xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision d0b2dbfa0ecf2bbc9709efc5e20baf8e4b44bbbf)
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 #include "opt_inet.h"
42 #include "opt_inet6.h"
43 #include "opt_bpf.h"
44 #include "opt_pf.h"
45 
46 #include <sys/param.h>
47 #include <sys/_bitset.h>
48 #include <sys/bitset.h>
49 #include <sys/bus.h>
50 #include <sys/conf.h>
51 #include <sys/endian.h>
52 #include <sys/fcntl.h>
53 #include <sys/filio.h>
54 #include <sys/hash.h>
55 #include <sys/interrupt.h>
56 #include <sys/jail.h>
57 #include <sys/kernel.h>
58 #include <sys/kthread.h>
59 #include <sys/lock.h>
60 #include <sys/mbuf.h>
61 #include <sys/module.h>
62 #include <sys/nv.h>
63 #include <sys/proc.h>
64 #include <sys/sdt.h>
65 #include <sys/smp.h>
66 #include <sys/socket.h>
67 #include <sys/sysctl.h>
68 #include <sys/md5.h>
69 #include <sys/ucred.h>
70 
71 #include <net/if.h>
72 #include <net/if_var.h>
73 #include <net/if_private.h>
74 #include <net/vnet.h>
75 #include <net/route.h>
76 #include <net/pfil.h>
77 #include <net/pfvar.h>
78 #include <net/if_pfsync.h>
79 #include <net/if_pflog.h>
80 
81 #include <netinet/in.h>
82 #include <netinet/ip.h>
83 #include <netinet/ip_var.h>
84 #include <netinet6/ip6_var.h>
85 #include <netinet/ip_icmp.h>
86 #include <netpfil/pf/pf_nv.h>
87 
88 #ifdef INET6
89 #include <netinet/ip6.h>
90 #endif /* INET6 */
91 
92 #ifdef ALTQ
93 #include <net/altq/altq.h>
94 #endif
95 
96 SDT_PROBE_DEFINE3(pf, ioctl, ioctl, error, "int", "int", "int");
97 SDT_PROBE_DEFINE3(pf, ioctl, function, error, "char *", "int", "int");
98 SDT_PROBE_DEFINE2(pf, ioctl, addrule, error, "int", "int");
99 SDT_PROBE_DEFINE2(pf, ioctl, nvchk, error, "int", "int");
100 
101 static struct pf_kpool	*pf_get_kpool(const char *, u_int32_t, u_int8_t,
102 			    u_int32_t, u_int8_t, u_int8_t, u_int8_t);
103 
104 static void		 pf_mv_kpool(struct pf_kpalist *, struct pf_kpalist *);
105 static void		 pf_empty_kpool(struct pf_kpalist *);
106 static int		 pfioctl(struct cdev *, u_long, caddr_t, int,
107 			    struct thread *);
108 static int		 pf_begin_eth(uint32_t *, const char *);
109 static void		 pf_rollback_eth_cb(struct epoch_context *);
110 static int		 pf_rollback_eth(uint32_t, const char *);
111 static int		 pf_commit_eth(uint32_t, const char *);
112 static void		 pf_free_eth_rule(struct pf_keth_rule *);
113 #ifdef ALTQ
114 static int		 pf_begin_altq(u_int32_t *);
115 static int		 pf_rollback_altq(u_int32_t);
116 static int		 pf_commit_altq(u_int32_t);
117 static int		 pf_enable_altq(struct pf_altq *);
118 static int		 pf_disable_altq(struct pf_altq *);
119 static uint16_t		 pf_qname2qid(const char *);
120 static void		 pf_qid_unref(uint16_t);
121 #endif /* ALTQ */
122 static int		 pf_begin_rules(u_int32_t *, int, const char *);
123 static int		 pf_rollback_rules(u_int32_t, int, char *);
124 static int		 pf_setup_pfsync_matching(struct pf_kruleset *);
125 static void		 pf_hash_rule_rolling(MD5_CTX *, struct pf_krule *);
126 static void		 pf_hash_rule(struct pf_krule *);
127 static void		 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
128 static int		 pf_commit_rules(u_int32_t, int, char *);
129 static int		 pf_addr_setup(struct pf_kruleset *,
130 			    struct pf_addr_wrap *, sa_family_t);
131 static void		 pf_addr_copyout(struct pf_addr_wrap *);
132 static void		 pf_src_node_copy(const struct pf_ksrc_node *,
133 			    struct pf_src_node *);
134 #ifdef ALTQ
135 static int		 pf_export_kaltq(struct pf_altq *,
136 			    struct pfioc_altq_v1 *, size_t);
137 static int		 pf_import_kaltq(struct pfioc_altq_v1 *,
138 			    struct pf_altq *, size_t);
139 #endif /* ALTQ */
140 
141 VNET_DEFINE(struct pf_krule,	pf_default_rule);
142 
143 static __inline int             pf_krule_compare(struct pf_krule *,
144 				    struct pf_krule *);
145 
146 RB_GENERATE(pf_krule_global, pf_krule, entry_global, pf_krule_compare);
147 
148 #ifdef ALTQ
149 VNET_DEFINE_STATIC(int,		pf_altq_running);
150 #define	V_pf_altq_running	VNET(pf_altq_running)
151 #endif
152 
153 #define	TAGID_MAX	 50000
154 struct pf_tagname {
155 	TAILQ_ENTRY(pf_tagname)	namehash_entries;
156 	TAILQ_ENTRY(pf_tagname)	taghash_entries;
157 	char			name[PF_TAG_NAME_SIZE];
158 	uint16_t		tag;
159 	int			ref;
160 };
161 
162 struct pf_tagset {
163 	TAILQ_HEAD(, pf_tagname)	*namehash;
164 	TAILQ_HEAD(, pf_tagname)	*taghash;
165 	unsigned int			 mask;
166 	uint32_t			 seed;
167 	BITSET_DEFINE(, TAGID_MAX)	 avail;
168 };
169 
170 VNET_DEFINE(struct pf_tagset, pf_tags);
171 #define	V_pf_tags	VNET(pf_tags)
172 static unsigned int	pf_rule_tag_hashsize;
173 #define	PF_RULE_TAG_HASH_SIZE_DEFAULT	128
174 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN,
175     &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT,
176     "Size of pf(4) rule tag hashtable");
177 
178 #ifdef ALTQ
179 VNET_DEFINE(struct pf_tagset, pf_qids);
180 #define	V_pf_qids	VNET(pf_qids)
181 static unsigned int	pf_queue_tag_hashsize;
182 #define	PF_QUEUE_TAG_HASH_SIZE_DEFAULT	128
183 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN,
184     &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT,
185     "Size of pf(4) queue tag hashtable");
186 #endif
187 VNET_DEFINE(uma_zone_t,	 pf_tag_z);
188 #define	V_pf_tag_z		 VNET(pf_tag_z)
189 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
190 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
191 
192 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
193 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
194 #endif
195 
196 VNET_DEFINE_STATIC(bool, pf_filter_local) = false;
197 #define V_pf_filter_local	VNET(pf_filter_local)
198 SYSCTL_BOOL(_net_pf, OID_AUTO, filter_local, CTLFLAG_VNET | CTLFLAG_RW,
199     &VNET_NAME(pf_filter_local), false,
200     "Enable filtering for packets delivered to local network stack");
201 
202 static void		 pf_init_tagset(struct pf_tagset *, unsigned int *,
203 			    unsigned int);
204 static void		 pf_cleanup_tagset(struct pf_tagset *);
205 static uint16_t		 tagname2hashindex(const struct pf_tagset *, const char *);
206 static uint16_t		 tag2hashindex(const struct pf_tagset *, uint16_t);
207 static u_int16_t	 tagname2tag(struct pf_tagset *, const char *);
208 static u_int16_t	 pf_tagname2tag(const char *);
209 static void		 tag_unref(struct pf_tagset *, u_int16_t);
210 
211 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
212 
213 struct cdev *pf_dev;
214 
215 /*
216  * XXX - These are new and need to be checked when moveing to a new version
217  */
218 static void		 pf_clear_all_states(void);
219 static unsigned int	 pf_clear_states(const struct pf_kstate_kill *);
220 static void		 pf_killstates(struct pf_kstate_kill *,
221 			    unsigned int *);
222 static int		 pf_killstates_row(struct pf_kstate_kill *,
223 			    struct pf_idhash *);
224 static int		 pf_killstates_nv(struct pfioc_nv *);
225 static int		 pf_clearstates_nv(struct pfioc_nv *);
226 static int		 pf_getstate(struct pfioc_nv *);
227 static int		 pf_getstatus(struct pfioc_nv *);
228 static int		 pf_clear_tables(void);
229 static void		 pf_clear_srcnodes(struct pf_ksrc_node *);
230 static void		 pf_kill_srcnodes(struct pfioc_src_node_kill *);
231 static int		 pf_keepcounters(struct pfioc_nv *);
232 static void		 pf_tbladdr_copyout(struct pf_addr_wrap *);
233 
234 /*
235  * Wrapper functions for pfil(9) hooks
236  */
237 static pfil_return_t pf_eth_check_in(struct mbuf **m, struct ifnet *ifp,
238     int flags, void *ruleset __unused, struct inpcb *inp);
239 static pfil_return_t pf_eth_check_out(struct mbuf **m, struct ifnet *ifp,
240     int flags, void *ruleset __unused, struct inpcb *inp);
241 #ifdef INET
242 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp,
243     int flags, void *ruleset __unused, struct inpcb *inp);
244 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp,
245     int flags, void *ruleset __unused, struct inpcb *inp);
246 #endif
247 #ifdef INET6
248 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp,
249     int flags, void *ruleset __unused, struct inpcb *inp);
250 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp,
251     int flags, void *ruleset __unused, struct inpcb *inp);
252 #endif
253 
254 static void		hook_pf_eth(void);
255 static void		hook_pf(void);
256 static void		dehook_pf_eth(void);
257 static void		dehook_pf(void);
258 static int		shutdown_pf(void);
259 static int		pf_load(void);
260 static void		pf_unload(void);
261 
262 static struct cdevsw pf_cdevsw = {
263 	.d_ioctl =	pfioctl,
264 	.d_name =	PF_NAME,
265 	.d_version =	D_VERSION,
266 };
267 
268 VNET_DEFINE_STATIC(bool, pf_pfil_hooked);
269 #define V_pf_pfil_hooked	VNET(pf_pfil_hooked)
270 VNET_DEFINE_STATIC(bool, pf_pfil_eth_hooked);
271 #define V_pf_pfil_eth_hooked	VNET(pf_pfil_eth_hooked)
272 
273 /*
274  * We need a flag that is neither hooked nor running to know when
275  * the VNET is "valid".  We primarily need this to control (global)
276  * external event, e.g., eventhandlers.
277  */
278 VNET_DEFINE(int, pf_vnet_active);
279 #define V_pf_vnet_active	VNET(pf_vnet_active)
280 
281 int pf_end_threads;
282 struct proc *pf_purge_proc;
283 
284 VNET_DEFINE(struct rmlock, pf_rules_lock);
285 VNET_DEFINE_STATIC(struct sx, pf_ioctl_lock);
286 #define	V_pf_ioctl_lock		VNET(pf_ioctl_lock)
287 struct sx			pf_end_lock;
288 
289 /* pfsync */
290 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
291 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
292 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
293 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
294 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
295 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
296 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
297 
298 /* pflog */
299 pflog_packet_t			*pflog_packet_ptr = NULL;
300 
301 /*
302  * Copy a user-provided string, returning an error if truncation would occur.
303  * Avoid scanning past "sz" bytes in the source string since there's no
304  * guarantee that it's nul-terminated.
305  */
306 static int
307 pf_user_strcpy(char *dst, const char *src, size_t sz)
308 {
309 	if (strnlen(src, sz) == sz)
310 		return (EINVAL);
311 	(void)strlcpy(dst, src, sz);
312 	return (0);
313 }
314 
315 static void
316 pfattach_vnet(void)
317 {
318 	u_int32_t *my_timeout = V_pf_default_rule.timeout;
319 
320 	bzero(&V_pf_status, sizeof(V_pf_status));
321 
322 	pf_initialize();
323 	pfr_initialize();
324 	pfi_initialize_vnet();
325 	pf_normalize_init();
326 	pf_syncookies_init();
327 
328 	V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
329 	V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
330 
331 	RB_INIT(&V_pf_anchors);
332 	pf_init_kruleset(&pf_main_ruleset);
333 
334 	pf_init_keth(V_pf_keth);
335 
336 	/* default rule should never be garbage collected */
337 	V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
338 #ifdef PF_DEFAULT_TO_DROP
339 	V_pf_default_rule.action = PF_DROP;
340 #else
341 	V_pf_default_rule.action = PF_PASS;
342 #endif
343 	V_pf_default_rule.nr = -1;
344 	V_pf_default_rule.rtableid = -1;
345 
346 	pf_counter_u64_init(&V_pf_default_rule.evaluations, M_WAITOK);
347 	for (int i = 0; i < 2; i++) {
348 		pf_counter_u64_init(&V_pf_default_rule.packets[i], M_WAITOK);
349 		pf_counter_u64_init(&V_pf_default_rule.bytes[i], M_WAITOK);
350 	}
351 	V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
352 	V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
353 	V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
354 
355 	V_pf_default_rule.timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
356 	    M_WAITOK | M_ZERO);
357 
358 #ifdef PF_WANT_32_TO_64_COUNTER
359 	V_pf_kifmarker = malloc(sizeof(*V_pf_kifmarker), PFI_MTYPE, M_WAITOK | M_ZERO);
360 	V_pf_rulemarker = malloc(sizeof(*V_pf_rulemarker), M_PFRULE, M_WAITOK | M_ZERO);
361 	PF_RULES_WLOCK();
362 	LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist);
363 	LIST_INSERT_HEAD(&V_pf_allrulelist, &V_pf_default_rule, allrulelist);
364 	V_pf_allrulecount++;
365 	LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist);
366 	PF_RULES_WUNLOCK();
367 #endif
368 
369 	/* initialize default timeouts */
370 	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
371 	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
372 	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
373 	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
374 	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
375 	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
376 	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
377 	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
378 	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
379 	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
380 	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
381 	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
382 	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
383 	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
384 	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
385 	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
386 	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
387 	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
388 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
389 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
390 
391 	V_pf_status.debug = PF_DEBUG_URGENT;
392 	/*
393 	 * XXX This is different than in OpenBSD where reassembly is enabled by
394 	 * defult. In FreeBSD we expect people to still use scrub rules and
395 	 * switch to the new syntax later. Only when they switch they must
396 	 * explicitly enable reassemle. We could change the default once the
397 	 * scrub rule functionality is hopefully removed some day in future.
398 	 */
399 	V_pf_status.reass = 0;
400 
401 	V_pf_pfil_hooked = false;
402 	V_pf_pfil_eth_hooked = false;
403 
404 	/* XXX do our best to avoid a conflict */
405 	V_pf_status.hostid = arc4random();
406 
407 	for (int i = 0; i < PFRES_MAX; i++)
408 		V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
409 	for (int i = 0; i < KLCNT_MAX; i++)
410 		V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
411 	for (int i = 0; i < FCNT_MAX; i++)
412 		pf_counter_u64_init(&V_pf_status.fcounters[i], M_WAITOK);
413 	for (int i = 0; i < SCNT_MAX; i++)
414 		V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
415 
416 	if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET,
417 	    INTR_MPSAFE, &V_pf_swi_cookie) != 0)
418 		/* XXXGL: leaked all above. */
419 		return;
420 }
421 
422 static struct pf_kpool *
423 pf_get_kpool(const char *anchor, u_int32_t ticket, u_int8_t rule_action,
424     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
425     u_int8_t check_ticket)
426 {
427 	struct pf_kruleset	*ruleset;
428 	struct pf_krule		*rule;
429 	int			 rs_num;
430 
431 	ruleset = pf_find_kruleset(anchor);
432 	if (ruleset == NULL)
433 		return (NULL);
434 	rs_num = pf_get_ruleset_number(rule_action);
435 	if (rs_num >= PF_RULESET_MAX)
436 		return (NULL);
437 	if (active) {
438 		if (check_ticket && ticket !=
439 		    ruleset->rules[rs_num].active.ticket)
440 			return (NULL);
441 		if (r_last)
442 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
443 			    pf_krulequeue);
444 		else
445 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
446 	} else {
447 		if (check_ticket && ticket !=
448 		    ruleset->rules[rs_num].inactive.ticket)
449 			return (NULL);
450 		if (r_last)
451 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
452 			    pf_krulequeue);
453 		else
454 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
455 	}
456 	if (!r_last) {
457 		while ((rule != NULL) && (rule->nr != rule_number))
458 			rule = TAILQ_NEXT(rule, entries);
459 	}
460 	if (rule == NULL)
461 		return (NULL);
462 
463 	return (&rule->rpool);
464 }
465 
466 static void
467 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb)
468 {
469 	struct pf_kpooladdr	*mv_pool_pa;
470 
471 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
472 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
473 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
474 	}
475 }
476 
477 static void
478 pf_empty_kpool(struct pf_kpalist *poola)
479 {
480 	struct pf_kpooladdr *pa;
481 
482 	while ((pa = TAILQ_FIRST(poola)) != NULL) {
483 		switch (pa->addr.type) {
484 		case PF_ADDR_DYNIFTL:
485 			pfi_dynaddr_remove(pa->addr.p.dyn);
486 			break;
487 		case PF_ADDR_TABLE:
488 			/* XXX: this could be unfinished pooladdr on pabuf */
489 			if (pa->addr.p.tbl != NULL)
490 				pfr_detach_table(pa->addr.p.tbl);
491 			break;
492 		}
493 		if (pa->kif)
494 			pfi_kkif_unref(pa->kif);
495 		TAILQ_REMOVE(poola, pa, entries);
496 		free(pa, M_PFRULE);
497 	}
498 }
499 
500 static void
501 pf_unlink_rule_locked(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
502 {
503 
504 	PF_RULES_WASSERT();
505 	PF_UNLNKDRULES_ASSERT();
506 
507 	TAILQ_REMOVE(rulequeue, rule, entries);
508 
509 	rule->rule_ref |= PFRULE_REFS;
510 	TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
511 }
512 
513 static void
514 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
515 {
516 
517 	PF_RULES_WASSERT();
518 
519 	PF_UNLNKDRULES_LOCK();
520 	pf_unlink_rule_locked(rulequeue, rule);
521 	PF_UNLNKDRULES_UNLOCK();
522 }
523 
524 static void
525 pf_free_eth_rule(struct pf_keth_rule *rule)
526 {
527 	PF_RULES_WASSERT();
528 
529 	if (rule == NULL)
530 		return;
531 
532 	if (rule->tag)
533 		tag_unref(&V_pf_tags, rule->tag);
534 	if (rule->match_tag)
535 		tag_unref(&V_pf_tags, rule->match_tag);
536 #ifdef ALTQ
537 	pf_qid_unref(rule->qid);
538 #endif
539 
540 	if (rule->bridge_to)
541 		pfi_kkif_unref(rule->bridge_to);
542 	if (rule->kif)
543 		pfi_kkif_unref(rule->kif);
544 
545 	if (rule->ipsrc.addr.type == PF_ADDR_TABLE)
546 		pfr_detach_table(rule->ipsrc.addr.p.tbl);
547 	if (rule->ipdst.addr.type == PF_ADDR_TABLE)
548 		pfr_detach_table(rule->ipdst.addr.p.tbl);
549 
550 	counter_u64_free(rule->evaluations);
551 	for (int i = 0; i < 2; i++) {
552 		counter_u64_free(rule->packets[i]);
553 		counter_u64_free(rule->bytes[i]);
554 	}
555 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
556 	pf_keth_anchor_remove(rule);
557 
558 	free(rule, M_PFRULE);
559 }
560 
561 void
562 pf_free_rule(struct pf_krule *rule)
563 {
564 
565 	PF_RULES_WASSERT();
566 	PF_CONFIG_ASSERT();
567 
568 	if (rule->tag)
569 		tag_unref(&V_pf_tags, rule->tag);
570 	if (rule->match_tag)
571 		tag_unref(&V_pf_tags, rule->match_tag);
572 #ifdef ALTQ
573 	if (rule->pqid != rule->qid)
574 		pf_qid_unref(rule->pqid);
575 	pf_qid_unref(rule->qid);
576 #endif
577 	switch (rule->src.addr.type) {
578 	case PF_ADDR_DYNIFTL:
579 		pfi_dynaddr_remove(rule->src.addr.p.dyn);
580 		break;
581 	case PF_ADDR_TABLE:
582 		pfr_detach_table(rule->src.addr.p.tbl);
583 		break;
584 	}
585 	switch (rule->dst.addr.type) {
586 	case PF_ADDR_DYNIFTL:
587 		pfi_dynaddr_remove(rule->dst.addr.p.dyn);
588 		break;
589 	case PF_ADDR_TABLE:
590 		pfr_detach_table(rule->dst.addr.p.tbl);
591 		break;
592 	}
593 	if (rule->overload_tbl)
594 		pfr_detach_table(rule->overload_tbl);
595 	if (rule->kif)
596 		pfi_kkif_unref(rule->kif);
597 	pf_kanchor_remove(rule);
598 	pf_empty_kpool(&rule->rpool.list);
599 
600 	pf_krule_free(rule);
601 }
602 
603 static void
604 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
605     unsigned int default_size)
606 {
607 	unsigned int i;
608 	unsigned int hashsize;
609 
610 	if (*tunable_size == 0 || !powerof2(*tunable_size))
611 		*tunable_size = default_size;
612 
613 	hashsize = *tunable_size;
614 	ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
615 	    M_WAITOK);
616 	ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
617 	    M_WAITOK);
618 	ts->mask = hashsize - 1;
619 	ts->seed = arc4random();
620 	for (i = 0; i < hashsize; i++) {
621 		TAILQ_INIT(&ts->namehash[i]);
622 		TAILQ_INIT(&ts->taghash[i]);
623 	}
624 	BIT_FILL(TAGID_MAX, &ts->avail);
625 }
626 
627 static void
628 pf_cleanup_tagset(struct pf_tagset *ts)
629 {
630 	unsigned int i;
631 	unsigned int hashsize;
632 	struct pf_tagname *t, *tmp;
633 
634 	/*
635 	 * Only need to clean up one of the hashes as each tag is hashed
636 	 * into each table.
637 	 */
638 	hashsize = ts->mask + 1;
639 	for (i = 0; i < hashsize; i++)
640 		TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
641 			uma_zfree(V_pf_tag_z, t);
642 
643 	free(ts->namehash, M_PFHASH);
644 	free(ts->taghash, M_PFHASH);
645 }
646 
647 static uint16_t
648 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
649 {
650 	size_t len;
651 
652 	len = strnlen(tagname, PF_TAG_NAME_SIZE - 1);
653 	return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask);
654 }
655 
656 static uint16_t
657 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
658 {
659 
660 	return (tag & ts->mask);
661 }
662 
663 static u_int16_t
664 tagname2tag(struct pf_tagset *ts, const char *tagname)
665 {
666 	struct pf_tagname	*tag;
667 	u_int32_t		 index;
668 	u_int16_t		 new_tagid;
669 
670 	PF_RULES_WASSERT();
671 
672 	index = tagname2hashindex(ts, tagname);
673 	TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
674 		if (strcmp(tagname, tag->name) == 0) {
675 			tag->ref++;
676 			return (tag->tag);
677 		}
678 
679 	/*
680 	 * new entry
681 	 *
682 	 * to avoid fragmentation, we do a linear search from the beginning
683 	 * and take the first free slot we find.
684 	 */
685 	new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
686 	/*
687 	 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
688 	 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
689 	 * set.  It may also return a bit number greater than TAGID_MAX due
690 	 * to rounding of the number of bits in the vector up to a multiple
691 	 * of the vector word size at declaration/allocation time.
692 	 */
693 	if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
694 		return (0);
695 
696 	/* Mark the tag as in use.  Bits are 0-based for BIT_CLR() */
697 	BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
698 
699 	/* allocate and fill new struct pf_tagname */
700 	tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
701 	if (tag == NULL)
702 		return (0);
703 	strlcpy(tag->name, tagname, sizeof(tag->name));
704 	tag->tag = new_tagid;
705 	tag->ref = 1;
706 
707 	/* Insert into namehash */
708 	TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
709 
710 	/* Insert into taghash */
711 	index = tag2hashindex(ts, new_tagid);
712 	TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
713 
714 	return (tag->tag);
715 }
716 
717 static void
718 tag_unref(struct pf_tagset *ts, u_int16_t tag)
719 {
720 	struct pf_tagname	*t;
721 	uint16_t		 index;
722 
723 	PF_RULES_WASSERT();
724 
725 	index = tag2hashindex(ts, tag);
726 	TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
727 		if (tag == t->tag) {
728 			if (--t->ref == 0) {
729 				TAILQ_REMOVE(&ts->taghash[index], t,
730 				    taghash_entries);
731 				index = tagname2hashindex(ts, t->name);
732 				TAILQ_REMOVE(&ts->namehash[index], t,
733 				    namehash_entries);
734 				/* Bits are 0-based for BIT_SET() */
735 				BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
736 				uma_zfree(V_pf_tag_z, t);
737 			}
738 			break;
739 		}
740 }
741 
742 static uint16_t
743 pf_tagname2tag(const char *tagname)
744 {
745 	return (tagname2tag(&V_pf_tags, tagname));
746 }
747 
748 static int
749 pf_begin_eth(uint32_t *ticket, const char *anchor)
750 {
751 	struct pf_keth_rule *rule, *tmp;
752 	struct pf_keth_ruleset *rs;
753 
754 	PF_RULES_WASSERT();
755 
756 	rs = pf_find_or_create_keth_ruleset(anchor);
757 	if (rs == NULL)
758 		return (EINVAL);
759 
760 	/* Purge old inactive rules. */
761 	TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
762 	    tmp) {
763 		TAILQ_REMOVE(rs->inactive.rules, rule,
764 		    entries);
765 		pf_free_eth_rule(rule);
766 	}
767 
768 	*ticket = ++rs->inactive.ticket;
769 	rs->inactive.open = 1;
770 
771 	return (0);
772 }
773 
774 static void
775 pf_rollback_eth_cb(struct epoch_context *ctx)
776 {
777 	struct pf_keth_ruleset *rs;
778 
779 	rs = __containerof(ctx, struct pf_keth_ruleset, epoch_ctx);
780 
781 	CURVNET_SET(rs->vnet);
782 
783 	PF_RULES_WLOCK();
784 	pf_rollback_eth(rs->inactive.ticket,
785 	    rs->anchor ? rs->anchor->path : "");
786 	PF_RULES_WUNLOCK();
787 
788 	CURVNET_RESTORE();
789 }
790 
791 static int
792 pf_rollback_eth(uint32_t ticket, const char *anchor)
793 {
794 	struct pf_keth_rule *rule, *tmp;
795 	struct pf_keth_ruleset *rs;
796 
797 	PF_RULES_WASSERT();
798 
799 	rs = pf_find_keth_ruleset(anchor);
800 	if (rs == NULL)
801 		return (EINVAL);
802 
803 	if (!rs->inactive.open ||
804 	    ticket != rs->inactive.ticket)
805 		return (0);
806 
807 	/* Purge old inactive rules. */
808 	TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
809 	    tmp) {
810 		TAILQ_REMOVE(rs->inactive.rules, rule, entries);
811 		pf_free_eth_rule(rule);
812 	}
813 
814 	rs->inactive.open = 0;
815 
816 	pf_remove_if_empty_keth_ruleset(rs);
817 
818 	return (0);
819 }
820 
821 #define	PF_SET_SKIP_STEPS(i)					\
822 	do {							\
823 		while (head[i] != cur) {			\
824 			head[i]->skip[i].ptr = cur;		\
825 			head[i] = TAILQ_NEXT(head[i], entries);	\
826 		}						\
827 	} while (0)
828 
829 static void
830 pf_eth_calc_skip_steps(struct pf_keth_ruleq *rules)
831 {
832 	struct pf_keth_rule *cur, *prev, *head[PFE_SKIP_COUNT];
833 	int i;
834 
835 	cur = TAILQ_FIRST(rules);
836 	prev = cur;
837 	for (i = 0; i < PFE_SKIP_COUNT; ++i)
838 		head[i] = cur;
839 	while (cur != NULL) {
840 		if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
841 			PF_SET_SKIP_STEPS(PFE_SKIP_IFP);
842 		if (cur->direction != prev->direction)
843 			PF_SET_SKIP_STEPS(PFE_SKIP_DIR);
844 		if (cur->proto != prev->proto)
845 			PF_SET_SKIP_STEPS(PFE_SKIP_PROTO);
846 		if (memcmp(&cur->src, &prev->src, sizeof(cur->src)) != 0)
847 			PF_SET_SKIP_STEPS(PFE_SKIP_SRC_ADDR);
848 		if (memcmp(&cur->dst, &prev->dst, sizeof(cur->dst)) != 0)
849 			PF_SET_SKIP_STEPS(PFE_SKIP_DST_ADDR);
850 		if (cur->ipsrc.neg != prev->ipsrc.neg ||
851 		    pf_addr_wrap_neq(&cur->ipsrc.addr, &prev->ipsrc.addr))
852 			PF_SET_SKIP_STEPS(PFE_SKIP_SRC_IP_ADDR);
853 		if (cur->ipdst.neg != prev->ipdst.neg ||
854 		    pf_addr_wrap_neq(&cur->ipdst.addr, &prev->ipdst.addr))
855 			PF_SET_SKIP_STEPS(PFE_SKIP_DST_IP_ADDR);
856 
857 		prev = cur;
858 		cur = TAILQ_NEXT(cur, entries);
859 	}
860 	for (i = 0; i < PFE_SKIP_COUNT; ++i)
861 		PF_SET_SKIP_STEPS(i);
862 }
863 
864 static int
865 pf_commit_eth(uint32_t ticket, const char *anchor)
866 {
867 	struct pf_keth_ruleq *rules;
868 	struct pf_keth_ruleset *rs;
869 
870 	rs = pf_find_keth_ruleset(anchor);
871 	if (rs == NULL) {
872 		return (EINVAL);
873 	}
874 
875 	if (!rs->inactive.open ||
876 	    ticket != rs->inactive.ticket)
877 		return (EBUSY);
878 
879 	PF_RULES_WASSERT();
880 
881 	pf_eth_calc_skip_steps(rs->inactive.rules);
882 
883 	rules = rs->active.rules;
884 	ck_pr_store_ptr(&rs->active.rules, rs->inactive.rules);
885 	rs->inactive.rules = rules;
886 	rs->inactive.ticket = rs->active.ticket;
887 
888 	/* Clean up inactive rules (i.e. previously active rules), only when
889 	 * we're sure they're no longer used. */
890 	NET_EPOCH_CALL(pf_rollback_eth_cb, &rs->epoch_ctx);
891 
892 	return (0);
893 }
894 
895 #ifdef ALTQ
896 static uint16_t
897 pf_qname2qid(const char *qname)
898 {
899 	return (tagname2tag(&V_pf_qids, qname));
900 }
901 
902 static void
903 pf_qid_unref(uint16_t qid)
904 {
905 	tag_unref(&V_pf_qids, qid);
906 }
907 
908 static int
909 pf_begin_altq(u_int32_t *ticket)
910 {
911 	struct pf_altq	*altq, *tmp;
912 	int		 error = 0;
913 
914 	PF_RULES_WASSERT();
915 
916 	/* Purge the old altq lists */
917 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
918 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
919 			/* detach and destroy the discipline */
920 			error = altq_remove(altq);
921 		}
922 		free(altq, M_PFALTQ);
923 	}
924 	TAILQ_INIT(V_pf_altq_ifs_inactive);
925 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
926 		pf_qid_unref(altq->qid);
927 		free(altq, M_PFALTQ);
928 	}
929 	TAILQ_INIT(V_pf_altqs_inactive);
930 	if (error)
931 		return (error);
932 	*ticket = ++V_ticket_altqs_inactive;
933 	V_altqs_inactive_open = 1;
934 	return (0);
935 }
936 
937 static int
938 pf_rollback_altq(u_int32_t ticket)
939 {
940 	struct pf_altq	*altq, *tmp;
941 	int		 error = 0;
942 
943 	PF_RULES_WASSERT();
944 
945 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
946 		return (0);
947 	/* Purge the old altq lists */
948 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
949 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
950 			/* detach and destroy the discipline */
951 			error = altq_remove(altq);
952 		}
953 		free(altq, M_PFALTQ);
954 	}
955 	TAILQ_INIT(V_pf_altq_ifs_inactive);
956 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
957 		pf_qid_unref(altq->qid);
958 		free(altq, M_PFALTQ);
959 	}
960 	TAILQ_INIT(V_pf_altqs_inactive);
961 	V_altqs_inactive_open = 0;
962 	return (error);
963 }
964 
965 static int
966 pf_commit_altq(u_int32_t ticket)
967 {
968 	struct pf_altqqueue	*old_altqs, *old_altq_ifs;
969 	struct pf_altq		*altq, *tmp;
970 	int			 err, error = 0;
971 
972 	PF_RULES_WASSERT();
973 
974 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
975 		return (EBUSY);
976 
977 	/* swap altqs, keep the old. */
978 	old_altqs = V_pf_altqs_active;
979 	old_altq_ifs = V_pf_altq_ifs_active;
980 	V_pf_altqs_active = V_pf_altqs_inactive;
981 	V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
982 	V_pf_altqs_inactive = old_altqs;
983 	V_pf_altq_ifs_inactive = old_altq_ifs;
984 	V_ticket_altqs_active = V_ticket_altqs_inactive;
985 
986 	/* Attach new disciplines */
987 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
988 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
989 			/* attach the discipline */
990 			error = altq_pfattach(altq);
991 			if (error == 0 && V_pf_altq_running)
992 				error = pf_enable_altq(altq);
993 			if (error != 0)
994 				return (error);
995 		}
996 	}
997 
998 	/* Purge the old altq lists */
999 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
1000 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
1001 			/* detach and destroy the discipline */
1002 			if (V_pf_altq_running)
1003 				error = pf_disable_altq(altq);
1004 			err = altq_pfdetach(altq);
1005 			if (err != 0 && error == 0)
1006 				error = err;
1007 			err = altq_remove(altq);
1008 			if (err != 0 && error == 0)
1009 				error = err;
1010 		}
1011 		free(altq, M_PFALTQ);
1012 	}
1013 	TAILQ_INIT(V_pf_altq_ifs_inactive);
1014 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
1015 		pf_qid_unref(altq->qid);
1016 		free(altq, M_PFALTQ);
1017 	}
1018 	TAILQ_INIT(V_pf_altqs_inactive);
1019 
1020 	V_altqs_inactive_open = 0;
1021 	return (error);
1022 }
1023 
1024 static int
1025 pf_enable_altq(struct pf_altq *altq)
1026 {
1027 	struct ifnet		*ifp;
1028 	struct tb_profile	 tb;
1029 	int			 error = 0;
1030 
1031 	if ((ifp = ifunit(altq->ifname)) == NULL)
1032 		return (EINVAL);
1033 
1034 	if (ifp->if_snd.altq_type != ALTQT_NONE)
1035 		error = altq_enable(&ifp->if_snd);
1036 
1037 	/* set tokenbucket regulator */
1038 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
1039 		tb.rate = altq->ifbandwidth;
1040 		tb.depth = altq->tbrsize;
1041 		error = tbr_set(&ifp->if_snd, &tb);
1042 	}
1043 
1044 	return (error);
1045 }
1046 
1047 static int
1048 pf_disable_altq(struct pf_altq *altq)
1049 {
1050 	struct ifnet		*ifp;
1051 	struct tb_profile	 tb;
1052 	int			 error;
1053 
1054 	if ((ifp = ifunit(altq->ifname)) == NULL)
1055 		return (EINVAL);
1056 
1057 	/*
1058 	 * when the discipline is no longer referenced, it was overridden
1059 	 * by a new one.  if so, just return.
