xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision 5c6935a645604b1e39b2cf6dbc23f66ec58f3e54)
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(MD5_CTX *, struct pf_krule *);
127 static void		 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
128 static int		 pf_commit_rules(u_int32_t, int, char *);
129 static int		 pf_addr_setup(struct pf_kruleset *,
130 			    struct pf_addr_wrap *, sa_family_t);
131 static void		 pf_addr_copyout(struct pf_addr_wrap *);
132 static void		 pf_src_node_copy(const struct pf_ksrc_node *,
133 			    struct pf_src_node *);
134 #ifdef ALTQ
135 static int		 pf_export_kaltq(struct pf_altq *,
136 			    struct pfioc_altq_v1 *, size_t);
137 static int		 pf_import_kaltq(struct pfioc_altq_v1 *,
138 			    struct pf_altq *, size_t);
139 #endif /* ALTQ */
140 
141 VNET_DEFINE(struct pf_krule,	pf_default_rule);
142 
143 #ifdef ALTQ
144 VNET_DEFINE_STATIC(int,		pf_altq_running);
145 #define	V_pf_altq_running	VNET(pf_altq_running)
146 #endif
147 
148 #define	TAGID_MAX	 50000
149 struct pf_tagname {
150 	TAILQ_ENTRY(pf_tagname)	namehash_entries;
151 	TAILQ_ENTRY(pf_tagname)	taghash_entries;
152 	char			name[PF_TAG_NAME_SIZE];
153 	uint16_t		tag;
154 	int			ref;
155 };
156 
157 struct pf_tagset {
158 	TAILQ_HEAD(, pf_tagname)	*namehash;
159 	TAILQ_HEAD(, pf_tagname)	*taghash;
160 	unsigned int			 mask;
161 	uint32_t			 seed;
162 	BITSET_DEFINE(, TAGID_MAX)	 avail;
163 };
164 
165 VNET_DEFINE(struct pf_tagset, pf_tags);
166 #define	V_pf_tags	VNET(pf_tags)
167 static unsigned int	pf_rule_tag_hashsize;
168 #define	PF_RULE_TAG_HASH_SIZE_DEFAULT	128
169 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN,
170     &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT,
171     "Size of pf(4) rule tag hashtable");
172 
173 #ifdef ALTQ
174 VNET_DEFINE(struct pf_tagset, pf_qids);
175 #define	V_pf_qids	VNET(pf_qids)
176 static unsigned int	pf_queue_tag_hashsize;
177 #define	PF_QUEUE_TAG_HASH_SIZE_DEFAULT	128
178 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN,
179     &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT,
180     "Size of pf(4) queue tag hashtable");
181 #endif
182 VNET_DEFINE(uma_zone_t,	 pf_tag_z);
183 #define	V_pf_tag_z		 VNET(pf_tag_z)
184 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
185 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
186 
187 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
188 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
189 #endif
190 
191 static void		 pf_init_tagset(struct pf_tagset *, unsigned int *,
192 			    unsigned int);
193 static void		 pf_cleanup_tagset(struct pf_tagset *);
194 static uint16_t		 tagname2hashindex(const struct pf_tagset *, const char *);
195 static uint16_t		 tag2hashindex(const struct pf_tagset *, uint16_t);
196 static u_int16_t	 tagname2tag(struct pf_tagset *, const char *);
197 static u_int16_t	 pf_tagname2tag(const char *);
198 static void		 tag_unref(struct pf_tagset *, u_int16_t);
199 
200 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
201 
202 struct cdev *pf_dev;
203 
204 /*
205  * XXX - These are new and need to be checked when moveing to a new version
206  */
207 static void		 pf_clear_all_states(void);
208 static unsigned int	 pf_clear_states(const struct pf_kstate_kill *);
209 static void		 pf_killstates(struct pf_kstate_kill *,
210 			    unsigned int *);
211 static int		 pf_killstates_row(struct pf_kstate_kill *,
212 			    struct pf_idhash *);
213 static int		 pf_killstates_nv(struct pfioc_nv *);
214 static int		 pf_clearstates_nv(struct pfioc_nv *);
215 static int		 pf_getstate(struct pfioc_nv *);
216 static int		 pf_getstatus(struct pfioc_nv *);
217 static int		 pf_clear_tables(void);
218 static void		 pf_clear_srcnodes(struct pf_ksrc_node *);
219 static void		 pf_kill_srcnodes(struct pfioc_src_node_kill *);
220 static int		 pf_keepcounters(struct pfioc_nv *);
221 static void		 pf_tbladdr_copyout(struct pf_addr_wrap *);
222 
223 /*
224  * Wrapper functions for pfil(9) hooks
225  */
226 static pfil_return_t pf_eth_check_in(struct mbuf **m, struct ifnet *ifp,
227     int flags, void *ruleset __unused, struct inpcb *inp);
228 static pfil_return_t pf_eth_check_out(struct mbuf **m, struct ifnet *ifp,
229     int flags, void *ruleset __unused, struct inpcb *inp);
230 #ifdef INET
231 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp,
232     int flags, void *ruleset __unused, struct inpcb *inp);
233 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp,
234     int flags, void *ruleset __unused, struct inpcb *inp);
235 #endif
236 #ifdef INET6
237 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp,
238     int flags, void *ruleset __unused, struct inpcb *inp);
239 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp,
240     int flags, void *ruleset __unused, struct inpcb *inp);
241 #endif
242 
243 static void		hook_pf_eth(void);
244 static void		hook_pf(void);
245 static void		dehook_pf_eth(void);
246 static void		dehook_pf(void);
247 static int		shutdown_pf(void);
248 static int		pf_load(void);
249 static void		pf_unload(void);
250 
251 static struct cdevsw pf_cdevsw = {
252 	.d_ioctl =	pfioctl,
253 	.d_name =	PF_NAME,
254 	.d_version =	D_VERSION,
255 };
256 
257 volatile VNET_DEFINE_STATIC(int, pf_pfil_hooked);
258 #define V_pf_pfil_hooked	VNET(pf_pfil_hooked)
259 volatile VNET_DEFINE_STATIC(int, pf_pfil_eth_hooked);
260 #define V_pf_pfil_eth_hooked	VNET(pf_pfil_eth_hooked)
261 
262 /*
263  * We need a flag that is neither hooked nor running to know when
264  * the VNET is "valid".  We primarily need this to control (global)
265  * external event, e.g., eventhandlers.
266  */
267 VNET_DEFINE(int, pf_vnet_active);
268 #define V_pf_vnet_active	VNET(pf_vnet_active)
269 
270 int pf_end_threads;
271 struct proc *pf_purge_proc;
272 
273 struct rmlock			pf_rules_lock;
274 struct sx			pf_ioctl_lock;
275 struct sx			pf_end_lock;
276 
277 /* pfsync */
278 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
279 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
280 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
281 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
282 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
283 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
284 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
285 
286 /* pflog */
287 pflog_packet_t			*pflog_packet_ptr = NULL;
288 
289 /*
290  * Copy a user-provided string, returning an error if truncation would occur.
291  * Avoid scanning past "sz" bytes in the source string since there's no
292  * guarantee that it's nul-terminated.
293  */
294 static int
295 pf_user_strcpy(char *dst, const char *src, size_t sz)
296 {
297 	if (strnlen(src, sz) == sz)
298 		return (EINVAL);
299 	(void)strlcpy(dst, src, sz);
300 	return (0);
301 }
302 
303 static void
304 pfattach_vnet(void)
305 {
306 	u_int32_t *my_timeout = V_pf_default_rule.timeout;
307 
308 	pf_initialize();
309 	pfr_initialize();
310 	pfi_initialize_vnet();
311 	pf_normalize_init();
312 	pf_syncookies_init();
313 
314 	V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
315 	V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
316 
317 	RB_INIT(&V_pf_anchors);
318 	pf_init_kruleset(&pf_main_ruleset);
319 
320 	pf_init_keth(V_pf_keth);
321 
322 	/* default rule should never be garbage collected */
323 	V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
324 #ifdef PF_DEFAULT_TO_DROP
325 	V_pf_default_rule.action = PF_DROP;
326 #else
327 	V_pf_default_rule.action = PF_PASS;
328 #endif
329 	V_pf_default_rule.nr = -1;
330 	V_pf_default_rule.rtableid = -1;
331 
332 	pf_counter_u64_init(&V_pf_default_rule.evaluations, M_WAITOK);
333 	for (int i = 0; i < 2; i++) {
334 		pf_counter_u64_init(&V_pf_default_rule.packets[i], M_WAITOK);
335 		pf_counter_u64_init(&V_pf_default_rule.bytes[i], M_WAITOK);
336 	}
337 	V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
338 	V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
339 	V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
340 
341 #ifdef PF_WANT_32_TO_64_COUNTER
342 	V_pf_kifmarker = malloc(sizeof(*V_pf_kifmarker), PFI_MTYPE, M_WAITOK | M_ZERO);
343 	V_pf_rulemarker = malloc(sizeof(*V_pf_rulemarker), M_PFRULE, M_WAITOK | M_ZERO);
344 	PF_RULES_WLOCK();
345 	LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist);
346 	LIST_INSERT_HEAD(&V_pf_allrulelist, &V_pf_default_rule, allrulelist);
347 	V_pf_allrulecount++;
348 	LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist);
349 	PF_RULES_WUNLOCK();
350 #endif
351 
352 	/* initialize default timeouts */
353 	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
354 	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
355 	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
356 	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
357 	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
358 	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
359 	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
360 	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
361 	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
362 	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
363 	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
364 	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
365 	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
366 	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
367 	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
368 	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
369 	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
370 	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
371 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
372 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
373 
374 	bzero(&V_pf_status, sizeof(V_pf_status));
375 	V_pf_status.debug = PF_DEBUG_URGENT;
376 
377 	V_pf_pfil_hooked = 0;
378 	V_pf_pfil_eth_hooked = 0;
379 
380 	/* XXX do our best to avoid a conflict */
381 	V_pf_status.hostid = arc4random();
382 
383 	for (int i = 0; i < PFRES_MAX; i++)
384 		V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
385 	for (int i = 0; i < KLCNT_MAX; i++)
386 		V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
387 	for (int i = 0; i < FCNT_MAX; i++)
388 		pf_counter_u64_init(&V_pf_status.fcounters[i], M_WAITOK);
389 	for (int i = 0; i < SCNT_MAX; i++)
390 		V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
391 
392 	if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET,
393 	    INTR_MPSAFE, &V_pf_swi_cookie) != 0)
394 		/* XXXGL: leaked all above. */
395 		return;
396 }
397 
398 static struct pf_kpool *
399 pf_get_kpool(const char *anchor, u_int32_t ticket, u_int8_t rule_action,
400     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
401     u_int8_t check_ticket)
402 {
403 	struct pf_kruleset	*ruleset;
404 	struct pf_krule		*rule;
405 	int			 rs_num;
406 
407 	ruleset = pf_find_kruleset(anchor);
408 	if (ruleset == NULL)
409 		return (NULL);
410 	rs_num = pf_get_ruleset_number(rule_action);
411 	if (rs_num >= PF_RULESET_MAX)
412 		return (NULL);
413 	if (active) {
414 		if (check_ticket && ticket !=
415 		    ruleset->rules[rs_num].active.ticket)
416 			return (NULL);
417 		if (r_last)
418 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
419 			    pf_krulequeue);
420 		else
421 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
422 	} else {
423 		if (check_ticket && ticket !=
424 		    ruleset->rules[rs_num].inactive.ticket)
425 			return (NULL);
426 		if (r_last)
427 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
428 			    pf_krulequeue);
429 		else
430 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
431 	}
432 	if (!r_last) {
433 		while ((rule != NULL) && (rule->nr != rule_number))
434 			rule = TAILQ_NEXT(rule, entries);
435 	}
436 	if (rule == NULL)
437 		return (NULL);
438 
439 	return (&rule->rpool);
440 }
441 
442 static void
443 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb)
444 {
445 	struct pf_kpooladdr	*mv_pool_pa;
446 
447 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
448 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
449 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
450 	}
451 }
452 
453 static void
454 pf_empty_kpool(struct pf_kpalist *poola)
455 {
456 	struct pf_kpooladdr *pa;
457 
458 	while ((pa = TAILQ_FIRST(poola)) != NULL) {
459 		switch (pa->addr.type) {
460 		case PF_ADDR_DYNIFTL:
461 			pfi_dynaddr_remove(pa->addr.p.dyn);
462 			break;
463 		case PF_ADDR_TABLE:
464 			/* XXX: this could be unfinished pooladdr on pabuf */
465 			if (pa->addr.p.tbl != NULL)
466 				pfr_detach_table(pa->addr.p.tbl);
467 			break;
468 		}
469 		if (pa->kif)
470 			pfi_kkif_unref(pa->kif);
471 		TAILQ_REMOVE(poola, pa, entries);
472 		free(pa, M_PFRULE);
473 	}
474 }
475 
476 static void
477 pf_unlink_rule_locked(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
478 {
479 
480 	PF_RULES_WASSERT();
481 	PF_UNLNKDRULES_ASSERT();
482 
483 	TAILQ_REMOVE(rulequeue, rule, entries);
484 
485 	rule->rule_ref |= PFRULE_REFS;
486 	TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
487 }
488 
489 static void
490 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
491 {
492 
493 	PF_RULES_WASSERT();
494 
495 	PF_UNLNKDRULES_LOCK();
496 	pf_unlink_rule_locked(rulequeue, rule);
497 	PF_UNLNKDRULES_UNLOCK();
498 }
499 
500 static void
501 pf_free_eth_rule(struct pf_keth_rule *rule)
502 {
503 	PF_RULES_WASSERT();
504 
505 	if (rule == NULL)
506 		return;
507 
508 	if (rule->tag)
509 		tag_unref(&V_pf_tags, rule->tag);
510 #ifdef ALTQ
511 	pf_qid_unref(rule->qid);
512 #endif
513 
514 	if (rule->kif)
515 		pfi_kkif_unref(rule->kif);
516 
517 	counter_u64_free(rule->evaluations);
518 	for (int i = 0; i < 2; i++) {
519 		counter_u64_free(rule->packets[i]);
520 		counter_u64_free(rule->bytes[i]);
521 	}
522 	pf_keth_anchor_remove(rule);
523 
524 	free(rule, M_PFRULE);
525 }
526 
527 void
528 pf_free_rule(struct pf_krule *rule)
529 {
530 
531 	PF_RULES_WASSERT();
532 
533 	if (rule->tag)
534 		tag_unref(&V_pf_tags, rule->tag);
535 	if (rule->match_tag)
536 		tag_unref(&V_pf_tags, rule->match_tag);
537 #ifdef ALTQ
538 	if (rule->pqid != rule->qid)
539 		pf_qid_unref(rule->pqid);
540 	pf_qid_unref(rule->qid);
541 #endif
542 	switch (rule->src.addr.type) {
543 	case PF_ADDR_DYNIFTL:
544 		pfi_dynaddr_remove(rule->src.addr.p.dyn);
545 		break;
546 	case PF_ADDR_TABLE:
547 		pfr_detach_table(rule->src.addr.p.tbl);
548 		break;
549 	}
550 	switch (rule->dst.addr.type) {
551 	case PF_ADDR_DYNIFTL:
552 		pfi_dynaddr_remove(rule->dst.addr.p.dyn);
553 		break;
554 	case PF_ADDR_TABLE:
555 		pfr_detach_table(rule->dst.addr.p.tbl);
556 		break;
557 	}
558 	if (rule->overload_tbl)
559 		pfr_detach_table(rule->overload_tbl);
560 	if (rule->kif)
561 		pfi_kkif_unref(rule->kif);
562 	pf_kanchor_remove(rule);
563 	pf_empty_kpool(&rule->rpool.list);
564 
565 	pf_krule_free(rule);
566 }
567 
568 static void
569 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
570     unsigned int default_size)
571 {
572 	unsigned int i;
573 	unsigned int hashsize;
574 
575 	if (*tunable_size == 0 || !powerof2(*tunable_size))
576 		*tunable_size = default_size;
577 
578 	hashsize = *tunable_size;
579 	ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
580 	    M_WAITOK);
581 	ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
582 	    M_WAITOK);
583 	ts->mask = hashsize - 1;
584 	ts->seed = arc4random();
585 	for (i = 0; i < hashsize; i++) {
586 		TAILQ_INIT(&ts->namehash[i]);
587 		TAILQ_INIT(&ts->taghash[i]);
588 	}
589 	BIT_FILL(TAGID_MAX, &ts->avail);
590 }
591 
592 static void
593 pf_cleanup_tagset(struct pf_tagset *ts)
594 {
595 	unsigned int i;
596 	unsigned int hashsize;
597 	struct pf_tagname *t, *tmp;
598 
599 	/*
600 	 * Only need to clean up one of the hashes as each tag is hashed
601 	 * into each table.
602 	 */
603 	hashsize = ts->mask + 1;
604 	for (i = 0; i < hashsize; i++)
605 		TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
606 			uma_zfree(V_pf_tag_z, t);
607 
608 	free(ts->namehash, M_PFHASH);
609 	free(ts->taghash, M_PFHASH);
610 }
611 
612 static uint16_t
613 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
614 {
615 	size_t len;
616 
617 	len = strnlen(tagname, PF_TAG_NAME_SIZE - 1);
618 	return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask);
619 }
620 
621 static uint16_t
622 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
623 {
624 
625 	return (tag & ts->mask);
626 }
627 
628 static u_int16_t
629 tagname2tag(struct pf_tagset *ts, const char *tagname)
630 {
631 	struct pf_tagname	*tag;
632 	u_int32_t		 index;
633 	u_int16_t		 new_tagid;
634 
635 	PF_RULES_WASSERT();
636 
637 	index = tagname2hashindex(ts, tagname);
638 	TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
639 		if (strcmp(tagname, tag->name) == 0) {
640 			tag->ref++;
641 			return (tag->tag);
642 		}
643 
644 	/*
645 	 * new entry
646 	 *
647 	 * to avoid fragmentation, we do a linear search from the beginning
648 	 * and take the first free slot we find.
649 	 */
650 	new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
651 	/*
652 	 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
653 	 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
654 	 * set.  It may also return a bit number greater than TAGID_MAX due
655 	 * to rounding of the number of bits in the vector up to a multiple
656 	 * of the vector word size at declaration/allocation time.
657 	 */
658 	if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
659 		return (0);
660 
661 	/* Mark the tag as in use.  Bits are 0-based for BIT_CLR() */
662 	BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
663 
664 	/* allocate and fill new struct pf_tagname */
665 	tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
666 	if (tag == NULL)
667 		return (0);
668 	strlcpy(tag->name, tagname, sizeof(tag->name));
669 	tag->tag = new_tagid;
670 	tag->ref = 1;
671 
672 	/* Insert into namehash */
673 	TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
674 
675 	/* Insert into taghash */
676 	index = tag2hashindex(ts, new_tagid);
677 	TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
678 
679 	return (tag->tag);
680 }
681 
682 static void
683 tag_unref(struct pf_tagset *ts, u_int16_t tag)
684 {
685 	struct pf_tagname	*t;
686 	uint16_t		 index;
687 
688 	PF_RULES_WASSERT();
689 
690 	index = tag2hashindex(ts, tag);
691 	TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
692 		if (tag == t->tag) {
693 			if (--t->ref == 0) {
694 				TAILQ_REMOVE(&ts->taghash[index], t,
695 				    taghash_entries);
696 				index = tagname2hashindex(ts, t->name);
697 				TAILQ_REMOVE(&ts->namehash[index], t,
698 				    namehash_entries);
699 				/* Bits are 0-based for BIT_SET() */
700 				BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
701 				uma_zfree(V_pf_tag_z, t);
702 			}
703 			break;
704 		}
705 }
706 
707 static uint16_t
708 pf_tagname2tag(const char *tagname)
709 {
710 	return (tagname2tag(&V_pf_tags, tagname));
711 }
712 
713 static int
714 pf_begin_eth(uint32_t *ticket, const char *anchor)
715 {
716 	struct pf_keth_rule *rule, *tmp;
717 	struct pf_keth_ruleset *rs;
718 
719 	PF_RULES_WASSERT();
720 
721 	rs = pf_find_or_create_keth_ruleset(anchor);
722 	if (rs == NULL)
723 		return (EINVAL);
724 
725 	if (rs->inactive.open)
726 		/* We may be waiting for NET_EPOCH_CALL(pf_rollback_eth_cb) to
727 		 * finish. */
728 		return (EBUSY);
729 
730 	/* Purge old inactive rules. */
731 	TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
732 	    tmp) {
733 		TAILQ_REMOVE(rs->inactive.rules, rule,
734 		    entries);
735 		pf_free_eth_rule(rule);
736 	}
737 
738 	*ticket = ++rs->inactive.ticket;
739 	rs->inactive.open = 1;
740 
741 	return (0);
742 }
743 
744 static void
745 pf_rollback_eth_cb(struct epoch_context *ctx)
746 {
747 	struct pf_keth_ruleset *rs;
748 
749 	rs = __containerof(ctx, struct pf_keth_ruleset, epoch_ctx);
750 
751 	CURVNET_SET(rs->vnet);
752 
753 	PF_RULES_WLOCK();
754 	pf_rollback_eth(rs->inactive.ticket,
755 	    rs->anchor ? rs->anchor->path : "");
756 	PF_RULES_WUNLOCK();
757 
758 	CURVNET_RESTORE();
759 }
760 
761 static int
762 pf_rollback_eth(uint32_t ticket, const char *anchor)
763 {
764 	struct pf_keth_rule *rule, *tmp;
765 	struct pf_keth_ruleset *rs;
766 
767 	PF_RULES_WASSERT();
768 
769 	rs = pf_find_keth_ruleset(anchor);
770 	if (rs == NULL)
771 		return (EINVAL);
772 
773 	if (!rs->inactive.open ||
774 	    ticket != rs->inactive.ticket)
775 		return (0);
776 
777 	/* Purge old inactive rules. */
778 	TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
779 	    tmp) {
780 		TAILQ_REMOVE(rs->inactive.rules, rule, entries);
781 		pf_free_eth_rule(rule);
782 	}
783 
784 	rs->inactive.open = 0;
785 
786 	pf_remove_if_empty_keth_ruleset(rs);
787 
788 	return (0);
789 }
790 
791 #define	PF_SET_SKIP_STEPS(i)					\
792 	do {							\
793 		while (head[i] != cur) {			\
794 			head[i]->skip[i].ptr = cur;		\
795 			head[i] = TAILQ_NEXT(head[i], entries);	\
796 		}						\
797 	} while (0)
798 
799 static void
800 pf_eth_calc_skip_steps(struct pf_keth_ruleq *rules)
801 {
802 	struct pf_keth_rule *cur, *prev, *head[PFE_SKIP_COUNT];
803 	int i;
804 
805 	cur = TAILQ_FIRST(rules);
806 	prev = cur;
807 	for (i = 0; i < PFE_SKIP_COUNT; ++i)
808 		head[i] = cur;
809 	while (cur != NULL) {
810 		if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
811 			PF_SET_SKIP_STEPS(PFE_SKIP_IFP);
812 		if (cur->direction != prev->direction)
813 			PF_SET_SKIP_STEPS(PFE_SKIP_DIR);
814 		if (cur->proto != prev->proto)
815 			PF_SET_SKIP_STEPS(PFE_SKIP_PROTO);
816 		if (memcmp(&cur->src, &prev->src, sizeof(cur->src)) != 0)
817 			PF_SET_SKIP_STEPS(PFE_SKIP_SRC_ADDR);
818 		if (memcmp(&cur->dst, &prev->dst, sizeof(cur->dst)) != 0)
819 			PF_SET_SKIP_STEPS(PFE_SKIP_DST_ADDR);
820 
821 		prev = cur;
822 		cur = TAILQ_NEXT(cur, entries);
823 	}
824 	for (i = 0; i < PFE_SKIP_COUNT; ++i)
825 		PF_SET_SKIP_STEPS(i);
826 }
827 
828 static int
829 pf_commit_eth(uint32_t ticket, const char *anchor)
830 {
831 	struct pf_keth_ruleq *rules;
832 	struct pf_keth_ruleset *rs;
833 
834 	rs = pf_find_keth_ruleset(anchor);
835 	if (rs == NULL) {
836 		return (EINVAL);
837 	}
838 
839 	if (!rs->inactive.open ||
840 	    ticket != rs->inactive.ticket)
841 		return (EBUSY);
842 
843 	PF_RULES_WASSERT();
844 
845 	pf_eth_calc_skip_steps(rs->inactive.rules);
846 
847 	rules = rs->active.rules;
848 	ck_pr_store_ptr(&rs->active.rules, rs->inactive.rules);
849 	rs->inactive.rules = rules;
850 	rs->inactive.ticket = rs->active.ticket;
851 
852 	/* Clean up inactive rules (i.e. previously active rules), only when
853 	 * we're sure they're no longer used. */
854 	NET_EPOCH_CALL(pf_rollback_eth_cb, &rs->epoch_ctx);
855 
856 	return (0);
857 }
858 
859 #ifdef ALTQ
860 static uint16_t
861 pf_qname2qid(const char *qname)
862 {
863 	return (tagname2tag(&V_pf_qids, qname));
864 }
865 
866 static void
867 pf_qid_unref(uint16_t qid)
868 {
869 	tag_unref(&V_pf_qids, qid);
870 }
871 
872 static int
873 pf_begin_altq(u_int32_t *ticket)
874 {
875 	struct pf_altq	*altq, *tmp;
876 	int		 error = 0;
877 
878 	PF_RULES_WASSERT();
879 
880 	/* Purge the old altq lists */
881 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
882 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
883 			/* detach and destroy the discipline */
884 			error = altq_remove(altq);
885 		}
886 		free(altq, M_PFALTQ);
887 	}
888 	TAILQ_INIT(V_pf_altq_ifs_inactive);
889 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
890 		pf_qid_unref(altq->qid);
891 		free(altq, M_PFALTQ);
892 	}
893 	TAILQ_INIT(V_pf_altqs_inactive);
894 	if (error)
895 		return (error);
896 	*ticket = ++V_ticket_altqs_inactive;
897 	V_altqs_inactive_open = 1;
898 	return (0);
899 }
900 
901 static int
902 pf_rollback_altq(u_int32_t ticket)
903 {
904 	struct pf_altq	*altq, *tmp;
905 	int		 error = 0;
906 
907 	PF_RULES_WASSERT();
908 
909 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
910 		return (0);
911 	/* Purge the old altq lists */
912 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
913 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
914 			/* detach and destroy the discipline */
915 			error = altq_remove(altq);
916 		}
917 		free(altq, M_PFALTQ);
918 	}
919 	TAILQ_INIT(V_pf_altq_ifs_inactive);
920 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
921 		pf_qid_unref(altq->qid);
922 		free(altq, M_PFALTQ);
923 	}
924 	TAILQ_INIT(V_pf_altqs_inactive);
925 	V_altqs_inactive_open = 0;
926 	return (error);
927 }
928 
929 static int
930 pf_commit_altq(u_int32_t ticket)
931 {
932 	struct pf_altqqueue	*old_altqs, *old_altq_ifs;
933 	struct pf_altq		*altq, *tmp;
934 	int			 err, error = 0;
935 
936 	PF_RULES_WASSERT();
937 
938 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
939 		return (EBUSY);
940 
941 	/* swap altqs, keep the old. */
942 	old_altqs = V_pf_altqs_active;
943 	old_altq_ifs = V_pf_altq_ifs_active;
944 	V_pf_altqs_active = V_pf_altqs_inactive;
945 	V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
946 	V_pf_altqs_inactive = old_altqs;
947 	V_pf_altq_ifs_inactive = old_altq_ifs;
948 	V_ticket_altqs_active = V_ticket_altqs_inactive;
949 
950 	/* Attach new disciplines */
951 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
952 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
953 			/* attach the discipline */
954 			error = altq_pfattach(altq);
955 			if (error == 0 && V_pf_altq_running)
956 				error = pf_enable_altq(altq);
957 			if (error != 0)
958 				return (error);
959 		}
960 	}
961 
962 	/* Purge the old altq lists */
963 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
964 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
965 			/* detach and destroy the discipline */
966 			if (V_pf_altq_running)
967 				error = pf_disable_altq(altq);
968 			err = altq_pfdetach(altq);
969 			if (err != 0 && error == 0)
970 				error = err;
971 			err = altq_remove(altq);
972 			if (err != 0 && error == 0)
973 				error = err;
974 		}
975 		free(altq, M_PFALTQ);
976 	}
977 	TAILQ_INIT(V_pf_altq_ifs_inactive);
978 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
979 		pf_qid_unref(altq->qid);
980 		free(altq, M_PFALTQ);
981 	}
982 	TAILQ_INIT(V_pf_altqs_inactive);
983 
984 	V_altqs_inactive_open = 0;
985 	return (error);
986 }
987 
988 static int
989 pf_enable_altq(struct pf_altq *altq)
990 {
991 	struct ifnet		*ifp;
992 	struct tb_profile	 tb;
993 	int			 error = 0;
994 
995 	if ((ifp = ifunit(altq->ifname)) == NULL)
996 		return (EINVAL);
997 
998 	if (ifp->if_snd.altq_type != ALTQT_NONE)
999 		error = altq_enable(&ifp->if_snd);
1000 
1001 	/* set tokenbucket regulator */
1002 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
1003 		tb.rate = altq->ifbandwidth;
1004 		tb.depth = altq->tbrsize;
1005 		error = tbr_set(&ifp->if_snd, &tb);
1006 	}
1007 
1008 	return (error);
1009 }
1010 
1011 static int
1012 pf_disable_altq(struct pf_altq *altq)
1013 {
1014 	struct ifnet		*ifp;
1015 	struct tb_profile	 tb;
1016 	int			 error;
1017 
1018 	if ((ifp = ifunit(altq->ifname)) == NULL)
1019 		return (EINVAL);
1020 
1021 	/*
1022 	 * when the discipline is no longer referenced, it was overridden
1023 	 * by a new one.  if so, just return.
