xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision 2e620256bd76c449c835c604e404483437743011)
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_nl.h>
87 #include <netpfil/pf/pf_nv.h>
88 
89 #ifdef INET6
90 #include <netinet/ip6.h>
91 #endif /* INET6 */
92 
93 #ifdef ALTQ
94 #include <net/altq/altq.h>
95 #endif
96 
97 SDT_PROBE_DEFINE3(pf, ioctl, ioctl, error, "int", "int", "int");
98 SDT_PROBE_DEFINE3(pf, ioctl, function, error, "char *", "int", "int");
99 SDT_PROBE_DEFINE2(pf, ioctl, addrule, error, "int", "int");
100 SDT_PROBE_DEFINE2(pf, ioctl, nvchk, error, "int", "int");
101 
102 static struct pf_kpool	*pf_get_kpool(const char *, u_int32_t, u_int8_t,
103 			    u_int32_t, u_int8_t, u_int8_t, u_int8_t);
104 
105 static void		 pf_mv_kpool(struct pf_kpalist *, struct pf_kpalist *);
106 static void		 pf_empty_kpool(struct pf_kpalist *);
107 static int		 pfioctl(struct cdev *, u_long, caddr_t, int,
108 			    struct thread *);
109 static int		 pf_begin_eth(uint32_t *, const char *);
110 static void		 pf_rollback_eth_cb(struct epoch_context *);
111 static int		 pf_rollback_eth(uint32_t, const char *);
112 static int		 pf_commit_eth(uint32_t, const char *);
113 static void		 pf_free_eth_rule(struct pf_keth_rule *);
114 #ifdef ALTQ
115 static int		 pf_begin_altq(u_int32_t *);
116 static int		 pf_rollback_altq(u_int32_t);
117 static int		 pf_commit_altq(u_int32_t);
118 static int		 pf_enable_altq(struct pf_altq *);
119 static int		 pf_disable_altq(struct pf_altq *);
120 static uint16_t		 pf_qname2qid(const char *);
121 static void		 pf_qid_unref(uint16_t);
122 #endif /* ALTQ */
123 static int		 pf_begin_rules(u_int32_t *, int, const char *);
124 static int		 pf_rollback_rules(u_int32_t, int, char *);
125 static int		 pf_setup_pfsync_matching(struct pf_kruleset *);
126 static void		 pf_hash_rule_rolling(MD5_CTX *, struct pf_krule *);
127 static void		 pf_hash_rule(struct pf_krule *);
128 static void		 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
129 static int		 pf_commit_rules(u_int32_t, int, char *);
130 static int		 pf_addr_setup(struct pf_kruleset *,
131 			    struct pf_addr_wrap *, sa_family_t);
132 static void		 pf_addr_copyout(struct pf_addr_wrap *);
133 static void		 pf_src_node_copy(const struct pf_ksrc_node *,
134 			    struct pf_src_node *);
135 #ifdef ALTQ
136 static int		 pf_export_kaltq(struct pf_altq *,
137 			    struct pfioc_altq_v1 *, size_t);
138 static int		 pf_import_kaltq(struct pfioc_altq_v1 *,
139 			    struct pf_altq *, size_t);
140 #endif /* ALTQ */
141 
142 VNET_DEFINE(struct pf_krule,	pf_default_rule);
143 
144 static __inline int             pf_krule_compare(struct pf_krule *,
145 				    struct pf_krule *);
146 
147 RB_GENERATE(pf_krule_global, pf_krule, entry_global, pf_krule_compare);
148 
149 #ifdef ALTQ
150 VNET_DEFINE_STATIC(int,		pf_altq_running);
151 #define	V_pf_altq_running	VNET(pf_altq_running)
152 #endif
153 
154 #define	TAGID_MAX	 50000
155 struct pf_tagname {
156 	TAILQ_ENTRY(pf_tagname)	namehash_entries;
157 	TAILQ_ENTRY(pf_tagname)	taghash_entries;
158 	char			name[PF_TAG_NAME_SIZE];
159 	uint16_t		tag;
160 	int			ref;
161 };
162 
163 struct pf_tagset {
164 	TAILQ_HEAD(, pf_tagname)	*namehash;
165 	TAILQ_HEAD(, pf_tagname)	*taghash;
166 	unsigned int			 mask;
167 	uint32_t			 seed;
168 	BITSET_DEFINE(, TAGID_MAX)	 avail;
169 };
170 
171 VNET_DEFINE(struct pf_tagset, pf_tags);
172 #define	V_pf_tags	VNET(pf_tags)
173 static unsigned int	pf_rule_tag_hashsize;
174 #define	PF_RULE_TAG_HASH_SIZE_DEFAULT	128
175 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN,
176     &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT,
177     "Size of pf(4) rule tag hashtable");
178 
179 #ifdef ALTQ
180 VNET_DEFINE(struct pf_tagset, pf_qids);
181 #define	V_pf_qids	VNET(pf_qids)
182 static unsigned int	pf_queue_tag_hashsize;
183 #define	PF_QUEUE_TAG_HASH_SIZE_DEFAULT	128
184 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN,
185     &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT,
186     "Size of pf(4) queue tag hashtable");
187 #endif
188 VNET_DEFINE(uma_zone_t,	 pf_tag_z);
189 #define	V_pf_tag_z		 VNET(pf_tag_z)
190 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
191 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
192 
193 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
194 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
195 #endif
196 
197 VNET_DEFINE_STATIC(bool, pf_filter_local) = false;
198 #define V_pf_filter_local	VNET(pf_filter_local)
199 SYSCTL_BOOL(_net_pf, OID_AUTO, filter_local, CTLFLAG_VNET | CTLFLAG_RW,
200     &VNET_NAME(pf_filter_local), false,
201     "Enable filtering for packets delivered to local network stack");
202 
203 #ifdef PF_DEFAULT_TO_DROP
204 VNET_DEFINE_STATIC(bool, default_to_drop) = true;
205 #else
206 VNET_DEFINE_STATIC(bool, default_to_drop);
207 #endif
208 #define	V_default_to_drop VNET(default_to_drop)
209 SYSCTL_BOOL(_net_pf, OID_AUTO, default_to_drop, CTLFLAG_RDTUN | CTLFLAG_VNET,
210     &VNET_NAME(default_to_drop), false,
211     "Make the default rule drop all packets.");
212 
213 static void		 pf_init_tagset(struct pf_tagset *, unsigned int *,
214 			    unsigned int);
215 static void		 pf_cleanup_tagset(struct pf_tagset *);
216 static uint16_t		 tagname2hashindex(const struct pf_tagset *, const char *);
217 static uint16_t		 tag2hashindex(const struct pf_tagset *, uint16_t);
218 static u_int16_t	 tagname2tag(struct pf_tagset *, const char *);
219 static u_int16_t	 pf_tagname2tag(const char *);
220 static void		 tag_unref(struct pf_tagset *, u_int16_t);
221 
222 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
223 
224 struct cdev *pf_dev;
225 
226 /*
227  * XXX - These are new and need to be checked when moveing to a new version
228  */
229 static void		 pf_clear_all_states(void);
230 static unsigned int	 pf_clear_states(const struct pf_kstate_kill *);
231 static void		 pf_killstates(struct pf_kstate_kill *,
232 			    unsigned int *);
233 static int		 pf_killstates_row(struct pf_kstate_kill *,
234 			    struct pf_idhash *);
235 static int		 pf_killstates_nv(struct pfioc_nv *);
236 static int		 pf_clearstates_nv(struct pfioc_nv *);
237 static int		 pf_getstate(struct pfioc_nv *);
238 static int		 pf_getstatus(struct pfioc_nv *);
239 static int		 pf_clear_tables(void);
240 static void		 pf_clear_srcnodes(struct pf_ksrc_node *);
241 static void		 pf_kill_srcnodes(struct pfioc_src_node_kill *);
242 static int		 pf_keepcounters(struct pfioc_nv *);
243 static void		 pf_tbladdr_copyout(struct pf_addr_wrap *);
244 
245 /*
246  * Wrapper functions for pfil(9) hooks
247  */
248 static pfil_return_t pf_eth_check_in(struct mbuf **m, struct ifnet *ifp,
249     int flags, void *ruleset __unused, struct inpcb *inp);
250 static pfil_return_t pf_eth_check_out(struct mbuf **m, struct ifnet *ifp,
251     int flags, void *ruleset __unused, struct inpcb *inp);
252 #ifdef INET
253 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp,
254     int flags, void *ruleset __unused, struct inpcb *inp);
255 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp,
256     int flags, void *ruleset __unused, struct inpcb *inp);
257 #endif
258 #ifdef INET6
259 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp,
260     int flags, void *ruleset __unused, struct inpcb *inp);
261 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp,
262     int flags, void *ruleset __unused, struct inpcb *inp);
263 #endif
264 
265 static void		hook_pf_eth(void);
266 static void		hook_pf(void);
267 static void		dehook_pf_eth(void);
268 static void		dehook_pf(void);
269 static int		shutdown_pf(void);
270 static int		pf_load(void);
271 static void		pf_unload(void);
272 
273 static struct cdevsw pf_cdevsw = {
274 	.d_ioctl =	pfioctl,
275 	.d_name =	PF_NAME,
276 	.d_version =	D_VERSION,
277 };
278 
279 VNET_DEFINE_STATIC(bool, pf_pfil_hooked);
280 #define V_pf_pfil_hooked	VNET(pf_pfil_hooked)
281 VNET_DEFINE_STATIC(bool, pf_pfil_eth_hooked);
282 #define V_pf_pfil_eth_hooked	VNET(pf_pfil_eth_hooked)
283 
284 /*
285  * We need a flag that is neither hooked nor running to know when
286  * the VNET is "valid".  We primarily need this to control (global)
287  * external event, e.g., eventhandlers.
288  */
289 VNET_DEFINE(int, pf_vnet_active);
290 #define V_pf_vnet_active	VNET(pf_vnet_active)
291 
292 int pf_end_threads;
293 struct proc *pf_purge_proc;
294 
295 VNET_DEFINE(struct rmlock, pf_rules_lock);
296 VNET_DEFINE_STATIC(struct sx, pf_ioctl_lock);
297 #define	V_pf_ioctl_lock		VNET(pf_ioctl_lock)
298 struct sx			pf_end_lock;
299 
300 /* pfsync */
301 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
302 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
303 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
304 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
305 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
306 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
307 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
308 
309 /* pflog */
310 pflog_packet_t			*pflog_packet_ptr = NULL;
311 
312 /*
313  * Copy a user-provided string, returning an error if truncation would occur.
314  * Avoid scanning past "sz" bytes in the source string since there's no
315  * guarantee that it's nul-terminated.
316  */
317 static int
318 pf_user_strcpy(char *dst, const char *src, size_t sz)
319 {
320 	if (strnlen(src, sz) == sz)
321 		return (EINVAL);
322 	(void)strlcpy(dst, src, sz);
323 	return (0);
324 }
325 
326 static void
327 pfattach_vnet(void)
328 {
329 	u_int32_t *my_timeout = V_pf_default_rule.timeout;
330 
331 	bzero(&V_pf_status, sizeof(V_pf_status));
332 
333 	pf_initialize();
334 	pfr_initialize();
335 	pfi_initialize_vnet();
336 	pf_normalize_init();
337 	pf_syncookies_init();
338 
339 	V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
340 	V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
341 
342 	RB_INIT(&V_pf_anchors);
343 	pf_init_kruleset(&pf_main_ruleset);
344 
345 	pf_init_keth(V_pf_keth);
346 
347 	/* default rule should never be garbage collected */
348 	V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
349 	V_pf_default_rule.action = V_default_to_drop ? PF_DROP : PF_PASS;
350 	V_pf_default_rule.nr = -1;
351 	V_pf_default_rule.rtableid = -1;
352 
353 	pf_counter_u64_init(&V_pf_default_rule.evaluations, M_WAITOK);
354 	for (int i = 0; i < 2; i++) {
355 		pf_counter_u64_init(&V_pf_default_rule.packets[i], M_WAITOK);
356 		pf_counter_u64_init(&V_pf_default_rule.bytes[i], M_WAITOK);
357 	}
358 	V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
359 	V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
360 	V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
361 
362 	V_pf_default_rule.timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
363 	    M_WAITOK | M_ZERO);
364 
365 #ifdef PF_WANT_32_TO_64_COUNTER
366 	V_pf_kifmarker = malloc(sizeof(*V_pf_kifmarker), PFI_MTYPE, M_WAITOK | M_ZERO);
367 	V_pf_rulemarker = malloc(sizeof(*V_pf_rulemarker), M_PFRULE, M_WAITOK | M_ZERO);
368 	PF_RULES_WLOCK();
369 	LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist);
370 	LIST_INSERT_HEAD(&V_pf_allrulelist, &V_pf_default_rule, allrulelist);
371 	V_pf_allrulecount++;
372 	LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist);
373 	PF_RULES_WUNLOCK();
374 #endif
375 
376 	/* initialize default timeouts */
377 	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
378 	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
379 	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
380 	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
381 	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
382 	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
383 	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
384 	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
385 	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
386 	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
387 	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
388 	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
389 	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
390 	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
391 	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
392 	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
393 	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
394 	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
395 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
396 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
397 
398 	V_pf_status.debug = PF_DEBUG_URGENT;
399 	/*
400 	 * XXX This is different than in OpenBSD where reassembly is enabled by
401 	 * defult. In FreeBSD we expect people to still use scrub rules and
402 	 * switch to the new syntax later. Only when they switch they must
403 	 * explicitly enable reassemle. We could change the default once the
404 	 * scrub rule functionality is hopefully removed some day in future.
405 	 */
406 	V_pf_status.reass = 0;
407 
408 	V_pf_pfil_hooked = false;
409 	V_pf_pfil_eth_hooked = false;
410 
411 	/* XXX do our best to avoid a conflict */
412 	V_pf_status.hostid = arc4random();
413 
414 	for (int i = 0; i < PFRES_MAX; i++)
415 		V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
416 	for (int i = 0; i < KLCNT_MAX; i++)
417 		V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
418 	for (int i = 0; i < FCNT_MAX; i++)
419 		pf_counter_u64_init(&V_pf_status.fcounters[i], M_WAITOK);
420 	for (int i = 0; i < SCNT_MAX; i++)
421 		V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
422 
423 	if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET,
424 	    INTR_MPSAFE, &V_pf_swi_cookie) != 0)
425 		/* XXXGL: leaked all above. */
426 		return;
427 }
428 
429 static struct pf_kpool *
430 pf_get_kpool(const char *anchor, u_int32_t ticket, u_int8_t rule_action,
431     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
432     u_int8_t check_ticket)
433 {
434 	struct pf_kruleset	*ruleset;
435 	struct pf_krule		*rule;
436 	int			 rs_num;
437 
438 	ruleset = pf_find_kruleset(anchor);
439 	if (ruleset == NULL)
440 		return (NULL);
441 	rs_num = pf_get_ruleset_number(rule_action);
442 	if (rs_num >= PF_RULESET_MAX)
443 		return (NULL);
444 	if (active) {
445 		if (check_ticket && ticket !=
446 		    ruleset->rules[rs_num].active.ticket)
447 			return (NULL);
448 		if (r_last)
449 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
450 			    pf_krulequeue);
451 		else
452 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
453 	} else {
454 		if (check_ticket && ticket !=
455 		    ruleset->rules[rs_num].inactive.ticket)
456 			return (NULL);
457 		if (r_last)
458 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
459 			    pf_krulequeue);
460 		else
461 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
462 	}
463 	if (!r_last) {
464 		while ((rule != NULL) && (rule->nr != rule_number))
465 			rule = TAILQ_NEXT(rule, entries);
466 	}
467 	if (rule == NULL)
468 		return (NULL);
469 
470 	return (&rule->rpool);
471 }
472 
473 static void
474 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb)
475 {
476 	struct pf_kpooladdr	*mv_pool_pa;
477 
478 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
479 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
480 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
481 	}
482 }
483 
484 static void
485 pf_empty_kpool(struct pf_kpalist *poola)
486 {
487 	struct pf_kpooladdr *pa;
488 
489 	while ((pa = TAILQ_FIRST(poola)) != NULL) {
490 		switch (pa->addr.type) {
491 		case PF_ADDR_DYNIFTL:
492 			pfi_dynaddr_remove(pa->addr.p.dyn);
493 			break;
494 		case PF_ADDR_TABLE:
495 			/* XXX: this could be unfinished pooladdr on pabuf */
496 			if (pa->addr.p.tbl != NULL)
497 				pfr_detach_table(pa->addr.p.tbl);
498 			break;
499 		}
500 		if (pa->kif)
501 			pfi_kkif_unref(pa->kif);
502 		TAILQ_REMOVE(poola, pa, entries);
503 		free(pa, M_PFRULE);
504 	}
505 }
506 
507 static void
508 pf_unlink_rule_locked(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
509 {
510 
511 	PF_RULES_WASSERT();
512 	PF_UNLNKDRULES_ASSERT();
513 
514 	TAILQ_REMOVE(rulequeue, rule, entries);
515 
516 	rule->rule_ref |= PFRULE_REFS;
517 	TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
518 }
519 
520 static void
521 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
522 {
523 
524 	PF_RULES_WASSERT();
525 
526 	PF_UNLNKDRULES_LOCK();
527 	pf_unlink_rule_locked(rulequeue, rule);
528 	PF_UNLNKDRULES_UNLOCK();
529 }
530 
531 static void
532 pf_free_eth_rule(struct pf_keth_rule *rule)
533 {
534 	PF_RULES_WASSERT();
535 
536 	if (rule == NULL)
537 		return;
538 
539 	if (rule->tag)
540 		tag_unref(&V_pf_tags, rule->tag);
541 	if (rule->match_tag)
542 		tag_unref(&V_pf_tags, rule->match_tag);
543 #ifdef ALTQ
544 	pf_qid_unref(rule->qid);
545 #endif
546 
547 	if (rule->bridge_to)
548 		pfi_kkif_unref(rule->bridge_to);
549 	if (rule->kif)
550 		pfi_kkif_unref(rule->kif);
551 
552 	if (rule->ipsrc.addr.type == PF_ADDR_TABLE)
553 		pfr_detach_table(rule->ipsrc.addr.p.tbl);
554 	if (rule->ipdst.addr.type == PF_ADDR_TABLE)
555 		pfr_detach_table(rule->ipdst.addr.p.tbl);
556 
557 	counter_u64_free(rule->evaluations);
558 	for (int i = 0; i < 2; i++) {
559 		counter_u64_free(rule->packets[i]);
560 		counter_u64_free(rule->bytes[i]);
561 	}
562 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
563 	pf_keth_anchor_remove(rule);
564 
565 	free(rule, M_PFRULE);
566 }
567 
568 void
569 pf_free_rule(struct pf_krule *rule)
570 {
571 
572 	PF_RULES_WASSERT();
573 	PF_CONFIG_ASSERT();
574 
575 	if (rule->tag)
576 		tag_unref(&V_pf_tags, rule->tag);
577 	if (rule->match_tag)
578 		tag_unref(&V_pf_tags, rule->match_tag);
579 #ifdef ALTQ
580 	if (rule->pqid != rule->qid)
581 		pf_qid_unref(rule->pqid);
582 	pf_qid_unref(rule->qid);
583 #endif
584 	switch (rule->src.addr.type) {
585 	case PF_ADDR_DYNIFTL:
586 		pfi_dynaddr_remove(rule->src.addr.p.dyn);
587 		break;
588 	case PF_ADDR_TABLE:
589 		pfr_detach_table(rule->src.addr.p.tbl);
590 		break;
591 	}
592 	switch (rule->dst.addr.type) {
593 	case PF_ADDR_DYNIFTL:
594 		pfi_dynaddr_remove(rule->dst.addr.p.dyn);
595 		break;
596 	case PF_ADDR_TABLE:
597 		pfr_detach_table(rule->dst.addr.p.tbl);
598 		break;
599 	}
600 	if (rule->overload_tbl)
601 		pfr_detach_table(rule->overload_tbl);
602 	if (rule->kif)
603 		pfi_kkif_unref(rule->kif);
604 	pf_kanchor_remove(rule);
605 	pf_empty_kpool(&rule->rpool.list);
606 
607 	pf_krule_free(rule);
608 }
609 
610 static void
611 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
612     unsigned int default_size)
613 {
614 	unsigned int i;
615 	unsigned int hashsize;
616 
617 	if (*tunable_size == 0 || !powerof2(*tunable_size))
618 		*tunable_size = default_size;
619 
620 	hashsize = *tunable_size;
621 	ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
622 	    M_WAITOK);
623 	ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
624 	    M_WAITOK);
625 	ts->mask = hashsize - 1;
626 	ts->seed = arc4random();
627 	for (i = 0; i < hashsize; i++) {
628 		TAILQ_INIT(&ts->namehash[i]);
629 		TAILQ_INIT(&ts->taghash[i]);
630 	}
631 	BIT_FILL(TAGID_MAX, &ts->avail);
632 }
633 
634 static void
635 pf_cleanup_tagset(struct pf_tagset *ts)
636 {
637 	unsigned int i;
638 	unsigned int hashsize;
639 	struct pf_tagname *t, *tmp;
640 
641 	/*
642 	 * Only need to clean up one of the hashes as each tag is hashed
643 	 * into each table.
644 	 */
645 	hashsize = ts->mask + 1;
646 	for (i = 0; i < hashsize; i++)
647 		TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
648 			uma_zfree(V_pf_tag_z, t);
649 
650 	free(ts->namehash, M_PFHASH);
651 	free(ts->taghash, M_PFHASH);
652 }
653 
654 static uint16_t
655 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
656 {
657 	size_t len;
658 
659 	len = strnlen(tagname, PF_TAG_NAME_SIZE - 1);
660 	return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask);
661 }
662 
663 static uint16_t
664 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
665 {
666 
667 	return (tag & ts->mask);
668 }
669 
670 static u_int16_t
671 tagname2tag(struct pf_tagset *ts, const char *tagname)
672 {
673 	struct pf_tagname	*tag;
674 	u_int32_t		 index;
675 	u_int16_t		 new_tagid;
676 
677 	PF_RULES_WASSERT();
678 
679 	index = tagname2hashindex(ts, tagname);
680 	TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
681 		if (strcmp(tagname, tag->name) == 0) {
682 			tag->ref++;
683 			return (tag->tag);
684 		}
685 
686 	/*
687 	 * new entry
688 	 *
689 	 * to avoid fragmentation, we do a linear search from the beginning
690 	 * and take the first free slot we find.
691 	 */
692 	new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
693 	/*
694 	 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
695 	 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
696 	 * set.  It may also return a bit number greater than TAGID_MAX due
697 	 * to rounding of the number of bits in the vector up to a multiple
698 	 * of the vector word size at declaration/allocation time.
699 	 */
700 	if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
701 		return (0);
702 
703 	/* Mark the tag as in use.  Bits are 0-based for BIT_CLR() */
704 	BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
705 
706 	/* allocate and fill new struct pf_tagname */
707 	tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
708 	if (tag == NULL)
709 		return (0);
710 	strlcpy(tag->name, tagname, sizeof(tag->name));
711 	tag->tag = new_tagid;
712 	tag->ref = 1;
713 
714 	/* Insert into namehash */
715 	TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
716 
717 	/* Insert into taghash */
718 	index = tag2hashindex(ts, new_tagid);
719 	TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
720 
721 	return (tag->tag);
722 }
723 
724 static void
725 tag_unref(struct pf_tagset *ts, u_int16_t tag)
726 {
727 	struct pf_tagname	*t;
728 	uint16_t		 index;
729 
730 	PF_RULES_WASSERT();
731 
732 	index = tag2hashindex(ts, tag);
733 	TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
734 		if (tag == t->tag) {
735 			if (--t->ref == 0) {
736 				TAILQ_REMOVE(&ts->taghash[index], t,
737 				    taghash_entries);
738 				index = tagname2hashindex(ts, t->name);
739 				TAILQ_REMOVE(&ts->namehash[index], t,
740 				    namehash_entries);
741 				/* Bits are 0-based for BIT_SET() */
742 				BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
743 				uma_zfree(V_pf_tag_z, t);
744 			}
745 			break;
746 		}
747 }
748 
749 static uint16_t
750 pf_tagname2tag(const char *tagname)
751 {
752 	return (tagname2tag(&V_pf_tags, tagname));
753 }
754 
755 static int
756 pf_begin_eth(uint32_t *ticket, const char *anchor)
757 {
758 	struct pf_keth_rule *rule, *tmp;
759 	struct pf_keth_ruleset *rs;
760 
761 	PF_RULES_WASSERT();
762 
763 	rs = pf_find_or_create_keth_ruleset(anchor);
764 	if (rs == NULL)
765 		return (EINVAL);
766 
767 	/* Purge old inactive rules. */
768 	TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
769 	    tmp) {
770 		TAILQ_REMOVE(rs->inactive.rules, rule,
771 		    entries);
772 		pf_free_eth_rule(rule);
773 	}
774 
775 	*ticket = ++rs->inactive.ticket;
776 	rs->inactive.open = 1;
777 
778 	return (0);
779 }
780 
781 static void
782 pf_rollback_eth_cb(struct epoch_context *ctx)
783 {
784 	struct pf_keth_ruleset *rs;
785 
786 	rs = __containerof(ctx, struct pf_keth_ruleset, epoch_ctx);
787 
788 	CURVNET_SET(rs->vnet);
789 
790 	PF_RULES_WLOCK();
791 	pf_rollback_eth(rs->inactive.ticket,
792 	    rs->anchor ? rs->anchor->path : "");
793 	PF_RULES_WUNLOCK();
794 
795 	CURVNET_RESTORE();
796 }
797 
798 static int
799 pf_rollback_eth(uint32_t ticket, const char *anchor)
800 {
801 	struct pf_keth_rule *rule, *tmp;
802 	struct pf_keth_ruleset *rs;
803 
804 	PF_RULES_WASSERT();
805 
806 	rs = pf_find_keth_ruleset(anchor);
807 	if (rs == NULL)
808 		return (EINVAL);
809 
810 	if (!rs->inactive.open ||
811 	    ticket != rs->inactive.ticket)
812 		return (0);
813 
814 	/* Purge old inactive rules. */
815 	TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
816 	    tmp) {
817 		TAILQ_REMOVE(rs->inactive.rules, rule, entries);
818 		pf_free_eth_rule(rule);
819 	}
820 
821 	rs->inactive.open = 0;
822 
823 	pf_remove_if_empty_keth_ruleset(rs);
824 
825 	return (0);
826 }
827 
828 #define	PF_SET_SKIP_STEPS(i)					\
829 	do {							\
830 		while (head[i] != cur) {			\
831 			head[i]->skip[i].ptr = cur;		\
832 			head[i] = TAILQ_NEXT(head[i], entries);	\
833 		}						\
834 	} while (0)
835 
836 static void
837 pf_eth_calc_skip_steps(struct pf_keth_ruleq *rules)
838 {
839 	struct pf_keth_rule *cur, *prev, *head[PFE_SKIP_COUNT];
840 	int i;
841 
842 	cur = TAILQ_FIRST(rules);
843 	prev = cur;
844 	for (i = 0; i < PFE_SKIP_COUNT; ++i)
845 		head[i] = cur;
846 	while (cur != NULL) {
847 		if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
848 			PF_SET_SKIP_STEPS(PFE_SKIP_IFP);
849 		if (cur->direction != prev->direction)
850 			PF_SET_SKIP_STEPS(PFE_SKIP_DIR);
851 		if (cur->proto != prev->proto)
852 			PF_SET_SKIP_STEPS(PFE_SKIP_PROTO);
853 		if (memcmp(&cur->src, &prev->src, sizeof(cur->src)) != 0)
854 			PF_SET_SKIP_STEPS(PFE_SKIP_SRC_ADDR);
855 		if (memcmp(&cur->dst, &prev->dst, sizeof(cur->dst)) != 0)
856 			PF_SET_SKIP_STEPS(PFE_SKIP_DST_ADDR);
857 		if (cur->ipsrc.neg != prev->ipsrc.neg ||
858 		    pf_addr_wrap_neq(&cur->ipsrc.addr, &prev->ipsrc.addr))
859 			PF_SET_SKIP_STEPS(PFE_SKIP_SRC_IP_ADDR);
860 		if (cur->ipdst.neg != prev->ipdst.neg ||
861 		    pf_addr_wrap_neq(&cur->ipdst.addr, &prev->ipdst.addr))
862 			PF_SET_SKIP_STEPS(PFE_SKIP_DST_IP_ADDR);
863 
864 		prev = cur;
865 		cur = TAILQ_NEXT(cur, entries);
866 	}
867 	for (i = 0; i < PFE_SKIP_COUNT; ++i)
868 		PF_SET_SKIP_STEPS(i);
869 }
870 
871 static int
872 pf_commit_eth(uint32_t ticket, const char *anchor)
873 {
874 	struct pf_keth_ruleq *rules;
875 	struct pf_keth_ruleset *rs;
876 
877 	rs = pf_find_keth_ruleset(anchor);
878 	if (rs == NULL) {
879 		return (EINVAL);
880 	}
881 
882 	if (!rs->inactive.open ||
883 	    ticket != rs->inactive.ticket)
884 		return (EBUSY);
885 
886 	PF_RULES_WASSERT();
887 
888 	pf_eth_calc_skip_steps(rs->inactive.rules);
889 
890 	rules = rs->active.rules;
891 	ck_pr_store_ptr(&rs->active.rules, rs->inactive.rules);
892 	rs->inactive.rules = rules;
893 	rs->inactive.ticket = rs->active.ticket;
894 
895 	/* Clean up inactive rules (i.e. previously active rules), only when
896 	 * we're sure they're no longer used. */
897 	NET_EPOCH_CALL(pf_rollback_eth_cb, &rs->epoch_ctx);
898 
899 	return (0);
900 }
901 
902 #ifdef ALTQ
903 static uint16_t
904 pf_qname2qid(const char *qname)
905 {
906 	return (tagname2tag(&V_pf_qids, qname));
907 }
908 
909 static void
910 pf_qid_unref(uint16_t qid)
911 {
912 	tag_unref(&V_pf_qids, qid);
913 }
914 
915 static int
916 pf_begin_altq(u_int32_t *ticket)
917 {
918 	struct pf_altq	*altq, *tmp;
919 	int		 error = 0;
920 
921 	PF_RULES_WASSERT();
922 
923 	/* Purge the old altq lists */
924 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
925 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
926 			/* detach and destroy the discipline */
927 			error = altq_remove(altq);
928 		}
929 		free(altq, M_PFALTQ);
930 	}
931 	TAILQ_INIT(V_pf_altq_ifs_inactive);
932 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
933 		pf_qid_unref(altq->qid);
934 		free(altq, M_PFALTQ);
935 	}
936 	TAILQ_INIT(V_pf_altqs_inactive);
937 	if (error)
938 		return (error);
939 	*ticket = ++V_ticket_altqs_inactive;
940 	V_altqs_inactive_open = 1;
941 	return (0);
942 }
943 
944 static int
945 pf_rollback_altq(u_int32_t ticket)
946 {
947 	struct pf_altq	*altq, *tmp;
948 	int		 error = 0;
949 
950 	PF_RULES_WASSERT();
951 
952 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
953 		return (0);
954 	/* Purge the old altq lists */
955 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
956 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
957 			/* detach and destroy the discipline */
958 			error = altq_remove(altq);
959 		}
960 		free(altq, M_PFALTQ);
961 	}
962 	TAILQ_INIT(V_pf_altq_ifs_inactive);
963 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
964 		pf_qid_unref(altq->qid);
965 		free(altq, M_PFALTQ);
966 	}
967 	TAILQ_INIT(V_pf_altqs_inactive);
968 	V_altqs_inactive_open = 0;
969 	return (error);
970 }
971 
972 static int
973 pf_commit_altq(u_int32_t ticket)
974 {
975 	struct pf_altqqueue	*old_altqs, *old_altq_ifs;
976 	struct pf_altq		*altq, *tmp;
977 	int			 err, error = 0;
978 
979 	PF_RULES_WASSERT();
980 
981 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
982 		return (EBUSY);
983 
984 	/* swap altqs, keep the old. */
985 	old_altqs = V_pf_altqs_active;
986 	old_altq_ifs = V_pf_altq_ifs_active;
987 	V_pf_altqs_active = V_pf_altqs_inactive;
988 	V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
989 	V_pf_altqs_inactive = old_altqs;
990 	V_pf_altq_ifs_inactive = old_altq_ifs;
991 	V_ticket_altqs_active = V_ticket_altqs_inactive;
992 
993 	/* Attach new disciplines */
994 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
995 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
996 			/* attach the discipline */
997 			error = altq_pfattach(altq);
998 			if (error == 0 && V_pf_altq_running)
999 				error = pf_enable_altq(altq);
1000 			if (error != 0)
1001 				return (error);
1002 		}
1003 	}
1004 
1005 	/* Purge the old altq lists */
1006 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
1007 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
1008 			/* detach and destroy the discipline */
1009 			if (V_pf_altq_running)
1010 				error = pf_disable_altq(altq);
1011 			err = altq_pfdetach(altq);
1012 			if (err != 0 && error == 0)
1013 				error = err;
1014 			err = altq_remove(altq);
1015 			if (err != 0 && error == 0)
1016 				error = err;
1017 		}
1018 		free(altq, M_PFALTQ);
1019 	}
1020 	TAILQ_INIT(V_pf_altq_ifs_inactive);
1021 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
1022 		pf_qid_unref(altq->qid);
1023 		free(altq, M_PFALTQ);
1024 	}
1025 	TAILQ_INIT(V_pf_altqs_inactive);
1026 
1027 	V_altqs_inactive_open = 0;
1028 	return (error);
1029 }
1030 
1031 static int
1032 pf_enable_altq(struct pf_altq *altq)
1033 {
1034 	struct ifnet		*ifp;
1035 	struct tb_profile	 tb;
1036 	int			 error = 0;
1037 
1038 	if ((ifp = ifunit(altq->ifname)) == NULL)
1039 		return (EINVAL);
1040 
1041 	if (ifp->if_snd.altq_type != ALTQT_NONE)
1042 		error = altq_enable(&ifp->if_snd);
1043 
1044 	/* set tokenbucket regulator */
1045 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
1046 		tb.rate = altq->ifbandwidth;
1047 		tb.depth = altq->tbrsize;
1048 		error = tbr_set(&ifp->if_snd, &tb);
1049 	}
1050 
1051 	return (error);
1052 }
1053 
1054 static int
1055 pf_disable_altq(struct pf_altq *altq)
1056 {
1057 	struct ifnet		*ifp;
1058 	struct tb_profile	 tb;
1059 	int			 error;
1060 
1061 	if ((ifp = ifunit(altq->ifname)) == NULL)
1062 		return (EINVAL);
1063 
1064 	/*
1065 	 * when the discipline is no longer referenced, it was overridden
1066 	 * by a new one.  if so, just return.
