xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision 500f4659d7c8947082dba040a1d58e7d228f8d44)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2001 Daniel Hartmeier
5  * Copyright (c) 2002,2003 Henning Brauer
6  * Copyright (c) 2012 Gleb Smirnoff <glebius@FreeBSD.org>
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  *
13  *    - Redistributions of source code must retain the above copyright
14  *      notice, this list of conditions and the following disclaimer.
15  *    - Redistributions in binary form must reproduce the above
16  *      copyright notice, this list of conditions and the following
17  *      disclaimer in the documentation and/or other materials provided
18  *      with the distribution.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
26  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
28  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
30  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
31  * POSSIBILITY OF SUCH DAMAGE.
32  *
33  * Effort sponsored in part by the Defense Advanced Research Projects
34  * Agency (DARPA) and Air Force Research Laboratory, Air Force
35  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
36  *
37  *	$OpenBSD: pf_ioctl.c,v 1.213 2009/02/15 21:46:12 mbalmer Exp $
38  */
39 
40 #include <sys/cdefs.h>
41 __FBSDID("$FreeBSD$");
42 
43 #include "opt_inet.h"
44 #include "opt_inet6.h"
45 #include "opt_bpf.h"
46 #include "opt_pf.h"
47 
48 #include <sys/param.h>
49 #include <sys/_bitset.h>
50 #include <sys/bitset.h>
51 #include <sys/bus.h>
52 #include <sys/conf.h>
53 #include <sys/endian.h>
54 #include <sys/fcntl.h>
55 #include <sys/filio.h>
56 #include <sys/hash.h>
57 #include <sys/interrupt.h>
58 #include <sys/jail.h>
59 #include <sys/kernel.h>
60 #include <sys/kthread.h>
61 #include <sys/lock.h>
62 #include <sys/mbuf.h>
63 #include <sys/module.h>
64 #include <sys/proc.h>
65 #include <sys/smp.h>
66 #include <sys/socket.h>
67 #include <sys/sysctl.h>
68 #include <sys/md5.h>
69 #include <sys/ucred.h>
70 
71 #include <net/if.h>
72 #include <net/if_var.h>
73 #include <net/vnet.h>
74 #include <net/route.h>
75 #include <net/pfil.h>
76 #include <net/pfvar.h>
77 #include <net/if_pfsync.h>
78 #include <net/if_pflog.h>
79 
80 #include <netinet/in.h>
81 #include <netinet/ip.h>
82 #include <netinet/ip_var.h>
83 #include <netinet6/ip6_var.h>
84 #include <netinet/ip_icmp.h>
85 
86 #ifdef INET6
87 #include <netinet/ip6.h>
88 #endif /* INET6 */
89 
90 #ifdef ALTQ
91 #include <net/altq/altq.h>
92 #endif
93 
94 static struct pf_kpool	*pf_get_kpool(char *, u_int32_t, u_int8_t, u_int32_t,
95 			    u_int8_t, u_int8_t, u_int8_t);
96 
97 static void		 pf_mv_kpool(struct pf_kpalist *, struct pf_kpalist *);
98 static void		 pf_empty_kpool(struct pf_kpalist *);
99 static int		 pfioctl(struct cdev *, u_long, caddr_t, int,
100 			    struct thread *);
101 #ifdef ALTQ
102 static int		 pf_begin_altq(u_int32_t *);
103 static int		 pf_rollback_altq(u_int32_t);
104 static int		 pf_commit_altq(u_int32_t);
105 static int		 pf_enable_altq(struct pf_altq *);
106 static int		 pf_disable_altq(struct pf_altq *);
107 static u_int32_t	 pf_qname2qid(char *);
108 static void		 pf_qid_unref(u_int32_t);
109 #endif /* ALTQ */
110 static int		 pf_begin_rules(u_int32_t *, int, const char *);
111 static int		 pf_rollback_rules(u_int32_t, int, char *);
112 static int		 pf_setup_pfsync_matching(struct pf_kruleset *);
113 static void		 pf_hash_rule(MD5_CTX *, struct pf_krule *);
114 static void		 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
115 static int		 pf_commit_rules(u_int32_t, int, char *);
116 static int		 pf_addr_setup(struct pf_kruleset *,
117 			    struct pf_addr_wrap *, sa_family_t);
118 static void		 pf_addr_copyout(struct pf_addr_wrap *);
119 static void		 pf_src_node_copy(const struct pf_ksrc_node *,
120 			    struct pf_src_node *);
121 #ifdef ALTQ
122 static int		 pf_export_kaltq(struct pf_altq *,
123 			    struct pfioc_altq_v1 *, size_t);
124 static int		 pf_import_kaltq(struct pfioc_altq_v1 *,
125 			    struct pf_altq *, size_t);
126 #endif /* ALTQ */
127 
128 VNET_DEFINE(struct pf_krule,	pf_default_rule);
129 
130 #ifdef ALTQ
131 VNET_DEFINE_STATIC(int,		pf_altq_running);
132 #define	V_pf_altq_running	VNET(pf_altq_running)
133 #endif
134 
135 #define	TAGID_MAX	 50000
136 struct pf_tagname {
137 	TAILQ_ENTRY(pf_tagname)	namehash_entries;
138 	TAILQ_ENTRY(pf_tagname)	taghash_entries;
139 	char			name[PF_TAG_NAME_SIZE];
140 	uint16_t		tag;
141 	int			ref;
142 };
143 
144 struct pf_tagset {
145 	TAILQ_HEAD(, pf_tagname)	*namehash;
146 	TAILQ_HEAD(, pf_tagname)	*taghash;
147 	unsigned int			 mask;
148 	uint32_t			 seed;
149 	BITSET_DEFINE(, TAGID_MAX)	 avail;
150 };
151 
152 VNET_DEFINE(struct pf_tagset, pf_tags);
153 #define	V_pf_tags	VNET(pf_tags)
154 static unsigned int	pf_rule_tag_hashsize;
155 #define	PF_RULE_TAG_HASH_SIZE_DEFAULT	128
156 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN,
157     &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT,
158     "Size of pf(4) rule tag hashtable");
159 
160 #ifdef ALTQ
161 VNET_DEFINE(struct pf_tagset, pf_qids);
162 #define	V_pf_qids	VNET(pf_qids)
163 static unsigned int	pf_queue_tag_hashsize;
164 #define	PF_QUEUE_TAG_HASH_SIZE_DEFAULT	128
165 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN,
166     &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT,
167     "Size of pf(4) queue tag hashtable");
168 #endif
169 VNET_DEFINE(uma_zone_t,	 pf_tag_z);
170 #define	V_pf_tag_z		 VNET(pf_tag_z)
171 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
172 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
173 
174 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
175 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
176 #endif
177 
178 static void		 pf_init_tagset(struct pf_tagset *, unsigned int *,
179 			    unsigned int);
180 static void		 pf_cleanup_tagset(struct pf_tagset *);
181 static uint16_t		 tagname2hashindex(const struct pf_tagset *, const char *);
182 static uint16_t		 tag2hashindex(const struct pf_tagset *, uint16_t);
183 static u_int16_t	 tagname2tag(struct pf_tagset *, char *);
184 static u_int16_t	 pf_tagname2tag(char *);
185 static void		 tag_unref(struct pf_tagset *, u_int16_t);
186 
187 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
188 
189 struct cdev *pf_dev;
190 
191 /*
192  * XXX - These are new and need to be checked when moveing to a new version
193  */
194 static void		 pf_clear_states(void);
195 static int		 pf_clear_tables(void);
196 static void		 pf_clear_srcnodes(struct pf_ksrc_node *);
197 static void		 pf_kill_srcnodes(struct pfioc_src_node_kill *);
198 static void		 pf_tbladdr_copyout(struct pf_addr_wrap *);
199 
200 /*
201  * Wrapper functions for pfil(9) hooks
202  */
203 #ifdef INET
204 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp,
205     int flags, void *ruleset __unused, struct inpcb *inp);
206 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp,
207     int flags, void *ruleset __unused, struct inpcb *inp);
208 #endif
209 #ifdef INET6
210 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp,
211     int flags, void *ruleset __unused, struct inpcb *inp);
212 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp,
213     int flags, void *ruleset __unused, struct inpcb *inp);
214 #endif
215 
216 static void		hook_pf(void);
217 static void		dehook_pf(void);
218 static int		shutdown_pf(void);
219 static int		pf_load(void);
220 static void		pf_unload(void);
221 
222 static struct cdevsw pf_cdevsw = {
223 	.d_ioctl =	pfioctl,
224 	.d_name =	PF_NAME,
225 	.d_version =	D_VERSION,
226 };
227 
228 volatile VNET_DEFINE_STATIC(int, pf_pfil_hooked);
229 #define V_pf_pfil_hooked	VNET(pf_pfil_hooked)
230 
231 /*
232  * We need a flag that is neither hooked nor running to know when
233  * the VNET is "valid".  We primarily need this to control (global)
234  * external event, e.g., eventhandlers.
235  */
236 VNET_DEFINE(int, pf_vnet_active);
237 #define V_pf_vnet_active	VNET(pf_vnet_active)
238 
239 int pf_end_threads;
240 struct proc *pf_purge_proc;
241 
242 struct rmlock			pf_rules_lock;
243 struct sx			pf_ioctl_lock;
244 struct sx			pf_end_lock;
245 
246 /* pfsync */
247 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
248 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
249 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
250 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
251 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
252 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
253 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
254 
255 /* pflog */
256 pflog_packet_t			*pflog_packet_ptr = NULL;
257 
258 extern u_long	pf_ioctl_maxcount;
259 
260 static void
261 pfattach_vnet(void)
262 {
263 	u_int32_t *my_timeout = V_pf_default_rule.timeout;
264 
265 	pf_initialize();
266 	pfr_initialize();
267 	pfi_initialize_vnet();
268 	pf_normalize_init();
269 
270 	V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
271 	V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
272 
273 	RB_INIT(&V_pf_anchors);
274 	pf_init_kruleset(&pf_main_ruleset);
275 
276 	/* default rule should never be garbage collected */
277 	V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
278 #ifdef PF_DEFAULT_TO_DROP
279 	V_pf_default_rule.action = PF_DROP;
280 #else
281 	V_pf_default_rule.action = PF_PASS;
282 #endif
283 	V_pf_default_rule.nr = -1;
284 	V_pf_default_rule.rtableid = -1;
285 
286 	V_pf_default_rule.evaluations = counter_u64_alloc(M_WAITOK);
287 	for (int i = 0; i < 2; i++) {
288 		V_pf_default_rule.packets[i] = counter_u64_alloc(M_WAITOK);
289 		V_pf_default_rule.bytes[i] = counter_u64_alloc(M_WAITOK);
290 	}
291 	V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
292 	V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
293 	V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
294 
295 	/* initialize default timeouts */
296 	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
297 	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
298 	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
299 	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
300 	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
301 	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
302 	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
303 	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
304 	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
305 	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
306 	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
307 	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
308 	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
309 	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
310 	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
311 	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
312 	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
313 	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
314 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
315 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
316 
317 	bzero(&V_pf_status, sizeof(V_pf_status));
318 	V_pf_status.debug = PF_DEBUG_URGENT;
319 
320 	V_pf_pfil_hooked = 0;
321 
322 	/* XXX do our best to avoid a conflict */
323 	V_pf_status.hostid = arc4random();
324 
325 	for (int i = 0; i < PFRES_MAX; i++)
326 		V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
327 	for (int i = 0; i < LCNT_MAX; i++)
328 		V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
329 	for (int i = 0; i < FCNT_MAX; i++)
330 		V_pf_status.fcounters[i] = counter_u64_alloc(M_WAITOK);
331 	for (int i = 0; i < SCNT_MAX; i++)
332 		V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
333 
334 	if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET,
335 	    INTR_MPSAFE, &V_pf_swi_cookie) != 0)
336 		/* XXXGL: leaked all above. */
337 		return;
338 }
339 
340 static struct pf_kpool *
341 pf_get_kpool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
342     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
343     u_int8_t check_ticket)
344 {
345 	struct pf_kruleset	*ruleset;
346 	struct pf_krule		*rule;
347 	int			 rs_num;
348 
349 	ruleset = pf_find_kruleset(anchor);
350 	if (ruleset == NULL)
351 		return (NULL);
352 	rs_num = pf_get_ruleset_number(rule_action);
353 	if (rs_num >= PF_RULESET_MAX)
354 		return (NULL);
355 	if (active) {
356 		if (check_ticket && ticket !=
357 		    ruleset->rules[rs_num].active.ticket)
358 			return (NULL);
359 		if (r_last)
360 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
361 			    pf_krulequeue);
362 		else
363 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
364 	} else {
365 		if (check_ticket && ticket !=
366 		    ruleset->rules[rs_num].inactive.ticket)
367 			return (NULL);
368 		if (r_last)
369 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
370 			    pf_krulequeue);
371 		else
372 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
373 	}
374 	if (!r_last) {
375 		while ((rule != NULL) && (rule->nr != rule_number))
376 			rule = TAILQ_NEXT(rule, entries);
377 	}
378 	if (rule == NULL)
379 		return (NULL);
380 
381 	return (&rule->rpool);
382 }
383 
384 static void
385 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb)
386 {
387 	struct pf_kpooladdr	*mv_pool_pa;
388 
389 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
390 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
391 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
392 	}
393 }
394 
395 static void
396 pf_empty_kpool(struct pf_kpalist *poola)
397 {
398 	struct pf_kpooladdr *pa;
399 
400 	while ((pa = TAILQ_FIRST(poola)) != NULL) {
401 		switch (pa->addr.type) {
402 		case PF_ADDR_DYNIFTL:
403 			pfi_dynaddr_remove(pa->addr.p.dyn);
404 			break;
405 		case PF_ADDR_TABLE:
406 			/* XXX: this could be unfinished pooladdr on pabuf */
407 			if (pa->addr.p.tbl != NULL)
408 				pfr_detach_table(pa->addr.p.tbl);
409 			break;
410 		}
411 		if (pa->kif)
412 			pfi_kkif_unref(pa->kif);
413 		TAILQ_REMOVE(poola, pa, entries);
414 		free(pa, M_PFRULE);
415 	}
416 }
417 
418 static void
419 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
420 {
421 
422 	PF_RULES_WASSERT();
423 
424 	TAILQ_REMOVE(rulequeue, rule, entries);
425 
426 	PF_UNLNKDRULES_LOCK();
427 	rule->rule_flag |= PFRULE_REFS;
428 	TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
429 	PF_UNLNKDRULES_UNLOCK();
430 }
431 
432 void
433 pf_free_rule(struct pf_krule *rule)
434 {
435 
436 	PF_RULES_WASSERT();
437 
438 	if (rule->tag)
439 		tag_unref(&V_pf_tags, rule->tag);
440 	if (rule->match_tag)
441 		tag_unref(&V_pf_tags, rule->match_tag);
442 #ifdef ALTQ
443 	if (rule->pqid != rule->qid)
444 		pf_qid_unref(rule->pqid);
445 	pf_qid_unref(rule->qid);
446 #endif
447 	switch (rule->src.addr.type) {
448 	case PF_ADDR_DYNIFTL:
449 		pfi_dynaddr_remove(rule->src.addr.p.dyn);
450 		break;
451 	case PF_ADDR_TABLE:
452 		pfr_detach_table(rule->src.addr.p.tbl);
453 		break;
454 	}
455 	switch (rule->dst.addr.type) {
456 	case PF_ADDR_DYNIFTL:
457 		pfi_dynaddr_remove(rule->dst.addr.p.dyn);
458 		break;
459 	case PF_ADDR_TABLE:
460 		pfr_detach_table(rule->dst.addr.p.tbl);
461 		break;
462 	}
463 	if (rule->overload_tbl)
464 		pfr_detach_table(rule->overload_tbl);
465 	if (rule->kif)
466 		pfi_kkif_unref(rule->kif);
467 	pf_kanchor_remove(rule);
468 	pf_empty_kpool(&rule->rpool.list);
469 
470 	pf_krule_free(rule);
471 }
472 
473 static void
474 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
475     unsigned int default_size)
476 {
477 	unsigned int i;
478 	unsigned int hashsize;
479 
480 	if (*tunable_size == 0 || !powerof2(*tunable_size))
481 		*tunable_size = default_size;
482 
483 	hashsize = *tunable_size;
484 	ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
485 	    M_WAITOK);
486 	ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
487 	    M_WAITOK);
488 	ts->mask = hashsize - 1;
489 	ts->seed = arc4random();
490 	for (i = 0; i < hashsize; i++) {
491 		TAILQ_INIT(&ts->namehash[i]);
492 		TAILQ_INIT(&ts->taghash[i]);
493 	}
494 	BIT_FILL(TAGID_MAX, &ts->avail);
495 }
496 
497 static void
498 pf_cleanup_tagset(struct pf_tagset *ts)
499 {
500 	unsigned int i;
501 	unsigned int hashsize;
502 	struct pf_tagname *t, *tmp;
503 
504 	/*
505 	 * Only need to clean up one of the hashes as each tag is hashed
506 	 * into each table.
507 	 */
508 	hashsize = ts->mask + 1;
509 	for (i = 0; i < hashsize; i++)
510 		TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
511 			uma_zfree(V_pf_tag_z, t);
512 
513 	free(ts->namehash, M_PFHASH);
514 	free(ts->taghash, M_PFHASH);
515 }
516 
517 static uint16_t
518 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
519 {
520 	size_t len;
521 
522 	len = strnlen(tagname, PF_TAG_NAME_SIZE - 1);
523 	return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask);
524 }
525 
526 static uint16_t
527 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
528 {
529 
530 	return (tag & ts->mask);
531 }
532 
533 static u_int16_t
534 tagname2tag(struct pf_tagset *ts, char *tagname)
535 {
536 	struct pf_tagname	*tag;
537 	u_int32_t		 index;
538 	u_int16_t		 new_tagid;
539 
540 	PF_RULES_WASSERT();
541 
542 	index = tagname2hashindex(ts, tagname);
543 	TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
544 		if (strcmp(tagname, tag->name) == 0) {
545 			tag->ref++;
546 			return (tag->tag);
547 		}
548 
549 	/*
550 	 * new entry
551 	 *
552 	 * to avoid fragmentation, we do a linear search from the beginning
553 	 * and take the first free slot we find.
554 	 */
555 	new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
556 	/*
557 	 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
558 	 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
559 	 * set.  It may also return a bit number greater than TAGID_MAX due
560 	 * to rounding of the number of bits in the vector up to a multiple
561 	 * of the vector word size at declaration/allocation time.
