xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision bdd1243df58e60e85101c09001d9812a789b6bc4)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2001 Daniel Hartmeier
5  * Copyright (c) 2002,2003 Henning Brauer
6  * Copyright (c) 2012 Gleb Smirnoff <glebius@FreeBSD.org>
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  *
13  *    - Redistributions of source code must retain the above copyright
14  *      notice, this list of conditions and the following disclaimer.
15  *    - Redistributions in binary form must reproduce the above
16  *      copyright notice, this list of conditions and the following
17  *      disclaimer in the documentation and/or other materials provided
18  *      with the distribution.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
26  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
28  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
30  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
31  * POSSIBILITY OF SUCH DAMAGE.
32  *
33  * Effort sponsored in part by the Defense Advanced Research Projects
34  * Agency (DARPA) and Air Force Research Laboratory, Air Force
35  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
36  *
37  *	$OpenBSD: pf_ioctl.c,v 1.213 2009/02/15 21:46:12 mbalmer Exp $
38  */
39 
40 #include <sys/cdefs.h>
41 __FBSDID("$FreeBSD$");
42 
43 #include "opt_inet.h"
44 #include "opt_inet6.h"
45 #include "opt_bpf.h"
46 #include "opt_pf.h"
47 
48 #include <sys/param.h>
49 #include <sys/_bitset.h>
50 #include <sys/bitset.h>
51 #include <sys/bus.h>
52 #include <sys/conf.h>
53 #include <sys/endian.h>
54 #include <sys/fcntl.h>
55 #include <sys/filio.h>
56 #include <sys/hash.h>
57 #include <sys/interrupt.h>
58 #include <sys/jail.h>
59 #include <sys/kernel.h>
60 #include <sys/kthread.h>
61 #include <sys/lock.h>
62 #include <sys/mbuf.h>
63 #include <sys/module.h>
64 #include <sys/nv.h>
65 #include <sys/proc.h>
66 #include <sys/sdt.h>
67 #include <sys/smp.h>
68 #include <sys/socket.h>
69 #include <sys/sysctl.h>
70 #include <sys/md5.h>
71 #include <sys/ucred.h>
72 
73 #include <net/if.h>
74 #include <net/if_var.h>
75 #include <net/if_private.h>
76 #include <net/vnet.h>
77 #include <net/route.h>
78 #include <net/pfil.h>
79 #include <net/pfvar.h>
80 #include <net/if_pfsync.h>
81 #include <net/if_pflog.h>
82 
83 #include <netinet/in.h>
84 #include <netinet/ip.h>
85 #include <netinet/ip_var.h>
86 #include <netinet6/ip6_var.h>
87 #include <netinet/ip_icmp.h>
88 #include <netpfil/pf/pf_nv.h>
89 
90 #ifdef INET6
91 #include <netinet/ip6.h>
92 #endif /* INET6 */
93 
94 #ifdef ALTQ
95 #include <net/altq/altq.h>
96 #endif
97 
98 SDT_PROBE_DEFINE3(pf, ioctl, ioctl, error, "int", "int", "int");
99 SDT_PROBE_DEFINE3(pf, ioctl, function, error, "char *", "int", "int");
100 SDT_PROBE_DEFINE2(pf, ioctl, addrule, error, "int", "int");
101 SDT_PROBE_DEFINE2(pf, ioctl, nvchk, error, "int", "int");
102 
103 static struct pf_kpool	*pf_get_kpool(const char *, u_int32_t, u_int8_t,
104 			    u_int32_t, u_int8_t, u_int8_t, u_int8_t);
105 
106 static void		 pf_mv_kpool(struct pf_kpalist *, struct pf_kpalist *);
107 static void		 pf_empty_kpool(struct pf_kpalist *);
108 static int		 pfioctl(struct cdev *, u_long, caddr_t, int,
109 			    struct thread *);
110 static int		 pf_begin_eth(uint32_t *, const char *);
111 static void		 pf_rollback_eth_cb(struct epoch_context *);
112 static int		 pf_rollback_eth(uint32_t, const char *);
113 static int		 pf_commit_eth(uint32_t, const char *);
114 static void		 pf_free_eth_rule(struct pf_keth_rule *);
115 #ifdef ALTQ
116 static int		 pf_begin_altq(u_int32_t *);
117 static int		 pf_rollback_altq(u_int32_t);
118 static int		 pf_commit_altq(u_int32_t);
119 static int		 pf_enable_altq(struct pf_altq *);
120 static int		 pf_disable_altq(struct pf_altq *);
121 static uint16_t		 pf_qname2qid(const char *);
122 static void		 pf_qid_unref(uint16_t);
123 #endif /* ALTQ */
124 static int		 pf_begin_rules(u_int32_t *, int, const char *);
125 static int		 pf_rollback_rules(u_int32_t, int, char *);
126 static int		 pf_setup_pfsync_matching(struct pf_kruleset *);
127 static void		 pf_hash_rule_rolling(MD5_CTX *, struct pf_krule *);
128 static void		 pf_hash_rule(struct pf_krule *);
129 static void		 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
130 static int		 pf_commit_rules(u_int32_t, int, char *);
131 static int		 pf_addr_setup(struct pf_kruleset *,
132 			    struct pf_addr_wrap *, sa_family_t);
133 static void		 pf_addr_copyout(struct pf_addr_wrap *);
134 static void		 pf_src_node_copy(const struct pf_ksrc_node *,
135 			    struct pf_src_node *);
136 #ifdef ALTQ
137 static int		 pf_export_kaltq(struct pf_altq *,
138 			    struct pfioc_altq_v1 *, size_t);
139 static int		 pf_import_kaltq(struct pfioc_altq_v1 *,
140 			    struct pf_altq *, size_t);
141 #endif /* ALTQ */
142 
143 VNET_DEFINE(struct pf_krule,	pf_default_rule);
144 
145 static __inline int             pf_krule_compare(struct pf_krule *,
146 				    struct pf_krule *);
147 
148 RB_GENERATE(pf_krule_global, pf_krule, entry_global, pf_krule_compare);
149 
150 #ifdef ALTQ
151 VNET_DEFINE_STATIC(int,		pf_altq_running);
152 #define	V_pf_altq_running	VNET(pf_altq_running)
153 #endif
154 
155 #define	TAGID_MAX	 50000
156 struct pf_tagname {
157 	TAILQ_ENTRY(pf_tagname)	namehash_entries;
158 	TAILQ_ENTRY(pf_tagname)	taghash_entries;
159 	char			name[PF_TAG_NAME_SIZE];
160 	uint16_t		tag;
161 	int			ref;
162 };
163 
164 struct pf_tagset {
165 	TAILQ_HEAD(, pf_tagname)	*namehash;
166 	TAILQ_HEAD(, pf_tagname)	*taghash;
167 	unsigned int			 mask;
168 	uint32_t			 seed;
169 	BITSET_DEFINE(, TAGID_MAX)	 avail;
170 };
171 
172 VNET_DEFINE(struct pf_tagset, pf_tags);
173 #define	V_pf_tags	VNET(pf_tags)
174 static unsigned int	pf_rule_tag_hashsize;
175 #define	PF_RULE_TAG_HASH_SIZE_DEFAULT	128
176 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN,
177     &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT,
178     "Size of pf(4) rule tag hashtable");
179 
180 #ifdef ALTQ
181 VNET_DEFINE(struct pf_tagset, pf_qids);
182 #define	V_pf_qids	VNET(pf_qids)
183 static unsigned int	pf_queue_tag_hashsize;
184 #define	PF_QUEUE_TAG_HASH_SIZE_DEFAULT	128
185 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN,
186     &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT,
187     "Size of pf(4) queue tag hashtable");
188 #endif
189 VNET_DEFINE(uma_zone_t,	 pf_tag_z);
190 #define	V_pf_tag_z		 VNET(pf_tag_z)
191 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
192 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
193 
194 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
195 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
196 #endif
197 
198 VNET_DEFINE_STATIC(bool, pf_filter_local) = false;
199 #define V_pf_filter_local	VNET(pf_filter_local)
200 SYSCTL_BOOL(_net_pf, OID_AUTO, filter_local, CTLFLAG_VNET | CTLFLAG_RW,
201     &VNET_NAME(pf_filter_local), false,
202     "Enable filtering for packets delivered to local network stack");
203 
204 static void		 pf_init_tagset(struct pf_tagset *, unsigned int *,
205 			    unsigned int);
206 static void		 pf_cleanup_tagset(struct pf_tagset *);
207 static uint16_t		 tagname2hashindex(const struct pf_tagset *, const char *);
208 static uint16_t		 tag2hashindex(const struct pf_tagset *, uint16_t);
209 static u_int16_t	 tagname2tag(struct pf_tagset *, const char *);
210 static u_int16_t	 pf_tagname2tag(const char *);
211 static void		 tag_unref(struct pf_tagset *, u_int16_t);
212 
213 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
214 
215 struct cdev *pf_dev;
216 
217 /*
218  * XXX - These are new and need to be checked when moveing to a new version
219  */
220 static void		 pf_clear_all_states(void);
221 static unsigned int	 pf_clear_states(const struct pf_kstate_kill *);
222 static void		 pf_killstates(struct pf_kstate_kill *,
223 			    unsigned int *);
224 static int		 pf_killstates_row(struct pf_kstate_kill *,
225 			    struct pf_idhash *);
226 static int		 pf_killstates_nv(struct pfioc_nv *);
227 static int		 pf_clearstates_nv(struct pfioc_nv *);
228 static int		 pf_getstate(struct pfioc_nv *);
229 static int		 pf_getstatus(struct pfioc_nv *);
230 static int		 pf_clear_tables(void);
231 static void		 pf_clear_srcnodes(struct pf_ksrc_node *);
232 static void		 pf_kill_srcnodes(struct pfioc_src_node_kill *);
233 static int		 pf_keepcounters(struct pfioc_nv *);
234 static void		 pf_tbladdr_copyout(struct pf_addr_wrap *);
235 
236 /*
237  * Wrapper functions for pfil(9) hooks
238  */
239 static pfil_return_t pf_eth_check_in(struct mbuf **m, struct ifnet *ifp,
240     int flags, void *ruleset __unused, struct inpcb *inp);
241 static pfil_return_t pf_eth_check_out(struct mbuf **m, struct ifnet *ifp,
242     int flags, void *ruleset __unused, struct inpcb *inp);
243 #ifdef INET
244 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp,
245     int flags, void *ruleset __unused, struct inpcb *inp);
246 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp,
247     int flags, void *ruleset __unused, struct inpcb *inp);
248 #endif
249 #ifdef INET6
250 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp,
251     int flags, void *ruleset __unused, struct inpcb *inp);
252 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp,
253     int flags, void *ruleset __unused, struct inpcb *inp);
254 #endif
255 
256 static void		hook_pf_eth(void);
257 static void		hook_pf(void);
258 static void		dehook_pf_eth(void);
259 static void		dehook_pf(void);
260 static int		shutdown_pf(void);
261 static int		pf_load(void);
262 static void		pf_unload(void);
263 
264 static struct cdevsw pf_cdevsw = {
265 	.d_ioctl =	pfioctl,
266 	.d_name =	PF_NAME,
267 	.d_version =	D_VERSION,
268 };
269 
270 VNET_DEFINE_STATIC(bool, pf_pfil_hooked);
271 #define V_pf_pfil_hooked	VNET(pf_pfil_hooked)
272 VNET_DEFINE_STATIC(bool, pf_pfil_eth_hooked);
273 #define V_pf_pfil_eth_hooked	VNET(pf_pfil_eth_hooked)
274 
275 /*
276  * We need a flag that is neither hooked nor running to know when
277  * the VNET is "valid".  We primarily need this to control (global)
278  * external event, e.g., eventhandlers.
279  */
280 VNET_DEFINE(int, pf_vnet_active);
281 #define V_pf_vnet_active	VNET(pf_vnet_active)
282 
283 int pf_end_threads;
284 struct proc *pf_purge_proc;
285 
286 VNET_DEFINE(struct rmlock, pf_rules_lock);
287 VNET_DEFINE_STATIC(struct sx, pf_ioctl_lock);
288 #define	V_pf_ioctl_lock		VNET(pf_ioctl_lock)
289 struct sx			pf_end_lock;
290 
291 /* pfsync */
292 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
293 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
294 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
295 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
296 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
297 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
298 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
299 
300 /* pflog */
301 pflog_packet_t			*pflog_packet_ptr = NULL;
302 
303 /*
304  * Copy a user-provided string, returning an error if truncation would occur.
305  * Avoid scanning past "sz" bytes in the source string since there's no
306  * guarantee that it's nul-terminated.
307  */
308 static int
309 pf_user_strcpy(char *dst, const char *src, size_t sz)
310 {
311 	if (strnlen(src, sz) == sz)
312 		return (EINVAL);
313 	(void)strlcpy(dst, src, sz);
314 	return (0);
315 }
316 
317 static void
318 pfattach_vnet(void)
319 {
320 	u_int32_t *my_timeout = V_pf_default_rule.timeout;
321 
322 	bzero(&V_pf_status, sizeof(V_pf_status));
323 
324 	pf_initialize();
325 	pfr_initialize();
326 	pfi_initialize_vnet();
327 	pf_normalize_init();
328 	pf_syncookies_init();
329 
330 	V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
331 	V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
332 
333 	RB_INIT(&V_pf_anchors);
334 	pf_init_kruleset(&pf_main_ruleset);
335 
336 	pf_init_keth(V_pf_keth);
337 
338 	/* default rule should never be garbage collected */
339 	V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
340 #ifdef PF_DEFAULT_TO_DROP
341 	V_pf_default_rule.action = PF_DROP;
342 #else
343 	V_pf_default_rule.action = PF_PASS;
344 #endif
345 	V_pf_default_rule.nr = -1;
346 	V_pf_default_rule.rtableid = -1;
347 
348 	pf_counter_u64_init(&V_pf_default_rule.evaluations, M_WAITOK);
349 	for (int i = 0; i < 2; i++) {
350 		pf_counter_u64_init(&V_pf_default_rule.packets[i], M_WAITOK);
351 		pf_counter_u64_init(&V_pf_default_rule.bytes[i], M_WAITOK);
352 	}
353 	V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
354 	V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
355 	V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
356 
357 	V_pf_default_rule.timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
358 	    M_WAITOK | M_ZERO);
359 
360 #ifdef PF_WANT_32_TO_64_COUNTER
361 	V_pf_kifmarker = malloc(sizeof(*V_pf_kifmarker), PFI_MTYPE, M_WAITOK | M_ZERO);
362 	V_pf_rulemarker = malloc(sizeof(*V_pf_rulemarker), M_PFRULE, M_WAITOK | M_ZERO);
363 	PF_RULES_WLOCK();
364 	LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist);
365 	LIST_INSERT_HEAD(&V_pf_allrulelist, &V_pf_default_rule, allrulelist);
366 	V_pf_allrulecount++;
367 	LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist);
368 	PF_RULES_WUNLOCK();
369 #endif
370 
371 	/* initialize default timeouts */
372 	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
373 	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
374 	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
375 	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
376 	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
377 	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
378 	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
379 	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
380 	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
381 	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
382 	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
383 	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
384 	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
385 	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
386 	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
387 	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
388 	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
389 	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
390 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
391 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
392 
393 	V_pf_status.debug = PF_DEBUG_URGENT;
394 	/*
395 	 * XXX This is different than in OpenBSD where reassembly is enabled by
396 	 * defult. In FreeBSD we expect people to still use scrub rules and
397 	 * switch to the new syntax later. Only when they switch they must
398 	 * explicitly enable reassemle. We could change the default once the
399 	 * scrub rule functionality is hopefully removed some day in future.
400 	 */
401 	V_pf_status.reass = 0;
402 
403 	V_pf_pfil_hooked = false;
404 	V_pf_pfil_eth_hooked = false;
405 
406 	/* XXX do our best to avoid a conflict */
407 	V_pf_status.hostid = arc4random();
408 
409 	for (int i = 0; i < PFRES_MAX; i++)
410 		V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
411 	for (int i = 0; i < KLCNT_MAX; i++)
412 		V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
413 	for (int i = 0; i < FCNT_MAX; i++)
414 		pf_counter_u64_init(&V_pf_status.fcounters[i], M_WAITOK);
415 	for (int i = 0; i < SCNT_MAX; i++)
416 		V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
417 
418 	if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET,
419 	    INTR_MPSAFE, &V_pf_swi_cookie) != 0)
420 		/* XXXGL: leaked all above. */
421 		return;
422 }
423 
424 static struct pf_kpool *
425 pf_get_kpool(const char *anchor, u_int32_t ticket, u_int8_t rule_action,
426     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
427     u_int8_t check_ticket)
428 {
429 	struct pf_kruleset	*ruleset;
430 	struct pf_krule		*rule;
431 	int			 rs_num;
432 
433 	ruleset = pf_find_kruleset(anchor);
434 	if (ruleset == NULL)
435 		return (NULL);
436 	rs_num = pf_get_ruleset_number(rule_action);
437 	if (rs_num >= PF_RULESET_MAX)
438 		return (NULL);
439 	if (active) {
440 		if (check_ticket && ticket !=
441 		    ruleset->rules[rs_num].active.ticket)
442 			return (NULL);
443 		if (r_last)
444 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
445 			    pf_krulequeue);
446 		else
447 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
448 	} else {
449 		if (check_ticket && ticket !=
450 		    ruleset->rules[rs_num].inactive.ticket)
451 			return (NULL);
452 		if (r_last)
453 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
454 			    pf_krulequeue);
455 		else
456 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
457 	}
458 	if (!r_last) {
459 		while ((rule != NULL) && (rule->nr != rule_number))
460 			rule = TAILQ_NEXT(rule, entries);
461 	}
462 	if (rule == NULL)
463 		return (NULL);
464 
465 	return (&rule->rpool);
466 }
467 
468 static void
469 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb)
470 {
471 	struct pf_kpooladdr	*mv_pool_pa;
472 
473 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
474 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
475 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
476 	}
477 }
478 
479 static void
480 pf_empty_kpool(struct pf_kpalist *poola)
481 {
482 	struct pf_kpooladdr *pa;
483 
484 	while ((pa = TAILQ_FIRST(poola)) != NULL) {
485 		switch (pa->addr.type) {
486 		case PF_ADDR_DYNIFTL:
487 			pfi_dynaddr_remove(pa->addr.p.dyn);
488 			break;
489 		case PF_ADDR_TABLE:
490 			/* XXX: this could be unfinished pooladdr on pabuf */
491 			if (pa->addr.p.tbl != NULL)
492 				pfr_detach_table(pa->addr.p.tbl);
493 			break;
494 		}
495 		if (pa->kif)
496 			pfi_kkif_unref(pa->kif);
497 		TAILQ_REMOVE(poola, pa, entries);
498 		free(pa, M_PFRULE);
499 	}
500 }
501 
502 static void
503 pf_unlink_rule_locked(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
504 {
505 
506 	PF_RULES_WASSERT();
507 	PF_UNLNKDRULES_ASSERT();
508 
509 	TAILQ_REMOVE(rulequeue, rule, entries);
510 
511 	rule->rule_ref |= PFRULE_REFS;
512 	TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
513 }
514 
515 static void
516 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
517 {
518 
519 	PF_RULES_WASSERT();
520 
521 	PF_UNLNKDRULES_LOCK();
522 	pf_unlink_rule_locked(rulequeue, rule);
523 	PF_UNLNKDRULES_UNLOCK();
524 }
525 
526 static void
527 pf_free_eth_rule(struct pf_keth_rule *rule)
528 {
529 	PF_RULES_WASSERT();
530 
531 	if (rule == NULL)
532 		return;
533 
534 	if (rule->tag)
535 		tag_unref(&V_pf_tags, rule->tag);
536 	if (rule->match_tag)
537 		tag_unref(&V_pf_tags, rule->match_tag);
538 #ifdef ALTQ
539 	pf_qid_unref(rule->qid);
540 #endif
541 
542 	if (rule->bridge_to)
543 		pfi_kkif_unref(rule->bridge_to);
544 	if (rule->kif)
545 		pfi_kkif_unref(rule->kif);
546 
547 	if (rule->ipsrc.addr.type == PF_ADDR_TABLE)
548 		pfr_detach_table(rule->ipsrc.addr.p.tbl);
549 	if (rule->ipdst.addr.type == PF_ADDR_TABLE)
550 		pfr_detach_table(rule->ipdst.addr.p.tbl);
551 
552 	counter_u64_free(rule->evaluations);
553 	for (int i = 0; i < 2; i++) {
554 		counter_u64_free(rule->packets[i]);
555 		counter_u64_free(rule->bytes[i]);
556 	}
557 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
558 	pf_keth_anchor_remove(rule);
559 
560 	free(rule, M_PFRULE);
561 }
562 
563 void
564 pf_free_rule(struct pf_krule *rule)
565 {
566 
567 	PF_RULES_WASSERT();
568 	PF_CONFIG_ASSERT();
569 
570 	if (rule->tag)
571 		tag_unref(&V_pf_tags, rule->tag);
572 	if (rule->match_tag)
573 		tag_unref(&V_pf_tags, rule->match_tag);
574 #ifdef ALTQ
575 	if (rule->pqid != rule->qid)
576 		pf_qid_unref(rule->pqid);
577 	pf_qid_unref(rule->qid);
578 #endif
579 	switch (rule->src.addr.type) {
580 	case PF_ADDR_DYNIFTL:
581 		pfi_dynaddr_remove(rule->src.addr.p.dyn);
582 		break;
583 	case PF_ADDR_TABLE:
584 		pfr_detach_table(rule->src.addr.p.tbl);
585 		break;
586 	}
587 	switch (rule->dst.addr.type) {
588 	case PF_ADDR_DYNIFTL:
589 		pfi_dynaddr_remove(rule->dst.addr.p.dyn);
590 		break;
591 	case PF_ADDR_TABLE:
592 		pfr_detach_table(rule->dst.addr.p.tbl);
593 		break;
594 	}
595 	if (rule->overload_tbl)
596 		pfr_detach_table(rule->overload_tbl);
597 	if (rule->kif)
598 		pfi_kkif_unref(rule->kif);
599 	pf_kanchor_remove(rule);
600 	pf_empty_kpool(&rule->rpool.list);
601 
602 	pf_krule_free(rule);
603 }
604 
605 static void
606 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
607     unsigned int default_size)
608 {
609 	unsigned int i;
610 	unsigned int hashsize;
611 
612 	if (*tunable_size == 0 || !powerof2(*tunable_size))
613 		*tunable_size = default_size;
614 
615 	hashsize = *tunable_size;
616 	ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
617 	    M_WAITOK);
618 	ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
619 	    M_WAITOK);
620 	ts->mask = hashsize - 1;
621 	ts->seed = arc4random();
622 	for (i = 0; i < hashsize; i++) {
623 		TAILQ_INIT(&ts->namehash[i]);
624 		TAILQ_INIT(&ts->taghash[i]);
625 	}
626 	BIT_FILL(TAGID_MAX, &ts->avail);
627 }
628 
629 static void
630 pf_cleanup_tagset(struct pf_tagset *ts)
631 {
632 	unsigned int i;
633 	unsigned int hashsize;
634 	struct pf_tagname *t, *tmp;
635 
636 	/*
637 	 * Only need to clean up one of the hashes as each tag is hashed
638 	 * into each table.
639 	 */
640 	hashsize = ts->mask + 1;
641 	for (i = 0; i < hashsize; i++)
642 		TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
643 			uma_zfree(V_pf_tag_z, t);
644 
645 	free(ts->namehash, M_PFHASH);
646 	free(ts->taghash, M_PFHASH);
647 }
648 
649 static uint16_t
650 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
651 {
652 	size_t len;
653 
654 	len = strnlen(tagname, PF_TAG_NAME_SIZE - 1);
655 	return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask);
656 }
657 
658 static uint16_t
659 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
660 {
661 
662 	return (tag & ts->mask);
663 }
664 
665 static u_int16_t
666 tagname2tag(struct pf_tagset *ts, const char *tagname)
667 {
668 	struct pf_tagname	*tag;
669 	u_int32_t		 index;
670 	u_int16_t		 new_tagid;
671 
672 	PF_RULES_WASSERT();
673 
674 	index = tagname2hashindex(ts, tagname);
675 	TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
676 		if (strcmp(tagname, tag->name) == 0) {
677 			tag->ref++;
678 			return (tag->tag);
679 		}
680 
681 	/*
682 	 * new entry
683 	 *
684 	 * to avoid fragmentation, we do a linear search from the beginning
685 	 * and take the first free slot we find.
686 	 */
687 	new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
688 	/*
689 	 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
690 	 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
691 	 * set.  It may also return a bit number greater than TAGID_MAX due
692 	 * to rounding of the number of bits in the vector up to a multiple
693 	 * of the vector word size at declaration/allocation time.
