xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision b1c5f60ce87cc2f179dfb81de507d9b7bf59564c)
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/nv.h>
65 #include <sys/proc.h>
66 #include <sys/sdt.h>
67 #include <sys/smp.h>
68 #include <sys/socket.h>
69 #include <sys/sysctl.h>
70 #include <sys/md5.h>
71 #include <sys/ucred.h>
72 
73 #include <net/if.h>
74 #include <net/if_var.h>
75 #include <net/vnet.h>
76 #include <net/route.h>
77 #include <net/pfil.h>
78 #include <net/pfvar.h>
79 #include <net/if_pfsync.h>
80 #include <net/if_pflog.h>
81 
82 #include <netinet/in.h>
83 #include <netinet/ip.h>
84 #include <netinet/ip_var.h>
85 #include <netinet6/ip6_var.h>
86 #include <netinet/ip_icmp.h>
87 #include <netpfil/pf/pf_nv.h>
88 
89 #ifdef INET6
90 #include <netinet/ip6.h>
91 #endif /* INET6 */
92 
93 #ifdef ALTQ
94 #include <net/altq/altq.h>
95 #endif
96 
97 SDT_PROBE_DEFINE3(pf, ioctl, ioctl, error, "int", "int", "int");
98 SDT_PROBE_DEFINE3(pf, ioctl, function, error, "char *", "int", "int");
99 SDT_PROBE_DEFINE2(pf, ioctl, addrule, error, "int", "int");
100 SDT_PROBE_DEFINE2(pf, ioctl, nvchk, error, "int", "int");
101 
102 static struct pf_kpool	*pf_get_kpool(const char *, u_int32_t, u_int8_t,
103 			    u_int32_t, u_int8_t, u_int8_t, u_int8_t);
104 
105 static void		 pf_mv_kpool(struct pf_kpalist *, struct pf_kpalist *);
106 static void		 pf_empty_kpool(struct pf_kpalist *);
107 static int		 pfioctl(struct cdev *, u_long, caddr_t, int,
108 			    struct thread *);
109 static int		 pf_begin_eth(uint32_t *, const char *);
110 static void		 pf_rollback_eth_cb(struct epoch_context *);
111 static int		 pf_rollback_eth(uint32_t, const char *);
112 static int		 pf_commit_eth(uint32_t, const char *);
113 static void		 pf_free_eth_rule(struct pf_keth_rule *);
114 #ifdef ALTQ
115 static int		 pf_begin_altq(u_int32_t *);
116 static int		 pf_rollback_altq(u_int32_t);
117 static int		 pf_commit_altq(u_int32_t);
118 static int		 pf_enable_altq(struct pf_altq *);
119 static int		 pf_disable_altq(struct pf_altq *);
120 static uint16_t		 pf_qname2qid(const char *);
121 static void		 pf_qid_unref(uint16_t);
122 #endif /* ALTQ */
123 static int		 pf_begin_rules(u_int32_t *, int, const char *);
124 static int		 pf_rollback_rules(u_int32_t, int, char *);
125 static int		 pf_setup_pfsync_matching(struct pf_kruleset *);
126 static void		 pf_hash_rule_rolling(MD5_CTX *, struct pf_krule *);
127 static void		 pf_hash_rule(struct pf_krule *);
128 static void		 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
129 static int		 pf_commit_rules(u_int32_t, int, char *);
130 static int		 pf_addr_setup(struct pf_kruleset *,
131 			    struct pf_addr_wrap *, sa_family_t);
132 static void		 pf_addr_copyout(struct pf_addr_wrap *);
133 static void		 pf_src_node_copy(const struct pf_ksrc_node *,
134 			    struct pf_src_node *);
135 #ifdef ALTQ
136 static int		 pf_export_kaltq(struct pf_altq *,
137 			    struct pfioc_altq_v1 *, size_t);
138 static int		 pf_import_kaltq(struct pfioc_altq_v1 *,
139 			    struct pf_altq *, size_t);
140 #endif /* ALTQ */
141 
142 VNET_DEFINE(struct pf_krule,	pf_default_rule);
143 
144 static __inline int             pf_krule_compare(struct pf_krule *,
145 				    struct pf_krule *);
146 
147 RB_GENERATE(pf_krule_global, pf_krule, entry_global, pf_krule_compare);
148 
149 #ifdef ALTQ
150 VNET_DEFINE_STATIC(int,		pf_altq_running);
151 #define	V_pf_altq_running	VNET(pf_altq_running)
152 #endif
153 
154 #define	TAGID_MAX	 50000
155 struct pf_tagname {
156 	TAILQ_ENTRY(pf_tagname)	namehash_entries;
157 	TAILQ_ENTRY(pf_tagname)	taghash_entries;
158 	char			name[PF_TAG_NAME_SIZE];
159 	uint16_t		tag;
160 	int			ref;
161 };
162 
163 struct pf_tagset {
164 	TAILQ_HEAD(, pf_tagname)	*namehash;
165 	TAILQ_HEAD(, pf_tagname)	*taghash;
166 	unsigned int			 mask;
167 	uint32_t			 seed;
168 	BITSET_DEFINE(, TAGID_MAX)	 avail;
169 };
170 
171 VNET_DEFINE(struct pf_tagset, pf_tags);
172 #define	V_pf_tags	VNET(pf_tags)
173 static unsigned int	pf_rule_tag_hashsize;
174 #define	PF_RULE_TAG_HASH_SIZE_DEFAULT	128
175 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN,
176     &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT,
177     "Size of pf(4) rule tag hashtable");
178 
179 #ifdef ALTQ
180 VNET_DEFINE(struct pf_tagset, pf_qids);
181 #define	V_pf_qids	VNET(pf_qids)
182 static unsigned int	pf_queue_tag_hashsize;
183 #define	PF_QUEUE_TAG_HASH_SIZE_DEFAULT	128
184 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN,
185     &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT,
186     "Size of pf(4) queue tag hashtable");
187 #endif
188 VNET_DEFINE(uma_zone_t,	 pf_tag_z);
189 #define	V_pf_tag_z		 VNET(pf_tag_z)
190 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
191 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
192 
193 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
194 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
195 #endif
196 
197 static void		 pf_init_tagset(struct pf_tagset *, unsigned int *,
198 			    unsigned int);
199 static void		 pf_cleanup_tagset(struct pf_tagset *);
200 static uint16_t		 tagname2hashindex(const struct pf_tagset *, const char *);
201 static uint16_t		 tag2hashindex(const struct pf_tagset *, uint16_t);
202 static u_int16_t	 tagname2tag(struct pf_tagset *, const char *);
203 static u_int16_t	 pf_tagname2tag(const char *);
204 static void		 tag_unref(struct pf_tagset *, u_int16_t);
205 
206 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
207 
208 struct cdev *pf_dev;
209 
210 /*
211  * XXX - These are new and need to be checked when moveing to a new version
212  */
213 static void		 pf_clear_all_states(void);
214 static unsigned int	 pf_clear_states(const struct pf_kstate_kill *);
215 static void		 pf_killstates(struct pf_kstate_kill *,
216 			    unsigned int *);
217 static int		 pf_killstates_row(struct pf_kstate_kill *,
218 			    struct pf_idhash *);
219 static int		 pf_killstates_nv(struct pfioc_nv *);
220 static int		 pf_clearstates_nv(struct pfioc_nv *);
221 static int		 pf_getstate(struct pfioc_nv *);
222 static int		 pf_getstatus(struct pfioc_nv *);
223 static int		 pf_clear_tables(void);
224 static void		 pf_clear_srcnodes(struct pf_ksrc_node *);
225 static void		 pf_kill_srcnodes(struct pfioc_src_node_kill *);
226 static int		 pf_keepcounters(struct pfioc_nv *);
227 static void		 pf_tbladdr_copyout(struct pf_addr_wrap *);
228 
229 /*
230  * Wrapper functions for pfil(9) hooks
231  */
232 static pfil_return_t pf_eth_check_in(struct mbuf **m, struct ifnet *ifp,
233     int flags, void *ruleset __unused, struct inpcb *inp);
234 static pfil_return_t pf_eth_check_out(struct mbuf **m, struct ifnet *ifp,
235     int flags, void *ruleset __unused, struct inpcb *inp);
236 #ifdef INET
237 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp,
238     int flags, void *ruleset __unused, struct inpcb *inp);
239 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp,
240     int flags, void *ruleset __unused, struct inpcb *inp);
241 #endif
242 #ifdef INET6
243 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp,
244     int flags, void *ruleset __unused, struct inpcb *inp);
245 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp,
246     int flags, void *ruleset __unused, struct inpcb *inp);
247 #endif
248 
249 static void		hook_pf_eth(void);
250 static void		hook_pf(void);
251 static void		dehook_pf_eth(void);
252 static void		dehook_pf(void);
253 static int		shutdown_pf(void);
254 static int		pf_load(void);
255 static void		pf_unload(void);
256 
257 static struct cdevsw pf_cdevsw = {
258 	.d_ioctl =	pfioctl,
259 	.d_name =	PF_NAME,
260 	.d_version =	D_VERSION,
261 };
262 
263 volatile VNET_DEFINE_STATIC(int, pf_pfil_hooked);
264 #define V_pf_pfil_hooked	VNET(pf_pfil_hooked)
265 volatile VNET_DEFINE_STATIC(int, pf_pfil_eth_hooked);
266 #define V_pf_pfil_eth_hooked	VNET(pf_pfil_eth_hooked)
267 
268 /*
269  * We need a flag that is neither hooked nor running to know when
270  * the VNET is "valid".  We primarily need this to control (global)
271  * external event, e.g., eventhandlers.
272  */
273 VNET_DEFINE(int, pf_vnet_active);
274 #define V_pf_vnet_active	VNET(pf_vnet_active)
275 
276 int pf_end_threads;
277 struct proc *pf_purge_proc;
278 
279 struct rmlock			pf_rules_lock;
280 struct sx			pf_ioctl_lock;
281 struct sx			pf_end_lock;
282 
283 /* pfsync */
284 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
285 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
286 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
287 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
288 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
289 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
290 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
291 
292 /* pflog */
293 pflog_packet_t			*pflog_packet_ptr = NULL;
294 
295 /*
296  * Copy a user-provided string, returning an error if truncation would occur.
297  * Avoid scanning past "sz" bytes in the source string since there's no
298  * guarantee that it's nul-terminated.
299  */
300 static int
301 pf_user_strcpy(char *dst, const char *src, size_t sz)
302 {
303 	if (strnlen(src, sz) == sz)
304 		return (EINVAL);
305 	(void)strlcpy(dst, src, sz);
306 	return (0);
307 }
308 
309 static void
310 pfattach_vnet(void)
311 {
312 	u_int32_t *my_timeout = V_pf_default_rule.timeout;
313 
314 	pf_initialize();
315 	pfr_initialize();
316 	pfi_initialize_vnet();
317 	pf_normalize_init();
318 	pf_syncookies_init();
319 
320 	V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
321 	V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
322 
323 	RB_INIT(&V_pf_anchors);
324 	pf_init_kruleset(&pf_main_ruleset);
325 
326 	pf_init_keth(V_pf_keth);
327 
328 	/* default rule should never be garbage collected */
329 	V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
330 #ifdef PF_DEFAULT_TO_DROP
331 	V_pf_default_rule.action = PF_DROP;
332 #else
333 	V_pf_default_rule.action = PF_PASS;
334 #endif
335 	V_pf_default_rule.nr = -1;
336 	V_pf_default_rule.rtableid = -1;
337 
338 	pf_counter_u64_init(&V_pf_default_rule.evaluations, M_WAITOK);
339 	for (int i = 0; i < 2; i++) {
340 		pf_counter_u64_init(&V_pf_default_rule.packets[i], M_WAITOK);
341 		pf_counter_u64_init(&V_pf_default_rule.bytes[i], M_WAITOK);
342 	}
343 	V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
344 	V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
345 	V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
346 
347 #ifdef PF_WANT_32_TO_64_COUNTER
348 	V_pf_kifmarker = malloc(sizeof(*V_pf_kifmarker), PFI_MTYPE, M_WAITOK | M_ZERO);
349 	V_pf_rulemarker = malloc(sizeof(*V_pf_rulemarker), M_PFRULE, M_WAITOK | M_ZERO);
350 	PF_RULES_WLOCK();
351 	LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist);
352 	LIST_INSERT_HEAD(&V_pf_allrulelist, &V_pf_default_rule, allrulelist);
353 	V_pf_allrulecount++;
354 	LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist);
355 	PF_RULES_WUNLOCK();
356 #endif
357 
358 	/* initialize default timeouts */
359 	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
360 	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
361 	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
362 	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
363 	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
364 	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
365 	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
366 	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
367 	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
368 	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
369 	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
370 	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
371 	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
372 	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
373 	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
374 	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
375 	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
376 	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
377 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
378 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
379 
380 	bzero(&V_pf_status, sizeof(V_pf_status));
381 	V_pf_status.debug = PF_DEBUG_URGENT;
382 
383 	V_pf_pfil_hooked = 0;
384 	V_pf_pfil_eth_hooked = 0;
385 
386 	/* XXX do our best to avoid a conflict */
387 	V_pf_status.hostid = arc4random();
388 
389 	for (int i = 0; i < PFRES_MAX; i++)
390 		V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
391 	for (int i = 0; i < KLCNT_MAX; i++)
392 		V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
393 	for (int i = 0; i < FCNT_MAX; i++)
394 		pf_counter_u64_init(&V_pf_status.fcounters[i], M_WAITOK);
395 	for (int i = 0; i < SCNT_MAX; i++)
396 		V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
397 
398 	if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET,
399 	    INTR_MPSAFE, &V_pf_swi_cookie) != 0)
400 		/* XXXGL: leaked all above. */
401 		return;
402 }
403 
404 static struct pf_kpool *
405 pf_get_kpool(const char *anchor, u_int32_t ticket, u_int8_t rule_action,
406     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
407     u_int8_t check_ticket)
408 {
409 	struct pf_kruleset	*ruleset;
410 	struct pf_krule		*rule;
411 	int			 rs_num;
412 
413 	ruleset = pf_find_kruleset(anchor);
414 	if (ruleset == NULL)
415 		return (NULL);
416 	rs_num = pf_get_ruleset_number(rule_action);
417 	if (rs_num >= PF_RULESET_MAX)
418 		return (NULL);
419 	if (active) {
420 		if (check_ticket && ticket !=
421 		    ruleset->rules[rs_num].active.ticket)
422 			return (NULL);
423 		if (r_last)
424 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
425 			    pf_krulequeue);
426 		else
427 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
428 	} else {
429 		if (check_ticket && ticket !=
430 		    ruleset->rules[rs_num].inactive.ticket)
431 			return (NULL);
432 		if (r_last)
433 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
434 			    pf_krulequeue);
435 		else
436 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
437 	}
438 	if (!r_last) {
439 		while ((rule != NULL) && (rule->nr != rule_number))
440 			rule = TAILQ_NEXT(rule, entries);
441 	}
442 	if (rule == NULL)
443 		return (NULL);
444 
445 	return (&rule->rpool);
446 }
447 
448 static void
449 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb)
450 {
451 	struct pf_kpooladdr	*mv_pool_pa;
452 
453 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
454 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
455 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
456 	}
457 }
458 
459 static void
460 pf_empty_kpool(struct pf_kpalist *poola)
461 {
462 	struct pf_kpooladdr *pa;
463 
464 	while ((pa = TAILQ_FIRST(poola)) != NULL) {
465 		switch (pa->addr.type) {
466 		case PF_ADDR_DYNIFTL:
467 			pfi_dynaddr_remove(pa->addr.p.dyn);
468 			break;
469 		case PF_ADDR_TABLE:
470 			/* XXX: this could be unfinished pooladdr on pabuf */
471 			if (pa->addr.p.tbl != NULL)
472 				pfr_detach_table(pa->addr.p.tbl);
473 			break;
474 		}
475 		if (pa->kif)
476 			pfi_kkif_unref(pa->kif);
477 		TAILQ_REMOVE(poola, pa, entries);
478 		free(pa, M_PFRULE);
479 	}
480 }
481 
482 static void
483 pf_unlink_rule_locked(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
484 {
485 
486 	PF_RULES_WASSERT();
487 	PF_UNLNKDRULES_ASSERT();
488 
489 	TAILQ_REMOVE(rulequeue, rule, entries);
490 
491 	rule->rule_ref |= PFRULE_REFS;
492 	TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
493 }
494 
495 static void
496 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
497 {
498 
499 	PF_RULES_WASSERT();
500 
501 	PF_UNLNKDRULES_LOCK();
502 	pf_unlink_rule_locked(rulequeue, rule);
503 	PF_UNLNKDRULES_UNLOCK();
504 }
505 
506 static void
507 pf_free_eth_rule(struct pf_keth_rule *rule)
508 {
509 	PF_RULES_WASSERT();
510 
511 	if (rule == NULL)
512 		return;
513 
514 	if (rule->tag)
515 		tag_unref(&V_pf_tags, rule->tag);
516 #ifdef ALTQ
517 	pf_qid_unref(rule->qid);
518 #endif
519 
520 	if (rule->kif)
521 		pfi_kkif_unref(rule->kif);
522 
523 	counter_u64_free(rule->evaluations);
524 	for (int i = 0; i < 2; i++) {
525 		counter_u64_free(rule->packets[i]);
526 		counter_u64_free(rule->bytes[i]);
527 	}
528 	pf_keth_anchor_remove(rule);
529 
530 	free(rule, M_PFRULE);
531 }
532 
533 void
534 pf_free_rule(struct pf_krule *rule)
535 {
536 
537 	PF_RULES_WASSERT();
538 	PF_CONFIG_ASSERT();
539 
540 	if (rule->tag)
541 		tag_unref(&V_pf_tags, rule->tag);
542 	if (rule->match_tag)
543 		tag_unref(&V_pf_tags, rule->match_tag);
544 #ifdef ALTQ
545 	if (rule->pqid != rule->qid)
546 		pf_qid_unref(rule->pqid);
547 	pf_qid_unref(rule->qid);
548 #endif
549 	switch (rule->src.addr.type) {
550 	case PF_ADDR_DYNIFTL:
551 		pfi_dynaddr_remove(rule->src.addr.p.dyn);
552 		break;
553 	case PF_ADDR_TABLE:
554 		pfr_detach_table(rule->src.addr.p.tbl);
555 		break;
556 	}
557 	switch (rule->dst.addr.type) {
558 	case PF_ADDR_DYNIFTL:
559 		pfi_dynaddr_remove(rule->dst.addr.p.dyn);
560 		break;
561 	case PF_ADDR_TABLE:
562 		pfr_detach_table(rule->dst.addr.p.tbl);
563 		break;
564 	}
565 	if (rule->overload_tbl)
566 		pfr_detach_table(rule->overload_tbl);
567 	if (rule->kif)
568 		pfi_kkif_unref(rule->kif);
569 	pf_kanchor_remove(rule);
570 	pf_empty_kpool(&rule->rpool.list);
571 
572 	pf_krule_free(rule);
573 }
574 
575 static void
576 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
577     unsigned int default_size)
578 {
579 	unsigned int i;
580 	unsigned int hashsize;
581 
582 	if (*tunable_size == 0 || !powerof2(*tunable_size))
583 		*tunable_size = default_size;
584 
585 	hashsize = *tunable_size;
586 	ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
587 	    M_WAITOK);
588 	ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
589 	    M_WAITOK);
590 	ts->mask = hashsize - 1;
591 	ts->seed = arc4random();
592 	for (i = 0; i < hashsize; i++) {
593 		TAILQ_INIT(&ts->namehash[i]);
594 		TAILQ_INIT(&ts->taghash[i]);
595 	}
596 	BIT_FILL(TAGID_MAX, &ts->avail);
597 }
598 
599 static void
600 pf_cleanup_tagset(struct pf_tagset *ts)
601 {
602 	unsigned int i;
603 	unsigned int hashsize;
604 	struct pf_tagname *t, *tmp;
605 
606 	/*
607 	 * Only need to clean up one of the hashes as each tag is hashed
608 	 * into each table.
609 	 */
610 	hashsize = ts->mask + 1;
611 	for (i = 0; i < hashsize; i++)
612 		TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
613 			uma_zfree(V_pf_tag_z, t);
614 
615 	free(ts->namehash, M_PFHASH);
616 	free(ts->taghash, M_PFHASH);
617 }
618 
619 static uint16_t
620 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
621 {
622 	size_t len;
623 
624 	len = strnlen(tagname, PF_TAG_NAME_SIZE - 1);
625 	return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask);
626 }
627 
628 static uint16_t
629 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
630 {
631 
632 	return (tag & ts->mask);
633 }
634 
635 static u_int16_t
636 tagname2tag(struct pf_tagset *ts, const char *tagname)
637 {
638 	struct pf_tagname	*tag;
639 	u_int32_t		 index;
640 	u_int16_t		 new_tagid;
641 
642 	PF_RULES_WASSERT();
643 
644 	index = tagname2hashindex(ts, tagname);
645 	TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
646 		if (strcmp(tagname, tag->name) == 0) {
647 			tag->ref++;
648 			return (tag->tag);
649 		}
650 
651 	/*
652 	 * new entry
653 	 *
654 	 * to avoid fragmentation, we do a linear search from the beginning
655 	 * and take the first free slot we find.
656 	 */
657 	new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
658 	/*
659 	 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
660 	 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
661 	 * set.  It may also return a bit number greater than TAGID_MAX due
662 	 * to rounding of the number of bits in the vector up to a multiple
663 	 * of the vector word size at declaration/allocation time.
664 	 */
665 	if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
666 		return (0);
667 
668 	/* Mark the tag as in use.  Bits are 0-based for BIT_CLR() */
669 	BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
670 
671 	/* allocate and fill new struct pf_tagname */
672 	tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
673 	if (tag == NULL)
674 		return (0);
675 	strlcpy(tag->name, tagname, sizeof(tag->name));
676 	tag->tag = new_tagid;
677 	tag->ref = 1;
678 
679 	/* Insert into namehash */
680 	TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
681 
682 	/* Insert into taghash */
683 	index = tag2hashindex(ts, new_tagid);
684 	TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
685 
686 	return (tag->tag);
687 }
688 
689 static void
690 tag_unref(struct pf_tagset *ts, u_int16_t tag)
691 {
692 	struct pf_tagname	*t;
693 	uint16_t		 index;
694 
695 	PF_RULES_WASSERT();
696 
697 	index = tag2hashindex(ts, tag);
698 	TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
699 		if (tag == t->tag) {
700 			if (--t->ref == 0) {
701 				TAILQ_REMOVE(&ts->taghash[index], t,
702 				    taghash_entries);
703 				index = tagname2hashindex(ts, t->name);
704 				TAILQ_REMOVE(&ts->namehash[index], t,
705 				    namehash_entries);
706 				/* Bits are 0-based for BIT_SET() */
707 				BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
708 				uma_zfree(V_pf_tag_z, t);
709 			}
710 			break;
711 		}
712 }
713 
714 static uint16_t
715 pf_tagname2tag(const char *tagname)
716 {
717 	return (tagname2tag(&V_pf_tags, tagname));
718 }
719 
720 static int
721 pf_begin_eth(uint32_t *ticket, const char *anchor)
722 {
723 	struct pf_keth_rule *rule, *tmp;
724 	struct pf_keth_ruleset *rs;
725 
726 	PF_RULES_WASSERT();
727 
728 	rs = pf_find_or_create_keth_ruleset(anchor);
729 	if (rs == NULL)
730 		return (EINVAL);
731 
732 	if (rs->inactive.open)
733 		/* We may be waiting for NET_EPOCH_CALL(pf_rollback_eth_cb) to
734 		 * finish. */
735 		return (EBUSY);
736 
737 	/* Purge old inactive rules. */
738 	TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
739 	    tmp) {
740 		TAILQ_REMOVE(rs->inactive.rules, rule,
741 		    entries);
742 		pf_free_eth_rule(rule);
743 	}
744 
745 	*ticket = ++rs->inactive.ticket;
746 	rs->inactive.open = 1;
747 
748 	return (0);
749 }
750 
751 static void
752 pf_rollback_eth_cb(struct epoch_context *ctx)
753 {
754 	struct pf_keth_ruleset *rs;
755 
756 	rs = __containerof(ctx, struct pf_keth_ruleset, epoch_ctx);
757 
758 	CURVNET_SET(rs->vnet);
759 
760 	PF_RULES_WLOCK();
761 	pf_rollback_eth(rs->inactive.ticket,
762 	    rs->anchor ? rs->anchor->path : "");
763 	PF_RULES_WUNLOCK();
764 
765 	CURVNET_RESTORE();
766 }
767 
768 static int
769 pf_rollback_eth(uint32_t ticket, const char *anchor)
770 {
771 	struct pf_keth_rule *rule, *tmp;
772 	struct pf_keth_ruleset *rs;
773 
774 	PF_RULES_WASSERT();
775 
776 	rs = pf_find_keth_ruleset(anchor);
777 	if (rs == NULL)
778 		return (EINVAL);
779 
780 	if (!rs->inactive.open ||
781 	    ticket != rs->inactive.ticket)
782 		return (0);
783 
784 	/* Purge old inactive rules. */
785 	TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
786 	    tmp) {
787 		TAILQ_REMOVE(rs->inactive.rules, rule, entries);
788 		pf_free_eth_rule(rule);
789 	}
790 
791 	rs->inactive.open = 0;
792 
793 	pf_remove_if_empty_keth_ruleset(rs);
794 
795 	return (0);
796 }
797 
798 #define	PF_SET_SKIP_STEPS(i)					\
799 	do {							\
800 		while (head[i] != cur) {			\
801 			head[i]->skip[i].ptr = cur;		\
802 			head[i] = TAILQ_NEXT(head[i], entries);	\
803 		}						\
804 	} while (0)
805 
806 static void
807 pf_eth_calc_skip_steps(struct pf_keth_ruleq *rules)
808 {
809 	struct pf_keth_rule *cur, *prev, *head[PFE_SKIP_COUNT];
810 	int i;
811 
812 	cur = TAILQ_FIRST(rules);
813 	prev = cur;
814 	for (i = 0; i < PFE_SKIP_COUNT; ++i)
815 		head[i] = cur;
816 	while (cur != NULL) {
817 		if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
818 			PF_SET_SKIP_STEPS(PFE_SKIP_IFP);
819 		if (cur->direction != prev->direction)
820 			PF_SET_SKIP_STEPS(PFE_SKIP_DIR);
821 		if (cur->proto != prev->proto)
822 			PF_SET_SKIP_STEPS(PFE_SKIP_PROTO);
823 		if (memcmp(&cur->src, &prev->src, sizeof(cur->src)) != 0)
824 			PF_SET_SKIP_STEPS(PFE_SKIP_SRC_ADDR);
825 		if (memcmp(&cur->dst, &prev->dst, sizeof(cur->dst)) != 0)
826 			PF_SET_SKIP_STEPS(PFE_SKIP_DST_ADDR);
827 
828 		prev = cur;
829 		cur = TAILQ_NEXT(cur, entries);
830 	}
831 	for (i = 0; i < PFE_SKIP_COUNT; ++i)
832 		PF_SET_SKIP_STEPS(i);
833 }
834 
835 static int
836 pf_commit_eth(uint32_t ticket, const char *anchor)
837 {
838 	struct pf_keth_ruleq *rules;
839 	struct pf_keth_ruleset *rs;
840 
841 	rs = pf_find_keth_ruleset(anchor);
842 	if (rs == NULL) {
843 		return (EINVAL);
844 	}
845 
846 	if (!rs->inactive.open ||
847 	    ticket != rs->inactive.ticket)
848 		return (EBUSY);
849 
850 	PF_RULES_WASSERT();
851 
852 	pf_eth_calc_skip_steps(rs->inactive.rules);
853 
854 	rules = rs->active.rules;
855 	ck_pr_store_ptr(&rs->active.rules, rs->inactive.rules);
856 	rs->inactive.rules = rules;
857 	rs->inactive.ticket = rs->active.ticket;
858 
859 	/* Clean up inactive rules (i.e. previously active rules), only when
860 	 * we're sure they're no longer used. */
861 	NET_EPOCH_CALL(pf_rollback_eth_cb, &rs->epoch_ctx);
862 
863 	return (0);
864 }
865 
866 #ifdef ALTQ
867 static uint16_t
868 pf_qname2qid(const char *qname)
869 {
870 	return (tagname2tag(&V_pf_qids, qname));
871 }
872 
873 static void
874 pf_qid_unref(uint16_t qid)
875 {
876 	tag_unref(&V_pf_qids, qid);
877 }
878 
879 static int
880 pf_begin_altq(u_int32_t *ticket)
881 {
882 	struct pf_altq	*altq, *tmp;
883 	int		 error = 0;
884 
885 	PF_RULES_WASSERT();
886 
887 	/* Purge the old altq lists */
888 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
889 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
890 			/* detach and destroy the discipline */
891 			error = altq_remove(altq);
892 		}
893 		free(altq, M_PFALTQ);
894 	}
895 	TAILQ_INIT(V_pf_altq_ifs_inactive);
896 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
897 		pf_qid_unref(altq->qid);
898 		free(altq, M_PFALTQ);
899 	}
900 	TAILQ_INIT(V_pf_altqs_inactive);
901 	if (error)
902 		return (error);
903 	*ticket = ++V_ticket_altqs_inactive;
904 	V_altqs_inactive_open = 1;
905 	return (0);
906 }
907 
908 static int
909 pf_rollback_altq(u_int32_t ticket)
910 {
911 	struct pf_altq	*altq, *tmp;
912 	int		 error = 0;
913 
914 	PF_RULES_WASSERT();
915 
916 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
917 		return (0);
918 	/* Purge the old altq lists */
919 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
920 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
921 			/* detach and destroy the discipline */
922 			error = altq_remove(altq);
923 		}
924 		free(altq, M_PFALTQ);
925 	}
926 	TAILQ_INIT(V_pf_altq_ifs_inactive);
927 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
928 		pf_qid_unref(altq->qid);
929 		free(altq, M_PFALTQ);
930 	}
931 	TAILQ_INIT(V_pf_altqs_inactive);
932 	V_altqs_inactive_open = 0;
933 	return (error);
934 }
935 
936 static int
937 pf_commit_altq(u_int32_t ticket)
938 {
939 	struct pf_altqqueue	*old_altqs, *old_altq_ifs;
940 	struct pf_altq		*altq, *tmp;
941 	int			 err, error = 0;
942 
943 	PF_RULES_WASSERT();
944 
945 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
946 		return (EBUSY);
947 
948 	/* swap altqs, keep the old. */
949 	old_altqs = V_pf_altqs_active;
950 	old_altq_ifs = V_pf_altq_ifs_active;
951 	V_pf_altqs_active = V_pf_altqs_inactive;
952 	V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
953 	V_pf_altqs_inactive = old_altqs;
954 	V_pf_altq_ifs_inactive = old_altq_ifs;
955 	V_ticket_altqs_active = V_ticket_altqs_inactive;
956 
957 	/* Attach new disciplines */
958 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
959 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
960 			/* attach the discipline */
961 			error = altq_pfattach(altq);
962 			if (error == 0 && V_pf_altq_running)
963 				error = pf_enable_altq(altq);
964 			if (error != 0)
965 				return (error);
966 		}
967 	}
968 
969 	/* Purge the old altq lists */
970 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
971 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
972 			/* detach and destroy the discipline */
973 			if (V_pf_altq_running)
974 				error = pf_disable_altq(altq);
975 			err = altq_pfdetach(altq);
976 			if (err != 0 && error == 0)
977 				error = err;
978 			err = altq_remove(altq);
979 			if (err != 0 && error == 0)
980 				error = err;
981 		}
982 		free(altq, M_PFALTQ);
983 	}
984 	TAILQ_INIT(V_pf_altq_ifs_inactive);
985 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
986 		pf_qid_unref(altq->qid);
987 		free(altq, M_PFALTQ);
988 	}
989 	TAILQ_INIT(V_pf_altqs_inactive);
990 
991 	V_altqs_inactive_open = 0;
992 	return (error);
993 }
994 
995 static int
996 pf_enable_altq(struct pf_altq *altq)
997 {
998 	struct ifnet		*ifp;
999 	struct tb_profile	 tb;
1000 	int			 error = 0;
1001 
1002 	if ((ifp = ifunit(altq->ifname)) == NULL)
1003 		return (EINVAL);
1004 
1005 	if (ifp->if_snd.altq_type != ALTQT_NONE)
1006 		error = altq_enable(&ifp->if_snd);
1007 
1008 	/* set tokenbucket regulator */
1009 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
1010 		tb.rate = altq->ifbandwidth;
1011 		tb.depth = altq->tbrsize;
1012 		error = tbr_set(&ifp->if_snd, &tb);
1013 	}
1014 
1015 	return (error);
1016 }
1017 
1018 static int
1019 pf_disable_altq(struct pf_altq *altq)
1020 {
1021 	struct ifnet		*ifp;
1022 	struct tb_profile	 tb;
1023 	int			 error;
1024 
1025 	if ((ifp = ifunit(altq->ifname)) == NULL)
1026 		return (EINVAL);
1027 
1028 	/*
1029 	 * when the discipline is no longer referenced, it was overridden
1030 	 * by a new one.  if so, just return.
