xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision ec4deee4e4f2aef1b97d9424f25d04e91fd7dc10)
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_pool	*pf_get_pool(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_pool(struct pf_palist *, struct pf_palist *);
98 static void		 pf_empty_pool(struct pf_palist *);
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_ruleset *);
113 static void		 pf_hash_rule(MD5_CTX *, struct pf_rule *);
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_ruleset *,
117 			    struct pf_addr_wrap *, sa_family_t);
118 static void		 pf_addr_copyout(struct pf_addr_wrap *);
119 #ifdef ALTQ
120 static int		 pf_export_kaltq(struct pf_altq *,
121 			    struct pfioc_altq_v1 *, size_t);
122 static int		 pf_import_kaltq(struct pfioc_altq_v1 *,
123 			    struct pf_altq *, size_t);
124 #endif /* ALTQ */
125 
126 VNET_DEFINE(struct pf_rule,	pf_default_rule);
127 
128 #ifdef ALTQ
129 VNET_DEFINE_STATIC(int,		pf_altq_running);
130 #define	V_pf_altq_running	VNET(pf_altq_running)
131 #endif
132 
133 #define	TAGID_MAX	 50000
134 struct pf_tagname {
135 	TAILQ_ENTRY(pf_tagname)	namehash_entries;
136 	TAILQ_ENTRY(pf_tagname)	taghash_entries;
137 	char			name[PF_TAG_NAME_SIZE];
138 	uint16_t		tag;
139 	int			ref;
140 };
141 
142 struct pf_tagset {
143 	TAILQ_HEAD(, pf_tagname)	*namehash;
144 	TAILQ_HEAD(, pf_tagname)	*taghash;
145 	unsigned int			 mask;
146 	uint32_t			 seed;
147 	BITSET_DEFINE(, TAGID_MAX)	 avail;
148 };
149 
150 VNET_DEFINE(struct pf_tagset, pf_tags);
151 #define	V_pf_tags	VNET(pf_tags)
152 static unsigned int	pf_rule_tag_hashsize;
153 #define	PF_RULE_TAG_HASH_SIZE_DEFAULT	128
154 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN,
155     &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT,
156     "Size of pf(4) rule tag hashtable");
157 
158 #ifdef ALTQ
159 VNET_DEFINE(struct pf_tagset, pf_qids);
160 #define	V_pf_qids	VNET(pf_qids)
161 static unsigned int	pf_queue_tag_hashsize;
162 #define	PF_QUEUE_TAG_HASH_SIZE_DEFAULT	128
163 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN,
164     &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT,
165     "Size of pf(4) queue tag hashtable");
166 #endif
167 VNET_DEFINE(uma_zone_t,	 pf_tag_z);
168 #define	V_pf_tag_z		 VNET(pf_tag_z)
169 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
170 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
171 
172 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
173 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
174 #endif
175 
176 static void		 pf_init_tagset(struct pf_tagset *, unsigned int *,
177 			    unsigned int);
178 static void		 pf_cleanup_tagset(struct pf_tagset *);
179 static uint16_t		 tagname2hashindex(const struct pf_tagset *, const char *);
180 static uint16_t		 tag2hashindex(const struct pf_tagset *, uint16_t);
181 static u_int16_t	 tagname2tag(struct pf_tagset *, char *);
182 static u_int16_t	 pf_tagname2tag(char *);
183 static void		 tag_unref(struct pf_tagset *, u_int16_t);
184 
185 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
186 
187 struct cdev *pf_dev;
188 
189 /*
190  * XXX - These are new and need to be checked when moveing to a new version
191  */
192 static void		 pf_clear_states(void);
193 static int		 pf_clear_tables(void);
194 static void		 pf_clear_srcnodes(struct pf_src_node *);
195 static void		 pf_kill_srcnodes(struct pfioc_src_node_kill *);
196 static void		 pf_tbladdr_copyout(struct pf_addr_wrap *);
197 
198 /*
199  * Wrapper functions for pfil(9) hooks
200  */
201 #ifdef INET
202 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp,
203     int flags, void *ruleset __unused, struct inpcb *inp);
204 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp,
205     int flags, void *ruleset __unused, struct inpcb *inp);
206 #endif
207 #ifdef INET6
208 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp,
209     int flags, void *ruleset __unused, struct inpcb *inp);
210 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp,
211     int flags, void *ruleset __unused, struct inpcb *inp);
212 #endif
213 
214 static int		hook_pf(void);
215 static int		dehook_pf(void);
216 static int		shutdown_pf(void);
217 static int		pf_load(void);
218 static void		pf_unload(void);
219 
220 static struct cdevsw pf_cdevsw = {
221 	.d_ioctl =	pfioctl,
222 	.d_name =	PF_NAME,
223 	.d_version =	D_VERSION,
224 };
225 
226 volatile VNET_DEFINE_STATIC(int, pf_pfil_hooked);
227 #define V_pf_pfil_hooked	VNET(pf_pfil_hooked)
228 
229 /*
230  * We need a flag that is neither hooked nor running to know when
231  * the VNET is "valid".  We primarily need this to control (global)
232  * external event, e.g., eventhandlers.
233  */
234 VNET_DEFINE(int, pf_vnet_active);
235 #define V_pf_vnet_active	VNET(pf_vnet_active)
236 
237 int pf_end_threads;
238 struct proc *pf_purge_proc;
239 
240 struct rmlock			pf_rules_lock;
241 struct sx			pf_ioctl_lock;
242 struct sx			pf_end_lock;
243 
244 /* pfsync */
245 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
246 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
247 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
248 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
249 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
250 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
251 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
252 
253 /* pflog */
254 pflog_packet_t			*pflog_packet_ptr = NULL;
255 
256 extern u_long	pf_ioctl_maxcount;
257 
258 static void
259 pfattach_vnet(void)
260 {
261 	u_int32_t *my_timeout = V_pf_default_rule.timeout;
262 
263 	pf_initialize();
264 	pfr_initialize();
265 	pfi_initialize_vnet();
266 	pf_normalize_init();
267 
268 	V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
269 	V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
270 
271 	RB_INIT(&V_pf_anchors);
272 	pf_init_ruleset(&pf_main_ruleset);
273 
274 	/* default rule should never be garbage collected */
275 	V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
276 #ifdef PF_DEFAULT_TO_DROP
277 	V_pf_default_rule.action = PF_DROP;
278 #else
279 	V_pf_default_rule.action = PF_PASS;
280 #endif
281 	V_pf_default_rule.nr = -1;
282 	V_pf_default_rule.rtableid = -1;
283 
284 	V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
285 	V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
286 	V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
287 
288 	/* initialize default timeouts */
289 	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
290 	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
291 	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
292 	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
293 	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
294 	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
295 	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
296 	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
297 	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
298 	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
299 	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
300 	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
301 	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
302 	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
303 	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
304 	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
305 	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
306 	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
307 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
308 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
309 
310 	bzero(&V_pf_status, sizeof(V_pf_status));
311 	V_pf_status.debug = PF_DEBUG_URGENT;
312 
313 	V_pf_pfil_hooked = 0;
314 
315 	/* XXX do our best to avoid a conflict */
316 	V_pf_status.hostid = arc4random();
317 
318 	for (int i = 0; i < PFRES_MAX; i++)
319 		V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
320 	for (int i = 0; i < LCNT_MAX; i++)
321 		V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
322 	for (int i = 0; i < FCNT_MAX; i++)
323 		V_pf_status.fcounters[i] = counter_u64_alloc(M_WAITOK);
324 	for (int i = 0; i < SCNT_MAX; i++)
325 		V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
326 
327 	if (swi_add(NULL, "pf send", pf_intr, curvnet, SWI_NET,
328 	    INTR_MPSAFE, &V_pf_swi_cookie) != 0)
329 		/* XXXGL: leaked all above. */
330 		return;
331 }
332 
333 
334 static struct pf_pool *
335 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
336     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
337     u_int8_t check_ticket)
338 {
339 	struct pf_ruleset	*ruleset;
340 	struct pf_rule		*rule;
341 	int			 rs_num;
342 
343 	ruleset = pf_find_ruleset(anchor);
344 	if (ruleset == NULL)
345 		return (NULL);
346 	rs_num = pf_get_ruleset_number(rule_action);
347 	if (rs_num >= PF_RULESET_MAX)
348 		return (NULL);
349 	if (active) {
350 		if (check_ticket && ticket !=
351 		    ruleset->rules[rs_num].active.ticket)
352 			return (NULL);
353 		if (r_last)
354 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
355 			    pf_rulequeue);
356 		else
357 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
358 	} else {
359 		if (check_ticket && ticket !=
360 		    ruleset->rules[rs_num].inactive.ticket)
361 			return (NULL);
362 		if (r_last)
363 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
364 			    pf_rulequeue);
365 		else
366 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
367 	}
368 	if (!r_last) {
369 		while ((rule != NULL) && (rule->nr != rule_number))
370 			rule = TAILQ_NEXT(rule, entries);
371 	}
372 	if (rule == NULL)
373 		return (NULL);
374 
375 	return (&rule->rpool);
376 }
377 
378 static void
379 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
380 {
381 	struct pf_pooladdr	*mv_pool_pa;
382 
383 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
384 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
385 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
386 	}
387 }
388 
389 static void
390 pf_empty_pool(struct pf_palist *poola)
391 {
392 	struct pf_pooladdr *pa;
393 
394 	while ((pa = TAILQ_FIRST(poola)) != NULL) {
395 		switch (pa->addr.type) {
396 		case PF_ADDR_DYNIFTL:
397 			pfi_dynaddr_remove(pa->addr.p.dyn);
398 			break;
399 		case PF_ADDR_TABLE:
400 			/* XXX: this could be unfinished pooladdr on pabuf */
401 			if (pa->addr.p.tbl != NULL)
402 				pfr_detach_table(pa->addr.p.tbl);
403 			break;
404 		}
405 		if (pa->kif)
406 			pfi_kif_unref(pa->kif);
407 		TAILQ_REMOVE(poola, pa, entries);
408 		free(pa, M_PFRULE);
409 	}
410 }
411 
412 static void
413 pf_unlink_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
414 {
415 
416 	PF_RULES_WASSERT();
417 
418 	TAILQ_REMOVE(rulequeue, rule, entries);
419 
420 	PF_UNLNKDRULES_LOCK();
421 	rule->rule_flag |= PFRULE_REFS;
422 	TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
423 	PF_UNLNKDRULES_UNLOCK();
424 }
425 
426 void
427 pf_free_rule(struct pf_rule *rule)
428 {
429 
430 	PF_RULES_WASSERT();
431 
432 	if (rule->tag)
433 		tag_unref(&V_pf_tags, rule->tag);
434 	if (rule->match_tag)
435 		tag_unref(&V_pf_tags, rule->match_tag);
436 #ifdef ALTQ
437 	if (rule->pqid != rule->qid)
438 		pf_qid_unref(rule->pqid);
439 	pf_qid_unref(rule->qid);
440 #endif
441 	switch (rule->src.addr.type) {
442 	case PF_ADDR_DYNIFTL:
443 		pfi_dynaddr_remove(rule->src.addr.p.dyn);
444 		break;
445 	case PF_ADDR_TABLE:
446 		pfr_detach_table(rule->src.addr.p.tbl);
447 		break;
448 	}
449 	switch (rule->dst.addr.type) {
450 	case PF_ADDR_DYNIFTL:
451 		pfi_dynaddr_remove(rule->dst.addr.p.dyn);
452 		break;
453 	case PF_ADDR_TABLE:
454 		pfr_detach_table(rule->dst.addr.p.tbl);
455 		break;
456 	}
457 	if (rule->overload_tbl)
458 		pfr_detach_table(rule->overload_tbl);
459 	if (rule->kif)
460 		pfi_kif_unref(rule->kif);
461 	pf_anchor_remove(rule);
462 	pf_empty_pool(&rule->rpool.list);
463 	counter_u64_free(rule->states_cur);
464 	counter_u64_free(rule->states_tot);
465 	counter_u64_free(rule->src_nodes);
466 	free(rule, M_PFRULE);
467 }
468 
469 static void
470 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
471     unsigned int default_size)
472 {
473 	unsigned int i;
474 	unsigned int hashsize;
475 
476 	if (*tunable_size == 0 || !powerof2(*tunable_size))
477 		*tunable_size = default_size;
478 
479 	hashsize = *tunable_size;
480 	ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
481 	    M_WAITOK);
482 	ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
483 	    M_WAITOK);
484 	ts->mask = hashsize - 1;
485 	ts->seed = arc4random();
486 	for (i = 0; i < hashsize; i++) {
487 		TAILQ_INIT(&ts->namehash[i]);
488 		TAILQ_INIT(&ts->taghash[i]);
489 	}
490 	BIT_FILL(TAGID_MAX, &ts->avail);
491 }
492 
493 static void
494 pf_cleanup_tagset(struct pf_tagset *ts)
495 {
496 	unsigned int i;
497 	unsigned int hashsize;
498 	struct pf_tagname *t, *tmp;
499 
500 	/*
501 	 * Only need to clean up one of the hashes as each tag is hashed
502 	 * into each table.
503 	 */
504 	hashsize = ts->mask + 1;
505 	for (i = 0; i < hashsize; i++)
506 		TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
507 			uma_zfree(V_pf_tag_z, t);
508 
509 	free(ts->namehash, M_PFHASH);
510 	free(ts->taghash, M_PFHASH);
511 }
512 
513 static uint16_t
514 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
515 {
516 
517 	return (murmur3_32_hash(tagname, strlen(tagname), ts->seed) & ts->mask);
518 }
519 
520 static uint16_t
521 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
522 {
523 
524 	return (tag & ts->mask);
525 }
526 
527 static u_int16_t
528 tagname2tag(struct pf_tagset *ts, char *tagname)
529 {
530 	struct pf_tagname	*tag;
531 	u_int32_t		 index;
532 	u_int16_t		 new_tagid;
533 
534 	PF_RULES_WASSERT();
535 
536 	index = tagname2hashindex(ts, tagname);
537 	TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
538 		if (strcmp(tagname, tag->name) == 0) {
539 			tag->ref++;
540 			return (tag->tag);
541 		}
542 
543 	/*
544 	 * new entry
545 	 *
546 	 * to avoid fragmentation, we do a linear search from the beginning
547 	 * and take the first free slot we find.
548 	 */
549 	new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
550 	/*
551 	 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
552 	 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
553 	 * set.  It may also return a bit number greater than TAGID_MAX due
554 	 * to rounding of the number of bits in the vector up to a multiple
555 	 * of the vector word size at declaration/allocation time.
556 	 */
557 	if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
558 		return (0);
559 
560 	/* Mark the tag as in use.  Bits are 0-based for BIT_CLR() */
561 	BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
562 
563 	/* allocate and fill new struct pf_tagname */
564 	tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
565 	if (tag == NULL)
566 		return (0);
567 	strlcpy(tag->name, tagname, sizeof(tag->name));
568 	tag->tag = new_tagid;
569 	tag->ref = 1;
570 
571 	/* Insert into namehash */
572 	TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
573 
574 	/* Insert into taghash */
575 	index = tag2hashindex(ts, new_tagid);
576 	TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
577 
578 	return (tag->tag);
579 }
580 
581 static void
582 tag_unref(struct pf_tagset *ts, u_int16_t tag)
583 {
584 	struct pf_tagname	*t;
585 	uint16_t		 index;
586 
587 	PF_RULES_WASSERT();
588 
589 	index = tag2hashindex(ts, tag);
590 	TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
591 		if (tag == t->tag) {
592 			if (--t->ref == 0) {
593 				TAILQ_REMOVE(&ts->taghash[index], t,
594 				    taghash_entries);
595 				index = tagname2hashindex(ts, t->name);
596 				TAILQ_REMOVE(&ts->namehash[index], t,
597 				    namehash_entries);
598 				/* Bits are 0-based for BIT_SET() */
599 				BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
600 				uma_zfree(V_pf_tag_z, t);
601 			}
602 			break;
603 		}
604 }
605 
606 static u_int16_t
607 pf_tagname2tag(char *tagname)
608 {
609 	return (tagname2tag(&V_pf_tags, tagname));
610 }
611 
612 #ifdef ALTQ
613 static u_int32_t
614 pf_qname2qid(char *qname)
615 {
616 	return ((u_int32_t)tagname2tag(&V_pf_qids, qname));
617 }
618 
619 static void
620 pf_qid_unref(u_int32_t qid)
621 {
622 	tag_unref(&V_pf_qids, (u_int16_t)qid);
623 }
624 
625 static int
626 pf_begin_altq(u_int32_t *ticket)
627 {
628 	struct pf_altq	*altq, *tmp;
629 	int		 error = 0;
630 
631 	PF_RULES_WASSERT();
632 
633 	/* Purge the old altq lists */
634 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
635 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
636 			/* detach and destroy the discipline */
637 			error = altq_remove(altq);
638 		}
639 		free(altq, M_PFALTQ);
640 	}
641 	TAILQ_INIT(V_pf_altq_ifs_inactive);
642 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
643 		pf_qid_unref(altq->qid);
644 		free(altq, M_PFALTQ);
645 	}
646 	TAILQ_INIT(V_pf_altqs_inactive);
647 	if (error)
648 		return (error);
649 	*ticket = ++V_ticket_altqs_inactive;
650 	V_altqs_inactive_open = 1;
651 	return (0);
652 }
653 
654 static int
655 pf_rollback_altq(u_int32_t ticket)
656 {
657 	struct pf_altq	*altq, *tmp;
658 	int		 error = 0;
659 
660 	PF_RULES_WASSERT();
661 
662 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
663 		return (0);
664 	/* Purge the old altq lists */
665 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
666 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
667 			/* detach and destroy the discipline */
668 			error = altq_remove(altq);
669 		}
670 		free(altq, M_PFALTQ);
671 	}
672 	TAILQ_INIT(V_pf_altq_ifs_inactive);
673 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
674 		pf_qid_unref(altq->qid);
675 		free(altq, M_PFALTQ);
676 	}
677 	TAILQ_INIT(V_pf_altqs_inactive);
678 	V_altqs_inactive_open = 0;
679 	return (error);
680 }
681 
682 static int
683 pf_commit_altq(u_int32_t ticket)
684 {
685 	struct pf_altqqueue	*old_altqs, *old_altq_ifs;
686 	struct pf_altq		*altq, *tmp;
687 	int			 err, error = 0;
688 
689 	PF_RULES_WASSERT();
690 
691 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
692 		return (EBUSY);
693 
694 	/* swap altqs, keep the old. */
695 	old_altqs = V_pf_altqs_active;
696 	old_altq_ifs = V_pf_altq_ifs_active;
697 	V_pf_altqs_active = V_pf_altqs_inactive;
698 	V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
699 	V_pf_altqs_inactive = old_altqs;
700 	V_pf_altq_ifs_inactive = old_altq_ifs;
701 	V_ticket_altqs_active = V_ticket_altqs_inactive;
702 
703 	/* Attach new disciplines */
704 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
705 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
706 			/* attach the discipline */
707 			error = altq_pfattach(altq);
708 			if (error == 0 && V_pf_altq_running)
709 				error = pf_enable_altq(altq);
710 			if (error != 0)
711 				return (error);
712 		}
713 	}
714 
715 	/* Purge the old altq lists */
716 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
717 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
718 			/* detach and destroy the discipline */
719 			if (V_pf_altq_running)
720 				error = pf_disable_altq(altq);
721 			err = altq_pfdetach(altq);
722 			if (err != 0 && error == 0)
723 				error = err;
724 			err = altq_remove(altq);
725 			if (err != 0 && error == 0)
726 				error = err;
727 		}
728 		free(altq, M_PFALTQ);
729 	}
730 	TAILQ_INIT(V_pf_altq_ifs_inactive);
731 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
732 		pf_qid_unref(altq->qid);
733 		free(altq, M_PFALTQ);
734 	}
735 	TAILQ_INIT(V_pf_altqs_inactive);
736 
737 	V_altqs_inactive_open = 0;
738 	return (error);
739 }
740 
741 static int
742 pf_enable_altq(struct pf_altq *altq)
743 {
744 	struct ifnet		*ifp;
745 	struct tb_profile	 tb;
746 	int			 error = 0;
747 
748 	if ((ifp = ifunit(altq->ifname)) == NULL)
749 		return (EINVAL);
750 
751 	if (ifp->if_snd.altq_type != ALTQT_NONE)
752 		error = altq_enable(&ifp->if_snd);
753 
754 	/* set tokenbucket regulator */
755 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
756 		tb.rate = altq->ifbandwidth;
757 		tb.depth = altq->tbrsize;
758 		error = tbr_set(&ifp->if_snd, &tb);
759 	}
760 
761 	return (error);
762 }
763 
764 static int
765 pf_disable_altq(struct pf_altq *altq)
766 {
767 	struct ifnet		*ifp;
768 	struct tb_profile	 tb;
769 	int			 error;
770 
771 	if ((ifp = ifunit(altq->ifname)) == NULL)
772 		return (EINVAL);
773 
774 	/*
775 	 * when the discipline is no longer referenced, it was overridden
776 	 * by a new one.  if so, just return.