1067 	 */
1068 	if (altq->altq_disc != ifp->if_snd.altq_disc)
1069 		return (0);
1070 
1071 	error = altq_disable(&ifp->if_snd);
1072 
1073 	if (error == 0) {
1074 		/* clear tokenbucket regulator */
1075 		tb.rate = 0;
1076 		error = tbr_set(&ifp->if_snd, &tb);
1077 	}
1078 
1079 	return (error);
1080 }
1081 
1082 static int
1083 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
1084     struct pf_altq *altq)
1085 {
1086 	struct ifnet	*ifp1;
1087 	int		 error = 0;
1088 
1089 	/* Deactivate the interface in question */
1090 	altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
1091 	if ((ifp1 = ifunit(altq->ifname)) == NULL ||
1092 	    (remove && ifp1 == ifp)) {
1093 		altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
1094 	} else {
1095 		error = altq_add(ifp1, altq);
1096 
1097 		if (ticket != V_ticket_altqs_inactive)
1098 			error = EBUSY;
1099 
1100 		if (error)
1101 			free(altq, M_PFALTQ);
1102 	}
1103 
1104 	return (error);
1105 }
1106 
1107 void
1108 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
1109 {
1110 	struct pf_altq	*a1, *a2, *a3;
1111 	u_int32_t	 ticket;
1112 	int		 error = 0;
1113 
1114 	/*
1115 	 * No need to re-evaluate the configuration for events on interfaces
1116 	 * that do not support ALTQ, as it's not possible for such
1117 	 * interfaces to be part of the configuration.
1118 	 */
1119 	if (!ALTQ_IS_READY(&ifp->if_snd))
1120 		return;
1121 
1122 	/* Interrupt userland queue modifications */
1123 	if (V_altqs_inactive_open)
1124 		pf_rollback_altq(V_ticket_altqs_inactive);
1125 
1126 	/* Start new altq ruleset */
1127 	if (pf_begin_altq(&ticket))
1128 		return;
1129 
1130 	/* Copy the current active set */
1131 	TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
1132 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1133 		if (a2 == NULL) {
1134 			error = ENOMEM;
1135 			break;
1136 		}
1137 		bcopy(a1, a2, sizeof(struct pf_altq));
1138 
1139 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1140 		if (error)
1141 			break;
1142 
1143 		TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
1144 	}
1145 	if (error)
1146 		goto out;
1147 	TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
1148 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1149 		if (a2 == NULL) {
1150 			error = ENOMEM;
1151 			break;
1152 		}
1153 		bcopy(a1, a2, sizeof(struct pf_altq));
1154 
1155 		if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
1156 			error = EBUSY;
1157 			free(a2, M_PFALTQ);
1158 			break;
1159 		}
1160 		a2->altq_disc = NULL;
1161 		TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
1162 			if (strncmp(a3->ifname, a2->ifname,
1163 				IFNAMSIZ) == 0) {
1164 				a2->altq_disc = a3->altq_disc;
1165 				break;
1166 			}
1167 		}
1168 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1169 		if (error)
1170 			break;
1171 
1172 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
1173 	}
1174 
1175 out:
1176 	if (error != 0)
1177 		pf_rollback_altq(ticket);
1178 	else
1179 		pf_commit_altq(ticket);
1180 }
1181 #endif /* ALTQ */
1182 
1183 static struct pf_krule_global *
1184 pf_rule_tree_alloc(int flags)
1185 {
1186 	struct pf_krule_global *tree;
1187 
1188 	tree = malloc(sizeof(struct pf_krule_global), M_TEMP, flags);
1189 	if (tree == NULL)
1190 		return (NULL);
1191 	RB_INIT(tree);
1192 	return (tree);
1193 }
1194 
1195 static void
1196 pf_rule_tree_free(struct pf_krule_global *tree)
1197 {
1198 
1199 	free(tree, M_TEMP);
1200 }
1201 
1202 static int
1203 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
1204 {
1205 	struct pf_krule_global *tree;
1206 	struct pf_kruleset	*rs;
1207 	struct pf_krule		*rule;
1208 
1209 	PF_RULES_WASSERT();
1210 
1211 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1212 		return (EINVAL);
1213 	tree = pf_rule_tree_alloc(M_NOWAIT);
1214 	if (tree == NULL)
1215 		return (ENOMEM);
1216 	rs = pf_find_or_create_kruleset(anchor);
1217 	if (rs == NULL) {
1218 		free(tree, M_TEMP);
1219 		return (EINVAL);
1220 	}
1221 	pf_rule_tree_free(rs->rules[rs_num].inactive.tree);
1222 	rs->rules[rs_num].inactive.tree = tree;
1223 
1224 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1225 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1226 		rs->rules[rs_num].inactive.rcount--;
1227 	}
1228 	*ticket = ++rs->rules[rs_num].inactive.ticket;
1229 	rs->rules[rs_num].inactive.open = 1;
1230 	return (0);
1231 }
1232 
1233 static int
1234 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
1235 {
1236 	struct pf_kruleset	*rs;
1237 	struct pf_krule		*rule;
1238 
1239 	PF_RULES_WASSERT();
1240 
1241 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1242 		return (EINVAL);
1243 	rs = pf_find_kruleset(anchor);
1244 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1245 	    rs->rules[rs_num].inactive.ticket != ticket)
1246 		return (0);
1247 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1248 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1249 		rs->rules[rs_num].inactive.rcount--;
1250 	}
1251 	rs->rules[rs_num].inactive.open = 0;
1252 	return (0);
1253 }
1254 
1255 #define PF_MD5_UPD(st, elm)						\
1256 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
1257 
1258 #define PF_MD5_UPD_STR(st, elm)						\
1259 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
1260 
1261 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
1262 		(stor) = htonl((st)->elm);				\
1263 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
1264 } while (0)
1265 
1266 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
1267 		(stor) = htons((st)->elm);				\
1268 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
1269 } while (0)
1270 
1271 static void
1272 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
1273 {
1274 	PF_MD5_UPD(pfr, addr.type);
1275 	switch (pfr->addr.type) {
1276 		case PF_ADDR_DYNIFTL:
1277 			PF_MD5_UPD(pfr, addr.v.ifname);
1278 			PF_MD5_UPD(pfr, addr.iflags);
1279 			break;
1280 		case PF_ADDR_TABLE:
1281 			PF_MD5_UPD(pfr, addr.v.tblname);
1282 			break;
1283 		case PF_ADDR_ADDRMASK:
1284 			/* XXX ignore af? */
1285 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
1286 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
1287 			break;
1288 	}
1289 
1290 	PF_MD5_UPD(pfr, port[0]);
1291 	PF_MD5_UPD(pfr, port[1]);
1292 	PF_MD5_UPD(pfr, neg);
1293 	PF_MD5_UPD(pfr, port_op);
1294 }
1295 
1296 static void
1297 pf_hash_rule_rolling(MD5_CTX *ctx, struct pf_krule *rule)
1298 {
1299 	u_int16_t x;
1300 	u_int32_t y;
1301 
1302 	pf_hash_rule_addr(ctx, &rule->src);
1303 	pf_hash_rule_addr(ctx, &rule->dst);
1304 	for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++)
1305 		PF_MD5_UPD_STR(rule, label[i]);
1306 	PF_MD5_UPD_STR(rule, ifname);
1307 	PF_MD5_UPD_STR(rule, match_tagname);
1308 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
1309 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
1310 	PF_MD5_UPD_HTONL(rule, prob, y);
1311 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
1312 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1313 	PF_MD5_UPD(rule, uid.op);
1314 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1315 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1316 	PF_MD5_UPD(rule, gid.op);
1317 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
1318 	PF_MD5_UPD(rule, action);
1319 	PF_MD5_UPD(rule, direction);
1320 	PF_MD5_UPD(rule, af);
1321 	PF_MD5_UPD(rule, quick);
1322 	PF_MD5_UPD(rule, ifnot);
1323 	PF_MD5_UPD(rule, match_tag_not);
1324 	PF_MD5_UPD(rule, natpass);
1325 	PF_MD5_UPD(rule, keep_state);
1326 	PF_MD5_UPD(rule, proto);
1327 	PF_MD5_UPD(rule, type);
1328 	PF_MD5_UPD(rule, code);
1329 	PF_MD5_UPD(rule, flags);
1330 	PF_MD5_UPD(rule, flagset);
1331 	PF_MD5_UPD(rule, allow_opts);
1332 	PF_MD5_UPD(rule, rt);
1333 	PF_MD5_UPD(rule, tos);
1334 	PF_MD5_UPD(rule, scrub_flags);
1335 	PF_MD5_UPD(rule, min_ttl);
1336 	PF_MD5_UPD(rule, set_tos);
1337 	if (rule->anchor != NULL)
1338 		PF_MD5_UPD_STR(rule, anchor->path);
1339 }
1340 
1341 static void
1342 pf_hash_rule(struct pf_krule *rule)
1343 {
1344 	MD5_CTX		ctx;
1345 
1346 	MD5Init(&ctx);
1347 	pf_hash_rule_rolling(&ctx, rule);
1348 	MD5Final(rule->md5sum, &ctx);
1349 }
1350 
1351 static int
1352 pf_krule_compare(struct pf_krule *a, struct pf_krule *b)
1353 {
1354 
1355 	return (memcmp(a->md5sum, b->md5sum, PF_MD5_DIGEST_LENGTH));
1356 }
1357 
1358 static int
1359 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1360 {
1361 	struct pf_kruleset	*rs;
1362 	struct pf_krule		*rule, **old_array, *old_rule;
1363 	struct pf_krulequeue	*old_rules;
1364 	struct pf_krule_global  *old_tree;
1365 	int			 error;
1366 	u_int32_t		 old_rcount;
1367 
1368 	PF_RULES_WASSERT();
1369 
1370 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1371 		return (EINVAL);
1372 	rs = pf_find_kruleset(anchor);
1373 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1374 	    ticket != rs->rules[rs_num].inactive.ticket)
1375 		return (EBUSY);
1376 
1377 	/* Calculate checksum for the main ruleset */
1378 	if (rs == &pf_main_ruleset) {
1379 		error = pf_setup_pfsync_matching(rs);
1380 		if (error != 0)
1381 			return (error);
1382 	}
1383 
1384 	/* Swap rules, keep the old. */
1385 	old_rules = rs->rules[rs_num].active.ptr;
1386 	old_rcount = rs->rules[rs_num].active.rcount;
1387 	old_array = rs->rules[rs_num].active.ptr_array;
1388 	old_tree = rs->rules[rs_num].active.tree;
1389 
1390 	rs->rules[rs_num].active.ptr =
1391 	    rs->rules[rs_num].inactive.ptr;
1392 	rs->rules[rs_num].active.ptr_array =
1393 	    rs->rules[rs_num].inactive.ptr_array;
1394 	rs->rules[rs_num].active.tree =
1395 	    rs->rules[rs_num].inactive.tree;
1396 	rs->rules[rs_num].active.rcount =
1397 	    rs->rules[rs_num].inactive.rcount;
1398 
1399 	/* Attempt to preserve counter information. */
1400 	if (V_pf_status.keep_counters && old_tree != NULL) {
1401 		TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr,
1402 		    entries) {
1403 			old_rule = RB_FIND(pf_krule_global, old_tree, rule);
1404 			if (old_rule == NULL) {
1405 				continue;
1406 			}
1407 			pf_counter_u64_critical_enter();
1408 			pf_counter_u64_add_protected(&rule->evaluations,
1409 			    pf_counter_u64_fetch(&old_rule->evaluations));
1410 			pf_counter_u64_add_protected(&rule->packets[0],
1411 			    pf_counter_u64_fetch(&old_rule->packets[0]));
1412 			pf_counter_u64_add_protected(&rule->packets[1],
1413 			    pf_counter_u64_fetch(&old_rule->packets[1]));
1414 			pf_counter_u64_add_protected(&rule->bytes[0],
1415 			    pf_counter_u64_fetch(&old_rule->bytes[0]));
1416 			pf_counter_u64_add_protected(&rule->bytes[1],
1417 			    pf_counter_u64_fetch(&old_rule->bytes[1]));
1418 			pf_counter_u64_critical_exit();
1419 		}
1420 	}
1421 
1422 	rs->rules[rs_num].inactive.ptr = old_rules;
1423 	rs->rules[rs_num].inactive.ptr_array = old_array;
1424 	rs->rules[rs_num].inactive.tree = NULL; /* important for pf_ioctl_addrule */
1425 	rs->rules[rs_num].inactive.rcount = old_rcount;
1426 
1427 	rs->rules[rs_num].active.ticket =
1428 	    rs->rules[rs_num].inactive.ticket;
1429 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1430 
1431 	/* Purge the old rule list. */
1432 	PF_UNLNKDRULES_LOCK();
1433 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1434 		pf_unlink_rule_locked(old_rules, rule);
1435 	PF_UNLNKDRULES_UNLOCK();
1436 	if (rs->rules[rs_num].inactive.ptr_array)
1437 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1438 	rs->rules[rs_num].inactive.ptr_array = NULL;
1439 	rs->rules[rs_num].inactive.rcount = 0;
1440 	rs->rules[rs_num].inactive.open = 0;
1441 	pf_remove_if_empty_kruleset(rs);
1442 	free(old_tree, M_TEMP);
1443 
1444 	return (0);
1445 }
1446 
1447 static int
1448 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1449 {
1450 	MD5_CTX			 ctx;
1451 	struct pf_krule		*rule;
1452 	int			 rs_cnt;
1453 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
1454 
1455 	MD5Init(&ctx);
1456 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1457 		/* XXX PF_RULESET_SCRUB as well? */
1458 		if (rs_cnt == PF_RULESET_SCRUB)
1459 			continue;
1460 
1461 		if (rs->rules[rs_cnt].inactive.ptr_array)
1462 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1463 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
1464 
1465 		if (rs->rules[rs_cnt].inactive.rcount) {
1466 			rs->rules[rs_cnt].inactive.ptr_array =
1467 			    mallocarray(rs->rules[rs_cnt].inactive.rcount,
1468 			    sizeof(struct pf_rule **),
1469 			    M_TEMP, M_NOWAIT);
1470 
1471 			if (!rs->rules[rs_cnt].inactive.ptr_array)
1472 				return (ENOMEM);
1473 		}
1474 
1475 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1476 		    entries) {
1477 			pf_hash_rule_rolling(&ctx, rule);
1478 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1479 		}
1480 	}
1481 
1482 	MD5Final(digest, &ctx);
1483 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1484 	return (0);
1485 }
1486 
1487 static int
1488 pf_eth_addr_setup(struct pf_keth_ruleset *ruleset, struct pf_addr_wrap *addr)
1489 {
1490 	int error = 0;
1491 
1492 	switch (addr->type) {
1493 	case PF_ADDR_TABLE:
1494 		addr->p.tbl = pfr_eth_attach_table(ruleset, addr->v.tblname);
1495 		if (addr->p.tbl == NULL)
1496 			error = ENOMEM;
1497 		break;
1498 	default:
1499 		error = EINVAL;
1500 	}
1501 
1502 	return (error);
1503 }
1504 
1505 static int
1506 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1507     sa_family_t af)
1508 {
1509 	int error = 0;
1510 
1511 	switch (addr->type) {
1512 	case PF_ADDR_TABLE:
1513 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1514 		if (addr->p.tbl == NULL)
1515 			error = ENOMEM;
1516 		break;
1517 	case PF_ADDR_DYNIFTL:
1518 		error = pfi_dynaddr_setup(addr, af);
1519 		break;
1520 	}
1521 
1522 	return (error);
1523 }
1524 
1525 static void
1526 pf_addr_copyout(struct pf_addr_wrap *addr)
1527 {
1528 
1529 	switch (addr->type) {
1530 	case PF_ADDR_DYNIFTL:
1531 		pfi_dynaddr_copyout(addr);
1532 		break;
1533 	case PF_ADDR_TABLE:
1534 		pf_tbladdr_copyout(addr);
1535 		break;
1536 	}
1537 }
1538 
1539 static void
1540 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
1541 {
1542 	int	secs = time_uptime, diff;
1543 
1544 	bzero(out, sizeof(struct pf_src_node));
1545 
1546 	bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
1547 	bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
1548 
1549 	if (in->rule.ptr != NULL)
1550 		out->rule.nr = in->rule.ptr->nr;
1551 
1552 	for (int i = 0; i < 2; i++) {
1553 		out->bytes[i] = counter_u64_fetch(in->bytes[i]);
1554 		out->packets[i] = counter_u64_fetch(in->packets[i]);
1555 	}
1556 
1557 	out->states = in->states;
1558 	out->conn = in->conn;
1559 	out->af = in->af;
1560 	out->ruletype = in->ruletype;
1561 
1562 	out->creation = secs - in->creation;
1563 	if (out->expire > secs)
1564 		out->expire -= secs;
1565 	else
1566 		out->expire = 0;
1567 
1568 	/* Adjust the connection rate estimate. */
1569 	diff = secs - in->conn_rate.last;
1570 	if (diff >= in->conn_rate.seconds)
1571 		out->conn_rate.count = 0;
1572 	else
1573 		out->conn_rate.count -=
1574 		    in->conn_rate.count * diff /
1575 		    in->conn_rate.seconds;
1576 }
1577 
1578 #ifdef ALTQ
1579 /*
1580  * Handle export of struct pf_kaltq to user binaries that may be using any
1581  * version of struct pf_altq.
1582  */
1583 static int
1584 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1585 {
1586 	u_int32_t version;
1587 
1588 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1589 		version = 0;
1590 	else
1591 		version = pa->version;
1592 
1593 	if (version > PFIOC_ALTQ_VERSION)
1594 		return (EINVAL);
1595 
1596 #define ASSIGN(x) exported_q->x = q->x
1597 #define COPY(x) \
1598 	bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1599 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1600 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1601 
1602 	switch (version) {
1603 	case 0: {
1604 		struct pf_altq_v0 *exported_q =
1605 		    &((struct pfioc_altq_v0 *)pa)->altq;
1606 
1607 		COPY(ifname);
1608 
1609 		ASSIGN(scheduler);
1610 		ASSIGN(tbrsize);
1611 		exported_q->tbrsize = SATU16(q->tbrsize);
1612 		exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1613 
1614 		COPY(qname);
1615 		COPY(parent);
1616 		ASSIGN(parent_qid);
1617 		exported_q->bandwidth = SATU32(q->bandwidth);
1618 		ASSIGN(priority);
1619 		ASSIGN(local_flags);
1620 
1621 		ASSIGN(qlimit);
1622 		ASSIGN(flags);
1623 
1624 		if (q->scheduler == ALTQT_HFSC) {
1625 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1626 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1627 			    SATU32(q->pq_u.hfsc_opts.x)
1628 
1629 			ASSIGN_OPT_SATU32(rtsc_m1);
1630 			ASSIGN_OPT(rtsc_d);
1631 			ASSIGN_OPT_SATU32(rtsc_m2);
1632 
1633 			ASSIGN_OPT_SATU32(lssc_m1);
1634 			ASSIGN_OPT(lssc_d);
1635 			ASSIGN_OPT_SATU32(lssc_m2);
1636 
1637 			ASSIGN_OPT_SATU32(ulsc_m1);
1638 			ASSIGN_OPT(ulsc_d);
1639 			ASSIGN_OPT_SATU32(ulsc_m2);
1640 
1641 			ASSIGN_OPT(flags);
1642 
1643 #undef ASSIGN_OPT
1644 #undef ASSIGN_OPT_SATU32
1645 		} else
1646 			COPY(pq_u);
1647 
1648 		ASSIGN(qid);
1649 		break;
1650 	}
1651 	case 1:	{
1652 		struct pf_altq_v1 *exported_q =
1653 		    &((struct pfioc_altq_v1 *)pa)->altq;
1654 
1655 		COPY(ifname);
1656 
1657 		ASSIGN(scheduler);
1658 		ASSIGN(tbrsize);
1659 		ASSIGN(ifbandwidth);
1660 
1661 		COPY(qname);
1662 		COPY(parent);
1663 		ASSIGN(parent_qid);
1664 		ASSIGN(bandwidth);
1665 		ASSIGN(priority);
1666 		ASSIGN(local_flags);
1667 
1668 		ASSIGN(qlimit);
1669 		ASSIGN(flags);
1670 		COPY(pq_u);
1671 
1672 		ASSIGN(qid);
1673 		break;
1674 	}
1675 	default:
1676 		panic("%s: unhandled struct pfioc_altq version", __func__);
1677 		break;
1678 	}
1679 
1680 #undef ASSIGN
1681 #undef COPY
1682 #undef SATU16
1683 #undef SATU32
1684 
1685 	return (0);
1686 }
1687 
1688 /*
1689  * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1690  * that may be using any version of it.
1691  */
1692 static int
1693 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1694 {
1695 	u_int32_t version;
1696 
1697 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1698 		version = 0;
1699 	else
1700 		version = pa->version;
1701 
1702 	if (version > PFIOC_ALTQ_VERSION)
1703 		return (EINVAL);
1704 
1705 #define ASSIGN(x) q->x = imported_q->x
1706 #define COPY(x) \
1707 	bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1708 
1709 	switch (version) {
1710 	case 0: {
1711 		struct pf_altq_v0 *imported_q =
1712 		    &((struct pfioc_altq_v0 *)pa)->altq;
1713 
1714 		COPY(ifname);
1715 
1716 		ASSIGN(scheduler);
1717 		ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1718 		ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1719 
1720 		COPY(qname);
1721 		COPY(parent);
1722 		ASSIGN(parent_qid);
1723 		ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1724 		ASSIGN(priority);
1725 		ASSIGN(local_flags);
1726 
1727 		ASSIGN(qlimit);
1728 		ASSIGN(flags);
1729 
1730 		if (imported_q->scheduler == ALTQT_HFSC) {
1731 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1732 
1733 			/*
1734 			 * The m1 and m2 parameters are being copied from
1735 			 * 32-bit to 64-bit.