1031 	 */
1032 	if (altq->altq_disc != ifp->if_snd.altq_disc)
1033 		return (0);
1034 
1035 	error = altq_disable(&ifp->if_snd);
1036 
1037 	if (error == 0) {
1038 		/* clear tokenbucket regulator */
1039 		tb.rate = 0;
1040 		error = tbr_set(&ifp->if_snd, &tb);
1041 	}
1042 
1043 	return (error);
1044 }
1045 
1046 static int
1047 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
1048     struct pf_altq *altq)
1049 {
1050 	struct ifnet	*ifp1;
1051 	int		 error = 0;
1052 
1053 	/* Deactivate the interface in question */
1054 	altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
1055 	if ((ifp1 = ifunit(altq->ifname)) == NULL ||
1056 	    (remove && ifp1 == ifp)) {
1057 		altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
1058 	} else {
1059 		error = altq_add(ifp1, altq);
1060 
1061 		if (ticket != V_ticket_altqs_inactive)
1062 			error = EBUSY;
1063 
1064 		if (error)
1065 			free(altq, M_PFALTQ);
1066 	}
1067 
1068 	return (error);
1069 }
1070 
1071 void
1072 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
1073 {
1074 	struct pf_altq	*a1, *a2, *a3;
1075 	u_int32_t	 ticket;
1076 	int		 error = 0;
1077 
1078 	/*
1079 	 * No need to re-evaluate the configuration for events on interfaces
1080 	 * that do not support ALTQ, as it's not possible for such
1081 	 * interfaces to be part of the configuration.
1082 	 */
1083 	if (!ALTQ_IS_READY(&ifp->if_snd))
1084 		return;
1085 
1086 	/* Interrupt userland queue modifications */
1087 	if (V_altqs_inactive_open)
1088 		pf_rollback_altq(V_ticket_altqs_inactive);
1089 
1090 	/* Start new altq ruleset */
1091 	if (pf_begin_altq(&ticket))
1092 		return;
1093 
1094 	/* Copy the current active set */
1095 	TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
1096 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1097 		if (a2 == NULL) {
1098 			error = ENOMEM;
1099 			break;
1100 		}
1101 		bcopy(a1, a2, sizeof(struct pf_altq));
1102 
1103 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1104 		if (error)
1105 			break;
1106 
1107 		TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
1108 	}
1109 	if (error)
1110 		goto out;
1111 	TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
1112 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1113 		if (a2 == NULL) {
1114 			error = ENOMEM;
1115 			break;
1116 		}
1117 		bcopy(a1, a2, sizeof(struct pf_altq));
1118 
1119 		if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
1120 			error = EBUSY;
1121 			free(a2, M_PFALTQ);
1122 			break;
1123 		}
1124 		a2->altq_disc = NULL;
1125 		TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
1126 			if (strncmp(a3->ifname, a2->ifname,
1127 				IFNAMSIZ) == 0) {
1128 				a2->altq_disc = a3->altq_disc;
1129 				break;
1130 			}
1131 		}
1132 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1133 		if (error)
1134 			break;
1135 
1136 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
1137 	}
1138 
1139 out:
1140 	if (error != 0)
1141 		pf_rollback_altq(ticket);
1142 	else
1143 		pf_commit_altq(ticket);
1144 }
1145 #endif /* ALTQ */
1146 
1147 static int
1148 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
1149 {
1150 	struct pf_krule_global *tree;
1151 	struct pf_kruleset	*rs;
1152 	struct pf_krule		*rule;
1153 
1154 	PF_RULES_WASSERT();
1155 
1156 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1157 		return (EINVAL);
1158 	tree = malloc(sizeof(struct pf_krule_global), M_TEMP, M_NOWAIT);
1159 	if (tree == NULL)
1160 		return (ENOMEM);
1161 	RB_INIT(tree);
1162 	rs = pf_find_or_create_kruleset(anchor);
1163 	if (rs == NULL) {
1164 		free(tree, M_TEMP);
1165 		return (EINVAL);
1166 	}
1167 	free(rs->rules[rs_num].inactive.tree, M_TEMP);
1168 	rs->rules[rs_num].inactive.tree = tree;
1169 
1170 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1171 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1172 		rs->rules[rs_num].inactive.rcount--;
1173 	}
1174 	*ticket = ++rs->rules[rs_num].inactive.ticket;
1175 	rs->rules[rs_num].inactive.open = 1;
1176 	return (0);
1177 }
1178 
1179 static int
1180 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
1181 {
1182 	struct pf_kruleset	*rs;
1183 	struct pf_krule		*rule;
1184 
1185 	PF_RULES_WASSERT();
1186 
1187 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1188 		return (EINVAL);
1189 	rs = pf_find_kruleset(anchor);
1190 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1191 	    rs->rules[rs_num].inactive.ticket != ticket)
1192 		return (0);
1193 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1194 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1195 		rs->rules[rs_num].inactive.rcount--;
1196 	}
1197 	rs->rules[rs_num].inactive.open = 0;
1198 	return (0);
1199 }
1200 
1201 #define PF_MD5_UPD(st, elm)						\
1202 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
1203 
1204 #define PF_MD5_UPD_STR(st, elm)						\
1205 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
1206 
1207 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
1208 		(stor) = htonl((st)->elm);				\
1209 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
1210 } while (0)
1211 
1212 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
1213 		(stor) = htons((st)->elm);				\
1214 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
1215 } while (0)
1216 
1217 static void
1218 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
1219 {
1220 	PF_MD5_UPD(pfr, addr.type);
1221 	switch (pfr->addr.type) {
1222 		case PF_ADDR_DYNIFTL:
1223 			PF_MD5_UPD(pfr, addr.v.ifname);
1224 			PF_MD5_UPD(pfr, addr.iflags);
1225 			break;
1226 		case PF_ADDR_TABLE:
1227 			PF_MD5_UPD(pfr, addr.v.tblname);
1228 			break;
1229 		case PF_ADDR_ADDRMASK:
1230 			/* XXX ignore af? */
1231 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
1232 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
1233 			break;
1234 	}
1235 
1236 	PF_MD5_UPD(pfr, port[0]);
1237 	PF_MD5_UPD(pfr, port[1]);
1238 	PF_MD5_UPD(pfr, neg);
1239 	PF_MD5_UPD(pfr, port_op);
1240 }
1241 
1242 static void
1243 pf_hash_rule_rolling(MD5_CTX *ctx, struct pf_krule *rule)
1244 {
1245 	u_int16_t x;
1246 	u_int32_t y;
1247 
1248 	pf_hash_rule_addr(ctx, &rule->src);
1249 	pf_hash_rule_addr(ctx, &rule->dst);
1250 	for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++)
1251 		PF_MD5_UPD_STR(rule, label[i]);
1252 	PF_MD5_UPD_STR(rule, ifname);
1253 	PF_MD5_UPD_STR(rule, match_tagname);
1254 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
1255 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
1256 	PF_MD5_UPD_HTONL(rule, prob, y);
1257 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
1258 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1259 	PF_MD5_UPD(rule, uid.op);
1260 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1261 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1262 	PF_MD5_UPD(rule, gid.op);
1263 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
1264 	PF_MD5_UPD(rule, action);
1265 	PF_MD5_UPD(rule, direction);
1266 	PF_MD5_UPD(rule, af);
1267 	PF_MD5_UPD(rule, quick);
1268 	PF_MD5_UPD(rule, ifnot);
1269 	PF_MD5_UPD(rule, match_tag_not);
1270 	PF_MD5_UPD(rule, natpass);
1271 	PF_MD5_UPD(rule, keep_state);
1272 	PF_MD5_UPD(rule, proto);
1273 	PF_MD5_UPD(rule, type);
1274 	PF_MD5_UPD(rule, code);
1275 	PF_MD5_UPD(rule, flags);
1276 	PF_MD5_UPD(rule, flagset);
1277 	PF_MD5_UPD(rule, allow_opts);
1278 	PF_MD5_UPD(rule, rt);
1279 	PF_MD5_UPD(rule, tos);
1280 	if (rule->anchor != NULL)
1281 		PF_MD5_UPD_STR(rule, anchor->path);
1282 }
1283 
1284 static void
1285 pf_hash_rule(struct pf_krule *rule)
1286 {
1287 	MD5_CTX		ctx;
1288 
1289 	MD5Init(&ctx);
1290 	pf_hash_rule_rolling(&ctx, rule);
1291 	MD5Final(rule->md5sum, &ctx);
1292 }
1293 
1294 static int
1295 pf_krule_compare(struct pf_krule *a, struct pf_krule *b)
1296 {
1297 
1298 	return (memcmp(a->md5sum, b->md5sum, PF_MD5_DIGEST_LENGTH));
1299 }
1300 
1301 static int
1302 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1303 {
1304 	struct pf_kruleset	*rs;
1305 	struct pf_krule		*rule, **old_array, *old_rule;
1306 	struct pf_krulequeue	*old_rules;
1307 	struct pf_krule_global  *old_tree;
1308 	int			 error;
1309 	u_int32_t		 old_rcount;
1310 
1311 	PF_RULES_WASSERT();
1312 
1313 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1314 		return (EINVAL);
1315 	rs = pf_find_kruleset(anchor);
1316 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1317 	    ticket != rs->rules[rs_num].inactive.ticket)
1318 		return (EBUSY);
1319 
1320 	/* Calculate checksum for the main ruleset */
1321 	if (rs == &pf_main_ruleset) {
1322 		error = pf_setup_pfsync_matching(rs);
1323 		if (error != 0)
1324 			return (error);
1325 	}
1326 
1327 	/* Swap rules, keep the old. */
1328 	old_rules = rs->rules[rs_num].active.ptr;
1329 	old_rcount = rs->rules[rs_num].active.rcount;
1330 	old_array = rs->rules[rs_num].active.ptr_array;
1331 	old_tree = rs->rules[rs_num].active.tree;
1332 
1333 	rs->rules[rs_num].active.ptr =
1334 	    rs->rules[rs_num].inactive.ptr;
1335 	rs->rules[rs_num].active.ptr_array =
1336 	    rs->rules[rs_num].inactive.ptr_array;
1337 	rs->rules[rs_num].active.tree =
1338 	    rs->rules[rs_num].inactive.tree;
1339 	rs->rules[rs_num].active.rcount =
1340 	    rs->rules[rs_num].inactive.rcount;
1341 
1342 	/* Attempt to preserve counter information. */
1343 	if (V_pf_status.keep_counters && old_tree != NULL) {
1344 		TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr,
1345 		    entries) {
1346 			old_rule = RB_FIND(pf_krule_global, old_tree, rule);
1347 			if (old_rule == NULL) {
1348 				continue;
1349 			}
1350 			pf_counter_u64_critical_enter();
1351 			pf_counter_u64_add_protected(&rule->evaluations,
1352 			    pf_counter_u64_fetch(&old_rule->evaluations));
1353 			pf_counter_u64_add_protected(&rule->packets[0],
1354 			    pf_counter_u64_fetch(&old_rule->packets[0]));
1355 			pf_counter_u64_add_protected(&rule->packets[1],
1356 			    pf_counter_u64_fetch(&old_rule->packets[1]));
1357 			pf_counter_u64_add_protected(&rule->bytes[0],
1358 			    pf_counter_u64_fetch(&old_rule->bytes[0]));
1359 			pf_counter_u64_add_protected(&rule->bytes[1],
1360 			    pf_counter_u64_fetch(&old_rule->bytes[1]));
1361 			pf_counter_u64_critical_exit();
1362 		}
1363 	}
1364 
1365 	rs->rules[rs_num].inactive.ptr = old_rules;
1366 	rs->rules[rs_num].inactive.ptr_array = old_array;
1367 	rs->rules[rs_num].inactive.tree = NULL;
1368 	rs->rules[rs_num].inactive.rcount = old_rcount;
1369 
1370 	rs->rules[rs_num].active.ticket =
1371 	    rs->rules[rs_num].inactive.ticket;
1372 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1373 
1374 	/* Purge the old rule list. */
1375 	PF_UNLNKDRULES_LOCK();
1376 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1377 		pf_unlink_rule_locked(old_rules, rule);
1378 	PF_UNLNKDRULES_UNLOCK();
1379 	if (rs->rules[rs_num].inactive.ptr_array)
1380 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1381 	rs->rules[rs_num].inactive.ptr_array = NULL;
1382 	rs->rules[rs_num].inactive.rcount = 0;
1383 	rs->rules[rs_num].inactive.open = 0;
1384 	pf_remove_if_empty_kruleset(rs);
1385 	free(old_tree, M_TEMP);
1386 
1387 	return (0);
1388 }
1389 
1390 static int
1391 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1392 {
1393 	MD5_CTX			 ctx;
1394 	struct pf_krule		*rule;
1395 	int			 rs_cnt;
1396 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
1397 
1398 	MD5Init(&ctx);
1399 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1400 		/* XXX PF_RULESET_SCRUB as well? */
1401 		if (rs_cnt == PF_RULESET_SCRUB)
1402 			continue;
1403 
1404 		if (rs->rules[rs_cnt].inactive.ptr_array)
1405 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1406 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
1407 
1408 		if (rs->rules[rs_cnt].inactive.rcount) {
1409 			rs->rules[rs_cnt].inactive.ptr_array =
1410 			    malloc(sizeof(caddr_t) *
1411 			    rs->rules[rs_cnt].inactive.rcount,
1412 			    M_TEMP, M_NOWAIT);
1413 
1414 			if (!rs->rules[rs_cnt].inactive.ptr_array)
1415 				return (ENOMEM);
1416 		}
1417 
1418 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1419 		    entries) {
1420 			pf_hash_rule_rolling(&ctx, rule);
1421 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1422 		}
1423 	}
1424 
1425 	MD5Final(digest, &ctx);
1426 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1427 	return (0);
1428 }
1429 
1430 static int
1431 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1432     sa_family_t af)
1433 {
1434 	int error = 0;
1435 
1436 	switch (addr->type) {
1437 	case PF_ADDR_TABLE:
1438 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1439 		if (addr->p.tbl == NULL)
1440 			error = ENOMEM;
1441 		break;
1442 	case PF_ADDR_DYNIFTL:
1443 		error = pfi_dynaddr_setup(addr, af);
1444 		break;
1445 	}
1446 
1447 	return (error);
1448 }
1449 
1450 static void
1451 pf_addr_copyout(struct pf_addr_wrap *addr)
1452 {
1453 
1454 	switch (addr->type) {
1455 	case PF_ADDR_DYNIFTL:
1456 		pfi_dynaddr_copyout(addr);
1457 		break;
1458 	case PF_ADDR_TABLE:
1459 		pf_tbladdr_copyout(addr);
1460 		break;
1461 	}
1462 }
1463 
1464 static void
1465 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
1466 {
1467 	int	secs = time_uptime, diff;
1468 
1469 	bzero(out, sizeof(struct pf_src_node));
1470 
1471 	bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
1472 	bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
1473 
1474 	if (in->rule.ptr != NULL)
1475 		out->rule.nr = in->rule.ptr->nr;
1476 
1477 	for (int i = 0; i < 2; i++) {
1478 		out->bytes[i] = counter_u64_fetch(in->bytes[i]);
1479 		out->packets[i] = counter_u64_fetch(in->packets[i]);
1480 	}
1481 
1482 	out->states = in->states;
1483 	out->conn = in->conn;
1484 	out->af = in->af;
1485 	out->ruletype = in->ruletype;
1486 
1487 	out->creation = secs - in->creation;
1488 	if (out->expire > secs)
1489 		out->expire -= secs;
1490 	else
1491 		out->expire = 0;
1492 
1493 	/* Adjust the connection rate estimate. */
1494 	diff = secs - in->conn_rate.last;
1495 	if (diff >= in->conn_rate.seconds)
1496 		out->conn_rate.count = 0;
1497 	else
1498 		out->conn_rate.count -=
1499 		    in->conn_rate.count * diff /
1500 		    in->conn_rate.seconds;
1501 }
1502 
1503 #ifdef ALTQ
1504 /*
1505  * Handle export of struct pf_kaltq to user binaries that may be using any
1506  * version of struct pf_altq.
1507  */
1508 static int
1509 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1510 {
1511 	u_int32_t version;
1512 
1513 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1514 		version = 0;
1515 	else
1516 		version = pa->version;
1517 
1518 	if (version > PFIOC_ALTQ_VERSION)
1519 		return (EINVAL);
1520 
1521 #define ASSIGN(x) exported_q->x = q->x
1522 #define COPY(x) \
1523 	bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1524 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1525 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1526 
1527 	switch (version) {
1528 	case 0: {
1529 		struct pf_altq_v0 *exported_q =
1530 		    &((struct pfioc_altq_v0 *)pa)->altq;
1531 
1532 		COPY(ifname);
1533 
1534 		ASSIGN(scheduler);
1535 		ASSIGN(tbrsize);
1536 		exported_q->tbrsize = SATU16(q->tbrsize);
1537 		exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1538 
1539 		COPY(qname);
1540 		COPY(parent);
1541 		ASSIGN(parent_qid);
1542 		exported_q->bandwidth = SATU32(q->bandwidth);
1543 		ASSIGN(priority);
1544 		ASSIGN(local_flags);
1545 
1546 		ASSIGN(qlimit);
1547 		ASSIGN(flags);
1548 
1549 		if (q->scheduler == ALTQT_HFSC) {
1550 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1551 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1552 			    SATU32(q->pq_u.hfsc_opts.x)
1553 
1554 			ASSIGN_OPT_SATU32(rtsc_m1);
1555 			ASSIGN_OPT(rtsc_d);
1556 			ASSIGN_OPT_SATU32(rtsc_m2);
1557 
1558 			ASSIGN_OPT_SATU32(lssc_m1);
1559 			ASSIGN_OPT(lssc_d);
1560 			ASSIGN_OPT_SATU32(lssc_m2);
1561 
1562 			ASSIGN_OPT_SATU32(ulsc_m1);
1563 			ASSIGN_OPT(ulsc_d);
1564 			ASSIGN_OPT_SATU32(ulsc_m2);
1565 
1566 			ASSIGN_OPT(flags);
1567 
1568 #undef ASSIGN_OPT
1569 #undef ASSIGN_OPT_SATU32
1570 		} else
1571 			COPY(pq_u);
1572 
1573 		ASSIGN(qid);
1574 		break;
1575 	}
1576 	case 1:	{
1577 		struct pf_altq_v1 *exported_q =
1578 		    &((struct pfioc_altq_v1 *)pa)->altq;
1579 
1580 		COPY(ifname);
1581 
1582 		ASSIGN(scheduler);
1583 		ASSIGN(tbrsize);
1584 		ASSIGN(ifbandwidth);
1585 
1586 		COPY(qname);
1587 		COPY(parent);
1588 		ASSIGN(parent_qid);
1589 		ASSIGN(bandwidth);
1590 		ASSIGN(priority);
1591 		ASSIGN(local_flags);
1592 
1593 		ASSIGN(qlimit);
1594 		ASSIGN(flags);
1595 		COPY(pq_u);
1596 
1597 		ASSIGN(qid);
1598 		break;
1599 	}
1600 	default:
1601 		panic("%s: unhandled struct pfioc_altq version", __func__);
1602 		break;
1603 	}
1604 
1605 #undef ASSIGN
1606 #undef COPY
1607 #undef SATU16
1608 #undef SATU32
1609 
1610 	return (0);
1611 }
1612 
1613 /*
1614  * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1615  * that may be using any version of it.
1616  */
1617 static int
1618 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1619 {
1620 	u_int32_t version;
1621 
1622 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1623 		version = 0;
1624 	else
1625 		version = pa->version;
1626 
1627 	if (version > PFIOC_ALTQ_VERSION)
1628 		return (EINVAL);
1629 
1630 #define ASSIGN(x) q->x = imported_q->x
1631 #define COPY(x) \
1632 	bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1633 
1634 	switch (version) {
1635 	case 0: {
1636 		struct pf_altq_v0 *imported_q =
1637 		    &((struct pfioc_altq_v0 *)pa)->altq;
1638 
1639 		COPY(ifname);
1640 
1641 		ASSIGN(scheduler);
1642 		ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1643 		ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1644 
1645 		COPY(qname);
1646 		COPY(parent);
1647 		ASSIGN(parent_qid);
1648 		ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1649 		ASSIGN(priority);
1650 		ASSIGN(local_flags);
1651 
1652 		ASSIGN(qlimit);
1653 		ASSIGN(flags);
1654 
1655 		if (imported_q->scheduler == ALTQT_HFSC) {
1656 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1657 
1658 			/*
1659 			 * The m1 and m2 parameters are being copied from
1660 			 * 32-bit to 64-bit.
