xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision 43cd6bbba053f02999d846ef297655ebec2d218b)
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 static int
2341 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
2342 {
2343 	int			 error = 0;
2344 	PF_RULES_RLOCK_TRACKER;
2345 
2346 #define	ERROUT_IOCTL(target, x)					\
2347     do {								\
2348 	    error = (x);						\
2349 	    SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__);	\
2350 	    goto target;						\
2351     } while (0)
2352 
2353 
2354 	/* XXX keep in sync with switch() below */
2355 	if (securelevel_gt(td->td_ucred, 2))
2356 		switch (cmd) {
2357 		case DIOCGETRULES:
2358 		case DIOCGETRULENV:
2359 		case DIOCGETADDRS:
2360 		case DIOCGETADDR:
2361 		case DIOCGETSTATE:
2362 		case DIOCGETSTATENV:
2363 		case DIOCSETSTATUSIF:
2364 		case DIOCGETSTATUSNV:
2365 		case DIOCCLRSTATUS:
2366 		case DIOCNATLOOK:
2367 		case DIOCSETDEBUG:
2368 #ifdef COMPAT_FREEBSD14
2369 		case DIOCGETSTATES:
2370 		case DIOCGETSTATESV2:
2371 #endif
2372 		case DIOCGETTIMEOUT:
2373 		case DIOCCLRRULECTRS:
2374 		case DIOCGETLIMIT:
2375 		case DIOCGETALTQSV0:
2376 		case DIOCGETALTQSV1:
2377 		case DIOCGETALTQV0:
2378 		case DIOCGETALTQV1:
2379 		case DIOCGETQSTATSV0:
2380 		case DIOCGETQSTATSV1:
2381 		case DIOCGETRULESETS:
2382 		case DIOCGETRULESET:
2383 		case DIOCRGETTABLES:
2384 		case DIOCRGETTSTATS:
2385 		case DIOCRCLRTSTATS:
2386 		case DIOCRCLRADDRS:
2387 		case DIOCRADDADDRS:
2388 		case DIOCRDELADDRS:
2389 		case DIOCRSETADDRS:
2390 		case DIOCRGETADDRS:
2391 		case DIOCRGETASTATS:
2392 		case DIOCRCLRASTATS:
2393 		case DIOCRTSTADDRS:
2394 		case DIOCOSFPGET:
2395 		case DIOCGETSRCNODES:
2396 		case DIOCCLRSRCNODES:
2397 		case DIOCGETSYNCOOKIES:
2398 		case DIOCIGETIFACES:
2399 		case DIOCGIFSPEEDV0:
2400 		case DIOCGIFSPEEDV1:
2401 		case DIOCSETIFFLAG:
2402 		case DIOCCLRIFFLAG:
2403 		case DIOCGETETHRULES:
2404 		case DIOCGETETHRULE:
2405 		case DIOCGETETHRULESETS:
2406 		case DIOCGETETHRULESET:
2407 			break;
2408 		case DIOCRCLRTABLES:
2409 		case DIOCRADDTABLES:
2410 		case DIOCRDELTABLES:
2411 		case DIOCRSETTFLAGS:
2412 			if (((struct pfioc_table *)addr)->pfrio_flags &
2413 			    PFR_FLAG_DUMMY)
2414 				break; /* dummy operation ok */
2415 			return (EPERM);
2416 		default:
2417 			return (EPERM);
2418 		}
2419 
2420 	if (!(flags & FWRITE))
2421 		switch (cmd) {
2422 		case DIOCGETRULES:
2423 		case DIOCGETADDRS:
2424 		case DIOCGETADDR:
2425 		case DIOCGETSTATE:
2426 		case DIOCGETSTATENV:
2427 		case DIOCGETSTATUSNV:
2428 #ifdef COMPAT_FREEBSD14
2429 		case DIOCGETSTATES:
2430 		case DIOCGETSTATESV2:
2431 #endif
2432 		case DIOCGETTIMEOUT:
2433 		case DIOCGETLIMIT:
2434 		case DIOCGETALTQSV0:
2435 		case DIOCGETALTQSV1:
2436 		case DIOCGETALTQV0:
2437 		case DIOCGETALTQV1:
2438 		case DIOCGETQSTATSV0:
2439 		case DIOCGETQSTATSV1:
2440 		case DIOCGETRULESETS:
2441 		case DIOCGETRULESET:
2442 		case DIOCNATLOOK:
2443 		case DIOCRGETTABLES:
2444 		case DIOCRGETTSTATS:
2445 		case DIOCRGETADDRS:
2446 		case DIOCRGETASTATS:
2447 		case DIOCRTSTADDRS:
2448 		case DIOCOSFPGET:
2449 		case DIOCGETSRCNODES:
2450 		case DIOCGETSYNCOOKIES:
2451 		case DIOCIGETIFACES:
2452 		case DIOCGIFSPEEDV1:
2453 		case DIOCGIFSPEEDV0:
2454 		case DIOCGETRULENV:
2455 		case DIOCGETETHRULES:
2456 		case DIOCGETETHRULE:
2457 		case DIOCGETETHRULESETS:
2458 		case DIOCGETETHRULESET:
2459 			break;
2460 		case DIOCRCLRTABLES:
2461 		case DIOCRADDTABLES:
2462 		case DIOCRDELTABLES:
2463 		case DIOCRCLRTSTATS:
2464 		case DIOCRCLRADDRS:
2465 		case DIOCRADDADDRS:
2466 		case DIOCRDELADDRS:
2467 		case DIOCRSETADDRS:
2468 		case DIOCRSETTFLAGS:
2469 			if (((struct pfioc_table *)addr)->pfrio_flags &
2470 			    PFR_FLAG_DUMMY) {
2471 				flags |= FWRITE; /* need write lock for dummy */
2472 				break; /* dummy operation ok */
2473 			}
2474 			return (EACCES);
2475 		default:
2476 			return (EACCES);
2477 		}
2478 
2479 	CURVNET_SET(TD_TO_VNET(td));
2480 
2481 	switch (cmd) {
2482 	case DIOCSTART:
2483 		sx_xlock(&V_pf_ioctl_lock);
2484 		if (V_pf_status.running)
2485 			error = EEXIST;
2486 		else {
2487 			hook_pf();
2488 			if (! TAILQ_EMPTY(V_pf_keth->active.rules))
2489 				hook_pf_eth();
2490 			V_pf_status.running = 1;
2491 			V_pf_status.since = time_second;
2492 			new_unrhdr64(&V_pf_stateid, time_second);
2493 
2494 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
2495 		}
2496 		break;
2497 
2498 	case DIOCSTOP:
2499 		sx_xlock(&V_pf_ioctl_lock);
2500 		if (!V_pf_status.running)
2501 			error = ENOENT;
2502 		else {
2503 			V_pf_status.running = 0;
2504 			dehook_pf();
2505 			dehook_pf_eth();
2506 			V_pf_status.since = time_second;
2507 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
2508 		}
2509 		break;
2510 
2511 	case DIOCGETETHRULES: {
2512 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2513 		nvlist_t		*nvl;
2514 		void			*packed;
2515 		struct pf_keth_rule	*tail;
2516 		struct pf_keth_ruleset	*rs;
2517 		u_int32_t		 ticket, nr;
2518 		const char		*anchor = "";
2519 
2520 		nvl = NULL;
2521 		packed = NULL;
2522 
2523 #define	ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULES_error, x)
2524 
2525 		if (nv->len > pf_ioctl_maxcount)
2526 			ERROUT(ENOMEM);
2527 
2528 		/* Copy the request in */
2529 		packed = malloc(nv->len, M_NVLIST, M_WAITOK);
2530 		if (packed == NULL)
2531 			ERROUT(ENOMEM);
2532 
2533 		error = copyin(nv->data, packed, nv->len);
2534 		if (error)
2535 			ERROUT(error);
2536 
2537 		nvl = nvlist_unpack(packed, nv->len, 0);
2538 		if (nvl == NULL)
2539 			ERROUT(EBADMSG);
2540 
2541 		if (! nvlist_exists_string(nvl, "anchor"))
2542 			ERROUT(EBADMSG);
2543 
2544 		anchor = nvlist_get_string(nvl, "anchor");
2545 
2546 		rs = pf_find_keth_ruleset(anchor);
2547 
2548 		nvlist_destroy(nvl);
2549 		nvl = NULL;
2550 		free(packed, M_NVLIST);
2551 		packed = NULL;
2552 
2553 		if (rs == NULL)
2554 			ERROUT(ENOENT);
2555 
2556 		/* Reply */
2557 		nvl = nvlist_create(0);
2558 		if (nvl == NULL)
2559 			ERROUT(ENOMEM);
2560 
2561 		PF_RULES_RLOCK();
2562 
2563 		ticket = rs->active.ticket;
2564 		tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq);
2565 		if (tail)
2566 			nr = tail->nr + 1;
2567 		else
2568 			nr = 0;
2569 
2570 		PF_RULES_RUNLOCK();
2571 
2572 		nvlist_add_number(nvl, "ticket", ticket);
2573 		nvlist_add_number(nvl, "nr", nr);
2574 
2575 		packed = nvlist_pack(nvl, &nv->len);
2576 		if (packed == NULL)
2577 			ERROUT(ENOMEM);
2578 
2579 		if (nv->size == 0)
2580 			ERROUT(0);
2581 		else if (nv->size < nv->len)
2582 			ERROUT(ENOSPC);
2583 
2584 		error = copyout(packed, nv->data, nv->len);
2585 
2586 #undef ERROUT
2587 DIOCGETETHRULES_error:
2588 		free(packed, M_NVLIST);
2589 		nvlist_destroy(nvl);
2590 		break;
2591 	}
2592 
2593 	case DIOCGETETHRULE: {
2594 		struct epoch_tracker	 et;
2595 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2596 		nvlist_t		*nvl = NULL;
2597 		void			*nvlpacked = NULL;
2598 		struct pf_keth_rule	*rule = NULL;
2599 		struct pf_keth_ruleset	*rs;
2600 		u_int32_t		 ticket, nr;
2601 		bool			 clear = false;
2602 		const char		*anchor;
2603 
2604 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULE_error, x)
2605 
2606 		if (nv->len > pf_ioctl_maxcount)
2607 			ERROUT(ENOMEM);
2608 
2609 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2610 		if (nvlpacked == NULL)
2611 			ERROUT(ENOMEM);
2612 
2613 		error = copyin(nv->data, nvlpacked, nv->len);
2614 		if (error)
2615 			ERROUT(error);
2616 
2617 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2618 		if (nvl == NULL)
2619 			ERROUT(EBADMSG);
2620 		if (! nvlist_exists_number(nvl, "ticket"))
2621 			ERROUT(EBADMSG);
2622 		ticket = nvlist_get_number(nvl, "ticket");
2623 		if (! nvlist_exists_string(nvl, "anchor"))
2624 			ERROUT(EBADMSG);
2625 		anchor = nvlist_get_string(nvl, "anchor");
2626 
2627 		if (nvlist_exists_bool(nvl, "clear"))
2628 			clear = nvlist_get_bool(nvl, "clear");
2629 
2630 		if (clear && !(flags & FWRITE))
2631 			ERROUT(EACCES);
2632 
2633 		if (! nvlist_exists_number(nvl, "nr"))
2634 			ERROUT(EBADMSG);
2635 		nr = nvlist_get_number(nvl, "nr");
2636 
2637 		PF_RULES_RLOCK();
2638 		rs = pf_find_keth_ruleset(anchor);
2639 		if (rs == NULL) {
2640 			PF_RULES_RUNLOCK();
2641 			ERROUT(ENOENT);
2642 		}
2643 		if (ticket != rs->active.ticket) {
2644 			PF_RULES_RUNLOCK();
2645 			ERROUT(EBUSY);
2646 		}
2647 
2648 		nvlist_destroy(nvl);
2649 		nvl = NULL;
2650 		free(nvlpacked, M_NVLIST);
2651 		nvlpacked = NULL;
2652 
2653 		rule = TAILQ_FIRST(rs->active.rules);
2654 		while ((rule != NULL) && (rule->nr != nr))
2655 			rule = TAILQ_NEXT(rule, entries);
2656 		if (rule == NULL) {
2657 			PF_RULES_RUNLOCK();
2658 			ERROUT(ENOENT);
2659 		}
2660 		/* Make sure rule can't go away. */
2661 		NET_EPOCH_ENTER(et);
2662 		PF_RULES_RUNLOCK();
2663 		nvl = pf_keth_rule_to_nveth_rule(rule);
2664 		if (pf_keth_anchor_nvcopyout(rs, rule, nvl))
2665 			ERROUT(EBUSY);
2666 		NET_EPOCH_EXIT(et);
2667 		if (nvl == NULL)
2668 			ERROUT(ENOMEM);
2669 
2670 		nvlpacked = nvlist_pack(nvl, &nv->len);
2671 		if (nvlpacked == NULL)
2672 			ERROUT(ENOMEM);
2673 
2674 		if (nv->size == 0)
2675 			ERROUT(0);
2676 		else if (nv->size < nv->len)
2677 			ERROUT(ENOSPC);
2678 
2679 		error = copyout(nvlpacked, nv->data, nv->len);
2680 		if (error == 0 && clear) {
2681 			counter_u64_zero(rule->evaluations);
2682 			for (int i = 0; i < 2; i++) {
2683 				counter_u64_zero(rule->packets[i]);
2684 				counter_u64_zero(rule->bytes[i]);
2685 			}
2686 		}
2687 
2688 #undef ERROUT
2689 DIOCGETETHRULE_error:
2690 		free(nvlpacked, M_NVLIST);
2691 		nvlist_destroy(nvl);
2692 		break;
2693 	}
2694 
2695 	case DIOCADDETHRULE: {
2696 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2697 		nvlist_t		*nvl = NULL;
2698 		void			*nvlpacked = NULL;
2699 		struct pf_keth_rule	*rule = NULL, *tail = NULL;
2700 		struct pf_keth_ruleset	*ruleset = NULL;
2701 		struct pfi_kkif		*kif = NULL, *bridge_to_kif = NULL;
2702 		const char		*anchor = "", *anchor_call = "";
2703 
2704 #define ERROUT(x)	ERROUT_IOCTL(DIOCADDETHRULE_error, x)
2705 
2706 		if (nv->len > pf_ioctl_maxcount)
2707 			ERROUT(ENOMEM);
2708 
2709 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2710 		if (nvlpacked == NULL)
2711 			ERROUT(ENOMEM);
2712 
2713 		error = copyin(nv->data, nvlpacked, nv->len);
2714 		if (error)
2715 			ERROUT(error);
2716 
2717 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2718 		if (nvl == NULL)
2719 			ERROUT(EBADMSG);
2720 
2721 		if (! nvlist_exists_number(nvl, "ticket"))
2722 			ERROUT(EBADMSG);
2723 
2724 		if (nvlist_exists_string(nvl, "anchor"))
2725 			anchor = nvlist_get_string(nvl, "anchor");
2726 		if (nvlist_exists_string(nvl, "anchor_call"))
2727 			anchor_call = nvlist_get_string(nvl, "anchor_call");
2728 
2729 		ruleset = pf_find_keth_ruleset(anchor);
2730 		if (ruleset == NULL)
2731 			ERROUT(EINVAL);
2732 
2733 		if (nvlist_get_number(nvl, "ticket") !=
2734 		    ruleset->inactive.ticket) {
2735 			DPFPRINTF(PF_DEBUG_MISC,
2736 			    ("ticket: %d != %d\n",
2737 			    (u_int32_t)nvlist_get_number(nvl, "ticket"),
2738 			    ruleset->inactive.ticket));
2739 			ERROUT(EBUSY);
2740 		}
2741 
2742 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
2743 		if (rule == NULL)
2744 			ERROUT(ENOMEM);
2745 		rule->timestamp = NULL;
2746 
2747 		error = pf_nveth_rule_to_keth_rule(nvl, rule);
2748 		if (error != 0)
2749 			ERROUT(error);
2750 
2751 		if (rule->ifname[0])
2752 			kif = pf_kkif_create(M_WAITOK);
2753 		if (rule->bridge_to_name[0])
2754 			bridge_to_kif = pf_kkif_create(M_WAITOK);
2755 		rule->evaluations = counter_u64_alloc(M_WAITOK);
2756 		for (int i = 0; i < 2; i++) {
2757 			rule->packets[i] = counter_u64_alloc(M_WAITOK);
2758 			rule->bytes[i] = counter_u64_alloc(M_WAITOK);
2759 		}
2760 		rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
2761 		    M_WAITOK | M_ZERO);
2762 
2763 		PF_RULES_WLOCK();
2764 
2765 		if (rule->ifname[0]) {
2766 			rule->kif = pfi_kkif_attach(kif, rule->ifname);
2767 			pfi_kkif_ref(rule->kif);
2768 		} else
2769 			rule->kif = NULL;
2770 		if (rule->bridge_to_name[0]) {
2771 			rule->bridge_to = pfi_kkif_attach(bridge_to_kif,
2772 			    rule->bridge_to_name);
2773 			pfi_kkif_ref(rule->bridge_to);
2774 		} else
2775 			rule->bridge_to = NULL;
2776 
2777 #ifdef ALTQ
2778 		/* set queue IDs */
2779 		if (rule->qname[0] != 0) {
2780 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2781 				error = EBUSY;
2782 			else
2783 				rule->qid = rule->qid;
2784 		}
2785 #endif
2786 		if (rule->tagname[0])
2787 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2788 				error = EBUSY;
2789 		if (rule->match_tagname[0])
2790 			if ((rule->match_tag = pf_tagname2tag(
2791 			    rule->match_tagname)) == 0)
2792 				error = EBUSY;
2793 
2794 		if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE)
2795 			error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr);
2796 		if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE)
2797 			error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr);
2798 
2799 		if (error) {
2800 			pf_free_eth_rule(rule);
2801 			PF_RULES_WUNLOCK();
2802 			ERROUT(error);
2803 		}
2804 
2805 		if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) {
2806 			pf_free_eth_rule(rule);
2807 			PF_RULES_WUNLOCK();
2808 			ERROUT(EINVAL);
2809 		}
2810 
2811 		tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq);
2812 		if (tail)
2813 			rule->nr = tail->nr + 1;
2814 		else
2815 			rule->nr = 0;
2816 
2817 		TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries);
2818 
2819 		PF_RULES_WUNLOCK();
2820 
2821 #undef ERROUT
2822 DIOCADDETHRULE_error:
2823 		nvlist_destroy(nvl);
2824 		free(nvlpacked, M_NVLIST);
2825 		break;
2826 	}
2827 
2828 	case DIOCGETETHRULESETS: {
2829 		struct epoch_tracker	 et;
2830 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2831 		nvlist_t		*nvl = NULL;
2832 		void			*nvlpacked = NULL;
2833 		struct pf_keth_ruleset	*ruleset;
