xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision 45dd2eaac379e5576f745380260470204c49beac)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2001 Daniel Hartmeier
5  * Copyright (c) 2002,2003 Henning Brauer
6  * Copyright (c) 2012 Gleb Smirnoff <glebius@FreeBSD.org>
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  *
13  *    - Redistributions of source code must retain the above copyright
14  *      notice, this list of conditions and the following disclaimer.
15  *    - Redistributions in binary form must reproduce the above
16  *      copyright notice, this list of conditions and the following
17  *      disclaimer in the documentation and/or other materials provided
18  *      with the distribution.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
26  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
28  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
30  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
31  * POSSIBILITY OF SUCH DAMAGE.
32  *
33  * Effort sponsored in part by the Defense Advanced Research Projects
34  * Agency (DARPA) and Air Force Research Laboratory, Air Force
35  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
36  *
37  *	$OpenBSD: pf_ioctl.c,v 1.213 2009/02/15 21:46:12 mbalmer Exp $
38  */
39 
40 #include <sys/cdefs.h>
41 __FBSDID("$FreeBSD$");
42 
43 #include "opt_inet.h"
44 #include "opt_inet6.h"
45 #include "opt_bpf.h"
46 #include "opt_pf.h"
47 
48 #include <sys/param.h>
49 #include <sys/_bitset.h>
50 #include <sys/bitset.h>
51 #include <sys/bus.h>
52 #include <sys/conf.h>
53 #include <sys/endian.h>
54 #include <sys/fcntl.h>
55 #include <sys/filio.h>
56 #include <sys/hash.h>
57 #include <sys/interrupt.h>
58 #include <sys/jail.h>
59 #include <sys/kernel.h>
60 #include <sys/kthread.h>
61 #include <sys/lock.h>
62 #include <sys/mbuf.h>
63 #include <sys/module.h>
64 #include <sys/nv.h>
65 #include <sys/proc.h>
66 #include <sys/sdt.h>
67 #include <sys/smp.h>
68 #include <sys/socket.h>
69 #include <sys/sysctl.h>
70 #include <sys/md5.h>
71 #include <sys/ucred.h>
72 
73 #include <net/if.h>
74 #include <net/if_var.h>
75 #include <net/vnet.h>
76 #include <net/route.h>
77 #include <net/pfil.h>
78 #include <net/pfvar.h>
79 #include <net/if_pfsync.h>
80 #include <net/if_pflog.h>
81 
82 #include <netinet/in.h>
83 #include <netinet/ip.h>
84 #include <netinet/ip_var.h>
85 #include <netinet6/ip6_var.h>
86 #include <netinet/ip_icmp.h>
87 #include <netpfil/pf/pf_nv.h>
88 
89 #ifdef INET6
90 #include <netinet/ip6.h>
91 #endif /* INET6 */
92 
93 #ifdef ALTQ
94 #include <net/altq/altq.h>
95 #endif
96 
97 SDT_PROBE_DEFINE3(pf, ioctl, ioctl, error, "int", "int", "int");
98 SDT_PROBE_DEFINE3(pf, ioctl, function, error, "char *", "int", "int");
99 SDT_PROBE_DEFINE2(pf, ioctl, addrule, error, "int", "int");
100 SDT_PROBE_DEFINE2(pf, ioctl, nvchk, error, "int", "int");
101 
102 static struct pf_kpool	*pf_get_kpool(const char *, u_int32_t, u_int8_t,
103 			    u_int32_t, u_int8_t, u_int8_t, u_int8_t);
104 
105 static void		 pf_mv_kpool(struct pf_kpalist *, struct pf_kpalist *);
106 static void		 pf_empty_kpool(struct pf_kpalist *);
107 static int		 pfioctl(struct cdev *, u_long, caddr_t, int,
108 			    struct thread *);
109 static int		 pf_begin_eth(uint32_t *, const char *);
110 static void		 pf_rollback_eth_cb(struct epoch_context *);
111 static int		 pf_rollback_eth(uint32_t, const char *);
112 static int		 pf_commit_eth(uint32_t, const char *);
113 static void		 pf_free_eth_rule(struct pf_keth_rule *);
114 #ifdef ALTQ
115 static int		 pf_begin_altq(u_int32_t *);
116 static int		 pf_rollback_altq(u_int32_t);
117 static int		 pf_commit_altq(u_int32_t);
118 static int		 pf_enable_altq(struct pf_altq *);
119 static int		 pf_disable_altq(struct pf_altq *);
120 static uint16_t		 pf_qname2qid(const char *);
121 static void		 pf_qid_unref(uint16_t);
122 #endif /* ALTQ */
123 static int		 pf_begin_rules(u_int32_t *, int, const char *);
124 static int		 pf_rollback_rules(u_int32_t, int, char *);
125 static int		 pf_setup_pfsync_matching(struct pf_kruleset *);
126 static void		 pf_hash_rule_rolling(MD5_CTX *, struct pf_krule *);
127 static void		 pf_hash_rule(struct pf_krule *);
128 static void		 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
129 static int		 pf_commit_rules(u_int32_t, int, char *);
130 static int		 pf_addr_setup(struct pf_kruleset *,
131 			    struct pf_addr_wrap *, sa_family_t);
132 static void		 pf_addr_copyout(struct pf_addr_wrap *);
133 static void		 pf_src_node_copy(const struct pf_ksrc_node *,
134 			    struct pf_src_node *);
135 #ifdef ALTQ
136 static int		 pf_export_kaltq(struct pf_altq *,
137 			    struct pfioc_altq_v1 *, size_t);
138 static int		 pf_import_kaltq(struct pfioc_altq_v1 *,
139 			    struct pf_altq *, size_t);
140 #endif /* ALTQ */
141 
142 VNET_DEFINE(struct pf_krule,	pf_default_rule);
143 
144 static __inline int             pf_krule_compare(struct pf_krule *,
145 				    struct pf_krule *);
146 
147 RB_GENERATE(pf_krule_global, pf_krule, entry_global, pf_krule_compare);
148 
149 #ifdef ALTQ
150 VNET_DEFINE_STATIC(int,		pf_altq_running);
151 #define	V_pf_altq_running	VNET(pf_altq_running)
152 #endif
153 
154 #define	TAGID_MAX	 50000
155 struct pf_tagname {
156 	TAILQ_ENTRY(pf_tagname)	namehash_entries;
157 	TAILQ_ENTRY(pf_tagname)	taghash_entries;
158 	char			name[PF_TAG_NAME_SIZE];
159 	uint16_t		tag;
160 	int			ref;
161 };
162 
163 struct pf_tagset {
164 	TAILQ_HEAD(, pf_tagname)	*namehash;
165 	TAILQ_HEAD(, pf_tagname)	*taghash;
166 	unsigned int			 mask;
167 	uint32_t			 seed;
168 	BITSET_DEFINE(, TAGID_MAX)	 avail;
169 };
170 
171 VNET_DEFINE(struct pf_tagset, pf_tags);
172 #define	V_pf_tags	VNET(pf_tags)
173 static unsigned int	pf_rule_tag_hashsize;
174 #define	PF_RULE_TAG_HASH_SIZE_DEFAULT	128
175 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN,
176     &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT,
177     "Size of pf(4) rule tag hashtable");
178 
179 #ifdef ALTQ
180 VNET_DEFINE(struct pf_tagset, pf_qids);
181 #define	V_pf_qids	VNET(pf_qids)
182 static unsigned int	pf_queue_tag_hashsize;
183 #define	PF_QUEUE_TAG_HASH_SIZE_DEFAULT	128
184 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN,
185     &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT,
186     "Size of pf(4) queue tag hashtable");
187 #endif
188 VNET_DEFINE(uma_zone_t,	 pf_tag_z);
189 #define	V_pf_tag_z		 VNET(pf_tag_z)
190 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
191 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
192 
193 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
194 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
195 #endif
196 
197 static void		 pf_init_tagset(struct pf_tagset *, unsigned int *,
198 			    unsigned int);
199 static void		 pf_cleanup_tagset(struct pf_tagset *);
200 static uint16_t		 tagname2hashindex(const struct pf_tagset *, const char *);
201 static uint16_t		 tag2hashindex(const struct pf_tagset *, uint16_t);
202 static u_int16_t	 tagname2tag(struct pf_tagset *, const char *);
203 static u_int16_t	 pf_tagname2tag(const char *);
204 static void		 tag_unref(struct pf_tagset *, u_int16_t);
205 
206 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
207 
208 struct cdev *pf_dev;
209 
210 /*
211  * XXX - These are new and need to be checked when moveing to a new version
212  */
213 static void		 pf_clear_all_states(void);
214 static unsigned int	 pf_clear_states(const struct pf_kstate_kill *);
215 static void		 pf_killstates(struct pf_kstate_kill *,
216 			    unsigned int *);
217 static int		 pf_killstates_row(struct pf_kstate_kill *,
218 			    struct pf_idhash *);
219 static int		 pf_killstates_nv(struct pfioc_nv *);
220 static int		 pf_clearstates_nv(struct pfioc_nv *);
221 static int		 pf_getstate(struct pfioc_nv *);
222 static int		 pf_getstatus(struct pfioc_nv *);
223 static int		 pf_clear_tables(void);
224 static void		 pf_clear_srcnodes(struct pf_ksrc_node *);
225 static void		 pf_kill_srcnodes(struct pfioc_src_node_kill *);
226 static int		 pf_keepcounters(struct pfioc_nv *);
227 static void		 pf_tbladdr_copyout(struct pf_addr_wrap *);
228 
229 /*
230  * Wrapper functions for pfil(9) hooks
231  */
232 static pfil_return_t pf_eth_check_in(struct mbuf **m, struct ifnet *ifp,
233     int flags, void *ruleset __unused, struct inpcb *inp);
234 static pfil_return_t pf_eth_check_out(struct mbuf **m, struct ifnet *ifp,
235     int flags, void *ruleset __unused, struct inpcb *inp);
236 #ifdef INET
237 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp,
238     int flags, void *ruleset __unused, struct inpcb *inp);
239 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp,
240     int flags, void *ruleset __unused, struct inpcb *inp);
241 #endif
242 #ifdef INET6
243 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp,
244     int flags, void *ruleset __unused, struct inpcb *inp);
245 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp,
246     int flags, void *ruleset __unused, struct inpcb *inp);
247 #endif
248 
249 static void		hook_pf_eth(void);
250 static void		hook_pf(void);
251 static void		dehook_pf_eth(void);
252 static void		dehook_pf(void);
253 static int		shutdown_pf(void);
254 static int		pf_load(void);
255 static void		pf_unload(void);
256 
257 static struct cdevsw pf_cdevsw = {
258 	.d_ioctl =	pfioctl,
259 	.d_name =	PF_NAME,
260 	.d_version =	D_VERSION,
261 };
262 
263 volatile VNET_DEFINE_STATIC(int, pf_pfil_hooked);
264 #define V_pf_pfil_hooked	VNET(pf_pfil_hooked)
265 volatile VNET_DEFINE_STATIC(int, pf_pfil_eth_hooked);
266 #define V_pf_pfil_eth_hooked	VNET(pf_pfil_eth_hooked)
267 
268 /*
269  * We need a flag that is neither hooked nor running to know when
270  * the VNET is "valid".  We primarily need this to control (global)
271  * external event, e.g., eventhandlers.
272  */
273 VNET_DEFINE(int, pf_vnet_active);
274 #define V_pf_vnet_active	VNET(pf_vnet_active)
275 
276 int pf_end_threads;
277 struct proc *pf_purge_proc;
278 
279 struct rmlock			pf_rules_lock;
280 struct sx			pf_ioctl_lock;
281 struct sx			pf_end_lock;
282 
283 /* pfsync */
284 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
285 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
286 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
287 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
288 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
289 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
290 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
291 
292 /* pflog */
293 pflog_packet_t			*pflog_packet_ptr = NULL;
294 
295 /*
296  * Copy a user-provided string, returning an error if truncation would occur.
297  * Avoid scanning past "sz" bytes in the source string since there's no
298  * guarantee that it's nul-terminated.
299  */
300 static int
301 pf_user_strcpy(char *dst, const char *src, size_t sz)
302 {
303 	if (strnlen(src, sz) == sz)
304 		return (EINVAL);
305 	(void)strlcpy(dst, src, sz);
306 	return (0);
307 }
308 
309 static void
310 pfattach_vnet(void)
311 {
312 	u_int32_t *my_timeout = V_pf_default_rule.timeout;
313 
314 	pf_initialize();
315 	pfr_initialize();
316 	pfi_initialize_vnet();
317 	pf_normalize_init();
318 	pf_syncookies_init();
319 
320 	V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
321 	V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
322 
323 	RB_INIT(&V_pf_anchors);
324 	pf_init_kruleset(&pf_main_ruleset);
325 
326 	pf_init_keth(V_pf_keth);
327 
328 	/* default rule should never be garbage collected */
329 	V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
330 #ifdef PF_DEFAULT_TO_DROP
331 	V_pf_default_rule.action = PF_DROP;
332 #else
333 	V_pf_default_rule.action = PF_PASS;
334 #endif
335 	V_pf_default_rule.nr = -1;
336 	V_pf_default_rule.rtableid = -1;
337 
338 	pf_counter_u64_init(&V_pf_default_rule.evaluations, M_WAITOK);
339 	for (int i = 0; i < 2; i++) {
340 		pf_counter_u64_init(&V_pf_default_rule.packets[i], M_WAITOK);
341 		pf_counter_u64_init(&V_pf_default_rule.bytes[i], M_WAITOK);
342 	}
343 	V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
344 	V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
345 	V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
346 
347 #ifdef PF_WANT_32_TO_64_COUNTER
348 	V_pf_kifmarker = malloc(sizeof(*V_pf_kifmarker), PFI_MTYPE, M_WAITOK | M_ZERO);
349 	V_pf_rulemarker = malloc(sizeof(*V_pf_rulemarker), M_PFRULE, M_WAITOK | M_ZERO);
350 	PF_RULES_WLOCK();
351 	LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist);
352 	LIST_INSERT_HEAD(&V_pf_allrulelist, &V_pf_default_rule, allrulelist);
353 	V_pf_allrulecount++;
354 	LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist);
355 	PF_RULES_WUNLOCK();
356 #endif
357 
358 	/* initialize default timeouts */
359 	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
360 	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
361 	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
362 	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
363 	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
364 	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
365 	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
366 	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
367 	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
368 	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
369 	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
370 	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
371 	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
372 	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
373 	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
374 	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
375 	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
376 	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
377 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
378 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
379 
380 	bzero(&V_pf_status, sizeof(V_pf_status));
381 	V_pf_status.debug = PF_DEBUG_URGENT;
382 
383 	V_pf_pfil_hooked = 0;
384 	V_pf_pfil_eth_hooked = 0;
385 
386 	/* XXX do our best to avoid a conflict */
387 	V_pf_status.hostid = arc4random();
388 
389 	for (int i = 0; i < PFRES_MAX; i++)
390 		V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
391 	for (int i = 0; i < KLCNT_MAX; i++)
392 		V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
393 	for (int i = 0; i < FCNT_MAX; i++)
394 		pf_counter_u64_init(&V_pf_status.fcounters[i], M_WAITOK);
395 	for (int i = 0; i < SCNT_MAX; i++)
396 		V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
397 
398 	if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET,
399 	    INTR_MPSAFE, &V_pf_swi_cookie) != 0)
400 		/* XXXGL: leaked all above. */
401 		return;
402 }
403 
404 static struct pf_kpool *
405 pf_get_kpool(const char *anchor, u_int32_t ticket, u_int8_t rule_action,
406     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
407     u_int8_t check_ticket)
408 {
409 	struct pf_kruleset	*ruleset;
410 	struct pf_krule		*rule;
411 	int			 rs_num;
412 
413 	ruleset = pf_find_kruleset(anchor);
414 	if (ruleset == NULL)
415 		return (NULL);
416 	rs_num = pf_get_ruleset_number(rule_action);
417 	if (rs_num >= PF_RULESET_MAX)
418 		return (NULL);
419 	if (active) {
420 		if (check_ticket && ticket !=
421 		    ruleset->rules[rs_num].active.ticket)
422 			return (NULL);
423 		if (r_last)
424 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
425 			    pf_krulequeue);
426 		else
427 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
428 	} else {
429 		if (check_ticket && ticket !=
430 		    ruleset->rules[rs_num].inactive.ticket)
431 			return (NULL);
432 		if (r_last)
433 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
434 			    pf_krulequeue);
435 		else
436 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
437 	}
438 	if (!r_last) {
439 		while ((rule != NULL) && (rule->nr != rule_number))
440 			rule = TAILQ_NEXT(rule, entries);
441 	}
442 	if (rule == NULL)
443 		return (NULL);
444 
445 	return (&rule->rpool);
446 }
447 
448 static void
449 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb)
450 {
451 	struct pf_kpooladdr	*mv_pool_pa;
452 
453 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
454 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
455 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
456 	}
457 }
458 
459 static void
460 pf_empty_kpool(struct pf_kpalist *poola)
461 {
462 	struct pf_kpooladdr *pa;
463 
464 	while ((pa = TAILQ_FIRST(poola)) != NULL) {
465 		switch (pa->addr.type) {
466 		case PF_ADDR_DYNIFTL:
467 			pfi_dynaddr_remove(pa->addr.p.dyn);
468 			break;
469 		case PF_ADDR_TABLE:
470 			/* XXX: this could be unfinished pooladdr on pabuf */
471 			if (pa->addr.p.tbl != NULL)
472 				pfr_detach_table(pa->addr.p.tbl);
473 			break;
474 		}
475 		if (pa->kif)
476 			pfi_kkif_unref(pa->kif);
477 		TAILQ_REMOVE(poola, pa, entries);
478 		free(pa, M_PFRULE);
479 	}
480 }
481 
482 static void
483 pf_unlink_rule_locked(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
484 {
485 
486 	PF_RULES_WASSERT();
487 	PF_UNLNKDRULES_ASSERT();
488 
489 	TAILQ_REMOVE(rulequeue, rule, entries);
490 
491 	rule->rule_ref |= PFRULE_REFS;
492 	TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
493 }
494 
495 static void
496 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
497 {
498 
499 	PF_RULES_WASSERT();
500 
501 	PF_UNLNKDRULES_LOCK();
502 	pf_unlink_rule_locked(rulequeue, rule);
503 	PF_UNLNKDRULES_UNLOCK();
504 }
505 
506 static void
507 pf_free_eth_rule(struct pf_keth_rule *rule)
508 {
509 	PF_RULES_WASSERT();
510 
511 	if (rule == NULL)
512 		return;
513 
514 	if (rule->tag)
515 		tag_unref(&V_pf_tags, rule->tag);
516 #ifdef ALTQ
517 	pf_qid_unref(rule->qid);
518 #endif
519 
520 	if (rule->kif)
521 		pfi_kkif_unref(rule->kif);
522 
523 	if (rule->ipsrc.addr.type == PF_ADDR_TABLE)
524 		pfr_detach_table(rule->ipsrc.addr.p.tbl);
525 	if (rule->ipdst.addr.type == PF_ADDR_TABLE)
526 		pfr_detach_table(rule->ipdst.addr.p.tbl);
527 
528 	counter_u64_free(rule->evaluations);
529 	for (int i = 0; i < 2; i++) {
530 		counter_u64_free(rule->packets[i]);
531 		counter_u64_free(rule->bytes[i]);
532 	}
533 	pf_keth_anchor_remove(rule);
534 
535 	free(rule, M_PFRULE);
536 }
537 
538 void
539 pf_free_rule(struct pf_krule *rule)
540 {
541 
542 	PF_RULES_WASSERT();
543 	PF_CONFIG_ASSERT();
544 
545 	if (rule->tag)
546 		tag_unref(&V_pf_tags, rule->tag);
547 	if (rule->match_tag)
548 		tag_unref(&V_pf_tags, rule->match_tag);
549 #ifdef ALTQ
550 	if (rule->pqid != rule->qid)
551 		pf_qid_unref(rule->pqid);
552 	pf_qid_unref(rule->qid);
553 #endif
554 	switch (rule->src.addr.type) {
555 	case PF_ADDR_DYNIFTL:
556 		pfi_dynaddr_remove(rule->src.addr.p.dyn);
557 		break;
558 	case PF_ADDR_TABLE:
559 		pfr_detach_table(rule->src.addr.p.tbl);
560 		break;
561 	}
562 	switch (rule->dst.addr.type) {
563 	case PF_ADDR_DYNIFTL:
564 		pfi_dynaddr_remove(rule->dst.addr.p.dyn);
565 		break;
566 	case PF_ADDR_TABLE:
567 		pfr_detach_table(rule->dst.addr.p.tbl);
568 		break;
569 	}
570 	if (rule->overload_tbl)
571 		pfr_detach_table(rule->overload_tbl);
572 	if (rule->kif)
573 		pfi_kkif_unref(rule->kif);
574 	pf_kanchor_remove(rule);
575 	pf_empty_kpool(&rule->rpool.list);
576 
577 	pf_krule_free(rule);
578 }
579 
580 static void
581 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
582     unsigned int default_size)
583 {
584 	unsigned int i;
585 	unsigned int hashsize;
586 
587 	if (*tunable_size == 0 || !powerof2(*tunable_size))
588 		*tunable_size = default_size;
589 
590 	hashsize = *tunable_size;
591 	ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
592 	    M_WAITOK);
593 	ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
594 	    M_WAITOK);
595 	ts->mask = hashsize - 1;
596 	ts->seed = arc4random();
597 	for (i = 0; i < hashsize; i++) {
598 		TAILQ_INIT(&ts->namehash[i]);
599 		TAILQ_INIT(&ts->taghash[i]);
600 	}
601 	BIT_FILL(TAGID_MAX, &ts->avail);
602 }
603 
604 static void
605 pf_cleanup_tagset(struct pf_tagset *ts)
606 {
607 	unsigned int i;
608 	unsigned int hashsize;
609 	struct pf_tagname *t, *tmp;
610 
611 	/*
612 	 * Only need to clean up one of the hashes as each tag is hashed
613 	 * into each table.
614 	 */
615 	hashsize = ts->mask + 1;
616 	for (i = 0; i < hashsize; i++)
617 		TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
618 			uma_zfree(V_pf_tag_z, t);
619 
620 	free(ts->namehash, M_PFHASH);
621 	free(ts->taghash, M_PFHASH);
622 }
623 
624 static uint16_t
625 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
626 {
627 	size_t len;
628 
629 	len = strnlen(tagname, PF_TAG_NAME_SIZE - 1);
630 	return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask);
631 }
632 
633 static uint16_t
634 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
635 {
636 
637 	return (tag & ts->mask);
638 }
639 
640 static u_int16_t
641 tagname2tag(struct pf_tagset *ts, const char *tagname)
642 {
643 	struct pf_tagname	*tag;
644 	u_int32_t		 index;
645 	u_int16_t		 new_tagid;
646 
647 	PF_RULES_WASSERT();
648 
649 	index = tagname2hashindex(ts, tagname);
650 	TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
651 		if (strcmp(tagname, tag->name) == 0) {
652 			tag->ref++;
653 			return (tag->tag);
654 		}
655 
656 	/*
657 	 * new entry
658 	 *
659 	 * to avoid fragmentation, we do a linear search from the beginning
660 	 * and take the first free slot we find.
661 	 */
662 	new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
663 	/*
664 	 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
665 	 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
666 	 * set.  It may also return a bit number greater than TAGID_MAX due
667 	 * to rounding of the number of bits in the vector up to a multiple
668 	 * of the vector word size at declaration/allocation time.