1661 			 */
1662 			ASSIGN_OPT(rtsc_m1);
1663 			ASSIGN_OPT(rtsc_d);
1664 			ASSIGN_OPT(rtsc_m2);
1665 
1666 			ASSIGN_OPT(lssc_m1);
1667 			ASSIGN_OPT(lssc_d);
1668 			ASSIGN_OPT(lssc_m2);
1669 
1670 			ASSIGN_OPT(ulsc_m1);
1671 			ASSIGN_OPT(ulsc_d);
1672 			ASSIGN_OPT(ulsc_m2);
1673 
1674 			ASSIGN_OPT(flags);
1675 
1676 #undef ASSIGN_OPT
1677 		} else
1678 			COPY(pq_u);
1679 
1680 		ASSIGN(qid);
1681 		break;
1682 	}
1683 	case 1: {
1684 		struct pf_altq_v1 *imported_q =
1685 		    &((struct pfioc_altq_v1 *)pa)->altq;
1686 
1687 		COPY(ifname);
1688 
1689 		ASSIGN(scheduler);
1690 		ASSIGN(tbrsize);
1691 		ASSIGN(ifbandwidth);
1692 
1693 		COPY(qname);
1694 		COPY(parent);
1695 		ASSIGN(parent_qid);
1696 		ASSIGN(bandwidth);
1697 		ASSIGN(priority);
1698 		ASSIGN(local_flags);
1699 
1700 		ASSIGN(qlimit);
1701 		ASSIGN(flags);
1702 		COPY(pq_u);
1703 
1704 		ASSIGN(qid);
1705 		break;
1706 	}
1707 	default:
1708 		panic("%s: unhandled struct pfioc_altq version", __func__);
1709 		break;
1710 	}
1711 
1712 #undef ASSIGN
1713 #undef COPY
1714 
1715 	return (0);
1716 }
1717 
1718 static struct pf_altq *
1719 pf_altq_get_nth_active(u_int32_t n)
1720 {
1721 	struct pf_altq		*altq;
1722 	u_int32_t		 nr;
1723 
1724 	nr = 0;
1725 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1726 		if (nr == n)
1727 			return (altq);
1728 		nr++;
1729 	}
1730 
1731 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1732 		if (nr == n)
1733 			return (altq);
1734 		nr++;
1735 	}
1736 
1737 	return (NULL);
1738 }
1739 #endif /* ALTQ */
1740 
1741 struct pf_krule *
1742 pf_krule_alloc(void)
1743 {
1744 	struct pf_krule *rule;
1745 
1746 	rule = malloc(sizeof(struct pf_krule), M_PFRULE, M_WAITOK | M_ZERO);
1747 	mtx_init(&rule->rpool.mtx, "pf_krule_pool", NULL, MTX_DEF);
1748 	return (rule);
1749 }
1750 
1751 void
1752 pf_krule_free(struct pf_krule *rule)
1753 {
1754 #ifdef PF_WANT_32_TO_64_COUNTER
1755 	bool wowned;
1756 #endif
1757 
1758 	if (rule == NULL)
1759 		return;
1760 
1761 #ifdef PF_WANT_32_TO_64_COUNTER
1762 	if (rule->allrulelinked) {
1763 		wowned = PF_RULES_WOWNED();
1764 		if (!wowned)
1765 			PF_RULES_WLOCK();
1766 		LIST_REMOVE(rule, allrulelist);
1767 		V_pf_allrulecount--;
1768 		if (!wowned)
1769 			PF_RULES_WUNLOCK();
1770 	}
1771 #endif
1772 
1773 	pf_counter_u64_deinit(&rule->evaluations);
1774 	for (int i = 0; i < 2; i++) {
1775 		pf_counter_u64_deinit(&rule->packets[i]);
1776 		pf_counter_u64_deinit(&rule->bytes[i]);
1777 	}
1778 	counter_u64_free(rule->states_cur);
1779 	counter_u64_free(rule->states_tot);
1780 	counter_u64_free(rule->src_nodes);
1781 
1782 	mtx_destroy(&rule->rpool.mtx);
1783 	free(rule, M_PFRULE);
1784 }
1785 
1786 static void
1787 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool,
1788     struct pf_pooladdr *pool)
1789 {
1790 
1791 	bzero(pool, sizeof(*pool));
1792 	bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr));
1793 	strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname));
1794 }
1795 
1796 static int
1797 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool,
1798     struct pf_kpooladdr *kpool)
1799 {
1800 	int ret;
1801 
1802 	bzero(kpool, sizeof(*kpool));
1803 	bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr));
1804 	ret = pf_user_strcpy(kpool->ifname, pool->ifname,
1805 	    sizeof(kpool->ifname));
1806 	return (ret);
1807 }
1808 
1809 static void
1810 pf_kpool_to_pool(const struct pf_kpool *kpool, struct pf_pool *pool)
1811 {
1812 	bzero(pool, sizeof(*pool));
1813 
1814 	bcopy(&kpool->key, &pool->key, sizeof(pool->key));
1815 	bcopy(&kpool->counter, &pool->counter, sizeof(pool->counter));
1816 
1817 	pool->tblidx = kpool->tblidx;
1818 	pool->proxy_port[0] = kpool->proxy_port[0];
1819 	pool->proxy_port[1] = kpool->proxy_port[1];
1820 	pool->opts = kpool->opts;
1821 }
1822 
1823 static void
1824 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool)
1825 {
1826 	_Static_assert(sizeof(pool->key) == sizeof(kpool->key), "");
1827 	_Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), "");
1828 
1829 	bcopy(&pool->key, &kpool->key, sizeof(kpool->key));
1830 	bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter));
1831 
1832 	kpool->tblidx = pool->tblidx;
1833 	kpool->proxy_port[0] = pool->proxy_port[0];
1834 	kpool->proxy_port[1] = pool->proxy_port[1];
1835 	kpool->opts = pool->opts;
1836 }
1837 
1838 static void
1839 pf_krule_to_rule(const struct pf_krule *krule, struct pf_rule *rule)
1840 {
1841 
1842 	bzero(rule, sizeof(*rule));
1843 
1844 	bcopy(&krule->src, &rule->src, sizeof(rule->src));
1845 	bcopy(&krule->dst, &rule->dst, sizeof(rule->dst));
1846 
1847 	for (int i = 0; i < PF_SKIP_COUNT; ++i) {
1848 		if (rule->skip[i].ptr == NULL)
1849 			rule->skip[i].nr = -1;
1850 		else
1851 			rule->skip[i].nr = krule->skip[i].ptr->nr;
1852 	}
1853 
1854 	strlcpy(rule->label, krule->label[0], sizeof(rule->label));
1855 	strlcpy(rule->ifname, krule->ifname, sizeof(rule->ifname));
1856 	strlcpy(rule->qname, krule->qname, sizeof(rule->qname));
1857 	strlcpy(rule->pqname, krule->pqname, sizeof(rule->pqname));
1858 	strlcpy(rule->tagname, krule->tagname, sizeof(rule->tagname));
1859 	strlcpy(rule->match_tagname, krule->match_tagname,
1860 	    sizeof(rule->match_tagname));
1861 	strlcpy(rule->overload_tblname, krule->overload_tblname,
1862 	    sizeof(rule->overload_tblname));
1863 
1864 	pf_kpool_to_pool(&krule->rpool, &rule->rpool);
1865 
1866 	rule->evaluations = pf_counter_u64_fetch(&krule->evaluations);
1867 	for (int i = 0; i < 2; i++) {
1868 		rule->packets[i] = pf_counter_u64_fetch(&krule->packets[i]);
1869 		rule->bytes[i] = pf_counter_u64_fetch(&krule->bytes[i]);
1870 	}
1871 
1872 	/* kif, anchor, overload_tbl are not copied over. */
1873 
1874 	rule->os_fingerprint = krule->os_fingerprint;
1875 
1876 	rule->rtableid = krule->rtableid;
1877 	bcopy(krule->timeout, rule->timeout, sizeof(krule->timeout));
1878 	rule->max_states = krule->max_states;
1879 	rule->max_src_nodes = krule->max_src_nodes;
1880 	rule->max_src_states = krule->max_src_states;
1881 	rule->max_src_conn = krule->max_src_conn;
1882 	rule->max_src_conn_rate.limit = krule->max_src_conn_rate.limit;
1883 	rule->max_src_conn_rate.seconds = krule->max_src_conn_rate.seconds;
1884 	rule->qid = krule->qid;
1885 	rule->pqid = krule->pqid;
1886 	rule->nr = krule->nr;
1887 	rule->prob = krule->prob;
1888 	rule->cuid = krule->cuid;
1889 	rule->cpid = krule->cpid;
1890 
1891 	rule->return_icmp = krule->return_icmp;
1892 	rule->return_icmp6 = krule->return_icmp6;
1893 	rule->max_mss = krule->max_mss;
1894 	rule->tag = krule->tag;
1895 	rule->match_tag = krule->match_tag;
1896 	rule->scrub_flags = krule->scrub_flags;
1897 
1898 	bcopy(&krule->uid, &rule->uid, sizeof(krule->uid));
1899 	bcopy(&krule->gid, &rule->gid, sizeof(krule->gid));
1900 
1901 	rule->rule_flag = krule->rule_flag;
1902 	rule->action = krule->action;
1903 	rule->direction = krule->direction;
1904 	rule->log = krule->log;
1905 	rule->logif = krule->logif;
1906 	rule->quick = krule->quick;
1907 	rule->ifnot = krule->ifnot;
1908 	rule->match_tag_not = krule->match_tag_not;
1909 	rule->natpass = krule->natpass;
1910 
1911 	rule->keep_state = krule->keep_state;
1912 	rule->af = krule->af;
1913 	rule->proto = krule->proto;
1914 	rule->type = krule->type;
1915 	rule->code = krule->code;
1916 	rule->flags = krule->flags;
1917 	rule->flagset = krule->flagset;
1918 	rule->min_ttl = krule->min_ttl;
1919 	rule->allow_opts = krule->allow_opts;
1920 	rule->rt = krule->rt;
1921 	rule->return_ttl = krule->return_ttl;
1922 	rule->tos = krule->tos;
1923 	rule->set_tos = krule->set_tos;
1924 	rule->anchor_relative = krule->anchor_relative;
1925 	rule->anchor_wildcard = krule->anchor_wildcard;
1926 
1927 	rule->flush = krule->flush;
1928 	rule->prio = krule->prio;
1929 	rule->set_prio[0] = krule->set_prio[0];
1930 	rule->set_prio[1] = krule->set_prio[1];
1931 
1932 	bcopy(&krule->divert, &rule->divert, sizeof(krule->divert));
1933 
1934 	rule->u_states_cur = counter_u64_fetch(krule->states_cur);
1935 	rule->u_states_tot = counter_u64_fetch(krule->states_tot);
1936 	rule->u_src_nodes = counter_u64_fetch(krule->src_nodes);
1937 }
1938 
1939 static int
1940 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
1941 {
1942 	int ret;
1943 
1944 #ifndef INET
1945 	if (rule->af == AF_INET) {
1946 		return (EAFNOSUPPORT);
1947 	}
1948 #endif /* INET */
1949 #ifndef INET6
1950 	if (rule->af == AF_INET6) {
1951 		return (EAFNOSUPPORT);
1952 	}
1953 #endif /* INET6 */
1954 
1955 	ret = pf_check_rule_addr(&rule->src);
1956 	if (ret != 0)
1957 		return (ret);
1958 	ret = pf_check_rule_addr(&rule->dst);
1959 	if (ret != 0)
1960 		return (ret);
1961 
1962 	bcopy(&rule->src, &krule->src, sizeof(rule->src));
1963 	bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
1964 
1965 	ret = pf_user_strcpy(krule->label[0], rule->label, sizeof(rule->label));
1966 	if (ret != 0)
1967 		return (ret);
1968 	ret = pf_user_strcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
1969 	if (ret != 0)
1970 		return (ret);
1971 	ret = pf_user_strcpy(krule->qname, rule->qname, sizeof(rule->qname));
1972 	if (ret != 0)
1973 		return (ret);
1974 	ret = pf_user_strcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
1975 	if (ret != 0)
1976 		return (ret);
1977 	ret = pf_user_strcpy(krule->tagname, rule->tagname,
1978 	    sizeof(rule->tagname));
1979 	if (ret != 0)
1980 		return (ret);
1981 	ret = pf_user_strcpy(krule->match_tagname, rule->match_tagname,
1982 	    sizeof(rule->match_tagname));
1983 	if (ret != 0)
1984 		return (ret);
1985 	ret = pf_user_strcpy(krule->overload_tblname, rule->overload_tblname,
1986 	    sizeof(rule->overload_tblname));
1987 	if (ret != 0)
1988 		return (ret);
1989 
1990 	pf_pool_to_kpool(&rule->rpool, &krule->rpool);
1991 
1992 	/* Don't allow userspace to set evaulations, packets or bytes. */
1993 	/* kif, anchor, overload_tbl are not copied over. */
1994 
1995 	krule->os_fingerprint = rule->os_fingerprint;
1996 
1997 	krule->rtableid = rule->rtableid;
1998 	bcopy(rule->timeout, krule->timeout, sizeof(krule->timeout));
1999 	krule->max_states = rule->max_states;
2000 	krule->max_src_nodes = rule->max_src_nodes;
2001 	krule->max_src_states = rule->max_src_states;
2002 	krule->max_src_conn = rule->max_src_conn;
2003 	krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
2004 	krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
2005 	krule->qid = rule->qid;
2006 	krule->pqid = rule->pqid;
2007 	krule->nr = rule->nr;
2008 	krule->prob = rule->prob;
2009 	krule->cuid = rule->cuid;
2010 	krule->cpid = rule->cpid;
2011 
2012 	krule->return_icmp = rule->return_icmp;
2013 	krule->return_icmp6 = rule->return_icmp6;
2014 	krule->max_mss = rule->max_mss;
2015 	krule->tag = rule->tag;
2016 	krule->match_tag = rule->match_tag;
2017 	krule->scrub_flags = rule->scrub_flags;
2018 
2019 	bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
2020 	bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
2021 
2022 	krule->rule_flag = rule->rule_flag;
2023 	krule->action = rule->action;
2024 	krule->direction = rule->direction;
2025 	krule->log = rule->log;
2026 	krule->logif = rule->logif;
2027 	krule->quick = rule->quick;
2028 	krule->ifnot = rule->ifnot;
2029 	krule->match_tag_not = rule->match_tag_not;
2030 	krule->natpass = rule->natpass;
2031 
2032 	krule->keep_state = rule->keep_state;
2033 	krule->af = rule->af;
2034 	krule->proto = rule->proto;
2035 	krule->type = rule->type;
2036 	krule->code = rule->code;
2037 	krule->flags = rule->flags;
2038 	krule->flagset = rule->flagset;
2039 	krule->min_ttl = rule->min_ttl;
2040 	krule->allow_opts = rule->allow_opts;
2041 	krule->rt = rule->rt;
2042 	krule->return_ttl = rule->return_ttl;
2043 	krule->tos = rule->tos;
2044 	krule->set_tos = rule->set_tos;
2045 
2046 	krule->flush = rule->flush;
2047 	krule->prio = rule->prio;
2048 	krule->set_prio[0] = rule->set_prio[0];
2049 	krule->set_prio[1] = rule->set_prio[1];
2050 
2051 	bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
2052 
2053 	return (0);
2054 }
2055 
2056 static int
2057 pf_state_kill_to_kstate_kill(const struct pfioc_state_kill *psk,
2058     struct pf_kstate_kill *kill)
2059 {
2060 	int ret;
2061 
2062 	bzero(kill, sizeof(*kill));
2063 
2064 	bcopy(&psk->psk_pfcmp, &kill->psk_pfcmp, sizeof(kill->psk_pfcmp));
2065 	kill->psk_af = psk->psk_af;
2066 	kill->psk_proto = psk->psk_proto;
2067 	bcopy(&psk->psk_src, &kill->psk_src, sizeof(kill->psk_src));
2068 	bcopy(&psk->psk_dst, &kill->psk_dst, sizeof(kill->psk_dst));
2069 	ret = pf_user_strcpy(kill->psk_ifname, psk->psk_ifname,
2070 	    sizeof(kill->psk_ifname));
2071 	if (ret != 0)
2072 		return (ret);
2073 	ret = pf_user_strcpy(kill->psk_label, psk->psk_label,
2074 	    sizeof(kill->psk_label));
2075 	if (ret != 0)
2076 		return (ret);
2077 
2078 	return (0);
2079 }
2080 
2081 static int
2082 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket,
2083     uint32_t pool_ticket, const char *anchor, const char *anchor_call,
2084     struct thread *td)
2085 {
2086 	struct pf_kruleset	*ruleset;
2087 	struct pf_krule		*tail;
2088 	struct pf_kpooladdr	*pa;
2089 	struct pfi_kkif		*kif = NULL;
2090 	int			 rs_num;
2091 	int			 error = 0;
2092 
2093 	if ((rule->return_icmp >> 8) > ICMP_MAXTYPE) {
2094 		error = EINVAL;
2095 		goto errout_unlocked;
2096 	}
2097 
2098 #define	ERROUT(x)	ERROUT_FUNCTION(errout, x)
2099 
2100 	if (rule->ifname[0])
2101 		kif = pf_kkif_create(M_WAITOK);
2102 	pf_counter_u64_init(&rule->evaluations, M_WAITOK);
2103 	for (int i = 0; i < 2; i++) {
2104 		pf_counter_u64_init(&rule->packets[i], M_WAITOK);
2105 		pf_counter_u64_init(&rule->bytes[i], M_WAITOK);
2106 	}
2107 	rule->states_cur = counter_u64_alloc(M_WAITOK);
2108 	rule->states_tot = counter_u64_alloc(M_WAITOK);
2109 	rule->src_nodes = counter_u64_alloc(M_WAITOK);
2110 	rule->cuid = td->td_ucred->cr_ruid;
2111 	rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
2112 	TAILQ_INIT(&rule->rpool.list);
2113 
2114 	PF_CONFIG_LOCK();
2115 	PF_RULES_WLOCK();
2116 #ifdef PF_WANT_32_TO_64_COUNTER
2117 	LIST_INSERT_HEAD(&V_pf_allrulelist, rule, allrulelist);
2118 	MPASS(!rule->allrulelinked);
2119 	rule->allrulelinked = true;
2120 	V_pf_allrulecount++;
2121 #endif
2122 	ruleset = pf_find_kruleset(anchor);
2123 	if (ruleset == NULL)
2124 		ERROUT(EINVAL);
2125 	rs_num = pf_get_ruleset_number(rule->action);
2126 	if (rs_num >= PF_RULESET_MAX)
2127 		ERROUT(EINVAL);
2128 	if (ticket != ruleset->rules[rs_num].inactive.ticket) {
2129 		DPFPRINTF(PF_DEBUG_MISC,
2130 		    ("ticket: %d != [%d]%d\n", ticket, rs_num,
2131 		    ruleset->rules[rs_num].inactive.ticket));
2132 		ERROUT(EBUSY);
2133 	}
2134 	if (pool_ticket != V_ticket_pabuf) {
2135 		DPFPRINTF(PF_DEBUG_MISC,
2136 		    ("pool_ticket: %d != %d\n", pool_ticket,
2137 		    V_ticket_pabuf));
2138 		ERROUT(EBUSY);
2139 	}
2140 
2141 	tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
2142 	    pf_krulequeue);
2143 	if (tail)
2144 		rule->nr = tail->nr + 1;
2145 	else
2146 		rule->nr = 0;
2147 	if (rule->ifname[0]) {
2148 		rule->kif = pfi_kkif_attach(kif, rule->ifname);
2149 		kif = NULL;
2150 		pfi_kkif_ref(rule->kif);
2151 	} else
2152 		rule->kif = NULL;
2153 
2154 	if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
2155 		error = EBUSY;
2156 
2157 #ifdef ALTQ
2158 	/* set queue IDs */
2159 	if (rule->qname[0] != 0) {
2160 		if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2161 			error = EBUSY;
2162 		else if (rule->pqname[0] != 0) {
2163 			if ((rule->pqid =
2164 			    pf_qname2qid(rule->pqname)) == 0)
2165 				error = EBUSY;
2166 		} else
2167 			rule->pqid = rule->qid;
2168 	}
2169 #endif
2170 	if (rule->tagname[0])
2171 		if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2172 			error = EBUSY;
2173 	if (rule->match_tagname[0])
2174 		if ((rule->match_tag =
2175 		    pf_tagname2tag(rule->match_tagname)) == 0)
2176 			error = EBUSY;
2177 	if (rule->rt && !rule->direction)
2178 		error = EINVAL;
2179 	if (!rule->log)
2180 		rule->logif = 0;
2181 	if (rule->logif >= PFLOGIFS_MAX)
2182 		error = EINVAL;
2183 	if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
2184 		error = ENOMEM;
2185 	if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
2186 		error = ENOMEM;
2187 	if (pf_kanchor_setup(rule, ruleset, anchor_call))
2188 		error = EINVAL;
2189 	if (rule->scrub_flags & PFSTATE_SETPRIO &&
2190 	    (rule->set_prio[0] > PF_PRIO_MAX ||
2191 	    rule->set_prio[1] > PF_PRIO_MAX))
2192 		error = EINVAL;
2193 	TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
2194 		if (pa->addr.type == PF_ADDR_TABLE) {
2195 			pa->addr.p.tbl = pfr_attach_table(ruleset,
2196 			    pa->addr.v.tblname);
2197 			if (pa->addr.p.tbl == NULL)
2198 				error = ENOMEM;
2199 		}
2200 
2201 	rule->overload_tbl = NULL;
2202 	if (rule->overload_tblname[0]) {
2203 		if ((rule->overload_tbl = pfr_attach_table(ruleset,
2204 		    rule->overload_tblname)) == NULL)
2205 			error = EINVAL;
2206 		else
2207 			rule->overload_tbl->pfrkt_flags |=
2208 			    PFR_TFLAG_ACTIVE;
2209 	}
2210 
2211 	pf_mv_kpool(&V_pf_pabuf, &rule->rpool.list);
2212 	if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
2213 	    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
2214 	    (rule->rt > PF_NOPFROUTE)) &&
2215 	    (TAILQ_FIRST(&rule->rpool.list) == NULL))
2216 		error = EINVAL;
2217 
2218 	if (error) {
2219 		pf_free_rule(rule);
2220 		rule = NULL;
2221 		ERROUT(error);
2222 	}
2223 
2224 	rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
2225 	TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
2226 	    rule, entries);
2227 	ruleset->rules[rs_num].inactive.rcount++;
2228 
2229 	PF_RULES_WUNLOCK();
2230 	pf_hash_rule(rule);
2231 	if (RB_INSERT(pf_krule_global, ruleset->rules[rs_num].inactive.tree, rule) != NULL) {
2232 		PF_RULES_WLOCK();
2233 		pf_free_rule(rule);
2234 		rule = NULL;
2235 		error = EINVAL;
2236 		ERROUT(error);
2237 	}
2238 	PF_CONFIG_UNLOCK();
2239 
2240 	return (0);
2241 
2242 #undef ERROUT
2243 errout:
2244 	PF_RULES_WUNLOCK();
2245 	PF_CONFIG_UNLOCK();
2246 errout_unlocked:
2247 	pf_kkif_free(kif);
2248 	pf_krule_free(rule);
2249 	return (error);
2250 }
2251 
2252 static bool
2253 pf_label_match(const struct pf_krule *rule, const char *label)
2254 {
2255 	int i = 0;
2256 
2257 	while (*rule->label[i]) {
2258 		if (strcmp(rule->label[i], label) == 0)
2259 			return (true);
2260 		i++;
2261 	}
2262 
2263 	return (false);
2264 }
2265 
2266 static unsigned int
2267 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir)
2268 {
2269 	struct pf_kstate *s;
2270 	int more = 0;
2271 
2272 	s = pf_find_state_all(key, dir, &more);
2273 	if (s == NULL)
2274 		return (0);
2275 
2276 	if (more) {
2277 		PF_STATE_UNLOCK(s);
2278 		return (0);
2279 	}
2280 
2281 	pf_unlink_state(s);
2282 	return (1);
2283 }
2284 
2285 static int
2286 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih)
2287 {
2288 	struct pf_kstate	*s;
2289 	struct pf_state_key	*sk;
2290 	struct pf_addr		*srcaddr, *dstaddr;
2291 	struct pf_state_key_cmp	 match_key;
2292 	int			 idx, killed = 0;
2293 	unsigned int		 dir;
2294 	u_int16_t		 srcport, dstport;
2295 	struct pfi_kkif		*kif;
2296 
2297 relock_DIOCKILLSTATES:
2298 	PF_HASHROW_LOCK(ih);
2299 	LIST_FOREACH(s, &ih->states, entry) {
2300 		/* For floating states look at the original kif. */
2301 		kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
2302 
2303 		sk = s->key[PF_SK_WIRE];
2304 		if (s->direction == PF_OUT) {
2305 			srcaddr = &sk->addr[1];
2306 			dstaddr = &sk->addr[0];
2307 			srcport = sk->port[1];
2308 			dstport = sk->port[0];
2309 		} else {
2310 			srcaddr = &sk->addr[0];
2311 			dstaddr = &sk->addr[1];
2312 			srcport = sk->port[0];
2313 			dstport = sk->port[1];
2314 		}
2315 
2316 		if (psk->psk_af && sk->af != psk->psk_af)
2317 			continue;
2318 
2319 		if (psk->psk_proto && psk->psk_proto != sk->proto)
2320 			continue;
2321 
2322 		if (! PF_MATCHA(psk->psk_src.neg, &psk->psk_src.addr.v.a.addr,
2323 		    &psk->psk_src.addr.v.a.mask, srcaddr, sk->af))
2324 			continue;
2325 
2326 		if (! PF_MATCHA(psk->psk_dst.neg, &psk->psk_dst.addr.v.a.addr,
2327 		    &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af))
2328 			continue;
2329 
2330 		if (!  PF_MATCHA(psk->psk_rt_addr.neg,
2331 		    &psk->psk_rt_addr.addr.v.a.addr,
2332 		    &psk->psk_rt_addr.addr.v.a.mask,
2333 		    &s->rt_addr, sk->af))
2334 			continue;
2335 
2336 		if (psk->psk_src.port_op != 0 &&
2337 		    ! pf_match_port(psk->psk_src.port_op,
2338 		    psk->psk_src.port[0], psk->psk_src.port[1], srcport))
2339 			continue;
2340 
2341 		if (psk->psk_dst.port_op != 0 &&
2342 		    ! pf_match_port(psk->psk_dst.port_op,
2343 		    psk->psk_dst.port[0], psk->psk_dst.port[1], dstport))
2344 			continue;
2345 
2346 		if (psk->psk_label[0] &&
2347 		    ! pf_label_match(s->rule.ptr, psk->psk_label))
2348 			continue;
2349 
2350 		if (psk->psk_ifname[0] && strcmp(psk->psk_ifname,
2351 		    kif->pfik_name))
2352 			continue;
2353 
2354 		if (psk->psk_kill_match) {
2355 			/* Create the key to find matching states, with lock
2356 			 * held. */
2357 
2358 			bzero(&match_key, sizeof(match_key));
2359 
2360 			if (s->direction == PF_OUT) {
2361 				dir = PF_IN;
2362 				idx = PF_SK_STACK;
2363 			} else {
2364 				dir = PF_OUT;
2365 				idx = PF_SK_WIRE;
2366 			}
2367 
2368 			match_key.af = s->key[idx]->af;
2369 			match_key.proto = s->key[idx]->proto;
2370 			PF_ACPY(&match_key.addr[0],
2371 			    &s->key[idx]->addr[1], match_key.af);
2372 			match_key.port[0] = s->key[idx]->port[1];
2373 			PF_ACPY(&match_key.addr[1],
2374 			    &s->key[idx]->addr[0], match_key.af);
2375 			match_key.port[1] = s->key[idx]->port[0];
2376 		}
2377 
2378 		pf_unlink_state(s);
2379 		killed++;
2380 
2381 		if (psk->psk_kill_match)
2382 			killed += pf_kill_matching_state(&match_key, dir);
2383 
2384 		goto relock_DIOCKILLSTATES;
2385 	}
2386 	PF_HASHROW_UNLOCK(ih);
2387 
2388 	return (killed);
2389 }
2390 
2391 static int
2392 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
2393 {
2394 	int			 error = 0;
2395 	PF_RULES_RLOCK_TRACKER;
2396 
2397 #define	ERROUT_IOCTL(target, x)					\
2398     do {								\
2399 	    error = (x);						\
2400 	    SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__);	\
2401 	    goto target;						\
2402     } while (0)
2403 
2404 
2405 	/* XXX keep in sync with switch() below */
2406 	if (securelevel_gt(td->td_ucred, 2))
2407 		switch (cmd) {
2408 		case DIOCGETRULES:
2409 		case DIOCGETRULE:
2410 		case DIOCGETRULENV:
2411 		case DIOCGETADDRS:
2412 		case DIOCGETADDR:
2413 		case DIOCGETSTATE:
2414 		case DIOCGETSTATENV:
2415 		case DIOCSETSTATUSIF:
2416 		case DIOCGETSTATUS:
2417 		case DIOCGETSTATUSNV:
2418 		case DIOCCLRSTATUS:
2419 		case DIOCNATLOOK:
2420 		case DIOCSETDEBUG:
2421 		case DIOCGETSTATES:
2422 		case DIOCGETSTATESV2:
2423 		case DIOCGETTIMEOUT:
2424 		case DIOCCLRRULECTRS:
2425 		case DIOCGETLIMIT:
2426 		case DIOCGETALTQSV0:
2427 		case DIOCGETALTQSV1:
2428 		case DIOCGETALTQV0:
2429 		case DIOCGETALTQV1:
2430 		case DIOCGETQSTATSV0:
2431 		case DIOCGETQSTATSV1:
2432 		case DIOCGETRULESETS:
2433 		case DIOCGETRULESET:
2434 		case DIOCRGETTABLES:
2435 		case DIOCRGETTSTATS:
2436 		case DIOCRCLRTSTATS:
2437 		case DIOCRCLRADDRS:
2438 		case DIOCRADDADDRS:
2439 		case DIOCRDELADDRS:
2440 		case DIOCRSETADDRS:
2441 		case DIOCRGETADDRS:
2442 		case DIOCRGETASTATS:
2443 		case DIOCRCLRASTATS:
2444 		case DIOCRTSTADDRS:
2445 		case DIOCOSFPGET:
2446 		case DIOCGETSRCNODES:
2447 		case DIOCCLRSRCNODES:
2448 		case DIOCGETSYNCOOKIES:
2449 		case DIOCIGETIFACES:
2450 		case DIOCGIFSPEEDV0:
2451 		case DIOCGIFSPEEDV1:
2452 		case DIOCSETIFFLAG:
2453 		case DIOCCLRIFFLAG:
2454 		case DIOCGETETHRULES:
2455 		case DIOCGETETHRULE:
2456 			break;
2457 		case DIOCRCLRTABLES:
2458 		case DIOCRADDTABLES:
2459 		case DIOCRDELTABLES:
2460 		case DIOCRSETTFLAGS:
2461 			if (((struct pfioc_table *)addr)->pfrio_flags &
2462 			    PFR_FLAG_DUMMY)
2463 				break; /* dummy operation ok */
