xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision f15e18a642cb3f7ebc747f8e9cdf11274140107d)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2001 Daniel Hartmeier
5  * Copyright (c) 2002,2003 Henning Brauer
6  * Copyright (c) 2012 Gleb Smirnoff <glebius@FreeBSD.org>
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  *
13  *    - Redistributions of source code must retain the above copyright
14  *      notice, this list of conditions and the following disclaimer.
15  *    - Redistributions in binary form must reproduce the above
16  *      copyright notice, this list of conditions and the following
17  *      disclaimer in the documentation and/or other materials provided
18  *      with the distribution.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
26  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
28  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
30  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
31  * POSSIBILITY OF SUCH DAMAGE.
32  *
33  * Effort sponsored in part by the Defense Advanced Research Projects
34  * Agency (DARPA) and Air Force Research Laboratory, Air Force
35  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
36  *
37  *	$OpenBSD: pf_ioctl.c,v 1.213 2009/02/15 21:46:12 mbalmer Exp $
38  */
39 
40 #include <sys/cdefs.h>
41 __FBSDID("$FreeBSD$");
42 
43 #include "opt_inet.h"
44 #include "opt_inet6.h"
45 #include "opt_bpf.h"
46 #include "opt_pf.h"
47 
48 #include <sys/param.h>
49 #include <sys/_bitset.h>
50 #include <sys/bitset.h>
51 #include <sys/bus.h>
52 #include <sys/conf.h>
53 #include <sys/endian.h>
54 #include <sys/fcntl.h>
55 #include <sys/filio.h>
56 #include <sys/hash.h>
57 #include <sys/interrupt.h>
58 #include <sys/jail.h>
59 #include <sys/kernel.h>
60 #include <sys/kthread.h>
61 #include <sys/lock.h>
62 #include <sys/mbuf.h>
63 #include <sys/module.h>
64 #include <sys/proc.h>
65 #include <sys/smp.h>
66 #include <sys/socket.h>
67 #include <sys/sysctl.h>
68 #include <sys/md5.h>
69 #include <sys/ucred.h>
70 
71 #include <net/if.h>
72 #include <net/if_var.h>
73 #include <net/vnet.h>
74 #include <net/route.h>
75 #include <net/pfil.h>
76 #include <net/pfvar.h>
77 #include <net/if_pfsync.h>
78 #include <net/if_pflog.h>
79 
80 #include <netinet/in.h>
81 #include <netinet/ip.h>
82 #include <netinet/ip_var.h>
83 #include <netinet6/ip6_var.h>
84 #include <netinet/ip_icmp.h>
85 
86 #ifdef INET6
87 #include <netinet/ip6.h>
88 #endif /* INET6 */
89 
90 #ifdef ALTQ
91 #include <net/altq/altq.h>
92 #endif
93 
94 static struct pf_kpool	*pf_get_kpool(char *, u_int32_t, u_int8_t, u_int32_t,
95 			    u_int8_t, u_int8_t, u_int8_t);
96 
97 static void		 pf_mv_kpool(struct pf_kpalist *, struct pf_kpalist *);
98 static void		 pf_empty_kpool(struct pf_kpalist *);
99 static int		 pfioctl(struct cdev *, u_long, caddr_t, int,
100 			    struct thread *);
101 #ifdef ALTQ
102 static int		 pf_begin_altq(u_int32_t *);
103 static int		 pf_rollback_altq(u_int32_t);
104 static int		 pf_commit_altq(u_int32_t);
105 static int		 pf_enable_altq(struct pf_altq *);
106 static int		 pf_disable_altq(struct pf_altq *);
107 static u_int32_t	 pf_qname2qid(char *);
108 static void		 pf_qid_unref(u_int32_t);
109 #endif /* ALTQ */
110 static int		 pf_begin_rules(u_int32_t *, int, const char *);
111 static int		 pf_rollback_rules(u_int32_t, int, char *);
112 static int		 pf_setup_pfsync_matching(struct pf_kruleset *);
113 static void		 pf_hash_rule(MD5_CTX *, struct pf_krule *);
114 static void		 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
115 static int		 pf_commit_rules(u_int32_t, int, char *);
116 static int		 pf_addr_setup(struct pf_kruleset *,
117 			    struct pf_addr_wrap *, sa_family_t);
118 static void		 pf_addr_copyout(struct pf_addr_wrap *);
119 static void		 pf_src_node_copy(const struct pf_ksrc_node *,
120 			    struct pf_src_node *);
121 #ifdef ALTQ
122 static int		 pf_export_kaltq(struct pf_altq *,
123 			    struct pfioc_altq_v1 *, size_t);
124 static int		 pf_import_kaltq(struct pfioc_altq_v1 *,
125 			    struct pf_altq *, size_t);
126 #endif /* ALTQ */
127 
128 VNET_DEFINE(struct pf_krule,	pf_default_rule);
129 
130 #ifdef ALTQ
131 VNET_DEFINE_STATIC(int,		pf_altq_running);
132 #define	V_pf_altq_running	VNET(pf_altq_running)
133 #endif
134 
135 #define	TAGID_MAX	 50000
136 struct pf_tagname {
137 	TAILQ_ENTRY(pf_tagname)	namehash_entries;
138 	TAILQ_ENTRY(pf_tagname)	taghash_entries;
139 	char			name[PF_TAG_NAME_SIZE];
140 	uint16_t		tag;
141 	int			ref;
142 };
143 
144 struct pf_tagset {
145 	TAILQ_HEAD(, pf_tagname)	*namehash;
146 	TAILQ_HEAD(, pf_tagname)	*taghash;
147 	unsigned int			 mask;
148 	uint32_t			 seed;
149 	BITSET_DEFINE(, TAGID_MAX)	 avail;
150 };
151 
152 VNET_DEFINE(struct pf_tagset, pf_tags);
153 #define	V_pf_tags	VNET(pf_tags)
154 static unsigned int	pf_rule_tag_hashsize;
155 #define	PF_RULE_TAG_HASH_SIZE_DEFAULT	128
156 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN,
157     &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT,
158     "Size of pf(4) rule tag hashtable");
159 
160 #ifdef ALTQ
161 VNET_DEFINE(struct pf_tagset, pf_qids);
162 #define	V_pf_qids	VNET(pf_qids)
163 static unsigned int	pf_queue_tag_hashsize;
164 #define	PF_QUEUE_TAG_HASH_SIZE_DEFAULT	128
165 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN,
166     &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT,
167     "Size of pf(4) queue tag hashtable");
168 #endif
169 VNET_DEFINE(uma_zone_t,	 pf_tag_z);
170 #define	V_pf_tag_z		 VNET(pf_tag_z)
171 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
172 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
173 
174 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
175 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
176 #endif
177 
178 static void		 pf_init_tagset(struct pf_tagset *, unsigned int *,
179 			    unsigned int);
180 static void		 pf_cleanup_tagset(struct pf_tagset *);
181 static uint16_t		 tagname2hashindex(const struct pf_tagset *, const char *);
182 static uint16_t		 tag2hashindex(const struct pf_tagset *, uint16_t);
183 static u_int16_t	 tagname2tag(struct pf_tagset *, char *);
184 static u_int16_t	 pf_tagname2tag(char *);
185 static void		 tag_unref(struct pf_tagset *, u_int16_t);
186 
187 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
188 
189 struct cdev *pf_dev;
190 
191 /*
192  * XXX - These are new and need to be checked when moveing to a new version
193  */
194 static void		 pf_clear_states(void);
195 static int		 pf_clear_tables(void);
196 static void		 pf_clear_srcnodes(struct pf_ksrc_node *);
197 static void		 pf_kill_srcnodes(struct pfioc_src_node_kill *);
198 static void		 pf_tbladdr_copyout(struct pf_addr_wrap *);
199 
200 /*
201  * Wrapper functions for pfil(9) hooks
202  */
203 #ifdef INET
204 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp,
205     int flags, void *ruleset __unused, struct inpcb *inp);
206 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp,
207     int flags, void *ruleset __unused, struct inpcb *inp);
208 #endif
209 #ifdef INET6
210 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp,
211     int flags, void *ruleset __unused, struct inpcb *inp);
212 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp,
213     int flags, void *ruleset __unused, struct inpcb *inp);
214 #endif
215 
216 static int		hook_pf(void);
217 static int		dehook_pf(void);
218 static int		shutdown_pf(void);
219 static int		pf_load(void);
220 static void		pf_unload(void);
221 
222 static struct cdevsw pf_cdevsw = {
223 	.d_ioctl =	pfioctl,
224 	.d_name =	PF_NAME,
225 	.d_version =	D_VERSION,
226 };
227 
228 volatile VNET_DEFINE_STATIC(int, pf_pfil_hooked);
229 #define V_pf_pfil_hooked	VNET(pf_pfil_hooked)
230 
231 /*
232  * We need a flag that is neither hooked nor running to know when
233  * the VNET is "valid".  We primarily need this to control (global)
234  * external event, e.g., eventhandlers.
235  */
236 VNET_DEFINE(int, pf_vnet_active);
237 #define V_pf_vnet_active	VNET(pf_vnet_active)
238 
239 int pf_end_threads;
240 struct proc *pf_purge_proc;
241 
242 struct rmlock			pf_rules_lock;
243 struct sx			pf_ioctl_lock;
244 struct sx			pf_end_lock;
245 
246 /* pfsync */
247 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
248 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
249 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
250 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
251 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
252 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
253 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
254 
255 /* pflog */
256 pflog_packet_t			*pflog_packet_ptr = NULL;
257 
258 extern u_long	pf_ioctl_maxcount;
259 
260 static void
261 pfattach_vnet(void)
262 {
263 	u_int32_t *my_timeout = V_pf_default_rule.timeout;
264 
265 	pf_initialize();
266 	pfr_initialize();
267 	pfi_initialize_vnet();
268 	pf_normalize_init();
269 
270 	V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
271 	V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
272 
273 	RB_INIT(&V_pf_anchors);
274 	pf_init_kruleset(&pf_main_ruleset);
275 
276 	/* default rule should never be garbage collected */
277 	V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
278 #ifdef PF_DEFAULT_TO_DROP
279 	V_pf_default_rule.action = PF_DROP;
280 #else
281 	V_pf_default_rule.action = PF_PASS;
282 #endif
283 	V_pf_default_rule.nr = -1;
284 	V_pf_default_rule.rtableid = -1;
285 
286 	V_pf_default_rule.evaluations = counter_u64_alloc(M_WAITOK);
287 	for (int i = 0; i < 2; i++) {
288 		V_pf_default_rule.packets[i] = counter_u64_alloc(M_WAITOK);
289 		V_pf_default_rule.bytes[i] = counter_u64_alloc(M_WAITOK);
290 	}
291 	V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
292 	V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
293 	V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
294 
295 	/* initialize default timeouts */
296 	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
297 	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
298 	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
299 	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
300 	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
301 	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
302 	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
303 	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
304 	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
305 	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
306 	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
307 	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
308 	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
309 	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
310 	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
311 	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
312 	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
313 	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
314 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
315 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
316 
317 	bzero(&V_pf_status, sizeof(V_pf_status));
318 	V_pf_status.debug = PF_DEBUG_URGENT;
319 
320 	V_pf_pfil_hooked = 0;
321 
322 	/* XXX do our best to avoid a conflict */
323 	V_pf_status.hostid = arc4random();
324 
325 	for (int i = 0; i < PFRES_MAX; i++)
326 		V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
327 	for (int i = 0; i < LCNT_MAX; i++)
328 		V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
329 	for (int i = 0; i < FCNT_MAX; i++)
330 		V_pf_status.fcounters[i] = counter_u64_alloc(M_WAITOK);
331 	for (int i = 0; i < SCNT_MAX; i++)
332 		V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
333 
334 	if (swi_add(NULL, "pf send", pf_intr, curvnet, SWI_NET,
335 	    INTR_MPSAFE, &V_pf_swi_cookie) != 0)
336 		/* XXXGL: leaked all above. */
337 		return;
338 }
339 
340 static struct pf_kpool *
341 pf_get_kpool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
342     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
343     u_int8_t check_ticket)
344 {
345 	struct pf_kruleset	*ruleset;
346 	struct pf_krule		*rule;
347 	int			 rs_num;
348 
349 	ruleset = pf_find_kruleset(anchor);
350 	if (ruleset == NULL)
351 		return (NULL);
352 	rs_num = pf_get_ruleset_number(rule_action);
353 	if (rs_num >= PF_RULESET_MAX)
354 		return (NULL);
355 	if (active) {
356 		if (check_ticket && ticket !=
357 		    ruleset->rules[rs_num].active.ticket)
358 			return (NULL);
359 		if (r_last)
360 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
361 			    pf_krulequeue);
362 		else
363 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
364 	} else {
365 		if (check_ticket && ticket !=
366 		    ruleset->rules[rs_num].inactive.ticket)
367 			return (NULL);
368 		if (r_last)
369 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
370 			    pf_krulequeue);
371 		else
372 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
373 	}
374 	if (!r_last) {
375 		while ((rule != NULL) && (rule->nr != rule_number))
376 			rule = TAILQ_NEXT(rule, entries);
377 	}
378 	if (rule == NULL)
379 		return (NULL);
380 
381 	return (&rule->rpool);
382 }
383 
384 static void
385 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb)
386 {
387 	struct pf_kpooladdr	*mv_pool_pa;
388 
389 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
390 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
391 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
392 	}
393 }
394 
395 static void
396 pf_empty_kpool(struct pf_kpalist *poola)
397 {
398 	struct pf_kpooladdr *pa;
399 
400 	while ((pa = TAILQ_FIRST(poola)) != NULL) {
401 		switch (pa->addr.type) {
402 		case PF_ADDR_DYNIFTL:
403 			pfi_dynaddr_remove(pa->addr.p.dyn);
404 			break;
405 		case PF_ADDR_TABLE:
406 			/* XXX: this could be unfinished pooladdr on pabuf */
407 			if (pa->addr.p.tbl != NULL)
408 				pfr_detach_table(pa->addr.p.tbl);
409 			break;
410 		}
411 		if (pa->kif)
412 			pfi_kkif_unref(pa->kif);
413 		TAILQ_REMOVE(poola, pa, entries);
414 		free(pa, M_PFRULE);
415 	}
416 }
417 
418 static void
419 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
420 {
421 
422 	PF_RULES_WASSERT();
423 
424 	TAILQ_REMOVE(rulequeue, rule, entries);
425 
426 	PF_UNLNKDRULES_LOCK();
427 	rule->rule_flag |= PFRULE_REFS;
428 	TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
429 	PF_UNLNKDRULES_UNLOCK();
430 }
431 
432 void
433 pf_free_rule(struct pf_krule *rule)
434 {
435 
436 	PF_RULES_WASSERT();
437 
438 	if (rule->tag)
439 		tag_unref(&V_pf_tags, rule->tag);
440 	if (rule->match_tag)
441 		tag_unref(&V_pf_tags, rule->match_tag);
442 #ifdef ALTQ
443 	if (rule->pqid != rule->qid)
444 		pf_qid_unref(rule->pqid);
445 	pf_qid_unref(rule->qid);
446 #endif
447 	switch (rule->src.addr.type) {
448 	case PF_ADDR_DYNIFTL:
449 		pfi_dynaddr_remove(rule->src.addr.p.dyn);
450 		break;
451 	case PF_ADDR_TABLE:
452 		pfr_detach_table(rule->src.addr.p.tbl);
453 		break;
454 	}
455 	switch (rule->dst.addr.type) {
456 	case PF_ADDR_DYNIFTL:
457 		pfi_dynaddr_remove(rule->dst.addr.p.dyn);
458 		break;
459 	case PF_ADDR_TABLE:
460 		pfr_detach_table(rule->dst.addr.p.tbl);
461 		break;
462 	}
463 	if (rule->overload_tbl)
464 		pfr_detach_table(rule->overload_tbl);
465 	if (rule->kif)
466 		pfi_kkif_unref(rule->kif);
467 	pf_kanchor_remove(rule);
468 	pf_empty_kpool(&rule->rpool.list);
469 	counter_u64_free(rule->evaluations);
470 	for (int i = 0; i < 2; i++) {
471 		counter_u64_free(rule->packets[i]);
472 		counter_u64_free(rule->bytes[i]);
473 	}
474 	counter_u64_free(rule->states_cur);
475 	counter_u64_free(rule->states_tot);
476 	counter_u64_free(rule->src_nodes);
477 	free(rule, M_PFRULE);
478 }
479 
480 static void
481 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
482     unsigned int default_size)
483 {
484 	unsigned int i;
485 	unsigned int hashsize;
486 
487 	if (*tunable_size == 0 || !powerof2(*tunable_size))
488 		*tunable_size = default_size;
489 
490 	hashsize = *tunable_size;
491 	ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
492 	    M_WAITOK);
493 	ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
494 	    M_WAITOK);
495 	ts->mask = hashsize - 1;
496 	ts->seed = arc4random();
497 	for (i = 0; i < hashsize; i++) {
498 		TAILQ_INIT(&ts->namehash[i]);
499 		TAILQ_INIT(&ts->taghash[i]);
500 	}
501 	BIT_FILL(TAGID_MAX, &ts->avail);
502 }
503 
504 static void
505 pf_cleanup_tagset(struct pf_tagset *ts)
506 {
507 	unsigned int i;
508 	unsigned int hashsize;
509 	struct pf_tagname *t, *tmp;
510 
511 	/*
512 	 * Only need to clean up one of the hashes as each tag is hashed
513 	 * into each table.
514 	 */
515 	hashsize = ts->mask + 1;
516 	for (i = 0; i < hashsize; i++)
517 		TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
518 			uma_zfree(V_pf_tag_z, t);
519 
520 	free(ts->namehash, M_PFHASH);
521 	free(ts->taghash, M_PFHASH);
522 }
523 
524 static uint16_t
525 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
526 {
527 	size_t len;
528 
529 	len = strnlen(tagname, PF_TAG_NAME_SIZE - 1);
530 	return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask);
531 }
532 
533 static uint16_t
534 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
535 {
536 
537 	return (tag & ts->mask);
538 }
539 
540 static u_int16_t
541 tagname2tag(struct pf_tagset *ts, char *tagname)
542 {
543 	struct pf_tagname	*tag;
544 	u_int32_t		 index;
545 	u_int16_t		 new_tagid;
546 
547 	PF_RULES_WASSERT();
548 
549 	index = tagname2hashindex(ts, tagname);
550 	TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
551 		if (strcmp(tagname, tag->name) == 0) {
552 			tag->ref++;
553 			return (tag->tag);
554 		}
555 
556 	/*
557 	 * new entry
558 	 *
559 	 * to avoid fragmentation, we do a linear search from the beginning
560 	 * and take the first free slot we find.
561 	 */
562 	new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
563 	/*
564 	 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
565 	 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
566 	 * set.  It may also return a bit number greater than TAGID_MAX due
567 	 * to rounding of the number of bits in the vector up to a multiple
568 	 * of the vector word size at declaration/allocation time.
