xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision 82c57e2a75c7134b70eb2138ff6bbc6e55bef130)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2001 Daniel Hartmeier
5  * Copyright (c) 2002,2003 Henning Brauer
6  * Copyright (c) 2012 Gleb Smirnoff <glebius@FreeBSD.org>
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  *
13  *    - Redistributions of source code must retain the above copyright
14  *      notice, this list of conditions and the following disclaimer.
15  *    - Redistributions in binary form must reproduce the above
16  *      copyright notice, this list of conditions and the following
17  *      disclaimer in the documentation and/or other materials provided
18  *      with the distribution.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
26  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
28  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
30  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
31  * POSSIBILITY OF SUCH DAMAGE.
32  *
33  * Effort sponsored in part by the Defense Advanced Research Projects
34  * Agency (DARPA) and Air Force Research Laboratory, Air Force
35  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
36  *
37  *	$OpenBSD: pf_ioctl.c,v 1.213 2009/02/15 21:46:12 mbalmer Exp $
38  */
39 
40 #include <sys/cdefs.h>
41 #include "opt_inet.h"
42 #include "opt_inet6.h"
43 #include "opt_bpf.h"
44 #include "opt_pf.h"
45 
46 #include <sys/param.h>
47 #include <sys/_bitset.h>
48 #include <sys/bitset.h>
49 #include <sys/bus.h>
50 #include <sys/conf.h>
51 #include <sys/endian.h>
52 #include <sys/fcntl.h>
53 #include <sys/filio.h>
54 #include <sys/hash.h>
55 #include <sys/interrupt.h>
56 #include <sys/jail.h>
57 #include <sys/kernel.h>
58 #include <sys/kthread.h>
59 #include <sys/lock.h>
60 #include <sys/mbuf.h>
61 #include <sys/module.h>
62 #include <sys/nv.h>
63 #include <sys/proc.h>
64 #include <sys/sdt.h>
65 #include <sys/smp.h>
66 #include <sys/socket.h>
67 #include <sys/sysctl.h>
68 #include <sys/md5.h>
69 #include <sys/ucred.h>
70 
71 #include <net/if.h>
72 #include <net/if_var.h>
73 #include <net/if_private.h>
74 #include <net/vnet.h>
75 #include <net/route.h>
76 #include <net/pfil.h>
77 #include <net/pfvar.h>
78 #include <net/if_pfsync.h>
79 #include <net/if_pflog.h>
80 
81 #include <netinet/in.h>
82 #include <netinet/ip.h>
83 #include <netinet/ip_var.h>
84 #include <netinet6/ip6_var.h>
85 #include <netinet/ip_icmp.h>
86 #include <netpfil/pf/pf_nv.h>
87 
88 #ifdef INET6
89 #include <netinet/ip6.h>
90 #endif /* INET6 */
91 
92 #ifdef ALTQ
93 #include <net/altq/altq.h>
94 #endif
95 
96 SDT_PROBE_DEFINE3(pf, ioctl, ioctl, error, "int", "int", "int");
97 SDT_PROBE_DEFINE3(pf, ioctl, function, error, "char *", "int", "int");
98 SDT_PROBE_DEFINE2(pf, ioctl, addrule, error, "int", "int");
99 SDT_PROBE_DEFINE2(pf, ioctl, nvchk, error, "int", "int");
100 
101 static struct pf_kpool	*pf_get_kpool(const char *, u_int32_t, u_int8_t,
102 			    u_int32_t, u_int8_t, u_int8_t, u_int8_t);
103 
104 static void		 pf_mv_kpool(struct pf_kpalist *, struct pf_kpalist *);
105 static void		 pf_empty_kpool(struct pf_kpalist *);
106 static int		 pfioctl(struct cdev *, u_long, caddr_t, int,
107 			    struct thread *);
108 static int		 pf_begin_eth(uint32_t *, const char *);
109 static void		 pf_rollback_eth_cb(struct epoch_context *);
110 static int		 pf_rollback_eth(uint32_t, const char *);
111 static int		 pf_commit_eth(uint32_t, const char *);
112 static void		 pf_free_eth_rule(struct pf_keth_rule *);
113 #ifdef ALTQ
114 static int		 pf_begin_altq(u_int32_t *);
115 static int		 pf_rollback_altq(u_int32_t);
116 static int		 pf_commit_altq(u_int32_t);
117 static int		 pf_enable_altq(struct pf_altq *);
118 static int		 pf_disable_altq(struct pf_altq *);
119 static uint16_t		 pf_qname2qid(const char *);
120 static void		 pf_qid_unref(uint16_t);
121 #endif /* ALTQ */
122 static int		 pf_begin_rules(u_int32_t *, int, const char *);
123 static int		 pf_rollback_rules(u_int32_t, int, char *);
124 static int		 pf_setup_pfsync_matching(struct pf_kruleset *);
125 static void		 pf_hash_rule_rolling(MD5_CTX *, struct pf_krule *);
126 static void		 pf_hash_rule(struct pf_krule *);
127 static void		 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
128 static int		 pf_commit_rules(u_int32_t, int, char *);
129 static int		 pf_addr_setup(struct pf_kruleset *,
130 			    struct pf_addr_wrap *, sa_family_t);
131 static void		 pf_addr_copyout(struct pf_addr_wrap *);
132 static void		 pf_src_node_copy(const struct pf_ksrc_node *,
133 			    struct pf_src_node *);
134 #ifdef ALTQ
135 static int		 pf_export_kaltq(struct pf_altq *,
136 			    struct pfioc_altq_v1 *, size_t);
137 static int		 pf_import_kaltq(struct pfioc_altq_v1 *,
138 			    struct pf_altq *, size_t);
139 #endif /* ALTQ */
140 
141 VNET_DEFINE(struct pf_krule,	pf_default_rule);
142 
143 static __inline int             pf_krule_compare(struct pf_krule *,
144 				    struct pf_krule *);
145 
146 RB_GENERATE(pf_krule_global, pf_krule, entry_global, pf_krule_compare);
147 
148 #ifdef ALTQ
149 VNET_DEFINE_STATIC(int,		pf_altq_running);
150 #define	V_pf_altq_running	VNET(pf_altq_running)
151 #endif
152 
153 #define	TAGID_MAX	 50000
154 struct pf_tagname {
155 	TAILQ_ENTRY(pf_tagname)	namehash_entries;
156 	TAILQ_ENTRY(pf_tagname)	taghash_entries;
157 	char			name[PF_TAG_NAME_SIZE];
158 	uint16_t		tag;
159 	int			ref;
160 };
161 
162 struct pf_tagset {
163 	TAILQ_HEAD(, pf_tagname)	*namehash;
164 	TAILQ_HEAD(, pf_tagname)	*taghash;
165 	unsigned int			 mask;
166 	uint32_t			 seed;
167 	BITSET_DEFINE(, TAGID_MAX)	 avail;
168 };
169 
170 VNET_DEFINE(struct pf_tagset, pf_tags);
171 #define	V_pf_tags	VNET(pf_tags)
172 static unsigned int	pf_rule_tag_hashsize;
173 #define	PF_RULE_TAG_HASH_SIZE_DEFAULT	128
174 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN,
175     &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT,
176     "Size of pf(4) rule tag hashtable");
177 
178 #ifdef ALTQ
179 VNET_DEFINE(struct pf_tagset, pf_qids);
180 #define	V_pf_qids	VNET(pf_qids)
181 static unsigned int	pf_queue_tag_hashsize;
182 #define	PF_QUEUE_TAG_HASH_SIZE_DEFAULT	128
183 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN,
184     &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT,
185     "Size of pf(4) queue tag hashtable");
186 #endif
187 VNET_DEFINE(uma_zone_t,	 pf_tag_z);
188 #define	V_pf_tag_z		 VNET(pf_tag_z)
189 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
190 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
191 
192 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
193 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
194 #endif
195 
196 VNET_DEFINE_STATIC(bool, pf_filter_local) = false;
197 #define V_pf_filter_local	VNET(pf_filter_local)
198 SYSCTL_BOOL(_net_pf, OID_AUTO, filter_local, CTLFLAG_VNET | CTLFLAG_RW,
199     &VNET_NAME(pf_filter_local), false,
200     "Enable filtering for packets delivered to local network stack");
201 
202 static void		 pf_init_tagset(struct pf_tagset *, unsigned int *,
203 			    unsigned int);
204 static void		 pf_cleanup_tagset(struct pf_tagset *);
205 static uint16_t		 tagname2hashindex(const struct pf_tagset *, const char *);
206 static uint16_t		 tag2hashindex(const struct pf_tagset *, uint16_t);
207 static u_int16_t	 tagname2tag(struct pf_tagset *, const char *);
208 static u_int16_t	 pf_tagname2tag(const char *);
209 static void		 tag_unref(struct pf_tagset *, u_int16_t);
210 
211 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
212 
213 struct cdev *pf_dev;
214 
215 /*
216  * XXX - These are new and need to be checked when moveing to a new version
217  */
218 static void		 pf_clear_all_states(void);
219 static unsigned int	 pf_clear_states(const struct pf_kstate_kill *);
220 static void		 pf_killstates(struct pf_kstate_kill *,
221 			    unsigned int *);
222 static int		 pf_killstates_row(struct pf_kstate_kill *,
223 			    struct pf_idhash *);
224 static int		 pf_killstates_nv(struct pfioc_nv *);
225 static int		 pf_clearstates_nv(struct pfioc_nv *);
226 static int		 pf_getstate(struct pfioc_nv *);
227 static int		 pf_getstatus(struct pfioc_nv *);
228 static int		 pf_clear_tables(void);
229 static void		 pf_clear_srcnodes(struct pf_ksrc_node *);
230 static void		 pf_kill_srcnodes(struct pfioc_src_node_kill *);
231 static int		 pf_keepcounters(struct pfioc_nv *);
232 static void		 pf_tbladdr_copyout(struct pf_addr_wrap *);
233 
234 /*
235  * Wrapper functions for pfil(9) hooks
236  */
237 static pfil_return_t pf_eth_check_in(struct mbuf **m, struct ifnet *ifp,
238     int flags, void *ruleset __unused, struct inpcb *inp);
239 static pfil_return_t pf_eth_check_out(struct mbuf **m, struct ifnet *ifp,
240     int flags, void *ruleset __unused, struct inpcb *inp);
241 #ifdef INET
242 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp,
243     int flags, void *ruleset __unused, struct inpcb *inp);
244 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp,
245     int flags, void *ruleset __unused, struct inpcb *inp);
246 #endif
247 #ifdef INET6
248 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp,
249     int flags, void *ruleset __unused, struct inpcb *inp);
250 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp,
251     int flags, void *ruleset __unused, struct inpcb *inp);
252 #endif
253 
254 static void		hook_pf_eth(void);
255 static void		hook_pf(void);
256 static void		dehook_pf_eth(void);
257 static void		dehook_pf(void);
258 static int		shutdown_pf(void);
259 static int		pf_load(void);
260 static void		pf_unload(void);
261 
262 static struct cdevsw pf_cdevsw = {
263 	.d_ioctl =	pfioctl,
264 	.d_name =	PF_NAME,
265 	.d_version =	D_VERSION,
266 };
267 
268 VNET_DEFINE_STATIC(bool, pf_pfil_hooked);
269 #define V_pf_pfil_hooked	VNET(pf_pfil_hooked)
270 VNET_DEFINE_STATIC(bool, pf_pfil_eth_hooked);
271 #define V_pf_pfil_eth_hooked	VNET(pf_pfil_eth_hooked)
272 
273 /*
274  * We need a flag that is neither hooked nor running to know when
275  * the VNET is "valid".  We primarily need this to control (global)
276  * external event, e.g., eventhandlers.
277  */
278 VNET_DEFINE(int, pf_vnet_active);
279 #define V_pf_vnet_active	VNET(pf_vnet_active)
280 
281 int pf_end_threads;
282 struct proc *pf_purge_proc;
283 
284 VNET_DEFINE(struct rmlock, pf_rules_lock);
285 VNET_DEFINE_STATIC(struct sx, pf_ioctl_lock);
286 #define	V_pf_ioctl_lock		VNET(pf_ioctl_lock)
287 struct sx			pf_end_lock;
288 
289 /* pfsync */
290 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
291 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
292 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
293 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
294 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
295 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
296 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
297 
298 /* pflog */
299 pflog_packet_t			*pflog_packet_ptr = NULL;
300 
301 /*
302  * Copy a user-provided string, returning an error if truncation would occur.
303  * Avoid scanning past "sz" bytes in the source string since there's no
304  * guarantee that it's nul-terminated.
305  */
306 static int
307 pf_user_strcpy(char *dst, const char *src, size_t sz)
308 {
309 	if (strnlen(src, sz) == sz)
310 		return (EINVAL);
311 	(void)strlcpy(dst, src, sz);
312 	return (0);
313 }
314 
315 static void
316 pfattach_vnet(void)
317 {
318 	u_int32_t *my_timeout = V_pf_default_rule.timeout;
319 
320 	bzero(&V_pf_status, sizeof(V_pf_status));
321 
322 	pf_initialize();
323 	pfr_initialize();
324 	pfi_initialize_vnet();
325 	pf_normalize_init();
326 	pf_syncookies_init();
327 
328 	V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
329 	V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
330 
331 	RB_INIT(&V_pf_anchors);
332 	pf_init_kruleset(&pf_main_ruleset);
333 
334 	pf_init_keth(V_pf_keth);
335 
336 	/* default rule should never be garbage collected */
337 	V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
338 #ifdef PF_DEFAULT_TO_DROP
339 	V_pf_default_rule.action = PF_DROP;
340 #else
341 	V_pf_default_rule.action = PF_PASS;
342 #endif
343 	V_pf_default_rule.nr = -1;
344 	V_pf_default_rule.rtableid = -1;
345 
346 	pf_counter_u64_init(&V_pf_default_rule.evaluations, M_WAITOK);
347 	for (int i = 0; i < 2; i++) {
348 		pf_counter_u64_init(&V_pf_default_rule.packets[i], M_WAITOK);
349 		pf_counter_u64_init(&V_pf_default_rule.bytes[i], M_WAITOK);
350 	}
351 	V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
352 	V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
353 	V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
354 
355 	V_pf_default_rule.timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
356 	    M_WAITOK | M_ZERO);
357 
358 #ifdef PF_WANT_32_TO_64_COUNTER
359 	V_pf_kifmarker = malloc(sizeof(*V_pf_kifmarker), PFI_MTYPE, M_WAITOK | M_ZERO);
360 	V_pf_rulemarker = malloc(sizeof(*V_pf_rulemarker), M_PFRULE, M_WAITOK | M_ZERO);
361 	PF_RULES_WLOCK();
362 	LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist);
363 	LIST_INSERT_HEAD(&V_pf_allrulelist, &V_pf_default_rule, allrulelist);
364 	V_pf_allrulecount++;
365 	LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist);
366 	PF_RULES_WUNLOCK();
367 #endif
368 
369 	/* initialize default timeouts */
370 	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
371 	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
372 	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
373 	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
374 	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
375 	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
376 	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
377 	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
378 	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
379 	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
380 	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
381 	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
382 	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
383 	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
384 	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
385 	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
386 	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
387 	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
388 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
389 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
390 
391 	V_pf_status.debug = PF_DEBUG_URGENT;
392 	/*
393 	 * XXX This is different than in OpenBSD where reassembly is enabled by
394 	 * defult. In FreeBSD we expect people to still use scrub rules and
395 	 * switch to the new syntax later. Only when they switch they must
396 	 * explicitly enable reassemle. We could change the default once the
397 	 * scrub rule functionality is hopefully removed some day in future.
398 	 */
399 	V_pf_status.reass = 0;
400 
401 	V_pf_pfil_hooked = false;
402 	V_pf_pfil_eth_hooked = false;
403 
404 	/* XXX do our best to avoid a conflict */
405 	V_pf_status.hostid = arc4random();
406 
407 	for (int i = 0; i < PFRES_MAX; i++)
408 		V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
409 	for (int i = 0; i < KLCNT_MAX; i++)
410 		V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
411 	for (int i = 0; i < FCNT_MAX; i++)
412 		pf_counter_u64_init(&V_pf_status.fcounters[i], M_WAITOK);
413 	for (int i = 0; i < SCNT_MAX; i++)
414 		V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
415 
416 	if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET,
417 	    INTR_MPSAFE, &V_pf_swi_cookie) != 0)
418 		/* XXXGL: leaked all above. */
419 		return;
420 }
421 
422 static struct pf_kpool *
423 pf_get_kpool(const char *anchor, u_int32_t ticket, u_int8_t rule_action,
424     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
425     u_int8_t check_ticket)
426 {
427 	struct pf_kruleset	*ruleset;
428 	struct pf_krule		*rule;
429 	int			 rs_num;
430 
431 	ruleset = pf_find_kruleset(anchor);
432 	if (ruleset == NULL)
433 		return (NULL);
434 	rs_num = pf_get_ruleset_number(rule_action);
435 	if (rs_num >= PF_RULESET_MAX)
436 		return (NULL);
437 	if (active) {
438 		if (check_ticket && ticket !=
439 		    ruleset->rules[rs_num].active.ticket)
440 			return (NULL);
441 		if (r_last)
442 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
443 			    pf_krulequeue);
444 		else
445 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
446 	} else {
447 		if (check_ticket && ticket !=
448 		    ruleset->rules[rs_num].inactive.ticket)
449 			return (NULL);
450 		if (r_last)
451 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
452 			    pf_krulequeue);
453 		else
454 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
455 	}
456 	if (!r_last) {
457 		while ((rule != NULL) && (rule->nr != rule_number))
458 			rule = TAILQ_NEXT(rule, entries);
459 	}
460 	if (rule == NULL)
461 		return (NULL);
462 
463 	return (&rule->rpool);
464 }
465 
466 static void
467 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb)
468 {
469 	struct pf_kpooladdr	*mv_pool_pa;
470 
471 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
472 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
473 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
474 	}
475 }
476 
477 static void
478 pf_empty_kpool(struct pf_kpalist *poola)
479 {
480 	struct pf_kpooladdr *pa;
481 
482 	while ((pa = TAILQ_FIRST(poola)) != NULL) {
483 		switch (pa->addr.type) {
484 		case PF_ADDR_DYNIFTL:
485 			pfi_dynaddr_remove(pa->addr.p.dyn);
486 			break;
487 		case PF_ADDR_TABLE:
488 			/* XXX: this could be unfinished pooladdr on pabuf */
489 			if (pa->addr.p.tbl != NULL)
490 				pfr_detach_table(pa->addr.p.tbl);
491 			break;
492 		}
493 		if (pa->kif)
494 			pfi_kkif_unref(pa->kif);
495 		TAILQ_REMOVE(poola, pa, entries);
496 		free(pa, M_PFRULE);
497 	}
498 }
499 
500 static void
501 pf_unlink_rule_locked(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
502 {
503 
504 	PF_RULES_WASSERT();
505 	PF_UNLNKDRULES_ASSERT();
506 
507 	TAILQ_REMOVE(rulequeue, rule, entries);
508 
509 	rule->rule_ref |= PFRULE_REFS;
510 	TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
511 }
512 
513 static void
514 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
515 {
516 
517 	PF_RULES_WASSERT();
518 
519 	PF_UNLNKDRULES_LOCK();
520 	pf_unlink_rule_locked(rulequeue, rule);
521 	PF_UNLNKDRULES_UNLOCK();
522 }
523 
524 static void
525 pf_free_eth_rule(struct pf_keth_rule *rule)
526 {
527 	PF_RULES_WASSERT();
528 
529 	if (rule == NULL)
530 		return;
531 
532 	if (rule->tag)
533 		tag_unref(&V_pf_tags, rule->tag);
534 	if (rule->match_tag)
535 		tag_unref(&V_pf_tags, rule->match_tag);
536 #ifdef ALTQ
537 	pf_qid_unref(rule->qid);
538 #endif
539 
540 	if (rule->bridge_to)
541 		pfi_kkif_unref(rule->bridge_to);
542 	if (rule->kif)
543 		pfi_kkif_unref(rule->kif);
544 
545 	if (rule->ipsrc.addr.type == PF_ADDR_TABLE)
546 		pfr_detach_table(rule->ipsrc.addr.p.tbl);
547 	if (rule->ipdst.addr.type == PF_ADDR_TABLE)
548 		pfr_detach_table(rule->ipdst.addr.p.tbl);
549 
550 	counter_u64_free(rule->evaluations);
551 	for (int i = 0; i < 2; i++) {
552 		counter_u64_free(rule->packets[i]);
553 		counter_u64_free(rule->bytes[i]);
554 	}
555 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
556 	pf_keth_anchor_remove(rule);
557 
558 	free(rule, M_PFRULE);
559 }
560 
561 void
562 pf_free_rule(struct pf_krule *rule)
563 {
564 
565 	PF_RULES_WASSERT();
566 	PF_CONFIG_ASSERT();
567 
568 	if (rule->tag)
569 		tag_unref(&V_pf_tags, rule->tag);
570 	if (rule->match_tag)
571 		tag_unref(&V_pf_tags, rule->match_tag);
572 #ifdef ALTQ
573 	if (rule->pqid != rule->qid)
574 		pf_qid_unref(rule->pqid);
575 	pf_qid_unref(rule->qid);
576 #endif
577 	switch (rule->src.addr.type) {
578 	case PF_ADDR_DYNIFTL:
579 		pfi_dynaddr_remove(rule->src.addr.p.dyn);
580 		break;
581 	case PF_ADDR_TABLE:
582 		pfr_detach_table(rule->src.addr.p.tbl);
583 		break;
584 	}
585 	switch (rule->dst.addr.type) {
586 	case PF_ADDR_DYNIFTL:
587 		pfi_dynaddr_remove(rule->dst.addr.p.dyn);
588 		break;
589 	case PF_ADDR_TABLE:
590 		pfr_detach_table(rule->dst.addr.p.tbl);
591 		break;
592 	}
593 	if (rule->overload_tbl)
594 		pfr_detach_table(rule->overload_tbl);
595 	if (rule->kif)
596 		pfi_kkif_unref(rule->kif);
597 	pf_kanchor_remove(rule);
598 	pf_empty_kpool(&rule->rpool.list);
599 
600 	pf_krule_free(rule);
601 }
602 
603 static void
604 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
605     unsigned int default_size)
606 {
607 	unsigned int i;
608 	unsigned int hashsize;
609 
610 	if (*tunable_size == 0 || !powerof2(*tunable_size))
611 		*tunable_size = default_size;
612 
613 	hashsize = *tunable_size;
614 	ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
615 	    M_WAITOK);
616 	ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
617 	    M_WAITOK);
618 	ts->mask = hashsize - 1;
619 	ts->seed = arc4random();
620 	for (i = 0; i < hashsize; i++) {
621 		TAILQ_INIT(&ts->namehash[i]);
622 		TAILQ_INIT(&ts->taghash[i]);
623 	}
624 	BIT_FILL(TAGID_MAX, &ts->avail);
625 }
626 
627 static void
628 pf_cleanup_tagset(struct pf_tagset *ts)
629 {
630 	unsigned int i;
631 	unsigned int hashsize;
632 	struct pf_tagname *t, *tmp;
633 
634 	/*
635 	 * Only need to clean up one of the hashes as each tag is hashed
636 	 * into each table.
637 	 */
638 	hashsize = ts->mask + 1;
639 	for (i = 0; i < hashsize; i++)
640 		TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
641 			uma_zfree(V_pf_tag_z, t);
642 
643 	free(ts->namehash, M_PFHASH);
644 	free(ts->taghash, M_PFHASH);
645 }
646 
647 static uint16_t
648 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
649 {
650 	size_t len;
651 
652 	len = strnlen(tagname, PF_TAG_NAME_SIZE - 1);
653 	return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask);
654 }
655 
656 static uint16_t
657 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
658 {
659 
660 	return (tag & ts->mask);
661 }
662 
663 static u_int16_t
664 tagname2tag(struct pf_tagset *ts, const char *tagname)
665 {
666 	struct pf_tagname	*tag;
667 	u_int32_t		 index;
668 	u_int16_t		 new_tagid;
669 
670 	PF_RULES_WASSERT();
671 
672 	index = tagname2hashindex(ts, tagname);
673 	TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
674 		if (strcmp(tagname, tag->name) == 0) {
675 			tag->ref++;
676 			return (tag->tag);
677 		}
678 
679 	/*
680 	 * new entry
681 	 *
682 	 * to avoid fragmentation, we do a linear search from the beginning
683 	 * and take the first free slot we find.
684 	 */
685 	new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
686 	/*
687 	 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
688 	 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
689 	 * set.  It may also return a bit number greater than TAGID_MAX due
690 	 * to rounding of the number of bits in the vector up to a multiple
691 	 * of the vector word size at declaration/allocation time.