2464 			return (EPERM);
2465 		default:
2466 			return (EPERM);
2467 		}
2468 
2469 	if (!(flags & FWRITE))
2470 		switch (cmd) {
2471 		case DIOCGETRULES:
2472 		case DIOCGETADDRS:
2473 		case DIOCGETADDR:
2474 		case DIOCGETSTATE:
2475 		case DIOCGETSTATENV:
2476 		case DIOCGETSTATUS:
2477 		case DIOCGETSTATUSNV:
2478 		case DIOCGETSTATES:
2479 		case DIOCGETSTATESV2:
2480 		case DIOCGETTIMEOUT:
2481 		case DIOCGETLIMIT:
2482 		case DIOCGETALTQSV0:
2483 		case DIOCGETALTQSV1:
2484 		case DIOCGETALTQV0:
2485 		case DIOCGETALTQV1:
2486 		case DIOCGETQSTATSV0:
2487 		case DIOCGETQSTATSV1:
2488 		case DIOCGETRULESETS:
2489 		case DIOCGETRULESET:
2490 		case DIOCNATLOOK:
2491 		case DIOCRGETTABLES:
2492 		case DIOCRGETTSTATS:
2493 		case DIOCRGETADDRS:
2494 		case DIOCRGETASTATS:
2495 		case DIOCRTSTADDRS:
2496 		case DIOCOSFPGET:
2497 		case DIOCGETSRCNODES:
2498 		case DIOCGETSYNCOOKIES:
2499 		case DIOCIGETIFACES:
2500 		case DIOCGIFSPEEDV1:
2501 		case DIOCGIFSPEEDV0:
2502 		case DIOCGETRULENV:
2503 		case DIOCGETETHRULES:
2504 		case DIOCGETETHRULE:
2505 			break;
2506 		case DIOCRCLRTABLES:
2507 		case DIOCRADDTABLES:
2508 		case DIOCRDELTABLES:
2509 		case DIOCRCLRTSTATS:
2510 		case DIOCRCLRADDRS:
2511 		case DIOCRADDADDRS:
2512 		case DIOCRDELADDRS:
2513 		case DIOCRSETADDRS:
2514 		case DIOCRSETTFLAGS:
2515 			if (((struct pfioc_table *)addr)->pfrio_flags &
2516 			    PFR_FLAG_DUMMY) {
2517 				flags |= FWRITE; /* need write lock for dummy */
2518 				break; /* dummy operation ok */
2519 			}
2520 			return (EACCES);
2521 		case DIOCGETRULE:
2522 			if (((struct pfioc_rule *)addr)->action ==
2523 			    PF_GET_CLR_CNTR)
2524 				return (EACCES);
2525 			break;
2526 		default:
2527 			return (EACCES);
2528 		}
2529 
2530 	CURVNET_SET(TD_TO_VNET(td));
2531 
2532 	switch (cmd) {
2533 	case DIOCSTART:
2534 		sx_xlock(&pf_ioctl_lock);
2535 		if (V_pf_status.running)
2536 			error = EEXIST;
2537 		else {
2538 			int cpu;
2539 
2540 			hook_pf();
2541 			if (! TAILQ_EMPTY(V_pf_keth->active.rules))
2542 				hook_pf_eth();
2543 			V_pf_status.running = 1;
2544 			V_pf_status.since = time_second;
2545 
2546 			CPU_FOREACH(cpu)
2547 				V_pf_stateid[cpu] = time_second;
2548 
2549 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
2550 		}
2551 		break;
2552 
2553 	case DIOCSTOP:
2554 		sx_xlock(&pf_ioctl_lock);
2555 		if (!V_pf_status.running)
2556 			error = ENOENT;
2557 		else {
2558 			V_pf_status.running = 0;
2559 			dehook_pf();
2560 			dehook_pf_eth();
2561 			V_pf_status.since = time_second;
2562 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
2563 		}
2564 		break;
2565 
2566 	case DIOCGETETHRULES: {
2567 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2568 		nvlist_t		*nvl;
2569 		void			*packed;
2570 		struct pf_keth_rule	*tail;
2571 		struct pf_keth_ruleset	*rs;
2572 		u_int32_t		 ticket, nr;
2573 		const char		*anchor = "";
2574 
2575 		nvl = NULL;
2576 		packed = NULL;
2577 
2578 #define	ERROUT(x)	do { error = (x); goto DIOCGETETHRULES_error; } while (0)
2579 
2580 		if (nv->len > pf_ioctl_maxcount)
2581 			ERROUT(ENOMEM);
2582 
2583 		/* Copy the request in */
2584 		packed = malloc(nv->len, M_NVLIST, M_WAITOK);
2585 		if (packed == NULL)
2586 			ERROUT(ENOMEM);
2587 
2588 		error = copyin(nv->data, packed, nv->len);
2589 		if (error)
2590 			ERROUT(error);
2591 
2592 		nvl = nvlist_unpack(packed, nv->len, 0);
2593 		if (nvl == NULL)
2594 			ERROUT(EBADMSG);
2595 
2596 		if (! nvlist_exists_string(nvl, "anchor"))
2597 			ERROUT(EBADMSG);
2598 
2599 		anchor = nvlist_get_string(nvl, "anchor");
2600 
2601 		rs = pf_find_keth_ruleset(anchor);
2602 
2603 		nvlist_destroy(nvl);
2604 		nvl = NULL;
2605 		free(packed, M_NVLIST);
2606 		packed = NULL;
2607 
2608 		if (rs == NULL)
2609 			ERROUT(ENOENT);
2610 
2611 		/* Reply */
2612 		nvl = nvlist_create(0);
2613 		if (nvl == NULL)
2614 			ERROUT(ENOMEM);
2615 
2616 		PF_RULES_RLOCK();
2617 
2618 		ticket = rs->active.ticket;
2619 		tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq);
2620 		if (tail)
2621 			nr = tail->nr + 1;
2622 		else
2623 			nr = 0;
2624 
2625 		PF_RULES_RUNLOCK();
2626 
2627 		nvlist_add_number(nvl, "ticket", ticket);
2628 		nvlist_add_number(nvl, "nr", nr);
2629 
2630 		packed = nvlist_pack(nvl, &nv->len);
2631 		if (packed == NULL)
2632 			ERROUT(ENOMEM);
2633 
2634 		if (nv->size == 0)
2635 			ERROUT(0);
2636 		else if (nv->size < nv->len)
2637 			ERROUT(ENOSPC);
2638 
2639 		error = copyout(packed, nv->data, nv->len);
2640 
2641 #undef ERROUT
2642 DIOCGETETHRULES_error:
2643 		free(packed, M_NVLIST);
2644 		nvlist_destroy(nvl);
2645 		break;
2646 	}
2647 
2648 	case DIOCGETETHRULE: {
2649 		struct epoch_tracker	 et;
2650 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2651 		nvlist_t		*nvl = NULL;
2652 		void			*nvlpacked = NULL;
2653 		struct pf_keth_rule	*rule = NULL;
2654 		struct pf_keth_ruleset	*rs;
2655 		u_int32_t		 ticket, nr;
2656 		bool			 clear = false;
2657 		const char		*anchor;
2658 
2659 #define ERROUT(x)	do { error = (x); goto DIOCGETETHRULE_error; } while (0)
2660 
2661 		nvlpacked = malloc(nv->len, M_TEMP, M_WAITOK);
2662 		if (nvlpacked == NULL)
2663 			ERROUT(ENOMEM);
2664 
2665 		error = copyin(nv->data, nvlpacked, nv->len);
2666 		if (error)
2667 			ERROUT(error);
2668 
2669 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2670 		if (! nvlist_exists_number(nvl, "ticket"))
2671 			ERROUT(EBADMSG);
2672 		ticket = nvlist_get_number(nvl, "ticket");
2673 		if (! nvlist_exists_string(nvl, "anchor"))
2674 			ERROUT(EBADMSG);
2675 		anchor = nvlist_get_string(nvl, "anchor");
2676 
2677 		if (nvlist_exists_bool(nvl, "clear"))
2678 			clear = nvlist_get_bool(nvl, "clear");
2679 
2680 		if (clear && !(flags & FWRITE))
2681 			ERROUT(EACCES);
2682 
2683 		if (! nvlist_exists_number(nvl, "nr"))
2684 			ERROUT(EBADMSG);
2685 		nr = nvlist_get_number(nvl, "nr");
2686 
2687 		PF_RULES_RLOCK();
2688 		rs = pf_find_keth_ruleset(anchor);
2689 		if (rs == NULL) {
2690 			PF_RULES_RUNLOCK();
2691 			ERROUT(ENOENT);
2692 		}
2693 		if (ticket != rs->active.ticket) {
2694 			PF_RULES_RUNLOCK();
2695 			ERROUT(EBUSY);
2696 		}
2697 
2698 		nvlist_destroy(nvl);
2699 		nvl = NULL;
2700 		free(nvlpacked, M_TEMP);
2701 		nvlpacked = NULL;
2702 
2703 		nvl = nvlist_create(0);
2704 
2705 		rule = TAILQ_FIRST(rs->active.rules);
2706 		while ((rule != NULL) && (rule->nr != nr))
2707 			rule = TAILQ_NEXT(rule, entries);
2708 		if (rule == NULL) {
2709 			PF_RULES_RUNLOCK();
2710 			ERROUT(ENOENT);
2711 		}
2712 		/* Make sure rule can't go away. */
2713 		NET_EPOCH_ENTER(et);
2714 		PF_RULES_RUNLOCK();
2715 		nvl = pf_keth_rule_to_nveth_rule(rule);
2716 		if (pf_keth_anchor_nvcopyout(rs, rule, nvl))
2717 			ERROUT(EBUSY);
2718 		NET_EPOCH_EXIT(et);
2719 		if (nvl == NULL)
2720 			ERROUT(ENOMEM);
2721 
2722 		nvlpacked = nvlist_pack(nvl, &nv->len);
2723 		if (nvlpacked == NULL)
2724 			ERROUT(ENOMEM);
2725 
2726 		if (nv->size == 0)
2727 			ERROUT(0);
2728 		else if (nv->size < nv->len)
2729 			ERROUT(ENOSPC);
2730 
2731 		error = copyout(nvlpacked, nv->data, nv->len);
2732 		if (error == 0 && clear) {
2733 			counter_u64_zero(rule->evaluations);
2734 			for (int i = 0; i < 2; i++) {
2735 				counter_u64_zero(rule->packets[i]);
2736 				counter_u64_zero(rule->bytes[i]);
2737 			}
2738 		}
2739 
2740 #undef ERROUT
2741 DIOCGETETHRULE_error:
2742 		free(nvlpacked, M_TEMP);
2743 		nvlist_destroy(nvl);
2744 		break;
2745 	}
2746 
2747 	case DIOCADDETHRULE: {
2748 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2749 		nvlist_t		*nvl = NULL;
2750 		void			*nvlpacked = NULL;
2751 		struct pf_keth_rule	*rule = NULL, *tail = NULL;
2752 		struct pf_keth_ruleset	*ruleset = NULL;
2753 		struct pfi_kkif		*kif = NULL;
2754 		const char		*anchor = "", *anchor_call = "";
2755 
2756 #define ERROUT(x)	do { error = (x); goto DIOCADDETHRULE_error; } while (0)
2757 
2758 		nvlpacked = malloc(nv->len, M_TEMP, M_WAITOK);
2759 		if (nvlpacked == NULL)
2760 			ERROUT(ENOMEM);
2761 
2762 		error = copyin(nv->data, nvlpacked, nv->len);
2763 		if (error)
2764 			ERROUT(error);
2765 
2766 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2767 		if (nvl == NULL)
2768 			ERROUT(EBADMSG);
2769 
2770 		if (! nvlist_exists_number(nvl, "ticket"))
2771 			ERROUT(EBADMSG);
2772 
2773 		if (nvlist_exists_string(nvl, "anchor"))
2774 			anchor = nvlist_get_string(nvl, "anchor");
2775 		if (nvlist_exists_string(nvl, "anchor_call"))
2776 			anchor_call = nvlist_get_string(nvl, "anchor_call");
2777 
2778 		ruleset = pf_find_keth_ruleset(anchor);
2779 		if (ruleset == NULL)
2780 			ERROUT(EINVAL);
2781 
2782 		if (nvlist_get_number(nvl, "ticket") !=
2783 		    ruleset->inactive.ticket) {
2784 			DPFPRINTF(PF_DEBUG_MISC,
2785 			    ("ticket: %d != %d\n",
2786 			    (u_int32_t)nvlist_get_number(nvl, "ticket"),
2787 			    ruleset->inactive.ticket));
2788 			ERROUT(EBUSY);
2789 		}
2790 
2791 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
2792 		if (rule == NULL)
2793 			ERROUT(ENOMEM);
2794 
2795 		error = pf_nveth_rule_to_keth_rule(nvl, rule);
2796 		if (error != 0)
2797 			ERROUT(error);
2798 
2799 		if (rule->ifname[0])
2800 			kif = pf_kkif_create(M_WAITOK);
2801 		rule->evaluations = counter_u64_alloc(M_WAITOK);
2802 		for (int i = 0; i < 2; i++) {
2803 			rule->packets[i] = counter_u64_alloc(M_WAITOK);
2804 			rule->bytes[i] = counter_u64_alloc(M_WAITOK);
2805 		}
2806 
2807 		PF_RULES_WLOCK();
2808 
2809 		if (rule->ifname[0]) {
2810 			rule->kif = pfi_kkif_attach(kif, rule->ifname);
2811 			pfi_kkif_ref(rule->kif);
2812 		} else
2813 			rule->kif = NULL;
2814 
2815 #ifdef ALTQ
2816 		/* set queue IDs */
2817 		if (rule->qname[0] != 0) {
2818 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2819 				error = EBUSY;
2820 			else
2821 				rule->qid = rule->qid;
2822 		}
2823 #endif
2824 		if (rule->tagname[0])
2825 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2826 				error = EBUSY;
2827 
2828 		if (error) {
2829 			pf_free_eth_rule(rule);
2830 			PF_RULES_WUNLOCK();
2831 			ERROUT(error);
2832 		}
2833 
2834 		if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) {
2835 			pf_free_eth_rule(rule);
2836 			PF_RULES_WUNLOCK();
2837 			ERROUT(EINVAL);
2838 		}
2839 
2840 		tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq);
2841 		if (tail)
2842 			rule->nr = tail->nr + 1;
2843 		else
2844 			rule->nr = 0;
2845 
2846 		TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries);
2847 
2848 		PF_RULES_WUNLOCK();
2849 
2850 #undef ERROUT
2851 DIOCADDETHRULE_error:
2852 		nvlist_destroy(nvl);
2853 		free(nvlpacked, M_TEMP);
2854 		break;
2855 	}
2856 
2857 	case DIOCADDRULENV: {
2858 		struct pfioc_nv	*nv = (struct pfioc_nv *)addr;
2859 		nvlist_t	*nvl = NULL;
2860 		void		*nvlpacked = NULL;
2861 		struct pf_krule	*rule = NULL;
2862 		const char	*anchor = "", *anchor_call = "";
2863 		uint32_t	 ticket = 0, pool_ticket = 0;
2864 
2865 #define	ERROUT(x)	ERROUT_IOCTL(DIOCADDRULENV_error, x)
2866 
2867 		if (nv->len > pf_ioctl_maxcount)
2868 			ERROUT(ENOMEM);
2869 
2870 		nvlpacked = malloc(nv->len, M_TEMP, M_WAITOK);
2871 		error = copyin(nv->data, nvlpacked, nv->len);
2872 		if (error)
2873 			ERROUT(error);
2874 
2875 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2876 		if (nvl == NULL)
2877 			ERROUT(EBADMSG);
2878 
2879 		if (! nvlist_exists_number(nvl, "ticket"))
2880 			ERROUT(EINVAL);
2881 		ticket = nvlist_get_number(nvl, "ticket");
2882 
2883 		if (! nvlist_exists_number(nvl, "pool_ticket"))
2884 			ERROUT(EINVAL);
2885 		pool_ticket = nvlist_get_number(nvl, "pool_ticket");
2886 
2887 		if (! nvlist_exists_nvlist(nvl, "rule"))
2888 			ERROUT(EINVAL);
2889 
2890 		rule = pf_krule_alloc();
2891 		error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"),
2892 		    rule);
2893 		if (error)
2894 			ERROUT(error);
2895 
2896 		if (nvlist_exists_string(nvl, "anchor"))
2897 			anchor = nvlist_get_string(nvl, "anchor");
2898 		if (nvlist_exists_string(nvl, "anchor_call"))
2899 			anchor_call = nvlist_get_string(nvl, "anchor_call");
2900 
2901 		if ((error = nvlist_error(nvl)))
2902 			ERROUT(error);
2903 
2904 		/* Frees rule on error */
2905 		error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor,
2906 		    anchor_call, td);
2907 
2908 		nvlist_destroy(nvl);
2909 		free(nvlpacked, M_TEMP);
2910 		break;
2911 #undef ERROUT
2912 DIOCADDRULENV_error:
2913 		pf_krule_free(rule);
2914 		nvlist_destroy(nvl);
2915 		free(nvlpacked, M_TEMP);
2916 
2917 		break;
2918 	}
2919 	case DIOCADDRULE: {
2920 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
2921 		struct pf_krule		*rule;
2922 
2923 		rule = pf_krule_alloc();
2924 		error = pf_rule_to_krule(&pr->rule, rule);
2925 		if (error != 0) {
2926 			pf_krule_free(rule);
2927 			break;
2928 		}
2929 
2930 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
2931 
2932 		/* Frees rule on error */
2933 		error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket,
2934 		    pr->anchor, pr->anchor_call, td);
2935 		break;
2936 	}
2937 
2938 	case DIOCGETRULES: {
2939 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
2940 		struct pf_kruleset	*ruleset;
2941 		struct pf_krule		*tail;
2942 		int			 rs_num;
2943 
2944 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
2945 
2946 		PF_RULES_WLOCK();
2947 		ruleset = pf_find_kruleset(pr->anchor);
2948 		if (ruleset == NULL) {
2949 			PF_RULES_WUNLOCK();
2950 			error = EINVAL;
2951 			break;
2952 		}
2953 		rs_num = pf_get_ruleset_number(pr->rule.action);
2954 		if (rs_num >= PF_RULESET_MAX) {
2955 			PF_RULES_WUNLOCK();
2956 			error = EINVAL;
2957 			break;
2958 		}
2959 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
2960 		    pf_krulequeue);
2961 		if (tail)
2962 			pr->nr = tail->nr + 1;
2963 		else
2964 			pr->nr = 0;
2965 		pr->ticket = ruleset->rules[rs_num].active.ticket;
2966 		PF_RULES_WUNLOCK();
2967 		break;
2968 	}
2969 
2970 	case DIOCGETRULE: {
2971 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
2972 		struct pf_kruleset	*ruleset;
2973 		struct pf_krule		*rule;
2974 		int			 rs_num;
2975 
2976 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
2977 
2978 		PF_RULES_WLOCK();
2979 		ruleset = pf_find_kruleset(pr->anchor);
2980 		if (ruleset == NULL) {
2981 			PF_RULES_WUNLOCK();
2982 			error = EINVAL;
2983 			break;
2984 		}
2985 		rs_num = pf_get_ruleset_number(pr->rule.action);
2986 		if (rs_num >= PF_RULESET_MAX) {
2987 			PF_RULES_WUNLOCK();
2988 			error = EINVAL;
2989 			break;
2990 		}
2991 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
2992 			PF_RULES_WUNLOCK();
2993 			error = EBUSY;
2994 			break;
2995 		}
2996 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
2997 		while ((rule != NULL) && (rule->nr != pr->nr))
2998 			rule = TAILQ_NEXT(rule, entries);
2999 		if (rule == NULL) {
3000 			PF_RULES_WUNLOCK();
3001 			error = EBUSY;
3002 			break;
3003 		}
3004 
3005 		pf_krule_to_rule(rule, &pr->rule);
3006 
3007 		if (pf_kanchor_copyout(ruleset, rule, pr)) {
3008 			PF_RULES_WUNLOCK();
3009 			error = EBUSY;
3010 			break;
3011 		}
3012 		pf_addr_copyout(&pr->rule.src.addr);
3013 		pf_addr_copyout(&pr->rule.dst.addr);
3014 
3015 		if (pr->action == PF_GET_CLR_CNTR) {
3016 			pf_counter_u64_zero(&rule->evaluations);
3017 			for (int i = 0; i < 2; i++) {
3018 				pf_counter_u64_zero(&rule->packets[i]);
3019 				pf_counter_u64_zero(&rule->bytes[i]);
3020 			}
3021 			counter_u64_zero(rule->states_tot);
3022 		}
3023 		PF_RULES_WUNLOCK();
3024 		break;
3025 	}
3026 
3027 	case DIOCGETRULENV: {
3028 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
3029 		nvlist_t		*nvrule = NULL;
3030 		nvlist_t		*nvl = NULL;
3031 		struct pf_kruleset	*ruleset;
3032 		struct pf_krule		*rule;
3033 		void			*nvlpacked = NULL;
3034 		int			 rs_num, nr;
3035 		bool			 clear_counter = false;
3036 
3037 #define	ERROUT(x)	ERROUT_IOCTL(DIOCGETRULENV_error, x)
3038 
3039 		if (nv->len > pf_ioctl_maxcount)
3040 			ERROUT(ENOMEM);
3041 
3042 		/* Copy the request in */
3043 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3044 		if (nvlpacked == NULL)
3045 			ERROUT(ENOMEM);
3046 
3047 		error = copyin(nv->data, nvlpacked, nv->len);
3048 		if (error)
3049 			ERROUT(error);
3050 
3051 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3052 		if (nvl == NULL)
3053 			ERROUT(EBADMSG);
3054 
3055 		if (! nvlist_exists_string(nvl, "anchor"))
3056 			ERROUT(EBADMSG);
3057 		if (! nvlist_exists_number(nvl, "ruleset"))
3058 			ERROUT(EBADMSG);
3059 		if (! nvlist_exists_number(nvl, "ticket"))
3060 			ERROUT(EBADMSG);
3061 		if (! nvlist_exists_number(nvl, "nr"))
3062 			ERROUT(EBADMSG);
3063 
3064 		if (nvlist_exists_bool(nvl, "clear_counter"))
3065 			clear_counter = nvlist_get_bool(nvl, "clear_counter");
3066 
3067 		if (clear_counter && !(flags & FWRITE))
3068 			ERROUT(EACCES);
3069 
3070 		nr = nvlist_get_number(nvl, "nr");
3071 
3072 		PF_RULES_WLOCK();
3073 		ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor"));
3074 		if (ruleset == NULL) {
3075 			PF_RULES_WUNLOCK();
3076 			ERROUT(ENOENT);
3077 		}
3078 
3079 		rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset"));
3080 		if (rs_num >= PF_RULESET_MAX) {
3081 			PF_RULES_WUNLOCK();
3082 			ERROUT(EINVAL);
3083 		}
3084 
3085 		if (nvlist_get_number(nvl, "ticket") !=
3086 		    ruleset->rules[rs_num].active.ticket) {
3087 			PF_RULES_WUNLOCK();
3088 			ERROUT(EBUSY);
3089 		}
3090 
3091 		if ((error = nvlist_error(nvl))) {
3092 			PF_RULES_WUNLOCK();
3093 			ERROUT(error);
3094 		}
3095 
3096 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
3097 		while ((rule != NULL) && (rule->nr != nr))
3098 			rule = TAILQ_NEXT(rule, entries);
3099 		if (rule == NULL) {
3100 			PF_RULES_WUNLOCK();
3101 			ERROUT(EBUSY);
3102 		}
3103 
3104 		nvrule = pf_krule_to_nvrule(rule);
3105 
3106 		nvlist_destroy(nvl);
3107 		nvl = nvlist_create(0);
3108 		if (nvl == NULL) {
3109 			PF_RULES_WUNLOCK();
3110 			ERROUT(ENOMEM);
3111 		}
3112 		nvlist_add_number(nvl, "nr", nr);
3113 		nvlist_add_nvlist(nvl, "rule", nvrule);
3114 		nvlist_destroy(nvrule);
3115 		nvrule = NULL;
3116 		if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) {
3117 			PF_RULES_WUNLOCK();
3118 			ERROUT(EBUSY);
3119 		}
3120 
3121 		free(nvlpacked, M_NVLIST);
3122 		nvlpacked = nvlist_pack(nvl, &nv->len);
3123 		if (nvlpacked == NULL) {
3124 			PF_RULES_WUNLOCK();
3125 			ERROUT(ENOMEM);
3126 		}
3127 
3128 		if (nv->size == 0) {
3129 			PF_RULES_WUNLOCK();
3130 			ERROUT(0);
3131 		}
3132 		else if (nv->size < nv->len) {
3133 			PF_RULES_WUNLOCK();
3134 			ERROUT(ENOSPC);
3135 		}
3136 
3137 		if (clear_counter) {
3138 			pf_counter_u64_zero(&rule->evaluations);
3139 			for (int i = 0; i < 2; i++) {
3140 				pf_counter_u64_zero(&rule->packets[i]);
3141 				pf_counter_u64_zero(&rule->bytes[i]);
3142 			}
3143 			counter_u64_zero(rule->states_tot);
3144 		}
3145 		PF_RULES_WUNLOCK();
3146 
3147 		error = copyout(nvlpacked, nv->data, nv->len);
3148 
3149 #undef ERROUT
3150 DIOCGETRULENV_error:
3151 		free(nvlpacked, M_NVLIST);
3152 		nvlist_destroy(nvrule);
3153 		nvlist_destroy(nvl);
3154 
3155 		break;
3156 	}
3157 
3158 	case DIOCCHANGERULE: {
3159 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
3160 		struct pf_kruleset	*ruleset;
3161 		struct pf_krule		*oldrule = NULL, *newrule = NULL;
3162 		struct pfi_kkif		*kif = NULL;
3163 		struct pf_kpooladdr	*pa;
3164 		u_int32_t		 nr = 0;
3165 		int			 rs_num;
3166 
3167 		pcr->anchor[sizeof(pcr->anchor) - 1] = 0;
3168 
3169 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
3170 		    pcr->action > PF_CHANGE_GET_TICKET) {
3171 			error = EINVAL;
3172 			break;
3173 		}
3174 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
3175 			error = EINVAL;
3176 			break;
3177 		}
3178 
3179 		if (pcr->action != PF_CHANGE_REMOVE) {
3180 			newrule = pf_krule_alloc();
3181 			error = pf_rule_to_krule(&pcr->rule, newrule);
3182 			if (error != 0) {
3183 				free(newrule, M_PFRULE);
3184 				break;
3185 			}
3186 
3187 			if (newrule->ifname[0])
3188 				kif = pf_kkif_create(M_WAITOK);
3189 			pf_counter_u64_init(&newrule->evaluations, M_WAITOK);
3190 			for (int i = 0; i < 2; i++) {
3191 				pf_counter_u64_init(&newrule->packets[i], M_WAITOK);
3192 				pf_counter_u64_init(&newrule->bytes[i], M_WAITOK);
3193 			}
3194 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
3195 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
3196 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
3197 			newrule->cuid = td->td_ucred->cr_ruid;
3198 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
3199 			TAILQ_INIT(&newrule->rpool.list);
3200 		}
3201 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGERULE_error; }
3202 
3203 		PF_RULES_WLOCK();
3204 #ifdef PF_WANT_32_TO_64_COUNTER
3205 		if (newrule != NULL) {
3206 			LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist);
3207 			newrule->allrulelinked = true;
3208 			V_pf_allrulecount++;
3209 		}
3210 #endif
3211 
3212 		if (!(pcr->action == PF_CHANGE_REMOVE ||
3213 		    pcr->action == PF_CHANGE_GET_TICKET) &&
3214 		    pcr->pool_ticket != V_ticket_pabuf)
3215 			ERROUT(EBUSY);
3216 
3217 		ruleset = pf_find_kruleset(pcr->anchor);
3218 		if (ruleset == NULL)
3219 			ERROUT(EINVAL);
3220 
3221 		rs_num = pf_get_ruleset_number(pcr->rule.action);
3222 		if (rs_num >= PF_RULESET_MAX)
3223 			ERROUT(EINVAL);
3224 
3225 		if (pcr->action == PF_CHANGE_GET_TICKET) {
3226 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
3227 			ERROUT(0);
3228 		} else if (pcr->ticket !=
3229 			    ruleset->rules[rs_num].active.ticket)
3230 				ERROUT(EINVAL);
3231 
3232 		if (pcr->action != PF_CHANGE_REMOVE) {
3233 			if (newrule->ifname[0]) {
3234 				newrule->kif = pfi_kkif_attach(kif,
3235 				    newrule->ifname);
3236 				kif = NULL;
3237 				pfi_kkif_ref(newrule->kif);
3238 			} else
3239 				newrule->kif = NULL;
3240 
3241 			if (newrule->rtableid > 0 &&
3242 			    newrule->rtableid >= rt_numfibs)
3243 				error = EBUSY;
3244 
3245 #ifdef ALTQ
3246 			/* set queue IDs */
3247 			if (newrule->qname[0] != 0) {
3248 				if ((newrule->qid =
3249 				    pf_qname2qid(newrule->qname)) == 0)
3250 					error = EBUSY;
3251 				else if (newrule->pqname[0] != 0) {
3252 					if ((newrule->pqid =
3253 					    pf_qname2qid(newrule->pqname)) == 0)
3254 						error = EBUSY;
3255 				} else
3256 					newrule->pqid = newrule->qid;
3257 			}
3258 #endif /* ALTQ */
3259 			if (newrule->tagname[0])
3260 				if ((newrule->tag =
3261 				    pf_tagname2tag(newrule->tagname)) == 0)
3262 					error = EBUSY;
3263 			if (newrule->match_tagname[0])
3264 				if ((newrule->match_tag = pf_tagname2tag(
3265 				    newrule->match_tagname)) == 0)
3266 					error = EBUSY;
3267 			if (newrule->rt && !newrule->direction)
3268 				error = EINVAL;
3269 			if (!newrule->log)
3270 				newrule->logif = 0;
3271 			if (newrule->logif >= PFLOGIFS_MAX)
3272 				error = EINVAL;
3273 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
3274 				error = ENOMEM;
3275 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
3276 				error = ENOMEM;
3277 			if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
3278 				error = EINVAL;
3279 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
3280 				if (pa->addr.type == PF_ADDR_TABLE) {
3281 					pa->addr.p.tbl =
3282 					    pfr_attach_table(ruleset,
3283 					    pa->addr.v.tblname);
3284 					if (pa->addr.p.tbl == NULL)
3285 						error = ENOMEM;
3286 				}
3287 
3288 			newrule->overload_tbl = NULL;
3289 			if (newrule->overload_tblname[0]) {
3290 				if ((newrule->overload_tbl = pfr_attach_table(
3291 				    ruleset, newrule->overload_tblname)) ==
3292 				    NULL)
3293 					error = EINVAL;
3294 				else
3295 					newrule->overload_tbl->pfrkt_flags |=
3296 					    PFR_TFLAG_ACTIVE;
3297 			}
3298 
3299 			pf_mv_kpool(&V_pf_pabuf, &newrule->rpool.list);
3300 			if (((((newrule->action == PF_NAT) ||
3301 			    (newrule->action == PF_RDR) ||
3302 			    (newrule->action == PF_BINAT) ||
