xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision 69c5fa5cd1ec9b09ed88a086607a8a0993818db9)
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_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
1562 {
1563 
1564 #ifndef INET
1565 	if (rule->af == AF_INET) {
1566 		return (EAFNOSUPPORT);
1567 	}
1568 #endif /* INET */
1569 #ifndef INET6
1570 	if (rule->af == AF_INET6) {
1571 		return (EAFNOSUPPORT);
1572 	}
1573 #endif /* INET6 */
1574 
1575 	if (rule->src.addr.type != PF_ADDR_ADDRMASK &&
1576 	    rule->src.addr.type != PF_ADDR_DYNIFTL &&
1577 	    rule->src.addr.type != PF_ADDR_TABLE) {
1578 		return (EINVAL);
1579 	}
1580 	if (rule->src.addr.p.dyn != NULL) {
1581 		return (EINVAL);
1582 	}
1583 
1584 	if (rule->dst.addr.type != PF_ADDR_ADDRMASK &&
1585 	    rule->dst.addr.type != PF_ADDR_DYNIFTL &&
1586 	    rule->dst.addr.type != PF_ADDR_TABLE) {
1587 		return (EINVAL);
1588 	}
1589 	if (rule->dst.addr.p.dyn != NULL) {
1590 		return (EINVAL);
1591 	}
1592 
1593 	bzero(krule, sizeof(*krule));
1594 
1595 	bcopy(&rule->src, &krule->src, sizeof(rule->src));
1596 	bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
1597 
1598 	strlcpy(krule->label, rule->label, sizeof(rule->label));
1599 	strlcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
1600 	strlcpy(krule->qname, rule->qname, sizeof(rule->qname));
1601 	strlcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
1602 	strlcpy(krule->tagname, rule->tagname, sizeof(rule->tagname));
1603 	strlcpy(krule->match_tagname, rule->match_tagname,
1604 	    sizeof(rule->match_tagname));
1605 	strlcpy(krule->overload_tblname, rule->overload_tblname,
1606 	    sizeof(rule->overload_tblname));
1607 
1608 	bcopy(&rule->rpool, &krule->rpool, sizeof(krule->rpool));
1609 
1610 	/* Don't allow userspace to set evaulations, packets or bytes. */
1611 	/* kif, anchor, overload_tbl are not copied over. */
1612 
1613 	krule->os_fingerprint = krule->os_fingerprint;
1614 
1615 	krule->rtableid = rule->rtableid;
1616 	bcopy(rule->timeout, krule->timeout, sizeof(krule->timeout));
1617 	krule->max_states = rule->max_states;
1618 	krule->max_src_nodes = rule->max_src_nodes;
1619 	krule->max_src_states = rule->max_src_states;
1620 	krule->max_src_conn = rule->max_src_conn;
1621 	krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
1622 	krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
1623 	krule->qid = rule->qid;
1624 	krule->pqid = rule->pqid;
1625 	krule->rt_listid = rule->rt_listid;
1626 	krule->nr = rule->nr;
1627 	krule->prob = rule->prob;
1628 	krule->cuid = rule->cuid;
1629 	krule->cpid = rule->cpid;
1630 
1631 	krule->return_icmp = rule->return_icmp;
1632 	krule->return_icmp6 = rule->return_icmp6;
1633 	krule->max_mss = rule->max_mss;
1634 	krule->tag = rule->tag;
1635 	krule->match_tag = rule->match_tag;
1636 	krule->scrub_flags = rule->scrub_flags;
1637 
1638 	bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
1639 	bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
1640 
1641 	krule->rule_flag = rule->rule_flag;
1642 	krule->action = rule->action;
1643 	krule->direction = rule->direction;
1644 	krule->log = rule->log;
1645 	krule->logif = rule->logif;
1646 	krule->quick = rule->quick;
1647 	krule->ifnot = rule->ifnot;
1648 	krule->match_tag_not = rule->match_tag_not;
1649 	krule->natpass = rule->natpass;
1650 
1651 	krule->keep_state = rule->keep_state;
1652 	krule->af = rule->af;
1653 	krule->proto = rule->proto;
1654 	krule->type = rule->type;
1655 	krule->code = rule->code;
1656 	krule->flags = rule->flags;
1657 	krule->flagset = rule->flagset;
1658 	krule->min_ttl = rule->min_ttl;
1659 	krule->allow_opts = rule->allow_opts;
1660 	krule->rt = rule->rt;
1661 	krule->return_ttl = rule->return_ttl;
1662 	krule->tos = rule->tos;
1663 	krule->set_tos = rule->set_tos;
1664 	krule->anchor_relative = rule->anchor_relative;
1665 	krule->anchor_wildcard = rule->anchor_wildcard;
1666 
1667 	krule->flush = rule->flush;
1668 	krule->prio = rule->prio;
1669 	krule->set_prio[0] = rule->set_prio[0];
1670 	krule->set_prio[1] = rule->set_prio[1];
1671 
1672 	bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
1673 
1674 	return (0);
1675 }
1676 
1677 static int
1678 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
1679 {
1680 	int			 error = 0;
1681 	PF_RULES_RLOCK_TRACKER;
1682 
1683 	/* XXX keep in sync with switch() below */
1684 	if (securelevel_gt(td->td_ucred, 2))
1685 		switch (cmd) {
1686 		case DIOCGETRULES:
1687 		case DIOCGETRULE:
1688 		case DIOCGETADDRS:
1689 		case DIOCGETADDR:
1690 		case DIOCGETSTATE:
1691 		case DIOCSETSTATUSIF:
1692 		case DIOCGETSTATUS:
1693 		case DIOCCLRSTATUS:
1694 		case DIOCNATLOOK:
1695 		case DIOCSETDEBUG:
1696 		case DIOCGETSTATES:
1697 		case DIOCGETTIMEOUT:
1698 		case DIOCCLRRULECTRS:
1699 		case DIOCGETLIMIT:
1700 		case DIOCGETALTQSV0:
1701 		case DIOCGETALTQSV1:
1702 		case DIOCGETALTQV0:
1703 		case DIOCGETALTQV1:
1704 		case DIOCGETQSTATSV0:
1705 		case DIOCGETQSTATSV1:
1706 		case DIOCGETRULESETS:
1707 		case DIOCGETRULESET:
1708 		case DIOCRGETTABLES:
1709 		case DIOCRGETTSTATS:
1710 		case DIOCRCLRTSTATS:
1711 		case DIOCRCLRADDRS:
1712 		case DIOCRADDADDRS:
1713 		case DIOCRDELADDRS:
1714 		case DIOCRSETADDRS:
1715 		case DIOCRGETADDRS:
1716 		case DIOCRGETASTATS:
1717 		case DIOCRCLRASTATS:
1718 		case DIOCRTSTADDRS:
1719 		case DIOCOSFPGET:
1720 		case DIOCGETSRCNODES:
1721 		case DIOCCLRSRCNODES:
1722 		case DIOCIGETIFACES:
1723 		case DIOCGIFSPEEDV0:
1724 		case DIOCGIFSPEEDV1:
1725 		case DIOCSETIFFLAG:
1726 		case DIOCCLRIFFLAG:
1727 			break;
1728 		case DIOCRCLRTABLES:
1729 		case DIOCRADDTABLES:
1730 		case DIOCRDELTABLES:
1731 		case DIOCRSETTFLAGS:
1732 			if (((struct pfioc_table *)addr)->pfrio_flags &
1733 			    PFR_FLAG_DUMMY)
1734 				break; /* dummy operation ok */
1735 			return (EPERM);
1736 		default:
1737 			return (EPERM);
1738 		}
1739 
1740 	if (!(flags & FWRITE))
1741 		switch (cmd) {
1742 		case DIOCGETRULES:
1743 		case DIOCGETADDRS:
1744 		case DIOCGETADDR:
1745 		case DIOCGETSTATE:
1746 		case DIOCGETSTATUS:
1747 		case DIOCGETSTATES:
1748 		case DIOCGETTIMEOUT:
1749 		case DIOCGETLIMIT:
1750 		case DIOCGETALTQSV0:
1751 		case DIOCGETALTQSV1:
1752 		case DIOCGETALTQV0:
1753 		case DIOCGETALTQV1:
1754 		case DIOCGETQSTATSV0:
1755 		case DIOCGETQSTATSV1:
1756 		case DIOCGETRULESETS:
1757 		case DIOCGETRULESET:
1758 		case DIOCNATLOOK:
1759 		case DIOCRGETTABLES:
1760 		case DIOCRGETTSTATS:
1761 		case DIOCRGETADDRS:
1762 		case DIOCRGETASTATS:
1763 		case DIOCRTSTADDRS:
1764 		case DIOCOSFPGET:
1765 		case DIOCGETSRCNODES:
1766 		case DIOCIGETIFACES:
1767 		case DIOCGIFSPEEDV1:
1768 		case DIOCGIFSPEEDV0:
1769 			break;
1770 		case DIOCRCLRTABLES:
1771 		case DIOCRADDTABLES:
1772 		case DIOCRDELTABLES:
1773 		case DIOCRCLRTSTATS:
1774 		case DIOCRCLRADDRS:
1775 		case DIOCRADDADDRS:
1776 		case DIOCRDELADDRS:
1777 		case DIOCRSETADDRS:
1778 		case DIOCRSETTFLAGS:
1779 			if (((struct pfioc_table *)addr)->pfrio_flags &
1780 			    PFR_FLAG_DUMMY) {
1781 				flags |= FWRITE; /* need write lock for dummy */
1782 				break; /* dummy operation ok */
1783 			}
1784 			return (EACCES);
1785 		case DIOCGETRULE:
1786 			if (((struct pfioc_rule *)addr)->action ==
1787 			    PF_GET_CLR_CNTR)
1788 				return (EACCES);
1789 			break;
1790 		default:
1791 			return (EACCES);
1792 		}
1793 
1794 	CURVNET_SET(TD_TO_VNET(td));
1795 
1796 	switch (cmd) {
1797 	case DIOCSTART:
1798 		sx_xlock(&pf_ioctl_lock);
1799 		if (V_pf_status.running)
1800 			error = EEXIST;
1801 		else {
1802 			int cpu;
1803 
1804 			error = hook_pf();
1805 			if (error) {
1806 				DPFPRINTF(PF_DEBUG_MISC,
1807 				    ("pf: pfil registration failed\n"));
1808 				break;
1809 			}
1810 			V_pf_status.running = 1;
1811 			V_pf_status.since = time_second;
1812 
1813 			CPU_FOREACH(cpu)
1814 				V_pf_stateid[cpu] = time_second;
1815 
1816 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1817 		}
1818 		break;
1819 
1820 	case DIOCSTOP:
1821 		sx_xlock(&pf_ioctl_lock);
1822 		if (!V_pf_status.running)
1823 			error = ENOENT;
1824 		else {
1825 			V_pf_status.running = 0;
1826 			error = dehook_pf();
1827 			if (error) {
1828 				V_pf_status.running = 1;
1829 				DPFPRINTF(PF_DEBUG_MISC,
1830 				    ("pf: pfil unregistration failed\n"));
1831 			}
1832 			V_pf_status.since = time_second;
1833 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1834 		}
1835 		break;
1836 
1837 	case DIOCADDRULE: {
1838 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1839 		struct pf_kruleset	*ruleset;
1840 		struct pf_krule		*rule, *tail;
1841 		struct pf_kpooladdr	*pa;
1842 		struct pfi_kkif		*kif = NULL;
1843 		int			 rs_num;
1844 
1845 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1846 			error = EINVAL;
1847 			break;
1848 		}
1849 
1850 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
1851 		error = pf_rule_to_krule(&pr->rule, rule);
1852 		if (error != 0) {
1853 			free(rule, M_PFRULE);
1854 			break;
1855 		}
1856 
1857 		if (rule->ifname[0])
1858 			kif = pf_kkif_create(M_WAITOK);
1859 		rule->evaluations = counter_u64_alloc(M_WAITOK);
1860 		for (int i = 0; i < 2; i++) {
1861 			rule->packets[i] = counter_u64_alloc(M_WAITOK);
1862 			rule->bytes[i] = counter_u64_alloc(M_WAITOK);
1863 		}
1864 		rule->states_cur = counter_u64_alloc(M_WAITOK);
1865 		rule->states_tot = counter_u64_alloc(M_WAITOK);
1866 		rule->src_nodes = counter_u64_alloc(M_WAITOK);
1867 		rule->cuid = td->td_ucred->cr_ruid;
1868 		rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1869 		TAILQ_INIT(&rule->rpool.list);
1870 #define	ERROUT(x)	{ error = (x); goto DIOCADDRULE_error; }
1871 
1872 		PF_RULES_WLOCK();
1873 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1874 		ruleset = pf_find_kruleset(pr->anchor);
1875 		if (ruleset == NULL)
1876 			ERROUT(EINVAL);
1877 		rs_num = pf_get_ruleset_number(pr->rule.action);
1878 		if (rs_num >= PF_RULESET_MAX)
1879 			ERROUT(EINVAL);
1880 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1881 			DPFPRINTF(PF_DEBUG_MISC,
1882 			    ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
1883 			    ruleset->rules[rs_num].inactive.ticket));
1884 			ERROUT(EBUSY);
1885 		}
1886 		if (pr->pool_ticket != V_ticket_pabuf) {
1887 			DPFPRINTF(PF_DEBUG_MISC,
1888 			    ("pool_ticket: %d != %d\n", pr->pool_ticket,
1889 			    V_ticket_pabuf));
1890 			ERROUT(EBUSY);
1891 		}
1892 
1893 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1894 		    pf_krulequeue);
1895 		if (tail)
1896 			rule->nr = tail->nr + 1;
1897 		else
1898 			rule->nr = 0;
1899 		if (rule->ifname[0]) {
1900 			rule->kif = pfi_kkif_attach(kif, rule->ifname);
1901 			pfi_kkif_ref(rule->kif);
1902 		} else
1903 			rule->kif = NULL;
1904 
1905 		if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
1906 			error = EBUSY;
1907 
1908 #ifdef ALTQ
1909 		/* set queue IDs */
1910 		if (rule->qname[0] != 0) {
1911 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1912 				error = EBUSY;
1913 			else if (rule->pqname[0] != 0) {
1914 				if ((rule->pqid =
1915 				    pf_qname2qid(rule->pqname)) == 0)
1916 					error = EBUSY;
1917 			} else
1918 				rule->pqid = rule->qid;
1919 		}
1920 #endif
1921 		if (rule->tagname[0])
1922 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1923 				error = EBUSY;
1924 		if (rule->match_tagname[0])
1925 			if ((rule->match_tag =
1926 			    pf_tagname2tag(rule->match_tagname)) == 0)
1927 				error = EBUSY;
1928 		if (rule->rt && !rule->direction)
1929 			error = EINVAL;
1930 		if (!rule->log)
1931 			rule->logif = 0;
1932 		if (rule->logif >= PFLOGIFS_MAX)
1933 			error = EINVAL;
1934 		if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1935 			error = ENOMEM;
1936 		if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1937 			error = ENOMEM;
1938 		if (pf_kanchor_setup(rule, ruleset, pr->anchor_call))
1939 			error = EINVAL;
1940 		if (rule->scrub_flags & PFSTATE_SETPRIO &&
1941 		    (rule->set_prio[0] > PF_PRIO_MAX ||
1942 		    rule->set_prio[1] > PF_PRIO_MAX))
1943 			error = EINVAL;