2834 		struct pf_keth_anchor	*anchor;
2835 		int			 nr = 0;
2836 
2837 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULESETS_error, x)
2838 
2839 		if (nv->len > pf_ioctl_maxcount)
2840 			ERROUT(ENOMEM);
2841 
2842 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2843 		if (nvlpacked == NULL)
2844 			ERROUT(ENOMEM);
2845 
2846 		error = copyin(nv->data, nvlpacked, nv->len);
2847 		if (error)
2848 			ERROUT(error);
2849 
2850 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2851 		if (nvl == NULL)
2852 			ERROUT(EBADMSG);
2853 		if (! nvlist_exists_string(nvl, "path"))
2854 			ERROUT(EBADMSG);
2855 
2856 		NET_EPOCH_ENTER(et);
2857 
2858 		if ((ruleset = pf_find_keth_ruleset(
2859 		    nvlist_get_string(nvl, "path"))) == NULL) {
2860 			NET_EPOCH_EXIT(et);
2861 			ERROUT(ENOENT);
2862 		}
2863 
2864 		if (ruleset->anchor == NULL) {
2865 			RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors)
2866 				if (anchor->parent == NULL)
2867 					nr++;
2868 		} else {
2869 			RB_FOREACH(anchor, pf_keth_anchor_node,
2870 			    &ruleset->anchor->children)
2871 				nr++;
2872 		}
2873 
2874 		NET_EPOCH_EXIT(et);
2875 
2876 		nvlist_destroy(nvl);
2877 		nvl = NULL;
2878 		free(nvlpacked, M_NVLIST);
2879 		nvlpacked = NULL;
2880 
2881 		nvl = nvlist_create(0);
2882 		if (nvl == NULL)
2883 			ERROUT(ENOMEM);
2884 
2885 		nvlist_add_number(nvl, "nr", nr);
2886 
2887 		nvlpacked = nvlist_pack(nvl, &nv->len);
2888 		if (nvlpacked == NULL)
2889 			ERROUT(ENOMEM);
2890 
2891 		if (nv->size == 0)
2892 			ERROUT(0);
2893 		else if (nv->size < nv->len)
2894 			ERROUT(ENOSPC);
2895 
2896 		error = copyout(nvlpacked, nv->data, nv->len);
2897 
2898 #undef ERROUT
2899 DIOCGETETHRULESETS_error:
2900 		free(nvlpacked, M_NVLIST);
2901 		nvlist_destroy(nvl);
2902 		break;
2903 	}
2904 
2905 	case DIOCGETETHRULESET: {
2906 		struct epoch_tracker	 et;
2907 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2908 		nvlist_t		*nvl = NULL;
2909 		void			*nvlpacked = NULL;
2910 		struct pf_keth_ruleset	*ruleset;
2911 		struct pf_keth_anchor	*anchor;
2912 		int			 nr = 0, req_nr = 0;
2913 		bool			 found = false;
2914 
2915 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULESET_error, x)
2916 
2917 		if (nv->len > pf_ioctl_maxcount)
2918 			ERROUT(ENOMEM);
2919 
2920 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2921 		if (nvlpacked == NULL)
2922 			ERROUT(ENOMEM);
2923 
2924 		error = copyin(nv->data, nvlpacked, nv->len);
2925 		if (error)
2926 			ERROUT(error);
2927 
2928 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2929 		if (nvl == NULL)
2930 			ERROUT(EBADMSG);
2931 		if (! nvlist_exists_string(nvl, "path"))
2932 			ERROUT(EBADMSG);
2933 		if (! nvlist_exists_number(nvl, "nr"))
2934 			ERROUT(EBADMSG);
2935 
2936 		req_nr = nvlist_get_number(nvl, "nr");
2937 
2938 		NET_EPOCH_ENTER(et);
2939 
2940 		if ((ruleset = pf_find_keth_ruleset(
2941 		    nvlist_get_string(nvl, "path"))) == NULL) {
2942 			NET_EPOCH_EXIT(et);
2943 			ERROUT(ENOENT);
2944 		}
2945 
2946 		nvlist_destroy(nvl);
2947 		nvl = NULL;
2948 		free(nvlpacked, M_NVLIST);
2949 		nvlpacked = NULL;
2950 
2951 		nvl = nvlist_create(0);
2952 		if (nvl == NULL) {
2953 			NET_EPOCH_EXIT(et);
2954 			ERROUT(ENOMEM);
2955 		}
2956 
2957 		if (ruleset->anchor == NULL) {
2958 			RB_FOREACH(anchor, pf_keth_anchor_global,
2959 			    &V_pf_keth_anchors) {
2960 				if (anchor->parent == NULL && nr++ == req_nr) {
2961 					found = true;
2962 					break;
2963 				}
2964 			}
2965 		} else {
2966 			RB_FOREACH(anchor, pf_keth_anchor_node,
2967 			     &ruleset->anchor->children) {
2968 				if (nr++ == req_nr) {
2969 					found = true;
2970 					break;
2971 				}
2972 			}
2973 		}
2974 
2975 		NET_EPOCH_EXIT(et);
2976 		if (found) {
2977 			nvlist_add_number(nvl, "nr", nr);
2978 			nvlist_add_string(nvl, "name", anchor->name);
2979 			if (ruleset->anchor)
2980 				nvlist_add_string(nvl, "path",
2981 				    ruleset->anchor->path);
2982 			else
2983 				nvlist_add_string(nvl, "path", "");
2984 		} else {
2985 			ERROUT(EBUSY);
2986 		}
2987 
2988 		nvlpacked = nvlist_pack(nvl, &nv->len);
2989 		if (nvlpacked == NULL)
2990 			ERROUT(ENOMEM);
2991 
2992 		if (nv->size == 0)
2993 			ERROUT(0);
2994 		else if (nv->size < nv->len)
2995 			ERROUT(ENOSPC);
2996 
2997 		error = copyout(nvlpacked, nv->data, nv->len);
2998 
2999 #undef ERROUT
3000 DIOCGETETHRULESET_error:
3001 		free(nvlpacked, M_NVLIST);
3002 		nvlist_destroy(nvl);
3003 		break;
3004 	}
3005 
3006 	case DIOCADDRULENV: {
3007 		struct pfioc_nv	*nv = (struct pfioc_nv *)addr;
3008 		nvlist_t	*nvl = NULL;
3009 		void		*nvlpacked = NULL;
3010 		struct pf_krule	*rule = NULL;
3011 		const char	*anchor = "", *anchor_call = "";
3012 		uint32_t	 ticket = 0, pool_ticket = 0;
3013 
3014 #define	ERROUT(x)	ERROUT_IOCTL(DIOCADDRULENV_error, x)
3015 
3016 		if (nv->len > pf_ioctl_maxcount)
3017 			ERROUT(ENOMEM);
3018 
3019 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3020 		error = copyin(nv->data, nvlpacked, nv->len);
3021 		if (error)
3022 			ERROUT(error);
3023 
3024 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3025 		if (nvl == NULL)
3026 			ERROUT(EBADMSG);
3027 
3028 		if (! nvlist_exists_number(nvl, "ticket"))
3029 			ERROUT(EINVAL);
3030 		ticket = nvlist_get_number(nvl, "ticket");
3031 
3032 		if (! nvlist_exists_number(nvl, "pool_ticket"))
3033 			ERROUT(EINVAL);
3034 		pool_ticket = nvlist_get_number(nvl, "pool_ticket");
3035 
3036 		if (! nvlist_exists_nvlist(nvl, "rule"))
3037 			ERROUT(EINVAL);
3038 
3039 		rule = pf_krule_alloc();
3040 		error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"),
3041 		    rule);
3042 		if (error)
3043 			ERROUT(error);
3044 
3045 		if (nvlist_exists_string(nvl, "anchor"))
3046 			anchor = nvlist_get_string(nvl, "anchor");
3047 		if (nvlist_exists_string(nvl, "anchor_call"))
3048 			anchor_call = nvlist_get_string(nvl, "anchor_call");
3049 
3050 		if ((error = nvlist_error(nvl)))
3051 			ERROUT(error);
3052 
3053 		/* Frees rule on error */
3054 		error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor,
3055 		    anchor_call, td);
3056 
3057 		nvlist_destroy(nvl);
3058 		free(nvlpacked, M_NVLIST);
3059 		break;
3060 #undef ERROUT
3061 DIOCADDRULENV_error:
3062 		pf_krule_free(rule);
3063 		nvlist_destroy(nvl);
3064 		free(nvlpacked, M_NVLIST);
3065 
3066 		break;
3067 	}
3068 	case DIOCADDRULE: {
3069 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
3070 		struct pf_krule		*rule;
3071 
3072 		rule = pf_krule_alloc();
3073 		error = pf_rule_to_krule(&pr->rule, rule);
3074 		if (error != 0) {
3075 			pf_krule_free(rule);
3076 			break;
3077 		}
3078 
3079 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
3080 
3081 		/* Frees rule on error */
3082 		error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket,
3083 		    pr->anchor, pr->anchor_call, td);
3084 		break;
3085 	}
3086 
3087 	case DIOCGETRULES: {
3088 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
3089 		struct pf_kruleset	*ruleset;
3090 		struct pf_krule		*tail;
3091 		int			 rs_num;
3092 
3093 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
3094 
3095 		PF_RULES_WLOCK();
3096 		ruleset = pf_find_kruleset(pr->anchor);
3097 		if (ruleset == NULL) {
3098 			PF_RULES_WUNLOCK();
3099 			error = EINVAL;
3100 			break;
3101 		}
3102 		rs_num = pf_get_ruleset_number(pr->rule.action);
3103 		if (rs_num >= PF_RULESET_MAX) {
3104 			PF_RULES_WUNLOCK();
3105 			error = EINVAL;
3106 			break;
3107 		}
3108 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
3109 		    pf_krulequeue);
3110 		if (tail)
3111 			pr->nr = tail->nr + 1;
3112 		else
3113 			pr->nr = 0;
3114 		pr->ticket = ruleset->rules[rs_num].active.ticket;
3115 		PF_RULES_WUNLOCK();
3116 		break;
3117 	}
3118 
3119 	case DIOCGETRULENV: {
3120 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
3121 		nvlist_t		*nvrule = NULL;
3122 		nvlist_t		*nvl = NULL;
3123 		struct pf_kruleset	*ruleset;
3124 		struct pf_krule		*rule;
3125 		void			*nvlpacked = NULL;
3126 		int			 rs_num, nr;
3127 		bool			 clear_counter = false;
3128 
3129 #define	ERROUT(x)	ERROUT_IOCTL(DIOCGETRULENV_error, x)
3130 
3131 		if (nv->len > pf_ioctl_maxcount)
3132 			ERROUT(ENOMEM);
3133 
3134 		/* Copy the request in */
3135 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3136 		if (nvlpacked == NULL)
3137 			ERROUT(ENOMEM);
3138 
3139 		error = copyin(nv->data, nvlpacked, nv->len);
3140 		if (error)
3141 			ERROUT(error);
3142 
3143 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3144 		if (nvl == NULL)
3145 			ERROUT(EBADMSG);
3146 
3147 		if (! nvlist_exists_string(nvl, "anchor"))
3148 			ERROUT(EBADMSG);
3149 		if (! nvlist_exists_number(nvl, "ruleset"))
3150 			ERROUT(EBADMSG);
3151 		if (! nvlist_exists_number(nvl, "ticket"))
3152 			ERROUT(EBADMSG);
3153 		if (! nvlist_exists_number(nvl, "nr"))
3154 			ERROUT(EBADMSG);
3155 
3156 		if (nvlist_exists_bool(nvl, "clear_counter"))
3157 			clear_counter = nvlist_get_bool(nvl, "clear_counter");
3158 
3159 		if (clear_counter && !(flags & FWRITE))
3160 			ERROUT(EACCES);
3161 
3162 		nr = nvlist_get_number(nvl, "nr");
3163 
3164 		PF_RULES_WLOCK();
3165 		ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor"));
3166 		if (ruleset == NULL) {
3167 			PF_RULES_WUNLOCK();
3168 			ERROUT(ENOENT);
3169 		}
3170 
3171 		rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset"));
3172 		if (rs_num >= PF_RULESET_MAX) {
3173 			PF_RULES_WUNLOCK();
3174 			ERROUT(EINVAL);
3175 		}
3176 
3177 		if (nvlist_get_number(nvl, "ticket") !=
3178 		    ruleset->rules[rs_num].active.ticket) {
3179 			PF_RULES_WUNLOCK();
3180 			ERROUT(EBUSY);
3181 		}
3182 
3183 		if ((error = nvlist_error(nvl))) {
3184 			PF_RULES_WUNLOCK();
3185 			ERROUT(error);
3186 		}
3187 
3188 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
3189 		while ((rule != NULL) && (rule->nr != nr))
3190 			rule = TAILQ_NEXT(rule, entries);
3191 		if (rule == NULL) {
3192 			PF_RULES_WUNLOCK();
3193 			ERROUT(EBUSY);
3194 		}
3195 
3196 		nvrule = pf_krule_to_nvrule(rule);
3197 
3198 		nvlist_destroy(nvl);
3199 		nvl = nvlist_create(0);
3200 		if (nvl == NULL) {
3201 			PF_RULES_WUNLOCK();
3202 			ERROUT(ENOMEM);
3203 		}
3204 		nvlist_add_number(nvl, "nr", nr);
3205 		nvlist_add_nvlist(nvl, "rule", nvrule);
3206 		nvlist_destroy(nvrule);
3207 		nvrule = NULL;
3208 		if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) {
3209 			PF_RULES_WUNLOCK();
3210 			ERROUT(EBUSY);
3211 		}
3212 
3213 		free(nvlpacked, M_NVLIST);
3214 		nvlpacked = nvlist_pack(nvl, &nv->len);
3215 		if (nvlpacked == NULL) {
3216 			PF_RULES_WUNLOCK();
3217 			ERROUT(ENOMEM);
3218 		}
3219 
3220 		if (nv->size == 0) {
3221 			PF_RULES_WUNLOCK();
3222 			ERROUT(0);
3223 		}
3224 		else if (nv->size < nv->len) {
3225 			PF_RULES_WUNLOCK();
3226 			ERROUT(ENOSPC);
3227 		}
3228 
3229 		if (clear_counter) {
3230 			pf_counter_u64_zero(&rule->evaluations);
3231 			for (int i = 0; i < 2; i++) {
3232 				pf_counter_u64_zero(&rule->packets[i]);
3233 				pf_counter_u64_zero(&rule->bytes[i]);
3234 			}
3235 			counter_u64_zero(rule->states_tot);
3236 		}
3237 		PF_RULES_WUNLOCK();
3238 
3239 		error = copyout(nvlpacked, nv->data, nv->len);
3240 
3241 #undef ERROUT
3242 DIOCGETRULENV_error:
3243 		free(nvlpacked, M_NVLIST);
3244 		nvlist_destroy(nvrule);
3245 		nvlist_destroy(nvl);
3246 
3247 		break;
3248 	}
3249 
3250 	case DIOCCHANGERULE: {
3251 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
3252 		struct pf_kruleset	*ruleset;
3253 		struct pf_krule		*oldrule = NULL, *newrule = NULL;
3254 		struct pfi_kkif		*kif = NULL;
3255 		struct pf_kpooladdr	*pa;
3256 		u_int32_t		 nr = 0;
3257 		int			 rs_num;
3258 
3259 		pcr->anchor[sizeof(pcr->anchor) - 1] = 0;
3260 
3261 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
3262 		    pcr->action > PF_CHANGE_GET_TICKET) {
3263 			error = EINVAL;
3264 			break;
3265 		}
3266 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
3267 			error = EINVAL;
3268 			break;
3269 		}
3270 
3271 		if (pcr->action != PF_CHANGE_REMOVE) {
3272 			newrule = pf_krule_alloc();
3273 			error = pf_rule_to_krule(&pcr->rule, newrule);
3274 			if (error != 0) {
3275 				pf_krule_free(newrule);
3276 				break;
3277 			}
3278 
3279 			if (newrule->ifname[0])
3280 				kif = pf_kkif_create(M_WAITOK);
3281 			pf_counter_u64_init(&newrule->evaluations, M_WAITOK);
3282 			for (int i = 0; i < 2; i++) {
3283 				pf_counter_u64_init(&newrule->packets[i], M_WAITOK);
3284 				pf_counter_u64_init(&newrule->bytes[i], M_WAITOK);
3285 			}
3286 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
3287 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
3288 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
3289 			newrule->cuid = td->td_ucred->cr_ruid;
3290 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
3291 			TAILQ_INIT(&newrule->rpool.list);
3292 		}
3293 #define	ERROUT(x)	ERROUT_IOCTL(DIOCCHANGERULE_error, x)
3294 
3295 		PF_CONFIG_LOCK();
3296 		PF_RULES_WLOCK();
3297 #ifdef PF_WANT_32_TO_64_COUNTER
3298 		if (newrule != NULL) {
3299 			LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist);
3300 			newrule->allrulelinked = true;
3301 			V_pf_allrulecount++;
3302 		}
3303 #endif
3304 
3305 		if (!(pcr->action == PF_CHANGE_REMOVE ||
3306 		    pcr->action == PF_CHANGE_GET_TICKET) &&
3307 		    pcr->pool_ticket != V_ticket_pabuf)
3308 			ERROUT(EBUSY);
3309 
3310 		ruleset = pf_find_kruleset(pcr->anchor);
3311 		if (ruleset == NULL)
3312 			ERROUT(EINVAL);
3313 
3314 		rs_num = pf_get_ruleset_number(pcr->rule.action);
3315 		if (rs_num >= PF_RULESET_MAX)
3316 			ERROUT(EINVAL);
3317 
3318 		/*
3319 		 * XXXMJG: there is no guarantee that the ruleset was
3320 		 * created by the usual route of calling DIOCXBEGIN.
3321 		 * As a result it is possible the rule tree will not
3322 		 * be allocated yet. Hack around it by doing it here.
3323 		 * Note it is fine to let the tree persist in case of
3324 		 * error as it will be freed down the road on future
3325 		 * updates (if need be).