669 	 */
670 	if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
671 		return (0);
672 
673 	/* Mark the tag as in use.  Bits are 0-based for BIT_CLR() */
674 	BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
675 
676 	/* allocate and fill new struct pf_tagname */
677 	tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
678 	if (tag == NULL)
679 		return (0);
680 	strlcpy(tag->name, tagname, sizeof(tag->name));
681 	tag->tag = new_tagid;
682 	tag->ref = 1;
683 
684 	/* Insert into namehash */
685 	TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
686 
687 	/* Insert into taghash */
688 	index = tag2hashindex(ts, new_tagid);
689 	TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
690 
691 	return (tag->tag);
692 }
693 
694 static void
695 tag_unref(struct pf_tagset *ts, u_int16_t tag)
696 {
697 	struct pf_tagname	*t;
698 	uint16_t		 index;
699 
700 	PF_RULES_WASSERT();
701 
702 	index = tag2hashindex(ts, tag);
703 	TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
704 		if (tag == t->tag) {
705 			if (--t->ref == 0) {
706 				TAILQ_REMOVE(&ts->taghash[index], t,
707 				    taghash_entries);
708 				index = tagname2hashindex(ts, t->name);
709 				TAILQ_REMOVE(&ts->namehash[index], t,
710 				    namehash_entries);
711 				/* Bits are 0-based for BIT_SET() */
712 				BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
713 				uma_zfree(V_pf_tag_z, t);
714 			}
715 			break;
716 		}
717 }
718 
719 static uint16_t
720 pf_tagname2tag(const char *tagname)
721 {
722 	return (tagname2tag(&V_pf_tags, tagname));
723 }
724 
725 static int
726 pf_begin_eth(uint32_t *ticket, const char *anchor)
727 {
728 	struct pf_keth_rule *rule, *tmp;
729 	struct pf_keth_ruleset *rs;
730 
731 	PF_RULES_WASSERT();
732 
733 	rs = pf_find_or_create_keth_ruleset(anchor);
734 	if (rs == NULL)
735 		return (EINVAL);
736 
737 	if (rs->inactive.open)
738 		/* We may be waiting for NET_EPOCH_CALL(pf_rollback_eth_cb) to
739 		 * finish. */
740 		return (EBUSY);
741 
742 	/* Purge old inactive rules. */
743 	TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
744 	    tmp) {
745 		TAILQ_REMOVE(rs->inactive.rules, rule,
746 		    entries);
747 		pf_free_eth_rule(rule);
748 	}
749 
750 	*ticket = ++rs->inactive.ticket;
751 	rs->inactive.open = 1;
752 
753 	return (0);
754 }
755 
756 static void
757 pf_rollback_eth_cb(struct epoch_context *ctx)
758 {
759 	struct pf_keth_ruleset *rs;
760 
761 	rs = __containerof(ctx, struct pf_keth_ruleset, epoch_ctx);
762 
763 	CURVNET_SET(rs->vnet);
764 
765 	PF_RULES_WLOCK();
766 	pf_rollback_eth(rs->inactive.ticket,
767 	    rs->anchor ? rs->anchor->path : "");
768 	PF_RULES_WUNLOCK();
769 
770 	CURVNET_RESTORE();
771 }
772 
773 static int
774 pf_rollback_eth(uint32_t ticket, const char *anchor)
775 {
776 	struct pf_keth_rule *rule, *tmp;
777 	struct pf_keth_ruleset *rs;
778 
779 	PF_RULES_WASSERT();
780 
781 	rs = pf_find_keth_ruleset(anchor);
782 	if (rs == NULL)
783 		return (EINVAL);
784 
785 	if (!rs->inactive.open ||
786 	    ticket != rs->inactive.ticket)
787 		return (0);
788 
789 	/* Purge old inactive rules. */
790 	TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
791 	    tmp) {
792 		TAILQ_REMOVE(rs->inactive.rules, rule, entries);
793 		pf_free_eth_rule(rule);
794 	}
795 
796 	rs->inactive.open = 0;
797 
798 	pf_remove_if_empty_keth_ruleset(rs);
799 
800 	return (0);
801 }
802 
803 #define	PF_SET_SKIP_STEPS(i)					\
804 	do {							\
805 		while (head[i] != cur) {			\
806 			head[i]->skip[i].ptr = cur;		\
807 			head[i] = TAILQ_NEXT(head[i], entries);	\
808 		}						\
809 	} while (0)
810 
811 static void
812 pf_eth_calc_skip_steps(struct pf_keth_ruleq *rules)
813 {
814 	struct pf_keth_rule *cur, *prev, *head[PFE_SKIP_COUNT];
815 	int i;
816 
817 	cur = TAILQ_FIRST(rules);
818 	prev = cur;
819 	for (i = 0; i < PFE_SKIP_COUNT; ++i)
820 		head[i] = cur;
821 	while (cur != NULL) {
822 		if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
823 			PF_SET_SKIP_STEPS(PFE_SKIP_IFP);
824 		if (cur->direction != prev->direction)
825 			PF_SET_SKIP_STEPS(PFE_SKIP_DIR);
826 		if (cur->proto != prev->proto)
827 			PF_SET_SKIP_STEPS(PFE_SKIP_PROTO);
828 		if (memcmp(&cur->src, &prev->src, sizeof(cur->src)) != 0)
829 			PF_SET_SKIP_STEPS(PFE_SKIP_SRC_ADDR);
830 		if (memcmp(&cur->dst, &prev->dst, sizeof(cur->dst)) != 0)
831 			PF_SET_SKIP_STEPS(PFE_SKIP_DST_ADDR);
832 
833 		prev = cur;
834 		cur = TAILQ_NEXT(cur, entries);
835 	}
836 	for (i = 0; i < PFE_SKIP_COUNT; ++i)
837 		PF_SET_SKIP_STEPS(i);
838 }
839 
840 static int
841 pf_commit_eth(uint32_t ticket, const char *anchor)
842 {
843 	struct pf_keth_ruleq *rules;
844 	struct pf_keth_ruleset *rs;
845 
846 	rs = pf_find_keth_ruleset(anchor);
847 	if (rs == NULL) {
848 		return (EINVAL);
849 	}
850 
851 	if (!rs->inactive.open ||
852 	    ticket != rs->inactive.ticket)
853 		return (EBUSY);
854 
855 	PF_RULES_WASSERT();
856 
857 	pf_eth_calc_skip_steps(rs->inactive.rules);
858 
859 	rules = rs->active.rules;
860 	ck_pr_store_ptr(&rs->active.rules, rs->inactive.rules);
861 	rs->inactive.rules = rules;
862 	rs->inactive.ticket = rs->active.ticket;
863 
864 	/* Clean up inactive rules (i.e. previously active rules), only when
865 	 * we're sure they're no longer used. */
866 	NET_EPOCH_CALL(pf_rollback_eth_cb, &rs->epoch_ctx);
867 
868 	return (0);
869 }
870 
871 #ifdef ALTQ
872 static uint16_t
873 pf_qname2qid(const char *qname)
874 {
875 	return (tagname2tag(&V_pf_qids, qname));
876 }
877 
878 static void
879 pf_qid_unref(uint16_t qid)
880 {
881 	tag_unref(&V_pf_qids, qid);
882 }
883 
884 static int
885 pf_begin_altq(u_int32_t *ticket)
886 {
887 	struct pf_altq	*altq, *tmp;
888 	int		 error = 0;
889 
890 	PF_RULES_WASSERT();
891 
892 	/* Purge the old altq lists */
893 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
894 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
895 			/* detach and destroy the discipline */
896 			error = altq_remove(altq);
897 		}
898 		free(altq, M_PFALTQ);
899 	}
900 	TAILQ_INIT(V_pf_altq_ifs_inactive);
901 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
902 		pf_qid_unref(altq->qid);
903 		free(altq, M_PFALTQ);
904 	}
905 	TAILQ_INIT(V_pf_altqs_inactive);
906 	if (error)
907 		return (error);
908 	*ticket = ++V_ticket_altqs_inactive;
909 	V_altqs_inactive_open = 1;
910 	return (0);
911 }
912 
913 static int
914 pf_rollback_altq(u_int32_t ticket)
915 {
916 	struct pf_altq	*altq, *tmp;
917 	int		 error = 0;
918 
919 	PF_RULES_WASSERT();
920 
921 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
922 		return (0);
923 	/* Purge the old altq lists */
924 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
925 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
926 			/* detach and destroy the discipline */
927 			error = altq_remove(altq);
928 		}
929 		free(altq, M_PFALTQ);
930 	}
931 	TAILQ_INIT(V_pf_altq_ifs_inactive);
932 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
933 		pf_qid_unref(altq->qid);
934 		free(altq, M_PFALTQ);
935 	}
936 	TAILQ_INIT(V_pf_altqs_inactive);
937 	V_altqs_inactive_open = 0;
938 	return (error);
939 }
940 
941 static int
942 pf_commit_altq(u_int32_t ticket)
943 {
944 	struct pf_altqqueue	*old_altqs, *old_altq_ifs;
945 	struct pf_altq		*altq, *tmp;
946 	int			 err, error = 0;
947 
948 	PF_RULES_WASSERT();
949 
950 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
951 		return (EBUSY);
952 
953 	/* swap altqs, keep the old. */
954 	old_altqs = V_pf_altqs_active;
955 	old_altq_ifs = V_pf_altq_ifs_active;
956 	V_pf_altqs_active = V_pf_altqs_inactive;
957 	V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
958 	V_pf_altqs_inactive = old_altqs;
959 	V_pf_altq_ifs_inactive = old_altq_ifs;
960 	V_ticket_altqs_active = V_ticket_altqs_inactive;
961 
962 	/* Attach new disciplines */
963 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
964 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
965 			/* attach the discipline */
966 			error = altq_pfattach(altq);
967 			if (error == 0 && V_pf_altq_running)
968 				error = pf_enable_altq(altq);
969 			if (error != 0)
970 				return (error);
971 		}
972 	}
973 
974 	/* Purge the old altq lists */
975 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
976 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
977 			/* detach and destroy the discipline */
978 			if (V_pf_altq_running)
979 				error = pf_disable_altq(altq);
980 			err = altq_pfdetach(altq);
981 			if (err != 0 && error == 0)
982 				error = err;
983 			err = altq_remove(altq);
984 			if (err != 0 && error == 0)
985 				error = err;
986 		}
987 		free(altq, M_PFALTQ);
988 	}
989 	TAILQ_INIT(V_pf_altq_ifs_inactive);
990 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
991 		pf_qid_unref(altq->qid);
992 		free(altq, M_PFALTQ);
993 	}
994 	TAILQ_INIT(V_pf_altqs_inactive);
995 
996 	V_altqs_inactive_open = 0;
997 	return (error);
998 }
999 
1000 static int
1001 pf_enable_altq(struct pf_altq *altq)
1002 {
1003 	struct ifnet		*ifp;
1004 	struct tb_profile	 tb;
1005 	int			 error = 0;
1006 
1007 	if ((ifp = ifunit(altq->ifname)) == NULL)
1008 		return (EINVAL);
1009 
1010 	if (ifp->if_snd.altq_type != ALTQT_NONE)
1011 		error = altq_enable(&ifp->if_snd);
1012 
1013 	/* set tokenbucket regulator */
1014 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
1015 		tb.rate = altq->ifbandwidth;
1016 		tb.depth = altq->tbrsize;
1017 		error = tbr_set(&ifp->if_snd, &tb);
1018 	}
1019 
1020 	return (error);
1021 }
1022 
1023 static int
1024 pf_disable_altq(struct pf_altq *altq)
1025 {
1026 	struct ifnet		*ifp;
1027 	struct tb_profile	 tb;
1028 	int			 error;
1029 
1030 	if ((ifp = ifunit(altq->ifname)) == NULL)
1031 		return (EINVAL);
1032 
1033 	/*
1034 	 * when the discipline is no longer referenced, it was overridden
1035 	 * by a new one.  if so, just return.
1036 	 */
1037 	if (altq->altq_disc != ifp->if_snd.altq_disc)
1038 		return (0);
1039 
1040 	error = altq_disable(&ifp->if_snd);
1041 
1042 	if (error == 0) {
1043 		/* clear tokenbucket regulator */
1044 		tb.rate = 0;
1045 		error = tbr_set(&ifp->if_snd, &tb);
1046 	}
1047 
1048 	return (error);
1049 }
1050 
1051 static int
1052 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
1053     struct pf_altq *altq)
1054 {
1055 	struct ifnet	*ifp1;
1056 	int		 error = 0;
1057 
1058 	/* Deactivate the interface in question */
1059 	altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
1060 	if ((ifp1 = ifunit(altq->ifname)) == NULL ||
1061 	    (remove && ifp1 == ifp)) {
1062 		altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
1063 	} else {
1064 		error = altq_add(ifp1, altq);
1065 
1066 		if (ticket != V_ticket_altqs_inactive)
1067 			error = EBUSY;
1068 
1069 		if (error)
1070 			free(altq, M_PFALTQ);
1071 	}
1072 
1073 	return (error);
1074 }
1075 
1076 void
1077 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
1078 {
1079 	struct pf_altq	*a1, *a2, *a3;
1080 	u_int32_t	 ticket;
1081 	int		 error = 0;
1082 
1083 	/*
1084 	 * No need to re-evaluate the configuration for events on interfaces
1085 	 * that do not support ALTQ, as it's not possible for such
1086 	 * interfaces to be part of the configuration.
1087 	 */
1088 	if (!ALTQ_IS_READY(&ifp->if_snd))
1089 		return;
1090 
1091 	/* Interrupt userland queue modifications */
1092 	if (V_altqs_inactive_open)
1093 		pf_rollback_altq(V_ticket_altqs_inactive);
1094 
1095 	/* Start new altq ruleset */
1096 	if (pf_begin_altq(&ticket))
1097 		return;
1098 
1099 	/* Copy the current active set */
1100 	TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
1101 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1102 		if (a2 == NULL) {
1103 			error = ENOMEM;
1104 			break;
1105 		}
1106 		bcopy(a1, a2, sizeof(struct pf_altq));
1107 
1108 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1109 		if (error)
1110 			break;
1111 
1112 		TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
1113 	}
1114 	if (error)
1115 		goto out;
1116 	TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
1117 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1118 		if (a2 == NULL) {
1119 			error = ENOMEM;
1120 			break;
1121 		}
1122 		bcopy(a1, a2, sizeof(struct pf_altq));
1123 
1124 		if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
1125 			error = EBUSY;
1126 			free(a2, M_PFALTQ);
1127 			break;
1128 		}
1129 		a2->altq_disc = NULL;
1130 		TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
1131 			if (strncmp(a3->ifname, a2->ifname,
1132 				IFNAMSIZ) == 0) {
1133 				a2->altq_disc = a3->altq_disc;
1134 				break;
1135 			}
1136 		}
1137 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1138 		if (error)
1139 			break;
1140 
1141 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
1142 	}
1143 
1144 out:
1145 	if (error != 0)
1146 		pf_rollback_altq(ticket);
1147 	else
1148 		pf_commit_altq(ticket);
1149 }
1150 #endif /* ALTQ */
1151 
1152 static int
1153 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
1154 {
1155 	struct pf_krule_global *tree;
1156 	struct pf_kruleset	*rs;
1157 	struct pf_krule		*rule;
1158 
1159 	PF_RULES_WASSERT();
1160 
1161 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1162 		return (EINVAL);
1163 	tree = malloc(sizeof(struct pf_krule_global), M_TEMP, M_NOWAIT);
1164 	if (tree == NULL)
1165 		return (ENOMEM);
1166 	RB_INIT(tree);
1167 	rs = pf_find_or_create_kruleset(anchor);
1168 	if (rs == NULL) {
1169 		free(tree, M_TEMP);
1170 		return (EINVAL);
1171 	}
1172 	free(rs->rules[rs_num].inactive.tree, M_TEMP);
1173 	rs->rules[rs_num].inactive.tree = tree;
1174 
1175 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1176 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1177 		rs->rules[rs_num].inactive.rcount--;
1178 	}
1179 	*ticket = ++rs->rules[rs_num].inactive.ticket;
1180 	rs->rules[rs_num].inactive.open = 1;
1181 	return (0);
1182 }
1183 
1184 static int
1185 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
1186 {
1187 	struct pf_kruleset	*rs;
1188 	struct pf_krule		*rule;
1189 
1190 	PF_RULES_WASSERT();
1191 
1192 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1193 		return (EINVAL);
1194 	rs = pf_find_kruleset(anchor);
1195 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1196 	    rs->rules[rs_num].inactive.ticket != ticket)
1197 		return (0);
1198 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1199 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1200 		rs->rules[rs_num].inactive.rcount--;
1201 	}
1202 	rs->rules[rs_num].inactive.open = 0;
1203 	return (0);
1204 }
1205 
1206 #define PF_MD5_UPD(st, elm)						\
1207 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
1208 
1209 #define PF_MD5_UPD_STR(st, elm)						\
1210 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
1211 
1212 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
1213 		(stor) = htonl((st)->elm);				\
1214 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
1215 } while (0)
1216 
1217 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
1218 		(stor) = htons((st)->elm);				\
1219 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
1220 } while (0)
1221 
1222 static void
1223 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
1224 {
1225 	PF_MD5_UPD(pfr, addr.type);
1226 	switch (pfr->addr.type) {
1227 		case PF_ADDR_DYNIFTL:
1228 			PF_MD5_UPD(pfr, addr.v.ifname);
1229 			PF_MD5_UPD(pfr, addr.iflags);
1230 			break;
1231 		case PF_ADDR_TABLE:
1232 			PF_MD5_UPD(pfr, addr.v.tblname);
1233 			break;
1234 		case PF_ADDR_ADDRMASK:
1235 			/* XXX ignore af? */
1236 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
1237 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
1238 			break;
1239 	}
1240 
1241 	PF_MD5_UPD(pfr, port[0]);
1242 	PF_MD5_UPD(pfr, port[1]);
1243 	PF_MD5_UPD(pfr, neg);
1244 	PF_MD5_UPD(pfr, port_op);
1245 }
1246 
1247 static void
1248 pf_hash_rule_rolling(MD5_CTX *ctx, struct pf_krule *rule)
1249 {
1250 	u_int16_t x;
1251 	u_int32_t y;
1252 
1253 	pf_hash_rule_addr(ctx, &rule->src);
1254 	pf_hash_rule_addr(ctx, &rule->dst);
1255 	for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++)
1256 		PF_MD5_UPD_STR(rule, label[i]);
1257 	PF_MD5_UPD_STR(rule, ifname);
1258 	PF_MD5_UPD_STR(rule, match_tagname);
1259 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
1260 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
1261 	PF_MD5_UPD_HTONL(rule, prob, y);
1262 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
1263 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1264 	PF_MD5_UPD(rule, uid.op);
1265 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1266 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1267 	PF_MD5_UPD(rule, gid.op);
1268 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
1269 	PF_MD5_UPD(rule, action);
1270 	PF_MD5_UPD(rule, direction);
1271 	PF_MD5_UPD(rule, af);
1272 	PF_MD5_UPD(rule, quick);
1273 	PF_MD5_UPD(rule, ifnot);
1274 	PF_MD5_UPD(rule, match_tag_not);
1275 	PF_MD5_UPD(rule, natpass);
1276 	PF_MD5_UPD(rule, keep_state);
1277 	PF_MD5_UPD(rule, proto);
1278 	PF_MD5_UPD(rule, type);
1279 	PF_MD5_UPD(rule, code);
1280 	PF_MD5_UPD(rule, flags);
1281 	PF_MD5_UPD(rule, flagset);
1282 	PF_MD5_UPD(rule, allow_opts);
1283 	PF_MD5_UPD(rule, rt);
1284 	PF_MD5_UPD(rule, tos);
1285 	if (rule->anchor != NULL)
1286 		PF_MD5_UPD_STR(rule, anchor->path);
1287 }
1288 
1289 static void
1290 pf_hash_rule(struct pf_krule *rule)
1291 {
1292 	MD5_CTX		ctx;
1293 
1294 	MD5Init(&ctx);
1295 	pf_hash_rule_rolling(&ctx, rule);
1296 	MD5Final(rule->md5sum, &ctx);
1297 }
1298 
1299 static int
1300 pf_krule_compare(struct pf_krule *a, struct pf_krule *b)
1301 {
1302 
1303 	return (memcmp(a->md5sum, b->md5sum, PF_MD5_DIGEST_LENGTH));
1304 }
1305 
1306 static int
1307 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1308 {
1309 	struct pf_kruleset	*rs;
1310 	struct pf_krule		*rule, **old_array, *old_rule;
1311 	struct pf_krulequeue	*old_rules;
1312 	struct pf_krule_global  *old_tree;
1313 	int			 error;
1314 	u_int32_t		 old_rcount;
1315 
1316 	PF_RULES_WASSERT();
1317 
1318 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1319 		return (EINVAL);
1320 	rs = pf_find_kruleset(anchor);
1321 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1322 	    ticket != rs->rules[rs_num].inactive.ticket)
1323 		return (EBUSY);
1324 
1325 	/* Calculate checksum for the main ruleset */
1326 	if (rs == &pf_main_ruleset) {
1327 		error = pf_setup_pfsync_matching(rs);
1328 		if (error != 0)
1329 			return (error);
1330 	}
1331 
1332 	/* Swap rules, keep the old. */
1333 	old_rules = rs->rules[rs_num].active.ptr;
1334 	old_rcount = rs->rules[rs_num].active.rcount;
1335 	old_array = rs->rules[rs_num].active.ptr_array;
1336 	old_tree = rs->rules[rs_num].active.tree;
1337 
1338 	rs->rules[rs_num].active.ptr =
1339 	    rs->rules[rs_num].inactive.ptr;
1340 	rs->rules[rs_num].active.ptr_array =
1341 	    rs->rules[rs_num].inactive.ptr_array;
1342 	rs->rules[rs_num].active.tree =
1343 	    rs->rules[rs_num].inactive.tree;
1344 	rs->rules[rs_num].active.rcount =
1345 	    rs->rules[rs_num].inactive.rcount;
1346 
1347 	/* Attempt to preserve counter information. */
1348 	if (V_pf_status.keep_counters && old_tree != NULL) {
1349 		TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr,
1350 		    entries) {
1351 			old_rule = RB_FIND(pf_krule_global, old_tree, rule);
1352 			if (old_rule == NULL) {
1353 				continue;
1354 			}
1355 			pf_counter_u64_critical_enter();
1356 			pf_counter_u64_add_protected(&rule->evaluations,
1357 			    pf_counter_u64_fetch(&old_rule->evaluations));
1358 			pf_counter_u64_add_protected(&rule->packets[0],
1359 			    pf_counter_u64_fetch(&old_rule->packets[0]));
1360 			pf_counter_u64_add_protected(&rule->packets[1],
1361 			    pf_counter_u64_fetch(&old_rule->packets[1]));
1362 			pf_counter_u64_add_protected(&rule->bytes[0],
1363 			    pf_counter_u64_fetch(&old_rule->bytes[0]));
1364 			pf_counter_u64_add_protected(&rule->bytes[1],
1365 			    pf_counter_u64_fetch(&old_rule->bytes[1]));
1366 			pf_counter_u64_critical_exit();
1367 		}
1368 	}
1369 
1370 	rs->rules[rs_num].inactive.ptr = old_rules;
1371 	rs->rules[rs_num].inactive.ptr_array = old_array;
1372 	rs->rules[rs_num].inactive.tree = NULL; /* important for pf_ioctl_addrule */
1373 	rs->rules[rs_num].inactive.rcount = old_rcount;
1374 
1375 	rs->rules[rs_num].active.ticket =
1376 	    rs->rules[rs_num].inactive.ticket;
1377 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1378 
1379 	/* Purge the old rule list. */
1380 	PF_UNLNKDRULES_LOCK();
1381 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1382 		pf_unlink_rule_locked(old_rules, rule);
1383 	PF_UNLNKDRULES_UNLOCK();
1384 	if (rs->rules[rs_num].inactive.ptr_array)
1385 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1386 	rs->rules[rs_num].inactive.ptr_array = NULL;
1387 	rs->rules[rs_num].inactive.rcount = 0;
1388 	rs->rules[rs_num].inactive.open = 0;
1389 	pf_remove_if_empty_kruleset(rs);
1390 	free(old_tree, M_TEMP);
1391 
1392 	return (0);
1393 }
1394 
1395 static int
1396 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1397 {
1398 	MD5_CTX			 ctx;
1399 	struct pf_krule		*rule;
1400 	int			 rs_cnt;
1401 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
1402 
1403 	MD5Init(&ctx);
1404 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1405 		/* XXX PF_RULESET_SCRUB as well? */
1406 		if (rs_cnt == PF_RULESET_SCRUB)
1407 			continue;
1408 
1409 		if (rs->rules[rs_cnt].inactive.ptr_array)
1410 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1411 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
1412 
1413 		if (rs->rules[rs_cnt].inactive.rcount) {
1414 			rs->rules[rs_cnt].inactive.ptr_array =
1415 			    malloc(sizeof(caddr_t) *
1416 			    rs->rules[rs_cnt].inactive.rcount,
1417 			    M_TEMP, M_NOWAIT);
1418 
1419 			if (!rs->rules[rs_cnt].inactive.ptr_array)
1420 				return (ENOMEM);
1421 		}
1422 
1423 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1424 		    entries) {
1425 			pf_hash_rule_rolling(&ctx, rule);
1426 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1427 		}
1428 	}
1429 
1430 	MD5Final(digest, &ctx);
1431 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1432 	return (0);
1433 }
1434 
1435 static int
1436 pf_eth_addr_setup(struct pf_keth_ruleset *ruleset, struct pf_addr_wrap *addr)
1437 {
1438 	int error = 0;
1439 
1440 	switch (addr->type) {
1441 	case PF_ADDR_TABLE:
1442 		addr->p.tbl = pfr_eth_attach_table(ruleset, addr->v.tblname);
1443 		if (addr->p.tbl == NULL)
1444 			error = ENOMEM;
1445 		break;
1446 	default:
1447 		error = EINVAL;
1448 	}
1449 
1450 	return (error);
1451 }
1452 
1453 static int
1454 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1455     sa_family_t af)
1456 {
1457 	int error = 0;
1458 
1459 	switch (addr->type) {
1460 	case PF_ADDR_TABLE:
1461 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1462 		if (addr->p.tbl == NULL)
1463 			error = ENOMEM;
1464 		break;
1465 	case PF_ADDR_DYNIFTL:
1466 		error = pfi_dynaddr_setup(addr, af);
1467 		break;
1468 	}
1469 
1470 	return (error);
1471 }
1472 
1473 static void
1474 pf_addr_copyout(struct pf_addr_wrap *addr)
1475 {
1476 
1477 	switch (addr->type) {
1478 	case PF_ADDR_DYNIFTL:
1479 		pfi_dynaddr_copyout(addr);
1480 		break;
1481 	case PF_ADDR_TABLE:
1482 		pf_tbladdr_copyout(addr);
1483 		break;
1484 	}
1485 }
1486 
1487 static void
1488 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
1489 {
1490 	int	secs = time_uptime, diff;
1491 
1492 	bzero(out, sizeof(struct pf_src_node));
1493 
1494 	bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
1495 	bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
1496 
1497 	if (in->rule.ptr != NULL)
1498 		out->rule.nr = in->rule.ptr->nr;
1499 
1500 	for (int i = 0; i < 2; i++) {
1501 		out->bytes[i] = counter_u64_fetch(in->bytes[i]);
1502 		out->packets[i] = counter_u64_fetch(in->packets[i]);
1503 	}
1504 
1505 	out->states = in->states;
1506 	out->conn = in->conn;
1507 	out->af = in->af;
1508 	out->ruletype = in->ruletype;
1509 
1510 	out->creation = secs - in->creation;
1511 	if (out->expire > secs)
1512 		out->expire -= secs;
1513 	else
1514 		out->expire = 0;
1515 
1516 	/* Adjust the connection rate estimate. */
1517 	diff = secs - in->conn_rate.last;
1518 	if (diff >= in->conn_rate.seconds)
1519 		out->conn_rate.count = 0;
1520 	else
1521 		out->conn_rate.count -=
1522 		    in->conn_rate.count * diff /
1523 		    in->conn_rate.seconds;
1524 }
1525 
1526 #ifdef ALTQ
1527 /*
1528  * Handle export of struct pf_kaltq to user binaries that may be using any
1529  * version of struct pf_altq.
1530  */
1531 static int
1532 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1533 {
1534 	u_int32_t version;
1535 
1536 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1537 		version = 0;
1538 	else
1539 		version = pa->version;
1540 
1541 	if (version > PFIOC_ALTQ_VERSION)
1542 		return (EINVAL);
1543 
1544 #define ASSIGN(x) exported_q->x = q->x
1545 #define COPY(x) \
1546 	bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1547 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1548 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1549 
1550 	switch (version) {
1551 	case 0: {
1552 		struct pf_altq_v0 *exported_q =
1553 		    &((struct pfioc_altq_v0 *)pa)->altq;
1554 
1555 		COPY(ifname);
1556 
1557 		ASSIGN(scheduler);
1558 		ASSIGN(tbrsize);
1559 		exported_q->tbrsize = SATU16(q->tbrsize);
1560 		exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1561 
1562 		COPY(qname);
1563 		COPY(parent);
1564 		ASSIGN(parent_qid);
1565 		exported_q->bandwidth = SATU32(q->bandwidth);
1566 		ASSIGN(priority);
1567 		ASSIGN(local_flags);
1568 
1569 		ASSIGN(qlimit);
1570 		ASSIGN(flags);
1571 
1572 		if (q->scheduler == ALTQT_HFSC) {
1573 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1574 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1575 			    SATU32(q->pq_u.hfsc_opts.x)
1576 
1577 			ASSIGN_OPT_SATU32(rtsc_m1);
1578 			ASSIGN_OPT(rtsc_d);
1579 			ASSIGN_OPT_SATU32(rtsc_m2);
1580 
1581 			ASSIGN_OPT_SATU32(lssc_m1);
1582 			ASSIGN_OPT(lssc_d);
1583 			ASSIGN_OPT_SATU32(lssc_m2);
1584 
1585 			ASSIGN_OPT_SATU32(ulsc_m1);
1586 			ASSIGN_OPT(ulsc_d);
1587 			ASSIGN_OPT_SATU32(ulsc_m2);
1588 
1589 			ASSIGN_OPT(flags);
1590 
1591 #undef ASSIGN_OPT
1592 #undef ASSIGN_OPT_SATU32
1593 		} else
1594 			COPY(pq_u);
1595 
1596 		ASSIGN(qid);
1597 		break;
1598 	}
1599 	case 1:	{
1600 		struct pf_altq_v1 *exported_q =
1601 		    &((struct pfioc_altq_v1 *)pa)->altq;
1602 
1603 		COPY(ifname);
1604 
1605 		ASSIGN(scheduler);
1606 		ASSIGN(tbrsize);
1607 		ASSIGN(ifbandwidth);
1608 
1609 		COPY(qname);
1610 		COPY(parent);
1611 		ASSIGN(parent_qid);
1612 		ASSIGN(bandwidth);
1613 		ASSIGN(priority);
1614 		ASSIGN(local_flags);
1615 
1616 		ASSIGN(qlimit);
1617 		ASSIGN(flags);
1618 		COPY(pq_u);
1619 
1620 		ASSIGN(qid);
1621 		break;
1622 	}
1623 	default:
1624 		panic("%s: unhandled struct pfioc_altq version", __func__);
1625 		break;
1626 	}
1627 
1628 #undef ASSIGN
1629 #undef COPY
1630 #undef SATU16
1631 #undef SATU32
1632 
1633 	return (0);
1634 }
1635 
1636 /*
1637  * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1638  * that may be using any version of it.
1639  */
1640 static int
1641 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1642 {
1643 	u_int32_t version;
1644 
1645 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1646 		version = 0;
1647 	else
1648 		version = pa->version;
1649 
1650 	if (version > PFIOC_ALTQ_VERSION)
1651 		return (EINVAL);
1652 
1653 #define ASSIGN(x) q->x = imported_q->x
1654 #define COPY(x) \
1655 	bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1656 
1657 	switch (version) {
1658 	case 0: {
1659 		struct pf_altq_v0 *imported_q =
1660 		    &((struct pfioc_altq_v0 *)pa)->altq;
1661 
1662 		COPY(ifname);
1663 
1664 		ASSIGN(scheduler);
1665 		ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1666 		ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1667 
1668 		COPY(qname);
1669 		COPY(parent);
1670 		ASSIGN(parent_qid);
1671 		ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1672 		ASSIGN(priority);
1673 		ASSIGN(local_flags);
1674 
1675 		ASSIGN(qlimit);
1676 		ASSIGN(flags);
1677 
1678 		if (imported_q->scheduler == ALTQT_HFSC) {
1679 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1680 
1681 			/*
1682 			 * The m1 and m2 parameters are being copied from
1683 			 * 32-bit to 64-bit.