1024 	 */
1025 	if (altq->altq_disc != ifp->if_snd.altq_disc)
1026 		return (0);
1027 
1028 	error = altq_disable(&ifp->if_snd);
1029 
1030 	if (error == 0) {
1031 		/* clear tokenbucket regulator */
1032 		tb.rate = 0;
1033 		error = tbr_set(&ifp->if_snd, &tb);
1034 	}
1035 
1036 	return (error);
1037 }
1038 
1039 static int
1040 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
1041     struct pf_altq *altq)
1042 {
1043 	struct ifnet	*ifp1;
1044 	int		 error = 0;
1045 
1046 	/* Deactivate the interface in question */
1047 	altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
1048 	if ((ifp1 = ifunit(altq->ifname)) == NULL ||
1049 	    (remove && ifp1 == ifp)) {
1050 		altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
1051 	} else {
1052 		error = altq_add(ifp1, altq);
1053 
1054 		if (ticket != V_ticket_altqs_inactive)
1055 			error = EBUSY;
1056 
1057 		if (error)
1058 			free(altq, M_PFALTQ);
1059 	}
1060 
1061 	return (error);
1062 }
1063 
1064 void
1065 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
1066 {
1067 	struct pf_altq	*a1, *a2, *a3;
1068 	u_int32_t	 ticket;
1069 	int		 error = 0;
1070 
1071 	/*
1072 	 * No need to re-evaluate the configuration for events on interfaces
1073 	 * that do not support ALTQ, as it's not possible for such
1074 	 * interfaces to be part of the configuration.
1075 	 */
1076 	if (!ALTQ_IS_READY(&ifp->if_snd))
1077 		return;
1078 
1079 	/* Interrupt userland queue modifications */
1080 	if (V_altqs_inactive_open)
1081 		pf_rollback_altq(V_ticket_altqs_inactive);
1082 
1083 	/* Start new altq ruleset */
1084 	if (pf_begin_altq(&ticket))
1085 		return;
1086 
1087 	/* Copy the current active set */
1088 	TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
1089 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1090 		if (a2 == NULL) {
1091 			error = ENOMEM;
1092 			break;
1093 		}
1094 		bcopy(a1, a2, sizeof(struct pf_altq));
1095 
1096 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1097 		if (error)
1098 			break;
1099 
1100 		TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
1101 	}
1102 	if (error)
1103 		goto out;
1104 	TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
1105 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1106 		if (a2 == NULL) {
1107 			error = ENOMEM;
1108 			break;
1109 		}
1110 		bcopy(a1, a2, sizeof(struct pf_altq));
1111 
1112 		if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
1113 			error = EBUSY;
1114 			free(a2, M_PFALTQ);
1115 			break;
1116 		}
1117 		a2->altq_disc = NULL;
1118 		TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
1119 			if (strncmp(a3->ifname, a2->ifname,
1120 				IFNAMSIZ) == 0) {
1121 				a2->altq_disc = a3->altq_disc;
1122 				break;
1123 			}
1124 		}
1125 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1126 		if (error)
1127 			break;
1128 
1129 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
1130 	}
1131 
1132 out:
1133 	if (error != 0)
1134 		pf_rollback_altq(ticket);
1135 	else
1136 		pf_commit_altq(ticket);
1137 }
1138 #endif /* ALTQ */
1139 
1140 static int
1141 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
1142 {
1143 	struct pf_kruleset	*rs;
1144 	struct pf_krule		*rule;
1145 
1146 	PF_RULES_WASSERT();
1147 
1148 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1149 		return (EINVAL);
1150 	rs = pf_find_or_create_kruleset(anchor);
1151 	if (rs == NULL)
1152 		return (EINVAL);
1153 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1154 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1155 		rs->rules[rs_num].inactive.rcount--;
1156 	}
1157 	*ticket = ++rs->rules[rs_num].inactive.ticket;
1158 	rs->rules[rs_num].inactive.open = 1;
1159 	return (0);
1160 }
1161 
1162 static int
1163 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
1164 {
1165 	struct pf_kruleset	*rs;
1166 	struct pf_krule		*rule;
1167 
1168 	PF_RULES_WASSERT();
1169 
1170 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1171 		return (EINVAL);
1172 	rs = pf_find_kruleset(anchor);
1173 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1174 	    rs->rules[rs_num].inactive.ticket != ticket)
1175 		return (0);
1176 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1177 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1178 		rs->rules[rs_num].inactive.rcount--;
1179 	}
1180 	rs->rules[rs_num].inactive.open = 0;
1181 	return (0);
1182 }
1183 
1184 #define PF_MD5_UPD(st, elm)						\
1185 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
1186 
1187 #define PF_MD5_UPD_STR(st, elm)						\
1188 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
1189 
1190 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
1191 		(stor) = htonl((st)->elm);				\
1192 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
1193 } while (0)
1194 
1195 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
1196 		(stor) = htons((st)->elm);				\
1197 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
1198 } while (0)
1199 
1200 static void
1201 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
1202 {
1203 	PF_MD5_UPD(pfr, addr.type);
1204 	switch (pfr->addr.type) {
1205 		case PF_ADDR_DYNIFTL:
1206 			PF_MD5_UPD(pfr, addr.v.ifname);
1207 			PF_MD5_UPD(pfr, addr.iflags);
1208 			break;
1209 		case PF_ADDR_TABLE:
1210 			PF_MD5_UPD(pfr, addr.v.tblname);
1211 			break;
1212 		case PF_ADDR_ADDRMASK:
1213 			/* XXX ignore af? */
1214 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
1215 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
1216 			break;
1217 	}
1218 
1219 	PF_MD5_UPD(pfr, port[0]);
1220 	PF_MD5_UPD(pfr, port[1]);
1221 	PF_MD5_UPD(pfr, neg);
1222 	PF_MD5_UPD(pfr, port_op);
1223 }
1224 
1225 static void
1226 pf_hash_rule(MD5_CTX *ctx, struct pf_krule *rule)
1227 {
1228 	u_int16_t x;
1229 	u_int32_t y;
1230 
1231 	pf_hash_rule_addr(ctx, &rule->src);
1232 	pf_hash_rule_addr(ctx, &rule->dst);
1233 	for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++)
1234 		PF_MD5_UPD_STR(rule, label[i]);
1235 	PF_MD5_UPD_STR(rule, ifname);
1236 	PF_MD5_UPD_STR(rule, match_tagname);
1237 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
1238 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
1239 	PF_MD5_UPD_HTONL(rule, prob, y);
1240 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
1241 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1242 	PF_MD5_UPD(rule, uid.op);
1243 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1244 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1245 	PF_MD5_UPD(rule, gid.op);
1246 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
1247 	PF_MD5_UPD(rule, action);
1248 	PF_MD5_UPD(rule, direction);
1249 	PF_MD5_UPD(rule, af);
1250 	PF_MD5_UPD(rule, quick);
1251 	PF_MD5_UPD(rule, ifnot);
1252 	PF_MD5_UPD(rule, match_tag_not);
1253 	PF_MD5_UPD(rule, natpass);
1254 	PF_MD5_UPD(rule, keep_state);
1255 	PF_MD5_UPD(rule, proto);
1256 	PF_MD5_UPD(rule, type);
1257 	PF_MD5_UPD(rule, code);
1258 	PF_MD5_UPD(rule, flags);
1259 	PF_MD5_UPD(rule, flagset);
1260 	PF_MD5_UPD(rule, allow_opts);
1261 	PF_MD5_UPD(rule, rt);
1262 	PF_MD5_UPD(rule, tos);
1263 }
1264 
1265 static bool
1266 pf_krule_compare(struct pf_krule *a, struct pf_krule *b)
1267 {
1268 	MD5_CTX		ctx[2];
1269 	u_int8_t	digest[2][PF_MD5_DIGEST_LENGTH];
1270 
1271 	MD5Init(&ctx[0]);
1272 	MD5Init(&ctx[1]);
1273 	pf_hash_rule(&ctx[0], a);
1274 	pf_hash_rule(&ctx[1], b);
1275 	MD5Final(digest[0], &ctx[0]);
1276 	MD5Final(digest[1], &ctx[1]);
1277 
1278 	return (memcmp(digest[0], digest[1], PF_MD5_DIGEST_LENGTH) == 0);
1279 }
1280 
1281 static int
1282 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1283 {
1284 	struct pf_kruleset	*rs;
1285 	struct pf_krule		*rule, **old_array, *tail;
1286 	struct pf_krulequeue	*old_rules;
1287 	int			 error;
1288 	u_int32_t		 old_rcount;
1289 
1290 	PF_RULES_WASSERT();
1291 
1292 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1293 		return (EINVAL);
1294 	rs = pf_find_kruleset(anchor);
1295 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1296 	    ticket != rs->rules[rs_num].inactive.ticket)
1297 		return (EBUSY);
1298 
1299 	/* Calculate checksum for the main ruleset */
1300 	if (rs == &pf_main_ruleset) {
1301 		error = pf_setup_pfsync_matching(rs);
1302 		if (error != 0)
1303 			return (error);
1304 	}
1305 
1306 	/* Swap rules, keep the old. */
1307 	old_rules = rs->rules[rs_num].active.ptr;
1308 	old_rcount = rs->rules[rs_num].active.rcount;
1309 	old_array = rs->rules[rs_num].active.ptr_array;
1310 
1311 	rs->rules[rs_num].active.ptr =
1312 	    rs->rules[rs_num].inactive.ptr;
1313 	rs->rules[rs_num].active.ptr_array =
1314 	    rs->rules[rs_num].inactive.ptr_array;
1315 	rs->rules[rs_num].active.rcount =
1316 	    rs->rules[rs_num].inactive.rcount;
1317 
1318 	/* Attempt to preserve counter information. */
1319 	if (V_pf_status.keep_counters) {
1320 		TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr,
1321 		    entries) {
1322 			tail = TAILQ_FIRST(old_rules);
1323 			while ((tail != NULL) && ! pf_krule_compare(tail, rule))
1324 				tail = TAILQ_NEXT(tail, entries);
1325 			if (tail != NULL) {
1326 				pf_counter_u64_critical_enter();
1327 				pf_counter_u64_add_protected(&rule->evaluations,
1328 				    pf_counter_u64_fetch(&tail->evaluations));
1329 				pf_counter_u64_add_protected(&rule->packets[0],
1330 				    pf_counter_u64_fetch(&tail->packets[0]));
1331 				pf_counter_u64_add_protected(&rule->packets[1],
1332 				    pf_counter_u64_fetch(&tail->packets[1]));
1333 				pf_counter_u64_add_protected(&rule->bytes[0],
1334 				    pf_counter_u64_fetch(&tail->bytes[0]));
1335 				pf_counter_u64_add_protected(&rule->bytes[1],
1336 				    pf_counter_u64_fetch(&tail->bytes[1]));
1337 				pf_counter_u64_critical_exit();
1338 			}
1339 		}
1340 	}
1341 
1342 	rs->rules[rs_num].inactive.ptr = old_rules;
1343 	rs->rules[rs_num].inactive.ptr_array = old_array;
1344 	rs->rules[rs_num].inactive.rcount = old_rcount;
1345 
1346 	rs->rules[rs_num].active.ticket =
1347 	    rs->rules[rs_num].inactive.ticket;
1348 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1349 
1350 	/* Purge the old rule list. */
1351 	PF_UNLNKDRULES_LOCK();
1352 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1353 		pf_unlink_rule_locked(old_rules, rule);
1354 	PF_UNLNKDRULES_UNLOCK();
1355 	if (rs->rules[rs_num].inactive.ptr_array)
1356 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1357 	rs->rules[rs_num].inactive.ptr_array = NULL;
1358 	rs->rules[rs_num].inactive.rcount = 0;
1359 	rs->rules[rs_num].inactive.open = 0;
1360 	pf_remove_if_empty_kruleset(rs);
1361 
1362 	return (0);
1363 }
1364 
1365 static int
1366 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1367 {
1368 	MD5_CTX			 ctx;
1369 	struct pf_krule		*rule;
1370 	int			 rs_cnt;
1371 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
1372 
1373 	MD5Init(&ctx);
1374 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1375 		/* XXX PF_RULESET_SCRUB as well? */
1376 		if (rs_cnt == PF_RULESET_SCRUB)
1377 			continue;
1378 
1379 		if (rs->rules[rs_cnt].inactive.ptr_array)
1380 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1381 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
1382 
1383 		if (rs->rules[rs_cnt].inactive.rcount) {
1384 			rs->rules[rs_cnt].inactive.ptr_array =
1385 			    malloc(sizeof(caddr_t) *
1386 			    rs->rules[rs_cnt].inactive.rcount,
1387 			    M_TEMP, M_NOWAIT);
1388 
1389 			if (!rs->rules[rs_cnt].inactive.ptr_array)
1390 				return (ENOMEM);
1391 		}
1392 
1393 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1394 		    entries) {
1395 			pf_hash_rule(&ctx, rule);
1396 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1397 		}
1398 	}
1399 
1400 	MD5Final(digest, &ctx);
1401 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1402 	return (0);
1403 }
1404 
1405 static int
1406 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1407     sa_family_t af)
1408 {
1409 	int error = 0;
1410 
1411 	switch (addr->type) {
1412 	case PF_ADDR_TABLE:
1413 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1414 		if (addr->p.tbl == NULL)
1415 			error = ENOMEM;
1416 		break;
1417 	case PF_ADDR_DYNIFTL:
1418 		error = pfi_dynaddr_setup(addr, af);
1419 		break;
1420 	}
1421 
1422 	return (error);
1423 }
1424 
1425 static void
1426 pf_addr_copyout(struct pf_addr_wrap *addr)
1427 {
1428 
1429 	switch (addr->type) {
1430 	case PF_ADDR_DYNIFTL:
1431 		pfi_dynaddr_copyout(addr);
1432 		break;
1433 	case PF_ADDR_TABLE:
1434 		pf_tbladdr_copyout(addr);
1435 		break;
1436 	}
1437 }
1438 
1439 static void
1440 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
1441 {
1442 	int	secs = time_uptime, diff;
1443 
1444 	bzero(out, sizeof(struct pf_src_node));
1445 
1446 	bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
1447 	bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
1448 
1449 	if (in->rule.ptr != NULL)
1450 		out->rule.nr = in->rule.ptr->nr;
1451 
1452 	for (int i = 0; i < 2; i++) {
1453 		out->bytes[i] = counter_u64_fetch(in->bytes[i]);
1454 		out->packets[i] = counter_u64_fetch(in->packets[i]);
1455 	}
1456 
1457 	out->states = in->states;
1458 	out->conn = in->conn;
1459 	out->af = in->af;
1460 	out->ruletype = in->ruletype;
1461 
1462 	out->creation = secs - in->creation;
1463 	if (out->expire > secs)
1464 		out->expire -= secs;
1465 	else
1466 		out->expire = 0;
1467 
1468 	/* Adjust the connection rate estimate. */
1469 	diff = secs - in->conn_rate.last;
1470 	if (diff >= in->conn_rate.seconds)
1471 		out->conn_rate.count = 0;
1472 	else
1473 		out->conn_rate.count -=
1474 		    in->conn_rate.count * diff /
1475 		    in->conn_rate.seconds;
1476 }
1477 
1478 #ifdef ALTQ
1479 /*
1480  * Handle export of struct pf_kaltq to user binaries that may be using any
1481  * version of struct pf_altq.
1482  */
1483 static int
1484 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1485 {
1486 	u_int32_t version;
1487 
1488 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1489 		version = 0;
1490 	else
1491 		version = pa->version;
1492 
1493 	if (version > PFIOC_ALTQ_VERSION)
1494 		return (EINVAL);
1495 
1496 #define ASSIGN(x) exported_q->x = q->x
1497 #define COPY(x) \
1498 	bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1499 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1500 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1501 
1502 	switch (version) {
1503 	case 0: {
1504 		struct pf_altq_v0 *exported_q =
1505 		    &((struct pfioc_altq_v0 *)pa)->altq;
1506 
1507 		COPY(ifname);
1508 
1509 		ASSIGN(scheduler);
1510 		ASSIGN(tbrsize);
1511 		exported_q->tbrsize = SATU16(q->tbrsize);
1512 		exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1513 
1514 		COPY(qname);
1515 		COPY(parent);
1516 		ASSIGN(parent_qid);
1517 		exported_q->bandwidth = SATU32(q->bandwidth);
1518 		ASSIGN(priority);
1519 		ASSIGN(local_flags);
1520 
1521 		ASSIGN(qlimit);
1522 		ASSIGN(flags);
1523 
1524 		if (q->scheduler == ALTQT_HFSC) {
1525 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1526 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1527 			    SATU32(q->pq_u.hfsc_opts.x)
1528 
1529 			ASSIGN_OPT_SATU32(rtsc_m1);
1530 			ASSIGN_OPT(rtsc_d);
1531 			ASSIGN_OPT_SATU32(rtsc_m2);
1532 
1533 			ASSIGN_OPT_SATU32(lssc_m1);
1534 			ASSIGN_OPT(lssc_d);
1535 			ASSIGN_OPT_SATU32(lssc_m2);
1536 
1537 			ASSIGN_OPT_SATU32(ulsc_m1);
1538 			ASSIGN_OPT(ulsc_d);
1539 			ASSIGN_OPT_SATU32(ulsc_m2);
1540 
1541 			ASSIGN_OPT(flags);
1542 
1543 #undef ASSIGN_OPT
1544 #undef ASSIGN_OPT_SATU32
1545 		} else
1546 			COPY(pq_u);
1547 
1548 		ASSIGN(qid);
1549 		break;
1550 	}
1551 	case 1:	{
1552 		struct pf_altq_v1 *exported_q =
1553 		    &((struct pfioc_altq_v1 *)pa)->altq;
1554 
1555 		COPY(ifname);
1556 
1557 		ASSIGN(scheduler);
1558 		ASSIGN(tbrsize);
1559 		ASSIGN(ifbandwidth);
1560 
1561 		COPY(qname);
1562 		COPY(parent);
1563 		ASSIGN(parent_qid);
1564 		ASSIGN(bandwidth);
1565 		ASSIGN(priority);
1566 		ASSIGN(local_flags);
1567 
1568 		ASSIGN(qlimit);
1569 		ASSIGN(flags);
1570 		COPY(pq_u);
1571 
1572 		ASSIGN(qid);
1573 		break;
1574 	}
1575 	default:
1576 		panic("%s: unhandled struct pfioc_altq version", __func__);
1577 		break;
1578 	}
1579 
1580 #undef ASSIGN
1581 #undef COPY
1582 #undef SATU16
1583 #undef SATU32
1584 
1585 	return (0);
1586 }
1587 
1588 /*
1589  * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1590  * that may be using any version of it.
1591  */
1592 static int
1593 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1594 {
1595 	u_int32_t version;
1596 
1597 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1598 		version = 0;
1599 	else
1600 		version = pa->version;
1601 
1602 	if (version > PFIOC_ALTQ_VERSION)
1603 		return (EINVAL);
1604 
1605 #define ASSIGN(x) q->x = imported_q->x
1606 #define COPY(x) \
1607 	bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1608 
1609 	switch (version) {
1610 	case 0: {
1611 		struct pf_altq_v0 *imported_q =
1612 		    &((struct pfioc_altq_v0 *)pa)->altq;
1613 
1614 		COPY(ifname);
1615 
1616 		ASSIGN(scheduler);
1617 		ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1618 		ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1619 
1620 		COPY(qname);
1621 		COPY(parent);
1622 		ASSIGN(parent_qid);
1623 		ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1624 		ASSIGN(priority);
1625 		ASSIGN(local_flags);
1626 
1627 		ASSIGN(qlimit);
1628 		ASSIGN(flags);
1629 
1630 		if (imported_q->scheduler == ALTQT_HFSC) {
1631 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1632 
1633 			/*
1634 			 * The m1 and m2 parameters are being copied from
1635 			 * 32-bit to 64-bit.