777 	 */
778 	if (altq->altq_disc != ifp->if_snd.altq_disc)
779 		return (0);
780 
781 	error = altq_disable(&ifp->if_snd);
782 
783 	if (error == 0) {
784 		/* clear tokenbucket regulator */
785 		tb.rate = 0;
786 		error = tbr_set(&ifp->if_snd, &tb);
787 	}
788 
789 	return (error);
790 }
791 
792 static int
793 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
794     struct pf_altq *altq)
795 {
796 	struct ifnet	*ifp1;
797 	int		 error = 0;
798 
799 	/* Deactivate the interface in question */
800 	altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
801 	if ((ifp1 = ifunit(altq->ifname)) == NULL ||
802 	    (remove && ifp1 == ifp)) {
803 		altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
804 	} else {
805 		error = altq_add(ifp1, altq);
806 
807 		if (ticket != V_ticket_altqs_inactive)
808 			error = EBUSY;
809 
810 		if (error)
811 			free(altq, M_PFALTQ);
812 	}
813 
814 	return (error);
815 }
816 
817 void
818 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
819 {
820 	struct pf_altq	*a1, *a2, *a3;
821 	u_int32_t	 ticket;
822 	int		 error = 0;
823 
824 	/*
825 	 * No need to re-evaluate the configuration for events on interfaces
826 	 * that do not support ALTQ, as it's not possible for such
827 	 * interfaces to be part of the configuration.
828 	 */
829 	if (!ALTQ_IS_READY(&ifp->if_snd))
830 		return;
831 
832 	/* Interrupt userland queue modifications */
833 	if (V_altqs_inactive_open)
834 		pf_rollback_altq(V_ticket_altqs_inactive);
835 
836 	/* Start new altq ruleset */
837 	if (pf_begin_altq(&ticket))
838 		return;
839 
840 	/* Copy the current active set */
841 	TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
842 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
843 		if (a2 == NULL) {
844 			error = ENOMEM;
845 			break;
846 		}
847 		bcopy(a1, a2, sizeof(struct pf_altq));
848 
849 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
850 		if (error)
851 			break;
852 
853 		TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
854 	}
855 	if (error)
856 		goto out;
857 	TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
858 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
859 		if (a2 == NULL) {
860 			error = ENOMEM;
861 			break;
862 		}
863 		bcopy(a1, a2, sizeof(struct pf_altq));
864 
865 		if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
866 			error = EBUSY;
867 			free(a2, M_PFALTQ);
868 			break;
869 		}
870 		a2->altq_disc = NULL;
871 		TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
872 			if (strncmp(a3->ifname, a2->ifname,
873 				IFNAMSIZ) == 0) {
874 				a2->altq_disc = a3->altq_disc;
875 				break;
876 			}
877 		}
878 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
879 		if (error)
880 			break;
881 
882 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
883 	}
884 
885 out:
886 	if (error != 0)
887 		pf_rollback_altq(ticket);
888 	else
889 		pf_commit_altq(ticket);
890 }
891 #endif /* ALTQ */
892 
893 static int
894 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
895 {
896 	struct pf_ruleset	*rs;
897 	struct pf_rule		*rule;
898 
899 	PF_RULES_WASSERT();
900 
901 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
902 		return (EINVAL);
903 	rs = pf_find_or_create_ruleset(anchor);
904 	if (rs == NULL)
905 		return (EINVAL);
906 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
907 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
908 		rs->rules[rs_num].inactive.rcount--;
909 	}
910 	*ticket = ++rs->rules[rs_num].inactive.ticket;
911 	rs->rules[rs_num].inactive.open = 1;
912 	return (0);
913 }
914 
915 static int
916 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
917 {
918 	struct pf_ruleset	*rs;
919 	struct pf_rule		*rule;
920 
921 	PF_RULES_WASSERT();
922 
923 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
924 		return (EINVAL);
925 	rs = pf_find_ruleset(anchor);
926 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
927 	    rs->rules[rs_num].inactive.ticket != ticket)
928 		return (0);
929 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
930 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
931 		rs->rules[rs_num].inactive.rcount--;
932 	}
933 	rs->rules[rs_num].inactive.open = 0;
934 	return (0);
935 }
936 
937 #define PF_MD5_UPD(st, elm)						\
938 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
939 
940 #define PF_MD5_UPD_STR(st, elm)						\
941 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
942 
943 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
944 		(stor) = htonl((st)->elm);				\
945 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
946 } while (0)
947 
948 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
949 		(stor) = htons((st)->elm);				\
950 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
951 } while (0)
952 
953 static void
954 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
955 {
956 	PF_MD5_UPD(pfr, addr.type);
957 	switch (pfr->addr.type) {
958 		case PF_ADDR_DYNIFTL:
959 			PF_MD5_UPD(pfr, addr.v.ifname);
960 			PF_MD5_UPD(pfr, addr.iflags);
961 			break;
962 		case PF_ADDR_TABLE:
963 			PF_MD5_UPD(pfr, addr.v.tblname);
964 			break;
965 		case PF_ADDR_ADDRMASK:
966 			/* XXX ignore af? */
967 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
968 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
969 			break;
970 	}
971 
972 	PF_MD5_UPD(pfr, port[0]);
973 	PF_MD5_UPD(pfr, port[1]);
974 	PF_MD5_UPD(pfr, neg);
975 	PF_MD5_UPD(pfr, port_op);
976 }
977 
978 static void
979 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
980 {
981 	u_int16_t x;
982 	u_int32_t y;
983 
984 	pf_hash_rule_addr(ctx, &rule->src);
985 	pf_hash_rule_addr(ctx, &rule->dst);
986 	PF_MD5_UPD_STR(rule, label);
987 	PF_MD5_UPD_STR(rule, ifname);
988 	PF_MD5_UPD_STR(rule, match_tagname);
989 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
990 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
991 	PF_MD5_UPD_HTONL(rule, prob, y);
992 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
993 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
994 	PF_MD5_UPD(rule, uid.op);
995 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
996 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
997 	PF_MD5_UPD(rule, gid.op);
998 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
999 	PF_MD5_UPD(rule, action);
1000 	PF_MD5_UPD(rule, direction);
1001 	PF_MD5_UPD(rule, af);
1002 	PF_MD5_UPD(rule, quick);
1003 	PF_MD5_UPD(rule, ifnot);
1004 	PF_MD5_UPD(rule, match_tag_not);
1005 	PF_MD5_UPD(rule, natpass);
1006 	PF_MD5_UPD(rule, keep_state);
1007 	PF_MD5_UPD(rule, proto);
1008 	PF_MD5_UPD(rule, type);
1009 	PF_MD5_UPD(rule, code);
1010 	PF_MD5_UPD(rule, flags);
1011 	PF_MD5_UPD(rule, flagset);
1012 	PF_MD5_UPD(rule, allow_opts);
1013 	PF_MD5_UPD(rule, rt);
1014 	PF_MD5_UPD(rule, tos);
1015 }
1016 
1017 static int
1018 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1019 {
1020 	struct pf_ruleset	*rs;
1021 	struct pf_rule		*rule, **old_array;
1022 	struct pf_rulequeue	*old_rules;
1023 	int			 error;
1024 	u_int32_t		 old_rcount;
1025 
1026 	PF_RULES_WASSERT();
1027 
1028 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1029 		return (EINVAL);
1030 	rs = pf_find_ruleset(anchor);
1031 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1032 	    ticket != rs->rules[rs_num].inactive.ticket)
1033 		return (EBUSY);
1034 
1035 	/* Calculate checksum for the main ruleset */
1036 	if (rs == &pf_main_ruleset) {
1037 		error = pf_setup_pfsync_matching(rs);
1038 		if (error != 0)
1039 			return (error);
1040 	}
1041 
1042 	/* Swap rules, keep the old. */
1043 	old_rules = rs->rules[rs_num].active.ptr;
1044 	old_rcount = rs->rules[rs_num].active.rcount;
1045 	old_array = rs->rules[rs_num].active.ptr_array;
1046 
1047 	rs->rules[rs_num].active.ptr =
1048 	    rs->rules[rs_num].inactive.ptr;
1049 	rs->rules[rs_num].active.ptr_array =
1050 	    rs->rules[rs_num].inactive.ptr_array;
1051 	rs->rules[rs_num].active.rcount =
1052 	    rs->rules[rs_num].inactive.rcount;
1053 	rs->rules[rs_num].inactive.ptr = old_rules;
1054 	rs->rules[rs_num].inactive.ptr_array = old_array;
1055 	rs->rules[rs_num].inactive.rcount = old_rcount;
1056 
1057 	rs->rules[rs_num].active.ticket =
1058 	    rs->rules[rs_num].inactive.ticket;
1059 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1060 
1061 
1062 	/* Purge the old rule list. */
1063 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1064 		pf_unlink_rule(old_rules, rule);
1065 	if (rs->rules[rs_num].inactive.ptr_array)
1066 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1067 	rs->rules[rs_num].inactive.ptr_array = NULL;
1068 	rs->rules[rs_num].inactive.rcount = 0;
1069 	rs->rules[rs_num].inactive.open = 0;
1070 	pf_remove_if_empty_ruleset(rs);
1071 
1072 	return (0);
1073 }
1074 
1075 static int
1076 pf_setup_pfsync_matching(struct pf_ruleset *rs)
1077 {
1078 	MD5_CTX			 ctx;
1079 	struct pf_rule		*rule;
1080 	int			 rs_cnt;
1081 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
1082 
1083 	MD5Init(&ctx);
1084 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1085 		/* XXX PF_RULESET_SCRUB as well? */
1086 		if (rs_cnt == PF_RULESET_SCRUB)
1087 			continue;
1088 
1089 		if (rs->rules[rs_cnt].inactive.ptr_array)
1090 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1091 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
1092 
1093 		if (rs->rules[rs_cnt].inactive.rcount) {
1094 			rs->rules[rs_cnt].inactive.ptr_array =
1095 			    malloc(sizeof(caddr_t) *
1096 			    rs->rules[rs_cnt].inactive.rcount,
1097 			    M_TEMP, M_NOWAIT);
1098 
1099 			if (!rs->rules[rs_cnt].inactive.ptr_array)
1100 				return (ENOMEM);
1101 		}
1102 
1103 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1104 		    entries) {
1105 			pf_hash_rule(&ctx, rule);
1106 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1107 		}
1108 	}
1109 
1110 	MD5Final(digest, &ctx);
1111 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1112 	return (0);
1113 }
1114 
1115 static int
1116 pf_addr_setup(struct pf_ruleset *ruleset, struct pf_addr_wrap *addr,
1117     sa_family_t af)
1118 {
1119 	int error = 0;
1120 
1121 	switch (addr->type) {
1122 	case PF_ADDR_TABLE:
1123 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1124 		if (addr->p.tbl == NULL)
1125 			error = ENOMEM;
1126 		break;
1127 	case PF_ADDR_DYNIFTL:
1128 		error = pfi_dynaddr_setup(addr, af);
1129 		break;
1130 	}
1131 
1132 	return (error);
1133 }
1134 
1135 static void
1136 pf_addr_copyout(struct pf_addr_wrap *addr)
1137 {
1138 
1139 	switch (addr->type) {
1140 	case PF_ADDR_DYNIFTL:
1141 		pfi_dynaddr_copyout(addr);
1142 		break;
1143 	case PF_ADDR_TABLE:
1144 		pf_tbladdr_copyout(addr);
1145 		break;
1146 	}
1147 }
1148 
1149 #ifdef ALTQ
1150 /*
1151  * Handle export of struct pf_kaltq to user binaries that may be using any
1152  * version of struct pf_altq.
1153  */
1154 static int
1155 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1156 {
1157 	u_int32_t version;
1158 
1159 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1160 		version = 0;
1161 	else
1162 		version = pa->version;
1163 
1164 	if (version > PFIOC_ALTQ_VERSION)
1165 		return (EINVAL);
1166 
1167 #define ASSIGN(x) exported_q->x = q->x
1168 #define COPY(x) \
1169 	bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1170 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1171 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1172 
1173 	switch (version) {
1174 	case 0: {
1175 		struct pf_altq_v0 *exported_q =
1176 		    &((struct pfioc_altq_v0 *)pa)->altq;
1177 
1178 		COPY(ifname);
1179 
1180 		ASSIGN(scheduler);
1181 		ASSIGN(tbrsize);
1182 		exported_q->tbrsize = SATU16(q->tbrsize);
1183 		exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1184 
1185 		COPY(qname);
1186 		COPY(parent);
1187 		ASSIGN(parent_qid);
1188 		exported_q->bandwidth = SATU32(q->bandwidth);
1189 		ASSIGN(priority);
1190 		ASSIGN(local_flags);
1191 
1192 		ASSIGN(qlimit);
1193 		ASSIGN(flags);
1194 
1195 		if (q->scheduler == ALTQT_HFSC) {
1196 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1197 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1198 			    SATU32(q->pq_u.hfsc_opts.x)
1199 
1200 			ASSIGN_OPT_SATU32(rtsc_m1);
1201 			ASSIGN_OPT(rtsc_d);
1202 			ASSIGN_OPT_SATU32(rtsc_m2);
1203 
1204 			ASSIGN_OPT_SATU32(lssc_m1);
1205 			ASSIGN_OPT(lssc_d);
1206 			ASSIGN_OPT_SATU32(lssc_m2);
1207 
1208 			ASSIGN_OPT_SATU32(ulsc_m1);
1209 			ASSIGN_OPT(ulsc_d);
1210 			ASSIGN_OPT_SATU32(ulsc_m2);
1211 
1212 			ASSIGN_OPT(flags);
1213 
1214 #undef ASSIGN_OPT
1215 #undef ASSIGN_OPT_SATU32
1216 		} else
1217 			COPY(pq_u);
1218 
1219 		ASSIGN(qid);
1220 		break;
1221 	}
1222 	case 1:	{
1223 		struct pf_altq_v1 *exported_q =
1224 		    &((struct pfioc_altq_v1 *)pa)->altq;
1225 
1226 		COPY(ifname);
1227 
1228 		ASSIGN(scheduler);
1229 		ASSIGN(tbrsize);
1230 		ASSIGN(ifbandwidth);
1231 
1232 		COPY(qname);
1233 		COPY(parent);
1234 		ASSIGN(parent_qid);
1235 		ASSIGN(bandwidth);
1236 		ASSIGN(priority);
1237 		ASSIGN(local_flags);
1238 
1239 		ASSIGN(qlimit);
1240 		ASSIGN(flags);
1241 		COPY(pq_u);
1242 
1243 		ASSIGN(qid);
1244 		break;
1245 	}
1246 	default:
1247 		panic("%s: unhandled struct pfioc_altq version", __func__);
1248 		break;
1249 	}
1250 
1251 #undef ASSIGN
1252 #undef COPY
1253 #undef SATU16
1254 #undef SATU32
1255 
1256 	return (0);
1257 }
1258 
1259 /*
1260  * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1261  * that may be using any version of it.