569 	 */
570 	if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
571 		return (0);
572 
573 	/* Mark the tag as in use.  Bits are 0-based for BIT_CLR() */
574 	BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
575 
576 	/* allocate and fill new struct pf_tagname */
577 	tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
578 	if (tag == NULL)
579 		return (0);
580 	strlcpy(tag->name, tagname, sizeof(tag->name));
581 	tag->tag = new_tagid;
582 	tag->ref = 1;
583 
584 	/* Insert into namehash */
585 	TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
586 
587 	/* Insert into taghash */
588 	index = tag2hashindex(ts, new_tagid);
589 	TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
590 
591 	return (tag->tag);
592 }
593 
594 static void
595 tag_unref(struct pf_tagset *ts, u_int16_t tag)
596 {
597 	struct pf_tagname	*t;
598 	uint16_t		 index;
599 
600 	PF_RULES_WASSERT();
601 
602 	index = tag2hashindex(ts, tag);
603 	TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
604 		if (tag == t->tag) {
605 			if (--t->ref == 0) {
606 				TAILQ_REMOVE(&ts->taghash[index], t,
607 				    taghash_entries);
608 				index = tagname2hashindex(ts, t->name);
609 				TAILQ_REMOVE(&ts->namehash[index], t,
610 				    namehash_entries);
611 				/* Bits are 0-based for BIT_SET() */
612 				BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
613 				uma_zfree(V_pf_tag_z, t);
614 			}
615 			break;
616 		}
617 }
618 
619 static u_int16_t
620 pf_tagname2tag(char *tagname)
621 {
622 	return (tagname2tag(&V_pf_tags, tagname));
623 }
624 
625 #ifdef ALTQ
626 static u_int32_t
627 pf_qname2qid(char *qname)
628 {
629 	return ((u_int32_t)tagname2tag(&V_pf_qids, qname));
630 }
631 
632 static void
633 pf_qid_unref(u_int32_t qid)
634 {
635 	tag_unref(&V_pf_qids, (u_int16_t)qid);
636 }
637 
638 static int
639 pf_begin_altq(u_int32_t *ticket)
640 {
641 	struct pf_altq	*altq, *tmp;
642 	int		 error = 0;
643 
644 	PF_RULES_WASSERT();
645 
646 	/* Purge the old altq lists */
647 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
648 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
649 			/* detach and destroy the discipline */
650 			error = altq_remove(altq);
651 		}
652 		free(altq, M_PFALTQ);
653 	}
654 	TAILQ_INIT(V_pf_altq_ifs_inactive);
655 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
656 		pf_qid_unref(altq->qid);
657 		free(altq, M_PFALTQ);
658 	}
659 	TAILQ_INIT(V_pf_altqs_inactive);
660 	if (error)
661 		return (error);
662 	*ticket = ++V_ticket_altqs_inactive;
663 	V_altqs_inactive_open = 1;
664 	return (0);
665 }
666 
667 static int
668 pf_rollback_altq(u_int32_t ticket)
669 {
670 	struct pf_altq	*altq, *tmp;
671 	int		 error = 0;
672 
673 	PF_RULES_WASSERT();
674 
675 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
676 		return (0);
677 	/* Purge the old altq lists */
678 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
679 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
680 			/* detach and destroy the discipline */
681 			error = altq_remove(altq);
682 		}
683 		free(altq, M_PFALTQ);
684 	}
685 	TAILQ_INIT(V_pf_altq_ifs_inactive);
686 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
687 		pf_qid_unref(altq->qid);
688 		free(altq, M_PFALTQ);
689 	}
690 	TAILQ_INIT(V_pf_altqs_inactive);
691 	V_altqs_inactive_open = 0;
692 	return (error);
693 }
694 
695 static int
696 pf_commit_altq(u_int32_t ticket)
697 {
698 	struct pf_altqqueue	*old_altqs, *old_altq_ifs;
699 	struct pf_altq		*altq, *tmp;
700 	int			 err, error = 0;
701 
702 	PF_RULES_WASSERT();
703 
704 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
705 		return (EBUSY);
706 
707 	/* swap altqs, keep the old. */
708 	old_altqs = V_pf_altqs_active;
709 	old_altq_ifs = V_pf_altq_ifs_active;
710 	V_pf_altqs_active = V_pf_altqs_inactive;
711 	V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
712 	V_pf_altqs_inactive = old_altqs;
713 	V_pf_altq_ifs_inactive = old_altq_ifs;
714 	V_ticket_altqs_active = V_ticket_altqs_inactive;
715 
716 	/* Attach new disciplines */
717 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
718 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
719 			/* attach the discipline */
720 			error = altq_pfattach(altq);
721 			if (error == 0 && V_pf_altq_running)
722 				error = pf_enable_altq(altq);
723 			if (error != 0)
724 				return (error);
725 		}
726 	}
727 
728 	/* Purge the old altq lists */
729 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
730 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
731 			/* detach and destroy the discipline */
732 			if (V_pf_altq_running)
733 				error = pf_disable_altq(altq);
734 			err = altq_pfdetach(altq);
735 			if (err != 0 && error == 0)
736 				error = err;
737 			err = altq_remove(altq);
738 			if (err != 0 && error == 0)
739 				error = err;
740 		}
741 		free(altq, M_PFALTQ);
742 	}
743 	TAILQ_INIT(V_pf_altq_ifs_inactive);
744 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
745 		pf_qid_unref(altq->qid);
746 		free(altq, M_PFALTQ);
747 	}
748 	TAILQ_INIT(V_pf_altqs_inactive);
749 
750 	V_altqs_inactive_open = 0;
751 	return (error);
752 }
753 
754 static int
755 pf_enable_altq(struct pf_altq *altq)
756 {
757 	struct ifnet		*ifp;
758 	struct tb_profile	 tb;
759 	int			 error = 0;
760 
761 	if ((ifp = ifunit(altq->ifname)) == NULL)
762 		return (EINVAL);
763 
764 	if (ifp->if_snd.altq_type != ALTQT_NONE)
765 		error = altq_enable(&ifp->if_snd);
766 
767 	/* set tokenbucket regulator */
768 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
769 		tb.rate = altq->ifbandwidth;
770 		tb.depth = altq->tbrsize;
771 		error = tbr_set(&ifp->if_snd, &tb);
772 	}
773 
774 	return (error);
775 }
776 
777 static int
778 pf_disable_altq(struct pf_altq *altq)
779 {
780 	struct ifnet		*ifp;
781 	struct tb_profile	 tb;
782 	int			 error;
783 
784 	if ((ifp = ifunit(altq->ifname)) == NULL)
785 		return (EINVAL);
786 
787 	/*
788 	 * when the discipline is no longer referenced, it was overridden
789 	 * by a new one.  if so, just return.
790 	 */
791 	if (altq->altq_disc != ifp->if_snd.altq_disc)
792 		return (0);
793 
794 	error = altq_disable(&ifp->if_snd);
795 
796 	if (error == 0) {
797 		/* clear tokenbucket regulator */
798 		tb.rate = 0;
799 		error = tbr_set(&ifp->if_snd, &tb);
800 	}
801 
802 	return (error);
803 }
804 
805 static int
806 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
807     struct pf_altq *altq)
808 {
809 	struct ifnet	*ifp1;
810 	int		 error = 0;
811 
812 	/* Deactivate the interface in question */
813 	altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
814 	if ((ifp1 = ifunit(altq->ifname)) == NULL ||
815 	    (remove && ifp1 == ifp)) {
816 		altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
817 	} else {
818 		error = altq_add(ifp1, altq);
819 
820 		if (ticket != V_ticket_altqs_inactive)
821 			error = EBUSY;
822 
823 		if (error)
824 			free(altq, M_PFALTQ);
825 	}
826 
827 	return (error);
828 }
829 
830 void
831 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
832 {
833 	struct pf_altq	*a1, *a2, *a3;
834 	u_int32_t	 ticket;
835 	int		 error = 0;
836 
837 	/*
838 	 * No need to re-evaluate the configuration for events on interfaces
839 	 * that do not support ALTQ, as it's not possible for such
840 	 * interfaces to be part of the configuration.
841 	 */
842 	if (!ALTQ_IS_READY(&ifp->if_snd))
843 		return;
844 
845 	/* Interrupt userland queue modifications */
846 	if (V_altqs_inactive_open)
847 		pf_rollback_altq(V_ticket_altqs_inactive);
848 
849 	/* Start new altq ruleset */
850 	if (pf_begin_altq(&ticket))
851 		return;
852 
853 	/* Copy the current active set */
854 	TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
855 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
856 		if (a2 == NULL) {
857 			error = ENOMEM;
858 			break;
859 		}
860 		bcopy(a1, a2, sizeof(struct pf_altq));
861 
862 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
863 		if (error)
864 			break;
865 
866 		TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
867 	}
868 	if (error)
869 		goto out;
870 	TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
871 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
872 		if (a2 == NULL) {
873 			error = ENOMEM;
874 			break;
875 		}
876 		bcopy(a1, a2, sizeof(struct pf_altq));
877 
878 		if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
879 			error = EBUSY;
880 			free(a2, M_PFALTQ);
881 			break;
882 		}
883 		a2->altq_disc = NULL;
884 		TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
885 			if (strncmp(a3->ifname, a2->ifname,
886 				IFNAMSIZ) == 0) {
887 				a2->altq_disc = a3->altq_disc;
888 				break;
889 			}
890 		}
891 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
892 		if (error)
893 			break;
894 
895 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
896 	}
897 
898 out:
899 	if (error != 0)
900 		pf_rollback_altq(ticket);
901 	else
902 		pf_commit_altq(ticket);
903 }
904 #endif /* ALTQ */
905 
906 static int
907 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
908 {
909 	struct pf_kruleset	*rs;
910 	struct pf_krule		*rule;
911 
912 	PF_RULES_WASSERT();
913 
914 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
915 		return (EINVAL);
916 	rs = pf_find_or_create_kruleset(anchor);
917 	if (rs == NULL)
918 		return (EINVAL);
919 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
920 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
921 		rs->rules[rs_num].inactive.rcount--;
922 	}
923 	*ticket = ++rs->rules[rs_num].inactive.ticket;
924 	rs->rules[rs_num].inactive.open = 1;
925 	return (0);
926 }
927 
928 static int
929 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
930 {
931 	struct pf_kruleset	*rs;
932 	struct pf_krule		*rule;
933 
934 	PF_RULES_WASSERT();
935 
936 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
937 		return (EINVAL);
938 	rs = pf_find_kruleset(anchor);
939 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
940 	    rs->rules[rs_num].inactive.ticket != ticket)
941 		return (0);
942 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
943 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
944 		rs->rules[rs_num].inactive.rcount--;
945 	}
946 	rs->rules[rs_num].inactive.open = 0;
947 	return (0);
948 }
949 
950 #define PF_MD5_UPD(st, elm)						\
951 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
952 
953 #define PF_MD5_UPD_STR(st, elm)						\
954 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
955 
956 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
957 		(stor) = htonl((st)->elm);				\
958 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
959 } while (0)
960 
961 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
962 		(stor) = htons((st)->elm);				\
963 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
964 } while (0)
965 
966 static void
967 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
968 {
969 	PF_MD5_UPD(pfr, addr.type);
970 	switch (pfr->addr.type) {
971 		case PF_ADDR_DYNIFTL:
972 			PF_MD5_UPD(pfr, addr.v.ifname);
973 			PF_MD5_UPD(pfr, addr.iflags);
974 			break;
975 		case PF_ADDR_TABLE:
976 			PF_MD5_UPD(pfr, addr.v.tblname);
977 			break;
978 		case PF_ADDR_ADDRMASK:
979 			/* XXX ignore af? */
980 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
981 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
982 			break;
983 	}
984 
985 	PF_MD5_UPD(pfr, port[0]);
986 	PF_MD5_UPD(pfr, port[1]);
987 	PF_MD5_UPD(pfr, neg);
988 	PF_MD5_UPD(pfr, port_op);
989 }
990 
991 static void
992 pf_hash_rule(MD5_CTX *ctx, struct pf_krule *rule)
993 {
994 	u_int16_t x;
995 	u_int32_t y;
996 
997 	pf_hash_rule_addr(ctx, &rule->src);
998 	pf_hash_rule_addr(ctx, &rule->dst);
999 	PF_MD5_UPD_STR(rule, label);
1000 	PF_MD5_UPD_STR(rule, ifname);
1001 	PF_MD5_UPD_STR(rule, match_tagname);
1002 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
1003 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
1004 	PF_MD5_UPD_HTONL(rule, prob, y);
1005 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
1006 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1007 	PF_MD5_UPD(rule, uid.op);
1008 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1009 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1010 	PF_MD5_UPD(rule, gid.op);
1011 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
1012 	PF_MD5_UPD(rule, action);
1013 	PF_MD5_UPD(rule, direction);
1014 	PF_MD5_UPD(rule, af);
1015 	PF_MD5_UPD(rule, quick);
1016 	PF_MD5_UPD(rule, ifnot);
1017 	PF_MD5_UPD(rule, match_tag_not);
1018 	PF_MD5_UPD(rule, natpass);
1019 	PF_MD5_UPD(rule, keep_state);
1020 	PF_MD5_UPD(rule, proto);
1021 	PF_MD5_UPD(rule, type);
1022 	PF_MD5_UPD(rule, code);
1023 	PF_MD5_UPD(rule, flags);
1024 	PF_MD5_UPD(rule, flagset);
1025 	PF_MD5_UPD(rule, allow_opts);
1026 	PF_MD5_UPD(rule, rt);
1027 	PF_MD5_UPD(rule, tos);
1028 }
1029 
1030 static int
1031 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1032 {
1033 	struct pf_kruleset	*rs;
1034 	struct pf_krule		*rule, **old_array;
1035 	struct pf_krulequeue	*old_rules;
1036 	int			 error;
1037 	u_int32_t		 old_rcount;
1038 
1039 	PF_RULES_WASSERT();
1040 
1041 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1042 		return (EINVAL);
1043 	rs = pf_find_kruleset(anchor);
1044 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1045 	    ticket != rs->rules[rs_num].inactive.ticket)
1046 		return (EBUSY);
1047 
1048 	/* Calculate checksum for the main ruleset */
1049 	if (rs == &pf_main_ruleset) {
1050 		error = pf_setup_pfsync_matching(rs);
1051 		if (error != 0)
1052 			return (error);
1053 	}
1054 
1055 	/* Swap rules, keep the old. */
1056 	old_rules = rs->rules[rs_num].active.ptr;
1057 	old_rcount = rs->rules[rs_num].active.rcount;
1058 	old_array = rs->rules[rs_num].active.ptr_array;
1059 
1060 	rs->rules[rs_num].active.ptr =
1061 	    rs->rules[rs_num].inactive.ptr;
1062 	rs->rules[rs_num].active.ptr_array =
1063 	    rs->rules[rs_num].inactive.ptr_array;
1064 	rs->rules[rs_num].active.rcount =
1065 	    rs->rules[rs_num].inactive.rcount;
1066 	rs->rules[rs_num].inactive.ptr = old_rules;
1067 	rs->rules[rs_num].inactive.ptr_array = old_array;
1068 	rs->rules[rs_num].inactive.rcount = old_rcount;
1069 
1070 	rs->rules[rs_num].active.ticket =
1071 	    rs->rules[rs_num].inactive.ticket;
1072 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1073 
1074 	/* Purge the old rule list. */
1075 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1076 		pf_unlink_rule(old_rules, rule);
1077 	if (rs->rules[rs_num].inactive.ptr_array)
1078 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1079 	rs->rules[rs_num].inactive.ptr_array = NULL;
1080 	rs->rules[rs_num].inactive.rcount = 0;
1081 	rs->rules[rs_num].inactive.open = 0;
1082 	pf_remove_if_empty_kruleset(rs);
1083 
1084 	return (0);
1085 }
1086 
1087 static int
1088 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1089 {
1090 	MD5_CTX			 ctx;
1091 	struct pf_krule		*rule;
1092 	int			 rs_cnt;
1093 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
1094 
1095 	MD5Init(&ctx);
1096 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1097 		/* XXX PF_RULESET_SCRUB as well? */
1098 		if (rs_cnt == PF_RULESET_SCRUB)
1099 			continue;
1100 
1101 		if (rs->rules[rs_cnt].inactive.ptr_array)
1102 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1103 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
1104 
1105 		if (rs->rules[rs_cnt].inactive.rcount) {
1106 			rs->rules[rs_cnt].inactive.ptr_array =
1107 			    malloc(sizeof(caddr_t) *
1108 			    rs->rules[rs_cnt].inactive.rcount,
1109 			    M_TEMP, M_NOWAIT);
1110 
1111 			if (!rs->rules[rs_cnt].inactive.ptr_array)
1112 				return (ENOMEM);
1113 		}
1114 
1115 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1116 		    entries) {
1117 			pf_hash_rule(&ctx, rule);
1118 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1119 		}
1120 	}
1121 
1122 	MD5Final(digest, &ctx);
1123 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1124 	return (0);
1125 }
1126 
1127 static int
1128 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1129     sa_family_t af)
1130 {
1131 	int error = 0;
1132 
1133 	switch (addr->type) {
1134 	case PF_ADDR_TABLE:
1135 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1136 		if (addr->p.tbl == NULL)
1137 			error = ENOMEM;
1138 		break;
1139 	case PF_ADDR_DYNIFTL:
1140 		error = pfi_dynaddr_setup(addr, af);
1141 		break;
1142 	}
1143 
1144 	return (error);
1145 }
1146 
1147 static void
1148 pf_addr_copyout(struct pf_addr_wrap *addr)
1149 {
1150 
1151 	switch (addr->type) {
1152 	case PF_ADDR_DYNIFTL:
1153 		pfi_dynaddr_copyout(addr);
1154 		break;
1155 	case PF_ADDR_TABLE:
1156 		pf_tbladdr_copyout(addr);
1157 		break;
1158 	}
1159 }
1160 
1161 static void
1162 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
1163 {
1164 	int	secs = time_uptime, diff;
1165 
1166 	bzero(out, sizeof(struct pf_src_node));
1167 
1168 	bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
1169 	bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
1170 
1171 	if (in->rule.ptr != NULL)
1172 		out->rule.nr = in->rule.ptr->nr;
1173 
1174 	for (int i = 0; i < 2; i++) {
1175 		out->bytes[i] = counter_u64_fetch(in->bytes[i]);
1176 		out->packets[i] = counter_u64_fetch(in->packets[i]);
1177 	}
1178 
1179 	out->states = in->states;
1180 	out->conn = in->conn;
1181 	out->af = in->af;
1182 	out->ruletype = in->ruletype;
1183 
1184 	out->creation = secs - in->creation;
1185 	if (out->expire > secs)
1186 		out->expire -= secs;
1187 	else
1188 		out->expire = 0;
1189 
1190 	/* Adjust the connection rate estimate. */
1191 	diff = secs - in->conn_rate.last;
1192 	if (diff >= in->conn_rate.seconds)
1193 		out->conn_rate.count = 0;
1194 	else
1195 		out->conn_rate.count -=
1196 		    in->conn_rate.count * diff /
1197 		    in->conn_rate.seconds;
1198 }
1199 
1200 #ifdef ALTQ
1201 /*
1202  * Handle export of struct pf_kaltq to user binaries that may be using any
1203  * version of struct pf_altq.
1204  */
1205 static int
1206 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1207 {
1208 	u_int32_t version;
1209 
1210 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1211 		version = 0;
1212 	else
1213 		version = pa->version;
1214 
1215 	if (version > PFIOC_ALTQ_VERSION)
1216 		return (EINVAL);
1217 
1218 #define ASSIGN(x) exported_q->x = q->x
1219 #define COPY(x) \
1220 	bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1221 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1222 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1223 
1224 	switch (version) {
1225 	case 0: {
1226 		struct pf_altq_v0 *exported_q =
1227 		    &((struct pfioc_altq_v0 *)pa)->altq;
1228 
1229 		COPY(ifname);
1230 
1231 		ASSIGN(scheduler);
1232 		ASSIGN(tbrsize);
1233 		exported_q->tbrsize = SATU16(q->tbrsize);
1234 		exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1235 
1236 		COPY(qname);
1237 		COPY(parent);
1238 		ASSIGN(parent_qid);
1239 		exported_q->bandwidth = SATU32(q->bandwidth);
1240 		ASSIGN(priority);
1241 		ASSIGN(local_flags);
1242 
1243 		ASSIGN(qlimit);
1244 		ASSIGN(flags);
1245 
1246 		if (q->scheduler == ALTQT_HFSC) {
1247 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1248 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1249 			    SATU32(q->pq_u.hfsc_opts.x)
1250 
1251 			ASSIGN_OPT_SATU32(rtsc_m1);
1252 			ASSIGN_OPT(rtsc_d);
1253 			ASSIGN_OPT_SATU32(rtsc_m2);
1254 
1255 			ASSIGN_OPT_SATU32(lssc_m1);
1256 			ASSIGN_OPT(lssc_d);
1257 			ASSIGN_OPT_SATU32(lssc_m2);
1258 
1259 			ASSIGN_OPT_SATU32(ulsc_m1);
1260 			ASSIGN_OPT(ulsc_d);
1261 			ASSIGN_OPT_SATU32(ulsc_m2);
1262 
1263 			ASSIGN_OPT(flags);
1264 
1265 #undef ASSIGN_OPT
1266 #undef ASSIGN_OPT_SATU32
1267 		} else
1268 			COPY(pq_u);
1269 
1270 		ASSIGN(qid);
1271 		break;
1272 	}
1273 	case 1:	{
1274 		struct pf_altq_v1 *exported_q =
1275 		    &((struct pfioc_altq_v1 *)pa)->altq;
1276 
1277 		COPY(ifname);
1278 
1279 		ASSIGN(scheduler);
1280 		ASSIGN(tbrsize);
1281 		ASSIGN(ifbandwidth);
1282 
1283 		COPY(qname);
1284 		COPY(parent);
1285 		ASSIGN(parent_qid);
1286 		ASSIGN(bandwidth);
1287 		ASSIGN(priority);
1288 		ASSIGN(local_flags);
1289 
1290 		ASSIGN(qlimit);
1291 		ASSIGN(flags);
1292 		COPY(pq_u);
1293 
1294 		ASSIGN(qid);
1295 		break;
1296 	}
1297 	default:
1298 		panic("%s: unhandled struct pfioc_altq version", __func__);
1299 		break;
1300 	}
1301 
1302 #undef ASSIGN
1303 #undef COPY
1304 #undef SATU16
1305 #undef SATU32
1306 
1307 	return (0);
1308 }
1309 
1310 /*
1311  * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1312  * that may be using any version of it.