562 	 */
563 	if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
564 		return (0);
565 
566 	/* Mark the tag as in use.  Bits are 0-based for BIT_CLR() */
567 	BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
568 
569 	/* allocate and fill new struct pf_tagname */
570 	tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
571 	if (tag == NULL)
572 		return (0);
573 	strlcpy(tag->name, tagname, sizeof(tag->name));
574 	tag->tag = new_tagid;
575 	tag->ref = 1;
576 
577 	/* Insert into namehash */
578 	TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
579 
580 	/* Insert into taghash */
581 	index = tag2hashindex(ts, new_tagid);
582 	TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
583 
584 	return (tag->tag);
585 }
586 
587 static void
588 tag_unref(struct pf_tagset *ts, u_int16_t tag)
589 {
590 	struct pf_tagname	*t;
591 	uint16_t		 index;
592 
593 	PF_RULES_WASSERT();
594 
595 	index = tag2hashindex(ts, tag);
596 	TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
597 		if (tag == t->tag) {
598 			if (--t->ref == 0) {
599 				TAILQ_REMOVE(&ts->taghash[index], t,
600 				    taghash_entries);
601 				index = tagname2hashindex(ts, t->name);
602 				TAILQ_REMOVE(&ts->namehash[index], t,
603 				    namehash_entries);
604 				/* Bits are 0-based for BIT_SET() */
605 				BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
606 				uma_zfree(V_pf_tag_z, t);
607 			}
608 			break;
609 		}
610 }
611 
612 static u_int16_t
613 pf_tagname2tag(char *tagname)
614 {
615 	return (tagname2tag(&V_pf_tags, tagname));
616 }
617 
618 #ifdef ALTQ
619 static u_int32_t
620 pf_qname2qid(char *qname)
621 {
622 	return ((u_int32_t)tagname2tag(&V_pf_qids, qname));
623 }
624 
625 static void
626 pf_qid_unref(u_int32_t qid)
627 {
628 	tag_unref(&V_pf_qids, (u_int16_t)qid);
629 }
630 
631 static int
632 pf_begin_altq(u_int32_t *ticket)
633 {
634 	struct pf_altq	*altq, *tmp;
635 	int		 error = 0;
636 
637 	PF_RULES_WASSERT();
638 
639 	/* Purge the old altq lists */
640 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
641 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
642 			/* detach and destroy the discipline */
643 			error = altq_remove(altq);
644 		}
645 		free(altq, M_PFALTQ);
646 	}
647 	TAILQ_INIT(V_pf_altq_ifs_inactive);
648 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
649 		pf_qid_unref(altq->qid);
650 		free(altq, M_PFALTQ);
651 	}
652 	TAILQ_INIT(V_pf_altqs_inactive);
653 	if (error)
654 		return (error);
655 	*ticket = ++V_ticket_altqs_inactive;
656 	V_altqs_inactive_open = 1;
657 	return (0);
658 }
659 
660 static int
661 pf_rollback_altq(u_int32_t ticket)
662 {
663 	struct pf_altq	*altq, *tmp;
664 	int		 error = 0;
665 
666 	PF_RULES_WASSERT();
667 
668 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
669 		return (0);
670 	/* Purge the old altq lists */
671 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
672 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
673 			/* detach and destroy the discipline */
674 			error = altq_remove(altq);
675 		}
676 		free(altq, M_PFALTQ);
677 	}
678 	TAILQ_INIT(V_pf_altq_ifs_inactive);
679 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
680 		pf_qid_unref(altq->qid);
681 		free(altq, M_PFALTQ);
682 	}
683 	TAILQ_INIT(V_pf_altqs_inactive);
684 	V_altqs_inactive_open = 0;
685 	return (error);
686 }
687 
688 static int
689 pf_commit_altq(u_int32_t ticket)
690 {
691 	struct pf_altqqueue	*old_altqs, *old_altq_ifs;
692 	struct pf_altq		*altq, *tmp;
693 	int			 err, error = 0;
694 
695 	PF_RULES_WASSERT();
696 
697 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
698 		return (EBUSY);
699 
700 	/* swap altqs, keep the old. */
701 	old_altqs = V_pf_altqs_active;
702 	old_altq_ifs = V_pf_altq_ifs_active;
703 	V_pf_altqs_active = V_pf_altqs_inactive;
704 	V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
705 	V_pf_altqs_inactive = old_altqs;
706 	V_pf_altq_ifs_inactive = old_altq_ifs;
707 	V_ticket_altqs_active = V_ticket_altqs_inactive;
708 
709 	/* Attach new disciplines */
710 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
711 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
712 			/* attach the discipline */
713 			error = altq_pfattach(altq);
714 			if (error == 0 && V_pf_altq_running)
715 				error = pf_enable_altq(altq);
716 			if (error != 0)
717 				return (error);
718 		}
719 	}
720 
721 	/* Purge the old altq lists */
722 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
723 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
724 			/* detach and destroy the discipline */
725 			if (V_pf_altq_running)
726 				error = pf_disable_altq(altq);
727 			err = altq_pfdetach(altq);
728 			if (err != 0 && error == 0)
729 				error = err;
730 			err = altq_remove(altq);
731 			if (err != 0 && error == 0)
732 				error = err;
733 		}
734 		free(altq, M_PFALTQ);
735 	}
736 	TAILQ_INIT(V_pf_altq_ifs_inactive);
737 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
738 		pf_qid_unref(altq->qid);
739 		free(altq, M_PFALTQ);
740 	}
741 	TAILQ_INIT(V_pf_altqs_inactive);
742 
743 	V_altqs_inactive_open = 0;
744 	return (error);
745 }
746 
747 static int
748 pf_enable_altq(struct pf_altq *altq)
749 {
750 	struct ifnet		*ifp;
751 	struct tb_profile	 tb;
752 	int			 error = 0;
753 
754 	if ((ifp = ifunit(altq->ifname)) == NULL)
755 		return (EINVAL);
756 
757 	if (ifp->if_snd.altq_type != ALTQT_NONE)
758 		error = altq_enable(&ifp->if_snd);
759 
760 	/* set tokenbucket regulator */
761 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
762 		tb.rate = altq->ifbandwidth;
763 		tb.depth = altq->tbrsize;
764 		error = tbr_set(&ifp->if_snd, &tb);
765 	}
766 
767 	return (error);
768 }
769 
770 static int
771 pf_disable_altq(struct pf_altq *altq)
772 {
773 	struct ifnet		*ifp;
774 	struct tb_profile	 tb;
775 	int			 error;
776 
777 	if ((ifp = ifunit(altq->ifname)) == NULL)
778 		return (EINVAL);
779 
780 	/*
781 	 * when the discipline is no longer referenced, it was overridden
782 	 * by a new one.  if so, just return.
783 	 */
784 	if (altq->altq_disc != ifp->if_snd.altq_disc)
785 		return (0);
786 
787 	error = altq_disable(&ifp->if_snd);
788 
789 	if (error == 0) {
790 		/* clear tokenbucket regulator */
791 		tb.rate = 0;
792 		error = tbr_set(&ifp->if_snd, &tb);
793 	}
794 
795 	return (error);
796 }
797 
798 static int
799 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
800     struct pf_altq *altq)
801 {
802 	struct ifnet	*ifp1;
803 	int		 error = 0;
804 
805 	/* Deactivate the interface in question */
806 	altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
807 	if ((ifp1 = ifunit(altq->ifname)) == NULL ||
808 	    (remove && ifp1 == ifp)) {
809 		altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
810 	} else {
811 		error = altq_add(ifp1, altq);
812 
813 		if (ticket != V_ticket_altqs_inactive)
814 			error = EBUSY;
815 
816 		if (error)
817 			free(altq, M_PFALTQ);
818 	}
819 
820 	return (error);
821 }
822 
823 void
824 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
825 {
826 	struct pf_altq	*a1, *a2, *a3;
827 	u_int32_t	 ticket;
828 	int		 error = 0;
829 
830 	/*
831 	 * No need to re-evaluate the configuration for events on interfaces
832 	 * that do not support ALTQ, as it's not possible for such
833 	 * interfaces to be part of the configuration.
834 	 */
835 	if (!ALTQ_IS_READY(&ifp->if_snd))
836 		return;
837 
838 	/* Interrupt userland queue modifications */
839 	if (V_altqs_inactive_open)
840 		pf_rollback_altq(V_ticket_altqs_inactive);
841 
842 	/* Start new altq ruleset */
843 	if (pf_begin_altq(&ticket))
844 		return;
845 
846 	/* Copy the current active set */
847 	TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
848 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
849 		if (a2 == NULL) {
850 			error = ENOMEM;
851 			break;
852 		}
853 		bcopy(a1, a2, sizeof(struct pf_altq));
854 
855 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
856 		if (error)
857 			break;
858 
859 		TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
860 	}
861 	if (error)
862 		goto out;
863 	TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
864 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
865 		if (a2 == NULL) {
866 			error = ENOMEM;
867 			break;
868 		}
869 		bcopy(a1, a2, sizeof(struct pf_altq));
870 
871 		if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
872 			error = EBUSY;
873 			free(a2, M_PFALTQ);
874 			break;
875 		}
876 		a2->altq_disc = NULL;
877 		TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
878 			if (strncmp(a3->ifname, a2->ifname,
879 				IFNAMSIZ) == 0) {
880 				a2->altq_disc = a3->altq_disc;
881 				break;
882 			}
883 		}
884 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
885 		if (error)
886 			break;
887 
888 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
889 	}
890 
891 out:
892 	if (error != 0)
893 		pf_rollback_altq(ticket);
894 	else
895 		pf_commit_altq(ticket);
896 }
897 #endif /* ALTQ */
898 
899 static int
900 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
901 {
902 	struct pf_kruleset	*rs;
903 	struct pf_krule		*rule;
904 
905 	PF_RULES_WASSERT();
906 
907 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
908 		return (EINVAL);
909 	rs = pf_find_or_create_kruleset(anchor);
910 	if (rs == NULL)
911 		return (EINVAL);
912 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
913 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
914 		rs->rules[rs_num].inactive.rcount--;
915 	}
916 	*ticket = ++rs->rules[rs_num].inactive.ticket;
917 	rs->rules[rs_num].inactive.open = 1;
918 	return (0);
919 }
920 
921 static int
922 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
923 {
924 	struct pf_kruleset	*rs;
925 	struct pf_krule		*rule;
926 
927 	PF_RULES_WASSERT();
928 
929 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
930 		return (EINVAL);
931 	rs = pf_find_kruleset(anchor);
932 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
933 	    rs->rules[rs_num].inactive.ticket != ticket)
934 		return (0);
935 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
936 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
937 		rs->rules[rs_num].inactive.rcount--;
938 	}
939 	rs->rules[rs_num].inactive.open = 0;
940 	return (0);
941 }
942 
943 #define PF_MD5_UPD(st, elm)						\
944 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
945 
946 #define PF_MD5_UPD_STR(st, elm)						\
947 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
948 
949 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
950 		(stor) = htonl((st)->elm);				\
951 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
952 } while (0)
953 
954 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
955 		(stor) = htons((st)->elm);				\
956 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
957 } while (0)
958 
959 static void
960 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
961 {
962 	PF_MD5_UPD(pfr, addr.type);
963 	switch (pfr->addr.type) {
964 		case PF_ADDR_DYNIFTL:
965 			PF_MD5_UPD(pfr, addr.v.ifname);
966 			PF_MD5_UPD(pfr, addr.iflags);
967 			break;
968 		case PF_ADDR_TABLE:
969 			PF_MD5_UPD(pfr, addr.v.tblname);
970 			break;
971 		case PF_ADDR_ADDRMASK:
972 			/* XXX ignore af? */
973 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
974 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
975 			break;
976 	}
977 
978 	PF_MD5_UPD(pfr, port[0]);
979 	PF_MD5_UPD(pfr, port[1]);
980 	PF_MD5_UPD(pfr, neg);
981 	PF_MD5_UPD(pfr, port_op);
982 }
983 
984 static void
985 pf_hash_rule(MD5_CTX *ctx, struct pf_krule *rule)
986 {
987 	u_int16_t x;
988 	u_int32_t y;
989 
990 	pf_hash_rule_addr(ctx, &rule->src);
991 	pf_hash_rule_addr(ctx, &rule->dst);
992 	PF_MD5_UPD_STR(rule, label);
993 	PF_MD5_UPD_STR(rule, ifname);
994 	PF_MD5_UPD_STR(rule, match_tagname);
995 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
996 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
997 	PF_MD5_UPD_HTONL(rule, prob, y);
998 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
999 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1000 	PF_MD5_UPD(rule, uid.op);
1001 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1002 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1003 	PF_MD5_UPD(rule, gid.op);
1004 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
1005 	PF_MD5_UPD(rule, action);
1006 	PF_MD5_UPD(rule, direction);
1007 	PF_MD5_UPD(rule, af);
1008 	PF_MD5_UPD(rule, quick);
1009 	PF_MD5_UPD(rule, ifnot);
1010 	PF_MD5_UPD(rule, match_tag_not);
1011 	PF_MD5_UPD(rule, natpass);
1012 	PF_MD5_UPD(rule, keep_state);
1013 	PF_MD5_UPD(rule, proto);
1014 	PF_MD5_UPD(rule, type);
1015 	PF_MD5_UPD(rule, code);
1016 	PF_MD5_UPD(rule, flags);
1017 	PF_MD5_UPD(rule, flagset);
1018 	PF_MD5_UPD(rule, allow_opts);
1019 	PF_MD5_UPD(rule, rt);
1020 	PF_MD5_UPD(rule, tos);
1021 }
1022 
1023 static int
1024 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1025 {
1026 	struct pf_kruleset	*rs;
1027 	struct pf_krule		*rule, **old_array;
1028 	struct pf_krulequeue	*old_rules;
1029 	int			 error;
1030 	u_int32_t		 old_rcount;
1031 
1032 	PF_RULES_WASSERT();
1033 
1034 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1035 		return (EINVAL);
1036 	rs = pf_find_kruleset(anchor);
1037 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1038 	    ticket != rs->rules[rs_num].inactive.ticket)
1039 		return (EBUSY);
1040 
1041 	/* Calculate checksum for the main ruleset */
1042 	if (rs == &pf_main_ruleset) {
1043 		error = pf_setup_pfsync_matching(rs);
1044 		if (error != 0)
1045 			return (error);
1046 	}
1047 
1048 	/* Swap rules, keep the old. */
1049 	old_rules = rs->rules[rs_num].active.ptr;
1050 	old_rcount = rs->rules[rs_num].active.rcount;
1051 	old_array = rs->rules[rs_num].active.ptr_array;
1052 
1053 	rs->rules[rs_num].active.ptr =
1054 	    rs->rules[rs_num].inactive.ptr;
1055 	rs->rules[rs_num].active.ptr_array =
1056 	    rs->rules[rs_num].inactive.ptr_array;
1057 	rs->rules[rs_num].active.rcount =
1058 	    rs->rules[rs_num].inactive.rcount;
1059 	rs->rules[rs_num].inactive.ptr = old_rules;
1060 	rs->rules[rs_num].inactive.ptr_array = old_array;
1061 	rs->rules[rs_num].inactive.rcount = old_rcount;
1062 
1063 	rs->rules[rs_num].active.ticket =
1064 	    rs->rules[rs_num].inactive.ticket;
1065 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1066 
1067 	/* Purge the old rule list. */
1068 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1069 		pf_unlink_rule(old_rules, rule);
1070 	if (rs->rules[rs_num].inactive.ptr_array)
1071 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1072 	rs->rules[rs_num].inactive.ptr_array = NULL;
1073 	rs->rules[rs_num].inactive.rcount = 0;
1074 	rs->rules[rs_num].inactive.open = 0;
1075 	pf_remove_if_empty_kruleset(rs);
1076 
1077 	return (0);
1078 }
1079 
1080 static int
1081 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1082 {
1083 	MD5_CTX			 ctx;
1084 	struct pf_krule		*rule;
1085 	int			 rs_cnt;
1086 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
1087 
1088 	MD5Init(&ctx);
1089 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1090 		/* XXX PF_RULESET_SCRUB as well? */
1091 		if (rs_cnt == PF_RULESET_SCRUB)
1092 			continue;
1093 
1094 		if (rs->rules[rs_cnt].inactive.ptr_array)
1095 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1096 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
1097 
1098 		if (rs->rules[rs_cnt].inactive.rcount) {
1099 			rs->rules[rs_cnt].inactive.ptr_array =
1100 			    malloc(sizeof(caddr_t) *
1101 			    rs->rules[rs_cnt].inactive.rcount,
1102 			    M_TEMP, M_NOWAIT);
1103 
1104 			if (!rs->rules[rs_cnt].inactive.ptr_array)
1105 				return (ENOMEM);
1106 		}
1107 
1108 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1109 		    entries) {
1110 			pf_hash_rule(&ctx, rule);
1111 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1112 		}
1113 	}
1114 
1115 	MD5Final(digest, &ctx);
1116 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1117 	return (0);
1118 }
1119 
1120 static int
1121 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1122     sa_family_t af)
1123 {
1124 	int error = 0;
1125 
1126 	switch (addr->type) {
1127 	case PF_ADDR_TABLE:
1128 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1129 		if (addr->p.tbl == NULL)
1130 			error = ENOMEM;
1131 		break;
1132 	case PF_ADDR_DYNIFTL:
1133 		error = pfi_dynaddr_setup(addr, af);
1134 		break;
1135 	}
1136 
1137 	return (error);
1138 }
1139 
1140 static void
1141 pf_addr_copyout(struct pf_addr_wrap *addr)
1142 {
1143 
1144 	switch (addr->type) {
1145 	case PF_ADDR_DYNIFTL:
1146 		pfi_dynaddr_copyout(addr);
1147 		break;
1148 	case PF_ADDR_TABLE:
1149 		pf_tbladdr_copyout(addr);
1150 		break;
1151 	}
1152 }
1153 
1154 static void
1155 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
1156 {
1157 	int	secs = time_uptime, diff;
1158 
1159 	bzero(out, sizeof(struct pf_src_node));
1160 
1161 	bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
1162 	bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
1163 
1164 	if (in->rule.ptr != NULL)
1165 		out->rule.nr = in->rule.ptr->nr;
1166 
1167 	for (int i = 0; i < 2; i++) {
1168 		out->bytes[i] = counter_u64_fetch(in->bytes[i]);
1169 		out->packets[i] = counter_u64_fetch(in->packets[i]);
1170 	}
1171 
1172 	out->states = in->states;
1173 	out->conn = in->conn;
1174 	out->af = in->af;
1175 	out->ruletype = in->ruletype;
1176 
1177 	out->creation = secs - in->creation;
1178 	if (out->expire > secs)
1179 		out->expire -= secs;
1180 	else
1181 		out->expire = 0;
1182 
1183 	/* Adjust the connection rate estimate. */
1184 	diff = secs - in->conn_rate.last;
1185 	if (diff >= in->conn_rate.seconds)
1186 		out->conn_rate.count = 0;
1187 	else
1188 		out->conn_rate.count -=
1189 		    in->conn_rate.count * diff /
1190 		    in->conn_rate.seconds;
1191 }
1192 
1193 #ifdef ALTQ
1194 /*
1195  * Handle export of struct pf_kaltq to user binaries that may be using any
1196  * version of struct pf_altq.
1197  */
1198 static int
1199 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1200 {
1201 	u_int32_t version;
1202 
1203 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1204 		version = 0;
1205 	else
1206 		version = pa->version;
1207 
1208 	if (version > PFIOC_ALTQ_VERSION)
1209 		return (EINVAL);
1210 
1211 #define ASSIGN(x) exported_q->x = q->x
1212 #define COPY(x) \
1213 	bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1214 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1215 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1216 
1217 	switch (version) {
1218 	case 0: {
1219 		struct pf_altq_v0 *exported_q =
1220 		    &((struct pfioc_altq_v0 *)pa)->altq;
1221 
1222 		COPY(ifname);
1223 
1224 		ASSIGN(scheduler);
1225 		ASSIGN(tbrsize);
1226 		exported_q->tbrsize = SATU16(q->tbrsize);
1227 		exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1228 
1229 		COPY(qname);
1230 		COPY(parent);
1231 		ASSIGN(parent_qid);
1232 		exported_q->bandwidth = SATU32(q->bandwidth);
1233 		ASSIGN(priority);
1234 		ASSIGN(local_flags);
1235 
1236 		ASSIGN(qlimit);
1237 		ASSIGN(flags);
1238 
1239 		if (q->scheduler == ALTQT_HFSC) {
1240 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1241 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1242 			    SATU32(q->pq_u.hfsc_opts.x)
1243 
1244 			ASSIGN_OPT_SATU32(rtsc_m1);
1245 			ASSIGN_OPT(rtsc_d);
1246 			ASSIGN_OPT_SATU32(rtsc_m2);
1247 
1248 			ASSIGN_OPT_SATU32(lssc_m1);
1249 			ASSIGN_OPT(lssc_d);
1250 			ASSIGN_OPT_SATU32(lssc_m2);
1251 
1252 			ASSIGN_OPT_SATU32(ulsc_m1);
1253 			ASSIGN_OPT(ulsc_d);
1254 			ASSIGN_OPT_SATU32(ulsc_m2);
1255 
1256 			ASSIGN_OPT(flags);
1257 
1258 #undef ASSIGN_OPT
1259 #undef ASSIGN_OPT_SATU32
1260 		} else
1261 			COPY(pq_u);
1262 
1263 		ASSIGN(qid);
1264 		break;
1265 	}
1266 	case 1:	{
1267 		struct pf_altq_v1 *exported_q =
1268 		    &((struct pfioc_altq_v1 *)pa)->altq;
1269 
1270 		COPY(ifname);
1271 
1272 		ASSIGN(scheduler);
1273 		ASSIGN(tbrsize);
1274 		ASSIGN(ifbandwidth);
1275 
1276 		COPY(qname);
1277 		COPY(parent);
1278 		ASSIGN(parent_qid);
1279 		ASSIGN(bandwidth);
1280 		ASSIGN(priority);
1281 		ASSIGN(local_flags);
1282 
1283 		ASSIGN(qlimit);
1284 		ASSIGN(flags);
1285 		COPY(pq_u);
1286 
1287 		ASSIGN(qid);
1288 		break;
1289 	}
1290 	default:
1291 		panic("%s: unhandled struct pfioc_altq version", __func__);
1292 		break;
1293 	}
1294 
1295 #undef ASSIGN
1296 #undef COPY
1297 #undef SATU16
1298 #undef SATU32
1299 
1300 	return (0);
1301 }
1302 
1303 /*
1304  * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1305  * that may be using any version of it.