1262  */
1263 static int
1264 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1265 {
1266 	u_int32_t version;
1267 
1268 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1269 		version = 0;
1270 	else
1271 		version = pa->version;
1272 
1273 	if (version > PFIOC_ALTQ_VERSION)
1274 		return (EINVAL);
1275 
1276 #define ASSIGN(x) q->x = imported_q->x
1277 #define COPY(x) \
1278 	bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1279 
1280 	switch (version) {
1281 	case 0: {
1282 		struct pf_altq_v0 *imported_q =
1283 		    &((struct pfioc_altq_v0 *)pa)->altq;
1284 
1285 		COPY(ifname);
1286 
1287 		ASSIGN(scheduler);
1288 		ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1289 		ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1290 
1291 		COPY(qname);
1292 		COPY(parent);
1293 		ASSIGN(parent_qid);
1294 		ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1295 		ASSIGN(priority);
1296 		ASSIGN(local_flags);
1297 
1298 		ASSIGN(qlimit);
1299 		ASSIGN(flags);
1300 
1301 		if (imported_q->scheduler == ALTQT_HFSC) {
1302 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1303 
1304 			/*
1305 			 * The m1 and m2 parameters are being copied from
1306 			 * 32-bit to 64-bit.
1307 			 */
1308 			ASSIGN_OPT(rtsc_m1);
1309 			ASSIGN_OPT(rtsc_d);
1310 			ASSIGN_OPT(rtsc_m2);
1311 
1312 			ASSIGN_OPT(lssc_m1);
1313 			ASSIGN_OPT(lssc_d);
1314 			ASSIGN_OPT(lssc_m2);
1315 
1316 			ASSIGN_OPT(ulsc_m1);
1317 			ASSIGN_OPT(ulsc_d);
1318 			ASSIGN_OPT(ulsc_m2);
1319 
1320 			ASSIGN_OPT(flags);
1321 
1322 #undef ASSIGN_OPT
1323 		} else
1324 			COPY(pq_u);
1325 
1326 		ASSIGN(qid);
1327 		break;
1328 	}
1329 	case 1: {
1330 		struct pf_altq_v1 *imported_q =
1331 		    &((struct pfioc_altq_v1 *)pa)->altq;
1332 
1333 		COPY(ifname);
1334 
1335 		ASSIGN(scheduler);
1336 		ASSIGN(tbrsize);
1337 		ASSIGN(ifbandwidth);
1338 
1339 		COPY(qname);
1340 		COPY(parent);
1341 		ASSIGN(parent_qid);
1342 		ASSIGN(bandwidth);
1343 		ASSIGN(priority);
1344 		ASSIGN(local_flags);
1345 
1346 		ASSIGN(qlimit);
1347 		ASSIGN(flags);
1348 		COPY(pq_u);
1349 
1350 		ASSIGN(qid);
1351 		break;
1352 	}
1353 	default:
1354 		panic("%s: unhandled struct pfioc_altq version", __func__);
1355 		break;
1356 	}
1357 
1358 #undef ASSIGN
1359 #undef COPY
1360 
1361 	return (0);
1362 }
1363 
1364 static struct pf_altq *
1365 pf_altq_get_nth_active(u_int32_t n)
1366 {
1367 	struct pf_altq		*altq;
1368 	u_int32_t		 nr;
1369 
1370 	nr = 0;
1371 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1372 		if (nr == n)
1373 			return (altq);
1374 		nr++;
1375 	}
1376 
1377 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1378 		if (nr == n)
1379 			return (altq);
1380 		nr++;
1381 	}
1382 
1383 	return (NULL);
1384 }
1385 #endif /* ALTQ */
1386 
1387 static int
1388 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
1389 {
1390 	int			 error = 0;
1391 	PF_RULES_RLOCK_TRACKER;
1392 
1393 	/* XXX keep in sync with switch() below */
1394 	if (securelevel_gt(td->td_ucred, 2))
1395 		switch (cmd) {
1396 		case DIOCGETRULES:
1397 		case DIOCGETRULE:
1398 		case DIOCGETADDRS:
1399 		case DIOCGETADDR:
1400 		case DIOCGETSTATE:
1401 		case DIOCSETSTATUSIF:
1402 		case DIOCGETSTATUS:
1403 		case DIOCCLRSTATUS:
1404 		case DIOCNATLOOK:
1405 		case DIOCSETDEBUG:
1406 		case DIOCGETSTATES:
1407 		case DIOCGETTIMEOUT:
1408 		case DIOCCLRRULECTRS:
1409 		case DIOCGETLIMIT:
1410 		case DIOCGETALTQSV0:
1411 		case DIOCGETALTQSV1:
1412 		case DIOCGETALTQV0:
1413 		case DIOCGETALTQV1:
1414 		case DIOCGETQSTATSV0:
1415 		case DIOCGETQSTATSV1:
1416 		case DIOCGETRULESETS:
1417 		case DIOCGETRULESET:
1418 		case DIOCRGETTABLES:
1419 		case DIOCRGETTSTATS:
1420 		case DIOCRCLRTSTATS:
1421 		case DIOCRCLRADDRS:
1422 		case DIOCRADDADDRS:
1423 		case DIOCRDELADDRS:
1424 		case DIOCRSETADDRS:
1425 		case DIOCRGETADDRS:
1426 		case DIOCRGETASTATS:
1427 		case DIOCRCLRASTATS:
1428 		case DIOCRTSTADDRS:
1429 		case DIOCOSFPGET:
1430 		case DIOCGETSRCNODES:
1431 		case DIOCCLRSRCNODES:
1432 		case DIOCIGETIFACES:
1433 		case DIOCGIFSPEEDV0:
1434 		case DIOCGIFSPEEDV1:
1435 		case DIOCSETIFFLAG:
1436 		case DIOCCLRIFFLAG:
1437 			break;
1438 		case DIOCRCLRTABLES:
1439 		case DIOCRADDTABLES:
1440 		case DIOCRDELTABLES:
1441 		case DIOCRSETTFLAGS:
1442 			if (((struct pfioc_table *)addr)->pfrio_flags &
1443 			    PFR_FLAG_DUMMY)
1444 				break; /* dummy operation ok */
1445 			return (EPERM);
1446 		default:
1447 			return (EPERM);
1448 		}
1449 
1450 	if (!(flags & FWRITE))
1451 		switch (cmd) {
1452 		case DIOCGETRULES:
1453 		case DIOCGETADDRS:
1454 		case DIOCGETADDR:
1455 		case DIOCGETSTATE:
1456 		case DIOCGETSTATUS:
1457 		case DIOCGETSTATES:
1458 		case DIOCGETTIMEOUT:
1459 		case DIOCGETLIMIT:
1460 		case DIOCGETALTQSV0:
1461 		case DIOCGETALTQSV1:
1462 		case DIOCGETALTQV0:
1463 		case DIOCGETALTQV1:
1464 		case DIOCGETQSTATSV0:
1465 		case DIOCGETQSTATSV1:
1466 		case DIOCGETRULESETS:
1467 		case DIOCGETRULESET:
1468 		case DIOCNATLOOK:
1469 		case DIOCRGETTABLES:
1470 		case DIOCRGETTSTATS:
1471 		case DIOCRGETADDRS:
1472 		case DIOCRGETASTATS:
1473 		case DIOCRTSTADDRS:
1474 		case DIOCOSFPGET:
1475 		case DIOCGETSRCNODES:
1476 		case DIOCIGETIFACES:
1477 		case DIOCGIFSPEEDV1:
1478 		case DIOCGIFSPEEDV0:
1479 			break;
1480 		case DIOCRCLRTABLES:
1481 		case DIOCRADDTABLES:
1482 		case DIOCRDELTABLES:
1483 		case DIOCRCLRTSTATS:
1484 		case DIOCRCLRADDRS:
1485 		case DIOCRADDADDRS:
1486 		case DIOCRDELADDRS:
1487 		case DIOCRSETADDRS:
1488 		case DIOCRSETTFLAGS:
1489 			if (((struct pfioc_table *)addr)->pfrio_flags &
1490 			    PFR_FLAG_DUMMY) {
1491 				flags |= FWRITE; /* need write lock for dummy */
1492 				break; /* dummy operation ok */
1493 			}
1494 			return (EACCES);
1495 		case DIOCGETRULE:
1496 			if (((struct pfioc_rule *)addr)->action ==
1497 			    PF_GET_CLR_CNTR)
1498 				return (EACCES);
1499 			break;
1500 		default:
1501 			return (EACCES);
1502 		}
1503 
1504 	CURVNET_SET(TD_TO_VNET(td));
1505 
1506 	switch (cmd) {
1507 	case DIOCSTART:
1508 		sx_xlock(&pf_ioctl_lock);
1509 		if (V_pf_status.running)
1510 			error = EEXIST;
1511 		else {
1512 			int cpu;
1513 
1514 			error = hook_pf();
1515 			if (error) {
1516 				DPFPRINTF(PF_DEBUG_MISC,
1517 				    ("pf: pfil registration failed\n"));
1518 				break;
1519 			}
1520 			V_pf_status.running = 1;
1521 			V_pf_status.since = time_second;
1522 
1523 			CPU_FOREACH(cpu)
1524 				V_pf_stateid[cpu] = time_second;
1525 
1526 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1527 		}
1528 		break;
1529 
1530 	case DIOCSTOP:
1531 		sx_xlock(&pf_ioctl_lock);
1532 		if (!V_pf_status.running)
1533 			error = ENOENT;
1534 		else {
1535 			V_pf_status.running = 0;
1536 			error = dehook_pf();
1537 			if (error) {
1538 				V_pf_status.running = 1;
1539 				DPFPRINTF(PF_DEBUG_MISC,
1540 				    ("pf: pfil unregistration failed\n"));
1541 			}
1542 			V_pf_status.since = time_second;
1543 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1544 		}
1545 		break;
1546 
1547 	case DIOCADDRULE: {
1548 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1549 		struct pf_ruleset	*ruleset;
1550 		struct pf_rule		*rule, *tail;
1551 		struct pf_pooladdr	*pa;
1552 		struct pfi_kif		*kif = NULL;
1553 		int			 rs_num;
1554 
1555 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1556 			error = EINVAL;
1557 			break;
1558 		}
1559 		if (pr->rule.src.addr.p.dyn != NULL ||
1560 		    pr->rule.dst.addr.p.dyn != NULL) {
1561 			error = EINVAL;
1562 			break;
1563 		}
1564 #ifndef INET
1565 		if (pr->rule.af == AF_INET) {
1566 			error = EAFNOSUPPORT;
1567 			break;
1568 		}
1569 #endif /* INET */
1570 #ifndef INET6
1571 		if (pr->rule.af == AF_INET6) {
1572 			error = EAFNOSUPPORT;
1573 			break;
1574 		}
1575 #endif /* INET6 */
1576 
1577 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
1578 		bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1579 		if (rule->ifname[0])
1580 			kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1581 		rule->states_cur = counter_u64_alloc(M_WAITOK);
1582 		rule->states_tot = counter_u64_alloc(M_WAITOK);
1583 		rule->src_nodes = counter_u64_alloc(M_WAITOK);
1584 		rule->cuid = td->td_ucred->cr_ruid;
1585 		rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1586 		TAILQ_INIT(&rule->rpool.list);
1587 
1588 #define	ERROUT(x)	{ error = (x); goto DIOCADDRULE_error; }
1589 
1590 		PF_RULES_WLOCK();
1591 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1592 		ruleset = pf_find_ruleset(pr->anchor);
1593 		if (ruleset == NULL)
1594 			ERROUT(EINVAL);
1595 		rs_num = pf_get_ruleset_number(pr->rule.action);
1596 		if (rs_num >= PF_RULESET_MAX)
1597 			ERROUT(EINVAL);
1598 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1599 			DPFPRINTF(PF_DEBUG_MISC,
1600 			    ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
1601 			    ruleset->rules[rs_num].inactive.ticket));
1602 			ERROUT(EBUSY);
1603 		}
1604 		if (pr->pool_ticket != V_ticket_pabuf) {
1605 			DPFPRINTF(PF_DEBUG_MISC,
1606 			    ("pool_ticket: %d != %d\n", pr->pool_ticket,
1607 			    V_ticket_pabuf));
1608 			ERROUT(EBUSY);
1609 		}
1610 
1611 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1612 		    pf_rulequeue);
1613 		if (tail)
1614 			rule->nr = tail->nr + 1;
1615 		else
1616 			rule->nr = 0;
1617 		if (rule->ifname[0]) {
1618 			rule->kif = pfi_kif_attach(kif, rule->ifname);
1619 			pfi_kif_ref(rule->kif);
1620 		} else
1621 			rule->kif = NULL;
1622 
1623 		if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
1624 			error = EBUSY;
1625 
1626 #ifdef ALTQ
1627 		/* set queue IDs */
1628 		if (rule->qname[0] != 0) {
1629 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1630 				error = EBUSY;
1631 			else if (rule->pqname[0] != 0) {
1632 				if ((rule->pqid =
1633 				    pf_qname2qid(rule->pqname)) == 0)
1634 					error = EBUSY;
1635 			} else
1636 				rule->pqid = rule->qid;
1637 		}
1638 #endif
1639 		if (rule->tagname[0])
1640 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1641 				error = EBUSY;
1642 		if (rule->match_tagname[0])
1643 			if ((rule->match_tag =
1644 			    pf_tagname2tag(rule->match_tagname)) == 0)
1645 				error = EBUSY;
1646 		if (rule->rt && !rule->direction)
1647 			error = EINVAL;
1648 		if (!rule->log)
1649 			rule->logif = 0;
1650 		if (rule->logif >= PFLOGIFS_MAX)
1651 			error = EINVAL;
1652 		if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1653 			error = ENOMEM;
1654 		if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1655 			error = ENOMEM;
1656 		if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1657 			error = EINVAL;
1658 		if (rule->scrub_flags & PFSTATE_SETPRIO &&
1659 		    (rule->set_prio[0] > PF_PRIO_MAX ||
1660 		    rule->set_prio[1] > PF_PRIO_MAX))
1661 			error = EINVAL;
1662 		TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1663 			if (pa->addr.type == PF_ADDR_TABLE) {
1664 				pa->addr.p.tbl = pfr_attach_table(ruleset,
1665 				    pa->addr.v.tblname);
1666 				if (pa->addr.p.tbl == NULL)
1667 					error = ENOMEM;
1668 			}
1669 
1670 		rule->overload_tbl = NULL;
1671 		if (rule->overload_tblname[0]) {
1672 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1673 			    rule->overload_tblname)) == NULL)
1674 				error = EINVAL;
1675 			else
1676 				rule->overload_tbl->pfrkt_flags |=
1677 				    PFR_TFLAG_ACTIVE;
1678 		}
1679 
1680 		pf_mv_pool(&V_pf_pabuf, &rule->rpool.list);
1681 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1682 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1683 		    (rule->rt > PF_NOPFROUTE)) &&
1684 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1685 			error = EINVAL;
1686 
1687 		if (error) {
1688 			pf_free_rule(rule);
1689 			PF_RULES_WUNLOCK();
1690 			break;
1691 		}
1692 
1693 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1694 		rule->evaluations = rule->packets[0] = rule->packets[1] =
1695 		    rule->bytes[0] = rule->bytes[1] = 0;
1696 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1697 		    rule, entries);
1698 		ruleset->rules[rs_num].inactive.rcount++;
1699 		PF_RULES_WUNLOCK();
1700 		break;
1701 
1702 #undef ERROUT
1703 DIOCADDRULE_error:
1704 		PF_RULES_WUNLOCK();
1705 		counter_u64_free(rule->states_cur);
1706 		counter_u64_free(rule->states_tot);
1707 		counter_u64_free(rule->src_nodes);
1708 		free(rule, M_PFRULE);
1709 		if (kif)
1710 			free(kif, PFI_MTYPE);
1711 		break;
1712 	}
1713 
1714 	case DIOCGETRULES: {
1715 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1716 		struct pf_ruleset	*ruleset;
1717 		struct pf_rule		*tail;
1718 		int			 rs_num;
1719 
1720 		PF_RULES_WLOCK();
1721 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1722 		ruleset = pf_find_ruleset(pr->anchor);
1723 		if (ruleset == NULL) {
1724 			PF_RULES_WUNLOCK();
1725 			error = EINVAL;
1726 			break;
1727 		}
1728 		rs_num = pf_get_ruleset_number(pr->rule.action);
1729 		if (rs_num >= PF_RULESET_MAX) {
1730 			PF_RULES_WUNLOCK();
1731 			error = EINVAL;
1732 			break;
1733 		}
1734 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1735 		    pf_rulequeue);
1736 		if (tail)
1737 			pr->nr = tail->nr + 1;
1738 		else
1739 			pr->nr = 0;
1740 		pr->ticket = ruleset->rules[rs_num].active.ticket;
1741 		PF_RULES_WUNLOCK();
1742 		break;
1743 	}
1744 
1745 	case DIOCGETRULE: {
1746 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1747 		struct pf_ruleset	*ruleset;
1748 		struct pf_rule		*rule;
1749 		int			 rs_num, i;
1750 
1751 		PF_RULES_WLOCK();
1752 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1753 		ruleset = pf_find_ruleset(pr->anchor);
1754 		if (ruleset == NULL) {
1755 			PF_RULES_WUNLOCK();
1756 			error = EINVAL;
1757 			break;
1758 		}
1759 		rs_num = pf_get_ruleset_number(pr->rule.action);
1760 		if (rs_num >= PF_RULESET_MAX) {
1761 			PF_RULES_WUNLOCK();
1762 			error = EINVAL;
1763 			break;
1764 		}
1765 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1766 			PF_RULES_WUNLOCK();
1767 			error = EBUSY;
1768 			break;
1769 		}
1770 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1771 		while ((rule != NULL) && (rule->nr != pr->nr))
1772 			rule = TAILQ_NEXT(rule, entries);
1773 		if (rule == NULL) {
1774 			PF_RULES_WUNLOCK();
1775 			error = EBUSY;
1776 			break;
1777 		}
1778 		bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1779 		pr->rule.u_states_cur = counter_u64_fetch(rule->states_cur);
1780 		pr->rule.u_states_tot = counter_u64_fetch(rule->states_tot);
1781 		pr->rule.u_src_nodes = counter_u64_fetch(rule->src_nodes);
1782 		if (pf_anchor_copyout(ruleset, rule, pr)) {
1783 			PF_RULES_WUNLOCK();
1784 			error = EBUSY;
1785 			break;
1786 		}
1787 		pf_addr_copyout(&pr->rule.src.addr);
1788 		pf_addr_copyout(&pr->rule.dst.addr);
1789 		for (i = 0; i < PF_SKIP_COUNT; ++i)
1790 			if (rule->skip[i].ptr == NULL)
1791 				pr->rule.skip[i].nr = -1;
1792 			else
1793 				pr->rule.skip[i].nr =
1794 				    rule->skip[i].ptr->nr;
1795 
1796 		if (pr->action == PF_GET_CLR_CNTR) {
1797 			rule->evaluations = 0;
1798 			rule->packets[0] = rule->packets[1] = 0;
1799 			rule->bytes[0] = rule->bytes[1] = 0;
1800 			counter_u64_zero(rule->states_tot);
1801 		}
1802 		PF_RULES_WUNLOCK();
1803 		break;
1804 	}
1805 
1806 	case DIOCCHANGERULE: {
1807 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1808 		struct pf_ruleset	*ruleset;
1809 		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1810 		struct pfi_kif		*kif = NULL;
1811 		struct pf_pooladdr	*pa;
1812 		u_int32_t		 nr = 0;
1813 		int			 rs_num;
1814 
1815 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1816 		    pcr->action > PF_CHANGE_GET_TICKET) {
1817 			error = EINVAL;
1818 			break;
1819 		}
1820 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1821 			error = EINVAL;
1822 			break;
1823 		}
1824 
1825 		if (pcr->action != PF_CHANGE_REMOVE) {
1826 #ifndef INET
1827 			if (pcr->rule.af == AF_INET) {
