xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision 0d7056458db5b5dd7fdc5ccd8abab73e3ee76a20)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2001 Daniel Hartmeier
5  * Copyright (c) 2002,2003 Henning Brauer
6  * Copyright (c) 2012 Gleb Smirnoff <glebius@FreeBSD.org>
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  *
13  *    - Redistributions of source code must retain the above copyright
14  *      notice, this list of conditions and the following disclaimer.
15  *    - Redistributions in binary form must reproduce the above
16  *      copyright notice, this list of conditions and the following
17  *      disclaimer in the documentation and/or other materials provided
18  *      with the distribution.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
26  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
28  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
30  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
31  * POSSIBILITY OF SUCH DAMAGE.
32  *
33  * Effort sponsored in part by the Defense Advanced Research Projects
34  * Agency (DARPA) and Air Force Research Laboratory, Air Force
35  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
36  *
37  *	$OpenBSD: pf_ioctl.c,v 1.213 2009/02/15 21:46:12 mbalmer Exp $
38  */
39 
40 #include <sys/cdefs.h>
41 #include "opt_inet.h"
42 #include "opt_inet6.h"
43 #include "opt_bpf.h"
44 #include "opt_pf.h"
45 
46 #include <sys/param.h>
47 #include <sys/_bitset.h>
48 #include <sys/bitset.h>
49 #include <sys/bus.h>
50 #include <sys/conf.h>
51 #include <sys/endian.h>
52 #include <sys/fcntl.h>
53 #include <sys/filio.h>
54 #include <sys/hash.h>
55 #include <sys/interrupt.h>
56 #include <sys/jail.h>
57 #include <sys/kernel.h>
58 #include <sys/kthread.h>
59 #include <sys/lock.h>
60 #include <sys/mbuf.h>
61 #include <sys/module.h>
62 #include <sys/nv.h>
63 #include <sys/proc.h>
64 #include <sys/sdt.h>
65 #include <sys/smp.h>
66 #include <sys/socket.h>
67 #include <sys/sysctl.h>
68 #include <sys/md5.h>
69 #include <sys/ucred.h>
70 
71 #include <net/if.h>
72 #include <net/if_var.h>
73 #include <net/if_private.h>
74 #include <net/vnet.h>
75 #include <net/route.h>
76 #include <net/pfil.h>
77 #include <net/pfvar.h>
78 #include <net/if_pfsync.h>
79 #include <net/if_pflog.h>
80 
81 #include <netinet/in.h>
82 #include <netinet/ip.h>
83 #include <netinet/ip_var.h>
84 #include <netinet6/ip6_var.h>
85 #include <netinet/ip_icmp.h>
86 #include <netpfil/pf/pf_nl.h>
87 #include <netpfil/pf/pf_nv.h>
88 
89 #ifdef INET6
90 #include <netinet/ip6.h>
91 #endif /* INET6 */
92 
93 #ifdef ALTQ
94 #include <net/altq/altq.h>
95 #endif
96 
97 SDT_PROBE_DEFINE3(pf, ioctl, ioctl, error, "int", "int", "int");
98 SDT_PROBE_DEFINE3(pf, ioctl, function, error, "char *", "int", "int");
99 SDT_PROBE_DEFINE2(pf, ioctl, addrule, error, "int", "int");
100 SDT_PROBE_DEFINE2(pf, ioctl, nvchk, error, "int", "int");
101 
102 static struct pf_kpool	*pf_get_kpool(const char *, u_int32_t, u_int8_t,
103 			    u_int32_t, u_int8_t, u_int8_t, u_int8_t);
104 
105 static void		 pf_mv_kpool(struct pf_kpalist *, struct pf_kpalist *);
106 static void		 pf_empty_kpool(struct pf_kpalist *);
107 static int		 pfioctl(struct cdev *, u_long, caddr_t, int,
108 			    struct thread *);
109 static int		 pf_begin_eth(uint32_t *, const char *);
110 static void		 pf_rollback_eth_cb(struct epoch_context *);
111 static int		 pf_rollback_eth(uint32_t, const char *);
112 static int		 pf_commit_eth(uint32_t, const char *);
113 static void		 pf_free_eth_rule(struct pf_keth_rule *);
114 #ifdef ALTQ
115 static int		 pf_begin_altq(u_int32_t *);
116 static int		 pf_rollback_altq(u_int32_t);
117 static int		 pf_commit_altq(u_int32_t);
118 static int		 pf_enable_altq(struct pf_altq *);
119 static int		 pf_disable_altq(struct pf_altq *);
120 static uint16_t		 pf_qname2qid(const char *);
121 static void		 pf_qid_unref(uint16_t);
122 #endif /* ALTQ */
123 static int		 pf_begin_rules(u_int32_t *, int, const char *);
124 static int		 pf_rollback_rules(u_int32_t, int, char *);
125 static int		 pf_setup_pfsync_matching(struct pf_kruleset *);
126 static void		 pf_hash_rule_rolling(MD5_CTX *, struct pf_krule *);
127 static void		 pf_hash_rule(struct pf_krule *);
128 static void		 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
129 static int		 pf_commit_rules(u_int32_t, int, char *);
130 static int		 pf_addr_setup(struct pf_kruleset *,
131 			    struct pf_addr_wrap *, sa_family_t);
132 static void		 pf_addr_copyout(struct pf_addr_wrap *);
133 static void		 pf_src_node_copy(const struct pf_ksrc_node *,
134 			    struct pf_src_node *);
135 #ifdef ALTQ
136 static int		 pf_export_kaltq(struct pf_altq *,
137 			    struct pfioc_altq_v1 *, size_t);
138 static int		 pf_import_kaltq(struct pfioc_altq_v1 *,
139 			    struct pf_altq *, size_t);
140 #endif /* ALTQ */
141 
142 VNET_DEFINE(struct pf_krule,	pf_default_rule);
143 
144 static __inline int             pf_krule_compare(struct pf_krule *,
145 				    struct pf_krule *);
146 
147 RB_GENERATE(pf_krule_global, pf_krule, entry_global, pf_krule_compare);
148 
149 #ifdef ALTQ
150 VNET_DEFINE_STATIC(int,		pf_altq_running);
151 #define	V_pf_altq_running	VNET(pf_altq_running)
152 #endif
153 
154 #define	TAGID_MAX	 50000
155 struct pf_tagname {
156 	TAILQ_ENTRY(pf_tagname)	namehash_entries;
157 	TAILQ_ENTRY(pf_tagname)	taghash_entries;
158 	char			name[PF_TAG_NAME_SIZE];
159 	uint16_t		tag;
160 	int			ref;
161 };
162 
163 struct pf_tagset {
164 	TAILQ_HEAD(, pf_tagname)	*namehash;
165 	TAILQ_HEAD(, pf_tagname)	*taghash;
166 	unsigned int			 mask;
167 	uint32_t			 seed;
168 	BITSET_DEFINE(, TAGID_MAX)	 avail;
169 };
170 
171 VNET_DEFINE(struct pf_tagset, pf_tags);
172 #define	V_pf_tags	VNET(pf_tags)
173 static unsigned int	pf_rule_tag_hashsize;
174 #define	PF_RULE_TAG_HASH_SIZE_DEFAULT	128
175 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN,
176     &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT,
177     "Size of pf(4) rule tag hashtable");
178 
179 #ifdef ALTQ
180 VNET_DEFINE(struct pf_tagset, pf_qids);
181 #define	V_pf_qids	VNET(pf_qids)
182 static unsigned int	pf_queue_tag_hashsize;
183 #define	PF_QUEUE_TAG_HASH_SIZE_DEFAULT	128
184 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN,
185     &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT,
186     "Size of pf(4) queue tag hashtable");
187 #endif
188 VNET_DEFINE(uma_zone_t,	 pf_tag_z);
189 #define	V_pf_tag_z		 VNET(pf_tag_z)
190 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
191 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
192 
193 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
194 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
195 #endif
196 
197 VNET_DEFINE_STATIC(bool, pf_filter_local) = false;
198 #define V_pf_filter_local	VNET(pf_filter_local)
199 SYSCTL_BOOL(_net_pf, OID_AUTO, filter_local, CTLFLAG_VNET | CTLFLAG_RW,
200     &VNET_NAME(pf_filter_local), false,
201     "Enable filtering for packets delivered to local network stack");
202 
203 #ifdef PF_DEFAULT_TO_DROP
204 VNET_DEFINE_STATIC(bool, default_to_drop) = true;
205 #else
206 VNET_DEFINE_STATIC(bool, default_to_drop);
207 #endif
208 #define	V_default_to_drop VNET(default_to_drop)
209 SYSCTL_BOOL(_net_pf, OID_AUTO, default_to_drop, CTLFLAG_RDTUN | CTLFLAG_VNET,
210     &VNET_NAME(default_to_drop), false,
211     "Make the default rule drop all packets.");
212 
213 static void		 pf_init_tagset(struct pf_tagset *, unsigned int *,
214 			    unsigned int);
215 static void		 pf_cleanup_tagset(struct pf_tagset *);
216 static uint16_t		 tagname2hashindex(const struct pf_tagset *, const char *);
217 static uint16_t		 tag2hashindex(const struct pf_tagset *, uint16_t);
218 static u_int16_t	 tagname2tag(struct pf_tagset *, const char *);
219 static u_int16_t	 pf_tagname2tag(const char *);
220 static void		 tag_unref(struct pf_tagset *, u_int16_t);
221 
222 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
223 
224 struct cdev *pf_dev;
225 
226 /*
227  * XXX - These are new and need to be checked when moveing to a new version
228  */
229 static void		 pf_clear_all_states(void);
230 static int		 pf_killstates_row(struct pf_kstate_kill *,
231 			    struct pf_idhash *);
232 static int		 pf_killstates_nv(struct pfioc_nv *);
233 static int		 pf_clearstates_nv(struct pfioc_nv *);
234 static int		 pf_getstate(struct pfioc_nv *);
235 static int		 pf_getstatus(struct pfioc_nv *);
236 static int		 pf_clear_tables(void);
237 static void		 pf_clear_srcnodes(struct pf_ksrc_node *);
238 static void		 pf_kill_srcnodes(struct pfioc_src_node_kill *);
239 static int		 pf_keepcounters(struct pfioc_nv *);
240 static void		 pf_tbladdr_copyout(struct pf_addr_wrap *);
241 
242 /*
243  * Wrapper functions for pfil(9) hooks
244  */
245 static pfil_return_t pf_eth_check_in(struct mbuf **m, struct ifnet *ifp,
246     int flags, void *ruleset __unused, struct inpcb *inp);
247 static pfil_return_t pf_eth_check_out(struct mbuf **m, struct ifnet *ifp,
248     int flags, void *ruleset __unused, struct inpcb *inp);
249 #ifdef INET
250 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp,
251     int flags, void *ruleset __unused, struct inpcb *inp);
252 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp,
253     int flags, void *ruleset __unused, struct inpcb *inp);
254 #endif
255 #ifdef INET6
256 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp,
257     int flags, void *ruleset __unused, struct inpcb *inp);
258 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp,
259     int flags, void *ruleset __unused, struct inpcb *inp);
260 #endif
261 
262 static void		hook_pf_eth(void);
263 static void		hook_pf(void);
264 static void		dehook_pf_eth(void);
265 static void		dehook_pf(void);
266 static int		shutdown_pf(void);
267 static int		pf_load(void);
268 static void		pf_unload(void);
269 
270 static struct cdevsw pf_cdevsw = {
271 	.d_ioctl =	pfioctl,
272 	.d_name =	PF_NAME,
273 	.d_version =	D_VERSION,
274 };
275 
276 VNET_DEFINE_STATIC(bool, pf_pfil_hooked);
277 #define V_pf_pfil_hooked	VNET(pf_pfil_hooked)
278 VNET_DEFINE_STATIC(bool, pf_pfil_eth_hooked);
279 #define V_pf_pfil_eth_hooked	VNET(pf_pfil_eth_hooked)
280 
281 /*
282  * We need a flag that is neither hooked nor running to know when
283  * the VNET is "valid".  We primarily need this to control (global)
284  * external event, e.g., eventhandlers.
285  */
286 VNET_DEFINE(int, pf_vnet_active);
287 #define V_pf_vnet_active	VNET(pf_vnet_active)
288 
289 int pf_end_threads;
290 struct proc *pf_purge_proc;
291 
292 VNET_DEFINE(struct rmlock, pf_rules_lock);
293 VNET_DEFINE_STATIC(struct sx, pf_ioctl_lock);
294 #define	V_pf_ioctl_lock		VNET(pf_ioctl_lock)
295 struct sx			pf_end_lock;
296 
297 /* pfsync */
298 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
299 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
300 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
301 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
302 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
303 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
304 VNET_DEFINE(pflow_export_state_t *, pflow_export_state_ptr);
305 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
306 
307 /* pflog */
308 pflog_packet_t			*pflog_packet_ptr = NULL;
309 
310 /*
311  * Copy a user-provided string, returning an error if truncation would occur.
312  * Avoid scanning past "sz" bytes in the source string since there's no
313  * guarantee that it's nul-terminated.
