xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision 3331975ab18f88a519b16e5c4781f5924b191eb7)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2001 Daniel Hartmeier
5  * Copyright (c) 2002,2003 Henning Brauer
6  * Copyright (c) 2012 Gleb Smirnoff <glebius@FreeBSD.org>
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  *
13  *    - Redistributions of source code must retain the above copyright
14  *      notice, this list of conditions and the following disclaimer.
15  *    - Redistributions in binary form must reproduce the above
16  *      copyright notice, this list of conditions and the following
17  *      disclaimer in the documentation and/or other materials provided
18  *      with the distribution.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
26  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
28  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
30  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
31  * POSSIBILITY OF SUCH DAMAGE.
32  *
33  * Effort sponsored in part by the Defense Advanced Research Projects
34  * Agency (DARPA) and Air Force Research Laboratory, Air Force
35  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
36  *
37  *	$OpenBSD: pf_ioctl.c,v 1.213 2009/02/15 21:46:12 mbalmer Exp $
38  */
39 
40 #include <sys/cdefs.h>
41 #include "opt_inet.h"
42 #include "opt_inet6.h"
43 #include "opt_bpf.h"
44 #include "opt_pf.h"
45 
46 #include <sys/param.h>
47 #include <sys/_bitset.h>
48 #include <sys/bitset.h>
49 #include <sys/bus.h>
50 #include <sys/conf.h>
51 #include <sys/endian.h>
52 #include <sys/fcntl.h>
53 #include <sys/filio.h>
54 #include <sys/hash.h>
55 #include <sys/interrupt.h>
56 #include <sys/jail.h>
57 #include <sys/kernel.h>
58 #include <sys/kthread.h>
59 #include <sys/lock.h>
60 #include <sys/mbuf.h>
61 #include <sys/module.h>
62 #include <sys/nv.h>
63 #include <sys/proc.h>
64 #include <sys/sdt.h>
65 #include <sys/smp.h>
66 #include <sys/socket.h>
67 #include <sys/sysctl.h>
68 #include <sys/md5.h>
69 #include <sys/ucred.h>
70 
71 #include <net/if.h>
72 #include <net/if_var.h>
73 #include <net/if_private.h>
74 #include <net/vnet.h>
75 #include <net/route.h>
76 #include <net/pfil.h>
77 #include <net/pfvar.h>
78 #include <net/if_pfsync.h>
79 #include <net/if_pflog.h>
80 
81 #include <netinet/in.h>
82 #include <netinet/ip.h>
83 #include <netinet/ip_var.h>
84 #include <netinet6/ip6_var.h>
85 #include <netinet/ip_icmp.h>
86 #include <netpfil/pf/pf_nl.h>
87 #include <netpfil/pf/pf_nv.h>
88 
89 #ifdef INET6
90 #include <netinet/ip6.h>
91 #endif /* INET6 */
92 
93 #ifdef ALTQ
94 #include <net/altq/altq.h>
95 #endif
96 
97 SDT_PROBE_DEFINE3(pf, ioctl, ioctl, error, "int", "int", "int");
98 SDT_PROBE_DEFINE3(pf, ioctl, function, error, "char *", "int", "int");
99 SDT_PROBE_DEFINE2(pf, ioctl, addrule, error, "int", "int");
100 SDT_PROBE_DEFINE2(pf, ioctl, nvchk, error, "int", "int");
101 
102 static struct pf_kpool	*pf_get_kpool(const char *, u_int32_t, u_int8_t,
103 			    u_int32_t, u_int8_t, u_int8_t, u_int8_t, int);
104 
105 static void		 pf_mv_kpool(struct pf_kpalist *, struct pf_kpalist *);
106 static void		 pf_empty_kpool(struct pf_kpalist *);
107 static int		 pfioctl(struct cdev *, u_long, caddr_t, int,
108 			    struct thread *);
109 static int		 pf_begin_eth(uint32_t *, const char *);
110 static void		 pf_rollback_eth_cb(struct epoch_context *);
111 static int		 pf_rollback_eth(uint32_t, const char *);
112 static int		 pf_commit_eth(uint32_t, const char *);
113 static void		 pf_free_eth_rule(struct pf_keth_rule *);
114 #ifdef ALTQ
115 static int		 pf_begin_altq(u_int32_t *);
116 static int		 pf_rollback_altq(u_int32_t);
117 static int		 pf_commit_altq(u_int32_t);
118 static int		 pf_enable_altq(struct pf_altq *);
119 static int		 pf_disable_altq(struct pf_altq *);
120 static uint16_t		 pf_qname2qid(const char *);
121 static void		 pf_qid_unref(uint16_t);
122 #endif /* ALTQ */
123 static int		 pf_begin_rules(u_int32_t *, int, const char *);
124 static int		 pf_rollback_rules(u_int32_t, int, char *);
125 static int		 pf_setup_pfsync_matching(struct pf_kruleset *);
126 static void		 pf_hash_rule_rolling(MD5_CTX *, struct pf_krule *);
127 static void		 pf_hash_rule(struct pf_krule *);
128 static void		 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
129 static int		 pf_commit_rules(u_int32_t, int, char *);
130 static int		 pf_addr_setup(struct pf_kruleset *,
131 			    struct pf_addr_wrap *, sa_family_t);
132 static void		 pf_src_node_copy(const struct pf_ksrc_node *,
133 			    struct pf_src_node *);
134 #ifdef ALTQ
135 static int		 pf_export_kaltq(struct pf_altq *,
136 			    struct pfioc_altq_v1 *, size_t);
137 static int		 pf_import_kaltq(struct pfioc_altq_v1 *,
138 			    struct pf_altq *, size_t);
139 #endif /* ALTQ */
140 
141 VNET_DEFINE(struct pf_krule,	pf_default_rule);
142 
143 static __inline int             pf_krule_compare(struct pf_krule *,
144 				    struct pf_krule *);
145 
146 RB_GENERATE(pf_krule_global, pf_krule, entry_global, pf_krule_compare);
147 
148 #ifdef ALTQ
149 VNET_DEFINE_STATIC(int,		pf_altq_running);
150 #define	V_pf_altq_running	VNET(pf_altq_running)
151 #endif
152 
153 #define	TAGID_MAX	 50000
154 struct pf_tagname {
155 	TAILQ_ENTRY(pf_tagname)	namehash_entries;
156 	TAILQ_ENTRY(pf_tagname)	taghash_entries;
157 	char			name[PF_TAG_NAME_SIZE];
158 	uint16_t		tag;
159 	int			ref;
160 };
161 
162 struct pf_tagset {
163 	TAILQ_HEAD(, pf_tagname)	*namehash;
164 	TAILQ_HEAD(, pf_tagname)	*taghash;
165 	unsigned int			 mask;
166 	uint32_t			 seed;
167 	BITSET_DEFINE(, TAGID_MAX)	 avail;
168 };
169 
170 VNET_DEFINE(struct pf_tagset, pf_tags);
171 #define	V_pf_tags	VNET(pf_tags)
172 static unsigned int	pf_rule_tag_hashsize;
173 #define	PF_RULE_TAG_HASH_SIZE_DEFAULT	128
174 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN,
175     &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT,
176     "Size of pf(4) rule tag hashtable");
177 
178 #ifdef ALTQ
179 VNET_DEFINE(struct pf_tagset, pf_qids);
180 #define	V_pf_qids	VNET(pf_qids)
181 static unsigned int	pf_queue_tag_hashsize;
182 #define	PF_QUEUE_TAG_HASH_SIZE_DEFAULT	128
183 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN,
184     &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT,
185     "Size of pf(4) queue tag hashtable");
186 #endif
187 VNET_DEFINE(uma_zone_t,	 pf_tag_z);
188 #define	V_pf_tag_z		 VNET(pf_tag_z)
189 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
190 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
191 
192 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
193 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
194 #endif
195 
196 VNET_DEFINE_STATIC(bool, pf_filter_local) = false;
197 #define V_pf_filter_local	VNET(pf_filter_local)
198 SYSCTL_BOOL(_net_pf, OID_AUTO, filter_local, CTLFLAG_VNET | CTLFLAG_RW,
199     &VNET_NAME(pf_filter_local), false,
200     "Enable filtering for packets delivered to local network stack");
201 
202 #ifdef PF_DEFAULT_TO_DROP
203 VNET_DEFINE_STATIC(bool, default_to_drop) = true;
204 #else
205 VNET_DEFINE_STATIC(bool, default_to_drop);
206 #endif
207 #define	V_default_to_drop VNET(default_to_drop)
208 SYSCTL_BOOL(_net_pf, OID_AUTO, default_to_drop, CTLFLAG_RDTUN | CTLFLAG_VNET,
209     &VNET_NAME(default_to_drop), false,
210     "Make the default rule drop all packets.");
211 
212 static void		 pf_init_tagset(struct pf_tagset *, unsigned int *,
213 			    unsigned int);
214 static void		 pf_cleanup_tagset(struct pf_tagset *);
215 static uint16_t		 tagname2hashindex(const struct pf_tagset *, const char *);
216 static uint16_t		 tag2hashindex(const struct pf_tagset *, uint16_t);
217 static u_int16_t	 tagname2tag(struct pf_tagset *, const char *);
218 static u_int16_t	 pf_tagname2tag(const char *);
219 static void		 tag_unref(struct pf_tagset *, u_int16_t);
220 
221 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
222 
223 struct cdev *pf_dev;
224 
225 /*
226  * XXX - These are new and need to be checked when moveing to a new version
227  */
228 static void		 pf_clear_all_states(void);
229 static int		 pf_killstates_row(struct pf_kstate_kill *,
230 			    struct pf_idhash *);
231 static int		 pf_killstates_nv(struct pfioc_nv *);
232 static int		 pf_clearstates_nv(struct pfioc_nv *);
233 static int		 pf_getstate(struct pfioc_nv *);
234 static int		 pf_getstatus(struct pfioc_nv *);
235 static int		 pf_clear_tables(void);
236 static void		 pf_kill_srcnodes(struct pfioc_src_node_kill *);
237 static int		 pf_keepcounters(struct pfioc_nv *);
238 static void		 pf_tbladdr_copyout(struct pf_addr_wrap *);
239 
240 /*
241  * Wrapper functions for pfil(9) hooks
242  */
243 static pfil_return_t pf_eth_check_in(struct mbuf **m, struct ifnet *ifp,
244     int flags, void *ruleset __unused, struct inpcb *inp);
245 static pfil_return_t pf_eth_check_out(struct mbuf **m, struct ifnet *ifp,
246     int flags, void *ruleset __unused, struct inpcb *inp);
247 #ifdef INET
248 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp,
249     int flags, void *ruleset __unused, struct inpcb *inp);
250 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp,
251     int flags, void *ruleset __unused, struct inpcb *inp);
252 #endif
253 #ifdef INET6
254 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp,
255     int flags, void *ruleset __unused, struct inpcb *inp);
256 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp,
257     int flags, void *ruleset __unused, struct inpcb *inp);
258 #endif
259 
260 static void		hook_pf_eth(void);
261 static void		hook_pf(void);
262 static void		dehook_pf_eth(void);
263 static void		dehook_pf(void);
264 static int		shutdown_pf(void);
265 static int		pf_load(void);
266 static void		pf_unload(void);
267 
268 static struct cdevsw pf_cdevsw = {
269 	.d_ioctl =	pfioctl,
270 	.d_name =	PF_NAME,
271 	.d_version =	D_VERSION,
272 };
273 
274 VNET_DEFINE_STATIC(bool, pf_pfil_hooked);
275 #define V_pf_pfil_hooked	VNET(pf_pfil_hooked)
276 VNET_DEFINE_STATIC(bool, pf_pfil_eth_hooked);
277 #define V_pf_pfil_eth_hooked	VNET(pf_pfil_eth_hooked)
278 
279 /*
280  * We need a flag that is neither hooked nor running to know when
281  * the VNET is "valid".  We primarily need this to control (global)
282  * external event, e.g., eventhandlers.
283  */
284 VNET_DEFINE(int, pf_vnet_active);
285 #define V_pf_vnet_active	VNET(pf_vnet_active)
286 
287 int pf_end_threads;
288 struct proc *pf_purge_proc;
289 
290 VNET_DEFINE(struct rmlock, pf_rules_lock);
291 VNET_DEFINE_STATIC(struct sx, pf_ioctl_lock);
292 #define	V_pf_ioctl_lock		VNET(pf_ioctl_lock)
293 struct sx			pf_end_lock;
294 
295 /* pfsync */
296 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
297 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
298 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
299 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
300 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
301 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
302 VNET_DEFINE(pflow_export_state_t *, pflow_export_state_ptr);
303 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
304 
305 /* pflog */
306 pflog_packet_t			*pflog_packet_ptr = NULL;
307 
308 /*
309  * Copy a user-provided string, returning an error if truncation would occur.
310  * Avoid scanning past "sz" bytes in the source string since there's no
311  * guarantee that it's nul-terminated.
312  */
313 static int
314 pf_user_strcpy(char *dst, const char *src, size_t sz)
315 {
316 	if (strnlen(src, sz) == sz)
317 		return (EINVAL);
318 	(void)strlcpy(dst, src, sz);
319 	return (0);
320 }
321 
322 static void
323 pfattach_vnet(void)
324 {
325 	u_int32_t *my_timeout = V_pf_default_rule.timeout;
326 
327 	bzero(&V_pf_status, sizeof(V_pf_status));
328 
329 	pf_initialize();
330 	pfr_initialize();
331 	pfi_initialize_vnet();
332 	pf_normalize_init();
333 	pf_syncookies_init();
334 
335 	V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
336 	V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
337 
338 	RB_INIT(&V_pf_anchors);
339 	pf_init_kruleset(&pf_main_ruleset);
340 
341 	pf_init_keth(V_pf_keth);
342 
343 	/* default rule should never be garbage collected */
344 	V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
345 	V_pf_default_rule.action = V_default_to_drop ? PF_DROP : PF_PASS;
346 	V_pf_default_rule.nr = -1;
347 	V_pf_default_rule.rtableid = -1;
348 
349 	pf_counter_u64_init(&V_pf_default_rule.evaluations, M_WAITOK);
350 	for (int i = 0; i < 2; i++) {
351 		pf_counter_u64_init(&V_pf_default_rule.packets[i], M_WAITOK);
352 		pf_counter_u64_init(&V_pf_default_rule.bytes[i], M_WAITOK);
353 	}
354 	V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
355 	V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
356 	V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
357 
358 	V_pf_default_rule.timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
359 	    M_WAITOK | M_ZERO);
360 
361 #ifdef PF_WANT_32_TO_64_COUNTER
362 	V_pf_kifmarker = malloc(sizeof(*V_pf_kifmarker), PFI_MTYPE, M_WAITOK | M_ZERO);
363 	V_pf_rulemarker = malloc(sizeof(*V_pf_rulemarker), M_PFRULE, M_WAITOK | M_ZERO);
364 	PF_RULES_WLOCK();
365 	LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist);
366 	LIST_INSERT_HEAD(&V_pf_allrulelist, &V_pf_default_rule, allrulelist);
367 	V_pf_allrulecount++;
368 	LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist);
369 	PF_RULES_WUNLOCK();
370 #endif
371 
372 	/* initialize default timeouts */
373 	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
374 	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
375 	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
376 	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
377 	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
378 	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
379 	my_timeout[PFTM_SCTP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
380 	my_timeout[PFTM_SCTP_OPENING] = PFTM_TCP_OPENING_VAL;
381 	my_timeout[PFTM_SCTP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
382 	my_timeout[PFTM_SCTP_CLOSING] = PFTM_TCP_CLOSING_VAL;
383 	my_timeout[PFTM_SCTP_CLOSED] = PFTM_TCP_CLOSED_VAL;
384 	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
385 	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
386 	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
387 	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
388 	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
389 	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
390 	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
391 	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
392 	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
393 	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
394 	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
395 	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
396 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
397 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
398 
399 	V_pf_status.debug = PF_DEBUG_URGENT;
400 	/*
401 	 * XXX This is different than in OpenBSD where reassembly is enabled by
402 	 * defult. In FreeBSD we expect people to still use scrub rules and
403 	 * switch to the new syntax later. Only when they switch they must
404 	 * explicitly enable reassemle. We could change the default once the
405 	 * scrub rule functionality is hopefully removed some day in future.
406 	 */
407 	V_pf_status.reass = 0;
408 
409 	V_pf_pfil_hooked = false;
410 	V_pf_pfil_eth_hooked = false;
411 
412 	/* XXX do our best to avoid a conflict */
413 	V_pf_status.hostid = arc4random();
414 
415 	for (int i = 0; i < PFRES_MAX; i++)
416 		V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
417 	for (int i = 0; i < KLCNT_MAX; i++)
418 		V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
419 	for (int i = 0; i < FCNT_MAX; i++)
420 		pf_counter_u64_init(&V_pf_status.fcounters[i], M_WAITOK);
421 	for (int i = 0; i < SCNT_MAX; i++)
422 		V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
423 
424 	if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET,
425 	    INTR_MPSAFE, &V_pf_swi_cookie) != 0)
426 		/* XXXGL: leaked all above. */
427 		return;
428 }
429 
430 static struct pf_kpool *
431 pf_get_kpool(const char *anchor, u_int32_t ticket, u_int8_t rule_action,
432     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
433     u_int8_t check_ticket, int which)
434 {
435 	struct pf_kruleset	*ruleset;
436 	struct pf_krule		*rule;
437 	int			 rs_num;
438 
439 	MPASS(which == PF_RDR || which == PF_NAT);
440 
441 	ruleset = pf_find_kruleset(anchor);
442 	if (ruleset == NULL)
443 		return (NULL);
444 	rs_num = pf_get_ruleset_number(rule_action);
445 	if (rs_num >= PF_RULESET_MAX)
446 		return (NULL);
447 	if (active) {
448 		if (check_ticket && ticket !=
449 		    ruleset->rules[rs_num].active.ticket)
450 			return (NULL);
451 		if (r_last)
452 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
453 			    pf_krulequeue);
454 		else
455 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
456 	} else {
457 		if (check_ticket && ticket !=
458 		    ruleset->rules[rs_num].inactive.ticket)
459 			return (NULL);
460 		if (r_last)
461 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
462 			    pf_krulequeue);
463 		else
464 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
465 	}
466 	if (!r_last) {
467 		while ((rule != NULL) && (rule->nr != rule_number))
468 			rule = TAILQ_NEXT(rule, entries);
469 	}
470 	if (rule == NULL)
471 		return (NULL);
472 
473 	if (which == PF_NAT)
474 		return (&rule->nat);
475 	else
476 		return (&rule->rdr);
477 }
478 
479 static void
480 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb)
481 {
482 	struct pf_kpooladdr	*mv_pool_pa;
483 
484 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
485 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
486 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
487 	}
488 }
489 
490 static void
491 pf_empty_kpool(struct pf_kpalist *poola)
492 {
493 	struct pf_kpooladdr *pa;
494 
495 	while ((pa = TAILQ_FIRST(poola)) != NULL) {
496 		switch (pa->addr.type) {
497 		case PF_ADDR_DYNIFTL:
498 			pfi_dynaddr_remove(pa->addr.p.dyn);
499 			break;
500 		case PF_ADDR_TABLE:
501 			/* XXX: this could be unfinished pooladdr on pabuf */
502 			if (pa->addr.p.tbl != NULL)
503 				pfr_detach_table(pa->addr.p.tbl);
504 			break;
505 		}
506 		if (pa->kif)
507 			pfi_kkif_unref(pa->kif);
508 		TAILQ_REMOVE(poola, pa, entries);
509 		free(pa, M_PFRULE);
510 	}
511 }
512 
513 static void
514 pf_unlink_rule_locked(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
515 {
516 
517 	PF_RULES_WASSERT();
518 	PF_UNLNKDRULES_ASSERT();
519 
520 	TAILQ_REMOVE(rulequeue, rule, entries);
521 
522 	rule->rule_ref |= PFRULE_REFS;
523 	TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
524 }
525 
526 static void
527 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
528 {
529 
530 	PF_RULES_WASSERT();
531 
532 	PF_UNLNKDRULES_LOCK();
533 	pf_unlink_rule_locked(rulequeue, rule);
534 	PF_UNLNKDRULES_UNLOCK();
535 }
536 
537 static void
538 pf_free_eth_rule(struct pf_keth_rule *rule)
539 {
540 	PF_RULES_WASSERT();
541 
542 	if (rule == NULL)
543 		return;
544 
545 	if (rule->tag)
546 		tag_unref(&V_pf_tags, rule->tag);
547 	if (rule->match_tag)
548 		tag_unref(&V_pf_tags, rule->match_tag);
549 #ifdef ALTQ
550 	pf_qid_unref(rule->qid);
551 #endif
552 
553 	if (rule->bridge_to)
554 		pfi_kkif_unref(rule->bridge_to);
555 	if (rule->kif)
556 		pfi_kkif_unref(rule->kif);
557 
558 	if (rule->ipsrc.addr.type == PF_ADDR_TABLE)
559 		pfr_detach_table(rule->ipsrc.addr.p.tbl);
560 	if (rule->ipdst.addr.type == PF_ADDR_TABLE)
561 		pfr_detach_table(rule->ipdst.addr.p.tbl);
562 
563 	counter_u64_free(rule->evaluations);
564 	for (int i = 0; i < 2; i++) {
565 		counter_u64_free(rule->packets[i]);
566 		counter_u64_free(rule->bytes[i]);
567 	}
568 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
569 	pf_keth_anchor_remove(rule);
570 
571 	free(rule, M_PFRULE);
572 }
573 
574 void
575 pf_free_rule(struct pf_krule *rule)
576 {
577 
578 	PF_RULES_WASSERT();
579 	PF_CONFIG_ASSERT();
580 
581 	if (rule->tag)
582 		tag_unref(&V_pf_tags, rule->tag);
583 	if (rule->match_tag)
584 		tag_unref(&V_pf_tags, rule->match_tag);
585 #ifdef ALTQ
586 	if (rule->pqid != rule->qid)
587 		pf_qid_unref(rule->pqid);
588 	pf_qid_unref(rule->qid);
589 #endif
590 	switch (rule->src.addr.type) {
591 	case PF_ADDR_DYNIFTL:
592 		pfi_dynaddr_remove(rule->src.addr.p.dyn);
593 		break;
594 	case PF_ADDR_TABLE:
595 		pfr_detach_table(rule->src.addr.p.tbl);
596 		break;
597 	}
598 	switch (rule->dst.addr.type) {
599 	case PF_ADDR_DYNIFTL:
600 		pfi_dynaddr_remove(rule->dst.addr.p.dyn);
601 		break;
602 	case PF_ADDR_TABLE:
603 		pfr_detach_table(rule->dst.addr.p.tbl);
604 		break;
605 	}
606 	if (rule->overload_tbl)
607 		pfr_detach_table(rule->overload_tbl);
608 	if (rule->kif)
609 		pfi_kkif_unref(rule->kif);
610 	if (rule->rcv_kif)
611 		pfi_kkif_unref(rule->rcv_kif);
612 	pf_kanchor_remove(rule);
613 	pf_empty_kpool(&rule->rdr.list);
614 	pf_empty_kpool(&rule->nat.list);
615 
616 	pf_krule_free(rule);
617 }
618 
619 static void
620 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
621     unsigned int default_size)
622 {
623 	unsigned int i;
624 	unsigned int hashsize;
625 
626 	if (*tunable_size == 0 || !powerof2(*tunable_size))
627 		*tunable_size = default_size;
628 
629 	hashsize = *tunable_size;
630 	ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
631 	    M_WAITOK);
632 	ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
633 	    M_WAITOK);
634 	ts->mask = hashsize - 1;
635 	ts->seed = arc4random();
636 	for (i = 0; i < hashsize; i++) {
637 		TAILQ_INIT(&ts->namehash[i]);
638 		TAILQ_INIT(&ts->taghash[i]);
639 	}
640 	BIT_FILL(TAGID_MAX, &ts->avail);
641 }
642 
643 static void
644 pf_cleanup_tagset(struct pf_tagset *ts)
645 {
646 	unsigned int i;
647 	unsigned int hashsize;
648 	struct pf_tagname *t, *tmp;
649 
650 	/*
651 	 * Only need to clean up one of the hashes as each tag is hashed
652 	 * into each table.
653 	 */
654 	hashsize = ts->mask + 1;
655 	for (i = 0; i < hashsize; i++)
656 		TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
657 			uma_zfree(V_pf_tag_z, t);
658 
659 	free(ts->namehash, M_PFHASH);
660 	free(ts->taghash, M_PFHASH);
661 }
662 
663 static uint16_t
664 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
665 {
666 	size_t len;
667 
668 	len = strnlen(tagname, PF_TAG_NAME_SIZE - 1);
669 	return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask);
670 }
671 
672 static uint16_t
673 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
674 {
675 
676 	return (tag & ts->mask);
677 }
678 
679 static u_int16_t
680 tagname2tag(struct pf_tagset *ts, const char *tagname)
681 {
682 	struct pf_tagname	*tag;
683 	u_int32_t		 index;
684 	u_int16_t		 new_tagid;
685 
686 	PF_RULES_WASSERT();
687 
688 	index = tagname2hashindex(ts, tagname);
689 	TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
690 		if (strcmp(tagname, tag->name) == 0) {
691 			tag->ref++;
692 			return (tag->tag);
693 		}
694 
695 	/*
696 	 * new entry
697 	 *
698 	 * to avoid fragmentation, we do a linear search from the beginning
699 	 * and take the first free slot we find.
700 	 */
701 	new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
702 	/*
703 	 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
704 	 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
705 	 * set.  It may also return a bit number greater than TAGID_MAX due
706 	 * to rounding of the number of bits in the vector up to a multiple
707 	 * of the vector word size at declaration/allocation time.