1944 		TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1945 			if (pa->addr.type == PF_ADDR_TABLE) {
1946 				pa->addr.p.tbl = pfr_attach_table(ruleset,
1947 				    pa->addr.v.tblname);
1948 				if (pa->addr.p.tbl == NULL)
1949 					error = ENOMEM;
1950 			}
1951 
1952 		rule->overload_tbl = NULL;
1953 		if (rule->overload_tblname[0]) {
1954 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1955 			    rule->overload_tblname)) == NULL)
1956 				error = EINVAL;
1957 			else
1958 				rule->overload_tbl->pfrkt_flags |=
1959 				    PFR_TFLAG_ACTIVE;
1960 		}
1961 
1962 		pf_mv_kpool(&V_pf_pabuf, &rule->rpool.list);
1963 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1964 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1965 		    (rule->rt > PF_NOPFROUTE)) &&
1966 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1967 			error = EINVAL;
1968 
1969 		if (error) {
1970 			pf_free_rule(rule);
1971 			PF_RULES_WUNLOCK();
1972 			break;
1973 		}
1974 
1975 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1976 		counter_u64_zero(rule->evaluations);
1977 		for (int i = 0; i < 2; i++) {
1978 			counter_u64_zero(rule->packets[i]);
1979 			counter_u64_zero(rule->bytes[i]);
1980 		}
1981 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1982 		    rule, entries);
1983 		ruleset->rules[rs_num].inactive.rcount++;
1984 		PF_RULES_WUNLOCK();
1985 		break;
1986 
1987 #undef ERROUT
1988 DIOCADDRULE_error:
1989 		PF_RULES_WUNLOCK();
1990 		counter_u64_free(rule->evaluations);
1991 		for (int i = 0; i < 2; i++) {
1992 			counter_u64_free(rule->packets[i]);
1993 			counter_u64_free(rule->bytes[i]);
1994 		}
1995 		counter_u64_free(rule->states_cur);
1996 		counter_u64_free(rule->states_tot);
1997 		counter_u64_free(rule->src_nodes);
1998 		free(rule, M_PFRULE);
1999 		if (kif)
2000 			pf_kkif_free(kif);
2001 		break;
2002 	}
2003 
2004 	case DIOCGETRULES: {
2005 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
2006 		struct pf_kruleset	*ruleset;
2007 		struct pf_krule		*tail;
2008 		int			 rs_num;
2009 
2010 		PF_RULES_WLOCK();
2011 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
2012 		ruleset = pf_find_kruleset(pr->anchor);
2013 		if (ruleset == NULL) {
2014 			PF_RULES_WUNLOCK();
2015 			error = EINVAL;
2016 			break;
2017 		}
2018 		rs_num = pf_get_ruleset_number(pr->rule.action);
2019 		if (rs_num >= PF_RULESET_MAX) {
2020 			PF_RULES_WUNLOCK();
2021 			error = EINVAL;
2022 			break;
2023 		}
2024 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
2025 		    pf_krulequeue);
2026 		if (tail)
2027 			pr->nr = tail->nr + 1;
2028 		else
2029 			pr->nr = 0;
2030 		pr->ticket = ruleset->rules[rs_num].active.ticket;
2031 		PF_RULES_WUNLOCK();
2032 		break;
2033 	}
2034 
2035 	case DIOCGETRULE: {
2036 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
2037 		struct pf_kruleset	*ruleset;
2038 		struct pf_krule		*rule;
2039 		int			 rs_num;
2040 
2041 		PF_RULES_WLOCK();
2042 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
2043 		ruleset = pf_find_kruleset(pr->anchor);
2044 		if (ruleset == NULL) {
2045 			PF_RULES_WUNLOCK();
2046 			error = EINVAL;
2047 			break;
2048 		}
2049 		rs_num = pf_get_ruleset_number(pr->rule.action);
2050 		if (rs_num >= PF_RULESET_MAX) {
2051 			PF_RULES_WUNLOCK();
2052 			error = EINVAL;
2053 			break;
2054 		}
2055 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
2056 			PF_RULES_WUNLOCK();
2057 			error = EBUSY;
2058 			break;
2059 		}
2060 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
2061 		while ((rule != NULL) && (rule->nr != pr->nr))
2062 			rule = TAILQ_NEXT(rule, entries);
2063 		if (rule == NULL) {
2064 			PF_RULES_WUNLOCK();
2065 			error = EBUSY;
2066 			break;
2067 		}
2068 
2069 		pf_krule_to_rule(rule, &pr->rule);
2070 
2071 		if (pf_kanchor_copyout(ruleset, rule, pr)) {
2072 			PF_RULES_WUNLOCK();
2073 			error = EBUSY;
2074 			break;
2075 		}
2076 		pf_addr_copyout(&pr->rule.src.addr);
2077 		pf_addr_copyout(&pr->rule.dst.addr);
2078 
2079 		if (pr->action == PF_GET_CLR_CNTR) {
2080 			counter_u64_zero(rule->evaluations);
2081 			for (int i = 0; i < 2; i++) {
2082 				counter_u64_zero(rule->packets[i]);
2083 				counter_u64_zero(rule->bytes[i]);
2084 			}
2085 			counter_u64_zero(rule->states_tot);
2086 		}
2087 		PF_RULES_WUNLOCK();
2088 		break;
2089 	}
2090 
2091 	case DIOCCHANGERULE: {
2092 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
2093 		struct pf_kruleset	*ruleset;
2094 		struct pf_krule		*oldrule = NULL, *newrule = NULL;
2095 		struct pfi_kkif		*kif = NULL;
2096 		struct pf_kpooladdr	*pa;
2097 		u_int32_t		 nr = 0;
2098 		int			 rs_num;
2099 
2100 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
2101 		    pcr->action > PF_CHANGE_GET_TICKET) {
2102 			error = EINVAL;
2103 			break;
2104 		}
2105 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
2106 			error = EINVAL;
2107 			break;
2108 		}
2109 
2110 		if (pcr->action != PF_CHANGE_REMOVE) {
2111 			newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
2112 			error = pf_rule_to_krule(&pcr->rule, newrule);
2113 			if (error != 0) {
2114 				free(newrule, M_PFRULE);
2115 				break;
2116 			}
2117 
2118 			if (newrule->ifname[0])
2119 				kif = pf_kkif_create(M_WAITOK);
2120 			newrule->evaluations = counter_u64_alloc(M_WAITOK);
2121 			for (int i = 0; i < 2; i++) {
2122 				newrule->packets[i] =
2123 				    counter_u64_alloc(M_WAITOK);
2124 				newrule->bytes[i] =
2125 				    counter_u64_alloc(M_WAITOK);
2126 			}
2127 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
2128 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
2129 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
2130 			newrule->cuid = td->td_ucred->cr_ruid;
2131 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
2132 			TAILQ_INIT(&newrule->rpool.list);
2133 		}
2134 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGERULE_error; }
2135 
2136 		PF_RULES_WLOCK();
2137 		if (!(pcr->action == PF_CHANGE_REMOVE ||
2138 		    pcr->action == PF_CHANGE_GET_TICKET) &&
2139 		    pcr->pool_ticket != V_ticket_pabuf)
2140 			ERROUT(EBUSY);
2141 
2142 		ruleset = pf_find_kruleset(pcr->anchor);
2143 		if (ruleset == NULL)
2144 			ERROUT(EINVAL);
2145 
2146 		rs_num = pf_get_ruleset_number(pcr->rule.action);
2147 		if (rs_num >= PF_RULESET_MAX)
2148 			ERROUT(EINVAL);
2149 
2150 		if (pcr->action == PF_CHANGE_GET_TICKET) {
2151 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
2152 			ERROUT(0);
2153 		} else if (pcr->ticket !=
2154 			    ruleset->rules[rs_num].active.ticket)
2155 				ERROUT(EINVAL);
2156 
2157 		if (pcr->action != PF_CHANGE_REMOVE) {
2158 			if (newrule->ifname[0]) {
2159 				newrule->kif = pfi_kkif_attach(kif,
2160 				    newrule->ifname);
2161 				pfi_kkif_ref(newrule->kif);
2162 			} else
2163 				newrule->kif = NULL;
2164 
2165 			if (newrule->rtableid > 0 &&
2166 			    newrule->rtableid >= rt_numfibs)
2167 				error = EBUSY;
2168 
2169 #ifdef ALTQ
2170 			/* set queue IDs */
2171 			if (newrule->qname[0] != 0) {
2172 				if ((newrule->qid =
2173 				    pf_qname2qid(newrule->qname)) == 0)
2174 					error = EBUSY;
2175 				else if (newrule->pqname[0] != 0) {
2176 					if ((newrule->pqid =
2177 					    pf_qname2qid(newrule->pqname)) == 0)
2178 						error = EBUSY;
2179 				} else
2180 					newrule->pqid = newrule->qid;
2181 			}
2182 #endif /* ALTQ */
2183 			if (newrule->tagname[0])
2184 				if ((newrule->tag =
2185 				    pf_tagname2tag(newrule->tagname)) == 0)
2186 					error = EBUSY;
2187 			if (newrule->match_tagname[0])
2188 				if ((newrule->match_tag = pf_tagname2tag(
2189 				    newrule->match_tagname)) == 0)
2190 					error = EBUSY;
2191 			if (newrule->rt && !newrule->direction)
2192 				error = EINVAL;
2193 			if (!newrule->log)
2194 				newrule->logif = 0;
2195 			if (newrule->logif >= PFLOGIFS_MAX)
2196 				error = EINVAL;
2197 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
2198 				error = ENOMEM;
2199 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
2200 				error = ENOMEM;
2201 			if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
2202 				error = EINVAL;
2203 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
2204 				if (pa->addr.type == PF_ADDR_TABLE) {
2205 					pa->addr.p.tbl =
2206 					    pfr_attach_table(ruleset,
2207 					    pa->addr.v.tblname);
2208 					if (pa->addr.p.tbl == NULL)
2209 						error = ENOMEM;
2210 				}
2211 
2212 			newrule->overload_tbl = NULL;
2213 			if (newrule->overload_tblname[0]) {
2214 				if ((newrule->overload_tbl = pfr_attach_table(
2215 				    ruleset, newrule->overload_tblname)) ==
2216 				    NULL)
2217 					error = EINVAL;
2218 				else
2219 					newrule->overload_tbl->pfrkt_flags |=
2220 					    PFR_TFLAG_ACTIVE;
2221 			}
2222 
2223 			pf_mv_kpool(&V_pf_pabuf, &newrule->rpool.list);
2224 			if (((((newrule->action == PF_NAT) ||
2225 			    (newrule->action == PF_RDR) ||
2226 			    (newrule->action == PF_BINAT) ||
2227 			    (newrule->rt > PF_NOPFROUTE)) &&
2228 			    !newrule->anchor)) &&
2229 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
2230 				error = EINVAL;
2231 
2232 			if (error) {
2233 				pf_free_rule(newrule);
2234 				PF_RULES_WUNLOCK();
2235 				break;
2236 			}
2237 
2238 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
2239 		}
2240 		pf_empty_kpool(&V_pf_pabuf);
2241 
2242 		if (pcr->action == PF_CHANGE_ADD_HEAD)
2243 			oldrule = TAILQ_FIRST(
2244 			    ruleset->rules[rs_num].active.ptr);
2245 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
2246 			oldrule = TAILQ_LAST(
2247 			    ruleset->rules[rs_num].active.ptr, pf_krulequeue);
2248 		else {
2249 			oldrule = TAILQ_FIRST(
2250 			    ruleset->rules[rs_num].active.ptr);
2251 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
2252 				oldrule = TAILQ_NEXT(oldrule, entries);
2253 			if (oldrule == NULL) {
2254 				if (newrule != NULL)
2255 					pf_free_rule(newrule);
2256 				PF_RULES_WUNLOCK();
2257 				error = EINVAL;
2258 				break;
2259 			}
2260 		}
2261 
2262 		if (pcr->action == PF_CHANGE_REMOVE) {
2263 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
2264 			    oldrule);
2265 			ruleset->rules[rs_num].active.rcount--;
2266 		} else {
2267 			if (oldrule == NULL)
2268 				TAILQ_INSERT_TAIL(
2269 				    ruleset->rules[rs_num].active.ptr,
2270 				    newrule, entries);
2271 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
2272 			    pcr->action == PF_CHANGE_ADD_BEFORE)
2273 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
2274 			else
2275 				TAILQ_INSERT_AFTER(
2276 				    ruleset->rules[rs_num].active.ptr,
2277 				    oldrule, newrule, entries);
2278 			ruleset->rules[rs_num].active.rcount++;
2279 		}
2280 
2281 		nr = 0;
2282 		TAILQ_FOREACH(oldrule,
2283 		    ruleset->rules[rs_num].active.ptr, entries)
2284 			oldrule->nr = nr++;
2285 
2286 		ruleset->rules[rs_num].active.ticket++;
2287 
2288 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
2289 		pf_remove_if_empty_kruleset(ruleset);
2290 
2291 		PF_RULES_WUNLOCK();
2292 		break;
2293 
2294 #undef ERROUT
2295 DIOCCHANGERULE_error:
2296 		PF_RULES_WUNLOCK();
2297 		if (newrule != NULL) {
2298 			counter_u64_free(newrule->evaluations);
2299 			for (int i = 0; i < 2; i++) {
2300 				counter_u64_free(newrule->packets[i]);
2301 				counter_u64_free(newrule->bytes[i]);
2302 			}
2303 			counter_u64_free(newrule->states_cur);
2304 			counter_u64_free(newrule->states_tot);
2305 			counter_u64_free(newrule->src_nodes);
2306 			free(newrule, M_PFRULE);
2307 		}
2308 		if (kif != NULL)
2309 			pf_kkif_free(kif);
2310 		break;
2311 	}
2312 
2313 	case DIOCCLRSTATES: {
2314 		struct pf_state		*s;
2315 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
2316 		u_int			 i, killed = 0;
2317 
2318 		for (i = 0; i <= pf_hashmask; i++) {
2319 			struct pf_idhash *ih = &V_pf_idhash[i];
2320 
2321 relock_DIOCCLRSTATES:
2322 			PF_HASHROW_LOCK(ih);
2323 			LIST_FOREACH(s, &ih->states, entry)
2324 				if (!psk->psk_ifname[0] ||
2325 				    !strcmp(psk->psk_ifname,
2326 				    s->kif->pfik_name)) {
2327 					/*
2328 					 * Don't send out individual
2329 					 * delete messages.