1313  */
1314 static int
1315 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1316 {
1317 	u_int32_t version;
1318 
1319 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1320 		version = 0;
1321 	else
1322 		version = pa->version;
1323 
1324 	if (version > PFIOC_ALTQ_VERSION)
1325 		return (EINVAL);
1326 
1327 #define ASSIGN(x) q->x = imported_q->x
1328 #define COPY(x) \
1329 	bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1330 
1331 	switch (version) {
1332 	case 0: {
1333 		struct pf_altq_v0 *imported_q =
1334 		    &((struct pfioc_altq_v0 *)pa)->altq;
1335 
1336 		COPY(ifname);
1337 
1338 		ASSIGN(scheduler);
1339 		ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1340 		ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1341 
1342 		COPY(qname);
1343 		COPY(parent);
1344 		ASSIGN(parent_qid);
1345 		ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1346 		ASSIGN(priority);
1347 		ASSIGN(local_flags);
1348 
1349 		ASSIGN(qlimit);
1350 		ASSIGN(flags);
1351 
1352 		if (imported_q->scheduler == ALTQT_HFSC) {
1353 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1354 
1355 			/*
1356 			 * The m1 and m2 parameters are being copied from
1357 			 * 32-bit to 64-bit.
1358 			 */
1359 			ASSIGN_OPT(rtsc_m1);
1360 			ASSIGN_OPT(rtsc_d);
1361 			ASSIGN_OPT(rtsc_m2);
1362 
1363 			ASSIGN_OPT(lssc_m1);
1364 			ASSIGN_OPT(lssc_d);
1365 			ASSIGN_OPT(lssc_m2);
1366 
1367 			ASSIGN_OPT(ulsc_m1);
1368 			ASSIGN_OPT(ulsc_d);
1369 			ASSIGN_OPT(ulsc_m2);
1370 
1371 			ASSIGN_OPT(flags);
1372 
1373 #undef ASSIGN_OPT
1374 		} else
1375 			COPY(pq_u);
1376 
1377 		ASSIGN(qid);
1378 		break;
1379 	}
1380 	case 1: {
1381 		struct pf_altq_v1 *imported_q =
1382 		    &((struct pfioc_altq_v1 *)pa)->altq;
1383 
1384 		COPY(ifname);
1385 
1386 		ASSIGN(scheduler);
1387 		ASSIGN(tbrsize);
1388 		ASSIGN(ifbandwidth);
1389 
1390 		COPY(qname);
1391 		COPY(parent);
1392 		ASSIGN(parent_qid);
1393 		ASSIGN(bandwidth);
1394 		ASSIGN(priority);
1395 		ASSIGN(local_flags);
1396 
1397 		ASSIGN(qlimit);
1398 		ASSIGN(flags);
1399 		COPY(pq_u);
1400 
1401 		ASSIGN(qid);
1402 		break;
1403 	}
1404 	default:
1405 		panic("%s: unhandled struct pfioc_altq version", __func__);
1406 		break;
1407 	}
1408 
1409 #undef ASSIGN
1410 #undef COPY
1411 
1412 	return (0);
1413 }
1414 
1415 static struct pf_altq *
1416 pf_altq_get_nth_active(u_int32_t n)
1417 {
1418 	struct pf_altq		*altq;
1419 	u_int32_t		 nr;
1420 
1421 	nr = 0;
1422 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1423 		if (nr == n)
1424 			return (altq);
1425 		nr++;
1426 	}
1427 
1428 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1429 		if (nr == n)
1430 			return (altq);
1431 		nr++;
1432 	}
1433 
1434 	return (NULL);
1435 }
1436 #endif /* ALTQ */
1437 
1438 static void
1439 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool,
1440     struct pf_pooladdr *pool)
1441 {
1442 
1443 	bzero(pool, sizeof(*pool));
1444 	bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr));
1445 	strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname));
1446 }
1447 
1448 static void
1449 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool,
1450     struct pf_kpooladdr *kpool)
1451 {
1452 
1453 	bzero(kpool, sizeof(*kpool));
1454 	bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr));
1455 	strlcpy(kpool->ifname, pool->ifname, sizeof(kpool->ifname));
1456 }
1457 
1458 static void
1459 pf_krule_to_rule(const struct pf_krule *krule, struct pf_rule *rule)
1460 {
1461 
1462 	bzero(rule, sizeof(*rule));
1463 
1464 	bcopy(&krule->src, &rule->src, sizeof(rule->src));
1465 	bcopy(&krule->dst, &rule->dst, sizeof(rule->dst));
1466 
1467 	for (int i = 0; i < PF_SKIP_COUNT; ++i) {
1468 		if (rule->skip[i].ptr == NULL)
1469 			rule->skip[i].nr = -1;
1470 		else
1471 			rule->skip[i].nr = krule->skip[i].ptr->nr;
1472 	}
1473 
1474 	strlcpy(rule->label, krule->label, sizeof(rule->label));
1475 	strlcpy(rule->ifname, krule->ifname, sizeof(rule->ifname));
1476 	strlcpy(rule->qname, krule->qname, sizeof(rule->qname));
1477 	strlcpy(rule->pqname, krule->pqname, sizeof(rule->pqname));
1478 	strlcpy(rule->tagname, krule->tagname, sizeof(rule->tagname));
1479 	strlcpy(rule->match_tagname, krule->match_tagname,
1480 	    sizeof(rule->match_tagname));
1481 	strlcpy(rule->overload_tblname, krule->overload_tblname,
1482 	    sizeof(rule->overload_tblname));
1483 
1484 	bcopy(&krule->rpool, &rule->rpool, sizeof(krule->rpool));
1485 
1486 	rule->evaluations = counter_u64_fetch(krule->evaluations);
1487 	for (int i = 0; i < 2; i++) {
1488 		rule->packets[i] = counter_u64_fetch(krule->packets[i]);
1489 		rule->bytes[i] = counter_u64_fetch(krule->bytes[i]);
1490 	}
1491 
1492 	/* kif, anchor, overload_tbl are not copied over. */
1493 
1494 	rule->os_fingerprint = krule->os_fingerprint;
1495 
1496 	rule->rtableid = krule->rtableid;
1497 	bcopy(krule->timeout, rule->timeout, sizeof(krule->timeout));
1498 	rule->max_states = krule->max_states;
1499 	rule->max_src_nodes = krule->max_src_nodes;
1500 	rule->max_src_states = krule->max_src_states;
1501 	rule->max_src_conn = krule->max_src_conn;
1502 	rule->max_src_conn_rate.limit = krule->max_src_conn_rate.limit;
1503 	rule->max_src_conn_rate.seconds = krule->max_src_conn_rate.seconds;
1504 	rule->qid = krule->qid;
1505 	rule->pqid = krule->pqid;
1506 	rule->rt_listid = krule->rt_listid;
1507 	rule->nr = krule->nr;
1508 	rule->prob = krule->prob;
1509 	rule->cuid = krule->cuid;
1510 	rule->cpid = krule->cpid;
1511 
1512 	rule->return_icmp = krule->return_icmp;
1513 	rule->return_icmp6 = krule->return_icmp6;
1514 	rule->max_mss = krule->max_mss;
1515 	rule->tag = krule->tag;
1516 	rule->match_tag = krule->match_tag;
1517 	rule->scrub_flags = krule->scrub_flags;
1518 
1519 	bcopy(&krule->uid, &rule->uid, sizeof(krule->uid));
1520 	bcopy(&krule->gid, &rule->gid, sizeof(krule->gid));
1521 
1522 	rule->rule_flag = krule->rule_flag;
1523 	rule->action = krule->action;
1524 	rule->direction = krule->direction;
1525 	rule->log = krule->log;
1526 	rule->logif = krule->logif;
1527 	rule->quick = krule->quick;
1528 	rule->ifnot = krule->ifnot;
1529 	rule->match_tag_not = krule->match_tag_not;
1530 	rule->natpass = krule->natpass;
1531 
1532 	rule->keep_state = krule->keep_state;
1533 	rule->af = krule->af;
1534 	rule->proto = krule->proto;
1535 	rule->type = krule->type;
1536 	rule->code = krule->code;
1537 	rule->flags = krule->flags;
1538 	rule->flagset = krule->flagset;
1539 	rule->min_ttl = krule->min_ttl;
1540 	rule->allow_opts = krule->allow_opts;
1541 	rule->rt = krule->rt;
1542 	rule->return_ttl = krule->return_ttl;
1543 	rule->tos = krule->tos;
1544 	rule->set_tos = krule->set_tos;
1545 	rule->anchor_relative = krule->anchor_relative;
1546 	rule->anchor_wildcard = krule->anchor_wildcard;
1547 
1548 	rule->flush = krule->flush;
1549 	rule->prio = krule->prio;
1550 	rule->set_prio[0] = krule->set_prio[0];
1551 	rule->set_prio[1] = krule->set_prio[1];
1552 
1553 	bcopy(&krule->divert, &rule->divert, sizeof(krule->divert));
1554 
1555 	rule->u_states_cur = counter_u64_fetch(krule->states_cur);
1556 	rule->u_states_tot = counter_u64_fetch(krule->states_tot);
1557 	rule->u_src_nodes = counter_u64_fetch(krule->src_nodes);
1558 }
1559 
1560 static int
1561 pf_check_rule_addr(const struct pf_rule_addr *addr)
1562 {
1563 
1564 	switch (addr->addr.type) {
1565 	case PF_ADDR_ADDRMASK:
1566 	case PF_ADDR_NOROUTE:
1567 	case PF_ADDR_DYNIFTL:
1568 	case PF_ADDR_TABLE:
1569 	case PF_ADDR_URPFFAILED:
1570 	case PF_ADDR_RANGE:
1571 		break;
1572 	default:
1573 		return (EINVAL);
1574 	}
1575 
1576 	if (addr->addr.p.dyn != NULL) {
1577 		return (EINVAL);
1578 	}
1579 
1580 	return (0);
1581 }
1582 
1583 static int
1584 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
1585 {
1586 	int ret;
1587 
1588 #ifndef INET
1589 	if (rule->af == AF_INET) {
1590 		return (EAFNOSUPPORT);
1591 	}
1592 #endif /* INET */
1593 #ifndef INET6
1594 	if (rule->af == AF_INET6) {
1595 		return (EAFNOSUPPORT);
1596 	}
1597 #endif /* INET6 */
1598 
1599 	ret = pf_check_rule_addr(&rule->src);
1600 	if (ret != 0)
1601 		return (ret);
1602 	ret = pf_check_rule_addr(&rule->dst);
1603 	if (ret != 0)
1604 		return (ret);
1605 
1606 	bzero(krule, sizeof(*krule));
1607 
1608 	bcopy(&rule->src, &krule->src, sizeof(rule->src));
1609 	bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
1610 
1611 	strlcpy(krule->label, rule->label, sizeof(rule->label));
1612 	strlcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
1613 	strlcpy(krule->qname, rule->qname, sizeof(rule->qname));
1614 	strlcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
1615 	strlcpy(krule->tagname, rule->tagname, sizeof(rule->tagname));
1616 	strlcpy(krule->match_tagname, rule->match_tagname,
1617 	    sizeof(rule->match_tagname));
1618 	strlcpy(krule->overload_tblname, rule->overload_tblname,
1619 	    sizeof(rule->overload_tblname));
1620 
1621 	bcopy(&rule->rpool, &krule->rpool, sizeof(krule->rpool));
1622 
1623 	/* Don't allow userspace to set evaulations, packets or bytes. */
1624 	/* kif, anchor, overload_tbl are not copied over. */
1625 
1626 	krule->os_fingerprint = krule->os_fingerprint;
1627 
1628 	krule->rtableid = rule->rtableid;
1629 	bcopy(rule->timeout, krule->timeout, sizeof(krule->timeout));
1630 	krule->max_states = rule->max_states;
1631 	krule->max_src_nodes = rule->max_src_nodes;
1632 	krule->max_src_states = rule->max_src_states;
1633 	krule->max_src_conn = rule->max_src_conn;
1634 	krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
1635 	krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
1636 	krule->qid = rule->qid;
1637 	krule->pqid = rule->pqid;
1638 	krule->rt_listid = rule->rt_listid;
1639 	krule->nr = rule->nr;
1640 	krule->prob = rule->prob;
1641 	krule->cuid = rule->cuid;
1642 	krule->cpid = rule->cpid;
1643 
1644 	krule->return_icmp = rule->return_icmp;
1645 	krule->return_icmp6 = rule->return_icmp6;
1646 	krule->max_mss = rule->max_mss;
1647 	krule->tag = rule->tag;
1648 	krule->match_tag = rule->match_tag;
1649 	krule->scrub_flags = rule->scrub_flags;
1650 
1651 	bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
1652 	bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
1653 
1654 	krule->rule_flag = rule->rule_flag;
1655 	krule->action = rule->action;
1656 	krule->direction = rule->direction;
1657 	krule->log = rule->log;
1658 	krule->logif = rule->logif;
1659 	krule->quick = rule->quick;
1660 	krule->ifnot = rule->ifnot;
1661 	krule->match_tag_not = rule->match_tag_not;
1662 	krule->natpass = rule->natpass;
1663 
1664 	krule->keep_state = rule->keep_state;
1665 	krule->af = rule->af;
1666 	krule->proto = rule->proto;
1667 	krule->type = rule->type;
1668 	krule->code = rule->code;
1669 	krule->flags = rule->flags;
1670 	krule->flagset = rule->flagset;
1671 	krule->min_ttl = rule->min_ttl;
1672 	krule->allow_opts = rule->allow_opts;
1673 	krule->rt = rule->rt;
1674 	krule->return_ttl = rule->return_ttl;
1675 	krule->tos = rule->tos;
1676 	krule->set_tos = rule->set_tos;
1677 	krule->anchor_relative = rule->anchor_relative;
1678 	krule->anchor_wildcard = rule->anchor_wildcard;
1679 
1680 	krule->flush = rule->flush;
1681 	krule->prio = rule->prio;
1682 	krule->set_prio[0] = rule->set_prio[0];
1683 	krule->set_prio[1] = rule->set_prio[1];
1684 
1685 	bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
1686 
1687 	return (0);
1688 }
1689 
1690 static int
1691 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
1692 {
1693 	int			 error = 0;
1694 	PF_RULES_RLOCK_TRACKER;
1695 
1696 	/* XXX keep in sync with switch() below */
1697 	if (securelevel_gt(td->td_ucred, 2))
1698 		switch (cmd) {
1699 		case DIOCGETRULES:
1700 		case DIOCGETRULE:
1701 		case DIOCGETADDRS:
1702 		case DIOCGETADDR:
1703 		case DIOCGETSTATE:
1704 		case DIOCSETSTATUSIF:
1705 		case DIOCGETSTATUS:
1706 		case DIOCCLRSTATUS:
1707 		case DIOCNATLOOK:
1708 		case DIOCSETDEBUG:
1709 		case DIOCGETSTATES:
1710 		case DIOCGETTIMEOUT:
1711 		case DIOCCLRRULECTRS:
1712 		case DIOCGETLIMIT:
1713 		case DIOCGETALTQSV0:
1714 		case DIOCGETALTQSV1:
1715 		case DIOCGETALTQV0:
1716 		case DIOCGETALTQV1:
1717 		case DIOCGETQSTATSV0:
1718 		case DIOCGETQSTATSV1:
1719 		case DIOCGETRULESETS:
1720 		case DIOCGETRULESET:
1721 		case DIOCRGETTABLES:
1722 		case DIOCRGETTSTATS:
1723 		case DIOCRCLRTSTATS:
1724 		case DIOCRCLRADDRS:
1725 		case DIOCRADDADDRS:
1726 		case DIOCRDELADDRS:
1727 		case DIOCRSETADDRS:
1728 		case DIOCRGETADDRS:
1729 		case DIOCRGETASTATS:
1730 		case DIOCRCLRASTATS:
1731 		case DIOCRTSTADDRS:
1732 		case DIOCOSFPGET:
1733 		case DIOCGETSRCNODES:
1734 		case DIOCCLRSRCNODES:
1735 		case DIOCIGETIFACES:
1736 		case DIOCGIFSPEEDV0:
1737 		case DIOCGIFSPEEDV1:
1738 		case DIOCSETIFFLAG:
1739 		case DIOCCLRIFFLAG:
1740 			break;
1741 		case DIOCRCLRTABLES:
1742 		case DIOCRADDTABLES:
1743 		case DIOCRDELTABLES:
1744 		case DIOCRSETTFLAGS:
1745 			if (((struct pfioc_table *)addr)->pfrio_flags &
1746 			    PFR_FLAG_DUMMY)
1747 				break; /* dummy operation ok */
1748 			return (EPERM);
1749 		default:
1750 			return (EPERM);
1751 		}
1752 
1753 	if (!(flags & FWRITE))
1754 		switch (cmd) {
1755 		case DIOCGETRULES:
1756 		case DIOCGETADDRS:
1757 		case DIOCGETADDR:
1758 		case DIOCGETSTATE:
1759 		case DIOCGETSTATUS:
1760 		case DIOCGETSTATES:
1761 		case DIOCGETTIMEOUT:
1762 		case DIOCGETLIMIT:
1763 		case DIOCGETALTQSV0:
1764 		case DIOCGETALTQSV1:
1765 		case DIOCGETALTQV0:
1766 		case DIOCGETALTQV1:
1767 		case DIOCGETQSTATSV0:
1768 		case DIOCGETQSTATSV1:
1769 		case DIOCGETRULESETS:
1770 		case DIOCGETRULESET:
1771 		case DIOCNATLOOK:
1772 		case DIOCRGETTABLES:
1773 		case DIOCRGETTSTATS:
1774 		case DIOCRGETADDRS:
1775 		case DIOCRGETASTATS:
1776 		case DIOCRTSTADDRS:
1777 		case DIOCOSFPGET:
1778 		case DIOCGETSRCNODES:
1779 		case DIOCIGETIFACES:
1780 		case DIOCGIFSPEEDV1:
1781 		case DIOCGIFSPEEDV0:
1782 			break;
1783 		case DIOCRCLRTABLES:
1784 		case DIOCRADDTABLES:
1785 		case DIOCRDELTABLES:
1786 		case DIOCRCLRTSTATS:
1787 		case DIOCRCLRADDRS:
1788 		case DIOCRADDADDRS:
1789 		case DIOCRDELADDRS:
1790 		case DIOCRSETADDRS:
1791 		case DIOCRSETTFLAGS:
1792 			if (((struct pfioc_table *)addr)->pfrio_flags &
1793 			    PFR_FLAG_DUMMY) {
1794 				flags |= FWRITE; /* need write lock for dummy */
1795 				break; /* dummy operation ok */
1796 			}
1797 			return (EACCES);
1798 		case DIOCGETRULE:
1799 			if (((struct pfioc_rule *)addr)->action ==
1800 			    PF_GET_CLR_CNTR)
1801 				return (EACCES);
1802 			break;
1803 		default:
1804 			return (EACCES);
1805 		}
1806 
1807 	CURVNET_SET(TD_TO_VNET(td));
1808 
1809 	switch (cmd) {
1810 	case DIOCSTART:
1811 		sx_xlock(&pf_ioctl_lock);
1812 		if (V_pf_status.running)
1813 			error = EEXIST;
1814 		else {
1815 			int cpu;
1816 
1817 			error = hook_pf();
1818 			if (error) {
1819 				DPFPRINTF(PF_DEBUG_MISC,
1820 				    ("pf: pfil registration failed\n"));
1821 				break;
1822 			}
1823 			V_pf_status.running = 1;
1824 			V_pf_status.since = time_second;
1825 
1826 			CPU_FOREACH(cpu)
1827 				V_pf_stateid[cpu] = time_second;
1828 
1829 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1830 		}
1831 		break;
1832 
1833 	case DIOCSTOP:
1834 		sx_xlock(&pf_ioctl_lock);
1835 		if (!V_pf_status.running)
1836 			error = ENOENT;
1837 		else {
1838 			V_pf_status.running = 0;
1839 			error = dehook_pf();
1840 			if (error) {
1841 				V_pf_status.running = 1;
1842 				DPFPRINTF(PF_DEBUG_MISC,
1843 				    ("pf: pfil unregistration failed\n"));
1844 			}
1845 			V_pf_status.since = time_second;
1846 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1847 		}
1848 		break;
1849 
1850 	case DIOCADDRULE: {