1828 				error = EAFNOSUPPORT;
1829 				break;
1830 			}
1831 #endif /* INET */
1832 #ifndef INET6
1833 			if (pcr->rule.af == AF_INET6) {
1834 				error = EAFNOSUPPORT;
1835 				break;
1836 			}
1837 #endif /* INET6 */
1838 			newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
1839 			bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1840 			if (newrule->ifname[0])
1841 				kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1842 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
1843 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
1844 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
1845 			newrule->cuid = td->td_ucred->cr_ruid;
1846 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1847 			TAILQ_INIT(&newrule->rpool.list);
1848 		}
1849 
1850 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGERULE_error; }
1851 
1852 		PF_RULES_WLOCK();
1853 		if (!(pcr->action == PF_CHANGE_REMOVE ||
1854 		    pcr->action == PF_CHANGE_GET_TICKET) &&
1855 		    pcr->pool_ticket != V_ticket_pabuf)
1856 			ERROUT(EBUSY);
1857 
1858 		ruleset = pf_find_ruleset(pcr->anchor);
1859 		if (ruleset == NULL)
1860 			ERROUT(EINVAL);
1861 
1862 		rs_num = pf_get_ruleset_number(pcr->rule.action);
1863 		if (rs_num >= PF_RULESET_MAX)
1864 			ERROUT(EINVAL);
1865 
1866 		if (pcr->action == PF_CHANGE_GET_TICKET) {
1867 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1868 			ERROUT(0);
1869 		} else if (pcr->ticket !=
1870 			    ruleset->rules[rs_num].active.ticket)
1871 				ERROUT(EINVAL);
1872 
1873 		if (pcr->action != PF_CHANGE_REMOVE) {
1874 			if (newrule->ifname[0]) {
1875 				newrule->kif = pfi_kif_attach(kif,
1876 				    newrule->ifname);
1877 				pfi_kif_ref(newrule->kif);
1878 			} else
1879 				newrule->kif = NULL;
1880 
1881 			if (newrule->rtableid > 0 &&
1882 			    newrule->rtableid >= rt_numfibs)
1883 				error = EBUSY;
1884 
1885 #ifdef ALTQ
1886 			/* set queue IDs */
1887 			if (newrule->qname[0] != 0) {
1888 				if ((newrule->qid =
1889 				    pf_qname2qid(newrule->qname)) == 0)
1890 					error = EBUSY;
1891 				else if (newrule->pqname[0] != 0) {
1892 					if ((newrule->pqid =
1893 					    pf_qname2qid(newrule->pqname)) == 0)
1894 						error = EBUSY;
1895 				} else
1896 					newrule->pqid = newrule->qid;
1897 			}
1898 #endif /* ALTQ */
1899 			if (newrule->tagname[0])
1900 				if ((newrule->tag =
1901 				    pf_tagname2tag(newrule->tagname)) == 0)
1902 					error = EBUSY;
1903 			if (newrule->match_tagname[0])
1904 				if ((newrule->match_tag = pf_tagname2tag(
1905 				    newrule->match_tagname)) == 0)
1906 					error = EBUSY;
1907 			if (newrule->rt && !newrule->direction)
1908 				error = EINVAL;
1909 			if (!newrule->log)
1910 				newrule->logif = 0;
1911 			if (newrule->logif >= PFLOGIFS_MAX)
1912 				error = EINVAL;
1913 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
1914 				error = ENOMEM;
1915 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
1916 				error = ENOMEM;
1917 			if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1918 				error = EINVAL;
1919 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1920 				if (pa->addr.type == PF_ADDR_TABLE) {
1921 					pa->addr.p.tbl =
1922 					    pfr_attach_table(ruleset,
1923 					    pa->addr.v.tblname);
1924 					if (pa->addr.p.tbl == NULL)
1925 						error = ENOMEM;
1926 				}
1927 
1928 			newrule->overload_tbl = NULL;
1929 			if (newrule->overload_tblname[0]) {
1930 				if ((newrule->overload_tbl = pfr_attach_table(
1931 				    ruleset, newrule->overload_tblname)) ==
1932 				    NULL)
1933 					error = EINVAL;
1934 				else
1935 					newrule->overload_tbl->pfrkt_flags |=
1936 					    PFR_TFLAG_ACTIVE;
1937 			}
1938 
1939 			pf_mv_pool(&V_pf_pabuf, &newrule->rpool.list);
1940 			if (((((newrule->action == PF_NAT) ||
1941 			    (newrule->action == PF_RDR) ||
1942 			    (newrule->action == PF_BINAT) ||
1943 			    (newrule->rt > PF_NOPFROUTE)) &&
1944 			    !newrule->anchor)) &&
1945 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1946 				error = EINVAL;
1947 
1948 			if (error) {
1949 				pf_free_rule(newrule);
1950 				PF_RULES_WUNLOCK();
1951 				break;
1952 			}
1953 
1954 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1955 			newrule->evaluations = 0;
1956 			newrule->packets[0] = newrule->packets[1] = 0;
1957 			newrule->bytes[0] = newrule->bytes[1] = 0;
1958 		}
1959 		pf_empty_pool(&V_pf_pabuf);
1960 
1961 		if (pcr->action == PF_CHANGE_ADD_HEAD)
1962 			oldrule = TAILQ_FIRST(
1963 			    ruleset->rules[rs_num].active.ptr);
1964 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1965 			oldrule = TAILQ_LAST(
1966 			    ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1967 		else {
1968 			oldrule = TAILQ_FIRST(
1969 			    ruleset->rules[rs_num].active.ptr);
1970 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1971 				oldrule = TAILQ_NEXT(oldrule, entries);
1972 			if (oldrule == NULL) {
1973 				if (newrule != NULL)
1974 					pf_free_rule(newrule);
1975 				PF_RULES_WUNLOCK();
1976 				error = EINVAL;
1977 				break;
1978 			}
1979 		}
1980 
1981 		if (pcr->action == PF_CHANGE_REMOVE) {
1982 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
1983 			    oldrule);
1984 			ruleset->rules[rs_num].active.rcount--;
1985 		} else {
1986 			if (oldrule == NULL)
1987 				TAILQ_INSERT_TAIL(
1988 				    ruleset->rules[rs_num].active.ptr,
1989 				    newrule, entries);
1990 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1991 			    pcr->action == PF_CHANGE_ADD_BEFORE)
1992 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1993 			else
1994 				TAILQ_INSERT_AFTER(
1995 				    ruleset->rules[rs_num].active.ptr,
1996 				    oldrule, newrule, entries);
1997 			ruleset->rules[rs_num].active.rcount++;
1998 		}
1999 
2000 		nr = 0;
2001 		TAILQ_FOREACH(oldrule,
2002 		    ruleset->rules[rs_num].active.ptr, entries)
2003 			oldrule->nr = nr++;
2004 
2005 		ruleset->rules[rs_num].active.ticket++;
2006 
2007 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
2008 		pf_remove_if_empty_ruleset(ruleset);
2009 
2010 		PF_RULES_WUNLOCK();
2011 		break;
2012 
2013 #undef ERROUT
2014 DIOCCHANGERULE_error:
2015 		PF_RULES_WUNLOCK();
2016 		if (newrule != NULL) {
2017 			counter_u64_free(newrule->states_cur);
2018 			counter_u64_free(newrule->states_tot);
2019 			counter_u64_free(newrule->src_nodes);
2020 			free(newrule, M_PFRULE);
2021 		}
2022 		if (kif != NULL)
2023 			free(kif, PFI_MTYPE);
2024 		break;
2025 	}
2026 
2027 	case DIOCCLRSTATES: {
2028 		struct pf_state		*s;
2029 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
2030 		u_int			 i, killed = 0;
2031 
2032 		for (i = 0; i <= pf_hashmask; i++) {
2033 			struct pf_idhash *ih = &V_pf_idhash[i];
2034 
2035 relock_DIOCCLRSTATES:
2036 			PF_HASHROW_LOCK(ih);
2037 			LIST_FOREACH(s, &ih->states, entry)
2038 				if (!psk->psk_ifname[0] ||
2039 				    !strcmp(psk->psk_ifname,
2040 				    s->kif->pfik_name)) {
2041 					/*
2042 					 * Don't send out individual
2043 					 * delete messages.
2044 					 */
2045 					s->state_flags |= PFSTATE_NOSYNC;
2046 					pf_unlink_state(s, PF_ENTER_LOCKED);
2047 					killed++;
2048 					goto relock_DIOCCLRSTATES;
2049 				}
2050 			PF_HASHROW_UNLOCK(ih);
2051 		}
2052 		psk->psk_killed = killed;
2053 		if (V_pfsync_clear_states_ptr != NULL)
2054 			V_pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
2055 		break;
2056 	}
2057 
2058 	case DIOCKILLSTATES: {
2059 		struct pf_state		*s;
2060 		struct pf_state_key	*sk;
2061 		struct pf_addr		*srcaddr, *dstaddr;
2062 		u_int16_t		 srcport, dstport;
2063 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
2064 		u_int			 i, killed = 0;
2065 
2066 		if (psk->psk_pfcmp.id) {
2067 			if (psk->psk_pfcmp.creatorid == 0)
2068 				psk->psk_pfcmp.creatorid = V_pf_status.hostid;
2069 			if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
2070 			    psk->psk_pfcmp.creatorid))) {
2071 				pf_unlink_state(s, PF_ENTER_LOCKED);
2072 				psk->psk_killed = 1;
2073 			}
2074 			break;
2075 		}
2076 
2077 		for (i = 0; i <= pf_hashmask; i++) {
2078 			struct pf_idhash *ih = &V_pf_idhash[i];
2079 
2080 relock_DIOCKILLSTATES:
2081 			PF_HASHROW_LOCK(ih);
2082 			LIST_FOREACH(s, &ih->states, entry) {
2083 				sk = s->key[PF_SK_WIRE];
2084 				if (s->direction == PF_OUT) {
2085 					srcaddr = &sk->addr[1];
2086 					dstaddr = &sk->addr[0];
2087 					srcport = sk->port[1];
2088 					dstport = sk->port[0];
2089 				} else {
2090 					srcaddr = &sk->addr[0];
2091 					dstaddr = &sk->addr[1];
2092 					srcport = sk->port[0];
2093 					dstport = sk->port[1];
2094 				}
2095 
2096 				if ((!psk->psk_af || sk->af == psk->psk_af)
2097 				    && (!psk->psk_proto || psk->psk_proto ==
2098 				    sk->proto) &&
2099 				    PF_MATCHA(psk->psk_src.neg,
2100 				    &psk->psk_src.addr.v.a.addr,
2101 				    &psk->psk_src.addr.v.a.mask,
2102 				    srcaddr, sk->af) &&
2103 				    PF_MATCHA(psk->psk_dst.neg,
2104 				    &psk->psk_dst.addr.v.a.addr,
2105 				    &psk->psk_dst.addr.v.a.mask,
2106 				    dstaddr, sk->af) &&
2107 				    (psk->psk_src.port_op == 0 ||
2108 				    pf_match_port(psk->psk_src.port_op,
2109 				    psk->psk_src.port[0], psk->psk_src.port[1],
2110 				    srcport)) &&
2111 				    (psk->psk_dst.port_op == 0 ||
2112 				    pf_match_port(psk->psk_dst.port_op,
2113 				    psk->psk_dst.port[0], psk->psk_dst.port[1],
2114 				    dstport)) &&
2115 				    (!psk->psk_label[0] ||
2116 				    (s->rule.ptr->label[0] &&
2117 				    !strcmp(psk->psk_label,
2118 				    s->rule.ptr->label))) &&
2119 				    (!psk->psk_ifname[0] ||
2120 				    !strcmp(psk->psk_ifname,
2121 				    s->kif->pfik_name))) {
2122 					pf_unlink_state(s, PF_ENTER_LOCKED);
2123 					killed++;
2124 					goto relock_DIOCKILLSTATES;
2125 				}
2126 			}
2127 			PF_HASHROW_UNLOCK(ih);
2128 		}
2129 		psk->psk_killed = killed;
2130 		break;
2131 	}
2132 
2133 	case DIOCADDSTATE: {
2134 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
2135 		struct pfsync_state	*sp = &ps->state;
2136 
2137 		if (sp->timeout >= PFTM_MAX) {
2138 			error = EINVAL;
2139 			break;
2140 		}
2141 		if (V_pfsync_state_import_ptr != NULL) {
2142 			PF_RULES_RLOCK();
2143 			error = V_pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
2144 			PF_RULES_RUNLOCK();
2145 		} else
2146 			error = EOPNOTSUPP;
2147 		break;
2148 	}
2149 
2150 	case DIOCGETSTATE: {
2151 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
2152 		struct pf_state		*s;
2153 
2154 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
2155 		if (s == NULL) {
2156 			error = ENOENT;
2157 			break;
2158 		}
2159 
2160 		pfsync_state_export(&ps->state, s);
2161 		PF_STATE_UNLOCK(s);
2162 		break;
2163 	}
2164 
2165 	case DIOCGETSTATES: {
2166 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
2167 		struct pf_state		*s;
2168 		struct pfsync_state	*pstore, *p;
2169 		int i, nr;
2170 
2171 		if (ps->ps_len <= 0) {
2172 			nr = uma_zone_get_cur(V_pf_state_z);
2173 			ps->ps_len = sizeof(struct pfsync_state) * nr;
2174 			break;
2175 		}
2176 
2177 		p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK | M_ZERO);
2178 		nr = 0;
2179 
2180 		for (i = 0; i <= pf_hashmask; i++) {
2181 			struct pf_idhash *ih = &V_pf_idhash[i];
2182 
2183 			PF_HASHROW_LOCK(ih);
2184 			LIST_FOREACH(s, &ih->states, entry) {
2185 
2186 				if (s->timeout == PFTM_UNLINKED)
2187 					continue;
2188 
2189 				if ((nr+1) * sizeof(*p) > ps->ps_len) {
2190 					PF_HASHROW_UNLOCK(ih);
2191 					goto DIOCGETSTATES_full;
2192 				}
2193 				pfsync_state_export(p, s);
2194 				p++;
2195 				nr++;
2196 			}
2197 			PF_HASHROW_UNLOCK(ih);
2198 		}
2199 DIOCGETSTATES_full:
2200 		error = copyout(pstore, ps->ps_states,
2201 		    sizeof(struct pfsync_state) * nr);
2202 		if (error) {
2203 			free(pstore, M_TEMP);
2204 			break;
2205 		}
2206 		ps->ps_len = sizeof(struct pfsync_state) * nr;
2207 		free(pstore, M_TEMP);
2208 
2209 		break;
2210 	}
2211 
2212 	case DIOCGETSTATUS: {
2213 		struct pf_status *s = (struct pf_status *)addr;
2214 
2215 		PF_RULES_RLOCK();
2216 		s->running = V_pf_status.running;
2217 		s->since   = V_pf_status.since;
2218 		s->debug   = V_pf_status.debug;
2219 		s->hostid  = V_pf_status.hostid;
2220 		s->states  = V_pf_status.states;
2221 		s->src_nodes = V_pf_status.src_nodes;
2222 
2223 		for (int i = 0; i < PFRES_MAX; i++)
2224 			s->counters[i] =
2225 			    counter_u64_fetch(V_pf_status.counters[i]);
2226 		for (int i = 0; i < LCNT_MAX; i++)
2227 			s->lcounters[i] =
2228 			    counter_u64_fetch(V_pf_status.lcounters[i]);
2229 		for (int i = 0; i < FCNT_MAX; i++)
2230 			s->fcounters[i] =
2231 			    counter_u64_fetch(V_pf_status.fcounters[i]);
2232 		for (int i = 0; i < SCNT_MAX; i++)
2233 			s->scounters[i] =
2234 			    counter_u64_fetch(V_pf_status.scounters[i]);
2235 
2236 		bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ);
2237 		bcopy(V_pf_status.pf_chksum, s->pf_chksum,
2238 		    PF_MD5_DIGEST_LENGTH);
2239 
2240 		pfi_update_status(s->ifname, s);
2241 		PF_RULES_RUNLOCK();
2242 		break;
2243 	}
2244 
2245 	case DIOCSETSTATUSIF: {
2246 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
2247 
2248 		if (pi->ifname[0] == 0) {
2249 			bzero(V_pf_status.ifname, IFNAMSIZ);
2250 			break;
2251 		}
2252 		PF_RULES_WLOCK();
2253 		strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
2254 		PF_RULES_WUNLOCK();
2255 		break;
2256 	}
2257 
2258 	case DIOCCLRSTATUS: {
2259 		PF_RULES_WLOCK();
2260 		for (int i = 0; i < PFRES_MAX; i++)
2261 			counter_u64_zero(V_pf_status.counters[i]);
2262 		for (int i = 0; i < FCNT_MAX; i++)
2263 			counter_u64_zero(V_pf_status.fcounters[i]);
2264 		for (int i = 0; i < SCNT_MAX; i++)
2265 			counter_u64_zero(V_pf_status.scounters[i]);
2266 		for (int i = 0; i < LCNT_MAX; i++)
2267 			counter_u64_zero(V_pf_status.lcounters[i]);
2268 		V_pf_status.since = time_second;
2269 		if (*V_pf_status.ifname)
2270 			pfi_update_status(V_pf_status.ifname, NULL);
2271 		PF_RULES_WUNLOCK();
2272 		break;
2273 	}
2274 
2275 	case DIOCNATLOOK: {
2276 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
2277 		struct pf_state_key	*sk;
2278 		struct pf_state		*state;
2279 		struct pf_state_key_cmp	 key;
2280 		int			 m = 0, direction = pnl->direction;
2281 		int			 sidx, didx;
2282 
2283 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
2284 		sidx = (direction == PF_IN) ? 1 : 0;
2285 		didx = (direction == PF_IN) ? 0 : 1;
2286 
2287 		if (!pnl->proto ||
2288 		    PF_AZERO(&pnl->saddr, pnl->af) ||
2289 		    PF_AZERO(&pnl->daddr, pnl->af) ||
2290 		    ((pnl->proto == IPPROTO_TCP ||
2291 		    pnl->proto == IPPROTO_UDP) &&
2292 		    (!pnl->dport || !pnl->sport)))
2293 			error = EINVAL;
2294 		else {
2295 			bzero(&key, sizeof(key));
2296 			key.af = pnl->af;
2297 			key.proto = pnl->proto;
2298 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
2299 			key.port[sidx] = pnl->sport;
2300 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
2301 			key.port[didx] = pnl->dport;
2302 
2303 			state = pf_find_state_all(&key, direction, &m);
2304 
2305 			if (m > 1)
2306 				error = E2BIG;	/* more than one state */
2307 			else if (state != NULL) {
2308 				/* XXXGL: not locked read */
2309 				sk = state->key[sidx];
2310 				PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
2311 				pnl->rsport = sk->port[sidx];
2312 				PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
2313 				pnl->rdport = sk->port[didx];
2314 			} else
2315 				error = ENOENT;
2316 		}
2317 		break;
2318 	}
2319 
2320 	case DIOCSETTIMEOUT: {
2321 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2322 		int		 old;
2323 
2324 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
2325 		    pt->seconds < 0) {
2326 			error = EINVAL;
2327 			break;
2328 		}
2329 		PF_RULES_WLOCK();
2330 		old = V_pf_default_rule.timeout[pt->timeout];
2331 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