2330 					 */
2331 					s->state_flags |= PFSTATE_NOSYNC;
2332 					pf_unlink_state(s, PF_ENTER_LOCKED);
2333 					killed++;
2334 					goto relock_DIOCCLRSTATES;
2335 				}
2336 			PF_HASHROW_UNLOCK(ih);
2337 		}
2338 		psk->psk_killed = killed;
2339 		if (V_pfsync_clear_states_ptr != NULL)
2340 			V_pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
2341 		break;
2342 	}
2343 
2344 	case DIOCKILLSTATES: {
2345 		struct pf_state		*s;
2346 		struct pf_state_key	*sk;
2347 		struct pf_addr		*srcaddr, *dstaddr;
2348 		u_int16_t		 srcport, dstport;
2349 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
2350 		u_int			 i, killed = 0;
2351 
2352 		if (psk->psk_pfcmp.id) {
2353 			if (psk->psk_pfcmp.creatorid == 0)
2354 				psk->psk_pfcmp.creatorid = V_pf_status.hostid;
2355 			if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
2356 			    psk->psk_pfcmp.creatorid))) {
2357 				pf_unlink_state(s, PF_ENTER_LOCKED);
2358 				psk->psk_killed = 1;
2359 			}
2360 			break;
2361 		}
2362 
2363 		for (i = 0; i <= pf_hashmask; i++) {
2364 			struct pf_idhash *ih = &V_pf_idhash[i];
2365 
2366 relock_DIOCKILLSTATES:
2367 			PF_HASHROW_LOCK(ih);
2368 			LIST_FOREACH(s, &ih->states, entry) {
2369 				sk = s->key[PF_SK_WIRE];
2370 				if (s->direction == PF_OUT) {
2371 					srcaddr = &sk->addr[1];
2372 					dstaddr = &sk->addr[0];
2373 					srcport = sk->port[1];
2374 					dstport = sk->port[0];
2375 				} else {
2376 					srcaddr = &sk->addr[0];
2377 					dstaddr = &sk->addr[1];
2378 					srcport = sk->port[0];
2379 					dstport = sk->port[1];
2380 				}
2381 
2382 				if ((!psk->psk_af || sk->af == psk->psk_af)
2383 				    && (!psk->psk_proto || psk->psk_proto ==
2384 				    sk->proto) &&
2385 				    PF_MATCHA(psk->psk_src.neg,
2386 				    &psk->psk_src.addr.v.a.addr,
2387 				    &psk->psk_src.addr.v.a.mask,
2388 				    srcaddr, sk->af) &&
2389 				    PF_MATCHA(psk->psk_dst.neg,
2390 				    &psk->psk_dst.addr.v.a.addr,
2391 				    &psk->psk_dst.addr.v.a.mask,
2392 				    dstaddr, sk->af) &&
2393 				    (psk->psk_src.port_op == 0 ||
2394 				    pf_match_port(psk->psk_src.port_op,
2395 				    psk->psk_src.port[0], psk->psk_src.port[1],
2396 				    srcport)) &&
2397 				    (psk->psk_dst.port_op == 0 ||
2398 				    pf_match_port(psk->psk_dst.port_op,
2399 				    psk->psk_dst.port[0], psk->psk_dst.port[1],
2400 				    dstport)) &&
2401 				    (!psk->psk_label[0] ||
2402 				    (s->rule.ptr->label[0] &&
2403 				    !strcmp(psk->psk_label,
2404 				    s->rule.ptr->label))) &&
2405 				    (!psk->psk_ifname[0] ||
2406 				    !strcmp(psk->psk_ifname,
2407 				    s->kif->pfik_name))) {
2408 					pf_unlink_state(s, PF_ENTER_LOCKED);
2409 					killed++;
2410 					goto relock_DIOCKILLSTATES;
2411 				}
2412 			}
2413 			PF_HASHROW_UNLOCK(ih);
2414 		}
2415 		psk->psk_killed = killed;
2416 		break;
2417 	}
2418 
2419 	case DIOCADDSTATE: {
2420 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
2421 		struct pfsync_state	*sp = &ps->state;
2422 
2423 		if (sp->timeout >= PFTM_MAX) {
2424 			error = EINVAL;
2425 			break;
2426 		}
2427 		if (V_pfsync_state_import_ptr != NULL) {
2428 			PF_RULES_RLOCK();
2429 			error = V_pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
2430 			PF_RULES_RUNLOCK();
2431 		} else
2432 			error = EOPNOTSUPP;
2433 		break;
2434 	}
2435 
2436 	case DIOCGETSTATE: {
2437 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
2438 		struct pf_state		*s;
2439 
2440 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
2441 		if (s == NULL) {
2442 			error = ENOENT;
2443 			break;
2444 		}
2445 
2446 		pfsync_state_export(&ps->state, s);
2447 		PF_STATE_UNLOCK(s);
2448 		break;
2449 	}
2450 
2451 	case DIOCGETSTATES: {
2452 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
2453 		struct pf_state		*s;
2454 		struct pfsync_state	*pstore, *p;
2455 		int i, nr;
2456 
2457 		if (ps->ps_len <= 0) {
2458 			nr = uma_zone_get_cur(V_pf_state_z);
2459 			ps->ps_len = sizeof(struct pfsync_state) * nr;
2460 			break;
2461 		}
2462 
2463 		p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK | M_ZERO);
2464 		nr = 0;
2465 
2466 		for (i = 0; i <= pf_hashmask; i++) {
2467 			struct pf_idhash *ih = &V_pf_idhash[i];
2468 
2469 			PF_HASHROW_LOCK(ih);
2470 			LIST_FOREACH(s, &ih->states, entry) {
2471 				if (s->timeout == PFTM_UNLINKED)
2472 					continue;
2473 
2474 				if ((nr+1) * sizeof(*p) > ps->ps_len) {
2475 					PF_HASHROW_UNLOCK(ih);
2476 					goto DIOCGETSTATES_full;
2477 				}
2478 				pfsync_state_export(p, s);
2479 				p++;
2480 				nr++;
2481 			}
2482 			PF_HASHROW_UNLOCK(ih);
2483 		}
2484 DIOCGETSTATES_full:
2485 		error = copyout(pstore, ps->ps_states,
2486 		    sizeof(struct pfsync_state) * nr);
2487 		if (error) {
2488 			free(pstore, M_TEMP);
2489 			break;
2490 		}
2491 		ps->ps_len = sizeof(struct pfsync_state) * nr;
2492 		free(pstore, M_TEMP);
2493 
2494 		break;
2495 	}
2496 
2497 	case DIOCGETSTATUS: {
2498 		struct pf_status *s = (struct pf_status *)addr;
2499 
2500 		PF_RULES_RLOCK();
2501 		s->running = V_pf_status.running;
2502 		s->since   = V_pf_status.since;
2503 		s->debug   = V_pf_status.debug;
2504 		s->hostid  = V_pf_status.hostid;
2505 		s->states  = V_pf_status.states;
2506 		s->src_nodes = V_pf_status.src_nodes;
2507 
2508 		for (int i = 0; i < PFRES_MAX; i++)
2509 			s->counters[i] =
2510 			    counter_u64_fetch(V_pf_status.counters[i]);
2511 		for (int i = 0; i < LCNT_MAX; i++)
2512 			s->lcounters[i] =
2513 			    counter_u64_fetch(V_pf_status.lcounters[i]);
2514 		for (int i = 0; i < FCNT_MAX; i++)
2515 			s->fcounters[i] =
2516 			    counter_u64_fetch(V_pf_status.fcounters[i]);
2517 		for (int i = 0; i < SCNT_MAX; i++)
2518 			s->scounters[i] =
2519 			    counter_u64_fetch(V_pf_status.scounters[i]);
2520 
2521 		bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ);
2522 		bcopy(V_pf_status.pf_chksum, s->pf_chksum,
2523 		    PF_MD5_DIGEST_LENGTH);
2524 
2525 		pfi_update_status(s->ifname, s);
2526 		PF_RULES_RUNLOCK();
2527 		break;
2528 	}
2529 
2530 	case DIOCSETSTATUSIF: {
2531 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
2532 
2533 		if (pi->ifname[0] == 0) {
2534 			bzero(V_pf_status.ifname, IFNAMSIZ);
2535 			break;
2536 		}
2537 		PF_RULES_WLOCK();
2538 		strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
2539 		PF_RULES_WUNLOCK();
2540 		break;
2541 	}
2542 
2543 	case DIOCCLRSTATUS: {
2544 		PF_RULES_WLOCK();
2545 		for (int i = 0; i < PFRES_MAX; i++)
2546 			counter_u64_zero(V_pf_status.counters[i]);
2547 		for (int i = 0; i < FCNT_MAX; i++)
2548 			counter_u64_zero(V_pf_status.fcounters[i]);
2549 		for (int i = 0; i < SCNT_MAX; i++)
2550 			counter_u64_zero(V_pf_status.scounters[i]);
2551 		for (int i = 0; i < LCNT_MAX; i++)
2552 			counter_u64_zero(V_pf_status.lcounters[i]);
2553 		V_pf_status.since = time_second;
2554 		if (*V_pf_status.ifname)
2555 			pfi_update_status(V_pf_status.ifname, NULL);
2556 		PF_RULES_WUNLOCK();
2557 		break;
2558 	}
2559 
2560 	case DIOCNATLOOK: {
2561 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
2562 		struct pf_state_key	*sk;
2563 		struct pf_state		*state;
2564 		struct pf_state_key_cmp	 key;
2565 		int			 m = 0, direction = pnl->direction;
2566 		int			 sidx, didx;
2567 
2568 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
2569 		sidx = (direction == PF_IN) ? 1 : 0;
2570 		didx = (direction == PF_IN) ? 0 : 1;
2571 
2572 		if (!pnl->proto ||
2573 		    PF_AZERO(&pnl->saddr, pnl->af) ||
2574 		    PF_AZERO(&pnl->daddr, pnl->af) ||
2575 		    ((pnl->proto == IPPROTO_TCP ||
2576 		    pnl->proto == IPPROTO_UDP) &&
2577 		    (!pnl->dport || !pnl->sport)))
2578 			error = EINVAL;
2579 		else {
2580 			bzero(&key, sizeof(key));
2581 			key.af = pnl->af;
2582 			key.proto = pnl->proto;
2583 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
2584 			key.port[sidx] = pnl->sport;
2585 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
2586 			key.port[didx] = pnl->dport;
2587 
2588 			state = pf_find_state_all(&key, direction, &m);
2589 
2590 			if (m > 1)
2591 				error = E2BIG;	/* more than one state */
2592 			else if (state != NULL) {
2593 				/* XXXGL: not locked read */
2594 				sk = state->key[sidx];
2595 				PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
2596 				pnl->rsport = sk->port[sidx];
2597 				PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
2598 				pnl->rdport = sk->port[didx];
2599 			} else
2600 				error = ENOENT;
2601 		}
2602 		break;
2603 	}
2604 
2605 	case DIOCSETTIMEOUT: {
2606 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2607 		int		 old;
2608 
2609 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
2610 		    pt->seconds < 0) {
2611 			error = EINVAL;
2612 			break;
2613 		}
2614 		PF_RULES_WLOCK();
2615 		old = V_pf_default_rule.timeout[pt->timeout];
2616 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
2617 			pt->seconds = 1;
2618 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
2619 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
2620 			wakeup(pf_purge_thread);
2621 		pt->seconds = old;
2622 		PF_RULES_WUNLOCK();
2623 		break;
2624 	}
2625 
2626 	case DIOCGETTIMEOUT: {
2627 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2628 
2629 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
2630 			error = EINVAL;
2631 			break;
2632 		}
2633 		PF_RULES_RLOCK();
2634 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
2635 		PF_RULES_RUNLOCK();
2636 		break;
2637 	}
2638 
2639 	case DIOCGETLIMIT: {
2640 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2641 
2642 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
2643 			error = EINVAL;
2644 			break;
2645 		}
2646 		PF_RULES_RLOCK();
2647 		pl->limit = V_pf_limits[pl->index].limit;
2648 		PF_RULES_RUNLOCK();
2649 		break;
2650 	}
2651 
2652 	case DIOCSETLIMIT: {
2653 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2654 		int			 old_limit;
2655 
2656 		PF_RULES_WLOCK();
2657 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
2658 		    V_pf_limits[pl->index].zone == NULL) {
2659 			PF_RULES_WUNLOCK();
2660 			error = EINVAL;
2661 			break;
2662 		}
2663 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
2664 		old_limit = V_pf_limits[pl->index].limit;
