xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision 6f63e88c0166ed3e5f2805a9e667c7d24d304cf1)
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 #ifndef INET
1560 		if (pr->rule.af == AF_INET) {
1561 			error = EAFNOSUPPORT;
1562 			break;
1563 		}
1564 #endif /* INET */
1565 #ifndef INET6
1566 		if (pr->rule.af == AF_INET6) {
1567 			error = EAFNOSUPPORT;
1568 			break;
1569 		}
1570 #endif /* INET6 */
1571 
1572 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
1573 		bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1574 		if (rule->ifname[0])
1575 			kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1576 		rule->states_cur = counter_u64_alloc(M_WAITOK);
1577 		rule->states_tot = counter_u64_alloc(M_WAITOK);
1578 		rule->src_nodes = counter_u64_alloc(M_WAITOK);
1579 		rule->cuid = td->td_ucred->cr_ruid;
1580 		rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1581 		TAILQ_INIT(&rule->rpool.list);
1582 
1583 #define	ERROUT(x)	{ error = (x); goto DIOCADDRULE_error; }
1584 
1585 		PF_RULES_WLOCK();
1586 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1587 		ruleset = pf_find_ruleset(pr->anchor);
1588 		if (ruleset == NULL)
1589 			ERROUT(EINVAL);
1590 		rs_num = pf_get_ruleset_number(pr->rule.action);
1591 		if (rs_num >= PF_RULESET_MAX)
1592 			ERROUT(EINVAL);
1593 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1594 			DPFPRINTF(PF_DEBUG_MISC,
1595 			    ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
1596 			    ruleset->rules[rs_num].inactive.ticket));
1597 			ERROUT(EBUSY);
1598 		}
1599 		if (pr->pool_ticket != V_ticket_pabuf) {
1600 			DPFPRINTF(PF_DEBUG_MISC,
1601 			    ("pool_ticket: %d != %d\n", pr->pool_ticket,
1602 			    V_ticket_pabuf));
1603 			ERROUT(EBUSY);
1604 		}
1605 
1606 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1607 		    pf_rulequeue);
1608 		if (tail)
1609 			rule->nr = tail->nr + 1;
1610 		else
1611 			rule->nr = 0;
1612 		if (rule->ifname[0]) {
1613 			rule->kif = pfi_kif_attach(kif, rule->ifname);
1614 			pfi_kif_ref(rule->kif);
1615 		} else
1616 			rule->kif = NULL;
1617 
1618 		if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
1619 			error = EBUSY;
1620 
1621 #ifdef ALTQ
1622 		/* set queue IDs */
1623 		if (rule->qname[0] != 0) {
1624 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1625 				error = EBUSY;
1626 			else if (rule->pqname[0] != 0) {
1627 				if ((rule->pqid =
1628 				    pf_qname2qid(rule->pqname)) == 0)
1629 					error = EBUSY;
1630 			} else
1631 				rule->pqid = rule->qid;
1632 		}
1633 #endif
1634 		if (rule->tagname[0])
1635 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1636 				error = EBUSY;
1637 		if (rule->match_tagname[0])
1638 			if ((rule->match_tag =
1639 			    pf_tagname2tag(rule->match_tagname)) == 0)
1640 				error = EBUSY;
1641 		if (rule->rt && !rule->direction)
1642 			error = EINVAL;
1643 		if (!rule->log)
1644 			rule->logif = 0;
1645 		if (rule->logif >= PFLOGIFS_MAX)
1646 			error = EINVAL;
1647 		if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1648 			error = ENOMEM;
1649 		if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1650 			error = ENOMEM;
1651 		if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1652 			error = EINVAL;
1653 		if (rule->scrub_flags & PFSTATE_SETPRIO &&
1654 		    (rule->set_prio[0] > PF_PRIO_MAX ||
1655 		    rule->set_prio[1] > PF_PRIO_MAX))
1656 			error = EINVAL;
1657 		TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1658 			if (pa->addr.type == PF_ADDR_TABLE) {
1659 				pa->addr.p.tbl = pfr_attach_table(ruleset,
1660 				    pa->addr.v.tblname);
1661 				if (pa->addr.p.tbl == NULL)
1662 					error = ENOMEM;
1663 			}
1664 
1665 		rule->overload_tbl = NULL;
1666 		if (rule->overload_tblname[0]) {
1667 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1668 			    rule->overload_tblname)) == NULL)
1669 				error = EINVAL;
1670 			else
1671 				rule->overload_tbl->pfrkt_flags |=
1672 				    PFR_TFLAG_ACTIVE;
1673 		}
1674 
1675 		pf_mv_pool(&V_pf_pabuf, &rule->rpool.list);
1676 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1677 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1678 		    (rule->rt > PF_NOPFROUTE)) &&
1679 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1680 			error = EINVAL;
1681 
1682 		if (error) {
1683 			pf_free_rule(rule);
1684 			PF_RULES_WUNLOCK();
1685 			break;
1686 		}
1687 
1688 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1689 		rule->evaluations = rule->packets[0] = rule->packets[1] =
1690 		    rule->bytes[0] = rule->bytes[1] = 0;
1691 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1692 		    rule, entries);
1693 		ruleset->rules[rs_num].inactive.rcount++;
1694 		PF_RULES_WUNLOCK();
1695 		break;
1696 
1697 #undef ERROUT
1698 DIOCADDRULE_error:
1699 		PF_RULES_WUNLOCK();
1700 		counter_u64_free(rule->states_cur);
1701 		counter_u64_free(rule->states_tot);
1702 		counter_u64_free(rule->src_nodes);
1703 		free(rule, M_PFRULE);
1704 		if (kif)
1705 			free(kif, PFI_MTYPE);
1706 		break;
1707 	}
1708 
1709 	case DIOCGETRULES: {
1710 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1711 		struct pf_ruleset	*ruleset;
1712 		struct pf_rule		*tail;
1713 		int			 rs_num;
1714 
1715 		PF_RULES_WLOCK();
1716 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1717 		ruleset = pf_find_ruleset(pr->anchor);
1718 		if (ruleset == NULL) {
1719 			PF_RULES_WUNLOCK();
1720 			error = EINVAL;
1721 			break;
1722 		}
1723 		rs_num = pf_get_ruleset_number(pr->rule.action);
1724 		if (rs_num >= PF_RULESET_MAX) {
1725 			PF_RULES_WUNLOCK();
1726 			error = EINVAL;
1727 			break;
1728 		}
1729 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1730 		    pf_rulequeue);
1731 		if (tail)
1732 			pr->nr = tail->nr + 1;
1733 		else
1734 			pr->nr = 0;
1735 		pr->ticket = ruleset->rules[rs_num].active.ticket;
1736 		PF_RULES_WUNLOCK();
1737 		break;
1738 	}
1739 
1740 	case DIOCGETRULE: {
1741 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1742 		struct pf_ruleset	*ruleset;
1743 		struct pf_rule		*rule;
1744 		int			 rs_num, i;
1745 
1746 		PF_RULES_WLOCK();
1747 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1748 		ruleset = pf_find_ruleset(pr->anchor);
1749 		if (ruleset == NULL) {
1750 			PF_RULES_WUNLOCK();
1751 			error = EINVAL;
1752 			break;
1753 		}
1754 		rs_num = pf_get_ruleset_number(pr->rule.action);
1755 		if (rs_num >= PF_RULESET_MAX) {
1756 			PF_RULES_WUNLOCK();
1757 			error = EINVAL;
1758 			break;
1759 		}
1760 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1761 			PF_RULES_WUNLOCK();
1762 			error = EBUSY;
1763 			break;
1764 		}
1765 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1766 		while ((rule != NULL) && (rule->nr != pr->nr))
1767 			rule = TAILQ_NEXT(rule, entries);
1768 		if (rule == NULL) {
1769 			PF_RULES_WUNLOCK();
1770 			error = EBUSY;
1771 			break;
1772 		}
1773 		bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1774 		pr->rule.u_states_cur = counter_u64_fetch(rule->states_cur);
1775 		pr->rule.u_states_tot = counter_u64_fetch(rule->states_tot);
1776 		pr->rule.u_src_nodes = counter_u64_fetch(rule->src_nodes);
1777 		if (pf_anchor_copyout(ruleset, rule, pr)) {
1778 			PF_RULES_WUNLOCK();
1779 			error = EBUSY;
1780 			break;
1781 		}
1782 		pf_addr_copyout(&pr->rule.src.addr);
1783 		pf_addr_copyout(&pr->rule.dst.addr);
1784 		for (i = 0; i < PF_SKIP_COUNT; ++i)
1785 			if (rule->skip[i].ptr == NULL)
1786 				pr->rule.skip[i].nr = -1;
1787 			else
1788 				pr->rule.skip[i].nr =
1789 				    rule->skip[i].ptr->nr;
1790 
1791 		if (pr->action == PF_GET_CLR_CNTR) {
1792 			rule->evaluations = 0;
1793 			rule->packets[0] = rule->packets[1] = 0;
1794 			rule->bytes[0] = rule->bytes[1] = 0;
1795 			counter_u64_zero(rule->states_tot);
1796 		}
1797 		PF_RULES_WUNLOCK();
1798 		break;
1799 	}
1800 
1801 	case DIOCCHANGERULE: {
1802 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1803 		struct pf_ruleset	*ruleset;
1804 		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1805 		struct pfi_kif		*kif = NULL;
1806 		struct pf_pooladdr	*pa;
1807 		u_int32_t		 nr = 0;
1808 		int			 rs_num;
1809 
1810 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1811 		    pcr->action > PF_CHANGE_GET_TICKET) {
1812 			error = EINVAL;
1813 			break;
1814 		}
1815 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1816 			error = EINVAL;
1817 			break;
1818 		}
1819 
1820 		if (pcr->action != PF_CHANGE_REMOVE) {
1821 #ifndef INET
1822 			if (pcr->rule.af == AF_INET) {
1823 				error = EAFNOSUPPORT;
1824 				break;
1825 			}
1826 #endif /* INET */
1827 #ifndef INET6
1828 			if (pcr->rule.af == AF_INET6) {
1829 				error = EAFNOSUPPORT;
1830 				break;
1831 			}
1832 #endif /* INET6 */
1833 			newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
1834 			bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1835 			if (newrule->ifname[0])
1836 				kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1837 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
1838 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
1839 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
1840 			newrule->cuid = td->td_ucred->cr_ruid;
1841 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1842 			TAILQ_INIT(&newrule->rpool.list);
1843 		}
1844 
1845 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGERULE_error; }
1846 
1847 		PF_RULES_WLOCK();
1848 		if (!(pcr->action == PF_CHANGE_REMOVE ||
1849 		    pcr->action == PF_CHANGE_GET_TICKET) &&
1850 		    pcr->pool_ticket != V_ticket_pabuf)
1851 			ERROUT(EBUSY);
1852 
1853 		ruleset = pf_find_ruleset(pcr->anchor);
1854 		if (ruleset == NULL)
1855 			ERROUT(EINVAL);
1856 
1857 		rs_num = pf_get_ruleset_number(pcr->rule.action);
1858 		if (rs_num >= PF_RULESET_MAX)
1859 			ERROUT(EINVAL);
1860 
1861 		if (pcr->action == PF_CHANGE_GET_TICKET) {
1862 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1863 			ERROUT(0);
1864 		} else if (pcr->ticket !=
1865 			    ruleset->rules[rs_num].active.ticket)
1866 				ERROUT(EINVAL);
1867 
1868 		if (pcr->action != PF_CHANGE_REMOVE) {
1869 			if (newrule->ifname[0]) {
1870 				newrule->kif = pfi_kif_attach(kif,
1871 				    newrule->ifname);
1872 				pfi_kif_ref(newrule->kif);
1873 			} else
1874 				newrule->kif = NULL;
1875 
1876 			if (newrule->rtableid > 0 &&
1877 			    newrule->rtableid >= rt_numfibs)
1878 				error = EBUSY;
1879 
1880 #ifdef ALTQ
1881 			/* set queue IDs */
1882 			if (newrule->qname[0] != 0) {
1883 				if ((newrule->qid =
1884 				    pf_qname2qid(newrule->qname)) == 0)
1885 					error = EBUSY;
1886 				else if (newrule->pqname[0] != 0) {
1887 					if ((newrule->pqid =
1888 					    pf_qname2qid(newrule->pqname)) == 0)
1889 						error = EBUSY;
1890 				} else