3303 			    (newrule->rt > PF_NOPFROUTE)) &&
3304 			    !newrule->anchor)) &&
3305 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
3306 				error = EINVAL;
3307 
3308 			if (error) {
3309 				pf_free_rule(newrule);
3310 				PF_RULES_WUNLOCK();
3311 				break;
3312 			}
3313 
3314 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
3315 		}
3316 		pf_empty_kpool(&V_pf_pabuf);
3317 
3318 		if (pcr->action == PF_CHANGE_ADD_HEAD)
3319 			oldrule = TAILQ_FIRST(
3320 			    ruleset->rules[rs_num].active.ptr);
3321 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
3322 			oldrule = TAILQ_LAST(
3323 			    ruleset->rules[rs_num].active.ptr, pf_krulequeue);
3324 		else {
3325 			oldrule = TAILQ_FIRST(
3326 			    ruleset->rules[rs_num].active.ptr);
3327 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
3328 				oldrule = TAILQ_NEXT(oldrule, entries);
3329 			if (oldrule == NULL) {
3330 				if (newrule != NULL)
3331 					pf_free_rule(newrule);
3332 				PF_RULES_WUNLOCK();
3333 				error = EINVAL;
3334 				break;
3335 			}
3336 		}
3337 
3338 		if (pcr->action == PF_CHANGE_REMOVE) {
3339 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
3340 			    oldrule);
3341 			ruleset->rules[rs_num].active.rcount--;
3342 		} else {
3343 			if (oldrule == NULL)
3344 				TAILQ_INSERT_TAIL(
3345 				    ruleset->rules[rs_num].active.ptr,
3346 				    newrule, entries);
3347 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
3348 			    pcr->action == PF_CHANGE_ADD_BEFORE)
3349 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
3350 			else
3351 				TAILQ_INSERT_AFTER(
3352 				    ruleset->rules[rs_num].active.ptr,
3353 				    oldrule, newrule, entries);
3354 			ruleset->rules[rs_num].active.rcount++;
3355 		}
3356 
3357 		nr = 0;
3358 		TAILQ_FOREACH(oldrule,
3359 		    ruleset->rules[rs_num].active.ptr, entries)
3360 			oldrule->nr = nr++;
3361 
3362 		ruleset->rules[rs_num].active.ticket++;
3363 
3364 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
3365 		pf_remove_if_empty_kruleset(ruleset);
3366 
3367 		PF_RULES_WUNLOCK();
3368 		break;
3369 
3370 #undef ERROUT
3371 DIOCCHANGERULE_error:
3372 		PF_RULES_WUNLOCK();
3373 		pf_krule_free(newrule);
3374 		pf_kkif_free(kif);
3375 		break;
3376 	}
3377 
3378 	case DIOCCLRSTATES: {
3379 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
3380 		struct pf_kstate_kill	 kill;
3381 
3382 		error = pf_state_kill_to_kstate_kill(psk, &kill);
3383 		if (error)
3384 			break;
3385 
3386 		psk->psk_killed = pf_clear_states(&kill);
3387 		break;
3388 	}
3389 
3390 	case DIOCCLRSTATESNV: {
3391 		error = pf_clearstates_nv((struct pfioc_nv *)addr);
3392 		break;
3393 	}
3394 
3395 	case DIOCKILLSTATES: {
3396 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
3397 		struct pf_kstate_kill	 kill;
3398 
3399 		error = pf_state_kill_to_kstate_kill(psk, &kill);
3400 		if (error)
3401 			break;
3402 
3403 		psk->psk_killed = 0;
3404 		pf_killstates(&kill, &psk->psk_killed);
3405 		break;
3406 	}
3407 
3408 	case DIOCKILLSTATESNV: {
3409 		error = pf_killstates_nv((struct pfioc_nv *)addr);
3410 		break;
3411 	}
3412 
3413 	case DIOCADDSTATE: {
3414 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
3415 		struct pfsync_state	*sp = &ps->state;
3416 
3417 		if (sp->timeout >= PFTM_MAX) {
3418 			error = EINVAL;
3419 			break;
3420 		}
3421 		if (V_pfsync_state_import_ptr != NULL) {
3422 			PF_RULES_RLOCK();
3423 			error = V_pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
3424 			PF_RULES_RUNLOCK();
3425 		} else
3426 			error = EOPNOTSUPP;
3427 		break;
3428 	}
3429 
3430 	case DIOCGETSTATE: {
3431 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
3432 		struct pf_kstate	*s;
3433 
3434 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
3435 		if (s == NULL) {
3436 			error = ENOENT;
3437 			break;
3438 		}
3439 
3440 		pfsync_state_export(&ps->state, s);
3441 		PF_STATE_UNLOCK(s);
3442 		break;
3443 	}
3444 
3445 	case DIOCGETSTATENV: {
3446 		error = pf_getstate((struct pfioc_nv *)addr);
3447 		break;
3448 	}
3449 
3450 	case DIOCGETSTATES: {
3451 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
3452 		struct pf_kstate	*s;
3453 		struct pfsync_state	*pstore, *p;
3454 		int			 i, nr;
3455 		size_t			 slice_count = 16, count;
3456 		void			*out;
3457 
3458 		if (ps->ps_len <= 0) {
3459 			nr = uma_zone_get_cur(V_pf_state_z);
3460 			ps->ps_len = sizeof(struct pfsync_state) * nr;
3461 			break;
3462 		}
3463 
3464 		out = ps->ps_states;
3465 		pstore = mallocarray(slice_count,
3466 		    sizeof(struct pfsync_state), M_TEMP, M_WAITOK | M_ZERO);
3467 		nr = 0;
3468 
3469 		for (i = 0; i <= pf_hashmask; i++) {
3470 			struct pf_idhash *ih = &V_pf_idhash[i];
3471 
3472 DIOCGETSTATES_retry:
3473 			p = pstore;
3474 
3475 			if (LIST_EMPTY(&ih->states))
3476 				continue;
3477 
3478 			PF_HASHROW_LOCK(ih);
3479 			count = 0;
3480 			LIST_FOREACH(s, &ih->states, entry) {
3481 				if (s->timeout == PFTM_UNLINKED)
3482 					continue;
3483 				count++;
3484 			}
3485 
3486 			if (count > slice_count) {
3487 				PF_HASHROW_UNLOCK(ih);
3488 				free(pstore, M_TEMP);
3489 				slice_count = count * 2;
3490 				pstore = mallocarray(slice_count,
3491 				    sizeof(struct pfsync_state), M_TEMP,
3492 				    M_WAITOK | M_ZERO);
3493 				goto DIOCGETSTATES_retry;
3494 			}
3495 
3496 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
3497 				PF_HASHROW_UNLOCK(ih);
3498 				goto DIOCGETSTATES_full;
3499 			}
3500 
3501 			LIST_FOREACH(s, &ih->states, entry) {
3502 				if (s->timeout == PFTM_UNLINKED)
3503 					continue;
3504 
3505 				pfsync_state_export(p, s);
3506 				p++;
3507 				nr++;
3508 			}
3509 			PF_HASHROW_UNLOCK(ih);
3510 			error = copyout(pstore, out,
3511 			    sizeof(struct pfsync_state) * count);
3512 			if (error)
3513 				break;
3514 			out = ps->ps_states + nr;
3515 		}
3516 DIOCGETSTATES_full:
3517 		ps->ps_len = sizeof(struct pfsync_state) * nr;
3518 		free(pstore, M_TEMP);
3519 
3520 		break;
3521 	}
3522 
3523 	case DIOCGETSTATESV2: {
3524 		struct pfioc_states_v2	*ps = (struct pfioc_states_v2 *)addr;
3525 		struct pf_kstate	*s;
3526 		struct pf_state_export	*pstore, *p;
3527 		int i, nr;
3528 		size_t slice_count = 16, count;
3529 		void *out;
3530 
3531 		if (ps->ps_req_version > PF_STATE_VERSION) {
3532 			error = ENOTSUP;
3533 			break;
3534 		}
3535 
3536 		if (ps->ps_len <= 0) {
3537 			nr = uma_zone_get_cur(V_pf_state_z);
3538 			ps->ps_len = sizeof(struct pf_state_export) * nr;
3539 			break;
3540 		}
3541 
3542 		out = ps->ps_states;
3543 		pstore = mallocarray(slice_count,
3544 		    sizeof(struct pf_state_export), M_TEMP, M_WAITOK | M_ZERO);
3545 		nr = 0;
3546 
3547 		for (i = 0; i <= pf_hashmask; i++) {
3548 			struct pf_idhash *ih = &V_pf_idhash[i];
3549 
3550 DIOCGETSTATESV2_retry:
3551 			p = pstore;
3552 
3553 			if (LIST_EMPTY(&ih->states))
3554 				continue;
3555 
3556 			PF_HASHROW_LOCK(ih);
3557 			count = 0;
3558 			LIST_FOREACH(s, &ih->states, entry) {
3559 				if (s->timeout == PFTM_UNLINKED)
3560 					continue;
3561 				count++;
3562 			}
3563 
3564 			if (count > slice_count) {
3565 				PF_HASHROW_UNLOCK(ih);
3566 				free(pstore, M_TEMP);
3567 				slice_count = count * 2;
3568 				pstore = mallocarray(slice_count,
3569 				    sizeof(struct pf_state_export), M_TEMP,
3570 				    M_WAITOK | M_ZERO);
3571 				goto DIOCGETSTATESV2_retry;
3572 			}
3573 
3574 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
3575 				PF_HASHROW_UNLOCK(ih);
3576 				goto DIOCGETSTATESV2_full;
3577 			}
3578 
3579 			LIST_FOREACH(s, &ih->states, entry) {
3580 				if (s->timeout == PFTM_UNLINKED)
3581 					continue;
3582 
3583 				pf_state_export(p, s);
3584 				p++;
3585 				nr++;
3586 			}
3587 			PF_HASHROW_UNLOCK(ih);
3588 			error = copyout(pstore, out,
3589 			    sizeof(struct pf_state_export) * count);
3590 			if (error)
3591 				break;
3592 			out = ps->ps_states + nr;
3593 		}
3594 DIOCGETSTATESV2_full:
3595 		ps->ps_len = nr * sizeof(struct pf_state_export);
3596 		free(pstore, M_TEMP);
3597 
3598 		break;
3599 	}
3600 
3601 	case DIOCGETSTATUS: {
3602 		struct pf_status *s = (struct pf_status *)addr;
3603 
3604 		PF_RULES_RLOCK();
3605 		s->running = V_pf_status.running;
3606 		s->since   = V_pf_status.since;
3607 		s->debug   = V_pf_status.debug;
3608 		s->hostid  = V_pf_status.hostid;
3609 		s->states  = V_pf_status.states;
3610 		s->src_nodes = V_pf_status.src_nodes;
3611 
3612 		for (int i = 0; i < PFRES_MAX; i++)
3613 			s->counters[i] =
3614 			    counter_u64_fetch(V_pf_status.counters[i]);
3615 		for (int i = 0; i < LCNT_MAX; i++)
3616 			s->lcounters[i] =
3617 			    counter_u64_fetch(V_pf_status.lcounters[i]);
3618 		for (int i = 0; i < FCNT_MAX; i++)
3619 			s->fcounters[i] =
3620 			    pf_counter_u64_fetch(&V_pf_status.fcounters[i]);
3621 		for (int i = 0; i < SCNT_MAX; i++)
3622 			s->scounters[i] =
3623 			    counter_u64_fetch(V_pf_status.scounters[i]);
3624 
3625 		bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ);
3626 		bcopy(V_pf_status.pf_chksum, s->pf_chksum,
3627 		    PF_MD5_DIGEST_LENGTH);
3628 
3629 		pfi_update_status(s->ifname, s);
3630 		PF_RULES_RUNLOCK();
3631 		break;
3632 	}
3633 
3634 	case DIOCGETSTATUSNV: {
3635 		error = pf_getstatus((struct pfioc_nv *)addr);
3636 		break;
3637 	}
3638 
3639 	case DIOCSETSTATUSIF: {
3640 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
3641 
3642 		if (pi->ifname[0] == 0) {
3643 			bzero(V_pf_status.ifname, IFNAMSIZ);
3644 			break;
3645 		}
3646 		PF_RULES_WLOCK();
3647 		error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
3648 		PF_RULES_WUNLOCK();
3649 		break;
3650 	}
3651 
3652 	case DIOCCLRSTATUS: {
3653 		PF_RULES_WLOCK();
3654 		for (int i = 0; i < PFRES_MAX; i++)
3655 			counter_u64_zero(V_pf_status.counters[i]);
3656 		for (int i = 0; i < FCNT_MAX; i++)
3657 			pf_counter_u64_zero(&V_pf_status.fcounters[i]);
3658 		for (int i = 0; i < SCNT_MAX; i++)
3659 			counter_u64_zero(V_pf_status.scounters[i]);
3660 		for (int i = 0; i < KLCNT_MAX; i++)
3661 			counter_u64_zero(V_pf_status.lcounters[i]);
3662 		V_pf_status.since = time_second;
3663 		if (*V_pf_status.ifname)
3664 			pfi_update_status(V_pf_status.ifname, NULL);
3665 		PF_RULES_WUNLOCK();
3666 		break;
3667 	}
3668 
3669 	case DIOCNATLOOK: {
3670 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
3671 		struct pf_state_key	*sk;
3672 		struct pf_kstate	*state;
3673 		struct pf_state_key_cmp	 key;
3674 		int			 m = 0, direction = pnl->direction;
3675 		int			 sidx, didx;
3676 
3677 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
3678 		sidx = (direction == PF_IN) ? 1 : 0;
3679 		didx = (direction == PF_IN) ? 0 : 1;
3680 
3681 		if (!pnl->proto ||
3682 		    PF_AZERO(&pnl->saddr, pnl->af) ||
3683 		    PF_AZERO(&pnl->daddr, pnl->af) ||
3684 		    ((pnl->proto == IPPROTO_TCP ||
3685 		    pnl->proto == IPPROTO_UDP) &&
3686 		    (!pnl->dport || !pnl->sport)))
3687 			error = EINVAL;
3688 		else {
3689 			bzero(&key, sizeof(key));
3690 			key.af = pnl->af;
3691 			key.proto = pnl->proto;
3692 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
3693 			key.port[sidx] = pnl->sport;
3694 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
3695 			key.port[didx] = pnl->dport;
3696 
3697 			state = pf_find_state_all(&key, direction, &m);
3698 			if (state == NULL) {
3699 				error = ENOENT;
3700 			} else {
3701 				if (m > 1) {
3702 					PF_STATE_UNLOCK(state);
3703 					error = E2BIG;	/* more than one state */
3704 				} else {
3705 					sk = state->key[sidx];
3706 					PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
3707 					pnl->rsport = sk->port[sidx];
3708 					PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
3709 					pnl->rdport = sk->port[didx];
3710 					PF_STATE_UNLOCK(state);
3711 				}
3712 			}
3713 		}
3714 		break;
3715 	}
3716 
3717 	case DIOCSETTIMEOUT: {
3718 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
3719 		int		 old;
3720 
3721 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
3722 		    pt->seconds < 0) {
3723 			error = EINVAL;
3724 			break;
3725 		}
3726 		PF_RULES_WLOCK();
3727 		old = V_pf_default_rule.timeout[pt->timeout];
3728 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
3729 			pt->seconds = 1;
3730 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
3731 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
3732 			wakeup(pf_purge_thread);
3733 		pt->seconds = old;
3734 		PF_RULES_WUNLOCK();
3735 		break;
3736 	}
3737 
3738 	case DIOCGETTIMEOUT: {
3739 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
3740 
3741 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
3742 			error = EINVAL;
3743 			break;
3744 		}
3745 		PF_RULES_RLOCK();
3746 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
3747 		PF_RULES_RUNLOCK();
3748 		break;
3749 	}
3750 
3751 	case DIOCGETLIMIT: {
3752 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
3753 
3754 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
3755 			error = EINVAL;
3756 			break;
3757 		}
3758 		PF_RULES_RLOCK();
3759 		pl->limit = V_pf_limits[pl->index].limit;
3760 		PF_RULES_RUNLOCK();
3761 		break;
3762 	}
3763 
3764 	case DIOCSETLIMIT: {
3765 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
3766 		int			 old_limit;
3767 
3768 		PF_RULES_WLOCK();
3769 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
3770 		    V_pf_limits[pl->index].zone == NULL) {
3771 			PF_RULES_WUNLOCK();
3772 			error = EINVAL;
3773 			break;
3774 		}
3775 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
3776 		old_limit = V_pf_limits[pl->index].limit;
3777 		V_pf_limits[pl->index].limit = pl->limit;
3778 		pl->limit = old_limit;
3779 		PF_RULES_WUNLOCK();
3780 		break;
3781 	}
3782 
3783 	case DIOCSETDEBUG: {
3784 		u_int32_t	*level = (u_int32_t *)addr;
3785 
3786 		PF_RULES_WLOCK();
3787 		V_pf_status.debug = *level;
3788 		PF_RULES_WUNLOCK();
3789 		break;
3790 	}
3791 
3792 	case DIOCCLRRULECTRS: {
3793 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
3794 		struct pf_kruleset	*ruleset = &pf_main_ruleset;
3795 		struct pf_krule		*rule;
3796 
3797 		PF_RULES_WLOCK();
3798 		TAILQ_FOREACH(rule,
3799 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
3800 			pf_counter_u64_zero(&rule->evaluations);
3801 			for (int i = 0; i < 2; i++) {
3802 				pf_counter_u64_zero(&rule->packets[i]);
3803 				pf_counter_u64_zero(&rule->bytes[i]);
3804 			}
3805 		}
3806 		PF_RULES_WUNLOCK();
3807 		break;
3808 	}
3809 
3810 	case DIOCGIFSPEEDV0:
3811 	case DIOCGIFSPEEDV1: {
3812 		struct pf_ifspeed_v1	*psp = (struct pf_ifspeed_v1 *)addr;
3813 		struct pf_ifspeed_v1	ps;
3814 		struct ifnet		*ifp;
3815 
3816 		if (psp->ifname[0] == '\0') {
3817 			error = EINVAL;
3818 			break;
3819 		}
3820 
3821 		error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ);
3822 		if (error != 0)
3823 			break;
3824 		ifp = ifunit(ps.ifname);
3825 		if (ifp != NULL) {
3826 			psp->baudrate32 =
3827 			    (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
3828 			if (cmd == DIOCGIFSPEEDV1)
3829 				psp->baudrate = ifp->if_baudrate;
3830 		} else {
3831 			error = EINVAL;
3832 		}
3833 		break;
3834 	}
3835 
3836 #ifdef ALTQ
3837 	case DIOCSTARTALTQ: {
3838 		struct pf_altq		*altq;
3839 
3840 		PF_RULES_WLOCK();
3841 		/* enable all altq interfaces on active list */
3842 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
3843 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
3844 				error = pf_enable_altq(altq);
3845 				if (error != 0)
3846 					break;
3847 			}
3848 		}
3849 		if (error == 0)
3850 			V_pf_altq_running = 1;
3851 		PF_RULES_WUNLOCK();
3852 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
3853 		break;
3854 	}
3855 
3856 	case DIOCSTOPALTQ: {
3857 		struct pf_altq		*altq;
3858 
3859 		PF_RULES_WLOCK();
3860 		/* disable all altq interfaces on active list */
3861 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
3862 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
3863 				error = pf_disable_altq(altq);
3864 				if (error != 0)
3865 					break;
3866 			}
3867 		}
3868 		if (error == 0)
3869 			V_pf_altq_running = 0;
3870 		PF_RULES_WUNLOCK();
3871 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
3872 		break;
3873 	}
3874 
3875 	case DIOCADDALTQV0:
3876 	case DIOCADDALTQV1: {
3877 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
3878 		struct pf_altq		*altq, *a;
3879 		struct ifnet		*ifp;
3880 
3881 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
3882 		error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
3883 		if (error)
3884 			break;
3885 		altq->local_flags = 0;
3886 
3887 		PF_RULES_WLOCK();
3888 		if (pa->ticket != V_ticket_altqs_inactive) {
3889 			PF_RULES_WUNLOCK();
3890 			free(altq, M_PFALTQ);
3891 			error = EBUSY;
3892 			break;
3893 		}
3894 
3895 		/*
3896 		 * if this is for a queue, find the discipline and
3897 		 * copy the necessary fields
3898 		 */
3899 		if (altq->qname[0] != 0) {
3900 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
3901 				PF_RULES_WUNLOCK();
3902 				error = EBUSY;
3903 				free(altq, M_PFALTQ);
3904 				break;
3905 			}
3906 			altq->altq_disc = NULL;
3907 			TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
3908 				if (strncmp(a->ifname, altq->ifname,
3909 				    IFNAMSIZ) == 0) {
3910 					altq->altq_disc = a->altq_disc;
3911 					break;
3912 				}
3913 			}
3914 		}
3915 
3916 		if ((ifp = ifunit(altq->ifname)) == NULL)
3917 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
3918 		else
3919 			error = altq_add(ifp, altq);
3920 
3921 		if (error) {
3922 			PF_RULES_WUNLOCK();
3923 			free(altq, M_PFALTQ);
3924 			break;
3925 		}
3926 
3927 		if (altq->qname[0] != 0)
3928 			TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
3929 		else
3930 			TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
3931 		/* version error check done on import above */
3932 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
3933 		PF_RULES_WUNLOCK();
3934 		break;
3935 	}
3936 
3937 	case DIOCGETALTQSV0:
3938 	case DIOCGETALTQSV1: {
3939 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
3940 		struct pf_altq		*altq;
3941 
3942 		PF_RULES_RLOCK();
3943 		pa->nr = 0;
3944 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
3945 			pa->nr++;
3946 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
3947 			pa->nr++;
3948 		pa->ticket = V_ticket_altqs_active;
3949 		PF_RULES_RUNLOCK();
3950 		break;
3951 	}
3952 
3953 	case DIOCGETALTQV0:
3954 	case DIOCGETALTQV1: {
3955 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
3956 		struct pf_altq		*altq;
3957 
3958 		PF_RULES_RLOCK();
3959 		if (pa->ticket != V_ticket_altqs_active) {
3960 			PF_RULES_RUNLOCK();
3961 			error = EBUSY;
3962 			break;
3963 		}
3964 		altq = pf_altq_get_nth_active(pa->nr);
3965 		if (altq == NULL) {
3966 			PF_RULES_RUNLOCK();
3967 			error = EBUSY;
3968 			break;
3969 		}
3970 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
3971 		PF_RULES_RUNLOCK();
3972 		break;
3973 	}
3974 
3975 	case DIOCCHANGEALTQV0:
3976 	case DIOCCHANGEALTQV1:
3977 		/* CHANGEALTQ not supported yet! */
3978 		error = ENODEV;
3979 		break;
3980 
3981 	case DIOCGETQSTATSV0:
3982 	case DIOCGETQSTATSV1: {
3983 		struct pfioc_qstats_v1	*pq = (struct pfioc_qstats_v1 *)addr;
3984 		struct pf_altq		*altq;
3985 		int			 nbytes;
3986 		u_int32_t		 version;
3987 
3988 		PF_RULES_RLOCK();
3989 		if (pq->ticket != V_ticket_altqs_active) {
3990 			PF_RULES_RUNLOCK();
3991 			error = EBUSY;
3992 			break;
3993 		}
3994 		nbytes = pq->nbytes;
3995 		altq = pf_altq_get_nth_active(pq->nr);
3996 		if (altq == NULL) {
3997 			PF_RULES_RUNLOCK();
3998 			error = EBUSY;
3999 			break;
4000 		}
4001 
4002 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
4003 			PF_RULES_RUNLOCK();
4004 			error = ENXIO;
4005 			break;
4006 		}
4007 		PF_RULES_RUNLOCK();
4008 		if (cmd == DIOCGETQSTATSV0)
4009 			version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
4010 		else
4011 			version = pq->version;
4012 		error = altq_getqstats(altq, pq->buf, &nbytes, version);
4013 		if (error == 0) {
4014 			pq->scheduler = altq->scheduler;
4015 			pq->nbytes = nbytes;
4016 		}
4017 		break;
4018 	}
4019 #endif /* ALTQ */
4020 
4021 	case DIOCBEGINADDRS: {
4022 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4023 
4024 		PF_RULES_WLOCK();
4025 		pf_empty_kpool(&V_pf_pabuf);
4026 		pp->ticket = ++V_ticket_pabuf;
4027 		PF_RULES_WUNLOCK();
4028 		break;
4029 	}
4030 
4031 	case DIOCADDADDR: {
4032 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4033 		struct pf_kpooladdr	*pa;
4034 		struct pfi_kkif		*kif = NULL;
4035 
4036 #ifndef INET
4037 		if (pp->af == AF_INET) {
4038 			error = EAFNOSUPPORT;
4039 			break;
4040 		}
4041 #endif /* INET */
4042 #ifndef INET6
4043 		if (pp->af == AF_INET6) {
4044 			error = EAFNOSUPPORT;
4045 			break;
4046 		}
4047 #endif /* INET6 */
4048 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
4049 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
4050 		    pp->addr.addr.type != PF_ADDR_TABLE) {
4051 			error = EINVAL;
4052 			break;
4053 		}
4054 		if (pp->addr.addr.p.dyn != NULL) {
4055 			error = EINVAL;
4056 			break;
4057 		}
4058 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
4059 		error = pf_pooladdr_to_kpooladdr(&pp->addr, pa);
4060 		if (error != 0)
4061 			break;
4062 		if (pa->ifname[0])
4063 			kif = pf_kkif_create(M_WAITOK);
4064 		PF_RULES_WLOCK();
4065 		if (pp->ticket != V_ticket_pabuf) {
4066 			PF_RULES_WUNLOCK();
4067 			if (pa->ifname[0])
4068 				pf_kkif_free(kif);
4069 			free(pa, M_PFRULE);
4070 			error = EBUSY;
4071 			break;
4072 		}
4073 		if (pa->ifname[0]) {
4074 			pa->kif = pfi_kkif_attach(kif, pa->ifname);
4075 			kif = NULL;
4076 			pfi_kkif_ref(pa->kif);
4077 		} else
4078 			pa->kif = NULL;
4079 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
4080 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
4081 			if (pa->ifname[0])
4082 				pfi_kkif_unref(pa->kif);
4083 			PF_RULES_WUNLOCK();
4084 			free(pa, M_PFRULE);
4085 			break;
4086 		}
4087 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
4088 		PF_RULES_WUNLOCK();
4089 		break;
4090 	}
4091 
4092 	case DIOCGETADDRS: {
4093 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4094 		struct pf_kpool		*pool;
4095 		struct pf_kpooladdr	*pa;
4096 
4097 		pp->anchor[sizeof(pp->anchor) - 1] = 0;
4098 		pp->nr = 0;
4099 
4100 		PF_RULES_RLOCK();
4101 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
4102 		    pp->r_num, 0, 1, 0);
4103 		if (pool == NULL) {
4104 			PF_RULES_RUNLOCK();
4105 			error = EBUSY;
4106 			break;
4107 		}
4108 		TAILQ_FOREACH(pa, &pool->list, entries)
4109 			pp->nr++;
4110 		PF_RULES_RUNLOCK();
4111 		break;
4112 	}
4113 
4114 	case DIOCGETADDR: {
4115 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4116 		struct pf_kpool		*pool;
4117 		struct pf_kpooladdr	*pa;
4118 		u_int32_t		 nr = 0;
4119 
4120 		pp->anchor[sizeof(pp->anchor) - 1] = 0;
4121 
4122 		PF_RULES_RLOCK();
4123 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
4124 		    pp->r_num, 0, 1, 1);
4125 		if (pool == NULL) {
4126 			PF_RULES_RUNLOCK();
4127 			error = EBUSY;
4128 			break;
4129 		}
4130 		pa = TAILQ_FIRST(&pool->list);
4131 		while ((pa != NULL) && (nr < pp->nr)) {
4132 			pa = TAILQ_NEXT(pa, entries);
4133 			nr++;
4134 		}
4135 		if (pa == NULL) {
4136 			PF_RULES_RUNLOCK();
4137 			error = EBUSY;
4138 			break;
4139 		}
4140 		pf_kpooladdr_to_pooladdr(pa, &pp->addr);
4141 		pf_addr_copyout(&pp->addr.addr);
4142 		PF_RULES_RUNLOCK();
4143 		break;
4144 	}
4145 
4146 	case DIOCCHANGEADDR: {
4147 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
4148 		struct pf_kpool		*pool;
4149 		struct pf_kpooladdr	*oldpa = NULL, *newpa = NULL;
4150 		struct pf_kruleset	*ruleset;
4151 		struct pfi_kkif		*kif = NULL;
4152 
4153 		pca->anchor[sizeof(pca->anchor) - 1] = 0;
4154 
4155 		if (pca->action < PF_CHANGE_ADD_HEAD ||
4156 		    pca->action > PF_CHANGE_REMOVE) {
4157 			error = EINVAL;
4158 			break;
4159 		}
4160 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
