xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision eac7052fdebb90caf2f653e06187bdbca837b9c7)
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 static struct pf_pool *
334 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
335     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
336     u_int8_t check_ticket)
337 {
338 	struct pf_ruleset	*ruleset;
339 	struct pf_rule		*rule;
340 	int			 rs_num;
341 
342 	ruleset = pf_find_ruleset(anchor);
343 	if (ruleset == NULL)
344 		return (NULL);
345 	rs_num = pf_get_ruleset_number(rule_action);
346 	if (rs_num >= PF_RULESET_MAX)
347 		return (NULL);
348 	if (active) {
349 		if (check_ticket && ticket !=
350 		    ruleset->rules[rs_num].active.ticket)
351 			return (NULL);
352 		if (r_last)
353 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
354 			    pf_rulequeue);
355 		else
356 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
357 	} else {
358 		if (check_ticket && ticket !=
359 		    ruleset->rules[rs_num].inactive.ticket)
360 			return (NULL);
361 		if (r_last)
362 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
363 			    pf_rulequeue);
364 		else
365 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
366 	}
367 	if (!r_last) {
368 		while ((rule != NULL) && (rule->nr != rule_number))
369 			rule = TAILQ_NEXT(rule, entries);
370 	}
371 	if (rule == NULL)
372 		return (NULL);
373 
374 	return (&rule->rpool);
375 }
376 
377 static void
378 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
379 {
380 	struct pf_pooladdr	*mv_pool_pa;
381 
382 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
383 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
384 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
385 	}
386 }
387 
388 static void
389 pf_empty_pool(struct pf_palist *poola)
390 {
391 	struct pf_pooladdr *pa;
392 
393 	while ((pa = TAILQ_FIRST(poola)) != NULL) {
394 		switch (pa->addr.type) {
395 		case PF_ADDR_DYNIFTL:
396 			pfi_dynaddr_remove(pa->addr.p.dyn);
397 			break;
398 		case PF_ADDR_TABLE:
399 			/* XXX: this could be unfinished pooladdr on pabuf */
400 			if (pa->addr.p.tbl != NULL)
401 				pfr_detach_table(pa->addr.p.tbl);
402 			break;
403 		}
404 		if (pa->kif)
405 			pfi_kif_unref(pa->kif);
406 		TAILQ_REMOVE(poola, pa, entries);
407 		free(pa, M_PFRULE);
408 	}
409 }
410 
411 static void
412 pf_unlink_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
413 {
414 
415 	PF_RULES_WASSERT();
416 
417 	TAILQ_REMOVE(rulequeue, rule, entries);
418 
419 	PF_UNLNKDRULES_LOCK();
420 	rule->rule_flag |= PFRULE_REFS;
421 	TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
422 	PF_UNLNKDRULES_UNLOCK();
423 }
424 
425 void
426 pf_free_rule(struct pf_rule *rule)
427 {
428 
429 	PF_RULES_WASSERT();
430 
431 	if (rule->tag)
432 		tag_unref(&V_pf_tags, rule->tag);
433 	if (rule->match_tag)
434 		tag_unref(&V_pf_tags, rule->match_tag);
435 #ifdef ALTQ
436 	if (rule->pqid != rule->qid)
437 		pf_qid_unref(rule->pqid);
438 	pf_qid_unref(rule->qid);
439 #endif
440 	switch (rule->src.addr.type) {
441 	case PF_ADDR_DYNIFTL:
442 		pfi_dynaddr_remove(rule->src.addr.p.dyn);
443 		break;
444 	case PF_ADDR_TABLE:
445 		pfr_detach_table(rule->src.addr.p.tbl);
446 		break;
447 	}
448 	switch (rule->dst.addr.type) {
449 	case PF_ADDR_DYNIFTL:
450 		pfi_dynaddr_remove(rule->dst.addr.p.dyn);
451 		break;
452 	case PF_ADDR_TABLE:
453 		pfr_detach_table(rule->dst.addr.p.tbl);
454 		break;
455 	}
456 	if (rule->overload_tbl)
457 		pfr_detach_table(rule->overload_tbl);
458 	if (rule->kif)
459 		pfi_kif_unref(rule->kif);
460 	pf_anchor_remove(rule);
461 	pf_empty_pool(&rule->rpool.list);
462 	counter_u64_free(rule->states_cur);
463 	counter_u64_free(rule->states_tot);
464 	counter_u64_free(rule->src_nodes);
465 	free(rule, M_PFRULE);
466 }
467 
468 static void
469 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
470     unsigned int default_size)
471 {
472 	unsigned int i;
473 	unsigned int hashsize;
474 
475 	if (*tunable_size == 0 || !powerof2(*tunable_size))
476 		*tunable_size = default_size;
477 
478 	hashsize = *tunable_size;
479 	ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
480 	    M_WAITOK);
481 	ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
482 	    M_WAITOK);
483 	ts->mask = hashsize - 1;
484 	ts->seed = arc4random();
485 	for (i = 0; i < hashsize; i++) {
486 		TAILQ_INIT(&ts->namehash[i]);
487 		TAILQ_INIT(&ts->taghash[i]);
488 	}
489 	BIT_FILL(TAGID_MAX, &ts->avail);
490 }
491 
492 static void
493 pf_cleanup_tagset(struct pf_tagset *ts)
494 {
495 	unsigned int i;
496 	unsigned int hashsize;
497 	struct pf_tagname *t, *tmp;
498 
499 	/*
500 	 * Only need to clean up one of the hashes as each tag is hashed
501 	 * into each table.
502 	 */
503 	hashsize = ts->mask + 1;
504 	for (i = 0; i < hashsize; i++)
505 		TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
506 			uma_zfree(V_pf_tag_z, t);
507 
508 	free(ts->namehash, M_PFHASH);
509 	free(ts->taghash, M_PFHASH);
510 }
511 
512 static uint16_t
513 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
514 {
515 
516 	return (murmur3_32_hash(tagname, strlen(tagname), ts->seed) & ts->mask);
517 }
518 
519 static uint16_t
520 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
521 {
522 
523 	return (tag & ts->mask);
524 }
525 
526 static u_int16_t
527 tagname2tag(struct pf_tagset *ts, char *tagname)
528 {
529 	struct pf_tagname	*tag;
530 	u_int32_t		 index;
531 	u_int16_t		 new_tagid;
532 
533 	PF_RULES_WASSERT();
534 
535 	index = tagname2hashindex(ts, tagname);
536 	TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
537 		if (strcmp(tagname, tag->name) == 0) {
538 			tag->ref++;
539 			return (tag->tag);
540 		}
541 
542 	/*
543 	 * new entry
544 	 *
545 	 * to avoid fragmentation, we do a linear search from the beginning
546 	 * and take the first free slot we find.
547 	 */
548 	new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
549 	/*
550 	 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
551 	 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
552 	 * set.  It may also return a bit number greater than TAGID_MAX due
553 	 * to rounding of the number of bits in the vector up to a multiple
554 	 * of the vector word size at declaration/allocation time.
555 	 */
556 	if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
557 		return (0);
558 
559 	/* Mark the tag as in use.  Bits are 0-based for BIT_CLR() */
560 	BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
561 
562 	/* allocate and fill new struct pf_tagname */
563 	tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
564 	if (tag == NULL)
565 		return (0);
566 	strlcpy(tag->name, tagname, sizeof(tag->name));
567 	tag->tag = new_tagid;
568 	tag->ref = 1;
569 
570 	/* Insert into namehash */
571 	TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
572 
573 	/* Insert into taghash */
574 	index = tag2hashindex(ts, new_tagid);
575 	TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
576 
577 	return (tag->tag);
578 }
579 
580 static void
581 tag_unref(struct pf_tagset *ts, u_int16_t tag)
582 {
583 	struct pf_tagname	*t;
584 	uint16_t		 index;
585 
586 	PF_RULES_WASSERT();
587 
588 	index = tag2hashindex(ts, tag);
589 	TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
590 		if (tag == t->tag) {
591 			if (--t->ref == 0) {
592 				TAILQ_REMOVE(&ts->taghash[index], t,
593 				    taghash_entries);
594 				index = tagname2hashindex(ts, t->name);
595 				TAILQ_REMOVE(&ts->namehash[index], t,
596 				    namehash_entries);
597 				/* Bits are 0-based for BIT_SET() */
598 				BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
599 				uma_zfree(V_pf_tag_z, t);
600 			}
601 			break;
602 		}
603 }
604 
605 static u_int16_t
606 pf_tagname2tag(char *tagname)
607 {
608 	return (tagname2tag(&V_pf_tags, tagname));
609 }
610 
611 #ifdef ALTQ
612 static u_int32_t
613 pf_qname2qid(char *qname)
614 {
615 	return ((u_int32_t)tagname2tag(&V_pf_qids, qname));
616 }
617 
618 static void
619 pf_qid_unref(u_int32_t qid)
620 {
621 	tag_unref(&V_pf_qids, (u_int16_t)qid);
622 }
623 
624 static int
625 pf_begin_altq(u_int32_t *ticket)
626 {
627 	struct pf_altq	*altq, *tmp;
628 	int		 error = 0;
629 
630 	PF_RULES_WASSERT();
631 
632 	/* Purge the old altq lists */
633 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
634 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
635 			/* detach and destroy the discipline */
636 			error = altq_remove(altq);
637 		}
638 		free(altq, M_PFALTQ);
639 	}
640 	TAILQ_INIT(V_pf_altq_ifs_inactive);
641 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
642 		pf_qid_unref(altq->qid);
643 		free(altq, M_PFALTQ);
644 	}
645 	TAILQ_INIT(V_pf_altqs_inactive);
646 	if (error)
647 		return (error);
648 	*ticket = ++V_ticket_altqs_inactive;
649 	V_altqs_inactive_open = 1;
650 	return (0);
651 }
652 
653 static int
654 pf_rollback_altq(u_int32_t ticket)
655 {
656 	struct pf_altq	*altq, *tmp;
657 	int		 error = 0;
658 
659 	PF_RULES_WASSERT();
660 
661 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
662 		return (0);
663 	/* Purge the old altq lists */
664 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
665 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
666 			/* detach and destroy the discipline */
667 			error = altq_remove(altq);
668 		}
669 		free(altq, M_PFALTQ);
670 	}
671 	TAILQ_INIT(V_pf_altq_ifs_inactive);
672 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
673 		pf_qid_unref(altq->qid);
674 		free(altq, M_PFALTQ);
675 	}
676 	TAILQ_INIT(V_pf_altqs_inactive);
677 	V_altqs_inactive_open = 0;
678 	return (error);
679 }
680 
681 static int
682 pf_commit_altq(u_int32_t ticket)
683 {
684 	struct pf_altqqueue	*old_altqs, *old_altq_ifs;
685 	struct pf_altq		*altq, *tmp;
686 	int			 err, error = 0;
687 
688 	PF_RULES_WASSERT();
689 
690 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
691 		return (EBUSY);
692 
693 	/* swap altqs, keep the old. */
694 	old_altqs = V_pf_altqs_active;
695 	old_altq_ifs = V_pf_altq_ifs_active;
696 	V_pf_altqs_active = V_pf_altqs_inactive;
697 	V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
698 	V_pf_altqs_inactive = old_altqs;
699 	V_pf_altq_ifs_inactive = old_altq_ifs;
700 	V_ticket_altqs_active = V_ticket_altqs_inactive;
701 
702 	/* Attach new disciplines */
703 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
704 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
705 			/* attach the discipline */
706 			error = altq_pfattach(altq);
707 			if (error == 0 && V_pf_altq_running)
708 				error = pf_enable_altq(altq);
709 			if (error != 0)
710 				return (error);
711 		}
712 	}
713 
714 	/* Purge the old altq lists */
715 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
716 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
717 			/* detach and destroy the discipline */
718 			if (V_pf_altq_running)
719 				error = pf_disable_altq(altq);
720 			err = altq_pfdetach(altq);
721 			if (err != 0 && error == 0)
722 				error = err;
723 			err = altq_remove(altq);
724 			if (err != 0 && error == 0)
725 				error = err;
726 		}
727 		free(altq, M_PFALTQ);
728 	}
729 	TAILQ_INIT(V_pf_altq_ifs_inactive);
730 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
731 		pf_qid_unref(altq->qid);
732 		free(altq, M_PFALTQ);
733 	}
734 	TAILQ_INIT(V_pf_altqs_inactive);
735 
736 	V_altqs_inactive_open = 0;
737 	return (error);
738 }
739 
740 static int
741 pf_enable_altq(struct pf_altq *altq)
742 {
743 	struct ifnet		*ifp;
744 	struct tb_profile	 tb;
745 	int			 error = 0;
746 
747 	if ((ifp = ifunit(altq->ifname)) == NULL)
748 		return (EINVAL);
749 
750 	if (ifp->if_snd.altq_type != ALTQT_NONE)
751 		error = altq_enable(&ifp->if_snd);
752 
753 	/* set tokenbucket regulator */
754 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
755 		tb.rate = altq->ifbandwidth;
756 		tb.depth = altq->tbrsize;
757 		error = tbr_set(&ifp->if_snd, &tb);
758 	}
759 
760 	return (error);
761 }
762 
763 static int
764 pf_disable_altq(struct pf_altq *altq)
765 {
766 	struct ifnet		*ifp;
767 	struct tb_profile	 tb;
768 	int			 error;
769 
770 	if ((ifp = ifunit(altq->ifname)) == NULL)
771 		return (EINVAL);
772 
773 	/*
774 	 * when the discipline is no longer referenced, it was overridden
775 	 * by a new one.  if so, just return.
776 	 */
777 	if (altq->altq_disc != ifp->if_snd.altq_disc)
778 		return (0);
779 
780 	error = altq_disable(&ifp->if_snd);
781 
782 	if (error == 0) {
783 		/* clear tokenbucket regulator */
784 		tb.rate = 0;
785 		error = tbr_set(&ifp->if_snd, &tb);
786 	}
787 
788 	return (error);
789 }
790 
791 static int
792 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
793     struct pf_altq *altq)
794 {
795 	struct ifnet	*ifp1;
796 	int		 error = 0;
797 
798 	/* Deactivate the interface in question */
799 	altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
800 	if ((ifp1 = ifunit(altq->ifname)) == NULL ||
801 	    (remove && ifp1 == ifp)) {
802 		altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
803 	} else {
804 		error = altq_add(ifp1, altq);
805 
806 		if (ticket != V_ticket_altqs_inactive)
807 			error = EBUSY;
808 
809 		if (error)
810 			free(altq, M_PFALTQ);
811 	}
812 
813 	return (error);
814 }
815 
816 void
817 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
818 {
819 	struct pf_altq	*a1, *a2, *a3;
820 	u_int32_t	 ticket;
821 	int		 error = 0;
822 
823 	/*
824 	 * No need to re-evaluate the configuration for events on interfaces
825 	 * that do not support ALTQ, as it's not possible for such
826 	 * interfaces to be part of the configuration.