2665 		V_pf_limits[pl->index].limit = pl->limit;
2666 		pl->limit = old_limit;
2667 		PF_RULES_WUNLOCK();
2668 		break;
2669 	}
2670 
2671 	case DIOCSETDEBUG: {
2672 		u_int32_t	*level = (u_int32_t *)addr;
2673 
2674 		PF_RULES_WLOCK();
2675 		V_pf_status.debug = *level;
2676 		PF_RULES_WUNLOCK();
2677 		break;
2678 	}
2679 
2680 	case DIOCCLRRULECTRS: {
2681 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
2682 		struct pf_kruleset	*ruleset = &pf_main_ruleset;
2683 		struct pf_krule		*rule;
2684 
2685 		PF_RULES_WLOCK();
2686 		TAILQ_FOREACH(rule,
2687 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
2688 			counter_u64_zero(rule->evaluations);
2689 			for (int i = 0; i < 2; i++) {
2690 				counter_u64_zero(rule->packets[i]);
2691 				counter_u64_zero(rule->bytes[i]);
2692 			}
2693 		}
2694 		PF_RULES_WUNLOCK();
2695 		break;
2696 	}
2697 
2698 	case DIOCGIFSPEEDV0:
2699 	case DIOCGIFSPEEDV1: {
2700 		struct pf_ifspeed_v1	*psp = (struct pf_ifspeed_v1 *)addr;
2701 		struct pf_ifspeed_v1	ps;
2702 		struct ifnet		*ifp;
2703 
2704 		if (psp->ifname[0] != 0) {
2705 			/* Can we completely trust user-land? */
2706 			strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
2707 			ifp = ifunit(ps.ifname);
2708 			if (ifp != NULL) {
2709 				psp->baudrate32 =
2710 				    (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
2711 				if (cmd == DIOCGIFSPEEDV1)
2712 					psp->baudrate = ifp->if_baudrate;
2713 			} else
2714 				error = EINVAL;
2715 		} else
2716 			error = EINVAL;
2717 		break;
2718 	}
2719 
2720 #ifdef ALTQ
2721 	case DIOCSTARTALTQ: {
2722 		struct pf_altq		*altq;
2723 
2724 		PF_RULES_WLOCK();
2725 		/* enable all altq interfaces on active list */
2726 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
2727 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
2728 				error = pf_enable_altq(altq);
2729 				if (error != 0)
2730 					break;
2731 			}
2732 		}
2733 		if (error == 0)
2734 			V_pf_altq_running = 1;
2735 		PF_RULES_WUNLOCK();
2736 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
2737 		break;
2738 	}
2739 
2740 	case DIOCSTOPALTQ: {
2741 		struct pf_altq		*altq;
2742 
2743 		PF_RULES_WLOCK();
2744 		/* disable all altq interfaces on active list */
2745 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
2746 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
2747 				error = pf_disable_altq(altq);
2748 				if (error != 0)
2749 					break;
2750 			}
2751 		}
2752 		if (error == 0)
2753 			V_pf_altq_running = 0;
2754 		PF_RULES_WUNLOCK();
2755 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2756 		break;
2757 	}
2758 
2759 	case DIOCADDALTQV0:
2760 	case DIOCADDALTQV1: {
2761 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2762 		struct pf_altq		*altq, *a;
2763 		struct ifnet		*ifp;
2764 
2765 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
2766 		error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
2767 		if (error)
2768 			break;
2769 		altq->local_flags = 0;
2770 
2771 		PF_RULES_WLOCK();
2772 		if (pa->ticket != V_ticket_altqs_inactive) {
2773 			PF_RULES_WUNLOCK();
2774 			free(altq, M_PFALTQ);
2775 			error = EBUSY;
2776 			break;
2777 		}
2778 
2779 		/*
2780 		 * if this is for a queue, find the discipline and
2781 		 * copy the necessary fields
2782 		 */
2783 		if (altq->qname[0] != 0) {
2784 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2785 				PF_RULES_WUNLOCK();
2786 				error = EBUSY;
2787 				free(altq, M_PFALTQ);
2788 				break;
2789 			}
2790 			altq->altq_disc = NULL;
2791 			TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
2792 				if (strncmp(a->ifname, altq->ifname,
2793 				    IFNAMSIZ) == 0) {
2794 					altq->altq_disc = a->altq_disc;
2795 					break;
2796 				}
2797 			}
2798 		}
2799 
2800 		if ((ifp = ifunit(altq->ifname)) == NULL)
2801 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
2802 		else
2803 			error = altq_add(ifp, altq);
2804 
2805 		if (error) {
2806 			PF_RULES_WUNLOCK();
2807 			free(altq, M_PFALTQ);
2808 			break;
2809 		}
2810 
2811 		if (altq->qname[0] != 0)
2812 			TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
2813 		else
2814 			TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
2815 		/* version error check done on import above */
2816 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
2817 		PF_RULES_WUNLOCK();
2818 		break;
2819 	}
2820 
2821 	case DIOCGETALTQSV0:
2822 	case DIOCGETALTQSV1: {
2823 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2824 		struct pf_altq		*altq;
2825 
2826 		PF_RULES_RLOCK();
2827 		pa->nr = 0;
2828 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
2829 			pa->nr++;
2830 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
2831 			pa->nr++;
2832 		pa->ticket = V_ticket_altqs_active;
2833 		PF_RULES_RUNLOCK();
2834 		break;
2835 	}
2836 
2837 	case DIOCGETALTQV0:
2838 	case DIOCGETALTQV1: {
2839 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2840 		struct pf_altq		*altq;
2841 
2842 		PF_RULES_RLOCK();
2843 		if (pa->ticket != V_ticket_altqs_active) {
2844 			PF_RULES_RUNLOCK();
2845 			error = EBUSY;
2846 			break;
2847 		}
2848 		altq = pf_altq_get_nth_active(pa->nr);
2849 		if (altq == NULL) {
2850 			PF_RULES_RUNLOCK();
2851 			error = EBUSY;
2852 			break;
2853 		}
2854 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
2855 		PF_RULES_RUNLOCK();
2856 		break;
2857 	}
2858 
2859 	case DIOCCHANGEALTQV0:
2860 	case DIOCCHANGEALTQV1:
2861 		/* CHANGEALTQ not supported yet! */
2862 		error = ENODEV;
2863 		break;
2864 
2865 	case DIOCGETQSTATSV0:
2866 	case DIOCGETQSTATSV1: {
2867 		struct pfioc_qstats_v1	*pq = (struct pfioc_qstats_v1 *)addr;
2868 		struct pf_altq		*altq;
2869 		int			 nbytes;
2870 		u_int32_t		 version;
2871 
2872 		PF_RULES_RLOCK();
2873 		if (pq->ticket != V_ticket_altqs_active) {
2874 			PF_RULES_RUNLOCK();
2875 			error = EBUSY;
2876 			break;
2877 		}
2878 		nbytes = pq->nbytes;
2879 		altq = pf_altq_get_nth_active(pq->nr);
2880 		if (altq == NULL) {
2881 			PF_RULES_RUNLOCK();
2882 			error = EBUSY;
2883 			break;
2884 		}
2885 
2886 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
2887 			PF_RULES_RUNLOCK();
2888 			error = ENXIO;
2889 			break;
2890 		}
2891 		PF_RULES_RUNLOCK();
2892 		if (cmd == DIOCGETQSTATSV0)
2893 			version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
2894 		else
2895 			version = pq->version;
2896 		error = altq_getqstats(altq, pq->buf, &nbytes, version);
2897 		if (error == 0) {
2898 			pq->scheduler = altq->scheduler;
2899 			pq->nbytes = nbytes;
2900 		}
2901 		break;
2902 	}
2903 #endif /* ALTQ */
2904 
2905 	case DIOCBEGINADDRS: {
2906 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2907 
2908 		PF_RULES_WLOCK();
2909 		pf_empty_kpool(&V_pf_pabuf);
2910 		pp->ticket = ++V_ticket_pabuf;
2911 		PF_RULES_WUNLOCK();
2912 		break;
2913 	}
2914 
2915 	case DIOCADDADDR: {
2916 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2917 		struct pf_kpooladdr	*pa;
2918 		struct pfi_kkif		*kif = NULL;
2919 
2920 #ifndef INET
2921 		if (pp->af == AF_INET) {
2922 			error = EAFNOSUPPORT;
2923 			break;
2924 		}
2925 #endif /* INET */
2926 #ifndef INET6
2927 		if (pp->af == AF_INET6) {
2928 			error = EAFNOSUPPORT;
2929 			break;
2930 		}
2931 #endif /* INET6 */
2932 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2933 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2934 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2935 			error = EINVAL;
2936 			break;
2937 		}
2938 		if (pp->addr.addr.p.dyn != NULL) {
2939 			error = EINVAL;
2940 			break;
2941 		}
2942 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2943 		pf_pooladdr_to_kpooladdr(&pp->addr, pa);
2944 		if (pa->ifname[0])
2945 			kif = pf_kkif_create(M_WAITOK);
2946 		PF_RULES_WLOCK();
2947 		if (pp->ticket != V_ticket_pabuf) {
2948 			PF_RULES_WUNLOCK();
2949 			if (pa->ifname[0])
2950 				pf_kkif_free(kif);
2951 			free(pa, M_PFRULE);
2952 			error = EBUSY;
2953 			break;
2954 		}
2955 		if (pa->ifname[0]) {
2956 			pa->kif = pfi_kkif_attach(kif, pa->ifname);
2957 			pfi_kkif_ref(pa->kif);
2958 		} else
2959 			pa->kif = NULL;
2960 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2961 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2962 			if (pa->ifname[0])
2963 				pfi_kkif_unref(pa->kif);
2964 			PF_RULES_WUNLOCK();
2965 			free(pa, M_PFRULE);
2966 			break;
2967 		}
2968 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2969 		PF_RULES_WUNLOCK();
2970 		break;
2971 	}
2972 
2973 	case DIOCGETADDRS: {
2974 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2975 		struct pf_kpool		*pool;
2976 		struct pf_kpooladdr	*pa;
2977 
2978 		PF_RULES_RLOCK();
2979 		pp->nr = 0;
2980 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2981 		    pp->r_num, 0, 1, 0);
2982 		if (pool == NULL) {
2983 			PF_RULES_RUNLOCK();
2984 			error = EBUSY;
2985 			break;
2986 		}
2987 		TAILQ_FOREACH(pa, &pool->list, entries)
2988 			pp->nr++;
2989 		PF_RULES_RUNLOCK();
2990 		break;
2991 	}
2992 
2993 	case DIOCGETADDR: {
2994 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2995 		struct pf_kpool		*pool;
2996 		struct pf_kpooladdr	*pa;
2997 		u_int32_t		 nr = 0;
2998 
2999 		PF_RULES_RLOCK();
3000 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
3001 		    pp->r_num, 0, 1, 1);
3002 		if (pool == NULL) {
3003 			PF_RULES_RUNLOCK();
3004 			error = EBUSY;
3005 			break;
3006 		}
3007 		pa = TAILQ_FIRST(&pool->list);
3008 		while ((pa != NULL) && (nr < pp->nr)) {
3009 			pa = TAILQ_NEXT(pa, entries);
3010 			nr++;
3011 		}
3012 		if (pa == NULL) {
3013 			PF_RULES_RUNLOCK();
3014 			error = EBUSY;
3015 			break;
3016 		}
3017 		pf_kpooladdr_to_pooladdr(pa, &pp->addr);
3018 		pf_addr_copyout(&pp->addr.addr);
3019 		PF_RULES_RUNLOCK();
3020 		break;
3021 	}
3022 
3023 	case DIOCCHANGEADDR: {
3024 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
3025 		struct pf_kpool		*pool;
3026 		struct pf_kpooladdr	*oldpa = NULL, *newpa = NULL;
3027 		struct pf_kruleset	*ruleset;
3028 		struct pfi_kkif		*kif = NULL;
3029 
3030 		if (pca->action < PF_CHANGE_ADD_HEAD ||
3031 		    pca->action > PF_CHANGE_REMOVE) {
3032 			error = EINVAL;
3033 			break;
3034 		}
3035 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