2332 			pt->seconds = 1;
2333 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
2334 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
2335 			wakeup(pf_purge_thread);
2336 		pt->seconds = old;
2337 		PF_RULES_WUNLOCK();
2338 		break;
2339 	}
2340 
2341 	case DIOCGETTIMEOUT: {
2342 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2343 
2344 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
2345 			error = EINVAL;
2346 			break;
2347 		}
2348 		PF_RULES_RLOCK();
2349 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
2350 		PF_RULES_RUNLOCK();
2351 		break;
2352 	}
2353 
2354 	case DIOCGETLIMIT: {
2355 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2356 
2357 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
2358 			error = EINVAL;
2359 			break;
2360 		}
2361 		PF_RULES_RLOCK();
2362 		pl->limit = V_pf_limits[pl->index].limit;
2363 		PF_RULES_RUNLOCK();
2364 		break;
2365 	}
2366 
2367 	case DIOCSETLIMIT: {
2368 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2369 		int			 old_limit;
2370 
2371 		PF_RULES_WLOCK();
2372 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
2373 		    V_pf_limits[pl->index].zone == NULL) {
2374 			PF_RULES_WUNLOCK();
2375 			error = EINVAL;
2376 			break;
2377 		}
2378 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
2379 		old_limit = V_pf_limits[pl->index].limit;
2380 		V_pf_limits[pl->index].limit = pl->limit;
2381 		pl->limit = old_limit;
2382 		PF_RULES_WUNLOCK();
2383 		break;
2384 	}
2385 
2386 	case DIOCSETDEBUG: {
2387 		u_int32_t	*level = (u_int32_t *)addr;
2388 
2389 		PF_RULES_WLOCK();
2390 		V_pf_status.debug = *level;
2391 		PF_RULES_WUNLOCK();
2392 		break;
2393 	}
2394 
2395 	case DIOCCLRRULECTRS: {
2396 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
2397 		struct pf_ruleset	*ruleset = &pf_main_ruleset;
2398 		struct pf_rule		*rule;
2399 
2400 		PF_RULES_WLOCK();
2401 		TAILQ_FOREACH(rule,
2402 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
2403 			rule->evaluations = 0;
2404 			rule->packets[0] = rule->packets[1] = 0;
2405 			rule->bytes[0] = rule->bytes[1] = 0;
2406 		}
2407 		PF_RULES_WUNLOCK();
2408 		break;
2409 	}
2410 
2411 	case DIOCGIFSPEEDV0:
2412 	case DIOCGIFSPEEDV1: {
2413 		struct pf_ifspeed_v1	*psp = (struct pf_ifspeed_v1 *)addr;
2414 		struct pf_ifspeed_v1	ps;
2415 		struct ifnet		*ifp;
2416 
2417 		if (psp->ifname[0] != 0) {
2418 			/* Can we completely trust user-land? */
2419 			strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
2420 			ifp = ifunit(ps.ifname);
2421 			if (ifp != NULL) {
2422 				psp->baudrate32 =
2423 				    (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
2424 				if (cmd == DIOCGIFSPEEDV1)
2425 					psp->baudrate = ifp->if_baudrate;
2426 			} else
2427 				error = EINVAL;
2428 		} else
2429 			error = EINVAL;
2430 		break;
2431 	}
2432 
2433 #ifdef ALTQ
2434 	case DIOCSTARTALTQ: {
2435 		struct pf_altq		*altq;
2436 
2437 		PF_RULES_WLOCK();
2438 		/* enable all altq interfaces on active list */
2439 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
2440 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
2441 				error = pf_enable_altq(altq);
2442 				if (error != 0)
2443 					break;
2444 			}
2445 		}
2446 		if (error == 0)
2447 			V_pf_altq_running = 1;
2448 		PF_RULES_WUNLOCK();
2449 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
2450 		break;
2451 	}
2452 
2453 	case DIOCSTOPALTQ: {
2454 		struct pf_altq		*altq;
2455 
2456 		PF_RULES_WLOCK();
2457 		/* disable all altq interfaces on active list */
2458 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
2459 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
2460 				error = pf_disable_altq(altq);
2461 				if (error != 0)
2462 					break;
2463 			}
2464 		}
2465 		if (error == 0)
2466 			V_pf_altq_running = 0;
2467 		PF_RULES_WUNLOCK();
2468 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2469 		break;
2470 	}
2471 
2472 	case DIOCADDALTQV0:
2473 	case DIOCADDALTQV1: {
2474 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2475 		struct pf_altq		*altq, *a;
2476 		struct ifnet		*ifp;
2477 
2478 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
2479 		error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
2480 		if (error)
2481 			break;
2482 		altq->local_flags = 0;
2483 
2484 		PF_RULES_WLOCK();
2485 		if (pa->ticket != V_ticket_altqs_inactive) {
2486 			PF_RULES_WUNLOCK();
2487 			free(altq, M_PFALTQ);
2488 			error = EBUSY;
2489 			break;
2490 		}
2491 
2492 		/*
2493 		 * if this is for a queue, find the discipline and
2494 		 * copy the necessary fields
2495 		 */
2496 		if (altq->qname[0] != 0) {
2497 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2498 				PF_RULES_WUNLOCK();
2499 				error = EBUSY;
2500 				free(altq, M_PFALTQ);
2501 				break;
2502 			}
2503 			altq->altq_disc = NULL;
2504 			TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
2505 				if (strncmp(a->ifname, altq->ifname,
2506 				    IFNAMSIZ) == 0) {
2507 					altq->altq_disc = a->altq_disc;
2508 					break;
2509 				}
2510 			}
2511 		}
2512 
2513 		if ((ifp = ifunit(altq->ifname)) == NULL)
2514 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
2515 		else
2516 			error = altq_add(ifp, altq);
2517 
2518 		if (error) {
2519 			PF_RULES_WUNLOCK();
2520 			free(altq, M_PFALTQ);
2521 			break;
2522 		}
2523 
2524 		if (altq->qname[0] != 0)
2525 			TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
2526 		else
2527 			TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
2528 		/* version error check done on import above */
2529 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
2530 		PF_RULES_WUNLOCK();
2531 		break;
2532 	}
2533 
2534 	case DIOCGETALTQSV0:
2535 	case DIOCGETALTQSV1: {
2536 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2537 		struct pf_altq		*altq;
2538 
2539 		PF_RULES_RLOCK();
2540 		pa->nr = 0;
2541 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
2542 			pa->nr++;
2543 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
2544 			pa->nr++;
2545 		pa->ticket = V_ticket_altqs_active;
2546 		PF_RULES_RUNLOCK();
2547 		break;
2548 	}
2549 
2550 	case DIOCGETALTQV0:
2551 	case DIOCGETALTQV1: {
2552 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2553 		struct pf_altq		*altq;
2554 
2555 		PF_RULES_RLOCK();
2556 		if (pa->ticket != V_ticket_altqs_active) {
2557 			PF_RULES_RUNLOCK();
2558 			error = EBUSY;
2559 			break;
2560 		}
2561 		altq = pf_altq_get_nth_active(pa->nr);
2562 		if (altq == NULL) {
2563 			PF_RULES_RUNLOCK();
2564 			error = EBUSY;
2565 			break;
2566 		}
2567 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
2568 		PF_RULES_RUNLOCK();
2569 		break;
2570 	}
2571 
2572 	case DIOCCHANGEALTQV0:
2573 	case DIOCCHANGEALTQV1:
2574 		/* CHANGEALTQ not supported yet! */
2575 		error = ENODEV;
2576 		break;
2577 
2578 	case DIOCGETQSTATSV0:
2579 	case DIOCGETQSTATSV1: {
2580 		struct pfioc_qstats_v1	*pq = (struct pfioc_qstats_v1 *)addr;
2581 		struct pf_altq		*altq;
2582 		int			 nbytes;
2583 		u_int32_t		 version;
2584 
2585 		PF_RULES_RLOCK();
2586 		if (pq->ticket != V_ticket_altqs_active) {
2587 			PF_RULES_RUNLOCK();
2588 			error = EBUSY;
2589 			break;
2590 		}
2591 		nbytes = pq->nbytes;
2592 		altq = pf_altq_get_nth_active(pq->nr);
2593 		if (altq == NULL) {
2594 			PF_RULES_RUNLOCK();
2595 			error = EBUSY;
2596 			break;
2597 		}
2598 
2599 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
2600 			PF_RULES_RUNLOCK();
2601 			error = ENXIO;
2602 			break;
2603 		}
2604 		PF_RULES_RUNLOCK();
2605 		if (cmd == DIOCGETQSTATSV0)
2606 			version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
2607 		else
2608 			version = pq->version;
2609 		error = altq_getqstats(altq, pq->buf, &nbytes, version);
2610 		if (error == 0) {
2611 			pq->scheduler = altq->scheduler;
2612 			pq->nbytes = nbytes;
2613 		}
2614 		break;
2615 	}
2616 #endif /* ALTQ */
2617 
2618 	case DIOCBEGINADDRS: {
2619 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2620 
2621 		PF_RULES_WLOCK();
2622 		pf_empty_pool(&V_pf_pabuf);
2623 		pp->ticket = ++V_ticket_pabuf;
2624 		PF_RULES_WUNLOCK();
2625 		break;
2626 	}
2627 
2628 	case DIOCADDADDR: {
2629 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2630 		struct pf_pooladdr	*pa;
2631 		struct pfi_kif		*kif = NULL;
2632 
2633 #ifndef INET
2634 		if (pp->af == AF_INET) {
2635 			error = EAFNOSUPPORT;
2636 			break;
2637 		}
2638 #endif /* INET */
2639 #ifndef INET6
2640 		if (pp->af == AF_INET6) {
2641 			error = EAFNOSUPPORT;
2642 			break;
2643 		}
2644 #endif /* INET6 */
2645 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2646 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2647 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2648 			error = EINVAL;
2649 			break;
2650 		}
2651 		if (pp->addr.addr.p.dyn != NULL) {
2652 			error = EINVAL;
2653 			break;
2654 		}
2655 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2656 		bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2657 		if (pa->ifname[0])
2658 			kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2659 		PF_RULES_WLOCK();
2660 		if (pp->ticket != V_ticket_pabuf) {
2661 			PF_RULES_WUNLOCK();
2662 			if (pa->ifname[0])
2663 				free(kif, PFI_MTYPE);
2664 			free(pa, M_PFRULE);
2665 			error = EBUSY;
2666 			break;
2667 		}
2668 		if (pa->ifname[0]) {
2669 			pa->kif = pfi_kif_attach(kif, pa->ifname);
2670 			pfi_kif_ref(pa->kif);
2671 		} else
2672 			pa->kif = NULL;
2673 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2674 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2675 			if (pa->ifname[0])
2676 				pfi_kif_unref(pa->kif);
2677 			PF_RULES_WUNLOCK();
2678 			free(pa, M_PFRULE);
2679 			break;
2680 		}
2681 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2682 		PF_RULES_WUNLOCK();
2683 		break;
2684 	}
2685 
2686 	case DIOCGETADDRS: {
2687 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2688 		struct pf_pool		*pool;
2689 		struct pf_pooladdr	*pa;
2690 
2691 		PF_RULES_RLOCK();
2692 		pp->nr = 0;
2693 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2694 		    pp->r_num, 0, 1, 0);
2695 		if (pool == NULL) {
2696 			PF_RULES_RUNLOCK();
2697 			error = EBUSY;
2698 			break;
2699 		}
2700 		TAILQ_FOREACH(pa, &pool->list, entries)
2701 			pp->nr++;
2702 		PF_RULES_RUNLOCK();
2703 		break;
2704 	}
2705 
2706 	case DIOCGETADDR: {
2707 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2708 		struct pf_pool		*pool;
2709 		struct pf_pooladdr	*pa;
2710 		u_int32_t		 nr = 0;
2711 
2712 		PF_RULES_RLOCK();
2713 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2714 		    pp->r_num, 0, 1, 1);
2715 		if (pool == NULL) {
2716 			PF_RULES_RUNLOCK();
2717 			error = EBUSY;
2718 			break;
2719 		}
2720 		pa = TAILQ_FIRST(&pool->list);
2721 		while ((pa != NULL) && (nr < pp->nr)) {
2722 			pa = TAILQ_NEXT(pa, entries);
2723 			nr++;
2724 		}
2725 		if (pa == NULL) {
2726 			PF_RULES_RUNLOCK();
2727 			error = EBUSY;
2728 			break;
2729 		}
2730 		bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2731 		pf_addr_copyout(&pp->addr.addr);
2732 		PF_RULES_RUNLOCK();
2733 		break;
2734 	}
2735 
2736 	case DIOCCHANGEADDR: {
2737 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
2738 		struct pf_pool		*pool;
2739 		struct pf_pooladdr	*oldpa = NULL, *newpa = NULL;
2740 		struct pf_ruleset	*ruleset;
2741 		struct pfi_kif		*kif = NULL;
2742 
2743 		if (pca->action < PF_CHANGE_ADD_HEAD ||
2744 		    pca->action > PF_CHANGE_REMOVE) {
2745 			error = EINVAL;
2746 			break;
2747 		}
2748 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2749 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2750 		    pca->addr.addr.type != PF_ADDR_TABLE) {
2751 			error = EINVAL;
2752 			break;
2753 		}
2754 		if (pca->addr.addr.p.dyn != NULL) {
2755 			error = EINVAL;
2756 			break;
2757 		}
2758 
2759 		if (pca->action != PF_CHANGE_REMOVE) {
2760 #ifndef INET
2761 			if (pca->af == AF_INET) {
2762 				error = EAFNOSUPPORT;
2763 				break;
2764 			}
2765 #endif /* INET */
2766 #ifndef INET6
2767 			if (pca->af == AF_INET6) {
2768 				error = EAFNOSUPPORT;
2769 				break;
2770 			}
2771 #endif /* INET6 */
2772 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
2773 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2774 			if (newpa->ifname[0])
2775 				kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2776 			newpa->kif = NULL;
2777 		}
2778 
2779 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGEADDR_error; }
2780 		PF_RULES_WLOCK();
2781 		ruleset = pf_find_ruleset(pca->anchor);
2782 		if (ruleset == NULL)
2783 			ERROUT(EBUSY);
2784 
2785 		pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2786 		    pca->r_num, pca->r_last, 1, 1);
2787 		if (pool == NULL)
2788 			ERROUT(EBUSY);
2789 
2790 		if (pca->action != PF_CHANGE_REMOVE) {
2791 			if (newpa->ifname[0]) {
2792 				newpa->kif = pfi_kif_attach(kif, newpa->ifname);
2793 				pfi_kif_ref(newpa->kif);
2794 				kif = NULL;
2795 			}
2796 
2797 			switch (newpa->addr.type) {
2798 			case PF_ADDR_DYNIFTL:
2799 				error = pfi_dynaddr_setup(&newpa->addr,
2800 				    pca->af);
2801 				break;
2802 			case PF_ADDR_TABLE:
2803 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
2804 				    newpa->addr.v.tblname);
2805 				if (newpa->addr.p.tbl == NULL)
2806 					error = ENOMEM;
2807 				break;
2808 			}
2809 			if (error)
2810 				goto DIOCCHANGEADDR_error;
2811 		}
2812 
2813 		switch (pca->action) {
2814 		case PF_CHANGE_ADD_HEAD:
2815 			oldpa = TAILQ_FIRST(&pool->list);
2816 			break;
2817 		case PF_CHANGE_ADD_TAIL:
2818 			oldpa = TAILQ_LAST(&pool->list, pf_palist);
2819 			break;
2820 		default:
2821 			oldpa = TAILQ_FIRST(&pool->list);
2822 			for (int i = 0; oldpa && i < pca->nr; i++)
2823 				oldpa = TAILQ_NEXT(oldpa, entries);
2824 
2825 			if (oldpa == NULL)
2826 				ERROUT(EINVAL);
2827 		}
2828 
2829 		if (pca->action == PF_CHANGE_REMOVE) {
2830 			TAILQ_REMOVE(&pool->list, oldpa, entries);
2831 			switch (oldpa->addr.type) {
2832 			case PF_ADDR_DYNIFTL:
2833 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
2834 				break;
2835 			case PF_ADDR_TABLE:
2836 				pfr_detach_table(oldpa->addr.p.tbl);
2837 				break;
2838 			}
2839 			if (oldpa->kif)
2840 				pfi_kif_unref(oldpa->kif);
2841 			free(oldpa, M_PFRULE);
2842 		} else {
2843 			if (oldpa == NULL)
2844 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2845 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
2846 			    pca->action == PF_CHANGE_ADD_BEFORE)
2847 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2848 			else
2849 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
2850 				    newpa, entries);
2851 		}
2852 
2853 		pool->cur = TAILQ_FIRST(&pool->list);
2854 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
2855 		PF_RULES_WUNLOCK();
2856 		break;
2857 
2858 #undef ERROUT
2859 DIOCCHANGEADDR_error:
2860 		if (newpa != NULL) {
2861 			if (newpa->kif)
2862 				pfi_kif_unref(newpa->kif);
2863 			free(newpa, M_PFRULE);
2864 		}
2865 		PF_RULES_WUNLOCK();
2866 		if (kif != NULL)
2867 			free(kif, PFI_MTYPE);
2868 		break;
2869 	}
2870 
2871 	case DIOCGETRULESETS: {
2872 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2873 		struct pf_ruleset	*ruleset;
2874 		struct pf_anchor	*anchor;
2875 
2876 		PF_RULES_RLOCK();
2877 		pr->path[sizeof(pr->path) - 1] = 0;
2878 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2879 			PF_RULES_RUNLOCK();
2880 			error = ENOENT;
2881 			break;
2882 		}
2883 		pr->nr = 0;
2884 		if (ruleset->anchor == NULL) {
2885 			/* XXX kludge for pf_main_ruleset */
2886 			RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2887 				if (anchor->parent == NULL)
2888 					pr->nr++;
2889 		} else {
2890 			RB_FOREACH(anchor, pf_anchor_node,
2891 			    &ruleset->anchor->children)
2892 				pr->nr++;
2893 		}
2894 		PF_RULES_RUNLOCK();
2895 		break;
2896 	}
2897 