1851 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1852 		struct pf_kruleset	*ruleset;
1853 		struct pf_krule		*rule, *tail;
1854 		struct pf_kpooladdr	*pa;
1855 		struct pfi_kkif		*kif = NULL;
1856 		int			 rs_num;
1857 
1858 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1859 			error = EINVAL;
1860 			break;
1861 		}
1862 
1863 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
1864 		error = pf_rule_to_krule(&pr->rule, rule);
1865 		if (error != 0) {
1866 			free(rule, M_PFRULE);
1867 			break;
1868 		}
1869 
1870 		if (rule->ifname[0])
1871 			kif = pf_kkif_create(M_WAITOK);
1872 		rule->evaluations = counter_u64_alloc(M_WAITOK);
1873 		for (int i = 0; i < 2; i++) {
1874 			rule->packets[i] = counter_u64_alloc(M_WAITOK);
1875 			rule->bytes[i] = counter_u64_alloc(M_WAITOK);
1876 		}
1877 		rule->states_cur = counter_u64_alloc(M_WAITOK);
1878 		rule->states_tot = counter_u64_alloc(M_WAITOK);
1879 		rule->src_nodes = counter_u64_alloc(M_WAITOK);
1880 		rule->cuid = td->td_ucred->cr_ruid;
1881 		rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1882 		TAILQ_INIT(&rule->rpool.list);
1883 #define	ERROUT(x)	{ error = (x); goto DIOCADDRULE_error; }
1884 
1885 		PF_RULES_WLOCK();
1886 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1887 		ruleset = pf_find_kruleset(pr->anchor);
1888 		if (ruleset == NULL)
1889 			ERROUT(EINVAL);
1890 		rs_num = pf_get_ruleset_number(pr->rule.action);
1891 		if (rs_num >= PF_RULESET_MAX)
1892 			ERROUT(EINVAL);
1893 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1894 			DPFPRINTF(PF_DEBUG_MISC,
1895 			    ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
1896 			    ruleset->rules[rs_num].inactive.ticket));
1897 			ERROUT(EBUSY);
1898 		}
1899 		if (pr->pool_ticket != V_ticket_pabuf) {
1900 			DPFPRINTF(PF_DEBUG_MISC,
1901 			    ("pool_ticket: %d != %d\n", pr->pool_ticket,
1902 			    V_ticket_pabuf));
1903 			ERROUT(EBUSY);
1904 		}
1905 
1906 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1907 		    pf_krulequeue);
1908 		if (tail)
1909 			rule->nr = tail->nr + 1;
1910 		else
1911 			rule->nr = 0;
1912 		if (rule->ifname[0]) {
1913 			rule->kif = pfi_kkif_attach(kif, rule->ifname);
1914 			pfi_kkif_ref(rule->kif);
1915 		} else
1916 			rule->kif = NULL;
1917 
1918 		if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
1919 			error = EBUSY;
1920 
1921 #ifdef ALTQ
1922 		/* set queue IDs */
1923 		if (rule->qname[0] != 0) {
1924 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1925 				error = EBUSY;
1926 			else if (rule->pqname[0] != 0) {
1927 				if ((rule->pqid =
1928 				    pf_qname2qid(rule->pqname)) == 0)
1929 					error = EBUSY;
1930 			} else
1931 				rule->pqid = rule->qid;
1932 		}
1933 #endif
1934 		if (rule->tagname[0])
1935 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1936 				error = EBUSY;
1937 		if (rule->match_tagname[0])
1938 			if ((rule->match_tag =
1939 			    pf_tagname2tag(rule->match_tagname)) == 0)
1940 				error = EBUSY;
1941 		if (rule->rt && !rule->direction)
1942 			error = EINVAL;
1943 		if (!rule->log)
1944 			rule->logif = 0;
1945 		if (rule->logif >= PFLOGIFS_MAX)
1946 			error = EINVAL;
1947 		if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1948 			error = ENOMEM;
1949 		if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1950 			error = ENOMEM;
1951 		if (pf_kanchor_setup(rule, ruleset, pr->anchor_call))
1952 			error = EINVAL;
1953 		if (rule->scrub_flags & PFSTATE_SETPRIO &&
1954 		    (rule->set_prio[0] > PF_PRIO_MAX ||
1955 		    rule->set_prio[1] > PF_PRIO_MAX))
1956 			error = EINVAL;
1957 		TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1958 			if (pa->addr.type == PF_ADDR_TABLE) {
1959 				pa->addr.p.tbl = pfr_attach_table(ruleset,
1960 				    pa->addr.v.tblname);
1961 				if (pa->addr.p.tbl == NULL)
1962 					error = ENOMEM;
1963 			}
1964 
1965 		rule->overload_tbl = NULL;
1966 		if (rule->overload_tblname[0]) {
1967 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1968 			    rule->overload_tblname)) == NULL)
1969 				error = EINVAL;
1970 			else
1971 				rule->overload_tbl->pfrkt_flags |=
1972 				    PFR_TFLAG_ACTIVE;
1973 		}
1974 
1975 		pf_mv_kpool(&V_pf_pabuf, &rule->rpool.list);
1976 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1977 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1978 		    (rule->rt > PF_NOPFROUTE)) &&
1979 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1980 			error = EINVAL;
1981 
1982 		if (error) {
1983 			pf_free_rule(rule);
1984 			PF_RULES_WUNLOCK();
1985 			break;
1986 		}
1987 
1988 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1989 		counter_u64_zero(rule->evaluations);
1990 		for (int i = 0; i < 2; i++) {
1991 			counter_u64_zero(rule->packets[i]);
1992 			counter_u64_zero(rule->bytes[i]);
1993 		}
1994 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1995 		    rule, entries);
1996 		ruleset->rules[rs_num].inactive.rcount++;
1997 		PF_RULES_WUNLOCK();
1998 		break;
1999 
2000 #undef ERROUT
2001 DIOCADDRULE_error:
2002 		PF_RULES_WUNLOCK();
2003 		counter_u64_free(rule->evaluations);
2004 		for (int i = 0; i < 2; i++) {
2005 			counter_u64_free(rule->packets[i]);
2006 			counter_u64_free(rule->bytes[i]);
2007 		}
2008 		counter_u64_free(rule->states_cur);
2009 		counter_u64_free(rule->states_tot);
2010 		counter_u64_free(rule->src_nodes);
2011 		free(rule, M_PFRULE);
2012 		if (kif)
2013 			pf_kkif_free(kif);
2014 		break;
2015 	}
2016 
2017 	case DIOCGETRULES: {
2018 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
2019 		struct pf_kruleset	*ruleset;
2020 		struct pf_krule		*tail;
2021 		int			 rs_num;
2022 
2023 		PF_RULES_WLOCK();
2024 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
2025 		ruleset = pf_find_kruleset(pr->anchor);
2026 		if (ruleset == NULL) {
2027 			PF_RULES_WUNLOCK();
2028 			error = EINVAL;
2029 			break;
2030 		}
2031 		rs_num = pf_get_ruleset_number(pr->rule.action);
2032 		if (rs_num >= PF_RULESET_MAX) {
2033 			PF_RULES_WUNLOCK();
2034 			error = EINVAL;
2035 			break;
2036 		}
2037 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
2038 		    pf_krulequeue);
2039 		if (tail)
2040 			pr->nr = tail->nr + 1;
2041 		else
2042 			pr->nr = 0;
2043 		pr->ticket = ruleset->rules[rs_num].active.ticket;
2044 		PF_RULES_WUNLOCK();
2045 		break;
2046 	}
2047 
2048 	case DIOCGETRULE: {
2049 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
2050 		struct pf_kruleset	*ruleset;
2051 		struct pf_krule		*rule;
2052 		int			 rs_num;
2053 
2054 		PF_RULES_WLOCK();
2055 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
2056 		ruleset = pf_find_kruleset(pr->anchor);
2057 		if (ruleset == NULL) {
2058 			PF_RULES_WUNLOCK();
2059 			error = EINVAL;
2060 			break;
2061 		}
2062 		rs_num = pf_get_ruleset_number(pr->rule.action);
2063 		if (rs_num >= PF_RULESET_MAX) {
2064 			PF_RULES_WUNLOCK();
2065 			error = EINVAL;
2066 			break;
2067 		}
2068 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
2069 			PF_RULES_WUNLOCK();
2070 			error = EBUSY;
2071 			break;
2072 		}
2073 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
2074 		while ((rule != NULL) && (rule->nr != pr->nr))
2075 			rule = TAILQ_NEXT(rule, entries);
2076 		if (rule == NULL) {
2077 			PF_RULES_WUNLOCK();
2078 			error = EBUSY;
2079 			break;
2080 		}
2081 
2082 		pf_krule_to_rule(rule, &pr->rule);
2083 
2084 		if (pf_kanchor_copyout(ruleset, rule, pr)) {
2085 			PF_RULES_WUNLOCK();
2086 			error = EBUSY;
2087 			break;
2088 		}
2089 		pf_addr_copyout(&pr->rule.src.addr);
2090 		pf_addr_copyout(&pr->rule.dst.addr);
2091 
2092 		if (pr->action == PF_GET_CLR_CNTR) {
2093 			counter_u64_zero(rule->evaluations);
2094 			for (int i = 0; i < 2; i++) {
2095 				counter_u64_zero(rule->packets[i]);
2096 				counter_u64_zero(rule->bytes[i]);
2097 			}
2098 			counter_u64_zero(rule->states_tot);
2099 		}
2100 		PF_RULES_WUNLOCK();
2101 		break;
2102 	}
2103 
2104 	case DIOCCHANGERULE: {
2105 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
2106 		struct pf_kruleset	*ruleset;
2107 		struct pf_krule		*oldrule = NULL, *newrule = NULL;
2108 		struct pfi_kkif		*kif = NULL;
2109 		struct pf_kpooladdr	*pa;
2110 		u_int32_t		 nr = 0;
2111 		int			 rs_num;
2112 
2113 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
2114 		    pcr->action > PF_CHANGE_GET_TICKET) {
2115 			error = EINVAL;
2116 			break;
2117 		}
2118 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
2119 			error = EINVAL;
2120 			break;
2121 		}
2122 
2123 		if (pcr->action != PF_CHANGE_REMOVE) {
2124 			newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
2125 			error = pf_rule_to_krule(&pcr->rule, newrule);
2126 			if (error != 0) {
2127 				free(newrule, M_PFRULE);
2128 				break;
2129 			}
2130 
2131 			if (newrule->ifname[0])
2132 				kif = pf_kkif_create(M_WAITOK);
2133 			newrule->evaluations = counter_u64_alloc(M_WAITOK);
2134 			for (int i = 0; i < 2; i++) {
2135 				newrule->packets[i] =
2136 				    counter_u64_alloc(M_WAITOK);
2137 				newrule->bytes[i] =
2138 				    counter_u64_alloc(M_WAITOK);
2139 			}
2140 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
2141 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
2142 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
2143 			newrule->cuid = td->td_ucred->cr_ruid;
2144 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
2145 			TAILQ_INIT(&newrule->rpool.list);
2146 		}
2147 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGERULE_error; }
2148 
2149 		PF_RULES_WLOCK();
2150 		if (!(pcr->action == PF_CHANGE_REMOVE ||
2151 		    pcr->action == PF_CHANGE_GET_TICKET) &&
2152 		    pcr->pool_ticket != V_ticket_pabuf)
2153 			ERROUT(EBUSY);
2154 
2155 		ruleset = pf_find_kruleset(pcr->anchor);
2156 		if (ruleset == NULL)
2157 			ERROUT(EINVAL);
2158 
2159 		rs_num = pf_get_ruleset_number(pcr->rule.action);
2160 		if (rs_num >= PF_RULESET_MAX)
2161 			ERROUT(EINVAL);
2162 
2163 		if (pcr->action == PF_CHANGE_GET_TICKET) {
2164 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
2165 			ERROUT(0);
2166 		} else if (pcr->ticket !=
2167 			    ruleset->rules[rs_num].active.ticket)
2168 				ERROUT(EINVAL);
2169 
2170 		if (pcr->action != PF_CHANGE_REMOVE) {
2171 			if (newrule->ifname[0]) {
2172 				newrule->kif = pfi_kkif_attach(kif,
2173 				    newrule->ifname);
2174 				pfi_kkif_ref(newrule->kif);
2175 			} else
2176 				newrule->kif = NULL;
2177 
2178 			if (newrule->rtableid > 0 &&
2179 			    newrule->rtableid >= rt_numfibs)
2180 				error = EBUSY;
2181 
2182 #ifdef ALTQ
2183 			/* set queue IDs */
2184 			if (newrule->qname[0] != 0) {
2185 				if ((newrule->qid =
2186 				    pf_qname2qid(newrule->qname)) == 0)
2187 					error = EBUSY;
2188 				else if (newrule->pqname[0] != 0) {
2189 					if ((newrule->pqid =
2190 					    pf_qname2qid(newrule->pqname)) == 0)
2191 						error = EBUSY;
2192 				} else
2193 					newrule->pqid = newrule->qid;
2194 			}
2195 #endif /* ALTQ */
2196 			if (newrule->tagname[0])
2197 				if ((newrule->tag =
2198 				    pf_tagname2tag(newrule->tagname)) == 0)
2199 					error = EBUSY;
2200 			if (newrule->match_tagname[0])
2201 				if ((newrule->match_tag = pf_tagname2tag(
2202 				    newrule->match_tagname)) == 0)
2203 					error = EBUSY;
2204 			if (newrule->rt && !newrule->direction)
2205 				error = EINVAL;
2206 			if (!newrule->log)
2207 				newrule->logif = 0;
2208 			if (newrule->logif >= PFLOGIFS_MAX)
2209 				error = EINVAL;
2210 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
2211 				error = ENOMEM;
2212 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
2213 				error = ENOMEM;
2214 			if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
2215 				error = EINVAL;
2216 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
2217 				if (pa->addr.type == PF_ADDR_TABLE) {
2218 					pa->addr.p.tbl =
2219 					    pfr_attach_table(ruleset,
2220 					    pa->addr.v.tblname);
2221 					if (pa->addr.p.tbl == NULL)
2222 						error = ENOMEM;
2223 				}
2224 
2225 			newrule->overload_tbl = NULL;
2226 			if (newrule->overload_tblname[0]) {
2227 				if ((newrule->overload_tbl = pfr_attach_table(
2228 				    ruleset, newrule->overload_tblname)) ==
2229 				    NULL)
2230 					error = EINVAL;
2231 				else
2232 					newrule->overload_tbl->pfrkt_flags |=
2233 					    PFR_TFLAG_ACTIVE;
2234 			}
2235 
2236 			pf_mv_kpool(&V_pf_pabuf, &newrule->rpool.list);
2237 			if (((((newrule->action == PF_NAT) ||
2238 			    (newrule->action == PF_RDR) ||
2239 			    (newrule->action == PF_BINAT) ||
2240 			    (newrule->rt > PF_NOPFROUTE)) &&
2241 			    !newrule->anchor)) &&
2242 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
2243 				error = EINVAL;
2244 
2245 			if (error) {
2246 				pf_free_rule(newrule);
2247 				PF_RULES_WUNLOCK();
2248 				break;
2249 			}
2250 
2251 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
2252 		}
2253 		pf_empty_kpool(&V_pf_pabuf);
2254 
2255 		if (pcr->action == PF_CHANGE_ADD_HEAD)
2256 			oldrule = TAILQ_FIRST(
2257 			    ruleset->rules[rs_num].active.ptr);
2258 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
2259 			oldrule = TAILQ_LAST(
2260 			    ruleset->rules[rs_num].active.ptr, pf_krulequeue);
2261 		else {
2262 			oldrule = TAILQ_FIRST(
2263 			    ruleset->rules[rs_num].active.ptr);
2264 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
2265 				oldrule = TAILQ_NEXT(oldrule, entries);
2266 			if (oldrule == NULL) {
2267 				if (newrule != NULL)
2268 					pf_free_rule(newrule);
2269 				PF_RULES_WUNLOCK();
2270 				error = EINVAL;
2271 				break;
2272 			}
2273 		}
2274 
2275 		if (pcr->action == PF_CHANGE_REMOVE) {
2276 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
2277 			    oldrule);
2278 			ruleset->rules[rs_num].active.rcount--;
2279 		} else {
2280 			if (oldrule == NULL)
2281 				TAILQ_INSERT_TAIL(
2282 				    ruleset->rules[rs_num].active.ptr,
2283 				    newrule, entries);
2284 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
2285 			    pcr->action == PF_CHANGE_ADD_BEFORE)
2286 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
2287 			else
2288 				TAILQ_INSERT_AFTER(
2289 				    ruleset->rules[rs_num].active.ptr,
2290 				    oldrule, newrule, entries);
2291 			ruleset->rules[rs_num].active.rcount++;
2292 		}
2293 
2294 		nr = 0;
2295 		TAILQ_FOREACH(oldrule,
2296 		    ruleset->rules[rs_num].active.ptr, entries)
2297 			oldrule->nr = nr++;
2298 
2299 		ruleset->rules[rs_num].active.ticket++;
2300 
2301 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
2302 		pf_remove_if_empty_kruleset(ruleset);
2303 
2304 		PF_RULES_WUNLOCK();
2305 		break;
2306 
2307 #undef ERROUT
2308 DIOCCHANGERULE_error:
2309 		PF_RULES_WUNLOCK();
2310 		if (newrule != NULL) {
2311 			counter_u64_free(newrule->evaluations);
2312 			for (int i = 0; i < 2; i++) {
2313 				counter_u64_free(newrule->packets[i]);
2314 				counter_u64_free(newrule->bytes[i]);
2315 			}
2316 			counter_u64_free(newrule->states_cur);
2317 			counter_u64_free(newrule->states_tot);
2318 			counter_u64_free(newrule->src_nodes);
2319 			free(newrule, M_PFRULE);
2320 		}
2321 		if (kif != NULL)
2322 			pf_kkif_free(kif);
2323 		break;
2324 	}
2325 
2326 	case DIOCCLRSTATES: {
2327 		struct pf_state		*s;
2328 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
2329 		u_int			 i, killed = 0;
2330 
2331 		for (i = 0; i <= pf_hashmask; i++) {
2332 			struct pf_idhash *ih = &V_pf_idhash[i];
2333 
2334 relock_DIOCCLRSTATES:
2335 			PF_HASHROW_LOCK(ih);
2336 			LIST_FOREACH(s, &ih->states, entry)
2337 				if (!psk->psk_ifname[0] ||
2338 				    !strcmp(psk->psk_ifname,
2339 				    s->kif->pfik_name)) {
2340 					/*
2341 					 * Don't send out individual
2342 					 * delete messages.