1684 			 */
1685 			ASSIGN_OPT(rtsc_m1);
1686 			ASSIGN_OPT(rtsc_d);
1687 			ASSIGN_OPT(rtsc_m2);
1688 
1689 			ASSIGN_OPT(lssc_m1);
1690 			ASSIGN_OPT(lssc_d);
1691 			ASSIGN_OPT(lssc_m2);
1692 
1693 			ASSIGN_OPT(ulsc_m1);
1694 			ASSIGN_OPT(ulsc_d);
1695 			ASSIGN_OPT(ulsc_m2);
1696 
1697 			ASSIGN_OPT(flags);
1698 
1699 #undef ASSIGN_OPT
1700 		} else
1701 			COPY(pq_u);
1702 
1703 		ASSIGN(qid);
1704 		break;
1705 	}
1706 	case 1: {
1707 		struct pf_altq_v1 *imported_q =
1708 		    &((struct pfioc_altq_v1 *)pa)->altq;
1709 
1710 		COPY(ifname);
1711 
1712 		ASSIGN(scheduler);
1713 		ASSIGN(tbrsize);
1714 		ASSIGN(ifbandwidth);
1715 
1716 		COPY(qname);
1717 		COPY(parent);
1718 		ASSIGN(parent_qid);
1719 		ASSIGN(bandwidth);
1720 		ASSIGN(priority);
1721 		ASSIGN(local_flags);
1722 
1723 		ASSIGN(qlimit);
1724 		ASSIGN(flags);
1725 		COPY(pq_u);
1726 
1727 		ASSIGN(qid);
1728 		break;
1729 	}
1730 	default:
1731 		panic("%s: unhandled struct pfioc_altq version", __func__);
1732 		break;
1733 	}
1734 
1735 #undef ASSIGN
1736 #undef COPY
1737 
1738 	return (0);
1739 }
1740 
1741 static struct pf_altq *
1742 pf_altq_get_nth_active(u_int32_t n)
1743 {
1744 	struct pf_altq		*altq;
1745 	u_int32_t		 nr;
1746 
1747 	nr = 0;
1748 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1749 		if (nr == n)
1750 			return (altq);
1751 		nr++;
1752 	}
1753 
1754 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1755 		if (nr == n)
1756 			return (altq);
1757 		nr++;
1758 	}
1759 
1760 	return (NULL);
1761 }
1762 #endif /* ALTQ */
1763 
1764 struct pf_krule *
1765 pf_krule_alloc(void)
1766 {
1767 	struct pf_krule *rule;
1768 
1769 	rule = malloc(sizeof(struct pf_krule), M_PFRULE, M_WAITOK | M_ZERO);
1770 	mtx_init(&rule->rpool.mtx, "pf_krule_pool", NULL, MTX_DEF);
1771 	return (rule);
1772 }
1773 
1774 void
1775 pf_krule_free(struct pf_krule *rule)
1776 {
1777 #ifdef PF_WANT_32_TO_64_COUNTER
1778 	bool wowned;
1779 #endif
1780 
1781 	if (rule == NULL)
1782 		return;
1783 
1784 #ifdef PF_WANT_32_TO_64_COUNTER
1785 	if (rule->allrulelinked) {
1786 		wowned = PF_RULES_WOWNED();
1787 		if (!wowned)
1788 			PF_RULES_WLOCK();
1789 		LIST_REMOVE(rule, allrulelist);
1790 		V_pf_allrulecount--;
1791 		if (!wowned)
1792 			PF_RULES_WUNLOCK();
1793 	}
1794 #endif
1795 
1796 	pf_counter_u64_deinit(&rule->evaluations);
1797 	for (int i = 0; i < 2; i++) {
1798 		pf_counter_u64_deinit(&rule->packets[i]);
1799 		pf_counter_u64_deinit(&rule->bytes[i]);
1800 	}
1801 	counter_u64_free(rule->states_cur);
1802 	counter_u64_free(rule->states_tot);
1803 	counter_u64_free(rule->src_nodes);
1804 
1805 	mtx_destroy(&rule->rpool.mtx);
1806 	free(rule, M_PFRULE);
1807 }
1808 
1809 static void
1810 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool,
1811     struct pf_pooladdr *pool)
1812 {
1813 
1814 	bzero(pool, sizeof(*pool));
1815 	bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr));
1816 	strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname));
1817 }
1818 
1819 static int
1820 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool,
1821     struct pf_kpooladdr *kpool)
1822 {
1823 	int ret;
1824 
1825 	bzero(kpool, sizeof(*kpool));
1826 	bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr));
1827 	ret = pf_user_strcpy(kpool->ifname, pool->ifname,
1828 	    sizeof(kpool->ifname));
1829 	return (ret);
1830 }
1831 
1832 static void
1833 pf_kpool_to_pool(const struct pf_kpool *kpool, struct pf_pool *pool)
1834 {
1835 	bzero(pool, sizeof(*pool));
1836 
1837 	bcopy(&kpool->key, &pool->key, sizeof(pool->key));
1838 	bcopy(&kpool->counter, &pool->counter, sizeof(pool->counter));
1839 
1840 	pool->tblidx = kpool->tblidx;
1841 	pool->proxy_port[0] = kpool->proxy_port[0];
1842 	pool->proxy_port[1] = kpool->proxy_port[1];
1843 	pool->opts = kpool->opts;
1844 }
1845 
1846 static void
1847 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool)
1848 {
1849 	_Static_assert(sizeof(pool->key) == sizeof(kpool->key), "");
1850 	_Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), "");
1851 
1852 	bcopy(&pool->key, &kpool->key, sizeof(kpool->key));
1853 	bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter));
1854 
1855 	kpool->tblidx = pool->tblidx;
1856 	kpool->proxy_port[0] = pool->proxy_port[0];
1857 	kpool->proxy_port[1] = pool->proxy_port[1];
1858 	kpool->opts = pool->opts;
1859 }
1860 
1861 static void
1862 pf_krule_to_rule(const struct pf_krule *krule, struct pf_rule *rule)
1863 {
1864 
1865 	bzero(rule, sizeof(*rule));
1866 
1867 	bcopy(&krule->src, &rule->src, sizeof(rule->src));
1868 	bcopy(&krule->dst, &rule->dst, sizeof(rule->dst));
1869 
1870 	for (int i = 0; i < PF_SKIP_COUNT; ++i) {
1871 		if (rule->skip[i].ptr == NULL)
1872 			rule->skip[i].nr = -1;
1873 		else
1874 			rule->skip[i].nr = krule->skip[i].ptr->nr;
1875 	}
1876 
1877 	strlcpy(rule->label, krule->label[0], sizeof(rule->label));
1878 	strlcpy(rule->ifname, krule->ifname, sizeof(rule->ifname));
1879 	strlcpy(rule->qname, krule->qname, sizeof(rule->qname));
1880 	strlcpy(rule->pqname, krule->pqname, sizeof(rule->pqname));
1881 	strlcpy(rule->tagname, krule->tagname, sizeof(rule->tagname));
1882 	strlcpy(rule->match_tagname, krule->match_tagname,
1883 	    sizeof(rule->match_tagname));
1884 	strlcpy(rule->overload_tblname, krule->overload_tblname,
1885 	    sizeof(rule->overload_tblname));
1886 
1887 	pf_kpool_to_pool(&krule->rpool, &rule->rpool);
1888 
1889 	rule->evaluations = pf_counter_u64_fetch(&krule->evaluations);
1890 	for (int i = 0; i < 2; i++) {
1891 		rule->packets[i] = pf_counter_u64_fetch(&krule->packets[i]);
1892 		rule->bytes[i] = pf_counter_u64_fetch(&krule->bytes[i]);
1893 	}
1894 
1895 	/* kif, anchor, overload_tbl are not copied over. */
1896 
1897 	rule->os_fingerprint = krule->os_fingerprint;
1898 
1899 	rule->rtableid = krule->rtableid;
1900 	bcopy(krule->timeout, rule->timeout, sizeof(krule->timeout));
1901 	rule->max_states = krule->max_states;
1902 	rule->max_src_nodes = krule->max_src_nodes;
1903 	rule->max_src_states = krule->max_src_states;
1904 	rule->max_src_conn = krule->max_src_conn;
1905 	rule->max_src_conn_rate.limit = krule->max_src_conn_rate.limit;
1906 	rule->max_src_conn_rate.seconds = krule->max_src_conn_rate.seconds;
1907 	rule->qid = krule->qid;
1908 	rule->pqid = krule->pqid;
1909 	rule->nr = krule->nr;
1910 	rule->prob = krule->prob;
1911 	rule->cuid = krule->cuid;
1912 	rule->cpid = krule->cpid;
1913 
1914 	rule->return_icmp = krule->return_icmp;
1915 	rule->return_icmp6 = krule->return_icmp6;
1916 	rule->max_mss = krule->max_mss;
1917 	rule->tag = krule->tag;
1918 	rule->match_tag = krule->match_tag;
1919 	rule->scrub_flags = krule->scrub_flags;
1920 
1921 	bcopy(&krule->uid, &rule->uid, sizeof(krule->uid));
1922 	bcopy(&krule->gid, &rule->gid, sizeof(krule->gid));
1923 
1924 	rule->rule_flag = krule->rule_flag;
1925 	rule->action = krule->action;
1926 	rule->direction = krule->direction;
1927 	rule->log = krule->log;
1928 	rule->logif = krule->logif;
1929 	rule->quick = krule->quick;
1930 	rule->ifnot = krule->ifnot;
1931 	rule->match_tag_not = krule->match_tag_not;
1932 	rule->natpass = krule->natpass;
1933 
1934 	rule->keep_state = krule->keep_state;
1935 	rule->af = krule->af;
1936 	rule->proto = krule->proto;
1937 	rule->type = krule->type;
1938 	rule->code = krule->code;
1939 	rule->flags = krule->flags;
1940 	rule->flagset = krule->flagset;
1941 	rule->min_ttl = krule->min_ttl;
1942 	rule->allow_opts = krule->allow_opts;
1943 	rule->rt = krule->rt;
1944 	rule->return_ttl = krule->return_ttl;
1945 	rule->tos = krule->tos;
1946 	rule->set_tos = krule->set_tos;
1947 	rule->anchor_relative = krule->anchor_relative;
1948 	rule->anchor_wildcard = krule->anchor_wildcard;
1949 
1950 	rule->flush = krule->flush;
1951 	rule->prio = krule->prio;
1952 	rule->set_prio[0] = krule->set_prio[0];
1953 	rule->set_prio[1] = krule->set_prio[1];
1954 
1955 	bcopy(&krule->divert, &rule->divert, sizeof(krule->divert));
1956 
1957 	rule->u_states_cur = counter_u64_fetch(krule->states_cur);
1958 	rule->u_states_tot = counter_u64_fetch(krule->states_tot);
1959 	rule->u_src_nodes = counter_u64_fetch(krule->src_nodes);
1960 }
1961 
1962 static int
1963 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
1964 {
1965 	int ret;
1966 
1967 #ifndef INET
1968 	if (rule->af == AF_INET) {
1969 		return (EAFNOSUPPORT);
1970 	}
1971 #endif /* INET */
1972 #ifndef INET6
1973 	if (rule->af == AF_INET6) {
1974 		return (EAFNOSUPPORT);
1975 	}
1976 #endif /* INET6 */
1977 
1978 	ret = pf_check_rule_addr(&rule->src);
1979 	if (ret != 0)
1980 		return (ret);
1981 	ret = pf_check_rule_addr(&rule->dst);
1982 	if (ret != 0)
1983 		return (ret);
1984 
1985 	bcopy(&rule->src, &krule->src, sizeof(rule->src));
1986 	bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
1987 
1988 	ret = pf_user_strcpy(krule->label[0], rule->label, sizeof(rule->label));
1989 	if (ret != 0)
1990 		return (ret);
1991 	ret = pf_user_strcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
1992 	if (ret != 0)
1993 		return (ret);
1994 	ret = pf_user_strcpy(krule->qname, rule->qname, sizeof(rule->qname));
1995 	if (ret != 0)
1996 		return (ret);
1997 	ret = pf_user_strcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
1998 	if (ret != 0)
1999 		return (ret);
2000 	ret = pf_user_strcpy(krule->tagname, rule->tagname,
2001 	    sizeof(rule->tagname));
2002 	if (ret != 0)
2003 		return (ret);
2004 	ret = pf_user_strcpy(krule->match_tagname, rule->match_tagname,
2005 	    sizeof(rule->match_tagname));
2006 	if (ret != 0)
2007 		return (ret);
2008 	ret = pf_user_strcpy(krule->overload_tblname, rule->overload_tblname,
2009 	    sizeof(rule->overload_tblname));
2010 	if (ret != 0)
2011 		return (ret);
2012 
2013 	pf_pool_to_kpool(&rule->rpool, &krule->rpool);
2014 
2015 	/* Don't allow userspace to set evaulations, packets or bytes. */
2016 	/* kif, anchor, overload_tbl are not copied over. */
2017 
2018 	krule->os_fingerprint = rule->os_fingerprint;
2019 
2020 	krule->rtableid = rule->rtableid;
2021 	bcopy(rule->timeout, krule->timeout, sizeof(krule->timeout));
2022 	krule->max_states = rule->max_states;
2023 	krule->max_src_nodes = rule->max_src_nodes;
2024 	krule->max_src_states = rule->max_src_states;
2025 	krule->max_src_conn = rule->max_src_conn;
2026 	krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
2027 	krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
2028 	krule->qid = rule->qid;
2029 	krule->pqid = rule->pqid;
2030 	krule->nr = rule->nr;
2031 	krule->prob = rule->prob;
2032 	krule->cuid = rule->cuid;
2033 	krule->cpid = rule->cpid;
2034 
2035 	krule->return_icmp = rule->return_icmp;
2036 	krule->return_icmp6 = rule->return_icmp6;
2037 	krule->max_mss = rule->max_mss;
2038 	krule->tag = rule->tag;
2039 	krule->match_tag = rule->match_tag;
2040 	krule->scrub_flags = rule->scrub_flags;
2041 
2042 	bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
2043 	bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
2044 
2045 	krule->rule_flag = rule->rule_flag;
2046 	krule->action = rule->action;
2047 	krule->direction = rule->direction;
2048 	krule->log = rule->log;
2049 	krule->logif = rule->logif;
2050 	krule->quick = rule->quick;
2051 	krule->ifnot = rule->ifnot;
2052 	krule->match_tag_not = rule->match_tag_not;
2053 	krule->natpass = rule->natpass;
2054 
2055 	krule->keep_state = rule->keep_state;
2056 	krule->af = rule->af;
2057 	krule->proto = rule->proto;
2058 	krule->type = rule->type;
2059 	krule->code = rule->code;
2060 	krule->flags = rule->flags;
2061 	krule->flagset = rule->flagset;
2062 	krule->min_ttl = rule->min_ttl;
2063 	krule->allow_opts = rule->allow_opts;
2064 	krule->rt = rule->rt;
2065 	krule->return_ttl = rule->return_ttl;
2066 	krule->tos = rule->tos;
2067 	krule->set_tos = rule->set_tos;
2068 
2069 	krule->flush = rule->flush;
2070 	krule->prio = rule->prio;
2071 	krule->set_prio[0] = rule->set_prio[0];
2072 	krule->set_prio[1] = rule->set_prio[1];
2073 
2074 	bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
2075 
2076 	return (0);
2077 }
2078 
2079 static int
2080 pf_state_kill_to_kstate_kill(const struct pfioc_state_kill *psk,
2081     struct pf_kstate_kill *kill)
2082 {
2083 	int ret;
2084 
2085 	bzero(kill, sizeof(*kill));
2086 
2087 	bcopy(&psk->psk_pfcmp, &kill->psk_pfcmp, sizeof(kill->psk_pfcmp));
2088 	kill->psk_af = psk->psk_af;
2089 	kill->psk_proto = psk->psk_proto;
2090 	bcopy(&psk->psk_src, &kill->psk_src, sizeof(kill->psk_src));
2091 	bcopy(&psk->psk_dst, &kill->psk_dst, sizeof(kill->psk_dst));
2092 	ret = pf_user_strcpy(kill->psk_ifname, psk->psk_ifname,
2093 	    sizeof(kill->psk_ifname));
2094 	if (ret != 0)
2095 		return (ret);
2096 	ret = pf_user_strcpy(kill->psk_label, psk->psk_label,
2097 	    sizeof(kill->psk_label));
2098 	if (ret != 0)
2099 		return (ret);
2100 
2101 	return (0);
2102 }
2103 
2104 static int
2105 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket,
2106     uint32_t pool_ticket, const char *anchor, const char *anchor_call,
2107     struct thread *td)
2108 {
2109 	struct pf_kruleset	*ruleset;
2110 	struct pf_krule		*tail;
2111 	struct pf_kpooladdr	*pa;
2112 	struct pfi_kkif		*kif = NULL;
2113 	int			 rs_num;
2114 	int			 error = 0;
2115 
2116 	if ((rule->return_icmp >> 8) > ICMP_MAXTYPE) {
2117 		error = EINVAL;
2118 		goto errout_unlocked;
2119 	}
2120 
2121 #define	ERROUT(x)	ERROUT_FUNCTION(errout, x)
2122 
2123 	if (rule->ifname[0])
2124 		kif = pf_kkif_create(M_WAITOK);
2125 	pf_counter_u64_init(&rule->evaluations, M_WAITOK);
2126 	for (int i = 0; i < 2; i++) {
2127 		pf_counter_u64_init(&rule->packets[i], M_WAITOK);
2128 		pf_counter_u64_init(&rule->bytes[i], M_WAITOK);
2129 	}
2130 	rule->states_cur = counter_u64_alloc(M_WAITOK);
2131 	rule->states_tot = counter_u64_alloc(M_WAITOK);
2132 	rule->src_nodes = counter_u64_alloc(M_WAITOK);
2133 	rule->cuid = td->td_ucred->cr_ruid;
2134 	rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
2135 	TAILQ_INIT(&rule->rpool.list);
2136 
2137 	PF_CONFIG_LOCK();
2138 	PF_RULES_WLOCK();
2139 #ifdef PF_WANT_32_TO_64_COUNTER
2140 	LIST_INSERT_HEAD(&V_pf_allrulelist, rule, allrulelist);
2141 	MPASS(!rule->allrulelinked);
2142 	rule->allrulelinked = true;
2143 	V_pf_allrulecount++;
2144 #endif
2145 	ruleset = pf_find_kruleset(anchor);
2146 	if (ruleset == NULL)
2147 		ERROUT(EINVAL);
2148 	rs_num = pf_get_ruleset_number(rule->action);
2149 	if (rs_num >= PF_RULESET_MAX)
2150 		ERROUT(EINVAL);
2151 	if (ticket != ruleset->rules[rs_num].inactive.ticket) {
2152 		DPFPRINTF(PF_DEBUG_MISC,
2153 		    ("ticket: %d != [%d]%d\n", ticket, rs_num,
2154 		    ruleset->rules[rs_num].inactive.ticket));
2155 		ERROUT(EBUSY);
2156 	}
2157 	if (pool_ticket != V_ticket_pabuf) {
2158 		DPFPRINTF(PF_DEBUG_MISC,
2159 		    ("pool_ticket: %d != %d\n", pool_ticket,
2160 		    V_ticket_pabuf));
2161 		ERROUT(EBUSY);
2162 	}
2163 	/*
2164 	 * XXXMJG hack: there is no mechanism to ensure they started the
2165 	 * transaction. Ticket checked above may happen to match by accident,
2166 	 * even if nobody called DIOCXBEGIN, let alone this process.
2167 	 * Partially work around it by checking if the RB tree got allocated,
2168 	 * see pf_begin_rules.
2169 	 */
2170 	if (ruleset->rules[rs_num].inactive.tree == NULL) {
2171 		ERROUT(EINVAL);
2172 	}
2173 
2174 	tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
2175 	    pf_krulequeue);
2176 	if (tail)
2177 		rule->nr = tail->nr + 1;
2178 	else
2179 		rule->nr = 0;
2180 	if (rule->ifname[0]) {
2181 		rule->kif = pfi_kkif_attach(kif, rule->ifname);
2182 		kif = NULL;
2183 		pfi_kkif_ref(rule->kif);
2184 	} else
2185 		rule->kif = NULL;
2186 
2187 	if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
2188 		error = EBUSY;
2189 
2190 #ifdef ALTQ
2191 	/* set queue IDs */
2192 	if (rule->qname[0] != 0) {
2193 		if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2194 			error = EBUSY;
2195 		else if (rule->pqname[0] != 0) {
2196 			if ((rule->pqid =
2197 			    pf_qname2qid(rule->pqname)) == 0)
2198 				error = EBUSY;
2199 		} else
2200 			rule->pqid = rule->qid;
2201 	}
2202 #endif
2203 	if (rule->tagname[0])
2204 		if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2205 			error = EBUSY;
2206 	if (rule->match_tagname[0])
2207 		if ((rule->match_tag =
2208 		    pf_tagname2tag(rule->match_tagname)) == 0)
2209 			error = EBUSY;
2210 	if (rule->rt && !rule->direction)
2211 		error = EINVAL;
2212 	if (!rule->log)
2213 		rule->logif = 0;
2214 	if (rule->logif >= PFLOGIFS_MAX)
2215 		error = EINVAL;
2216 	if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
2217 		error = ENOMEM;
2218 	if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
2219 		error = ENOMEM;
2220 	if (pf_kanchor_setup(rule, ruleset, anchor_call))
2221 		error = EINVAL;
2222 	if (rule->scrub_flags & PFSTATE_SETPRIO &&
2223 	    (rule->set_prio[0] > PF_PRIO_MAX ||
2224 	    rule->set_prio[1] > PF_PRIO_MAX))
2225 		error = EINVAL;
2226 	TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
2227 		if (pa->addr.type == PF_ADDR_TABLE) {
2228 			pa->addr.p.tbl = pfr_attach_table(ruleset,
2229 			    pa->addr.v.tblname);
2230 			if (pa->addr.p.tbl == NULL)
2231 				error = ENOMEM;
2232 		}
2233 
2234 	rule->overload_tbl = NULL;
2235 	if (rule->overload_tblname[0]) {
2236 		if ((rule->overload_tbl = pfr_attach_table(ruleset,
2237 		    rule->overload_tblname)) == NULL)
2238 			error = EINVAL;
2239 		else
2240 			rule->overload_tbl->pfrkt_flags |=
2241 			    PFR_TFLAG_ACTIVE;
2242 	}
2243 
2244 	pf_mv_kpool(&V_pf_pabuf, &rule->rpool.list);
2245 	if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
2246 	    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
2247 	    (rule->rt > PF_NOPFROUTE)) &&
2248 	    (TAILQ_FIRST(&rule->rpool.list) == NULL))
2249 		error = EINVAL;
2250 
2251 	if (error) {
2252 		pf_free_rule(rule);
2253 		rule = NULL;
2254 		ERROUT(error);
2255 	}
2256 
2257 	rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
2258 	TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
2259 	    rule, entries);
2260 	ruleset->rules[rs_num].inactive.rcount++;
2261 
2262 	PF_RULES_WUNLOCK();
2263 	pf_hash_rule(rule);
2264 	if (RB_INSERT(pf_krule_global, ruleset->rules[rs_num].inactive.tree, rule) != NULL) {
2265 		PF_RULES_WLOCK();
2266 		TAILQ_REMOVE(ruleset->rules[rs_num].inactive.ptr, rule, entries);
2267 		ruleset->rules[rs_num].inactive.rcount--;
2268 		pf_free_rule(rule);
2269 		rule = NULL;
2270 		ERROUT(EEXIST);
2271 	}
2272 	PF_CONFIG_UNLOCK();
2273 
2274 	return (0);
2275 
2276 #undef ERROUT
2277 errout:
2278 	PF_RULES_WUNLOCK();
2279 	PF_CONFIG_UNLOCK();
2280 errout_unlocked:
2281 	pf_kkif_free(kif);
2282 	pf_krule_free(rule);
2283 	return (error);
2284 }
2285 
2286 static bool
2287 pf_label_match(const struct pf_krule *rule, const char *label)
2288 {
2289 	int i = 0;
2290 
2291 	while (*rule->label[i]) {
2292 		if (strcmp(rule->label[i], label) == 0)
2293 			return (true);
2294 		i++;
2295 	}
2296 
2297 	return (false);
2298 }
2299 
2300 static unsigned int
2301 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir)
2302 {
2303 	struct pf_kstate *s;
2304 	int more = 0;
2305 
2306 	s = pf_find_state_all(key, dir, &more);
2307 	if (s == NULL)
2308 		return (0);
2309 
2310 	if (more) {
2311 		PF_STATE_UNLOCK(s);
2312 		return (0);
2313 	}
2314 
2315 	pf_unlink_state(s);
2316 	return (1);
2317 }
2318 
2319 static int
2320 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih)
2321 {
2322 	struct pf_kstate	*s;
2323 	struct pf_state_key	*sk;
2324 	struct pf_addr		*srcaddr, *dstaddr;
2325 	struct pf_state_key_cmp	 match_key;
2326 	int			 idx, killed = 0;
2327 	unsigned int		 dir;
2328 	u_int16_t		 srcport, dstport;
2329 	struct pfi_kkif		*kif;
2330 
2331 relock_DIOCKILLSTATES:
2332 	PF_HASHROW_LOCK(ih);
2333 	LIST_FOREACH(s, &ih->states, entry) {
2334 		/* For floating states look at the original kif. */
2335 		kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
2336 
2337 		sk = s->key[PF_SK_WIRE];
2338 		if (s->direction == PF_OUT) {
2339 			srcaddr = &sk->addr[1];
2340 			dstaddr = &sk->addr[0];
2341 			srcport = sk->port[1];
2342 			dstport = sk->port[0];
2343 		} else {
2344 			srcaddr = &sk->addr[0];
2345 			dstaddr = &sk->addr[1];
2346 			srcport = sk->port[0];
2347 			dstport = sk->port[1];
2348 		}
2349 
2350 		if (psk->psk_af && sk->af != psk->psk_af)
2351 			continue;
2352 
2353 		if (psk->psk_proto && psk->psk_proto != sk->proto)
2354 			continue;
2355 
2356 		if (! PF_MATCHA(psk->psk_src.neg, &psk->psk_src.addr.v.a.addr,
2357 		    &psk->psk_src.addr.v.a.mask, srcaddr, sk->af))
2358 			continue;
2359 
2360 		if (! PF_MATCHA(psk->psk_dst.neg, &psk->psk_dst.addr.v.a.addr,
2361 		    &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af))
2362 			continue;
2363 
2364 		if (!  PF_MATCHA(psk->psk_rt_addr.neg,
2365 		    &psk->psk_rt_addr.addr.v.a.addr,
2366 		    &psk->psk_rt_addr.addr.v.a.mask,
2367 		    &s->rt_addr, sk->af))
2368 			continue;
2369 
2370 		if (psk->psk_src.port_op != 0 &&
2371 		    ! pf_match_port(psk->psk_src.port_op,
2372 		    psk->psk_src.port[0], psk->psk_src.port[1], srcport))
2373 			continue;
2374 
2375 		if (psk->psk_dst.port_op != 0 &&
2376 		    ! pf_match_port(psk->psk_dst.port_op,
2377 		    psk->psk_dst.port[0], psk->psk_dst.port[1], dstport))
2378 			continue;
2379 
2380 		if (psk->psk_label[0] &&
2381 		    ! pf_label_match(s->rule.ptr, psk->psk_label))
2382 			continue;
2383 
2384 		if (psk->psk_ifname[0] && strcmp(psk->psk_ifname,
2385 		    kif->pfik_name))
2386 			continue;
2387 
2388 		if (psk->psk_kill_match) {
2389 			/* Create the key to find matching states, with lock
2390 			 * held. */
2391 
2392 			bzero(&match_key, sizeof(match_key));
2393 
2394 			if (s->direction == PF_OUT) {
2395 				dir = PF_IN;
2396 				idx = PF_SK_STACK;
2397 			} else {
2398 				dir = PF_OUT;
2399 				idx = PF_SK_WIRE;
2400 			}
2401 
2402 			match_key.af = s->key[idx]->af;
2403 			match_key.proto = s->key[idx]->proto;
2404 			PF_ACPY(&match_key.addr[0],
2405 			    &s->key[idx]->addr[1], match_key.af);
2406 			match_key.port[0] = s->key[idx]->port[1];
2407 			PF_ACPY(&match_key.addr[1],
2408 			    &s->key[idx]->addr[0], match_key.af);
2409 			match_key.port[1] = s->key[idx]->port[0];
2410 		}
2411 
2412 		pf_unlink_state(s);
2413 		killed++;
2414 
2415 		if (psk->psk_kill_match)
2416 			killed += pf_kill_matching_state(&match_key, dir);
2417 
2418 		goto relock_DIOCKILLSTATES;
2419 	}
2420 	PF_HASHROW_UNLOCK(ih);
2421 
2422 	return (killed);
2423 }
2424 
2425 static int
2426 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
2427 {
2428 	int			 error = 0;
2429 	PF_RULES_RLOCK_TRACKER;
2430 
2431 #define	ERROUT_IOCTL(target, x)					\
2432     do {								\
2433 	    error = (x);						\
2434 	    SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__);	\
2435 	    goto target;						\
2436     } while (0)
2437 
2438 
2439 	/* XXX keep in sync with switch() below */
2440 	if (securelevel_gt(td->td_ucred, 2))
2441 		switch (cmd) {
2442 		case DIOCGETRULES:
2443 		case DIOCGETRULE:
2444 		case DIOCGETRULENV:
2445 		case DIOCGETADDRS:
2446 		case DIOCGETADDR:
2447 		case DIOCGETSTATE:
2448 		case DIOCGETSTATENV:
2449 		case DIOCSETSTATUSIF:
2450 		case DIOCGETSTATUS:
2451 		case DIOCGETSTATUSNV:
2452 		case DIOCCLRSTATUS:
2453 		case DIOCNATLOOK:
2454 		case DIOCSETDEBUG:
2455 		case DIOCGETSTATES:
2456 		case DIOCGETSTATESV2:
2457 		case DIOCGETTIMEOUT:
2458 		case DIOCCLRRULECTRS:
2459 		case DIOCGETLIMIT:
2460 		case DIOCGETALTQSV0:
2461 		case DIOCGETALTQSV1:
2462 		case DIOCGETALTQV0:
2463 		case DIOCGETALTQV1:
2464 		case DIOCGETQSTATSV0:
2465 		case DIOCGETQSTATSV1:
2466 		case DIOCGETRULESETS:
2467 		case DIOCGETRULESET:
2468 		case DIOCRGETTABLES:
2469 		case DIOCRGETTSTATS:
2470 		case DIOCRCLRTSTATS:
2471 		case DIOCRCLRADDRS:
2472 		case DIOCRADDADDRS:
2473 		case DIOCRDELADDRS:
2474 		case DIOCRSETADDRS:
2475 		case DIOCRGETADDRS:
2476 		case DIOCRGETASTATS:
2477 		case DIOCRCLRASTATS:
2478 		case DIOCRTSTADDRS:
2479 		case DIOCOSFPGET:
2480 		case DIOCGETSRCNODES:
2481 		case DIOCCLRSRCNODES:
2482 		case DIOCGETSYNCOOKIES:
2483 		case DIOCIGETIFACES:
2484 		case DIOCGIFSPEEDV0:
2485 		case DIOCGIFSPEEDV1:
2486 		case DIOCSETIFFLAG:
2487 		case DIOCCLRIFFLAG:
2488 		case DIOCGETETHRULES:
2489 		case DIOCGETETHRULE:
2490 		case DIOCGETETHRULESETS:
2491 		case DIOCGETETHRULESET:
2492 			break;
2493 		case DIOCRCLRTABLES:
2494 		case DIOCRADDTABLES:
2495 		case DIOCRDELTABLES:
2496 		case DIOCRSETTFLAGS:
2497 			if (((struct pfioc_table *)addr)->pfrio_flags &
2498 			    PFR_FLAG_DUMMY)
2499 				break; /* dummy operation ok */
2500 			return (EPERM);
2501 		default:
2502 			return (EPERM);
2503 		}
2504 
2505 	if (!(flags & FWRITE))
2506 		switch (cmd) {
2507 		case DIOCGETRULES:
2508 		case DIOCGETADDRS:
2509 		case DIOCGETADDR:
2510 		case DIOCGETSTATE:
2511 		case DIOCGETSTATENV:
2512 		case DIOCGETSTATUS:
2513 		case DIOCGETSTATUSNV:
2514 		case DIOCGETSTATES:
2515 		case DIOCGETSTATESV2:
2516 		case DIOCGETTIMEOUT:
2517 		case DIOCGETLIMIT:
2518 		case DIOCGETALTQSV0:
2519 		case DIOCGETALTQSV1:
2520 		case DIOCGETALTQV0:
2521 		case DIOCGETALTQV1:
2522 		case DIOCGETQSTATSV0:
2523 		case DIOCGETQSTATSV1:
2524 		case DIOCGETRULESETS:
2525 		case DIOCGETRULESET:
2526 		case DIOCNATLOOK:
2527 		case DIOCRGETTABLES:
2528 		case DIOCRGETTSTATS:
2529 		case DIOCRGETADDRS:
2530 		case DIOCRGETASTATS:
2531 		case DIOCRTSTADDRS:
2532 		case DIOCOSFPGET:
2533 		case DIOCGETSRCNODES:
2534 		case DIOCGETSYNCOOKIES:
2535 		case DIOCIGETIFACES:
2536 		case DIOCGIFSPEEDV1:
2537 		case DIOCGIFSPEEDV0:
2538 		case DIOCGETRULENV:
2539 		case DIOCGETETHRULES:
2540 		case DIOCGETETHRULE:
2541 		case DIOCGETETHRULESETS:
2542 		case DIOCGETETHRULESET:
2543 			break;
2544 		case DIOCRCLRTABLES:
2545 		case DIOCRADDTABLES:
2546 		case DIOCRDELTABLES:
2547 		case DIOCRCLRTSTATS:
2548 		case DIOCRCLRADDRS:
2549 		case DIOCRADDADDRS:
2550 		case DIOCRDELADDRS:
2551 		case DIOCRSETADDRS:
2552 		case DIOCRSETTFLAGS:
2553 			if (((struct pfioc_table *)addr)->pfrio_flags &
2554 			    PFR_FLAG_DUMMY) {
2555 				flags |= FWRITE; /* need write lock for dummy */
2556 				break; /* dummy operation ok */
2557 			}
2558 			return (EACCES);
2559 		case DIOCGETRULE:
2560 			if (((struct pfioc_rule *)addr)->action ==
2561 			    PF_GET_CLR_CNTR)
2562 				return (EACCES);
2563 			break;
2564 		default:
2565 			return (EACCES);
2566 		}
2567 
2568 	CURVNET_SET(TD_TO_VNET(td));
2569 
2570 	switch (cmd) {
2571 	case DIOCSTART:
2572 		sx_xlock(&pf_ioctl_lock);
2573 		if (V_pf_status.running)
2574 			error = EEXIST;
2575 		else {
2576 			int cpu;
2577 
2578 			hook_pf();
2579 			if (! TAILQ_EMPTY(V_pf_keth->active.rules))
2580 				hook_pf_eth();
2581 			V_pf_status.running = 1;
2582 			V_pf_status.since = time_second;
2583 
2584 			CPU_FOREACH(cpu)
2585 				V_pf_stateid[cpu] = time_second;
2586 
2587 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
2588 		}
2589 		break;
2590 
2591 	case DIOCSTOP:
2592 		sx_xlock(&pf_ioctl_lock);
2593 		if (!V_pf_status.running)
2594 			error = ENOENT;
2595 		else {
2596 			V_pf_status.running = 0;
2597 			dehook_pf();
2598 			dehook_pf_eth();
2599 			V_pf_status.since = time_second;
2600 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
2601 		}
2602 		break;
2603 
2604 	case DIOCGETETHRULES: {
2605 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2606 		nvlist_t		*nvl;
2607 		void			*packed;
2608 		struct pf_keth_rule	*tail;
2609 		struct pf_keth_ruleset	*rs;
2610 		u_int32_t		 ticket, nr;
2611 		const char		*anchor = "";
2612 
2613 		nvl = NULL;
2614 		packed = NULL;
2615 
2616 #define	ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULES_error, x)
2617 
2618 		if (nv->len > pf_ioctl_maxcount)
2619 			ERROUT(ENOMEM);
2620 
2621 		/* Copy the request in */
2622 		packed = malloc(nv->len, M_NVLIST, M_WAITOK);
2623 		if (packed == NULL)
2624 			ERROUT(ENOMEM);
2625 
2626 		error = copyin(nv->data, packed, nv->len);
2627 		if (error)
2628 			ERROUT(error);
2629 
2630 		nvl = nvlist_unpack(packed, nv->len, 0);
2631 		if (nvl == NULL)
2632 			ERROUT(EBADMSG);
2633 
2634 		if (! nvlist_exists_string(nvl, "anchor"))
2635 			ERROUT(EBADMSG);
2636 
2637 		anchor = nvlist_get_string(nvl, "anchor");
2638 
2639 		rs = pf_find_keth_ruleset(anchor);
2640 
2641 		nvlist_destroy(nvl);
2642 		nvl = NULL;
2643 		free(packed, M_NVLIST);
2644 		packed = NULL;
2645 
2646 		if (rs == NULL)
2647 			ERROUT(ENOENT);
2648 
2649 		/* Reply */
2650 		nvl = nvlist_create(0);
2651 		if (nvl == NULL)
2652 			ERROUT(ENOMEM);
2653 
2654 		PF_RULES_RLOCK();
2655 
2656 		ticket = rs->active.ticket;
2657 		tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq);
2658 		if (tail)
2659 			nr = tail->nr + 1;
2660 		else
2661 			nr = 0;
2662 
2663 		PF_RULES_RUNLOCK();
2664 
2665 		nvlist_add_number(nvl, "ticket", ticket);
2666 		nvlist_add_number(nvl, "nr", nr);
2667 
2668 		packed = nvlist_pack(nvl, &nv->len);
2669 		if (packed == NULL)
2670 			ERROUT(ENOMEM);
2671 
2672 		if (nv->size == 0)
2673 			ERROUT(0);
2674 		else if (nv->size < nv->len)
2675 			ERROUT(ENOSPC);
2676 
2677 		error = copyout(packed, nv->data, nv->len);
2678 
2679 #undef ERROUT
2680 DIOCGETETHRULES_error:
2681 		free(packed, M_NVLIST);
2682 		nvlist_destroy(nvl);
2683 		break;
2684 	}
2685 
2686 	case DIOCGETETHRULE: {
2687 		struct epoch_tracker	 et;
2688 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2689 		nvlist_t		*nvl = NULL;
2690 		void			*nvlpacked = NULL;
2691 		struct pf_keth_rule	*rule = NULL;
2692 		struct pf_keth_ruleset	*rs;
2693 		u_int32_t		 ticket, nr;
2694 		bool			 clear = false;
2695 		const char		*anchor;
2696 
2697 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULE_error, x)
2698 
2699 		if (nv->len > pf_ioctl_maxcount)
2700 			ERROUT(ENOMEM);
2701 
2702 		nvlpacked = malloc(nv->len, M_TEMP, M_WAITOK);
2703 		if (nvlpacked == NULL)
2704 			ERROUT(ENOMEM);
2705 
2706 		error = copyin(nv->data, nvlpacked, nv->len);
2707 		if (error)
2708 			ERROUT(error);
2709 
2710 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2711 		if (nvl == NULL)
2712 			ERROUT(EBADMSG);
2713 		if (! nvlist_exists_number(nvl, "ticket"))
2714 			ERROUT(EBADMSG);
2715 		ticket = nvlist_get_number(nvl, "ticket");
2716 		if (! nvlist_exists_string(nvl, "anchor"))
2717 			ERROUT(EBADMSG);
2718 		anchor = nvlist_get_string(nvl, "anchor");
2719 
2720 		if (nvlist_exists_bool(nvl, "clear"))
2721 			clear = nvlist_get_bool(nvl, "clear");
2722 
2723 		if (clear && !(flags & FWRITE))
2724 			ERROUT(EACCES);
2725 
2726 		if (! nvlist_exists_number(nvl, "nr"))
2727 			ERROUT(EBADMSG);
2728 		nr = nvlist_get_number(nvl, "nr");
2729 
2730 		PF_RULES_RLOCK();
2731 		rs = pf_find_keth_ruleset(anchor);
2732 		if (rs == NULL) {
2733 			PF_RULES_RUNLOCK();
2734 			ERROUT(ENOENT);
2735 		}
2736 		if (ticket != rs->active.ticket) {
2737 			PF_RULES_RUNLOCK();
2738 			ERROUT(EBUSY);
2739 		}
2740 
2741 		nvlist_destroy(nvl);
2742 		nvl = NULL;
2743 		free(nvlpacked, M_TEMP);
2744 		nvlpacked = NULL;
2745 
2746 		rule = TAILQ_FIRST(rs->active.rules);
2747 		while ((rule != NULL) && (rule->nr != nr))
2748 			rule = TAILQ_NEXT(rule, entries);
2749 		if (rule == NULL) {
2750 			PF_RULES_RUNLOCK();
2751 			ERROUT(ENOENT);
2752 		}
2753 		/* Make sure rule can't go away. */
2754 		NET_EPOCH_ENTER(et);
2755 		PF_RULES_RUNLOCK();
2756 		nvl = pf_keth_rule_to_nveth_rule(rule);
2757 		if (pf_keth_anchor_nvcopyout(rs, rule, nvl))
2758 			ERROUT(EBUSY);
2759 		NET_EPOCH_EXIT(et);
2760 		if (nvl == NULL)
2761 			ERROUT(ENOMEM);
2762 
2763 		nvlpacked = nvlist_pack(nvl, &nv->len);
2764 		if (nvlpacked == NULL)
2765 			ERROUT(ENOMEM);
2766 
2767 		if (nv->size == 0)
2768 			ERROUT(0);
2769 		else if (nv->size < nv->len)
2770 			ERROUT(ENOSPC);
2771 
2772 		error = copyout(nvlpacked, nv->data, nv->len);
2773 		if (error == 0 && clear) {
2774 			counter_u64_zero(rule->evaluations);
2775 			for (int i = 0; i < 2; i++) {
2776 				counter_u64_zero(rule->packets[i]);
2777 				counter_u64_zero(rule->bytes[i]);
2778 			}
2779 		}
2780 
2781 #undef ERROUT
2782 DIOCGETETHRULE_error:
2783 		free(nvlpacked, M_TEMP);
2784 		nvlist_destroy(nvl);
2785 		break;
2786 	}
2787 
2788 	case DIOCADDETHRULE: {
2789 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2790 		nvlist_t		*nvl = NULL;
2791 		void			*nvlpacked = NULL;
2792 		struct pf_keth_rule	*rule = NULL, *tail = NULL;
2793 		struct pf_keth_ruleset	*ruleset = NULL;
2794 		struct pfi_kkif		*kif = NULL;
2795 		const char		*anchor = "", *anchor_call = "";
2796 
2797 #define ERROUT(x)	ERROUT_IOCTL(DIOCADDETHRULE_error, x)
2798 
2799 		nvlpacked = malloc(nv->len, M_TEMP, M_WAITOK);
2800 		if (nvlpacked == NULL)
2801 			ERROUT(ENOMEM);
2802 
2803 		error = copyin(nv->data, nvlpacked, nv->len);
2804 		if (error)
2805 			ERROUT(error);
2806 
2807 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2808 		if (nvl == NULL)
2809 			ERROUT(EBADMSG);
2810 
2811 		if (! nvlist_exists_number(nvl, "ticket"))
2812 			ERROUT(EBADMSG);
2813 
2814 		if (nvlist_exists_string(nvl, "anchor"))
2815 			anchor = nvlist_get_string(nvl, "anchor");
2816 		if (nvlist_exists_string(nvl, "anchor_call"))
2817 			anchor_call = nvlist_get_string(nvl, "anchor_call");
2818 
2819 		ruleset = pf_find_keth_ruleset(anchor);
2820 		if (ruleset == NULL)
2821 			ERROUT(EINVAL);
2822 
2823 		if (nvlist_get_number(nvl, "ticket") !=
2824 		    ruleset->inactive.ticket) {
2825 			DPFPRINTF(PF_DEBUG_MISC,
2826 			    ("ticket: %d != %d\n",
2827 			    (u_int32_t)nvlist_get_number(nvl, "ticket"),
2828 			    ruleset->inactive.ticket));
2829 			ERROUT(EBUSY);
2830 		}
2831 
2832 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
2833 		if (rule == NULL)
2834 			ERROUT(ENOMEM);
2835 
2836 		error = pf_nveth_rule_to_keth_rule(nvl, rule);
2837 		if (error != 0)
2838 			ERROUT(error);
2839 
2840 		if (rule->ifname[0])
2841 			kif = pf_kkif_create(M_WAITOK);
2842 		rule->evaluations = counter_u64_alloc(M_WAITOK);
2843 		for (int i = 0; i < 2; i++) {
2844 			rule->packets[i] = counter_u64_alloc(M_WAITOK);
2845 			rule->bytes[i] = counter_u64_alloc(M_WAITOK);
2846 		}
2847 
2848 		PF_RULES_WLOCK();
2849 
2850 		if (rule->ifname[0]) {
2851 			rule->kif = pfi_kkif_attach(kif, rule->ifname);
2852 			pfi_kkif_ref(rule->kif);
2853 		} else
2854 			rule->kif = NULL;
2855 
2856 #ifdef ALTQ
2857 		/* set queue IDs */
2858 		if (rule->qname[0] != 0) {
2859 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2860 				error = EBUSY;
2861 			else
2862 				rule->qid = rule->qid;
2863 		}
2864 #endif
2865 		if (rule->tagname[0])
2866 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2867 				error = EBUSY;
2868 
2869 		if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE)
2870 			error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr);
2871 		if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE)
2872 			error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr);
2873 
2874 		if (error) {
2875 			pf_free_eth_rule(rule);
2876 			PF_RULES_WUNLOCK();
2877 			ERROUT(error);
2878 		}
2879 
2880 		if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) {
2881 			pf_free_eth_rule(rule);
2882 			PF_RULES_WUNLOCK();
2883 			ERROUT(EINVAL);
2884 		}
2885 
2886 		tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq);
2887 		if (tail)
2888 			rule->nr = tail->nr + 1;
2889 		else
2890 			rule->nr = 0;
2891 
2892 		TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries);
2893 
2894 		PF_RULES_WUNLOCK();
2895 
2896 #undef ERROUT
2897 DIOCADDETHRULE_error:
2898 		nvlist_destroy(nvl);
2899 		free(nvlpacked, M_TEMP);
2900 		break;
2901 	}
2902 
2903 	case DIOCGETETHRULESETS: {
2904 		struct epoch_tracker	 et;
2905 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2906 		nvlist_t		*nvl = NULL;
2907 		void			*nvlpacked = NULL;
2908 		struct pf_keth_ruleset	*ruleset;
2909 		struct pf_keth_anchor	*anchor;
2910 		int			 nr = 0;
2911 
2912 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULESETS_error, x)
2913 
2914 		if (nv->len > pf_ioctl_maxcount)
2915 			ERROUT(ENOMEM);
2916 
2917 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2918 		if (nvlpacked == NULL)
2919 			ERROUT(ENOMEM);
2920 
2921 		error = copyin(nv->data, nvlpacked, nv->len);
2922 		if (error)
2923 			ERROUT(error);
2924 
2925 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2926 		if (nvl == NULL)
2927 			ERROUT(EBADMSG);
2928 		if (! nvlist_exists_string(nvl, "path"))
2929 			ERROUT(EBADMSG);
2930 
2931 		NET_EPOCH_ENTER(et);
2932 
2933 		if ((ruleset = pf_find_keth_ruleset(
2934 		    nvlist_get_string(nvl, "path"))) == NULL) {
2935 			NET_EPOCH_EXIT(et);
2936 			ERROUT(ENOENT);
2937 		}
2938 
2939 		if (ruleset->anchor == NULL) {
2940 			RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors)
2941 				if (anchor->parent == NULL)
2942 					nr++;
2943 		} else {
2944 			RB_FOREACH(anchor, pf_keth_anchor_node,
2945 			    &ruleset->anchor->children)
2946 				nr++;
2947 		}
2948 
2949 		NET_EPOCH_EXIT(et);
2950 
2951 		nvlist_destroy(nvl);
2952 		nvl = NULL;
2953 		free(nvlpacked, M_NVLIST);
2954 		nvlpacked = NULL;
2955 
2956 		nvl = nvlist_create(0);
2957 		if (nvl == NULL)
2958 			ERROUT(ENOMEM);
2959 
2960 		nvlist_add_number(nvl, "nr", nr);
2961 
2962 		nvlpacked = nvlist_pack(nvl, &nv->len);
2963 		if (nvlpacked == NULL)
2964 			ERROUT(ENOMEM);
2965 
2966 		if (nv->size == 0)
2967 			ERROUT(0);
2968 		else if (nv->size < nv->len)
2969 			ERROUT(ENOSPC);
2970 
2971 		error = copyout(nvlpacked, nv->data, nv->len);
2972 
2973 #undef ERROUT
2974 DIOCGETETHRULESETS_error:
2975 		free(nvlpacked, M_NVLIST);
2976 		nvlist_destroy(nvl);
2977 		break;
2978 	}
2979 
2980 	case DIOCGETETHRULESET: {
2981 		struct epoch_tracker	 et;
2982 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2983 		nvlist_t		*nvl = NULL;
2984 		void			*nvlpacked = NULL;
2985 		struct pf_keth_ruleset	*ruleset;
2986 		struct pf_keth_anchor	*anchor;
2987 		int			 nr = 0, req_nr = 0;
2988 		bool			 found = false;
2989 
2990 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULESET_error, x)
2991 
2992 		if (nv->len > pf_ioctl_maxcount)
2993 			ERROUT(ENOMEM);
2994 
2995 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2996 		if (nvlpacked == NULL)
2997 			ERROUT(ENOMEM);
2998 
2999 		error = copyin(nv->data, nvlpacked, nv->len);
3000 		if (error)
3001 			ERROUT(error);
3002 
3003 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3004 		if (nvl == NULL)
3005 			ERROUT(EBADMSG);
3006 		if (! nvlist_exists_string(nvl, "path"))
3007 			ERROUT(EBADMSG);
3008 		if (! nvlist_exists_number(nvl, "nr"))
3009 			ERROUT(EBADMSG);
3010 
3011 		req_nr = nvlist_get_number(nvl, "nr");
3012 
3013 		NET_EPOCH_ENTER(et);
3014 
3015 		if ((ruleset = pf_find_keth_ruleset(
3016 		    nvlist_get_string(nvl, "path"))) == NULL) {
3017 			NET_EPOCH_EXIT(et);
3018 			ERROUT(ENOENT);
3019 		}
3020 
3021 		nvlist_destroy(nvl);
3022 		nvl = NULL;
3023 		free(nvlpacked, M_NVLIST);
3024 		nvlpacked = NULL;
3025 
3026 		nvl = nvlist_create(0);
3027 		if (nvl == NULL) {
3028 			NET_EPOCH_EXIT(et);
3029 			ERROUT(ENOMEM);
3030 		}
3031 
3032 		if (ruleset->anchor == NULL) {
3033 			RB_FOREACH(anchor, pf_keth_anchor_global,
3034 			    &V_pf_keth_anchors) {
3035 				if (anchor->parent == NULL && nr++ == req_nr) {
3036 					found = true;
3037 					break;
3038 				}
3039 			}
3040 		} else {
3041 			RB_FOREACH(anchor, pf_keth_anchor_node,
3042 			     &ruleset->anchor->children) {
3043 				if (nr++ == req_nr) {
3044 					found = true;
3045 					break;
3046 				}
3047 			}
3048 		}
3049 
3050 		NET_EPOCH_EXIT(et);
3051 		if (found) {
3052 			nvlist_add_number(nvl, "nr", nr);
3053 			nvlist_add_string(nvl, "name", anchor->name);
3054 			if (ruleset->anchor)
3055 				nvlist_add_string(nvl, "path",
3056 				    ruleset->anchor->path);
3057 			else
3058 				nvlist_add_string(nvl, "path", "");
3059 		} else {
3060 			ERROUT(EBUSY);
3061 		}
3062 
3063 		nvlpacked = nvlist_pack(nvl, &nv->len);
3064 		if (nvlpacked == NULL)
3065 			ERROUT(ENOMEM);
3066 
3067 		if (nv->size == 0)
3068 			ERROUT(0);
3069 		else if (nv->size < nv->len)
3070 			ERROUT(ENOSPC);
3071 
3072 		error = copyout(nvlpacked, nv->data, nv->len);
3073 
3074 #undef ERROUT
3075 DIOCGETETHRULESET_error:
3076 		free(nvlpacked, M_NVLIST);
3077 		nvlist_destroy(nvl);
3078 		break;
3079 	}
3080 
3081 	case DIOCADDRULENV: {
3082 		struct pfioc_nv	*nv = (struct pfioc_nv *)addr;
3083 		nvlist_t	*nvl = NULL;
3084 		void		*nvlpacked = NULL;
3085 		struct pf_krule	*rule = NULL;
3086 		const char	*anchor = "", *anchor_call = "";
3087 		uint32_t	 ticket = 0, pool_ticket = 0;
3088 
3089 #define	ERROUT(x)	ERROUT_IOCTL(DIOCADDRULENV_error, x)
3090 
3091 		if (nv->len > pf_ioctl_maxcount)
3092 			ERROUT(ENOMEM);
3093 
3094 		nvlpacked = malloc(nv->len, M_TEMP, M_WAITOK);
3095 		error = copyin(nv->data, nvlpacked, nv->len);
3096 		if (error)
3097 			ERROUT(error);
3098 
3099 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3100 		if (nvl == NULL)
3101 			ERROUT(EBADMSG);
3102 
3103 		if (! nvlist_exists_number(nvl, "ticket"))
3104 			ERROUT(EINVAL);
3105 		ticket = nvlist_get_number(nvl, "ticket");
3106 
3107 		if (! nvlist_exists_number(nvl, "pool_ticket"))
3108 			ERROUT(EINVAL);
3109 		pool_ticket = nvlist_get_number(nvl, "pool_ticket");
3110 
3111 		if (! nvlist_exists_nvlist(nvl, "rule"))
3112 			ERROUT(EINVAL);
3113 
3114 		rule = pf_krule_alloc();
3115 		error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"),
3116 		    rule);
3117 		if (error)
3118 			ERROUT(error);
3119 
3120 		if (nvlist_exists_string(nvl, "anchor"))
3121 			anchor = nvlist_get_string(nvl, "anchor");
3122 		if (nvlist_exists_string(nvl, "anchor_call"))
3123 			anchor_call = nvlist_get_string(nvl, "anchor_call");
3124 
3125 		if ((error = nvlist_error(nvl)))
3126 			ERROUT(error);
3127 
3128 		/* Frees rule on error */
3129 		error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor,
3130 		    anchor_call, td);
3131 
3132 		nvlist_destroy(nvl);
3133 		free(nvlpacked, M_TEMP);
3134 		break;
3135 #undef ERROUT
3136 DIOCADDRULENV_error:
3137 		pf_krule_free(rule);
3138 		nvlist_destroy(nvl);
3139 		free(nvlpacked, M_TEMP);
3140 
3141 		break;
3142 	}
3143 	case DIOCADDRULE: {
3144 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
3145 		struct pf_krule		*rule;
3146 
3147 		rule = pf_krule_alloc();
3148 		error = pf_rule_to_krule(&pr->rule, rule);
3149 		if (error != 0) {
3150 			pf_krule_free(rule);
3151 			break;
3152 		}
3153 
3154 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
3155 
3156 		/* Frees rule on error */
3157 		error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket,
3158 		    pr->anchor, pr->anchor_call, td);
3159 		break;
3160 	}
3161 
3162 	case DIOCGETRULES: {
3163 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
3164 		struct pf_kruleset	*ruleset;
3165 		struct pf_krule		*tail;
3166 		int			 rs_num;
3167 
3168 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
3169 
3170 		PF_RULES_WLOCK();
3171 		ruleset = pf_find_kruleset(pr->anchor);
3172 		if (ruleset == NULL) {
3173 			PF_RULES_WUNLOCK();
3174 			error = EINVAL;
3175 			break;
3176 		}
3177 		rs_num = pf_get_ruleset_number(pr->rule.action);
3178 		if (rs_num >= PF_RULESET_MAX) {
3179 			PF_RULES_WUNLOCK();
3180 			error = EINVAL;
3181 			break;
3182 		}
3183 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
3184 		    pf_krulequeue);
3185 		if (tail)
3186 			pr->nr = tail->nr + 1;
3187 		else
3188 			pr->nr = 0;
3189 		pr->ticket = ruleset->rules[rs_num].active.ticket;
3190 		PF_RULES_WUNLOCK();
3191 		break;
3192 	}
3193 
3194 	case DIOCGETRULE: {
3195 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
3196 		struct pf_kruleset	*ruleset;
3197 		struct pf_krule		*rule;
3198 		int			 rs_num;
3199 
3200 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
3201 
3202 		PF_RULES_WLOCK();
3203 		ruleset = pf_find_kruleset(pr->anchor);
3204 		if (ruleset == NULL) {
3205 			PF_RULES_WUNLOCK();
3206 			error = EINVAL;
3207 			break;
3208 		}
3209 		rs_num = pf_get_ruleset_number(pr->rule.action);
3210 		if (rs_num >= PF_RULESET_MAX) {
3211 			PF_RULES_WUNLOCK();
3212 			error = EINVAL;
3213 			break;
3214 		}
3215 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
3216 			PF_RULES_WUNLOCK();
3217 			error = EBUSY;
3218 			break;
3219 		}
3220 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
3221 		while ((rule != NULL) && (rule->nr != pr->nr))
3222 			rule = TAILQ_NEXT(rule, entries);
3223 		if (rule == NULL) {
3224 			PF_RULES_WUNLOCK();
3225 			error = EBUSY;
3226 			break;
3227 		}
3228 
3229 		pf_krule_to_rule(rule, &pr->rule);
3230 
3231 		if (pf_kanchor_copyout(ruleset, rule, pr)) {
3232 			PF_RULES_WUNLOCK();
3233 			error = EBUSY;
3234 			break;
3235 		}
3236 		pf_addr_copyout(&pr->rule.src.addr);
3237 		pf_addr_copyout(&pr->rule.dst.addr);
3238 
3239 		if (pr->action == PF_GET_CLR_CNTR) {
3240 			pf_counter_u64_zero(&rule->evaluations);
3241 			for (int i = 0; i < 2; i++) {
3242 				pf_counter_u64_zero(&rule->packets[i]);
3243 				pf_counter_u64_zero(&rule->bytes[i]);
3244 			}
3245 			counter_u64_zero(rule->states_tot);
3246 		}
3247 		PF_RULES_WUNLOCK();
3248 		break;
3249 	}
3250 
3251 	case DIOCGETRULENV: {
3252 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
3253 		nvlist_t		*nvrule = NULL;
3254 		nvlist_t		*nvl = NULL;
3255 		struct pf_kruleset	*ruleset;
3256 		struct pf_krule		*rule;
3257 		void			*nvlpacked = NULL;
3258 		int			 rs_num, nr;
3259 		bool			 clear_counter = false;
3260 
3261 #define	ERROUT(x)	ERROUT_IOCTL(DIOCGETRULENV_error, x)
3262 
3263 		if (nv->len > pf_ioctl_maxcount)
3264 			ERROUT(ENOMEM);
3265 
3266 		/* Copy the request in */
3267 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3268 		if (nvlpacked == NULL)
3269 			ERROUT(ENOMEM);
3270 
3271 		error = copyin(nv->data, nvlpacked, nv->len);
3272 		if (error)
3273 			ERROUT(error);
3274 
3275 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3276 		if (nvl == NULL)
3277 			ERROUT(EBADMSG);
3278 
3279 		if (! nvlist_exists_string(nvl, "anchor"))
3280 			ERROUT(EBADMSG);
3281 		if (! nvlist_exists_number(nvl, "ruleset"))
3282 			ERROUT(EBADMSG);
3283 		if (! nvlist_exists_number(nvl, "ticket"))
3284 			ERROUT(EBADMSG);
3285 		if (! nvlist_exists_number(nvl, "nr"))
3286 			ERROUT(EBADMSG);
3287 
3288 		if (nvlist_exists_bool(nvl, "clear_counter"))
3289 			clear_counter = nvlist_get_bool(nvl, "clear_counter");
3290 
3291 		if (clear_counter && !(flags & FWRITE))
3292 			ERROUT(EACCES);
3293 
3294 		nr = nvlist_get_number(nvl, "nr");
3295 
3296 		PF_RULES_WLOCK();
3297 		ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor"));
3298 		if (ruleset == NULL) {
3299 			PF_RULES_WUNLOCK();
3300 			ERROUT(ENOENT);
3301 		}
3302 
3303 		rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset"));
3304 		if (rs_num >= PF_RULESET_MAX) {
3305 			PF_RULES_WUNLOCK();
3306 			ERROUT(EINVAL);
3307 		}
3308 
3309 		if (nvlist_get_number(nvl, "ticket") !=
3310 		    ruleset->rules[rs_num].active.ticket) {
3311 			PF_RULES_WUNLOCK();
3312 			ERROUT(EBUSY);
3313 		}
3314 
3315 		if ((error = nvlist_error(nvl))) {
3316 			PF_RULES_WUNLOCK();
3317 			ERROUT(error);
3318 		}
3319 
3320 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
3321 		while ((rule != NULL) && (rule->nr != nr))
3322 			rule = TAILQ_NEXT(rule, entries);
3323 		if (rule == NULL) {
3324 			PF_RULES_WUNLOCK();
3325 			ERROUT(EBUSY);
3326 		}
3327 
3328 		nvrule = pf_krule_to_nvrule(rule);
3329 
3330 		nvlist_destroy(nvl);
3331 		nvl = nvlist_create(0);
3332 		if (nvl == NULL) {
3333 			PF_RULES_WUNLOCK();
3334 			ERROUT(ENOMEM);
3335 		}
3336 		nvlist_add_number(nvl, "nr", nr);
3337 		nvlist_add_nvlist(nvl, "rule", nvrule);
3338 		nvlist_destroy(nvrule);
3339 		nvrule = NULL;
3340 		if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) {
3341 			PF_RULES_WUNLOCK();
3342 			ERROUT(EBUSY);
3343 		}
3344 
3345 		free(nvlpacked, M_NVLIST);
3346 		nvlpacked = nvlist_pack(nvl, &nv->len);
3347 		if (nvlpacked == NULL) {
3348 			PF_RULES_WUNLOCK();
3349 			ERROUT(ENOMEM);
3350 		}
3351 
3352 		if (nv->size == 0) {
3353 			PF_RULES_WUNLOCK();