3036 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
3037 		    pca->addr.addr.type != PF_ADDR_TABLE) {
3038 			error = EINVAL;
3039 			break;
3040 		}
3041 		if (pca->addr.addr.p.dyn != NULL) {
3042 			error = EINVAL;
3043 			break;
3044 		}
3045 
3046 		if (pca->action != PF_CHANGE_REMOVE) {
3047 #ifndef INET
3048 			if (pca->af == AF_INET) {
3049 				error = EAFNOSUPPORT;
3050 				break;
3051 			}
3052 #endif /* INET */
3053 #ifndef INET6
3054 			if (pca->af == AF_INET6) {
3055 				error = EAFNOSUPPORT;
3056 				break;
3057 			}
3058 #endif /* INET6 */
3059 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
3060 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
3061 			if (newpa->ifname[0])
3062 				kif = pf_kkif_create(M_WAITOK);
3063 			newpa->kif = NULL;
3064 		}
3065 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGEADDR_error; }
3066 		PF_RULES_WLOCK();
3067 		ruleset = pf_find_kruleset(pca->anchor);
3068 		if (ruleset == NULL)
3069 			ERROUT(EBUSY);
3070 
3071 		pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
3072 		    pca->r_num, pca->r_last, 1, 1);
3073 		if (pool == NULL)
3074 			ERROUT(EBUSY);
3075 
3076 		if (pca->action != PF_CHANGE_REMOVE) {
3077 			if (newpa->ifname[0]) {
3078 				newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
3079 				pfi_kkif_ref(newpa->kif);
3080 				kif = NULL;
3081 			}
3082 
3083 			switch (newpa->addr.type) {
3084 			case PF_ADDR_DYNIFTL:
3085 				error = pfi_dynaddr_setup(&newpa->addr,
3086 				    pca->af);
3087 				break;
3088 			case PF_ADDR_TABLE:
3089 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
3090 				    newpa->addr.v.tblname);
3091 				if (newpa->addr.p.tbl == NULL)
3092 					error = ENOMEM;
3093 				break;
3094 			}
3095 			if (error)
3096 				goto DIOCCHANGEADDR_error;
3097 		}
3098 
3099 		switch (pca->action) {
3100 		case PF_CHANGE_ADD_HEAD:
3101 			oldpa = TAILQ_FIRST(&pool->list);
3102 			break;
3103 		case PF_CHANGE_ADD_TAIL:
3104 			oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
3105 			break;
3106 		default:
3107 			oldpa = TAILQ_FIRST(&pool->list);
3108 			for (int i = 0; oldpa && i < pca->nr; i++)
3109 				oldpa = TAILQ_NEXT(oldpa, entries);
3110 
3111 			if (oldpa == NULL)
3112 				ERROUT(EINVAL);
3113 		}
3114 
3115 		if (pca->action == PF_CHANGE_REMOVE) {
3116 			TAILQ_REMOVE(&pool->list, oldpa, entries);
3117 			switch (oldpa->addr.type) {
3118 			case PF_ADDR_DYNIFTL:
3119 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
3120 				break;
3121 			case PF_ADDR_TABLE:
3122 				pfr_detach_table(oldpa->addr.p.tbl);
3123 				break;
3124 			}
3125 			if (oldpa->kif)
3126 				pfi_kkif_unref(oldpa->kif);
3127 			free(oldpa, M_PFRULE);
3128 		} else {
3129 			if (oldpa == NULL)
3130 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
3131 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
3132 			    pca->action == PF_CHANGE_ADD_BEFORE)
3133 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
3134 			else
3135 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
3136 				    newpa, entries);
3137 		}
3138 
3139 		pool->cur = TAILQ_FIRST(&pool->list);
3140 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
3141 		PF_RULES_WUNLOCK();
3142 		break;
3143 
3144 #undef ERROUT
3145 DIOCCHANGEADDR_error:
3146 		if (newpa != NULL) {
3147 			if (newpa->kif)
3148 				pfi_kkif_unref(newpa->kif);
3149 			free(newpa, M_PFRULE);
3150 		}
3151 		PF_RULES_WUNLOCK();
3152 		if (kif != NULL)
3153 			pf_kkif_free(kif);
3154 		break;
3155 	}
3156 
3157 	case DIOCGETRULESETS: {
3158 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
3159 		struct pf_kruleset	*ruleset;
3160 		struct pf_kanchor	*anchor;
3161 
3162 		PF_RULES_RLOCK();
3163 		pr->path[sizeof(pr->path) - 1] = 0;
3164 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
3165 			PF_RULES_RUNLOCK();
3166 			error = ENOENT;
3167 			break;
3168 		}
3169 		pr->nr = 0;
3170 		if (ruleset->anchor == NULL) {
3171 			/* XXX kludge for pf_main_ruleset */
3172 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
3173 				if (anchor->parent == NULL)
3174 					pr->nr++;
3175 		} else {
3176 			RB_FOREACH(anchor, pf_kanchor_node,
3177 			    &ruleset->anchor->children)
3178 				pr->nr++;
3179 		}
3180 		PF_RULES_RUNLOCK();
3181 		break;
3182 	}
3183 
3184 	case DIOCGETRULESET: {
3185 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
3186 		struct pf_kruleset	*ruleset;
3187 		struct pf_kanchor	*anchor;
3188 		u_int32_t		 nr = 0;
3189 
3190 		PF_RULES_RLOCK();
3191 		pr->path[sizeof(pr->path) - 1] = 0;
3192 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
3193 			PF_RULES_RUNLOCK();
3194 			error = ENOENT;
3195 			break;
3196 		}
3197 		pr->name[0] = 0;
3198 		if (ruleset->anchor == NULL) {
3199 			/* XXX kludge for pf_main_ruleset */
3200 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
3201 				if (anchor->parent == NULL && nr++ == pr->nr) {
3202 					strlcpy(pr->name, anchor->name,
3203 					    sizeof(pr->name));
3204 					break;
3205 				}
3206 		} else {
3207 			RB_FOREACH(anchor, pf_kanchor_node,
3208 			    &ruleset->anchor->children)
3209 				if (nr++ == pr->nr) {
3210 					strlcpy(pr->name, anchor->name,
3211 					    sizeof(pr->name));
3212 					break;
3213 				}
3214 		}
3215 		if (!pr->name[0])
3216 			error = EBUSY;
3217 		PF_RULES_RUNLOCK();
3218 		break;
3219 	}
3220 
3221 	case DIOCRCLRTABLES: {
3222 		struct pfioc_table *io = (struct pfioc_table *)addr;
3223 
3224 		if (io->pfrio_esize != 0) {
3225 			error = ENODEV;
3226 			break;
3227 		}
3228 		PF_RULES_WLOCK();
3229 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
3230 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
3231 		PF_RULES_WUNLOCK();
3232 		break;
3233 	}
3234 
3235 	case DIOCRADDTABLES: {
3236 		struct pfioc_table *io = (struct pfioc_table *)addr;
3237 		struct pfr_table *pfrts;
3238 		size_t totlen;
3239 
3240 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3241 			error = ENODEV;
3242 			break;
3243 		}
3244 
3245 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
3246 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
3247 			error = ENOMEM;
3248 			break;
3249 		}
3250 
3251 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3252 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3253 		    M_TEMP, M_WAITOK);
3254 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3255 		if (error) {
3256 			free(pfrts, M_TEMP);
3257 			break;
3258 		}
3259 		PF_RULES_WLOCK();
3260 		error = pfr_add_tables(pfrts, io->pfrio_size,
3261 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3262 		PF_RULES_WUNLOCK();
3263 		free(pfrts, M_TEMP);
3264 		break;
3265 	}
3266 
3267 	case DIOCRDELTABLES: {
3268 		struct pfioc_table *io = (struct pfioc_table *)addr;
3269 		struct pfr_table *pfrts;
3270 		size_t totlen;
3271 
3272 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3273 			error = ENODEV;
3274 			break;
3275 		}
3276 
3277 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
3278 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
3279 			error = ENOMEM;
3280 			break;
3281 		}
3282 
3283 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3284 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3285 		    M_TEMP, M_WAITOK);
3286 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3287 		if (error) {
3288 			free(pfrts, M_TEMP);
3289 			break;
3290 		}
3291 		PF_RULES_WLOCK();
3292 		error = pfr_del_tables(pfrts, io->pfrio_size,
3293 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3294 		PF_RULES_WUNLOCK();
3295 		free(pfrts, M_TEMP);
3296 		break;
3297 	}
3298 
3299 	case DIOCRGETTABLES: {
3300 		struct pfioc_table *io = (struct pfioc_table *)addr;
3301 		struct pfr_table *pfrts;
3302 		size_t totlen;
3303 		int n;
3304 
3305 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3306 			error = ENODEV;
3307 			break;
3308 		}
3309 		PF_RULES_RLOCK();
3310 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3311 		if (n < 0) {
3312 			PF_RULES_RUNLOCK();
3313 			error = EINVAL;
3314 			break;
3315 		}
3316 		io->pfrio_size = min(io->pfrio_size, n);
3317 
3318 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3319 
3320 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3321 		    M_TEMP, M_NOWAIT);
3322 		if (pfrts == NULL) {
3323 			error = ENOMEM;
3324 			PF_RULES_RUNLOCK();
3325 			break;
3326 		}
3327 		error = pfr_get_tables(&io->pfrio_table, pfrts,
3328 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3329 		PF_RULES_RUNLOCK();
3330 		if (error == 0)
3331 			error = copyout(pfrts, io->pfrio_buffer, totlen);
3332 		free(pfrts, M_TEMP);
3333 		break;
3334 	}
3335 
3336 	case DIOCRGETTSTATS: {
3337 		struct pfioc_table *io = (struct pfioc_table *)addr;
3338 		struct pfr_tstats *pfrtstats;
3339 		size_t totlen;
3340 		int n;
3341 
3342 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
3343 			error = ENODEV;
3344 			break;
3345 		}
3346 		PF_RULES_WLOCK();
3347 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3348 		if (n < 0) {
3349 			PF_RULES_WUNLOCK();
3350 			error = EINVAL;
3351 			break;
3352 		}
3353 		io->pfrio_size = min(io->pfrio_size, n);
3354 
3355 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
3356 		pfrtstats = mallocarray(io->pfrio_size,
3357 		    sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT);
3358 		if (pfrtstats == NULL) {
3359 			error = ENOMEM;
3360 			PF_RULES_WUNLOCK();
3361 			break;
3362 		}
3363 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
3364 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3365 		PF_RULES_WUNLOCK();
3366 		if (error == 0)
3367 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
3368 		free(pfrtstats, M_TEMP);
3369 		break;
3370 	}
3371 
3372 	case DIOCRCLRTSTATS: {
3373 		struct pfioc_table *io = (struct pfioc_table *)addr;
3374 		struct pfr_table *pfrts;
3375 		size_t totlen;
3376 
3377 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3378 			error = ENODEV;
3379 			break;
3380 		}
3381 
3382 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
3383 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
3384 			/* We used to count tables and use the minimum required
3385 			 * size, so we didn't fail on overly large requests.