3326 		 */
3327 		if (ruleset->rules[rs_num].active.tree == NULL) {
3328 			ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT);
3329 			if (ruleset->rules[rs_num].active.tree == NULL) {
3330 				ERROUT(ENOMEM);
3331 			}
3332 		}
3333 
3334 		if (pcr->action == PF_CHANGE_GET_TICKET) {
3335 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
3336 			ERROUT(0);
3337 		} else if (pcr->ticket !=
3338 			    ruleset->rules[rs_num].active.ticket)
3339 				ERROUT(EINVAL);
3340 
3341 		if (pcr->action != PF_CHANGE_REMOVE) {
3342 			if (newrule->ifname[0]) {
3343 				newrule->kif = pfi_kkif_attach(kif,
3344 				    newrule->ifname);
3345 				kif = NULL;
3346 				pfi_kkif_ref(newrule->kif);
3347 			} else
3348 				newrule->kif = NULL;
3349 
3350 			if (newrule->rtableid > 0 &&
3351 			    newrule->rtableid >= rt_numfibs)
3352 				error = EBUSY;
3353 
3354 #ifdef ALTQ
3355 			/* set queue IDs */
3356 			if (newrule->qname[0] != 0) {
3357 				if ((newrule->qid =
3358 				    pf_qname2qid(newrule->qname)) == 0)
3359 					error = EBUSY;
3360 				else if (newrule->pqname[0] != 0) {
3361 					if ((newrule->pqid =
3362 					    pf_qname2qid(newrule->pqname)) == 0)
3363 						error = EBUSY;
3364 				} else
3365 					newrule->pqid = newrule->qid;
3366 			}
3367 #endif /* ALTQ */
3368 			if (newrule->tagname[0])
3369 				if ((newrule->tag =
3370 				    pf_tagname2tag(newrule->tagname)) == 0)
3371 					error = EBUSY;
3372 			if (newrule->match_tagname[0])
3373 				if ((newrule->match_tag = pf_tagname2tag(
3374 				    newrule->match_tagname)) == 0)
3375 					error = EBUSY;
3376 			if (newrule->rt && !newrule->direction)
3377 				error = EINVAL;
3378 			if (!newrule->log)
3379 				newrule->logif = 0;
3380 			if (newrule->logif >= PFLOGIFS_MAX)
3381 				error = EINVAL;
3382 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
3383 				error = ENOMEM;
3384 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
3385 				error = ENOMEM;
3386 			if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
3387 				error = EINVAL;
3388 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
3389 				if (pa->addr.type == PF_ADDR_TABLE) {
3390 					pa->addr.p.tbl =
3391 					    pfr_attach_table(ruleset,
3392 					    pa->addr.v.tblname);
3393 					if (pa->addr.p.tbl == NULL)
3394 						error = ENOMEM;
3395 				}
3396 
3397 			newrule->overload_tbl = NULL;
3398 			if (newrule->overload_tblname[0]) {
3399 				if ((newrule->overload_tbl = pfr_attach_table(
3400 				    ruleset, newrule->overload_tblname)) ==
3401 				    NULL)
3402 					error = EINVAL;
3403 				else
3404 					newrule->overload_tbl->pfrkt_flags |=
3405 					    PFR_TFLAG_ACTIVE;
3406 			}
3407 
3408 			pf_mv_kpool(&V_pf_pabuf, &newrule->rpool.list);
3409 			if (((((newrule->action == PF_NAT) ||
3410 			    (newrule->action == PF_RDR) ||
3411 			    (newrule->action == PF_BINAT) ||
3412 			    (newrule->rt > PF_NOPFROUTE)) &&
3413 			    !newrule->anchor)) &&
3414 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
3415 				error = EINVAL;
3416 
3417 			if (error) {
3418 				pf_free_rule(newrule);
3419 				PF_RULES_WUNLOCK();
3420 				PF_CONFIG_UNLOCK();
3421 				break;
3422 			}
3423 
3424 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
3425 		}
3426 		pf_empty_kpool(&V_pf_pabuf);
3427 
3428 		if (pcr->action == PF_CHANGE_ADD_HEAD)
3429 			oldrule = TAILQ_FIRST(
3430 			    ruleset->rules[rs_num].active.ptr);
3431 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
3432 			oldrule = TAILQ_LAST(
3433 			    ruleset->rules[rs_num].active.ptr, pf_krulequeue);
3434 		else {
3435 			oldrule = TAILQ_FIRST(
3436 			    ruleset->rules[rs_num].active.ptr);
3437 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
3438 				oldrule = TAILQ_NEXT(oldrule, entries);
3439 			if (oldrule == NULL) {
3440 				if (newrule != NULL)
3441 					pf_free_rule(newrule);
3442 				PF_RULES_WUNLOCK();
3443 				PF_CONFIG_UNLOCK();
3444 				error = EINVAL;
3445 				break;
3446 			}
3447 		}
3448 
3449 		if (pcr->action == PF_CHANGE_REMOVE) {
3450 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
3451 			    oldrule);
3452 			RB_REMOVE(pf_krule_global,
3453 			    ruleset->rules[rs_num].active.tree, oldrule);
3454 			ruleset->rules[rs_num].active.rcount--;
3455 		} else {
3456 			pf_hash_rule(newrule);
3457 			if (RB_INSERT(pf_krule_global,
3458 			    ruleset->rules[rs_num].active.tree, newrule) != NULL) {
3459 				pf_free_rule(newrule);
3460 				PF_RULES_WUNLOCK();
3461 				PF_CONFIG_UNLOCK();
3462 				error = EEXIST;
3463 				break;
3464 			}
3465 
3466 			if (oldrule == NULL)
3467 				TAILQ_INSERT_TAIL(
3468 				    ruleset->rules[rs_num].active.ptr,
3469 				    newrule, entries);
3470 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
3471 			    pcr->action == PF_CHANGE_ADD_BEFORE)
3472 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
3473 			else
3474 				TAILQ_INSERT_AFTER(
3475 				    ruleset->rules[rs_num].active.ptr,
3476 				    oldrule, newrule, entries);
3477 			ruleset->rules[rs_num].active.rcount++;
3478 		}
3479 
3480 		nr = 0;
3481 		TAILQ_FOREACH(oldrule,
3482 		    ruleset->rules[rs_num].active.ptr, entries)
3483 			oldrule->nr = nr++;
3484 
3485 		ruleset->rules[rs_num].active.ticket++;
3486 
3487 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
3488 		pf_remove_if_empty_kruleset(ruleset);
3489 
3490 		PF_RULES_WUNLOCK();
3491 		PF_CONFIG_UNLOCK();
3492 		break;
3493 
3494 #undef ERROUT
3495 DIOCCHANGERULE_error:
3496 		PF_RULES_WUNLOCK();
3497 		PF_CONFIG_UNLOCK();
3498 		pf_krule_free(newrule);
3499 		pf_kkif_free(kif);
3500 		break;
3501 	}
3502 
3503 	case DIOCCLRSTATESNV: {
3504 		error = pf_clearstates_nv((struct pfioc_nv *)addr);
3505 		break;
3506 	}
3507 
3508 	case DIOCKILLSTATESNV: {
3509 		error = pf_killstates_nv((struct pfioc_nv *)addr);
3510 		break;
3511 	}
3512 
3513 	case DIOCADDSTATE: {
3514 		struct pfioc_state		*ps = (struct pfioc_state *)addr;
3515 		struct pfsync_state_1301	*sp = &ps->state;
3516 
3517 		if (sp->timeout >= PFTM_MAX) {
3518 			error = EINVAL;
3519 			break;
3520 		}
3521 		if (V_pfsync_state_import_ptr != NULL) {
3522 			PF_RULES_RLOCK();
3523 			error = V_pfsync_state_import_ptr(
3524 			    (union pfsync_state_union *)sp, PFSYNC_SI_IOCTL,
3525 			    PFSYNC_MSG_VERSION_1301);
3526 			PF_RULES_RUNLOCK();
3527 		} else
3528 			error = EOPNOTSUPP;
3529 		break;
3530 	}
3531 
3532 	case DIOCGETSTATE: {
3533 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
3534 		struct pf_kstate	*s;
3535 
3536 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
3537 		if (s == NULL) {
3538 			error = ENOENT;
3539 			break;
3540 		}
3541 
3542 		pfsync_state_export((union pfsync_state_union*)&ps->state,
3543 		    s, PFSYNC_MSG_VERSION_1301);
3544 		PF_STATE_UNLOCK(s);
3545 		break;
3546 	}
3547 
3548 	case DIOCGETSTATENV: {
3549 		error = pf_getstate((struct pfioc_nv *)addr);
3550 		break;
3551 	}
3552 
3553 #ifdef COMPAT_FREEBSD14
3554 	case DIOCGETSTATES: {
3555 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
3556 		struct pf_kstate	*s;
3557 		struct pfsync_state_1301	*pstore, *p;
3558 		int			 i, nr;
3559 		size_t			 slice_count = 16, count;
3560 		void			*out;
3561 
3562 		if (ps->ps_len <= 0) {
3563 			nr = uma_zone_get_cur(V_pf_state_z);
3564 			ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3565 			break;
3566 		}
3567 
3568 		out = ps->ps_states;
3569 		pstore = mallocarray(slice_count,
3570 		    sizeof(struct pfsync_state_1301), M_TEMP, M_WAITOK | M_ZERO);
3571 		nr = 0;
3572 
3573 		for (i = 0; i <= pf_hashmask; i++) {
3574 			struct pf_idhash *ih = &V_pf_idhash[i];
3575 
3576 DIOCGETSTATES_retry:
3577 			p = pstore;
3578 
3579 			if (LIST_EMPTY(&ih->states))
3580 				continue;
3581 
3582 			PF_HASHROW_LOCK(ih);
3583 			count = 0;
3584 			LIST_FOREACH(s, &ih->states, entry) {
3585 				if (s->timeout == PFTM_UNLINKED)
3586 					continue;
3587 				count++;
3588 			}
3589 
3590 			if (count > slice_count) {
3591 				PF_HASHROW_UNLOCK(ih);
3592 				free(pstore, M_TEMP);
3593 				slice_count = count * 2;
3594 				pstore = mallocarray(slice_count,
3595 				    sizeof(struct pfsync_state_1301), M_TEMP,
3596 				    M_WAITOK | M_ZERO);
3597 				goto DIOCGETSTATES_retry;
3598 			}
3599 
3600 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
3601 				PF_HASHROW_UNLOCK(ih);
3602 				goto DIOCGETSTATES_full;
3603 			}
3604 
3605 			LIST_FOREACH(s, &ih->states, entry) {
3606 				if (s->timeout == PFTM_UNLINKED)
3607 					continue;
3608 
3609 				pfsync_state_export((union pfsync_state_union*)p,
3610 				    s, PFSYNC_MSG_VERSION_1301);
3611 				p++;
3612 				nr++;
3613 			}
3614 			PF_HASHROW_UNLOCK(ih);
3615 			error = copyout(pstore, out,
3616 			    sizeof(struct pfsync_state_1301) * count);
3617 			if (error)
3618 				break;
3619 			out = ps->ps_states + nr;
3620 		}
3621 DIOCGETSTATES_full:
3622 		ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3623 		free(pstore, M_TEMP);
3624 
3625 		break;
3626 	}
3627 
3628 	case DIOCGETSTATESV2: {
3629 		struct pfioc_states_v2	*ps = (struct pfioc_states_v2 *)addr;
3630 		struct pf_kstate	*s;
3631 		struct pf_state_export	*pstore, *p;
3632 		int i, nr;
3633 		size_t slice_count = 16, count;
3634 		void *out;
3635 
3636 		if (ps->ps_req_version > PF_STATE_VERSION) {
3637 			error = ENOTSUP;
3638 			break;
3639 		}
3640 
3641 		if (ps->ps_len <= 0) {
3642 			nr = uma_zone_get_cur(V_pf_state_z);
3643 			ps->ps_len = sizeof(struct pf_state_export) * nr;
3644 			break;
3645 		}
3646 
3647 		out = ps->ps_states;
3648 		pstore = mallocarray(slice_count,
3649 		    sizeof(struct pf_state_export), M_TEMP, M_WAITOK | M_ZERO);
3650 		nr = 0;
3651 
3652 		for (i = 0; i <= pf_hashmask; i++) {
3653 			struct pf_idhash *ih = &V_pf_idhash[i];
3654 
3655 DIOCGETSTATESV2_retry:
3656 			p = pstore;
3657 
3658 			if (LIST_EMPTY(&ih->states))
3659 				continue;
3660 
3661 			PF_HASHROW_LOCK(ih);
3662 			count = 0;
3663 			LIST_FOREACH(s, &ih->states, entry) {
3664 				if (s->timeout == PFTM_UNLINKED)
3665 					continue;
3666 				count++;
3667 			}
3668 
3669 			if (count > slice_count) {
3670 				PF_HASHROW_UNLOCK(ih);
3671 				free(pstore, M_TEMP);
3672 				slice_count = count * 2;
3673 				pstore = mallocarray(slice_count,
3674 				    sizeof(struct pf_state_export), M_TEMP,
3675 				    M_WAITOK | M_ZERO);
3676 				goto DIOCGETSTATESV2_retry;
3677 			}
3678 
3679 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
3680 				PF_HASHROW_UNLOCK(ih);
3681 				goto DIOCGETSTATESV2_full;
3682 			}
3683 
3684 			LIST_FOREACH(s, &ih->states, entry) {
3685 				if (s->timeout == PFTM_UNLINKED)
3686 					continue;
3687 
3688 				pf_state_export(p, s);
3689 				p++;
3690 				nr++;
3691 			}
3692 			PF_HASHROW_UNLOCK(ih);
3693 			error = copyout(pstore, out,
3694 			    sizeof(struct pf_state_export) * count);
3695 			if (error)
3696 				break;
3697 			out = ps->ps_states + nr;
3698 		}
3699 DIOCGETSTATESV2_full:
3700 		ps->ps_len = nr * sizeof(struct pf_state_export);
3701 		free(pstore, M_TEMP);
3702 
3703 		break;
3704 	}
3705 #endif
3706 	case DIOCGETSTATUSNV: {
3707 		error = pf_getstatus((struct pfioc_nv *)addr);
3708 		break;
3709 	}
3710 
3711 	case DIOCSETSTATUSIF: {
3712 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
3713 
3714 		if (pi->ifname[0] == 0) {
3715 			bzero(V_pf_status.ifname, IFNAMSIZ);
3716 			break;
3717 		}
3718 		PF_RULES_WLOCK();
3719 		error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
3720 		PF_RULES_WUNLOCK();
3721 		break;
3722 	}
3723 
3724 	case DIOCCLRSTATUS: {
3725 		PF_RULES_WLOCK();
3726 		for (int i = 0; i < PFRES_MAX; i++)
3727 			counter_u64_zero(V_pf_status.counters[i]);
3728 		for (int i = 0; i < FCNT_MAX; i++)
3729 			pf_counter_u64_zero(&V_pf_status.fcounters[i]);
3730 		for (int i = 0; i < SCNT_MAX; i++)
3731 			counter_u64_zero(V_pf_status.scounters[i]);
3732 		for (int i = 0; i < KLCNT_MAX; i++)
3733 			counter_u64_zero(V_pf_status.lcounters[i]);
3734 		V_pf_status.since = time_second;
3735 		if (*V_pf_status.ifname)
3736 			pfi_update_status(V_pf_status.ifname, NULL);
3737 		PF_RULES_WUNLOCK();
3738 		break;
3739 	}
3740 
3741 	case DIOCNATLOOK: {
3742 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
3743 		struct pf_state_key	*sk;
3744 		struct pf_kstate	*state;
3745 		struct pf_state_key_cmp	 key;
3746 		int			 m = 0, direction = pnl->direction;
3747 		int			 sidx, didx;
3748 
3749 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
3750 		sidx = (direction == PF_IN) ? 1 : 0;
3751 		didx = (direction == PF_IN) ? 0 : 1;
3752 
3753 		if (!pnl->proto ||
3754 		    PF_AZERO(&pnl->saddr, pnl->af) ||
3755 		    PF_AZERO(&pnl->daddr, pnl->af) ||
3756 		    ((pnl->proto == IPPROTO_TCP ||
3757 		    pnl->proto == IPPROTO_UDP) &&
3758 		    (!pnl->dport || !pnl->sport)))
3759 			error = EINVAL;
3760 		else {
3761 			bzero(&key, sizeof(key));
3762 			key.af = pnl->af;
3763 			key.proto = pnl->proto;
3764 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
3765 			key.port[sidx] = pnl->sport;
3766 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
3767 			key.port[didx] = pnl->dport;
3768 
3769 			state = pf_find_state_all(&key, direction, &m);
3770 			if (state == NULL) {
3771 				error = ENOENT;
3772 			} else {
3773 				if (m > 1) {
3774 					PF_STATE_UNLOCK(state);
3775 					error = E2BIG;	/* more than one state */
3776 				} else {
3777 					sk = state->key[sidx];
3778 					PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
3779 					pnl->rsport = sk->port[sidx];
3780 					PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
3781 					pnl->rdport = sk->port[didx];
3782 					PF_STATE_UNLOCK(state);
3783 				}
3784 			}
3785 		}
3786 		break;
3787 	}
3788 
3789 	case DIOCSETTIMEOUT: {
3790 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
3791 		int		 old;
3792 
3793 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
3794 		    pt->seconds < 0) {
3795 			error = EINVAL;
3796 			break;
3797 		}
3798 		PF_RULES_WLOCK();
3799 		old = V_pf_default_rule.timeout[pt->timeout];
3800 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
3801 			pt->seconds = 1;
3802 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
3803 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
3804 			wakeup(pf_purge_thread);
3805 		pt->seconds = old;
3806 		PF_RULES_WUNLOCK();
3807 		break;
3808 	}
3809 
3810 	case DIOCGETTIMEOUT: {
3811 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
3812 
3813 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
3814 			error = EINVAL;
3815 			break;
3816 		}
3817 		PF_RULES_RLOCK();
3818 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
3819 		PF_RULES_RUNLOCK();
3820 		break;
3821 	}
3822 
3823 	case DIOCGETLIMIT: {
3824 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
3825 
3826 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
3827 			error = EINVAL;
3828 			break;
3829 		}
3830 		PF_RULES_RLOCK();
3831 		pl->limit = V_pf_limits[pl->index].limit;
3832 		PF_RULES_RUNLOCK();
3833 		break;
3834 	}
3835 
3836 	case DIOCSETLIMIT: {
3837 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
3838 		int			 old_limit;
3839 
3840 		PF_RULES_WLOCK();
3841 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
3842 		    V_pf_limits[pl->index].zone == NULL) {
3843 			PF_RULES_WUNLOCK();
3844 			error = EINVAL;
3845 			break;
3846 		}
3847 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
3848 		old_limit = V_pf_limits[pl->index].limit;
3849 		V_pf_limits[pl->index].limit = pl->limit;
3850 		pl->limit = old_limit;
3851 		PF_RULES_WUNLOCK();
3852 		break;
3853 	}
3854 
3855 	case DIOCSETDEBUG: {
3856 		u_int32_t	*level = (u_int32_t *)addr;
3857 
3858 		PF_RULES_WLOCK();
3859 		V_pf_status.debug = *level;
3860 		PF_RULES_WUNLOCK();
3861 		break;
3862 	}
3863 
3864 	case DIOCCLRRULECTRS: {
3865 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
3866 		struct pf_kruleset	*ruleset = &pf_main_ruleset;
3867 		struct pf_krule		*rule;
3868 
3869 		PF_RULES_WLOCK();
3870 		TAILQ_FOREACH(rule,
3871 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
3872 			pf_counter_u64_zero(&rule->evaluations);
3873 			for (int i = 0; i < 2; i++) {
3874 				pf_counter_u64_zero(&rule->packets[i]);
3875 				pf_counter_u64_zero(&rule->bytes[i]);
3876 			}
3877 		}
3878 		PF_RULES_WUNLOCK();
3879 		break;
3880 	}
3881 
3882 	case DIOCGIFSPEEDV0:
3883 	case DIOCGIFSPEEDV1: {
3884 		struct pf_ifspeed_v1	*psp = (struct pf_ifspeed_v1 *)addr;
3885 		struct pf_ifspeed_v1	ps;
3886 		struct ifnet		*ifp;
3887 
3888 		if (psp->ifname[0] == '\0') {
3889 			error = EINVAL;
3890 			break;
3891 		}
3892 
3893 		error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ);
3894 		if (error != 0)
3895 			break;
3896 		ifp = ifunit(ps.ifname);
3897 		if (ifp != NULL) {
3898 			psp->baudrate32 =
3899 			    (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
3900 			if (cmd == DIOCGIFSPEEDV1)
3901 				psp->baudrate = ifp->if_baudrate;
3902 		} else {
3903 			error = EINVAL;
3904 		}
3905 		break;
3906 	}
3907 
3908 #ifdef ALTQ
3909 	case DIOCSTARTALTQ: {
3910 		struct pf_altq		*altq;
3911 
3912 		PF_RULES_WLOCK();
3913 		/* enable all altq interfaces on active list */
3914 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
3915 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
3916 				error = pf_enable_altq(altq);
3917 				if (error != 0)
3918 					break;
3919 			}
3920 		}
3921 		if (error == 0)
3922 			V_pf_altq_running = 1;
3923 		PF_RULES_WUNLOCK();
3924 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
3925 		break;
3926 	}
3927 
3928 	case DIOCSTOPALTQ: {
3929 		struct pf_altq		*altq;
3930 
3931 		PF_RULES_WLOCK();
3932 		/* disable all altq interfaces on active list */
3933 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
3934 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
3935 				error = pf_disable_altq(altq);
3936 				if (error != 0)
3937 					break;
3938 			}
3939 		}
3940 		if (error == 0)
3941 			V_pf_altq_running = 0;
3942 		PF_RULES_WUNLOCK();
3943 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
3944 		break;
3945 	}
3946 
3947 	case DIOCADDALTQV0:
3948 	case DIOCADDALTQV1: {
3949 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
3950 		struct pf_altq		*altq, *a;
3951 		struct ifnet		*ifp;
3952 
3953 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
3954 		error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
3955 		if (error)
3956 			break;
3957 		altq->local_flags = 0;
3958 
3959 		PF_RULES_WLOCK();
3960 		if (pa->ticket != V_ticket_altqs_inactive) {
3961 			PF_RULES_WUNLOCK();
3962 			free(altq, M_PFALTQ);
3963 			error = EBUSY;
3964 			break;
3965 		}
3966 
3967 		/*
3968 		 * if this is for a queue, find the discipline and
3969 		 * copy the necessary fields
3970 		 */
3971 		if (altq->qname[0] != 0) {
3972 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
3973 				PF_RULES_WUNLOCK();
3974 				error = EBUSY;
3975 				free(altq, M_PFALTQ);
3976 				break;
3977 			}
3978 			altq->altq_disc = NULL;
3979 			TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
3980 				if (strncmp(a->ifname, altq->ifname,
3981 				    IFNAMSIZ) == 0) {
3982 					altq->altq_disc = a->altq_disc;
3983 					break;
3984 				}
3985 			}
3986 		}
3987 
3988 		if ((ifp = ifunit(altq->ifname)) == NULL)
3989 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
3990 		else
3991 			error = altq_add(ifp, altq);
3992 
3993 		if (error) {
3994 			PF_RULES_WUNLOCK();
3995 			free(altq, M_PFALTQ);
3996 			break;
3997 		}
3998 
3999 		if (altq->qname[0] != 0)
4000 			TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
4001 		else
4002 			TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
4003 		/* version error check done on import above */
4004 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4005 		PF_RULES_WUNLOCK();
4006 		break;
4007 	}
4008 
4009 	case DIOCGETALTQSV0:
4010 	case DIOCGETALTQSV1: {
4011 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4012 		struct pf_altq		*altq;
4013 
4014 		PF_RULES_RLOCK();
4015 		pa->nr = 0;
4016 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
4017 			pa->nr++;
4018 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
4019 			pa->nr++;
4020 		pa->ticket = V_ticket_altqs_active;
4021 		PF_RULES_RUNLOCK();
4022 		break;
4023 	}
4024 
4025 	case DIOCGETALTQV0:
4026 	case DIOCGETALTQV1: {
4027 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4028 		struct pf_altq		*altq;
4029 
4030 		PF_RULES_RLOCK();
4031 		if (pa->ticket != V_ticket_altqs_active) {
4032 			PF_RULES_RUNLOCK();
4033 			error = EBUSY;
4034 			break;
4035 		}
4036 		altq = pf_altq_get_nth_active(pa->nr);
4037 		if (altq == NULL) {
4038 			PF_RULES_RUNLOCK();
4039 			error = EBUSY;
4040 			break;
4041 		}
4042 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4043 		PF_RULES_RUNLOCK();
4044 		break;
4045 	}
4046 
4047 	case DIOCCHANGEALTQV0:
4048 	case DIOCCHANGEALTQV1:
4049 		/* CHANGEALTQ not supported yet! */
4050 		error = ENODEV;
4051 		break;
4052 
4053 	case DIOCGETQSTATSV0:
4054 	case DIOCGETQSTATSV1: {
4055 		struct pfioc_qstats_v1	*pq = (struct pfioc_qstats_v1 *)addr;
4056 		struct pf_altq		*altq;
4057 		int			 nbytes;
4058 		u_int32_t		 version;
4059 
4060 		PF_RULES_RLOCK();
4061 		if (pq->ticket != V_ticket_altqs_active) {
4062 			PF_RULES_RUNLOCK();
4063 			error = EBUSY;
4064 			break;
4065 		}
4066 		nbytes = pq->nbytes;
4067 		altq = pf_altq_get_nth_active(pq->nr);
4068 		if (altq == NULL) {
4069 			PF_RULES_RUNLOCK();
4070 			error = EBUSY;
4071 			break;
4072 		}
4073 
4074 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
4075 			PF_RULES_RUNLOCK();
4076 			error = ENXIO;
4077 			break;
4078 		}
4079 		PF_RULES_RUNLOCK();
4080 		if (cmd == DIOCGETQSTATSV0)
4081 			version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
4082 		else
4083 			version = pq->version;
4084 		error = altq_getqstats(altq, pq->buf, &nbytes, version);
4085 		if (error == 0) {
4086 			pq->scheduler = altq->scheduler;
4087 			pq->nbytes = nbytes;
4088 		}
4089 		break;
4090 	}
4091 #endif /* ALTQ */
4092 
4093 	case DIOCBEGINADDRS: {
4094 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4095 
4096 		PF_RULES_WLOCK();
4097 		pf_empty_kpool(&V_pf_pabuf);
4098 		pp->ticket = ++V_ticket_pabuf;
4099 		PF_RULES_WUNLOCK();
4100 		break;
4101 	}
4102 
4103 	case DIOCADDADDR: {
4104 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4105 		struct pf_kpooladdr	*pa;
4106 		struct pfi_kkif		*kif = NULL;
4107 
4108 #ifndef INET
4109 		if (pp->af == AF_INET) {
4110 			error = EAFNOSUPPORT;
4111 			break;
4112 		}
4113 #endif /* INET */
4114 #ifndef INET6
4115 		if (pp->af == AF_INET6) {
4116 			error = EAFNOSUPPORT;
4117 			break;
4118 		}
4119 #endif /* INET6 */
4120 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
4121 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
4122 		    pp->addr.addr.type != PF_ADDR_TABLE) {
4123 			error = EINVAL;
4124 			break;
4125 		}
4126 		if (pp->addr.addr.p.dyn != NULL) {
4127 			error = EINVAL;
4128 			break;
4129 		}
4130 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
4131 		error = pf_pooladdr_to_kpooladdr(&pp->addr, pa);
4132 		if (error != 0)
4133 			break;
4134 		if (pa->ifname[0])
4135 			kif = pf_kkif_create(M_WAITOK);
4136 		PF_RULES_WLOCK();
4137 		if (pp->ticket != V_ticket_pabuf) {
4138 			PF_RULES_WUNLOCK();
4139 			if (pa->ifname[0])
4140 				pf_kkif_free(kif);
4141 			free(pa, M_PFRULE);
4142 			error = EBUSY;
4143 			break;
4144 		}
4145 		if (pa->ifname[0]) {
4146 			pa->kif = pfi_kkif_attach(kif, pa->ifname);
4147 			kif = NULL;
4148 			pfi_kkif_ref(pa->kif);
4149 		} else
4150 			pa->kif = NULL;
4151 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
4152 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
4153 			if (pa->ifname[0])
4154 				pfi_kkif_unref(pa->kif);
4155 			PF_RULES_WUNLOCK();
4156 			free(pa, M_PFRULE);
4157 			break;
4158 		}
4159 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
4160 		PF_RULES_WUNLOCK();
4161 		break;
4162 	}
4163 
4164 	case DIOCGETADDRS: {
4165 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4166 		struct pf_kpool		*pool;
4167 		struct pf_kpooladdr	*pa;
4168 
4169 		pp->anchor[sizeof(pp->anchor) - 1] = 0;
4170 		pp->nr = 0;
4171 
4172 		PF_RULES_RLOCK();
4173 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
4174 		    pp->r_num, 0, 1, 0);
4175 		if (pool == NULL) {
4176 			PF_RULES_RUNLOCK();
4177 			error = EBUSY;
4178 			break;
4179 		}
4180 		TAILQ_FOREACH(pa, &pool->list, entries)
4181 			pp->nr++;
4182 		PF_RULES_RUNLOCK();
4183 		break;
4184 	}
4185 
4186 	case DIOCGETADDR: {
4187 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4188 		struct pf_kpool		*pool;
4189 		struct pf_kpooladdr	*pa;
4190 		u_int32_t		 nr = 0;
4191 
4192 		pp->anchor[sizeof(pp->anchor) - 1] = 0;
4193 
4194 		PF_RULES_RLOCK();
4195 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
4196 		    pp->r_num, 0, 1, 1);
4197 		if (pool == NULL) {
4198 			PF_RULES_RUNLOCK();
4199 			error = EBUSY;
4200 			break;
4201 		}
4202 		pa = TAILQ_FIRST(&pool->list);
4203 		while ((pa != NULL) && (nr < pp->nr)) {
4204 			pa = TAILQ_NEXT(pa, entries);
4205 			nr++;
4206 		}
4207 		if (pa == NULL) {
4208 			PF_RULES_RUNLOCK();
4209 			error = EBUSY;
4210 			break;
4211 		}
4212 		pf_kpooladdr_to_pooladdr(pa, &pp->addr);
4213 		pf_addr_copyout(&pp->addr.addr);
4214 		PF_RULES_RUNLOCK();
4215 		break;
4216 	}
4217 
4218 	case DIOCCHANGEADDR: {
4219 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
4220 		struct pf_kpool		*pool;
4221 		struct pf_kpooladdr	*oldpa = NULL, *newpa = NULL;
4222 		struct pf_kruleset	*ruleset;
4223 		struct pfi_kkif		*kif = NULL;
4224 
4225 		pca->anchor[sizeof(pca->anchor) - 1] = 0;
4226 
4227 		if (pca->action < PF_CHANGE_ADD_HEAD ||
4228 		    pca->action > PF_CHANGE_REMOVE) {
4229 			error = EINVAL;
4230 			break;
4231 		}
4232 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
4233 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
4234 		    pca->addr.addr.type != PF_ADDR_TABLE) {
4235 			error = EINVAL;
4236 			break;
4237 		}
4238 		if (pca->addr.addr.p.dyn != NULL) {
4239 			error = EINVAL;
4240 			break;
4241 		}
4242 
4243 		if (pca->action != PF_CHANGE_REMOVE) {
4244 #ifndef INET
4245 			if (pca->af == AF_INET) {
4246 				error = EAFNOSUPPORT;
4247 				break;
4248 			}
4249 #endif /* INET */
4250 #ifndef INET6
4251 			if (pca->af == AF_INET6) {
4252 				error = EAFNOSUPPORT;
4253 				break;
4254 			}
4255 #endif /* INET6 */
4256 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
4257 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
4258 			if (newpa->ifname[0])
4259 				kif = pf_kkif_create(M_WAITOK);
4260 			newpa->kif = NULL;
4261 		}
4262 #define	ERROUT(x)	ERROUT_IOCTL(DIOCCHANGEADDR_error, x)
4263 		PF_RULES_WLOCK();
4264 		ruleset = pf_find_kruleset(pca->anchor);
4265 		if (ruleset == NULL)
4266 			ERROUT(EBUSY);
4267 
4268 		pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
4269 		    pca->r_num, pca->r_last, 1, 1);
4270 		if (pool == NULL)
4271 			ERROUT(EBUSY);
4272 
4273 		if (pca->action != PF_CHANGE_REMOVE) {
4274 			if (newpa->ifname[0]) {
4275 				newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
4276 				pfi_kkif_ref(newpa->kif);
4277 				kif = NULL;
4278 			}
4279 
4280 			switch (newpa->addr.type) {
4281 			case PF_ADDR_DYNIFTL:
4282 				error = pfi_dynaddr_setup(&newpa->addr,
4283 				    pca->af);
4284 				break;
4285 			case PF_ADDR_TABLE:
4286 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
4287 				    newpa->addr.v.tblname);
4288 				if (newpa->addr.p.tbl == NULL)
4289 					error = ENOMEM;
4290 				break;
4291 			}
4292 			if (error)
4293 				goto DIOCCHANGEADDR_error;
4294 		}
4295 
4296 		switch (pca->action) {
4297 		case PF_CHANGE_ADD_HEAD:
4298 			oldpa = TAILQ_FIRST(&pool->list);
4299 			break;
4300 		case PF_CHANGE_ADD_TAIL:
4301 			oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
4302 			break;
4303 		default:
4304 			oldpa = TAILQ_FIRST(&pool->list);
4305 			for (int i = 0; oldpa && i < pca->nr; i++)
4306 				oldpa = TAILQ_NEXT(oldpa, entries);
4307 
4308 			if (oldpa == NULL)
4309 				ERROUT(EINVAL);
4310 		}
4311 
4312 		if (pca->action == PF_CHANGE_REMOVE) {
4313 			TAILQ_REMOVE(&pool->list, oldpa, entries);
4314 			switch (oldpa->addr.type) {
4315 			case PF_ADDR_DYNIFTL:
4316 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
4317 				break;
4318 			case PF_ADDR_TABLE:
4319 				pfr_detach_table(oldpa->addr.p.tbl);
4320 				break;
4321 			}
4322 			if (oldpa->kif)
4323 				pfi_kkif_unref(oldpa->kif);
4324 			free(oldpa, M_PFRULE);
4325 		} else {
4326 			if (oldpa == NULL)
4327 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
4328 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
4329 			    pca->action == PF_CHANGE_ADD_BEFORE)
4330 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
4331 			else
4332 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
4333 				    newpa, entries);
4334 		}
4335 
4336 		pool->cur = TAILQ_FIRST(&pool->list);
4337 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
4338 		PF_RULES_WUNLOCK();
4339 		break;
4340 
4341 #undef ERROUT
4342 DIOCCHANGEADDR_error:
4343 		if (newpa != NULL) {
4344 			if (newpa->kif)
4345 				pfi_kkif_unref(newpa->kif);
4346 			free(newpa, M_PFRULE);
4347 		}
4348 		PF_RULES_WUNLOCK();
4349 		pf_kkif_free(kif);
4350 		break;
4351 	}
4352 
4353 	case DIOCGETRULESETS: {
4354 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
4355 		struct pf_kruleset	*ruleset;
4356 		struct pf_kanchor	*anchor;
4357 
4358 		pr->path[sizeof(pr->path) - 1] = 0;
4359 
4360 		PF_RULES_RLOCK();
4361 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4362 			PF_RULES_RUNLOCK();
4363 			error = ENOENT;
4364 			break;
4365 		}
4366 		pr->nr = 0;
4367 		if (ruleset->anchor == NULL) {
4368 			/* XXX kludge for pf_main_ruleset */
4369 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4370 				if (anchor->parent == NULL)
4371 					pr->nr++;
4372 		} else {
4373 			RB_FOREACH(anchor, pf_kanchor_node,
4374 			    &ruleset->anchor->children)
4375 				pr->nr++;
4376 		}
4377 		PF_RULES_RUNLOCK();
4378 		break;
4379 	}
4380 
4381 	case DIOCGETRULESET: {
4382 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
4383 		struct pf_kruleset	*ruleset;
4384 		struct pf_kanchor	*anchor;
4385 		u_int32_t		 nr = 0;
4386 
4387 		pr->path[sizeof(pr->path) - 1] = 0;
4388 
4389 		PF_RULES_RLOCK();
4390 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4391 			PF_RULES_RUNLOCK();
4392 			error = ENOENT;
4393 			break;
4394 		}
4395 		pr->name[0] = 0;
4396 		if (ruleset->anchor == NULL) {
4397 			/* XXX kludge for pf_main_ruleset */
4398 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4399 				if (anchor->parent == NULL && nr++ == pr->nr) {
4400 					strlcpy(pr->name, anchor->name,
4401 					    sizeof(pr->name));
4402 					break;
4403 				}
4404 		} else {
4405 			RB_FOREACH(anchor, pf_kanchor_node,
4406 			    &ruleset->anchor->children)
4407 				if (nr++ == pr->nr) {
4408 					strlcpy(pr->name, anchor->name,
4409 					    sizeof(pr->name));
4410 					break;
4411 				}
4412 		}
4413 		if (!pr->name[0])
4414 			error = EBUSY;
4415 		PF_RULES_RUNLOCK();
4416 		break;
4417 	}
4418 
4419 	case DIOCRCLRTABLES: {
4420 		struct pfioc_table *io = (struct pfioc_table *)addr;
4421 
4422 		if (io->pfrio_esize != 0) {
4423 			error = ENODEV;
4424 			break;
4425 		}
4426 		PF_RULES_WLOCK();
4427 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
4428 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4429 		PF_RULES_WUNLOCK();
4430 		break;
4431 	}
4432 
4433 	case DIOCRADDTABLES: {
4434 		struct pfioc_table *io = (struct pfioc_table *)addr;
4435 		struct pfr_table *pfrts;
4436 		size_t totlen;
4437 
4438 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4439 			error = ENODEV;
4440 			break;
4441 		}
4442 
4443 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4444 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4445 			error = ENOMEM;
4446 			break;
4447 		}
4448 
4449 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4450 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4451 		    M_TEMP, M_WAITOK);
4452 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4453 		if (error) {
4454 			free(pfrts, M_TEMP);
4455 			break;
4456 		}
4457 		PF_RULES_WLOCK();
4458 		error = pfr_add_tables(pfrts, io->pfrio_size,
4459 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4460 		PF_RULES_WUNLOCK();
4461 		free(pfrts, M_TEMP);
4462 		break;
4463 	}
4464 
4465 	case DIOCRDELTABLES: {
4466 		struct pfioc_table *io = (struct pfioc_table *)addr;
4467 		struct pfr_table *pfrts;
4468 		size_t totlen;
4469 
4470 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4471 			error = ENODEV;
4472 			break;
4473 		}
4474 
4475 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4476 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4477 			error = ENOMEM;
4478 			break;
4479 		}
4480 
4481 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4482 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4483 		    M_TEMP, M_WAITOK);
4484 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4485 		if (error) {
4486 			free(pfrts, M_TEMP);
4487 			break;
4488 		}
4489 		PF_RULES_WLOCK();
4490 		error = pfr_del_tables(pfrts, io->pfrio_size,
4491 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4492 		PF_RULES_WUNLOCK();
4493 		free(pfrts, M_TEMP);
4494 		break;
4495 	}
4496 
4497 	case DIOCRGETTABLES: {
4498 		struct pfioc_table *io = (struct pfioc_table *)addr;
4499 		struct pfr_table *pfrts;
4500 		size_t totlen;
4501 		int n;
4502 
4503 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4504 			error = ENODEV;
4505 			break;
4506 		}
4507 		PF_RULES_RLOCK();
4508 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4509 		if (n < 0) {
4510 			PF_RULES_RUNLOCK();
4511 			error = EINVAL;
4512 			break;
4513 		}
4514 		io->pfrio_size = min(io->pfrio_size, n);
4515 
4516 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4517 
4518 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4519 		    M_TEMP, M_NOWAIT | M_ZERO);
4520 		if (pfrts == NULL) {
4521 			error = ENOMEM;
4522 			PF_RULES_RUNLOCK();
4523 			break;
4524 		}
4525 		error = pfr_get_tables(&io->pfrio_table, pfrts,
4526 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4527 		PF_RULES_RUNLOCK();
4528 		if (error == 0)
4529 			error = copyout(pfrts, io->pfrio_buffer, totlen);
4530 		free(pfrts, M_TEMP);
4531 		break;
4532 	}
4533 
4534 	case DIOCRGETTSTATS: {
4535 		struct pfioc_table *io = (struct pfioc_table *)addr;
4536 		struct pfr_tstats *pfrtstats;
4537 		size_t totlen;
4538 		int n;
4539 
4540 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
4541 			error = ENODEV;
4542 			break;
4543 		}
4544 		PF_TABLE_STATS_LOCK();
4545 		PF_RULES_RLOCK();
4546 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4547 		if (n < 0) {
4548 			PF_RULES_RUNLOCK();
4549 			PF_TABLE_STATS_UNLOCK();
4550 			error = EINVAL;
4551 			break;
4552 		}
4553 		io->pfrio_size = min(io->pfrio_size, n);
4554 
4555 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
4556 		pfrtstats = mallocarray(io->pfrio_size,
4557 		    sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT | M_ZERO);
4558 		if (pfrtstats == NULL) {
4559 			error = ENOMEM;
4560 			PF_RULES_RUNLOCK();
4561 			PF_TABLE_STATS_UNLOCK();
4562 			break;
4563 		}
4564 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
4565 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4566 		PF_RULES_RUNLOCK();
4567 		PF_TABLE_STATS_UNLOCK();
4568 		if (error == 0)
4569 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
4570 		free(pfrtstats, M_TEMP);
4571 		break;
4572 	}
4573 
4574 	case DIOCRCLRTSTATS: {
4575 		struct pfioc_table *io = (struct pfioc_table *)addr;
4576 		struct pfr_table *pfrts;
4577 		size_t totlen;
4578 
4579 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4580 			error = ENODEV;
4581 			break;
4582 		}
4583 
4584 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4585 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4586 			/* We used to count tables and use the minimum required
4587 			 * size, so we didn't fail on overly large requests.