4161 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
4162 		    pca->addr.addr.type != PF_ADDR_TABLE) {
4163 			error = EINVAL;
4164 			break;
4165 		}
4166 		if (pca->addr.addr.p.dyn != NULL) {
4167 			error = EINVAL;
4168 			break;
4169 		}
4170 
4171 		if (pca->action != PF_CHANGE_REMOVE) {
4172 #ifndef INET
4173 			if (pca->af == AF_INET) {
4174 				error = EAFNOSUPPORT;
4175 				break;
4176 			}
4177 #endif /* INET */
4178 #ifndef INET6
4179 			if (pca->af == AF_INET6) {
4180 				error = EAFNOSUPPORT;
4181 				break;
4182 			}
4183 #endif /* INET6 */
4184 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
4185 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
4186 			if (newpa->ifname[0])
4187 				kif = pf_kkif_create(M_WAITOK);
4188 			newpa->kif = NULL;
4189 		}
4190 #define	ERROUT(x)	ERROUT_IOCTL(DIOCCHANGEADDR_error, x)
4191 		PF_RULES_WLOCK();
4192 		ruleset = pf_find_kruleset(pca->anchor);
4193 		if (ruleset == NULL)
4194 			ERROUT(EBUSY);
4195 
4196 		pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
4197 		    pca->r_num, pca->r_last, 1, 1);
4198 		if (pool == NULL)
4199 			ERROUT(EBUSY);
4200 
4201 		if (pca->action != PF_CHANGE_REMOVE) {
4202 			if (newpa->ifname[0]) {
4203 				newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
4204 				pfi_kkif_ref(newpa->kif);
4205 				kif = NULL;
4206 			}
4207 
4208 			switch (newpa->addr.type) {
4209 			case PF_ADDR_DYNIFTL:
4210 				error = pfi_dynaddr_setup(&newpa->addr,
4211 				    pca->af);
4212 				break;
4213 			case PF_ADDR_TABLE:
4214 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
4215 				    newpa->addr.v.tblname);
4216 				if (newpa->addr.p.tbl == NULL)
4217 					error = ENOMEM;
4218 				break;
4219 			}
4220 			if (error)
4221 				goto DIOCCHANGEADDR_error;
4222 		}
4223 
4224 		switch (pca->action) {
4225 		case PF_CHANGE_ADD_HEAD:
4226 			oldpa = TAILQ_FIRST(&pool->list);
4227 			break;
4228 		case PF_CHANGE_ADD_TAIL:
4229 			oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
4230 			break;
4231 		default:
4232 			oldpa = TAILQ_FIRST(&pool->list);
4233 			for (int i = 0; oldpa && i < pca->nr; i++)
4234 				oldpa = TAILQ_NEXT(oldpa, entries);
4235 
4236 			if (oldpa == NULL)
4237 				ERROUT(EINVAL);
4238 		}
4239 
4240 		if (pca->action == PF_CHANGE_REMOVE) {
4241 			TAILQ_REMOVE(&pool->list, oldpa, entries);
4242 			switch (oldpa->addr.type) {
4243 			case PF_ADDR_DYNIFTL:
4244 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
4245 				break;
4246 			case PF_ADDR_TABLE:
4247 				pfr_detach_table(oldpa->addr.p.tbl);
4248 				break;
4249 			}
4250 			if (oldpa->kif)
4251 				pfi_kkif_unref(oldpa->kif);
4252 			free(oldpa, M_PFRULE);
4253 		} else {
4254 			if (oldpa == NULL)
4255 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
4256 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
4257 			    pca->action == PF_CHANGE_ADD_BEFORE)
4258 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
4259 			else
4260 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
4261 				    newpa, entries);
4262 		}
4263 
4264 		pool->cur = TAILQ_FIRST(&pool->list);
4265 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
4266 		PF_RULES_WUNLOCK();
4267 		break;
4268 
4269 #undef ERROUT
4270 DIOCCHANGEADDR_error:
4271 		if (newpa != NULL) {
4272 			if (newpa->kif)
4273 				pfi_kkif_unref(newpa->kif);
4274 			free(newpa, M_PFRULE);
4275 		}
4276 		PF_RULES_WUNLOCK();
4277 		pf_kkif_free(kif);
4278 		break;
4279 	}
4280 
4281 	case DIOCGETRULESETS: {
4282 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
4283 		struct pf_kruleset	*ruleset;
4284 		struct pf_kanchor	*anchor;
4285 
4286 		pr->path[sizeof(pr->path) - 1] = 0;
4287 
4288 		PF_RULES_RLOCK();
4289 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4290 			PF_RULES_RUNLOCK();
4291 			error = ENOENT;
4292 			break;
4293 		}
4294 		pr->nr = 0;
4295 		if (ruleset->anchor == NULL) {
4296 			/* XXX kludge for pf_main_ruleset */
4297 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4298 				if (anchor->parent == NULL)
4299 					pr->nr++;
4300 		} else {
4301 			RB_FOREACH(anchor, pf_kanchor_node,
4302 			    &ruleset->anchor->children)
4303 				pr->nr++;
4304 		}
4305 		PF_RULES_RUNLOCK();
4306 		break;
4307 	}
4308 
4309 	case DIOCGETRULESET: {
4310 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
4311 		struct pf_kruleset	*ruleset;
4312 		struct pf_kanchor	*anchor;
4313 		u_int32_t		 nr = 0;
4314 
4315 		pr->path[sizeof(pr->path) - 1] = 0;
4316 
4317 		PF_RULES_RLOCK();
4318 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4319 			PF_RULES_RUNLOCK();
4320 			error = ENOENT;
4321 			break;
4322 		}
4323 		pr->name[0] = 0;
4324 		if (ruleset->anchor == NULL) {
4325 			/* XXX kludge for pf_main_ruleset */
4326 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4327 				if (anchor->parent == NULL && nr++ == pr->nr) {
4328 					strlcpy(pr->name, anchor->name,
4329 					    sizeof(pr->name));
4330 					break;
4331 				}
4332 		} else {
4333 			RB_FOREACH(anchor, pf_kanchor_node,
4334 			    &ruleset->anchor->children)
4335 				if (nr++ == pr->nr) {
4336 					strlcpy(pr->name, anchor->name,
4337 					    sizeof(pr->name));
4338 					break;
4339 				}
4340 		}
4341 		if (!pr->name[0])
4342 			error = EBUSY;
4343 		PF_RULES_RUNLOCK();
4344 		break;
4345 	}
4346 
4347 	case DIOCRCLRTABLES: {
4348 		struct pfioc_table *io = (struct pfioc_table *)addr;
4349 
4350 		if (io->pfrio_esize != 0) {
4351 			error = ENODEV;
4352 			break;
4353 		}
4354 		PF_RULES_WLOCK();
4355 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
4356 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4357 		PF_RULES_WUNLOCK();
4358 		break;
4359 	}
4360 
4361 	case DIOCRADDTABLES: {
4362 		struct pfioc_table *io = (struct pfioc_table *)addr;
4363 		struct pfr_table *pfrts;
4364 		size_t totlen;
4365 
4366 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4367 			error = ENODEV;
4368 			break;
4369 		}
4370 
4371 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4372 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4373 			error = ENOMEM;
4374 			break;
4375 		}
4376 
4377 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4378 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4379 		    M_TEMP, M_WAITOK);
4380 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4381 		if (error) {
4382 			free(pfrts, M_TEMP);
4383 			break;
4384 		}
4385 		PF_RULES_WLOCK();
4386 		error = pfr_add_tables(pfrts, io->pfrio_size,
4387 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4388 		PF_RULES_WUNLOCK();
4389 		free(pfrts, M_TEMP);
4390 		break;
4391 	}
4392 
4393 	case DIOCRDELTABLES: {
4394 		struct pfioc_table *io = (struct pfioc_table *)addr;
4395 		struct pfr_table *pfrts;
4396 		size_t totlen;
4397 
4398 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4399 			error = ENODEV;
4400 			break;
4401 		}
4402 
4403 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4404 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4405 			error = ENOMEM;
4406 			break;
4407 		}
4408 
4409 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4410 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4411 		    M_TEMP, M_WAITOK);
4412 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4413 		if (error) {
4414 			free(pfrts, M_TEMP);
4415 			break;
4416 		}
4417 		PF_RULES_WLOCK();
4418 		error = pfr_del_tables(pfrts, io->pfrio_size,
4419 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4420 		PF_RULES_WUNLOCK();
4421 		free(pfrts, M_TEMP);
4422 		break;
4423 	}
4424 
4425 	case DIOCRGETTABLES: {
4426 		struct pfioc_table *io = (struct pfioc_table *)addr;
4427 		struct pfr_table *pfrts;
4428 		size_t totlen;
4429 		int n;
4430 
4431 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4432 			error = ENODEV;
4433 			break;
4434 		}
4435 		PF_RULES_RLOCK();
4436 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4437 		if (n < 0) {
4438 			PF_RULES_RUNLOCK();
4439 			error = EINVAL;
4440 			break;
4441 		}
4442 		io->pfrio_size = min(io->pfrio_size, n);
4443 
4444 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4445 
4446 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4447 		    M_TEMP, M_NOWAIT | M_ZERO);
4448 		if (pfrts == NULL) {
4449 			error = ENOMEM;
4450 			PF_RULES_RUNLOCK();
4451 			break;
4452 		}
4453 		error = pfr_get_tables(&io->pfrio_table, pfrts,
4454 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4455 		PF_RULES_RUNLOCK();
4456 		if (error == 0)
4457 			error = copyout(pfrts, io->pfrio_buffer, totlen);
4458 		free(pfrts, M_TEMP);
4459 		break;
4460 	}
4461 
4462 	case DIOCRGETTSTATS: {
4463 		struct pfioc_table *io = (struct pfioc_table *)addr;
4464 		struct pfr_tstats *pfrtstats;
4465 		size_t totlen;
4466 		int n;
4467 
4468 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
4469 			error = ENODEV;
4470 			break;
4471 		}
4472 		PF_TABLE_STATS_LOCK();
4473 		PF_RULES_RLOCK();
4474 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4475 		if (n < 0) {
4476 			PF_RULES_RUNLOCK();
4477 			PF_TABLE_STATS_UNLOCK();
4478 			error = EINVAL;
4479 			break;
4480 		}
4481 		io->pfrio_size = min(io->pfrio_size, n);
4482 
4483 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
4484 		pfrtstats = mallocarray(io->pfrio_size,
4485 		    sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT | M_ZERO);
4486 		if (pfrtstats == NULL) {
4487 			error = ENOMEM;
4488 			PF_RULES_RUNLOCK();
4489 			PF_TABLE_STATS_UNLOCK();
4490 			break;
4491 		}
4492 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
4493 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4494 		PF_RULES_RUNLOCK();
4495 		PF_TABLE_STATS_UNLOCK();
4496 		if (error == 0)
4497 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
4498 		free(pfrtstats, M_TEMP);
4499 		break;
4500 	}
4501 
4502 	case DIOCRCLRTSTATS: {
4503 		struct pfioc_table *io = (struct pfioc_table *)addr;
4504 		struct pfr_table *pfrts;
4505 		size_t totlen;
4506 
4507 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4508 			error = ENODEV;
4509 			break;
4510 		}
4511 
4512 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4513 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4514 			/* We used to count tables and use the minimum required
4515 			 * size, so we didn't fail on overly large requests.
4516 			 * Keep doing so. */
4517 			io->pfrio_size = pf_ioctl_maxcount;
4518 			break;
4519 		}
4520 
4521 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4522 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4523 		    M_TEMP, M_WAITOK);
4524 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4525 		if (error) {
4526 			free(pfrts, M_TEMP);
4527 			break;
4528 		}
4529 
4530 		PF_TABLE_STATS_LOCK();
4531 		PF_RULES_RLOCK();
4532 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
4533 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4534 		PF_RULES_RUNLOCK();
4535 		PF_TABLE_STATS_UNLOCK();
4536 		free(pfrts, M_TEMP);
4537 		break;
4538 	}
4539 
4540 	case DIOCRSETTFLAGS: {
4541 		struct pfioc_table *io = (struct pfioc_table *)addr;
4542 		struct pfr_table *pfrts;
4543 		size_t totlen;
4544 		int n;
4545 
4546 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4547 			error = ENODEV;
4548 			break;
4549 		}
4550 
4551 		PF_RULES_RLOCK();
4552 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4553 		if (n < 0) {
4554 			PF_RULES_RUNLOCK();
4555 			error = EINVAL;
4556 			break;
4557 		}
4558 
4559 		io->pfrio_size = min(io->pfrio_size, n);
4560 		PF_RULES_RUNLOCK();
4561 
4562 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4563 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4564 		    M_TEMP, M_WAITOK);
4565 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4566 		if (error) {
4567 			free(pfrts, M_TEMP);
4568 			break;
4569 		}
4570 		PF_RULES_WLOCK();
4571 		error = pfr_set_tflags(pfrts, io->pfrio_size,
4572 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
4573 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4574 		PF_RULES_WUNLOCK();
4575 		free(pfrts, M_TEMP);
4576 		break;
4577 	}
4578 
4579 	case DIOCRCLRADDRS: {
4580 		struct pfioc_table *io = (struct pfioc_table *)addr;
4581 
4582 		if (io->pfrio_esize != 0) {
4583 			error = ENODEV;
4584 			break;
4585 		}
4586 		PF_RULES_WLOCK();
4587 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
4588 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4589 		PF_RULES_WUNLOCK();
4590 		break;
4591 	}
4592 
4593 	case DIOCRADDADDRS: {
4594 		struct pfioc_table *io = (struct pfioc_table *)addr;
4595 		struct pfr_addr *pfras;
4596 		size_t totlen;
4597 
4598 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4599 			error = ENODEV;
4600 			break;
4601 		}
4602 		if (io->pfrio_size < 0 ||
4603 		    io->pfrio_size > pf_ioctl_maxcount ||
4604 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4605 			error = EINVAL;
4606 			break;
4607 		}
4608 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4609 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4610 		    M_TEMP, M_WAITOK);
4611 		error = copyin(io->pfrio_buffer, pfras, totlen);
4612 		if (error) {
4613 			free(pfras, M_TEMP);
4614 			break;
4615 		}
4616 		PF_RULES_WLOCK();
4617 		error = pfr_add_addrs(&io->pfrio_table, pfras,
4618 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
4619 		    PFR_FLAG_USERIOCTL);
4620 		PF_RULES_WUNLOCK();
4621 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4622 			error = copyout(pfras, io->pfrio_buffer, totlen);
4623 		free(pfras, M_TEMP);
4624 		break;
4625 	}
4626 
4627 	case DIOCRDELADDRS: {
4628 		struct pfioc_table *io = (struct pfioc_table *)addr;
4629 		struct pfr_addr *pfras;
4630 		size_t totlen;
4631 
4632 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4633 			error = ENODEV;
4634 			break;
4635 		}
4636 		if (io->pfrio_size < 0 ||
4637 		    io->pfrio_size > pf_ioctl_maxcount ||
4638 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4639 			error = EINVAL;
4640 			break;
4641 		}
4642 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4643 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4644 		    M_TEMP, M_WAITOK);
4645 		error = copyin(io->pfrio_buffer, pfras, totlen);
4646 		if (error) {
4647 			free(pfras, M_TEMP);
4648 			break;
4649 		}
4650 		PF_RULES_WLOCK();
4651 		error = pfr_del_addrs(&io->pfrio_table, pfras,
4652 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
4653 		    PFR_FLAG_USERIOCTL);
4654 		PF_RULES_WUNLOCK();
4655 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4656 			error = copyout(pfras, io->pfrio_buffer, totlen);
4657 		free(pfras, M_TEMP);
4658 		break;
4659 	}
4660 
4661 	case DIOCRSETADDRS: {
4662 		struct pfioc_table *io = (struct pfioc_table *)addr;
4663 		struct pfr_addr *pfras;
4664 		size_t totlen, count;
4665 
4666 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4667 			error = ENODEV;
4668 			break;
4669 		}
4670 		if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
4671 			error = EINVAL;
4672 			break;
4673 		}
4674 		count = max(io->pfrio_size, io->pfrio_size2);
4675 		if (count > pf_ioctl_maxcount ||
4676 		    WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
4677 			error = EINVAL;
4678 			break;
4679 		}
4680 		totlen = count * sizeof(struct pfr_addr);
4681 		pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
4682 		    M_WAITOK);
4683 		error = copyin(io->pfrio_buffer, pfras, totlen);
4684 		if (error) {
4685 			free(pfras, M_TEMP);
4686 			break;
4687 		}
4688 		PF_RULES_WLOCK();
4689 		error = pfr_set_addrs(&io->pfrio_table, pfras,
4690 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
4691 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
4692 		    PFR_FLAG_USERIOCTL, 0);
4693 		PF_RULES_WUNLOCK();
4694 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4695 			error = copyout(pfras, io->pfrio_buffer, totlen);
4696 		free(pfras, M_TEMP);
4697 		break;
4698 	}
4699 
4700 	case DIOCRGETADDRS: {
4701 		struct pfioc_table *io = (struct pfioc_table *)addr;
4702 		struct pfr_addr *pfras;
4703 		size_t totlen;
4704 
4705 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4706 			error = ENODEV;
4707 			break;
4708 		}
4709 		if (io->pfrio_size < 0 ||
4710 		    io->pfrio_size > pf_ioctl_maxcount ||
4711 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4712 			error = EINVAL;
4713 			break;
4714 		}
4715 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4716 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4717 		    M_TEMP, M_WAITOK | M_ZERO);
4718 		PF_RULES_RLOCK();
4719 		error = pfr_get_addrs(&io->pfrio_table, pfras,
4720 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4721 		PF_RULES_RUNLOCK();
4722 		if (error == 0)
4723 			error = copyout(pfras, io->pfrio_buffer, totlen);
4724 		free(pfras, M_TEMP);
4725 		break;
4726 	}
4727 
4728 	case DIOCRGETASTATS: {
4729 		struct pfioc_table *io = (struct pfioc_table *)addr;
4730 		struct pfr_astats *pfrastats;
4731 		size_t totlen;
4732 
4733 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
4734 			error = ENODEV;
4735 			break;
4736 		}
4737 		if (io->pfrio_size < 0 ||
4738 		    io->pfrio_size > pf_ioctl_maxcount ||
4739 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
4740 			error = EINVAL;
4741 			break;
4742 		}
4743 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
4744 		pfrastats = mallocarray(io->pfrio_size,
4745 		    sizeof(struct pfr_astats), M_TEMP, M_WAITOK | M_ZERO);
4746 		PF_RULES_RLOCK();
4747 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
4748 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4749 		PF_RULES_RUNLOCK();
4750 		if (error == 0)
4751 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
4752 		free(pfrastats, M_TEMP);
4753 		break;
4754 	}
4755 
4756 	case DIOCRCLRASTATS: {
4757 		struct pfioc_table *io = (struct pfioc_table *)addr;
4758 		struct pfr_addr *pfras;
4759 		size_t totlen;
4760 
4761 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4762 			error = ENODEV;
4763 			break;
4764 		}
4765 		if (io->pfrio_size < 0 ||
4766 		    io->pfrio_size > pf_ioctl_maxcount ||
4767 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4768 			error = EINVAL;
4769 			break;
4770 		}
4771 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4772 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4773 		    M_TEMP, M_WAITOK);
4774 		error = copyin(io->pfrio_buffer, pfras, totlen);
4775 		if (error) {
4776 			free(pfras, M_TEMP);
4777 			break;
4778 		}
4779 		PF_RULES_WLOCK();
4780 		error = pfr_clr_astats(&io->pfrio_table, pfras,
4781 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
4782 		    PFR_FLAG_USERIOCTL);
4783 		PF_RULES_WUNLOCK();
4784 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4785 			error = copyout(pfras, io->pfrio_buffer, totlen);
4786 		free(pfras, M_TEMP);
4787 		break;
4788 	}
4789 
4790 	case DIOCRTSTADDRS: {
4791 		struct pfioc_table *io = (struct pfioc_table *)addr;
4792 		struct pfr_addr *pfras;
4793 		size_t totlen;
4794 
4795 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4796 			error = ENODEV;
4797 			break;
4798 		}
4799 		if (io->pfrio_size < 0 ||
4800 		    io->pfrio_size > pf_ioctl_maxcount ||
4801 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4802 			error = EINVAL;
4803 			break;
4804 		}
4805 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4806 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4807 		    M_TEMP, M_WAITOK);
4808 		error = copyin(io->pfrio_buffer, pfras, totlen);
4809 		if (error) {
4810 			free(pfras, M_TEMP);
4811 			break;
4812 		}
4813 		PF_RULES_RLOCK();
4814 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
4815 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
4816 		    PFR_FLAG_USERIOCTL);
4817 		PF_RULES_RUNLOCK();
4818 		if (error == 0)
4819 			error = copyout(pfras, io->pfrio_buffer, totlen);
4820 		free(pfras, M_TEMP);
4821 		break;
4822 	}
4823 
4824 	case DIOCRINADEFINE: {
4825 		struct pfioc_table *io = (struct pfioc_table *)addr;
4826 		struct pfr_addr *pfras;
4827 		size_t totlen;
4828 
4829 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4830 			error = ENODEV;
4831 			break;
4832 		}
4833 		if (io->pfrio_size < 0 ||
4834 		    io->pfrio_size > pf_ioctl_maxcount ||
4835 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4836 			error = EINVAL;
4837 			break;
4838 		}
4839 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4840 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4841 		    M_TEMP, M_WAITOK);
4842 		error = copyin(io->pfrio_buffer, pfras, totlen);
4843 		if (error) {
4844 			free(pfras, M_TEMP);
4845 			break;
4846 		}
4847 		PF_RULES_WLOCK();
4848 		error = pfr_ina_define(&io->pfrio_table, pfras,
4849 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
4850 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4851 		PF_RULES_WUNLOCK();
4852 		free(pfras, M_TEMP);
4853 		break;
4854 	}
4855 
4856 	case DIOCOSFPADD: {
4857 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
4858 		PF_RULES_WLOCK();
4859 		error = pf_osfp_add(io);
4860 		PF_RULES_WUNLOCK();
4861 		break;
4862 	}
4863 
4864 	case DIOCOSFPGET: {
4865 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
4866 		PF_RULES_RLOCK();
4867 		error = pf_osfp_get(io);
4868 		PF_RULES_RUNLOCK();
4869 		break;
4870 	}
4871 
4872 	case DIOCXBEGIN: {
4873 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
4874 		struct pfioc_trans_e	*ioes, *ioe;
4875 		size_t			 totlen;
4876 		int			 i;
4877 
4878 		if (io->esize != sizeof(*ioe)) {
4879 			error = ENODEV;
4880 			break;
4881 		}
4882 		if (io->size < 0 ||
4883 		    io->size > pf_ioctl_maxcount ||
4884 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
4885 			error = EINVAL;
4886 			break;
4887 		}
4888 		totlen = sizeof(struct pfioc_trans_e) * io->size;
4889 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
4890 		    M_TEMP, M_WAITOK);
4891 		error = copyin(io->array, ioes, totlen);
4892 		if (error) {
4893 			free(ioes, M_TEMP);
4894 			break;
4895 		}
4896 		PF_RULES_WLOCK();
4897 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
4898 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
4899 			switch (ioe->rs_num) {
4900 			case PF_RULESET_ETH:
4901 				if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) {
4902 					PF_RULES_WUNLOCK();
4903 					free(ioes, M_TEMP);
4904 					goto fail;
4905 				}
4906 				break;
4907 #ifdef ALTQ
4908 			case PF_RULESET_ALTQ:
4909 				if (ioe->anchor[0]) {
4910 					PF_RULES_WUNLOCK();
4911 					free(ioes, M_TEMP);
4912 					error = EINVAL;
4913 					goto fail;
4914 				}
4915 				if ((error = pf_begin_altq(&ioe->ticket))) {
4916 					PF_RULES_WUNLOCK();
4917 					free(ioes, M_TEMP);
4918 					goto fail;
4919 				}
4920 				break;
4921 #endif /* ALTQ */
4922 			case PF_RULESET_TABLE:
4923 			    {
4924 				struct pfr_table table;
4925 
4926 				bzero(&table, sizeof(table));
4927 				strlcpy(table.pfrt_anchor, ioe->anchor,
4928 				    sizeof(table.pfrt_anchor));
4929 				if ((error = pfr_ina_begin(&table,
4930 				    &ioe->ticket, NULL, 0))) {
4931 					PF_RULES_WUNLOCK();
4932 					free(ioes, M_TEMP);
4933 					goto fail;
4934 				}
4935 				break;
4936 			    }
4937 			default:
4938 				if ((error = pf_begin_rules(&ioe->ticket,
4939 				    ioe->rs_num, ioe->anchor))) {
4940 					PF_RULES_WUNLOCK();
4941 					free(ioes, M_TEMP);
4942 					goto fail;
4943 				}
4944 				break;
4945 			}
4946 		}
4947 		PF_RULES_WUNLOCK();
4948 		error = copyout(ioes, io->array, totlen);
4949 		free(ioes, M_TEMP);
4950 		break;
4951 	}
4952 
4953 	case DIOCXROLLBACK: {
4954 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
4955 		struct pfioc_trans_e	*ioe, *ioes;
4956 		size_t			 totlen;
4957 		int			 i;
4958 
4959 		if (io->esize != sizeof(*ioe)) {
4960 			error = ENODEV;
4961 			break;
4962 		}
4963 		if (io->size < 0 ||
4964 		    io->size > pf_ioctl_maxcount ||
4965 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
4966 			error = EINVAL;
4967 			break;
4968 		}
4969 		totlen = sizeof(struct pfioc_trans_e) * io->size;