3354 			ERROUT(0);
3355 		}
3356 		else if (nv->size < nv->len) {
3357 			PF_RULES_WUNLOCK();
3358 			ERROUT(ENOSPC);
3359 		}
3360 
3361 		if (clear_counter) {
3362 			pf_counter_u64_zero(&rule->evaluations);
3363 			for (int i = 0; i < 2; i++) {
3364 				pf_counter_u64_zero(&rule->packets[i]);
3365 				pf_counter_u64_zero(&rule->bytes[i]);
3366 			}
3367 			counter_u64_zero(rule->states_tot);
3368 		}
3369 		PF_RULES_WUNLOCK();
3370 
3371 		error = copyout(nvlpacked, nv->data, nv->len);
3372 
3373 #undef ERROUT
3374 DIOCGETRULENV_error:
3375 		free(nvlpacked, M_NVLIST);
3376 		nvlist_destroy(nvrule);
3377 		nvlist_destroy(nvl);
3378 
3379 		break;
3380 	}
3381 
3382 	case DIOCCHANGERULE: {
3383 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
3384 		struct pf_kruleset	*ruleset;
3385 		struct pf_krule		*oldrule = NULL, *newrule = NULL;
3386 		struct pfi_kkif		*kif = NULL;
3387 		struct pf_kpooladdr	*pa;
3388 		u_int32_t		 nr = 0;
3389 		int			 rs_num;
3390 
3391 		pcr->anchor[sizeof(pcr->anchor) - 1] = 0;
3392 
3393 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
3394 		    pcr->action > PF_CHANGE_GET_TICKET) {
3395 			error = EINVAL;
3396 			break;
3397 		}
3398 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
3399 			error = EINVAL;
3400 			break;
3401 		}
3402 
3403 		if (pcr->action != PF_CHANGE_REMOVE) {
3404 			newrule = pf_krule_alloc();
3405 			error = pf_rule_to_krule(&pcr->rule, newrule);
3406 			if (error != 0) {
3407 				free(newrule, M_PFRULE);
3408 				break;
3409 			}
3410 
3411 			if (newrule->ifname[0])
3412 				kif = pf_kkif_create(M_WAITOK);
3413 			pf_counter_u64_init(&newrule->evaluations, M_WAITOK);
3414 			for (int i = 0; i < 2; i++) {
3415 				pf_counter_u64_init(&newrule->packets[i], M_WAITOK);
3416 				pf_counter_u64_init(&newrule->bytes[i], M_WAITOK);
3417 			}
3418 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
3419 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
3420 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
3421 			newrule->cuid = td->td_ucred->cr_ruid;
3422 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
3423 			TAILQ_INIT(&newrule->rpool.list);
3424 		}
3425 #define	ERROUT(x)	ERROUT_IOCTL(DIOCCHANGERULE_error, x)
3426 
3427 		PF_RULES_WLOCK();
3428 #ifdef PF_WANT_32_TO_64_COUNTER
3429 		if (newrule != NULL) {
3430 			LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist);
3431 			newrule->allrulelinked = true;
3432 			V_pf_allrulecount++;
3433 		}
3434 #endif
3435 
3436 		if (!(pcr->action == PF_CHANGE_REMOVE ||
3437 		    pcr->action == PF_CHANGE_GET_TICKET) &&
3438 		    pcr->pool_ticket != V_ticket_pabuf)
3439 			ERROUT(EBUSY);
3440 
3441 		ruleset = pf_find_kruleset(pcr->anchor);
3442 		if (ruleset == NULL)
3443 			ERROUT(EINVAL);
3444 
3445 		rs_num = pf_get_ruleset_number(pcr->rule.action);
3446 		if (rs_num >= PF_RULESET_MAX)
3447 			ERROUT(EINVAL);
3448 
3449 		if (pcr->action == PF_CHANGE_GET_TICKET) {
3450 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
3451 			ERROUT(0);
3452 		} else if (pcr->ticket !=
3453 			    ruleset->rules[rs_num].active.ticket)
3454 				ERROUT(EINVAL);
3455 
3456 		if (pcr->action != PF_CHANGE_REMOVE) {
3457 			if (newrule->ifname[0]) {
3458 				newrule->kif = pfi_kkif_attach(kif,
3459 				    newrule->ifname);
3460 				kif = NULL;
3461 				pfi_kkif_ref(newrule->kif);
3462 			} else
3463 				newrule->kif = NULL;
3464 
3465 			if (newrule->rtableid > 0 &&
3466 			    newrule->rtableid >= rt_numfibs)
3467 				error = EBUSY;
3468 
3469 #ifdef ALTQ
3470 			/* set queue IDs */
3471 			if (newrule->qname[0] != 0) {
3472 				if ((newrule->qid =
3473 				    pf_qname2qid(newrule->qname)) == 0)
3474 					error = EBUSY;
3475 				else if (newrule->pqname[0] != 0) {
3476 					if ((newrule->pqid =
3477 					    pf_qname2qid(newrule->pqname)) == 0)
3478 						error = EBUSY;
3479 				} else
3480 					newrule->pqid = newrule->qid;
3481 			}
3482 #endif /* ALTQ */
3483 			if (newrule->tagname[0])
3484 				if ((newrule->tag =
3485 				    pf_tagname2tag(newrule->tagname)) == 0)
3486 					error = EBUSY;
3487 			if (newrule->match_tagname[0])
3488 				if ((newrule->match_tag = pf_tagname2tag(
3489 				    newrule->match_tagname)) == 0)
3490 					error = EBUSY;
3491 			if (newrule->rt && !newrule->direction)
3492 				error = EINVAL;
3493 			if (!newrule->log)
3494 				newrule->logif = 0;
3495 			if (newrule->logif >= PFLOGIFS_MAX)
3496 				error = EINVAL;
3497 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
3498 				error = ENOMEM;
3499 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
3500 				error = ENOMEM;
3501 			if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
3502 				error = EINVAL;
3503 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
3504 				if (pa->addr.type == PF_ADDR_TABLE) {
3505 					pa->addr.p.tbl =
3506 					    pfr_attach_table(ruleset,
3507 					    pa->addr.v.tblname);
3508 					if (pa->addr.p.tbl == NULL)
3509 						error = ENOMEM;
3510 				}
3511 
3512 			newrule->overload_tbl = NULL;
3513 			if (newrule->overload_tblname[0]) {
3514 				if ((newrule->overload_tbl = pfr_attach_table(
3515 				    ruleset, newrule->overload_tblname)) ==
3516 				    NULL)
3517 					error = EINVAL;
3518 				else
3519 					newrule->overload_tbl->pfrkt_flags |=
3520 					    PFR_TFLAG_ACTIVE;
3521 			}
3522 
3523 			pf_mv_kpool(&V_pf_pabuf, &newrule->rpool.list);
3524 			if (((((newrule->action == PF_NAT) ||
3525 			    (newrule->action == PF_RDR) ||
3526 			    (newrule->action == PF_BINAT) ||
3527 			    (newrule->rt > PF_NOPFROUTE)) &&
3528 			    !newrule->anchor)) &&
3529 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
3530 				error = EINVAL;
3531 
3532 			if (error) {
3533 				pf_free_rule(newrule);
3534 				PF_RULES_WUNLOCK();
3535 				break;
3536 			}
3537 
3538 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
3539 		}
3540 		pf_empty_kpool(&V_pf_pabuf);
3541 
3542 		if (pcr->action == PF_CHANGE_ADD_HEAD)
3543 			oldrule = TAILQ_FIRST(
3544 			    ruleset->rules[rs_num].active.ptr);
3545 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
3546 			oldrule = TAILQ_LAST(
3547 			    ruleset->rules[rs_num].active.ptr, pf_krulequeue);
3548 		else {
3549 			oldrule = TAILQ_FIRST(
3550 			    ruleset->rules[rs_num].active.ptr);
3551 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
3552 				oldrule = TAILQ_NEXT(oldrule, entries);
3553 			if (oldrule == NULL) {
3554 				if (newrule != NULL)
3555 					pf_free_rule(newrule);
3556 				PF_RULES_WUNLOCK();
3557 				error = EINVAL;
3558 				break;
3559 			}
3560 		}
3561 
3562 		if (pcr->action == PF_CHANGE_REMOVE) {
3563 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
3564 			    oldrule);
3565 			ruleset->rules[rs_num].active.rcount--;
3566 		} else {
3567 			if (oldrule == NULL)
3568 				TAILQ_INSERT_TAIL(
3569 				    ruleset->rules[rs_num].active.ptr,
3570 				    newrule, entries);
3571 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
3572 			    pcr->action == PF_CHANGE_ADD_BEFORE)
3573 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
3574 			else
3575 				TAILQ_INSERT_AFTER(
3576 				    ruleset->rules[rs_num].active.ptr,
3577 				    oldrule, newrule, entries);
3578 			ruleset->rules[rs_num].active.rcount++;
3579 		}
3580 
3581 		nr = 0;
3582 		TAILQ_FOREACH(oldrule,
3583 		    ruleset->rules[rs_num].active.ptr, entries)
3584 			oldrule->nr = nr++;
3585 
3586 		ruleset->rules[rs_num].active.ticket++;
3587 
3588 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
3589 		pf_remove_if_empty_kruleset(ruleset);
3590 
3591 		PF_RULES_WUNLOCK();
3592 		break;
3593 
3594 #undef ERROUT
3595 DIOCCHANGERULE_error:
3596 		PF_RULES_WUNLOCK();
3597 		pf_krule_free(newrule);
3598 		pf_kkif_free(kif);
3599 		break;
3600 	}
3601 
3602 	case DIOCCLRSTATES: {
3603 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
3604 		struct pf_kstate_kill	 kill;
3605 
3606 		error = pf_state_kill_to_kstate_kill(psk, &kill);
3607 		if (error)
3608 			break;
3609 
3610 		psk->psk_killed = pf_clear_states(&kill);
3611 		break;
3612 	}
3613 
3614 	case DIOCCLRSTATESNV: {
3615 		error = pf_clearstates_nv((struct pfioc_nv *)addr);
3616 		break;
3617 	}
3618 
3619 	case DIOCKILLSTATES: {
3620 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
3621 		struct pf_kstate_kill	 kill;
3622 
3623 		error = pf_state_kill_to_kstate_kill(psk, &kill);
3624 		if (error)
3625 			break;
3626 
3627 		psk->psk_killed = 0;
3628 		pf_killstates(&kill, &psk->psk_killed);
3629 		break;
3630 	}
3631 
3632 	case DIOCKILLSTATESNV: {
3633 		error = pf_killstates_nv((struct pfioc_nv *)addr);
3634 		break;
3635 	}
3636 
3637 	case DIOCADDSTATE: {
3638 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
3639 		struct pfsync_state	*sp = &ps->state;
3640 
3641 		if (sp->timeout >= PFTM_MAX) {
3642 			error = EINVAL;
3643 			break;
3644 		}
3645 		if (V_pfsync_state_import_ptr != NULL) {
3646 			PF_RULES_RLOCK();
3647 			error = V_pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
3648 			PF_RULES_RUNLOCK();
3649 		} else
3650 			error = EOPNOTSUPP;
3651 		break;
3652 	}
3653 
3654 	case DIOCGETSTATE: {
3655 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
3656 		struct pf_kstate	*s;
3657 
3658 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
3659 		if (s == NULL) {
3660 			error = ENOENT;
3661 			break;
3662 		}
3663 
3664 		pfsync_state_export(&ps->state, s);
3665 		PF_STATE_UNLOCK(s);
3666 		break;
3667 	}
3668 
3669 	case DIOCGETSTATENV: {
3670 		error = pf_getstate((struct pfioc_nv *)addr);
3671 		break;
3672 	}
3673 
3674 	case DIOCGETSTATES: {
3675 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
3676 		struct pf_kstate	*s;
3677 		struct pfsync_state	*pstore, *p;
3678 		int			 i, nr;
3679 		size_t			 slice_count = 16, count;
3680 		void			*out;
3681 
3682 		if (ps->ps_len <= 0) {
3683 			nr = uma_zone_get_cur(V_pf_state_z);
3684 			ps->ps_len = sizeof(struct pfsync_state) * nr;
3685 			break;
3686 		}
3687 
3688 		out = ps->ps_states;
3689 		pstore = mallocarray(slice_count,
3690 		    sizeof(struct pfsync_state), M_TEMP, M_WAITOK | M_ZERO);
3691 		nr = 0;
3692 
3693 		for (i = 0; i <= pf_hashmask; i++) {
3694 			struct pf_idhash *ih = &V_pf_idhash[i];
3695 
3696 DIOCGETSTATES_retry:
3697 			p = pstore;
3698 
3699 			if (LIST_EMPTY(&ih->states))
3700 				continue;
3701 
3702 			PF_HASHROW_LOCK(ih);
3703 			count = 0;
3704 			LIST_FOREACH(s, &ih->states, entry) {
3705 				if (s->timeout == PFTM_UNLINKED)
3706 					continue;
3707 				count++;
3708 			}
3709 
3710 			if (count > slice_count) {
3711 				PF_HASHROW_UNLOCK(ih);
3712 				free(pstore, M_TEMP);
3713 				slice_count = count * 2;
3714 				pstore = mallocarray(slice_count,
3715 				    sizeof(struct pfsync_state), M_TEMP,
3716 				    M_WAITOK | M_ZERO);
3717 				goto DIOCGETSTATES_retry;
3718 			}
3719 
3720 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
3721 				PF_HASHROW_UNLOCK(ih);
3722 				goto DIOCGETSTATES_full;
3723 			}
3724 
3725 			LIST_FOREACH(s, &ih->states, entry) {
3726 				if (s->timeout == PFTM_UNLINKED)
3727 					continue;
3728 
3729 				pfsync_state_export(p, s);
3730 				p++;
3731 				nr++;
3732 			}
3733 			PF_HASHROW_UNLOCK(ih);
3734 			error = copyout(pstore, out,
3735 			    sizeof(struct pfsync_state) * count);
3736 			if (error)
3737 				break;
3738 			out = ps->ps_states + nr;
3739 		}
3740 DIOCGETSTATES_full:
3741 		ps->ps_len = sizeof(struct pfsync_state) * nr;
3742 		free(pstore, M_TEMP);
3743 
3744 		break;
3745 	}
3746 
3747 	case DIOCGETSTATESV2: {
3748 		struct pfioc_states_v2	*ps = (struct pfioc_states_v2 *)addr;
3749 		struct pf_kstate	*s;
3750 		struct pf_state_export	*pstore, *p;
3751 		int i, nr;
3752 		size_t slice_count = 16, count;
3753 		void *out;
3754 
3755 		if (ps->ps_req_version > PF_STATE_VERSION) {
3756 			error = ENOTSUP;
3757 			break;
3758 		}
3759 
3760 		if (ps->ps_len <= 0) {
3761 			nr = uma_zone_get_cur(V_pf_state_z);
3762 			ps->ps_len = sizeof(struct pf_state_export) * nr;
3763 			break;
3764 		}
3765 
3766 		out = ps->ps_states;
3767 		pstore = mallocarray(slice_count,
3768 		    sizeof(struct pf_state_export), M_TEMP, M_WAITOK | M_ZERO);
3769 		nr = 0;
3770 
3771 		for (i = 0; i <= pf_hashmask; i++) {
3772 			struct pf_idhash *ih = &V_pf_idhash[i];
3773 
3774 DIOCGETSTATESV2_retry:
3775 			p = pstore;
3776 
3777 			if (LIST_EMPTY(&ih->states))
3778 				continue;
3779 
3780 			PF_HASHROW_LOCK(ih);
3781 			count = 0;
3782 			LIST_FOREACH(s, &ih->states, entry) {
3783 				if (s->timeout == PFTM_UNLINKED)
3784 					continue;
3785 				count++;
3786 			}
3787 
3788 			if (count > slice_count) {
3789 				PF_HASHROW_UNLOCK(ih);
3790 				free(pstore, M_TEMP);
3791 				slice_count = count * 2;
3792 				pstore = mallocarray(slice_count,
3793 				    sizeof(struct pf_state_export), M_TEMP,
3794 				    M_WAITOK | M_ZERO);
3795 				goto DIOCGETSTATESV2_retry;
3796 			}
3797 
3798 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
3799 				PF_HASHROW_UNLOCK(ih);
3800 				goto DIOCGETSTATESV2_full;
3801 			}
3802 
3803 			LIST_FOREACH(s, &ih->states, entry) {
3804 				if (s->timeout == PFTM_UNLINKED)
3805 					continue;
3806 
3807 				pf_state_export(p, s);
3808 				p++;
3809 				nr++;
3810 			}
3811 			PF_HASHROW_UNLOCK(ih);
3812 			error = copyout(pstore, out,
3813 			    sizeof(struct pf_state_export) * count);
3814 			if (error)
3815 				break;
3816 			out = ps->ps_states + nr;
3817 		}
3818 DIOCGETSTATESV2_full:
3819 		ps->ps_len = nr * sizeof(struct pf_state_export);
3820 		free(pstore, M_TEMP);
3821 
3822 		break;
3823 	}
3824 
3825 	case DIOCGETSTATUS: {
3826 		struct pf_status *s = (struct pf_status *)addr;
3827 
3828 		PF_RULES_RLOCK();
3829 		s->running = V_pf_status.running;
3830 		s->since   = V_pf_status.since;
3831 		s->debug   = V_pf_status.debug;
3832 		s->hostid  = V_pf_status.hostid;
3833 		s->states  = V_pf_status.states;
3834 		s->src_nodes = V_pf_status.src_nodes;
3835 
3836 		for (int i = 0; i < PFRES_MAX; i++)
3837 			s->counters[i] =
3838 			    counter_u64_fetch(V_pf_status.counters[i]);
3839 		for (int i = 0; i < LCNT_MAX; i++)
3840 			s->lcounters[i] =
3841 			    counter_u64_fetch(V_pf_status.lcounters[i]);
3842 		for (int i = 0; i < FCNT_MAX; i++)
3843 			s->fcounters[i] =
3844 			    pf_counter_u64_fetch(&V_pf_status.fcounters[i]);
3845 		for (int i = 0; i < SCNT_MAX; i++)
3846 			s->scounters[i] =
3847 			    counter_u64_fetch(V_pf_status.scounters[i]);
3848 
3849 		bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ);
3850 		bcopy(V_pf_status.pf_chksum, s->pf_chksum,
3851 		    PF_MD5_DIGEST_LENGTH);
3852 
3853 		pfi_update_status(s->ifname, s);
3854 		PF_RULES_RUNLOCK();
3855 		break;
3856 	}
3857 
3858 	case DIOCGETSTATUSNV: {
3859 		error = pf_getstatus((struct pfioc_nv *)addr);
3860 		break;
3861 	}
3862 
3863 	case DIOCSETSTATUSIF: {
3864 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
3865 
3866 		if (pi->ifname[0] == 0) {
3867 			bzero(V_pf_status.ifname, IFNAMSIZ);
3868 			break;
3869 		}
3870 		PF_RULES_WLOCK();
3871 		error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
3872 		PF_RULES_WUNLOCK();
3873 		break;
3874 	}
3875 
3876 	case DIOCCLRSTATUS: {
3877 		PF_RULES_WLOCK();
3878 		for (int i = 0; i < PFRES_MAX; i++)
3879 			counter_u64_zero(V_pf_status.counters[i]);
3880 		for (int i = 0; i < FCNT_MAX; i++)
3881 			pf_counter_u64_zero(&V_pf_status.fcounters[i]);
3882 		for (int i = 0; i < SCNT_MAX; i++)
3883 			counter_u64_zero(V_pf_status.scounters[i]);
3884 		for (int i = 0; i < KLCNT_MAX; i++)
3885 			counter_u64_zero(V_pf_status.lcounters[i]);
3886 		V_pf_status.since = time_second;
3887 		if (*V_pf_status.ifname)
3888 			pfi_update_status(V_pf_status.ifname, NULL);
3889 		PF_RULES_WUNLOCK();
3890 		break;
3891 	}
3892 
3893 	case DIOCNATLOOK: {
3894 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
3895 		struct pf_state_key	*sk;
3896 		struct pf_kstate	*state;
3897 		struct pf_state_key_cmp	 key;
3898 		int			 m = 0, direction = pnl->direction;
3899 		int			 sidx, didx;
3900 
3901 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
3902 		sidx = (direction == PF_IN) ? 1 : 0;
3903 		didx = (direction == PF_IN) ? 0 : 1;
3904 
3905 		if (!pnl->proto ||
3906 		    PF_AZERO(&pnl->saddr, pnl->af) ||
3907 		    PF_AZERO(&pnl->daddr, pnl->af) ||
3908 		    ((pnl->proto == IPPROTO_TCP ||
3909 		    pnl->proto == IPPROTO_UDP) &&
3910 		    (!pnl->dport || !pnl->sport)))
3911 			error = EINVAL;
3912 		else {
3913 			bzero(&key, sizeof(key));
3914 			key.af = pnl->af;
3915 			key.proto = pnl->proto;
3916 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
3917 			key.port[sidx] = pnl->sport;
3918 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
3919 			key.port[didx] = pnl->dport;
3920 
3921 			state = pf_find_state_all(&key, direction, &m);
3922 			if (state == NULL) {
3923 				error = ENOENT;
3924 			} else {
3925 				if (m > 1) {
3926 					PF_STATE_UNLOCK(state);
3927 					error = E2BIG;	/* more than one state */
3928 				} else {
3929 					sk = state->key[sidx];
3930 					PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
3931 					pnl->rsport = sk->port[sidx];
3932 					PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
3933 					pnl->rdport = sk->port[didx];
3934 					PF_STATE_UNLOCK(state);
3935 				}
3936 			}
3937 		}
3938 		break;
3939 	}
3940 
3941 	case DIOCSETTIMEOUT: {
3942 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
3943 		int		 old;
3944 
3945 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
3946 		    pt->seconds < 0) {
3947 			error = EINVAL;
3948 			break;
3949 		}
3950 		PF_RULES_WLOCK();
3951 		old = V_pf_default_rule.timeout[pt->timeout];
3952 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
3953 			pt->seconds = 1;
3954 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
3955 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
3956 			wakeup(pf_purge_thread);
3957 		pt->seconds = old;
3958 		PF_RULES_WUNLOCK();
3959 		break;
3960 	}
3961 
3962 	case DIOCGETTIMEOUT: {
3963 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
3964 
3965 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
3966 			error = EINVAL;
3967 			break;
3968 		}
3969 		PF_RULES_RLOCK();
3970 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
3971 		PF_RULES_RUNLOCK();
3972 		break;
3973 	}
3974 
3975 	case DIOCGETLIMIT: {
3976 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
3977 
3978 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
3979 			error = EINVAL;
3980 			break;
3981 		}
3982 		PF_RULES_RLOCK();
3983 		pl->limit = V_pf_limits[pl->index].limit;
3984 		PF_RULES_RUNLOCK();
3985 		break;
3986 	}
3987 
3988 	case DIOCSETLIMIT: {
3989 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
3990 		int			 old_limit;
3991 
3992 		PF_RULES_WLOCK();
3993 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
3994 		    V_pf_limits[pl->index].zone == NULL) {
3995 			PF_RULES_WUNLOCK();
3996 			error = EINVAL;
3997 			break;
3998 		}
3999 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
4000 		old_limit = V_pf_limits[pl->index].limit;
4001 		V_pf_limits[pl->index].limit = pl->limit;
4002 		pl->limit = old_limit;
4003 		PF_RULES_WUNLOCK();
4004 		break;
4005 	}
4006 
4007 	case DIOCSETDEBUG: {
4008 		u_int32_t	*level = (u_int32_t *)addr;
4009 
4010 		PF_RULES_WLOCK();
4011 		V_pf_status.debug = *level;
4012 		PF_RULES_WUNLOCK();
4013 		break;
4014 	}
4015 
4016 	case DIOCCLRRULECTRS: {
4017 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
4018 		struct pf_kruleset	*ruleset = &pf_main_ruleset;
4019 		struct pf_krule		*rule;
4020 
4021 		PF_RULES_WLOCK();
4022 		TAILQ_FOREACH(rule,
4023 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
4024 			pf_counter_u64_zero(&rule->evaluations);
4025 			for (int i = 0; i < 2; i++) {
4026 				pf_counter_u64_zero(&rule->packets[i]);
4027 				pf_counter_u64_zero(&rule->bytes[i]);
4028 			}
4029 		}
4030 		PF_RULES_WUNLOCK();
4031 		break;
4032 	}
4033 
4034 	case DIOCGIFSPEEDV0:
4035 	case DIOCGIFSPEEDV1: {
4036 		struct pf_ifspeed_v1	*psp = (struct pf_ifspeed_v1 *)addr;
4037 		struct pf_ifspeed_v1	ps;
4038 		struct ifnet		*ifp;
4039 
4040 		if (psp->ifname[0] == '\0') {
4041 			error = EINVAL;
4042 			break;
4043 		}
4044 
4045 		error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ);
4046 		if (error != 0)
4047 			break;
4048 		ifp = ifunit(ps.ifname);
4049 		if (ifp != NULL) {
4050 			psp->baudrate32 =
4051 			    (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
4052 			if (cmd == DIOCGIFSPEEDV1)
4053 				psp->baudrate = ifp->if_baudrate;
4054 		} else {
4055 			error = EINVAL;
4056 		}
4057 		break;
4058 	}
4059 
4060 #ifdef ALTQ
4061 	case DIOCSTARTALTQ: {
4062 		struct pf_altq		*altq;
4063 
4064 		PF_RULES_WLOCK();
4065 		/* enable all altq interfaces on active list */
4066 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4067 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4068 				error = pf_enable_altq(altq);
4069 				if (error != 0)
4070 					break;
4071 			}
4072 		}
4073 		if (error == 0)
4074 			V_pf_altq_running = 1;
4075 		PF_RULES_WUNLOCK();
4076 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
4077 		break;
4078 	}
4079 
4080 	case DIOCSTOPALTQ: {
4081 		struct pf_altq		*altq;
4082 
4083 		PF_RULES_WLOCK();
4084 		/* disable all altq interfaces on active list */
4085 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4086 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4087 				error = pf_disable_altq(altq);
4088 				if (error != 0)
4089 					break;
4090 			}
4091 		}
4092 		if (error == 0)
4093 			V_pf_altq_running = 0;
4094 		PF_RULES_WUNLOCK();
4095 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
4096 		break;
4097 	}
4098 
4099 	case DIOCADDALTQV0:
4100 	case DIOCADDALTQV1: {
4101 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4102 		struct pf_altq		*altq, *a;
4103 		struct ifnet		*ifp;
4104 
4105 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
4106 		error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
4107 		if (error)
4108 			break;
4109 		altq->local_flags = 0;
4110 
4111 		PF_RULES_WLOCK();
4112 		if (pa->ticket != V_ticket_altqs_inactive) {
4113 			PF_RULES_WUNLOCK();
4114 			free(altq, M_PFALTQ);
4115 			error = EBUSY;
4116 			break;
4117 		}
4118 
4119 		/*
4120 		 * if this is for a queue, find the discipline and
4121 		 * copy the necessary fields
4122 		 */
4123 		if (altq->qname[0] != 0) {
4124 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
4125 				PF_RULES_WUNLOCK();
4126 				error = EBUSY;
4127 				free(altq, M_PFALTQ);
4128 				break;
4129 			}
4130 			altq->altq_disc = NULL;
4131 			TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
4132 				if (strncmp(a->ifname, altq->ifname,
4133 				    IFNAMSIZ) == 0) {
4134 					altq->altq_disc = a->altq_disc;
4135 					break;
4136 				}
4137 			}
4138 		}
4139 
4140 		if ((ifp = ifunit(altq->ifname)) == NULL)
4141 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
4142 		else
4143 			error = altq_add(ifp, altq);
4144 
4145 		if (error) {
4146 			PF_RULES_WUNLOCK();
4147 			free(altq, M_PFALTQ);
4148 			break;
4149 		}
4150 
4151 		if (altq->qname[0] != 0)
4152 			TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
4153 		else
4154 			TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
4155 		/* version error check done on import above */
4156 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4157 		PF_RULES_WUNLOCK();
4158 		break;
4159 	}
4160 
4161 	case DIOCGETALTQSV0:
4162 	case DIOCGETALTQSV1: {
4163 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4164 		struct pf_altq		*altq;
4165 
4166 		PF_RULES_RLOCK();
4167 		pa->nr = 0;
4168 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
4169 			pa->nr++;
4170 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
4171 			pa->nr++;
4172 		pa->ticket = V_ticket_altqs_active;
4173 		PF_RULES_RUNLOCK();
4174 		break;
4175 	}
4176 
4177 	case DIOCGETALTQV0:
4178 	case DIOCGETALTQV1: {
4179 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4180 		struct pf_altq		*altq;
4181 
4182 		PF_RULES_RLOCK();