692 	 */
693 	if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
694 		return (0);
695 
696 	/* Mark the tag as in use.  Bits are 0-based for BIT_CLR() */
697 	BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
698 
699 	/* allocate and fill new struct pf_tagname */
700 	tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
701 	if (tag == NULL)
702 		return (0);
703 	strlcpy(tag->name, tagname, sizeof(tag->name));
704 	tag->tag = new_tagid;
705 	tag->ref = 1;
706 
707 	/* Insert into namehash */
708 	TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
709 
710 	/* Insert into taghash */
711 	index = tag2hashindex(ts, new_tagid);
712 	TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
713 
714 	return (tag->tag);
715 }
716 
717 static void
718 tag_unref(struct pf_tagset *ts, u_int16_t tag)
719 {
720 	struct pf_tagname	*t;
721 	uint16_t		 index;
722 
723 	PF_RULES_WASSERT();
724 
725 	index = tag2hashindex(ts, tag);
726 	TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
727 		if (tag == t->tag) {
728 			if (--t->ref == 0) {
729 				TAILQ_REMOVE(&ts->taghash[index], t,
730 				    taghash_entries);
731 				index = tagname2hashindex(ts, t->name);
732 				TAILQ_REMOVE(&ts->namehash[index], t,
733 				    namehash_entries);
734 				/* Bits are 0-based for BIT_SET() */
735 				BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
736 				uma_zfree(V_pf_tag_z, t);
737 			}
738 			break;
739 		}
740 }
741 
742 static uint16_t
743 pf_tagname2tag(const char *tagname)
744 {
745 	return (tagname2tag(&V_pf_tags, tagname));
746 }
747 
748 static int
749 pf_begin_eth(uint32_t *ticket, const char *anchor)
750 {
751 	struct pf_keth_rule *rule, *tmp;
752 	struct pf_keth_ruleset *rs;
753 
754 	PF_RULES_WASSERT();
755 
756 	rs = pf_find_or_create_keth_ruleset(anchor);
757 	if (rs == NULL)
758 		return (EINVAL);
759 
760 	/* Purge old inactive rules. */
761 	TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
762 	    tmp) {
763 		TAILQ_REMOVE(rs->inactive.rules, rule,
764 		    entries);
765 		pf_free_eth_rule(rule);
766 	}
767 
768 	*ticket = ++rs->inactive.ticket;
769 	rs->inactive.open = 1;
770 
771 	return (0);
772 }
773 
774 static void
775 pf_rollback_eth_cb(struct epoch_context *ctx)
776 {
777 	struct pf_keth_ruleset *rs;
778 
779 	rs = __containerof(ctx, struct pf_keth_ruleset, epoch_ctx);
780 
781 	CURVNET_SET(rs->vnet);
782 
783 	PF_RULES_WLOCK();
784 	pf_rollback_eth(rs->inactive.ticket,
785 	    rs->anchor ? rs->anchor->path : "");
786 	PF_RULES_WUNLOCK();
787 
788 	CURVNET_RESTORE();
789 }
790 
791 static int
792 pf_rollback_eth(uint32_t ticket, const char *anchor)
793 {
794 	struct pf_keth_rule *rule, *tmp;
795 	struct pf_keth_ruleset *rs;
796 
797 	PF_RULES_WASSERT();
798 
799 	rs = pf_find_keth_ruleset(anchor);
800 	if (rs == NULL)
801 		return (EINVAL);
802 
803 	if (!rs->inactive.open ||
804 	    ticket != rs->inactive.ticket)
805 		return (0);
806 
807 	/* Purge old inactive rules. */
808 	TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
809 	    tmp) {
810 		TAILQ_REMOVE(rs->inactive.rules, rule, entries);
811 		pf_free_eth_rule(rule);
812 	}
813 
814 	rs->inactive.open = 0;
815 
816 	pf_remove_if_empty_keth_ruleset(rs);
817 
818 	return (0);
819 }
820 
821 #define	PF_SET_SKIP_STEPS(i)					\
822 	do {							\
823 		while (head[i] != cur) {			\
824 			head[i]->skip[i].ptr = cur;		\
825 			head[i] = TAILQ_NEXT(head[i], entries);	\
826 		}						\
827 	} while (0)
828 
829 static void
830 pf_eth_calc_skip_steps(struct pf_keth_ruleq *rules)
831 {
832 	struct pf_keth_rule *cur, *prev, *head[PFE_SKIP_COUNT];
833 	int i;
834 
835 	cur = TAILQ_FIRST(rules);
836 	prev = cur;
837 	for (i = 0; i < PFE_SKIP_COUNT; ++i)
838 		head[i] = cur;
839 	while (cur != NULL) {
840 		if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
841 			PF_SET_SKIP_STEPS(PFE_SKIP_IFP);
842 		if (cur->direction != prev->direction)
843 			PF_SET_SKIP_STEPS(PFE_SKIP_DIR);
844 		if (cur->proto != prev->proto)
845 			PF_SET_SKIP_STEPS(PFE_SKIP_PROTO);
846 		if (memcmp(&cur->src, &prev->src, sizeof(cur->src)) != 0)
847 			PF_SET_SKIP_STEPS(PFE_SKIP_SRC_ADDR);
848 		if (memcmp(&cur->dst, &prev->dst, sizeof(cur->dst)) != 0)
849 			PF_SET_SKIP_STEPS(PFE_SKIP_DST_ADDR);
850 		if (cur->ipsrc.neg != prev->ipsrc.neg ||
851 		    pf_addr_wrap_neq(&cur->ipsrc.addr, &prev->ipsrc.addr))
852 			PF_SET_SKIP_STEPS(PFE_SKIP_SRC_IP_ADDR);
853 		if (cur->ipdst.neg != prev->ipdst.neg ||
854 		    pf_addr_wrap_neq(&cur->ipdst.addr, &prev->ipdst.addr))
855 			PF_SET_SKIP_STEPS(PFE_SKIP_DST_IP_ADDR);
856 
857 		prev = cur;
858 		cur = TAILQ_NEXT(cur, entries);
859 	}
860 	for (i = 0; i < PFE_SKIP_COUNT; ++i)
861 		PF_SET_SKIP_STEPS(i);
862 }
863 
864 static int
865 pf_commit_eth(uint32_t ticket, const char *anchor)
866 {
867 	struct pf_keth_ruleq *rules;
868 	struct pf_keth_ruleset *rs;
869 
870 	rs = pf_find_keth_ruleset(anchor);
871 	if (rs == NULL) {
872 		return (EINVAL);
873 	}
874 
875 	if (!rs->inactive.open ||
876 	    ticket != rs->inactive.ticket)
877 		return (EBUSY);
878 
879 	PF_RULES_WASSERT();
880 
881 	pf_eth_calc_skip_steps(rs->inactive.rules);
882 
883 	rules = rs->active.rules;
884 	ck_pr_store_ptr(&rs->active.rules, rs->inactive.rules);
885 	rs->inactive.rules = rules;
886 	rs->inactive.ticket = rs->active.ticket;
887 
888 	/* Clean up inactive rules (i.e. previously active rules), only when
889 	 * we're sure they're no longer used. */
890 	NET_EPOCH_CALL(pf_rollback_eth_cb, &rs->epoch_ctx);
891 
892 	return (0);
893 }
894 
895 #ifdef ALTQ
896 static uint16_t
897 pf_qname2qid(const char *qname)
898 {
899 	return (tagname2tag(&V_pf_qids, qname));
900 }
901 
902 static void
903 pf_qid_unref(uint16_t qid)
904 {
905 	tag_unref(&V_pf_qids, qid);
906 }
907 
908 static int
909 pf_begin_altq(u_int32_t *ticket)
910 {
911 	struct pf_altq	*altq, *tmp;
912 	int		 error = 0;
913 
914 	PF_RULES_WASSERT();
915 
916 	/* Purge the old altq lists */
917 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
918 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
919 			/* detach and destroy the discipline */
920 			error = altq_remove(altq);
921 		}
922 		free(altq, M_PFALTQ);
923 	}
924 	TAILQ_INIT(V_pf_altq_ifs_inactive);
925 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
926 		pf_qid_unref(altq->qid);
927 		free(altq, M_PFALTQ);
928 	}
929 	TAILQ_INIT(V_pf_altqs_inactive);
930 	if (error)
931 		return (error);
932 	*ticket = ++V_ticket_altqs_inactive;
933 	V_altqs_inactive_open = 1;
934 	return (0);
935 }
936 
937 static int
938 pf_rollback_altq(u_int32_t ticket)
939 {
940 	struct pf_altq	*altq, *tmp;
941 	int		 error = 0;
942 
943 	PF_RULES_WASSERT();
944 
945 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
946 		return (0);
947 	/* Purge the old altq lists */
948 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
949 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
950 			/* detach and destroy the discipline */
951 			error = altq_remove(altq);
952 		}
953 		free(altq, M_PFALTQ);
954 	}
955 	TAILQ_INIT(V_pf_altq_ifs_inactive);
956 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
957 		pf_qid_unref(altq->qid);
958 		free(altq, M_PFALTQ);
959 	}
960 	TAILQ_INIT(V_pf_altqs_inactive);
961 	V_altqs_inactive_open = 0;
962 	return (error);
963 }
964 
965 static int
966 pf_commit_altq(u_int32_t ticket)
967 {
968 	struct pf_altqqueue	*old_altqs, *old_altq_ifs;
969 	struct pf_altq		*altq, *tmp;
970 	int			 err, error = 0;
971 
972 	PF_RULES_WASSERT();
973 
974 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
975 		return (EBUSY);
976 
977 	/* swap altqs, keep the old. */
978 	old_altqs = V_pf_altqs_active;
979 	old_altq_ifs = V_pf_altq_ifs_active;
980 	V_pf_altqs_active = V_pf_altqs_inactive;
981 	V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
982 	V_pf_altqs_inactive = old_altqs;
983 	V_pf_altq_ifs_inactive = old_altq_ifs;
984 	V_ticket_altqs_active = V_ticket_altqs_inactive;
985 
986 	/* Attach new disciplines */
987 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
988 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
989 			/* attach the discipline */
990 			error = altq_pfattach(altq);
991 			if (error == 0 && V_pf_altq_running)
992 				error = pf_enable_altq(altq);
993 			if (error != 0)
994 				return (error);
995 		}
996 	}
997 
998 	/* Purge the old altq lists */
999 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
1000 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
1001 			/* detach and destroy the discipline */
1002 			if (V_pf_altq_running)
1003 				error = pf_disable_altq(altq);
1004 			err = altq_pfdetach(altq);
1005 			if (err != 0 && error == 0)
1006 				error = err;
1007 			err = altq_remove(altq);
1008 			if (err != 0 && error == 0)
1009 				error = err;
1010 		}
1011 		free(altq, M_PFALTQ);
1012 	}
1013 	TAILQ_INIT(V_pf_altq_ifs_inactive);
1014 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
1015 		pf_qid_unref(altq->qid);
1016 		free(altq, M_PFALTQ);
1017 	}
1018 	TAILQ_INIT(V_pf_altqs_inactive);
1019 
1020 	V_altqs_inactive_open = 0;
1021 	return (error);
1022 }
1023 
1024 static int
1025 pf_enable_altq(struct pf_altq *altq)
1026 {
1027 	struct ifnet		*ifp;
1028 	struct tb_profile	 tb;
1029 	int			 error = 0;
1030 
1031 	if ((ifp = ifunit(altq->ifname)) == NULL)
1032 		return (EINVAL);
1033 
1034 	if (ifp->if_snd.altq_type != ALTQT_NONE)
1035 		error = altq_enable(&ifp->if_snd);
1036 
1037 	/* set tokenbucket regulator */
1038 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
1039 		tb.rate = altq->ifbandwidth;
1040 		tb.depth = altq->tbrsize;
1041 		error = tbr_set(&ifp->if_snd, &tb);
1042 	}
1043 
1044 	return (error);
1045 }
1046 
1047 static int
1048 pf_disable_altq(struct pf_altq *altq)
1049 {
1050 	struct ifnet		*ifp;
1051 	struct tb_profile	 tb;
1052 	int			 error;
1053 
1054 	if ((ifp = ifunit(altq->ifname)) == NULL)
1055 		return (EINVAL);
1056 
1057 	/*
1058 	 * when the discipline is no longer referenced, it was overridden
1059 	 * by a new one.  if so, just return.
1060 	 */
1061 	if (altq->altq_disc != ifp->if_snd.altq_disc)
1062 		return (0);
1063 
1064 	error = altq_disable(&ifp->if_snd);
1065 
1066 	if (error == 0) {
1067 		/* clear tokenbucket regulator */
1068 		tb.rate = 0;
1069 		error = tbr_set(&ifp->if_snd, &tb);
1070 	}
1071 
1072 	return (error);
1073 }
1074 
1075 static int
1076 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
1077     struct pf_altq *altq)
1078 {
1079 	struct ifnet	*ifp1;
1080 	int		 error = 0;
1081 
1082 	/* Deactivate the interface in question */
1083 	altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
1084 	if ((ifp1 = ifunit(altq->ifname)) == NULL ||
1085 	    (remove && ifp1 == ifp)) {
1086 		altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
1087 	} else {
1088 		error = altq_add(ifp1, altq);
1089 
1090 		if (ticket != V_ticket_altqs_inactive)
1091 			error = EBUSY;
1092 
1093 		if (error)
1094 			free(altq, M_PFALTQ);
1095 	}
1096 
1097 	return (error);
1098 }
1099 
1100 void
1101 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
1102 {
1103 	struct pf_altq	*a1, *a2, *a3;
1104 	u_int32_t	 ticket;
1105 	int		 error = 0;
1106 
1107 	/*
1108 	 * No need to re-evaluate the configuration for events on interfaces
1109 	 * that do not support ALTQ, as it's not possible for such
1110 	 * interfaces to be part of the configuration.
1111 	 */
1112 	if (!ALTQ_IS_READY(&ifp->if_snd))
1113 		return;
1114 
1115 	/* Interrupt userland queue modifications */
1116 	if (V_altqs_inactive_open)
1117 		pf_rollback_altq(V_ticket_altqs_inactive);
1118 
1119 	/* Start new altq ruleset */
1120 	if (pf_begin_altq(&ticket))
1121 		return;
1122 
1123 	/* Copy the current active set */
1124 	TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
1125 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1126 		if (a2 == NULL) {
1127 			error = ENOMEM;
1128 			break;
1129 		}
1130 		bcopy(a1, a2, sizeof(struct pf_altq));
1131 
1132 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1133 		if (error)
1134 			break;
1135 
1136 		TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
1137 	}
1138 	if (error)
1139 		goto out;
1140 	TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
1141 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1142 		if (a2 == NULL) {
1143 			error = ENOMEM;
1144 			break;
1145 		}
1146 		bcopy(a1, a2, sizeof(struct pf_altq));
1147 
1148 		if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
1149 			error = EBUSY;
1150 			free(a2, M_PFALTQ);
1151 			break;
1152 		}
1153 		a2->altq_disc = NULL;
1154 		TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
1155 			if (strncmp(a3->ifname, a2->ifname,
1156 				IFNAMSIZ) == 0) {
1157 				a2->altq_disc = a3->altq_disc;
1158 				break;
1159 			}
1160 		}
1161 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1162 		if (error)
1163 			break;
1164 
1165 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
1166 	}
1167 
1168 out:
1169 	if (error != 0)
1170 		pf_rollback_altq(ticket);
1171 	else
1172 		pf_commit_altq(ticket);
1173 }
1174 #endif /* ALTQ */
1175 
1176 static struct pf_krule_global *
1177 pf_rule_tree_alloc(int flags)
1178 {
1179 	struct pf_krule_global *tree;
1180 
1181 	tree = malloc(sizeof(struct pf_krule_global), M_TEMP, flags);
1182 	if (tree == NULL)
1183 		return (NULL);
1184 	RB_INIT(tree);
1185 	return (tree);
1186 }
1187 
1188 static void
1189 pf_rule_tree_free(struct pf_krule_global *tree)
1190 {
1191 
1192 	free(tree, M_TEMP);
1193 }
1194 
1195 static int
1196 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
1197 {
1198 	struct pf_krule_global *tree;
1199 	struct pf_kruleset	*rs;
1200 	struct pf_krule		*rule;
1201 
1202 	PF_RULES_WASSERT();
1203 
1204 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1205 		return (EINVAL);
1206 	tree = pf_rule_tree_alloc(M_NOWAIT);
1207 	if (tree == NULL)
1208 		return (ENOMEM);
1209 	rs = pf_find_or_create_kruleset(anchor);
1210 	if (rs == NULL) {
1211 		free(tree, M_TEMP);
1212 		return (EINVAL);
1213 	}
1214 	pf_rule_tree_free(rs->rules[rs_num].inactive.tree);
1215 	rs->rules[rs_num].inactive.tree = tree;
1216 
1217 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1218 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1219 		rs->rules[rs_num].inactive.rcount--;
1220 	}
1221 	*ticket = ++rs->rules[rs_num].inactive.ticket;
1222 	rs->rules[rs_num].inactive.open = 1;
1223 	return (0);
1224 }
1225 
1226 static int
1227 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
1228 {
1229 	struct pf_kruleset	*rs;
1230 	struct pf_krule		*rule;
1231 
1232 	PF_RULES_WASSERT();
1233 
1234 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1235 		return (EINVAL);
1236 	rs = pf_find_kruleset(anchor);
1237 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1238 	    rs->rules[rs_num].inactive.ticket != ticket)
1239 		return (0);
1240 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1241 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1242 		rs->rules[rs_num].inactive.rcount--;
1243 	}
1244 	rs->rules[rs_num].inactive.open = 0;
1245 	return (0);
1246 }
1247 
1248 #define PF_MD5_UPD(st, elm)						\
1249 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
1250 
1251 #define PF_MD5_UPD_STR(st, elm)						\
1252 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
1253 
1254 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
1255 		(stor) = htonl((st)->elm);				\
1256 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
1257 } while (0)
1258 
1259 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
1260 		(stor) = htons((st)->elm);				\
1261 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
1262 } while (0)
1263 
1264 static void
1265 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
1266 {
1267 	PF_MD5_UPD(pfr, addr.type);
1268 	switch (pfr->addr.type) {
1269 		case PF_ADDR_DYNIFTL:
1270 			PF_MD5_UPD(pfr, addr.v.ifname);
1271 			PF_MD5_UPD(pfr, addr.iflags);
1272 			break;
1273 		case PF_ADDR_TABLE:
1274 			PF_MD5_UPD(pfr, addr.v.tblname);
1275 			break;
1276 		case PF_ADDR_ADDRMASK:
1277 			/* XXX ignore af? */
1278 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
1279 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
1280 			break;
1281 	}
1282 
1283 	PF_MD5_UPD(pfr, port[0]);
1284 	PF_MD5_UPD(pfr, port[1]);
1285 	PF_MD5_UPD(pfr, neg);
1286 	PF_MD5_UPD(pfr, port_op);
1287 }
1288 
1289 static void
1290 pf_hash_rule_rolling(MD5_CTX *ctx, struct pf_krule *rule)
1291 {
1292 	u_int16_t x;
1293 	u_int32_t y;
1294 
1295 	pf_hash_rule_addr(ctx, &rule->src);
1296 	pf_hash_rule_addr(ctx, &rule->dst);
1297 	for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++)
1298 		PF_MD5_UPD_STR(rule, label[i]);
1299 	PF_MD5_UPD_STR(rule, ifname);
1300 	PF_MD5_UPD_STR(rule, match_tagname);
1301 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
1302 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
1303 	PF_MD5_UPD_HTONL(rule, prob, y);
1304 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
1305 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1306 	PF_MD5_UPD(rule, uid.op);
1307 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1308 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1309 	PF_MD5_UPD(rule, gid.op);
1310 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
1311 	PF_MD5_UPD(rule, action);
1312 	PF_MD5_UPD(rule, direction);
1313 	PF_MD5_UPD(rule, af);
1314 	PF_MD5_UPD(rule, quick);
1315 	PF_MD5_UPD(rule, ifnot);
1316 	PF_MD5_UPD(rule, match_tag_not);
1317 	PF_MD5_UPD(rule, natpass);
1318 	PF_MD5_UPD(rule, keep_state);
1319 	PF_MD5_UPD(rule, proto);
1320 	PF_MD5_UPD(rule, type);
1321 	PF_MD5_UPD(rule, code);
1322 	PF_MD5_UPD(rule, flags);
1323 	PF_MD5_UPD(rule, flagset);
1324 	PF_MD5_UPD(rule, allow_opts);
1325 	PF_MD5_UPD(rule, rt);
1326 	PF_MD5_UPD(rule, tos);
1327 	PF_MD5_UPD(rule, scrub_flags);
1328 	PF_MD5_UPD(rule, min_ttl);
1329 	PF_MD5_UPD(rule, set_tos);
1330 	if (rule->anchor != NULL)
1331 		PF_MD5_UPD_STR(rule, anchor->path);
1332 }
1333 
1334 static void
1335 pf_hash_rule(struct pf_krule *rule)
1336 {
1337 	MD5_CTX		ctx;
1338 
1339 	MD5Init(&ctx);
1340 	pf_hash_rule_rolling(&ctx, rule);
1341 	MD5Final(rule->md5sum, &ctx);
1342 }
1343 
1344 static int
1345 pf_krule_compare(struct pf_krule *a, struct pf_krule *b)
1346 {
1347 
1348 	return (memcmp(a->md5sum, b->md5sum, PF_MD5_DIGEST_LENGTH));
1349 }
1350 
1351 static int
1352 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1353 {
1354 	struct pf_kruleset	*rs;
1355 	struct pf_krule		*rule, **old_array, *old_rule;
1356 	struct pf_krulequeue	*old_rules;
1357 	struct pf_krule_global  *old_tree;
1358 	int			 error;
1359 	u_int32_t		 old_rcount;
1360 
1361 	PF_RULES_WASSERT();
1362 
1363 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1364 		return (EINVAL);
1365 	rs = pf_find_kruleset(anchor);
1366 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1367 	    ticket != rs->rules[rs_num].inactive.ticket)
1368 		return (EBUSY);
1369 
1370 	/* Calculate checksum for the main ruleset */
1371 	if (rs == &pf_main_ruleset) {
1372 		error = pf_setup_pfsync_matching(rs);
1373 		if (error != 0)
1374 			return (error);
1375 	}
1376 
1377 	/* Swap rules, keep the old. */
1378 	old_rules = rs->rules[rs_num].active.ptr;
1379 	old_rcount = rs->rules[rs_num].active.rcount;
1380 	old_array = rs->rules[rs_num].active.ptr_array;
1381 	old_tree = rs->rules[rs_num].active.tree;
1382 
1383 	rs->rules[rs_num].active.ptr =
1384 	    rs->rules[rs_num].inactive.ptr;
1385 	rs->rules[rs_num].active.ptr_array =
1386 	    rs->rules[rs_num].inactive.ptr_array;
1387 	rs->rules[rs_num].active.tree =
1388 	    rs->rules[rs_num].inactive.tree;
1389 	rs->rules[rs_num].active.rcount =
1390 	    rs->rules[rs_num].inactive.rcount;
1391 
1392 	/* Attempt to preserve counter information. */
1393 	if (V_pf_status.keep_counters && old_tree != NULL) {
1394 		TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr,
1395 		    entries) {
1396 			old_rule = RB_FIND(pf_krule_global, old_tree, rule);
1397 			if (old_rule == NULL) {
1398 				continue;
1399 			}
1400 			pf_counter_u64_critical_enter();
1401 			pf_counter_u64_add_protected(&rule->evaluations,
1402 			    pf_counter_u64_fetch(&old_rule->evaluations));
1403 			pf_counter_u64_add_protected(&rule->packets[0],
1404 			    pf_counter_u64_fetch(&old_rule->packets[0]));
1405 			pf_counter_u64_add_protected(&rule->packets[1],
1406 			    pf_counter_u64_fetch(&old_rule->packets[1]));
1407 			pf_counter_u64_add_protected(&rule->bytes[0],
1408 			    pf_counter_u64_fetch(&old_rule->bytes[0]));
1409 			pf_counter_u64_add_protected(&rule->bytes[1],
1410 			    pf_counter_u64_fetch(&old_rule->bytes[1]));
1411 			pf_counter_u64_critical_exit();
1412 		}
1413 	}
1414 
1415 	rs->rules[rs_num].inactive.ptr = old_rules;
1416 	rs->rules[rs_num].inactive.ptr_array = old_array;
1417 	rs->rules[rs_num].inactive.tree = NULL; /* important for pf_ioctl_addrule */
1418 	rs->rules[rs_num].inactive.rcount = old_rcount;
1419 
1420 	rs->rules[rs_num].active.ticket =
1421 	    rs->rules[rs_num].inactive.ticket;
1422 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1423 
1424 	/* Purge the old rule list. */
1425 	PF_UNLNKDRULES_LOCK();
1426 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1427 		pf_unlink_rule_locked(old_rules, rule);
1428 	PF_UNLNKDRULES_UNLOCK();
1429 	if (rs->rules[rs_num].inactive.ptr_array)
1430 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1431 	rs->rules[rs_num].inactive.ptr_array = NULL;
1432 	rs->rules[rs_num].inactive.rcount = 0;
1433 	rs->rules[rs_num].inactive.open = 0;
1434 	pf_remove_if_empty_kruleset(rs);
1435 	free(old_tree, M_TEMP);
1436 
1437 	return (0);
1438 }
1439 
1440 static int
1441 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1442 {
1443 	MD5_CTX			 ctx;
1444 	struct pf_krule		*rule;
1445 	int			 rs_cnt;
1446 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
1447 
1448 	MD5Init(&ctx);
1449 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1450 		/* XXX PF_RULESET_SCRUB as well? */
1451 		if (rs_cnt == PF_RULESET_SCRUB)
1452 			continue;
1453 
1454 		if (rs->rules[rs_cnt].inactive.ptr_array)
1455 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1456 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
1457 
1458 		if (rs->rules[rs_cnt].inactive.rcount) {
1459 			rs->rules[rs_cnt].inactive.ptr_array =
1460 			    mallocarray(rs->rules[rs_cnt].inactive.rcount,
1461 			    sizeof(struct pf_rule **),
1462 			    M_TEMP, M_NOWAIT);
1463 
1464 			if (!rs->rules[rs_cnt].inactive.ptr_array)
1465 				return (ENOMEM);
1466 		}
1467 
1468 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1469 		    entries) {
1470 			pf_hash_rule_rolling(&ctx, rule);
1471 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1472 		}
1473 	}
1474 
1475 	MD5Final(digest, &ctx);
1476 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1477 	return (0);
1478 }
1479 
1480 static int
1481 pf_eth_addr_setup(struct pf_keth_ruleset *ruleset, struct pf_addr_wrap *addr)
1482 {
1483 	int error = 0;
1484 
1485 	switch (addr->type) {
1486 	case PF_ADDR_TABLE:
1487 		addr->p.tbl = pfr_eth_attach_table(ruleset, addr->v.tblname);
1488 		if (addr->p.tbl == NULL)
1489 			error = ENOMEM;
1490 		break;
1491 	default:
1492 		error = EINVAL;
1493 	}
1494 
1495 	return (error);
1496 }
1497 
1498 static int
1499 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1500     sa_family_t af)
1501 {
1502 	int error = 0;
1503 
1504 	switch (addr->type) {
1505 	case PF_ADDR_TABLE:
1506 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1507 		if (addr->p.tbl == NULL)
1508 			error = ENOMEM;
1509 		break;
1510 	case PF_ADDR_DYNIFTL:
1511 		error = pfi_dynaddr_setup(addr, af);
1512 		break;
1513 	}
1514 
1515 	return (error);
1516 }
1517 
1518 static void
1519 pf_addr_copyout(struct pf_addr_wrap *addr)
1520 {
1521 
1522 	switch (addr->type) {
1523 	case PF_ADDR_DYNIFTL:
1524 		pfi_dynaddr_copyout(addr);
1525 		break;
1526 	case PF_ADDR_TABLE:
1527 		pf_tbladdr_copyout(addr);
1528 		break;
1529 	}
1530 }
1531 
1532 static void
1533 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
1534 {
1535 	int	secs = time_uptime, diff;
1536 
1537 	bzero(out, sizeof(struct pf_src_node));
1538 
1539 	bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
1540 	bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
1541 
1542 	if (in->rule.ptr != NULL)
1543 		out->rule.nr = in->rule.ptr->nr;
1544 
1545 	for (int i = 0; i < 2; i++) {
1546 		out->bytes[i] = counter_u64_fetch(in->bytes[i]);
1547 		out->packets[i] = counter_u64_fetch(in->packets[i]);
1548 	}
1549 
1550 	out->states = in->states;
1551 	out->conn = in->conn;
1552 	out->af = in->af;
1553 	out->ruletype = in->ruletype;
1554 
1555 	out->creation = secs - in->creation;
1556 	if (out->expire > secs)
1557 		out->expire -= secs;
1558 	else
1559 		out->expire = 0;
1560 
1561 	/* Adjust the connection rate estimate. */
1562 	diff = secs - in->conn_rate.last;
1563 	if (diff >= in->conn_rate.seconds)
1564 		out->conn_rate.count = 0;
1565 	else
1566 		out->conn_rate.count -=
1567 		    in->conn_rate.count * diff /
1568 		    in->conn_rate.seconds;
1569 }
1570 
1571 #ifdef ALTQ
1572 /*
1573  * Handle export of struct pf_kaltq to user binaries that may be using any
1574  * version of struct pf_altq.
1575  */
1576 static int
1577 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1578 {
1579 	u_int32_t version;
1580 
1581 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1582 		version = 0;
1583 	else
1584 		version = pa->version;
1585 
1586 	if (version > PFIOC_ALTQ_VERSION)
1587 		return (EINVAL);
1588 
1589 #define ASSIGN(x) exported_q->x = q->x
1590 #define COPY(x) \
1591 	bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1592 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1593 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1594 
1595 	switch (version) {
1596 	case 0: {
1597 		struct pf_altq_v0 *exported_q =
1598 		    &((struct pfioc_altq_v0 *)pa)->altq;
1599 
1600 		COPY(ifname);
1601 
1602 		ASSIGN(scheduler);
1603 		ASSIGN(tbrsize);
1604 		exported_q->tbrsize = SATU16(q->tbrsize);
1605 		exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1606 
1607 		COPY(qname);
1608 		COPY(parent);
1609 		ASSIGN(parent_qid);
1610 		exported_q->bandwidth = SATU32(q->bandwidth);
1611 		ASSIGN(priority);
1612 		ASSIGN(local_flags);
1613 
1614 		ASSIGN(qlimit);
1615 		ASSIGN(flags);
1616 
1617 		if (q->scheduler == ALTQT_HFSC) {
1618 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1619 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1620 			    SATU32(q->pq_u.hfsc_opts.x)
1621 
1622 			ASSIGN_OPT_SATU32(rtsc_m1);
1623 			ASSIGN_OPT(rtsc_d);
1624 			ASSIGN_OPT_SATU32(rtsc_m2);
1625 
1626 			ASSIGN_OPT_SATU32(lssc_m1);
1627 			ASSIGN_OPT(lssc_d);
1628 			ASSIGN_OPT_SATU32(lssc_m2);
1629 
1630 			ASSIGN_OPT_SATU32(ulsc_m1);
1631 			ASSIGN_OPT(ulsc_d);
1632 			ASSIGN_OPT_SATU32(ulsc_m2);
1633 
1634 			ASSIGN_OPT(flags);
1635 
1636 #undef ASSIGN_OPT
1637 #undef ASSIGN_OPT_SATU32
1638 		} else
1639 			COPY(pq_u);
1640 
1641 		ASSIGN(qid);
1642 		break;
1643 	}
1644 	case 1:	{
1645 		struct pf_altq_v1 *exported_q =
1646 		    &((struct pfioc_altq_v1 *)pa)->altq;
1647 
1648 		COPY(ifname);
1649 
1650 		ASSIGN(scheduler);
1651 		ASSIGN(tbrsize);
1652 		ASSIGN(ifbandwidth);
1653 
1654 		COPY(qname);
1655 		COPY(parent);
1656 		ASSIGN(parent_qid);
1657 		ASSIGN(bandwidth);
1658 		ASSIGN(priority);
1659 		ASSIGN(local_flags);
1660 
1661 		ASSIGN(qlimit);
1662 		ASSIGN(flags);
1663 		COPY(pq_u);
1664 
1665 		ASSIGN(qid);
1666 		break;
1667 	}
1668 	default:
1669 		panic("%s: unhandled struct pfioc_altq version", __func__);
1670 		break;
1671 	}
1672 
1673 #undef ASSIGN
1674 #undef COPY
1675 #undef SATU16
1676 #undef SATU32
1677 
1678 	return (0);
1679 }
1680 
1681 /*
1682  * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1683  * that may be using any version of it.
1684  */
1685 static int
1686 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1687 {
1688 	u_int32_t version;
1689 
1690 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1691 		version = 0;
1692 	else
1693 		version = pa->version;
1694 
1695 	if (version > PFIOC_ALTQ_VERSION)
1696 		return (EINVAL);
1697 
1698 #define ASSIGN(x) q->x = imported_q->x
1699 #define COPY(x) \
1700 	bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1701 
1702 	switch (version) {
1703 	case 0: {
1704 		struct pf_altq_v0 *imported_q =
1705 		    &((struct pfioc_altq_v0 *)pa)->altq;
1706 
1707 		COPY(ifname);
1708 
1709 		ASSIGN(scheduler);
1710 		ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1711 		ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1712 
1713 		COPY(qname);
1714 		COPY(parent);
1715 		ASSIGN(parent_qid);
1716 		ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1717 		ASSIGN(priority);
1718 		ASSIGN(local_flags);
1719 
1720 		ASSIGN(qlimit);
1721 		ASSIGN(flags);
1722 
1723 		if (imported_q->scheduler == ALTQT_HFSC) {
1724 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1725 
1726 			/*
1727 			 * The m1 and m2 parameters are being copied from
1728 			 * 32-bit to 64-bit.