2898 	case DIOCGETRULESET: {
2899 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2900 		struct pf_ruleset	*ruleset;
2901 		struct pf_anchor	*anchor;
2902 		u_int32_t		 nr = 0;
2903 
2904 		PF_RULES_RLOCK();
2905 		pr->path[sizeof(pr->path) - 1] = 0;
2906 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2907 			PF_RULES_RUNLOCK();
2908 			error = ENOENT;
2909 			break;
2910 		}
2911 		pr->name[0] = 0;
2912 		if (ruleset->anchor == NULL) {
2913 			/* XXX kludge for pf_main_ruleset */
2914 			RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2915 				if (anchor->parent == NULL && nr++ == pr->nr) {
2916 					strlcpy(pr->name, anchor->name,
2917 					    sizeof(pr->name));
2918 					break;
2919 				}
2920 		} else {
2921 			RB_FOREACH(anchor, pf_anchor_node,
2922 			    &ruleset->anchor->children)
2923 				if (nr++ == pr->nr) {
2924 					strlcpy(pr->name, anchor->name,
2925 					    sizeof(pr->name));
2926 					break;
2927 				}
2928 		}
2929 		if (!pr->name[0])
2930 			error = EBUSY;
2931 		PF_RULES_RUNLOCK();
2932 		break;
2933 	}
2934 
2935 	case DIOCRCLRTABLES: {
2936 		struct pfioc_table *io = (struct pfioc_table *)addr;
2937 
2938 		if (io->pfrio_esize != 0) {
2939 			error = ENODEV;
2940 			break;
2941 		}
2942 		PF_RULES_WLOCK();
2943 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2944 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2945 		PF_RULES_WUNLOCK();
2946 		break;
2947 	}
2948 
2949 	case DIOCRADDTABLES: {
2950 		struct pfioc_table *io = (struct pfioc_table *)addr;
2951 		struct pfr_table *pfrts;
2952 		size_t totlen;
2953 
2954 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2955 			error = ENODEV;
2956 			break;
2957 		}
2958 
2959 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
2960 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
2961 			error = ENOMEM;
2962 			break;
2963 		}
2964 
2965 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2966 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
2967 		    M_TEMP, M_WAITOK);
2968 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2969 		if (error) {
2970 			free(pfrts, M_TEMP);
2971 			break;
2972 		}
2973 		PF_RULES_WLOCK();
2974 		error = pfr_add_tables(pfrts, io->pfrio_size,
2975 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2976 		PF_RULES_WUNLOCK();
2977 		free(pfrts, M_TEMP);
2978 		break;
2979 	}
2980 
2981 	case DIOCRDELTABLES: {
2982 		struct pfioc_table *io = (struct pfioc_table *)addr;
2983 		struct pfr_table *pfrts;
2984 		size_t totlen;
2985 
2986 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2987 			error = ENODEV;
2988 			break;
2989 		}
2990 
2991 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
2992 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
2993 			error = ENOMEM;
2994 			break;
2995 		}
2996 
2997 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2998 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
2999 		    M_TEMP, M_WAITOK);
3000 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3001 		if (error) {
3002 			free(pfrts, M_TEMP);
3003 			break;
3004 		}
3005 		PF_RULES_WLOCK();
3006 		error = pfr_del_tables(pfrts, io->pfrio_size,
3007 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3008 		PF_RULES_WUNLOCK();
3009 		free(pfrts, M_TEMP);
3010 		break;
3011 	}
3012 
3013 	case DIOCRGETTABLES: {
3014 		struct pfioc_table *io = (struct pfioc_table *)addr;
3015 		struct pfr_table *pfrts;
3016 		size_t totlen;
3017 		int n;
3018 
3019 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3020 			error = ENODEV;
3021 			break;
3022 		}
3023 		PF_RULES_RLOCK();
3024 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3025 		if (n < 0) {
3026 			PF_RULES_RUNLOCK();
3027 			error = EINVAL;
3028 			break;
3029 		}
3030 		io->pfrio_size = min(io->pfrio_size, n);
3031 
3032 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3033 
3034 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3035 		    M_TEMP, M_NOWAIT);
3036 		if (pfrts == NULL) {
3037 			error = ENOMEM;
3038 			PF_RULES_RUNLOCK();
3039 			break;
3040 		}
3041 		error = pfr_get_tables(&io->pfrio_table, pfrts,
3042 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3043 		PF_RULES_RUNLOCK();
3044 		if (error == 0)
3045 			error = copyout(pfrts, io->pfrio_buffer, totlen);
3046 		free(pfrts, M_TEMP);
3047 		break;
3048 	}
3049 
3050 	case DIOCRGETTSTATS: {
3051 		struct pfioc_table *io = (struct pfioc_table *)addr;
3052 		struct pfr_tstats *pfrtstats;
3053 		size_t totlen;
3054 		int n;
3055 
3056 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
3057 			error = ENODEV;
3058 			break;
3059 		}
3060 		PF_RULES_WLOCK();
3061 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3062 		if (n < 0) {
3063 			PF_RULES_WUNLOCK();
3064 			error = EINVAL;
3065 			break;
3066 		}
3067 		io->pfrio_size = min(io->pfrio_size, n);
3068 
3069 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
3070 		pfrtstats = mallocarray(io->pfrio_size,
3071 		    sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT);
3072 		if (pfrtstats == NULL) {
3073 			error = ENOMEM;
3074 			PF_RULES_WUNLOCK();
3075 			break;
3076 		}
3077 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
3078 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3079 		PF_RULES_WUNLOCK();
3080 		if (error == 0)
3081 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
3082 		free(pfrtstats, M_TEMP);
3083 		break;
3084 	}
3085 
3086 	case DIOCRCLRTSTATS: {
3087 		struct pfioc_table *io = (struct pfioc_table *)addr;
3088 		struct pfr_table *pfrts;
3089 		size_t totlen;
3090 		int n;
3091 
3092 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3093 			error = ENODEV;
3094 			break;
3095 		}
3096 
3097 		PF_RULES_WLOCK();
3098 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3099 		if (n < 0) {
3100 			PF_RULES_WUNLOCK();
3101 			error = EINVAL;
3102 			break;
3103 		}
3104 		io->pfrio_size = min(io->pfrio_size, n);
3105 
3106 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3107 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3108 		    M_TEMP, M_NOWAIT);
3109 		if (pfrts == NULL) {
3110 			error = ENOMEM;
3111 			PF_RULES_WUNLOCK();
3112 			break;
3113 		}
3114 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3115 		if (error) {
3116 			free(pfrts, M_TEMP);
3117 			PF_RULES_WUNLOCK();
3118 			break;
3119 		}
3120 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
3121 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3122 		PF_RULES_WUNLOCK();
3123 		free(pfrts, M_TEMP);
3124 		break;
3125 	}
3126 
3127 	case DIOCRSETTFLAGS: {
3128 		struct pfioc_table *io = (struct pfioc_table *)addr;
3129 		struct pfr_table *pfrts;
3130 		size_t totlen;
3131 		int n;
3132 
3133 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3134 			error = ENODEV;
3135 			break;
3136 		}
3137 
3138 		PF_RULES_RLOCK();
3139 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3140 		if (n < 0) {
3141 			PF_RULES_RUNLOCK();
3142 			error = EINVAL;
3143 			break;
3144 		}
3145 
3146 		io->pfrio_size = min(io->pfrio_size, n);
3147 		PF_RULES_RUNLOCK();
3148 
3149 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3150 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3151 		    M_TEMP, M_WAITOK);
3152 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3153 		if (error) {
3154 			free(pfrts, M_TEMP);
3155 			break;
3156 		}
3157 		PF_RULES_WLOCK();
3158 		error = pfr_set_tflags(pfrts, io->pfrio_size,
3159 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
3160 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3161 		PF_RULES_WUNLOCK();
3162 		free(pfrts, M_TEMP);
3163 		break;
3164 	}
3165 
3166 	case DIOCRCLRADDRS: {
3167 		struct pfioc_table *io = (struct pfioc_table *)addr;
3168 
3169 		if (io->pfrio_esize != 0) {
3170 			error = ENODEV;
3171 			break;
3172 		}
3173 		PF_RULES_WLOCK();
3174 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
3175 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
3176 		PF_RULES_WUNLOCK();
3177 		break;
3178 	}
3179 
3180 	case DIOCRADDADDRS: {
3181 		struct pfioc_table *io = (struct pfioc_table *)addr;
3182 		struct pfr_addr *pfras;
3183 		size_t totlen;
3184 
3185 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3186 			error = ENODEV;
3187 			break;
3188 		}
3189 		if (io->pfrio_size < 0 ||
3190 		    io->pfrio_size > pf_ioctl_maxcount ||
3191 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3192 			error = EINVAL;
3193 			break;
3194 		}
3195 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3196 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3197 		    M_TEMP, M_NOWAIT);
3198 		if (! pfras) {
3199 			error = ENOMEM;
3200 			break;
3201 		}
3202 		error = copyin(io->pfrio_buffer, pfras, totlen);
3203 		if (error) {
3204 			free(pfras, M_TEMP);
3205 			break;
3206 		}
3207 		PF_RULES_WLOCK();
3208 		error = pfr_add_addrs(&io->pfrio_table, pfras,
3209 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
3210 		    PFR_FLAG_USERIOCTL);
3211 		PF_RULES_WUNLOCK();
3212 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3213 			error = copyout(pfras, io->pfrio_buffer, totlen);
3214 		free(pfras, M_TEMP);
3215 		break;
3216 	}
3217 
3218 	case DIOCRDELADDRS: {
3219 		struct pfioc_table *io = (struct pfioc_table *)addr;
3220 		struct pfr_addr *pfras;
3221 		size_t totlen;
3222 
3223 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3224 			error = ENODEV;
3225 			break;
3226 		}
3227 		if (io->pfrio_size < 0 ||
3228 		    io->pfrio_size > pf_ioctl_maxcount ||
3229 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3230 			error = EINVAL;
3231 			break;
3232 		}
3233 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3234 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3235 		    M_TEMP, M_NOWAIT);
3236 		if (! pfras) {
3237 			error = ENOMEM;
3238 			break;
3239 		}
3240 		error = copyin(io->pfrio_buffer, pfras, totlen);
3241 		if (error) {
3242 			free(pfras, M_TEMP);
3243 			break;
3244 		}
3245 		PF_RULES_WLOCK();
3246 		error = pfr_del_addrs(&io->pfrio_table, pfras,
3247 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
3248 		    PFR_FLAG_USERIOCTL);
3249 		PF_RULES_WUNLOCK();
3250 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3251 			error = copyout(pfras, io->pfrio_buffer, totlen);
3252 		free(pfras, M_TEMP);
3253 		break;
3254 	}
3255 
3256 	case DIOCRSETADDRS: {
3257 		struct pfioc_table *io = (struct pfioc_table *)addr;
3258 		struct pfr_addr *pfras;
3259 		size_t totlen, count;
3260 
3261 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3262 			error = ENODEV;
3263 			break;
3264 		}
3265 		if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
3266 			error = EINVAL;
3267 			break;
3268 		}
3269 		count = max(io->pfrio_size, io->pfrio_size2);
3270 		if (count > pf_ioctl_maxcount ||
3271 		    WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
3272 			error = EINVAL;
3273 			break;
3274 		}
3275 		totlen = count * sizeof(struct pfr_addr);
3276 		pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
3277 		    M_NOWAIT);
3278 		if (! pfras) {
3279 			error = ENOMEM;
3280 			break;
3281 		}
3282 		error = copyin(io->pfrio_buffer, pfras, totlen);
3283 		if (error) {
3284 			free(pfras, M_TEMP);
3285 			break;
3286 		}
3287 		PF_RULES_WLOCK();
3288 		error = pfr_set_addrs(&io->pfrio_table, pfras,
3289 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
3290 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
3291 		    PFR_FLAG_USERIOCTL, 0);
3292 		PF_RULES_WUNLOCK();
3293 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3294 			error = copyout(pfras, io->pfrio_buffer, totlen);
3295 		free(pfras, M_TEMP);
3296 		break;
3297 	}
3298 
3299 	case DIOCRGETADDRS: {
3300 		struct pfioc_table *io = (struct pfioc_table *)addr;
3301 		struct pfr_addr *pfras;
3302 		size_t totlen;
3303 
3304 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3305 			error = ENODEV;
3306 			break;
3307 		}
3308 		if (io->pfrio_size < 0 ||
3309 		    io->pfrio_size > pf_ioctl_maxcount ||
3310 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3311 			error = EINVAL;
3312 			break;
3313 		}
3314 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3315 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3316 		    M_TEMP, M_NOWAIT);
3317 		if (! pfras) {
3318 			error = ENOMEM;
3319 			break;
3320 		}
3321 		PF_RULES_RLOCK();
3322 		error = pfr_get_addrs(&io->pfrio_table, pfras,
3323 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3324 		PF_RULES_RUNLOCK();
3325 		if (error == 0)
3326 			error = copyout(pfras, io->pfrio_buffer, totlen);
3327 		free(pfras, M_TEMP);
3328 		break;
3329 	}
3330 
3331 	case DIOCRGETASTATS: {
3332 		struct pfioc_table *io = (struct pfioc_table *)addr;
3333 		struct pfr_astats *pfrastats;
3334 		size_t totlen;
3335 
3336 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
3337 			error = ENODEV;
3338 			break;
3339 		}
3340 		if (io->pfrio_size < 0 ||
3341 		    io->pfrio_size > pf_ioctl_maxcount ||
3342 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
3343 			error = EINVAL;
3344 			break;
3345 		}
3346 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
3347 		pfrastats = mallocarray(io->pfrio_size,
3348 		    sizeof(struct pfr_astats), M_TEMP, M_NOWAIT);
3349 		if (! pfrastats) {
3350 			error = ENOMEM;
3351 			break;
3352 		}
3353 		PF_RULES_RLOCK();
3354 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
3355 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3356 		PF_RULES_RUNLOCK();
3357 		if (error == 0)
3358 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
3359 		free(pfrastats, M_TEMP);
3360 		break;
3361 	}
3362 
3363 	case DIOCRCLRASTATS: {
3364 		struct pfioc_table *io = (struct pfioc_table *)addr;
3365 		struct pfr_addr *pfras;
3366 		size_t totlen;
3367 
3368 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3369 			error = ENODEV;
3370 			break;
3371 		}
3372 		if (io->pfrio_size < 0 ||
3373 		    io->pfrio_size > pf_ioctl_maxcount ||
3374 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3375 			error = EINVAL;
3376 			break;
3377 		}
3378 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3379 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3380 		    M_TEMP, M_NOWAIT);
3381 		if (! pfras) {
3382 			error = ENOMEM;
3383 			break;
3384 		}
3385 		error = copyin(io->pfrio_buffer, pfras, totlen);
3386 		if (error) {
3387 			free(pfras, M_TEMP);
3388 			break;
3389 		}
3390 		PF_RULES_WLOCK();
3391 		error = pfr_clr_astats(&io->pfrio_table, pfras,
3392 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
3393 		    PFR_FLAG_USERIOCTL);
3394 		PF_RULES_WUNLOCK();
3395 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3396 			error = copyout(pfras, io->pfrio_buffer, totlen);
3397 		free(pfras, M_TEMP);
3398 		break;
3399 	}
3400 
3401 	case DIOCRTSTADDRS: {
3402 		struct pfioc_table *io = (struct pfioc_table *)addr;
3403 		struct pfr_addr *pfras;
3404 		size_t totlen;
3405 
3406 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3407 			error = ENODEV;
3408 			break;
3409 		}
3410 		if (io->pfrio_size < 0 ||
3411 		    io->pfrio_size > pf_ioctl_maxcount ||
3412 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3413 			error = EINVAL;
3414 			break;
3415 		}
3416 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3417 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3418 		    M_TEMP, M_NOWAIT);
3419 		if (! pfras) {
3420 			error = ENOMEM;
3421 			break;
3422 		}
3423 		error = copyin(io->pfrio_buffer, pfras, totlen);
3424 		if (error) {
3425 			free(pfras, M_TEMP);
3426 			break;
3427 		}
3428 		PF_RULES_RLOCK();
3429 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
3430 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
3431 		    PFR_FLAG_USERIOCTL);
3432 		PF_RULES_RUNLOCK();
3433 		if (error == 0)
3434 			error = copyout(pfras, io->pfrio_buffer, totlen);
3435 		free(pfras, M_TEMP);
3436 		break;
3437 	}
3438 