314  */
315 static int
316 pf_user_strcpy(char *dst, const char *src, size_t sz)
317 {
318 	if (strnlen(src, sz) == sz)
319 		return (EINVAL);
320 	(void)strlcpy(dst, src, sz);
321 	return (0);
322 }
323 
324 static void
325 pfattach_vnet(void)
326 {
327 	u_int32_t *my_timeout = V_pf_default_rule.timeout;
328 
329 	bzero(&V_pf_status, sizeof(V_pf_status));
330 
331 	pf_initialize();
332 	pfr_initialize();
333 	pfi_initialize_vnet();
334 	pf_normalize_init();
335 	pf_syncookies_init();
336 
337 	V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
338 	V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
339 
340 	RB_INIT(&V_pf_anchors);
341 	pf_init_kruleset(&pf_main_ruleset);
342 
343 	pf_init_keth(V_pf_keth);
344 
345 	/* default rule should never be garbage collected */
346 	V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
347 	V_pf_default_rule.action = V_default_to_drop ? PF_DROP : PF_PASS;
348 	V_pf_default_rule.nr = -1;
349 	V_pf_default_rule.rtableid = -1;
350 
351 	pf_counter_u64_init(&V_pf_default_rule.evaluations, M_WAITOK);
352 	for (int i = 0; i < 2; i++) {
353 		pf_counter_u64_init(&V_pf_default_rule.packets[i], M_WAITOK);
354 		pf_counter_u64_init(&V_pf_default_rule.bytes[i], M_WAITOK);
355 	}
356 	V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
357 	V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
358 	V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
359 
360 	V_pf_default_rule.timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
361 	    M_WAITOK | M_ZERO);
362 
363 #ifdef PF_WANT_32_TO_64_COUNTER
364 	V_pf_kifmarker = malloc(sizeof(*V_pf_kifmarker), PFI_MTYPE, M_WAITOK | M_ZERO);
365 	V_pf_rulemarker = malloc(sizeof(*V_pf_rulemarker), M_PFRULE, M_WAITOK | M_ZERO);
366 	PF_RULES_WLOCK();
367 	LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist);
368 	LIST_INSERT_HEAD(&V_pf_allrulelist, &V_pf_default_rule, allrulelist);
369 	V_pf_allrulecount++;
370 	LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist);
371 	PF_RULES_WUNLOCK();
372 #endif
373 
374 	/* initialize default timeouts */
375 	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
376 	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
377 	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
378 	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
379 	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
380 	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
381 	my_timeout[PFTM_SCTP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
382 	my_timeout[PFTM_SCTP_OPENING] = PFTM_TCP_OPENING_VAL;
383 	my_timeout[PFTM_SCTP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
384 	my_timeout[PFTM_SCTP_CLOSING] = PFTM_TCP_CLOSING_VAL;
385 	my_timeout[PFTM_SCTP_CLOSED] = PFTM_TCP_CLOSED_VAL;
386 	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
387 	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
388 	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
389 	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
390 	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
391 	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
392 	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
393 	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
394 	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
395 	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
396 	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
397 	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
398 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
399 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
400 
401 	V_pf_status.debug = PF_DEBUG_URGENT;
402 	/*
403 	 * XXX This is different than in OpenBSD where reassembly is enabled by
404 	 * defult. In FreeBSD we expect people to still use scrub rules and
405 	 * switch to the new syntax later. Only when they switch they must
406 	 * explicitly enable reassemle. We could change the default once the
407 	 * scrub rule functionality is hopefully removed some day in future.
408 	 */
409 	V_pf_status.reass = 0;
410 
411 	V_pf_pfil_hooked = false;
412 	V_pf_pfil_eth_hooked = false;
413 
414 	/* XXX do our best to avoid a conflict */
415 	V_pf_status.hostid = arc4random();
416 
417 	for (int i = 0; i < PFRES_MAX; i++)
418 		V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
419 	for (int i = 0; i < KLCNT_MAX; i++)
420 		V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
421 	for (int i = 0; i < FCNT_MAX; i++)
422 		pf_counter_u64_init(&V_pf_status.fcounters[i], M_WAITOK);
423 	for (int i = 0; i < SCNT_MAX; i++)
424 		V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
425 
426 	if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET,
427 	    INTR_MPSAFE, &V_pf_swi_cookie) != 0)
428 		/* XXXGL: leaked all above. */
429 		return;
430 }
431 
432 static struct pf_kpool *
433 pf_get_kpool(const char *anchor, u_int32_t ticket, u_int8_t rule_action,
434     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
435     u_int8_t check_ticket)
436 {
437 	struct pf_kruleset	*ruleset;
438 	struct pf_krule		*rule;
439 	int			 rs_num;
440 
441 	ruleset = pf_find_kruleset(anchor);
442 	if (ruleset == NULL)
443 		return (NULL);
444 	rs_num = pf_get_ruleset_number(rule_action);
445 	if (rs_num >= PF_RULESET_MAX)
446 		return (NULL);
447 	if (active) {
448 		if (check_ticket && ticket !=
449 		    ruleset->rules[rs_num].active.ticket)
450 			return (NULL);
451 		if (r_last)
452 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
453 			    pf_krulequeue);
454 		else
455 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
456 	} else {
457 		if (check_ticket && ticket !=
458 		    ruleset->rules[rs_num].inactive.ticket)
459 			return (NULL);
460 		if (r_last)
461 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
462 			    pf_krulequeue);
463 		else
464 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
465 	}
466 	if (!r_last) {
467 		while ((rule != NULL) && (rule->nr != rule_number))
468 			rule = TAILQ_NEXT(rule, entries);
469 	}
470 	if (rule == NULL)
471 		return (NULL);
472 
473 	return (&rule->rpool);
474 }
475 
476 static void
477 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb)
478 {
479 	struct pf_kpooladdr	*mv_pool_pa;
480 
481 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
482 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
483 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
484 	}
485 }
486 
487 static void
488 pf_empty_kpool(struct pf_kpalist *poola)
489 {
490 	struct pf_kpooladdr *pa;
491 
492 	while ((pa = TAILQ_FIRST(poola)) != NULL) {
493 		switch (pa->addr.type) {
494 		case PF_ADDR_DYNIFTL:
495 			pfi_dynaddr_remove(pa->addr.p.dyn);
496 			break;
497 		case PF_ADDR_TABLE:
498 			/* XXX: this could be unfinished pooladdr on pabuf */
499 			if (pa->addr.p.tbl != NULL)
500 				pfr_detach_table(pa->addr.p.tbl);
501 			break;
502 		}
503 		if (pa->kif)
504 			pfi_kkif_unref(pa->kif);
505 		TAILQ_REMOVE(poola, pa, entries);
506 		free(pa, M_PFRULE);
507 	}
508 }
509 
510 static void
511 pf_unlink_rule_locked(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
512 {
513 
514 	PF_RULES_WASSERT();
515 	PF_UNLNKDRULES_ASSERT();
516 
517 	TAILQ_REMOVE(rulequeue, rule, entries);
518 
519 	rule->rule_ref |= PFRULE_REFS;
520 	TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
521 }
522 
523 static void
524 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
525 {
526 
527 	PF_RULES_WASSERT();
528 
529 	PF_UNLNKDRULES_LOCK();
530 	pf_unlink_rule_locked(rulequeue, rule);
531 	PF_UNLNKDRULES_UNLOCK();
532 }
533 
534 static void
535 pf_free_eth_rule(struct pf_keth_rule *rule)
536 {
537 	PF_RULES_WASSERT();
538 
539 	if (rule == NULL)
540 		return;
541 
542 	if (rule->tag)
543 		tag_unref(&V_pf_tags, rule->tag);
544 	if (rule->match_tag)
545 		tag_unref(&V_pf_tags, rule->match_tag);
546 #ifdef ALTQ
547 	pf_qid_unref(rule->qid);
548 #endif
549 
550 	if (rule->bridge_to)
551 		pfi_kkif_unref(rule->bridge_to);
552 	if (rule->kif)
553 		pfi_kkif_unref(rule->kif);
554 
555 	if (rule->ipsrc.addr.type == PF_ADDR_TABLE)
556 		pfr_detach_table(rule->ipsrc.addr.p.tbl);
557 	if (rule->ipdst.addr.type == PF_ADDR_TABLE)
558 		pfr_detach_table(rule->ipdst.addr.p.tbl);
559 
560 	counter_u64_free(rule->evaluations);
561 	for (int i = 0; i < 2; i++) {
562 		counter_u64_free(rule->packets[i]);
563 		counter_u64_free(rule->bytes[i]);
564 	}
565 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
566 	pf_keth_anchor_remove(rule);
567 
568 	free(rule, M_PFRULE);
569 }
570 
571 void
572 pf_free_rule(struct pf_krule *rule)
573 {
574 
575 	PF_RULES_WASSERT();
576 	PF_CONFIG_ASSERT();
577 
578 	if (rule->tag)
579 		tag_unref(&V_pf_tags, rule->tag);
580 	if (rule->match_tag)
581 		tag_unref(&V_pf_tags, rule->match_tag);
582 #ifdef ALTQ
583 	if (rule->pqid != rule->qid)
584 		pf_qid_unref(rule->pqid);
585 	pf_qid_unref(rule->qid);
586 #endif
587 	switch (rule->src.addr.type) {
588 	case PF_ADDR_DYNIFTL:
589 		pfi_dynaddr_remove(rule->src.addr.p.dyn);
590 		break;
591 	case PF_ADDR_TABLE:
592 		pfr_detach_table(rule->src.addr.p.tbl);
593 		break;
594 	}
595 	switch (rule->dst.addr.type) {
596 	case PF_ADDR_DYNIFTL:
597 		pfi_dynaddr_remove(rule->dst.addr.p.dyn);
598 		break;
599 	case PF_ADDR_TABLE:
600 		pfr_detach_table(rule->dst.addr.p.tbl);
601 		break;
602 	}
603 	if (rule->overload_tbl)
604 		pfr_detach_table(rule->overload_tbl);
605 	if (rule->kif)
606 		pfi_kkif_unref(rule->kif);
607 	pf_kanchor_remove(rule);
608 	pf_empty_kpool(&rule->rpool.list);
609 
610 	pf_krule_free(rule);
611 }
612 
613 static void
614 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
615     unsigned int default_size)
616 {
617 	unsigned int i;
618 	unsigned int hashsize;
619 
620 	if (*tunable_size == 0 || !powerof2(*tunable_size))
621 		*tunable_size = default_size;
622 
623 	hashsize = *tunable_size;
624 	ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
625 	    M_WAITOK);
626 	ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
627 	    M_WAITOK);
628 	ts->mask = hashsize - 1;
629 	ts->seed = arc4random();
630 	for (i = 0; i < hashsize; i++) {
631 		TAILQ_INIT(&ts->namehash[i]);
632 		TAILQ_INIT(&ts->taghash[i]);
633 	}
634 	BIT_FILL(TAGID_MAX, &ts->avail);
635 }
636 
637 static void
638 pf_cleanup_tagset(struct pf_tagset *ts)
639 {
640 	unsigned int i;
641 	unsigned int hashsize;
642 	struct pf_tagname *t, *tmp;
643 
644 	/*
645 	 * Only need to clean up one of the hashes as each tag is hashed
646 	 * into each table.
647 	 */
648 	hashsize = ts->mask + 1;
649 	for (i = 0; i < hashsize; i++)
650 		TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
651 			uma_zfree(V_pf_tag_z, t);
652 
653 	free(ts->namehash, M_PFHASH);
654 	free(ts->taghash, M_PFHASH);
655 }
656 
657 static uint16_t
658 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
659 {
660 	size_t len;
661 
662 	len = strnlen(tagname, PF_TAG_NAME_SIZE - 1);
663 	return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask);
664 }
665 
666 static uint16_t
667 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
668 {
669 
670 	return (tag & ts->mask);
671 }
672 
673 static u_int16_t
674 tagname2tag(struct pf_tagset *ts, const char *tagname)
675 {
676 	struct pf_tagname	*tag;
677 	u_int32_t		 index;
678 	u_int16_t		 new_tagid;
679 
680 	PF_RULES_WASSERT();
681 
682 	index = tagname2hashindex(ts, tagname);
683 	TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
684 		if (strcmp(tagname, tag->name) == 0) {
685 			tag->ref++;
686 			return (tag->tag);
687 		}
688 
689 	/*
690 	 * new entry
691 	 *
692 	 * to avoid fragmentation, we do a linear search from the beginning
693 	 * and take the first free slot we find.
694 	 */
695 	new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
696 	/*
697 	 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
698 	 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
699 	 * set.  It may also return a bit number greater than TAGID_MAX due
700 	 * to rounding of the number of bits in the vector up to a multiple
701 	 * of the vector word size at declaration/allocation time.