1060 	 */
1061 	if (altq->altq_disc != ifp->if_snd.altq_disc)
1062 		return (0);
1063 
1064 	error = altq_disable(&ifp->if_snd);
1065 
1066 	if (error == 0) {
1067 		/* clear tokenbucket regulator */
1068 		tb.rate = 0;
1069 		error = tbr_set(&ifp->if_snd, &tb);
1070 	}
1071 
1072 	return (error);
1073 }
1074 
1075 static int
1076 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
1077     struct pf_altq *altq)
1078 {
1079 	struct ifnet	*ifp1;
1080 	int		 error = 0;
1081 
1082 	/* Deactivate the interface in question */
1083 	altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
1084 	if ((ifp1 = ifunit(altq->ifname)) == NULL ||
1085 	    (remove && ifp1 == ifp)) {
1086 		altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
1087 	} else {
1088 		error = altq_add(ifp1, altq);
1089 
1090 		if (ticket != V_ticket_altqs_inactive)
1091 			error = EBUSY;
1092 
1093 		if (error)
1094 			free(altq, M_PFALTQ);
1095 	}
1096 
1097 	return (error);
1098 }
1099 
1100 void
1101 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
1102 {
1103 	struct pf_altq	*a1, *a2, *a3;
1104 	u_int32_t	 ticket;
1105 	int		 error = 0;
1106 
1107 	/*
1108 	 * No need to re-evaluate the configuration for events on interfaces
1109 	 * that do not support ALTQ, as it's not possible for such
1110 	 * interfaces to be part of the configuration.
1111 	 */
1112 	if (!ALTQ_IS_READY(&ifp->if_snd))
1113 		return;
1114 
1115 	/* Interrupt userland queue modifications */
1116 	if (V_altqs_inactive_open)
1117 		pf_rollback_altq(V_ticket_altqs_inactive);
1118 
1119 	/* Start new altq ruleset */
1120 	if (pf_begin_altq(&ticket))
1121 		return;
1122 
1123 	/* Copy the current active set */
1124 	TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
1125 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1126 		if (a2 == NULL) {
1127 			error = ENOMEM;
1128 			break;
1129 		}
1130 		bcopy(a1, a2, sizeof(struct pf_altq));
1131 
1132 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1133 		if (error)
1134 			break;
1135 
1136 		TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
1137 	}
1138 	if (error)
1139 		goto out;
1140 	TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
1141 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1142 		if (a2 == NULL) {
1143 			error = ENOMEM;
1144 			break;
1145 		}
1146 		bcopy(a1, a2, sizeof(struct pf_altq));
1147 
1148 		if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
1149 			error = EBUSY;
1150 			free(a2, M_PFALTQ);
1151 			break;
1152 		}
1153 		a2->altq_disc = NULL;
1154 		TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
1155 			if (strncmp(a3->ifname, a2->ifname,
1156 				IFNAMSIZ) == 0) {
1157 				a2->altq_disc = a3->altq_disc;
1158 				break;
1159 			}
1160 		}
1161 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1162 		if (error)
1163 			break;
1164 
1165 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
1166 	}
1167 
1168 out:
1169 	if (error != 0)
1170 		pf_rollback_altq(ticket);
1171 	else
1172 		pf_commit_altq(ticket);
1173 }
1174 #endif /* ALTQ */
1175 
1176 static struct pf_krule_global *
1177 pf_rule_tree_alloc(int flags)
1178 {
1179 	struct pf_krule_global *tree;
1180 
1181 	tree = malloc(sizeof(struct pf_krule_global), M_TEMP, flags);
1182 	if (tree == NULL)
1183 		return (NULL);
1184 	RB_INIT(tree);
1185 	return (tree);
1186 }
1187 
1188 static void
1189 pf_rule_tree_free(struct pf_krule_global *tree)
1190 {
1191 
1192 	free(tree, M_TEMP);
1193 }
1194 
1195 static int
1196 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
1197 {
1198 	struct pf_krule_global *tree;
1199 	struct pf_kruleset	*rs;
1200 	struct pf_krule		*rule;
1201 
1202 	PF_RULES_WASSERT();
1203 
1204 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1205 		return (EINVAL);
1206 	tree = pf_rule_tree_alloc(M_NOWAIT);
1207 	if (tree == NULL)
1208 		return (ENOMEM);
1209 	rs = pf_find_or_create_kruleset(anchor);
1210 	if (rs == NULL) {
1211 		free(tree, M_TEMP);
1212 		return (EINVAL);
1213 	}
1214 	pf_rule_tree_free(rs->rules[rs_num].inactive.tree);
1215 	rs->rules[rs_num].inactive.tree = tree;
1216 
1217 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1218 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1219 		rs->rules[rs_num].inactive.rcount--;
1220 	}
1221 	*ticket = ++rs->rules[rs_num].inactive.ticket;
1222 	rs->rules[rs_num].inactive.open = 1;
1223 	return (0);
1224 }
1225 
1226 static int
1227 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
1228 {
1229 	struct pf_kruleset	*rs;
1230 	struct pf_krule		*rule;
1231 
1232 	PF_RULES_WASSERT();
1233 
1234 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1235 		return (EINVAL);
1236 	rs = pf_find_kruleset(anchor);
1237 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1238 	    rs->rules[rs_num].inactive.ticket != ticket)
1239 		return (0);
1240 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1241 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1242 		rs->rules[rs_num].inactive.rcount--;
1243 	}
1244 	rs->rules[rs_num].inactive.open = 0;
1245 	return (0);
1246 }
1247 
1248 #define PF_MD5_UPD(st, elm)						\
1249 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
1250 
1251 #define PF_MD5_UPD_STR(st, elm)						\
1252 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
1253 
1254 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
1255 		(stor) = htonl((st)->elm);				\
1256 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
1257 } while (0)
1258 
1259 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
1260 		(stor) = htons((st)->elm);				\
1261 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
1262 } while (0)
1263 
1264 static void
1265 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
1266 {
1267 	PF_MD5_UPD(pfr, addr.type);
1268 	switch (pfr->addr.type) {
1269 		case PF_ADDR_DYNIFTL:
1270 			PF_MD5_UPD(pfr, addr.v.ifname);
1271 			PF_MD5_UPD(pfr, addr.iflags);
1272 			break;
1273 		case PF_ADDR_TABLE:
1274 			PF_MD5_UPD(pfr, addr.v.tblname);
1275 			break;
1276 		case PF_ADDR_ADDRMASK:
1277 			/* XXX ignore af? */
1278 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
1279 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
1280 			break;
1281 	}
1282 
1283 	PF_MD5_UPD(pfr, port[0]);
1284 	PF_MD5_UPD(pfr, port[1]);
1285 	PF_MD5_UPD(pfr, neg);
1286 	PF_MD5_UPD(pfr, port_op);
1287 }
1288 
1289 static void
1290 pf_hash_rule_rolling(MD5_CTX *ctx, struct pf_krule *rule)
1291 {
1292 	u_int16_t x;
1293 	u_int32_t y;
1294 
1295 	pf_hash_rule_addr(ctx, &rule->src);
1296 	pf_hash_rule_addr(ctx, &rule->dst);
1297 	for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++)
1298 		PF_MD5_UPD_STR(rule, label[i]);
1299 	PF_MD5_UPD_STR(rule, ifname);
1300 	PF_MD5_UPD_STR(rule, match_tagname);
1301 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
1302 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
1303 	PF_MD5_UPD_HTONL(rule, prob, y);
1304 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
1305 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1306 	PF_MD5_UPD(rule, uid.op);
1307 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1308 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1309 	PF_MD5_UPD(rule, gid.op);
1310 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
1311 	PF_MD5_UPD(rule, action);
1312 	PF_MD5_UPD(rule, direction);
1313 	PF_MD5_UPD(rule, af);
1314 	PF_MD5_UPD(rule, quick);
1315 	PF_MD5_UPD(rule, ifnot);
1316 	PF_MD5_UPD(rule, match_tag_not);
1317 	PF_MD5_UPD(rule, natpass);
1318 	PF_MD5_UPD(rule, keep_state);
1319 	PF_MD5_UPD(rule, proto);
1320 	PF_MD5_UPD(rule, type);
1321 	PF_MD5_UPD(rule, code);
1322 	PF_MD5_UPD(rule, flags);
1323 	PF_MD5_UPD(rule, flagset);
1324 	PF_MD5_UPD(rule, allow_opts);
1325 	PF_MD5_UPD(rule, rt);
1326 	PF_MD5_UPD(rule, tos);
1327 	PF_MD5_UPD(rule, scrub_flags);
1328 	PF_MD5_UPD(rule, min_ttl);
1329 	PF_MD5_UPD(rule, set_tos);
1330 	if (rule->anchor != NULL)
1331 		PF_MD5_UPD_STR(rule, anchor->path);
1332 }
1333 
1334 static void
1335 pf_hash_rule(struct pf_krule *rule)
1336 {
1337 	MD5_CTX		ctx;
1338 
1339 	MD5Init(&ctx);
1340 	pf_hash_rule_rolling(&ctx, rule);
1341 	MD5Final(rule->md5sum, &ctx);
1342 }
1343 
1344 static int
1345 pf_krule_compare(struct pf_krule *a, struct pf_krule *b)
1346 {
1347 
1348 	return (memcmp(a->md5sum, b->md5sum, PF_MD5_DIGEST_LENGTH));
1349 }
1350 
1351 static int
1352 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1353 {
1354 	struct pf_kruleset	*rs;
1355 	struct pf_krule		*rule, **old_array, *old_rule;
1356 	struct pf_krulequeue	*old_rules;
1357 	struct pf_krule_global  *old_tree;
1358 	int			 error;
1359 	u_int32_t		 old_rcount;
1360 
1361 	PF_RULES_WASSERT();
1362 
1363 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1364 		return (EINVAL);
1365 	rs = pf_find_kruleset(anchor);
1366 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1367 	    ticket != rs->rules[rs_num].inactive.ticket)
1368 		return (EBUSY);
1369 
1370 	/* Calculate checksum for the main ruleset */
1371 	if (rs == &pf_main_ruleset) {
1372 		error = pf_setup_pfsync_matching(rs);
1373 		if (error != 0)
1374 			return (error);
1375 	}
1376 
1377 	/* Swap rules, keep the old. */
1378 	old_rules = rs->rules[rs_num].active.ptr;
1379 	old_rcount = rs->rules[rs_num].active.rcount;
1380 	old_array = rs->rules[rs_num].active.ptr_array;
1381 	old_tree = rs->rules[rs_num].active.tree;
1382 
1383 	rs->rules[rs_num].active.ptr =
1384 	    rs->rules[rs_num].inactive.ptr;
1385 	rs->rules[rs_num].active.ptr_array =
1386 	    rs->rules[rs_num].inactive.ptr_array;
1387 	rs->rules[rs_num].active.tree =
1388 	    rs->rules[rs_num].inactive.tree;
1389 	rs->rules[rs_num].active.rcount =
1390 	    rs->rules[rs_num].inactive.rcount;
1391 
1392 	/* Attempt to preserve counter information. */
1393 	if (V_pf_status.keep_counters && old_tree != NULL) {
1394 		TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr,
1395 		    entries) {
1396 			old_rule = RB_FIND(pf_krule_global, old_tree, rule);
1397 			if (old_rule == NULL) {
1398 				continue;
1399 			}
1400 			pf_counter_u64_critical_enter();
1401 			pf_counter_u64_add_protected(&rule->evaluations,
1402 			    pf_counter_u64_fetch(&old_rule->evaluations));
1403 			pf_counter_u64_add_protected(&rule->packets[0],
1404 			    pf_counter_u64_fetch(&old_rule->packets[0]));
1405 			pf_counter_u64_add_protected(&rule->packets[1],
1406 			    pf_counter_u64_fetch(&old_rule->packets[1]));
1407 			pf_counter_u64_add_protected(&rule->bytes[0],
1408 			    pf_counter_u64_fetch(&old_rule->bytes[0]));
1409 			pf_counter_u64_add_protected(&rule->bytes[1],
1410 			    pf_counter_u64_fetch(&old_rule->bytes[1]));
1411 			pf_counter_u64_critical_exit();
1412 		}
1413 	}
1414 
1415 	rs->rules[rs_num].inactive.ptr = old_rules;
1416 	rs->rules[rs_num].inactive.ptr_array = old_array;
1417 	rs->rules[rs_num].inactive.tree = NULL; /* important for pf_ioctl_addrule */
1418 	rs->rules[rs_num].inactive.rcount = old_rcount;
1419 
1420 	rs->rules[rs_num].active.ticket =
1421 	    rs->rules[rs_num].inactive.ticket;
1422 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1423 
1424 	/* Purge the old rule list. */
1425 	PF_UNLNKDRULES_LOCK();
1426 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1427 		pf_unlink_rule_locked(old_rules, rule);
1428 	PF_UNLNKDRULES_UNLOCK();
1429 	if (rs->rules[rs_num].inactive.ptr_array)
1430 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1431 	rs->rules[rs_num].inactive.ptr_array = NULL;
1432 	rs->rules[rs_num].inactive.rcount = 0;
1433 	rs->rules[rs_num].inactive.open = 0;
1434 	pf_remove_if_empty_kruleset(rs);
1435 	free(old_tree, M_TEMP);
1436 
1437 	return (0);
1438 }
1439 
1440 static int
1441 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1442 {
1443 	MD5_CTX			 ctx;
1444 	struct pf_krule		*rule;
1445 	int			 rs_cnt;
1446 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
1447 
1448 	MD5Init(&ctx);
1449 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1450 		/* XXX PF_RULESET_SCRUB as well? */
1451 		if (rs_cnt == PF_RULESET_SCRUB)
1452 			continue;
1453 
1454 		if (rs->rules[rs_cnt].inactive.ptr_array)
1455 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1456 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
1457 
1458 		if (rs->rules[rs_cnt].inactive.rcount) {
1459 			rs->rules[rs_cnt].inactive.ptr_array =
1460 			    mallocarray(rs->rules[rs_cnt].inactive.rcount,
1461 			    sizeof(struct pf_rule **),
1462 			    M_TEMP, M_NOWAIT);
1463 
1464 			if (!rs->rules[rs_cnt].inactive.ptr_array)
1465 				return (ENOMEM);
1466 		}
1467 
1468 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1469 		    entries) {
1470 			pf_hash_rule_rolling(&ctx, rule);
1471 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1472 		}
1473 	}
1474 
1475 	MD5Final(digest, &ctx);
1476 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1477 	return (0);
1478 }
1479 
1480 static int
1481 pf_eth_addr_setup(struct pf_keth_ruleset *ruleset, struct pf_addr_wrap *addr)
1482 {
1483 	int error = 0;
1484 
1485 	switch (addr->type) {
1486 	case PF_ADDR_TABLE:
1487 		addr->p.tbl = pfr_eth_attach_table(ruleset, addr->v.tblname);
1488 		if (addr->p.tbl == NULL)
1489 			error = ENOMEM;
1490 		break;
1491 	default:
1492 		error = EINVAL;
1493 	}
1494 
1495 	return (error);
1496 }
1497 
1498 static int
1499 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1500     sa_family_t af)
1501 {
1502 	int error = 0;
1503 
1504 	switch (addr->type) {
1505 	case PF_ADDR_TABLE:
1506 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1507 		if (addr->p.tbl == NULL)
1508 			error = ENOMEM;
1509 		break;
1510 	case PF_ADDR_DYNIFTL:
1511 		error = pfi_dynaddr_setup(addr, af);
1512 		break;
1513 	}
1514 
1515 	return (error);
1516 }
1517 
1518 static void
1519 pf_addr_copyout(struct pf_addr_wrap *addr)
1520 {
1521 
1522 	switch (addr->type) {
1523 	case PF_ADDR_DYNIFTL:
1524 		pfi_dynaddr_copyout(addr);
1525 		break;
1526 	case PF_ADDR_TABLE:
1527 		pf_tbladdr_copyout(addr);
1528 		break;
1529 	}
1530 }
1531 
1532 static void
1533 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
1534 {
1535 	int	secs = time_uptime, diff;
1536 
1537 	bzero(out, sizeof(struct pf_src_node));
1538 
1539 	bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
1540 	bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
1541 
1542 	if (in->rule.ptr != NULL)
1543 		out->rule.nr = in->rule.ptr->nr;
1544 
1545 	for (int i = 0; i < 2; i++) {
1546 		out->bytes[i] = counter_u64_fetch(in->bytes[i]);
1547 		out->packets[i] = counter_u64_fetch(in->packets[i]);
1548 	}
1549 
1550 	out->states = in->states;
1551 	out->conn = in->conn;
1552 	out->af = in->af;
1553 	out->ruletype = in->ruletype;
1554 
1555 	out->creation = secs - in->creation;
1556 	if (out->expire > secs)
1557 		out->expire -= secs;
1558 	else
1559 		out->expire = 0;
1560 
1561 	/* Adjust the connection rate estimate. */
1562 	diff = secs - in->conn_rate.last;
1563 	if (diff >= in->conn_rate.seconds)
1564 		out->conn_rate.count = 0;
1565 	else
1566 		out->conn_rate.count -=
1567 		    in->conn_rate.count * diff /
1568 		    in->conn_rate.seconds;
1569 }
1570 
1571 #ifdef ALTQ
1572 /*
1573  * Handle export of struct pf_kaltq to user binaries that may be using any
1574  * version of struct pf_altq.
1575  */
1576 static int
1577 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1578 {
1579 	u_int32_t version;
1580 
1581 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1582 		version = 0;
1583 	else
1584 		version = pa->version;
1585 
1586 	if (version > PFIOC_ALTQ_VERSION)
1587 		return (EINVAL);
1588 
1589 #define ASSIGN(x) exported_q->x = q->x
1590 #define COPY(x) \
1591 	bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1592 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1593 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1594 
1595 	switch (version) {
1596 	case 0: {
1597 		struct pf_altq_v0 *exported_q =
1598 		    &((struct pfioc_altq_v0 *)pa)->altq;
1599 
1600 		COPY(ifname);
1601 
1602 		ASSIGN(scheduler);
1603 		ASSIGN(tbrsize);
1604 		exported_q->tbrsize = SATU16(q->tbrsize);
1605 		exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1606 
1607 		COPY(qname);
1608 		COPY(parent);
1609 		ASSIGN(parent_qid);
1610 		exported_q->bandwidth = SATU32(q->bandwidth);
1611 		ASSIGN(priority);
1612 		ASSIGN(local_flags);
1613 
1614 		ASSIGN(qlimit);
1615 		ASSIGN(flags);
1616 
1617 		if (q->scheduler == ALTQT_HFSC) {
1618 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1619 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1620 			    SATU32(q->pq_u.hfsc_opts.x)
1621 
1622 			ASSIGN_OPT_SATU32(rtsc_m1);
1623 			ASSIGN_OPT(rtsc_d);
1624 			ASSIGN_OPT_SATU32(rtsc_m2);
1625 
1626 			ASSIGN_OPT_SATU32(lssc_m1);
1627 			ASSIGN_OPT(lssc_d);
1628 			ASSIGN_OPT_SATU32(lssc_m2);
1629 
1630 			ASSIGN_OPT_SATU32(ulsc_m1);
1631 			ASSIGN_OPT(ulsc_d);
1632 			ASSIGN_OPT_SATU32(ulsc_m2);
1633 
1634 			ASSIGN_OPT(flags);
1635 
1636 #undef ASSIGN_OPT
1637 #undef ASSIGN_OPT_SATU32
1638 		} else
1639 			COPY(pq_u);
1640 
1641 		ASSIGN(qid);
1642 		break;
1643 	}
1644 	case 1:	{
1645 		struct pf_altq_v1 *exported_q =
1646 		    &((struct pfioc_altq_v1 *)pa)->altq;
1647 
1648 		COPY(ifname);
1649 
1650 		ASSIGN(scheduler);
1651 		ASSIGN(tbrsize);
1652 		ASSIGN(ifbandwidth);
1653 
1654 		COPY(qname);
1655 		COPY(parent);
1656 		ASSIGN(parent_qid);
1657 		ASSIGN(bandwidth);
1658 		ASSIGN(priority);
1659 		ASSIGN(local_flags);
1660 
1661 		ASSIGN(qlimit);
1662 		ASSIGN(flags);
1663 		COPY(pq_u);
1664 
1665 		ASSIGN(qid);
1666 		break;
1667 	}
1668 	default:
1669 		panic("%s: unhandled struct pfioc_altq version", __func__);
1670 		break;
1671 	}
1672 
1673 #undef ASSIGN
1674 #undef COPY
1675 #undef SATU16
1676 #undef SATU32
1677 
1678 	return (0);
1679 }
1680 
1681 /*
1682  * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1683  * that may be using any version of it.
1684  */
1685 static int
1686 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1687 {
1688 	u_int32_t version;
1689 
1690 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1691 		version = 0;
1692 	else
1693 		version = pa->version;
1694 
1695 	if (version > PFIOC_ALTQ_VERSION)
1696 		return (EINVAL);
1697 
1698 #define ASSIGN(x) q->x = imported_q->x
1699 #define COPY(x) \
1700 	bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1701 
1702 	switch (version) {
1703 	case 0: {
1704 		struct pf_altq_v0 *imported_q =
1705 		    &((struct pfioc_altq_v0 *)pa)->altq;
1706 
1707 		COPY(ifname);
1708 
1709 		ASSIGN(scheduler);
1710 		ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1711 		ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1712 
1713 		COPY(qname);
1714 		COPY(parent);
1715 		ASSIGN(parent_qid);
1716 		ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1717 		ASSIGN(priority);
1718 		ASSIGN(local_flags);
1719 
1720 		ASSIGN(qlimit);
1721 		ASSIGN(flags);
1722 
1723 		if (imported_q->scheduler == ALTQT_HFSC) {
1724 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1725 
1726 			/*
1727 			 * The m1 and m2 parameters are being copied from
1728 			 * 32-bit to 64-bit.