2343 					 */
2344 					s->state_flags |= PFSTATE_NOSYNC;
2345 					pf_unlink_state(s, PF_ENTER_LOCKED);
2346 					killed++;
2347 					goto relock_DIOCCLRSTATES;
2348 				}
2349 			PF_HASHROW_UNLOCK(ih);
2350 		}
2351 		psk->psk_killed = killed;
2352 		if (V_pfsync_clear_states_ptr != NULL)
2353 			V_pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
2354 		break;
2355 	}
2356 
2357 	case DIOCKILLSTATES: {
2358 		struct pf_state		*s;
2359 		struct pf_state_key	*sk;
2360 		struct pf_addr		*srcaddr, *dstaddr;
2361 		u_int16_t		 srcport, dstport;
2362 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
2363 		u_int			 i, killed = 0;
2364 
2365 		if (psk->psk_pfcmp.id) {
2366 			if (psk->psk_pfcmp.creatorid == 0)
2367 				psk->psk_pfcmp.creatorid = V_pf_status.hostid;
2368 			if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
2369 			    psk->psk_pfcmp.creatorid))) {
2370 				pf_unlink_state(s, PF_ENTER_LOCKED);
2371 				psk->psk_killed = 1;
2372 			}
2373 			break;
2374 		}
2375 
2376 		for (i = 0; i <= pf_hashmask; i++) {
2377 			struct pf_idhash *ih = &V_pf_idhash[i];
2378 
2379 relock_DIOCKILLSTATES:
2380 			PF_HASHROW_LOCK(ih);
2381 			LIST_FOREACH(s, &ih->states, entry) {
2382 				sk = s->key[PF_SK_WIRE];
2383 				if (s->direction == PF_OUT) {
2384 					srcaddr = &sk->addr[1];
2385 					dstaddr = &sk->addr[0];
2386 					srcport = sk->port[1];
2387 					dstport = sk->port[0];
2388 				} else {
2389 					srcaddr = &sk->addr[0];
2390 					dstaddr = &sk->addr[1];
2391 					srcport = sk->port[0];
2392 					dstport = sk->port[1];
2393 				}
2394 
2395 				if ((!psk->psk_af || sk->af == psk->psk_af)
2396 				    && (!psk->psk_proto || psk->psk_proto ==
2397 				    sk->proto) &&
2398 				    PF_MATCHA(psk->psk_src.neg,
2399 				    &psk->psk_src.addr.v.a.addr,
2400 				    &psk->psk_src.addr.v.a.mask,
2401 				    srcaddr, sk->af) &&
2402 				    PF_MATCHA(psk->psk_dst.neg,
2403 				    &psk->psk_dst.addr.v.a.addr,
2404 				    &psk->psk_dst.addr.v.a.mask,
2405 				    dstaddr, sk->af) &&
2406 				    (psk->psk_src.port_op == 0 ||
2407 				    pf_match_port(psk->psk_src.port_op,
2408 				    psk->psk_src.port[0], psk->psk_src.port[1],
2409 				    srcport)) &&
2410 				    (psk->psk_dst.port_op == 0 ||
2411 				    pf_match_port(psk->psk_dst.port_op,
2412 				    psk->psk_dst.port[0], psk->psk_dst.port[1],
2413 				    dstport)) &&
2414 				    (!psk->psk_label[0] ||
2415 				    (s->rule.ptr->label[0] &&
2416 				    !strcmp(psk->psk_label,
2417 				    s->rule.ptr->label))) &&
2418 				    (!psk->psk_ifname[0] ||
2419 				    !strcmp(psk->psk_ifname,
2420 				    s->kif->pfik_name))) {
2421 					pf_unlink_state(s, PF_ENTER_LOCKED);
2422 					killed++;
2423 					goto relock_DIOCKILLSTATES;
2424 				}
2425 			}
2426 			PF_HASHROW_UNLOCK(ih);
2427 		}
2428 		psk->psk_killed = killed;
2429 		break;
2430 	}
2431 
2432 	case DIOCADDSTATE: {
2433 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
2434 		struct pfsync_state	*sp = &ps->state;
2435 
2436 		if (sp->timeout >= PFTM_MAX) {
2437 			error = EINVAL;
2438 			break;
2439 		}
2440 		if (V_pfsync_state_import_ptr != NULL) {
2441 			PF_RULES_RLOCK();
2442 			error = V_pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
2443 			PF_RULES_RUNLOCK();
2444 		} else
2445 			error = EOPNOTSUPP;
2446 		break;
2447 	}
2448 
2449 	case DIOCGETSTATE: {
2450 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
2451 		struct pf_state		*s;
2452 
2453 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
2454 		if (s == NULL) {
2455 			error = ENOENT;
2456 			break;
2457 		}
2458 
2459 		pfsync_state_export(&ps->state, s);
2460 		PF_STATE_UNLOCK(s);
2461 		break;
2462 	}
2463 
2464 	case DIOCGETSTATES: {
2465 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
2466 		struct pf_state		*s;
2467 		struct pfsync_state	*pstore, *p;
2468 		int i, nr;
2469 
2470 		if (ps->ps_len <= 0) {
2471 			nr = uma_zone_get_cur(V_pf_state_z);
2472 			ps->ps_len = sizeof(struct pfsync_state) * nr;
2473 			break;
2474 		}
2475 
2476 		p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK | M_ZERO);
2477 		nr = 0;
2478 
2479 		for (i = 0; i <= pf_hashmask; i++) {
2480 			struct pf_idhash *ih = &V_pf_idhash[i];
2481 
2482 			PF_HASHROW_LOCK(ih);
2483 			LIST_FOREACH(s, &ih->states, entry) {
2484 				if (s->timeout == PFTM_UNLINKED)
2485 					continue;
2486 
2487 				if ((nr+1) * sizeof(*p) > ps->ps_len) {
2488 					PF_HASHROW_UNLOCK(ih);
2489 					goto DIOCGETSTATES_full;
2490 				}
2491 				pfsync_state_export(p, s);
2492 				p++;
2493 				nr++;
2494 			}
2495 			PF_HASHROW_UNLOCK(ih);
2496 		}
2497 DIOCGETSTATES_full:
2498 		error = copyout(pstore, ps->ps_states,
2499 		    sizeof(struct pfsync_state) * nr);
2500 		if (error) {
2501 			free(pstore, M_TEMP);
2502 			break;
2503 		}
2504 		ps->ps_len = sizeof(struct pfsync_state) * nr;
2505 		free(pstore, M_TEMP);
2506 
2507 		break;
2508 	}
2509 
2510 	case DIOCGETSTATUS: {
2511 		struct pf_status *s = (struct pf_status *)addr;
2512 
2513 		PF_RULES_RLOCK();
2514 		s->running = V_pf_status.running;
2515 		s->since   = V_pf_status.since;
2516 		s->debug   = V_pf_status.debug;
2517 		s->hostid  = V_pf_status.hostid;
2518 		s->states  = V_pf_status.states;
2519 		s->src_nodes = V_pf_status.src_nodes;
2520 
2521 		for (int i = 0; i < PFRES_MAX; i++)
2522 			s->counters[i] =
2523 			    counter_u64_fetch(V_pf_status.counters[i]);
2524 		for (int i = 0; i < LCNT_MAX; i++)
2525 			s->lcounters[i] =
2526 			    counter_u64_fetch(V_pf_status.lcounters[i]);
2527 		for (int i = 0; i < FCNT_MAX; i++)
2528 			s->fcounters[i] =
2529 			    counter_u64_fetch(V_pf_status.fcounters[i]);
2530 		for (int i = 0; i < SCNT_MAX; i++)
2531 			s->scounters[i] =
2532 			    counter_u64_fetch(V_pf_status.scounters[i]);
2533 
2534 		bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ);
2535 		bcopy(V_pf_status.pf_chksum, s->pf_chksum,
2536 		    PF_MD5_DIGEST_LENGTH);
2537 
2538 		pfi_update_status(s->ifname, s);
2539 		PF_RULES_RUNLOCK();
2540 		break;
2541 	}
2542 
2543 	case DIOCSETSTATUSIF: {
2544 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
2545 
2546 		if (pi->ifname[0] == 0) {
2547 			bzero(V_pf_status.ifname, IFNAMSIZ);
2548 			break;
2549 		}
2550 		PF_RULES_WLOCK();
2551 		strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
2552 		PF_RULES_WUNLOCK();
2553 		break;
2554 	}
2555 
2556 	case DIOCCLRSTATUS: {
2557 		PF_RULES_WLOCK();
2558 		for (int i = 0; i < PFRES_MAX; i++)
2559 			counter_u64_zero(V_pf_status.counters[i]);
2560 		for (int i = 0; i < FCNT_MAX; i++)
2561 			counter_u64_zero(V_pf_status.fcounters[i]);
2562 		for (int i = 0; i < SCNT_MAX; i++)
2563 			counter_u64_zero(V_pf_status.scounters[i]);
2564 		for (int i = 0; i < LCNT_MAX; i++)
2565 			counter_u64_zero(V_pf_status.lcounters[i]);
2566 		V_pf_status.since = time_second;
2567 		if (*V_pf_status.ifname)
2568 			pfi_update_status(V_pf_status.ifname, NULL);
2569 		PF_RULES_WUNLOCK();
2570 		break;
2571 	}
2572 
2573 	case DIOCNATLOOK: {
2574 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
2575 		struct pf_state_key	*sk;
2576 		struct pf_state		*state;
2577 		struct pf_state_key_cmp	 key;
2578 		int			 m = 0, direction = pnl->direction;
2579 		int			 sidx, didx;
2580 
2581 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
2582 		sidx = (direction == PF_IN) ? 1 : 0;
2583 		didx = (direction == PF_IN) ? 0 : 1;
2584 
2585 		if (!pnl->proto ||
2586 		    PF_AZERO(&pnl->saddr, pnl->af) ||
2587 		    PF_AZERO(&pnl->daddr, pnl->af) ||
2588 		    ((pnl->proto == IPPROTO_TCP ||
2589 		    pnl->proto == IPPROTO_UDP) &&
2590 		    (!pnl->dport || !pnl->sport)))
2591 			error = EINVAL;
2592 		else {
2593 			bzero(&key, sizeof(key));
2594 			key.af = pnl->af;
2595 			key.proto = pnl->proto;
2596 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
2597 			key.port[sidx] = pnl->sport;
2598 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
2599 			key.port[didx] = pnl->dport;
2600 
2601 			state = pf_find_state_all(&key, direction, &m);
2602 
2603 			if (m > 1)
2604 				error = E2BIG;	/* more than one state */
2605 			else if (state != NULL) {
2606 				/* XXXGL: not locked read */
2607 				sk = state->key[sidx];
2608 				PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
2609 				pnl->rsport = sk->port[sidx];
2610 				PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
2611 				pnl->rdport = sk->port[didx];
2612 			} else
2613 				error = ENOENT;
2614 		}
2615 		break;
2616 	}
2617 
2618 	case DIOCSETTIMEOUT: {
2619 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2620 		int		 old;
2621 
2622 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
2623 		    pt->seconds < 0) {
2624 			error = EINVAL;
2625 			break;
2626 		}
2627 		PF_RULES_WLOCK();
2628 		old = V_pf_default_rule.timeout[pt->timeout];
2629 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
2630 			pt->seconds = 1;
2631 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
2632 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
2633 			wakeup(pf_purge_thread);
2634 		pt->seconds = old;
2635 		PF_RULES_WUNLOCK();
2636 		break;
2637 	}
2638 
2639 	case DIOCGETTIMEOUT: {
2640 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2641 
2642 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
2643 			error = EINVAL;
2644 			break;
2645 		}
2646 		PF_RULES_RLOCK();
2647 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
2648 		PF_RULES_RUNLOCK();
2649 		break;
2650 	}
2651 
2652 	case DIOCGETLIMIT: {
2653 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2654 
2655 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
2656 			error = EINVAL;
2657 			break;
2658 		}
2659 		PF_RULES_RLOCK();
2660 		pl->limit = V_pf_limits[pl->index].limit;
2661 		PF_RULES_RUNLOCK();
2662 		break;
2663 	}
2664 
2665 	case DIOCSETLIMIT: {
2666 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2667 		int			 old_limit;
2668 
2669 		PF_RULES_WLOCK();
2670 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
2671 		    V_pf_limits[pl->index].zone == NULL) {
2672 			PF_RULES_WUNLOCK();
2673 			error = EINVAL;
2674 			break;
2675 		}
2676 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
2677 		old_limit = V_pf_limits[pl->index].limit;
2678 		V_pf_limits[pl->index].limit = pl->limit;
2679 		pl->limit = old_limit;
2680 		PF_RULES_WUNLOCK();
2681 		break;
2682 	}
2683 
2684 	case DIOCSETDEBUG: {
2685 		u_int32_t	*level = (u_int32_t *)addr;
2686 
2687 		PF_RULES_WLOCK();
2688 		V_pf_status.debug = *level;
2689 		PF_RULES_WUNLOCK();
2690 		break;
2691 	}
2692 
2693 	case DIOCCLRRULECTRS: {
2694 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
2695 		struct pf_kruleset	*ruleset = &pf_main_ruleset;
2696 		struct pf_krule		*rule;
2697 
2698 		PF_RULES_WLOCK();
2699 		TAILQ_FOREACH(rule,
2700 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
2701 			counter_u64_zero(rule->evaluations);
2702 			for (int i = 0; i < 2; i++) {
2703 				counter_u64_zero(rule->packets[i]);
2704 				counter_u64_zero(rule->bytes[i]);
2705 			}
2706 		}
2707 		PF_RULES_WUNLOCK();
2708 		break;
2709 	}
2710 
2711 	case DIOCGIFSPEEDV0:
2712 	case DIOCGIFSPEEDV1: {
2713 		struct pf_ifspeed_v1	*psp = (struct pf_ifspeed_v1 *)addr;
2714 		struct pf_ifspeed_v1	ps;
2715 		struct ifnet		*ifp;
2716 
2717 		if (psp->ifname[0] != 0) {
2718 			/* Can we completely trust user-land? */
2719 			strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
2720 			ifp = ifunit(ps.ifname);
2721 			if (ifp != NULL) {
2722 				psp->baudrate32 =
2723 				    (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
2724 				if (cmd == DIOCGIFSPEEDV1)
2725 					psp->baudrate = ifp->if_baudrate;
2726 			} else
2727 				error = EINVAL;
2728 		} else
2729 			error = EINVAL;
2730 		break;
2731 	}
2732 
2733 #ifdef ALTQ
2734 	case DIOCSTARTALTQ: {
2735 		struct pf_altq		*altq;
2736 
2737 		PF_RULES_WLOCK();
2738 		/* enable all altq interfaces on active list */
2739 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
2740 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
2741 				error = pf_enable_altq(altq);
2742 				if (error != 0)
2743 					break;
2744 			}
2745 		}
2746 		if (error == 0)
2747 			V_pf_altq_running = 1;
2748 		PF_RULES_WUNLOCK();
2749 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
2750 		break;
2751 	}
2752 
2753 	case DIOCSTOPALTQ: {
2754 		struct pf_altq		*altq;
2755 
2756 		PF_RULES_WLOCK();
2757 		/* disable all altq interfaces on active list */
2758 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
2759 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
2760 				error = pf_disable_altq(altq);
2761 				if (error != 0)
2762 					break;
2763 			}
2764 		}
2765 		if (error == 0)
2766 			V_pf_altq_running = 0;
2767 		PF_RULES_WUNLOCK();
2768 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2769 		break;
2770 	}
2771 
2772 	case DIOCADDALTQV0:
2773 	case DIOCADDALTQV1: {
2774 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2775 		struct pf_altq		*altq, *a;
2776 		struct ifnet		*ifp;
2777 
2778 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
2779 		error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
2780 		if (error)
2781 			break;
2782 		altq->local_flags = 0;
2783 
2784 		PF_RULES_WLOCK();
2785 		if (pa->ticket != V_ticket_altqs_inactive) {
2786 			PF_RULES_WUNLOCK();
2787 			free(altq, M_PFALTQ);
2788 			error = EBUSY;
2789 			break;
2790 		}
2791 
2792 		/*
2793 		 * if this is for a queue, find the discipline and
2794 		 * copy the necessary fields
2795 		 */
2796 		if (altq->qname[0] != 0) {
2797 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2798 				PF_RULES_WUNLOCK();
2799 				error = EBUSY;
2800 				free(altq, M_PFALTQ);
2801 				break;
2802 			}
2803 			altq->altq_disc = NULL;
2804 			TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
2805 				if (strncmp(a->ifname, altq->ifname,
2806 				    IFNAMSIZ) == 0) {
2807 					altq->altq_disc = a->altq_disc;
2808 					break;
2809 				}
2810 			}
2811 		}
2812 
2813 		if ((ifp = ifunit(altq->ifname)) == NULL)
2814 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
2815 		else
2816 			error = altq_add(ifp, altq);
2817 
2818 		if (error) {
2819 			PF_RULES_WUNLOCK();
2820 			free(altq, M_PFALTQ);
2821 			break;
2822 		}
2823 
2824 		if (altq->qname[0] != 0)
2825 			TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
2826 		else
2827 			TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
2828 		/* version error check done on import above */
2829 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
2830 		PF_RULES_WUNLOCK();
2831 		break;
2832 	}
2833 
2834 	case DIOCGETALTQSV0:
2835 	case DIOCGETALTQSV1: {
2836 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2837 		struct pf_altq		*altq;
2838 
2839 		PF_RULES_RLOCK();
2840 		pa->nr = 0;
2841 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
2842 			pa->nr++;
2843 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
2844 			pa->nr++;
2845 		pa->ticket = V_ticket_altqs_active;
2846 		PF_RULES_RUNLOCK();
2847 		break;
2848 	}
2849 
2850 	case DIOCGETALTQV0:
2851 	case DIOCGETALTQV1: {
2852 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2853 		struct pf_altq		*altq;
2854 
2855 		PF_RULES_RLOCK();
2856 		if (pa->ticket != V_ticket_altqs_active) {
2857 			PF_RULES_RUNLOCK();
2858 			error = EBUSY;
2859 			break;