1736 			 */
1737 			ASSIGN_OPT(rtsc_m1);
1738 			ASSIGN_OPT(rtsc_d);
1739 			ASSIGN_OPT(rtsc_m2);
1740 
1741 			ASSIGN_OPT(lssc_m1);
1742 			ASSIGN_OPT(lssc_d);
1743 			ASSIGN_OPT(lssc_m2);
1744 
1745 			ASSIGN_OPT(ulsc_m1);
1746 			ASSIGN_OPT(ulsc_d);
1747 			ASSIGN_OPT(ulsc_m2);
1748 
1749 			ASSIGN_OPT(flags);
1750 
1751 #undef ASSIGN_OPT
1752 		} else
1753 			COPY(pq_u);
1754 
1755 		ASSIGN(qid);
1756 		break;
1757 	}
1758 	case 1: {
1759 		struct pf_altq_v1 *imported_q =
1760 		    &((struct pfioc_altq_v1 *)pa)->altq;
1761 
1762 		COPY(ifname);
1763 
1764 		ASSIGN(scheduler);
1765 		ASSIGN(tbrsize);
1766 		ASSIGN(ifbandwidth);
1767 
1768 		COPY(qname);
1769 		COPY(parent);
1770 		ASSIGN(parent_qid);
1771 		ASSIGN(bandwidth);
1772 		ASSIGN(priority);
1773 		ASSIGN(local_flags);
1774 
1775 		ASSIGN(qlimit);
1776 		ASSIGN(flags);
1777 		COPY(pq_u);
1778 
1779 		ASSIGN(qid);
1780 		break;
1781 	}
1782 	default:
1783 		panic("%s: unhandled struct pfioc_altq version", __func__);
1784 		break;
1785 	}
1786 
1787 #undef ASSIGN
1788 #undef COPY
1789 
1790 	return (0);
1791 }
1792 
1793 static struct pf_altq *
1794 pf_altq_get_nth_active(u_int32_t n)
1795 {
1796 	struct pf_altq		*altq;
1797 	u_int32_t		 nr;
1798 
1799 	nr = 0;
1800 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1801 		if (nr == n)
1802 			return (altq);
1803 		nr++;
1804 	}
1805 
1806 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1807 		if (nr == n)
1808 			return (altq);
1809 		nr++;
1810 	}
1811 
1812 	return (NULL);
1813 }
1814 #endif /* ALTQ */
1815 
1816 struct pf_krule *
1817 pf_krule_alloc(void)
1818 {
1819 	struct pf_krule *rule;
1820 
1821 	rule = malloc(sizeof(struct pf_krule), M_PFRULE, M_WAITOK | M_ZERO);
1822 	mtx_init(&rule->rpool.mtx, "pf_krule_pool", NULL, MTX_DEF);
1823 	rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
1824 	    M_WAITOK | M_ZERO);
1825 	return (rule);
1826 }
1827 
1828 void
1829 pf_krule_free(struct pf_krule *rule)
1830 {
1831 #ifdef PF_WANT_32_TO_64_COUNTER
1832 	bool wowned;
1833 #endif
1834 
1835 	if (rule == NULL)
1836 		return;
1837 
1838 #ifdef PF_WANT_32_TO_64_COUNTER
1839 	if (rule->allrulelinked) {
1840 		wowned = PF_RULES_WOWNED();
1841 		if (!wowned)
1842 			PF_RULES_WLOCK();
1843 		LIST_REMOVE(rule, allrulelist);
1844 		V_pf_allrulecount--;
1845 		if (!wowned)
1846 			PF_RULES_WUNLOCK();
1847 	}
1848 #endif
1849 
1850 	pf_counter_u64_deinit(&rule->evaluations);
1851 	for (int i = 0; i < 2; i++) {
1852 		pf_counter_u64_deinit(&rule->packets[i]);
1853 		pf_counter_u64_deinit(&rule->bytes[i]);
1854 	}
1855 	counter_u64_free(rule->states_cur);
1856 	counter_u64_free(rule->states_tot);
1857 	counter_u64_free(rule->src_nodes);
1858 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
1859 
1860 	mtx_destroy(&rule->rpool.mtx);
1861 	free(rule, M_PFRULE);
1862 }
1863 
1864 static void
1865 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool,
1866     struct pf_pooladdr *pool)
1867 {
1868 
1869 	bzero(pool, sizeof(*pool));
1870 	bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr));
1871 	strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname));
1872 }
1873 
1874 static int
1875 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool,
1876     struct pf_kpooladdr *kpool)
1877 {
1878 	int ret;
1879 
1880 	bzero(kpool, sizeof(*kpool));
1881 	bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr));
1882 	ret = pf_user_strcpy(kpool->ifname, pool->ifname,
1883 	    sizeof(kpool->ifname));
1884 	return (ret);
1885 }
1886 
1887 static void
1888 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool)
1889 {
1890 	_Static_assert(sizeof(pool->key) == sizeof(kpool->key), "");
1891 	_Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), "");
1892 
1893 	bcopy(&pool->key, &kpool->key, sizeof(kpool->key));
1894 	bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter));
1895 
1896 	kpool->tblidx = pool->tblidx;
1897 	kpool->proxy_port[0] = pool->proxy_port[0];
1898 	kpool->proxy_port[1] = pool->proxy_port[1];
1899 	kpool->opts = pool->opts;
1900 }
1901 
1902 static int
1903 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
1904 {
1905 	int ret;
1906 
1907 #ifndef INET
1908 	if (rule->af == AF_INET) {
1909 		return (EAFNOSUPPORT);
1910 	}
1911 #endif /* INET */
1912 #ifndef INET6
1913 	if (rule->af == AF_INET6) {
1914 		return (EAFNOSUPPORT);
1915 	}
1916 #endif /* INET6 */
1917 
1918 	ret = pf_check_rule_addr(&rule->src);
1919 	if (ret != 0)
1920 		return (ret);
1921 	ret = pf_check_rule_addr(&rule->dst);
1922 	if (ret != 0)
1923 		return (ret);
1924 
1925 	bcopy(&rule->src, &krule->src, sizeof(rule->src));
1926 	bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
1927 
1928 	ret = pf_user_strcpy(krule->label[0], rule->label, sizeof(rule->label));
1929 	if (ret != 0)
1930 		return (ret);
1931 	ret = pf_user_strcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
1932 	if (ret != 0)
1933 		return (ret);
1934 	ret = pf_user_strcpy(krule->qname, rule->qname, sizeof(rule->qname));
1935 	if (ret != 0)
1936 		return (ret);
1937 	ret = pf_user_strcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
1938 	if (ret != 0)
1939 		return (ret);
1940 	ret = pf_user_strcpy(krule->tagname, rule->tagname,
1941 	    sizeof(rule->tagname));
1942 	if (ret != 0)
1943 		return (ret);
1944 	ret = pf_user_strcpy(krule->match_tagname, rule->match_tagname,
1945 	    sizeof(rule->match_tagname));
1946 	if (ret != 0)
1947 		return (ret);
1948 	ret = pf_user_strcpy(krule->overload_tblname, rule->overload_tblname,
1949 	    sizeof(rule->overload_tblname));
1950 	if (ret != 0)
1951 		return (ret);
1952 
1953 	pf_pool_to_kpool(&rule->rpool, &krule->rpool);
1954 
1955 	/* Don't allow userspace to set evaluations, packets or bytes. */
1956 	/* kif, anchor, overload_tbl are not copied over. */
1957 
1958 	krule->os_fingerprint = rule->os_fingerprint;
1959 
1960 	krule->rtableid = rule->rtableid;
1961 	bcopy(rule->timeout, krule->timeout, sizeof(krule->timeout));
1962 	krule->max_states = rule->max_states;
1963 	krule->max_src_nodes = rule->max_src_nodes;
1964 	krule->max_src_states = rule->max_src_states;
1965 	krule->max_src_conn = rule->max_src_conn;
1966 	krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
1967 	krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
1968 	krule->qid = rule->qid;
1969 	krule->pqid = rule->pqid;
1970 	krule->nr = rule->nr;
1971 	krule->prob = rule->prob;
1972 	krule->cuid = rule->cuid;
1973 	krule->cpid = rule->cpid;
1974 
1975 	krule->return_icmp = rule->return_icmp;
1976 	krule->return_icmp6 = rule->return_icmp6;
1977 	krule->max_mss = rule->max_mss;
1978 	krule->tag = rule->tag;
1979 	krule->match_tag = rule->match_tag;
1980 	krule->scrub_flags = rule->scrub_flags;
1981 
1982 	bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
1983 	bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
1984 
1985 	krule->rule_flag = rule->rule_flag;
1986 	krule->action = rule->action;
1987 	krule->direction = rule->direction;
1988 	krule->log = rule->log;
1989 	krule->logif = rule->logif;
1990 	krule->quick = rule->quick;
1991 	krule->ifnot = rule->ifnot;
1992 	krule->match_tag_not = rule->match_tag_not;
1993 	krule->natpass = rule->natpass;
1994 
1995 	krule->keep_state = rule->keep_state;
1996 	krule->af = rule->af;
1997 	krule->proto = rule->proto;
1998 	krule->type = rule->type;
1999 	krule->code = rule->code;
2000 	krule->flags = rule->flags;
2001 	krule->flagset = rule->flagset;
2002 	krule->min_ttl = rule->min_ttl;
2003 	krule->allow_opts = rule->allow_opts;
2004 	krule->rt = rule->rt;
2005 	krule->return_ttl = rule->return_ttl;
2006 	krule->tos = rule->tos;
2007 	krule->set_tos = rule->set_tos;
2008 
2009 	krule->flush = rule->flush;
2010 	krule->prio = rule->prio;
2011 	krule->set_prio[0] = rule->set_prio[0];
2012 	krule->set_prio[1] = rule->set_prio[1];
2013 
2014 	bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
2015 
2016 	return (0);
2017 }
2018 
2019 static int
2020 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket,
2021     uint32_t pool_ticket, const char *anchor, const char *anchor_call,
2022     struct thread *td)
2023 {
2024 	struct pf_kruleset	*ruleset;
2025 	struct pf_krule		*tail;
2026 	struct pf_kpooladdr	*pa;
2027 	struct pfi_kkif		*kif = NULL;
2028 	int			 rs_num;
2029 	int			 error = 0;
2030 
2031 	if ((rule->return_icmp >> 8) > ICMP_MAXTYPE) {
2032 		error = EINVAL;
2033 		goto errout_unlocked;
2034 	}
2035 
2036 #define	ERROUT(x)	ERROUT_FUNCTION(errout, x)
2037 
2038 	if (rule->ifname[0])
2039 		kif = pf_kkif_create(M_WAITOK);
2040 	pf_counter_u64_init(&rule->evaluations, M_WAITOK);
2041 	for (int i = 0; i < 2; i++) {
2042 		pf_counter_u64_init(&rule->packets[i], M_WAITOK);
2043 		pf_counter_u64_init(&rule->bytes[i], M_WAITOK);
2044 	}
2045 	rule->states_cur = counter_u64_alloc(M_WAITOK);
2046 	rule->states_tot = counter_u64_alloc(M_WAITOK);
2047 	rule->src_nodes = counter_u64_alloc(M_WAITOK);
2048 	rule->cuid = td->td_ucred->cr_ruid;
2049 	rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
2050 	TAILQ_INIT(&rule->rpool.list);
2051 
2052 	PF_CONFIG_LOCK();
2053 	PF_RULES_WLOCK();
2054 #ifdef PF_WANT_32_TO_64_COUNTER
2055 	LIST_INSERT_HEAD(&V_pf_allrulelist, rule, allrulelist);
2056 	MPASS(!rule->allrulelinked);
2057 	rule->allrulelinked = true;
2058 	V_pf_allrulecount++;
2059 #endif
2060 	ruleset = pf_find_kruleset(anchor);
2061 	if (ruleset == NULL)
2062 		ERROUT(EINVAL);
2063 	rs_num = pf_get_ruleset_number(rule->action);
2064 	if (rs_num >= PF_RULESET_MAX)
2065 		ERROUT(EINVAL);
2066 	if (ticket != ruleset->rules[rs_num].inactive.ticket) {
2067 		DPFPRINTF(PF_DEBUG_MISC,
2068 		    ("ticket: %d != [%d]%d\n", ticket, rs_num,
2069 		    ruleset->rules[rs_num].inactive.ticket));
2070 		ERROUT(EBUSY);
2071 	}
2072 	if (pool_ticket != V_ticket_pabuf) {
2073 		DPFPRINTF(PF_DEBUG_MISC,
2074 		    ("pool_ticket: %d != %d\n", pool_ticket,
2075 		    V_ticket_pabuf));
2076 		ERROUT(EBUSY);
2077 	}
2078 	/*
2079 	 * XXXMJG hack: there is no mechanism to ensure they started the
2080 	 * transaction. Ticket checked above may happen to match by accident,
2081 	 * even if nobody called DIOCXBEGIN, let alone this process.
2082 	 * Partially work around it by checking if the RB tree got allocated,
2083 	 * see pf_begin_rules.
2084 	 */
2085 	if (ruleset->rules[rs_num].inactive.tree == NULL) {
2086 		ERROUT(EINVAL);
2087 	}
2088 
2089 	tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
2090 	    pf_krulequeue);
2091 	if (tail)
2092 		rule->nr = tail->nr + 1;
2093 	else
2094 		rule->nr = 0;
2095 	if (rule->ifname[0]) {
2096 		rule->kif = pfi_kkif_attach(kif, rule->ifname);
2097 		kif = NULL;
2098 		pfi_kkif_ref(rule->kif);
2099 	} else
2100 		rule->kif = NULL;
2101 
2102 	if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
2103 		error = EBUSY;
2104 
2105 #ifdef ALTQ
2106 	/* set queue IDs */
2107 	if (rule->qname[0] != 0) {
2108 		if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2109 			error = EBUSY;
2110 		else if (rule->pqname[0] != 0) {
2111 			if ((rule->pqid =
2112 			    pf_qname2qid(rule->pqname)) == 0)
2113 				error = EBUSY;
2114 		} else
2115 			rule->pqid = rule->qid;
2116 	}
2117 #endif
2118 	if (rule->tagname[0])
2119 		if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2120 			error = EBUSY;
2121 	if (rule->match_tagname[0])
2122 		if ((rule->match_tag =
2123 		    pf_tagname2tag(rule->match_tagname)) == 0)
2124 			error = EBUSY;
2125 	if (rule->rt && !rule->direction)
2126 		error = EINVAL;
2127 	if (!rule->log)
2128 		rule->logif = 0;
2129 	if (rule->logif >= PFLOGIFS_MAX)
2130 		error = EINVAL;
2131 	if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
2132 		error = ENOMEM;
2133 	if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
2134 		error = ENOMEM;
2135 	if (pf_kanchor_setup(rule, ruleset, anchor_call))
2136 		error = EINVAL;
2137 	if (rule->scrub_flags & PFSTATE_SETPRIO &&
2138 	    (rule->set_prio[0] > PF_PRIO_MAX ||
2139 	    rule->set_prio[1] > PF_PRIO_MAX))
2140 		error = EINVAL;
2141 	TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
2142 		if (pa->addr.type == PF_ADDR_TABLE) {
2143 			pa->addr.p.tbl = pfr_attach_table(ruleset,
2144 			    pa->addr.v.tblname);
2145 			if (pa->addr.p.tbl == NULL)
2146 				error = ENOMEM;
2147 		}
2148 
2149 	rule->overload_tbl = NULL;
2150 	if (rule->overload_tblname[0]) {
2151 		if ((rule->overload_tbl = pfr_attach_table(ruleset,
2152 		    rule->overload_tblname)) == NULL)
2153 			error = EINVAL;
2154 		else
2155 			rule->overload_tbl->pfrkt_flags |=
2156 			    PFR_TFLAG_ACTIVE;
2157 	}
2158 
2159 	pf_mv_kpool(&V_pf_pabuf, &rule->rpool.list);
2160 	if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
2161 	    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
2162 	    (rule->rt > PF_NOPFROUTE)) &&
2163 	    (TAILQ_FIRST(&rule->rpool.list) == NULL))
2164 		error = EINVAL;
2165 
2166 	if (error) {
2167 		pf_free_rule(rule);
2168 		rule = NULL;
2169 		ERROUT(error);
2170 	}
2171 
2172 	rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
2173 	TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
2174 	    rule, entries);
2175 	ruleset->rules[rs_num].inactive.rcount++;
2176 
2177 	PF_RULES_WUNLOCK();
2178 	pf_hash_rule(rule);
2179 	if (RB_INSERT(pf_krule_global, ruleset->rules[rs_num].inactive.tree, rule) != NULL) {
2180 		PF_RULES_WLOCK();
2181 		TAILQ_REMOVE(ruleset->rules[rs_num].inactive.ptr, rule, entries);
2182 		ruleset->rules[rs_num].inactive.rcount--;
2183 		pf_free_rule(rule);
2184 		rule = NULL;
2185 		ERROUT(EEXIST);
2186 	}
2187 	PF_CONFIG_UNLOCK();
2188 
2189 	return (0);
2190 
2191 #undef ERROUT
2192 errout:
2193 	PF_RULES_WUNLOCK();
2194 	PF_CONFIG_UNLOCK();
2195 errout_unlocked:
2196 	pf_kkif_free(kif);
2197 	pf_krule_free(rule);
2198 	return (error);
2199 }
2200 
2201 static bool
2202 pf_label_match(const struct pf_krule *rule, const char *label)
2203 {
2204 	int i = 0;
2205 
2206 	while (*rule->label[i]) {
2207 		if (strcmp(rule->label[i], label) == 0)
2208 			return (true);
2209 		i++;
2210 	}
2211 
2212 	return (false);
2213 }
2214 
2215 static unsigned int
2216 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir)
2217 {
2218 	struct pf_kstate *s;
2219 	int more = 0;
2220 
2221 	s = pf_find_state_all(key, dir, &more);
2222 	if (s == NULL)
2223 		return (0);
2224 
2225 	if (more) {
2226 		PF_STATE_UNLOCK(s);
2227 		return (0);
2228 	}
2229 
2230 	pf_unlink_state(s);
2231 	return (1);
2232 }
2233 
2234 static int
2235 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih)
2236 {
2237 	struct pf_kstate	*s;
2238 	struct pf_state_key	*sk;
2239 	struct pf_addr		*srcaddr, *dstaddr;
2240 	struct pf_state_key_cmp	 match_key;
2241 	int			 idx, killed = 0;
2242 	unsigned int		 dir;
2243 	u_int16_t		 srcport, dstport;
2244 	struct pfi_kkif		*kif;
2245 
2246 relock_DIOCKILLSTATES:
2247 	PF_HASHROW_LOCK(ih);
2248 	LIST_FOREACH(s, &ih->states, entry) {
2249 		/* For floating states look at the original kif. */
2250 		kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
2251 
2252 		sk = s->key[PF_SK_WIRE];
2253 		if (s->direction == PF_OUT) {
2254 			srcaddr = &sk->addr[1];
2255 			dstaddr = &sk->addr[0];
2256 			srcport = sk->port[1];
2257 			dstport = sk->port[0];
2258 		} else {
2259 			srcaddr = &sk->addr[0];
2260 			dstaddr = &sk->addr[1];
2261 			srcport = sk->port[0];
2262 			dstport = sk->port[1];
2263 		}
2264 
2265 		if (psk->psk_af && sk->af != psk->psk_af)
2266 			continue;
2267 
2268 		if (psk->psk_proto && psk->psk_proto != sk->proto)
2269 			continue;
2270 
2271 		if (! PF_MATCHA(psk->psk_src.neg, &psk->psk_src.addr.v.a.addr,
2272 		    &psk->psk_src.addr.v.a.mask, srcaddr, sk->af))
2273 			continue;
2274 
2275 		if (! PF_MATCHA(psk->psk_dst.neg, &psk->psk_dst.addr.v.a.addr,
2276 		    &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af))
2277 			continue;
2278 
2279 		if (!  PF_MATCHA(psk->psk_rt_addr.neg,
2280 		    &psk->psk_rt_addr.addr.v.a.addr,
2281 		    &psk->psk_rt_addr.addr.v.a.mask,
2282 		    &s->rt_addr, sk->af))
2283 			continue;
2284 
2285 		if (psk->psk_src.port_op != 0 &&
2286 		    ! pf_match_port(psk->psk_src.port_op,
2287 		    psk->psk_src.port[0], psk->psk_src.port[1], srcport))
2288 			continue;
2289 
2290 		if (psk->psk_dst.port_op != 0 &&
2291 		    ! pf_match_port(psk->psk_dst.port_op,
2292 		    psk->psk_dst.port[0], psk->psk_dst.port[1], dstport))
2293 			continue;
2294 
2295 		if (psk->psk_label[0] &&
2296 		    ! pf_label_match(s->rule.ptr, psk->psk_label))
2297 			continue;
2298 
2299 		if (psk->psk_ifname[0] && strcmp(psk->psk_ifname,
2300 		    kif->pfik_name))
2301 			continue;
2302 
2303 		if (psk->psk_kill_match) {
2304 			/* Create the key to find matching states, with lock
2305 			 * held. */
2306 
2307 			bzero(&match_key, sizeof(match_key));
2308 
2309 			if (s->direction == PF_OUT) {
2310 				dir = PF_IN;
2311 				idx = PF_SK_STACK;
2312 			} else {
2313 				dir = PF_OUT;
2314 				idx = PF_SK_WIRE;
2315 			}
2316 
2317 			match_key.af = s->key[idx]->af;
2318 			match_key.proto = s->key[idx]->proto;
2319 			PF_ACPY(&match_key.addr[0],
2320 			    &s->key[idx]->addr[1], match_key.af);
2321 			match_key.port[0] = s->key[idx]->port[1];
2322 			PF_ACPY(&match_key.addr[1],
2323 			    &s->key[idx]->addr[0], match_key.af);
2324 			match_key.port[1] = s->key[idx]->port[0];
2325 		}
2326 
2327 		pf_unlink_state(s);
2328 		killed++;
2329 
2330 		if (psk->psk_kill_match)
2331 			killed += pf_kill_matching_state(&match_key, dir);
2332 
2333 		goto relock_DIOCKILLSTATES;
2334 	}
2335 	PF_HASHROW_UNLOCK(ih);
2336 
2337 	return (killed);
2338 }
2339 
2340 int
2341 pf_start(void)
2342 {
2343 	int error = 0;
2344 
2345 	sx_xlock(&V_pf_ioctl_lock);
2346 	if (V_pf_status.running)
2347 		error = EEXIST;
2348 	else {
2349 		hook_pf();
2350 		if (! TAILQ_EMPTY(V_pf_keth->active.rules))
2351 			hook_pf_eth();
2352 		V_pf_status.running = 1;
2353 		V_pf_status.since = time_second;
2354 		new_unrhdr64(&V_pf_stateid, time_second);
2355 
2356 		DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
2357 	}
2358 	sx_xunlock(&V_pf_ioctl_lock);
2359 
2360 	return (error);
2361 }
2362 
2363 int
2364 pf_stop(void)
2365 {
2366 	int error = 0;
2367 
2368 	sx_xlock(&V_pf_ioctl_lock);
2369 	if (!V_pf_status.running)
2370 		error = ENOENT;
2371 	else {
2372 		V_pf_status.running = 0;
2373 		dehook_pf();
2374 		dehook_pf_eth();
2375 		V_pf_status.since = time_second;
2376 		DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
2377 	}
2378 	sx_xunlock(&V_pf_ioctl_lock);
2379 
2380 	return (error);
2381 }
2382 
2383 static int
2384 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
2385 {
2386 	int			 error = 0;
2387 	PF_RULES_RLOCK_TRACKER;
2388 
2389 #define	ERROUT_IOCTL(target, x)					\
2390     do {								\
2391 	    error = (x);						\
2392 	    SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__);	\
2393 	    goto target;						\
2394     } while (0)
2395 
2396 
2397 	/* XXX keep in sync with switch() below */
2398 	if (securelevel_gt(td->td_ucred, 2))
2399 		switch (cmd) {
2400 		case DIOCGETRULES:
2401 		case DIOCGETRULENV:
2402 		case DIOCGETADDRS:
2403 		case DIOCGETADDR:
2404 		case DIOCGETSTATE:
2405 		case DIOCGETSTATENV:
2406 		case DIOCSETSTATUSIF:
2407 		case DIOCGETSTATUSNV:
2408 		case DIOCCLRSTATUS:
2409 		case DIOCNATLOOK:
2410 		case DIOCSETDEBUG:
2411 #ifdef COMPAT_FREEBSD14
2412 		case DIOCGETSTATES:
2413 		case DIOCGETSTATESV2:
2414 #endif
2415 		case DIOCGETTIMEOUT:
2416 		case DIOCCLRRULECTRS:
2417 		case DIOCGETLIMIT:
2418 		case DIOCGETALTQSV0:
2419 		case DIOCGETALTQSV1:
2420 		case DIOCGETALTQV0:
2421 		case DIOCGETALTQV1:
2422 		case DIOCGETQSTATSV0:
2423 		case DIOCGETQSTATSV1:
2424 		case DIOCGETRULESETS:
2425 		case DIOCGETRULESET:
2426 		case DIOCRGETTABLES:
2427 		case DIOCRGETTSTATS:
2428 		case DIOCRCLRTSTATS:
2429 		case DIOCRCLRADDRS:
2430 		case DIOCRADDADDRS:
2431 		case DIOCRDELADDRS:
2432 		case DIOCRSETADDRS:
2433 		case DIOCRGETADDRS:
2434 		case DIOCRGETASTATS:
2435 		case DIOCRCLRASTATS:
2436 		case DIOCRTSTADDRS:
2437 		case DIOCOSFPGET:
2438 		case DIOCGETSRCNODES:
2439 		case DIOCCLRSRCNODES:
2440 		case DIOCGETSYNCOOKIES:
2441 		case DIOCIGETIFACES:
2442 		case DIOCGIFSPEEDV0:
2443 		case DIOCGIFSPEEDV1:
2444 		case DIOCSETIFFLAG:
2445 		case DIOCCLRIFFLAG:
2446 		case DIOCGETETHRULES:
2447 		case DIOCGETETHRULE:
2448 		case DIOCGETETHRULESETS:
2449 		case DIOCGETETHRULESET:
2450 			break;
2451 		case DIOCRCLRTABLES:
2452 		case DIOCRADDTABLES:
2453 		case DIOCRDELTABLES:
2454 		case DIOCRSETTFLAGS:
2455 			if (((struct pfioc_table *)addr)->pfrio_flags &
2456 			    PFR_FLAG_DUMMY)
2457 				break; /* dummy operation ok */
2458 			return (EPERM);
2459 		default:
2460 			return (EPERM);
2461 		}
2462 
2463 	if (!(flags & FWRITE))
2464 		switch (cmd) {
2465 		case DIOCGETRULES:
2466 		case DIOCGETADDRS:
2467 		case DIOCGETADDR:
2468 		case DIOCGETSTATE:
2469 		case DIOCGETSTATENV:
2470 		case DIOCGETSTATUSNV:
2471 #ifdef COMPAT_FREEBSD14
2472 		case DIOCGETSTATES:
2473 		case DIOCGETSTATESV2:
2474 #endif
2475 		case DIOCGETTIMEOUT:
2476 		case DIOCGETLIMIT:
2477 		case DIOCGETALTQSV0:
2478 		case DIOCGETALTQSV1:
2479 		case DIOCGETALTQV0:
2480 		case DIOCGETALTQV1:
2481 		case DIOCGETQSTATSV0:
2482 		case DIOCGETQSTATSV1:
2483 		case DIOCGETRULESETS:
2484 		case DIOCGETRULESET:
2485 		case DIOCNATLOOK:
2486 		case DIOCRGETTABLES:
2487 		case DIOCRGETTSTATS:
2488 		case DIOCRGETADDRS:
2489 		case DIOCRGETASTATS:
2490 		case DIOCRTSTADDRS:
2491 		case DIOCOSFPGET:
2492 		case DIOCGETSRCNODES:
2493 		case DIOCGETSYNCOOKIES:
2494 		case DIOCIGETIFACES:
2495 		case DIOCGIFSPEEDV1:
2496 		case DIOCGIFSPEEDV0:
2497 		case DIOCGETRULENV:
2498 		case DIOCGETETHRULES:
2499 		case DIOCGETETHRULE:
2500 		case DIOCGETETHRULESETS:
2501 		case DIOCGETETHRULESET:
2502 			break;
2503 		case DIOCRCLRTABLES:
2504 		case DIOCRADDTABLES:
2505 		case DIOCRDELTABLES:
2506 		case DIOCRCLRTSTATS:
2507 		case DIOCRCLRADDRS:
2508 		case DIOCRADDADDRS:
2509 		case DIOCRDELADDRS:
2510 		case DIOCRSETADDRS:
2511 		case DIOCRSETTFLAGS:
2512 			if (((struct pfioc_table *)addr)->pfrio_flags &
2513 			    PFR_FLAG_DUMMY) {
2514 				flags |= FWRITE; /* need write lock for dummy */
2515 				break; /* dummy operation ok */
2516 			}
2517 			return (EACCES);
2518 		default:
2519 			return (EACCES);
2520 		}
2521 
2522 	CURVNET_SET(TD_TO_VNET(td));
2523 
2524 	switch (cmd) {
2525 #ifdef COMPAT_FREEBSD14
2526 	case DIOCSTART:
2527 		error = pf_start();
2528 		break;
2529 
2530 	case DIOCSTOP:
2531 		error = pf_stop();
2532 		break;
2533 #endif
2534 
2535 	case DIOCGETETHRULES: {
2536 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2537 		nvlist_t		*nvl;
2538 		void			*packed;
2539 		struct pf_keth_rule	*tail;
2540 		struct pf_keth_ruleset	*rs;