694 	 */
695 	if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
696 		return (0);
697 
698 	/* Mark the tag as in use.  Bits are 0-based for BIT_CLR() */
699 	BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
700 
701 	/* allocate and fill new struct pf_tagname */
702 	tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
703 	if (tag == NULL)
704 		return (0);
705 	strlcpy(tag->name, tagname, sizeof(tag->name));
706 	tag->tag = new_tagid;
707 	tag->ref = 1;
708 
709 	/* Insert into namehash */
710 	TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
711 
712 	/* Insert into taghash */
713 	index = tag2hashindex(ts, new_tagid);
714 	TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
715 
716 	return (tag->tag);
717 }
718 
719 static void
720 tag_unref(struct pf_tagset *ts, u_int16_t tag)
721 {
722 	struct pf_tagname	*t;
723 	uint16_t		 index;
724 
725 	PF_RULES_WASSERT();
726 
727 	index = tag2hashindex(ts, tag);
728 	TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
729 		if (tag == t->tag) {
730 			if (--t->ref == 0) {
731 				TAILQ_REMOVE(&ts->taghash[index], t,
732 				    taghash_entries);
733 				index = tagname2hashindex(ts, t->name);
734 				TAILQ_REMOVE(&ts->namehash[index], t,
735 				    namehash_entries);
736 				/* Bits are 0-based for BIT_SET() */
737 				BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
738 				uma_zfree(V_pf_tag_z, t);
739 			}
740 			break;
741 		}
742 }
743 
744 static uint16_t
745 pf_tagname2tag(const char *tagname)
746 {
747 	return (tagname2tag(&V_pf_tags, tagname));
748 }
749 
750 static int
751 pf_begin_eth(uint32_t *ticket, const char *anchor)
752 {
753 	struct pf_keth_rule *rule, *tmp;
754 	struct pf_keth_ruleset *rs;
755 
756 	PF_RULES_WASSERT();
757 
758 	rs = pf_find_or_create_keth_ruleset(anchor);
759 	if (rs == NULL)
760 		return (EINVAL);
761 
762 	/* Purge old inactive rules. */
763 	TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
764 	    tmp) {
765 		TAILQ_REMOVE(rs->inactive.rules, rule,
766 		    entries);
767 		pf_free_eth_rule(rule);
768 	}
769 
770 	*ticket = ++rs->inactive.ticket;
771 	rs->inactive.open = 1;
772 
773 	return (0);
774 }
775 
776 static void
777 pf_rollback_eth_cb(struct epoch_context *ctx)
778 {
779 	struct pf_keth_ruleset *rs;
780 
781 	rs = __containerof(ctx, struct pf_keth_ruleset, epoch_ctx);
782 
783 	CURVNET_SET(rs->vnet);
784 
785 	PF_RULES_WLOCK();
786 	pf_rollback_eth(rs->inactive.ticket,
787 	    rs->anchor ? rs->anchor->path : "");
788 	PF_RULES_WUNLOCK();
789 
790 	CURVNET_RESTORE();
791 }
792 
793 static int
794 pf_rollback_eth(uint32_t ticket, const char *anchor)
795 {
796 	struct pf_keth_rule *rule, *tmp;
797 	struct pf_keth_ruleset *rs;
798 
799 	PF_RULES_WASSERT();
800 
801 	rs = pf_find_keth_ruleset(anchor);
802 	if (rs == NULL)
803 		return (EINVAL);
804 
805 	if (!rs->inactive.open ||
806 	    ticket != rs->inactive.ticket)
807 		return (0);
808 
809 	/* Purge old inactive rules. */
810 	TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
811 	    tmp) {
812 		TAILQ_REMOVE(rs->inactive.rules, rule, entries);
813 		pf_free_eth_rule(rule);
814 	}
815 
816 	rs->inactive.open = 0;
817 
818 	pf_remove_if_empty_keth_ruleset(rs);
819 
820 	return (0);
821 }
822 
823 #define	PF_SET_SKIP_STEPS(i)					\
824 	do {							\
825 		while (head[i] != cur) {			\
826 			head[i]->skip[i].ptr = cur;		\
827 			head[i] = TAILQ_NEXT(head[i], entries);	\
828 		}						\
829 	} while (0)
830 
831 static void
832 pf_eth_calc_skip_steps(struct pf_keth_ruleq *rules)
833 {
834 	struct pf_keth_rule *cur, *prev, *head[PFE_SKIP_COUNT];
835 	int i;
836 
837 	cur = TAILQ_FIRST(rules);
838 	prev = cur;
839 	for (i = 0; i < PFE_SKIP_COUNT; ++i)
840 		head[i] = cur;
841 	while (cur != NULL) {
842 		if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
843 			PF_SET_SKIP_STEPS(PFE_SKIP_IFP);
844 		if (cur->direction != prev->direction)
845 			PF_SET_SKIP_STEPS(PFE_SKIP_DIR);
846 		if (cur->proto != prev->proto)
847 			PF_SET_SKIP_STEPS(PFE_SKIP_PROTO);
848 		if (memcmp(&cur->src, &prev->src, sizeof(cur->src)) != 0)
849 			PF_SET_SKIP_STEPS(PFE_SKIP_SRC_ADDR);
850 		if (memcmp(&cur->dst, &prev->dst, sizeof(cur->dst)) != 0)
851 			PF_SET_SKIP_STEPS(PFE_SKIP_DST_ADDR);
852 		if (cur->ipsrc.neg != prev->ipsrc.neg ||
853 		    pf_addr_wrap_neq(&cur->ipsrc.addr, &prev->ipsrc.addr))
854 			PF_SET_SKIP_STEPS(PFE_SKIP_SRC_IP_ADDR);
855 		if (cur->ipdst.neg != prev->ipdst.neg ||
856 		    pf_addr_wrap_neq(&cur->ipdst.addr, &prev->ipdst.addr))
857 			PF_SET_SKIP_STEPS(PFE_SKIP_DST_IP_ADDR);
858 
859 		prev = cur;
860 		cur = TAILQ_NEXT(cur, entries);
861 	}
862 	for (i = 0; i < PFE_SKIP_COUNT; ++i)
863 		PF_SET_SKIP_STEPS(i);
864 }
865 
866 static int
867 pf_commit_eth(uint32_t ticket, const char *anchor)
868 {
869 	struct pf_keth_ruleq *rules;
870 	struct pf_keth_ruleset *rs;
871 
872 	rs = pf_find_keth_ruleset(anchor);
873 	if (rs == NULL) {
874 		return (EINVAL);
875 	}
876 
877 	if (!rs->inactive.open ||
878 	    ticket != rs->inactive.ticket)
879 		return (EBUSY);
880 
881 	PF_RULES_WASSERT();
882 
883 	pf_eth_calc_skip_steps(rs->inactive.rules);
884 
885 	rules = rs->active.rules;
886 	ck_pr_store_ptr(&rs->active.rules, rs->inactive.rules);
887 	rs->inactive.rules = rules;
888 	rs->inactive.ticket = rs->active.ticket;
889 
890 	/* Clean up inactive rules (i.e. previously active rules), only when
891 	 * we're sure they're no longer used. */
892 	NET_EPOCH_CALL(pf_rollback_eth_cb, &rs->epoch_ctx);
893 
894 	return (0);
895 }
896 
897 #ifdef ALTQ
898 static uint16_t
899 pf_qname2qid(const char *qname)
900 {
901 	return (tagname2tag(&V_pf_qids, qname));
902 }
903 
904 static void
905 pf_qid_unref(uint16_t qid)
906 {
907 	tag_unref(&V_pf_qids, qid);
908 }
909 
910 static int
911 pf_begin_altq(u_int32_t *ticket)
912 {
913 	struct pf_altq	*altq, *tmp;
914 	int		 error = 0;
915 
916 	PF_RULES_WASSERT();
917 
918 	/* Purge the old altq lists */
919 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
920 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
921 			/* detach and destroy the discipline */
922 			error = altq_remove(altq);
923 		}
924 		free(altq, M_PFALTQ);
925 	}
926 	TAILQ_INIT(V_pf_altq_ifs_inactive);
927 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
928 		pf_qid_unref(altq->qid);
929 		free(altq, M_PFALTQ);
930 	}
931 	TAILQ_INIT(V_pf_altqs_inactive);
932 	if (error)
933 		return (error);
934 	*ticket = ++V_ticket_altqs_inactive;
935 	V_altqs_inactive_open = 1;
936 	return (0);
937 }
938 
939 static int
940 pf_rollback_altq(u_int32_t ticket)
941 {
942 	struct pf_altq	*altq, *tmp;
943 	int		 error = 0;
944 
945 	PF_RULES_WASSERT();
946 
947 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
948 		return (0);
949 	/* Purge the old altq lists */
950 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
951 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
952 			/* detach and destroy the discipline */
953 			error = altq_remove(altq);
954 		}
955 		free(altq, M_PFALTQ);
956 	}
957 	TAILQ_INIT(V_pf_altq_ifs_inactive);
958 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
959 		pf_qid_unref(altq->qid);
960 		free(altq, M_PFALTQ);
961 	}
962 	TAILQ_INIT(V_pf_altqs_inactive);
963 	V_altqs_inactive_open = 0;
964 	return (error);
965 }
966 
967 static int
968 pf_commit_altq(u_int32_t ticket)
969 {
970 	struct pf_altqqueue	*old_altqs, *old_altq_ifs;
971 	struct pf_altq		*altq, *tmp;
972 	int			 err, error = 0;
973 
974 	PF_RULES_WASSERT();
975 
976 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
977 		return (EBUSY);
978 
979 	/* swap altqs, keep the old. */
980 	old_altqs = V_pf_altqs_active;
981 	old_altq_ifs = V_pf_altq_ifs_active;
982 	V_pf_altqs_active = V_pf_altqs_inactive;
983 	V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
984 	V_pf_altqs_inactive = old_altqs;
985 	V_pf_altq_ifs_inactive = old_altq_ifs;
986 	V_ticket_altqs_active = V_ticket_altqs_inactive;
987 
988 	/* Attach new disciplines */
989 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
990 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
991 			/* attach the discipline */
992 			error = altq_pfattach(altq);
993 			if (error == 0 && V_pf_altq_running)
994 				error = pf_enable_altq(altq);
995 			if (error != 0)
996 				return (error);
997 		}
998 	}
999 
1000 	/* Purge the old altq lists */
1001 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
1002 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
1003 			/* detach and destroy the discipline */
1004 			if (V_pf_altq_running)
1005 				error = pf_disable_altq(altq);
1006 			err = altq_pfdetach(altq);
1007 			if (err != 0 && error == 0)
1008 				error = err;
1009 			err = altq_remove(altq);
1010 			if (err != 0 && error == 0)
1011 				error = err;
1012 		}
1013 		free(altq, M_PFALTQ);
1014 	}
1015 	TAILQ_INIT(V_pf_altq_ifs_inactive);
1016 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
1017 		pf_qid_unref(altq->qid);
1018 		free(altq, M_PFALTQ);
1019 	}
1020 	TAILQ_INIT(V_pf_altqs_inactive);
1021 
1022 	V_altqs_inactive_open = 0;
1023 	return (error);
1024 }
1025 
1026 static int
1027 pf_enable_altq(struct pf_altq *altq)
1028 {
1029 	struct ifnet		*ifp;
1030 	struct tb_profile	 tb;
1031 	int			 error = 0;
1032 
1033 	if ((ifp = ifunit(altq->ifname)) == NULL)
1034 		return (EINVAL);
1035 
1036 	if (ifp->if_snd.altq_type != ALTQT_NONE)
1037 		error = altq_enable(&ifp->if_snd);
1038 
1039 	/* set tokenbucket regulator */
1040 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
1041 		tb.rate = altq->ifbandwidth;
1042 		tb.depth = altq->tbrsize;
1043 		error = tbr_set(&ifp->if_snd, &tb);
1044 	}
1045 
1046 	return (error);
1047 }
1048 
1049 static int
1050 pf_disable_altq(struct pf_altq *altq)
1051 {
1052 	struct ifnet		*ifp;
1053 	struct tb_profile	 tb;
1054 	int			 error;
1055 
1056 	if ((ifp = ifunit(altq->ifname)) == NULL)
1057 		return (EINVAL);
1058 
1059 	/*
1060 	 * when the discipline is no longer referenced, it was overridden
1061 	 * by a new one.  if so, just return.
1062 	 */
1063 	if (altq->altq_disc != ifp->if_snd.altq_disc)
1064 		return (0);
1065 
1066 	error = altq_disable(&ifp->if_snd);
1067 
1068 	if (error == 0) {
1069 		/* clear tokenbucket regulator */
1070 		tb.rate = 0;
1071 		error = tbr_set(&ifp->if_snd, &tb);
1072 	}
1073 
1074 	return (error);
1075 }
1076 
1077 static int
1078 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
1079     struct pf_altq *altq)
1080 {
1081 	struct ifnet	*ifp1;
1082 	int		 error = 0;
1083 
1084 	/* Deactivate the interface in question */
1085 	altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
1086 	if ((ifp1 = ifunit(altq->ifname)) == NULL ||
1087 	    (remove && ifp1 == ifp)) {
1088 		altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
1089 	} else {
1090 		error = altq_add(ifp1, altq);
1091 
1092 		if (ticket != V_ticket_altqs_inactive)
1093 			error = EBUSY;
1094 
1095 		if (error)
1096 			free(altq, M_PFALTQ);
1097 	}
1098 
1099 	return (error);
1100 }
1101 
1102 void
1103 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
1104 {
1105 	struct pf_altq	*a1, *a2, *a3;
1106 	u_int32_t	 ticket;
1107 	int		 error = 0;
1108 
1109 	/*
1110 	 * No need to re-evaluate the configuration for events on interfaces
1111 	 * that do not support ALTQ, as it's not possible for such
1112 	 * interfaces to be part of the configuration.
1113 	 */
1114 	if (!ALTQ_IS_READY(&ifp->if_snd))
1115 		return;
1116 
1117 	/* Interrupt userland queue modifications */
1118 	if (V_altqs_inactive_open)
1119 		pf_rollback_altq(V_ticket_altqs_inactive);
1120 
1121 	/* Start new altq ruleset */
1122 	if (pf_begin_altq(&ticket))
1123 		return;
1124 
1125 	/* Copy the current active set */
1126 	TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
1127 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1128 		if (a2 == NULL) {
1129 			error = ENOMEM;
1130 			break;
1131 		}
1132 		bcopy(a1, a2, sizeof(struct pf_altq));
1133 
1134 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1135 		if (error)
1136 			break;
1137 
1138 		TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
1139 	}
1140 	if (error)
1141 		goto out;
1142 	TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
1143 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1144 		if (a2 == NULL) {
1145 			error = ENOMEM;
1146 			break;
1147 		}
1148 		bcopy(a1, a2, sizeof(struct pf_altq));
1149 
1150 		if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
1151 			error = EBUSY;
1152 			free(a2, M_PFALTQ);
1153 			break;
1154 		}
1155 		a2->altq_disc = NULL;
1156 		TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
1157 			if (strncmp(a3->ifname, a2->ifname,
1158 				IFNAMSIZ) == 0) {
1159 				a2->altq_disc = a3->altq_disc;
1160 				break;
1161 			}
1162 		}
1163 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1164 		if (error)
1165 			break;
1166 
1167 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
1168 	}
1169 
1170 out:
1171 	if (error != 0)
1172 		pf_rollback_altq(ticket);
1173 	else
1174 		pf_commit_altq(ticket);
1175 }
1176 #endif /* ALTQ */
1177 
1178 static struct pf_krule_global *
1179 pf_rule_tree_alloc(int flags)
1180 {
1181 	struct pf_krule_global *tree;
1182 
1183 	tree = malloc(sizeof(struct pf_krule_global), M_TEMP, flags);
1184 	if (tree == NULL)
1185 		return (NULL);
1186 	RB_INIT(tree);
1187 	return (tree);
1188 }
1189 
1190 static void
1191 pf_rule_tree_free(struct pf_krule_global *tree)
1192 {
1193 
1194 	free(tree, M_TEMP);
1195 }
1196 
1197 static int
1198 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
1199 {
1200 	struct pf_krule_global *tree;
1201 	struct pf_kruleset	*rs;
1202 	struct pf_krule		*rule;
1203 
1204 	PF_RULES_WASSERT();
1205 
1206 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1207 		return (EINVAL);
1208 	tree = pf_rule_tree_alloc(M_NOWAIT);
1209 	if (tree == NULL)
1210 		return (ENOMEM);
1211 	rs = pf_find_or_create_kruleset(anchor);
1212 	if (rs == NULL) {
1213 		free(tree, M_TEMP);
1214 		return (EINVAL);
1215 	}
1216 	pf_rule_tree_free(rs->rules[rs_num].inactive.tree);
1217 	rs->rules[rs_num].inactive.tree = tree;
1218 
1219 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1220 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1221 		rs->rules[rs_num].inactive.rcount--;
1222 	}
1223 	*ticket = ++rs->rules[rs_num].inactive.ticket;
1224 	rs->rules[rs_num].inactive.open = 1;
1225 	return (0);
1226 }
1227 
1228 static int
1229 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
1230 {
1231 	struct pf_kruleset	*rs;
1232 	struct pf_krule		*rule;
1233 
1234 	PF_RULES_WASSERT();
1235 
1236 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1237 		return (EINVAL);
1238 	rs = pf_find_kruleset(anchor);
1239 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1240 	    rs->rules[rs_num].inactive.ticket != ticket)
1241 		return (0);
1242 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1243 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1244 		rs->rules[rs_num].inactive.rcount--;
1245 	}
1246 	rs->rules[rs_num].inactive.open = 0;
1247 	return (0);
1248 }
1249 
1250 #define PF_MD5_UPD(st, elm)						\
1251 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
1252 
1253 #define PF_MD5_UPD_STR(st, elm)						\
1254 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
1255 
1256 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
1257 		(stor) = htonl((st)->elm);				\
1258 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
1259 } while (0)
1260 
1261 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
1262 		(stor) = htons((st)->elm);				\
1263 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
1264 } while (0)
1265 
1266 static void
1267 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
1268 {
1269 	PF_MD5_UPD(pfr, addr.type);
1270 	switch (pfr->addr.type) {
1271 		case PF_ADDR_DYNIFTL:
1272 			PF_MD5_UPD(pfr, addr.v.ifname);
1273 			PF_MD5_UPD(pfr, addr.iflags);
1274 			break;
1275 		case PF_ADDR_TABLE:
1276 			PF_MD5_UPD(pfr, addr.v.tblname);
1277 			break;
1278 		case PF_ADDR_ADDRMASK:
1279 			/* XXX ignore af? */
1280 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
1281 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
1282 			break;
1283 	}
1284 
1285 	PF_MD5_UPD(pfr, port[0]);
1286 	PF_MD5_UPD(pfr, port[1]);
1287 	PF_MD5_UPD(pfr, neg);
1288 	PF_MD5_UPD(pfr, port_op);
1289 }
1290 
1291 static void
1292 pf_hash_rule_rolling(MD5_CTX *ctx, struct pf_krule *rule)
1293 {
1294 	u_int16_t x;
1295 	u_int32_t y;
1296 
1297 	pf_hash_rule_addr(ctx, &rule->src);
1298 	pf_hash_rule_addr(ctx, &rule->dst);
1299 	for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++)
1300 		PF_MD5_UPD_STR(rule, label[i]);
1301 	PF_MD5_UPD_STR(rule, ifname);
1302 	PF_MD5_UPD_STR(rule, match_tagname);
1303 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
1304 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
1305 	PF_MD5_UPD_HTONL(rule, prob, y);
1306 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
1307 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1308 	PF_MD5_UPD(rule, uid.op);
1309 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1310 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1311 	PF_MD5_UPD(rule, gid.op);
1312 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
1313 	PF_MD5_UPD(rule, action);
1314 	PF_MD5_UPD(rule, direction);
1315 	PF_MD5_UPD(rule, af);
1316 	PF_MD5_UPD(rule, quick);
1317 	PF_MD5_UPD(rule, ifnot);
1318 	PF_MD5_UPD(rule, match_tag_not);
1319 	PF_MD5_UPD(rule, natpass);
1320 	PF_MD5_UPD(rule, keep_state);
1321 	PF_MD5_UPD(rule, proto);
1322 	PF_MD5_UPD(rule, type);
1323 	PF_MD5_UPD(rule, code);
1324 	PF_MD5_UPD(rule, flags);
1325 	PF_MD5_UPD(rule, flagset);
1326 	PF_MD5_UPD(rule, allow_opts);
1327 	PF_MD5_UPD(rule, rt);
1328 	PF_MD5_UPD(rule, tos);
1329 	PF_MD5_UPD(rule, scrub_flags);
1330 	PF_MD5_UPD(rule, min_ttl);
1331 	PF_MD5_UPD(rule, set_tos);
1332 	if (rule->anchor != NULL)
1333 		PF_MD5_UPD_STR(rule, anchor->path);
1334 }
1335 
1336 static void
1337 pf_hash_rule(struct pf_krule *rule)
1338 {
1339 	MD5_CTX		ctx;
1340 
1341 	MD5Init(&ctx);
1342 	pf_hash_rule_rolling(&ctx, rule);
1343 	MD5Final(rule->md5sum, &ctx);
1344 }
1345 
1346 static int
1347 pf_krule_compare(struct pf_krule *a, struct pf_krule *b)
1348 {
1349 
1350 	return (memcmp(a->md5sum, b->md5sum, PF_MD5_DIGEST_LENGTH));
1351 }
1352 
1353 static int
1354 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1355 {
1356 	struct pf_kruleset	*rs;
1357 	struct pf_krule		*rule, **old_array, *old_rule;
1358 	struct pf_krulequeue	*old_rules;
1359 	struct pf_krule_global  *old_tree;
1360 	int			 error;
1361 	u_int32_t		 old_rcount;
1362 
1363 	PF_RULES_WASSERT();
1364 
1365 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1366 		return (EINVAL);
1367 	rs = pf_find_kruleset(anchor);
1368 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1369 	    ticket != rs->rules[rs_num].inactive.ticket)
1370 		return (EBUSY);
1371 
1372 	/* Calculate checksum for the main ruleset */
1373 	if (rs == &pf_main_ruleset) {
1374 		error = pf_setup_pfsync_matching(rs);
1375 		if (error != 0)
1376 			return (error);
1377 	}
1378 
1379 	/* Swap rules, keep the old. */
1380 	old_rules = rs->rules[rs_num].active.ptr;
1381 	old_rcount = rs->rules[rs_num].active.rcount;
1382 	old_array = rs->rules[rs_num].active.ptr_array;
1383 	old_tree = rs->rules[rs_num].active.tree;
1384 
1385 	rs->rules[rs_num].active.ptr =
1386 	    rs->rules[rs_num].inactive.ptr;
1387 	rs->rules[rs_num].active.ptr_array =
1388 	    rs->rules[rs_num].inactive.ptr_array;
1389 	rs->rules[rs_num].active.tree =
1390 	    rs->rules[rs_num].inactive.tree;
1391 	rs->rules[rs_num].active.rcount =
1392 	    rs->rules[rs_num].inactive.rcount;
1393 
1394 	/* Attempt to preserve counter information. */
1395 	if (V_pf_status.keep_counters && old_tree != NULL) {
1396 		TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr,
1397 		    entries) {
1398 			old_rule = RB_FIND(pf_krule_global, old_tree, rule);
1399 			if (old_rule == NULL) {
1400 				continue;
1401 			}
1402 			pf_counter_u64_critical_enter();
1403 			pf_counter_u64_add_protected(&rule->evaluations,
1404 			    pf_counter_u64_fetch(&old_rule->evaluations));
1405 			pf_counter_u64_add_protected(&rule->packets[0],
1406 			    pf_counter_u64_fetch(&old_rule->packets[0]));
1407 			pf_counter_u64_add_protected(&rule->packets[1],
1408 			    pf_counter_u64_fetch(&old_rule->packets[1]));
1409 			pf_counter_u64_add_protected(&rule->bytes[0],
1410 			    pf_counter_u64_fetch(&old_rule->bytes[0]));
1411 			pf_counter_u64_add_protected(&rule->bytes[1],
1412 			    pf_counter_u64_fetch(&old_rule->bytes[1]));
1413 			pf_counter_u64_critical_exit();
1414 		}
1415 	}
1416 
1417 	rs->rules[rs_num].inactive.ptr = old_rules;
1418 	rs->rules[rs_num].inactive.ptr_array = old_array;
1419 	rs->rules[rs_num].inactive.tree = NULL; /* important for pf_ioctl_addrule */
1420 	rs->rules[rs_num].inactive.rcount = old_rcount;
1421 
1422 	rs->rules[rs_num].active.ticket =
1423 	    rs->rules[rs_num].inactive.ticket;
1424 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1425 
1426 	/* Purge the old rule list. */
1427 	PF_UNLNKDRULES_LOCK();
1428 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1429 		pf_unlink_rule_locked(old_rules, rule);
1430 	PF_UNLNKDRULES_UNLOCK();
1431 	if (rs->rules[rs_num].inactive.ptr_array)
1432 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1433 	rs->rules[rs_num].inactive.ptr_array = NULL;
1434 	rs->rules[rs_num].inactive.rcount = 0;
1435 	rs->rules[rs_num].inactive.open = 0;
1436 	pf_remove_if_empty_kruleset(rs);
1437 	free(old_tree, M_TEMP);
1438 
1439 	return (0);
1440 }
1441 
1442 static int
1443 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1444 {
1445 	MD5_CTX			 ctx;
1446 	struct pf_krule		*rule;
1447 	int			 rs_cnt;
1448 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
1449 
1450 	MD5Init(&ctx);
1451 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1452 		/* XXX PF_RULESET_SCRUB as well? */
1453 		if (rs_cnt == PF_RULESET_SCRUB)
1454 			continue;
1455 
1456 		if (rs->rules[rs_cnt].inactive.ptr_array)
1457 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1458 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
1459 
1460 		if (rs->rules[rs_cnt].inactive.rcount) {
1461 			rs->rules[rs_cnt].inactive.ptr_array =
1462 			    mallocarray(rs->rules[rs_cnt].inactive.rcount,
1463 			    sizeof(struct pf_rule **),
1464 			    M_TEMP, M_NOWAIT);
1465 
1466 			if (!rs->rules[rs_cnt].inactive.ptr_array)
1467 				return (ENOMEM);
1468 		}
1469 
1470 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1471 		    entries) {
1472 			pf_hash_rule_rolling(&ctx, rule);
1473 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1474 		}
1475 	}
1476 
1477 	MD5Final(digest, &ctx);
1478 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1479 	return (0);
1480 }
1481 
1482 static int
1483 pf_eth_addr_setup(struct pf_keth_ruleset *ruleset, struct pf_addr_wrap *addr)
1484 {
1485 	int error = 0;
1486 
1487 	switch (addr->type) {
1488 	case PF_ADDR_TABLE:
1489 		addr->p.tbl = pfr_eth_attach_table(ruleset, addr->v.tblname);
1490 		if (addr->p.tbl == NULL)
1491 			error = ENOMEM;
1492 		break;
1493 	default:
1494 		error = EINVAL;
1495 	}
1496 
1497 	return (error);
1498 }
1499 
1500 static int
1501 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1502     sa_family_t af)
1503 {
1504 	int error = 0;
1505 
1506 	switch (addr->type) {
1507 	case PF_ADDR_TABLE:
1508 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1509 		if (addr->p.tbl == NULL)
1510 			error = ENOMEM;
1511 		break;
1512 	case PF_ADDR_DYNIFTL:
1513 		error = pfi_dynaddr_setup(addr, af);
1514 		break;
1515 	}
1516 
1517 	return (error);
1518 }
1519 
1520 static void
1521 pf_addr_copyout(struct pf_addr_wrap *addr)
1522 {
1523 
1524 	switch (addr->type) {
1525 	case PF_ADDR_DYNIFTL:
1526 		pfi_dynaddr_copyout(addr);
1527 		break;
1528 	case PF_ADDR_TABLE:
1529 		pf_tbladdr_copyout(addr);
1530 		break;
1531 	}
1532 }
1533 
1534 static void
1535 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
1536 {
1537 	int	secs = time_uptime, diff;
1538 
1539 	bzero(out, sizeof(struct pf_src_node));
1540 
1541 	bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
1542 	bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
1543 
1544 	if (in->rule.ptr != NULL)
1545 		out->rule.nr = in->rule.ptr->nr;
1546 
1547 	for (int i = 0; i < 2; i++) {
1548 		out->bytes[i] = counter_u64_fetch(in->bytes[i]);
1549 		out->packets[i] = counter_u64_fetch(in->packets[i]);
1550 	}
1551 
1552 	out->states = in->states;
1553 	out->conn = in->conn;
1554 	out->af = in->af;
1555 	out->ruletype = in->ruletype;
1556 
1557 	out->creation = secs - in->creation;
1558 	if (out->expire > secs)
1559 		out->expire -= secs;
1560 	else
1561 		out->expire = 0;
1562 
1563 	/* Adjust the connection rate estimate. */
1564 	diff = secs - in->conn_rate.last;
1565 	if (diff >= in->conn_rate.seconds)
1566 		out->conn_rate.count = 0;
1567 	else
1568 		out->conn_rate.count -=
1569 		    in->conn_rate.count * diff /
1570 		    in->conn_rate.seconds;
1571 }
1572 
1573 #ifdef ALTQ
1574 /*
1575  * Handle export of struct pf_kaltq to user binaries that may be using any
1576  * version of struct pf_altq.
1577  */
1578 static int
1579 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1580 {
1581 	u_int32_t version;
1582 
1583 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1584 		version = 0;
1585 	else
1586 		version = pa->version;
1587 
1588 	if (version > PFIOC_ALTQ_VERSION)
1589 		return (EINVAL);
1590 
1591 #define ASSIGN(x) exported_q->x = q->x
1592 #define COPY(x) \
1593 	bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1594 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1595 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1596 
1597 	switch (version) {
1598 	case 0: {
1599 		struct pf_altq_v0 *exported_q =
1600 		    &((struct pfioc_altq_v0 *)pa)->altq;
1601 
1602 		COPY(ifname);
1603 
1604 		ASSIGN(scheduler);
1605 		ASSIGN(tbrsize);
1606 		exported_q->tbrsize = SATU16(q->tbrsize);
1607 		exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1608 
1609 		COPY(qname);
1610 		COPY(parent);
1611 		ASSIGN(parent_qid);
1612 		exported_q->bandwidth = SATU32(q->bandwidth);
1613 		ASSIGN(priority);
1614 		ASSIGN(local_flags);
1615 
1616 		ASSIGN(qlimit);
1617 		ASSIGN(flags);
1618 
1619 		if (q->scheduler == ALTQT_HFSC) {
1620 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1621 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1622 			    SATU32(q->pq_u.hfsc_opts.x)
1623 
1624 			ASSIGN_OPT_SATU32(rtsc_m1);
1625 			ASSIGN_OPT(rtsc_d);
1626 			ASSIGN_OPT_SATU32(rtsc_m2);
1627 
1628 			ASSIGN_OPT_SATU32(lssc_m1);
1629 			ASSIGN_OPT(lssc_d);
1630 			ASSIGN_OPT_SATU32(lssc_m2);
1631 
1632 			ASSIGN_OPT_SATU32(ulsc_m1);
1633 			ASSIGN_OPT(ulsc_d);
1634 			ASSIGN_OPT_SATU32(ulsc_m2);
1635 
1636 			ASSIGN_OPT(flags);
1637 
1638 #undef ASSIGN_OPT
1639 #undef ASSIGN_OPT_SATU32
1640 		} else
1641 			COPY(pq_u);
1642 
1643 		ASSIGN(qid);
1644 		break;
1645 	}
1646 	case 1:	{
1647 		struct pf_altq_v1 *exported_q =
1648 		    &((struct pfioc_altq_v1 *)pa)->altq;
1649 
1650 		COPY(ifname);
1651 
1652 		ASSIGN(scheduler);
1653 		ASSIGN(tbrsize);
1654 		ASSIGN(ifbandwidth);
1655 
1656 		COPY(qname);
1657 		COPY(parent);
1658 		ASSIGN(parent_qid);
1659 		ASSIGN(bandwidth);
1660 		ASSIGN(priority);
1661 		ASSIGN(local_flags);
1662 
1663 		ASSIGN(qlimit);
1664 		ASSIGN(flags);
1665 		COPY(pq_u);
1666 
1667 		ASSIGN(qid);
1668 		break;
1669 	}
1670 	default:
1671 		panic("%s: unhandled struct pfioc_altq version", __func__);
1672 		break;
1673 	}
1674 
1675 #undef ASSIGN
1676 #undef COPY
1677 #undef SATU16
1678 #undef SATU32
1679 
1680 	return (0);
1681 }
1682 
1683 /*
1684  * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1685  * that may be using any version of it.
1686  */
1687 static int
1688 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1689 {
1690 	u_int32_t version;
1691 
1692 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1693 		version = 0;
1694 	else
1695 		version = pa->version;
1696 
1697 	if (version > PFIOC_ALTQ_VERSION)
1698 		return (EINVAL);
1699 
1700 #define ASSIGN(x) q->x = imported_q->x
1701 #define COPY(x) \
1702 	bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1703 
1704 	switch (version) {
1705 	case 0: {
1706 		struct pf_altq_v0 *imported_q =
1707 		    &((struct pfioc_altq_v0 *)pa)->altq;
1708 
1709 		COPY(ifname);
1710 
1711 		ASSIGN(scheduler);
1712 		ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1713 		ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1714 
1715 		COPY(qname);
1716 		COPY(parent);
1717 		ASSIGN(parent_qid);
1718 		ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1719 		ASSIGN(priority);
1720 		ASSIGN(local_flags);
1721 
1722 		ASSIGN(qlimit);
1723 		ASSIGN(flags);
1724 
1725 		if (imported_q->scheduler == ALTQT_HFSC) {
1726 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1727 
1728 			/*
1729 			 * The m1 and m2 parameters are being copied from
1730 			 * 32-bit to 64-bit.