1729 			 */
1730 			ASSIGN_OPT(rtsc_m1);
1731 			ASSIGN_OPT(rtsc_d);
1732 			ASSIGN_OPT(rtsc_m2);
1733 
1734 			ASSIGN_OPT(lssc_m1);
1735 			ASSIGN_OPT(lssc_d);
1736 			ASSIGN_OPT(lssc_m2);
1737 
1738 			ASSIGN_OPT(ulsc_m1);
1739 			ASSIGN_OPT(ulsc_d);
1740 			ASSIGN_OPT(ulsc_m2);
1741 
1742 			ASSIGN_OPT(flags);
1743 
1744 #undef ASSIGN_OPT
1745 		} else
1746 			COPY(pq_u);
1747 
1748 		ASSIGN(qid);
1749 		break;
1750 	}
1751 	case 1: {
1752 		struct pf_altq_v1 *imported_q =
1753 		    &((struct pfioc_altq_v1 *)pa)->altq;
1754 
1755 		COPY(ifname);
1756 
1757 		ASSIGN(scheduler);
1758 		ASSIGN(tbrsize);
1759 		ASSIGN(ifbandwidth);
1760 
1761 		COPY(qname);
1762 		COPY(parent);
1763 		ASSIGN(parent_qid);
1764 		ASSIGN(bandwidth);
1765 		ASSIGN(priority);
1766 		ASSIGN(local_flags);
1767 
1768 		ASSIGN(qlimit);
1769 		ASSIGN(flags);
1770 		COPY(pq_u);
1771 
1772 		ASSIGN(qid);
1773 		break;
1774 	}
1775 	default:
1776 		panic("%s: unhandled struct pfioc_altq version", __func__);
1777 		break;
1778 	}
1779 
1780 #undef ASSIGN
1781 #undef COPY
1782 
1783 	return (0);
1784 }
1785 
1786 static struct pf_altq *
1787 pf_altq_get_nth_active(u_int32_t n)
1788 {
1789 	struct pf_altq		*altq;
1790 	u_int32_t		 nr;
1791 
1792 	nr = 0;
1793 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1794 		if (nr == n)
1795 			return (altq);
1796 		nr++;
1797 	}
1798 
1799 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1800 		if (nr == n)
1801 			return (altq);
1802 		nr++;
1803 	}
1804 
1805 	return (NULL);
1806 }
1807 #endif /* ALTQ */
1808 
1809 struct pf_krule *
1810 pf_krule_alloc(void)
1811 {
1812 	struct pf_krule *rule;
1813 
1814 	rule = malloc(sizeof(struct pf_krule), M_PFRULE, M_WAITOK | M_ZERO);
1815 	mtx_init(&rule->rpool.mtx, "pf_krule_pool", NULL, MTX_DEF);
1816 	rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
1817 	    M_WAITOK | M_ZERO);
1818 	return (rule);
1819 }
1820 
1821 void
1822 pf_krule_free(struct pf_krule *rule)
1823 {
1824 #ifdef PF_WANT_32_TO_64_COUNTER
1825 	bool wowned;
1826 #endif
1827 
1828 	if (rule == NULL)
1829 		return;
1830 
1831 #ifdef PF_WANT_32_TO_64_COUNTER
1832 	if (rule->allrulelinked) {
1833 		wowned = PF_RULES_WOWNED();
1834 		if (!wowned)
1835 			PF_RULES_WLOCK();
1836 		LIST_REMOVE(rule, allrulelist);
1837 		V_pf_allrulecount--;
1838 		if (!wowned)
1839 			PF_RULES_WUNLOCK();
1840 	}
1841 #endif
1842 
1843 	pf_counter_u64_deinit(&rule->evaluations);
1844 	for (int i = 0; i < 2; i++) {
1845 		pf_counter_u64_deinit(&rule->packets[i]);
1846 		pf_counter_u64_deinit(&rule->bytes[i]);
1847 	}
1848 	counter_u64_free(rule->states_cur);
1849 	counter_u64_free(rule->states_tot);
1850 	counter_u64_free(rule->src_nodes);
1851 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
1852 
1853 	mtx_destroy(&rule->rpool.mtx);
1854 	free(rule, M_PFRULE);
1855 }
1856 
1857 static void
1858 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool,
1859     struct pf_pooladdr *pool)
1860 {
1861 
1862 	bzero(pool, sizeof(*pool));
1863 	bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr));
1864 	strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname));
1865 }
1866 
1867 static int
1868 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool,
1869     struct pf_kpooladdr *kpool)
1870 {
1871 	int ret;
1872 
1873 	bzero(kpool, sizeof(*kpool));
1874 	bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr));
1875 	ret = pf_user_strcpy(kpool->ifname, pool->ifname,
1876 	    sizeof(kpool->ifname));
1877 	return (ret);
1878 }
1879 
1880 static void
1881 pf_kpool_to_pool(const struct pf_kpool *kpool, struct pf_pool *pool)
1882 {
1883 	bzero(pool, sizeof(*pool));
1884 
1885 	bcopy(&kpool->key, &pool->key, sizeof(pool->key));
1886 	bcopy(&kpool->counter, &pool->counter, sizeof(pool->counter));
1887 
1888 	pool->tblidx = kpool->tblidx;
1889 	pool->proxy_port[0] = kpool->proxy_port[0];
1890 	pool->proxy_port[1] = kpool->proxy_port[1];
1891 	pool->opts = kpool->opts;
1892 }
1893 
1894 static void
1895 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool)
1896 {
1897 	_Static_assert(sizeof(pool->key) == sizeof(kpool->key), "");
1898 	_Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), "");
1899 
1900 	bcopy(&pool->key, &kpool->key, sizeof(kpool->key));
1901 	bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter));
1902 
1903 	kpool->tblidx = pool->tblidx;
1904 	kpool->proxy_port[0] = pool->proxy_port[0];
1905 	kpool->proxy_port[1] = pool->proxy_port[1];
1906 	kpool->opts = pool->opts;
1907 }
1908 
1909 static void
1910 pf_krule_to_rule(const struct pf_krule *krule, struct pf_rule *rule)
1911 {
1912 
1913 	bzero(rule, sizeof(*rule));
1914 
1915 	bcopy(&krule->src, &rule->src, sizeof(rule->src));
1916 	bcopy(&krule->dst, &rule->dst, sizeof(rule->dst));
1917 
1918 	for (int i = 0; i < PF_SKIP_COUNT; ++i) {
1919 		if (rule->skip[i].ptr == NULL)
1920 			rule->skip[i].nr = -1;
1921 		else
1922 			rule->skip[i].nr = krule->skip[i].ptr->nr;
1923 	}
1924 
1925 	strlcpy(rule->label, krule->label[0], sizeof(rule->label));
1926 	strlcpy(rule->ifname, krule->ifname, sizeof(rule->ifname));
1927 	strlcpy(rule->qname, krule->qname, sizeof(rule->qname));
1928 	strlcpy(rule->pqname, krule->pqname, sizeof(rule->pqname));
1929 	strlcpy(rule->tagname, krule->tagname, sizeof(rule->tagname));
1930 	strlcpy(rule->match_tagname, krule->match_tagname,
1931 	    sizeof(rule->match_tagname));
1932 	strlcpy(rule->overload_tblname, krule->overload_tblname,
1933 	    sizeof(rule->overload_tblname));
1934 
1935 	pf_kpool_to_pool(&krule->rpool, &rule->rpool);
1936 
1937 	rule->evaluations = pf_counter_u64_fetch(&krule->evaluations);
1938 	for (int i = 0; i < 2; i++) {
1939 		rule->packets[i] = pf_counter_u64_fetch(&krule->packets[i]);
1940 		rule->bytes[i] = pf_counter_u64_fetch(&krule->bytes[i]);
1941 	}
1942 
1943 	/* kif, anchor, overload_tbl are not copied over. */
1944 
1945 	rule->os_fingerprint = krule->os_fingerprint;
1946 
1947 	rule->rtableid = krule->rtableid;
1948 	bcopy(krule->timeout, rule->timeout, sizeof(krule->timeout));
1949 	rule->max_states = krule->max_states;
1950 	rule->max_src_nodes = krule->max_src_nodes;
1951 	rule->max_src_states = krule->max_src_states;
1952 	rule->max_src_conn = krule->max_src_conn;
1953 	rule->max_src_conn_rate.limit = krule->max_src_conn_rate.limit;
1954 	rule->max_src_conn_rate.seconds = krule->max_src_conn_rate.seconds;
1955 	rule->qid = krule->qid;
1956 	rule->pqid = krule->pqid;
1957 	rule->nr = krule->nr;
1958 	rule->prob = krule->prob;
1959 	rule->cuid = krule->cuid;
1960 	rule->cpid = krule->cpid;
1961 
1962 	rule->return_icmp = krule->return_icmp;
1963 	rule->return_icmp6 = krule->return_icmp6;
1964 	rule->max_mss = krule->max_mss;
1965 	rule->tag = krule->tag;
1966 	rule->match_tag = krule->match_tag;
1967 	rule->scrub_flags = krule->scrub_flags;
1968 
1969 	bcopy(&krule->uid, &rule->uid, sizeof(krule->uid));
1970 	bcopy(&krule->gid, &rule->gid, sizeof(krule->gid));
1971 
1972 	rule->rule_flag = krule->rule_flag;
1973 	rule->action = krule->action;
1974 	rule->direction = krule->direction;
1975 	rule->log = krule->log;
1976 	rule->logif = krule->logif;
1977 	rule->quick = krule->quick;
1978 	rule->ifnot = krule->ifnot;
1979 	rule->match_tag_not = krule->match_tag_not;
1980 	rule->natpass = krule->natpass;
1981 
1982 	rule->keep_state = krule->keep_state;
1983 	rule->af = krule->af;
1984 	rule->proto = krule->proto;
1985 	rule->type = krule->type;
1986 	rule->code = krule->code;
1987 	rule->flags = krule->flags;
1988 	rule->flagset = krule->flagset;
1989 	rule->min_ttl = krule->min_ttl;
1990 	rule->allow_opts = krule->allow_opts;
1991 	rule->rt = krule->rt;
1992 	rule->return_ttl = krule->return_ttl;
1993 	rule->tos = krule->tos;
1994 	rule->set_tos = krule->set_tos;
1995 	rule->anchor_relative = krule->anchor_relative;
1996 	rule->anchor_wildcard = krule->anchor_wildcard;
1997 
1998 	rule->flush = krule->flush;
1999 	rule->prio = krule->prio;
2000 	rule->set_prio[0] = krule->set_prio[0];
2001 	rule->set_prio[1] = krule->set_prio[1];
2002 
2003 	bcopy(&krule->divert, &rule->divert, sizeof(krule->divert));
2004 
2005 	rule->u_states_cur = counter_u64_fetch(krule->states_cur);
2006 	rule->u_states_tot = counter_u64_fetch(krule->states_tot);
2007 	rule->u_src_nodes = counter_u64_fetch(krule->src_nodes);
2008 }
2009 
2010 static int
2011 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
2012 {
2013 	int ret;
2014 
2015 #ifndef INET
2016 	if (rule->af == AF_INET) {
2017 		return (EAFNOSUPPORT);
2018 	}
2019 #endif /* INET */
2020 #ifndef INET6
2021 	if (rule->af == AF_INET6) {
2022 		return (EAFNOSUPPORT);
2023 	}
2024 #endif /* INET6 */
2025 
2026 	ret = pf_check_rule_addr(&rule->src);
2027 	if (ret != 0)
2028 		return (ret);
2029 	ret = pf_check_rule_addr(&rule->dst);
2030 	if (ret != 0)
2031 		return (ret);
2032 
2033 	bcopy(&rule->src, &krule->src, sizeof(rule->src));
2034 	bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
2035 
2036 	ret = pf_user_strcpy(krule->label[0], rule->label, sizeof(rule->label));
2037 	if (ret != 0)
2038 		return (ret);
2039 	ret = pf_user_strcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
2040 	if (ret != 0)
2041 		return (ret);
2042 	ret = pf_user_strcpy(krule->qname, rule->qname, sizeof(rule->qname));
2043 	if (ret != 0)
2044 		return (ret);
2045 	ret = pf_user_strcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
2046 	if (ret != 0)
2047 		return (ret);
2048 	ret = pf_user_strcpy(krule->tagname, rule->tagname,
2049 	    sizeof(rule->tagname));
2050 	if (ret != 0)
2051 		return (ret);
2052 	ret = pf_user_strcpy(krule->match_tagname, rule->match_tagname,
2053 	    sizeof(rule->match_tagname));
2054 	if (ret != 0)
2055 		return (ret);
2056 	ret = pf_user_strcpy(krule->overload_tblname, rule->overload_tblname,
2057 	    sizeof(rule->overload_tblname));
2058 	if (ret != 0)
2059 		return (ret);
2060 
2061 	pf_pool_to_kpool(&rule->rpool, &krule->rpool);
2062 
2063 	/* Don't allow userspace to set evaluations, packets or bytes. */
2064 	/* kif, anchor, overload_tbl are not copied over. */
2065 
2066 	krule->os_fingerprint = rule->os_fingerprint;
2067 
2068 	krule->rtableid = rule->rtableid;
2069 	bcopy(rule->timeout, krule->timeout, sizeof(krule->timeout));
2070 	krule->max_states = rule->max_states;
2071 	krule->max_src_nodes = rule->max_src_nodes;
2072 	krule->max_src_states = rule->max_src_states;
2073 	krule->max_src_conn = rule->max_src_conn;
2074 	krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
2075 	krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
2076 	krule->qid = rule->qid;
2077 	krule->pqid = rule->pqid;
2078 	krule->nr = rule->nr;
2079 	krule->prob = rule->prob;
2080 	krule->cuid = rule->cuid;
2081 	krule->cpid = rule->cpid;
2082 
2083 	krule->return_icmp = rule->return_icmp;
2084 	krule->return_icmp6 = rule->return_icmp6;
2085 	krule->max_mss = rule->max_mss;
2086 	krule->tag = rule->tag;
2087 	krule->match_tag = rule->match_tag;
2088 	krule->scrub_flags = rule->scrub_flags;
2089 
2090 	bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
2091 	bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
2092 
2093 	krule->rule_flag = rule->rule_flag;
2094 	krule->action = rule->action;
2095 	krule->direction = rule->direction;
2096 	krule->log = rule->log;
2097 	krule->logif = rule->logif;
2098 	krule->quick = rule->quick;
2099 	krule->ifnot = rule->ifnot;
2100 	krule->match_tag_not = rule->match_tag_not;
2101 	krule->natpass = rule->natpass;
2102 
2103 	krule->keep_state = rule->keep_state;
2104 	krule->af = rule->af;
2105 	krule->proto = rule->proto;
2106 	krule->type = rule->type;
2107 	krule->code = rule->code;
2108 	krule->flags = rule->flags;
2109 	krule->flagset = rule->flagset;
2110 	krule->min_ttl = rule->min_ttl;
2111 	krule->allow_opts = rule->allow_opts;
2112 	krule->rt = rule->rt;
2113 	krule->return_ttl = rule->return_ttl;
2114 	krule->tos = rule->tos;
2115 	krule->set_tos = rule->set_tos;
2116 
2117 	krule->flush = rule->flush;
2118 	krule->prio = rule->prio;
2119 	krule->set_prio[0] = rule->set_prio[0];
2120 	krule->set_prio[1] = rule->set_prio[1];
2121 
2122 	bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
2123 
2124 	return (0);
2125 }
2126 
2127 static int
2128 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket,
2129     uint32_t pool_ticket, const char *anchor, const char *anchor_call,
2130     struct thread *td)
2131 {
2132 	struct pf_kruleset	*ruleset;
2133 	struct pf_krule		*tail;
2134 	struct pf_kpooladdr	*pa;
2135 	struct pfi_kkif		*kif = NULL;
2136 	int			 rs_num;
2137 	int			 error = 0;
2138 
2139 	if ((rule->return_icmp >> 8) > ICMP_MAXTYPE) {
2140 		error = EINVAL;
2141 		goto errout_unlocked;
2142 	}
2143 
2144 #define	ERROUT(x)	ERROUT_FUNCTION(errout, x)
2145 
2146 	if (rule->ifname[0])
2147 		kif = pf_kkif_create(M_WAITOK);
2148 	pf_counter_u64_init(&rule->evaluations, M_WAITOK);
2149 	for (int i = 0; i < 2; i++) {
2150 		pf_counter_u64_init(&rule->packets[i], M_WAITOK);
2151 		pf_counter_u64_init(&rule->bytes[i], M_WAITOK);
2152 	}
2153 	rule->states_cur = counter_u64_alloc(M_WAITOK);
2154 	rule->states_tot = counter_u64_alloc(M_WAITOK);
2155 	rule->src_nodes = counter_u64_alloc(M_WAITOK);
2156 	rule->cuid = td->td_ucred->cr_ruid;
2157 	rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
2158 	TAILQ_INIT(&rule->rpool.list);
2159 
2160 	PF_CONFIG_LOCK();
2161 	PF_RULES_WLOCK();
2162 #ifdef PF_WANT_32_TO_64_COUNTER
2163 	LIST_INSERT_HEAD(&V_pf_allrulelist, rule, allrulelist);
2164 	MPASS(!rule->allrulelinked);
2165 	rule->allrulelinked = true;
2166 	V_pf_allrulecount++;
2167 #endif
2168 	ruleset = pf_find_kruleset(anchor);
2169 	if (ruleset == NULL)
2170 		ERROUT(EINVAL);
2171 	rs_num = pf_get_ruleset_number(rule->action);
2172 	if (rs_num >= PF_RULESET_MAX)
2173 		ERROUT(EINVAL);
2174 	if (ticket != ruleset->rules[rs_num].inactive.ticket) {
2175 		DPFPRINTF(PF_DEBUG_MISC,
2176 		    ("ticket: %d != [%d]%d\n", ticket, rs_num,
2177 		    ruleset->rules[rs_num].inactive.ticket));
2178 		ERROUT(EBUSY);
2179 	}
2180 	if (pool_ticket != V_ticket_pabuf) {
2181 		DPFPRINTF(PF_DEBUG_MISC,
2182 		    ("pool_ticket: %d != %d\n", pool_ticket,
2183 		    V_ticket_pabuf));
2184 		ERROUT(EBUSY);
2185 	}
2186 	/*
2187 	 * XXXMJG hack: there is no mechanism to ensure they started the
2188 	 * transaction. Ticket checked above may happen to match by accident,
2189 	 * even if nobody called DIOCXBEGIN, let alone this process.
2190 	 * Partially work around it by checking if the RB tree got allocated,
2191 	 * see pf_begin_rules.
2192 	 */
2193 	if (ruleset->rules[rs_num].inactive.tree == NULL) {
2194 		ERROUT(EINVAL);
2195 	}
2196 
2197 	tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
2198 	    pf_krulequeue);
2199 	if (tail)
2200 		rule->nr = tail->nr + 1;
2201 	else
2202 		rule->nr = 0;
2203 	if (rule->ifname[0]) {
2204 		rule->kif = pfi_kkif_attach(kif, rule->ifname);
2205 		kif = NULL;
2206 		pfi_kkif_ref(rule->kif);
2207 	} else
2208 		rule->kif = NULL;
2209 
2210 	if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
2211 		error = EBUSY;
2212 
2213 #ifdef ALTQ
2214 	/* set queue IDs */
2215 	if (rule->qname[0] != 0) {
2216 		if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2217 			error = EBUSY;
2218 		else if (rule->pqname[0] != 0) {
2219 			if ((rule->pqid =
2220 			    pf_qname2qid(rule->pqname)) == 0)
2221 				error = EBUSY;
2222 		} else
2223 			rule->pqid = rule->qid;
2224 	}
2225 #endif
2226 	if (rule->tagname[0])
2227 		if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2228 			error = EBUSY;
2229 	if (rule->match_tagname[0])
2230 		if ((rule->match_tag =
2231 		    pf_tagname2tag(rule->match_tagname)) == 0)
2232 			error = EBUSY;
2233 	if (rule->rt && !rule->direction)
2234 		error = EINVAL;
2235 	if (!rule->log)
2236 		rule->logif = 0;
2237 	if (rule->logif >= PFLOGIFS_MAX)
2238 		error = EINVAL;
2239 	if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
2240 		error = ENOMEM;
2241 	if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
2242 		error = ENOMEM;
2243 	if (pf_kanchor_setup(rule, ruleset, anchor_call))
2244 		error = EINVAL;
2245 	if (rule->scrub_flags & PFSTATE_SETPRIO &&
2246 	    (rule->set_prio[0] > PF_PRIO_MAX ||
2247 	    rule->set_prio[1] > PF_PRIO_MAX))
2248 		error = EINVAL;
2249 	TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
2250 		if (pa->addr.type == PF_ADDR_TABLE) {
2251 			pa->addr.p.tbl = pfr_attach_table(ruleset,
2252 			    pa->addr.v.tblname);
2253 			if (pa->addr.p.tbl == NULL)
2254 				error = ENOMEM;
2255 		}
2256 
2257 	rule->overload_tbl = NULL;
2258 	if (rule->overload_tblname[0]) {
2259 		if ((rule->overload_tbl = pfr_attach_table(ruleset,
2260 		    rule->overload_tblname)) == NULL)
2261 			error = EINVAL;
2262 		else
2263 			rule->overload_tbl->pfrkt_flags |=
2264 			    PFR_TFLAG_ACTIVE;
2265 	}
2266 
2267 	pf_mv_kpool(&V_pf_pabuf, &rule->rpool.list);
2268 	if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
2269 	    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
2270 	    (rule->rt > PF_NOPFROUTE)) &&
2271 	    (TAILQ_FIRST(&rule->rpool.list) == NULL))
2272 		error = EINVAL;
2273 
2274 	if (error) {
2275 		pf_free_rule(rule);
2276 		rule = NULL;
2277 		ERROUT(error);
2278 	}
2279 
2280 	rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
2281 	TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
2282 	    rule, entries);
2283 	ruleset->rules[rs_num].inactive.rcount++;
2284 
2285 	PF_RULES_WUNLOCK();
2286 	pf_hash_rule(rule);
2287 	if (RB_INSERT(pf_krule_global, ruleset->rules[rs_num].inactive.tree, rule) != NULL) {
2288 		PF_RULES_WLOCK();
2289 		TAILQ_REMOVE(ruleset->rules[rs_num].inactive.ptr, rule, entries);
2290 		ruleset->rules[rs_num].inactive.rcount--;
2291 		pf_free_rule(rule);
2292 		rule = NULL;
2293 		ERROUT(EEXIST);
2294 	}
2295 	PF_CONFIG_UNLOCK();
2296 
2297 	return (0);
2298 
2299 #undef ERROUT
2300 errout:
2301 	PF_RULES_WUNLOCK();
2302 	PF_CONFIG_UNLOCK();
2303 errout_unlocked:
2304 	pf_kkif_free(kif);
2305 	pf_krule_free(rule);
2306 	return (error);
2307 }
2308 
2309 static bool
2310 pf_label_match(const struct pf_krule *rule, const char *label)
2311 {
2312 	int i = 0;
2313 
2314 	while (*rule->label[i]) {
2315 		if (strcmp(rule->label[i], label) == 0)
2316 			return (true);
2317 		i++;
2318 	}
2319 
2320 	return (false);
2321 }
2322 
2323 static unsigned int
2324 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir)
2325 {
2326 	struct pf_kstate *s;
2327 	int more = 0;
2328 
2329 	s = pf_find_state_all(key, dir, &more);
2330 	if (s == NULL)
2331 		return (0);
2332 
2333 	if (more) {
2334 		PF_STATE_UNLOCK(s);
2335 		return (0);
2336 	}
2337 
2338 	pf_unlink_state(s);
2339 	return (1);
2340 }
2341 
2342 static int
2343 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih)
2344 {
2345 	struct pf_kstate	*s;
2346 	struct pf_state_key	*sk;
2347 	struct pf_addr		*srcaddr, *dstaddr;
2348 	struct pf_state_key_cmp	 match_key;
2349 	int			 idx, killed = 0;
2350 	unsigned int		 dir;
2351 	u_int16_t		 srcport, dstport;
2352 	struct pfi_kkif		*kif;
2353 
2354 relock_DIOCKILLSTATES:
2355 	PF_HASHROW_LOCK(ih);
2356 	LIST_FOREACH(s, &ih->states, entry) {
2357 		/* For floating states look at the original kif. */
2358 		kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
2359 
2360 		sk = s->key[PF_SK_WIRE];
2361 		if (s->direction == PF_OUT) {
2362 			srcaddr = &sk->addr[1];
2363 			dstaddr = &sk->addr[0];
2364 			srcport = sk->port[1];
2365 			dstport = sk->port[0];
2366 		} else {
2367 			srcaddr = &sk->addr[0];
2368 			dstaddr = &sk->addr[1];
2369 			srcport = sk->port[0];
2370 			dstport = sk->port[1];
2371 		}
2372 
2373 		if (psk->psk_af && sk->af != psk->psk_af)
2374 			continue;
2375 
2376 		if (psk->psk_proto && psk->psk_proto != sk->proto)
2377 			continue;
2378 
2379 		if (! PF_MATCHA(psk->psk_src.neg, &psk->psk_src.addr.v.a.addr,
2380 		    &psk->psk_src.addr.v.a.mask, srcaddr, sk->af))
2381 			continue;
2382 
2383 		if (! PF_MATCHA(psk->psk_dst.neg, &psk->psk_dst.addr.v.a.addr,
2384 		    &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af))
2385 			continue;
2386 
2387 		if (!  PF_MATCHA(psk->psk_rt_addr.neg,
2388 		    &psk->psk_rt_addr.addr.v.a.addr,
2389 		    &psk->psk_rt_addr.addr.v.a.mask,
2390 		    &s->rt_addr, sk->af))
2391 			continue;
2392 
2393 		if (psk->psk_src.port_op != 0 &&
2394 		    ! pf_match_port(psk->psk_src.port_op,
2395 		    psk->psk_src.port[0], psk->psk_src.port[1], srcport))
2396 			continue;
2397 
2398 		if (psk->psk_dst.port_op != 0 &&
2399 		    ! pf_match_port(psk->psk_dst.port_op,
2400 		    psk->psk_dst.port[0], psk->psk_dst.port[1], dstport))
2401 			continue;
2402 
2403 		if (psk->psk_label[0] &&
2404 		    ! pf_label_match(s->rule.ptr, psk->psk_label))
2405 			continue;
2406 
2407 		if (psk->psk_ifname[0] && strcmp(psk->psk_ifname,
2408 		    kif->pfik_name))
2409 			continue;
2410 
2411 		if (psk->psk_kill_match) {
2412 			/* Create the key to find matching states, with lock
2413 			 * held. */
2414 
2415 			bzero(&match_key, sizeof(match_key));
2416 
2417 			if (s->direction == PF_OUT) {
2418 				dir = PF_IN;
2419 				idx = PF_SK_STACK;
2420 			} else {
2421 				dir = PF_OUT;
2422 				idx = PF_SK_WIRE;
2423 			}
2424 
2425 			match_key.af = s->key[idx]->af;
2426 			match_key.proto = s->key[idx]->proto;
2427 			PF_ACPY(&match_key.addr[0],
2428 			    &s->key[idx]->addr[1], match_key.af);
2429 			match_key.port[0] = s->key[idx]->port[1];
2430 			PF_ACPY(&match_key.addr[1],
2431 			    &s->key[idx]->addr[0], match_key.af);
2432 			match_key.port[1] = s->key[idx]->port[0];
2433 		}
2434 
2435 		pf_unlink_state(s);
2436 		killed++;
2437 
2438 		if (psk->psk_kill_match)
2439 			killed += pf_kill_matching_state(&match_key, dir);
2440 
2441 		goto relock_DIOCKILLSTATES;
2442 	}
2443 	PF_HASHROW_UNLOCK(ih);
2444 
2445 	return (killed);
2446 }
2447 
2448 static int
2449 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
2450 {
2451 	int			 error = 0;
2452 	PF_RULES_RLOCK_TRACKER;
2453 
2454 #define	ERROUT_IOCTL(target, x)					\
2455     do {								\
2456 	    error = (x);						\
2457 	    SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__);	\
2458 	    goto target;						\
2459     } while (0)
2460 
2461 
2462 	/* XXX keep in sync with switch() below */
2463 	if (securelevel_gt(td->td_ucred, 2))
2464 		switch (cmd) {
2465 		case DIOCGETRULES:
2466 		case DIOCGETRULENV:
2467 		case DIOCGETADDRS:
2468 		case DIOCGETADDR:
2469 		case DIOCGETSTATE:
2470 		case DIOCGETSTATENV:
2471 		case DIOCSETSTATUSIF:
2472 		case DIOCGETSTATUSNV:
2473 		case DIOCCLRSTATUS:
2474 		case DIOCNATLOOK:
2475 		case DIOCSETDEBUG:
2476 		case DIOCGETSTATES:
2477 		case DIOCGETSTATESV2:
2478 		case DIOCGETTIMEOUT:
2479 		case DIOCCLRRULECTRS:
2480 		case DIOCGETLIMIT:
2481 		case DIOCGETALTQSV0:
2482 		case DIOCGETALTQSV1:
2483 		case DIOCGETALTQV0:
2484 		case DIOCGETALTQV1:
2485 		case DIOCGETQSTATSV0:
2486 		case DIOCGETQSTATSV1:
2487 		case DIOCGETRULESETS:
2488 		case DIOCGETRULESET:
2489 		case DIOCRGETTABLES:
2490 		case DIOCRGETTSTATS:
2491 		case DIOCRCLRTSTATS:
2492 		case DIOCRCLRADDRS:
2493 		case DIOCRADDADDRS:
2494 		case DIOCRDELADDRS:
2495 		case DIOCRSETADDRS:
2496 		case DIOCRGETADDRS:
2497 		case DIOCRGETASTATS:
2498 		case DIOCRCLRASTATS:
2499 		case DIOCRTSTADDRS:
2500 		case DIOCOSFPGET:
2501 		case DIOCGETSRCNODES:
2502 		case DIOCCLRSRCNODES:
2503 		case DIOCGETSYNCOOKIES:
2504 		case DIOCIGETIFACES:
2505 		case DIOCGIFSPEEDV0:
2506 		case DIOCGIFSPEEDV1:
2507 		case DIOCSETIFFLAG:
2508 		case DIOCCLRIFFLAG:
2509 		case DIOCGETETHRULES:
2510 		case DIOCGETETHRULE:
2511 		case DIOCGETETHRULESETS:
2512 		case DIOCGETETHRULESET:
2513 			break;
2514 		case DIOCRCLRTABLES:
2515 		case DIOCRADDTABLES:
2516 		case DIOCRDELTABLES:
2517 		case DIOCRSETTFLAGS:
2518 			if (((struct pfioc_table *)addr)->pfrio_flags &
2519 			    PFR_FLAG_DUMMY)
2520 				break; /* dummy operation ok */
2521 			return (EPERM);
2522 		default:
2523 			return (EPERM);
2524 		}
2525 
2526 	if (!(flags & FWRITE))
2527 		switch (cmd) {
2528 		case DIOCGETRULES:
2529 		case DIOCGETADDRS:
2530 		case DIOCGETADDR:
2531 		case DIOCGETSTATE:
2532 		case DIOCGETSTATENV:
2533 		case DIOCGETSTATUSNV:
2534 		case DIOCGETSTATES:
2535 		case DIOCGETSTATESV2:
2536 		case DIOCGETTIMEOUT:
2537 		case DIOCGETLIMIT:
2538 		case DIOCGETALTQSV0:
2539 		case DIOCGETALTQSV1:
2540 		case DIOCGETALTQV0:
2541 		case DIOCGETALTQV1:
2542 		case DIOCGETQSTATSV0:
2543 		case DIOCGETQSTATSV1:
2544 		case DIOCGETRULESETS:
2545 		case DIOCGETRULESET:
2546 		case DIOCNATLOOK:
2547 		case DIOCRGETTABLES:
2548 		case DIOCRGETTSTATS:
2549 		case DIOCRGETADDRS:
2550 		case DIOCRGETASTATS:
2551 		case DIOCRTSTADDRS:
2552 		case DIOCOSFPGET:
2553 		case DIOCGETSRCNODES:
2554 		case DIOCGETSYNCOOKIES:
2555 		case DIOCIGETIFACES:
2556 		case DIOCGIFSPEEDV1:
2557 		case DIOCGIFSPEEDV0:
2558 		case DIOCGETRULENV:
2559 		case DIOCGETETHRULES:
2560 		case DIOCGETETHRULE:
2561 		case DIOCGETETHRULESETS:
2562 		case DIOCGETETHRULESET:
2563 			break;
2564 		case DIOCRCLRTABLES:
2565 		case DIOCRADDTABLES:
2566 		case DIOCRDELTABLES:
2567 		case DIOCRCLRTSTATS:
2568 		case DIOCRCLRADDRS:
2569 		case DIOCRADDADDRS:
2570 		case DIOCRDELADDRS:
2571 		case DIOCRSETADDRS:
2572 		case DIOCRSETTFLAGS:
2573 			if (((struct pfioc_table *)addr)->pfrio_flags &
2574 			    PFR_FLAG_DUMMY) {
2575 				flags |= FWRITE; /* need write lock for dummy */
2576 				break; /* dummy operation ok */
2577 			}
2578 			return (EACCES);
2579 		default:
2580 			return (EACCES);
2581 		}
2582 
2583 	CURVNET_SET(TD_TO_VNET(td));
2584 
2585 	switch (cmd) {
2586 	case DIOCSTART:
2587 		sx_xlock(&V_pf_ioctl_lock);
2588 		if (V_pf_status.running)
2589 			error = EEXIST;
2590 		else {
2591 			hook_pf();
2592 			if (! TAILQ_EMPTY(V_pf_keth->active.rules))
2593 				hook_pf_eth();
2594 			V_pf_status.running = 1;
2595 			V_pf_status.since = time_second;
2596 			new_unrhdr64(&V_pf_stateid, time_second);
2597 
2598 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
2599 		}
2600 		break;
2601 
2602 	case DIOCSTOP:
2603 		sx_xlock(&V_pf_ioctl_lock);
2604 		if (!V_pf_status.running)
2605 			error = ENOENT;
2606 		else {
2607 			V_pf_status.running = 0;
2608 			dehook_pf();
2609 			dehook_pf_eth();
2610 			V_pf_status.since = time_second;
2611 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
2612 		}
2613 		break;
2614 
2615 	case DIOCGETETHRULES: {
2616 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2617 		nvlist_t		*nvl;
2618 		void			*packed;
2619 		struct pf_keth_rule	*tail;
2620 		struct pf_keth_ruleset	*rs;
2621 		u_int32_t		 ticket, nr;
2622 		const char		*anchor = "";
2623 
2624 		nvl = NULL;
2625 		packed = NULL;
2626 
2627 #define	ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULES_error, x)
2628 
2629 		if (nv->len > pf_ioctl_maxcount)
2630 			ERROUT(ENOMEM);
2631 
2632 		/* Copy the request in */
2633 		packed = malloc(nv->len, M_NVLIST, M_WAITOK);
2634 		if (packed == NULL)
2635 			ERROUT(ENOMEM);
2636 
2637 		error = copyin(nv->data, packed, nv->len);
2638 		if (error)
2639 			ERROUT(error);
2640 
2641 		nvl = nvlist_unpack(packed, nv->len, 0);
2642 		if (nvl == NULL)
2643 			ERROUT(EBADMSG);
2644 
2645 		if (! nvlist_exists_string(nvl, "anchor"))
2646 			ERROUT(EBADMSG);
2647 
2648 		anchor = nvlist_get_string(nvl, "anchor");
2649 
2650 		rs = pf_find_keth_ruleset(anchor);
2651 
2652 		nvlist_destroy(nvl);
2653 		nvl = NULL;
2654 		free(packed, M_NVLIST);
2655 		packed = NULL;
2656 
2657 		if (rs == NULL)
2658 			ERROUT(ENOENT);
2659 
2660 		/* Reply */
2661 		nvl = nvlist_create(0);
2662 		if (nvl == NULL)
2663 			ERROUT(ENOMEM);
2664 
2665 		PF_RULES_RLOCK();
2666 
2667 		ticket = rs->active.ticket;
2668 		tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq);
2669 		if (tail)
2670 			nr = tail->nr + 1;
2671 		else
2672 			nr = 0;
2673 
2674 		PF_RULES_RUNLOCK();
2675 
2676 		nvlist_add_number(nvl, "ticket", ticket);
2677 		nvlist_add_number(nvl, "nr", nr);
2678 
2679 		packed = nvlist_pack(nvl, &nv->len);
2680 		if (packed == NULL)
2681 			ERROUT(ENOMEM);
2682 
2683 		if (nv->size == 0)
2684 			ERROUT(0);
2685 		else if (nv->size < nv->len)
2686 			ERROUT(ENOSPC);
2687 
2688 		error = copyout(packed, nv->data, nv->len);
2689 
2690 #undef ERROUT
2691 DIOCGETETHRULES_error:
2692 		free(packed, M_NVLIST);
2693 		nvlist_destroy(nvl);
2694 		break;
2695 	}
2696 
2697 	case DIOCGETETHRULE: {
2698 		struct epoch_tracker	 et;
2699 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2700 		nvlist_t		*nvl = NULL;
2701 		void			*nvlpacked = NULL;
2702 		struct pf_keth_rule	*rule = NULL;
2703 		struct pf_keth_ruleset	*rs;
2704 		u_int32_t		 ticket, nr;
2705 		bool			 clear = false;
2706 		const char		*anchor;
2707 
2708 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULE_error, x)
2709 
2710 		if (nv->len > pf_ioctl_maxcount)
2711 			ERROUT(ENOMEM);
2712 
2713 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2714 		if (nvlpacked == NULL)
2715 			ERROUT(ENOMEM);
2716 
2717 		error = copyin(nv->data, nvlpacked, nv->len);
2718 		if (error)
2719 			ERROUT(error);
2720 
2721 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2722 		if (nvl == NULL)
2723 			ERROUT(EBADMSG);
2724 		if (! nvlist_exists_number(nvl, "ticket"))
2725 			ERROUT(EBADMSG);
2726 		ticket = nvlist_get_number(nvl, "ticket");
2727 		if (! nvlist_exists_string(nvl, "anchor"))
2728 			ERROUT(EBADMSG);
2729 		anchor = nvlist_get_string(nvl, "anchor");
2730 
2731 		if (nvlist_exists_bool(nvl, "clear"))
2732 			clear = nvlist_get_bool(nvl, "clear");
2733 
2734 		if (clear && !(flags & FWRITE))
2735 			ERROUT(EACCES);
2736 
2737 		if (! nvlist_exists_number(nvl, "nr"))
2738 			ERROUT(EBADMSG);
2739 		nr = nvlist_get_number(nvl, "nr");
2740 
2741 		PF_RULES_RLOCK();
2742 		rs = pf_find_keth_ruleset(anchor);
2743 		if (rs == NULL) {
2744 			PF_RULES_RUNLOCK();
2745 			ERROUT(ENOENT);
2746 		}
2747 		if (ticket != rs->active.ticket) {
2748 			PF_RULES_RUNLOCK();
2749 			ERROUT(EBUSY);
2750 		}
2751 
2752 		nvlist_destroy(nvl);
2753 		nvl = NULL;
2754 		free(nvlpacked, M_NVLIST);
2755 		nvlpacked = NULL;
2756 
2757 		rule = TAILQ_FIRST(rs->active.rules);
2758 		while ((rule != NULL) && (rule->nr != nr))
2759 			rule = TAILQ_NEXT(rule, entries);
2760 		if (rule == NULL) {
2761 			PF_RULES_RUNLOCK();
2762 			ERROUT(ENOENT);
2763 		}
2764 		/* Make sure rule can't go away. */
2765 		NET_EPOCH_ENTER(et);
2766 		PF_RULES_RUNLOCK();
2767 		nvl = pf_keth_rule_to_nveth_rule(rule);
2768 		if (pf_keth_anchor_nvcopyout(rs, rule, nvl))
2769 			ERROUT(EBUSY);
2770 		NET_EPOCH_EXIT(et);
2771 		if (nvl == NULL)
2772 			ERROUT(ENOMEM);
2773 
2774 		nvlpacked = nvlist_pack(nvl, &nv->len);
2775 		if (nvlpacked == NULL)
2776 			ERROUT(ENOMEM);
2777 
2778 		if (nv->size == 0)
2779 			ERROUT(0);
2780 		else if (nv->size < nv->len)
2781 			ERROUT(ENOSPC);
2782 
2783 		error = copyout(nvlpacked, nv->data, nv->len);
2784 		if (error == 0 && clear) {
2785 			counter_u64_zero(rule->evaluations);
2786 			for (int i = 0; i < 2; i++) {
2787 				counter_u64_zero(rule->packets[i]);
2788 				counter_u64_zero(rule->bytes[i]);
2789 			}
2790 		}
2791 
2792 #undef ERROUT
2793 DIOCGETETHRULE_error:
2794 		free(nvlpacked, M_NVLIST);
2795 		nvlist_destroy(nvl);
2796 		break;
2797 	}
2798 
2799 	case DIOCADDETHRULE: {
2800 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2801 		nvlist_t		*nvl = NULL;
2802 		void			*nvlpacked = NULL;
2803 		struct pf_keth_rule	*rule = NULL, *tail = NULL;
2804 		struct pf_keth_ruleset	*ruleset = NULL;
2805 		struct pfi_kkif		*kif = NULL, *bridge_to_kif = NULL;
2806 		const char		*anchor = "", *anchor_call = "";
2807 
2808 #define ERROUT(x)	ERROUT_IOCTL(DIOCADDETHRULE_error, x)
2809 
2810 		if (nv->len > pf_ioctl_maxcount)
2811 			ERROUT(ENOMEM);
2812 
2813 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2814 		if (nvlpacked == NULL)
2815 			ERROUT(ENOMEM);
2816 
2817 		error = copyin(nv->data, nvlpacked, nv->len);
2818 		if (error)
2819 			ERROUT(error);
2820 
2821 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2822 		if (nvl == NULL)
2823 			ERROUT(EBADMSG);
2824 
2825 		if (! nvlist_exists_number(nvl, "ticket"))
2826 			ERROUT(EBADMSG);
2827 
2828 		if (nvlist_exists_string(nvl, "anchor"))
2829 			anchor = nvlist_get_string(nvl, "anchor");
2830 		if (nvlist_exists_string(nvl, "anchor_call"))
2831 			anchor_call = nvlist_get_string(nvl, "anchor_call");
2832 
2833 		ruleset = pf_find_keth_ruleset(anchor);
2834 		if (ruleset == NULL)
2835 			ERROUT(EINVAL);
2836 
2837 		if (nvlist_get_number(nvl, "ticket") !=
2838 		    ruleset->inactive.ticket) {
2839 			DPFPRINTF(PF_DEBUG_MISC,
2840 			    ("ticket: %d != %d\n",
2841 			    (u_int32_t)nvlist_get_number(nvl, "ticket"),
2842 			    ruleset->inactive.ticket));
2843 			ERROUT(EBUSY);
2844 		}
2845 
2846 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
2847 		if (rule == NULL)
2848 			ERROUT(ENOMEM);
2849 		rule->timestamp = NULL;
2850 
2851 		error = pf_nveth_rule_to_keth_rule(nvl, rule);
2852 		if (error != 0)
2853 			ERROUT(error);
2854 
2855 		if (rule->ifname[0])
2856 			kif = pf_kkif_create(M_WAITOK);
2857 		if (rule->bridge_to_name[0])
2858 			bridge_to_kif = pf_kkif_create(M_WAITOK);
2859 		rule->evaluations = counter_u64_alloc(M_WAITOK);
2860 		for (int i = 0; i < 2; i++) {
2861 			rule->packets[i] = counter_u64_alloc(M_WAITOK);
2862 			rule->bytes[i] = counter_u64_alloc(M_WAITOK);
2863 		}
2864 		rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
2865 		    M_WAITOK | M_ZERO);
2866 
2867 		PF_RULES_WLOCK();
2868 
2869 		if (rule->ifname[0]) {
2870 			rule->kif = pfi_kkif_attach(kif, rule->ifname);
2871 			pfi_kkif_ref(rule->kif);
2872 		} else
2873 			rule->kif = NULL;
2874 		if (rule->bridge_to_name[0]) {
2875 			rule->bridge_to = pfi_kkif_attach(bridge_to_kif,
2876 			    rule->bridge_to_name);
2877 			pfi_kkif_ref(rule->bridge_to);
2878 		} else
2879 			rule->bridge_to = NULL;
2880 
2881 #ifdef ALTQ
2882 		/* set queue IDs */
2883 		if (rule->qname[0] != 0) {
2884 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2885 				error = EBUSY;
2886 			else
2887 				rule->qid = rule->qid;
2888 		}
2889 #endif
2890 		if (rule->tagname[0])
2891 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2892 				error = EBUSY;
2893 		if (rule->match_tagname[0])
2894 			if ((rule->match_tag = pf_tagname2tag(
2895 			    rule->match_tagname)) == 0)
2896 				error = EBUSY;
2897 
2898 		if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE)
2899 			error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr);
2900 		if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE)
2901 			error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr);
2902 
2903 		if (error) {
2904 			pf_free_eth_rule(rule);
2905 			PF_RULES_WUNLOCK();
2906 			ERROUT(error);
2907 		}
2908 
2909 		if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) {
2910 			pf_free_eth_rule(rule);
2911 			PF_RULES_WUNLOCK();
2912 			ERROUT(EINVAL);
2913 		}
2914 
2915 		tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq);
2916 		if (tail)
2917 			rule->nr = tail->nr + 1;
2918 		else
2919 			rule->nr = 0;
2920 
2921 		TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries);
2922 
2923 		PF_RULES_WUNLOCK();
2924 
2925 #undef ERROUT
2926 DIOCADDETHRULE_error:
2927 		nvlist_destroy(nvl);
2928 		free(nvlpacked, M_NVLIST);
2929 		break;
2930 	}
2931 
2932 	case DIOCGETETHRULESETS: {
2933 		struct epoch_tracker	 et;
2934 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2935 		nvlist_t		*nvl = NULL;
2936 		void			*nvlpacked = NULL;
2937 		struct pf_keth_ruleset	*ruleset;
2938 		struct pf_keth_anchor	*anchor;
2939 		int			 nr = 0;
2940 
2941 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULESETS_error, x)
2942 
2943 		if (nv->len > pf_ioctl_maxcount)
2944 			ERROUT(ENOMEM);
2945 
2946 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2947 		if (nvlpacked == NULL)
2948 			ERROUT(ENOMEM);
2949 
2950 		error = copyin(nv->data, nvlpacked, nv->len);
2951 		if (error)
2952 			ERROUT(error);
2953 
2954 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2955 		if (nvl == NULL)
2956 			ERROUT(EBADMSG);
2957 		if (! nvlist_exists_string(nvl, "path"))
2958 			ERROUT(EBADMSG);
2959 
2960 		NET_EPOCH_ENTER(et);
2961 
2962 		if ((ruleset = pf_find_keth_ruleset(
2963 		    nvlist_get_string(nvl, "path"))) == NULL) {
2964 			NET_EPOCH_EXIT(et);
2965 			ERROUT(ENOENT);
2966 		}
2967 
2968 		if (ruleset->anchor == NULL) {
2969 			RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors)
2970 				if (anchor->parent == NULL)
2971 					nr++;
2972 		} else {
2973 			RB_FOREACH(anchor, pf_keth_anchor_node,
2974 			    &ruleset->anchor->children)
2975 				nr++;
2976 		}
2977 
2978 		NET_EPOCH_EXIT(et);
2979 
2980 		nvlist_destroy(nvl);
2981 		nvl = NULL;
2982 		free(nvlpacked, M_NVLIST);
2983 		nvlpacked = NULL;
2984 
2985 		nvl = nvlist_create(0);
2986 		if (nvl == NULL)
2987 			ERROUT(ENOMEM);
2988 
2989 		nvlist_add_number(nvl, "nr", nr);
2990 
2991 		nvlpacked = nvlist_pack(nvl, &nv->len);
2992 		if (nvlpacked == NULL)
2993 			ERROUT(ENOMEM);
2994 
2995 		if (nv->size == 0)
2996 			ERROUT(0);
2997 		else if (nv->size < nv->len)
2998 			ERROUT(ENOSPC);
2999 
3000 		error = copyout(nvlpacked, nv->data, nv->len);
3001 
3002 #undef ERROUT
3003 DIOCGETETHRULESETS_error:
3004 		free(nvlpacked, M_NVLIST);
3005 		nvlist_destroy(nvl);
3006 		break;
3007 	}
3008 
3009 	case DIOCGETETHRULESET: {
3010 		struct epoch_tracker	 et;
3011 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
3012 		nvlist_t		*nvl = NULL;
3013 		void			*nvlpacked = NULL;
3014 		struct pf_keth_ruleset	*ruleset;
3015 		struct pf_keth_anchor	*anchor;
3016 		int			 nr = 0, req_nr = 0;
3017 		bool			 found = false;
3018 
3019 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULESET_error, x)
3020 
3021 		if (nv->len > pf_ioctl_maxcount)
3022 			ERROUT(ENOMEM);
3023 
3024 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3025 		if (nvlpacked == NULL)
3026 			ERROUT(ENOMEM);
3027 
3028 		error = copyin(nv->data, nvlpacked, nv->len);
3029 		if (error)
3030 			ERROUT(error);
3031 
3032 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3033 		if (nvl == NULL)
3034 			ERROUT(EBADMSG);
3035 		if (! nvlist_exists_string(nvl, "path"))
3036 			ERROUT(EBADMSG);
3037 		if (! nvlist_exists_number(nvl, "nr"))
3038 			ERROUT(EBADMSG);
3039 
3040 		req_nr = nvlist_get_number(nvl, "nr");
3041 
3042 		NET_EPOCH_ENTER(et);
3043 
3044 		if ((ruleset = pf_find_keth_ruleset(
3045 		    nvlist_get_string(nvl, "path"))) == NULL) {
3046 			NET_EPOCH_EXIT(et);
3047 			ERROUT(ENOENT);
3048 		}
3049 
3050 		nvlist_destroy(nvl);
3051 		nvl = NULL;
3052 		free(nvlpacked, M_NVLIST);
3053 		nvlpacked = NULL;
3054 
3055 		nvl = nvlist_create(0);
3056 		if (nvl == NULL) {
3057 			NET_EPOCH_EXIT(et);
3058 			ERROUT(ENOMEM);
3059 		}
3060 
3061 		if (ruleset->anchor == NULL) {
3062 			RB_FOREACH(anchor, pf_keth_anchor_global,
3063 			    &V_pf_keth_anchors) {
3064 				if (anchor->parent == NULL && nr++ == req_nr) {
3065 					found = true;
3066 					break;
3067 				}
3068 			}
3069 		} else {
3070 			RB_FOREACH(anchor, pf_keth_anchor_node,
3071 			     &ruleset->anchor->children) {
3072 				if (nr++ == req_nr) {
3073 					found = true;
3074 					break;
3075 				}
3076 			}
3077 		}
3078 
3079 		NET_EPOCH_EXIT(et);
3080 		if (found) {
3081 			nvlist_add_number(nvl, "nr", nr);
3082 			nvlist_add_string(nvl, "name", anchor->name);
3083 			if (ruleset->anchor)
3084 				nvlist_add_string(nvl, "path",
3085 				    ruleset->anchor->path);
3086 			else
3087 				nvlist_add_string(nvl, "path", "");
3088 		} else {
3089 			ERROUT(EBUSY);
3090 		}
3091 
3092 		nvlpacked = nvlist_pack(nvl, &nv->len);
3093 		if (nvlpacked == NULL)
3094 			ERROUT(ENOMEM);
3095 
3096 		if (nv->size == 0)
3097 			ERROUT(0);
3098 		else if (nv->size < nv->len)
3099 			ERROUT(ENOSPC);
3100 
3101 		error = copyout(nvlpacked, nv->data, nv->len);
3102 
3103 #undef ERROUT
3104 DIOCGETETHRULESET_error:
3105 		free(nvlpacked, M_NVLIST);
3106 		nvlist_destroy(nvl);
3107 		break;
3108 	}
3109 
3110 	case DIOCADDRULENV: {
3111 		struct pfioc_nv	*nv = (struct pfioc_nv *)addr;
3112 		nvlist_t	*nvl = NULL;
3113 		void		*nvlpacked = NULL;
3114 		struct pf_krule	*rule = NULL;
3115 		const char	*anchor = "", *anchor_call = "";
3116 		uint32_t	 ticket = 0, pool_ticket = 0;
3117 
3118 #define	ERROUT(x)	ERROUT_IOCTL(DIOCADDRULENV_error, x)
3119 
3120 		if (nv->len > pf_ioctl_maxcount)
3121 			ERROUT(ENOMEM);
3122 
3123 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3124 		error = copyin(nv->data, nvlpacked, nv->len);
3125 		if (error)
3126 			ERROUT(error);
3127 
3128 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3129 		if (nvl == NULL)
3130 			ERROUT(EBADMSG);
3131 
3132 		if (! nvlist_exists_number(nvl, "ticket"))
3133 			ERROUT(EINVAL);
3134 		ticket = nvlist_get_number(nvl, "ticket");
3135 
3136 		if (! nvlist_exists_number(nvl, "pool_ticket"))
3137 			ERROUT(EINVAL);
3138 		pool_ticket = nvlist_get_number(nvl, "pool_ticket");
3139 
3140 		if (! nvlist_exists_nvlist(nvl, "rule"))
3141 			ERROUT(EINVAL);
3142 
3143 		rule = pf_krule_alloc();
3144 		error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"),
3145 		    rule);
3146 		if (error)
3147 			ERROUT(error);
3148 
3149 		if (nvlist_exists_string(nvl, "anchor"))
3150 			anchor = nvlist_get_string(nvl, "anchor");
3151 		if (nvlist_exists_string(nvl, "anchor_call"))
3152 			anchor_call = nvlist_get_string(nvl, "anchor_call");
3153 
3154 		if ((error = nvlist_error(nvl)))
3155 			ERROUT(error);
3156 
3157 		/* Frees rule on error */
3158 		error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor,
3159 		    anchor_call, td);
3160 
3161 		nvlist_destroy(nvl);
3162 		free(nvlpacked, M_NVLIST);
3163 		break;
3164 #undef ERROUT
3165 DIOCADDRULENV_error:
3166 		pf_krule_free(rule);
3167 		nvlist_destroy(nvl);
3168 		free(nvlpacked, M_NVLIST);
3169 
3170 		break;
3171 	}
3172 	case DIOCADDRULE: {
3173 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
3174 		struct pf_krule		*rule;
3175 
3176 		rule = pf_krule_alloc();
3177 		error = pf_rule_to_krule(&pr->rule, rule);
3178 		if (error != 0) {
3179 			pf_krule_free(rule);
3180 			break;
3181 		}
3182 
3183 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
3184 
3185 		/* Frees rule on error */
3186 		error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket,
3187 		    pr->anchor, pr->anchor_call, td);
3188 		break;
3189 	}
3190 
3191 	case DIOCGETRULES: {
3192 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
3193 		struct pf_kruleset	*ruleset;
3194 		struct pf_krule		*tail;
3195 		int			 rs_num;
3196 
3197 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
3198 
3199 		PF_RULES_WLOCK();
3200 		ruleset = pf_find_kruleset(pr->anchor);
3201 		if (ruleset == NULL) {
3202 			PF_RULES_WUNLOCK();
3203 			error = EINVAL;
3204 			break;
3205 		}
3206 		rs_num = pf_get_ruleset_number(pr->rule.action);
3207 		if (rs_num >= PF_RULESET_MAX) {
3208 			PF_RULES_WUNLOCK();
3209 			error = EINVAL;
3210 			break;
3211 		}
3212 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
3213 		    pf_krulequeue);
3214 		if (tail)
3215 			pr->nr = tail->nr + 1;
3216 		else
3217 			pr->nr = 0;
3218 		pr->ticket = ruleset->rules[rs_num].active.ticket;
3219 		PF_RULES_WUNLOCK();
3220 		break;
3221 	}
3222 
3223 	case DIOCGETRULENV: {
3224 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
3225 		nvlist_t		*nvrule = NULL;
3226 		nvlist_t		*nvl = NULL;
3227 		struct pf_kruleset	*ruleset;
3228 		struct pf_krule		*rule;
3229 		void			*nvlpacked = NULL;
3230 		int			 rs_num, nr;
3231 		bool			 clear_counter = false;
3232 
3233 #define	ERROUT(x)	ERROUT_IOCTL(DIOCGETRULENV_error, x)
3234 
3235 		if (nv->len > pf_ioctl_maxcount)
3236 			ERROUT(ENOMEM);
3237 
3238 		/* Copy the request in */
3239 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3240 		if (nvlpacked == NULL)
3241 			ERROUT(ENOMEM);
3242 
3243 		error = copyin(nv->data, nvlpacked, nv->len);
3244 		if (error)
3245 			ERROUT(error);
3246 
3247 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3248 		if (nvl == NULL)
3249 			ERROUT(EBADMSG);
3250 
3251 		if (! nvlist_exists_string(nvl, "anchor"))
3252 			ERROUT(EBADMSG);
3253 		if (! nvlist_exists_number(nvl, "ruleset"))
3254 			ERROUT(EBADMSG);
3255 		if (! nvlist_exists_number(nvl, "ticket"))
3256 			ERROUT(EBADMSG);
3257 		if (! nvlist_exists_number(nvl, "nr"))
3258 			ERROUT(EBADMSG);
3259 
3260 		if (nvlist_exists_bool(nvl, "clear_counter"))
3261 			clear_counter = nvlist_get_bool(nvl, "clear_counter");
3262 
3263 		if (clear_counter && !(flags & FWRITE))
3264 			ERROUT(EACCES);
3265 
3266 		nr = nvlist_get_number(nvl, "nr");
3267 
3268 		PF_RULES_WLOCK();
3269 		ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor"));
3270 		if (ruleset == NULL) {
3271 			PF_RULES_WUNLOCK();
3272 			ERROUT(ENOENT);
3273 		}
3274 
3275 		rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset"));
3276 		if (rs_num >= PF_RULESET_MAX) {
3277 			PF_RULES_WUNLOCK();
3278 			ERROUT(EINVAL);
3279 		}
3280 
3281 		if (nvlist_get_number(nvl, "ticket") !=
3282 		    ruleset->rules[rs_num].active.ticket) {
3283 			PF_RULES_WUNLOCK();
3284 			ERROUT(EBUSY);
3285 		}
3286 
3287 		if ((error = nvlist_error(nvl))) {
3288 			PF_RULES_WUNLOCK();
3289 			ERROUT(error);
3290 		}
3291 
3292 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
3293 		while ((rule != NULL) && (rule->nr != nr))
3294 			rule = TAILQ_NEXT(rule, entries);
3295 		if (rule == NULL) {
3296 			PF_RULES_WUNLOCK();
3297 			ERROUT(EBUSY);
3298 		}
3299 
3300 		nvrule = pf_krule_to_nvrule(rule);
3301 
3302 		nvlist_destroy(nvl);
3303 		nvl = nvlist_create(0);
3304 		if (nvl == NULL) {
3305 			PF_RULES_WUNLOCK();
3306 			ERROUT(ENOMEM);
3307 		}
3308 		nvlist_add_number(nvl, "nr", nr);
3309 		nvlist_add_nvlist(nvl, "rule", nvrule);
3310 		nvlist_destroy(nvrule);
3311 		nvrule = NULL;
3312 		if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) {
3313 			PF_RULES_WUNLOCK();
3314 			ERROUT(EBUSY);
3315 		}
3316 
3317 		free(nvlpacked, M_NVLIST);
3318 		nvlpacked = nvlist_pack(nvl, &nv->len);
3319 		if (nvlpacked == NULL) {
3320 			PF_RULES_WUNLOCK();
3321 			ERROUT(ENOMEM);
3322 		}
3323 
3324 		if (nv->size == 0) {
3325 			PF_RULES_WUNLOCK();
3326 			ERROUT(0);
3327 		}
3328 		else if (nv->size < nv->len) {
3329 			PF_RULES_WUNLOCK();
3330 			ERROUT(ENOSPC);
3331 		}
3332 
3333 		if (clear_counter) {
3334 			pf_counter_u64_zero(&rule->evaluations);
3335 			for (int i = 0; i < 2; i++) {
3336 				pf_counter_u64_zero(&rule->packets[i]);
3337 				pf_counter_u64_zero(&rule->bytes[i]);
3338 			}
3339 			counter_u64_zero(rule->states_tot);
3340 		}
3341 		PF_RULES_WUNLOCK();
3342 
3343 		error = copyout(nvlpacked, nv->data, nv->len);
3344 
3345 #undef ERROUT
3346 DIOCGETRULENV_error:
3347 		free(nvlpacked, M_NVLIST);
3348 		nvlist_destroy(nvrule);
3349 		nvlist_destroy(nvl);
3350 
3351 		break;
3352 	}
3353 
3354 	case DIOCCHANGERULE: {
3355 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
3356 		struct pf_kruleset	*ruleset;
3357 		struct pf_krule		*oldrule = NULL, *newrule = NULL;
3358 		struct pfi_kkif		*kif = NULL;
3359 		struct pf_kpooladdr	*pa;
3360 		u_int32_t		 nr = 0;
3361 		int			 rs_num;
3362 
3363 		pcr->anchor[sizeof(pcr->anchor) - 1] = 0;
3364 
3365 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
3366 		    pcr->action > PF_CHANGE_GET_TICKET) {
3367 			error = EINVAL;
3368 			break;
3369 		}
3370 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
3371 			error = EINVAL;
3372 			break;
3373 		}
3374 
3375 		if (pcr->action != PF_CHANGE_REMOVE) {
3376 			newrule = pf_krule_alloc();
3377 			error = pf_rule_to_krule(&pcr->rule, newrule);
3378 			if (error != 0) {
3379 				pf_krule_free(newrule);
3380 				break;
3381 			}
3382 
3383 			if (newrule->ifname[0])
3384 				kif = pf_kkif_create(M_WAITOK);
3385 			pf_counter_u64_init(&newrule->evaluations, M_WAITOK);
3386 			for (int i = 0; i < 2; i++) {
3387 				pf_counter_u64_init(&newrule->packets[i], M_WAITOK);
3388 				pf_counter_u64_init(&newrule->bytes[i], M_WAITOK);
3389 			}
3390 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
3391 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
3392 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
3393 			newrule->cuid = td->td_ucred->cr_ruid;
3394 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
3395 			TAILQ_INIT(&newrule->rpool.list);
3396 		}
3397 #define	ERROUT(x)	ERROUT_IOCTL(DIOCCHANGERULE_error, x)
3398 
3399 		PF_CONFIG_LOCK();
3400 		PF_RULES_WLOCK();
3401 #ifdef PF_WANT_32_TO_64_COUNTER
3402 		if (newrule != NULL) {
3403 			LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist);
3404 			newrule->allrulelinked = true;
3405 			V_pf_allrulecount++;
3406 		}
3407 #endif
3408 
3409 		if (!(pcr->action == PF_CHANGE_REMOVE ||
3410 		    pcr->action == PF_CHANGE_GET_TICKET) &&
3411 		    pcr->pool_ticket != V_ticket_pabuf)
3412 			ERROUT(EBUSY);
3413 
3414 		ruleset = pf_find_kruleset(pcr->anchor);
3415 		if (ruleset == NULL)
3416 			ERROUT(EINVAL);
3417 
3418 		rs_num = pf_get_ruleset_number(pcr->rule.action);
3419 		if (rs_num >= PF_RULESET_MAX)
3420 			ERROUT(EINVAL);
3421 
3422 		/*
3423 		 * XXXMJG: there is no guarantee that the ruleset was
3424 		 * created by the usual route of calling DIOCXBEGIN.