827 	 */
828 	if (!ALTQ_IS_READY(&ifp->if_snd))
829 		return;
830 
831 	/* Interrupt userland queue modifications */
832 	if (V_altqs_inactive_open)
833 		pf_rollback_altq(V_ticket_altqs_inactive);
834 
835 	/* Start new altq ruleset */
836 	if (pf_begin_altq(&ticket))
837 		return;
838 
839 	/* Copy the current active set */
840 	TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
841 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
842 		if (a2 == NULL) {
843 			error = ENOMEM;
844 			break;
845 		}
846 		bcopy(a1, a2, sizeof(struct pf_altq));
847 
848 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
849 		if (error)
850 			break;
851 
852 		TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
853 	}
854 	if (error)
855 		goto out;
856 	TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
857 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
858 		if (a2 == NULL) {
859 			error = ENOMEM;
860 			break;
861 		}
862 		bcopy(a1, a2, sizeof(struct pf_altq));
863 
864 		if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
865 			error = EBUSY;
866 			free(a2, M_PFALTQ);
867 			break;
868 		}
869 		a2->altq_disc = NULL;
870 		TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
871 			if (strncmp(a3->ifname, a2->ifname,
872 				IFNAMSIZ) == 0) {
873 				a2->altq_disc = a3->altq_disc;
874 				break;
875 			}
876 		}
877 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
878 		if (error)
879 			break;
880 
881 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
882 	}
883 
884 out:
885 	if (error != 0)
886 		pf_rollback_altq(ticket);
887 	else
888 		pf_commit_altq(ticket);
889 }
890 #endif /* ALTQ */
891 
892 static int
893 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
894 {
895 	struct pf_ruleset	*rs;
896 	struct pf_rule		*rule;
897 
898 	PF_RULES_WASSERT();
899 
900 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
901 		return (EINVAL);
902 	rs = pf_find_or_create_ruleset(anchor);
903 	if (rs == NULL)
904 		return (EINVAL);
905 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
906 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
907 		rs->rules[rs_num].inactive.rcount--;
908 	}
909 	*ticket = ++rs->rules[rs_num].inactive.ticket;
910 	rs->rules[rs_num].inactive.open = 1;
911 	return (0);
912 }
913 
914 static int
915 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
916 {
917 	struct pf_ruleset	*rs;
918 	struct pf_rule		*rule;
919 
920 	PF_RULES_WASSERT();
921 
922 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
923 		return (EINVAL);
924 	rs = pf_find_ruleset(anchor);
925 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
926 	    rs->rules[rs_num].inactive.ticket != ticket)
927 		return (0);
928 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
929 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
930 		rs->rules[rs_num].inactive.rcount--;
931 	}
932 	rs->rules[rs_num].inactive.open = 0;
933 	return (0);
934 }
935 
936 #define PF_MD5_UPD(st, elm)						\
937 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
938 
939 #define PF_MD5_UPD_STR(st, elm)						\
940 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
941 
942 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
943 		(stor) = htonl((st)->elm);				\
944 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
945 } while (0)
946 
947 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
948 		(stor) = htons((st)->elm);				\
949 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
950 } while (0)
951 
952 static void
953 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
954 {
955 	PF_MD5_UPD(pfr, addr.type);
956 	switch (pfr->addr.type) {
957 		case PF_ADDR_DYNIFTL:
958 			PF_MD5_UPD(pfr, addr.v.ifname);
959 			PF_MD5_UPD(pfr, addr.iflags);
960 			break;
961 		case PF_ADDR_TABLE:
962 			PF_MD5_UPD(pfr, addr.v.tblname);
963 			break;
964 		case PF_ADDR_ADDRMASK:
965 			/* XXX ignore af? */
966 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
967 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
968 			break;
969 	}
970 
971 	PF_MD5_UPD(pfr, port[0]);
972 	PF_MD5_UPD(pfr, port[1]);
973 	PF_MD5_UPD(pfr, neg);
974 	PF_MD5_UPD(pfr, port_op);
975 }
976 
977 static void
978 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
979 {
980 	u_int16_t x;
981 	u_int32_t y;
982 
983 	pf_hash_rule_addr(ctx, &rule->src);
984 	pf_hash_rule_addr(ctx, &rule->dst);
985 	PF_MD5_UPD_STR(rule, label);
986 	PF_MD5_UPD_STR(rule, ifname);
987 	PF_MD5_UPD_STR(rule, match_tagname);
988 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
989 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
990 	PF_MD5_UPD_HTONL(rule, prob, y);
991 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
992 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
993 	PF_MD5_UPD(rule, uid.op);
994 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
995 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
996 	PF_MD5_UPD(rule, gid.op);
997 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
998 	PF_MD5_UPD(rule, action);
999 	PF_MD5_UPD(rule, direction);
1000 	PF_MD5_UPD(rule, af);
1001 	PF_MD5_UPD(rule, quick);
1002 	PF_MD5_UPD(rule, ifnot);
1003 	PF_MD5_UPD(rule, match_tag_not);
1004 	PF_MD5_UPD(rule, natpass);
1005 	PF_MD5_UPD(rule, keep_state);
1006 	PF_MD5_UPD(rule, proto);
1007 	PF_MD5_UPD(rule, type);
1008 	PF_MD5_UPD(rule, code);
1009 	PF_MD5_UPD(rule, flags);
1010 	PF_MD5_UPD(rule, flagset);
1011 	PF_MD5_UPD(rule, allow_opts);
1012 	PF_MD5_UPD(rule, rt);
1013 	PF_MD5_UPD(rule, tos);
1014 }
1015 
1016 static int
1017 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1018 {
1019 	struct pf_ruleset	*rs;
1020 	struct pf_rule		*rule, **old_array;
1021 	struct pf_rulequeue	*old_rules;
1022 	int			 error;
1023 	u_int32_t		 old_rcount;
1024 
1025 	PF_RULES_WASSERT();
1026 
1027 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1028 		return (EINVAL);
1029 	rs = pf_find_ruleset(anchor);
1030 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1031 	    ticket != rs->rules[rs_num].inactive.ticket)
1032 		return (EBUSY);
1033 
1034 	/* Calculate checksum for the main ruleset */
1035 	if (rs == &pf_main_ruleset) {
1036 		error = pf_setup_pfsync_matching(rs);
1037 		if (error != 0)
1038 			return (error);
1039 	}
1040 
1041 	/* Swap rules, keep the old. */
1042 	old_rules = rs->rules[rs_num].active.ptr;
1043 	old_rcount = rs->rules[rs_num].active.rcount;
1044 	old_array = rs->rules[rs_num].active.ptr_array;
1045 
1046 	rs->rules[rs_num].active.ptr =
1047 	    rs->rules[rs_num].inactive.ptr;
1048 	rs->rules[rs_num].active.ptr_array =
1049 	    rs->rules[rs_num].inactive.ptr_array;
1050 	rs->rules[rs_num].active.rcount =
1051 	    rs->rules[rs_num].inactive.rcount;
1052 	rs->rules[rs_num].inactive.ptr = old_rules;
1053 	rs->rules[rs_num].inactive.ptr_array = old_array;
1054 	rs->rules[rs_num].inactive.rcount = old_rcount;
1055 
1056 	rs->rules[rs_num].active.ticket =
1057 	    rs->rules[rs_num].inactive.ticket;
1058 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1059 
1060 	/* Purge the old rule list. */
1061 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1062 		pf_unlink_rule(old_rules, rule);
1063 	if (rs->rules[rs_num].inactive.ptr_array)
1064 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1065 	rs->rules[rs_num].inactive.ptr_array = NULL;
1066 	rs->rules[rs_num].inactive.rcount = 0;
1067 	rs->rules[rs_num].inactive.open = 0;
1068 	pf_remove_if_empty_ruleset(rs);
1069 
1070 	return (0);
1071 }
1072 
1073 static int
1074 pf_setup_pfsync_matching(struct pf_ruleset *rs)
1075 {
1076 	MD5_CTX			 ctx;
1077 	struct pf_rule		*rule;
1078 	int			 rs_cnt;
1079 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
1080 
1081 	MD5Init(&ctx);
1082 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1083 		/* XXX PF_RULESET_SCRUB as well? */
1084 		if (rs_cnt == PF_RULESET_SCRUB)
1085 			continue;
1086 
1087 		if (rs->rules[rs_cnt].inactive.ptr_array)
1088 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1089 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
1090 
1091 		if (rs->rules[rs_cnt].inactive.rcount) {
1092 			rs->rules[rs_cnt].inactive.ptr_array =
1093 			    malloc(sizeof(caddr_t) *
1094 			    rs->rules[rs_cnt].inactive.rcount,
1095 			    M_TEMP, M_NOWAIT);
1096 
1097 			if (!rs->rules[rs_cnt].inactive.ptr_array)
1098 				return (ENOMEM);
1099 		}
1100 
1101 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1102 		    entries) {
1103 			pf_hash_rule(&ctx, rule);
1104 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1105 		}
1106 	}
1107 
1108 	MD5Final(digest, &ctx);
1109 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1110 	return (0);
1111 }
1112 
1113 static int
1114 pf_addr_setup(struct pf_ruleset *ruleset, struct pf_addr_wrap *addr,
1115     sa_family_t af)
1116 {
1117 	int error = 0;
1118 
1119 	switch (addr->type) {
1120 	case PF_ADDR_TABLE:
1121 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1122 		if (addr->p.tbl == NULL)
1123 			error = ENOMEM;
1124 		break;
1125 	case PF_ADDR_DYNIFTL:
1126 		error = pfi_dynaddr_setup(addr, af);
1127 		break;
1128 	}
1129 
1130 	return (error);
1131 }
1132 
1133 static void
1134 pf_addr_copyout(struct pf_addr_wrap *addr)
1135 {
1136 
1137 	switch (addr->type) {
1138 	case PF_ADDR_DYNIFTL:
1139 		pfi_dynaddr_copyout(addr);
1140 		break;
1141 	case PF_ADDR_TABLE:
1142 		pf_tbladdr_copyout(addr);
1143 		break;
1144 	}
1145 }
1146 
1147 #ifdef ALTQ
1148 /*
1149  * Handle export of struct pf_kaltq to user binaries that may be using any
1150  * version of struct pf_altq.
1151  */
1152 static int
1153 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1154 {
1155 	u_int32_t version;
1156 
1157 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1158 		version = 0;
1159 	else
1160 		version = pa->version;
1161 
1162 	if (version > PFIOC_ALTQ_VERSION)
1163 		return (EINVAL);
1164 
1165 #define ASSIGN(x) exported_q->x = q->x
1166 #define COPY(x) \
1167 	bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1168 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1169 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1170 
1171 	switch (version) {
1172 	case 0: {
1173 		struct pf_altq_v0 *exported_q =
1174 		    &((struct pfioc_altq_v0 *)pa)->altq;
1175 
1176 		COPY(ifname);
1177 
1178 		ASSIGN(scheduler);
1179 		ASSIGN(tbrsize);
1180 		exported_q->tbrsize = SATU16(q->tbrsize);
1181 		exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1182 
1183 		COPY(qname);
1184 		COPY(parent);
1185 		ASSIGN(parent_qid);
1186 		exported_q->bandwidth = SATU32(q->bandwidth);
1187 		ASSIGN(priority);
1188 		ASSIGN(local_flags);
1189 
1190 		ASSIGN(qlimit);
1191 		ASSIGN(flags);
1192 
1193 		if (q->scheduler == ALTQT_HFSC) {
1194 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1195 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1196 			    SATU32(q->pq_u.hfsc_opts.x)
1197 
1198 			ASSIGN_OPT_SATU32(rtsc_m1);
1199 			ASSIGN_OPT(rtsc_d);
1200 			ASSIGN_OPT_SATU32(rtsc_m2);
1201 
1202 			ASSIGN_OPT_SATU32(lssc_m1);
1203 			ASSIGN_OPT(lssc_d);
1204 			ASSIGN_OPT_SATU32(lssc_m2);
1205 
1206 			ASSIGN_OPT_SATU32(ulsc_m1);
1207 			ASSIGN_OPT(ulsc_d);
1208 			ASSIGN_OPT_SATU32(ulsc_m2);
1209 
1210 			ASSIGN_OPT(flags);
1211 
1212 #undef ASSIGN_OPT
1213 #undef ASSIGN_OPT_SATU32
1214 		} else
1215 			COPY(pq_u);
1216 
1217 		ASSIGN(qid);
1218 		break;
1219 	}
1220 	case 1:	{
1221 		struct pf_altq_v1 *exported_q =
1222 		    &((struct pfioc_altq_v1 *)pa)->altq;
1223 
1224 		COPY(ifname);
1225 
1226 		ASSIGN(scheduler);
1227 		ASSIGN(tbrsize);
1228 		ASSIGN(ifbandwidth);
1229 
1230 		COPY(qname);
1231 		COPY(parent);
1232 		ASSIGN(parent_qid);
1233 		ASSIGN(bandwidth);
1234 		ASSIGN(priority);
1235 		ASSIGN(local_flags);
1236 
1237 		ASSIGN(qlimit);
1238 		ASSIGN(flags);
1239 		COPY(pq_u);
1240 
1241 		ASSIGN(qid);
1242 		break;
1243 	}
1244 	default:
1245 		panic("%s: unhandled struct pfioc_altq version", __func__);
1246 		break;
1247 	}
1248 
1249 #undef ASSIGN
1250 #undef COPY
1251 #undef SATU16
1252 #undef SATU32
1253 
1254 	return (0);
1255 }
1256 
1257 /*
1258  * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1259  * that may be using any version of it.
1260  */
1261 static int
1262 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1263 {
1264 	u_int32_t version;
1265 
1266 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1267 		version = 0;
1268 	else
1269 		version = pa->version;
1270 
1271 	if (version > PFIOC_ALTQ_VERSION)
1272 		return (EINVAL);
1273 
1274 #define ASSIGN(x) q->x = imported_q->x
1275 #define COPY(x) \
1276 	bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1277 
1278 	switch (version) {
1279 	case 0: {
1280 		struct pf_altq_v0 *imported_q =
1281 		    &((struct pfioc_altq_v0 *)pa)->altq;
1282 
1283 		COPY(ifname);
1284 
1285 		ASSIGN(scheduler);
1286 		ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1287 		ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1288 
1289 		COPY(qname);
1290 		COPY(parent);
1291 		ASSIGN(parent_qid);
1292 		ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1293 		ASSIGN(priority);
1294 		ASSIGN(local_flags);
1295 
1296 		ASSIGN(qlimit);
1297 		ASSIGN(flags);
1298 
1299 		if (imported_q->scheduler == ALTQT_HFSC) {
1300 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1301 
1302 			/*
1303 			 * The m1 and m2 parameters are being copied from
1304 			 * 32-bit to 64-bit.