3386 			 * Keep doing so. */
3387 			io->pfrio_size = pf_ioctl_maxcount;
3388 			break;
3389 		}
3390 
3391 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3392 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3393 		    M_TEMP, M_NOWAIT);
3394 		if (pfrts == NULL) {
3395 			error = ENOMEM;
3396 			break;
3397 		}
3398 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3399 		if (error) {
3400 			free(pfrts, M_TEMP);
3401 			break;
3402 		}
3403 
3404 		PF_RULES_WLOCK();
3405 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
3406 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3407 		PF_RULES_WUNLOCK();
3408 		free(pfrts, M_TEMP);
3409 		break;
3410 	}
3411 
3412 	case DIOCRSETTFLAGS: {
3413 		struct pfioc_table *io = (struct pfioc_table *)addr;
3414 		struct pfr_table *pfrts;
3415 		size_t totlen;
3416 		int n;
3417 
3418 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3419 			error = ENODEV;
3420 			break;
3421 		}
3422 
3423 		PF_RULES_RLOCK();
3424 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3425 		if (n < 0) {
3426 			PF_RULES_RUNLOCK();
3427 			error = EINVAL;
3428 			break;
3429 		}
3430 
3431 		io->pfrio_size = min(io->pfrio_size, n);
3432 		PF_RULES_RUNLOCK();
3433 
3434 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3435 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3436 		    M_TEMP, M_WAITOK);
3437 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3438 		if (error) {
3439 			free(pfrts, M_TEMP);
3440 			break;
3441 		}
3442 		PF_RULES_WLOCK();
3443 		error = pfr_set_tflags(pfrts, io->pfrio_size,
3444 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
3445 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3446 		PF_RULES_WUNLOCK();
3447 		free(pfrts, M_TEMP);
3448 		break;
3449 	}
3450 
3451 	case DIOCRCLRADDRS: {
3452 		struct pfioc_table *io = (struct pfioc_table *)addr;
3453 
3454 		if (io->pfrio_esize != 0) {
3455 			error = ENODEV;
3456 			break;
3457 		}
3458 		PF_RULES_WLOCK();
3459 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
3460 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
3461 		PF_RULES_WUNLOCK();
3462 		break;
3463 	}
3464 
3465 	case DIOCRADDADDRS: {
3466 		struct pfioc_table *io = (struct pfioc_table *)addr;
3467 		struct pfr_addr *pfras;
3468 		size_t totlen;
3469 
3470 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3471 			error = ENODEV;
3472 			break;
3473 		}
3474 		if (io->pfrio_size < 0 ||
3475 		    io->pfrio_size > pf_ioctl_maxcount ||
3476 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3477 			error = EINVAL;
3478 			break;
3479 		}
3480 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3481 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3482 		    M_TEMP, M_NOWAIT);
3483 		if (! pfras) {
3484 			error = ENOMEM;
3485 			break;
3486 		}
3487 		error = copyin(io->pfrio_buffer, pfras, totlen);
3488 		if (error) {
3489 			free(pfras, M_TEMP);
3490 			break;
3491 		}
3492 		PF_RULES_WLOCK();
3493 		error = pfr_add_addrs(&io->pfrio_table, pfras,
3494 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
3495 		    PFR_FLAG_USERIOCTL);
3496 		PF_RULES_WUNLOCK();
3497 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3498 			error = copyout(pfras, io->pfrio_buffer, totlen);
3499 		free(pfras, M_TEMP);
3500 		break;
3501 	}
3502 
3503 	case DIOCRDELADDRS: {
3504 		struct pfioc_table *io = (struct pfioc_table *)addr;
3505 		struct pfr_addr *pfras;
3506 		size_t totlen;
3507 
3508 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3509 			error = ENODEV;
3510 			break;
3511 		}
3512 		if (io->pfrio_size < 0 ||
3513 		    io->pfrio_size > pf_ioctl_maxcount ||
3514 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3515 			error = EINVAL;
3516 			break;
3517 		}
3518 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3519 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3520 		    M_TEMP, M_NOWAIT);
3521 		if (! pfras) {
3522 			error = ENOMEM;
3523 			break;
3524 		}
3525 		error = copyin(io->pfrio_buffer, pfras, totlen);
3526 		if (error) {
3527 			free(pfras, M_TEMP);
3528 			break;
3529 		}
3530 		PF_RULES_WLOCK();
3531 		error = pfr_del_addrs(&io->pfrio_table, pfras,
3532 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
3533 		    PFR_FLAG_USERIOCTL);
3534 		PF_RULES_WUNLOCK();
3535 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3536 			error = copyout(pfras, io->pfrio_buffer, totlen);
3537 		free(pfras, M_TEMP);
3538 		break;
3539 	}
3540 
3541 	case DIOCRSETADDRS: {
3542 		struct pfioc_table *io = (struct pfioc_table *)addr;
3543 		struct pfr_addr *pfras;
3544 		size_t totlen, count;
3545 
3546 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3547 			error = ENODEV;
3548 			break;
3549 		}
3550 		if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
3551 			error = EINVAL;
3552 			break;
3553 		}
3554 		count = max(io->pfrio_size, io->pfrio_size2);
3555 		if (count > pf_ioctl_maxcount ||
3556 		    WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
3557 			error = EINVAL;
3558 			break;
3559 		}
3560 		totlen = count * sizeof(struct pfr_addr);
3561 		pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
3562 		    M_NOWAIT);
3563 		if (! pfras) {
3564 			error = ENOMEM;
3565 			break;
3566 		}
3567 		error = copyin(io->pfrio_buffer, pfras, totlen);
3568 		if (error) {
3569 			free(pfras, M_TEMP);
3570 			break;
3571 		}
3572 		PF_RULES_WLOCK();
3573 		error = pfr_set_addrs(&io->pfrio_table, pfras,
3574 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
3575 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
3576 		    PFR_FLAG_USERIOCTL, 0);
3577 		PF_RULES_WUNLOCK();
3578 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3579 			error = copyout(pfras, io->pfrio_buffer, totlen);
3580 		free(pfras, M_TEMP);
3581 		break;
3582 	}
3583 
3584 	case DIOCRGETADDRS: {
3585 		struct pfioc_table *io = (struct pfioc_table *)addr;
3586 		struct pfr_addr *pfras;
3587 		size_t totlen;
3588 
3589 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3590 			error = ENODEV;
3591 			break;
3592 		}
3593 		if (io->pfrio_size < 0 ||
3594 		    io->pfrio_size > pf_ioctl_maxcount ||
3595 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3596 			error = EINVAL;
3597 			break;
3598 		}
3599 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3600 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3601 		    M_TEMP, M_NOWAIT);
3602 		if (! pfras) {
3603 			error = ENOMEM;
3604 			break;
3605 		}
3606 		PF_RULES_RLOCK();
3607 		error = pfr_get_addrs(&io->pfrio_table, pfras,
3608 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3609 		PF_RULES_RUNLOCK();
3610 		if (error == 0)
3611 			error = copyout(pfras, io->pfrio_buffer, totlen);
3612 		free(pfras, M_TEMP);
3613 		break;
3614 	}
3615 
3616 	case DIOCRGETASTATS: {
3617 		struct pfioc_table *io = (struct pfioc_table *)addr;
3618 		struct pfr_astats *pfrastats;
3619 		size_t totlen;
3620 
3621 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
3622 			error = ENODEV;
3623 			break;
3624 		}
3625 		if (io->pfrio_size < 0 ||
3626 		    io->pfrio_size > pf_ioctl_maxcount ||
3627 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
3628 			error = EINVAL;
3629 			break;
3630 		}
3631 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
3632 		pfrastats = mallocarray(io->pfrio_size,
3633 		    sizeof(struct pfr_astats), M_TEMP, M_NOWAIT);
3634 		if (! pfrastats) {
3635 			error = ENOMEM;
3636 			break;
3637 		}
3638 		PF_RULES_RLOCK();
3639 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
3640 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3641 		PF_RULES_RUNLOCK();
3642 		if (error == 0)
3643 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
3644 		free(pfrastats, M_TEMP);
3645 		break;
3646 	}
3647 
3648 	case DIOCRCLRASTATS: {
3649 		struct pfioc_table *io = (struct pfioc_table *)addr;
3650 		struct pfr_addr *pfras;
3651 		size_t totlen;
3652 
3653 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3654 			error = ENODEV;
3655 			break;
3656 		}
3657 		if (io->pfrio_size < 0 ||
3658 		    io->pfrio_size > pf_ioctl_maxcount ||
3659 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3660 			error = EINVAL;
3661 			break;
3662 		}
3663 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3664 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3665 		    M_TEMP, M_NOWAIT);
3666 		if (! pfras) {
3667 			error = ENOMEM;
3668 			break;
3669 		}
3670 		error = copyin(io->pfrio_buffer, pfras, totlen);
3671 		if (error) {
3672 			free(pfras, M_TEMP);
3673 			break;
3674 		}
3675 		PF_RULES_WLOCK();
3676 		error = pfr_clr_astats(&io->pfrio_table, pfras,
3677 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
3678 		    PFR_FLAG_USERIOCTL);
3679 		PF_RULES_WUNLOCK();
3680 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3681 			error = copyout(pfras, io->pfrio_buffer, totlen);
3682 		free(pfras, M_TEMP);
3683 		break;
3684 	}
3685 
3686 	case DIOCRTSTADDRS: {
3687 		struct pfioc_table *io = (struct pfioc_table *)addr;
3688 		struct pfr_addr *pfras;
3689 		size_t totlen;
3690 
3691 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3692 			error = ENODEV;
3693 			break;
3694 		}
3695 		if (io->pfrio_size < 0 ||
3696 		    io->pfrio_size > pf_ioctl_maxcount ||
3697 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3698 			error = EINVAL;
3699 			break;
3700 		}
3701 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3702 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3703 		    M_TEMP, M_NOWAIT);
3704 		if (! pfras) {
3705 			error = ENOMEM;
3706 			break;
3707 		}
3708 		error = copyin(io->pfrio_buffer, pfras, totlen);
3709 		if (error) {
3710 			free(pfras, M_TEMP);
3711 			break;
3712 		}
3713 		PF_RULES_RLOCK();
3714 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
3715 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
3716 		    PFR_FLAG_USERIOCTL);
3717 		PF_RULES_RUNLOCK();
3718 		if (error == 0)
3719 			error = copyout(pfras, io->pfrio_buffer, totlen);
3720 		free(pfras, M_TEMP);
3721 		break;
3722 	}
3723 
3724 	case DIOCRINADEFINE: {
3725 		struct pfioc_table *io = (struct pfioc_table *)addr;
3726 		struct pfr_addr *pfras;
3727 		size_t totlen;
3728 
3729 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3730 			error = ENODEV;
3731 			break;
3732 		}
3733 		if (io->pfrio_size < 0 ||
3734 		    io->pfrio_size > pf_ioctl_maxcount ||
3735 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3736 			error = EINVAL;
3737 			break;
3738 		}
3739 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3740 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3741 		    M_TEMP, M_NOWAIT);
3742 		if (! pfras) {
3743 			error = ENOMEM;
3744 			break;
3745 		}
3746 		error = copyin(io->pfrio_buffer, pfras, totlen);
3747 		if (error) {
3748 			free(pfras, M_TEMP);
3749 			break;
3750 		}
3751 		PF_RULES_WLOCK();
3752 		error = pfr_ina_define(&io->pfrio_table, pfras,
3753 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
3754 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3755 		PF_RULES_WUNLOCK();
3756 		free(pfras, M_TEMP);
3757 		break;
3758 	}
3759 
3760 	case DIOCOSFPADD: {
3761 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
3762 		PF_RULES_WLOCK();
3763 		error = pf_osfp_add(io);
3764 		PF_RULES_WUNLOCK();
3765 		break;
3766 	}
3767 
3768 	case DIOCOSFPGET: {
3769 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
3770 		PF_RULES_RLOCK();
3771 		error = pf_osfp_get(io);
3772 		PF_RULES_RUNLOCK();
3773 		break;
3774 	}
3775 
3776 	case DIOCXBEGIN: {
3777 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3778 		struct pfioc_trans_e	*ioes, *ioe;
3779 		size_t			 totlen;
3780 		int			 i;
3781 
3782 		if (io->esize != sizeof(*ioe)) {
3783 			error = ENODEV;
3784 			break;
3785 		}
3786 		if (io->size < 0 ||
3787 		    io->size > pf_ioctl_maxcount ||
3788 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3789 			error = EINVAL;
3790 			break;
3791 		}
3792 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3793 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3794 		    M_TEMP, M_NOWAIT);
3795 		if (! ioes) {
3796 			error = ENOMEM;
3797 			break;
3798 		}
3799 		error = copyin(io->array, ioes, totlen);
3800 		if (error) {
3801 			free(ioes, M_TEMP);
3802 			break;
3803 		}
3804 		PF_RULES_WLOCK();
3805 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3806 			switch (ioe->rs_num) {
3807 #ifdef ALTQ
3808 			case PF_RULESET_ALTQ:
3809 				if (ioe->anchor[0]) {
3810 					PF_RULES_WUNLOCK();
3811 					free(ioes, M_TEMP);
3812 					error = EINVAL;
3813 					goto fail;
3814 				}
3815 				if ((error = pf_begin_altq(&ioe->ticket))) {
3816 					PF_RULES_WUNLOCK();
3817 					free(ioes, M_TEMP);
3818 					goto fail;
3819 				}
3820 				break;
3821 #endif /* ALTQ */
3822 			case PF_RULESET_TABLE:
3823 			    {
3824 				struct pfr_table table;
3825 
3826 				bzero(&table, sizeof(table));
3827 				strlcpy(table.pfrt_anchor, ioe->anchor,
3828 				    sizeof(table.pfrt_anchor));
3829 				if ((error = pfr_ina_begin(&table,
3830 				    &ioe->ticket, NULL, 0))) {
3831 					PF_RULES_WUNLOCK();
3832 					free(ioes, M_TEMP);
3833 					goto fail;
3834 				}
3835 				break;
3836 			    }
3837 			default:
3838 				if ((error = pf_begin_rules(&ioe->ticket,
3839 				    ioe->rs_num, ioe->anchor))) {
3840 					PF_RULES_WUNLOCK();
3841 					free(ioes, M_TEMP);
3842 					goto fail;
3843 				}
3844 				break;
3845 			}
3846 		}
3847 		PF_RULES_WUNLOCK();
3848 		error = copyout(ioes, io->array, totlen);
3849 		free(ioes, M_TEMP);
3850 		break;
3851 	}
3852 
3853 	case DIOCXROLLBACK: {
3854 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3855 		struct pfioc_trans_e	*ioe, *ioes;
3856 		size_t			 totlen;
3857 		int			 i;
3858 
3859 		if (io->esize != sizeof(*ioe)) {
3860 			error = ENODEV;
3861 			break;
3862 		}
3863 		if (io->size < 0 ||
3864 		    io->size > pf_ioctl_maxcount ||
3865 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3866 			error = EINVAL;
3867 			break;
3868 		}
3869 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3870 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3871 		    M_TEMP, M_NOWAIT);
3872 		if (! ioes) {
3873 			error = ENOMEM;
3874 			break;
3875 		}
3876 		error = copyin(io->array, ioes, totlen);
3877 		if (error) {
3878 			free(ioes, M_TEMP);
3879 			break;
3880 		}