2860 		}
2861 		altq = pf_altq_get_nth_active(pa->nr);
2862 		if (altq == NULL) {
2863 			PF_RULES_RUNLOCK();
2864 			error = EBUSY;
2865 			break;
2866 		}
2867 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
2868 		PF_RULES_RUNLOCK();
2869 		break;
2870 	}
2871 
2872 	case DIOCCHANGEALTQV0:
2873 	case DIOCCHANGEALTQV1:
2874 		/* CHANGEALTQ not supported yet! */
2875 		error = ENODEV;
2876 		break;
2877 
2878 	case DIOCGETQSTATSV0:
2879 	case DIOCGETQSTATSV1: {
2880 		struct pfioc_qstats_v1	*pq = (struct pfioc_qstats_v1 *)addr;
2881 		struct pf_altq		*altq;
2882 		int			 nbytes;
2883 		u_int32_t		 version;
2884 
2885 		PF_RULES_RLOCK();
2886 		if (pq->ticket != V_ticket_altqs_active) {
2887 			PF_RULES_RUNLOCK();
2888 			error = EBUSY;
2889 			break;
2890 		}
2891 		nbytes = pq->nbytes;
2892 		altq = pf_altq_get_nth_active(pq->nr);
2893 		if (altq == NULL) {
2894 			PF_RULES_RUNLOCK();
2895 			error = EBUSY;
2896 			break;
2897 		}
2898 
2899 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
2900 			PF_RULES_RUNLOCK();
2901 			error = ENXIO;
2902 			break;
2903 		}
2904 		PF_RULES_RUNLOCK();
2905 		if (cmd == DIOCGETQSTATSV0)
2906 			version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
2907 		else
2908 			version = pq->version;
2909 		error = altq_getqstats(altq, pq->buf, &nbytes, version);
2910 		if (error == 0) {
2911 			pq->scheduler = altq->scheduler;
2912 			pq->nbytes = nbytes;
2913 		}
2914 		break;
2915 	}
2916 #endif /* ALTQ */
2917 
2918 	case DIOCBEGINADDRS: {
2919 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2920 
2921 		PF_RULES_WLOCK();
2922 		pf_empty_kpool(&V_pf_pabuf);
2923 		pp->ticket = ++V_ticket_pabuf;
2924 		PF_RULES_WUNLOCK();
2925 		break;
2926 	}
2927 
2928 	case DIOCADDADDR: {
2929 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2930 		struct pf_kpooladdr	*pa;
2931 		struct pfi_kkif		*kif = NULL;
2932 
2933 #ifndef INET
2934 		if (pp->af == AF_INET) {
2935 			error = EAFNOSUPPORT;
2936 			break;
2937 		}
2938 #endif /* INET */
2939 #ifndef INET6
2940 		if (pp->af == AF_INET6) {
2941 			error = EAFNOSUPPORT;
2942 			break;
2943 		}
2944 #endif /* INET6 */
2945 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2946 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2947 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2948 			error = EINVAL;
2949 			break;
2950 		}
2951 		if (pp->addr.addr.p.dyn != NULL) {
2952 			error = EINVAL;
2953 			break;
2954 		}
2955 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2956 		pf_pooladdr_to_kpooladdr(&pp->addr, pa);
2957 		if (pa->ifname[0])
2958 			kif = pf_kkif_create(M_WAITOK);
2959 		PF_RULES_WLOCK();
2960 		if (pp->ticket != V_ticket_pabuf) {
2961 			PF_RULES_WUNLOCK();
2962 			if (pa->ifname[0])
2963 				pf_kkif_free(kif);
2964 			free(pa, M_PFRULE);
2965 			error = EBUSY;
2966 			break;
2967 		}
2968 		if (pa->ifname[0]) {
2969 			pa->kif = pfi_kkif_attach(kif, pa->ifname);
2970 			pfi_kkif_ref(pa->kif);
2971 		} else
2972 			pa->kif = NULL;
2973 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2974 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2975 			if (pa->ifname[0])
2976 				pfi_kkif_unref(pa->kif);
2977 			PF_RULES_WUNLOCK();
2978 			free(pa, M_PFRULE);
2979 			break;
2980 		}
2981 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2982 		PF_RULES_WUNLOCK();
2983 		break;
2984 	}
2985 
2986 	case DIOCGETADDRS: {
2987 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2988 		struct pf_kpool		*pool;
2989 		struct pf_kpooladdr	*pa;
2990 
2991 		PF_RULES_RLOCK();
2992 		pp->nr = 0;
2993 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2994 		    pp->r_num, 0, 1, 0);
2995 		if (pool == NULL) {
2996 			PF_RULES_RUNLOCK();
2997 			error = EBUSY;
2998 			break;
2999 		}
3000 		TAILQ_FOREACH(pa, &pool->list, entries)
3001 			pp->nr++;
3002 		PF_RULES_RUNLOCK();
3003 		break;
3004 	}
3005 
3006 	case DIOCGETADDR: {
3007 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
3008 		struct pf_kpool		*pool;
3009 		struct pf_kpooladdr	*pa;
3010 		u_int32_t		 nr = 0;
3011 
3012 		PF_RULES_RLOCK();
3013 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
3014 		    pp->r_num, 0, 1, 1);
3015 		if (pool == NULL) {
3016 			PF_RULES_RUNLOCK();
3017 			error = EBUSY;
3018 			break;
3019 		}
3020 		pa = TAILQ_FIRST(&pool->list);
3021 		while ((pa != NULL) && (nr < pp->nr)) {
3022 			pa = TAILQ_NEXT(pa, entries);
3023 			nr++;
3024 		}
3025 		if (pa == NULL) {
3026 			PF_RULES_RUNLOCK();
3027 			error = EBUSY;
3028 			break;
3029 		}
3030 		pf_kpooladdr_to_pooladdr(pa, &pp->addr);
3031 		pf_addr_copyout(&pp->addr.addr);
3032 		PF_RULES_RUNLOCK();
3033 		break;
3034 	}
3035 
3036 	case DIOCCHANGEADDR: {
3037 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
3038 		struct pf_kpool		*pool;
3039 		struct pf_kpooladdr	*oldpa = NULL, *newpa = NULL;
3040 		struct pf_kruleset	*ruleset;
3041 		struct pfi_kkif		*kif = NULL;
3042 
3043 		if (pca->action < PF_CHANGE_ADD_HEAD ||
3044 		    pca->action > PF_CHANGE_REMOVE) {
3045 			error = EINVAL;
3046 			break;
3047 		}
3048 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
3049 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
3050 		    pca->addr.addr.type != PF_ADDR_TABLE) {
3051 			error = EINVAL;
3052 			break;
3053 		}
3054 		if (pca->addr.addr.p.dyn != NULL) {
3055 			error = EINVAL;
3056 			break;
3057 		}
3058 
3059 		if (pca->action != PF_CHANGE_REMOVE) {
3060 #ifndef INET
3061 			if (pca->af == AF_INET) {
3062 				error = EAFNOSUPPORT;
3063 				break;
3064 			}
3065 #endif /* INET */
3066 #ifndef INET6
3067 			if (pca->af == AF_INET6) {
3068 				error = EAFNOSUPPORT;
3069 				break;
3070 			}
3071 #endif /* INET6 */
3072 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
3073 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
3074 			if (newpa->ifname[0])
3075 				kif = pf_kkif_create(M_WAITOK);
3076 			newpa->kif = NULL;
3077 		}
3078 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGEADDR_error; }
3079 		PF_RULES_WLOCK();
3080 		ruleset = pf_find_kruleset(pca->anchor);
3081 		if (ruleset == NULL)
3082 			ERROUT(EBUSY);
3083 
3084 		pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
3085 		    pca->r_num, pca->r_last, 1, 1);
3086 		if (pool == NULL)
3087 			ERROUT(EBUSY);
3088 
3089 		if (pca->action != PF_CHANGE_REMOVE) {
3090 			if (newpa->ifname[0]) {
3091 				newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
3092 				pfi_kkif_ref(newpa->kif);
3093 				kif = NULL;
3094 			}
3095 
3096 			switch (newpa->addr.type) {
3097 			case PF_ADDR_DYNIFTL:
3098 				error = pfi_dynaddr_setup(&newpa->addr,
3099 				    pca->af);
3100 				break;
3101 			case PF_ADDR_TABLE:
3102 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
3103 				    newpa->addr.v.tblname);
3104 				if (newpa->addr.p.tbl == NULL)
3105 					error = ENOMEM;
3106 				break;
3107 			}
3108 			if (error)
3109 				goto DIOCCHANGEADDR_error;
3110 		}
3111 
3112 		switch (pca->action) {
3113 		case PF_CHANGE_ADD_HEAD:
3114 			oldpa = TAILQ_FIRST(&pool->list);
3115 			break;
3116 		case PF_CHANGE_ADD_TAIL:
3117 			oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
3118 			break;
3119 		default:
3120 			oldpa = TAILQ_FIRST(&pool->list);
3121 			for (int i = 0; oldpa && i < pca->nr; i++)
3122 				oldpa = TAILQ_NEXT(oldpa, entries);
3123 
3124 			if (oldpa == NULL)
3125 				ERROUT(EINVAL);
3126 		}
3127 
3128 		if (pca->action == PF_CHANGE_REMOVE) {
3129 			TAILQ_REMOVE(&pool->list, oldpa, entries);
3130 			switch (oldpa->addr.type) {
3131 			case PF_ADDR_DYNIFTL:
3132 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
3133 				break;
3134 			case PF_ADDR_TABLE:
3135 				pfr_detach_table(oldpa->addr.p.tbl);
3136 				break;
3137 			}
3138 			if (oldpa->kif)
3139 				pfi_kkif_unref(oldpa->kif);
3140 			free(oldpa, M_PFRULE);
3141 		} else {
3142 			if (oldpa == NULL)
3143 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
3144 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
3145 			    pca->action == PF_CHANGE_ADD_BEFORE)
3146 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
3147 			else
3148 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
3149 				    newpa, entries);
3150 		}
3151 
3152 		pool->cur = TAILQ_FIRST(&pool->list);
3153 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
3154 		PF_RULES_WUNLOCK();
3155 		break;
3156 
3157 #undef ERROUT
3158 DIOCCHANGEADDR_error:
3159 		if (newpa != NULL) {
3160 			if (newpa->kif)
3161 				pfi_kkif_unref(newpa->kif);
3162 			free(newpa, M_PFRULE);
3163 		}
3164 		PF_RULES_WUNLOCK();
3165 		if (kif != NULL)
3166 			pf_kkif_free(kif);
3167 		break;
3168 	}
3169 
3170 	case DIOCGETRULESETS: {
3171 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
3172 		struct pf_kruleset	*ruleset;
3173 		struct pf_kanchor	*anchor;
3174 
3175 		PF_RULES_RLOCK();
3176 		pr->path[sizeof(pr->path) - 1] = 0;
3177 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
3178 			PF_RULES_RUNLOCK();
3179 			error = ENOENT;
3180 			break;
3181 		}
3182 		pr->nr = 0;
3183 		if (ruleset->anchor == NULL) {
3184 			/* XXX kludge for pf_main_ruleset */
3185 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
3186 				if (anchor->parent == NULL)
3187 					pr->nr++;
3188 		} else {
3189 			RB_FOREACH(anchor, pf_kanchor_node,
3190 			    &ruleset->anchor->children)
3191 				pr->nr++;
3192 		}
3193 		PF_RULES_RUNLOCK();
3194 		break;
3195 	}
3196 
3197 	case DIOCGETRULESET: {
3198 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
3199 		struct pf_kruleset	*ruleset;
3200 		struct pf_kanchor	*anchor;
3201 		u_int32_t		 nr = 0;
3202 
3203 		PF_RULES_RLOCK();
3204 		pr->path[sizeof(pr->path) - 1] = 0;
3205 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
3206 			PF_RULES_RUNLOCK();
3207 			error = ENOENT;
3208 			break;
3209 		}
3210 		pr->name[0] = 0;
3211 		if (ruleset->anchor == NULL) {
3212 			/* XXX kludge for pf_main_ruleset */
3213 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
3214 				if (anchor->parent == NULL && nr++ == pr->nr) {
3215 					strlcpy(pr->name, anchor->name,
3216 					    sizeof(pr->name));
3217 					break;
3218 				}
3219 		} else {
3220 			RB_FOREACH(anchor, pf_kanchor_node,
3221 			    &ruleset->anchor->children)
3222 				if (nr++ == pr->nr) {
3223 					strlcpy(pr->name, anchor->name,
3224 					    sizeof(pr->name));
3225 					break;
3226 				}
3227 		}
3228 		if (!pr->name[0])
3229 			error = EBUSY;
3230 		PF_RULES_RUNLOCK();
3231 		break;
3232 	}
3233 
3234 	case DIOCRCLRTABLES: {
3235 		struct pfioc_table *io = (struct pfioc_table *)addr;
3236 
3237 		if (io->pfrio_esize != 0) {
3238 			error = ENODEV;
3239 			break;
3240 		}
3241 		PF_RULES_WLOCK();
3242 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
3243 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
3244 		PF_RULES_WUNLOCK();
3245 		break;
3246 	}
3247 
3248 	case DIOCRADDTABLES: {
3249 		struct pfioc_table *io = (struct pfioc_table *)addr;
3250 		struct pfr_table *pfrts;
3251 		size_t totlen;
3252 
3253 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3254 			error = ENODEV;
3255 			break;
3256 		}
3257 
3258 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
3259 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
3260 			error = ENOMEM;
3261 			break;
3262 		}
3263 
3264 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3265 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3266 		    M_TEMP, M_WAITOK);
3267 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3268 		if (error) {
3269 			free(pfrts, M_TEMP);
3270 			break;
3271 		}
3272 		PF_RULES_WLOCK();
3273 		error = pfr_add_tables(pfrts, io->pfrio_size,
3274 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3275 		PF_RULES_WUNLOCK();
3276 		free(pfrts, M_TEMP);
3277 		break;
3278 	}
3279 
3280 	case DIOCRDELTABLES: {
3281 		struct pfioc_table *io = (struct pfioc_table *)addr;
3282 		struct pfr_table *pfrts;
3283 		size_t totlen;
3284 
3285 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3286 			error = ENODEV;
3287 			break;
3288 		}
3289 
3290 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
3291 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
3292 			error = ENOMEM;
3293 			break;
3294 		}
3295 
3296 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3297 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3298 		    M_TEMP, M_WAITOK);
3299 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3300 		if (error) {
3301 			free(pfrts, M_TEMP);
3302 			break;
3303 		}
3304 		PF_RULES_WLOCK();
3305 		error = pfr_del_tables(pfrts, io->pfrio_size,
3306 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3307 		PF_RULES_WUNLOCK();
3308 		free(pfrts, M_TEMP);
3309 		break;
3310 	}
3311 
3312 	case DIOCRGETTABLES: {
3313 		struct pfioc_table *io = (struct pfioc_table *)addr;
3314 		struct pfr_table *pfrts;
3315 		size_t totlen;
3316 		int n;
3317 
3318 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3319 			error = ENODEV;
3320 			break;
3321 		}
3322 		PF_RULES_RLOCK();
3323 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3324 		if (n < 0) {
3325 			PF_RULES_RUNLOCK();
3326 			error = EINVAL;
3327 			break;
3328 		}
3329 		io->pfrio_size = min(io->pfrio_size, n);
3330 
3331 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3332 
3333 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3334 		    M_TEMP, M_NOWAIT);
3335 		if (pfrts == NULL) {
3336 			error = ENOMEM;
3337 			PF_RULES_RUNLOCK();
3338 			break;
3339 		}
3340 		error = pfr_get_tables(&io->pfrio_table, pfrts,
3341 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3342 		PF_RULES_RUNLOCK();
3343 		if (error == 0)
3344 			error = copyout(pfrts, io->pfrio_buffer, totlen);
3345 		free(pfrts, M_TEMP);
3346 		break;
3347 	}
3348 
3349 	case DIOCRGETTSTATS: {
3350 		struct pfioc_table *io = (struct pfioc_table *)addr;
3351 		struct pfr_tstats *pfrtstats;
3352 		size_t totlen;
3353 		int n;
3354 
3355 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
3356 			error = ENODEV;
3357 			break;
3358 		}
3359 		PF_RULES_WLOCK();
3360 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3361 		if (n < 0) {
3362 			PF_RULES_WUNLOCK();
3363 			error = EINVAL;
3364 			break;
3365 		}
3366 		io->pfrio_size = min(io->pfrio_size, n);
3367 
3368 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
3369 		pfrtstats = mallocarray(io->pfrio_size,
3370 		    sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT);
3371 		if (pfrtstats == NULL) {
3372 			error = ENOMEM;
3373 			PF_RULES_WUNLOCK();
3374 			break;
3375 		}
3376 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
3377 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3378 		PF_RULES_WUNLOCK();
3379 		if (error == 0)
3380 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
3381 		free(pfrtstats, M_TEMP);
3382 		break;
3383 	}
3384 
3385 	case DIOCRCLRTSTATS: {
3386 		struct pfioc_table *io = (struct pfioc_table *)addr;
3387 		struct pfr_table *pfrts;
3388 		size_t totlen;
3389 
3390 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3391 			error = ENODEV;
3392 			break;
3393 		}
3394 
3395 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
3396 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
3397 			/* We used to count tables and use the minimum required
3398 			 * size, so we didn't fail on overly large requests.