4588 			 * Keep doing so. */
4589 			io->pfrio_size = pf_ioctl_maxcount;
4590 			break;
4591 		}
4592 
4593 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4594 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4595 		    M_TEMP, M_WAITOK);
4596 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4597 		if (error) {
4598 			free(pfrts, M_TEMP);
4599 			break;
4600 		}
4601 
4602 		PF_TABLE_STATS_LOCK();
4603 		PF_RULES_RLOCK();
4604 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
4605 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4606 		PF_RULES_RUNLOCK();
4607 		PF_TABLE_STATS_UNLOCK();
4608 		free(pfrts, M_TEMP);
4609 		break;
4610 	}
4611 
4612 	case DIOCRSETTFLAGS: {
4613 		struct pfioc_table *io = (struct pfioc_table *)addr;
4614 		struct pfr_table *pfrts;
4615 		size_t totlen;
4616 		int n;
4617 
4618 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4619 			error = ENODEV;
4620 			break;
4621 		}
4622 
4623 		PF_RULES_RLOCK();
4624 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4625 		if (n < 0) {
4626 			PF_RULES_RUNLOCK();
4627 			error = EINVAL;
4628 			break;
4629 		}
4630 
4631 		io->pfrio_size = min(io->pfrio_size, n);
4632 		PF_RULES_RUNLOCK();
4633 
4634 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4635 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4636 		    M_TEMP, M_WAITOK);
4637 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4638 		if (error) {
4639 			free(pfrts, M_TEMP);
4640 			break;
4641 		}
4642 		PF_RULES_WLOCK();
4643 		error = pfr_set_tflags(pfrts, io->pfrio_size,
4644 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
4645 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4646 		PF_RULES_WUNLOCK();
4647 		free(pfrts, M_TEMP);
4648 		break;
4649 	}
4650 
4651 	case DIOCRCLRADDRS: {
4652 		struct pfioc_table *io = (struct pfioc_table *)addr;
4653 
4654 		if (io->pfrio_esize != 0) {
4655 			error = ENODEV;
4656 			break;
4657 		}
4658 		PF_RULES_WLOCK();
4659 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
4660 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4661 		PF_RULES_WUNLOCK();
4662 		break;
4663 	}
4664 
4665 	case DIOCRADDADDRS: {
4666 		struct pfioc_table *io = (struct pfioc_table *)addr;
4667 		struct pfr_addr *pfras;
4668 		size_t totlen;
4669 
4670 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4671 			error = ENODEV;
4672 			break;
4673 		}
4674 		if (io->pfrio_size < 0 ||
4675 		    io->pfrio_size > pf_ioctl_maxcount ||
4676 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4677 			error = EINVAL;
4678 			break;
4679 		}
4680 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4681 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4682 		    M_TEMP, M_WAITOK);
4683 		error = copyin(io->pfrio_buffer, pfras, totlen);
4684 		if (error) {
4685 			free(pfras, M_TEMP);
4686 			break;
4687 		}
4688 		PF_RULES_WLOCK();
4689 		error = pfr_add_addrs(&io->pfrio_table, pfras,
4690 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
4691 		    PFR_FLAG_USERIOCTL);
4692 		PF_RULES_WUNLOCK();
4693 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4694 			error = copyout(pfras, io->pfrio_buffer, totlen);
4695 		free(pfras, M_TEMP);
4696 		break;
4697 	}
4698 
4699 	case DIOCRDELADDRS: {
4700 		struct pfioc_table *io = (struct pfioc_table *)addr;
4701 		struct pfr_addr *pfras;
4702 		size_t totlen;
4703 
4704 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4705 			error = ENODEV;
4706 			break;
4707 		}
4708 		if (io->pfrio_size < 0 ||
4709 		    io->pfrio_size > pf_ioctl_maxcount ||
4710 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4711 			error = EINVAL;
4712 			break;
4713 		}
4714 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4715 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4716 		    M_TEMP, M_WAITOK);
4717 		error = copyin(io->pfrio_buffer, pfras, totlen);
4718 		if (error) {
4719 			free(pfras, M_TEMP);
4720 			break;
4721 		}
4722 		PF_RULES_WLOCK();
4723 		error = pfr_del_addrs(&io->pfrio_table, pfras,
4724 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
4725 		    PFR_FLAG_USERIOCTL);
4726 		PF_RULES_WUNLOCK();
4727 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4728 			error = copyout(pfras, io->pfrio_buffer, totlen);
4729 		free(pfras, M_TEMP);
4730 		break;
4731 	}
4732 
4733 	case DIOCRSETADDRS: {
4734 		struct pfioc_table *io = (struct pfioc_table *)addr;
4735 		struct pfr_addr *pfras;
4736 		size_t totlen, count;
4737 
4738 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4739 			error = ENODEV;
4740 			break;
4741 		}
4742 		if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
4743 			error = EINVAL;
4744 			break;
4745 		}
4746 		count = max(io->pfrio_size, io->pfrio_size2);
4747 		if (count > pf_ioctl_maxcount ||
4748 		    WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
4749 			error = EINVAL;
4750 			break;
4751 		}
4752 		totlen = count * sizeof(struct pfr_addr);
4753 		pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
4754 		    M_WAITOK);
4755 		error = copyin(io->pfrio_buffer, pfras, totlen);
4756 		if (error) {
4757 			free(pfras, M_TEMP);
4758 			break;
4759 		}
4760 		PF_RULES_WLOCK();
4761 		error = pfr_set_addrs(&io->pfrio_table, pfras,
4762 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
4763 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
4764 		    PFR_FLAG_USERIOCTL, 0);
4765 		PF_RULES_WUNLOCK();
4766 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4767 			error = copyout(pfras, io->pfrio_buffer, totlen);
4768 		free(pfras, M_TEMP);
4769 		break;
4770 	}
4771 
4772 	case DIOCRGETADDRS: {
4773 		struct pfioc_table *io = (struct pfioc_table *)addr;
4774 		struct pfr_addr *pfras;
4775 		size_t totlen;
4776 
4777 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4778 			error = ENODEV;
4779 			break;
4780 		}
4781 		if (io->pfrio_size < 0 ||
4782 		    io->pfrio_size > pf_ioctl_maxcount ||
4783 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4784 			error = EINVAL;
4785 			break;
4786 		}
4787 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4788 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4789 		    M_TEMP, M_WAITOK | M_ZERO);
4790 		PF_RULES_RLOCK();
4791 		error = pfr_get_addrs(&io->pfrio_table, pfras,
4792 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4793 		PF_RULES_RUNLOCK();
4794 		if (error == 0)
4795 			error = copyout(pfras, io->pfrio_buffer, totlen);
4796 		free(pfras, M_TEMP);
4797 		break;
4798 	}
4799 
4800 	case DIOCRGETASTATS: {
4801 		struct pfioc_table *io = (struct pfioc_table *)addr;
4802 		struct pfr_astats *pfrastats;
4803 		size_t totlen;
4804 
4805 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
4806 			error = ENODEV;
4807 			break;
4808 		}
4809 		if (io->pfrio_size < 0 ||
4810 		    io->pfrio_size > pf_ioctl_maxcount ||
4811 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
4812 			error = EINVAL;
4813 			break;
4814 		}
4815 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
4816 		pfrastats = mallocarray(io->pfrio_size,
4817 		    sizeof(struct pfr_astats), M_TEMP, M_WAITOK | M_ZERO);
4818 		PF_RULES_RLOCK();
4819 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
4820 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4821 		PF_RULES_RUNLOCK();
4822 		if (error == 0)
4823 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
4824 		free(pfrastats, M_TEMP);
4825 		break;
4826 	}
4827 
4828 	case DIOCRCLRASTATS: {
4829 		struct pfioc_table *io = (struct pfioc_table *)addr;
4830 		struct pfr_addr *pfras;
4831 		size_t totlen;
4832 
4833 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4834 			error = ENODEV;
4835 			break;
4836 		}
4837 		if (io->pfrio_size < 0 ||
4838 		    io->pfrio_size > pf_ioctl_maxcount ||
4839 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4840 			error = EINVAL;
4841 			break;
4842 		}
4843 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4844 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4845 		    M_TEMP, M_WAITOK);
4846 		error = copyin(io->pfrio_buffer, pfras, totlen);
4847 		if (error) {
4848 			free(pfras, M_TEMP);
4849 			break;
4850 		}
4851 		PF_RULES_WLOCK();
4852 		error = pfr_clr_astats(&io->pfrio_table, pfras,
4853 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
4854 		    PFR_FLAG_USERIOCTL);
4855 		PF_RULES_WUNLOCK();
4856 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4857 			error = copyout(pfras, io->pfrio_buffer, totlen);
4858 		free(pfras, M_TEMP);
4859 		break;
4860 	}
4861 
4862 	case DIOCRTSTADDRS: {
4863 		struct pfioc_table *io = (struct pfioc_table *)addr;
4864 		struct pfr_addr *pfras;
4865 		size_t totlen;
4866 
4867 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4868 			error = ENODEV;
4869 			break;
4870 		}
4871 		if (io->pfrio_size < 0 ||
4872 		    io->pfrio_size > pf_ioctl_maxcount ||
4873 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4874 			error = EINVAL;
4875 			break;
4876 		}
4877 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4878 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4879 		    M_TEMP, M_WAITOK);
4880 		error = copyin(io->pfrio_buffer, pfras, totlen);
4881 		if (error) {
4882 			free(pfras, M_TEMP);
4883 			break;
4884 		}
4885 		PF_RULES_RLOCK();
4886 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
4887 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
4888 		    PFR_FLAG_USERIOCTL);
4889 		PF_RULES_RUNLOCK();
4890 		if (error == 0)
4891 			error = copyout(pfras, io->pfrio_buffer, totlen);
4892 		free(pfras, M_TEMP);
4893 		break;
4894 	}
4895 
4896 	case DIOCRINADEFINE: {
4897 		struct pfioc_table *io = (struct pfioc_table *)addr;
4898 		struct pfr_addr *pfras;
4899 		size_t totlen;
4900 
4901 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4902 			error = ENODEV;
4903 			break;
4904 		}
4905 		if (io->pfrio_size < 0 ||
4906 		    io->pfrio_size > pf_ioctl_maxcount ||
4907 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4908 			error = EINVAL;
4909 			break;
4910 		}
4911 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4912 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4913 		    M_TEMP, M_WAITOK);
4914 		error = copyin(io->pfrio_buffer, pfras, totlen);
4915 		if (error) {
4916 			free(pfras, M_TEMP);
4917 			break;
4918 		}
4919 		PF_RULES_WLOCK();
4920 		error = pfr_ina_define(&io->pfrio_table, pfras,
4921 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
4922 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4923 		PF_RULES_WUNLOCK();
4924 		free(pfras, M_TEMP);
4925 		break;
4926 	}
4927 
4928 	case DIOCOSFPADD: {
4929 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
4930 		PF_RULES_WLOCK();
4931 		error = pf_osfp_add(io);
4932 		PF_RULES_WUNLOCK();
4933 		break;
4934 	}
4935 
4936 	case DIOCOSFPGET: {
4937 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
4938 		PF_RULES_RLOCK();
4939 		error = pf_osfp_get(io);
4940 		PF_RULES_RUNLOCK();
4941 		break;
4942 	}
4943 
4944 	case DIOCXBEGIN: {
4945 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
4946 		struct pfioc_trans_e	*ioes, *ioe;
4947 		size_t			 totlen;
4948 		int			 i;
4949 
4950 		if (io->esize != sizeof(*ioe)) {
4951 			error = ENODEV;
4952 			break;
4953 		}
4954 		if (io->size < 0 ||
4955 		    io->size > pf_ioctl_maxcount ||
4956 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
4957 			error = EINVAL;
4958 			break;
4959 		}
4960 		totlen = sizeof(struct pfioc_trans_e) * io->size;
4961 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
4962 		    M_TEMP, M_WAITOK);
4963 		error = copyin(io->array, ioes, totlen);
4964 		if (error) {
4965 			free(ioes, M_TEMP);
4966 			break;
4967 		}
4968 		/* Ensure there's no more ethernet rules to clean up. */
4969 		NET_EPOCH_DRAIN_CALLBACKS();
4970 		PF_RULES_WLOCK();
4971 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
4972 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
4973 			switch (ioe->rs_num) {
4974 			case PF_RULESET_ETH:
4975 				if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) {
4976 					PF_RULES_WUNLOCK();
4977 					free(ioes, M_TEMP);
4978 					goto fail;
4979 				}
4980 				break;
4981 #ifdef ALTQ
4982 			case PF_RULESET_ALTQ:
4983 				if (ioe->anchor[0]) {
4984 					PF_RULES_WUNLOCK();
4985 					free(ioes, M_TEMP);
4986 					error = EINVAL;
4987 					goto fail;
4988 				}
4989 				if ((error = pf_begin_altq(&ioe->ticket))) {
4990 					PF_RULES_WUNLOCK();
4991 					free(ioes, M_TEMP);
4992 					goto fail;
4993 				}
4994 				break;
4995 #endif /* ALTQ */
4996 			case PF_RULESET_TABLE:
4997 			    {
4998 				struct pfr_table table;
4999 
5000 				bzero(&table, sizeof(table));
5001 				strlcpy(table.pfrt_anchor, ioe->anchor,
5002 				    sizeof(table.pfrt_anchor));
5003 				if ((error = pfr_ina_begin(&table,
5004 				    &ioe->ticket, NULL, 0))) {
5005 					PF_RULES_WUNLOCK();
5006 					free(ioes, M_TEMP);
5007 					goto fail;
5008 				}
5009 				break;
5010 			    }
5011 			default:
5012 				if ((error = pf_begin_rules(&ioe->ticket,
5013 				    ioe->rs_num, ioe->anchor))) {
5014 					PF_RULES_WUNLOCK();
5015 					free(ioes, M_TEMP);
5016 					goto fail;
5017 				}
5018 				break;
5019 			}
5020 		}
5021 		PF_RULES_WUNLOCK();
5022 		error = copyout(ioes, io->array, totlen);
5023 		free(ioes, M_TEMP);
5024 		break;
5025 	}
5026 
5027 	case DIOCXROLLBACK: {
5028 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5029 		struct pfioc_trans_e	*ioe, *ioes;
5030 		size_t			 totlen;
5031 		int			 i;
5032 
5033 		if (io->esize != sizeof(*ioe)) {
5034 			error = ENODEV;
5035 			break;
5036 		}
5037 		if (io->size < 0 ||
5038 		    io->size > pf_ioctl_maxcount ||
5039 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5040 			error = EINVAL;
5041 			break;
5042 		}
5043 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5044 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5045 		    M_TEMP, M_WAITOK);
5046 		error = copyin(io->array, ioes, totlen);
5047 		if (error) {
5048 			free(ioes, M_TEMP);
5049 			break;
5050 		}
5051 		PF_RULES_WLOCK();
5052 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5053 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5054 			switch (ioe->rs_num) {
5055 			case PF_RULESET_ETH:
5056 				if ((error = pf_rollback_eth(ioe->ticket,
5057 				    ioe->anchor))) {
5058 					PF_RULES_WUNLOCK();
5059 					free(ioes, M_TEMP);
5060 					goto fail; /* really bad */
5061 				}
5062 				break;
5063 #ifdef ALTQ
5064 			case PF_RULESET_ALTQ:
5065 				if (ioe->anchor[0]) {
5066 					PF_RULES_WUNLOCK();
5067 					free(ioes, M_TEMP);
5068 					error = EINVAL;
5069 					goto fail;
5070 				}
5071 				if ((error = pf_rollback_altq(ioe->ticket))) {
5072 					PF_RULES_WUNLOCK();
5073 					free(ioes, M_TEMP);
5074 					goto fail; /* really bad */
5075 				}
5076 				break;
5077 #endif /* ALTQ */
5078 			case PF_RULESET_TABLE:
5079 			    {
5080 				struct pfr_table table;
5081 
5082 				bzero(&table, sizeof(table));
5083 				strlcpy(table.pfrt_anchor, ioe->anchor,
5084 				    sizeof(table.pfrt_anchor));
5085 				if ((error = pfr_ina_rollback(&table,
5086 				    ioe->ticket, NULL, 0))) {
5087 					PF_RULES_WUNLOCK();
5088 					free(ioes, M_TEMP);
5089 					goto fail; /* really bad */
5090 				}
5091 				break;
5092 			    }
5093 			default:
5094 				if ((error = pf_rollback_rules(ioe->ticket,
5095 				    ioe->rs_num, ioe->anchor))) {
5096 					PF_RULES_WUNLOCK();
5097 					free(ioes, M_TEMP);
5098 					goto fail; /* really bad */
5099 				}
5100 				break;
5101 			}
5102 		}
5103 		PF_RULES_WUNLOCK();
5104 		free(ioes, M_TEMP);
5105 		break;
5106 	}
5107 
5108 	case DIOCXCOMMIT: {
5109 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5110 		struct pfioc_trans_e	*ioe, *ioes;
5111 		struct pf_kruleset	*rs;
5112 		struct pf_keth_ruleset	*ers;
5113 		size_t			 totlen;
5114 		int			 i;
5115 
5116 		if (io->esize != sizeof(*ioe)) {
5117 			error = ENODEV;
5118 			break;
5119 		}
5120 
5121 		if (io->size < 0 ||
5122 		    io->size > pf_ioctl_maxcount ||
5123 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5124 			error = EINVAL;
5125 			break;
5126 		}
5127 
5128 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5129 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5130 		    M_TEMP, M_WAITOK);
5131 		error = copyin(io->array, ioes, totlen);
5132 		if (error) {
5133 			free(ioes, M_TEMP);
5134 			break;
5135 		}
5136 		PF_RULES_WLOCK();
5137 		/* First makes sure everything will succeed. */
5138 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5139 			ioe->anchor[sizeof(ioe->anchor) - 1] = 0;
5140 			switch (ioe->rs_num) {
5141 			case PF_RULESET_ETH:
5142 				ers = pf_find_keth_ruleset(ioe->anchor);
5143 				if (ers == NULL || ioe->ticket == 0 ||
5144 				    ioe->ticket != ers->inactive.ticket) {
5145 					PF_RULES_WUNLOCK();
5146 					free(ioes, M_TEMP);
5147 					error = EINVAL;
5148 					goto fail;
5149 				}
5150 				break;
5151 #ifdef ALTQ
5152 			case PF_RULESET_ALTQ:
5153 				if (ioe->anchor[0]) {
5154 					PF_RULES_WUNLOCK();
5155 					free(ioes, M_TEMP);
5156 					error = EINVAL;
5157 					goto fail;
5158 				}
5159 				if (!V_altqs_inactive_open || ioe->ticket !=
5160 				    V_ticket_altqs_inactive) {
5161 					PF_RULES_WUNLOCK();
5162 					free(ioes, M_TEMP);
5163 					error = EBUSY;
5164 					goto fail;
5165 				}
5166 				break;
5167 #endif /* ALTQ */
5168 			case PF_RULESET_TABLE:
5169 				rs = pf_find_kruleset(ioe->anchor);
5170 				if (rs == NULL || !rs->topen || ioe->ticket !=
5171 				    rs->tticket) {
5172 					PF_RULES_WUNLOCK();
5173 					free(ioes, M_TEMP);
5174 					error = EBUSY;
5175 					goto fail;
5176 				}
5177 				break;
5178 			default:
5179 				if (ioe->rs_num < 0 || ioe->rs_num >=
5180 				    PF_RULESET_MAX) {
5181 					PF_RULES_WUNLOCK();
5182 					free(ioes, M_TEMP);
5183 					error = EINVAL;
5184 					goto fail;
5185 				}
5186 				rs = pf_find_kruleset(ioe->anchor);
5187 				if (rs == NULL ||
5188 				    !rs->rules[ioe->rs_num].inactive.open ||
5189 				    rs->rules[ioe->rs_num].inactive.ticket !=
5190 				    ioe->ticket) {
5191 					PF_RULES_WUNLOCK();
5192 					free(ioes, M_TEMP);
5193 					error = EBUSY;
5194 					goto fail;
5195 				}
5196 				break;
5197 			}
5198 		}
5199 		/* Now do the commit - no errors should happen here. */
5200 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5201 			switch (ioe->rs_num) {
5202 			case PF_RULESET_ETH:
5203 				if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) {
5204 					PF_RULES_WUNLOCK();
5205 					free(ioes, M_TEMP);
5206 					goto fail; /* really bad */
5207 				}
5208 				break;
5209 #ifdef ALTQ
5210 			case PF_RULESET_ALTQ:
5211 				if ((error = pf_commit_altq(ioe->ticket))) {
5212 					PF_RULES_WUNLOCK();
5213 					free(ioes, M_TEMP);
5214 					goto fail; /* really bad */
5215 				}
5216 				break;
5217 #endif /* ALTQ */
5218 			case PF_RULESET_TABLE:
5219 			    {
5220 				struct pfr_table table;
5221 
5222 				bzero(&table, sizeof(table));
5223 				(void)strlcpy(table.pfrt_anchor, ioe->anchor,
5224 				    sizeof(table.pfrt_anchor));
5225 				if ((error = pfr_ina_commit(&table,
5226 				    ioe->ticket, NULL, NULL, 0))) {
5227 					PF_RULES_WUNLOCK();
5228 					free(ioes, M_TEMP);
5229 					goto fail; /* really bad */
5230 				}
5231 				break;
5232 			    }
5233 			default:
5234 				if ((error = pf_commit_rules(ioe->ticket,
5235 				    ioe->rs_num, ioe->anchor))) {
5236 					PF_RULES_WUNLOCK();
5237 					free(ioes, M_TEMP);
5238 					goto fail; /* really bad */
5239 				}
5240 				break;
5241 			}
5242 		}
5243 		PF_RULES_WUNLOCK();
5244 
5245 		/* Only hook into EtherNet taffic if we've got rules for it. */
5246 		if (! TAILQ_EMPTY(V_pf_keth->active.rules))
5247 			hook_pf_eth();
5248 		else
5249 			dehook_pf_eth();
5250 
5251 		free(ioes, M_TEMP);
5252 		break;
5253 	}
5254 
5255 	case DIOCGETSRCNODES: {
5256 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
5257 		struct pf_srchash	*sh;
5258 		struct pf_ksrc_node	*n;
5259 		struct pf_src_node	*p, *pstore;
5260 		uint32_t		 i, nr = 0;
5261 
5262 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5263 				i++, sh++) {
5264 			PF_HASHROW_LOCK(sh);
5265 			LIST_FOREACH(n, &sh->nodes, entry)
5266 				nr++;
5267 			PF_HASHROW_UNLOCK(sh);
5268 		}
5269 
5270 		psn->psn_len = min(psn->psn_len,
5271 		    sizeof(struct pf_src_node) * nr);
5272 
5273 		if (psn->psn_len == 0) {
5274 			psn->psn_len = sizeof(struct pf_src_node) * nr;
5275 			break;
5276 		}
5277 
5278 		nr = 0;
5279 
5280 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
5281 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5282 		    i++, sh++) {
5283 		    PF_HASHROW_LOCK(sh);
5284 		    LIST_FOREACH(n, &sh->nodes, entry) {
5285 
5286 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
5287 				break;
5288 
5289 			pf_src_node_copy(n, p);
5290 
5291 			p++;
5292 			nr++;
5293 		    }
5294 		    PF_HASHROW_UNLOCK(sh);
5295 		}
5296 		error = copyout(pstore, psn->psn_src_nodes,
5297 		    sizeof(struct pf_src_node) * nr);
5298 		if (error) {
5299 			free(pstore, M_TEMP);
5300 			break;
5301 		}
5302 		psn->psn_len = sizeof(struct pf_src_node) * nr;
5303 		free(pstore, M_TEMP);
5304 		break;
5305 	}
5306 
5307 	case DIOCCLRSRCNODES: {
5308 		pf_clear_srcnodes(NULL);
5309 		pf_purge_expired_src_nodes();
5310 		break;
5311 	}
5312 
5313 	case DIOCKILLSRCNODES:
5314 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
5315 		break;
5316 
5317 #ifdef COMPAT_FREEBSD13
5318 	case DIOCKEEPCOUNTERS_FREEBSD13:
5319 #endif
5320 	case DIOCKEEPCOUNTERS:
5321 		error = pf_keepcounters((struct pfioc_nv *)addr);
5322 		break;
5323 
5324 	case DIOCGETSYNCOOKIES:
5325 		error = pf_get_syncookies((struct pfioc_nv *)addr);
5326 		break;
5327 
5328 	case DIOCSETSYNCOOKIES:
5329 		error = pf_set_syncookies((struct pfioc_nv *)addr);
5330 		break;
5331 
5332 	case DIOCSETHOSTID: {
5333 		u_int32_t	*hostid = (u_int32_t *)addr;
5334 
5335 		PF_RULES_WLOCK();
5336 		if (*hostid == 0)
5337 			V_pf_status.hostid = arc4random();
5338 		else
5339 			V_pf_status.hostid = *hostid;
5340 		PF_RULES_WUNLOCK();
5341 		break;
5342 	}
5343 
5344 	case DIOCOSFPFLUSH:
5345 		PF_RULES_WLOCK();
5346 		pf_osfp_flush();
5347 		PF_RULES_WUNLOCK();
5348 		break;
5349 
5350 	case DIOCIGETIFACES: {
5351 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5352 		struct pfi_kif *ifstore;
5353 		size_t bufsiz;
5354 
5355 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
5356 			error = ENODEV;
5357 			break;
5358 		}
5359 
5360 		if (io->pfiio_size < 0 ||
5361 		    io->pfiio_size > pf_ioctl_maxcount ||
5362 		    WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
5363 			error = EINVAL;
5364 			break;
5365 		}
5366 
5367 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5368 
5369 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
5370 		ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
5371 		    M_TEMP, M_WAITOK | M_ZERO);
5372 
5373 		PF_RULES_RLOCK();
5374 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
5375 		PF_RULES_RUNLOCK();
5376 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
5377 		free(ifstore, M_TEMP);
5378 		break;
5379 	}
5380 
5381 	case DIOCSETIFFLAG: {
5382 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5383 
5384 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5385 
5386 		PF_RULES_WLOCK();
5387 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
5388 		PF_RULES_WUNLOCK();
5389 		break;
5390 	}
5391 
5392 	case DIOCCLRIFFLAG: {
5393 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5394 
5395 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5396 
5397 		PF_RULES_WLOCK();
5398 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
5399 		PF_RULES_WUNLOCK();
5400 		break;
5401 	}
5402 
5403 	case DIOCSETREASS: {
5404 		u_int32_t	*reass = (u_int32_t *)addr;
5405 
5406 		V_pf_status.reass = *reass & (PF_REASS_ENABLED|PF_REASS_NODF);
5407 		/* Removal of DF flag without reassembly enabled is not a
5408 		 * valid combination. Disable reassembly in such case. */
5409 		if (!(V_pf_status.reass & PF_REASS_ENABLED))
5410 			V_pf_status.reass = 0;
5411 		break;
5412 	}
5413 
5414 	default:
5415 		error = ENODEV;
5416 		break;
5417 	}
5418 fail:
5419 	if (sx_xlocked(&V_pf_ioctl_lock))
5420 		sx_xunlock(&V_pf_ioctl_lock);
5421 	CURVNET_RESTORE();
5422 
5423 #undef ERROUT_IOCTL
5424 
5425 	return (error);
5426 }
5427 
5428 void
5429 pfsync_state_export(union pfsync_state_union *sp, struct pf_kstate *st, int msg_version)
5430 {
5431 	bzero(sp, sizeof(union pfsync_state_union));
5432 
5433 	/* copy from state key */
5434 	sp->pfs_1301.key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5435 	sp->pfs_1301.key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5436 	sp->pfs_1301.key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5437 	sp->pfs_1301.key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5438 	sp->pfs_1301.key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5439 	sp->pfs_1301.key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5440 	sp->pfs_1301.key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5441 	sp->pfs_1301.key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5442 	sp->pfs_1301.proto = st->key[PF_SK_WIRE]->proto;
5443 	sp->pfs_1301.af = st->key[PF_SK_WIRE]->af;
5444 
5445 	/* copy from state */
5446 	strlcpy(sp->pfs_1301.ifname, st->kif->pfik_name, sizeof(sp->pfs_1301.ifname));
5447 	bcopy(&st->rt_addr, &sp->pfs_1301.rt_addr, sizeof(sp->pfs_1301.rt_addr));
5448 	sp->pfs_1301.creation = htonl(time_uptime - st->creation);
5449 	sp->pfs_1301.expire = pf_state_expires(st);
5450 	if (sp->pfs_1301.expire <= time_uptime)
5451 		sp->pfs_1301.expire = htonl(0);
5452 	else
5453 		sp->pfs_1301.expire = htonl(sp->pfs_1301.expire - time_uptime);
5454 
5455 	sp->pfs_1301.direction = st->direction;
5456 	sp->pfs_1301.log = st->act.log;
5457 	sp->pfs_1301.timeout = st->timeout;
5458 
5459 	switch (msg_version) {
5460 		case PFSYNC_MSG_VERSION_1301:
5461 			sp->pfs_1301.state_flags = st->state_flags;
5462 			break;
5463 		case PFSYNC_MSG_VERSION_1400:
5464 			sp->pfs_1400.state_flags = htons(st->state_flags);
5465 			sp->pfs_1400.qid = htons(st->act.qid);
5466 			sp->pfs_1400.pqid = htons(st->act.pqid);
5467 			sp->pfs_1400.dnpipe = htons(st->act.dnpipe);
5468 			sp->pfs_1400.dnrpipe = htons(st->act.dnrpipe);
5469 			sp->pfs_1400.rtableid = htonl(st->act.rtableid);
5470 			sp->pfs_1400.min_ttl = st->act.min_ttl;
5471 			sp->pfs_1400.set_tos = st->act.set_tos;
5472 			sp->pfs_1400.max_mss = htons(st->act.max_mss);
5473 			sp->pfs_1400.set_prio[0] = st->act.set_prio[0];
5474 			sp->pfs_1400.set_prio[1] = st->act.set_prio[1];
5475 			sp->pfs_1400.rt = st->rt;
5476 			if (st->rt_kif)
5477 				strlcpy(sp->pfs_1400.rt_ifname,
5478 				    st->rt_kif->pfik_name,
5479 				    sizeof(sp->pfs_1400.rt_ifname));
5480 			break;
5481 		default:
5482 			panic("%s: Unsupported pfsync_msg_version %d",
5483 			    __func__, msg_version);
5484 	}
5485 
5486 	if (st->src_node)
5487 		sp->pfs_1301.sync_flags |= PFSYNC_FLAG_SRCNODE;
5488 	if (st->nat_src_node)
5489 		sp->pfs_1301.sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5490 
5491 	sp->pfs_1301.id = st->id;
5492 	sp->pfs_1301.creatorid = st->creatorid;
5493 	pf_state_peer_hton(&st->src, &sp->pfs_1301.src);
5494 	pf_state_peer_hton(&st->dst, &sp->pfs_1301.dst);
5495 
5496 	if (st->rule.ptr == NULL)
5497 		sp->pfs_1301.rule = htonl(-1);
5498 	else
5499 		sp->pfs_1301.rule = htonl(st->rule.ptr->nr);
5500 	if (st->anchor.ptr == NULL)
5501 		sp->pfs_1301.anchor = htonl(-1);
5502 	else
5503 		sp->pfs_1301.anchor = htonl(st->anchor.ptr->nr);
5504 	if (st->nat_rule.ptr == NULL)
5505 		sp->pfs_1301.nat_rule = htonl(-1);
5506 	else
5507 		sp->pfs_1301.nat_rule = htonl(st->nat_rule.ptr->nr);
5508 
5509 	pf_state_counter_hton(st->packets[0], sp->pfs_1301.packets[0]);
5510 	pf_state_counter_hton(st->packets[1], sp->pfs_1301.packets[1]);
5511 	pf_state_counter_hton(st->bytes[0], sp->pfs_1301.bytes[0]);
5512 	pf_state_counter_hton(st->bytes[1], sp->pfs_1301.bytes[1]);
5513 }
5514 
5515 void
5516 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st)
5517 {
5518 	bzero(sp, sizeof(*sp));
5519 
5520 	sp->version = PF_STATE_VERSION;
5521 
5522 	/* copy from state key */
5523 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5524 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5525 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5526 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5527 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5528 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5529 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5530 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5531 	sp->proto = st->key[PF_SK_WIRE]->proto;
5532 	sp->af = st->key[PF_SK_WIRE]->af;
5533 
5534 	/* copy from state */
5535 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
5536 	strlcpy(sp->orig_ifname, st->orig_kif->pfik_name,
5537 	    sizeof(sp->orig_ifname));
5538 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
5539 	sp->creation = htonl(time_uptime - st->creation);
5540 	sp->expire = pf_state_expires(st);
5541 	if (sp->expire <= time_uptime)
5542 		sp->expire = htonl(0);
5543 	else
5544 		sp->expire = htonl(sp->expire - time_uptime);
5545 
5546 	sp->direction = st->direction;
5547 	sp->log = st->act.log;
5548 	sp->timeout = st->timeout;
5549 	/* 8 bits for the old libpfctl, 16 bits for the new libpfctl */
5550 	sp->state_flags_compat = st->state_flags;
5551 	sp->state_flags = htons(st->state_flags);
5552 	if (st->src_node)
5553 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
5554 	if (st->nat_src_node)
5555 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5556 
5557 	sp->id = st->id;
5558 	sp->creatorid = st->creatorid;
5559 	pf_state_peer_hton(&st->src, &sp->src);
5560 	pf_state_peer_hton(&st->dst, &sp->dst);
5561 
5562 	if (st->rule.ptr == NULL)
5563 		sp->rule = htonl(-1);
5564 	else
5565 		sp->rule = htonl(st->rule.ptr->nr);
5566 	if (st->anchor.ptr == NULL)
5567 		sp->anchor = htonl(-1);
5568 	else
5569 		sp->anchor = htonl(st->anchor.ptr->nr);
5570 	if (st->nat_rule.ptr == NULL)
5571 		sp->nat_rule = htonl(-1);
5572 	else
5573 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
5574 
5575 	sp->packets[0] = st->packets[0];
5576 	sp->packets[1] = st->packets[1];
5577 	sp->bytes[0] = st->bytes[0];
5578 	sp->bytes[1] = st->bytes[1];
5579 
5580 	sp->qid = htons(st->act.qid);
5581 	sp->pqid = htons(st->act.pqid);
5582 	sp->dnpipe = htons(st->act.dnpipe);
5583 	sp->dnrpipe = htons(st->act.dnrpipe);
5584 	sp->rtableid = htonl(st->act.rtableid);
5585 	sp->min_ttl = st->act.min_ttl;
5586 	sp->set_tos = st->act.set_tos;
5587 	sp->max_mss = htons(st->act.max_mss);
5588 	sp->rt = st->rt;
5589 	if (st->rt_kif)
5590 		strlcpy(sp->rt_ifname, st->rt_kif->pfik_name,
5591 		    sizeof(sp->rt_ifname));
5592 	sp->set_prio[0] = st->act.set_prio[0];
5593 	sp->set_prio[1] = st->act.set_prio[1];
5594 
5595 }
5596 
5597 static void
5598 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
5599 {
5600 	struct pfr_ktable *kt;
5601 
5602 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
5603 
5604 	kt = aw->p.tbl;
5605 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
5606 		kt = kt->pfrkt_root;
5607 	aw->p.tbl = NULL;
5608 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
5609 		kt->pfrkt_cnt : -1;
5610 }
5611 
5612 static int
5613 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters,
5614     size_t number, char **names)
5615 {
5616 	nvlist_t        *nvc;
5617 
5618 	nvc = nvlist_create(0);
5619 	if (nvc == NULL)
5620 		return (ENOMEM);
5621 
5622 	for (int i = 0; i < number; i++) {
5623 		nvlist_append_number_array(nvc, "counters",
5624 		    counter_u64_fetch(counters[i]));
5625 		nvlist_append_string_array(nvc, "names",
5626 		    names[i]);
5627 		nvlist_append_number_array(nvc, "ids",
5628 		    i);
5629 	}
5630 	nvlist_add_nvlist(nvl, name, nvc);
5631 	nvlist_destroy(nvc);
5632 
5633 	return (0);
5634 }
5635 
5636 static int
5637 pf_getstatus(struct pfioc_nv *nv)
5638 {
5639 	nvlist_t        *nvl = NULL, *nvc = NULL;
5640 	void            *nvlpacked = NULL;
5641 	int              error;
5642 	struct pf_status s;
5643 	char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES;
5644 	char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES;
5645 	char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES;
5646 	PF_RULES_RLOCK_TRACKER;
5647 
5648 #define ERROUT(x)      ERROUT_FUNCTION(errout, x)
5649 
5650 	PF_RULES_RLOCK();
5651 
5652 	nvl = nvlist_create(0);
5653 	if (nvl == NULL)
5654 		ERROUT(ENOMEM);
5655 
5656 	nvlist_add_bool(nvl, "running", V_pf_status.running);
5657 	nvlist_add_number(nvl, "since", V_pf_status.since);
5658 	nvlist_add_number(nvl, "debug", V_pf_status.debug);
5659 	nvlist_add_number(nvl, "hostid", V_pf_status.hostid);
5660 	nvlist_add_number(nvl, "states", V_pf_status.states);
5661 	nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes);
5662 	nvlist_add_number(nvl, "reass", V_pf_status.reass);
5663 	nvlist_add_bool(nvl, "syncookies_active",
5664 	    V_pf_status.syncookies_active);
5665 
5666 	/* counters */
5667 	error = pf_add_status_counters(nvl, "counters", V_pf_status.counters,
5668 	    PFRES_MAX, pf_reasons);
5669 	if (error != 0)
5670 		ERROUT(error);
5671 
5672 	/* lcounters */
5673 	error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters,
5674 	    KLCNT_MAX, pf_lcounter);
5675 	if (error != 0)
5676 		ERROUT(error);
5677 
5678 	/* fcounters */
5679 	nvc = nvlist_create(0);
5680 	if (nvc == NULL)
5681 		ERROUT(ENOMEM);
5682 
5683 	for (int i = 0; i < FCNT_MAX; i++) {
5684 		nvlist_append_number_array(nvc, "counters",
5685 		    pf_counter_u64_fetch(&V_pf_status.fcounters[i]));
5686 		nvlist_append_string_array(nvc, "names",
5687 		    pf_fcounter[i]);
5688 		nvlist_append_number_array(nvc, "ids",
5689 		    i);
5690 	}
5691 	nvlist_add_nvlist(nvl, "fcounters", nvc);
5692 	nvlist_destroy(nvc);
5693 	nvc = NULL;
5694 
5695 	/* scounters */
5696 	error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters,
5697 	    SCNT_MAX, pf_fcounter);
5698 	if (error != 0)
5699 		ERROUT(error);
5700 
5701 	nvlist_add_string(nvl, "ifname", V_pf_status.ifname);
5702 	nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum,
5703 	    PF_MD5_DIGEST_LENGTH);
5704 
5705 	pfi_update_status(V_pf_status.ifname, &s);
5706 
5707 	/* pcounters / bcounters */
5708 	for (int i = 0; i < 2; i++) {
5709 		for (int j = 0; j < 2; j++) {
5710 			for (int k = 0; k < 2; k++) {
5711 				nvlist_append_number_array(nvl, "pcounters",
5712 				    s.pcounters[i][j][k]);
5713 			}
5714 			nvlist_append_number_array(nvl, "bcounters",
5715 			    s.bcounters[i][j]);
5716 		}
5717 	}
5718 
5719 	nvlpacked = nvlist_pack(nvl, &nv->len);
5720 	if (nvlpacked == NULL)
5721 		ERROUT(ENOMEM);
5722 
5723 	if (nv->size == 0)
5724 		ERROUT(0);
5725 	else if (nv->size < nv->len)
5726 		ERROUT(ENOSPC);
5727 
5728 	PF_RULES_RUNLOCK();
5729 	error = copyout(nvlpacked, nv->data, nv->len);
5730 	goto done;
5731 
5732 #undef ERROUT
5733 errout:
5734 	PF_RULES_RUNLOCK();
5735 done:
5736 	free(nvlpacked, M_NVLIST);
5737 	nvlist_destroy(nvc);
5738 	nvlist_destroy(nvl);
5739 
5740 	return (error);
5741 }
5742 
5743 /*
5744  * XXX - Check for version mismatch!!!