1305 			 */
1306 			ASSIGN_OPT(rtsc_m1);
1307 			ASSIGN_OPT(rtsc_d);
1308 			ASSIGN_OPT(rtsc_m2);
1309 
1310 			ASSIGN_OPT(lssc_m1);
1311 			ASSIGN_OPT(lssc_d);
1312 			ASSIGN_OPT(lssc_m2);
1313 
1314 			ASSIGN_OPT(ulsc_m1);
1315 			ASSIGN_OPT(ulsc_d);
1316 			ASSIGN_OPT(ulsc_m2);
1317 
1318 			ASSIGN_OPT(flags);
1319 
1320 #undef ASSIGN_OPT
1321 		} else
1322 			COPY(pq_u);
1323 
1324 		ASSIGN(qid);
1325 		break;
1326 	}
1327 	case 1: {
1328 		struct pf_altq_v1 *imported_q =
1329 		    &((struct pfioc_altq_v1 *)pa)->altq;
1330 
1331 		COPY(ifname);
1332 
1333 		ASSIGN(scheduler);
1334 		ASSIGN(tbrsize);
1335 		ASSIGN(ifbandwidth);
1336 
1337 		COPY(qname);
1338 		COPY(parent);
1339 		ASSIGN(parent_qid);
1340 		ASSIGN(bandwidth);
1341 		ASSIGN(priority);
1342 		ASSIGN(local_flags);
1343 
1344 		ASSIGN(qlimit);
1345 		ASSIGN(flags);
1346 		COPY(pq_u);
1347 
1348 		ASSIGN(qid);
1349 		break;
1350 	}
1351 	default:
1352 		panic("%s: unhandled struct pfioc_altq version", __func__);
1353 		break;
1354 	}
1355 
1356 #undef ASSIGN
1357 #undef COPY
1358 
1359 	return (0);
1360 }
1361 
1362 static struct pf_altq *
1363 pf_altq_get_nth_active(u_int32_t n)
1364 {
1365 	struct pf_altq		*altq;
1366 	u_int32_t		 nr;
1367 
1368 	nr = 0;
1369 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1370 		if (nr == n)
1371 			return (altq);
1372 		nr++;
1373 	}
1374 
1375 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1376 		if (nr == n)
1377 			return (altq);
1378 		nr++;
1379 	}
1380 
1381 	return (NULL);
1382 }
1383 #endif /* ALTQ */
1384 
1385 static int
1386 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
1387 {
1388 	int			 error = 0;
1389 	PF_RULES_RLOCK_TRACKER;
1390 
1391 	/* XXX keep in sync with switch() below */
1392 	if (securelevel_gt(td->td_ucred, 2))
1393 		switch (cmd) {
1394 		case DIOCGETRULES:
1395 		case DIOCGETRULE:
1396 		case DIOCGETADDRS:
1397 		case DIOCGETADDR:
1398 		case DIOCGETSTATE:
1399 		case DIOCSETSTATUSIF:
1400 		case DIOCGETSTATUS:
1401 		case DIOCCLRSTATUS:
1402 		case DIOCNATLOOK:
1403 		case DIOCSETDEBUG:
1404 		case DIOCGETSTATES:
1405 		case DIOCGETTIMEOUT:
1406 		case DIOCCLRRULECTRS:
1407 		case DIOCGETLIMIT:
1408 		case DIOCGETALTQSV0:
1409 		case DIOCGETALTQSV1:
1410 		case DIOCGETALTQV0:
1411 		case DIOCGETALTQV1:
1412 		case DIOCGETQSTATSV0:
1413 		case DIOCGETQSTATSV1:
1414 		case DIOCGETRULESETS:
1415 		case DIOCGETRULESET:
1416 		case DIOCRGETTABLES:
1417 		case DIOCRGETTSTATS:
1418 		case DIOCRCLRTSTATS:
1419 		case DIOCRCLRADDRS:
1420 		case DIOCRADDADDRS:
1421 		case DIOCRDELADDRS:
1422 		case DIOCRSETADDRS:
1423 		case DIOCRGETADDRS:
1424 		case DIOCRGETASTATS:
1425 		case DIOCRCLRASTATS:
1426 		case DIOCRTSTADDRS:
1427 		case DIOCOSFPGET:
1428 		case DIOCGETSRCNODES:
1429 		case DIOCCLRSRCNODES:
1430 		case DIOCIGETIFACES:
1431 		case DIOCGIFSPEEDV0:
1432 		case DIOCGIFSPEEDV1:
1433 		case DIOCSETIFFLAG:
1434 		case DIOCCLRIFFLAG:
1435 			break;
1436 		case DIOCRCLRTABLES:
1437 		case DIOCRADDTABLES:
1438 		case DIOCRDELTABLES:
1439 		case DIOCRSETTFLAGS:
1440 			if (((struct pfioc_table *)addr)->pfrio_flags &
1441 			    PFR_FLAG_DUMMY)
1442 				break; /* dummy operation ok */
1443 			return (EPERM);
1444 		default:
1445 			return (EPERM);
1446 		}
1447 
1448 	if (!(flags & FWRITE))
1449 		switch (cmd) {
1450 		case DIOCGETRULES:
1451 		case DIOCGETADDRS:
1452 		case DIOCGETADDR:
1453 		case DIOCGETSTATE:
1454 		case DIOCGETSTATUS:
1455 		case DIOCGETSTATES:
1456 		case DIOCGETTIMEOUT:
1457 		case DIOCGETLIMIT:
1458 		case DIOCGETALTQSV0:
1459 		case DIOCGETALTQSV1:
1460 		case DIOCGETALTQV0:
1461 		case DIOCGETALTQV1:
1462 		case DIOCGETQSTATSV0:
1463 		case DIOCGETQSTATSV1:
1464 		case DIOCGETRULESETS:
1465 		case DIOCGETRULESET:
1466 		case DIOCNATLOOK:
1467 		case DIOCRGETTABLES:
1468 		case DIOCRGETTSTATS:
1469 		case DIOCRGETADDRS:
1470 		case DIOCRGETASTATS:
1471 		case DIOCRTSTADDRS:
1472 		case DIOCOSFPGET:
1473 		case DIOCGETSRCNODES:
1474 		case DIOCIGETIFACES:
1475 		case DIOCGIFSPEEDV1:
1476 		case DIOCGIFSPEEDV0:
1477 			break;
1478 		case DIOCRCLRTABLES:
1479 		case DIOCRADDTABLES:
1480 		case DIOCRDELTABLES:
1481 		case DIOCRCLRTSTATS:
1482 		case DIOCRCLRADDRS:
1483 		case DIOCRADDADDRS:
1484 		case DIOCRDELADDRS:
1485 		case DIOCRSETADDRS:
1486 		case DIOCRSETTFLAGS:
1487 			if (((struct pfioc_table *)addr)->pfrio_flags &
1488 			    PFR_FLAG_DUMMY) {
1489 				flags |= FWRITE; /* need write lock for dummy */
1490 				break; /* dummy operation ok */
1491 			}
1492 			return (EACCES);
1493 		case DIOCGETRULE:
1494 			if (((struct pfioc_rule *)addr)->action ==
1495 			    PF_GET_CLR_CNTR)
1496 				return (EACCES);
1497 			break;
1498 		default:
1499 			return (EACCES);
1500 		}
1501 
1502 	CURVNET_SET(TD_TO_VNET(td));
1503 
1504 	switch (cmd) {
1505 	case DIOCSTART:
1506 		sx_xlock(&pf_ioctl_lock);
1507 		if (V_pf_status.running)
1508 			error = EEXIST;
1509 		else {
1510 			int cpu;
1511 
1512 			error = hook_pf();
1513 			if (error) {
1514 				DPFPRINTF(PF_DEBUG_MISC,
1515 				    ("pf: pfil registration failed\n"));
1516 				break;
1517 			}
1518 			V_pf_status.running = 1;
1519 			V_pf_status.since = time_second;
1520 
1521 			CPU_FOREACH(cpu)
1522 				V_pf_stateid[cpu] = time_second;
1523 
1524 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1525 		}
1526 		break;
1527 
1528 	case DIOCSTOP:
1529 		sx_xlock(&pf_ioctl_lock);
1530 		if (!V_pf_status.running)
1531 			error = ENOENT;
1532 		else {
1533 			V_pf_status.running = 0;
1534 			error = dehook_pf();
1535 			if (error) {
1536 				V_pf_status.running = 1;
1537 				DPFPRINTF(PF_DEBUG_MISC,
1538 				    ("pf: pfil unregistration failed\n"));
1539 			}
1540 			V_pf_status.since = time_second;
1541 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1542 		}
1543 		break;
1544 
1545 	case DIOCADDRULE: {
1546 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1547 		struct pf_ruleset	*ruleset;
1548 		struct pf_rule		*rule, *tail;
1549 		struct pf_pooladdr	*pa;
1550 		struct pfi_kif		*kif = NULL;
1551 		int			 rs_num;
1552 
1553 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1554 			error = EINVAL;
1555 			break;
1556 		}
1557 		if (pr->rule.src.addr.p.dyn != NULL ||
1558 		    pr->rule.dst.addr.p.dyn != NULL) {
1559 			error = EINVAL;
1560 			break;
1561 		}
1562 #ifndef INET
1563 		if (pr->rule.af == AF_INET) {
1564 			error = EAFNOSUPPORT;
1565 			break;
1566 		}
1567 #endif /* INET */
1568 #ifndef INET6
1569 		if (pr->rule.af == AF_INET6) {
1570 			error = EAFNOSUPPORT;
1571 			break;
1572 		}
1573 #endif /* INET6 */
1574 
1575 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
1576 		bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1577 		if (rule->ifname[0])
1578 			kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1579 		rule->states_cur = counter_u64_alloc(M_WAITOK);
1580 		rule->states_tot = counter_u64_alloc(M_WAITOK);
1581 		rule->src_nodes = counter_u64_alloc(M_WAITOK);
1582 		rule->cuid = td->td_ucred->cr_ruid;
1583 		rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1584 		TAILQ_INIT(&rule->rpool.list);
1585 #define	ERROUT(x)	{ error = (x); goto DIOCADDRULE_error; }
1586 
1587 		PF_RULES_WLOCK();
1588 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1589 		ruleset = pf_find_ruleset(pr->anchor);
1590 		if (ruleset == NULL)
1591 			ERROUT(EINVAL);
1592 		rs_num = pf_get_ruleset_number(pr->rule.action);
1593 		if (rs_num >= PF_RULESET_MAX)
1594 			ERROUT(EINVAL);
1595 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1596 			DPFPRINTF(PF_DEBUG_MISC,
1597 			    ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
1598 			    ruleset->rules[rs_num].inactive.ticket));
1599 			ERROUT(EBUSY);
1600 		}
1601 		if (pr->pool_ticket != V_ticket_pabuf) {
1602 			DPFPRINTF(PF_DEBUG_MISC,
1603 			    ("pool_ticket: %d != %d\n", pr->pool_ticket,
1604 			    V_ticket_pabuf));
1605 			ERROUT(EBUSY);
1606 		}
1607 
1608 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1609 		    pf_rulequeue);
1610 		if (tail)
1611 			rule->nr = tail->nr + 1;
1612 		else
1613 			rule->nr = 0;
1614 		if (rule->ifname[0]) {
1615 			rule->kif = pfi_kif_attach(kif, rule->ifname);
1616 			pfi_kif_ref(rule->kif);
1617 		} else
1618 			rule->kif = NULL;
1619 
1620 		if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
1621 			error = EBUSY;
1622 
1623 #ifdef ALTQ
1624 		/* set queue IDs */
1625 		if (rule->qname[0] != 0) {
1626 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1627 				error = EBUSY;
1628 			else if (rule->pqname[0] != 0) {
1629 				if ((rule->pqid =
1630 				    pf_qname2qid(rule->pqname)) == 0)
1631 					error = EBUSY;
1632 			} else
1633 				rule->pqid = rule->qid;
1634 		}
1635 #endif
1636 		if (rule->tagname[0])
1637 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1638 				error = EBUSY;
1639 		if (rule->match_tagname[0])
1640 			if ((rule->match_tag =
1641 			    pf_tagname2tag(rule->match_tagname)) == 0)
1642 				error = EBUSY;
1643 		if (rule->rt && !rule->direction)
1644 			error = EINVAL;
1645 		if (!rule->log)
1646 			rule->logif = 0;
1647 		if (rule->logif >= PFLOGIFS_MAX)
1648 			error = EINVAL;
1649 		if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1650 			error = ENOMEM;
1651 		if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1652 			error = ENOMEM;
1653 		if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1654 			error = EINVAL;
1655 		if (rule->scrub_flags & PFSTATE_SETPRIO &&
1656 		    (rule->set_prio[0] > PF_PRIO_MAX ||
1657 		    rule->set_prio[1] > PF_PRIO_MAX))
1658 			error = EINVAL;
1659 		TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1660 			if (pa->addr.type == PF_ADDR_TABLE) {
1661 				pa->addr.p.tbl = pfr_attach_table(ruleset,
1662 				    pa->addr.v.tblname);
1663 				if (pa->addr.p.tbl == NULL)
1664 					error = ENOMEM;
1665 			}
1666 
1667 		rule->overload_tbl = NULL;
1668 		if (rule->overload_tblname[0]) {
1669 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1670 			    rule->overload_tblname)) == NULL)
1671 				error = EINVAL;
1672 			else
1673 				rule->overload_tbl->pfrkt_flags |=
1674 				    PFR_TFLAG_ACTIVE;
1675 		}
1676 
1677 		pf_mv_pool(&V_pf_pabuf, &rule->rpool.list);
1678 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1679 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1680 		    (rule->rt > PF_NOPFROUTE)) &&
1681 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1682 			error = EINVAL;
1683 
1684 		if (error) {
1685 			pf_free_rule(rule);
1686 			PF_RULES_WUNLOCK();
1687 			break;
1688 		}
1689 
1690 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1691 		rule->evaluations = rule->packets[0] = rule->packets[1] =
1692 		    rule->bytes[0] = rule->bytes[1] = 0;
1693 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1694 		    rule, entries);
1695 		ruleset->rules[rs_num].inactive.rcount++;