3399 			 * Keep doing so. */
3400 			io->pfrio_size = pf_ioctl_maxcount;
3401 			break;
3402 		}
3403 
3404 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3405 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3406 		    M_TEMP, M_NOWAIT);
3407 		if (pfrts == NULL) {
3408 			error = ENOMEM;
3409 			break;
3410 		}
3411 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3412 		if (error) {
3413 			free(pfrts, M_TEMP);
3414 			break;
3415 		}
3416 
3417 		PF_RULES_WLOCK();
3418 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
3419 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3420 		PF_RULES_WUNLOCK();
3421 		free(pfrts, M_TEMP);
3422 		break;
3423 	}
3424 
3425 	case DIOCRSETTFLAGS: {
3426 		struct pfioc_table *io = (struct pfioc_table *)addr;
3427 		struct pfr_table *pfrts;
3428 		size_t totlen;
3429 		int n;
3430 
3431 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3432 			error = ENODEV;
3433 			break;
3434 		}
3435 
3436 		PF_RULES_RLOCK();
3437 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3438 		if (n < 0) {
3439 			PF_RULES_RUNLOCK();
3440 			error = EINVAL;
3441 			break;
3442 		}
3443 
3444 		io->pfrio_size = min(io->pfrio_size, n);
3445 		PF_RULES_RUNLOCK();
3446 
3447 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3448 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3449 		    M_TEMP, M_WAITOK);
3450 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3451 		if (error) {
3452 			free(pfrts, M_TEMP);
3453 			break;
3454 		}
3455 		PF_RULES_WLOCK();
3456 		error = pfr_set_tflags(pfrts, io->pfrio_size,
3457 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
3458 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3459 		PF_RULES_WUNLOCK();
3460 		free(pfrts, M_TEMP);
3461 		break;
3462 	}
3463 
3464 	case DIOCRCLRADDRS: {
3465 		struct pfioc_table *io = (struct pfioc_table *)addr;
3466 
3467 		if (io->pfrio_esize != 0) {
3468 			error = ENODEV;
3469 			break;
3470 		}
3471 		PF_RULES_WLOCK();
3472 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
3473 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
3474 		PF_RULES_WUNLOCK();
3475 		break;
3476 	}
3477 
3478 	case DIOCRADDADDRS: {
3479 		struct pfioc_table *io = (struct pfioc_table *)addr;
3480 		struct pfr_addr *pfras;
3481 		size_t totlen;
3482 
3483 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3484 			error = ENODEV;
3485 			break;
3486 		}
3487 		if (io->pfrio_size < 0 ||
3488 		    io->pfrio_size > pf_ioctl_maxcount ||
3489 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3490 			error = EINVAL;
3491 			break;
3492 		}
3493 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3494 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3495 		    M_TEMP, M_NOWAIT);
3496 		if (! pfras) {
3497 			error = ENOMEM;
3498 			break;
3499 		}
3500 		error = copyin(io->pfrio_buffer, pfras, totlen);
3501 		if (error) {
3502 			free(pfras, M_TEMP);
3503 			break;
3504 		}
3505 		PF_RULES_WLOCK();
3506 		error = pfr_add_addrs(&io->pfrio_table, pfras,
3507 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
3508 		    PFR_FLAG_USERIOCTL);
3509 		PF_RULES_WUNLOCK();
3510 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3511 			error = copyout(pfras, io->pfrio_buffer, totlen);
3512 		free(pfras, M_TEMP);
3513 		break;
3514 	}
3515 
3516 	case DIOCRDELADDRS: {
3517 		struct pfioc_table *io = (struct pfioc_table *)addr;
3518 		struct pfr_addr *pfras;
3519 		size_t totlen;
3520 
3521 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3522 			error = ENODEV;
3523 			break;
3524 		}
3525 		if (io->pfrio_size < 0 ||
3526 		    io->pfrio_size > pf_ioctl_maxcount ||
3527 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3528 			error = EINVAL;
3529 			break;
3530 		}
3531 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3532 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3533 		    M_TEMP, M_NOWAIT);
3534 		if (! pfras) {
3535 			error = ENOMEM;
3536 			break;
3537 		}
3538 		error = copyin(io->pfrio_buffer, pfras, totlen);
3539 		if (error) {
3540 			free(pfras, M_TEMP);
3541 			break;
3542 		}
3543 		PF_RULES_WLOCK();
3544 		error = pfr_del_addrs(&io->pfrio_table, pfras,
3545 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
3546 		    PFR_FLAG_USERIOCTL);
3547 		PF_RULES_WUNLOCK();
3548 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3549 			error = copyout(pfras, io->pfrio_buffer, totlen);
3550 		free(pfras, M_TEMP);
3551 		break;
3552 	}
3553 
3554 	case DIOCRSETADDRS: {
3555 		struct pfioc_table *io = (struct pfioc_table *)addr;
3556 		struct pfr_addr *pfras;
3557 		size_t totlen, count;
3558 
3559 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3560 			error = ENODEV;
3561 			break;
3562 		}
3563 		if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
3564 			error = EINVAL;
3565 			break;
3566 		}
3567 		count = max(io->pfrio_size, io->pfrio_size2);
3568 		if (count > pf_ioctl_maxcount ||
3569 		    WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
3570 			error = EINVAL;
3571 			break;
3572 		}
3573 		totlen = count * sizeof(struct pfr_addr);
3574 		pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
3575 		    M_NOWAIT);
3576 		if (! pfras) {
3577 			error = ENOMEM;
3578 			break;
3579 		}
3580 		error = copyin(io->pfrio_buffer, pfras, totlen);
3581 		if (error) {
3582 			free(pfras, M_TEMP);
3583 			break;
3584 		}
3585 		PF_RULES_WLOCK();
3586 		error = pfr_set_addrs(&io->pfrio_table, pfras,
3587 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
3588 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
3589 		    PFR_FLAG_USERIOCTL, 0);
3590 		PF_RULES_WUNLOCK();
3591 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3592 			error = copyout(pfras, io->pfrio_buffer, totlen);
3593 		free(pfras, M_TEMP);
3594 		break;
3595 	}
3596 
3597 	case DIOCRGETADDRS: {
3598 		struct pfioc_table *io = (struct pfioc_table *)addr;
3599 		struct pfr_addr *pfras;
3600 		size_t totlen;
3601 
3602 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3603 			error = ENODEV;
3604 			break;
3605 		}
3606 		if (io->pfrio_size < 0 ||
3607 		    io->pfrio_size > pf_ioctl_maxcount ||
3608 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3609 			error = EINVAL;
3610 			break;
3611 		}
3612 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3613 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3614 		    M_TEMP, M_NOWAIT);
3615 		if (! pfras) {
3616 			error = ENOMEM;
3617 			break;
3618 		}
3619 		PF_RULES_RLOCK();
3620 		error = pfr_get_addrs(&io->pfrio_table, pfras,
3621 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3622 		PF_RULES_RUNLOCK();
3623 		if (error == 0)
3624 			error = copyout(pfras, io->pfrio_buffer, totlen);
3625 		free(pfras, M_TEMP);
3626 		break;
3627 	}
3628 
3629 	case DIOCRGETASTATS: {
3630 		struct pfioc_table *io = (struct pfioc_table *)addr;
3631 		struct pfr_astats *pfrastats;
3632 		size_t totlen;
3633 
3634 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
3635 			error = ENODEV;
3636 			break;
3637 		}
3638 		if (io->pfrio_size < 0 ||
3639 		    io->pfrio_size > pf_ioctl_maxcount ||
3640 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
3641 			error = EINVAL;
3642 			break;
3643 		}
3644 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
3645 		pfrastats = mallocarray(io->pfrio_size,
3646 		    sizeof(struct pfr_astats), M_TEMP, M_NOWAIT);
3647 		if (! pfrastats) {
3648 			error = ENOMEM;
3649 			break;
3650 		}
3651 		PF_RULES_RLOCK();
3652 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
3653 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3654 		PF_RULES_RUNLOCK();
3655 		if (error == 0)
3656 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
3657 		free(pfrastats, M_TEMP);
3658 		break;
3659 	}
3660 
3661 	case DIOCRCLRASTATS: {
3662 		struct pfioc_table *io = (struct pfioc_table *)addr;
3663 		struct pfr_addr *pfras;
3664 		size_t totlen;
3665 
3666 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3667 			error = ENODEV;
3668 			break;
3669 		}
3670 		if (io->pfrio_size < 0 ||
3671 		    io->pfrio_size > pf_ioctl_maxcount ||
3672 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3673 			error = EINVAL;
3674 			break;
3675 		}
3676 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3677 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3678 		    M_TEMP, M_NOWAIT);
3679 		if (! pfras) {
3680 			error = ENOMEM;
3681 			break;
3682 		}
3683 		error = copyin(io->pfrio_buffer, pfras, totlen);
3684 		if (error) {
3685 			free(pfras, M_TEMP);
3686 			break;
3687 		}
3688 		PF_RULES_WLOCK();
3689 		error = pfr_clr_astats(&io->pfrio_table, pfras,
3690 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
3691 		    PFR_FLAG_USERIOCTL);
3692 		PF_RULES_WUNLOCK();
3693 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3694 			error = copyout(pfras, io->pfrio_buffer, totlen);
3695 		free(pfras, M_TEMP);
3696 		break;
3697 	}
3698 
3699 	case DIOCRTSTADDRS: {
3700 		struct pfioc_table *io = (struct pfioc_table *)addr;
3701 		struct pfr_addr *pfras;
3702 		size_t totlen;
3703 
3704 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3705 			error = ENODEV;
3706 			break;
3707 		}
3708 		if (io->pfrio_size < 0 ||
3709 		    io->pfrio_size > pf_ioctl_maxcount ||
3710 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3711 			error = EINVAL;
3712 			break;
3713 		}
3714 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3715 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3716 		    M_TEMP, M_NOWAIT);
3717 		if (! pfras) {
3718 			error = ENOMEM;
3719 			break;
3720 		}
3721 		error = copyin(io->pfrio_buffer, pfras, totlen);
3722 		if (error) {
3723 			free(pfras, M_TEMP);
3724 			break;
3725 		}
3726 		PF_RULES_RLOCK();
3727 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
3728 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
3729 		    PFR_FLAG_USERIOCTL);
3730 		PF_RULES_RUNLOCK();
3731 		if (error == 0)
3732 			error = copyout(pfras, io->pfrio_buffer, totlen);
3733 		free(pfras, M_TEMP);
3734 		break;
3735 	}
3736 
3737 	case DIOCRINADEFINE: {
3738 		struct pfioc_table *io = (struct pfioc_table *)addr;
3739 		struct pfr_addr *pfras;
3740 		size_t totlen;
3741 
3742 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3743 			error = ENODEV;
3744 			break;
3745 		}
3746 		if (io->pfrio_size < 0 ||
3747 		    io->pfrio_size > pf_ioctl_maxcount ||
3748 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3749 			error = EINVAL;
3750 			break;
3751 		}
3752 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3753 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3754 		    M_TEMP, M_NOWAIT);
3755 		if (! pfras) {
3756 			error = ENOMEM;
3757 			break;
3758 		}
3759 		error = copyin(io->pfrio_buffer, pfras, totlen);
3760 		if (error) {
3761 			free(pfras, M_TEMP);
3762 			break;
3763 		}
3764 		PF_RULES_WLOCK();
3765 		error = pfr_ina_define(&io->pfrio_table, pfras,
3766 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
3767 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3768 		PF_RULES_WUNLOCK();
3769 		free(pfras, M_TEMP);
3770 		break;
3771 	}
3772 
3773 	case DIOCOSFPADD: {
3774 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
3775 		PF_RULES_WLOCK();
3776 		error = pf_osfp_add(io);
3777 		PF_RULES_WUNLOCK();
3778 		break;
3779 	}
3780 
3781 	case DIOCOSFPGET: {
3782 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
3783 		PF_RULES_RLOCK();
3784 		error = pf_osfp_get(io);
3785 		PF_RULES_RUNLOCK();
3786 		break;
3787 	}
3788 
3789 	case DIOCXBEGIN: {
3790 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3791 		struct pfioc_trans_e	*ioes, *ioe;
3792 		size_t			 totlen;
3793 		int			 i;
3794 
3795 		if (io->esize != sizeof(*ioe)) {
3796 			error = ENODEV;
3797 			break;
3798 		}
3799 		if (io->size < 0 ||
3800 		    io->size > pf_ioctl_maxcount ||
3801 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3802 			error = EINVAL;
3803 			break;
3804 		}
3805 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3806 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3807 		    M_TEMP, M_NOWAIT);
3808 		if (! ioes) {
3809 			error = ENOMEM;
3810 			break;
3811 		}
3812 		error = copyin(io->array, ioes, totlen);
3813 		if (error) {
3814 			free(ioes, M_TEMP);
3815 			break;
3816 		}
3817 		PF_RULES_WLOCK();
3818 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3819 			switch (ioe->rs_num) {
3820 #ifdef ALTQ
3821 			case PF_RULESET_ALTQ:
3822 				if (ioe->anchor[0]) {
3823 					PF_RULES_WUNLOCK();
3824 					free(ioes, M_TEMP);
3825 					error = EINVAL;
3826 					goto fail;
3827 				}
3828 				if ((error = pf_begin_altq(&ioe->ticket))) {
3829 					PF_RULES_WUNLOCK();
3830 					free(ioes, M_TEMP);
3831 					goto fail;
3832 				}
3833 				break;
3834 #endif /* ALTQ */
3835 			case PF_RULESET_TABLE:
3836 			    {
3837 				struct pfr_table table;
3838 
3839 				bzero(&table, sizeof(table));
3840 				strlcpy(table.pfrt_anchor, ioe->anchor,
3841 				    sizeof(table.pfrt_anchor));
3842 				if ((error = pfr_ina_begin(&table,
3843 				    &ioe->ticket, NULL, 0))) {
3844 					PF_RULES_WUNLOCK();
3845 					free(ioes, M_TEMP);
3846 					goto fail;
3847 				}
3848 				break;
3849 			    }
3850 			default:
3851 				if ((error = pf_begin_rules(&ioe->ticket,
3852 				    ioe->rs_num, ioe->anchor))) {
3853 					PF_RULES_WUNLOCK();
3854 					free(ioes, M_TEMP);
3855 					goto fail;
3856 				}
3857 				break;
3858 			}
3859 		}
3860 		PF_RULES_WUNLOCK();
3861 		error = copyout(ioes, io->array, totlen);
3862 		free(ioes, M_TEMP);
3863 		break;
3864 	}
3865 
3866 	case DIOCXROLLBACK: {
3867 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3868 		struct pfioc_trans_e	*ioe, *ioes;
3869 		size_t			 totlen;
3870 		int			 i;
3871 
3872 		if (io->esize != sizeof(*ioe)) {
3873 			error = ENODEV;
3874 			break;
3875 		}
3876 		if (io->size < 0 ||
3877 		    io->size > pf_ioctl_maxcount ||
3878 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3879 			error = EINVAL;
3880 			break;
3881 		}
3882 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3883 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3884 		    M_TEMP, M_NOWAIT);
3885 		if (! ioes) {
3886 			error = ENOMEM;
3887 			break;
3888 		}
3889 		error = copyin(io->array, ioes, totlen);
3890 		if (error) {
3891 			free(ioes, M_TEMP);
3892 			break;
3893 		}
3894 		PF_RULES_WLOCK();
3895 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3896 			switch (ioe->rs_num) {
3897 #ifdef ALTQ
3898 			case PF_RULESET_ALTQ:
3899 				if (ioe->anchor[0]) {
3900 					PF_RULES_WUNLOCK();
3901 					free(ioes, M_TEMP);
3902 					error = EINVAL;
3903 					goto fail;
3904 				}
3905 				if ((error = pf_rollback_altq(ioe->ticket))) {
3906 					PF_RULES_WUNLOCK();
3907 					free(ioes, M_TEMP);
3908 					goto fail; /* really bad */
3909 				}
3910 				break;
3911 #endif /* ALTQ */
3912 			case PF_RULESET_TABLE:
3913 			    {
3914 				struct pfr_table table;
3915 
3916 				bzero(&table, sizeof(table));
3917 				strlcpy(table.pfrt_anchor, ioe->anchor,
3918 				    sizeof(table.pfrt_anchor));
3919 				if ((error = pfr_ina_rollback(&table,
3920 				    ioe->ticket, NULL, 0))) {
3921 					PF_RULES_WUNLOCK();
3922 					free(ioes, M_TEMP);
3923 					goto fail; /* really bad */
3924 				}
3925 				break;
3926 			    }
3927 			default:
3928 				if ((error = pf_rollback_rules(ioe->ticket,
3929 				    ioe->rs_num, ioe->anchor))) {
3930 					PF_RULES_WUNLOCK();
3931 					free(ioes, M_TEMP);
3932 					goto fail; /* really bad */
3933 				}
3934 				break;
3935 			}
3936 		}
3937 		PF_RULES_WUNLOCK();
3938 		free(ioes, M_TEMP);
3939 		break;
3940 	}
3941 
3942 	case DIOCXCOMMIT: {
3943 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3944 		struct pfioc_trans_e	*ioe, *ioes;
3945 		struct pf_kruleset	*rs;
3946 		size_t			 totlen;
3947 		int			 i;
3948 
3949 		if (io->esize != sizeof(*ioe)) {
3950 			error = ENODEV;
3951 			break;
3952 		}
3953 
3954 		if (io->size < 0 ||
3955 		    io->size > pf_ioctl_maxcount ||
3956 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3957 			error = EINVAL;
3958 			break;
3959 		}
3960 
3961 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3962 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3963 		    M_TEMP, M_NOWAIT);
3964 		if (ioes == NULL) {
3965 			error = ENOMEM;
3966 			break;
3967 		}
3968 		error = copyin(io->array, ioes, totlen);
3969 		if (error) {
3970 			free(ioes, M_TEMP);
3971 			break;
3972 		}
3973 		PF_RULES_WLOCK();
3974 		/* First makes sure everything will succeed. */
3975 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3976 			switch (ioe->rs_num) {
3977 #ifdef ALTQ
3978 			case PF_RULESET_ALTQ:
3979 				if (ioe->anchor[0]) {
3980 					PF_RULES_WUNLOCK();
3981 					free(ioes, M_TEMP);
3982 					error = EINVAL;
3983 					goto fail;
3984 				}
3985 				if (!V_altqs_inactive_open || ioe->ticket !=
3986 				    V_ticket_altqs_inactive) {
3987 					PF_RULES_WUNLOCK();
3988 					free(ioes, M_TEMP);
3989 					error = EBUSY;
3990 					goto fail;
3991 				}
3992 				break;
3993 #endif /* ALTQ */
3994 			case PF_RULESET_TABLE:
3995 				rs = pf_find_kruleset(ioe->anchor);
3996 				if (rs == NULL || !rs->topen || ioe->ticket !=
3997 				    rs->tticket) {
3998 					PF_RULES_WUNLOCK();
3999 					free(ioes, M_TEMP);
4000 					error = EBUSY;
4001 					goto fail;
4002 				}
4003 				break;
4004 			default:
4005 				if (ioe->rs_num < 0 || ioe->rs_num >=
4006 				    PF_RULESET_MAX) {
4007 					PF_RULES_WUNLOCK();
4008 					free(ioes, M_TEMP);
4009 					error = EINVAL;
4010 					goto fail;
4011 				}
4012 				rs = pf_find_kruleset(ioe->anchor);
4013 				if (rs == NULL ||
4014 				    !rs->rules[ioe->rs_num].inactive.open ||
4015 				    rs->rules[ioe->rs_num].inactive.ticket !=
4016 				    ioe->ticket) {
4017 					PF_RULES_WUNLOCK();
4018 					free(ioes, M_TEMP);
4019 					error = EBUSY;
4020 					goto fail;
4021 				}
4022 				break;
4023 			}
4024 		}
4025 		/* Now do the commit - no errors should happen here. */
4026 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
4027 			switch (ioe->rs_num) {
4028 #ifdef ALTQ
4029 			case PF_RULESET_ALTQ:
4030 				if ((error = pf_commit_altq(ioe->ticket))) {
4031 					PF_RULES_WUNLOCK();
4032 					free(ioes, M_TEMP);
4033 					goto fail; /* really bad */
4034 				}
4035 				break;
4036 #endif /* ALTQ */
4037 			case PF_RULESET_TABLE:
4038 			    {
4039 				struct pfr_table table;
4040 
4041 				bzero(&table, sizeof(table));
4042 				strlcpy(table.pfrt_anchor, ioe->anchor,
4043 				    sizeof(table.pfrt_anchor));
4044 				if ((error = pfr_ina_commit(&table,
4045 				    ioe->ticket, NULL, NULL, 0))) {
4046 					PF_RULES_WUNLOCK();
4047 					free(ioes, M_TEMP);
4048 					goto fail; /* really bad */
4049 				}
4050 				break;
4051 			    }
4052 			default:
4053 				if ((error = pf_commit_rules(ioe->ticket,
4054 				    ioe->rs_num, ioe->anchor))) {
4055 					PF_RULES_WUNLOCK();
4056 					free(ioes, M_TEMP);
4057 					goto fail; /* really bad */
4058 				}
4059 				break;
4060 			}
4061 		}
4062 		PF_RULES_WUNLOCK();
4063 		free(ioes, M_TEMP);
4064 		break;
4065 	}
4066 
4067 	case DIOCGETSRCNODES: {
4068 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
4069 		struct pf_srchash	*sh;
4070 		struct pf_ksrc_node	*n;
4071 		struct pf_src_node	*p, *pstore;
4072 		uint32_t		 i, nr = 0;
4073 
4074 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4075 				i++, sh++) {
4076 			PF_HASHROW_LOCK(sh);
4077 			LIST_FOREACH(n, &sh->nodes, entry)
4078 				nr++;
4079 			PF_HASHROW_UNLOCK(sh);
4080 		}
4081 
4082 		psn->psn_len = min(psn->psn_len,
4083 		    sizeof(struct pf_src_node) * nr);
4084 
4085 		if (psn->psn_len == 0) {
4086 			psn->psn_len = sizeof(struct pf_src_node) * nr;
4087 			break;
4088 		}
4089 
4090 		nr = 0;
4091 
4092 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
4093 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4094 		    i++, sh++) {
4095 		    PF_HASHROW_LOCK(sh);
4096 		    LIST_FOREACH(n, &sh->nodes, entry) {
4097 
4098 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
4099 				break;
4100 
4101 			pf_src_node_copy(n, p);
4102 
4103 			p++;
4104 			nr++;
4105 		    }
4106 		    PF_HASHROW_UNLOCK(sh);
4107 		}
4108 		error = copyout(pstore, psn->psn_src_nodes,
4109 		    sizeof(struct pf_src_node) * nr);
4110 		if (error) {
4111 			free(pstore, M_TEMP);
4112 			break;
4113 		}
4114 		psn->psn_len = sizeof(struct pf_src_node) * nr;
4115 		free(pstore, M_TEMP);
4116 		break;
4117 	}
4118 
4119 	case DIOCCLRSRCNODES: {
4120 		pf_clear_srcnodes(NULL);
4121 		pf_purge_expired_src_nodes();
4122 		break;
4123 	}
4124 
4125 	case DIOCKILLSRCNODES:
4126 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
4127 		break;
4128 
4129 	case DIOCSETHOSTID: {
4130 		u_int32_t	*hostid = (u_int32_t *)addr;
4131 
4132 		PF_RULES_WLOCK();
4133 		if (*hostid == 0)
4134 			V_pf_status.hostid = arc4random();
4135 		else
4136 			V_pf_status.hostid = *hostid;
4137 		PF_RULES_WUNLOCK();
4138 		break;
4139 	}
4140 
4141 	case DIOCOSFPFLUSH:
4142 		PF_RULES_WLOCK();
4143 		pf_osfp_flush();
4144 		PF_RULES_WUNLOCK();
4145 		break;
4146 
4147 	case DIOCIGETIFACES: {
4148 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
4149 		struct pfi_kif *ifstore;
4150 		size_t bufsiz;
4151 
4152 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
4153 			error = ENODEV;
4154 			break;
4155 		}
4156 
4157 		if (io->pfiio_size < 0 ||
4158 		    io->pfiio_size > pf_ioctl_maxcount ||
4159 		    WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
4160 			error = EINVAL;
4161 			break;
4162 		}
4163 
4164 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
4165 		ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
4166 		    M_TEMP, M_NOWAIT);
4167 		if (ifstore == NULL) {
4168 			error = ENOMEM;
4169 			break;
4170 		}
4171 
4172 		PF_RULES_RLOCK();
4173 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
4174 		PF_RULES_RUNLOCK();
4175 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
4176 		free(ifstore, M_TEMP);
4177 		break;
4178 	}
4179 
4180 	case DIOCSETIFFLAG: {
4181 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
4182 
4183 		PF_RULES_WLOCK();
4184 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
4185 		PF_RULES_WUNLOCK();
4186 		break;
4187 	}
4188 
4189 	case DIOCCLRIFFLAG: {
4190 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
4191 
4192 		PF_RULES_WLOCK();
4193 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
4194 		PF_RULES_WUNLOCK();
4195 		break;
4196 	}
4197 
4198 	default:
4199 		error = ENODEV;
4200 		break;
4201 	}
4202 fail:
4203 	if (sx_xlocked(&pf_ioctl_lock))
4204 		sx_xunlock(&pf_ioctl_lock);
4205 	CURVNET_RESTORE();
4206 
4207 	return (error);
4208 }
4209 
4210 void
4211 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
4212 {
4213 	bzero(sp, sizeof(struct pfsync_state));
4214 
4215 	/* copy from state key */
4216 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
4217 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
4218 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
4219 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
4220 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
4221 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
4222 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
4223 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
4224 	sp->proto = st->key[PF_SK_WIRE]->proto;
4225 	sp->af = st->key[PF_SK_WIRE]->af;
4226 
4227 	/* copy from state */
4228 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
4229 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
4230 	sp->creation = htonl(time_uptime - st->creation);
4231 	sp->expire = pf_state_expires(st);
4232 	if (sp->expire <= time_uptime)
4233 		sp->expire = htonl(0);
4234 	else
4235 		sp->expire = htonl(sp->expire - time_uptime);
4236 
4237 	sp->direction = st->direction;
4238 	sp->log = st->log;
4239 	sp->timeout = st->timeout;
4240 	sp->state_flags = st->state_flags;
4241 	if (st->src_node)
4242 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
4243 	if (st->nat_src_node)
4244 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
4245 
4246 	sp->id = st->id;
4247 	sp->creatorid = st->creatorid;
4248 	pf_state_peer_hton(&st->src, &sp->src);
4249 	pf_state_peer_hton(&st->dst, &sp->dst);
4250 
4251 	if (st->rule.ptr == NULL)
4252 		sp->rule = htonl(-1);
4253 	else
4254 		sp->rule = htonl(st->rule.ptr->nr);
4255 	if (st->anchor.ptr == NULL)
4256 		sp->anchor = htonl(-1);
4257 	else
4258 		sp->anchor = htonl(st->anchor.ptr->nr);
4259 	if (st->nat_rule.ptr == NULL)
4260 		sp->nat_rule = htonl(-1);
4261 	else
4262 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
4263 
4264 	pf_state_counter_hton(counter_u64_fetch(st->packets[0]),
4265 	    sp->packets[0]);
4266 	pf_state_counter_hton(counter_u64_fetch(st->packets[1]),
4267 	    sp->packets[1]);
4268 	pf_state_counter_hton(counter_u64_fetch(st->bytes[0]), sp->bytes[0]);
4269 	pf_state_counter_hton(counter_u64_fetch(st->bytes[1]), sp->bytes[1]);
4270 
4271 }
4272 
4273 static void
4274 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
4275 {
4276 	struct pfr_ktable *kt;
4277 
4278 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
4279 
4280 	kt = aw->p.tbl;
4281 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
4282 		kt = kt->pfrkt_root;
4283 	aw->p.tbl = NULL;
4284 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
4285 		kt->pfrkt_cnt : -1;
4286 }
4287 
4288 /*
4289  * XXX - Check for version missmatch!!!