1891 					newrule->pqid = newrule->qid;
1892 			}
1893 #endif /* ALTQ */
1894 			if (newrule->tagname[0])
1895 				if ((newrule->tag =
1896 				    pf_tagname2tag(newrule->tagname)) == 0)
1897 					error = EBUSY;
1898 			if (newrule->match_tagname[0])
1899 				if ((newrule->match_tag = pf_tagname2tag(
1900 				    newrule->match_tagname)) == 0)
1901 					error = EBUSY;
1902 			if (newrule->rt && !newrule->direction)
1903 				error = EINVAL;
1904 			if (!newrule->log)
1905 				newrule->logif = 0;
1906 			if (newrule->logif >= PFLOGIFS_MAX)
1907 				error = EINVAL;
1908 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
1909 				error = ENOMEM;
1910 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
1911 				error = ENOMEM;
1912 			if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1913 				error = EINVAL;
1914 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1915 				if (pa->addr.type == PF_ADDR_TABLE) {
1916 					pa->addr.p.tbl =
1917 					    pfr_attach_table(ruleset,
1918 					    pa->addr.v.tblname);
1919 					if (pa->addr.p.tbl == NULL)
1920 						error = ENOMEM;
1921 				}
1922 
1923 			newrule->overload_tbl = NULL;
1924 			if (newrule->overload_tblname[0]) {
1925 				if ((newrule->overload_tbl = pfr_attach_table(
1926 				    ruleset, newrule->overload_tblname)) ==
1927 				    NULL)
1928 					error = EINVAL;
1929 				else
1930 					newrule->overload_tbl->pfrkt_flags |=
1931 					    PFR_TFLAG_ACTIVE;
1932 			}
1933 
1934 			pf_mv_pool(&V_pf_pabuf, &newrule->rpool.list);
1935 			if (((((newrule->action == PF_NAT) ||
1936 			    (newrule->action == PF_RDR) ||
1937 			    (newrule->action == PF_BINAT) ||
1938 			    (newrule->rt > PF_NOPFROUTE)) &&
1939 			    !newrule->anchor)) &&
1940 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1941 				error = EINVAL;
1942 
1943 			if (error) {
1944 				pf_free_rule(newrule);
1945 				PF_RULES_WUNLOCK();
1946 				break;
1947 			}
1948 
1949 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1950 			newrule->evaluations = 0;
1951 			newrule->packets[0] = newrule->packets[1] = 0;
1952 			newrule->bytes[0] = newrule->bytes[1] = 0;
1953 		}
1954 		pf_empty_pool(&V_pf_pabuf);
1955 
1956 		if (pcr->action == PF_CHANGE_ADD_HEAD)
1957 			oldrule = TAILQ_FIRST(
1958 			    ruleset->rules[rs_num].active.ptr);
1959 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1960 			oldrule = TAILQ_LAST(
1961 			    ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1962 		else {
1963 			oldrule = TAILQ_FIRST(
1964 			    ruleset->rules[rs_num].active.ptr);
1965 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1966 				oldrule = TAILQ_NEXT(oldrule, entries);
1967 			if (oldrule == NULL) {
1968 				if (newrule != NULL)
1969 					pf_free_rule(newrule);
1970 				PF_RULES_WUNLOCK();
1971 				error = EINVAL;
1972 				break;
1973 			}
1974 		}
1975 
1976 		if (pcr->action == PF_CHANGE_REMOVE) {
1977 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
1978 			    oldrule);
1979 			ruleset->rules[rs_num].active.rcount--;
1980 		} else {
1981 			if (oldrule == NULL)
1982 				TAILQ_INSERT_TAIL(
1983 				    ruleset->rules[rs_num].active.ptr,
1984 				    newrule, entries);
1985 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1986 			    pcr->action == PF_CHANGE_ADD_BEFORE)
1987 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1988 			else
1989 				TAILQ_INSERT_AFTER(
1990 				    ruleset->rules[rs_num].active.ptr,
1991 				    oldrule, newrule, entries);
1992 			ruleset->rules[rs_num].active.rcount++;
1993 		}
1994 
1995 		nr = 0;
1996 		TAILQ_FOREACH(oldrule,
1997 		    ruleset->rules[rs_num].active.ptr, entries)
1998 			oldrule->nr = nr++;
1999 
2000 		ruleset->rules[rs_num].active.ticket++;
2001 
2002 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
2003 		pf_remove_if_empty_ruleset(ruleset);
2004 
2005 		PF_RULES_WUNLOCK();
2006 		break;
2007 
2008 #undef ERROUT
2009 DIOCCHANGERULE_error:
2010 		PF_RULES_WUNLOCK();
2011 		if (newrule != NULL) {
2012 			counter_u64_free(newrule->states_cur);
2013 			counter_u64_free(newrule->states_tot);
2014 			counter_u64_free(newrule->src_nodes);
2015 			free(newrule, M_PFRULE);
2016 		}
2017 		if (kif != NULL)
2018 			free(kif, PFI_MTYPE);
2019 		break;
2020 	}
2021 
2022 	case DIOCCLRSTATES: {
2023 		struct pf_state		*s;
2024 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
2025 		u_int			 i, killed = 0;
2026 
2027 		for (i = 0; i <= pf_hashmask; i++) {
2028 			struct pf_idhash *ih = &V_pf_idhash[i];
2029 
2030 relock_DIOCCLRSTATES:
2031 			PF_HASHROW_LOCK(ih);
2032 			LIST_FOREACH(s, &ih->states, entry)
2033 				if (!psk->psk_ifname[0] ||
2034 				    !strcmp(psk->psk_ifname,
2035 				    s->kif->pfik_name)) {
2036 					/*
2037 					 * Don't send out individual
2038 					 * delete messages.
2039 					 */
2040 					s->state_flags |= PFSTATE_NOSYNC;
2041 					pf_unlink_state(s, PF_ENTER_LOCKED);
2042 					killed++;
2043 					goto relock_DIOCCLRSTATES;
2044 				}
2045 			PF_HASHROW_UNLOCK(ih);
2046 		}
2047 		psk->psk_killed = killed;
2048 		if (V_pfsync_clear_states_ptr != NULL)
2049 			V_pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
2050 		break;
2051 	}
2052 
2053 	case DIOCKILLSTATES: {
2054 		struct pf_state		*s;
2055 		struct pf_state_key	*sk;
2056 		struct pf_addr		*srcaddr, *dstaddr;
2057 		u_int16_t		 srcport, dstport;
2058 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
2059 		u_int			 i, killed = 0;
2060 
2061 		if (psk->psk_pfcmp.id) {
2062 			if (psk->psk_pfcmp.creatorid == 0)
2063 				psk->psk_pfcmp.creatorid = V_pf_status.hostid;
2064 			if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
2065 			    psk->psk_pfcmp.creatorid))) {
2066 				pf_unlink_state(s, PF_ENTER_LOCKED);
2067 				psk->psk_killed = 1;
2068 			}
2069 			break;
2070 		}
2071 
2072 		for (i = 0; i <= pf_hashmask; i++) {
2073 			struct pf_idhash *ih = &V_pf_idhash[i];
2074 
2075 relock_DIOCKILLSTATES:
2076 			PF_HASHROW_LOCK(ih);
2077 			LIST_FOREACH(s, &ih->states, entry) {
2078 				sk = s->key[PF_SK_WIRE];
2079 				if (s->direction == PF_OUT) {
2080 					srcaddr = &sk->addr[1];
2081 					dstaddr = &sk->addr[0];
2082 					srcport = sk->port[1];
2083 					dstport = sk->port[0];
2084 				} else {
2085 					srcaddr = &sk->addr[0];
2086 					dstaddr = &sk->addr[1];
2087 					srcport = sk->port[0];
2088 					dstport = sk->port[1];
2089 				}
2090 
2091 				if ((!psk->psk_af || sk->af == psk->psk_af)
2092 				    && (!psk->psk_proto || psk->psk_proto ==
2093 				    sk->proto) &&
2094 				    PF_MATCHA(psk->psk_src.neg,
2095 				    &psk->psk_src.addr.v.a.addr,
2096 				    &psk->psk_src.addr.v.a.mask,
2097 				    srcaddr, sk->af) &&
2098 				    PF_MATCHA(psk->psk_dst.neg,
2099 				    &psk->psk_dst.addr.v.a.addr,
2100 				    &psk->psk_dst.addr.v.a.mask,
2101 				    dstaddr, sk->af) &&
2102 				    (psk->psk_src.port_op == 0 ||
2103 				    pf_match_port(psk->psk_src.port_op,
2104 				    psk->psk_src.port[0], psk->psk_src.port[1],
2105 				    srcport)) &&
2106 				    (psk->psk_dst.port_op == 0 ||
2107 				    pf_match_port(psk->psk_dst.port_op,
2108 				    psk->psk_dst.port[0], psk->psk_dst.port[1],
2109 				    dstport)) &&
2110 				    (!psk->psk_label[0] ||
2111 				    (s->rule.ptr->label[0] &&
2112 				    !strcmp(psk->psk_label,
2113 				    s->rule.ptr->label))) &&
2114 				    (!psk->psk_ifname[0] ||
2115 				    !strcmp(psk->psk_ifname,
2116 				    s->kif->pfik_name))) {
2117 					pf_unlink_state(s, PF_ENTER_LOCKED);
2118 					killed++;
2119 					goto relock_DIOCKILLSTATES;
2120 				}
2121 			}
2122 			PF_HASHROW_UNLOCK(ih);
2123 		}
2124 		psk->psk_killed = killed;
2125 		break;
2126 	}
2127 
2128 	case DIOCADDSTATE: {
2129 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
2130 		struct pfsync_state	*sp = &ps->state;
2131 
2132 		if (sp->timeout >= PFTM_MAX) {
2133 			error = EINVAL;
2134 			break;
2135 		}
2136 		if (V_pfsync_state_import_ptr != NULL) {
2137 			PF_RULES_RLOCK();
2138 			error = V_pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
2139 			PF_RULES_RUNLOCK();
2140 		} else
2141 			error = EOPNOTSUPP;
2142 		break;
2143 	}
2144 
2145 	case DIOCGETSTATE: {
2146 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
2147 		struct pf_state		*s;
2148 
2149 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
2150 		if (s == NULL) {
2151 			error = ENOENT;
2152 			break;
2153 		}
2154 
2155 		pfsync_state_export(&ps->state, s);
2156 		PF_STATE_UNLOCK(s);
2157 		break;
2158 	}
2159 
2160 	case DIOCGETSTATES: {
2161 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
2162 		struct pf_state		*s;
2163 		struct pfsync_state	*pstore, *p;
2164 		int i, nr;
2165 
2166 		if (ps->ps_len <= 0) {
2167 			nr = uma_zone_get_cur(V_pf_state_z);
2168 			ps->ps_len = sizeof(struct pfsync_state) * nr;
2169 			break;
2170 		}
2171 
2172 		p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK | M_ZERO);
2173 		nr = 0;
2174 
2175 		for (i = 0; i <= pf_hashmask; i++) {
2176 			struct pf_idhash *ih = &V_pf_idhash[i];
2177 
2178 			PF_HASHROW_LOCK(ih);
2179 			LIST_FOREACH(s, &ih->states, entry) {
2180 
2181 				if (s->timeout == PFTM_UNLINKED)
2182 					continue;
2183 
2184 				if ((nr+1) * sizeof(*p) > ps->ps_len) {
2185 					PF_HASHROW_UNLOCK(ih);
2186 					goto DIOCGETSTATES_full;
2187 				}
2188 				pfsync_state_export(p, s);
2189 				p++;
2190 				nr++;
2191 			}
2192 			PF_HASHROW_UNLOCK(ih);
2193 		}
2194 DIOCGETSTATES_full:
2195 		error = copyout(pstore, ps->ps_states,
2196 		    sizeof(struct pfsync_state) * nr);
2197 		if (error) {
2198 			free(pstore, M_TEMP);
2199 			break;
2200 		}
2201 		ps->ps_len = sizeof(struct pfsync_state) * nr;
2202 		free(pstore, M_TEMP);
2203 
2204 		break;
2205 	}
2206 
2207 	case DIOCGETSTATUS: {
2208 		struct pf_status *s = (struct pf_status *)addr;
2209 
2210 		PF_RULES_RLOCK();
2211 		s->running = V_pf_status.running;
2212 		s->since   = V_pf_status.since;
2213 		s->debug   = V_pf_status.debug;
2214 		s->hostid  = V_pf_status.hostid;
2215 		s->states  = V_pf_status.states;
2216 		s->src_nodes = V_pf_status.src_nodes;
2217 
2218 		for (int i = 0; i < PFRES_MAX; i++)
2219 			s->counters[i] =
2220 			    counter_u64_fetch(V_pf_status.counters[i]);
2221 		for (int i = 0; i < LCNT_MAX; i++)
2222 			s->lcounters[i] =
2223 			    counter_u64_fetch(V_pf_status.lcounters[i]);
2224 		for (int i = 0; i < FCNT_MAX; i++)
2225 			s->fcounters[i] =
2226 			    counter_u64_fetch(V_pf_status.fcounters[i]);
2227 		for (int i = 0; i < SCNT_MAX; i++)
2228 			s->scounters[i] =
2229 			    counter_u64_fetch(V_pf_status.scounters[i]);
2230 
2231 		bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ);
2232 		bcopy(V_pf_status.pf_chksum, s->pf_chksum,
2233 		    PF_MD5_DIGEST_LENGTH);
2234 
2235 		pfi_update_status(s->ifname, s);