1729 			 */
1730 			ASSIGN_OPT(rtsc_m1);
1731 			ASSIGN_OPT(rtsc_d);
1732 			ASSIGN_OPT(rtsc_m2);
1733 
1734 			ASSIGN_OPT(lssc_m1);
1735 			ASSIGN_OPT(lssc_d);
1736 			ASSIGN_OPT(lssc_m2);
1737 
1738 			ASSIGN_OPT(ulsc_m1);
1739 			ASSIGN_OPT(ulsc_d);
1740 			ASSIGN_OPT(ulsc_m2);
1741 
1742 			ASSIGN_OPT(flags);
1743 
1744 #undef ASSIGN_OPT
1745 		} else
1746 			COPY(pq_u);
1747 
1748 		ASSIGN(qid);
1749 		break;
1750 	}
1751 	case 1: {
1752 		struct pf_altq_v1 *imported_q =
1753 		    &((struct pfioc_altq_v1 *)pa)->altq;
1754 
1755 		COPY(ifname);
1756 
1757 		ASSIGN(scheduler);
1758 		ASSIGN(tbrsize);
1759 		ASSIGN(ifbandwidth);
1760 
1761 		COPY(qname);
1762 		COPY(parent);
1763 		ASSIGN(parent_qid);
1764 		ASSIGN(bandwidth);
1765 		ASSIGN(priority);
1766 		ASSIGN(local_flags);
1767 
1768 		ASSIGN(qlimit);
1769 		ASSIGN(flags);
1770 		COPY(pq_u);
1771 
1772 		ASSIGN(qid);
1773 		break;
1774 	}
1775 	default:
1776 		panic("%s: unhandled struct pfioc_altq version", __func__);
1777 		break;
1778 	}
1779 
1780 #undef ASSIGN
1781 #undef COPY
1782 
1783 	return (0);
1784 }
1785 
1786 static struct pf_altq *
1787 pf_altq_get_nth_active(u_int32_t n)
1788 {
1789 	struct pf_altq		*altq;
1790 	u_int32_t		 nr;
1791 
1792 	nr = 0;
1793 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1794 		if (nr == n)
1795 			return (altq);
1796 		nr++;
1797 	}
1798 
1799 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1800 		if (nr == n)
1801 			return (altq);
1802 		nr++;
1803 	}
1804 
1805 	return (NULL);
1806 }
1807 #endif /* ALTQ */
1808 
1809 struct pf_krule *
1810 pf_krule_alloc(void)
1811 {
1812 	struct pf_krule *rule;
1813 
1814 	rule = malloc(sizeof(struct pf_krule), M_PFRULE, M_WAITOK | M_ZERO);
1815 	mtx_init(&rule->rpool.mtx, "pf_krule_pool", NULL, MTX_DEF);
1816 	rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
1817 	    M_WAITOK | M_ZERO);
1818 	return (rule);
1819 }
1820 
1821 void
1822 pf_krule_free(struct pf_krule *rule)
1823 {
1824 #ifdef PF_WANT_32_TO_64_COUNTER
1825 	bool wowned;
1826 #endif
1827 
1828 	if (rule == NULL)
1829 		return;
1830 
1831 #ifdef PF_WANT_32_TO_64_COUNTER
1832 	if (rule->allrulelinked) {
1833 		wowned = PF_RULES_WOWNED();
1834 		if (!wowned)
1835 			PF_RULES_WLOCK();
1836 		LIST_REMOVE(rule, allrulelist);
1837 		V_pf_allrulecount--;
1838 		if (!wowned)
1839 			PF_RULES_WUNLOCK();
1840 	}
1841 #endif
1842 
1843 	pf_counter_u64_deinit(&rule->evaluations);
1844 	for (int i = 0; i < 2; i++) {
1845 		pf_counter_u64_deinit(&rule->packets[i]);
1846 		pf_counter_u64_deinit(&rule->bytes[i]);
1847 	}
1848 	counter_u64_free(rule->states_cur);
1849 	counter_u64_free(rule->states_tot);
1850 	counter_u64_free(rule->src_nodes);
1851 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
1852 
1853 	mtx_destroy(&rule->rpool.mtx);
1854 	free(rule, M_PFRULE);
1855 }
1856 
1857 static void
1858 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool,
1859     struct pf_pooladdr *pool)
1860 {
1861 
1862 	bzero(pool, sizeof(*pool));
1863 	bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr));
1864 	strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname));
1865 }
1866 
1867 static int
1868 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool,
1869     struct pf_kpooladdr *kpool)
1870 {
1871 	int ret;
1872 
1873 	bzero(kpool, sizeof(*kpool));
1874 	bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr));
1875 	ret = pf_user_strcpy(kpool->ifname, pool->ifname,
1876 	    sizeof(kpool->ifname));
1877 	return (ret);
1878 }
1879 
1880 static void
1881 pf_kpool_to_pool(const struct pf_kpool *kpool, struct pf_pool *pool)
1882 {
1883 	bzero(pool, sizeof(*pool));
1884 
1885 	bcopy(&kpool->key, &pool->key, sizeof(pool->key));
1886 	bcopy(&kpool->counter, &pool->counter, sizeof(pool->counter));
1887 
1888 	pool->tblidx = kpool->tblidx;
1889 	pool->proxy_port[0] = kpool->proxy_port[0];
1890 	pool->proxy_port[1] = kpool->proxy_port[1];
1891 	pool->opts = kpool->opts;
1892 }
1893 
1894 static void
1895 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool)
1896 {
1897 	_Static_assert(sizeof(pool->key) == sizeof(kpool->key), "");
1898 	_Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), "");
1899 
1900 	bcopy(&pool->key, &kpool->key, sizeof(kpool->key));
1901 	bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter));
1902 
1903 	kpool->tblidx = pool->tblidx;
1904 	kpool->proxy_port[0] = pool->proxy_port[0];
1905 	kpool->proxy_port[1] = pool->proxy_port[1];
1906 	kpool->opts = pool->opts;
1907 }
1908 
1909 static void
1910 pf_krule_to_rule(const struct pf_krule *krule, struct pf_rule *rule)
1911 {
1912 
1913 	bzero(rule, sizeof(*rule));
1914 
1915 	bcopy(&krule->src, &rule->src, sizeof(rule->src));
1916 	bcopy(&krule->dst, &rule->dst, sizeof(rule->dst));
1917 
1918 	for (int i = 0; i < PF_SKIP_COUNT; ++i) {
1919 		if (rule->skip[i].ptr == NULL)
1920 			rule->skip[i].nr = -1;
1921 		else
1922 			rule->skip[i].nr = krule->skip[i].ptr->nr;
1923 	}
1924 
1925 	strlcpy(rule->label, krule->label[0], sizeof(rule->label));
1926 	strlcpy(rule->ifname, krule->ifname, sizeof(rule->ifname));
1927 	strlcpy(rule->qname, krule->qname, sizeof(rule->qname));
1928 	strlcpy(rule->pqname, krule->pqname, sizeof(rule->pqname));
1929 	strlcpy(rule->tagname, krule->tagname, sizeof(rule->tagname));
1930 	strlcpy(rule->match_tagname, krule->match_tagname,
1931 	    sizeof(rule->match_tagname));
1932 	strlcpy(rule->overload_tblname, krule->overload_tblname,
1933 	    sizeof(rule->overload_tblname));
1934 
1935 	pf_kpool_to_pool(&krule->rpool, &rule->rpool);
1936 
1937 	rule->evaluations = pf_counter_u64_fetch(&krule->evaluations);
1938 	for (int i = 0; i < 2; i++) {
1939 		rule->packets[i] = pf_counter_u64_fetch(&krule->packets[i]);
1940 		rule->bytes[i] = pf_counter_u64_fetch(&krule->bytes[i]);
1941 	}
1942 
1943 	/* kif, anchor, overload_tbl are not copied over. */
1944 
1945 	rule->os_fingerprint = krule->os_fingerprint;
1946 
1947 	rule->rtableid = krule->rtableid;
1948 	bcopy(krule->timeout, rule->timeout, sizeof(krule->timeout));
1949 	rule->max_states = krule->max_states;
1950 	rule->max_src_nodes = krule->max_src_nodes;
1951 	rule->max_src_states = krule->max_src_states;
1952 	rule->max_src_conn = krule->max_src_conn;
1953 	rule->max_src_conn_rate.limit = krule->max_src_conn_rate.limit;
1954 	rule->max_src_conn_rate.seconds = krule->max_src_conn_rate.seconds;
1955 	rule->qid = krule->qid;
1956 	rule->pqid = krule->pqid;
1957 	rule->nr = krule->nr;
1958 	rule->prob = krule->prob;
1959 	rule->cuid = krule->cuid;
1960 	rule->cpid = krule->cpid;
1961 
1962 	rule->return_icmp = krule->return_icmp;
1963 	rule->return_icmp6 = krule->return_icmp6;
1964 	rule->max_mss = krule->max_mss;
1965 	rule->tag = krule->tag;
1966 	rule->match_tag = krule->match_tag;
1967 	rule->scrub_flags = krule->scrub_flags;
1968 
1969 	bcopy(&krule->uid, &rule->uid, sizeof(krule->uid));
1970 	bcopy(&krule->gid, &rule->gid, sizeof(krule->gid));
1971 
1972 	rule->rule_flag = krule->rule_flag;
1973 	rule->action = krule->action;
1974 	rule->direction = krule->direction;
1975 	rule->log = krule->log;
1976 	rule->logif = krule->logif;
1977 	rule->quick = krule->quick;
1978 	rule->ifnot = krule->ifnot;
1979 	rule->match_tag_not = krule->match_tag_not;
1980 	rule->natpass = krule->natpass;
1981 
1982 	rule->keep_state = krule->keep_state;
1983 	rule->af = krule->af;
1984 	rule->proto = krule->proto;
1985 	rule->type = krule->type;
1986 	rule->code = krule->code;
1987 	rule->flags = krule->flags;
1988 	rule->flagset = krule->flagset;
1989 	rule->min_ttl = krule->min_ttl;
1990 	rule->allow_opts = krule->allow_opts;
1991 	rule->rt = krule->rt;
1992 	rule->return_ttl = krule->return_ttl;
1993 	rule->tos = krule->tos;
1994 	rule->set_tos = krule->set_tos;
1995 	rule->anchor_relative = krule->anchor_relative;
1996 	rule->anchor_wildcard = krule->anchor_wildcard;
1997 
1998 	rule->flush = krule->flush;
1999 	rule->prio = krule->prio;
2000 	rule->set_prio[0] = krule->set_prio[0];
2001 	rule->set_prio[1] = krule->set_prio[1];
2002 
2003 	bcopy(&krule->divert, &rule->divert, sizeof(krule->divert));
2004 
2005 	rule->u_states_cur = counter_u64_fetch(krule->states_cur);
2006 	rule->u_states_tot = counter_u64_fetch(krule->states_tot);
2007 	rule->u_src_nodes = counter_u64_fetch(krule->src_nodes);
2008 }
2009 
2010 static int
2011 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
2012 {
2013 	int ret;
2014 
2015 #ifndef INET
2016 	if (rule->af == AF_INET) {
2017 		return (EAFNOSUPPORT);
2018 	}
2019 #endif /* INET */
2020 #ifndef INET6
2021 	if (rule->af == AF_INET6) {
2022 		return (EAFNOSUPPORT);
2023 	}
2024 #endif /* INET6 */
2025 
2026 	ret = pf_check_rule_addr(&rule->src);
2027 	if (ret != 0)
2028 		return (ret);
2029 	ret = pf_check_rule_addr(&rule->dst);
2030 	if (ret != 0)
2031 		return (ret);
2032 
2033 	bcopy(&rule->src, &krule->src, sizeof(rule->src));
2034 	bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
2035 
2036 	ret = pf_user_strcpy(krule->label[0], rule->label, sizeof(rule->label));
2037 	if (ret != 0)
2038 		return (ret);
2039 	ret = pf_user_strcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
2040 	if (ret != 0)
2041 		return (ret);
2042 	ret = pf_user_strcpy(krule->qname, rule->qname, sizeof(rule->qname));
2043 	if (ret != 0)
2044 		return (ret);
2045 	ret = pf_user_strcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
2046 	if (ret != 0)
2047 		return (ret);
2048 	ret = pf_user_strcpy(krule->tagname, rule->tagname,
2049 	    sizeof(rule->tagname));
2050 	if (ret != 0)
2051 		return (ret);
2052 	ret = pf_user_strcpy(krule->match_tagname, rule->match_tagname,
2053 	    sizeof(rule->match_tagname));
2054 	if (ret != 0)
2055 		return (ret);
2056 	ret = pf_user_strcpy(krule->overload_tblname, rule->overload_tblname,
2057 	    sizeof(rule->overload_tblname));
2058 	if (ret != 0)
2059 		return (ret);
2060 
2061 	pf_pool_to_kpool(&rule->rpool, &krule->rpool);
2062 
2063 	/* Don't allow userspace to set evaluations, packets or bytes. */
2064 	/* kif, anchor, overload_tbl are not copied over. */
2065 
2066 	krule->os_fingerprint = rule->os_fingerprint;
2067 
2068 	krule->rtableid = rule->rtableid;
2069 	bcopy(rule->timeout, krule->timeout, sizeof(krule->timeout));
2070 	krule->max_states = rule->max_states;
2071 	krule->max_src_nodes = rule->max_src_nodes;
2072 	krule->max_src_states = rule->max_src_states;
2073 	krule->max_src_conn = rule->max_src_conn;
2074 	krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
2075 	krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
2076 	krule->qid = rule->qid;
2077 	krule->pqid = rule->pqid;
2078 	krule->nr = rule->nr;
2079 	krule->prob = rule->prob;
2080 	krule->cuid = rule->cuid;
2081 	krule->cpid = rule->cpid;
2082 
2083 	krule->return_icmp = rule->return_icmp;
2084 	krule->return_icmp6 = rule->return_icmp6;
2085 	krule->max_mss = rule->max_mss;
2086 	krule->tag = rule->tag;
2087 	krule->match_tag = rule->match_tag;
2088 	krule->scrub_flags = rule->scrub_flags;
2089 
2090 	bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
2091 	bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
2092 
2093 	krule->rule_flag = rule->rule_flag;
2094 	krule->action = rule->action;
2095 	krule->direction = rule->direction;
2096 	krule->log = rule->log;
2097 	krule->logif = rule->logif;
2098 	krule->quick = rule->quick;
2099 	krule->ifnot = rule->ifnot;
2100 	krule->match_tag_not = rule->match_tag_not;
2101 	krule->natpass = rule->natpass;
2102 
2103 	krule->keep_state = rule->keep_state;
2104 	krule->af = rule->af;
2105 	krule->proto = rule->proto;
2106 	krule->type = rule->type;
2107 	krule->code = rule->code;
2108 	krule->flags = rule->flags;
2109 	krule->flagset = rule->flagset;
2110 	krule->min_ttl = rule->min_ttl;
2111 	krule->allow_opts = rule->allow_opts;
2112 	krule->rt = rule->rt;
2113 	krule->return_ttl = rule->return_ttl;
2114 	krule->tos = rule->tos;
2115 	krule->set_tos = rule->set_tos;
2116 
2117 	krule->flush = rule->flush;
2118 	krule->prio = rule->prio;
2119 	krule->set_prio[0] = rule->set_prio[0];
2120 	krule->set_prio[1] = rule->set_prio[1];
2121 
2122 	bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
2123 
2124 	return (0);
2125 }
2126 
2127 static int
2128 pf_state_kill_to_kstate_kill(const struct pfioc_state_kill *psk,
2129     struct pf_kstate_kill *kill)
2130 {
2131 	int ret;
2132 
2133 	bzero(kill, sizeof(*kill));
2134 
2135 	bcopy(&psk->psk_pfcmp, &kill->psk_pfcmp, sizeof(kill->psk_pfcmp));
2136 	kill->psk_af = psk->psk_af;
2137 	kill->psk_proto = psk->psk_proto;
2138 	bcopy(&psk->psk_src, &kill->psk_src, sizeof(kill->psk_src));
2139 	bcopy(&psk->psk_dst, &kill->psk_dst, sizeof(kill->psk_dst));
2140 	ret = pf_user_strcpy(kill->psk_ifname, psk->psk_ifname,
2141 	    sizeof(kill->psk_ifname));
2142 	if (ret != 0)
2143 		return (ret);
2144 	ret = pf_user_strcpy(kill->psk_label, psk->psk_label,
2145 	    sizeof(kill->psk_label));
2146 	if (ret != 0)
2147 		return (ret);
2148 
2149 	return (0);
2150 }
2151 
2152 static int
2153 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket,
2154     uint32_t pool_ticket, const char *anchor, const char *anchor_call,
2155     struct thread *td)
2156 {
2157 	struct pf_kruleset	*ruleset;
2158 	struct pf_krule		*tail;
2159 	struct pf_kpooladdr	*pa;
2160 	struct pfi_kkif		*kif = NULL;
2161 	int			 rs_num;
2162 	int			 error = 0;
2163 
2164 	if ((rule->return_icmp >> 8) > ICMP_MAXTYPE) {
2165 		error = EINVAL;
2166 		goto errout_unlocked;
2167 	}
2168 
2169 #define	ERROUT(x)	ERROUT_FUNCTION(errout, x)
2170 
2171 	if (rule->ifname[0])
2172 		kif = pf_kkif_create(M_WAITOK);
2173 	pf_counter_u64_init(&rule->evaluations, M_WAITOK);
2174 	for (int i = 0; i < 2; i++) {
2175 		pf_counter_u64_init(&rule->packets[i], M_WAITOK);
2176 		pf_counter_u64_init(&rule->bytes[i], M_WAITOK);
2177 	}
2178 	rule->states_cur = counter_u64_alloc(M_WAITOK);
2179 	rule->states_tot = counter_u64_alloc(M_WAITOK);
2180 	rule->src_nodes = counter_u64_alloc(M_WAITOK);
2181 	rule->cuid = td->td_ucred->cr_ruid;
2182 	rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
2183 	TAILQ_INIT(&rule->rpool.list);
2184 
2185 	PF_CONFIG_LOCK();
2186 	PF_RULES_WLOCK();
2187 #ifdef PF_WANT_32_TO_64_COUNTER
2188 	LIST_INSERT_HEAD(&V_pf_allrulelist, rule, allrulelist);
2189 	MPASS(!rule->allrulelinked);
2190 	rule->allrulelinked = true;
2191 	V_pf_allrulecount++;
2192 #endif
2193 	ruleset = pf_find_kruleset(anchor);
2194 	if (ruleset == NULL)
2195 		ERROUT(EINVAL);
2196 	rs_num = pf_get_ruleset_number(rule->action);
2197 	if (rs_num >= PF_RULESET_MAX)
2198 		ERROUT(EINVAL);
2199 	if (ticket != ruleset->rules[rs_num].inactive.ticket) {
2200 		DPFPRINTF(PF_DEBUG_MISC,
2201 		    ("ticket: %d != [%d]%d\n", ticket, rs_num,
2202 		    ruleset->rules[rs_num].inactive.ticket));
2203 		ERROUT(EBUSY);
2204 	}
2205 	if (pool_ticket != V_ticket_pabuf) {
2206 		DPFPRINTF(PF_DEBUG_MISC,
2207 		    ("pool_ticket: %d != %d\n", pool_ticket,
2208 		    V_ticket_pabuf));
2209 		ERROUT(EBUSY);
2210 	}
2211 	/*
2212 	 * XXXMJG hack: there is no mechanism to ensure they started the
2213 	 * transaction. Ticket checked above may happen to match by accident,
2214 	 * even if nobody called DIOCXBEGIN, let alone this process.
2215 	 * Partially work around it by checking if the RB tree got allocated,
2216 	 * see pf_begin_rules.
2217 	 */
2218 	if (ruleset->rules[rs_num].inactive.tree == NULL) {
2219 		ERROUT(EINVAL);
2220 	}
2221 
2222 	tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
2223 	    pf_krulequeue);
2224 	if (tail)
2225 		rule->nr = tail->nr + 1;
2226 	else
2227 		rule->nr = 0;
2228 	if (rule->ifname[0]) {
2229 		rule->kif = pfi_kkif_attach(kif, rule->ifname);
2230 		kif = NULL;
2231 		pfi_kkif_ref(rule->kif);
2232 	} else
2233 		rule->kif = NULL;
2234 
2235 	if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
2236 		error = EBUSY;
2237 
2238 #ifdef ALTQ
2239 	/* set queue IDs */
2240 	if (rule->qname[0] != 0) {
2241 		if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2242 			error = EBUSY;
2243 		else if (rule->pqname[0] != 0) {
2244 			if ((rule->pqid =
2245 			    pf_qname2qid(rule->pqname)) == 0)
2246 				error = EBUSY;
2247 		} else
2248 			rule->pqid = rule->qid;
2249 	}
2250 #endif
2251 	if (rule->tagname[0])
2252 		if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2253 			error = EBUSY;
2254 	if (rule->match_tagname[0])
2255 		if ((rule->match_tag =
2256 		    pf_tagname2tag(rule->match_tagname)) == 0)
2257 			error = EBUSY;
2258 	if (rule->rt && !rule->direction)
2259 		error = EINVAL;
2260 	if (!rule->log)
2261 		rule->logif = 0;
2262 	if (rule->logif >= PFLOGIFS_MAX)
2263 		error = EINVAL;
2264 	if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
2265 		error = ENOMEM;
2266 	if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
2267 		error = ENOMEM;
2268 	if (pf_kanchor_setup(rule, ruleset, anchor_call))
2269 		error = EINVAL;
2270 	if (rule->scrub_flags & PFSTATE_SETPRIO &&
2271 	    (rule->set_prio[0] > PF_PRIO_MAX ||
2272 	    rule->set_prio[1] > PF_PRIO_MAX))
2273 		error = EINVAL;
2274 	TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
2275 		if (pa->addr.type == PF_ADDR_TABLE) {
2276 			pa->addr.p.tbl = pfr_attach_table(ruleset,
2277 			    pa->addr.v.tblname);
2278 			if (pa->addr.p.tbl == NULL)
2279 				error = ENOMEM;
2280 		}
2281 
2282 	rule->overload_tbl = NULL;
2283 	if (rule->overload_tblname[0]) {
2284 		if ((rule->overload_tbl = pfr_attach_table(ruleset,
2285 		    rule->overload_tblname)) == NULL)
2286 			error = EINVAL;
2287 		else
2288 			rule->overload_tbl->pfrkt_flags |=
2289 			    PFR_TFLAG_ACTIVE;
2290 	}
2291 
2292 	pf_mv_kpool(&V_pf_pabuf, &rule->rpool.list);
2293 	if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
2294 	    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
2295 	    (rule->rt > PF_NOPFROUTE)) &&
2296 	    (TAILQ_FIRST(&rule->rpool.list) == NULL))
2297 		error = EINVAL;
2298 
2299 	if (error) {
2300 		pf_free_rule(rule);
2301 		rule = NULL;
2302 		ERROUT(error);
2303 	}
2304 
2305 	rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
2306 	TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
2307 	    rule, entries);
2308 	ruleset->rules[rs_num].inactive.rcount++;
2309 
2310 	PF_RULES_WUNLOCK();
2311 	pf_hash_rule(rule);
2312 	if (RB_INSERT(pf_krule_global, ruleset->rules[rs_num].inactive.tree, rule) != NULL) {
2313 		PF_RULES_WLOCK();
2314 		TAILQ_REMOVE(ruleset->rules[rs_num].inactive.ptr, rule, entries);
2315 		ruleset->rules[rs_num].inactive.rcount--;
2316 		pf_free_rule(rule);
2317 		rule = NULL;
2318 		ERROUT(EEXIST);
2319 	}
2320 	PF_CONFIG_UNLOCK();
2321 
2322 	return (0);
2323 
2324 #undef ERROUT
2325 errout:
2326 	PF_RULES_WUNLOCK();
2327 	PF_CONFIG_UNLOCK();
2328 errout_unlocked:
2329 	pf_kkif_free(kif);
2330 	pf_krule_free(rule);
2331 	return (error);
2332 }
2333 
2334 static bool
2335 pf_label_match(const struct pf_krule *rule, const char *label)
2336 {
2337 	int i = 0;
2338 
2339 	while (*rule->label[i]) {
2340 		if (strcmp(rule->label[i], label) == 0)
2341 			return (true);
2342 		i++;
2343 	}
2344 
2345 	return (false);
2346 }
2347 
2348 static unsigned int
2349 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir)
2350 {
2351 	struct pf_kstate *s;
2352 	int more = 0;
2353 
2354 	s = pf_find_state_all(key, dir, &more);
2355 	if (s == NULL)
2356 		return (0);
2357 
2358 	if (more) {
2359 		PF_STATE_UNLOCK(s);
2360 		return (0);
2361 	}
2362 
2363 	pf_unlink_state(s);
2364 	return (1);
2365 }
2366 
2367 static int
2368 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih)
2369 {
2370 	struct pf_kstate	*s;
2371 	struct pf_state_key	*sk;
2372 	struct pf_addr		*srcaddr, *dstaddr;
2373 	struct pf_state_key_cmp	 match_key;
2374 	int			 idx, killed = 0;
2375 	unsigned int		 dir;
2376 	u_int16_t		 srcport, dstport;
2377 	struct pfi_kkif		*kif;
2378 
2379 relock_DIOCKILLSTATES:
2380 	PF_HASHROW_LOCK(ih);
2381 	LIST_FOREACH(s, &ih->states, entry) {
2382 		/* For floating states look at the original kif. */
2383 		kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
2384 
2385 		sk = s->key[PF_SK_WIRE];
2386 		if (s->direction == PF_OUT) {
2387 			srcaddr = &sk->addr[1];
2388 			dstaddr = &sk->addr[0];
2389 			srcport = sk->port[1];
2390 			dstport = sk->port[0];
2391 		} else {
2392 			srcaddr = &sk->addr[0];
2393 			dstaddr = &sk->addr[1];
2394 			srcport = sk->port[0];
2395 			dstport = sk->port[1];
2396 		}
2397 
2398 		if (psk->psk_af && sk->af != psk->psk_af)
2399 			continue;
2400 
2401 		if (psk->psk_proto && psk->psk_proto != sk->proto)
2402 			continue;
2403 
2404 		if (! PF_MATCHA(psk->psk_src.neg, &psk->psk_src.addr.v.a.addr,
2405 		    &psk->psk_src.addr.v.a.mask, srcaddr, sk->af))
2406 			continue;
2407 
2408 		if (! PF_MATCHA(psk->psk_dst.neg, &psk->psk_dst.addr.v.a.addr,
2409 		    &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af))
2410 			continue;
2411 
2412 		if (!  PF_MATCHA(psk->psk_rt_addr.neg,
2413 		    &psk->psk_rt_addr.addr.v.a.addr,
2414 		    &psk->psk_rt_addr.addr.v.a.mask,
2415 		    &s->rt_addr, sk->af))
2416 			continue;
2417 
2418 		if (psk->psk_src.port_op != 0 &&
2419 		    ! pf_match_port(psk->psk_src.port_op,
2420 		    psk->psk_src.port[0], psk->psk_src.port[1], srcport))
2421 			continue;
2422 
2423 		if (psk->psk_dst.port_op != 0 &&
2424 		    ! pf_match_port(psk->psk_dst.port_op,
2425 		    psk->psk_dst.port[0], psk->psk_dst.port[1], dstport))
2426 			continue;
2427 
2428 		if (psk->psk_label[0] &&
2429 		    ! pf_label_match(s->rule.ptr, psk->psk_label))
2430 			continue;
2431 
2432 		if (psk->psk_ifname[0] && strcmp(psk->psk_ifname,
2433 		    kif->pfik_name))
2434 			continue;
2435 
2436 		if (psk->psk_kill_match) {
2437 			/* Create the key to find matching states, with lock
2438 			 * held. */
2439 
2440 			bzero(&match_key, sizeof(match_key));
2441 
2442 			if (s->direction == PF_OUT) {
2443 				dir = PF_IN;
2444 				idx = PF_SK_STACK;
2445 			} else {
2446 				dir = PF_OUT;
2447 				idx = PF_SK_WIRE;
2448 			}
2449 
2450 			match_key.af = s->key[idx]->af;
2451 			match_key.proto = s->key[idx]->proto;
2452 			PF_ACPY(&match_key.addr[0],
2453 			    &s->key[idx]->addr[1], match_key.af);
2454 			match_key.port[0] = s->key[idx]->port[1];
2455 			PF_ACPY(&match_key.addr[1],
2456 			    &s->key[idx]->addr[0], match_key.af);
2457 			match_key.port[1] = s->key[idx]->port[0];
2458 		}
2459 
2460 		pf_unlink_state(s);
2461 		killed++;
2462 
2463 		if (psk->psk_kill_match)
2464 			killed += pf_kill_matching_state(&match_key, dir);
2465 
2466 		goto relock_DIOCKILLSTATES;
2467 	}
2468 	PF_HASHROW_UNLOCK(ih);
2469 
2470 	return (killed);
2471 }
2472 
2473 static int
2474 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
2475 {
2476 	int			 error = 0;
2477 	PF_RULES_RLOCK_TRACKER;
2478 
2479 #define	ERROUT_IOCTL(target, x)					\
2480     do {								\
2481 	    error = (x);						\
2482 	    SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__);	\
2483 	    goto target;						\
2484     } while (0)
2485 
2486 
2487 	/* XXX keep in sync with switch() below */
2488 	if (securelevel_gt(td->td_ucred, 2))
2489 		switch (cmd) {
2490 		case DIOCGETRULES:
2491 		case DIOCGETRULE:
2492 		case DIOCGETRULENV:
2493 		case DIOCGETADDRS:
2494 		case DIOCGETADDR:
2495 		case DIOCGETSTATE:
2496 		case DIOCGETSTATENV:
2497 		case DIOCSETSTATUSIF:
2498 		case DIOCGETSTATUS:
2499 		case DIOCGETSTATUSNV:
2500 		case DIOCCLRSTATUS:
2501 		case DIOCNATLOOK:
2502 		case DIOCSETDEBUG:
2503 		case DIOCGETSTATES:
2504 		case DIOCGETSTATESV2:
2505 		case DIOCGETTIMEOUT:
2506 		case DIOCCLRRULECTRS:
2507 		case DIOCGETLIMIT:
2508 		case DIOCGETALTQSV0:
2509 		case DIOCGETALTQSV1:
2510 		case DIOCGETALTQV0:
2511 		case DIOCGETALTQV1:
2512 		case DIOCGETQSTATSV0:
2513 		case DIOCGETQSTATSV1:
2514 		case DIOCGETRULESETS:
2515 		case DIOCGETRULESET:
2516 		case DIOCRGETTABLES:
2517 		case DIOCRGETTSTATS:
2518 		case DIOCRCLRTSTATS:
2519 		case DIOCRCLRADDRS:
2520 		case DIOCRADDADDRS:
2521 		case DIOCRDELADDRS:
2522 		case DIOCRSETADDRS:
2523 		case DIOCRGETADDRS:
2524 		case DIOCRGETASTATS:
2525 		case DIOCRCLRASTATS:
2526 		case DIOCRTSTADDRS:
2527 		case DIOCOSFPGET:
2528 		case DIOCGETSRCNODES:
2529 		case DIOCCLRSRCNODES:
2530 		case DIOCGETSYNCOOKIES:
2531 		case DIOCIGETIFACES:
2532 		case DIOCGIFSPEEDV0:
2533 		case DIOCGIFSPEEDV1:
2534 		case DIOCSETIFFLAG:
2535 		case DIOCCLRIFFLAG:
2536 		case DIOCGETETHRULES:
2537 		case DIOCGETETHRULE:
2538 		case DIOCGETETHRULESETS:
2539 		case DIOCGETETHRULESET:
2540 			break;
2541 		case DIOCRCLRTABLES:
2542 		case DIOCRADDTABLES:
2543 		case DIOCRDELTABLES:
2544 		case DIOCRSETTFLAGS:
2545 			if (((struct pfioc_table *)addr)->pfrio_flags &
2546 			    PFR_FLAG_DUMMY)
2547 				break; /* dummy operation ok */
2548 			return (EPERM);
2549 		default:
2550 			return (EPERM);
2551 		}
2552 
2553 	if (!(flags & FWRITE))
2554 		switch (cmd) {
2555 		case DIOCGETRULES:
2556 		case DIOCGETADDRS:
2557 		case DIOCGETADDR:
2558 		case DIOCGETSTATE:
2559 		case DIOCGETSTATENV:
2560 		case DIOCGETSTATUS:
2561 		case DIOCGETSTATUSNV:
2562 		case DIOCGETSTATES:
2563 		case DIOCGETSTATESV2:
2564 		case DIOCGETTIMEOUT:
2565 		case DIOCGETLIMIT:
2566 		case DIOCGETALTQSV0:
2567 		case DIOCGETALTQSV1:
2568 		case DIOCGETALTQV0:
2569 		case DIOCGETALTQV1:
2570 		case DIOCGETQSTATSV0:
2571 		case DIOCGETQSTATSV1:
2572 		case DIOCGETRULESETS:
2573 		case DIOCGETRULESET:
2574 		case DIOCNATLOOK:
2575 		case DIOCRGETTABLES:
2576 		case DIOCRGETTSTATS:
2577 		case DIOCRGETADDRS:
2578 		case DIOCRGETASTATS:
2579 		case DIOCRTSTADDRS:
2580 		case DIOCOSFPGET:
2581 		case DIOCGETSRCNODES:
2582 		case DIOCGETSYNCOOKIES:
2583 		case DIOCIGETIFACES:
2584 		case DIOCGIFSPEEDV1:
2585 		case DIOCGIFSPEEDV0:
2586 		case DIOCGETRULENV:
2587 		case DIOCGETETHRULES:
2588 		case DIOCGETETHRULE:
2589 		case DIOCGETETHRULESETS:
2590 		case DIOCGETETHRULESET:
2591 			break;
2592 		case DIOCRCLRTABLES:
2593 		case DIOCRADDTABLES:
2594 		case DIOCRDELTABLES:
2595 		case DIOCRCLRTSTATS:
2596 		case DIOCRCLRADDRS:
2597 		case DIOCRADDADDRS:
2598 		case DIOCRDELADDRS:
2599 		case DIOCRSETADDRS:
2600 		case DIOCRSETTFLAGS:
2601 			if (((struct pfioc_table *)addr)->pfrio_flags &
2602 			    PFR_FLAG_DUMMY) {
2603 				flags |= FWRITE; /* need write lock for dummy */
2604 				break; /* dummy operation ok */
2605 			}
2606 			return (EACCES);
2607 		case DIOCGETRULE:
2608 			if (((struct pfioc_rule *)addr)->action ==
2609 			    PF_GET_CLR_CNTR)
2610 				return (EACCES);
2611 			break;
2612 		default:
2613 			return (EACCES);
2614 		}
2615 
2616 	CURVNET_SET(TD_TO_VNET(td));
2617 
2618 	switch (cmd) {
2619 	case DIOCSTART:
2620 		sx_xlock(&V_pf_ioctl_lock);
2621 		if (V_pf_status.running)
2622 			error = EEXIST;
2623 		else {
2624 			hook_pf();
2625 			if (! TAILQ_EMPTY(V_pf_keth->active.rules))
2626 				hook_pf_eth();
2627 			V_pf_status.running = 1;
2628 			V_pf_status.since = time_second;
2629 			new_unrhdr64(&V_pf_stateid, time_second);
2630 
2631 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
2632 		}
2633 		break;
2634 
2635 	case DIOCSTOP:
2636 		sx_xlock(&V_pf_ioctl_lock);
2637 		if (!V_pf_status.running)
2638 			error = ENOENT;
2639 		else {
2640 			V_pf_status.running = 0;
2641 			dehook_pf();
2642 			dehook_pf_eth();
2643 			V_pf_status.since = time_second;
2644 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
2645 		}
2646 		break;
2647 
2648 	case DIOCGETETHRULES: {
2649 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2650 		nvlist_t		*nvl;
2651 		void			*packed;
2652 		struct pf_keth_rule	*tail;
2653 		struct pf_keth_ruleset	*rs;
2654 		u_int32_t		 ticket, nr;
2655 		const char		*anchor = "";
2656 
2657 		nvl = NULL;
2658 		packed = NULL;
2659 
2660 #define	ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULES_error, x)
2661 
2662 		if (nv->len > pf_ioctl_maxcount)
2663 			ERROUT(ENOMEM);
2664 
2665 		/* Copy the request in */
2666 		packed = malloc(nv->len, M_NVLIST, M_WAITOK);
2667 		if (packed == NULL)
2668 			ERROUT(ENOMEM);
2669 
2670 		error = copyin(nv->data, packed, nv->len);
2671 		if (error)
2672 			ERROUT(error);
2673 
2674 		nvl = nvlist_unpack(packed, nv->len, 0);
2675 		if (nvl == NULL)
2676 			ERROUT(EBADMSG);
2677 
2678 		if (! nvlist_exists_string(nvl, "anchor"))
2679 			ERROUT(EBADMSG);
2680 
2681 		anchor = nvlist_get_string(nvl, "anchor");
2682 
2683 		rs = pf_find_keth_ruleset(anchor);
2684 
2685 		nvlist_destroy(nvl);
2686 		nvl = NULL;
2687 		free(packed, M_NVLIST);
2688 		packed = NULL;
2689 
2690 		if (rs == NULL)
2691 			ERROUT(ENOENT);
2692 
2693 		/* Reply */
2694 		nvl = nvlist_create(0);
2695 		if (nvl == NULL)
2696 			ERROUT(ENOMEM);
2697 
2698 		PF_RULES_RLOCK();
2699 
2700 		ticket = rs->active.ticket;
2701 		tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq);
2702 		if (tail)
2703 			nr = tail->nr + 1;
2704 		else
2705 			nr = 0;
2706 
2707 		PF_RULES_RUNLOCK();
2708 
2709 		nvlist_add_number(nvl, "ticket", ticket);
2710 		nvlist_add_number(nvl, "nr", nr);
2711 
2712 		packed = nvlist_pack(nvl, &nv->len);
2713 		if (packed == NULL)
2714 			ERROUT(ENOMEM);
2715 
2716 		if (nv->size == 0)
2717 			ERROUT(0);
2718 		else if (nv->size < nv->len)
2719 			ERROUT(ENOSPC);
2720 
2721 		error = copyout(packed, nv->data, nv->len);
2722 
2723 #undef ERROUT
2724 DIOCGETETHRULES_error:
2725 		free(packed, M_NVLIST);
2726 		nvlist_destroy(nvl);
2727 		break;
2728 	}
2729 
2730 	case DIOCGETETHRULE: {
2731 		struct epoch_tracker	 et;
2732 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2733 		nvlist_t		*nvl = NULL;
2734 		void			*nvlpacked = NULL;
2735 		struct pf_keth_rule	*rule = NULL;
2736 		struct pf_keth_ruleset	*rs;
2737 		u_int32_t		 ticket, nr;
2738 		bool			 clear = false;
2739 		const char		*anchor;
2740 
2741 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULE_error, x)
2742 
2743 		if (nv->len > pf_ioctl_maxcount)
2744 			ERROUT(ENOMEM);
2745 
2746 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2747 		if (nvlpacked == NULL)
2748 			ERROUT(ENOMEM);
2749 
2750 		error = copyin(nv->data, nvlpacked, nv->len);
2751 		if (error)
2752 			ERROUT(error);
2753 
2754 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2755 		if (nvl == NULL)
2756 			ERROUT(EBADMSG);
2757 		if (! nvlist_exists_number(nvl, "ticket"))
2758 			ERROUT(EBADMSG);
2759 		ticket = nvlist_get_number(nvl, "ticket");
2760 		if (! nvlist_exists_string(nvl, "anchor"))
2761 			ERROUT(EBADMSG);
2762 		anchor = nvlist_get_string(nvl, "anchor");
2763 
2764 		if (nvlist_exists_bool(nvl, "clear"))
2765 			clear = nvlist_get_bool(nvl, "clear");
2766 
2767 		if (clear && !(flags & FWRITE))
2768 			ERROUT(EACCES);
2769 
2770 		if (! nvlist_exists_number(nvl, "nr"))
2771 			ERROUT(EBADMSG);
2772 		nr = nvlist_get_number(nvl, "nr");
2773 
2774 		PF_RULES_RLOCK();
2775 		rs = pf_find_keth_ruleset(anchor);
2776 		if (rs == NULL) {
2777 			PF_RULES_RUNLOCK();
2778 			ERROUT(ENOENT);
2779 		}
2780 		if (ticket != rs->active.ticket) {
2781 			PF_RULES_RUNLOCK();
2782 			ERROUT(EBUSY);
2783 		}
2784 
2785 		nvlist_destroy(nvl);
2786 		nvl = NULL;
2787 		free(nvlpacked, M_NVLIST);
2788 		nvlpacked = NULL;
2789 
2790 		rule = TAILQ_FIRST(rs->active.rules);
2791 		while ((rule != NULL) && (rule->nr != nr))
2792 			rule = TAILQ_NEXT(rule, entries);
2793 		if (rule == NULL) {
2794 			PF_RULES_RUNLOCK();
2795 			ERROUT(ENOENT);
2796 		}
2797 		/* Make sure rule can't go away. */
2798 		NET_EPOCH_ENTER(et);
2799 		PF_RULES_RUNLOCK();
2800 		nvl = pf_keth_rule_to_nveth_rule(rule);
2801 		if (pf_keth_anchor_nvcopyout(rs, rule, nvl))
2802 			ERROUT(EBUSY);
2803 		NET_EPOCH_EXIT(et);
2804 		if (nvl == NULL)
2805 			ERROUT(ENOMEM);
2806 
2807 		nvlpacked = nvlist_pack(nvl, &nv->len);
2808 		if (nvlpacked == NULL)
2809 			ERROUT(ENOMEM);
2810 
2811 		if (nv->size == 0)
2812 			ERROUT(0);
2813 		else if (nv->size < nv->len)
2814 			ERROUT(ENOSPC);
2815 
2816 		error = copyout(nvlpacked, nv->data, nv->len);
2817 		if (error == 0 && clear) {
2818 			counter_u64_zero(rule->evaluations);
2819 			for (int i = 0; i < 2; i++) {
2820 				counter_u64_zero(rule->packets[i]);
2821 				counter_u64_zero(rule->bytes[i]);
2822 			}
2823 		}
2824 
2825 #undef ERROUT
2826 DIOCGETETHRULE_error:
2827 		free(nvlpacked, M_NVLIST);
2828 		nvlist_destroy(nvl);
2829 		break;
2830 	}
2831 
2832 	case DIOCADDETHRULE: {
2833 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2834 		nvlist_t		*nvl = NULL;
2835 		void			*nvlpacked = NULL;
2836 		struct pf_keth_rule	*rule = NULL, *tail = NULL;
2837 		struct pf_keth_ruleset	*ruleset = NULL;
2838 		struct pfi_kkif		*kif = NULL, *bridge_to_kif = NULL;
2839 		const char		*anchor = "", *anchor_call = "";
2840 
2841 #define ERROUT(x)	ERROUT_IOCTL(DIOCADDETHRULE_error, x)
2842 
2843 		if (nv->len > pf_ioctl_maxcount)
2844 			ERROUT(ENOMEM);
2845 
2846 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2847 		if (nvlpacked == NULL)
2848 			ERROUT(ENOMEM);
2849 
2850 		error = copyin(nv->data, nvlpacked, nv->len);
2851 		if (error)
2852 			ERROUT(error);
2853 
2854 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2855 		if (nvl == NULL)
2856 			ERROUT(EBADMSG);
2857 
2858 		if (! nvlist_exists_number(nvl, "ticket"))
2859 			ERROUT(EBADMSG);
2860 
2861 		if (nvlist_exists_string(nvl, "anchor"))
2862 			anchor = nvlist_get_string(nvl, "anchor");
2863 		if (nvlist_exists_string(nvl, "anchor_call"))
2864 			anchor_call = nvlist_get_string(nvl, "anchor_call");
2865 
2866 		ruleset = pf_find_keth_ruleset(anchor);
2867 		if (ruleset == NULL)
2868 			ERROUT(EINVAL);
2869 
2870 		if (nvlist_get_number(nvl, "ticket") !=
2871 		    ruleset->inactive.ticket) {
2872 			DPFPRINTF(PF_DEBUG_MISC,
2873 			    ("ticket: %d != %d\n",
2874 			    (u_int32_t)nvlist_get_number(nvl, "ticket"),
2875 			    ruleset->inactive.ticket));
2876 			ERROUT(EBUSY);
2877 		}
2878 
2879 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
2880 		if (rule == NULL)
2881 			ERROUT(ENOMEM);
2882 		rule->timestamp = NULL;
2883 
2884 		error = pf_nveth_rule_to_keth_rule(nvl, rule);
2885 		if (error != 0)
2886 			ERROUT(error);
2887 
2888 		if (rule->ifname[0])
2889 			kif = pf_kkif_create(M_WAITOK);
2890 		if (rule->bridge_to_name[0])
2891 			bridge_to_kif = pf_kkif_create(M_WAITOK);
2892 		rule->evaluations = counter_u64_alloc(M_WAITOK);
2893 		for (int i = 0; i < 2; i++) {
2894 			rule->packets[i] = counter_u64_alloc(M_WAITOK);
2895 			rule->bytes[i] = counter_u64_alloc(M_WAITOK);
2896 		}
2897 		rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
2898 		    M_WAITOK | M_ZERO);
2899 
2900 		PF_RULES_WLOCK();
2901 
2902 		if (rule->ifname[0]) {
2903 			rule->kif = pfi_kkif_attach(kif, rule->ifname);
2904 			pfi_kkif_ref(rule->kif);
2905 		} else
2906 			rule->kif = NULL;
2907 		if (rule->bridge_to_name[0]) {
2908 			rule->bridge_to = pfi_kkif_attach(bridge_to_kif,
2909 			    rule->bridge_to_name);
2910 			pfi_kkif_ref(rule->bridge_to);
2911 		} else
2912 			rule->bridge_to = NULL;
2913 
2914 #ifdef ALTQ
2915 		/* set queue IDs */
2916 		if (rule->qname[0] != 0) {
2917 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2918 				error = EBUSY;
2919 			else
2920 				rule->qid = rule->qid;
2921 		}
2922 #endif
2923 		if (rule->tagname[0])
2924 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2925 				error = EBUSY;
2926 		if (rule->match_tagname[0])
2927 			if ((rule->match_tag = pf_tagname2tag(
2928 			    rule->match_tagname)) == 0)
2929 				error = EBUSY;
2930 
2931 		if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE)
2932 			error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr);
2933 		if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE)
2934 			error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr);
2935 
2936 		if (error) {
2937 			pf_free_eth_rule(rule);
2938 			PF_RULES_WUNLOCK();
2939 			ERROUT(error);
2940 		}
2941 
2942 		if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) {
2943 			pf_free_eth_rule(rule);
2944 			PF_RULES_WUNLOCK();
2945 			ERROUT(EINVAL);
2946 		}
2947 
2948 		tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq);
2949 		if (tail)
2950 			rule->nr = tail->nr + 1;
2951 		else
2952 			rule->nr = 0;
2953 
2954 		TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries);
2955 
2956 		PF_RULES_WUNLOCK();
2957 
2958 #undef ERROUT
2959 DIOCADDETHRULE_error:
2960 		nvlist_destroy(nvl);
2961 		free(nvlpacked, M_NVLIST);
2962 		break;
2963 	}
2964 
2965 	case DIOCGETETHRULESETS: {
2966 		struct epoch_tracker	 et;
2967 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2968 		nvlist_t		*nvl = NULL;
2969 		void			*nvlpacked = NULL;
2970 		struct pf_keth_ruleset	*ruleset;
2971 		struct pf_keth_anchor	*anchor;
2972 		int			 nr = 0;
2973 
2974 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULESETS_error, x)
2975 
2976 		if (nv->len > pf_ioctl_maxcount)
2977 			ERROUT(ENOMEM);
2978 
2979 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2980 		if (nvlpacked == NULL)
2981 			ERROUT(ENOMEM);
2982 
2983 		error = copyin(nv->data, nvlpacked, nv->len);
2984 		if (error)
2985 			ERROUT(error);
2986 
2987 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2988 		if (nvl == NULL)
2989 			ERROUT(EBADMSG);
2990 		if (! nvlist_exists_string(nvl, "path"))
2991 			ERROUT(EBADMSG);
2992 
2993 		NET_EPOCH_ENTER(et);
2994 
2995 		if ((ruleset = pf_find_keth_ruleset(
2996 		    nvlist_get_string(nvl, "path"))) == NULL) {
2997 			NET_EPOCH_EXIT(et);
2998 			ERROUT(ENOENT);
2999 		}
3000 
3001 		if (ruleset->anchor == NULL) {
3002 			RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors)
3003 				if (anchor->parent == NULL)
3004 					nr++;
3005 		} else {
3006 			RB_FOREACH(anchor, pf_keth_anchor_node,
3007 			    &ruleset->anchor->children)
3008 				nr++;
3009 		}
3010 
3011 		NET_EPOCH_EXIT(et);
3012 
3013 		nvlist_destroy(nvl);
3014 		nvl = NULL;
3015 		free(nvlpacked, M_NVLIST);
3016 		nvlpacked = NULL;
3017 
3018 		nvl = nvlist_create(0);
3019 		if (nvl == NULL)
3020 			ERROUT(ENOMEM);
3021 
3022 		nvlist_add_number(nvl, "nr", nr);
3023 
3024 		nvlpacked = nvlist_pack(nvl, &nv->len);
3025 		if (nvlpacked == NULL)
3026 			ERROUT(ENOMEM);
3027 
3028 		if (nv->size == 0)
3029 			ERROUT(0);
3030 		else if (nv->size < nv->len)
3031 			ERROUT(ENOSPC);
3032 
3033 		error = copyout(nvlpacked, nv->data, nv->len);
3034 
3035 #undef ERROUT
3036 DIOCGETETHRULESETS_error:
3037 		free(nvlpacked, M_NVLIST);
3038 		nvlist_destroy(nvl);
3039 		break;
3040 	}
3041 
3042 	case DIOCGETETHRULESET: {
3043 		struct epoch_tracker	 et;
3044 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
3045 		nvlist_t		*nvl = NULL;
3046 		void			*nvlpacked = NULL;
3047 		struct pf_keth_ruleset	*ruleset;
3048 		struct pf_keth_anchor	*anchor;
3049 		int			 nr = 0, req_nr = 0;
3050 		bool			 found = false;
3051 
3052 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULESET_error, x)
3053 
3054 		if (nv->len > pf_ioctl_maxcount)
3055 			ERROUT(ENOMEM);
3056 
3057 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3058 		if (nvlpacked == NULL)
3059 			ERROUT(ENOMEM);
3060 
3061 		error = copyin(nv->data, nvlpacked, nv->len);
3062 		if (error)
3063 			ERROUT(error);
3064 
3065 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3066 		if (nvl == NULL)
3067 			ERROUT(EBADMSG);
3068 		if (! nvlist_exists_string(nvl, "path"))
3069 			ERROUT(EBADMSG);
3070 		if (! nvlist_exists_number(nvl, "nr"))
3071 			ERROUT(EBADMSG);
3072 
3073 		req_nr = nvlist_get_number(nvl, "nr");
3074 
3075 		NET_EPOCH_ENTER(et);
3076 
3077 		if ((ruleset = pf_find_keth_ruleset(
3078 		    nvlist_get_string(nvl, "path"))) == NULL) {
3079 			NET_EPOCH_EXIT(et);
3080 			ERROUT(ENOENT);
3081 		}
3082 
3083 		nvlist_destroy(nvl);
3084 		nvl = NULL;
3085 		free(nvlpacked, M_NVLIST);
3086 		nvlpacked = NULL;
3087 
3088 		nvl = nvlist_create(0);
3089 		if (nvl == NULL) {
3090 			NET_EPOCH_EXIT(et);
3091 			ERROUT(ENOMEM);
3092 		}
3093 
3094 		if (ruleset->anchor == NULL) {
3095 			RB_FOREACH(anchor, pf_keth_anchor_global,
3096 			    &V_pf_keth_anchors) {
3097 				if (anchor->parent == NULL && nr++ == req_nr) {
3098 					found = true;
3099 					break;
3100 				}
3101 			}
3102 		} else {
3103 			RB_FOREACH(anchor, pf_keth_anchor_node,
3104 			     &ruleset->anchor->children) {
3105 				if (nr++ == req_nr) {
3106 					found = true;
3107 					break;
3108 				}
3109 			}
3110 		}
3111 
3112 		NET_EPOCH_EXIT(et);
3113 		if (found) {
3114 			nvlist_add_number(nvl, "nr", nr);
3115 			nvlist_add_string(nvl, "name", anchor->name);
3116 			if (ruleset->anchor)
3117 				nvlist_add_string(nvl, "path",
3118 				    ruleset->anchor->path);
3119 			else
3120 				nvlist_add_string(nvl, "path", "");
3121 		} else {
3122 			ERROUT(EBUSY);
3123 		}
3124 
3125 		nvlpacked = nvlist_pack(nvl, &nv->len);
3126 		if (nvlpacked == NULL)
3127 			ERROUT(ENOMEM);
3128 
3129 		if (nv->size == 0)
3130 			ERROUT(0);
3131 		else if (nv->size < nv->len)
3132 			ERROUT(ENOSPC);
3133 
3134 		error = copyout(nvlpacked, nv->data, nv->len);
3135 
3136 #undef ERROUT
3137 DIOCGETETHRULESET_error:
3138 		free(nvlpacked, M_NVLIST);
3139 		nvlist_destroy(nvl);
3140 		break;
3141 	}
3142 
3143 	case DIOCADDRULENV: {
3144 		struct pfioc_nv	*nv = (struct pfioc_nv *)addr;
3145 		nvlist_t	*nvl = NULL;
3146 		void		*nvlpacked = NULL;
3147 		struct pf_krule	*rule = NULL;
3148 		const char	*anchor = "", *anchor_call = "";
3149 		uint32_t	 ticket = 0, pool_ticket = 0;
3150 
3151 #define	ERROUT(x)	ERROUT_IOCTL(DIOCADDRULENV_error, x)
3152 
3153 		if (nv->len > pf_ioctl_maxcount)
3154 			ERROUT(ENOMEM);
3155 
3156 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3157 		error = copyin(nv->data, nvlpacked, nv->len);
3158 		if (error)
3159 			ERROUT(error);
3160 
3161 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3162 		if (nvl == NULL)
3163 			ERROUT(EBADMSG);
3164 
3165 		if (! nvlist_exists_number(nvl, "ticket"))
3166 			ERROUT(EINVAL);
3167 		ticket = nvlist_get_number(nvl, "ticket");
3168 
3169 		if (! nvlist_exists_number(nvl, "pool_ticket"))
3170 			ERROUT(EINVAL);
3171 		pool_ticket = nvlist_get_number(nvl, "pool_ticket");
3172 
3173 		if (! nvlist_exists_nvlist(nvl, "rule"))
3174 			ERROUT(EINVAL);
3175 
3176 		rule = pf_krule_alloc();
3177 		error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"),
3178 		    rule);
3179 		if (error)
3180 			ERROUT(error);
3181 
3182 		if (nvlist_exists_string(nvl, "anchor"))
3183 			anchor = nvlist_get_string(nvl, "anchor");
3184 		if (nvlist_exists_string(nvl, "anchor_call"))
3185 			anchor_call = nvlist_get_string(nvl, "anchor_call");
3186 
3187 		if ((error = nvlist_error(nvl)))
3188 			ERROUT(error);
3189 
3190 		/* Frees rule on error */
3191 		error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor,
3192 		    anchor_call, td);
3193 
3194 		nvlist_destroy(nvl);
3195 		free(nvlpacked, M_NVLIST);
3196 		break;
3197 #undef ERROUT
3198 DIOCADDRULENV_error:
3199 		pf_krule_free(rule);
3200 		nvlist_destroy(nvl);
3201 		free(nvlpacked, M_NVLIST);
3202 
3203 		break;
3204 	}
3205 	case DIOCADDRULE: {
3206 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
3207 		struct pf_krule		*rule;
3208 
3209 		rule = pf_krule_alloc();
3210 		error = pf_rule_to_krule(&pr->rule, rule);
3211 		if (error != 0) {
3212 			pf_krule_free(rule);
3213 			break;
3214 		}
3215 
3216 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
3217 
3218 		/* Frees rule on error */
3219 		error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket,
3220 		    pr->anchor, pr->anchor_call, td);
3221 		break;
3222 	}
3223 
3224 	case DIOCGETRULES: {
3225 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
3226 		struct pf_kruleset	*ruleset;
3227 		struct pf_krule		*tail;
3228 		int			 rs_num;
3229 
3230 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
3231 
3232 		PF_RULES_WLOCK();
3233 		ruleset = pf_find_kruleset(pr->anchor);
3234 		if (ruleset == NULL) {
3235 			PF_RULES_WUNLOCK();
3236 			error = EINVAL;
3237 			break;
3238 		}
3239 		rs_num = pf_get_ruleset_number(pr->rule.action);
3240 		if (rs_num >= PF_RULESET_MAX) {
3241 			PF_RULES_WUNLOCK();
3242 			error = EINVAL;
3243 			break;
3244 		}
3245 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
3246 		    pf_krulequeue);
3247 		if (tail)
3248 			pr->nr = tail->nr + 1;
3249 		else
3250 			pr->nr = 0;
3251 		pr->ticket = ruleset->rules[rs_num].active.ticket;
3252 		PF_RULES_WUNLOCK();
3253 		break;
3254 	}
3255 
3256 	case DIOCGETRULE: {
3257 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
3258 		struct pf_kruleset	*ruleset;
3259 		struct pf_krule		*rule;
3260 		int			 rs_num;
3261 
3262 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
3263 
3264 		PF_RULES_WLOCK();
3265 		ruleset = pf_find_kruleset(pr->anchor);
3266 		if (ruleset == NULL) {
3267 			PF_RULES_WUNLOCK();
3268 			error = EINVAL;
3269 			break;
3270 		}
3271 		rs_num = pf_get_ruleset_number(pr->rule.action);
3272 		if (rs_num >= PF_RULESET_MAX) {
3273 			PF_RULES_WUNLOCK();
3274 			error = EINVAL;
3275 			break;
3276 		}
3277 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
3278 			PF_RULES_WUNLOCK();
3279 			error = EBUSY;
3280 			break;
3281 		}
3282 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
3283 		while ((rule != NULL) && (rule->nr != pr->nr))
3284 			rule = TAILQ_NEXT(rule, entries);
3285 		if (rule == NULL) {
3286 			PF_RULES_WUNLOCK();
3287 			error = EBUSY;
3288 			break;
3289 		}
3290 
3291 		pf_krule_to_rule(rule, &pr->rule);
3292 
3293 		if (pf_kanchor_copyout(ruleset, rule, pr)) {
3294 			PF_RULES_WUNLOCK();
3295 			error = EBUSY;
3296 			break;
3297 		}
3298 		pf_addr_copyout(&pr->rule.src.addr);
3299 		pf_addr_copyout(&pr->rule.dst.addr);
3300 
3301 		if (pr->action == PF_GET_CLR_CNTR) {
3302 			pf_counter_u64_zero(&rule->evaluations);
3303 			for (int i = 0; i < 2; i++) {
3304 				pf_counter_u64_zero(&rule->packets[i]);
3305 				pf_counter_u64_zero(&rule->bytes[i]);
3306 			}
3307 			counter_u64_zero(rule->states_tot);
3308 		}
3309 		PF_RULES_WUNLOCK();
3310 		break;
3311 	}
3312 
3313 	case DIOCGETRULENV: {
3314 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
3315 		nvlist_t		*nvrule = NULL;
3316 		nvlist_t		*nvl = NULL;
3317 		struct pf_kruleset	*ruleset;
3318 		struct pf_krule		*rule;
3319 		void			*nvlpacked = NULL;
3320 		int			 rs_num, nr;
3321 		bool			 clear_counter = false;
3322 
3323 #define	ERROUT(x)	ERROUT_IOCTL(DIOCGETRULENV_error, x)
3324 
3325 		if (nv->len > pf_ioctl_maxcount)
3326 			ERROUT(ENOMEM);
3327 
3328 		/* Copy the request in */
3329 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3330 		if (nvlpacked == NULL)
3331 			ERROUT(ENOMEM);
3332 
3333 		error = copyin(nv->data, nvlpacked, nv->len);
3334 		if (error)
3335 			ERROUT(error);
3336 
3337 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3338 		if (nvl == NULL)
3339 			ERROUT(EBADMSG);
3340 
3341 		if (! nvlist_exists_string(nvl, "anchor"))
3342 			ERROUT(EBADMSG);
3343 		if (! nvlist_exists_number(nvl, "ruleset"))
3344 			ERROUT(EBADMSG);
3345 		if (! nvlist_exists_number(nvl, "ticket"))
3346 			ERROUT(EBADMSG);
3347 		if (! nvlist_exists_number(nvl, "nr"))
3348 			ERROUT(EBADMSG);
3349 
3350 		if (nvlist_exists_bool(nvl, "clear_counter"))
3351 			clear_counter = nvlist_get_bool(nvl, "clear_counter");
3352 
3353 		if (clear_counter && !(flags & FWRITE))
3354 			ERROUT(EACCES);
3355 
3356 		nr = nvlist_get_number(nvl, "nr");
3357 
3358 		PF_RULES_WLOCK();
3359 		ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor"));
3360 		if (ruleset == NULL) {
3361 			PF_RULES_WUNLOCK();
3362 			ERROUT(ENOENT);
3363 		}
3364 
3365 		rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset"));
3366 		if (rs_num >= PF_RULESET_MAX) {
3367 			PF_RULES_WUNLOCK();
3368 			ERROUT(EINVAL);
3369 		}
3370 
3371 		if (nvlist_get_number(nvl, "ticket") !=
3372 		    ruleset->rules[rs_num].active.ticket) {
3373 			PF_RULES_WUNLOCK();
3374 			ERROUT(EBUSY);
3375 		}
3376 
3377 		if ((error = nvlist_error(nvl))) {
3378 			PF_RULES_WUNLOCK();
3379 			ERROUT(error);
3380 		}
3381 
3382 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
3383 		while ((rule != NULL) && (rule->nr != nr))
3384 			rule = TAILQ_NEXT(rule, entries);
3385 		if (rule == NULL) {
3386 			PF_RULES_WUNLOCK();
3387 			ERROUT(EBUSY);
3388 		}
3389 
3390 		nvrule = pf_krule_to_nvrule(rule);
3391 
3392 		nvlist_destroy(nvl);
3393 		nvl = nvlist_create(0);
3394 		if (nvl == NULL) {
3395 			PF_RULES_WUNLOCK();
3396 			ERROUT(ENOMEM);
3397 		}
3398 		nvlist_add_number(nvl, "nr", nr);
3399 		nvlist_add_nvlist(nvl, "rule", nvrule);
3400 		nvlist_destroy(nvrule);
3401 		nvrule = NULL;
3402 		if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) {
3403 			PF_RULES_WUNLOCK();
3404 			ERROUT(EBUSY);
3405 		}
3406 
3407 		free(nvlpacked, M_NVLIST);
3408 		nvlpacked = nvlist_pack(nvl, &nv->len);
3409 		if (nvlpacked == NULL) {
3410 			PF_RULES_WUNLOCK();
3411 			ERROUT(ENOMEM);
3412 		}
3413 
3414 		if (nv->size == 0) {
3415 			PF_RULES_WUNLOCK();
3416 			ERROUT(0);
3417 		}
3418 		else if (nv->size < nv->len) {
3419 			PF_RULES_WUNLOCK();
3420 			ERROUT(ENOSPC);
3421 		}
3422 
3423 		if (clear_counter) {
3424 			pf_counter_u64_zero(&rule->evaluations);
3425 			for (int i = 0; i < 2; i++) {
3426 				pf_counter_u64_zero(&rule->packets[i]);
3427 				pf_counter_u64_zero(&rule->bytes[i]);
3428 			}
3429 			counter_u64_zero(rule->states_tot);
3430 		}
3431 		PF_RULES_WUNLOCK();
3432 
3433 		error = copyout(nvlpacked, nv->data, nv->len);
3434 
3435 #undef ERROUT
3436 DIOCGETRULENV_error:
3437 		free(nvlpacked, M_NVLIST);
3438 		nvlist_destroy(nvrule);
3439 		nvlist_destroy(nvl);
3440 
3441 		break;
3442 	}
3443 
3444 	case DIOCCHANGERULE: {
3445 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
3446 		struct pf_kruleset	*ruleset;
3447 		struct pf_krule		*oldrule = NULL, *newrule = NULL;
3448 		struct pfi_kkif		*kif = NULL;
3449 		struct pf_kpooladdr	*pa;
3450 		u_int32_t		 nr = 0;
3451 		int			 rs_num;
3452 
3453 		pcr->anchor[sizeof(pcr->anchor) - 1] = 0;
3454 
3455 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
3456 		    pcr->action > PF_CHANGE_GET_TICKET) {
3457 			error = EINVAL;
3458 			break;
3459 		}
3460 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
3461 			error = EINVAL;
3462 			break;
3463 		}
3464 
3465 		if (pcr->action != PF_CHANGE_REMOVE) {
3466 			newrule = pf_krule_alloc();
3467 			error = pf_rule_to_krule(&pcr->rule, newrule);
3468 			if (error != 0) {
3469 				pf_krule_free(newrule);
3470 				break;
3471 			}
3472 
3473 			if (newrule->ifname[0])
3474 				kif = pf_kkif_create(M_WAITOK);
3475 			pf_counter_u64_init(&newrule->evaluations, M_WAITOK);
3476 			for (int i = 0; i < 2; i++) {
3477 				pf_counter_u64_init(&newrule->packets[i], M_WAITOK);
3478 				pf_counter_u64_init(&newrule->bytes[i], M_WAITOK);
3479 			}
3480 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
3481 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
3482 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
3483 			newrule->cuid = td->td_ucred->cr_ruid;
3484 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
3485 			TAILQ_INIT(&newrule->rpool.list);
3486 		}
3487 #define	ERROUT(x)	ERROUT_IOCTL(DIOCCHANGERULE_error, x)
3488 
3489 		PF_CONFIG_LOCK();
3490 		PF_RULES_WLOCK();
3491 #ifdef PF_WANT_32_TO_64_COUNTER
3492 		if (newrule != NULL) {
3493 			LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist);
3494 			newrule->allrulelinked = true;
3495 			V_pf_allrulecount++;
3496 		}
3497 #endif
3498 
3499 		if (!(pcr->action == PF_CHANGE_REMOVE ||
3500 		    pcr->action == PF_CHANGE_GET_TICKET) &&
3501 		    pcr->pool_ticket != V_ticket_pabuf)
3502 			ERROUT(EBUSY);
3503 
3504 		ruleset = pf_find_kruleset(pcr->anchor);
3505 		if (ruleset == NULL)
3506 			ERROUT(EINVAL);
3507 
3508 		rs_num = pf_get_ruleset_number(pcr->rule.action);
3509 		if (rs_num >= PF_RULESET_MAX)
3510 			ERROUT(EINVAL);
3511 
3512 		/*
3513 		 * XXXMJG: there is no guarantee that the ruleset was
3514 		 * created by the usual route of calling DIOCXBEGIN.