1731 			 */
1732 			ASSIGN_OPT(rtsc_m1);
1733 			ASSIGN_OPT(rtsc_d);
1734 			ASSIGN_OPT(rtsc_m2);
1735 
1736 			ASSIGN_OPT(lssc_m1);
1737 			ASSIGN_OPT(lssc_d);
1738 			ASSIGN_OPT(lssc_m2);
1739 
1740 			ASSIGN_OPT(ulsc_m1);
1741 			ASSIGN_OPT(ulsc_d);
1742 			ASSIGN_OPT(ulsc_m2);
1743 
1744 			ASSIGN_OPT(flags);
1745 
1746 #undef ASSIGN_OPT
1747 		} else
1748 			COPY(pq_u);
1749 
1750 		ASSIGN(qid);
1751 		break;
1752 	}
1753 	case 1: {
1754 		struct pf_altq_v1 *imported_q =
1755 		    &((struct pfioc_altq_v1 *)pa)->altq;
1756 
1757 		COPY(ifname);
1758 
1759 		ASSIGN(scheduler);
1760 		ASSIGN(tbrsize);
1761 		ASSIGN(ifbandwidth);
1762 
1763 		COPY(qname);
1764 		COPY(parent);
1765 		ASSIGN(parent_qid);
1766 		ASSIGN(bandwidth);
1767 		ASSIGN(priority);
1768 		ASSIGN(local_flags);
1769 
1770 		ASSIGN(qlimit);
1771 		ASSIGN(flags);
1772 		COPY(pq_u);
1773 
1774 		ASSIGN(qid);
1775 		break;
1776 	}
1777 	default:
1778 		panic("%s: unhandled struct pfioc_altq version", __func__);
1779 		break;
1780 	}
1781 
1782 #undef ASSIGN
1783 #undef COPY
1784 
1785 	return (0);
1786 }
1787 
1788 static struct pf_altq *
1789 pf_altq_get_nth_active(u_int32_t n)
1790 {
1791 	struct pf_altq		*altq;
1792 	u_int32_t		 nr;
1793 
1794 	nr = 0;
1795 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1796 		if (nr == n)
1797 			return (altq);
1798 		nr++;
1799 	}
1800 
1801 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1802 		if (nr == n)
1803 			return (altq);
1804 		nr++;
1805 	}
1806 
1807 	return (NULL);
1808 }
1809 #endif /* ALTQ */
1810 
1811 struct pf_krule *
1812 pf_krule_alloc(void)
1813 {
1814 	struct pf_krule *rule;
1815 
1816 	rule = malloc(sizeof(struct pf_krule), M_PFRULE, M_WAITOK | M_ZERO);
1817 	mtx_init(&rule->rpool.mtx, "pf_krule_pool", NULL, MTX_DEF);
1818 	rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
1819 	    M_WAITOK | M_ZERO);
1820 	return (rule);
1821 }
1822 
1823 void
1824 pf_krule_free(struct pf_krule *rule)
1825 {
1826 #ifdef PF_WANT_32_TO_64_COUNTER
1827 	bool wowned;
1828 #endif
1829 
1830 	if (rule == NULL)
1831 		return;
1832 
1833 #ifdef PF_WANT_32_TO_64_COUNTER
1834 	if (rule->allrulelinked) {
1835 		wowned = PF_RULES_WOWNED();
1836 		if (!wowned)
1837 			PF_RULES_WLOCK();
1838 		LIST_REMOVE(rule, allrulelist);
1839 		V_pf_allrulecount--;
1840 		if (!wowned)
1841 			PF_RULES_WUNLOCK();
1842 	}
1843 #endif
1844 
1845 	pf_counter_u64_deinit(&rule->evaluations);
1846 	for (int i = 0; i < 2; i++) {
1847 		pf_counter_u64_deinit(&rule->packets[i]);
1848 		pf_counter_u64_deinit(&rule->bytes[i]);
1849 	}
1850 	counter_u64_free(rule->states_cur);
1851 	counter_u64_free(rule->states_tot);
1852 	counter_u64_free(rule->src_nodes);
1853 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
1854 
1855 	mtx_destroy(&rule->rpool.mtx);
1856 	free(rule, M_PFRULE);
1857 }
1858 
1859 static void
1860 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool,
1861     struct pf_pooladdr *pool)
1862 {
1863 
1864 	bzero(pool, sizeof(*pool));
1865 	bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr));
1866 	strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname));
1867 }
1868 
1869 static int
1870 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool,
1871     struct pf_kpooladdr *kpool)
1872 {
1873 	int ret;
1874 
1875 	bzero(kpool, sizeof(*kpool));
1876 	bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr));
1877 	ret = pf_user_strcpy(kpool->ifname, pool->ifname,
1878 	    sizeof(kpool->ifname));
1879 	return (ret);
1880 }
1881 
1882 static void
1883 pf_kpool_to_pool(const struct pf_kpool *kpool, struct pf_pool *pool)
1884 {
1885 	bzero(pool, sizeof(*pool));
1886 
1887 	bcopy(&kpool->key, &pool->key, sizeof(pool->key));
1888 	bcopy(&kpool->counter, &pool->counter, sizeof(pool->counter));
1889 
1890 	pool->tblidx = kpool->tblidx;
1891 	pool->proxy_port[0] = kpool->proxy_port[0];
1892 	pool->proxy_port[1] = kpool->proxy_port[1];
1893 	pool->opts = kpool->opts;
1894 }
1895 
1896 static void
1897 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool)
1898 {
1899 	_Static_assert(sizeof(pool->key) == sizeof(kpool->key), "");
1900 	_Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), "");
1901 
1902 	bcopy(&pool->key, &kpool->key, sizeof(kpool->key));
1903 	bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter));
1904 
1905 	kpool->tblidx = pool->tblidx;
1906 	kpool->proxy_port[0] = pool->proxy_port[0];
1907 	kpool->proxy_port[1] = pool->proxy_port[1];
1908 	kpool->opts = pool->opts;
1909 }
1910 
1911 static void
1912 pf_krule_to_rule(const struct pf_krule *krule, struct pf_rule *rule)
1913 {
1914 
1915 	bzero(rule, sizeof(*rule));
1916 
1917 	bcopy(&krule->src, &rule->src, sizeof(rule->src));
1918 	bcopy(&krule->dst, &rule->dst, sizeof(rule->dst));
1919 
1920 	for (int i = 0; i < PF_SKIP_COUNT; ++i) {
1921 		if (rule->skip[i].ptr == NULL)
1922 			rule->skip[i].nr = -1;
1923 		else
1924 			rule->skip[i].nr = krule->skip[i].ptr->nr;
1925 	}
1926 
1927 	strlcpy(rule->label, krule->label[0], sizeof(rule->label));
1928 	strlcpy(rule->ifname, krule->ifname, sizeof(rule->ifname));
1929 	strlcpy(rule->qname, krule->qname, sizeof(rule->qname));
1930 	strlcpy(rule->pqname, krule->pqname, sizeof(rule->pqname));
1931 	strlcpy(rule->tagname, krule->tagname, sizeof(rule->tagname));
1932 	strlcpy(rule->match_tagname, krule->match_tagname,
1933 	    sizeof(rule->match_tagname));
1934 	strlcpy(rule->overload_tblname, krule->overload_tblname,
1935 	    sizeof(rule->overload_tblname));
1936 
1937 	pf_kpool_to_pool(&krule->rpool, &rule->rpool);
1938 
1939 	rule->evaluations = pf_counter_u64_fetch(&krule->evaluations);
1940 	for (int i = 0; i < 2; i++) {
1941 		rule->packets[i] = pf_counter_u64_fetch(&krule->packets[i]);
1942 		rule->bytes[i] = pf_counter_u64_fetch(&krule->bytes[i]);
1943 	}
1944 
1945 	/* kif, anchor, overload_tbl are not copied over. */
1946 
1947 	rule->os_fingerprint = krule->os_fingerprint;
1948 
1949 	rule->rtableid = krule->rtableid;
1950 	bcopy(krule->timeout, rule->timeout, sizeof(krule->timeout));
1951 	rule->max_states = krule->max_states;
1952 	rule->max_src_nodes = krule->max_src_nodes;
1953 	rule->max_src_states = krule->max_src_states;
1954 	rule->max_src_conn = krule->max_src_conn;
1955 	rule->max_src_conn_rate.limit = krule->max_src_conn_rate.limit;
1956 	rule->max_src_conn_rate.seconds = krule->max_src_conn_rate.seconds;
1957 	rule->qid = krule->qid;
1958 	rule->pqid = krule->pqid;
1959 	rule->nr = krule->nr;
1960 	rule->prob = krule->prob;
1961 	rule->cuid = krule->cuid;
1962 	rule->cpid = krule->cpid;
1963 
1964 	rule->return_icmp = krule->return_icmp;
1965 	rule->return_icmp6 = krule->return_icmp6;
1966 	rule->max_mss = krule->max_mss;
1967 	rule->tag = krule->tag;
1968 	rule->match_tag = krule->match_tag;
1969 	rule->scrub_flags = krule->scrub_flags;
1970 
1971 	bcopy(&krule->uid, &rule->uid, sizeof(krule->uid));
1972 	bcopy(&krule->gid, &rule->gid, sizeof(krule->gid));
1973 
1974 	rule->rule_flag = krule->rule_flag;
1975 	rule->action = krule->action;
1976 	rule->direction = krule->direction;
1977 	rule->log = krule->log;
1978 	rule->logif = krule->logif;
1979 	rule->quick = krule->quick;
1980 	rule->ifnot = krule->ifnot;
1981 	rule->match_tag_not = krule->match_tag_not;
1982 	rule->natpass = krule->natpass;
1983 
1984 	rule->keep_state = krule->keep_state;
1985 	rule->af = krule->af;
1986 	rule->proto = krule->proto;
1987 	rule->type = krule->type;
1988 	rule->code = krule->code;
1989 	rule->flags = krule->flags;
1990 	rule->flagset = krule->flagset;
1991 	rule->min_ttl = krule->min_ttl;
1992 	rule->allow_opts = krule->allow_opts;
1993 	rule->rt = krule->rt;
1994 	rule->return_ttl = krule->return_ttl;
1995 	rule->tos = krule->tos;
1996 	rule->set_tos = krule->set_tos;
1997 	rule->anchor_relative = krule->anchor_relative;
1998 	rule->anchor_wildcard = krule->anchor_wildcard;
1999 
2000 	rule->flush = krule->flush;
2001 	rule->prio = krule->prio;
2002 	rule->set_prio[0] = krule->set_prio[0];
2003 	rule->set_prio[1] = krule->set_prio[1];
2004 
2005 	bcopy(&krule->divert, &rule->divert, sizeof(krule->divert));
2006 
2007 	rule->u_states_cur = counter_u64_fetch(krule->states_cur);
2008 	rule->u_states_tot = counter_u64_fetch(krule->states_tot);
2009 	rule->u_src_nodes = counter_u64_fetch(krule->src_nodes);
2010 }
2011 
2012 static int
2013 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
2014 {
2015 	int ret;
2016 
2017 #ifndef INET
2018 	if (rule->af == AF_INET) {
2019 		return (EAFNOSUPPORT);
2020 	}
2021 #endif /* INET */
2022 #ifndef INET6
2023 	if (rule->af == AF_INET6) {
2024 		return (EAFNOSUPPORT);
2025 	}
2026 #endif /* INET6 */
2027 
2028 	ret = pf_check_rule_addr(&rule->src);
2029 	if (ret != 0)
2030 		return (ret);
2031 	ret = pf_check_rule_addr(&rule->dst);
2032 	if (ret != 0)
2033 		return (ret);
2034 
2035 	bcopy(&rule->src, &krule->src, sizeof(rule->src));
2036 	bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
2037 
2038 	ret = pf_user_strcpy(krule->label[0], rule->label, sizeof(rule->label));
2039 	if (ret != 0)
2040 		return (ret);
2041 	ret = pf_user_strcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
2042 	if (ret != 0)
2043 		return (ret);
2044 	ret = pf_user_strcpy(krule->qname, rule->qname, sizeof(rule->qname));
2045 	if (ret != 0)
2046 		return (ret);
2047 	ret = pf_user_strcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
2048 	if (ret != 0)
2049 		return (ret);
2050 	ret = pf_user_strcpy(krule->tagname, rule->tagname,
2051 	    sizeof(rule->tagname));
2052 	if (ret != 0)
2053 		return (ret);
2054 	ret = pf_user_strcpy(krule->match_tagname, rule->match_tagname,
2055 	    sizeof(rule->match_tagname));
2056 	if (ret != 0)
2057 		return (ret);
2058 	ret = pf_user_strcpy(krule->overload_tblname, rule->overload_tblname,
2059 	    sizeof(rule->overload_tblname));
2060 	if (ret != 0)
2061 		return (ret);
2062 
2063 	pf_pool_to_kpool(&rule->rpool, &krule->rpool);
2064 
2065 	/* Don't allow userspace to set evaluations, packets or bytes. */
2066 	/* kif, anchor, overload_tbl are not copied over. */
2067 
2068 	krule->os_fingerprint = rule->os_fingerprint;
2069 
2070 	krule->rtableid = rule->rtableid;
2071 	bcopy(rule->timeout, krule->timeout, sizeof(krule->timeout));
2072 	krule->max_states = rule->max_states;
2073 	krule->max_src_nodes = rule->max_src_nodes;
2074 	krule->max_src_states = rule->max_src_states;
2075 	krule->max_src_conn = rule->max_src_conn;
2076 	krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
2077 	krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
2078 	krule->qid = rule->qid;
2079 	krule->pqid = rule->pqid;
2080 	krule->nr = rule->nr;
2081 	krule->prob = rule->prob;
2082 	krule->cuid = rule->cuid;
2083 	krule->cpid = rule->cpid;
2084 
2085 	krule->return_icmp = rule->return_icmp;
2086 	krule->return_icmp6 = rule->return_icmp6;
2087 	krule->max_mss = rule->max_mss;
2088 	krule->tag = rule->tag;
2089 	krule->match_tag = rule->match_tag;
2090 	krule->scrub_flags = rule->scrub_flags;
2091 
2092 	bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
2093 	bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
2094 
2095 	krule->rule_flag = rule->rule_flag;
2096 	krule->action = rule->action;
2097 	krule->direction = rule->direction;
2098 	krule->log = rule->log;
2099 	krule->logif = rule->logif;
2100 	krule->quick = rule->quick;
2101 	krule->ifnot = rule->ifnot;
2102 	krule->match_tag_not = rule->match_tag_not;
2103 	krule->natpass = rule->natpass;
2104 
2105 	krule->keep_state = rule->keep_state;
2106 	krule->af = rule->af;
2107 	krule->proto = rule->proto;
2108 	krule->type = rule->type;
2109 	krule->code = rule->code;
2110 	krule->flags = rule->flags;
2111 	krule->flagset = rule->flagset;
2112 	krule->min_ttl = rule->min_ttl;
2113 	krule->allow_opts = rule->allow_opts;
2114 	krule->rt = rule->rt;
2115 	krule->return_ttl = rule->return_ttl;
2116 	krule->tos = rule->tos;
2117 	krule->set_tos = rule->set_tos;
2118 
2119 	krule->flush = rule->flush;
2120 	krule->prio = rule->prio;
2121 	krule->set_prio[0] = rule->set_prio[0];
2122 	krule->set_prio[1] = rule->set_prio[1];
2123 
2124 	bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
2125 
2126 	return (0);
2127 }
2128 
2129 static int
2130 pf_state_kill_to_kstate_kill(const struct pfioc_state_kill *psk,
2131     struct pf_kstate_kill *kill)
2132 {
2133 	int ret;
2134 
2135 	bzero(kill, sizeof(*kill));
2136 
2137 	bcopy(&psk->psk_pfcmp, &kill->psk_pfcmp, sizeof(kill->psk_pfcmp));
2138 	kill->psk_af = psk->psk_af;
2139 	kill->psk_proto = psk->psk_proto;
2140 	bcopy(&psk->psk_src, &kill->psk_src, sizeof(kill->psk_src));
2141 	bcopy(&psk->psk_dst, &kill->psk_dst, sizeof(kill->psk_dst));
2142 	ret = pf_user_strcpy(kill->psk_ifname, psk->psk_ifname,
2143 	    sizeof(kill->psk_ifname));
2144 	if (ret != 0)
2145 		return (ret);
2146 	ret = pf_user_strcpy(kill->psk_label, psk->psk_label,
2147 	    sizeof(kill->psk_label));
2148 	if (ret != 0)
2149 		return (ret);
2150 
2151 	return (0);
2152 }
2153 
2154 static int
2155 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket,
2156     uint32_t pool_ticket, const char *anchor, const char *anchor_call,
2157     struct thread *td)
2158 {
2159 	struct pf_kruleset	*ruleset;
2160 	struct pf_krule		*tail;
2161 	struct pf_kpooladdr	*pa;
2162 	struct pfi_kkif		*kif = NULL;
2163 	int			 rs_num;
2164 	int			 error = 0;
2165 
2166 	if ((rule->return_icmp >> 8) > ICMP_MAXTYPE) {
2167 		error = EINVAL;
2168 		goto errout_unlocked;
2169 	}
2170 
2171 #define	ERROUT(x)	ERROUT_FUNCTION(errout, x)
2172 
2173 	if (rule->ifname[0])
2174 		kif = pf_kkif_create(M_WAITOK);
2175 	pf_counter_u64_init(&rule->evaluations, M_WAITOK);
2176 	for (int i = 0; i < 2; i++) {
2177 		pf_counter_u64_init(&rule->packets[i], M_WAITOK);
2178 		pf_counter_u64_init(&rule->bytes[i], M_WAITOK);
2179 	}
2180 	rule->states_cur = counter_u64_alloc(M_WAITOK);
2181 	rule->states_tot = counter_u64_alloc(M_WAITOK);
2182 	rule->src_nodes = counter_u64_alloc(M_WAITOK);
2183 	rule->cuid = td->td_ucred->cr_ruid;
2184 	rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
2185 	TAILQ_INIT(&rule->rpool.list);
2186 
2187 	PF_CONFIG_LOCK();
2188 	PF_RULES_WLOCK();
2189 #ifdef PF_WANT_32_TO_64_COUNTER
2190 	LIST_INSERT_HEAD(&V_pf_allrulelist, rule, allrulelist);
2191 	MPASS(!rule->allrulelinked);
2192 	rule->allrulelinked = true;
2193 	V_pf_allrulecount++;
2194 #endif
2195 	ruleset = pf_find_kruleset(anchor);
2196 	if (ruleset == NULL)
2197 		ERROUT(EINVAL);
2198 	rs_num = pf_get_ruleset_number(rule->action);
2199 	if (rs_num >= PF_RULESET_MAX)
2200 		ERROUT(EINVAL);
2201 	if (ticket != ruleset->rules[rs_num].inactive.ticket) {
2202 		DPFPRINTF(PF_DEBUG_MISC,
2203 		    ("ticket: %d != [%d]%d\n", ticket, rs_num,
2204 		    ruleset->rules[rs_num].inactive.ticket));
2205 		ERROUT(EBUSY);
2206 	}
2207 	if (pool_ticket != V_ticket_pabuf) {
2208 		DPFPRINTF(PF_DEBUG_MISC,
2209 		    ("pool_ticket: %d != %d\n", pool_ticket,
2210 		    V_ticket_pabuf));
2211 		ERROUT(EBUSY);
2212 	}
2213 	/*
2214 	 * XXXMJG hack: there is no mechanism to ensure they started the
2215 	 * transaction. Ticket checked above may happen to match by accident,
2216 	 * even if nobody called DIOCXBEGIN, let alone this process.
2217 	 * Partially work around it by checking if the RB tree got allocated,
2218 	 * see pf_begin_rules.
2219 	 */
2220 	if (ruleset->rules[rs_num].inactive.tree == NULL) {
2221 		ERROUT(EINVAL);
2222 	}
2223 
2224 	tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
2225 	    pf_krulequeue);
2226 	if (tail)
2227 		rule->nr = tail->nr + 1;
2228 	else
2229 		rule->nr = 0;
2230 	if (rule->ifname[0]) {
2231 		rule->kif = pfi_kkif_attach(kif, rule->ifname);
2232 		kif = NULL;
2233 		pfi_kkif_ref(rule->kif);
2234 	} else
2235 		rule->kif = NULL;
2236 
2237 	if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
2238 		error = EBUSY;
2239 
2240 #ifdef ALTQ
2241 	/* set queue IDs */
2242 	if (rule->qname[0] != 0) {
2243 		if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2244 			error = EBUSY;
2245 		else if (rule->pqname[0] != 0) {
2246 			if ((rule->pqid =
2247 			    pf_qname2qid(rule->pqname)) == 0)
2248 				error = EBUSY;
2249 		} else
2250 			rule->pqid = rule->qid;
2251 	}
2252 #endif
2253 	if (rule->tagname[0])
2254 		if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2255 			error = EBUSY;
2256 	if (rule->match_tagname[0])
2257 		if ((rule->match_tag =
2258 		    pf_tagname2tag(rule->match_tagname)) == 0)
2259 			error = EBUSY;
2260 	if (rule->rt && !rule->direction)
2261 		error = EINVAL;
2262 	if (!rule->log)
2263 		rule->logif = 0;
2264 	if (rule->logif >= PFLOGIFS_MAX)
2265 		error = EINVAL;
2266 	if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
2267 		error = ENOMEM;
2268 	if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
2269 		error = ENOMEM;
2270 	if (pf_kanchor_setup(rule, ruleset, anchor_call))
2271 		error = EINVAL;
2272 	if (rule->scrub_flags & PFSTATE_SETPRIO &&
2273 	    (rule->set_prio[0] > PF_PRIO_MAX ||
2274 	    rule->set_prio[1] > PF_PRIO_MAX))
2275 		error = EINVAL;
2276 	TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
2277 		if (pa->addr.type == PF_ADDR_TABLE) {
2278 			pa->addr.p.tbl = pfr_attach_table(ruleset,
2279 			    pa->addr.v.tblname);
2280 			if (pa->addr.p.tbl == NULL)
2281 				error = ENOMEM;
2282 		}
2283 
2284 	rule->overload_tbl = NULL;
2285 	if (rule->overload_tblname[0]) {
2286 		if ((rule->overload_tbl = pfr_attach_table(ruleset,
2287 		    rule->overload_tblname)) == NULL)
2288 			error = EINVAL;
2289 		else
2290 			rule->overload_tbl->pfrkt_flags |=
2291 			    PFR_TFLAG_ACTIVE;
2292 	}
2293 
2294 	pf_mv_kpool(&V_pf_pabuf, &rule->rpool.list);
2295 	if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
2296 	    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
2297 	    (rule->rt > PF_NOPFROUTE)) &&
2298 	    (TAILQ_FIRST(&rule->rpool.list) == NULL))
2299 		error = EINVAL;
2300 
2301 	if (error) {
2302 		pf_free_rule(rule);
2303 		rule = NULL;
2304 		ERROUT(error);
2305 	}
2306 
2307 	rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
2308 	TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
2309 	    rule, entries);
2310 	ruleset->rules[rs_num].inactive.rcount++;
2311 
2312 	PF_RULES_WUNLOCK();
2313 	pf_hash_rule(rule);
2314 	if (RB_INSERT(pf_krule_global, ruleset->rules[rs_num].inactive.tree, rule) != NULL) {
2315 		PF_RULES_WLOCK();
2316 		TAILQ_REMOVE(ruleset->rules[rs_num].inactive.ptr, rule, entries);
2317 		ruleset->rules[rs_num].inactive.rcount--;
2318 		pf_free_rule(rule);
2319 		rule = NULL;
2320 		ERROUT(EEXIST);
2321 	}
2322 	PF_CONFIG_UNLOCK();
2323 
2324 	return (0);
2325 
2326 #undef ERROUT
2327 errout:
2328 	PF_RULES_WUNLOCK();
2329 	PF_CONFIG_UNLOCK();
2330 errout_unlocked:
2331 	pf_kkif_free(kif);
2332 	pf_krule_free(rule);
2333 	return (error);
2334 }
2335 
2336 static bool
2337 pf_label_match(const struct pf_krule *rule, const char *label)
2338 {
2339 	int i = 0;
2340 
2341 	while (*rule->label[i]) {
2342 		if (strcmp(rule->label[i], label) == 0)
2343 			return (true);
2344 		i++;
2345 	}
2346 
2347 	return (false);
2348 }
2349 
2350 static unsigned int
2351 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir)
2352 {
2353 	struct pf_kstate *s;
2354 	int more = 0;
2355 
2356 	s = pf_find_state_all(key, dir, &more);
2357 	if (s == NULL)
2358 		return (0);
2359 
2360 	if (more) {
2361 		PF_STATE_UNLOCK(s);
2362 		return (0);
2363 	}
2364 
2365 	pf_unlink_state(s);
2366 	return (1);
2367 }
2368 
2369 static int
2370 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih)
2371 {
2372 	struct pf_kstate	*s;
2373 	struct pf_state_key	*sk;
2374 	struct pf_addr		*srcaddr, *dstaddr;
2375 	struct pf_state_key_cmp	 match_key;
2376 	int			 idx, killed = 0;
2377 	unsigned int		 dir;
2378 	u_int16_t		 srcport, dstport;
2379 	struct pfi_kkif		*kif;
2380 
2381 relock_DIOCKILLSTATES:
2382 	PF_HASHROW_LOCK(ih);
2383 	LIST_FOREACH(s, &ih->states, entry) {
2384 		/* For floating states look at the original kif. */
2385 		kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
2386 
2387 		sk = s->key[PF_SK_WIRE];
2388 		if (s->direction == PF_OUT) {
2389 			srcaddr = &sk->addr[1];
2390 			dstaddr = &sk->addr[0];
2391 			srcport = sk->port[1];
2392 			dstport = sk->port[0];
2393 		} else {
2394 			srcaddr = &sk->addr[0];
2395 			dstaddr = &sk->addr[1];
2396 			srcport = sk->port[0];
2397 			dstport = sk->port[1];
2398 		}
2399 
2400 		if (psk->psk_af && sk->af != psk->psk_af)
2401 			continue;
2402 
2403 		if (psk->psk_proto && psk->psk_proto != sk->proto)
2404 			continue;
2405 
2406 		if (! PF_MATCHA(psk->psk_src.neg, &psk->psk_src.addr.v.a.addr,
2407 		    &psk->psk_src.addr.v.a.mask, srcaddr, sk->af))
2408 			continue;
2409 
2410 		if (! PF_MATCHA(psk->psk_dst.neg, &psk->psk_dst.addr.v.a.addr,
2411 		    &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af))
2412 			continue;
2413 
2414 		if (!  PF_MATCHA(psk->psk_rt_addr.neg,
2415 		    &psk->psk_rt_addr.addr.v.a.addr,
2416 		    &psk->psk_rt_addr.addr.v.a.mask,
2417 		    &s->rt_addr, sk->af))
2418 			continue;
2419 
2420 		if (psk->psk_src.port_op != 0 &&
2421 		    ! pf_match_port(psk->psk_src.port_op,
2422 		    psk->psk_src.port[0], psk->psk_src.port[1], srcport))
2423 			continue;
2424 
2425 		if (psk->psk_dst.port_op != 0 &&
2426 		    ! pf_match_port(psk->psk_dst.port_op,
2427 		    psk->psk_dst.port[0], psk->psk_dst.port[1], dstport))
2428 			continue;
2429 
2430 		if (psk->psk_label[0] &&
2431 		    ! pf_label_match(s->rule.ptr, psk->psk_label))
2432 			continue;
2433 
2434 		if (psk->psk_ifname[0] && strcmp(psk->psk_ifname,
2435 		    kif->pfik_name))
2436 			continue;
2437 
2438 		if (psk->psk_kill_match) {
2439 			/* Create the key to find matching states, with lock
2440 			 * held. */
2441 
2442 			bzero(&match_key, sizeof(match_key));
2443 
2444 			if (s->direction == PF_OUT) {
2445 				dir = PF_IN;
2446 				idx = PF_SK_STACK;
2447 			} else {
2448 				dir = PF_OUT;
2449 				idx = PF_SK_WIRE;
2450 			}
2451 
2452 			match_key.af = s->key[idx]->af;
2453 			match_key.proto = s->key[idx]->proto;
2454 			PF_ACPY(&match_key.addr[0],
2455 			    &s->key[idx]->addr[1], match_key.af);
2456 			match_key.port[0] = s->key[idx]->port[1];
2457 			PF_ACPY(&match_key.addr[1],
2458 			    &s->key[idx]->addr[0], match_key.af);
2459 			match_key.port[1] = s->key[idx]->port[0];
2460 		}
2461 
2462 		pf_unlink_state(s);
2463 		killed++;
2464 
2465 		if (psk->psk_kill_match)
2466 			killed += pf_kill_matching_state(&match_key, dir);
2467 
2468 		goto relock_DIOCKILLSTATES;
2469 	}
2470 	PF_HASHROW_UNLOCK(ih);
2471 
2472 	return (killed);
2473 }
2474 
2475 static int
2476 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
2477 {
2478 	int			 error = 0;
2479 	PF_RULES_RLOCK_TRACKER;
2480 
2481 #define	ERROUT_IOCTL(target, x)					\
2482     do {								\
2483 	    error = (x);						\
2484 	    SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__);	\
2485 	    goto target;						\
2486     } while (0)
2487 
2488 
2489 	/* XXX keep in sync with switch() below */
2490 	if (securelevel_gt(td->td_ucred, 2))
2491 		switch (cmd) {
2492 		case DIOCGETRULES:
2493 		case DIOCGETRULE:
2494 		case DIOCGETRULENV:
2495 		case DIOCGETADDRS:
2496 		case DIOCGETADDR:
2497 		case DIOCGETSTATE:
2498 		case DIOCGETSTATENV:
2499 		case DIOCSETSTATUSIF:
2500 		case DIOCGETSTATUS:
2501 		case DIOCGETSTATUSNV:
2502 		case DIOCCLRSTATUS:
2503 		case DIOCNATLOOK:
2504 		case DIOCSETDEBUG:
2505 		case DIOCGETSTATES:
2506 		case DIOCGETSTATESV2:
2507 		case DIOCGETTIMEOUT:
2508 		case DIOCCLRRULECTRS:
2509 		case DIOCGETLIMIT:
2510 		case DIOCGETALTQSV0:
2511 		case DIOCGETALTQSV1:
2512 		case DIOCGETALTQV0:
2513 		case DIOCGETALTQV1:
2514 		case DIOCGETQSTATSV0:
2515 		case DIOCGETQSTATSV1:
2516 		case DIOCGETRULESETS:
2517 		case DIOCGETRULESET:
2518 		case DIOCRGETTABLES:
2519 		case DIOCRGETTSTATS:
2520 		case DIOCRCLRTSTATS:
2521 		case DIOCRCLRADDRS:
2522 		case DIOCRADDADDRS:
2523 		case DIOCRDELADDRS:
2524 		case DIOCRSETADDRS:
2525 		case DIOCRGETADDRS:
2526 		case DIOCRGETASTATS:
2527 		case DIOCRCLRASTATS:
2528 		case DIOCRTSTADDRS:
2529 		case DIOCOSFPGET:
2530 		case DIOCGETSRCNODES:
2531 		case DIOCCLRSRCNODES:
2532 		case DIOCGETSYNCOOKIES:
2533 		case DIOCIGETIFACES:
2534 		case DIOCGIFSPEEDV0:
2535 		case DIOCGIFSPEEDV1:
2536 		case DIOCSETIFFLAG:
2537 		case DIOCCLRIFFLAG:
2538 		case DIOCGETETHRULES:
2539 		case DIOCGETETHRULE:
2540 		case DIOCGETETHRULESETS:
2541 		case DIOCGETETHRULESET:
2542 			break;
2543 		case DIOCRCLRTABLES:
2544 		case DIOCRADDTABLES:
2545 		case DIOCRDELTABLES:
2546 		case DIOCRSETTFLAGS:
2547 			if (((struct pfioc_table *)addr)->pfrio_flags &
2548 			    PFR_FLAG_DUMMY)
2549 				break; /* dummy operation ok */
2550 			return (EPERM);
2551 		default:
2552 			return (EPERM);
2553 		}
2554 
2555 	if (!(flags & FWRITE))
2556 		switch (cmd) {
2557 		case DIOCGETRULES:
2558 		case DIOCGETADDRS:
2559 		case DIOCGETADDR:
2560 		case DIOCGETSTATE:
2561 		case DIOCGETSTATENV:
2562 		case DIOCGETSTATUS:
2563 		case DIOCGETSTATUSNV:
2564 		case DIOCGETSTATES:
2565 		case DIOCGETSTATESV2:
2566 		case DIOCGETTIMEOUT:
2567 		case DIOCGETLIMIT:
2568 		case DIOCGETALTQSV0:
2569 		case DIOCGETALTQSV1:
2570 		case DIOCGETALTQV0:
2571 		case DIOCGETALTQV1:
2572 		case DIOCGETQSTATSV0:
2573 		case DIOCGETQSTATSV1:
2574 		case DIOCGETRULESETS:
2575 		case DIOCGETRULESET:
2576 		case DIOCNATLOOK:
2577 		case DIOCRGETTABLES:
2578 		case DIOCRGETTSTATS:
2579 		case DIOCRGETADDRS:
2580 		case DIOCRGETASTATS:
2581 		case DIOCRTSTADDRS:
2582 		case DIOCOSFPGET:
2583 		case DIOCGETSRCNODES:
2584 		case DIOCGETSYNCOOKIES:
2585 		case DIOCIGETIFACES:
2586 		case DIOCGIFSPEEDV1:
2587 		case DIOCGIFSPEEDV0:
2588 		case DIOCGETRULENV:
2589 		case DIOCGETETHRULES:
2590 		case DIOCGETETHRULE:
2591 		case DIOCGETETHRULESETS:
2592 		case DIOCGETETHRULESET:
2593 			break;
2594 		case DIOCRCLRTABLES:
2595 		case DIOCRADDTABLES:
2596 		case DIOCRDELTABLES:
2597 		case DIOCRCLRTSTATS:
2598 		case DIOCRCLRADDRS:
2599 		case DIOCRADDADDRS:
2600 		case DIOCRDELADDRS:
2601 		case DIOCRSETADDRS:
2602 		case DIOCRSETTFLAGS:
2603 			if (((struct pfioc_table *)addr)->pfrio_flags &
2604 			    PFR_FLAG_DUMMY) {
2605 				flags |= FWRITE; /* need write lock for dummy */
2606 				break; /* dummy operation ok */
2607 			}
2608 			return (EACCES);
2609 		case DIOCGETRULE:
2610 			if (((struct pfioc_rule *)addr)->action ==
2611 			    PF_GET_CLR_CNTR)
2612 				return (EACCES);
2613 			break;
2614 		default:
2615 			return (EACCES);
2616 		}
2617 
2618 	CURVNET_SET(TD_TO_VNET(td));
2619 
2620 	switch (cmd) {
2621 	case DIOCSTART:
2622 		sx_xlock(&V_pf_ioctl_lock);
2623 		if (V_pf_status.running)
2624 			error = EEXIST;
2625 		else {
2626 			hook_pf();
2627 			if (! TAILQ_EMPTY(V_pf_keth->active.rules))
2628 				hook_pf_eth();
2629 			V_pf_status.running = 1;
2630 			V_pf_status.since = time_second;
2631 			new_unrhdr64(&V_pf_stateid, time_second);
2632 
2633 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
2634 		}
2635 		break;
2636 
2637 	case DIOCSTOP:
2638 		sx_xlock(&V_pf_ioctl_lock);
2639 		if (!V_pf_status.running)
2640 			error = ENOENT;
2641 		else {
2642 			V_pf_status.running = 0;
2643 			dehook_pf();
2644 			dehook_pf_eth();
2645 			V_pf_status.since = time_second;
2646 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
2647 		}
2648 		break;
2649 
2650 	case DIOCGETETHRULES: {
2651 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2652 		nvlist_t		*nvl;
2653 		void			*packed;
2654 		struct pf_keth_rule	*tail;
2655 		struct pf_keth_ruleset	*rs;
2656 		u_int32_t		 ticket, nr;
2657 		const char		*anchor = "";
2658 
2659 		nvl = NULL;
2660 		packed = NULL;
2661 
2662 #define	ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULES_error, x)
2663 
2664 		if (nv->len > pf_ioctl_maxcount)
2665 			ERROUT(ENOMEM);
2666 
2667 		/* Copy the request in */
2668 		packed = malloc(nv->len, M_NVLIST, M_WAITOK);
2669 		if (packed == NULL)
2670 			ERROUT(ENOMEM);
2671 
2672 		error = copyin(nv->data, packed, nv->len);
2673 		if (error)
2674 			ERROUT(error);
2675 
2676 		nvl = nvlist_unpack(packed, nv->len, 0);
2677 		if (nvl == NULL)
2678 			ERROUT(EBADMSG);
2679 
2680 		if (! nvlist_exists_string(nvl, "anchor"))
2681 			ERROUT(EBADMSG);
2682 
2683 		anchor = nvlist_get_string(nvl, "anchor");
2684 
2685 		rs = pf_find_keth_ruleset(anchor);
2686 
2687 		nvlist_destroy(nvl);
2688 		nvl = NULL;
2689 		free(packed, M_NVLIST);
2690 		packed = NULL;
2691 
2692 		if (rs == NULL)
2693 			ERROUT(ENOENT);
2694 
2695 		/* Reply */
2696 		nvl = nvlist_create(0);
2697 		if (nvl == NULL)
2698 			ERROUT(ENOMEM);
2699 
2700 		PF_RULES_RLOCK();
2701 
2702 		ticket = rs->active.ticket;
2703 		tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq);
2704 		if (tail)
2705 			nr = tail->nr + 1;
2706 		else
2707 			nr = 0;
2708 
2709 		PF_RULES_RUNLOCK();
2710 
2711 		nvlist_add_number(nvl, "ticket", ticket);
2712 		nvlist_add_number(nvl, "nr", nr);
2713 
2714 		packed = nvlist_pack(nvl, &nv->len);
2715 		if (packed == NULL)
2716 			ERROUT(ENOMEM);
2717 
2718 		if (nv->size == 0)
2719 			ERROUT(0);
2720 		else if (nv->size < nv->len)
2721 			ERROUT(ENOSPC);
2722 
2723 		error = copyout(packed, nv->data, nv->len);
2724 
2725 #undef ERROUT
2726 DIOCGETETHRULES_error:
2727 		free(packed, M_NVLIST);
2728 		nvlist_destroy(nvl);
2729 		break;
2730 	}
2731 
2732 	case DIOCGETETHRULE: {
2733 		struct epoch_tracker	 et;
2734 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2735 		nvlist_t		*nvl = NULL;
2736 		void			*nvlpacked = NULL;
2737 		struct pf_keth_rule	*rule = NULL;
2738 		struct pf_keth_ruleset	*rs;
2739 		u_int32_t		 ticket, nr;
2740 		bool			 clear = false;
2741 		const char		*anchor;
2742 
2743 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULE_error, x)
2744 
2745 		if (nv->len > pf_ioctl_maxcount)
2746 			ERROUT(ENOMEM);
2747 
2748 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2749 		if (nvlpacked == NULL)
2750 			ERROUT(ENOMEM);
2751 
2752 		error = copyin(nv->data, nvlpacked, nv->len);
2753 		if (error)
2754 			ERROUT(error);
2755 
2756 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2757 		if (nvl == NULL)
2758 			ERROUT(EBADMSG);
2759 		if (! nvlist_exists_number(nvl, "ticket"))
2760 			ERROUT(EBADMSG);
2761 		ticket = nvlist_get_number(nvl, "ticket");
2762 		if (! nvlist_exists_string(nvl, "anchor"))
2763 			ERROUT(EBADMSG);
2764 		anchor = nvlist_get_string(nvl, "anchor");
2765 
2766 		if (nvlist_exists_bool(nvl, "clear"))
2767 			clear = nvlist_get_bool(nvl, "clear");
2768 
2769 		if (clear && !(flags & FWRITE))
2770 			ERROUT(EACCES);
2771 
2772 		if (! nvlist_exists_number(nvl, "nr"))
2773 			ERROUT(EBADMSG);
2774 		nr = nvlist_get_number(nvl, "nr");
2775 
2776 		PF_RULES_RLOCK();
2777 		rs = pf_find_keth_ruleset(anchor);
2778 		if (rs == NULL) {
2779 			PF_RULES_RUNLOCK();
2780 			ERROUT(ENOENT);
2781 		}
2782 		if (ticket != rs->active.ticket) {
2783 			PF_RULES_RUNLOCK();
2784 			ERROUT(EBUSY);
2785 		}
2786 
2787 		nvlist_destroy(nvl);
2788 		nvl = NULL;
2789 		free(nvlpacked, M_NVLIST);
2790 		nvlpacked = NULL;
2791 
2792 		rule = TAILQ_FIRST(rs->active.rules);
2793 		while ((rule != NULL) && (rule->nr != nr))
2794 			rule = TAILQ_NEXT(rule, entries);
2795 		if (rule == NULL) {
2796 			PF_RULES_RUNLOCK();
2797 			ERROUT(ENOENT);
2798 		}
2799 		/* Make sure rule can't go away. */
2800 		NET_EPOCH_ENTER(et);
2801 		PF_RULES_RUNLOCK();
2802 		nvl = pf_keth_rule_to_nveth_rule(rule);
2803 		if (pf_keth_anchor_nvcopyout(rs, rule, nvl))
2804 			ERROUT(EBUSY);
2805 		NET_EPOCH_EXIT(et);
2806 		if (nvl == NULL)
2807 			ERROUT(ENOMEM);
2808 
2809 		nvlpacked = nvlist_pack(nvl, &nv->len);
2810 		if (nvlpacked == NULL)
2811 			ERROUT(ENOMEM);
2812 
2813 		if (nv->size == 0)
2814 			ERROUT(0);
2815 		else if (nv->size < nv->len)
2816 			ERROUT(ENOSPC);
2817 
2818 		error = copyout(nvlpacked, nv->data, nv->len);
2819 		if (error == 0 && clear) {
2820 			counter_u64_zero(rule->evaluations);
2821 			for (int i = 0; i < 2; i++) {
2822 				counter_u64_zero(rule->packets[i]);
2823 				counter_u64_zero(rule->bytes[i]);
2824 			}
2825 		}
2826 
2827 #undef ERROUT
2828 DIOCGETETHRULE_error:
2829 		free(nvlpacked, M_NVLIST);
2830 		nvlist_destroy(nvl);
2831 		break;
2832 	}
2833 
2834 	case DIOCADDETHRULE: {
2835 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2836 		nvlist_t		*nvl = NULL;
2837 		void			*nvlpacked = NULL;
2838 		struct pf_keth_rule	*rule = NULL, *tail = NULL;
2839 		struct pf_keth_ruleset	*ruleset = NULL;
2840 		struct pfi_kkif		*kif = NULL, *bridge_to_kif = NULL;
2841 		const char		*anchor = "", *anchor_call = "";
2842 
2843 #define ERROUT(x)	ERROUT_IOCTL(DIOCADDETHRULE_error, x)
2844 
2845 		if (nv->len > pf_ioctl_maxcount)
2846 			ERROUT(ENOMEM);
2847 
2848 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2849 		if (nvlpacked == NULL)
2850 			ERROUT(ENOMEM);
2851 
2852 		error = copyin(nv->data, nvlpacked, nv->len);
2853 		if (error)
2854 			ERROUT(error);
2855 
2856 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2857 		if (nvl == NULL)
2858 			ERROUT(EBADMSG);
2859 
2860 		if (! nvlist_exists_number(nvl, "ticket"))
2861 			ERROUT(EBADMSG);
2862 
2863 		if (nvlist_exists_string(nvl, "anchor"))
2864 			anchor = nvlist_get_string(nvl, "anchor");
2865 		if (nvlist_exists_string(nvl, "anchor_call"))
2866 			anchor_call = nvlist_get_string(nvl, "anchor_call");
2867 
2868 		ruleset = pf_find_keth_ruleset(anchor);
2869 		if (ruleset == NULL)
2870 			ERROUT(EINVAL);
2871 
2872 		if (nvlist_get_number(nvl, "ticket") !=
2873 		    ruleset->inactive.ticket) {
2874 			DPFPRINTF(PF_DEBUG_MISC,
2875 			    ("ticket: %d != %d\n",
2876 			    (u_int32_t)nvlist_get_number(nvl, "ticket"),
2877 			    ruleset->inactive.ticket));
2878 			ERROUT(EBUSY);
2879 		}
2880 
2881 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
2882 		if (rule == NULL)
2883 			ERROUT(ENOMEM);
2884 		rule->timestamp = NULL;
2885 
2886 		error = pf_nveth_rule_to_keth_rule(nvl, rule);
2887 		if (error != 0)
2888 			ERROUT(error);
2889 
2890 		if (rule->ifname[0])
2891 			kif = pf_kkif_create(M_WAITOK);
2892 		if (rule->bridge_to_name[0])
2893 			bridge_to_kif = pf_kkif_create(M_WAITOK);
2894 		rule->evaluations = counter_u64_alloc(M_WAITOK);
2895 		for (int i = 0; i < 2; i++) {
2896 			rule->packets[i] = counter_u64_alloc(M_WAITOK);
2897 			rule->bytes[i] = counter_u64_alloc(M_WAITOK);
2898 		}
2899 		rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
2900 		    M_WAITOK | M_ZERO);
2901 
2902 		PF_RULES_WLOCK();
2903 
2904 		if (rule->ifname[0]) {
2905 			rule->kif = pfi_kkif_attach(kif, rule->ifname);
2906 			pfi_kkif_ref(rule->kif);
2907 		} else
2908 			rule->kif = NULL;
2909 		if (rule->bridge_to_name[0]) {
2910 			rule->bridge_to = pfi_kkif_attach(bridge_to_kif,
2911 			    rule->bridge_to_name);
2912 			pfi_kkif_ref(rule->bridge_to);
2913 		} else
2914 			rule->bridge_to = NULL;
2915 
2916 #ifdef ALTQ
2917 		/* set queue IDs */
2918 		if (rule->qname[0] != 0) {
2919 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2920 				error = EBUSY;
2921 			else
2922 				rule->qid = rule->qid;
2923 		}
2924 #endif
2925 		if (rule->tagname[0])
2926 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2927 				error = EBUSY;
2928 		if (rule->match_tagname[0])
2929 			if ((rule->match_tag = pf_tagname2tag(
2930 			    rule->match_tagname)) == 0)
2931 				error = EBUSY;
2932 
2933 		if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE)
2934 			error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr);
2935 		if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE)
2936 			error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr);
2937 
2938 		if (error) {
2939 			pf_free_eth_rule(rule);
2940 			PF_RULES_WUNLOCK();
2941 			ERROUT(error);
2942 		}
2943 
2944 		if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) {
2945 			pf_free_eth_rule(rule);
2946 			PF_RULES_WUNLOCK();
2947 			ERROUT(EINVAL);
2948 		}
2949 
2950 		tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq);
2951 		if (tail)
2952 			rule->nr = tail->nr + 1;
2953 		else
2954 			rule->nr = 0;
2955 
2956 		TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries);
2957 
2958 		PF_RULES_WUNLOCK();
2959 
2960 #undef ERROUT
2961 DIOCADDETHRULE_error:
2962 		nvlist_destroy(nvl);
2963 		free(nvlpacked, M_NVLIST);
2964 		break;
2965 	}
2966 
2967 	case DIOCGETETHRULESETS: {
2968 		struct epoch_tracker	 et;
2969 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2970 		nvlist_t		*nvl = NULL;
2971 		void			*nvlpacked = NULL;
2972 		struct pf_keth_ruleset	*ruleset;
2973 		struct pf_keth_anchor	*anchor;
2974 		int			 nr = 0;
2975 
2976 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULESETS_error, x)
2977 
2978 		if (nv->len > pf_ioctl_maxcount)
2979 			ERROUT(ENOMEM);
2980 
2981 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2982 		if (nvlpacked == NULL)
2983 			ERROUT(ENOMEM);
2984 
2985 		error = copyin(nv->data, nvlpacked, nv->len);
2986 		if (error)
2987 			ERROUT(error);
2988 
2989 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2990 		if (nvl == NULL)
2991 			ERROUT(EBADMSG);
2992 		if (! nvlist_exists_string(nvl, "path"))
2993 			ERROUT(EBADMSG);
2994 
2995 		NET_EPOCH_ENTER(et);
2996 
2997 		if ((ruleset = pf_find_keth_ruleset(
2998 		    nvlist_get_string(nvl, "path"))) == NULL) {
2999 			NET_EPOCH_EXIT(et);
3000 			ERROUT(ENOENT);
3001 		}
3002 
3003 		if (ruleset->anchor == NULL) {
3004 			RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors)
3005 				if (anchor->parent == NULL)
3006 					nr++;
3007 		} else {
3008 			RB_FOREACH(anchor, pf_keth_anchor_node,
3009 			    &ruleset->anchor->children)
3010 				nr++;
3011 		}
3012 
3013 		NET_EPOCH_EXIT(et);
3014 
3015 		nvlist_destroy(nvl);
3016 		nvl = NULL;
3017 		free(nvlpacked, M_NVLIST);
3018 		nvlpacked = NULL;
3019 
3020 		nvl = nvlist_create(0);
3021 		if (nvl == NULL)
3022 			ERROUT(ENOMEM);
3023 
3024 		nvlist_add_number(nvl, "nr", nr);
3025 
3026 		nvlpacked = nvlist_pack(nvl, &nv->len);
3027 		if (nvlpacked == NULL)
3028 			ERROUT(ENOMEM);
3029 
3030 		if (nv->size == 0)
3031 			ERROUT(0);
3032 		else if (nv->size < nv->len)
3033 			ERROUT(ENOSPC);
3034 
3035 		error = copyout(nvlpacked, nv->data, nv->len);
3036 
3037 #undef ERROUT
3038 DIOCGETETHRULESETS_error:
3039 		free(nvlpacked, M_NVLIST);
3040 		nvlist_destroy(nvl);
3041 		break;
3042 	}
3043 
3044 	case DIOCGETETHRULESET: {
3045 		struct epoch_tracker	 et;
3046 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
3047 		nvlist_t		*nvl = NULL;
3048 		void			*nvlpacked = NULL;
3049 		struct pf_keth_ruleset	*ruleset;
3050 		struct pf_keth_anchor	*anchor;
3051 		int			 nr = 0, req_nr = 0;
3052 		bool			 found = false;
3053 
3054 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULESET_error, x)
3055 
3056 		if (nv->len > pf_ioctl_maxcount)
3057 			ERROUT(ENOMEM);
3058 
3059 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3060 		if (nvlpacked == NULL)
3061 			ERROUT(ENOMEM);
3062 
3063 		error = copyin(nv->data, nvlpacked, nv->len);
3064 		if (error)
3065 			ERROUT(error);
3066 
3067 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3068 		if (nvl == NULL)
3069 			ERROUT(EBADMSG);
3070 		if (! nvlist_exists_string(nvl, "path"))
3071 			ERROUT(EBADMSG);
3072 		if (! nvlist_exists_number(nvl, "nr"))
3073 			ERROUT(EBADMSG);
3074 
3075 		req_nr = nvlist_get_number(nvl, "nr");
3076 
3077 		NET_EPOCH_ENTER(et);
3078 
3079 		if ((ruleset = pf_find_keth_ruleset(
3080 		    nvlist_get_string(nvl, "path"))) == NULL) {
3081 			NET_EPOCH_EXIT(et);
3082 			ERROUT(ENOENT);
3083 		}
3084 
3085 		nvlist_destroy(nvl);
3086 		nvl = NULL;
3087 		free(nvlpacked, M_NVLIST);
3088 		nvlpacked = NULL;
3089 
3090 		nvl = nvlist_create(0);
3091 		if (nvl == NULL) {
3092 			NET_EPOCH_EXIT(et);
3093 			ERROUT(ENOMEM);
3094 		}
3095 
3096 		if (ruleset->anchor == NULL) {
3097 			RB_FOREACH(anchor, pf_keth_anchor_global,
3098 			    &V_pf_keth_anchors) {
3099 				if (anchor->parent == NULL && nr++ == req_nr) {
3100 					found = true;
3101 					break;
3102 				}
3103 			}
3104 		} else {
3105 			RB_FOREACH(anchor, pf_keth_anchor_node,
3106 			     &ruleset->anchor->children) {
3107 				if (nr++ == req_nr) {
3108 					found = true;
3109 					break;
3110 				}
3111 			}
3112 		}
3113 
3114 		NET_EPOCH_EXIT(et);
3115 		if (found) {
3116 			nvlist_add_number(nvl, "nr", nr);
3117 			nvlist_add_string(nvl, "name", anchor->name);
3118 			if (ruleset->anchor)
3119 				nvlist_add_string(nvl, "path",
3120 				    ruleset->anchor->path);
3121 			else
3122 				nvlist_add_string(nvl, "path", "");
3123 		} else {
3124 			ERROUT(EBUSY);
3125 		}
3126 
3127 		nvlpacked = nvlist_pack(nvl, &nv->len);
3128 		if (nvlpacked == NULL)
3129 			ERROUT(ENOMEM);
3130 
3131 		if (nv->size == 0)
3132 			ERROUT(0);
3133 		else if (nv->size < nv->len)
3134 			ERROUT(ENOSPC);
3135 
3136 		error = copyout(nvlpacked, nv->data, nv->len);
3137 
3138 #undef ERROUT
3139 DIOCGETETHRULESET_error:
3140 		free(nvlpacked, M_NVLIST);
3141 		nvlist_destroy(nvl);
3142 		break;
3143 	}
3144 
3145 	case DIOCADDRULENV: {
3146 		struct pfioc_nv	*nv = (struct pfioc_nv *)addr;
3147 		nvlist_t	*nvl = NULL;
3148 		void		*nvlpacked = NULL;
3149 		struct pf_krule	*rule = NULL;
3150 		const char	*anchor = "", *anchor_call = "";
3151 		uint32_t	 ticket = 0, pool_ticket = 0;
3152 
3153 #define	ERROUT(x)	ERROUT_IOCTL(DIOCADDRULENV_error, x)
3154 
3155 		if (nv->len > pf_ioctl_maxcount)
3156 			ERROUT(ENOMEM);
3157 
3158 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3159 		error = copyin(nv->data, nvlpacked, nv->len);
3160 		if (error)
3161 			ERROUT(error);
3162 
3163 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3164 		if (nvl == NULL)
3165 			ERROUT(EBADMSG);
3166 
3167 		if (! nvlist_exists_number(nvl, "ticket"))
3168 			ERROUT(EINVAL);
3169 		ticket = nvlist_get_number(nvl, "ticket");
3170 
3171 		if (! nvlist_exists_number(nvl, "pool_ticket"))
3172 			ERROUT(EINVAL);
3173 		pool_ticket = nvlist_get_number(nvl, "pool_ticket");
3174 
3175 		if (! nvlist_exists_nvlist(nvl, "rule"))
3176 			ERROUT(EINVAL);
3177 
3178 		rule = pf_krule_alloc();
3179 		error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"),
3180 		    rule);
3181 		if (error)
3182 			ERROUT(error);
3183 
3184 		if (nvlist_exists_string(nvl, "anchor"))
3185 			anchor = nvlist_get_string(nvl, "anchor");
3186 		if (nvlist_exists_string(nvl, "anchor_call"))
3187 			anchor_call = nvlist_get_string(nvl, "anchor_call");
3188 
3189 		if ((error = nvlist_error(nvl)))
3190 			ERROUT(error);
3191 
3192 		/* Frees rule on error */
3193 		error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor,
3194 		    anchor_call, td);
3195 
3196 		nvlist_destroy(nvl);
3197 		free(nvlpacked, M_NVLIST);
3198 		break;
3199 #undef ERROUT
3200 DIOCADDRULENV_error:
3201 		pf_krule_free(rule);
3202 		nvlist_destroy(nvl);
3203 		free(nvlpacked, M_NVLIST);
3204 
3205 		break;
3206 	}
3207 	case DIOCADDRULE: {
3208 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
3209 		struct pf_krule		*rule;
3210 
3211 		rule = pf_krule_alloc();
3212 		error = pf_rule_to_krule(&pr->rule, rule);
3213 		if (error != 0) {
3214 			pf_krule_free(rule);
3215 			break;
3216 		}
3217 
3218 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
3219 
3220 		/* Frees rule on error */
3221 		error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket,
3222 		    pr->anchor, pr->anchor_call, td);
3223 		break;
3224 	}
3225 
3226 	case DIOCGETRULES: {
3227 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
3228 		struct pf_kruleset	*ruleset;
3229 		struct pf_krule		*tail;
3230 		int			 rs_num;
3231 
3232 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
3233 
3234 		PF_RULES_WLOCK();
3235 		ruleset = pf_find_kruleset(pr->anchor);
3236 		if (ruleset == NULL) {
3237 			PF_RULES_WUNLOCK();
3238 			error = EINVAL;
3239 			break;
3240 		}
3241 		rs_num = pf_get_ruleset_number(pr->rule.action);
3242 		if (rs_num >= PF_RULESET_MAX) {
3243 			PF_RULES_WUNLOCK();
3244 			error = EINVAL;
3245 			break;
3246 		}
3247 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
3248 		    pf_krulequeue);
3249 		if (tail)
3250 			pr->nr = tail->nr + 1;
3251 		else
3252 			pr->nr = 0;
3253 		pr->ticket = ruleset->rules[rs_num].active.ticket;
3254 		PF_RULES_WUNLOCK();
3255 		break;
3256 	}
3257 
3258 	case DIOCGETRULE: {
3259 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
3260 		struct pf_kruleset	*ruleset;
3261 		struct pf_krule		*rule;
3262 		int			 rs_num;
3263 
3264 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
3265 
3266 		PF_RULES_WLOCK();
3267 		ruleset = pf_find_kruleset(pr->anchor);
3268 		if (ruleset == NULL) {
3269 			PF_RULES_WUNLOCK();
3270 			error = EINVAL;
3271 			break;
3272 		}
3273 		rs_num = pf_get_ruleset_number(pr->rule.action);
3274 		if (rs_num >= PF_RULESET_MAX) {
3275 			PF_RULES_WUNLOCK();
3276 			error = EINVAL;
3277 			break;
3278 		}
3279 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
3280 			PF_RULES_WUNLOCK();
3281 			error = EBUSY;
3282 			break;
3283 		}
3284 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
3285 		while ((rule != NULL) && (rule->nr != pr->nr))
3286 			rule = TAILQ_NEXT(rule, entries);
3287 		if (rule == NULL) {
3288 			PF_RULES_WUNLOCK();
3289 			error = EBUSY;
3290 			break;
3291 		}
3292 
3293 		pf_krule_to_rule(rule, &pr->rule);
3294 
3295 		if (pf_kanchor_copyout(ruleset, rule, pr)) {
3296 			PF_RULES_WUNLOCK();
3297 			error = EBUSY;
3298 			break;
3299 		}
3300 		pf_addr_copyout(&pr->rule.src.addr);
3301 		pf_addr_copyout(&pr->rule.dst.addr);
3302 
3303 		if (pr->action == PF_GET_CLR_CNTR) {
3304 			pf_counter_u64_zero(&rule->evaluations);
3305 			for (int i = 0; i < 2; i++) {
3306 				pf_counter_u64_zero(&rule->packets[i]);
3307 				pf_counter_u64_zero(&rule->bytes[i]);
3308 			}
3309 			counter_u64_zero(rule->states_tot);
3310 		}
3311 		PF_RULES_WUNLOCK();
3312 		break;
3313 	}
3314 
3315 	case DIOCGETRULENV: {
3316 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
3317 		nvlist_t		*nvrule = NULL;
3318 		nvlist_t		*nvl = NULL;
3319 		struct pf_kruleset	*ruleset;
3320 		struct pf_krule		*rule;
3321 		void			*nvlpacked = NULL;
3322 		int			 rs_num, nr;
3323 		bool			 clear_counter = false;
3324 
3325 #define	ERROUT(x)	ERROUT_IOCTL(DIOCGETRULENV_error, x)
3326 
3327 		if (nv->len > pf_ioctl_maxcount)
3328 			ERROUT(ENOMEM);
3329 
3330 		/* Copy the request in */
3331 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3332 		if (nvlpacked == NULL)
3333 			ERROUT(ENOMEM);
3334 
3335 		error = copyin(nv->data, nvlpacked, nv->len);
3336 		if (error)
3337 			ERROUT(error);
3338 
3339 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3340 		if (nvl == NULL)
3341 			ERROUT(EBADMSG);
3342 
3343 		if (! nvlist_exists_string(nvl, "anchor"))
3344 			ERROUT(EBADMSG);
3345 		if (! nvlist_exists_number(nvl, "ruleset"))
3346 			ERROUT(EBADMSG);
3347 		if (! nvlist_exists_number(nvl, "ticket"))
3348 			ERROUT(EBADMSG);
3349 		if (! nvlist_exists_number(nvl, "nr"))
3350 			ERROUT(EBADMSG);
3351 
3352 		if (nvlist_exists_bool(nvl, "clear_counter"))
3353 			clear_counter = nvlist_get_bool(nvl, "clear_counter");
3354 
3355 		if (clear_counter && !(flags & FWRITE))
3356 			ERROUT(EACCES);
3357 
3358 		nr = nvlist_get_number(nvl, "nr");
3359 
3360 		PF_RULES_WLOCK();
3361 		ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor"));
3362 		if (ruleset == NULL) {
3363 			PF_RULES_WUNLOCK();
3364 			ERROUT(ENOENT);
3365 		}
3366 
3367 		rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset"));
3368 		if (rs_num >= PF_RULESET_MAX) {
3369 			PF_RULES_WUNLOCK();
3370 			ERROUT(EINVAL);
3371 		}
3372 
3373 		if (nvlist_get_number(nvl, "ticket") !=
3374 		    ruleset->rules[rs_num].active.ticket) {
3375 			PF_RULES_WUNLOCK();
3376 			ERROUT(EBUSY);
3377 		}
3378 
3379 		if ((error = nvlist_error(nvl))) {
3380 			PF_RULES_WUNLOCK();
3381 			ERROUT(error);
3382 		}
3383 
3384 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
3385 		while ((rule != NULL) && (rule->nr != nr))
3386 			rule = TAILQ_NEXT(rule, entries);
3387 		if (rule == NULL) {
3388 			PF_RULES_WUNLOCK();
3389 			ERROUT(EBUSY);
3390 		}
3391 
3392 		nvrule = pf_krule_to_nvrule(rule);
3393 
3394 		nvlist_destroy(nvl);
3395 		nvl = nvlist_create(0);
3396 		if (nvl == NULL) {
3397 			PF_RULES_WUNLOCK();
3398 			ERROUT(ENOMEM);
3399 		}
3400 		nvlist_add_number(nvl, "nr", nr);
3401 		nvlist_add_nvlist(nvl, "rule", nvrule);
3402 		nvlist_destroy(nvrule);
3403 		nvrule = NULL;
3404 		if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) {
3405 			PF_RULES_WUNLOCK();
3406 			ERROUT(EBUSY);
3407 		}
3408 
3409 		free(nvlpacked, M_NVLIST);
3410 		nvlpacked = nvlist_pack(nvl, &nv->len);
3411 		if (nvlpacked == NULL) {
3412 			PF_RULES_WUNLOCK();
3413 			ERROUT(ENOMEM);
3414 		}
3415 
3416 		if (nv->size == 0) {
3417 			PF_RULES_WUNLOCK();
3418 			ERROUT(0);
3419 		}
3420 		else if (nv->size < nv->len) {
3421 			PF_RULES_WUNLOCK();
3422 			ERROUT(ENOSPC);
3423 		}
3424 
3425 		if (clear_counter) {
3426 			pf_counter_u64_zero(&rule->evaluations);
3427 			for (int i = 0; i < 2; i++) {
3428 				pf_counter_u64_zero(&rule->packets[i]);
3429 				pf_counter_u64_zero(&rule->bytes[i]);
3430 			}
3431 			counter_u64_zero(rule->states_tot);
3432 		}
3433 		PF_RULES_WUNLOCK();
3434 
3435 		error = copyout(nvlpacked, nv->data, nv->len);
3436 
3437 #undef ERROUT
3438 DIOCGETRULENV_error:
3439 		free(nvlpacked, M_NVLIST);
3440 		nvlist_destroy(nvrule);
3441 		nvlist_destroy(nvl);
3442 
3443 		break;
3444 	}
3445 
3446 	case DIOCCHANGERULE: {
3447 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
3448 		struct pf_kruleset	*ruleset;
3449 		struct pf_krule		*oldrule = NULL, *newrule = NULL;
3450 		struct pfi_kkif		*kif = NULL;
3451 		struct pf_kpooladdr	*pa;
3452 		u_int32_t		 nr = 0;
3453 		int			 rs_num;
3454 
3455 		pcr->anchor[sizeof(pcr->anchor) - 1] = 0;
3456 
3457 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
3458 		    pcr->action > PF_CHANGE_GET_TICKET) {
3459 			error = EINVAL;
3460 			break;
3461 		}
3462 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
3463 			error = EINVAL;
3464 			break;
3465 		}
3466 
3467 		if (pcr->action != PF_CHANGE_REMOVE) {
3468 			newrule = pf_krule_alloc();
3469 			error = pf_rule_to_krule(&pcr->rule, newrule);
3470 			if (error != 0) {
3471 				pf_krule_free(newrule);
3472 				break;
3473 			}
3474 
3475 			if (newrule->ifname[0])
3476 				kif = pf_kkif_create(M_WAITOK);
3477 			pf_counter_u64_init(&newrule->evaluations, M_WAITOK);
3478 			for (int i = 0; i < 2; i++) {
3479 				pf_counter_u64_init(&newrule->packets[i], M_WAITOK);
3480 				pf_counter_u64_init(&newrule->bytes[i], M_WAITOK);
3481 			}
3482 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
3483 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
3484 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
3485 			newrule->cuid = td->td_ucred->cr_ruid;
3486 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
3487 			TAILQ_INIT(&newrule->rpool.list);
3488 		}
3489 #define	ERROUT(x)	ERROUT_IOCTL(DIOCCHANGERULE_error, x)
3490 
3491 		PF_CONFIG_LOCK();
3492 		PF_RULES_WLOCK();
3493 #ifdef PF_WANT_32_TO_64_COUNTER
3494 		if (newrule != NULL) {
3495 			LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist);
3496 			newrule->allrulelinked = true;
3497 			V_pf_allrulecount++;
3498 		}
3499 #endif
3500 
3501 		if (!(pcr->action == PF_CHANGE_REMOVE ||
3502 		    pcr->action == PF_CHANGE_GET_TICKET) &&
3503 		    pcr->pool_ticket != V_ticket_pabuf)
3504 			ERROUT(EBUSY);
3505 
3506 		ruleset = pf_find_kruleset(pcr->anchor);
3507 		if (ruleset == NULL)
3508 			ERROUT(EINVAL);
3509 
3510 		rs_num = pf_get_ruleset_number(pcr->rule.action);
3511 		if (rs_num >= PF_RULESET_MAX)
3512 			ERROUT(EINVAL);
3513 
3514 		/*
3515 		 * XXXMJG: there is no guarantee that the ruleset was
3516 		 * created by the usual route of calling DIOCXBEGIN.