1306  */
1307 static int
1308 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1309 {
1310 	u_int32_t version;
1311 
1312 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1313 		version = 0;
1314 	else
1315 		version = pa->version;
1316 
1317 	if (version > PFIOC_ALTQ_VERSION)
1318 		return (EINVAL);
1319 
1320 #define ASSIGN(x) q->x = imported_q->x
1321 #define COPY(x) \
1322 	bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1323 
1324 	switch (version) {
1325 	case 0: {
1326 		struct pf_altq_v0 *imported_q =
1327 		    &((struct pfioc_altq_v0 *)pa)->altq;
1328 
1329 		COPY(ifname);
1330 
1331 		ASSIGN(scheduler);
1332 		ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1333 		ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1334 
1335 		COPY(qname);
1336 		COPY(parent);
1337 		ASSIGN(parent_qid);
1338 		ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1339 		ASSIGN(priority);
1340 		ASSIGN(local_flags);
1341 
1342 		ASSIGN(qlimit);
1343 		ASSIGN(flags);
1344 
1345 		if (imported_q->scheduler == ALTQT_HFSC) {
1346 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1347 
1348 			/*
1349 			 * The m1 and m2 parameters are being copied from
1350 			 * 32-bit to 64-bit.
1351 			 */
1352 			ASSIGN_OPT(rtsc_m1);
1353 			ASSIGN_OPT(rtsc_d);
1354 			ASSIGN_OPT(rtsc_m2);
1355 
1356 			ASSIGN_OPT(lssc_m1);
1357 			ASSIGN_OPT(lssc_d);
1358 			ASSIGN_OPT(lssc_m2);
1359 
1360 			ASSIGN_OPT(ulsc_m1);
1361 			ASSIGN_OPT(ulsc_d);
1362 			ASSIGN_OPT(ulsc_m2);
1363 
1364 			ASSIGN_OPT(flags);
1365 
1366 #undef ASSIGN_OPT
1367 		} else
1368 			COPY(pq_u);
1369 
1370 		ASSIGN(qid);
1371 		break;
1372 	}
1373 	case 1: {
1374 		struct pf_altq_v1 *imported_q =
1375 		    &((struct pfioc_altq_v1 *)pa)->altq;
1376 
1377 		COPY(ifname);
1378 
1379 		ASSIGN(scheduler);
1380 		ASSIGN(tbrsize);
1381 		ASSIGN(ifbandwidth);
1382 
1383 		COPY(qname);
1384 		COPY(parent);
1385 		ASSIGN(parent_qid);
1386 		ASSIGN(bandwidth);
1387 		ASSIGN(priority);
1388 		ASSIGN(local_flags);
1389 
1390 		ASSIGN(qlimit);
1391 		ASSIGN(flags);
1392 		COPY(pq_u);
1393 
1394 		ASSIGN(qid);
1395 		break;
1396 	}
1397 	default:
1398 		panic("%s: unhandled struct pfioc_altq version", __func__);
1399 		break;
1400 	}
1401 
1402 #undef ASSIGN
1403 #undef COPY
1404 
1405 	return (0);
1406 }
1407 
1408 static struct pf_altq *
1409 pf_altq_get_nth_active(u_int32_t n)
1410 {
1411 	struct pf_altq		*altq;
1412 	u_int32_t		 nr;
1413 
1414 	nr = 0;
1415 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1416 		if (nr == n)
1417 			return (altq);
1418 		nr++;
1419 	}
1420 
1421 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1422 		if (nr == n)
1423 			return (altq);
1424 		nr++;
1425 	}
1426 
1427 	return (NULL);
1428 }
1429 #endif /* ALTQ */
1430 
1431 void
1432 pf_krule_free(struct pf_krule *rule)
1433 {
1434 	if (rule == NULL)
1435 		return;
1436 
1437 	counter_u64_free(rule->evaluations);
1438 	for (int i = 0; i < 2; i++) {
1439 		counter_u64_free(rule->packets[i]);
1440 		counter_u64_free(rule->bytes[i]);
1441 	}
1442 	counter_u64_free(rule->states_cur);
1443 	counter_u64_free(rule->states_tot);
1444 	counter_u64_free(rule->src_nodes);
1445 	free(rule, M_PFRULE);
1446 }
1447 
1448 static void
1449 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool,
1450     struct pf_pooladdr *pool)
1451 {
1452 
1453 	bzero(pool, sizeof(*pool));
1454 	bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr));
1455 	strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname));
1456 }
1457 
1458 static void
1459 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool,
1460     struct pf_kpooladdr *kpool)
1461 {
1462 
1463 	bzero(kpool, sizeof(*kpool));
1464 	bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr));
1465 	strlcpy(kpool->ifname, pool->ifname, sizeof(kpool->ifname));
1466 }
1467 
1468 static void
1469 pf_krule_to_rule(const struct pf_krule *krule, struct pf_rule *rule)
1470 {
1471 
1472 	bzero(rule, sizeof(*rule));
1473 
1474 	bcopy(&krule->src, &rule->src, sizeof(rule->src));
1475 	bcopy(&krule->dst, &rule->dst, sizeof(rule->dst));
1476 
1477 	for (int i = 0; i < PF_SKIP_COUNT; ++i) {
1478 		if (rule->skip[i].ptr == NULL)
1479 			rule->skip[i].nr = -1;
1480 		else
1481 			rule->skip[i].nr = krule->skip[i].ptr->nr;
1482 	}
1483 
1484 	strlcpy(rule->label, krule->label, sizeof(rule->label));
1485 	strlcpy(rule->ifname, krule->ifname, sizeof(rule->ifname));
1486 	strlcpy(rule->qname, krule->qname, sizeof(rule->qname));
1487 	strlcpy(rule->pqname, krule->pqname, sizeof(rule->pqname));
1488 	strlcpy(rule->tagname, krule->tagname, sizeof(rule->tagname));
1489 	strlcpy(rule->match_tagname, krule->match_tagname,
1490 	    sizeof(rule->match_tagname));
1491 	strlcpy(rule->overload_tblname, krule->overload_tblname,
1492 	    sizeof(rule->overload_tblname));
1493 
1494 	bcopy(&krule->rpool, &rule->rpool, sizeof(krule->rpool));
1495 
1496 	rule->evaluations = counter_u64_fetch(krule->evaluations);
1497 	for (int i = 0; i < 2; i++) {
1498 		rule->packets[i] = counter_u64_fetch(krule->packets[i]);
1499 		rule->bytes[i] = counter_u64_fetch(krule->bytes[i]);
1500 	}
1501 
1502 	/* kif, anchor, overload_tbl are not copied over. */
1503 
1504 	rule->os_fingerprint = krule->os_fingerprint;
1505 
1506 	rule->rtableid = krule->rtableid;
1507 	bcopy(krule->timeout, rule->timeout, sizeof(krule->timeout));
1508 	rule->max_states = krule->max_states;
1509 	rule->max_src_nodes = krule->max_src_nodes;
1510 	rule->max_src_states = krule->max_src_states;
1511 	rule->max_src_conn = krule->max_src_conn;
1512 	rule->max_src_conn_rate.limit = krule->max_src_conn_rate.limit;
1513 	rule->max_src_conn_rate.seconds = krule->max_src_conn_rate.seconds;
1514 	rule->qid = krule->qid;
1515 	rule->pqid = krule->pqid;
1516 	rule->rt_listid = krule->rt_listid;
1517 	rule->nr = krule->nr;
1518 	rule->prob = krule->prob;
1519 	rule->cuid = krule->cuid;
1520 	rule->cpid = krule->cpid;
1521 
1522 	rule->return_icmp = krule->return_icmp;
1523 	rule->return_icmp6 = krule->return_icmp6;
1524 	rule->max_mss = krule->max_mss;
1525 	rule->tag = krule->tag;
1526 	rule->match_tag = krule->match_tag;
1527 	rule->scrub_flags = krule->scrub_flags;
1528 
1529 	bcopy(&krule->uid, &rule->uid, sizeof(krule->uid));
1530 	bcopy(&krule->gid, &rule->gid, sizeof(krule->gid));
1531 
1532 	rule->rule_flag = krule->rule_flag;
1533 	rule->action = krule->action;
1534 	rule->direction = krule->direction;
1535 	rule->log = krule->log;
1536 	rule->logif = krule->logif;
1537 	rule->quick = krule->quick;
1538 	rule->ifnot = krule->ifnot;
1539 	rule->match_tag_not = krule->match_tag_not;
1540 	rule->natpass = krule->natpass;
1541 
1542 	rule->keep_state = krule->keep_state;
1543 	rule->af = krule->af;
1544 	rule->proto = krule->proto;
1545 	rule->type = krule->type;
1546 	rule->code = krule->code;
1547 	rule->flags = krule->flags;
1548 	rule->flagset = krule->flagset;
1549 	rule->min_ttl = krule->min_ttl;
1550 	rule->allow_opts = krule->allow_opts;
1551 	rule->rt = krule->rt;
1552 	rule->return_ttl = krule->return_ttl;
1553 	rule->tos = krule->tos;
1554 	rule->set_tos = krule->set_tos;
1555 	rule->anchor_relative = krule->anchor_relative;
1556 	rule->anchor_wildcard = krule->anchor_wildcard;
1557 
1558 	rule->flush = krule->flush;
1559 	rule->prio = krule->prio;
1560 	rule->set_prio[0] = krule->set_prio[0];
1561 	rule->set_prio[1] = krule->set_prio[1];
1562 
1563 	bcopy(&krule->divert, &rule->divert, sizeof(krule->divert));
1564 
1565 	rule->u_states_cur = counter_u64_fetch(krule->states_cur);
1566 	rule->u_states_tot = counter_u64_fetch(krule->states_tot);
1567 	rule->u_src_nodes = counter_u64_fetch(krule->src_nodes);
1568 }
1569 
1570 static int
1571 pf_check_rule_addr(const struct pf_rule_addr *addr)
1572 {
1573 
1574 	switch (addr->addr.type) {
1575 	case PF_ADDR_ADDRMASK:
1576 	case PF_ADDR_NOROUTE:
1577 	case PF_ADDR_DYNIFTL:
1578 	case PF_ADDR_TABLE:
1579 	case PF_ADDR_URPFFAILED:
1580 	case PF_ADDR_RANGE:
1581 		break;
1582 	default:
1583 		return (EINVAL);
1584 	}
1585 
1586 	if (addr->addr.p.dyn != NULL) {
1587 		return (EINVAL);
1588 	}
1589 
1590 	return (0);
1591 }
1592 
1593 static int
1594 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
1595 {
1596 	int ret;
1597 
1598 #ifndef INET
1599 	if (rule->af == AF_INET) {
1600 		return (EAFNOSUPPORT);
1601 	}
1602 #endif /* INET */
1603 #ifndef INET6
1604 	if (rule->af == AF_INET6) {
1605 		return (EAFNOSUPPORT);
1606 	}
1607 #endif /* INET6 */
1608 
1609 	ret = pf_check_rule_addr(&rule->src);
1610 	if (ret != 0)
1611 		return (ret);
1612 	ret = pf_check_rule_addr(&rule->dst);
1613 	if (ret != 0)
1614 		return (ret);
1615 
1616 	bzero(krule, sizeof(*krule));
1617 
1618 	bcopy(&rule->src, &krule->src, sizeof(rule->src));
1619 	bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
1620 
1621 	strlcpy(krule->label, rule->label, sizeof(rule->label));
1622 	strlcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
1623 	strlcpy(krule->qname, rule->qname, sizeof(rule->qname));
1624 	strlcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
1625 	strlcpy(krule->tagname, rule->tagname, sizeof(rule->tagname));
1626 	strlcpy(krule->match_tagname, rule->match_tagname,
1627 	    sizeof(rule->match_tagname));
1628 	strlcpy(krule->overload_tblname, rule->overload_tblname,
1629 	    sizeof(rule->overload_tblname));
1630 
1631 	bcopy(&rule->rpool, &krule->rpool, sizeof(krule->rpool));
1632 
1633 	/* Don't allow userspace to set evaulations, packets or bytes. */
1634 	/* kif, anchor, overload_tbl are not copied over. */
1635 
1636 	krule->os_fingerprint = rule->os_fingerprint;
1637 
1638 	krule->rtableid = rule->rtableid;
1639 	bcopy(rule->timeout, krule->timeout, sizeof(krule->timeout));
1640 	krule->max_states = rule->max_states;
1641 	krule->max_src_nodes = rule->max_src_nodes;
1642 	krule->max_src_states = rule->max_src_states;
1643 	krule->max_src_conn = rule->max_src_conn;
1644 	krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
1645 	krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
1646 	krule->qid = rule->qid;
1647 	krule->pqid = rule->pqid;
1648 	krule->rt_listid = rule->rt_listid;
1649 	krule->nr = rule->nr;
1650 	krule->prob = rule->prob;
1651 	krule->cuid = rule->cuid;
1652 	krule->cpid = rule->cpid;
1653 
1654 	krule->return_icmp = rule->return_icmp;
1655 	krule->return_icmp6 = rule->return_icmp6;
1656 	krule->max_mss = rule->max_mss;
1657 	krule->tag = rule->tag;
1658 	krule->match_tag = rule->match_tag;
1659 	krule->scrub_flags = rule->scrub_flags;
1660 
1661 	bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
1662 	bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
1663 
1664 	krule->rule_flag = rule->rule_flag;
1665 	krule->action = rule->action;
1666 	krule->direction = rule->direction;
1667 	krule->log = rule->log;
1668 	krule->logif = rule->logif;
1669 	krule->quick = rule->quick;
1670 	krule->ifnot = rule->ifnot;
1671 	krule->match_tag_not = rule->match_tag_not;
1672 	krule->natpass = rule->natpass;
1673 
1674 	krule->keep_state = rule->keep_state;
1675 	krule->af = rule->af;
1676 	krule->proto = rule->proto;
1677 	krule->type = rule->type;
1678 	krule->code = rule->code;
1679 	krule->flags = rule->flags;
1680 	krule->flagset = rule->flagset;
1681 	krule->min_ttl = rule->min_ttl;
1682 	krule->allow_opts = rule->allow_opts;
1683 	krule->rt = rule->rt;
1684 	krule->return_ttl = rule->return_ttl;
1685 	krule->tos = rule->tos;
1686 	krule->set_tos = rule->set_tos;
1687 	krule->anchor_relative = rule->anchor_relative;
1688 	krule->anchor_wildcard = rule->anchor_wildcard;
1689 
1690 	krule->flush = rule->flush;
1691 	krule->prio = rule->prio;
1692 	krule->set_prio[0] = rule->set_prio[0];
1693 	krule->set_prio[1] = rule->set_prio[1];
1694 
1695 	bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
1696 
1697 	return (0);
1698 }
1699 
1700 static int
1701 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
1702 {
1703 	int			 error = 0;
1704 	PF_RULES_RLOCK_TRACKER;
1705 
1706 	/* XXX keep in sync with switch() below */
1707 	if (securelevel_gt(td->td_ucred, 2))
1708 		switch (cmd) {
1709 		case DIOCGETRULES:
1710 		case DIOCGETRULE:
1711 		case DIOCGETADDRS:
1712 		case DIOCGETADDR:
1713 		case DIOCGETSTATE:
1714 		case DIOCSETSTATUSIF:
1715 		case DIOCGETSTATUS:
1716 		case DIOCCLRSTATUS:
1717 		case DIOCNATLOOK:
1718 		case DIOCSETDEBUG:
1719 		case DIOCGETSTATES:
1720 		case DIOCGETTIMEOUT:
1721 		case DIOCCLRRULECTRS:
1722 		case DIOCGETLIMIT:
1723 		case DIOCGETALTQSV0:
1724 		case DIOCGETALTQSV1:
1725 		case DIOCGETALTQV0:
1726 		case DIOCGETALTQV1:
1727 		case DIOCGETQSTATSV0:
1728 		case DIOCGETQSTATSV1:
1729 		case DIOCGETRULESETS:
1730 		case DIOCGETRULESET:
1731 		case DIOCRGETTABLES:
1732 		case DIOCRGETTSTATS:
1733 		case DIOCRCLRTSTATS:
1734 		case DIOCRCLRADDRS:
1735 		case DIOCRADDADDRS:
1736 		case DIOCRDELADDRS:
1737 		case DIOCRSETADDRS:
1738 		case DIOCRGETADDRS:
1739 		case DIOCRGETASTATS:
1740 		case DIOCRCLRASTATS:
1741 		case DIOCRTSTADDRS:
1742 		case DIOCOSFPGET:
1743 		case DIOCGETSRCNODES:
1744 		case DIOCCLRSRCNODES:
1745 		case DIOCIGETIFACES:
1746 		case DIOCGIFSPEEDV0:
1747 		case DIOCGIFSPEEDV1:
1748 		case DIOCSETIFFLAG:
1749 		case DIOCCLRIFFLAG:
1750 			break;
1751 		case DIOCRCLRTABLES:
1752 		case DIOCRADDTABLES:
1753 		case DIOCRDELTABLES:
1754 		case DIOCRSETTFLAGS:
1755 			if (((struct pfioc_table *)addr)->pfrio_flags &
1756 			    PFR_FLAG_DUMMY)
1757 				break; /* dummy operation ok */
1758 			return (EPERM);
1759 		default:
1760 			return (EPERM);
1761 		}
1762 
1763 	if (!(flags & FWRITE))
1764 		switch (cmd) {
1765 		case DIOCGETRULES:
1766 		case DIOCGETADDRS:
1767 		case DIOCGETADDR:
1768 		case DIOCGETSTATE:
1769 		case DIOCGETSTATUS:
1770 		case DIOCGETSTATES:
1771 		case DIOCGETTIMEOUT:
1772 		case DIOCGETLIMIT:
1773 		case DIOCGETALTQSV0:
1774 		case DIOCGETALTQSV1:
1775 		case DIOCGETALTQV0:
1776 		case DIOCGETALTQV1:
1777 		case DIOCGETQSTATSV0:
1778 		case DIOCGETQSTATSV1:
1779 		case DIOCGETRULESETS:
1780 		case DIOCGETRULESET:
1781 		case DIOCNATLOOK:
1782 		case DIOCRGETTABLES:
1783 		case DIOCRGETTSTATS:
1784 		case DIOCRGETADDRS:
1785 		case DIOCRGETASTATS:
1786 		case DIOCRTSTADDRS:
1787 		case DIOCOSFPGET:
1788 		case DIOCGETSRCNODES:
1789 		case DIOCIGETIFACES:
1790 		case DIOCGIFSPEEDV1:
1791 		case DIOCGIFSPEEDV0:
1792 			break;
1793 		case DIOCRCLRTABLES:
1794 		case DIOCRADDTABLES:
1795 		case DIOCRDELTABLES:
1796 		case DIOCRCLRTSTATS:
1797 		case DIOCRCLRADDRS:
1798 		case DIOCRADDADDRS:
1799 		case DIOCRDELADDRS:
1800 		case DIOCRSETADDRS:
1801 		case DIOCRSETTFLAGS:
1802 			if (((struct pfioc_table *)addr)->pfrio_flags &
1803 			    PFR_FLAG_DUMMY) {
1804 				flags |= FWRITE; /* need write lock for dummy */
1805 				break; /* dummy operation ok */
1806 			}
1807 			return (EACCES);
1808 		case DIOCGETRULE:
1809 			if (((struct pfioc_rule *)addr)->action ==
1810 			    PF_GET_CLR_CNTR)
1811 				return (EACCES);
1812 			break;
1813 		default:
1814 			return (EACCES);
1815 		}
1816 
1817 	CURVNET_SET(TD_TO_VNET(td));
1818 
1819 	switch (cmd) {
1820 	case DIOCSTART:
1821 		sx_xlock(&pf_ioctl_lock);
1822 		if (V_pf_status.running)
1823 			error = EEXIST;
1824 		else {
1825 			int cpu;
1826 
1827 			hook_pf();
1828 			V_pf_status.running = 1;
1829 			V_pf_status.since = time_second;
1830 
1831 			CPU_FOREACH(cpu)
1832 				V_pf_stateid[cpu] = time_second;
1833 