3515 		 * As a result it is possible the rule tree will not
3516 		 * be allocated yet. Hack around it by doing it here.
3517 		 * Note it is fine to let the tree persist in case of
3518 		 * error as it will be freed down the road on future
3519 		 * updates (if need be).
3520 		 */
3521 		if (ruleset->rules[rs_num].active.tree == NULL) {
3522 			ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT);
3523 			if (ruleset->rules[rs_num].active.tree == NULL) {
3524 				ERROUT(ENOMEM);
3525 			}
3526 		}
3527 
3528 		if (pcr->action == PF_CHANGE_GET_TICKET) {
3529 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
3530 			ERROUT(0);
3531 		} else if (pcr->ticket !=
3532 			    ruleset->rules[rs_num].active.ticket)
3533 				ERROUT(EINVAL);
3534 
3535 		if (pcr->action != PF_CHANGE_REMOVE) {
3536 			if (newrule->ifname[0]) {
3537 				newrule->kif = pfi_kkif_attach(kif,
3538 				    newrule->ifname);
3539 				kif = NULL;
3540 				pfi_kkif_ref(newrule->kif);
3541 			} else
3542 				newrule->kif = NULL;
3543 
3544 			if (newrule->rtableid > 0 &&
3545 			    newrule->rtableid >= rt_numfibs)
3546 				error = EBUSY;
3547 
3548 #ifdef ALTQ
3549 			/* set queue IDs */
3550 			if (newrule->qname[0] != 0) {
3551 				if ((newrule->qid =
3552 				    pf_qname2qid(newrule->qname)) == 0)
3553 					error = EBUSY;
3554 				else if (newrule->pqname[0] != 0) {
3555 					if ((newrule->pqid =
3556 					    pf_qname2qid(newrule->pqname)) == 0)
3557 						error = EBUSY;
3558 				} else
3559 					newrule->pqid = newrule->qid;
3560 			}
3561 #endif /* ALTQ */
3562 			if (newrule->tagname[0])
3563 				if ((newrule->tag =
3564 				    pf_tagname2tag(newrule->tagname)) == 0)
3565 					error = EBUSY;
3566 			if (newrule->match_tagname[0])
3567 				if ((newrule->match_tag = pf_tagname2tag(
3568 				    newrule->match_tagname)) == 0)
3569 					error = EBUSY;
3570 			if (newrule->rt && !newrule->direction)
3571 				error = EINVAL;
3572 			if (!newrule->log)
3573 				newrule->logif = 0;
3574 			if (newrule->logif >= PFLOGIFS_MAX)
3575 				error = EINVAL;
3576 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
3577 				error = ENOMEM;
3578 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
3579 				error = ENOMEM;
3580 			if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
3581 				error = EINVAL;
3582 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
3583 				if (pa->addr.type == PF_ADDR_TABLE) {
3584 					pa->addr.p.tbl =
3585 					    pfr_attach_table(ruleset,
3586 					    pa->addr.v.tblname);
3587 					if (pa->addr.p.tbl == NULL)
3588 						error = ENOMEM;
3589 				}
3590 
3591 			newrule->overload_tbl = NULL;
3592 			if (newrule->overload_tblname[0]) {
3593 				if ((newrule->overload_tbl = pfr_attach_table(
3594 				    ruleset, newrule->overload_tblname)) ==
3595 				    NULL)
3596 					error = EINVAL;
3597 				else
3598 					newrule->overload_tbl->pfrkt_flags |=
3599 					    PFR_TFLAG_ACTIVE;
3600 			}
3601 
3602 			pf_mv_kpool(&V_pf_pabuf, &newrule->rpool.list);
3603 			if (((((newrule->action == PF_NAT) ||
3604 			    (newrule->action == PF_RDR) ||
3605 			    (newrule->action == PF_BINAT) ||
3606 			    (newrule->rt > PF_NOPFROUTE)) &&
3607 			    !newrule->anchor)) &&
3608 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
3609 				error = EINVAL;
3610 
3611 			if (error) {
3612 				pf_free_rule(newrule);
3613 				PF_RULES_WUNLOCK();
3614 				PF_CONFIG_UNLOCK();
3615 				break;
3616 			}
3617 
3618 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
3619 		}
3620 		pf_empty_kpool(&V_pf_pabuf);
3621 
3622 		if (pcr->action == PF_CHANGE_ADD_HEAD)
3623 			oldrule = TAILQ_FIRST(
3624 			    ruleset->rules[rs_num].active.ptr);
3625 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
3626 			oldrule = TAILQ_LAST(
3627 			    ruleset->rules[rs_num].active.ptr, pf_krulequeue);
3628 		else {
3629 			oldrule = TAILQ_FIRST(
3630 			    ruleset->rules[rs_num].active.ptr);
3631 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
3632 				oldrule = TAILQ_NEXT(oldrule, entries);
3633 			if (oldrule == NULL) {
3634 				if (newrule != NULL)
3635 					pf_free_rule(newrule);
3636 				PF_RULES_WUNLOCK();
3637 				PF_CONFIG_UNLOCK();
3638 				error = EINVAL;
3639 				break;
3640 			}
3641 		}
3642 
3643 		if (pcr->action == PF_CHANGE_REMOVE) {
3644 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
3645 			    oldrule);
3646 			RB_REMOVE(pf_krule_global,
3647 			    ruleset->rules[rs_num].active.tree, oldrule);
3648 			ruleset->rules[rs_num].active.rcount--;
3649 		} else {
3650 			pf_hash_rule(newrule);
3651 			if (RB_INSERT(pf_krule_global,
3652 			    ruleset->rules[rs_num].active.tree, newrule) != NULL) {
3653 				pf_free_rule(newrule);
3654 				PF_RULES_WUNLOCK();
3655 				PF_CONFIG_UNLOCK();
3656 				error = EEXIST;
3657 				break;
3658 			}
3659 
3660 			if (oldrule == NULL)
3661 				TAILQ_INSERT_TAIL(
3662 				    ruleset->rules[rs_num].active.ptr,
3663 				    newrule, entries);
3664 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
3665 			    pcr->action == PF_CHANGE_ADD_BEFORE)
3666 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
3667 			else
3668 				TAILQ_INSERT_AFTER(
3669 				    ruleset->rules[rs_num].active.ptr,
3670 				    oldrule, newrule, entries);
3671 			ruleset->rules[rs_num].active.rcount++;
3672 		}
3673 
3674 		nr = 0;
3675 		TAILQ_FOREACH(oldrule,
3676 		    ruleset->rules[rs_num].active.ptr, entries)
3677 			oldrule->nr = nr++;
3678 
3679 		ruleset->rules[rs_num].active.ticket++;
3680 
3681 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
3682 		pf_remove_if_empty_kruleset(ruleset);
3683 
3684 		PF_RULES_WUNLOCK();
3685 		PF_CONFIG_UNLOCK();
3686 		break;
3687 
3688 #undef ERROUT
3689 DIOCCHANGERULE_error:
3690 		PF_RULES_WUNLOCK();
3691 		PF_CONFIG_UNLOCK();
3692 		pf_krule_free(newrule);
3693 		pf_kkif_free(kif);
3694 		break;
3695 	}
3696 
3697 	case DIOCCLRSTATES: {
3698 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
3699 		struct pf_kstate_kill	 kill;
3700 
3701 		error = pf_state_kill_to_kstate_kill(psk, &kill);
3702 		if (error)
3703 			break;
3704 
3705 		psk->psk_killed = pf_clear_states(&kill);
3706 		break;
3707 	}
3708 
3709 	case DIOCCLRSTATESNV: {
3710 		error = pf_clearstates_nv((struct pfioc_nv *)addr);
3711 		break;
3712 	}
3713 
3714 	case DIOCKILLSTATES: {
3715 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
3716 		struct pf_kstate_kill	 kill;
3717 
3718 		error = pf_state_kill_to_kstate_kill(psk, &kill);
3719 		if (error)
3720 			break;
3721 
3722 		psk->psk_killed = 0;
3723 		pf_killstates(&kill, &psk->psk_killed);
3724 		break;
3725 	}
3726 
3727 	case DIOCKILLSTATESNV: {
3728 		error = pf_killstates_nv((struct pfioc_nv *)addr);
3729 		break;
3730 	}
3731 
3732 	case DIOCADDSTATE: {
3733 		struct pfioc_state		*ps = (struct pfioc_state *)addr;
3734 		struct pfsync_state_1301	*sp = &ps->state;
3735 
3736 		if (sp->timeout >= PFTM_MAX) {
3737 			error = EINVAL;
3738 			break;
3739 		}
3740 		if (V_pfsync_state_import_ptr != NULL) {
3741 			PF_RULES_RLOCK();
3742 			error = V_pfsync_state_import_ptr(
3743 			    (union pfsync_state_union *)sp, PFSYNC_SI_IOCTL,
3744 			    PFSYNC_MSG_VERSION_1301);
3745 			PF_RULES_RUNLOCK();
3746 		} else
3747 			error = EOPNOTSUPP;
3748 		break;
3749 	}
3750 
3751 	case DIOCGETSTATE: {
3752 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
3753 		struct pf_kstate	*s;
3754 
3755 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
3756 		if (s == NULL) {
3757 			error = ENOENT;
3758 			break;
3759 		}
3760 
3761 		pfsync_state_export((union pfsync_state_union*)&ps->state,
3762 		    s, PFSYNC_MSG_VERSION_1301);
3763 		PF_STATE_UNLOCK(s);
3764 		break;
3765 	}
3766 
3767 	case DIOCGETSTATENV: {
3768 		error = pf_getstate((struct pfioc_nv *)addr);
3769 		break;
3770 	}
3771 
3772 	case DIOCGETSTATES: {
3773 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
3774 		struct pf_kstate	*s;
3775 		struct pfsync_state_1301	*pstore, *p;
3776 		int			 i, nr;
3777 		size_t			 slice_count = 16, count;
3778 		void			*out;
3779 
3780 		if (ps->ps_len <= 0) {
3781 			nr = uma_zone_get_cur(V_pf_state_z);
3782 			ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3783 			break;
3784 		}
3785 
3786 		out = ps->ps_states;
3787 		pstore = mallocarray(slice_count,
3788 		    sizeof(struct pfsync_state_1301), M_TEMP, M_WAITOK | M_ZERO);
3789 		nr = 0;
3790 
3791 		for (i = 0; i <= pf_hashmask; i++) {
3792 			struct pf_idhash *ih = &V_pf_idhash[i];
3793 
3794 DIOCGETSTATES_retry:
3795 			p = pstore;
3796 
3797 			if (LIST_EMPTY(&ih->states))
3798 				continue;
3799 
3800 			PF_HASHROW_LOCK(ih);
3801 			count = 0;
3802 			LIST_FOREACH(s, &ih->states, entry) {
3803 				if (s->timeout == PFTM_UNLINKED)
3804 					continue;
3805 				count++;
3806 			}
3807 
3808 			if (count > slice_count) {
3809 				PF_HASHROW_UNLOCK(ih);
3810 				free(pstore, M_TEMP);
3811 				slice_count = count * 2;
3812 				pstore = mallocarray(slice_count,
3813 				    sizeof(struct pfsync_state_1301), M_TEMP,
3814 				    M_WAITOK | M_ZERO);
3815 				goto DIOCGETSTATES_retry;
3816 			}
3817 
3818 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
3819 				PF_HASHROW_UNLOCK(ih);
3820 				goto DIOCGETSTATES_full;
3821 			}
3822 
3823 			LIST_FOREACH(s, &ih->states, entry) {
3824 				if (s->timeout == PFTM_UNLINKED)
3825 					continue;
3826 
3827 				pfsync_state_export((union pfsync_state_union*)p,
3828 				    s, PFSYNC_MSG_VERSION_1301);
3829 				p++;
3830 				nr++;
3831 			}
3832 			PF_HASHROW_UNLOCK(ih);
3833 			error = copyout(pstore, out,
3834 			    sizeof(struct pfsync_state_1301) * count);
3835 			if (error)
3836 				break;
3837 			out = ps->ps_states + nr;
3838 		}
3839 DIOCGETSTATES_full:
3840 		ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3841 		free(pstore, M_TEMP);
3842 
3843 		break;
3844 	}
3845 
3846 	case DIOCGETSTATESV2: {
3847 		struct pfioc_states_v2	*ps = (struct pfioc_states_v2 *)addr;
3848 		struct pf_kstate	*s;
3849 		struct pf_state_export	*pstore, *p;
3850 		int i, nr;
3851 		size_t slice_count = 16, count;
3852 		void *out;
3853 
3854 		if (ps->ps_req_version > PF_STATE_VERSION) {
3855 			error = ENOTSUP;
3856 			break;
3857 		}
3858 
3859 		if (ps->ps_len <= 0) {
3860 			nr = uma_zone_get_cur(V_pf_state_z);
3861 			ps->ps_len = sizeof(struct pf_state_export) * nr;
3862 			break;
3863 		}
3864 
3865 		out = ps->ps_states;
3866 		pstore = mallocarray(slice_count,
3867 		    sizeof(struct pf_state_export), M_TEMP, M_WAITOK | M_ZERO);
3868 		nr = 0;
3869 
3870 		for (i = 0; i <= pf_hashmask; i++) {
3871 			struct pf_idhash *ih = &V_pf_idhash[i];
3872 
3873 DIOCGETSTATESV2_retry:
3874 			p = pstore;
3875 
3876 			if (LIST_EMPTY(&ih->states))
3877 				continue;
3878 
3879 			PF_HASHROW_LOCK(ih);
3880 			count = 0;
3881 			LIST_FOREACH(s, &ih->states, entry) {
3882 				if (s->timeout == PFTM_UNLINKED)
3883 					continue;
3884 				count++;
3885 			}
3886 
3887 			if (count > slice_count) {
3888 				PF_HASHROW_UNLOCK(ih);
3889 				free(pstore, M_TEMP);
3890 				slice_count = count * 2;
3891 				pstore = mallocarray(slice_count,
3892 				    sizeof(struct pf_state_export), M_TEMP,
3893 				    M_WAITOK | M_ZERO);
3894 				goto DIOCGETSTATESV2_retry;
3895 			}
3896 
3897 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
3898 				PF_HASHROW_UNLOCK(ih);
3899 				goto DIOCGETSTATESV2_full;
3900 			}
3901 
3902 			LIST_FOREACH(s, &ih->states, entry) {
3903 				if (s->timeout == PFTM_UNLINKED)
3904 					continue;
3905 
3906 				pf_state_export(p, s);
3907 				p++;
3908 				nr++;
3909 			}
3910 			PF_HASHROW_UNLOCK(ih);
3911 			error = copyout(pstore, out,
3912 			    sizeof(struct pf_state_export) * count);
3913 			if (error)
3914 				break;
3915 			out = ps->ps_states + nr;
3916 		}
3917 DIOCGETSTATESV2_full:
3918 		ps->ps_len = nr * sizeof(struct pf_state_export);
3919 		free(pstore, M_TEMP);
3920 
3921 		break;
3922 	}
3923 
3924 	case DIOCGETSTATUS: {
3925 		struct pf_status *s = (struct pf_status *)addr;
3926 
3927 		PF_RULES_RLOCK();
3928 		s->running = V_pf_status.running;
3929 		s->since   = V_pf_status.since;
3930 		s->debug   = V_pf_status.debug;
3931 		s->hostid  = V_pf_status.hostid;
3932 		s->states  = V_pf_status.states;
3933 		s->src_nodes = V_pf_status.src_nodes;
3934 
3935 		for (int i = 0; i < PFRES_MAX; i++)
3936 			s->counters[i] =
3937 			    counter_u64_fetch(V_pf_status.counters[i]);
3938 		for (int i = 0; i < LCNT_MAX; i++)
3939 			s->lcounters[i] =
3940 			    counter_u64_fetch(V_pf_status.lcounters[i]);
3941 		for (int i = 0; i < FCNT_MAX; i++)
3942 			s->fcounters[i] =
3943 			    pf_counter_u64_fetch(&V_pf_status.fcounters[i]);
3944 		for (int i = 0; i < SCNT_MAX; i++)
3945 			s->scounters[i] =
3946 			    counter_u64_fetch(V_pf_status.scounters[i]);
3947 
3948 		bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ);
3949 		bcopy(V_pf_status.pf_chksum, s->pf_chksum,
3950 		    PF_MD5_DIGEST_LENGTH);
3951 
3952 		pfi_update_status(s->ifname, s);
3953 		PF_RULES_RUNLOCK();
3954 		break;
3955 	}
3956 
3957 	case DIOCGETSTATUSNV: {
3958 		error = pf_getstatus((struct pfioc_nv *)addr);
3959 		break;
3960 	}
3961 
3962 	case DIOCSETSTATUSIF: {
3963 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
3964 
3965 		if (pi->ifname[0] == 0) {
3966 			bzero(V_pf_status.ifname, IFNAMSIZ);
3967 			break;
3968 		}
3969 		PF_RULES_WLOCK();
3970 		error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
3971 		PF_RULES_WUNLOCK();
3972 		break;
3973 	}
3974 
3975 	case DIOCCLRSTATUS: {
3976 		PF_RULES_WLOCK();
3977 		for (int i = 0; i < PFRES_MAX; i++)
3978 			counter_u64_zero(V_pf_status.counters[i]);
3979 		for (int i = 0; i < FCNT_MAX; i++)
3980 			pf_counter_u64_zero(&V_pf_status.fcounters[i]);
3981 		for (int i = 0; i < SCNT_MAX; i++)
3982 			counter_u64_zero(V_pf_status.scounters[i]);
3983 		for (int i = 0; i < KLCNT_MAX; i++)
3984 			counter_u64_zero(V_pf_status.lcounters[i]);
3985 		V_pf_status.since = time_second;
3986 		if (*V_pf_status.ifname)
3987 			pfi_update_status(V_pf_status.ifname, NULL);
3988 		PF_RULES_WUNLOCK();
3989 		break;
3990 	}
3991 
3992 	case DIOCNATLOOK: {
3993 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
3994 		struct pf_state_key	*sk;
3995 		struct pf_kstate	*state;
3996 		struct pf_state_key_cmp	 key;
3997 		int			 m = 0, direction = pnl->direction;
3998 		int			 sidx, didx;
3999 
4000 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
4001 		sidx = (direction == PF_IN) ? 1 : 0;
4002 		didx = (direction == PF_IN) ? 0 : 1;
4003 
4004 		if (!pnl->proto ||
4005 		    PF_AZERO(&pnl->saddr, pnl->af) ||
4006 		    PF_AZERO(&pnl->daddr, pnl->af) ||
4007 		    ((pnl->proto == IPPROTO_TCP ||
4008 		    pnl->proto == IPPROTO_UDP) &&
4009 		    (!pnl->dport || !pnl->sport)))
4010 			error = EINVAL;
4011 		else {
4012 			bzero(&key, sizeof(key));
4013 			key.af = pnl->af;
4014 			key.proto = pnl->proto;
4015 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
4016 			key.port[sidx] = pnl->sport;
4017 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
4018 			key.port[didx] = pnl->dport;
4019 
4020 			state = pf_find_state_all(&key, direction, &m);
4021 			if (state == NULL) {
4022 				error = ENOENT;
4023 			} else {
4024 				if (m > 1) {
4025 					PF_STATE_UNLOCK(state);
4026 					error = E2BIG;	/* more than one state */
4027 				} else {
4028 					sk = state->key[sidx];
4029 					PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
4030 					pnl->rsport = sk->port[sidx];
4031 					PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
4032 					pnl->rdport = sk->port[didx];
4033 					PF_STATE_UNLOCK(state);
4034 				}
4035 			}
4036 		}
4037 		break;
4038 	}
4039 
4040 	case DIOCSETTIMEOUT: {
4041 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
4042 		int		 old;
4043 
4044 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
4045 		    pt->seconds < 0) {
4046 			error = EINVAL;
4047 			break;
4048 		}
4049 		PF_RULES_WLOCK();
4050 		old = V_pf_default_rule.timeout[pt->timeout];
4051 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
4052 			pt->seconds = 1;
4053 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
4054 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
4055 			wakeup(pf_purge_thread);
4056 		pt->seconds = old;
4057 		PF_RULES_WUNLOCK();
4058 		break;
4059 	}
4060 
4061 	case DIOCGETTIMEOUT: {
4062 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
4063 
4064 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
4065 			error = EINVAL;
4066 			break;
4067 		}
4068 		PF_RULES_RLOCK();
4069 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
4070 		PF_RULES_RUNLOCK();
4071 		break;
4072 	}
4073 
4074 	case DIOCGETLIMIT: {
4075 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
4076 
4077 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
4078 			error = EINVAL;
4079 			break;
4080 		}
4081 		PF_RULES_RLOCK();
4082 		pl->limit = V_pf_limits[pl->index].limit;
4083 		PF_RULES_RUNLOCK();
4084 		break;
4085 	}
4086 
4087 	case DIOCSETLIMIT: {
4088 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
4089 		int			 old_limit;
4090 
4091 		PF_RULES_WLOCK();
4092 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
4093 		    V_pf_limits[pl->index].zone == NULL) {
4094 			PF_RULES_WUNLOCK();
4095 			error = EINVAL;
4096 			break;
4097 		}
4098 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
4099 		old_limit = V_pf_limits[pl->index].limit;
4100 		V_pf_limits[pl->index].limit = pl->limit;
4101 		pl->limit = old_limit;
4102 		PF_RULES_WUNLOCK();
4103 		break;
4104 	}
4105 
4106 	case DIOCSETDEBUG: {
4107 		u_int32_t	*level = (u_int32_t *)addr;
4108 
4109 		PF_RULES_WLOCK();
4110 		V_pf_status.debug = *level;
4111 		PF_RULES_WUNLOCK();
4112 		break;
4113 	}
4114 
4115 	case DIOCCLRRULECTRS: {
4116 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
4117 		struct pf_kruleset	*ruleset = &pf_main_ruleset;
4118 		struct pf_krule		*rule;
4119 
4120 		PF_RULES_WLOCK();
4121 		TAILQ_FOREACH(rule,
4122 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
4123 			pf_counter_u64_zero(&rule->evaluations);
4124 			for (int i = 0; i < 2; i++) {
4125 				pf_counter_u64_zero(&rule->packets[i]);
4126 				pf_counter_u64_zero(&rule->bytes[i]);
4127 			}
4128 		}
4129 		PF_RULES_WUNLOCK();
4130 		break;
4131 	}
4132 
4133 	case DIOCGIFSPEEDV0:
4134 	case DIOCGIFSPEEDV1: {
4135 		struct pf_ifspeed_v1	*psp = (struct pf_ifspeed_v1 *)addr;
4136 		struct pf_ifspeed_v1	ps;
4137 		struct ifnet		*ifp;
4138 
4139 		if (psp->ifname[0] == '\0') {
4140 			error = EINVAL;
4141 			break;
4142 		}
4143 
4144 		error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ);
4145 		if (error != 0)
4146 			break;
4147 		ifp = ifunit(ps.ifname);
4148 		if (ifp != NULL) {
4149 			psp->baudrate32 =
4150 			    (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
4151 			if (cmd == DIOCGIFSPEEDV1)
4152 				psp->baudrate = ifp->if_baudrate;
4153 		} else {
4154 			error = EINVAL;
4155 		}
4156 		break;
4157 	}
4158 
4159 #ifdef ALTQ
4160 	case DIOCSTARTALTQ: {
4161 		struct pf_altq		*altq;
4162 
4163 		PF_RULES_WLOCK();
4164 		/* enable all altq interfaces on active list */
4165 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4166 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4167 				error = pf_enable_altq(altq);
4168 				if (error != 0)
4169 					break;
4170 			}
4171 		}
4172 		if (error == 0)
4173 			V_pf_altq_running = 1;
4174 		PF_RULES_WUNLOCK();
4175 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
4176 		break;
4177 	}
4178 
4179 	case DIOCSTOPALTQ: {
4180 		struct pf_altq		*altq;
4181 
4182 		PF_RULES_WLOCK();
4183 		/* disable all altq interfaces on active list */
4184 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4185 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4186 				error = pf_disable_altq(altq);
4187 				if (error != 0)
4188 					break;
4189 			}
4190 		}
4191 		if (error == 0)
4192 			V_pf_altq_running = 0;
4193 		PF_RULES_WUNLOCK();
4194 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
4195 		break;
4196 	}
4197 
4198 	case DIOCADDALTQV0:
4199 	case DIOCADDALTQV1: {
4200 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4201 		struct pf_altq		*altq, *a;
4202 		struct ifnet		*ifp;
4203 
4204 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
4205 		error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
4206 		if (error)
4207 			break;
4208 		altq->local_flags = 0;
4209 
4210 		PF_RULES_WLOCK();
4211 		if (pa->ticket != V_ticket_altqs_inactive) {
4212 			PF_RULES_WUNLOCK();
4213 			free(altq, M_PFALTQ);
4214 			error = EBUSY;
4215 			break;
4216 		}
4217 
4218 		/*
4219 		 * if this is for a queue, find the discipline and
4220 		 * copy the necessary fields
4221 		 */
4222 		if (altq->qname[0] != 0) {
4223 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
4224 				PF_RULES_WUNLOCK();
4225 				error = EBUSY;
4226 				free(altq, M_PFALTQ);
4227 				break;
4228 			}
4229 			altq->altq_disc = NULL;
4230 			TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
4231 				if (strncmp(a->ifname, altq->ifname,
4232 				    IFNAMSIZ) == 0) {
4233 					altq->altq_disc = a->altq_disc;
4234 					break;
4235 				}
4236 			}
4237 		}
4238 
4239 		if ((ifp = ifunit(altq->ifname)) == NULL)
4240 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
4241 		else
4242 			error = altq_add(ifp, altq);
4243 
4244 		if (error) {
4245 			PF_RULES_WUNLOCK();
4246 			free(altq, M_PFALTQ);
4247 			break;
4248 		}
4249 
4250 		if (altq->qname[0] != 0)
4251 			TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
4252 		else
4253 			TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
4254 		/* version error check done on import above */
4255 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4256 		PF_RULES_WUNLOCK();
4257 		break;
4258 	}
4259 
4260 	case DIOCGETALTQSV0:
4261 	case DIOCGETALTQSV1: {
4262 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4263 		struct pf_altq		*altq;
4264 
4265 		PF_RULES_RLOCK();
4266 		pa->nr = 0;
4267 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
4268 			pa->nr++;
4269 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
4270 			pa->nr++;
4271 		pa->ticket = V_ticket_altqs_active;
4272 		PF_RULES_RUNLOCK();
4273 		break;
4274 	}
4275 
4276 	case DIOCGETALTQV0:
4277 	case DIOCGETALTQV1: {
4278 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4279 		struct pf_altq		*altq;
4280 
4281 		PF_RULES_RLOCK();
4282 		if (pa->ticket != V_ticket_altqs_active) {
4283 			PF_RULES_RUNLOCK();
4284 			error = EBUSY;
4285 			break;
4286 		}
4287 		altq = pf_altq_get_nth_active(pa->nr);
4288 		if (altq == NULL) {
4289 			PF_RULES_RUNLOCK();
4290 			error = EBUSY;
4291 			break;
4292 		}
4293 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4294 		PF_RULES_RUNLOCK();
4295 		break;
4296 	}
4297 
4298 	case DIOCCHANGEALTQV0:
4299 	case DIOCCHANGEALTQV1:
4300 		/* CHANGEALTQ not supported yet! */
4301 		error = ENODEV;
4302 		break;
4303 
4304 	case DIOCGETQSTATSV0:
4305 	case DIOCGETQSTATSV1: {
4306 		struct pfioc_qstats_v1	*pq = (struct pfioc_qstats_v1 *)addr;
4307 		struct pf_altq		*altq;
4308 		int			 nbytes;
4309 		u_int32_t		 version;
4310 
4311 		PF_RULES_RLOCK();
4312 		if (pq->ticket != V_ticket_altqs_active) {
4313 			PF_RULES_RUNLOCK();
4314 			error = EBUSY;
4315 			break;
4316 		}
4317 		nbytes = pq->nbytes;
4318 		altq = pf_altq_get_nth_active(pq->nr);
4319 		if (altq == NULL) {
4320 			PF_RULES_RUNLOCK();
4321 			error = EBUSY;
4322 			break;
4323 		}
4324 
4325 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
4326 			PF_RULES_RUNLOCK();
4327 			error = ENXIO;
4328 			break;
4329 		}
4330 		PF_RULES_RUNLOCK();
4331 		if (cmd == DIOCGETQSTATSV0)
4332 			version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
4333 		else
4334 			version = pq->version;
4335 		error = altq_getqstats(altq, pq->buf, &nbytes, version);
4336 		if (error == 0) {
4337 			pq->scheduler = altq->scheduler;
4338 			pq->nbytes = nbytes;
4339 		}
4340 		break;
4341 	}
4342 #endif /* ALTQ */
4343 
4344 	case DIOCBEGINADDRS: {
4345 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4346 
4347 		PF_RULES_WLOCK();
4348 		pf_empty_kpool(&V_pf_pabuf);
4349 		pp->ticket = ++V_ticket_pabuf;
4350 		PF_RULES_WUNLOCK();
4351 		break;
4352 	}
4353 
4354 	case DIOCADDADDR: {
4355 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4356 		struct pf_kpooladdr	*pa;
4357 		struct pfi_kkif		*kif = NULL;
4358 
4359 #ifndef INET
4360 		if (pp->af == AF_INET) {