3881 		PF_RULES_WLOCK();
3882 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3883 			switch (ioe->rs_num) {
3884 #ifdef ALTQ
3885 			case PF_RULESET_ALTQ:
3886 				if (ioe->anchor[0]) {
3887 					PF_RULES_WUNLOCK();
3888 					free(ioes, M_TEMP);
3889 					error = EINVAL;
3890 					goto fail;
3891 				}
3892 				if ((error = pf_rollback_altq(ioe->ticket))) {
3893 					PF_RULES_WUNLOCK();
3894 					free(ioes, M_TEMP);
3895 					goto fail; /* really bad */
3896 				}
3897 				break;
3898 #endif /* ALTQ */
3899 			case PF_RULESET_TABLE:
3900 			    {
3901 				struct pfr_table table;
3902 
3903 				bzero(&table, sizeof(table));
3904 				strlcpy(table.pfrt_anchor, ioe->anchor,
3905 				    sizeof(table.pfrt_anchor));
3906 				if ((error = pfr_ina_rollback(&table,
3907 				    ioe->ticket, NULL, 0))) {
3908 					PF_RULES_WUNLOCK();
3909 					free(ioes, M_TEMP);
3910 					goto fail; /* really bad */
3911 				}
3912 				break;
3913 			    }
3914 			default:
3915 				if ((error = pf_rollback_rules(ioe->ticket,
3916 				    ioe->rs_num, ioe->anchor))) {
3917 					PF_RULES_WUNLOCK();
3918 					free(ioes, M_TEMP);
3919 					goto fail; /* really bad */
3920 				}
3921 				break;
3922 			}
3923 		}
3924 		PF_RULES_WUNLOCK();
3925 		free(ioes, M_TEMP);
3926 		break;
3927 	}
3928 
3929 	case DIOCXCOMMIT: {
3930 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3931 		struct pfioc_trans_e	*ioe, *ioes;
3932 		struct pf_kruleset	*rs;
3933 		size_t			 totlen;
3934 		int			 i;
3935 
3936 		if (io->esize != sizeof(*ioe)) {
3937 			error = ENODEV;
3938 			break;
3939 		}
3940 
3941 		if (io->size < 0 ||
3942 		    io->size > pf_ioctl_maxcount ||
3943 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3944 			error = EINVAL;
3945 			break;
3946 		}
3947 
3948 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3949 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3950 		    M_TEMP, M_NOWAIT);
3951 		if (ioes == NULL) {
3952 			error = ENOMEM;
3953 			break;
3954 		}
3955 		error = copyin(io->array, ioes, totlen);
3956 		if (error) {
3957 			free(ioes, M_TEMP);
3958 			break;
3959 		}
3960 		PF_RULES_WLOCK();
3961 		/* First makes sure everything will succeed. */
3962 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3963 			switch (ioe->rs_num) {
3964 #ifdef ALTQ
3965 			case PF_RULESET_ALTQ:
3966 				if (ioe->anchor[0]) {
3967 					PF_RULES_WUNLOCK();
3968 					free(ioes, M_TEMP);
3969 					error = EINVAL;
3970 					goto fail;
3971 				}
3972 				if (!V_altqs_inactive_open || ioe->ticket !=
3973 				    V_ticket_altqs_inactive) {
3974 					PF_RULES_WUNLOCK();
3975 					free(ioes, M_TEMP);
3976 					error = EBUSY;
3977 					goto fail;
3978 				}
3979 				break;
3980 #endif /* ALTQ */
3981 			case PF_RULESET_TABLE:
3982 				rs = pf_find_kruleset(ioe->anchor);
3983 				if (rs == NULL || !rs->topen || ioe->ticket !=
3984 				    rs->tticket) {
3985 					PF_RULES_WUNLOCK();
3986 					free(ioes, M_TEMP);
3987 					error = EBUSY;
3988 					goto fail;
3989 				}
3990 				break;
3991 			default:
3992 				if (ioe->rs_num < 0 || ioe->rs_num >=
3993 				    PF_RULESET_MAX) {
3994 					PF_RULES_WUNLOCK();
3995 					free(ioes, M_TEMP);
3996 					error = EINVAL;
3997 					goto fail;
3998 				}
3999 				rs = pf_find_kruleset(ioe->anchor);
4000 				if (rs == NULL ||
4001 				    !rs->rules[ioe->rs_num].inactive.open ||
4002 				    rs->rules[ioe->rs_num].inactive.ticket !=
4003 				    ioe->ticket) {
4004 					PF_RULES_WUNLOCK();
4005 					free(ioes, M_TEMP);
4006 					error = EBUSY;
4007 					goto fail;
4008 				}
4009 				break;
4010 			}
4011 		}
4012 		/* Now do the commit - no errors should happen here. */
4013 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
4014 			switch (ioe->rs_num) {
4015 #ifdef ALTQ
4016 			case PF_RULESET_ALTQ:
4017 				if ((error = pf_commit_altq(ioe->ticket))) {
4018 					PF_RULES_WUNLOCK();
4019 					free(ioes, M_TEMP);
4020 					goto fail; /* really bad */
4021 				}
4022 				break;
4023 #endif /* ALTQ */
4024 			case PF_RULESET_TABLE:
4025 			    {
4026 				struct pfr_table table;
4027 
4028 				bzero(&table, sizeof(table));
4029 				strlcpy(table.pfrt_anchor, ioe->anchor,
4030 				    sizeof(table.pfrt_anchor));
4031 				if ((error = pfr_ina_commit(&table,
4032 				    ioe->ticket, NULL, NULL, 0))) {
4033 					PF_RULES_WUNLOCK();
4034 					free(ioes, M_TEMP);
4035 					goto fail; /* really bad */
4036 				}
4037 				break;
4038 			    }
4039 			default:
4040 				if ((error = pf_commit_rules(ioe->ticket,
4041 				    ioe->rs_num, ioe->anchor))) {
4042 					PF_RULES_WUNLOCK();
4043 					free(ioes, M_TEMP);
4044 					goto fail; /* really bad */
4045 				}
4046 				break;
4047 			}
4048 		}
4049 		PF_RULES_WUNLOCK();
4050 		free(ioes, M_TEMP);
4051 		break;
4052 	}
4053 
4054 	case DIOCGETSRCNODES: {
4055 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
4056 		struct pf_srchash	*sh;
4057 		struct pf_ksrc_node	*n;
4058 		struct pf_src_node	*p, *pstore;
4059 		uint32_t		 i, nr = 0;
4060 
4061 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4062 				i++, sh++) {
4063 			PF_HASHROW_LOCK(sh);
4064 			LIST_FOREACH(n, &sh->nodes, entry)
4065 				nr++;
4066 			PF_HASHROW_UNLOCK(sh);
4067 		}
4068 
4069 		psn->psn_len = min(psn->psn_len,
4070 		    sizeof(struct pf_src_node) * nr);
4071 
4072 		if (psn->psn_len == 0) {
4073 			psn->psn_len = sizeof(struct pf_src_node) * nr;
4074 			break;
4075 		}
4076 
4077 		nr = 0;
4078 
4079 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
4080 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4081 		    i++, sh++) {
4082 		    PF_HASHROW_LOCK(sh);
4083 		    LIST_FOREACH(n, &sh->nodes, entry) {
4084 
4085 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
4086 				break;
4087 
4088 			pf_src_node_copy(n, p);
4089 
4090 			p++;
4091 			nr++;
4092 		    }
4093 		    PF_HASHROW_UNLOCK(sh);
4094 		}
4095 		error = copyout(pstore, psn->psn_src_nodes,
4096 		    sizeof(struct pf_src_node) * nr);
4097 		if (error) {
4098 			free(pstore, M_TEMP);
4099 			break;
4100 		}
4101 		psn->psn_len = sizeof(struct pf_src_node) * nr;
4102 		free(pstore, M_TEMP);
4103 		break;
4104 	}
4105 
4106 	case DIOCCLRSRCNODES: {
4107 		pf_clear_srcnodes(NULL);
4108 		pf_purge_expired_src_nodes();
4109 		break;
4110 	}
4111 
4112 	case DIOCKILLSRCNODES:
4113 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
4114 		break;
4115 
4116 	case DIOCSETHOSTID: {
4117 		u_int32_t	*hostid = (u_int32_t *)addr;
4118 
4119 		PF_RULES_WLOCK();
4120 		if (*hostid == 0)
4121 			V_pf_status.hostid = arc4random();
4122 		else
4123 			V_pf_status.hostid = *hostid;
4124 		PF_RULES_WUNLOCK();
4125 		break;
4126 	}
4127 
4128 	case DIOCOSFPFLUSH:
4129 		PF_RULES_WLOCK();
4130 		pf_osfp_flush();
4131 		PF_RULES_WUNLOCK();
4132 		break;
4133 
4134 	case DIOCIGETIFACES: {
4135 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
4136 		struct pfi_kif *ifstore;
4137 		size_t bufsiz;
4138 
4139 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
4140 			error = ENODEV;
4141 			break;
4142 		}
4143 
4144 		if (io->pfiio_size < 0 ||
4145 		    io->pfiio_size > pf_ioctl_maxcount ||
4146 		    WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
4147 			error = EINVAL;
4148 			break;
4149 		}
4150 
4151 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
4152 		ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
4153 		    M_TEMP, M_NOWAIT);
4154 		if (ifstore == NULL) {
4155 			error = ENOMEM;
4156 			break;
4157 		}
4158 
4159 		PF_RULES_RLOCK();
4160 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
4161 		PF_RULES_RUNLOCK();
4162 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
4163 		free(ifstore, M_TEMP);
4164 		break;
4165 	}
4166 
4167 	case DIOCSETIFFLAG: {
4168 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
4169 
4170 		PF_RULES_WLOCK();
4171 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
4172 		PF_RULES_WUNLOCK();
4173 		break;
4174 	}
4175 
4176 	case DIOCCLRIFFLAG: {
4177 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
4178 
4179 		PF_RULES_WLOCK();
4180 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
4181 		PF_RULES_WUNLOCK();
4182 		break;
4183 	}
4184 
4185 	default:
4186 		error = ENODEV;
4187 		break;
4188 	}
4189 fail:
4190 	if (sx_xlocked(&pf_ioctl_lock))
4191 		sx_xunlock(&pf_ioctl_lock);
4192 	CURVNET_RESTORE();
4193 
4194 	return (error);
4195 }
4196 
4197 void
4198 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
4199 {
4200 	bzero(sp, sizeof(struct pfsync_state));
4201 
4202 	/* copy from state key */
4203 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
4204 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
4205 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
4206 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
4207 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
4208 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
4209 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
4210 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
4211 	sp->proto = st->key[PF_SK_WIRE]->proto;
4212 	sp->af = st->key[PF_SK_WIRE]->af;
4213 
4214 	/* copy from state */
4215 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
4216 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
4217 	sp->creation = htonl(time_uptime - st->creation);
4218 	sp->expire = pf_state_expires(st);
4219 	if (sp->expire <= time_uptime)
4220 		sp->expire = htonl(0);
4221 	else
4222 		sp->expire = htonl(sp->expire - time_uptime);
4223 
4224 	sp->direction = st->direction;
4225 	sp->log = st->log;
4226 	sp->timeout = st->timeout;
4227 	sp->state_flags = st->state_flags;
4228 	if (st->src_node)
4229 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
4230 	if (st->nat_src_node)
4231 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
4232 
4233 	sp->id = st->id;
4234 	sp->creatorid = st->creatorid;
4235 	pf_state_peer_hton(&st->src, &sp->src);
4236 	pf_state_peer_hton(&st->dst, &sp->dst);
4237 
4238 	if (st->rule.ptr == NULL)
4239 		sp->rule = htonl(-1);
4240 	else
4241 		sp->rule = htonl(st->rule.ptr->nr);
4242 	if (st->anchor.ptr == NULL)
4243 		sp->anchor = htonl(-1);
4244 	else
4245 		sp->anchor = htonl(st->anchor.ptr->nr);
4246 	if (st->nat_rule.ptr == NULL)
4247 		sp->nat_rule = htonl(-1);
4248 	else
4249 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
4250 
4251 	pf_state_counter_hton(counter_u64_fetch(st->packets[0]),
4252 	    sp->packets[0]);
4253 	pf_state_counter_hton(counter_u64_fetch(st->packets[1]),
4254 	    sp->packets[1]);
4255 	pf_state_counter_hton(counter_u64_fetch(st->bytes[0]), sp->bytes[0]);
4256 	pf_state_counter_hton(counter_u64_fetch(st->bytes[1]), sp->bytes[1]);
4257 
4258 }
4259 
4260 static void
4261 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
4262 {
4263 	struct pfr_ktable *kt;
4264 
4265 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
4266 
4267 	kt = aw->p.tbl;
4268 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
4269 		kt = kt->pfrkt_root;
4270 	aw->p.tbl = NULL;
4271 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
4272 		kt->pfrkt_cnt : -1;
4273 }
4274 
4275 /*
4276  * XXX - Check for version missmatch!!!
4277  */
4278 static void
4279 pf_clear_states(void)
4280 {
4281 	struct pf_state	*s;
4282 	u_int i;
4283 
4284 	for (i = 0; i <= pf_hashmask; i++) {
4285 		struct pf_idhash *ih = &V_pf_idhash[i];
4286 relock:
4287 		PF_HASHROW_LOCK(ih);
4288 		LIST_FOREACH(s, &ih->states, entry) {
4289 			s->timeout = PFTM_PURGE;
4290 			/* Don't send out individual delete messages. */
4291 			s->state_flags |= PFSTATE_NOSYNC;
4292 			pf_unlink_state(s, PF_ENTER_LOCKED);
4293 			goto relock;
4294 		}
4295 		PF_HASHROW_UNLOCK(ih);
4296 	}
4297 }
4298 
4299 static int
4300 pf_clear_tables(void)
4301 {
4302 	struct pfioc_table io;
4303 	int error;
4304 
4305 	bzero(&io, sizeof(io));
4306 
4307 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
4308 	    io.pfrio_flags);
4309 
4310 	return (error);
4311 }
4312 
4313 static void
4314 pf_clear_srcnodes(struct pf_ksrc_node *n)
4315 {
4316 	struct pf_state *s;
4317 	int i;
4318 
4319 	for (i = 0; i <= pf_hashmask; i++) {
4320 		struct pf_idhash *ih = &V_pf_idhash[i];
4321 
4322 		PF_HASHROW_LOCK(ih);
4323 		LIST_FOREACH(s, &ih->states, entry) {
4324 			if (n == NULL || n == s->src_node)
4325 				s->src_node = NULL;
4326 			if (n == NULL || n == s->nat_src_node)
4327 				s->nat_src_node = NULL;
4328 		}
4329 		PF_HASHROW_UNLOCK(ih);
4330 	}
4331 
4332 	if (n == NULL) {
4333 		struct pf_srchash *sh;
4334 
4335 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4336 		    i++, sh++) {
4337 			PF_HASHROW_LOCK(sh);
4338 			LIST_FOREACH(n, &sh->nodes, entry) {
4339 				n->expire = 1;
4340 				n->states = 0;
4341 			}
4342 			PF_HASHROW_UNLOCK(sh);
4343 		}
4344 	} else {
4345 		/* XXX: hash slot should already be locked here. */
4346 		n->expire = 1;
4347 		n->states = 0;
4348 	}
4349 }
4350 
4351 static void
4352 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
4353 {
4354 	struct pf_ksrc_node_list	 kill;
4355 
4356 	LIST_INIT(&kill);
4357 	for (int i = 0; i <= pf_srchashmask; i++) {
4358 		struct pf_srchash *sh = &V_pf_srchash[i];
4359 		struct pf_ksrc_node *sn, *tmp;
4360 
4361 		PF_HASHROW_LOCK(sh);
4362 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
4363 			if (PF_MATCHA(psnk->psnk_src.neg,
4364 			      &psnk->psnk_src.addr.v.a.addr,
4365 			      &psnk->psnk_src.addr.v.a.mask,
4366 			      &sn->addr, sn->af) &&
4367 			    PF_MATCHA(psnk->psnk_dst.neg,
4368 			      &psnk->psnk_dst.addr.v.a.addr,
4369 			      &psnk->psnk_dst.addr.v.a.mask,
4370 			      &sn->raddr, sn->af)) {
4371 				pf_unlink_src_node(sn);
4372 				LIST_INSERT_HEAD(&kill, sn, entry);
4373 				sn->expire = 1;
4374 			}
4375 		PF_HASHROW_UNLOCK(sh);
4376 	}
4377 
4378 	for (int i = 0; i <= pf_hashmask; i++) {
4379 		struct pf_idhash *ih = &V_pf_idhash[i];
4380 		struct pf_state *s;
4381 
4382 		PF_HASHROW_LOCK(ih);
4383 		LIST_FOREACH(s, &ih->states, entry) {
4384 			if (s->src_node && s->src_node->expire == 1)
4385 				s->src_node = NULL;
4386 			if (s->nat_src_node && s->nat_src_node->expire == 1)
4387 				s->nat_src_node = NULL;
4388 		}
4389 		PF_HASHROW_UNLOCK(ih);
4390 	}
4391 
4392 	psnk->psnk_killed = pf_free_src_nodes(&kill);
4393 }
4394 
4395 /*
4396  * XXX - Check for version missmatch!!!