3517 		 * As a result it is possible the rule tree will not
3518 		 * be allocated yet. Hack around it by doing it here.
3519 		 * Note it is fine to let the tree persist in case of
3520 		 * error as it will be freed down the road on future
3521 		 * updates (if need be).
3522 		 */
3523 		if (ruleset->rules[rs_num].active.tree == NULL) {
3524 			ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT);
3525 			if (ruleset->rules[rs_num].active.tree == NULL) {
3526 				ERROUT(ENOMEM);
3527 			}
3528 		}
3529 
3530 		if (pcr->action == PF_CHANGE_GET_TICKET) {
3531 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
3532 			ERROUT(0);
3533 		} else if (pcr->ticket !=
3534 			    ruleset->rules[rs_num].active.ticket)
3535 				ERROUT(EINVAL);
3536 
3537 		if (pcr->action != PF_CHANGE_REMOVE) {
3538 			if (newrule->ifname[0]) {
3539 				newrule->kif = pfi_kkif_attach(kif,
3540 				    newrule->ifname);
3541 				kif = NULL;
3542 				pfi_kkif_ref(newrule->kif);
3543 			} else
3544 				newrule->kif = NULL;
3545 
3546 			if (newrule->rtableid > 0 &&
3547 			    newrule->rtableid >= rt_numfibs)
3548 				error = EBUSY;
3549 
3550 #ifdef ALTQ
3551 			/* set queue IDs */
3552 			if (newrule->qname[0] != 0) {
3553 				if ((newrule->qid =
3554 				    pf_qname2qid(newrule->qname)) == 0)
3555 					error = EBUSY;
3556 				else if (newrule->pqname[0] != 0) {
3557 					if ((newrule->pqid =
3558 					    pf_qname2qid(newrule->pqname)) == 0)
3559 						error = EBUSY;
3560 				} else
3561 					newrule->pqid = newrule->qid;
3562 			}
3563 #endif /* ALTQ */
3564 			if (newrule->tagname[0])
3565 				if ((newrule->tag =
3566 				    pf_tagname2tag(newrule->tagname)) == 0)
3567 					error = EBUSY;
3568 			if (newrule->match_tagname[0])
3569 				if ((newrule->match_tag = pf_tagname2tag(
3570 				    newrule->match_tagname)) == 0)
3571 					error = EBUSY;
3572 			if (newrule->rt && !newrule->direction)
3573 				error = EINVAL;
3574 			if (!newrule->log)
3575 				newrule->logif = 0;
3576 			if (newrule->logif >= PFLOGIFS_MAX)
3577 				error = EINVAL;
3578 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
3579 				error = ENOMEM;
3580 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
3581 				error = ENOMEM;
3582 			if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
3583 				error = EINVAL;
3584 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
3585 				if (pa->addr.type == PF_ADDR_TABLE) {
3586 					pa->addr.p.tbl =
3587 					    pfr_attach_table(ruleset,
3588 					    pa->addr.v.tblname);
3589 					if (pa->addr.p.tbl == NULL)
3590 						error = ENOMEM;
3591 				}
3592 
3593 			newrule->overload_tbl = NULL;
3594 			if (newrule->overload_tblname[0]) {
3595 				if ((newrule->overload_tbl = pfr_attach_table(
3596 				    ruleset, newrule->overload_tblname)) ==
3597 				    NULL)
3598 					error = EINVAL;
3599 				else
3600 					newrule->overload_tbl->pfrkt_flags |=
3601 					    PFR_TFLAG_ACTIVE;
3602 			}
3603 
3604 			pf_mv_kpool(&V_pf_pabuf, &newrule->rpool.list);
3605 			if (((((newrule->action == PF_NAT) ||
3606 			    (newrule->action == PF_RDR) ||
3607 			    (newrule->action == PF_BINAT) ||
3608 			    (newrule->rt > PF_NOPFROUTE)) &&
3609 			    !newrule->anchor)) &&
3610 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
3611 				error = EINVAL;
3612 
3613 			if (error) {
3614 				pf_free_rule(newrule);
3615 				PF_RULES_WUNLOCK();
3616 				PF_CONFIG_UNLOCK();
3617 				break;
3618 			}
3619 
3620 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
3621 		}
3622 		pf_empty_kpool(&V_pf_pabuf);
3623 
3624 		if (pcr->action == PF_CHANGE_ADD_HEAD)
3625 			oldrule = TAILQ_FIRST(
3626 			    ruleset->rules[rs_num].active.ptr);
3627 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
3628 			oldrule = TAILQ_LAST(
3629 			    ruleset->rules[rs_num].active.ptr, pf_krulequeue);
3630 		else {
3631 			oldrule = TAILQ_FIRST(
3632 			    ruleset->rules[rs_num].active.ptr);
3633 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
3634 				oldrule = TAILQ_NEXT(oldrule, entries);
3635 			if (oldrule == NULL) {
3636 				if (newrule != NULL)
3637 					pf_free_rule(newrule);
3638 				PF_RULES_WUNLOCK();
3639 				PF_CONFIG_UNLOCK();
3640 				error = EINVAL;
3641 				break;
3642 			}
3643 		}
3644 
3645 		if (pcr->action == PF_CHANGE_REMOVE) {
3646 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
3647 			    oldrule);
3648 			RB_REMOVE(pf_krule_global,
3649 			    ruleset->rules[rs_num].active.tree, oldrule);
3650 			ruleset->rules[rs_num].active.rcount--;
3651 		} else {
3652 			pf_hash_rule(newrule);
3653 			if (RB_INSERT(pf_krule_global,
3654 			    ruleset->rules[rs_num].active.tree, newrule) != NULL) {
3655 				pf_free_rule(newrule);
3656 				PF_RULES_WUNLOCK();
3657 				PF_CONFIG_UNLOCK();
3658 				error = EEXIST;
3659 				break;
3660 			}
3661 
3662 			if (oldrule == NULL)
3663 				TAILQ_INSERT_TAIL(
3664 				    ruleset->rules[rs_num].active.ptr,
3665 				    newrule, entries);
3666 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
3667 			    pcr->action == PF_CHANGE_ADD_BEFORE)
3668 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
3669 			else
3670 				TAILQ_INSERT_AFTER(
3671 				    ruleset->rules[rs_num].active.ptr,
3672 				    oldrule, newrule, entries);
3673 			ruleset->rules[rs_num].active.rcount++;
3674 		}
3675 
3676 		nr = 0;
3677 		TAILQ_FOREACH(oldrule,
3678 		    ruleset->rules[rs_num].active.ptr, entries)
3679 			oldrule->nr = nr++;
3680 
3681 		ruleset->rules[rs_num].active.ticket++;
3682 
3683 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
3684 		pf_remove_if_empty_kruleset(ruleset);
3685 
3686 		PF_RULES_WUNLOCK();
3687 		PF_CONFIG_UNLOCK();
3688 		break;
3689 
3690 #undef ERROUT
3691 DIOCCHANGERULE_error:
3692 		PF_RULES_WUNLOCK();
3693 		PF_CONFIG_UNLOCK();
3694 		pf_krule_free(newrule);
3695 		pf_kkif_free(kif);
3696 		break;
3697 	}
3698 
3699 	case DIOCCLRSTATES: {
3700 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
3701 		struct pf_kstate_kill	 kill;
3702 
3703 		error = pf_state_kill_to_kstate_kill(psk, &kill);
3704 		if (error)
3705 			break;
3706 
3707 		psk->psk_killed = pf_clear_states(&kill);
3708 		break;
3709 	}
3710 
3711 	case DIOCCLRSTATESNV: {
3712 		error = pf_clearstates_nv((struct pfioc_nv *)addr);
3713 		break;
3714 	}
3715 
3716 	case DIOCKILLSTATES: {
3717 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
3718 		struct pf_kstate_kill	 kill;
3719 
3720 		error = pf_state_kill_to_kstate_kill(psk, &kill);
3721 		if (error)
3722 			break;
3723 
3724 		psk->psk_killed = 0;
3725 		pf_killstates(&kill, &psk->psk_killed);
3726 		break;
3727 	}
3728 
3729 	case DIOCKILLSTATESNV: {
3730 		error = pf_killstates_nv((struct pfioc_nv *)addr);
3731 		break;
3732 	}
3733 
3734 	case DIOCADDSTATE: {
3735 		struct pfioc_state		*ps = (struct pfioc_state *)addr;
3736 		struct pfsync_state_1301	*sp = &ps->state;
3737 
3738 		if (sp->timeout >= PFTM_MAX) {
3739 			error = EINVAL;
3740 			break;
3741 		}
3742 		if (V_pfsync_state_import_ptr != NULL) {
3743 			PF_RULES_RLOCK();
3744 			error = V_pfsync_state_import_ptr(
3745 			    (union pfsync_state_union *)sp, PFSYNC_SI_IOCTL,
3746 			    PFSYNC_MSG_VERSION_1301);
3747 			PF_RULES_RUNLOCK();
3748 		} else
3749 			error = EOPNOTSUPP;
3750 		break;
3751 	}
3752 
3753 	case DIOCGETSTATE: {
3754 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
3755 		struct pf_kstate	*s;
3756 
3757 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
3758 		if (s == NULL) {
3759 			error = ENOENT;
3760 			break;
3761 		}
3762 
3763 		pfsync_state_export((union pfsync_state_union*)&ps->state,
3764 		    s, PFSYNC_MSG_VERSION_1301);
3765 		PF_STATE_UNLOCK(s);
3766 		break;
3767 	}
3768 
3769 	case DIOCGETSTATENV: {
3770 		error = pf_getstate((struct pfioc_nv *)addr);
3771 		break;
3772 	}
3773 
3774 	case DIOCGETSTATES: {
3775 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
3776 		struct pf_kstate	*s;
3777 		struct pfsync_state_1301	*pstore, *p;
3778 		int			 i, nr;
3779 		size_t			 slice_count = 16, count;
3780 		void			*out;
3781 
3782 		if (ps->ps_len <= 0) {
3783 			nr = uma_zone_get_cur(V_pf_state_z);
3784 			ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3785 			break;
3786 		}
3787 
3788 		out = ps->ps_states;
3789 		pstore = mallocarray(slice_count,
3790 		    sizeof(struct pfsync_state_1301), M_TEMP, M_WAITOK | M_ZERO);
3791 		nr = 0;
3792 
3793 		for (i = 0; i <= pf_hashmask; i++) {
3794 			struct pf_idhash *ih = &V_pf_idhash[i];
3795 
3796 DIOCGETSTATES_retry:
3797 			p = pstore;
3798 
3799 			if (LIST_EMPTY(&ih->states))
3800 				continue;
3801 
3802 			PF_HASHROW_LOCK(ih);
3803 			count = 0;
3804 			LIST_FOREACH(s, &ih->states, entry) {
3805 				if (s->timeout == PFTM_UNLINKED)
3806 					continue;
3807 				count++;
3808 			}
3809 
3810 			if (count > slice_count) {
3811 				PF_HASHROW_UNLOCK(ih);
3812 				free(pstore, M_TEMP);
3813 				slice_count = count * 2;
3814 				pstore = mallocarray(slice_count,
3815 				    sizeof(struct pfsync_state_1301), M_TEMP,
3816 				    M_WAITOK | M_ZERO);
3817 				goto DIOCGETSTATES_retry;
3818 			}
3819 
3820 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
3821 				PF_HASHROW_UNLOCK(ih);
3822 				goto DIOCGETSTATES_full;
3823 			}
3824 
3825 			LIST_FOREACH(s, &ih->states, entry) {
3826 				if (s->timeout == PFTM_UNLINKED)
3827 					continue;
3828 
3829 				pfsync_state_export((union pfsync_state_union*)p,
3830 				    s, PFSYNC_MSG_VERSION_1301);
3831 				p++;
3832 				nr++;
3833 			}
3834 			PF_HASHROW_UNLOCK(ih);
3835 			error = copyout(pstore, out,
3836 			    sizeof(struct pfsync_state_1301) * count);
3837 			if (error)
3838 				break;
3839 			out = ps->ps_states + nr;
3840 		}
3841 DIOCGETSTATES_full:
3842 		ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3843 		free(pstore, M_TEMP);
3844 
3845 		break;
3846 	}
3847 
3848 	case DIOCGETSTATESV2: {
3849 		struct pfioc_states_v2	*ps = (struct pfioc_states_v2 *)addr;
3850 		struct pf_kstate	*s;
3851 		struct pf_state_export	*pstore, *p;
3852 		int i, nr;
3853 		size_t slice_count = 16, count;
3854 		void *out;
3855 
3856 		if (ps->ps_req_version > PF_STATE_VERSION) {
3857 			error = ENOTSUP;
3858 			break;
3859 		}
3860 
3861 		if (ps->ps_len <= 0) {
3862 			nr = uma_zone_get_cur(V_pf_state_z);
3863 			ps->ps_len = sizeof(struct pf_state_export) * nr;
3864 			break;
3865 		}
3866 
3867 		out = ps->ps_states;
3868 		pstore = mallocarray(slice_count,
3869 		    sizeof(struct pf_state_export), M_TEMP, M_WAITOK | M_ZERO);
3870 		nr = 0;
3871 
3872 		for (i = 0; i <= pf_hashmask; i++) {
3873 			struct pf_idhash *ih = &V_pf_idhash[i];
3874 
3875 DIOCGETSTATESV2_retry:
3876 			p = pstore;
3877 
3878 			if (LIST_EMPTY(&ih->states))
3879 				continue;
3880 
3881 			PF_HASHROW_LOCK(ih);
3882 			count = 0;
3883 			LIST_FOREACH(s, &ih->states, entry) {
3884 				if (s->timeout == PFTM_UNLINKED)
3885 					continue;
3886 				count++;
3887 			}
3888 
3889 			if (count > slice_count) {
3890 				PF_HASHROW_UNLOCK(ih);
3891 				free(pstore, M_TEMP);
3892 				slice_count = count * 2;
3893 				pstore = mallocarray(slice_count,
3894 				    sizeof(struct pf_state_export), M_TEMP,
3895 				    M_WAITOK | M_ZERO);
3896 				goto DIOCGETSTATESV2_retry;
3897 			}
3898 
3899 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
3900 				PF_HASHROW_UNLOCK(ih);
3901 				goto DIOCGETSTATESV2_full;
3902 			}
3903 
3904 			LIST_FOREACH(s, &ih->states, entry) {
3905 				if (s->timeout == PFTM_UNLINKED)
3906 					continue;
3907 
3908 				pf_state_export(p, s);
3909 				p++;
3910 				nr++;
3911 			}
3912 			PF_HASHROW_UNLOCK(ih);
3913 			error = copyout(pstore, out,
3914 			    sizeof(struct pf_state_export) * count);
3915 			if (error)
3916 				break;
3917 			out = ps->ps_states + nr;
3918 		}
3919 DIOCGETSTATESV2_full:
3920 		ps->ps_len = nr * sizeof(struct pf_state_export);
3921 		free(pstore, M_TEMP);
3922 
3923 		break;
3924 	}
3925 
3926 	case DIOCGETSTATUS: {
3927 		struct pf_status *s = (struct pf_status *)addr;
3928 
3929 		PF_RULES_RLOCK();
3930 		s->running = V_pf_status.running;
3931 		s->since   = V_pf_status.since;
3932 		s->debug   = V_pf_status.debug;
3933 		s->hostid  = V_pf_status.hostid;
3934 		s->states  = V_pf_status.states;
3935 		s->src_nodes = V_pf_status.src_nodes;
3936 
3937 		for (int i = 0; i < PFRES_MAX; i++)
3938 			s->counters[i] =
3939 			    counter_u64_fetch(V_pf_status.counters[i]);
3940 		for (int i = 0; i < LCNT_MAX; i++)
3941 			s->lcounters[i] =
3942 			    counter_u64_fetch(V_pf_status.lcounters[i]);
3943 		for (int i = 0; i < FCNT_MAX; i++)
3944 			s->fcounters[i] =
3945 			    pf_counter_u64_fetch(&V_pf_status.fcounters[i]);
3946 		for (int i = 0; i < SCNT_MAX; i++)
3947 			s->scounters[i] =
3948 			    counter_u64_fetch(V_pf_status.scounters[i]);
3949 
3950 		bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ);
3951 		bcopy(V_pf_status.pf_chksum, s->pf_chksum,
3952 		    PF_MD5_DIGEST_LENGTH);
3953 
3954 		pfi_update_status(s->ifname, s);
3955 		PF_RULES_RUNLOCK();
3956 		break;
3957 	}
3958 
3959 	case DIOCGETSTATUSNV: {
3960 		error = pf_getstatus((struct pfioc_nv *)addr);
3961 		break;
3962 	}
3963 
3964 	case DIOCSETSTATUSIF: {
3965 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
3966 
3967 		if (pi->ifname[0] == 0) {
3968 			bzero(V_pf_status.ifname, IFNAMSIZ);
3969 			break;
3970 		}
3971 		PF_RULES_WLOCK();
3972 		error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
3973 		PF_RULES_WUNLOCK();
3974 		break;
3975 	}
3976 
3977 	case DIOCCLRSTATUS: {
3978 		PF_RULES_WLOCK();
3979 		for (int i = 0; i < PFRES_MAX; i++)
3980 			counter_u64_zero(V_pf_status.counters[i]);
3981 		for (int i = 0; i < FCNT_MAX; i++)
3982 			pf_counter_u64_zero(&V_pf_status.fcounters[i]);
3983 		for (int i = 0; i < SCNT_MAX; i++)
3984 			counter_u64_zero(V_pf_status.scounters[i]);
3985 		for (int i = 0; i < KLCNT_MAX; i++)
3986 			counter_u64_zero(V_pf_status.lcounters[i]);
3987 		V_pf_status.since = time_second;
3988 		if (*V_pf_status.ifname)
3989 			pfi_update_status(V_pf_status.ifname, NULL);
3990 		PF_RULES_WUNLOCK();
3991 		break;
3992 	}
3993 
3994 	case DIOCNATLOOK: {
3995 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
3996 		struct pf_state_key	*sk;
3997 		struct pf_kstate	*state;
3998 		struct pf_state_key_cmp	 key;
3999 		int			 m = 0, direction = pnl->direction;
4000 		int			 sidx, didx;
4001 
4002 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
4003 		sidx = (direction == PF_IN) ? 1 : 0;
4004 		didx = (direction == PF_IN) ? 0 : 1;
4005 
4006 		if (!pnl->proto ||
4007 		    PF_AZERO(&pnl->saddr, pnl->af) ||
4008 		    PF_AZERO(&pnl->daddr, pnl->af) ||
4009 		    ((pnl->proto == IPPROTO_TCP ||
4010 		    pnl->proto == IPPROTO_UDP) &&
4011 		    (!pnl->dport || !pnl->sport)))
4012 			error = EINVAL;
4013 		else {
4014 			bzero(&key, sizeof(key));
4015 			key.af = pnl->af;
4016 			key.proto = pnl->proto;
4017 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
4018 			key.port[sidx] = pnl->sport;
4019 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
4020 			key.port[didx] = pnl->dport;
4021 
4022 			state = pf_find_state_all(&key, direction, &m);
4023 			if (state == NULL) {
4024 				error = ENOENT;
4025 			} else {
4026 				if (m > 1) {
4027 					PF_STATE_UNLOCK(state);
4028 					error = E2BIG;	/* more than one state */
4029 				} else {
4030 					sk = state->key[sidx];
4031 					PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
4032 					pnl->rsport = sk->port[sidx];
4033 					PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
4034 					pnl->rdport = sk->port[didx];
4035 					PF_STATE_UNLOCK(state);
4036 				}
4037 			}
4038 		}
4039 		break;
4040 	}
4041 
4042 	case DIOCSETTIMEOUT: {
4043 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
4044 		int		 old;
4045 
4046 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
4047 		    pt->seconds < 0) {
4048 			error = EINVAL;
4049 			break;
4050 		}
4051 		PF_RULES_WLOCK();
4052 		old = V_pf_default_rule.timeout[pt->timeout];
4053 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
4054 			pt->seconds = 1;
4055 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
4056 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
4057 			wakeup(pf_purge_thread);
4058 		pt->seconds = old;
4059 		PF_RULES_WUNLOCK();
4060 		break;
4061 	}
4062 
4063 	case DIOCGETTIMEOUT: {
4064 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
4065 
4066 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
4067 			error = EINVAL;
4068 			break;
4069 		}
4070 		PF_RULES_RLOCK();
4071 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
4072 		PF_RULES_RUNLOCK();
4073 		break;
4074 	}
4075 
4076 	case DIOCGETLIMIT: {
4077 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
4078 
4079 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
4080 			error = EINVAL;
4081 			break;
4082 		}
4083 		PF_RULES_RLOCK();
4084 		pl->limit = V_pf_limits[pl->index].limit;
4085 		PF_RULES_RUNLOCK();
4086 		break;
4087 	}
4088 
4089 	case DIOCSETLIMIT: {
4090 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
4091 		int			 old_limit;
4092 
4093 		PF_RULES_WLOCK();
4094 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
4095 		    V_pf_limits[pl->index].zone == NULL) {
4096 			PF_RULES_WUNLOCK();
4097 			error = EINVAL;
4098 			break;
4099 		}
4100 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
4101 		old_limit = V_pf_limits[pl->index].limit;
4102 		V_pf_limits[pl->index].limit = pl->limit;
4103 		pl->limit = old_limit;
4104 		PF_RULES_WUNLOCK();
4105 		break;
4106 	}
4107 
4108 	case DIOCSETDEBUG: {
4109 		u_int32_t	*level = (u_int32_t *)addr;
4110 
4111 		PF_RULES_WLOCK();
4112 		V_pf_status.debug = *level;
4113 		PF_RULES_WUNLOCK();
4114 		break;
4115 	}
4116 
4117 	case DIOCCLRRULECTRS: {
4118 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
4119 		struct pf_kruleset	*ruleset = &pf_main_ruleset;
4120 		struct pf_krule		*rule;
4121 
4122 		PF_RULES_WLOCK();
4123 		TAILQ_FOREACH(rule,
4124 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
4125 			pf_counter_u64_zero(&rule->evaluations);
4126 			for (int i = 0; i < 2; i++) {
4127 				pf_counter_u64_zero(&rule->packets[i]);
4128 				pf_counter_u64_zero(&rule->bytes[i]);
4129 			}
4130 		}
4131 		PF_RULES_WUNLOCK();
4132 		break;
4133 	}
4134 
4135 	case DIOCGIFSPEEDV0:
4136 	case DIOCGIFSPEEDV1: {
4137 		struct pf_ifspeed_v1	*psp = (struct pf_ifspeed_v1 *)addr;
4138 		struct pf_ifspeed_v1	ps;
4139 		struct ifnet		*ifp;
4140 
4141 		if (psp->ifname[0] == '\0') {
4142 			error = EINVAL;
4143 			break;
4144 		}
4145 
4146 		error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ);
4147 		if (error != 0)
4148 			break;
4149 		ifp = ifunit(ps.ifname);
4150 		if (ifp != NULL) {
4151 			psp->baudrate32 =
4152 			    (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
4153 			if (cmd == DIOCGIFSPEEDV1)
4154 				psp->baudrate = ifp->if_baudrate;
4155 		} else {
4156 			error = EINVAL;
4157 		}
4158 		break;
4159 	}
4160 
4161 #ifdef ALTQ
4162 	case DIOCSTARTALTQ: {
4163 		struct pf_altq		*altq;
4164 
4165 		PF_RULES_WLOCK();
4166 		/* enable all altq interfaces on active list */
4167 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4168 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4169 				error = pf_enable_altq(altq);
4170 				if (error != 0)
4171 					break;
4172 			}
4173 		}
4174 		if (error == 0)
4175 			V_pf_altq_running = 1;
4176 		PF_RULES_WUNLOCK();
4177 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
4178 		break;
4179 	}
4180 
4181 	case DIOCSTOPALTQ: {
4182 		struct pf_altq		*altq;
4183 
4184 		PF_RULES_WLOCK();
4185 		/* disable all altq interfaces on active list */
4186 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4187 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4188 				error = pf_disable_altq(altq);
4189 				if (error != 0)
4190 					break;
4191 			}
4192 		}
4193 		if (error == 0)
4194 			V_pf_altq_running = 0;
4195 		PF_RULES_WUNLOCK();
4196 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
4197 		break;
4198 	}
4199 
4200 	case DIOCADDALTQV0:
4201 	case DIOCADDALTQV1: {
4202 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4203 		struct pf_altq		*altq, *a;
4204 		struct ifnet		*ifp;
4205 
4206 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
4207 		error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
4208 		if (error)
4209 			break;
4210 		altq->local_flags = 0;
4211 
4212 		PF_RULES_WLOCK();
4213 		if (pa->ticket != V_ticket_altqs_inactive) {
4214 			PF_RULES_WUNLOCK();
4215 			free(altq, M_PFALTQ);
4216 			error = EBUSY;
4217 			break;
4218 		}
4219 
4220 		/*
4221 		 * if this is for a queue, find the discipline and
4222 		 * copy the necessary fields
4223 		 */
4224 		if (altq->qname[0] != 0) {
4225 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
4226 				PF_RULES_WUNLOCK();
4227 				error = EBUSY;
4228 				free(altq, M_PFALTQ);
4229 				break;
4230 			}
4231 			altq->altq_disc = NULL;
4232 			TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
4233 				if (strncmp(a->ifname, altq->ifname,
4234 				    IFNAMSIZ) == 0) {
4235 					altq->altq_disc = a->altq_disc;
4236 					break;
4237 				}
4238 			}
4239 		}
4240 
4241 		if ((ifp = ifunit(altq->ifname)) == NULL)
4242 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
4243 		else
4244 			error = altq_add(ifp, altq);
4245 
4246 		if (error) {
4247 			PF_RULES_WUNLOCK();
4248 			free(altq, M_PFALTQ);
4249 			break;
4250 		}
4251 
4252 		if (altq->qname[0] != 0)
4253 			TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
4254 		else
4255 			TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
4256 		/* version error check done on import above */
4257 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4258 		PF_RULES_WUNLOCK();
4259 		break;
4260 	}
4261 
4262 	case DIOCGETALTQSV0:
4263 	case DIOCGETALTQSV1: {
4264 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4265 		struct pf_altq		*altq;
4266 
4267 		PF_RULES_RLOCK();
4268 		pa->nr = 0;
4269 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
4270 			pa->nr++;
4271 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
4272 			pa->nr++;
4273 		pa->ticket = V_ticket_altqs_active;
4274 		PF_RULES_RUNLOCK();
4275 		break;
4276 	}
4277 
4278 	case DIOCGETALTQV0:
4279 	case DIOCGETALTQV1: {
4280 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4281 		struct pf_altq		*altq;
4282 
4283 		PF_RULES_RLOCK();
4284 		if (pa->ticket != V_ticket_altqs_active) {
4285 			PF_RULES_RUNLOCK();
4286 			error = EBUSY;
4287 			break;
4288 		}
4289 		altq = pf_altq_get_nth_active(pa->nr);
4290 		if (altq == NULL) {
4291 			PF_RULES_RUNLOCK();
4292 			error = EBUSY;
4293 			break;
4294 		}
4295 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4296 		PF_RULES_RUNLOCK();
4297 		break;
4298 	}
4299 
4300 	case DIOCCHANGEALTQV0:
4301 	case DIOCCHANGEALTQV1:
4302 		/* CHANGEALTQ not supported yet! */
4303 		error = ENODEV;
4304 		break;
4305 
4306 	case DIOCGETQSTATSV0:
4307 	case DIOCGETQSTATSV1: {
4308 		struct pfioc_qstats_v1	*pq = (struct pfioc_qstats_v1 *)addr;
4309 		struct pf_altq		*altq;
4310 		int			 nbytes;
4311 		u_int32_t		 version;
4312 
4313 		PF_RULES_RLOCK();
4314 		if (pq->ticket != V_ticket_altqs_active) {
4315 			PF_RULES_RUNLOCK();
4316 			error = EBUSY;
4317 			break;
4318 		}
4319 		nbytes = pq->nbytes;
4320 		altq = pf_altq_get_nth_active(pq->nr);
4321 		if (altq == NULL) {
4322 			PF_RULES_RUNLOCK();
4323 			error = EBUSY;
4324 			break;
4325 		}
4326 
4327 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
4328 			PF_RULES_RUNLOCK();
4329 			error = ENXIO;
4330 			break;
4331 		}
4332 		PF_RULES_RUNLOCK();
4333 		if (cmd == DIOCGETQSTATSV0)
4334 			version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
4335 		else
4336 			version = pq->version;
4337 		error = altq_getqstats(altq, pq->buf, &nbytes, version);
4338 		if (error == 0) {
4339 			pq->scheduler = altq->scheduler;
4340 			pq->nbytes = nbytes;
4341 		}
4342 		break;
4343 	}
4344 #endif /* ALTQ */
4345 
4346 	case DIOCBEGINADDRS: {
4347 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4348 
4349 		PF_RULES_WLOCK();
4350 		pf_empty_kpool(&V_pf_pabuf);
4351 		pp->ticket = ++V_ticket_pabuf;
4352 		PF_RULES_WUNLOCK();
4353 		break;
4354 	}
4355 
4356 	case DIOCADDADDR: {
4357 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4358 		struct pf_kpooladdr	*pa;
4359 		struct pfi_kkif		*kif = NULL;
4360 
4361 #ifndef INET