4361 			error = EAFNOSUPPORT;
4362 			break;
4363 		}
4364 #endif /* INET */
4365 #ifndef INET6
4366 		if (pp->af == AF_INET6) {
4367 			error = EAFNOSUPPORT;
4368 			break;
4369 		}
4370 #endif /* INET6 */
4371 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
4372 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
4373 		    pp->addr.addr.type != PF_ADDR_TABLE) {
4374 			error = EINVAL;
4375 			break;
4376 		}
4377 		if (pp->addr.addr.p.dyn != NULL) {
4378 			error = EINVAL;
4379 			break;
4380 		}
4381 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
4382 		error = pf_pooladdr_to_kpooladdr(&pp->addr, pa);
4383 		if (error != 0)
4384 			break;
4385 		if (pa->ifname[0])
4386 			kif = pf_kkif_create(M_WAITOK);
4387 		PF_RULES_WLOCK();
4388 		if (pp->ticket != V_ticket_pabuf) {
4389 			PF_RULES_WUNLOCK();
4390 			if (pa->ifname[0])
4391 				pf_kkif_free(kif);
4392 			free(pa, M_PFRULE);
4393 			error = EBUSY;
4394 			break;
4395 		}
4396 		if (pa->ifname[0]) {
4397 			pa->kif = pfi_kkif_attach(kif, pa->ifname);
4398 			kif = NULL;
4399 			pfi_kkif_ref(pa->kif);
4400 		} else
4401 			pa->kif = NULL;
4402 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
4403 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
4404 			if (pa->ifname[0])
4405 				pfi_kkif_unref(pa->kif);
4406 			PF_RULES_WUNLOCK();
4407 			free(pa, M_PFRULE);
4408 			break;
4409 		}
4410 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
4411 		PF_RULES_WUNLOCK();
4412 		break;
4413 	}
4414 
4415 	case DIOCGETADDRS: {
4416 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4417 		struct pf_kpool		*pool;
4418 		struct pf_kpooladdr	*pa;
4419 
4420 		pp->anchor[sizeof(pp->anchor) - 1] = 0;
4421 		pp->nr = 0;
4422 
4423 		PF_RULES_RLOCK();
4424 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
4425 		    pp->r_num, 0, 1, 0);
4426 		if (pool == NULL) {
4427 			PF_RULES_RUNLOCK();
4428 			error = EBUSY;
4429 			break;
4430 		}
4431 		TAILQ_FOREACH(pa, &pool->list, entries)
4432 			pp->nr++;
4433 		PF_RULES_RUNLOCK();
4434 		break;
4435 	}
4436 
4437 	case DIOCGETADDR: {
4438 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4439 		struct pf_kpool		*pool;
4440 		struct pf_kpooladdr	*pa;
4441 		u_int32_t		 nr = 0;
4442 
4443 		pp->anchor[sizeof(pp->anchor) - 1] = 0;
4444 
4445 		PF_RULES_RLOCK();
4446 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
4447 		    pp->r_num, 0, 1, 1);
4448 		if (pool == NULL) {
4449 			PF_RULES_RUNLOCK();
4450 			error = EBUSY;
4451 			break;
4452 		}
4453 		pa = TAILQ_FIRST(&pool->list);
4454 		while ((pa != NULL) && (nr < pp->nr)) {
4455 			pa = TAILQ_NEXT(pa, entries);
4456 			nr++;
4457 		}
4458 		if (pa == NULL) {
4459 			PF_RULES_RUNLOCK();
4460 			error = EBUSY;
4461 			break;
4462 		}
4463 		pf_kpooladdr_to_pooladdr(pa, &pp->addr);
4464 		pf_addr_copyout(&pp->addr.addr);
4465 		PF_RULES_RUNLOCK();
4466 		break;
4467 	}
4468 
4469 	case DIOCCHANGEADDR: {
4470 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
4471 		struct pf_kpool		*pool;
4472 		struct pf_kpooladdr	*oldpa = NULL, *newpa = NULL;
4473 		struct pf_kruleset	*ruleset;
4474 		struct pfi_kkif		*kif = NULL;
4475 
4476 		pca->anchor[sizeof(pca->anchor) - 1] = 0;
4477 
4478 		if (pca->action < PF_CHANGE_ADD_HEAD ||
4479 		    pca->action > PF_CHANGE_REMOVE) {
4480 			error = EINVAL;
4481 			break;
4482 		}
4483 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
4484 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
4485 		    pca->addr.addr.type != PF_ADDR_TABLE) {
4486 			error = EINVAL;
4487 			break;
4488 		}
4489 		if (pca->addr.addr.p.dyn != NULL) {
4490 			error = EINVAL;
4491 			break;
4492 		}
4493 
4494 		if (pca->action != PF_CHANGE_REMOVE) {
4495 #ifndef INET
4496 			if (pca->af == AF_INET) {
4497 				error = EAFNOSUPPORT;
4498 				break;
4499 			}
4500 #endif /* INET */
4501 #ifndef INET6
4502 			if (pca->af == AF_INET6) {
4503 				error = EAFNOSUPPORT;
4504 				break;
4505 			}
4506 #endif /* INET6 */
4507 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
4508 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
4509 			if (newpa->ifname[0])
4510 				kif = pf_kkif_create(M_WAITOK);
4511 			newpa->kif = NULL;
4512 		}
4513 #define	ERROUT(x)	ERROUT_IOCTL(DIOCCHANGEADDR_error, x)
4514 		PF_RULES_WLOCK();
4515 		ruleset = pf_find_kruleset(pca->anchor);
4516 		if (ruleset == NULL)
4517 			ERROUT(EBUSY);
4518 
4519 		pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
4520 		    pca->r_num, pca->r_last, 1, 1);
4521 		if (pool == NULL)
4522 			ERROUT(EBUSY);
4523 
4524 		if (pca->action != PF_CHANGE_REMOVE) {
4525 			if (newpa->ifname[0]) {
4526 				newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
4527 				pfi_kkif_ref(newpa->kif);
4528 				kif = NULL;
4529 			}
4530 
4531 			switch (newpa->addr.type) {
4532 			case PF_ADDR_DYNIFTL:
4533 				error = pfi_dynaddr_setup(&newpa->addr,
4534 				    pca->af);
4535 				break;
4536 			case PF_ADDR_TABLE:
4537 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
4538 				    newpa->addr.v.tblname);
4539 				if (newpa->addr.p.tbl == NULL)
4540 					error = ENOMEM;
4541 				break;
4542 			}
4543 			if (error)
4544 				goto DIOCCHANGEADDR_error;
4545 		}
4546 
4547 		switch (pca->action) {
4548 		case PF_CHANGE_ADD_HEAD:
4549 			oldpa = TAILQ_FIRST(&pool->list);
4550 			break;
4551 		case PF_CHANGE_ADD_TAIL:
4552 			oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
4553 			break;
4554 		default:
4555 			oldpa = TAILQ_FIRST(&pool->list);
4556 			for (int i = 0; oldpa && i < pca->nr; i++)
4557 				oldpa = TAILQ_NEXT(oldpa, entries);
4558 
4559 			if (oldpa == NULL)
4560 				ERROUT(EINVAL);
4561 		}
4562 
4563 		if (pca->action == PF_CHANGE_REMOVE) {
4564 			TAILQ_REMOVE(&pool->list, oldpa, entries);
4565 			switch (oldpa->addr.type) {
4566 			case PF_ADDR_DYNIFTL:
4567 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
4568 				break;
4569 			case PF_ADDR_TABLE:
4570 				pfr_detach_table(oldpa->addr.p.tbl);
4571 				break;
4572 			}
4573 			if (oldpa->kif)
4574 				pfi_kkif_unref(oldpa->kif);
4575 			free(oldpa, M_PFRULE);
4576 		} else {
4577 			if (oldpa == NULL)
4578 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
4579 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
4580 			    pca->action == PF_CHANGE_ADD_BEFORE)
4581 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
4582 			else
4583 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
4584 				    newpa, entries);
4585 		}
4586 
4587 		pool->cur = TAILQ_FIRST(&pool->list);
4588 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
4589 		PF_RULES_WUNLOCK();
4590 		break;
4591 
4592 #undef ERROUT
4593 DIOCCHANGEADDR_error:
4594 		if (newpa != NULL) {
4595 			if (newpa->kif)
4596 				pfi_kkif_unref(newpa->kif);
4597 			free(newpa, M_PFRULE);
4598 		}
4599 		PF_RULES_WUNLOCK();
4600 		pf_kkif_free(kif);
4601 		break;
4602 	}
4603 
4604 	case DIOCGETRULESETS: {
4605 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
4606 		struct pf_kruleset	*ruleset;
4607 		struct pf_kanchor	*anchor;
4608 
4609 		pr->path[sizeof(pr->path) - 1] = 0;
4610 
4611 		PF_RULES_RLOCK();
4612 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4613 			PF_RULES_RUNLOCK();
4614 			error = ENOENT;
4615 			break;
4616 		}
4617 		pr->nr = 0;
4618 		if (ruleset->anchor == NULL) {
4619 			/* XXX kludge for pf_main_ruleset */
4620 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4621 				if (anchor->parent == NULL)
4622 					pr->nr++;
4623 		} else {
4624 			RB_FOREACH(anchor, pf_kanchor_node,
4625 			    &ruleset->anchor->children)
4626 				pr->nr++;
4627 		}
4628 		PF_RULES_RUNLOCK();
4629 		break;
4630 	}
4631 
4632 	case DIOCGETRULESET: {
4633 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
4634 		struct pf_kruleset	*ruleset;
4635 		struct pf_kanchor	*anchor;
4636 		u_int32_t		 nr = 0;
4637 
4638 		pr->path[sizeof(pr->path) - 1] = 0;
4639 
4640 		PF_RULES_RLOCK();
4641 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4642 			PF_RULES_RUNLOCK();
4643 			error = ENOENT;
4644 			break;
4645 		}
4646 		pr->name[0] = 0;
4647 		if (ruleset->anchor == NULL) {
4648 			/* XXX kludge for pf_main_ruleset */
4649 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4650 				if (anchor->parent == NULL && nr++ == pr->nr) {
4651 					strlcpy(pr->name, anchor->name,
4652 					    sizeof(pr->name));
4653 					break;
4654 				}
4655 		} else {
4656 			RB_FOREACH(anchor, pf_kanchor_node,
4657 			    &ruleset->anchor->children)
4658 				if (nr++ == pr->nr) {
4659 					strlcpy(pr->name, anchor->name,
4660 					    sizeof(pr->name));
4661 					break;
4662 				}
4663 		}
4664 		if (!pr->name[0])
4665 			error = EBUSY;
4666 		PF_RULES_RUNLOCK();
4667 		break;
4668 	}
4669 
4670 	case DIOCRCLRTABLES: {
4671 		struct pfioc_table *io = (struct pfioc_table *)addr;
4672 
4673 		if (io->pfrio_esize != 0) {
4674 			error = ENODEV;
4675 			break;
4676 		}
4677 		PF_RULES_WLOCK();
4678 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
4679 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4680 		PF_RULES_WUNLOCK();
4681 		break;
4682 	}
4683 
4684 	case DIOCRADDTABLES: {
4685 		struct pfioc_table *io = (struct pfioc_table *)addr;
4686 		struct pfr_table *pfrts;
4687 		size_t totlen;
4688 
4689 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4690 			error = ENODEV;
4691 			break;
4692 		}
4693 
4694 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4695 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4696 			error = ENOMEM;
4697 			break;
4698 		}
4699 
4700 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4701 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4702 		    M_TEMP, M_WAITOK);
4703 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4704 		if (error) {
4705 			free(pfrts, M_TEMP);
4706 			break;
4707 		}
4708 		PF_RULES_WLOCK();
4709 		error = pfr_add_tables(pfrts, io->pfrio_size,
4710 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4711 		PF_RULES_WUNLOCK();
4712 		free(pfrts, M_TEMP);
4713 		break;
4714 	}
4715 
4716 	case DIOCRDELTABLES: {
4717 		struct pfioc_table *io = (struct pfioc_table *)addr;
4718 		struct pfr_table *pfrts;
4719 		size_t totlen;
4720 
4721 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4722 			error = ENODEV;
4723 			break;
4724 		}
4725 
4726 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4727 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4728 			error = ENOMEM;
4729 			break;
4730 		}
4731 
4732 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4733 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4734 		    M_TEMP, M_WAITOK);
4735 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4736 		if (error) {
4737 			free(pfrts, M_TEMP);
4738 			break;
4739 		}
4740 		PF_RULES_WLOCK();
4741 		error = pfr_del_tables(pfrts, io->pfrio_size,
4742 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4743 		PF_RULES_WUNLOCK();
4744 		free(pfrts, M_TEMP);
4745 		break;
4746 	}
4747 
4748 	case DIOCRGETTABLES: {
4749 		struct pfioc_table *io = (struct pfioc_table *)addr;
4750 		struct pfr_table *pfrts;
4751 		size_t totlen;
4752 		int n;
4753 
4754 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4755 			error = ENODEV;
4756 			break;
4757 		}
4758 		PF_RULES_RLOCK();
4759 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4760 		if (n < 0) {
4761 			PF_RULES_RUNLOCK();
4762 			error = EINVAL;
4763 			break;
4764 		}
4765 		io->pfrio_size = min(io->pfrio_size, n);
4766 
4767 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4768 
4769 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4770 		    M_TEMP, M_NOWAIT | M_ZERO);
4771 		if (pfrts == NULL) {
4772 			error = ENOMEM;
4773 			PF_RULES_RUNLOCK();
4774 			break;
4775 		}
4776 		error = pfr_get_tables(&io->pfrio_table, pfrts,
4777 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4778 		PF_RULES_RUNLOCK();
4779 		if (error == 0)
4780 			error = copyout(pfrts, io->pfrio_buffer, totlen);
4781 		free(pfrts, M_TEMP);
4782 		break;
4783 	}
4784 
4785 	case DIOCRGETTSTATS: {
4786 		struct pfioc_table *io = (struct pfioc_table *)addr;
4787 		struct pfr_tstats *pfrtstats;
4788 		size_t totlen;
4789 		int n;
4790 
4791 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
4792 			error = ENODEV;
4793 			break;
4794 		}
4795 		PF_TABLE_STATS_LOCK();
4796 		PF_RULES_RLOCK();
4797 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4798 		if (n < 0) {
4799 			PF_RULES_RUNLOCK();
4800 			PF_TABLE_STATS_UNLOCK();
4801 			error = EINVAL;
4802 			break;
4803 		}
4804 		io->pfrio_size = min(io->pfrio_size, n);
4805 
4806 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
4807 		pfrtstats = mallocarray(io->pfrio_size,
4808 		    sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT | M_ZERO);
4809 		if (pfrtstats == NULL) {
4810 			error = ENOMEM;
4811 			PF_RULES_RUNLOCK();
4812 			PF_TABLE_STATS_UNLOCK();
4813 			break;
4814 		}
4815 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
4816 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4817 		PF_RULES_RUNLOCK();
4818 		PF_TABLE_STATS_UNLOCK();
4819 		if (error == 0)
4820 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
4821 		free(pfrtstats, M_TEMP);
4822 		break;
4823 	}
4824 
4825 	case DIOCRCLRTSTATS: {
4826 		struct pfioc_table *io = (struct pfioc_table *)addr;
4827 		struct pfr_table *pfrts;
4828 		size_t totlen;
4829 
4830 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4831 			error = ENODEV;
4832 			break;
4833 		}
4834 
4835 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4836 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4837 			/* We used to count tables and use the minimum required
4838 			 * size, so we didn't fail on overly large requests.
4839 			 * Keep doing so. */
4840 			io->pfrio_size = pf_ioctl_maxcount;
4841 			break;
4842 		}
4843 
4844 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4845 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4846 		    M_TEMP, M_WAITOK);
4847 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4848 		if (error) {
4849 			free(pfrts, M_TEMP);
4850 			break;
4851 		}
4852 
4853 		PF_TABLE_STATS_LOCK();
4854 		PF_RULES_RLOCK();
4855 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
4856 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4857 		PF_RULES_RUNLOCK();
4858 		PF_TABLE_STATS_UNLOCK();
4859 		free(pfrts, M_TEMP);
4860 		break;
4861 	}
4862 
4863 	case DIOCRSETTFLAGS: {
4864 		struct pfioc_table *io = (struct pfioc_table *)addr;
4865 		struct pfr_table *pfrts;
4866 		size_t totlen;
4867 		int n;
4868 
4869 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4870 			error = ENODEV;
4871 			break;
4872 		}
4873 
4874 		PF_RULES_RLOCK();
4875 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4876 		if (n < 0) {
4877 			PF_RULES_RUNLOCK();
4878 			error = EINVAL;
4879 			break;
4880 		}
4881 
4882 		io->pfrio_size = min(io->pfrio_size, n);
4883 		PF_RULES_RUNLOCK();
4884 
4885 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4886 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4887 		    M_TEMP, M_WAITOK);
4888 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4889 		if (error) {
4890 			free(pfrts, M_TEMP);
4891 			break;
4892 		}
4893 		PF_RULES_WLOCK();
4894 		error = pfr_set_tflags(pfrts, io->pfrio_size,
4895 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
4896 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4897 		PF_RULES_WUNLOCK();
4898 		free(pfrts, M_TEMP);
4899 		break;
4900 	}
4901 
4902 	case DIOCRCLRADDRS: {
4903 		struct pfioc_table *io = (struct pfioc_table *)addr;
4904 
4905 		if (io->pfrio_esize != 0) {
4906 			error = ENODEV;
4907 			break;
4908 		}
4909 		PF_RULES_WLOCK();
4910 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
4911 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4912 		PF_RULES_WUNLOCK();
4913 		break;
4914 	}
4915 
4916 	case DIOCRADDADDRS: {
4917 		struct pfioc_table *io = (struct pfioc_table *)addr;
4918 		struct pfr_addr *pfras;
4919 		size_t totlen;
4920 
4921 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4922 			error = ENODEV;
4923 			break;
4924 		}
4925 		if (io->pfrio_size < 0 ||
4926 		    io->pfrio_size > pf_ioctl_maxcount ||
4927 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4928 			error = EINVAL;
4929 			break;
4930 		}
4931 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4932 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4933 		    M_TEMP, M_WAITOK);
4934 		error = copyin(io->pfrio_buffer, pfras, totlen);
4935 		if (error) {
4936 			free(pfras, M_TEMP);
4937 			break;
4938 		}
4939 		PF_RULES_WLOCK();
4940 		error = pfr_add_addrs(&io->pfrio_table, pfras,
4941 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
4942 		    PFR_FLAG_USERIOCTL);
4943 		PF_RULES_WUNLOCK();
4944 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4945 			error = copyout(pfras, io->pfrio_buffer, totlen);
4946 		free(pfras, M_TEMP);
4947 		break;
4948 	}
4949 
4950 	case DIOCRDELADDRS: {
4951 		struct pfioc_table *io = (struct pfioc_table *)addr;
4952 		struct pfr_addr *pfras;
4953 		size_t totlen;
4954 
4955 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4956 			error = ENODEV;
4957 			break;
4958 		}
4959 		if (io->pfrio_size < 0 ||
4960 		    io->pfrio_size > pf_ioctl_maxcount ||
4961 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4962 			error = EINVAL;
4963 			break;
4964 		}
4965 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4966 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4967 		    M_TEMP, M_WAITOK);
4968 		error = copyin(io->pfrio_buffer, pfras, totlen);
4969 		if (error) {
4970 			free(pfras, M_TEMP);
4971 			break;
4972 		}
4973 		PF_RULES_WLOCK();
4974 		error = pfr_del_addrs(&io->pfrio_table, pfras,
4975 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
4976 		    PFR_FLAG_USERIOCTL);
4977 		PF_RULES_WUNLOCK();
4978 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4979 			error = copyout(pfras, io->pfrio_buffer, totlen);
4980 		free(pfras, M_TEMP);
4981 		break;
4982 	}
4983 
4984 	case DIOCRSETADDRS: {
4985 		struct pfioc_table *io = (struct pfioc_table *)addr;
4986 		struct pfr_addr *pfras;
4987 		size_t totlen, count;
4988 
4989 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4990 			error = ENODEV;
4991 			break;
4992 		}
4993 		if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
4994 			error = EINVAL;
4995 			break;
4996 		}
4997 		count = max(io->pfrio_size, io->pfrio_size2);
4998 		if (count > pf_ioctl_maxcount ||
4999 		    WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
5000 			error = EINVAL;
5001 			break;
5002 		}
5003 		totlen = count * sizeof(struct pfr_addr);
5004 		pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
5005 		    M_WAITOK);
5006 		error = copyin(io->pfrio_buffer, pfras, totlen);
5007 		if (error) {
5008 			free(pfras, M_TEMP);
5009 			break;
5010 		}
5011 		PF_RULES_WLOCK();
5012 		error = pfr_set_addrs(&io->pfrio_table, pfras,
5013 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
5014 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
5015 		    PFR_FLAG_USERIOCTL, 0);
5016 		PF_RULES_WUNLOCK();
5017 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
5018 			error = copyout(pfras, io->pfrio_buffer, totlen);
5019 		free(pfras, M_TEMP);
5020 		break;
5021 	}
5022 
5023 	case DIOCRGETADDRS: {
5024 		struct pfioc_table *io = (struct pfioc_table *)addr;
5025 		struct pfr_addr *pfras;
5026 		size_t totlen;
5027 
5028 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5029 			error = ENODEV;
5030 			break;
5031 		}
5032 		if (io->pfrio_size < 0 ||
5033 		    io->pfrio_size > pf_ioctl_maxcount ||
5034 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5035 			error = EINVAL;
5036 			break;
5037 		}
5038 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
5039 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5040 		    M_TEMP, M_WAITOK | M_ZERO);
5041 		PF_RULES_RLOCK();
5042 		error = pfr_get_addrs(&io->pfrio_table, pfras,
5043 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5044 		PF_RULES_RUNLOCK();
5045 		if (error == 0)
5046 			error = copyout(pfras, io->pfrio_buffer, totlen);
5047 		free(pfras, M_TEMP);
5048 		break;
5049 	}
5050 
5051 	case DIOCRGETASTATS: {
5052 		struct pfioc_table *io = (struct pfioc_table *)addr;
5053 		struct pfr_astats *pfrastats;
5054 		size_t totlen;
5055 
5056 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
5057 			error = ENODEV;
5058 			break;
5059 		}
5060 		if (io->pfrio_size < 0 ||
5061 		    io->pfrio_size > pf_ioctl_maxcount ||
5062 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
5063 			error = EINVAL;
5064 			break;
5065 		}
5066 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
5067 		pfrastats = mallocarray(io->pfrio_size,
5068 		    sizeof(struct pfr_astats), M_TEMP, M_WAITOK | M_ZERO);
5069 		PF_RULES_RLOCK();
5070 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
5071 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5072 		PF_RULES_RUNLOCK();
5073 		if (error == 0)
5074 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
5075 		free(pfrastats, M_TEMP);
5076 		break;
5077 	}
5078 
5079 	case DIOCRCLRASTATS: {
5080 		struct pfioc_table *io = (struct pfioc_table *)addr;
5081 		struct pfr_addr *pfras;
5082 		size_t totlen;
5083 
5084 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5085 			error = ENODEV;
5086 			break;
5087 		}
5088 		if (io->pfrio_size < 0 ||
5089 		    io->pfrio_size > pf_ioctl_maxcount ||
5090 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5091 			error = EINVAL;
5092 			break;
5093 		}
5094 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
5095 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5096 		    M_TEMP, M_WAITOK);
5097 		error = copyin(io->pfrio_buffer, pfras, totlen);
5098 		if (error) {
5099 			free(pfras, M_TEMP);
5100 			break;
5101 		}
5102 		PF_RULES_WLOCK();
5103 		error = pfr_clr_astats(&io->pfrio_table, pfras,
5104 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
5105 		    PFR_FLAG_USERIOCTL);
5106 		PF_RULES_WUNLOCK();
5107 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
5108 			error = copyout(pfras, io->pfrio_buffer, totlen);
5109 		free(pfras, M_TEMP);
5110 		break;
5111 	}
5112 
5113 	case DIOCRTSTADDRS: {
5114 		struct pfioc_table *io = (struct pfioc_table *)addr;
5115 		struct pfr_addr *pfras;
5116 		size_t totlen;
5117 
5118 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5119 			error = ENODEV;
5120 			break;
5121 		}
5122 		if (io->pfrio_size < 0 ||
5123 		    io->pfrio_size > pf_ioctl_maxcount ||
5124 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5125 			error = EINVAL;
5126 			break;
5127 		}
5128 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
5129 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5130 		    M_TEMP, M_WAITOK);
5131 		error = copyin(io->pfrio_buffer, pfras, totlen);
5132 		if (error) {
5133 			free(pfras, M_TEMP);
5134 			break;
5135 		}
5136 		PF_RULES_RLOCK();
5137 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
5138 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
5139 		    PFR_FLAG_USERIOCTL);
5140 		PF_RULES_RUNLOCK();
5141 		if (error == 0)
5142 			error = copyout(pfras, io->pfrio_buffer, totlen);
5143 		free(pfras, M_TEMP);
5144 		break;
5145 	}
5146 
5147 	case DIOCRINADEFINE: {
5148 		struct pfioc_table *io = (struct pfioc_table *)addr;
5149 		struct pfr_addr *pfras;
5150 		size_t totlen;
5151 
5152 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5153 			error = ENODEV;
5154 			break;
5155 		}
5156 		if (io->pfrio_size < 0 ||
5157 		    io->pfrio_size > pf_ioctl_maxcount ||
5158 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5159 			error = EINVAL;
5160 			break;
5161 		}
5162 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
5163 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5164 		    M_TEMP, M_WAITOK);
5165 		error = copyin(io->pfrio_buffer, pfras, totlen);
5166 		if (error) {
5167 			free(pfras, M_TEMP);
5168 			break;
5169 		}
5170 		PF_RULES_WLOCK();
5171 		error = pfr_ina_define(&io->pfrio_table, pfras,
5172 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
5173 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5174 		PF_RULES_WUNLOCK();
5175 		free(pfras, M_TEMP);
5176 		break;
5177 	}
5178 
5179 	case DIOCOSFPADD: {
5180 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5181 		PF_RULES_WLOCK();
5182 		error = pf_osfp_add(io);
5183 		PF_RULES_WUNLOCK();
5184 		break;
5185 	}
5186 
5187 	case DIOCOSFPGET: {