1834 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1835 		}
1836 		break;
1837 
1838 	case DIOCSTOP:
1839 		sx_xlock(&pf_ioctl_lock);
1840 		if (!V_pf_status.running)
1841 			error = ENOENT;
1842 		else {
1843 			V_pf_status.running = 0;
1844 			dehook_pf();
1845 			V_pf_status.since = time_second;
1846 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1847 		}
1848 		break;
1849 
1850 	case DIOCADDRULE: {
1851 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1852 		struct pf_kruleset	*ruleset;
1853 		struct pf_krule		*rule, *tail;
1854 		struct pf_kpooladdr	*pa;
1855 		struct pfi_kkif		*kif = NULL;
1856 		int			 rs_num;
1857 
1858 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1859 			error = EINVAL;
1860 			break;
1861 		}
1862 
1863 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
1864 		error = pf_rule_to_krule(&pr->rule, rule);
1865 		if (error != 0) {
1866 			free(rule, M_PFRULE);
1867 			break;
1868 		}
1869 
1870 		if (rule->ifname[0])
1871 			kif = pf_kkif_create(M_WAITOK);
1872 		rule->evaluations = counter_u64_alloc(M_WAITOK);
1873 		for (int i = 0; i < 2; i++) {
1874 			rule->packets[i] = counter_u64_alloc(M_WAITOK);
1875 			rule->bytes[i] = counter_u64_alloc(M_WAITOK);
1876 		}
1877 		rule->states_cur = counter_u64_alloc(M_WAITOK);
1878 		rule->states_tot = counter_u64_alloc(M_WAITOK);
1879 		rule->src_nodes = counter_u64_alloc(M_WAITOK);
1880 		rule->cuid = td->td_ucred->cr_ruid;
1881 		rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1882 		TAILQ_INIT(&rule->rpool.list);
1883 #define	ERROUT(x)	{ error = (x); goto DIOCADDRULE_error; }
1884 
1885 		PF_RULES_WLOCK();
1886 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1887 		ruleset = pf_find_kruleset(pr->anchor);
1888 		if (ruleset == NULL)
1889 			ERROUT(EINVAL);
1890 		rs_num = pf_get_ruleset_number(pr->rule.action);
1891 		if (rs_num >= PF_RULESET_MAX)
1892 			ERROUT(EINVAL);
1893 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1894 			DPFPRINTF(PF_DEBUG_MISC,
1895 			    ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
1896 			    ruleset->rules[rs_num].inactive.ticket));
1897 			ERROUT(EBUSY);
1898 		}
1899 		if (pr->pool_ticket != V_ticket_pabuf) {
1900 			DPFPRINTF(PF_DEBUG_MISC,
1901 			    ("pool_ticket: %d != %d\n", pr->pool_ticket,
1902 			    V_ticket_pabuf));
1903 			ERROUT(EBUSY);
1904 		}
1905 
1906 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1907 		    pf_krulequeue);
1908 		if (tail)
1909 			rule->nr = tail->nr + 1;
1910 		else
1911 			rule->nr = 0;
1912 		if (rule->ifname[0]) {
1913 			rule->kif = pfi_kkif_attach(kif, rule->ifname);
1914 			pfi_kkif_ref(rule->kif);
1915 		} else
1916 			rule->kif = NULL;
1917 
1918 		if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
1919 			error = EBUSY;
1920 
1921 #ifdef ALTQ
1922 		/* set queue IDs */
1923 		if (rule->qname[0] != 0) {
1924 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1925 				error = EBUSY;
1926 			else if (rule->pqname[0] != 0) {
1927 				if ((rule->pqid =
1928 				    pf_qname2qid(rule->pqname)) == 0)
1929 					error = EBUSY;
1930 			} else
1931 				rule->pqid = rule->qid;
1932 		}
1933 #endif
1934 		if (rule->tagname[0])
1935 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1936 				error = EBUSY;
1937 		if (rule->match_tagname[0])
1938 			if ((rule->match_tag =
1939 			    pf_tagname2tag(rule->match_tagname)) == 0)
1940 				error = EBUSY;
1941 		if (rule->rt && !rule->direction)
1942 			error = EINVAL;
1943 		if (!rule->log)
1944 			rule->logif = 0;
1945 		if (rule->logif >= PFLOGIFS_MAX)
1946 			error = EINVAL;
1947 		if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1948 			error = ENOMEM;
1949 		if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1950 			error = ENOMEM;
1951 		if (pf_kanchor_setup(rule, ruleset, pr->anchor_call))
1952 			error = EINVAL;
1953 		if (rule->scrub_flags & PFSTATE_SETPRIO &&
1954 		    (rule->set_prio[0] > PF_PRIO_MAX ||
1955 		    rule->set_prio[1] > PF_PRIO_MAX))
1956 			error = EINVAL;
1957 		TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1958 			if (pa->addr.type == PF_ADDR_TABLE) {
1959 				pa->addr.p.tbl = pfr_attach_table(ruleset,
1960 				    pa->addr.v.tblname);
1961 				if (pa->addr.p.tbl == NULL)
1962 					error = ENOMEM;
1963 			}
1964 
1965 		rule->overload_tbl = NULL;
1966 		if (rule->overload_tblname[0]) {
1967 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1968 			    rule->overload_tblname)) == NULL)
1969 				error = EINVAL;
1970 			else
1971 				rule->overload_tbl->pfrkt_flags |=
1972 				    PFR_TFLAG_ACTIVE;
1973 		}
1974 
1975 		pf_mv_kpool(&V_pf_pabuf, &rule->rpool.list);
1976 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1977 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1978 		    (rule->rt > PF_NOPFROUTE)) &&
1979 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1980 			error = EINVAL;
1981 
1982 		if (error) {
1983 			pf_free_rule(rule);
1984 			PF_RULES_WUNLOCK();
1985 			break;
1986 		}
1987 
1988 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1989 		counter_u64_zero(rule->evaluations);
1990 		for (int i = 0; i < 2; i++) {
1991 			counter_u64_zero(rule->packets[i]);
1992 			counter_u64_zero(rule->bytes[i]);
1993 		}
1994 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1995 		    rule, entries);
1996 		ruleset->rules[rs_num].inactive.rcount++;
1997 		PF_RULES_WUNLOCK();
1998 		break;
1999 
2000 #undef ERROUT
2001 DIOCADDRULE_error:
2002 		PF_RULES_WUNLOCK();
2003 		pf_krule_free(rule);
2004 		pf_kkif_free(kif);
2005 		break;
2006 	}
2007 
2008 	case DIOCGETRULES: {
2009 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
2010 		struct pf_kruleset	*ruleset;
2011 		struct pf_krule		*tail;
2012 		int			 rs_num;
2013 
2014 		PF_RULES_WLOCK();
2015 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
2016 		ruleset = pf_find_kruleset(pr->anchor);
2017 		if (ruleset == NULL) {
2018 			PF_RULES_WUNLOCK();
2019 			error = EINVAL;
2020 			break;
2021 		}
2022 		rs_num = pf_get_ruleset_number(pr->rule.action);
2023 		if (rs_num >= PF_RULESET_MAX) {
2024 			PF_RULES_WUNLOCK();
2025 			error = EINVAL;
2026 			break;
2027 		}
2028 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
2029 		    pf_krulequeue);
2030 		if (tail)
2031 			pr->nr = tail->nr + 1;
2032 		else
2033 			pr->nr = 0;
2034 		pr->ticket = ruleset->rules[rs_num].active.ticket;
2035 		PF_RULES_WUNLOCK();
2036 		break;
2037 	}
2038 
2039 	case DIOCGETRULE: {
2040 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
2041 		struct pf_kruleset	*ruleset;
2042 		struct pf_krule		*rule;
2043 		int			 rs_num;
2044 
2045 		PF_RULES_WLOCK();
2046 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
2047 		ruleset = pf_find_kruleset(pr->anchor);
2048 		if (ruleset == NULL) {
2049 			PF_RULES_WUNLOCK();
2050 			error = EINVAL;
2051 			break;
2052 		}
2053 		rs_num = pf_get_ruleset_number(pr->rule.action);
2054 		if (rs_num >= PF_RULESET_MAX) {
2055 			PF_RULES_WUNLOCK();
2056 			error = EINVAL;
2057 			break;
2058 		}
2059 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
2060 			PF_RULES_WUNLOCK();
2061 			error = EBUSY;
2062 			break;
2063 		}
2064 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
2065 		while ((rule != NULL) && (rule->nr != pr->nr))
2066 			rule = TAILQ_NEXT(rule, entries);
2067 		if (rule == NULL) {
2068 			PF_RULES_WUNLOCK();
2069 			error = EBUSY;
2070 			break;
2071 		}
2072 
2073 		pf_krule_to_rule(rule, &pr->rule);
2074 
2075 		if (pf_kanchor_copyout(ruleset, rule, pr)) {
2076 			PF_RULES_WUNLOCK();
2077 			error = EBUSY;
2078 			break;
2079 		}
2080 		pf_addr_copyout(&pr->rule.src.addr);
2081 		pf_addr_copyout(&pr->rule.dst.addr);
2082 
2083 		if (pr->action == PF_GET_CLR_CNTR) {
2084 			counter_u64_zero(rule->evaluations);
2085 			for (int i = 0; i < 2; i++) {
2086 				counter_u64_zero(rule->packets[i]);
2087 				counter_u64_zero(rule->bytes[i]);
2088 			}
2089 			counter_u64_zero(rule->states_tot);
2090 		}
2091 		PF_RULES_WUNLOCK();
2092 		break;
2093 	}
2094 
2095 	case DIOCCHANGERULE: {
2096 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
2097 		struct pf_kruleset	*ruleset;
2098 		struct pf_krule		*oldrule = NULL, *newrule = NULL;
2099 		struct pfi_kkif		*kif = NULL;
2100 		struct pf_kpooladdr	*pa;
2101 		u_int32_t		 nr = 0;
2102 		int			 rs_num;
2103 
2104 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
2105 		    pcr->action > PF_CHANGE_GET_TICKET) {
2106 			error = EINVAL;
2107 			break;
2108 		}
2109 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
2110 			error = EINVAL;
2111 			break;
2112 		}
2113 
2114 		if (pcr->action != PF_CHANGE_REMOVE) {
2115 			newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK);
2116 			error = pf_rule_to_krule(&pcr->rule, newrule);
2117 			if (error != 0) {
2118 				free(newrule, M_PFRULE);
2119 				break;
2120 			}
2121 
2122 			if (newrule->ifname[0])
2123 				kif = pf_kkif_create(M_WAITOK);
2124 			newrule->evaluations = counter_u64_alloc(M_WAITOK);
2125 			for (int i = 0; i < 2; i++) {
2126 				newrule->packets[i] =
2127 				    counter_u64_alloc(M_WAITOK);
2128 				newrule->bytes[i] =
2129 				    counter_u64_alloc(M_WAITOK);
2130 			}
2131 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
2132 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
2133 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
2134 			newrule->cuid = td->td_ucred->cr_ruid;
2135 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
2136 			TAILQ_INIT(&newrule->rpool.list);
2137 		}
2138 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGERULE_error; }
2139 
2140 		PF_RULES_WLOCK();
2141 		if (!(pcr->action == PF_CHANGE_REMOVE ||
2142 		    pcr->action == PF_CHANGE_GET_TICKET) &&
2143 		    pcr->pool_ticket != V_ticket_pabuf)
2144 			ERROUT(EBUSY);
2145 
2146 		ruleset = pf_find_kruleset(pcr->anchor);
2147 		if (ruleset == NULL)
2148 			ERROUT(EINVAL);
2149 
2150 		rs_num = pf_get_ruleset_number(pcr->rule.action);
2151 		if (rs_num >= PF_RULESET_MAX)
2152 			ERROUT(EINVAL);
2153 
2154 		if (pcr->action == PF_CHANGE_GET_TICKET) {
2155 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
2156 			ERROUT(0);
2157 		} else if (pcr->ticket !=
2158 			    ruleset->rules[rs_num].active.ticket)
2159 				ERROUT(EINVAL);
2160 
2161 		if (pcr->action != PF_CHANGE_REMOVE) {
2162 			if (newrule->ifname[0]) {
2163 				newrule->kif = pfi_kkif_attach(kif,
2164 				    newrule->ifname);
2165 				pfi_kkif_ref(newrule->kif);
2166 			} else
2167 				newrule->kif = NULL;
2168 
2169 			if (newrule->rtableid > 0 &&
2170 			    newrule->rtableid >= rt_numfibs)
2171 				error = EBUSY;
2172 
2173 #ifdef ALTQ
2174 			/* set queue IDs */
2175 			if (newrule->qname[0] != 0) {
2176 				if ((newrule->qid =
2177 				    pf_qname2qid(newrule->qname)) == 0)
2178 					error = EBUSY;
2179 				else if (newrule->pqname[0] != 0) {
2180 					if ((newrule->pqid =
2181 					    pf_qname2qid(newrule->pqname)) == 0)
2182 						error = EBUSY;
2183 				} else
2184 					newrule->pqid = newrule->qid;
2185 			}
2186 #endif /* ALTQ */
2187 			if (newrule->tagname[0])
2188 				if ((newrule->tag =
2189 				    pf_tagname2tag(newrule->tagname)) == 0)
2190 					error = EBUSY;
2191 			if (newrule->match_tagname[0])
2192 				if ((newrule->match_tag = pf_tagname2tag(
2193 				    newrule->match_tagname)) == 0)
2194 					error = EBUSY;
2195 			if (newrule->rt && !newrule->direction)
2196 				error = EINVAL;
2197 			if (!newrule->log)
2198 				newrule->logif = 0;
2199 			if (newrule->logif >= PFLOGIFS_MAX)
2200 				error = EINVAL;
2201 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
2202 				error = ENOMEM;
2203 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
2204 				error = ENOMEM;
2205 			if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
2206 				error = EINVAL;
2207 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
2208 				if (pa->addr.type == PF_ADDR_TABLE) {
2209 					pa->addr.p.tbl =
2210 					    pfr_attach_table(ruleset,
2211 					    pa->addr.v.tblname);
2212 					if (pa->addr.p.tbl == NULL)
2213 						error = ENOMEM;
2214 				}
2215 
2216 			newrule->overload_tbl = NULL;
2217 			if (newrule->overload_tblname[0]) {
2218 				if ((newrule->overload_tbl = pfr_attach_table(
2219 				    ruleset, newrule->overload_tblname)) ==
2220 				    NULL)
2221 					error = EINVAL;
2222 				else
2223 					newrule->overload_tbl->pfrkt_flags |=
2224 					    PFR_TFLAG_ACTIVE;
2225 			}
2226 
2227 			pf_mv_kpool(&V_pf_pabuf, &newrule->rpool.list);
2228 			if (((((newrule->action == PF_NAT) ||
2229 			    (newrule->action == PF_RDR) ||
2230 			    (newrule->action == PF_BINAT) ||
2231 			    (newrule->rt > PF_NOPFROUTE)) &&
2232 			    !newrule->anchor)) &&
2233 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
2234 				error = EINVAL;
2235 
2236 			if (error) {
2237 				pf_free_rule(newrule);
2238 				PF_RULES_WUNLOCK();
2239 				break;
2240 			}
2241 
2242 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
2243 		}
2244 		pf_empty_kpool(&V_pf_pabuf);
2245 
2246 		if (pcr->action == PF_CHANGE_ADD_HEAD)
2247 			oldrule = TAILQ_FIRST(
2248 			    ruleset->rules[rs_num].active.ptr);
2249 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
2250 			oldrule = TAILQ_LAST(
2251 			    ruleset->rules[rs_num].active.ptr, pf_krulequeue);
2252 		else {
2253 			oldrule = TAILQ_FIRST(
2254 			    ruleset->rules[rs_num].active.ptr);
2255 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
2256 				oldrule = TAILQ_NEXT(oldrule, entries);
2257 			if (oldrule == NULL) {
2258 				if (newrule != NULL)
2259 					pf_free_rule(newrule);
2260 				PF_RULES_WUNLOCK();
2261 				error = EINVAL;
2262 				break;
2263 			}
2264 		}
2265 
2266 		if (pcr->action == PF_CHANGE_REMOVE) {
2267 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
2268 			    oldrule);
2269 			ruleset->rules[rs_num].active.rcount--;
2270 		} else {
2271 			if (oldrule == NULL)
2272 				TAILQ_INSERT_TAIL(
2273 				    ruleset->rules[rs_num].active.ptr,
2274 				    newrule, entries);
2275 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
2276 			    pcr->action == PF_CHANGE_ADD_BEFORE)
2277 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
2278 			else
2279 				TAILQ_INSERT_AFTER(
2280 				    ruleset->rules[rs_num].active.ptr,
2281 				    oldrule, newrule, entries);
2282 			ruleset->rules[rs_num].active.rcount++;
2283 		}
2284 
2285 		nr = 0;
2286 		TAILQ_FOREACH(oldrule,
2287 		    ruleset->rules[rs_num].active.ptr, entries)
2288 			oldrule->nr = nr++;
2289 
2290 		ruleset->rules[rs_num].active.ticket++;
2291 
2292 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
2293 		pf_remove_if_empty_kruleset(ruleset);
2294 
2295 		PF_RULES_WUNLOCK();
2296 		break;
2297 
2298 #undef ERROUT
2299 DIOCCHANGERULE_error:
2300 		PF_RULES_WUNLOCK();
2301 		pf_krule_free(newrule);
2302 		pf_kkif_free(kif);
2303 		break;
2304 	}
2305 
2306 	case DIOCCLRSTATES: {
2307 		struct pf_state		*s;
2308 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
2309 		u_int			 i, killed = 0;
2310 
2311 		for (i = 0; i <= pf_hashmask; i++) {
2312 			struct pf_idhash *ih = &V_pf_idhash[i];
2313 
2314 relock_DIOCCLRSTATES:
2315 			PF_HASHROW_LOCK(ih);
2316 			LIST_FOREACH(s, &ih->states, entry)
2317 				if (!psk->psk_ifname[0] ||
2318 				    !strcmp(psk->psk_ifname,
2319 				    s->kif->pfik_name)) {
2320 					/*
2321 					 * Don't send out individual
2322 					 * delete messages.