702 	 */
703 	if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
704 		return (0);
705 
706 	/* Mark the tag as in use.  Bits are 0-based for BIT_CLR() */
707 	BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
708 
709 	/* allocate and fill new struct pf_tagname */
710 	tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
711 	if (tag == NULL)
712 		return (0);
713 	strlcpy(tag->name, tagname, sizeof(tag->name));
714 	tag->tag = new_tagid;
715 	tag->ref = 1;
716 
717 	/* Insert into namehash */
718 	TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
719 
720 	/* Insert into taghash */
721 	index = tag2hashindex(ts, new_tagid);
722 	TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
723 
724 	return (tag->tag);
725 }
726 
727 static void
728 tag_unref(struct pf_tagset *ts, u_int16_t tag)
729 {
730 	struct pf_tagname	*t;
731 	uint16_t		 index;
732 
733 	PF_RULES_WASSERT();
734 
735 	index = tag2hashindex(ts, tag);
736 	TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
737 		if (tag == t->tag) {
738 			if (--t->ref == 0) {
739 				TAILQ_REMOVE(&ts->taghash[index], t,
740 				    taghash_entries);
741 				index = tagname2hashindex(ts, t->name);
742 				TAILQ_REMOVE(&ts->namehash[index], t,
743 				    namehash_entries);
744 				/* Bits are 0-based for BIT_SET() */
745 				BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
746 				uma_zfree(V_pf_tag_z, t);
747 			}
748 			break;
749 		}
750 }
751 
752 static uint16_t
753 pf_tagname2tag(const char *tagname)
754 {
755 	return (tagname2tag(&V_pf_tags, tagname));
756 }
757 
758 static int
759 pf_begin_eth(uint32_t *ticket, const char *anchor)
760 {
761 	struct pf_keth_rule *rule, *tmp;
762 	struct pf_keth_ruleset *rs;
763 
764 	PF_RULES_WASSERT();
765 
766 	rs = pf_find_or_create_keth_ruleset(anchor);
767 	if (rs == NULL)
768 		return (EINVAL);
769 
770 	/* Purge old inactive rules. */
771 	TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
772 	    tmp) {
773 		TAILQ_REMOVE(rs->inactive.rules, rule,
774 		    entries);
775 		pf_free_eth_rule(rule);
776 	}
777 
778 	*ticket = ++rs->inactive.ticket;
779 	rs->inactive.open = 1;
780 
781 	return (0);
782 }
783 
784 static void
785 pf_rollback_eth_cb(struct epoch_context *ctx)
786 {
787 	struct pf_keth_ruleset *rs;
788 
789 	rs = __containerof(ctx, struct pf_keth_ruleset, epoch_ctx);
790 
791 	CURVNET_SET(rs->vnet);
792 
793 	PF_RULES_WLOCK();
794 	pf_rollback_eth(rs->inactive.ticket,
795 	    rs->anchor ? rs->anchor->path : "");
796 	PF_RULES_WUNLOCK();
797 
798 	CURVNET_RESTORE();
799 }
800 
801 static int
802 pf_rollback_eth(uint32_t ticket, const char *anchor)
803 {
804 	struct pf_keth_rule *rule, *tmp;
805 	struct pf_keth_ruleset *rs;
806 
807 	PF_RULES_WASSERT();
808 
809 	rs = pf_find_keth_ruleset(anchor);
810 	if (rs == NULL)
811 		return (EINVAL);
812 
813 	if (!rs->inactive.open ||
814 	    ticket != rs->inactive.ticket)
815 		return (0);
816 
817 	/* Purge old inactive rules. */
818 	TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
819 	    tmp) {
820 		TAILQ_REMOVE(rs->inactive.rules, rule, entries);
821 		pf_free_eth_rule(rule);
822 	}
823 
824 	rs->inactive.open = 0;
825 
826 	pf_remove_if_empty_keth_ruleset(rs);
827 
828 	return (0);
829 }
830 
831 #define	PF_SET_SKIP_STEPS(i)					\
832 	do {							\
833 		while (head[i] != cur) {			\
834 			head[i]->skip[i].ptr = cur;		\
835 			head[i] = TAILQ_NEXT(head[i], entries);	\
836 		}						\
837 	} while (0)
838 
839 static void
840 pf_eth_calc_skip_steps(struct pf_keth_ruleq *rules)
841 {
842 	struct pf_keth_rule *cur, *prev, *head[PFE_SKIP_COUNT];
843 	int i;
844 
845 	cur = TAILQ_FIRST(rules);
846 	prev = cur;
847 	for (i = 0; i < PFE_SKIP_COUNT; ++i)
848 		head[i] = cur;
849 	while (cur != NULL) {
850 		if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
851 			PF_SET_SKIP_STEPS(PFE_SKIP_IFP);
852 		if (cur->direction != prev->direction)
853 			PF_SET_SKIP_STEPS(PFE_SKIP_DIR);
854 		if (cur->proto != prev->proto)
855 			PF_SET_SKIP_STEPS(PFE_SKIP_PROTO);
856 		if (memcmp(&cur->src, &prev->src, sizeof(cur->src)) != 0)
857 			PF_SET_SKIP_STEPS(PFE_SKIP_SRC_ADDR);
858 		if (memcmp(&cur->dst, &prev->dst, sizeof(cur->dst)) != 0)
859 			PF_SET_SKIP_STEPS(PFE_SKIP_DST_ADDR);
860 		if (cur->ipsrc.neg != prev->ipsrc.neg ||
861 		    pf_addr_wrap_neq(&cur->ipsrc.addr, &prev->ipsrc.addr))
862 			PF_SET_SKIP_STEPS(PFE_SKIP_SRC_IP_ADDR);
863 		if (cur->ipdst.neg != prev->ipdst.neg ||
864 		    pf_addr_wrap_neq(&cur->ipdst.addr, &prev->ipdst.addr))
865 			PF_SET_SKIP_STEPS(PFE_SKIP_DST_IP_ADDR);
866 
867 		prev = cur;
868 		cur = TAILQ_NEXT(cur, entries);
869 	}
870 	for (i = 0; i < PFE_SKIP_COUNT; ++i)
871 		PF_SET_SKIP_STEPS(i);
872 }
873 
874 static int
875 pf_commit_eth(uint32_t ticket, const char *anchor)
876 {
877 	struct pf_keth_ruleq *rules;
878 	struct pf_keth_ruleset *rs;
879 
880 	rs = pf_find_keth_ruleset(anchor);
881 	if (rs == NULL) {
882 		return (EINVAL);
883 	}
884 
885 	if (!rs->inactive.open ||
886 	    ticket != rs->inactive.ticket)
887 		return (EBUSY);
888 
889 	PF_RULES_WASSERT();
890 
891 	pf_eth_calc_skip_steps(rs->inactive.rules);
892 
893 	rules = rs->active.rules;
894 	ck_pr_store_ptr(&rs->active.rules, rs->inactive.rules);
895 	rs->inactive.rules = rules;
896 	rs->inactive.ticket = rs->active.ticket;
897 
898 	/* Clean up inactive rules (i.e. previously active rules), only when
899 	 * we're sure they're no longer used. */
900 	NET_EPOCH_CALL(pf_rollback_eth_cb, &rs->epoch_ctx);
901 
902 	return (0);
903 }
904 
905 #ifdef ALTQ
906 static uint16_t
907 pf_qname2qid(const char *qname)
908 {
909 	return (tagname2tag(&V_pf_qids, qname));
910 }
911 
912 static void
913 pf_qid_unref(uint16_t qid)
914 {
915 	tag_unref(&V_pf_qids, qid);
916 }
917 
918 static int
919 pf_begin_altq(u_int32_t *ticket)
920 {
921 	struct pf_altq	*altq, *tmp;
922 	int		 error = 0;
923 
924 	PF_RULES_WASSERT();
925 
926 	/* Purge the old altq lists */
927 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
928 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
929 			/* detach and destroy the discipline */
930 			error = altq_remove(altq);
931 		}
932 		free(altq, M_PFALTQ);
933 	}
934 	TAILQ_INIT(V_pf_altq_ifs_inactive);
935 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
936 		pf_qid_unref(altq->qid);
937 		free(altq, M_PFALTQ);
938 	}
939 	TAILQ_INIT(V_pf_altqs_inactive);
940 	if (error)
941 		return (error);
942 	*ticket = ++V_ticket_altqs_inactive;
943 	V_altqs_inactive_open = 1;
944 	return (0);
945 }
946 
947 static int
948 pf_rollback_altq(u_int32_t ticket)
949 {
950 	struct pf_altq	*altq, *tmp;
951 	int		 error = 0;
952 
953 	PF_RULES_WASSERT();
954 
955 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
956 		return (0);
957 	/* Purge the old altq lists */
958 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
959 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
960 			/* detach and destroy the discipline */
961 			error = altq_remove(altq);
962 		}
963 		free(altq, M_PFALTQ);
964 	}
965 	TAILQ_INIT(V_pf_altq_ifs_inactive);
966 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
967 		pf_qid_unref(altq->qid);
968 		free(altq, M_PFALTQ);
969 	}
970 	TAILQ_INIT(V_pf_altqs_inactive);
971 	V_altqs_inactive_open = 0;
972 	return (error);
973 }
974 
975 static int
976 pf_commit_altq(u_int32_t ticket)
977 {
978 	struct pf_altqqueue	*old_altqs, *old_altq_ifs;
979 	struct pf_altq		*altq, *tmp;
980 	int			 err, error = 0;
981 
982 	PF_RULES_WASSERT();
983 
984 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
985 		return (EBUSY);
986 
987 	/* swap altqs, keep the old. */
988 	old_altqs = V_pf_altqs_active;
989 	old_altq_ifs = V_pf_altq_ifs_active;
990 	V_pf_altqs_active = V_pf_altqs_inactive;
991 	V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
992 	V_pf_altqs_inactive = old_altqs;
993 	V_pf_altq_ifs_inactive = old_altq_ifs;
994 	V_ticket_altqs_active = V_ticket_altqs_inactive;
995 
996 	/* Attach new disciplines */
997 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
998 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
999 			/* attach the discipline */
1000 			error = altq_pfattach(altq);
1001 			if (error == 0 && V_pf_altq_running)
1002 				error = pf_enable_altq(altq);
1003 			if (error != 0)
1004 				return (error);
1005 		}
1006 	}
1007 
1008 	/* Purge the old altq lists */
1009 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
1010 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
1011 			/* detach and destroy the discipline */
1012 			if (V_pf_altq_running)
1013 				error = pf_disable_altq(altq);
1014 			err = altq_pfdetach(altq);
1015 			if (err != 0 && error == 0)
1016 				error = err;
1017 			err = altq_remove(altq);
1018 			if (err != 0 && error == 0)
1019 				error = err;
1020 		}
1021 		free(altq, M_PFALTQ);
1022 	}
1023 	TAILQ_INIT(V_pf_altq_ifs_inactive);
1024 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
1025 		pf_qid_unref(altq->qid);
1026 		free(altq, M_PFALTQ);
1027 	}
1028 	TAILQ_INIT(V_pf_altqs_inactive);
1029 
1030 	V_altqs_inactive_open = 0;
1031 	return (error);
1032 }
1033 
1034 static int
1035 pf_enable_altq(struct pf_altq *altq)
1036 {
1037 	struct ifnet		*ifp;
1038 	struct tb_profile	 tb;
1039 	int			 error = 0;
1040 
1041 	if ((ifp = ifunit(altq->ifname)) == NULL)
1042 		return (EINVAL);
1043 
1044 	if (ifp->if_snd.altq_type != ALTQT_NONE)
1045 		error = altq_enable(&ifp->if_snd);
1046 
1047 	/* set tokenbucket regulator */
1048 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
1049 		tb.rate = altq->ifbandwidth;
1050 		tb.depth = altq->tbrsize;
1051 		error = tbr_set(&ifp->if_snd, &tb);
1052 	}
1053 
1054 	return (error);
1055 }
1056 
1057 static int
1058 pf_disable_altq(struct pf_altq *altq)
1059 {
1060 	struct ifnet		*ifp;
1061 	struct tb_profile	 tb;
1062 	int			 error;
1063 
1064 	if ((ifp = ifunit(altq->ifname)) == NULL)
1065 		return (EINVAL);
1066 
1067 	/*
1068 	 * when the discipline is no longer referenced, it was overridden
1069 	 * by a new one.  if so, just return.
1070 	 */
1071 	if (altq->altq_disc != ifp->if_snd.altq_disc)
1072 		return (0);
1073 
1074 	error = altq_disable(&ifp->if_snd);
1075 
1076 	if (error == 0) {
1077 		/* clear tokenbucket regulator */
1078 		tb.rate = 0;
1079 		error = tbr_set(&ifp->if_snd, &tb);
1080 	}
1081 
1082 	return (error);
1083 }
1084 
1085 static int
1086 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
1087     struct pf_altq *altq)
1088 {
1089 	struct ifnet	*ifp1;
1090 	int		 error = 0;
1091 
1092 	/* Deactivate the interface in question */
1093 	altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
1094 	if ((ifp1 = ifunit(altq->ifname)) == NULL ||
1095 	    (remove && ifp1 == ifp)) {
1096 		altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
1097 	} else {
1098 		error = altq_add(ifp1, altq);
1099 
1100 		if (ticket != V_ticket_altqs_inactive)
1101 			error = EBUSY;
1102 
1103 		if (error)
1104 			free(altq, M_PFALTQ);
1105 	}
1106 
1107 	return (error);
1108 }
1109 
1110 void
1111 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
1112 {
1113 	struct pf_altq	*a1, *a2, *a3;
1114 	u_int32_t	 ticket;
1115 	int		 error = 0;
1116 
1117 	/*
1118 	 * No need to re-evaluate the configuration for events on interfaces
1119 	 * that do not support ALTQ, as it's not possible for such
1120 	 * interfaces to be part of the configuration.