1696 		PF_RULES_WUNLOCK();
1697 		break;
1698 
1699 #undef ERROUT
1700 DIOCADDRULE_error:
1701 		PF_RULES_WUNLOCK();
1702 		counter_u64_free(rule->states_cur);
1703 		counter_u64_free(rule->states_tot);
1704 		counter_u64_free(rule->src_nodes);
1705 		free(rule, M_PFRULE);
1706 		if (kif)
1707 			free(kif, PFI_MTYPE);
1708 		break;
1709 	}
1710 
1711 	case DIOCGETRULES: {
1712 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1713 		struct pf_ruleset	*ruleset;
1714 		struct pf_rule		*tail;
1715 		int			 rs_num;
1716 
1717 		PF_RULES_WLOCK();
1718 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1719 		ruleset = pf_find_ruleset(pr->anchor);
1720 		if (ruleset == NULL) {
1721 			PF_RULES_WUNLOCK();
1722 			error = EINVAL;
1723 			break;
1724 		}
1725 		rs_num = pf_get_ruleset_number(pr->rule.action);
1726 		if (rs_num >= PF_RULESET_MAX) {
1727 			PF_RULES_WUNLOCK();
1728 			error = EINVAL;
1729 			break;
1730 		}
1731 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1732 		    pf_rulequeue);
1733 		if (tail)
1734 			pr->nr = tail->nr + 1;
1735 		else
1736 			pr->nr = 0;
1737 		pr->ticket = ruleset->rules[rs_num].active.ticket;
1738 		PF_RULES_WUNLOCK();
1739 		break;
1740 	}
1741 
1742 	case DIOCGETRULE: {
1743 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1744 		struct pf_ruleset	*ruleset;
1745 		struct pf_rule		*rule;
1746 		int			 rs_num, i;
1747 
1748 		PF_RULES_WLOCK();
1749 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1750 		ruleset = pf_find_ruleset(pr->anchor);
1751 		if (ruleset == NULL) {
1752 			PF_RULES_WUNLOCK();
1753 			error = EINVAL;
1754 			break;
1755 		}
1756 		rs_num = pf_get_ruleset_number(pr->rule.action);
1757 		if (rs_num >= PF_RULESET_MAX) {
1758 			PF_RULES_WUNLOCK();
1759 			error = EINVAL;
1760 			break;
1761 		}
1762 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1763 			PF_RULES_WUNLOCK();
1764 			error = EBUSY;
1765 			break;
1766 		}
1767 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1768 		while ((rule != NULL) && (rule->nr != pr->nr))
1769 			rule = TAILQ_NEXT(rule, entries);
1770 		if (rule == NULL) {
1771 			PF_RULES_WUNLOCK();
1772 			error = EBUSY;
1773 			break;
1774 		}
1775 		bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1776 		pr->rule.u_states_cur = counter_u64_fetch(rule->states_cur);
1777 		pr->rule.u_states_tot = counter_u64_fetch(rule->states_tot);
1778 		pr->rule.u_src_nodes = counter_u64_fetch(rule->src_nodes);
1779 		if (pf_anchor_copyout(ruleset, rule, pr)) {
1780 			PF_RULES_WUNLOCK();
1781 			error = EBUSY;
1782 			break;
1783 		}
1784 		pf_addr_copyout(&pr->rule.src.addr);
1785 		pf_addr_copyout(&pr->rule.dst.addr);
1786 		for (i = 0; i < PF_SKIP_COUNT; ++i)
1787 			if (rule->skip[i].ptr == NULL)
1788 				pr->rule.skip[i].nr = -1;
1789 			else
1790 				pr->rule.skip[i].nr =
1791 				    rule->skip[i].ptr->nr;
1792 
1793 		if (pr->action == PF_GET_CLR_CNTR) {
1794 			rule->evaluations = 0;
1795 			rule->packets[0] = rule->packets[1] = 0;
1796 			rule->bytes[0] = rule->bytes[1] = 0;
1797 			counter_u64_zero(rule->states_tot);
1798 		}
1799 		PF_RULES_WUNLOCK();
1800 		break;
1801 	}
1802 
1803 	case DIOCCHANGERULE: {
1804 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1805 		struct pf_ruleset	*ruleset;
1806 		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1807 		struct pfi_kif		*kif = NULL;
1808 		struct pf_pooladdr	*pa;
1809 		u_int32_t		 nr = 0;
1810 		int			 rs_num;
1811 
1812 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1813 		    pcr->action > PF_CHANGE_GET_TICKET) {
1814 			error = EINVAL;
1815 			break;
1816 		}
1817 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1818 			error = EINVAL;
1819 			break;
1820 		}
1821 
1822 		if (pcr->action != PF_CHANGE_REMOVE) {
1823 #ifndef INET
1824 			if (pcr->rule.af == AF_INET) {
1825 				error = EAFNOSUPPORT;
1826 				break;
1827 			}
1828 #endif /* INET */
1829 #ifndef INET6
1830 			if (pcr->rule.af == AF_INET6) {
1831 				error = EAFNOSUPPORT;
1832 				break;
1833 			}
1834 #endif /* INET6 */
1835 			newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
1836 			bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1837 			if (newrule->ifname[0])
1838 				kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK);
1839 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
1840 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
1841 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
1842 			newrule->cuid = td->td_ucred->cr_ruid;
1843 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1844 			TAILQ_INIT(&newrule->rpool.list);
1845 		}
1846 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGERULE_error; }
1847 
1848 		PF_RULES_WLOCK();
1849 		if (!(pcr->action == PF_CHANGE_REMOVE ||
1850 		    pcr->action == PF_CHANGE_GET_TICKET) &&
1851 		    pcr->pool_ticket != V_ticket_pabuf)
1852 			ERROUT(EBUSY);
1853 
1854 		ruleset = pf_find_ruleset(pcr->anchor);
1855 		if (ruleset == NULL)
1856 			ERROUT(EINVAL);
1857 
1858 		rs_num = pf_get_ruleset_number(pcr->rule.action);
1859 		if (rs_num >= PF_RULESET_MAX)
1860 			ERROUT(EINVAL);
1861 
1862 		if (pcr->action == PF_CHANGE_GET_TICKET) {
1863 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1864 			ERROUT(0);
1865 		} else if (pcr->ticket !=
1866 			    ruleset->rules[rs_num].active.ticket)
1867 				ERROUT(EINVAL);
1868 
1869 		if (pcr->action != PF_CHANGE_REMOVE) {
1870 			if (newrule->ifname[0]) {
1871 				newrule->kif = pfi_kif_attach(kif,
1872 				    newrule->ifname);
1873 				pfi_kif_ref(newrule->kif);
1874 			} else
1875 				newrule->kif = NULL;
1876 
1877 			if (newrule->rtableid > 0 &&
1878 			    newrule->rtableid >= rt_numfibs)
1879 				error = EBUSY;
1880 
1881 #ifdef ALTQ
1882 			/* set queue IDs */
1883 			if (newrule->qname[0] != 0) {
1884 				if ((newrule->qid =
1885 				    pf_qname2qid(newrule->qname)) == 0)
1886 					error = EBUSY;
1887 				else if (newrule->pqname[0] != 0) {
1888 					if ((newrule->pqid =
1889 					    pf_qname2qid(newrule->pqname)) == 0)
1890 						error = EBUSY;
1891 				} else
1892 					newrule->pqid = newrule->qid;
1893 			}
1894 #endif /* ALTQ */
1895 			if (newrule->tagname[0])
1896 				if ((newrule->tag =
1897 				    pf_tagname2tag(newrule->tagname)) == 0)
1898 					error = EBUSY;
1899 			if (newrule->match_tagname[0])
1900 				if ((newrule->match_tag = pf_tagname2tag(
1901 				    newrule->match_tagname)) == 0)
1902 					error = EBUSY;
1903 			if (newrule->rt && !newrule->direction)
1904 				error = EINVAL;
1905 			if (!newrule->log)
1906 				newrule->logif = 0;
1907 			if (newrule->logif >= PFLOGIFS_MAX)
1908 				error = EINVAL;
1909 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
1910 				error = ENOMEM;
1911 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
1912 				error = ENOMEM;
1913 			if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1914 				error = EINVAL;
1915 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1916 				if (pa->addr.type == PF_ADDR_TABLE) {
1917 					pa->addr.p.tbl =
1918 					    pfr_attach_table(ruleset,
1919 					    pa->addr.v.tblname);
1920 					if (pa->addr.p.tbl == NULL)
1921 						error = ENOMEM;
1922 				}
1923 
1924 			newrule->overload_tbl = NULL;
1925 			if (newrule->overload_tblname[0]) {
1926 				if ((newrule->overload_tbl = pfr_attach_table(
1927 				    ruleset, newrule->overload_tblname)) ==
1928 				    NULL)
1929 					error = EINVAL;
1930 				else
1931 					newrule->overload_tbl->pfrkt_flags |=
1932 					    PFR_TFLAG_ACTIVE;
1933 			}
1934 
1935 			pf_mv_pool(&V_pf_pabuf, &newrule->rpool.list);
1936 			if (((((newrule->action == PF_NAT) ||
1937 			    (newrule->action == PF_RDR) ||
1938 			    (newrule->action == PF_BINAT) ||
1939 			    (newrule->rt > PF_NOPFROUTE)) &&
1940 			    !newrule->anchor)) &&
1941 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1942 				error = EINVAL;
1943 
1944 			if (error) {
1945 				pf_free_rule(newrule);
1946 				PF_RULES_WUNLOCK();
1947 				break;
1948 			}
1949 
1950 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1951 			newrule->evaluations = 0;
1952 			newrule->packets[0] = newrule->packets[1] = 0;
1953 			newrule->bytes[0] = newrule->bytes[1] = 0;
1954 		}
1955 		pf_empty_pool(&V_pf_pabuf);
1956 
1957 		if (pcr->action == PF_CHANGE_ADD_HEAD)
1958 			oldrule = TAILQ_FIRST(
1959 			    ruleset->rules[rs_num].active.ptr);
1960 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1961 			oldrule = TAILQ_LAST(
1962 			    ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1963 		else {
1964 			oldrule = TAILQ_FIRST(
1965 			    ruleset->rules[rs_num].active.ptr);
1966 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1967 				oldrule = TAILQ_NEXT(oldrule, entries);
1968 			if (oldrule == NULL) {
1969 				if (newrule != NULL)
1970 					pf_free_rule(newrule);
1971 				PF_RULES_WUNLOCK();
1972 				error = EINVAL;
1973 				break;
1974 			}
1975 		}
1976 
1977 		if (pcr->action == PF_CHANGE_REMOVE) {
1978 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
1979 			    oldrule);
1980 			ruleset->rules[rs_num].active.rcount--;
1981 		} else {
1982 			if (oldrule == NULL)
1983 				TAILQ_INSERT_TAIL(
1984 				    ruleset->rules[rs_num].active.ptr,
1985 				    newrule, entries);
1986 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1987 			    pcr->action == PF_CHANGE_ADD_BEFORE)
1988 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1989 			else
1990 				TAILQ_INSERT_AFTER(
1991 				    ruleset->rules[rs_num].active.ptr,
1992 				    oldrule, newrule, entries);
1993 			ruleset->rules[rs_num].active.rcount++;
1994 		}
1995 
1996 		nr = 0;
1997 		TAILQ_FOREACH(oldrule,
1998 		    ruleset->rules[rs_num].active.ptr, entries)
1999 			oldrule->nr = nr++;
2000 
2001 		ruleset->rules[rs_num].active.ticket++;
2002 
2003 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
2004 		pf_remove_if_empty_ruleset(ruleset);
2005 
2006 		PF_RULES_WUNLOCK();
2007 		break;
2008 
2009 #undef ERROUT
2010 DIOCCHANGERULE_error:
2011 		PF_RULES_WUNLOCK();
2012 		if (newrule != NULL) {
2013 			counter_u64_free(newrule->states_cur);
2014 			counter_u64_free(newrule->states_tot);
2015 			counter_u64_free(newrule->src_nodes);
2016 			free(newrule, M_PFRULE);
2017 		}
2018 		if (kif != NULL)
2019 			free(kif, PFI_MTYPE);
2020 		break;
2021 	}
2022 
2023 	case DIOCCLRSTATES: {
2024 		struct pf_state		*s;
2025 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
2026 		u_int			 i, killed = 0;
2027 
2028 		for (i = 0; i <= pf_hashmask; i++) {
2029 			struct pf_idhash *ih = &V_pf_idhash[i];
2030 
2031 relock_DIOCCLRSTATES:
2032 			PF_HASHROW_LOCK(ih);
2033 			LIST_FOREACH(s, &ih->states, entry)
2034 				if (!psk->psk_ifname[0] ||
2035 				    !strcmp(psk->psk_ifname,
2036 				    s->kif->pfik_name)) {
2037 					/*
2038 					 * Don't send out individual
2039 					 * delete messages.