1636 			 */
1637 			ASSIGN_OPT(rtsc_m1);
1638 			ASSIGN_OPT(rtsc_d);
1639 			ASSIGN_OPT(rtsc_m2);
1640 
1641 			ASSIGN_OPT(lssc_m1);
1642 			ASSIGN_OPT(lssc_d);
1643 			ASSIGN_OPT(lssc_m2);
1644 
1645 			ASSIGN_OPT(ulsc_m1);
1646 			ASSIGN_OPT(ulsc_d);
1647 			ASSIGN_OPT(ulsc_m2);
1648 
1649 			ASSIGN_OPT(flags);
1650 
1651 #undef ASSIGN_OPT
1652 		} else
1653 			COPY(pq_u);
1654 
1655 		ASSIGN(qid);
1656 		break;
1657 	}
1658 	case 1: {
1659 		struct pf_altq_v1 *imported_q =
1660 		    &((struct pfioc_altq_v1 *)pa)->altq;
1661 
1662 		COPY(ifname);
1663 
1664 		ASSIGN(scheduler);
1665 		ASSIGN(tbrsize);
1666 		ASSIGN(ifbandwidth);
1667 
1668 		COPY(qname);
1669 		COPY(parent);
1670 		ASSIGN(parent_qid);
1671 		ASSIGN(bandwidth);
1672 		ASSIGN(priority);
1673 		ASSIGN(local_flags);
1674 
1675 		ASSIGN(qlimit);
1676 		ASSIGN(flags);
1677 		COPY(pq_u);
1678 
1679 		ASSIGN(qid);
1680 		break;
1681 	}
1682 	default:
1683 		panic("%s: unhandled struct pfioc_altq version", __func__);
1684 		break;
1685 	}
1686 
1687 #undef ASSIGN
1688 #undef COPY
1689 
1690 	return (0);
1691 }
1692 
1693 static struct pf_altq *
1694 pf_altq_get_nth_active(u_int32_t n)
1695 {
1696 	struct pf_altq		*altq;
1697 	u_int32_t		 nr;
1698 
1699 	nr = 0;
1700 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1701 		if (nr == n)
1702 			return (altq);
1703 		nr++;
1704 	}
1705 
1706 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1707 		if (nr == n)
1708 			return (altq);
1709 		nr++;
1710 	}
1711 
1712 	return (NULL);
1713 }
1714 #endif /* ALTQ */
1715 
1716 struct pf_krule *
1717 pf_krule_alloc(void)
1718 {
1719 	struct pf_krule *rule;
1720 
1721 	rule = malloc(sizeof(struct pf_krule), M_PFRULE, M_WAITOK | M_ZERO);
1722 	mtx_init(&rule->rpool.mtx, "pf_krule_pool", NULL, MTX_DEF);
1723 	return (rule);
1724 }
1725 
1726 void
1727 pf_krule_free(struct pf_krule *rule)
1728 {
1729 #ifdef PF_WANT_32_TO_64_COUNTER
1730 	bool wowned;
1731 #endif
1732 
1733 	if (rule == NULL)
1734 		return;
1735 
1736 #ifdef PF_WANT_32_TO_64_COUNTER
1737 	if (rule->allrulelinked) {
1738 		wowned = PF_RULES_WOWNED();
1739 		if (!wowned)
1740 			PF_RULES_WLOCK();
1741 		LIST_REMOVE(rule, allrulelist);
1742 		V_pf_allrulecount--;
1743 		if (!wowned)
1744 			PF_RULES_WUNLOCK();
1745 	}
1746 #endif
1747 
1748 	pf_counter_u64_deinit(&rule->evaluations);
1749 	for (int i = 0; i < 2; i++) {
1750 		pf_counter_u64_deinit(&rule->packets[i]);
1751 		pf_counter_u64_deinit(&rule->bytes[i]);
1752 	}
1753 	counter_u64_free(rule->states_cur);
1754 	counter_u64_free(rule->states_tot);
1755 	counter_u64_free(rule->src_nodes);
1756 
1757 	mtx_destroy(&rule->rpool.mtx);
1758 	free(rule, M_PFRULE);
1759 }
1760 
1761 static void
1762 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool,
1763     struct pf_pooladdr *pool)
1764 {
1765 
1766 	bzero(pool, sizeof(*pool));
1767 	bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr));
1768 	strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname));
1769 }
1770 
1771 static int
1772 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool,
1773     struct pf_kpooladdr *kpool)
1774 {
1775 	int ret;
1776 
1777 	bzero(kpool, sizeof(*kpool));
1778 	bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr));
1779 	ret = pf_user_strcpy(kpool->ifname, pool->ifname,
1780 	    sizeof(kpool->ifname));
1781 	return (ret);
1782 }
1783 
1784 static void
1785 pf_kpool_to_pool(const struct pf_kpool *kpool, struct pf_pool *pool)
1786 {
1787 	bzero(pool, sizeof(*pool));
1788 
1789 	bcopy(&kpool->key, &pool->key, sizeof(pool->key));
1790 	bcopy(&kpool->counter, &pool->counter, sizeof(pool->counter));
1791 
1792 	pool->tblidx = kpool->tblidx;
1793 	pool->proxy_port[0] = kpool->proxy_port[0];
1794 	pool->proxy_port[1] = kpool->proxy_port[1];
1795 	pool->opts = kpool->opts;
1796 }
1797 
1798 static void
1799 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool)
1800 {
1801 	_Static_assert(sizeof(pool->key) == sizeof(kpool->key), "");
1802 	_Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), "");
1803 
1804 	bcopy(&pool->key, &kpool->key, sizeof(kpool->key));
1805 	bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter));
1806 
1807 	kpool->tblidx = pool->tblidx;
1808 	kpool->proxy_port[0] = pool->proxy_port[0];
1809 	kpool->proxy_port[1] = pool->proxy_port[1];
1810 	kpool->opts = pool->opts;
1811 }
1812 
1813 static void
1814 pf_krule_to_rule(const struct pf_krule *krule, struct pf_rule *rule)
1815 {
1816 
1817 	bzero(rule, sizeof(*rule));
1818 
1819 	bcopy(&krule->src, &rule->src, sizeof(rule->src));
1820 	bcopy(&krule->dst, &rule->dst, sizeof(rule->dst));
1821 
1822 	for (int i = 0; i < PF_SKIP_COUNT; ++i) {
1823 		if (rule->skip[i].ptr == NULL)
1824 			rule->skip[i].nr = -1;
1825 		else
1826 			rule->skip[i].nr = krule->skip[i].ptr->nr;
1827 	}
1828 
1829 	strlcpy(rule->label, krule->label[0], sizeof(rule->label));
1830 	strlcpy(rule->ifname, krule->ifname, sizeof(rule->ifname));
1831 	strlcpy(rule->qname, krule->qname, sizeof(rule->qname));
1832 	strlcpy(rule->pqname, krule->pqname, sizeof(rule->pqname));
1833 	strlcpy(rule->tagname, krule->tagname, sizeof(rule->tagname));
1834 	strlcpy(rule->match_tagname, krule->match_tagname,
1835 	    sizeof(rule->match_tagname));
1836 	strlcpy(rule->overload_tblname, krule->overload_tblname,
1837 	    sizeof(rule->overload_tblname));
1838 
1839 	pf_kpool_to_pool(&krule->rpool, &rule->rpool);
1840 
1841 	rule->evaluations = pf_counter_u64_fetch(&krule->evaluations);
1842 	for (int i = 0; i < 2; i++) {
1843 		rule->packets[i] = pf_counter_u64_fetch(&krule->packets[i]);
1844 		rule->bytes[i] = pf_counter_u64_fetch(&krule->bytes[i]);
1845 	}
1846 
1847 	/* kif, anchor, overload_tbl are not copied over. */
1848 
1849 	rule->os_fingerprint = krule->os_fingerprint;
1850 
1851 	rule->rtableid = krule->rtableid;
1852 	bcopy(krule->timeout, rule->timeout, sizeof(krule->timeout));
1853 	rule->max_states = krule->max_states;
1854 	rule->max_src_nodes = krule->max_src_nodes;
1855 	rule->max_src_states = krule->max_src_states;
1856 	rule->max_src_conn = krule->max_src_conn;
1857 	rule->max_src_conn_rate.limit = krule->max_src_conn_rate.limit;
1858 	rule->max_src_conn_rate.seconds = krule->max_src_conn_rate.seconds;
1859 	rule->qid = krule->qid;
1860 	rule->pqid = krule->pqid;
1861 	rule->nr = krule->nr;
1862 	rule->prob = krule->prob;
1863 	rule->cuid = krule->cuid;
1864 	rule->cpid = krule->cpid;
1865 
1866 	rule->return_icmp = krule->return_icmp;
1867 	rule->return_icmp6 = krule->return_icmp6;
1868 	rule->max_mss = krule->max_mss;
1869 	rule->tag = krule->tag;
1870 	rule->match_tag = krule->match_tag;
1871 	rule->scrub_flags = krule->scrub_flags;
1872 
1873 	bcopy(&krule->uid, &rule->uid, sizeof(krule->uid));
1874 	bcopy(&krule->gid, &rule->gid, sizeof(krule->gid));
1875 
1876 	rule->rule_flag = krule->rule_flag;
1877 	rule->action = krule->action;
1878 	rule->direction = krule->direction;
1879 	rule->log = krule->log;
1880 	rule->logif = krule->logif;
1881 	rule->quick = krule->quick;
1882 	rule->ifnot = krule->ifnot;
1883 	rule->match_tag_not = krule->match_tag_not;
1884 	rule->natpass = krule->natpass;
1885 
1886 	rule->keep_state = krule->keep_state;
1887 	rule->af = krule->af;
1888 	rule->proto = krule->proto;
1889 	rule->type = krule->type;
1890 	rule->code = krule->code;
1891 	rule->flags = krule->flags;
1892 	rule->flagset = krule->flagset;
1893 	rule->min_ttl = krule->min_ttl;
1894 	rule->allow_opts = krule->allow_opts;
1895 	rule->rt = krule->rt;
1896 	rule->return_ttl = krule->return_ttl;
1897 	rule->tos = krule->tos;
1898 	rule->set_tos = krule->set_tos;
1899 	rule->anchor_relative = krule->anchor_relative;
1900 	rule->anchor_wildcard = krule->anchor_wildcard;
1901 
1902 	rule->flush = krule->flush;
1903 	rule->prio = krule->prio;
1904 	rule->set_prio[0] = krule->set_prio[0];
1905 	rule->set_prio[1] = krule->set_prio[1];
1906 
1907 	bcopy(&krule->divert, &rule->divert, sizeof(krule->divert));
1908 
1909 	rule->u_states_cur = counter_u64_fetch(krule->states_cur);
1910 	rule->u_states_tot = counter_u64_fetch(krule->states_tot);
1911 	rule->u_src_nodes = counter_u64_fetch(krule->src_nodes);
1912 }
1913 
1914 static int
1915 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
1916 {
1917 	int ret;
1918 
1919 #ifndef INET
1920 	if (rule->af == AF_INET) {
1921 		return (EAFNOSUPPORT);
1922 	}
1923 #endif /* INET */
1924 #ifndef INET6
1925 	if (rule->af == AF_INET6) {
1926 		return (EAFNOSUPPORT);
1927 	}
1928 #endif /* INET6 */
1929 
1930 	ret = pf_check_rule_addr(&rule->src);
1931 	if (ret != 0)
1932 		return (ret);
1933 	ret = pf_check_rule_addr(&rule->dst);
1934 	if (ret != 0)
1935 		return (ret);
1936 
1937 	bcopy(&rule->src, &krule->src, sizeof(rule->src));
1938 	bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
1939 
1940 	ret = pf_user_strcpy(krule->label[0], rule->label, sizeof(rule->label));
1941 	if (ret != 0)
1942 		return (ret);
1943 	ret = pf_user_strcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
1944 	if (ret != 0)
1945 		return (ret);
1946 	ret = pf_user_strcpy(krule->qname, rule->qname, sizeof(rule->qname));
1947 	if (ret != 0)
1948 		return (ret);
1949 	ret = pf_user_strcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
1950 	if (ret != 0)
1951 		return (ret);
1952 	ret = pf_user_strcpy(krule->tagname, rule->tagname,
1953 	    sizeof(rule->tagname));
1954 	if (ret != 0)
1955 		return (ret);
1956 	ret = pf_user_strcpy(krule->match_tagname, rule->match_tagname,
1957 	    sizeof(rule->match_tagname));
1958 	if (ret != 0)
1959 		return (ret);
1960 	ret = pf_user_strcpy(krule->overload_tblname, rule->overload_tblname,
1961 	    sizeof(rule->overload_tblname));
1962 	if (ret != 0)
1963 		return (ret);
1964 
1965 	pf_pool_to_kpool(&rule->rpool, &krule->rpool);
1966 
1967 	/* Don't allow userspace to set evaulations, packets or bytes. */
1968 	/* kif, anchor, overload_tbl are not copied over. */
1969 
1970 	krule->os_fingerprint = rule->os_fingerprint;
1971 
1972 	krule->rtableid = rule->rtableid;
1973 	bcopy(rule->timeout, krule->timeout, sizeof(krule->timeout));
1974 	krule->max_states = rule->max_states;
1975 	krule->max_src_nodes = rule->max_src_nodes;
1976 	krule->max_src_states = rule->max_src_states;
1977 	krule->max_src_conn = rule->max_src_conn;
1978 	krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
1979 	krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
1980 	krule->qid = rule->qid;
1981 	krule->pqid = rule->pqid;
1982 	krule->nr = rule->nr;
1983 	krule->prob = rule->prob;
1984 	krule->cuid = rule->cuid;
1985 	krule->cpid = rule->cpid;
1986 
1987 	krule->return_icmp = rule->return_icmp;
1988 	krule->return_icmp6 = rule->return_icmp6;
1989 	krule->max_mss = rule->max_mss;
1990 	krule->tag = rule->tag;
1991 	krule->match_tag = rule->match_tag;
1992 	krule->scrub_flags = rule->scrub_flags;
1993 
1994 	bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
1995 	bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
1996 
1997 	krule->rule_flag = rule->rule_flag;
1998 	krule->action = rule->action;
1999 	krule->direction = rule->direction;
2000 	krule->log = rule->log;
2001 	krule->logif = rule->logif;
2002 	krule->quick = rule->quick;
2003 	krule->ifnot = rule->ifnot;
2004 	krule->match_tag_not = rule->match_tag_not;
2005 	krule->natpass = rule->natpass;
2006 
2007 	krule->keep_state = rule->keep_state;
2008 	krule->af = rule->af;
2009 	krule->proto = rule->proto;
2010 	krule->type = rule->type;
2011 	krule->code = rule->code;
2012 	krule->flags = rule->flags;
2013 	krule->flagset = rule->flagset;
2014 	krule->min_ttl = rule->min_ttl;
2015 	krule->allow_opts = rule->allow_opts;
2016 	krule->rt = rule->rt;
2017 	krule->return_ttl = rule->return_ttl;
2018 	krule->tos = rule->tos;
2019 	krule->set_tos = rule->set_tos;
2020 
2021 	krule->flush = rule->flush;
2022 	krule->prio = rule->prio;
2023 	krule->set_prio[0] = rule->set_prio[0];
2024 	krule->set_prio[1] = rule->set_prio[1];
2025 
2026 	bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
2027 
2028 	return (0);
2029 }
2030 
2031 static int
2032 pf_state_kill_to_kstate_kill(const struct pfioc_state_kill *psk,
2033     struct pf_kstate_kill *kill)
2034 {
2035 	int ret;
2036 
2037 	bzero(kill, sizeof(*kill));
2038 
2039 	bcopy(&psk->psk_pfcmp, &kill->psk_pfcmp, sizeof(kill->psk_pfcmp));
2040 	kill->psk_af = psk->psk_af;
2041 	kill->psk_proto = psk->psk_proto;
2042 	bcopy(&psk->psk_src, &kill->psk_src, sizeof(kill->psk_src));
2043 	bcopy(&psk->psk_dst, &kill->psk_dst, sizeof(kill->psk_dst));
2044 	ret = pf_user_strcpy(kill->psk_ifname, psk->psk_ifname,
2045 	    sizeof(kill->psk_ifname));
2046 	if (ret != 0)
2047 		return (ret);
2048 	ret = pf_user_strcpy(kill->psk_label, psk->psk_label,
2049 	    sizeof(kill->psk_label));
2050 	if (ret != 0)
2051 		return (ret);
2052 
2053 	return (0);
2054 }
2055 
2056 static int
2057 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket,
2058     uint32_t pool_ticket, const char *anchor, const char *anchor_call,
2059     struct thread *td)
2060 {
2061 	struct pf_kruleset	*ruleset;
2062 	struct pf_krule		*tail;
2063 	struct pf_kpooladdr	*pa;
2064 	struct pfi_kkif		*kif = NULL;
2065 	int			 rs_num;
2066 	int			 error = 0;
2067 
2068 	if ((rule->return_icmp >> 8) > ICMP_MAXTYPE) {
2069 		error = EINVAL;
2070 		goto errout_unlocked;
2071 	}
2072 
2073 #define	ERROUT(x)	ERROUT_FUNCTION(errout, x)
2074 
2075 	if (rule->ifname[0])
2076 		kif = pf_kkif_create(M_WAITOK);
2077 	pf_counter_u64_init(&rule->evaluations, M_WAITOK);
2078 	for (int i = 0; i < 2; i++) {
2079 		pf_counter_u64_init(&rule->packets[i], M_WAITOK);
2080 		pf_counter_u64_init(&rule->bytes[i], M_WAITOK);
2081 	}
2082 	rule->states_cur = counter_u64_alloc(M_WAITOK);
2083 	rule->states_tot = counter_u64_alloc(M_WAITOK);
2084 	rule->src_nodes = counter_u64_alloc(M_WAITOK);
2085 	rule->cuid = td->td_ucred->cr_ruid;
2086 	rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
2087 	TAILQ_INIT(&rule->rpool.list);
2088 
2089 	PF_RULES_WLOCK();
2090 #ifdef PF_WANT_32_TO_64_COUNTER
2091 	LIST_INSERT_HEAD(&V_pf_allrulelist, rule, allrulelist);
2092 	MPASS(!rule->allrulelinked);
2093 	rule->allrulelinked = true;
2094 	V_pf_allrulecount++;
2095 #endif
2096 	ruleset = pf_find_kruleset(anchor);
2097 	if (ruleset == NULL)
2098 		ERROUT(EINVAL);
2099 	rs_num = pf_get_ruleset_number(rule->action);
2100 	if (rs_num >= PF_RULESET_MAX)
2101 		ERROUT(EINVAL);
2102 	if (ticket != ruleset->rules[rs_num].inactive.ticket) {
2103 		DPFPRINTF(PF_DEBUG_MISC,
2104 		    ("ticket: %d != [%d]%d\n", ticket, rs_num,
2105 		    ruleset->rules[rs_num].inactive.ticket));
2106 		ERROUT(EBUSY);
2107 	}
2108 	if (pool_ticket != V_ticket_pabuf) {
2109 		DPFPRINTF(PF_DEBUG_MISC,
2110 		    ("pool_ticket: %d != %d\n", pool_ticket,
2111 		    V_ticket_pabuf));
2112 		ERROUT(EBUSY);
2113 	}
2114 
2115 	tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
2116 	    pf_krulequeue);
2117 	if (tail)
2118 		rule->nr = tail->nr + 1;
2119 	else
2120 		rule->nr = 0;
2121 	if (rule->ifname[0]) {
2122 		rule->kif = pfi_kkif_attach(kif, rule->ifname);
2123 		kif = NULL;
2124 		pfi_kkif_ref(rule->kif);
2125 	} else
2126 		rule->kif = NULL;
2127 
2128 	if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
2129 		error = EBUSY;
2130 
2131 #ifdef ALTQ
2132 	/* set queue IDs */
2133 	if (rule->qname[0] != 0) {
2134 		if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2135 			error = EBUSY;
2136 		else if (rule->pqname[0] != 0) {
2137 			if ((rule->pqid =
2138 			    pf_qname2qid(rule->pqname)) == 0)
2139 				error = EBUSY;
2140 		} else
2141 			rule->pqid = rule->qid;
2142 	}
2143 #endif
2144 	if (rule->tagname[0])
2145 		if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2146 			error = EBUSY;
2147 	if (rule->match_tagname[0])
2148 		if ((rule->match_tag =
2149 		    pf_tagname2tag(rule->match_tagname)) == 0)
2150 			error = EBUSY;
2151 	if (rule->rt && !rule->direction)
2152 		error = EINVAL;
2153 	if (!rule->log)
2154 		rule->logif = 0;
2155 	if (rule->logif >= PFLOGIFS_MAX)
2156 		error = EINVAL;
2157 	if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
2158 		error = ENOMEM;
2159 	if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
2160 		error = ENOMEM;
2161 	if (pf_kanchor_setup(rule, ruleset, anchor_call))
2162 		error = EINVAL;
2163 	if (rule->scrub_flags & PFSTATE_SETPRIO &&
2164 	    (rule->set_prio[0] > PF_PRIO_MAX ||
2165 	    rule->set_prio[1] > PF_PRIO_MAX))
2166 		error = EINVAL;
2167 	TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
2168 		if (pa->addr.type == PF_ADDR_TABLE) {
2169 			pa->addr.p.tbl = pfr_attach_table(ruleset,
2170 			    pa->addr.v.tblname);
2171 			if (pa->addr.p.tbl == NULL)
2172 				error = ENOMEM;
2173 		}
2174 
2175 	rule->overload_tbl = NULL;
2176 	if (rule->overload_tblname[0]) {
2177 		if ((rule->overload_tbl = pfr_attach_table(ruleset,
2178 		    rule->overload_tblname)) == NULL)
2179 			error = EINVAL;
2180 		else
2181 			rule->overload_tbl->pfrkt_flags |=
2182 			    PFR_TFLAG_ACTIVE;
2183 	}
2184 
2185 	pf_mv_kpool(&V_pf_pabuf, &rule->rpool.list);
2186 	if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
2187 	    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
2188 	    (rule->rt > PF_NOPFROUTE)) &&
2189 	    (TAILQ_FIRST(&rule->rpool.list) == NULL))
2190 		error = EINVAL;
2191 
2192 	if (error) {
2193 		pf_free_rule(rule);
2194 		rule = NULL;
2195 		ERROUT(error);
2196 	}
2197 
2198 	rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
2199 	TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
2200 	    rule, entries);
2201 	ruleset->rules[rs_num].inactive.rcount++;
2202 
2203 	PF_RULES_WUNLOCK();
2204 
2205 	return (0);
2206 
2207 #undef ERROUT
2208 errout:
2209 	PF_RULES_WUNLOCK();
2210 errout_unlocked:
2211 	pf_kkif_free(kif);
2212 	pf_krule_free(rule);
2213 	return (error);
2214 }
2215 
2216 static bool
2217 pf_label_match(const struct pf_krule *rule, const char *label)
2218 {
2219 	int i = 0;
2220 
2221 	while (*rule->label[i]) {
2222 		if (strcmp(rule->label[i], label) == 0)
2223 			return (true);
2224 		i++;
2225 	}
2226 
2227 	return (false);
2228 }
2229 
2230 static unsigned int
2231 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir)
2232 {
2233 	struct pf_kstate *s;
2234 	int more = 0;
2235 
2236 	s = pf_find_state_all(key, dir, &more);
2237 	if (s == NULL)
2238 		return (0);
2239 
2240 	if (more) {
2241 		PF_STATE_UNLOCK(s);
2242 		return (0);
2243 	}
2244 
2245 	pf_unlink_state(s);
2246 	return (1);
2247 }
2248 
2249 static int
2250 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih)
2251 {
2252 	struct pf_kstate	*s;
2253 	struct pf_state_key	*sk;
2254 	struct pf_addr		*srcaddr, *dstaddr;
2255 	struct pf_state_key_cmp	 match_key;
2256 	int			 idx, killed = 0;
2257 	unsigned int		 dir;
2258 	u_int16_t		 srcport, dstport;
2259 	struct pfi_kkif		*kif;
2260 
2261 relock_DIOCKILLSTATES:
2262 	PF_HASHROW_LOCK(ih);
2263 	LIST_FOREACH(s, &ih->states, entry) {
2264 		/* For floating states look at the original kif. */
2265 		kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
2266 
2267 		sk = s->key[PF_SK_WIRE];
2268 		if (s->direction == PF_OUT) {
2269 			srcaddr = &sk->addr[1];
2270 			dstaddr = &sk->addr[0];
2271 			srcport = sk->port[1];
2272 			dstport = sk->port[0];
2273 		} else {
2274 			srcaddr = &sk->addr[0];
2275 			dstaddr = &sk->addr[1];
2276 			srcport = sk->port[0];
2277 			dstport = sk->port[1];
2278 		}
2279 
2280 		if (psk->psk_af && sk->af != psk->psk_af)
2281 			continue;
2282 
2283 		if (psk->psk_proto && psk->psk_proto != sk->proto)
2284 			continue;
2285 
2286 		if (! PF_MATCHA(psk->psk_src.neg, &psk->psk_src.addr.v.a.addr,
2287 		    &psk->psk_src.addr.v.a.mask, srcaddr, sk->af))
2288 			continue;
2289 
2290 		if (! PF_MATCHA(psk->psk_dst.neg, &psk->psk_dst.addr.v.a.addr,
2291 		    &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af))
2292 			continue;
2293 
2294 		if (!  PF_MATCHA(psk->psk_rt_addr.neg,
2295 		    &psk->psk_rt_addr.addr.v.a.addr,
2296 		    &psk->psk_rt_addr.addr.v.a.mask,
2297 		    &s->rt_addr, sk->af))
2298 			continue;
2299 
2300 		if (psk->psk_src.port_op != 0 &&
2301 		    ! pf_match_port(psk->psk_src.port_op,
2302 		    psk->psk_src.port[0], psk->psk_src.port[1], srcport))
2303 			continue;
2304 
2305 		if (psk->psk_dst.port_op != 0 &&
2306 		    ! pf_match_port(psk->psk_dst.port_op,
2307 		    psk->psk_dst.port[0], psk->psk_dst.port[1], dstport))
2308 			continue;
2309 
2310 		if (psk->psk_label[0] &&
2311 		    ! pf_label_match(s->rule.ptr, psk->psk_label))
2312 			continue;
2313 
2314 		if (psk->psk_ifname[0] && strcmp(psk->psk_ifname,
2315 		    kif->pfik_name))
2316 			continue;
2317 
2318 		if (psk->psk_kill_match) {
2319 			/* Create the key to find matching states, with lock
2320 			 * held. */
2321 
2322 			bzero(&match_key, sizeof(match_key));
2323 
2324 			if (s->direction == PF_OUT) {
2325 				dir = PF_IN;
2326 				idx = PF_SK_STACK;
2327 			} else {
2328 				dir = PF_OUT;
2329 				idx = PF_SK_WIRE;
2330 			}
2331 
2332 			match_key.af = s->key[idx]->af;
2333 			match_key.proto = s->key[idx]->proto;
2334 			PF_ACPY(&match_key.addr[0],
2335 			    &s->key[idx]->addr[1], match_key.af);
2336 			match_key.port[0] = s->key[idx]->port[1];
2337 			PF_ACPY(&match_key.addr[1],
2338 			    &s->key[idx]->addr[0], match_key.af);
2339 			match_key.port[1] = s->key[idx]->port[0];
2340 		}
2341 
2342 		pf_unlink_state(s);
2343 		killed++;
2344 
2345 		if (psk->psk_kill_match)
2346 			killed += pf_kill_matching_state(&match_key, dir);
2347 
2348 		goto relock_DIOCKILLSTATES;
2349 	}
2350 	PF_HASHROW_UNLOCK(ih);
2351 
2352 	return (killed);
2353 }
2354 
2355 static int
2356 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
2357 {
2358 	int			 error = 0;
2359 	PF_RULES_RLOCK_TRACKER;
2360 
2361 #define	ERROUT_IOCTL(target, x)					\
2362     do {								\
2363 	    error = (x);						\
2364 	    SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__);	\
2365 	    goto target;						\
2366     } while (0)
2367 
2368 
2369 	/* XXX keep in sync with switch() below */
2370 	if (securelevel_gt(td->td_ucred, 2))
2371 		switch (cmd) {
2372 		case DIOCGETRULES:
2373 		case DIOCGETRULE:
2374 		case DIOCGETRULENV:
2375 		case DIOCGETADDRS:
2376 		case DIOCGETADDR:
2377 		case DIOCGETSTATE:
2378 		case DIOCGETSTATENV:
2379 		case DIOCSETSTATUSIF:
2380 		case DIOCGETSTATUS:
2381 		case DIOCGETSTATUSNV:
2382 		case DIOCCLRSTATUS:
2383 		case DIOCNATLOOK:
2384 		case DIOCSETDEBUG:
2385 		case DIOCGETSTATES:
2386 		case DIOCGETSTATESV2:
2387 		case DIOCGETTIMEOUT:
2388 		case DIOCCLRRULECTRS:
2389 		case DIOCGETLIMIT:
2390 		case DIOCGETALTQSV0:
2391 		case DIOCGETALTQSV1:
2392 		case DIOCGETALTQV0:
2393 		case DIOCGETALTQV1:
2394 		case DIOCGETQSTATSV0:
2395 		case DIOCGETQSTATSV1:
2396 		case DIOCGETRULESETS:
2397 		case DIOCGETRULESET:
2398 		case DIOCRGETTABLES:
2399 		case DIOCRGETTSTATS:
2400 		case DIOCRCLRTSTATS:
2401 		case DIOCRCLRADDRS:
2402 		case DIOCRADDADDRS:
2403 		case DIOCRDELADDRS:
2404 		case DIOCRSETADDRS:
2405 		case DIOCRGETADDRS:
2406 		case DIOCRGETASTATS:
2407 		case DIOCRCLRASTATS:
2408 		case DIOCRTSTADDRS:
2409 		case DIOCOSFPGET:
2410 		case DIOCGETSRCNODES:
2411 		case DIOCCLRSRCNODES:
2412 		case DIOCGETSYNCOOKIES:
2413 		case DIOCIGETIFACES:
2414 		case DIOCGIFSPEEDV0:
2415 		case DIOCGIFSPEEDV1:
2416 		case DIOCSETIFFLAG:
2417 		case DIOCCLRIFFLAG:
2418 		case DIOCGETETHRULES:
2419 		case DIOCGETETHRULE:
2420 			break;
2421 		case DIOCRCLRTABLES:
2422 		case DIOCRADDTABLES:
2423 		case DIOCRDELTABLES:
2424 		case DIOCRSETTFLAGS:
2425 			if (((struct pfioc_table *)addr)->pfrio_flags &
2426 			    PFR_FLAG_DUMMY)
2427 				break; /* dummy operation ok */
2428 			return (EPERM);
2429 		default:
2430 			return (EPERM);
2431 		}
2432 
2433 	if (!(flags & FWRITE))
2434 		switch (cmd) {
2435 		case DIOCGETRULES:
2436 		case DIOCGETADDRS:
2437 		case DIOCGETADDR:
2438 		case DIOCGETSTATE:
2439 		case DIOCGETSTATENV:
2440 		case DIOCGETSTATUS:
2441 		case DIOCGETSTATUSNV:
2442 		case DIOCGETSTATES:
2443 		case DIOCGETSTATESV2:
2444 		case DIOCGETTIMEOUT:
2445 		case DIOCGETLIMIT:
2446 		case DIOCGETALTQSV0:
2447 		case DIOCGETALTQSV1:
2448 		case DIOCGETALTQV0:
2449 		case DIOCGETALTQV1:
2450 		case DIOCGETQSTATSV0:
2451 		case DIOCGETQSTATSV1:
2452 		case DIOCGETRULESETS:
2453 		case DIOCGETRULESET:
2454 		case DIOCNATLOOK:
2455 		case DIOCRGETTABLES:
2456 		case DIOCRGETTSTATS:
2457 		case DIOCRGETADDRS:
2458 		case DIOCRGETASTATS:
2459 		case DIOCRTSTADDRS:
2460 		case DIOCOSFPGET:
2461 		case DIOCGETSRCNODES:
2462 		case DIOCGETSYNCOOKIES:
2463 		case DIOCIGETIFACES:
2464 		case DIOCGIFSPEEDV1:
2465 		case DIOCGIFSPEEDV0:
2466 		case DIOCGETRULENV:
2467 		case DIOCGETETHRULES:
2468 		case DIOCGETETHRULE:
2469 			break;
2470 		case DIOCRCLRTABLES:
2471 		case DIOCRADDTABLES:
2472 		case DIOCRDELTABLES:
2473 		case DIOCRCLRTSTATS:
2474 		case DIOCRCLRADDRS:
2475 		case DIOCRADDADDRS:
2476 		case DIOCRDELADDRS:
2477 		case DIOCRSETADDRS:
2478 		case DIOCRSETTFLAGS:
2479 			if (((struct pfioc_table *)addr)->pfrio_flags &
2480 			    PFR_FLAG_DUMMY) {
2481 				flags |= FWRITE; /* need write lock for dummy */
2482 				break; /* dummy operation ok */
2483 			}
2484 			return (EACCES);
2485 		case DIOCGETRULE:
2486 			if (((struct pfioc_rule *)addr)->action ==
2487 			    PF_GET_CLR_CNTR)
2488 				return (EACCES);
2489 			break;
2490 		default:
2491 			return (EACCES);
2492 		}
2493 
2494 	CURVNET_SET(TD_TO_VNET(td));
2495 
2496 	switch (cmd) {
2497 	case DIOCSTART:
2498 		sx_xlock(&pf_ioctl_lock);
2499 		if (V_pf_status.running)
2500 			error = EEXIST;
2501 		else {
2502 			int cpu;
2503 
2504 			hook_pf();
2505 			if (! TAILQ_EMPTY(V_pf_keth->active.rules))
2506 				hook_pf_eth();
2507 			V_pf_status.running = 1;
2508 			V_pf_status.since = time_second;
2509 
2510 			CPU_FOREACH(cpu)
2511 				V_pf_stateid[cpu] = time_second;
2512 
2513 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
2514 		}
2515 		break;
2516 
2517 	case DIOCSTOP:
2518 		sx_xlock(&pf_ioctl_lock);
2519 		if (!V_pf_status.running)
2520 			error = ENOENT;
2521 		else {
2522 			V_pf_status.running = 0;
2523 			dehook_pf();
2524 			dehook_pf_eth();
2525 			V_pf_status.since = time_second;
2526 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
2527 		}
2528 		break;
2529 
2530 	case DIOCGETETHRULES: {
2531 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2532 		nvlist_t		*nvl;
2533 		void			*packed;
2534 		struct pf_keth_rule	*tail;
2535 		struct pf_keth_ruleset	*rs;
2536 		u_int32_t		 ticket, nr;
2537 		const char		*anchor = "";
2538 
2539 		nvl = NULL;
2540 		packed = NULL;
2541 
2542 #define	ERROUT(x)	do { error = (x); goto DIOCGETETHRULES_error; } while (0)
2543 
2544 		if (nv->len > pf_ioctl_maxcount)
2545 			ERROUT(ENOMEM);
2546 
2547 		/* Copy the request in */
2548 		packed = malloc(nv->len, M_NVLIST, M_WAITOK);
2549 		if (packed == NULL)
2550 			ERROUT(ENOMEM);
2551 
2552 		error = copyin(nv->data, packed, nv->len);
2553 		if (error)
2554 			ERROUT(error);
2555 
2556 		nvl = nvlist_unpack(packed, nv->len, 0);
2557 		if (nvl == NULL)
2558 			ERROUT(EBADMSG);
2559 
2560 		if (! nvlist_exists_string(nvl, "anchor"))
2561 			ERROUT(EBADMSG);
2562 
2563 		anchor = nvlist_get_string(nvl, "anchor");
2564 
2565 		rs = pf_find_keth_ruleset(anchor);
2566 
2567 		nvlist_destroy(nvl);
2568 		nvl = NULL;
2569 		free(packed, M_NVLIST);
2570 		packed = NULL;
2571 
2572 		if (rs == NULL)
2573 			ERROUT(ENOENT);
2574 
2575 		/* Reply */
2576 		nvl = nvlist_create(0);
2577 		if (nvl == NULL)
2578 			ERROUT(ENOMEM);
2579 
2580 		PF_RULES_RLOCK();
2581 
2582 		ticket = rs->active.ticket;
2583 		tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq);
2584 		if (tail)
2585 			nr = tail->nr + 1;
2586 		else
2587 			nr = 0;
2588 
2589 		PF_RULES_RUNLOCK();
2590 
2591 		nvlist_add_number(nvl, "ticket", ticket);
2592 		nvlist_add_number(nvl, "nr", nr);
2593 
2594 		packed = nvlist_pack(nvl, &nv->len);
2595 		if (packed == NULL)
2596 			ERROUT(ENOMEM);
2597 
2598 		if (nv->size == 0)
2599 			ERROUT(0);
2600 		else if (nv->size < nv->len)
2601 			ERROUT(ENOSPC);
2602 
2603 		error = copyout(packed, nv->data, nv->len);
2604 
2605 #undef ERROUT
2606 DIOCGETETHRULES_error:
2607 		free(packed, M_NVLIST);
2608 		nvlist_destroy(nvl);
2609 		break;
2610 	}
2611 
2612 	case DIOCGETETHRULE: {
2613 		struct epoch_tracker	 et;
2614 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2615 		nvlist_t		*nvl = NULL;
2616 		void			*nvlpacked = NULL;
2617 		struct pf_keth_rule	*rule = NULL;
2618 		struct pf_keth_ruleset	*rs;
2619 		u_int32_t		 ticket, nr;
2620 		bool			 clear = false;
2621 		const char		*anchor;
2622 
2623 #define ERROUT(x)	do { error = (x); goto DIOCGETETHRULE_error; } while (0)
2624 
2625 		nvlpacked = malloc(nv->len, M_TEMP, M_WAITOK);
2626 		if (nvlpacked == NULL)
2627 			ERROUT(ENOMEM);
2628 
2629 		error = copyin(nv->data, nvlpacked, nv->len);
2630 		if (error)
2631 			ERROUT(error);
2632 
2633 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2634 		if (! nvlist_exists_number(nvl, "ticket"))
2635 			ERROUT(EBADMSG);
2636 		ticket = nvlist_get_number(nvl, "ticket");
2637 		if (! nvlist_exists_string(nvl, "anchor"))
2638 			ERROUT(EBADMSG);
2639 		anchor = nvlist_get_string(nvl, "anchor");
2640 
2641 		if (nvlist_exists_bool(nvl, "clear"))
2642 			clear = nvlist_get_bool(nvl, "clear");
2643 
2644 		if (clear && !(flags & FWRITE))
2645 			ERROUT(EACCES);
2646 
2647 		if (! nvlist_exists_number(nvl, "nr"))
2648 			ERROUT(EBADMSG);
2649 		nr = nvlist_get_number(nvl, "nr");
2650 
2651 		PF_RULES_RLOCK();
2652 		rs = pf_find_keth_ruleset(anchor);
2653 		if (rs == NULL) {
2654 			PF_RULES_RUNLOCK();
2655 			ERROUT(ENOENT);
2656 		}
2657 		if (ticket != rs->active.ticket) {
2658 			PF_RULES_RUNLOCK();
2659 			ERROUT(EBUSY);
2660 		}
2661 
2662 		nvlist_destroy(nvl);
2663 		nvl = NULL;
2664 		free(nvlpacked, M_TEMP);
2665 		nvlpacked = NULL;
2666 
2667 		nvl = nvlist_create(0);
2668 
2669 		rule = TAILQ_FIRST(rs->active.rules);
2670 		while ((rule != NULL) && (rule->nr != nr))
2671 			rule = TAILQ_NEXT(rule, entries);
2672 		if (rule == NULL) {
2673 			PF_RULES_RUNLOCK();
2674 			ERROUT(ENOENT);
2675 		}
2676 		/* Make sure rule can't go away. */
2677 		NET_EPOCH_ENTER(et);
2678 		PF_RULES_RUNLOCK();
2679 		nvl = pf_keth_rule_to_nveth_rule(rule);
2680 		if (pf_keth_anchor_nvcopyout(rs, rule, nvl))
2681 			ERROUT(EBUSY);
2682 		NET_EPOCH_EXIT(et);
2683 		if (nvl == NULL)
2684 			ERROUT(ENOMEM);
2685 
2686 		nvlpacked = nvlist_pack(nvl, &nv->len);
2687 		if (nvlpacked == NULL)
2688 			ERROUT(ENOMEM);
2689 
2690 		if (nv->size == 0)
2691 			ERROUT(0);
2692 		else if (nv->size < nv->len)
2693 			ERROUT(ENOSPC);
2694 
2695 		error = copyout(nvlpacked, nv->data, nv->len);
2696 		if (error == 0 && clear) {
2697 			counter_u64_zero(rule->evaluations);
2698 			for (int i = 0; i < 2; i++) {
2699 				counter_u64_zero(rule->packets[i]);
2700 				counter_u64_zero(rule->bytes[i]);
2701 			}
2702 		}
2703 
2704 #undef ERROUT
2705 DIOCGETETHRULE_error:
2706 		free(nvlpacked, M_TEMP);
2707 		nvlist_destroy(nvl);
2708 		break;
2709 	}
2710 
2711 	case DIOCADDETHRULE: {
2712 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2713 		nvlist_t		*nvl = NULL;
2714 		void			*nvlpacked = NULL;
2715 		struct pf_keth_rule	*rule = NULL, *tail = NULL;
2716 		struct pf_keth_ruleset	*ruleset = NULL;
2717 		struct pfi_kkif		*kif = NULL;
2718 		const char		*anchor = "", *anchor_call = "";
2719 
2720 #define ERROUT(x)	do { error = (x); goto DIOCADDETHRULE_error; } while (0)
2721 
2722 		nvlpacked = malloc(nv->len, M_TEMP, M_WAITOK);
2723 		if (nvlpacked == NULL)
2724 			ERROUT(ENOMEM);
2725 
2726 		error = copyin(nv->data, nvlpacked, nv->len);
2727 		if (error)
2728 			ERROUT(error);
2729 
2730 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2731 		if (nvl == NULL)
2732 			ERROUT(EBADMSG);
2733 
2734 		if (! nvlist_exists_number(nvl, "ticket"))
2735 			ERROUT(EBADMSG);
2736 
2737 		if (nvlist_exists_string(nvl, "anchor"))
2738 			anchor = nvlist_get_string(nvl, "anchor");
2739 		if (nvlist_exists_string(nvl, "anchor_call"))
2740 			anchor_call = nvlist_get_string(nvl, "anchor_call");
2741 
2742 		ruleset = pf_find_keth_ruleset(anchor);
2743 		if (ruleset == NULL)
2744 			ERROUT(EINVAL);
2745 
2746 		if (nvlist_get_number(nvl, "ticket") !=
2747 		    ruleset->inactive.ticket) {
2748 			DPFPRINTF(PF_DEBUG_MISC,
2749 			    ("ticket: %d != %d\n",
2750 			    (u_int32_t)nvlist_get_number(nvl, "ticket"),
2751 			    ruleset->inactive.ticket));
2752 			ERROUT(EBUSY);
2753 		}
2754 
2755 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
2756 		if (rule == NULL)
2757 			ERROUT(ENOMEM);
2758 
2759 		error = pf_nveth_rule_to_keth_rule(nvl, rule);
2760 		if (error != 0)
2761 			ERROUT(error);
2762 
2763 		if (rule->ifname[0])
2764 			kif = pf_kkif_create(M_WAITOK);
2765 		rule->evaluations = counter_u64_alloc(M_WAITOK);
2766 		for (int i = 0; i < 2; i++) {
2767 			rule->packets[i] = counter_u64_alloc(M_WAITOK);
2768 			rule->bytes[i] = counter_u64_alloc(M_WAITOK);
2769 		}
2770 
2771 		PF_RULES_WLOCK();
2772 
2773 		if (rule->ifname[0]) {
2774 			rule->kif = pfi_kkif_attach(kif, rule->ifname);
2775 			pfi_kkif_ref(rule->kif);
2776 		} else
2777 			rule->kif = NULL;
2778 
2779 #ifdef ALTQ
2780 		/* set queue IDs */
2781 		if (rule->qname[0] != 0) {
2782 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2783 				error = EBUSY;
2784 			else
2785 				rule->qid = rule->qid;
2786 		}
2787 #endif
2788 		if (rule->tagname[0])
2789 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2790 				error = EBUSY;
2791 
2792 		if (error) {
2793 			pf_free_eth_rule(rule);
2794 			PF_RULES_WUNLOCK();
2795 			ERROUT(error);
2796 		}
2797 
2798 		if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) {
2799 			pf_free_eth_rule(rule);
2800 			PF_RULES_WUNLOCK();
2801 			ERROUT(EINVAL);
2802 		}
2803 
2804 		tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq);
2805 		if (tail)
2806 			rule->nr = tail->nr + 1;
2807 		else
2808 			rule->nr = 0;
2809 
2810 		TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries);
2811 
2812 		PF_RULES_WUNLOCK();
2813 
2814 #undef ERROUT
2815 DIOCADDETHRULE_error:
2816 		nvlist_destroy(nvl);
2817 		free(nvlpacked, M_TEMP);
2818 		break;
2819 	}
2820 
2821 	case DIOCADDRULENV: {
2822 		struct pfioc_nv	*nv = (struct pfioc_nv *)addr;
2823 		nvlist_t	*nvl = NULL;
2824 		void		*nvlpacked = NULL;
2825 		struct pf_krule	*rule = NULL;
2826 		const char	*anchor = "", *anchor_call = "";
2827 		uint32_t	 ticket = 0, pool_ticket = 0;
2828 
2829 #define	ERROUT(x)	ERROUT_IOCTL(DIOCADDRULENV_error, x)
2830 
2831 		if (nv->len > pf_ioctl_maxcount)
2832 			ERROUT(ENOMEM);
2833 
2834 		nvlpacked = malloc(nv->len, M_TEMP, M_WAITOK);
2835 		error = copyin(nv->data, nvlpacked, nv->len);
2836 		if (error)
2837 			ERROUT(error);
2838 
2839 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2840 		if (nvl == NULL)
2841 			ERROUT(EBADMSG);
2842 
2843 		if (! nvlist_exists_number(nvl, "ticket"))
2844 			ERROUT(EINVAL);
2845 		ticket = nvlist_get_number(nvl, "ticket");
2846 
2847 		if (! nvlist_exists_number(nvl, "pool_ticket"))
2848 			ERROUT(EINVAL);
2849 		pool_ticket = nvlist_get_number(nvl, "pool_ticket");
2850 
2851 		if (! nvlist_exists_nvlist(nvl, "rule"))
2852 			ERROUT(EINVAL);
2853 
2854 		rule = pf_krule_alloc();
2855 		error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"),
2856 		    rule);
2857 		if (error)
2858 			ERROUT(error);
2859 
2860 		if (nvlist_exists_string(nvl, "anchor"))
2861 			anchor = nvlist_get_string(nvl, "anchor");
2862 		if (nvlist_exists_string(nvl, "anchor_call"))
2863 			anchor_call = nvlist_get_string(nvl, "anchor_call");
2864 
2865 		if ((error = nvlist_error(nvl)))
2866 			ERROUT(error);
2867 
2868 		/* Frees rule on error */
2869 		error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor,
2870 		    anchor_call, td);
2871 
2872 		nvlist_destroy(nvl);
2873 		free(nvlpacked, M_TEMP);
2874 		break;
2875 #undef ERROUT
2876 DIOCADDRULENV_error:
2877 		pf_krule_free(rule);
2878 		nvlist_destroy(nvl);
2879 		free(nvlpacked, M_TEMP);
2880 
2881 		break;
2882 	}
2883 	case DIOCADDRULE: {
2884 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
2885 		struct pf_krule		*rule;
2886 
2887 		rule = pf_krule_alloc();
2888 		error = pf_rule_to_krule(&pr->rule, rule);
2889 		if (error != 0) {
2890 			pf_krule_free(rule);
2891 			break;
2892 		}
2893 
2894 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
2895 
2896 		/* Frees rule on error */
2897 		error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket,
2898 		    pr->anchor, pr->anchor_call, td);
2899 		break;
2900 	}
2901 
2902 	case DIOCGETRULES: {
2903 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
2904 		struct pf_kruleset	*ruleset;
2905 		struct pf_krule		*tail;
2906 		int			 rs_num;
2907 
2908 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
2909 
2910 		PF_RULES_WLOCK();
2911 		ruleset = pf_find_kruleset(pr->anchor);
2912 		if (ruleset == NULL) {
2913 			PF_RULES_WUNLOCK();
2914 			error = EINVAL;
2915 			break;
2916 		}
2917 		rs_num = pf_get_ruleset_number(pr->rule.action);
2918 		if (rs_num >= PF_RULESET_MAX) {
2919 			PF_RULES_WUNLOCK();
2920 			error = EINVAL;
2921 			break;
2922 		}
2923 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
2924 		    pf_krulequeue);
2925 		if (tail)
2926 			pr->nr = tail->nr + 1;
2927 		else
2928 			pr->nr = 0;
2929 		pr->ticket = ruleset->rules[rs_num].active.ticket;
2930 		PF_RULES_WUNLOCK();
2931 		break;
2932 	}
2933 
2934 	case DIOCGETRULE: {
2935 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
2936 		struct pf_kruleset	*ruleset;
2937 		struct pf_krule		*rule;
2938 		int			 rs_num;
2939 
2940 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
2941 
2942 		PF_RULES_WLOCK();
2943 		ruleset = pf_find_kruleset(pr->anchor);
2944 		if (ruleset == NULL) {
2945 			PF_RULES_WUNLOCK();
2946 			error = EINVAL;
2947 			break;
2948 		}
2949 		rs_num = pf_get_ruleset_number(pr->rule.action);
2950 		if (rs_num >= PF_RULESET_MAX) {
2951 			PF_RULES_WUNLOCK();
2952 			error = EINVAL;
2953 			break;
2954 		}
2955 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
2956 			PF_RULES_WUNLOCK();
2957 			error = EBUSY;
2958 			break;
2959 		}
2960 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
2961 		while ((rule != NULL) && (rule->nr != pr->nr))
2962 			rule = TAILQ_NEXT(rule, entries);
2963 		if (rule == NULL) {
2964 			PF_RULES_WUNLOCK();
2965 			error = EBUSY;
2966 			break;
2967 		}
2968 
2969 		pf_krule_to_rule(rule, &pr->rule);
2970 
2971 		if (pf_kanchor_copyout(ruleset, rule, pr)) {
2972 			PF_RULES_WUNLOCK();
2973 			error = EBUSY;
2974 			break;
2975 		}
2976 		pf_addr_copyout(&pr->rule.src.addr);
2977 		pf_addr_copyout(&pr->rule.dst.addr);
2978 
2979 		if (pr->action == PF_GET_CLR_CNTR) {
2980 			pf_counter_u64_zero(&rule->evaluations);
2981 			for (int i = 0; i < 2; i++) {
2982 				pf_counter_u64_zero(&rule->packets[i]);
2983 				pf_counter_u64_zero(&rule->bytes[i]);
2984 			}
2985 			counter_u64_zero(rule->states_tot);
2986 		}
2987 		PF_RULES_WUNLOCK();
2988 		break;
2989 	}
2990 
2991 	case DIOCGETRULENV: {
2992 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2993 		nvlist_t		*nvrule = NULL;
2994 		nvlist_t		*nvl = NULL;
2995 		struct pf_kruleset	*ruleset;
2996 		struct pf_krule		*rule;
2997 		void			*nvlpacked = NULL;
2998 		int			 rs_num, nr;
2999 		bool			 clear_counter = false;
3000 
3001 #define	ERROUT(x)	ERROUT_IOCTL(DIOCGETRULENV_error, x)
3002 
3003 		if (nv->len > pf_ioctl_maxcount)
3004 			ERROUT(ENOMEM);
3005 
3006 		/* Copy the request in */
3007 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3008 		if (nvlpacked == NULL)
3009 			ERROUT(ENOMEM);
3010 
3011 		error = copyin(nv->data, nvlpacked, nv->len);
3012 		if (error)
3013 			ERROUT(error);
3014 
3015 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3016 		if (nvl == NULL)
3017 			ERROUT(EBADMSG);
3018 
3019 		if (! nvlist_exists_string(nvl, "anchor"))
3020 			ERROUT(EBADMSG);
3021 		if (! nvlist_exists_number(nvl, "ruleset"))
3022 			ERROUT(EBADMSG);
3023 		if (! nvlist_exists_number(nvl, "ticket"))
3024 			ERROUT(EBADMSG);
3025 		if (! nvlist_exists_number(nvl, "nr"))
3026 			ERROUT(EBADMSG);
3027 
3028 		if (nvlist_exists_bool(nvl, "clear_counter"))
3029 			clear_counter = nvlist_get_bool(nvl, "clear_counter");
3030 
3031 		if (clear_counter && !(flags & FWRITE))
3032 			ERROUT(EACCES);
3033 
3034 		nr = nvlist_get_number(nvl, "nr");
3035 
3036 		PF_RULES_WLOCK();
3037 		ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor"));
3038 		if (ruleset == NULL) {
3039 			PF_RULES_WUNLOCK();
3040 			ERROUT(ENOENT);
3041 		}
3042 
3043 		rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset"));
3044 		if (rs_num >= PF_RULESET_MAX) {
3045 			PF_RULES_WUNLOCK();
3046 			ERROUT(EINVAL);
3047 		}
3048 
3049 		if (nvlist_get_number(nvl, "ticket") !=
3050 		    ruleset->rules[rs_num].active.ticket) {
3051 			PF_RULES_WUNLOCK();
3052 			ERROUT(EBUSY);
3053 		}
3054 
3055 		if ((error = nvlist_error(nvl))) {
3056 			PF_RULES_WUNLOCK();
3057 			ERROUT(error);
3058 		}
3059 
3060 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
3061 		while ((rule != NULL) && (rule->nr != nr))
3062 			rule = TAILQ_NEXT(rule, entries);
3063 		if (rule == NULL) {
3064 			PF_RULES_WUNLOCK();
3065 			ERROUT(EBUSY);
3066 		}
3067 
3068 		nvrule = pf_krule_to_nvrule(rule);
3069 
3070 		nvlist_destroy(nvl);
3071 		nvl = nvlist_create(0);
3072 		if (nvl == NULL) {
3073 			PF_RULES_WUNLOCK();
3074 			ERROUT(ENOMEM);
3075 		}
3076 		nvlist_add_number(nvl, "nr", nr);
3077 		nvlist_add_nvlist(nvl, "rule", nvrule);
3078 		nvlist_destroy(nvrule);
3079 		nvrule = NULL;
3080 		if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) {
3081 			PF_RULES_WUNLOCK();
3082 			ERROUT(EBUSY);
3083 		}
3084 
3085 		free(nvlpacked, M_NVLIST);
3086 		nvlpacked = nvlist_pack(nvl, &nv->len);
3087 		if (nvlpacked == NULL) {
3088 			PF_RULES_WUNLOCK();
3089 			ERROUT(ENOMEM);
3090 		}
3091 
3092 		if (nv->size == 0) {
3093 			PF_RULES_WUNLOCK();
3094 			ERROUT(0);
3095 		}
3096 		else if (nv->size < nv->len) {
3097 			PF_RULES_WUNLOCK();
3098 			ERROUT(ENOSPC);
3099 		}
3100 
3101 		if (clear_counter) {
3102 			pf_counter_u64_zero(&rule->evaluations);
3103 			for (int i = 0; i < 2; i++) {
3104 				pf_counter_u64_zero(&rule->packets[i]);
3105 				pf_counter_u64_zero(&rule->bytes[i]);
3106 			}
3107 			counter_u64_zero(rule->states_tot);
3108 		}
3109 		PF_RULES_WUNLOCK();
3110 
3111 		error = copyout(nvlpacked, nv->data, nv->len);
3112 
3113 #undef ERROUT
3114 DIOCGETRULENV_error:
3115 		free(nvlpacked, M_NVLIST);
3116 		nvlist_destroy(nvrule);
3117 		nvlist_destroy(nvl);
3118 
3119 		break;
3120 	}
3121 
3122 	case DIOCCHANGERULE: {
3123 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
3124 		struct pf_kruleset	*ruleset;
3125 		struct pf_krule		*oldrule = NULL, *newrule = NULL;
3126 		struct pfi_kkif		*kif = NULL;
3127 		struct pf_kpooladdr	*pa;
3128 		u_int32_t		 nr = 0;
3129 		int			 rs_num;
3130 
3131 		pcr->anchor[sizeof(pcr->anchor) - 1] = 0;
3132 
3133 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
3134 		    pcr->action > PF_CHANGE_GET_TICKET) {
3135 			error = EINVAL;
3136 			break;
3137 		}
3138 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
3139 			error = EINVAL;
3140 			break;
3141 		}
3142 
3143 		if (pcr->action != PF_CHANGE_REMOVE) {
3144 			newrule = pf_krule_alloc();
3145 			error = pf_rule_to_krule(&pcr->rule, newrule);
3146 			if (error != 0) {
3147 				free(newrule, M_PFRULE);
3148 				break;
3149 			}
3150 
3151 			if (newrule->ifname[0])
3152 				kif = pf_kkif_create(M_WAITOK);
3153 			pf_counter_u64_init(&newrule->evaluations, M_WAITOK);
3154 			for (int i = 0; i < 2; i++) {
3155 				pf_counter_u64_init(&newrule->packets[i], M_WAITOK);
3156 				pf_counter_u64_init(&newrule->bytes[i], M_WAITOK);
3157 			}
3158 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
3159 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
3160 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
3161 			newrule->cuid = td->td_ucred->cr_ruid;
3162 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
3163 			TAILQ_INIT(&newrule->rpool.list);
3164 		}
3165 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGERULE_error; }
3166 
3167 		PF_RULES_WLOCK();
3168 #ifdef PF_WANT_32_TO_64_COUNTER
3169 		if (newrule != NULL) {
3170 			LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist);
3171 			newrule->allrulelinked = true;
3172 			V_pf_allrulecount++;
3173 		}
3174 #endif
3175 
3176 		if (!(pcr->action == PF_CHANGE_REMOVE ||
3177 		    pcr->action == PF_CHANGE_GET_TICKET) &&
3178 		    pcr->pool_ticket != V_ticket_pabuf)
3179 			ERROUT(EBUSY);
3180 
3181 		ruleset = pf_find_kruleset(pcr->anchor);
3182 		if (ruleset == NULL)
3183 			ERROUT(EINVAL);
3184 
3185 		rs_num = pf_get_ruleset_number(pcr->rule.action);
3186 		if (rs_num >= PF_RULESET_MAX)
3187 			ERROUT(EINVAL);
3188 
3189 		if (pcr->action == PF_CHANGE_GET_TICKET) {
3190 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
3191 			ERROUT(0);
3192 		} else if (pcr->ticket !=
3193 			    ruleset->rules[rs_num].active.ticket)
3194 				ERROUT(EINVAL);
3195 
3196 		if (pcr->action != PF_CHANGE_REMOVE) {
3197 			if (newrule->ifname[0]) {
3198 				newrule->kif = pfi_kkif_attach(kif,
3199 				    newrule->ifname);
3200 				kif = NULL;
3201 				pfi_kkif_ref(newrule->kif);
3202 			} else
3203 				newrule->kif = NULL;
3204 
3205 			if (newrule->rtableid > 0 &&
3206 			    newrule->rtableid >= rt_numfibs)
3207 				error = EBUSY;
3208 
3209 #ifdef ALTQ
3210 			/* set queue IDs */
3211 			if (newrule->qname[0] != 0) {
3212 				if ((newrule->qid =
3213 				    pf_qname2qid(newrule->qname)) == 0)
3214 					error = EBUSY;
3215 				else if (newrule->pqname[0] != 0) {
3216 					if ((newrule->pqid =
3217 					    pf_qname2qid(newrule->pqname)) == 0)
3218 						error = EBUSY;
3219 				} else
3220 					newrule->pqid = newrule->qid;
3221 			}
3222 #endif /* ALTQ */
3223 			if (newrule->tagname[0])
3224 				if ((newrule->tag =
3225 				    pf_tagname2tag(newrule->tagname)) == 0)
3226 					error = EBUSY;
3227 			if (newrule->match_tagname[0])
3228 				if ((newrule->match_tag = pf_tagname2tag(
3229 				    newrule->match_tagname)) == 0)
3230 					error = EBUSY;
3231 			if (newrule->rt && !newrule->direction)
3232 				error = EINVAL;
3233 			if (!newrule->log)
3234 				newrule->logif = 0;
3235 			if (newrule->logif >= PFLOGIFS_MAX)
3236 				error = EINVAL;
3237 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
3238 				error = ENOMEM;
3239 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
3240 				error = ENOMEM;
3241 			if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
3242 				error = EINVAL;
3243 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
3244 				if (pa->addr.type == PF_ADDR_TABLE) {
3245 					pa->addr.p.tbl =
3246 					    pfr_attach_table(ruleset,
3247 					    pa->addr.v.tblname);
3248 					if (pa->addr.p.tbl == NULL)
3249 						error = ENOMEM;
3250 				}
3251 
3252 			newrule->overload_tbl = NULL;
3253 			if (newrule->overload_tblname[0]) {
3254 				if ((newrule->overload_tbl = pfr_attach_table(
3255 				    ruleset, newrule->overload_tblname)) ==
3256 				    NULL)
3257 					error = EINVAL;
3258 				else
3259 					newrule->overload_tbl->pfrkt_flags |=
3260 					    PFR_TFLAG_ACTIVE;
3261 			}
3262 
3263 			pf_mv_kpool(&V_pf_pabuf, &newrule->rpool.list);
3264 			if (((((newrule->action == PF_NAT) ||
3265 			    (newrule->action == PF_RDR) ||
3266 			    (newrule->action == PF_BINAT) ||
3267 			    (newrule->rt > PF_NOPFROUTE)) &&
3268 			    !newrule->anchor)) &&
3269 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
3270 				error = EINVAL;
3271 
3272 			if (error) {
3273 				pf_free_rule(newrule);