4970 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
4971 		    M_TEMP, M_WAITOK);
4972 		error = copyin(io->array, ioes, totlen);
4973 		if (error) {
4974 			free(ioes, M_TEMP);
4975 			break;
4976 		}
4977 		PF_RULES_WLOCK();
4978 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
4979 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
4980 			switch (ioe->rs_num) {
4981 			case PF_RULESET_ETH:
4982 				if ((error = pf_rollback_eth(ioe->ticket,
4983 				    ioe->anchor))) {
4984 					PF_RULES_WUNLOCK();
4985 					free(ioes, M_TEMP);
4986 					goto fail; /* really bad */
4987 				}
4988 				break;
4989 #ifdef ALTQ
4990 			case PF_RULESET_ALTQ:
4991 				if (ioe->anchor[0]) {
4992 					PF_RULES_WUNLOCK();
4993 					free(ioes, M_TEMP);
4994 					error = EINVAL;
4995 					goto fail;
4996 				}
4997 				if ((error = pf_rollback_altq(ioe->ticket))) {
4998 					PF_RULES_WUNLOCK();
4999 					free(ioes, M_TEMP);
5000 					goto fail; /* really bad */
5001 				}
5002 				break;
5003 #endif /* ALTQ */
5004 			case PF_RULESET_TABLE:
5005 			    {
5006 				struct pfr_table table;
5007 
5008 				bzero(&table, sizeof(table));
5009 				strlcpy(table.pfrt_anchor, ioe->anchor,
5010 				    sizeof(table.pfrt_anchor));
5011 				if ((error = pfr_ina_rollback(&table,
5012 				    ioe->ticket, NULL, 0))) {
5013 					PF_RULES_WUNLOCK();
5014 					free(ioes, M_TEMP);
5015 					goto fail; /* really bad */
5016 				}
5017 				break;
5018 			    }
5019 			default:
5020 				if ((error = pf_rollback_rules(ioe->ticket,
5021 				    ioe->rs_num, ioe->anchor))) {
5022 					PF_RULES_WUNLOCK();
5023 					free(ioes, M_TEMP);
5024 					goto fail; /* really bad */
5025 				}
5026 				break;
5027 			}
5028 		}
5029 		PF_RULES_WUNLOCK();
5030 		free(ioes, M_TEMP);
5031 		break;
5032 	}
5033 
5034 	case DIOCXCOMMIT: {
5035 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5036 		struct pfioc_trans_e	*ioe, *ioes;
5037 		struct pf_kruleset	*rs;
5038 		struct pf_keth_ruleset	*ers;
5039 		size_t			 totlen;
5040 		int			 i;
5041 
5042 		if (io->esize != sizeof(*ioe)) {
5043 			error = ENODEV;
5044 			break;
5045 		}
5046 
5047 		if (io->size < 0 ||
5048 		    io->size > pf_ioctl_maxcount ||
5049 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5050 			error = EINVAL;
5051 			break;
5052 		}
5053 
5054 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5055 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5056 		    M_TEMP, M_WAITOK);
5057 		error = copyin(io->array, ioes, totlen);
5058 		if (error) {
5059 			free(ioes, M_TEMP);
5060 			break;
5061 		}
5062 		PF_RULES_WLOCK();
5063 		/* First makes sure everything will succeed. */
5064 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5065 			ioe->anchor[sizeof(ioe->anchor) - 1] = 0;
5066 			switch (ioe->rs_num) {
5067 			case PF_RULESET_ETH:
5068 				ers = pf_find_keth_ruleset(ioe->anchor);
5069 				if (ers == NULL || ioe->ticket == 0 ||
5070 				    ioe->ticket != ers->inactive.ticket) {
5071 					PF_RULES_WUNLOCK();
5072 					free(ioes, M_TEMP);
5073 					error = EINVAL;
5074 					goto fail;
5075 				}
5076 				break;
5077 #ifdef ALTQ
5078 			case PF_RULESET_ALTQ:
5079 				if (ioe->anchor[0]) {
5080 					PF_RULES_WUNLOCK();
5081 					free(ioes, M_TEMP);
5082 					error = EINVAL;
5083 					goto fail;
5084 				}
5085 				if (!V_altqs_inactive_open || ioe->ticket !=
5086 				    V_ticket_altqs_inactive) {
5087 					PF_RULES_WUNLOCK();
5088 					free(ioes, M_TEMP);
5089 					error = EBUSY;
5090 					goto fail;
5091 				}
5092 				break;
5093 #endif /* ALTQ */
5094 			case PF_RULESET_TABLE:
5095 				rs = pf_find_kruleset(ioe->anchor);
5096 				if (rs == NULL || !rs->topen || ioe->ticket !=
5097 				    rs->tticket) {
5098 					PF_RULES_WUNLOCK();
5099 					free(ioes, M_TEMP);
5100 					error = EBUSY;
5101 					goto fail;
5102 				}
5103 				break;
5104 			default:
5105 				if (ioe->rs_num < 0 || ioe->rs_num >=
5106 				    PF_RULESET_MAX) {
5107 					PF_RULES_WUNLOCK();
5108 					free(ioes, M_TEMP);
5109 					error = EINVAL;
5110 					goto fail;
5111 				}
5112 				rs = pf_find_kruleset(ioe->anchor);
5113 				if (rs == NULL ||
5114 				    !rs->rules[ioe->rs_num].inactive.open ||
5115 				    rs->rules[ioe->rs_num].inactive.ticket !=
5116 				    ioe->ticket) {
5117 					PF_RULES_WUNLOCK();
5118 					free(ioes, M_TEMP);
5119 					error = EBUSY;
5120 					goto fail;
5121 				}
5122 				break;
5123 			}
5124 		}
5125 		/* Now do the commit - no errors should happen here. */
5126 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5127 			switch (ioe->rs_num) {
5128 			case PF_RULESET_ETH:
5129 				if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) {
5130 					PF_RULES_WUNLOCK();
5131 					free(ioes, M_TEMP);
5132 					goto fail; /* really bad */
5133 				}
5134 				break;
5135 #ifdef ALTQ
5136 			case PF_RULESET_ALTQ:
5137 				if ((error = pf_commit_altq(ioe->ticket))) {
5138 					PF_RULES_WUNLOCK();
5139 					free(ioes, M_TEMP);
5140 					goto fail; /* really bad */
5141 				}
5142 				break;
5143 #endif /* ALTQ */
5144 			case PF_RULESET_TABLE:
5145 			    {
5146 				struct pfr_table table;
5147 
5148 				bzero(&table, sizeof(table));
5149 				(void)strlcpy(table.pfrt_anchor, ioe->anchor,
5150 				    sizeof(table.pfrt_anchor));
5151 				if ((error = pfr_ina_commit(&table,
5152 				    ioe->ticket, NULL, NULL, 0))) {
5153 					PF_RULES_WUNLOCK();
5154 					free(ioes, M_TEMP);
5155 					goto fail; /* really bad */
5156 				}
5157 				break;
5158 			    }
5159 			default:
5160 				if ((error = pf_commit_rules(ioe->ticket,
5161 				    ioe->rs_num, ioe->anchor))) {
5162 					PF_RULES_WUNLOCK();
5163 					free(ioes, M_TEMP);
5164 					goto fail; /* really bad */
5165 				}
5166 				break;
5167 			}
5168 		}
5169 		PF_RULES_WUNLOCK();
5170 
5171 		/* Only hook into EtherNet taffic if we've got rules for it. */
5172 		if (! TAILQ_EMPTY(V_pf_keth->active.rules))
5173 			hook_pf_eth();
5174 		else
5175 			dehook_pf_eth();
5176 
5177 		free(ioes, M_TEMP);
5178 		break;
5179 	}
5180 
5181 	case DIOCGETSRCNODES: {
5182 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
5183 		struct pf_srchash	*sh;
5184 		struct pf_ksrc_node	*n;
5185 		struct pf_src_node	*p, *pstore;
5186 		uint32_t		 i, nr = 0;
5187 
5188 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5189 				i++, sh++) {
5190 			PF_HASHROW_LOCK(sh);
5191 			LIST_FOREACH(n, &sh->nodes, entry)
5192 				nr++;
5193 			PF_HASHROW_UNLOCK(sh);
5194 		}
5195 
5196 		psn->psn_len = min(psn->psn_len,
5197 		    sizeof(struct pf_src_node) * nr);
5198 
5199 		if (psn->psn_len == 0) {
5200 			psn->psn_len = sizeof(struct pf_src_node) * nr;
5201 			break;
5202 		}
5203 
5204 		nr = 0;
5205 
5206 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
5207 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5208 		    i++, sh++) {
5209 		    PF_HASHROW_LOCK(sh);
5210 		    LIST_FOREACH(n, &sh->nodes, entry) {
5211 
5212 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
5213 				break;
5214 
5215 			pf_src_node_copy(n, p);
5216 
5217 			p++;
5218 			nr++;
5219 		    }
5220 		    PF_HASHROW_UNLOCK(sh);
5221 		}
5222 		error = copyout(pstore, psn->psn_src_nodes,
5223 		    sizeof(struct pf_src_node) * nr);
5224 		if (error) {
5225 			free(pstore, M_TEMP);
5226 			break;
5227 		}
5228 		psn->psn_len = sizeof(struct pf_src_node) * nr;
5229 		free(pstore, M_TEMP);
5230 		break;
5231 	}
5232 
5233 	case DIOCCLRSRCNODES: {
5234 		pf_clear_srcnodes(NULL);
5235 		pf_purge_expired_src_nodes();
5236 		break;
5237 	}
5238 
5239 	case DIOCKILLSRCNODES:
5240 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
5241 		break;
5242 
5243 #ifdef COMPAT_FREEBSD13
5244 	case DIOCKEEPCOUNTERS_FREEBSD13:
5245 #endif
5246 	case DIOCKEEPCOUNTERS:
5247 		error = pf_keepcounters((struct pfioc_nv *)addr);
5248 		break;
5249 
5250 	case DIOCGETSYNCOOKIES:
5251 		error = pf_get_syncookies((struct pfioc_nv *)addr);
5252 		break;
5253 
5254 	case DIOCSETSYNCOOKIES:
5255 		error = pf_set_syncookies((struct pfioc_nv *)addr);
5256 		break;
5257 
5258 	case DIOCSETHOSTID: {
5259 		u_int32_t	*hostid = (u_int32_t *)addr;
5260 
5261 		PF_RULES_WLOCK();
5262 		if (*hostid == 0)
5263 			V_pf_status.hostid = arc4random();
5264 		else
5265 			V_pf_status.hostid = *hostid;
5266 		PF_RULES_WUNLOCK();
5267 		break;
5268 	}
5269 
5270 	case DIOCOSFPFLUSH:
5271 		PF_RULES_WLOCK();
5272 		pf_osfp_flush();
5273 		PF_RULES_WUNLOCK();
5274 		break;
5275 
5276 	case DIOCIGETIFACES: {
5277 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5278 		struct pfi_kif *ifstore;
5279 		size_t bufsiz;
5280 
5281 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
5282 			error = ENODEV;
5283 			break;
5284 		}
5285 
5286 		if (io->pfiio_size < 0 ||
5287 		    io->pfiio_size > pf_ioctl_maxcount ||
5288 		    WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
5289 			error = EINVAL;
5290 			break;
5291 		}
5292 
5293 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
5294 		ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
5295 		    M_TEMP, M_WAITOK | M_ZERO);
5296 
5297 		PF_RULES_RLOCK();
5298 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
5299 		PF_RULES_RUNLOCK();
5300 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
5301 		free(ifstore, M_TEMP);
5302 		break;
5303 	}
5304 
5305 	case DIOCSETIFFLAG: {
5306 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5307 
5308 		PF_RULES_WLOCK();
5309 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
5310 		PF_RULES_WUNLOCK();
5311 		break;
5312 	}
5313 
5314 	case DIOCCLRIFFLAG: {
5315 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5316 
5317 		PF_RULES_WLOCK();
5318 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
5319 		PF_RULES_WUNLOCK();
5320 		break;
5321 	}
5322 
5323 	default:
5324 		error = ENODEV;
5325 		break;
5326 	}
5327 fail:
5328 	if (sx_xlocked(&pf_ioctl_lock))
5329 		sx_xunlock(&pf_ioctl_lock);
5330 	CURVNET_RESTORE();
5331 
5332 #undef ERROUT_IOCTL
5333 
5334 	return (error);
5335 }
5336 
5337 void
5338 pfsync_state_export(struct pfsync_state *sp, struct pf_kstate *st)
5339 {
5340 	bzero(sp, sizeof(struct pfsync_state));
5341 
5342 	/* copy from state key */
5343 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5344 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5345 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5346 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5347 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5348 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5349 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5350 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5351 	sp->proto = st->key[PF_SK_WIRE]->proto;
5352 	sp->af = st->key[PF_SK_WIRE]->af;
5353 
5354 	/* copy from state */
5355 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
5356 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
5357 	sp->creation = htonl(time_uptime - st->creation);
5358 	sp->expire = pf_state_expires(st);
5359 	if (sp->expire <= time_uptime)
5360 		sp->expire = htonl(0);
5361 	else
5362 		sp->expire = htonl(sp->expire - time_uptime);
5363 
5364 	sp->direction = st->direction;
5365 	sp->log = st->log;
5366 	sp->timeout = st->timeout;
5367 	sp->state_flags = st->state_flags;
5368 	if (st->src_node)
5369 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
5370 	if (st->nat_src_node)
5371 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5372 
5373 	sp->id = st->id;
5374 	sp->creatorid = st->creatorid;
5375 	pf_state_peer_hton(&st->src, &sp->src);
5376 	pf_state_peer_hton(&st->dst, &sp->dst);
5377 
5378 	if (st->rule.ptr == NULL)
5379 		sp->rule = htonl(-1);
5380 	else
5381 		sp->rule = htonl(st->rule.ptr->nr);
5382 	if (st->anchor.ptr == NULL)
5383 		sp->anchor = htonl(-1);
5384 	else
5385 		sp->anchor = htonl(st->anchor.ptr->nr);
5386 	if (st->nat_rule.ptr == NULL)
5387 		sp->nat_rule = htonl(-1);
5388 	else
5389 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
5390 
5391 	pf_state_counter_hton(st->packets[0], sp->packets[0]);
5392 	pf_state_counter_hton(st->packets[1], sp->packets[1]);
5393 	pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
5394 	pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
5395 }
5396 
5397 void
5398 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st)
5399 {
5400 	bzero(sp, sizeof(*sp));
5401 
5402 	sp->version = PF_STATE_VERSION;
5403 
5404 	/* copy from state key */
5405 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5406 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5407 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5408 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5409 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5410 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5411 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5412 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5413 	sp->proto = st->key[PF_SK_WIRE]->proto;
5414 	sp->af = st->key[PF_SK_WIRE]->af;
5415 
5416 	/* copy from state */
5417 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
5418 	strlcpy(sp->orig_ifname, st->orig_kif->pfik_name,
5419 	    sizeof(sp->orig_ifname));
5420 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
5421 	sp->creation = htonl(time_uptime - st->creation);
5422 	sp->expire = pf_state_expires(st);
5423 	if (sp->expire <= time_uptime)
5424 		sp->expire = htonl(0);
5425 	else
5426 		sp->expire = htonl(sp->expire - time_uptime);
5427 
5428 	sp->direction = st->direction;
5429 	sp->log = st->log;
5430 	sp->timeout = st->timeout;
5431 	sp->state_flags = st->state_flags;
5432 	if (st->src_node)
5433 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
5434 	if (st->nat_src_node)
5435 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5436 
5437 	sp->id = st->id;
5438 	sp->creatorid = st->creatorid;
5439 	pf_state_peer_hton(&st->src, &sp->src);
5440 	pf_state_peer_hton(&st->dst, &sp->dst);
5441 
5442 	if (st->rule.ptr == NULL)
5443 		sp->rule = htonl(-1);
5444 	else
5445 		sp->rule = htonl(st->rule.ptr->nr);
5446 	if (st->anchor.ptr == NULL)
5447 		sp->anchor = htonl(-1);
5448 	else
5449 		sp->anchor = htonl(st->anchor.ptr->nr);
5450 	if (st->nat_rule.ptr == NULL)
5451 		sp->nat_rule = htonl(-1);
5452 	else
5453 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
5454 
5455 	sp->packets[0] = st->packets[0];
5456 	sp->packets[1] = st->packets[1];
5457 	sp->bytes[0] = st->bytes[0];
5458 	sp->bytes[1] = st->bytes[1];
5459 }
5460 
5461 static void
5462 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
5463 {
5464 	struct pfr_ktable *kt;
5465 
5466 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
5467 
5468 	kt = aw->p.tbl;
5469 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
5470 		kt = kt->pfrkt_root;
5471 	aw->p.tbl = NULL;
5472 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
5473 		kt->pfrkt_cnt : -1;
5474 }
5475 
5476 static int
5477 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters,
5478     size_t number, char **names)
5479 {
5480 	nvlist_t        *nvc;
5481 
5482 	nvc = nvlist_create(0);
5483 	if (nvc == NULL)
5484 		return (ENOMEM);
5485 
5486 	for (int i = 0; i < number; i++) {
5487 		nvlist_append_number_array(nvc, "counters",
5488 		    counter_u64_fetch(counters[i]));
5489 		nvlist_append_string_array(nvc, "names",
5490 		    names[i]);
5491 		nvlist_append_number_array(nvc, "ids",
5492 		    i);
5493 	}
5494 	nvlist_add_nvlist(nvl, name, nvc);
5495 	nvlist_destroy(nvc);
5496 
5497 	return (0);
5498 }
5499 
5500 static int
5501 pf_getstatus(struct pfioc_nv *nv)
5502 {
5503 	nvlist_t        *nvl = NULL, *nvc = NULL;
5504 	void            *nvlpacked = NULL;
5505 	int              error;
5506 	struct pf_status s;
5507 	char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES;
5508 	char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES;
5509 	char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES;
5510 	PF_RULES_RLOCK_TRACKER;
5511 
5512 #define ERROUT(x)      ERROUT_FUNCTION(errout, x)
5513 
5514 	PF_RULES_RLOCK();
5515 
5516 	nvl = nvlist_create(0);
5517 	if (nvl == NULL)
5518 		ERROUT(ENOMEM);
5519 
5520 	nvlist_add_bool(nvl, "running", V_pf_status.running);
5521 	nvlist_add_number(nvl, "since", V_pf_status.since);
5522 	nvlist_add_number(nvl, "debug", V_pf_status.debug);
5523 	nvlist_add_number(nvl, "hostid", V_pf_status.hostid);
5524 	nvlist_add_number(nvl, "states", V_pf_status.states);
5525 	nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes);
5526 
5527 	/* counters */
5528 	error = pf_add_status_counters(nvl, "counters", V_pf_status.counters,
5529 	    PFRES_MAX, pf_reasons);
5530 	if (error != 0)
5531 		ERROUT(error);
5532 
5533 	/* lcounters */
5534 	error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters,
5535 	    KLCNT_MAX, pf_lcounter);
5536 	if (error != 0)
5537 		ERROUT(error);
5538 
5539 	/* fcounters */
5540 	nvc = nvlist_create(0);
5541 	if (nvc == NULL)
5542 		ERROUT(ENOMEM);
5543 
5544 	for (int i = 0; i < FCNT_MAX; i++) {
5545 		nvlist_append_number_array(nvc, "counters",
5546 		    pf_counter_u64_fetch(&V_pf_status.fcounters[i]));
5547 		nvlist_append_string_array(nvc, "names",
5548 		    pf_fcounter[i]);
5549 		nvlist_append_number_array(nvc, "ids",
5550 		    i);
5551 	}
5552 	nvlist_add_nvlist(nvl, "fcounters", nvc);
5553 	nvlist_destroy(nvc);
5554 	nvc = NULL;
5555 
5556 	/* scounters */
5557 	error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters,
5558 	    SCNT_MAX, pf_fcounter);
5559 	if (error != 0)
5560 		ERROUT(error);
5561 
5562 	nvlist_add_string(nvl, "ifname", V_pf_status.ifname);
5563 	nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum,
5564 	    PF_MD5_DIGEST_LENGTH);
5565 
5566 	pfi_update_status(V_pf_status.ifname, &s);
5567 
5568 	/* pcounters / bcounters */
5569 	for (int i = 0; i < 2; i++) {
5570 		for (int j = 0; j < 2; j++) {
5571 			for (int k = 0; k < 2; k++) {
5572 				nvlist_append_number_array(nvl, "pcounters",
5573 				    s.pcounters[i][j][k]);
5574 			}
5575 			nvlist_append_number_array(nvl, "bcounters",
5576 			    s.bcounters[i][j]);
5577 		}
5578 	}
5579 
5580 	nvlpacked = nvlist_pack(nvl, &nv->len);
5581 	if (nvlpacked == NULL)
5582 		ERROUT(ENOMEM);
5583 
5584 	if (nv->size == 0)
5585 		ERROUT(0);
5586 	else if (nv->size < nv->len)
5587 		ERROUT(ENOSPC);
5588 
5589 	PF_RULES_RUNLOCK();
5590 	error = copyout(nvlpacked, nv->data, nv->len);
5591 	goto done;
5592 
5593 #undef ERROUT
5594 errout:
5595 	PF_RULES_RUNLOCK();
5596 done:
5597 	free(nvlpacked, M_NVLIST);
5598 	nvlist_destroy(nvc);
5599 	nvlist_destroy(nvl);
5600 
5601 	return (error);
5602 }
5603 
5604 /*
5605  * XXX - Check for version missmatch!!!
5606  */
5607 static void
5608 pf_clear_all_states(void)
5609 {
5610 	struct pf_kstate	*s;
5611 	u_int i;
5612 
5613 	for (i = 0; i <= pf_hashmask; i++) {
5614 		struct pf_idhash *ih = &V_pf_idhash[i];
5615 relock:
5616 		PF_HASHROW_LOCK(ih);
5617 		LIST_FOREACH(s, &ih->states, entry) {
5618 			s->timeout = PFTM_PURGE;
5619 			/* Don't send out individual delete messages. */
5620 			s->state_flags |= PFSTATE_NOSYNC;
5621 			pf_unlink_state(s);
5622 			goto relock;
5623 		}
5624 		PF_HASHROW_UNLOCK(ih);
5625 	}
5626 }
5627 
5628 static int
5629 pf_clear_tables(void)
5630 {
5631 	struct pfioc_table io;
5632 	int error;
5633 
5634 	bzero(&io, sizeof(io));
5635 
5636 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
5637 	    io.pfrio_flags);
5638 
5639 	return (error);
5640 }
5641 
5642 static void
5643 pf_clear_srcnodes(struct pf_ksrc_node *n)
5644 {
5645 	struct pf_kstate *s;
5646 	int i;
5647 
5648 	for (i = 0; i <= pf_hashmask; i++) {
5649 		struct pf_idhash *ih = &V_pf_idhash[i];
5650 
5651 		PF_HASHROW_LOCK(ih);
5652 		LIST_FOREACH(s, &ih->states, entry) {
5653 			if (n == NULL || n == s->src_node)
5654 				s->src_node = NULL;
5655 			if (n == NULL || n == s->nat_src_node)
5656 				s->nat_src_node = NULL;
5657 		}
5658 		PF_HASHROW_UNLOCK(ih);
5659 	}
5660 
5661 	if (n == NULL) {
5662 		struct pf_srchash *sh;
5663 
5664 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5665 		    i++, sh++) {
5666 			PF_HASHROW_LOCK(sh);
5667 			LIST_FOREACH(n, &sh->nodes, entry) {
5668 				n->expire = 1;
5669 				n->states = 0;
5670 			}
5671 			PF_HASHROW_UNLOCK(sh);
5672 		}
5673 	} else {
5674 		/* XXX: hash slot should already be locked here. */
5675 		n->expire = 1;
5676 		n->states = 0;
5677 	}
5678 }
5679 
5680 static void
5681 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
5682 {
5683 	struct pf_ksrc_node_list	 kill;
5684 
5685 	LIST_INIT(&kill);
5686 	for (int i = 0; i <= pf_srchashmask; i++) {
5687 		struct pf_srchash *sh = &V_pf_srchash[i];
5688 		struct pf_ksrc_node *sn, *tmp;
5689 
5690 		PF_HASHROW_LOCK(sh);
5691 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
5692 			if (PF_MATCHA(psnk->psnk_src.neg,
5693 			      &psnk->psnk_src.addr.v.a.addr,
5694 			      &psnk->psnk_src.addr.v.a.mask,
5695 			      &sn->addr, sn->af) &&
5696 			    PF_MATCHA(psnk->psnk_dst.neg,
5697 			      &psnk->psnk_dst.addr.v.a.addr,
5698 			      &psnk->psnk_dst.addr.v.a.mask,
5699 			      &sn->raddr, sn->af)) {
5700 				pf_unlink_src_node(sn);
5701 				LIST_INSERT_HEAD(&kill, sn, entry);
5702 				sn->expire = 1;
5703 			}
5704 		PF_HASHROW_UNLOCK(sh);
5705 	}
5706 
5707 	for (int i = 0; i <= pf_hashmask; i++) {
5708 		struct pf_idhash *ih = &V_pf_idhash[i];
5709 		struct pf_kstate *s;
5710 
5711 		PF_HASHROW_LOCK(ih);
5712 		LIST_FOREACH(s, &ih->states, entry) {
5713 			if (s->src_node && s->src_node->expire == 1)
5714 				s->src_node = NULL;
5715 			if (s->nat_src_node && s->nat_src_node->expire == 1)
5716 				s->nat_src_node = NULL;
5717 		}
5718 		PF_HASHROW_UNLOCK(ih);
5719 	}
5720 
5721 	psnk->psnk_killed = pf_free_src_nodes(&kill);
5722 }
5723 
5724 static int
5725 pf_keepcounters(struct pfioc_nv *nv)
5726 {
5727 	nvlist_t	*nvl = NULL;
5728 	void		*nvlpacked = NULL;
5729 	int		 error = 0;
5730 
5731 #define	ERROUT(x)	ERROUT_FUNCTION(on_error, x)
5732 
5733 	if (nv->len > pf_ioctl_maxcount)
5734 		ERROUT(ENOMEM);
5735 
5736 	nvlpacked = malloc(nv->len, M_TEMP, M_WAITOK);
5737 	if (nvlpacked == NULL)
5738 		ERROUT(ENOMEM);
5739 
5740 	error = copyin(nv->data, nvlpacked, nv->len);
5741 	if (error)
5742 		ERROUT(error);
5743 
5744 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
5745 	if (nvl == NULL)
5746 		ERROUT(EBADMSG);
5747 
5748 	if (! nvlist_exists_bool(nvl, "keep_counters"))
5749 		ERROUT(EBADMSG);
5750 
5751 	V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters");
5752 
5753 on_error:
5754 	nvlist_destroy(nvl);
5755 	free(nvlpacked, M_TEMP);
5756 	return (error);
5757 }
5758 
5759 static unsigned int
5760 pf_clear_states(const struct pf_kstate_kill *kill)
5761 {
5762 	struct pf_state_key_cmp	 match_key;
5763 	struct pf_kstate	*s;
5764 	struct pfi_kkif	*kif;
5765 	int		 idx;
5766 	unsigned int	 killed = 0, dir;
5767 
5768 	for (unsigned int i = 0; i <= pf_hashmask; i++) {
5769 		struct pf_idhash *ih = &V_pf_idhash[i];
5770 
5771 relock_DIOCCLRSTATES:
5772 		PF_HASHROW_LOCK(ih);
5773 		LIST_FOREACH(s, &ih->states, entry) {
5774 			/* For floating states look at the original kif. */
5775 			kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
5776 
5777 			if (kill->psk_ifname[0] &&
5778 			    strcmp(kill->psk_ifname,
5779 			    kif->pfik_name))
5780 				continue;
5781 
5782 			if (kill->psk_kill_match) {
5783 				bzero(&match_key, sizeof(match_key));
5784 
5785 				if (s->direction == PF_OUT) {
5786 					dir = PF_IN;
5787 					idx = PF_SK_STACK;
5788 				} else {
5789 					dir = PF_OUT;
5790 					idx = PF_SK_WIRE;
5791 				}
5792 
5793 				match_key.af = s->key[idx]->af;
5794 				match_key.proto = s->key[idx]->proto;
5795 				PF_ACPY(&match_key.addr[0],
5796 				    &s->key[idx]->addr[1], match_key.af);
5797 				match_key.port[0] = s->key[idx]->port[1];
5798 				PF_ACPY(&match_key.addr[1],
5799 				    &s->key[idx]->addr[0], match_key.af);
5800 				match_key.port[1] = s->key[idx]->port[0];
5801 			}
5802 
5803 			/*
5804 			 * Don't send out individual
5805 			 * delete messages.