708 	 */
709 	if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
710 		return (0);
711 
712 	/* Mark the tag as in use.  Bits are 0-based for BIT_CLR() */
713 	BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
714 
715 	/* allocate and fill new struct pf_tagname */
716 	tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
717 	if (tag == NULL)
718 		return (0);
719 	strlcpy(tag->name, tagname, sizeof(tag->name));
720 	tag->tag = new_tagid;
721 	tag->ref = 1;
722 
723 	/* Insert into namehash */
724 	TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
725 
726 	/* Insert into taghash */
727 	index = tag2hashindex(ts, new_tagid);
728 	TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
729 
730 	return (tag->tag);
731 }
732 
733 static void
734 tag_unref(struct pf_tagset *ts, u_int16_t tag)
735 {
736 	struct pf_tagname	*t;
737 	uint16_t		 index;
738 
739 	PF_RULES_WASSERT();
740 
741 	index = tag2hashindex(ts, tag);
742 	TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
743 		if (tag == t->tag) {
744 			if (--t->ref == 0) {
745 				TAILQ_REMOVE(&ts->taghash[index], t,
746 				    taghash_entries);
747 				index = tagname2hashindex(ts, t->name);
748 				TAILQ_REMOVE(&ts->namehash[index], t,
749 				    namehash_entries);
750 				/* Bits are 0-based for BIT_SET() */
751 				BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
752 				uma_zfree(V_pf_tag_z, t);
753 			}
754 			break;
755 		}
756 }
757 
758 static uint16_t
759 pf_tagname2tag(const char *tagname)
760 {
761 	return (tagname2tag(&V_pf_tags, tagname));
762 }
763 
764 static int
765 pf_begin_eth(uint32_t *ticket, const char *anchor)
766 {
767 	struct pf_keth_rule *rule, *tmp;
768 	struct pf_keth_ruleset *rs;
769 
770 	PF_RULES_WASSERT();
771 
772 	rs = pf_find_or_create_keth_ruleset(anchor);
773 	if (rs == NULL)
774 		return (EINVAL);
775 
776 	/* Purge old inactive rules. */
777 	TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
778 	    tmp) {
779 		TAILQ_REMOVE(rs->inactive.rules, rule,
780 		    entries);
781 		pf_free_eth_rule(rule);
782 	}
783 
784 	*ticket = ++rs->inactive.ticket;
785 	rs->inactive.open = 1;
786 
787 	return (0);
788 }
789 
790 static void
791 pf_rollback_eth_cb(struct epoch_context *ctx)
792 {
793 	struct pf_keth_ruleset *rs;
794 
795 	rs = __containerof(ctx, struct pf_keth_ruleset, epoch_ctx);
796 
797 	CURVNET_SET(rs->vnet);
798 
799 	PF_RULES_WLOCK();
800 	pf_rollback_eth(rs->inactive.ticket,
801 	    rs->anchor ? rs->anchor->path : "");
802 	PF_RULES_WUNLOCK();
803 
804 	CURVNET_RESTORE();
805 }
806 
807 static int
808 pf_rollback_eth(uint32_t ticket, const char *anchor)
809 {
810 	struct pf_keth_rule *rule, *tmp;
811 	struct pf_keth_ruleset *rs;
812 
813 	PF_RULES_WASSERT();
814 
815 	rs = pf_find_keth_ruleset(anchor);
816 	if (rs == NULL)
817 		return (EINVAL);
818 
819 	if (!rs->inactive.open ||
820 	    ticket != rs->inactive.ticket)
821 		return (0);
822 
823 	/* Purge old inactive rules. */
824 	TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
825 	    tmp) {
826 		TAILQ_REMOVE(rs->inactive.rules, rule, entries);
827 		pf_free_eth_rule(rule);
828 	}
829 
830 	rs->inactive.open = 0;
831 
832 	pf_remove_if_empty_keth_ruleset(rs);
833 
834 	return (0);
835 }
836 
837 #define	PF_SET_SKIP_STEPS(i)					\
838 	do {							\
839 		while (head[i] != cur) {			\
840 			head[i]->skip[i].ptr = cur;		\
841 			head[i] = TAILQ_NEXT(head[i], entries);	\
842 		}						\
843 	} while (0)
844 
845 static void
846 pf_eth_calc_skip_steps(struct pf_keth_ruleq *rules)
847 {
848 	struct pf_keth_rule *cur, *prev, *head[PFE_SKIP_COUNT];
849 	int i;
850 
851 	cur = TAILQ_FIRST(rules);
852 	prev = cur;
853 	for (i = 0; i < PFE_SKIP_COUNT; ++i)
854 		head[i] = cur;
855 	while (cur != NULL) {
856 		if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
857 			PF_SET_SKIP_STEPS(PFE_SKIP_IFP);
858 		if (cur->direction != prev->direction)
859 			PF_SET_SKIP_STEPS(PFE_SKIP_DIR);
860 		if (cur->proto != prev->proto)
861 			PF_SET_SKIP_STEPS(PFE_SKIP_PROTO);
862 		if (memcmp(&cur->src, &prev->src, sizeof(cur->src)) != 0)
863 			PF_SET_SKIP_STEPS(PFE_SKIP_SRC_ADDR);
864 		if (memcmp(&cur->dst, &prev->dst, sizeof(cur->dst)) != 0)
865 			PF_SET_SKIP_STEPS(PFE_SKIP_DST_ADDR);
866 		if (cur->ipsrc.neg != prev->ipsrc.neg ||
867 		    pf_addr_wrap_neq(&cur->ipsrc.addr, &prev->ipsrc.addr))
868 			PF_SET_SKIP_STEPS(PFE_SKIP_SRC_IP_ADDR);
869 		if (cur->ipdst.neg != prev->ipdst.neg ||
870 		    pf_addr_wrap_neq(&cur->ipdst.addr, &prev->ipdst.addr))
871 			PF_SET_SKIP_STEPS(PFE_SKIP_DST_IP_ADDR);
872 
873 		prev = cur;
874 		cur = TAILQ_NEXT(cur, entries);
875 	}
876 	for (i = 0; i < PFE_SKIP_COUNT; ++i)
877 		PF_SET_SKIP_STEPS(i);
878 }
879 
880 static int
881 pf_commit_eth(uint32_t ticket, const char *anchor)
882 {
883 	struct pf_keth_ruleq *rules;
884 	struct pf_keth_ruleset *rs;
885 
886 	rs = pf_find_keth_ruleset(anchor);
887 	if (rs == NULL) {
888 		return (EINVAL);
889 	}
890 
891 	if (!rs->inactive.open ||
892 	    ticket != rs->inactive.ticket)
893 		return (EBUSY);
894 
895 	PF_RULES_WASSERT();
896 
897 	pf_eth_calc_skip_steps(rs->inactive.rules);
898 
899 	rules = rs->active.rules;
900 	ck_pr_store_ptr(&rs->active.rules, rs->inactive.rules);
901 	rs->inactive.rules = rules;
902 	rs->inactive.ticket = rs->active.ticket;
903 
904 	/* Clean up inactive rules (i.e. previously active rules), only when
905 	 * we're sure they're no longer used. */
906 	NET_EPOCH_CALL(pf_rollback_eth_cb, &rs->epoch_ctx);
907 
908 	return (0);
909 }
910 
911 #ifdef ALTQ
912 static uint16_t
913 pf_qname2qid(const char *qname)
914 {
915 	return (tagname2tag(&V_pf_qids, qname));
916 }
917 
918 static void
919 pf_qid_unref(uint16_t qid)
920 {
921 	tag_unref(&V_pf_qids, qid);
922 }
923 
924 static int
925 pf_begin_altq(u_int32_t *ticket)
926 {
927 	struct pf_altq	*altq, *tmp;
928 	int		 error = 0;
929 
930 	PF_RULES_WASSERT();
931 
932 	/* Purge the old altq lists */
933 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
934 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
935 			/* detach and destroy the discipline */
936 			error = altq_remove(altq);
937 		}
938 		free(altq, M_PFALTQ);
939 	}
940 	TAILQ_INIT(V_pf_altq_ifs_inactive);
941 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
942 		pf_qid_unref(altq->qid);
943 		free(altq, M_PFALTQ);
944 	}
945 	TAILQ_INIT(V_pf_altqs_inactive);
946 	if (error)
947 		return (error);
948 	*ticket = ++V_ticket_altqs_inactive;
949 	V_altqs_inactive_open = 1;
950 	return (0);
951 }
952 
953 static int
954 pf_rollback_altq(u_int32_t ticket)
955 {
956 	struct pf_altq	*altq, *tmp;
957 	int		 error = 0;
958 
959 	PF_RULES_WASSERT();
960 
961 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
962 		return (0);
963 	/* Purge the old altq lists */
964 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
965 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
966 			/* detach and destroy the discipline */
967 			error = altq_remove(altq);
968 		}
969 		free(altq, M_PFALTQ);
970 	}
971 	TAILQ_INIT(V_pf_altq_ifs_inactive);
972 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
973 		pf_qid_unref(altq->qid);
974 		free(altq, M_PFALTQ);
975 	}
976 	TAILQ_INIT(V_pf_altqs_inactive);
977 	V_altqs_inactive_open = 0;
978 	return (error);
979 }
980 
981 static int
982 pf_commit_altq(u_int32_t ticket)
983 {
984 	struct pf_altqqueue	*old_altqs, *old_altq_ifs;
985 	struct pf_altq		*altq, *tmp;
986 	int			 err, error = 0;
987 
988 	PF_RULES_WASSERT();
989 
990 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
991 		return (EBUSY);
992 
993 	/* swap altqs, keep the old. */
994 	old_altqs = V_pf_altqs_active;
995 	old_altq_ifs = V_pf_altq_ifs_active;
996 	V_pf_altqs_active = V_pf_altqs_inactive;
997 	V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
998 	V_pf_altqs_inactive = old_altqs;
999 	V_pf_altq_ifs_inactive = old_altq_ifs;
1000 	V_ticket_altqs_active = V_ticket_altqs_inactive;
1001 
1002 	/* Attach new disciplines */
1003 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1004 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
1005 			/* attach the discipline */
1006 			error = altq_pfattach(altq);
1007 			if (error == 0 && V_pf_altq_running)
1008 				error = pf_enable_altq(altq);
1009 			if (error != 0)
1010 				return (error);
1011 		}
1012 	}
1013 
1014 	/* Purge the old altq lists */
1015 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
1016 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
1017 			/* detach and destroy the discipline */
1018 			if (V_pf_altq_running)
1019 				error = pf_disable_altq(altq);
1020 			err = altq_pfdetach(altq);
1021 			if (err != 0 && error == 0)
1022 				error = err;
1023 			err = altq_remove(altq);
1024 			if (err != 0 && error == 0)
1025 				error = err;
1026 		}
1027 		free(altq, M_PFALTQ);
1028 	}
1029 	TAILQ_INIT(V_pf_altq_ifs_inactive);
1030 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
1031 		pf_qid_unref(altq->qid);
1032 		free(altq, M_PFALTQ);
1033 	}
1034 	TAILQ_INIT(V_pf_altqs_inactive);
1035 
1036 	V_altqs_inactive_open = 0;
1037 	return (error);
1038 }
1039 
1040 static int
1041 pf_enable_altq(struct pf_altq *altq)
1042 {
1043 	struct ifnet		*ifp;
1044 	struct tb_profile	 tb;
1045 	int			 error = 0;
1046 
1047 	if ((ifp = ifunit(altq->ifname)) == NULL)
1048 		return (EINVAL);
1049 
1050 	if (ifp->if_snd.altq_type != ALTQT_NONE)
1051 		error = altq_enable(&ifp->if_snd);
1052 
1053 	/* set tokenbucket regulator */
1054 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
1055 		tb.rate = altq->ifbandwidth;
1056 		tb.depth = altq->tbrsize;
1057 		error = tbr_set(&ifp->if_snd, &tb);
1058 	}
1059 
1060 	return (error);
1061 }
1062 
1063 static int
1064 pf_disable_altq(struct pf_altq *altq)
1065 {
1066 	struct ifnet		*ifp;
1067 	struct tb_profile	 tb;
1068 	int			 error;
1069 
1070 	if ((ifp = ifunit(altq->ifname)) == NULL)
1071 		return (EINVAL);
1072 
1073 	/*
1074 	 * when the discipline is no longer referenced, it was overridden
1075 	 * by a new one.  if so, just return.
1076 	 */
1077 	if (altq->altq_disc != ifp->if_snd.altq_disc)
1078 		return (0);
1079 
1080 	error = altq_disable(&ifp->if_snd);
1081 
1082 	if (error == 0) {
1083 		/* clear tokenbucket regulator */
1084 		tb.rate = 0;
1085 		error = tbr_set(&ifp->if_snd, &tb);
1086 	}
1087 
1088 	return (error);
1089 }
1090 
1091 static int
1092 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
1093     struct pf_altq *altq)
1094 {
1095 	struct ifnet	*ifp1;
1096 	int		 error = 0;
1097 
1098 	/* Deactivate the interface in question */
1099 	altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
1100 	if ((ifp1 = ifunit(altq->ifname)) == NULL ||
1101 	    (remove && ifp1 == ifp)) {
1102 		altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
1103 	} else {
1104 		error = altq_add(ifp1, altq);
1105 
1106 		if (ticket != V_ticket_altqs_inactive)
1107 			error = EBUSY;
1108 
1109 		if (error)
1110 			free(altq, M_PFALTQ);
1111 	}
1112 
1113 	return (error);
1114 }
1115 
1116 void
1117 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
1118 {
1119 	struct pf_altq	*a1, *a2, *a3;
1120 	u_int32_t	 ticket;
1121 	int		 error = 0;
1122 
1123 	/*
1124 	 * No need to re-evaluate the configuration for events on interfaces
1125 	 * that do not support ALTQ, as it's not possible for such
1126 	 * interfaces to be part of the configuration.
1127 	 */
1128 	if (!ALTQ_IS_READY(&ifp->if_snd))
1129 		return;
1130 
1131 	/* Interrupt userland queue modifications */
1132 	if (V_altqs_inactive_open)
1133 		pf_rollback_altq(V_ticket_altqs_inactive);
1134 
1135 	/* Start new altq ruleset */
1136 	if (pf_begin_altq(&ticket))
1137 		return;
1138 
1139 	/* Copy the current active set */
1140 	TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
1141 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1142 		if (a2 == NULL) {
1143 			error = ENOMEM;
1144 			break;
1145 		}
1146 		bcopy(a1, a2, sizeof(struct pf_altq));
1147 
1148 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1149 		if (error)
1150 			break;
1151 
1152 		TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
1153 	}
1154 	if (error)
1155 		goto out;
1156 	TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
1157 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1158 		if (a2 == NULL) {
1159 			error = ENOMEM;
1160 			break;
1161 		}
1162 		bcopy(a1, a2, sizeof(struct pf_altq));
1163 
1164 		if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
1165 			error = EBUSY;
1166 			free(a2, M_PFALTQ);
1167 			break;
1168 		}
1169 		a2->altq_disc = NULL;
1170 		TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
1171 			if (strncmp(a3->ifname, a2->ifname,
1172 				IFNAMSIZ) == 0) {
1173 				a2->altq_disc = a3->altq_disc;
1174 				break;
1175 			}
1176 		}
1177 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1178 		if (error)
1179 			break;
1180 
1181 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
1182 	}
1183 
1184 out:
1185 	if (error != 0)
1186 		pf_rollback_altq(ticket);
1187 	else
1188 		pf_commit_altq(ticket);
1189 }
1190 #endif /* ALTQ */
1191 
1192 static struct pf_krule_global *
1193 pf_rule_tree_alloc(int flags)
1194 {
1195 	struct pf_krule_global *tree;
1196 
1197 	tree = malloc(sizeof(struct pf_krule_global), M_TEMP, flags);
1198 	if (tree == NULL)
1199 		return (NULL);
1200 	RB_INIT(tree);
1201 	return (tree);
1202 }
1203 
1204 static void
1205 pf_rule_tree_free(struct pf_krule_global *tree)
1206 {
1207 
1208 	free(tree, M_TEMP);
1209 }
1210 
1211 static int
1212 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
1213 {
1214 	struct pf_krule_global *tree;
1215 	struct pf_kruleset	*rs;
1216 	struct pf_krule		*rule;
1217 
1218 	PF_RULES_WASSERT();
1219 
1220 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1221 		return (EINVAL);
1222 	tree = pf_rule_tree_alloc(M_NOWAIT);
1223 	if (tree == NULL)
1224 		return (ENOMEM);
1225 	rs = pf_find_or_create_kruleset(anchor);
1226 	if (rs == NULL) {
1227 		free(tree, M_TEMP);
1228 		return (EINVAL);
1229 	}
1230 	pf_rule_tree_free(rs->rules[rs_num].inactive.tree);
1231 	rs->rules[rs_num].inactive.tree = tree;
1232 
1233 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1234 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1235 		rs->rules[rs_num].inactive.rcount--;
1236 	}
1237 	*ticket = ++rs->rules[rs_num].inactive.ticket;
1238 	rs->rules[rs_num].inactive.open = 1;
1239 	return (0);
1240 }
1241 
1242 static int
1243 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
1244 {
1245 	struct pf_kruleset	*rs;
1246 	struct pf_krule		*rule;
1247 
1248 	PF_RULES_WASSERT();
1249 
1250 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1251 		return (EINVAL);
1252 	rs = pf_find_kruleset(anchor);
1253 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1254 	    rs->rules[rs_num].inactive.ticket != ticket)
1255 		return (0);
1256 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1257 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1258 		rs->rules[rs_num].inactive.rcount--;
1259 	}
1260 	rs->rules[rs_num].inactive.open = 0;
1261 	return (0);
1262 }
1263 
1264 #define PF_MD5_UPD(st, elm)						\
1265 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
1266 
1267 #define PF_MD5_UPD_STR(st, elm)						\
1268 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
1269 
1270 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
1271 		(stor) = htonl((st)->elm);				\
1272 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
1273 } while (0)
1274 
1275 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
1276 		(stor) = htons((st)->elm);				\
1277 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
1278 } while (0)
1279 
1280 static void
1281 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
1282 {
1283 	PF_MD5_UPD(pfr, addr.type);
1284 	switch (pfr->addr.type) {
1285 		case PF_ADDR_DYNIFTL:
1286 			PF_MD5_UPD(pfr, addr.v.ifname);
1287 			PF_MD5_UPD(pfr, addr.iflags);
1288 			break;
1289 		case PF_ADDR_TABLE:
1290 			PF_MD5_UPD(pfr, addr.v.tblname);
1291 			break;
1292 		case PF_ADDR_ADDRMASK:
1293 			/* XXX ignore af? */
1294 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
1295 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
1296 			break;
1297 	}
1298 
1299 	PF_MD5_UPD(pfr, port[0]);
1300 	PF_MD5_UPD(pfr, port[1]);
1301 	PF_MD5_UPD(pfr, neg);
1302 	PF_MD5_UPD(pfr, port_op);
1303 }
1304 
1305 static void
1306 pf_hash_rule_rolling(MD5_CTX *ctx, struct pf_krule *rule)
1307 {
1308 	u_int16_t x;
1309 	u_int32_t y;
1310 
1311 	pf_hash_rule_addr(ctx, &rule->src);
1312 	pf_hash_rule_addr(ctx, &rule->dst);
1313 	for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++)
1314 		PF_MD5_UPD_STR(rule, label[i]);
1315 	PF_MD5_UPD_STR(rule, ifname);
1316 	PF_MD5_UPD_STR(rule, rcv_ifname);
1317 	PF_MD5_UPD_STR(rule, match_tagname);
1318 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
1319 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
1320 	PF_MD5_UPD_HTONL(rule, prob, y);
1321 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
1322 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1323 	PF_MD5_UPD(rule, uid.op);
1324 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1325 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1326 	PF_MD5_UPD(rule, gid.op);
1327 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
1328 	PF_MD5_UPD(rule, action);
1329 	PF_MD5_UPD(rule, direction);
1330 	PF_MD5_UPD(rule, af);
1331 	PF_MD5_UPD(rule, quick);
1332 	PF_MD5_UPD(rule, ifnot);
1333 	PF_MD5_UPD(rule, match_tag_not);
1334 	PF_MD5_UPD(rule, natpass);
1335 	PF_MD5_UPD(rule, keep_state);
1336 	PF_MD5_UPD(rule, proto);
1337 	PF_MD5_UPD(rule, type);
1338 	PF_MD5_UPD(rule, code);
1339 	PF_MD5_UPD(rule, flags);
1340 	PF_MD5_UPD(rule, flagset);
1341 	PF_MD5_UPD(rule, allow_opts);
1342 	PF_MD5_UPD(rule, rt);
1343 	PF_MD5_UPD(rule, tos);
1344 	PF_MD5_UPD(rule, scrub_flags);
1345 	PF_MD5_UPD(rule, min_ttl);
1346 	PF_MD5_UPD(rule, set_tos);
1347 	if (rule->anchor != NULL)
1348 		PF_MD5_UPD_STR(rule, anchor->path);
1349 }
1350 
1351 static void
1352 pf_hash_rule(struct pf_krule *rule)
1353 {
1354 	MD5_CTX		ctx;
1355 
1356 	MD5Init(&ctx);
1357 	pf_hash_rule_rolling(&ctx, rule);
1358 	MD5Final(rule->md5sum, &ctx);
1359 }
1360 
1361 static int
1362 pf_krule_compare(struct pf_krule *a, struct pf_krule *b)
1363 {
1364 
1365 	return (memcmp(a->md5sum, b->md5sum, PF_MD5_DIGEST_LENGTH));
1366 }
1367 
1368 static int
1369 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1370 {
1371 	struct pf_kruleset	*rs;
1372 	struct pf_krule		*rule, **old_array, *old_rule;
1373 	struct pf_krulequeue	*old_rules;
1374 	struct pf_krule_global  *old_tree;
1375 	int			 error;
1376 	u_int32_t		 old_rcount;
1377 
1378 	PF_RULES_WASSERT();
1379 
1380 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1381 		return (EINVAL);
1382 	rs = pf_find_kruleset(anchor);
1383 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1384 	    ticket != rs->rules[rs_num].inactive.ticket)
1385 		return (EBUSY);
1386 
1387 	/* Calculate checksum for the main ruleset */
1388 	if (rs == &pf_main_ruleset) {
1389 		error = pf_setup_pfsync_matching(rs);
1390 		if (error != 0)
1391 			return (error);
1392 	}
1393 
1394 	/* Swap rules, keep the old. */
1395 	old_rules = rs->rules[rs_num].active.ptr;
1396 	old_rcount = rs->rules[rs_num].active.rcount;
1397 	old_array = rs->rules[rs_num].active.ptr_array;
1398 	old_tree = rs->rules[rs_num].active.tree;
1399 
1400 	rs->rules[rs_num].active.ptr =
1401 	    rs->rules[rs_num].inactive.ptr;
1402 	rs->rules[rs_num].active.ptr_array =
1403 	    rs->rules[rs_num].inactive.ptr_array;
1404 	rs->rules[rs_num].active.tree =
1405 	    rs->rules[rs_num].inactive.tree;
1406 	rs->rules[rs_num].active.rcount =
1407 	    rs->rules[rs_num].inactive.rcount;
1408 
1409 	/* Attempt to preserve counter information. */
1410 	if (V_pf_status.keep_counters && old_tree != NULL) {
1411 		TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr,
1412 		    entries) {
1413 			old_rule = RB_FIND(pf_krule_global, old_tree, rule);
1414 			if (old_rule == NULL) {
1415 				continue;
1416 			}
1417 			pf_counter_u64_critical_enter();
1418 			pf_counter_u64_rollup_protected(&rule->evaluations,
1419 			    pf_counter_u64_fetch(&old_rule->evaluations));
1420 			pf_counter_u64_rollup_protected(&rule->packets[0],
1421 			    pf_counter_u64_fetch(&old_rule->packets[0]));
1422 			pf_counter_u64_rollup_protected(&rule->packets[1],
1423 			    pf_counter_u64_fetch(&old_rule->packets[1]));
1424 			pf_counter_u64_rollup_protected(&rule->bytes[0],
1425 			    pf_counter_u64_fetch(&old_rule->bytes[0]));
1426 			pf_counter_u64_rollup_protected(&rule->bytes[1],
1427 			    pf_counter_u64_fetch(&old_rule->bytes[1]));
1428 			pf_counter_u64_critical_exit();
1429 		}
1430 	}
1431 
1432 	rs->rules[rs_num].inactive.ptr = old_rules;
1433 	rs->rules[rs_num].inactive.ptr_array = old_array;
1434 	rs->rules[rs_num].inactive.tree = NULL; /* important for pf_ioctl_addrule */
1435 	rs->rules[rs_num].inactive.rcount = old_rcount;
1436 
1437 	rs->rules[rs_num].active.ticket =
1438 	    rs->rules[rs_num].inactive.ticket;
1439 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1440 
1441 	/* Purge the old rule list. */
1442 	PF_UNLNKDRULES_LOCK();
1443 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1444 		pf_unlink_rule_locked(old_rules, rule);
1445 	PF_UNLNKDRULES_UNLOCK();
1446 	if (rs->rules[rs_num].inactive.ptr_array)
1447 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1448 	rs->rules[rs_num].inactive.ptr_array = NULL;
1449 	rs->rules[rs_num].inactive.rcount = 0;
1450 	rs->rules[rs_num].inactive.open = 0;
1451 	pf_remove_if_empty_kruleset(rs);
1452 	free(old_tree, M_TEMP);
1453 
1454 	return (0);
1455 }
1456 
1457 static int
1458 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1459 {
1460 	MD5_CTX			 ctx;
1461 	struct pf_krule		*rule;
1462 	int			 rs_cnt;
1463 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
1464 
1465 	MD5Init(&ctx);
1466 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1467 		/* XXX PF_RULESET_SCRUB as well? */
1468 		if (rs_cnt == PF_RULESET_SCRUB)
1469 			continue;
1470 
1471 		if (rs->rules[rs_cnt].inactive.ptr_array)
1472 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1473 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
1474 
1475 		if (rs->rules[rs_cnt].inactive.rcount) {
1476 			rs->rules[rs_cnt].inactive.ptr_array =
1477 			    mallocarray(rs->rules[rs_cnt].inactive.rcount,
1478 			    sizeof(struct pf_rule **),
1479 			    M_TEMP, M_NOWAIT);
1480 
1481 			if (!rs->rules[rs_cnt].inactive.ptr_array)
1482 				return (ENOMEM);
1483 		}
1484 
1485 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1486 		    entries) {
1487 			pf_hash_rule_rolling(&ctx, rule);
1488 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1489 		}
1490 	}
1491 
1492 	MD5Final(digest, &ctx);
1493 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1494 	return (0);
1495 }
1496 
1497 static int
1498 pf_eth_addr_setup(struct pf_keth_ruleset *ruleset, struct pf_addr_wrap *addr)
1499 {
1500 	int error = 0;
1501 
1502 	switch (addr->type) {
1503 	case PF_ADDR_TABLE:
1504 		addr->p.tbl = pfr_eth_attach_table(ruleset, addr->v.tblname);
1505 		if (addr->p.tbl == NULL)
1506 			error = ENOMEM;
1507 		break;
1508 	default:
1509 		error = EINVAL;
1510 	}
1511 
1512 	return (error);
1513 }
1514 
1515 static int
1516 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1517     sa_family_t af)
1518 {
1519 	int error = 0;
1520 
1521 	switch (addr->type) {
1522 	case PF_ADDR_TABLE:
1523 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1524 		if (addr->p.tbl == NULL)
1525 			error = ENOMEM;
1526 		break;
1527 	case PF_ADDR_DYNIFTL:
1528 		error = pfi_dynaddr_setup(addr, af);
1529 		break;
1530 	}
1531 
1532 	return (error);
1533 }
1534 
1535 void
1536 pf_addr_copyout(struct pf_addr_wrap *addr)
1537 {
1538 
1539 	switch (addr->type) {
1540 	case PF_ADDR_DYNIFTL:
1541 		pfi_dynaddr_copyout(addr);
1542 		break;
1543 	case PF_ADDR_TABLE:
1544 		pf_tbladdr_copyout(addr);
1545 		break;
1546 	}
1547 }
1548 
1549 static void
1550 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
1551 {
1552 	int	secs = time_uptime, diff;
1553 
1554 	bzero(out, sizeof(struct pf_src_node));
1555 
1556 	bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
1557 	bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
1558 
1559 	if (in->rule != NULL)
1560 		out->rule.nr = in->rule->nr;
1561 
1562 	for (int i = 0; i < 2; i++) {
1563 		out->bytes[i] = counter_u64_fetch(in->bytes[i]);
1564 		out->packets[i] = counter_u64_fetch(in->packets[i]);
1565 	}
1566 
1567 	out->states = in->states;
1568 	out->conn = in->conn;
1569 	out->af = in->af;
1570 	out->ruletype = in->ruletype;
1571 
1572 	out->creation = secs - in->creation;
1573 	if (out->expire > secs)
1574 		out->expire -= secs;
1575 	else
1576 		out->expire = 0;
1577 
1578 	/* Adjust the connection rate estimate. */
1579 	out->conn_rate = in->conn_rate;
1580 	diff = secs - in->conn_rate.last;
1581 	if (diff >= in->conn_rate.seconds)
1582 		out->conn_rate.count = 0;
1583 	else
1584 		out->conn_rate.count -=
1585 		    in->conn_rate.count * diff /
1586 		    in->conn_rate.seconds;
1587 }
1588 
1589 #ifdef ALTQ
1590 /*
1591  * Handle export of struct pf_kaltq to user binaries that may be using any
1592  * version of struct pf_altq.
1593  */
1594 static int
1595 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1596 {
1597 	u_int32_t version;
1598 
1599 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1600 		version = 0;
1601 	else
1602 		version = pa->version;
1603 
1604 	if (version > PFIOC_ALTQ_VERSION)
1605 		return (EINVAL);
1606 
1607 #define ASSIGN(x) exported_q->x = q->x
1608 #define COPY(x) \
1609 	bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1610 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1611 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1612 
1613 	switch (version) {
1614 	case 0: {
1615 		struct pf_altq_v0 *exported_q =
1616 		    &((struct pfioc_altq_v0 *)pa)->altq;
1617 
1618 		COPY(ifname);
1619 
1620 		ASSIGN(scheduler);
1621 		ASSIGN(tbrsize);
1622 		exported_q->tbrsize = SATU16(q->tbrsize);
1623 		exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1624 
1625 		COPY(qname);
1626 		COPY(parent);
1627 		ASSIGN(parent_qid);
1628 		exported_q->bandwidth = SATU32(q->bandwidth);
1629 		ASSIGN(priority);
1630 		ASSIGN(local_flags);
1631 
1632 		ASSIGN(qlimit);
1633 		ASSIGN(flags);
1634 
1635 		if (q->scheduler == ALTQT_HFSC) {
1636 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1637 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1638 			    SATU32(q->pq_u.hfsc_opts.x)
1639 
1640 			ASSIGN_OPT_SATU32(rtsc_m1);
1641 			ASSIGN_OPT(rtsc_d);
1642 			ASSIGN_OPT_SATU32(rtsc_m2);
1643 
1644 			ASSIGN_OPT_SATU32(lssc_m1);
1645 			ASSIGN_OPT(lssc_d);
1646 			ASSIGN_OPT_SATU32(lssc_m2);
1647 
1648 			ASSIGN_OPT_SATU32(ulsc_m1);
1649 			ASSIGN_OPT(ulsc_d);
1650 			ASSIGN_OPT_SATU32(ulsc_m2);
1651 
1652 			ASSIGN_OPT(flags);
1653 
1654 #undef ASSIGN_OPT
1655 #undef ASSIGN_OPT_SATU32
1656 		} else
1657 			COPY(pq_u);
1658 
1659 		ASSIGN(qid);
1660 		break;
1661 	}
1662 	case 1:	{
1663 		struct pf_altq_v1 *exported_q =
1664 		    &((struct pfioc_altq_v1 *)pa)->altq;
1665 
1666 		COPY(ifname);
1667 
1668 		ASSIGN(scheduler);
1669 		ASSIGN(tbrsize);
1670 		ASSIGN(ifbandwidth);
1671 
1672 		COPY(qname);
1673 		COPY(parent);
1674 		ASSIGN(parent_qid);
1675 		ASSIGN(bandwidth);
1676 		ASSIGN(priority);
1677 		ASSIGN(local_flags);
1678 
1679 		ASSIGN(qlimit);
1680 		ASSIGN(flags);
1681 		COPY(pq_u);
1682 
1683 		ASSIGN(qid);
1684 		break;
1685 	}
1686 	default:
1687 		panic("%s: unhandled struct pfioc_altq version", __func__);
1688 		break;
1689 	}
1690 
1691 #undef ASSIGN
1692 #undef COPY
1693 #undef SATU16
1694 #undef SATU32
1695 
1696 	return (0);
1697 }
1698 
1699 /*
1700  * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1701  * that may be using any version of it.
1702  */
1703 static int
1704 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1705 {
1706 	u_int32_t version;
1707 
1708 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1709 		version = 0;
1710 	else
1711 		version = pa->version;
1712 
1713 	if (version > PFIOC_ALTQ_VERSION)
1714 		return (EINVAL);
1715 
1716 #define ASSIGN(x) q->x = imported_q->x
1717 #define COPY(x) \
1718 	bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1719 
1720 	switch (version) {
1721 	case 0: {
1722 		struct pf_altq_v0 *imported_q =
1723 		    &((struct pfioc_altq_v0 *)pa)->altq;
1724 
1725 		COPY(ifname);
1726 
1727 		ASSIGN(scheduler);
1728 		ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1729 		ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1730 
1731 		COPY(qname);
1732 		COPY(parent);
1733 		ASSIGN(parent_qid);
1734 		ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1735 		ASSIGN(priority);
1736 		ASSIGN(local_flags);
1737 
1738 		ASSIGN(qlimit);
1739 		ASSIGN(flags);
1740 
1741 		if (imported_q->scheduler == ALTQT_HFSC) {
1742 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1743 
1744 			/*
1745 			 * The m1 and m2 parameters are being copied from
1746 			 * 32-bit to 64-bit.