2236 		PF_RULES_RUNLOCK();
2237 		break;
2238 	}
2239 
2240 	case DIOCSETSTATUSIF: {
2241 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
2242 
2243 		if (pi->ifname[0] == 0) {
2244 			bzero(V_pf_status.ifname, IFNAMSIZ);
2245 			break;
2246 		}
2247 		PF_RULES_WLOCK();
2248 		strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
2249 		PF_RULES_WUNLOCK();
2250 		break;
2251 	}
2252 
2253 	case DIOCCLRSTATUS: {
2254 		PF_RULES_WLOCK();
2255 		for (int i = 0; i < PFRES_MAX; i++)
2256 			counter_u64_zero(V_pf_status.counters[i]);
2257 		for (int i = 0; i < FCNT_MAX; i++)
2258 			counter_u64_zero(V_pf_status.fcounters[i]);
2259 		for (int i = 0; i < SCNT_MAX; i++)
2260 			counter_u64_zero(V_pf_status.scounters[i]);
2261 		for (int i = 0; i < LCNT_MAX; i++)
2262 			counter_u64_zero(V_pf_status.lcounters[i]);
2263 		V_pf_status.since = time_second;
2264 		if (*V_pf_status.ifname)
2265 			pfi_update_status(V_pf_status.ifname, NULL);
2266 		PF_RULES_WUNLOCK();
2267 		break;
2268 	}
2269 
2270 	case DIOCNATLOOK: {
2271 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
2272 		struct pf_state_key	*sk;
2273 		struct pf_state		*state;
2274 		struct pf_state_key_cmp	 key;
2275 		int			 m = 0, direction = pnl->direction;
2276 		int			 sidx, didx;
2277 
2278 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
2279 		sidx = (direction == PF_IN) ? 1 : 0;
2280 		didx = (direction == PF_IN) ? 0 : 1;
2281 
2282 		if (!pnl->proto ||
2283 		    PF_AZERO(&pnl->saddr, pnl->af) ||
2284 		    PF_AZERO(&pnl->daddr, pnl->af) ||
2285 		    ((pnl->proto == IPPROTO_TCP ||
2286 		    pnl->proto == IPPROTO_UDP) &&
2287 		    (!pnl->dport || !pnl->sport)))
2288 			error = EINVAL;
2289 		else {
2290 			bzero(&key, sizeof(key));
2291 			key.af = pnl->af;
2292 			key.proto = pnl->proto;
2293 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
2294 			key.port[sidx] = pnl->sport;
2295 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
2296 			key.port[didx] = pnl->dport;
2297 
2298 			state = pf_find_state_all(&key, direction, &m);
2299 
2300 			if (m > 1)
2301 				error = E2BIG;	/* more than one state */
2302 			else if (state != NULL) {
2303 				/* XXXGL: not locked read */
2304 				sk = state->key[sidx];
2305 				PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
2306 				pnl->rsport = sk->port[sidx];
2307 				PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
2308 				pnl->rdport = sk->port[didx];
2309 			} else
2310 				error = ENOENT;
2311 		}
2312 		break;
2313 	}
2314 
2315 	case DIOCSETTIMEOUT: {
2316 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2317 		int		 old;
2318 
2319 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
2320 		    pt->seconds < 0) {
2321 			error = EINVAL;
2322 			break;
2323 		}
2324 		PF_RULES_WLOCK();
2325 		old = V_pf_default_rule.timeout[pt->timeout];
2326 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
2327 			pt->seconds = 1;
2328 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
2329 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
2330 			wakeup(pf_purge_thread);
2331 		pt->seconds = old;
2332 		PF_RULES_WUNLOCK();
2333 		break;
2334 	}
2335 
2336 	case DIOCGETTIMEOUT: {
2337 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2338 
2339 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
2340 			error = EINVAL;
2341 			break;
2342 		}
2343 		PF_RULES_RLOCK();
2344 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
2345 		PF_RULES_RUNLOCK();
2346 		break;
2347 	}
2348 
2349 	case DIOCGETLIMIT: {
2350 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2351 
2352 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
2353 			error = EINVAL;
2354 			break;
2355 		}
2356 		PF_RULES_RLOCK();
2357 		pl->limit = V_pf_limits[pl->index].limit;
2358 		PF_RULES_RUNLOCK();
2359 		break;
2360 	}
2361 
2362 	case DIOCSETLIMIT: {
2363 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2364 		int			 old_limit;
2365 
2366 		PF_RULES_WLOCK();
2367 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
2368 		    V_pf_limits[pl->index].zone == NULL) {
2369 			PF_RULES_WUNLOCK();
2370 			error = EINVAL;
2371 			break;
2372 		}
2373 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
2374 		old_limit = V_pf_limits[pl->index].limit;
2375 		V_pf_limits[pl->index].limit = pl->limit;
2376 		pl->limit = old_limit;
2377 		PF_RULES_WUNLOCK();
2378 		break;
2379 	}
2380 
2381 	case DIOCSETDEBUG: {
2382 		u_int32_t	*level = (u_int32_t *)addr;
2383 
2384 		PF_RULES_WLOCK();
2385 		V_pf_status.debug = *level;
2386 		PF_RULES_WUNLOCK();
2387 		break;
2388 	}
2389 
2390 	case DIOCCLRRULECTRS: {
2391 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
2392 		struct pf_ruleset	*ruleset = &pf_main_ruleset;
2393 		struct pf_rule		*rule;
2394 
2395 		PF_RULES_WLOCK();
2396 		TAILQ_FOREACH(rule,
2397 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
2398 			rule->evaluations = 0;
2399 			rule->packets[0] = rule->packets[1] = 0;
2400 			rule->bytes[0] = rule->bytes[1] = 0;
2401 		}
2402 		PF_RULES_WUNLOCK();
2403 		break;
2404 	}
2405 
2406 	case DIOCGIFSPEEDV0:
2407 	case DIOCGIFSPEEDV1: {
2408 		struct pf_ifspeed_v1	*psp = (struct pf_ifspeed_v1 *)addr;
2409 		struct pf_ifspeed_v1	ps;
2410 		struct ifnet		*ifp;
2411 
2412 		if (psp->ifname[0] != 0) {
2413 			/* Can we completely trust user-land? */
2414 			strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
2415 			ifp = ifunit(ps.ifname);
2416 			if (ifp != NULL) {
2417 				psp->baudrate32 =
2418 				    (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
2419 				if (cmd == DIOCGIFSPEEDV1)
2420 					psp->baudrate = ifp->if_baudrate;
2421 			} else
2422 				error = EINVAL;
2423 		} else
2424 			error = EINVAL;
2425 		break;
2426 	}
2427 
2428 #ifdef ALTQ
2429 	case DIOCSTARTALTQ: {
2430 		struct pf_altq		*altq;
2431 
2432 		PF_RULES_WLOCK();
2433 		/* enable all altq interfaces on active list */
2434 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
2435 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
2436 				error = pf_enable_altq(altq);
2437 				if (error != 0)
2438 					break;
2439 			}
2440 		}
2441 		if (error == 0)
2442 			V_pf_altq_running = 1;
2443 		PF_RULES_WUNLOCK();
2444 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
2445 		break;
2446 	}
2447 
2448 	case DIOCSTOPALTQ: {
2449 		struct pf_altq		*altq;
2450 
2451 		PF_RULES_WLOCK();
2452 		/* disable all altq interfaces on active list */
2453 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
2454 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
2455 				error = pf_disable_altq(altq);
2456 				if (error != 0)
2457 					break;
2458 			}
2459 		}
2460 		if (error == 0)
2461 			V_pf_altq_running = 0;
2462 		PF_RULES_WUNLOCK();
2463 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2464 		break;
2465 	}
2466 
2467 	case DIOCADDALTQV0:
2468 	case DIOCADDALTQV1: {
2469 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2470 		struct pf_altq		*altq, *a;
2471 		struct ifnet		*ifp;
2472 
2473 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
2474 		error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
2475 		if (error)
2476 			break;
2477 		altq->local_flags = 0;
2478 
2479 		PF_RULES_WLOCK();
2480 		if (pa->ticket != V_ticket_altqs_inactive) {
2481 			PF_RULES_WUNLOCK();
2482 			free(altq, M_PFALTQ);
2483 			error = EBUSY;
2484 			break;
2485 		}
2486 
2487 		/*
2488 		 * if this is for a queue, find the discipline and
2489 		 * copy the necessary fields
2490 		 */
2491 		if (altq->qname[0] != 0) {
2492 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2493 				PF_RULES_WUNLOCK();
2494 				error = EBUSY;
2495 				free(altq, M_PFALTQ);
2496 				break;
2497 			}
2498 			altq->altq_disc = NULL;
2499 			TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
2500 				if (strncmp(a->ifname, altq->ifname,
2501 				    IFNAMSIZ) == 0) {
2502 					altq->altq_disc = a->altq_disc;
2503 					break;
2504 				}
2505 			}
2506 		}
2507 
2508 		if ((ifp = ifunit(altq->ifname)) == NULL)
2509 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
2510 		else
2511 			error = altq_add(ifp, altq);
2512 
2513 		if (error) {
2514 			PF_RULES_WUNLOCK();
2515 			free(altq, M_PFALTQ);
2516 			break;
2517 		}
2518 
2519 		if (altq->qname[0] != 0)
2520 			TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
2521 		else
2522 			TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
2523 		/* version error check done on import above */
2524 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
2525 		PF_RULES_WUNLOCK();
2526 		break;
2527 	}
2528 
2529 	case DIOCGETALTQSV0:
2530 	case DIOCGETALTQSV1: {
2531 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2532 		struct pf_altq		*altq;
2533 
2534 		PF_RULES_RLOCK();
2535 		pa->nr = 0;
2536 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
2537 			pa->nr++;
2538 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
2539 			pa->nr++;
2540 		pa->ticket = V_ticket_altqs_active;
2541 		PF_RULES_RUNLOCK();
2542 		break;
2543 	}
2544 
2545 	case DIOCGETALTQV0:
2546 	case DIOCGETALTQV1: {
2547 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2548 		struct pf_altq		*altq;
2549 
2550 		PF_RULES_RLOCK();
2551 		if (pa->ticket != V_ticket_altqs_active) {
2552 			PF_RULES_RUNLOCK();
2553 			error = EBUSY;
2554 			break;
2555 		}
2556 		altq = pf_altq_get_nth_active(pa->nr);
2557 		if (altq == NULL) {
2558 			PF_RULES_RUNLOCK();
2559 			error = EBUSY;
2560 			break;
2561 		}
2562 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
2563 		PF_RULES_RUNLOCK();
2564 		break;
2565 	}
2566 
2567 	case DIOCCHANGEALTQV0:
2568 	case DIOCCHANGEALTQV1:
2569 		/* CHANGEALTQ not supported yet! */
2570 		error = ENODEV;
2571 		break;
2572 
2573 	case DIOCGETQSTATSV0:
2574 	case DIOCGETQSTATSV1: {
2575 		struct pfioc_qstats_v1	*pq = (struct pfioc_qstats_v1 *)addr;
2576 		struct pf_altq		*altq;
2577 		int			 nbytes;
2578 		u_int32_t		 version;
2579 
2580 		PF_RULES_RLOCK();
2581 		if (pq->ticket != V_ticket_altqs_active) {
2582 			PF_RULES_RUNLOCK();
2583 			error = EBUSY;
2584 			break;
2585 		}
2586 		nbytes = pq->nbytes;
2587 		altq = pf_altq_get_nth_active(pq->nr);
2588 		if (altq == NULL) {
2589 			PF_RULES_RUNLOCK();
2590 			error = EBUSY;
2591 			break;
2592 		}
2593 
2594 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
2595 			PF_RULES_RUNLOCK();
2596 			error = ENXIO;
2597 			break;
2598 		}
2599 		PF_RULES_RUNLOCK();
2600 		if (cmd == DIOCGETQSTATSV0)
2601 			version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
2602 		else
2603 			version = pq->version;
2604 		error = altq_getqstats(altq, pq->buf, &nbytes, version);
2605 		if (error == 0) {
2606 			pq->scheduler = altq->scheduler;
2607 			pq->nbytes = nbytes;
2608 		}
2609 		break;