5188 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5189 		PF_RULES_RLOCK();
5190 		error = pf_osfp_get(io);
5191 		PF_RULES_RUNLOCK();
5192 		break;
5193 	}
5194 
5195 	case DIOCXBEGIN: {
5196 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5197 		struct pfioc_trans_e	*ioes, *ioe;
5198 		size_t			 totlen;
5199 		int			 i;
5200 
5201 		if (io->esize != sizeof(*ioe)) {
5202 			error = ENODEV;
5203 			break;
5204 		}
5205 		if (io->size < 0 ||
5206 		    io->size > pf_ioctl_maxcount ||
5207 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5208 			error = EINVAL;
5209 			break;
5210 		}
5211 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5212 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5213 		    M_TEMP, M_WAITOK);
5214 		error = copyin(io->array, ioes, totlen);
5215 		if (error) {
5216 			free(ioes, M_TEMP);
5217 			break;
5218 		}
5219 		/* Ensure there's no more ethernet rules to clean up. */
5220 		NET_EPOCH_DRAIN_CALLBACKS();
5221 		PF_RULES_WLOCK();
5222 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5223 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5224 			switch (ioe->rs_num) {
5225 			case PF_RULESET_ETH:
5226 				if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) {
5227 					PF_RULES_WUNLOCK();
5228 					free(ioes, M_TEMP);
5229 					goto fail;
5230 				}
5231 				break;
5232 #ifdef ALTQ
5233 			case PF_RULESET_ALTQ:
5234 				if (ioe->anchor[0]) {
5235 					PF_RULES_WUNLOCK();
5236 					free(ioes, M_TEMP);
5237 					error = EINVAL;
5238 					goto fail;
5239 				}
5240 				if ((error = pf_begin_altq(&ioe->ticket))) {
5241 					PF_RULES_WUNLOCK();
5242 					free(ioes, M_TEMP);
5243 					goto fail;
5244 				}
5245 				break;
5246 #endif /* ALTQ */
5247 			case PF_RULESET_TABLE:
5248 			    {
5249 				struct pfr_table table;
5250 
5251 				bzero(&table, sizeof(table));
5252 				strlcpy(table.pfrt_anchor, ioe->anchor,
5253 				    sizeof(table.pfrt_anchor));
5254 				if ((error = pfr_ina_begin(&table,
5255 				    &ioe->ticket, NULL, 0))) {
5256 					PF_RULES_WUNLOCK();
5257 					free(ioes, M_TEMP);
5258 					goto fail;
5259 				}
5260 				break;
5261 			    }
5262 			default:
5263 				if ((error = pf_begin_rules(&ioe->ticket,
5264 				    ioe->rs_num, ioe->anchor))) {
5265 					PF_RULES_WUNLOCK();
5266 					free(ioes, M_TEMP);
5267 					goto fail;
5268 				}
5269 				break;
5270 			}
5271 		}
5272 		PF_RULES_WUNLOCK();
5273 		error = copyout(ioes, io->array, totlen);
5274 		free(ioes, M_TEMP);
5275 		break;
5276 	}
5277 
5278 	case DIOCXROLLBACK: {
5279 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5280 		struct pfioc_trans_e	*ioe, *ioes;
5281 		size_t			 totlen;
5282 		int			 i;
5283 
5284 		if (io->esize != sizeof(*ioe)) {
5285 			error = ENODEV;
5286 			break;
5287 		}
5288 		if (io->size < 0 ||
5289 		    io->size > pf_ioctl_maxcount ||
5290 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5291 			error = EINVAL;
5292 			break;
5293 		}
5294 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5295 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5296 		    M_TEMP, M_WAITOK);
5297 		error = copyin(io->array, ioes, totlen);
5298 		if (error) {
5299 			free(ioes, M_TEMP);
5300 			break;
5301 		}
5302 		PF_RULES_WLOCK();
5303 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5304 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5305 			switch (ioe->rs_num) {
5306 			case PF_RULESET_ETH:
5307 				if ((error = pf_rollback_eth(ioe->ticket,
5308 				    ioe->anchor))) {
5309 					PF_RULES_WUNLOCK();
5310 					free(ioes, M_TEMP);
5311 					goto fail; /* really bad */
5312 				}
5313 				break;
5314 #ifdef ALTQ
5315 			case PF_RULESET_ALTQ:
5316 				if (ioe->anchor[0]) {
5317 					PF_RULES_WUNLOCK();
5318 					free(ioes, M_TEMP);
5319 					error = EINVAL;
5320 					goto fail;
5321 				}
5322 				if ((error = pf_rollback_altq(ioe->ticket))) {
5323 					PF_RULES_WUNLOCK();
5324 					free(ioes, M_TEMP);
5325 					goto fail; /* really bad */
5326 				}
5327 				break;
5328 #endif /* ALTQ */
5329 			case PF_RULESET_TABLE:
5330 			    {
5331 				struct pfr_table table;
5332 
5333 				bzero(&table, sizeof(table));
5334 				strlcpy(table.pfrt_anchor, ioe->anchor,
5335 				    sizeof(table.pfrt_anchor));
5336 				if ((error = pfr_ina_rollback(&table,
5337 				    ioe->ticket, NULL, 0))) {
5338 					PF_RULES_WUNLOCK();
5339 					free(ioes, M_TEMP);
5340 					goto fail; /* really bad */
5341 				}
5342 				break;
5343 			    }
5344 			default:
5345 				if ((error = pf_rollback_rules(ioe->ticket,
5346 				    ioe->rs_num, ioe->anchor))) {
5347 					PF_RULES_WUNLOCK();
5348 					free(ioes, M_TEMP);
5349 					goto fail; /* really bad */
5350 				}
5351 				break;
5352 			}
5353 		}
5354 		PF_RULES_WUNLOCK();
5355 		free(ioes, M_TEMP);
5356 		break;
5357 	}
5358 
5359 	case DIOCXCOMMIT: {
5360 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5361 		struct pfioc_trans_e	*ioe, *ioes;
5362 		struct pf_kruleset	*rs;
5363 		struct pf_keth_ruleset	*ers;
5364 		size_t			 totlen;
5365 		int			 i;
5366 
5367 		if (io->esize != sizeof(*ioe)) {
5368 			error = ENODEV;
5369 			break;
5370 		}
5371 
5372 		if (io->size < 0 ||
5373 		    io->size > pf_ioctl_maxcount ||
5374 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5375 			error = EINVAL;
5376 			break;
5377 		}
5378 
5379 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5380 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5381 		    M_TEMP, M_WAITOK);
5382 		error = copyin(io->array, ioes, totlen);
5383 		if (error) {
5384 			free(ioes, M_TEMP);
5385 			break;
5386 		}
5387 		PF_RULES_WLOCK();
5388 		/* First makes sure everything will succeed. */
5389 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5390 			ioe->anchor[sizeof(ioe->anchor) - 1] = 0;
5391 			switch (ioe->rs_num) {
5392 			case PF_RULESET_ETH:
5393 				ers = pf_find_keth_ruleset(ioe->anchor);
5394 				if (ers == NULL || ioe->ticket == 0 ||
5395 				    ioe->ticket != ers->inactive.ticket) {
5396 					PF_RULES_WUNLOCK();
5397 					free(ioes, M_TEMP);
5398 					error = EINVAL;
5399 					goto fail;
5400 				}
5401 				break;
5402 #ifdef ALTQ
5403 			case PF_RULESET_ALTQ:
5404 				if (ioe->anchor[0]) {
5405 					PF_RULES_WUNLOCK();
5406 					free(ioes, M_TEMP);
5407 					error = EINVAL;
5408 					goto fail;
5409 				}
5410 				if (!V_altqs_inactive_open || ioe->ticket !=
5411 				    V_ticket_altqs_inactive) {
5412 					PF_RULES_WUNLOCK();
5413 					free(ioes, M_TEMP);
5414 					error = EBUSY;
5415 					goto fail;
5416 				}
5417 				break;
5418 #endif /* ALTQ */
5419 			case PF_RULESET_TABLE:
5420 				rs = pf_find_kruleset(ioe->anchor);
5421 				if (rs == NULL || !rs->topen || ioe->ticket !=
5422 				    rs->tticket) {
5423 					PF_RULES_WUNLOCK();
5424 					free(ioes, M_TEMP);
5425 					error = EBUSY;
5426 					goto fail;
5427 				}
5428 				break;
5429 			default:
5430 				if (ioe->rs_num < 0 || ioe->rs_num >=
5431 				    PF_RULESET_MAX) {
5432 					PF_RULES_WUNLOCK();
5433 					free(ioes, M_TEMP);
5434 					error = EINVAL;
5435 					goto fail;
5436 				}
5437 				rs = pf_find_kruleset(ioe->anchor);
5438 				if (rs == NULL ||
5439 				    !rs->rules[ioe->rs_num].inactive.open ||
5440 				    rs->rules[ioe->rs_num].inactive.ticket !=
5441 				    ioe->ticket) {
5442 					PF_RULES_WUNLOCK();
5443 					free(ioes, M_TEMP);
5444 					error = EBUSY;
5445 					goto fail;
5446 				}
5447 				break;
5448 			}
5449 		}
5450 		/* Now do the commit - no errors should happen here. */
5451 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5452 			switch (ioe->rs_num) {
5453 			case PF_RULESET_ETH:
5454 				if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) {
5455 					PF_RULES_WUNLOCK();
5456 					free(ioes, M_TEMP);
5457 					goto fail; /* really bad */
5458 				}
5459 				break;
5460 #ifdef ALTQ
5461 			case PF_RULESET_ALTQ:
5462 				if ((error = pf_commit_altq(ioe->ticket))) {
5463 					PF_RULES_WUNLOCK();
5464 					free(ioes, M_TEMP);
5465 					goto fail; /* really bad */
5466 				}
5467 				break;
5468 #endif /* ALTQ */
5469 			case PF_RULESET_TABLE:
5470 			    {
5471 				struct pfr_table table;
5472 
5473 				bzero(&table, sizeof(table));
5474 				(void)strlcpy(table.pfrt_anchor, ioe->anchor,
5475 				    sizeof(table.pfrt_anchor));
5476 				if ((error = pfr_ina_commit(&table,
5477 				    ioe->ticket, NULL, NULL, 0))) {
5478 					PF_RULES_WUNLOCK();
5479 					free(ioes, M_TEMP);
5480 					goto fail; /* really bad */
5481 				}
5482 				break;
5483 			    }
5484 			default:
5485 				if ((error = pf_commit_rules(ioe->ticket,
5486 				    ioe->rs_num, ioe->anchor))) {
5487 					PF_RULES_WUNLOCK();
5488 					free(ioes, M_TEMP);
5489 					goto fail; /* really bad */
5490 				}
5491 				break;
5492 			}
5493 		}
5494 		PF_RULES_WUNLOCK();
5495 
5496 		/* Only hook into EtherNet taffic if we've got rules for it. */
5497 		if (! TAILQ_EMPTY(V_pf_keth->active.rules))
5498 			hook_pf_eth();
5499 		else
5500 			dehook_pf_eth();
5501 
5502 		free(ioes, M_TEMP);
5503 		break;
5504 	}
5505 
5506 	case DIOCGETSRCNODES: {
5507 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
5508 		struct pf_srchash	*sh;
5509 		struct pf_ksrc_node	*n;
5510 		struct pf_src_node	*p, *pstore;
5511 		uint32_t		 i, nr = 0;
5512 
5513 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5514 				i++, sh++) {
5515 			PF_HASHROW_LOCK(sh);
5516 			LIST_FOREACH(n, &sh->nodes, entry)
5517 				nr++;
5518 			PF_HASHROW_UNLOCK(sh);
5519 		}
5520 
5521 		psn->psn_len = min(psn->psn_len,
5522 		    sizeof(struct pf_src_node) * nr);
5523 
5524 		if (psn->psn_len == 0) {
5525 			psn->psn_len = sizeof(struct pf_src_node) * nr;
5526 			break;
5527 		}
5528 
5529 		nr = 0;
5530 
5531 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
5532 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5533 		    i++, sh++) {
5534 		    PF_HASHROW_LOCK(sh);
5535 		    LIST_FOREACH(n, &sh->nodes, entry) {
5536 
5537 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
5538 				break;
5539 
5540 			pf_src_node_copy(n, p);
5541 
5542 			p++;
5543 			nr++;
5544 		    }
5545 		    PF_HASHROW_UNLOCK(sh);
5546 		}
5547 		error = copyout(pstore, psn->psn_src_nodes,
5548 		    sizeof(struct pf_src_node) * nr);
5549 		if (error) {
5550 			free(pstore, M_TEMP);
5551 			break;
5552 		}
5553 		psn->psn_len = sizeof(struct pf_src_node) * nr;
5554 		free(pstore, M_TEMP);
5555 		break;
5556 	}
5557 
5558 	case DIOCCLRSRCNODES: {
5559 		pf_clear_srcnodes(NULL);
5560 		pf_purge_expired_src_nodes();
5561 		break;
5562 	}
5563 
5564 	case DIOCKILLSRCNODES:
5565 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
5566 		break;
5567 
5568 #ifdef COMPAT_FREEBSD13
5569 	case DIOCKEEPCOUNTERS_FREEBSD13:
5570 #endif
5571 	case DIOCKEEPCOUNTERS:
5572 		error = pf_keepcounters((struct pfioc_nv *)addr);
5573 		break;
5574 
5575 	case DIOCGETSYNCOOKIES:
5576 		error = pf_get_syncookies((struct pfioc_nv *)addr);
5577 		break;
5578 
5579 	case DIOCSETSYNCOOKIES:
5580 		error = pf_set_syncookies((struct pfioc_nv *)addr);
5581 		break;
5582 
5583 	case DIOCSETHOSTID: {
5584 		u_int32_t	*hostid = (u_int32_t *)addr;
5585 
5586 		PF_RULES_WLOCK();
5587 		if (*hostid == 0)
5588 			V_pf_status.hostid = arc4random();
5589 		else
5590 			V_pf_status.hostid = *hostid;
5591 		PF_RULES_WUNLOCK();
5592 		break;
5593 	}
5594 
5595 	case DIOCOSFPFLUSH:
5596 		PF_RULES_WLOCK();
5597 		pf_osfp_flush();
5598 		PF_RULES_WUNLOCK();
5599 		break;
5600 
5601 	case DIOCIGETIFACES: {
5602 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5603 		struct pfi_kif *ifstore;
5604 		size_t bufsiz;
5605 
5606 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
5607 			error = ENODEV;
5608 			break;
5609 		}
5610 
5611 		if (io->pfiio_size < 0 ||
5612 		    io->pfiio_size > pf_ioctl_maxcount ||
5613 		    WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
5614 			error = EINVAL;
5615 			break;
5616 		}
5617 
5618 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5619 
5620 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
5621 		ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
5622 		    M_TEMP, M_WAITOK | M_ZERO);
5623 
5624 		PF_RULES_RLOCK();
5625 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
5626 		PF_RULES_RUNLOCK();
5627 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
5628 		free(ifstore, M_TEMP);
5629 		break;
5630 	}
5631 
5632 	case DIOCSETIFFLAG: {
5633 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5634 
5635 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5636 
5637 		PF_RULES_WLOCK();
5638 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
5639 		PF_RULES_WUNLOCK();
5640 		break;
5641 	}
5642 
5643 	case DIOCCLRIFFLAG: {
5644 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5645 
5646 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5647 
5648 		PF_RULES_WLOCK();
5649 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
5650 		PF_RULES_WUNLOCK();
5651 		break;
5652 	}
5653 
5654 	case DIOCSETREASS: {
5655 		u_int32_t	*reass = (u_int32_t *)addr;
5656 
5657 		V_pf_status.reass = *reass & (PF_REASS_ENABLED|PF_REASS_NODF);
5658 		/* Removal of DF flag without reassembly enabled is not a
5659 		 * valid combination. Disable reassembly in such case. */
5660 		if (!(V_pf_status.reass & PF_REASS_ENABLED))
5661 			V_pf_status.reass = 0;
5662 		break;
5663 	}
5664 
5665 	default:
5666 		error = ENODEV;
5667 		break;
5668 	}
5669 fail:
5670 	if (sx_xlocked(&V_pf_ioctl_lock))
5671 		sx_xunlock(&V_pf_ioctl_lock);
5672 	CURVNET_RESTORE();
5673 
5674 #undef ERROUT_IOCTL
5675 
5676 	return (error);
5677 }
5678 
5679 void
5680 pfsync_state_export(union pfsync_state_union *sp, struct pf_kstate *st, int msg_version)
5681 {
5682 	bzero(sp, sizeof(union pfsync_state_union));
5683 
5684 	/* copy from state key */
5685 	sp->pfs_1301.key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5686 	sp->pfs_1301.key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5687 	sp->pfs_1301.key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5688 	sp->pfs_1301.key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5689 	sp->pfs_1301.key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5690 	sp->pfs_1301.key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5691 	sp->pfs_1301.key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5692 	sp->pfs_1301.key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5693 	sp->pfs_1301.proto = st->key[PF_SK_WIRE]->proto;
5694 	sp->pfs_1301.af = st->key[PF_SK_WIRE]->af;
5695 
5696 	/* copy from state */
5697 	strlcpy(sp->pfs_1301.ifname, st->kif->pfik_name, sizeof(sp->pfs_1301.ifname));
5698 	bcopy(&st->rt_addr, &sp->pfs_1301.rt_addr, sizeof(sp->pfs_1301.rt_addr));
5699 	sp->pfs_1301.creation = htonl(time_uptime - st->creation);
5700 	sp->pfs_1301.expire = pf_state_expires(st);
5701 	if (sp->pfs_1301.expire <= time_uptime)
5702 		sp->pfs_1301.expire = htonl(0);
5703 	else
5704 		sp->pfs_1301.expire = htonl(sp->pfs_1301.expire - time_uptime);
5705 
5706 	sp->pfs_1301.direction = st->direction;
5707 	sp->pfs_1301.log = st->act.log;
5708 	sp->pfs_1301.timeout = st->timeout;
5709 
5710 	switch (msg_version) {
5711 		case PFSYNC_MSG_VERSION_1301:
5712 			sp->pfs_1301.state_flags = st->state_flags;
5713 			break;
5714 		case PFSYNC_MSG_VERSION_1400:
5715 			sp->pfs_1400.state_flags = htons(st->state_flags);
5716 			sp->pfs_1400.qid = htons(st->act.qid);
5717 			sp->pfs_1400.pqid = htons(st->act.pqid);
5718 			sp->pfs_1400.dnpipe = htons(st->act.dnpipe);
5719 			sp->pfs_1400.dnrpipe = htons(st->act.dnrpipe);
5720 			sp->pfs_1400.rtableid = htonl(st->act.rtableid);
5721 			sp->pfs_1400.min_ttl = st->act.min_ttl;
5722 			sp->pfs_1400.set_tos = st->act.set_tos;
5723 			sp->pfs_1400.max_mss = htons(st->act.max_mss);
5724 			sp->pfs_1400.set_prio[0] = st->act.set_prio[0];
5725 			sp->pfs_1400.set_prio[1] = st->act.set_prio[1];
5726 			sp->pfs_1400.rt = st->rt;
5727 			if (st->rt_kif)
5728 				strlcpy(sp->pfs_1400.rt_ifname,
5729 				    st->rt_kif->pfik_name,
5730 				    sizeof(sp->pfs_1400.rt_ifname));
5731 			break;
5732 		default:
5733 			panic("%s: Unsupported pfsync_msg_version %d",
5734 			    __func__, msg_version);
5735 	}
5736 
5737 	if (st->src_node)
5738 		sp->pfs_1301.sync_flags |= PFSYNC_FLAG_SRCNODE;
5739 	if (st->nat_src_node)
5740 		sp->pfs_1301.sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5741 
5742 	sp->pfs_1301.id = st->id;
5743 	sp->pfs_1301.creatorid = st->creatorid;
5744 	pf_state_peer_hton(&st->src, &sp->pfs_1301.src);
5745 	pf_state_peer_hton(&st->dst, &sp->pfs_1301.dst);
5746 
5747 	if (st->rule.ptr == NULL)
5748 		sp->pfs_1301.rule = htonl(-1);
5749 	else
5750 		sp->pfs_1301.rule = htonl(st->rule.ptr->nr);
5751 	if (st->anchor.ptr == NULL)
5752 		sp->pfs_1301.anchor = htonl(-1);
5753 	else
5754 		sp->pfs_1301.anchor = htonl(st->anchor.ptr->nr);
5755 	if (st->nat_rule.ptr == NULL)
5756 		sp->pfs_1301.nat_rule = htonl(-1);
5757 	else
5758 		sp->pfs_1301.nat_rule = htonl(st->nat_rule.ptr->nr);
5759 
5760 	pf_state_counter_hton(st->packets[0], sp->pfs_1301.packets[0]);
5761 	pf_state_counter_hton(st->packets[1], sp->pfs_1301.packets[1]);
5762 	pf_state_counter_hton(st->bytes[0], sp->pfs_1301.bytes[0]);
5763 	pf_state_counter_hton(st->bytes[1], sp->pfs_1301.bytes[1]);
5764 }
5765 
5766 void
5767 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st)
5768 {
5769 	bzero(sp, sizeof(*sp));
5770 
5771 	sp->version = PF_STATE_VERSION;
5772 
5773 	/* copy from state key */
5774 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5775 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5776 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5777 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5778 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5779 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5780 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5781 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5782 	sp->proto = st->key[PF_SK_WIRE]->proto;
5783 	sp->af = st->key[PF_SK_WIRE]->af;
5784 
5785 	/* copy from state */
5786 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
5787 	strlcpy(sp->orig_ifname, st->orig_kif->pfik_name,
5788 	    sizeof(sp->orig_ifname));
5789 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
5790 	sp->creation = htonl(time_uptime - st->creation);
5791 	sp->expire = pf_state_expires(st);
5792 	if (sp->expire <= time_uptime)
5793 		sp->expire = htonl(0);
5794 	else
5795 		sp->expire = htonl(sp->expire - time_uptime);
5796 
5797 	sp->direction = st->direction;
5798 	sp->log = st->act.log;
5799 	sp->timeout = st->timeout;
5800 	/* 8 bits for the old libpfctl, 16 bits for the new libpfctl */
5801 	sp->state_flags_compat = st->state_flags;
5802 	sp->state_flags = htons(st->state_flags);
5803 	if (st->src_node)
5804 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
5805 	if (st->nat_src_node)
5806 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5807 
5808 	sp->id = st->id;
5809 	sp->creatorid = st->creatorid;
5810 	pf_state_peer_hton(&st->src, &sp->src);
5811 	pf_state_peer_hton(&st->dst, &sp->dst);
5812 
5813 	if (st->rule.ptr == NULL)
5814 		sp->rule = htonl(-1);
5815 	else
5816 		sp->rule = htonl(st->rule.ptr->nr);
5817 	if (st->anchor.ptr == NULL)
5818 		sp->anchor = htonl(-1);
5819 	else
5820 		sp->anchor = htonl(st->anchor.ptr->nr);
5821 	if (st->nat_rule.ptr == NULL)
5822 		sp->nat_rule = htonl(-1);
5823 	else
5824 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
5825 
5826 	sp->packets[0] = st->packets[0];
5827 	sp->packets[1] = st->packets[1];
5828 	sp->bytes[0] = st->bytes[0];
5829 	sp->bytes[1] = st->bytes[1];
5830 
5831 	sp->qid = htons(st->act.qid);
5832 	sp->pqid = htons(st->act.pqid);
5833 	sp->dnpipe = htons(st->act.dnpipe);
5834 	sp->dnrpipe = htons(st->act.dnrpipe);
5835 	sp->rtableid = htonl(st->act.rtableid);
5836 	sp->min_ttl = st->act.min_ttl;
5837 	sp->set_tos = st->act.set_tos;
5838 	sp->max_mss = htons(st->act.max_mss);
5839 	sp->rt = st->rt;
5840 	if (st->rt_kif)
5841 		strlcpy(sp->rt_ifname, st->rt_kif->pfik_name,
5842 		    sizeof(sp->rt_ifname));
5843 	sp->set_prio[0] = st->act.set_prio[0];
5844 	sp->set_prio[1] = st->act.set_prio[1];
5845 
5846 }
5847 
5848 static void
5849 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
5850 {
5851 	struct pfr_ktable *kt;
5852 
5853 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
5854 
5855 	kt = aw->p.tbl;
5856 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
5857 		kt = kt->pfrkt_root;
5858 	aw->p.tbl = NULL;
5859 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
5860 		kt->pfrkt_cnt : -1;
5861 }
5862 
5863 static int
5864 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters,
5865     size_t number, char **names)
5866 {
5867 	nvlist_t        *nvc;
5868 
5869 	nvc = nvlist_create(0);
5870 	if (nvc == NULL)
5871 		return (ENOMEM);
5872 
5873 	for (int i = 0; i < number; i++) {
5874 		nvlist_append_number_array(nvc, "counters",
5875 		    counter_u64_fetch(counters[i]));
5876 		nvlist_append_string_array(nvc, "names",
5877 		    names[i]);
5878 		nvlist_append_number_array(nvc, "ids",
5879 		    i);
5880 	}
5881 	nvlist_add_nvlist(nvl, name, nvc);
5882 	nvlist_destroy(nvc);
5883 
5884 	return (0);
5885 }
5886 
5887 static int
5888 pf_getstatus(struct pfioc_nv *nv)
5889 {
5890 	nvlist_t        *nvl = NULL, *nvc = NULL;
5891 	void            *nvlpacked = NULL;
5892 	int              error;
5893 	struct pf_status s;
5894 	char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES;
5895 	char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES;
5896 	char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES;
5897 	PF_RULES_RLOCK_TRACKER;
5898 
5899 #define ERROUT(x)      ERROUT_FUNCTION(errout, x)
5900 
5901 	PF_RULES_RLOCK();
5902 
5903 	nvl = nvlist_create(0);
5904 	if (nvl == NULL)
5905 		ERROUT(ENOMEM);
5906 
5907 	nvlist_add_bool(nvl, "running", V_pf_status.running);
5908 	nvlist_add_number(nvl, "since", V_pf_status.since);
5909 	nvlist_add_number(nvl, "debug", V_pf_status.debug);
5910 	nvlist_add_number(nvl, "hostid", V_pf_status.hostid);
5911 	nvlist_add_number(nvl, "states", V_pf_status.states);
5912 	nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes);
5913 	nvlist_add_number(nvl, "reass", V_pf_status.reass);
5914 	nvlist_add_bool(nvl, "syncookies_active",
5915 	    V_pf_status.syncookies_active);
5916 
5917 	/* counters */
5918 	error = pf_add_status_counters(nvl, "counters", V_pf_status.counters,
5919 	    PFRES_MAX, pf_reasons);
5920 	if (error != 0)
5921 		ERROUT(error);
5922 
5923 	/* lcounters */
5924 	error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters,
5925 	    KLCNT_MAX, pf_lcounter);
5926 	if (error != 0)
5927 		ERROUT(error);
5928 
5929 	/* fcounters */
5930 	nvc = nvlist_create(0);
5931 	if (nvc == NULL)
5932 		ERROUT(ENOMEM);
5933 
5934 	for (int i = 0; i < FCNT_MAX; i++) {
5935 		nvlist_append_number_array(nvc, "counters",
5936 		    pf_counter_u64_fetch(&V_pf_status.fcounters[i]));
5937 		nvlist_append_string_array(nvc, "names",
5938 		    pf_fcounter[i]);
5939 		nvlist_append_number_array(nvc, "ids",
5940 		    i);
5941 	}
5942 	nvlist_add_nvlist(nvl, "fcounters", nvc);
5943 	nvlist_destroy(nvc);
5944 	nvc = NULL;
5945 
5946 	/* scounters */
5947 	error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters,
5948 	    SCNT_MAX, pf_fcounter);
5949 	if (error != 0)
5950 		ERROUT(error);
5951 
5952 	nvlist_add_string(nvl, "ifname", V_pf_status.ifname);
5953 	nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum,
5954 	    PF_MD5_DIGEST_LENGTH);
5955 
5956 	pfi_update_status(V_pf_status.ifname, &s);
5957 
5958 	/* pcounters / bcounters */
5959 	for (int i = 0; i < 2; i++) {
5960 		for (int j = 0; j < 2; j++) {
5961 			for (int k = 0; k < 2; k++) {
5962 				nvlist_append_number_array(nvl, "pcounters",
5963 				    s.pcounters[i][j][k]);
5964 			}
5965 			nvlist_append_number_array(nvl, "bcounters",
5966 			    s.bcounters[i][j]);
5967 		}
5968 	}
5969 
5970 	nvlpacked = nvlist_pack(nvl, &nv->len);
5971 	if (nvlpacked == NULL)
5972 		ERROUT(ENOMEM);
5973 
5974 	if (nv->size == 0)
5975 		ERROUT(0);
5976 	else if (nv->size < nv->len)
5977 		ERROUT(ENOSPC);
5978 
5979 	PF_RULES_RUNLOCK();
5980 	error = copyout(nvlpacked, nv->data, nv->len);
5981 	goto done;
5982 
5983 #undef ERROUT
5984 errout:
5985 	PF_RULES_RUNLOCK();
5986 done:
5987 	free(nvlpacked, M_NVLIST);
5988 	nvlist_destroy(nvc);
5989 	nvlist_destroy(nvl);
5990 
5991 	return (error);
5992 }
5993 
5994 /*
5995  * XXX - Check for version mismatch!!!