2541 		u_int32_t		 ticket, nr;
2542 		const char		*anchor = "";
2543 
2544 		nvl = NULL;
2545 		packed = NULL;
2546 
2547 #define	ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULES_error, x)
2548 
2549 		if (nv->len > pf_ioctl_maxcount)
2550 			ERROUT(ENOMEM);
2551 
2552 		/* Copy the request in */
2553 		packed = malloc(nv->len, M_NVLIST, M_WAITOK);
2554 		if (packed == NULL)
2555 			ERROUT(ENOMEM);
2556 
2557 		error = copyin(nv->data, packed, nv->len);
2558 		if (error)
2559 			ERROUT(error);
2560 
2561 		nvl = nvlist_unpack(packed, nv->len, 0);
2562 		if (nvl == NULL)
2563 			ERROUT(EBADMSG);
2564 
2565 		if (! nvlist_exists_string(nvl, "anchor"))
2566 			ERROUT(EBADMSG);
2567 
2568 		anchor = nvlist_get_string(nvl, "anchor");
2569 
2570 		rs = pf_find_keth_ruleset(anchor);
2571 
2572 		nvlist_destroy(nvl);
2573 		nvl = NULL;
2574 		free(packed, M_NVLIST);
2575 		packed = NULL;
2576 
2577 		if (rs == NULL)
2578 			ERROUT(ENOENT);
2579 
2580 		/* Reply */
2581 		nvl = nvlist_create(0);
2582 		if (nvl == NULL)
2583 			ERROUT(ENOMEM);
2584 
2585 		PF_RULES_RLOCK();
2586 
2587 		ticket = rs->active.ticket;
2588 		tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq);
2589 		if (tail)
2590 			nr = tail->nr + 1;
2591 		else
2592 			nr = 0;
2593 
2594 		PF_RULES_RUNLOCK();
2595 
2596 		nvlist_add_number(nvl, "ticket", ticket);
2597 		nvlist_add_number(nvl, "nr", nr);
2598 
2599 		packed = nvlist_pack(nvl, &nv->len);
2600 		if (packed == NULL)
2601 			ERROUT(ENOMEM);
2602 
2603 		if (nv->size == 0)
2604 			ERROUT(0);
2605 		else if (nv->size < nv->len)
2606 			ERROUT(ENOSPC);
2607 
2608 		error = copyout(packed, nv->data, nv->len);
2609 
2610 #undef ERROUT
2611 DIOCGETETHRULES_error:
2612 		free(packed, M_NVLIST);
2613 		nvlist_destroy(nvl);
2614 		break;
2615 	}
2616 
2617 	case DIOCGETETHRULE: {
2618 		struct epoch_tracker	 et;
2619 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2620 		nvlist_t		*nvl = NULL;
2621 		void			*nvlpacked = NULL;
2622 		struct pf_keth_rule	*rule = NULL;
2623 		struct pf_keth_ruleset	*rs;
2624 		u_int32_t		 ticket, nr;
2625 		bool			 clear = false;
2626 		const char		*anchor;
2627 
2628 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULE_error, x)
2629 
2630 		if (nv->len > pf_ioctl_maxcount)
2631 			ERROUT(ENOMEM);
2632 
2633 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2634 		if (nvlpacked == NULL)
2635 			ERROUT(ENOMEM);
2636 
2637 		error = copyin(nv->data, nvlpacked, nv->len);
2638 		if (error)
2639 			ERROUT(error);
2640 
2641 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2642 		if (nvl == NULL)
2643 			ERROUT(EBADMSG);
2644 		if (! nvlist_exists_number(nvl, "ticket"))
2645 			ERROUT(EBADMSG);
2646 		ticket = nvlist_get_number(nvl, "ticket");
2647 		if (! nvlist_exists_string(nvl, "anchor"))
2648 			ERROUT(EBADMSG);
2649 		anchor = nvlist_get_string(nvl, "anchor");
2650 
2651 		if (nvlist_exists_bool(nvl, "clear"))
2652 			clear = nvlist_get_bool(nvl, "clear");
2653 
2654 		if (clear && !(flags & FWRITE))
2655 			ERROUT(EACCES);
2656 
2657 		if (! nvlist_exists_number(nvl, "nr"))
2658 			ERROUT(EBADMSG);
2659 		nr = nvlist_get_number(nvl, "nr");
2660 
2661 		PF_RULES_RLOCK();
2662 		rs = pf_find_keth_ruleset(anchor);
2663 		if (rs == NULL) {
2664 			PF_RULES_RUNLOCK();
2665 			ERROUT(ENOENT);
2666 		}
2667 		if (ticket != rs->active.ticket) {
2668 			PF_RULES_RUNLOCK();
2669 			ERROUT(EBUSY);
2670 		}
2671 
2672 		nvlist_destroy(nvl);
2673 		nvl = NULL;
2674 		free(nvlpacked, M_NVLIST);
2675 		nvlpacked = NULL;
2676 
2677 		rule = TAILQ_FIRST(rs->active.rules);
2678 		while ((rule != NULL) && (rule->nr != nr))
2679 			rule = TAILQ_NEXT(rule, entries);
2680 		if (rule == NULL) {
2681 			PF_RULES_RUNLOCK();
2682 			ERROUT(ENOENT);
2683 		}
2684 		/* Make sure rule can't go away. */
2685 		NET_EPOCH_ENTER(et);
2686 		PF_RULES_RUNLOCK();
2687 		nvl = pf_keth_rule_to_nveth_rule(rule);
2688 		if (pf_keth_anchor_nvcopyout(rs, rule, nvl))
2689 			ERROUT(EBUSY);
2690 		NET_EPOCH_EXIT(et);
2691 		if (nvl == NULL)
2692 			ERROUT(ENOMEM);
2693 
2694 		nvlpacked = nvlist_pack(nvl, &nv->len);
2695 		if (nvlpacked == NULL)
2696 			ERROUT(ENOMEM);
2697 
2698 		if (nv->size == 0)
2699 			ERROUT(0);
2700 		else if (nv->size < nv->len)
2701 			ERROUT(ENOSPC);
2702 
2703 		error = copyout(nvlpacked, nv->data, nv->len);
2704 		if (error == 0 && clear) {
2705 			counter_u64_zero(rule->evaluations);
2706 			for (int i = 0; i < 2; i++) {
2707 				counter_u64_zero(rule->packets[i]);
2708 				counter_u64_zero(rule->bytes[i]);
2709 			}
2710 		}
2711 
2712 #undef ERROUT
2713 DIOCGETETHRULE_error:
2714 		free(nvlpacked, M_NVLIST);
2715 		nvlist_destroy(nvl);
2716 		break;
2717 	}
2718 
2719 	case DIOCADDETHRULE: {
2720 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2721 		nvlist_t		*nvl = NULL;
2722 		void			*nvlpacked = NULL;
2723 		struct pf_keth_rule	*rule = NULL, *tail = NULL;
2724 		struct pf_keth_ruleset	*ruleset = NULL;
2725 		struct pfi_kkif		*kif = NULL, *bridge_to_kif = NULL;
2726 		const char		*anchor = "", *anchor_call = "";
2727 
2728 #define ERROUT(x)	ERROUT_IOCTL(DIOCADDETHRULE_error, x)
2729 
2730 		if (nv->len > pf_ioctl_maxcount)
2731 			ERROUT(ENOMEM);
2732 
2733 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2734 		if (nvlpacked == NULL)
2735 			ERROUT(ENOMEM);
2736 
2737 		error = copyin(nv->data, nvlpacked, nv->len);
2738 		if (error)
2739 			ERROUT(error);
2740 
2741 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2742 		if (nvl == NULL)
2743 			ERROUT(EBADMSG);
2744 
2745 		if (! nvlist_exists_number(nvl, "ticket"))
2746 			ERROUT(EBADMSG);
2747 
2748 		if (nvlist_exists_string(nvl, "anchor"))
2749 			anchor = nvlist_get_string(nvl, "anchor");
2750 		if (nvlist_exists_string(nvl, "anchor_call"))
2751 			anchor_call = nvlist_get_string(nvl, "anchor_call");
2752 
2753 		ruleset = pf_find_keth_ruleset(anchor);
2754 		if (ruleset == NULL)
2755 			ERROUT(EINVAL);
2756 
2757 		if (nvlist_get_number(nvl, "ticket") !=
2758 		    ruleset->inactive.ticket) {
2759 			DPFPRINTF(PF_DEBUG_MISC,
2760 			    ("ticket: %d != %d\n",
2761 			    (u_int32_t)nvlist_get_number(nvl, "ticket"),
2762 			    ruleset->inactive.ticket));
2763 			ERROUT(EBUSY);
2764 		}
2765 
2766 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
2767 		if (rule == NULL)
2768 			ERROUT(ENOMEM);
2769 		rule->timestamp = NULL;
2770 
2771 		error = pf_nveth_rule_to_keth_rule(nvl, rule);
2772 		if (error != 0)
2773 			ERROUT(error);
2774 
2775 		if (rule->ifname[0])
2776 			kif = pf_kkif_create(M_WAITOK);
2777 		if (rule->bridge_to_name[0])
2778 			bridge_to_kif = pf_kkif_create(M_WAITOK);
2779 		rule->evaluations = counter_u64_alloc(M_WAITOK);
2780 		for (int i = 0; i < 2; i++) {
2781 			rule->packets[i] = counter_u64_alloc(M_WAITOK);
2782 			rule->bytes[i] = counter_u64_alloc(M_WAITOK);
2783 		}
2784 		rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
2785 		    M_WAITOK | M_ZERO);
2786 
2787 		PF_RULES_WLOCK();
2788 
2789 		if (rule->ifname[0]) {
2790 			rule->kif = pfi_kkif_attach(kif, rule->ifname);
2791 			pfi_kkif_ref(rule->kif);
2792 		} else
2793 			rule->kif = NULL;
2794 		if (rule->bridge_to_name[0]) {
2795 			rule->bridge_to = pfi_kkif_attach(bridge_to_kif,
2796 			    rule->bridge_to_name);
2797 			pfi_kkif_ref(rule->bridge_to);
2798 		} else
2799 			rule->bridge_to = NULL;
2800 
2801 #ifdef ALTQ
2802 		/* set queue IDs */
2803 		if (rule->qname[0] != 0) {
2804 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2805 				error = EBUSY;
2806 			else
2807 				rule->qid = rule->qid;
2808 		}
2809 #endif
2810 		if (rule->tagname[0])
2811 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2812 				error = EBUSY;
2813 		if (rule->match_tagname[0])
2814 			if ((rule->match_tag = pf_tagname2tag(
2815 			    rule->match_tagname)) == 0)
2816 				error = EBUSY;
2817 
2818 		if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE)
2819 			error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr);
2820 		if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE)
2821 			error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr);
2822 
2823 		if (error) {
2824 			pf_free_eth_rule(rule);
2825 			PF_RULES_WUNLOCK();
2826 			ERROUT(error);
2827 		}
2828 
2829 		if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) {
2830 			pf_free_eth_rule(rule);
2831 			PF_RULES_WUNLOCK();
2832 			ERROUT(EINVAL);
2833 		}
2834 
2835 		tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq);
2836 		if (tail)
2837 			rule->nr = tail->nr + 1;
2838 		else
2839 			rule->nr = 0;
2840 
2841 		TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries);
2842 
2843 		PF_RULES_WUNLOCK();
2844 
2845 #undef ERROUT
2846 DIOCADDETHRULE_error:
2847 		nvlist_destroy(nvl);
2848 		free(nvlpacked, M_NVLIST);
2849 		break;
2850 	}
2851 
2852 	case DIOCGETETHRULESETS: {
2853 		struct epoch_tracker	 et;
2854 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2855 		nvlist_t		*nvl = NULL;
2856 		void			*nvlpacked = NULL;
2857 		struct pf_keth_ruleset	*ruleset;
2858 		struct pf_keth_anchor	*anchor;
2859 		int			 nr = 0;
2860 
2861 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULESETS_error, x)
2862 
2863 		if (nv->len > pf_ioctl_maxcount)
2864 			ERROUT(ENOMEM);
2865 
2866 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2867 		if (nvlpacked == NULL)
2868 			ERROUT(ENOMEM);
2869 
2870 		error = copyin(nv->data, nvlpacked, nv->len);
2871 		if (error)
2872 			ERROUT(error);
2873 
2874 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2875 		if (nvl == NULL)
2876 			ERROUT(EBADMSG);
2877 		if (! nvlist_exists_string(nvl, "path"))
2878 			ERROUT(EBADMSG);
2879 
2880 		NET_EPOCH_ENTER(et);
2881 
2882 		if ((ruleset = pf_find_keth_ruleset(
2883 		    nvlist_get_string(nvl, "path"))) == NULL) {
2884 			NET_EPOCH_EXIT(et);
2885 			ERROUT(ENOENT);
2886 		}
2887 
2888 		if (ruleset->anchor == NULL) {
2889 			RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors)
2890 				if (anchor->parent == NULL)
2891 					nr++;
2892 		} else {
2893 			RB_FOREACH(anchor, pf_keth_anchor_node,
2894 			    &ruleset->anchor->children)
2895 				nr++;
2896 		}
2897 
2898 		NET_EPOCH_EXIT(et);
2899 
2900 		nvlist_destroy(nvl);
2901 		nvl = NULL;
2902 		free(nvlpacked, M_NVLIST);
2903 		nvlpacked = NULL;
2904 
2905 		nvl = nvlist_create(0);
2906 		if (nvl == NULL)
2907 			ERROUT(ENOMEM);
2908 
2909 		nvlist_add_number(nvl, "nr", nr);
2910 
2911 		nvlpacked = nvlist_pack(nvl, &nv->len);
2912 		if (nvlpacked == NULL)
2913 			ERROUT(ENOMEM);
2914 
2915 		if (nv->size == 0)
2916 			ERROUT(0);
2917 		else if (nv->size < nv->len)
2918 			ERROUT(ENOSPC);
2919 
2920 		error = copyout(nvlpacked, nv->data, nv->len);
2921 
2922 #undef ERROUT
2923 DIOCGETETHRULESETS_error:
2924 		free(nvlpacked, M_NVLIST);
2925 		nvlist_destroy(nvl);
2926 		break;
2927 	}
2928 
2929 	case DIOCGETETHRULESET: {
2930 		struct epoch_tracker	 et;
2931 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2932 		nvlist_t		*nvl = NULL;
2933 		void			*nvlpacked = NULL;
2934 		struct pf_keth_ruleset	*ruleset;
2935 		struct pf_keth_anchor	*anchor;
2936 		int			 nr = 0, req_nr = 0;
2937 		bool			 found = false;
2938 
2939 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULESET_error, x)
2940 
2941 		if (nv->len > pf_ioctl_maxcount)
2942 			ERROUT(ENOMEM);
2943 
2944 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2945 		if (nvlpacked == NULL)
2946 			ERROUT(ENOMEM);
2947 
2948 		error = copyin(nv->data, nvlpacked, nv->len);
2949 		if (error)
2950 			ERROUT(error);
2951 
2952 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2953 		if (nvl == NULL)
2954 			ERROUT(EBADMSG);
2955 		if (! nvlist_exists_string(nvl, "path"))
2956 			ERROUT(EBADMSG);
2957 		if (! nvlist_exists_number(nvl, "nr"))
2958 			ERROUT(EBADMSG);
2959 
2960 		req_nr = nvlist_get_number(nvl, "nr");
2961 
2962 		NET_EPOCH_ENTER(et);
2963 
2964 		if ((ruleset = pf_find_keth_ruleset(
2965 		    nvlist_get_string(nvl, "path"))) == NULL) {
2966 			NET_EPOCH_EXIT(et);
2967 			ERROUT(ENOENT);
2968 		}
2969 
2970 		nvlist_destroy(nvl);
2971 		nvl = NULL;
2972 		free(nvlpacked, M_NVLIST);
2973 		nvlpacked = NULL;
2974 
2975 		nvl = nvlist_create(0);
2976 		if (nvl == NULL) {
2977 			NET_EPOCH_EXIT(et);
2978 			ERROUT(ENOMEM);
2979 		}
2980 
2981 		if (ruleset->anchor == NULL) {
2982 			RB_FOREACH(anchor, pf_keth_anchor_global,
2983 			    &V_pf_keth_anchors) {
2984 				if (anchor->parent == NULL && nr++ == req_nr) {
2985 					found = true;
2986 					break;
2987 				}
2988 			}
2989 		} else {
2990 			RB_FOREACH(anchor, pf_keth_anchor_node,
2991 			     &ruleset->anchor->children) {
2992 				if (nr++ == req_nr) {
2993 					found = true;
2994 					break;
2995 				}
2996 			}
2997 		}
2998 
2999 		NET_EPOCH_EXIT(et);
3000 		if (found) {
3001 			nvlist_add_number(nvl, "nr", nr);
3002 			nvlist_add_string(nvl, "name", anchor->name);
3003 			if (ruleset->anchor)
3004 				nvlist_add_string(nvl, "path",
3005 				    ruleset->anchor->path);
3006 			else
3007 				nvlist_add_string(nvl, "path", "");
3008 		} else {
3009 			ERROUT(EBUSY);
3010 		}
3011 
3012 		nvlpacked = nvlist_pack(nvl, &nv->len);
3013 		if (nvlpacked == NULL)
3014 			ERROUT(ENOMEM);
3015 
3016 		if (nv->size == 0)
3017 			ERROUT(0);
3018 		else if (nv->size < nv->len)
3019 			ERROUT(ENOSPC);
3020 
3021 		error = copyout(nvlpacked, nv->data, nv->len);
3022 
3023 #undef ERROUT
3024 DIOCGETETHRULESET_error:
3025 		free(nvlpacked, M_NVLIST);
3026 		nvlist_destroy(nvl);
3027 		break;
3028 	}
3029 
3030 	case DIOCADDRULENV: {
3031 		struct pfioc_nv	*nv = (struct pfioc_nv *)addr;
3032 		nvlist_t	*nvl = NULL;
3033 		void		*nvlpacked = NULL;
3034 		struct pf_krule	*rule = NULL;
3035 		const char	*anchor = "", *anchor_call = "";
3036 		uint32_t	 ticket = 0, pool_ticket = 0;
3037 
3038 #define	ERROUT(x)	ERROUT_IOCTL(DIOCADDRULENV_error, x)
3039 
3040 		if (nv->len > pf_ioctl_maxcount)
3041 			ERROUT(ENOMEM);
3042 
3043 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3044 		error = copyin(nv->data, nvlpacked, nv->len);
3045 		if (error)
3046 			ERROUT(error);
3047 
3048 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3049 		if (nvl == NULL)
3050 			ERROUT(EBADMSG);
3051 
3052 		if (! nvlist_exists_number(nvl, "ticket"))
3053 			ERROUT(EINVAL);
3054 		ticket = nvlist_get_number(nvl, "ticket");
3055 
3056 		if (! nvlist_exists_number(nvl, "pool_ticket"))
3057 			ERROUT(EINVAL);
3058 		pool_ticket = nvlist_get_number(nvl, "pool_ticket");
3059 
3060 		if (! nvlist_exists_nvlist(nvl, "rule"))
3061 			ERROUT(EINVAL);
3062 
3063 		rule = pf_krule_alloc();
3064 		error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"),
3065 		    rule);
3066 		if (error)
3067 			ERROUT(error);
3068 
3069 		if (nvlist_exists_string(nvl, "anchor"))
3070 			anchor = nvlist_get_string(nvl, "anchor");
3071 		if (nvlist_exists_string(nvl, "anchor_call"))
3072 			anchor_call = nvlist_get_string(nvl, "anchor_call");
3073 
3074 		if ((error = nvlist_error(nvl)))
3075 			ERROUT(error);
3076 
3077 		/* Frees rule on error */
3078 		error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor,
3079 		    anchor_call, td);
3080 
3081 		nvlist_destroy(nvl);
3082 		free(nvlpacked, M_NVLIST);
3083 		break;
3084 #undef ERROUT
3085 DIOCADDRULENV_error:
3086 		pf_krule_free(rule);
3087 		nvlist_destroy(nvl);
3088 		free(nvlpacked, M_NVLIST);
3089 
3090 		break;
3091 	}
3092 	case DIOCADDRULE: {
3093 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
3094 		struct pf_krule		*rule;
3095 
3096 		rule = pf_krule_alloc();
3097 		error = pf_rule_to_krule(&pr->rule, rule);
3098 		if (error != 0) {
3099 			pf_krule_free(rule);
3100 			break;
3101 		}
3102 
3103 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
3104 
3105 		/* Frees rule on error */
3106 		error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket,
3107 		    pr->anchor, pr->anchor_call, td);
3108 		break;
3109 	}
3110 
3111 	case DIOCGETRULES: {
3112 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
3113 		struct pf_kruleset	*ruleset;
3114 		struct pf_krule		*tail;
3115 		int			 rs_num;
3116 
3117 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
3118 
3119 		PF_RULES_WLOCK();
3120 		ruleset = pf_find_kruleset(pr->anchor);
3121 		if (ruleset == NULL) {
3122 			PF_RULES_WUNLOCK();
3123 			error = EINVAL;
3124 			break;
3125 		}
3126 		rs_num = pf_get_ruleset_number(pr->rule.action);
3127 		if (rs_num >= PF_RULESET_MAX) {
3128 			PF_RULES_WUNLOCK();
3129 			error = EINVAL;
3130 			break;
3131 		}
3132 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
3133 		    pf_krulequeue);
3134 		if (tail)
3135 			pr->nr = tail->nr + 1;
3136 		else
3137 			pr->nr = 0;
3138 		pr->ticket = ruleset->rules[rs_num].active.ticket;
3139 		PF_RULES_WUNLOCK();
3140 		break;
3141 	}
3142 
3143 	case DIOCGETRULENV: {
3144 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
3145 		nvlist_t		*nvrule = NULL;
3146 		nvlist_t		*nvl = NULL;
3147 		struct pf_kruleset	*ruleset;
3148 		struct pf_krule		*rule;
3149 		void			*nvlpacked = NULL;
3150 		int			 rs_num, nr;
3151 		bool			 clear_counter = false;
3152 
3153 #define	ERROUT(x)	ERROUT_IOCTL(DIOCGETRULENV_error, x)
3154 
3155 		if (nv->len > pf_ioctl_maxcount)
3156 			ERROUT(ENOMEM);
3157 
3158 		/* Copy the request in */
3159 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3160 		if (nvlpacked == NULL)
3161 			ERROUT(ENOMEM);
3162 
3163 		error = copyin(nv->data, nvlpacked, nv->len);
3164 		if (error)
3165 			ERROUT(error);
3166 
3167 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3168 		if (nvl == NULL)
3169 			ERROUT(EBADMSG);
3170 
3171 		if (! nvlist_exists_string(nvl, "anchor"))
3172 			ERROUT(EBADMSG);
3173 		if (! nvlist_exists_number(nvl, "ruleset"))
3174 			ERROUT(EBADMSG);
3175 		if (! nvlist_exists_number(nvl, "ticket"))
3176 			ERROUT(EBADMSG);
3177 		if (! nvlist_exists_number(nvl, "nr"))
3178 			ERROUT(EBADMSG);
3179 
3180 		if (nvlist_exists_bool(nvl, "clear_counter"))
3181 			clear_counter = nvlist_get_bool(nvl, "clear_counter");
3182 
3183 		if (clear_counter && !(flags & FWRITE))
3184 			ERROUT(EACCES);
3185 
3186 		nr = nvlist_get_number(nvl, "nr");
3187 
3188 		PF_RULES_WLOCK();
3189 		ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor"));
3190 		if (ruleset == NULL) {
3191 			PF_RULES_WUNLOCK();
3192 			ERROUT(ENOENT);
3193 		}
3194 
3195 		rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset"));
3196 		if (rs_num >= PF_RULESET_MAX) {
3197 			PF_RULES_WUNLOCK();
3198 			ERROUT(EINVAL);
3199 		}
3200 
3201 		if (nvlist_get_number(nvl, "ticket") !=
3202 		    ruleset->rules[rs_num].active.ticket) {
3203 			PF_RULES_WUNLOCK();
3204 			ERROUT(EBUSY);
3205 		}
3206 
3207 		if ((error = nvlist_error(nvl))) {
3208 			PF_RULES_WUNLOCK();
3209 			ERROUT(error);
3210 		}
3211 
3212 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
3213 		while ((rule != NULL) && (rule->nr != nr))
3214 			rule = TAILQ_NEXT(rule, entries);
3215 		if (rule == NULL) {
3216 			PF_RULES_WUNLOCK();
3217 			ERROUT(EBUSY);
3218 		}
3219 
3220 		nvrule = pf_krule_to_nvrule(rule);
3221 
3222 		nvlist_destroy(nvl);
3223 		nvl = nvlist_create(0);
3224 		if (nvl == NULL) {
3225 			PF_RULES_WUNLOCK();
3226 			ERROUT(ENOMEM);
3227 		}
3228 		nvlist_add_number(nvl, "nr", nr);
3229 		nvlist_add_nvlist(nvl, "rule", nvrule);
3230 		nvlist_destroy(nvrule);
3231 		nvrule = NULL;
3232 		if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) {
3233 			PF_RULES_WUNLOCK();
3234 			ERROUT(EBUSY);
3235 		}
3236 
3237 		free(nvlpacked, M_NVLIST);
3238 		nvlpacked = nvlist_pack(nvl, &nv->len);
3239 		if (nvlpacked == NULL) {
3240 			PF_RULES_WUNLOCK();
3241 			ERROUT(ENOMEM);
3242 		}
3243 
3244 		if (nv->size == 0) {
3245 			PF_RULES_WUNLOCK();
3246 			ERROUT(0);
3247 		}
3248 		else if (nv->size < nv->len) {
3249 			PF_RULES_WUNLOCK();
3250 			ERROUT(ENOSPC);
3251 		}
3252 
3253 		if (clear_counter) {
3254 			pf_counter_u64_zero(&rule->evaluations);
3255 			for (int i = 0; i < 2; i++) {
3256 				pf_counter_u64_zero(&rule->packets[i]);
3257 				pf_counter_u64_zero(&rule->bytes[i]);
3258 			}
3259 			counter_u64_zero(rule->states_tot);
3260 		}
3261 		PF_RULES_WUNLOCK();
3262 
3263 		error = copyout(nvlpacked, nv->data, nv->len);
3264 
3265 #undef ERROUT
3266 DIOCGETRULENV_error:
3267 		free(nvlpacked, M_NVLIST);
3268 		nvlist_destroy(nvrule);
3269 		nvlist_destroy(nvl);
3270 
3271 		break;
3272 	}
3273 
3274 	case DIOCCHANGERULE: {
3275 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
3276 		struct pf_kruleset	*ruleset;
3277 		struct pf_krule		*oldrule = NULL, *newrule = NULL;
3278 		struct pfi_kkif		*kif = NULL;
3279 		struct pf_kpooladdr	*pa;
3280 		u_int32_t		 nr = 0;
3281 		int			 rs_num;
3282 
3283 		pcr->anchor[sizeof(pcr->anchor) - 1] = 0;
3284 
3285 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
3286 		    pcr->action > PF_CHANGE_GET_TICKET) {
3287 			error = EINVAL;
3288 			break;
3289 		}
3290 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
3291 			error = EINVAL;
3292 			break;
3293 		}
3294 
3295 		if (pcr->action != PF_CHANGE_REMOVE) {
3296 			newrule = pf_krule_alloc();
3297 			error = pf_rule_to_krule(&pcr->rule, newrule);
3298 			if (error != 0) {
3299 				pf_krule_free(newrule);
3300 				break;
3301 			}
3302 
3303 			if (newrule->ifname[0])
3304 				kif = pf_kkif_create(M_WAITOK);
3305 			pf_counter_u64_init(&newrule->evaluations, M_WAITOK);
3306 			for (int i = 0; i < 2; i++) {
3307 				pf_counter_u64_init(&newrule->packets[i], M_WAITOK);
3308 				pf_counter_u64_init(&newrule->bytes[i], M_WAITOK);
3309 			}
3310 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
3311 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
3312 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
3313 			newrule->cuid = td->td_ucred->cr_ruid;
3314 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
3315 			TAILQ_INIT(&newrule->rpool.list);
3316 		}
3317 #define	ERROUT(x)	ERROUT_IOCTL(DIOCCHANGERULE_error, x)
3318 
3319 		PF_CONFIG_LOCK();
3320 		PF_RULES_WLOCK();
3321 #ifdef PF_WANT_32_TO_64_COUNTER
3322 		if (newrule != NULL) {
3323 			LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist);
3324 			newrule->allrulelinked = true;
3325 			V_pf_allrulecount++;
3326 		}
3327 #endif
3328 
3329 		if (!(pcr->action == PF_CHANGE_REMOVE ||
3330 		    pcr->action == PF_CHANGE_GET_TICKET) &&
3331 		    pcr->pool_ticket != V_ticket_pabuf)
3332 			ERROUT(EBUSY);
3333 
3334 		ruleset = pf_find_kruleset(pcr->anchor);
3335 		if (ruleset == NULL)
3336 			ERROUT(EINVAL);
3337 
3338 		rs_num = pf_get_ruleset_number(pcr->rule.action);
3339 		if (rs_num >= PF_RULESET_MAX)
3340 			ERROUT(EINVAL);
3341 
3342 		/*
3343 		 * XXXMJG: there is no guarantee that the ruleset was
3344 		 * created by the usual route of calling DIOCXBEGIN.