4362 		if (pp->af == AF_INET) {
4363 			error = EAFNOSUPPORT;
4364 			break;
4365 		}
4366 #endif /* INET */
4367 #ifndef INET6
4368 		if (pp->af == AF_INET6) {
4369 			error = EAFNOSUPPORT;
4370 			break;
4371 		}
4372 #endif /* INET6 */
4373 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
4374 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
4375 		    pp->addr.addr.type != PF_ADDR_TABLE) {
4376 			error = EINVAL;
4377 			break;
4378 		}
4379 		if (pp->addr.addr.p.dyn != NULL) {
4380 			error = EINVAL;
4381 			break;
4382 		}
4383 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
4384 		error = pf_pooladdr_to_kpooladdr(&pp->addr, pa);
4385 		if (error != 0)
4386 			break;
4387 		if (pa->ifname[0])
4388 			kif = pf_kkif_create(M_WAITOK);
4389 		PF_RULES_WLOCK();
4390 		if (pp->ticket != V_ticket_pabuf) {
4391 			PF_RULES_WUNLOCK();
4392 			if (pa->ifname[0])
4393 				pf_kkif_free(kif);
4394 			free(pa, M_PFRULE);
4395 			error = EBUSY;
4396 			break;
4397 		}
4398 		if (pa->ifname[0]) {
4399 			pa->kif = pfi_kkif_attach(kif, pa->ifname);
4400 			kif = NULL;
4401 			pfi_kkif_ref(pa->kif);
4402 		} else
4403 			pa->kif = NULL;
4404 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
4405 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
4406 			if (pa->ifname[0])
4407 				pfi_kkif_unref(pa->kif);
4408 			PF_RULES_WUNLOCK();
4409 			free(pa, M_PFRULE);
4410 			break;
4411 		}
4412 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
4413 		PF_RULES_WUNLOCK();
4414 		break;
4415 	}
4416 
4417 	case DIOCGETADDRS: {
4418 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4419 		struct pf_kpool		*pool;
4420 		struct pf_kpooladdr	*pa;
4421 
4422 		pp->anchor[sizeof(pp->anchor) - 1] = 0;
4423 		pp->nr = 0;
4424 
4425 		PF_RULES_RLOCK();
4426 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
4427 		    pp->r_num, 0, 1, 0);
4428 		if (pool == NULL) {
4429 			PF_RULES_RUNLOCK();
4430 			error = EBUSY;
4431 			break;
4432 		}
4433 		TAILQ_FOREACH(pa, &pool->list, entries)
4434 			pp->nr++;
4435 		PF_RULES_RUNLOCK();
4436 		break;
4437 	}
4438 
4439 	case DIOCGETADDR: {
4440 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4441 		struct pf_kpool		*pool;
4442 		struct pf_kpooladdr	*pa;
4443 		u_int32_t		 nr = 0;
4444 
4445 		pp->anchor[sizeof(pp->anchor) - 1] = 0;
4446 
4447 		PF_RULES_RLOCK();
4448 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
4449 		    pp->r_num, 0, 1, 1);
4450 		if (pool == NULL) {
4451 			PF_RULES_RUNLOCK();
4452 			error = EBUSY;
4453 			break;
4454 		}
4455 		pa = TAILQ_FIRST(&pool->list);
4456 		while ((pa != NULL) && (nr < pp->nr)) {
4457 			pa = TAILQ_NEXT(pa, entries);
4458 			nr++;
4459 		}
4460 		if (pa == NULL) {
4461 			PF_RULES_RUNLOCK();
4462 			error = EBUSY;
4463 			break;
4464 		}
4465 		pf_kpooladdr_to_pooladdr(pa, &pp->addr);
4466 		pf_addr_copyout(&pp->addr.addr);
4467 		PF_RULES_RUNLOCK();
4468 		break;
4469 	}
4470 
4471 	case DIOCCHANGEADDR: {
4472 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
4473 		struct pf_kpool		*pool;
4474 		struct pf_kpooladdr	*oldpa = NULL, *newpa = NULL;
4475 		struct pf_kruleset	*ruleset;
4476 		struct pfi_kkif		*kif = NULL;
4477 
4478 		pca->anchor[sizeof(pca->anchor) - 1] = 0;
4479 
4480 		if (pca->action < PF_CHANGE_ADD_HEAD ||
4481 		    pca->action > PF_CHANGE_REMOVE) {
4482 			error = EINVAL;
4483 			break;
4484 		}
4485 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
4486 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
4487 		    pca->addr.addr.type != PF_ADDR_TABLE) {
4488 			error = EINVAL;
4489 			break;
4490 		}
4491 		if (pca->addr.addr.p.dyn != NULL) {
4492 			error = EINVAL;
4493 			break;
4494 		}
4495 
4496 		if (pca->action != PF_CHANGE_REMOVE) {
4497 #ifndef INET
4498 			if (pca->af == AF_INET) {
4499 				error = EAFNOSUPPORT;
4500 				break;
4501 			}
4502 #endif /* INET */
4503 #ifndef INET6
4504 			if (pca->af == AF_INET6) {
4505 				error = EAFNOSUPPORT;
4506 				break;
4507 			}
4508 #endif /* INET6 */
4509 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
4510 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
4511 			if (newpa->ifname[0])
4512 				kif = pf_kkif_create(M_WAITOK);
4513 			newpa->kif = NULL;
4514 		}
4515 #define	ERROUT(x)	ERROUT_IOCTL(DIOCCHANGEADDR_error, x)
4516 		PF_RULES_WLOCK();
4517 		ruleset = pf_find_kruleset(pca->anchor);
4518 		if (ruleset == NULL)
4519 			ERROUT(EBUSY);
4520 
4521 		pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
4522 		    pca->r_num, pca->r_last, 1, 1);
4523 		if (pool == NULL)
4524 			ERROUT(EBUSY);
4525 
4526 		if (pca->action != PF_CHANGE_REMOVE) {
4527 			if (newpa->ifname[0]) {
4528 				newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
4529 				pfi_kkif_ref(newpa->kif);
4530 				kif = NULL;
4531 			}
4532 
4533 			switch (newpa->addr.type) {
4534 			case PF_ADDR_DYNIFTL:
4535 				error = pfi_dynaddr_setup(&newpa->addr,
4536 				    pca->af);
4537 				break;
4538 			case PF_ADDR_TABLE:
4539 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
4540 				    newpa->addr.v.tblname);
4541 				if (newpa->addr.p.tbl == NULL)
4542 					error = ENOMEM;
4543 				break;
4544 			}
4545 			if (error)
4546 				goto DIOCCHANGEADDR_error;
4547 		}
4548 
4549 		switch (pca->action) {
4550 		case PF_CHANGE_ADD_HEAD:
4551 			oldpa = TAILQ_FIRST(&pool->list);
4552 			break;
4553 		case PF_CHANGE_ADD_TAIL:
4554 			oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
4555 			break;
4556 		default:
4557 			oldpa = TAILQ_FIRST(&pool->list);
4558 			for (int i = 0; oldpa && i < pca->nr; i++)
4559 				oldpa = TAILQ_NEXT(oldpa, entries);
4560 
4561 			if (oldpa == NULL)
4562 				ERROUT(EINVAL);
4563 		}
4564 
4565 		if (pca->action == PF_CHANGE_REMOVE) {
4566 			TAILQ_REMOVE(&pool->list, oldpa, entries);
4567 			switch (oldpa->addr.type) {
4568 			case PF_ADDR_DYNIFTL:
4569 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
4570 				break;
4571 			case PF_ADDR_TABLE:
4572 				pfr_detach_table(oldpa->addr.p.tbl);
4573 				break;
4574 			}
4575 			if (oldpa->kif)
4576 				pfi_kkif_unref(oldpa->kif);
4577 			free(oldpa, M_PFRULE);
4578 		} else {
4579 			if (oldpa == NULL)
4580 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
4581 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
4582 			    pca->action == PF_CHANGE_ADD_BEFORE)
4583 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
4584 			else
4585 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
4586 				    newpa, entries);
4587 		}
4588 
4589 		pool->cur = TAILQ_FIRST(&pool->list);
4590 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
4591 		PF_RULES_WUNLOCK();
4592 		break;
4593 
4594 #undef ERROUT
4595 DIOCCHANGEADDR_error:
4596 		if (newpa != NULL) {
4597 			if (newpa->kif)
4598 				pfi_kkif_unref(newpa->kif);
4599 			free(newpa, M_PFRULE);
4600 		}
4601 		PF_RULES_WUNLOCK();
4602 		pf_kkif_free(kif);
4603 		break;
4604 	}
4605 
4606 	case DIOCGETRULESETS: {
4607 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
4608 		struct pf_kruleset	*ruleset;
4609 		struct pf_kanchor	*anchor;
4610 
4611 		pr->path[sizeof(pr->path) - 1] = 0;
4612 
4613 		PF_RULES_RLOCK();
4614 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4615 			PF_RULES_RUNLOCK();
4616 			error = ENOENT;
4617 			break;
4618 		}
4619 		pr->nr = 0;
4620 		if (ruleset->anchor == NULL) {
4621 			/* XXX kludge for pf_main_ruleset */
4622 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4623 				if (anchor->parent == NULL)
4624 					pr->nr++;
4625 		} else {
4626 			RB_FOREACH(anchor, pf_kanchor_node,
4627 			    &ruleset->anchor->children)
4628 				pr->nr++;
4629 		}
4630 		PF_RULES_RUNLOCK();
4631 		break;
4632 	}
4633 
4634 	case DIOCGETRULESET: {
4635 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
4636 		struct pf_kruleset	*ruleset;
4637 		struct pf_kanchor	*anchor;
4638 		u_int32_t		 nr = 0;
4639 
4640 		pr->path[sizeof(pr->path) - 1] = 0;
4641 
4642 		PF_RULES_RLOCK();
4643 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4644 			PF_RULES_RUNLOCK();
4645 			error = ENOENT;
4646 			break;
4647 		}
4648 		pr->name[0] = 0;
4649 		if (ruleset->anchor == NULL) {
4650 			/* XXX kludge for pf_main_ruleset */
4651 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4652 				if (anchor->parent == NULL && nr++ == pr->nr) {
4653 					strlcpy(pr->name, anchor->name,
4654 					    sizeof(pr->name));
4655 					break;
4656 				}
4657 		} else {
4658 			RB_FOREACH(anchor, pf_kanchor_node,
4659 			    &ruleset->anchor->children)
4660 				if (nr++ == pr->nr) {
4661 					strlcpy(pr->name, anchor->name,
4662 					    sizeof(pr->name));
4663 					break;
4664 				}
4665 		}
4666 		if (!pr->name[0])
4667 			error = EBUSY;
4668 		PF_RULES_RUNLOCK();
4669 		break;
4670 	}
4671 
4672 	case DIOCRCLRTABLES: {
4673 		struct pfioc_table *io = (struct pfioc_table *)addr;
4674 
4675 		if (io->pfrio_esize != 0) {
4676 			error = ENODEV;
4677 			break;
4678 		}
4679 		PF_RULES_WLOCK();
4680 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
4681 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4682 		PF_RULES_WUNLOCK();
4683 		break;
4684 	}
4685 
4686 	case DIOCRADDTABLES: {
4687 		struct pfioc_table *io = (struct pfioc_table *)addr;
4688 		struct pfr_table *pfrts;
4689 		size_t totlen;
4690 
4691 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4692 			error = ENODEV;
4693 			break;
4694 		}
4695 
4696 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4697 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4698 			error = ENOMEM;
4699 			break;
4700 		}
4701 
4702 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4703 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4704 		    M_TEMP, M_WAITOK);
4705 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4706 		if (error) {
4707 			free(pfrts, M_TEMP);
4708 			break;
4709 		}
4710 		PF_RULES_WLOCK();
4711 		error = pfr_add_tables(pfrts, io->pfrio_size,
4712 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4713 		PF_RULES_WUNLOCK();
4714 		free(pfrts, M_TEMP);
4715 		break;
4716 	}
4717 
4718 	case DIOCRDELTABLES: {
4719 		struct pfioc_table *io = (struct pfioc_table *)addr;
4720 		struct pfr_table *pfrts;
4721 		size_t totlen;
4722 
4723 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4724 			error = ENODEV;
4725 			break;
4726 		}
4727 
4728 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4729 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4730 			error = ENOMEM;
4731 			break;
4732 		}
4733 
4734 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4735 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4736 		    M_TEMP, M_WAITOK);
4737 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4738 		if (error) {
4739 			free(pfrts, M_TEMP);
4740 			break;
4741 		}
4742 		PF_RULES_WLOCK();
4743 		error = pfr_del_tables(pfrts, io->pfrio_size,
4744 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4745 		PF_RULES_WUNLOCK();
4746 		free(pfrts, M_TEMP);
4747 		break;
4748 	}
4749 
4750 	case DIOCRGETTABLES: {
4751 		struct pfioc_table *io = (struct pfioc_table *)addr;
4752 		struct pfr_table *pfrts;
4753 		size_t totlen;
4754 		int n;
4755 
4756 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4757 			error = ENODEV;
4758 			break;
4759 		}
4760 		PF_RULES_RLOCK();
4761 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4762 		if (n < 0) {
4763 			PF_RULES_RUNLOCK();
4764 			error = EINVAL;
4765 			break;
4766 		}
4767 		io->pfrio_size = min(io->pfrio_size, n);
4768 
4769 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4770 
4771 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4772 		    M_TEMP, M_NOWAIT | M_ZERO);
4773 		if (pfrts == NULL) {
4774 			error = ENOMEM;
4775 			PF_RULES_RUNLOCK();
4776 			break;
4777 		}
4778 		error = pfr_get_tables(&io->pfrio_table, pfrts,
4779 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4780 		PF_RULES_RUNLOCK();
4781 		if (error == 0)
4782 			error = copyout(pfrts, io->pfrio_buffer, totlen);
4783 		free(pfrts, M_TEMP);
4784 		break;
4785 	}
4786 
4787 	case DIOCRGETTSTATS: {
4788 		struct pfioc_table *io = (struct pfioc_table *)addr;
4789 		struct pfr_tstats *pfrtstats;
4790 		size_t totlen;
4791 		int n;
4792 
4793 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
4794 			error = ENODEV;
4795 			break;
4796 		}
4797 		PF_TABLE_STATS_LOCK();
4798 		PF_RULES_RLOCK();
4799 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4800 		if (n < 0) {
4801 			PF_RULES_RUNLOCK();
4802 			PF_TABLE_STATS_UNLOCK();
4803 			error = EINVAL;
4804 			break;
4805 		}
4806 		io->pfrio_size = min(io->pfrio_size, n);
4807 
4808 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
4809 		pfrtstats = mallocarray(io->pfrio_size,
4810 		    sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT | M_ZERO);
4811 		if (pfrtstats == NULL) {
4812 			error = ENOMEM;
4813 			PF_RULES_RUNLOCK();
4814 			PF_TABLE_STATS_UNLOCK();
4815 			break;
4816 		}
4817 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
4818 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4819 		PF_RULES_RUNLOCK();
4820 		PF_TABLE_STATS_UNLOCK();
4821 		if (error == 0)
4822 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
4823 		free(pfrtstats, M_TEMP);
4824 		break;
4825 	}
4826 
4827 	case DIOCRCLRTSTATS: {
4828 		struct pfioc_table *io = (struct pfioc_table *)addr;
4829 		struct pfr_table *pfrts;
4830 		size_t totlen;
4831 
4832 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4833 			error = ENODEV;
4834 			break;
4835 		}
4836 
4837 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4838 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4839 			/* We used to count tables and use the minimum required
4840 			 * size, so we didn't fail on overly large requests.
4841 			 * Keep doing so. */
4842 			io->pfrio_size = pf_ioctl_maxcount;
4843 			break;
4844 		}
4845 
4846 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4847 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4848 		    M_TEMP, M_WAITOK);
4849 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4850 		if (error) {
4851 			free(pfrts, M_TEMP);
4852 			break;
4853 		}
4854 
4855 		PF_TABLE_STATS_LOCK();
4856 		PF_RULES_RLOCK();
4857 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
4858 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4859 		PF_RULES_RUNLOCK();
4860 		PF_TABLE_STATS_UNLOCK();
4861 		free(pfrts, M_TEMP);
4862 		break;
4863 	}
4864 
4865 	case DIOCRSETTFLAGS: {
4866 		struct pfioc_table *io = (struct pfioc_table *)addr;
4867 		struct pfr_table *pfrts;
4868 		size_t totlen;
4869 		int n;
4870 
4871 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4872 			error = ENODEV;
4873 			break;
4874 		}
4875 
4876 		PF_RULES_RLOCK();
4877 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4878 		if (n < 0) {
4879 			PF_RULES_RUNLOCK();
4880 			error = EINVAL;
4881 			break;
4882 		}
4883 
4884 		io->pfrio_size = min(io->pfrio_size, n);
4885 		PF_RULES_RUNLOCK();
4886 
4887 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4888 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4889 		    M_TEMP, M_WAITOK);
4890 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4891 		if (error) {
4892 			free(pfrts, M_TEMP);
4893 			break;
4894 		}
4895 		PF_RULES_WLOCK();
4896 		error = pfr_set_tflags(pfrts, io->pfrio_size,
4897 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
4898 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4899 		PF_RULES_WUNLOCK();
4900 		free(pfrts, M_TEMP);
4901 		break;
4902 	}
4903 
4904 	case DIOCRCLRADDRS: {
4905 		struct pfioc_table *io = (struct pfioc_table *)addr;
4906 
4907 		if (io->pfrio_esize != 0) {
4908 			error = ENODEV;
4909 			break;
4910 		}
4911 		PF_RULES_WLOCK();
4912 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
4913 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4914 		PF_RULES_WUNLOCK();
4915 		break;
4916 	}
4917 
4918 	case DIOCRADDADDRS: {
4919 		struct pfioc_table *io = (struct pfioc_table *)addr;
4920 		struct pfr_addr *pfras;
4921 		size_t totlen;
4922 
4923 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4924 			error = ENODEV;
4925 			break;
4926 		}
4927 		if (io->pfrio_size < 0 ||
4928 		    io->pfrio_size > pf_ioctl_maxcount ||
4929 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4930 			error = EINVAL;
4931 			break;
4932 		}
4933 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4934 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4935 		    M_TEMP, M_WAITOK);
4936 		error = copyin(io->pfrio_buffer, pfras, totlen);
4937 		if (error) {
4938 			free(pfras, M_TEMP);
4939 			break;
4940 		}
4941 		PF_RULES_WLOCK();
4942 		error = pfr_add_addrs(&io->pfrio_table, pfras,
4943 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
4944 		    PFR_FLAG_USERIOCTL);
4945 		PF_RULES_WUNLOCK();
4946 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4947 			error = copyout(pfras, io->pfrio_buffer, totlen);
4948 		free(pfras, M_TEMP);
4949 		break;
4950 	}
4951 
4952 	case DIOCRDELADDRS: {
4953 		struct pfioc_table *io = (struct pfioc_table *)addr;
4954 		struct pfr_addr *pfras;
4955 		size_t totlen;
4956 
4957 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4958 			error = ENODEV;
4959 			break;
4960 		}
4961 		if (io->pfrio_size < 0 ||
4962 		    io->pfrio_size > pf_ioctl_maxcount ||
4963 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4964 			error = EINVAL;
4965 			break;
4966 		}
4967 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4968 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4969 		    M_TEMP, M_WAITOK);
4970 		error = copyin(io->pfrio_buffer, pfras, totlen);
4971 		if (error) {
4972 			free(pfras, M_TEMP);
4973 			break;
4974 		}
4975 		PF_RULES_WLOCK();
4976 		error = pfr_del_addrs(&io->pfrio_table, pfras,
4977 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
4978 		    PFR_FLAG_USERIOCTL);
4979 		PF_RULES_WUNLOCK();
4980 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4981 			error = copyout(pfras, io->pfrio_buffer, totlen);
4982 		free(pfras, M_TEMP);
4983 		break;
4984 	}
4985 
4986 	case DIOCRSETADDRS: {
4987 		struct pfioc_table *io = (struct pfioc_table *)addr;
4988 		struct pfr_addr *pfras;
4989 		size_t totlen, count;
4990 
4991 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4992 			error = ENODEV;
4993 			break;
4994 		}
4995 		if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
4996 			error = EINVAL;
4997 			break;
4998 		}
4999 		count = max(io->pfrio_size, io->pfrio_size2);
5000 		if (count > pf_ioctl_maxcount ||
5001 		    WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
5002 			error = EINVAL;
5003 			break;
5004 		}
5005 		totlen = count * sizeof(struct pfr_addr);
5006 		pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
5007 		    M_WAITOK);
5008 		error = copyin(io->pfrio_buffer, pfras, totlen);
5009 		if (error) {
5010 			free(pfras, M_TEMP);
5011 			break;
5012 		}
5013 		PF_RULES_WLOCK();
5014 		error = pfr_set_addrs(&io->pfrio_table, pfras,
5015 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
5016 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
5017 		    PFR_FLAG_USERIOCTL, 0);
5018 		PF_RULES_WUNLOCK();
5019 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
5020 			error = copyout(pfras, io->pfrio_buffer, totlen);
5021 		free(pfras, M_TEMP);
5022 		break;
5023 	}
5024 
5025 	case DIOCRGETADDRS: {
5026 		struct pfioc_table *io = (struct pfioc_table *)addr;
5027 		struct pfr_addr *pfras;
5028 		size_t totlen;
5029 
5030 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5031 			error = ENODEV;
5032 			break;
5033 		}
5034 		if (io->pfrio_size < 0 ||
5035 		    io->pfrio_size > pf_ioctl_maxcount ||
5036 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5037 			error = EINVAL;
5038 			break;
5039 		}
5040 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
5041 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5042 		    M_TEMP, M_WAITOK | M_ZERO);
5043 		PF_RULES_RLOCK();
5044 		error = pfr_get_addrs(&io->pfrio_table, pfras,
5045 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5046 		PF_RULES_RUNLOCK();
5047 		if (error == 0)
5048 			error = copyout(pfras, io->pfrio_buffer, totlen);
5049 		free(pfras, M_TEMP);
5050 		break;
5051 	}
5052 
5053 	case DIOCRGETASTATS: {
5054 		struct pfioc_table *io = (struct pfioc_table *)addr;
5055 		struct pfr_astats *pfrastats;
5056 		size_t totlen;
5057 
5058 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
5059 			error = ENODEV;
5060 			break;
5061 		}
5062 		if (io->pfrio_size < 0 ||
5063 		    io->pfrio_size > pf_ioctl_maxcount ||
5064 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
5065 			error = EINVAL;
5066 			break;
5067 		}
5068 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
5069 		pfrastats = mallocarray(io->pfrio_size,
5070 		    sizeof(struct pfr_astats), M_TEMP, M_WAITOK | M_ZERO);
5071 		PF_RULES_RLOCK();
5072 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
5073 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5074 		PF_RULES_RUNLOCK();
5075 		if (error == 0)
5076 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
5077 		free(pfrastats, M_TEMP);
5078 		break;
5079 	}
5080 
5081 	case DIOCRCLRASTATS: {
5082 		struct pfioc_table *io = (struct pfioc_table *)addr;
5083 		struct pfr_addr *pfras;
5084 		size_t totlen;
5085 
5086 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5087 			error = ENODEV;
5088 			break;
5089 		}
5090 		if (io->pfrio_size < 0 ||
5091 		    io->pfrio_size > pf_ioctl_maxcount ||
5092 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5093 			error = EINVAL;
5094 			break;
5095 		}
5096 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
5097 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5098 		    M_TEMP, M_WAITOK);
5099 		error = copyin(io->pfrio_buffer, pfras, totlen);
5100 		if (error) {
5101 			free(pfras, M_TEMP);
5102 			break;
5103 		}
5104 		PF_RULES_WLOCK();
5105 		error = pfr_clr_astats(&io->pfrio_table, pfras,
5106 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
5107 		    PFR_FLAG_USERIOCTL);
5108 		PF_RULES_WUNLOCK();
5109 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
5110 			error = copyout(pfras, io->pfrio_buffer, totlen);
5111 		free(pfras, M_TEMP);
5112 		break;
5113 	}
5114 
5115 	case DIOCRTSTADDRS: {
5116 		struct pfioc_table *io = (struct pfioc_table *)addr;
5117 		struct pfr_addr *pfras;
5118 		size_t totlen;
5119 
5120 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5121 			error = ENODEV;
5122 			break;
5123 		}
5124 		if (io->pfrio_size < 0 ||
5125 		    io->pfrio_size > pf_ioctl_maxcount ||
5126 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5127 			error = EINVAL;
5128 			break;
5129 		}
5130 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
5131 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5132 		    M_TEMP, M_WAITOK);
5133 		error = copyin(io->pfrio_buffer, pfras, totlen);
5134 		if (error) {
5135 			free(pfras, M_TEMP);
5136 			break;
5137 		}
5138 		PF_RULES_RLOCK();
5139 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
5140 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
5141 		    PFR_FLAG_USERIOCTL);
5142 		PF_RULES_RUNLOCK();
5143 		if (error == 0)
5144 			error = copyout(pfras, io->pfrio_buffer, totlen);
5145 		free(pfras, M_TEMP);
5146 		break;
5147 	}
5148 
5149 	case DIOCRINADEFINE: {
5150 		struct pfioc_table *io = (struct pfioc_table *)addr;
5151 		struct pfr_addr *pfras;
5152 		size_t totlen;
5153 
5154 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5155 			error = ENODEV;
5156 			break;
5157 		}
5158 		if (io->pfrio_size < 0 ||
5159 		    io->pfrio_size > pf_ioctl_maxcount ||
5160 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5161 			error = EINVAL;
5162 			break;
5163 		}
5164 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
5165 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5166 		    M_TEMP, M_WAITOK);
5167 		error = copyin(io->pfrio_buffer, pfras, totlen);
5168 		if (error) {
5169 			free(pfras, M_TEMP);
5170 			break;
5171 		}
5172 		PF_RULES_WLOCK();
5173 		error = pfr_ina_define(&io->pfrio_table, pfras,
5174 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
5175 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5176 		PF_RULES_WUNLOCK();
5177 		free(pfras, M_TEMP);
5178 		break;
5179 	}
5180 
5181 	case DIOCOSFPADD: {
5182 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5183 		PF_RULES_WLOCK();
5184 		error = pf_osfp_add(io);
5185 		PF_RULES_WUNLOCK();
5186 		break;
5187 	}