1121 	 */
1122 	if (!ALTQ_IS_READY(&ifp->if_snd))
1123 		return;
1124 
1125 	/* Interrupt userland queue modifications */
1126 	if (V_altqs_inactive_open)
1127 		pf_rollback_altq(V_ticket_altqs_inactive);
1128 
1129 	/* Start new altq ruleset */
1130 	if (pf_begin_altq(&ticket))
1131 		return;
1132 
1133 	/* Copy the current active set */
1134 	TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
1135 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1136 		if (a2 == NULL) {
1137 			error = ENOMEM;
1138 			break;
1139 		}
1140 		bcopy(a1, a2, sizeof(struct pf_altq));
1141 
1142 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1143 		if (error)
1144 			break;
1145 
1146 		TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
1147 	}
1148 	if (error)
1149 		goto out;
1150 	TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
1151 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1152 		if (a2 == NULL) {
1153 			error = ENOMEM;
1154 			break;
1155 		}
1156 		bcopy(a1, a2, sizeof(struct pf_altq));
1157 
1158 		if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
1159 			error = EBUSY;
1160 			free(a2, M_PFALTQ);
1161 			break;
1162 		}
1163 		a2->altq_disc = NULL;
1164 		TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
1165 			if (strncmp(a3->ifname, a2->ifname,
1166 				IFNAMSIZ) == 0) {
1167 				a2->altq_disc = a3->altq_disc;
1168 				break;
1169 			}
1170 		}
1171 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1172 		if (error)
1173 			break;
1174 
1175 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
1176 	}
1177 
1178 out:
1179 	if (error != 0)
1180 		pf_rollback_altq(ticket);
1181 	else
1182 		pf_commit_altq(ticket);
1183 }
1184 #endif /* ALTQ */
1185 
1186 static struct pf_krule_global *
1187 pf_rule_tree_alloc(int flags)
1188 {
1189 	struct pf_krule_global *tree;
1190 
1191 	tree = malloc(sizeof(struct pf_krule_global), M_TEMP, flags);
1192 	if (tree == NULL)
1193 		return (NULL);
1194 	RB_INIT(tree);
1195 	return (tree);
1196 }
1197 
1198 static void
1199 pf_rule_tree_free(struct pf_krule_global *tree)
1200 {
1201 
1202 	free(tree, M_TEMP);
1203 }
1204 
1205 static int
1206 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
1207 {
1208 	struct pf_krule_global *tree;
1209 	struct pf_kruleset	*rs;
1210 	struct pf_krule		*rule;
1211 
1212 	PF_RULES_WASSERT();
1213 
1214 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1215 		return (EINVAL);
1216 	tree = pf_rule_tree_alloc(M_NOWAIT);
1217 	if (tree == NULL)
1218 		return (ENOMEM);
1219 	rs = pf_find_or_create_kruleset(anchor);
1220 	if (rs == NULL) {
1221 		free(tree, M_TEMP);
1222 		return (EINVAL);
1223 	}
1224 	pf_rule_tree_free(rs->rules[rs_num].inactive.tree);
1225 	rs->rules[rs_num].inactive.tree = tree;
1226 
1227 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1228 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1229 		rs->rules[rs_num].inactive.rcount--;
1230 	}
1231 	*ticket = ++rs->rules[rs_num].inactive.ticket;
1232 	rs->rules[rs_num].inactive.open = 1;
1233 	return (0);
1234 }
1235 
1236 static int
1237 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
1238 {
1239 	struct pf_kruleset	*rs;
1240 	struct pf_krule		*rule;
1241 
1242 	PF_RULES_WASSERT();
1243 
1244 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1245 		return (EINVAL);
1246 	rs = pf_find_kruleset(anchor);
1247 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1248 	    rs->rules[rs_num].inactive.ticket != ticket)
1249 		return (0);
1250 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1251 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1252 		rs->rules[rs_num].inactive.rcount--;
1253 	}
1254 	rs->rules[rs_num].inactive.open = 0;
1255 	return (0);
1256 }
1257 
1258 #define PF_MD5_UPD(st, elm)						\
1259 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
1260 
1261 #define PF_MD5_UPD_STR(st, elm)						\
1262 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
1263 
1264 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
1265 		(stor) = htonl((st)->elm);				\
1266 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
1267 } while (0)
1268 
1269 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
1270 		(stor) = htons((st)->elm);				\
1271 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
1272 } while (0)
1273 
1274 static void
1275 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
1276 {
1277 	PF_MD5_UPD(pfr, addr.type);
1278 	switch (pfr->addr.type) {
1279 		case PF_ADDR_DYNIFTL:
1280 			PF_MD5_UPD(pfr, addr.v.ifname);
1281 			PF_MD5_UPD(pfr, addr.iflags);
1282 			break;
1283 		case PF_ADDR_TABLE:
1284 			PF_MD5_UPD(pfr, addr.v.tblname);
1285 			break;
1286 		case PF_ADDR_ADDRMASK:
1287 			/* XXX ignore af? */
1288 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
1289 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
1290 			break;
1291 	}
1292 
1293 	PF_MD5_UPD(pfr, port[0]);
1294 	PF_MD5_UPD(pfr, port[1]);
1295 	PF_MD5_UPD(pfr, neg);
1296 	PF_MD5_UPD(pfr, port_op);
1297 }
1298 
1299 static void
1300 pf_hash_rule_rolling(MD5_CTX *ctx, struct pf_krule *rule)
1301 {
1302 	u_int16_t x;
1303 	u_int32_t y;
1304 
1305 	pf_hash_rule_addr(ctx, &rule->src);
1306 	pf_hash_rule_addr(ctx, &rule->dst);
1307 	for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++)
1308 		PF_MD5_UPD_STR(rule, label[i]);
1309 	PF_MD5_UPD_STR(rule, ifname);
1310 	PF_MD5_UPD_STR(rule, match_tagname);
1311 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
1312 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
1313 	PF_MD5_UPD_HTONL(rule, prob, y);
1314 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
1315 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1316 	PF_MD5_UPD(rule, uid.op);
1317 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1318 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1319 	PF_MD5_UPD(rule, gid.op);
1320 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
1321 	PF_MD5_UPD(rule, action);
1322 	PF_MD5_UPD(rule, direction);
1323 	PF_MD5_UPD(rule, af);
1324 	PF_MD5_UPD(rule, quick);
1325 	PF_MD5_UPD(rule, ifnot);
1326 	PF_MD5_UPD(rule, match_tag_not);
1327 	PF_MD5_UPD(rule, natpass);
1328 	PF_MD5_UPD(rule, keep_state);
1329 	PF_MD5_UPD(rule, proto);
1330 	PF_MD5_UPD(rule, type);
1331 	PF_MD5_UPD(rule, code);
1332 	PF_MD5_UPD(rule, flags);
1333 	PF_MD5_UPD(rule, flagset);
1334 	PF_MD5_UPD(rule, allow_opts);
1335 	PF_MD5_UPD(rule, rt);
1336 	PF_MD5_UPD(rule, tos);
1337 	PF_MD5_UPD(rule, scrub_flags);
1338 	PF_MD5_UPD(rule, min_ttl);
1339 	PF_MD5_UPD(rule, set_tos);
1340 	if (rule->anchor != NULL)
1341 		PF_MD5_UPD_STR(rule, anchor->path);
1342 }
1343 
1344 static void
1345 pf_hash_rule(struct pf_krule *rule)
1346 {
1347 	MD5_CTX		ctx;
1348 
1349 	MD5Init(&ctx);
1350 	pf_hash_rule_rolling(&ctx, rule);
1351 	MD5Final(rule->md5sum, &ctx);
1352 }
1353 
1354 static int
1355 pf_krule_compare(struct pf_krule *a, struct pf_krule *b)
1356 {
1357 
1358 	return (memcmp(a->md5sum, b->md5sum, PF_MD5_DIGEST_LENGTH));
1359 }
1360 
1361 static int
1362 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1363 {
1364 	struct pf_kruleset	*rs;
1365 	struct pf_krule		*rule, **old_array, *old_rule;
1366 	struct pf_krulequeue	*old_rules;
1367 	struct pf_krule_global  *old_tree;
1368 	int			 error;
1369 	u_int32_t		 old_rcount;
1370 
1371 	PF_RULES_WASSERT();
1372 
1373 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1374 		return (EINVAL);
1375 	rs = pf_find_kruleset(anchor);
1376 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1377 	    ticket != rs->rules[rs_num].inactive.ticket)
1378 		return (EBUSY);
1379 
1380 	/* Calculate checksum for the main ruleset */
1381 	if (rs == &pf_main_ruleset) {
1382 		error = pf_setup_pfsync_matching(rs);
1383 		if (error != 0)
1384 			return (error);
1385 	}
1386 
1387 	/* Swap rules, keep the old. */
1388 	old_rules = rs->rules[rs_num].active.ptr;
1389 	old_rcount = rs->rules[rs_num].active.rcount;
1390 	old_array = rs->rules[rs_num].active.ptr_array;
1391 	old_tree = rs->rules[rs_num].active.tree;
1392 
1393 	rs->rules[rs_num].active.ptr =
1394 	    rs->rules[rs_num].inactive.ptr;
1395 	rs->rules[rs_num].active.ptr_array =
1396 	    rs->rules[rs_num].inactive.ptr_array;
1397 	rs->rules[rs_num].active.tree =
1398 	    rs->rules[rs_num].inactive.tree;
1399 	rs->rules[rs_num].active.rcount =
1400 	    rs->rules[rs_num].inactive.rcount;
1401 
1402 	/* Attempt to preserve counter information. */
1403 	if (V_pf_status.keep_counters && old_tree != NULL) {
1404 		TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr,
1405 		    entries) {
1406 			old_rule = RB_FIND(pf_krule_global, old_tree, rule);
1407 			if (old_rule == NULL) {
1408 				continue;
1409 			}
1410 			pf_counter_u64_critical_enter();
1411 			pf_counter_u64_add_protected(&rule->evaluations,
1412 			    pf_counter_u64_fetch(&old_rule->evaluations));
1413 			pf_counter_u64_add_protected(&rule->packets[0],
1414 			    pf_counter_u64_fetch(&old_rule->packets[0]));
1415 			pf_counter_u64_add_protected(&rule->packets[1],
1416 			    pf_counter_u64_fetch(&old_rule->packets[1]));
1417 			pf_counter_u64_add_protected(&rule->bytes[0],
1418 			    pf_counter_u64_fetch(&old_rule->bytes[0]));
1419 			pf_counter_u64_add_protected(&rule->bytes[1],
1420 			    pf_counter_u64_fetch(&old_rule->bytes[1]));
1421 			pf_counter_u64_critical_exit();
1422 		}
1423 	}
1424 
1425 	rs->rules[rs_num].inactive.ptr = old_rules;
1426 	rs->rules[rs_num].inactive.ptr_array = old_array;
1427 	rs->rules[rs_num].inactive.tree = NULL; /* important for pf_ioctl_addrule */
1428 	rs->rules[rs_num].inactive.rcount = old_rcount;
1429 
1430 	rs->rules[rs_num].active.ticket =
1431 	    rs->rules[rs_num].inactive.ticket;
1432 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1433 
1434 	/* Purge the old rule list. */
1435 	PF_UNLNKDRULES_LOCK();
1436 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1437 		pf_unlink_rule_locked(old_rules, rule);
1438 	PF_UNLNKDRULES_UNLOCK();
1439 	if (rs->rules[rs_num].inactive.ptr_array)
1440 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1441 	rs->rules[rs_num].inactive.ptr_array = NULL;
1442 	rs->rules[rs_num].inactive.rcount = 0;
1443 	rs->rules[rs_num].inactive.open = 0;
1444 	pf_remove_if_empty_kruleset(rs);
1445 	free(old_tree, M_TEMP);
1446 
1447 	return (0);
1448 }
1449 
1450 static int
1451 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1452 {
1453 	MD5_CTX			 ctx;
1454 	struct pf_krule		*rule;
1455 	int			 rs_cnt;
1456 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
1457 
1458 	MD5Init(&ctx);
1459 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1460 		/* XXX PF_RULESET_SCRUB as well? */
1461 		if (rs_cnt == PF_RULESET_SCRUB)
1462 			continue;
1463 
1464 		if (rs->rules[rs_cnt].inactive.ptr_array)
1465 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1466 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
1467 
1468 		if (rs->rules[rs_cnt].inactive.rcount) {
1469 			rs->rules[rs_cnt].inactive.ptr_array =
1470 			    mallocarray(rs->rules[rs_cnt].inactive.rcount,
1471 			    sizeof(struct pf_rule **),
1472 			    M_TEMP, M_NOWAIT);
1473 
1474 			if (!rs->rules[rs_cnt].inactive.ptr_array)
1475 				return (ENOMEM);
1476 		}
1477 
1478 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1479 		    entries) {
1480 			pf_hash_rule_rolling(&ctx, rule);
1481 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1482 		}
1483 	}
1484 
1485 	MD5Final(digest, &ctx);
1486 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1487 	return (0);
1488 }
1489 
1490 static int
1491 pf_eth_addr_setup(struct pf_keth_ruleset *ruleset, struct pf_addr_wrap *addr)
1492 {
1493 	int error = 0;
1494 
1495 	switch (addr->type) {
1496 	case PF_ADDR_TABLE:
1497 		addr->p.tbl = pfr_eth_attach_table(ruleset, addr->v.tblname);
1498 		if (addr->p.tbl == NULL)
1499 			error = ENOMEM;
1500 		break;
1501 	default:
1502 		error = EINVAL;
1503 	}
1504 
1505 	return (error);
1506 }
1507 
1508 static int
1509 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1510     sa_family_t af)
1511 {
1512 	int error = 0;
1513 
1514 	switch (addr->type) {
1515 	case PF_ADDR_TABLE:
1516 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1517 		if (addr->p.tbl == NULL)
1518 			error = ENOMEM;
1519 		break;
1520 	case PF_ADDR_DYNIFTL:
1521 		error = pfi_dynaddr_setup(addr, af);
1522 		break;
1523 	}
1524 
1525 	return (error);
1526 }
1527 
1528 static void
1529 pf_addr_copyout(struct pf_addr_wrap *addr)
1530 {
1531 
1532 	switch (addr->type) {
1533 	case PF_ADDR_DYNIFTL:
1534 		pfi_dynaddr_copyout(addr);
1535 		break;
1536 	case PF_ADDR_TABLE:
1537 		pf_tbladdr_copyout(addr);
1538 		break;
1539 	}
1540 }
1541 
1542 static void
1543 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
1544 {
1545 	int	secs = time_uptime, diff;
1546 
1547 	bzero(out, sizeof(struct pf_src_node));
1548 
1549 	bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
1550 	bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
1551 
1552 	if (in->rule.ptr != NULL)
1553 		out->rule.nr = in->rule.ptr->nr;
1554 
1555 	for (int i = 0; i < 2; i++) {
1556 		out->bytes[i] = counter_u64_fetch(in->bytes[i]);
1557 		out->packets[i] = counter_u64_fetch(in->packets[i]);
1558 	}
1559 
1560 	out->states = in->states;
1561 	out->conn = in->conn;
1562 	out->af = in->af;
1563 	out->ruletype = in->ruletype;
1564 
1565 	out->creation = secs - in->creation;
1566 	if (out->expire > secs)
1567 		out->expire -= secs;
1568 	else
1569 		out->expire = 0;
1570 
1571 	/* Adjust the connection rate estimate. */
1572 	diff = secs - in->conn_rate.last;
1573 	if (diff >= in->conn_rate.seconds)
1574 		out->conn_rate.count = 0;
1575 	else
1576 		out->conn_rate.count -=
1577 		    in->conn_rate.count * diff /
1578 		    in->conn_rate.seconds;
1579 }
1580 
1581 #ifdef ALTQ
1582 /*
1583  * Handle export of struct pf_kaltq to user binaries that may be using any
1584  * version of struct pf_altq.
1585  */
1586 static int
1587 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1588 {
1589 	u_int32_t version;
1590 
1591 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1592 		version = 0;
1593 	else
1594 		version = pa->version;
1595 
1596 	if (version > PFIOC_ALTQ_VERSION)
1597 		return (EINVAL);
1598 
1599 #define ASSIGN(x) exported_q->x = q->x
1600 #define COPY(x) \
1601 	bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1602 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1603 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1604 
1605 	switch (version) {
1606 	case 0: {
1607 		struct pf_altq_v0 *exported_q =
1608 		    &((struct pfioc_altq_v0 *)pa)->altq;
1609 
1610 		COPY(ifname);
1611 
1612 		ASSIGN(scheduler);
1613 		ASSIGN(tbrsize);
1614 		exported_q->tbrsize = SATU16(q->tbrsize);
1615 		exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1616 
1617 		COPY(qname);
1618 		COPY(parent);
1619 		ASSIGN(parent_qid);
1620 		exported_q->bandwidth = SATU32(q->bandwidth);
1621 		ASSIGN(priority);
1622 		ASSIGN(local_flags);
1623 
1624 		ASSIGN(qlimit);
1625 		ASSIGN(flags);
1626 
1627 		if (q->scheduler == ALTQT_HFSC) {
1628 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1629 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1630 			    SATU32(q->pq_u.hfsc_opts.x)
1631 
1632 			ASSIGN_OPT_SATU32(rtsc_m1);
1633 			ASSIGN_OPT(rtsc_d);
1634 			ASSIGN_OPT_SATU32(rtsc_m2);
1635 
1636 			ASSIGN_OPT_SATU32(lssc_m1);
1637 			ASSIGN_OPT(lssc_d);
1638 			ASSIGN_OPT_SATU32(lssc_m2);
1639 
1640 			ASSIGN_OPT_SATU32(ulsc_m1);
1641 			ASSIGN_OPT(ulsc_d);
1642 			ASSIGN_OPT_SATU32(ulsc_m2);
1643 
1644 			ASSIGN_OPT(flags);
1645 
1646 #undef ASSIGN_OPT
1647 #undef ASSIGN_OPT_SATU32
1648 		} else
1649 			COPY(pq_u);
1650 
1651 		ASSIGN(qid);
1652 		break;
1653 	}
1654 	case 1:	{
1655 		struct pf_altq_v1 *exported_q =
1656 		    &((struct pfioc_altq_v1 *)pa)->altq;
1657 
1658 		COPY(ifname);
1659 
1660 		ASSIGN(scheduler);
1661 		ASSIGN(tbrsize);
1662 		ASSIGN(ifbandwidth);
1663 
1664 		COPY(qname);
1665 		COPY(parent);
1666 		ASSIGN(parent_qid);
1667 		ASSIGN(bandwidth);
1668 		ASSIGN(priority);
1669 		ASSIGN(local_flags);
1670 
1671 		ASSIGN(qlimit);
1672 		ASSIGN(flags);
1673 		COPY(pq_u);
1674 
1675 		ASSIGN(qid);
1676 		break;
1677 	}
1678 	default:
1679 		panic("%s: unhandled struct pfioc_altq version", __func__);
1680 		break;
1681 	}
1682 
1683 #undef ASSIGN
1684 #undef COPY
1685 #undef SATU16
1686 #undef SATU32
1687 
1688 	return (0);
1689 }
1690 
1691 /*
1692  * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1693  * that may be using any version of it.