2040 					 */
2041 					s->state_flags |= PFSTATE_NOSYNC;
2042 					pf_unlink_state(s, PF_ENTER_LOCKED);
2043 					killed++;
2044 					goto relock_DIOCCLRSTATES;
2045 				}
2046 			PF_HASHROW_UNLOCK(ih);
2047 		}
2048 		psk->psk_killed = killed;
2049 		if (V_pfsync_clear_states_ptr != NULL)
2050 			V_pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
2051 		break;
2052 	}
2053 
2054 	case DIOCKILLSTATES: {
2055 		struct pf_state		*s;
2056 		struct pf_state_key	*sk;
2057 		struct pf_addr		*srcaddr, *dstaddr;
2058 		u_int16_t		 srcport, dstport;
2059 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
2060 		u_int			 i, killed = 0;
2061 
2062 		if (psk->psk_pfcmp.id) {
2063 			if (psk->psk_pfcmp.creatorid == 0)
2064 				psk->psk_pfcmp.creatorid = V_pf_status.hostid;
2065 			if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
2066 			    psk->psk_pfcmp.creatorid))) {
2067 				pf_unlink_state(s, PF_ENTER_LOCKED);
2068 				psk->psk_killed = 1;
2069 			}
2070 			break;
2071 		}
2072 
2073 		for (i = 0; i <= pf_hashmask; i++) {
2074 			struct pf_idhash *ih = &V_pf_idhash[i];
2075 
2076 relock_DIOCKILLSTATES:
2077 			PF_HASHROW_LOCK(ih);
2078 			LIST_FOREACH(s, &ih->states, entry) {
2079 				sk = s->key[PF_SK_WIRE];
2080 				if (s->direction == PF_OUT) {
2081 					srcaddr = &sk->addr[1];
2082 					dstaddr = &sk->addr[0];
2083 					srcport = sk->port[1];
2084 					dstport = sk->port[0];
2085 				} else {
2086 					srcaddr = &sk->addr[0];
2087 					dstaddr = &sk->addr[1];
2088 					srcport = sk->port[0];
2089 					dstport = sk->port[1];
2090 				}
2091 
2092 				if ((!psk->psk_af || sk->af == psk->psk_af)
2093 				    && (!psk->psk_proto || psk->psk_proto ==
2094 				    sk->proto) &&
2095 				    PF_MATCHA(psk->psk_src.neg,
2096 				    &psk->psk_src.addr.v.a.addr,
2097 				    &psk->psk_src.addr.v.a.mask,
2098 				    srcaddr, sk->af) &&
2099 				    PF_MATCHA(psk->psk_dst.neg,
2100 				    &psk->psk_dst.addr.v.a.addr,
2101 				    &psk->psk_dst.addr.v.a.mask,
2102 				    dstaddr, sk->af) &&
2103 				    (psk->psk_src.port_op == 0 ||
2104 				    pf_match_port(psk->psk_src.port_op,
2105 				    psk->psk_src.port[0], psk->psk_src.port[1],
2106 				    srcport)) &&
2107 				    (psk->psk_dst.port_op == 0 ||
2108 				    pf_match_port(psk->psk_dst.port_op,
2109 				    psk->psk_dst.port[0], psk->psk_dst.port[1],
2110 				    dstport)) &&
2111 				    (!psk->psk_label[0] ||
2112 				    (s->rule.ptr->label[0] &&
2113 				    !strcmp(psk->psk_label,
2114 				    s->rule.ptr->label))) &&
2115 				    (!psk->psk_ifname[0] ||
2116 				    !strcmp(psk->psk_ifname,
2117 				    s->kif->pfik_name))) {
2118 					pf_unlink_state(s, PF_ENTER_LOCKED);
2119 					killed++;
2120 					goto relock_DIOCKILLSTATES;
2121 				}
2122 			}
2123 			PF_HASHROW_UNLOCK(ih);
2124 		}
2125 		psk->psk_killed = killed;
2126 		break;
2127 	}
2128 
2129 	case DIOCADDSTATE: {
2130 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
2131 		struct pfsync_state	*sp = &ps->state;
2132 
2133 		if (sp->timeout >= PFTM_MAX) {
2134 			error = EINVAL;
2135 			break;
2136 		}
2137 		if (V_pfsync_state_import_ptr != NULL) {
2138 			PF_RULES_RLOCK();
2139 			error = V_pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
2140 			PF_RULES_RUNLOCK();
2141 		} else
2142 			error = EOPNOTSUPP;
2143 		break;
2144 	}
2145 
2146 	case DIOCGETSTATE: {
2147 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
2148 		struct pf_state		*s;
2149 
2150 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
2151 		if (s == NULL) {
2152 			error = ENOENT;
2153 			break;
2154 		}
2155 
2156 		pfsync_state_export(&ps->state, s);
2157 		PF_STATE_UNLOCK(s);
2158 		break;
2159 	}
2160 
2161 	case DIOCGETSTATES: {
2162 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
2163 		struct pf_state		*s;
2164 		struct pfsync_state	*pstore, *p;
2165 		int i, nr;
2166 
2167 		if (ps->ps_len <= 0) {
2168 			nr = uma_zone_get_cur(V_pf_state_z);
2169 			ps->ps_len = sizeof(struct pfsync_state) * nr;
2170 			break;
2171 		}
2172 
2173 		p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK | M_ZERO);
2174 		nr = 0;
2175 
2176 		for (i = 0; i <= pf_hashmask; i++) {
2177 			struct pf_idhash *ih = &V_pf_idhash[i];
2178 
2179 			PF_HASHROW_LOCK(ih);
2180 			LIST_FOREACH(s, &ih->states, entry) {
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 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGEADDR_error; }
2774 		PF_RULES_WLOCK();
2775 		ruleset = pf_find_ruleset(pca->anchor);
2776 		if (ruleset == NULL)
2777 			ERROUT(EBUSY);
2778 
2779 		pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2780 		    pca->r_num, pca->r_last, 1, 1);
2781 		if (pool == NULL)
2782 			ERROUT(EBUSY);
2783 
2784 		if (pca->action != PF_CHANGE_REMOVE) {
2785 			if (newpa->ifname[0]) {
2786 				newpa->kif = pfi_kif_attach(kif, newpa->ifname);
2787 				pfi_kif_ref(newpa->kif);
2788 				kif = NULL;
2789 			}
2790 
2791 			switch (newpa->addr.type) {
2792 			case PF_ADDR_DYNIFTL:
2793 				error = pfi_dynaddr_setup(&newpa->addr,
2794 				    pca->af);
2795 				break;
2796 			case PF_ADDR_TABLE:
2797 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
2798 				    newpa->addr.v.tblname);
2799 				if (newpa->addr.p.tbl == NULL)
2800 					error = ENOMEM;
2801 				break;
2802 			}
2803 			if (error)
2804 				goto DIOCCHANGEADDR_error;
2805 		}
2806 
2807 		switch (pca->action) {
2808 		case PF_CHANGE_ADD_HEAD:
2809 			oldpa = TAILQ_FIRST(&pool->list);
2810 			break;
2811 		case PF_CHANGE_ADD_TAIL:
2812 			oldpa = TAILQ_LAST(&pool->list, pf_palist);
2813 			break;
2814 		default:
2815 			oldpa = TAILQ_FIRST(&pool->list);
2816 			for (int i = 0; oldpa && i < pca->nr; i++)
2817 				oldpa = TAILQ_NEXT(oldpa, entries);
2818 
2819 			if (oldpa == NULL)
2820 				ERROUT(EINVAL);
2821 		}
2822 
2823 		if (pca->action == PF_CHANGE_REMOVE) {
2824 			TAILQ_REMOVE(&pool->list, oldpa, entries);
2825 			switch (oldpa->addr.type) {
2826 			case PF_ADDR_DYNIFTL:
2827 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
2828 				break;
2829 			case PF_ADDR_TABLE:
2830 				pfr_detach_table(oldpa->addr.p.tbl);
2831 				break;
2832 			}
2833 			if (oldpa->kif)
2834 				pfi_kif_unref(oldpa->kif);
2835 			free(oldpa, M_PFRULE);
2836 		} else {
2837 			if (oldpa == NULL)
2838 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2839 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
2840 			    pca->action == PF_CHANGE_ADD_BEFORE)
2841 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2842 			else
2843 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
2844 				    newpa, entries);
2845 		}
2846 
2847 		pool->cur = TAILQ_FIRST(&pool->list);
2848 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
2849 		PF_RULES_WUNLOCK();
2850 		break;
2851 
2852 #undef ERROUT
2853 DIOCCHANGEADDR_error:
2854 		if (newpa != NULL) {
2855 			if (newpa->kif)
2856 				pfi_kif_unref(newpa->kif);
2857 			free(newpa, M_PFRULE);
2858 		}
2859 		PF_RULES_WUNLOCK();
2860 		if (kif != NULL)
2861 			free(kif, PFI_MTYPE);
2862 		break;
2863 	}
2864 
2865 	case DIOCGETRULESETS: {
2866 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2867 		struct pf_ruleset	*ruleset;
2868 		struct pf_anchor	*anchor;
2869 
2870 		PF_RULES_RLOCK();
2871 		pr->path[sizeof(pr->path) - 1] = 0;
2872 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2873 			PF_RULES_RUNLOCK();
2874 			error = ENOENT;
2875 			break;
2876 		}
2877 		pr->nr = 0;
2878 		if (ruleset->anchor == NULL) {
2879 			/* XXX kludge for pf_main_ruleset */
2880 			RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2881 				if (anchor->parent == NULL)
2882 					pr->nr++;
2883 		} else {
2884 			RB_FOREACH(anchor, pf_anchor_node,
2885 			    &ruleset->anchor->children)
2886 				pr->nr++;
2887 		}
2888 		PF_RULES_RUNLOCK();
2889 		break;
2890 	}
2891 
2892 	case DIOCGETRULESET: {
2893 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2894 		struct pf_ruleset	*ruleset;
2895 		struct pf_anchor	*anchor;
2896 		u_int32_t		 nr = 0;
2897 
2898 		PF_RULES_RLOCK();
2899 		pr->path[sizeof(pr->path) - 1] = 0;
2900 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2901 			PF_RULES_RUNLOCK();
2902 			error = ENOENT;
2903 			break;
2904 		}
2905 		pr->name[0] = 0;
2906 		if (ruleset->anchor == NULL) {
2907 			/* XXX kludge for pf_main_ruleset */
2908 			RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors)
2909 				if (anchor->parent == NULL && nr++ == pr->nr) {
2910 					strlcpy(pr->name, anchor->name,
2911 					    sizeof(pr->name));
2912 					break;
2913 				}
2914 		} else {
2915 			RB_FOREACH(anchor, pf_anchor_node,
2916 			    &ruleset->anchor->children)
2917 				if (nr++ == pr->nr) {
2918 					strlcpy(pr->name, anchor->name,
2919 					    sizeof(pr->name));
2920 					break;
2921 				}
2922 		}
2923 		if (!pr->name[0])
2924 			error = EBUSY;
2925 		PF_RULES_RUNLOCK();
2926 		break;
2927 	}
2928 
2929 	case DIOCRCLRTABLES: {
2930 		struct pfioc_table *io = (struct pfioc_table *)addr;
2931 
2932 		if (io->pfrio_esize != 0) {
2933 			error = ENODEV;
2934 			break;
2935 		}
2936 		PF_RULES_WLOCK();
2937 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2938 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2939 		PF_RULES_WUNLOCK();
2940 		break;
2941 	}
2942 
2943 	case DIOCRADDTABLES: {
2944 		struct pfioc_table *io = (struct pfioc_table *)addr;
2945 		struct pfr_table *pfrts;
2946 		size_t totlen;
2947 
2948 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2949 			error = ENODEV;
2950 			break;
2951 		}
2952 
2953 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
2954 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
2955 			error = ENOMEM;
2956 			break;
2957 		}
2958 
2959 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2960 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
2961 		    M_TEMP, M_WAITOK);
2962 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2963 		if (error) {
2964 			free(pfrts, M_TEMP);
2965 			break;
2966 		}
2967 		PF_RULES_WLOCK();
2968 		error = pfr_add_tables(pfrts, io->pfrio_size,
2969 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2970 		PF_RULES_WUNLOCK();
2971 		free(pfrts, M_TEMP);
2972 		break;
2973 	}
2974 
2975 	case DIOCRDELTABLES: {
2976 		struct pfioc_table *io = (struct pfioc_table *)addr;
2977 		struct pfr_table *pfrts;
2978 		size_t totlen;
2979 
2980 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2981 			error = ENODEV;
2982 			break;
2983 		}
2984 
2985 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
2986 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
2987 			error = ENOMEM;
2988 			break;
2989 		}
2990 
2991 		totlen = io->pfrio_size * sizeof(struct pfr_table);
2992 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
2993 		    M_TEMP, M_WAITOK);
2994 		error = copyin(io->pfrio_buffer, pfrts, totlen);
2995 		if (error) {
2996 			free(pfrts, M_TEMP);
2997 			break;
2998 		}
2999 		PF_RULES_WLOCK();
3000 		error = pfr_del_tables(pfrts, io->pfrio_size,
3001 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3002 		PF_RULES_WUNLOCK();
3003 		free(pfrts, M_TEMP);
3004 		break;
3005 	}
3006 
3007 	case DIOCRGETTABLES: {
3008 		struct pfioc_table *io = (struct pfioc_table *)addr;
3009 		struct pfr_table *pfrts;
3010 		size_t totlen;
3011 		int n;
3012 
3013 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3014 			error = ENODEV;
3015 			break;
3016 		}
3017 		PF_RULES_RLOCK();
3018 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3019 		if (n < 0) {
3020 			PF_RULES_RUNLOCK();
3021 			error = EINVAL;
3022 			break;
3023 		}
3024 		io->pfrio_size = min(io->pfrio_size, n);
3025 
3026 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3027 
3028 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3029 		    M_TEMP, M_NOWAIT);
3030 		if (pfrts == NULL) {
3031 			error = ENOMEM;
3032 			PF_RULES_RUNLOCK();
3033 			break;
3034 		}
3035 		error = pfr_get_tables(&io->pfrio_table, pfrts,
3036 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3037 		PF_RULES_RUNLOCK();
3038 		if (error == 0)
3039 			error = copyout(pfrts, io->pfrio_buffer, totlen);
3040 		free(pfrts, M_TEMP);
3041 		break;
3042 	}
3043 
3044 	case DIOCRGETTSTATS: {
3045 		struct pfioc_table *io = (struct pfioc_table *)addr;
3046 		struct pfr_tstats *pfrtstats;
3047 		size_t totlen;
3048 		int n;
3049 
3050 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
3051 			error = ENODEV;
3052 			break;
3053 		}
3054 		PF_RULES_WLOCK();
3055 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3056 		if (n < 0) {
3057 			PF_RULES_WUNLOCK();
3058 			error = EINVAL;
3059 			break;
3060 		}
3061 		io->pfrio_size = min(io->pfrio_size, n);
3062 
3063 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
3064 		pfrtstats = mallocarray(io->pfrio_size,
3065 		    sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT);
3066 		if (pfrtstats == NULL) {
3067 			error = ENOMEM;
3068 			PF_RULES_WUNLOCK();
3069 			break;
3070 		}
3071 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
3072 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3073 		PF_RULES_WUNLOCK();
3074 		if (error == 0)
3075 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
3076 		free(pfrtstats, M_TEMP);
3077 		break;
3078 	}
3079 
3080 	case DIOCRCLRTSTATS: {
3081 		struct pfioc_table *io = (struct pfioc_table *)addr;
3082 		struct pfr_table *pfrts;
3083 		size_t totlen;
3084 		int n;
3085 
3086 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3087 			error = ENODEV;
3088 			break;
3089 		}
3090 
3091 		PF_RULES_WLOCK();
3092 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3093 		if (n < 0) {
3094 			PF_RULES_WUNLOCK();
3095 			error = EINVAL;
3096 			break;
3097 		}
3098 		io->pfrio_size = min(io->pfrio_size, n);
3099 
3100 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3101 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3102 		    M_TEMP, M_NOWAIT);
3103 		if (pfrts == NULL) {
3104 			error = ENOMEM;
3105 			PF_RULES_WUNLOCK();
3106 			break;
3107 		}
3108 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3109 		if (error) {
3110 			free(pfrts, M_TEMP);
3111 			PF_RULES_WUNLOCK();
3112 			break;
3113 		}
3114 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
3115 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3116 		PF_RULES_WUNLOCK();