3439 	case DIOCRINADEFINE: {
3440 		struct pfioc_table *io = (struct pfioc_table *)addr;
3441 		struct pfr_addr *pfras;
3442 		size_t totlen;
3443 
3444 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3445 			error = ENODEV;
3446 			break;
3447 		}
3448 		if (io->pfrio_size < 0 ||
3449 		    io->pfrio_size > pf_ioctl_maxcount ||
3450 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3451 			error = EINVAL;
3452 			break;
3453 		}
3454 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3455 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3456 		    M_TEMP, M_NOWAIT);
3457 		if (! pfras) {
3458 			error = ENOMEM;
3459 			break;
3460 		}
3461 		error = copyin(io->pfrio_buffer, pfras, totlen);
3462 		if (error) {
3463 			free(pfras, M_TEMP);
3464 			break;
3465 		}
3466 		PF_RULES_WLOCK();
3467 		error = pfr_ina_define(&io->pfrio_table, pfras,
3468 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
3469 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3470 		PF_RULES_WUNLOCK();
3471 		free(pfras, M_TEMP);
3472 		break;
3473 	}
3474 
3475 	case DIOCOSFPADD: {
3476 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
3477 		PF_RULES_WLOCK();
3478 		error = pf_osfp_add(io);
3479 		PF_RULES_WUNLOCK();
3480 		break;
3481 	}
3482 
3483 	case DIOCOSFPGET: {
3484 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
3485 		PF_RULES_RLOCK();
3486 		error = pf_osfp_get(io);
3487 		PF_RULES_RUNLOCK();
3488 		break;
3489 	}
3490 
3491 	case DIOCXBEGIN: {
3492 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3493 		struct pfioc_trans_e	*ioes, *ioe;
3494 		size_t			 totlen;
3495 		int			 i;
3496 
3497 		if (io->esize != sizeof(*ioe)) {
3498 			error = ENODEV;
3499 			break;
3500 		}
3501 		if (io->size < 0 ||
3502 		    io->size > pf_ioctl_maxcount ||
3503 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3504 			error = EINVAL;
3505 			break;
3506 		}
3507 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3508 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3509 		    M_TEMP, M_NOWAIT);
3510 		if (! ioes) {
3511 			error = ENOMEM;
3512 			break;
3513 		}
3514 		error = copyin(io->array, ioes, totlen);
3515 		if (error) {
3516 			free(ioes, M_TEMP);
3517 			break;
3518 		}
3519 		PF_RULES_WLOCK();
3520 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3521 			switch (ioe->rs_num) {
3522 #ifdef ALTQ
3523 			case PF_RULESET_ALTQ:
3524 				if (ioe->anchor[0]) {
3525 					PF_RULES_WUNLOCK();
3526 					free(ioes, M_TEMP);
3527 					error = EINVAL;
3528 					goto fail;
3529 				}
3530 				if ((error = pf_begin_altq(&ioe->ticket))) {
3531 					PF_RULES_WUNLOCK();
3532 					free(ioes, M_TEMP);
3533 					goto fail;
3534 				}
3535 				break;
3536 #endif /* ALTQ */
3537 			case PF_RULESET_TABLE:
3538 			    {
3539 				struct pfr_table table;
3540 
3541 				bzero(&table, sizeof(table));
3542 				strlcpy(table.pfrt_anchor, ioe->anchor,
3543 				    sizeof(table.pfrt_anchor));
3544 				if ((error = pfr_ina_begin(&table,
3545 				    &ioe->ticket, NULL, 0))) {
3546 					PF_RULES_WUNLOCK();
3547 					free(ioes, M_TEMP);
3548 					goto fail;
3549 				}
3550 				break;
3551 			    }
3552 			default:
3553 				if ((error = pf_begin_rules(&ioe->ticket,
3554 				    ioe->rs_num, ioe->anchor))) {
3555 					PF_RULES_WUNLOCK();
3556 					free(ioes, M_TEMP);
3557 					goto fail;
3558 				}
3559 				break;
3560 			}
3561 		}
3562 		PF_RULES_WUNLOCK();
3563 		error = copyout(ioes, io->array, totlen);
3564 		free(ioes, M_TEMP);
3565 		break;
3566 	}
3567 
3568 	case DIOCXROLLBACK: {
3569 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3570 		struct pfioc_trans_e	*ioe, *ioes;
3571 		size_t			 totlen;
3572 		int			 i;
3573 
3574 		if (io->esize != sizeof(*ioe)) {
3575 			error = ENODEV;
3576 			break;
3577 		}
3578 		if (io->size < 0 ||
3579 		    io->size > pf_ioctl_maxcount ||
3580 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3581 			error = EINVAL;
3582 			break;
3583 		}
3584 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3585 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3586 		    M_TEMP, M_NOWAIT);
3587 		if (! ioes) {
3588 			error = ENOMEM;
3589 			break;
3590 		}
3591 		error = copyin(io->array, ioes, totlen);
3592 		if (error) {
3593 			free(ioes, M_TEMP);
3594 			break;
3595 		}
3596 		PF_RULES_WLOCK();
3597 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3598 			switch (ioe->rs_num) {
3599 #ifdef ALTQ
3600 			case PF_RULESET_ALTQ:
3601 				if (ioe->anchor[0]) {
3602 					PF_RULES_WUNLOCK();
3603 					free(ioes, M_TEMP);
3604 					error = EINVAL;
3605 					goto fail;
3606 				}
3607 				if ((error = pf_rollback_altq(ioe->ticket))) {
3608 					PF_RULES_WUNLOCK();
3609 					free(ioes, M_TEMP);
3610 					goto fail; /* really bad */
3611 				}
3612 				break;
3613 #endif /* ALTQ */
3614 			case PF_RULESET_TABLE:
3615 			    {
3616 				struct pfr_table table;
3617 
3618 				bzero(&table, sizeof(table));
3619 				strlcpy(table.pfrt_anchor, ioe->anchor,
3620 				    sizeof(table.pfrt_anchor));
3621 				if ((error = pfr_ina_rollback(&table,
3622 				    ioe->ticket, NULL, 0))) {
3623 					PF_RULES_WUNLOCK();
3624 					free(ioes, M_TEMP);
3625 					goto fail; /* really bad */
3626 				}
3627 				break;
3628 			    }
3629 			default:
3630 				if ((error = pf_rollback_rules(ioe->ticket,
3631 				    ioe->rs_num, ioe->anchor))) {
3632 					PF_RULES_WUNLOCK();
3633 					free(ioes, M_TEMP);
3634 					goto fail; /* really bad */
3635 				}
3636 				break;
3637 			}
3638 		}
3639 		PF_RULES_WUNLOCK();
3640 		free(ioes, M_TEMP);
3641 		break;
3642 	}
3643 
3644 	case DIOCXCOMMIT: {
3645 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3646 		struct pfioc_trans_e	*ioe, *ioes;
3647 		struct pf_ruleset	*rs;
3648 		size_t			 totlen;
3649 		int			 i;
3650 
3651 		if (io->esize != sizeof(*ioe)) {
3652 			error = ENODEV;
3653 			break;
3654 		}
3655 
3656 		if (io->size < 0 ||
3657 		    io->size > pf_ioctl_maxcount ||
3658 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3659 			error = EINVAL;
3660 			break;
3661 		}
3662 
3663 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3664 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3665 		    M_TEMP, M_NOWAIT);
3666 		if (ioes == NULL) {
3667 			error = ENOMEM;
3668 			break;
3669 		}
3670 		error = copyin(io->array, ioes, totlen);
3671 		if (error) {
3672 			free(ioes, M_TEMP);
3673 			break;
3674 		}
3675 		PF_RULES_WLOCK();
3676 		/* First makes sure everything will succeed. */
3677 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3678 			switch (ioe->rs_num) {
3679 #ifdef ALTQ
3680 			case PF_RULESET_ALTQ:
3681 				if (ioe->anchor[0]) {
3682 					PF_RULES_WUNLOCK();
3683 					free(ioes, M_TEMP);
3684 					error = EINVAL;
3685 					goto fail;
3686 				}
3687 				if (!V_altqs_inactive_open || ioe->ticket !=
3688 				    V_ticket_altqs_inactive) {
3689 					PF_RULES_WUNLOCK();
3690 					free(ioes, M_TEMP);
3691 					error = EBUSY;
3692 					goto fail;
3693 				}
3694 				break;
3695 #endif /* ALTQ */
3696 			case PF_RULESET_TABLE:
3697 				rs = pf_find_ruleset(ioe->anchor);
3698 				if (rs == NULL || !rs->topen || ioe->ticket !=
3699 				    rs->tticket) {
3700 					PF_RULES_WUNLOCK();
3701 					free(ioes, M_TEMP);
3702 					error = EBUSY;
3703 					goto fail;
3704 				}
3705 				break;
3706 			default:
3707 				if (ioe->rs_num < 0 || ioe->rs_num >=
3708 				    PF_RULESET_MAX) {
3709 					PF_RULES_WUNLOCK();
3710 					free(ioes, M_TEMP);
3711 					error = EINVAL;
3712 					goto fail;
3713 				}
3714 				rs = pf_find_ruleset(ioe->anchor);
3715 				if (rs == NULL ||
3716 				    !rs->rules[ioe->rs_num].inactive.open ||
3717 				    rs->rules[ioe->rs_num].inactive.ticket !=
3718 				    ioe->ticket) {
3719 					PF_RULES_WUNLOCK();
3720 					free(ioes, M_TEMP);
3721 					error = EBUSY;
3722 					goto fail;
3723 				}
3724 				break;
3725 			}
3726 		}
3727 		/* Now do the commit - no errors should happen here. */
3728 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3729 			switch (ioe->rs_num) {
3730 #ifdef ALTQ
3731 			case PF_RULESET_ALTQ:
3732 				if ((error = pf_commit_altq(ioe->ticket))) {
3733 					PF_RULES_WUNLOCK();
3734 					free(ioes, M_TEMP);
3735 					goto fail; /* really bad */
3736 				}
3737 				break;
3738 #endif /* ALTQ */
3739 			case PF_RULESET_TABLE:
3740 			    {
3741 				struct pfr_table table;
3742 
3743 				bzero(&table, sizeof(table));
3744 				strlcpy(table.pfrt_anchor, ioe->anchor,
3745 				    sizeof(table.pfrt_anchor));
3746 				if ((error = pfr_ina_commit(&table,
3747 				    ioe->ticket, NULL, NULL, 0))) {
3748 					PF_RULES_WUNLOCK();
3749 					free(ioes, M_TEMP);
3750 					goto fail; /* really bad */
3751 				}
3752 				break;
3753 			    }
3754 			default:
3755 				if ((error = pf_commit_rules(ioe->ticket,
3756 				    ioe->rs_num, ioe->anchor))) {
3757 					PF_RULES_WUNLOCK();
3758 					free(ioes, M_TEMP);
3759 					goto fail; /* really bad */
3760 				}
3761 				break;
3762 			}
3763 		}
3764 		PF_RULES_WUNLOCK();
3765 		free(ioes, M_TEMP);
3766 		break;
3767 	}
3768 
3769 	case DIOCGETSRCNODES: {
3770 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
3771 		struct pf_srchash	*sh;
3772 		struct pf_src_node	*n, *p, *pstore;
3773 		uint32_t		 i, nr = 0;
3774 
3775 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3776 				i++, sh++) {
3777 			PF_HASHROW_LOCK(sh);
3778 			LIST_FOREACH(n, &sh->nodes, entry)
3779 				nr++;
3780 			PF_HASHROW_UNLOCK(sh);
3781 		}
3782 
3783 		psn->psn_len = min(psn->psn_len,
3784 		    sizeof(struct pf_src_node) * nr);
3785 
3786 		if (psn->psn_len == 0) {
3787 			psn->psn_len = sizeof(struct pf_src_node) * nr;
3788 			break;
3789 		}
3790 
3791 		nr = 0;
3792 
3793 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
3794 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3795 		    i++, sh++) {
3796 		    PF_HASHROW_LOCK(sh);
3797 		    LIST_FOREACH(n, &sh->nodes, entry) {
3798 			int	secs = time_uptime, diff;
3799 
3800 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
3801 				break;
3802 
3803 			bcopy(n, p, sizeof(struct pf_src_node));
3804 			if (n->rule.ptr != NULL)
3805 				p->rule.nr = n->rule.ptr->nr;
3806 			p->creation = secs - p->creation;
3807 			if (p->expire > secs)
3808 				p->expire -= secs;
3809 			else
3810 				p->expire = 0;
3811 
3812 			/* Adjust the connection rate estimate. */
3813 			diff = secs - n->conn_rate.last;
3814 			if (diff >= n->conn_rate.seconds)
3815 				p->conn_rate.count = 0;
3816 			else
3817 				p->conn_rate.count -=
3818 				    n->conn_rate.count * diff /
3819 				    n->conn_rate.seconds;
3820 			p++;
3821 			nr++;
3822 		    }
3823 		    PF_HASHROW_UNLOCK(sh);
3824 		}
3825 		error = copyout(pstore, psn->psn_src_nodes,
3826 		    sizeof(struct pf_src_node) * nr);
3827 		if (error) {
3828 			free(pstore, M_TEMP);
3829 			break;
3830 		}
3831 		psn->psn_len = sizeof(struct pf_src_node) * nr;
3832 		free(pstore, M_TEMP);
3833 		break;
3834 	}
3835 
3836 	case DIOCCLRSRCNODES: {
3837 
3838 		pf_clear_srcnodes(NULL);
3839 		pf_purge_expired_src_nodes();
3840 		break;
3841 	}
3842 
3843 	case DIOCKILLSRCNODES:
3844 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
3845 		break;
3846 
3847 	case DIOCSETHOSTID: {
3848 		u_int32_t	*hostid = (u_int32_t *)addr;
3849 
3850 		PF_RULES_WLOCK();
3851 		if (*hostid == 0)
3852 			V_pf_status.hostid = arc4random();
3853 		else
3854 			V_pf_status.hostid = *hostid;
3855 		PF_RULES_WUNLOCK();
3856 		break;
3857 	}
3858 
3859 	case DIOCOSFPFLUSH:
3860 		PF_RULES_WLOCK();
3861 		pf_osfp_flush();
3862 		PF_RULES_WUNLOCK();
3863 		break;
3864 
3865 	case DIOCIGETIFACES: {
3866 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3867 		struct pfi_kif *ifstore;
3868 		size_t bufsiz;
3869 
3870 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
3871 			error = ENODEV;
3872 			break;
3873 		}
3874 
3875 		if (io->pfiio_size < 0 ||
3876 		    io->pfiio_size > pf_ioctl_maxcount ||
3877 		    WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
3878 			error = EINVAL;
3879 			break;
3880 		}
3881 
3882 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
3883 		ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
3884 		    M_TEMP, M_NOWAIT);
3885 		if (ifstore == NULL) {
3886 			error = ENOMEM;
3887 			break;
3888 		}
3889 
3890 		PF_RULES_RLOCK();
3891 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
3892 		PF_RULES_RUNLOCK();
3893 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
3894 		free(ifstore, M_TEMP);
3895 		break;
3896 	}
3897 
3898 	case DIOCSETIFFLAG: {
3899 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3900 
3901 		PF_RULES_WLOCK();
3902 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3903 		PF_RULES_WUNLOCK();
3904 		break;
3905 	}
3906 
3907 	case DIOCCLRIFFLAG: {
3908 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3909 
3910 		PF_RULES_WLOCK();
3911 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3912 		PF_RULES_WUNLOCK();
3913 		break;
3914 	}
3915 
3916 	default:
3917 		error = ENODEV;
3918 		break;
3919 	}
3920 fail:
3921 	if (sx_xlocked(&pf_ioctl_lock))
3922 		sx_xunlock(&pf_ioctl_lock);
3923 	CURVNET_RESTORE();
3924 
3925 	return (error);
3926 }
3927 
3928 void
3929 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
3930 {
3931 	bzero(sp, sizeof(struct pfsync_state));
3932 
3933 	/* copy from state key */
3934 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
3935 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
3936 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
3937 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
3938 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
3939 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
3940 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
3941 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
3942 	sp->proto = st->key[PF_SK_WIRE]->proto;
3943 	sp->af = st->key[PF_SK_WIRE]->af;
3944 
3945 	/* copy from state */
3946 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
3947 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
3948 	sp->creation = htonl(time_uptime - st->creation);
3949 	sp->expire = pf_state_expires(st);
3950 	if (sp->expire <= time_uptime)
3951 		sp->expire = htonl(0);
3952 	else
3953 		sp->expire = htonl(sp->expire - time_uptime);
3954 
3955 	sp->direction = st->direction;
3956 	sp->log = st->log;
3957 	sp->timeout = st->timeout;
3958 	sp->state_flags = st->state_flags;
3959 	if (st->src_node)
3960 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
3961 	if (st->nat_src_node)
3962 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
3963 
3964 	sp->id = st->id;
3965 	sp->creatorid = st->creatorid;
3966 	pf_state_peer_hton(&st->src, &sp->src);
3967 	pf_state_peer_hton(&st->dst, &sp->dst);
3968 
3969 	if (st->rule.ptr == NULL)
3970 		sp->rule = htonl(-1);
3971 	else
3972 		sp->rule = htonl(st->rule.ptr->nr);
3973 	if (st->anchor.ptr == NULL)
3974 		sp->anchor = htonl(-1);
3975 	else
3976 		sp->anchor = htonl(st->anchor.ptr->nr);
3977 	if (st->nat_rule.ptr == NULL)
3978 		sp->nat_rule = htonl(-1);
3979 	else
3980 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
3981 
3982 	pf_state_counter_hton(st->packets[0], sp->packets[0]);
3983 	pf_state_counter_hton(st->packets[1], sp->packets[1]);
3984 	pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
3985 	pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
3986 
3987 }
3988 
3989 static void
3990 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
3991 {
3992 	struct pfr_ktable *kt;
3993 
3994 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
3995 
3996 	kt = aw->p.tbl;
3997 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
3998 		kt = kt->pfrkt_root;
3999 	aw->p.tbl = NULL;
4000 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
4001 		kt->pfrkt_cnt : -1;
4002 }
4003 
4004 /*
4005  * XXX - Check for version missmatch!!!