3345 		 * As a result it is possible the rule tree will not
3346 		 * be allocated yet. Hack around it by doing it here.
3347 		 * Note it is fine to let the tree persist in case of
3348 		 * error as it will be freed down the road on future
3349 		 * updates (if need be).
3350 		 */
3351 		if (ruleset->rules[rs_num].active.tree == NULL) {
3352 			ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT);
3353 			if (ruleset->rules[rs_num].active.tree == NULL) {
3354 				ERROUT(ENOMEM);
3355 			}
3356 		}
3357 
3358 		if (pcr->action == PF_CHANGE_GET_TICKET) {
3359 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
3360 			ERROUT(0);
3361 		} else if (pcr->ticket !=
3362 			    ruleset->rules[rs_num].active.ticket)
3363 				ERROUT(EINVAL);
3364 
3365 		if (pcr->action != PF_CHANGE_REMOVE) {
3366 			if (newrule->ifname[0]) {
3367 				newrule->kif = pfi_kkif_attach(kif,
3368 				    newrule->ifname);
3369 				kif = NULL;
3370 				pfi_kkif_ref(newrule->kif);
3371 			} else
3372 				newrule->kif = NULL;
3373 
3374 			if (newrule->rtableid > 0 &&
3375 			    newrule->rtableid >= rt_numfibs)
3376 				error = EBUSY;
3377 
3378 #ifdef ALTQ
3379 			/* set queue IDs */
3380 			if (newrule->qname[0] != 0) {
3381 				if ((newrule->qid =
3382 				    pf_qname2qid(newrule->qname)) == 0)
3383 					error = EBUSY;
3384 				else if (newrule->pqname[0] != 0) {
3385 					if ((newrule->pqid =
3386 					    pf_qname2qid(newrule->pqname)) == 0)
3387 						error = EBUSY;
3388 				} else
3389 					newrule->pqid = newrule->qid;
3390 			}
3391 #endif /* ALTQ */
3392 			if (newrule->tagname[0])
3393 				if ((newrule->tag =
3394 				    pf_tagname2tag(newrule->tagname)) == 0)
3395 					error = EBUSY;
3396 			if (newrule->match_tagname[0])
3397 				if ((newrule->match_tag = pf_tagname2tag(
3398 				    newrule->match_tagname)) == 0)
3399 					error = EBUSY;
3400 			if (newrule->rt && !newrule->direction)
3401 				error = EINVAL;
3402 			if (!newrule->log)
3403 				newrule->logif = 0;
3404 			if (newrule->logif >= PFLOGIFS_MAX)
3405 				error = EINVAL;
3406 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
3407 				error = ENOMEM;
3408 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
3409 				error = ENOMEM;
3410 			if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
3411 				error = EINVAL;
3412 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
3413 				if (pa->addr.type == PF_ADDR_TABLE) {
3414 					pa->addr.p.tbl =
3415 					    pfr_attach_table(ruleset,
3416 					    pa->addr.v.tblname);
3417 					if (pa->addr.p.tbl == NULL)
3418 						error = ENOMEM;
3419 				}
3420 
3421 			newrule->overload_tbl = NULL;
3422 			if (newrule->overload_tblname[0]) {
3423 				if ((newrule->overload_tbl = pfr_attach_table(
3424 				    ruleset, newrule->overload_tblname)) ==
3425 				    NULL)
3426 					error = EINVAL;
3427 				else
3428 					newrule->overload_tbl->pfrkt_flags |=
3429 					    PFR_TFLAG_ACTIVE;
3430 			}
3431 
3432 			pf_mv_kpool(&V_pf_pabuf, &newrule->rpool.list);
3433 			if (((((newrule->action == PF_NAT) ||
3434 			    (newrule->action == PF_RDR) ||
3435 			    (newrule->action == PF_BINAT) ||
3436 			    (newrule->rt > PF_NOPFROUTE)) &&
3437 			    !newrule->anchor)) &&
3438 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
3439 				error = EINVAL;
3440 
3441 			if (error) {
3442 				pf_free_rule(newrule);
3443 				PF_RULES_WUNLOCK();
3444 				PF_CONFIG_UNLOCK();
3445 				break;
3446 			}
3447 
3448 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
3449 		}
3450 		pf_empty_kpool(&V_pf_pabuf);
3451 
3452 		if (pcr->action == PF_CHANGE_ADD_HEAD)
3453 			oldrule = TAILQ_FIRST(
3454 			    ruleset->rules[rs_num].active.ptr);
3455 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
3456 			oldrule = TAILQ_LAST(
3457 			    ruleset->rules[rs_num].active.ptr, pf_krulequeue);
3458 		else {
3459 			oldrule = TAILQ_FIRST(
3460 			    ruleset->rules[rs_num].active.ptr);
3461 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
3462 				oldrule = TAILQ_NEXT(oldrule, entries);
3463 			if (oldrule == NULL) {
3464 				if (newrule != NULL)
3465 					pf_free_rule(newrule);
3466 				PF_RULES_WUNLOCK();
3467 				PF_CONFIG_UNLOCK();
3468 				error = EINVAL;
3469 				break;
3470 			}
3471 		}
3472 
3473 		if (pcr->action == PF_CHANGE_REMOVE) {
3474 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
3475 			    oldrule);
3476 			RB_REMOVE(pf_krule_global,
3477 			    ruleset->rules[rs_num].active.tree, oldrule);
3478 			ruleset->rules[rs_num].active.rcount--;
3479 		} else {
3480 			pf_hash_rule(newrule);
3481 			if (RB_INSERT(pf_krule_global,
3482 			    ruleset->rules[rs_num].active.tree, newrule) != NULL) {
3483 				pf_free_rule(newrule);
3484 				PF_RULES_WUNLOCK();
3485 				PF_CONFIG_UNLOCK();
3486 				error = EEXIST;
3487 				break;
3488 			}
3489 
3490 			if (oldrule == NULL)
3491 				TAILQ_INSERT_TAIL(
3492 				    ruleset->rules[rs_num].active.ptr,
3493 				    newrule, entries);
3494 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
3495 			    pcr->action == PF_CHANGE_ADD_BEFORE)
3496 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
3497 			else
3498 				TAILQ_INSERT_AFTER(
3499 				    ruleset->rules[rs_num].active.ptr,
3500 				    oldrule, newrule, entries);
3501 			ruleset->rules[rs_num].active.rcount++;
3502 		}
3503 
3504 		nr = 0;
3505 		TAILQ_FOREACH(oldrule,
3506 		    ruleset->rules[rs_num].active.ptr, entries)
3507 			oldrule->nr = nr++;
3508 
3509 		ruleset->rules[rs_num].active.ticket++;
3510 
3511 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
3512 		pf_remove_if_empty_kruleset(ruleset);
3513 
3514 		PF_RULES_WUNLOCK();
3515 		PF_CONFIG_UNLOCK();
3516 		break;
3517 
3518 #undef ERROUT
3519 DIOCCHANGERULE_error:
3520 		PF_RULES_WUNLOCK();
3521 		PF_CONFIG_UNLOCK();
3522 		pf_krule_free(newrule);
3523 		pf_kkif_free(kif);
3524 		break;
3525 	}
3526 
3527 	case DIOCCLRSTATESNV: {
3528 		error = pf_clearstates_nv((struct pfioc_nv *)addr);
3529 		break;
3530 	}
3531 
3532 	case DIOCKILLSTATESNV: {
3533 		error = pf_killstates_nv((struct pfioc_nv *)addr);
3534 		break;
3535 	}
3536 
3537 	case DIOCADDSTATE: {
3538 		struct pfioc_state		*ps = (struct pfioc_state *)addr;
3539 		struct pfsync_state_1301	*sp = &ps->state;
3540 
3541 		if (sp->timeout >= PFTM_MAX) {
3542 			error = EINVAL;
3543 			break;
3544 		}
3545 		if (V_pfsync_state_import_ptr != NULL) {
3546 			PF_RULES_RLOCK();
3547 			error = V_pfsync_state_import_ptr(
3548 			    (union pfsync_state_union *)sp, PFSYNC_SI_IOCTL,
3549 			    PFSYNC_MSG_VERSION_1301);
3550 			PF_RULES_RUNLOCK();
3551 		} else
3552 			error = EOPNOTSUPP;
3553 		break;
3554 	}
3555 
3556 	case DIOCGETSTATE: {
3557 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
3558 		struct pf_kstate	*s;
3559 
3560 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
3561 		if (s == NULL) {
3562 			error = ENOENT;
3563 			break;
3564 		}
3565 
3566 		pfsync_state_export((union pfsync_state_union*)&ps->state,
3567 		    s, PFSYNC_MSG_VERSION_1301);
3568 		PF_STATE_UNLOCK(s);
3569 		break;
3570 	}
3571 
3572 	case DIOCGETSTATENV: {
3573 		error = pf_getstate((struct pfioc_nv *)addr);
3574 		break;
3575 	}
3576 
3577 #ifdef COMPAT_FREEBSD14
3578 	case DIOCGETSTATES: {
3579 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
3580 		struct pf_kstate	*s;
3581 		struct pfsync_state_1301	*pstore, *p;
3582 		int			 i, nr;
3583 		size_t			 slice_count = 16, count;
3584 		void			*out;
3585 
3586 		if (ps->ps_len <= 0) {
3587 			nr = uma_zone_get_cur(V_pf_state_z);
3588 			ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3589 			break;
3590 		}
3591 
3592 		out = ps->ps_states;
3593 		pstore = mallocarray(slice_count,
3594 		    sizeof(struct pfsync_state_1301), M_TEMP, M_WAITOK | M_ZERO);
3595 		nr = 0;
3596 
3597 		for (i = 0; i <= pf_hashmask; i++) {
3598 			struct pf_idhash *ih = &V_pf_idhash[i];
3599 
3600 DIOCGETSTATES_retry:
3601 			p = pstore;
3602 
3603 			if (LIST_EMPTY(&ih->states))
3604 				continue;
3605 
3606 			PF_HASHROW_LOCK(ih);
3607 			count = 0;
3608 			LIST_FOREACH(s, &ih->states, entry) {
3609 				if (s->timeout == PFTM_UNLINKED)
3610 					continue;
3611 				count++;
3612 			}
3613 
3614 			if (count > slice_count) {
3615 				PF_HASHROW_UNLOCK(ih);
3616 				free(pstore, M_TEMP);
3617 				slice_count = count * 2;
3618 				pstore = mallocarray(slice_count,
3619 				    sizeof(struct pfsync_state_1301), M_TEMP,
3620 				    M_WAITOK | M_ZERO);
3621 				goto DIOCGETSTATES_retry;
3622 			}
3623 
3624 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
3625 				PF_HASHROW_UNLOCK(ih);
3626 				goto DIOCGETSTATES_full;
3627 			}
3628 
3629 			LIST_FOREACH(s, &ih->states, entry) {
3630 				if (s->timeout == PFTM_UNLINKED)
3631 					continue;
3632 
3633 				pfsync_state_export((union pfsync_state_union*)p,
3634 				    s, PFSYNC_MSG_VERSION_1301);
3635 				p++;
3636 				nr++;
3637 			}
3638 			PF_HASHROW_UNLOCK(ih);
3639 			error = copyout(pstore, out,
3640 			    sizeof(struct pfsync_state_1301) * count);
3641 			if (error)
3642 				break;
3643 			out = ps->ps_states + nr;
3644 		}
3645 DIOCGETSTATES_full:
3646 		ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3647 		free(pstore, M_TEMP);
3648 
3649 		break;
3650 	}
3651 
3652 	case DIOCGETSTATESV2: {
3653 		struct pfioc_states_v2	*ps = (struct pfioc_states_v2 *)addr;
3654 		struct pf_kstate	*s;
3655 		struct pf_state_export	*pstore, *p;
3656 		int i, nr;
3657 		size_t slice_count = 16, count;
3658 		void *out;
3659 
3660 		if (ps->ps_req_version > PF_STATE_VERSION) {
3661 			error = ENOTSUP;
3662 			break;
3663 		}
3664 
3665 		if (ps->ps_len <= 0) {
3666 			nr = uma_zone_get_cur(V_pf_state_z);
3667 			ps->ps_len = sizeof(struct pf_state_export) * nr;
3668 			break;
3669 		}
3670 
3671 		out = ps->ps_states;
3672 		pstore = mallocarray(slice_count,
3673 		    sizeof(struct pf_state_export), M_TEMP, M_WAITOK | M_ZERO);
3674 		nr = 0;
3675 
3676 		for (i = 0; i <= pf_hashmask; i++) {
3677 			struct pf_idhash *ih = &V_pf_idhash[i];
3678 
3679 DIOCGETSTATESV2_retry:
3680 			p = pstore;
3681 
3682 			if (LIST_EMPTY(&ih->states))
3683 				continue;
3684 
3685 			PF_HASHROW_LOCK(ih);
3686 			count = 0;
3687 			LIST_FOREACH(s, &ih->states, entry) {
3688 				if (s->timeout == PFTM_UNLINKED)
3689 					continue;
3690 				count++;
3691 			}
3692 
3693 			if (count > slice_count) {
3694 				PF_HASHROW_UNLOCK(ih);
3695 				free(pstore, M_TEMP);
3696 				slice_count = count * 2;
3697 				pstore = mallocarray(slice_count,
3698 				    sizeof(struct pf_state_export), M_TEMP,
3699 				    M_WAITOK | M_ZERO);
3700 				goto DIOCGETSTATESV2_retry;
3701 			}
3702 
3703 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
3704 				PF_HASHROW_UNLOCK(ih);
3705 				goto DIOCGETSTATESV2_full;
3706 			}
3707 
3708 			LIST_FOREACH(s, &ih->states, entry) {
3709 				if (s->timeout == PFTM_UNLINKED)
3710 					continue;
3711 
3712 				pf_state_export(p, s);
3713 				p++;
3714 				nr++;
3715 			}
3716 			PF_HASHROW_UNLOCK(ih);
3717 			error = copyout(pstore, out,
3718 			    sizeof(struct pf_state_export) * count);
3719 			if (error)
3720 				break;
3721 			out = ps->ps_states + nr;
3722 		}
3723 DIOCGETSTATESV2_full:
3724 		ps->ps_len = nr * sizeof(struct pf_state_export);
3725 		free(pstore, M_TEMP);
3726 
3727 		break;
3728 	}
3729 #endif
3730 	case DIOCGETSTATUSNV: {
3731 		error = pf_getstatus((struct pfioc_nv *)addr);
3732 		break;
3733 	}
3734 
3735 	case DIOCSETSTATUSIF: {
3736 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
3737 
3738 		if (pi->ifname[0] == 0) {
3739 			bzero(V_pf_status.ifname, IFNAMSIZ);
3740 			break;
3741 		}
3742 		PF_RULES_WLOCK();
3743 		error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
3744 		PF_RULES_WUNLOCK();
3745 		break;
3746 	}
3747 
3748 	case DIOCCLRSTATUS: {
3749 		PF_RULES_WLOCK();
3750 		for (int i = 0; i < PFRES_MAX; i++)
3751 			counter_u64_zero(V_pf_status.counters[i]);
3752 		for (int i = 0; i < FCNT_MAX; i++)
3753 			pf_counter_u64_zero(&V_pf_status.fcounters[i]);
3754 		for (int i = 0; i < SCNT_MAX; i++)
3755 			counter_u64_zero(V_pf_status.scounters[i]);
3756 		for (int i = 0; i < KLCNT_MAX; i++)
3757 			counter_u64_zero(V_pf_status.lcounters[i]);
3758 		V_pf_status.since = time_second;
3759 		if (*V_pf_status.ifname)
3760 			pfi_update_status(V_pf_status.ifname, NULL);
3761 		PF_RULES_WUNLOCK();
3762 		break;
3763 	}
3764 
3765 	case DIOCNATLOOK: {
3766 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
3767 		struct pf_state_key	*sk;
3768 		struct pf_kstate	*state;
3769 		struct pf_state_key_cmp	 key;
3770 		int			 m = 0, direction = pnl->direction;
3771 		int			 sidx, didx;
3772 
3773 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
3774 		sidx = (direction == PF_IN) ? 1 : 0;
3775 		didx = (direction == PF_IN) ? 0 : 1;
3776 
3777 		if (!pnl->proto ||
3778 		    PF_AZERO(&pnl->saddr, pnl->af) ||
3779 		    PF_AZERO(&pnl->daddr, pnl->af) ||
3780 		    ((pnl->proto == IPPROTO_TCP ||
3781 		    pnl->proto == IPPROTO_UDP) &&
3782 		    (!pnl->dport || !pnl->sport)))
3783 			error = EINVAL;
3784 		else {
3785 			bzero(&key, sizeof(key));
3786 			key.af = pnl->af;
3787 			key.proto = pnl->proto;
3788 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
3789 			key.port[sidx] = pnl->sport;
3790 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
3791 			key.port[didx] = pnl->dport;
3792 
3793 			state = pf_find_state_all(&key, direction, &m);
3794 			if (state == NULL) {
3795 				error = ENOENT;
3796 			} else {
3797 				if (m > 1) {
3798 					PF_STATE_UNLOCK(state);
3799 					error = E2BIG;	/* more than one state */
3800 				} else {
3801 					sk = state->key[sidx];
3802 					PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
3803 					pnl->rsport = sk->port[sidx];
3804 					PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
3805 					pnl->rdport = sk->port[didx];
3806 					PF_STATE_UNLOCK(state);
3807 				}
3808 			}
3809 		}
3810 		break;
3811 	}
3812 
3813 	case DIOCSETTIMEOUT: {
3814 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
3815 		int		 old;
3816 
3817 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
3818 		    pt->seconds < 0) {
3819 			error = EINVAL;
3820 			break;
3821 		}
3822 		PF_RULES_WLOCK();
3823 		old = V_pf_default_rule.timeout[pt->timeout];
3824 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
3825 			pt->seconds = 1;
3826 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
3827 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
3828 			wakeup(pf_purge_thread);
3829 		pt->seconds = old;
3830 		PF_RULES_WUNLOCK();
3831 		break;
3832 	}
3833 
3834 	case DIOCGETTIMEOUT: {
3835 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
3836 
3837 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
3838 			error = EINVAL;
3839 			break;
3840 		}
3841 		PF_RULES_RLOCK();
3842 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
3843 		PF_RULES_RUNLOCK();
3844 		break;
3845 	}
3846 
3847 	case DIOCGETLIMIT: {
3848 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
3849 
3850 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
3851 			error = EINVAL;
3852 			break;
3853 		}
3854 		PF_RULES_RLOCK();
3855 		pl->limit = V_pf_limits[pl->index].limit;
3856 		PF_RULES_RUNLOCK();
3857 		break;
3858 	}
3859 
3860 	case DIOCSETLIMIT: {
3861 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
3862 		int			 old_limit;
3863 
3864 		PF_RULES_WLOCK();
3865 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
3866 		    V_pf_limits[pl->index].zone == NULL) {
3867 			PF_RULES_WUNLOCK();
3868 			error = EINVAL;
3869 			break;
3870 		}
3871 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
3872 		old_limit = V_pf_limits[pl->index].limit;
3873 		V_pf_limits[pl->index].limit = pl->limit;
3874 		pl->limit = old_limit;
3875 		PF_RULES_WUNLOCK();
3876 		break;
3877 	}
3878 
3879 	case DIOCSETDEBUG: {
3880 		u_int32_t	*level = (u_int32_t *)addr;
3881 
3882 		PF_RULES_WLOCK();
3883 		V_pf_status.debug = *level;
3884 		PF_RULES_WUNLOCK();
3885 		break;
3886 	}
3887 
3888 	case DIOCCLRRULECTRS: {
3889 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
3890 		struct pf_kruleset	*ruleset = &pf_main_ruleset;
3891 		struct pf_krule		*rule;
3892 
3893 		PF_RULES_WLOCK();
3894 		TAILQ_FOREACH(rule,
3895 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
3896 			pf_counter_u64_zero(&rule->evaluations);
3897 			for (int i = 0; i < 2; i++) {
3898 				pf_counter_u64_zero(&rule->packets[i]);
3899 				pf_counter_u64_zero(&rule->bytes[i]);
3900 			}
3901 		}
3902 		PF_RULES_WUNLOCK();
3903 		break;
3904 	}
3905 
3906 	case DIOCGIFSPEEDV0:
3907 	case DIOCGIFSPEEDV1: {
3908 		struct pf_ifspeed_v1	*psp = (struct pf_ifspeed_v1 *)addr;
3909 		struct pf_ifspeed_v1	ps;
3910 		struct ifnet		*ifp;
3911 
3912 		if (psp->ifname[0] == '\0') {
3913 			error = EINVAL;
3914 			break;
3915 		}
3916 
3917 		error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ);
3918 		if (error != 0)
3919 			break;
3920 		ifp = ifunit(ps.ifname);
3921 		if (ifp != NULL) {
3922 			psp->baudrate32 =
3923 			    (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
3924 			if (cmd == DIOCGIFSPEEDV1)
3925 				psp->baudrate = ifp->if_baudrate;
3926 		} else {
3927 			error = EINVAL;
3928 		}
3929 		break;
3930 	}
3931 
3932 #ifdef ALTQ
3933 	case DIOCSTARTALTQ: {
3934 		struct pf_altq		*altq;
3935 
3936 		PF_RULES_WLOCK();
3937 		/* enable all altq interfaces on active list */
3938 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
3939 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
3940 				error = pf_enable_altq(altq);
3941 				if (error != 0)
3942 					break;
3943 			}
3944 		}
3945 		if (error == 0)
3946 			V_pf_altq_running = 1;
3947 		PF_RULES_WUNLOCK();
3948 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
3949 		break;
3950 	}
3951 
3952 	case DIOCSTOPALTQ: {
3953 		struct pf_altq		*altq;
3954 
3955 		PF_RULES_WLOCK();
3956 		/* disable all altq interfaces on active list */
3957 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
3958 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
3959 				error = pf_disable_altq(altq);
3960 				if (error != 0)
3961 					break;
3962 			}
3963 		}
3964 		if (error == 0)
3965 			V_pf_altq_running = 0;
3966 		PF_RULES_WUNLOCK();
3967 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
3968 		break;
3969 	}
3970 
3971 	case DIOCADDALTQV0:
3972 	case DIOCADDALTQV1: {
3973 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
3974 		struct pf_altq		*altq, *a;
3975 		struct ifnet		*ifp;
3976 
3977 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
3978 		error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
3979 		if (error)
3980 			break;
3981 		altq->local_flags = 0;
3982 
3983 		PF_RULES_WLOCK();
3984 		if (pa->ticket != V_ticket_altqs_inactive) {
3985 			PF_RULES_WUNLOCK();
3986 			free(altq, M_PFALTQ);
3987 			error = EBUSY;
3988 			break;
3989 		}
3990 
3991 		/*
3992 		 * if this is for a queue, find the discipline and
3993 		 * copy the necessary fields
3994 		 */
3995 		if (altq->qname[0] != 0) {
3996 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
3997 				PF_RULES_WUNLOCK();
3998 				error = EBUSY;
3999 				free(altq, M_PFALTQ);
4000 				break;
4001 			}
4002 			altq->altq_disc = NULL;
4003 			TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
4004 				if (strncmp(a->ifname, altq->ifname,
4005 				    IFNAMSIZ) == 0) {
4006 					altq->altq_disc = a->altq_disc;
4007 					break;
4008 				}
4009 			}
4010 		}
4011 
4012 		if ((ifp = ifunit(altq->ifname)) == NULL)
4013 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
4014 		else
4015 			error = altq_add(ifp, altq);
4016 
4017 		if (error) {
4018 			PF_RULES_WUNLOCK();
4019 			free(altq, M_PFALTQ);
4020 			break;
4021 		}
4022 
4023 		if (altq->qname[0] != 0)
4024 			TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
4025 		else
4026 			TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
4027 		/* version error check done on import above */
4028 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4029 		PF_RULES_WUNLOCK();
4030 		break;
4031 	}
4032 
4033 	case DIOCGETALTQSV0:
4034 	case DIOCGETALTQSV1: {
4035 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4036 		struct pf_altq		*altq;
4037 
4038 		PF_RULES_RLOCK();
4039 		pa->nr = 0;
4040 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
4041 			pa->nr++;
4042 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
4043 			pa->nr++;
4044 		pa->ticket = V_ticket_altqs_active;
4045 		PF_RULES_RUNLOCK();
4046 		break;
4047 	}
4048 
4049 	case DIOCGETALTQV0:
4050 	case DIOCGETALTQV1: {
4051 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4052 		struct pf_altq		*altq;
4053 
4054 		PF_RULES_RLOCK();
4055 		if (pa->ticket != V_ticket_altqs_active) {
4056 			PF_RULES_RUNLOCK();
4057 			error = EBUSY;
4058 			break;
4059 		}
4060 		altq = pf_altq_get_nth_active(pa->nr);
4061 		if (altq == NULL) {
4062 			PF_RULES_RUNLOCK();
4063 			error = EBUSY;
4064 			break;
4065 		}
4066 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4067 		PF_RULES_RUNLOCK();
4068 		break;
4069 	}
4070 
4071 	case DIOCCHANGEALTQV0:
4072 	case DIOCCHANGEALTQV1:
4073 		/* CHANGEALTQ not supported yet! */
4074 		error = ENODEV;
4075 		break;
4076 
4077 	case DIOCGETQSTATSV0:
4078 	case DIOCGETQSTATSV1: {
4079 		struct pfioc_qstats_v1	*pq = (struct pfioc_qstats_v1 *)addr;
4080 		struct pf_altq		*altq;
4081 		int			 nbytes;
4082 		u_int32_t		 version;
4083 
4084 		PF_RULES_RLOCK();
4085 		if (pq->ticket != V_ticket_altqs_active) {
4086 			PF_RULES_RUNLOCK();
4087 			error = EBUSY;
4088 			break;
4089 		}
4090 		nbytes = pq->nbytes;
4091 		altq = pf_altq_get_nth_active(pq->nr);
4092 		if (altq == NULL) {
4093 			PF_RULES_RUNLOCK();
4094 			error = EBUSY;
4095 			break;
4096 		}
4097 
4098 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
4099 			PF_RULES_RUNLOCK();
4100 			error = ENXIO;
4101 			break;
4102 		}
4103 		PF_RULES_RUNLOCK();
4104 		if (cmd == DIOCGETQSTATSV0)
4105 			version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
4106 		else
4107 			version = pq->version;
4108 		error = altq_getqstats(altq, pq->buf, &nbytes, version);
4109 		if (error == 0) {
4110 			pq->scheduler = altq->scheduler;
4111 			pq->nbytes = nbytes;
4112 		}
4113 		break;
4114 	}
4115 #endif /* ALTQ */
4116 
4117 	case DIOCBEGINADDRS: {
4118 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4119 
4120 		PF_RULES_WLOCK();
4121 		pf_empty_kpool(&V_pf_pabuf);
4122 		pp->ticket = ++V_ticket_pabuf;
4123 		PF_RULES_WUNLOCK();
4124 		break;
4125 	}
4126 
4127 	case DIOCADDADDR: {
4128 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4129 		struct pf_kpooladdr	*pa;
4130 		struct pfi_kkif		*kif = NULL;
4131 
4132 #ifndef INET
4133 		if (pp->af == AF_INET) {
4134 			error = EAFNOSUPPORT;
4135 			break;
4136 		}
4137 #endif /* INET */
4138 #ifndef INET6
4139 		if (pp->af == AF_INET6) {
4140 			error = EAFNOSUPPORT;
4141 			break;
4142 		}
4143 #endif /* INET6 */
4144 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
4145 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
4146 		    pp->addr.addr.type != PF_ADDR_TABLE) {
4147 			error = EINVAL;
4148 			break;
4149 		}
4150 		if (pp->addr.addr.p.dyn != NULL) {
4151 			error = EINVAL;
4152 			break;
4153 		}
4154 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
4155 		error = pf_pooladdr_to_kpooladdr(&pp->addr, pa);
4156 		if (error != 0)
4157 			break;
4158 		if (pa->ifname[0])
4159 			kif = pf_kkif_create(M_WAITOK);
4160 		PF_RULES_WLOCK();
4161 		if (pp->ticket != V_ticket_pabuf) {
4162 			PF_RULES_WUNLOCK();
4163 			if (pa->ifname[0])
4164 				pf_kkif_free(kif);
4165 			free(pa, M_PFRULE);
4166 			error = EBUSY;
4167 			break;
4168 		}
4169 		if (pa->ifname[0]) {
4170 			pa->kif = pfi_kkif_attach(kif, pa->ifname);
4171 			kif = NULL;
4172 			pfi_kkif_ref(pa->kif);
4173 		} else
4174 			pa->kif = NULL;
4175 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
4176 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
4177 			if (pa->ifname[0])
4178 				pfi_kkif_unref(pa->kif);
4179 			PF_RULES_WUNLOCK();
4180 			free(pa, M_PFRULE);
4181 			break;
4182 		}
4183 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
4184 		PF_RULES_WUNLOCK();
4185 		break;
4186 	}
4187 
4188 	case DIOCGETADDRS: {
4189 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4190 		struct pf_kpool		*pool;
4191 		struct pf_kpooladdr	*pa;
4192 
4193 		pp->anchor[sizeof(pp->anchor) - 1] = 0;
4194 		pp->nr = 0;
4195 
4196 		PF_RULES_RLOCK();
4197 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
4198 		    pp->r_num, 0, 1, 0);
4199 		if (pool == NULL) {
4200 			PF_RULES_RUNLOCK();
4201 			error = EBUSY;
4202 			break;
4203 		}
4204 		TAILQ_FOREACH(pa, &pool->list, entries)
4205 			pp->nr++;
4206 		PF_RULES_RUNLOCK();
4207 		break;
4208 	}
4209 
4210 	case DIOCGETADDR: {
4211 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4212 		struct pf_kpool		*pool;
4213 		struct pf_kpooladdr	*pa;
4214 		u_int32_t		 nr = 0;
4215 
4216 		pp->anchor[sizeof(pp->anchor) - 1] = 0;
4217 
4218 		PF_RULES_RLOCK();
4219 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
4220 		    pp->r_num, 0, 1, 1);
4221 		if (pool == NULL) {
4222 			PF_RULES_RUNLOCK();
4223 			error = EBUSY;
4224 			break;
4225 		}
4226 		pa = TAILQ_FIRST(&pool->list);
4227 		while ((pa != NULL) && (nr < pp->nr)) {
4228 			pa = TAILQ_NEXT(pa, entries);
4229 			nr++;
4230 		}
4231 		if (pa == NULL) {
4232 			PF_RULES_RUNLOCK();
4233 			error = EBUSY;
4234 			break;
4235 		}
4236 		pf_kpooladdr_to_pooladdr(pa, &pp->addr);
4237 		pf_addr_copyout(&pp->addr.addr);
4238 		PF_RULES_RUNLOCK();
4239 		break;
4240 	}
4241 
4242 	case DIOCCHANGEADDR: {