3117 		free(pfrts, M_TEMP);
3118 		break;
3119 	}
3120 
3121 	case DIOCRSETTFLAGS: {
3122 		struct pfioc_table *io = (struct pfioc_table *)addr;
3123 		struct pfr_table *pfrts;
3124 		size_t totlen;
3125 		int n;
3126 
3127 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3128 			error = ENODEV;
3129 			break;
3130 		}
3131 
3132 		PF_RULES_RLOCK();
3133 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3134 		if (n < 0) {
3135 			PF_RULES_RUNLOCK();
3136 			error = EINVAL;
3137 			break;
3138 		}
3139 
3140 		io->pfrio_size = min(io->pfrio_size, n);
3141 		PF_RULES_RUNLOCK();
3142 
3143 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3144 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3145 		    M_TEMP, M_WAITOK);
3146 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3147 		if (error) {
3148 			free(pfrts, M_TEMP);
3149 			break;
3150 		}
3151 		PF_RULES_WLOCK();
3152 		error = pfr_set_tflags(pfrts, io->pfrio_size,
3153 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
3154 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3155 		PF_RULES_WUNLOCK();
3156 		free(pfrts, M_TEMP);
3157 		break;
3158 	}
3159 
3160 	case DIOCRCLRADDRS: {
3161 		struct pfioc_table *io = (struct pfioc_table *)addr;
3162 
3163 		if (io->pfrio_esize != 0) {
3164 			error = ENODEV;
3165 			break;
3166 		}
3167 		PF_RULES_WLOCK();
3168 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
3169 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
3170 		PF_RULES_WUNLOCK();
3171 		break;
3172 	}
3173 
3174 	case DIOCRADDADDRS: {
3175 		struct pfioc_table *io = (struct pfioc_table *)addr;
3176 		struct pfr_addr *pfras;
3177 		size_t totlen;
3178 
3179 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3180 			error = ENODEV;
3181 			break;
3182 		}
3183 		if (io->pfrio_size < 0 ||
3184 		    io->pfrio_size > pf_ioctl_maxcount ||
3185 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3186 			error = EINVAL;
3187 			break;
3188 		}
3189 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3190 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3191 		    M_TEMP, M_NOWAIT);
3192 		if (! pfras) {
3193 			error = ENOMEM;
3194 			break;
3195 		}
3196 		error = copyin(io->pfrio_buffer, pfras, totlen);
3197 		if (error) {
3198 			free(pfras, M_TEMP);
3199 			break;
3200 		}
3201 		PF_RULES_WLOCK();
3202 		error = pfr_add_addrs(&io->pfrio_table, pfras,
3203 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
3204 		    PFR_FLAG_USERIOCTL);
3205 		PF_RULES_WUNLOCK();
3206 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3207 			error = copyout(pfras, io->pfrio_buffer, totlen);
3208 		free(pfras, M_TEMP);
3209 		break;
3210 	}
3211 
3212 	case DIOCRDELADDRS: {
3213 		struct pfioc_table *io = (struct pfioc_table *)addr;
3214 		struct pfr_addr *pfras;
3215 		size_t totlen;
3216 
3217 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3218 			error = ENODEV;
3219 			break;
3220 		}
3221 		if (io->pfrio_size < 0 ||
3222 		    io->pfrio_size > pf_ioctl_maxcount ||
3223 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3224 			error = EINVAL;
3225 			break;
3226 		}
3227 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3228 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3229 		    M_TEMP, M_NOWAIT);
3230 		if (! pfras) {
3231 			error = ENOMEM;
3232 			break;
3233 		}
3234 		error = copyin(io->pfrio_buffer, pfras, totlen);
3235 		if (error) {
3236 			free(pfras, M_TEMP);
3237 			break;
3238 		}
3239 		PF_RULES_WLOCK();
3240 		error = pfr_del_addrs(&io->pfrio_table, pfras,
3241 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
3242 		    PFR_FLAG_USERIOCTL);
3243 		PF_RULES_WUNLOCK();
3244 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3245 			error = copyout(pfras, io->pfrio_buffer, totlen);
3246 		free(pfras, M_TEMP);
3247 		break;
3248 	}
3249 
3250 	case DIOCRSETADDRS: {
3251 		struct pfioc_table *io = (struct pfioc_table *)addr;
3252 		struct pfr_addr *pfras;
3253 		size_t totlen, count;
3254 
3255 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3256 			error = ENODEV;
3257 			break;
3258 		}
3259 		if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
3260 			error = EINVAL;
3261 			break;
3262 		}
3263 		count = max(io->pfrio_size, io->pfrio_size2);
3264 		if (count > pf_ioctl_maxcount ||
3265 		    WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
3266 			error = EINVAL;
3267 			break;
3268 		}
3269 		totlen = count * sizeof(struct pfr_addr);
3270 		pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
3271 		    M_NOWAIT);
3272 		if (! pfras) {
3273 			error = ENOMEM;
3274 			break;
3275 		}
3276 		error = copyin(io->pfrio_buffer, pfras, totlen);
3277 		if (error) {
3278 			free(pfras, M_TEMP);
3279 			break;
3280 		}
3281 		PF_RULES_WLOCK();
3282 		error = pfr_set_addrs(&io->pfrio_table, pfras,
3283 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
3284 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
3285 		    PFR_FLAG_USERIOCTL, 0);
3286 		PF_RULES_WUNLOCK();
3287 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3288 			error = copyout(pfras, io->pfrio_buffer, totlen);
3289 		free(pfras, M_TEMP);
3290 		break;
3291 	}
3292 
3293 	case DIOCRGETADDRS: {
3294 		struct pfioc_table *io = (struct pfioc_table *)addr;
3295 		struct pfr_addr *pfras;
3296 		size_t totlen;
3297 
3298 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3299 			error = ENODEV;
3300 			break;
3301 		}
3302 		if (io->pfrio_size < 0 ||
3303 		    io->pfrio_size > pf_ioctl_maxcount ||
3304 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3305 			error = EINVAL;
3306 			break;
3307 		}
3308 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3309 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3310 		    M_TEMP, M_NOWAIT);
3311 		if (! pfras) {
3312 			error = ENOMEM;
3313 			break;
3314 		}
3315 		PF_RULES_RLOCK();
3316 		error = pfr_get_addrs(&io->pfrio_table, pfras,
3317 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3318 		PF_RULES_RUNLOCK();
3319 		if (error == 0)
3320 			error = copyout(pfras, io->pfrio_buffer, totlen);
3321 		free(pfras, M_TEMP);
3322 		break;
3323 	}
3324 
3325 	case DIOCRGETASTATS: {
3326 		struct pfioc_table *io = (struct pfioc_table *)addr;
3327 		struct pfr_astats *pfrastats;
3328 		size_t totlen;
3329 
3330 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
3331 			error = ENODEV;
3332 			break;
3333 		}
3334 		if (io->pfrio_size < 0 ||
3335 		    io->pfrio_size > pf_ioctl_maxcount ||
3336 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
3337 			error = EINVAL;
3338 			break;
3339 		}
3340 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
3341 		pfrastats = mallocarray(io->pfrio_size,
3342 		    sizeof(struct pfr_astats), M_TEMP, M_NOWAIT);
3343 		if (! pfrastats) {
3344 			error = ENOMEM;
3345 			break;
3346 		}
3347 		PF_RULES_RLOCK();
3348 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
3349 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3350 		PF_RULES_RUNLOCK();
3351 		if (error == 0)
3352 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
3353 		free(pfrastats, M_TEMP);
3354 		break;
3355 	}
3356 
3357 	case DIOCRCLRASTATS: {
3358 		struct pfioc_table *io = (struct pfioc_table *)addr;
3359 		struct pfr_addr *pfras;
3360 		size_t totlen;
3361 
3362 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3363 			error = ENODEV;
3364 			break;
3365 		}
3366 		if (io->pfrio_size < 0 ||
3367 		    io->pfrio_size > pf_ioctl_maxcount ||
3368 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3369 			error = EINVAL;
3370 			break;
3371 		}
3372 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3373 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3374 		    M_TEMP, M_NOWAIT);
3375 		if (! pfras) {
3376 			error = ENOMEM;
3377 			break;
3378 		}
3379 		error = copyin(io->pfrio_buffer, pfras, totlen);
3380 		if (error) {
3381 			free(pfras, M_TEMP);
3382 			break;
3383 		}
3384 		PF_RULES_WLOCK();
3385 		error = pfr_clr_astats(&io->pfrio_table, pfras,
3386 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
3387 		    PFR_FLAG_USERIOCTL);
3388 		PF_RULES_WUNLOCK();
3389 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3390 			error = copyout(pfras, io->pfrio_buffer, totlen);
3391 		free(pfras, M_TEMP);
3392 		break;
3393 	}
3394 
3395 	case DIOCRTSTADDRS: {
3396 		struct pfioc_table *io = (struct pfioc_table *)addr;
3397 		struct pfr_addr *pfras;
3398 		size_t totlen;
3399 
3400 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3401 			error = ENODEV;
3402 			break;
3403 		}
3404 		if (io->pfrio_size < 0 ||
3405 		    io->pfrio_size > pf_ioctl_maxcount ||
3406 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3407 			error = EINVAL;
3408 			break;
3409 		}
3410 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3411 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3412 		    M_TEMP, M_NOWAIT);
3413 		if (! pfras) {
3414 			error = ENOMEM;
3415 			break;
3416 		}
3417 		error = copyin(io->pfrio_buffer, pfras, totlen);
3418 		if (error) {
3419 			free(pfras, M_TEMP);
3420 			break;
3421 		}
3422 		PF_RULES_RLOCK();
3423 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
3424 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
3425 		    PFR_FLAG_USERIOCTL);
3426 		PF_RULES_RUNLOCK();
3427 		if (error == 0)
3428 			error = copyout(pfras, io->pfrio_buffer, totlen);
3429 		free(pfras, M_TEMP);
3430 		break;
3431 	}
3432 
3433 	case DIOCRINADEFINE: {
3434 		struct pfioc_table *io = (struct pfioc_table *)addr;
3435 		struct pfr_addr *pfras;
3436 		size_t totlen;
3437 
3438 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3439 			error = ENODEV;
3440 			break;
3441 		}
3442 		if (io->pfrio_size < 0 ||
3443 		    io->pfrio_size > pf_ioctl_maxcount ||
3444 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3445 			error = EINVAL;
3446 			break;
3447 		}
3448 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3449 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3450 		    M_TEMP, M_NOWAIT);
3451 		if (! pfras) {
3452 			error = ENOMEM;
3453 			break;
3454 		}
3455 		error = copyin(io->pfrio_buffer, pfras, totlen);
3456 		if (error) {
3457 			free(pfras, M_TEMP);
3458 			break;
3459 		}
3460 		PF_RULES_WLOCK();
3461 		error = pfr_ina_define(&io->pfrio_table, pfras,
3462 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
3463 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3464 		PF_RULES_WUNLOCK();
3465 		free(pfras, M_TEMP);
3466 		break;
3467 	}
3468 
3469 	case DIOCOSFPADD: {
3470 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
3471 		PF_RULES_WLOCK();
3472 		error = pf_osfp_add(io);
3473 		PF_RULES_WUNLOCK();
3474 		break;
3475 	}
3476 
3477 	case DIOCOSFPGET: {
3478 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
3479 		PF_RULES_RLOCK();
3480 		error = pf_osfp_get(io);
3481 		PF_RULES_RUNLOCK();
3482 		break;
3483 	}
3484 
3485 	case DIOCXBEGIN: {
3486 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3487 		struct pfioc_trans_e	*ioes, *ioe;
3488 		size_t			 totlen;
3489 		int			 i;
3490 
3491 		if (io->esize != sizeof(*ioe)) {
3492 			error = ENODEV;
3493 			break;
3494 		}
3495 		if (io->size < 0 ||
3496 		    io->size > pf_ioctl_maxcount ||
3497 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3498 			error = EINVAL;
3499 			break;
3500 		}
3501 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3502 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3503 		    M_TEMP, M_NOWAIT);
3504 		if (! ioes) {
3505 			error = ENOMEM;
3506 			break;
3507 		}
3508 		error = copyin(io->array, ioes, totlen);
3509 		if (error) {
3510 			free(ioes, M_TEMP);
3511 			break;
3512 		}
3513 		PF_RULES_WLOCK();
3514 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3515 			switch (ioe->rs_num) {
3516 #ifdef ALTQ
3517 			case PF_RULESET_ALTQ:
3518 				if (ioe->anchor[0]) {
3519 					PF_RULES_WUNLOCK();
3520 					free(ioes, M_TEMP);
3521 					error = EINVAL;
3522 					goto fail;
3523 				}
3524 				if ((error = pf_begin_altq(&ioe->ticket))) {
3525 					PF_RULES_WUNLOCK();
3526 					free(ioes, M_TEMP);
3527 					goto fail;
3528 				}
3529 				break;
3530 #endif /* ALTQ */
3531 			case PF_RULESET_TABLE:
3532 			    {
3533 				struct pfr_table table;
3534 
3535 				bzero(&table, sizeof(table));
3536 				strlcpy(table.pfrt_anchor, ioe->anchor,
3537 				    sizeof(table.pfrt_anchor));
3538 				if ((error = pfr_ina_begin(&table,
3539 				    &ioe->ticket, NULL, 0))) {
3540 					PF_RULES_WUNLOCK();
3541 					free(ioes, M_TEMP);
3542 					goto fail;
3543 				}
3544 				break;
3545 			    }
3546 			default:
3547 				if ((error = pf_begin_rules(&ioe->ticket,
3548 				    ioe->rs_num, ioe->anchor))) {
3549 					PF_RULES_WUNLOCK();
3550 					free(ioes, M_TEMP);
3551 					goto fail;
3552 				}
3553 				break;
3554 			}
3555 		}
3556 		PF_RULES_WUNLOCK();
3557 		error = copyout(ioes, io->array, totlen);
3558 		free(ioes, M_TEMP);
3559 		break;
3560 	}
3561 
3562 	case DIOCXROLLBACK: {
3563 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3564 		struct pfioc_trans_e	*ioe, *ioes;
3565 		size_t			 totlen;
3566 		int			 i;
3567 
3568 		if (io->esize != sizeof(*ioe)) {
3569 			error = ENODEV;
3570 			break;
3571 		}
3572 		if (io->size < 0 ||
3573 		    io->size > pf_ioctl_maxcount ||
3574 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3575 			error = EINVAL;
3576 			break;
3577 		}
3578 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3579 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3580 		    M_TEMP, M_NOWAIT);
3581 		if (! ioes) {
3582 			error = ENOMEM;
3583 			break;
3584 		}
3585 		error = copyin(io->array, ioes, totlen);
3586 		if (error) {
3587 			free(ioes, M_TEMP);
3588 			break;
3589 		}
3590 		PF_RULES_WLOCK();
3591 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3592 			switch (ioe->rs_num) {
3593 #ifdef ALTQ
3594 			case PF_RULESET_ALTQ:
3595 				if (ioe->anchor[0]) {
3596 					PF_RULES_WUNLOCK();
3597 					free(ioes, M_TEMP);
3598 					error = EINVAL;
3599 					goto fail;
3600 				}
3601 				if ((error = pf_rollback_altq(ioe->ticket))) {
3602 					PF_RULES_WUNLOCK();
3603 					free(ioes, M_TEMP);
3604 					goto fail; /* really bad */
3605 				}
3606 				break;
3607 #endif /* ALTQ */