1747 			 */
1748 			ASSIGN_OPT(rtsc_m1);
1749 			ASSIGN_OPT(rtsc_d);
1750 			ASSIGN_OPT(rtsc_m2);
1751 
1752 			ASSIGN_OPT(lssc_m1);
1753 			ASSIGN_OPT(lssc_d);
1754 			ASSIGN_OPT(lssc_m2);
1755 
1756 			ASSIGN_OPT(ulsc_m1);
1757 			ASSIGN_OPT(ulsc_d);
1758 			ASSIGN_OPT(ulsc_m2);
1759 
1760 			ASSIGN_OPT(flags);
1761 
1762 #undef ASSIGN_OPT
1763 		} else
1764 			COPY(pq_u);
1765 
1766 		ASSIGN(qid);
1767 		break;
1768 	}
1769 	case 1: {
1770 		struct pf_altq_v1 *imported_q =
1771 		    &((struct pfioc_altq_v1 *)pa)->altq;
1772 
1773 		COPY(ifname);
1774 
1775 		ASSIGN(scheduler);
1776 		ASSIGN(tbrsize);
1777 		ASSIGN(ifbandwidth);
1778 
1779 		COPY(qname);
1780 		COPY(parent);
1781 		ASSIGN(parent_qid);
1782 		ASSIGN(bandwidth);
1783 		ASSIGN(priority);
1784 		ASSIGN(local_flags);
1785 
1786 		ASSIGN(qlimit);
1787 		ASSIGN(flags);
1788 		COPY(pq_u);
1789 
1790 		ASSIGN(qid);
1791 		break;
1792 	}
1793 	default:
1794 		panic("%s: unhandled struct pfioc_altq version", __func__);
1795 		break;
1796 	}
1797 
1798 #undef ASSIGN
1799 #undef COPY
1800 
1801 	return (0);
1802 }
1803 
1804 static struct pf_altq *
1805 pf_altq_get_nth_active(u_int32_t n)
1806 {
1807 	struct pf_altq		*altq;
1808 	u_int32_t		 nr;
1809 
1810 	nr = 0;
1811 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1812 		if (nr == n)
1813 			return (altq);
1814 		nr++;
1815 	}
1816 
1817 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1818 		if (nr == n)
1819 			return (altq);
1820 		nr++;
1821 	}
1822 
1823 	return (NULL);
1824 }
1825 #endif /* ALTQ */
1826 
1827 struct pf_krule *
1828 pf_krule_alloc(void)
1829 {
1830 	struct pf_krule *rule;
1831 
1832 	rule = malloc(sizeof(struct pf_krule), M_PFRULE, M_WAITOK | M_ZERO);
1833 	mtx_init(&rule->nat.mtx, "pf_krule_nat_pool", NULL, MTX_DEF);
1834 	mtx_init(&rule->rdr.mtx, "pf_krule_rdr_pool", NULL, MTX_DEF);
1835 	rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
1836 	    M_WAITOK | M_ZERO);
1837 	return (rule);
1838 }
1839 
1840 void
1841 pf_krule_free(struct pf_krule *rule)
1842 {
1843 #ifdef PF_WANT_32_TO_64_COUNTER
1844 	bool wowned;
1845 #endif
1846 
1847 	if (rule == NULL)
1848 		return;
1849 
1850 #ifdef PF_WANT_32_TO_64_COUNTER
1851 	if (rule->allrulelinked) {
1852 		wowned = PF_RULES_WOWNED();
1853 		if (!wowned)
1854 			PF_RULES_WLOCK();
1855 		LIST_REMOVE(rule, allrulelist);
1856 		V_pf_allrulecount--;
1857 		if (!wowned)
1858 			PF_RULES_WUNLOCK();
1859 	}
1860 #endif
1861 
1862 	pf_counter_u64_deinit(&rule->evaluations);
1863 	for (int i = 0; i < 2; i++) {
1864 		pf_counter_u64_deinit(&rule->packets[i]);
1865 		pf_counter_u64_deinit(&rule->bytes[i]);
1866 	}
1867 	counter_u64_free(rule->states_cur);
1868 	counter_u64_free(rule->states_tot);
1869 	counter_u64_free(rule->src_nodes);
1870 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
1871 
1872 	mtx_destroy(&rule->nat.mtx);
1873 	mtx_destroy(&rule->rdr.mtx);
1874 	free(rule, M_PFRULE);
1875 }
1876 
1877 void
1878 pf_krule_clear_counters(struct pf_krule *rule)
1879 {
1880 	pf_counter_u64_zero(&rule->evaluations);
1881 	for (int i = 0; i < 2; i++) {
1882 		pf_counter_u64_zero(&rule->packets[i]);
1883 		pf_counter_u64_zero(&rule->bytes[i]);
1884 	}
1885 	counter_u64_zero(rule->states_tot);
1886 }
1887 
1888 static void
1889 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool,
1890     struct pf_pooladdr *pool)
1891 {
1892 
1893 	bzero(pool, sizeof(*pool));
1894 	bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr));
1895 	strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname));
1896 }
1897 
1898 static int
1899 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool,
1900     struct pf_kpooladdr *kpool)
1901 {
1902 	int ret;
1903 
1904 	bzero(kpool, sizeof(*kpool));
1905 	bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr));
1906 	ret = pf_user_strcpy(kpool->ifname, pool->ifname,
1907 	    sizeof(kpool->ifname));
1908 	return (ret);
1909 }
1910 
1911 static void
1912 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool)
1913 {
1914 	_Static_assert(sizeof(pool->key) == sizeof(kpool->key), "");
1915 	_Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), "");
1916 
1917 	bcopy(&pool->key, &kpool->key, sizeof(kpool->key));
1918 	bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter));
1919 
1920 	kpool->tblidx = pool->tblidx;
1921 	kpool->proxy_port[0] = pool->proxy_port[0];
1922 	kpool->proxy_port[1] = pool->proxy_port[1];
1923 	kpool->opts = pool->opts;
1924 }
1925 
1926 static int
1927 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
1928 {
1929 	int ret;
1930 
1931 #ifndef INET
1932 	if (rule->af == AF_INET) {
1933 		return (EAFNOSUPPORT);
1934 	}
1935 #endif /* INET */
1936 #ifndef INET6
1937 	if (rule->af == AF_INET6) {
1938 		return (EAFNOSUPPORT);
1939 	}
1940 #endif /* INET6 */
1941 
1942 	ret = pf_check_rule_addr(&rule->src);
1943 	if (ret != 0)
1944 		return (ret);
1945 	ret = pf_check_rule_addr(&rule->dst);
1946 	if (ret != 0)
1947 		return (ret);
1948 
1949 	bcopy(&rule->src, &krule->src, sizeof(rule->src));
1950 	bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
1951 
1952 	ret = pf_user_strcpy(krule->label[0], rule->label, sizeof(rule->label));
1953 	if (ret != 0)
1954 		return (ret);
1955 	ret = pf_user_strcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
1956 	if (ret != 0)
1957 		return (ret);
1958 	ret = pf_user_strcpy(krule->qname, rule->qname, sizeof(rule->qname));
1959 	if (ret != 0)
1960 		return (ret);
1961 	ret = pf_user_strcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
1962 	if (ret != 0)
1963 		return (ret);
1964 	ret = pf_user_strcpy(krule->tagname, rule->tagname,
1965 	    sizeof(rule->tagname));
1966 	if (ret != 0)
1967 		return (ret);
1968 	ret = pf_user_strcpy(krule->match_tagname, rule->match_tagname,
1969 	    sizeof(rule->match_tagname));
1970 	if (ret != 0)
1971 		return (ret);
1972 	ret = pf_user_strcpy(krule->overload_tblname, rule->overload_tblname,
1973 	    sizeof(rule->overload_tblname));
1974 	if (ret != 0)
1975 		return (ret);
1976 
1977 	pf_pool_to_kpool(&rule->rpool, &krule->rdr);
1978 
1979 	/* Don't allow userspace to set evaluations, packets or bytes. */
1980 	/* kif, anchor, overload_tbl are not copied over. */
1981 
1982 	krule->os_fingerprint = rule->os_fingerprint;
1983 
1984 	krule->rtableid = rule->rtableid;
1985 	/* pf_rule->timeout is smaller than pf_krule->timeout */
1986 	bcopy(rule->timeout, krule->timeout, sizeof(rule->timeout));
1987 	krule->max_states = rule->max_states;
1988 	krule->max_src_nodes = rule->max_src_nodes;
1989 	krule->max_src_states = rule->max_src_states;
1990 	krule->max_src_conn = rule->max_src_conn;
1991 	krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
1992 	krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
1993 	krule->qid = rule->qid;
1994 	krule->pqid = rule->pqid;
1995 	krule->nr = rule->nr;
1996 	krule->prob = rule->prob;
1997 	krule->cuid = rule->cuid;
1998 	krule->cpid = rule->cpid;
1999 
2000 	krule->return_icmp = rule->return_icmp;
2001 	krule->return_icmp6 = rule->return_icmp6;
2002 	krule->max_mss = rule->max_mss;
2003 	krule->tag = rule->tag;
2004 	krule->match_tag = rule->match_tag;
2005 	krule->scrub_flags = rule->scrub_flags;
2006 
2007 	bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
2008 	bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
2009 
2010 	krule->rule_flag = rule->rule_flag;
2011 	krule->action = rule->action;
2012 	krule->direction = rule->direction;
2013 	krule->log = rule->log;
2014 	krule->logif = rule->logif;
2015 	krule->quick = rule->quick;
2016 	krule->ifnot = rule->ifnot;
2017 	krule->match_tag_not = rule->match_tag_not;
2018 	krule->natpass = rule->natpass;
2019 
2020 	krule->keep_state = rule->keep_state;
2021 	krule->af = rule->af;
2022 	krule->proto = rule->proto;
2023 	krule->type = rule->type;
2024 	krule->code = rule->code;
2025 	krule->flags = rule->flags;
2026 	krule->flagset = rule->flagset;
2027 	krule->min_ttl = rule->min_ttl;
2028 	krule->allow_opts = rule->allow_opts;
2029 	krule->rt = rule->rt;
2030 	krule->return_ttl = rule->return_ttl;
2031 	krule->tos = rule->tos;
2032 	krule->set_tos = rule->set_tos;
2033 
2034 	krule->flush = rule->flush;
2035 	krule->prio = rule->prio;
2036 	krule->set_prio[0] = rule->set_prio[0];
2037 	krule->set_prio[1] = rule->set_prio[1];
2038 
2039 	bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
2040 
2041 	return (0);
2042 }
2043 
2044 int
2045 pf_ioctl_getrules(struct pfioc_rule *pr)
2046 {
2047 	struct pf_kruleset	*ruleset;
2048 	struct pf_krule		*tail;
2049 	int			 rs_num;
2050 
2051 	PF_RULES_WLOCK();
2052 	ruleset = pf_find_kruleset(pr->anchor);
2053 	if (ruleset == NULL) {
2054 		PF_RULES_WUNLOCK();
2055 		return (EINVAL);
2056 	}
2057 	rs_num = pf_get_ruleset_number(pr->rule.action);
2058 	if (rs_num >= PF_RULESET_MAX) {
2059 		PF_RULES_WUNLOCK();
2060 		return (EINVAL);
2061 	}
2062 	tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
2063 	    pf_krulequeue);
2064 	if (tail)
2065 		pr->nr = tail->nr + 1;
2066 	else
2067 		pr->nr = 0;
2068 	pr->ticket = ruleset->rules[rs_num].active.ticket;
2069 	PF_RULES_WUNLOCK();
2070 
2071 	return (0);
2072 }
2073 
2074 int
2075 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket,
2076     uint32_t pool_ticket, const char *anchor, const char *anchor_call,
2077     uid_t uid, pid_t pid)
2078 {
2079 	struct pf_kruleset	*ruleset;
2080 	struct pf_krule		*tail;
2081 	struct pf_kpooladdr	*pa;
2082 	struct pfi_kkif		*kif = NULL, *rcv_kif = NULL;
2083 	int			 rs_num;
2084 	int			 error = 0;
2085 
2086 	if ((rule->return_icmp >> 8) > ICMP_MAXTYPE) {
2087 		error = EINVAL;
2088 		goto errout_unlocked;
2089 	}
2090 
2091 #define	ERROUT(x)	ERROUT_FUNCTION(errout, x)
2092 
2093 	if (rule->ifname[0])
2094 		kif = pf_kkif_create(M_WAITOK);
2095 	if (rule->rcv_ifname[0])
2096 		rcv_kif = pf_kkif_create(M_WAITOK);
2097 	pf_counter_u64_init(&rule->evaluations, M_WAITOK);
2098 	for (int i = 0; i < 2; i++) {
2099 		pf_counter_u64_init(&rule->packets[i], M_WAITOK);
2100 		pf_counter_u64_init(&rule->bytes[i], M_WAITOK);
2101 	}
2102 	rule->states_cur = counter_u64_alloc(M_WAITOK);
2103 	rule->states_tot = counter_u64_alloc(M_WAITOK);
2104 	rule->src_nodes = counter_u64_alloc(M_WAITOK);
2105 	rule->cuid = uid;
2106 	rule->cpid = pid;
2107 	TAILQ_INIT(&rule->rdr.list);
2108 	TAILQ_INIT(&rule->nat.list);
2109 
2110 	PF_CONFIG_LOCK();
2111 	PF_RULES_WLOCK();
2112 #ifdef PF_WANT_32_TO_64_COUNTER
2113 	LIST_INSERT_HEAD(&V_pf_allrulelist, rule, allrulelist);
2114 	MPASS(!rule->allrulelinked);
2115 	rule->allrulelinked = true;
2116 	V_pf_allrulecount++;
2117 #endif
2118 	ruleset = pf_find_kruleset(anchor);
2119 	if (ruleset == NULL)
2120 		ERROUT(EINVAL);
2121 	rs_num = pf_get_ruleset_number(rule->action);
2122 	if (rs_num >= PF_RULESET_MAX)
2123 		ERROUT(EINVAL);
2124 	if (ticket != ruleset->rules[rs_num].inactive.ticket) {
2125 		DPFPRINTF(PF_DEBUG_MISC,
2126 		    ("ticket: %d != [%d]%d\n", ticket, rs_num,
2127 		    ruleset->rules[rs_num].inactive.ticket));
2128 		ERROUT(EBUSY);
2129 	}
2130 	if (pool_ticket != V_ticket_pabuf) {
2131 		DPFPRINTF(PF_DEBUG_MISC,
2132 		    ("pool_ticket: %d != %d\n", pool_ticket,
2133 		    V_ticket_pabuf));
2134 		ERROUT(EBUSY);
2135 	}
2136 	/*
2137 	 * XXXMJG hack: there is no mechanism to ensure they started the
2138 	 * transaction. Ticket checked above may happen to match by accident,
2139 	 * even if nobody called DIOCXBEGIN, let alone this process.
2140 	 * Partially work around it by checking if the RB tree got allocated,
2141 	 * see pf_begin_rules.
2142 	 */
2143 	if (ruleset->rules[rs_num].inactive.tree == NULL) {
2144 		ERROUT(EINVAL);
2145 	}
2146 
2147 	tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
2148 	    pf_krulequeue);
2149 	if (tail)
2150 		rule->nr = tail->nr + 1;
2151 	else
2152 		rule->nr = 0;
2153 	if (rule->ifname[0]) {
2154 		rule->kif = pfi_kkif_attach(kif, rule->ifname);
2155 		kif = NULL;
2156 		pfi_kkif_ref(rule->kif);
2157 	} else
2158 		rule->kif = NULL;
2159 
2160 	if (rule->rcv_ifname[0]) {
2161 		rule->rcv_kif = pfi_kkif_attach(rcv_kif, rule->rcv_ifname);
2162 		rcv_kif = NULL;
2163 		pfi_kkif_ref(rule->rcv_kif);
2164 	} else
2165 		rule->rcv_kif = NULL;
2166 
2167 	if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
2168 		error = EBUSY;
2169 
2170 #ifdef ALTQ
2171 	/* set queue IDs */
2172 	if (rule->qname[0] != 0) {
2173 		if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2174 			error = EBUSY;
2175 		else if (rule->pqname[0] != 0) {
2176 			if ((rule->pqid =
2177 			    pf_qname2qid(rule->pqname)) == 0)
2178 				error = EBUSY;
2179 		} else
2180 			rule->pqid = rule->qid;
2181 	}
2182 #endif
2183 	if (rule->tagname[0])
2184 		if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2185 			error = EBUSY;
2186 	if (rule->match_tagname[0])
2187 		if ((rule->match_tag =
2188 		    pf_tagname2tag(rule->match_tagname)) == 0)
2189 			error = EBUSY;
2190 	if (rule->rt && !rule->direction)
2191 		error = EINVAL;
2192 	if (!rule->log)
2193 		rule->logif = 0;
2194 	if (rule->logif >= PFLOGIFS_MAX)
2195 		error = EINVAL;
2196 	if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
2197 		error = ENOMEM;
2198 	if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
2199 		error = ENOMEM;
2200 	if (pf_kanchor_setup(rule, ruleset, anchor_call))
2201 		error = EINVAL;
2202 	if (rule->scrub_flags & PFSTATE_SETPRIO &&
2203 	    (rule->set_prio[0] > PF_PRIO_MAX ||
2204 	    rule->set_prio[1] > PF_PRIO_MAX))
2205 		error = EINVAL;
2206 	for (int i = 0; i < 2; i++) {
2207 		TAILQ_FOREACH(pa, &V_pf_pabuf[i], entries)
2208 			if (pa->addr.type == PF_ADDR_TABLE) {
2209 				pa->addr.p.tbl = pfr_attach_table(ruleset,
2210 				    pa->addr.v.tblname);
2211 				if (pa->addr.p.tbl == NULL)
2212 					error = ENOMEM;
2213 			}
2214 	}
2215 
2216 	rule->overload_tbl = NULL;
2217 	if (rule->overload_tblname[0]) {
2218 		if ((rule->overload_tbl = pfr_attach_table(ruleset,
2219 		    rule->overload_tblname)) == NULL)
2220 			error = EINVAL;
2221 		else
2222 			rule->overload_tbl->pfrkt_flags |=
2223 			    PFR_TFLAG_ACTIVE;
2224 	}
2225 
2226 	pf_mv_kpool(&V_pf_pabuf[0], &rule->nat.list);
2227 	pf_mv_kpool(&V_pf_pabuf[1], &rule->rdr.list);
2228 	if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
2229 	    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
2230 	    (rule->rt > PF_NOPFROUTE)) &&
2231 	    (TAILQ_FIRST(&rule->rdr.list) == NULL))
2232 		error = EINVAL;
2233 
2234 	if (rule->action == PF_PASS && rule->rdr.opts & PF_POOL_STICKYADDR &&
2235 	    !rule->keep_state) {
2236 		error = EINVAL;
2237 	}
2238 
2239 	if (error) {
2240 		pf_free_rule(rule);
2241 		rule = NULL;
2242 		ERROUT(error);
2243 	}
2244 
2245 	rule->nat.cur = TAILQ_FIRST(&rule->nat.list);
2246 	rule->rdr.cur = TAILQ_FIRST(&rule->rdr.list);
2247 	TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
2248 	    rule, entries);
2249 	ruleset->rules[rs_num].inactive.rcount++;
2250 
2251 	PF_RULES_WUNLOCK();
2252 	pf_hash_rule(rule);
2253 	if (RB_INSERT(pf_krule_global, ruleset->rules[rs_num].inactive.tree, rule) != NULL) {
2254 		PF_RULES_WLOCK();
2255 		TAILQ_REMOVE(ruleset->rules[rs_num].inactive.ptr, rule, entries);
2256 		ruleset->rules[rs_num].inactive.rcount--;
2257 		pf_free_rule(rule);
2258 		rule = NULL;
2259 		ERROUT(EEXIST);
2260 	}
2261 	PF_CONFIG_UNLOCK();
2262 
2263 	return (0);
2264 
2265 #undef ERROUT
2266 errout:
2267 	PF_RULES_WUNLOCK();
2268 	PF_CONFIG_UNLOCK();
2269 errout_unlocked:
2270 	pf_kkif_free(rcv_kif);
2271 	pf_kkif_free(kif);
2272 	pf_krule_free(rule);
2273 	return (error);
2274 }
2275 
2276 static bool
2277 pf_label_match(const struct pf_krule *rule, const char *label)
2278 {
2279 	int i = 0;
2280 
2281 	while (*rule->label[i]) {
2282 		if (strcmp(rule->label[i], label) == 0)
2283 			return (true);
2284 		i++;
2285 	}
2286 
2287 	return (false);
2288 }
2289 
2290 static unsigned int
2291 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir)
2292 {
2293 	struct pf_kstate *s;
2294 	int more = 0;
2295 
2296 	s = pf_find_state_all(key, dir, &more);
2297 	if (s == NULL)
2298 		return (0);
2299 
2300 	if (more) {
2301 		PF_STATE_UNLOCK(s);
2302 		return (0);
2303 	}
2304 
2305 	pf_unlink_state(s);
2306 	return (1);
2307 }
2308 
2309 static int
2310 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih)
2311 {
2312 	struct pf_kstate	*s;
2313 	struct pf_state_key	*sk;
2314 	struct pf_addr		*srcaddr, *dstaddr;
2315 	struct pf_state_key_cmp	 match_key;
2316 	int			 idx, killed = 0;
2317 	unsigned int		 dir;
2318 	u_int16_t		 srcport, dstport;
2319 	struct pfi_kkif		*kif;
2320 
2321 relock_DIOCKILLSTATES:
2322 	PF_HASHROW_LOCK(ih);
2323 	LIST_FOREACH(s, &ih->states, entry) {
2324 		/* For floating states look at the original kif. */
2325 		kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
2326 
2327 		sk = s->key[psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE];
2328 		if (s->direction == PF_OUT) {
2329 			srcaddr = &sk->addr[1];
2330 			dstaddr = &sk->addr[0];
2331 			srcport = sk->port[1];
2332 			dstport = sk->port[0];
2333 		} else {
2334 			srcaddr = &sk->addr[0];
2335 			dstaddr = &sk->addr[1];
2336 			srcport = sk->port[0];
2337 			dstport = sk->port[1];
2338 		}
2339 
2340 		if (psk->psk_af && sk->af != psk->psk_af)
2341 			continue;
2342 
2343 		if (psk->psk_proto && psk->psk_proto != sk->proto)
2344 			continue;
2345 
2346 		if (! PF_MATCHA(psk->psk_src.neg, &psk->psk_src.addr.v.a.addr,
2347 		    &psk->psk_src.addr.v.a.mask, srcaddr, sk->af))
2348 			continue;
2349 
2350 		if (! PF_MATCHA(psk->psk_dst.neg, &psk->psk_dst.addr.v.a.addr,
2351 		    &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af))
2352 			continue;
2353 
2354 		if (!  PF_MATCHA(psk->psk_rt_addr.neg,
2355 		    &psk->psk_rt_addr.addr.v.a.addr,
2356 		    &psk->psk_rt_addr.addr.v.a.mask,
2357 		    &s->act.rt_addr, sk->af))
2358 			continue;
2359 
2360 		if (psk->psk_src.port_op != 0 &&
2361 		    ! pf_match_port(psk->psk_src.port_op,
2362 		    psk->psk_src.port[0], psk->psk_src.port[1], srcport))
2363 			continue;
2364 
2365 		if (psk->psk_dst.port_op != 0 &&
2366 		    ! pf_match_port(psk->psk_dst.port_op,
2367 		    psk->psk_dst.port[0], psk->psk_dst.port[1], dstport))
2368 			continue;
2369 
2370 		if (psk->psk_label[0] &&
2371 		    ! pf_label_match(s->rule, psk->psk_label))
2372 			continue;
2373 
2374 		if (psk->psk_ifname[0] && strcmp(psk->psk_ifname,
2375 		    kif->pfik_name))
2376 			continue;
2377 
2378 		if (psk->psk_kill_match) {
2379 			/* Create the key to find matching states, with lock
2380 			 * held. */
2381 
2382 			bzero(&match_key, sizeof(match_key));
2383 
2384 			if (s->direction == PF_OUT) {
2385 				dir = PF_IN;
2386 				idx = psk->psk_nat ? PF_SK_WIRE : PF_SK_STACK;
2387 			} else {
2388 				dir = PF_OUT;
2389 				idx = psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE;
2390 			}
2391 
2392 			match_key.af = s->key[idx]->af;
2393 			match_key.proto = s->key[idx]->proto;
2394 			PF_ACPY(&match_key.addr[0],
2395 			    &s->key[idx]->addr[1], match_key.af);
2396 			match_key.port[0] = s->key[idx]->port[1];
2397 			PF_ACPY(&match_key.addr[1],
2398 			    &s->key[idx]->addr[0], match_key.af);
2399 			match_key.port[1] = s->key[idx]->port[0];
2400 		}
2401 
2402 		pf_unlink_state(s);
2403 		killed++;
2404 
2405 		if (psk->psk_kill_match)
2406 			killed += pf_kill_matching_state(&match_key, dir);
2407 
2408 		goto relock_DIOCKILLSTATES;
2409 	}
2410 	PF_HASHROW_UNLOCK(ih);
2411 
2412 	return (killed);
2413 }
2414 
2415 int
2416 pf_start(void)
2417 {
2418 	int error = 0;
2419 
2420 	sx_xlock(&V_pf_ioctl_lock);
2421 	if (V_pf_status.running)
2422 		error = EEXIST;
2423 	else {
2424 		hook_pf();
2425 		if (! TAILQ_EMPTY(V_pf_keth->active.rules))
2426 			hook_pf_eth();
2427 		V_pf_status.running = 1;
2428 		V_pf_status.since = time_second;
2429 		new_unrhdr64(&V_pf_stateid, time_second);
2430 
2431 		DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
2432 	}
2433 	sx_xunlock(&V_pf_ioctl_lock);
2434 
2435 	return (error);
2436 }
2437 
2438 int
2439 pf_stop(void)
2440 {
2441 	int error = 0;
2442 
2443 	sx_xlock(&V_pf_ioctl_lock);
2444 	if (!V_pf_status.running)
2445 		error = ENOENT;
2446 	else {
2447 		V_pf_status.running = 0;
2448 		dehook_pf();
2449 		dehook_pf_eth();
2450 		V_pf_status.since = time_second;
2451 		DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
2452 	}
2453 	sx_xunlock(&V_pf_ioctl_lock);
2454 
2455 	return (error);
2456 }
2457 
2458 void
2459 pf_ioctl_clear_status(void)
2460 {
2461 	PF_RULES_WLOCK();
2462 	for (int i = 0; i < PFRES_MAX; i++)
2463 		counter_u64_zero(V_pf_status.counters[i]);
2464 	for (int i = 0; i < FCNT_MAX; i++)
2465 		pf_counter_u64_zero(&V_pf_status.fcounters[i]);
2466 	for (int i = 0; i < SCNT_MAX; i++)
2467 		counter_u64_zero(V_pf_status.scounters[i]);
2468 	for (int i = 0; i < KLCNT_MAX; i++)
2469 		counter_u64_zero(V_pf_status.lcounters[i]);
2470 	V_pf_status.since = time_second;
2471 	if (*V_pf_status.ifname)
2472 		pfi_update_status(V_pf_status.ifname, NULL);
2473 	PF_RULES_WUNLOCK();
2474 }
2475 
2476 int
2477 pf_ioctl_set_timeout(int timeout, int seconds, int *prev_seconds)
2478 {
2479 	uint32_t old;
2480 
2481 	if (timeout < 0 || timeout >= PFTM_MAX ||
2482 	    seconds < 0)
2483 		return (EINVAL);
2484 
2485 	PF_RULES_WLOCK();
2486 	old = V_pf_default_rule.timeout[timeout];
2487 	if (timeout == PFTM_INTERVAL && seconds == 0)
2488 		seconds = 1;
2489 	V_pf_default_rule.timeout[timeout] = seconds;
2490 	if (timeout == PFTM_INTERVAL && seconds < old)
2491 		wakeup(pf_purge_thread);
2492 
2493 	if (prev_seconds != NULL)
2494 		*prev_seconds = old;
2495 
2496 	PF_RULES_WUNLOCK();
2497 
2498 	return (0);
2499 }
2500 
2501 int
2502 pf_ioctl_get_timeout(int timeout, int *seconds)
2503 {
2504 	PF_RULES_RLOCK_TRACKER;
2505 
2506 	if (timeout < 0 || timeout >= PFTM_MAX)
2507 		return (EINVAL);
2508 
2509 	PF_RULES_RLOCK();
2510 	*seconds = V_pf_default_rule.timeout[timeout];
2511 	PF_RULES_RUNLOCK();
2512 
2513 	return (0);
2514 }
2515 
2516 int
2517 pf_ioctl_set_limit(int index, unsigned int limit, unsigned int *old_limit)
2518 {
2519 
2520 	PF_RULES_WLOCK();
2521 	if (index < 0 || index >= PF_LIMIT_MAX ||
2522 	    V_pf_limits[index].zone == NULL) {
2523 		PF_RULES_WUNLOCK();
2524 		return (EINVAL);
2525 	}
2526 	uma_zone_set_max(V_pf_limits[index].zone, limit);
2527 	if (old_limit != NULL)
2528 		*old_limit = V_pf_limits[index].limit;
2529 	V_pf_limits[index].limit = limit;
2530 	PF_RULES_WUNLOCK();
2531 
2532 	return (0);
2533 }
2534 
2535 int
2536 pf_ioctl_get_limit(int index, unsigned int *limit)
2537 {
2538 	PF_RULES_RLOCK_TRACKER;
2539 
2540 	if (index < 0 || index >= PF_LIMIT_MAX)
2541 		return (EINVAL);
2542 
2543 	PF_RULES_RLOCK();
2544 	*limit = V_pf_limits[index].limit;
2545 	PF_RULES_RUNLOCK();
2546 
2547 	return (0);
2548 }
2549 
2550 int
2551 pf_ioctl_begin_addrs(uint32_t *ticket)
2552 {
2553 	PF_RULES_WLOCK();
2554 	pf_empty_kpool(&V_pf_pabuf[0]);
2555 	pf_empty_kpool(&V_pf_pabuf[1]);
2556 	*ticket = ++V_ticket_pabuf;
2557 	PF_RULES_WUNLOCK();
2558 
2559 	return (0);
2560 }
2561 
2562 int
2563 pf_ioctl_add_addr(struct pf_nl_pooladdr *pp)
2564 {
2565 	struct pf_kpooladdr	*pa = NULL;
2566 	struct pfi_kkif		*kif = NULL;
2567 	int error;
2568 
2569 	if (pp->which != PF_RDR && pp->which != PF_NAT)
2570 		return (EINVAL);
2571 
2572 #ifndef INET
2573 	if (pp->af == AF_INET)
2574 		return (EAFNOSUPPORT);
2575 #endif /* INET */
2576 #ifndef INET6
2577 	if (pp->af == AF_INET6)
2578 		return (EAFNOSUPPORT);
2579 #endif /* INET6 */
2580 
2581 	if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2582 	    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2583 	    pp->addr.addr.type != PF_ADDR_TABLE)
2584 		return (EINVAL);
2585 
2586 	if (pp->addr.addr.p.dyn != NULL)
2587 		return (EINVAL);
2588 
2589 	pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2590 	error = pf_pooladdr_to_kpooladdr(&pp->addr, pa);
2591 	if (error != 0)
2592 		goto out;
2593 	if (pa->ifname[0])
2594 		kif = pf_kkif_create(M_WAITOK);
2595 	PF_RULES_WLOCK();
2596 	if (pp->ticket != V_ticket_pabuf) {
2597 		PF_RULES_WUNLOCK();
2598 		if (pa->ifname[0])
2599 			pf_kkif_free(kif);
2600 		error = EBUSY;
2601 		goto out;
2602 	}
2603 	if (pa->ifname[0]) {
2604 		pa->kif = pfi_kkif_attach(kif, pa->ifname);
2605 		kif = NULL;
2606 		pfi_kkif_ref(pa->kif);
2607 	} else
2608 		pa->kif = NULL;
2609 	if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2610 	    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2611 		if (pa->ifname[0])
2612 			pfi_kkif_unref(pa->kif);
2613 		PF_RULES_WUNLOCK();
2614 		goto out;
2615 	}
2616 	TAILQ_INSERT_TAIL(&V_pf_pabuf[pp->which == PF_RDR ? 1 : 0],
2617 	    pa, entries);
2618 	PF_RULES_WUNLOCK();
2619 
2620 	return (0);
2621 
2622 out:
2623 	free(pa, M_PFRULE);
2624 	return (error);
2625 }
2626 
2627 int
2628 pf_ioctl_get_addrs(struct pf_nl_pooladdr *pp)
2629 {
2630 	struct pf_kpool		*pool;
2631 	struct pf_kpooladdr	*pa;
2632 
2633 	PF_RULES_RLOCK_TRACKER;
2634 
2635 	if (pp->which != PF_RDR && pp->which != PF_NAT)
2636 		return (EINVAL);
2637 