3425 		 * As a result it is possible the rule tree will not
3426 		 * be allocated yet. Hack around it by doing it here.
3427 		 * Note it is fine to let the tree persist in case of
3428 		 * error as it will be freed down the road on future
3429 		 * updates (if need be).
3430 		 */
3431 		if (ruleset->rules[rs_num].active.tree == NULL) {
3432 			ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT);
3433 			if (ruleset->rules[rs_num].active.tree == NULL) {
3434 				ERROUT(ENOMEM);
3435 			}
3436 		}
3437 
3438 		if (pcr->action == PF_CHANGE_GET_TICKET) {
3439 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
3440 			ERROUT(0);
3441 		} else if (pcr->ticket !=
3442 			    ruleset->rules[rs_num].active.ticket)
3443 				ERROUT(EINVAL);
3444 
3445 		if (pcr->action != PF_CHANGE_REMOVE) {
3446 			if (newrule->ifname[0]) {
3447 				newrule->kif = pfi_kkif_attach(kif,
3448 				    newrule->ifname);
3449 				kif = NULL;
3450 				pfi_kkif_ref(newrule->kif);
3451 			} else
3452 				newrule->kif = NULL;
3453 
3454 			if (newrule->rtableid > 0 &&
3455 			    newrule->rtableid >= rt_numfibs)
3456 				error = EBUSY;
3457 
3458 #ifdef ALTQ
3459 			/* set queue IDs */
3460 			if (newrule->qname[0] != 0) {
3461 				if ((newrule->qid =
3462 				    pf_qname2qid(newrule->qname)) == 0)
3463 					error = EBUSY;
3464 				else if (newrule->pqname[0] != 0) {
3465 					if ((newrule->pqid =
3466 					    pf_qname2qid(newrule->pqname)) == 0)
3467 						error = EBUSY;
3468 				} else
3469 					newrule->pqid = newrule->qid;
3470 			}
3471 #endif /* ALTQ */
3472 			if (newrule->tagname[0])
3473 				if ((newrule->tag =
3474 				    pf_tagname2tag(newrule->tagname)) == 0)
3475 					error = EBUSY;
3476 			if (newrule->match_tagname[0])
3477 				if ((newrule->match_tag = pf_tagname2tag(
3478 				    newrule->match_tagname)) == 0)
3479 					error = EBUSY;
3480 			if (newrule->rt && !newrule->direction)
3481 				error = EINVAL;
3482 			if (!newrule->log)
3483 				newrule->logif = 0;
3484 			if (newrule->logif >= PFLOGIFS_MAX)
3485 				error = EINVAL;
3486 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
3487 				error = ENOMEM;
3488 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
3489 				error = ENOMEM;
3490 			if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
3491 				error = EINVAL;
3492 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
3493 				if (pa->addr.type == PF_ADDR_TABLE) {
3494 					pa->addr.p.tbl =
3495 					    pfr_attach_table(ruleset,
3496 					    pa->addr.v.tblname);
3497 					if (pa->addr.p.tbl == NULL)
3498 						error = ENOMEM;
3499 				}
3500 
3501 			newrule->overload_tbl = NULL;
3502 			if (newrule->overload_tblname[0]) {
3503 				if ((newrule->overload_tbl = pfr_attach_table(
3504 				    ruleset, newrule->overload_tblname)) ==
3505 				    NULL)
3506 					error = EINVAL;
3507 				else
3508 					newrule->overload_tbl->pfrkt_flags |=
3509 					    PFR_TFLAG_ACTIVE;
3510 			}
3511 
3512 			pf_mv_kpool(&V_pf_pabuf, &newrule->rpool.list);
3513 			if (((((newrule->action == PF_NAT) ||
3514 			    (newrule->action == PF_RDR) ||
3515 			    (newrule->action == PF_BINAT) ||
3516 			    (newrule->rt > PF_NOPFROUTE)) &&
3517 			    !newrule->anchor)) &&
3518 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
3519 				error = EINVAL;
3520 
3521 			if (error) {
3522 				pf_free_rule(newrule);
3523 				PF_RULES_WUNLOCK();
3524 				PF_CONFIG_UNLOCK();
3525 				break;
3526 			}
3527 
3528 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
3529 		}
3530 		pf_empty_kpool(&V_pf_pabuf);
3531 
3532 		if (pcr->action == PF_CHANGE_ADD_HEAD)
3533 			oldrule = TAILQ_FIRST(
3534 			    ruleset->rules[rs_num].active.ptr);
3535 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
3536 			oldrule = TAILQ_LAST(
3537 			    ruleset->rules[rs_num].active.ptr, pf_krulequeue);
3538 		else {
3539 			oldrule = TAILQ_FIRST(
3540 			    ruleset->rules[rs_num].active.ptr);
3541 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
3542 				oldrule = TAILQ_NEXT(oldrule, entries);
3543 			if (oldrule == NULL) {
3544 				if (newrule != NULL)
3545 					pf_free_rule(newrule);
3546 				PF_RULES_WUNLOCK();
3547 				PF_CONFIG_UNLOCK();
3548 				error = EINVAL;
3549 				break;
3550 			}
3551 		}
3552 
3553 		if (pcr->action == PF_CHANGE_REMOVE) {
3554 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
3555 			    oldrule);
3556 			RB_REMOVE(pf_krule_global,
3557 			    ruleset->rules[rs_num].active.tree, oldrule);
3558 			ruleset->rules[rs_num].active.rcount--;
3559 		} else {
3560 			pf_hash_rule(newrule);
3561 			if (RB_INSERT(pf_krule_global,
3562 			    ruleset->rules[rs_num].active.tree, newrule) != NULL) {
3563 				pf_free_rule(newrule);
3564 				PF_RULES_WUNLOCK();
3565 				PF_CONFIG_UNLOCK();
3566 				error = EEXIST;
3567 				break;
3568 			}
3569 
3570 			if (oldrule == NULL)
3571 				TAILQ_INSERT_TAIL(
3572 				    ruleset->rules[rs_num].active.ptr,
3573 				    newrule, entries);
3574 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
3575 			    pcr->action == PF_CHANGE_ADD_BEFORE)
3576 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
3577 			else
3578 				TAILQ_INSERT_AFTER(
3579 				    ruleset->rules[rs_num].active.ptr,
3580 				    oldrule, newrule, entries);
3581 			ruleset->rules[rs_num].active.rcount++;
3582 		}
3583 
3584 		nr = 0;
3585 		TAILQ_FOREACH(oldrule,
3586 		    ruleset->rules[rs_num].active.ptr, entries)
3587 			oldrule->nr = nr++;
3588 
3589 		ruleset->rules[rs_num].active.ticket++;
3590 
3591 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
3592 		pf_remove_if_empty_kruleset(ruleset);
3593 
3594 		PF_RULES_WUNLOCK();
3595 		PF_CONFIG_UNLOCK();
3596 		break;
3597 
3598 #undef ERROUT
3599 DIOCCHANGERULE_error:
3600 		PF_RULES_WUNLOCK();
3601 		PF_CONFIG_UNLOCK();
3602 		pf_krule_free(newrule);
3603 		pf_kkif_free(kif);
3604 		break;
3605 	}
3606 
3607 	case DIOCCLRSTATESNV: {
3608 		error = pf_clearstates_nv((struct pfioc_nv *)addr);
3609 		break;
3610 	}
3611 
3612 	case DIOCKILLSTATESNV: {
3613 		error = pf_killstates_nv((struct pfioc_nv *)addr);
3614 		break;
3615 	}
3616 
3617 	case DIOCADDSTATE: {
3618 		struct pfioc_state		*ps = (struct pfioc_state *)addr;
3619 		struct pfsync_state_1301	*sp = &ps->state;
3620 
3621 		if (sp->timeout >= PFTM_MAX) {
3622 			error = EINVAL;
3623 			break;
3624 		}
3625 		if (V_pfsync_state_import_ptr != NULL) {
3626 			PF_RULES_RLOCK();
3627 			error = V_pfsync_state_import_ptr(
3628 			    (union pfsync_state_union *)sp, PFSYNC_SI_IOCTL,
3629 			    PFSYNC_MSG_VERSION_1301);
3630 			PF_RULES_RUNLOCK();
3631 		} else
3632 			error = EOPNOTSUPP;
3633 		break;
3634 	}
3635 
3636 	case DIOCGETSTATE: {
3637 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
3638 		struct pf_kstate	*s;
3639 
3640 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
3641 		if (s == NULL) {
3642 			error = ENOENT;
3643 			break;
3644 		}
3645 
3646 		pfsync_state_export((union pfsync_state_union*)&ps->state,
3647 		    s, PFSYNC_MSG_VERSION_1301);
3648 		PF_STATE_UNLOCK(s);
3649 		break;
3650 	}
3651 
3652 	case DIOCGETSTATENV: {
3653 		error = pf_getstate((struct pfioc_nv *)addr);
3654 		break;
3655 	}
3656 
3657 	case DIOCGETSTATES: {
3658 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
3659 		struct pf_kstate	*s;
3660 		struct pfsync_state_1301	*pstore, *p;
3661 		int			 i, nr;
3662 		size_t			 slice_count = 16, count;
3663 		void			*out;
3664 
3665 		if (ps->ps_len <= 0) {
3666 			nr = uma_zone_get_cur(V_pf_state_z);
3667 			ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3668 			break;
3669 		}
3670 
3671 		out = ps->ps_states;
3672 		pstore = mallocarray(slice_count,
3673 		    sizeof(struct pfsync_state_1301), M_TEMP, M_WAITOK | M_ZERO);
3674 		nr = 0;
3675 
3676 		for (i = 0; i <= pf_hashmask; i++) {
3677 			struct pf_idhash *ih = &V_pf_idhash[i];
3678 
3679 DIOCGETSTATES_retry:
3680 			p = pstore;
3681 
3682 			if (LIST_EMPTY(&ih->states))
3683 				continue;
3684 
3685 			PF_HASHROW_LOCK(ih);
3686 			count = 0;
3687 			LIST_FOREACH(s, &ih->states, entry) {
3688 				if (s->timeout == PFTM_UNLINKED)
3689 					continue;
3690 				count++;
3691 			}
3692 
3693 			if (count > slice_count) {
3694 				PF_HASHROW_UNLOCK(ih);
3695 				free(pstore, M_TEMP);
3696 				slice_count = count * 2;
3697 				pstore = mallocarray(slice_count,
3698 				    sizeof(struct pfsync_state_1301), M_TEMP,
3699 				    M_WAITOK | M_ZERO);
3700 				goto DIOCGETSTATES_retry;
3701 			}
3702 
3703 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
3704 				PF_HASHROW_UNLOCK(ih);
3705 				goto DIOCGETSTATES_full;
3706 			}
3707 
3708 			LIST_FOREACH(s, &ih->states, entry) {
3709 				if (s->timeout == PFTM_UNLINKED)
3710 					continue;
3711 
3712 				pfsync_state_export((union pfsync_state_union*)p,
3713 				    s, PFSYNC_MSG_VERSION_1301);
3714 				p++;
3715 				nr++;
3716 			}
3717 			PF_HASHROW_UNLOCK(ih);
3718 			error = copyout(pstore, out,
3719 			    sizeof(struct pfsync_state_1301) * count);
3720 			if (error)
3721 				break;
3722 			out = ps->ps_states + nr;
3723 		}
3724 DIOCGETSTATES_full:
3725 		ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3726 		free(pstore, M_TEMP);
3727 
3728 		break;
3729 	}
3730 
3731 	case DIOCGETSTATESV2: {
3732 		struct pfioc_states_v2	*ps = (struct pfioc_states_v2 *)addr;
3733 		struct pf_kstate	*s;
3734 		struct pf_state_export	*pstore, *p;
3735 		int i, nr;
3736 		size_t slice_count = 16, count;
3737 		void *out;
3738 
3739 		if (ps->ps_req_version > PF_STATE_VERSION) {
3740 			error = ENOTSUP;
3741 			break;
3742 		}
3743 
3744 		if (ps->ps_len <= 0) {
3745 			nr = uma_zone_get_cur(V_pf_state_z);
3746 			ps->ps_len = sizeof(struct pf_state_export) * nr;
3747 			break;
3748 		}
3749 
3750 		out = ps->ps_states;
3751 		pstore = mallocarray(slice_count,
3752 		    sizeof(struct pf_state_export), M_TEMP, M_WAITOK | M_ZERO);
3753 		nr = 0;
3754 
3755 		for (i = 0; i <= pf_hashmask; i++) {
3756 			struct pf_idhash *ih = &V_pf_idhash[i];
3757 
3758 DIOCGETSTATESV2_retry:
3759 			p = pstore;
3760 
3761 			if (LIST_EMPTY(&ih->states))
3762 				continue;
3763 
3764 			PF_HASHROW_LOCK(ih);
3765 			count = 0;
3766 			LIST_FOREACH(s, &ih->states, entry) {
3767 				if (s->timeout == PFTM_UNLINKED)
3768 					continue;
3769 				count++;
3770 			}
3771 
3772 			if (count > slice_count) {
3773 				PF_HASHROW_UNLOCK(ih);
3774 				free(pstore, M_TEMP);
3775 				slice_count = count * 2;
3776 				pstore = mallocarray(slice_count,
3777 				    sizeof(struct pf_state_export), M_TEMP,
3778 				    M_WAITOK | M_ZERO);
3779 				goto DIOCGETSTATESV2_retry;
3780 			}
3781 
3782 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
3783 				PF_HASHROW_UNLOCK(ih);
3784 				goto DIOCGETSTATESV2_full;
3785 			}
3786 
3787 			LIST_FOREACH(s, &ih->states, entry) {
3788 				if (s->timeout == PFTM_UNLINKED)
3789 					continue;
3790 
3791 				pf_state_export(p, s);
3792 				p++;
3793 				nr++;
3794 			}
3795 			PF_HASHROW_UNLOCK(ih);
3796 			error = copyout(pstore, out,
3797 			    sizeof(struct pf_state_export) * count);
3798 			if (error)
3799 				break;
3800 			out = ps->ps_states + nr;
3801 		}
3802 DIOCGETSTATESV2_full:
3803 		ps->ps_len = nr * sizeof(struct pf_state_export);
3804 		free(pstore, M_TEMP);
3805 
3806 		break;
3807 	}
3808 
3809 	case DIOCGETSTATUSNV: {
3810 		error = pf_getstatus((struct pfioc_nv *)addr);
3811 		break;
3812 	}
3813 
3814 	case DIOCSETSTATUSIF: {
3815 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
3816 
3817 		if (pi->ifname[0] == 0) {
3818 			bzero(V_pf_status.ifname, IFNAMSIZ);
3819 			break;
3820 		}
3821 		PF_RULES_WLOCK();
3822 		error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
3823 		PF_RULES_WUNLOCK();
3824 		break;
3825 	}
3826 
3827 	case DIOCCLRSTATUS: {
3828 		PF_RULES_WLOCK();
3829 		for (int i = 0; i < PFRES_MAX; i++)
3830 			counter_u64_zero(V_pf_status.counters[i]);
3831 		for (int i = 0; i < FCNT_MAX; i++)
3832 			pf_counter_u64_zero(&V_pf_status.fcounters[i]);
3833 		for (int i = 0; i < SCNT_MAX; i++)
3834 			counter_u64_zero(V_pf_status.scounters[i]);
3835 		for (int i = 0; i < KLCNT_MAX; i++)
3836 			counter_u64_zero(V_pf_status.lcounters[i]);
3837 		V_pf_status.since = time_second;
3838 		if (*V_pf_status.ifname)
3839 			pfi_update_status(V_pf_status.ifname, NULL);
3840 		PF_RULES_WUNLOCK();
3841 		break;
3842 	}
3843 
3844 	case DIOCNATLOOK: {
3845 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
3846 		struct pf_state_key	*sk;
3847 		struct pf_kstate	*state;
3848 		struct pf_state_key_cmp	 key;
3849 		int			 m = 0, direction = pnl->direction;
3850 		int			 sidx, didx;
3851 
3852 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
3853 		sidx = (direction == PF_IN) ? 1 : 0;
3854 		didx = (direction == PF_IN) ? 0 : 1;
3855 
3856 		if (!pnl->proto ||
3857 		    PF_AZERO(&pnl->saddr, pnl->af) ||
3858 		    PF_AZERO(&pnl->daddr, pnl->af) ||
3859 		    ((pnl->proto == IPPROTO_TCP ||
3860 		    pnl->proto == IPPROTO_UDP) &&
3861 		    (!pnl->dport || !pnl->sport)))
3862 			error = EINVAL;
3863 		else {
3864 			bzero(&key, sizeof(key));
3865 			key.af = pnl->af;
3866 			key.proto = pnl->proto;
3867 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
3868 			key.port[sidx] = pnl->sport;
3869 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
3870 			key.port[didx] = pnl->dport;
3871 
3872 			state = pf_find_state_all(&key, direction, &m);
3873 			if (state == NULL) {
3874 				error = ENOENT;
3875 			} else {
3876 				if (m > 1) {
3877 					PF_STATE_UNLOCK(state);
3878 					error = E2BIG;	/* more than one state */
3879 				} else {
3880 					sk = state->key[sidx];
3881 					PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
3882 					pnl->rsport = sk->port[sidx];
3883 					PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
3884 					pnl->rdport = sk->port[didx];
3885 					PF_STATE_UNLOCK(state);
3886 				}
3887 			}
3888 		}
3889 		break;
3890 	}
3891 
3892 	case DIOCSETTIMEOUT: {
3893 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
3894 		int		 old;
3895 
3896 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
3897 		    pt->seconds < 0) {
3898 			error = EINVAL;
3899 			break;
3900 		}
3901 		PF_RULES_WLOCK();
3902 		old = V_pf_default_rule.timeout[pt->timeout];
3903 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
3904 			pt->seconds = 1;
3905 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
3906 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
3907 			wakeup(pf_purge_thread);
3908 		pt->seconds = old;
3909 		PF_RULES_WUNLOCK();
3910 		break;
3911 	}
3912 
3913 	case DIOCGETTIMEOUT: {
3914 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
3915 
3916 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
3917 			error = EINVAL;
3918 			break;
3919 		}
3920 		PF_RULES_RLOCK();
3921 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
3922 		PF_RULES_RUNLOCK();
3923 		break;
3924 	}
3925 
3926 	case DIOCGETLIMIT: {
3927 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
3928 
3929 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
3930 			error = EINVAL;
3931 			break;
3932 		}
3933 		PF_RULES_RLOCK();
3934 		pl->limit = V_pf_limits[pl->index].limit;
3935 		PF_RULES_RUNLOCK();
3936 		break;
3937 	}
3938 
3939 	case DIOCSETLIMIT: {
3940 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
3941 		int			 old_limit;
3942 
3943 		PF_RULES_WLOCK();
3944 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
3945 		    V_pf_limits[pl->index].zone == NULL) {
3946 			PF_RULES_WUNLOCK();
3947 			error = EINVAL;
3948 			break;
3949 		}
3950 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
3951 		old_limit = V_pf_limits[pl->index].limit;
3952 		V_pf_limits[pl->index].limit = pl->limit;
3953 		pl->limit = old_limit;
3954 		PF_RULES_WUNLOCK();
3955 		break;
3956 	}
3957 
3958 	case DIOCSETDEBUG: {
3959 		u_int32_t	*level = (u_int32_t *)addr;
3960 
3961 		PF_RULES_WLOCK();
3962 		V_pf_status.debug = *level;
3963 		PF_RULES_WUNLOCK();
3964 		break;
3965 	}
3966 
3967 	case DIOCCLRRULECTRS: {
3968 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
3969 		struct pf_kruleset	*ruleset = &pf_main_ruleset;
3970 		struct pf_krule		*rule;
3971 
3972 		PF_RULES_WLOCK();
3973 		TAILQ_FOREACH(rule,
3974 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
3975 			pf_counter_u64_zero(&rule->evaluations);
3976 			for (int i = 0; i < 2; i++) {
3977 				pf_counter_u64_zero(&rule->packets[i]);
3978 				pf_counter_u64_zero(&rule->bytes[i]);
3979 			}
3980 		}
3981 		PF_RULES_WUNLOCK();
3982 		break;
3983 	}
3984 
3985 	case DIOCGIFSPEEDV0:
3986 	case DIOCGIFSPEEDV1: {
3987 		struct pf_ifspeed_v1	*psp = (struct pf_ifspeed_v1 *)addr;
3988 		struct pf_ifspeed_v1	ps;
3989 		struct ifnet		*ifp;
3990 
3991 		if (psp->ifname[0] == '\0') {
3992 			error = EINVAL;
3993 			break;
3994 		}
3995 
3996 		error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ);
3997 		if (error != 0)
3998 			break;
3999 		ifp = ifunit(ps.ifname);
4000 		if (ifp != NULL) {
4001 			psp->baudrate32 =
4002 			    (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
4003 			if (cmd == DIOCGIFSPEEDV1)
4004 				psp->baudrate = ifp->if_baudrate;
4005 		} else {
4006 			error = EINVAL;
4007 		}
4008 		break;
4009 	}
4010 
4011 #ifdef ALTQ
4012 	case DIOCSTARTALTQ: {
4013 		struct pf_altq		*altq;
4014 
4015 		PF_RULES_WLOCK();
4016 		/* enable all altq interfaces on active list */
4017 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4018 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4019 				error = pf_enable_altq(altq);
4020 				if (error != 0)
4021 					break;
4022 			}
4023 		}
4024 		if (error == 0)
4025 			V_pf_altq_running = 1;
4026 		PF_RULES_WUNLOCK();
4027 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
4028 		break;
4029 	}
4030 
4031 	case DIOCSTOPALTQ: {
4032 		struct pf_altq		*altq;
4033 
4034 		PF_RULES_WLOCK();
4035 		/* disable all altq interfaces on active list */
4036 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4037 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4038 				error = pf_disable_altq(altq);
4039 				if (error != 0)
4040 					break;
4041 			}
4042 		}
4043 		if (error == 0)
4044 			V_pf_altq_running = 0;
4045 		PF_RULES_WUNLOCK();
4046 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
4047 		break;
4048 	}
4049 
4050 	case DIOCADDALTQV0:
4051 	case DIOCADDALTQV1: {
4052 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4053 		struct pf_altq		*altq, *a;
4054 		struct ifnet		*ifp;
4055 
4056 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
4057 		error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
4058 		if (error)
4059 			break;
4060 		altq->local_flags = 0;
4061 
4062 		PF_RULES_WLOCK();
4063 		if (pa->ticket != V_ticket_altqs_inactive) {
4064 			PF_RULES_WUNLOCK();
4065 			free(altq, M_PFALTQ);
4066 			error = EBUSY;
4067 			break;
4068 		}
4069 
4070 		/*
4071 		 * if this is for a queue, find the discipline and
4072 		 * copy the necessary fields
4073 		 */
4074 		if (altq->qname[0] != 0) {
4075 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
4076 				PF_RULES_WUNLOCK();
4077 				error = EBUSY;
4078 				free(altq, M_PFALTQ);
4079 				break;
4080 			}
4081 			altq->altq_disc = NULL;
4082 			TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
4083 				if (strncmp(a->ifname, altq->ifname,
4084 				    IFNAMSIZ) == 0) {
4085 					altq->altq_disc = a->altq_disc;
4086 					break;
4087 				}
4088 			}
4089 		}
4090 
4091 		if ((ifp = ifunit(altq->ifname)) == NULL)
4092 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
4093 		else
4094 			error = altq_add(ifp, altq);
4095 
4096 		if (error) {
4097 			PF_RULES_WUNLOCK();
4098 			free(altq, M_PFALTQ);
4099 			break;
4100 		}
4101 
4102 		if (altq->qname[0] != 0)
4103 			TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
4104 		else
4105 			TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
4106 		/* version error check done on import above */
4107 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4108 		PF_RULES_WUNLOCK();
4109 		break;
4110 	}
4111 
4112 	case DIOCGETALTQSV0:
4113 	case DIOCGETALTQSV1: {
4114 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4115 		struct pf_altq		*altq;
4116 
4117 		PF_RULES_RLOCK();
4118 		pa->nr = 0;
4119 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
4120 			pa->nr++;
4121 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
4122 			pa->nr++;
4123 		pa->ticket = V_ticket_altqs_active;
4124 		PF_RULES_RUNLOCK();
4125 		break;
4126 	}
4127 
4128 	case DIOCGETALTQV0:
4129 	case DIOCGETALTQV1: {
4130 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4131 		struct pf_altq		*altq;
4132 
4133 		PF_RULES_RLOCK();
4134 		if (pa->ticket != V_ticket_altqs_active) {
4135 			PF_RULES_RUNLOCK();
4136 			error = EBUSY;
4137 			break;
4138 		}
4139 		altq = pf_altq_get_nth_active(pa->nr);
4140 		if (altq == NULL) {
4141 			PF_RULES_RUNLOCK();
4142 			error = EBUSY;
4143 			break;
4144 		}
4145 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4146 		PF_RULES_RUNLOCK();
4147 		break;
4148 	}
4149 
4150 	case DIOCCHANGEALTQV0:
4151 	case DIOCCHANGEALTQV1:
4152 		/* CHANGEALTQ not supported yet! */
4153 		error = ENODEV;
4154 		break;
4155 
4156 	case DIOCGETQSTATSV0:
4157 	case DIOCGETQSTATSV1: {
4158 		struct pfioc_qstats_v1	*pq = (struct pfioc_qstats_v1 *)addr;
4159 		struct pf_altq		*altq;
4160 		int			 nbytes;
4161 		u_int32_t		 version;
4162 
4163 		PF_RULES_RLOCK();
4164 		if (pq->ticket != V_ticket_altqs_active) {
4165 			PF_RULES_RUNLOCK();
4166 			error = EBUSY;
4167 			break;
4168 		}
4169 		nbytes = pq->nbytes;
4170 		altq = pf_altq_get_nth_active(pq->nr);
4171 		if (altq == NULL) {
4172 			PF_RULES_RUNLOCK();
4173 			error = EBUSY;
4174 			break;
4175 		}
4176 
4177 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
4178 			PF_RULES_RUNLOCK();
4179 			error = ENXIO;
4180 			break;
4181 		}
4182 		PF_RULES_RUNLOCK();
4183 		if (cmd == DIOCGETQSTATSV0)
4184 			version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
4185 		else
4186 			version = pq->version;
4187 		error = altq_getqstats(altq, pq->buf, &nbytes, version);
4188 		if (error == 0) {
4189 			pq->scheduler = altq->scheduler;
4190 			pq->nbytes = nbytes;
4191 		}
4192 		break;
4193 	}
4194 #endif /* ALTQ */
4195 
4196 	case DIOCBEGINADDRS: {
4197 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4198 
4199 		PF_RULES_WLOCK();
4200 		pf_empty_kpool(&V_pf_pabuf);
4201 		pp->ticket = ++V_ticket_pabuf;
4202 		PF_RULES_WUNLOCK();
4203 		break;
4204 	}
4205 
4206 	case DIOCADDADDR: {
4207 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4208 		struct pf_kpooladdr	*pa;
4209 		struct pfi_kkif		*kif = NULL;
4210 
4211 #ifndef INET
4212 		if (pp->af == AF_INET) {
4213 			error = EAFNOSUPPORT;
4214 			break;
4215 		}
4216 #endif /* INET */
4217 #ifndef INET6
4218 		if (pp->af == AF_INET6) {
4219 			error = EAFNOSUPPORT;
4220 			break;
4221 		}
4222 #endif /* INET6 */
4223 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
4224 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
4225 		    pp->addr.addr.type != PF_ADDR_TABLE) {
4226 			error = EINVAL;
4227 			break;
4228 		}
4229 		if (pp->addr.addr.p.dyn != NULL) {
4230 			error = EINVAL;
4231 			break;
4232 		}
4233 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
4234 		error = pf_pooladdr_to_kpooladdr(&pp->addr, pa);
4235 		if (error != 0)
4236 			break;
4237 		if (pa->ifname[0])
4238 			kif = pf_kkif_create(M_WAITOK);
4239 		PF_RULES_WLOCK();
4240 		if (pp->ticket != V_ticket_pabuf) {
4241 			PF_RULES_WUNLOCK();
4242 			if (pa->ifname[0])
4243 				pf_kkif_free(kif);
4244 			free(pa, M_PFRULE);
4245 			error = EBUSY;
4246 			break;
4247 		}
4248 		if (pa->ifname[0]) {
4249 			pa->kif = pfi_kkif_attach(kif, pa->ifname);
4250 			kif = NULL;
4251 			pfi_kkif_ref(pa->kif);
4252 		} else
4253 			pa->kif = NULL;