5745  */
5746 static void
5747 pf_clear_all_states(void)
5748 {
5749 	struct pf_kstate	*s;
5750 	u_int i;
5751 
5752 	for (i = 0; i <= pf_hashmask; i++) {
5753 		struct pf_idhash *ih = &V_pf_idhash[i];
5754 relock:
5755 		PF_HASHROW_LOCK(ih);
5756 		LIST_FOREACH(s, &ih->states, entry) {
5757 			s->timeout = PFTM_PURGE;
5758 			/* Don't send out individual delete messages. */
5759 			s->state_flags |= PFSTATE_NOSYNC;
5760 			pf_unlink_state(s);
5761 			goto relock;
5762 		}
5763 		PF_HASHROW_UNLOCK(ih);
5764 	}
5765 }
5766 
5767 static int
5768 pf_clear_tables(void)
5769 {
5770 	struct pfioc_table io;
5771 	int error;
5772 
5773 	bzero(&io, sizeof(io));
5774 
5775 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
5776 	    io.pfrio_flags);
5777 
5778 	return (error);
5779 }
5780 
5781 static void
5782 pf_clear_srcnodes(struct pf_ksrc_node *n)
5783 {
5784 	struct pf_kstate *s;
5785 	int i;
5786 
5787 	for (i = 0; i <= pf_hashmask; i++) {
5788 		struct pf_idhash *ih = &V_pf_idhash[i];
5789 
5790 		PF_HASHROW_LOCK(ih);
5791 		LIST_FOREACH(s, &ih->states, entry) {
5792 			if (n == NULL || n == s->src_node)
5793 				s->src_node = NULL;
5794 			if (n == NULL || n == s->nat_src_node)
5795 				s->nat_src_node = NULL;
5796 		}
5797 		PF_HASHROW_UNLOCK(ih);
5798 	}
5799 
5800 	if (n == NULL) {
5801 		struct pf_srchash *sh;
5802 
5803 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5804 		    i++, sh++) {
5805 			PF_HASHROW_LOCK(sh);
5806 			LIST_FOREACH(n, &sh->nodes, entry) {
5807 				n->expire = 1;
5808 				n->states = 0;
5809 			}
5810 			PF_HASHROW_UNLOCK(sh);
5811 		}
5812 	} else {
5813 		/* XXX: hash slot should already be locked here. */
5814 		n->expire = 1;
5815 		n->states = 0;
5816 	}
5817 }
5818 
5819 static void
5820 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
5821 {
5822 	struct pf_ksrc_node_list	 kill;
5823 
5824 	LIST_INIT(&kill);
5825 	for (int i = 0; i <= pf_srchashmask; i++) {
5826 		struct pf_srchash *sh = &V_pf_srchash[i];
5827 		struct pf_ksrc_node *sn, *tmp;
5828 
5829 		PF_HASHROW_LOCK(sh);
5830 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
5831 			if (PF_MATCHA(psnk->psnk_src.neg,
5832 			      &psnk->psnk_src.addr.v.a.addr,
5833 			      &psnk->psnk_src.addr.v.a.mask,
5834 			      &sn->addr, sn->af) &&
5835 			    PF_MATCHA(psnk->psnk_dst.neg,
5836 			      &psnk->psnk_dst.addr.v.a.addr,
5837 			      &psnk->psnk_dst.addr.v.a.mask,
5838 			      &sn->raddr, sn->af)) {
5839 				pf_unlink_src_node(sn);
5840 				LIST_INSERT_HEAD(&kill, sn, entry);
5841 				sn->expire = 1;
5842 			}
5843 		PF_HASHROW_UNLOCK(sh);
5844 	}
5845 
5846 	for (int i = 0; i <= pf_hashmask; i++) {
5847 		struct pf_idhash *ih = &V_pf_idhash[i];
5848 		struct pf_kstate *s;
5849 
5850 		PF_HASHROW_LOCK(ih);
5851 		LIST_FOREACH(s, &ih->states, entry) {
5852 			if (s->src_node && s->src_node->expire == 1)
5853 				s->src_node = NULL;
5854 			if (s->nat_src_node && s->nat_src_node->expire == 1)
5855 				s->nat_src_node = NULL;
5856 		}
5857 		PF_HASHROW_UNLOCK(ih);
5858 	}
5859 
5860 	psnk->psnk_killed = pf_free_src_nodes(&kill);
5861 }
5862 
5863 static int
5864 pf_keepcounters(struct pfioc_nv *nv)
5865 {
5866 	nvlist_t	*nvl = NULL;
5867 	void		*nvlpacked = NULL;
5868 	int		 error = 0;
5869 
5870 #define	ERROUT(x)	ERROUT_FUNCTION(on_error, x)
5871 
5872 	if (nv->len > pf_ioctl_maxcount)
5873 		ERROUT(ENOMEM);
5874 
5875 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
5876 	if (nvlpacked == NULL)
5877 		ERROUT(ENOMEM);
5878 
5879 	error = copyin(nv->data, nvlpacked, nv->len);
5880 	if (error)
5881 		ERROUT(error);
5882 
5883 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
5884 	if (nvl == NULL)
5885 		ERROUT(EBADMSG);
5886 
5887 	if (! nvlist_exists_bool(nvl, "keep_counters"))
5888 		ERROUT(EBADMSG);
5889 
5890 	V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters");
5891 
5892 on_error:
5893 	nvlist_destroy(nvl);
5894 	free(nvlpacked, M_NVLIST);
5895 	return (error);
5896 }
5897 
5898 static unsigned int
5899 pf_clear_states(const struct pf_kstate_kill *kill)
5900 {
5901 	struct pf_state_key_cmp	 match_key;
5902 	struct pf_kstate	*s;
5903 	struct pfi_kkif	*kif;
5904 	int		 idx;
5905 	unsigned int	 killed = 0, dir;
5906 
5907 	for (unsigned int i = 0; i <= pf_hashmask; i++) {
5908 		struct pf_idhash *ih = &V_pf_idhash[i];
5909 
5910 relock_DIOCCLRSTATES:
5911 		PF_HASHROW_LOCK(ih);
5912 		LIST_FOREACH(s, &ih->states, entry) {
5913 			/* For floating states look at the original kif. */
5914 			kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
5915 
5916 			if (kill->psk_ifname[0] &&
5917 			    strcmp(kill->psk_ifname,
5918 			    kif->pfik_name))
5919 				continue;
5920 
5921 			if (kill->psk_kill_match) {
5922 				bzero(&match_key, sizeof(match_key));
5923 
5924 				if (s->direction == PF_OUT) {
5925 					dir = PF_IN;
5926 					idx = PF_SK_STACK;
5927 				} else {
5928 					dir = PF_OUT;
5929 					idx = PF_SK_WIRE;
5930 				}
5931 
5932 				match_key.af = s->key[idx]->af;
5933 				match_key.proto = s->key[idx]->proto;
5934 				PF_ACPY(&match_key.addr[0],
5935 				    &s->key[idx]->addr[1], match_key.af);
5936 				match_key.port[0] = s->key[idx]->port[1];
5937 				PF_ACPY(&match_key.addr[1],
5938 				    &s->key[idx]->addr[0], match_key.af);
5939 				match_key.port[1] = s->key[idx]->port[0];
5940 			}
5941 
5942 			/*
5943 			 * Don't send out individual
5944 			 * delete messages.
5945 			 */
5946 			s->state_flags |= PFSTATE_NOSYNC;
5947 			pf_unlink_state(s);
5948 			killed++;
5949 
5950 			if (kill->psk_kill_match)
5951 				killed += pf_kill_matching_state(&match_key,
5952 				    dir);
5953 
5954 			goto relock_DIOCCLRSTATES;
5955 		}
5956 		PF_HASHROW_UNLOCK(ih);
5957 	}
5958 
5959 	if (V_pfsync_clear_states_ptr != NULL)
5960 		V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname);
5961 
5962 	return (killed);
5963 }
5964 
5965 static void
5966 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed)
5967 {
5968 	struct pf_kstate	*s;
5969 
5970 	if (kill->psk_pfcmp.id) {
5971 		if (kill->psk_pfcmp.creatorid == 0)
5972 			kill->psk_pfcmp.creatorid = V_pf_status.hostid;
5973 		if ((s = pf_find_state_byid(kill->psk_pfcmp.id,
5974 		    kill->psk_pfcmp.creatorid))) {
5975 			pf_unlink_state(s);
5976 			*killed = 1;
5977 		}
5978 		return;
5979 	}
5980 
5981 	for (unsigned int i = 0; i <= pf_hashmask; i++)
5982 		*killed += pf_killstates_row(kill, &V_pf_idhash[i]);
5983 
5984 	return;
5985 }
5986 
5987 static int
5988 pf_killstates_nv(struct pfioc_nv *nv)
5989 {
5990 	struct pf_kstate_kill	 kill;
5991 	nvlist_t		*nvl = NULL;
5992 	void			*nvlpacked = NULL;
5993 	int			 error = 0;
5994 	unsigned int		 killed = 0;
5995 
5996 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
5997 
5998 	if (nv->len > pf_ioctl_maxcount)
5999 		ERROUT(ENOMEM);
6000 
6001 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6002 	if (nvlpacked == NULL)
6003 		ERROUT(ENOMEM);
6004 
6005 	error = copyin(nv->data, nvlpacked, nv->len);
6006 	if (error)
6007 		ERROUT(error);
6008 
6009 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6010 	if (nvl == NULL)
6011 		ERROUT(EBADMSG);
6012 
6013 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6014 	if (error)
6015 		ERROUT(error);
6016 
6017 	pf_killstates(&kill, &killed);
6018 
6019 	free(nvlpacked, M_NVLIST);
6020 	nvlpacked = NULL;
6021 	nvlist_destroy(nvl);
6022 	nvl = nvlist_create(0);
6023 	if (nvl == NULL)
6024 		ERROUT(ENOMEM);
6025 
6026 	nvlist_add_number(nvl, "killed", killed);
6027 
6028 	nvlpacked = nvlist_pack(nvl, &nv->len);
6029 	if (nvlpacked == NULL)
6030 		ERROUT(ENOMEM);
6031 
6032 	if (nv->size == 0)
6033 		ERROUT(0);
6034 	else if (nv->size < nv->len)
6035 		ERROUT(ENOSPC);
6036 
6037 	error = copyout(nvlpacked, nv->data, nv->len);
6038 
6039 on_error:
6040 	nvlist_destroy(nvl);
6041 	free(nvlpacked, M_NVLIST);
6042 	return (error);
6043 }
6044 
6045 static int
6046 pf_clearstates_nv(struct pfioc_nv *nv)
6047 {
6048 	struct pf_kstate_kill	 kill;
6049 	nvlist_t		*nvl = NULL;
6050 	void			*nvlpacked = NULL;
6051 	int			 error = 0;
6052 	unsigned int		 killed;
6053 
6054 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
6055 
6056 	if (nv->len > pf_ioctl_maxcount)
6057 		ERROUT(ENOMEM);
6058 
6059 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6060 	if (nvlpacked == NULL)
6061 		ERROUT(ENOMEM);
6062 
6063 	error = copyin(nv->data, nvlpacked, nv->len);
6064 	if (error)
6065 		ERROUT(error);
6066 
6067 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6068 	if (nvl == NULL)
6069 		ERROUT(EBADMSG);
6070 
6071 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6072 	if (error)
6073 		ERROUT(error);
6074 
6075 	killed = pf_clear_states(&kill);
6076 
6077 	free(nvlpacked, M_NVLIST);
6078 	nvlpacked = NULL;
6079 	nvlist_destroy(nvl);
6080 	nvl = nvlist_create(0);
6081 	if (nvl == NULL)
6082 		ERROUT(ENOMEM);
6083 
6084 	nvlist_add_number(nvl, "killed", killed);
6085 
6086 	nvlpacked = nvlist_pack(nvl, &nv->len);
6087 	if (nvlpacked == NULL)
6088 		ERROUT(ENOMEM);
6089 
6090 	if (nv->size == 0)
6091 		ERROUT(0);
6092 	else if (nv->size < nv->len)
6093 		ERROUT(ENOSPC);
6094 
6095 	error = copyout(nvlpacked, nv->data, nv->len);
6096 
6097 #undef ERROUT
6098 on_error:
6099 	nvlist_destroy(nvl);
6100 	free(nvlpacked, M_NVLIST);
6101 	return (error);
6102 }
6103 
6104 static int
6105 pf_getstate(struct pfioc_nv *nv)
6106 {
6107 	nvlist_t		*nvl = NULL, *nvls;
6108 	void			*nvlpacked = NULL;
6109 	struct pf_kstate	*s = NULL;
6110 	int			 error = 0;
6111 	uint64_t		 id, creatorid;
6112 
6113 #define ERROUT(x)	ERROUT_FUNCTION(errout, x)
6114 
6115 	if (nv->len > pf_ioctl_maxcount)
6116 		ERROUT(ENOMEM);
6117 
6118 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6119 	if (nvlpacked == NULL)
6120 		ERROUT(ENOMEM);
6121 
6122 	error = copyin(nv->data, nvlpacked, nv->len);
6123 	if (error)
6124 		ERROUT(error);
6125 
6126 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6127 	if (nvl == NULL)
6128 		ERROUT(EBADMSG);
6129 
6130 	PFNV_CHK(pf_nvuint64(nvl, "id", &id));
6131 	PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid));
6132 
6133 	s = pf_find_state_byid(id, creatorid);
6134 	if (s == NULL)
6135 		ERROUT(ENOENT);
6136 
6137 	free(nvlpacked, M_NVLIST);
6138 	nvlpacked = NULL;
6139 	nvlist_destroy(nvl);
6140 	nvl = nvlist_create(0);
6141 	if (nvl == NULL)
6142 		ERROUT(ENOMEM);
6143 
6144 	nvls = pf_state_to_nvstate(s);
6145 	if (nvls == NULL)
6146 		ERROUT(ENOMEM);
6147 
6148 	nvlist_add_nvlist(nvl, "state", nvls);
6149 	nvlist_destroy(nvls);
6150 
6151 	nvlpacked = nvlist_pack(nvl, &nv->len);
6152 	if (nvlpacked == NULL)
6153 		ERROUT(ENOMEM);
6154 
6155 	if (nv->size == 0)
6156 		ERROUT(0);
6157 	else if (nv->size < nv->len)
6158 		ERROUT(ENOSPC);
6159 
6160 	error = copyout(nvlpacked, nv->data, nv->len);
6161 
6162 #undef ERROUT
6163 errout:
6164 	if (s != NULL)
6165 		PF_STATE_UNLOCK(s);
6166 	free(nvlpacked, M_NVLIST);
6167 	nvlist_destroy(nvl);
6168 	return (error);
6169 }
6170 
6171 /*
6172  * XXX - Check for version mismatch!!!