1694  */
1695 static int
1696 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1697 {
1698 	u_int32_t version;
1699 
1700 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1701 		version = 0;
1702 	else
1703 		version = pa->version;
1704 
1705 	if (version > PFIOC_ALTQ_VERSION)
1706 		return (EINVAL);
1707 
1708 #define ASSIGN(x) q->x = imported_q->x
1709 #define COPY(x) \
1710 	bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1711 
1712 	switch (version) {
1713 	case 0: {
1714 		struct pf_altq_v0 *imported_q =
1715 		    &((struct pfioc_altq_v0 *)pa)->altq;
1716 
1717 		COPY(ifname);
1718 
1719 		ASSIGN(scheduler);
1720 		ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1721 		ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1722 
1723 		COPY(qname);
1724 		COPY(parent);
1725 		ASSIGN(parent_qid);
1726 		ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1727 		ASSIGN(priority);
1728 		ASSIGN(local_flags);
1729 
1730 		ASSIGN(qlimit);
1731 		ASSIGN(flags);
1732 
1733 		if (imported_q->scheduler == ALTQT_HFSC) {
1734 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1735 
1736 			/*
1737 			 * The m1 and m2 parameters are being copied from
1738 			 * 32-bit to 64-bit.
1739 			 */
1740 			ASSIGN_OPT(rtsc_m1);
1741 			ASSIGN_OPT(rtsc_d);
1742 			ASSIGN_OPT(rtsc_m2);
1743 
1744 			ASSIGN_OPT(lssc_m1);
1745 			ASSIGN_OPT(lssc_d);
1746 			ASSIGN_OPT(lssc_m2);
1747 
1748 			ASSIGN_OPT(ulsc_m1);
1749 			ASSIGN_OPT(ulsc_d);
1750 			ASSIGN_OPT(ulsc_m2);
1751 
1752 			ASSIGN_OPT(flags);
1753 
1754 #undef ASSIGN_OPT
1755 		} else
1756 			COPY(pq_u);
1757 
1758 		ASSIGN(qid);
1759 		break;
1760 	}
1761 	case 1: {
1762 		struct pf_altq_v1 *imported_q =
1763 		    &((struct pfioc_altq_v1 *)pa)->altq;
1764 
1765 		COPY(ifname);
1766 
1767 		ASSIGN(scheduler);
1768 		ASSIGN(tbrsize);
1769 		ASSIGN(ifbandwidth);
1770 
1771 		COPY(qname);
1772 		COPY(parent);
1773 		ASSIGN(parent_qid);
1774 		ASSIGN(bandwidth);
1775 		ASSIGN(priority);
1776 		ASSIGN(local_flags);
1777 
1778 		ASSIGN(qlimit);
1779 		ASSIGN(flags);
1780 		COPY(pq_u);
1781 
1782 		ASSIGN(qid);
1783 		break;
1784 	}
1785 	default:
1786 		panic("%s: unhandled struct pfioc_altq version", __func__);
1787 		break;
1788 	}
1789 
1790 #undef ASSIGN
1791 #undef COPY
1792 
1793 	return (0);
1794 }
1795 
1796 static struct pf_altq *
1797 pf_altq_get_nth_active(u_int32_t n)
1798 {
1799 	struct pf_altq		*altq;
1800 	u_int32_t		 nr;
1801 
1802 	nr = 0;
1803 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1804 		if (nr == n)
1805 			return (altq);
1806 		nr++;
1807 	}
1808 
1809 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1810 		if (nr == n)
1811 			return (altq);
1812 		nr++;
1813 	}
1814 
1815 	return (NULL);
1816 }
1817 #endif /* ALTQ */
1818 
1819 struct pf_krule *
1820 pf_krule_alloc(void)
1821 {
1822 	struct pf_krule *rule;
1823 
1824 	rule = malloc(sizeof(struct pf_krule), M_PFRULE, M_WAITOK | M_ZERO);
1825 	mtx_init(&rule->rpool.mtx, "pf_krule_pool", NULL, MTX_DEF);
1826 	rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
1827 	    M_WAITOK | M_ZERO);
1828 	return (rule);
1829 }
1830 
1831 void
1832 pf_krule_free(struct pf_krule *rule)
1833 {
1834 #ifdef PF_WANT_32_TO_64_COUNTER
1835 	bool wowned;
1836 #endif
1837 
1838 	if (rule == NULL)
1839 		return;
1840 
1841 #ifdef PF_WANT_32_TO_64_COUNTER
1842 	if (rule->allrulelinked) {
1843 		wowned = PF_RULES_WOWNED();
1844 		if (!wowned)
1845 			PF_RULES_WLOCK();
1846 		LIST_REMOVE(rule, allrulelist);
1847 		V_pf_allrulecount--;
1848 		if (!wowned)
1849 			PF_RULES_WUNLOCK();
1850 	}
1851 #endif
1852 
1853 	pf_counter_u64_deinit(&rule->evaluations);
1854 	for (int i = 0; i < 2; i++) {
1855 		pf_counter_u64_deinit(&rule->packets[i]);
1856 		pf_counter_u64_deinit(&rule->bytes[i]);
1857 	}
1858 	counter_u64_free(rule->states_cur);
1859 	counter_u64_free(rule->states_tot);
1860 	counter_u64_free(rule->src_nodes);
1861 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
1862 
1863 	mtx_destroy(&rule->rpool.mtx);
1864 	free(rule, M_PFRULE);
1865 }
1866 
1867 void
1868 pf_krule_clear_counters(struct pf_krule *rule)
1869 {
1870 	pf_counter_u64_zero(&rule->evaluations);
1871 	for (int i = 0; i < 2; i++) {
1872 		pf_counter_u64_zero(&rule->packets[i]);
1873 		pf_counter_u64_zero(&rule->bytes[i]);
1874 	}
1875 	counter_u64_zero(rule->states_tot);
1876 }
1877 
1878 static void
1879 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool,
1880     struct pf_pooladdr *pool)
1881 {
1882 
1883 	bzero(pool, sizeof(*pool));
1884 	bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr));
1885 	strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname));
1886 }
1887 
1888 static int
1889 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool,
1890     struct pf_kpooladdr *kpool)
1891 {
1892 	int ret;
1893 
1894 	bzero(kpool, sizeof(*kpool));
1895 	bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr));
1896 	ret = pf_user_strcpy(kpool->ifname, pool->ifname,
1897 	    sizeof(kpool->ifname));
1898 	return (ret);
1899 }
1900 
1901 static void
1902 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool)
1903 {
1904 	_Static_assert(sizeof(pool->key) == sizeof(kpool->key), "");
1905 	_Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), "");
1906 
1907 	bcopy(&pool->key, &kpool->key, sizeof(kpool->key));
1908 	bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter));
1909 
1910 	kpool->tblidx = pool->tblidx;
1911 	kpool->proxy_port[0] = pool->proxy_port[0];
1912 	kpool->proxy_port[1] = pool->proxy_port[1];
1913 	kpool->opts = pool->opts;
1914 }
1915 
1916 static int
1917 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
1918 {
1919 	int ret;
1920 
1921 #ifndef INET
1922 	if (rule->af == AF_INET) {
1923 		return (EAFNOSUPPORT);
1924 	}
1925 #endif /* INET */
1926 #ifndef INET6
1927 	if (rule->af == AF_INET6) {
1928 		return (EAFNOSUPPORT);
1929 	}
1930 #endif /* INET6 */
1931 
1932 	ret = pf_check_rule_addr(&rule->src);
1933 	if (ret != 0)
1934 		return (ret);
1935 	ret = pf_check_rule_addr(&rule->dst);
1936 	if (ret != 0)
1937 		return (ret);
1938 
1939 	bcopy(&rule->src, &krule->src, sizeof(rule->src));
1940 	bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
1941 
1942 	ret = pf_user_strcpy(krule->label[0], rule->label, sizeof(rule->label));
1943 	if (ret != 0)
1944 		return (ret);
1945 	ret = pf_user_strcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
1946 	if (ret != 0)
1947 		return (ret);
1948 	ret = pf_user_strcpy(krule->qname, rule->qname, sizeof(rule->qname));
1949 	if (ret != 0)
1950 		return (ret);
1951 	ret = pf_user_strcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
1952 	if (ret != 0)
1953 		return (ret);
1954 	ret = pf_user_strcpy(krule->tagname, rule->tagname,
1955 	    sizeof(rule->tagname));
1956 	if (ret != 0)
1957 		return (ret);
1958 	ret = pf_user_strcpy(krule->match_tagname, rule->match_tagname,
1959 	    sizeof(rule->match_tagname));
1960 	if (ret != 0)
1961 		return (ret);
1962 	ret = pf_user_strcpy(krule->overload_tblname, rule->overload_tblname,
1963 	    sizeof(rule->overload_tblname));
1964 	if (ret != 0)
1965 		return (ret);
1966 
1967 	pf_pool_to_kpool(&rule->rpool, &krule->rpool);
1968 
1969 	/* Don't allow userspace to set evaluations, packets or bytes. */
1970 	/* kif, anchor, overload_tbl are not copied over. */
1971 
1972 	krule->os_fingerprint = rule->os_fingerprint;
1973 
1974 	krule->rtableid = rule->rtableid;
1975 	/* pf_rule->timeout is smaller than pf_krule->timeout */
1976 	bcopy(rule->timeout, krule->timeout, sizeof(rule->timeout));
1977 	krule->max_states = rule->max_states;
1978 	krule->max_src_nodes = rule->max_src_nodes;
1979 	krule->max_src_states = rule->max_src_states;
1980 	krule->max_src_conn = rule->max_src_conn;
1981 	krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
1982 	krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
1983 	krule->qid = rule->qid;
1984 	krule->pqid = rule->pqid;
1985 	krule->nr = rule->nr;
1986 	krule->prob = rule->prob;
1987 	krule->cuid = rule->cuid;
1988 	krule->cpid = rule->cpid;
1989 
1990 	krule->return_icmp = rule->return_icmp;
1991 	krule->return_icmp6 = rule->return_icmp6;
1992 	krule->max_mss = rule->max_mss;
1993 	krule->tag = rule->tag;
1994 	krule->match_tag = rule->match_tag;
1995 	krule->scrub_flags = rule->scrub_flags;
1996 
1997 	bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
1998 	bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
1999 
2000 	krule->rule_flag = rule->rule_flag;
2001 	krule->action = rule->action;
2002 	krule->direction = rule->direction;
2003 	krule->log = rule->log;
2004 	krule->logif = rule->logif;
2005 	krule->quick = rule->quick;
2006 	krule->ifnot = rule->ifnot;
2007 	krule->match_tag_not = rule->match_tag_not;
2008 	krule->natpass = rule->natpass;
2009 
2010 	krule->keep_state = rule->keep_state;
2011 	krule->af = rule->af;
2012 	krule->proto = rule->proto;
2013 	krule->type = rule->type;
2014 	krule->code = rule->code;
2015 	krule->flags = rule->flags;
2016 	krule->flagset = rule->flagset;
2017 	krule->min_ttl = rule->min_ttl;
2018 	krule->allow_opts = rule->allow_opts;
2019 	krule->rt = rule->rt;
2020 	krule->return_ttl = rule->return_ttl;
2021 	krule->tos = rule->tos;
2022 	krule->set_tos = rule->set_tos;
2023 
2024 	krule->flush = rule->flush;
2025 	krule->prio = rule->prio;
2026 	krule->set_prio[0] = rule->set_prio[0];
2027 	krule->set_prio[1] = rule->set_prio[1];
2028 
2029 	bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
2030 
2031 	return (0);
2032 }
2033 
2034 int
2035 pf_ioctl_getrules(struct pfioc_rule *pr)
2036 {
2037 	struct pf_kruleset	*ruleset;
2038 	struct pf_krule		*tail;
2039 	int			 rs_num;
2040 
2041 	PF_RULES_WLOCK();
2042 	ruleset = pf_find_kruleset(pr->anchor);
2043 	if (ruleset == NULL) {
2044 		PF_RULES_WUNLOCK();
2045 		return (EINVAL);
2046 	}
2047 	rs_num = pf_get_ruleset_number(pr->rule.action);
2048 	if (rs_num >= PF_RULESET_MAX) {
2049 		PF_RULES_WUNLOCK();
2050 		return (EINVAL);
2051 	}
2052 	tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
2053 	    pf_krulequeue);
2054 	if (tail)
2055 		pr->nr = tail->nr + 1;
2056 	else
2057 		pr->nr = 0;
2058 	pr->ticket = ruleset->rules[rs_num].active.ticket;
2059 	PF_RULES_WUNLOCK();
2060 
2061 	return (0);
2062 }
2063 
2064 int
2065 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket,
2066     uint32_t pool_ticket, const char *anchor, const char *anchor_call,
2067     uid_t uid, pid_t pid)
2068 {
2069 	struct pf_kruleset	*ruleset;
2070 	struct pf_krule		*tail;
2071 	struct pf_kpooladdr	*pa;
2072 	struct pfi_kkif		*kif = NULL;
2073 	int			 rs_num;
2074 	int			 error = 0;
2075 
2076 	if ((rule->return_icmp >> 8) > ICMP_MAXTYPE) {
2077 		error = EINVAL;
2078 		goto errout_unlocked;
2079 	}
2080 
2081 #define	ERROUT(x)	ERROUT_FUNCTION(errout, x)
2082 
2083 	if (rule->ifname[0])
2084 		kif = pf_kkif_create(M_WAITOK);
2085 	pf_counter_u64_init(&rule->evaluations, M_WAITOK);
2086 	for (int i = 0; i < 2; i++) {
2087 		pf_counter_u64_init(&rule->packets[i], M_WAITOK);
2088 		pf_counter_u64_init(&rule->bytes[i], M_WAITOK);
2089 	}
2090 	rule->states_cur = counter_u64_alloc(M_WAITOK);
2091 	rule->states_tot = counter_u64_alloc(M_WAITOK);
2092 	rule->src_nodes = counter_u64_alloc(M_WAITOK);
2093 	rule->cuid = uid;
2094 	rule->cpid = pid;
2095 	TAILQ_INIT(&rule->rpool.list);
2096 
2097 	PF_CONFIG_LOCK();
2098 	PF_RULES_WLOCK();
2099 #ifdef PF_WANT_32_TO_64_COUNTER
2100 	LIST_INSERT_HEAD(&V_pf_allrulelist, rule, allrulelist);
2101 	MPASS(!rule->allrulelinked);
2102 	rule->allrulelinked = true;
2103 	V_pf_allrulecount++;
2104 #endif
2105 	ruleset = pf_find_kruleset(anchor);
2106 	if (ruleset == NULL)
2107 		ERROUT(EINVAL);
2108 	rs_num = pf_get_ruleset_number(rule->action);
2109 	if (rs_num >= PF_RULESET_MAX)
2110 		ERROUT(EINVAL);
2111 	if (ticket != ruleset->rules[rs_num].inactive.ticket) {
2112 		DPFPRINTF(PF_DEBUG_MISC,
2113 		    ("ticket: %d != [%d]%d\n", ticket, rs_num,
2114 		    ruleset->rules[rs_num].inactive.ticket));
2115 		ERROUT(EBUSY);
2116 	}
2117 	if (pool_ticket != V_ticket_pabuf) {
2118 		DPFPRINTF(PF_DEBUG_MISC,
2119 		    ("pool_ticket: %d != %d\n", pool_ticket,
2120 		    V_ticket_pabuf));
2121 		ERROUT(EBUSY);
2122 	}
2123 	/*
2124 	 * XXXMJG hack: there is no mechanism to ensure they started the
2125 	 * transaction. Ticket checked above may happen to match by accident,
2126 	 * even if nobody called DIOCXBEGIN, let alone this process.