4254 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
4255 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
4256 			if (pa->ifname[0])
4257 				pfi_kkif_unref(pa->kif);
4258 			PF_RULES_WUNLOCK();
4259 			free(pa, M_PFRULE);
4260 			break;
4261 		}
4262 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
4263 		PF_RULES_WUNLOCK();
4264 		break;
4265 	}
4266 
4267 	case DIOCGETADDRS: {
4268 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4269 		struct pf_kpool		*pool;
4270 		struct pf_kpooladdr	*pa;
4271 
4272 		pp->anchor[sizeof(pp->anchor) - 1] = 0;
4273 		pp->nr = 0;
4274 
4275 		PF_RULES_RLOCK();
4276 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
4277 		    pp->r_num, 0, 1, 0);
4278 		if (pool == NULL) {
4279 			PF_RULES_RUNLOCK();
4280 			error = EBUSY;
4281 			break;
4282 		}
4283 		TAILQ_FOREACH(pa, &pool->list, entries)
4284 			pp->nr++;
4285 		PF_RULES_RUNLOCK();
4286 		break;
4287 	}
4288 
4289 	case DIOCGETADDR: {
4290 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4291 		struct pf_kpool		*pool;
4292 		struct pf_kpooladdr	*pa;
4293 		u_int32_t		 nr = 0;
4294 
4295 		pp->anchor[sizeof(pp->anchor) - 1] = 0;
4296 
4297 		PF_RULES_RLOCK();
4298 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
4299 		    pp->r_num, 0, 1, 1);
4300 		if (pool == NULL) {
4301 			PF_RULES_RUNLOCK();
4302 			error = EBUSY;
4303 			break;
4304 		}
4305 		pa = TAILQ_FIRST(&pool->list);
4306 		while ((pa != NULL) && (nr < pp->nr)) {
4307 			pa = TAILQ_NEXT(pa, entries);
4308 			nr++;
4309 		}
4310 		if (pa == NULL) {
4311 			PF_RULES_RUNLOCK();
4312 			error = EBUSY;
4313 			break;
4314 		}
4315 		pf_kpooladdr_to_pooladdr(pa, &pp->addr);
4316 		pf_addr_copyout(&pp->addr.addr);
4317 		PF_RULES_RUNLOCK();
4318 		break;
4319 	}
4320 
4321 	case DIOCCHANGEADDR: {
4322 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
4323 		struct pf_kpool		*pool;
4324 		struct pf_kpooladdr	*oldpa = NULL, *newpa = NULL;
4325 		struct pf_kruleset	*ruleset;
4326 		struct pfi_kkif		*kif = NULL;
4327 
4328 		pca->anchor[sizeof(pca->anchor) - 1] = 0;
4329 
4330 		if (pca->action < PF_CHANGE_ADD_HEAD ||
4331 		    pca->action > PF_CHANGE_REMOVE) {
4332 			error = EINVAL;
4333 			break;
4334 		}
4335 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
4336 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
4337 		    pca->addr.addr.type != PF_ADDR_TABLE) {
4338 			error = EINVAL;
4339 			break;
4340 		}
4341 		if (pca->addr.addr.p.dyn != NULL) {
4342 			error = EINVAL;
4343 			break;
4344 		}
4345 
4346 		if (pca->action != PF_CHANGE_REMOVE) {
4347 #ifndef INET
4348 			if (pca->af == AF_INET) {
4349 				error = EAFNOSUPPORT;
4350 				break;
4351 			}
4352 #endif /* INET */
4353 #ifndef INET6
4354 			if (pca->af == AF_INET6) {
4355 				error = EAFNOSUPPORT;
4356 				break;
4357 			}
4358 #endif /* INET6 */
4359 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
4360 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
4361 			if (newpa->ifname[0])
4362 				kif = pf_kkif_create(M_WAITOK);
4363 			newpa->kif = NULL;
4364 		}
4365 #define	ERROUT(x)	ERROUT_IOCTL(DIOCCHANGEADDR_error, x)
4366 		PF_RULES_WLOCK();
4367 		ruleset = pf_find_kruleset(pca->anchor);
4368 		if (ruleset == NULL)
4369 			ERROUT(EBUSY);
4370 
4371 		pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
4372 		    pca->r_num, pca->r_last, 1, 1);
4373 		if (pool == NULL)
4374 			ERROUT(EBUSY);
4375 
4376 		if (pca->action != PF_CHANGE_REMOVE) {
4377 			if (newpa->ifname[0]) {
4378 				newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
4379 				pfi_kkif_ref(newpa->kif);
4380 				kif = NULL;
4381 			}
4382 
4383 			switch (newpa->addr.type) {
4384 			case PF_ADDR_DYNIFTL:
4385 				error = pfi_dynaddr_setup(&newpa->addr,
4386 				    pca->af);
4387 				break;
4388 			case PF_ADDR_TABLE:
4389 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
4390 				    newpa->addr.v.tblname);
4391 				if (newpa->addr.p.tbl == NULL)
4392 					error = ENOMEM;
4393 				break;
4394 			}
4395 			if (error)
4396 				goto DIOCCHANGEADDR_error;
4397 		}
4398 
4399 		switch (pca->action) {
4400 		case PF_CHANGE_ADD_HEAD:
4401 			oldpa = TAILQ_FIRST(&pool->list);
4402 			break;
4403 		case PF_CHANGE_ADD_TAIL:
4404 			oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
4405 			break;
4406 		default:
4407 			oldpa = TAILQ_FIRST(&pool->list);
4408 			for (int i = 0; oldpa && i < pca->nr; i++)
4409 				oldpa = TAILQ_NEXT(oldpa, entries);
4410 
4411 			if (oldpa == NULL)
4412 				ERROUT(EINVAL);
4413 		}
4414 
4415 		if (pca->action == PF_CHANGE_REMOVE) {
4416 			TAILQ_REMOVE(&pool->list, oldpa, entries);
4417 			switch (oldpa->addr.type) {
4418 			case PF_ADDR_DYNIFTL:
4419 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
4420 				break;
4421 			case PF_ADDR_TABLE:
4422 				pfr_detach_table(oldpa->addr.p.tbl);
4423 				break;
4424 			}
4425 			if (oldpa->kif)
4426 				pfi_kkif_unref(oldpa->kif);
4427 			free(oldpa, M_PFRULE);
4428 		} else {
4429 			if (oldpa == NULL)
4430 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
4431 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
4432 			    pca->action == PF_CHANGE_ADD_BEFORE)
4433 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
4434 			else
4435 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
4436 				    newpa, entries);
4437 		}
4438 
4439 		pool->cur = TAILQ_FIRST(&pool->list);
4440 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
4441 		PF_RULES_WUNLOCK();
4442 		break;
4443 
4444 #undef ERROUT
4445 DIOCCHANGEADDR_error:
4446 		if (newpa != NULL) {
4447 			if (newpa->kif)
4448 				pfi_kkif_unref(newpa->kif);
4449 			free(newpa, M_PFRULE);
4450 		}
4451 		PF_RULES_WUNLOCK();
4452 		pf_kkif_free(kif);
4453 		break;
4454 	}
4455 
4456 	case DIOCGETRULESETS: {
4457 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
4458 		struct pf_kruleset	*ruleset;
4459 		struct pf_kanchor	*anchor;
4460 
4461 		pr->path[sizeof(pr->path) - 1] = 0;
4462 
4463 		PF_RULES_RLOCK();
4464 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4465 			PF_RULES_RUNLOCK();
4466 			error = ENOENT;
4467 			break;
4468 		}
4469 		pr->nr = 0;
4470 		if (ruleset->anchor == NULL) {
4471 			/* XXX kludge for pf_main_ruleset */
4472 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4473 				if (anchor->parent == NULL)
4474 					pr->nr++;
4475 		} else {
4476 			RB_FOREACH(anchor, pf_kanchor_node,
4477 			    &ruleset->anchor->children)
4478 				pr->nr++;
4479 		}
4480 		PF_RULES_RUNLOCK();
4481 		break;
4482 	}
4483 
4484 	case DIOCGETRULESET: {
4485 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
4486 		struct pf_kruleset	*ruleset;
4487 		struct pf_kanchor	*anchor;
4488 		u_int32_t		 nr = 0;
4489 
4490 		pr->path[sizeof(pr->path) - 1] = 0;
4491 
4492 		PF_RULES_RLOCK();
4493 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4494 			PF_RULES_RUNLOCK();
4495 			error = ENOENT;
4496 			break;
4497 		}
4498 		pr->name[0] = 0;
4499 		if (ruleset->anchor == NULL) {
4500 			/* XXX kludge for pf_main_ruleset */
4501 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4502 				if (anchor->parent == NULL && nr++ == pr->nr) {
4503 					strlcpy(pr->name, anchor->name,
4504 					    sizeof(pr->name));
4505 					break;
4506 				}
4507 		} else {
4508 			RB_FOREACH(anchor, pf_kanchor_node,
4509 			    &ruleset->anchor->children)
4510 				if (nr++ == pr->nr) {
4511 					strlcpy(pr->name, anchor->name,
4512 					    sizeof(pr->name));
4513 					break;
4514 				}
4515 		}
4516 		if (!pr->name[0])
4517 			error = EBUSY;
4518 		PF_RULES_RUNLOCK();
4519 		break;
4520 	}
4521 
4522 	case DIOCRCLRTABLES: {
4523 		struct pfioc_table *io = (struct pfioc_table *)addr;
4524 
4525 		if (io->pfrio_esize != 0) {
4526 			error = ENODEV;
4527 			break;
4528 		}
4529 		PF_RULES_WLOCK();
4530 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
4531 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4532 		PF_RULES_WUNLOCK();
4533 		break;
4534 	}
4535 
4536 	case DIOCRADDTABLES: {
4537 		struct pfioc_table *io = (struct pfioc_table *)addr;
4538 		struct pfr_table *pfrts;
4539 		size_t totlen;
4540 
4541 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4542 			error = ENODEV;
4543 			break;
4544 		}
4545 
4546 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4547 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4548 			error = ENOMEM;
4549 			break;
4550 		}
4551 
4552 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4553 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4554 		    M_TEMP, M_WAITOK);
4555 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4556 		if (error) {
4557 			free(pfrts, M_TEMP);
4558 			break;
4559 		}
4560 		PF_RULES_WLOCK();
4561 		error = pfr_add_tables(pfrts, io->pfrio_size,
4562 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4563 		PF_RULES_WUNLOCK();
4564 		free(pfrts, M_TEMP);
4565 		break;
4566 	}
4567 
4568 	case DIOCRDELTABLES: {
4569 		struct pfioc_table *io = (struct pfioc_table *)addr;
4570 		struct pfr_table *pfrts;
4571 		size_t totlen;
4572 
4573 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4574 			error = ENODEV;
4575 			break;
4576 		}
4577 
4578 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4579 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4580 			error = ENOMEM;
4581 			break;
4582 		}
4583 
4584 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4585 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4586 		    M_TEMP, M_WAITOK);
4587 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4588 		if (error) {
4589 			free(pfrts, M_TEMP);
4590 			break;
4591 		}
4592 		PF_RULES_WLOCK();
4593 		error = pfr_del_tables(pfrts, io->pfrio_size,
4594 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4595 		PF_RULES_WUNLOCK();
4596 		free(pfrts, M_TEMP);
4597 		break;
4598 	}
4599 
4600 	case DIOCRGETTABLES: {
4601 		struct pfioc_table *io = (struct pfioc_table *)addr;
4602 		struct pfr_table *pfrts;
4603 		size_t totlen;
4604 		int n;
4605 
4606 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4607 			error = ENODEV;
4608 			break;
4609 		}
4610 		PF_RULES_RLOCK();
4611 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4612 		if (n < 0) {
4613 			PF_RULES_RUNLOCK();
4614 			error = EINVAL;
4615 			break;
4616 		}
4617 		io->pfrio_size = min(io->pfrio_size, n);
4618 
4619 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4620 
4621 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4622 		    M_TEMP, M_NOWAIT | M_ZERO);
4623 		if (pfrts == NULL) {
4624 			error = ENOMEM;
4625 			PF_RULES_RUNLOCK();
4626 			break;
4627 		}
4628 		error = pfr_get_tables(&io->pfrio_table, pfrts,
4629 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4630 		PF_RULES_RUNLOCK();
4631 		if (error == 0)
4632 			error = copyout(pfrts, io->pfrio_buffer, totlen);
4633 		free(pfrts, M_TEMP);
4634 		break;
4635 	}
4636 
4637 	case DIOCRGETTSTATS: {
4638 		struct pfioc_table *io = (struct pfioc_table *)addr;
4639 		struct pfr_tstats *pfrtstats;
4640 		size_t totlen;
4641 		int n;
4642 
4643 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
4644 			error = ENODEV;
4645 			break;
4646 		}
4647 		PF_TABLE_STATS_LOCK();
4648 		PF_RULES_RLOCK();
4649 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4650 		if (n < 0) {
4651 			PF_RULES_RUNLOCK();
4652 			PF_TABLE_STATS_UNLOCK();
4653 			error = EINVAL;
4654 			break;
4655 		}
4656 		io->pfrio_size = min(io->pfrio_size, n);
4657 
4658 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
4659 		pfrtstats = mallocarray(io->pfrio_size,
4660 		    sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT | M_ZERO);
4661 		if (pfrtstats == NULL) {
4662 			error = ENOMEM;
4663 			PF_RULES_RUNLOCK();
4664 			PF_TABLE_STATS_UNLOCK();
4665 			break;
4666 		}
4667 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
4668 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4669 		PF_RULES_RUNLOCK();
4670 		PF_TABLE_STATS_UNLOCK();
4671 		if (error == 0)
4672 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
4673 		free(pfrtstats, M_TEMP);
4674 		break;
4675 	}
4676 
4677 	case DIOCRCLRTSTATS: {
4678 		struct pfioc_table *io = (struct pfioc_table *)addr;
4679 		struct pfr_table *pfrts;
4680 		size_t totlen;
4681 
4682 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4683 			error = ENODEV;
4684 			break;
4685 		}
4686 
4687 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4688 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4689 			/* We used to count tables and use the minimum required
4690 			 * size, so we didn't fail on overly large requests.
4691 			 * Keep doing so. */
4692 			io->pfrio_size = pf_ioctl_maxcount;
4693 			break;
4694 		}
4695 
4696 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4697 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4698 		    M_TEMP, M_WAITOK);
4699 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4700 		if (error) {
4701 			free(pfrts, M_TEMP);
4702 			break;
4703 		}
4704 
4705 		PF_TABLE_STATS_LOCK();
4706 		PF_RULES_RLOCK();
4707 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
4708 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4709 		PF_RULES_RUNLOCK();
4710 		PF_TABLE_STATS_UNLOCK();
4711 		free(pfrts, M_TEMP);
4712 		break;
4713 	}
4714 
4715 	case DIOCRSETTFLAGS: {
4716 		struct pfioc_table *io = (struct pfioc_table *)addr;
4717 		struct pfr_table *pfrts;
4718 		size_t totlen;
4719 		int n;
4720 
4721 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4722 			error = ENODEV;
4723 			break;
4724 		}
4725 
4726 		PF_RULES_RLOCK();
4727 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4728 		if (n < 0) {
4729 			PF_RULES_RUNLOCK();
4730 			error = EINVAL;
4731 			break;
4732 		}
4733 
4734 		io->pfrio_size = min(io->pfrio_size, n);
4735 		PF_RULES_RUNLOCK();
4736 
4737 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4738 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4739 		    M_TEMP, M_WAITOK);
4740 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4741 		if (error) {
4742 			free(pfrts, M_TEMP);
4743 			break;
4744 		}
4745 		PF_RULES_WLOCK();
4746 		error = pfr_set_tflags(pfrts, io->pfrio_size,
4747 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
4748 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4749 		PF_RULES_WUNLOCK();
4750 		free(pfrts, M_TEMP);
4751 		break;
4752 	}
4753 
4754 	case DIOCRCLRADDRS: {
4755 		struct pfioc_table *io = (struct pfioc_table *)addr;
4756 
4757 		if (io->pfrio_esize != 0) {
4758 			error = ENODEV;
4759 			break;
4760 		}
4761 		PF_RULES_WLOCK();
4762 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
4763 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4764 		PF_RULES_WUNLOCK();
4765 		break;
4766 	}
4767 
4768 	case DIOCRADDADDRS: {
4769 		struct pfioc_table *io = (struct pfioc_table *)addr;
4770 		struct pfr_addr *pfras;
4771 		size_t totlen;
4772 
4773 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4774 			error = ENODEV;
4775 			break;
4776 		}
4777 		if (io->pfrio_size < 0 ||
4778 		    io->pfrio_size > pf_ioctl_maxcount ||
4779 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4780 			error = EINVAL;
4781 			break;
4782 		}
4783 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4784 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4785 		    M_TEMP, M_WAITOK);
4786 		error = copyin(io->pfrio_buffer, pfras, totlen);
4787 		if (error) {
4788 			free(pfras, M_TEMP);
4789 			break;
4790 		}
4791 		PF_RULES_WLOCK();
4792 		error = pfr_add_addrs(&io->pfrio_table, pfras,
4793 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
4794 		    PFR_FLAG_USERIOCTL);
4795 		PF_RULES_WUNLOCK();
4796 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4797 			error = copyout(pfras, io->pfrio_buffer, totlen);
4798 		free(pfras, M_TEMP);
4799 		break;
4800 	}
4801 
4802 	case DIOCRDELADDRS: {
4803 		struct pfioc_table *io = (struct pfioc_table *)addr;
4804 		struct pfr_addr *pfras;
4805 		size_t totlen;
4806 
4807 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4808 			error = ENODEV;
4809 			break;
4810 		}
4811 		if (io->pfrio_size < 0 ||
4812 		    io->pfrio_size > pf_ioctl_maxcount ||
4813 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4814 			error = EINVAL;
4815 			break;
4816 		}
4817 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4818 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4819 		    M_TEMP, M_WAITOK);
4820 		error = copyin(io->pfrio_buffer, pfras, totlen);
4821 		if (error) {
4822 			free(pfras, M_TEMP);
4823 			break;
4824 		}
4825 		PF_RULES_WLOCK();
4826 		error = pfr_del_addrs(&io->pfrio_table, pfras,
4827 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
4828 		    PFR_FLAG_USERIOCTL);
4829 		PF_RULES_WUNLOCK();
4830 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4831 			error = copyout(pfras, io->pfrio_buffer, totlen);
4832 		free(pfras, M_TEMP);
4833 		break;
4834 	}
4835 
4836 	case DIOCRSETADDRS: {
4837 		struct pfioc_table *io = (struct pfioc_table *)addr;
4838 		struct pfr_addr *pfras;
4839 		size_t totlen, count;
4840 
4841 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4842 			error = ENODEV;
4843 			break;
4844 		}
4845 		if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
4846 			error = EINVAL;
4847 			break;
4848 		}
4849 		count = max(io->pfrio_size, io->pfrio_size2);
4850 		if (count > pf_ioctl_maxcount ||
4851 		    WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
4852 			error = EINVAL;
4853 			break;
4854 		}
4855 		totlen = count * sizeof(struct pfr_addr);
4856 		pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
4857 		    M_WAITOK);
4858 		error = copyin(io->pfrio_buffer, pfras, totlen);
4859 		if (error) {
4860 			free(pfras, M_TEMP);
4861 			break;
4862 		}
4863 		PF_RULES_WLOCK();
4864 		error = pfr_set_addrs(&io->pfrio_table, pfras,
4865 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
4866 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
4867 		    PFR_FLAG_USERIOCTL, 0);
4868 		PF_RULES_WUNLOCK();
4869 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4870 			error = copyout(pfras, io->pfrio_buffer, totlen);
4871 		free(pfras, M_TEMP);
4872 		break;
4873 	}
4874 
4875 	case DIOCRGETADDRS: {
4876 		struct pfioc_table *io = (struct pfioc_table *)addr;
4877 		struct pfr_addr *pfras;
4878 		size_t totlen;
4879 
4880 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4881 			error = ENODEV;
4882 			break;
4883 		}
4884 		if (io->pfrio_size < 0 ||
4885 		    io->pfrio_size > pf_ioctl_maxcount ||
4886 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4887 			error = EINVAL;
4888 			break;
4889 		}
4890 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4891 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4892 		    M_TEMP, M_WAITOK | M_ZERO);
4893 		PF_RULES_RLOCK();
4894 		error = pfr_get_addrs(&io->pfrio_table, pfras,
4895 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4896 		PF_RULES_RUNLOCK();
4897 		if (error == 0)
4898 			error = copyout(pfras, io->pfrio_buffer, totlen);
4899 		free(pfras, M_TEMP);
4900 		break;
4901 	}
4902 
4903 	case DIOCRGETASTATS: {
4904 		struct pfioc_table *io = (struct pfioc_table *)addr;
4905 		struct pfr_astats *pfrastats;
4906 		size_t totlen;
4907 
4908 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
4909 			error = ENODEV;
4910 			break;
4911 		}
4912 		if (io->pfrio_size < 0 ||
4913 		    io->pfrio_size > pf_ioctl_maxcount ||
4914 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
4915 			error = EINVAL;
4916 			break;
4917 		}
4918 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
4919 		pfrastats = mallocarray(io->pfrio_size,
4920 		    sizeof(struct pfr_astats), M_TEMP, M_WAITOK | M_ZERO);
4921 		PF_RULES_RLOCK();
4922 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
4923 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4924 		PF_RULES_RUNLOCK();
4925 		if (error == 0)
4926 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
4927 		free(pfrastats, M_TEMP);
4928 		break;
4929 	}
4930 
4931 	case DIOCRCLRASTATS: {
4932 		struct pfioc_table *io = (struct pfioc_table *)addr;
4933 		struct pfr_addr *pfras;
4934 		size_t totlen;
4935 
4936 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4937 			error = ENODEV;
4938 			break;
4939 		}
4940 		if (io->pfrio_size < 0 ||
4941 		    io->pfrio_size > pf_ioctl_maxcount ||
4942 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4943 			error = EINVAL;
4944 			break;
4945 		}
4946 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4947 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4948 		    M_TEMP, M_WAITOK);
4949 		error = copyin(io->pfrio_buffer, pfras, totlen);
4950 		if (error) {
4951 			free(pfras, M_TEMP);
4952 			break;
4953 		}
4954 		PF_RULES_WLOCK();
4955 		error = pfr_clr_astats(&io->pfrio_table, pfras,
4956 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
4957 		    PFR_FLAG_USERIOCTL);
4958 		PF_RULES_WUNLOCK();
4959 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4960 			error = copyout(pfras, io->pfrio_buffer, totlen);
4961 		free(pfras, M_TEMP);
4962 		break;
4963 	}
4964 
4965 	case DIOCRTSTADDRS: {
4966 		struct pfioc_table *io = (struct pfioc_table *)addr;
4967 		struct pfr_addr *pfras;
4968 		size_t totlen;
4969 
4970 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4971 			error = ENODEV;
4972 			break;
4973 		}
4974 		if (io->pfrio_size < 0 ||
4975 		    io->pfrio_size > pf_ioctl_maxcount ||
4976 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4977 			error = EINVAL;
4978 			break;
4979 		}
4980 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4981 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4982 		    M_TEMP, M_WAITOK);
4983 		error = copyin(io->pfrio_buffer, pfras, totlen);
4984 		if (error) {
4985 			free(pfras, M_TEMP);
4986 			break;
4987 		}
4988 		PF_RULES_RLOCK();
4989 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
4990 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
4991 		    PFR_FLAG_USERIOCTL);
4992 		PF_RULES_RUNLOCK();
4993 		if (error == 0)
4994 			error = copyout(pfras, io->pfrio_buffer, totlen);
4995 		free(pfras, M_TEMP);
4996 		break;
4997 	}
4998 
4999 	case DIOCRINADEFINE: {
5000 		struct pfioc_table *io = (struct pfioc_table *)addr;
5001 		struct pfr_addr *pfras;
5002 		size_t totlen;
5003 
5004 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5005 			error = ENODEV;
5006 			break;
5007 		}
5008 		if (io->pfrio_size < 0 ||
5009 		    io->pfrio_size > pf_ioctl_maxcount ||
5010 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5011 			error = EINVAL;
5012 			break;
5013 		}
5014 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
5015 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5016 		    M_TEMP, M_WAITOK);
5017 		error = copyin(io->pfrio_buffer, pfras, totlen);
5018 		if (error) {
5019 			free(pfras, M_TEMP);
5020 			break;
5021 		}
5022 		PF_RULES_WLOCK();
5023 		error = pfr_ina_define(&io->pfrio_table, pfras,
5024 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
5025 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5026 		PF_RULES_WUNLOCK();
5027 		free(pfras, M_TEMP);
5028 		break;
5029 	}
5030 
5031 	case DIOCOSFPADD: {
5032 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5033 		PF_RULES_WLOCK();
5034 		error = pf_osfp_add(io);
5035 		PF_RULES_WUNLOCK();
5036 		break;
5037 	}
5038 
5039 	case DIOCOSFPGET: {
5040 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5041 		PF_RULES_RLOCK();
5042 		error = pf_osfp_get(io);
5043 		PF_RULES_RUNLOCK();
5044 		break;
5045 	}
5046 
5047 	case DIOCXBEGIN: {
5048 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5049 		struct pfioc_trans_e	*ioes, *ioe;
5050 		size_t			 totlen;
5051 		int			 i;
5052 
5053 		if (io->esize != sizeof(*ioe)) {
5054 			error = ENODEV;
5055 			break;
5056 		}
5057 		if (io->size < 0 ||
5058 		    io->size > pf_ioctl_maxcount ||