2638 	pp->anchor[sizeof(pp->anchor) - 1] = 0;
2639 	pp->nr = 0;
2640 
2641 	PF_RULES_RLOCK();
2642 	pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2643 	    pp->r_num, 0, 1, 0, pp->which);
2644 	if (pool == NULL) {
2645 		PF_RULES_RUNLOCK();
2646 		return (EBUSY);
2647 	}
2648 	TAILQ_FOREACH(pa, &pool->list, entries)
2649 		pp->nr++;
2650 	PF_RULES_RUNLOCK();
2651 
2652 	return (0);
2653 }
2654 
2655 int
2656 pf_ioctl_get_addr(struct pf_nl_pooladdr *pp)
2657 {
2658 	struct pf_kpool		*pool;
2659 	struct pf_kpooladdr	*pa;
2660 	u_int32_t		 nr = 0;
2661 
2662 	if (pp->which != PF_RDR && pp->which != PF_NAT)
2663 		return (EINVAL);
2664 
2665 	PF_RULES_RLOCK_TRACKER;
2666 
2667 	pp->anchor[sizeof(pp->anchor) - 1] = 0;
2668 
2669 	PF_RULES_RLOCK();
2670 	pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2671 	    pp->r_num, 0, 1, 1, pp->which);
2672 	if (pool == NULL) {
2673 		PF_RULES_RUNLOCK();
2674 		return (EBUSY);
2675 	}
2676 	pa = TAILQ_FIRST(&pool->list);
2677 	while ((pa != NULL) && (nr < pp->nr)) {
2678 		pa = TAILQ_NEXT(pa, entries);
2679 		nr++;
2680 	}
2681 	if (pa == NULL) {
2682 		PF_RULES_RUNLOCK();
2683 		return (EBUSY);
2684 	}
2685 	pf_kpooladdr_to_pooladdr(pa, &pp->addr);
2686 	pf_addr_copyout(&pp->addr.addr);
2687 	PF_RULES_RUNLOCK();
2688 
2689 	return (0);
2690 }
2691 
2692 int
2693 pf_ioctl_get_rulesets(struct pfioc_ruleset *pr)
2694 {
2695 	struct pf_kruleset	*ruleset;
2696 	struct pf_kanchor	*anchor;
2697 
2698 	PF_RULES_RLOCK_TRACKER;
2699 
2700 	pr->path[sizeof(pr->path) - 1] = 0;
2701 
2702 	PF_RULES_RLOCK();
2703 	if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
2704 		PF_RULES_RUNLOCK();
2705 		return (ENOENT);
2706 	}
2707 	pr->nr = 0;
2708 	if (ruleset->anchor == NULL) {
2709 		/* XXX kludge for pf_main_ruleset */
2710 		RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
2711 			if (anchor->parent == NULL)
2712 				pr->nr++;
2713 	} else {
2714 		RB_FOREACH(anchor, pf_kanchor_node,
2715 		    &ruleset->anchor->children)
2716 			pr->nr++;
2717 	}
2718 	PF_RULES_RUNLOCK();
2719 
2720 	return (0);
2721 }
2722 
2723 int
2724 pf_ioctl_get_ruleset(struct pfioc_ruleset *pr)
2725 {
2726 	struct pf_kruleset	*ruleset;
2727 	struct pf_kanchor	*anchor;
2728 	u_int32_t		 nr = 0;
2729 	int			 error = 0;
2730 
2731 	PF_RULES_RLOCK_TRACKER;
2732 
2733 	PF_RULES_RLOCK();
2734 	if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
2735 		PF_RULES_RUNLOCK();
2736 		return (ENOENT);
2737 	}
2738 
2739 	pr->name[0] = 0;
2740 	if (ruleset->anchor == NULL) {
2741 		/* XXX kludge for pf_main_ruleset */
2742 		RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
2743 			if (anchor->parent == NULL && nr++ == pr->nr) {
2744 				strlcpy(pr->name, anchor->name,
2745 				    sizeof(pr->name));
2746 				break;
2747 			}
2748 	} else {
2749 		RB_FOREACH(anchor, pf_kanchor_node,
2750 		    &ruleset->anchor->children)
2751 			if (nr++ == pr->nr) {
2752 				strlcpy(pr->name, anchor->name,
2753 				    sizeof(pr->name));
2754 				break;
2755 			}
2756 	}
2757 	if (!pr->name[0])
2758 		error = EBUSY;
2759 	PF_RULES_RUNLOCK();
2760 
2761 	return (error);
2762 }
2763 
2764 static int
2765 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
2766 {
2767 	int			 error = 0;
2768 	PF_RULES_RLOCK_TRACKER;
2769 
2770 #define	ERROUT_IOCTL(target, x)					\
2771     do {								\
2772 	    error = (x);						\
2773 	    SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__);	\
2774 	    goto target;						\
2775     } while (0)
2776 
2777 
2778 	/* XXX keep in sync with switch() below */
2779 	if (securelevel_gt(td->td_ucred, 2))
2780 		switch (cmd) {
2781 		case DIOCGETRULES:
2782 		case DIOCGETRULENV:
2783 		case DIOCGETADDRS:
2784 		case DIOCGETADDR:
2785 		case DIOCGETSTATE:
2786 		case DIOCGETSTATENV:
2787 		case DIOCSETSTATUSIF:
2788 		case DIOCGETSTATUSNV:
2789 		case DIOCCLRSTATUS:
2790 		case DIOCNATLOOK:
2791 		case DIOCSETDEBUG:
2792 #ifdef COMPAT_FREEBSD14
2793 		case DIOCGETSTATES:
2794 		case DIOCGETSTATESV2:
2795 #endif
2796 		case DIOCGETTIMEOUT:
2797 		case DIOCCLRRULECTRS:
2798 		case DIOCGETLIMIT:
2799 		case DIOCGETALTQSV0:
2800 		case DIOCGETALTQSV1:
2801 		case DIOCGETALTQV0:
2802 		case DIOCGETALTQV1:
2803 		case DIOCGETQSTATSV0:
2804 		case DIOCGETQSTATSV1:
2805 		case DIOCGETRULESETS:
2806 		case DIOCGETRULESET:
2807 		case DIOCRGETTABLES:
2808 		case DIOCRGETTSTATS:
2809 		case DIOCRCLRTSTATS:
2810 		case DIOCRCLRADDRS:
2811 		case DIOCRADDADDRS:
2812 		case DIOCRDELADDRS:
2813 		case DIOCRSETADDRS:
2814 		case DIOCRGETADDRS:
2815 		case DIOCRGETASTATS:
2816 		case DIOCRCLRASTATS:
2817 		case DIOCRTSTADDRS:
2818 		case DIOCOSFPGET:
2819 		case DIOCGETSRCNODES:
2820 		case DIOCCLRSRCNODES:
2821 		case DIOCGETSYNCOOKIES:
2822 		case DIOCIGETIFACES:
2823 		case DIOCGIFSPEEDV0:
2824 		case DIOCGIFSPEEDV1:
2825 		case DIOCSETIFFLAG:
2826 		case DIOCCLRIFFLAG:
2827 		case DIOCGETETHRULES:
2828 		case DIOCGETETHRULE:
2829 		case DIOCGETETHRULESETS:
2830 		case DIOCGETETHRULESET:
2831 			break;
2832 		case DIOCRCLRTABLES:
2833 		case DIOCRADDTABLES:
2834 		case DIOCRDELTABLES:
2835 		case DIOCRSETTFLAGS:
2836 			if (((struct pfioc_table *)addr)->pfrio_flags &
2837 			    PFR_FLAG_DUMMY)
2838 				break; /* dummy operation ok */
2839 			return (EPERM);
2840 		default:
2841 			return (EPERM);
2842 		}
2843 
2844 	if (!(flags & FWRITE))
2845 		switch (cmd) {
2846 		case DIOCGETRULES:
2847 		case DIOCGETADDRS:
2848 		case DIOCGETADDR:
2849 		case DIOCGETSTATE:
2850 		case DIOCGETSTATENV:
2851 		case DIOCGETSTATUSNV:
2852 #ifdef COMPAT_FREEBSD14
2853 		case DIOCGETSTATES:
2854 		case DIOCGETSTATESV2:
2855 #endif
2856 		case DIOCGETTIMEOUT:
2857 		case DIOCGETLIMIT:
2858 		case DIOCGETALTQSV0:
2859 		case DIOCGETALTQSV1:
2860 		case DIOCGETALTQV0:
2861 		case DIOCGETALTQV1:
2862 		case DIOCGETQSTATSV0:
2863 		case DIOCGETQSTATSV1:
2864 		case DIOCGETRULESETS:
2865 		case DIOCGETRULESET:
2866 		case DIOCNATLOOK:
2867 		case DIOCRGETTABLES:
2868 		case DIOCRGETTSTATS:
2869 		case DIOCRGETADDRS:
2870 		case DIOCRGETASTATS:
2871 		case DIOCRTSTADDRS:
2872 		case DIOCOSFPGET:
2873 		case DIOCGETSRCNODES:
2874 		case DIOCGETSYNCOOKIES:
2875 		case DIOCIGETIFACES:
2876 		case DIOCGIFSPEEDV1:
2877 		case DIOCGIFSPEEDV0:
2878 		case DIOCGETRULENV:
2879 		case DIOCGETETHRULES:
2880 		case DIOCGETETHRULE:
2881 		case DIOCGETETHRULESETS:
2882 		case DIOCGETETHRULESET:
2883 			break;
2884 		case DIOCRCLRTABLES:
2885 		case DIOCRADDTABLES:
2886 		case DIOCRDELTABLES:
2887 		case DIOCRCLRTSTATS:
2888 		case DIOCRCLRADDRS:
2889 		case DIOCRADDADDRS:
2890 		case DIOCRDELADDRS:
2891 		case DIOCRSETADDRS:
2892 		case DIOCRSETTFLAGS:
2893 			if (((struct pfioc_table *)addr)->pfrio_flags &
2894 			    PFR_FLAG_DUMMY) {
2895 				flags |= FWRITE; /* need write lock for dummy */
2896 				break; /* dummy operation ok */
2897 			}
2898 			return (EACCES);
2899 		default:
2900 			return (EACCES);
2901 		}
2902 
2903 	CURVNET_SET(TD_TO_VNET(td));
2904 
2905 	switch (cmd) {
2906 #ifdef COMPAT_FREEBSD14
2907 	case DIOCSTART:
2908 		error = pf_start();
2909 		break;
2910 
2911 	case DIOCSTOP:
2912 		error = pf_stop();
2913 		break;
2914 #endif
2915 
2916 	case DIOCGETETHRULES: {
2917 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2918 		nvlist_t		*nvl;
2919 		void			*packed;
2920 		struct pf_keth_rule	*tail;
2921 		struct pf_keth_ruleset	*rs;
2922 		u_int32_t		 ticket, nr;
2923 		const char		*anchor = "";
2924 
2925 		nvl = NULL;
2926 		packed = NULL;
2927 
2928 #define	ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULES_error, x)
2929 
2930 		if (nv->len > pf_ioctl_maxcount)
2931 			ERROUT(ENOMEM);
2932 
2933 		/* Copy the request in */
2934 		packed = malloc(nv->len, M_NVLIST, M_WAITOK);
2935 		error = copyin(nv->data, packed, nv->len);
2936 		if (error)
2937 			ERROUT(error);
2938 
2939 		nvl = nvlist_unpack(packed, nv->len, 0);
2940 		if (nvl == NULL)
2941 			ERROUT(EBADMSG);
2942 
2943 		if (! nvlist_exists_string(nvl, "anchor"))
2944 			ERROUT(EBADMSG);
2945 
2946 		anchor = nvlist_get_string(nvl, "anchor");
2947 
2948 		rs = pf_find_keth_ruleset(anchor);
2949 
2950 		nvlist_destroy(nvl);
2951 		nvl = NULL;
2952 		free(packed, M_NVLIST);
2953 		packed = NULL;
2954 
2955 		if (rs == NULL)
2956 			ERROUT(ENOENT);
2957 
2958 		/* Reply */
2959 		nvl = nvlist_create(0);
2960 		if (nvl == NULL)
2961 			ERROUT(ENOMEM);
2962 
2963 		PF_RULES_RLOCK();
2964 
2965 		ticket = rs->active.ticket;
2966 		tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq);
2967 		if (tail)
2968 			nr = tail->nr + 1;
2969 		else
2970 			nr = 0;
2971 
2972 		PF_RULES_RUNLOCK();
2973 
2974 		nvlist_add_number(nvl, "ticket", ticket);
2975 		nvlist_add_number(nvl, "nr", nr);
2976 
2977 		packed = nvlist_pack(nvl, &nv->len);
2978 		if (packed == NULL)
2979 			ERROUT(ENOMEM);
2980 
2981 		if (nv->size == 0)
2982 			ERROUT(0);
2983 		else if (nv->size < nv->len)
2984 			ERROUT(ENOSPC);
2985 
2986 		error = copyout(packed, nv->data, nv->len);
2987 
2988 #undef ERROUT
2989 DIOCGETETHRULES_error:
2990 		free(packed, M_NVLIST);
2991 		nvlist_destroy(nvl);
2992 		break;
2993 	}
2994 
2995 	case DIOCGETETHRULE: {
2996 		struct epoch_tracker	 et;
2997 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2998 		nvlist_t		*nvl = NULL;
2999 		void			*nvlpacked = NULL;
3000 		struct pf_keth_rule	*rule = NULL;
3001 		struct pf_keth_ruleset	*rs;
3002 		u_int32_t		 ticket, nr;
3003 		bool			 clear = false;
3004 		const char		*anchor;
3005 
3006 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULE_error, x)
3007 
3008 		if (nv->len > pf_ioctl_maxcount)
3009 			ERROUT(ENOMEM);
3010 
3011 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3012 		error = copyin(nv->data, nvlpacked, nv->len);
3013 		if (error)
3014 			ERROUT(error);
3015 
3016 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3017 		if (nvl == NULL)
3018 			ERROUT(EBADMSG);
3019 		if (! nvlist_exists_number(nvl, "ticket"))
3020 			ERROUT(EBADMSG);
3021 		ticket = nvlist_get_number(nvl, "ticket");
3022 		if (! nvlist_exists_string(nvl, "anchor"))
3023 			ERROUT(EBADMSG);
3024 		anchor = nvlist_get_string(nvl, "anchor");
3025 
3026 		if (nvlist_exists_bool(nvl, "clear"))
3027 			clear = nvlist_get_bool(nvl, "clear");
3028 
3029 		if (clear && !(flags & FWRITE))
3030 			ERROUT(EACCES);
3031 
3032 		if (! nvlist_exists_number(nvl, "nr"))
3033 			ERROUT(EBADMSG);
3034 		nr = nvlist_get_number(nvl, "nr");
3035 
3036 		PF_RULES_RLOCK();
3037 		rs = pf_find_keth_ruleset(anchor);
3038 		if (rs == NULL) {
3039 			PF_RULES_RUNLOCK();
3040 			ERROUT(ENOENT);
3041 		}
3042 		if (ticket != rs->active.ticket) {
3043 			PF_RULES_RUNLOCK();
3044 			ERROUT(EBUSY);
3045 		}
3046 
3047 		nvlist_destroy(nvl);
3048 		nvl = NULL;
3049 		free(nvlpacked, M_NVLIST);
3050 		nvlpacked = NULL;
3051 
3052 		rule = TAILQ_FIRST(rs->active.rules);
3053 		while ((rule != NULL) && (rule->nr != nr))
3054 			rule = TAILQ_NEXT(rule, entries);
3055 		if (rule == NULL) {
3056 			PF_RULES_RUNLOCK();
3057 			ERROUT(ENOENT);
3058 		}
3059 		/* Make sure rule can't go away. */
3060 		NET_EPOCH_ENTER(et);
3061 		PF_RULES_RUNLOCK();
3062 		nvl = pf_keth_rule_to_nveth_rule(rule);
3063 		if (pf_keth_anchor_nvcopyout(rs, rule, nvl)) {
3064 			NET_EPOCH_EXIT(et);
3065 			ERROUT(EBUSY);
3066 		}
3067 		NET_EPOCH_EXIT(et);
3068 		if (nvl == NULL)
3069 			ERROUT(ENOMEM);
3070 
3071 		nvlpacked = nvlist_pack(nvl, &nv->len);
3072 		if (nvlpacked == NULL)
3073 			ERROUT(ENOMEM);
3074 
3075 		if (nv->size == 0)
3076 			ERROUT(0);
3077 		else if (nv->size < nv->len)
3078 			ERROUT(ENOSPC);
3079 
3080 		error = copyout(nvlpacked, nv->data, nv->len);
3081 		if (error == 0 && clear) {
3082 			counter_u64_zero(rule->evaluations);
3083 			for (int i = 0; i < 2; i++) {
3084 				counter_u64_zero(rule->packets[i]);
3085 				counter_u64_zero(rule->bytes[i]);
3086 			}
3087 		}
3088 
3089 #undef ERROUT
3090 DIOCGETETHRULE_error:
3091 		free(nvlpacked, M_NVLIST);
3092 		nvlist_destroy(nvl);
3093 		break;
3094 	}
3095 
3096 	case DIOCADDETHRULE: {
3097 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
3098 		nvlist_t		*nvl = NULL;
3099 		void			*nvlpacked = NULL;
3100 		struct pf_keth_rule	*rule = NULL, *tail = NULL;
3101 		struct pf_keth_ruleset	*ruleset = NULL;
3102 		struct pfi_kkif		*kif = NULL, *bridge_to_kif = NULL;
3103 		const char		*anchor = "", *anchor_call = "";
3104 
3105 #define ERROUT(x)	ERROUT_IOCTL(DIOCADDETHRULE_error, x)
3106 
3107 		if (nv->len > pf_ioctl_maxcount)
3108 			ERROUT(ENOMEM);
3109 
3110 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3111 		error = copyin(nv->data, nvlpacked, nv->len);
3112 		if (error)
3113 			ERROUT(error);
3114 
3115 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3116 		if (nvl == NULL)
3117 			ERROUT(EBADMSG);
3118 
3119 		if (! nvlist_exists_number(nvl, "ticket"))
3120 			ERROUT(EBADMSG);
3121 
3122 		if (nvlist_exists_string(nvl, "anchor"))
3123 			anchor = nvlist_get_string(nvl, "anchor");
3124 		if (nvlist_exists_string(nvl, "anchor_call"))
3125 			anchor_call = nvlist_get_string(nvl, "anchor_call");
3126 
3127 		ruleset = pf_find_keth_ruleset(anchor);
3128 		if (ruleset == NULL)
3129 			ERROUT(EINVAL);
3130 
3131 		if (nvlist_get_number(nvl, "ticket") !=
3132 		    ruleset->inactive.ticket) {
3133 			DPFPRINTF(PF_DEBUG_MISC,
3134 			    ("ticket: %d != %d\n",
3135 			    (u_int32_t)nvlist_get_number(nvl, "ticket"),
3136 			    ruleset->inactive.ticket));
3137 			ERROUT(EBUSY);
3138 		}
3139 
3140 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
3141 		rule->timestamp = NULL;
3142 
3143 		error = pf_nveth_rule_to_keth_rule(nvl, rule);
3144 		if (error != 0)
3145 			ERROUT(error);
3146 
3147 		if (rule->ifname[0])
3148 			kif = pf_kkif_create(M_WAITOK);
3149 		if (rule->bridge_to_name[0])
3150 			bridge_to_kif = pf_kkif_create(M_WAITOK);
3151 		rule->evaluations = counter_u64_alloc(M_WAITOK);
3152 		for (int i = 0; i < 2; i++) {
3153 			rule->packets[i] = counter_u64_alloc(M_WAITOK);
3154 			rule->bytes[i] = counter_u64_alloc(M_WAITOK);
3155 		}
3156 		rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
3157 		    M_WAITOK | M_ZERO);
3158 
3159 		PF_RULES_WLOCK();
3160 
3161 		if (rule->ifname[0]) {
3162 			rule->kif = pfi_kkif_attach(kif, rule->ifname);
3163 			pfi_kkif_ref(rule->kif);
3164 		} else
3165 			rule->kif = NULL;
3166 		if (rule->bridge_to_name[0]) {
3167 			rule->bridge_to = pfi_kkif_attach(bridge_to_kif,
3168 			    rule->bridge_to_name);
3169 			pfi_kkif_ref(rule->bridge_to);
3170 		} else
3171 			rule->bridge_to = NULL;
3172 
3173 #ifdef ALTQ
3174 		/* set queue IDs */
3175 		if (rule->qname[0] != 0) {
3176 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
3177 				error = EBUSY;
3178 			else
3179 				rule->qid = rule->qid;
3180 		}
3181 #endif
3182 		if (rule->tagname[0])
3183 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
3184 				error = EBUSY;
3185 		if (rule->match_tagname[0])
3186 			if ((rule->match_tag = pf_tagname2tag(
3187 			    rule->match_tagname)) == 0)
3188 				error = EBUSY;
3189 
3190 		if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE)
3191 			error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr);
3192 		if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE)
3193 			error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr);
3194 
3195 		if (error) {
3196 			pf_free_eth_rule(rule);
3197 			PF_RULES_WUNLOCK();
3198 			ERROUT(error);
3199 		}
3200 
3201 		if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) {
3202 			pf_free_eth_rule(rule);
3203 			PF_RULES_WUNLOCK();
3204 			ERROUT(EINVAL);
3205 		}
3206 
3207 		tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq);
3208 		if (tail)
3209 			rule->nr = tail->nr + 1;
3210 		else
3211 			rule->nr = 0;
3212 
3213 		TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries);
3214 
3215 		PF_RULES_WUNLOCK();
3216 
3217 #undef ERROUT
3218 DIOCADDETHRULE_error:
3219 		nvlist_destroy(nvl);
3220 		free(nvlpacked, M_NVLIST);
3221 		break;
3222 	}
3223 
3224 	case DIOCGETETHRULESETS: {
3225 		struct epoch_tracker	 et;
3226 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
3227 		nvlist_t		*nvl = NULL;
3228 		void			*nvlpacked = NULL;
3229 		struct pf_keth_ruleset	*ruleset;
3230 		struct pf_keth_anchor	*anchor;
3231 		int			 nr = 0;
3232 
3233 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULESETS_error, x)
3234 
3235 		if (nv->len > pf_ioctl_maxcount)
3236 			ERROUT(ENOMEM);
3237 
3238 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3239 		error = copyin(nv->data, nvlpacked, nv->len);
3240 		if (error)
3241 			ERROUT(error);
3242 
3243 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3244 		if (nvl == NULL)
3245 			ERROUT(EBADMSG);
3246 		if (! nvlist_exists_string(nvl, "path"))
3247 			ERROUT(EBADMSG);
3248 
3249 		NET_EPOCH_ENTER(et);
3250 
3251 		if ((ruleset = pf_find_keth_ruleset(
3252 		    nvlist_get_string(nvl, "path"))) == NULL) {
3253 			NET_EPOCH_EXIT(et);
3254 			ERROUT(ENOENT);
3255 		}
3256 
3257 		if (ruleset->anchor == NULL) {
3258 			RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors)
3259 				if (anchor->parent == NULL)
3260 					nr++;
3261 		} else {
3262 			RB_FOREACH(anchor, pf_keth_anchor_node,
3263 			    &ruleset->anchor->children)
3264 				nr++;
3265 		}
3266 
3267 		NET_EPOCH_EXIT(et);
3268 
3269 		nvlist_destroy(nvl);
3270 		nvl = NULL;
3271 		free(nvlpacked, M_NVLIST);
3272 		nvlpacked = NULL;
3273 
3274 		nvl = nvlist_create(0);
3275 		if (nvl == NULL)
3276 			ERROUT(ENOMEM);
3277 
3278 		nvlist_add_number(nvl, "nr", nr);
3279 
3280 		nvlpacked = nvlist_pack(nvl, &nv->len);
3281 		if (nvlpacked == NULL)
3282 			ERROUT(ENOMEM);
3283 
3284 		if (nv->size == 0)
3285 			ERROUT(0);
3286 		else if (nv->size < nv->len)
3287 			ERROUT(ENOSPC);
3288 
3289 		error = copyout(nvlpacked, nv->data, nv->len);
3290 
3291 #undef ERROUT
3292 DIOCGETETHRULESETS_error:
3293 		free(nvlpacked, M_NVLIST);
3294 		nvlist_destroy(nvl);
3295 		break;
3296 	}
3297 
3298 	case DIOCGETETHRULESET: {
3299 		struct epoch_tracker	 et;
3300 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
3301 		nvlist_t		*nvl = NULL;
3302 		void			*nvlpacked = NULL;
3303 		struct pf_keth_ruleset	*ruleset;
3304 		struct pf_keth_anchor	*anchor;
3305 		int			 nr = 0, req_nr = 0;
3306 		bool			 found = false;
3307 
3308 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULESET_error, x)
3309 
3310 		if (nv->len > pf_ioctl_maxcount)
3311 			ERROUT(ENOMEM);
3312 
3313 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3314 		error = copyin(nv->data, nvlpacked, nv->len);
3315 		if (error)
3316 			ERROUT(error);
3317 
3318 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3319 		if (nvl == NULL)
3320 			ERROUT(EBADMSG);
3321 		if (! nvlist_exists_string(nvl, "path"))
3322 			ERROUT(EBADMSG);
3323 		if (! nvlist_exists_number(nvl, "nr"))
3324 			ERROUT(EBADMSG);
3325 
3326 		req_nr = nvlist_get_number(nvl, "nr");
3327 
3328 		NET_EPOCH_ENTER(et);
3329 
3330 		if ((ruleset = pf_find_keth_ruleset(
3331 		    nvlist_get_string(nvl, "path"))) == NULL) {
3332 			NET_EPOCH_EXIT(et);
3333 			ERROUT(ENOENT);
3334 		}
3335 
3336 		nvlist_destroy(nvl);
3337 		nvl = NULL;
3338 		free(nvlpacked, M_NVLIST);
3339 		nvlpacked = NULL;
3340 
3341 		nvl = nvlist_create(0);
3342 		if (nvl == NULL) {
3343 			NET_EPOCH_EXIT(et);
3344 			ERROUT(ENOMEM);
3345 		}
3346 
3347 		if (ruleset->anchor == NULL) {
3348 			RB_FOREACH(anchor, pf_keth_anchor_global,
3349 			    &V_pf_keth_anchors) {
3350 				if (anchor->parent == NULL && nr++ == req_nr) {
3351 					found = true;
3352 					break;
3353 				}
3354 			}
3355 		} else {
3356 			RB_FOREACH(anchor, pf_keth_anchor_node,
3357 			     &ruleset->anchor->children) {
3358 				if (nr++ == req_nr) {
3359 					found = true;
3360 					break;
3361 				}
3362 			}
3363 		}
3364 
3365 		NET_EPOCH_EXIT(et);
3366 		if (found) {
3367 			nvlist_add_number(nvl, "nr", nr);
3368 			nvlist_add_string(nvl, "name", anchor->name);
3369 			if (ruleset->anchor)
3370 				nvlist_add_string(nvl, "path",
3371 				    ruleset->anchor->path);
3372 			else
3373 				nvlist_add_string(nvl, "path", "");
3374 		} else {
3375 			ERROUT(EBUSY);
3376 		}
3377 
3378 		nvlpacked = nvlist_pack(nvl, &nv->len);
3379 		if (nvlpacked == NULL)
3380 			ERROUT(ENOMEM);
3381 
3382 		if (nv->size == 0)
3383 			ERROUT(0);
3384 		else if (nv->size < nv->len)
3385 			ERROUT(ENOSPC);
3386 
3387 		error = copyout(nvlpacked, nv->data, nv->len);
3388 
3389 #undef ERROUT
3390 DIOCGETETHRULESET_error:
3391 		free(nvlpacked, M_NVLIST);
3392 		nvlist_destroy(nvl);
3393 		break;
3394 	}
3395 
3396 	case DIOCADDRULENV: {
3397 		struct pfioc_nv	*nv = (struct pfioc_nv *)addr;
3398 		nvlist_t	*nvl = NULL;
3399 		void		*nvlpacked = NULL;
3400 		struct pf_krule	*rule = NULL;
3401 		const char	*anchor = "", *anchor_call = "";
3402 		uint32_t	 ticket = 0, pool_ticket = 0;
3403 
3404 #define	ERROUT(x)	ERROUT_IOCTL(DIOCADDRULENV_error, x)
3405 
3406 		if (nv->len > pf_ioctl_maxcount)
3407 			ERROUT(ENOMEM);
3408 
3409 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3410 		error = copyin(nv->data, nvlpacked, nv->len);
3411 		if (error)
3412 			ERROUT(error);
3413 
3414 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3415 		if (nvl == NULL)
3416 			ERROUT(EBADMSG);
3417 
3418 		if (! nvlist_exists_number(nvl, "ticket"))
3419 			ERROUT(EINVAL);
3420 		ticket = nvlist_get_number(nvl, "ticket");
3421 
3422 		if (! nvlist_exists_number(nvl, "pool_ticket"))
3423 			ERROUT(EINVAL);
3424 		pool_ticket = nvlist_get_number(nvl, "pool_ticket");
3425 
3426 		if (! nvlist_exists_nvlist(nvl, "rule"))
3427 			ERROUT(EINVAL);
3428 
3429 		rule = pf_krule_alloc();
3430 		error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"),
3431 		    rule);
3432 		if (error)
3433 			ERROUT(error);
3434 
3435 		if (nvlist_exists_string(nvl, "anchor"))
3436 			anchor = nvlist_get_string(nvl, "anchor");
3437 		if (nvlist_exists_string(nvl, "anchor_call"))
3438 			anchor_call = nvlist_get_string(nvl, "anchor_call");
3439 
3440 		if ((error = nvlist_error(nvl)))
3441 			ERROUT(error);
3442 
3443 		/* Frees rule on error */
3444 		error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor,
3445 		    anchor_call, td->td_ucred->cr_ruid,
3446 		    td->td_proc ? td->td_proc->p_pid : 0);
3447 
3448 		nvlist_destroy(nvl);
3449 		free(nvlpacked, M_NVLIST);
3450 		break;
3451 #undef ERROUT
3452 DIOCADDRULENV_error:
3453 		pf_krule_free(rule);
3454 		nvlist_destroy(nvl);
3455 		free(nvlpacked, M_NVLIST);
3456 
3457 		break;
3458 	}
3459 	case DIOCADDRULE: {
3460 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
3461 		struct pf_krule		*rule;
3462 
3463 		rule = pf_krule_alloc();
3464 		error = pf_rule_to_krule(&pr->rule, rule);
3465 		if (error != 0) {
3466 			pf_krule_free(rule);
3467 			break;
3468 		}
3469 
3470 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
3471 
3472 		/* Frees rule on error */
3473 		error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket,
3474 		    pr->anchor, pr->anchor_call, td->td_ucred->cr_ruid,
3475 		    td->td_proc ? td->td_proc->p_pid : 0);
3476 		break;
3477 	}
3478 
3479 	case DIOCGETRULES: {
3480 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
3481 
3482 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
3483 
3484 		error = pf_ioctl_getrules(pr);
3485 
3486 		break;
3487 	}
3488 
3489 	case DIOCGETRULENV: {
3490 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
3491 		nvlist_t		*nvrule = NULL;
3492 		nvlist_t		*nvl = NULL;
3493 		struct pf_kruleset	*ruleset;
3494 		struct pf_krule		*rule;
3495 		void			*nvlpacked = NULL;
3496 		int			 rs_num, nr;
3497 		bool			 clear_counter = false;
3498 
3499 #define	ERROUT(x)	ERROUT_IOCTL(DIOCGETRULENV_error, x)
3500 
3501 		if (nv->len > pf_ioctl_maxcount)
3502 			ERROUT(ENOMEM);
3503 
3504 		/* Copy the request in */
3505 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3506 		error = copyin(nv->data, nvlpacked, nv->len);
3507 		if (error)
3508 			ERROUT(error);
3509 
3510 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3511 		if (nvl == NULL)
3512 			ERROUT(EBADMSG);
3513 
3514 		if (! nvlist_exists_string(nvl, "anchor"))
3515 			ERROUT(EBADMSG);
3516 		if (! nvlist_exists_number(nvl, "ruleset"))
3517 			ERROUT(EBADMSG);
3518 		if (! nvlist_exists_number(nvl, "ticket"))
3519 			ERROUT(EBADMSG);
3520 		if (! nvlist_exists_number(nvl, "nr"))
3521 			ERROUT(EBADMSG);
3522 
3523 		if (nvlist_exists_bool(nvl, "clear_counter"))
3524 			clear_counter = nvlist_get_bool(nvl, "clear_counter");
3525 
3526 		if (clear_counter && !(flags & FWRITE))
3527 			ERROUT(EACCES);
3528 
3529 		nr = nvlist_get_number(nvl, "nr");
3530 
3531 		PF_RULES_WLOCK();
3532 		ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor"));
3533 		if (ruleset == NULL) {
3534 			PF_RULES_WUNLOCK();
3535 			ERROUT(ENOENT);
3536 		}
3537 
3538 		rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset"));
3539 		if (rs_num >= PF_RULESET_MAX) {
3540 			PF_RULES_WUNLOCK();
3541 			ERROUT(EINVAL);
3542 		}
3543 
3544 		if (nvlist_get_number(nvl, "ticket") !=
3545 		    ruleset->rules[rs_num].active.ticket) {