2323 					 */
2324 					s->state_flags |= PFSTATE_NOSYNC;
2325 					pf_unlink_state(s, PF_ENTER_LOCKED);
2326 					killed++;
2327 					goto relock_DIOCCLRSTATES;
2328 				}
2329 			PF_HASHROW_UNLOCK(ih);
2330 		}
2331 		psk->psk_killed = killed;
2332 		if (V_pfsync_clear_states_ptr != NULL)
2333 			V_pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname);
2334 		break;
2335 	}
2336 
2337 	case DIOCKILLSTATES: {
2338 		struct pf_state		*s;
2339 		struct pf_state_key	*sk;
2340 		struct pf_addr		*srcaddr, *dstaddr;
2341 		u_int16_t		 srcport, dstport;
2342 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
2343 		u_int			 i, killed = 0;
2344 
2345 		if (psk->psk_pfcmp.id) {
2346 			if (psk->psk_pfcmp.creatorid == 0)
2347 				psk->psk_pfcmp.creatorid = V_pf_status.hostid;
2348 			if ((s = pf_find_state_byid(psk->psk_pfcmp.id,
2349 			    psk->psk_pfcmp.creatorid))) {
2350 				pf_unlink_state(s, PF_ENTER_LOCKED);
2351 				psk->psk_killed = 1;
2352 			}
2353 			break;
2354 		}
2355 
2356 		for (i = 0; i <= pf_hashmask; i++) {
2357 			struct pf_idhash *ih = &V_pf_idhash[i];
2358 
2359 relock_DIOCKILLSTATES:
2360 			PF_HASHROW_LOCK(ih);
2361 			LIST_FOREACH(s, &ih->states, entry) {
2362 				sk = s->key[PF_SK_WIRE];
2363 				if (s->direction == PF_OUT) {
2364 					srcaddr = &sk->addr[1];
2365 					dstaddr = &sk->addr[0];
2366 					srcport = sk->port[1];
2367 					dstport = sk->port[0];
2368 				} else {
2369 					srcaddr = &sk->addr[0];
2370 					dstaddr = &sk->addr[1];
2371 					srcport = sk->port[0];
2372 					dstport = sk->port[1];
2373 				}
2374 
2375 				if ((!psk->psk_af || sk->af == psk->psk_af)
2376 				    && (!psk->psk_proto || psk->psk_proto ==
2377 				    sk->proto) &&
2378 				    PF_MATCHA(psk->psk_src.neg,
2379 				    &psk->psk_src.addr.v.a.addr,
2380 				    &psk->psk_src.addr.v.a.mask,
2381 				    srcaddr, sk->af) &&
2382 				    PF_MATCHA(psk->psk_dst.neg,
2383 				    &psk->psk_dst.addr.v.a.addr,
2384 				    &psk->psk_dst.addr.v.a.mask,
2385 				    dstaddr, sk->af) &&
2386 				    (psk->psk_src.port_op == 0 ||
2387 				    pf_match_port(psk->psk_src.port_op,
2388 				    psk->psk_src.port[0], psk->psk_src.port[1],
2389 				    srcport)) &&
2390 				    (psk->psk_dst.port_op == 0 ||
2391 				    pf_match_port(psk->psk_dst.port_op,
2392 				    psk->psk_dst.port[0], psk->psk_dst.port[1],
2393 				    dstport)) &&
2394 				    (!psk->psk_label[0] ||
2395 				    (s->rule.ptr->label[0] &&
2396 				    !strcmp(psk->psk_label,
2397 				    s->rule.ptr->label))) &&
2398 				    (!psk->psk_ifname[0] ||
2399 				    !strcmp(psk->psk_ifname,
2400 				    s->kif->pfik_name))) {
2401 					pf_unlink_state(s, PF_ENTER_LOCKED);
2402 					killed++;
2403 					goto relock_DIOCKILLSTATES;
2404 				}
2405 			}
2406 			PF_HASHROW_UNLOCK(ih);
2407 		}
2408 		psk->psk_killed = killed;
2409 		break;
2410 	}
2411 
2412 	case DIOCADDSTATE: {
2413 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
2414 		struct pfsync_state	*sp = &ps->state;
2415 
2416 		if (sp->timeout >= PFTM_MAX) {
2417 			error = EINVAL;
2418 			break;
2419 		}
2420 		if (V_pfsync_state_import_ptr != NULL) {
2421 			PF_RULES_RLOCK();
2422 			error = V_pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
2423 			PF_RULES_RUNLOCK();
2424 		} else
2425 			error = EOPNOTSUPP;
2426 		break;
2427 	}
2428 
2429 	case DIOCGETSTATE: {
2430 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
2431 		struct pf_state		*s;
2432 
2433 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
2434 		if (s == NULL) {
2435 			error = ENOENT;
2436 			break;
2437 		}
2438 
2439 		pfsync_state_export(&ps->state, s);
2440 		PF_STATE_UNLOCK(s);
2441 		break;
2442 	}
2443 
2444 	case DIOCGETSTATES: {
2445 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
2446 		struct pf_state		*s;
2447 		struct pfsync_state	*pstore, *p;
2448 		int i, nr;
2449 
2450 		if (ps->ps_len <= 0) {
2451 			nr = uma_zone_get_cur(V_pf_state_z);
2452 			ps->ps_len = sizeof(struct pfsync_state) * nr;
2453 			break;
2454 		}
2455 
2456 		p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK | M_ZERO);
2457 		nr = 0;
2458 
2459 		for (i = 0; i <= pf_hashmask; i++) {
2460 			struct pf_idhash *ih = &V_pf_idhash[i];
2461 
2462 			PF_HASHROW_LOCK(ih);
2463 			LIST_FOREACH(s, &ih->states, entry) {
2464 				if (s->timeout == PFTM_UNLINKED)
2465 					continue;
2466 
2467 				if ((nr+1) * sizeof(*p) > ps->ps_len) {
2468 					PF_HASHROW_UNLOCK(ih);
2469 					goto DIOCGETSTATES_full;
2470 				}
2471 				pfsync_state_export(p, s);
2472 				p++;
2473 				nr++;
2474 			}
2475 			PF_HASHROW_UNLOCK(ih);
2476 		}
2477 DIOCGETSTATES_full:
2478 		error = copyout(pstore, ps->ps_states,
2479 		    sizeof(struct pfsync_state) * nr);
2480 		if (error) {
2481 			free(pstore, M_TEMP);
2482 			break;
2483 		}
2484 		ps->ps_len = sizeof(struct pfsync_state) * nr;
2485 		free(pstore, M_TEMP);
2486 
2487 		break;
2488 	}
2489 
2490 	case DIOCGETSTATUS: {
2491 		struct pf_status *s = (struct pf_status *)addr;
2492 
2493 		PF_RULES_RLOCK();
2494 		s->running = V_pf_status.running;
2495 		s->since   = V_pf_status.since;
2496 		s->debug   = V_pf_status.debug;
2497 		s->hostid  = V_pf_status.hostid;
2498 		s->states  = V_pf_status.states;
2499 		s->src_nodes = V_pf_status.src_nodes;
2500 
2501 		for (int i = 0; i < PFRES_MAX; i++)
2502 			s->counters[i] =
2503 			    counter_u64_fetch(V_pf_status.counters[i]);
2504 		for (int i = 0; i < LCNT_MAX; i++)
2505 			s->lcounters[i] =
2506 			    counter_u64_fetch(V_pf_status.lcounters[i]);
2507 		for (int i = 0; i < FCNT_MAX; i++)
2508 			s->fcounters[i] =
2509 			    counter_u64_fetch(V_pf_status.fcounters[i]);
2510 		for (int i = 0; i < SCNT_MAX; i++)
2511 			s->scounters[i] =
2512 			    counter_u64_fetch(V_pf_status.scounters[i]);
2513 
2514 		bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ);
2515 		bcopy(V_pf_status.pf_chksum, s->pf_chksum,
2516 		    PF_MD5_DIGEST_LENGTH);
2517 
2518 		pfi_update_status(s->ifname, s);
2519 		PF_RULES_RUNLOCK();
2520 		break;
2521 	}
2522 
2523 	case DIOCSETSTATUSIF: {
2524 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
2525 
2526 		if (pi->ifname[0] == 0) {
2527 			bzero(V_pf_status.ifname, IFNAMSIZ);
2528 			break;
2529 		}
2530 		PF_RULES_WLOCK();
2531 		strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
2532 		PF_RULES_WUNLOCK();
2533 		break;
2534 	}
2535 
2536 	case DIOCCLRSTATUS: {
2537 		PF_RULES_WLOCK();
2538 		for (int i = 0; i < PFRES_MAX; i++)
2539 			counter_u64_zero(V_pf_status.counters[i]);
2540 		for (int i = 0; i < FCNT_MAX; i++)
2541 			counter_u64_zero(V_pf_status.fcounters[i]);
2542 		for (int i = 0; i < SCNT_MAX; i++)
2543 			counter_u64_zero(V_pf_status.scounters[i]);
2544 		for (int i = 0; i < LCNT_MAX; i++)
2545 			counter_u64_zero(V_pf_status.lcounters[i]);
2546 		V_pf_status.since = time_second;
2547 		if (*V_pf_status.ifname)
2548 			pfi_update_status(V_pf_status.ifname, NULL);
2549 		PF_RULES_WUNLOCK();
2550 		break;
2551 	}
2552 
2553 	case DIOCNATLOOK: {
2554 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
2555 		struct pf_state_key	*sk;
2556 		struct pf_state		*state;
2557 		struct pf_state_key_cmp	 key;
2558 		int			 m = 0, direction = pnl->direction;
2559 		int			 sidx, didx;
2560 
2561 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
2562 		sidx = (direction == PF_IN) ? 1 : 0;
2563 		didx = (direction == PF_IN) ? 0 : 1;
2564 
2565 		if (!pnl->proto ||
2566 		    PF_AZERO(&pnl->saddr, pnl->af) ||
2567 		    PF_AZERO(&pnl->daddr, pnl->af) ||
2568 		    ((pnl->proto == IPPROTO_TCP ||
2569 		    pnl->proto == IPPROTO_UDP) &&
2570 		    (!pnl->dport || !pnl->sport)))
2571 			error = EINVAL;
2572 		else {
2573 			bzero(&key, sizeof(key));
2574 			key.af = pnl->af;
2575 			key.proto = pnl->proto;
2576 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
2577 			key.port[sidx] = pnl->sport;
2578 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
2579 			key.port[didx] = pnl->dport;
2580 
2581 			state = pf_find_state_all(&key, direction, &m);
2582 
2583 			if (m > 1)
2584 				error = E2BIG;	/* more than one state */
2585 			else if (state != NULL) {
2586 				/* XXXGL: not locked read */
2587 				sk = state->key[sidx];
2588 				PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
2589 				pnl->rsport = sk->port[sidx];
2590 				PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
2591 				pnl->rdport = sk->port[didx];
2592 			} else
2593 				error = ENOENT;
2594 		}
2595 		break;
2596 	}
2597 
2598 	case DIOCSETTIMEOUT: {
2599 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2600 		int		 old;
2601 
2602 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
2603 		    pt->seconds < 0) {
2604 			error = EINVAL;
2605 			break;
2606 		}
2607 		PF_RULES_WLOCK();
2608 		old = V_pf_default_rule.timeout[pt->timeout];
2609 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
2610 			pt->seconds = 1;
2611 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
2612 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
2613 			wakeup(pf_purge_thread);
2614 		pt->seconds = old;
2615 		PF_RULES_WUNLOCK();
2616 		break;
2617 	}
2618 
2619 	case DIOCGETTIMEOUT: {
2620 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2621 
2622 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
2623 			error = EINVAL;
2624 			break;
2625 		}
2626 		PF_RULES_RLOCK();
2627 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
2628 		PF_RULES_RUNLOCK();
2629 		break;
2630 	}
2631 
2632 	case DIOCGETLIMIT: {
2633 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2634 
2635 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
2636 			error = EINVAL;
2637 			break;
2638 		}
2639 		PF_RULES_RLOCK();
2640 		pl->limit = V_pf_limits[pl->index].limit;
2641 		PF_RULES_RUNLOCK();
2642 		break;
2643 	}
2644 
2645 	case DIOCSETLIMIT: {
2646 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2647 		int			 old_limit;
2648 
2649 		PF_RULES_WLOCK();
2650 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
2651 		    V_pf_limits[pl->index].zone == NULL) {
2652 			PF_RULES_WUNLOCK();
2653 			error = EINVAL;
2654 			break;
2655 		}
2656 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
2657 		old_limit = V_pf_limits[pl->index].limit;
2658 		V_pf_limits[pl->index].limit = pl->limit;
2659 		pl->limit = old_limit;
2660 		PF_RULES_WUNLOCK();
2661 		break;
2662 	}
2663 
2664 	case DIOCSETDEBUG: {
2665 		u_int32_t	*level = (u_int32_t *)addr;
2666 
2667 		PF_RULES_WLOCK();
2668 		V_pf_status.debug = *level;
2669 		PF_RULES_WUNLOCK();
2670 		break;
2671 	}
2672 
2673 	case DIOCCLRRULECTRS: {
2674 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
2675 		struct pf_kruleset	*ruleset = &pf_main_ruleset;
2676 		struct pf_krule		*rule;
2677 
2678 		PF_RULES_WLOCK();
2679 		TAILQ_FOREACH(rule,
2680 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
2681 			counter_u64_zero(rule->evaluations);
2682 			for (int i = 0; i < 2; i++) {
2683 				counter_u64_zero(rule->packets[i]);
2684 				counter_u64_zero(rule->bytes[i]);
2685 			}
2686 		}
2687 		PF_RULES_WUNLOCK();
2688 		break;
2689 	}
2690 
2691 	case DIOCGIFSPEEDV0:
2692 	case DIOCGIFSPEEDV1: {
2693 		struct pf_ifspeed_v1	*psp = (struct pf_ifspeed_v1 *)addr;
2694 		struct pf_ifspeed_v1	ps;
2695 		struct ifnet		*ifp;
2696 
2697 		if (psp->ifname[0] != 0) {
2698 			/* Can we completely trust user-land? */
2699 			strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
2700 			ifp = ifunit(ps.ifname);
2701 			if (ifp != NULL) {
2702 				psp->baudrate32 =
2703 				    (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
2704 				if (cmd == DIOCGIFSPEEDV1)
2705 					psp->baudrate = ifp->if_baudrate;
2706 			} else
2707 				error = EINVAL;
2708 		} else
2709 			error = EINVAL;
2710 		break;
2711 	}
2712 
2713 #ifdef ALTQ
2714 	case DIOCSTARTALTQ: {
2715 		struct pf_altq		*altq;
2716 
2717 		PF_RULES_WLOCK();
2718 		/* enable all altq interfaces on active list */
2719 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
2720 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
2721 				error = pf_enable_altq(altq);
2722 				if (error != 0)
2723 					break;
2724 			}
2725 		}
2726 		if (error == 0)
2727 			V_pf_altq_running = 1;
2728 		PF_RULES_WUNLOCK();
2729 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
2730 		break;
2731 	}
2732 
2733 	case DIOCSTOPALTQ: {
2734 		struct pf_altq		*altq;
2735 
2736 		PF_RULES_WLOCK();
2737 		/* disable all altq interfaces on active list */
2738 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
2739 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
2740 				error = pf_disable_altq(altq);
2741 				if (error != 0)
2742 					break;
2743 			}
2744 		}
2745 		if (error == 0)
2746 			V_pf_altq_running = 0;
2747 		PF_RULES_WUNLOCK();
2748 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2749 		break;
2750 	}
2751 
2752 	case DIOCADDALTQV0:
2753 	case DIOCADDALTQV1: {
2754 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2755 		struct pf_altq		*altq, *a;
2756 		struct ifnet		*ifp;
2757 
2758 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
2759 		error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
2760 		if (error)
2761 			break;
2762 		altq->local_flags = 0;
2763 
2764 		PF_RULES_WLOCK();
2765 		if (pa->ticket != V_ticket_altqs_inactive) {
2766 			PF_RULES_WUNLOCK();
2767 			free(altq, M_PFALTQ);
2768 			error = EBUSY;
2769 			break;
2770 		}
2771 
2772 		/*
2773 		 * if this is for a queue, find the discipline and
2774 		 * copy the necessary fields
2775 		 */
2776 		if (altq->qname[0] != 0) {
2777 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2778 				PF_RULES_WUNLOCK();
2779 				error = EBUSY;
2780 				free(altq, M_PFALTQ);
2781 				break;
2782 			}
2783 			altq->altq_disc = NULL;
2784 			TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
2785 				if (strncmp(a->ifname, altq->ifname,
2786 				    IFNAMSIZ) == 0) {
2787 					altq->altq_disc = a->altq_disc;
2788 					break;
2789 				}
2790 			}
2791 		}
2792 
2793 		if ((ifp = ifunit(altq->ifname)) == NULL)
2794 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
2795 		else
2796 			error = altq_add(ifp, altq);
2797 
2798 		if (error) {
2799 			PF_RULES_WUNLOCK();
2800 			free(altq, M_PFALTQ);
2801 			break;
2802 		}
2803 
2804 		if (altq->qname[0] != 0)
2805 			TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
2806 		else
2807 			TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
2808 		/* version error check done on import above */
2809 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
2810 		PF_RULES_WUNLOCK();
2811 		break;
2812 	}
2813 
2814 	case DIOCGETALTQSV0:
2815 	case DIOCGETALTQSV1: {
2816 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2817 		struct pf_altq		*altq;
2818 
2819 		PF_RULES_RLOCK();
2820 		pa->nr = 0;
2821 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
2822 			pa->nr++;
2823 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
2824 			pa->nr++;
2825 		pa->ticket = V_ticket_altqs_active;
2826 		PF_RULES_RUNLOCK();
2827 		break;
2828 	}
2829 
2830 	case DIOCGETALTQV0:
2831 	case DIOCGETALTQV1: {
2832 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
2833 		struct pf_altq		*altq;
2834 
2835 		PF_RULES_RLOCK();
2836 		if (pa->ticket != V_ticket_altqs_active) {
2837 			PF_RULES_RUNLOCK();
2838 			error = EBUSY;