4243 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
4244 		struct pf_kpool		*pool;
4245 		struct pf_kpooladdr	*oldpa = NULL, *newpa = NULL;
4246 		struct pf_kruleset	*ruleset;
4247 		struct pfi_kkif		*kif = NULL;
4248 
4249 		pca->anchor[sizeof(pca->anchor) - 1] = 0;
4250 
4251 		if (pca->action < PF_CHANGE_ADD_HEAD ||
4252 		    pca->action > PF_CHANGE_REMOVE) {
4253 			error = EINVAL;
4254 			break;
4255 		}
4256 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
4257 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
4258 		    pca->addr.addr.type != PF_ADDR_TABLE) {
4259 			error = EINVAL;
4260 			break;
4261 		}
4262 		if (pca->addr.addr.p.dyn != NULL) {
4263 			error = EINVAL;
4264 			break;
4265 		}
4266 
4267 		if (pca->action != PF_CHANGE_REMOVE) {
4268 #ifndef INET
4269 			if (pca->af == AF_INET) {
4270 				error = EAFNOSUPPORT;
4271 				break;
4272 			}
4273 #endif /* INET */
4274 #ifndef INET6
4275 			if (pca->af == AF_INET6) {
4276 				error = EAFNOSUPPORT;
4277 				break;
4278 			}
4279 #endif /* INET6 */
4280 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
4281 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
4282 			if (newpa->ifname[0])
4283 				kif = pf_kkif_create(M_WAITOK);
4284 			newpa->kif = NULL;
4285 		}
4286 #define	ERROUT(x)	ERROUT_IOCTL(DIOCCHANGEADDR_error, x)
4287 		PF_RULES_WLOCK();
4288 		ruleset = pf_find_kruleset(pca->anchor);
4289 		if (ruleset == NULL)
4290 			ERROUT(EBUSY);
4291 
4292 		pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
4293 		    pca->r_num, pca->r_last, 1, 1);
4294 		if (pool == NULL)
4295 			ERROUT(EBUSY);
4296 
4297 		if (pca->action != PF_CHANGE_REMOVE) {
4298 			if (newpa->ifname[0]) {
4299 				newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
4300 				pfi_kkif_ref(newpa->kif);
4301 				kif = NULL;
4302 			}
4303 
4304 			switch (newpa->addr.type) {
4305 			case PF_ADDR_DYNIFTL:
4306 				error = pfi_dynaddr_setup(&newpa->addr,
4307 				    pca->af);
4308 				break;
4309 			case PF_ADDR_TABLE:
4310 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
4311 				    newpa->addr.v.tblname);
4312 				if (newpa->addr.p.tbl == NULL)
4313 					error = ENOMEM;
4314 				break;
4315 			}
4316 			if (error)
4317 				goto DIOCCHANGEADDR_error;
4318 		}
4319 
4320 		switch (pca->action) {
4321 		case PF_CHANGE_ADD_HEAD:
4322 			oldpa = TAILQ_FIRST(&pool->list);
4323 			break;
4324 		case PF_CHANGE_ADD_TAIL:
4325 			oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
4326 			break;
4327 		default:
4328 			oldpa = TAILQ_FIRST(&pool->list);
4329 			for (int i = 0; oldpa && i < pca->nr; i++)
4330 				oldpa = TAILQ_NEXT(oldpa, entries);
4331 
4332 			if (oldpa == NULL)
4333 				ERROUT(EINVAL);
4334 		}
4335 
4336 		if (pca->action == PF_CHANGE_REMOVE) {
4337 			TAILQ_REMOVE(&pool->list, oldpa, entries);
4338 			switch (oldpa->addr.type) {
4339 			case PF_ADDR_DYNIFTL:
4340 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
4341 				break;
4342 			case PF_ADDR_TABLE:
4343 				pfr_detach_table(oldpa->addr.p.tbl);
4344 				break;
4345 			}
4346 			if (oldpa->kif)
4347 				pfi_kkif_unref(oldpa->kif);
4348 			free(oldpa, M_PFRULE);
4349 		} else {
4350 			if (oldpa == NULL)
4351 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
4352 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
4353 			    pca->action == PF_CHANGE_ADD_BEFORE)
4354 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
4355 			else
4356 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
4357 				    newpa, entries);
4358 		}
4359 
4360 		pool->cur = TAILQ_FIRST(&pool->list);
4361 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
4362 		PF_RULES_WUNLOCK();
4363 		break;
4364 
4365 #undef ERROUT
4366 DIOCCHANGEADDR_error:
4367 		if (newpa != NULL) {
4368 			if (newpa->kif)
4369 				pfi_kkif_unref(newpa->kif);
4370 			free(newpa, M_PFRULE);
4371 		}
4372 		PF_RULES_WUNLOCK();
4373 		pf_kkif_free(kif);
4374 		break;
4375 	}
4376 
4377 	case DIOCGETRULESETS: {
4378 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
4379 		struct pf_kruleset	*ruleset;
4380 		struct pf_kanchor	*anchor;
4381 
4382 		pr->path[sizeof(pr->path) - 1] = 0;
4383 
4384 		PF_RULES_RLOCK();
4385 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4386 			PF_RULES_RUNLOCK();
4387 			error = ENOENT;
4388 			break;
4389 		}
4390 		pr->nr = 0;
4391 		if (ruleset->anchor == NULL) {
4392 			/* XXX kludge for pf_main_ruleset */
4393 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4394 				if (anchor->parent == NULL)
4395 					pr->nr++;
4396 		} else {
4397 			RB_FOREACH(anchor, pf_kanchor_node,
4398 			    &ruleset->anchor->children)
4399 				pr->nr++;
4400 		}
4401 		PF_RULES_RUNLOCK();
4402 		break;
4403 	}
4404 
4405 	case DIOCGETRULESET: {
4406 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
4407 		struct pf_kruleset	*ruleset;
4408 		struct pf_kanchor	*anchor;
4409 		u_int32_t		 nr = 0;
4410 
4411 		pr->path[sizeof(pr->path) - 1] = 0;
4412 
4413 		PF_RULES_RLOCK();
4414 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4415 			PF_RULES_RUNLOCK();
4416 			error = ENOENT;
4417 			break;
4418 		}
4419 		pr->name[0] = 0;
4420 		if (ruleset->anchor == NULL) {
4421 			/* XXX kludge for pf_main_ruleset */
4422 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4423 				if (anchor->parent == NULL && nr++ == pr->nr) {
4424 					strlcpy(pr->name, anchor->name,
4425 					    sizeof(pr->name));
4426 					break;
4427 				}
4428 		} else {
4429 			RB_FOREACH(anchor, pf_kanchor_node,
4430 			    &ruleset->anchor->children)
4431 				if (nr++ == pr->nr) {
4432 					strlcpy(pr->name, anchor->name,
4433 					    sizeof(pr->name));
4434 					break;
4435 				}
4436 		}
4437 		if (!pr->name[0])
4438 			error = EBUSY;
4439 		PF_RULES_RUNLOCK();
4440 		break;
4441 	}
4442 
4443 	case DIOCRCLRTABLES: {
4444 		struct pfioc_table *io = (struct pfioc_table *)addr;
4445 
4446 		if (io->pfrio_esize != 0) {
4447 			error = ENODEV;
4448 			break;
4449 		}
4450 		PF_RULES_WLOCK();
4451 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
4452 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4453 		PF_RULES_WUNLOCK();
4454 		break;
4455 	}
4456 
4457 	case DIOCRADDTABLES: {
4458 		struct pfioc_table *io = (struct pfioc_table *)addr;
4459 		struct pfr_table *pfrts;
4460 		size_t totlen;
4461 
4462 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4463 			error = ENODEV;
4464 			break;
4465 		}
4466 
4467 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4468 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4469 			error = ENOMEM;
4470 			break;
4471 		}
4472 
4473 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4474 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4475 		    M_TEMP, M_WAITOK);
4476 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4477 		if (error) {
4478 			free(pfrts, M_TEMP);
4479 			break;
4480 		}
4481 		PF_RULES_WLOCK();
4482 		error = pfr_add_tables(pfrts, io->pfrio_size,
4483 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4484 		PF_RULES_WUNLOCK();
4485 		free(pfrts, M_TEMP);
4486 		break;
4487 	}
4488 
4489 	case DIOCRDELTABLES: {
4490 		struct pfioc_table *io = (struct pfioc_table *)addr;
4491 		struct pfr_table *pfrts;
4492 		size_t totlen;
4493 
4494 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4495 			error = ENODEV;
4496 			break;
4497 		}
4498 
4499 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4500 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4501 			error = ENOMEM;
4502 			break;
4503 		}
4504 
4505 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4506 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4507 		    M_TEMP, M_WAITOK);
4508 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4509 		if (error) {
4510 			free(pfrts, M_TEMP);
4511 			break;
4512 		}
4513 		PF_RULES_WLOCK();
4514 		error = pfr_del_tables(pfrts, io->pfrio_size,
4515 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4516 		PF_RULES_WUNLOCK();
4517 		free(pfrts, M_TEMP);
4518 		break;
4519 	}
4520 
4521 	case DIOCRGETTABLES: {
4522 		struct pfioc_table *io = (struct pfioc_table *)addr;
4523 		struct pfr_table *pfrts;
4524 		size_t totlen;
4525 		int n;
4526 
4527 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4528 			error = ENODEV;
4529 			break;
4530 		}
4531 		PF_RULES_RLOCK();
4532 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4533 		if (n < 0) {
4534 			PF_RULES_RUNLOCK();
4535 			error = EINVAL;
4536 			break;
4537 		}
4538 		io->pfrio_size = min(io->pfrio_size, n);
4539 
4540 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4541 
4542 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4543 		    M_TEMP, M_NOWAIT | M_ZERO);
4544 		if (pfrts == NULL) {
4545 			error = ENOMEM;
4546 			PF_RULES_RUNLOCK();
4547 			break;
4548 		}
4549 		error = pfr_get_tables(&io->pfrio_table, pfrts,
4550 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4551 		PF_RULES_RUNLOCK();
4552 		if (error == 0)
4553 			error = copyout(pfrts, io->pfrio_buffer, totlen);
4554 		free(pfrts, M_TEMP);
4555 		break;
4556 	}
4557 
4558 	case DIOCRGETTSTATS: {
4559 		struct pfioc_table *io = (struct pfioc_table *)addr;
4560 		struct pfr_tstats *pfrtstats;
4561 		size_t totlen;
4562 		int n;
4563 
4564 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
4565 			error = ENODEV;
4566 			break;
4567 		}
4568 		PF_TABLE_STATS_LOCK();
4569 		PF_RULES_RLOCK();
4570 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4571 		if (n < 0) {
4572 			PF_RULES_RUNLOCK();
4573 			PF_TABLE_STATS_UNLOCK();
4574 			error = EINVAL;
4575 			break;
4576 		}
4577 		io->pfrio_size = min(io->pfrio_size, n);
4578 
4579 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
4580 		pfrtstats = mallocarray(io->pfrio_size,
4581 		    sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT | M_ZERO);
4582 		if (pfrtstats == NULL) {
4583 			error = ENOMEM;
4584 			PF_RULES_RUNLOCK();
4585 			PF_TABLE_STATS_UNLOCK();
4586 			break;
4587 		}
4588 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
4589 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4590 		PF_RULES_RUNLOCK();
4591 		PF_TABLE_STATS_UNLOCK();
4592 		if (error == 0)
4593 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
4594 		free(pfrtstats, M_TEMP);
4595 		break;
4596 	}
4597 
4598 	case DIOCRCLRTSTATS: {
4599 		struct pfioc_table *io = (struct pfioc_table *)addr;
4600 		struct pfr_table *pfrts;
4601 		size_t totlen;
4602 
4603 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4604 			error = ENODEV;
4605 			break;
4606 		}
4607 
4608 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4609 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4610 			/* We used to count tables and use the minimum required
4611 			 * size, so we didn't fail on overly large requests.
4612 			 * Keep doing so. */
4613 			io->pfrio_size = pf_ioctl_maxcount;
4614 			break;
4615 		}
4616 
4617 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4618 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4619 		    M_TEMP, M_WAITOK);
4620 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4621 		if (error) {
4622 			free(pfrts, M_TEMP);
4623 			break;
4624 		}
4625 
4626 		PF_TABLE_STATS_LOCK();
4627 		PF_RULES_RLOCK();
4628 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
4629 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4630 		PF_RULES_RUNLOCK();
4631 		PF_TABLE_STATS_UNLOCK();
4632 		free(pfrts, M_TEMP);
4633 		break;
4634 	}
4635 
4636 	case DIOCRSETTFLAGS: {
4637 		struct pfioc_table *io = (struct pfioc_table *)addr;
4638 		struct pfr_table *pfrts;
4639 		size_t totlen;
4640 		int n;
4641 
4642 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4643 			error = ENODEV;
4644 			break;
4645 		}
4646 
4647 		PF_RULES_RLOCK();
4648 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4649 		if (n < 0) {
4650 			PF_RULES_RUNLOCK();
4651 			error = EINVAL;
4652 			break;
4653 		}
4654 
4655 		io->pfrio_size = min(io->pfrio_size, n);
4656 		PF_RULES_RUNLOCK();
4657 
4658 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4659 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4660 		    M_TEMP, M_WAITOK);
4661 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4662 		if (error) {
4663 			free(pfrts, M_TEMP);
4664 			break;
4665 		}
4666 		PF_RULES_WLOCK();
4667 		error = pfr_set_tflags(pfrts, io->pfrio_size,
4668 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
4669 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4670 		PF_RULES_WUNLOCK();
4671 		free(pfrts, M_TEMP);
4672 		break;
4673 	}
4674 
4675 	case DIOCRCLRADDRS: {
4676 		struct pfioc_table *io = (struct pfioc_table *)addr;
4677 
4678 		if (io->pfrio_esize != 0) {
4679 			error = ENODEV;
4680 			break;
4681 		}
4682 		PF_RULES_WLOCK();
4683 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
4684 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4685 		PF_RULES_WUNLOCK();
4686 		break;
4687 	}
4688 
4689 	case DIOCRADDADDRS: {
4690 		struct pfioc_table *io = (struct pfioc_table *)addr;
4691 		struct pfr_addr *pfras;
4692 		size_t totlen;
4693 
4694 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4695 			error = ENODEV;
4696 			break;
4697 		}
4698 		if (io->pfrio_size < 0 ||
4699 		    io->pfrio_size > pf_ioctl_maxcount ||
4700 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4701 			error = EINVAL;
4702 			break;
4703 		}
4704 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4705 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4706 		    M_TEMP, M_WAITOK);
4707 		error = copyin(io->pfrio_buffer, pfras, totlen);
4708 		if (error) {
4709 			free(pfras, M_TEMP);
4710 			break;
4711 		}
4712 		PF_RULES_WLOCK();
4713 		error = pfr_add_addrs(&io->pfrio_table, pfras,
4714 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
4715 		    PFR_FLAG_USERIOCTL);
4716 		PF_RULES_WUNLOCK();
4717 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4718 			error = copyout(pfras, io->pfrio_buffer, totlen);
4719 		free(pfras, M_TEMP);
4720 		break;
4721 	}
4722 
4723 	case DIOCRDELADDRS: {
4724 		struct pfioc_table *io = (struct pfioc_table *)addr;
4725 		struct pfr_addr *pfras;
4726 		size_t totlen;
4727 
4728 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4729 			error = ENODEV;
4730 			break;
4731 		}
4732 		if (io->pfrio_size < 0 ||
4733 		    io->pfrio_size > pf_ioctl_maxcount ||
4734 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4735 			error = EINVAL;
4736 			break;
4737 		}
4738 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4739 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4740 		    M_TEMP, M_WAITOK);
4741 		error = copyin(io->pfrio_buffer, pfras, totlen);
4742 		if (error) {
4743 			free(pfras, M_TEMP);
4744 			break;
4745 		}
4746 		PF_RULES_WLOCK();
4747 		error = pfr_del_addrs(&io->pfrio_table, pfras,
4748 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
4749 		    PFR_FLAG_USERIOCTL);
4750 		PF_RULES_WUNLOCK();
4751 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4752 			error = copyout(pfras, io->pfrio_buffer, totlen);
4753 		free(pfras, M_TEMP);
4754 		break;
4755 	}
4756 
4757 	case DIOCRSETADDRS: {
4758 		struct pfioc_table *io = (struct pfioc_table *)addr;
4759 		struct pfr_addr *pfras;
4760 		size_t totlen, count;
4761 
4762 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4763 			error = ENODEV;
4764 			break;
4765 		}
4766 		if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
4767 			error = EINVAL;
4768 			break;
4769 		}
4770 		count = max(io->pfrio_size, io->pfrio_size2);
4771 		if (count > pf_ioctl_maxcount ||
4772 		    WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
4773 			error = EINVAL;
4774 			break;
4775 		}
4776 		totlen = count * sizeof(struct pfr_addr);
4777 		pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
4778 		    M_WAITOK);
4779 		error = copyin(io->pfrio_buffer, pfras, totlen);
4780 		if (error) {
4781 			free(pfras, M_TEMP);
4782 			break;
4783 		}
4784 		PF_RULES_WLOCK();
4785 		error = pfr_set_addrs(&io->pfrio_table, pfras,
4786 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
4787 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
4788 		    PFR_FLAG_USERIOCTL, 0);
4789 		PF_RULES_WUNLOCK();
4790 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4791 			error = copyout(pfras, io->pfrio_buffer, totlen);
4792 		free(pfras, M_TEMP);
4793 		break;
4794 	}
4795 
4796 	case DIOCRGETADDRS: {
4797 		struct pfioc_table *io = (struct pfioc_table *)addr;
4798 		struct pfr_addr *pfras;
4799 		size_t totlen;
4800 
4801 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4802 			error = ENODEV;
4803 			break;
4804 		}
4805 		if (io->pfrio_size < 0 ||
4806 		    io->pfrio_size > pf_ioctl_maxcount ||
4807 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4808 			error = EINVAL;
4809 			break;
4810 		}
4811 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4812 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4813 		    M_TEMP, M_WAITOK | M_ZERO);
4814 		PF_RULES_RLOCK();
4815 		error = pfr_get_addrs(&io->pfrio_table, pfras,
4816 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4817 		PF_RULES_RUNLOCK();
4818 		if (error == 0)
4819 			error = copyout(pfras, io->pfrio_buffer, totlen);
4820 		free(pfras, M_TEMP);
4821 		break;
4822 	}
4823 
4824 	case DIOCRGETASTATS: {
4825 		struct pfioc_table *io = (struct pfioc_table *)addr;
4826 		struct pfr_astats *pfrastats;
4827 		size_t totlen;
4828 
4829 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
4830 			error = ENODEV;
4831 			break;
4832 		}
4833 		if (io->pfrio_size < 0 ||
4834 		    io->pfrio_size > pf_ioctl_maxcount ||
4835 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
4836 			error = EINVAL;
4837 			break;
4838 		}
4839 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
4840 		pfrastats = mallocarray(io->pfrio_size,
4841 		    sizeof(struct pfr_astats), M_TEMP, M_WAITOK | M_ZERO);
4842 		PF_RULES_RLOCK();
4843 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
4844 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4845 		PF_RULES_RUNLOCK();
4846 		if (error == 0)
4847 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
4848 		free(pfrastats, M_TEMP);
4849 		break;
4850 	}
4851 
4852 	case DIOCRCLRASTATS: {
4853 		struct pfioc_table *io = (struct pfioc_table *)addr;
4854 		struct pfr_addr *pfras;
4855 		size_t totlen;
4856 
4857 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4858 			error = ENODEV;
4859 			break;
4860 		}
4861 		if (io->pfrio_size < 0 ||
4862 		    io->pfrio_size > pf_ioctl_maxcount ||
4863 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4864 			error = EINVAL;
4865 			break;
4866 		}
4867 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4868 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4869 		    M_TEMP, M_WAITOK);
4870 		error = copyin(io->pfrio_buffer, pfras, totlen);
4871 		if (error) {
4872 			free(pfras, M_TEMP);
4873 			break;
4874 		}
4875 		PF_RULES_WLOCK();
4876 		error = pfr_clr_astats(&io->pfrio_table, pfras,
4877 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
4878 		    PFR_FLAG_USERIOCTL);
4879 		PF_RULES_WUNLOCK();
4880 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4881 			error = copyout(pfras, io->pfrio_buffer, totlen);
4882 		free(pfras, M_TEMP);
4883 		break;
4884 	}
4885 
4886 	case DIOCRTSTADDRS: {
4887 		struct pfioc_table *io = (struct pfioc_table *)addr;
4888 		struct pfr_addr *pfras;
4889 		size_t totlen;
4890 
4891 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4892 			error = ENODEV;
4893 			break;
4894 		}
4895 		if (io->pfrio_size < 0 ||
4896 		    io->pfrio_size > pf_ioctl_maxcount ||
4897 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4898 			error = EINVAL;
4899 			break;
4900 		}
4901 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4902 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4903 		    M_TEMP, M_WAITOK);
4904 		error = copyin(io->pfrio_buffer, pfras, totlen);
4905 		if (error) {
4906 			free(pfras, M_TEMP);
4907 			break;
4908 		}
4909 		PF_RULES_RLOCK();
4910 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
4911 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
4912 		    PFR_FLAG_USERIOCTL);
4913 		PF_RULES_RUNLOCK();
4914 		if (error == 0)
4915 			error = copyout(pfras, io->pfrio_buffer, totlen);
4916 		free(pfras, M_TEMP);
4917 		break;
4918 	}
4919 
4920 	case DIOCRINADEFINE: {
4921 		struct pfioc_table *io = (struct pfioc_table *)addr;
4922 		struct pfr_addr *pfras;
4923 		size_t totlen;
4924 
4925 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4926 			error = ENODEV;
4927 			break;
4928 		}
4929 		if (io->pfrio_size < 0 ||
4930 		    io->pfrio_size > pf_ioctl_maxcount ||
4931 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4932 			error = EINVAL;
4933 			break;
4934 		}
4935 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4936 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4937 		    M_TEMP, M_WAITOK);
4938 		error = copyin(io->pfrio_buffer, pfras, totlen);
4939 		if (error) {
4940 			free(pfras, M_TEMP);
4941 			break;
4942 		}
4943 		PF_RULES_WLOCK();
4944 		error = pfr_ina_define(&io->pfrio_table, pfras,
4945 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
4946 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4947 		PF_RULES_WUNLOCK();
4948 		free(pfras, M_TEMP);
4949 		break;
4950 	}
4951 
4952 	case DIOCOSFPADD: {
4953 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
4954 		PF_RULES_WLOCK();
4955 		error = pf_osfp_add(io);
4956 		PF_RULES_WUNLOCK();
4957 		break;
4958 	}
4959 
4960 	case DIOCOSFPGET: {
4961 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
4962 		PF_RULES_RLOCK();
4963 		error = pf_osfp_get(io);
4964 		PF_RULES_RUNLOCK();
4965 		break;
4966 	}
4967 
4968 	case DIOCXBEGIN: {
4969 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
4970 		struct pfioc_trans_e	*ioes, *ioe;
4971 		size_t			 totlen;
4972 		int			 i;
4973 
4974 		if (io->esize != sizeof(*ioe)) {
4975 			error = ENODEV;
4976 			break;
4977 		}
4978 		if (io->size < 0 ||
4979 		    io->size > pf_ioctl_maxcount ||
4980 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
4981 			error = EINVAL;
4982 			break;
4983 		}
4984 		totlen = sizeof(struct pfioc_trans_e) * io->size;
4985 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
4986 		    M_TEMP, M_WAITOK);
4987 		error = copyin(io->array, ioes, totlen);
4988 		if (error) {
4989 			free(ioes, M_TEMP);
4990 			break;
4991 		}
4992 		/* Ensure there's no more ethernet rules to clean up. */
4993 		NET_EPOCH_DRAIN_CALLBACKS();
4994 		PF_RULES_WLOCK();
4995 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
4996 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
4997 			switch (ioe->rs_num) {
4998 			case PF_RULESET_ETH:
4999 				if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) {
5000 					PF_RULES_WUNLOCK();
5001 					free(ioes, M_TEMP);
5002 					goto fail;
5003 				}
5004 				break;
5005 #ifdef ALTQ
5006 			case PF_RULESET_ALTQ:
5007 				if (ioe->anchor[0]) {
5008 					PF_RULES_WUNLOCK();
5009 					free(ioes, M_TEMP);
5010 					error = EINVAL;
5011 					goto fail;
5012 				}
5013 				if ((error = pf_begin_altq(&ioe->ticket))) {
5014 					PF_RULES_WUNLOCK();
5015 					free(ioes, M_TEMP);
5016 					goto fail;
5017 				}
5018 				break;
5019 #endif /* ALTQ */
5020 			case PF_RULESET_TABLE:
5021 			    {
5022 				struct pfr_table table;
5023 
5024 				bzero(&table, sizeof(table));
5025 				strlcpy(table.pfrt_anchor, ioe->anchor,
5026 				    sizeof(table.pfrt_anchor));
5027 				if ((error = pfr_ina_begin(&table,
5028 				    &ioe->ticket, NULL, 0))) {
5029 					PF_RULES_WUNLOCK();
5030 					free(ioes, M_TEMP);
5031 					goto fail;
5032 				}
5033 				break;
5034 			    }
5035 			default:
5036 				if ((error = pf_begin_rules(&ioe->ticket,
5037 				    ioe->rs_num, ioe->anchor))) {
5038 					PF_RULES_WUNLOCK();
5039 					free(ioes, M_TEMP);
5040 					goto fail;
5041 				}
5042 				break;
5043 			}
5044 		}
5045 		PF_RULES_WUNLOCK();
5046 		error = copyout(ioes, io->array, totlen);
5047 		free(ioes, M_TEMP);
5048 		break;
5049 	}
5050 
5051 	case DIOCXROLLBACK: {
5052 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5053 		struct pfioc_trans_e	*ioe, *ioes;
5054 		size_t			 totlen;
5055 		int			 i;
5056 
5057 		if (io->esize != sizeof(*ioe)) {
5058 			error = ENODEV;
5059 			break;
5060 		}
5061 		if (io->size < 0 ||
5062 		    io->size > pf_ioctl_maxcount ||
5063 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5064 			error = EINVAL;
5065 			break;
5066 		}
5067 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5068 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5069 		    M_TEMP, M_WAITOK);
5070 		error = copyin(io->array, ioes, totlen);
5071 		if (error) {
5072 			free(ioes, M_TEMP);
5073 			break;
5074 		}
5075 		PF_RULES_WLOCK();
5076 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5077 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5078 			switch (ioe->rs_num) {
5079 			case PF_RULESET_ETH:
5080 				if ((error = pf_rollback_eth(ioe->ticket,
5081 				    ioe->anchor))) {
5082 					PF_RULES_WUNLOCK();
5083 					free(ioes, M_TEMP);
5084 					goto fail; /* really bad */
5085 				}
5086 				break;
5087 #ifdef ALTQ
5088 			case PF_RULESET_ALTQ:
5089 				if (ioe->anchor[0]) {
5090 					PF_RULES_WUNLOCK();
5091 					free(ioes, M_TEMP);
5092 					error = EINVAL;
5093 					goto fail;
5094 				}
5095 				if ((error = pf_rollback_altq(ioe->ticket))) {
5096 					PF_RULES_WUNLOCK();
5097 					free(ioes, M_TEMP);
5098 					goto fail; /* really bad */
5099 				}
5100 				break;
5101 #endif /* ALTQ */
5102 			case PF_RULESET_TABLE:
5103 			    {
5104 				struct pfr_table table;
5105 
5106 				bzero(&table, sizeof(table));