3274 				PF_RULES_WUNLOCK();
3275 				break;
3276 			}
3277 
3278 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
3279 		}
3280 		pf_empty_kpool(&V_pf_pabuf);
3281 
3282 		if (pcr->action == PF_CHANGE_ADD_HEAD)
3283 			oldrule = TAILQ_FIRST(
3284 			    ruleset->rules[rs_num].active.ptr);
3285 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
3286 			oldrule = TAILQ_LAST(
3287 			    ruleset->rules[rs_num].active.ptr, pf_krulequeue);
3288 		else {
3289 			oldrule = TAILQ_FIRST(
3290 			    ruleset->rules[rs_num].active.ptr);
3291 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
3292 				oldrule = TAILQ_NEXT(oldrule, entries);
3293 			if (oldrule == NULL) {
3294 				if (newrule != NULL)
3295 					pf_free_rule(newrule);
3296 				PF_RULES_WUNLOCK();
3297 				error = EINVAL;
3298 				break;
3299 			}
3300 		}
3301 
3302 		if (pcr->action == PF_CHANGE_REMOVE) {
3303 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
3304 			    oldrule);
3305 			ruleset->rules[rs_num].active.rcount--;
3306 		} else {
3307 			if (oldrule == NULL)
3308 				TAILQ_INSERT_TAIL(
3309 				    ruleset->rules[rs_num].active.ptr,
3310 				    newrule, entries);
3311 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
3312 			    pcr->action == PF_CHANGE_ADD_BEFORE)
3313 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
3314 			else
3315 				TAILQ_INSERT_AFTER(
3316 				    ruleset->rules[rs_num].active.ptr,
3317 				    oldrule, newrule, entries);
3318 			ruleset->rules[rs_num].active.rcount++;
3319 		}
3320 
3321 		nr = 0;
3322 		TAILQ_FOREACH(oldrule,
3323 		    ruleset->rules[rs_num].active.ptr, entries)
3324 			oldrule->nr = nr++;
3325 
3326 		ruleset->rules[rs_num].active.ticket++;
3327 
3328 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
3329 		pf_remove_if_empty_kruleset(ruleset);
3330 
3331 		PF_RULES_WUNLOCK();
3332 		break;
3333 
3334 #undef ERROUT
3335 DIOCCHANGERULE_error:
3336 		PF_RULES_WUNLOCK();
3337 		pf_krule_free(newrule);
3338 		pf_kkif_free(kif);
3339 		break;
3340 	}
3341 
3342 	case DIOCCLRSTATES: {
3343 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
3344 		struct pf_kstate_kill	 kill;
3345 
3346 		error = pf_state_kill_to_kstate_kill(psk, &kill);
3347 		if (error)
3348 			break;
3349 
3350 		psk->psk_killed = pf_clear_states(&kill);
3351 		break;
3352 	}
3353 
3354 	case DIOCCLRSTATESNV: {
3355 		error = pf_clearstates_nv((struct pfioc_nv *)addr);
3356 		break;
3357 	}
3358 
3359 	case DIOCKILLSTATES: {
3360 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
3361 		struct pf_kstate_kill	 kill;
3362 
3363 		error = pf_state_kill_to_kstate_kill(psk, &kill);
3364 		if (error)
3365 			break;
3366 
3367 		psk->psk_killed = 0;
3368 		pf_killstates(&kill, &psk->psk_killed);
3369 		break;
3370 	}
3371 
3372 	case DIOCKILLSTATESNV: {
3373 		error = pf_killstates_nv((struct pfioc_nv *)addr);
3374 		break;
3375 	}
3376 
3377 	case DIOCADDSTATE: {
3378 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
3379 		struct pfsync_state	*sp = &ps->state;
3380 
3381 		if (sp->timeout >= PFTM_MAX) {
3382 			error = EINVAL;
3383 			break;
3384 		}
3385 		if (V_pfsync_state_import_ptr != NULL) {
3386 			PF_RULES_RLOCK();
3387 			error = V_pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
3388 			PF_RULES_RUNLOCK();
3389 		} else
3390 			error = EOPNOTSUPP;
3391 		break;
3392 	}
3393 
3394 	case DIOCGETSTATE: {
3395 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
3396 		struct pf_kstate	*s;
3397 
3398 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
3399 		if (s == NULL) {
3400 			error = ENOENT;
3401 			break;
3402 		}
3403 
3404 		pfsync_state_export(&ps->state, s);
3405 		PF_STATE_UNLOCK(s);
3406 		break;
3407 	}
3408 
3409 	case DIOCGETSTATENV: {
3410 		error = pf_getstate((struct pfioc_nv *)addr);
3411 		break;
3412 	}
3413 
3414 	case DIOCGETSTATES: {
3415 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
3416 		struct pf_kstate	*s;
3417 		struct pfsync_state	*pstore, *p;
3418 		int			 i, nr;
3419 		size_t			 slice_count = 16, count;
3420 		void			*out;
3421 
3422 		if (ps->ps_len <= 0) {
3423 			nr = uma_zone_get_cur(V_pf_state_z);
3424 			ps->ps_len = sizeof(struct pfsync_state) * nr;
3425 			break;
3426 		}
3427 
3428 		out = ps->ps_states;
3429 		pstore = mallocarray(slice_count,
3430 		    sizeof(struct pfsync_state), M_TEMP, M_WAITOK | M_ZERO);
3431 		nr = 0;
3432 
3433 		for (i = 0; i <= pf_hashmask; i++) {
3434 			struct pf_idhash *ih = &V_pf_idhash[i];
3435 
3436 DIOCGETSTATES_retry:
3437 			p = pstore;
3438 
3439 			if (LIST_EMPTY(&ih->states))
3440 				continue;
3441 
3442 			PF_HASHROW_LOCK(ih);
3443 			count = 0;
3444 			LIST_FOREACH(s, &ih->states, entry) {
3445 				if (s->timeout == PFTM_UNLINKED)
3446 					continue;
3447 				count++;
3448 			}
3449 
3450 			if (count > slice_count) {
3451 				PF_HASHROW_UNLOCK(ih);
3452 				free(pstore, M_TEMP);
3453 				slice_count = count * 2;
3454 				pstore = mallocarray(slice_count,
3455 				    sizeof(struct pfsync_state), M_TEMP,
3456 				    M_WAITOK | M_ZERO);
3457 				goto DIOCGETSTATES_retry;
3458 			}
3459 
3460 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
3461 				PF_HASHROW_UNLOCK(ih);
3462 				goto DIOCGETSTATES_full;
3463 			}
3464 
3465 			LIST_FOREACH(s, &ih->states, entry) {
3466 				if (s->timeout == PFTM_UNLINKED)
3467 					continue;
3468 
3469 				pfsync_state_export(p, s);
3470 				p++;
3471 				nr++;
3472 			}
3473 			PF_HASHROW_UNLOCK(ih);
3474 			error = copyout(pstore, out,
3475 			    sizeof(struct pfsync_state) * count);
3476 			if (error)
3477 				break;
3478 			out = ps->ps_states + nr;
3479 		}
3480 DIOCGETSTATES_full:
3481 		ps->ps_len = sizeof(struct pfsync_state) * nr;
3482 		free(pstore, M_TEMP);
3483 
3484 		break;
3485 	}
3486 
3487 	case DIOCGETSTATESV2: {
3488 		struct pfioc_states_v2	*ps = (struct pfioc_states_v2 *)addr;
3489 		struct pf_kstate	*s;
3490 		struct pf_state_export	*pstore, *p;
3491 		int i, nr;
3492 		size_t slice_count = 16, count;
3493 		void *out;
3494 
3495 		if (ps->ps_req_version > PF_STATE_VERSION) {
3496 			error = ENOTSUP;
3497 			break;
3498 		}
3499 
3500 		if (ps->ps_len <= 0) {
3501 			nr = uma_zone_get_cur(V_pf_state_z);
3502 			ps->ps_len = sizeof(struct pf_state_export) * nr;
3503 			break;
3504 		}
3505 
3506 		out = ps->ps_states;
3507 		pstore = mallocarray(slice_count,
3508 		    sizeof(struct pf_state_export), M_TEMP, M_WAITOK | M_ZERO);
3509 		nr = 0;
3510 
3511 		for (i = 0; i <= pf_hashmask; i++) {
3512 			struct pf_idhash *ih = &V_pf_idhash[i];
3513 
3514 DIOCGETSTATESV2_retry:
3515 			p = pstore;
3516 
3517 			if (LIST_EMPTY(&ih->states))
3518 				continue;
3519 
3520 			PF_HASHROW_LOCK(ih);
3521 			count = 0;
3522 			LIST_FOREACH(s, &ih->states, entry) {
3523 				if (s->timeout == PFTM_UNLINKED)
3524 					continue;
3525 				count++;
3526 			}
3527 
3528 			if (count > slice_count) {
3529 				PF_HASHROW_UNLOCK(ih);
3530 				free(pstore, M_TEMP);
3531 				slice_count = count * 2;
3532 				pstore = mallocarray(slice_count,
3533 				    sizeof(struct pf_state_export), M_TEMP,
3534 				    M_WAITOK | M_ZERO);
3535 				goto DIOCGETSTATESV2_retry;
3536 			}
3537 
3538 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
3539 				PF_HASHROW_UNLOCK(ih);
3540 				goto DIOCGETSTATESV2_full;
3541 			}
3542 
3543 			LIST_FOREACH(s, &ih->states, entry) {
3544 				if (s->timeout == PFTM_UNLINKED)
3545 					continue;
3546 
3547 				pf_state_export(p, s);
3548 				p++;
3549 				nr++;
3550 			}
3551 			PF_HASHROW_UNLOCK(ih);
3552 			error = copyout(pstore, out,
3553 			    sizeof(struct pf_state_export) * count);
3554 			if (error)
3555 				break;
3556 			out = ps->ps_states + nr;
3557 		}
3558 DIOCGETSTATESV2_full:
3559 		ps->ps_len = nr * sizeof(struct pf_state_export);
3560 		free(pstore, M_TEMP);
3561 
3562 		break;
3563 	}
3564 
3565 	case DIOCGETSTATUS: {
3566 		struct pf_status *s = (struct pf_status *)addr;
3567 
3568 		PF_RULES_RLOCK();
3569 		s->running = V_pf_status.running;
3570 		s->since   = V_pf_status.since;
3571 		s->debug   = V_pf_status.debug;
3572 		s->hostid  = V_pf_status.hostid;
3573 		s->states  = V_pf_status.states;
3574 		s->src_nodes = V_pf_status.src_nodes;
3575 
3576 		for (int i = 0; i < PFRES_MAX; i++)
3577 			s->counters[i] =
3578 			    counter_u64_fetch(V_pf_status.counters[i]);
3579 		for (int i = 0; i < LCNT_MAX; i++)
3580 			s->lcounters[i] =
3581 			    counter_u64_fetch(V_pf_status.lcounters[i]);
3582 		for (int i = 0; i < FCNT_MAX; i++)
3583 			s->fcounters[i] =
3584 			    pf_counter_u64_fetch(&V_pf_status.fcounters[i]);
3585 		for (int i = 0; i < SCNT_MAX; i++)
3586 			s->scounters[i] =
3587 			    counter_u64_fetch(V_pf_status.scounters[i]);
3588 
3589 		bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ);
3590 		bcopy(V_pf_status.pf_chksum, s->pf_chksum,
3591 		    PF_MD5_DIGEST_LENGTH);
3592 
3593 		pfi_update_status(s->ifname, s);
3594 		PF_RULES_RUNLOCK();
3595 		break;
3596 	}
3597 
3598 	case DIOCGETSTATUSNV: {
3599 		error = pf_getstatus((struct pfioc_nv *)addr);
3600 		break;
3601 	}
3602 
3603 	case DIOCSETSTATUSIF: {
3604 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
3605 
3606 		if (pi->ifname[0] == 0) {
3607 			bzero(V_pf_status.ifname, IFNAMSIZ);
3608 			break;
3609 		}
3610 		PF_RULES_WLOCK();
3611 		error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
3612 		PF_RULES_WUNLOCK();
3613 		break;
3614 	}
3615 
3616 	case DIOCCLRSTATUS: {
3617 		PF_RULES_WLOCK();
3618 		for (int i = 0; i < PFRES_MAX; i++)
3619 			counter_u64_zero(V_pf_status.counters[i]);
3620 		for (int i = 0; i < FCNT_MAX; i++)
3621 			pf_counter_u64_zero(&V_pf_status.fcounters[i]);
3622 		for (int i = 0; i < SCNT_MAX; i++)
3623 			counter_u64_zero(V_pf_status.scounters[i]);
3624 		for (int i = 0; i < KLCNT_MAX; i++)
3625 			counter_u64_zero(V_pf_status.lcounters[i]);
3626 		V_pf_status.since = time_second;
3627 		if (*V_pf_status.ifname)
3628 			pfi_update_status(V_pf_status.ifname, NULL);
3629 		PF_RULES_WUNLOCK();
3630 		break;
3631 	}
3632 
3633 	case DIOCNATLOOK: {
3634 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
3635 		struct pf_state_key	*sk;
3636 		struct pf_kstate	*state;
3637 		struct pf_state_key_cmp	 key;
3638 		int			 m = 0, direction = pnl->direction;
3639 		int			 sidx, didx;
3640 
3641 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
3642 		sidx = (direction == PF_IN) ? 1 : 0;
3643 		didx = (direction == PF_IN) ? 0 : 1;
3644 
3645 		if (!pnl->proto ||
3646 		    PF_AZERO(&pnl->saddr, pnl->af) ||
3647 		    PF_AZERO(&pnl->daddr, pnl->af) ||
3648 		    ((pnl->proto == IPPROTO_TCP ||
3649 		    pnl->proto == IPPROTO_UDP) &&
3650 		    (!pnl->dport || !pnl->sport)))
3651 			error = EINVAL;
3652 		else {
3653 			bzero(&key, sizeof(key));
3654 			key.af = pnl->af;
3655 			key.proto = pnl->proto;
3656 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
3657 			key.port[sidx] = pnl->sport;
3658 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
3659 			key.port[didx] = pnl->dport;
3660 
3661 			state = pf_find_state_all(&key, direction, &m);
3662 			if (state == NULL) {
3663 				error = ENOENT;
3664 			} else {
3665 				if (m > 1) {
3666 					PF_STATE_UNLOCK(state);
3667 					error = E2BIG;	/* more than one state */
3668 				} else {
3669 					sk = state->key[sidx];
3670 					PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
3671 					pnl->rsport = sk->port[sidx];
3672 					PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
3673 					pnl->rdport = sk->port[didx];
3674 					PF_STATE_UNLOCK(state);
3675 				}
3676 			}
3677 		}
3678 		break;
3679 	}
3680 
3681 	case DIOCSETTIMEOUT: {
3682 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
3683 		int		 old;
3684 
3685 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
3686 		    pt->seconds < 0) {
3687 			error = EINVAL;
3688 			break;
3689 		}
3690 		PF_RULES_WLOCK();
3691 		old = V_pf_default_rule.timeout[pt->timeout];
3692 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
3693 			pt->seconds = 1;
3694 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
3695 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
3696 			wakeup(pf_purge_thread);
3697 		pt->seconds = old;
3698 		PF_RULES_WUNLOCK();
3699 		break;
3700 	}
3701 
3702 	case DIOCGETTIMEOUT: {
3703 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
3704 
3705 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
3706 			error = EINVAL;
3707 			break;
3708 		}
3709 		PF_RULES_RLOCK();
3710 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
3711 		PF_RULES_RUNLOCK();
3712 		break;
3713 	}
3714 
3715 	case DIOCGETLIMIT: {
3716 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
3717 
3718 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
3719 			error = EINVAL;
3720 			break;
3721 		}
3722 		PF_RULES_RLOCK();
3723 		pl->limit = V_pf_limits[pl->index].limit;
3724 		PF_RULES_RUNLOCK();
3725 		break;
3726 	}
3727 
3728 	case DIOCSETLIMIT: {
3729 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
3730 		int			 old_limit;
3731 
3732 		PF_RULES_WLOCK();
3733 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
3734 		    V_pf_limits[pl->index].zone == NULL) {
3735 			PF_RULES_WUNLOCK();
3736 			error = EINVAL;
3737 			break;
3738 		}
3739 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
3740 		old_limit = V_pf_limits[pl->index].limit;
3741 		V_pf_limits[pl->index].limit = pl->limit;
3742 		pl->limit = old_limit;
3743 		PF_RULES_WUNLOCK();
3744 		break;
3745 	}
3746 
3747 	case DIOCSETDEBUG: {
3748 		u_int32_t	*level = (u_int32_t *)addr;
3749 
3750 		PF_RULES_WLOCK();
3751 		V_pf_status.debug = *level;
3752 		PF_RULES_WUNLOCK();
3753 		break;
3754 	}
3755 
3756 	case DIOCCLRRULECTRS: {
3757 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
3758 		struct pf_kruleset	*ruleset = &pf_main_ruleset;
3759 		struct pf_krule		*rule;
3760 
3761 		PF_RULES_WLOCK();
3762 		TAILQ_FOREACH(rule,
3763 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
3764 			pf_counter_u64_zero(&rule->evaluations);
3765 			for (int i = 0; i < 2; i++) {
3766 				pf_counter_u64_zero(&rule->packets[i]);
3767 				pf_counter_u64_zero(&rule->bytes[i]);
3768 			}
3769 		}
3770 		PF_RULES_WUNLOCK();
3771 		break;
3772 	}
3773 
3774 	case DIOCGIFSPEEDV0:
3775 	case DIOCGIFSPEEDV1: {
3776 		struct pf_ifspeed_v1	*psp = (struct pf_ifspeed_v1 *)addr;
3777 		struct pf_ifspeed_v1	ps;
3778 		struct ifnet		*ifp;
3779 
3780 		if (psp->ifname[0] == '\0') {
3781 			error = EINVAL;
3782 			break;
3783 		}
3784 
3785 		error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ);
3786 		if (error != 0)
3787 			break;
3788 		ifp = ifunit(ps.ifname);
3789 		if (ifp != NULL) {
3790 			psp->baudrate32 =
3791 			    (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
3792 			if (cmd == DIOCGIFSPEEDV1)
3793 				psp->baudrate = ifp->if_baudrate;
3794 		} else {
3795 			error = EINVAL;
3796 		}
3797 		break;
3798 	}
3799 
3800 #ifdef ALTQ
3801 	case DIOCSTARTALTQ: {
3802 		struct pf_altq		*altq;
3803 
3804 		PF_RULES_WLOCK();
3805 		/* enable all altq interfaces on active list */
3806 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
3807 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
3808 				error = pf_enable_altq(altq);
3809 				if (error != 0)
3810 					break;
3811 			}
3812 		}
3813 		if (error == 0)
3814 			V_pf_altq_running = 1;
3815 		PF_RULES_WUNLOCK();
3816 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
3817 		break;
3818 	}
3819 
3820 	case DIOCSTOPALTQ: {
3821 		struct pf_altq		*altq;
3822 
3823 		PF_RULES_WLOCK();
3824 		/* disable all altq interfaces on active list */
3825 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
3826 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
3827 				error = pf_disable_altq(altq);
3828 				if (error != 0)
3829 					break;
3830 			}
3831 		}
3832 		if (error == 0)
3833 			V_pf_altq_running = 0;
3834 		PF_RULES_WUNLOCK();
3835 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
3836 		break;
3837 	}
3838 
3839 	case DIOCADDALTQV0:
3840 	case DIOCADDALTQV1: {
3841 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
3842 		struct pf_altq		*altq, *a;
3843 		struct ifnet		*ifp;
3844 
3845 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
3846 		error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
3847 		if (error)
3848 			break;
3849 		altq->local_flags = 0;
3850 
3851 		PF_RULES_WLOCK();
3852 		if (pa->ticket != V_ticket_altqs_inactive) {
3853 			PF_RULES_WUNLOCK();
3854 			free(altq, M_PFALTQ);
3855 			error = EBUSY;
3856 			break;
3857 		}
3858 
3859 		/*
3860 		 * if this is for a queue, find the discipline and
3861 		 * copy the necessary fields
3862 		 */
3863 		if (altq->qname[0] != 0) {
3864 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
3865 				PF_RULES_WUNLOCK();
3866 				error = EBUSY;
3867 				free(altq, M_PFALTQ);
3868 				break;
3869 			}
3870 			altq->altq_disc = NULL;
3871 			TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
3872 				if (strncmp(a->ifname, altq->ifname,
3873 				    IFNAMSIZ) == 0) {
3874 					altq->altq_disc = a->altq_disc;
3875 					break;
3876 				}
3877 			}
3878 		}
3879 
3880 		if ((ifp = ifunit(altq->ifname)) == NULL)
3881 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
3882 		else
3883 			error = altq_add(ifp, altq);
3884 
3885 		if (error) {
3886 			PF_RULES_WUNLOCK();
3887 			free(altq, M_PFALTQ);
3888 			break;
3889 		}
3890 
3891 		if (altq->qname[0] != 0)
3892 			TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
3893 		else
3894 			TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
3895 		/* version error check done on import above */
3896 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
3897 		PF_RULES_WUNLOCK();
3898 		break;
3899 	}
3900 
3901 	case DIOCGETALTQSV0:
3902 	case DIOCGETALTQSV1: {
3903 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
3904 		struct pf_altq		*altq;
3905 
3906 		PF_RULES_RLOCK();
3907 		pa->nr = 0;
3908 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
3909 			pa->nr++;
3910 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
3911 			pa->nr++;
3912 		pa->ticket = V_ticket_altqs_active;
3913 		PF_RULES_RUNLOCK();
3914 		break;
3915 	}
3916 
3917 	case DIOCGETALTQV0:
3918 	case DIOCGETALTQV1: {
3919 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
3920 		struct pf_altq		*altq;
3921 
3922 		PF_RULES_RLOCK();
3923 		if (pa->ticket != V_ticket_altqs_active) {
3924 			PF_RULES_RUNLOCK();
3925 			error = EBUSY;
3926 			break;
3927 		}
3928 		altq = pf_altq_get_nth_active(pa->nr);
3929 		if (altq == NULL) {
3930 			PF_RULES_RUNLOCK();
3931 			error = EBUSY;
3932 			break;
3933 		}
3934 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
3935 		PF_RULES_RUNLOCK();
3936 		break;
3937 	}
3938 
3939 	case DIOCCHANGEALTQV0:
3940 	case DIOCCHANGEALTQV1:
3941 		/* CHANGEALTQ not supported yet! */
3942 		error = ENODEV;
3943 		break;
3944 
3945 	case DIOCGETQSTATSV0:
3946 	case DIOCGETQSTATSV1: {
3947 		struct pfioc_qstats_v1	*pq = (struct pfioc_qstats_v1 *)addr;
3948 		struct pf_altq		*altq;
3949 		int			 nbytes;
3950 		u_int32_t		 version;
3951 
3952 		PF_RULES_RLOCK();
3953 		if (pq->ticket != V_ticket_altqs_active) {
3954 			PF_RULES_RUNLOCK();
3955 			error = EBUSY;
3956 			break;
3957 		}
3958 		nbytes = pq->nbytes;
3959 		altq = pf_altq_get_nth_active(pq->nr);
3960 		if (altq == NULL) {
3961 			PF_RULES_RUNLOCK();
3962 			error = EBUSY;
3963 			break;
3964 		}
3965 
3966 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
3967 			PF_RULES_RUNLOCK();
3968 			error = ENXIO;
3969 			break;
3970 		}
3971 		PF_RULES_RUNLOCK();
3972 		if (cmd == DIOCGETQSTATSV0)
3973 			version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
3974 		else
3975 			version = pq->version;
3976 		error = altq_getqstats(altq, pq->buf, &nbytes, version);
3977 		if (error == 0) {
3978 			pq->scheduler = altq->scheduler;
3979 			pq->nbytes = nbytes;
3980 		}
3981 		break;
3982 	}
3983 #endif /* ALTQ */
3984 
3985 	case DIOCBEGINADDRS: {
3986 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
3987 
3988 		PF_RULES_WLOCK();
3989 		pf_empty_kpool(&V_pf_pabuf);
3990 		pp->ticket = ++V_ticket_pabuf;
3991 		PF_RULES_WUNLOCK();
3992 		break;
3993 	}
3994 
3995 	case DIOCADDADDR: {
3996 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
3997 		struct pf_kpooladdr	*pa;
3998 		struct pfi_kkif		*kif = NULL;
3999 
4000 #ifndef INET
4001 		if (pp->af == AF_INET) {
4002 			error = EAFNOSUPPORT;
4003 			break;
4004 		}
4005 #endif /* INET */
4006 #ifndef INET6
4007 		if (pp->af == AF_INET6) {
4008 			error = EAFNOSUPPORT;
4009 			break;
4010 		}
4011 #endif /* INET6 */
4012 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
4013 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
4014 		    pp->addr.addr.type != PF_ADDR_TABLE) {
4015 			error = EINVAL;
4016 			break;
4017 		}
4018 		if (pp->addr.addr.p.dyn != NULL) {
4019 			error = EINVAL;
4020 			break;
4021 		}
4022 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
4023 		error = pf_pooladdr_to_kpooladdr(&pp->addr, pa);
4024 		if (error != 0)
4025 			break;
4026 		if (pa->ifname[0])
4027 			kif = pf_kkif_create(M_WAITOK);
4028 		PF_RULES_WLOCK();
4029 		if (pp->ticket != V_ticket_pabuf) {
4030 			PF_RULES_WUNLOCK();
4031 			if (pa->ifname[0])
4032 				pf_kkif_free(kif);
4033 			free(pa, M_PFRULE);
4034 			error = EBUSY;
4035 			break;
4036 		}
4037 		if (pa->ifname[0]) {
4038 			pa->kif = pfi_kkif_attach(kif, pa->ifname);
4039 			kif = NULL;
4040 			pfi_kkif_ref(pa->kif);
4041 		} else
4042 			pa->kif = NULL;
4043 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
4044 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
4045 			if (pa->ifname[0])
4046 				pfi_kkif_unref(pa->kif);
4047 			PF_RULES_WUNLOCK();
4048 			free(pa, M_PFRULE);
4049 			break;
4050 		}
4051 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
4052 		PF_RULES_WUNLOCK();
4053 		break;
4054 	}
4055 
4056 	case DIOCGETADDRS: {
4057 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4058 		struct pf_kpool		*pool;
4059 		struct pf_kpooladdr	*pa;
4060 
4061 		pp->anchor[sizeof(pp->anchor) - 1] = 0;
4062 		pp->nr = 0;
4063 
4064 		PF_RULES_RLOCK();
4065 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
4066 		    pp->r_num, 0, 1, 0);
4067 		if (pool == NULL) {
4068 			PF_RULES_RUNLOCK();
4069 			error = EBUSY;
4070 			break;
4071 		}
4072 		TAILQ_FOREACH(pa, &pool->list, entries)
4073 			pp->nr++;
4074 		PF_RULES_RUNLOCK();
4075 		break;
4076 	}
4077 
4078 	case DIOCGETADDR: {
4079 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4080 		struct pf_kpool		*pool;
4081 		struct pf_kpooladdr	*pa;
4082 		u_int32_t		 nr = 0;
4083 
4084 		pp->anchor[sizeof(pp->anchor) - 1] = 0;
4085 
4086 		PF_RULES_RLOCK();
4087 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
4088 		    pp->r_num, 0, 1, 1);
4089 		if (pool == NULL) {
4090 			PF_RULES_RUNLOCK();
4091 			error = EBUSY;
4092 			break;
4093 		}
4094 		pa = TAILQ_FIRST(&pool->list);
4095 		while ((pa != NULL) && (nr < pp->nr)) {
4096 			pa = TAILQ_NEXT(pa, entries);
4097 			nr++;
4098 		}
4099 		if (pa == NULL) {
4100 			PF_RULES_RUNLOCK();
4101 			error = EBUSY;
4102 			break;
4103 		}
4104 		pf_kpooladdr_to_pooladdr(pa, &pp->addr);
4105 		pf_addr_copyout(&pp->addr.addr);
4106 		PF_RULES_RUNLOCK();
4107 		break;
4108 	}
4109 
4110 	case DIOCCHANGEADDR: {
4111 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
4112 		struct pf_kpool		*pool;
4113 		struct pf_kpooladdr	*oldpa = NULL, *newpa = NULL;
4114 		struct pf_kruleset	*ruleset;
4115 		struct pfi_kkif		*kif = NULL;
4116 
4117 		pca->anchor[sizeof(pca->anchor) - 1] = 0;
4118 
4119 		if (pca->action < PF_CHANGE_ADD_HEAD ||
4120 		    pca->action > PF_CHANGE_REMOVE) {
4121 			error = EINVAL;
4122 			break;
4123 		}
4124 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
4125 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
4126 		    pca->addr.addr.type != PF_ADDR_TABLE) {