5806 			 */
5807 			s->state_flags |= PFSTATE_NOSYNC;
5808 			pf_unlink_state(s);
5809 			killed++;
5810 
5811 			if (kill->psk_kill_match)
5812 				killed += pf_kill_matching_state(&match_key,
5813 				    dir);
5814 
5815 			goto relock_DIOCCLRSTATES;
5816 		}
5817 		PF_HASHROW_UNLOCK(ih);
5818 	}
5819 
5820 	if (V_pfsync_clear_states_ptr != NULL)
5821 		V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname);
5822 
5823 	return (killed);
5824 }
5825 
5826 static void
5827 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed)
5828 {
5829 	struct pf_kstate	*s;
5830 
5831 	if (kill->psk_pfcmp.id) {
5832 		if (kill->psk_pfcmp.creatorid == 0)
5833 			kill->psk_pfcmp.creatorid = V_pf_status.hostid;
5834 		if ((s = pf_find_state_byid(kill->psk_pfcmp.id,
5835 		    kill->psk_pfcmp.creatorid))) {
5836 			pf_unlink_state(s);
5837 			*killed = 1;
5838 		}
5839 		return;
5840 	}
5841 
5842 	for (unsigned int i = 0; i <= pf_hashmask; i++)
5843 		*killed += pf_killstates_row(kill, &V_pf_idhash[i]);
5844 
5845 	return;
5846 }
5847 
5848 static int
5849 pf_killstates_nv(struct pfioc_nv *nv)
5850 {
5851 	struct pf_kstate_kill	 kill;
5852 	nvlist_t		*nvl = NULL;
5853 	void			*nvlpacked = NULL;
5854 	int			 error = 0;
5855 	unsigned int		 killed = 0;
5856 
5857 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
5858 
5859 	if (nv->len > pf_ioctl_maxcount)
5860 		ERROUT(ENOMEM);
5861 
5862 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
5863 	if (nvlpacked == NULL)
5864 		ERROUT(ENOMEM);
5865 
5866 	error = copyin(nv->data, nvlpacked, nv->len);
5867 	if (error)
5868 		ERROUT(error);
5869 
5870 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
5871 	if (nvl == NULL)
5872 		ERROUT(EBADMSG);
5873 
5874 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
5875 	if (error)
5876 		ERROUT(error);
5877 
5878 	pf_killstates(&kill, &killed);
5879 
5880 	free(nvlpacked, M_NVLIST);
5881 	nvlpacked = NULL;
5882 	nvlist_destroy(nvl);
5883 	nvl = nvlist_create(0);
5884 	if (nvl == NULL)
5885 		ERROUT(ENOMEM);
5886 
5887 	nvlist_add_number(nvl, "killed", killed);
5888 
5889 	nvlpacked = nvlist_pack(nvl, &nv->len);
5890 	if (nvlpacked == NULL)
5891 		ERROUT(ENOMEM);
5892 
5893 	if (nv->size == 0)
5894 		ERROUT(0);
5895 	else if (nv->size < nv->len)
5896 		ERROUT(ENOSPC);
5897 
5898 	error = copyout(nvlpacked, nv->data, nv->len);
5899 
5900 on_error:
5901 	nvlist_destroy(nvl);
5902 	free(nvlpacked, M_NVLIST);
5903 	return (error);
5904 }
5905 
5906 static int
5907 pf_clearstates_nv(struct pfioc_nv *nv)
5908 {
5909 	struct pf_kstate_kill	 kill;
5910 	nvlist_t		*nvl = NULL;
5911 	void			*nvlpacked = NULL;
5912 	int			 error = 0;
5913 	unsigned int		 killed;
5914 
5915 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
5916 
5917 	if (nv->len > pf_ioctl_maxcount)
5918 		ERROUT(ENOMEM);
5919 
5920 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
5921 	if (nvlpacked == NULL)
5922 		ERROUT(ENOMEM);
5923 
5924 	error = copyin(nv->data, nvlpacked, nv->len);
5925 	if (error)
5926 		ERROUT(error);
5927 
5928 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
5929 	if (nvl == NULL)
5930 		ERROUT(EBADMSG);
5931 
5932 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
5933 	if (error)
5934 		ERROUT(error);
5935 
5936 	killed = pf_clear_states(&kill);
5937 
5938 	free(nvlpacked, M_NVLIST);
5939 	nvlpacked = NULL;
5940 	nvlist_destroy(nvl);
5941 	nvl = nvlist_create(0);
5942 	if (nvl == NULL)
5943 		ERROUT(ENOMEM);
5944 
5945 	nvlist_add_number(nvl, "killed", killed);
5946 
5947 	nvlpacked = nvlist_pack(nvl, &nv->len);
5948 	if (nvlpacked == NULL)
5949 		ERROUT(ENOMEM);
5950 
5951 	if (nv->size == 0)
5952 		ERROUT(0);
5953 	else if (nv->size < nv->len)
5954 		ERROUT(ENOSPC);
5955 
5956 	error = copyout(nvlpacked, nv->data, nv->len);
5957 
5958 #undef ERROUT
5959 on_error:
5960 	nvlist_destroy(nvl);
5961 	free(nvlpacked, M_NVLIST);
5962 	return (error);
5963 }
5964 
5965 static int
5966 pf_getstate(struct pfioc_nv *nv)
5967 {
5968 	nvlist_t		*nvl = NULL, *nvls;
5969 	void			*nvlpacked = NULL;
5970 	struct pf_kstate	*s = NULL;
5971 	int			 error = 0;
5972 	uint64_t		 id, creatorid;
5973 
5974 #define ERROUT(x)	ERROUT_FUNCTION(errout, x)
5975 
5976 	if (nv->len > pf_ioctl_maxcount)
5977 		ERROUT(ENOMEM);
5978 
5979 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
5980 	if (nvlpacked == NULL)
5981 		ERROUT(ENOMEM);
5982 
5983 	error = copyin(nv->data, nvlpacked, nv->len);
5984 	if (error)
5985 		ERROUT(error);
5986 
5987 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
5988 	if (nvl == NULL)
5989 		ERROUT(EBADMSG);
5990 
5991 	PFNV_CHK(pf_nvuint64(nvl, "id", &id));
5992 	PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid));
5993 
5994 	s = pf_find_state_byid(id, creatorid);
5995 	if (s == NULL)
5996 		ERROUT(ENOENT);
5997 
5998 	free(nvlpacked, M_NVLIST);
5999 	nvlpacked = NULL;
6000 	nvlist_destroy(nvl);
6001 	nvl = nvlist_create(0);
6002 	if (nvl == NULL)
6003 		ERROUT(ENOMEM);
6004 
6005 	nvls = pf_state_to_nvstate(s);
6006 	if (nvls == NULL)
6007 		ERROUT(ENOMEM);
6008 
6009 	nvlist_add_nvlist(nvl, "state", nvls);
6010 	nvlist_destroy(nvls);
6011 
6012 	nvlpacked = nvlist_pack(nvl, &nv->len);
6013 	if (nvlpacked == NULL)
6014 		ERROUT(ENOMEM);
6015 
6016 	if (nv->size == 0)
6017 		ERROUT(0);
6018 	else if (nv->size < nv->len)
6019 		ERROUT(ENOSPC);
6020 
6021 	error = copyout(nvlpacked, nv->data, nv->len);
6022 
6023 #undef ERROUT
6024 errout:
6025 	if (s != NULL)
6026 		PF_STATE_UNLOCK(s);
6027 	free(nvlpacked, M_NVLIST);
6028 	nvlist_destroy(nvl);
6029 	return (error);
6030 }
6031 
6032 /*
6033  * XXX - Check for version missmatch!!!
6034  */
6035 
6036 /*
6037  * Duplicate pfctl -Fa operation to get rid of as much as we can.
6038  */
6039 static int
6040 shutdown_pf(void)
6041 {
6042 	int error = 0;
6043 	u_int32_t t[5];
6044 	char nn = '\0';
6045 
6046 	do {
6047 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
6048 		    != 0) {
6049 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
6050 			break;
6051 		}
6052 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
6053 		    != 0) {
6054 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
6055 			break;		/* XXX: rollback? */
6056 		}
6057 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
6058 		    != 0) {
6059 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
6060 			break;		/* XXX: rollback? */
6061 		}
6062 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
6063 		    != 0) {
6064 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
6065 			break;		/* XXX: rollback? */
6066 		}
6067 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
6068 		    != 0) {
6069 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
6070 			break;		/* XXX: rollback? */
6071 		}
6072 
6073 		/* XXX: these should always succeed here */
6074 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
6075 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
6076 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
6077 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
6078 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
6079 
6080 		if ((error = pf_clear_tables()) != 0)
6081 			break;
6082 
6083 		if ((error = pf_begin_eth(&t[0], &nn)) != 0) {
6084 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth\n"));
6085 			break;
6086 		}
6087 		pf_commit_eth(t[0], &nn);
6088 
6089 #ifdef ALTQ
6090 		if ((error = pf_begin_altq(&t[0])) != 0) {
6091 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
6092 			break;
6093 		}
6094 		pf_commit_altq(t[0]);
6095 #endif
6096 
6097 		pf_clear_all_states();
6098 
6099 		pf_clear_srcnodes(NULL);
6100 
6101 		/* status does not use malloced mem so no need to cleanup */
6102 		/* fingerprints and interfaces have their own cleanup code */
6103 	} while(0);
6104 
6105 	return (error);
6106 }
6107 
6108 static pfil_return_t
6109 pf_check_return(int chk, struct mbuf **m)
6110 {
6111 
6112 	switch (chk) {
6113 	case PF_PASS:
6114 		if (*m == NULL)
6115 			return (PFIL_CONSUMED);
6116 		else
6117 			return (PFIL_PASS);
6118 		break;
6119 	default:
6120 		if (*m != NULL) {
6121 			m_freem(*m);
6122 			*m = NULL;
6123 		}
6124 		return (PFIL_DROPPED);
6125 	}
6126 }
6127 
6128 static pfil_return_t
6129 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6130     void *ruleset __unused, struct inpcb *inp)
6131 {
6132 	int chk;
6133 
6134 	chk = pf_test_eth(PF_IN, flags, ifp, m, inp);
6135 
6136 	return (pf_check_return(chk, m));
6137 }
6138 
6139 static pfil_return_t
6140 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6141     void *ruleset __unused, struct inpcb *inp)
6142 {
6143 	int chk;
6144 
6145 	chk = pf_test_eth(PF_OUT, flags, ifp, m, inp);
6146 
6147 	return (pf_check_return(chk, m));
6148 }
6149 
6150 #ifdef INET
6151 static pfil_return_t
6152 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6153     void *ruleset __unused, struct inpcb *inp)
6154 {
6155 	int chk;
6156 
6157 	chk = pf_test(PF_IN, flags, ifp, m, inp);
6158 
6159 	return (pf_check_return(chk, m));
6160 }
6161 
6162 static pfil_return_t
6163 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6164     void *ruleset __unused,  struct inpcb *inp)
6165 {
6166 	int chk;
6167 
6168 	chk = pf_test(PF_OUT, flags, ifp, m, inp);
6169 
6170 	return (pf_check_return(chk, m));
6171 }
6172 #endif
6173 
6174 #ifdef INET6
6175 static pfil_return_t
6176 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
6177     void *ruleset __unused,  struct inpcb *inp)
6178 {
6179 	int chk;
6180 
6181 	/*
6182 	 * In case of loopback traffic IPv6 uses the real interface in
6183 	 * order to support scoped addresses. In order to support stateful
6184 	 * filtering we have change this to lo0 as it is the case in IPv4.
6185 	 */
6186 	CURVNET_SET(ifp->if_vnet);
6187 	chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
6188 	CURVNET_RESTORE();
6189 
6190 	return (pf_check_return(chk, m));
6191 }
6192 
6193 static pfil_return_t
6194 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
6195     void *ruleset __unused,  struct inpcb *inp)
6196 {
6197 	int chk;
6198 
6199 	CURVNET_SET(ifp->if_vnet);
6200 	chk = pf_test6(PF_OUT, flags, ifp, m, inp);
6201 	CURVNET_RESTORE();
6202 
6203 	return (pf_check_return(chk, m));
6204 }
6205 #endif /* INET6 */
6206 
6207 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook);
6208 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook);
6209 #define	V_pf_eth_in_hook	VNET(pf_eth_in_hook)
6210 #define	V_pf_eth_out_hook	VNET(pf_eth_out_hook)
6211 
6212 #ifdef INET
6213 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
6214 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
6215 #define	V_pf_ip4_in_hook	VNET(pf_ip4_in_hook)
6216 #define	V_pf_ip4_out_hook	VNET(pf_ip4_out_hook)
6217 #endif
6218 #ifdef INET6
6219 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
6220 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
6221 #define	V_pf_ip6_in_hook	VNET(pf_ip6_in_hook)
6222 #define	V_pf_ip6_out_hook	VNET(pf_ip6_out_hook)
6223 #endif
6224 
6225 static void
6226 hook_pf_eth(void)
6227 {
6228 	struct pfil_hook_args pha;
6229 	struct pfil_link_args pla;
6230 	int ret __diagused;
6231 
6232 	if (V_pf_pfil_eth_hooked)
6233 		return;
6234 
6235 	pha.pa_version = PFIL_VERSION;
6236 	pha.pa_modname = "pf";
6237 	pha.pa_ruleset = NULL;
6238 
6239 	pla.pa_version = PFIL_VERSION;
6240 
6241 	pha.pa_type = PFIL_TYPE_ETHERNET;
6242 	pha.pa_func = pf_eth_check_in;
6243 	pha.pa_flags = PFIL_IN;
6244 	pha.pa_rulname = "eth-in";
6245 	V_pf_eth_in_hook = pfil_add_hook(&pha);
6246 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6247 	pla.pa_head = V_link_pfil_head;
6248 	pla.pa_hook = V_pf_eth_in_hook;
6249 	ret = pfil_link(&pla);
6250 	MPASS(ret == 0);
6251 	pha.pa_func = pf_eth_check_out;
6252 	pha.pa_flags = PFIL_OUT;
6253 	pha.pa_rulname = "eth-out";
6254 	V_pf_eth_out_hook = pfil_add_hook(&pha);
6255 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6256 	pla.pa_head = V_link_pfil_head;
6257 	pla.pa_hook = V_pf_eth_out_hook;
6258 	ret = pfil_link(&pla);
6259 	MPASS(ret == 0);
6260 
6261 	V_pf_pfil_eth_hooked = 1;
6262 }
6263 
6264 static void
6265 hook_pf(void)
6266 {
6267 	struct pfil_hook_args pha;
6268 	struct pfil_link_args pla;
6269 	int ret;
6270 
6271 	if (V_pf_pfil_hooked)
6272 		return;
6273 
6274 	pha.pa_version = PFIL_VERSION;
6275 	pha.pa_modname = "pf";
6276 	pha.pa_ruleset = NULL;
6277 
6278 	pla.pa_version = PFIL_VERSION;
6279 
6280 #ifdef INET
6281 	pha.pa_type = PFIL_TYPE_IP4;
6282 	pha.pa_func = pf_check_in;
6283 	pha.pa_flags = PFIL_IN;
6284 	pha.pa_rulname = "default-in";
6285 	V_pf_ip4_in_hook = pfil_add_hook(&pha);
6286 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6287 	pla.pa_head = V_inet_pfil_head;
6288 	pla.pa_hook = V_pf_ip4_in_hook;
6289 	ret = pfil_link(&pla);
6290 	MPASS(ret == 0);
6291 	pha.pa_func = pf_check_out;
6292 	pha.pa_flags = PFIL_OUT;
6293 	pha.pa_rulname = "default-out";
6294 	V_pf_ip4_out_hook = pfil_add_hook(&pha);
6295 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6296 	pla.pa_head = V_inet_pfil_head;
6297 	pla.pa_hook = V_pf_ip4_out_hook;
6298 	ret = pfil_link(&pla);
6299 	MPASS(ret == 0);
6300 #endif
6301 #ifdef INET6
6302 	pha.pa_type = PFIL_TYPE_IP6;
6303 	pha.pa_func = pf_check6_in;
6304 	pha.pa_flags = PFIL_IN;
6305 	pha.pa_rulname = "default-in6";
6306 	V_pf_ip6_in_hook = pfil_add_hook(&pha);
6307 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6308 	pla.pa_head = V_inet6_pfil_head;
6309 	pla.pa_hook = V_pf_ip6_in_hook;
6310 	ret = pfil_link(&pla);
6311 	MPASS(ret == 0);
6312 	pha.pa_func = pf_check6_out;
6313 	pha.pa_rulname = "default-out6";
6314 	pha.pa_flags = PFIL_OUT;
6315 	V_pf_ip6_out_hook = pfil_add_hook(&pha);
6316 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6317 	pla.pa_head = V_inet6_pfil_head;
6318 	pla.pa_hook = V_pf_ip6_out_hook;
6319 	ret = pfil_link(&pla);
6320 	MPASS(ret == 0);
6321 #endif
6322 
6323 	V_pf_pfil_hooked = 1;
6324 }
6325 
6326 static void
6327 dehook_pf_eth(void)
6328 {
6329 
6330 	if (V_pf_pfil_eth_hooked == 0)
6331 		return;
6332 
6333 	pfil_remove_hook(V_pf_eth_in_hook);
6334 	pfil_remove_hook(V_pf_eth_out_hook);
6335 
6336 	V_pf_pfil_eth_hooked = 0;
6337 }
6338 
6339 static void
6340 dehook_pf(void)
6341 {
6342 
6343 	if (V_pf_pfil_hooked == 0)
6344 		return;
6345 
6346 #ifdef INET
6347 	pfil_remove_hook(V_pf_ip4_in_hook);
6348 	pfil_remove_hook(V_pf_ip4_out_hook);
6349 #endif
6350 #ifdef INET6
6351 	pfil_remove_hook(V_pf_ip6_in_hook);
6352 	pfil_remove_hook(V_pf_ip6_out_hook);
6353 #endif
6354 
6355 	V_pf_pfil_hooked = 0;
6356 }
6357 
6358 static void
6359 pf_load_vnet(void)
6360 {
6361 	V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
6362 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
6363 
6364 	pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
6365 	    PF_RULE_TAG_HASH_SIZE_DEFAULT);
6366 #ifdef ALTQ
6367 	pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
6368 	    PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
6369 #endif
6370 
6371 	V_pf_keth = &V_pf_main_keth_anchor.ruleset;
6372 
6373 	pfattach_vnet();
6374 	V_pf_vnet_active = 1;
6375 }
6376 
6377 static int
6378 pf_load(void)
6379 {
6380 	int error;
6381 
6382 	rm_init_flags(&pf_rules_lock, "pf rulesets", RM_RECURSE);
6383 	sx_init(&pf_ioctl_lock, "pf ioctl");
6384 	sx_init(&pf_end_lock, "pf end thread");
6385 
6386 	pf_mtag_initialize();
6387 
6388 	pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
6389 	if (pf_dev == NULL)
6390 		return (ENOMEM);
6391 
6392 	pf_end_threads = 0;
6393 	error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
6394 	if (error != 0)
6395 		return (error);
6396 
6397 	pfi_initialize();
6398 
6399 	return (0);
6400 }
6401 
6402 static void
6403 pf_unload_vnet(void)
6404 {
6405 	int ret;
6406 
6407 	V_pf_vnet_active = 0;
6408 	V_pf_status.running = 0;
6409 	dehook_pf();
6410 	dehook_pf_eth();
6411 
6412 	PF_RULES_WLOCK();
6413 	pf_syncookies_cleanup();
6414 	shutdown_pf();
6415 	PF_RULES_WUNLOCK();
6416 
6417 	ret = swi_remove(V_pf_swi_cookie);
6418 	MPASS(ret == 0);
6419 	ret = intr_event_destroy(V_pf_swi_ie);
6420 	MPASS(ret == 0);
6421 
6422 	pf_unload_vnet_purge();
6423 
6424 	pf_normalize_cleanup();
6425 	PF_RULES_WLOCK();
6426 	pfi_cleanup_vnet();
6427 	PF_RULES_WUNLOCK();
6428 	pfr_cleanup();
6429 	pf_osfp_flush();
6430 	pf_cleanup();
6431 	if (IS_DEFAULT_VNET(curvnet))
6432 		pf_mtag_cleanup();
6433 
6434 	pf_cleanup_tagset(&V_pf_tags);
6435 #ifdef ALTQ
6436 	pf_cleanup_tagset(&V_pf_qids);
6437 #endif
6438 	uma_zdestroy(V_pf_tag_z);
6439 
6440 #ifdef PF_WANT_32_TO_64_COUNTER
6441 	PF_RULES_WLOCK();
6442 	LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist);
6443 
6444 	MPASS(LIST_EMPTY(&V_pf_allkiflist));
6445 	MPASS(V_pf_allkifcount == 0);
6446 
6447 	LIST_REMOVE(&V_pf_default_rule, allrulelist);
6448 	V_pf_allrulecount--;
6449 	LIST_REMOVE(V_pf_rulemarker, allrulelist);
6450 
6451 	/*
6452 	 * There are known pf rule leaks when running the test suite.
6453 	 */
6454 #ifdef notyet
6455 	MPASS(LIST_EMPTY(&V_pf_allrulelist));
6456 	MPASS(V_pf_allrulecount == 0);
6457 #endif
6458 
6459 	PF_RULES_WUNLOCK();
6460 
6461 	free(V_pf_kifmarker, PFI_MTYPE);
6462 	free(V_pf_rulemarker, M_PFRULE);
6463 #endif
6464 
6465 	/* Free counters last as we updated them during shutdown. */
6466 	pf_counter_u64_deinit(&V_pf_default_rule.evaluations);
6467 	for (int i = 0; i < 2; i++) {
6468 		pf_counter_u64_deinit(&V_pf_default_rule.packets[i]);
6469 		pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]);
6470 	}
6471 	counter_u64_free(V_pf_default_rule.states_cur);
6472 	counter_u64_free(V_pf_default_rule.states_tot);
6473 	counter_u64_free(V_pf_default_rule.src_nodes);
6474 
6475 	for (int i = 0; i < PFRES_MAX; i++)
6476 		counter_u64_free(V_pf_status.counters[i]);
6477 	for (int i = 0; i < KLCNT_MAX; i++)
6478 		counter_u64_free(V_pf_status.lcounters[i]);
6479 	for (int i = 0; i < FCNT_MAX; i++)
6480 		pf_counter_u64_deinit(&V_pf_status.fcounters[i]);
6481 	for (int i = 0; i < SCNT_MAX; i++)
6482 		counter_u64_free(V_pf_status.scounters[i]);
6483 }
6484 
6485 static void
6486 pf_unload(void)
6487 {
6488 
6489 	sx_xlock(&pf_end_lock);
6490 	pf_end_threads = 1;
6491 	while (pf_end_threads < 2) {
6492 		wakeup_one(pf_purge_thread);
6493 		sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
6494 	}
6495 	sx_xunlock(&pf_end_lock);
6496 
6497 	if (pf_dev != NULL)
6498 		destroy_dev(pf_dev);
6499 
6500 	pfi_cleanup();
6501 
6502 	rm_destroy(&pf_rules_lock);
6503 	sx_destroy(&pf_ioctl_lock);
6504 	sx_destroy(&pf_end_lock);
6505 }
6506 
6507 static void
6508 vnet_pf_init(void *unused __unused)
6509 {
6510 
6511 	pf_load_vnet();
6512 }
6513 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6514     vnet_pf_init, NULL);
6515 
6516 static void
6517 vnet_pf_uninit(const void *unused __unused)
6518 {
6519 
6520 	pf_unload_vnet();
6521 }
6522 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
6523 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6524     vnet_pf_uninit, NULL);
6525 
6526 static int
6527 pf_modevent(module_t mod, int type, void *data)
6528 {
6529 	int error = 0;
6530 
6531 	switch(type) {
6532 	case MOD_LOAD:
6533 		error = pf_load();
6534 		break;
6535 	case MOD_UNLOAD:
6536 		/* Handled in SYSUNINIT(pf_unload) to ensure it's done after
6537 		 * the vnet_pf_uninit()s */
6538 		break;
6539 	default:
6540 		error = EINVAL;
6541 		break;
6542 	}
6543 
6544 	return (error);
6545 }
6546 
6547 static moduledata_t pf_mod = {
6548 	"pf",
6549 	pf_modevent,
6550 	0
6551 };
6552 
6553 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
6554 MODULE_VERSION(pf, PF_MODVER);
6555