2610 	}
2611 #endif /* ALTQ */
2612 
2613 	case DIOCBEGINADDRS: {
2614 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2615 
2616 		PF_RULES_WLOCK();
2617 		pf_empty_pool(&V_pf_pabuf);
2618 		pp->ticket = ++V_ticket_pabuf;
2619 		PF_RULES_WUNLOCK();
2620 		break;
2621 	}
2622 
2623 	case DIOCADDADDR: {
2624 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2625 		struct pf_pooladdr	*pa;
2626 		struct pfi_kif		*kif = NULL;
2627 
2628 #ifndef INET
2629 		if (pp->af == AF_INET) {
2630 			error = EAFNOSUPPORT;
2631 			break;
2632 		}
2633 #endif /* INET */
2634 #ifndef INET6
2635 		if (pp->af == AF_INET6) {
2636 			error = EAFNOSUPPORT;
2637 			break;
2638 		}
2639 #endif /* INET6 */
2640 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2641 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2642 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2643 			error = EINVAL;
2644 			break;
2645 		}
2646 		if (pp->addr.addr.p.dyn != NULL) {
2647 			error = EINVAL;
2648 			break;
2649 		}
2650 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2651 		bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2652 		if (pa->ifname[0])
2653 			kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2654 		PF_RULES_WLOCK();
2655 		if (pp->ticket != V_ticket_pabuf) {
2656 			PF_RULES_WUNLOCK();
2657 			if (pa->ifname[0])
2658 				free(kif, PFI_MTYPE);
2659 			free(pa, M_PFRULE);
2660 			error = EBUSY;
2661 			break;
2662 		}
2663 		if (pa->ifname[0]) {
2664 			pa->kif = pfi_kif_attach(kif, pa->ifname);
2665 			pfi_kif_ref(pa->kif);
2666 		} else
2667 			pa->kif = NULL;
2668 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2669 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2670 			if (pa->ifname[0])
2671 				pfi_kif_unref(pa->kif);
2672 			PF_RULES_WUNLOCK();
2673 			free(pa, M_PFRULE);
2674 			break;
2675 		}
2676 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2677 		PF_RULES_WUNLOCK();
2678 		break;
2679 	}
2680 
2681 	case DIOCGETADDRS: {
2682 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2683 		struct pf_pool		*pool;
2684 		struct pf_pooladdr	*pa;
2685 
2686 		PF_RULES_RLOCK();
2687 		pp->nr = 0;
2688 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2689 		    pp->r_num, 0, 1, 0);
2690 		if (pool == NULL) {
2691 			PF_RULES_RUNLOCK();
2692 			error = EBUSY;
2693 			break;
2694 		}
2695 		TAILQ_FOREACH(pa, &pool->list, entries)
2696 			pp->nr++;
2697 		PF_RULES_RUNLOCK();
2698 		break;
2699 	}
2700 
2701 	case DIOCGETADDR: {
2702 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2703 		struct pf_pool		*pool;
2704 		struct pf_pooladdr	*pa;
2705 		u_int32_t		 nr = 0;
2706 
2707 		PF_RULES_RLOCK();
2708 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2709 		    pp->r_num, 0, 1, 1);
2710 		if (pool == NULL) {
2711 			PF_RULES_RUNLOCK();
2712 			error = EBUSY;
2713 			break;
2714 		}
2715 		pa = TAILQ_FIRST(&pool->list);
2716 		while ((pa != NULL) && (nr < pp->nr)) {
2717 			pa = TAILQ_NEXT(pa, entries);
2718 			nr++;
2719 		}
2720 		if (pa == NULL) {
2721 			PF_RULES_RUNLOCK();
2722 			error = EBUSY;
2723 			break;
2724 		}
2725 		bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2726 		pf_addr_copyout(&pp->addr.addr);
2727 		PF_RULES_RUNLOCK();
2728 		break;
2729 	}
2730 
2731 	case DIOCCHANGEADDR: {
2732 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
2733 		struct pf_pool		*pool;
2734 		struct pf_pooladdr	*oldpa = NULL, *newpa = NULL;
2735 		struct pf_ruleset	*ruleset;
2736 		struct pfi_kif		*kif = NULL;
2737 
2738 		if (pca->action < PF_CHANGE_ADD_HEAD ||
2739 		    pca->action > PF_CHANGE_REMOVE) {
2740 			error = EINVAL;
2741 			break;
2742 		}
2743 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2744 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2745 		    pca->addr.addr.type != PF_ADDR_TABLE) {
2746 			error = EINVAL;
2747 			break;
2748 		}
2749 		if (pca->addr.addr.p.dyn != NULL) {
2750 			error = EINVAL;
2751 			break;
2752 		}
2753 
2754 		if (pca->action != PF_CHANGE_REMOVE) {
2755 #ifndef INET
2756 			if (pca->af == AF_INET) {
2757 				error = EAFNOSUPPORT;
2758 				break;
2759 			}
2760 #endif /* INET */
2761 #ifndef INET6
2762 			if (pca->af == AF_INET6) {
2763 				error = EAFNOSUPPORT;
2764 				break;
2765 			}
2766 #endif /* INET6 */
2767 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
2768 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2769 			if (newpa->ifname[0])
2770 				kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
2771 			newpa->kif = NULL;
2772 		}
2773 
2774 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGEADDR_error; }
2775 		PF_RULES_WLOCK();
2776 		ruleset = pf_find_ruleset(pca->anchor);
2777 		if (ruleset == NULL)
2778 			ERROUT(EBUSY);
2779 
2780 		pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2781 		    pca->r_num, pca->r_last, 1, 1);
2782 		if (pool == NULL)
2783 			ERROUT(EBUSY);
2784 
2785 		if (pca->action != PF_CHANGE_REMOVE) {
2786 			if (newpa->ifname[0]) {
2787 				newpa->kif = pfi_kif_attach(kif, newpa->ifname);
2788 				pfi_kif_ref(newpa->kif);
2789 				kif = NULL;
2790 			}
2791 
2792 			switch (newpa->addr.type) {
2793 			case PF_ADDR_DYNIFTL:
2794 				error = pfi_dynaddr_setup(&newpa->addr,
2795 				    pca->af);
2796 				break;
2797 			case PF_ADDR_TABLE:
2798 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
2799 				    newpa->addr.v.tblname);
2800 				if (newpa->addr.p.tbl == NULL)
2801 					error = ENOMEM;
2802 				break;
2803 			}
2804 			if (error)
2805 				goto DIOCCHANGEADDR_error;
2806 		}
2807 
2808 		switch (pca->action) {
2809 		case PF_CHANGE_ADD_HEAD:
2810 			oldpa = TAILQ_FIRST(&pool->list);
2811 			break;
2812 		case PF_CHANGE_ADD_TAIL:
2813 			oldpa = TAILQ_LAST(&pool->list, pf_palist);
2814 			break;
2815 		default:
2816 			oldpa = TAILQ_FIRST(&pool->list);
2817 			for (int i = 0; oldpa && i < pca->nr; i++)
2818 				oldpa = TAILQ_NEXT(oldpa, entries);
2819 
2820 			if (oldpa == NULL)
2821 				ERROUT(EINVAL);
2822 		}
2823 
2824 		if (pca->action == PF_CHANGE_REMOVE) {
2825 			TAILQ_REMOVE(&pool->list, oldpa, entries);
2826 			switch (oldpa->addr.type) {
2827 			case PF_ADDR_DYNIFTL:
2828 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
2829 				break;
2830 			case PF_ADDR_TABLE:
2831 				pfr_detach_table(oldpa->addr.p.tbl);
2832 				break;
2833 			}
2834 			if (oldpa->kif)
2835 				pfi_kif_unref(oldpa->kif);
2836 			free(oldpa, M_PFRULE);
2837 		} else {
2838 			if (oldpa == NULL)
2839 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2840 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
2841 			    pca->action == PF_CHANGE_ADD_BEFORE)
2842 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2843 			else
2844 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
2845 				    newpa, entries);
2846 		}
2847 
2848 		pool->cur = TAILQ_FIRST(&pool->list);
2849 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
2850 		PF_RULES_WUNLOCK();
2851 		break;
2852 
2853 #undef ERROUT
2854 DIOCCHANGEADDR_error:
2855 		if (newpa != NULL) {
2856 			if (newpa->kif)
2857 				pfi_kif_unref(newpa->kif);
2858 			free(newpa, M_PFRULE);
2859 		}
2860 		PF_RULES_WUNLOCK();
2861 		if (kif != NULL)
2862 			free(kif, PFI_MTYPE);
2863 		break;
2864 	}
2865 
2866 	case DIOCGETRULESETS: {
2867 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2868 		struct pf_ruleset	*ruleset;
2869 		struct pf_anchor	*anchor;
2870 
2871 		PF_RULES_RLOCK();
2872 		pr->path[sizeof(pr->path) - 1] = 0;
2873 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2874 			PF_RULES_RUNLOCK();
2875 			error = ENOENT;
2876 			break;
2877 		}
2878 		pr->nr = 0;
2879 		if (ruleset->anchor == NULL) {
2880 			/* XXX kludge for pf_main_ruleset */
2881 			RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2882 				if (anchor->parent == NULL)
2883 					pr->nr++;
2884 		} else {
2885 			RB_FOREACH(anchor, pf_anchor_node,
2886 			    &ruleset->anchor->children)
2887 				pr->nr++;
2888 		}
2889 		PF_RULES_RUNLOCK();
2890 		break;
2891 	}
2892 
2893 	case DIOCGETRULESET: {
2894 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2895 		struct pf_ruleset	*ruleset;
2896 		struct pf_anchor	*anchor;
2897 		u_int32_t		 nr = 0;
2898 
2899 		PF_RULES_RLOCK();
2900 		pr->path[sizeof(pr->path) - 1] = 0;
2901 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2902 			PF_RULES_RUNLOCK();
2903 			error = ENOENT;
2904 			break;
2905 		}
2906 		pr->name[0] = 0;
2907 		if (ruleset->anchor == NULL) {
2908 			/* XXX kludge for pf_main_ruleset */
2909 			RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2910 				if (anchor->parent == NULL && nr++ == pr->nr) {
2911 					strlcpy(pr->name, anchor->name,
2912 					    sizeof(pr->name));
2913 					break;
2914 				}
2915 		} else {
2916 			RB_FOREACH(anchor, pf_anchor_node,
2917 			    &ruleset->anchor->children)
2918 				if (nr++ == pr->nr) {
2919 					strlcpy(pr->name, anchor->name,
2920 					    sizeof(pr->name));
2921 					break;
2922 				}
2923 		}
2924 		if (!pr->name[0])
2925 			error = EBUSY;
2926 		PF_RULES_RUNLOCK();
2927 		break;
2928 	}
2929 
2930 	case DIOCRCLRTABLES: {
2931 		struct pfioc_table *io = (struct pfioc_table *)addr;
2932 
2933 		if (io->pfrio_esize != 0) {
2934 			error = ENODEV;
2935 			break;
2936 		}
2937 		PF_RULES_WLOCK();
2938 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2939 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2940 		PF_RULES_WUNLOCK();
2941 		break;
2942 	}
2943 
2944 	case DIOCRADDTABLES: {
2945 		struct pfioc_table *io = (struct pfioc_table *)addr;
2946 		struct pfr_table *pfrts;
2947 		size_t totlen;
2948 
2949 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2950 			error = ENODEV;
2951 			break;
2952 		}
2953 
2954 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
2955 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
2956 			error = ENOMEM;
2957 			break;
2958 		}
2959 
2960 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2961 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
2962 		    M_TEMP, M_WAITOK);
2963 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2964 		if (error) {
2965 			free(pfrts, M_TEMP);
2966 			break;
2967 		}
2968 		PF_RULES_WLOCK();
2969 		error = pfr_add_tables(pfrts, io->pfrio_size,
2970 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2971 		PF_RULES_WUNLOCK();
2972 		free(pfrts, M_TEMP);
2973 		break;
2974 	}
2975 
2976 	case DIOCRDELTABLES: {
2977 		struct pfioc_table *io = (struct pfioc_table *)addr;
2978 		struct pfr_table *pfrts;
2979 		size_t totlen;
2980 
2981 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2982 			error = ENODEV;
2983 			break;
2984 		}
2985 
2986 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
2987 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