2839 			break;
2840 		}
2841 		altq = pf_altq_get_nth_active(pa->nr);
2842 		if (altq == NULL) {
2843 			PF_RULES_RUNLOCK();
2844 			error = EBUSY;
2845 			break;
2846 		}
2847 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
2848 		PF_RULES_RUNLOCK();
2849 		break;
2850 	}
2851 
2852 	case DIOCCHANGEALTQV0:
2853 	case DIOCCHANGEALTQV1:
2854 		/* CHANGEALTQ not supported yet! */
2855 		error = ENODEV;
2856 		break;
2857 
2858 	case DIOCGETQSTATSV0:
2859 	case DIOCGETQSTATSV1: {
2860 		struct pfioc_qstats_v1	*pq = (struct pfioc_qstats_v1 *)addr;
2861 		struct pf_altq		*altq;
2862 		int			 nbytes;
2863 		u_int32_t		 version;
2864 
2865 		PF_RULES_RLOCK();
2866 		if (pq->ticket != V_ticket_altqs_active) {
2867 			PF_RULES_RUNLOCK();
2868 			error = EBUSY;
2869 			break;
2870 		}
2871 		nbytes = pq->nbytes;
2872 		altq = pf_altq_get_nth_active(pq->nr);
2873 		if (altq == NULL) {
2874 			PF_RULES_RUNLOCK();
2875 			error = EBUSY;
2876 			break;
2877 		}
2878 
2879 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
2880 			PF_RULES_RUNLOCK();
2881 			error = ENXIO;
2882 			break;
2883 		}
2884 		PF_RULES_RUNLOCK();
2885 		if (cmd == DIOCGETQSTATSV0)
2886 			version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
2887 		else
2888 			version = pq->version;
2889 		error = altq_getqstats(altq, pq->buf, &nbytes, version);
2890 		if (error == 0) {
2891 			pq->scheduler = altq->scheduler;
2892 			pq->nbytes = nbytes;
2893 		}
2894 		break;
2895 	}
2896 #endif /* ALTQ */
2897 
2898 	case DIOCBEGINADDRS: {
2899 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2900 
2901 		PF_RULES_WLOCK();
2902 		pf_empty_kpool(&V_pf_pabuf);
2903 		pp->ticket = ++V_ticket_pabuf;
2904 		PF_RULES_WUNLOCK();
2905 		break;
2906 	}
2907 
2908 	case DIOCADDADDR: {
2909 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2910 		struct pf_kpooladdr	*pa;
2911 		struct pfi_kkif		*kif = NULL;
2912 
2913 #ifndef INET
2914 		if (pp->af == AF_INET) {
2915 			error = EAFNOSUPPORT;
2916 			break;
2917 		}
2918 #endif /* INET */
2919 #ifndef INET6
2920 		if (pp->af == AF_INET6) {
2921 			error = EAFNOSUPPORT;
2922 			break;
2923 		}
2924 #endif /* INET6 */
2925 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2926 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2927 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2928 			error = EINVAL;
2929 			break;
2930 		}
2931 		if (pp->addr.addr.p.dyn != NULL) {
2932 			error = EINVAL;
2933 			break;
2934 		}
2935 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2936 		pf_pooladdr_to_kpooladdr(&pp->addr, pa);
2937 		if (pa->ifname[0])
2938 			kif = pf_kkif_create(M_WAITOK);
2939 		PF_RULES_WLOCK();
2940 		if (pp->ticket != V_ticket_pabuf) {
2941 			PF_RULES_WUNLOCK();
2942 			if (pa->ifname[0])
2943 				pf_kkif_free(kif);
2944 			free(pa, M_PFRULE);
2945 			error = EBUSY;
2946 			break;
2947 		}
2948 		if (pa->ifname[0]) {
2949 			pa->kif = pfi_kkif_attach(kif, pa->ifname);
2950 			pfi_kkif_ref(pa->kif);
2951 		} else
2952 			pa->kif = NULL;
2953 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2954 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2955 			if (pa->ifname[0])
2956 				pfi_kkif_unref(pa->kif);
2957 			PF_RULES_WUNLOCK();
2958 			free(pa, M_PFRULE);
2959 			break;
2960 		}
2961 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2962 		PF_RULES_WUNLOCK();
2963 		break;
2964 	}
2965 
2966 	case DIOCGETADDRS: {
2967 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2968 		struct pf_kpool		*pool;
2969 		struct pf_kpooladdr	*pa;
2970 
2971 		PF_RULES_RLOCK();
2972 		pp->nr = 0;
2973 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2974 		    pp->r_num, 0, 1, 0);
2975 		if (pool == NULL) {
2976 			PF_RULES_RUNLOCK();
2977 			error = EBUSY;
2978 			break;
2979 		}
2980 		TAILQ_FOREACH(pa, &pool->list, entries)
2981 			pp->nr++;
2982 		PF_RULES_RUNLOCK();
2983 		break;
2984 	}
2985 
2986 	case DIOCGETADDR: {
2987 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2988 		struct pf_kpool		*pool;
2989 		struct pf_kpooladdr	*pa;
2990 		u_int32_t		 nr = 0;
2991 
2992 		PF_RULES_RLOCK();
2993 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2994 		    pp->r_num, 0, 1, 1);
2995 		if (pool == NULL) {
2996 			PF_RULES_RUNLOCK();
2997 			error = EBUSY;
2998 			break;
2999 		}
3000 		pa = TAILQ_FIRST(&pool->list);
3001 		while ((pa != NULL) && (nr < pp->nr)) {
3002 			pa = TAILQ_NEXT(pa, entries);
3003 			nr++;
3004 		}
3005 		if (pa == NULL) {
3006 			PF_RULES_RUNLOCK();
3007 			error = EBUSY;
3008 			break;
3009 		}
3010 		pf_kpooladdr_to_pooladdr(pa, &pp->addr);
3011 		pf_addr_copyout(&pp->addr.addr);
3012 		PF_RULES_RUNLOCK();
3013 		break;
3014 	}
3015 
3016 	case DIOCCHANGEADDR: {
3017 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
3018 		struct pf_kpool		*pool;
3019 		struct pf_kpooladdr	*oldpa = NULL, *newpa = NULL;
3020 		struct pf_kruleset	*ruleset;
3021 		struct pfi_kkif		*kif = NULL;
3022 
3023 		if (pca->action < PF_CHANGE_ADD_HEAD ||
3024 		    pca->action > PF_CHANGE_REMOVE) {
3025 			error = EINVAL;
3026 			break;
3027 		}
3028 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
3029 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
3030 		    pca->addr.addr.type != PF_ADDR_TABLE) {
3031 			error = EINVAL;
3032 			break;
3033 		}
3034 		if (pca->addr.addr.p.dyn != NULL) {
3035 			error = EINVAL;
3036 			break;
3037 		}
3038 
3039 		if (pca->action != PF_CHANGE_REMOVE) {
3040 #ifndef INET
3041 			if (pca->af == AF_INET) {
3042 				error = EAFNOSUPPORT;
3043 				break;
3044 			}
3045 #endif /* INET */
3046 #ifndef INET6
3047 			if (pca->af == AF_INET6) {
3048 				error = EAFNOSUPPORT;
3049 				break;
3050 			}
3051 #endif /* INET6 */
3052 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
3053 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
3054 			if (newpa->ifname[0])
3055 				kif = pf_kkif_create(M_WAITOK);
3056 			newpa->kif = NULL;
3057 		}
3058 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGEADDR_error; }
3059 		PF_RULES_WLOCK();
3060 		ruleset = pf_find_kruleset(pca->anchor);
3061 		if (ruleset == NULL)
3062 			ERROUT(EBUSY);
3063 
3064 		pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
3065 		    pca->r_num, pca->r_last, 1, 1);
3066 		if (pool == NULL)
3067 			ERROUT(EBUSY);
3068 
3069 		if (pca->action != PF_CHANGE_REMOVE) {
3070 			if (newpa->ifname[0]) {
3071 				newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
3072 				pfi_kkif_ref(newpa->kif);
3073 				kif = NULL;
3074 			}
3075 
3076 			switch (newpa->addr.type) {
3077 			case PF_ADDR_DYNIFTL:
3078 				error = pfi_dynaddr_setup(&newpa->addr,
3079 				    pca->af);
3080 				break;
3081 			case PF_ADDR_TABLE:
3082 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
3083 				    newpa->addr.v.tblname);
3084 				if (newpa->addr.p.tbl == NULL)
3085 					error = ENOMEM;
3086 				break;
3087 			}
3088 			if (error)
3089 				goto DIOCCHANGEADDR_error;
3090 		}
3091 
3092 		switch (pca->action) {
3093 		case PF_CHANGE_ADD_HEAD:
3094 			oldpa = TAILQ_FIRST(&pool->list);
3095 			break;
3096 		case PF_CHANGE_ADD_TAIL:
3097 			oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
3098 			break;
3099 		default:
3100 			oldpa = TAILQ_FIRST(&pool->list);
3101 			for (int i = 0; oldpa && i < pca->nr; i++)
3102 				oldpa = TAILQ_NEXT(oldpa, entries);
3103 
3104 			if (oldpa == NULL)
3105 				ERROUT(EINVAL);
3106 		}
3107 
3108 		if (pca->action == PF_CHANGE_REMOVE) {
3109 			TAILQ_REMOVE(&pool->list, oldpa, entries);
3110 			switch (oldpa->addr.type) {
3111 			case PF_ADDR_DYNIFTL:
3112 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
3113 				break;
3114 			case PF_ADDR_TABLE:
3115 				pfr_detach_table(oldpa->addr.p.tbl);
3116 				break;
3117 			}
3118 			if (oldpa->kif)
3119 				pfi_kkif_unref(oldpa->kif);
3120 			free(oldpa, M_PFRULE);
3121 		} else {
3122 			if (oldpa == NULL)
3123 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
3124 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
3125 			    pca->action == PF_CHANGE_ADD_BEFORE)
3126 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
3127 			else
3128 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
3129 				    newpa, entries);
3130 		}
3131 
3132 		pool->cur = TAILQ_FIRST(&pool->list);
3133 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
3134 		PF_RULES_WUNLOCK();
3135 		break;
3136 
3137 #undef ERROUT
3138 DIOCCHANGEADDR_error:
3139 		if (newpa != NULL) {
3140 			if (newpa->kif)
3141 				pfi_kkif_unref(newpa->kif);
3142 			free(newpa, M_PFRULE);
3143 		}
3144 		PF_RULES_WUNLOCK();
3145 		pf_kkif_free(kif);
3146 		break;
3147 	}
3148 
3149 	case DIOCGETRULESETS: {
3150 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
3151 		struct pf_kruleset	*ruleset;
3152 		struct pf_kanchor	*anchor;
3153 
3154 		PF_RULES_RLOCK();
3155 		pr->path[sizeof(pr->path) - 1] = 0;
3156 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
3157 			PF_RULES_RUNLOCK();
3158 			error = ENOENT;
3159 			break;
3160 		}
3161 		pr->nr = 0;
3162 		if (ruleset->anchor == NULL) {
3163 			/* XXX kludge for pf_main_ruleset */
3164 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
3165 				if (anchor->parent == NULL)
3166 					pr->nr++;
3167 		} else {
3168 			RB_FOREACH(anchor, pf_kanchor_node,
3169 			    &ruleset->anchor->children)
3170 				pr->nr++;
3171 		}
3172 		PF_RULES_RUNLOCK();
3173 		break;
3174 	}
3175 
3176 	case DIOCGETRULESET: {
3177 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
3178 		struct pf_kruleset	*ruleset;
3179 		struct pf_kanchor	*anchor;
3180 		u_int32_t		 nr = 0;
3181 
3182 		PF_RULES_RLOCK();
3183 		pr->path[sizeof(pr->path) - 1] = 0;
3184 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
3185 			PF_RULES_RUNLOCK();
3186 			error = ENOENT;
3187 			break;
3188 		}
3189 		pr->name[0] = 0;
3190 		if (ruleset->anchor == NULL) {
3191 			/* XXX kludge for pf_main_ruleset */
3192 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
3193 				if (anchor->parent == NULL && nr++ == pr->nr) {
3194 					strlcpy(pr->name, anchor->name,
3195 					    sizeof(pr->name));
3196 					break;
3197 				}
3198 		} else {
3199 			RB_FOREACH(anchor, pf_kanchor_node,
3200 			    &ruleset->anchor->children)
3201 				if (nr++ == pr->nr) {
3202 					strlcpy(pr->name, anchor->name,
3203 					    sizeof(pr->name));
3204 					break;
3205 				}
3206 		}
3207 		if (!pr->name[0])
3208 			error = EBUSY;
3209 		PF_RULES_RUNLOCK();
3210 		break;
3211 	}
3212 
3213 	case DIOCRCLRTABLES: {
3214 		struct pfioc_table *io = (struct pfioc_table *)addr;
3215 
3216 		if (io->pfrio_esize != 0) {
3217 			error = ENODEV;
3218 			break;
3219 		}
3220 		PF_RULES_WLOCK();
3221 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
3222 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
3223 		PF_RULES_WUNLOCK();
3224 		break;
3225 	}
3226 
3227 	case DIOCRADDTABLES: {
3228 		struct pfioc_table *io = (struct pfioc_table *)addr;
3229 		struct pfr_table *pfrts;
3230 		size_t totlen;
3231 
3232 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3233 			error = ENODEV;
3234 			break;
3235 		}
3236 
3237 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
3238 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
3239 			error = ENOMEM;
3240 			break;
3241 		}
3242 
3243 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3244 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3245 		    M_TEMP, M_WAITOK);
3246 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3247 		if (error) {
3248 			free(pfrts, M_TEMP);
3249 			break;
3250 		}
3251 		PF_RULES_WLOCK();
3252 		error = pfr_add_tables(pfrts, io->pfrio_size,
3253 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3254 		PF_RULES_WUNLOCK();
3255 		free(pfrts, M_TEMP);
3256 		break;
3257 	}
3258 
3259 	case DIOCRDELTABLES: {
3260 		struct pfioc_table *io = (struct pfioc_table *)addr;
3261 		struct pfr_table *pfrts;
3262 		size_t totlen;
3263 
3264 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3265 			error = ENODEV;
3266 			break;
3267 		}
3268 
3269 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
3270 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
3271 			error = ENOMEM;
3272 			break;
3273 		}
3274 
3275 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3276 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3277 		    M_TEMP, M_WAITOK);
3278 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3279 		if (error) {
3280 			free(pfrts, M_TEMP);
3281 			break;
3282 		}
3283 		PF_RULES_WLOCK();
3284 		error = pfr_del_tables(pfrts, io->pfrio_size,
3285 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3286 		PF_RULES_WUNLOCK();
3287 		free(pfrts, M_TEMP);
3288 		break;
3289 	}
3290 
3291 	case DIOCRGETTABLES: {
3292 		struct pfioc_table *io = (struct pfioc_table *)addr;
3293 		struct pfr_table *pfrts;
3294 		size_t totlen;
3295 		int n;
3296 
3297 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3298 			error = ENODEV;
3299 			break;
3300 		}
3301 		PF_RULES_RLOCK();
3302 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3303 		if (n < 0) {
3304 			PF_RULES_RUNLOCK();
3305 			error = EINVAL;
3306 			break;
3307 		}
3308 		io->pfrio_size = min(io->pfrio_size, n);
3309 
3310 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3311 
3312 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3313 		    M_TEMP, M_NOWAIT);
3314 		if (pfrts == NULL) {
3315 			error = ENOMEM;
3316 			PF_RULES_RUNLOCK();
3317 			break;
3318 		}
3319 		error = pfr_get_tables(&io->pfrio_table, pfrts,
3320 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3321 		PF_RULES_RUNLOCK();
3322 		if (error == 0)
3323 			error = copyout(pfrts, io->pfrio_buffer, totlen);
3324 		free(pfrts, M_TEMP);
3325 		break;
3326 	}
3327 
3328 	case DIOCRGETTSTATS: {
3329 		struct pfioc_table *io = (struct pfioc_table *)addr;
3330 		struct pfr_tstats *pfrtstats;
3331 		size_t totlen;
3332 		int n;
3333 
3334 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
3335 			error = ENODEV;
3336 			break;
3337 		}
3338 		PF_RULES_WLOCK();
3339 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3340 		if (n < 0) {
3341 			PF_RULES_WUNLOCK();
3342 			error = EINVAL;
3343 			break;
3344 		}
3345 		io->pfrio_size = min(io->pfrio_size, n);
3346 
3347 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
3348 		pfrtstats = mallocarray(io->pfrio_size,
3349 		    sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT);
3350 		if (pfrtstats == NULL) {
3351 			error = ENOMEM;
3352 			PF_RULES_WUNLOCK();
3353 			break;
3354 		}
3355 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
3356 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3357 		PF_RULES_WUNLOCK();
3358 		if (error == 0)
3359 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
3360 		free(pfrtstats, M_TEMP);
3361 		break;
3362 	}
3363 
3364 	case DIOCRCLRTSTATS: {
3365 		struct pfioc_table *io = (struct pfioc_table *)addr;
3366 		struct pfr_table *pfrts;
3367 		size_t totlen;
3368 
3369 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3370 			error = ENODEV;
3371 			break;
3372 		}
3373 
3374 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
3375 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
3376 			/* We used to count tables and use the minimum required
3377 			 * size, so we didn't fail on overly large requests.