3608 			case PF_RULESET_TABLE:
3609 			    {
3610 				struct pfr_table table;
3611 
3612 				bzero(&table, sizeof(table));
3613 				strlcpy(table.pfrt_anchor, ioe->anchor,
3614 				    sizeof(table.pfrt_anchor));
3615 				if ((error = pfr_ina_rollback(&table,
3616 				    ioe->ticket, NULL, 0))) {
3617 					PF_RULES_WUNLOCK();
3618 					free(ioes, M_TEMP);
3619 					goto fail; /* really bad */
3620 				}
3621 				break;
3622 			    }
3623 			default:
3624 				if ((error = pf_rollback_rules(ioe->ticket,
3625 				    ioe->rs_num, ioe->anchor))) {
3626 					PF_RULES_WUNLOCK();
3627 					free(ioes, M_TEMP);
3628 					goto fail; /* really bad */
3629 				}
3630 				break;
3631 			}
3632 		}
3633 		PF_RULES_WUNLOCK();
3634 		free(ioes, M_TEMP);
3635 		break;
3636 	}
3637 
3638 	case DIOCXCOMMIT: {
3639 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3640 		struct pfioc_trans_e	*ioe, *ioes;
3641 		struct pf_ruleset	*rs;
3642 		size_t			 totlen;
3643 		int			 i;
3644 
3645 		if (io->esize != sizeof(*ioe)) {
3646 			error = ENODEV;
3647 			break;
3648 		}
3649 
3650 		if (io->size < 0 ||
3651 		    io->size > pf_ioctl_maxcount ||
3652 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3653 			error = EINVAL;
3654 			break;
3655 		}
3656 
3657 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3658 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3659 		    M_TEMP, M_NOWAIT);
3660 		if (ioes == NULL) {
3661 			error = ENOMEM;
3662 			break;
3663 		}
3664 		error = copyin(io->array, ioes, totlen);
3665 		if (error) {
3666 			free(ioes, M_TEMP);
3667 			break;
3668 		}
3669 		PF_RULES_WLOCK();
3670 		/* First makes sure everything will succeed. */
3671 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3672 			switch (ioe->rs_num) {
3673 #ifdef ALTQ
3674 			case PF_RULESET_ALTQ:
3675 				if (ioe->anchor[0]) {
3676 					PF_RULES_WUNLOCK();
3677 					free(ioes, M_TEMP);
3678 					error = EINVAL;
3679 					goto fail;
3680 				}
3681 				if (!V_altqs_inactive_open || ioe->ticket !=
3682 				    V_ticket_altqs_inactive) {
3683 					PF_RULES_WUNLOCK();
3684 					free(ioes, M_TEMP);
3685 					error = EBUSY;
3686 					goto fail;
3687 				}
3688 				break;
3689 #endif /* ALTQ */
3690 			case PF_RULESET_TABLE:
3691 				rs = pf_find_ruleset(ioe->anchor);
3692 				if (rs == NULL || !rs->topen || ioe->ticket !=
3693 				    rs->tticket) {
3694 					PF_RULES_WUNLOCK();
3695 					free(ioes, M_TEMP);
3696 					error = EBUSY;
3697 					goto fail;
3698 				}
3699 				break;
3700 			default:
3701 				if (ioe->rs_num < 0 || ioe->rs_num >=
3702 				    PF_RULESET_MAX) {
3703 					PF_RULES_WUNLOCK();
3704 					free(ioes, M_TEMP);
3705 					error = EINVAL;
3706 					goto fail;
3707 				}
3708 				rs = pf_find_ruleset(ioe->anchor);
3709 				if (rs == NULL ||
3710 				    !rs->rules[ioe->rs_num].inactive.open ||
3711 				    rs->rules[ioe->rs_num].inactive.ticket !=
3712 				    ioe->ticket) {
3713 					PF_RULES_WUNLOCK();
3714 					free(ioes, M_TEMP);
3715 					error = EBUSY;
3716 					goto fail;
3717 				}
3718 				break;
3719 			}
3720 		}
3721 		/* Now do the commit - no errors should happen here. */
3722 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3723 			switch (ioe->rs_num) {
3724 #ifdef ALTQ
3725 			case PF_RULESET_ALTQ:
3726 				if ((error = pf_commit_altq(ioe->ticket))) {
3727 					PF_RULES_WUNLOCK();
3728 					free(ioes, M_TEMP);
3729 					goto fail; /* really bad */
3730 				}
3731 				break;
3732 #endif /* ALTQ */
3733 			case PF_RULESET_TABLE:
3734 			    {
3735 				struct pfr_table table;
3736 
3737 				bzero(&table, sizeof(table));
3738 				strlcpy(table.pfrt_anchor, ioe->anchor,
3739 				    sizeof(table.pfrt_anchor));
3740 				if ((error = pfr_ina_commit(&table,
3741 				    ioe->ticket, NULL, NULL, 0))) {
3742 					PF_RULES_WUNLOCK();
3743 					free(ioes, M_TEMP);
3744 					goto fail; /* really bad */
3745 				}
3746 				break;
3747 			    }
3748 			default:
3749 				if ((error = pf_commit_rules(ioe->ticket,
3750 				    ioe->rs_num, ioe->anchor))) {
3751 					PF_RULES_WUNLOCK();
3752 					free(ioes, M_TEMP);
3753 					goto fail; /* really bad */
3754 				}
3755 				break;
3756 			}
3757 		}
3758 		PF_RULES_WUNLOCK();
3759 		free(ioes, M_TEMP);
3760 		break;
3761 	}
3762 
3763 	case DIOCGETSRCNODES: {
3764 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
3765 		struct pf_srchash	*sh;
3766 		struct pf_src_node	*n, *p, *pstore;
3767 		uint32_t		 i, nr = 0;
3768 
3769 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3770 				i++, sh++) {
3771 			PF_HASHROW_LOCK(sh);
3772 			LIST_FOREACH(n, &sh->nodes, entry)
3773 				nr++;
3774 			PF_HASHROW_UNLOCK(sh);
3775 		}
3776 
3777 		psn->psn_len = min(psn->psn_len,
3778 		    sizeof(struct pf_src_node) * nr);
3779 
3780 		if (psn->psn_len == 0) {
3781 			psn->psn_len = sizeof(struct pf_src_node) * nr;
3782 			break;
3783 		}
3784 
3785 		nr = 0;
3786 
3787 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
3788 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
3789 		    i++, sh++) {
3790 		    PF_HASHROW_LOCK(sh);
3791 		    LIST_FOREACH(n, &sh->nodes, entry) {
3792 			int	secs = time_uptime, diff;
3793 
3794 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
3795 				break;
3796 
3797 			bcopy(n, p, sizeof(struct pf_src_node));
3798 			if (n->rule.ptr != NULL)
3799 				p->rule.nr = n->rule.ptr->nr;
3800 			p->creation = secs - p->creation;
3801 			if (p->expire > secs)
3802 				p->expire -= secs;
3803 			else
3804 				p->expire = 0;
3805 
3806 			/* Adjust the connection rate estimate. */
3807 			diff = secs - n->conn_rate.last;
3808 			if (diff >= n->conn_rate.seconds)
3809 				p->conn_rate.count = 0;
3810 			else
3811 				p->conn_rate.count -=
3812 				    n->conn_rate.count * diff /
3813 				    n->conn_rate.seconds;
3814 			p++;
3815 			nr++;
3816 		    }
3817 		    PF_HASHROW_UNLOCK(sh);
3818 		}
3819 		error = copyout(pstore, psn->psn_src_nodes,
3820 		    sizeof(struct pf_src_node) * nr);
3821 		if (error) {
3822 			free(pstore, M_TEMP);
3823 			break;
3824 		}
3825 		psn->psn_len = sizeof(struct pf_src_node) * nr;
3826 		free(pstore, M_TEMP);
3827 		break;
3828 	}
3829 
3830 	case DIOCCLRSRCNODES: {
3831 		pf_clear_srcnodes(NULL);
3832 		pf_purge_expired_src_nodes();
3833 		break;
3834 	}
3835 
3836 	case DIOCKILLSRCNODES:
3837 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
3838 		break;
3839 
3840 	case DIOCSETHOSTID: {
3841 		u_int32_t	*hostid = (u_int32_t *)addr;
3842 
3843 		PF_RULES_WLOCK();
3844 		if (*hostid == 0)
3845 			V_pf_status.hostid = arc4random();
3846 		else
3847 			V_pf_status.hostid = *hostid;
3848 		PF_RULES_WUNLOCK();
3849 		break;
3850 	}
3851 
3852 	case DIOCOSFPFLUSH:
3853 		PF_RULES_WLOCK();
3854 		pf_osfp_flush();
3855 		PF_RULES_WUNLOCK();
3856 		break;
3857 
3858 	case DIOCIGETIFACES: {
3859 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3860 		struct pfi_kif *ifstore;
3861 		size_t bufsiz;
3862 
3863 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
3864 			error = ENODEV;
3865 			break;
3866 		}
3867 
3868 		if (io->pfiio_size < 0 ||
3869 		    io->pfiio_size > pf_ioctl_maxcount ||
3870 		    WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
3871 			error = EINVAL;
3872 			break;
3873 		}
3874 
3875 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
3876 		ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
3877 		    M_TEMP, M_NOWAIT);
3878 		if (ifstore == NULL) {
3879 			error = ENOMEM;
3880 			break;
3881 		}
3882 
3883 		PF_RULES_RLOCK();
3884 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
3885 		PF_RULES_RUNLOCK();
3886 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
3887 		free(ifstore, M_TEMP);
3888 		break;
3889 	}
3890 
3891 	case DIOCSETIFFLAG: {
3892 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3893 
3894 		PF_RULES_WLOCK();
3895 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3896 		PF_RULES_WUNLOCK();
3897 		break;
3898 	}
3899 
3900 	case DIOCCLRIFFLAG: {
3901 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3902 
3903 		PF_RULES_WLOCK();
3904 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3905 		PF_RULES_WUNLOCK();
3906 		break;
3907 	}
3908 
3909 	default:
3910 		error = ENODEV;
3911 		break;
3912 	}
3913 fail:
3914 	if (sx_xlocked(&pf_ioctl_lock))
3915 		sx_xunlock(&pf_ioctl_lock);
3916 	CURVNET_RESTORE();
3917 
3918 	return (error);
3919 }
3920 
3921 void
3922 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
3923 {
3924 	bzero(sp, sizeof(struct pfsync_state));
3925 
3926 	/* copy from state key */
3927 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
3928 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
3929 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
3930 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
3931 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
3932 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
3933 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
3934 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
3935 	sp->proto = st->key[PF_SK_WIRE]->proto;
3936 	sp->af = st->key[PF_SK_WIRE]->af;
3937 
3938 	/* copy from state */
3939 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
3940 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
3941 	sp->creation = htonl(time_uptime - st->creation);
3942 	sp->expire = pf_state_expires(st);
3943 	if (sp->expire <= time_uptime)
3944 		sp->expire = htonl(0);
3945 	else
3946 		sp->expire = htonl(sp->expire - time_uptime);
3947 
3948 	sp->direction = st->direction;
3949 	sp->log = st->log;
3950 	sp->timeout = st->timeout;
3951 	sp->state_flags = st->state_flags;
3952 	if (st->src_node)
3953 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
3954 	if (st->nat_src_node)
3955 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
3956 
3957 	sp->id = st->id;
3958 	sp->creatorid = st->creatorid;
3959 	pf_state_peer_hton(&st->src, &sp->src);
3960 	pf_state_peer_hton(&st->dst, &sp->dst);
3961 
3962 	if (st->rule.ptr == NULL)
3963 		sp->rule = htonl(-1);
3964 	else
3965 		sp->rule = htonl(st->rule.ptr->nr);
3966 	if (st->anchor.ptr == NULL)
3967 		sp->anchor = htonl(-1);
3968 	else
3969 		sp->anchor = htonl(st->anchor.ptr->nr);
3970 	if (st->nat_rule.ptr == NULL)
3971 		sp->nat_rule = htonl(-1);
3972 	else
3973 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
3974 
3975 	pf_state_counter_hton(st->packets[0], sp->packets[0]);
3976 	pf_state_counter_hton(st->packets[1], sp->packets[1]);
3977 	pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
3978 	pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
3979 
3980 }
3981 
3982 static void
3983 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
3984 {
3985 	struct pfr_ktable *kt;
3986 
3987 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
3988 
3989 	kt = aw->p.tbl;
3990 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
3991 		kt = kt->pfrkt_root;
3992 	aw->p.tbl = NULL;
3993 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
3994 		kt->pfrkt_cnt : -1;
3995 }
3996 
3997 /*
3998  * XXX - Check for version missmatch!!!
3999  */
4000 static void
4001 pf_clear_states(void)
4002 {
4003 	struct pf_state	*s;
4004 	u_int i;
4005 
4006 	for (i = 0; i <= pf_hashmask; i++) {
4007 		struct pf_idhash *ih = &V_pf_idhash[i];
4008 relock:
4009 		PF_HASHROW_LOCK(ih);
4010 		LIST_FOREACH(s, &ih->states, entry) {
4011 			s->timeout = PFTM_PURGE;
4012 			/* Don't send out individual delete messages. */
4013 			s->state_flags |= PFSTATE_NOSYNC;
4014 			pf_unlink_state(s, PF_ENTER_LOCKED);
4015 			goto relock;
4016 		}
4017 		PF_HASHROW_UNLOCK(ih);
4018 	}
4019 }
4020 
4021 static int
4022 pf_clear_tables(void)
4023 {
4024 	struct pfioc_table io;
4025 	int error;
4026 
4027 	bzero(&io, sizeof(io));
4028 
4029 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
4030 	    io.pfrio_flags);
4031 
4032 	return (error);
4033 }
4034 
4035 static void
4036 pf_clear_srcnodes(struct pf_src_node *n)
4037 {
4038 	struct pf_state *s;
4039 	int i;
4040 
4041 	for (i = 0; i <= pf_hashmask; i++) {
4042 		struct pf_idhash *ih = &V_pf_idhash[i];
4043 
4044 		PF_HASHROW_LOCK(ih);
4045 		LIST_FOREACH(s, &ih->states, entry) {
4046 			if (n == NULL || n == s->src_node)
4047 				s->src_node = NULL;
4048 			if (n == NULL || n == s->nat_src_node)
4049 				s->nat_src_node = NULL;
4050 		}
4051 		PF_HASHROW_UNLOCK(ih);
4052 	}
4053 
4054 	if (n == NULL) {
4055 		struct pf_srchash *sh;
4056 
4057 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4058 		    i++, sh++) {
4059 			PF_HASHROW_LOCK(sh);
4060 			LIST_FOREACH(n, &sh->nodes, entry) {
4061 				n->expire = 1;
4062 				n->states = 0;
4063 			}
4064 			PF_HASHROW_UNLOCK(sh);
4065 		}
4066 	} else {
4067 		/* XXX: hash slot should already be locked here. */
4068 		n->expire = 1;
4069 		n->states = 0;
4070 	}
4071 }
4072 
4073 static void
4074 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
4075 {
4076 	struct pf_src_node_list	 kill;
4077 
4078 	LIST_INIT(&kill);
4079 	for (int i = 0; i <= pf_srchashmask; i++) {
4080 		struct pf_srchash *sh = &V_pf_srchash[i];
4081 		struct pf_src_node *sn, *tmp;
4082 
4083 		PF_HASHROW_LOCK(sh);
4084 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
4085 			if (PF_MATCHA(psnk->psnk_src.neg,
4086 			      &psnk->psnk_src.addr.v.a.addr,
4087 			      &psnk->psnk_src.addr.v.a.mask,
4088 			      &sn->addr, sn->af) &&
4089 			    PF_MATCHA(psnk->psnk_dst.neg,
4090 			      &psnk->psnk_dst.addr.v.a.addr,
4091 			      &psnk->psnk_dst.addr.v.a.mask,
4092 			      &sn->raddr, sn->af)) {
4093 				pf_unlink_src_node(sn);
4094 				LIST_INSERT_HEAD(&kill, sn, entry);
4095 				sn->expire = 1;
4096 			}
4097 		PF_HASHROW_UNLOCK(sh);
4098 	}
4099 
4100 	for (int i = 0; i <= pf_hashmask; i++) {
4101 		struct pf_idhash *ih = &V_pf_idhash[i];
4102 		struct pf_state *s;
4103 
4104 		PF_HASHROW_LOCK(ih);
4105 		LIST_FOREACH(s, &ih->states, entry) {
4106 			if (s->src_node && s->src_node->expire == 1)
4107 				s->src_node = NULL;
4108 			if (s->nat_src_node && s->nat_src_node->expire == 1)
4109 				s->nat_src_node = NULL;
4110 		}
4111 		PF_HASHROW_UNLOCK(ih);
4112 	}
4113 
4114 	psnk->psnk_killed = pf_free_src_nodes(&kill);
4115 }
4116 
4117 /*
4118  * XXX - Check for version missmatch!!!