5059 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5060 			error = EINVAL;
5061 			break;
5062 		}
5063 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5064 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5065 		    M_TEMP, M_WAITOK);
5066 		error = copyin(io->array, ioes, totlen);
5067 		if (error) {
5068 			free(ioes, M_TEMP);
5069 			break;
5070 		}
5071 		/* Ensure there's no more ethernet rules to clean up. */
5072 		NET_EPOCH_DRAIN_CALLBACKS();
5073 		PF_RULES_WLOCK();
5074 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5075 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5076 			switch (ioe->rs_num) {
5077 			case PF_RULESET_ETH:
5078 				if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) {
5079 					PF_RULES_WUNLOCK();
5080 					free(ioes, M_TEMP);
5081 					goto fail;
5082 				}
5083 				break;
5084 #ifdef ALTQ
5085 			case PF_RULESET_ALTQ:
5086 				if (ioe->anchor[0]) {
5087 					PF_RULES_WUNLOCK();
5088 					free(ioes, M_TEMP);
5089 					error = EINVAL;
5090 					goto fail;
5091 				}
5092 				if ((error = pf_begin_altq(&ioe->ticket))) {
5093 					PF_RULES_WUNLOCK();
5094 					free(ioes, M_TEMP);
5095 					goto fail;
5096 				}
5097 				break;
5098 #endif /* ALTQ */
5099 			case PF_RULESET_TABLE:
5100 			    {
5101 				struct pfr_table table;
5102 
5103 				bzero(&table, sizeof(table));
5104 				strlcpy(table.pfrt_anchor, ioe->anchor,
5105 				    sizeof(table.pfrt_anchor));
5106 				if ((error = pfr_ina_begin(&table,
5107 				    &ioe->ticket, NULL, 0))) {
5108 					PF_RULES_WUNLOCK();
5109 					free(ioes, M_TEMP);
5110 					goto fail;
5111 				}
5112 				break;
5113 			    }
5114 			default:
5115 				if ((error = pf_begin_rules(&ioe->ticket,
5116 				    ioe->rs_num, ioe->anchor))) {
5117 					PF_RULES_WUNLOCK();
5118 					free(ioes, M_TEMP);
5119 					goto fail;
5120 				}
5121 				break;
5122 			}
5123 		}
5124 		PF_RULES_WUNLOCK();
5125 		error = copyout(ioes, io->array, totlen);
5126 		free(ioes, M_TEMP);
5127 		break;
5128 	}
5129 
5130 	case DIOCXROLLBACK: {
5131 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5132 		struct pfioc_trans_e	*ioe, *ioes;
5133 		size_t			 totlen;
5134 		int			 i;
5135 
5136 		if (io->esize != sizeof(*ioe)) {
5137 			error = ENODEV;
5138 			break;
5139 		}
5140 		if (io->size < 0 ||
5141 		    io->size > pf_ioctl_maxcount ||
5142 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5143 			error = EINVAL;
5144 			break;
5145 		}
5146 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5147 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5148 		    M_TEMP, M_WAITOK);
5149 		error = copyin(io->array, ioes, totlen);
5150 		if (error) {
5151 			free(ioes, M_TEMP);
5152 			break;
5153 		}
5154 		PF_RULES_WLOCK();
5155 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5156 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5157 			switch (ioe->rs_num) {
5158 			case PF_RULESET_ETH:
5159 				if ((error = pf_rollback_eth(ioe->ticket,
5160 				    ioe->anchor))) {
5161 					PF_RULES_WUNLOCK();
5162 					free(ioes, M_TEMP);
5163 					goto fail; /* really bad */
5164 				}
5165 				break;
5166 #ifdef ALTQ
5167 			case PF_RULESET_ALTQ:
5168 				if (ioe->anchor[0]) {
5169 					PF_RULES_WUNLOCK();
5170 					free(ioes, M_TEMP);
5171 					error = EINVAL;
5172 					goto fail;
5173 				}
5174 				if ((error = pf_rollback_altq(ioe->ticket))) {
5175 					PF_RULES_WUNLOCK();
5176 					free(ioes, M_TEMP);
5177 					goto fail; /* really bad */
5178 				}
5179 				break;
5180 #endif /* ALTQ */
5181 			case PF_RULESET_TABLE:
5182 			    {
5183 				struct pfr_table table;
5184 
5185 				bzero(&table, sizeof(table));
5186 				strlcpy(table.pfrt_anchor, ioe->anchor,
5187 				    sizeof(table.pfrt_anchor));
5188 				if ((error = pfr_ina_rollback(&table,
5189 				    ioe->ticket, NULL, 0))) {
5190 					PF_RULES_WUNLOCK();
5191 					free(ioes, M_TEMP);
5192 					goto fail; /* really bad */
5193 				}
5194 				break;
5195 			    }
5196 			default:
5197 				if ((error = pf_rollback_rules(ioe->ticket,
5198 				    ioe->rs_num, ioe->anchor))) {
5199 					PF_RULES_WUNLOCK();
5200 					free(ioes, M_TEMP);
5201 					goto fail; /* really bad */
5202 				}
5203 				break;
5204 			}
5205 		}
5206 		PF_RULES_WUNLOCK();
5207 		free(ioes, M_TEMP);
5208 		break;
5209 	}
5210 
5211 	case DIOCXCOMMIT: {
5212 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5213 		struct pfioc_trans_e	*ioe, *ioes;
5214 		struct pf_kruleset	*rs;
5215 		struct pf_keth_ruleset	*ers;
5216 		size_t			 totlen;
5217 		int			 i;
5218 
5219 		if (io->esize != sizeof(*ioe)) {
5220 			error = ENODEV;
5221 			break;
5222 		}
5223 
5224 		if (io->size < 0 ||
5225 		    io->size > pf_ioctl_maxcount ||
5226 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5227 			error = EINVAL;
5228 			break;
5229 		}
5230 
5231 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5232 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5233 		    M_TEMP, M_WAITOK);
5234 		error = copyin(io->array, ioes, totlen);
5235 		if (error) {
5236 			free(ioes, M_TEMP);
5237 			break;
5238 		}
5239 		PF_RULES_WLOCK();
5240 		/* First makes sure everything will succeed. */
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 				ers = pf_find_keth_ruleset(ioe->anchor);
5246 				if (ers == NULL || ioe->ticket == 0 ||
5247 				    ioe->ticket != ers->inactive.ticket) {
5248 					PF_RULES_WUNLOCK();
5249 					free(ioes, M_TEMP);
5250 					error = EINVAL;
5251 					goto fail;
5252 				}
5253 				break;
5254 #ifdef ALTQ
5255 			case PF_RULESET_ALTQ:
5256 				if (ioe->anchor[0]) {
5257 					PF_RULES_WUNLOCK();
5258 					free(ioes, M_TEMP);
5259 					error = EINVAL;
5260 					goto fail;
5261 				}
5262 				if (!V_altqs_inactive_open || ioe->ticket !=
5263 				    V_ticket_altqs_inactive) {
5264 					PF_RULES_WUNLOCK();
5265 					free(ioes, M_TEMP);
5266 					error = EBUSY;
5267 					goto fail;
5268 				}
5269 				break;
5270 #endif /* ALTQ */
5271 			case PF_RULESET_TABLE:
5272 				rs = pf_find_kruleset(ioe->anchor);
5273 				if (rs == NULL || !rs->topen || ioe->ticket !=
5274 				    rs->tticket) {
5275 					PF_RULES_WUNLOCK();
5276 					free(ioes, M_TEMP);
5277 					error = EBUSY;
5278 					goto fail;
5279 				}
5280 				break;
5281 			default:
5282 				if (ioe->rs_num < 0 || ioe->rs_num >=
5283 				    PF_RULESET_MAX) {
5284 					PF_RULES_WUNLOCK();
5285 					free(ioes, M_TEMP);
5286 					error = EINVAL;
5287 					goto fail;
5288 				}
5289 				rs = pf_find_kruleset(ioe->anchor);
5290 				if (rs == NULL ||
5291 				    !rs->rules[ioe->rs_num].inactive.open ||
5292 				    rs->rules[ioe->rs_num].inactive.ticket !=
5293 				    ioe->ticket) {
5294 					PF_RULES_WUNLOCK();
5295 					free(ioes, M_TEMP);
5296 					error = EBUSY;
5297 					goto fail;
5298 				}
5299 				break;
5300 			}
5301 		}
5302 		/* Now do the commit - no errors should happen here. */
5303 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5304 			switch (ioe->rs_num) {
5305 			case PF_RULESET_ETH:
5306 				if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) {
5307 					PF_RULES_WUNLOCK();
5308 					free(ioes, M_TEMP);
5309 					goto fail; /* really bad */
5310 				}
5311 				break;
5312 #ifdef ALTQ
5313 			case PF_RULESET_ALTQ:
5314 				if ((error = pf_commit_altq(ioe->ticket))) {
5315 					PF_RULES_WUNLOCK();
5316 					free(ioes, M_TEMP);
5317 					goto fail; /* really bad */
5318 				}
5319 				break;
5320 #endif /* ALTQ */
5321 			case PF_RULESET_TABLE:
5322 			    {
5323 				struct pfr_table table;
5324 
5325 				bzero(&table, sizeof(table));
5326 				(void)strlcpy(table.pfrt_anchor, ioe->anchor,
5327 				    sizeof(table.pfrt_anchor));
5328 				if ((error = pfr_ina_commit(&table,
5329 				    ioe->ticket, NULL, NULL, 0))) {
5330 					PF_RULES_WUNLOCK();
5331 					free(ioes, M_TEMP);
5332 					goto fail; /* really bad */
5333 				}
5334 				break;
5335 			    }
5336 			default:
5337 				if ((error = pf_commit_rules(ioe->ticket,
5338 				    ioe->rs_num, ioe->anchor))) {
5339 					PF_RULES_WUNLOCK();
5340 					free(ioes, M_TEMP);
5341 					goto fail; /* really bad */
5342 				}
5343 				break;
5344 			}
5345 		}
5346 		PF_RULES_WUNLOCK();
5347 
5348 		/* Only hook into EtherNet taffic if we've got rules for it. */
5349 		if (! TAILQ_EMPTY(V_pf_keth->active.rules))
5350 			hook_pf_eth();
5351 		else
5352 			dehook_pf_eth();
5353 
5354 		free(ioes, M_TEMP);
5355 		break;
5356 	}
5357 
5358 	case DIOCGETSRCNODES: {
5359 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
5360 		struct pf_srchash	*sh;
5361 		struct pf_ksrc_node	*n;
5362 		struct pf_src_node	*p, *pstore;
5363 		uint32_t		 i, nr = 0;
5364 
5365 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5366 				i++, sh++) {
5367 			PF_HASHROW_LOCK(sh);
5368 			LIST_FOREACH(n, &sh->nodes, entry)
5369 				nr++;
5370 			PF_HASHROW_UNLOCK(sh);
5371 		}
5372 
5373 		psn->psn_len = min(psn->psn_len,
5374 		    sizeof(struct pf_src_node) * nr);
5375 
5376 		if (psn->psn_len == 0) {
5377 			psn->psn_len = sizeof(struct pf_src_node) * nr;
5378 			break;
5379 		}
5380 
5381 		nr = 0;
5382 
5383 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
5384 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5385 		    i++, sh++) {
5386 		    PF_HASHROW_LOCK(sh);
5387 		    LIST_FOREACH(n, &sh->nodes, entry) {
5388 
5389 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
5390 				break;
5391 
5392 			pf_src_node_copy(n, p);
5393 
5394 			p++;
5395 			nr++;
5396 		    }
5397 		    PF_HASHROW_UNLOCK(sh);
5398 		}
5399 		error = copyout(pstore, psn->psn_src_nodes,
5400 		    sizeof(struct pf_src_node) * nr);
5401 		if (error) {
5402 			free(pstore, M_TEMP);
5403 			break;
5404 		}
5405 		psn->psn_len = sizeof(struct pf_src_node) * nr;
5406 		free(pstore, M_TEMP);
5407 		break;
5408 	}
5409 
5410 	case DIOCCLRSRCNODES: {
5411 		pf_clear_srcnodes(NULL);
5412 		pf_purge_expired_src_nodes();
5413 		break;
5414 	}
5415 
5416 	case DIOCKILLSRCNODES:
5417 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
5418 		break;
5419 
5420 #ifdef COMPAT_FREEBSD13
5421 	case DIOCKEEPCOUNTERS_FREEBSD13:
5422 #endif
5423 	case DIOCKEEPCOUNTERS:
5424 		error = pf_keepcounters((struct pfioc_nv *)addr);
5425 		break;
5426 
5427 	case DIOCGETSYNCOOKIES:
5428 		error = pf_get_syncookies((struct pfioc_nv *)addr);
5429 		break;
5430 
5431 	case DIOCSETSYNCOOKIES:
5432 		error = pf_set_syncookies((struct pfioc_nv *)addr);
5433 		break;
5434 
5435 	case DIOCSETHOSTID: {
5436 		u_int32_t	*hostid = (u_int32_t *)addr;
5437 
5438 		PF_RULES_WLOCK();
5439 		if (*hostid == 0)
5440 			V_pf_status.hostid = arc4random();
5441 		else
5442 			V_pf_status.hostid = *hostid;
5443 		PF_RULES_WUNLOCK();
5444 		break;
5445 	}
5446 
5447 	case DIOCOSFPFLUSH:
5448 		PF_RULES_WLOCK();
5449 		pf_osfp_flush();
5450 		PF_RULES_WUNLOCK();
5451 		break;
5452 
5453 	case DIOCIGETIFACES: {
5454 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5455 		struct pfi_kif *ifstore;
5456 		size_t bufsiz;
5457 
5458 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
5459 			error = ENODEV;
5460 			break;
5461 		}
5462 
5463 		if (io->pfiio_size < 0 ||
5464 		    io->pfiio_size > pf_ioctl_maxcount ||
5465 		    WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
5466 			error = EINVAL;
5467 			break;
5468 		}
5469 
5470 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5471 
5472 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
5473 		ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
5474 		    M_TEMP, M_WAITOK | M_ZERO);
5475 
5476 		PF_RULES_RLOCK();
5477 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
5478 		PF_RULES_RUNLOCK();
5479 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
5480 		free(ifstore, M_TEMP);
5481 		break;
5482 	}
5483 
5484 	case DIOCSETIFFLAG: {
5485 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5486 
5487 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5488 
5489 		PF_RULES_WLOCK();
5490 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
5491 		PF_RULES_WUNLOCK();
5492 		break;
5493 	}
5494 
5495 	case DIOCCLRIFFLAG: {
5496 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5497 
5498 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5499 
5500 		PF_RULES_WLOCK();
5501 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
5502 		PF_RULES_WUNLOCK();
5503 		break;
5504 	}
5505 
5506 	case DIOCSETREASS: {
5507 		u_int32_t	*reass = (u_int32_t *)addr;
5508 
5509 		V_pf_status.reass = *reass & (PF_REASS_ENABLED|PF_REASS_NODF);
5510 		/* Removal of DF flag without reassembly enabled is not a
5511 		 * valid combination. Disable reassembly in such case. */
5512 		if (!(V_pf_status.reass & PF_REASS_ENABLED))
5513 			V_pf_status.reass = 0;
5514 		break;
5515 	}
5516 
5517 	default:
5518 		error = ENODEV;
5519 		break;
5520 	}
5521 fail:
5522 	if (sx_xlocked(&V_pf_ioctl_lock))
5523 		sx_xunlock(&V_pf_ioctl_lock);
5524 	CURVNET_RESTORE();
5525 
5526 #undef ERROUT_IOCTL
5527 
5528 	return (error);
5529 }
5530 
5531 void
5532 pfsync_state_export(union pfsync_state_union *sp, struct pf_kstate *st, int msg_version)
5533 {
5534 	bzero(sp, sizeof(union pfsync_state_union));
5535 
5536 	/* copy from state key */
5537 	sp->pfs_1301.key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5538 	sp->pfs_1301.key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5539 	sp->pfs_1301.key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5540 	sp->pfs_1301.key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5541 	sp->pfs_1301.key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5542 	sp->pfs_1301.key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5543 	sp->pfs_1301.key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5544 	sp->pfs_1301.key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5545 	sp->pfs_1301.proto = st->key[PF_SK_WIRE]->proto;
5546 	sp->pfs_1301.af = st->key[PF_SK_WIRE]->af;
5547 
5548 	/* copy from state */
5549 	strlcpy(sp->pfs_1301.ifname, st->kif->pfik_name, sizeof(sp->pfs_1301.ifname));
5550 	bcopy(&st->rt_addr, &sp->pfs_1301.rt_addr, sizeof(sp->pfs_1301.rt_addr));
5551 	sp->pfs_1301.creation = htonl(time_uptime - st->creation);
5552 	sp->pfs_1301.expire = pf_state_expires(st);
5553 	if (sp->pfs_1301.expire <= time_uptime)
5554 		sp->pfs_1301.expire = htonl(0);
5555 	else
5556 		sp->pfs_1301.expire = htonl(sp->pfs_1301.expire - time_uptime);
5557 
5558 	sp->pfs_1301.direction = st->direction;
5559 	sp->pfs_1301.log = st->act.log;
5560 	sp->pfs_1301.timeout = st->timeout;
5561 
5562 	switch (msg_version) {
5563 		case PFSYNC_MSG_VERSION_1301:
5564 			sp->pfs_1301.state_flags = st->state_flags;
5565 			break;
5566 		case PFSYNC_MSG_VERSION_1400:
5567 			sp->pfs_1400.state_flags = htons(st->state_flags);
5568 			sp->pfs_1400.qid = htons(st->act.qid);
5569 			sp->pfs_1400.pqid = htons(st->act.pqid);
5570 			sp->pfs_1400.dnpipe = htons(st->act.dnpipe);
5571 			sp->pfs_1400.dnrpipe = htons(st->act.dnrpipe);
5572 			sp->pfs_1400.rtableid = htonl(st->act.rtableid);
5573 			sp->pfs_1400.min_ttl = st->act.min_ttl;
5574 			sp->pfs_1400.set_tos = st->act.set_tos;
5575 			sp->pfs_1400.max_mss = htons(st->act.max_mss);
5576 			sp->pfs_1400.set_prio[0] = st->act.set_prio[0];
5577 			sp->pfs_1400.set_prio[1] = st->act.set_prio[1];
5578 			sp->pfs_1400.rt = st->rt;
5579 			if (st->rt_kif)
5580 				strlcpy(sp->pfs_1400.rt_ifname,
5581 				    st->rt_kif->pfik_name,
5582 				    sizeof(sp->pfs_1400.rt_ifname));
5583 			break;
5584 		default:
5585 			panic("%s: Unsupported pfsync_msg_version %d",
5586 			    __func__, msg_version);
5587 	}
5588 
5589 	if (st->src_node)
5590 		sp->pfs_1301.sync_flags |= PFSYNC_FLAG_SRCNODE;
5591 	if (st->nat_src_node)
5592 		sp->pfs_1301.sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5593 
5594 	sp->pfs_1301.id = st->id;
5595 	sp->pfs_1301.creatorid = st->creatorid;
5596 	pf_state_peer_hton(&st->src, &sp->pfs_1301.src);
5597 	pf_state_peer_hton(&st->dst, &sp->pfs_1301.dst);
5598 
5599 	if (st->rule.ptr == NULL)
5600 		sp->pfs_1301.rule = htonl(-1);
5601 	else
5602 		sp->pfs_1301.rule = htonl(st->rule.ptr->nr);
5603 	if (st->anchor.ptr == NULL)
5604 		sp->pfs_1301.anchor = htonl(-1);
5605 	else
5606 		sp->pfs_1301.anchor = htonl(st->anchor.ptr->nr);
5607 	if (st->nat_rule.ptr == NULL)
5608 		sp->pfs_1301.nat_rule = htonl(-1);
5609 	else
5610 		sp->pfs_1301.nat_rule = htonl(st->nat_rule.ptr->nr);
5611 
5612 	pf_state_counter_hton(st->packets[0], sp->pfs_1301.packets[0]);
5613 	pf_state_counter_hton(st->packets[1], sp->pfs_1301.packets[1]);
5614 	pf_state_counter_hton(st->bytes[0], sp->pfs_1301.bytes[0]);
5615 	pf_state_counter_hton(st->bytes[1], sp->pfs_1301.bytes[1]);
5616 }
5617 
5618 void
5619 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st)
5620 {
5621 	bzero(sp, sizeof(*sp));
5622 
5623 	sp->version = PF_STATE_VERSION;
5624 
5625 	/* copy from state key */
5626 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5627 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5628 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5629 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5630 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5631 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5632 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5633 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5634 	sp->proto = st->key[PF_SK_WIRE]->proto;
5635 	sp->af = st->key[PF_SK_WIRE]->af;
5636 
5637 	/* copy from state */
5638 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
5639 	strlcpy(sp->orig_ifname, st->orig_kif->pfik_name,
5640 	    sizeof(sp->orig_ifname));
5641 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
5642 	sp->creation = htonl(time_uptime - st->creation);
5643 	sp->expire = pf_state_expires(st);
5644 	if (sp->expire <= time_uptime)
5645 		sp->expire = htonl(0);
5646 	else
5647 		sp->expire = htonl(sp->expire - time_uptime);
5648 
5649 	sp->direction = st->direction;
5650 	sp->log = st->act.log;
5651 	sp->timeout = st->timeout;
5652 	/* 8 bits for the old libpfctl, 16 bits for the new libpfctl */
5653 	sp->state_flags_compat = st->state_flags;
5654 	sp->state_flags = htons(st->state_flags);
5655 	if (st->src_node)
5656 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
5657 	if (st->nat_src_node)
5658 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5659 
5660 	sp->id = st->id;
5661 	sp->creatorid = st->creatorid;
5662 	pf_state_peer_hton(&st->src, &sp->src);
5663 	pf_state_peer_hton(&st->dst, &sp->dst);
5664 
5665 	if (st->rule.ptr == NULL)
5666 		sp->rule = htonl(-1);
5667 	else
5668 		sp->rule = htonl(st->rule.ptr->nr);
5669 	if (st->anchor.ptr == NULL)
5670 		sp->anchor = htonl(-1);
5671 	else
5672 		sp->anchor = htonl(st->anchor.ptr->nr);
5673 	if (st->nat_rule.ptr == NULL)
5674 		sp->nat_rule = htonl(-1);
5675 	else
5676 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
5677 
5678 	sp->packets[0] = st->packets[0];
5679 	sp->packets[1] = st->packets[1];
5680 	sp->bytes[0] = st->bytes[0];
5681 	sp->bytes[1] = st->bytes[1];
5682 
5683 	sp->qid = htons(st->act.qid);
5684 	sp->pqid = htons(st->act.pqid);
5685 	sp->dnpipe = htons(st->act.dnpipe);
5686 	sp->dnrpipe = htons(st->act.dnrpipe);
5687 	sp->rtableid = htonl(st->act.rtableid);
5688 	sp->min_ttl = st->act.min_ttl;
5689 	sp->set_tos = st->act.set_tos;
5690 	sp->max_mss = htons(st->act.max_mss);
5691 	sp->rt = st->rt;
5692 	if (st->rt_kif)
5693 		strlcpy(sp->rt_ifname, st->rt_kif->pfik_name,
5694 		    sizeof(sp->rt_ifname));
5695 	sp->set_prio[0] = st->act.set_prio[0];
5696 	sp->set_prio[1] = st->act.set_prio[1];
5697 
5698 }
5699 
5700 static void
5701 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
5702 {
5703 	struct pfr_ktable *kt;
5704 
5705 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
5706 
5707 	kt = aw->p.tbl;
5708 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
5709 		kt = kt->pfrkt_root;
5710 	aw->p.tbl = NULL;
5711 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
5712 		kt->pfrkt_cnt : -1;
5713 }
5714 
5715 static int
5716 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters,
5717     size_t number, char **names)
5718 {
5719 	nvlist_t        *nvc;
5720 
5721 	nvc = nvlist_create(0);
5722 	if (nvc == NULL)
5723 		return (ENOMEM);
5724 
5725 	for (int i = 0; i < number; i++) {
5726 		nvlist_append_number_array(nvc, "counters",
5727 		    counter_u64_fetch(counters[i]));
5728 		nvlist_append_string_array(nvc, "names",
5729 		    names[i]);
5730 		nvlist_append_number_array(nvc, "ids",
5731 		    i);
5732 	}
5733 	nvlist_add_nvlist(nvl, name, nvc);
5734 	nvlist_destroy(nvc);
5735 
5736 	return (0);
5737 }
5738 
5739 static int
5740 pf_getstatus(struct pfioc_nv *nv)
5741 {
5742 	nvlist_t        *nvl = NULL, *nvc = NULL;
5743 	void            *nvlpacked = NULL;
5744 	int              error;
5745 	struct pf_status s;
5746 	char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES;
5747 	char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES;
5748 	char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES;
5749 	PF_RULES_RLOCK_TRACKER;
5750 
5751 #define ERROUT(x)      ERROUT_FUNCTION(errout, x)
5752 
5753 	PF_RULES_RLOCK();
5754 
5755 	nvl = nvlist_create(0);
5756 	if (nvl == NULL)
5757 		ERROUT(ENOMEM);
5758 
5759 	nvlist_add_bool(nvl, "running", V_pf_status.running);
5760 	nvlist_add_number(nvl, "since", V_pf_status.since);
5761 	nvlist_add_number(nvl, "debug", V_pf_status.debug);
5762 	nvlist_add_number(nvl, "hostid", V_pf_status.hostid);
5763 	nvlist_add_number(nvl, "states", V_pf_status.states);
5764 	nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes);
5765 	nvlist_add_number(nvl, "reass", V_pf_status.reass);
5766 	nvlist_add_bool(nvl, "syncookies_active",
5767 	    V_pf_status.syncookies_active);
5768 
5769 	/* counters */
5770 	error = pf_add_status_counters(nvl, "counters", V_pf_status.counters,
5771 	    PFRES_MAX, pf_reasons);
5772 	if (error != 0)
5773 		ERROUT(error);
5774 
5775 	/* lcounters */
5776 	error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters,
5777 	    KLCNT_MAX, pf_lcounter);
5778 	if (error != 0)
5779 		ERROUT(error);
5780 
5781 	/* fcounters */
5782 	nvc = nvlist_create(0);
5783 	if (nvc == NULL)
5784 		ERROUT(ENOMEM);
5785 
5786 	for (int i = 0; i < FCNT_MAX; i++) {
5787 		nvlist_append_number_array(nvc, "counters",
5788 		    pf_counter_u64_fetch(&V_pf_status.fcounters[i]));
5789 		nvlist_append_string_array(nvc, "names",
5790 		    pf_fcounter[i]);
5791 		nvlist_append_number_array(nvc, "ids",
5792 		    i);
5793 	}
5794 	nvlist_add_nvlist(nvl, "fcounters", nvc);
5795 	nvlist_destroy(nvc);
5796 	nvc = NULL;
5797 
5798 	/* scounters */
5799 	error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters,
5800 	    SCNT_MAX, pf_fcounter);
5801 	if (error != 0)
5802 		ERROUT(error);
5803 
5804 	nvlist_add_string(nvl, "ifname", V_pf_status.ifname);
5805 	nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum,
5806 	    PF_MD5_DIGEST_LENGTH);
5807 
5808 	pfi_update_status(V_pf_status.ifname, &s);
5809 
5810 	/* pcounters / bcounters */
5811 	for (int i = 0; i < 2; i++) {
5812 		for (int j = 0; j < 2; j++) {
5813 			for (int k = 0; k < 2; k++) {
5814 				nvlist_append_number_array(nvl, "pcounters",
5815 				    s.pcounters[i][j][k]);
5816 			}
5817 			nvlist_append_number_array(nvl, "bcounters",
5818 			    s.bcounters[i][j]);
5819 		}
5820 	}
5821 
5822 	nvlpacked = nvlist_pack(nvl, &nv->len);
5823 	if (nvlpacked == NULL)
5824 		ERROUT(ENOMEM);
5825 
5826 	if (nv->size == 0)
5827 		ERROUT(0);
5828 	else if (nv->size < nv->len)
5829 		ERROUT(ENOSPC);
5830 
5831 	PF_RULES_RUNLOCK();
5832 	error = copyout(nvlpacked, nv->data, nv->len);
5833 	goto done;
5834 
5835 #undef ERROUT
5836 errout:
5837 	PF_RULES_RUNLOCK();
5838 done:
5839 	free(nvlpacked, M_NVLIST);
5840 	nvlist_destroy(nvc);
5841 	nvlist_destroy(nvl);
5842 
5843 	return (error);
5844 }
5845 
5846 /*
5847  * XXX - Check for version mismatch!!!