4006  */
4007 static void
4008 pf_clear_states(void)
4009 {
4010 	struct pf_state	*s;
4011 	u_int i;
4012 
4013 	for (i = 0; i <= pf_hashmask; i++) {
4014 		struct pf_idhash *ih = &V_pf_idhash[i];
4015 relock:
4016 		PF_HASHROW_LOCK(ih);
4017 		LIST_FOREACH(s, &ih->states, entry) {
4018 			s->timeout = PFTM_PURGE;
4019 			/* Don't send out individual delete messages. */
4020 			s->state_flags |= PFSTATE_NOSYNC;
4021 			pf_unlink_state(s, PF_ENTER_LOCKED);
4022 			goto relock;
4023 		}
4024 		PF_HASHROW_UNLOCK(ih);
4025 	}
4026 }
4027 
4028 static int
4029 pf_clear_tables(void)
4030 {
4031 	struct pfioc_table io;
4032 	int error;
4033 
4034 	bzero(&io, sizeof(io));
4035 
4036 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
4037 	    io.pfrio_flags);
4038 
4039 	return (error);
4040 }
4041 
4042 static void
4043 pf_clear_srcnodes(struct pf_src_node *n)
4044 {
4045 	struct pf_state *s;
4046 	int i;
4047 
4048 	for (i = 0; i <= pf_hashmask; i++) {
4049 		struct pf_idhash *ih = &V_pf_idhash[i];
4050 
4051 		PF_HASHROW_LOCK(ih);
4052 		LIST_FOREACH(s, &ih->states, entry) {
4053 			if (n == NULL || n == s->src_node)
4054 				s->src_node = NULL;
4055 			if (n == NULL || n == s->nat_src_node)
4056 				s->nat_src_node = NULL;
4057 		}
4058 		PF_HASHROW_UNLOCK(ih);
4059 	}
4060 
4061 	if (n == NULL) {
4062 		struct pf_srchash *sh;
4063 
4064 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4065 		    i++, sh++) {
4066 			PF_HASHROW_LOCK(sh);
4067 			LIST_FOREACH(n, &sh->nodes, entry) {
4068 				n->expire = 1;
4069 				n->states = 0;
4070 			}
4071 			PF_HASHROW_UNLOCK(sh);
4072 		}
4073 	} else {
4074 		/* XXX: hash slot should already be locked here. */
4075 		n->expire = 1;
4076 		n->states = 0;
4077 	}
4078 }
4079 
4080 static void
4081 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
4082 {
4083 	struct pf_src_node_list	 kill;
4084 
4085 	LIST_INIT(&kill);
4086 	for (int i = 0; i <= pf_srchashmask; i++) {
4087 		struct pf_srchash *sh = &V_pf_srchash[i];
4088 		struct pf_src_node *sn, *tmp;
4089 
4090 		PF_HASHROW_LOCK(sh);
4091 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
4092 			if (PF_MATCHA(psnk->psnk_src.neg,
4093 			      &psnk->psnk_src.addr.v.a.addr,
4094 			      &psnk->psnk_src.addr.v.a.mask,
4095 			      &sn->addr, sn->af) &&
4096 			    PF_MATCHA(psnk->psnk_dst.neg,
4097 			      &psnk->psnk_dst.addr.v.a.addr,
4098 			      &psnk->psnk_dst.addr.v.a.mask,
4099 			      &sn->raddr, sn->af)) {
4100 				pf_unlink_src_node(sn);
4101 				LIST_INSERT_HEAD(&kill, sn, entry);
4102 				sn->expire = 1;
4103 			}
4104 		PF_HASHROW_UNLOCK(sh);
4105 	}
4106 
4107 	for (int i = 0; i <= pf_hashmask; i++) {
4108 		struct pf_idhash *ih = &V_pf_idhash[i];
4109 		struct pf_state *s;
4110 
4111 		PF_HASHROW_LOCK(ih);
4112 		LIST_FOREACH(s, &ih->states, entry) {
4113 			if (s->src_node && s->src_node->expire == 1)
4114 				s->src_node = NULL;
4115 			if (s->nat_src_node && s->nat_src_node->expire == 1)
4116 				s->nat_src_node = NULL;
4117 		}
4118 		PF_HASHROW_UNLOCK(ih);
4119 	}
4120 
4121 	psnk->psnk_killed = pf_free_src_nodes(&kill);
4122 }
4123 
4124 /*
4125  * XXX - Check for version missmatch!!!
4126  */
4127 
4128 /*
4129  * Duplicate pfctl -Fa operation to get rid of as much as we can.
4130  */
4131 static int
4132 shutdown_pf(void)
4133 {
4134 	int error = 0;
4135 	u_int32_t t[5];
4136 	char nn = '\0';
4137 
4138 	do {
4139 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
4140 		    != 0) {
4141 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
4142 			break;
4143 		}
4144 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
4145 		    != 0) {
4146 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
4147 			break;		/* XXX: rollback? */
4148 		}
4149 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
4150 		    != 0) {
4151 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
4152 			break;		/* XXX: rollback? */
4153 		}
4154 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
4155 		    != 0) {
4156 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
4157 			break;		/* XXX: rollback? */
4158 		}
4159 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
4160 		    != 0) {
4161 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
4162 			break;		/* XXX: rollback? */
4163 		}
4164 
4165 		/* XXX: these should always succeed here */
4166 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
4167 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
4168 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
4169 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
4170 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
4171 
4172 		if ((error = pf_clear_tables()) != 0)
4173 			break;
4174 
4175 #ifdef ALTQ
4176 		if ((error = pf_begin_altq(&t[0])) != 0) {
4177 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
4178 			break;
4179 		}
4180 		pf_commit_altq(t[0]);
4181 #endif
4182 
4183 		pf_clear_states();
4184 
4185 		pf_clear_srcnodes(NULL);
4186 
4187 		/* status does not use malloced mem so no need to cleanup */
4188 		/* fingerprints and interfaces have their own cleanup code */
4189 	} while(0);
4190 
4191 	return (error);
4192 }
4193 
4194 static pfil_return_t
4195 pf_check_return(int chk, struct mbuf **m)
4196 {
4197 
4198 	switch (chk) {
4199 	case PF_PASS:
4200 		if (*m == NULL)
4201 			return (PFIL_CONSUMED);
4202 		else
4203 			return (PFIL_PASS);
4204 		break;
4205 	default:
4206 		if (*m != NULL) {
4207 			m_freem(*m);
4208 			*m = NULL;
4209 		}
4210 		return (PFIL_DROPPED);
4211 	}
4212 }
4213 
4214 #ifdef INET
4215 static pfil_return_t
4216 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
4217     void *ruleset __unused, struct inpcb *inp)
4218 {
4219 	int chk;
4220 
4221 	chk = pf_test(PF_IN, flags, ifp, m, inp);
4222 
4223 	return (pf_check_return(chk, m));
4224 }
4225 
4226 static pfil_return_t
4227 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
4228     void *ruleset __unused,  struct inpcb *inp)
4229 {
4230 	int chk;
4231 
4232 	chk = pf_test(PF_OUT, flags, ifp, m, inp);
4233 
4234 	return (pf_check_return(chk, m));
4235 }
4236 #endif
4237 
4238 #ifdef INET6
4239 static pfil_return_t
4240 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
4241     void *ruleset __unused,  struct inpcb *inp)
4242 {
4243 	int chk;
4244 
4245 	/*
4246 	 * In case of loopback traffic IPv6 uses the real interface in
4247 	 * order to support scoped addresses. In order to support stateful
4248 	 * filtering we have change this to lo0 as it is the case in IPv4.
4249 	 */
4250 	CURVNET_SET(ifp->if_vnet);
4251 	chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
4252 	CURVNET_RESTORE();
4253 
4254 	return (pf_check_return(chk, m));
4255 }
4256 
4257 static pfil_return_t
4258 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
4259     void *ruleset __unused,  struct inpcb *inp)
4260 {
4261 	int chk;
4262 
4263 	CURVNET_SET(ifp->if_vnet);
4264 	chk = pf_test6(PF_OUT, flags, ifp, m, inp);
4265 	CURVNET_RESTORE();
4266 
4267 	return (pf_check_return(chk, m));
4268 }
4269 #endif /* INET6 */
4270 
4271 #ifdef INET
4272 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
4273 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
4274 #define	V_pf_ip4_in_hook	VNET(pf_ip4_in_hook)
4275 #define	V_pf_ip4_out_hook	VNET(pf_ip4_out_hook)
4276 #endif
4277 #ifdef INET6
4278 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
4279 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
4280 #define	V_pf_ip6_in_hook	VNET(pf_ip6_in_hook)
4281 #define	V_pf_ip6_out_hook	VNET(pf_ip6_out_hook)
4282 #endif
4283 
4284 static int
4285 hook_pf(void)
4286 {
4287 	struct pfil_hook_args pha;
4288 	struct pfil_link_args pla;
4289 
4290 	if (V_pf_pfil_hooked)
4291 		return (0);
4292 
4293 	pha.pa_version = PFIL_VERSION;
4294 	pha.pa_modname = "pf";
4295 	pha.pa_ruleset = NULL;
4296 
4297 	pla.pa_version = PFIL_VERSION;
4298 
4299 #ifdef INET
4300 	pha.pa_type = PFIL_TYPE_IP4;
4301 	pha.pa_func = pf_check_in;
4302 	pha.pa_flags = PFIL_IN;
4303 	pha.pa_rulname = "default-in";
4304 	V_pf_ip4_in_hook = pfil_add_hook(&pha);
4305 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
4306 	pla.pa_head = V_inet_pfil_head;
4307 	pla.pa_hook = V_pf_ip4_in_hook;
4308 	(void)pfil_link(&pla);
4309 	pha.pa_func = pf_check_out;
4310 	pha.pa_flags = PFIL_OUT;
4311 	pha.pa_rulname = "default-out";
4312 	V_pf_ip4_out_hook = pfil_add_hook(&pha);
4313 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
4314 	pla.pa_head = V_inet_pfil_head;
4315 	pla.pa_hook = V_pf_ip4_out_hook;
4316 	(void)pfil_link(&pla);
4317 #endif
4318 #ifdef INET6
4319 	pha.pa_type = PFIL_TYPE_IP6;
4320 	pha.pa_func = pf_check6_in;
4321 	pha.pa_flags = PFIL_IN;
4322 	pha.pa_rulname = "default-in6";
4323 	V_pf_ip6_in_hook = pfil_add_hook(&pha);
4324 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
4325 	pla.pa_head = V_inet6_pfil_head;
4326 	pla.pa_hook = V_pf_ip6_in_hook;
4327 	(void)pfil_link(&pla);
4328 	pha.pa_func = pf_check6_out;
4329 	pha.pa_rulname = "default-out6";
4330 	pha.pa_flags = PFIL_OUT;
4331 	V_pf_ip6_out_hook = pfil_add_hook(&pha);
4332 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
4333 	pla.pa_head = V_inet6_pfil_head;
4334 	pla.pa_hook = V_pf_ip6_out_hook;
4335 	(void)pfil_link(&pla);
4336 #endif
4337 
4338 	V_pf_pfil_hooked = 1;
4339 	return (0);
4340 }
4341 
4342 static int
4343 dehook_pf(void)
4344 {
4345 
4346 	if (V_pf_pfil_hooked == 0)
4347 		return (0);
4348 
4349 #ifdef INET
4350 	pfil_remove_hook(V_pf_ip4_in_hook);
4351 	pfil_remove_hook(V_pf_ip4_out_hook);
4352 #endif
4353 #ifdef INET6
4354 	pfil_remove_hook(V_pf_ip6_in_hook);
4355 	pfil_remove_hook(V_pf_ip6_out_hook);
4356 #endif
4357 
4358 	V_pf_pfil_hooked = 0;
4359 	return (0);
4360 }
4361 
4362 static void
4363 pf_load_vnet(void)
4364 {
4365 	V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
4366 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
4367 
4368 	pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
4369 	    PF_RULE_TAG_HASH_SIZE_DEFAULT);
4370 #ifdef ALTQ
4371 	pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
4372 	    PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
4373 #endif
4374 
4375 	pfattach_vnet();
4376 	V_pf_vnet_active = 1;
4377 }
4378 
4379 static int
4380 pf_load(void)
4381 {
4382 	int error;
4383 
4384 	rm_init(&pf_rules_lock, "pf rulesets");
4385 	sx_init(&pf_ioctl_lock, "pf ioctl");
4386 	sx_init(&pf_end_lock, "pf end thread");
4387 
4388 	pf_mtag_initialize();
4389 
4390 	pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
4391 	if (pf_dev == NULL)
4392 		return (ENOMEM);
4393 
4394 	pf_end_threads = 0;
4395 	error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
4396 	if (error != 0)
4397 		return (error);
4398 
4399 	pfi_initialize();
4400 
4401 	return (0);
4402 }
4403 
4404 static void
4405 pf_unload_vnet(void)
4406 {
4407 	int error;
4408 
4409 	V_pf_vnet_active = 0;
4410 	V_pf_status.running = 0;
4411 	error = dehook_pf();
4412 	if (error) {
4413 		/*
4414 		 * Should not happen!
4415 		 * XXX Due to error code ESRCH, kldunload will show
4416 		 * a message like 'No such process'.
4417 		 */
4418 		printf("%s : pfil unregisteration fail\n", __FUNCTION__);
4419 		return;
4420 	}
4421 
4422 	PF_RULES_WLOCK();
4423 	shutdown_pf();
4424 	PF_RULES_WUNLOCK();
4425 
4426 	swi_remove(V_pf_swi_cookie);
4427 
4428 	pf_unload_vnet_purge();
4429 
4430 	pf_normalize_cleanup();
4431 	PF_RULES_WLOCK();
4432 	pfi_cleanup_vnet();
4433 	PF_RULES_WUNLOCK();
4434 	pfr_cleanup();
4435 	pf_osfp_flush();
4436 	pf_cleanup();
4437 	if (IS_DEFAULT_VNET(curvnet))
4438 		pf_mtag_cleanup();
4439 
4440 	pf_cleanup_tagset(&V_pf_tags);
4441 #ifdef ALTQ
4442 	pf_cleanup_tagset(&V_pf_qids);
4443 #endif
4444 	uma_zdestroy(V_pf_tag_z);
4445 
4446 	/* Free counters last as we updated them during shutdown. */
4447 	counter_u64_free(V_pf_default_rule.states_cur);
4448 	counter_u64_free(V_pf_default_rule.states_tot);
4449 	counter_u64_free(V_pf_default_rule.src_nodes);
4450 
4451 	for (int i = 0; i < PFRES_MAX; i++)
4452 		counter_u64_free(V_pf_status.counters[i]);
4453 	for (int i = 0; i < LCNT_MAX; i++)
4454 		counter_u64_free(V_pf_status.lcounters[i]);
4455 	for (int i = 0; i < FCNT_MAX; i++)
4456 		counter_u64_free(V_pf_status.fcounters[i]);
4457 	for (int i = 0; i < SCNT_MAX; i++)
4458 		counter_u64_free(V_pf_status.scounters[i]);
4459 }
4460 
4461 static void
4462 pf_unload(void)
4463 {
4464 
4465 	sx_xlock(&pf_end_lock);
4466 	pf_end_threads = 1;
4467 	while (pf_end_threads < 2) {
4468 		wakeup_one(pf_purge_thread);
4469 		sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
4470 	}
4471 	sx_xunlock(&pf_end_lock);
4472 
4473 	if (pf_dev != NULL)
4474 		destroy_dev(pf_dev);
4475 
4476 	pfi_cleanup();
4477 
4478 	rm_destroy(&pf_rules_lock);
4479 	sx_destroy(&pf_ioctl_lock);
4480 	sx_destroy(&pf_end_lock);
4481 }
4482 
4483 static void
4484 vnet_pf_init(void *unused __unused)
4485 {
4486 
4487 	pf_load_vnet();
4488 }
4489 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
4490     vnet_pf_init, NULL);
4491 
4492 static void
4493 vnet_pf_uninit(const void *unused __unused)
4494 {
4495 
4496 	pf_unload_vnet();
4497 }
4498 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
4499 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
4500     vnet_pf_uninit, NULL);
4501 
4502 
4503 static int
4504 pf_modevent(module_t mod, int type, void *data)
4505 {
4506 	int error = 0;
4507 
4508 	switch(type) {
4509 	case MOD_LOAD:
4510 		error = pf_load();
4511 		break;
4512 	case MOD_UNLOAD:
4513 		/* Handled in SYSUNINIT(pf_unload) to ensure it's done after
4514 		 * the vnet_pf_uninit()s */
4515 		break;
4516 	default:
4517 		error = EINVAL;
4518 		break;
4519 	}
4520 
4521 	return (error);
4522 }
4523 
4524 static moduledata_t pf_mod = {
4525 	"pf",
4526 	pf_modevent,
4527 	0
4528 };
4529 
4530 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
4531 MODULE_VERSION(pf, PF_MODVER);
4532