3546 			PF_RULES_WUNLOCK();
3547 			ERROUT(EBUSY);
3548 		}
3549 
3550 		if ((error = nvlist_error(nvl))) {
3551 			PF_RULES_WUNLOCK();
3552 			ERROUT(error);
3553 		}
3554 
3555 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
3556 		while ((rule != NULL) && (rule->nr != nr))
3557 			rule = TAILQ_NEXT(rule, entries);
3558 		if (rule == NULL) {
3559 			PF_RULES_WUNLOCK();
3560 			ERROUT(EBUSY);
3561 		}
3562 
3563 		nvrule = pf_krule_to_nvrule(rule);
3564 
3565 		nvlist_destroy(nvl);
3566 		nvl = nvlist_create(0);
3567 		if (nvl == NULL) {
3568 			PF_RULES_WUNLOCK();
3569 			ERROUT(ENOMEM);
3570 		}
3571 		nvlist_add_number(nvl, "nr", nr);
3572 		nvlist_add_nvlist(nvl, "rule", nvrule);
3573 		nvlist_destroy(nvrule);
3574 		nvrule = NULL;
3575 		if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) {
3576 			PF_RULES_WUNLOCK();
3577 			ERROUT(EBUSY);
3578 		}
3579 
3580 		free(nvlpacked, M_NVLIST);
3581 		nvlpacked = nvlist_pack(nvl, &nv->len);
3582 		if (nvlpacked == NULL) {
3583 			PF_RULES_WUNLOCK();
3584 			ERROUT(ENOMEM);
3585 		}
3586 
3587 		if (nv->size == 0) {
3588 			PF_RULES_WUNLOCK();
3589 			ERROUT(0);
3590 		}
3591 		else if (nv->size < nv->len) {
3592 			PF_RULES_WUNLOCK();
3593 			ERROUT(ENOSPC);
3594 		}
3595 
3596 		if (clear_counter)
3597 			pf_krule_clear_counters(rule);
3598 
3599 		PF_RULES_WUNLOCK();
3600 
3601 		error = copyout(nvlpacked, nv->data, nv->len);
3602 
3603 #undef ERROUT
3604 DIOCGETRULENV_error:
3605 		free(nvlpacked, M_NVLIST);
3606 		nvlist_destroy(nvrule);
3607 		nvlist_destroy(nvl);
3608 
3609 		break;
3610 	}
3611 
3612 	case DIOCCHANGERULE: {
3613 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
3614 		struct pf_kruleset	*ruleset;
3615 		struct pf_krule		*oldrule = NULL, *newrule = NULL;
3616 		struct pfi_kkif		*kif = NULL;
3617 		struct pf_kpooladdr	*pa;
3618 		u_int32_t		 nr = 0;
3619 		int			 rs_num;
3620 
3621 		pcr->anchor[sizeof(pcr->anchor) - 1] = 0;
3622 
3623 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
3624 		    pcr->action > PF_CHANGE_GET_TICKET) {
3625 			error = EINVAL;
3626 			break;
3627 		}
3628 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
3629 			error = EINVAL;
3630 			break;
3631 		}
3632 
3633 		if (pcr->action != PF_CHANGE_REMOVE) {
3634 			newrule = pf_krule_alloc();
3635 			error = pf_rule_to_krule(&pcr->rule, newrule);
3636 			if (error != 0) {
3637 				pf_krule_free(newrule);
3638 				break;
3639 			}
3640 
3641 			if (newrule->ifname[0])
3642 				kif = pf_kkif_create(M_WAITOK);
3643 			pf_counter_u64_init(&newrule->evaluations, M_WAITOK);
3644 			for (int i = 0; i < 2; i++) {
3645 				pf_counter_u64_init(&newrule->packets[i], M_WAITOK);
3646 				pf_counter_u64_init(&newrule->bytes[i], M_WAITOK);
3647 			}
3648 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
3649 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
3650 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
3651 			newrule->cuid = td->td_ucred->cr_ruid;
3652 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
3653 			TAILQ_INIT(&newrule->nat.list);
3654 			TAILQ_INIT(&newrule->rdr.list);
3655 		}
3656 #define	ERROUT(x)	ERROUT_IOCTL(DIOCCHANGERULE_error, x)
3657 
3658 		PF_CONFIG_LOCK();
3659 		PF_RULES_WLOCK();
3660 #ifdef PF_WANT_32_TO_64_COUNTER
3661 		if (newrule != NULL) {
3662 			LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist);
3663 			newrule->allrulelinked = true;
3664 			V_pf_allrulecount++;
3665 		}
3666 #endif
3667 
3668 		if (!(pcr->action == PF_CHANGE_REMOVE ||
3669 		    pcr->action == PF_CHANGE_GET_TICKET) &&
3670 		    pcr->pool_ticket != V_ticket_pabuf)
3671 			ERROUT(EBUSY);
3672 
3673 		ruleset = pf_find_kruleset(pcr->anchor);
3674 		if (ruleset == NULL)
3675 			ERROUT(EINVAL);
3676 
3677 		rs_num = pf_get_ruleset_number(pcr->rule.action);
3678 		if (rs_num >= PF_RULESET_MAX)
3679 			ERROUT(EINVAL);
3680 
3681 		/*
3682 		 * XXXMJG: there is no guarantee that the ruleset was
3683 		 * created by the usual route of calling DIOCXBEGIN.
3684 		 * As a result it is possible the rule tree will not
3685 		 * be allocated yet. Hack around it by doing it here.
3686 		 * Note it is fine to let the tree persist in case of
3687 		 * error as it will be freed down the road on future
3688 		 * updates (if need be).
3689 		 */
3690 		if (ruleset->rules[rs_num].active.tree == NULL) {
3691 			ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT);
3692 			if (ruleset->rules[rs_num].active.tree == NULL) {
3693 				ERROUT(ENOMEM);
3694 			}
3695 		}
3696 
3697 		if (pcr->action == PF_CHANGE_GET_TICKET) {
3698 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
3699 			ERROUT(0);
3700 		} else if (pcr->ticket !=
3701 			    ruleset->rules[rs_num].active.ticket)
3702 				ERROUT(EINVAL);
3703 
3704 		if (pcr->action != PF_CHANGE_REMOVE) {
3705 			if (newrule->ifname[0]) {
3706 				newrule->kif = pfi_kkif_attach(kif,
3707 				    newrule->ifname);
3708 				kif = NULL;
3709 				pfi_kkif_ref(newrule->kif);
3710 			} else
3711 				newrule->kif = NULL;
3712 
3713 			if (newrule->rtableid > 0 &&
3714 			    newrule->rtableid >= rt_numfibs)
3715 				error = EBUSY;
3716 
3717 #ifdef ALTQ
3718 			/* set queue IDs */
3719 			if (newrule->qname[0] != 0) {
3720 				if ((newrule->qid =
3721 				    pf_qname2qid(newrule->qname)) == 0)
3722 					error = EBUSY;
3723 				else if (newrule->pqname[0] != 0) {
3724 					if ((newrule->pqid =
3725 					    pf_qname2qid(newrule->pqname)) == 0)
3726 						error = EBUSY;
3727 				} else
3728 					newrule->pqid = newrule->qid;
3729 			}
3730 #endif /* ALTQ */
3731 			if (newrule->tagname[0])
3732 				if ((newrule->tag =
3733 				    pf_tagname2tag(newrule->tagname)) == 0)
3734 					error = EBUSY;
3735 			if (newrule->match_tagname[0])
3736 				if ((newrule->match_tag = pf_tagname2tag(
3737 				    newrule->match_tagname)) == 0)
3738 					error = EBUSY;
3739 			if (newrule->rt && !newrule->direction)
3740 				error = EINVAL;
3741 			if (!newrule->log)
3742 				newrule->logif = 0;
3743 			if (newrule->logif >= PFLOGIFS_MAX)
3744 				error = EINVAL;
3745 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
3746 				error = ENOMEM;
3747 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
3748 				error = ENOMEM;
3749 			if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
3750 				error = EINVAL;
3751 			for (int i = 0; i < 2; i++) {
3752 				TAILQ_FOREACH(pa, &V_pf_pabuf[i], entries)
3753 					if (pa->addr.type == PF_ADDR_TABLE) {
3754 						pa->addr.p.tbl =
3755 						    pfr_attach_table(ruleset,
3756 						    pa->addr.v.tblname);
3757 						if (pa->addr.p.tbl == NULL)
3758 							error = ENOMEM;
3759 					}
3760 			}
3761 
3762 			newrule->overload_tbl = NULL;
3763 			if (newrule->overload_tblname[0]) {
3764 				if ((newrule->overload_tbl = pfr_attach_table(
3765 				    ruleset, newrule->overload_tblname)) ==
3766 				    NULL)
3767 					error = EINVAL;
3768 				else
3769 					newrule->overload_tbl->pfrkt_flags |=
3770 					    PFR_TFLAG_ACTIVE;
3771 			}
3772 
3773 			pf_mv_kpool(&V_pf_pabuf[0], &newrule->nat.list);
3774 			pf_mv_kpool(&V_pf_pabuf[1], &newrule->rdr.list);
3775 			if (((((newrule->action == PF_NAT) ||
3776 			    (newrule->action == PF_RDR) ||
3777 			    (newrule->action == PF_BINAT) ||
3778 			    (newrule->rt > PF_NOPFROUTE)) &&
3779 			    !newrule->anchor)) &&
3780 			    (TAILQ_FIRST(&newrule->rdr.list) == NULL))
3781 				error = EINVAL;
3782 
3783 			if (error) {
3784 				pf_free_rule(newrule);
3785 				PF_RULES_WUNLOCK();
3786 				PF_CONFIG_UNLOCK();
3787 				break;
3788 			}
3789 
3790 			newrule->nat.cur = TAILQ_FIRST(&newrule->nat.list);
3791 			newrule->rdr.cur = TAILQ_FIRST(&newrule->rdr.list);
3792 		}
3793 		pf_empty_kpool(&V_pf_pabuf[0]);
3794 		pf_empty_kpool(&V_pf_pabuf[1]);
3795 
3796 		if (pcr->action == PF_CHANGE_ADD_HEAD)
3797 			oldrule = TAILQ_FIRST(
3798 			    ruleset->rules[rs_num].active.ptr);
3799 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
3800 			oldrule = TAILQ_LAST(
3801 			    ruleset->rules[rs_num].active.ptr, pf_krulequeue);
3802 		else {
3803 			oldrule = TAILQ_FIRST(
3804 			    ruleset->rules[rs_num].active.ptr);
3805 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
3806 				oldrule = TAILQ_NEXT(oldrule, entries);
3807 			if (oldrule == NULL) {
3808 				if (newrule != NULL)
3809 					pf_free_rule(newrule);
3810 				PF_RULES_WUNLOCK();
3811 				PF_CONFIG_UNLOCK();
3812 				error = EINVAL;
3813 				break;
3814 			}
3815 		}
3816 
3817 		if (pcr->action == PF_CHANGE_REMOVE) {
3818 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
3819 			    oldrule);
3820 			RB_REMOVE(pf_krule_global,
3821 			    ruleset->rules[rs_num].active.tree, oldrule);
3822 			ruleset->rules[rs_num].active.rcount--;
3823 		} else {
3824 			pf_hash_rule(newrule);
3825 			if (RB_INSERT(pf_krule_global,
3826 			    ruleset->rules[rs_num].active.tree, newrule) != NULL) {
3827 				pf_free_rule(newrule);
3828 				PF_RULES_WUNLOCK();
3829 				PF_CONFIG_UNLOCK();
3830 				error = EEXIST;
3831 				break;
3832 			}
3833 
3834 			if (oldrule == NULL)
3835 				TAILQ_INSERT_TAIL(
3836 				    ruleset->rules[rs_num].active.ptr,
3837 				    newrule, entries);
3838 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
3839 			    pcr->action == PF_CHANGE_ADD_BEFORE)
3840 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
3841 			else
3842 				TAILQ_INSERT_AFTER(
3843 				    ruleset->rules[rs_num].active.ptr,
3844 				    oldrule, newrule, entries);
3845 			ruleset->rules[rs_num].active.rcount++;
3846 		}
3847 
3848 		nr = 0;
3849 		TAILQ_FOREACH(oldrule,
3850 		    ruleset->rules[rs_num].active.ptr, entries)
3851 			oldrule->nr = nr++;
3852 
3853 		ruleset->rules[rs_num].active.ticket++;
3854 
3855 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
3856 		pf_remove_if_empty_kruleset(ruleset);
3857 
3858 		PF_RULES_WUNLOCK();
3859 		PF_CONFIG_UNLOCK();
3860 		break;
3861 
3862 #undef ERROUT
3863 DIOCCHANGERULE_error:
3864 		PF_RULES_WUNLOCK();
3865 		PF_CONFIG_UNLOCK();
3866 		pf_krule_free(newrule);
3867 		pf_kkif_free(kif);
3868 		break;
3869 	}
3870 
3871 	case DIOCCLRSTATESNV: {
3872 		error = pf_clearstates_nv((struct pfioc_nv *)addr);
3873 		break;
3874 	}
3875 
3876 	case DIOCKILLSTATESNV: {
3877 		error = pf_killstates_nv((struct pfioc_nv *)addr);
3878 		break;
3879 	}
3880 
3881 	case DIOCADDSTATE: {
3882 		struct pfioc_state		*ps = (struct pfioc_state *)addr;
3883 		struct pfsync_state_1301	*sp = &ps->state;
3884 
3885 		if (sp->timeout >= PFTM_MAX) {
3886 			error = EINVAL;
3887 			break;
3888 		}
3889 		if (V_pfsync_state_import_ptr != NULL) {
3890 			PF_RULES_RLOCK();
3891 			error = V_pfsync_state_import_ptr(
3892 			    (union pfsync_state_union *)sp, PFSYNC_SI_IOCTL,
3893 			    PFSYNC_MSG_VERSION_1301);
3894 			PF_RULES_RUNLOCK();
3895 		} else
3896 			error = EOPNOTSUPP;
3897 		break;
3898 	}
3899 
3900 	case DIOCGETSTATE: {
3901 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
3902 		struct pf_kstate	*s;
3903 
3904 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
3905 		if (s == NULL) {
3906 			error = ENOENT;
3907 			break;
3908 		}
3909 
3910 		pfsync_state_export((union pfsync_state_union*)&ps->state,
3911 		    s, PFSYNC_MSG_VERSION_1301);
3912 		PF_STATE_UNLOCK(s);
3913 		break;
3914 	}
3915 
3916 	case DIOCGETSTATENV: {
3917 		error = pf_getstate((struct pfioc_nv *)addr);
3918 		break;
3919 	}
3920 
3921 #ifdef COMPAT_FREEBSD14
3922 	case DIOCGETSTATES: {
3923 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
3924 		struct pf_kstate	*s;
3925 		struct pfsync_state_1301	*pstore, *p;
3926 		int			 i, nr;
3927 		size_t			 slice_count = 16, count;
3928 		void			*out;
3929 
3930 		if (ps->ps_len <= 0) {
3931 			nr = uma_zone_get_cur(V_pf_state_z);
3932 			ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3933 			break;
3934 		}
3935 
3936 		out = ps->ps_states;
3937 		pstore = mallocarray(slice_count,
3938 		    sizeof(struct pfsync_state_1301), M_TEMP, M_WAITOK | M_ZERO);
3939 		nr = 0;
3940 
3941 		for (i = 0; i <= V_pf_hashmask; i++) {
3942 			struct pf_idhash *ih = &V_pf_idhash[i];
3943 
3944 DIOCGETSTATES_retry:
3945 			p = pstore;
3946 
3947 			if (LIST_EMPTY(&ih->states))
3948 				continue;
3949 
3950 			PF_HASHROW_LOCK(ih);
3951 			count = 0;
3952 			LIST_FOREACH(s, &ih->states, entry) {
3953 				if (s->timeout == PFTM_UNLINKED)
3954 					continue;
3955 				count++;
3956 			}
3957 
3958 			if (count > slice_count) {
3959 				PF_HASHROW_UNLOCK(ih);
3960 				free(pstore, M_TEMP);
3961 				slice_count = count * 2;
3962 				pstore = mallocarray(slice_count,
3963 				    sizeof(struct pfsync_state_1301), M_TEMP,
3964 				    M_WAITOK | M_ZERO);
3965 				goto DIOCGETSTATES_retry;
3966 			}
3967 
3968 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
3969 				PF_HASHROW_UNLOCK(ih);
3970 				goto DIOCGETSTATES_full;
3971 			}
3972 
3973 			LIST_FOREACH(s, &ih->states, entry) {
3974 				if (s->timeout == PFTM_UNLINKED)
3975 					continue;
3976 
3977 				pfsync_state_export((union pfsync_state_union*)p,
3978 				    s, PFSYNC_MSG_VERSION_1301);
3979 				p++;
3980 				nr++;
3981 			}
3982 			PF_HASHROW_UNLOCK(ih);
3983 			error = copyout(pstore, out,
3984 			    sizeof(struct pfsync_state_1301) * count);
3985 			if (error)
3986 				break;
3987 			out = ps->ps_states + nr;
3988 		}
3989 DIOCGETSTATES_full:
3990 		ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3991 		free(pstore, M_TEMP);
3992 
3993 		break;
3994 	}
3995 
3996 	case DIOCGETSTATESV2: {
3997 		struct pfioc_states_v2	*ps = (struct pfioc_states_v2 *)addr;
3998 		struct pf_kstate	*s;
3999 		struct pf_state_export	*pstore, *p;
4000 		int i, nr;
4001 		size_t slice_count = 16, count;
4002 		void *out;
4003 
4004 		if (ps->ps_req_version > PF_STATE_VERSION) {
4005 			error = ENOTSUP;
4006 			break;
4007 		}
4008 
4009 		if (ps->ps_len <= 0) {
4010 			nr = uma_zone_get_cur(V_pf_state_z);
4011 			ps->ps_len = sizeof(struct pf_state_export) * nr;
4012 			break;
4013 		}
4014 
4015 		out = ps->ps_states;
4016 		pstore = mallocarray(slice_count,
4017 		    sizeof(struct pf_state_export), M_TEMP, M_WAITOK | M_ZERO);
4018 		nr = 0;
4019 
4020 		for (i = 0; i <= V_pf_hashmask; i++) {
4021 			struct pf_idhash *ih = &V_pf_idhash[i];
4022 
4023 DIOCGETSTATESV2_retry:
4024 			p = pstore;
4025 
4026 			if (LIST_EMPTY(&ih->states))
4027 				continue;
4028 
4029 			PF_HASHROW_LOCK(ih);
4030 			count = 0;
4031 			LIST_FOREACH(s, &ih->states, entry) {
4032 				if (s->timeout == PFTM_UNLINKED)
4033 					continue;
4034 				count++;
4035 			}
4036 
4037 			if (count > slice_count) {
4038 				PF_HASHROW_UNLOCK(ih);
4039 				free(pstore, M_TEMP);
4040 				slice_count = count * 2;
4041 				pstore = mallocarray(slice_count,
4042 				    sizeof(struct pf_state_export), M_TEMP,
4043 				    M_WAITOK | M_ZERO);
4044 				goto DIOCGETSTATESV2_retry;
4045 			}
4046 
4047 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
4048 				PF_HASHROW_UNLOCK(ih);
4049 				goto DIOCGETSTATESV2_full;
4050 			}
4051 
4052 			LIST_FOREACH(s, &ih->states, entry) {
4053 				if (s->timeout == PFTM_UNLINKED)
4054 					continue;
4055 
4056 				pf_state_export(p, s);
4057 				p++;
4058 				nr++;
4059 			}
4060 			PF_HASHROW_UNLOCK(ih);
4061 			error = copyout(pstore, out,
4062 			    sizeof(struct pf_state_export) * count);
4063 			if (error)
4064 				break;
4065 			out = ps->ps_states + nr;
4066 		}
4067 DIOCGETSTATESV2_full:
4068 		ps->ps_len = nr * sizeof(struct pf_state_export);
4069 		free(pstore, M_TEMP);
4070 
4071 		break;
4072 	}
4073 #endif
4074 	case DIOCGETSTATUSNV: {
4075 		error = pf_getstatus((struct pfioc_nv *)addr);
4076 		break;
4077 	}
4078 
4079 	case DIOCSETSTATUSIF: {
4080 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
4081 
4082 		if (pi->ifname[0] == 0) {
4083 			bzero(V_pf_status.ifname, IFNAMSIZ);
4084 			break;
4085 		}
4086 		PF_RULES_WLOCK();
4087 		error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
4088 		PF_RULES_WUNLOCK();
4089 		break;
4090 	}
4091 
4092 	case DIOCCLRSTATUS: {
4093 		pf_ioctl_clear_status();
4094 		break;
4095 	}
4096 
4097 	case DIOCNATLOOK: {
4098 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
4099 		struct pf_state_key	*sk;
4100 		struct pf_kstate	*state;
4101 		struct pf_state_key_cmp	 key;
4102 		int			 m = 0, direction = pnl->direction;
4103 		int			 sidx, didx;
4104 
4105 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
4106 		sidx = (direction == PF_IN) ? 1 : 0;
4107 		didx = (direction == PF_IN) ? 0 : 1;
4108 
4109 		if (!pnl->proto ||
4110 		    PF_AZERO(&pnl->saddr, pnl->af) ||
4111 		    PF_AZERO(&pnl->daddr, pnl->af) ||
4112 		    ((pnl->proto == IPPROTO_TCP ||
4113 		    pnl->proto == IPPROTO_UDP) &&
4114 		    (!pnl->dport || !pnl->sport)))
4115 			error = EINVAL;
4116 		else {
4117 			bzero(&key, sizeof(key));
4118 			key.af = pnl->af;
4119 			key.proto = pnl->proto;
4120 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
4121 			key.port[sidx] = pnl->sport;
4122 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
4123 			key.port[didx] = pnl->dport;
4124 
4125 			state = pf_find_state_all(&key, direction, &m);
4126 			if (state == NULL) {
4127 				error = ENOENT;
4128 			} else {
4129 				if (m > 1) {
4130 					PF_STATE_UNLOCK(state);
4131 					error = E2BIG;	/* more than one state */
4132 				} else {
4133 					sk = state->key[sidx];
4134 					PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
4135 					pnl->rsport = sk->port[sidx];
4136 					PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
4137 					pnl->rdport = sk->port[didx];
4138 					PF_STATE_UNLOCK(state);
4139 				}
4140 			}
4141 		}
4142 		break;
4143 	}
4144 
4145 	case DIOCSETTIMEOUT: {
4146 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
4147 
4148 		error = pf_ioctl_set_timeout(pt->timeout, pt->seconds,
4149 		    &pt->seconds);
4150 		break;
4151 	}
4152 
4153 	case DIOCGETTIMEOUT: {
4154 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
4155 
4156 		error = pf_ioctl_get_timeout(pt->timeout, &pt->seconds);
4157 		break;
4158 	}
4159 
4160 	case DIOCGETLIMIT: {
4161 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
4162 
4163 		error = pf_ioctl_get_limit(pl->index, &pl->limit);
4164 		break;
4165 	}
4166 
4167 	case DIOCSETLIMIT: {
4168 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
4169 		unsigned int old_limit;
4170 
4171 		error = pf_ioctl_set_limit(pl->index, pl->limit, &old_limit);
4172 		pl->limit = old_limit;
4173 		break;
4174 	}
4175 
4176 	case DIOCSETDEBUG: {
4177 		u_int32_t	*level = (u_int32_t *)addr;
4178 
4179 		PF_RULES_WLOCK();
4180 		V_pf_status.debug = *level;
4181 		PF_RULES_WUNLOCK();
4182 		break;
4183 	}
4184 
4185 	case DIOCCLRRULECTRS: {
4186 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
4187 		struct pf_kruleset	*ruleset = &pf_main_ruleset;
4188 		struct pf_krule		*rule;
4189 
4190 		PF_RULES_WLOCK();
4191 		TAILQ_FOREACH(rule,
4192 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
4193 			pf_counter_u64_zero(&rule->evaluations);
4194 			for (int i = 0; i < 2; i++) {
4195 				pf_counter_u64_zero(&rule->packets[i]);
4196 				pf_counter_u64_zero(&rule->bytes[i]);
4197 			}
4198 		}
4199 		PF_RULES_WUNLOCK();
4200 		break;
4201 	}
4202 
4203 	case DIOCGIFSPEEDV0:
4204 	case DIOCGIFSPEEDV1: {
4205 		struct pf_ifspeed_v1	*psp = (struct pf_ifspeed_v1 *)addr;
4206 		struct pf_ifspeed_v1	ps;
4207 		struct ifnet		*ifp;
4208 
4209 		if (psp->ifname[0] == '\0') {
4210 			error = EINVAL;
4211 			break;
4212 		}
4213 
4214 		error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ);
4215 		if (error != 0)
4216 			break;
4217 		ifp = ifunit(ps.ifname);
4218 		if (ifp != NULL) {
4219 			psp->baudrate32 =
4220 			    (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
4221 			if (cmd == DIOCGIFSPEEDV1)
4222 				psp->baudrate = ifp->if_baudrate;
4223 		} else {
4224 			error = EINVAL;
4225 		}
4226 		break;
4227 	}
4228 
4229 #ifdef ALTQ
4230 	case DIOCSTARTALTQ: {
4231 		struct pf_altq		*altq;
4232 
4233 		PF_RULES_WLOCK();
4234 		/* enable all altq interfaces on active list */
4235 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4236 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4237 				error = pf_enable_altq(altq);
4238 				if (error != 0)
4239 					break;
4240 			}
4241 		}
4242 		if (error == 0)
4243 			V_pf_altq_running = 1;
4244 		PF_RULES_WUNLOCK();
4245 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
4246 		break;
4247 	}
4248 
4249 	case DIOCSTOPALTQ: {
4250 		struct pf_altq		*altq;
4251 
4252 		PF_RULES_WLOCK();
4253 		/* disable all altq interfaces on active list */
4254 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4255 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4256 				error = pf_disable_altq(altq);
4257 				if (error != 0)
4258 					break;
4259 			}
4260 		}
4261 		if (error == 0)
4262 			V_pf_altq_running = 0;
4263 		PF_RULES_WUNLOCK();
4264 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
4265 		break;
4266 	}
4267 
4268 	case DIOCADDALTQV0:
4269 	case DIOCADDALTQV1: {
4270 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4271 		struct pf_altq		*altq, *a;
4272 		struct ifnet		*ifp;
4273 
4274 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
4275 		error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
4276 		if (error)
4277 			break;
4278 		altq->local_flags = 0;
4279 
4280 		PF_RULES_WLOCK();
4281 		if (pa->ticket != V_ticket_altqs_inactive) {
4282 			PF_RULES_WUNLOCK();
4283 			free(altq, M_PFALTQ);
4284 			error = EBUSY;
4285 			break;
4286 		}
4287 
4288 		/*
4289 		 * if this is for a queue, find the discipline and
4290 		 * copy the necessary fields
4291 		 */
4292 		if (altq->qname[0] != 0) {
4293 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
4294 				PF_RULES_WUNLOCK();
4295 				error = EBUSY;
4296 				free(altq, M_PFALTQ);
4297 				break;
4298 			}
4299 			altq->altq_disc = NULL;
4300 			TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
4301 				if (strncmp(a->ifname, altq->ifname,
4302 				    IFNAMSIZ) == 0) {
4303 					altq->altq_disc = a->altq_disc;
4304 					break;
4305 				}
4306 			}
4307 		}
4308 
4309 		if ((ifp = ifunit(altq->ifname)) == NULL)
4310 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
4311 		else
4312 			error = altq_add(ifp, altq);
4313 
4314 		if (error) {
4315 			PF_RULES_WUNLOCK();
4316 			free(altq, M_PFALTQ);
4317 			break;
4318 		}
4319 
4320 		if (altq->qname[0] != 0)
4321 			TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
4322 		else
4323 			TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
4324 		/* version error check done on import above */
4325 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4326 		PF_RULES_WUNLOCK();
4327 		break;
4328 	}
4329 
4330 	case DIOCGETALTQSV0:
4331 	case DIOCGETALTQSV1: {
4332 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4333 		struct pf_altq		*altq;
4334 
4335 		PF_RULES_RLOCK();
4336 		pa->nr = 0;
4337 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
4338 			pa->nr++;
4339 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
4340 			pa->nr++;
4341 		pa->ticket = V_ticket_altqs_active;
4342 		PF_RULES_RUNLOCK();
4343 		break;
4344 	}
4345 
4346 	case DIOCGETALTQV0:
4347 	case DIOCGETALTQV1: {
4348 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4349 		struct pf_altq		*altq;
4350 
4351 		PF_RULES_RLOCK();
4352 		if (pa->ticket != V_ticket_altqs_active) {
4353 			PF_RULES_RUNLOCK();
4354 			error = EBUSY;
4355 			break;
4356 		}
4357 		altq = pf_altq_get_nth_active(pa->nr);
4358 		if (altq == NULL) {
4359 			PF_RULES_RUNLOCK();
4360 			error = EBUSY;
4361 			break;
4362 		}
4363 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4364 		PF_RULES_RUNLOCK();
4365 		break;
4366 	}
4367 
4368 	case DIOCCHANGEALTQV0:
4369 	case DIOCCHANGEALTQV1:
4370 		/* CHANGEALTQ not supported yet! */
4371 		error = ENODEV;
4372 		break;
4373 
4374 	case DIOCGETQSTATSV0:
4375 	case DIOCGETQSTATSV1: {
4376 		struct pfioc_qstats_v1	*pq = (struct pfioc_qstats_v1 *)addr;
4377 		struct pf_altq		*altq;
4378 		int			 nbytes;
4379 		u_int32_t		 version;
4380 
4381 		PF_RULES_RLOCK();
4382 		if (pq->ticket != V_ticket_altqs_active) {
4383 			PF_RULES_RUNLOCK();
4384 			error = EBUSY;
4385 			break;
4386 		}
4387 		nbytes = pq->nbytes;
4388 		altq = pf_altq_get_nth_active(pq->nr);
4389 		if (altq == NULL) {
4390 			PF_RULES_RUNLOCK();
4391 			error = EBUSY;
4392 			break;
4393 		}
4394 
4395 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
4396 			PF_RULES_RUNLOCK();
4397 			error = ENXIO;
4398 			break;
4399 		}
4400 		PF_RULES_RUNLOCK();
4401 		if (cmd == DIOCGETQSTATSV0)
4402 			version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
4403 		else
4404 			version = pq->version;
4405 		error = altq_getqstats(altq, pq->buf, &nbytes, version);
4406 		if (error == 0) {
4407 			pq->scheduler = altq->scheduler;
4408 			pq->nbytes = nbytes;
4409 		}
4410 		break;
4411 	}
4412 #endif /* ALTQ */
4413 
4414 	case DIOCBEGINADDRS: {
4415 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4416 
4417 		error = pf_ioctl_begin_addrs(&pp->ticket);
4418 		break;