5996  */
5997 static void
5998 pf_clear_all_states(void)
5999 {
6000 	struct pf_kstate	*s;
6001 	u_int i;
6002 
6003 	for (i = 0; i <= pf_hashmask; i++) {
6004 		struct pf_idhash *ih = &V_pf_idhash[i];
6005 relock:
6006 		PF_HASHROW_LOCK(ih);
6007 		LIST_FOREACH(s, &ih->states, entry) {
6008 			s->timeout = PFTM_PURGE;
6009 			/* Don't send out individual delete messages. */
6010 			s->state_flags |= PFSTATE_NOSYNC;
6011 			pf_unlink_state(s);
6012 			goto relock;
6013 		}
6014 		PF_HASHROW_UNLOCK(ih);
6015 	}
6016 }
6017 
6018 static int
6019 pf_clear_tables(void)
6020 {
6021 	struct pfioc_table io;
6022 	int error;
6023 
6024 	bzero(&io, sizeof(io));
6025 
6026 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
6027 	    io.pfrio_flags);
6028 
6029 	return (error);
6030 }
6031 
6032 static void
6033 pf_clear_srcnodes(struct pf_ksrc_node *n)
6034 {
6035 	struct pf_kstate *s;
6036 	int i;
6037 
6038 	for (i = 0; i <= pf_hashmask; i++) {
6039 		struct pf_idhash *ih = &V_pf_idhash[i];
6040 
6041 		PF_HASHROW_LOCK(ih);
6042 		LIST_FOREACH(s, &ih->states, entry) {
6043 			if (n == NULL || n == s->src_node)
6044 				s->src_node = NULL;
6045 			if (n == NULL || n == s->nat_src_node)
6046 				s->nat_src_node = NULL;
6047 		}
6048 		PF_HASHROW_UNLOCK(ih);
6049 	}
6050 
6051 	if (n == NULL) {
6052 		struct pf_srchash *sh;
6053 
6054 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
6055 		    i++, sh++) {
6056 			PF_HASHROW_LOCK(sh);
6057 			LIST_FOREACH(n, &sh->nodes, entry) {
6058 				n->expire = 1;
6059 				n->states = 0;
6060 			}
6061 			PF_HASHROW_UNLOCK(sh);
6062 		}
6063 	} else {
6064 		/* XXX: hash slot should already be locked here. */
6065 		n->expire = 1;
6066 		n->states = 0;
6067 	}
6068 }
6069 
6070 static void
6071 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
6072 {
6073 	struct pf_ksrc_node_list	 kill;
6074 
6075 	LIST_INIT(&kill);
6076 	for (int i = 0; i <= pf_srchashmask; i++) {
6077 		struct pf_srchash *sh = &V_pf_srchash[i];
6078 		struct pf_ksrc_node *sn, *tmp;
6079 
6080 		PF_HASHROW_LOCK(sh);
6081 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
6082 			if (PF_MATCHA(psnk->psnk_src.neg,
6083 			      &psnk->psnk_src.addr.v.a.addr,
6084 			      &psnk->psnk_src.addr.v.a.mask,
6085 			      &sn->addr, sn->af) &&
6086 			    PF_MATCHA(psnk->psnk_dst.neg,
6087 			      &psnk->psnk_dst.addr.v.a.addr,
6088 			      &psnk->psnk_dst.addr.v.a.mask,
6089 			      &sn->raddr, sn->af)) {
6090 				pf_unlink_src_node(sn);
6091 				LIST_INSERT_HEAD(&kill, sn, entry);
6092 				sn->expire = 1;
6093 			}
6094 		PF_HASHROW_UNLOCK(sh);
6095 	}
6096 
6097 	for (int i = 0; i <= pf_hashmask; i++) {
6098 		struct pf_idhash *ih = &V_pf_idhash[i];
6099 		struct pf_kstate *s;
6100 
6101 		PF_HASHROW_LOCK(ih);
6102 		LIST_FOREACH(s, &ih->states, entry) {
6103 			if (s->src_node && s->src_node->expire == 1)
6104 				s->src_node = NULL;
6105 			if (s->nat_src_node && s->nat_src_node->expire == 1)
6106 				s->nat_src_node = NULL;
6107 		}
6108 		PF_HASHROW_UNLOCK(ih);
6109 	}
6110 
6111 	psnk->psnk_killed = pf_free_src_nodes(&kill);
6112 }
6113 
6114 static int
6115 pf_keepcounters(struct pfioc_nv *nv)
6116 {
6117 	nvlist_t	*nvl = NULL;
6118 	void		*nvlpacked = NULL;
6119 	int		 error = 0;
6120 
6121 #define	ERROUT(x)	ERROUT_FUNCTION(on_error, x)
6122 
6123 	if (nv->len > pf_ioctl_maxcount)
6124 		ERROUT(ENOMEM);
6125 
6126 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6127 	if (nvlpacked == NULL)
6128 		ERROUT(ENOMEM);
6129 
6130 	error = copyin(nv->data, nvlpacked, nv->len);
6131 	if (error)
6132 		ERROUT(error);
6133 
6134 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6135 	if (nvl == NULL)
6136 		ERROUT(EBADMSG);
6137 
6138 	if (! nvlist_exists_bool(nvl, "keep_counters"))
6139 		ERROUT(EBADMSG);
6140 
6141 	V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters");
6142 
6143 on_error:
6144 	nvlist_destroy(nvl);
6145 	free(nvlpacked, M_NVLIST);
6146 	return (error);
6147 }
6148 
6149 static unsigned int
6150 pf_clear_states(const struct pf_kstate_kill *kill)
6151 {
6152 	struct pf_state_key_cmp	 match_key;
6153 	struct pf_kstate	*s;
6154 	struct pfi_kkif	*kif;
6155 	int		 idx;
6156 	unsigned int	 killed = 0, dir;
6157 
6158 	for (unsigned int i = 0; i <= pf_hashmask; i++) {
6159 		struct pf_idhash *ih = &V_pf_idhash[i];
6160 
6161 relock_DIOCCLRSTATES:
6162 		PF_HASHROW_LOCK(ih);
6163 		LIST_FOREACH(s, &ih->states, entry) {
6164 			/* For floating states look at the original kif. */
6165 			kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
6166 
6167 			if (kill->psk_ifname[0] &&
6168 			    strcmp(kill->psk_ifname,
6169 			    kif->pfik_name))
6170 				continue;
6171 
6172 			if (kill->psk_kill_match) {
6173 				bzero(&match_key, sizeof(match_key));
6174 
6175 				if (s->direction == PF_OUT) {
6176 					dir = PF_IN;
6177 					idx = PF_SK_STACK;
6178 				} else {
6179 					dir = PF_OUT;
6180 					idx = PF_SK_WIRE;
6181 				}
6182 
6183 				match_key.af = s->key[idx]->af;
6184 				match_key.proto = s->key[idx]->proto;
6185 				PF_ACPY(&match_key.addr[0],
6186 				    &s->key[idx]->addr[1], match_key.af);
6187 				match_key.port[0] = s->key[idx]->port[1];
6188 				PF_ACPY(&match_key.addr[1],
6189 				    &s->key[idx]->addr[0], match_key.af);
6190 				match_key.port[1] = s->key[idx]->port[0];
6191 			}
6192 
6193 			/*
6194 			 * Don't send out individual
6195 			 * delete messages.
6196 			 */
6197 			s->state_flags |= PFSTATE_NOSYNC;
6198 			pf_unlink_state(s);
6199 			killed++;
6200 
6201 			if (kill->psk_kill_match)
6202 				killed += pf_kill_matching_state(&match_key,
6203 				    dir);
6204 
6205 			goto relock_DIOCCLRSTATES;
6206 		}
6207 		PF_HASHROW_UNLOCK(ih);
6208 	}
6209 
6210 	if (V_pfsync_clear_states_ptr != NULL)
6211 		V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname);
6212 
6213 	return (killed);
6214 }
6215 
6216 static void
6217 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed)
6218 {
6219 	struct pf_kstate	*s;
6220 
6221 	if (kill->psk_pfcmp.id) {
6222 		if (kill->psk_pfcmp.creatorid == 0)
6223 			kill->psk_pfcmp.creatorid = V_pf_status.hostid;
6224 		if ((s = pf_find_state_byid(kill->psk_pfcmp.id,
6225 		    kill->psk_pfcmp.creatorid))) {
6226 			pf_unlink_state(s);
6227 			*killed = 1;
6228 		}
6229 		return;
6230 	}
6231 
6232 	for (unsigned int i = 0; i <= pf_hashmask; i++)
6233 		*killed += pf_killstates_row(kill, &V_pf_idhash[i]);
6234 
6235 	return;
6236 }
6237 
6238 static int
6239 pf_killstates_nv(struct pfioc_nv *nv)
6240 {
6241 	struct pf_kstate_kill	 kill;
6242 	nvlist_t		*nvl = NULL;
6243 	void			*nvlpacked = NULL;
6244 	int			 error = 0;
6245 	unsigned int		 killed = 0;
6246 
6247 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
6248 
6249 	if (nv->len > pf_ioctl_maxcount)
6250 		ERROUT(ENOMEM);
6251 
6252 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6253 	if (nvlpacked == NULL)
6254 		ERROUT(ENOMEM);
6255 
6256 	error = copyin(nv->data, nvlpacked, nv->len);
6257 	if (error)
6258 		ERROUT(error);
6259 
6260 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6261 	if (nvl == NULL)
6262 		ERROUT(EBADMSG);
6263 
6264 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6265 	if (error)
6266 		ERROUT(error);
6267 
6268 	pf_killstates(&kill, &killed);
6269 
6270 	free(nvlpacked, M_NVLIST);
6271 	nvlpacked = NULL;
6272 	nvlist_destroy(nvl);
6273 	nvl = nvlist_create(0);
6274 	if (nvl == NULL)
6275 		ERROUT(ENOMEM);
6276 
6277 	nvlist_add_number(nvl, "killed", killed);
6278 
6279 	nvlpacked = nvlist_pack(nvl, &nv->len);
6280 	if (nvlpacked == NULL)
6281 		ERROUT(ENOMEM);
6282 
6283 	if (nv->size == 0)
6284 		ERROUT(0);
6285 	else if (nv->size < nv->len)
6286 		ERROUT(ENOSPC);
6287 
6288 	error = copyout(nvlpacked, nv->data, nv->len);
6289 
6290 on_error:
6291 	nvlist_destroy(nvl);
6292 	free(nvlpacked, M_NVLIST);
6293 	return (error);
6294 }
6295 
6296 static int
6297 pf_clearstates_nv(struct pfioc_nv *nv)
6298 {
6299 	struct pf_kstate_kill	 kill;
6300 	nvlist_t		*nvl = NULL;
6301 	void			*nvlpacked = NULL;
6302 	int			 error = 0;
6303 	unsigned int		 killed;
6304 
6305 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
6306 
6307 	if (nv->len > pf_ioctl_maxcount)
6308 		ERROUT(ENOMEM);
6309 
6310 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6311 	if (nvlpacked == NULL)
6312 		ERROUT(ENOMEM);
6313 
6314 	error = copyin(nv->data, nvlpacked, nv->len);
6315 	if (error)
6316 		ERROUT(error);
6317 
6318 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6319 	if (nvl == NULL)
6320 		ERROUT(EBADMSG);
6321 
6322 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6323 	if (error)
6324 		ERROUT(error);
6325 
6326 	killed = pf_clear_states(&kill);
6327 
6328 	free(nvlpacked, M_NVLIST);
6329 	nvlpacked = NULL;
6330 	nvlist_destroy(nvl);
6331 	nvl = nvlist_create(0);
6332 	if (nvl == NULL)
6333 		ERROUT(ENOMEM);
6334 
6335 	nvlist_add_number(nvl, "killed", killed);
6336 
6337 	nvlpacked = nvlist_pack(nvl, &nv->len);
6338 	if (nvlpacked == NULL)
6339 		ERROUT(ENOMEM);
6340 
6341 	if (nv->size == 0)
6342 		ERROUT(0);
6343 	else if (nv->size < nv->len)
6344 		ERROUT(ENOSPC);
6345 
6346 	error = copyout(nvlpacked, nv->data, nv->len);
6347 
6348 #undef ERROUT
6349 on_error:
6350 	nvlist_destroy(nvl);
6351 	free(nvlpacked, M_NVLIST);
6352 	return (error);
6353 }
6354 
6355 static int
6356 pf_getstate(struct pfioc_nv *nv)
6357 {
6358 	nvlist_t		*nvl = NULL, *nvls;
6359 	void			*nvlpacked = NULL;
6360 	struct pf_kstate	*s = NULL;
6361 	int			 error = 0;
6362 	uint64_t		 id, creatorid;
6363 
6364 #define ERROUT(x)	ERROUT_FUNCTION(errout, x)
6365 
6366 	if (nv->len > pf_ioctl_maxcount)
6367 		ERROUT(ENOMEM);
6368 
6369 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6370 	if (nvlpacked == NULL)
6371 		ERROUT(ENOMEM);
6372 
6373 	error = copyin(nv->data, nvlpacked, nv->len);
6374 	if (error)
6375 		ERROUT(error);
6376 
6377 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6378 	if (nvl == NULL)
6379 		ERROUT(EBADMSG);
6380 
6381 	PFNV_CHK(pf_nvuint64(nvl, "id", &id));
6382 	PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid));
6383 
6384 	s = pf_find_state_byid(id, creatorid);
6385 	if (s == NULL)
6386 		ERROUT(ENOENT);
6387 
6388 	free(nvlpacked, M_NVLIST);
6389 	nvlpacked = NULL;
6390 	nvlist_destroy(nvl);
6391 	nvl = nvlist_create(0);
6392 	if (nvl == NULL)
6393 		ERROUT(ENOMEM);
6394 
6395 	nvls = pf_state_to_nvstate(s);
6396 	if (nvls == NULL)
6397 		ERROUT(ENOMEM);
6398 
6399 	nvlist_add_nvlist(nvl, "state", nvls);
6400 	nvlist_destroy(nvls);
6401 
6402 	nvlpacked = nvlist_pack(nvl, &nv->len);
6403 	if (nvlpacked == NULL)
6404 		ERROUT(ENOMEM);
6405 
6406 	if (nv->size == 0)
6407 		ERROUT(0);
6408 	else if (nv->size < nv->len)
6409 		ERROUT(ENOSPC);
6410 
6411 	error = copyout(nvlpacked, nv->data, nv->len);
6412 
6413 #undef ERROUT
6414 errout:
6415 	if (s != NULL)
6416 		PF_STATE_UNLOCK(s);
6417 	free(nvlpacked, M_NVLIST);
6418 	nvlist_destroy(nvl);
6419 	return (error);
6420 }
6421 
6422 /*
6423  * XXX - Check for version mismatch!!!
6424  */
6425 
6426 /*
6427  * Duplicate pfctl -Fa operation to get rid of as much as we can.
6428  */
6429 static int
6430 shutdown_pf(void)
6431 {
6432 	int error = 0;
6433 	u_int32_t t[5];
6434 	char nn = '\0';
6435 
6436 	do {
6437 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
6438 		    != 0) {
6439 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
6440 			break;
6441 		}
6442 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
6443 		    != 0) {
6444 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
6445 			break;		/* XXX: rollback? */
6446 		}
6447 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
6448 		    != 0) {
6449 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
6450 			break;		/* XXX: rollback? */
6451 		}
6452 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
6453 		    != 0) {
6454 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
6455 			break;		/* XXX: rollback? */
6456 		}
6457 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
6458 		    != 0) {
6459 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
6460 			break;		/* XXX: rollback? */
6461 		}
6462 
6463 		/* XXX: these should always succeed here */
6464 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
6465 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
6466 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
6467 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
6468 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
6469 
6470 		if ((error = pf_clear_tables()) != 0)
6471 			break;
6472 
6473 		if ((error = pf_begin_eth(&t[0], &nn)) != 0) {
6474 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth\n"));
6475 			break;
6476 		}
6477 		pf_commit_eth(t[0], &nn);
6478 
6479 #ifdef ALTQ
6480 		if ((error = pf_begin_altq(&t[0])) != 0) {
6481 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
6482 			break;
6483 		}
6484 		pf_commit_altq(t[0]);
6485 #endif
6486 
6487 		pf_clear_all_states();
6488 
6489 		pf_clear_srcnodes(NULL);
6490 
6491 		/* status does not use malloced mem so no need to cleanup */
6492 		/* fingerprints and interfaces have their own cleanup code */
6493 	} while(0);
6494 
6495 	return (error);
6496 }
6497 
6498 static pfil_return_t
6499 pf_check_return(int chk, struct mbuf **m)
6500 {
6501 
6502 	switch (chk) {
6503 	case PF_PASS:
6504 		if (*m == NULL)
6505 			return (PFIL_CONSUMED);
6506 		else
6507 			return (PFIL_PASS);
6508 		break;
6509 	default:
6510 		if (*m != NULL) {
6511 			m_freem(*m);
6512 			*m = NULL;
6513 		}
6514 		return (PFIL_DROPPED);
6515 	}
6516 }
6517 
6518 static pfil_return_t
6519 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6520     void *ruleset __unused, struct inpcb *inp)
6521 {
6522 	int chk;
6523 
6524 	chk = pf_test_eth(PF_IN, flags, ifp, m, inp);
6525 
6526 	return (pf_check_return(chk, m));
6527 }
6528 
6529 static pfil_return_t
6530 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6531     void *ruleset __unused, struct inpcb *inp)
6532 {
6533 	int chk;
6534 
6535 	chk = pf_test_eth(PF_OUT, flags, ifp, m, inp);
6536 
6537 	return (pf_check_return(chk, m));
6538 }
6539 
6540 #ifdef INET
6541 static pfil_return_t
6542 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6543     void *ruleset __unused, struct inpcb *inp)
6544 {
6545 	int chk;
6546 
6547 	chk = pf_test(PF_IN, flags, ifp, m, inp, NULL);
6548 
6549 	return (pf_check_return(chk, m));
6550 }
6551 
6552 static pfil_return_t
6553 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6554     void *ruleset __unused,  struct inpcb *inp)
6555 {
6556 	int chk;
6557 
6558 	chk = pf_test(PF_OUT, flags, ifp, m, inp, NULL);
6559 
6560 	return (pf_check_return(chk, m));
6561 }
6562 #endif
6563 
6564 #ifdef INET6
6565 static pfil_return_t
6566 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
6567     void *ruleset __unused,  struct inpcb *inp)
6568 {
6569 	int chk;
6570 
6571 	/*
6572 	 * In case of loopback traffic IPv6 uses the real interface in
6573 	 * order to support scoped addresses. In order to support stateful
6574 	 * filtering we have change this to lo0 as it is the case in IPv4.
6575 	 */
6576 	CURVNET_SET(ifp->if_vnet);
6577 	chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp,
6578 	    m, inp, NULL);
6579 	CURVNET_RESTORE();
6580 
6581 	return (pf_check_return(chk, m));
6582 }
6583 
6584 static pfil_return_t
6585 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
6586     void *ruleset __unused,  struct inpcb *inp)
6587 {
6588 	int chk;
6589 
6590 	CURVNET_SET(ifp->if_vnet);
6591 	chk = pf_test6(PF_OUT, flags, ifp, m, inp, NULL);
6592 	CURVNET_RESTORE();
6593 
6594 	return (pf_check_return(chk, m));
6595 }
6596 #endif /* INET6 */
6597 
6598 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook);
6599 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook);
6600 #define	V_pf_eth_in_hook	VNET(pf_eth_in_hook)
6601 #define	V_pf_eth_out_hook	VNET(pf_eth_out_hook)
6602 
6603 #ifdef INET
6604 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
6605 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
6606 #define	V_pf_ip4_in_hook	VNET(pf_ip4_in_hook)
6607 #define	V_pf_ip4_out_hook	VNET(pf_ip4_out_hook)
6608 #endif
6609 #ifdef INET6
6610 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
6611 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
6612 #define	V_pf_ip6_in_hook	VNET(pf_ip6_in_hook)
6613 #define	V_pf_ip6_out_hook	VNET(pf_ip6_out_hook)
6614 #endif
6615 
6616 static void
6617 hook_pf_eth(void)
6618 {
6619 	struct pfil_hook_args pha = {
6620 		.pa_version = PFIL_VERSION,
6621 		.pa_modname = "pf",
6622 		.pa_type = PFIL_TYPE_ETHERNET,
6623 	};
6624 	struct pfil_link_args pla = {
6625 		.pa_version = PFIL_VERSION,
6626 	};
6627 	int ret __diagused;
6628 
6629 	if (atomic_load_bool(&V_pf_pfil_eth_hooked))
6630 		return;
6631 
6632 	pha.pa_mbuf_chk = pf_eth_check_in;
6633 	pha.pa_flags = PFIL_IN;
6634 	pha.pa_rulname = "eth-in";
6635 	V_pf_eth_in_hook = pfil_add_hook(&pha);
6636 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6637 	pla.pa_head = V_link_pfil_head;
6638 	pla.pa_hook = V_pf_eth_in_hook;
6639 	ret = pfil_link(&pla);
6640 	MPASS(ret == 0);
6641 	pha.pa_mbuf_chk = pf_eth_check_out;
6642 	pha.pa_flags = PFIL_OUT;
6643 	pha.pa_rulname = "eth-out";
6644 	V_pf_eth_out_hook = pfil_add_hook(&pha);
6645 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6646 	pla.pa_head = V_link_pfil_head;
6647 	pla.pa_hook = V_pf_eth_out_hook;
6648 	ret = pfil_link(&pla);
6649 	MPASS(ret == 0);
6650 
6651 	atomic_store_bool(&V_pf_pfil_eth_hooked, true);
6652 }
6653 
6654 static void
6655 hook_pf(void)
6656 {
6657 	struct pfil_hook_args pha = {
6658 		.pa_version = PFIL_VERSION,
6659 		.pa_modname = "pf",
6660 	};
6661 	struct pfil_link_args pla = {
6662 		.pa_version = PFIL_VERSION,
6663 	};
6664 	int ret __diagused;
6665 
6666 	if (atomic_load_bool(&V_pf_pfil_hooked))
6667 		return;
6668 
6669 #ifdef INET
6670 	pha.pa_type = PFIL_TYPE_IP4;
6671 	pha.pa_mbuf_chk = pf_check_in;
6672 	pha.pa_flags = PFIL_IN;
6673 	pha.pa_rulname = "default-in";
6674 	V_pf_ip4_in_hook = pfil_add_hook(&pha);
6675 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6676 	pla.pa_head = V_inet_pfil_head;
6677 	pla.pa_hook = V_pf_ip4_in_hook;
6678 	ret = pfil_link(&pla);
6679 	MPASS(ret == 0);
6680 	pha.pa_mbuf_chk = pf_check_out;
6681 	pha.pa_flags = PFIL_OUT;
6682 	pha.pa_rulname = "default-out";
6683 	V_pf_ip4_out_hook = pfil_add_hook(&pha);
6684 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6685 	pla.pa_head = V_inet_pfil_head;
6686 	pla.pa_hook = V_pf_ip4_out_hook;
6687 	ret = pfil_link(&pla);
6688 	MPASS(ret == 0);
6689 	if (V_pf_filter_local) {
6690 		pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6691 		pla.pa_head = V_inet_local_pfil_head;
6692 		pla.pa_hook = V_pf_ip4_out_hook;
6693 		ret = pfil_link(&pla);
6694 		MPASS(ret == 0);
6695 	}
6696 #endif
6697 #ifdef INET6
6698 	pha.pa_type = PFIL_TYPE_IP6;
6699 	pha.pa_mbuf_chk = pf_check6_in;
6700 	pha.pa_flags = PFIL_IN;
6701 	pha.pa_rulname = "default-in6";
6702 	V_pf_ip6_in_hook = pfil_add_hook(&pha);
6703 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6704 	pla.pa_head = V_inet6_pfil_head;
6705 	pla.pa_hook = V_pf_ip6_in_hook;
6706 	ret = pfil_link(&pla);
6707 	MPASS(ret == 0);
6708 	pha.pa_mbuf_chk = pf_check6_out;
6709 	pha.pa_rulname = "default-out6";
6710 	pha.pa_flags = PFIL_OUT;
6711 	V_pf_ip6_out_hook = pfil_add_hook(&pha);
6712 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6713 	pla.pa_head = V_inet6_pfil_head;
6714 	pla.pa_hook = V_pf_ip6_out_hook;
6715 	ret = pfil_link(&pla);
6716 	MPASS(ret == 0);
6717 	if (V_pf_filter_local) {
6718 		pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6719 		pla.pa_head = V_inet6_local_pfil_head;
6720 		pla.pa_hook = V_pf_ip6_out_hook;
6721 		ret = pfil_link(&pla);
6722 		MPASS(ret == 0);
6723 	}
6724 #endif
6725 
6726 	atomic_store_bool(&V_pf_pfil_hooked, true);
6727 }
6728 
6729 static void
6730 dehook_pf_eth(void)
6731 {
6732 
6733 	if (!atomic_load_bool(&V_pf_pfil_eth_hooked))
6734 		return;
6735 
6736 	pfil_remove_hook(V_pf_eth_in_hook);
6737 	pfil_remove_hook(V_pf_eth_out_hook);
6738 
6739 	atomic_store_bool(&V_pf_pfil_eth_hooked, false);
6740 }
6741 
6742 static void
6743 dehook_pf(void)
6744 {
6745 
6746 	if (!atomic_load_bool(&V_pf_pfil_hooked))
6747 		return;
6748 
6749 #ifdef INET
6750 	pfil_remove_hook(V_pf_ip4_in_hook);
6751 	pfil_remove_hook(V_pf_ip4_out_hook);
6752 #endif
6753 #ifdef INET6
6754 	pfil_remove_hook(V_pf_ip6_in_hook);
6755 	pfil_remove_hook(V_pf_ip6_out_hook);
6756 #endif
6757 
6758 	atomic_store_bool(&V_pf_pfil_hooked, false);
6759 }
6760 
6761 static void
6762 pf_load_vnet(void)
6763 {
6764 	V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
6765 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
6766 
6767 	rm_init_flags(&V_pf_rules_lock, "pf rulesets", RM_RECURSE);
6768 	sx_init(&V_pf_ioctl_lock, "pf ioctl");
6769 
6770 	pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
6771 	    PF_RULE_TAG_HASH_SIZE_DEFAULT);
6772 #ifdef ALTQ
6773 	pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
6774 	    PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
6775 #endif
6776 
6777 	V_pf_keth = &V_pf_main_keth_anchor.ruleset;
6778 
6779 	pfattach_vnet();
6780 	V_pf_vnet_active = 1;
6781 }
6782 
6783 static int
6784 pf_load(void)
6785 {
6786 	int error;
6787 
6788 	sx_init(&pf_end_lock, "pf end thread");
6789 
6790 	pf_mtag_initialize();
6791 
6792 	pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
6793 	if (pf_dev == NULL)
6794 		return (ENOMEM);
6795 
6796 	pf_end_threads = 0;
6797 	error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
6798 	if (error != 0)
6799 		return (error);
6800 
6801 	pfi_initialize();
6802 
6803 	return (0);
6804 }
6805 
6806 static void
6807 pf_unload_vnet(void)
6808 {
6809 	int ret __diagused;
6810 
6811 	V_pf_vnet_active = 0;
6812 	V_pf_status.running = 0;
6813 	dehook_pf();
6814 	dehook_pf_eth();
6815 
6816 	PF_RULES_WLOCK();
6817 	pf_syncookies_cleanup();
6818 	shutdown_pf();
6819 	PF_RULES_WUNLOCK();
6820 
6821 	/* Make sure we've cleaned up ethernet rules before we continue. */
6822 	NET_EPOCH_DRAIN_CALLBACKS();
6823 
6824 	ret = swi_remove(V_pf_swi_cookie);
6825 	MPASS(ret == 0);
6826 	ret = intr_event_destroy(V_pf_swi_ie);
6827 	MPASS(ret == 0);
6828 
6829 	pf_unload_vnet_purge();
6830 
6831 	pf_normalize_cleanup();
6832 	PF_RULES_WLOCK();
6833 	pfi_cleanup_vnet();
6834 	PF_RULES_WUNLOCK();
6835 	pfr_cleanup();
6836 	pf_osfp_flush();
6837 	pf_cleanup();
6838 	if (IS_DEFAULT_VNET(curvnet))
6839 		pf_mtag_cleanup();
6840 
6841 	pf_cleanup_tagset(&V_pf_tags);
6842 #ifdef ALTQ
6843 	pf_cleanup_tagset(&V_pf_qids);
6844 #endif
6845 	uma_zdestroy(V_pf_tag_z);
6846 
6847 #ifdef PF_WANT_32_TO_64_COUNTER
6848 	PF_RULES_WLOCK();
6849 	LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist);
6850 
6851 	MPASS(LIST_EMPTY(&V_pf_allkiflist));
6852 	MPASS(V_pf_allkifcount == 0);
6853 
6854 	LIST_REMOVE(&V_pf_default_rule, allrulelist);
6855 	V_pf_allrulecount--;
6856 	LIST_REMOVE(V_pf_rulemarker, allrulelist);
6857 
6858 	/*
6859 	 * There are known pf rule leaks when running the test suite.
6860 	 */
6861 #ifdef notyet
6862 	MPASS(LIST_EMPTY(&V_pf_allrulelist));
6863 	MPASS(V_pf_allrulecount == 0);
6864 #endif
6865 
6866 	PF_RULES_WUNLOCK();
6867 
6868 	free(V_pf_kifmarker, PFI_MTYPE);
6869 	free(V_pf_rulemarker, M_PFRULE);
6870 #endif
6871 
6872 	/* Free counters last as we updated them during shutdown. */
6873 	pf_counter_u64_deinit(&V_pf_default_rule.evaluations);
6874 	for (int i = 0; i < 2; i++) {
6875 		pf_counter_u64_deinit(&V_pf_default_rule.packets[i]);
6876 		pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]);
6877 	}
6878 	counter_u64_free(V_pf_default_rule.states_cur);
6879 	counter_u64_free(V_pf_default_rule.states_tot);
6880 	counter_u64_free(V_pf_default_rule.src_nodes);
6881 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, V_pf_default_rule.timestamp);
6882 
6883 	for (int i = 0; i < PFRES_MAX; i++)
6884 		counter_u64_free(V_pf_status.counters[i]);
6885 	for (int i = 0; i < KLCNT_MAX; i++)
6886 		counter_u64_free(V_pf_status.lcounters[i]);
6887 	for (int i = 0; i < FCNT_MAX; i++)
6888 		pf_counter_u64_deinit(&V_pf_status.fcounters[i]);
6889 	for (int i = 0; i < SCNT_MAX; i++)
6890 		counter_u64_free(V_pf_status.scounters[i]);
6891 
6892 	rm_destroy(&V_pf_rules_lock);
6893 	sx_destroy(&V_pf_ioctl_lock);
6894 }
6895 
6896 static void
6897 pf_unload(void)
6898 {
6899 
6900 	sx_xlock(&pf_end_lock);
6901 	pf_end_threads = 1;
6902 	while (pf_end_threads < 2) {
6903 		wakeup_one(pf_purge_thread);
6904 		sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
6905 	}
6906 	sx_xunlock(&pf_end_lock);
6907 
6908 	if (pf_dev != NULL)
6909 		destroy_dev(pf_dev);
6910 
6911 	pfi_cleanup();
6912 
6913 	sx_destroy(&pf_end_lock);
6914 }
6915 
6916 static void
6917 vnet_pf_init(void *unused __unused)
6918 {
6919 
6920 	pf_load_vnet();
6921 }
6922 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6923     vnet_pf_init, NULL);
6924 
6925 static void
6926 vnet_pf_uninit(const void *unused __unused)
6927 {
6928 
6929 	pf_unload_vnet();
6930 }
6931 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
6932 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6933     vnet_pf_uninit, NULL);
6934 
6935 static int
6936 pf_modevent(module_t mod, int type, void *data)
6937 {
6938 	int error = 0;
6939 
6940 	switch(type) {
6941 	case MOD_LOAD:
6942 		error = pf_load();
6943 		break;
6944 	case MOD_UNLOAD:
6945 		/* Handled in SYSUNINIT(pf_unload) to ensure it's done after
6946 		 * the vnet_pf_uninit()s */
6947 		break;
6948 	default:
6949 		error = EINVAL;
6950 		break;
6951 	}
6952 
6953 	return (error);
6954 }
6955 
6956 static moduledata_t pf_mod = {
6957 	"pf",
6958 	pf_modevent,
6959 	0
6960 };
6961 
6962 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
6963 MODULE_VERSION(pf, PF_MODVER);
6964