4119  */
4120 
4121 /*
4122  * Duplicate pfctl -Fa operation to get rid of as much as we can.
4123  */
4124 static int
4125 shutdown_pf(void)
4126 {
4127 	int error = 0;
4128 	u_int32_t t[5];
4129 	char nn = '\0';
4130 
4131 	do {
4132 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
4133 		    != 0) {
4134 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
4135 			break;
4136 		}
4137 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
4138 		    != 0) {
4139 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
4140 			break;		/* XXX: rollback? */
4141 		}
4142 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
4143 		    != 0) {
4144 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
4145 			break;		/* XXX: rollback? */
4146 		}
4147 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
4148 		    != 0) {
4149 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
4150 			break;		/* XXX: rollback? */
4151 		}
4152 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
4153 		    != 0) {
4154 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
4155 			break;		/* XXX: rollback? */
4156 		}
4157 
4158 		/* XXX: these should always succeed here */
4159 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
4160 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
4161 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
4162 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
4163 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
4164 
4165 		if ((error = pf_clear_tables()) != 0)
4166 			break;
4167 
4168 #ifdef ALTQ
4169 		if ((error = pf_begin_altq(&t[0])) != 0) {
4170 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
4171 			break;
4172 		}
4173 		pf_commit_altq(t[0]);
4174 #endif
4175 
4176 		pf_clear_states();
4177 
4178 		pf_clear_srcnodes(NULL);
4179 
4180 		/* status does not use malloced mem so no need to cleanup */
4181 		/* fingerprints and interfaces have their own cleanup code */
4182 	} while(0);
4183 
4184 	return (error);
4185 }
4186 
4187 static pfil_return_t
4188 pf_check_return(int chk, struct mbuf **m)
4189 {
4190 
4191 	switch (chk) {
4192 	case PF_PASS:
4193 		if (*m == NULL)
4194 			return (PFIL_CONSUMED);
4195 		else
4196 			return (PFIL_PASS);
4197 		break;
4198 	default:
4199 		if (*m != NULL) {
4200 			m_freem(*m);
4201 			*m = NULL;
4202 		}
4203 		return (PFIL_DROPPED);
4204 	}
4205 }
4206 
4207 #ifdef INET
4208 static pfil_return_t
4209 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
4210     void *ruleset __unused, struct inpcb *inp)
4211 {
4212 	int chk;
4213 
4214 	chk = pf_test(PF_IN, flags, ifp, m, inp);
4215 
4216 	return (pf_check_return(chk, m));
4217 }
4218 
4219 static pfil_return_t
4220 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
4221     void *ruleset __unused,  struct inpcb *inp)
4222 {
4223 	int chk;
4224 
4225 	chk = pf_test(PF_OUT, flags, ifp, m, inp);
4226 
4227 	return (pf_check_return(chk, m));
4228 }
4229 #endif
4230 
4231 #ifdef INET6
4232 static pfil_return_t
4233 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
4234     void *ruleset __unused,  struct inpcb *inp)
4235 {
4236 	int chk;
4237 
4238 	/*
4239 	 * In case of loopback traffic IPv6 uses the real interface in
4240 	 * order to support scoped addresses. In order to support stateful
4241 	 * filtering we have change this to lo0 as it is the case in IPv4.
4242 	 */
4243 	CURVNET_SET(ifp->if_vnet);
4244 	chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
4245 	CURVNET_RESTORE();
4246 
4247 	return (pf_check_return(chk, m));
4248 }
4249 
4250 static pfil_return_t
4251 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
4252     void *ruleset __unused,  struct inpcb *inp)
4253 {
4254 	int chk;
4255 
4256 	CURVNET_SET(ifp->if_vnet);
4257 	chk = pf_test6(PF_OUT, flags, ifp, m, inp);
4258 	CURVNET_RESTORE();
4259 
4260 	return (pf_check_return(chk, m));
4261 }
4262 #endif /* INET6 */
4263 
4264 #ifdef INET
4265 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
4266 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
4267 #define	V_pf_ip4_in_hook	VNET(pf_ip4_in_hook)
4268 #define	V_pf_ip4_out_hook	VNET(pf_ip4_out_hook)
4269 #endif
4270 #ifdef INET6
4271 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
4272 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
4273 #define	V_pf_ip6_in_hook	VNET(pf_ip6_in_hook)
4274 #define	V_pf_ip6_out_hook	VNET(pf_ip6_out_hook)
4275 #endif
4276 
4277 static int
4278 hook_pf(void)
4279 {
4280 	struct pfil_hook_args pha;
4281 	struct pfil_link_args pla;
4282 
4283 	if (V_pf_pfil_hooked)
4284 		return (0);
4285 
4286 	pha.pa_version = PFIL_VERSION;
4287 	pha.pa_modname = "pf";
4288 	pha.pa_ruleset = NULL;
4289 
4290 	pla.pa_version = PFIL_VERSION;
4291 
4292 #ifdef INET
4293 	pha.pa_type = PFIL_TYPE_IP4;
4294 	pha.pa_func = pf_check_in;
4295 	pha.pa_flags = PFIL_IN;
4296 	pha.pa_rulname = "default-in";
4297 	V_pf_ip4_in_hook = pfil_add_hook(&pha);
4298 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
4299 	pla.pa_head = V_inet_pfil_head;
4300 	pla.pa_hook = V_pf_ip4_in_hook;
4301 	(void)pfil_link(&pla);
4302 	pha.pa_func = pf_check_out;
4303 	pha.pa_flags = PFIL_OUT;
4304 	pha.pa_rulname = "default-out";
4305 	V_pf_ip4_out_hook = pfil_add_hook(&pha);
4306 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
4307 	pla.pa_head = V_inet_pfil_head;
4308 	pla.pa_hook = V_pf_ip4_out_hook;
4309 	(void)pfil_link(&pla);
4310 #endif
4311 #ifdef INET6
4312 	pha.pa_type = PFIL_TYPE_IP6;
4313 	pha.pa_func = pf_check6_in;
4314 	pha.pa_flags = PFIL_IN;
4315 	pha.pa_rulname = "default-in6";
4316 	V_pf_ip6_in_hook = pfil_add_hook(&pha);
4317 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
4318 	pla.pa_head = V_inet6_pfil_head;
4319 	pla.pa_hook = V_pf_ip6_in_hook;
4320 	(void)pfil_link(&pla);
4321 	pha.pa_func = pf_check6_out;
4322 	pha.pa_rulname = "default-out6";
4323 	pha.pa_flags = PFIL_OUT;
4324 	V_pf_ip6_out_hook = pfil_add_hook(&pha);
4325 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
4326 	pla.pa_head = V_inet6_pfil_head;
4327 	pla.pa_hook = V_pf_ip6_out_hook;
4328 	(void)pfil_link(&pla);
4329 #endif
4330 
4331 	V_pf_pfil_hooked = 1;
4332 	return (0);
4333 }
4334 
4335 static int
4336 dehook_pf(void)
4337 {
4338 
4339 	if (V_pf_pfil_hooked == 0)
4340 		return (0);
4341 
4342 #ifdef INET
4343 	pfil_remove_hook(V_pf_ip4_in_hook);
4344 	pfil_remove_hook(V_pf_ip4_out_hook);
4345 #endif
4346 #ifdef INET6
4347 	pfil_remove_hook(V_pf_ip6_in_hook);
4348 	pfil_remove_hook(V_pf_ip6_out_hook);
4349 #endif
4350 
4351 	V_pf_pfil_hooked = 0;
4352 	return (0);
4353 }
4354 
4355 static void
4356 pf_load_vnet(void)
4357 {
4358 	V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
4359 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
4360 
4361 	pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
4362 	    PF_RULE_TAG_HASH_SIZE_DEFAULT);
4363 #ifdef ALTQ
4364 	pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
4365 	    PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
4366 #endif
4367 
4368 	pfattach_vnet();
4369 	V_pf_vnet_active = 1;
4370 }
4371 
4372 static int
4373 pf_load(void)
4374 {
4375 	int error;
4376 
4377 	rm_init(&pf_rules_lock, "pf rulesets");
4378 	sx_init(&pf_ioctl_lock, "pf ioctl");
4379 	sx_init(&pf_end_lock, "pf end thread");
4380 
4381 	pf_mtag_initialize();
4382 
4383 	pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
4384 	if (pf_dev == NULL)
4385 		return (ENOMEM);
4386 
4387 	pf_end_threads = 0;
4388 	error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
4389 	if (error != 0)
4390 		return (error);
4391 
4392 	pfi_initialize();
4393 
4394 	return (0);
4395 }
4396 
4397 static void
4398 pf_unload_vnet(void)
4399 {
4400 	int error;
4401 
4402 	V_pf_vnet_active = 0;
4403 	V_pf_status.running = 0;
4404 	error = dehook_pf();
4405 	if (error) {
4406 		/*
4407 		 * Should not happen!
4408 		 * XXX Due to error code ESRCH, kldunload will show
4409 		 * a message like 'No such process'.
4410 		 */
4411 		printf("%s : pfil unregisteration fail\n", __FUNCTION__);
4412 		return;
4413 	}
4414 
4415 	PF_RULES_WLOCK();
4416 	shutdown_pf();
4417 	PF_RULES_WUNLOCK();
4418 
4419 	swi_remove(V_pf_swi_cookie);
4420 
4421 	pf_unload_vnet_purge();
4422 
4423 	pf_normalize_cleanup();
4424 	PF_RULES_WLOCK();
4425 	pfi_cleanup_vnet();
4426 	PF_RULES_WUNLOCK();
4427 	pfr_cleanup();
4428 	pf_osfp_flush();
4429 	pf_cleanup();
4430 	if (IS_DEFAULT_VNET(curvnet))
4431 		pf_mtag_cleanup();
4432 
4433 	pf_cleanup_tagset(&V_pf_tags);
4434 #ifdef ALTQ
4435 	pf_cleanup_tagset(&V_pf_qids);
4436 #endif
4437 	uma_zdestroy(V_pf_tag_z);
4438 
4439 	/* Free counters last as we updated them during shutdown. */
4440 	counter_u64_free(V_pf_default_rule.states_cur);
4441 	counter_u64_free(V_pf_default_rule.states_tot);
4442 	counter_u64_free(V_pf_default_rule.src_nodes);
4443 
4444 	for (int i = 0; i < PFRES_MAX; i++)
4445 		counter_u64_free(V_pf_status.counters[i]);
4446 	for (int i = 0; i < LCNT_MAX; i++)
4447 		counter_u64_free(V_pf_status.lcounters[i]);
4448 	for (int i = 0; i < FCNT_MAX; i++)
4449 		counter_u64_free(V_pf_status.fcounters[i]);
4450 	for (int i = 0; i < SCNT_MAX; i++)
4451 		counter_u64_free(V_pf_status.scounters[i]);
4452 }
4453 
4454 static void
4455 pf_unload(void)
4456 {
4457 
4458 	sx_xlock(&pf_end_lock);
4459 	pf_end_threads = 1;
4460 	while (pf_end_threads < 2) {
4461 		wakeup_one(pf_purge_thread);
4462 		sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
4463 	}
4464 	sx_xunlock(&pf_end_lock);
4465 
4466 	if (pf_dev != NULL)
4467 		destroy_dev(pf_dev);
4468 
4469 	pfi_cleanup();
4470 
4471 	rm_destroy(&pf_rules_lock);
4472 	sx_destroy(&pf_ioctl_lock);
4473 	sx_destroy(&pf_end_lock);
4474 }
4475 
4476 static void
4477 vnet_pf_init(void *unused __unused)
4478 {
4479 
4480 	pf_load_vnet();
4481 }
4482 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
4483     vnet_pf_init, NULL);
4484 
4485 static void
4486 vnet_pf_uninit(const void *unused __unused)
4487 {
4488 
4489 	pf_unload_vnet();
4490 }
4491 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
4492 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
4493     vnet_pf_uninit, NULL);
4494 
4495 static int
4496 pf_modevent(module_t mod, int type, void *data)
4497 {
4498 	int error = 0;
4499 
4500 	switch(type) {
4501 	case MOD_LOAD:
4502 		error = pf_load();
4503 		break;
4504 	case MOD_UNLOAD:
4505 		/* Handled in SYSUNINIT(pf_unload) to ensure it's done after
4506 		 * the vnet_pf_uninit()s */
4507 		break;
4508 	default:
4509 		error = EINVAL;
4510 		break;
4511 	}
4512 
4513 	return (error);
4514 }
4515 
4516 static moduledata_t pf_mod = {
4517 	"pf",
4518 	pf_modevent,
4519 	0
4520 };
4521 
4522 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
4523 MODULE_VERSION(pf, PF_MODVER);
4524