4290  */
4291 static void
4292 pf_clear_states(void)
4293 {
4294 	struct pf_state	*s;
4295 	u_int i;
4296 
4297 	for (i = 0; i <= pf_hashmask; i++) {
4298 		struct pf_idhash *ih = &V_pf_idhash[i];
4299 relock:
4300 		PF_HASHROW_LOCK(ih);
4301 		LIST_FOREACH(s, &ih->states, entry) {
4302 			s->timeout = PFTM_PURGE;
4303 			/* Don't send out individual delete messages. */
4304 			s->state_flags |= PFSTATE_NOSYNC;
4305 			pf_unlink_state(s, PF_ENTER_LOCKED);
4306 			goto relock;
4307 		}
4308 		PF_HASHROW_UNLOCK(ih);
4309 	}
4310 }
4311 
4312 static int
4313 pf_clear_tables(void)
4314 {
4315 	struct pfioc_table io;
4316 	int error;
4317 
4318 	bzero(&io, sizeof(io));
4319 
4320 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
4321 	    io.pfrio_flags);
4322 
4323 	return (error);
4324 }
4325 
4326 static void
4327 pf_clear_srcnodes(struct pf_ksrc_node *n)
4328 {
4329 	struct pf_state *s;
4330 	int i;
4331 
4332 	for (i = 0; i <= pf_hashmask; i++) {
4333 		struct pf_idhash *ih = &V_pf_idhash[i];
4334 
4335 		PF_HASHROW_LOCK(ih);
4336 		LIST_FOREACH(s, &ih->states, entry) {
4337 			if (n == NULL || n == s->src_node)
4338 				s->src_node = NULL;
4339 			if (n == NULL || n == s->nat_src_node)
4340 				s->nat_src_node = NULL;
4341 		}
4342 		PF_HASHROW_UNLOCK(ih);
4343 	}
4344 
4345 	if (n == NULL) {
4346 		struct pf_srchash *sh;
4347 
4348 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4349 		    i++, sh++) {
4350 			PF_HASHROW_LOCK(sh);
4351 			LIST_FOREACH(n, &sh->nodes, entry) {
4352 				n->expire = 1;
4353 				n->states = 0;
4354 			}
4355 			PF_HASHROW_UNLOCK(sh);
4356 		}
4357 	} else {
4358 		/* XXX: hash slot should already be locked here. */
4359 		n->expire = 1;
4360 		n->states = 0;
4361 	}
4362 }
4363 
4364 static void
4365 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
4366 {
4367 	struct pf_ksrc_node_list	 kill;
4368 
4369 	LIST_INIT(&kill);
4370 	for (int i = 0; i <= pf_srchashmask; i++) {
4371 		struct pf_srchash *sh = &V_pf_srchash[i];
4372 		struct pf_ksrc_node *sn, *tmp;
4373 
4374 		PF_HASHROW_LOCK(sh);
4375 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
4376 			if (PF_MATCHA(psnk->psnk_src.neg,
4377 			      &psnk->psnk_src.addr.v.a.addr,
4378 			      &psnk->psnk_src.addr.v.a.mask,
4379 			      &sn->addr, sn->af) &&
4380 			    PF_MATCHA(psnk->psnk_dst.neg,
4381 			      &psnk->psnk_dst.addr.v.a.addr,
4382 			      &psnk->psnk_dst.addr.v.a.mask,
4383 			      &sn->raddr, sn->af)) {
4384 				pf_unlink_src_node(sn);
4385 				LIST_INSERT_HEAD(&kill, sn, entry);
4386 				sn->expire = 1;
4387 			}
4388 		PF_HASHROW_UNLOCK(sh);
4389 	}
4390 
4391 	for (int i = 0; i <= pf_hashmask; i++) {
4392 		struct pf_idhash *ih = &V_pf_idhash[i];
4393 		struct pf_state *s;
4394 
4395 		PF_HASHROW_LOCK(ih);
4396 		LIST_FOREACH(s, &ih->states, entry) {
4397 			if (s->src_node && s->src_node->expire == 1)
4398 				s->src_node = NULL;
4399 			if (s->nat_src_node && s->nat_src_node->expire == 1)
4400 				s->nat_src_node = NULL;
4401 		}
4402 		PF_HASHROW_UNLOCK(ih);
4403 	}
4404 
4405 	psnk->psnk_killed = pf_free_src_nodes(&kill);
4406 }
4407 
4408 /*
4409  * XXX - Check for version missmatch!!!
4410  */
4411 
4412 /*
4413  * Duplicate pfctl -Fa operation to get rid of as much as we can.
4414  */
4415 static int
4416 shutdown_pf(void)
4417 {
4418 	int error = 0;
4419 	u_int32_t t[5];
4420 	char nn = '\0';
4421 
4422 	do {
4423 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
4424 		    != 0) {
4425 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
4426 			break;
4427 		}
4428 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
4429 		    != 0) {
4430 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
4431 			break;		/* XXX: rollback? */
4432 		}
4433 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
4434 		    != 0) {
4435 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
4436 			break;		/* XXX: rollback? */
4437 		}
4438 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
4439 		    != 0) {
4440 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
4441 			break;		/* XXX: rollback? */
4442 		}
4443 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
4444 		    != 0) {
4445 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
4446 			break;		/* XXX: rollback? */
4447 		}
4448 
4449 		/* XXX: these should always succeed here */
4450 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
4451 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
4452 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
4453 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
4454 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
4455 
4456 		if ((error = pf_clear_tables()) != 0)
4457 			break;
4458 
4459 #ifdef ALTQ
4460 		if ((error = pf_begin_altq(&t[0])) != 0) {
4461 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
4462 			break;
4463 		}
4464 		pf_commit_altq(t[0]);
4465 #endif
4466 
4467 		pf_clear_states();
4468 
4469 		pf_clear_srcnodes(NULL);
4470 
4471 		/* status does not use malloced mem so no need to cleanup */
4472 		/* fingerprints and interfaces have their own cleanup code */
4473 	} while(0);
4474 
4475 	return (error);
4476 }
4477 
4478 static pfil_return_t
4479 pf_check_return(int chk, struct mbuf **m)
4480 {
4481 
4482 	switch (chk) {
4483 	case PF_PASS:
4484 		if (*m == NULL)
4485 			return (PFIL_CONSUMED);
4486 		else
4487 			return (PFIL_PASS);
4488 		break;
4489 	default:
4490 		if (*m != NULL) {
4491 			m_freem(*m);
4492 			*m = NULL;
4493 		}
4494 		return (PFIL_DROPPED);
4495 	}
4496 }
4497 
4498 #ifdef INET
4499 static pfil_return_t
4500 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
4501     void *ruleset __unused, struct inpcb *inp)
4502 {
4503 	int chk;
4504 
4505 	chk = pf_test(PF_IN, flags, ifp, m, inp);
4506 
4507 	return (pf_check_return(chk, m));
4508 }
4509 
4510 static pfil_return_t
4511 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
4512     void *ruleset __unused,  struct inpcb *inp)
4513 {
4514 	int chk;
4515 
4516 	chk = pf_test(PF_OUT, flags, ifp, m, inp);
4517 
4518 	return (pf_check_return(chk, m));
4519 }
4520 #endif
4521 
4522 #ifdef INET6
4523 static pfil_return_t
4524 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
4525     void *ruleset __unused,  struct inpcb *inp)
4526 {
4527 	int chk;
4528 
4529 	/*
4530 	 * In case of loopback traffic IPv6 uses the real interface in
4531 	 * order to support scoped addresses. In order to support stateful
4532 	 * filtering we have change this to lo0 as it is the case in IPv4.
4533 	 */
4534 	CURVNET_SET(ifp->if_vnet);
4535 	chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
4536 	CURVNET_RESTORE();
4537 
4538 	return (pf_check_return(chk, m));
4539 }
4540 
4541 static pfil_return_t
4542 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
4543     void *ruleset __unused,  struct inpcb *inp)
4544 {
4545 	int chk;
4546 
4547 	CURVNET_SET(ifp->if_vnet);
4548 	chk = pf_test6(PF_OUT, flags, ifp, m, inp);
4549 	CURVNET_RESTORE();
4550 
4551 	return (pf_check_return(chk, m));
4552 }
4553 #endif /* INET6 */
4554 
4555 #ifdef INET
4556 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
4557 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
4558 #define	V_pf_ip4_in_hook	VNET(pf_ip4_in_hook)
4559 #define	V_pf_ip4_out_hook	VNET(pf_ip4_out_hook)
4560 #endif
4561 #ifdef INET6
4562 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
4563 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
4564 #define	V_pf_ip6_in_hook	VNET(pf_ip6_in_hook)
4565 #define	V_pf_ip6_out_hook	VNET(pf_ip6_out_hook)
4566 #endif
4567 
4568 static int
4569 hook_pf(void)
4570 {
4571 	struct pfil_hook_args pha;
4572 	struct pfil_link_args pla;
4573 
4574 	if (V_pf_pfil_hooked)
4575 		return (0);
4576 
4577 	pha.pa_version = PFIL_VERSION;
4578 	pha.pa_modname = "pf";
4579 	pha.pa_ruleset = NULL;
4580 
4581 	pla.pa_version = PFIL_VERSION;
4582 
4583 #ifdef INET
4584 	pha.pa_type = PFIL_TYPE_IP4;
4585 	pha.pa_func = pf_check_in;
4586 	pha.pa_flags = PFIL_IN;
4587 	pha.pa_rulname = "default-in";
4588 	V_pf_ip4_in_hook = pfil_add_hook(&pha);
4589 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
4590 	pla.pa_head = V_inet_pfil_head;
4591 	pla.pa_hook = V_pf_ip4_in_hook;
4592 	(void)pfil_link(&pla);
4593 	pha.pa_func = pf_check_out;
4594 	pha.pa_flags = PFIL_OUT;
4595 	pha.pa_rulname = "default-out";
4596 	V_pf_ip4_out_hook = pfil_add_hook(&pha);
4597 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
4598 	pla.pa_head = V_inet_pfil_head;
4599 	pla.pa_hook = V_pf_ip4_out_hook;
4600 	(void)pfil_link(&pla);
4601 #endif
4602 #ifdef INET6
4603 	pha.pa_type = PFIL_TYPE_IP6;
4604 	pha.pa_func = pf_check6_in;
4605 	pha.pa_flags = PFIL_IN;
4606 	pha.pa_rulname = "default-in6";
4607 	V_pf_ip6_in_hook = pfil_add_hook(&pha);
4608 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
4609 	pla.pa_head = V_inet6_pfil_head;
4610 	pla.pa_hook = V_pf_ip6_in_hook;
4611 	(void)pfil_link(&pla);
4612 	pha.pa_func = pf_check6_out;
4613 	pha.pa_rulname = "default-out6";
4614 	pha.pa_flags = PFIL_OUT;
4615 	V_pf_ip6_out_hook = pfil_add_hook(&pha);
4616 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
4617 	pla.pa_head = V_inet6_pfil_head;
4618 	pla.pa_hook = V_pf_ip6_out_hook;
4619 	(void)pfil_link(&pla);
4620 #endif
4621 
4622 	V_pf_pfil_hooked = 1;
4623 	return (0);
4624 }
4625 
4626 static int
4627 dehook_pf(void)
4628 {
4629 
4630 	if (V_pf_pfil_hooked == 0)
4631 		return (0);
4632 
4633 #ifdef INET
4634 	pfil_remove_hook(V_pf_ip4_in_hook);
4635 	pfil_remove_hook(V_pf_ip4_out_hook);
4636 #endif
4637 #ifdef INET6
4638 	pfil_remove_hook(V_pf_ip6_in_hook);
4639 	pfil_remove_hook(V_pf_ip6_out_hook);
4640 #endif
4641 
4642 	V_pf_pfil_hooked = 0;
4643 	return (0);
4644 }
4645 
4646 static void
4647 pf_load_vnet(void)
4648 {
4649 	V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
4650 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
4651 
4652 	pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
4653 	    PF_RULE_TAG_HASH_SIZE_DEFAULT);
4654 #ifdef ALTQ
4655 	pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
4656 	    PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
4657 #endif
4658 
4659 	pfattach_vnet();
4660 	V_pf_vnet_active = 1;
4661 }
4662 
4663 static int
4664 pf_load(void)
4665 {
4666 	int error;
4667 
4668 	rm_init(&pf_rules_lock, "pf rulesets");
4669 	sx_init(&pf_ioctl_lock, "pf ioctl");
4670 	sx_init(&pf_end_lock, "pf end thread");
4671 
4672 	pf_mtag_initialize();
4673 
4674 	pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
4675 	if (pf_dev == NULL)
4676 		return (ENOMEM);
4677 
4678 	pf_end_threads = 0;
4679 	error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
4680 	if (error != 0)
4681 		return (error);
4682 
4683 	pfi_initialize();
4684 
4685 	return (0);
4686 }
4687 
4688 static void
4689 pf_unload_vnet(void)
4690 {
4691 	int error;
4692 
4693 	V_pf_vnet_active = 0;
4694 	V_pf_status.running = 0;
4695 	error = dehook_pf();
4696 	if (error) {
4697 		/*
4698 		 * Should not happen!
4699 		 * XXX Due to error code ESRCH, kldunload will show
4700 		 * a message like 'No such process'.
4701 		 */
4702 		printf("%s : pfil unregisteration fail\n", __FUNCTION__);
4703 		return;
4704 	}
4705 
4706 	PF_RULES_WLOCK();
4707 	shutdown_pf();
4708 	PF_RULES_WUNLOCK();
4709 
4710 	swi_remove(V_pf_swi_cookie);
4711 
4712 	pf_unload_vnet_purge();
4713 
4714 	pf_normalize_cleanup();
4715 	PF_RULES_WLOCK();
4716 	pfi_cleanup_vnet();
4717 	PF_RULES_WUNLOCK();
4718 	pfr_cleanup();
4719 	pf_osfp_flush();
4720 	pf_cleanup();
4721 	if (IS_DEFAULT_VNET(curvnet))
4722 		pf_mtag_cleanup();
4723 
4724 	pf_cleanup_tagset(&V_pf_tags);
4725 #ifdef ALTQ
4726 	pf_cleanup_tagset(&V_pf_qids);
4727 #endif
4728 	uma_zdestroy(V_pf_tag_z);
4729 
4730 	/* Free counters last as we updated them during shutdown. */
4731 	counter_u64_free(V_pf_default_rule.evaluations);
4732 	for (int i = 0; i < 2; i++) {
4733 		counter_u64_free(V_pf_default_rule.packets[i]);
4734 		counter_u64_free(V_pf_default_rule.bytes[i]);
4735 	}
4736 	counter_u64_free(V_pf_default_rule.states_cur);
4737 	counter_u64_free(V_pf_default_rule.states_tot);
4738 	counter_u64_free(V_pf_default_rule.src_nodes);
4739 
4740 	for (int i = 0; i < PFRES_MAX; i++)
4741 		counter_u64_free(V_pf_status.counters[i]);
4742 	for (int i = 0; i < LCNT_MAX; i++)
4743 		counter_u64_free(V_pf_status.lcounters[i]);
4744 	for (int i = 0; i < FCNT_MAX; i++)
4745 		counter_u64_free(V_pf_status.fcounters[i]);
4746 	for (int i = 0; i < SCNT_MAX; i++)
4747 		counter_u64_free(V_pf_status.scounters[i]);
4748 }
4749 
4750 static void
4751 pf_unload(void)
4752 {
4753 
4754 	sx_xlock(&pf_end_lock);
4755 	pf_end_threads = 1;
4756 	while (pf_end_threads < 2) {
4757 		wakeup_one(pf_purge_thread);
4758 		sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
4759 	}
4760 	sx_xunlock(&pf_end_lock);
4761 
4762 	if (pf_dev != NULL)
4763 		destroy_dev(pf_dev);
4764 
4765 	pfi_cleanup();
4766 
4767 	rm_destroy(&pf_rules_lock);
4768 	sx_destroy(&pf_ioctl_lock);
4769 	sx_destroy(&pf_end_lock);
4770 }
4771 
4772 static void
4773 vnet_pf_init(void *unused __unused)
4774 {
4775 
4776 	pf_load_vnet();
4777 }
4778 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
4779     vnet_pf_init, NULL);
4780 
4781 static void
4782 vnet_pf_uninit(const void *unused __unused)
4783 {
4784 
4785 	pf_unload_vnet();
4786 }
4787 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
4788 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
4789     vnet_pf_uninit, NULL);
4790 
4791 static int
4792 pf_modevent(module_t mod, int type, void *data)
4793 {
4794 	int error = 0;
4795 
4796 	switch(type) {
4797 	case MOD_LOAD:
4798 		error = pf_load();
4799 		break;
4800 	case MOD_UNLOAD:
4801 		/* Handled in SYSUNINIT(pf_unload) to ensure it's done after
4802 		 * the vnet_pf_uninit()s */
4803 		break;
4804 	default:
4805 		error = EINVAL;
4806 		break;
4807 	}
4808 
4809 	return (error);
4810 }
4811 
4812 static moduledata_t pf_mod = {
4813 	"pf",
4814 	pf_modevent,
4815 	0
4816 };
4817 
4818 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
4819 MODULE_VERSION(pf, PF_MODVER);
4820