4183 		if (pa->ticket != V_ticket_altqs_active) {
4184 			PF_RULES_RUNLOCK();
4185 			error = EBUSY;
4186 			break;
4187 		}
4188 		altq = pf_altq_get_nth_active(pa->nr);
4189 		if (altq == NULL) {
4190 			PF_RULES_RUNLOCK();
4191 			error = EBUSY;
4192 			break;
4193 		}
4194 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4195 		PF_RULES_RUNLOCK();
4196 		break;
4197 	}
4198 
4199 	case DIOCCHANGEALTQV0:
4200 	case DIOCCHANGEALTQV1:
4201 		/* CHANGEALTQ not supported yet! */
4202 		error = ENODEV;
4203 		break;
4204 
4205 	case DIOCGETQSTATSV0:
4206 	case DIOCGETQSTATSV1: {
4207 		struct pfioc_qstats_v1	*pq = (struct pfioc_qstats_v1 *)addr;
4208 		struct pf_altq		*altq;
4209 		int			 nbytes;
4210 		u_int32_t		 version;
4211 
4212 		PF_RULES_RLOCK();
4213 		if (pq->ticket != V_ticket_altqs_active) {
4214 			PF_RULES_RUNLOCK();
4215 			error = EBUSY;
4216 			break;
4217 		}
4218 		nbytes = pq->nbytes;
4219 		altq = pf_altq_get_nth_active(pq->nr);
4220 		if (altq == NULL) {
4221 			PF_RULES_RUNLOCK();
4222 			error = EBUSY;
4223 			break;
4224 		}
4225 
4226 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
4227 			PF_RULES_RUNLOCK();
4228 			error = ENXIO;
4229 			break;
4230 		}
4231 		PF_RULES_RUNLOCK();
4232 		if (cmd == DIOCGETQSTATSV0)
4233 			version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
4234 		else
4235 			version = pq->version;
4236 		error = altq_getqstats(altq, pq->buf, &nbytes, version);
4237 		if (error == 0) {
4238 			pq->scheduler = altq->scheduler;
4239 			pq->nbytes = nbytes;
4240 		}
4241 		break;
4242 	}
4243 #endif /* ALTQ */
4244 
4245 	case DIOCBEGINADDRS: {
4246 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4247 
4248 		PF_RULES_WLOCK();
4249 		pf_empty_kpool(&V_pf_pabuf);
4250 		pp->ticket = ++V_ticket_pabuf;
4251 		PF_RULES_WUNLOCK();
4252 		break;
4253 	}
4254 
4255 	case DIOCADDADDR: {
4256 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4257 		struct pf_kpooladdr	*pa;
4258 		struct pfi_kkif		*kif = NULL;
4259 
4260 #ifndef INET
4261 		if (pp->af == AF_INET) {
4262 			error = EAFNOSUPPORT;
4263 			break;
4264 		}
4265 #endif /* INET */
4266 #ifndef INET6
4267 		if (pp->af == AF_INET6) {
4268 			error = EAFNOSUPPORT;
4269 			break;
4270 		}
4271 #endif /* INET6 */
4272 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
4273 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
4274 		    pp->addr.addr.type != PF_ADDR_TABLE) {
4275 			error = EINVAL;
4276 			break;
4277 		}
4278 		if (pp->addr.addr.p.dyn != NULL) {
4279 			error = EINVAL;
4280 			break;
4281 		}
4282 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
4283 		error = pf_pooladdr_to_kpooladdr(&pp->addr, pa);
4284 		if (error != 0)
4285 			break;
4286 		if (pa->ifname[0])
4287 			kif = pf_kkif_create(M_WAITOK);
4288 		PF_RULES_WLOCK();
4289 		if (pp->ticket != V_ticket_pabuf) {
4290 			PF_RULES_WUNLOCK();
4291 			if (pa->ifname[0])
4292 				pf_kkif_free(kif);
4293 			free(pa, M_PFRULE);
4294 			error = EBUSY;
4295 			break;
4296 		}
4297 		if (pa->ifname[0]) {
4298 			pa->kif = pfi_kkif_attach(kif, pa->ifname);
4299 			kif = NULL;
4300 			pfi_kkif_ref(pa->kif);
4301 		} else
4302 			pa->kif = NULL;
4303 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
4304 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
4305 			if (pa->ifname[0])
4306 				pfi_kkif_unref(pa->kif);
4307 			PF_RULES_WUNLOCK();
4308 			free(pa, M_PFRULE);
4309 			break;
4310 		}
4311 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
4312 		PF_RULES_WUNLOCK();
4313 		break;
4314 	}
4315 
4316 	case DIOCGETADDRS: {
4317 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4318 		struct pf_kpool		*pool;
4319 		struct pf_kpooladdr	*pa;
4320 
4321 		pp->anchor[sizeof(pp->anchor) - 1] = 0;
4322 		pp->nr = 0;
4323 
4324 		PF_RULES_RLOCK();
4325 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
4326 		    pp->r_num, 0, 1, 0);
4327 		if (pool == NULL) {
4328 			PF_RULES_RUNLOCK();
4329 			error = EBUSY;
4330 			break;
4331 		}
4332 		TAILQ_FOREACH(pa, &pool->list, entries)
4333 			pp->nr++;
4334 		PF_RULES_RUNLOCK();
4335 		break;
4336 	}
4337 
4338 	case DIOCGETADDR: {
4339 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4340 		struct pf_kpool		*pool;
4341 		struct pf_kpooladdr	*pa;
4342 		u_int32_t		 nr = 0;
4343 
4344 		pp->anchor[sizeof(pp->anchor) - 1] = 0;
4345 
4346 		PF_RULES_RLOCK();
4347 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
4348 		    pp->r_num, 0, 1, 1);
4349 		if (pool == NULL) {
4350 			PF_RULES_RUNLOCK();
4351 			error = EBUSY;
4352 			break;
4353 		}
4354 		pa = TAILQ_FIRST(&pool->list);
4355 		while ((pa != NULL) && (nr < pp->nr)) {
4356 			pa = TAILQ_NEXT(pa, entries);
4357 			nr++;
4358 		}
4359 		if (pa == NULL) {
4360 			PF_RULES_RUNLOCK();
4361 			error = EBUSY;
4362 			break;
4363 		}
4364 		pf_kpooladdr_to_pooladdr(pa, &pp->addr);
4365 		pf_addr_copyout(&pp->addr.addr);
4366 		PF_RULES_RUNLOCK();
4367 		break;
4368 	}
4369 
4370 	case DIOCCHANGEADDR: {
4371 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
4372 		struct pf_kpool		*pool;
4373 		struct pf_kpooladdr	*oldpa = NULL, *newpa = NULL;
4374 		struct pf_kruleset	*ruleset;
4375 		struct pfi_kkif		*kif = NULL;
4376 
4377 		pca->anchor[sizeof(pca->anchor) - 1] = 0;
4378 
4379 		if (pca->action < PF_CHANGE_ADD_HEAD ||
4380 		    pca->action > PF_CHANGE_REMOVE) {
4381 			error = EINVAL;
4382 			break;
4383 		}
4384 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
4385 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
4386 		    pca->addr.addr.type != PF_ADDR_TABLE) {
4387 			error = EINVAL;
4388 			break;
4389 		}
4390 		if (pca->addr.addr.p.dyn != NULL) {
4391 			error = EINVAL;
4392 			break;
4393 		}
4394 
4395 		if (pca->action != PF_CHANGE_REMOVE) {
4396 #ifndef INET
4397 			if (pca->af == AF_INET) {
4398 				error = EAFNOSUPPORT;
4399 				break;
4400 			}
4401 #endif /* INET */
4402 #ifndef INET6
4403 			if (pca->af == AF_INET6) {
4404 				error = EAFNOSUPPORT;
4405 				break;
4406 			}
4407 #endif /* INET6 */
4408 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
4409 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
4410 			if (newpa->ifname[0])
4411 				kif = pf_kkif_create(M_WAITOK);
4412 			newpa->kif = NULL;
4413 		}
4414 #define	ERROUT(x)	ERROUT_IOCTL(DIOCCHANGEADDR_error, x)
4415 		PF_RULES_WLOCK();
4416 		ruleset = pf_find_kruleset(pca->anchor);
4417 		if (ruleset == NULL)
4418 			ERROUT(EBUSY);
4419 
4420 		pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
4421 		    pca->r_num, pca->r_last, 1, 1);
4422 		if (pool == NULL)
4423 			ERROUT(EBUSY);
4424 
4425 		if (pca->action != PF_CHANGE_REMOVE) {
4426 			if (newpa->ifname[0]) {
4427 				newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
4428 				pfi_kkif_ref(newpa->kif);
4429 				kif = NULL;
4430 			}
4431 
4432 			switch (newpa->addr.type) {
4433 			case PF_ADDR_DYNIFTL:
4434 				error = pfi_dynaddr_setup(&newpa->addr,
4435 				    pca->af);
4436 				break;
4437 			case PF_ADDR_TABLE:
4438 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
4439 				    newpa->addr.v.tblname);
4440 				if (newpa->addr.p.tbl == NULL)
4441 					error = ENOMEM;
4442 				break;
4443 			}
4444 			if (error)
4445 				goto DIOCCHANGEADDR_error;
4446 		}
4447 
4448 		switch (pca->action) {
4449 		case PF_CHANGE_ADD_HEAD:
4450 			oldpa = TAILQ_FIRST(&pool->list);
4451 			break;
4452 		case PF_CHANGE_ADD_TAIL:
4453 			oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
4454 			break;
4455 		default:
4456 			oldpa = TAILQ_FIRST(&pool->list);
4457 			for (int i = 0; oldpa && i < pca->nr; i++)
4458 				oldpa = TAILQ_NEXT(oldpa, entries);
4459 
4460 			if (oldpa == NULL)
4461 				ERROUT(EINVAL);
4462 		}
4463 
4464 		if (pca->action == PF_CHANGE_REMOVE) {
4465 			TAILQ_REMOVE(&pool->list, oldpa, entries);
4466 			switch (oldpa->addr.type) {
4467 			case PF_ADDR_DYNIFTL:
4468 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
4469 				break;
4470 			case PF_ADDR_TABLE:
4471 				pfr_detach_table(oldpa->addr.p.tbl);
4472 				break;
4473 			}
4474 			if (oldpa->kif)
4475 				pfi_kkif_unref(oldpa->kif);
4476 			free(oldpa, M_PFRULE);
4477 		} else {
4478 			if (oldpa == NULL)
4479 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
4480 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
4481 			    pca->action == PF_CHANGE_ADD_BEFORE)
4482 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
4483 			else
4484 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
4485 				    newpa, entries);
4486 		}
4487 
4488 		pool->cur = TAILQ_FIRST(&pool->list);
4489 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
4490 		PF_RULES_WUNLOCK();
4491 		break;
4492 
4493 #undef ERROUT
4494 DIOCCHANGEADDR_error:
4495 		if (newpa != NULL) {
4496 			if (newpa->kif)
4497 				pfi_kkif_unref(newpa->kif);
4498 			free(newpa, M_PFRULE);
4499 		}
4500 		PF_RULES_WUNLOCK();
4501 		pf_kkif_free(kif);
4502 		break;
4503 	}
4504 
4505 	case DIOCGETRULESETS: {
4506 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
4507 		struct pf_kruleset	*ruleset;
4508 		struct pf_kanchor	*anchor;
4509 
4510 		pr->path[sizeof(pr->path) - 1] = 0;
4511 
4512 		PF_RULES_RLOCK();
4513 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4514 			PF_RULES_RUNLOCK();
4515 			error = ENOENT;
4516 			break;
4517 		}
4518 		pr->nr = 0;
4519 		if (ruleset->anchor == NULL) {
4520 			/* XXX kludge for pf_main_ruleset */
4521 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4522 				if (anchor->parent == NULL)
4523 					pr->nr++;
4524 		} else {
4525 			RB_FOREACH(anchor, pf_kanchor_node,
4526 			    &ruleset->anchor->children)
4527 				pr->nr++;
4528 		}
4529 		PF_RULES_RUNLOCK();
4530 		break;
4531 	}
4532 
4533 	case DIOCGETRULESET: {
4534 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
4535 		struct pf_kruleset	*ruleset;
4536 		struct pf_kanchor	*anchor;
4537 		u_int32_t		 nr = 0;
4538 
4539 		pr->path[sizeof(pr->path) - 1] = 0;
4540 
4541 		PF_RULES_RLOCK();
4542 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4543 			PF_RULES_RUNLOCK();
4544 			error = ENOENT;
4545 			break;
4546 		}
4547 		pr->name[0] = 0;
4548 		if (ruleset->anchor == NULL) {
4549 			/* XXX kludge for pf_main_ruleset */
4550 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4551 				if (anchor->parent == NULL && nr++ == pr->nr) {
4552 					strlcpy(pr->name, anchor->name,
4553 					    sizeof(pr->name));
4554 					break;
4555 				}
4556 		} else {
4557 			RB_FOREACH(anchor, pf_kanchor_node,
4558 			    &ruleset->anchor->children)
4559 				if (nr++ == pr->nr) {
4560 					strlcpy(pr->name, anchor->name,
4561 					    sizeof(pr->name));
4562 					break;
4563 				}
4564 		}
4565 		if (!pr->name[0])
4566 			error = EBUSY;
4567 		PF_RULES_RUNLOCK();
4568 		break;
4569 	}
4570 
4571 	case DIOCRCLRTABLES: {
4572 		struct pfioc_table *io = (struct pfioc_table *)addr;
4573 
4574 		if (io->pfrio_esize != 0) {
4575 			error = ENODEV;
4576 			break;
4577 		}
4578 		PF_RULES_WLOCK();
4579 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
4580 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4581 		PF_RULES_WUNLOCK();
4582 		break;
4583 	}
4584 
4585 	case DIOCRADDTABLES: {
4586 		struct pfioc_table *io = (struct pfioc_table *)addr;
4587 		struct pfr_table *pfrts;
4588 		size_t totlen;
4589 
4590 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4591 			error = ENODEV;
4592 			break;
4593 		}
4594 
4595 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4596 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4597 			error = ENOMEM;
4598 			break;
4599 		}
4600 
4601 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4602 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4603 		    M_TEMP, M_WAITOK);
4604 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4605 		if (error) {
4606 			free(pfrts, M_TEMP);
4607 			break;
4608 		}
4609 		PF_RULES_WLOCK();
4610 		error = pfr_add_tables(pfrts, io->pfrio_size,
4611 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4612 		PF_RULES_WUNLOCK();
4613 		free(pfrts, M_TEMP);
4614 		break;
4615 	}
4616 
4617 	case DIOCRDELTABLES: {
4618 		struct pfioc_table *io = (struct pfioc_table *)addr;
4619 		struct pfr_table *pfrts;
4620 		size_t totlen;
4621 
4622 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4623 			error = ENODEV;
4624 			break;
4625 		}
4626 
4627 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4628 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4629 			error = ENOMEM;
4630 			break;
4631 		}
4632 
4633 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4634 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4635 		    M_TEMP, M_WAITOK);
4636 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4637 		if (error) {
4638 			free(pfrts, M_TEMP);
4639 			break;
4640 		}
4641 		PF_RULES_WLOCK();
4642 		error = pfr_del_tables(pfrts, io->pfrio_size,
4643 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4644 		PF_RULES_WUNLOCK();
4645 		free(pfrts, M_TEMP);
4646 		break;
4647 	}
4648 
4649 	case DIOCRGETTABLES: {
4650 		struct pfioc_table *io = (struct pfioc_table *)addr;
4651 		struct pfr_table *pfrts;
4652 		size_t totlen;
4653 		int n;
4654 
4655 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4656 			error = ENODEV;
4657 			break;
4658 		}
4659 		PF_RULES_RLOCK();
4660 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4661 		if (n < 0) {
4662 			PF_RULES_RUNLOCK();
4663 			error = EINVAL;
4664 			break;
4665 		}
4666 		io->pfrio_size = min(io->pfrio_size, n);
4667 
4668 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4669 
4670 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4671 		    M_TEMP, M_NOWAIT | M_ZERO);
4672 		if (pfrts == NULL) {
4673 			error = ENOMEM;
4674 			PF_RULES_RUNLOCK();
4675 			break;
4676 		}
4677 		error = pfr_get_tables(&io->pfrio_table, pfrts,
4678 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4679 		PF_RULES_RUNLOCK();
4680 		if (error == 0)
4681 			error = copyout(pfrts, io->pfrio_buffer, totlen);
4682 		free(pfrts, M_TEMP);
4683 		break;
4684 	}
4685 
4686 	case DIOCRGETTSTATS: {
4687 		struct pfioc_table *io = (struct pfioc_table *)addr;
4688 		struct pfr_tstats *pfrtstats;
4689 		size_t totlen;
4690 		int n;
4691 
4692 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
4693 			error = ENODEV;
4694 			break;
4695 		}
4696 		PF_TABLE_STATS_LOCK();
4697 		PF_RULES_RLOCK();
4698 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4699 		if (n < 0) {
4700 			PF_RULES_RUNLOCK();
4701 			PF_TABLE_STATS_UNLOCK();
4702 			error = EINVAL;
4703 			break;
4704 		}
4705 		io->pfrio_size = min(io->pfrio_size, n);
4706 
4707 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
4708 		pfrtstats = mallocarray(io->pfrio_size,
4709 		    sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT | M_ZERO);
4710 		if (pfrtstats == NULL) {
4711 			error = ENOMEM;
4712 			PF_RULES_RUNLOCK();
4713 			PF_TABLE_STATS_UNLOCK();
4714 			break;
4715 		}
4716 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
4717 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4718 		PF_RULES_RUNLOCK();
4719 		PF_TABLE_STATS_UNLOCK();
4720 		if (error == 0)
4721 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
4722 		free(pfrtstats, M_TEMP);
4723 		break;
4724 	}
4725 
4726 	case DIOCRCLRTSTATS: {
4727 		struct pfioc_table *io = (struct pfioc_table *)addr;
4728 		struct pfr_table *pfrts;
4729 		size_t totlen;
4730 
4731 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4732 			error = ENODEV;
4733 			break;
4734 		}
4735 
4736 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4737 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4738 			/* We used to count tables and use the minimum required
4739 			 * size, so we didn't fail on overly large requests.
4740 			 * Keep doing so. */
4741 			io->pfrio_size = pf_ioctl_maxcount;
4742 			break;
4743 		}
4744 
4745 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4746 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4747 		    M_TEMP, M_WAITOK);
4748 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4749 		if (error) {
4750 			free(pfrts, M_TEMP);
4751 			break;
4752 		}
4753 
4754 		PF_TABLE_STATS_LOCK();
4755 		PF_RULES_RLOCK();
4756 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
4757 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4758 		PF_RULES_RUNLOCK();
4759 		PF_TABLE_STATS_UNLOCK();
4760 		free(pfrts, M_TEMP);
4761 		break;
4762 	}
4763 
4764 	case DIOCRSETTFLAGS: {
4765 		struct pfioc_table *io = (struct pfioc_table *)addr;
4766 		struct pfr_table *pfrts;
4767 		size_t totlen;
4768 		int n;
4769 
4770 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4771 			error = ENODEV;
4772 			break;
4773 		}
4774 
4775 		PF_RULES_RLOCK();
4776 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4777 		if (n < 0) {
4778 			PF_RULES_RUNLOCK();
4779 			error = EINVAL;
4780 			break;
4781 		}
4782 
4783 		io->pfrio_size = min(io->pfrio_size, n);
4784 		PF_RULES_RUNLOCK();
4785 
4786 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4787 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4788 		    M_TEMP, M_WAITOK);
4789 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4790 		if (error) {
4791 			free(pfrts, M_TEMP);
4792 			break;
4793 		}
4794 		PF_RULES_WLOCK();
4795 		error = pfr_set_tflags(pfrts, io->pfrio_size,
4796 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
4797 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4798 		PF_RULES_WUNLOCK();
4799 		free(pfrts, M_TEMP);
4800 		break;
4801 	}
4802 
4803 	case DIOCRCLRADDRS: {
4804 		struct pfioc_table *io = (struct pfioc_table *)addr;
4805 
4806 		if (io->pfrio_esize != 0) {
4807 			error = ENODEV;
4808 			break;
4809 		}
4810 		PF_RULES_WLOCK();
4811 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
4812 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4813 		PF_RULES_WUNLOCK();
4814 		break;
4815 	}
4816 
4817 	case DIOCRADDADDRS: {
4818 		struct pfioc_table *io = (struct pfioc_table *)addr;
4819 		struct pfr_addr *pfras;
4820 		size_t totlen;
4821 
4822 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4823 			error = ENODEV;
4824 			break;
4825 		}
4826 		if (io->pfrio_size < 0 ||
4827 		    io->pfrio_size > pf_ioctl_maxcount ||
4828 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4829 			error = EINVAL;
4830 			break;
4831 		}
4832 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4833 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4834 		    M_TEMP, M_WAITOK);
4835 		error = copyin(io->pfrio_buffer, pfras, totlen);
4836 		if (error) {
4837 			free(pfras, M_TEMP);
4838 			break;
4839 		}
4840 		PF_RULES_WLOCK();
4841 		error = pfr_add_addrs(&io->pfrio_table, pfras,
4842 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
4843 		    PFR_FLAG_USERIOCTL);
4844 		PF_RULES_WUNLOCK();
4845 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4846 			error = copyout(pfras, io->pfrio_buffer, totlen);
4847 		free(pfras, M_TEMP);
4848 		break;
4849 	}
4850 
4851 	case DIOCRDELADDRS: {
4852 		struct pfioc_table *io = (struct pfioc_table *)addr;
4853 		struct pfr_addr *pfras;
4854 		size_t totlen;
4855 
4856 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4857 			error = ENODEV;
4858 			break;
4859 		}
4860 		if (io->pfrio_size < 0 ||
4861 		    io->pfrio_size > pf_ioctl_maxcount ||
4862 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4863 			error = EINVAL;
4864 			break;
4865 		}
4866 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4867 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4868 		    M_TEMP, M_WAITOK);
4869 		error = copyin(io->pfrio_buffer, pfras, totlen);
4870 		if (error) {
4871 			free(pfras, M_TEMP);
4872 			break;
4873 		}
4874 		PF_RULES_WLOCK();
4875 		error = pfr_del_addrs(&io->pfrio_table, pfras,
4876 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
4877 		    PFR_FLAG_USERIOCTL);
4878 		PF_RULES_WUNLOCK();
4879 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4880 			error = copyout(pfras, io->pfrio_buffer, totlen);
4881 		free(pfras, M_TEMP);
4882 		break;
4883 	}
4884 
4885 	case DIOCRSETADDRS: {
4886 		struct pfioc_table *io = (struct pfioc_table *)addr;
4887 		struct pfr_addr *pfras;
4888 		size_t totlen, count;
4889 
4890 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4891 			error = ENODEV;
4892 			break;
4893 		}
4894 		if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
4895 			error = EINVAL;
4896 			break;
4897 		}
4898 		count = max(io->pfrio_size, io->pfrio_size2);
4899 		if (count > pf_ioctl_maxcount ||
4900 		    WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
4901 			error = EINVAL;
4902 			break;
4903 		}
4904 		totlen = count * sizeof(struct pfr_addr);
4905 		pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
4906 		    M_WAITOK);
4907 		error = copyin(io->pfrio_buffer, pfras, totlen);
4908 		if (error) {
4909 			free(pfras, M_TEMP);
4910 			break;
4911 		}
4912 		PF_RULES_WLOCK();
4913 		error = pfr_set_addrs(&io->pfrio_table, pfras,
4914 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
4915 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
4916 		    PFR_FLAG_USERIOCTL, 0);
4917 		PF_RULES_WUNLOCK();
4918 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4919 			error = copyout(pfras, io->pfrio_buffer, totlen);
4920 		free(pfras, M_TEMP);
4921 		break;
4922 	}
4923 
4924 	case DIOCRGETADDRS: {
4925 		struct pfioc_table *io = (struct pfioc_table *)addr;
4926 		struct pfr_addr *pfras;
4927 		size_t totlen;
4928 
4929 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4930 			error = ENODEV;
4931 			break;
4932 		}
4933 		if (io->pfrio_size < 0 ||
4934 		    io->pfrio_size > pf_ioctl_maxcount ||
4935 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4936 			error = EINVAL;
4937 			break;
4938 		}
4939 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4940 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4941 		    M_TEMP, M_WAITOK | M_ZERO);
4942 		PF_RULES_RLOCK();
4943 		error = pfr_get_addrs(&io->pfrio_table, pfras,
4944 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4945 		PF_RULES_RUNLOCK();
4946 		if (error == 0)
4947 			error = copyout(pfras, io->pfrio_buffer, totlen);
4948 		free(pfras, M_TEMP);
4949 		break;
4950 	}
4951 
4952 	case DIOCRGETASTATS: {
4953 		struct pfioc_table *io = (struct pfioc_table *)addr;
4954 		struct pfr_astats *pfrastats;
4955 		size_t totlen;
4956 
4957 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
4958 			error = ENODEV;
4959 			break;
4960 		}
4961 		if (io->pfrio_size < 0 ||
4962 		    io->pfrio_size > pf_ioctl_maxcount ||
4963 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
4964 			error = EINVAL;
4965 			break;
4966 		}
4967 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
4968 		pfrastats = mallocarray(io->pfrio_size,
4969 		    sizeof(struct pfr_astats), M_TEMP, M_WAITOK | M_ZERO);
4970 		PF_RULES_RLOCK();
4971 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
4972 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4973 		PF_RULES_RUNLOCK();
4974 		if (error == 0)
4975 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
4976 		free(pfrastats, M_TEMP);
4977 		break;
4978 	}
4979 
4980 	case DIOCRCLRASTATS: {
4981 		struct pfioc_table *io = (struct pfioc_table *)addr;
4982 		struct pfr_addr *pfras;
4983 		size_t totlen;
4984 
4985 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4986 			error = ENODEV;
4987 			break;
4988 		}
4989 		if (io->pfrio_size < 0 ||
4990 		    io->pfrio_size > pf_ioctl_maxcount ||
4991 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4992 			error = EINVAL;
4993 			break;
4994 		}
4995 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4996 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4997 		    M_TEMP, M_WAITOK);
4998 		error = copyin(io->pfrio_buffer, pfras, totlen);
4999 		if (error) {
5000 			free(pfras, M_TEMP);
5001 			break;
5002 		}
5003 		PF_RULES_WLOCK();
5004 		error = pfr_clr_astats(&io->pfrio_table, pfras,
5005 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
5006 		    PFR_FLAG_USERIOCTL);
5007 		PF_RULES_WUNLOCK();