5107 				strlcpy(table.pfrt_anchor, ioe->anchor,
5108 				    sizeof(table.pfrt_anchor));
5109 				if ((error = pfr_ina_rollback(&table,
5110 				    ioe->ticket, NULL, 0))) {
5111 					PF_RULES_WUNLOCK();
5112 					free(ioes, M_TEMP);
5113 					goto fail; /* really bad */
5114 				}
5115 				break;
5116 			    }
5117 			default:
5118 				if ((error = pf_rollback_rules(ioe->ticket,
5119 				    ioe->rs_num, ioe->anchor))) {
5120 					PF_RULES_WUNLOCK();
5121 					free(ioes, M_TEMP);
5122 					goto fail; /* really bad */
5123 				}
5124 				break;
5125 			}
5126 		}
5127 		PF_RULES_WUNLOCK();
5128 		free(ioes, M_TEMP);
5129 		break;
5130 	}
5131 
5132 	case DIOCXCOMMIT: {
5133 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5134 		struct pfioc_trans_e	*ioe, *ioes;
5135 		struct pf_kruleset	*rs;
5136 		struct pf_keth_ruleset	*ers;
5137 		size_t			 totlen;
5138 		int			 i;
5139 
5140 		if (io->esize != sizeof(*ioe)) {
5141 			error = ENODEV;
5142 			break;
5143 		}
5144 
5145 		if (io->size < 0 ||
5146 		    io->size > pf_ioctl_maxcount ||
5147 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5148 			error = EINVAL;
5149 			break;
5150 		}
5151 
5152 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5153 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5154 		    M_TEMP, M_WAITOK);
5155 		error = copyin(io->array, ioes, totlen);
5156 		if (error) {
5157 			free(ioes, M_TEMP);
5158 			break;
5159 		}
5160 		PF_RULES_WLOCK();
5161 		/* First makes sure everything will succeed. */
5162 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5163 			ioe->anchor[sizeof(ioe->anchor) - 1] = 0;
5164 			switch (ioe->rs_num) {
5165 			case PF_RULESET_ETH:
5166 				ers = pf_find_keth_ruleset(ioe->anchor);
5167 				if (ers == NULL || ioe->ticket == 0 ||
5168 				    ioe->ticket != ers->inactive.ticket) {
5169 					PF_RULES_WUNLOCK();
5170 					free(ioes, M_TEMP);
5171 					error = EINVAL;
5172 					goto fail;
5173 				}
5174 				break;
5175 #ifdef ALTQ
5176 			case PF_RULESET_ALTQ:
5177 				if (ioe->anchor[0]) {
5178 					PF_RULES_WUNLOCK();
5179 					free(ioes, M_TEMP);
5180 					error = EINVAL;
5181 					goto fail;
5182 				}
5183 				if (!V_altqs_inactive_open || ioe->ticket !=
5184 				    V_ticket_altqs_inactive) {
5185 					PF_RULES_WUNLOCK();
5186 					free(ioes, M_TEMP);
5187 					error = EBUSY;
5188 					goto fail;
5189 				}
5190 				break;
5191 #endif /* ALTQ */
5192 			case PF_RULESET_TABLE:
5193 				rs = pf_find_kruleset(ioe->anchor);
5194 				if (rs == NULL || !rs->topen || ioe->ticket !=
5195 				    rs->tticket) {
5196 					PF_RULES_WUNLOCK();
5197 					free(ioes, M_TEMP);
5198 					error = EBUSY;
5199 					goto fail;
5200 				}
5201 				break;
5202 			default:
5203 				if (ioe->rs_num < 0 || ioe->rs_num >=
5204 				    PF_RULESET_MAX) {
5205 					PF_RULES_WUNLOCK();
5206 					free(ioes, M_TEMP);
5207 					error = EINVAL;
5208 					goto fail;
5209 				}
5210 				rs = pf_find_kruleset(ioe->anchor);
5211 				if (rs == NULL ||
5212 				    !rs->rules[ioe->rs_num].inactive.open ||
5213 				    rs->rules[ioe->rs_num].inactive.ticket !=
5214 				    ioe->ticket) {
5215 					PF_RULES_WUNLOCK();
5216 					free(ioes, M_TEMP);
5217 					error = EBUSY;
5218 					goto fail;
5219 				}
5220 				break;
5221 			}
5222 		}
5223 		/* Now do the commit - no errors should happen here. */
5224 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5225 			switch (ioe->rs_num) {
5226 			case PF_RULESET_ETH:
5227 				if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) {
5228 					PF_RULES_WUNLOCK();
5229 					free(ioes, M_TEMP);
5230 					goto fail; /* really bad */
5231 				}
5232 				break;
5233 #ifdef ALTQ
5234 			case PF_RULESET_ALTQ:
5235 				if ((error = pf_commit_altq(ioe->ticket))) {
5236 					PF_RULES_WUNLOCK();
5237 					free(ioes, M_TEMP);
5238 					goto fail; /* really bad */
5239 				}
5240 				break;
5241 #endif /* ALTQ */
5242 			case PF_RULESET_TABLE:
5243 			    {
5244 				struct pfr_table table;
5245 
5246 				bzero(&table, sizeof(table));
5247 				(void)strlcpy(table.pfrt_anchor, ioe->anchor,
5248 				    sizeof(table.pfrt_anchor));
5249 				if ((error = pfr_ina_commit(&table,
5250 				    ioe->ticket, NULL, NULL, 0))) {
5251 					PF_RULES_WUNLOCK();
5252 					free(ioes, M_TEMP);
5253 					goto fail; /* really bad */
5254 				}
5255 				break;
5256 			    }
5257 			default:
5258 				if ((error = pf_commit_rules(ioe->ticket,
5259 				    ioe->rs_num, ioe->anchor))) {
5260 					PF_RULES_WUNLOCK();
5261 					free(ioes, M_TEMP);
5262 					goto fail; /* really bad */
5263 				}
5264 				break;
5265 			}
5266 		}
5267 		PF_RULES_WUNLOCK();
5268 
5269 		/* Only hook into EtherNet taffic if we've got rules for it. */
5270 		if (! TAILQ_EMPTY(V_pf_keth->active.rules))
5271 			hook_pf_eth();
5272 		else
5273 			dehook_pf_eth();
5274 
5275 		free(ioes, M_TEMP);
5276 		break;
5277 	}
5278 
5279 	case DIOCGETSRCNODES: {
5280 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
5281 		struct pf_srchash	*sh;
5282 		struct pf_ksrc_node	*n;
5283 		struct pf_src_node	*p, *pstore;
5284 		uint32_t		 i, nr = 0;
5285 
5286 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5287 				i++, sh++) {
5288 			PF_HASHROW_LOCK(sh);
5289 			LIST_FOREACH(n, &sh->nodes, entry)
5290 				nr++;
5291 			PF_HASHROW_UNLOCK(sh);
5292 		}
5293 
5294 		psn->psn_len = min(psn->psn_len,
5295 		    sizeof(struct pf_src_node) * nr);
5296 
5297 		if (psn->psn_len == 0) {
5298 			psn->psn_len = sizeof(struct pf_src_node) * nr;
5299 			break;
5300 		}
5301 
5302 		nr = 0;
5303 
5304 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
5305 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5306 		    i++, sh++) {
5307 		    PF_HASHROW_LOCK(sh);
5308 		    LIST_FOREACH(n, &sh->nodes, entry) {
5309 
5310 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
5311 				break;
5312 
5313 			pf_src_node_copy(n, p);
5314 
5315 			p++;
5316 			nr++;
5317 		    }
5318 		    PF_HASHROW_UNLOCK(sh);
5319 		}
5320 		error = copyout(pstore, psn->psn_src_nodes,
5321 		    sizeof(struct pf_src_node) * nr);
5322 		if (error) {
5323 			free(pstore, M_TEMP);
5324 			break;
5325 		}
5326 		psn->psn_len = sizeof(struct pf_src_node) * nr;
5327 		free(pstore, M_TEMP);
5328 		break;
5329 	}
5330 
5331 	case DIOCCLRSRCNODES: {
5332 		pf_clear_srcnodes(NULL);
5333 		pf_purge_expired_src_nodes();
5334 		break;
5335 	}
5336 
5337 	case DIOCKILLSRCNODES:
5338 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
5339 		break;
5340 
5341 #ifdef COMPAT_FREEBSD13
5342 	case DIOCKEEPCOUNTERS_FREEBSD13:
5343 #endif
5344 	case DIOCKEEPCOUNTERS:
5345 		error = pf_keepcounters((struct pfioc_nv *)addr);
5346 		break;
5347 
5348 	case DIOCGETSYNCOOKIES:
5349 		error = pf_get_syncookies((struct pfioc_nv *)addr);
5350 		break;
5351 
5352 	case DIOCSETSYNCOOKIES:
5353 		error = pf_set_syncookies((struct pfioc_nv *)addr);
5354 		break;
5355 
5356 	case DIOCSETHOSTID: {
5357 		u_int32_t	*hostid = (u_int32_t *)addr;
5358 
5359 		PF_RULES_WLOCK();
5360 		if (*hostid == 0)
5361 			V_pf_status.hostid = arc4random();
5362 		else
5363 			V_pf_status.hostid = *hostid;
5364 		PF_RULES_WUNLOCK();
5365 		break;
5366 	}
5367 
5368 	case DIOCOSFPFLUSH:
5369 		PF_RULES_WLOCK();
5370 		pf_osfp_flush();
5371 		PF_RULES_WUNLOCK();
5372 		break;
5373 
5374 	case DIOCIGETIFACES: {
5375 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5376 		struct pfi_kif *ifstore;
5377 		size_t bufsiz;
5378 
5379 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
5380 			error = ENODEV;
5381 			break;
5382 		}
5383 
5384 		if (io->pfiio_size < 0 ||
5385 		    io->pfiio_size > pf_ioctl_maxcount ||
5386 		    WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
5387 			error = EINVAL;
5388 			break;
5389 		}
5390 
5391 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5392 
5393 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
5394 		ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
5395 		    M_TEMP, M_WAITOK | M_ZERO);
5396 
5397 		PF_RULES_RLOCK();
5398 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
5399 		PF_RULES_RUNLOCK();
5400 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
5401 		free(ifstore, M_TEMP);
5402 		break;
5403 	}
5404 
5405 	case DIOCSETIFFLAG: {
5406 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5407 
5408 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5409 
5410 		PF_RULES_WLOCK();
5411 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
5412 		PF_RULES_WUNLOCK();
5413 		break;
5414 	}
5415 
5416 	case DIOCCLRIFFLAG: {
5417 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5418 
5419 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5420 
5421 		PF_RULES_WLOCK();
5422 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
5423 		PF_RULES_WUNLOCK();
5424 		break;
5425 	}
5426 
5427 	case DIOCSETREASS: {
5428 		u_int32_t	*reass = (u_int32_t *)addr;
5429 
5430 		V_pf_status.reass = *reass & (PF_REASS_ENABLED|PF_REASS_NODF);
5431 		/* Removal of DF flag without reassembly enabled is not a
5432 		 * valid combination. Disable reassembly in such case. */
5433 		if (!(V_pf_status.reass & PF_REASS_ENABLED))
5434 			V_pf_status.reass = 0;
5435 		break;
5436 	}
5437 
5438 	default:
5439 		error = ENODEV;
5440 		break;
5441 	}
5442 fail:
5443 	CURVNET_RESTORE();
5444 
5445 #undef ERROUT_IOCTL
5446 
5447 	return (error);
5448 }
5449 
5450 void
5451 pfsync_state_export(union pfsync_state_union *sp, struct pf_kstate *st, int msg_version)
5452 {
5453 	bzero(sp, sizeof(union pfsync_state_union));
5454 
5455 	/* copy from state key */
5456 	sp->pfs_1301.key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5457 	sp->pfs_1301.key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5458 	sp->pfs_1301.key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5459 	sp->pfs_1301.key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5460 	sp->pfs_1301.key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5461 	sp->pfs_1301.key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5462 	sp->pfs_1301.key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5463 	sp->pfs_1301.key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5464 	sp->pfs_1301.proto = st->key[PF_SK_WIRE]->proto;
5465 	sp->pfs_1301.af = st->key[PF_SK_WIRE]->af;
5466 
5467 	/* copy from state */
5468 	strlcpy(sp->pfs_1301.ifname, st->kif->pfik_name, sizeof(sp->pfs_1301.ifname));
5469 	bcopy(&st->rt_addr, &sp->pfs_1301.rt_addr, sizeof(sp->pfs_1301.rt_addr));
5470 	sp->pfs_1301.creation = htonl(time_uptime - st->creation);
5471 	sp->pfs_1301.expire = pf_state_expires(st);
5472 	if (sp->pfs_1301.expire <= time_uptime)
5473 		sp->pfs_1301.expire = htonl(0);
5474 	else
5475 		sp->pfs_1301.expire = htonl(sp->pfs_1301.expire - time_uptime);
5476 
5477 	sp->pfs_1301.direction = st->direction;
5478 	sp->pfs_1301.log = st->act.log;
5479 	sp->pfs_1301.timeout = st->timeout;
5480 
5481 	switch (msg_version) {
5482 		case PFSYNC_MSG_VERSION_1301:
5483 			sp->pfs_1301.state_flags = st->state_flags;
5484 			break;
5485 		case PFSYNC_MSG_VERSION_1400:
5486 			sp->pfs_1400.state_flags = htons(st->state_flags);
5487 			sp->pfs_1400.qid = htons(st->act.qid);
5488 			sp->pfs_1400.pqid = htons(st->act.pqid);
5489 			sp->pfs_1400.dnpipe = htons(st->act.dnpipe);
5490 			sp->pfs_1400.dnrpipe = htons(st->act.dnrpipe);
5491 			sp->pfs_1400.rtableid = htonl(st->act.rtableid);
5492 			sp->pfs_1400.min_ttl = st->act.min_ttl;
5493 			sp->pfs_1400.set_tos = st->act.set_tos;
5494 			sp->pfs_1400.max_mss = htons(st->act.max_mss);
5495 			sp->pfs_1400.set_prio[0] = st->act.set_prio[0];
5496 			sp->pfs_1400.set_prio[1] = st->act.set_prio[1];
5497 			sp->pfs_1400.rt = st->rt;
5498 			if (st->rt_kif)
5499 				strlcpy(sp->pfs_1400.rt_ifname,
5500 				    st->rt_kif->pfik_name,
5501 				    sizeof(sp->pfs_1400.rt_ifname));
5502 			break;
5503 		default:
5504 			panic("%s: Unsupported pfsync_msg_version %d",
5505 			    __func__, msg_version);
5506 	}
5507 
5508 	if (st->src_node)
5509 		sp->pfs_1301.sync_flags |= PFSYNC_FLAG_SRCNODE;
5510 	if (st->nat_src_node)
5511 		sp->pfs_1301.sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5512 
5513 	sp->pfs_1301.id = st->id;
5514 	sp->pfs_1301.creatorid = st->creatorid;
5515 	pf_state_peer_hton(&st->src, &sp->pfs_1301.src);
5516 	pf_state_peer_hton(&st->dst, &sp->pfs_1301.dst);
5517 
5518 	if (st->rule.ptr == NULL)
5519 		sp->pfs_1301.rule = htonl(-1);
5520 	else
5521 		sp->pfs_1301.rule = htonl(st->rule.ptr->nr);
5522 	if (st->anchor.ptr == NULL)
5523 		sp->pfs_1301.anchor = htonl(-1);
5524 	else
5525 		sp->pfs_1301.anchor = htonl(st->anchor.ptr->nr);
5526 	if (st->nat_rule.ptr == NULL)
5527 		sp->pfs_1301.nat_rule = htonl(-1);
5528 	else
5529 		sp->pfs_1301.nat_rule = htonl(st->nat_rule.ptr->nr);
5530 
5531 	pf_state_counter_hton(st->packets[0], sp->pfs_1301.packets[0]);
5532 	pf_state_counter_hton(st->packets[1], sp->pfs_1301.packets[1]);
5533 	pf_state_counter_hton(st->bytes[0], sp->pfs_1301.bytes[0]);
5534 	pf_state_counter_hton(st->bytes[1], sp->pfs_1301.bytes[1]);
5535 }
5536 
5537 void
5538 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st)
5539 {
5540 	bzero(sp, sizeof(*sp));
5541 
5542 	sp->version = PF_STATE_VERSION;
5543 
5544 	/* copy from state key */
5545 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5546 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5547 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5548 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5549 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5550 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5551 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5552 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5553 	sp->proto = st->key[PF_SK_WIRE]->proto;
5554 	sp->af = st->key[PF_SK_WIRE]->af;
5555 
5556 	/* copy from state */
5557 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
5558 	strlcpy(sp->orig_ifname, st->orig_kif->pfik_name,
5559 	    sizeof(sp->orig_ifname));
5560 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
5561 	sp->creation = htonl(time_uptime - st->creation);
5562 	sp->expire = pf_state_expires(st);
5563 	if (sp->expire <= time_uptime)
5564 		sp->expire = htonl(0);
5565 	else
5566 		sp->expire = htonl(sp->expire - time_uptime);
5567 
5568 	sp->direction = st->direction;
5569 	sp->log = st->act.log;
5570 	sp->timeout = st->timeout;
5571 	/* 8 bits for the old libpfctl, 16 bits for the new libpfctl */
5572 	sp->state_flags_compat = st->state_flags;
5573 	sp->state_flags = htons(st->state_flags);
5574 	if (st->src_node)
5575 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
5576 	if (st->nat_src_node)
5577 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5578 
5579 	sp->id = st->id;
5580 	sp->creatorid = st->creatorid;
5581 	pf_state_peer_hton(&st->src, &sp->src);
5582 	pf_state_peer_hton(&st->dst, &sp->dst);
5583 
5584 	if (st->rule.ptr == NULL)
5585 		sp->rule = htonl(-1);
5586 	else
5587 		sp->rule = htonl(st->rule.ptr->nr);
5588 	if (st->anchor.ptr == NULL)
5589 		sp->anchor = htonl(-1);
5590 	else
5591 		sp->anchor = htonl(st->anchor.ptr->nr);
5592 	if (st->nat_rule.ptr == NULL)
5593 		sp->nat_rule = htonl(-1);
5594 	else
5595 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
5596 
5597 	sp->packets[0] = st->packets[0];
5598 	sp->packets[1] = st->packets[1];
5599 	sp->bytes[0] = st->bytes[0];
5600 	sp->bytes[1] = st->bytes[1];
5601 
5602 	sp->qid = htons(st->act.qid);
5603 	sp->pqid = htons(st->act.pqid);
5604 	sp->dnpipe = htons(st->act.dnpipe);
5605 	sp->dnrpipe = htons(st->act.dnrpipe);
5606 	sp->rtableid = htonl(st->act.rtableid);
5607 	sp->min_ttl = st->act.min_ttl;
5608 	sp->set_tos = st->act.set_tos;
5609 	sp->max_mss = htons(st->act.max_mss);
5610 	sp->rt = st->rt;
5611 	if (st->rt_kif)
5612 		strlcpy(sp->rt_ifname, st->rt_kif->pfik_name,
5613 		    sizeof(sp->rt_ifname));
5614 	sp->set_prio[0] = st->act.set_prio[0];
5615 	sp->set_prio[1] = st->act.set_prio[1];
5616 
5617 }
5618 
5619 static void
5620 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
5621 {
5622 	struct pfr_ktable *kt;
5623 
5624 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
5625 
5626 	kt = aw->p.tbl;
5627 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
5628 		kt = kt->pfrkt_root;
5629 	aw->p.tbl = NULL;
5630 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
5631 		kt->pfrkt_cnt : -1;
5632 }
5633 
5634 static int
5635 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters,
5636     size_t number, char **names)
5637 {
5638 	nvlist_t        *nvc;
5639 
5640 	nvc = nvlist_create(0);
5641 	if (nvc == NULL)
5642 		return (ENOMEM);
5643 
5644 	for (int i = 0; i < number; i++) {
5645 		nvlist_append_number_array(nvc, "counters",
5646 		    counter_u64_fetch(counters[i]));
5647 		nvlist_append_string_array(nvc, "names",
5648 		    names[i]);
5649 		nvlist_append_number_array(nvc, "ids",
5650 		    i);
5651 	}
5652 	nvlist_add_nvlist(nvl, name, nvc);
5653 	nvlist_destroy(nvc);
5654 
5655 	return (0);
5656 }
5657 
5658 static int
5659 pf_getstatus(struct pfioc_nv *nv)
5660 {
5661 	nvlist_t        *nvl = NULL, *nvc = NULL;
5662 	void            *nvlpacked = NULL;
5663 	int              error;
5664 	struct pf_status s;
5665 	char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES;
5666 	char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES;
5667 	char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES;
5668 	PF_RULES_RLOCK_TRACKER;
5669 
5670 #define ERROUT(x)      ERROUT_FUNCTION(errout, x)
5671 
5672 	PF_RULES_RLOCK();
5673 
5674 	nvl = nvlist_create(0);
5675 	if (nvl == NULL)
5676 		ERROUT(ENOMEM);
5677 
5678 	nvlist_add_bool(nvl, "running", V_pf_status.running);
5679 	nvlist_add_number(nvl, "since", V_pf_status.since);
5680 	nvlist_add_number(nvl, "debug", V_pf_status.debug);
5681 	nvlist_add_number(nvl, "hostid", V_pf_status.hostid);
5682 	nvlist_add_number(nvl, "states", V_pf_status.states);
5683 	nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes);
5684 	nvlist_add_number(nvl, "reass", V_pf_status.reass);
5685 	nvlist_add_bool(nvl, "syncookies_active",
5686 	    V_pf_status.syncookies_active);
5687 
5688 	/* counters */
5689 	error = pf_add_status_counters(nvl, "counters", V_pf_status.counters,
5690 	    PFRES_MAX, pf_reasons);
5691 	if (error != 0)
5692 		ERROUT(error);
5693 
5694 	/* lcounters */
5695 	error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters,
5696 	    KLCNT_MAX, pf_lcounter);
5697 	if (error != 0)
5698 		ERROUT(error);
5699 
5700 	/* fcounters */
5701 	nvc = nvlist_create(0);
5702 	if (nvc == NULL)
5703 		ERROUT(ENOMEM);
5704 
5705 	for (int i = 0; i < FCNT_MAX; i++) {
5706 		nvlist_append_number_array(nvc, "counters",
5707 		    pf_counter_u64_fetch(&V_pf_status.fcounters[i]));
5708 		nvlist_append_string_array(nvc, "names",
5709 		    pf_fcounter[i]);
5710 		nvlist_append_number_array(nvc, "ids",
5711 		    i);
5712 	}
5713 	nvlist_add_nvlist(nvl, "fcounters", nvc);
5714 	nvlist_destroy(nvc);
5715 	nvc = NULL;
5716 
5717 	/* scounters */
5718 	error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters,
5719 	    SCNT_MAX, pf_fcounter);
5720 	if (error != 0)
5721 		ERROUT(error);
5722 
5723 	nvlist_add_string(nvl, "ifname", V_pf_status.ifname);
5724 	nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum,
5725 	    PF_MD5_DIGEST_LENGTH);
5726 
5727 	pfi_update_status(V_pf_status.ifname, &s);
5728 
5729 	/* pcounters / bcounters */
5730 	for (int i = 0; i < 2; i++) {
5731 		for (int j = 0; j < 2; j++) {
5732 			for (int k = 0; k < 2; k++) {
5733 				nvlist_append_number_array(nvl, "pcounters",
5734 				    s.pcounters[i][j][k]);
5735 			}
5736 			nvlist_append_number_array(nvl, "bcounters",
5737 			    s.bcounters[i][j]);
5738 		}
5739 	}
5740 
5741 	nvlpacked = nvlist_pack(nvl, &nv->len);
5742 	if (nvlpacked == NULL)
5743 		ERROUT(ENOMEM);
5744 
5745 	if (nv->size == 0)
5746 		ERROUT(0);
5747 	else if (nv->size < nv->len)
5748 		ERROUT(ENOSPC);
5749 
5750 	PF_RULES_RUNLOCK();
5751 	error = copyout(nvlpacked, nv->data, nv->len);
5752 	goto done;
5753 
5754 #undef ERROUT
5755 errout:
5756 	PF_RULES_RUNLOCK();
5757 done:
5758 	free(nvlpacked, M_NVLIST);
5759 	nvlist_destroy(nvc);
5760 	nvlist_destroy(nvl);
5761 
5762 	return (error);
5763 }
5764 
5765 /*
5766  * XXX - Check for version mismatch!!!
5767  */
5768 static void
5769 pf_clear_all_states(void)
5770 {
5771 	struct pf_kstate	*s;
5772 	u_int i;
5773 
5774 	for (i = 0; i <= pf_hashmask; i++) {
5775 		struct pf_idhash *ih = &V_pf_idhash[i];
5776 relock:
5777 		PF_HASHROW_LOCK(ih);
5778 		LIST_FOREACH(s, &ih->states, entry) {
5779 			s->timeout = PFTM_PURGE;
5780 			/* Don't send out individual delete messages. */
5781 			s->state_flags |= PFSTATE_NOSYNC;
5782 			pf_unlink_state(s);
5783 			goto relock;
5784 		}
5785 		PF_HASHROW_UNLOCK(ih);
5786 	}
5787 }
5788 
5789 static int
5790 pf_clear_tables(void)
5791 {
5792 	struct pfioc_table io;
5793 	int error;
5794 
5795 	bzero(&io, sizeof(io));
5796 
5797 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
5798 	    io.pfrio_flags);
5799 
5800 	return (error);
5801 }
5802 
5803 static void
5804 pf_clear_srcnodes(struct pf_ksrc_node *n)
5805 {
5806 	struct pf_kstate *s;
5807 	int i;
5808 
5809 	for (i = 0; i <= pf_hashmask; i++) {
5810 		struct pf_idhash *ih = &V_pf_idhash[i];
5811 
5812 		PF_HASHROW_LOCK(ih);
5813 		LIST_FOREACH(s, &ih->states, entry) {
5814 			if (n == NULL || n == s->src_node)
5815 				s->src_node = NULL;
5816 			if (n == NULL || n == s->nat_src_node)
5817 				s->nat_src_node = NULL;
5818 		}
5819 		PF_HASHROW_UNLOCK(ih);
5820 	}
5821 
5822 	if (n == NULL) {
5823 		struct pf_srchash *sh;
5824 
5825 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5826 		    i++, sh++) {
5827 			PF_HASHROW_LOCK(sh);
5828 			LIST_FOREACH(n, &sh->nodes, entry) {
5829 				n->expire = 1;
5830 				n->states = 0;
5831 			}
5832 			PF_HASHROW_UNLOCK(sh);
5833 		}
5834 	} else {
5835 		/* XXX: hash slot should already be locked here. */
5836 		n->expire = 1;
5837 		n->states = 0;
5838 	}
5839 }
5840 
5841 static void
5842 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
5843 {
5844 	struct pf_ksrc_node_list	 kill;
5845 
5846 	LIST_INIT(&kill);
5847 	for (int i = 0; i <= pf_srchashmask; i++) {
5848 		struct pf_srchash *sh = &V_pf_srchash[i];
5849 		struct pf_ksrc_node *sn, *tmp;
5850 
5851 		PF_HASHROW_LOCK(sh);
5852 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
5853 			if (PF_MATCHA(psnk->psnk_src.neg,
5854 			      &psnk->psnk_src.addr.v.a.addr,
5855 			      &psnk->psnk_src.addr.v.a.mask,
5856 			      &sn->addr, sn->af) &&
5857 			    PF_MATCHA(psnk->psnk_dst.neg,
5858 			      &psnk->psnk_dst.addr.v.a.addr,
5859 			      &psnk->psnk_dst.addr.v.a.mask,
5860 			      &sn->raddr, sn->af)) {
5861 				pf_unlink_src_node(sn);
5862 				LIST_INSERT_HEAD(&kill, sn, entry);
5863 				sn->expire = 1;
5864 			}
5865 		PF_HASHROW_UNLOCK(sh);
5866 	}
5867 
5868 	for (int i = 0; i <= pf_hashmask; i++) {
5869 		struct pf_idhash *ih = &V_pf_idhash[i];
5870 		struct pf_kstate *s;
5871 
5872 		PF_HASHROW_LOCK(ih);
5873 		LIST_FOREACH(s, &ih->states, entry) {
5874 			if (s->src_node && s->src_node->expire == 1)
5875 				s->src_node = NULL;
5876 			if (s->nat_src_node && s->nat_src_node->expire == 1)
5877 				s->nat_src_node = NULL;
5878 		}
5879 		PF_HASHROW_UNLOCK(ih);
5880 	}
5881 
5882 	psnk->psnk_killed = pf_free_src_nodes(&kill);
5883 }
5884 
5885 static int
5886 pf_keepcounters(struct pfioc_nv *nv)
5887 {
5888 	nvlist_t	*nvl = NULL;
5889 	void		*nvlpacked = NULL;
5890 	int		 error = 0;
5891 
5892 #define	ERROUT(x)	ERROUT_FUNCTION(on_error, x)
5893 
5894 	if (nv->len > pf_ioctl_maxcount)
5895 		ERROUT(ENOMEM);
5896 
5897 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
5898 	if (nvlpacked == NULL)
5899 		ERROUT(ENOMEM);
5900 
5901 	error = copyin(nv->data, nvlpacked, nv->len);
5902 	if (error)
5903 		ERROUT(error);
5904 
5905 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
5906 	if (nvl == NULL)
5907 		ERROUT(EBADMSG);
5908 
5909 	if (! nvlist_exists_bool(nvl, "keep_counters"))
5910 		ERROUT(EBADMSG);
5911 
5912 	V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters");
5913 
5914 on_error:
5915 	nvlist_destroy(nvl);
5916 	free(nvlpacked, M_NVLIST);
5917 	return (error);
5918 }
5919 
5920 static unsigned int
5921 pf_clear_states(const struct pf_kstate_kill *kill)
5922 {
5923 	struct pf_state_key_cmp	 match_key;
5924 	struct pf_kstate	*s;
5925 	struct pfi_kkif	*kif;
5926 	int		 idx;
5927 	unsigned int	 killed = 0, dir;
5928 
5929 	for (unsigned int i = 0; i <= pf_hashmask; i++) {
5930 		struct pf_idhash *ih = &V_pf_idhash[i];
5931 
5932 relock_DIOCCLRSTATES:
5933 		PF_HASHROW_LOCK(ih);
5934 		LIST_FOREACH(s, &ih->states, entry) {
5935 			/* For floating states look at the original kif. */
5936 			kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
5937 
5938 			if (kill->psk_ifname[0] &&
5939 			    strcmp(kill->psk_ifname,
5940 			    kif->pfik_name))
5941 				continue;
5942 
5943 			if (kill->psk_kill_match) {
5944 				bzero(&match_key, sizeof(match_key));
5945 
5946 				if (s->direction == PF_OUT) {
5947 					dir = PF_IN;
5948 					idx = PF_SK_STACK;
5949 				} else {
5950 					dir = PF_OUT;
5951 					idx = PF_SK_WIRE;
5952 				}
5953 
5954 				match_key.af = s->key[idx]->af;
5955 				match_key.proto = s->key[idx]->proto;
5956 				PF_ACPY(&match_key.addr[0],
5957 				    &s->key[idx]->addr[1], match_key.af);
5958 				match_key.port[0] = s->key[idx]->port[1];
5959 				PF_ACPY(&match_key.addr[1],
5960 				    &s->key[idx]->addr[0], match_key.af);
5961 				match_key.port[1] = s->key[idx]->port[0];
5962 			}
5963 
5964 			/*
5965 			 * Don't send out individual
5966 			 * delete messages.