5848  */
5849 static void
5850 pf_clear_all_states(void)
5851 {
5852 	struct pf_kstate	*s;
5853 	u_int i;
5854 
5855 	for (i = 0; i <= pf_hashmask; i++) {
5856 		struct pf_idhash *ih = &V_pf_idhash[i];
5857 relock:
5858 		PF_HASHROW_LOCK(ih);
5859 		LIST_FOREACH(s, &ih->states, entry) {
5860 			s->timeout = PFTM_PURGE;
5861 			/* Don't send out individual delete messages. */
5862 			s->state_flags |= PFSTATE_NOSYNC;
5863 			pf_unlink_state(s);
5864 			goto relock;
5865 		}
5866 		PF_HASHROW_UNLOCK(ih);
5867 	}
5868 }
5869 
5870 static int
5871 pf_clear_tables(void)
5872 {
5873 	struct pfioc_table io;
5874 	int error;
5875 
5876 	bzero(&io, sizeof(io));
5877 
5878 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
5879 	    io.pfrio_flags);
5880 
5881 	return (error);
5882 }
5883 
5884 static void
5885 pf_clear_srcnodes(struct pf_ksrc_node *n)
5886 {
5887 	struct pf_kstate *s;
5888 	int i;
5889 
5890 	for (i = 0; i <= pf_hashmask; i++) {
5891 		struct pf_idhash *ih = &V_pf_idhash[i];
5892 
5893 		PF_HASHROW_LOCK(ih);
5894 		LIST_FOREACH(s, &ih->states, entry) {
5895 			if (n == NULL || n == s->src_node)
5896 				s->src_node = NULL;
5897 			if (n == NULL || n == s->nat_src_node)
5898 				s->nat_src_node = NULL;
5899 		}
5900 		PF_HASHROW_UNLOCK(ih);
5901 	}
5902 
5903 	if (n == NULL) {
5904 		struct pf_srchash *sh;
5905 
5906 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5907 		    i++, sh++) {
5908 			PF_HASHROW_LOCK(sh);
5909 			LIST_FOREACH(n, &sh->nodes, entry) {
5910 				n->expire = 1;
5911 				n->states = 0;
5912 			}
5913 			PF_HASHROW_UNLOCK(sh);
5914 		}
5915 	} else {
5916 		/* XXX: hash slot should already be locked here. */
5917 		n->expire = 1;
5918 		n->states = 0;
5919 	}
5920 }
5921 
5922 static void
5923 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
5924 {
5925 	struct pf_ksrc_node_list	 kill;
5926 
5927 	LIST_INIT(&kill);
5928 	for (int i = 0; i <= pf_srchashmask; i++) {
5929 		struct pf_srchash *sh = &V_pf_srchash[i];
5930 		struct pf_ksrc_node *sn, *tmp;
5931 
5932 		PF_HASHROW_LOCK(sh);
5933 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
5934 			if (PF_MATCHA(psnk->psnk_src.neg,
5935 			      &psnk->psnk_src.addr.v.a.addr,
5936 			      &psnk->psnk_src.addr.v.a.mask,
5937 			      &sn->addr, sn->af) &&
5938 			    PF_MATCHA(psnk->psnk_dst.neg,
5939 			      &psnk->psnk_dst.addr.v.a.addr,
5940 			      &psnk->psnk_dst.addr.v.a.mask,
5941 			      &sn->raddr, sn->af)) {
5942 				pf_unlink_src_node(sn);
5943 				LIST_INSERT_HEAD(&kill, sn, entry);
5944 				sn->expire = 1;
5945 			}
5946 		PF_HASHROW_UNLOCK(sh);
5947 	}
5948 
5949 	for (int i = 0; i <= pf_hashmask; i++) {
5950 		struct pf_idhash *ih = &V_pf_idhash[i];
5951 		struct pf_kstate *s;
5952 
5953 		PF_HASHROW_LOCK(ih);
5954 		LIST_FOREACH(s, &ih->states, entry) {
5955 			if (s->src_node && s->src_node->expire == 1)
5956 				s->src_node = NULL;
5957 			if (s->nat_src_node && s->nat_src_node->expire == 1)
5958 				s->nat_src_node = NULL;
5959 		}
5960 		PF_HASHROW_UNLOCK(ih);
5961 	}
5962 
5963 	psnk->psnk_killed = pf_free_src_nodes(&kill);
5964 }
5965 
5966 static int
5967 pf_keepcounters(struct pfioc_nv *nv)
5968 {
5969 	nvlist_t	*nvl = NULL;
5970 	void		*nvlpacked = NULL;
5971 	int		 error = 0;
5972 
5973 #define	ERROUT(x)	ERROUT_FUNCTION(on_error, x)
5974 
5975 	if (nv->len > pf_ioctl_maxcount)
5976 		ERROUT(ENOMEM);
5977 
5978 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
5979 	if (nvlpacked == NULL)
5980 		ERROUT(ENOMEM);
5981 
5982 	error = copyin(nv->data, nvlpacked, nv->len);
5983 	if (error)
5984 		ERROUT(error);
5985 
5986 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
5987 	if (nvl == NULL)
5988 		ERROUT(EBADMSG);
5989 
5990 	if (! nvlist_exists_bool(nvl, "keep_counters"))
5991 		ERROUT(EBADMSG);
5992 
5993 	V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters");
5994 
5995 on_error:
5996 	nvlist_destroy(nvl);
5997 	free(nvlpacked, M_NVLIST);
5998 	return (error);
5999 }
6000 
6001 static unsigned int
6002 pf_clear_states(const struct pf_kstate_kill *kill)
6003 {
6004 	struct pf_state_key_cmp	 match_key;
6005 	struct pf_kstate	*s;
6006 	struct pfi_kkif	*kif;
6007 	int		 idx;
6008 	unsigned int	 killed = 0, dir;
6009 
6010 	for (unsigned int i = 0; i <= pf_hashmask; i++) {
6011 		struct pf_idhash *ih = &V_pf_idhash[i];
6012 
6013 relock_DIOCCLRSTATES:
6014 		PF_HASHROW_LOCK(ih);
6015 		LIST_FOREACH(s, &ih->states, entry) {
6016 			/* For floating states look at the original kif. */
6017 			kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
6018 
6019 			if (kill->psk_ifname[0] &&
6020 			    strcmp(kill->psk_ifname,
6021 			    kif->pfik_name))
6022 				continue;
6023 
6024 			if (kill->psk_kill_match) {
6025 				bzero(&match_key, sizeof(match_key));
6026 
6027 				if (s->direction == PF_OUT) {
6028 					dir = PF_IN;
6029 					idx = PF_SK_STACK;
6030 				} else {
6031 					dir = PF_OUT;
6032 					idx = PF_SK_WIRE;
6033 				}
6034 
6035 				match_key.af = s->key[idx]->af;
6036 				match_key.proto = s->key[idx]->proto;
6037 				PF_ACPY(&match_key.addr[0],
6038 				    &s->key[idx]->addr[1], match_key.af);
6039 				match_key.port[0] = s->key[idx]->port[1];
6040 				PF_ACPY(&match_key.addr[1],
6041 				    &s->key[idx]->addr[0], match_key.af);
6042 				match_key.port[1] = s->key[idx]->port[0];
6043 			}
6044 
6045 			/*
6046 			 * Don't send out individual
6047 			 * delete messages.
6048 			 */
6049 			s->state_flags |= PFSTATE_NOSYNC;
6050 			pf_unlink_state(s);
6051 			killed++;
6052 
6053 			if (kill->psk_kill_match)
6054 				killed += pf_kill_matching_state(&match_key,
6055 				    dir);
6056 
6057 			goto relock_DIOCCLRSTATES;
6058 		}
6059 		PF_HASHROW_UNLOCK(ih);
6060 	}
6061 
6062 	if (V_pfsync_clear_states_ptr != NULL)
6063 		V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname);
6064 
6065 	return (killed);
6066 }
6067 
6068 static void
6069 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed)
6070 {
6071 	struct pf_kstate	*s;
6072 
6073 	if (kill->psk_pfcmp.id) {
6074 		if (kill->psk_pfcmp.creatorid == 0)
6075 			kill->psk_pfcmp.creatorid = V_pf_status.hostid;
6076 		if ((s = pf_find_state_byid(kill->psk_pfcmp.id,
6077 		    kill->psk_pfcmp.creatorid))) {
6078 			pf_unlink_state(s);
6079 			*killed = 1;
6080 		}
6081 		return;
6082 	}
6083 
6084 	for (unsigned int i = 0; i <= pf_hashmask; i++)
6085 		*killed += pf_killstates_row(kill, &V_pf_idhash[i]);
6086 
6087 	return;
6088 }
6089 
6090 static int
6091 pf_killstates_nv(struct pfioc_nv *nv)
6092 {
6093 	struct pf_kstate_kill	 kill;
6094 	nvlist_t		*nvl = NULL;
6095 	void			*nvlpacked = NULL;
6096 	int			 error = 0;
6097 	unsigned int		 killed = 0;
6098 
6099 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
6100 
6101 	if (nv->len > pf_ioctl_maxcount)
6102 		ERROUT(ENOMEM);
6103 
6104 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6105 	if (nvlpacked == NULL)
6106 		ERROUT(ENOMEM);
6107 
6108 	error = copyin(nv->data, nvlpacked, nv->len);
6109 	if (error)
6110 		ERROUT(error);
6111 
6112 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6113 	if (nvl == NULL)
6114 		ERROUT(EBADMSG);
6115 
6116 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6117 	if (error)
6118 		ERROUT(error);
6119 
6120 	pf_killstates(&kill, &killed);
6121 
6122 	free(nvlpacked, M_NVLIST);
6123 	nvlpacked = NULL;
6124 	nvlist_destroy(nvl);
6125 	nvl = nvlist_create(0);
6126 	if (nvl == NULL)
6127 		ERROUT(ENOMEM);
6128 
6129 	nvlist_add_number(nvl, "killed", killed);
6130 
6131 	nvlpacked = nvlist_pack(nvl, &nv->len);
6132 	if (nvlpacked == NULL)
6133 		ERROUT(ENOMEM);
6134 
6135 	if (nv->size == 0)
6136 		ERROUT(0);
6137 	else if (nv->size < nv->len)
6138 		ERROUT(ENOSPC);
6139 
6140 	error = copyout(nvlpacked, nv->data, nv->len);
6141 
6142 on_error:
6143 	nvlist_destroy(nvl);
6144 	free(nvlpacked, M_NVLIST);
6145 	return (error);
6146 }
6147 
6148 static int
6149 pf_clearstates_nv(struct pfioc_nv *nv)
6150 {
6151 	struct pf_kstate_kill	 kill;
6152 	nvlist_t		*nvl = NULL;
6153 	void			*nvlpacked = NULL;
6154 	int			 error = 0;
6155 	unsigned int		 killed;
6156 
6157 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
6158 
6159 	if (nv->len > pf_ioctl_maxcount)
6160 		ERROUT(ENOMEM);
6161 
6162 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6163 	if (nvlpacked == NULL)
6164 		ERROUT(ENOMEM);
6165 
6166 	error = copyin(nv->data, nvlpacked, nv->len);
6167 	if (error)
6168 		ERROUT(error);
6169 
6170 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6171 	if (nvl == NULL)
6172 		ERROUT(EBADMSG);
6173 
6174 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6175 	if (error)
6176 		ERROUT(error);
6177 
6178 	killed = pf_clear_states(&kill);
6179 
6180 	free(nvlpacked, M_NVLIST);
6181 	nvlpacked = NULL;
6182 	nvlist_destroy(nvl);
6183 	nvl = nvlist_create(0);
6184 	if (nvl == NULL)
6185 		ERROUT(ENOMEM);
6186 
6187 	nvlist_add_number(nvl, "killed", killed);
6188 
6189 	nvlpacked = nvlist_pack(nvl, &nv->len);
6190 	if (nvlpacked == NULL)
6191 		ERROUT(ENOMEM);
6192 
6193 	if (nv->size == 0)
6194 		ERROUT(0);
6195 	else if (nv->size < nv->len)
6196 		ERROUT(ENOSPC);
6197 
6198 	error = copyout(nvlpacked, nv->data, nv->len);
6199 
6200 #undef ERROUT
6201 on_error:
6202 	nvlist_destroy(nvl);
6203 	free(nvlpacked, M_NVLIST);
6204 	return (error);
6205 }
6206 
6207 static int
6208 pf_getstate(struct pfioc_nv *nv)
6209 {
6210 	nvlist_t		*nvl = NULL, *nvls;
6211 	void			*nvlpacked = NULL;
6212 	struct pf_kstate	*s = NULL;
6213 	int			 error = 0;
6214 	uint64_t		 id, creatorid;
6215 
6216 #define ERROUT(x)	ERROUT_FUNCTION(errout, x)
6217 
6218 	if (nv->len > pf_ioctl_maxcount)
6219 		ERROUT(ENOMEM);
6220 
6221 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6222 	if (nvlpacked == NULL)
6223 		ERROUT(ENOMEM);
6224 
6225 	error = copyin(nv->data, nvlpacked, nv->len);
6226 	if (error)
6227 		ERROUT(error);
6228 
6229 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6230 	if (nvl == NULL)
6231 		ERROUT(EBADMSG);
6232 
6233 	PFNV_CHK(pf_nvuint64(nvl, "id", &id));
6234 	PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid));
6235 
6236 	s = pf_find_state_byid(id, creatorid);
6237 	if (s == NULL)
6238 		ERROUT(ENOENT);
6239 
6240 	free(nvlpacked, M_NVLIST);
6241 	nvlpacked = NULL;
6242 	nvlist_destroy(nvl);
6243 	nvl = nvlist_create(0);
6244 	if (nvl == NULL)
6245 		ERROUT(ENOMEM);
6246 
6247 	nvls = pf_state_to_nvstate(s);
6248 	if (nvls == NULL)
6249 		ERROUT(ENOMEM);
6250 
6251 	nvlist_add_nvlist(nvl, "state", nvls);
6252 	nvlist_destroy(nvls);
6253 
6254 	nvlpacked = nvlist_pack(nvl, &nv->len);
6255 	if (nvlpacked == NULL)
6256 		ERROUT(ENOMEM);
6257 
6258 	if (nv->size == 0)
6259 		ERROUT(0);
6260 	else if (nv->size < nv->len)
6261 		ERROUT(ENOSPC);
6262 
6263 	error = copyout(nvlpacked, nv->data, nv->len);
6264 
6265 #undef ERROUT
6266 errout:
6267 	if (s != NULL)
6268 		PF_STATE_UNLOCK(s);
6269 	free(nvlpacked, M_NVLIST);
6270 	nvlist_destroy(nvl);
6271 	return (error);
6272 }
6273 
6274 /*
6275  * XXX - Check for version mismatch!!!
6276  */
6277 
6278 /*
6279  * Duplicate pfctl -Fa operation to get rid of as much as we can.
6280  */
6281 static int
6282 shutdown_pf(void)
6283 {
6284 	int error = 0;
6285 	u_int32_t t[5];
6286 	char nn = '\0';
6287 
6288 	do {
6289 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
6290 		    != 0) {
6291 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
6292 			break;
6293 		}
6294 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
6295 		    != 0) {
6296 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
6297 			break;		/* XXX: rollback? */
6298 		}
6299 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
6300 		    != 0) {
6301 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
6302 			break;		/* XXX: rollback? */
6303 		}
6304 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
6305 		    != 0) {
6306 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
6307 			break;		/* XXX: rollback? */
6308 		}
6309 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
6310 		    != 0) {
6311 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
6312 			break;		/* XXX: rollback? */
6313 		}
6314 
6315 		/* XXX: these should always succeed here */
6316 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
6317 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
6318 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
6319 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
6320 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
6321 
6322 		if ((error = pf_clear_tables()) != 0)
6323 			break;
6324 
6325 		if ((error = pf_begin_eth(&t[0], &nn)) != 0) {
6326 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth\n"));
6327 			break;
6328 		}
6329 		pf_commit_eth(t[0], &nn);
6330 
6331 #ifdef ALTQ
6332 		if ((error = pf_begin_altq(&t[0])) != 0) {
6333 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
6334 			break;
6335 		}
6336 		pf_commit_altq(t[0]);
6337 #endif
6338 
6339 		pf_clear_all_states();
6340 
6341 		pf_clear_srcnodes(NULL);
6342 
6343 		/* status does not use malloced mem so no need to cleanup */
6344 		/* fingerprints and interfaces have their own cleanup code */
6345 	} while(0);
6346 
6347 	return (error);
6348 }
6349 
6350 static pfil_return_t
6351 pf_check_return(int chk, struct mbuf **m)
6352 {
6353 
6354 	switch (chk) {
6355 	case PF_PASS:
6356 		if (*m == NULL)
6357 			return (PFIL_CONSUMED);
6358 		else
6359 			return (PFIL_PASS);
6360 		break;
6361 	default:
6362 		if (*m != NULL) {
6363 			m_freem(*m);
6364 			*m = NULL;
6365 		}
6366 		return (PFIL_DROPPED);
6367 	}
6368 }
6369 
6370 static pfil_return_t
6371 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6372     void *ruleset __unused, struct inpcb *inp)
6373 {
6374 	int chk;
6375 
6376 	chk = pf_test_eth(PF_IN, flags, ifp, m, inp);
6377 
6378 	return (pf_check_return(chk, m));
6379 }
6380 
6381 static pfil_return_t
6382 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6383     void *ruleset __unused, struct inpcb *inp)
6384 {
6385 	int chk;
6386 
6387 	chk = pf_test_eth(PF_OUT, flags, ifp, m, inp);
6388 
6389 	return (pf_check_return(chk, m));
6390 }
6391 
6392 #ifdef INET
6393 static pfil_return_t
6394 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6395     void *ruleset __unused, struct inpcb *inp)
6396 {
6397 	int chk;
6398 
6399 	chk = pf_test(PF_IN, flags, ifp, m, inp, NULL);
6400 
6401 	return (pf_check_return(chk, m));
6402 }
6403 
6404 static pfil_return_t
6405 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6406     void *ruleset __unused,  struct inpcb *inp)
6407 {
6408 	int chk;
6409 
6410 	chk = pf_test(PF_OUT, flags, ifp, m, inp, NULL);
6411 
6412 	return (pf_check_return(chk, m));
6413 }
6414 #endif
6415 
6416 #ifdef INET6
6417 static pfil_return_t
6418 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
6419     void *ruleset __unused,  struct inpcb *inp)
6420 {
6421 	int chk;
6422 
6423 	/*
6424 	 * In case of loopback traffic IPv6 uses the real interface in
6425 	 * order to support scoped addresses. In order to support stateful
6426 	 * filtering we have change this to lo0 as it is the case in IPv4.
6427 	 */
6428 	CURVNET_SET(ifp->if_vnet);
6429 	chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp,
6430 	    m, inp, NULL);
6431 	CURVNET_RESTORE();
6432 
6433 	return (pf_check_return(chk, m));
6434 }
6435 
6436 static pfil_return_t
6437 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
6438     void *ruleset __unused,  struct inpcb *inp)
6439 {
6440 	int chk;
6441 
6442 	CURVNET_SET(ifp->if_vnet);
6443 	chk = pf_test6(PF_OUT, flags, ifp, m, inp, NULL);
6444 	CURVNET_RESTORE();
6445 
6446 	return (pf_check_return(chk, m));
6447 }
6448 #endif /* INET6 */
6449 
6450 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook);
6451 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook);
6452 #define	V_pf_eth_in_hook	VNET(pf_eth_in_hook)
6453 #define	V_pf_eth_out_hook	VNET(pf_eth_out_hook)
6454 
6455 #ifdef INET
6456 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
6457 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
6458 #define	V_pf_ip4_in_hook	VNET(pf_ip4_in_hook)
6459 #define	V_pf_ip4_out_hook	VNET(pf_ip4_out_hook)
6460 #endif
6461 #ifdef INET6
6462 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
6463 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
6464 #define	V_pf_ip6_in_hook	VNET(pf_ip6_in_hook)
6465 #define	V_pf_ip6_out_hook	VNET(pf_ip6_out_hook)
6466 #endif
6467 
6468 static void
6469 hook_pf_eth(void)
6470 {
6471 	struct pfil_hook_args pha = {
6472 		.pa_version = PFIL_VERSION,
6473 		.pa_modname = "pf",
6474 		.pa_type = PFIL_TYPE_ETHERNET,
6475 	};
6476 	struct pfil_link_args pla = {
6477 		.pa_version = PFIL_VERSION,
6478 	};
6479 	int ret __diagused;
6480 
6481 	if (atomic_load_bool(&V_pf_pfil_eth_hooked))
6482 		return;
6483 
6484 	pha.pa_mbuf_chk = pf_eth_check_in;
6485 	pha.pa_flags = PFIL_IN;
6486 	pha.pa_rulname = "eth-in";
6487 	V_pf_eth_in_hook = pfil_add_hook(&pha);
6488 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6489 	pla.pa_head = V_link_pfil_head;
6490 	pla.pa_hook = V_pf_eth_in_hook;
6491 	ret = pfil_link(&pla);
6492 	MPASS(ret == 0);
6493 	pha.pa_mbuf_chk = pf_eth_check_out;
6494 	pha.pa_flags = PFIL_OUT;
6495 	pha.pa_rulname = "eth-out";
6496 	V_pf_eth_out_hook = pfil_add_hook(&pha);
6497 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6498 	pla.pa_head = V_link_pfil_head;
6499 	pla.pa_hook = V_pf_eth_out_hook;
6500 	ret = pfil_link(&pla);
6501 	MPASS(ret == 0);
6502 
6503 	atomic_store_bool(&V_pf_pfil_eth_hooked, true);
6504 }
6505 
6506 static void
6507 hook_pf(void)
6508 {
6509 	struct pfil_hook_args pha = {
6510 		.pa_version = PFIL_VERSION,
6511 		.pa_modname = "pf",
6512 	};
6513 	struct pfil_link_args pla = {
6514 		.pa_version = PFIL_VERSION,
6515 	};
6516 	int ret __diagused;
6517 
6518 	if (atomic_load_bool(&V_pf_pfil_hooked))
6519 		return;
6520 
6521 #ifdef INET
6522 	pha.pa_type = PFIL_TYPE_IP4;
6523 	pha.pa_mbuf_chk = pf_check_in;
6524 	pha.pa_flags = PFIL_IN;
6525 	pha.pa_rulname = "default-in";
6526 	V_pf_ip4_in_hook = pfil_add_hook(&pha);
6527 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6528 	pla.pa_head = V_inet_pfil_head;
6529 	pla.pa_hook = V_pf_ip4_in_hook;
6530 	ret = pfil_link(&pla);
6531 	MPASS(ret == 0);
6532 	pha.pa_mbuf_chk = pf_check_out;
6533 	pha.pa_flags = PFIL_OUT;
6534 	pha.pa_rulname = "default-out";
6535 	V_pf_ip4_out_hook = pfil_add_hook(&pha);
6536 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6537 	pla.pa_head = V_inet_pfil_head;
6538 	pla.pa_hook = V_pf_ip4_out_hook;
6539 	ret = pfil_link(&pla);
6540 	MPASS(ret == 0);
6541 	if (V_pf_filter_local) {
6542 		pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6543 		pla.pa_head = V_inet_local_pfil_head;
6544 		pla.pa_hook = V_pf_ip4_out_hook;
6545 		ret = pfil_link(&pla);
6546 		MPASS(ret == 0);
6547 	}
6548 #endif
6549 #ifdef INET6
6550 	pha.pa_type = PFIL_TYPE_IP6;
6551 	pha.pa_mbuf_chk = pf_check6_in;
6552 	pha.pa_flags = PFIL_IN;
6553 	pha.pa_rulname = "default-in6";
6554 	V_pf_ip6_in_hook = pfil_add_hook(&pha);
6555 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6556 	pla.pa_head = V_inet6_pfil_head;
6557 	pla.pa_hook = V_pf_ip6_in_hook;
6558 	ret = pfil_link(&pla);
6559 	MPASS(ret == 0);
6560 	pha.pa_mbuf_chk = pf_check6_out;
6561 	pha.pa_rulname = "default-out6";
6562 	pha.pa_flags = PFIL_OUT;
6563 	V_pf_ip6_out_hook = pfil_add_hook(&pha);
6564 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6565 	pla.pa_head = V_inet6_pfil_head;
6566 	pla.pa_hook = V_pf_ip6_out_hook;
6567 	ret = pfil_link(&pla);
6568 	MPASS(ret == 0);
6569 	if (V_pf_filter_local) {
6570 		pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6571 		pla.pa_head = V_inet6_local_pfil_head;
6572 		pla.pa_hook = V_pf_ip6_out_hook;
6573 		ret = pfil_link(&pla);
6574 		MPASS(ret == 0);
6575 	}
6576 #endif
6577 
6578 	atomic_store_bool(&V_pf_pfil_hooked, true);
6579 }
6580 
6581 static void
6582 dehook_pf_eth(void)
6583 {
6584 
6585 	if (!atomic_load_bool(&V_pf_pfil_eth_hooked))
6586 		return;
6587 
6588 	pfil_remove_hook(V_pf_eth_in_hook);
6589 	pfil_remove_hook(V_pf_eth_out_hook);
6590 
6591 	atomic_store_bool(&V_pf_pfil_eth_hooked, false);
6592 }
6593 
6594 static void
6595 dehook_pf(void)
6596 {
6597 
6598 	if (!atomic_load_bool(&V_pf_pfil_hooked))
6599 		return;
6600 
6601 #ifdef INET
6602 	pfil_remove_hook(V_pf_ip4_in_hook);
6603 	pfil_remove_hook(V_pf_ip4_out_hook);
6604 #endif
6605 #ifdef INET6
6606 	pfil_remove_hook(V_pf_ip6_in_hook);
6607 	pfil_remove_hook(V_pf_ip6_out_hook);
6608 #endif
6609 
6610 	atomic_store_bool(&V_pf_pfil_hooked, false);
6611 }
6612 
6613 static void
6614 pf_load_vnet(void)
6615 {
6616 	V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
6617 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
6618 
6619 	rm_init_flags(&V_pf_rules_lock, "pf rulesets", RM_RECURSE);
6620 	sx_init(&V_pf_ioctl_lock, "pf ioctl");
6621 
6622 	pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
6623 	    PF_RULE_TAG_HASH_SIZE_DEFAULT);
6624 #ifdef ALTQ
6625 	pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
6626 	    PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
6627 #endif
6628 
6629 	V_pf_keth = &V_pf_main_keth_anchor.ruleset;
6630 
6631 	pfattach_vnet();
6632 	V_pf_vnet_active = 1;
6633 }
6634 
6635 static int
6636 pf_load(void)
6637 {
6638 	int error;
6639 
6640 	sx_init(&pf_end_lock, "pf end thread");
6641 
6642 	pf_mtag_initialize();
6643 
6644 	pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
6645 	if (pf_dev == NULL)
6646 		return (ENOMEM);
6647 
6648 	pf_end_threads = 0;
6649 	error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
6650 	if (error != 0)
6651 		return (error);
6652 
6653 	pfi_initialize();
6654 
6655 	return (0);
6656 }
6657 
6658 static void
6659 pf_unload_vnet(void)
6660 {
6661 	int ret __diagused;
6662 
6663 	V_pf_vnet_active = 0;
6664 	V_pf_status.running = 0;
6665 	dehook_pf();
6666 	dehook_pf_eth();
6667 
6668 	PF_RULES_WLOCK();
6669 	pf_syncookies_cleanup();
6670 	shutdown_pf();
6671 	PF_RULES_WUNLOCK();
6672 
6673 	/* Make sure we've cleaned up ethernet rules before we continue. */
6674 	NET_EPOCH_DRAIN_CALLBACKS();
6675 
6676 	ret = swi_remove(V_pf_swi_cookie);
6677 	MPASS(ret == 0);
6678 	ret = intr_event_destroy(V_pf_swi_ie);
6679 	MPASS(ret == 0);
6680 
6681 	pf_unload_vnet_purge();
6682 
6683 	pf_normalize_cleanup();
6684 	PF_RULES_WLOCK();
6685 	pfi_cleanup_vnet();
6686 	PF_RULES_WUNLOCK();
6687 	pfr_cleanup();
6688 	pf_osfp_flush();
6689 	pf_cleanup();
6690 	if (IS_DEFAULT_VNET(curvnet))
6691 		pf_mtag_cleanup();
6692 
6693 	pf_cleanup_tagset(&V_pf_tags);
6694 #ifdef ALTQ
6695 	pf_cleanup_tagset(&V_pf_qids);
6696 #endif
6697 	uma_zdestroy(V_pf_tag_z);
6698 
6699 #ifdef PF_WANT_32_TO_64_COUNTER
6700 	PF_RULES_WLOCK();
6701 	LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist);
6702 
6703 	MPASS(LIST_EMPTY(&V_pf_allkiflist));
6704 	MPASS(V_pf_allkifcount == 0);
6705 
6706 	LIST_REMOVE(&V_pf_default_rule, allrulelist);
6707 	V_pf_allrulecount--;
6708 	LIST_REMOVE(V_pf_rulemarker, allrulelist);
6709 
6710 	/*
6711 	 * There are known pf rule leaks when running the test suite.
6712 	 */
6713 #ifdef notyet
6714 	MPASS(LIST_EMPTY(&V_pf_allrulelist));
6715 	MPASS(V_pf_allrulecount == 0);
6716 #endif
6717 
6718 	PF_RULES_WUNLOCK();
6719 
6720 	free(V_pf_kifmarker, PFI_MTYPE);
6721 	free(V_pf_rulemarker, M_PFRULE);
6722 #endif
6723 
6724 	/* Free counters last as we updated them during shutdown. */
6725 	pf_counter_u64_deinit(&V_pf_default_rule.evaluations);
6726 	for (int i = 0; i < 2; i++) {
6727 		pf_counter_u64_deinit(&V_pf_default_rule.packets[i]);
6728 		pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]);
6729 	}
6730 	counter_u64_free(V_pf_default_rule.states_cur);
6731 	counter_u64_free(V_pf_default_rule.states_tot);
6732 	counter_u64_free(V_pf_default_rule.src_nodes);
6733 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, V_pf_default_rule.timestamp);
6734 
6735 	for (int i = 0; i < PFRES_MAX; i++)
6736 		counter_u64_free(V_pf_status.counters[i]);
6737 	for (int i = 0; i < KLCNT_MAX; i++)
6738 		counter_u64_free(V_pf_status.lcounters[i]);
6739 	for (int i = 0; i < FCNT_MAX; i++)
6740 		pf_counter_u64_deinit(&V_pf_status.fcounters[i]);
6741 	for (int i = 0; i < SCNT_MAX; i++)
6742 		counter_u64_free(V_pf_status.scounters[i]);
6743 
6744 	rm_destroy(&V_pf_rules_lock);
6745 	sx_destroy(&V_pf_ioctl_lock);
6746 }
6747 
6748 static void
6749 pf_unload(void)
6750 {
6751 
6752 	sx_xlock(&pf_end_lock);
6753 	pf_end_threads = 1;
6754 	while (pf_end_threads < 2) {
6755 		wakeup_one(pf_purge_thread);
6756 		sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
6757 	}
6758 	sx_xunlock(&pf_end_lock);
6759 
6760 	if (pf_dev != NULL)
6761 		destroy_dev(pf_dev);
6762 
6763 	pfi_cleanup();
6764 
6765 	sx_destroy(&pf_end_lock);
6766 }
6767 
6768 static void
6769 vnet_pf_init(void *unused __unused)
6770 {
6771 
6772 	pf_load_vnet();
6773 }
6774 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6775     vnet_pf_init, NULL);
6776 
6777 static void
6778 vnet_pf_uninit(const void *unused __unused)
6779 {
6780 
6781 	pf_unload_vnet();
6782 }
6783 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
6784 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6785     vnet_pf_uninit, NULL);
6786 
6787 static int
6788 pf_modevent(module_t mod, int type, void *data)
6789 {
6790 	int error = 0;
6791 
6792 	switch(type) {
6793 	case MOD_LOAD:
6794 		error = pf_load();
6795 		break;
6796 	case MOD_UNLOAD:
6797 		/* Handled in SYSUNINIT(pf_unload) to ensure it's done after
6798 		 * the vnet_pf_uninit()s */
6799 		break;
6800 	default:
6801 		error = EINVAL;
6802 		break;
6803 	}
6804 
6805 	return (error);
6806 }
6807 
6808 static moduledata_t pf_mod = {
6809 	"pf",
6810 	pf_modevent,
6811 	0
6812 };
6813 
6814 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
6815 MODULE_VERSION(pf, PF_MODVER);
6816