5188 
5189 	case DIOCOSFPGET: {
5190 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5191 		PF_RULES_RLOCK();
5192 		error = pf_osfp_get(io);
5193 		PF_RULES_RUNLOCK();
5194 		break;
5195 	}
5196 
5197 	case DIOCXBEGIN: {
5198 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5199 		struct pfioc_trans_e	*ioes, *ioe;
5200 		size_t			 totlen;
5201 		int			 i;
5202 
5203 		if (io->esize != sizeof(*ioe)) {
5204 			error = ENODEV;
5205 			break;
5206 		}
5207 		if (io->size < 0 ||
5208 		    io->size > pf_ioctl_maxcount ||
5209 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5210 			error = EINVAL;
5211 			break;
5212 		}
5213 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5214 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5215 		    M_TEMP, M_WAITOK);
5216 		error = copyin(io->array, ioes, totlen);
5217 		if (error) {
5218 			free(ioes, M_TEMP);
5219 			break;
5220 		}
5221 		/* Ensure there's no more ethernet rules to clean up. */
5222 		NET_EPOCH_DRAIN_CALLBACKS();
5223 		PF_RULES_WLOCK();
5224 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5225 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5226 			switch (ioe->rs_num) {
5227 			case PF_RULESET_ETH:
5228 				if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) {
5229 					PF_RULES_WUNLOCK();
5230 					free(ioes, M_TEMP);
5231 					goto fail;
5232 				}
5233 				break;
5234 #ifdef ALTQ
5235 			case PF_RULESET_ALTQ:
5236 				if (ioe->anchor[0]) {
5237 					PF_RULES_WUNLOCK();
5238 					free(ioes, M_TEMP);
5239 					error = EINVAL;
5240 					goto fail;
5241 				}
5242 				if ((error = pf_begin_altq(&ioe->ticket))) {
5243 					PF_RULES_WUNLOCK();
5244 					free(ioes, M_TEMP);
5245 					goto fail;
5246 				}
5247 				break;
5248 #endif /* ALTQ */
5249 			case PF_RULESET_TABLE:
5250 			    {
5251 				struct pfr_table table;
5252 
5253 				bzero(&table, sizeof(table));
5254 				strlcpy(table.pfrt_anchor, ioe->anchor,
5255 				    sizeof(table.pfrt_anchor));
5256 				if ((error = pfr_ina_begin(&table,
5257 				    &ioe->ticket, NULL, 0))) {
5258 					PF_RULES_WUNLOCK();
5259 					free(ioes, M_TEMP);
5260 					goto fail;
5261 				}
5262 				break;
5263 			    }
5264 			default:
5265 				if ((error = pf_begin_rules(&ioe->ticket,
5266 				    ioe->rs_num, ioe->anchor))) {
5267 					PF_RULES_WUNLOCK();
5268 					free(ioes, M_TEMP);
5269 					goto fail;
5270 				}
5271 				break;
5272 			}
5273 		}
5274 		PF_RULES_WUNLOCK();
5275 		error = copyout(ioes, io->array, totlen);
5276 		free(ioes, M_TEMP);
5277 		break;
5278 	}
5279 
5280 	case DIOCXROLLBACK: {
5281 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5282 		struct pfioc_trans_e	*ioe, *ioes;
5283 		size_t			 totlen;
5284 		int			 i;
5285 
5286 		if (io->esize != sizeof(*ioe)) {
5287 			error = ENODEV;
5288 			break;
5289 		}
5290 		if (io->size < 0 ||
5291 		    io->size > pf_ioctl_maxcount ||
5292 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5293 			error = EINVAL;
5294 			break;
5295 		}
5296 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5297 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5298 		    M_TEMP, M_WAITOK);
5299 		error = copyin(io->array, ioes, totlen);
5300 		if (error) {
5301 			free(ioes, M_TEMP);
5302 			break;
5303 		}
5304 		PF_RULES_WLOCK();
5305 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5306 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5307 			switch (ioe->rs_num) {
5308 			case PF_RULESET_ETH:
5309 				if ((error = pf_rollback_eth(ioe->ticket,
5310 				    ioe->anchor))) {
5311 					PF_RULES_WUNLOCK();
5312 					free(ioes, M_TEMP);
5313 					goto fail; /* really bad */
5314 				}
5315 				break;
5316 #ifdef ALTQ
5317 			case PF_RULESET_ALTQ:
5318 				if (ioe->anchor[0]) {
5319 					PF_RULES_WUNLOCK();
5320 					free(ioes, M_TEMP);
5321 					error = EINVAL;
5322 					goto fail;
5323 				}
5324 				if ((error = pf_rollback_altq(ioe->ticket))) {
5325 					PF_RULES_WUNLOCK();
5326 					free(ioes, M_TEMP);
5327 					goto fail; /* really bad */
5328 				}
5329 				break;
5330 #endif /* ALTQ */
5331 			case PF_RULESET_TABLE:
5332 			    {
5333 				struct pfr_table table;
5334 
5335 				bzero(&table, sizeof(table));
5336 				strlcpy(table.pfrt_anchor, ioe->anchor,
5337 				    sizeof(table.pfrt_anchor));
5338 				if ((error = pfr_ina_rollback(&table,
5339 				    ioe->ticket, NULL, 0))) {
5340 					PF_RULES_WUNLOCK();
5341 					free(ioes, M_TEMP);
5342 					goto fail; /* really bad */
5343 				}
5344 				break;
5345 			    }
5346 			default:
5347 				if ((error = pf_rollback_rules(ioe->ticket,
5348 				    ioe->rs_num, ioe->anchor))) {
5349 					PF_RULES_WUNLOCK();
5350 					free(ioes, M_TEMP);
5351 					goto fail; /* really bad */
5352 				}
5353 				break;
5354 			}
5355 		}
5356 		PF_RULES_WUNLOCK();
5357 		free(ioes, M_TEMP);
5358 		break;
5359 	}
5360 
5361 	case DIOCXCOMMIT: {
5362 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5363 		struct pfioc_trans_e	*ioe, *ioes;
5364 		struct pf_kruleset	*rs;
5365 		struct pf_keth_ruleset	*ers;
5366 		size_t			 totlen;
5367 		int			 i;
5368 
5369 		if (io->esize != sizeof(*ioe)) {
5370 			error = ENODEV;
5371 			break;
5372 		}
5373 
5374 		if (io->size < 0 ||
5375 		    io->size > pf_ioctl_maxcount ||
5376 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5377 			error = EINVAL;
5378 			break;
5379 		}
5380 
5381 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5382 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5383 		    M_TEMP, M_WAITOK);
5384 		error = copyin(io->array, ioes, totlen);
5385 		if (error) {
5386 			free(ioes, M_TEMP);
5387 			break;
5388 		}
5389 		PF_RULES_WLOCK();
5390 		/* First makes sure everything will succeed. */
5391 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5392 			ioe->anchor[sizeof(ioe->anchor) - 1] = 0;
5393 			switch (ioe->rs_num) {
5394 			case PF_RULESET_ETH:
5395 				ers = pf_find_keth_ruleset(ioe->anchor);
5396 				if (ers == NULL || ioe->ticket == 0 ||
5397 				    ioe->ticket != ers->inactive.ticket) {
5398 					PF_RULES_WUNLOCK();
5399 					free(ioes, M_TEMP);
5400 					error = EINVAL;
5401 					goto fail;
5402 				}
5403 				break;
5404 #ifdef ALTQ
5405 			case PF_RULESET_ALTQ:
5406 				if (ioe->anchor[0]) {
5407 					PF_RULES_WUNLOCK();
5408 					free(ioes, M_TEMP);
5409 					error = EINVAL;
5410 					goto fail;
5411 				}
5412 				if (!V_altqs_inactive_open || ioe->ticket !=
5413 				    V_ticket_altqs_inactive) {
5414 					PF_RULES_WUNLOCK();
5415 					free(ioes, M_TEMP);
5416 					error = EBUSY;
5417 					goto fail;
5418 				}
5419 				break;
5420 #endif /* ALTQ */
5421 			case PF_RULESET_TABLE:
5422 				rs = pf_find_kruleset(ioe->anchor);
5423 				if (rs == NULL || !rs->topen || ioe->ticket !=
5424 				    rs->tticket) {
5425 					PF_RULES_WUNLOCK();
5426 					free(ioes, M_TEMP);
5427 					error = EBUSY;
5428 					goto fail;
5429 				}
5430 				break;
5431 			default:
5432 				if (ioe->rs_num < 0 || ioe->rs_num >=
5433 				    PF_RULESET_MAX) {
5434 					PF_RULES_WUNLOCK();
5435 					free(ioes, M_TEMP);
5436 					error = EINVAL;
5437 					goto fail;
5438 				}
5439 				rs = pf_find_kruleset(ioe->anchor);
5440 				if (rs == NULL ||
5441 				    !rs->rules[ioe->rs_num].inactive.open ||
5442 				    rs->rules[ioe->rs_num].inactive.ticket !=
5443 				    ioe->ticket) {
5444 					PF_RULES_WUNLOCK();
5445 					free(ioes, M_TEMP);
5446 					error = EBUSY;
5447 					goto fail;
5448 				}
5449 				break;
5450 			}
5451 		}
5452 		/* Now do the commit - no errors should happen here. */
5453 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5454 			switch (ioe->rs_num) {
5455 			case PF_RULESET_ETH:
5456 				if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) {
5457 					PF_RULES_WUNLOCK();
5458 					free(ioes, M_TEMP);
5459 					goto fail; /* really bad */
5460 				}
5461 				break;
5462 #ifdef ALTQ
5463 			case PF_RULESET_ALTQ:
5464 				if ((error = pf_commit_altq(ioe->ticket))) {
5465 					PF_RULES_WUNLOCK();
5466 					free(ioes, M_TEMP);
5467 					goto fail; /* really bad */
5468 				}
5469 				break;
5470 #endif /* ALTQ */
5471 			case PF_RULESET_TABLE:
5472 			    {
5473 				struct pfr_table table;
5474 
5475 				bzero(&table, sizeof(table));
5476 				(void)strlcpy(table.pfrt_anchor, ioe->anchor,
5477 				    sizeof(table.pfrt_anchor));
5478 				if ((error = pfr_ina_commit(&table,
5479 				    ioe->ticket, NULL, NULL, 0))) {
5480 					PF_RULES_WUNLOCK();
5481 					free(ioes, M_TEMP);
5482 					goto fail; /* really bad */
5483 				}
5484 				break;
5485 			    }
5486 			default:
5487 				if ((error = pf_commit_rules(ioe->ticket,
5488 				    ioe->rs_num, ioe->anchor))) {
5489 					PF_RULES_WUNLOCK();
5490 					free(ioes, M_TEMP);
5491 					goto fail; /* really bad */
5492 				}
5493 				break;
5494 			}
5495 		}
5496 		PF_RULES_WUNLOCK();
5497 
5498 		/* Only hook into EtherNet taffic if we've got rules for it. */
5499 		if (! TAILQ_EMPTY(V_pf_keth->active.rules))
5500 			hook_pf_eth();
5501 		else
5502 			dehook_pf_eth();
5503 
5504 		free(ioes, M_TEMP);
5505 		break;
5506 	}
5507 
5508 	case DIOCGETSRCNODES: {
5509 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
5510 		struct pf_srchash	*sh;
5511 		struct pf_ksrc_node	*n;
5512 		struct pf_src_node	*p, *pstore;
5513 		uint32_t		 i, nr = 0;
5514 
5515 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5516 				i++, sh++) {
5517 			PF_HASHROW_LOCK(sh);
5518 			LIST_FOREACH(n, &sh->nodes, entry)
5519 				nr++;
5520 			PF_HASHROW_UNLOCK(sh);
5521 		}
5522 
5523 		psn->psn_len = min(psn->psn_len,
5524 		    sizeof(struct pf_src_node) * nr);
5525 
5526 		if (psn->psn_len == 0) {
5527 			psn->psn_len = sizeof(struct pf_src_node) * nr;
5528 			break;
5529 		}
5530 
5531 		nr = 0;
5532 
5533 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
5534 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5535 		    i++, sh++) {
5536 		    PF_HASHROW_LOCK(sh);
5537 		    LIST_FOREACH(n, &sh->nodes, entry) {
5538 
5539 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
5540 				break;
5541 
5542 			pf_src_node_copy(n, p);
5543 
5544 			p++;
5545 			nr++;
5546 		    }
5547 		    PF_HASHROW_UNLOCK(sh);
5548 		}
5549 		error = copyout(pstore, psn->psn_src_nodes,
5550 		    sizeof(struct pf_src_node) * nr);
5551 		if (error) {
5552 			free(pstore, M_TEMP);
5553 			break;
5554 		}
5555 		psn->psn_len = sizeof(struct pf_src_node) * nr;
5556 		free(pstore, M_TEMP);
5557 		break;
5558 	}
5559 
5560 	case DIOCCLRSRCNODES: {
5561 		pf_clear_srcnodes(NULL);
5562 		pf_purge_expired_src_nodes();
5563 		break;
5564 	}
5565 
5566 	case DIOCKILLSRCNODES:
5567 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
5568 		break;
5569 
5570 #ifdef COMPAT_FREEBSD13
5571 	case DIOCKEEPCOUNTERS_FREEBSD13:
5572 #endif
5573 	case DIOCKEEPCOUNTERS:
5574 		error = pf_keepcounters((struct pfioc_nv *)addr);
5575 		break;
5576 
5577 	case DIOCGETSYNCOOKIES:
5578 		error = pf_get_syncookies((struct pfioc_nv *)addr);
5579 		break;
5580 
5581 	case DIOCSETSYNCOOKIES:
5582 		error = pf_set_syncookies((struct pfioc_nv *)addr);
5583 		break;
5584 
5585 	case DIOCSETHOSTID: {
5586 		u_int32_t	*hostid = (u_int32_t *)addr;
5587 
5588 		PF_RULES_WLOCK();
5589 		if (*hostid == 0)
5590 			V_pf_status.hostid = arc4random();
5591 		else
5592 			V_pf_status.hostid = *hostid;
5593 		PF_RULES_WUNLOCK();
5594 		break;
5595 	}
5596 
5597 	case DIOCOSFPFLUSH:
5598 		PF_RULES_WLOCK();
5599 		pf_osfp_flush();
5600 		PF_RULES_WUNLOCK();
5601 		break;
5602 
5603 	case DIOCIGETIFACES: {
5604 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5605 		struct pfi_kif *ifstore;
5606 		size_t bufsiz;
5607 
5608 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
5609 			error = ENODEV;
5610 			break;
5611 		}
5612 
5613 		if (io->pfiio_size < 0 ||
5614 		    io->pfiio_size > pf_ioctl_maxcount ||
5615 		    WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
5616 			error = EINVAL;
5617 			break;
5618 		}
5619 
5620 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5621 
5622 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
5623 		ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
5624 		    M_TEMP, M_WAITOK | M_ZERO);
5625 
5626 		PF_RULES_RLOCK();
5627 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
5628 		PF_RULES_RUNLOCK();
5629 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
5630 		free(ifstore, M_TEMP);
5631 		break;
5632 	}
5633 
5634 	case DIOCSETIFFLAG: {
5635 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5636 
5637 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5638 
5639 		PF_RULES_WLOCK();
5640 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
5641 		PF_RULES_WUNLOCK();
5642 		break;
5643 	}
5644 
5645 	case DIOCCLRIFFLAG: {
5646 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5647 
5648 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5649 
5650 		PF_RULES_WLOCK();
5651 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
5652 		PF_RULES_WUNLOCK();
5653 		break;
5654 	}
5655 
5656 	case DIOCSETREASS: {
5657 		u_int32_t	*reass = (u_int32_t *)addr;
5658 
5659 		V_pf_status.reass = *reass & (PF_REASS_ENABLED|PF_REASS_NODF);
5660 		/* Removal of DF flag without reassembly enabled is not a
5661 		 * valid combination. Disable reassembly in such case. */
5662 		if (!(V_pf_status.reass & PF_REASS_ENABLED))
5663 			V_pf_status.reass = 0;
5664 		break;
5665 	}
5666 
5667 	default:
5668 		error = ENODEV;
5669 		break;
5670 	}
5671 fail:
5672 	if (sx_xlocked(&V_pf_ioctl_lock))
5673 		sx_xunlock(&V_pf_ioctl_lock);
5674 	CURVNET_RESTORE();
5675 
5676 #undef ERROUT_IOCTL
5677 
5678 	return (error);
5679 }
5680 
5681 void
5682 pfsync_state_export(union pfsync_state_union *sp, struct pf_kstate *st, int msg_version)
5683 {
5684 	bzero(sp, sizeof(union pfsync_state_union));
5685 
5686 	/* copy from state key */
5687 	sp->pfs_1301.key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5688 	sp->pfs_1301.key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5689 	sp->pfs_1301.key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5690 	sp->pfs_1301.key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5691 	sp->pfs_1301.key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5692 	sp->pfs_1301.key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5693 	sp->pfs_1301.key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5694 	sp->pfs_1301.key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5695 	sp->pfs_1301.proto = st->key[PF_SK_WIRE]->proto;
5696 	sp->pfs_1301.af = st->key[PF_SK_WIRE]->af;
5697 
5698 	/* copy from state */
5699 	strlcpy(sp->pfs_1301.ifname, st->kif->pfik_name, sizeof(sp->pfs_1301.ifname));
5700 	bcopy(&st->rt_addr, &sp->pfs_1301.rt_addr, sizeof(sp->pfs_1301.rt_addr));
5701 	sp->pfs_1301.creation = htonl(time_uptime - st->creation);
5702 	sp->pfs_1301.expire = pf_state_expires(st);
5703 	if (sp->pfs_1301.expire <= time_uptime)
5704 		sp->pfs_1301.expire = htonl(0);
5705 	else
5706 		sp->pfs_1301.expire = htonl(sp->pfs_1301.expire - time_uptime);
5707 
5708 	sp->pfs_1301.direction = st->direction;
5709 	sp->pfs_1301.log = st->log;
5710 	sp->pfs_1301.timeout = st->timeout;
5711 
5712 	switch (msg_version) {
5713 		case PFSYNC_MSG_VERSION_1301:
5714 			sp->pfs_1301.state_flags = st->state_flags;
5715 			break;
5716 		case PFSYNC_MSG_VERSION_1400:
5717 			sp->pfs_1400.state_flags = htons(st->state_flags);
5718 			sp->pfs_1400.qid = htons(st->qid);
5719 			sp->pfs_1400.pqid = htons(st->pqid);
5720 			sp->pfs_1400.dnpipe = htons(st->dnpipe);
5721 			sp->pfs_1400.dnrpipe = htons(st->dnrpipe);
5722 			sp->pfs_1400.rtableid = htonl(st->rtableid);
5723 			sp->pfs_1400.min_ttl = st->min_ttl;
5724 			sp->pfs_1400.set_tos = st->set_tos;
5725 			sp->pfs_1400.max_mss = htons(st->max_mss);
5726 			sp->pfs_1400.set_prio[0] = st->set_prio[0];
5727 			sp->pfs_1400.set_prio[1] = st->set_prio[1];
5728 			sp->pfs_1400.rt = st->rt;
5729 			if (st->rt_kif)
5730 				strlcpy(sp->pfs_1400.rt_ifname,
5731 				    st->rt_kif->pfik_name,
5732 				    sizeof(sp->pfs_1400.rt_ifname));
5733 			break;
5734 		default:
5735 			panic("%s: Unsupported pfsync_msg_version %d",
5736 			    __func__, msg_version);
5737 	}
5738 
5739 	if (st->src_node)
5740 		sp->pfs_1301.sync_flags |= PFSYNC_FLAG_SRCNODE;
5741 	if (st->nat_src_node)
5742 		sp->pfs_1301.sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5743 
5744 	sp->pfs_1301.id = st->id;
5745 	sp->pfs_1301.creatorid = st->creatorid;
5746 	pf_state_peer_hton(&st->src, &sp->pfs_1301.src);
5747 	pf_state_peer_hton(&st->dst, &sp->pfs_1301.dst);
5748 
5749 	if (st->rule.ptr == NULL)
5750 		sp->pfs_1301.rule = htonl(-1);
5751 	else
5752 		sp->pfs_1301.rule = htonl(st->rule.ptr->nr);
5753 	if (st->anchor.ptr == NULL)
5754 		sp->pfs_1301.anchor = htonl(-1);
5755 	else
5756 		sp->pfs_1301.anchor = htonl(st->anchor.ptr->nr);
5757 	if (st->nat_rule.ptr == NULL)
5758 		sp->pfs_1301.nat_rule = htonl(-1);
5759 	else
5760 		sp->pfs_1301.nat_rule = htonl(st->nat_rule.ptr->nr);
5761 
5762 	pf_state_counter_hton(st->packets[0], sp->pfs_1301.packets[0]);
5763 	pf_state_counter_hton(st->packets[1], sp->pfs_1301.packets[1]);
5764 	pf_state_counter_hton(st->bytes[0], sp->pfs_1301.bytes[0]);
5765 	pf_state_counter_hton(st->bytes[1], sp->pfs_1301.bytes[1]);
5766 }
5767 
5768 void
5769 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st)
5770 {
5771 	bzero(sp, sizeof(*sp));
5772 
5773 	sp->version = PF_STATE_VERSION;
5774 
5775 	/* copy from state key */
5776 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5777 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5778 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5779 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5780 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5781 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5782 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5783 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5784 	sp->proto = st->key[PF_SK_WIRE]->proto;
5785 	sp->af = st->key[PF_SK_WIRE]->af;
5786 
5787 	/* copy from state */
5788 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
5789 	strlcpy(sp->orig_ifname, st->orig_kif->pfik_name,
5790 	    sizeof(sp->orig_ifname));
5791 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
5792 	sp->creation = htonl(time_uptime - st->creation);
5793 	sp->expire = pf_state_expires(st);
5794 	if (sp->expire <= time_uptime)
5795 		sp->expire = htonl(0);
5796 	else
5797 		sp->expire = htonl(sp->expire - time_uptime);
5798 
5799 	sp->direction = st->direction;
5800 	sp->log = st->log;
5801 	sp->timeout = st->timeout;
5802 	/* 8 bits for the old libpfctl, 16 bits for the new libpfctl */
5803 	sp->state_flags_compat = st->state_flags;
5804 	sp->state_flags = htons(st->state_flags);
5805 	if (st->src_node)
5806 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
5807 	if (st->nat_src_node)
5808 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5809 
5810 	sp->id = st->id;
5811 	sp->creatorid = st->creatorid;
5812 	pf_state_peer_hton(&st->src, &sp->src);
5813 	pf_state_peer_hton(&st->dst, &sp->dst);
5814 
5815 	if (st->rule.ptr == NULL)
5816 		sp->rule = htonl(-1);
5817 	else
5818 		sp->rule = htonl(st->rule.ptr->nr);
5819 	if (st->anchor.ptr == NULL)
5820 		sp->anchor = htonl(-1);
5821 	else
5822 		sp->anchor = htonl(st->anchor.ptr->nr);
5823 	if (st->nat_rule.ptr == NULL)
5824 		sp->nat_rule = htonl(-1);
5825 	else
5826 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
5827 
5828 	sp->packets[0] = st->packets[0];
5829 	sp->packets[1] = st->packets[1];
5830 	sp->bytes[0] = st->bytes[0];
5831 	sp->bytes[1] = st->bytes[1];
5832 
5833 	sp->qid = htons(st->qid);
5834 	sp->pqid = htons(st->pqid);
5835 	sp->dnpipe = htons(st->dnpipe);
5836 	sp->dnrpipe = htons(st->dnrpipe);
5837 	sp->rtableid = htonl(st->rtableid);
5838 	sp->min_ttl = st->min_ttl;
5839 	sp->set_tos = st->set_tos;
5840 	sp->max_mss = htons(st->max_mss);
5841 	sp->rt = st->rt;
5842 	if (st->rt_kif)
5843 		strlcpy(sp->rt_ifname, st->rt_kif->pfik_name,
5844 		    sizeof(sp->rt_ifname));
5845 	sp->set_prio[0] = st->set_prio[0];
5846 	sp->set_prio[1] = st->set_prio[1];
5847 
5848 }
5849 
5850 static void
5851 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
5852 {
5853 	struct pfr_ktable *kt;
5854 
5855 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
5856 
5857 	kt = aw->p.tbl;
5858 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
5859 		kt = kt->pfrkt_root;
5860 	aw->p.tbl = NULL;
5861 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
5862 		kt->pfrkt_cnt : -1;
5863 }
5864 
5865 static int
5866 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters,
5867     size_t number, char **names)
5868 {
5869 	nvlist_t        *nvc;
5870 
5871 	nvc = nvlist_create(0);
5872 	if (nvc == NULL)
5873 		return (ENOMEM);
5874 
5875 	for (int i = 0; i < number; i++) {
5876 		nvlist_append_number_array(nvc, "counters",
5877 		    counter_u64_fetch(counters[i]));
5878 		nvlist_append_string_array(nvc, "names",
5879 		    names[i]);
5880 		nvlist_append_number_array(nvc, "ids",
5881 		    i);
5882 	}
5883 	nvlist_add_nvlist(nvl, name, nvc);
5884 	nvlist_destroy(nvc);
5885 
5886 	return (0);
5887 }
5888 
5889 static int
5890 pf_getstatus(struct pfioc_nv *nv)
5891 {
5892 	nvlist_t        *nvl = NULL, *nvc = NULL;
5893 	void            *nvlpacked = NULL;
5894 	int              error;
5895 	struct pf_status s;
5896 	char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES;
5897 	char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES;
5898 	char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES;
5899 	PF_RULES_RLOCK_TRACKER;
5900 
5901 #define ERROUT(x)      ERROUT_FUNCTION(errout, x)
5902 
5903 	PF_RULES_RLOCK();
5904 
5905 	nvl = nvlist_create(0);
5906 	if (nvl == NULL)
5907 		ERROUT(ENOMEM);
5908 
5909 	nvlist_add_bool(nvl, "running", V_pf_status.running);
5910 	nvlist_add_number(nvl, "since", V_pf_status.since);
5911 	nvlist_add_number(nvl, "debug", V_pf_status.debug);
5912 	nvlist_add_number(nvl, "hostid", V_pf_status.hostid);
5913 	nvlist_add_number(nvl, "states", V_pf_status.states);
5914 	nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes);
5915 	nvlist_add_number(nvl, "reass", V_pf_status.reass);
5916 	nvlist_add_bool(nvl, "syncookies_active",
5917 	    V_pf_status.syncookies_active);
5918 
5919 	/* counters */
5920 	error = pf_add_status_counters(nvl, "counters", V_pf_status.counters,
5921 	    PFRES_MAX, pf_reasons);
5922 	if (error != 0)
5923 		ERROUT(error);
5924 
5925 	/* lcounters */
5926 	error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters,
5927 	    KLCNT_MAX, pf_lcounter);
5928 	if (error != 0)
5929 		ERROUT(error);
5930 
5931 	/* fcounters */
5932 	nvc = nvlist_create(0);
5933 	if (nvc == NULL)
5934 		ERROUT(ENOMEM);
5935 
5936 	for (int i = 0; i < FCNT_MAX; i++) {
5937 		nvlist_append_number_array(nvc, "counters",
5938 		    pf_counter_u64_fetch(&V_pf_status.fcounters[i]));
5939 		nvlist_append_string_array(nvc, "names",
5940 		    pf_fcounter[i]);
5941 		nvlist_append_number_array(nvc, "ids",
5942 		    i);
5943 	}
5944 	nvlist_add_nvlist(nvl, "fcounters", nvc);
5945 	nvlist_destroy(nvc);
5946 	nvc = NULL;
5947 
5948 	/* scounters */
5949 	error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters,
5950 	    SCNT_MAX, pf_fcounter);
5951 	if (error != 0)
5952 		ERROUT(error);
5953 
5954 	nvlist_add_string(nvl, "ifname", V_pf_status.ifname);
5955 	nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum,
5956 	    PF_MD5_DIGEST_LENGTH);
5957 
5958 	pfi_update_status(V_pf_status.ifname, &s);
5959 
5960 	/* pcounters / bcounters */
5961 	for (int i = 0; i < 2; i++) {
5962 		for (int j = 0; j < 2; j++) {
5963 			for (int k = 0; k < 2; k++) {
5964 				nvlist_append_number_array(nvl, "pcounters",
5965 				    s.pcounters[i][j][k]);
5966 			}
5967 			nvlist_append_number_array(nvl, "bcounters",
5968 			    s.bcounters[i][j]);
5969 		}
5970 	}
5971 
5972 	nvlpacked = nvlist_pack(nvl, &nv->len);
5973 	if (nvlpacked == NULL)
5974 		ERROUT(ENOMEM);
5975 
5976 	if (nv->size == 0)
5977 		ERROUT(0);
5978 	else if (nv->size < nv->len)
5979 		ERROUT(ENOSPC);
5980 
5981 	PF_RULES_RUNLOCK();
5982 	error = copyout(nvlpacked, nv->data, nv->len);
5983 	goto done;
5984 
5985 #undef ERROUT
5986 errout:
5987 	PF_RULES_RUNLOCK();
5988 done:
5989 	free(nvlpacked, M_NVLIST);
5990 	nvlist_destroy(nvc);
5991 	nvlist_destroy(nvl);
5992 
5993 	return (error);
5994 }
5995 
5996 /*
5997  * XXX - Check for version mismatch!!!