4419 	}
4420 
4421 	case DIOCADDADDR: {
4422 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4423 		struct pf_nl_pooladdr npp = {};
4424 
4425 		npp.which = PF_RDR;
4426 		memcpy(&npp, pp, sizeof(*pp));
4427 		error = pf_ioctl_add_addr(&npp);
4428 		break;
4429 	}
4430 
4431 	case DIOCGETADDRS: {
4432 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4433 		struct pf_nl_pooladdr npp = {};
4434 
4435 		npp.which = PF_RDR;
4436 		memcpy(&npp, pp, sizeof(*pp));
4437 		error = pf_ioctl_get_addrs(&npp);
4438 		memcpy(pp, &npp, sizeof(*pp));
4439 
4440 		break;
4441 	}
4442 
4443 	case DIOCGETADDR: {
4444 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4445 		struct pf_nl_pooladdr npp = {};
4446 
4447 		npp.which = PF_RDR;
4448 		memcpy(&npp, pp, sizeof(*pp));
4449 		error = pf_ioctl_get_addr(&npp);
4450 		memcpy(pp, &npp, sizeof(*pp));
4451 
4452 		break;
4453 	}
4454 
4455 	case DIOCCHANGEADDR: {
4456 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
4457 		struct pf_kpool		*pool;
4458 		struct pf_kpooladdr	*oldpa = NULL, *newpa = NULL;
4459 		struct pf_kruleset	*ruleset;
4460 		struct pfi_kkif		*kif = NULL;
4461 
4462 		pca->anchor[sizeof(pca->anchor) - 1] = 0;
4463 
4464 		if (pca->action < PF_CHANGE_ADD_HEAD ||
4465 		    pca->action > PF_CHANGE_REMOVE) {
4466 			error = EINVAL;
4467 			break;
4468 		}
4469 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
4470 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
4471 		    pca->addr.addr.type != PF_ADDR_TABLE) {
4472 			error = EINVAL;
4473 			break;
4474 		}
4475 		if (pca->addr.addr.p.dyn != NULL) {
4476 			error = EINVAL;
4477 			break;
4478 		}
4479 
4480 		if (pca->action != PF_CHANGE_REMOVE) {
4481 #ifndef INET
4482 			if (pca->af == AF_INET) {
4483 				error = EAFNOSUPPORT;
4484 				break;
4485 			}
4486 #endif /* INET */
4487 #ifndef INET6
4488 			if (pca->af == AF_INET6) {
4489 				error = EAFNOSUPPORT;
4490 				break;
4491 			}
4492 #endif /* INET6 */
4493 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
4494 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
4495 			if (newpa->ifname[0])
4496 				kif = pf_kkif_create(M_WAITOK);
4497 			newpa->kif = NULL;
4498 		}
4499 #define	ERROUT(x)	ERROUT_IOCTL(DIOCCHANGEADDR_error, x)
4500 		PF_RULES_WLOCK();
4501 		ruleset = pf_find_kruleset(pca->anchor);
4502 		if (ruleset == NULL)
4503 			ERROUT(EBUSY);
4504 
4505 		pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
4506 		    pca->r_num, pca->r_last, 1, 1, PF_RDR);
4507 		if (pool == NULL)
4508 			ERROUT(EBUSY);
4509 
4510 		if (pca->action != PF_CHANGE_REMOVE) {
4511 			if (newpa->ifname[0]) {
4512 				newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
4513 				pfi_kkif_ref(newpa->kif);
4514 				kif = NULL;
4515 			}
4516 
4517 			switch (newpa->addr.type) {
4518 			case PF_ADDR_DYNIFTL:
4519 				error = pfi_dynaddr_setup(&newpa->addr,
4520 				    pca->af);
4521 				break;
4522 			case PF_ADDR_TABLE:
4523 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
4524 				    newpa->addr.v.tblname);
4525 				if (newpa->addr.p.tbl == NULL)
4526 					error = ENOMEM;
4527 				break;
4528 			}
4529 			if (error)
4530 				goto DIOCCHANGEADDR_error;
4531 		}
4532 
4533 		switch (pca->action) {
4534 		case PF_CHANGE_ADD_HEAD:
4535 			oldpa = TAILQ_FIRST(&pool->list);
4536 			break;
4537 		case PF_CHANGE_ADD_TAIL:
4538 			oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
4539 			break;
4540 		default:
4541 			oldpa = TAILQ_FIRST(&pool->list);
4542 			for (int i = 0; oldpa && i < pca->nr; i++)
4543 				oldpa = TAILQ_NEXT(oldpa, entries);
4544 
4545 			if (oldpa == NULL)
4546 				ERROUT(EINVAL);
4547 		}
4548 
4549 		if (pca->action == PF_CHANGE_REMOVE) {
4550 			TAILQ_REMOVE(&pool->list, oldpa, entries);
4551 			switch (oldpa->addr.type) {
4552 			case PF_ADDR_DYNIFTL:
4553 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
4554 				break;
4555 			case PF_ADDR_TABLE:
4556 				pfr_detach_table(oldpa->addr.p.tbl);
4557 				break;
4558 			}
4559 			if (oldpa->kif)
4560 				pfi_kkif_unref(oldpa->kif);
4561 			free(oldpa, M_PFRULE);
4562 		} else {
4563 			if (oldpa == NULL)
4564 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
4565 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
4566 			    pca->action == PF_CHANGE_ADD_BEFORE)
4567 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
4568 			else
4569 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
4570 				    newpa, entries);
4571 		}
4572 
4573 		pool->cur = TAILQ_FIRST(&pool->list);
4574 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
4575 		PF_RULES_WUNLOCK();
4576 		break;
4577 
4578 #undef ERROUT
4579 DIOCCHANGEADDR_error:
4580 		if (newpa != NULL) {
4581 			if (newpa->kif)
4582 				pfi_kkif_unref(newpa->kif);
4583 			free(newpa, M_PFRULE);
4584 		}
4585 		PF_RULES_WUNLOCK();
4586 		pf_kkif_free(kif);
4587 		break;
4588 	}
4589 
4590 	case DIOCGETRULESETS: {
4591 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
4592 
4593 		pr->path[sizeof(pr->path) - 1] = 0;
4594 
4595 		error = pf_ioctl_get_rulesets(pr);
4596 		break;
4597 	}
4598 
4599 	case DIOCGETRULESET: {
4600 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
4601 
4602 		pr->path[sizeof(pr->path) - 1] = 0;
4603 
4604 		error = pf_ioctl_get_ruleset(pr);
4605 		break;
4606 	}
4607 
4608 	case DIOCRCLRTABLES: {
4609 		struct pfioc_table *io = (struct pfioc_table *)addr;
4610 
4611 		if (io->pfrio_esize != 0) {
4612 			error = ENODEV;
4613 			break;
4614 		}
4615 		PF_RULES_WLOCK();
4616 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
4617 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4618 		PF_RULES_WUNLOCK();
4619 		break;
4620 	}
4621 
4622 	case DIOCRADDTABLES: {
4623 		struct pfioc_table *io = (struct pfioc_table *)addr;
4624 		struct pfr_table *pfrts;
4625 		size_t totlen;
4626 
4627 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4628 			error = ENODEV;
4629 			break;
4630 		}
4631 
4632 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4633 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4634 			error = ENOMEM;
4635 			break;
4636 		}
4637 
4638 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4639 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4640 		    M_TEMP, M_WAITOK);
4641 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4642 		if (error) {
4643 			free(pfrts, M_TEMP);
4644 			break;
4645 		}
4646 		PF_RULES_WLOCK();
4647 		error = pfr_add_tables(pfrts, io->pfrio_size,
4648 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4649 		PF_RULES_WUNLOCK();
4650 		free(pfrts, M_TEMP);
4651 		break;
4652 	}
4653 
4654 	case DIOCRDELTABLES: {
4655 		struct pfioc_table *io = (struct pfioc_table *)addr;
4656 		struct pfr_table *pfrts;
4657 		size_t totlen;
4658 
4659 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4660 			error = ENODEV;
4661 			break;
4662 		}
4663 
4664 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4665 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4666 			error = ENOMEM;
4667 			break;
4668 		}
4669 
4670 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4671 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4672 		    M_TEMP, M_WAITOK);
4673 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4674 		if (error) {
4675 			free(pfrts, M_TEMP);
4676 			break;
4677 		}
4678 		PF_RULES_WLOCK();
4679 		error = pfr_del_tables(pfrts, io->pfrio_size,
4680 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4681 		PF_RULES_WUNLOCK();
4682 		free(pfrts, M_TEMP);
4683 		break;
4684 	}
4685 
4686 	case DIOCRGETTABLES: {
4687 		struct pfioc_table *io = (struct pfioc_table *)addr;
4688 		struct pfr_table *pfrts;
4689 		size_t totlen;
4690 		int n;
4691 
4692 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4693 			error = ENODEV;
4694 			break;
4695 		}
4696 		PF_RULES_RLOCK();
4697 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4698 		if (n < 0) {
4699 			PF_RULES_RUNLOCK();
4700 			error = EINVAL;
4701 			break;
4702 		}
4703 		io->pfrio_size = min(io->pfrio_size, n);
4704 
4705 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4706 
4707 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4708 		    M_TEMP, M_NOWAIT | M_ZERO);
4709 		if (pfrts == NULL) {
4710 			error = ENOMEM;
4711 			PF_RULES_RUNLOCK();
4712 			break;
4713 		}
4714 		error = pfr_get_tables(&io->pfrio_table, pfrts,
4715 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4716 		PF_RULES_RUNLOCK();
4717 		if (error == 0)
4718 			error = copyout(pfrts, io->pfrio_buffer, totlen);
4719 		free(pfrts, M_TEMP);
4720 		break;
4721 	}
4722 
4723 	case DIOCRGETTSTATS: {
4724 		struct pfioc_table *io = (struct pfioc_table *)addr;
4725 		struct pfr_tstats *pfrtstats;
4726 		size_t totlen;
4727 		int n;
4728 
4729 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
4730 			error = ENODEV;
4731 			break;
4732 		}
4733 		PF_TABLE_STATS_LOCK();
4734 		PF_RULES_RLOCK();
4735 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4736 		if (n < 0) {
4737 			PF_RULES_RUNLOCK();
4738 			PF_TABLE_STATS_UNLOCK();
4739 			error = EINVAL;
4740 			break;
4741 		}
4742 		io->pfrio_size = min(io->pfrio_size, n);
4743 
4744 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
4745 		pfrtstats = mallocarray(io->pfrio_size,
4746 		    sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT | M_ZERO);
4747 		if (pfrtstats == NULL) {
4748 			error = ENOMEM;
4749 			PF_RULES_RUNLOCK();
4750 			PF_TABLE_STATS_UNLOCK();
4751 			break;
4752 		}
4753 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
4754 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4755 		PF_RULES_RUNLOCK();
4756 		PF_TABLE_STATS_UNLOCK();
4757 		if (error == 0)
4758 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
4759 		free(pfrtstats, M_TEMP);
4760 		break;
4761 	}
4762 
4763 	case DIOCRCLRTSTATS: {
4764 		struct pfioc_table *io = (struct pfioc_table *)addr;
4765 		struct pfr_table *pfrts;
4766 		size_t totlen;
4767 
4768 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4769 			error = ENODEV;
4770 			break;
4771 		}
4772 
4773 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4774 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4775 			/* We used to count tables and use the minimum required
4776 			 * size, so we didn't fail on overly large requests.
4777 			 * Keep doing so. */
4778 			io->pfrio_size = pf_ioctl_maxcount;
4779 			break;
4780 		}
4781 
4782 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4783 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4784 		    M_TEMP, M_WAITOK);
4785 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4786 		if (error) {
4787 			free(pfrts, M_TEMP);
4788 			break;
4789 		}
4790 
4791 		PF_TABLE_STATS_LOCK();
4792 		PF_RULES_RLOCK();
4793 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
4794 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4795 		PF_RULES_RUNLOCK();
4796 		PF_TABLE_STATS_UNLOCK();
4797 		free(pfrts, M_TEMP);
4798 		break;
4799 	}
4800 
4801 	case DIOCRSETTFLAGS: {
4802 		struct pfioc_table *io = (struct pfioc_table *)addr;
4803 		struct pfr_table *pfrts;
4804 		size_t totlen;
4805 		int n;
4806 
4807 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4808 			error = ENODEV;
4809 			break;
4810 		}
4811 
4812 		PF_RULES_RLOCK();
4813 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4814 		if (n < 0) {
4815 			PF_RULES_RUNLOCK();
4816 			error = EINVAL;
4817 			break;
4818 		}
4819 
4820 		io->pfrio_size = min(io->pfrio_size, n);
4821 		PF_RULES_RUNLOCK();
4822 
4823 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4824 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4825 		    M_TEMP, M_WAITOK);
4826 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4827 		if (error) {
4828 			free(pfrts, M_TEMP);
4829 			break;
4830 		}
4831 		PF_RULES_WLOCK();
4832 		error = pfr_set_tflags(pfrts, io->pfrio_size,
4833 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
4834 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4835 		PF_RULES_WUNLOCK();
4836 		free(pfrts, M_TEMP);
4837 		break;
4838 	}
4839 
4840 	case DIOCRCLRADDRS: {
4841 		struct pfioc_table *io = (struct pfioc_table *)addr;
4842 
4843 		if (io->pfrio_esize != 0) {
4844 			error = ENODEV;
4845 			break;
4846 		}
4847 		PF_RULES_WLOCK();
4848 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
4849 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4850 		PF_RULES_WUNLOCK();
4851 		break;
4852 	}
4853 
4854 	case DIOCRADDADDRS: {
4855 		struct pfioc_table *io = (struct pfioc_table *)addr;
4856 		struct pfr_addr *pfras;
4857 		size_t totlen;
4858 
4859 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4860 			error = ENODEV;
4861 			break;
4862 		}
4863 		if (io->pfrio_size < 0 ||
4864 		    io->pfrio_size > pf_ioctl_maxcount ||
4865 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4866 			error = EINVAL;
4867 			break;
4868 		}
4869 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4870 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4871 		    M_TEMP, M_WAITOK);
4872 		error = copyin(io->pfrio_buffer, pfras, totlen);
4873 		if (error) {
4874 			free(pfras, M_TEMP);
4875 			break;
4876 		}
4877 		PF_RULES_WLOCK();
4878 		error = pfr_add_addrs(&io->pfrio_table, pfras,
4879 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
4880 		    PFR_FLAG_USERIOCTL);
4881 		PF_RULES_WUNLOCK();
4882 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4883 			error = copyout(pfras, io->pfrio_buffer, totlen);
4884 		free(pfras, M_TEMP);
4885 		break;
4886 	}
4887 
4888 	case DIOCRDELADDRS: {
4889 		struct pfioc_table *io = (struct pfioc_table *)addr;
4890 		struct pfr_addr *pfras;
4891 		size_t totlen;
4892 
4893 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4894 			error = ENODEV;
4895 			break;
4896 		}
4897 		if (io->pfrio_size < 0 ||
4898 		    io->pfrio_size > pf_ioctl_maxcount ||
4899 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4900 			error = EINVAL;
4901 			break;
4902 		}
4903 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4904 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4905 		    M_TEMP, M_WAITOK);
4906 		error = copyin(io->pfrio_buffer, pfras, totlen);
4907 		if (error) {
4908 			free(pfras, M_TEMP);
4909 			break;
4910 		}
4911 		PF_RULES_WLOCK();
4912 		error = pfr_del_addrs(&io->pfrio_table, pfras,
4913 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
4914 		    PFR_FLAG_USERIOCTL);
4915 		PF_RULES_WUNLOCK();
4916 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4917 			error = copyout(pfras, io->pfrio_buffer, totlen);
4918 		free(pfras, M_TEMP);
4919 		break;
4920 	}
4921 
4922 	case DIOCRSETADDRS: {
4923 		struct pfioc_table *io = (struct pfioc_table *)addr;
4924 		struct pfr_addr *pfras;
4925 		size_t totlen, count;
4926 
4927 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4928 			error = ENODEV;
4929 			break;
4930 		}
4931 		if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
4932 			error = EINVAL;
4933 			break;
4934 		}
4935 		count = max(io->pfrio_size, io->pfrio_size2);
4936 		if (count > pf_ioctl_maxcount ||
4937 		    WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
4938 			error = EINVAL;
4939 			break;
4940 		}
4941 		totlen = count * sizeof(struct pfr_addr);
4942 		pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
4943 		    M_WAITOK);
4944 		error = copyin(io->pfrio_buffer, pfras, totlen);
4945 		if (error) {
4946 			free(pfras, M_TEMP);
4947 			break;
4948 		}
4949 		PF_RULES_WLOCK();
4950 		error = pfr_set_addrs(&io->pfrio_table, pfras,
4951 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
4952 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
4953 		    PFR_FLAG_USERIOCTL, 0);
4954 		PF_RULES_WUNLOCK();
4955 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4956 			error = copyout(pfras, io->pfrio_buffer, totlen);
4957 		free(pfras, M_TEMP);
4958 		break;
4959 	}
4960 
4961 	case DIOCRGETADDRS: {
4962 		struct pfioc_table *io = (struct pfioc_table *)addr;
4963 		struct pfr_addr *pfras;
4964 		size_t totlen;
4965 
4966 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4967 			error = ENODEV;
4968 			break;
4969 		}
4970 		if (io->pfrio_size < 0 ||
4971 		    io->pfrio_size > pf_ioctl_maxcount ||
4972 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4973 			error = EINVAL;
4974 			break;
4975 		}
4976 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4977 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4978 		    M_TEMP, M_WAITOK | M_ZERO);
4979 		PF_RULES_RLOCK();
4980 		error = pfr_get_addrs(&io->pfrio_table, pfras,
4981 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4982 		PF_RULES_RUNLOCK();
4983 		if (error == 0)
4984 			error = copyout(pfras, io->pfrio_buffer, totlen);
4985 		free(pfras, M_TEMP);
4986 		break;
4987 	}
4988 
4989 	case DIOCRGETASTATS: {
4990 		struct pfioc_table *io = (struct pfioc_table *)addr;
4991 		struct pfr_astats *pfrastats;
4992 		size_t totlen;
4993 
4994 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
4995 			error = ENODEV;
4996 			break;
4997 		}
4998 		if (io->pfrio_size < 0 ||
4999 		    io->pfrio_size > pf_ioctl_maxcount ||
5000 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
5001 			error = EINVAL;
5002 			break;
5003 		}
5004 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
5005 		pfrastats = mallocarray(io->pfrio_size,
5006 		    sizeof(struct pfr_astats), M_TEMP, M_WAITOK | M_ZERO);
5007 		PF_RULES_RLOCK();
5008 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
5009 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5010 		PF_RULES_RUNLOCK();
5011 		if (error == 0)
5012 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
5013 		free(pfrastats, M_TEMP);
5014 		break;
5015 	}
5016 
5017 	case DIOCRCLRASTATS: {
5018 		struct pfioc_table *io = (struct pfioc_table *)addr;
5019 		struct pfr_addr *pfras;
5020 		size_t totlen;
5021 
5022 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5023 			error = ENODEV;
5024 			break;
5025 		}
5026 		if (io->pfrio_size < 0 ||
5027 		    io->pfrio_size > pf_ioctl_maxcount ||
5028 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5029 			error = EINVAL;
5030 			break;
5031 		}
5032 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
5033 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5034 		    M_TEMP, M_WAITOK);
5035 		error = copyin(io->pfrio_buffer, pfras, totlen);
5036 		if (error) {
5037 			free(pfras, M_TEMP);
5038 			break;
5039 		}
5040 		PF_RULES_WLOCK();
5041 		error = pfr_clr_astats(&io->pfrio_table, pfras,
5042 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
5043 		    PFR_FLAG_USERIOCTL);
5044 		PF_RULES_WUNLOCK();
5045 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
5046 			error = copyout(pfras, io->pfrio_buffer, totlen);
5047 		free(pfras, M_TEMP);
5048 		break;
5049 	}
5050 
5051 	case DIOCRTSTADDRS: {
5052 		struct pfioc_table *io = (struct pfioc_table *)addr;
5053 		struct pfr_addr *pfras;
5054 		size_t totlen;
5055 
5056 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5057 			error = ENODEV;
5058 			break;
5059 		}
5060 		if (io->pfrio_size < 0 ||
5061 		    io->pfrio_size > pf_ioctl_maxcount ||
5062 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5063 			error = EINVAL;
5064 			break;
5065 		}
5066 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
5067 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5068 		    M_TEMP, M_WAITOK);
5069 		error = copyin(io->pfrio_buffer, pfras, totlen);
5070 		if (error) {
5071 			free(pfras, M_TEMP);
5072 			break;
5073 		}
5074 		PF_RULES_RLOCK();
5075 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
5076 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
5077 		    PFR_FLAG_USERIOCTL);
5078 		PF_RULES_RUNLOCK();
5079 		if (error == 0)
5080 			error = copyout(pfras, io->pfrio_buffer, totlen);
5081 		free(pfras, M_TEMP);
5082 		break;
5083 	}
5084 
5085 	case DIOCRINADEFINE: {
5086 		struct pfioc_table *io = (struct pfioc_table *)addr;
5087 		struct pfr_addr *pfras;
5088 		size_t totlen;
5089 
5090 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5091 			error = ENODEV;
5092 			break;
5093 		}
5094 		if (io->pfrio_size < 0 ||
5095 		    io->pfrio_size > pf_ioctl_maxcount ||
5096 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5097 			error = EINVAL;
5098 			break;
5099 		}
5100 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
5101 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5102 		    M_TEMP, M_WAITOK);
5103 		error = copyin(io->pfrio_buffer, pfras, totlen);
5104 		if (error) {
5105 			free(pfras, M_TEMP);
5106 			break;
5107 		}
5108 		PF_RULES_WLOCK();
5109 		error = pfr_ina_define(&io->pfrio_table, pfras,
5110 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
5111 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5112 		PF_RULES_WUNLOCK();
5113 		free(pfras, M_TEMP);
5114 		break;
5115 	}
5116 
5117 	case DIOCOSFPADD: {
5118 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5119 		PF_RULES_WLOCK();
5120 		error = pf_osfp_add(io);
5121 		PF_RULES_WUNLOCK();
5122 		break;
5123 	}
5124 
5125 	case DIOCOSFPGET: {
5126 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5127 		PF_RULES_RLOCK();
5128 		error = pf_osfp_get(io);
5129 		PF_RULES_RUNLOCK();
5130 		break;
5131 	}
5132 
5133 	case DIOCXBEGIN: {
5134 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5135 		struct pfioc_trans_e	*ioes, *ioe;
5136 		size_t			 totlen;
5137 		int			 i;
5138 
5139 		if (io->esize != sizeof(*ioe)) {
5140 			error = ENODEV;
5141 			break;
5142 		}
5143 		if (io->size < 0 ||
5144 		    io->size > pf_ioctl_maxcount ||
5145 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5146 			error = EINVAL;
5147 			break;
5148 		}
5149 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5150 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5151 		    M_TEMP, M_WAITOK);
5152 		error = copyin(io->array, ioes, totlen);
5153 		if (error) {
5154 			free(ioes, M_TEMP);
5155 			break;
5156 		}
5157 		/* Ensure there's no more ethernet rules to clean up. */
5158 		NET_EPOCH_DRAIN_CALLBACKS();
5159 		PF_RULES_WLOCK();
5160 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5161 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5162 			switch (ioe->rs_num) {
5163 			case PF_RULESET_ETH:
5164 				if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) {
5165 					PF_RULES_WUNLOCK();
5166 					free(ioes, M_TEMP);
5167 					goto fail;
5168 				}
5169 				break;
5170 #ifdef ALTQ
5171 			case PF_RULESET_ALTQ:
5172 				if (ioe->anchor[0]) {
5173 					PF_RULES_WUNLOCK();
5174 					free(ioes, M_TEMP);
5175 					error = EINVAL;
5176 					goto fail;
5177 				}
5178 				if ((error = pf_begin_altq(&ioe->ticket))) {
5179 					PF_RULES_WUNLOCK();
5180 					free(ioes, M_TEMP);
5181 					goto fail;
5182 				}
5183 				break;
5184 #endif /* ALTQ */
5185 			case PF_RULESET_TABLE:
5186 			    {
5187 				struct pfr_table table;
5188 
5189 				bzero(&table, sizeof(table));
5190 				strlcpy(table.pfrt_anchor, ioe->anchor,
5191 				    sizeof(table.pfrt_anchor));
5192 				if ((error = pfr_ina_begin(&table,
5193 				    &ioe->ticket, NULL, 0))) {
5194 					PF_RULES_WUNLOCK();
5195 					free(ioes, M_TEMP);
5196 					goto fail;
5197 				}
5198 				break;
5199 			    }
5200 			default:
5201 				if ((error = pf_begin_rules(&ioe->ticket,
5202 				    ioe->rs_num, ioe->anchor))) {
5203 					PF_RULES_WUNLOCK();
5204 					free(ioes, M_TEMP);
5205 					goto fail;
5206 				}
5207 				break;
5208 			}
5209 		}
5210 		PF_RULES_WUNLOCK();
5211 		error = copyout(ioes, io->array, totlen);
5212 		free(ioes, M_TEMP);
5213 		break;
5214 	}
5215 
5216 	case DIOCXROLLBACK: {
5217 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5218 		struct pfioc_trans_e	*ioe, *ioes;
5219 		size_t			 totlen;
5220 		int			 i;
5221 
5222 		if (io->esize != sizeof(*ioe)) {
5223 			error = ENODEV;
5224 			break;
5225 		}
5226 		if (io->size < 0 ||
5227 		    io->size > pf_ioctl_maxcount ||
5228 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5229 			error = EINVAL;
5230 			break;
5231 		}
5232 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5233 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5234 		    M_TEMP, M_WAITOK);
5235 		error = copyin(io->array, ioes, totlen);
5236 		if (error) {
5237 			free(ioes, M_TEMP);
5238 			break;
5239 		}
5240 		PF_RULES_WLOCK();
5241 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5242 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5243 			switch (ioe->rs_num) {
5244 			case PF_RULESET_ETH:
5245 				if ((error = pf_rollback_eth(ioe->ticket,
5246 				    ioe->anchor))) {
5247 					PF_RULES_WUNLOCK();
5248 					free(ioes, M_TEMP);
5249 					goto fail; /* really bad */
5250 				}
5251 				break;
5252 #ifdef ALTQ
5253 			case PF_RULESET_ALTQ:
5254 				if (ioe->anchor[0]) {
5255 					PF_RULES_WUNLOCK();
5256 					free(ioes, M_TEMP);
5257 					error = EINVAL;
5258 					goto fail;
5259 				}
5260 				if ((error = pf_rollback_altq(ioe->ticket))) {
5261 					PF_RULES_WUNLOCK();
5262 					free(ioes, M_TEMP);
5263 					goto fail; /* really bad */
5264 				}
5265 				break;
5266 #endif /* ALTQ */
5267 			case PF_RULESET_TABLE:
5268 			    {
5269 				struct pfr_table table;
5270 
5271 				bzero(&table, sizeof(table));
5272 				strlcpy(table.pfrt_anchor, ioe->anchor,
5273 				    sizeof(table.pfrt_anchor));
5274 				if ((error = pfr_ina_rollback(&table,
5275 				    ioe->ticket, NULL, 0))) {
5276 					PF_RULES_WUNLOCK();
5277 					free(ioes, M_TEMP);
5278 					goto fail; /* really bad */
5279 				}
5280 				break;
5281 			    }
5282 			default:
5283 				if ((error = pf_rollback_rules(ioe->ticket,
5284 				    ioe->rs_num, ioe->anchor))) {
5285 					PF_RULES_WUNLOCK();
5286 					free(ioes, M_TEMP);
5287 					goto fail; /* really bad */
5288 				}
5289 				break;
5290 			}
5291 		}
5292 		PF_RULES_WUNLOCK();
5293 		free(ioes, M_TEMP);
5294 		break;
5295 	}
5296 
5297 	case DIOCXCOMMIT: {
5298 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5299 		struct pfioc_trans_e	*ioe, *ioes;
5300 		struct pf_kruleset	*rs;
5301 		struct pf_keth_ruleset	*ers;
5302 		size_t			 totlen;
5303 		int			 i;
5304 
5305 		if (io->esize != sizeof(*ioe)) {
5306 			error = ENODEV;
5307 			break;
5308 		}
5309 
5310 		if (io->size < 0 ||
5311 		    io->size > pf_ioctl_maxcount ||
5312 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5313 			error = EINVAL;
5314 			break;
5315 		}
5316 
5317 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5318 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5319 		    M_TEMP, M_WAITOK);
5320 		error = copyin(io->array, ioes, totlen);
5321 		if (error) {
5322 			free(ioes, M_TEMP);
5323 			break;
5324 		}
5325 		PF_RULES_WLOCK();
5326 		/* First makes sure everything will succeed. */
5327 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5328 			ioe->anchor[sizeof(ioe->anchor) - 1] = 0;
5329 			switch (ioe->rs_num) {
5330 			case PF_RULESET_ETH:
5331 				ers = pf_find_keth_ruleset(ioe->anchor);
5332 				if (ers == NULL || ioe->ticket == 0 ||
5333 				    ioe->ticket != ers->inactive.ticket) {
5334 					PF_RULES_WUNLOCK();
5335 					free(ioes, M_TEMP);
5336 					error = EINVAL;
5337 					goto fail;
5338 				}
5339 				break;
5340 #ifdef ALTQ
5341 			case PF_RULESET_ALTQ:
5342 				if (ioe->anchor[0]) {
5343 					PF_RULES_WUNLOCK();
5344 					free(ioes, M_TEMP);
5345 					error = EINVAL;
5346 					goto fail;
5347 				}
5348 				if (!V_altqs_inactive_open || ioe->ticket !=
5349 				    V_ticket_altqs_inactive) {
5350 					PF_RULES_WUNLOCK();
5351 					free(ioes, M_TEMP);
5352 					error = EBUSY;
5353 					goto fail;
5354 				}
5355 				break;
5356 #endif /* ALTQ */
5357 			case PF_RULESET_TABLE:
5358 				rs = pf_find_kruleset(ioe->anchor);
5359 				if (rs == NULL || !rs->topen || ioe->ticket !=
5360 				    rs->tticket) {
5361 					PF_RULES_WUNLOCK();
5362 					free(ioes, M_TEMP);
5363 					error = EBUSY;
5364 					goto fail;
5365 				}
5366 				break;
5367 			default:
5368 				if (ioe->rs_num < 0 || ioe->rs_num >=
5369 				    PF_RULESET_MAX) {
5370 					PF_RULES_WUNLOCK();
5371 					free(ioes, M_TEMP);
5372 					error = EINVAL;
5373 					goto fail;
5374 				}
5375 				rs = pf_find_kruleset(ioe->anchor);
5376 				if (rs == NULL ||
5377 				    !rs->rules[ioe->rs_num].inactive.open ||
5378 				    rs->rules[ioe->rs_num].inactive.ticket !=
5379 				    ioe->ticket) {
5380 					PF_RULES_WUNLOCK();
5381 					free(ioes, M_TEMP);
5382 					error = EBUSY;
5383 					goto fail;
5384 				}
5385 				break;
5386 			}
5387 		}
5388 		/* Now do the commit - no errors should happen here. */
5389 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5390 			switch (ioe->rs_num) {
5391 			case PF_RULESET_ETH:
5392 				if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) {
5393 					PF_RULES_WUNLOCK();
5394 					free(ioes, M_TEMP);
5395 					goto fail; /* really bad */
5396 				}
5397 				break;
5398 #ifdef ALTQ
5399 			case PF_RULESET_ALTQ:
5400 				if ((error = pf_commit_altq(ioe->ticket))) {
5401 					PF_RULES_WUNLOCK();
5402 					free(ioes, M_TEMP);
5403 					goto fail; /* really bad */
5404 				}
5405 				break;
5406 #endif /* ALTQ */
5407 			case PF_RULESET_TABLE:
5408 			    {
5409 				struct pfr_table table;
5410 
5411 				bzero(&table, sizeof(table));
5412 				(void)strlcpy(table.pfrt_anchor, ioe->anchor,
5413 				    sizeof(table.pfrt_anchor));
5414 				if ((error = pfr_ina_commit(&table,
5415 				    ioe->ticket, NULL, NULL, 0))) {
5416 					PF_RULES_WUNLOCK();
5417 					free(ioes, M_TEMP);
5418 					goto fail; /* really bad */
5419 				}
5420 				break;
5421 			    }
5422 			default:
5423 				if ((error = pf_commit_rules(ioe->ticket,
5424 				    ioe->rs_num, ioe->anchor))) {
5425 					PF_RULES_WUNLOCK();
5426 					free(ioes, M_TEMP);
5427 					goto fail; /* really bad */
5428 				}
5429 				break;
5430 			}
5431 		}
5432 		PF_RULES_WUNLOCK();
5433 
5434 		/* Only hook into EtherNet taffic if we've got rules for it. */
5435 		if (! TAILQ_EMPTY(V_pf_keth->active.rules))
5436 			hook_pf_eth();
5437 		else
5438 			dehook_pf_eth();
5439 
5440 		free(ioes, M_TEMP);
5441 		break;
5442 	}
5443 
5444 	case DIOCGETSRCNODES: {
5445 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
5446 		struct pf_srchash	*sh;
5447 		struct pf_ksrc_node	*n;
5448 		struct pf_src_node	*p, *pstore;
5449 		uint32_t		 i, nr = 0;
5450 
5451 		for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5452 				i++, sh++) {
5453 			PF_HASHROW_LOCK(sh);
5454 			LIST_FOREACH(n, &sh->nodes, entry)
5455 				nr++;
5456 			PF_HASHROW_UNLOCK(sh);
5457 		}
5458 
5459 		psn->psn_len = min(psn->psn_len,
5460 		    sizeof(struct pf_src_node) * nr);
5461 
5462 		if (psn->psn_len == 0) {
5463 			psn->psn_len = sizeof(struct pf_src_node) * nr;
5464 			break;
5465 		}
5466 
5467 		nr = 0;
5468 
5469 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
5470 		for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5471 		    i++, sh++) {
5472 		    PF_HASHROW_LOCK(sh);
5473 		    LIST_FOREACH(n, &sh->nodes, entry) {
5474 
5475 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
5476 				break;
5477 
5478 			pf_src_node_copy(n, p);
5479 
5480 			p++;
5481 			nr++;
5482 		    }
5483 		    PF_HASHROW_UNLOCK(sh);
5484 		}
5485 		error = copyout(pstore, psn->psn_src_nodes,
5486 		    sizeof(struct pf_src_node) * nr);
5487 		if (error) {
5488 			free(pstore, M_TEMP);
5489 			break;
5490 		}
5491 		psn->psn_len = sizeof(struct pf_src_node) * nr;
5492 		free(pstore, M_TEMP);
5493 		break;
5494 	}
5495 
5496 	case DIOCCLRSRCNODES: {
5497 		pf_kill_srcnodes(NULL);
5498 		break;
5499 	}
5500 
5501 	case DIOCKILLSRCNODES:
5502 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
5503 		break;
5504 
5505 #ifdef COMPAT_FREEBSD13
5506 	case DIOCKEEPCOUNTERS_FREEBSD13:
5507 #endif
5508 	case DIOCKEEPCOUNTERS:
5509 		error = pf_keepcounters((struct pfioc_nv *)addr);
5510 		break;
5511 
5512 	case DIOCGETSYNCOOKIES:
5513 		error = pf_get_syncookies((struct pfioc_nv *)addr);
5514 		break;
5515 
5516 	case DIOCSETSYNCOOKIES:
5517 		error = pf_set_syncookies((struct pfioc_nv *)addr);
5518 		break;
5519 
5520 	case DIOCSETHOSTID: {
5521 		u_int32_t	*hostid = (u_int32_t *)addr;
5522 
5523 		PF_RULES_WLOCK();
5524 		if (*hostid == 0)
5525 			V_pf_status.hostid = arc4random();
5526 		else
5527 			V_pf_status.hostid = *hostid;
5528 		PF_RULES_WUNLOCK();
5529 		break;
5530 	}
5531 
5532 	case DIOCOSFPFLUSH:
5533 		PF_RULES_WLOCK();
5534 		pf_osfp_flush();
5535 		PF_RULES_WUNLOCK();
5536 		break;
5537 
5538 	case DIOCIGETIFACES: {
5539 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5540 		struct pfi_kif *ifstore;
5541 		size_t bufsiz;
5542 
5543 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
5544 			error = ENODEV;
5545 			break;
5546 		}
5547 
5548 		if (io->pfiio_size < 0 ||
5549 		    io->pfiio_size > pf_ioctl_maxcount ||
5550 		    WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
5551 			error = EINVAL;
5552 			break;
5553 		}
5554 
5555 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5556 
5557 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
5558 		ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
5559 		    M_TEMP, M_WAITOK | M_ZERO);
5560 
5561 		PF_RULES_RLOCK();
5562 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
5563 		PF_RULES_RUNLOCK();
5564 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
5565 		free(ifstore, M_TEMP);
5566 		break;
5567 	}
5568 
5569 	case DIOCSETIFFLAG: {
5570 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5571 
5572 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5573 
5574 		PF_RULES_WLOCK();
5575 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
5576 		PF_RULES_WUNLOCK();
5577 		break;
5578 	}
5579 
5580 	case DIOCCLRIFFLAG: {
5581 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5582 
5583 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5584 
5585 		PF_RULES_WLOCK();
5586 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
5587 		PF_RULES_WUNLOCK();
5588 		break;
5589 	}
5590 
5591 	case DIOCSETREASS: {
5592 		u_int32_t	*reass = (u_int32_t *)addr;
5593 
5594 		V_pf_status.reass = *reass & (PF_REASS_ENABLED|PF_REASS_NODF);
5595 		/* Removal of DF flag without reassembly enabled is not a
5596 		 * valid combination. Disable reassembly in such case. */
5597 		if (!(V_pf_status.reass & PF_REASS_ENABLED))
5598 			V_pf_status.reass = 0;
5599 		break;
5600 	}
5601 
5602 	default:
5603 		error = ENODEV;
5604 		break;
5605 	}
5606 fail:
5607 	CURVNET_RESTORE();
5608 
5609 #undef ERROUT_IOCTL
5610 
5611 	return (error);
5612 }
5613 
5614 void
5615 pfsync_state_export(union pfsync_state_union *sp, struct pf_kstate *st, int msg_version)
5616 {
5617 	bzero(sp, sizeof(union pfsync_state_union));
5618 
5619 	/* copy from state key */
5620 	sp->pfs_1301.key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5621 	sp->pfs_1301.key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5622 	sp->pfs_1301.key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5623 	sp->pfs_1301.key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5624 	sp->pfs_1301.key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5625 	sp->pfs_1301.key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5626 	sp->pfs_1301.key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5627 	sp->pfs_1301.key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5628 	sp->pfs_1301.proto = st->key[PF_SK_WIRE]->proto;
5629 	sp->pfs_1301.af = st->key[PF_SK_WIRE]->af;
5630 
5631 	/* copy from state */
5632 	strlcpy(sp->pfs_1301.ifname, st->kif->pfik_name, sizeof(sp->pfs_1301.ifname));
5633 	bcopy(&st->act.rt_addr, &sp->pfs_1301.rt_addr, sizeof(sp->pfs_1301.rt_addr));
5634 	sp->pfs_1301.creation = htonl(time_uptime - (st->creation / 1000));
5635 	sp->pfs_1301.expire = pf_state_expires(st);
5636 	if (sp->pfs_1301.expire <= time_uptime)
5637 		sp->pfs_1301.expire = htonl(0);
5638 	else
5639 		sp->pfs_1301.expire = htonl(sp->pfs_1301.expire - time_uptime);
5640 
5641 	sp->pfs_1301.direction = st->direction;
5642 	sp->pfs_1301.log = st->act.log;
5643 	sp->pfs_1301.timeout = st->timeout;
5644 
5645 	switch (msg_version) {
5646 		case PFSYNC_MSG_VERSION_1301:
5647 			sp->pfs_1301.state_flags = st->state_flags;
5648 			break;
5649 		case PFSYNC_MSG_VERSION_1400:
5650 			sp->pfs_1400.state_flags = htons(st->state_flags);
5651 			sp->pfs_1400.qid = htons(st->act.qid);
5652 			sp->pfs_1400.pqid = htons(st->act.pqid);
5653 			sp->pfs_1400.dnpipe = htons(st->act.dnpipe);
5654 			sp->pfs_1400.dnrpipe = htons(st->act.dnrpipe);
5655 			sp->pfs_1400.rtableid = htonl(st->act.rtableid);
5656 			sp->pfs_1400.min_ttl = st->act.min_ttl;
5657 			sp->pfs_1400.set_tos = st->act.set_tos;
5658 			sp->pfs_1400.max_mss = htons(st->act.max_mss);
5659 			sp->pfs_1400.set_prio[0] = st->act.set_prio[0];
5660 			sp->pfs_1400.set_prio[1] = st->act.set_prio[1];
5661 			sp->pfs_1400.rt = st->act.rt;
5662 			if (st->act.rt_kif)
5663 				strlcpy(sp->pfs_1400.rt_ifname,
5664 				    st->act.rt_kif->pfik_name,
5665 				    sizeof(sp->pfs_1400.rt_ifname));
5666 			break;
5667 		default:
5668 			panic("%s: Unsupported pfsync_msg_version %d",
5669 			    __func__, msg_version);
5670 	}
5671 
5672 	if (st->src_node)
5673 		sp->pfs_1301.sync_flags |= PFSYNC_FLAG_SRCNODE;
5674 	if (st->nat_src_node)
5675 		sp->pfs_1301.sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5676 
5677 	sp->pfs_1301.id = st->id;
5678 	sp->pfs_1301.creatorid = st->creatorid;
5679 	pf_state_peer_hton(&st->src, &sp->pfs_1301.src);
5680 	pf_state_peer_hton(&st->dst, &sp->pfs_1301.dst);
5681 
5682 	if (st->rule == NULL)
5683 		sp->pfs_1301.rule = htonl(-1);
5684 	else
5685 		sp->pfs_1301.rule = htonl(st->rule->nr);
5686 	if (st->anchor == NULL)
5687 		sp->pfs_1301.anchor = htonl(-1);
5688 	else
5689 		sp->pfs_1301.anchor = htonl(st->anchor->nr);
5690 	if (st->nat_rule == NULL)
5691 		sp->pfs_1301.nat_rule = htonl(-1);
5692 	else
5693 		sp->pfs_1301.nat_rule = htonl(st->nat_rule->nr);
5694 
5695 	pf_state_counter_hton(st->packets[0], sp->pfs_1301.packets[0]);
5696 	pf_state_counter_hton(st->packets[1], sp->pfs_1301.packets[1]);
5697 	pf_state_counter_hton(st->bytes[0], sp->pfs_1301.bytes[0]);
5698 	pf_state_counter_hton(st->bytes[1], sp->pfs_1301.bytes[1]);
5699 }
5700 
5701 void
5702 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st)
5703 {
5704 	bzero(sp, sizeof(*sp));
5705 
5706 	sp->version = PF_STATE_VERSION;
5707 
5708 	/* copy from state key */
5709 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5710 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5711 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5712 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5713 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5714 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5715 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5716 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5717 	sp->proto = st->key[PF_SK_WIRE]->proto;
5718 	sp->af = st->key[PF_SK_WIRE]->af;
5719 
5720 	/* copy from state */
5721 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
5722 	strlcpy(sp->orig_ifname, st->orig_kif->pfik_name,
5723 	    sizeof(sp->orig_ifname));
5724 	bcopy(&st->act.rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
5725 	sp->creation = htonl(time_uptime - (st->creation / 1000));
5726 	sp->expire = pf_state_expires(st);
5727 	if (sp->expire <= time_uptime)
5728 		sp->expire = htonl(0);
5729 	else
5730 		sp->expire = htonl(sp->expire - time_uptime);
5731 
5732 	sp->direction = st->direction;
5733 	sp->log = st->act.log;
5734 	sp->timeout = st->timeout;
5735 	/* 8 bits for the old libpfctl, 16 bits for the new libpfctl */
5736 	sp->state_flags_compat = st->state_flags;
5737 	sp->state_flags = htons(st->state_flags);
5738 	if (st->src_node)
5739 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
5740 	if (st->nat_src_node)
5741 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5742 
5743 	sp->id = st->id;
5744 	sp->creatorid = st->creatorid;
5745 	pf_state_peer_hton(&st->src, &sp->src);
5746 	pf_state_peer_hton(&st->dst, &sp->dst);
5747 
5748 	if (st->rule == NULL)
5749 		sp->rule = htonl(-1);
5750 	else
5751 		sp->rule = htonl(st->rule->nr);
5752 	if (st->anchor == NULL)
5753 		sp->anchor = htonl(-1);
5754 	else
5755 		sp->anchor = htonl(st->anchor->nr);
5756 	if (st->nat_rule == NULL)
5757 		sp->nat_rule = htonl(-1);
5758 	else
5759 		sp->nat_rule = htonl(st->nat_rule->nr);
5760 
5761 	sp->packets[0] = st->packets[0];
5762 	sp->packets[1] = st->packets[1];
5763 	sp->bytes[0] = st->bytes[0];
5764 	sp->bytes[1] = st->bytes[1];
5765 
5766 	sp->qid = htons(st->act.qid);
5767 	sp->pqid = htons(st->act.pqid);
5768 	sp->dnpipe = htons(st->act.dnpipe);
5769 	sp->dnrpipe = htons(st->act.dnrpipe);
5770 	sp->rtableid = htonl(st->act.rtableid);
5771 	sp->min_ttl = st->act.min_ttl;
5772 	sp->set_tos = st->act.set_tos;
5773 	sp->max_mss = htons(st->act.max_mss);
5774 	sp->rt = st->act.rt;
5775 	if (st->act.rt_kif)
5776 		strlcpy(sp->rt_ifname, st->act.rt_kif->pfik_name,
5777 		    sizeof(sp->rt_ifname));
5778 	sp->set_prio[0] = st->act.set_prio[0];
5779 	sp->set_prio[1] = st->act.set_prio[1];
5780 
5781 }
5782 
5783 static void
5784 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
5785 {
5786 	struct pfr_ktable *kt;
5787 
5788 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
5789 
5790 	kt = aw->p.tbl;
5791 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
5792 		kt = kt->pfrkt_root;
5793 	aw->p.tbl = NULL;
5794 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
5795 		kt->pfrkt_cnt : -1;
5796 }
5797 
5798 static int
5799 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters,
5800     size_t number, char **names)
5801 {
5802 	nvlist_t        *nvc;
5803 
5804 	nvc = nvlist_create(0);
5805 	if (nvc == NULL)
5806 		return (ENOMEM);
5807 
5808 	for (int i = 0; i < number; i++) {
5809 		nvlist_append_number_array(nvc, "counters",
5810 		    counter_u64_fetch(counters[i]));
5811 		nvlist_append_string_array(nvc, "names",
5812 		    names[i]);
5813 		nvlist_append_number_array(nvc, "ids",
5814 		    i);
5815 	}
5816 	nvlist_add_nvlist(nvl, name, nvc);
5817 	nvlist_destroy(nvc);
5818 
5819 	return (0);
5820 }
5821 
5822 static int
5823 pf_getstatus(struct pfioc_nv *nv)
5824 {
5825 	nvlist_t        *nvl = NULL, *nvc = NULL;
5826 	void            *nvlpacked = NULL;
5827 	int              error;
5828 	struct pf_status s;
5829 	char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES;
5830 	char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES;
5831 	char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES;
5832 	PF_RULES_RLOCK_TRACKER;
5833 
5834 #define ERROUT(x)      ERROUT_FUNCTION(errout, x)
5835 
5836 	PF_RULES_RLOCK();
5837 
5838 	nvl = nvlist_create(0);
5839 	if (nvl == NULL)
5840 		ERROUT(ENOMEM);
5841 
5842 	nvlist_add_bool(nvl, "running", V_pf_status.running);
5843 	nvlist_add_number(nvl, "since", V_pf_status.since);
5844 	nvlist_add_number(nvl, "debug", V_pf_status.debug);
5845 	nvlist_add_number(nvl, "hostid", V_pf_status.hostid);
5846 	nvlist_add_number(nvl, "states", V_pf_status.states);
5847 	nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes);
5848 	nvlist_add_number(nvl, "reass", V_pf_status.reass);
5849 	nvlist_add_bool(nvl, "syncookies_active",
5850 	    V_pf_status.syncookies_active);
5851 	nvlist_add_number(nvl, "halfopen_states", V_pf_status.states_halfopen);
5852 
5853 	/* counters */
5854 	error = pf_add_status_counters(nvl, "counters", V_pf_status.counters,
5855 	    PFRES_MAX, pf_reasons);
5856 	if (error != 0)
5857 		ERROUT(error);
5858 
5859 	/* lcounters */
5860 	error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters,
5861 	    KLCNT_MAX, pf_lcounter);
5862 	if (error != 0)
5863 		ERROUT(error);
5864 
5865 	/* fcounters */
5866 	nvc = nvlist_create(0);
5867 	if (nvc == NULL)
5868 		ERROUT(ENOMEM);
5869 
5870 	for (int i = 0; i < FCNT_MAX; i++) {
5871 		nvlist_append_number_array(nvc, "counters",
5872 		    pf_counter_u64_fetch(&V_pf_status.fcounters[i]));
5873 		nvlist_append_string_array(nvc, "names",
5874 		    pf_fcounter[i]);
5875 		nvlist_append_number_array(nvc, "ids",
5876 		    i);
5877 	}
5878 	nvlist_add_nvlist(nvl, "fcounters", nvc);
5879 	nvlist_destroy(nvc);
5880 	nvc = NULL;
5881 
5882 	/* scounters */
5883 	error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters,
5884 	    SCNT_MAX, pf_fcounter);
5885 	if (error != 0)
5886 		ERROUT(error);
5887 
5888 	nvlist_add_string(nvl, "ifname", V_pf_status.ifname);
5889 	nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum,
5890 	    PF_MD5_DIGEST_LENGTH);
5891 
5892 	pfi_update_status(V_pf_status.ifname, &s);
5893 
5894 	/* pcounters / bcounters */
5895 	for (int i = 0; i < 2; i++) {
5896 		for (int j = 0; j < 2; j++) {
5897 			for (int k = 0; k < 2; k++) {
5898 				nvlist_append_number_array(nvl, "pcounters",
5899 				    s.pcounters[i][j][k]);
5900 			}
5901 			nvlist_append_number_array(nvl, "bcounters",
5902 			    s.bcounters[i][j]);
5903 		}
5904 	}
5905 
5906 	nvlpacked = nvlist_pack(nvl, &nv->len);
5907 	if (nvlpacked == NULL)
5908 		ERROUT(ENOMEM);
5909 
5910 	if (nv->size == 0)
5911 		ERROUT(0);
5912 	else if (nv->size < nv->len)
5913 		ERROUT(ENOSPC);
5914 
5915 	PF_RULES_RUNLOCK();
5916 	error = copyout(nvlpacked, nv->data, nv->len);
5917 	goto done;
5918 
5919 #undef ERROUT
5920 errout:
5921 	PF_RULES_RUNLOCK();
5922 done:
5923 	free(nvlpacked, M_NVLIST);
5924 	nvlist_destroy(nvc);
5925 	nvlist_destroy(nvl);
5926 
5927 	return (error);
5928 }
5929 
5930 /*
5931  * XXX - Check for version mismatch!!!
5932  */
5933 static void
5934 pf_clear_all_states(void)
5935 {
5936 	struct epoch_tracker	 et;
5937 	struct pf_kstate	*s;
5938 	u_int i;
5939 
5940 	NET_EPOCH_ENTER(et);
5941 	for (i = 0; i <= V_pf_hashmask; i++) {
5942 		struct pf_idhash *ih = &V_pf_idhash[i];
5943 relock:
5944 		PF_HASHROW_LOCK(ih);
5945 		LIST_FOREACH(s, &ih->states, entry) {
5946 			s->timeout = PFTM_PURGE;
5947 			/* Don't send out individual delete messages. */
5948 			s->state_flags |= PFSTATE_NOSYNC;
5949 			pf_unlink_state(s);
5950 			goto relock;
5951 		}
5952 		PF_HASHROW_UNLOCK(ih);
5953 	}
5954 	NET_EPOCH_EXIT(et);
5955 }
5956 
5957 static int
5958 pf_clear_tables(void)
5959 {
5960 	struct pfioc_table io;
5961 	int error;
5962 
5963 	bzero(&io, sizeof(io));
5964 	io.pfrio_flags |= PFR_FLAG_ALLRSETS;
5965 
5966 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
5967 	    io.pfrio_flags);
5968 
5969 	return (error);
5970 }
5971 
5972 static void
5973 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
5974 {
5975 	struct pf_ksrc_node_list	 kill;
5976 	u_int 				 killed;
5977 
5978 	LIST_INIT(&kill);
5979 	for (int i = 0; i <= V_pf_srchashmask; i++) {
5980 		struct pf_srchash *sh = &V_pf_srchash[i];
5981 		struct pf_ksrc_node *sn, *tmp;
5982 
5983 		PF_HASHROW_LOCK(sh);
5984 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
5985 			if (psnk == NULL ||
5986 			    (PF_MATCHA(psnk->psnk_src.neg,
5987 			      &psnk->psnk_src.addr.v.a.addr,
5988 			      &psnk->psnk_src.addr.v.a.mask,
5989 			      &sn->addr, sn->af) &&
5990 			    PF_MATCHA(psnk->psnk_dst.neg,
5991 			      &psnk->psnk_dst.addr.v.a.addr,
5992 			      &psnk->psnk_dst.addr.v.a.mask,
5993 			      &sn->raddr, sn->af))) {
5994 				pf_unlink_src_node(sn);
5995 				LIST_INSERT_HEAD(&kill, sn, entry);
5996 				sn->expire = 1;
5997 			}
5998 		PF_HASHROW_UNLOCK(sh);
5999 	}
6000 
6001 	for (int i = 0; i <= V_pf_hashmask; i++) {
6002 		struct pf_idhash *ih = &V_pf_idhash[i];
6003 		struct pf_kstate *s;
6004 
6005 		PF_HASHROW_LOCK(ih);
6006 		LIST_FOREACH(s, &ih->states, entry) {
6007 			if (s->src_node && s->src_node->expire == 1)
6008 				s->src_node = NULL;
6009 			if (s->nat_src_node && s->nat_src_node->expire == 1)
6010 				s->nat_src_node = NULL;
6011 		}
6012 		PF_HASHROW_UNLOCK(ih);
6013 	}
6014 
6015 	killed = pf_free_src_nodes(&kill);
6016 
6017 	if (psnk != NULL)
6018 		psnk->psnk_killed = killed;
6019 }
6020 
6021 static int
6022 pf_keepcounters(struct pfioc_nv *nv)
6023 {
6024 	nvlist_t	*nvl = NULL;
6025 	void		*nvlpacked = NULL;
6026 	int		 error = 0;
6027 
6028 #define	ERROUT(x)	ERROUT_FUNCTION(on_error, x)
6029 
6030 	if (nv->len > pf_ioctl_maxcount)
6031 		ERROUT(ENOMEM);
6032 
6033 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6034 	error = copyin(nv->data, nvlpacked, nv->len);
6035 	if (error)
6036 		ERROUT(error);
6037 
6038 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6039 	if (nvl == NULL)
6040 		ERROUT(EBADMSG);
6041 
6042 	if (! nvlist_exists_bool(nvl, "keep_counters"))
6043 		ERROUT(EBADMSG);
6044 
6045 	V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters");
6046 
6047 on_error:
6048 	nvlist_destroy(nvl);
6049 	free(nvlpacked, M_NVLIST);
6050 	return (error);
6051 }
6052 
6053 unsigned int
6054 pf_clear_states(const struct pf_kstate_kill *kill)
6055 {
6056 	struct pf_state_key_cmp	 match_key;
6057 	struct pf_kstate	*s;
6058 	struct pfi_kkif	*kif;
6059 	int		 idx;
6060 	unsigned int	 killed = 0, dir;
6061 
6062 	NET_EPOCH_ASSERT();
6063 
6064 	for (unsigned int i = 0; i <= V_pf_hashmask; i++) {
6065 		struct pf_idhash *ih = &V_pf_idhash[i];
6066 
6067 relock_DIOCCLRSTATES:
6068 		PF_HASHROW_LOCK(ih);
6069 		LIST_FOREACH(s, &ih->states, entry) {
6070 			/* For floating states look at the original kif. */
6071 			kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
6072 
6073 			if (kill->psk_ifname[0] &&
6074 			    strcmp(kill->psk_ifname,
6075 			    kif->pfik_name))
6076 				continue;
6077 
6078 			if (kill->psk_kill_match) {
6079 				bzero(&match_key, sizeof(match_key));
6080 
6081 				if (s->direction == PF_OUT) {
6082 					dir = PF_IN;
6083 					idx = PF_SK_STACK;
6084 				} else {
6085 					dir = PF_OUT;
6086 					idx = PF_SK_WIRE;
6087 				}
6088 
6089 				match_key.af = s->key[idx]->af;
6090 				match_key.proto = s->key[idx]->proto;
6091 				PF_ACPY(&match_key.addr[0],
6092 				    &s->key[idx]->addr[1], match_key.af);
6093 				match_key.port[0] = s->key[idx]->port[1];
6094 				PF_ACPY(&match_key.addr[1],
6095 				    &s->key[idx]->addr[0], match_key.af);
6096 				match_key.port[1] = s->key[idx]->port[0];
6097 			}
6098 
6099 			/*
6100 			 * Don't send out individual
6101 			 * delete messages.