5008 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
5009 			error = copyout(pfras, io->pfrio_buffer, totlen);
5010 		free(pfras, M_TEMP);
5011 		break;
5012 	}
5013 
5014 	case DIOCRTSTADDRS: {
5015 		struct pfioc_table *io = (struct pfioc_table *)addr;
5016 		struct pfr_addr *pfras;
5017 		size_t totlen;
5018 
5019 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5020 			error = ENODEV;
5021 			break;
5022 		}
5023 		if (io->pfrio_size < 0 ||
5024 		    io->pfrio_size > pf_ioctl_maxcount ||
5025 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5026 			error = EINVAL;
5027 			break;
5028 		}
5029 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
5030 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5031 		    M_TEMP, M_WAITOK);
5032 		error = copyin(io->pfrio_buffer, pfras, totlen);
5033 		if (error) {
5034 			free(pfras, M_TEMP);
5035 			break;
5036 		}
5037 		PF_RULES_RLOCK();
5038 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
5039 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
5040 		    PFR_FLAG_USERIOCTL);
5041 		PF_RULES_RUNLOCK();
5042 		if (error == 0)
5043 			error = copyout(pfras, io->pfrio_buffer, totlen);
5044 		free(pfras, M_TEMP);
5045 		break;
5046 	}
5047 
5048 	case DIOCRINADEFINE: {
5049 		struct pfioc_table *io = (struct pfioc_table *)addr;
5050 		struct pfr_addr *pfras;
5051 		size_t totlen;
5052 
5053 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5054 			error = ENODEV;
5055 			break;
5056 		}
5057 		if (io->pfrio_size < 0 ||
5058 		    io->pfrio_size > pf_ioctl_maxcount ||
5059 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5060 			error = EINVAL;
5061 			break;
5062 		}
5063 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
5064 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5065 		    M_TEMP, M_WAITOK);
5066 		error = copyin(io->pfrio_buffer, pfras, totlen);
5067 		if (error) {
5068 			free(pfras, M_TEMP);
5069 			break;
5070 		}
5071 		PF_RULES_WLOCK();
5072 		error = pfr_ina_define(&io->pfrio_table, pfras,
5073 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
5074 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5075 		PF_RULES_WUNLOCK();
5076 		free(pfras, M_TEMP);
5077 		break;
5078 	}
5079 
5080 	case DIOCOSFPADD: {
5081 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5082 		PF_RULES_WLOCK();
5083 		error = pf_osfp_add(io);
5084 		PF_RULES_WUNLOCK();
5085 		break;
5086 	}
5087 
5088 	case DIOCOSFPGET: {
5089 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5090 		PF_RULES_RLOCK();
5091 		error = pf_osfp_get(io);
5092 		PF_RULES_RUNLOCK();
5093 		break;
5094 	}
5095 
5096 	case DIOCXBEGIN: {
5097 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5098 		struct pfioc_trans_e	*ioes, *ioe;
5099 		size_t			 totlen;
5100 		int			 i;
5101 
5102 		if (io->esize != sizeof(*ioe)) {
5103 			error = ENODEV;
5104 			break;
5105 		}
5106 		if (io->size < 0 ||
5107 		    io->size > pf_ioctl_maxcount ||
5108 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5109 			error = EINVAL;
5110 			break;
5111 		}
5112 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5113 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5114 		    M_TEMP, M_WAITOK);
5115 		error = copyin(io->array, ioes, totlen);
5116 		if (error) {
5117 			free(ioes, M_TEMP);
5118 			break;
5119 		}
5120 		/* Ensure there's no more ethernet rules to clean up. */
5121 		epoch_drain_callbacks(net_epoch_preempt);
5122 		PF_RULES_WLOCK();
5123 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5124 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5125 			switch (ioe->rs_num) {
5126 			case PF_RULESET_ETH:
5127 				if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) {
5128 					PF_RULES_WUNLOCK();
5129 					free(ioes, M_TEMP);
5130 					goto fail;
5131 				}
5132 				break;
5133 #ifdef ALTQ
5134 			case PF_RULESET_ALTQ:
5135 				if (ioe->anchor[0]) {
5136 					PF_RULES_WUNLOCK();
5137 					free(ioes, M_TEMP);
5138 					error = EINVAL;
5139 					goto fail;
5140 				}
5141 				if ((error = pf_begin_altq(&ioe->ticket))) {
5142 					PF_RULES_WUNLOCK();
5143 					free(ioes, M_TEMP);
5144 					goto fail;
5145 				}
5146 				break;
5147 #endif /* ALTQ */
5148 			case PF_RULESET_TABLE:
5149 			    {
5150 				struct pfr_table table;
5151 
5152 				bzero(&table, sizeof(table));
5153 				strlcpy(table.pfrt_anchor, ioe->anchor,
5154 				    sizeof(table.pfrt_anchor));
5155 				if ((error = pfr_ina_begin(&table,
5156 				    &ioe->ticket, NULL, 0))) {
5157 					PF_RULES_WUNLOCK();
5158 					free(ioes, M_TEMP);
5159 					goto fail;
5160 				}
5161 				break;
5162 			    }
5163 			default:
5164 				if ((error = pf_begin_rules(&ioe->ticket,
5165 				    ioe->rs_num, ioe->anchor))) {
5166 					PF_RULES_WUNLOCK();
5167 					free(ioes, M_TEMP);
5168 					goto fail;
5169 				}
5170 				break;
5171 			}
5172 		}
5173 		PF_RULES_WUNLOCK();
5174 		error = copyout(ioes, io->array, totlen);
5175 		free(ioes, M_TEMP);
5176 		break;
5177 	}
5178 
5179 	case DIOCXROLLBACK: {
5180 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5181 		struct pfioc_trans_e	*ioe, *ioes;
5182 		size_t			 totlen;
5183 		int			 i;
5184 
5185 		if (io->esize != sizeof(*ioe)) {
5186 			error = ENODEV;
5187 			break;
5188 		}
5189 		if (io->size < 0 ||
5190 		    io->size > pf_ioctl_maxcount ||
5191 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5192 			error = EINVAL;
5193 			break;
5194 		}
5195 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5196 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5197 		    M_TEMP, M_WAITOK);
5198 		error = copyin(io->array, ioes, totlen);
5199 		if (error) {
5200 			free(ioes, M_TEMP);
5201 			break;
5202 		}
5203 		PF_RULES_WLOCK();
5204 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5205 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5206 			switch (ioe->rs_num) {
5207 			case PF_RULESET_ETH:
5208 				if ((error = pf_rollback_eth(ioe->ticket,
5209 				    ioe->anchor))) {
5210 					PF_RULES_WUNLOCK();
5211 					free(ioes, M_TEMP);
5212 					goto fail; /* really bad */
5213 				}
5214 				break;
5215 #ifdef ALTQ
5216 			case PF_RULESET_ALTQ:
5217 				if (ioe->anchor[0]) {
5218 					PF_RULES_WUNLOCK();
5219 					free(ioes, M_TEMP);
5220 					error = EINVAL;
5221 					goto fail;
5222 				}
5223 				if ((error = pf_rollback_altq(ioe->ticket))) {
5224 					PF_RULES_WUNLOCK();
5225 					free(ioes, M_TEMP);
5226 					goto fail; /* really bad */
5227 				}
5228 				break;
5229 #endif /* ALTQ */
5230 			case PF_RULESET_TABLE:
5231 			    {
5232 				struct pfr_table table;
5233 
5234 				bzero(&table, sizeof(table));
5235 				strlcpy(table.pfrt_anchor, ioe->anchor,
5236 				    sizeof(table.pfrt_anchor));
5237 				if ((error = pfr_ina_rollback(&table,
5238 				    ioe->ticket, NULL, 0))) {
5239 					PF_RULES_WUNLOCK();
5240 					free(ioes, M_TEMP);
5241 					goto fail; /* really bad */
5242 				}
5243 				break;
5244 			    }
5245 			default:
5246 				if ((error = pf_rollback_rules(ioe->ticket,
5247 				    ioe->rs_num, ioe->anchor))) {
5248 					PF_RULES_WUNLOCK();
5249 					free(ioes, M_TEMP);
5250 					goto fail; /* really bad */
5251 				}
5252 				break;
5253 			}
5254 		}
5255 		PF_RULES_WUNLOCK();
5256 		free(ioes, M_TEMP);
5257 		break;
5258 	}
5259 
5260 	case DIOCXCOMMIT: {
5261 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5262 		struct pfioc_trans_e	*ioe, *ioes;
5263 		struct pf_kruleset	*rs;
5264 		struct pf_keth_ruleset	*ers;
5265 		size_t			 totlen;
5266 		int			 i;
5267 
5268 		if (io->esize != sizeof(*ioe)) {
5269 			error = ENODEV;
5270 			break;
5271 		}
5272 
5273 		if (io->size < 0 ||
5274 		    io->size > pf_ioctl_maxcount ||
5275 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5276 			error = EINVAL;
5277 			break;
5278 		}
5279 
5280 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5281 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5282 		    M_TEMP, M_WAITOK);
5283 		error = copyin(io->array, ioes, totlen);
5284 		if (error) {
5285 			free(ioes, M_TEMP);
5286 			break;
5287 		}
5288 		PF_RULES_WLOCK();
5289 		/* First makes sure everything will succeed. */
5290 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5291 			ioe->anchor[sizeof(ioe->anchor) - 1] = 0;
5292 			switch (ioe->rs_num) {
5293 			case PF_RULESET_ETH:
5294 				ers = pf_find_keth_ruleset(ioe->anchor);
5295 				if (ers == NULL || ioe->ticket == 0 ||
5296 				    ioe->ticket != ers->inactive.ticket) {
5297 					PF_RULES_WUNLOCK();
5298 					free(ioes, M_TEMP);
5299 					error = EINVAL;
5300 					goto fail;
5301 				}
5302 				break;
5303 #ifdef ALTQ
5304 			case PF_RULESET_ALTQ:
5305 				if (ioe->anchor[0]) {
5306 					PF_RULES_WUNLOCK();
5307 					free(ioes, M_TEMP);
5308 					error = EINVAL;
5309 					goto fail;
5310 				}
5311 				if (!V_altqs_inactive_open || ioe->ticket !=
5312 				    V_ticket_altqs_inactive) {
5313 					PF_RULES_WUNLOCK();
5314 					free(ioes, M_TEMP);
5315 					error = EBUSY;
5316 					goto fail;
5317 				}
5318 				break;
5319 #endif /* ALTQ */
5320 			case PF_RULESET_TABLE:
5321 				rs = pf_find_kruleset(ioe->anchor);
5322 				if (rs == NULL || !rs->topen || ioe->ticket !=
5323 				    rs->tticket) {
5324 					PF_RULES_WUNLOCK();
5325 					free(ioes, M_TEMP);
5326 					error = EBUSY;
5327 					goto fail;
5328 				}
5329 				break;
5330 			default:
5331 				if (ioe->rs_num < 0 || ioe->rs_num >=
5332 				    PF_RULESET_MAX) {
5333 					PF_RULES_WUNLOCK();
5334 					free(ioes, M_TEMP);
5335 					error = EINVAL;
5336 					goto fail;
5337 				}
5338 				rs = pf_find_kruleset(ioe->anchor);
5339 				if (rs == NULL ||
5340 				    !rs->rules[ioe->rs_num].inactive.open ||
5341 				    rs->rules[ioe->rs_num].inactive.ticket !=
5342 				    ioe->ticket) {
5343 					PF_RULES_WUNLOCK();
5344 					free(ioes, M_TEMP);
5345 					error = EBUSY;
5346 					goto fail;
5347 				}
5348 				break;
5349 			}
5350 		}
5351 		/* Now do the commit - no errors should happen here. */
5352 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5353 			switch (ioe->rs_num) {
5354 			case PF_RULESET_ETH:
5355 				if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) {
5356 					PF_RULES_WUNLOCK();
5357 					free(ioes, M_TEMP);
5358 					goto fail; /* really bad */
5359 				}
5360 				break;
5361 #ifdef ALTQ
5362 			case PF_RULESET_ALTQ:
5363 				if ((error = pf_commit_altq(ioe->ticket))) {
5364 					PF_RULES_WUNLOCK();
5365 					free(ioes, M_TEMP);
5366 					goto fail; /* really bad */
5367 				}
5368 				break;
5369 #endif /* ALTQ */
5370 			case PF_RULESET_TABLE:
5371 			    {
5372 				struct pfr_table table;
5373 
5374 				bzero(&table, sizeof(table));
5375 				(void)strlcpy(table.pfrt_anchor, ioe->anchor,
5376 				    sizeof(table.pfrt_anchor));
5377 				if ((error = pfr_ina_commit(&table,
5378 				    ioe->ticket, NULL, NULL, 0))) {
5379 					PF_RULES_WUNLOCK();
5380 					free(ioes, M_TEMP);
5381 					goto fail; /* really bad */
5382 				}
5383 				break;
5384 			    }
5385 			default:
5386 				if ((error = pf_commit_rules(ioe->ticket,
5387 				    ioe->rs_num, ioe->anchor))) {
5388 					PF_RULES_WUNLOCK();
5389 					free(ioes, M_TEMP);
5390 					goto fail; /* really bad */
5391 				}
5392 				break;
5393 			}
5394 		}
5395 		PF_RULES_WUNLOCK();
5396 
5397 		/* Only hook into EtherNet taffic if we've got rules for it. */
5398 		if (! TAILQ_EMPTY(V_pf_keth->active.rules))
5399 			hook_pf_eth();
5400 		else
5401 			dehook_pf_eth();
5402 
5403 		free(ioes, M_TEMP);
5404 		break;
5405 	}
5406 
5407 	case DIOCGETSRCNODES: {
5408 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
5409 		struct pf_srchash	*sh;
5410 		struct pf_ksrc_node	*n;
5411 		struct pf_src_node	*p, *pstore;
5412 		uint32_t		 i, nr = 0;
5413 
5414 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5415 				i++, sh++) {
5416 			PF_HASHROW_LOCK(sh);
5417 			LIST_FOREACH(n, &sh->nodes, entry)
5418 				nr++;
5419 			PF_HASHROW_UNLOCK(sh);
5420 		}
5421 
5422 		psn->psn_len = min(psn->psn_len,
5423 		    sizeof(struct pf_src_node) * nr);
5424 
5425 		if (psn->psn_len == 0) {
5426 			psn->psn_len = sizeof(struct pf_src_node) * nr;
5427 			break;
5428 		}
5429 
5430 		nr = 0;
5431 
5432 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
5433 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5434 		    i++, sh++) {
5435 		    PF_HASHROW_LOCK(sh);
5436 		    LIST_FOREACH(n, &sh->nodes, entry) {
5437 
5438 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
5439 				break;
5440 
5441 			pf_src_node_copy(n, p);
5442 
5443 			p++;
5444 			nr++;
5445 		    }
5446 		    PF_HASHROW_UNLOCK(sh);
5447 		}
5448 		error = copyout(pstore, psn->psn_src_nodes,
5449 		    sizeof(struct pf_src_node) * nr);
5450 		if (error) {
5451 			free(pstore, M_TEMP);
5452 			break;
5453 		}
5454 		psn->psn_len = sizeof(struct pf_src_node) * nr;
5455 		free(pstore, M_TEMP);
5456 		break;
5457 	}
5458 
5459 	case DIOCCLRSRCNODES: {
5460 		pf_clear_srcnodes(NULL);
5461 		pf_purge_expired_src_nodes();
5462 		break;
5463 	}
5464 
5465 	case DIOCKILLSRCNODES:
5466 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
5467 		break;
5468 
5469 #ifdef COMPAT_FREEBSD13
5470 	case DIOCKEEPCOUNTERS_FREEBSD13:
5471 #endif
5472 	case DIOCKEEPCOUNTERS:
5473 		error = pf_keepcounters((struct pfioc_nv *)addr);
5474 		break;
5475 
5476 	case DIOCGETSYNCOOKIES:
5477 		error = pf_get_syncookies((struct pfioc_nv *)addr);
5478 		break;
5479 
5480 	case DIOCSETSYNCOOKIES:
5481 		error = pf_set_syncookies((struct pfioc_nv *)addr);
5482 		break;
5483 
5484 	case DIOCSETHOSTID: {
5485 		u_int32_t	*hostid = (u_int32_t *)addr;
5486 
5487 		PF_RULES_WLOCK();
5488 		if (*hostid == 0)
5489 			V_pf_status.hostid = arc4random();
5490 		else
5491 			V_pf_status.hostid = *hostid;
5492 		PF_RULES_WUNLOCK();
5493 		break;
5494 	}
5495 
5496 	case DIOCOSFPFLUSH:
5497 		PF_RULES_WLOCK();
5498 		pf_osfp_flush();
5499 		PF_RULES_WUNLOCK();
5500 		break;
5501 
5502 	case DIOCIGETIFACES: {
5503 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5504 		struct pfi_kif *ifstore;
5505 		size_t bufsiz;
5506 
5507 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
5508 			error = ENODEV;
5509 			break;
5510 		}
5511 
5512 		if (io->pfiio_size < 0 ||
5513 		    io->pfiio_size > pf_ioctl_maxcount ||
5514 		    WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
5515 			error = EINVAL;
5516 			break;
5517 		}
5518 
5519 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
5520 		ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
5521 		    M_TEMP, M_WAITOK | M_ZERO);
5522 
5523 		PF_RULES_RLOCK();
5524 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
5525 		PF_RULES_RUNLOCK();
5526 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
5527 		free(ifstore, M_TEMP);
5528 		break;
5529 	}
5530 
5531 	case DIOCSETIFFLAG: {
5532 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5533 
5534 		PF_RULES_WLOCK();
5535 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
5536 		PF_RULES_WUNLOCK();
5537 		break;
5538 	}
5539 
5540 	case DIOCCLRIFFLAG: {
5541 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5542 
5543 		PF_RULES_WLOCK();
5544 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
5545 		PF_RULES_WUNLOCK();
5546 		break;
5547 	}
5548 
5549 	default:
5550 		error = ENODEV;
5551 		break;
5552 	}
5553 fail:
5554 	if (sx_xlocked(&pf_ioctl_lock))
5555 		sx_xunlock(&pf_ioctl_lock);
5556 	CURVNET_RESTORE();
5557 
5558 #undef ERROUT_IOCTL
5559 
5560 	return (error);
5561 }
5562 
5563 void
5564 pfsync_state_export(struct pfsync_state *sp, struct pf_kstate *st)
5565 {
5566 	bzero(sp, sizeof(struct pfsync_state));
5567 
5568 	/* copy from state key */
5569 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5570 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5571 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5572 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5573 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5574 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5575 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5576 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5577 	sp->proto = st->key[PF_SK_WIRE]->proto;
5578 	sp->af = st->key[PF_SK_WIRE]->af;
5579 
5580 	/* copy from state */
5581 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
5582 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
5583 	sp->creation = htonl(time_uptime - st->creation);
5584 	sp->expire = pf_state_expires(st);
5585 	if (sp->expire <= time_uptime)
5586 		sp->expire = htonl(0);
5587 	else
5588 		sp->expire = htonl(sp->expire - time_uptime);
5589 
5590 	sp->direction = st->direction;
5591 	sp->log = st->log;
5592 	sp->timeout = st->timeout;
5593 	sp->state_flags = st->state_flags;
5594 	if (st->src_node)
5595 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
5596 	if (st->nat_src_node)
5597 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5598 
5599 	sp->id = st->id;
5600 	sp->creatorid = st->creatorid;
5601 	pf_state_peer_hton(&st->src, &sp->src);
5602 	pf_state_peer_hton(&st->dst, &sp->dst);
5603 
5604 	if (st->rule.ptr == NULL)
5605 		sp->rule = htonl(-1);
5606 	else
5607 		sp->rule = htonl(st->rule.ptr->nr);
5608 	if (st->anchor.ptr == NULL)
5609 		sp->anchor = htonl(-1);
5610 	else
5611 		sp->anchor = htonl(st->anchor.ptr->nr);
5612 	if (st->nat_rule.ptr == NULL)
5613 		sp->nat_rule = htonl(-1);
5614 	else
5615 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
5616 
5617 	pf_state_counter_hton(st->packets[0], sp->packets[0]);
5618 	pf_state_counter_hton(st->packets[1], sp->packets[1]);
5619 	pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
5620 	pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
5621 }
5622 
5623 void
5624 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st)
5625 {
5626 	bzero(sp, sizeof(*sp));
5627 
5628 	sp->version = PF_STATE_VERSION;
5629 
5630 	/* copy from state key */
5631 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5632 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5633 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5634 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5635 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5636 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5637 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5638 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5639 	sp->proto = st->key[PF_SK_WIRE]->proto;
5640 	sp->af = st->key[PF_SK_WIRE]->af;
5641 
5642 	/* copy from state */
5643 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
5644 	strlcpy(sp->orig_ifname, st->orig_kif->pfik_name,
5645 	    sizeof(sp->orig_ifname));
5646 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
5647 	sp->creation = htonl(time_uptime - st->creation);
5648 	sp->expire = pf_state_expires(st);
5649 	if (sp->expire <= time_uptime)
5650 		sp->expire = htonl(0);
5651 	else
5652 		sp->expire = htonl(sp->expire - time_uptime);
5653 
5654 	sp->direction = st->direction;
5655 	sp->log = st->log;
5656 	sp->timeout = st->timeout;
5657 	sp->state_flags = st->state_flags;
5658 	if (st->src_node)
5659 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
5660 	if (st->nat_src_node)
5661 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5662 
5663 	sp->id = st->id;
5664 	sp->creatorid = st->creatorid;
5665 	pf_state_peer_hton(&st->src, &sp->src);
5666 	pf_state_peer_hton(&st->dst, &sp->dst);
5667 
5668 	if (st->rule.ptr == NULL)
5669 		sp->rule = htonl(-1);
5670 	else
5671 		sp->rule = htonl(st->rule.ptr->nr);
5672 	if (st->anchor.ptr == NULL)
5673 		sp->anchor = htonl(-1);
5674 	else
5675 		sp->anchor = htonl(st->anchor.ptr->nr);
5676 	if (st->nat_rule.ptr == NULL)
5677 		sp->nat_rule = htonl(-1);
5678 	else
5679 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
5680 
5681 	sp->packets[0] = st->packets[0];
5682 	sp->packets[1] = st->packets[1];
5683 	sp->bytes[0] = st->bytes[0];
5684 	sp->bytes[1] = st->bytes[1];
5685 }
5686 
5687 static void
5688 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
5689 {
5690 	struct pfr_ktable *kt;
5691 
5692 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
5693 
5694 	kt = aw->p.tbl;
5695 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
5696 		kt = kt->pfrkt_root;
5697 	aw->p.tbl = NULL;
5698 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
5699 		kt->pfrkt_cnt : -1;
5700 }
5701 
5702 static int
5703 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters,
5704     size_t number, char **names)
5705 {
5706 	nvlist_t        *nvc;
5707 
5708 	nvc = nvlist_create(0);
5709 	if (nvc == NULL)
5710 		return (ENOMEM);
5711 
5712 	for (int i = 0; i < number; i++) {
5713 		nvlist_append_number_array(nvc, "counters",
5714 		    counter_u64_fetch(counters[i]));
5715 		nvlist_append_string_array(nvc, "names",
5716 		    names[i]);
5717 		nvlist_append_number_array(nvc, "ids",
5718 		    i);
5719 	}
5720 	nvlist_add_nvlist(nvl, name, nvc);
5721 	nvlist_destroy(nvc);
5722 
5723 	return (0);
5724 }
5725 
5726 static int
5727 pf_getstatus(struct pfioc_nv *nv)
5728 {
5729 	nvlist_t        *nvl = NULL, *nvc = NULL;
5730 	void            *nvlpacked = NULL;
5731 	int              error;
5732 	struct pf_status s;
5733 	char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES;
5734 	char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES;
5735 	char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES;
5736 	PF_RULES_RLOCK_TRACKER;
5737 
5738 #define ERROUT(x)      ERROUT_FUNCTION(errout, x)
5739 
5740 	PF_RULES_RLOCK();
5741 
5742 	nvl = nvlist_create(0);
5743 	if (nvl == NULL)
5744 		ERROUT(ENOMEM);
5745 
5746 	nvlist_add_bool(nvl, "running", V_pf_status.running);
5747 	nvlist_add_number(nvl, "since", V_pf_status.since);
5748 	nvlist_add_number(nvl, "debug", V_pf_status.debug);
5749 	nvlist_add_number(nvl, "hostid", V_pf_status.hostid);
5750 	nvlist_add_number(nvl, "states", V_pf_status.states);
5751 	nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes);
5752 
5753 	/* counters */
5754 	error = pf_add_status_counters(nvl, "counters", V_pf_status.counters,
5755 	    PFRES_MAX, pf_reasons);
5756 	if (error != 0)
5757 		ERROUT(error);
5758 
5759 	/* lcounters */
5760 	error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters,
5761 	    KLCNT_MAX, pf_lcounter);
5762 	if (error != 0)
5763 		ERROUT(error);
5764 
5765 	/* fcounters */
5766 	nvc = nvlist_create(0);
5767 	if (nvc == NULL)
5768 		ERROUT(ENOMEM);
5769 
5770 	for (int i = 0; i < FCNT_MAX; i++) {
5771 		nvlist_append_number_array(nvc, "counters",
5772 		    pf_counter_u64_fetch(&V_pf_status.fcounters[i]));
5773 		nvlist_append_string_array(nvc, "names",
5774 		    pf_fcounter[i]);
5775 		nvlist_append_number_array(nvc, "ids",
5776 		    i);
5777 	}
5778 	nvlist_add_nvlist(nvl, "fcounters", nvc);
5779 	nvlist_destroy(nvc);
5780 	nvc = NULL;
5781 
5782 	/* scounters */
5783 	error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters,
5784 	    SCNT_MAX, pf_fcounter);
5785 	if (error != 0)
5786 		ERROUT(error);
5787 
5788 	nvlist_add_string(nvl, "ifname", V_pf_status.ifname);
5789 	nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum,
5790 	    PF_MD5_DIGEST_LENGTH);
5791 
5792 	pfi_update_status(V_pf_status.ifname, &s);
5793 
5794 	/* pcounters / bcounters */
5795 	for (int i = 0; i < 2; i++) {
5796 		for (int j = 0; j < 2; j++) {
5797 			for (int k = 0; k < 2; k++) {
5798 				nvlist_append_number_array(nvl, "pcounters",
5799 				    s.pcounters[i][j][k]);
5800 			}
5801 			nvlist_append_number_array(nvl, "bcounters",
5802 			    s.bcounters[i][j]);
5803 		}
5804 	}
5805 
5806 	nvlpacked = nvlist_pack(nvl, &nv->len);
5807 	if (nvlpacked == NULL)
5808 		ERROUT(ENOMEM);
5809 
5810 	if (nv->size == 0)
5811 		ERROUT(0);
5812 	else if (nv->size < nv->len)
5813 		ERROUT(ENOSPC);
5814 
5815 	PF_RULES_RUNLOCK();
5816 	error = copyout(nvlpacked, nv->data, nv->len);
5817 	goto done;
5818 
5819 #undef ERROUT
5820 errout:
5821 	PF_RULES_RUNLOCK();
5822 done:
5823 	free(nvlpacked, M_NVLIST);
5824 	nvlist_destroy(nvc);
5825 	nvlist_destroy(nvl);
5826 
5827 	return (error);
5828 }
5829 
5830 /*
5831  * XXX - Check for version missmatch!!!