6173  */
6174 
6175 /*
6176  * Duplicate pfctl -Fa operation to get rid of as much as we can.
6177  */
6178 static int
6179 shutdown_pf(void)
6180 {
6181 	int error = 0;
6182 	u_int32_t t[5];
6183 	char nn = '\0';
6184 
6185 	do {
6186 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
6187 		    != 0) {
6188 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
6189 			break;
6190 		}
6191 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
6192 		    != 0) {
6193 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
6194 			break;		/* XXX: rollback? */
6195 		}
6196 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
6197 		    != 0) {
6198 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
6199 			break;		/* XXX: rollback? */
6200 		}
6201 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
6202 		    != 0) {
6203 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
6204 			break;		/* XXX: rollback? */
6205 		}
6206 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
6207 		    != 0) {
6208 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
6209 			break;		/* XXX: rollback? */
6210 		}
6211 
6212 		/* XXX: these should always succeed here */
6213 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
6214 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
6215 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
6216 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
6217 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
6218 
6219 		if ((error = pf_clear_tables()) != 0)
6220 			break;
6221 
6222 		if ((error = pf_begin_eth(&t[0], &nn)) != 0) {
6223 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth\n"));
6224 			break;
6225 		}
6226 		pf_commit_eth(t[0], &nn);
6227 
6228 #ifdef ALTQ
6229 		if ((error = pf_begin_altq(&t[0])) != 0) {
6230 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
6231 			break;
6232 		}
6233 		pf_commit_altq(t[0]);
6234 #endif
6235 
6236 		pf_clear_all_states();
6237 
6238 		pf_clear_srcnodes(NULL);
6239 
6240 		/* status does not use malloced mem so no need to cleanup */
6241 		/* fingerprints and interfaces have their own cleanup code */
6242 	} while(0);
6243 
6244 	return (error);
6245 }
6246 
6247 static pfil_return_t
6248 pf_check_return(int chk, struct mbuf **m)
6249 {
6250 
6251 	switch (chk) {
6252 	case PF_PASS:
6253 		if (*m == NULL)
6254 			return (PFIL_CONSUMED);
6255 		else
6256 			return (PFIL_PASS);
6257 		break;
6258 	default:
6259 		if (*m != NULL) {
6260 			m_freem(*m);
6261 			*m = NULL;
6262 		}
6263 		return (PFIL_DROPPED);
6264 	}
6265 }
6266 
6267 static pfil_return_t
6268 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6269     void *ruleset __unused, struct inpcb *inp)
6270 {
6271 	int chk;
6272 
6273 	chk = pf_test_eth(PF_IN, flags, ifp, m, inp);
6274 
6275 	return (pf_check_return(chk, m));
6276 }
6277 
6278 static pfil_return_t
6279 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6280     void *ruleset __unused, struct inpcb *inp)
6281 {
6282 	int chk;
6283 
6284 	chk = pf_test_eth(PF_OUT, flags, ifp, m, inp);
6285 
6286 	return (pf_check_return(chk, m));
6287 }
6288 
6289 #ifdef INET
6290 static pfil_return_t
6291 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6292     void *ruleset __unused, struct inpcb *inp)
6293 {
6294 	int chk;
6295 
6296 	chk = pf_test(PF_IN, flags, ifp, m, inp, NULL);
6297 
6298 	return (pf_check_return(chk, m));
6299 }
6300 
6301 static pfil_return_t
6302 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6303     void *ruleset __unused,  struct inpcb *inp)
6304 {
6305 	int chk;
6306 
6307 	chk = pf_test(PF_OUT, flags, ifp, m, inp, NULL);
6308 
6309 	return (pf_check_return(chk, m));
6310 }
6311 #endif
6312 
6313 #ifdef INET6
6314 static pfil_return_t
6315 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
6316     void *ruleset __unused,  struct inpcb *inp)
6317 {
6318 	int chk;
6319 
6320 	/*
6321 	 * In case of loopback traffic IPv6 uses the real interface in
6322 	 * order to support scoped addresses. In order to support stateful
6323 	 * filtering we have change this to lo0 as it is the case in IPv4.
6324 	 */
6325 	CURVNET_SET(ifp->if_vnet);
6326 	chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp,
6327 	    m, inp, NULL);
6328 	CURVNET_RESTORE();
6329 
6330 	return (pf_check_return(chk, m));
6331 }
6332 
6333 static pfil_return_t
6334 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
6335     void *ruleset __unused,  struct inpcb *inp)
6336 {
6337 	int chk;
6338 
6339 	CURVNET_SET(ifp->if_vnet);
6340 	chk = pf_test6(PF_OUT, flags, ifp, m, inp, NULL);
6341 	CURVNET_RESTORE();
6342 
6343 	return (pf_check_return(chk, m));
6344 }
6345 #endif /* INET6 */
6346 
6347 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook);
6348 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook);
6349 #define	V_pf_eth_in_hook	VNET(pf_eth_in_hook)
6350 #define	V_pf_eth_out_hook	VNET(pf_eth_out_hook)
6351 
6352 #ifdef INET
6353 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
6354 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
6355 #define	V_pf_ip4_in_hook	VNET(pf_ip4_in_hook)
6356 #define	V_pf_ip4_out_hook	VNET(pf_ip4_out_hook)
6357 #endif
6358 #ifdef INET6
6359 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
6360 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
6361 #define	V_pf_ip6_in_hook	VNET(pf_ip6_in_hook)
6362 #define	V_pf_ip6_out_hook	VNET(pf_ip6_out_hook)
6363 #endif
6364 
6365 static void
6366 hook_pf_eth(void)
6367 {
6368 	struct pfil_hook_args pha = {
6369 		.pa_version = PFIL_VERSION,
6370 		.pa_modname = "pf",
6371 		.pa_type = PFIL_TYPE_ETHERNET,
6372 	};
6373 	struct pfil_link_args pla = {
6374 		.pa_version = PFIL_VERSION,
6375 	};
6376 	int ret __diagused;
6377 
6378 	if (atomic_load_bool(&V_pf_pfil_eth_hooked))
6379 		return;
6380 
6381 	pha.pa_mbuf_chk = pf_eth_check_in;
6382 	pha.pa_flags = PFIL_IN;
6383 	pha.pa_rulname = "eth-in";
6384 	V_pf_eth_in_hook = pfil_add_hook(&pha);
6385 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6386 	pla.pa_head = V_link_pfil_head;
6387 	pla.pa_hook = V_pf_eth_in_hook;
6388 	ret = pfil_link(&pla);
6389 	MPASS(ret == 0);
6390 	pha.pa_mbuf_chk = pf_eth_check_out;
6391 	pha.pa_flags = PFIL_OUT;
6392 	pha.pa_rulname = "eth-out";
6393 	V_pf_eth_out_hook = pfil_add_hook(&pha);
6394 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6395 	pla.pa_head = V_link_pfil_head;
6396 	pla.pa_hook = V_pf_eth_out_hook;
6397 	ret = pfil_link(&pla);
6398 	MPASS(ret == 0);
6399 
6400 	atomic_store_bool(&V_pf_pfil_eth_hooked, true);
6401 }
6402 
6403 static void
6404 hook_pf(void)
6405 {
6406 	struct pfil_hook_args pha = {
6407 		.pa_version = PFIL_VERSION,
6408 		.pa_modname = "pf",
6409 	};
6410 	struct pfil_link_args pla = {
6411 		.pa_version = PFIL_VERSION,
6412 	};
6413 	int ret __diagused;
6414 
6415 	if (atomic_load_bool(&V_pf_pfil_hooked))
6416 		return;
6417 
6418 #ifdef INET
6419 	pha.pa_type = PFIL_TYPE_IP4;
6420 	pha.pa_mbuf_chk = pf_check_in;
6421 	pha.pa_flags = PFIL_IN;
6422 	pha.pa_rulname = "default-in";
6423 	V_pf_ip4_in_hook = pfil_add_hook(&pha);
6424 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6425 	pla.pa_head = V_inet_pfil_head;
6426 	pla.pa_hook = V_pf_ip4_in_hook;
6427 	ret = pfil_link(&pla);
6428 	MPASS(ret == 0);
6429 	pha.pa_mbuf_chk = pf_check_out;
6430 	pha.pa_flags = PFIL_OUT;
6431 	pha.pa_rulname = "default-out";
6432 	V_pf_ip4_out_hook = pfil_add_hook(&pha);
6433 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6434 	pla.pa_head = V_inet_pfil_head;
6435 	pla.pa_hook = V_pf_ip4_out_hook;
6436 	ret = pfil_link(&pla);
6437 	MPASS(ret == 0);
6438 	if (V_pf_filter_local) {
6439 		pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6440 		pla.pa_head = V_inet_local_pfil_head;
6441 		pla.pa_hook = V_pf_ip4_out_hook;
6442 		ret = pfil_link(&pla);
6443 		MPASS(ret == 0);
6444 	}
6445 #endif
6446 #ifdef INET6
6447 	pha.pa_type = PFIL_TYPE_IP6;
6448 	pha.pa_mbuf_chk = pf_check6_in;
6449 	pha.pa_flags = PFIL_IN;
6450 	pha.pa_rulname = "default-in6";
6451 	V_pf_ip6_in_hook = pfil_add_hook(&pha);
6452 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6453 	pla.pa_head = V_inet6_pfil_head;
6454 	pla.pa_hook = V_pf_ip6_in_hook;
6455 	ret = pfil_link(&pla);
6456 	MPASS(ret == 0);
6457 	pha.pa_mbuf_chk = pf_check6_out;
6458 	pha.pa_rulname = "default-out6";
6459 	pha.pa_flags = PFIL_OUT;
6460 	V_pf_ip6_out_hook = pfil_add_hook(&pha);
6461 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6462 	pla.pa_head = V_inet6_pfil_head;
6463 	pla.pa_hook = V_pf_ip6_out_hook;
6464 	ret = pfil_link(&pla);
6465 	MPASS(ret == 0);
6466 	if (V_pf_filter_local) {
6467 		pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6468 		pla.pa_head = V_inet6_local_pfil_head;
6469 		pla.pa_hook = V_pf_ip6_out_hook;
6470 		ret = pfil_link(&pla);
6471 		MPASS(ret == 0);
6472 	}
6473 #endif
6474 
6475 	atomic_store_bool(&V_pf_pfil_hooked, true);
6476 }
6477 
6478 static void
6479 dehook_pf_eth(void)
6480 {
6481 
6482 	if (!atomic_load_bool(&V_pf_pfil_eth_hooked))
6483 		return;
6484 
6485 	pfil_remove_hook(V_pf_eth_in_hook);
6486 	pfil_remove_hook(V_pf_eth_out_hook);
6487 
6488 	atomic_store_bool(&V_pf_pfil_eth_hooked, false);
6489 }
6490 
6491 static void
6492 dehook_pf(void)
6493 {
6494 
6495 	if (!atomic_load_bool(&V_pf_pfil_hooked))
6496 		return;
6497 
6498 #ifdef INET
6499 	pfil_remove_hook(V_pf_ip4_in_hook);
6500 	pfil_remove_hook(V_pf_ip4_out_hook);
6501 #endif
6502 #ifdef INET6
6503 	pfil_remove_hook(V_pf_ip6_in_hook);
6504 	pfil_remove_hook(V_pf_ip6_out_hook);
6505 #endif
6506 
6507 	atomic_store_bool(&V_pf_pfil_hooked, false);
6508 }
6509 
6510 static void
6511 pf_load_vnet(void)
6512 {
6513 	V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
6514 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
6515 
6516 	rm_init_flags(&V_pf_rules_lock, "pf rulesets", RM_RECURSE);
6517 	sx_init(&V_pf_ioctl_lock, "pf ioctl");
6518 
6519 	pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
6520 	    PF_RULE_TAG_HASH_SIZE_DEFAULT);
6521 #ifdef ALTQ
6522 	pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
6523 	    PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
6524 #endif
6525 
6526 	V_pf_keth = &V_pf_main_keth_anchor.ruleset;
6527 
6528 	pfattach_vnet();
6529 	V_pf_vnet_active = 1;
6530 }
6531 
6532 static int
6533 pf_load(void)
6534 {
6535 	int error;
6536 
6537 	sx_init(&pf_end_lock, "pf end thread");
6538 
6539 	pf_mtag_initialize();
6540 
6541 	pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
6542 	if (pf_dev == NULL)
6543 		return (ENOMEM);
6544 
6545 	pf_end_threads = 0;
6546 	error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
6547 	if (error != 0)
6548 		return (error);
6549 
6550 	pfi_initialize();
6551 
6552 	return (0);
6553 }
6554 
6555 static void
6556 pf_unload_vnet(void)
6557 {
6558 	int ret __diagused;
6559 
6560 	V_pf_vnet_active = 0;
6561 	V_pf_status.running = 0;
6562 	dehook_pf();
6563 	dehook_pf_eth();
6564 
6565 	PF_RULES_WLOCK();
6566 	pf_syncookies_cleanup();
6567 	shutdown_pf();
6568 	PF_RULES_WUNLOCK();
6569 
6570 	/* Make sure we've cleaned up ethernet rules before we continue. */
6571 	NET_EPOCH_DRAIN_CALLBACKS();
6572 
6573 	ret = swi_remove(V_pf_swi_cookie);
6574 	MPASS(ret == 0);
6575 	ret = intr_event_destroy(V_pf_swi_ie);
6576 	MPASS(ret == 0);
6577 
6578 	pf_unload_vnet_purge();
6579 
6580 	pf_normalize_cleanup();
6581 	PF_RULES_WLOCK();
6582 	pfi_cleanup_vnet();
6583 	PF_RULES_WUNLOCK();
6584 	pfr_cleanup();
6585 	pf_osfp_flush();
6586 	pf_cleanup();
6587 	if (IS_DEFAULT_VNET(curvnet))
6588 		pf_mtag_cleanup();
6589 
6590 	pf_cleanup_tagset(&V_pf_tags);
6591 #ifdef ALTQ
6592 	pf_cleanup_tagset(&V_pf_qids);
6593 #endif
6594 	uma_zdestroy(V_pf_tag_z);
6595 
6596 #ifdef PF_WANT_32_TO_64_COUNTER
6597 	PF_RULES_WLOCK();
6598 	LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist);
6599 
6600 	MPASS(LIST_EMPTY(&V_pf_allkiflist));
6601 	MPASS(V_pf_allkifcount == 0);
6602 
6603 	LIST_REMOVE(&V_pf_default_rule, allrulelist);
6604 	V_pf_allrulecount--;
6605 	LIST_REMOVE(V_pf_rulemarker, allrulelist);
6606 
6607 	/*
6608 	 * There are known pf rule leaks when running the test suite.
6609 	 */
6610 #ifdef notyet
6611 	MPASS(LIST_EMPTY(&V_pf_allrulelist));
6612 	MPASS(V_pf_allrulecount == 0);
6613 #endif
6614 
6615 	PF_RULES_WUNLOCK();
6616 
6617 	free(V_pf_kifmarker, PFI_MTYPE);
6618 	free(V_pf_rulemarker, M_PFRULE);
6619 #endif
6620 
6621 	/* Free counters last as we updated them during shutdown. */
6622 	pf_counter_u64_deinit(&V_pf_default_rule.evaluations);
6623 	for (int i = 0; i < 2; i++) {
6624 		pf_counter_u64_deinit(&V_pf_default_rule.packets[i]);
6625 		pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]);
6626 	}
6627 	counter_u64_free(V_pf_default_rule.states_cur);
6628 	counter_u64_free(V_pf_default_rule.states_tot);
6629 	counter_u64_free(V_pf_default_rule.src_nodes);
6630 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, V_pf_default_rule.timestamp);
6631 
6632 	for (int i = 0; i < PFRES_MAX; i++)
6633 		counter_u64_free(V_pf_status.counters[i]);
6634 	for (int i = 0; i < KLCNT_MAX; i++)
6635 		counter_u64_free(V_pf_status.lcounters[i]);
6636 	for (int i = 0; i < FCNT_MAX; i++)
6637 		pf_counter_u64_deinit(&V_pf_status.fcounters[i]);
6638 	for (int i = 0; i < SCNT_MAX; i++)
6639 		counter_u64_free(V_pf_status.scounters[i]);
6640 
6641 	rm_destroy(&V_pf_rules_lock);
6642 	sx_destroy(&V_pf_ioctl_lock);
6643 }
6644 
6645 static void
6646 pf_unload(void)
6647 {
6648 
6649 	sx_xlock(&pf_end_lock);
6650 	pf_end_threads = 1;
6651 	while (pf_end_threads < 2) {
6652 		wakeup_one(pf_purge_thread);
6653 		sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
6654 	}
6655 	sx_xunlock(&pf_end_lock);
6656 
6657 	pf_nl_unregister();
6658 
6659 	if (pf_dev != NULL)
6660 		destroy_dev(pf_dev);
6661 
6662 	pfi_cleanup();
6663 
6664 	sx_destroy(&pf_end_lock);
6665 }
6666 
6667 static void
6668 vnet_pf_init(void *unused __unused)
6669 {
6670 
6671 	pf_load_vnet();
6672 }
6673 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6674     vnet_pf_init, NULL);
6675 
6676 static void
6677 vnet_pf_uninit(const void *unused __unused)
6678 {
6679 
6680 	pf_unload_vnet();
6681 }
6682 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
6683 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6684     vnet_pf_uninit, NULL);
6685 
6686 static int
6687 pf_modevent(module_t mod, int type, void *data)
6688 {
6689 	int error = 0;
6690 
6691 	switch(type) {
6692 	case MOD_LOAD:
6693 		error = pf_load();
6694 		pf_nl_register();
6695 		break;
6696 	case MOD_UNLOAD:
6697 		/* Handled in SYSUNINIT(pf_unload) to ensure it's done after
6698 		 * the vnet_pf_uninit()s */
6699 		break;
6700 	default:
6701 		error = EINVAL;
6702 		break;
6703 	}
6704 
6705 	return (error);
6706 }
6707 
6708 static moduledata_t pf_mod = {
6709 	"pf",
6710 	pf_modevent,
6711 	0
6712 };
6713 
6714 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
6715 MODULE_DEPEND(pf, netlink, 1, 1, 1);
6716 MODULE_VERSION(pf, PF_MODVER);
6717