3378 			 * Keep doing so. */
3379 			io->pfrio_size = pf_ioctl_maxcount;
3380 			break;
3381 		}
3382 
3383 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3384 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3385 		    M_TEMP, M_NOWAIT);
3386 		if (pfrts == NULL) {
3387 			error = ENOMEM;
3388 			break;
3389 		}
3390 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3391 		if (error) {
3392 			free(pfrts, M_TEMP);
3393 			break;
3394 		}
3395 
3396 		PF_RULES_WLOCK();
3397 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
3398 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3399 		PF_RULES_WUNLOCK();
3400 		free(pfrts, M_TEMP);
3401 		break;
3402 	}
3403 
3404 	case DIOCRSETTFLAGS: {
3405 		struct pfioc_table *io = (struct pfioc_table *)addr;
3406 		struct pfr_table *pfrts;
3407 		size_t totlen;
3408 		int n;
3409 
3410 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3411 			error = ENODEV;
3412 			break;
3413 		}
3414 
3415 		PF_RULES_RLOCK();
3416 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3417 		if (n < 0) {
3418 			PF_RULES_RUNLOCK();
3419 			error = EINVAL;
3420 			break;
3421 		}
3422 
3423 		io->pfrio_size = min(io->pfrio_size, n);
3424 		PF_RULES_RUNLOCK();
3425 
3426 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3427 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3428 		    M_TEMP, M_WAITOK);
3429 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3430 		if (error) {
3431 			free(pfrts, M_TEMP);
3432 			break;
3433 		}
3434 		PF_RULES_WLOCK();
3435 		error = pfr_set_tflags(pfrts, io->pfrio_size,
3436 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
3437 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3438 		PF_RULES_WUNLOCK();
3439 		free(pfrts, M_TEMP);
3440 		break;
3441 	}
3442 
3443 	case DIOCRCLRADDRS: {
3444 		struct pfioc_table *io = (struct pfioc_table *)addr;
3445 
3446 		if (io->pfrio_esize != 0) {
3447 			error = ENODEV;
3448 			break;
3449 		}
3450 		PF_RULES_WLOCK();
3451 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
3452 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
3453 		PF_RULES_WUNLOCK();
3454 		break;
3455 	}
3456 
3457 	case DIOCRADDADDRS: {
3458 		struct pfioc_table *io = (struct pfioc_table *)addr;
3459 		struct pfr_addr *pfras;
3460 		size_t totlen;
3461 
3462 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3463 			error = ENODEV;
3464 			break;
3465 		}
3466 		if (io->pfrio_size < 0 ||
3467 		    io->pfrio_size > pf_ioctl_maxcount ||
3468 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3469 			error = EINVAL;
3470 			break;
3471 		}
3472 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3473 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3474 		    M_TEMP, M_NOWAIT);
3475 		if (! pfras) {
3476 			error = ENOMEM;
3477 			break;
3478 		}
3479 		error = copyin(io->pfrio_buffer, pfras, totlen);
3480 		if (error) {
3481 			free(pfras, M_TEMP);
3482 			break;
3483 		}
3484 		PF_RULES_WLOCK();
3485 		error = pfr_add_addrs(&io->pfrio_table, pfras,
3486 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
3487 		    PFR_FLAG_USERIOCTL);
3488 		PF_RULES_WUNLOCK();
3489 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3490 			error = copyout(pfras, io->pfrio_buffer, totlen);
3491 		free(pfras, M_TEMP);
3492 		break;
3493 	}
3494 
3495 	case DIOCRDELADDRS: {
3496 		struct pfioc_table *io = (struct pfioc_table *)addr;
3497 		struct pfr_addr *pfras;
3498 		size_t totlen;
3499 
3500 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3501 			error = ENODEV;
3502 			break;
3503 		}
3504 		if (io->pfrio_size < 0 ||
3505 		    io->pfrio_size > pf_ioctl_maxcount ||
3506 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3507 			error = EINVAL;
3508 			break;
3509 		}
3510 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3511 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3512 		    M_TEMP, M_NOWAIT);
3513 		if (! pfras) {
3514 			error = ENOMEM;
3515 			break;
3516 		}
3517 		error = copyin(io->pfrio_buffer, pfras, totlen);
3518 		if (error) {
3519 			free(pfras, M_TEMP);
3520 			break;
3521 		}
3522 		PF_RULES_WLOCK();
3523 		error = pfr_del_addrs(&io->pfrio_table, pfras,
3524 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
3525 		    PFR_FLAG_USERIOCTL);
3526 		PF_RULES_WUNLOCK();
3527 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3528 			error = copyout(pfras, io->pfrio_buffer, totlen);
3529 		free(pfras, M_TEMP);
3530 		break;
3531 	}
3532 
3533 	case DIOCRSETADDRS: {
3534 		struct pfioc_table *io = (struct pfioc_table *)addr;
3535 		struct pfr_addr *pfras;
3536 		size_t totlen, count;
3537 
3538 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3539 			error = ENODEV;
3540 			break;
3541 		}
3542 		if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
3543 			error = EINVAL;
3544 			break;
3545 		}
3546 		count = max(io->pfrio_size, io->pfrio_size2);
3547 		if (count > pf_ioctl_maxcount ||
3548 		    WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
3549 			error = EINVAL;
3550 			break;
3551 		}
3552 		totlen = count * sizeof(struct pfr_addr);
3553 		pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
3554 		    M_NOWAIT);
3555 		if (! pfras) {
3556 			error = ENOMEM;
3557 			break;
3558 		}
3559 		error = copyin(io->pfrio_buffer, pfras, totlen);
3560 		if (error) {
3561 			free(pfras, M_TEMP);
3562 			break;
3563 		}
3564 		PF_RULES_WLOCK();
3565 		error = pfr_set_addrs(&io->pfrio_table, pfras,
3566 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
3567 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
3568 		    PFR_FLAG_USERIOCTL, 0);
3569 		PF_RULES_WUNLOCK();
3570 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3571 			error = copyout(pfras, io->pfrio_buffer, totlen);
3572 		free(pfras, M_TEMP);
3573 		break;
3574 	}
3575 
3576 	case DIOCRGETADDRS: {
3577 		struct pfioc_table *io = (struct pfioc_table *)addr;
3578 		struct pfr_addr *pfras;
3579 		size_t totlen;
3580 
3581 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3582 			error = ENODEV;
3583 			break;
3584 		}
3585 		if (io->pfrio_size < 0 ||
3586 		    io->pfrio_size > pf_ioctl_maxcount ||
3587 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3588 			error = EINVAL;
3589 			break;
3590 		}
3591 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3592 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3593 		    M_TEMP, M_NOWAIT);
3594 		if (! pfras) {
3595 			error = ENOMEM;
3596 			break;
3597 		}
3598 		PF_RULES_RLOCK();
3599 		error = pfr_get_addrs(&io->pfrio_table, pfras,
3600 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3601 		PF_RULES_RUNLOCK();
3602 		if (error == 0)
3603 			error = copyout(pfras, io->pfrio_buffer, totlen);
3604 		free(pfras, M_TEMP);
3605 		break;
3606 	}
3607 
3608 	case DIOCRGETASTATS: {
3609 		struct pfioc_table *io = (struct pfioc_table *)addr;
3610 		struct pfr_astats *pfrastats;
3611 		size_t totlen;
3612 
3613 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
3614 			error = ENODEV;
3615 			break;
3616 		}
3617 		if (io->pfrio_size < 0 ||
3618 		    io->pfrio_size > pf_ioctl_maxcount ||
3619 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
3620 			error = EINVAL;
3621 			break;
3622 		}
3623 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
3624 		pfrastats = mallocarray(io->pfrio_size,
3625 		    sizeof(struct pfr_astats), M_TEMP, M_NOWAIT);
3626 		if (! pfrastats) {
3627 			error = ENOMEM;
3628 			break;
3629 		}
3630 		PF_RULES_RLOCK();
3631 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
3632 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3633 		PF_RULES_RUNLOCK();
3634 		if (error == 0)
3635 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
3636 		free(pfrastats, M_TEMP);
3637 		break;
3638 	}
3639 
3640 	case DIOCRCLRASTATS: {
3641 		struct pfioc_table *io = (struct pfioc_table *)addr;
3642 		struct pfr_addr *pfras;
3643 		size_t totlen;
3644 
3645 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3646 			error = ENODEV;
3647 			break;
3648 		}
3649 		if (io->pfrio_size < 0 ||
3650 		    io->pfrio_size > pf_ioctl_maxcount ||
3651 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3652 			error = EINVAL;
3653 			break;
3654 		}
3655 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3656 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3657 		    M_TEMP, M_NOWAIT);
3658 		if (! pfras) {
3659 			error = ENOMEM;
3660 			break;
3661 		}
3662 		error = copyin(io->pfrio_buffer, pfras, totlen);
3663 		if (error) {
3664 			free(pfras, M_TEMP);
3665 			break;
3666 		}
3667 		PF_RULES_WLOCK();
3668 		error = pfr_clr_astats(&io->pfrio_table, pfras,
3669 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
3670 		    PFR_FLAG_USERIOCTL);
3671 		PF_RULES_WUNLOCK();
3672 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
3673 			error = copyout(pfras, io->pfrio_buffer, totlen);
3674 		free(pfras, M_TEMP);
3675 		break;
3676 	}
3677 
3678 	case DIOCRTSTADDRS: {
3679 		struct pfioc_table *io = (struct pfioc_table *)addr;
3680 		struct pfr_addr *pfras;
3681 		size_t totlen;
3682 
3683 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3684 			error = ENODEV;
3685 			break;
3686 		}
3687 		if (io->pfrio_size < 0 ||
3688 		    io->pfrio_size > pf_ioctl_maxcount ||
3689 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3690 			error = EINVAL;
3691 			break;
3692 		}
3693 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3694 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3695 		    M_TEMP, M_NOWAIT);
3696 		if (! pfras) {
3697 			error = ENOMEM;
3698 			break;
3699 		}
3700 		error = copyin(io->pfrio_buffer, pfras, totlen);
3701 		if (error) {
3702 			free(pfras, M_TEMP);
3703 			break;
3704 		}
3705 		PF_RULES_RLOCK();
3706 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
3707 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
3708 		    PFR_FLAG_USERIOCTL);
3709 		PF_RULES_RUNLOCK();
3710 		if (error == 0)
3711 			error = copyout(pfras, io->pfrio_buffer, totlen);
3712 		free(pfras, M_TEMP);
3713 		break;
3714 	}
3715 
3716 	case DIOCRINADEFINE: {
3717 		struct pfioc_table *io = (struct pfioc_table *)addr;
3718 		struct pfr_addr *pfras;
3719 		size_t totlen;
3720 
3721 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
3722 			error = ENODEV;
3723 			break;
3724 		}
3725 		if (io->pfrio_size < 0 ||
3726 		    io->pfrio_size > pf_ioctl_maxcount ||
3727 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
3728 			error = EINVAL;
3729 			break;
3730 		}
3731 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
3732 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
3733 		    M_TEMP, M_NOWAIT);
3734 		if (! pfras) {
3735 			error = ENOMEM;
3736 			break;
3737 		}
3738 		error = copyin(io->pfrio_buffer, pfras, totlen);
3739 		if (error) {
3740 			free(pfras, M_TEMP);
3741 			break;
3742 		}
3743 		PF_RULES_WLOCK();
3744 		error = pfr_ina_define(&io->pfrio_table, pfras,
3745 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
3746 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3747 		PF_RULES_WUNLOCK();
3748 		free(pfras, M_TEMP);
3749 		break;
3750 	}
3751 
3752 	case DIOCOSFPADD: {
3753 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
3754 		PF_RULES_WLOCK();
3755 		error = pf_osfp_add(io);
3756 		PF_RULES_WUNLOCK();
3757 		break;
3758 	}
3759 
3760 	case DIOCOSFPGET: {
3761 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
3762 		PF_RULES_RLOCK();
3763 		error = pf_osfp_get(io);
3764 		PF_RULES_RUNLOCK();
3765 		break;
3766 	}
3767 
3768 	case DIOCXBEGIN: {
3769 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3770 		struct pfioc_trans_e	*ioes, *ioe;
3771 		size_t			 totlen;
3772 		int			 i;
3773 
3774 		if (io->esize != sizeof(*ioe)) {
3775 			error = ENODEV;
3776 			break;
3777 		}
3778 		if (io->size < 0 ||
3779 		    io->size > pf_ioctl_maxcount ||
3780 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3781 			error = EINVAL;
3782 			break;
3783 		}
3784 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3785 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3786 		    M_TEMP, M_NOWAIT);
3787 		if (! ioes) {
3788 			error = ENOMEM;
3789 			break;
3790 		}
3791 		error = copyin(io->array, ioes, totlen);
3792 		if (error) {
3793 			free(ioes, M_TEMP);
3794 			break;
3795 		}
3796 		PF_RULES_WLOCK();
3797 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3798 			switch (ioe->rs_num) {
3799 #ifdef ALTQ
3800 			case PF_RULESET_ALTQ:
3801 				if (ioe->anchor[0]) {
3802 					PF_RULES_WUNLOCK();
3803 					free(ioes, M_TEMP);
3804 					error = EINVAL;
3805 					goto fail;
3806 				}
3807 				if ((error = pf_begin_altq(&ioe->ticket))) {
3808 					PF_RULES_WUNLOCK();
3809 					free(ioes, M_TEMP);
3810 					goto fail;
3811 				}
3812 				break;
3813 #endif /* ALTQ */
3814 			case PF_RULESET_TABLE:
3815 			    {
3816 				struct pfr_table table;
3817 
3818 				bzero(&table, sizeof(table));
3819 				strlcpy(table.pfrt_anchor, ioe->anchor,
3820 				    sizeof(table.pfrt_anchor));
3821 				if ((error = pfr_ina_begin(&table,
3822 				    &ioe->ticket, NULL, 0))) {
3823 					PF_RULES_WUNLOCK();
3824 					free(ioes, M_TEMP);
3825 					goto fail;
3826 				}
3827 				break;
3828 			    }
3829 			default:
3830 				if ((error = pf_begin_rules(&ioe->ticket,
3831 				    ioe->rs_num, ioe->anchor))) {
3832 					PF_RULES_WUNLOCK();
3833 					free(ioes, M_TEMP);
3834 					goto fail;
3835 				}
3836 				break;
3837 			}
3838 		}
3839 		PF_RULES_WUNLOCK();
3840 		error = copyout(ioes, io->array, totlen);
3841 		free(ioes, M_TEMP);
3842 		break;
3843 	}
3844 
3845 	case DIOCXROLLBACK: {
3846 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3847 		struct pfioc_trans_e	*ioe, *ioes;
3848 		size_t			 totlen;
3849 		int			 i;
3850 
3851 		if (io->esize != sizeof(*ioe)) {
3852 			error = ENODEV;
3853 			break;
3854 		}
3855 		if (io->size < 0 ||
3856 		    io->size > pf_ioctl_maxcount ||
3857 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3858 			error = EINVAL;
3859 			break;
3860 		}
3861 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3862 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3863 		    M_TEMP, M_NOWAIT);
3864 		if (! ioes) {
3865 			error = ENOMEM;
3866 			break;
3867 		}
3868 		error = copyin(io->array, ioes, totlen);
3869 		if (error) {
3870 			free(ioes, M_TEMP);
3871 			break;
3872 		}
3873 		PF_RULES_WLOCK();
3874 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3875 			switch (ioe->rs_num) {
3876 #ifdef ALTQ
3877 			case PF_RULESET_ALTQ:
3878 				if (ioe->anchor[0]) {
3879 					PF_RULES_WUNLOCK();
3880 					free(ioes, M_TEMP);
3881 					error = EINVAL;
3882 					goto fail;
3883 				}
3884 				if ((error = pf_rollback_altq(ioe->ticket))) {
3885 					PF_RULES_WUNLOCK();
3886 					free(ioes, M_TEMP);
3887 					goto fail; /* really bad */
3888 				}
3889 				break;
3890 #endif /* ALTQ */
3891 			case PF_RULESET_TABLE:
3892 			    {
3893 				struct pfr_table table;
3894 
3895 				bzero(&table, sizeof(table));
3896 				strlcpy(table.pfrt_anchor, ioe->anchor,
3897 				    sizeof(table.pfrt_anchor));
3898 				if ((error = pfr_ina_rollback(&table,
3899 				    ioe->ticket, NULL, 0))) {
3900 					PF_RULES_WUNLOCK();
3901 					free(ioes, M_TEMP);
3902 					goto fail; /* really bad */
3903 				}
3904 				break;
3905 			    }
3906 			default:
3907 				if ((error = pf_rollback_rules(ioe->ticket,
3908 				    ioe->rs_num, ioe->anchor))) {
3909 					PF_RULES_WUNLOCK();
3910 					free(ioes, M_TEMP);
3911 					goto fail; /* really bad */
3912 				}
3913 				break;
3914 			}
3915 		}
3916 		PF_RULES_WUNLOCK();
3917 		free(ioes, M_TEMP);
3918 		break;
3919 	}
3920 
3921 	case DIOCXCOMMIT: {
3922 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
3923 		struct pfioc_trans_e	*ioe, *ioes;
3924 		struct pf_kruleset	*rs;
3925 		size_t			 totlen;
3926 		int			 i;
3927 
3928 		if (io->esize != sizeof(*ioe)) {
3929 			error = ENODEV;
3930 			break;
3931 		}
3932 
3933 		if (io->size < 0 ||
3934 		    io->size > pf_ioctl_maxcount ||
3935 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
3936 			error = EINVAL;
3937 			break;
3938 		}
3939 
3940 		totlen = sizeof(struct pfioc_trans_e) * io->size;
3941 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
3942 		    M_TEMP, M_NOWAIT);
3943 		if (ioes == NULL) {
3944 			error = ENOMEM;
3945 			break;
3946 		}
3947 		error = copyin(io->array, ioes, totlen);
3948 		if (error) {
3949 			free(ioes, M_TEMP);
3950 			break;
3951 		}
3952 		PF_RULES_WLOCK();
3953 		/* First makes sure everything will succeed. */
3954 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
3955 			switch (ioe->rs_num) {
3956 #ifdef ALTQ
3957 			case PF_RULESET_ALTQ:
3958 				if (ioe->anchor[0]) {
3959 					PF_RULES_WUNLOCK();
3960 					free(ioes, M_TEMP);
3961 					error = EINVAL;
3962 					goto fail;
3963 				}
3964 				if (!V_altqs_inactive_open || ioe->ticket !=
3965 				    V_ticket_altqs_inactive) {
3966 					PF_RULES_WUNLOCK();
3967 					free(ioes, M_TEMP);
3968 					error = EBUSY;
3969 					goto fail;
3970 				}
3971 				break;
3972 #endif /* ALTQ */
3973 			case PF_RULESET_TABLE:
3974 				rs = pf_find_kruleset(ioe->anchor);
3975 				if (rs == NULL || !rs->topen || ioe->ticket !=
3976 				    rs->tticket) {
3977 					PF_RULES_WUNLOCK();
3978 					free(ioes, M_TEMP);
3979 					error = EBUSY;
3980 					goto fail;
3981 				}
3982 				break;
3983 			default:
3984 				if (ioe->rs_num < 0 || ioe->rs_num >=
3985 				    PF_RULESET_MAX) {
3986 					PF_RULES_WUNLOCK();
3987 					free(ioes, M_TEMP);
3988 					error = EINVAL;
3989 					goto fail;
3990 				}
3991 				rs = pf_find_kruleset(ioe->anchor);
3992 				if (rs == NULL ||
3993 				    !rs->rules[ioe->rs_num].inactive.open ||
3994 				    rs->rules[ioe->rs_num].inactive.ticket !=
3995 				    ioe->ticket) {
3996 					PF_RULES_WUNLOCK();
3997 					free(ioes, M_TEMP);
3998 					error = EBUSY;
3999 					goto fail;
4000 				}
4001 				break;
4002 			}
4003 		}
4004 		/* Now do the commit - no errors should happen here. */
4005 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
4006 			switch (ioe->rs_num) {
4007 #ifdef ALTQ
4008 			case PF_RULESET_ALTQ:
4009 				if ((error = pf_commit_altq(ioe->ticket))) {
4010 					PF_RULES_WUNLOCK();
4011 					free(ioes, M_TEMP);
4012 					goto fail; /* really bad */
4013 				}
4014 				break;
4015 #endif /* ALTQ */
4016 			case PF_RULESET_TABLE:
4017 			    {
4018 				struct pfr_table table;
4019 
4020 				bzero(&table, sizeof(table));
4021 				strlcpy(table.pfrt_anchor, ioe->anchor,
4022 				    sizeof(table.pfrt_anchor));
4023 				if ((error = pfr_ina_commit(&table,
4024 				    ioe->ticket, NULL, NULL, 0))) {
4025 					PF_RULES_WUNLOCK();
4026 					free(ioes, M_TEMP);
4027 					goto fail; /* really bad */
4028 				}
4029 				break;
4030 			    }
4031 			default:
4032 				if ((error = pf_commit_rules(ioe->ticket,
4033 				    ioe->rs_num, ioe->anchor))) {
4034 					PF_RULES_WUNLOCK();
4035 					free(ioes, M_TEMP);
4036 					goto fail; /* really bad */
4037 				}
4038 				break;
4039 			}
4040 		}
4041 		PF_RULES_WUNLOCK();
4042 		free(ioes, M_TEMP);
4043 		break;
4044 	}
4045 
4046 	case DIOCGETSRCNODES: {
4047 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
4048 		struct pf_srchash	*sh;
4049 		struct pf_ksrc_node	*n;
4050 		struct pf_src_node	*p, *pstore;
4051 		uint32_t		 i, nr = 0;
4052 
4053 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4054 				i++, sh++) {
4055 			PF_HASHROW_LOCK(sh);
4056 			LIST_FOREACH(n, &sh->nodes, entry)
4057 				nr++;
4058 			PF_HASHROW_UNLOCK(sh);
4059 		}
4060 
4061 		psn->psn_len = min(psn->psn_len,
4062 		    sizeof(struct pf_src_node) * nr);
4063 
4064 		if (psn->psn_len == 0) {
4065 			psn->psn_len = sizeof(struct pf_src_node) * nr;
4066 			break;
4067 		}
4068 
4069 		nr = 0;
4070 
4071 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
4072 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4073 		    i++, sh++) {
4074 		    PF_HASHROW_LOCK(sh);
4075 		    LIST_FOREACH(n, &sh->nodes, entry) {
4076 
4077 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
4078 				break;
4079 
4080 			pf_src_node_copy(n, p);
4081 
4082 			p++;
4083 			nr++;
4084 		    }
4085 		    PF_HASHROW_UNLOCK(sh);
4086 		}
4087 		error = copyout(pstore, psn->psn_src_nodes,
4088 		    sizeof(struct pf_src_node) * nr);
4089 		if (error) {
4090 			free(pstore, M_TEMP);
4091 			break;
4092 		}
4093 		psn->psn_len = sizeof(struct pf_src_node) * nr;
4094 		free(pstore, M_TEMP);
4095 		break;
4096 	}
4097 
4098 	case DIOCCLRSRCNODES: {
4099 		pf_clear_srcnodes(NULL);
4100 		pf_purge_expired_src_nodes();
4101 		break;
4102 	}
4103 
4104 	case DIOCKILLSRCNODES:
4105 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
4106 		break;
4107 
4108 	case DIOCSETHOSTID: {
4109 		u_int32_t	*hostid = (u_int32_t *)addr;
4110 
4111 		PF_RULES_WLOCK();
4112 		if (*hostid == 0)
4113 			V_pf_status.hostid = arc4random();
4114 		else
4115 			V_pf_status.hostid = *hostid;
4116 		PF_RULES_WUNLOCK();
4117 		break;
4118 	}
4119 
4120 	case DIOCOSFPFLUSH:
4121 		PF_RULES_WLOCK();
4122 		pf_osfp_flush();
4123 		PF_RULES_WUNLOCK();
4124 		break;
4125 
4126 	case DIOCIGETIFACES: {
4127 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
4128 		struct pfi_kif *ifstore;
4129 		size_t bufsiz;
4130 
4131 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
4132 			error = ENODEV;
4133 			break;
4134 		}
4135 
4136 		if (io->pfiio_size < 0 ||
4137 		    io->pfiio_size > pf_ioctl_maxcount ||
4138 		    WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
4139 			error = EINVAL;
4140 			break;
4141 		}
4142 
4143 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
4144 		ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
4145 		    M_TEMP, M_NOWAIT);
4146 		if (ifstore == NULL) {
4147 			error = ENOMEM;
4148 			break;
4149 		}
4150 
4151 		PF_RULES_RLOCK();
4152 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
4153 		PF_RULES_RUNLOCK();
4154 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
4155 		free(ifstore, M_TEMP);
4156 		break;
4157 	}
4158 
4159 	case DIOCSETIFFLAG: {
4160 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
4161 
4162 		PF_RULES_WLOCK();
4163 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
4164 		PF_RULES_WUNLOCK();
4165 		break;
4166 	}
4167 
4168 	case DIOCCLRIFFLAG: {
4169 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
4170 
4171 		PF_RULES_WLOCK();
4172 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
4173 		PF_RULES_WUNLOCK();
4174 		break;
4175 	}
4176 
4177 	default:
4178 		error = ENODEV;
4179 		break;
4180 	}
4181 fail:
4182 	if (sx_xlocked(&pf_ioctl_lock))
4183 		sx_xunlock(&pf_ioctl_lock);
4184 	CURVNET_RESTORE();
4185 
4186 	return (error);
4187 }
4188 
4189 void
4190 pfsync_state_export(struct pfsync_state *sp, struct pf_state *st)
4191 {
4192 	bzero(sp, sizeof(struct pfsync_state));
4193 
4194 	/* copy from state key */
4195 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
4196 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
4197 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
4198 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
4199 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
4200 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
4201 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
4202 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
4203 	sp->proto = st->key[PF_SK_WIRE]->proto;
4204 	sp->af = st->key[PF_SK_WIRE]->af;
4205 
4206 	/* copy from state */
4207 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
4208 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
4209 	sp->creation = htonl(time_uptime - st->creation);
4210 	sp->expire = pf_state_expires(st);
4211 	if (sp->expire <= time_uptime)
4212 		sp->expire = htonl(0);
4213 	else
4214 		sp->expire = htonl(sp->expire - time_uptime);
4215 
4216 	sp->direction = st->direction;
4217 	sp->log = st->log;
4218 	sp->timeout = st->timeout;
4219 	sp->state_flags = st->state_flags;
4220 	if (st->src_node)
4221 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
4222 	if (st->nat_src_node)
4223 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
4224 
4225 	sp->id = st->id;
4226 	sp->creatorid = st->creatorid;
4227 	pf_state_peer_hton(&st->src, &sp->src);
4228 	pf_state_peer_hton(&st->dst, &sp->dst);
4229 
4230 	if (st->rule.ptr == NULL)
4231 		sp->rule = htonl(-1);
4232 	else
4233 		sp->rule = htonl(st->rule.ptr->nr);
4234 	if (st->anchor.ptr == NULL)
4235 		sp->anchor = htonl(-1);
4236 	else
4237 		sp->anchor = htonl(st->anchor.ptr->nr);
4238 	if (st->nat_rule.ptr == NULL)
4239 		sp->nat_rule = htonl(-1);
4240 	else
4241 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
4242 
4243 	pf_state_counter_hton(counter_u64_fetch(st->packets[0]),
4244 	    sp->packets[0]);
4245 	pf_state_counter_hton(counter_u64_fetch(st->packets[1]),
4246 	    sp->packets[1]);
4247 	pf_state_counter_hton(counter_u64_fetch(st->bytes[0]), sp->bytes[0]);
4248 	pf_state_counter_hton(counter_u64_fetch(st->bytes[1]), sp->bytes[1]);
4249 
4250 }
4251 
4252 static void
4253 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
4254 {
4255 	struct pfr_ktable *kt;
4256 
4257 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
4258 
4259 	kt = aw->p.tbl;
4260 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
4261 		kt = kt->pfrkt_root;
4262 	aw->p.tbl = NULL;
4263 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
4264 		kt->pfrkt_cnt : -1;
4265 }
4266 
4267 /*
4268  * XXX - Check for version missmatch!!!