5832  */
5833 static void
5834 pf_clear_all_states(void)
5835 {
5836 	struct pf_kstate	*s;
5837 	u_int i;
5838 
5839 	for (i = 0; i <= pf_hashmask; i++) {
5840 		struct pf_idhash *ih = &V_pf_idhash[i];
5841 relock:
5842 		PF_HASHROW_LOCK(ih);
5843 		LIST_FOREACH(s, &ih->states, entry) {
5844 			s->timeout = PFTM_PURGE;
5845 			/* Don't send out individual delete messages. */
5846 			s->state_flags |= PFSTATE_NOSYNC;
5847 			pf_unlink_state(s);
5848 			goto relock;
5849 		}
5850 		PF_HASHROW_UNLOCK(ih);
5851 	}
5852 }
5853 
5854 static int
5855 pf_clear_tables(void)
5856 {
5857 	struct pfioc_table io;
5858 	int error;
5859 
5860 	bzero(&io, sizeof(io));
5861 
5862 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
5863 	    io.pfrio_flags);
5864 
5865 	return (error);
5866 }
5867 
5868 static void
5869 pf_clear_srcnodes(struct pf_ksrc_node *n)
5870 {
5871 	struct pf_kstate *s;
5872 	int i;
5873 
5874 	for (i = 0; i <= pf_hashmask; i++) {
5875 		struct pf_idhash *ih = &V_pf_idhash[i];
5876 
5877 		PF_HASHROW_LOCK(ih);
5878 		LIST_FOREACH(s, &ih->states, entry) {
5879 			if (n == NULL || n == s->src_node)
5880 				s->src_node = NULL;
5881 			if (n == NULL || n == s->nat_src_node)
5882 				s->nat_src_node = NULL;
5883 		}
5884 		PF_HASHROW_UNLOCK(ih);
5885 	}
5886 
5887 	if (n == NULL) {
5888 		struct pf_srchash *sh;
5889 
5890 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5891 		    i++, sh++) {
5892 			PF_HASHROW_LOCK(sh);
5893 			LIST_FOREACH(n, &sh->nodes, entry) {
5894 				n->expire = 1;
5895 				n->states = 0;
5896 			}
5897 			PF_HASHROW_UNLOCK(sh);
5898 		}
5899 	} else {
5900 		/* XXX: hash slot should already be locked here. */
5901 		n->expire = 1;
5902 		n->states = 0;
5903 	}
5904 }
5905 
5906 static void
5907 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
5908 {
5909 	struct pf_ksrc_node_list	 kill;
5910 
5911 	LIST_INIT(&kill);
5912 	for (int i = 0; i <= pf_srchashmask; i++) {
5913 		struct pf_srchash *sh = &V_pf_srchash[i];
5914 		struct pf_ksrc_node *sn, *tmp;
5915 
5916 		PF_HASHROW_LOCK(sh);
5917 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
5918 			if (PF_MATCHA(psnk->psnk_src.neg,
5919 			      &psnk->psnk_src.addr.v.a.addr,
5920 			      &psnk->psnk_src.addr.v.a.mask,
5921 			      &sn->addr, sn->af) &&
5922 			    PF_MATCHA(psnk->psnk_dst.neg,
5923 			      &psnk->psnk_dst.addr.v.a.addr,
5924 			      &psnk->psnk_dst.addr.v.a.mask,
5925 			      &sn->raddr, sn->af)) {
5926 				pf_unlink_src_node(sn);
5927 				LIST_INSERT_HEAD(&kill, sn, entry);
5928 				sn->expire = 1;
5929 			}
5930 		PF_HASHROW_UNLOCK(sh);
5931 	}
5932 
5933 	for (int i = 0; i <= pf_hashmask; i++) {
5934 		struct pf_idhash *ih = &V_pf_idhash[i];
5935 		struct pf_kstate *s;
5936 
5937 		PF_HASHROW_LOCK(ih);
5938 		LIST_FOREACH(s, &ih->states, entry) {
5939 			if (s->src_node && s->src_node->expire == 1)
5940 				s->src_node = NULL;
5941 			if (s->nat_src_node && s->nat_src_node->expire == 1)
5942 				s->nat_src_node = NULL;
5943 		}
5944 		PF_HASHROW_UNLOCK(ih);
5945 	}
5946 
5947 	psnk->psnk_killed = pf_free_src_nodes(&kill);
5948 }
5949 
5950 static int
5951 pf_keepcounters(struct pfioc_nv *nv)
5952 {
5953 	nvlist_t	*nvl = NULL;
5954 	void		*nvlpacked = NULL;
5955 	int		 error = 0;
5956 
5957 #define	ERROUT(x)	ERROUT_FUNCTION(on_error, x)
5958 
5959 	if (nv->len > pf_ioctl_maxcount)
5960 		ERROUT(ENOMEM);
5961 
5962 	nvlpacked = malloc(nv->len, M_TEMP, M_WAITOK);
5963 	if (nvlpacked == NULL)
5964 		ERROUT(ENOMEM);
5965 
5966 	error = copyin(nv->data, nvlpacked, nv->len);
5967 	if (error)
5968 		ERROUT(error);
5969 
5970 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
5971 	if (nvl == NULL)
5972 		ERROUT(EBADMSG);
5973 
5974 	if (! nvlist_exists_bool(nvl, "keep_counters"))
5975 		ERROUT(EBADMSG);
5976 
5977 	V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters");
5978 
5979 on_error:
5980 	nvlist_destroy(nvl);
5981 	free(nvlpacked, M_TEMP);
5982 	return (error);
5983 }
5984 
5985 static unsigned int
5986 pf_clear_states(const struct pf_kstate_kill *kill)
5987 {
5988 	struct pf_state_key_cmp	 match_key;
5989 	struct pf_kstate	*s;
5990 	struct pfi_kkif	*kif;
5991 	int		 idx;
5992 	unsigned int	 killed = 0, dir;
5993 
5994 	for (unsigned int i = 0; i <= pf_hashmask; i++) {
5995 		struct pf_idhash *ih = &V_pf_idhash[i];
5996 
5997 relock_DIOCCLRSTATES:
5998 		PF_HASHROW_LOCK(ih);
5999 		LIST_FOREACH(s, &ih->states, entry) {
6000 			/* For floating states look at the original kif. */
6001 			kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
6002 
6003 			if (kill->psk_ifname[0] &&
6004 			    strcmp(kill->psk_ifname,
6005 			    kif->pfik_name))
6006 				continue;
6007 
6008 			if (kill->psk_kill_match) {
6009 				bzero(&match_key, sizeof(match_key));
6010 
6011 				if (s->direction == PF_OUT) {
6012 					dir = PF_IN;
6013 					idx = PF_SK_STACK;
6014 				} else {
6015 					dir = PF_OUT;
6016 					idx = PF_SK_WIRE;
6017 				}
6018 
6019 				match_key.af = s->key[idx]->af;
6020 				match_key.proto = s->key[idx]->proto;
6021 				PF_ACPY(&match_key.addr[0],
6022 				    &s->key[idx]->addr[1], match_key.af);
6023 				match_key.port[0] = s->key[idx]->port[1];
6024 				PF_ACPY(&match_key.addr[1],
6025 				    &s->key[idx]->addr[0], match_key.af);
6026 				match_key.port[1] = s->key[idx]->port[0];
6027 			}
6028 
6029 			/*
6030 			 * Don't send out individual
6031 			 * delete messages.
6032 			 */
6033 			s->state_flags |= PFSTATE_NOSYNC;
6034 			pf_unlink_state(s);
6035 			killed++;
6036 
6037 			if (kill->psk_kill_match)
6038 				killed += pf_kill_matching_state(&match_key,
6039 				    dir);
6040 
6041 			goto relock_DIOCCLRSTATES;
6042 		}
6043 		PF_HASHROW_UNLOCK(ih);
6044 	}
6045 
6046 	if (V_pfsync_clear_states_ptr != NULL)
6047 		V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname);
6048 
6049 	return (killed);
6050 }
6051 
6052 static void
6053 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed)
6054 {
6055 	struct pf_kstate	*s;
6056 
6057 	if (kill->psk_pfcmp.id) {
6058 		if (kill->psk_pfcmp.creatorid == 0)
6059 			kill->psk_pfcmp.creatorid = V_pf_status.hostid;
6060 		if ((s = pf_find_state_byid(kill->psk_pfcmp.id,
6061 		    kill->psk_pfcmp.creatorid))) {
6062 			pf_unlink_state(s);
6063 			*killed = 1;
6064 		}
6065 		return;
6066 	}
6067 
6068 	for (unsigned int i = 0; i <= pf_hashmask; i++)
6069 		*killed += pf_killstates_row(kill, &V_pf_idhash[i]);
6070 
6071 	return;
6072 }
6073 
6074 static int
6075 pf_killstates_nv(struct pfioc_nv *nv)
6076 {
6077 	struct pf_kstate_kill	 kill;
6078 	nvlist_t		*nvl = NULL;
6079 	void			*nvlpacked = NULL;
6080 	int			 error = 0;
6081 	unsigned int		 killed = 0;
6082 
6083 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
6084 
6085 	if (nv->len > pf_ioctl_maxcount)
6086 		ERROUT(ENOMEM);
6087 
6088 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6089 	if (nvlpacked == NULL)
6090 		ERROUT(ENOMEM);
6091 
6092 	error = copyin(nv->data, nvlpacked, nv->len);
6093 	if (error)
6094 		ERROUT(error);
6095 
6096 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6097 	if (nvl == NULL)
6098 		ERROUT(EBADMSG);
6099 
6100 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6101 	if (error)
6102 		ERROUT(error);
6103 
6104 	pf_killstates(&kill, &killed);
6105 
6106 	free(nvlpacked, M_NVLIST);
6107 	nvlpacked = NULL;
6108 	nvlist_destroy(nvl);
6109 	nvl = nvlist_create(0);
6110 	if (nvl == NULL)
6111 		ERROUT(ENOMEM);
6112 
6113 	nvlist_add_number(nvl, "killed", killed);
6114 
6115 	nvlpacked = nvlist_pack(nvl, &nv->len);
6116 	if (nvlpacked == NULL)
6117 		ERROUT(ENOMEM);
6118 
6119 	if (nv->size == 0)
6120 		ERROUT(0);
6121 	else if (nv->size < nv->len)
6122 		ERROUT(ENOSPC);
6123 
6124 	error = copyout(nvlpacked, nv->data, nv->len);
6125 
6126 on_error:
6127 	nvlist_destroy(nvl);
6128 	free(nvlpacked, M_NVLIST);
6129 	return (error);
6130 }
6131 
6132 static int
6133 pf_clearstates_nv(struct pfioc_nv *nv)
6134 {
6135 	struct pf_kstate_kill	 kill;
6136 	nvlist_t		*nvl = NULL;
6137 	void			*nvlpacked = NULL;
6138 	int			 error = 0;
6139 	unsigned int		 killed;
6140 
6141 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
6142 
6143 	if (nv->len > pf_ioctl_maxcount)
6144 		ERROUT(ENOMEM);
6145 
6146 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6147 	if (nvlpacked == NULL)
6148 		ERROUT(ENOMEM);
6149 
6150 	error = copyin(nv->data, nvlpacked, nv->len);
6151 	if (error)
6152 		ERROUT(error);
6153 
6154 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6155 	if (nvl == NULL)
6156 		ERROUT(EBADMSG);
6157 
6158 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6159 	if (error)
6160 		ERROUT(error);
6161 
6162 	killed = pf_clear_states(&kill);
6163 
6164 	free(nvlpacked, M_NVLIST);
6165 	nvlpacked = NULL;
6166 	nvlist_destroy(nvl);
6167 	nvl = nvlist_create(0);
6168 	if (nvl == NULL)
6169 		ERROUT(ENOMEM);
6170 
6171 	nvlist_add_number(nvl, "killed", killed);
6172 
6173 	nvlpacked = nvlist_pack(nvl, &nv->len);
6174 	if (nvlpacked == NULL)
6175 		ERROUT(ENOMEM);
6176 
6177 	if (nv->size == 0)
6178 		ERROUT(0);
6179 	else if (nv->size < nv->len)
6180 		ERROUT(ENOSPC);
6181 
6182 	error = copyout(nvlpacked, nv->data, nv->len);
6183 
6184 #undef ERROUT
6185 on_error:
6186 	nvlist_destroy(nvl);
6187 	free(nvlpacked, M_NVLIST);
6188 	return (error);
6189 }
6190 
6191 static int
6192 pf_getstate(struct pfioc_nv *nv)
6193 {
6194 	nvlist_t		*nvl = NULL, *nvls;
6195 	void			*nvlpacked = NULL;
6196 	struct pf_kstate	*s = NULL;
6197 	int			 error = 0;
6198 	uint64_t		 id, creatorid;
6199 
6200 #define ERROUT(x)	ERROUT_FUNCTION(errout, x)
6201 
6202 	if (nv->len > pf_ioctl_maxcount)
6203 		ERROUT(ENOMEM);
6204 
6205 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6206 	if (nvlpacked == NULL)
6207 		ERROUT(ENOMEM);
6208 
6209 	error = copyin(nv->data, nvlpacked, nv->len);
6210 	if (error)
6211 		ERROUT(error);
6212 
6213 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6214 	if (nvl == NULL)
6215 		ERROUT(EBADMSG);
6216 
6217 	PFNV_CHK(pf_nvuint64(nvl, "id", &id));
6218 	PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid));
6219 
6220 	s = pf_find_state_byid(id, creatorid);
6221 	if (s == NULL)
6222 		ERROUT(ENOENT);
6223 
6224 	free(nvlpacked, M_NVLIST);
6225 	nvlpacked = NULL;
6226 	nvlist_destroy(nvl);
6227 	nvl = nvlist_create(0);
6228 	if (nvl == NULL)
6229 		ERROUT(ENOMEM);
6230 
6231 	nvls = pf_state_to_nvstate(s);
6232 	if (nvls == NULL)
6233 		ERROUT(ENOMEM);
6234 
6235 	nvlist_add_nvlist(nvl, "state", nvls);
6236 	nvlist_destroy(nvls);
6237 
6238 	nvlpacked = nvlist_pack(nvl, &nv->len);
6239 	if (nvlpacked == NULL)
6240 		ERROUT(ENOMEM);
6241 
6242 	if (nv->size == 0)
6243 		ERROUT(0);
6244 	else if (nv->size < nv->len)
6245 		ERROUT(ENOSPC);
6246 
6247 	error = copyout(nvlpacked, nv->data, nv->len);
6248 
6249 #undef ERROUT
6250 errout:
6251 	if (s != NULL)
6252 		PF_STATE_UNLOCK(s);
6253 	free(nvlpacked, M_NVLIST);
6254 	nvlist_destroy(nvl);
6255 	return (error);
6256 }
6257 
6258 /*
6259  * XXX - Check for version missmatch!!!
6260  */
6261 
6262 /*
6263  * Duplicate pfctl -Fa operation to get rid of as much as we can.
6264  */
6265 static int
6266 shutdown_pf(void)
6267 {
6268 	int error = 0;
6269 	u_int32_t t[5];
6270 	char nn = '\0';
6271 
6272 	do {
6273 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
6274 		    != 0) {
6275 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
6276 			break;
6277 		}
6278 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
6279 		    != 0) {
6280 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
6281 			break;		/* XXX: rollback? */
6282 		}
6283 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
6284 		    != 0) {
6285 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
6286 			break;		/* XXX: rollback? */
6287 		}
6288 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
6289 		    != 0) {
6290 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
6291 			break;		/* XXX: rollback? */
6292 		}
6293 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
6294 		    != 0) {
6295 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
6296 			break;		/* XXX: rollback? */
6297 		}
6298 
6299 		/* XXX: these should always succeed here */
6300 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
6301 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
6302 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
6303 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
6304 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
6305 
6306 		if ((error = pf_clear_tables()) != 0)
6307 			break;
6308 
6309 		if ((error = pf_begin_eth(&t[0], &nn)) != 0) {
6310 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth\n"));
6311 			break;
6312 		}
6313 		pf_commit_eth(t[0], &nn);
6314 
6315 #ifdef ALTQ
6316 		if ((error = pf_begin_altq(&t[0])) != 0) {
6317 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
6318 			break;
6319 		}
6320 		pf_commit_altq(t[0]);
6321 #endif
6322 
6323 		pf_clear_all_states();
6324 
6325 		pf_clear_srcnodes(NULL);
6326 
6327 		/* status does not use malloced mem so no need to cleanup */
6328 		/* fingerprints and interfaces have their own cleanup code */
6329 	} while(0);
6330 
6331 	return (error);
6332 }
6333 
6334 static pfil_return_t
6335 pf_check_return(int chk, struct mbuf **m)
6336 {
6337 
6338 	switch (chk) {
6339 	case PF_PASS:
6340 		if (*m == NULL)
6341 			return (PFIL_CONSUMED);
6342 		else
6343 			return (PFIL_PASS);
6344 		break;
6345 	default:
6346 		if (*m != NULL) {
6347 			m_freem(*m);
6348 			*m = NULL;
6349 		}
6350 		return (PFIL_DROPPED);
6351 	}
6352 }
6353 
6354 static pfil_return_t
6355 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6356     void *ruleset __unused, struct inpcb *inp)
6357 {
6358 	int chk;
6359 
6360 	chk = pf_test_eth(PF_IN, flags, ifp, m, inp);
6361 
6362 	return (pf_check_return(chk, m));
6363 }
6364 
6365 static pfil_return_t
6366 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6367     void *ruleset __unused, struct inpcb *inp)
6368 {
6369 	int chk;
6370 
6371 	chk = pf_test_eth(PF_OUT, flags, ifp, m, inp);
6372 
6373 	return (pf_check_return(chk, m));
6374 }
6375 
6376 #ifdef INET
6377 static pfil_return_t
6378 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6379     void *ruleset __unused, struct inpcb *inp)
6380 {
6381 	int chk;
6382 
6383 	chk = pf_test(PF_IN, flags, ifp, m, inp);
6384 
6385 	return (pf_check_return(chk, m));
6386 }
6387 
6388 static pfil_return_t
6389 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6390     void *ruleset __unused,  struct inpcb *inp)
6391 {
6392 	int chk;
6393 
6394 	chk = pf_test(PF_OUT, flags, ifp, m, inp);
6395 
6396 	return (pf_check_return(chk, m));
6397 }
6398 #endif
6399 
6400 #ifdef INET6
6401 static pfil_return_t
6402 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
6403     void *ruleset __unused,  struct inpcb *inp)
6404 {
6405 	int chk;
6406 
6407 	/*
6408 	 * In case of loopback traffic IPv6 uses the real interface in
6409 	 * order to support scoped addresses. In order to support stateful
6410 	 * filtering we have change this to lo0 as it is the case in IPv4.
6411 	 */
6412 	CURVNET_SET(ifp->if_vnet);
6413 	chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
6414 	CURVNET_RESTORE();
6415 
6416 	return (pf_check_return(chk, m));
6417 }
6418 
6419 static pfil_return_t
6420 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
6421     void *ruleset __unused,  struct inpcb *inp)
6422 {
6423 	int chk;
6424 
6425 	CURVNET_SET(ifp->if_vnet);
6426 	chk = pf_test6(PF_OUT, flags, ifp, m, inp);
6427 	CURVNET_RESTORE();
6428 
6429 	return (pf_check_return(chk, m));
6430 }
6431 #endif /* INET6 */
6432 
6433 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook);
6434 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook);
6435 #define	V_pf_eth_in_hook	VNET(pf_eth_in_hook)
6436 #define	V_pf_eth_out_hook	VNET(pf_eth_out_hook)
6437 
6438 #ifdef INET
6439 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
6440 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
6441 #define	V_pf_ip4_in_hook	VNET(pf_ip4_in_hook)
6442 #define	V_pf_ip4_out_hook	VNET(pf_ip4_out_hook)
6443 #endif
6444 #ifdef INET6
6445 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
6446 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
6447 #define	V_pf_ip6_in_hook	VNET(pf_ip6_in_hook)
6448 #define	V_pf_ip6_out_hook	VNET(pf_ip6_out_hook)
6449 #endif
6450 
6451 static void
6452 hook_pf_eth(void)
6453 {
6454 	struct pfil_hook_args pha;
6455 	struct pfil_link_args pla;
6456 	int ret __diagused;
6457 
6458 	if (V_pf_pfil_eth_hooked)
6459 		return;
6460 
6461 	pha.pa_version = PFIL_VERSION;
6462 	pha.pa_modname = "pf";
6463 	pha.pa_ruleset = NULL;
6464 
6465 	pla.pa_version = PFIL_VERSION;
6466 
6467 	pha.pa_type = PFIL_TYPE_ETHERNET;
6468 	pha.pa_func = pf_eth_check_in;
6469 	pha.pa_flags = PFIL_IN;
6470 	pha.pa_rulname = "eth-in";
6471 	V_pf_eth_in_hook = pfil_add_hook(&pha);
6472 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6473 	pla.pa_head = V_link_pfil_head;
6474 	pla.pa_hook = V_pf_eth_in_hook;
6475 	ret = pfil_link(&pla);
6476 	MPASS(ret == 0);
6477 	pha.pa_func = pf_eth_check_out;
6478 	pha.pa_flags = PFIL_OUT;
6479 	pha.pa_rulname = "eth-out";
6480 	V_pf_eth_out_hook = pfil_add_hook(&pha);
6481 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6482 	pla.pa_head = V_link_pfil_head;
6483 	pla.pa_hook = V_pf_eth_out_hook;
6484 	ret = pfil_link(&pla);
6485 	MPASS(ret == 0);
6486 
6487 	V_pf_pfil_eth_hooked = 1;
6488 }
6489 
6490 static void
6491 hook_pf(void)
6492 {
6493 	struct pfil_hook_args pha;
6494 	struct pfil_link_args pla;
6495 	int ret __diagused;
6496 
6497 	if (V_pf_pfil_hooked)
6498 		return;
6499 
6500 	pha.pa_version = PFIL_VERSION;
6501 	pha.pa_modname = "pf";
6502 	pha.pa_ruleset = NULL;
6503 
6504 	pla.pa_version = PFIL_VERSION;
6505 
6506 #ifdef INET
6507 	pha.pa_type = PFIL_TYPE_IP4;
6508 	pha.pa_func = pf_check_in;
6509 	pha.pa_flags = PFIL_IN;
6510 	pha.pa_rulname = "default-in";
6511 	V_pf_ip4_in_hook = pfil_add_hook(&pha);
6512 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6513 	pla.pa_head = V_inet_pfil_head;
6514 	pla.pa_hook = V_pf_ip4_in_hook;
6515 	ret = pfil_link(&pla);
6516 	MPASS(ret == 0);
6517 	pha.pa_func = pf_check_out;
6518 	pha.pa_flags = PFIL_OUT;
6519 	pha.pa_rulname = "default-out";
6520 	V_pf_ip4_out_hook = pfil_add_hook(&pha);
6521 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6522 	pla.pa_head = V_inet_pfil_head;
6523 	pla.pa_hook = V_pf_ip4_out_hook;
6524 	ret = pfil_link(&pla);
6525 	MPASS(ret == 0);
6526 #endif
6527 #ifdef INET6
6528 	pha.pa_type = PFIL_TYPE_IP6;
6529 	pha.pa_func = pf_check6_in;
6530 	pha.pa_flags = PFIL_IN;
6531 	pha.pa_rulname = "default-in6";
6532 	V_pf_ip6_in_hook = pfil_add_hook(&pha);
6533 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6534 	pla.pa_head = V_inet6_pfil_head;
6535 	pla.pa_hook = V_pf_ip6_in_hook;
6536 	ret = pfil_link(&pla);
6537 	MPASS(ret == 0);
6538 	pha.pa_func = pf_check6_out;
6539 	pha.pa_rulname = "default-out6";
6540 	pha.pa_flags = PFIL_OUT;
6541 	V_pf_ip6_out_hook = pfil_add_hook(&pha);
6542 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6543 	pla.pa_head = V_inet6_pfil_head;
6544 	pla.pa_hook = V_pf_ip6_out_hook;
6545 	ret = pfil_link(&pla);
6546 	MPASS(ret == 0);
6547 #endif
6548 
6549 	V_pf_pfil_hooked = 1;
6550 }
6551 
6552 static void
6553 dehook_pf_eth(void)
6554 {
6555 
6556 	if (V_pf_pfil_eth_hooked == 0)
6557 		return;
6558 
6559 	pfil_remove_hook(V_pf_eth_in_hook);
6560 	pfil_remove_hook(V_pf_eth_out_hook);
6561 
6562 	V_pf_pfil_eth_hooked = 0;
6563 }
6564 
6565 static void
6566 dehook_pf(void)
6567 {
6568 
6569 	if (V_pf_pfil_hooked == 0)
6570 		return;
6571 
6572 #ifdef INET
6573 	pfil_remove_hook(V_pf_ip4_in_hook);
6574 	pfil_remove_hook(V_pf_ip4_out_hook);
6575 #endif
6576 #ifdef INET6
6577 	pfil_remove_hook(V_pf_ip6_in_hook);
6578 	pfil_remove_hook(V_pf_ip6_out_hook);
6579 #endif
6580 
6581 	V_pf_pfil_hooked = 0;
6582 }
6583 
6584 static void
6585 pf_load_vnet(void)
6586 {
6587 	V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
6588 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
6589 
6590 	pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
6591 	    PF_RULE_TAG_HASH_SIZE_DEFAULT);
6592 #ifdef ALTQ
6593 	pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
6594 	    PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
6595 #endif
6596 
6597 	V_pf_keth = &V_pf_main_keth_anchor.ruleset;
6598 
6599 	pfattach_vnet();
6600 	V_pf_vnet_active = 1;
6601 }
6602 
6603 static int
6604 pf_load(void)
6605 {
6606 	int error;
6607 
6608 	rm_init_flags(&pf_rules_lock, "pf rulesets", RM_RECURSE);
6609 	sx_init(&pf_ioctl_lock, "pf ioctl");
6610 	sx_init(&pf_end_lock, "pf end thread");
6611 
6612 	pf_mtag_initialize();
6613 
6614 	pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
6615 	if (pf_dev == NULL)
6616 		return (ENOMEM);
6617 
6618 	pf_end_threads = 0;
6619 	error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
6620 	if (error != 0)
6621 		return (error);
6622 
6623 	pfi_initialize();
6624 
6625 	return (0);
6626 }
6627 
6628 static void
6629 pf_unload_vnet(void)
6630 {
6631 	int ret __diagused;
6632 
6633 	V_pf_vnet_active = 0;
6634 	V_pf_status.running = 0;
6635 	dehook_pf();
6636 	dehook_pf_eth();
6637 
6638 	PF_RULES_WLOCK();
6639 	pf_syncookies_cleanup();
6640 	shutdown_pf();
6641 	PF_RULES_WUNLOCK();
6642 
6643 	ret = swi_remove(V_pf_swi_cookie);
6644 	MPASS(ret == 0);
6645 	ret = intr_event_destroy(V_pf_swi_ie);
6646 	MPASS(ret == 0);
6647 
6648 	pf_unload_vnet_purge();
6649 
6650 	pf_normalize_cleanup();
6651 	PF_RULES_WLOCK();
6652 	pfi_cleanup_vnet();
6653 	PF_RULES_WUNLOCK();
6654 	pfr_cleanup();
6655 	pf_osfp_flush();
6656 	pf_cleanup();
6657 	if (IS_DEFAULT_VNET(curvnet))
6658 		pf_mtag_cleanup();
6659 
6660 	pf_cleanup_tagset(&V_pf_tags);
6661 #ifdef ALTQ
6662 	pf_cleanup_tagset(&V_pf_qids);
6663 #endif
6664 	uma_zdestroy(V_pf_tag_z);
6665 
6666 #ifdef PF_WANT_32_TO_64_COUNTER
6667 	PF_RULES_WLOCK();
6668 	LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist);
6669 
6670 	MPASS(LIST_EMPTY(&V_pf_allkiflist));
6671 	MPASS(V_pf_allkifcount == 0);
6672 
6673 	LIST_REMOVE(&V_pf_default_rule, allrulelist);
6674 	V_pf_allrulecount--;
6675 	LIST_REMOVE(V_pf_rulemarker, allrulelist);
6676 
6677 	/*
6678 	 * There are known pf rule leaks when running the test suite.
6679 	 */
6680 #ifdef notyet
6681 	MPASS(LIST_EMPTY(&V_pf_allrulelist));
6682 	MPASS(V_pf_allrulecount == 0);
6683 #endif
6684 
6685 	PF_RULES_WUNLOCK();
6686 
6687 	free(V_pf_kifmarker, PFI_MTYPE);
6688 	free(V_pf_rulemarker, M_PFRULE);
6689 #endif
6690 
6691 	/* Free counters last as we updated them during shutdown. */
6692 	pf_counter_u64_deinit(&V_pf_default_rule.evaluations);
6693 	for (int i = 0; i < 2; i++) {
6694 		pf_counter_u64_deinit(&V_pf_default_rule.packets[i]);
6695 		pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]);
6696 	}
6697 	counter_u64_free(V_pf_default_rule.states_cur);
6698 	counter_u64_free(V_pf_default_rule.states_tot);
6699 	counter_u64_free(V_pf_default_rule.src_nodes);
6700 
6701 	for (int i = 0; i < PFRES_MAX; i++)
6702 		counter_u64_free(V_pf_status.counters[i]);
6703 	for (int i = 0; i < KLCNT_MAX; i++)
6704 		counter_u64_free(V_pf_status.lcounters[i]);
6705 	for (int i = 0; i < FCNT_MAX; i++)
6706 		pf_counter_u64_deinit(&V_pf_status.fcounters[i]);
6707 	for (int i = 0; i < SCNT_MAX; i++)
6708 		counter_u64_free(V_pf_status.scounters[i]);
6709 }
6710 
6711 static void
6712 pf_unload(void)
6713 {
6714 
6715 	sx_xlock(&pf_end_lock);
6716 	pf_end_threads = 1;
6717 	while (pf_end_threads < 2) {
6718 		wakeup_one(pf_purge_thread);
6719 		sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
6720 	}
6721 	sx_xunlock(&pf_end_lock);
6722 
6723 	if (pf_dev != NULL)
6724 		destroy_dev(pf_dev);
6725 
6726 	pfi_cleanup();
6727 
6728 	rm_destroy(&pf_rules_lock);
6729 	sx_destroy(&pf_ioctl_lock);
6730 	sx_destroy(&pf_end_lock);
6731 }
6732 
6733 static void
6734 vnet_pf_init(void *unused __unused)
6735 {
6736 
6737 	pf_load_vnet();
6738 }
6739 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6740     vnet_pf_init, NULL);
6741 
6742 static void
6743 vnet_pf_uninit(const void *unused __unused)
6744 {
6745 
6746 	pf_unload_vnet();
6747 }
6748 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
6749 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6750     vnet_pf_uninit, NULL);
6751 
6752 static int
6753 pf_modevent(module_t mod, int type, void *data)
6754 {
6755 	int error = 0;
6756 
6757 	switch(type) {
6758 	case MOD_LOAD:
6759 		error = pf_load();
6760 		break;
6761 	case MOD_UNLOAD:
6762 		/* Handled in SYSUNINIT(pf_unload) to ensure it's done after
6763 		 * the vnet_pf_uninit()s */
6764 		break;
6765 	default:
6766 		error = EINVAL;
6767 		break;
6768 	}
6769 
6770 	return (error);
6771 }
6772 
6773 static moduledata_t pf_mod = {
6774 	"pf",
6775 	pf_modevent,
6776 	0
6777 };
6778 
6779 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
6780 MODULE_VERSION(pf, PF_MODVER);
6781