2988 			error = ENOMEM;
2989 			break;
2990 		}
2991 
2992 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2993 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
2994 		    M_TEMP, M_WAITOK);
2995 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2996 		if (error) {
2997 			free(pfrts, M_TEMP);
2998 			break;
2999 		}
3000 		PF_RULES_WLOCK();
3001 		error = pfr_del_tables(pfrts, io->pfrio_size,
3002 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3003 		PF_RULES_WUNLOCK();
3004 		free(pfrts, M_TEMP);
3005 		break;
3006 	}
3007 
3008 	case DIOCRGETTABLES: {
3009 		struct pfioc_table *io = (struct pfioc_table *)addr;
3010 		struct pfr_table *pfrts;
3011 		size_t totlen;
3012 		int n;
3013 
3014 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3015 			error = ENODEV;
3016 			break;
3017 		}
3018 		PF_RULES_RLOCK();
3019 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3020 		if (n < 0) {
3021 			PF_RULES_RUNLOCK();
3022 			error = EINVAL;
3023 			break;
3024 		}
3025 		io->pfrio_size = min(io->pfrio_size, n);
3026 
3027 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3028 
3029 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3030 		    M_TEMP, M_NOWAIT);
3031 		if (pfrts == NULL) {
3032 			error = ENOMEM;
3033 			PF_RULES_RUNLOCK();
3034 			break;
3035 		}
3036 		error = pfr_get_tables(&io->pfrio_table, pfrts,
3037 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3038 		PF_RULES_RUNLOCK();
3039 		if (error == 0)
3040 			error = copyout(pfrts, io->pfrio_buffer, totlen);
3041 		free(pfrts, M_TEMP);
3042 		break;
3043 	}
3044 
3045 	case DIOCRGETTSTATS: {
3046 		struct pfioc_table *io = (struct pfioc_table *)addr;
3047 		struct pfr_tstats *pfrtstats;
3048 		size_t totlen;
3049 		int n;
3050 
3051 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
3052 			error = ENODEV;
3053 			break;
3054 		}
3055 		PF_RULES_WLOCK();
3056 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3057 		if (n < 0) {
3058 			PF_RULES_WUNLOCK();
3059 			error = EINVAL;
3060 			break;
3061 		}
3062 		io->pfrio_size = min(io->pfrio_size, n);
3063 
3064 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
3065 		pfrtstats = mallocarray(io->pfrio_size,
3066 		    sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT);
3067 		if (pfrtstats == NULL) {
3068 			error = ENOMEM;
3069 			PF_RULES_WUNLOCK();
3070 			break;
3071 		}
3072 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
3073 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3074 		PF_RULES_WUNLOCK();
3075 		if (error == 0)
3076 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
3077 		free(pfrtstats, M_TEMP);
3078 		break;
3079 	}
3080 
3081 	case DIOCRCLRTSTATS: {
3082 		struct pfioc_table *io = (struct pfioc_table *)addr;
3083 		struct pfr_table *pfrts;
3084 		size_t totlen;
3085 		int n;
3086 
3087 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3088 			error = ENODEV;
3089 			break;
3090 		}
3091 
3092 		PF_RULES_WLOCK();
3093 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3094 		if (n < 0) {
3095 			PF_RULES_WUNLOCK();
3096 			error = EINVAL;
3097 			break;
3098 		}
3099 		io->pfrio_size = min(io->pfrio_size, n);
3100 
3101 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3102 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3103 		    M_TEMP, M_NOWAIT);
3104 		if (pfrts == NULL) {
3105 			error = ENOMEM;
3106 			PF_RULES_WUNLOCK();
3107 			break;
3108 		}
3109 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3110 		if (error) {
3111 			free(pfrts, M_TEMP);
3112 			PF_RULES_WUNLOCK();
3113 			break;
3114 		}
3115 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
3116 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3117 		PF_RULES_WUNLOCK();
3118 		free(pfrts, M_TEMP);
3119 		break;
3120 	}
3121 
3122 	case DIOCRSETTFLAGS: {
3123 		struct pfioc_table *io = (struct pfioc_table *)addr;
3124 		struct pfr_table *pfrts;
3125 		size_t totlen;
3126 		int n;
3127 
3128 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3129 			error = ENODEV;
3130 			break;
3131 		}
3132 
3133 		PF_RULES_RLOCK();
3134 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3135 		if (n < 0) {
3136 			PF_RULES_RUNLOCK();
3137 			error = EINVAL;
3138 			break;
3139 		}
3140 
3141 		io->pfrio_size = min(io->pfrio_size, n);
3142 		PF_RULES_RUNLOCK();
3143 
3144 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3145 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3146 		    M_TEMP, M_WAITOK);
3147 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3148 		if (error) {
3149 			free(pfrts, M_TEMP);
3150 			break;
3151 		}
3152 		PF_RULES_WLOCK();
3153 		error = pfr_set_tflags(pfrts, io->pfrio_size,
3154 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
3155 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3156 		PF_RULES_WUNLOCK();
3157 		free(pfrts, M_TEMP);
3158 		break;
3159 	}
3160 
3161 	case DIOCRCLRADDRS: {
3162 		struct pfioc_table *io = (struct pfioc_table *)addr;
3163 
3164 		if (io->pfrio_esize != 0) {
3165 			error = ENODEV;
3166 			break;
3167 		}
3168 		PF_RULES_WLOCK();
3169 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
3170 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
3171 		PF_RULES_WUNLOCK();
3172 		break;
3173 	}
3174 
3175 	case DIOCRADDADDRS: {
3176 		struct pfioc_table *io = (struct pfioc_table *)addr;
3177 		struct pfr_addr *pfras;
3178 		size_t totlen;
3179 
3180 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3181 			error = ENODEV;
3182 			break;
3183 		}
3184 		if (io->pfrio_size < 0 ||
3185 		    io->pfrio_size > pf_ioctl_maxcount ||
3186 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3187 			error = EINVAL;
3188 			break;
3189 		}
3190 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3191 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3192 		    M_TEMP, M_NOWAIT);
3193 		if (! pfras) {
3194 			error = ENOMEM;
3195 			break;
3196 		}
3197 		error = copyin(io->pfrio_buffer, pfras, totlen);
3198 		if (error) {
3199 			free(pfras, M_TEMP);
3200 			break;
3201 		}
3202 		PF_RULES_WLOCK();
3203 		error = pfr_add_addrs(&io->pfrio_table, pfras,
3204 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
3205 		    PFR_FLAG_USERIOCTL);
3206 		PF_RULES_WUNLOCK();
3207 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3208 			error = copyout(pfras, io->pfrio_buffer, totlen);
3209 		free(pfras, M_TEMP);
3210 		break;
3211 	}
3212 
3213 	case DIOCRDELADDRS: {
3214 		struct pfioc_table *io = (struct pfioc_table *)addr;
3215 		struct pfr_addr *pfras;
3216 		size_t totlen;
3217 
3218 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3219 			error = ENODEV;
3220 			break;
3221 		}
3222 		if (io->pfrio_size < 0 ||
3223 		    io->pfrio_size > pf_ioctl_maxcount ||
3224 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3225 			error = EINVAL;
3226 			break;
3227 		}
3228 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3229 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3230 		    M_TEMP, M_NOWAIT);
3231 		if (! pfras) {
3232 			error = ENOMEM;
3233 			break;
3234 		}
3235 		error = copyin(io->pfrio_buffer, pfras, totlen);
3236 		if (error) {
3237 			free(pfras, M_TEMP);
3238 			break;
3239 		}
3240 		PF_RULES_WLOCK();
3241 		error = pfr_del_addrs(&io->pfrio_table, pfras,
3242 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
3243 		    PFR_FLAG_USERIOCTL);
3244 		PF_RULES_WUNLOCK();
3245 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3246 			error = copyout(pfras, io->pfrio_buffer, totlen);
3247 		free(pfras, M_TEMP);
3248 		break;
3249 	}
3250 
3251 	case DIOCRSETADDRS: {
3252 		struct pfioc_table *io = (struct pfioc_table *)addr;
3253 		struct pfr_addr *pfras;
3254 		size_t totlen, count;
3255 
3256 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3257 			error = ENODEV;
3258 			break;
3259 		}
3260 		if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
3261 			error = EINVAL;
3262 			break;
3263 		}
3264 		count = max(io->pfrio_size, io->pfrio_size2);
3265 		if (count > pf_ioctl_maxcount ||
3266 		    WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
3267 			error = EINVAL;
3268 			break;
3269 		}
3270 		totlen = count * sizeof(struct pfr_addr);
3271 		pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
3272 		    M_NOWAIT);
3273 		if (! pfras) {
3274 			error = ENOMEM;
3275 			break;
3276 		}
3277 		error = copyin(io->pfrio_buffer, pfras, totlen);
3278 		if (error) {
3279 			free(pfras, M_TEMP);
3280 			break;
3281 		}
3282 		PF_RULES_WLOCK();
3283 		error = pfr_set_addrs(&io->pfrio_table, pfras,
3284 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
3285 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
3286 		    PFR_FLAG_USERIOCTL, 0);
3287 		PF_RULES_WUNLOCK();
3288 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3289 			error = copyout(pfras, io->pfrio_buffer, totlen);
3290 		free(pfras, M_TEMP);
3291 		break;
3292 	}
3293 
3294 	case DIOCRGETADDRS: {
3295 		struct pfioc_table *io = (struct pfioc_table *)addr;
3296 		struct pfr_addr *pfras;
3297 		size_t totlen;
3298 
3299 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3300 			error = ENODEV;
3301 			break;
3302 		}
3303 		if (io->pfrio_size < 0 ||
3304 		    io->pfrio_size > pf_ioctl_maxcount ||
3305 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3306 			error = EINVAL;
3307 			break;
3308 		}
3309 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3310 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3311 		    M_TEMP, M_NOWAIT);
3312 		if (! pfras) {
3313 			error = ENOMEM;
3314 			break;
3315 		}
3316 		PF_RULES_RLOCK();
3317 		error = pfr_get_addrs(&io->pfrio_table, pfras,
3318 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3319 		PF_RULES_RUNLOCK();
3320 		if (error == 0)
3321 			error = copyout(pfras, io->pfrio_buffer, totlen);
3322 		free(pfras, M_TEMP);
3323 		break;
3324 	}
3325 
3326 	case DIOCRGETASTATS: {
3327 		struct pfioc_table *io = (struct pfioc_table *)addr;
3328 		struct pfr_astats *pfrastats;
3329 		size_t totlen;
3330 
3331 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
3332 			error = ENODEV;
3333 			break;
3334 		}
3335 		if (io->pfrio_size < 0 ||
3336 		    io->pfrio_size > pf_ioctl_maxcount ||
3337 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
3338 			error = EINVAL;
3339 			break;
3340 		}
3341 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
3342 		pfrastats = mallocarray(io->pfrio_size,
3343 		    sizeof(struct pfr_astats), M_TEMP, M_NOWAIT);
3344 		if (! pfrastats) {
3345 			error = ENOMEM;
3346 			break;
3347 		}
3348 		PF_RULES_RLOCK();
3349 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
3350 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3351 		PF_RULES_RUNLOCK();
3352 		if (error == 0)
3353 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