5967 			 */
5968 			s->state_flags |= PFSTATE_NOSYNC;
5969 			pf_unlink_state(s);
5970 			killed++;
5971 
5972 			if (kill->psk_kill_match)
5973 				killed += pf_kill_matching_state(&match_key,
5974 				    dir);
5975 
5976 			goto relock_DIOCCLRSTATES;
5977 		}
5978 		PF_HASHROW_UNLOCK(ih);
5979 	}
5980 
5981 	if (V_pfsync_clear_states_ptr != NULL)
5982 		V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname);
5983 
5984 	return (killed);
5985 }
5986 
5987 static void
5988 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed)
5989 {
5990 	struct pf_kstate	*s;
5991 
5992 	if (kill->psk_pfcmp.id) {
5993 		if (kill->psk_pfcmp.creatorid == 0)
5994 			kill->psk_pfcmp.creatorid = V_pf_status.hostid;
5995 		if ((s = pf_find_state_byid(kill->psk_pfcmp.id,
5996 		    kill->psk_pfcmp.creatorid))) {
5997 			pf_unlink_state(s);
5998 			*killed = 1;
5999 		}
6000 		return;
6001 	}
6002 
6003 	for (unsigned int i = 0; i <= pf_hashmask; i++)
6004 		*killed += pf_killstates_row(kill, &V_pf_idhash[i]);
6005 
6006 	return;
6007 }
6008 
6009 static int
6010 pf_killstates_nv(struct pfioc_nv *nv)
6011 {
6012 	struct pf_kstate_kill	 kill;
6013 	nvlist_t		*nvl = NULL;
6014 	void			*nvlpacked = NULL;
6015 	int			 error = 0;
6016 	unsigned int		 killed = 0;
6017 
6018 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
6019 
6020 	if (nv->len > pf_ioctl_maxcount)
6021 		ERROUT(ENOMEM);
6022 
6023 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6024 	if (nvlpacked == NULL)
6025 		ERROUT(ENOMEM);
6026 
6027 	error = copyin(nv->data, nvlpacked, nv->len);
6028 	if (error)
6029 		ERROUT(error);
6030 
6031 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6032 	if (nvl == NULL)
6033 		ERROUT(EBADMSG);
6034 
6035 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6036 	if (error)
6037 		ERROUT(error);
6038 
6039 	pf_killstates(&kill, &killed);
6040 
6041 	free(nvlpacked, M_NVLIST);
6042 	nvlpacked = NULL;
6043 	nvlist_destroy(nvl);
6044 	nvl = nvlist_create(0);
6045 	if (nvl == NULL)
6046 		ERROUT(ENOMEM);
6047 
6048 	nvlist_add_number(nvl, "killed", killed);
6049 
6050 	nvlpacked = nvlist_pack(nvl, &nv->len);
6051 	if (nvlpacked == NULL)
6052 		ERROUT(ENOMEM);
6053 
6054 	if (nv->size == 0)
6055 		ERROUT(0);
6056 	else if (nv->size < nv->len)
6057 		ERROUT(ENOSPC);
6058 
6059 	error = copyout(nvlpacked, nv->data, nv->len);
6060 
6061 on_error:
6062 	nvlist_destroy(nvl);
6063 	free(nvlpacked, M_NVLIST);
6064 	return (error);
6065 }
6066 
6067 static int
6068 pf_clearstates_nv(struct pfioc_nv *nv)
6069 {
6070 	struct pf_kstate_kill	 kill;
6071 	nvlist_t		*nvl = NULL;
6072 	void			*nvlpacked = NULL;
6073 	int			 error = 0;
6074 	unsigned int		 killed;
6075 
6076 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
6077 
6078 	if (nv->len > pf_ioctl_maxcount)
6079 		ERROUT(ENOMEM);
6080 
6081 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6082 	if (nvlpacked == NULL)
6083 		ERROUT(ENOMEM);
6084 
6085 	error = copyin(nv->data, nvlpacked, nv->len);
6086 	if (error)
6087 		ERROUT(error);
6088 
6089 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6090 	if (nvl == NULL)
6091 		ERROUT(EBADMSG);
6092 
6093 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6094 	if (error)
6095 		ERROUT(error);
6096 
6097 	killed = pf_clear_states(&kill);
6098 
6099 	free(nvlpacked, M_NVLIST);
6100 	nvlpacked = NULL;
6101 	nvlist_destroy(nvl);
6102 	nvl = nvlist_create(0);
6103 	if (nvl == NULL)
6104 		ERROUT(ENOMEM);
6105 
6106 	nvlist_add_number(nvl, "killed", killed);
6107 
6108 	nvlpacked = nvlist_pack(nvl, &nv->len);
6109 	if (nvlpacked == NULL)
6110 		ERROUT(ENOMEM);
6111 
6112 	if (nv->size == 0)
6113 		ERROUT(0);
6114 	else if (nv->size < nv->len)
6115 		ERROUT(ENOSPC);
6116 
6117 	error = copyout(nvlpacked, nv->data, nv->len);
6118 
6119 #undef ERROUT
6120 on_error:
6121 	nvlist_destroy(nvl);
6122 	free(nvlpacked, M_NVLIST);
6123 	return (error);
6124 }
6125 
6126 static int
6127 pf_getstate(struct pfioc_nv *nv)
6128 {
6129 	nvlist_t		*nvl = NULL, *nvls;
6130 	void			*nvlpacked = NULL;
6131 	struct pf_kstate	*s = NULL;
6132 	int			 error = 0;
6133 	uint64_t		 id, creatorid;
6134 
6135 #define ERROUT(x)	ERROUT_FUNCTION(errout, x)
6136 
6137 	if (nv->len > pf_ioctl_maxcount)
6138 		ERROUT(ENOMEM);
6139 
6140 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6141 	if (nvlpacked == NULL)
6142 		ERROUT(ENOMEM);
6143 
6144 	error = copyin(nv->data, nvlpacked, nv->len);
6145 	if (error)
6146 		ERROUT(error);
6147 
6148 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6149 	if (nvl == NULL)
6150 		ERROUT(EBADMSG);
6151 
6152 	PFNV_CHK(pf_nvuint64(nvl, "id", &id));
6153 	PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid));
6154 
6155 	s = pf_find_state_byid(id, creatorid);
6156 	if (s == NULL)
6157 		ERROUT(ENOENT);
6158 
6159 	free(nvlpacked, M_NVLIST);
6160 	nvlpacked = NULL;
6161 	nvlist_destroy(nvl);
6162 	nvl = nvlist_create(0);
6163 	if (nvl == NULL)
6164 		ERROUT(ENOMEM);
6165 
6166 	nvls = pf_state_to_nvstate(s);
6167 	if (nvls == NULL)
6168 		ERROUT(ENOMEM);
6169 
6170 	nvlist_add_nvlist(nvl, "state", nvls);
6171 	nvlist_destroy(nvls);
6172 
6173 	nvlpacked = nvlist_pack(nvl, &nv->len);
6174 	if (nvlpacked == NULL)
6175 		ERROUT(ENOMEM);
6176 
6177 	if (nv->size == 0)
6178 		ERROUT(0);
6179 	else if (nv->size < nv->len)
6180 		ERROUT(ENOSPC);
6181 
6182 	error = copyout(nvlpacked, nv->data, nv->len);
6183 
6184 #undef ERROUT
6185 errout:
6186 	if (s != NULL)
6187 		PF_STATE_UNLOCK(s);
6188 	free(nvlpacked, M_NVLIST);
6189 	nvlist_destroy(nvl);
6190 	return (error);
6191 }
6192 
6193 /*
6194  * XXX - Check for version mismatch!!!
6195  */
6196 
6197 /*
6198  * Duplicate pfctl -Fa operation to get rid of as much as we can.
6199  */
6200 static int
6201 shutdown_pf(void)
6202 {
6203 	int error = 0;
6204 	u_int32_t t[5];
6205 	char nn = '\0';
6206 
6207 	do {
6208 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
6209 		    != 0) {
6210 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
6211 			break;
6212 		}
6213 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
6214 		    != 0) {
6215 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
6216 			break;		/* XXX: rollback? */
6217 		}
6218 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
6219 		    != 0) {
6220 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
6221 			break;		/* XXX: rollback? */
6222 		}
6223 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
6224 		    != 0) {
6225 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
6226 			break;		/* XXX: rollback? */
6227 		}
6228 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
6229 		    != 0) {
6230 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
6231 			break;		/* XXX: rollback? */
6232 		}
6233 
6234 		/* XXX: these should always succeed here */
6235 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
6236 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
6237 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
6238 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
6239 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
6240 
6241 		if ((error = pf_clear_tables()) != 0)
6242 			break;
6243 
6244 		if ((error = pf_begin_eth(&t[0], &nn)) != 0) {
6245 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth\n"));
6246 			break;
6247 		}
6248 		pf_commit_eth(t[0], &nn);
6249 
6250 #ifdef ALTQ
6251 		if ((error = pf_begin_altq(&t[0])) != 0) {
6252 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
6253 			break;
6254 		}
6255 		pf_commit_altq(t[0]);
6256 #endif
6257 
6258 		pf_clear_all_states();
6259 
6260 		pf_clear_srcnodes(NULL);
6261 
6262 		/* status does not use malloced mem so no need to cleanup */
6263 		/* fingerprints and interfaces have their own cleanup code */
6264 	} while(0);
6265 
6266 	return (error);
6267 }
6268 
6269 static pfil_return_t
6270 pf_check_return(int chk, struct mbuf **m)
6271 {
6272 
6273 	switch (chk) {
6274 	case PF_PASS:
6275 		if (*m == NULL)
6276 			return (PFIL_CONSUMED);
6277 		else
6278 			return (PFIL_PASS);
6279 		break;
6280 	default:
6281 		if (*m != NULL) {
6282 			m_freem(*m);
6283 			*m = NULL;
6284 		}
6285 		return (PFIL_DROPPED);
6286 	}
6287 }
6288 
6289 static pfil_return_t
6290 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6291     void *ruleset __unused, struct inpcb *inp)
6292 {
6293 	int chk;
6294 
6295 	chk = pf_test_eth(PF_IN, flags, ifp, m, inp);
6296 
6297 	return (pf_check_return(chk, m));
6298 }
6299 
6300 static pfil_return_t
6301 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6302     void *ruleset __unused, struct inpcb *inp)
6303 {
6304 	int chk;
6305 
6306 	chk = pf_test_eth(PF_OUT, flags, ifp, m, inp);
6307 
6308 	return (pf_check_return(chk, m));
6309 }
6310 
6311 #ifdef INET
6312 static pfil_return_t
6313 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6314     void *ruleset __unused, struct inpcb *inp)
6315 {
6316 	int chk;
6317 
6318 	chk = pf_test(PF_IN, flags, ifp, m, inp, NULL);
6319 
6320 	return (pf_check_return(chk, m));
6321 }
6322 
6323 static pfil_return_t
6324 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6325     void *ruleset __unused,  struct inpcb *inp)
6326 {
6327 	int chk;
6328 
6329 	chk = pf_test(PF_OUT, flags, ifp, m, inp, NULL);
6330 
6331 	return (pf_check_return(chk, m));
6332 }
6333 #endif
6334 
6335 #ifdef INET6
6336 static pfil_return_t
6337 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
6338     void *ruleset __unused,  struct inpcb *inp)
6339 {
6340 	int chk;
6341 
6342 	/*
6343 	 * In case of loopback traffic IPv6 uses the real interface in
6344 	 * order to support scoped addresses. In order to support stateful
6345 	 * filtering we have change this to lo0 as it is the case in IPv4.
6346 	 */
6347 	CURVNET_SET(ifp->if_vnet);
6348 	chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp,
6349 	    m, inp, NULL);
6350 	CURVNET_RESTORE();
6351 
6352 	return (pf_check_return(chk, m));
6353 }
6354 
6355 static pfil_return_t
6356 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
6357     void *ruleset __unused,  struct inpcb *inp)
6358 {
6359 	int chk;
6360 
6361 	CURVNET_SET(ifp->if_vnet);
6362 	chk = pf_test6(PF_OUT, flags, ifp, m, inp, NULL);
6363 	CURVNET_RESTORE();
6364 
6365 	return (pf_check_return(chk, m));
6366 }
6367 #endif /* INET6 */
6368 
6369 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook);
6370 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook);
6371 #define	V_pf_eth_in_hook	VNET(pf_eth_in_hook)
6372 #define	V_pf_eth_out_hook	VNET(pf_eth_out_hook)
6373 
6374 #ifdef INET
6375 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
6376 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
6377 #define	V_pf_ip4_in_hook	VNET(pf_ip4_in_hook)
6378 #define	V_pf_ip4_out_hook	VNET(pf_ip4_out_hook)
6379 #endif
6380 #ifdef INET6
6381 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
6382 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
6383 #define	V_pf_ip6_in_hook	VNET(pf_ip6_in_hook)
6384 #define	V_pf_ip6_out_hook	VNET(pf_ip6_out_hook)
6385 #endif
6386 
6387 static void
6388 hook_pf_eth(void)
6389 {
6390 	struct pfil_hook_args pha = {
6391 		.pa_version = PFIL_VERSION,
6392 		.pa_modname = "pf",
6393 		.pa_type = PFIL_TYPE_ETHERNET,
6394 	};
6395 	struct pfil_link_args pla = {
6396 		.pa_version = PFIL_VERSION,
6397 	};
6398 	int ret __diagused;
6399 
6400 	if (atomic_load_bool(&V_pf_pfil_eth_hooked))
6401 		return;
6402 
6403 	pha.pa_mbuf_chk = pf_eth_check_in;
6404 	pha.pa_flags = PFIL_IN;
6405 	pha.pa_rulname = "eth-in";
6406 	V_pf_eth_in_hook = pfil_add_hook(&pha);
6407 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6408 	pla.pa_head = V_link_pfil_head;
6409 	pla.pa_hook = V_pf_eth_in_hook;
6410 	ret = pfil_link(&pla);
6411 	MPASS(ret == 0);
6412 	pha.pa_mbuf_chk = pf_eth_check_out;
6413 	pha.pa_flags = PFIL_OUT;
6414 	pha.pa_rulname = "eth-out";
6415 	V_pf_eth_out_hook = pfil_add_hook(&pha);
6416 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6417 	pla.pa_head = V_link_pfil_head;
6418 	pla.pa_hook = V_pf_eth_out_hook;
6419 	ret = pfil_link(&pla);
6420 	MPASS(ret == 0);
6421 
6422 	atomic_store_bool(&V_pf_pfil_eth_hooked, true);
6423 }
6424 
6425 static void
6426 hook_pf(void)
6427 {
6428 	struct pfil_hook_args pha = {
6429 		.pa_version = PFIL_VERSION,
6430 		.pa_modname = "pf",
6431 	};
6432 	struct pfil_link_args pla = {
6433 		.pa_version = PFIL_VERSION,
6434 	};
6435 	int ret __diagused;
6436 
6437 	if (atomic_load_bool(&V_pf_pfil_hooked))
6438 		return;
6439 
6440 #ifdef INET
6441 	pha.pa_type = PFIL_TYPE_IP4;
6442 	pha.pa_mbuf_chk = pf_check_in;
6443 	pha.pa_flags = PFIL_IN;
6444 	pha.pa_rulname = "default-in";
6445 	V_pf_ip4_in_hook = pfil_add_hook(&pha);
6446 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6447 	pla.pa_head = V_inet_pfil_head;
6448 	pla.pa_hook = V_pf_ip4_in_hook;
6449 	ret = pfil_link(&pla);
6450 	MPASS(ret == 0);
6451 	pha.pa_mbuf_chk = pf_check_out;
6452 	pha.pa_flags = PFIL_OUT;
6453 	pha.pa_rulname = "default-out";
6454 	V_pf_ip4_out_hook = pfil_add_hook(&pha);
6455 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6456 	pla.pa_head = V_inet_pfil_head;
6457 	pla.pa_hook = V_pf_ip4_out_hook;
6458 	ret = pfil_link(&pla);
6459 	MPASS(ret == 0);
6460 	if (V_pf_filter_local) {
6461 		pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6462 		pla.pa_head = V_inet_local_pfil_head;
6463 		pla.pa_hook = V_pf_ip4_out_hook;
6464 		ret = pfil_link(&pla);
6465 		MPASS(ret == 0);
6466 	}
6467 #endif
6468 #ifdef INET6
6469 	pha.pa_type = PFIL_TYPE_IP6;
6470 	pha.pa_mbuf_chk = pf_check6_in;
6471 	pha.pa_flags = PFIL_IN;
6472 	pha.pa_rulname = "default-in6";
6473 	V_pf_ip6_in_hook = pfil_add_hook(&pha);
6474 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6475 	pla.pa_head = V_inet6_pfil_head;
6476 	pla.pa_hook = V_pf_ip6_in_hook;
6477 	ret = pfil_link(&pla);
6478 	MPASS(ret == 0);
6479 	pha.pa_mbuf_chk = pf_check6_out;
6480 	pha.pa_rulname = "default-out6";
6481 	pha.pa_flags = PFIL_OUT;
6482 	V_pf_ip6_out_hook = pfil_add_hook(&pha);
6483 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6484 	pla.pa_head = V_inet6_pfil_head;
6485 	pla.pa_hook = V_pf_ip6_out_hook;
6486 	ret = pfil_link(&pla);
6487 	MPASS(ret == 0);
6488 	if (V_pf_filter_local) {
6489 		pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6490 		pla.pa_head = V_inet6_local_pfil_head;
6491 		pla.pa_hook = V_pf_ip6_out_hook;
6492 		ret = pfil_link(&pla);
6493 		MPASS(ret == 0);
6494 	}
6495 #endif
6496 
6497 	atomic_store_bool(&V_pf_pfil_hooked, true);
6498 }
6499 
6500 static void
6501 dehook_pf_eth(void)
6502 {
6503 
6504 	if (!atomic_load_bool(&V_pf_pfil_eth_hooked))
6505 		return;
6506 
6507 	pfil_remove_hook(V_pf_eth_in_hook);
6508 	pfil_remove_hook(V_pf_eth_out_hook);
6509 
6510 	atomic_store_bool(&V_pf_pfil_eth_hooked, false);
6511 }
6512 
6513 static void
6514 dehook_pf(void)
6515 {
6516 
6517 	if (!atomic_load_bool(&V_pf_pfil_hooked))
6518 		return;
6519 
6520 #ifdef INET
6521 	pfil_remove_hook(V_pf_ip4_in_hook);
6522 	pfil_remove_hook(V_pf_ip4_out_hook);
6523 #endif
6524 #ifdef INET6
6525 	pfil_remove_hook(V_pf_ip6_in_hook);
6526 	pfil_remove_hook(V_pf_ip6_out_hook);
6527 #endif
6528 
6529 	atomic_store_bool(&V_pf_pfil_hooked, false);
6530 }
6531 
6532 static void
6533 pf_load_vnet(void)
6534 {
6535 	V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
6536 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
6537 
6538 	rm_init_flags(&V_pf_rules_lock, "pf rulesets", RM_RECURSE);
6539 	sx_init(&V_pf_ioctl_lock, "pf ioctl");
6540 
6541 	pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
6542 	    PF_RULE_TAG_HASH_SIZE_DEFAULT);
6543 #ifdef ALTQ
6544 	pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
6545 	    PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
6546 #endif
6547 
6548 	V_pf_keth = &V_pf_main_keth_anchor.ruleset;
6549 
6550 	pfattach_vnet();
6551 	V_pf_vnet_active = 1;
6552 }
6553 
6554 static int
6555 pf_load(void)
6556 {
6557 	int error;
6558 
6559 	sx_init(&pf_end_lock, "pf end thread");
6560 
6561 	pf_mtag_initialize();
6562 
6563 	pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
6564 	if (pf_dev == NULL)
6565 		return (ENOMEM);
6566 
6567 	pf_end_threads = 0;
6568 	error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
6569 	if (error != 0)
6570 		return (error);
6571 
6572 	pfi_initialize();
6573 
6574 	return (0);
6575 }
6576 
6577 static void
6578 pf_unload_vnet(void)
6579 {
6580 	int ret __diagused;
6581 
6582 	V_pf_vnet_active = 0;
6583 	V_pf_status.running = 0;
6584 	dehook_pf();
6585 	dehook_pf_eth();
6586 
6587 	PF_RULES_WLOCK();
6588 	pf_syncookies_cleanup();
6589 	shutdown_pf();
6590 	PF_RULES_WUNLOCK();
6591 
6592 	/* Make sure we've cleaned up ethernet rules before we continue. */
6593 	NET_EPOCH_DRAIN_CALLBACKS();
6594 
6595 	ret = swi_remove(V_pf_swi_cookie);
6596 	MPASS(ret == 0);
6597 	ret = intr_event_destroy(V_pf_swi_ie);
6598 	MPASS(ret == 0);
6599 
6600 	pf_unload_vnet_purge();
6601 
6602 	pf_normalize_cleanup();
6603 	PF_RULES_WLOCK();
6604 	pfi_cleanup_vnet();
6605 	PF_RULES_WUNLOCK();
6606 	pfr_cleanup();
6607 	pf_osfp_flush();
6608 	pf_cleanup();
6609 	if (IS_DEFAULT_VNET(curvnet))
6610 		pf_mtag_cleanup();
6611 
6612 	pf_cleanup_tagset(&V_pf_tags);
6613 #ifdef ALTQ
6614 	pf_cleanup_tagset(&V_pf_qids);
6615 #endif
6616 	uma_zdestroy(V_pf_tag_z);
6617 
6618 #ifdef PF_WANT_32_TO_64_COUNTER
6619 	PF_RULES_WLOCK();
6620 	LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist);
6621 
6622 	MPASS(LIST_EMPTY(&V_pf_allkiflist));
6623 	MPASS(V_pf_allkifcount == 0);
6624 
6625 	LIST_REMOVE(&V_pf_default_rule, allrulelist);
6626 	V_pf_allrulecount--;
6627 	LIST_REMOVE(V_pf_rulemarker, allrulelist);
6628 
6629 	/*
6630 	 * There are known pf rule leaks when running the test suite.
6631 	 */
6632 #ifdef notyet
6633 	MPASS(LIST_EMPTY(&V_pf_allrulelist));
6634 	MPASS(V_pf_allrulecount == 0);
6635 #endif
6636 
6637 	PF_RULES_WUNLOCK();
6638 
6639 	free(V_pf_kifmarker, PFI_MTYPE);
6640 	free(V_pf_rulemarker, M_PFRULE);
6641 #endif
6642 
6643 	/* Free counters last as we updated them during shutdown. */
6644 	pf_counter_u64_deinit(&V_pf_default_rule.evaluations);
6645 	for (int i = 0; i < 2; i++) {
6646 		pf_counter_u64_deinit(&V_pf_default_rule.packets[i]);
6647 		pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]);
6648 	}
6649 	counter_u64_free(V_pf_default_rule.states_cur);
6650 	counter_u64_free(V_pf_default_rule.states_tot);
6651 	counter_u64_free(V_pf_default_rule.src_nodes);
6652 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, V_pf_default_rule.timestamp);
6653 
6654 	for (int i = 0; i < PFRES_MAX; i++)
6655 		counter_u64_free(V_pf_status.counters[i]);
6656 	for (int i = 0; i < KLCNT_MAX; i++)
6657 		counter_u64_free(V_pf_status.lcounters[i]);
6658 	for (int i = 0; i < FCNT_MAX; i++)
6659 		pf_counter_u64_deinit(&V_pf_status.fcounters[i]);
6660 	for (int i = 0; i < SCNT_MAX; i++)
6661 		counter_u64_free(V_pf_status.scounters[i]);
6662 
6663 	rm_destroy(&V_pf_rules_lock);
6664 	sx_destroy(&V_pf_ioctl_lock);
6665 }
6666 
6667 static void
6668 pf_unload(void)
6669 {
6670 
6671 	sx_xlock(&pf_end_lock);
6672 	pf_end_threads = 1;
6673 	while (pf_end_threads < 2) {
6674 		wakeup_one(pf_purge_thread);
6675 		sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
6676 	}
6677 	sx_xunlock(&pf_end_lock);
6678 
6679 	pf_nl_unregister();
6680 
6681 	if (pf_dev != NULL)
6682 		destroy_dev(pf_dev);
6683 
6684 	pfi_cleanup();
6685 
6686 	sx_destroy(&pf_end_lock);
6687 }
6688 
6689 static void
6690 vnet_pf_init(void *unused __unused)
6691 {
6692 
6693 	pf_load_vnet();
6694 }
6695 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6696     vnet_pf_init, NULL);
6697 
6698 static void
6699 vnet_pf_uninit(const void *unused __unused)
6700 {
6701 
6702 	pf_unload_vnet();
6703 }
6704 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
6705 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6706     vnet_pf_uninit, NULL);
6707 
6708 static int
6709 pf_modevent(module_t mod, int type, void *data)
6710 {
6711 	int error = 0;
6712 
6713 	switch(type) {
6714 	case MOD_LOAD:
6715 		error = pf_load();
6716 		pf_nl_register();
6717 		break;
6718 	case MOD_UNLOAD:
6719 		/* Handled in SYSUNINIT(pf_unload) to ensure it's done after
6720 		 * the vnet_pf_uninit()s */
6721 		break;
6722 	default:
6723 		error = EINVAL;
6724 		break;
6725 	}
6726 
6727 	return (error);
6728 }
6729 
6730 static moduledata_t pf_mod = {
6731 	"pf",
6732 	pf_modevent,
6733 	0
6734 };
6735 
6736 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
6737 MODULE_DEPEND(pf, netlink, 1, 1, 1);
6738 MODULE_VERSION(pf, PF_MODVER);
6739