4127 			error = EINVAL;
4128 			break;
4129 		}
4130 		if (pca->addr.addr.p.dyn != NULL) {
4131 			error = EINVAL;
4132 			break;
4133 		}
4134 
4135 		if (pca->action != PF_CHANGE_REMOVE) {
4136 #ifndef INET
4137 			if (pca->af == AF_INET) {
4138 				error = EAFNOSUPPORT;
4139 				break;
4140 			}
4141 #endif /* INET */
4142 #ifndef INET6
4143 			if (pca->af == AF_INET6) {
4144 				error = EAFNOSUPPORT;
4145 				break;
4146 			}
4147 #endif /* INET6 */
4148 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
4149 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
4150 			if (newpa->ifname[0])
4151 				kif = pf_kkif_create(M_WAITOK);
4152 			newpa->kif = NULL;
4153 		}
4154 #define	ERROUT(x)	ERROUT_IOCTL(DIOCCHANGEADDR_error, x)
4155 		PF_RULES_WLOCK();
4156 		ruleset = pf_find_kruleset(pca->anchor);
4157 		if (ruleset == NULL)
4158 			ERROUT(EBUSY);
4159 
4160 		pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
4161 		    pca->r_num, pca->r_last, 1, 1);
4162 		if (pool == NULL)
4163 			ERROUT(EBUSY);
4164 
4165 		if (pca->action != PF_CHANGE_REMOVE) {
4166 			if (newpa->ifname[0]) {
4167 				newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
4168 				pfi_kkif_ref(newpa->kif);
4169 				kif = NULL;
4170 			}
4171 
4172 			switch (newpa->addr.type) {
4173 			case PF_ADDR_DYNIFTL:
4174 				error = pfi_dynaddr_setup(&newpa->addr,
4175 				    pca->af);
4176 				break;
4177 			case PF_ADDR_TABLE:
4178 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
4179 				    newpa->addr.v.tblname);
4180 				if (newpa->addr.p.tbl == NULL)
4181 					error = ENOMEM;
4182 				break;
4183 			}
4184 			if (error)
4185 				goto DIOCCHANGEADDR_error;
4186 		}
4187 
4188 		switch (pca->action) {
4189 		case PF_CHANGE_ADD_HEAD:
4190 			oldpa = TAILQ_FIRST(&pool->list);
4191 			break;
4192 		case PF_CHANGE_ADD_TAIL:
4193 			oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
4194 			break;
4195 		default:
4196 			oldpa = TAILQ_FIRST(&pool->list);
4197 			for (int i = 0; oldpa && i < pca->nr; i++)
4198 				oldpa = TAILQ_NEXT(oldpa, entries);
4199 
4200 			if (oldpa == NULL)
4201 				ERROUT(EINVAL);
4202 		}
4203 
4204 		if (pca->action == PF_CHANGE_REMOVE) {
4205 			TAILQ_REMOVE(&pool->list, oldpa, entries);
4206 			switch (oldpa->addr.type) {
4207 			case PF_ADDR_DYNIFTL:
4208 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
4209 				break;
4210 			case PF_ADDR_TABLE:
4211 				pfr_detach_table(oldpa->addr.p.tbl);
4212 				break;
4213 			}
4214 			if (oldpa->kif)
4215 				pfi_kkif_unref(oldpa->kif);
4216 			free(oldpa, M_PFRULE);
4217 		} else {
4218 			if (oldpa == NULL)
4219 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
4220 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
4221 			    pca->action == PF_CHANGE_ADD_BEFORE)
4222 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
4223 			else
4224 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
4225 				    newpa, entries);
4226 		}
4227 
4228 		pool->cur = TAILQ_FIRST(&pool->list);
4229 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
4230 		PF_RULES_WUNLOCK();
4231 		break;
4232 
4233 #undef ERROUT
4234 DIOCCHANGEADDR_error:
4235 		if (newpa != NULL) {
4236 			if (newpa->kif)
4237 				pfi_kkif_unref(newpa->kif);
4238 			free(newpa, M_PFRULE);
4239 		}
4240 		PF_RULES_WUNLOCK();
4241 		pf_kkif_free(kif);
4242 		break;
4243 	}
4244 
4245 	case DIOCGETRULESETS: {
4246 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
4247 		struct pf_kruleset	*ruleset;
4248 		struct pf_kanchor	*anchor;
4249 
4250 		pr->path[sizeof(pr->path) - 1] = 0;
4251 
4252 		PF_RULES_RLOCK();
4253 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4254 			PF_RULES_RUNLOCK();
4255 			error = ENOENT;
4256 			break;
4257 		}
4258 		pr->nr = 0;
4259 		if (ruleset->anchor == NULL) {
4260 			/* XXX kludge for pf_main_ruleset */
4261 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4262 				if (anchor->parent == NULL)
4263 					pr->nr++;
4264 		} else {
4265 			RB_FOREACH(anchor, pf_kanchor_node,
4266 			    &ruleset->anchor->children)
4267 				pr->nr++;
4268 		}
4269 		PF_RULES_RUNLOCK();
4270 		break;
4271 	}
4272 
4273 	case DIOCGETRULESET: {
4274 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
4275 		struct pf_kruleset	*ruleset;
4276 		struct pf_kanchor	*anchor;
4277 		u_int32_t		 nr = 0;
4278 
4279 		pr->path[sizeof(pr->path) - 1] = 0;
4280 
4281 		PF_RULES_RLOCK();
4282 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4283 			PF_RULES_RUNLOCK();
4284 			error = ENOENT;
4285 			break;
4286 		}
4287 		pr->name[0] = 0;
4288 		if (ruleset->anchor == NULL) {
4289 			/* XXX kludge for pf_main_ruleset */
4290 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4291 				if (anchor->parent == NULL && nr++ == pr->nr) {
4292 					strlcpy(pr->name, anchor->name,
4293 					    sizeof(pr->name));
4294 					break;
4295 				}
4296 		} else {
4297 			RB_FOREACH(anchor, pf_kanchor_node,
4298 			    &ruleset->anchor->children)
4299 				if (nr++ == pr->nr) {
4300 					strlcpy(pr->name, anchor->name,
4301 					    sizeof(pr->name));
4302 					break;
4303 				}
4304 		}
4305 		if (!pr->name[0])
4306 			error = EBUSY;
4307 		PF_RULES_RUNLOCK();
4308 		break;
4309 	}
4310 
4311 	case DIOCRCLRTABLES: {
4312 		struct pfioc_table *io = (struct pfioc_table *)addr;
4313 
4314 		if (io->pfrio_esize != 0) {
4315 			error = ENODEV;
4316 			break;
4317 		}
4318 		PF_RULES_WLOCK();
4319 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
4320 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4321 		PF_RULES_WUNLOCK();
4322 		break;
4323 	}
4324 
4325 	case DIOCRADDTABLES: {
4326 		struct pfioc_table *io = (struct pfioc_table *)addr;
4327 		struct pfr_table *pfrts;
4328 		size_t totlen;
4329 
4330 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4331 			error = ENODEV;
4332 			break;
4333 		}
4334 
4335 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4336 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4337 			error = ENOMEM;
4338 			break;
4339 		}
4340 
4341 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4342 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4343 		    M_TEMP, M_WAITOK);
4344 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4345 		if (error) {
4346 			free(pfrts, M_TEMP);
4347 			break;
4348 		}
4349 		PF_RULES_WLOCK();
4350 		error = pfr_add_tables(pfrts, io->pfrio_size,
4351 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4352 		PF_RULES_WUNLOCK();
4353 		free(pfrts, M_TEMP);
4354 		break;
4355 	}
4356 
4357 	case DIOCRDELTABLES: {
4358 		struct pfioc_table *io = (struct pfioc_table *)addr;
4359 		struct pfr_table *pfrts;
4360 		size_t totlen;
4361 
4362 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4363 			error = ENODEV;
4364 			break;
4365 		}
4366 
4367 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4368 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4369 			error = ENOMEM;
4370 			break;
4371 		}
4372 
4373 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4374 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4375 		    M_TEMP, M_WAITOK);
4376 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4377 		if (error) {
4378 			free(pfrts, M_TEMP);
4379 			break;
4380 		}
4381 		PF_RULES_WLOCK();
4382 		error = pfr_del_tables(pfrts, io->pfrio_size,
4383 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4384 		PF_RULES_WUNLOCK();
4385 		free(pfrts, M_TEMP);
4386 		break;
4387 	}
4388 
4389 	case DIOCRGETTABLES: {
4390 		struct pfioc_table *io = (struct pfioc_table *)addr;
4391 		struct pfr_table *pfrts;
4392 		size_t totlen;
4393 		int n;
4394 
4395 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4396 			error = ENODEV;
4397 			break;
4398 		}
4399 		PF_RULES_RLOCK();
4400 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4401 		if (n < 0) {
4402 			PF_RULES_RUNLOCK();
4403 			error = EINVAL;
4404 			break;
4405 		}
4406 		io->pfrio_size = min(io->pfrio_size, n);
4407 
4408 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4409 
4410 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4411 		    M_TEMP, M_NOWAIT | M_ZERO);
4412 		if (pfrts == NULL) {
4413 			error = ENOMEM;
4414 			PF_RULES_RUNLOCK();
4415 			break;
4416 		}
4417 		error = pfr_get_tables(&io->pfrio_table, pfrts,
4418 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4419 		PF_RULES_RUNLOCK();
4420 		if (error == 0)
4421 			error = copyout(pfrts, io->pfrio_buffer, totlen);
4422 		free(pfrts, M_TEMP);
4423 		break;
4424 	}
4425 
4426 	case DIOCRGETTSTATS: {
4427 		struct pfioc_table *io = (struct pfioc_table *)addr;
4428 		struct pfr_tstats *pfrtstats;
4429 		size_t totlen;
4430 		int n;
4431 
4432 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
4433 			error = ENODEV;
4434 			break;
4435 		}
4436 		PF_TABLE_STATS_LOCK();
4437 		PF_RULES_RLOCK();
4438 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4439 		if (n < 0) {
4440 			PF_RULES_RUNLOCK();
4441 			PF_TABLE_STATS_UNLOCK();
4442 			error = EINVAL;
4443 			break;
4444 		}
4445 		io->pfrio_size = min(io->pfrio_size, n);
4446 
4447 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
4448 		pfrtstats = mallocarray(io->pfrio_size,
4449 		    sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT | M_ZERO);
4450 		if (pfrtstats == NULL) {
4451 			error = ENOMEM;
4452 			PF_RULES_RUNLOCK();
4453 			PF_TABLE_STATS_UNLOCK();
4454 			break;
4455 		}
4456 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
4457 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4458 		PF_RULES_RUNLOCK();
4459 		PF_TABLE_STATS_UNLOCK();
4460 		if (error == 0)
4461 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
4462 		free(pfrtstats, M_TEMP);
4463 		break;
4464 	}
4465 
4466 	case DIOCRCLRTSTATS: {
4467 		struct pfioc_table *io = (struct pfioc_table *)addr;
4468 		struct pfr_table *pfrts;
4469 		size_t totlen;
4470 
4471 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4472 			error = ENODEV;
4473 			break;
4474 		}
4475 
4476 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4477 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4478 			/* We used to count tables and use the minimum required
4479 			 * size, so we didn't fail on overly large requests.
4480 			 * Keep doing so. */
4481 			io->pfrio_size = pf_ioctl_maxcount;
4482 			break;
4483 		}
4484 
4485 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4486 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4487 		    M_TEMP, M_WAITOK);
4488 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4489 		if (error) {
4490 			free(pfrts, M_TEMP);
4491 			break;
4492 		}
4493 
4494 		PF_TABLE_STATS_LOCK();
4495 		PF_RULES_RLOCK();
4496 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
4497 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4498 		PF_RULES_RUNLOCK();
4499 		PF_TABLE_STATS_UNLOCK();
4500 		free(pfrts, M_TEMP);
4501 		break;
4502 	}
4503 
4504 	case DIOCRSETTFLAGS: {
4505 		struct pfioc_table *io = (struct pfioc_table *)addr;
4506 		struct pfr_table *pfrts;
4507 		size_t totlen;
4508 		int n;
4509 
4510 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4511 			error = ENODEV;
4512 			break;
4513 		}
4514 
4515 		PF_RULES_RLOCK();
4516 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4517 		if (n < 0) {
4518 			PF_RULES_RUNLOCK();
4519 			error = EINVAL;
4520 			break;
4521 		}
4522 
4523 		io->pfrio_size = min(io->pfrio_size, n);
4524 		PF_RULES_RUNLOCK();
4525 
4526 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4527 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4528 		    M_TEMP, M_WAITOK);
4529 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4530 		if (error) {
4531 			free(pfrts, M_TEMP);
4532 			break;
4533 		}
4534 		PF_RULES_WLOCK();
4535 		error = pfr_set_tflags(pfrts, io->pfrio_size,
4536 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
4537 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4538 		PF_RULES_WUNLOCK();
4539 		free(pfrts, M_TEMP);
4540 		break;
4541 	}
4542 
4543 	case DIOCRCLRADDRS: {
4544 		struct pfioc_table *io = (struct pfioc_table *)addr;
4545 
4546 		if (io->pfrio_esize != 0) {
4547 			error = ENODEV;
4548 			break;
4549 		}
4550 		PF_RULES_WLOCK();
4551 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
4552 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4553 		PF_RULES_WUNLOCK();
4554 		break;
4555 	}
4556 
4557 	case DIOCRADDADDRS: {
4558 		struct pfioc_table *io = (struct pfioc_table *)addr;
4559 		struct pfr_addr *pfras;
4560 		size_t totlen;
4561 
4562 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4563 			error = ENODEV;
4564 			break;
4565 		}
4566 		if (io->pfrio_size < 0 ||
4567 		    io->pfrio_size > pf_ioctl_maxcount ||
4568 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4569 			error = EINVAL;
4570 			break;
4571 		}
4572 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4573 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4574 		    M_TEMP, M_WAITOK);
4575 		error = copyin(io->pfrio_buffer, pfras, totlen);
4576 		if (error) {
4577 			free(pfras, M_TEMP);
4578 			break;
4579 		}
4580 		PF_RULES_WLOCK();
4581 		error = pfr_add_addrs(&io->pfrio_table, pfras,
4582 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
4583 		    PFR_FLAG_USERIOCTL);
4584 		PF_RULES_WUNLOCK();
4585 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4586 			error = copyout(pfras, io->pfrio_buffer, totlen);
4587 		free(pfras, M_TEMP);
4588 		break;
4589 	}
4590 
4591 	case DIOCRDELADDRS: {
4592 		struct pfioc_table *io = (struct pfioc_table *)addr;
4593 		struct pfr_addr *pfras;
4594 		size_t totlen;
4595 
4596 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4597 			error = ENODEV;
4598 			break;
4599 		}
4600 		if (io->pfrio_size < 0 ||
4601 		    io->pfrio_size > pf_ioctl_maxcount ||
4602 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4603 			error = EINVAL;
4604 			break;
4605 		}
4606 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4607 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4608 		    M_TEMP, M_WAITOK);
4609 		error = copyin(io->pfrio_buffer, pfras, totlen);
4610 		if (error) {
4611 			free(pfras, M_TEMP);
4612 			break;
4613 		}
4614 		PF_RULES_WLOCK();
4615 		error = pfr_del_addrs(&io->pfrio_table, pfras,
4616 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
4617 		    PFR_FLAG_USERIOCTL);
4618 		PF_RULES_WUNLOCK();
4619 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4620 			error = copyout(pfras, io->pfrio_buffer, totlen);
4621 		free(pfras, M_TEMP);
4622 		break;
4623 	}
4624 
4625 	case DIOCRSETADDRS: {
4626 		struct pfioc_table *io = (struct pfioc_table *)addr;
4627 		struct pfr_addr *pfras;
4628 		size_t totlen, count;
4629 
4630 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4631 			error = ENODEV;
4632 			break;
4633 		}
4634 		if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
4635 			error = EINVAL;
4636 			break;
4637 		}
4638 		count = max(io->pfrio_size, io->pfrio_size2);
4639 		if (count > pf_ioctl_maxcount ||
4640 		    WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
4641 			error = EINVAL;
4642 			break;
4643 		}
4644 		totlen = count * sizeof(struct pfr_addr);
4645 		pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
4646 		    M_WAITOK);
4647 		error = copyin(io->pfrio_buffer, pfras, totlen);
4648 		if (error) {
4649 			free(pfras, M_TEMP);
4650 			break;
4651 		}
4652 		PF_RULES_WLOCK();
4653 		error = pfr_set_addrs(&io->pfrio_table, pfras,
4654 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
4655 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
4656 		    PFR_FLAG_USERIOCTL, 0);
4657 		PF_RULES_WUNLOCK();
4658 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4659 			error = copyout(pfras, io->pfrio_buffer, totlen);
4660 		free(pfras, M_TEMP);
4661 		break;
4662 	}
4663 
4664 	case DIOCRGETADDRS: {
4665 		struct pfioc_table *io = (struct pfioc_table *)addr;
4666 		struct pfr_addr *pfras;
4667 		size_t totlen;
4668 
4669 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4670 			error = ENODEV;
4671 			break;
4672 		}
4673 		if (io->pfrio_size < 0 ||
4674 		    io->pfrio_size > pf_ioctl_maxcount ||
4675 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4676 			error = EINVAL;
4677 			break;
4678 		}
4679 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4680 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4681 		    M_TEMP, M_WAITOK | M_ZERO);
4682 		PF_RULES_RLOCK();
4683 		error = pfr_get_addrs(&io->pfrio_table, pfras,
4684 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4685 		PF_RULES_RUNLOCK();
4686 		if (error == 0)
4687 			error = copyout(pfras, io->pfrio_buffer, totlen);
4688 		free(pfras, M_TEMP);
4689 		break;
4690 	}
4691 
4692 	case DIOCRGETASTATS: {
4693 		struct pfioc_table *io = (struct pfioc_table *)addr;
4694 		struct pfr_astats *pfrastats;
4695 		size_t totlen;
4696 
4697 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
4698 			error = ENODEV;
4699 			break;
4700 		}
4701 		if (io->pfrio_size < 0 ||
4702 		    io->pfrio_size > pf_ioctl_maxcount ||
4703 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
4704 			error = EINVAL;
4705 			break;
4706 		}
4707 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
4708 		pfrastats = mallocarray(io->pfrio_size,
4709 		    sizeof(struct pfr_astats), M_TEMP, M_WAITOK | M_ZERO);
4710 		PF_RULES_RLOCK();
4711 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
4712 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4713 		PF_RULES_RUNLOCK();
4714 		if (error == 0)
4715 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
4716 		free(pfrastats, M_TEMP);
4717 		break;
4718 	}
4719 
4720 	case DIOCRCLRASTATS: {
4721 		struct pfioc_table *io = (struct pfioc_table *)addr;
4722 		struct pfr_addr *pfras;
4723 		size_t totlen;
4724 
4725 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4726 			error = ENODEV;
4727 			break;
4728 		}
4729 		if (io->pfrio_size < 0 ||
4730 		    io->pfrio_size > pf_ioctl_maxcount ||
4731 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4732 			error = EINVAL;
4733 			break;
4734 		}
4735 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4736 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4737 		    M_TEMP, M_WAITOK);
4738 		error = copyin(io->pfrio_buffer, pfras, totlen);
4739 		if (error) {
4740 			free(pfras, M_TEMP);
4741 			break;
4742 		}
4743 		PF_RULES_WLOCK();
4744 		error = pfr_clr_astats(&io->pfrio_table, pfras,
4745 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
4746 		    PFR_FLAG_USERIOCTL);
4747 		PF_RULES_WUNLOCK();
4748 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4749 			error = copyout(pfras, io->pfrio_buffer, totlen);
4750 		free(pfras, M_TEMP);
4751 		break;
4752 	}
4753 
4754 	case DIOCRTSTADDRS: {
4755 		struct pfioc_table *io = (struct pfioc_table *)addr;
4756 		struct pfr_addr *pfras;
4757 		size_t totlen;
4758 
4759 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4760 			error = ENODEV;
4761 			break;
4762 		}
4763 		if (io->pfrio_size < 0 ||
4764 		    io->pfrio_size > pf_ioctl_maxcount ||
4765 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4766 			error = EINVAL;
4767 			break;
4768 		}
4769 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4770 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4771 		    M_TEMP, M_WAITOK);
4772 		error = copyin(io->pfrio_buffer, pfras, totlen);
4773 		if (error) {
4774 			free(pfras, M_TEMP);
4775 			break;
4776 		}
4777 		PF_RULES_RLOCK();
4778 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
4779 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
4780 		    PFR_FLAG_USERIOCTL);
4781 		PF_RULES_RUNLOCK();
4782 		if (error == 0)
4783 			error = copyout(pfras, io->pfrio_buffer, totlen);
4784 		free(pfras, M_TEMP);
4785 		break;
4786 	}
4787 
4788 	case DIOCRINADEFINE: {
4789 		struct pfioc_table *io = (struct pfioc_table *)addr;
4790 		struct pfr_addr *pfras;
4791 		size_t totlen;
4792 
4793 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4794 			error = ENODEV;
4795 			break;
4796 		}
4797 		if (io->pfrio_size < 0 ||
4798 		    io->pfrio_size > pf_ioctl_maxcount ||
4799 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4800 			error = EINVAL;
4801 			break;
4802 		}
4803 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4804 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4805 		    M_TEMP, M_WAITOK);
4806 		error = copyin(io->pfrio_buffer, pfras, totlen);
4807 		if (error) {
4808 			free(pfras, M_TEMP);
4809 			break;
4810 		}
4811 		PF_RULES_WLOCK();
4812 		error = pfr_ina_define(&io->pfrio_table, pfras,
4813 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
4814 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4815 		PF_RULES_WUNLOCK();
4816 		free(pfras, M_TEMP);
4817 		break;
4818 	}
4819 
4820 	case DIOCOSFPADD: {
4821 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
4822 		PF_RULES_WLOCK();
4823 		error = pf_osfp_add(io);
4824 		PF_RULES_WUNLOCK();
4825 		break;
4826 	}
4827 
4828 	case DIOCOSFPGET: {
4829 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
4830 		PF_RULES_RLOCK();
4831 		error = pf_osfp_get(io);
4832 		PF_RULES_RUNLOCK();
4833 		break;
4834 	}
4835 
4836 	case DIOCXBEGIN: {
4837 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
4838 		struct pfioc_trans_e	*ioes, *ioe;
4839 		size_t			 totlen;
4840 		int			 i;
4841 
4842 		if (io->esize != sizeof(*ioe)) {
4843 			error = ENODEV;
4844 			break;
4845 		}
4846 		if (io->size < 0 ||
4847 		    io->size > pf_ioctl_maxcount ||
4848 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
4849 			error = EINVAL;
4850 			break;
4851 		}
4852 		totlen = sizeof(struct pfioc_trans_e) * io->size;
4853 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
4854 		    M_TEMP, M_WAITOK);
4855 		error = copyin(io->array, ioes, totlen);
4856 		if (error) {
4857 			free(ioes, M_TEMP);
4858 			break;
4859 		}
4860 		PF_RULES_WLOCK();
4861 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
4862 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
4863 			switch (ioe->rs_num) {
4864 			case PF_RULESET_ETH:
4865 				if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) {
4866 					PF_RULES_WUNLOCK();
4867 					free(ioes, M_TEMP);
4868 					goto fail;
4869 				}
4870 				break;
4871 #ifdef ALTQ
4872 			case PF_RULESET_ALTQ:
4873 				if (ioe->anchor[0]) {
4874 					PF_RULES_WUNLOCK();
4875 					free(ioes, M_TEMP);
4876 					error = EINVAL;
4877 					goto fail;
4878 				}
4879 				if ((error = pf_begin_altq(&ioe->ticket))) {
4880 					PF_RULES_WUNLOCK();
4881 					free(ioes, M_TEMP);
4882 					goto fail;
4883 				}
4884 				break;
4885 #endif /* ALTQ */
4886 			case PF_RULESET_TABLE:
4887 			    {
4888 				struct pfr_table table;
4889 
4890 				bzero(&table, sizeof(table));
4891 				strlcpy(table.pfrt_anchor, ioe->anchor,
4892 				    sizeof(table.pfrt_anchor));
4893 				if ((error = pfr_ina_begin(&table,
4894 				    &ioe->ticket, NULL, 0))) {
4895 					PF_RULES_WUNLOCK();
4896 					free(ioes, M_TEMP);
4897 					goto fail;
4898 				}
4899 				break;
4900 			    }
4901 			default:
4902 				if ((error = pf_begin_rules(&ioe->ticket,
4903 				    ioe->rs_num, ioe->anchor))) {
4904 					PF_RULES_WUNLOCK();
4905 					free(ioes, M_TEMP);
4906 					goto fail;
4907 				}
4908 				break;
4909 			}
4910 		}
4911 		PF_RULES_WUNLOCK();
4912 		error = copyout(ioes, io->array, totlen);
4913 		free(ioes, M_TEMP);
4914 		break;
4915 	}
4916 
4917 	case DIOCXROLLBACK: {
4918 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
4919 		struct pfioc_trans_e	*ioe, *ioes;
4920 		size_t			 totlen;
4921 		int			 i;
4922 
4923 		if (io->esize != sizeof(*ioe)) {
4924 			error = ENODEV;
4925 			break;
4926 		}
4927 		if (io->size < 0 ||
4928 		    io->size > pf_ioctl_maxcount ||
4929 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
4930 			error = EINVAL;
4931 			break;
4932 		}
4933 		totlen = sizeof(struct pfioc_trans_e) * io->size;
4934 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
4935 		    M_TEMP, M_WAITOK);