3354 		free(pfrastats, M_TEMP);
3355 		break;
3356 	}
3357 
3358 	case DIOCRCLRASTATS: {
3359 		struct pfioc_table *io = (struct pfioc_table *)addr;
3360 		struct pfr_addr *pfras;
3361 		size_t totlen;
3362 
3363 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3364 			error = ENODEV;
3365 			break;
3366 		}
3367 		if (io->pfrio_size < 0 ||
3368 		    io->pfrio_size > pf_ioctl_maxcount ||
3369 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3370 			error = EINVAL;
3371 			break;
3372 		}
3373 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3374 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3375 		    M_TEMP, M_NOWAIT);
3376 		if (! pfras) {
3377 			error = ENOMEM;
3378 			break;
3379 		}
3380 		error = copyin(io->pfrio_buffer, pfras, totlen);
3381 		if (error) {
3382 			free(pfras, M_TEMP);
3383 			break;
3384 		}
3385 		PF_RULES_WLOCK();
3386 		error = pfr_clr_astats(&io->pfrio_table, pfras,
3387 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
3388 		    PFR_FLAG_USERIOCTL);
3389 		PF_RULES_WUNLOCK();
3390 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3391 			error = copyout(pfras, io->pfrio_buffer, totlen);
3392 		free(pfras, M_TEMP);
3393 		break;
3394 	}
3395 
3396 	case DIOCRTSTADDRS: {
3397 		struct pfioc_table *io = (struct pfioc_table *)addr;
3398 		struct pfr_addr *pfras;
3399 		size_t totlen;
3400 
3401 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3402 			error = ENODEV;
3403 			break;
3404 		}
3405 		if (io->pfrio_size < 0 ||
3406 		    io->pfrio_size > pf_ioctl_maxcount ||
3407 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3408 			error = EINVAL;
3409 			break;
3410 		}
3411 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3412 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3413 		    M_TEMP, M_NOWAIT);
3414 		if (! pfras) {
3415 			error = ENOMEM;
3416 			break;
3417 		}
3418 		error = copyin(io->pfrio_buffer, pfras, totlen);
3419 		if (error) {
3420 			free(pfras, M_TEMP);
3421 			break;
3422 		}
3423 		PF_RULES_RLOCK();
3424 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
3425 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
3426 		    PFR_FLAG_USERIOCTL);
3427 		PF_RULES_RUNLOCK();
3428 		if (error == 0)
3429 			error = copyout(pfras, io->pfrio_buffer, totlen);
3430 		free(pfras, M_TEMP);
3431 		break;
3432 	}
3433 
3434 	case DIOCRINADEFINE: {
3435 		struct pfioc_table *io = (struct pfioc_table *)addr;
3436 		struct pfr_addr *pfras;
3437 		size_t totlen;
3438 
3439 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3440 			error = ENODEV;
3441 			break;
3442 		}
3443 		if (io->pfrio_size < 0 ||
3444 		    io->pfrio_size > pf_ioctl_maxcount ||
3445 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3446 			error = EINVAL;
3447 			break;
3448 		}
3449 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3450 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3451 		    M_TEMP, M_NOWAIT);
3452 		if (! pfras) {
3453 			error = ENOMEM;
3454 			break;
3455 		}
3456 		error = copyin(io->pfrio_buffer, pfras, totlen);
3457 		if (error) {
3458 			free(pfras, M_TEMP);
3459 			break;
3460 		}
3461 		PF_RULES_WLOCK();
3462 		error = pfr_ina_define(&io->pfrio_table, pfras,
3463 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
3464 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3465 		PF_RULES_WUNLOCK();
3466 		free(pfras, M_TEMP);
3467 		break;
3468 	}
3469 
3470 	case DIOCOSFPADD: {
3471 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
3472 		PF_RULES_WLOCK();
3473 		error = pf_osfp_add(io);
3474 		PF_RULES_WUNLOCK();
3475 		break;
3476 	}
3477 
3478 	case DIOCOSFPGET: {
3479 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
3480 		PF_RULES_RLOCK();
3481 		error = pf_osfp_get(io);
3482 		PF_RULES_RUNLOCK();
3483 		break;
3484 	}
3485 
3486 	case DIOCXBEGIN: {
3487 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3488 		struct pfioc_trans_e	*ioes, *ioe;
3489 		size_t			 totlen;
3490 		int			 i;
3491 
3492 		if (io->esize != sizeof(*ioe)) {
3493 			error = ENODEV;
3494 			break;
3495 		}
3496 		if (io->size < 0 ||
3497 		    io->size > pf_ioctl_maxcount ||
3498 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3499 			error = EINVAL;
3500 			break;
3501 		}
3502 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3503 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3504 		    M_TEMP, M_NOWAIT);
3505 		if (! ioes) {
3506 			error = ENOMEM;
3507 			break;
3508 		}
3509 		error = copyin(io->array, ioes, totlen);
3510 		if (error) {
3511 			free(ioes, M_TEMP);
3512 			break;
3513 		}
3514 		PF_RULES_WLOCK();
3515 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3516 			switch (ioe->rs_num) {
3517 #ifdef ALTQ
3518 			case PF_RULESET_ALTQ:
3519 				if (ioe->anchor[0]) {
3520 					PF_RULES_WUNLOCK();
3521 					free(ioes, M_TEMP);
3522 					error = EINVAL;
3523 					goto fail;
3524 				}
3525 				if ((error = pf_begin_altq(&ioe->ticket))) {
3526 					PF_RULES_WUNLOCK();
3527 					free(ioes, M_TEMP);
3528 					goto fail;
3529 				}
3530 				break;
3531 #endif /* ALTQ */
3532 			case PF_RULESET_TABLE:
3533 			    {
3534 				struct pfr_table table;
3535 
3536 				bzero(&table, sizeof(table));
3537 				strlcpy(table.pfrt_anchor, ioe->anchor,
3538 				    sizeof(table.pfrt_anchor));
3539 				if ((error = pfr_ina_begin(&table,
3540 				    &ioe->ticket, NULL, 0))) {
3541 					PF_RULES_WUNLOCK();
3542 					free(ioes, M_TEMP);
3543 					goto fail;
3544 				}
3545 				break;
3546 			    }
3547 			default:
3548 				if ((error = pf_begin_rules(&ioe->ticket,
3549 				    ioe->rs_num, ioe->anchor))) {
3550 					PF_RULES_WUNLOCK();
3551 					free(ioes, M_TEMP);
3552 					goto fail;
3553 				}
3554 				break;
3555 			}
3556 		}
3557 		PF_RULES_WUNLOCK();
3558 		error = copyout(ioes, io->array, totlen);
3559 		free(ioes, M_TEMP);
3560 		break;
3561 	}
3562 
3563 	case DIOCXROLLBACK: {
3564 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3565 		struct pfioc_trans_e	*ioe, *ioes;
3566 		size_t			 totlen;
3567 		int			 i;
3568 
3569 		if (io->esize != sizeof(*ioe)) {
3570 			error = ENODEV;
3571 			break;
3572 		}
3573 		if (io->size < 0 ||
3574 		    io->size > pf_ioctl_maxcount ||
3575 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3576 			error = EINVAL;
3577 			break;
3578 		}
3579 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3580 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3581 		    M_TEMP, M_NOWAIT);
3582 		if (! ioes) {
3583 			error = ENOMEM;
3584 			break;
3585 		}
3586 		error = copyin(io->array, ioes, totlen);
3587 		if (error) {
3588 			free(ioes, M_TEMP);
3589 			break;
3590 		}
3591 		PF_RULES_WLOCK();
3592 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3593 			switch (ioe->rs_num) {
3594 #ifdef ALTQ
3595 			case PF_RULESET_ALTQ:
3596 				if (ioe->anchor[0]) {
3597 					PF_RULES_WUNLOCK();
3598 					free(ioes, M_TEMP);
3599 					error = EINVAL;
3600 					goto fail;
3601 				}
3602 				if ((error = pf_rollback_altq(ioe->ticket))) {
3603 					PF_RULES_WUNLOCK();
3604 					free(ioes, M_TEMP);
3605 					goto fail; /* really bad */
3606 				}
3607 				break;
3608 #endif /* ALTQ */
3609 			case PF_RULESET_TABLE:
3610 			    {
3611 				struct pfr_table table;
3612 
3613 				bzero(&table, sizeof(table));
3614 				strlcpy(table.pfrt_anchor, ioe->anchor,
3615 				    sizeof(table.pfrt_anchor));
3616 				if ((error = pfr_ina_rollback(&table,
3617 				    ioe->ticket, NULL, 0))) {
3618 					PF_RULES_WUNLOCK();
3619 					free(ioes, M_TEMP);
3620 					goto fail; /* really bad */
3621 				}
3622 				break;
3623 			    }
3624 			default:
3625 				if ((error = pf_rollback_rules(ioe->ticket,
3626 				    ioe->rs_num, ioe->anchor))) {
3627 					PF_RULES_WUNLOCK();
3628 					free(ioes, M_TEMP);
3629 					goto fail; /* really bad */
3630 				}
3631 				break;
3632 			}
3633 		}
3634 		PF_RULES_WUNLOCK();
3635 		free(ioes, M_TEMP);
3636 		break;
3637 	}
3638 
3639 	case DIOCXCOMMIT: {
3640 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3641 		struct pfioc_trans_e	*ioe, *ioes;
3642 		struct pf_ruleset	*rs;
3643 		size_t			 totlen;
3644 		int			 i;
3645 
3646 		if (io->esize != sizeof(*ioe)) {
3647 			error = ENODEV;
3648 			break;
3649 		}
3650 
3651 		if (io->size < 0 ||
3652 		    io->size > pf_ioctl_maxcount ||
3653 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3654 			error = EINVAL;
3655 			break;
3656 		}
3657 
3658 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3659 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3660 		    M_TEMP, M_NOWAIT);
3661 		if (ioes == NULL) {
3662 			error = ENOMEM;
3663 			break;
3664 		}
3665 		error = copyin(io->array, ioes, totlen);
3666 		if (error) {
3667 			free(ioes, M_TEMP);
3668 			break;
3669 		}
3670 		PF_RULES_WLOCK();
3671 		/* First makes sure everything will succeed. */
3672 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3673 			switch (ioe->rs_num) {
3674 #ifdef ALTQ
3675 			case PF_RULESET_ALTQ:
3676 				if (ioe->anchor[0]) {
3677 					PF_RULES_WUNLOCK();
3678 					free(ioes, M_TEMP);
3679 					error = EINVAL;
3680 					goto fail;
3681 				}
3682 				if (!V_altqs_inactive_open || ioe->ticket !=
3683 				    V_ticket_altqs_inactive) {
3684 					PF_RULES_WUNLOCK();
3685 					free(ioes, M_TEMP);
3686 					error = EBUSY;
3687 					goto fail;
3688 				}
3689 				break;
3690 #endif /* ALTQ */
3691 			case PF_RULESET_TABLE:
3692 				rs = pf_find_ruleset(ioe->anchor);
3693 				if (rs == NULL || !rs->topen || ioe->ticket !=
3694 				    rs->tticket) {
3695 					PF_RULES_WUNLOCK();
3696 					free(ioes, M_TEMP);
3697 					error = EBUSY;
3698 					goto fail;
3699 				}
3700 				break;
3701 			default:
3702 				if (ioe->rs_num < 0 || ioe->rs_num >=
3703 				    PF_RULESET_MAX) {
3704 					PF_RULES_WUNLOCK();
3705 					free(ioes, M_TEMP);
3706 					error = EINVAL;
3707 					goto fail;
3708 				}
3709 				rs = pf_find_ruleset(ioe->anchor);
3710 				if (rs == NULL ||
3711 				    !rs->rules[ioe->rs_num].inactive.open ||
3712 				    rs->rules[ioe->rs_num].inactive.ticket !=
3713 				    ioe->ticket) {
3714 					PF_RULES_WUNLOCK();
3715 					free(ioes, M_TEMP);
3716 					error = EBUSY;
3717 					goto fail;
3718 				}
3719 				break;
3720 			}
3721 		}
3722 		/* Now do the commit - no errors should happen here. */
3723 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3724 			switch (ioe->rs_num) {
3725 #ifdef ALTQ
3726 			case PF_RULESET_ALTQ:
3727 				if ((error = pf_commit_altq(ioe->ticket))) {
3728 					PF_RULES_WUNLOCK();
3729 					free(ioes, M_TEMP);
3730 					goto fail; /* really bad */
3731 				}
3732 				break;
3733 #endif /* ALTQ */
3734 			case PF_RULESET_TABLE:
3735 			    {
3736 				struct pfr_table table;
3737 
3738 				bzero(&table, sizeof(table));