2127 	 * Partially work around it by checking if the RB tree got allocated,
2128 	 * see pf_begin_rules.
2129 	 */
2130 	if (ruleset->rules[rs_num].inactive.tree == NULL) {
2131 		ERROUT(EINVAL);
2132 	}
2133 
2134 	tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
2135 	    pf_krulequeue);
2136 	if (tail)
2137 		rule->nr = tail->nr + 1;
2138 	else
2139 		rule->nr = 0;
2140 	if (rule->ifname[0]) {
2141 		rule->kif = pfi_kkif_attach(kif, rule->ifname);
2142 		kif = NULL;
2143 		pfi_kkif_ref(rule->kif);
2144 	} else
2145 		rule->kif = NULL;
2146 
2147 	if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
2148 		error = EBUSY;
2149 
2150 #ifdef ALTQ
2151 	/* set queue IDs */
2152 	if (rule->qname[0] != 0) {
2153 		if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2154 			error = EBUSY;
2155 		else if (rule->pqname[0] != 0) {
2156 			if ((rule->pqid =
2157 			    pf_qname2qid(rule->pqname)) == 0)
2158 				error = EBUSY;
2159 		} else
2160 			rule->pqid = rule->qid;
2161 	}
2162 #endif
2163 	if (rule->tagname[0])
2164 		if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2165 			error = EBUSY;
2166 	if (rule->match_tagname[0])
2167 		if ((rule->match_tag =
2168 		    pf_tagname2tag(rule->match_tagname)) == 0)
2169 			error = EBUSY;
2170 	if (rule->rt && !rule->direction)
2171 		error = EINVAL;
2172 	if (!rule->log)
2173 		rule->logif = 0;
2174 	if (rule->logif >= PFLOGIFS_MAX)
2175 		error = EINVAL;
2176 	if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
2177 		error = ENOMEM;
2178 	if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
2179 		error = ENOMEM;
2180 	if (pf_kanchor_setup(rule, ruleset, anchor_call))
2181 		error = EINVAL;
2182 	if (rule->scrub_flags & PFSTATE_SETPRIO &&
2183 	    (rule->set_prio[0] > PF_PRIO_MAX ||
2184 	    rule->set_prio[1] > PF_PRIO_MAX))
2185 		error = EINVAL;
2186 	TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
2187 		if (pa->addr.type == PF_ADDR_TABLE) {
2188 			pa->addr.p.tbl = pfr_attach_table(ruleset,
2189 			    pa->addr.v.tblname);
2190 			if (pa->addr.p.tbl == NULL)
2191 				error = ENOMEM;
2192 		}
2193 
2194 	rule->overload_tbl = NULL;
2195 	if (rule->overload_tblname[0]) {
2196 		if ((rule->overload_tbl = pfr_attach_table(ruleset,
2197 		    rule->overload_tblname)) == NULL)
2198 			error = EINVAL;
2199 		else
2200 			rule->overload_tbl->pfrkt_flags |=
2201 			    PFR_TFLAG_ACTIVE;
2202 	}
2203 
2204 	pf_mv_kpool(&V_pf_pabuf, &rule->rpool.list);
2205 	if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
2206 	    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
2207 	    (rule->rt > PF_NOPFROUTE)) &&
2208 	    (TAILQ_FIRST(&rule->rpool.list) == NULL))
2209 		error = EINVAL;
2210 
2211 	if (error) {
2212 		pf_free_rule(rule);
2213 		rule = NULL;
2214 		ERROUT(error);
2215 	}
2216 
2217 	rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
2218 	TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
2219 	    rule, entries);
2220 	ruleset->rules[rs_num].inactive.rcount++;
2221 
2222 	PF_RULES_WUNLOCK();
2223 	pf_hash_rule(rule);
2224 	if (RB_INSERT(pf_krule_global, ruleset->rules[rs_num].inactive.tree, rule) != NULL) {
2225 		PF_RULES_WLOCK();
2226 		TAILQ_REMOVE(ruleset->rules[rs_num].inactive.ptr, rule, entries);
2227 		ruleset->rules[rs_num].inactive.rcount--;
2228 		pf_free_rule(rule);
2229 		rule = NULL;
2230 		ERROUT(EEXIST);
2231 	}
2232 	PF_CONFIG_UNLOCK();
2233 
2234 	return (0);
2235 
2236 #undef ERROUT
2237 errout:
2238 	PF_RULES_WUNLOCK();
2239 	PF_CONFIG_UNLOCK();
2240 errout_unlocked:
2241 	pf_kkif_free(kif);
2242 	pf_krule_free(rule);
2243 	return (error);
2244 }
2245 
2246 static bool
2247 pf_label_match(const struct pf_krule *rule, const char *label)
2248 {
2249 	int i = 0;
2250 
2251 	while (*rule->label[i]) {
2252 		if (strcmp(rule->label[i], label) == 0)
2253 			return (true);
2254 		i++;
2255 	}
2256 
2257 	return (false);
2258 }
2259 
2260 static unsigned int
2261 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir)
2262 {
2263 	struct pf_kstate *s;
2264 	int more = 0;
2265 
2266 	s = pf_find_state_all(key, dir, &more);
2267 	if (s == NULL)
2268 		return (0);
2269 
2270 	if (more) {
2271 		PF_STATE_UNLOCK(s);
2272 		return (0);
2273 	}
2274 
2275 	pf_unlink_state(s);
2276 	return (1);
2277 }
2278 
2279 static int
2280 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih)
2281 {
2282 	struct pf_kstate	*s;
2283 	struct pf_state_key	*sk;
2284 	struct pf_addr		*srcaddr, *dstaddr;
2285 	struct pf_state_key_cmp	 match_key;
2286 	int			 idx, killed = 0;
2287 	unsigned int		 dir;
2288 	u_int16_t		 srcport, dstport;
2289 	struct pfi_kkif		*kif;
2290 
2291 relock_DIOCKILLSTATES:
2292 	PF_HASHROW_LOCK(ih);
2293 	LIST_FOREACH(s, &ih->states, entry) {
2294 		/* For floating states look at the original kif. */
2295 		kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
2296 
2297 		sk = s->key[psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE];
2298 		if (s->direction == PF_OUT) {
2299 			srcaddr = &sk->addr[1];
2300 			dstaddr = &sk->addr[0];
2301 			srcport = sk->port[1];
2302 			dstport = sk->port[0];
2303 		} else {
2304 			srcaddr = &sk->addr[0];
2305 			dstaddr = &sk->addr[1];
2306 			srcport = sk->port[0];
2307 			dstport = sk->port[1];
2308 		}
2309 
2310 		if (psk->psk_af && sk->af != psk->psk_af)
2311 			continue;
2312 
2313 		if (psk->psk_proto && psk->psk_proto != sk->proto)
2314 			continue;
2315 
2316 		if (! PF_MATCHA(psk->psk_src.neg, &psk->psk_src.addr.v.a.addr,
2317 		    &psk->psk_src.addr.v.a.mask, srcaddr, sk->af))
2318 			continue;
2319 
2320 		if (! PF_MATCHA(psk->psk_dst.neg, &psk->psk_dst.addr.v.a.addr,
2321 		    &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af))
2322 			continue;
2323 
2324 		if (!  PF_MATCHA(psk->psk_rt_addr.neg,
2325 		    &psk->psk_rt_addr.addr.v.a.addr,
2326 		    &psk->psk_rt_addr.addr.v.a.mask,
2327 		    &s->rt_addr, sk->af))
2328 			continue;
2329 
2330 		if (psk->psk_src.port_op != 0 &&
2331 		    ! pf_match_port(psk->psk_src.port_op,
2332 		    psk->psk_src.port[0], psk->psk_src.port[1], srcport))
2333 			continue;
2334 
2335 		if (psk->psk_dst.port_op != 0 &&
2336 		    ! pf_match_port(psk->psk_dst.port_op,
2337 		    psk->psk_dst.port[0], psk->psk_dst.port[1], dstport))
2338 			continue;
2339 
2340 		if (psk->psk_label[0] &&
2341 		    ! pf_label_match(s->rule.ptr, psk->psk_label))
2342 			continue;
2343 
2344 		if (psk->psk_ifname[0] && strcmp(psk->psk_ifname,
2345 		    kif->pfik_name))
2346 			continue;
2347 
2348 		if (psk->psk_kill_match) {
2349 			/* Create the key to find matching states, with lock
2350 			 * held. */
2351 
2352 			bzero(&match_key, sizeof(match_key));
2353 
2354 			if (s->direction == PF_OUT) {
2355 				dir = PF_IN;
2356 				idx = psk->psk_nat ? PF_SK_WIRE : PF_SK_STACK;
2357 			} else {
2358 				dir = PF_OUT;
2359 				idx = psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE;
2360 			}
2361 
2362 			match_key.af = s->key[idx]->af;
2363 			match_key.proto = s->key[idx]->proto;
2364 			PF_ACPY(&match_key.addr[0],
2365 			    &s->key[idx]->addr[1], match_key.af);
2366 			match_key.port[0] = s->key[idx]->port[1];
2367 			PF_ACPY(&match_key.addr[1],
2368 			    &s->key[idx]->addr[0], match_key.af);
2369 			match_key.port[1] = s->key[idx]->port[0];
2370 		}
2371 
2372 		pf_unlink_state(s);
2373 		killed++;
2374 
2375 		if (psk->psk_kill_match)
2376 			killed += pf_kill_matching_state(&match_key, dir);
2377 
2378 		goto relock_DIOCKILLSTATES;
2379 	}
2380 	PF_HASHROW_UNLOCK(ih);
2381 
2382 	return (killed);
2383 }
2384 
2385 int
2386 pf_start(void)
2387 {
2388 	int error = 0;
2389 
2390 	sx_xlock(&V_pf_ioctl_lock);
2391 	if (V_pf_status.running)
2392 		error = EEXIST;
2393 	else {
2394 		hook_pf();
2395 		if (! TAILQ_EMPTY(V_pf_keth->active.rules))
2396 			hook_pf_eth();
2397 		V_pf_status.running = 1;
2398 		V_pf_status.since = time_second;
2399 		new_unrhdr64(&V_pf_stateid, time_second);
2400 
2401 		DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
2402 	}
2403 	sx_xunlock(&V_pf_ioctl_lock);
2404 
2405 	return (error);
2406 }
2407 
2408 int
2409 pf_stop(void)
2410 {
2411 	int error = 0;
2412 
2413 	sx_xlock(&V_pf_ioctl_lock);
2414 	if (!V_pf_status.running)
2415 		error = ENOENT;
2416 	else {
2417 		V_pf_status.running = 0;
2418 		dehook_pf();
2419 		dehook_pf_eth();
2420 		V_pf_status.since = time_second;
2421 		DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
2422 	}
2423 	sx_xunlock(&V_pf_ioctl_lock);
2424 
2425 	return (error);
2426 }
2427 
2428 void
2429 pf_ioctl_clear_status(void)
2430 {
2431 	PF_RULES_WLOCK();
2432 	for (int i = 0; i < PFRES_MAX; i++)
2433 		counter_u64_zero(V_pf_status.counters[i]);
2434 	for (int i = 0; i < FCNT_MAX; i++)
2435 		pf_counter_u64_zero(&V_pf_status.fcounters[i]);
2436 	for (int i = 0; i < SCNT_MAX; i++)
2437 		counter_u64_zero(V_pf_status.scounters[i]);
2438 	for (int i = 0; i < KLCNT_MAX; i++)
2439 		counter_u64_zero(V_pf_status.lcounters[i]);
2440 	V_pf_status.since = time_second;
2441 	if (*V_pf_status.ifname)
2442 		pfi_update_status(V_pf_status.ifname, NULL);
2443 	PF_RULES_WUNLOCK();
2444 }
2445 
2446 static int
2447 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
2448 {
2449 	int			 error = 0;
2450 	PF_RULES_RLOCK_TRACKER;
2451 
2452 #define	ERROUT_IOCTL(target, x)					\
2453     do {								\
2454 	    error = (x);						\
2455 	    SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__);	\
2456 	    goto target;						\
2457     } while (0)
2458 
2459 
2460 	/* XXX keep in sync with switch() below */
2461 	if (securelevel_gt(td->td_ucred, 2))
2462 		switch (cmd) {
2463 		case DIOCGETRULES:
2464 		case DIOCGETRULENV:
2465 		case DIOCGETADDRS:
2466 		case DIOCGETADDR:
2467 		case DIOCGETSTATE:
2468 		case DIOCGETSTATENV:
2469 		case DIOCSETSTATUSIF:
2470 		case DIOCGETSTATUSNV:
2471 		case DIOCCLRSTATUS:
2472 		case DIOCNATLOOK:
2473 		case DIOCSETDEBUG:
2474 #ifdef COMPAT_FREEBSD14
2475 		case DIOCGETSTATES:
2476 		case DIOCGETSTATESV2:
2477 #endif
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 #ifdef COMPAT_FREEBSD14
2535 		case DIOCGETSTATES:
2536 		case DIOCGETSTATESV2:
2537 #endif
2538 		case DIOCGETTIMEOUT:
2539 		case DIOCGETLIMIT:
2540 		case DIOCGETALTQSV0:
2541 		case DIOCGETALTQSV1:
2542 		case DIOCGETALTQV0:
2543 		case DIOCGETALTQV1:
2544 		case DIOCGETQSTATSV0:
2545 		case DIOCGETQSTATSV1:
2546 		case DIOCGETRULESETS:
2547 		case DIOCGETRULESET:
2548 		case DIOCNATLOOK:
2549 		case DIOCRGETTABLES:
2550 		case DIOCRGETTSTATS:
2551 		case DIOCRGETADDRS:
2552 		case DIOCRGETASTATS:
2553 		case DIOCRTSTADDRS:
2554 		case DIOCOSFPGET:
2555 		case DIOCGETSRCNODES:
2556 		case DIOCGETSYNCOOKIES:
2557 		case DIOCIGETIFACES:
2558 		case DIOCGIFSPEEDV1:
2559 		case DIOCGIFSPEEDV0:
2560 		case DIOCGETRULENV:
2561 		case DIOCGETETHRULES:
2562 		case DIOCGETETHRULE:
2563 		case DIOCGETETHRULESETS:
2564 		case DIOCGETETHRULESET:
2565 			break;
2566 		case DIOCRCLRTABLES:
2567 		case DIOCRADDTABLES:
2568 		case DIOCRDELTABLES:
2569 		case DIOCRCLRTSTATS:
2570 		case DIOCRCLRADDRS:
2571 		case DIOCRADDADDRS:
2572 		case DIOCRDELADDRS:
2573 		case DIOCRSETADDRS:
2574 		case DIOCRSETTFLAGS:
2575 			if (((struct pfioc_table *)addr)->pfrio_flags &
2576 			    PFR_FLAG_DUMMY) {
2577 				flags |= FWRITE; /* need write lock for dummy */
2578 				break; /* dummy operation ok */
2579 			}
2580 			return (EACCES);
2581 		default:
2582 			return (EACCES);
2583 		}
2584 
2585 	CURVNET_SET(TD_TO_VNET(td));
2586 
2587 	switch (cmd) {
2588 #ifdef COMPAT_FREEBSD14
2589 	case DIOCSTART:
2590 		error = pf_start();
2591 		break;
2592 
2593 	case DIOCSTOP:
2594 		error = pf_stop();
2595 		break;
2596 #endif
2597 
2598 	case DIOCGETETHRULES: {
2599 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2600 		nvlist_t		*nvl;
2601 		void			*packed;
2602 		struct pf_keth_rule	*tail;
2603 		struct pf_keth_ruleset	*rs;
2604 		u_int32_t		 ticket, nr;
2605 		const char		*anchor = "";
2606 
2607 		nvl = NULL;
2608 		packed = NULL;
2609 
2610 #define	ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULES_error, x)
2611 
2612 		if (nv->len > pf_ioctl_maxcount)
2613 			ERROUT(ENOMEM);
2614 
2615 		/* Copy the request in */
2616 		packed = malloc(nv->len, M_NVLIST, M_WAITOK);
2617 		if (packed == NULL)
2618 			ERROUT(ENOMEM);
2619 
2620 		error = copyin(nv->data, packed, nv->len);
2621 		if (error)
2622 			ERROUT(error);
2623 
2624 		nvl = nvlist_unpack(packed, nv->len, 0);
2625 		if (nvl == NULL)
2626 			ERROUT(EBADMSG);
2627 
2628 		if (! nvlist_exists_string(nvl, "anchor"))
2629 			ERROUT(EBADMSG);
2630 
2631 		anchor = nvlist_get_string(nvl, "anchor");
2632 
2633 		rs = pf_find_keth_ruleset(anchor);
2634 
2635 		nvlist_destroy(nvl);
2636 		nvl = NULL;
2637 		free(packed, M_NVLIST);
2638 		packed = NULL;
2639 
2640 		if (rs == NULL)
2641 			ERROUT(ENOENT);
2642 
2643 		/* Reply */
2644 		nvl = nvlist_create(0);
2645 		if (nvl == NULL)
2646 			ERROUT(ENOMEM);
2647 
2648 		PF_RULES_RLOCK();
2649 
2650 		ticket = rs->active.ticket;
2651 		tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq);
2652 		if (tail)
2653 			nr = tail->nr + 1;
2654 		else
2655 			nr = 0;
2656 
2657 		PF_RULES_RUNLOCK();
2658 
2659 		nvlist_add_number(nvl, "ticket", ticket);
2660 		nvlist_add_number(nvl, "nr", nr);
2661 
2662 		packed = nvlist_pack(nvl, &nv->len);
2663 		if (packed == NULL)
2664 			ERROUT(ENOMEM);
2665 
2666 		if (nv->size == 0)
2667 			ERROUT(0);
2668 		else if (nv->size < nv->len)
2669 			ERROUT(ENOSPC);
2670 
2671 		error = copyout(packed, nv->data, nv->len);
2672 
2673 #undef ERROUT
2674 DIOCGETETHRULES_error:
2675 		free(packed, M_NVLIST);
2676 		nvlist_destroy(nvl);
2677 		break;
2678 	}
2679 
2680 	case DIOCGETETHRULE: {
2681 		struct epoch_tracker	 et;
2682 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2683 		nvlist_t		*nvl = NULL;
2684 		void			*nvlpacked = NULL;
2685 		struct pf_keth_rule	*rule = NULL;
2686 		struct pf_keth_ruleset	*rs;
2687 		u_int32_t		 ticket, nr;
2688 		bool			 clear = false;
2689 		const char		*anchor;
2690 
2691 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULE_error, x)
2692 
2693 		if (nv->len > pf_ioctl_maxcount)
2694 			ERROUT(ENOMEM);
2695 
2696 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2697 		if (nvlpacked == NULL)
2698 			ERROUT(ENOMEM);
2699 
2700 		error = copyin(nv->data, nvlpacked, nv->len);
2701 		if (error)
2702 			ERROUT(error);
2703 
2704 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2705 		if (nvl == NULL)
2706 			ERROUT(EBADMSG);
2707 		if (! nvlist_exists_number(nvl, "ticket"))
2708 			ERROUT(EBADMSG);
2709 		ticket = nvlist_get_number(nvl, "ticket");
2710 		if (! nvlist_exists_string(nvl, "anchor"))
2711 			ERROUT(EBADMSG);
2712 		anchor = nvlist_get_string(nvl, "anchor");
2713 
2714 		if (nvlist_exists_bool(nvl, "clear"))
2715 			clear = nvlist_get_bool(nvl, "clear");
2716 
2717 		if (clear && !(flags & FWRITE))
2718 			ERROUT(EACCES);
2719 
2720 		if (! nvlist_exists_number(nvl, "nr"))
2721 			ERROUT(EBADMSG);
2722 		nr = nvlist_get_number(nvl, "nr");
2723 
2724 		PF_RULES_RLOCK();
2725 		rs = pf_find_keth_ruleset(anchor);
2726 		if (rs == NULL) {
2727 			PF_RULES_RUNLOCK();
2728 			ERROUT(ENOENT);
2729 		}
2730 		if (ticket != rs->active.ticket) {
2731 			PF_RULES_RUNLOCK();
2732 			ERROUT(EBUSY);
2733 		}
2734 
2735 		nvlist_destroy(nvl);
2736 		nvl = NULL;
2737 		free(nvlpacked, M_NVLIST);
2738 		nvlpacked = NULL;
2739 
2740 		rule = TAILQ_FIRST(rs->active.rules);
2741 		while ((rule != NULL) && (rule->nr != nr))
2742 			rule = TAILQ_NEXT(rule, entries);
2743 		if (rule == NULL) {
2744 			PF_RULES_RUNLOCK();
2745 			ERROUT(ENOENT);
2746 		}
2747 		/* Make sure rule can't go away. */
2748 		NET_EPOCH_ENTER(et);
2749 		PF_RULES_RUNLOCK();
2750 		nvl = pf_keth_rule_to_nveth_rule(rule);
2751 		if (pf_keth_anchor_nvcopyout(rs, rule, nvl))
2752 			ERROUT(EBUSY);
2753 		NET_EPOCH_EXIT(et);
2754 		if (nvl == NULL)
2755 			ERROUT(ENOMEM);
2756 
2757 		nvlpacked = nvlist_pack(nvl, &nv->len);
2758 		if (nvlpacked == NULL)
2759 			ERROUT(ENOMEM);
2760 
2761 		if (nv->size == 0)
2762 			ERROUT(0);
2763 		else if (nv->size < nv->len)
2764 			ERROUT(ENOSPC);
2765 
2766 		error = copyout(nvlpacked, nv->data, nv->len);
2767 		if (error == 0 && clear) {
2768 			counter_u64_zero(rule->evaluations);
2769 			for (int i = 0; i < 2; i++) {
2770 				counter_u64_zero(rule->packets[i]);
2771 				counter_u64_zero(rule->bytes[i]);
2772 			}
2773 		}
2774 
2775 #undef ERROUT
2776 DIOCGETETHRULE_error:
2777 		free(nvlpacked, M_NVLIST);
2778 		nvlist_destroy(nvl);
2779 		break;
2780 	}
2781 
2782 	case DIOCADDETHRULE: {
2783 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2784 		nvlist_t		*nvl = NULL;
2785 		void			*nvlpacked = NULL;
2786 		struct pf_keth_rule	*rule = NULL, *tail = NULL;
2787 		struct pf_keth_ruleset	*ruleset = NULL;
2788 		struct pfi_kkif		*kif = NULL, *bridge_to_kif = NULL;
2789 		const char		*anchor = "", *anchor_call = "";
2790 
2791 #define ERROUT(x)	ERROUT_IOCTL(DIOCADDETHRULE_error, x)
2792 
2793 		if (nv->len > pf_ioctl_maxcount)
2794 			ERROUT(ENOMEM);
2795 
2796 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2797 		if (nvlpacked == NULL)
2798 			ERROUT(ENOMEM);
2799 
2800 		error = copyin(nv->data, nvlpacked, nv->len);
2801 		if (error)
2802 			ERROUT(error);
2803 
2804 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2805 		if (nvl == NULL)
2806 			ERROUT(EBADMSG);
2807 
2808 		if (! nvlist_exists_number(nvl, "ticket"))
2809 			ERROUT(EBADMSG);
2810 
2811 		if (nvlist_exists_string(nvl, "anchor"))
2812 			anchor = nvlist_get_string(nvl, "anchor");
2813 		if (nvlist_exists_string(nvl, "anchor_call"))
2814 			anchor_call = nvlist_get_string(nvl, "anchor_call");
2815 
2816 		ruleset = pf_find_keth_ruleset(anchor);
2817 		if (ruleset == NULL)
2818 			ERROUT(EINVAL);
2819 
2820 		if (nvlist_get_number(nvl, "ticket") !=
2821 		    ruleset->inactive.ticket) {
2822 			DPFPRINTF(PF_DEBUG_MISC,
2823 			    ("ticket: %d != %d\n",
2824 			    (u_int32_t)nvlist_get_number(nvl, "ticket"),
2825 			    ruleset->inactive.ticket));
2826 			ERROUT(EBUSY);
2827 		}
2828 
2829 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
2830 		if (rule == NULL)
2831 			ERROUT(ENOMEM);
2832 		rule->timestamp = NULL;
2833 
2834 		error = pf_nveth_rule_to_keth_rule(nvl, rule);
2835 		if (error != 0)
2836 			ERROUT(error);
2837 
2838 		if (rule->ifname[0])
2839 			kif = pf_kkif_create(M_WAITOK);
2840 		if (rule->bridge_to_name[0])
2841 			bridge_to_kif = pf_kkif_create(M_WAITOK);
2842 		rule->evaluations = counter_u64_alloc(M_WAITOK);
2843 		for (int i = 0; i < 2; i++) {
2844 			rule->packets[i] = counter_u64_alloc(M_WAITOK);
2845 			rule->bytes[i] = counter_u64_alloc(M_WAITOK);
2846 		}
2847 		rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
2848 		    M_WAITOK | M_ZERO);
2849 
2850 		PF_RULES_WLOCK();
2851 
2852 		if (rule->ifname[0]) {
2853 			rule->kif = pfi_kkif_attach(kif, rule->ifname);
2854 			pfi_kkif_ref(rule->kif);
2855 		} else
2856 			rule->kif = NULL;
2857 		if (rule->bridge_to_name[0]) {
2858 			rule->bridge_to = pfi_kkif_attach(bridge_to_kif,
2859 			    rule->bridge_to_name);
2860 			pfi_kkif_ref(rule->bridge_to);
2861 		} else
2862 			rule->bridge_to = NULL;
2863 
2864 #ifdef ALTQ
2865 		/* set queue IDs */
2866 		if (rule->qname[0] != 0) {
2867 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2868 				error = EBUSY;
2869 			else
2870 				rule->qid = rule->qid;
2871 		}
2872 #endif
2873 		if (rule->tagname[0])
2874 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2875 				error = EBUSY;
2876 		if (rule->match_tagname[0])
2877 			if ((rule->match_tag = pf_tagname2tag(
2878 			    rule->match_tagname)) == 0)
2879 				error = EBUSY;
2880 
2881 		if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE)
2882 			error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr);
2883 		if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE)
2884 			error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr);
2885 
2886 		if (error) {
2887 			pf_free_eth_rule(rule);
2888 			PF_RULES_WUNLOCK();
2889 			ERROUT(error);
2890 		}
2891 
2892 		if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) {
2893 			pf_free_eth_rule(rule);
2894 			PF_RULES_WUNLOCK();
2895 			ERROUT(EINVAL);
2896 		}
2897 
2898 		tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq);
2899 		if (tail)
2900 			rule->nr = tail->nr + 1;
2901 		else
2902 			rule->nr = 0;
2903 
2904 		TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries);
2905 
2906 		PF_RULES_WUNLOCK();
2907 
2908 #undef ERROUT
2909 DIOCADDETHRULE_error:
2910 		nvlist_destroy(nvl);
2911 		free(nvlpacked, M_NVLIST);
2912 		break;
2913 	}
2914 
2915 	case DIOCGETETHRULESETS: {
2916 		struct epoch_tracker	 et;
2917 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2918 		nvlist_t		*nvl = NULL;
2919 		void			*nvlpacked = NULL;
2920 		struct pf_keth_ruleset	*ruleset;
2921 		struct pf_keth_anchor	*anchor;
2922 		int			 nr = 0;
2923 
2924 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULESETS_error, x)
2925 
2926 		if (nv->len > pf_ioctl_maxcount)
2927 			ERROUT(ENOMEM);
2928 
2929 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2930 		if (nvlpacked == NULL)
2931 			ERROUT(ENOMEM);
2932 
2933 		error = copyin(nv->data, nvlpacked, nv->len);
2934 		if (error)
2935 			ERROUT(error);
2936 
2937 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2938 		if (nvl == NULL)
2939 			ERROUT(EBADMSG);
2940 		if (! nvlist_exists_string(nvl, "path"))
2941 			ERROUT(EBADMSG);
2942 
2943 		NET_EPOCH_ENTER(et);
2944 
2945 		if ((ruleset = pf_find_keth_ruleset(
2946 		    nvlist_get_string(nvl, "path"))) == NULL) {