4936 		error = copyin(io->array, ioes, totlen);
4937 		if (error) {
4938 			free(ioes, M_TEMP);
4939 			break;
4940 		}
4941 		PF_RULES_WLOCK();
4942 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
4943 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
4944 			switch (ioe->rs_num) {
4945 			case PF_RULESET_ETH:
4946 				if ((error = pf_rollback_eth(ioe->ticket,
4947 				    ioe->anchor))) {
4948 					PF_RULES_WUNLOCK();
4949 					free(ioes, M_TEMP);
4950 					goto fail; /* really bad */
4951 				}
4952 				break;
4953 #ifdef ALTQ
4954 			case PF_RULESET_ALTQ:
4955 				if (ioe->anchor[0]) {
4956 					PF_RULES_WUNLOCK();
4957 					free(ioes, M_TEMP);
4958 					error = EINVAL;
4959 					goto fail;
4960 				}
4961 				if ((error = pf_rollback_altq(ioe->ticket))) {
4962 					PF_RULES_WUNLOCK();
4963 					free(ioes, M_TEMP);
4964 					goto fail; /* really bad */
4965 				}
4966 				break;
4967 #endif /* ALTQ */
4968 			case PF_RULESET_TABLE:
4969 			    {
4970 				struct pfr_table table;
4971 
4972 				bzero(&table, sizeof(table));
4973 				strlcpy(table.pfrt_anchor, ioe->anchor,
4974 				    sizeof(table.pfrt_anchor));
4975 				if ((error = pfr_ina_rollback(&table,
4976 				    ioe->ticket, NULL, 0))) {
4977 					PF_RULES_WUNLOCK();
4978 					free(ioes, M_TEMP);
4979 					goto fail; /* really bad */
4980 				}
4981 				break;
4982 			    }
4983 			default:
4984 				if ((error = pf_rollback_rules(ioe->ticket,
4985 				    ioe->rs_num, ioe->anchor))) {
4986 					PF_RULES_WUNLOCK();
4987 					free(ioes, M_TEMP);
4988 					goto fail; /* really bad */
4989 				}
4990 				break;
4991 			}
4992 		}
4993 		PF_RULES_WUNLOCK();
4994 		free(ioes, M_TEMP);
4995 		break;
4996 	}
4997 
4998 	case DIOCXCOMMIT: {
4999 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5000 		struct pfioc_trans_e	*ioe, *ioes;
5001 		struct pf_kruleset	*rs;
5002 		struct pf_keth_ruleset	*ers;
5003 		size_t			 totlen;
5004 		int			 i;
5005 
5006 		if (io->esize != sizeof(*ioe)) {
5007 			error = ENODEV;
5008 			break;
5009 		}
5010 
5011 		if (io->size < 0 ||
5012 		    io->size > pf_ioctl_maxcount ||
5013 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5014 			error = EINVAL;
5015 			break;
5016 		}
5017 
5018 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5019 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5020 		    M_TEMP, M_WAITOK);
5021 		error = copyin(io->array, ioes, totlen);
5022 		if (error) {
5023 			free(ioes, M_TEMP);
5024 			break;
5025 		}
5026 		PF_RULES_WLOCK();
5027 		/* First makes sure everything will succeed. */
5028 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5029 			ioe->anchor[sizeof(ioe->anchor) - 1] = 0;
5030 			switch (ioe->rs_num) {
5031 			case PF_RULESET_ETH:
5032 				ers = pf_find_keth_ruleset(ioe->anchor);
5033 				if (ers == NULL || ioe->ticket == 0 ||
5034 				    ioe->ticket != ers->inactive.ticket) {
5035 					PF_RULES_WUNLOCK();
5036 					free(ioes, M_TEMP);
5037 					error = EINVAL;
5038 					goto fail;
5039 				}
5040 				break;
5041 #ifdef ALTQ
5042 			case PF_RULESET_ALTQ:
5043 				if (ioe->anchor[0]) {
5044 					PF_RULES_WUNLOCK();
5045 					free(ioes, M_TEMP);
5046 					error = EINVAL;
5047 					goto fail;
5048 				}
5049 				if (!V_altqs_inactive_open || ioe->ticket !=
5050 				    V_ticket_altqs_inactive) {
5051 					PF_RULES_WUNLOCK();
5052 					free(ioes, M_TEMP);
5053 					error = EBUSY;
5054 					goto fail;
5055 				}
5056 				break;
5057 #endif /* ALTQ */
5058 			case PF_RULESET_TABLE:
5059 				rs = pf_find_kruleset(ioe->anchor);
5060 				if (rs == NULL || !rs->topen || ioe->ticket !=
5061 				    rs->tticket) {
5062 					PF_RULES_WUNLOCK();
5063 					free(ioes, M_TEMP);
5064 					error = EBUSY;
5065 					goto fail;
5066 				}
5067 				break;
5068 			default:
5069 				if (ioe->rs_num < 0 || ioe->rs_num >=
5070 				    PF_RULESET_MAX) {
5071 					PF_RULES_WUNLOCK();
5072 					free(ioes, M_TEMP);
5073 					error = EINVAL;
5074 					goto fail;
5075 				}
5076 				rs = pf_find_kruleset(ioe->anchor);
5077 				if (rs == NULL ||
5078 				    !rs->rules[ioe->rs_num].inactive.open ||
5079 				    rs->rules[ioe->rs_num].inactive.ticket !=
5080 				    ioe->ticket) {
5081 					PF_RULES_WUNLOCK();
5082 					free(ioes, M_TEMP);
5083 					error = EBUSY;
5084 					goto fail;
5085 				}
5086 				break;
5087 			}
5088 		}
5089 		/* Now do the commit - no errors should happen here. */
5090 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5091 			switch (ioe->rs_num) {
5092 			case PF_RULESET_ETH:
5093 				if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) {
5094 					PF_RULES_WUNLOCK();
5095 					free(ioes, M_TEMP);
5096 					goto fail; /* really bad */
5097 				}
5098 				break;
5099 #ifdef ALTQ
5100 			case PF_RULESET_ALTQ:
5101 				if ((error = pf_commit_altq(ioe->ticket))) {
5102 					PF_RULES_WUNLOCK();
5103 					free(ioes, M_TEMP);
5104 					goto fail; /* really bad */
5105 				}
5106 				break;
5107 #endif /* ALTQ */
5108 			case PF_RULESET_TABLE:
5109 			    {
5110 				struct pfr_table table;
5111 
5112 				bzero(&table, sizeof(table));
5113 				(void)strlcpy(table.pfrt_anchor, ioe->anchor,
5114 				    sizeof(table.pfrt_anchor));
5115 				if ((error = pfr_ina_commit(&table,
5116 				    ioe->ticket, NULL, NULL, 0))) {
5117 					PF_RULES_WUNLOCK();
5118 					free(ioes, M_TEMP);
5119 					goto fail; /* really bad */
5120 				}
5121 				break;
5122 			    }
5123 			default:
5124 				if ((error = pf_commit_rules(ioe->ticket,
5125 				    ioe->rs_num, ioe->anchor))) {
5126 					PF_RULES_WUNLOCK();
5127 					free(ioes, M_TEMP);
5128 					goto fail; /* really bad */
5129 				}
5130 				break;
5131 			}
5132 		}
5133 		PF_RULES_WUNLOCK();
5134 
5135 		/* Only hook into EtherNet taffic if we've got rules for it. */
5136 		if (! TAILQ_EMPTY(V_pf_keth->active.rules))
5137 			hook_pf_eth();
5138 		else
5139 			dehook_pf_eth();
5140 
5141 		free(ioes, M_TEMP);
5142 		break;
5143 	}
5144 
5145 	case DIOCGETSRCNODES: {
5146 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
5147 		struct pf_srchash	*sh;
5148 		struct pf_ksrc_node	*n;
5149 		struct pf_src_node	*p, *pstore;
5150 		uint32_t		 i, nr = 0;
5151 
5152 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5153 				i++, sh++) {
5154 			PF_HASHROW_LOCK(sh);
5155 			LIST_FOREACH(n, &sh->nodes, entry)
5156 				nr++;
5157 			PF_HASHROW_UNLOCK(sh);
5158 		}
5159 
5160 		psn->psn_len = min(psn->psn_len,
5161 		    sizeof(struct pf_src_node) * nr);
5162 
5163 		if (psn->psn_len == 0) {
5164 			psn->psn_len = sizeof(struct pf_src_node) * nr;
5165 			break;
5166 		}
5167 
5168 		nr = 0;
5169 
5170 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
5171 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5172 		    i++, sh++) {
5173 		    PF_HASHROW_LOCK(sh);
5174 		    LIST_FOREACH(n, &sh->nodes, entry) {
5175 
5176 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
5177 				break;
5178 
5179 			pf_src_node_copy(n, p);
5180 
5181 			p++;
5182 			nr++;
5183 		    }
5184 		    PF_HASHROW_UNLOCK(sh);
5185 		}
5186 		error = copyout(pstore, psn->psn_src_nodes,
5187 		    sizeof(struct pf_src_node) * nr);
5188 		if (error) {
5189 			free(pstore, M_TEMP);
5190 			break;
5191 		}
5192 		psn->psn_len = sizeof(struct pf_src_node) * nr;
5193 		free(pstore, M_TEMP);
5194 		break;
5195 	}
5196 
5197 	case DIOCCLRSRCNODES: {
5198 		pf_clear_srcnodes(NULL);
5199 		pf_purge_expired_src_nodes();
5200 		break;
5201 	}
5202 
5203 	case DIOCKILLSRCNODES:
5204 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
5205 		break;
5206 
5207 #ifdef COMPAT_FREEBSD13
5208 	case DIOCKEEPCOUNTERS_FREEBSD13:
5209 #endif
5210 	case DIOCKEEPCOUNTERS:
5211 		error = pf_keepcounters((struct pfioc_nv *)addr);
5212 		break;
5213 
5214 	case DIOCGETSYNCOOKIES:
5215 		error = pf_get_syncookies((struct pfioc_nv *)addr);
5216 		break;
5217 
5218 	case DIOCSETSYNCOOKIES:
5219 		error = pf_set_syncookies((struct pfioc_nv *)addr);
5220 		break;
5221 
5222 	case DIOCSETHOSTID: {
5223 		u_int32_t	*hostid = (u_int32_t *)addr;
5224 
5225 		PF_RULES_WLOCK();
5226 		if (*hostid == 0)
5227 			V_pf_status.hostid = arc4random();
5228 		else
5229 			V_pf_status.hostid = *hostid;
5230 		PF_RULES_WUNLOCK();
5231 		break;
5232 	}
5233 
5234 	case DIOCOSFPFLUSH:
5235 		PF_RULES_WLOCK();
5236 		pf_osfp_flush();
5237 		PF_RULES_WUNLOCK();
5238 		break;
5239 
5240 	case DIOCIGETIFACES: {
5241 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5242 		struct pfi_kif *ifstore;
5243 		size_t bufsiz;
5244 
5245 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
5246 			error = ENODEV;
5247 			break;
5248 		}
5249 
5250 		if (io->pfiio_size < 0 ||
5251 		    io->pfiio_size > pf_ioctl_maxcount ||
5252 		    WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
5253 			error = EINVAL;
5254 			break;
5255 		}
5256 
5257 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
5258 		ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
5259 		    M_TEMP, M_WAITOK | M_ZERO);
5260 
5261 		PF_RULES_RLOCK();
5262 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
5263 		PF_RULES_RUNLOCK();
5264 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
5265 		free(ifstore, M_TEMP);
5266 		break;
5267 	}
5268 
5269 	case DIOCSETIFFLAG: {
5270 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5271 
5272 		PF_RULES_WLOCK();
5273 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
5274 		PF_RULES_WUNLOCK();
5275 		break;
5276 	}
5277 
5278 	case DIOCCLRIFFLAG: {
5279 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5280 
5281 		PF_RULES_WLOCK();
5282 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
5283 		PF_RULES_WUNLOCK();
5284 		break;
5285 	}
5286 
5287 	default:
5288 		error = ENODEV;
5289 		break;
5290 	}
5291 fail:
5292 	if (sx_xlocked(&pf_ioctl_lock))
5293 		sx_xunlock(&pf_ioctl_lock);
5294 	CURVNET_RESTORE();
5295 
5296 #undef ERROUT_IOCTL
5297 
5298 	return (error);
5299 }
5300 
5301 void
5302 pfsync_state_export(struct pfsync_state *sp, struct pf_kstate *st)
5303 {
5304 	bzero(sp, sizeof(struct pfsync_state));
5305 
5306 	/* copy from state key */
5307 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5308 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5309 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5310 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5311 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5312 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5313 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5314 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5315 	sp->proto = st->key[PF_SK_WIRE]->proto;
5316 	sp->af = st->key[PF_SK_WIRE]->af;
5317 
5318 	/* copy from state */
5319 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
5320 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
5321 	sp->creation = htonl(time_uptime - st->creation);
5322 	sp->expire = pf_state_expires(st);
5323 	if (sp->expire <= time_uptime)
5324 		sp->expire = htonl(0);
5325 	else
5326 		sp->expire = htonl(sp->expire - time_uptime);
5327 
5328 	sp->direction = st->direction;
5329 	sp->log = st->log;
5330 	sp->timeout = st->timeout;
5331 	sp->state_flags = st->state_flags;
5332 	if (st->src_node)
5333 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
5334 	if (st->nat_src_node)
5335 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5336 
5337 	sp->id = st->id;
5338 	sp->creatorid = st->creatorid;
5339 	pf_state_peer_hton(&st->src, &sp->src);
5340 	pf_state_peer_hton(&st->dst, &sp->dst);
5341 
5342 	if (st->rule.ptr == NULL)
5343 		sp->rule = htonl(-1);
5344 	else
5345 		sp->rule = htonl(st->rule.ptr->nr);
5346 	if (st->anchor.ptr == NULL)
5347 		sp->anchor = htonl(-1);
5348 	else
5349 		sp->anchor = htonl(st->anchor.ptr->nr);
5350 	if (st->nat_rule.ptr == NULL)
5351 		sp->nat_rule = htonl(-1);
5352 	else
5353 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
5354 
5355 	pf_state_counter_hton(st->packets[0], sp->packets[0]);
5356 	pf_state_counter_hton(st->packets[1], sp->packets[1]);
5357 	pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
5358 	pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
5359 }
5360 
5361 void
5362 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st)
5363 {
5364 	bzero(sp, sizeof(*sp));
5365 
5366 	sp->version = PF_STATE_VERSION;
5367 
5368 	/* copy from state key */
5369 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5370 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5371 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5372 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5373 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5374 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5375 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5376 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5377 	sp->proto = st->key[PF_SK_WIRE]->proto;
5378 	sp->af = st->key[PF_SK_WIRE]->af;
5379 
5380 	/* copy from state */
5381 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
5382 	strlcpy(sp->orig_ifname, st->orig_kif->pfik_name,
5383 	    sizeof(sp->orig_ifname));
5384 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
5385 	sp->creation = htonl(time_uptime - st->creation);
5386 	sp->expire = pf_state_expires(st);
5387 	if (sp->expire <= time_uptime)
5388 		sp->expire = htonl(0);
5389 	else
5390 		sp->expire = htonl(sp->expire - time_uptime);
5391 
5392 	sp->direction = st->direction;
5393 	sp->log = st->log;
5394 	sp->timeout = st->timeout;
5395 	sp->state_flags = st->state_flags;
5396 	if (st->src_node)
5397 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
5398 	if (st->nat_src_node)
5399 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5400 
5401 	sp->id = st->id;
5402 	sp->creatorid = st->creatorid;
5403 	pf_state_peer_hton(&st->src, &sp->src);
5404 	pf_state_peer_hton(&st->dst, &sp->dst);
5405 
5406 	if (st->rule.ptr == NULL)
5407 		sp->rule = htonl(-1);
5408 	else
5409 		sp->rule = htonl(st->rule.ptr->nr);
5410 	if (st->anchor.ptr == NULL)
5411 		sp->anchor = htonl(-1);
5412 	else
5413 		sp->anchor = htonl(st->anchor.ptr->nr);
5414 	if (st->nat_rule.ptr == NULL)
5415 		sp->nat_rule = htonl(-1);
5416 	else
5417 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
5418 
5419 	sp->packets[0] = st->packets[0];
5420 	sp->packets[1] = st->packets[1];
5421 	sp->bytes[0] = st->bytes[0];
5422 	sp->bytes[1] = st->bytes[1];
5423 }
5424 
5425 static void
5426 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
5427 {
5428 	struct pfr_ktable *kt;
5429 
5430 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
5431 
5432 	kt = aw->p.tbl;
5433 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
5434 		kt = kt->pfrkt_root;
5435 	aw->p.tbl = NULL;
5436 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
5437 		kt->pfrkt_cnt : -1;
5438 }
5439 
5440 static int
5441 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters,
5442     size_t number, char **names)
5443 {
5444 	nvlist_t        *nvc;
5445 
5446 	nvc = nvlist_create(0);
5447 	if (nvc == NULL)
5448 		return (ENOMEM);
5449 
5450 	for (int i = 0; i < number; i++) {
5451 		nvlist_append_number_array(nvc, "counters",
5452 		    counter_u64_fetch(counters[i]));
5453 		nvlist_append_string_array(nvc, "names",
5454 		    names[i]);
5455 		nvlist_append_number_array(nvc, "ids",
5456 		    i);
5457 	}
5458 	nvlist_add_nvlist(nvl, name, nvc);
5459 	nvlist_destroy(nvc);
5460 
5461 	return (0);
5462 }
5463 
5464 static int
5465 pf_getstatus(struct pfioc_nv *nv)
5466 {
5467 	nvlist_t        *nvl = NULL, *nvc = NULL;
5468 	void            *nvlpacked = NULL;
5469 	int              error;
5470 	struct pf_status s;
5471 	char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES;
5472 	char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES;
5473 	char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES;
5474 	PF_RULES_RLOCK_TRACKER;
5475 
5476 #define ERROUT(x)      ERROUT_FUNCTION(errout, x)
5477 
5478 	PF_RULES_RLOCK();
5479 
5480 	nvl = nvlist_create(0);
5481 	if (nvl == NULL)
5482 		ERROUT(ENOMEM);
5483 
5484 	nvlist_add_bool(nvl, "running", V_pf_status.running);
5485 	nvlist_add_number(nvl, "since", V_pf_status.since);
5486 	nvlist_add_number(nvl, "debug", V_pf_status.debug);
5487 	nvlist_add_number(nvl, "hostid", V_pf_status.hostid);
5488 	nvlist_add_number(nvl, "states", V_pf_status.states);
5489 	nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes);
5490 
5491 	/* counters */
5492 	error = pf_add_status_counters(nvl, "counters", V_pf_status.counters,
5493 	    PFRES_MAX, pf_reasons);
5494 	if (error != 0)
5495 		ERROUT(error);
5496 
5497 	/* lcounters */
5498 	error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters,
5499 	    KLCNT_MAX, pf_lcounter);
5500 	if (error != 0)
5501 		ERROUT(error);
5502 
5503 	/* fcounters */
5504 	nvc = nvlist_create(0);
5505 	if (nvc == NULL)
5506 		ERROUT(ENOMEM);
5507 
5508 	for (int i = 0; i < FCNT_MAX; i++) {
5509 		nvlist_append_number_array(nvc, "counters",
5510 		    pf_counter_u64_fetch(&V_pf_status.fcounters[i]));
5511 		nvlist_append_string_array(nvc, "names",
5512 		    pf_fcounter[i]);
5513 		nvlist_append_number_array(nvc, "ids",
5514 		    i);
5515 	}
5516 	nvlist_add_nvlist(nvl, "fcounters", nvc);
5517 	nvlist_destroy(nvc);
5518 	nvc = NULL;
5519 
5520 	/* scounters */
5521 	error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters,
5522 	    SCNT_MAX, pf_fcounter);
5523 	if (error != 0)
5524 		ERROUT(error);
5525 
5526 	nvlist_add_string(nvl, "ifname", V_pf_status.ifname);
5527 	nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum,
5528 	    PF_MD5_DIGEST_LENGTH);
5529 
5530 	pfi_update_status(V_pf_status.ifname, &s);
5531 
5532 	/* pcounters / bcounters */
5533 	for (int i = 0; i < 2; i++) {
5534 		for (int j = 0; j < 2; j++) {
5535 			for (int k = 0; k < 2; k++) {
5536 				nvlist_append_number_array(nvl, "pcounters",
5537 				    s.pcounters[i][j][k]);
5538 			}
5539 			nvlist_append_number_array(nvl, "bcounters",
5540 			    s.bcounters[i][j]);
5541 		}
5542 	}
5543 
5544 	nvlpacked = nvlist_pack(nvl, &nv->len);
5545 	if (nvlpacked == NULL)
5546 		ERROUT(ENOMEM);
5547 
5548 	if (nv->size == 0)
5549 		ERROUT(0);
5550 	else if (nv->size < nv->len)
5551 		ERROUT(ENOSPC);
5552 
5553 	PF_RULES_RUNLOCK();
5554 	error = copyout(nvlpacked, nv->data, nv->len);
5555 	goto done;
5556 
5557 #undef ERROUT
5558 errout:
5559 	PF_RULES_RUNLOCK();
5560 done:
5561 	free(nvlpacked, M_NVLIST);
5562 	nvlist_destroy(nvc);
5563 	nvlist_destroy(nvl);
5564 
5565 	return (error);
5566 }
5567 
5568 /*
5569  * XXX - Check for version missmatch!!!
5570  */
5571 static void
5572 pf_clear_all_states(void)
5573 {
5574 	struct pf_kstate	*s;
5575 	u_int i;
5576 
5577 	for (i = 0; i <= pf_hashmask; i++) {
5578 		struct pf_idhash *ih = &V_pf_idhash[i];
5579 relock:
5580 		PF_HASHROW_LOCK(ih);
5581 		LIST_FOREACH(s, &ih->states, entry) {
5582 			s->timeout = PFTM_PURGE;
5583 			/* Don't send out individual delete messages. */
5584 			s->state_flags |= PFSTATE_NOSYNC;
5585 			pf_unlink_state(s);
5586 			goto relock;
5587 		}
5588 		PF_HASHROW_UNLOCK(ih);
5589 	}
5590 }
5591 
5592 static int
5593 pf_clear_tables(void)
5594 {
5595 	struct pfioc_table io;
5596 	int error;
5597 
5598 	bzero(&io, sizeof(io));
5599 
5600 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
5601 	    io.pfrio_flags);
5602 
5603 	return (error);
5604 }
5605 
5606 static void
5607 pf_clear_srcnodes(struct pf_ksrc_node *n)
5608 {
5609 	struct pf_kstate *s;
5610 	int i;
5611 
5612 	for (i = 0; i <= pf_hashmask; i++) {
5613 		struct pf_idhash *ih = &V_pf_idhash[i];
5614 
5615 		PF_HASHROW_LOCK(ih);
5616 		LIST_FOREACH(s, &ih->states, entry) {
5617 			if (n == NULL || n == s->src_node)
5618 				s->src_node = NULL;
5619 			if (n == NULL || n == s->nat_src_node)
5620 				s->nat_src_node = NULL;
5621 		}
5622 		PF_HASHROW_UNLOCK(ih);
5623 	}
5624 
5625 	if (n == NULL) {
5626 		struct pf_srchash *sh;
5627 
5628 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5629 		    i++, sh++) {
5630 			PF_HASHROW_LOCK(sh);
5631 			LIST_FOREACH(n, &sh->nodes, entry) {
5632 				n->expire = 1;
5633 				n->states = 0;
5634 			}
5635 			PF_HASHROW_UNLOCK(sh);
5636 		}
5637 	} else {
5638 		/* XXX: hash slot should already be locked here. */
5639 		n->expire = 1;
5640 		n->states = 0;
5641 	}
5642 }
5643 
5644 static void
5645 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
5646 {
5647 	struct pf_ksrc_node_list	 kill;
5648 
5649 	LIST_INIT(&kill);
5650 	for (int i = 0; i <= pf_srchashmask; i++) {
5651 		struct pf_srchash *sh = &V_pf_srchash[i];
5652 		struct pf_ksrc_node *sn, *tmp;
5653 
5654 		PF_HASHROW_LOCK(sh);
5655 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
5656 			if (PF_MATCHA(psnk->psnk_src.neg,
5657 			      &psnk->psnk_src.addr.v.a.addr,
5658 			      &psnk->psnk_src.addr.v.a.mask,
5659 			      &sn->addr, sn->af) &&
5660 			    PF_MATCHA(psnk->psnk_dst.neg,
5661 			      &psnk->psnk_dst.addr.v.a.addr,
5662 			      &psnk->psnk_dst.addr.v.a.mask,
5663 			      &sn->raddr, sn->af)) {
5664 				pf_unlink_src_node(sn);
5665 				LIST_INSERT_HEAD(&kill, sn, entry);
5666 				sn->expire = 1;
5667 			}
5668 		PF_HASHROW_UNLOCK(sh);
5669 	}
5670 
5671 	for (int i = 0; i <= pf_hashmask; i++) {
5672 		struct pf_idhash *ih = &V_pf_idhash[i];
5673 		struct pf_kstate *s;
5674 
5675 		PF_HASHROW_LOCK(ih);
5676 		LIST_FOREACH(s, &ih->states, entry) {
5677 			if (s->src_node && s->src_node->expire == 1)
5678 				s->src_node = NULL;
5679 			if (s->nat_src_node && s->nat_src_node->expire == 1)
5680 				s->nat_src_node = NULL;
5681 		}
5682 		PF_HASHROW_UNLOCK(ih);
5683 	}
5684 
5685 	psnk->psnk_killed = pf_free_src_nodes(&kill);
5686 }
5687 
5688 static int
5689 pf_keepcounters(struct pfioc_nv *nv)
5690 {
5691 	nvlist_t	*nvl = NULL;
5692 	void		*nvlpacked = NULL;
5693 	int		 error = 0;
5694 
5695 #define	ERROUT(x)	ERROUT_FUNCTION(on_error, x)
5696 
5697 	if (nv->len > pf_ioctl_maxcount)
5698 		ERROUT(ENOMEM);
5699 
5700 	nvlpacked = malloc(nv->len, M_TEMP, M_WAITOK);
5701 	if (nvlpacked == NULL)
5702 		ERROUT(ENOMEM);
5703 
5704 	error = copyin(nv->data, nvlpacked, nv->len);
5705 	if (error)
5706 		ERROUT(error);
5707 
5708 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
5709 	if (nvl == NULL)
5710 		ERROUT(EBADMSG);
5711 
5712 	if (! nvlist_exists_bool(nvl, "keep_counters"))
5713 		ERROUT(EBADMSG);
5714 
5715 	V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters");
5716 
5717 on_error:
5718 	nvlist_destroy(nvl);
5719 	free(nvlpacked, M_TEMP);
5720 	return (error);
5721 }
5722 
5723 static unsigned int
5724 pf_clear_states(const struct pf_kstate_kill *kill)
5725 {
5726 	struct pf_state_key_cmp	 match_key;
5727 	struct pf_kstate	*s;
5728 	struct pfi_kkif	*kif;
5729 	int		 idx;
5730 	unsigned int	 killed = 0, dir;
5731 
5732 	for (unsigned int i = 0; i <= pf_hashmask; i++) {
5733 		struct pf_idhash *ih = &V_pf_idhash[i];
5734 
5735 relock_DIOCCLRSTATES:
5736 		PF_HASHROW_LOCK(ih);
5737 		LIST_FOREACH(s, &ih->states, entry) {
5738 			/* For floating states look at the original kif. */
5739 			kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
5740 
5741 			if (kill->psk_ifname[0] &&
5742 			    strcmp(kill->psk_ifname,
5743 			    kif->pfik_name))
5744 				continue;
5745 
5746 			if (kill->psk_kill_match) {
5747 				bzero(&match_key, sizeof(match_key));
5748 
5749 				if (s->direction == PF_OUT) {
5750 					dir = PF_IN;
5751 					idx = PF_SK_STACK;
5752 				} else {
5753 					dir = PF_OUT;
5754 					idx = PF_SK_WIRE;
5755 				}
5756 
5757 				match_key.af = s->key[idx]->af;
5758 				match_key.proto = s->key[idx]->proto;
5759 				PF_ACPY(&match_key.addr[0],
5760 				    &s->key[idx]->addr[1], match_key.af);
5761 				match_key.port[0] = s->key[idx]->port[1];
5762 				PF_ACPY(&match_key.addr[1],
5763 				    &s->key[idx]->addr[0], match_key.af);
5764 				match_key.port[1] = s->key[idx]->port[0];
5765 			}
5766 
5767 			/*
5768 			 * Don't send out individual
5769 			 * delete messages.