3739 				strlcpy(table.pfrt_anchor, ioe->anchor,
3740 				    sizeof(table.pfrt_anchor));
3741 				if ((error = pfr_ina_commit(&table,
3742 				    ioe->ticket, NULL, NULL, 0))) {
3743 					PF_RULES_WUNLOCK();
3744 					free(ioes, M_TEMP);
3745 					goto fail; /* really bad */
3746 				}
3747 				break;
3748 			    }
3749 			default:
3750 				if ((error = pf_commit_rules(ioe->ticket,
3751 				    ioe->rs_num, ioe->anchor))) {
3752 					PF_RULES_WUNLOCK();
3753 					free(ioes, M_TEMP);
3754 					goto fail; /* really bad */
3755 				}
3756 				break;
3757 			}
3758 		}
3759 		PF_RULES_WUNLOCK();
3760 		free(ioes, M_TEMP);
3761 		break;
3762 	}
3763 
3764 	case DIOCGETSRCNODES: {
3765 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
3766 		struct pf_srchash	*sh;
3767 		struct pf_src_node	*n, *p, *pstore;
3768 		uint32_t		 i, nr = 0;
3769 
3770 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3771 				i++, sh++) {
3772 			PF_HASHROW_LOCK(sh);
3773 			LIST_FOREACH(n, &sh->nodes, entry)
3774 				nr++;
3775 			PF_HASHROW_UNLOCK(sh);
3776 		}
3777 
3778 		psn->psn_len = min(psn->psn_len,
3779 		    sizeof(struct pf_src_node) * nr);
3780 
3781 		if (psn->psn_len == 0) {
3782 			psn->psn_len = sizeof(struct pf_src_node) * nr;
3783 			break;
3784 		}
3785 
3786 		nr = 0;
3787 
3788 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
3789 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3790 		    i++, sh++) {
3791 		    PF_HASHROW_LOCK(sh);
3792 		    LIST_FOREACH(n, &sh->nodes, entry) {
3793 			int	secs = time_uptime, diff;
3794 
3795 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
3796 				break;
3797 
3798 			bcopy(n, p, sizeof(struct pf_src_node));
3799 			if (n->rule.ptr != NULL)
3800 				p->rule.nr = n->rule.ptr->nr;
3801 			p->creation = secs - p->creation;
3802 			if (p->expire > secs)
3803 				p->expire -= secs;
3804 			else
3805 				p->expire = 0;
3806 
3807 			/* Adjust the connection rate estimate. */
3808 			diff = secs - n->conn_rate.last;
3809 			if (diff >= n->conn_rate.seconds)
3810 				p->conn_rate.count = 0;
3811 			else
3812 				p->conn_rate.count -=
3813 				    n->conn_rate.count * diff /
3814 				    n->conn_rate.seconds;
3815 			p++;
3816 			nr++;
3817 		    }
3818 		    PF_HASHROW_UNLOCK(sh);
3819 		}
3820 		error = copyout(pstore, psn->psn_src_nodes,
3821 		    sizeof(struct pf_src_node) * nr);
3822 		if (error) {
3823 			free(pstore, M_TEMP);
3824 			break;
3825 		}
3826 		psn->psn_len = sizeof(struct pf_src_node) * nr;
3827 		free(pstore, M_TEMP);
3828 		break;
3829 	}
3830 
3831 	case DIOCCLRSRCNODES: {
3832 
3833 		pf_clear_srcnodes(NULL);
3834 		pf_purge_expired_src_nodes();
3835 		break;
3836 	}
3837 
3838 	case DIOCKILLSRCNODES:
3839 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
3840 		break;
3841 
3842 	case DIOCSETHOSTID: {
3843 		u_int32_t	*hostid = (u_int32_t *)addr;
3844 
3845 		PF_RULES_WLOCK();
3846 		if (*hostid == 0)
3847 			V_pf_status.hostid = arc4random();
3848 		else
3849 			V_pf_status.hostid = *hostid;
3850 		PF_RULES_WUNLOCK();
3851 		break;
3852 	}
3853 
3854 	case DIOCOSFPFLUSH:
3855 		PF_RULES_WLOCK();
3856 		pf_osfp_flush();
3857 		PF_RULES_WUNLOCK();
3858 		break;
3859 
3860 	case DIOCIGETIFACES: {
3861 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3862 		struct pfi_kif *ifstore;
3863 		size_t bufsiz;
3864 
3865 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
3866 			error = ENODEV;
3867 			break;
3868 		}
3869 
3870 		if (io->pfiio_size < 0 ||
3871 		    io->pfiio_size > pf_ioctl_maxcount ||
3872 		    WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
3873 			error = EINVAL;
3874 			break;
3875 		}
3876 
3877 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
3878 		ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
3879 		    M_TEMP, M_NOWAIT);
3880 		if (ifstore == NULL) {
3881 			error = ENOMEM;
3882 			break;
3883 		}
3884 
3885 		PF_RULES_RLOCK();
3886 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
3887 		PF_RULES_RUNLOCK();
3888 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
3889 		free(ifstore, M_TEMP);
3890 		break;
3891 	}
3892 
3893 	case DIOCSETIFFLAG: {
3894 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3895 
3896 		PF_RULES_WLOCK();
3897 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3898 		PF_RULES_WUNLOCK();
3899 		break;
3900 	}
3901 
3902 	case DIOCCLRIFFLAG: {
3903 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3904 
3905 		PF_RULES_WLOCK();
3906 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3907 		PF_RULES_WUNLOCK();
3908 		break;
3909 	}
3910 
3911 	default:
3912 		error = ENODEV;
3913 		break;
3914 	}
3915 fail:
3916 	if (sx_xlocked(&pf_ioctl_lock))
3917 		sx_xunlock(&pf_ioctl_lock);
3918 	CURVNET_RESTORE();
3919 
3920 	return (error);
3921 }
3922 
3923 void
3924 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
3925 {
3926 	bzero(sp, sizeof(struct pfsync_state));
3927 
3928 	/* copy from state key */
3929 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
3930 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
3931 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
3932 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
3933 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
3934 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
3935 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
3936 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
3937 	sp->proto = st->key[PF_SK_WIRE]->proto;
3938 	sp->af = st->key[PF_SK_WIRE]->af;
3939 
3940 	/* copy from state */
3941 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
3942 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
3943 	sp->creation = htonl(time_uptime - st->creation);
3944 	sp->expire = pf_state_expires(st);
3945 	if (sp->expire <= time_uptime)
3946 		sp->expire = htonl(0);
3947 	else
3948 		sp->expire = htonl(sp->expire - time_uptime);
3949 
3950 	sp->direction = st->direction;
3951 	sp->log = st->log;
3952 	sp->timeout = st->timeout;
3953 	sp->state_flags = st->state_flags;
3954 	if (st->src_node)
3955 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
3956 	if (st->nat_src_node)
3957 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
3958 
3959 	sp->id = st->id;
3960 	sp->creatorid = st->creatorid;
3961 	pf_state_peer_hton(&st->src, &sp->src);
3962 	pf_state_peer_hton(&st->dst, &sp->dst);
3963 
3964 	if (st->rule.ptr == NULL)
3965 		sp->rule = htonl(-1);
3966 	else
3967 		sp->rule = htonl(st->rule.ptr->nr);
3968 	if (st->anchor.ptr == NULL)
3969 		sp->anchor = htonl(-1);
3970 	else
3971 		sp->anchor = htonl(st->anchor.ptr->nr);
3972 	if (st->nat_rule.ptr == NULL)
3973 		sp->nat_rule = htonl(-1);
3974 	else
3975 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
3976 
3977 	pf_state_counter_hton(st->packets[0], sp->packets[0]);
3978 	pf_state_counter_hton(st->packets[1], sp->packets[1]);
3979 	pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
3980 	pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
3981 
3982 }
3983 
3984 static void
3985 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
3986 {
3987 	struct pfr_ktable *kt;
3988 
3989 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
3990 
3991 	kt = aw->p.tbl;
3992 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
3993 		kt = kt->pfrkt_root;
3994 	aw->p.tbl = NULL;
3995 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
3996 		kt->pfrkt_cnt : -1;
3997 }
3998 
3999 /*
4000  * XXX - Check for version missmatch!!!
4001  */
4002 static void
4003 pf_clear_states(void)
4004 {
4005 	struct pf_state	*s;
4006 	u_int i;
4007 
4008 	for (i = 0; i <= pf_hashmask; i++) {
4009 		struct pf_idhash *ih = &V_pf_idhash[i];
4010 relock:
4011 		PF_HASHROW_LOCK(ih);
4012 		LIST_FOREACH(s, &ih->states, entry) {
4013 			s->timeout = PFTM_PURGE;
4014 			/* Don't send out individual delete messages. */
4015 			s->state_flags |= PFSTATE_NOSYNC;
4016 			pf_unlink_state(s, PF_ENTER_LOCKED);
4017 			goto relock;
4018 		}
4019 		PF_HASHROW_UNLOCK(ih);
4020 	}
4021 }
4022 
4023 static int
4024 pf_clear_tables(void)
4025 {
4026 	struct pfioc_table io;
4027 	int error;
4028 
4029 	bzero(&io, sizeof(io));
4030 
4031 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
4032 	    io.pfrio_flags);
4033 
4034 	return (error);
4035 }
4036 
4037 static void
4038 pf_clear_srcnodes(struct pf_src_node *n)
4039 {
4040 	struct pf_state *s;
4041 	int i;
4042 
4043 	for (i = 0; i <= pf_hashmask; i++) {
4044 		struct pf_idhash *ih = &V_pf_idhash[i];
4045 
4046 		PF_HASHROW_LOCK(ih);
4047 		LIST_FOREACH(s, &ih->states, entry) {
4048 			if (n == NULL || n == s->src_node)
4049 				s->src_node = NULL;
4050 			if (n == NULL || n == s->nat_src_node)
4051 				s->nat_src_node = NULL;
4052 		}
4053 		PF_HASHROW_UNLOCK(ih);
4054 	}
4055 
4056 	if (n == NULL) {
4057 		struct pf_srchash *sh;
4058 
4059 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4060 		    i++, sh++) {
4061 			PF_HASHROW_LOCK(sh);
4062 			LIST_FOREACH(n, &sh->nodes, entry) {
4063 				n->expire = 1;
4064 				n->states = 0;
4065 			}
4066 			PF_HASHROW_UNLOCK(sh);
4067 		}
4068 	} else {
4069 		/* XXX: hash slot should already be locked here. */
4070 		n->expire = 1;
4071 		n->states = 0;
4072 	}
4073 }
4074 
4075 static void
4076 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
4077 {
4078 	struct pf_src_node_list	 kill;
4079 
4080 	LIST_INIT(&kill);
4081 	for (int i = 0; i <= pf_srchashmask; i++) {
4082 		struct pf_srchash *sh = &V_pf_srchash[i];
4083 		struct pf_src_node *sn, *tmp;
4084 
4085 		PF_HASHROW_LOCK(sh);
4086 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
4087 			if (PF_MATCHA(psnk->psnk_src.neg,
4088 			      &psnk->psnk_src.addr.v.a.addr,
4089 			      &psnk->psnk_src.addr.v.a.mask,
4090 			      &sn->addr, sn->af) &&
4091 			    PF_MATCHA(psnk->psnk_dst.neg,
4092 			      &psnk->psnk_dst.addr.v.a.addr,
4093 			      &psnk->psnk_dst.addr.v.a.mask,
4094 			      &sn->raddr, sn->af)) {
4095 				pf_unlink_src_node(sn);
4096 				LIST_INSERT_HEAD(&kill, sn, entry);
4097 				sn->expire = 1;
4098 			}
4099 		PF_HASHROW_UNLOCK(sh);
4100 	}
4101 
4102 	for (int i = 0; i <= pf_hashmask; i++) {
4103 		struct pf_idhash *ih = &V_pf_idhash[i];
4104 		struct pf_state *s;
4105 
4106 		PF_HASHROW_LOCK(ih);
4107 		LIST_FOREACH(s, &ih->states, entry) {
4108 			if (s->src_node && s->src_node->expire == 1)
4109 				s->src_node = NULL;
4110 			if (s->nat_src_node && s->nat_src_node->expire == 1)
4111 				s->nat_src_node = NULL;
4112 		}
4113 		PF_HASHROW_UNLOCK(ih);
4114 	}
4115 
4116 	psnk->psnk_killed = pf_free_src_nodes(&kill);
4117 }
4118 
4119 /*
4120  * XXX - Check for version missmatch!!!
4121  */
4122 
4123 /*
4124  * Duplicate pfctl -Fa operation to get rid of as much as we can.