5998  */
5999 static void
6000 pf_clear_all_states(void)
6001 {
6002 	struct pf_kstate	*s;
6003 	u_int i;
6004 
6005 	for (i = 0; i <= pf_hashmask; i++) {
6006 		struct pf_idhash *ih = &V_pf_idhash[i];
6007 relock:
6008 		PF_HASHROW_LOCK(ih);
6009 		LIST_FOREACH(s, &ih->states, entry) {
6010 			s->timeout = PFTM_PURGE;
6011 			/* Don't send out individual delete messages. */
6012 			s->state_flags |= PFSTATE_NOSYNC;
6013 			pf_unlink_state(s);
6014 			goto relock;
6015 		}
6016 		PF_HASHROW_UNLOCK(ih);
6017 	}
6018 }
6019 
6020 static int
6021 pf_clear_tables(void)
6022 {
6023 	struct pfioc_table io;
6024 	int error;
6025 
6026 	bzero(&io, sizeof(io));
6027 
6028 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
6029 	    io.pfrio_flags);
6030 
6031 	return (error);
6032 }
6033 
6034 static void
6035 pf_clear_srcnodes(struct pf_ksrc_node *n)
6036 {
6037 	struct pf_kstate *s;
6038 	int i;
6039 
6040 	for (i = 0; i <= pf_hashmask; i++) {
6041 		struct pf_idhash *ih = &V_pf_idhash[i];
6042 
6043 		PF_HASHROW_LOCK(ih);
6044 		LIST_FOREACH(s, &ih->states, entry) {
6045 			if (n == NULL || n == s->src_node)
6046 				s->src_node = NULL;
6047 			if (n == NULL || n == s->nat_src_node)
6048 				s->nat_src_node = NULL;
6049 		}
6050 		PF_HASHROW_UNLOCK(ih);
6051 	}
6052 
6053 	if (n == NULL) {
6054 		struct pf_srchash *sh;
6055 
6056 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
6057 		    i++, sh++) {
6058 			PF_HASHROW_LOCK(sh);
6059 			LIST_FOREACH(n, &sh->nodes, entry) {
6060 				n->expire = 1;
6061 				n->states = 0;
6062 			}
6063 			PF_HASHROW_UNLOCK(sh);
6064 		}
6065 	} else {
6066 		/* XXX: hash slot should already be locked here. */
6067 		n->expire = 1;
6068 		n->states = 0;
6069 	}
6070 }
6071 
6072 static void
6073 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
6074 {
6075 	struct pf_ksrc_node_list	 kill;
6076 
6077 	LIST_INIT(&kill);
6078 	for (int i = 0; i <= pf_srchashmask; i++) {
6079 		struct pf_srchash *sh = &V_pf_srchash[i];
6080 		struct pf_ksrc_node *sn, *tmp;
6081 
6082 		PF_HASHROW_LOCK(sh);
6083 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
6084 			if (PF_MATCHA(psnk->psnk_src.neg,
6085 			      &psnk->psnk_src.addr.v.a.addr,
6086 			      &psnk->psnk_src.addr.v.a.mask,
6087 			      &sn->addr, sn->af) &&
6088 			    PF_MATCHA(psnk->psnk_dst.neg,
6089 			      &psnk->psnk_dst.addr.v.a.addr,
6090 			      &psnk->psnk_dst.addr.v.a.mask,
6091 			      &sn->raddr, sn->af)) {
6092 				pf_unlink_src_node(sn);
6093 				LIST_INSERT_HEAD(&kill, sn, entry);
6094 				sn->expire = 1;
6095 			}
6096 		PF_HASHROW_UNLOCK(sh);
6097 	}
6098 
6099 	for (int i = 0; i <= pf_hashmask; i++) {
6100 		struct pf_idhash *ih = &V_pf_idhash[i];
6101 		struct pf_kstate *s;
6102 
6103 		PF_HASHROW_LOCK(ih);
6104 		LIST_FOREACH(s, &ih->states, entry) {
6105 			if (s->src_node && s->src_node->expire == 1)
6106 				s->src_node = NULL;
6107 			if (s->nat_src_node && s->nat_src_node->expire == 1)
6108 				s->nat_src_node = NULL;
6109 		}
6110 		PF_HASHROW_UNLOCK(ih);
6111 	}
6112 
6113 	psnk->psnk_killed = pf_free_src_nodes(&kill);
6114 }
6115 
6116 static int
6117 pf_keepcounters(struct pfioc_nv *nv)
6118 {
6119 	nvlist_t	*nvl = NULL;
6120 	void		*nvlpacked = NULL;
6121 	int		 error = 0;
6122 
6123 #define	ERROUT(x)	ERROUT_FUNCTION(on_error, x)
6124 
6125 	if (nv->len > pf_ioctl_maxcount)
6126 		ERROUT(ENOMEM);
6127 
6128 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6129 	if (nvlpacked == NULL)
6130 		ERROUT(ENOMEM);
6131 
6132 	error = copyin(nv->data, nvlpacked, nv->len);
6133 	if (error)
6134 		ERROUT(error);
6135 
6136 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6137 	if (nvl == NULL)
6138 		ERROUT(EBADMSG);
6139 
6140 	if (! nvlist_exists_bool(nvl, "keep_counters"))
6141 		ERROUT(EBADMSG);
6142 
6143 	V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters");
6144 
6145 on_error:
6146 	nvlist_destroy(nvl);
6147 	free(nvlpacked, M_NVLIST);
6148 	return (error);
6149 }
6150 
6151 static unsigned int
6152 pf_clear_states(const struct pf_kstate_kill *kill)
6153 {
6154 	struct pf_state_key_cmp	 match_key;
6155 	struct pf_kstate	*s;
6156 	struct pfi_kkif	*kif;
6157 	int		 idx;
6158 	unsigned int	 killed = 0, dir;
6159 
6160 	for (unsigned int i = 0; i <= pf_hashmask; i++) {
6161 		struct pf_idhash *ih = &V_pf_idhash[i];
6162 
6163 relock_DIOCCLRSTATES:
6164 		PF_HASHROW_LOCK(ih);
6165 		LIST_FOREACH(s, &ih->states, entry) {
6166 			/* For floating states look at the original kif. */
6167 			kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
6168 
6169 			if (kill->psk_ifname[0] &&
6170 			    strcmp(kill->psk_ifname,
6171 			    kif->pfik_name))
6172 				continue;
6173 
6174 			if (kill->psk_kill_match) {
6175 				bzero(&match_key, sizeof(match_key));
6176 
6177 				if (s->direction == PF_OUT) {
6178 					dir = PF_IN;
6179 					idx = PF_SK_STACK;
6180 				} else {
6181 					dir = PF_OUT;
6182 					idx = PF_SK_WIRE;
6183 				}
6184 
6185 				match_key.af = s->key[idx]->af;
6186 				match_key.proto = s->key[idx]->proto;
6187 				PF_ACPY(&match_key.addr[0],
6188 				    &s->key[idx]->addr[1], match_key.af);
6189 				match_key.port[0] = s->key[idx]->port[1];
6190 				PF_ACPY(&match_key.addr[1],
6191 				    &s->key[idx]->addr[0], match_key.af);
6192 				match_key.port[1] = s->key[idx]->port[0];
6193 			}
6194 
6195 			/*
6196 			 * Don't send out individual
6197 			 * delete messages.
6198 			 */
6199 			s->state_flags |= PFSTATE_NOSYNC;
6200 			pf_unlink_state(s);
6201 			killed++;
6202 
6203 			if (kill->psk_kill_match)
6204 				killed += pf_kill_matching_state(&match_key,
6205 				    dir);
6206 
6207 			goto relock_DIOCCLRSTATES;
6208 		}
6209 		PF_HASHROW_UNLOCK(ih);
6210 	}
6211 
6212 	if (V_pfsync_clear_states_ptr != NULL)
6213 		V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname);
6214 
6215 	return (killed);
6216 }
6217 
6218 static void
6219 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed)
6220 {
6221 	struct pf_kstate	*s;
6222 
6223 	if (kill->psk_pfcmp.id) {
6224 		if (kill->psk_pfcmp.creatorid == 0)
6225 			kill->psk_pfcmp.creatorid = V_pf_status.hostid;
6226 		if ((s = pf_find_state_byid(kill->psk_pfcmp.id,
6227 		    kill->psk_pfcmp.creatorid))) {
6228 			pf_unlink_state(s);
6229 			*killed = 1;
6230 		}
6231 		return;
6232 	}
6233 
6234 	for (unsigned int i = 0; i <= pf_hashmask; i++)
6235 		*killed += pf_killstates_row(kill, &V_pf_idhash[i]);
6236 
6237 	return;
6238 }
6239 
6240 static int
6241 pf_killstates_nv(struct pfioc_nv *nv)
6242 {
6243 	struct pf_kstate_kill	 kill;
6244 	nvlist_t		*nvl = NULL;
6245 	void			*nvlpacked = NULL;
6246 	int			 error = 0;
6247 	unsigned int		 killed = 0;
6248 
6249 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
6250 
6251 	if (nv->len > pf_ioctl_maxcount)
6252 		ERROUT(ENOMEM);
6253 
6254 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6255 	if (nvlpacked == NULL)
6256 		ERROUT(ENOMEM);
6257 
6258 	error = copyin(nv->data, nvlpacked, nv->len);
6259 	if (error)
6260 		ERROUT(error);
6261 
6262 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6263 	if (nvl == NULL)
6264 		ERROUT(EBADMSG);
6265 
6266 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6267 	if (error)
6268 		ERROUT(error);
6269 
6270 	pf_killstates(&kill, &killed);
6271 
6272 	free(nvlpacked, M_NVLIST);
6273 	nvlpacked = NULL;
6274 	nvlist_destroy(nvl);
6275 	nvl = nvlist_create(0);
6276 	if (nvl == NULL)
6277 		ERROUT(ENOMEM);
6278 
6279 	nvlist_add_number(nvl, "killed", killed);
6280 
6281 	nvlpacked = nvlist_pack(nvl, &nv->len);
6282 	if (nvlpacked == NULL)
6283 		ERROUT(ENOMEM);
6284 
6285 	if (nv->size == 0)
6286 		ERROUT(0);
6287 	else if (nv->size < nv->len)
6288 		ERROUT(ENOSPC);
6289 
6290 	error = copyout(nvlpacked, nv->data, nv->len);
6291 
6292 on_error:
6293 	nvlist_destroy(nvl);
6294 	free(nvlpacked, M_NVLIST);
6295 	return (error);
6296 }
6297 
6298 static int
6299 pf_clearstates_nv(struct pfioc_nv *nv)
6300 {
6301 	struct pf_kstate_kill	 kill;
6302 	nvlist_t		*nvl = NULL;
6303 	void			*nvlpacked = NULL;
6304 	int			 error = 0;
6305 	unsigned int		 killed;
6306 
6307 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
6308 
6309 	if (nv->len > pf_ioctl_maxcount)
6310 		ERROUT(ENOMEM);
6311 
6312 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6313 	if (nvlpacked == NULL)
6314 		ERROUT(ENOMEM);
6315 
6316 	error = copyin(nv->data, nvlpacked, nv->len);
6317 	if (error)
6318 		ERROUT(error);
6319 
6320 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6321 	if (nvl == NULL)
6322 		ERROUT(EBADMSG);
6323 
6324 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6325 	if (error)
6326 		ERROUT(error);
6327 
6328 	killed = pf_clear_states(&kill);
6329 
6330 	free(nvlpacked, M_NVLIST);
6331 	nvlpacked = NULL;
6332 	nvlist_destroy(nvl);
6333 	nvl = nvlist_create(0);
6334 	if (nvl == NULL)
6335 		ERROUT(ENOMEM);
6336 
6337 	nvlist_add_number(nvl, "killed", killed);
6338 
6339 	nvlpacked = nvlist_pack(nvl, &nv->len);
6340 	if (nvlpacked == NULL)
6341 		ERROUT(ENOMEM);
6342 
6343 	if (nv->size == 0)
6344 		ERROUT(0);
6345 	else if (nv->size < nv->len)
6346 		ERROUT(ENOSPC);
6347 
6348 	error = copyout(nvlpacked, nv->data, nv->len);
6349 
6350 #undef ERROUT
6351 on_error:
6352 	nvlist_destroy(nvl);
6353 	free(nvlpacked, M_NVLIST);
6354 	return (error);
6355 }
6356 
6357 static int
6358 pf_getstate(struct pfioc_nv *nv)
6359 {
6360 	nvlist_t		*nvl = NULL, *nvls;
6361 	void			*nvlpacked = NULL;
6362 	struct pf_kstate	*s = NULL;
6363 	int			 error = 0;
6364 	uint64_t		 id, creatorid;
6365 
6366 #define ERROUT(x)	ERROUT_FUNCTION(errout, x)
6367 
6368 	if (nv->len > pf_ioctl_maxcount)
6369 		ERROUT(ENOMEM);
6370 
6371 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6372 	if (nvlpacked == NULL)
6373 		ERROUT(ENOMEM);
6374 
6375 	error = copyin(nv->data, nvlpacked, nv->len);
6376 	if (error)
6377 		ERROUT(error);
6378 
6379 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6380 	if (nvl == NULL)
6381 		ERROUT(EBADMSG);
6382 
6383 	PFNV_CHK(pf_nvuint64(nvl, "id", &id));
6384 	PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid));
6385 
6386 	s = pf_find_state_byid(id, creatorid);
6387 	if (s == NULL)
6388 		ERROUT(ENOENT);
6389 
6390 	free(nvlpacked, M_NVLIST);
6391 	nvlpacked = NULL;
6392 	nvlist_destroy(nvl);
6393 	nvl = nvlist_create(0);
6394 	if (nvl == NULL)
6395 		ERROUT(ENOMEM);
6396 
6397 	nvls = pf_state_to_nvstate(s);
6398 	if (nvls == NULL)
6399 		ERROUT(ENOMEM);
6400 
6401 	nvlist_add_nvlist(nvl, "state", nvls);
6402 	nvlist_destroy(nvls);
6403 
6404 	nvlpacked = nvlist_pack(nvl, &nv->len);
6405 	if (nvlpacked == NULL)
6406 		ERROUT(ENOMEM);
6407 
6408 	if (nv->size == 0)
6409 		ERROUT(0);
6410 	else if (nv->size < nv->len)
6411 		ERROUT(ENOSPC);
6412 
6413 	error = copyout(nvlpacked, nv->data, nv->len);
6414 
6415 #undef ERROUT
6416 errout:
6417 	if (s != NULL)
6418 		PF_STATE_UNLOCK(s);
6419 	free(nvlpacked, M_NVLIST);
6420 	nvlist_destroy(nvl);
6421 	return (error);
6422 }
6423 
6424 /*
6425  * XXX - Check for version mismatch!!!
6426  */
6427 
6428 /*
6429  * Duplicate pfctl -Fa operation to get rid of as much as we can.
6430  */
6431 static int
6432 shutdown_pf(void)
6433 {
6434 	int error = 0;
6435 	u_int32_t t[5];
6436 	char nn = '\0';
6437 
6438 	do {
6439 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
6440 		    != 0) {
6441 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
6442 			break;
6443 		}
6444 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
6445 		    != 0) {
6446 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
6447 			break;		/* XXX: rollback? */
6448 		}
6449 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
6450 		    != 0) {
6451 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
6452 			break;		/* XXX: rollback? */
6453 		}
6454 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
6455 		    != 0) {
6456 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
6457 			break;		/* XXX: rollback? */
6458 		}
6459 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
6460 		    != 0) {
6461 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
6462 			break;		/* XXX: rollback? */
6463 		}
6464 
6465 		/* XXX: these should always succeed here */
6466 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
6467 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
6468 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
6469 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
6470 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
6471 
6472 		if ((error = pf_clear_tables()) != 0)
6473 			break;
6474 
6475 		if ((error = pf_begin_eth(&t[0], &nn)) != 0) {
6476 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth\n"));
6477 			break;
6478 		}
6479 		pf_commit_eth(t[0], &nn);
6480 
6481 #ifdef ALTQ
6482 		if ((error = pf_begin_altq(&t[0])) != 0) {
6483 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
6484 			break;
6485 		}
6486 		pf_commit_altq(t[0]);
6487 #endif
6488 
6489 		pf_clear_all_states();
6490 
6491 		pf_clear_srcnodes(NULL);
6492 
6493 		/* status does not use malloced mem so no need to cleanup */
6494 		/* fingerprints and interfaces have their own cleanup code */
6495 	} while(0);
6496 
6497 	return (error);
6498 }
6499 
6500 static pfil_return_t
6501 pf_check_return(int chk, struct mbuf **m)
6502 {
6503 
6504 	switch (chk) {
6505 	case PF_PASS:
6506 		if (*m == NULL)
6507 			return (PFIL_CONSUMED);
6508 		else
6509 			return (PFIL_PASS);
6510 		break;
6511 	default:
6512 		if (*m != NULL) {
6513 			m_freem(*m);
6514 			*m = NULL;
6515 		}
6516 		return (PFIL_DROPPED);
6517 	}
6518 }
6519 
6520 static pfil_return_t
6521 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6522     void *ruleset __unused, struct inpcb *inp)
6523 {
6524 	int chk;
6525 
6526 	chk = pf_test_eth(PF_IN, flags, ifp, m, inp);
6527 
6528 	return (pf_check_return(chk, m));
6529 }
6530 
6531 static pfil_return_t
6532 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6533     void *ruleset __unused, struct inpcb *inp)
6534 {
6535 	int chk;
6536 
6537 	chk = pf_test_eth(PF_OUT, flags, ifp, m, inp);
6538 
6539 	return (pf_check_return(chk, m));
6540 }
6541 
6542 #ifdef INET
6543 static pfil_return_t
6544 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6545     void *ruleset __unused, struct inpcb *inp)
6546 {
6547 	int chk;
6548 
6549 	chk = pf_test(PF_IN, flags, ifp, m, inp, NULL);
6550 
6551 	return (pf_check_return(chk, m));
6552 }
6553 
6554 static pfil_return_t
6555 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6556     void *ruleset __unused,  struct inpcb *inp)
6557 {
6558 	int chk;
6559 
6560 	chk = pf_test(PF_OUT, flags, ifp, m, inp, NULL);
6561 
6562 	return (pf_check_return(chk, m));
6563 }
6564 #endif
6565 
6566 #ifdef INET6
6567 static pfil_return_t
6568 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
6569     void *ruleset __unused,  struct inpcb *inp)
6570 {
6571 	int chk;
6572 
6573 	/*
6574 	 * In case of loopback traffic IPv6 uses the real interface in
6575 	 * order to support scoped addresses. In order to support stateful
6576 	 * filtering we have change this to lo0 as it is the case in IPv4.
6577 	 */
6578 	CURVNET_SET(ifp->if_vnet);
6579 	chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp,
6580 	    m, inp, NULL);
6581 	CURVNET_RESTORE();
6582 
6583 	return (pf_check_return(chk, m));
6584 }
6585 
6586 static pfil_return_t
6587 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
6588     void *ruleset __unused,  struct inpcb *inp)
6589 {
6590 	int chk;
6591 
6592 	CURVNET_SET(ifp->if_vnet);
6593 	chk = pf_test6(PF_OUT, flags, ifp, m, inp, NULL);
6594 	CURVNET_RESTORE();
6595 
6596 	return (pf_check_return(chk, m));
6597 }
6598 #endif /* INET6 */
6599 
6600 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook);
6601 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook);
6602 #define	V_pf_eth_in_hook	VNET(pf_eth_in_hook)
6603 #define	V_pf_eth_out_hook	VNET(pf_eth_out_hook)
6604 
6605 #ifdef INET
6606 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
6607 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
6608 #define	V_pf_ip4_in_hook	VNET(pf_ip4_in_hook)
6609 #define	V_pf_ip4_out_hook	VNET(pf_ip4_out_hook)
6610 #endif
6611 #ifdef INET6
6612 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
6613 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
6614 #define	V_pf_ip6_in_hook	VNET(pf_ip6_in_hook)
6615 #define	V_pf_ip6_out_hook	VNET(pf_ip6_out_hook)
6616 #endif
6617 
6618 static void
6619 hook_pf_eth(void)
6620 {
6621 	struct pfil_hook_args pha = {
6622 		.pa_version = PFIL_VERSION,
6623 		.pa_modname = "pf",
6624 		.pa_type = PFIL_TYPE_ETHERNET,
6625 	};
6626 	struct pfil_link_args pla = {
6627 		.pa_version = PFIL_VERSION,
6628 	};
6629 	int ret __diagused;
6630 
6631 	if (atomic_load_bool(&V_pf_pfil_eth_hooked))
6632 		return;
6633 
6634 	pha.pa_mbuf_chk = pf_eth_check_in;
6635 	pha.pa_flags = PFIL_IN;
6636 	pha.pa_rulname = "eth-in";
6637 	V_pf_eth_in_hook = pfil_add_hook(&pha);
6638 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6639 	pla.pa_head = V_link_pfil_head;
6640 	pla.pa_hook = V_pf_eth_in_hook;
6641 	ret = pfil_link(&pla);
6642 	MPASS(ret == 0);
6643 	pha.pa_mbuf_chk = pf_eth_check_out;
6644 	pha.pa_flags = PFIL_OUT;
6645 	pha.pa_rulname = "eth-out";
6646 	V_pf_eth_out_hook = pfil_add_hook(&pha);
6647 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6648 	pla.pa_head = V_link_pfil_head;
6649 	pla.pa_hook = V_pf_eth_out_hook;
6650 	ret = pfil_link(&pla);
6651 	MPASS(ret == 0);
6652 
6653 	atomic_store_bool(&V_pf_pfil_eth_hooked, true);
6654 }
6655 
6656 static void
6657 hook_pf(void)
6658 {
6659 	struct pfil_hook_args pha = {
6660 		.pa_version = PFIL_VERSION,
6661 		.pa_modname = "pf",
6662 	};
6663 	struct pfil_link_args pla = {
6664 		.pa_version = PFIL_VERSION,
6665 	};
6666 	int ret __diagused;
6667 
6668 	if (atomic_load_bool(&V_pf_pfil_hooked))
6669 		return;
6670 
6671 #ifdef INET
6672 	pha.pa_type = PFIL_TYPE_IP4;
6673 	pha.pa_mbuf_chk = pf_check_in;
6674 	pha.pa_flags = PFIL_IN;
6675 	pha.pa_rulname = "default-in";
6676 	V_pf_ip4_in_hook = pfil_add_hook(&pha);
6677 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6678 	pla.pa_head = V_inet_pfil_head;
6679 	pla.pa_hook = V_pf_ip4_in_hook;
6680 	ret = pfil_link(&pla);
6681 	MPASS(ret == 0);
6682 	pha.pa_mbuf_chk = pf_check_out;
6683 	pha.pa_flags = PFIL_OUT;
6684 	pha.pa_rulname = "default-out";
6685 	V_pf_ip4_out_hook = pfil_add_hook(&pha);
6686 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6687 	pla.pa_head = V_inet_pfil_head;
6688 	pla.pa_hook = V_pf_ip4_out_hook;
6689 	ret = pfil_link(&pla);
6690 	MPASS(ret == 0);
6691 	if (V_pf_filter_local) {
6692 		pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6693 		pla.pa_head = V_inet_local_pfil_head;
6694 		pla.pa_hook = V_pf_ip4_out_hook;
6695 		ret = pfil_link(&pla);
6696 		MPASS(ret == 0);
6697 	}
6698 #endif
6699 #ifdef INET6
6700 	pha.pa_type = PFIL_TYPE_IP6;
6701 	pha.pa_mbuf_chk = pf_check6_in;
6702 	pha.pa_flags = PFIL_IN;
6703 	pha.pa_rulname = "default-in6";
6704 	V_pf_ip6_in_hook = pfil_add_hook(&pha);
6705 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6706 	pla.pa_head = V_inet6_pfil_head;
6707 	pla.pa_hook = V_pf_ip6_in_hook;
6708 	ret = pfil_link(&pla);
6709 	MPASS(ret == 0);
6710 	pha.pa_mbuf_chk = pf_check6_out;
6711 	pha.pa_rulname = "default-out6";
6712 	pha.pa_flags = PFIL_OUT;
6713 	V_pf_ip6_out_hook = pfil_add_hook(&pha);
6714 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6715 	pla.pa_head = V_inet6_pfil_head;
6716 	pla.pa_hook = V_pf_ip6_out_hook;
6717 	ret = pfil_link(&pla);
6718 	MPASS(ret == 0);
6719 	if (V_pf_filter_local) {
6720 		pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6721 		pla.pa_head = V_inet6_local_pfil_head;
6722 		pla.pa_hook = V_pf_ip6_out_hook;
6723 		ret = pfil_link(&pla);
6724 		MPASS(ret == 0);
6725 	}
6726 #endif
6727 
6728 	atomic_store_bool(&V_pf_pfil_hooked, true);
6729 }
6730 
6731 static void
6732 dehook_pf_eth(void)
6733 {
6734 
6735 	if (!atomic_load_bool(&V_pf_pfil_eth_hooked))
6736 		return;
6737 
6738 	pfil_remove_hook(V_pf_eth_in_hook);
6739 	pfil_remove_hook(V_pf_eth_out_hook);
6740 
6741 	atomic_store_bool(&V_pf_pfil_eth_hooked, false);
6742 }
6743 
6744 static void
6745 dehook_pf(void)
6746 {
6747 
6748 	if (!atomic_load_bool(&V_pf_pfil_hooked))
6749 		return;
6750 
6751 #ifdef INET
6752 	pfil_remove_hook(V_pf_ip4_in_hook);
6753 	pfil_remove_hook(V_pf_ip4_out_hook);
6754 #endif
6755 #ifdef INET6
6756 	pfil_remove_hook(V_pf_ip6_in_hook);
6757 	pfil_remove_hook(V_pf_ip6_out_hook);
6758 #endif
6759 
6760 	atomic_store_bool(&V_pf_pfil_hooked, false);
6761 }
6762 
6763 static void
6764 pf_load_vnet(void)
6765 {
6766 	V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
6767 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
6768 
6769 	rm_init_flags(&V_pf_rules_lock, "pf rulesets", RM_RECURSE);
6770 	sx_init(&V_pf_ioctl_lock, "pf ioctl");
6771 
6772 	pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
6773 	    PF_RULE_TAG_HASH_SIZE_DEFAULT);
6774 #ifdef ALTQ
6775 	pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
6776 	    PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
6777 #endif
6778 
6779 	V_pf_keth = &V_pf_main_keth_anchor.ruleset;
6780 
6781 	pfattach_vnet();
6782 	V_pf_vnet_active = 1;
6783 }
6784 
6785 static int
6786 pf_load(void)
6787 {
6788 	int error;
6789 
6790 	sx_init(&pf_end_lock, "pf end thread");
6791 
6792 	pf_mtag_initialize();
6793 
6794 	pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
6795 	if (pf_dev == NULL)
6796 		return (ENOMEM);
6797 
6798 	pf_end_threads = 0;
6799 	error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
6800 	if (error != 0)
6801 		return (error);
6802 
6803 	pfi_initialize();
6804 
6805 	return (0);
6806 }
6807 
6808 static void
6809 pf_unload_vnet(void)
6810 {
6811 	int ret __diagused;
6812 
6813 	V_pf_vnet_active = 0;
6814 	V_pf_status.running = 0;
6815 	dehook_pf();
6816 	dehook_pf_eth();
6817 
6818 	PF_RULES_WLOCK();
6819 	pf_syncookies_cleanup();
6820 	shutdown_pf();
6821 	PF_RULES_WUNLOCK();
6822 
6823 	/* Make sure we've cleaned up ethernet rules before we continue. */
6824 	NET_EPOCH_DRAIN_CALLBACKS();
6825 
6826 	ret = swi_remove(V_pf_swi_cookie);
6827 	MPASS(ret == 0);
6828 	ret = intr_event_destroy(V_pf_swi_ie);
6829 	MPASS(ret == 0);
6830 
6831 	pf_unload_vnet_purge();
6832 
6833 	pf_normalize_cleanup();
6834 	PF_RULES_WLOCK();
6835 	pfi_cleanup_vnet();
6836 	PF_RULES_WUNLOCK();
6837 	pfr_cleanup();
6838 	pf_osfp_flush();
6839 	pf_cleanup();
6840 	if (IS_DEFAULT_VNET(curvnet))
6841 		pf_mtag_cleanup();
6842 
6843 	pf_cleanup_tagset(&V_pf_tags);
6844 #ifdef ALTQ
6845 	pf_cleanup_tagset(&V_pf_qids);
6846 #endif
6847 	uma_zdestroy(V_pf_tag_z);
6848 
6849 #ifdef PF_WANT_32_TO_64_COUNTER
6850 	PF_RULES_WLOCK();
6851 	LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist);
6852 
6853 	MPASS(LIST_EMPTY(&V_pf_allkiflist));
6854 	MPASS(V_pf_allkifcount == 0);
6855 
6856 	LIST_REMOVE(&V_pf_default_rule, allrulelist);
6857 	V_pf_allrulecount--;
6858 	LIST_REMOVE(V_pf_rulemarker, allrulelist);
6859 
6860 	/*
6861 	 * There are known pf rule leaks when running the test suite.
6862 	 */
6863 #ifdef notyet
6864 	MPASS(LIST_EMPTY(&V_pf_allrulelist));
6865 	MPASS(V_pf_allrulecount == 0);
6866 #endif
6867 
6868 	PF_RULES_WUNLOCK();
6869 
6870 	free(V_pf_kifmarker, PFI_MTYPE);
6871 	free(V_pf_rulemarker, M_PFRULE);
6872 #endif
6873 
6874 	/* Free counters last as we updated them during shutdown. */
6875 	pf_counter_u64_deinit(&V_pf_default_rule.evaluations);
6876 	for (int i = 0; i < 2; i++) {
6877 		pf_counter_u64_deinit(&V_pf_default_rule.packets[i]);
6878 		pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]);
6879 	}
6880 	counter_u64_free(V_pf_default_rule.states_cur);
6881 	counter_u64_free(V_pf_default_rule.states_tot);
6882 	counter_u64_free(V_pf_default_rule.src_nodes);
6883 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, V_pf_default_rule.timestamp);
6884 
6885 	for (int i = 0; i < PFRES_MAX; i++)
6886 		counter_u64_free(V_pf_status.counters[i]);
6887 	for (int i = 0; i < KLCNT_MAX; i++)
6888 		counter_u64_free(V_pf_status.lcounters[i]);
6889 	for (int i = 0; i < FCNT_MAX; i++)
6890 		pf_counter_u64_deinit(&V_pf_status.fcounters[i]);
6891 	for (int i = 0; i < SCNT_MAX; i++)
6892 		counter_u64_free(V_pf_status.scounters[i]);
6893 
6894 	rm_destroy(&V_pf_rules_lock);
6895 	sx_destroy(&V_pf_ioctl_lock);
6896 }
6897 
6898 static void
6899 pf_unload(void)
6900 {
6901 
6902 	sx_xlock(&pf_end_lock);
6903 	pf_end_threads = 1;
6904 	while (pf_end_threads < 2) {
6905 		wakeup_one(pf_purge_thread);
6906 		sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
6907 	}
6908 	sx_xunlock(&pf_end_lock);
6909 
6910 	if (pf_dev != NULL)
6911 		destroy_dev(pf_dev);
6912 
6913 	pfi_cleanup();
6914 
6915 	sx_destroy(&pf_end_lock);
6916 }
6917 
6918 static void
6919 vnet_pf_init(void *unused __unused)
6920 {
6921 
6922 	pf_load_vnet();
6923 }
6924 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6925     vnet_pf_init, NULL);
6926 
6927 static void
6928 vnet_pf_uninit(const void *unused __unused)
6929 {
6930 
6931 	pf_unload_vnet();
6932 }
6933 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
6934 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6935     vnet_pf_uninit, NULL);
6936 
6937 static int
6938 pf_modevent(module_t mod, int type, void *data)
6939 {
6940 	int error = 0;
6941 
6942 	switch(type) {
6943 	case MOD_LOAD:
6944 		error = pf_load();
6945 		break;
6946 	case MOD_UNLOAD:
6947 		/* Handled in SYSUNINIT(pf_unload) to ensure it's done after
6948 		 * the vnet_pf_uninit()s */
6949 		break;
6950 	default:
6951 		error = EINVAL;
6952 		break;
6953 	}
6954 
6955 	return (error);
6956 }
6957 
6958 static moduledata_t pf_mod = {
6959 	"pf",
6960 	pf_modevent,
6961 	0
6962 };
6963 
6964 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
6965 MODULE_VERSION(pf, PF_MODVER);
6966