5770 			 */
5771 			s->state_flags |= PFSTATE_NOSYNC;
5772 			pf_unlink_state(s);
5773 			killed++;
5774 
5775 			if (kill->psk_kill_match)
5776 				killed += pf_kill_matching_state(&match_key,
5777 				    dir);
5778 
5779 			goto relock_DIOCCLRSTATES;
5780 		}
5781 		PF_HASHROW_UNLOCK(ih);
5782 	}
5783 
5784 	if (V_pfsync_clear_states_ptr != NULL)
5785 		V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname);
5786 
5787 	return (killed);
5788 }
5789 
5790 static void
5791 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed)
5792 {
5793 	struct pf_kstate	*s;
5794 
5795 	if (kill->psk_pfcmp.id) {
5796 		if (kill->psk_pfcmp.creatorid == 0)
5797 			kill->psk_pfcmp.creatorid = V_pf_status.hostid;
5798 		if ((s = pf_find_state_byid(kill->psk_pfcmp.id,
5799 		    kill->psk_pfcmp.creatorid))) {
5800 			pf_unlink_state(s);
5801 			*killed = 1;
5802 		}
5803 		return;
5804 	}
5805 
5806 	for (unsigned int i = 0; i <= pf_hashmask; i++)
5807 		*killed += pf_killstates_row(kill, &V_pf_idhash[i]);
5808 
5809 	return;
5810 }
5811 
5812 static int
5813 pf_killstates_nv(struct pfioc_nv *nv)
5814 {
5815 	struct pf_kstate_kill	 kill;
5816 	nvlist_t		*nvl = NULL;
5817 	void			*nvlpacked = NULL;
5818 	int			 error = 0;
5819 	unsigned int		 killed = 0;
5820 
5821 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
5822 
5823 	if (nv->len > pf_ioctl_maxcount)
5824 		ERROUT(ENOMEM);
5825 
5826 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
5827 	if (nvlpacked == NULL)
5828 		ERROUT(ENOMEM);
5829 
5830 	error = copyin(nv->data, nvlpacked, nv->len);
5831 	if (error)
5832 		ERROUT(error);
5833 
5834 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
5835 	if (nvl == NULL)
5836 		ERROUT(EBADMSG);
5837 
5838 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
5839 	if (error)
5840 		ERROUT(error);
5841 
5842 	pf_killstates(&kill, &killed);
5843 
5844 	free(nvlpacked, M_NVLIST);
5845 	nvlpacked = NULL;
5846 	nvlist_destroy(nvl);
5847 	nvl = nvlist_create(0);
5848 	if (nvl == NULL)
5849 		ERROUT(ENOMEM);
5850 
5851 	nvlist_add_number(nvl, "killed", killed);
5852 
5853 	nvlpacked = nvlist_pack(nvl, &nv->len);
5854 	if (nvlpacked == NULL)
5855 		ERROUT(ENOMEM);
5856 
5857 	if (nv->size == 0)
5858 		ERROUT(0);
5859 	else if (nv->size < nv->len)
5860 		ERROUT(ENOSPC);
5861 
5862 	error = copyout(nvlpacked, nv->data, nv->len);
5863 
5864 on_error:
5865 	nvlist_destroy(nvl);
5866 	free(nvlpacked, M_NVLIST);
5867 	return (error);
5868 }
5869 
5870 static int
5871 pf_clearstates_nv(struct pfioc_nv *nv)
5872 {
5873 	struct pf_kstate_kill	 kill;
5874 	nvlist_t		*nvl = NULL;
5875 	void			*nvlpacked = NULL;
5876 	int			 error = 0;
5877 	unsigned int		 killed;
5878 
5879 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
5880 
5881 	if (nv->len > pf_ioctl_maxcount)
5882 		ERROUT(ENOMEM);
5883 
5884 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
5885 	if (nvlpacked == NULL)
5886 		ERROUT(ENOMEM);
5887 
5888 	error = copyin(nv->data, nvlpacked, nv->len);
5889 	if (error)
5890 		ERROUT(error);
5891 
5892 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
5893 	if (nvl == NULL)
5894 		ERROUT(EBADMSG);
5895 
5896 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
5897 	if (error)
5898 		ERROUT(error);
5899 
5900 	killed = pf_clear_states(&kill);
5901 
5902 	free(nvlpacked, M_NVLIST);
5903 	nvlpacked = NULL;
5904 	nvlist_destroy(nvl);
5905 	nvl = nvlist_create(0);
5906 	if (nvl == NULL)
5907 		ERROUT(ENOMEM);
5908 
5909 	nvlist_add_number(nvl, "killed", killed);
5910 
5911 	nvlpacked = nvlist_pack(nvl, &nv->len);
5912 	if (nvlpacked == NULL)
5913 		ERROUT(ENOMEM);
5914 
5915 	if (nv->size == 0)
5916 		ERROUT(0);
5917 	else if (nv->size < nv->len)
5918 		ERROUT(ENOSPC);
5919 
5920 	error = copyout(nvlpacked, nv->data, nv->len);
5921 
5922 #undef ERROUT
5923 on_error:
5924 	nvlist_destroy(nvl);
5925 	free(nvlpacked, M_NVLIST);
5926 	return (error);
5927 }
5928 
5929 static int
5930 pf_getstate(struct pfioc_nv *nv)
5931 {
5932 	nvlist_t		*nvl = NULL, *nvls;
5933 	void			*nvlpacked = NULL;
5934 	struct pf_kstate	*s = NULL;
5935 	int			 error = 0;
5936 	uint64_t		 id, creatorid;
5937 
5938 #define ERROUT(x)	ERROUT_FUNCTION(errout, x)
5939 
5940 	if (nv->len > pf_ioctl_maxcount)
5941 		ERROUT(ENOMEM);
5942 
5943 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
5944 	if (nvlpacked == NULL)
5945 		ERROUT(ENOMEM);
5946 
5947 	error = copyin(nv->data, nvlpacked, nv->len);
5948 	if (error)
5949 		ERROUT(error);
5950 
5951 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
5952 	if (nvl == NULL)
5953 		ERROUT(EBADMSG);
5954 
5955 	PFNV_CHK(pf_nvuint64(nvl, "id", &id));
5956 	PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid));
5957 
5958 	s = pf_find_state_byid(id, creatorid);
5959 	if (s == NULL)
5960 		ERROUT(ENOENT);
5961 
5962 	free(nvlpacked, M_NVLIST);
5963 	nvlpacked = NULL;
5964 	nvlist_destroy(nvl);
5965 	nvl = nvlist_create(0);
5966 	if (nvl == NULL)
5967 		ERROUT(ENOMEM);
5968 
5969 	nvls = pf_state_to_nvstate(s);
5970 	if (nvls == NULL)
5971 		ERROUT(ENOMEM);
5972 
5973 	nvlist_add_nvlist(nvl, "state", nvls);
5974 	nvlist_destroy(nvls);
5975 
5976 	nvlpacked = nvlist_pack(nvl, &nv->len);
5977 	if (nvlpacked == NULL)
5978 		ERROUT(ENOMEM);
5979 
5980 	if (nv->size == 0)
5981 		ERROUT(0);
5982 	else if (nv->size < nv->len)
5983 		ERROUT(ENOSPC);
5984 
5985 	error = copyout(nvlpacked, nv->data, nv->len);
5986 
5987 #undef ERROUT
5988 errout:
5989 	if (s != NULL)
5990 		PF_STATE_UNLOCK(s);
5991 	free(nvlpacked, M_NVLIST);
5992 	nvlist_destroy(nvl);
5993 	return (error);
5994 }
5995 
5996 /*
5997  * XXX - Check for version missmatch!!!
5998  */
5999 
6000 /*
6001  * Duplicate pfctl -Fa operation to get rid of as much as we can.
6002  */
6003 static int
6004 shutdown_pf(void)
6005 {
6006 	int error = 0;
6007 	u_int32_t t[5];
6008 	char nn = '\0';
6009 
6010 	do {
6011 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
6012 		    != 0) {
6013 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
6014 			break;
6015 		}
6016 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
6017 		    != 0) {
6018 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
6019 			break;		/* XXX: rollback? */
6020 		}
6021 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
6022 		    != 0) {
6023 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
6024 			break;		/* XXX: rollback? */
6025 		}
6026 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
6027 		    != 0) {
6028 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
6029 			break;		/* XXX: rollback? */
6030 		}
6031 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
6032 		    != 0) {
6033 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
6034 			break;		/* XXX: rollback? */
6035 		}
6036 
6037 		/* XXX: these should always succeed here */
6038 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
6039 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
6040 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
6041 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
6042 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
6043 
6044 		if ((error = pf_clear_tables()) != 0)
6045 			break;
6046 
6047 		if ((error = pf_begin_eth(&t[0], &nn)) != 0) {
6048 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth\n"));
6049 			break;
6050 		}
6051 		pf_commit_eth(t[0], &nn);
6052 
6053 #ifdef ALTQ
6054 		if ((error = pf_begin_altq(&t[0])) != 0) {
6055 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
6056 			break;
6057 		}
6058 		pf_commit_altq(t[0]);
6059 #endif
6060 
6061 		pf_clear_all_states();
6062 
6063 		pf_clear_srcnodes(NULL);
6064 
6065 		/* status does not use malloced mem so no need to cleanup */
6066 		/* fingerprints and interfaces have their own cleanup code */
6067 	} while(0);
6068 
6069 	return (error);
6070 }
6071 
6072 static pfil_return_t
6073 pf_check_return(int chk, struct mbuf **m)
6074 {
6075 
6076 	switch (chk) {
6077 	case PF_PASS:
6078 		if (*m == NULL)
6079 			return (PFIL_CONSUMED);
6080 		else
6081 			return (PFIL_PASS);
6082 		break;
6083 	default:
6084 		if (*m != NULL) {
6085 			m_freem(*m);
6086 			*m = NULL;
6087 		}
6088 		return (PFIL_DROPPED);
6089 	}
6090 }
6091 
6092 static pfil_return_t
6093 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6094     void *ruleset __unused, struct inpcb *inp)
6095 {
6096 	int chk;
6097 
6098 	chk = pf_test_eth(PF_IN, flags, ifp, m, inp);
6099 
6100 	return (pf_check_return(chk, m));
6101 }
6102 
6103 static pfil_return_t
6104 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6105     void *ruleset __unused, struct inpcb *inp)
6106 {
6107 	int chk;
6108 
6109 	chk = pf_test_eth(PF_OUT, flags, ifp, m, inp);
6110 
6111 	return (pf_check_return(chk, m));
6112 }
6113 
6114 #ifdef INET
6115 static pfil_return_t
6116 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6117     void *ruleset __unused, struct inpcb *inp)
6118 {
6119 	int chk;
6120 
6121 	chk = pf_test(PF_IN, flags, ifp, m, inp);
6122 
6123 	return (pf_check_return(chk, m));
6124 }
6125 
6126 static pfil_return_t
6127 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6128     void *ruleset __unused,  struct inpcb *inp)
6129 {
6130 	int chk;
6131 
6132 	chk = pf_test(PF_OUT, flags, ifp, m, inp);
6133 
6134 	return (pf_check_return(chk, m));
6135 }
6136 #endif
6137 
6138 #ifdef INET6
6139 static pfil_return_t
6140 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
6141     void *ruleset __unused,  struct inpcb *inp)
6142 {
6143 	int chk;
6144 
6145 	/*
6146 	 * In case of loopback traffic IPv6 uses the real interface in
6147 	 * order to support scoped addresses. In order to support stateful
6148 	 * filtering we have change this to lo0 as it is the case in IPv4.
6149 	 */
6150 	CURVNET_SET(ifp->if_vnet);
6151 	chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
6152 	CURVNET_RESTORE();
6153 
6154 	return (pf_check_return(chk, m));
6155 }
6156 
6157 static pfil_return_t
6158 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
6159     void *ruleset __unused,  struct inpcb *inp)
6160 {
6161 	int chk;
6162 
6163 	CURVNET_SET(ifp->if_vnet);
6164 	chk = pf_test6(PF_OUT, flags, ifp, m, inp);
6165 	CURVNET_RESTORE();
6166 
6167 	return (pf_check_return(chk, m));
6168 }
6169 #endif /* INET6 */
6170 
6171 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook);
6172 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook);
6173 #define	V_pf_eth_in_hook	VNET(pf_eth_in_hook)
6174 #define	V_pf_eth_out_hook	VNET(pf_eth_out_hook)
6175 
6176 #ifdef INET
6177 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
6178 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
6179 #define	V_pf_ip4_in_hook	VNET(pf_ip4_in_hook)
6180 #define	V_pf_ip4_out_hook	VNET(pf_ip4_out_hook)
6181 #endif
6182 #ifdef INET6
6183 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
6184 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
6185 #define	V_pf_ip6_in_hook	VNET(pf_ip6_in_hook)
6186 #define	V_pf_ip6_out_hook	VNET(pf_ip6_out_hook)
6187 #endif
6188 
6189 static void
6190 hook_pf_eth(void)
6191 {
6192 	struct pfil_hook_args pha;
6193 	struct pfil_link_args pla;
6194 	int ret __diagused;
6195 
6196 	if (V_pf_pfil_eth_hooked)
6197 		return;
6198 
6199 	pha.pa_version = PFIL_VERSION;
6200 	pha.pa_modname = "pf";
6201 	pha.pa_ruleset = NULL;
6202 
6203 	pla.pa_version = PFIL_VERSION;
6204 
6205 	pha.pa_type = PFIL_TYPE_ETHERNET;
6206 	pha.pa_func = pf_eth_check_in;
6207 	pha.pa_flags = PFIL_IN;
6208 	pha.pa_rulname = "eth-in";
6209 	V_pf_eth_in_hook = pfil_add_hook(&pha);
6210 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6211 	pla.pa_head = V_link_pfil_head;
6212 	pla.pa_hook = V_pf_eth_in_hook;
6213 	ret = pfil_link(&pla);
6214 	MPASS(ret == 0);
6215 	pha.pa_func = pf_eth_check_out;
6216 	pha.pa_flags = PFIL_OUT;
6217 	pha.pa_rulname = "eth-out";
6218 	V_pf_eth_out_hook = pfil_add_hook(&pha);
6219 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6220 	pla.pa_head = V_link_pfil_head;
6221 	pla.pa_hook = V_pf_eth_out_hook;
6222 	ret = pfil_link(&pla);
6223 	MPASS(ret == 0);
6224 
6225 	V_pf_pfil_eth_hooked = 1;
6226 }
6227 
6228 static void
6229 hook_pf(void)
6230 {
6231 	struct pfil_hook_args pha;
6232 	struct pfil_link_args pla;
6233 	int ret;
6234 
6235 	if (V_pf_pfil_hooked)
6236 		return;
6237 
6238 	pha.pa_version = PFIL_VERSION;
6239 	pha.pa_modname = "pf";
6240 	pha.pa_ruleset = NULL;
6241 
6242 	pla.pa_version = PFIL_VERSION;
6243 
6244 #ifdef INET
6245 	pha.pa_type = PFIL_TYPE_IP4;
6246 	pha.pa_func = pf_check_in;
6247 	pha.pa_flags = PFIL_IN;
6248 	pha.pa_rulname = "default-in";
6249 	V_pf_ip4_in_hook = pfil_add_hook(&pha);
6250 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6251 	pla.pa_head = V_inet_pfil_head;
6252 	pla.pa_hook = V_pf_ip4_in_hook;
6253 	ret = pfil_link(&pla);
6254 	MPASS(ret == 0);
6255 	pha.pa_func = pf_check_out;
6256 	pha.pa_flags = PFIL_OUT;
6257 	pha.pa_rulname = "default-out";
6258 	V_pf_ip4_out_hook = pfil_add_hook(&pha);
6259 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6260 	pla.pa_head = V_inet_pfil_head;
6261 	pla.pa_hook = V_pf_ip4_out_hook;
6262 	ret = pfil_link(&pla);
6263 	MPASS(ret == 0);
6264 #endif
6265 #ifdef INET6
6266 	pha.pa_type = PFIL_TYPE_IP6;
6267 	pha.pa_func = pf_check6_in;
6268 	pha.pa_flags = PFIL_IN;
6269 	pha.pa_rulname = "default-in6";
6270 	V_pf_ip6_in_hook = pfil_add_hook(&pha);
6271 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6272 	pla.pa_head = V_inet6_pfil_head;
6273 	pla.pa_hook = V_pf_ip6_in_hook;
6274 	ret = pfil_link(&pla);
6275 	MPASS(ret == 0);
6276 	pha.pa_func = pf_check6_out;
6277 	pha.pa_rulname = "default-out6";
6278 	pha.pa_flags = PFIL_OUT;
6279 	V_pf_ip6_out_hook = pfil_add_hook(&pha);
6280 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6281 	pla.pa_head = V_inet6_pfil_head;
6282 	pla.pa_hook = V_pf_ip6_out_hook;
6283 	ret = pfil_link(&pla);
6284 	MPASS(ret == 0);
6285 #endif
6286 
6287 	V_pf_pfil_hooked = 1;
6288 }
6289 
6290 static void
6291 dehook_pf_eth(void)
6292 {
6293 
6294 	if (V_pf_pfil_eth_hooked == 0)
6295 		return;
6296 
6297 	pfil_remove_hook(V_pf_eth_in_hook);
6298 	pfil_remove_hook(V_pf_eth_out_hook);
6299 
6300 	V_pf_pfil_eth_hooked = 0;
6301 }
6302 
6303 static void
6304 dehook_pf(void)
6305 {
6306 
6307 	if (V_pf_pfil_hooked == 0)
6308 		return;
6309 
6310 #ifdef INET
6311 	pfil_remove_hook(V_pf_ip4_in_hook);
6312 	pfil_remove_hook(V_pf_ip4_out_hook);
6313 #endif
6314 #ifdef INET6
6315 	pfil_remove_hook(V_pf_ip6_in_hook);
6316 	pfil_remove_hook(V_pf_ip6_out_hook);
6317 #endif
6318 
6319 	V_pf_pfil_hooked = 0;
6320 }
6321 
6322 static void
6323 pf_load_vnet(void)
6324 {
6325 	V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
6326 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
6327 
6328 	pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
6329 	    PF_RULE_TAG_HASH_SIZE_DEFAULT);
6330 #ifdef ALTQ
6331 	pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
6332 	    PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
6333 #endif
6334 
6335 	V_pf_keth = &V_pf_main_keth_anchor.ruleset;
6336 
6337 	pfattach_vnet();
6338 	V_pf_vnet_active = 1;
6339 }
6340 
6341 static int
6342 pf_load(void)
6343 {
6344 	int error;
6345 
6346 	rm_init_flags(&pf_rules_lock, "pf rulesets", RM_RECURSE);
6347 	sx_init(&pf_ioctl_lock, "pf ioctl");
6348 	sx_init(&pf_end_lock, "pf end thread");
6349 
6350 	pf_mtag_initialize();
6351 
6352 	pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
6353 	if (pf_dev == NULL)
6354 		return (ENOMEM);
6355 
6356 	pf_end_threads = 0;
6357 	error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
6358 	if (error != 0)
6359 		return (error);
6360 
6361 	pfi_initialize();
6362 
6363 	return (0);
6364 }
6365 
6366 static void
6367 pf_unload_vnet(void)
6368 {
6369 	int ret;
6370 
6371 	V_pf_vnet_active = 0;
6372 	V_pf_status.running = 0;
6373 	dehook_pf();
6374 	dehook_pf_eth();
6375 
6376 	PF_RULES_WLOCK();
6377 	pf_syncookies_cleanup();
6378 	shutdown_pf();
6379 	PF_RULES_WUNLOCK();
6380 
6381 	ret = swi_remove(V_pf_swi_cookie);
6382 	MPASS(ret == 0);
6383 	ret = intr_event_destroy(V_pf_swi_ie);
6384 	MPASS(ret == 0);
6385 
6386 	pf_unload_vnet_purge();
6387 
6388 	pf_normalize_cleanup();
6389 	PF_RULES_WLOCK();
6390 	pfi_cleanup_vnet();
6391 	PF_RULES_WUNLOCK();
6392 	pfr_cleanup();
6393 	pf_osfp_flush();
6394 	pf_cleanup();
6395 	if (IS_DEFAULT_VNET(curvnet))
6396 		pf_mtag_cleanup();
6397 
6398 	pf_cleanup_tagset(&V_pf_tags);
6399 #ifdef ALTQ
6400 	pf_cleanup_tagset(&V_pf_qids);
6401 #endif
6402 	uma_zdestroy(V_pf_tag_z);
6403 
6404 #ifdef PF_WANT_32_TO_64_COUNTER
6405 	PF_RULES_WLOCK();
6406 	LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist);
6407 
6408 	MPASS(LIST_EMPTY(&V_pf_allkiflist));
6409 	MPASS(V_pf_allkifcount == 0);
6410 
6411 	LIST_REMOVE(&V_pf_default_rule, allrulelist);
6412 	V_pf_allrulecount--;
6413 	LIST_REMOVE(V_pf_rulemarker, allrulelist);
6414 
6415 	/*
6416 	 * There are known pf rule leaks when running the test suite.
6417 	 */
6418 #ifdef notyet
6419 	MPASS(LIST_EMPTY(&V_pf_allrulelist));
6420 	MPASS(V_pf_allrulecount == 0);
6421 #endif
6422 
6423 	PF_RULES_WUNLOCK();
6424 
6425 	free(V_pf_kifmarker, PFI_MTYPE);
6426 	free(V_pf_rulemarker, M_PFRULE);
6427 #endif
6428 
6429 	/* Free counters last as we updated them during shutdown. */
6430 	pf_counter_u64_deinit(&V_pf_default_rule.evaluations);
6431 	for (int i = 0; i < 2; i++) {
6432 		pf_counter_u64_deinit(&V_pf_default_rule.packets[i]);
6433 		pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]);
6434 	}
6435 	counter_u64_free(V_pf_default_rule.states_cur);
6436 	counter_u64_free(V_pf_default_rule.states_tot);
6437 	counter_u64_free(V_pf_default_rule.src_nodes);
6438 
6439 	for (int i = 0; i < PFRES_MAX; i++)
6440 		counter_u64_free(V_pf_status.counters[i]);
6441 	for (int i = 0; i < KLCNT_MAX; i++)
6442 		counter_u64_free(V_pf_status.lcounters[i]);
6443 	for (int i = 0; i < FCNT_MAX; i++)
6444 		pf_counter_u64_deinit(&V_pf_status.fcounters[i]);
6445 	for (int i = 0; i < SCNT_MAX; i++)
6446 		counter_u64_free(V_pf_status.scounters[i]);
6447 }
6448 
6449 static void
6450 pf_unload(void)
6451 {
6452 
6453 	sx_xlock(&pf_end_lock);
6454 	pf_end_threads = 1;
6455 	while (pf_end_threads < 2) {
6456 		wakeup_one(pf_purge_thread);
6457 		sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
6458 	}
6459 	sx_xunlock(&pf_end_lock);
6460 
6461 	if (pf_dev != NULL)
6462 		destroy_dev(pf_dev);
6463 
6464 	pfi_cleanup();
6465 
6466 	rm_destroy(&pf_rules_lock);
6467 	sx_destroy(&pf_ioctl_lock);
6468 	sx_destroy(&pf_end_lock);
6469 }
6470 
6471 static void
6472 vnet_pf_init(void *unused __unused)
6473 {
6474 
6475 	pf_load_vnet();
6476 }
6477 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6478     vnet_pf_init, NULL);
6479 
6480 static void
6481 vnet_pf_uninit(const void *unused __unused)
6482 {
6483 
6484 	pf_unload_vnet();
6485 }
6486 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
6487 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6488     vnet_pf_uninit, NULL);
6489 
6490 static int
6491 pf_modevent(module_t mod, int type, void *data)
6492 {
6493 	int error = 0;
6494 
6495 	switch(type) {
6496 	case MOD_LOAD:
6497 		error = pf_load();
6498 		break;
6499 	case MOD_UNLOAD:
6500 		/* Handled in SYSUNINIT(pf_unload) to ensure it's done after
6501 		 * the vnet_pf_uninit()s */
6502 		break;
6503 	default:
6504 		error = EINVAL;
6505 		break;
6506 	}
6507 
6508 	return (error);
6509 }
6510 
6511 static moduledata_t pf_mod = {
6512 	"pf",
6513 	pf_modevent,
6514 	0
6515 };
6516 
6517 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
6518 MODULE_VERSION(pf, PF_MODVER);
6519