;
4711 	V_pf_status.running = 0;
4712 	dehook_pf();
4713 
4714 	PF_RULES_WLOCK();
4715 	shutdown_pf();
4716 	PF_RULES_WUNLOCK();
4717 
4718 	ret = swi_remove(V_pf_swi_cookie);
4719 	MPASS(ret == 0);
4720 	ret = intr_event_destroy(V_pf_swi_ie);
4721 	MPASS(ret == 0);
4722 
4723 	pf_unload_vnet_purge();
4724 
4725 	pf_normalize_cleanup();
4726 	PF_RULES_WLOCK();
4727 	pfi_cleanup_vnet();
4728 	PF_RULES_WUNLOCK();
4729 	pfr_cleanup();
4730 	pf_osfp_flush();
4731 	pf_cleanup();
4732 	if (IS_DEFAULT_VNET(curvnet))
4733 		pf_mtag_cleanup();
4734 
4735 	pf_cleanup_tagset(&V_pf_tags);
4736 #ifdef ALTQ
4737 	pf_cleanup_tagset(&V_pf_qids);
4738 #endif
4739 	uma_zdestroy(V_pf_tag_z);
4740 
4741 	/* Free counters last as we updated them during shutdown. */
4742 	counter_u64_free(V_pf_default_rule.evaluations);
4743 	for (int i = 0; i < 2; i++) {
4744 		counter_u64_free(V_pf_default_rule.packets[i]);
4745 		counter_u64_free(V_pf_default_rule.bytes[i]);
4746 	}
4747 	counter_u64_free(V_pf_default_rule.states_cur);
4748 	counter_u64_free(V_pf_default_rule.states_tot);
4749 	counter_u64_free(V_pf_default_rule.src_nodes);
4750 
4751 	for (int i = 0; i < PFRES_MAX; i++)
4752 		counter_u64_free(V_pf_status.counters[i]);
4753 	for (int i = 0; i < LCNT_MAX; i++)
4754 		counter_u64_free(V_pf_status.lcounters[i]);
4755 	for (int i = 0; i < FCNT_MAX; i++)
4756 		counter_u64_free(V_pf_status.fcounters[i]);
4757 	for (int i = 0; i < SCNT_MAX; i++)
4758 		counter_u64_free(V_pf_status.scounters[i]);
4759 }
4760 
4761 static void
4762 pf_unload(void)
4763 {
4764 
4765 	sx_xlock(&pf_end_lock);
4766 	pf_end_threads = 1;
4767 	while (pf_end_threads < 2) {
4768 		wakeup_one(pf_purge_thread);
4769 		sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
4770 	}
4771 	sx_xunlock(&pf_end_lock);
4772 
4773 	if (pf_dev != NULL)
4774 		destroy_dev(pf_dev);
4775 
4776 	pfi_cleanup();
4777 
4778 	rm_destroy(&pf_rules_lock);
4779 	sx_destroy(&pf_ioctl_lock);
4780 	sx_destroy(&pf_end_lock);
4781 }
4782 
4783 static void
4784 vnet_pf_init(void *unused __unused)
4785 {
4786 
4787 	pf_load_vnet();
4788 }
4789 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
4790     vnet_pf_init, NULL);
4791 
4792 static void
4793 vnet_pf_uninit(const void *unused __unused)
4794 {
4795 
4796 	pf_unload_vnet();
4797 }
4798 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
4799 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
4800     vnet_pf_uninit, NULL);
4801 
4802 static int
4803 pf_modevent(module_t mod, int type, void *data)
4804 {
4805 	int error = 0;
4806 
4807 	switch(type) {
4808 	case MOD_LOAD:
4809 		error = pf_load();
4810 		break;
4811 	case MOD_UNLOAD:
4812 		/* Handled in SYSUNINIT(pf_unload) to ensure it's done after
4813 		 * the vnet_pf_uninit()s */
4814 		break;
4815 	default:
4816 		error = EINVAL;
4817 		break;
4818 	}
4819 
4820 	return (error);
4821 }
4822 
4823 static moduledata_t pf_mod = {
4824 	"pf",
4825 	pf_modevent,
4826 	0
4827 };
4828 
4829 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
4830 MODULE_VERSION(pf, PF_MODVER);
4831