4397  */
4398 
4399 /*
4400  * Duplicate pfctl -Fa operation to get rid of as much as we can.
4401  */
4402 static int
4403 shutdown_pf(void)
4404 {
4405 	int error = 0;
4406 	u_int32_t t[5];
4407 	char nn = '\0';
4408 
4409 	do {
4410 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
4411 		    != 0) {
4412 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
4413 			break;
4414 		}
4415 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
4416 		    != 0) {
4417 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
4418 			break;		/* XXX: rollback? */
4419 		}
4420 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
4421 		    != 0) {
4422 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
4423 			break;		/* XXX: rollback? */
4424 		}
4425 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
4426 		    != 0) {
4427 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
4428 			break;		/* XXX: rollback? */
4429 		}
4430 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
4431 		    != 0) {
4432 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
4433 			break;		/* XXX: rollback? */
4434 		}
4435 
4436 		/* XXX: these should always succeed here */
4437 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
4438 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
4439 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
4440 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
4441 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
4442 
4443 		if ((error = pf_clear_tables()) != 0)
4444 			break;
4445 
4446 #ifdef ALTQ
4447 		if ((error = pf_begin_altq(&t[0])) != 0) {
4448 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
4449 			break;
4450 		}
4451 		pf_commit_altq(t[0]);
4452 #endif
4453 
4454 		pf_clear_states();
4455 
4456 		pf_clear_srcnodes(NULL);
4457 
4458 		/* status does not use malloced mem so no need to cleanup */
4459 		/* fingerprints and interfaces have their own cleanup code */
4460 	} while(0);
4461 
4462 	return (error);
4463 }
4464 
4465 static pfil_return_t
4466 pf_check_return(int chk, struct mbuf **m)
4467 {
4468 
4469 	switch (chk) {
4470 	case PF_PASS:
4471 		if (*m == NULL)
4472 			return (PFIL_CONSUMED);
4473 		else
4474 			return (PFIL_PASS);
4475 		break;
4476 	default:
4477 		if (*m != NULL) {
4478 			m_freem(*m);
4479 			*m = NULL;
4480 		}
4481 		return (PFIL_DROPPED);
4482 	}
4483 }
4484 
4485 #ifdef INET
4486 static pfil_return_t
4487 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
4488     void *ruleset __unused, struct inpcb *inp)
4489 {
4490 	int chk;
4491 
4492 	chk = pf_test(PF_IN, flags, ifp, m, inp);
4493 
4494 	return (pf_check_return(chk, m));
4495 }
4496 
4497 static pfil_return_t
4498 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
4499     void *ruleset __unused,  struct inpcb *inp)
4500 {
4501 	int chk;
4502 
4503 	chk = pf_test(PF_OUT, flags, ifp, m, inp);
4504 
4505 	return (pf_check_return(chk, m));
4506 }
4507 #endif
4508 
4509 #ifdef INET6
4510 static pfil_return_t
4511 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
4512     void *ruleset __unused,  struct inpcb *inp)
4513 {
4514 	int chk;
4515 
4516 	/*
4517 	 * In case of loopback traffic IPv6 uses the real interface in
4518 	 * order to support scoped addresses. In order to support stateful
4519 	 * filtering we have change this to lo0 as it is the case in IPv4.
4520 	 */
4521 	CURVNET_SET(ifp->if_vnet);
4522 	chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
4523 	CURVNET_RESTORE();
4524 
4525 	return (pf_check_return(chk, m));
4526 }
4527 
4528 static pfil_return_t
4529 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
4530     void *ruleset __unused,  struct inpcb *inp)
4531 {
4532 	int chk;
4533 
4534 	CURVNET_SET(ifp->if_vnet);
4535 	chk = pf_test6(PF_OUT, flags, ifp, m, inp);
4536 	CURVNET_RESTORE();
4537 
4538 	return (pf_check_return(chk, m));
4539 }
4540 #endif /* INET6 */
4541 
4542 #ifdef INET
4543 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
4544 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
4545 #define	V_pf_ip4_in_hook	VNET(pf_ip4_in_hook)
4546 #define	V_pf_ip4_out_hook	VNET(pf_ip4_out_hook)
4547 #endif
4548 #ifdef INET6
4549 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
4550 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
4551 #define	V_pf_ip6_in_hook	VNET(pf_ip6_in_hook)
4552 #define	V_pf_ip6_out_hook	VNET(pf_ip6_out_hook)
4553 #endif
4554 
4555 static int
4556 hook_pf(void)
4557 {
4558 	struct pfil_hook_args pha;
4559 	struct pfil_link_args pla;
4560 
4561 	if (V_pf_pfil_hooked)
4562 		return (0);
4563 
4564 	pha.pa_version = PFIL_VERSION;
4565 	pha.pa_modname = "pf";
4566 	pha.pa_ruleset = NULL;
4567 
4568 	pla.pa_version = PFIL_VERSION;
4569 
4570 #ifdef INET
4571 	pha.pa_type = PFIL_TYPE_IP4;
4572 	pha.pa_func = pf_check_in;
4573 	pha.pa_flags = PFIL_IN;
4574 	pha.pa_rulname = "default-in";
4575 	V_pf_ip4_in_hook = pfil_add_hook(&pha);
4576 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
4577 	pla.pa_head = V_inet_pfil_head;
4578 	pla.pa_hook = V_pf_ip4_in_hook;
4579 	(void)pfil_link(&pla);
4580 	pha.pa_func = pf_check_out;
4581 	pha.pa_flags = PFIL_OUT;
4582 	pha.pa_rulname = "default-out";
4583 	V_pf_ip4_out_hook = pfil_add_hook(&pha);
4584 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
4585 	pla.pa_head = V_inet_pfil_head;
4586 	pla.pa_hook = V_pf_ip4_out_hook;
4587 	(void)pfil_link(&pla);
4588 #endif
4589 #ifdef INET6
4590 	pha.pa_type = PFIL_TYPE_IP6;
4591 	pha.pa_func = pf_check6_in;
4592 	pha.pa_flags = PFIL_IN;
4593 	pha.pa_rulname = "default-in6";
4594 	V_pf_ip6_in_hook = pfil_add_hook(&pha);
4595 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
4596 	pla.pa_head = V_inet6_pfil_head;
4597 	pla.pa_hook = V_pf_ip6_in_hook;
4598 	(void)pfil_link(&pla);
4599 	pha.pa_func = pf_check6_out;
4600 	pha.pa_rulname = "default-out6";
4601 	pha.pa_flags = PFIL_OUT;
4602 	V_pf_ip6_out_hook = pfil_add_hook(&pha);
4603 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
4604 	pla.pa_head = V_inet6_pfil_head;
4605 	pla.pa_hook = V_pf_ip6_out_hook;
4606 	(void)pfil_link(&pla);
4607 #endif
4608 
4609 	V_pf_pfil_hooked = 1;
4610 	return (0);
4611 }
4612 
4613 static int
4614 dehook_pf(void)
4615 {
4616 
4617 	if (V_pf_pfil_hooked == 0)
4618 		return (0);
4619 
4620 #ifdef INET
4621 	pfil_remove_hook(V_pf_ip4_in_hook);
4622 	pfil_remove_hook(V_pf_ip4_out_hook);
4623 #endif
4624 #ifdef INET6
4625 	pfil_remove_hook(V_pf_ip6_in_hook);
4626 	pfil_remove_hook(V_pf_ip6_out_hook);
4627 #endif
4628 
4629 	V_pf_pfil_hooked = 0;
4630 	return (0);
4631 }
4632 
4633 static void
4634 pf_load_vnet(void)
4635 {
4636 	V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
4637 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
4638 
4639 	pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
4640 	    PF_RULE_TAG_HASH_SIZE_DEFAULT);
4641 #ifdef ALTQ
4642 	pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
4643 	    PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
4644 #endif
4645 
4646 	pfattach_vnet();
4647 	V_pf_vnet_active = 1;
4648 }
4649 
4650 static int
4651 pf_load(void)
4652 {
4653 	int error;
4654 
4655 	rm_init(&pf_rules_lock, "pf rulesets");
4656 	sx_init(&pf_ioctl_lock, "pf ioctl");
4657 	sx_init(&pf_end_lock, "pf end thread");
4658 
4659 	pf_mtag_initialize();
4660 
4661 	pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
4662 	if (pf_dev == NULL)
4663 		return (ENOMEM);
4664 
4665 	pf_end_threads = 0;
4666 	error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
4667 	if (error != 0)
4668 		return (error);
4669 
4670 	pfi_initialize();
4671 
4672 	return (0);
4673 }
4674 
4675 static void
4676 pf_unload_vnet(void)
4677 {
4678 	int error;
4679 
4680 	V_pf_vnet_active = 0;
4681 	V_pf_status.running = 0;
4682 	error = dehook_pf();
4683 	if (error) {
4684 		/*
4685 		 * Should not happen!
4686 		 * XXX Due to error code ESRCH, kldunload will show
4687 		 * a message like 'No such process'.
4688 		 */
4689 		printf("%s : pfil unregisteration fail\n", __FUNCTION__);
4690 		return;
4691 	}
4692 
4693 	PF_RULES_WLOCK();
4694 	shutdown_pf();
4695 	PF_RULES_WUNLOCK();
4696 
4697 	swi_remove(V_pf_swi_cookie);
4698 
4699 	pf_unload_vnet_purge();
4700 
4701 	pf_normalize_cleanup();
4702 	PF_RULES_WLOCK();
4703 	pfi_cleanup_vnet();
4704 	PF_RULES_WUNLOCK();
4705 	pfr_cleanup();
4706 	pf_osfp_flush();
4707 	pf_cleanup();
4708 	if (IS_DEFAULT_VNET(curvnet))
4709 		pf_mtag_cleanup();
4710 
4711 	pf_cleanup_tagset(&V_pf_tags);
4712 #ifdef ALTQ
4713 	pf_cleanup_tagset(&V_pf_qids);
4714 #endif
4715 	uma_zdestroy(V_pf_tag_z);
4716 
4717 	/* Free counters last as we updated them during shutdown. */
4718 	counter_u64_free(V_pf_default_rule.evaluations);
4719 	for (int i = 0; i < 2; i++) {
4720 		counter_u64_free(V_pf_default_rule.packets[i]);
4721 		counter_u64_free(V_pf_default_rule.bytes[i]);
4722 	}
4723 	counter_u64_free(V_pf_default_rule.states_cur);
4724 	counter_u64_free(V_pf_default_rule.states_tot);
4725 	counter_u64_free(V_pf_default_rule.src_nodes);
4726 
4727 	for (int i = 0; i < PFRES_MAX; i++)
4728 		counter_u64_free(V_pf_status.counters[i]);
4729 	for (int i = 0; i < LCNT_MAX; i++)
4730 		counter_u64_free(V_pf_status.lcounters[i]);
4731 	for (int i = 0; i < FCNT_MAX; i++)
4732 		counter_u64_free(V_pf_status.fcounters[i]);
4733 	for (int i = 0; i < SCNT_MAX; i++)
4734 		counter_u64_free(V_pf_status.scounters[i]);
4735 }
4736 
4737 static void
4738 pf_unload(void)
4739 {
4740 
4741 	sx_xlock(&pf_end_lock);
4742 	pf_end_threads = 1;
4743 	while (pf_end_threads < 2) {
4744 		wakeup_one(pf_purge_thread);
4745 		sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
4746 	}
4747 	sx_xunlock(&pf_end_lock);
4748 
4749 	if (pf_dev != NULL)
4750 		destroy_dev(pf_dev);
4751 
4752 	pfi_cleanup();
4753 
4754 	rm_destroy(&pf_rules_lock);
4755 	sx_destroy(&pf_ioctl_lock);
4756 	sx_destroy(&pf_end_lock);
4757 }
4758 
4759 static void
4760 vnet_pf_init(void *unused __unused)
4761 {
4762 
4763 	pf_load_vnet();
4764 }
4765 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
4766     vnet_pf_init, NULL);
4767 
4768 static void
4769 vnet_pf_uninit(const void *unused __unused)
4770 {
4771 
4772 	pf_unload_vnet();
4773 }
4774 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
4775 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
4776     vnet_pf_uninit, NULL);
4777 
4778 static int
4779 pf_modevent(module_t mod, int type, void *data)
4780 {
4781 	int error = 0;
4782 
4783 	switch(type) {
4784 	case MOD_LOAD:
4785 		error = pf_load();
4786 		break;
4787 	case MOD_UNLOAD:
4788 		/* Handled in SYSUNINIT(pf_unload) to ensure it's done after
4789 		 * the vnet_pf_uninit()s */
4790 		break;
4791 	default:
4792 		error = EINVAL;
4793 		break;
4794 	}
4795 
4796 	return (error);
4797 }
4798 
4799 static moduledata_t pf_mod = {
4800 	"pf",
4801 	pf_modevent,
4802 	0
4803 };
4804 
4805 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
4806 MODULE_VERSION(pf, PF_MODVER);
4807