4125  */
4126 static int
4127 shutdown_pf(void)
4128 {
4129 	int error = 0;
4130 	u_int32_t t[5];
4131 	char nn = '\0';
4132 
4133 	do {
4134 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
4135 		    != 0) {
4136 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
4137 			break;
4138 		}
4139 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
4140 		    != 0) {
4141 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
4142 			break;		/* XXX: rollback? */
4143 		}
4144 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
4145 		    != 0) {
4146 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
4147 			break;		/* XXX: rollback? */
4148 		}
4149 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
4150 		    != 0) {
4151 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
4152 			break;		/* XXX: rollback? */
4153 		}
4154 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
4155 		    != 0) {
4156 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
4157 			break;		/* XXX: rollback? */
4158 		}
4159 
4160 		/* XXX: these should always succeed here */
4161 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
4162 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
4163 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
4164 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
4165 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
4166 
4167 		if ((error = pf_clear_tables()) != 0)
4168 			break;
4169 
4170 #ifdef ALTQ
4171 		if ((error = pf_begin_altq(&t[0])) != 0) {
4172 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
4173 			break;
4174 		}
4175 		pf_commit_altq(t[0]);
4176 #endif
4177 
4178 		pf_clear_states();
4179 
4180 		pf_clear_srcnodes(NULL);
4181 
4182 		/* status does not use malloced mem so no need to cleanup */
4183 		/* fingerprints and interfaces have their own cleanup code */
4184 	} while(0);
4185 
4186 	return (error);
4187 }
4188 
4189 static pfil_return_t
4190 pf_check_return(int chk, struct mbuf **m)
4191 {
4192 
4193 	switch (chk) {
4194 	case PF_PASS:
4195 		if (*m == NULL)
4196 			return (PFIL_CONSUMED);
4197 		else
4198 			return (PFIL_PASS);
4199 		break;
4200 	default:
4201 		if (*m != NULL) {
4202 			m_freem(*m);
4203 			*m = NULL;
4204 		}
4205 		return (PFIL_DROPPED);
4206 	}
4207 }
4208 
4209 #ifdef INET
4210 static pfil_return_t
4211 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
4212     void *ruleset __unused, struct inpcb *inp)
4213 {
4214 	int chk;
4215 
4216 	chk = pf_test(PF_IN, flags, ifp, m, inp);
4217 
4218 	return (pf_check_return(chk, m));
4219 }
4220 
4221 static pfil_return_t
4222 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
4223     void *ruleset __unused,  struct inpcb *inp)
4224 {
4225 	int chk;
4226 
4227 	chk = pf_test(PF_OUT, flags, ifp, m, inp);
4228 
4229 	return (pf_check_return(chk, m));
4230 }
4231 #endif
4232 
4233 #ifdef INET6
4234 static pfil_return_t
4235 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
4236     void *ruleset __unused,  struct inpcb *inp)
4237 {
4238 	int chk;
4239 
4240 	/*
4241 	 * In case of loopback traffic IPv6 uses the real interface in
4242 	 * order to support scoped addresses. In order to support stateful
4243 	 * filtering we have change this to lo0 as it is the case in IPv4.
4244 	 */
4245 	CURVNET_SET(ifp->if_vnet);
4246 	chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
4247 	CURVNET_RESTORE();
4248 
4249 	return (pf_check_return(chk, m));
4250 }
4251 
4252 static pfil_return_t
4253 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
4254     void *ruleset __unused,  struct inpcb *inp)
4255 {
4256 	int chk;
4257 
4258 	CURVNET_SET(ifp->if_vnet);
4259 	chk = pf_test6(PF_OUT, flags, ifp, m, inp);
4260 	CURVNET_RESTORE();
4261 
4262 	return (pf_check_return(chk, m));
4263 }
4264 #endif /* INET6 */
4265 
4266 #ifdef INET
4267 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
4268 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
4269 #define	V_pf_ip4_in_hook	VNET(pf_ip4_in_hook)
4270 #define	V_pf_ip4_out_hook	VNET(pf_ip4_out_hook)
4271 #endif
4272 #ifdef INET6
4273 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
4274 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
4275 #define	V_pf_ip6_in_hook	VNET(pf_ip6_in_hook)
4276 #define	V_pf_ip6_out_hook	VNET(pf_ip6_out_hook)
4277 #endif
4278 
4279 static int
4280 hook_pf(void)
4281 {
4282 	struct pfil_hook_args pha;
4283 	struct pfil_link_args pla;
4284 
4285 	if (V_pf_pfil_hooked)
4286 		return (0);
4287 
4288 	pha.pa_version = PFIL_VERSION;
4289 	pha.pa_modname = "pf";
4290 	pha.pa_ruleset = NULL;
4291 
4292 	pla.pa_version = PFIL_VERSION;
4293 
4294 #ifdef INET
4295 	pha.pa_type = PFIL_TYPE_IP4;
4296 	pha.pa_func = pf_check_in;
4297 	pha.pa_flags = PFIL_IN;
4298 	pha.pa_rulname = "default-in";
4299 	V_pf_ip4_in_hook = pfil_add_hook(&pha);
4300 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
4301 	pla.pa_head = V_inet_pfil_head;
4302 	pla.pa_hook = V_pf_ip4_in_hook;
4303 	(void)pfil_link(&pla);
4304 	pha.pa_func = pf_check_out;
4305 	pha.pa_flags = PFIL_OUT;
4306 	pha.pa_rulname = "default-out";
4307 	V_pf_ip4_out_hook = pfil_add_hook(&pha);
4308 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
4309 	pla.pa_head = V_inet_pfil_head;
4310 	pla.pa_hook = V_pf_ip4_out_hook;
4311 	(void)pfil_link(&pla);
4312 #endif
4313 #ifdef INET6
4314 	pha.pa_type = PFIL_TYPE_IP6;
4315 	pha.pa_func = pf_check6_in;
4316 	pha.pa_flags = PFIL_IN;
4317 	pha.pa_rulname = "default-in6";
4318 	V_pf_ip6_in_hook = pfil_add_hook(&pha);
4319 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
4320 	pla.pa_head = V_inet6_pfil_head;
4321 	pla.pa_hook = V_pf_ip6_in_hook;
4322 	(void)pfil_link(&pla);
4323 	pha.pa_func = pf_check6_out;
4324 	pha.pa_rulname = "default-out6";
4325 	pha.pa_flags = PFIL_OUT;
4326 	V_pf_ip6_out_hook = pfil_add_hook(&pha);
4327 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
4328 	pla.pa_head = V_inet6_pfil_head;
4329 	pla.pa_hook = V_pf_ip6_out_hook;
4330 	(void)pfil_link(&pla);
4331 #endif
4332 
4333 	V_pf_pfil_hooked = 1;
4334 	return (0);
4335 }
4336 
4337 static int
4338 dehook_pf(void)
4339 {
4340 
4341 	if (V_pf_pfil_hooked == 0)
4342 		return (0);
4343 
4344 #ifdef INET
4345 	pfil_remove_hook(V_pf_ip4_in_hook);
4346 	pfil_remove_hook(V_pf_ip4_out_hook);
4347 #endif
4348 #ifdef INET6
4349 	pfil_remove_hook(V_pf_ip6_in_hook);
4350 	pfil_remove_hook(V_pf_ip6_out_hook);
4351 #endif
4352 
4353 	V_pf_pfil_hooked = 0;
4354 	return (0);
4355 }
4356 
4357 static void
4358 pf_load_vnet(void)
4359 {
4360 	V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
4361 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
4362 
4363 	pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
4364 	    PF_RULE_TAG_HASH_SIZE_DEFAULT);
4365 #ifdef ALTQ
4366 	pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
4367 	    PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
4368 #endif
4369 
4370 	pfattach_vnet();
4371 	V_pf_vnet_active = 1;
4372 }
4373 
4374 static int
4375 pf_load(void)
4376 {
4377 	int error;
4378 
4379 	rm_init(&pf_rules_lock, "pf rulesets");
4380 	sx_init(&pf_ioctl_lock, "pf ioctl");
4381 	sx_init(&pf_end_lock, "pf end thread");
4382 
4383 	pf_mtag_initialize();
4384 
4385 	pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
4386 	if (pf_dev == NULL)
4387 		return (ENOMEM);
4388 
4389 	pf_end_threads = 0;
4390 	error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
4391 	if (error != 0)
4392 		return (error);
4393 
4394 	pfi_initialize();
4395 
4396 	return (0);
4397 }
4398 
4399 static void
4400 pf_unload_vnet(void)
4401 {
4402 	int error;
4403 
4404 	V_pf_vnet_active = 0;
4405 	V_pf_status.running = 0;
4406 	error = dehook_pf();
4407 	if (error) {
4408 		/*
4409 		 * Should not happen!
4410 		 * XXX Due to error code ESRCH, kldunload will show
4411 		 * a message like 'No such process'.
4412 		 */
4413 		printf("%s : pfil unregisteration fail\n", __FUNCTION__);
4414 		return;
4415 	}
4416 
4417 	PF_RULES_WLOCK();
4418 	shutdown_pf();
4419 	PF_RULES_WUNLOCK();
4420 
4421 	swi_remove(V_pf_swi_cookie);
4422 
4423 	pf_unload_vnet_purge();
4424 
4425 	pf_normalize_cleanup();
4426 	PF_RULES_WLOCK();
4427 	pfi_cleanup_vnet();
4428 	PF_RULES_WUNLOCK();
4429 	pfr_cleanup();
4430 	pf_osfp_flush();
4431 	pf_cleanup();
4432 	if (IS_DEFAULT_VNET(curvnet))
4433 		pf_mtag_cleanup();
4434 
4435 	pf_cleanup_tagset(&V_pf_tags);
4436 #ifdef ALTQ
4437 	pf_cleanup_tagset(&V_pf_qids);
4438 #endif
4439 	uma_zdestroy(V_pf_tag_z);
4440 
4441 	/* Free counters last as we updated them during shutdown. */
4442 	counter_u64_free(V_pf_default_rule.states_cur);
4443 	counter_u64_free(V_pf_default_rule.states_tot);
4444 	counter_u64_free(V_pf_default_rule.src_nodes);
4445 
4446 	for (int i = 0; i < PFRES_MAX; i++)
4447 		counter_u64_free(V_pf_status.counters[i]);
4448 	for (int i = 0; i < LCNT_MAX; i++)
4449 		counter_u64_free(V_pf_status.lcounters[i]);
4450 	for (int i = 0; i < FCNT_MAX; i++)
4451 		counter_u64_free(V_pf_status.fcounters[i]);
4452 	for (int i = 0; i < SCNT_MAX; i++)
4453 		counter_u64_free(V_pf_status.scounters[i]);
4454 }
4455 
4456 static void
4457 pf_unload(void)
4458 {
4459 
4460 	sx_xlock(&pf_end_lock);
4461 	pf_end_threads = 1;
4462 	while (pf_end_threads < 2) {
4463 		wakeup_one(pf_purge_thread);
4464 		sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
4465 	}
4466 	sx_xunlock(&pf_end_lock);
4467 
4468 	if (pf_dev != NULL)
4469 		destroy_dev(pf_dev);
4470 
4471 	pfi_cleanup();
4472 
4473 	rm_destroy(&pf_rules_lock);
4474 	sx_destroy(&pf_ioctl_lock);
4475 	sx_destroy(&pf_end_lock);
4476 }
4477 
4478 static void
4479 vnet_pf_init(void *unused __unused)
4480 {
4481 
4482 	pf_load_vnet();
4483 }
4484 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
4485     vnet_pf_init, NULL);
4486 
4487 static void
4488 vnet_pf_uninit(const void *unused __unused)
4489 {
4490 
4491 	pf_unload_vnet();
4492 }
4493 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
4494 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
4495     vnet_pf_uninit, NULL);
4496 
4497 
4498 static int
4499 pf_modevent(module_t mod, int type, void *data)
4500 {
4501 	int error = 0;
4502 
4503 	switch(type) {
4504 	case MOD_LOAD:
4505 		error = pf_load();
4506 		break;
4507 	case MOD_UNLOAD:
4508 		/* Handled in SYSUNINIT(pf_unload) to ensure it's done after
4509 		 * the vnet_pf_uninit()s */
4510 		break;
4511 	default:
4512 		error = EINVAL;
4513 		break;
4514 	}
4515 
4516 	return (error);
4517 }
4518 
4519 static moduledata_t pf_mod = {
4520 	"pf",
4521 	pf_modevent,
4522 	0
4523 };
4524 
4525 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
4526 MODULE_VERSION(pf, PF_MODVER);
4527