4269  */
4270 static void
4271 pf_clear_states(void)
4272 {
4273 	struct pf_state	*s;
4274 	u_int i;
4275 
4276 	for (i = 0; i <= pf_hashmask; i++) {
4277 		struct pf_idhash *ih = &V_pf_idhash[i];
4278 relock:
4279 		PF_HASHROW_LOCK(ih);
4280 		LIST_FOREACH(s, &ih->states, entry) {
4281 			s->timeout = PFTM_PURGE;
4282 			/* Don't send out individual delete messages. */
4283 			s->state_flags |= PFSTATE_NOSYNC;
4284 			pf_unlink_state(s, PF_ENTER_LOCKED);
4285 			goto relock;
4286 		}
4287 		PF_HASHROW_UNLOCK(ih);
4288 	}
4289 }
4290 
4291 static int
4292 pf_clear_tables(void)
4293 {
4294 	struct pfioc_table io;
4295 	int error;
4296 
4297 	bzero(&io, sizeof(io));
4298 
4299 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
4300 	    io.pfrio_flags);
4301 
4302 	return (error);
4303 }
4304 
4305 static void
4306 pf_clear_srcnodes(struct pf_ksrc_node *n)
4307 {
4308 	struct pf_state *s;
4309 	int i;
4310 
4311 	for (i = 0; i <= pf_hashmask; i++) {
4312 		struct pf_idhash *ih = &V_pf_idhash[i];
4313 
4314 		PF_HASHROW_LOCK(ih);
4315 		LIST_FOREACH(s, &ih->states, entry) {
4316 			if (n == NULL || n == s->src_node)
4317 				s->src_node = NULL;
4318 			if (n == NULL || n == s->nat_src_node)
4319 				s->nat_src_node = NULL;
4320 		}
4321 		PF_HASHROW_UNLOCK(ih);
4322 	}
4323 
4324 	if (n == NULL) {
4325 		struct pf_srchash *sh;
4326 
4327 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4328 		    i++, sh++) {
4329 			PF_HASHROW_LOCK(sh);
4330 			LIST_FOREACH(n, &sh->nodes, entry) {
4331 				n->expire = 1;
4332 				n->states = 0;
4333 			}
4334 			PF_HASHROW_UNLOCK(sh);
4335 		}
4336 	} else {
4337 		/* XXX: hash slot should already be locked here. */
4338 		n->expire = 1;
4339 		n->states = 0;
4340 	}
4341 }
4342 
4343 static void
4344 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
4345 {
4346 	struct pf_ksrc_node_list	 kill;
4347 
4348 	LIST_INIT(&kill);
4349 	for (int i = 0; i <= pf_srchashmask; i++) {
4350 		struct pf_srchash *sh = &V_pf_srchash[i];
4351 		struct pf_ksrc_node *sn, *tmp;
4352 
4353 		PF_HASHROW_LOCK(sh);
4354 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
4355 			if (PF_MATCHA(psnk->psnk_src.neg,
4356 			      &psnk->psnk_src.addr.v.a.addr,
4357 			      &psnk->psnk_src.addr.v.a.mask,
4358 			      &sn->addr, sn->af) &&
4359 			    PF_MATCHA(psnk->psnk_dst.neg,
4360 			      &psnk->psnk_dst.addr.v.a.addr,
4361 			      &psnk->psnk_dst.addr.v.a.mask,
4362 			      &sn->raddr, sn->af)) {
4363 				pf_unlink_src_node(sn);
4364 				LIST_INSERT_HEAD(&kill, sn, entry);
4365 				sn->expire = 1;
4366 			}
4367 		PF_HASHROW_UNLOCK(sh);
4368 	}
4369 
4370 	for (int i = 0; i <= pf_hashmask; i++) {
4371 		struct pf_idhash *ih = &V_pf_idhash[i];
4372 		struct pf_state *s;
4373 
4374 		PF_HASHROW_LOCK(ih);
4375 		LIST_FOREACH(s, &ih->states, entry) {
4376 			if (s->src_node && s->src_node->expire == 1)
4377 				s->src_node = NULL;
4378 			if (s->nat_src_node && s->nat_src_node->expire == 1)
4379 				s->nat_src_node = NULL;
4380 		}
4381 		PF_HASHROW_UNLOCK(ih);
4382 	}
4383 
4384 	psnk->psnk_killed = pf_free_src_nodes(&kill);
4385 }
4386 
4387 /*
4388  * XXX - Check for version missmatch!!!
4389  */
4390 
4391 /*
4392  * Duplicate pfctl -Fa operation to get rid of as much as we can.
4393  */
4394 static int
4395 shutdown_pf(void)
4396 {
4397 	int error = 0;
4398 	u_int32_t t[5];
4399 	char nn = '\0';
4400 
4401 	do {
4402 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
4403 		    != 0) {
4404 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
4405 			break;
4406 		}
4407 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
4408 		    != 0) {
4409 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
4410 			break;		/* XXX: rollback? */
4411 		}
4412 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
4413 		    != 0) {
4414 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
4415 			break;		/* XXX: rollback? */
4416 		}
4417 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
4418 		    != 0) {
4419 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
4420 			break;		/* XXX: rollback? */
4421 		}
4422 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
4423 		    != 0) {
4424 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
4425 			break;		/* XXX: rollback? */
4426 		}
4427 
4428 		/* XXX: these should always succeed here */
4429 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
4430 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
4431 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
4432 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
4433 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
4434 
4435 		if ((error = pf_clear_tables()) != 0)
4436 			break;
4437 
4438 #ifdef ALTQ
4439 		if ((error = pf_begin_altq(&t[0])) != 0) {
4440 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
4441 			break;
4442 		}
4443 		pf_commit_altq(t[0]);
4444 #endif
4445 
4446 		pf_clear_states();
4447 
4448 		pf_clear_srcnodes(NULL);
4449 
4450 		/* status does not use malloced mem so no need to cleanup */
4451 		/* fingerprints and interfaces have their own cleanup code */
4452 	} while(0);
4453 
4454 	return (error);
4455 }
4456 
4457 static pfil_return_t
4458 pf_check_return(int chk, struct mbuf **m)
4459 {
4460 
4461 	switch (chk) {
4462 	case PF_PASS:
4463 		if (*m == NULL)
4464 			return (PFIL_CONSUMED);
4465 		else
4466 			return (PFIL_PASS);
4467 		break;
4468 	default:
4469 		if (*m != NULL) {
4470 			m_freem(*m);
4471 			*m = NULL;
4472 		}
4473 		return (PFIL_DROPPED);
4474 	}
4475 }
4476 
4477 #ifdef INET
4478 static pfil_return_t
4479 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
4480     void *ruleset __unused, struct inpcb *inp)
4481 {
4482 	int chk;
4483 
4484 	chk = pf_test(PF_IN, flags, ifp, m, inp);
4485 
4486 	return (pf_check_return(chk, m));
4487 }
4488 
4489 static pfil_return_t
4490 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
4491     void *ruleset __unused,  struct inpcb *inp)
4492 {
4493 	int chk;
4494 
4495 	chk = pf_test(PF_OUT, flags, ifp, m, inp);
4496 
4497 	return (pf_check_return(chk, m));
4498 }
4499 #endif
4500 
4501 #ifdef INET6
4502 static pfil_return_t
4503 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
4504     void *ruleset __unused,  struct inpcb *inp)
4505 {
4506 	int chk;
4507 
4508 	/*
4509 	 * In case of loopback traffic IPv6 uses the real interface in
4510 	 * order to support scoped addresses. In order to support stateful
4511 	 * filtering we have change this to lo0 as it is the case in IPv4.
4512 	 */
4513 	CURVNET_SET(ifp->if_vnet);
4514 	chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
4515 	CURVNET_RESTORE();
4516 
4517 	return (pf_check_return(chk, m));
4518 }
4519 
4520 static pfil_return_t
4521 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
4522     void *ruleset __unused,  struct inpcb *inp)
4523 {
4524 	int chk;
4525 
4526 	CURVNET_SET(ifp->if_vnet);
4527 	chk = pf_test6(PF_OUT, flags, ifp, m, inp);
4528 	CURVNET_RESTORE();
4529 
4530 	return (pf_check_return(chk, m));
4531 }
4532 #endif /* INET6 */
4533 
4534 #ifdef INET
4535 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
4536 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
4537 #define	V_pf_ip4_in_hook	VNET(pf_ip4_in_hook)
4538 #define	V_pf_ip4_out_hook	VNET(pf_ip4_out_hook)
4539 #endif
4540 #ifdef INET6
4541 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
4542 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
4543 #define	V_pf_ip6_in_hook	VNET(pf_ip6_in_hook)
4544 #define	V_pf_ip6_out_hook	VNET(pf_ip6_out_hook)
4545 #endif
4546 
4547 static void
4548 hook_pf(void)
4549 {
4550 	struct pfil_hook_args pha;
4551 	struct pfil_link_args pla;
4552 	int ret;
4553 
4554 	if (V_pf_pfil_hooked)
4555 		return;
4556 
4557 	pha.pa_version = PFIL_VERSION;
4558 	pha.pa_modname = "pf";
4559 	pha.pa_ruleset = NULL;
4560 
4561 	pla.pa_version = PFIL_VERSION;
4562 
4563 #ifdef INET
4564 	pha.pa_type = PFIL_TYPE_IP4;
4565 	pha.pa_func = pf_check_in;
4566 	pha.pa_flags = PFIL_IN;
4567 	pha.pa_rulname = "default-in";
4568 	V_pf_ip4_in_hook = pfil_add_hook(&pha);
4569 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
4570 	pla.pa_head = V_inet_pfil_head;
4571 	pla.pa_hook = V_pf_ip4_in_hook;
4572 	ret = pfil_link(&pla);
4573 	MPASS(ret == 0);
4574 	pha.pa_func = pf_check_out;
4575 	pha.pa_flags = PFIL_OUT;
4576 	pha.pa_rulname = "default-out";
4577 	V_pf_ip4_out_hook = pfil_add_hook(&pha);
4578 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
4579 	pla.pa_head = V_inet_pfil_head;
4580 	pla.pa_hook = V_pf_ip4_out_hook;
4581 	ret = pfil_link(&pla);
4582 	MPASS(ret == 0);
4583 #endif
4584 #ifdef INET6
4585 	pha.pa_type = PFIL_TYPE_IP6;
4586 	pha.pa_func = pf_check6_in;
4587 	pha.pa_flags = PFIL_IN;
4588 	pha.pa_rulname = "default-in6";
4589 	V_pf_ip6_in_hook = pfil_add_hook(&pha);
4590 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
4591 	pla.pa_head = V_inet6_pfil_head;
4592 	pla.pa_hook = V_pf_ip6_in_hook;
4593 	ret = pfil_link(&pla);
4594 	MPASS(ret == 0);
4595 	pha.pa_func = pf_check6_out;
4596 	pha.pa_rulname = "default-out6";
4597 	pha.pa_flags = PFIL_OUT;
4598 	V_pf_ip6_out_hook = pfil_add_hook(&pha);
4599 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
4600 	pla.pa_head = V_inet6_pfil_head;
4601 	pla.pa_hook = V_pf_ip6_out_hook;
4602 	ret = pfil_link(&pla);
4603 	MPASS(ret == 0);
4604 #endif
4605 
4606 	V_pf_pfil_hooked = 1;
4607 }
4608 
4609 static void
4610 dehook_pf(void)
4611 {
4612 
4613 	if (V_pf_pfil_hooked == 0)
4614 		return;
4615 
4616 #ifdef INET
4617 	pfil_remove_hook(V_pf_ip4_in_hook);
4618 	pfil_remove_hook(V_pf_ip4_out_hook);
4619 #endif
4620 #ifdef INET6
4621 	pfil_remove_hook(V_pf_ip6_in_hook);
4622 	pfil_remove_hook(V_pf_ip6_out_hook);
4623 #endif
4624 
4625 	V_pf_pfil_hooked = 0;
4626 }
4627 
4628 static void
4629 pf_load_vnet(void)
4630 {
4631 	V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
4632 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
4633 
4634 	pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
4635 	    PF_RULE_TAG_HASH_SIZE_DEFAULT);
4636 #ifdef ALTQ
4637 	pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
4638 	    PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
4639 #endif
4640 
4641 	pfattach_vnet();
4642 	V_pf_vnet_active = 1;
4643 }
4644 
4645 static int
4646 pf_load(void)
4647 {
4648 	int error;
4649 
4650 	rm_init(&pf_rules_lock, "pf rulesets");
4651 	sx_init(&pf_ioctl_lock, "pf ioctl");
4652 	sx_init(&pf_end_lock, "pf end thread");
4653 
4654 	pf_mtag_initialize();
4655 
4656 	pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
4657 	if (pf_dev == NULL)
4658 		return (ENOMEM);
4659 
4660 	pf_end_threads = 0;
4661 	error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
4662 	if (error != 0)
4663 		return (error);
4664 
4665 	pfi_initialize();
4666 
4667 	return (0);
4668 }
4669 
4670 static void
4671 pf_unload_vnet(void)
4672 {
4673 	int ret;
4674 
4675 	V_pf_vnet_active = 0;
4676 	V_pf_status.running = 0;
4677 	dehook_pf();
4678 
4679 	PF_RULES_WLOCK();
4680 	shutdown_pf();
4681 	PF_RULES_WUNLOCK();
4682 
4683 	ret = swi_remove(V_pf_swi_cookie);
4684 	MPASS(ret == 0);
4685 	ret = intr_event_destroy(V_pf_swi_ie);
4686 	MPASS(ret == 0);
4687 
4688 	pf_unload_vnet_purge();
4689 
4690 	pf_normalize_cleanup();
4691 	PF_RULES_WLOCK();
4692 	pfi_cleanup_vnet();
4693 	PF_RULES_WUNLOCK();
4694 	pfr_cleanup();
4695 	pf_osfp_flush();
4696 	pf_cleanup();
4697 	if (IS_DEFAULT_VNET(curvnet))
4698 		pf_mtag_cleanup();
4699 
4700 	pf_cleanup_tagset(&V_pf_tags);
4701 #ifdef ALTQ
4702 	pf_cleanup_tagset(&V_pf_qids);
4703 #endif
4704 	uma_zdestroy(V_pf_tag_z);
4705 
4706 	/* Free counters last as we updated them during shutdown. */
4707 	counter_u64_free(V_pf_default_rule.evaluations);
4708 	for (int i = 0; i < 2; i++) {
4709 		counter_u64_free(V_pf_default_rule.packets[i]);
4710 		counter_u64_free(V_pf_default_rule.bytes[i]);
4711 	}
4712 	counter_u64_free(V_pf_default_rule.states_cur);
4713 	counter_u64_free(V_pf_default_rule.states_tot);
4714 	counter_u64_free(V_pf_default_rule.src_nodes);
4715 
4716 	for (int i = 0; i < PFRES_MAX; i++)
4717 		counter_u64_free(V_pf_status.counters[i]);
4718 	for (int i = 0; i < LCNT_MAX; i++)
4719 		counter_u64_free(V_pf_status.lcounters[i]);
4720 	for (int i = 0; i < FCNT_MAX; i++)
4721 		counter_u64_free(V_pf_status.fcounters[i]);
4722 	for (int i = 0; i < SCNT_MAX; i++)
4723 		counter_u64_free(V_pf_status.scounters[i]);
4724 }
4725 
4726 static void
4727 pf_unload(void)
4728 {
4729 
4730 	sx_xlock(&pf_end_lock);
4731 	pf_end_threads = 1;
4732 	while (pf_end_threads < 2) {
4733 		wakeup_one(pf_purge_thread);
4734 		sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
4735 	}
4736 	sx_xunlock(&pf_end_lock);
4737 
4738 	if (pf_dev != NULL)
4739 		destroy_dev(pf_dev);
4740 
4741 	pfi_cleanup();
4742 
4743 	rm_destroy(&pf_rules_lock);
4744 	sx_destroy(&pf_ioctl_lock);
4745 	sx_destroy(&pf_end_lock);
4746 }
4747 
4748 static void
4749 vnet_pf_init(void *unused __unused)
4750 {
4751 
4752 	pf_load_vnet();
4753 }
4754 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
4755     vnet_pf_init, NULL);
4756 
4757 static void
4758 vnet_pf_uninit(const void *unused __unused)
4759 {
4760 
4761 	pf_unload_vnet();
4762 }
4763 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
4764 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
4765     vnet_pf_uninit, NULL);
4766 
4767 static int
4768 pf_modevent(module_t mod, int type, void *data)
4769 {
4770 	int error = 0;
4771 
4772 	switch(type) {
4773 	case MOD_LOAD:
4774 		error = pf_load();
4775 		break;
4776 	case MOD_UNLOAD:
4777 		/* Handled in SYSUNINIT(pf_unload) to ensure it's done after
4778 		 * the vnet_pf_uninit()s */
4779 		break;
4780 	default:
4781 		error = EINVAL;
4782 		break;
4783 	}
4784 
4785 	return (error);
4786 }
4787 
4788 static moduledata_t pf_mod = {
4789 	"pf",
4790 	pf_modevent,
4791 	0
4792 };
4793 
4794 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
4795 MODULE_VERSION(pf, PF_MODVER);
4796