6102 			 */
6103 			s->state_flags |= PFSTATE_NOSYNC;
6104 			pf_unlink_state(s);
6105 			killed++;
6106 
6107 			if (kill->psk_kill_match)
6108 				killed += pf_kill_matching_state(&match_key,
6109 				    dir);
6110 
6111 			goto relock_DIOCCLRSTATES;
6112 		}
6113 		PF_HASHROW_UNLOCK(ih);
6114 	}
6115 
6116 	if (V_pfsync_clear_states_ptr != NULL)
6117 		V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname);
6118 
6119 	return (killed);
6120 }
6121 
6122 void
6123 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed)
6124 {
6125 	struct pf_kstate	*s;
6126 
6127 	NET_EPOCH_ASSERT();
6128 	if (kill->psk_pfcmp.id) {
6129 		if (kill->psk_pfcmp.creatorid == 0)
6130 			kill->psk_pfcmp.creatorid = V_pf_status.hostid;
6131 		if ((s = pf_find_state_byid(kill->psk_pfcmp.id,
6132 		    kill->psk_pfcmp.creatorid))) {
6133 			pf_unlink_state(s);
6134 			*killed = 1;
6135 		}
6136 		return;
6137 	}
6138 
6139 	for (unsigned int i = 0; i <= V_pf_hashmask; i++)
6140 		*killed += pf_killstates_row(kill, &V_pf_idhash[i]);
6141 }
6142 
6143 static int
6144 pf_killstates_nv(struct pfioc_nv *nv)
6145 {
6146 	struct pf_kstate_kill	 kill;
6147 	struct epoch_tracker	 et;
6148 	nvlist_t		*nvl = NULL;
6149 	void			*nvlpacked = NULL;
6150 	int			 error = 0;
6151 	unsigned int		 killed = 0;
6152 
6153 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
6154 
6155 	if (nv->len > pf_ioctl_maxcount)
6156 		ERROUT(ENOMEM);
6157 
6158 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6159 	error = copyin(nv->data, nvlpacked, nv->len);
6160 	if (error)
6161 		ERROUT(error);
6162 
6163 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6164 	if (nvl == NULL)
6165 		ERROUT(EBADMSG);
6166 
6167 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6168 	if (error)
6169 		ERROUT(error);
6170 
6171 	NET_EPOCH_ENTER(et);
6172 	pf_killstates(&kill, &killed);
6173 	NET_EPOCH_EXIT(et);
6174 
6175 	free(nvlpacked, M_NVLIST);
6176 	nvlpacked = NULL;
6177 	nvlist_destroy(nvl);
6178 	nvl = nvlist_create(0);
6179 	if (nvl == NULL)
6180 		ERROUT(ENOMEM);
6181 
6182 	nvlist_add_number(nvl, "killed", killed);
6183 
6184 	nvlpacked = nvlist_pack(nvl, &nv->len);
6185 	if (nvlpacked == NULL)
6186 		ERROUT(ENOMEM);
6187 
6188 	if (nv->size == 0)
6189 		ERROUT(0);
6190 	else if (nv->size < nv->len)
6191 		ERROUT(ENOSPC);
6192 
6193 	error = copyout(nvlpacked, nv->data, nv->len);
6194 
6195 on_error:
6196 	nvlist_destroy(nvl);
6197 	free(nvlpacked, M_NVLIST);
6198 	return (error);
6199 }
6200 
6201 static int
6202 pf_clearstates_nv(struct pfioc_nv *nv)
6203 {
6204 	struct pf_kstate_kill	 kill;
6205 	struct epoch_tracker	 et;
6206 	nvlist_t		*nvl = NULL;
6207 	void			*nvlpacked = NULL;
6208 	int			 error = 0;
6209 	unsigned int		 killed;
6210 
6211 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
6212 
6213 	if (nv->len > pf_ioctl_maxcount)
6214 		ERROUT(ENOMEM);
6215 
6216 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6217 	error = copyin(nv->data, nvlpacked, nv->len);
6218 	if (error)
6219 		ERROUT(error);
6220 
6221 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6222 	if (nvl == NULL)
6223 		ERROUT(EBADMSG);
6224 
6225 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6226 	if (error)
6227 		ERROUT(error);
6228 
6229 	NET_EPOCH_ENTER(et);
6230 	killed = pf_clear_states(&kill);
6231 	NET_EPOCH_EXIT(et);
6232 
6233 	free(nvlpacked, M_NVLIST);
6234 	nvlpacked = NULL;
6235 	nvlist_destroy(nvl);
6236 	nvl = nvlist_create(0);
6237 	if (nvl == NULL)
6238 		ERROUT(ENOMEM);
6239 
6240 	nvlist_add_number(nvl, "killed", killed);
6241 
6242 	nvlpacked = nvlist_pack(nvl, &nv->len);
6243 	if (nvlpacked == NULL)
6244 		ERROUT(ENOMEM);
6245 
6246 	if (nv->size == 0)
6247 		ERROUT(0);
6248 	else if (nv->size < nv->len)
6249 		ERROUT(ENOSPC);
6250 
6251 	error = copyout(nvlpacked, nv->data, nv->len);
6252 
6253 #undef ERROUT
6254 on_error:
6255 	nvlist_destroy(nvl);
6256 	free(nvlpacked, M_NVLIST);
6257 	return (error);
6258 }
6259 
6260 static int
6261 pf_getstate(struct pfioc_nv *nv)
6262 {
6263 	nvlist_t		*nvl = NULL, *nvls;
6264 	void			*nvlpacked = NULL;
6265 	struct pf_kstate	*s = NULL;
6266 	int			 error = 0;
6267 	uint64_t		 id, creatorid;
6268 
6269 #define ERROUT(x)	ERROUT_FUNCTION(errout, x)
6270 
6271 	if (nv->len > pf_ioctl_maxcount)
6272 		ERROUT(ENOMEM);
6273 
6274 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6275 	error = copyin(nv->data, nvlpacked, nv->len);
6276 	if (error)
6277 		ERROUT(error);
6278 
6279 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6280 	if (nvl == NULL)
6281 		ERROUT(EBADMSG);
6282 
6283 	PFNV_CHK(pf_nvuint64(nvl, "id", &id));
6284 	PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid));
6285 
6286 	s = pf_find_state_byid(id, creatorid);
6287 	if (s == NULL)
6288 		ERROUT(ENOENT);
6289 
6290 	free(nvlpacked, M_NVLIST);
6291 	nvlpacked = NULL;
6292 	nvlist_destroy(nvl);
6293 	nvl = nvlist_create(0);
6294 	if (nvl == NULL)
6295 		ERROUT(ENOMEM);
6296 
6297 	nvls = pf_state_to_nvstate(s);
6298 	if (nvls == NULL)
6299 		ERROUT(ENOMEM);
6300 
6301 	nvlist_add_nvlist(nvl, "state", nvls);
6302 	nvlist_destroy(nvls);
6303 
6304 	nvlpacked = nvlist_pack(nvl, &nv->len);
6305 	if (nvlpacked == NULL)
6306 		ERROUT(ENOMEM);
6307 
6308 	if (nv->size == 0)
6309 		ERROUT(0);
6310 	else if (nv->size < nv->len)
6311 		ERROUT(ENOSPC);
6312 
6313 	error = copyout(nvlpacked, nv->data, nv->len);
6314 
6315 #undef ERROUT
6316 errout:
6317 	if (s != NULL)
6318 		PF_STATE_UNLOCK(s);
6319 	free(nvlpacked, M_NVLIST);
6320 	nvlist_destroy(nvl);
6321 	return (error);
6322 }
6323 
6324 /*
6325  * XXX - Check for version mismatch!!!
6326  */
6327 
6328 /*
6329  * Duplicate pfctl -Fa operation to get rid of as much as we can.
6330  */
6331 static int
6332 shutdown_pf(void)
6333 {
6334 	int error = 0;
6335 	u_int32_t t[5];
6336 	char nn = '\0';
6337 	struct pf_kanchor *anchor;
6338 	struct pf_keth_anchor *eth_anchor;
6339 	int rs_num;
6340 
6341 	do {
6342 		/* Unlink rules of all user defined anchors */
6343 		RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) {
6344 			/* Wildcard based anchors may not have a respective
6345 			 * explicit anchor rule or they may be left empty
6346 			 * without rules. It leads to anchor.refcnt=0, and the
6347 			 * rest of the logic does not expect it. */
6348 			if (anchor->refcnt == 0)
6349 				anchor->refcnt = 1;
6350 			for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6351 				if ((error = pf_begin_rules(&t[rs_num], rs_num,
6352 				    anchor->path)) != 0) {
6353 					DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: "
6354 					    "anchor.path=%s rs_num=%d\n",
6355 					    anchor->path, rs_num));
6356 					goto error;	/* XXX: rollback? */
6357 				}
6358 			}
6359 			for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6360 				error = pf_commit_rules(t[rs_num], rs_num,
6361 				    anchor->path);
6362 				MPASS(error == 0);
6363 			}
6364 		}
6365 
6366 		/* Unlink rules of all user defined ether anchors */
6367 		RB_FOREACH(eth_anchor, pf_keth_anchor_global,
6368 		    &V_pf_keth_anchors) {
6369 			/* Wildcard based anchors may not have a respective
6370 			 * explicit anchor rule or they may be left empty
6371 			 * without rules. It leads to anchor.refcnt=0, and the
6372 			 * rest of the logic does not expect it. */
6373 			if (eth_anchor->refcnt == 0)
6374 				eth_anchor->refcnt = 1;
6375 			if ((error = pf_begin_eth(&t[0], eth_anchor->path))
6376 			    != 0) {
6377 				DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth "
6378 				    "anchor.path=%s\n", eth_anchor->path));
6379 				goto error;
6380 			}
6381 			error = pf_commit_eth(t[0], eth_anchor->path);
6382 			MPASS(error == 0);
6383 		}
6384 
6385 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
6386 		    != 0) {
6387 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
6388 			break;
6389 		}
6390 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
6391 		    != 0) {
6392 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
6393 			break;		/* XXX: rollback? */
6394 		}
6395 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
6396 		    != 0) {
6397 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
6398 			break;		/* XXX: rollback? */
6399 		}
6400 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
6401 		    != 0) {
6402 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
6403 			break;		/* XXX: rollback? */
6404 		}
6405 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
6406 		    != 0) {
6407 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
6408 			break;		/* XXX: rollback? */
6409 		}
6410 
6411 		error = pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
6412 		MPASS(error == 0);
6413 		error = pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
6414 		MPASS(error == 0);
6415 		error = pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
6416 		MPASS(error == 0);
6417 		error = pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
6418 		MPASS(error == 0);
6419 		error = pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
6420 		MPASS(error == 0);
6421 
6422 		if ((error = pf_clear_tables()) != 0)
6423 			break;
6424 
6425 		if ((error = pf_begin_eth(&t[0], &nn)) != 0) {
6426 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth\n"));
6427 			break;
6428 		}
6429 		error = pf_commit_eth(t[0], &nn);
6430 		MPASS(error == 0);
6431 
6432 #ifdef ALTQ
6433 		if ((error = pf_begin_altq(&t[0])) != 0) {
6434 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
6435 			break;
6436 		}
6437 		pf_commit_altq(t[0]);
6438 #endif
6439 
6440 		pf_clear_all_states();
6441 
6442 		pf_kill_srcnodes(NULL);
6443 
6444 		/* status does not use malloced mem so no need to cleanup */
6445 		/* fingerprints and interfaces have their own cleanup code */
6446 	} while(0);
6447 
6448 error:
6449 	return (error);
6450 }
6451 
6452 static pfil_return_t
6453 pf_check_return(int chk, struct mbuf **m)
6454 {
6455 
6456 	switch (chk) {
6457 	case PF_PASS:
6458 		if (*m == NULL)
6459 			return (PFIL_CONSUMED);
6460 		else
6461 			return (PFIL_PASS);
6462 		break;
6463 	default:
6464 		if (*m != NULL) {
6465 			m_freem(*m);
6466 			*m = NULL;
6467 		}
6468 		return (PFIL_DROPPED);
6469 	}
6470 }
6471 
6472 static pfil_return_t
6473 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6474     void *ruleset __unused, struct inpcb *inp)
6475 {
6476 	int chk;
6477 
6478 	CURVNET_ASSERT_SET();
6479 
6480 	chk = pf_test_eth(PF_IN, flags, ifp, m, inp);
6481 
6482 	return (pf_check_return(chk, m));
6483 }
6484 
6485 static pfil_return_t
6486 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6487     void *ruleset __unused, struct inpcb *inp)
6488 {
6489 	int chk;
6490 
6491 	CURVNET_ASSERT_SET();
6492 
6493 	chk = pf_test_eth(PF_OUT, flags, ifp, m, inp);
6494 
6495 	return (pf_check_return(chk, m));
6496 }
6497 
6498 #ifdef INET
6499 static pfil_return_t
6500 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6501     void *ruleset __unused, struct inpcb *inp)
6502 {
6503 	int chk;
6504 
6505 	CURVNET_ASSERT_SET();
6506 
6507 	chk = pf_test(AF_INET, PF_IN, flags, ifp, m, inp, NULL);
6508 
6509 	return (pf_check_return(chk, m));
6510 }
6511 
6512 static pfil_return_t
6513 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6514     void *ruleset __unused,  struct inpcb *inp)
6515 {
6516 	int chk;
6517 
6518 	CURVNET_ASSERT_SET();
6519 
6520 	chk = pf_test(AF_INET, PF_OUT, flags, ifp, m, inp, NULL);
6521 
6522 	return (pf_check_return(chk, m));
6523 }
6524 #endif
6525 
6526 #ifdef INET6
6527 static pfil_return_t
6528 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
6529     void *ruleset __unused,  struct inpcb *inp)
6530 {
6531 	int chk;
6532 
6533 	CURVNET_ASSERT_SET();
6534 
6535 	/*
6536 	 * In case of loopback traffic IPv6 uses the real interface in
6537 	 * order to support scoped addresses. In order to support stateful
6538 	 * filtering we have change this to lo0 as it is the case in IPv4.
6539 	 */
6540 	chk = pf_test(AF_INET6, PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp,
6541 	    m, inp, NULL);
6542 
6543 	return (pf_check_return(chk, m));
6544 }
6545 
6546 static pfil_return_t
6547 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
6548     void *ruleset __unused,  struct inpcb *inp)
6549 {
6550 	int chk;
6551 
6552 	CURVNET_ASSERT_SET();
6553 
6554 	chk = pf_test(AF_INET6, PF_OUT, flags, ifp, m, inp, NULL);
6555 
6556 	return (pf_check_return(chk, m));
6557 }
6558 #endif /* INET6 */
6559 
6560 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook);
6561 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook);
6562 #define	V_pf_eth_in_hook	VNET(pf_eth_in_hook)
6563 #define	V_pf_eth_out_hook	VNET(pf_eth_out_hook)
6564 
6565 #ifdef INET
6566 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
6567 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
6568 #define	V_pf_ip4_in_hook	VNET(pf_ip4_in_hook)
6569 #define	V_pf_ip4_out_hook	VNET(pf_ip4_out_hook)
6570 #endif
6571 #ifdef INET6
6572 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
6573 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
6574 #define	V_pf_ip6_in_hook	VNET(pf_ip6_in_hook)
6575 #define	V_pf_ip6_out_hook	VNET(pf_ip6_out_hook)
6576 #endif
6577 
6578 static void
6579 hook_pf_eth(void)
6580 {
6581 	struct pfil_hook_args pha = {
6582 		.pa_version = PFIL_VERSION,
6583 		.pa_modname = "pf",
6584 		.pa_type = PFIL_TYPE_ETHERNET,
6585 	};
6586 	struct pfil_link_args pla = {
6587 		.pa_version = PFIL_VERSION,
6588 	};
6589 	int ret __diagused;
6590 
6591 	if (atomic_load_bool(&V_pf_pfil_eth_hooked))
6592 		return;
6593 
6594 	pha.pa_mbuf_chk = pf_eth_check_in;
6595 	pha.pa_flags = PFIL_IN;
6596 	pha.pa_rulname = "eth-in";
6597 	V_pf_eth_in_hook = pfil_add_hook(&pha);
6598 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6599 	pla.pa_head = V_link_pfil_head;
6600 	pla.pa_hook = V_pf_eth_in_hook;
6601 	ret = pfil_link(&pla);
6602 	MPASS(ret == 0);
6603 	pha.pa_mbuf_chk = pf_eth_check_out;
6604 	pha.pa_flags = PFIL_OUT;
6605 	pha.pa_rulname = "eth-out";
6606 	V_pf_eth_out_hook = pfil_add_hook(&pha);
6607 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6608 	pla.pa_head = V_link_pfil_head;
6609 	pla.pa_hook = V_pf_eth_out_hook;
6610 	ret = pfil_link(&pla);
6611 	MPASS(ret == 0);
6612 
6613 	atomic_store_bool(&V_pf_pfil_eth_hooked, true);
6614 }
6615 
6616 static void
6617 hook_pf(void)
6618 {
6619 	struct pfil_hook_args pha = {
6620 		.pa_version = PFIL_VERSION,
6621 		.pa_modname = "pf",
6622 	};
6623 	struct pfil_link_args pla = {
6624 		.pa_version = PFIL_VERSION,
6625 	};
6626 	int ret __diagused;
6627 
6628 	if (atomic_load_bool(&V_pf_pfil_hooked))
6629 		return;
6630 
6631 #ifdef INET
6632 	pha.pa_type = PFIL_TYPE_IP4;
6633 	pha.pa_mbuf_chk = pf_check_in;
6634 	pha.pa_flags = PFIL_IN;
6635 	pha.pa_rulname = "default-in";
6636 	V_pf_ip4_in_hook = pfil_add_hook(&pha);
6637 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6638 	pla.pa_head = V_inet_pfil_head;
6639 	pla.pa_hook = V_pf_ip4_in_hook;
6640 	ret = pfil_link(&pla);
6641 	MPASS(ret == 0);
6642 	pha.pa_mbuf_chk = pf_check_out;
6643 	pha.pa_flags = PFIL_OUT;
6644 	pha.pa_rulname = "default-out";
6645 	V_pf_ip4_out_hook = pfil_add_hook(&pha);
6646 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6647 	pla.pa_head = V_inet_pfil_head;
6648 	pla.pa_hook = V_pf_ip4_out_hook;
6649 	ret = pfil_link(&pla);
6650 	MPASS(ret == 0);
6651 	if (V_pf_filter_local) {
6652 		pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6653 		pla.pa_head = V_inet_local_pfil_head;
6654 		pla.pa_hook = V_pf_ip4_out_hook;
6655 		ret = pfil_link(&pla);
6656 		MPASS(ret == 0);
6657 	}
6658 #endif
6659 #ifdef INET6
6660 	pha.pa_type = PFIL_TYPE_IP6;
6661 	pha.pa_mbuf_chk = pf_check6_in;
6662 	pha.pa_flags = PFIL_IN;
6663 	pha.pa_rulname = "default-in6";
6664 	V_pf_ip6_in_hook = pfil_add_hook(&pha);
6665 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6666 	pla.pa_head = V_inet6_pfil_head;
6667 	pla.pa_hook = V_pf_ip6_in_hook;
6668 	ret = pfil_link(&pla);
6669 	MPASS(ret == 0);
6670 	pha.pa_mbuf_chk = pf_check6_out;
6671 	pha.pa_rulname = "default-out6";
6672 	pha.pa_flags = PFIL_OUT;
6673 	V_pf_ip6_out_hook = pfil_add_hook(&pha);
6674 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6675 	pla.pa_head = V_inet6_pfil_head;
6676 	pla.pa_hook = V_pf_ip6_out_hook;
6677 	ret = pfil_link(&pla);
6678 	MPASS(ret == 0);
6679 	if (V_pf_filter_local) {
6680 		pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6681 		pla.pa_head = V_inet6_local_pfil_head;
6682 		pla.pa_hook = V_pf_ip6_out_hook;
6683 		ret = pfil_link(&pla);
6684 		MPASS(ret == 0);
6685 	}
6686 #endif
6687 
6688 	atomic_store_bool(&V_pf_pfil_hooked, true);
6689 }
6690 
6691 static void
6692 dehook_pf_eth(void)
6693 {
6694 
6695 	if (!atomic_load_bool(&V_pf_pfil_eth_hooked))
6696 		return;
6697 
6698 	pfil_remove_hook(V_pf_eth_in_hook);
6699 	pfil_remove_hook(V_pf_eth_out_hook);
6700 
6701 	atomic_store_bool(&V_pf_pfil_eth_hooked, false);
6702 }
6703 
6704 static void
6705 dehook_pf(void)
6706 {
6707 
6708 	if (!atomic_load_bool(&V_pf_pfil_hooked))
6709 		return;
6710 
6711 #ifdef INET
6712 	pfil_remove_hook(V_pf_ip4_in_hook);
6713 	pfil_remove_hook(V_pf_ip4_out_hook);
6714 #endif
6715 #ifdef INET6
6716 	pfil_remove_hook(V_pf_ip6_in_hook);
6717 	pfil_remove_hook(V_pf_ip6_out_hook);
6718 #endif
6719 
6720 	atomic_store_bool(&V_pf_pfil_hooked, false);
6721 }
6722 
6723 static void
6724 pf_load_vnet(void)
6725 {
6726 	V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
6727 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
6728 
6729 	rm_init_flags(&V_pf_rules_lock, "pf rulesets", RM_RECURSE);
6730 	sx_init(&V_pf_ioctl_lock, "pf ioctl");
6731 
6732 	pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
6733 	    PF_RULE_TAG_HASH_SIZE_DEFAULT);
6734 #ifdef ALTQ
6735 	pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
6736 	    PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
6737 #endif
6738 
6739 	V_pf_keth = &V_pf_main_keth_anchor.ruleset;
6740 
6741 	pfattach_vnet();
6742 	V_pf_vnet_active = 1;
6743 }
6744 
6745 static int
6746 pf_load(void)
6747 {
6748 	int error;
6749 
6750 	sx_init(&pf_end_lock, "pf end thread");
6751 
6752 	pf_mtag_initialize();
6753 
6754 	pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
6755 	if (pf_dev == NULL)
6756 		return (ENOMEM);
6757 
6758 	pf_end_threads = 0;
6759 	error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
6760 	if (error != 0)
6761 		return (error);
6762 
6763 	pfi_initialize();
6764 
6765 	return (0);
6766 }
6767 
6768 static void
6769 pf_unload_vnet(void)
6770 {
6771 	int ret __diagused;
6772 
6773 	V_pf_vnet_active = 0;
6774 	V_pf_status.running = 0;
6775 	dehook_pf();
6776 	dehook_pf_eth();
6777 
6778 	PF_RULES_WLOCK();
6779 	pf_syncookies_cleanup();
6780 	shutdown_pf();
6781 	PF_RULES_WUNLOCK();
6782 
6783 	/* Make sure we've cleaned up ethernet rules before we continue. */
6784 	NET_EPOCH_DRAIN_CALLBACKS();
6785 
6786 	ret = swi_remove(V_pf_swi_cookie);
6787 	MPASS(ret == 0);
6788 	ret = intr_event_destroy(V_pf_swi_ie);
6789 	MPASS(ret == 0);
6790 
6791 	pf_unload_vnet_purge();
6792 
6793 	pf_normalize_cleanup();
6794 	PF_RULES_WLOCK();
6795 	pfi_cleanup_vnet();
6796 	PF_RULES_WUNLOCK();
6797 	pfr_cleanup();
6798 	pf_osfp_flush();
6799 	pf_cleanup();
6800 	if (IS_DEFAULT_VNET(curvnet))
6801 		pf_mtag_cleanup();
6802 
6803 	pf_cleanup_tagset(&V_pf_tags);
6804 #ifdef ALTQ
6805 	pf_cleanup_tagset(&V_pf_qids);
6806 #endif
6807 	uma_zdestroy(V_pf_tag_z);
6808 
6809 #ifdef PF_WANT_32_TO_64_COUNTER
6810 	PF_RULES_WLOCK();
6811 	LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist);
6812 
6813 	MPASS(LIST_EMPTY(&V_pf_allkiflist));
6814 	MPASS(V_pf_allkifcount == 0);
6815 
6816 	LIST_REMOVE(&V_pf_default_rule, allrulelist);
6817 	V_pf_allrulecount--;
6818 	LIST_REMOVE(V_pf_rulemarker, allrulelist);
6819 
6820 	MPASS(LIST_EMPTY(&V_pf_allrulelist));
6821 	MPASS(V_pf_allrulecount == 0);
6822 
6823 	PF_RULES_WUNLOCK();
6824 
6825 	free(V_pf_kifmarker, PFI_MTYPE);
6826 	free(V_pf_rulemarker, M_PFRULE);
6827 #endif
6828 
6829 	/* Free counters last as we updated them during shutdown. */
6830 	pf_counter_u64_deinit(&V_pf_default_rule.evaluations);
6831 	for (int i = 0; i < 2; i++) {
6832 		pf_counter_u64_deinit(&V_pf_default_rule.packets[i]);
6833 		pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]);
6834 	}
6835 	counter_u64_free(V_pf_default_rule.states_cur);
6836 	counter_u64_free(V_pf_default_rule.states_tot);
6837 	counter_u64_free(V_pf_default_rule.src_nodes);
6838 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, V_pf_default_rule.timestamp);
6839 
6840 	for (int i = 0; i < PFRES_MAX; i++)
6841 		counter_u64_free(V_pf_status.counters[i]);
6842 	for (int i = 0; i < KLCNT_MAX; i++)
6843 		counter_u64_free(V_pf_status.lcounters[i]);
6844 	for (int i = 0; i < FCNT_MAX; i++)
6845 		pf_counter_u64_deinit(&V_pf_status.fcounters[i]);
6846 	for (int i = 0; i < SCNT_MAX; i++)
6847 		counter_u64_free(V_pf_status.scounters[i]);
6848 
6849 	rm_destroy(&V_pf_rules_lock);
6850 	sx_destroy(&V_pf_ioctl_lock);
6851 }
6852 
6853 static void
6854 pf_unload(void)
6855 {
6856 
6857 	sx_xlock(&pf_end_lock);
6858 	pf_end_threads = 1;
6859 	while (pf_end_threads < 2) {
6860 		wakeup_one(pf_purge_thread);
6861 		sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
6862 	}
6863 	sx_xunlock(&pf_end_lock);
6864 
6865 	pf_nl_unregister();
6866 
6867 	if (pf_dev != NULL)
6868 		destroy_dev(pf_dev);
6869 
6870 	pfi_cleanup();
6871 
6872 	sx_destroy(&pf_end_lock);
6873 }
6874 
6875 static void
6876 vnet_pf_init(void *unused __unused)
6877 {
6878 
6879 	pf_load_vnet();
6880 }
6881 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6882     vnet_pf_init, NULL);
6883 
6884 static void
6885 vnet_pf_uninit(const void *unused __unused)
6886 {
6887 
6888 	pf_unload_vnet();
6889 }
6890 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
6891 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6892     vnet_pf_uninit, NULL);
6893 
6894 static int
6895 pf_modevent(module_t mod, int type, void *data)
6896 {
6897 	int error = 0;
6898 
6899 	switch(type) {
6900 	case MOD_LOAD:
6901 		error = pf_load();
6902 		pf_nl_register();
6903 		break;
6904 	case MOD_UNLOAD:
6905 		/* Handled in SYSUNINIT(pf_unload) to ensure it's done after
6906 		 * the vnet_pf_uninit()s */
6907 		break;
6908 	default:
6909 		error = EINVAL;
6910 		break;
6911 	}
6912 
6913 	return (error);
6914 }
6915 
6916 static moduledata_t pf_mod = {
6917 	"pf",
6918 	pf_modevent,
6919 	0
6920 };
6921 
6922 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
6923 MODULE_DEPEND(pf, netlink, 1, 1, 1);
6924 MODULE_VERSION(pf, PF_MODVER);
6925