2947 			NET_EPOCH_EXIT(et);
2948 			ERROUT(ENOENT);
2949 		}
2950 
2951 		if (ruleset->anchor == NULL) {
2952 			RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors)
2953 				if (anchor->parent == NULL)
2954 					nr++;
2955 		} else {
2956 			RB_FOREACH(anchor, pf_keth_anchor_node,
2957 			    &ruleset->anchor->children)
2958 				nr++;
2959 		}
2960 
2961 		NET_EPOCH_EXIT(et);
2962 
2963 		nvlist_destroy(nvl);
2964 		nvl = NULL;
2965 		free(nvlpacked, M_NVLIST);
2966 		nvlpacked = NULL;
2967 
2968 		nvl = nvlist_create(0);
2969 		if (nvl == NULL)
2970 			ERROUT(ENOMEM);
2971 
2972 		nvlist_add_number(nvl, "nr", nr);
2973 
2974 		nvlpacked = nvlist_pack(nvl, &nv->len);
2975 		if (nvlpacked == NULL)
2976 			ERROUT(ENOMEM);
2977 
2978 		if (nv->size == 0)
2979 			ERROUT(0);
2980 		else if (nv->size < nv->len)
2981 			ERROUT(ENOSPC);
2982 
2983 		error = copyout(nvlpacked, nv->data, nv->len);
2984 
2985 #undef ERROUT
2986 DIOCGETETHRULESETS_error:
2987 		free(nvlpacked, M_NVLIST);
2988 		nvlist_destroy(nvl);
2989 		break;
2990 	}
2991 
2992 	case DIOCGETETHRULESET: {
2993 		struct epoch_tracker	 et;
2994 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2995 		nvlist_t		*nvl = NULL;
2996 		void			*nvlpacked = NULL;
2997 		struct pf_keth_ruleset	*ruleset;
2998 		struct pf_keth_anchor	*anchor;
2999 		int			 nr = 0, req_nr = 0;
3000 		bool			 found = false;
3001 
3002 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULESET_error, x)
3003 
3004 		if (nv->len > pf_ioctl_maxcount)
3005 			ERROUT(ENOMEM);
3006 
3007 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3008 		if (nvlpacked == NULL)
3009 			ERROUT(ENOMEM);
3010 
3011 		error = copyin(nv->data, nvlpacked, nv->len);
3012 		if (error)
3013 			ERROUT(error);
3014 
3015 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3016 		if (nvl == NULL)
3017 			ERROUT(EBADMSG);
3018 		if (! nvlist_exists_string(nvl, "path"))
3019 			ERROUT(EBADMSG);
3020 		if (! nvlist_exists_number(nvl, "nr"))
3021 			ERROUT(EBADMSG);
3022 
3023 		req_nr = nvlist_get_number(nvl, "nr");
3024 
3025 		NET_EPOCH_ENTER(et);
3026 
3027 		if ((ruleset = pf_find_keth_ruleset(
3028 		    nvlist_get_string(nvl, "path"))) == NULL) {
3029 			NET_EPOCH_EXIT(et);
3030 			ERROUT(ENOENT);
3031 		}
3032 
3033 		nvlist_destroy(nvl);
3034 		nvl = NULL;
3035 		free(nvlpacked, M_NVLIST);
3036 		nvlpacked = NULL;
3037 
3038 		nvl = nvlist_create(0);
3039 		if (nvl == NULL) {
3040 			NET_EPOCH_EXIT(et);
3041 			ERROUT(ENOMEM);
3042 		}
3043 
3044 		if (ruleset->anchor == NULL) {
3045 			RB_FOREACH(anchor, pf_keth_anchor_global,
3046 			    &V_pf_keth_anchors) {
3047 				if (anchor->parent == NULL && nr++ == req_nr) {
3048 					found = true;
3049 					break;
3050 				}
3051 			}
3052 		} else {
3053 			RB_FOREACH(anchor, pf_keth_anchor_node,
3054 			     &ruleset->anchor->children) {
3055 				if (nr++ == req_nr) {
3056 					found = true;
3057 					break;
3058 				}
3059 			}
3060 		}
3061 
3062 		NET_EPOCH_EXIT(et);
3063 		if (found) {
3064 			nvlist_add_number(nvl, "nr", nr);
3065 			nvlist_add_string(nvl, "name", anchor->name);
3066 			if (ruleset->anchor)
3067 				nvlist_add_string(nvl, "path",
3068 				    ruleset->anchor->path);
3069 			else
3070 				nvlist_add_string(nvl, "path", "");
3071 		} else {
3072 			ERROUT(EBUSY);
3073 		}
3074 
3075 		nvlpacked = nvlist_pack(nvl, &nv->len);
3076 		if (nvlpacked == NULL)
3077 			ERROUT(ENOMEM);
3078 
3079 		if (nv->size == 0)
3080 			ERROUT(0);
3081 		else if (nv->size < nv->len)
3082 			ERROUT(ENOSPC);
3083 
3084 		error = copyout(nvlpacked, nv->data, nv->len);
3085 
3086 #undef ERROUT
3087 DIOCGETETHRULESET_error:
3088 		free(nvlpacked, M_NVLIST);
3089 		nvlist_destroy(nvl);
3090 		break;
3091 	}
3092 
3093 	case DIOCADDRULENV: {
3094 		struct pfioc_nv	*nv = (struct pfioc_nv *)addr;
3095 		nvlist_t	*nvl = NULL;
3096 		void		*nvlpacked = NULL;
3097 		struct pf_krule	*rule = NULL;
3098 		const char	*anchor = "", *anchor_call = "";
3099 		uint32_t	 ticket = 0, pool_ticket = 0;
3100 
3101 #define	ERROUT(x)	ERROUT_IOCTL(DIOCADDRULENV_error, x)
3102 
3103 		if (nv->len > pf_ioctl_maxcount)
3104 			ERROUT(ENOMEM);
3105 
3106 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3107 		error = copyin(nv->data, nvlpacked, nv->len);
3108 		if (error)
3109 			ERROUT(error);
3110 
3111 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3112 		if (nvl == NULL)
3113 			ERROUT(EBADMSG);
3114 
3115 		if (! nvlist_exists_number(nvl, "ticket"))
3116 			ERROUT(EINVAL);
3117 		ticket = nvlist_get_number(nvl, "ticket");
3118 
3119 		if (! nvlist_exists_number(nvl, "pool_ticket"))
3120 			ERROUT(EINVAL);
3121 		pool_ticket = nvlist_get_number(nvl, "pool_ticket");
3122 
3123 		if (! nvlist_exists_nvlist(nvl, "rule"))
3124 			ERROUT(EINVAL);
3125 
3126 		rule = pf_krule_alloc();
3127 		error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"),
3128 		    rule);
3129 		if (error)
3130 			ERROUT(error);
3131 
3132 		if (nvlist_exists_string(nvl, "anchor"))
3133 			anchor = nvlist_get_string(nvl, "anchor");
3134 		if (nvlist_exists_string(nvl, "anchor_call"))
3135 			anchor_call = nvlist_get_string(nvl, "anchor_call");
3136 
3137 		if ((error = nvlist_error(nvl)))
3138 			ERROUT(error);
3139 
3140 		/* Frees rule on error */
3141 		error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor,
3142 		    anchor_call, td->td_ucred->cr_ruid,
3143 		    td->td_proc ? td->td_proc->p_pid : 0);
3144 
3145 		nvlist_destroy(nvl);
3146 		free(nvlpacked, M_NVLIST);
3147 		break;
3148 #undef ERROUT
3149 DIOCADDRULENV_error:
3150 		pf_krule_free(rule);
3151 		nvlist_destroy(nvl);
3152 		free(nvlpacked, M_NVLIST);
3153 
3154 		break;
3155 	}
3156 	case DIOCADDRULE: {
3157 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
3158 		struct pf_krule		*rule;
3159 
3160 		rule = pf_krule_alloc();
3161 		error = pf_rule_to_krule(&pr->rule, rule);
3162 		if (error != 0) {
3163 			pf_krule_free(rule);
3164 			break;
3165 		}
3166 
3167 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
3168 
3169 		/* Frees rule on error */
3170 		error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket,
3171 		    pr->anchor, pr->anchor_call, td->td_ucred->cr_ruid,
3172 		    td->td_proc ? td->td_proc->p_pid : 0);
3173 		break;
3174 	}
3175 
3176 	case DIOCGETRULES: {
3177 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
3178 
3179 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
3180 
3181 		error = pf_ioctl_getrules(pr);
3182 
3183 		break;
3184 	}
3185 
3186 	case DIOCGETRULENV: {
3187 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
3188 		nvlist_t		*nvrule = NULL;
3189 		nvlist_t		*nvl = NULL;
3190 		struct pf_kruleset	*ruleset;
3191 		struct pf_krule		*rule;
3192 		void			*nvlpacked = NULL;
3193 		int			 rs_num, nr;
3194 		bool			 clear_counter = false;
3195 
3196 #define	ERROUT(x)	ERROUT_IOCTL(DIOCGETRULENV_error, x)
3197 
3198 		if (nv->len > pf_ioctl_maxcount)
3199 			ERROUT(ENOMEM);
3200 
3201 		/* Copy the request in */
3202 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3203 		if (nvlpacked == NULL)
3204 			ERROUT(ENOMEM);
3205 
3206 		error = copyin(nv->data, nvlpacked, nv->len);
3207 		if (error)
3208 			ERROUT(error);
3209 
3210 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3211 		if (nvl == NULL)
3212 			ERROUT(EBADMSG);
3213 
3214 		if (! nvlist_exists_string(nvl, "anchor"))
3215 			ERROUT(EBADMSG);
3216 		if (! nvlist_exists_number(nvl, "ruleset"))
3217 			ERROUT(EBADMSG);
3218 		if (! nvlist_exists_number(nvl, "ticket"))
3219 			ERROUT(EBADMSG);
3220 		if (! nvlist_exists_number(nvl, "nr"))
3221 			ERROUT(EBADMSG);
3222 
3223 		if (nvlist_exists_bool(nvl, "clear_counter"))
3224 			clear_counter = nvlist_get_bool(nvl, "clear_counter");
3225 
3226 		if (clear_counter && !(flags & FWRITE))
3227 			ERROUT(EACCES);
3228 
3229 		nr = nvlist_get_number(nvl, "nr");
3230 
3231 		PF_RULES_WLOCK();
3232 		ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor"));
3233 		if (ruleset == NULL) {
3234 			PF_RULES_WUNLOCK();
3235 			ERROUT(ENOENT);
3236 		}
3237 
3238 		rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset"));
3239 		if (rs_num >= PF_RULESET_MAX) {
3240 			PF_RULES_WUNLOCK();
3241 			ERROUT(EINVAL);
3242 		}
3243 
3244 		if (nvlist_get_number(nvl, "ticket") !=
3245 		    ruleset->rules[rs_num].active.ticket) {
3246 			PF_RULES_WUNLOCK();
3247 			ERROUT(EBUSY);
3248 		}
3249 
3250 		if ((error = nvlist_error(nvl))) {
3251 			PF_RULES_WUNLOCK();
3252 			ERROUT(error);
3253 		}
3254 
3255 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
3256 		while ((rule != NULL) && (rule->nr != nr))
3257 			rule = TAILQ_NEXT(rule, entries);
3258 		if (rule == NULL) {
3259 			PF_RULES_WUNLOCK();
3260 			ERROUT(EBUSY);
3261 		}
3262 
3263 		nvrule = pf_krule_to_nvrule(rule);
3264 
3265 		nvlist_destroy(nvl);
3266 		nvl = nvlist_create(0);
3267 		if (nvl == NULL) {
3268 			PF_RULES_WUNLOCK();
3269 			ERROUT(ENOMEM);
3270 		}
3271 		nvlist_add_number(nvl, "nr", nr);
3272 		nvlist_add_nvlist(nvl, "rule", nvrule);
3273 		nvlist_destroy(nvrule);
3274 		nvrule = NULL;
3275 		if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) {
3276 			PF_RULES_WUNLOCK();
3277 			ERROUT(EBUSY);
3278 		}
3279 
3280 		free(nvlpacked, M_NVLIST);
3281 		nvlpacked = nvlist_pack(nvl, &nv->len);
3282 		if (nvlpacked == NULL) {
3283 			PF_RULES_WUNLOCK();
3284 			ERROUT(ENOMEM);
3285 		}
3286 
3287 		if (nv->size == 0) {
3288 			PF_RULES_WUNLOCK();
3289 			ERROUT(0);
3290 		}
3291 		else if (nv->size < nv->len) {
3292 			PF_RULES_WUNLOCK();
3293 			ERROUT(ENOSPC);
3294 		}
3295 
3296 		if (clear_counter)
3297 			pf_krule_clear_counters(rule);
3298 
3299 		PF_RULES_WUNLOCK();
3300 
3301 		error = copyout(nvlpacked, nv->data, nv->len);
3302 
3303 #undef ERROUT
3304 DIOCGETRULENV_error:
3305 		free(nvlpacked, M_NVLIST);
3306 		nvlist_destroy(nvrule);
3307 		nvlist_destroy(nvl);
3308 
3309 		break;
3310 	}
3311 
3312 	case DIOCCHANGERULE: {
3313 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
3314 		struct pf_kruleset	*ruleset;
3315 		struct pf_krule		*oldrule = NULL, *newrule = NULL;
3316 		struct pfi_kkif		*kif = NULL;
3317 		struct pf_kpooladdr	*pa;
3318 		u_int32_t		 nr = 0;
3319 		int			 rs_num;
3320 
3321 		pcr->anchor[sizeof(pcr->anchor) - 1] = 0;
3322 
3323 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
3324 		    pcr->action > PF_CHANGE_GET_TICKET) {
3325 			error = EINVAL;
3326 			break;
3327 		}
3328 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
3329 			error = EINVAL;
3330 			break;
3331 		}
3332 
3333 		if (pcr->action != PF_CHANGE_REMOVE) {
3334 			newrule = pf_krule_alloc();
3335 			error = pf_rule_to_krule(&pcr->rule, newrule);
3336 			if (error != 0) {
3337 				pf_krule_free(newrule);
3338 				break;
3339 			}
3340 
3341 			if (newrule->ifname[0])
3342 				kif = pf_kkif_create(M_WAITOK);
3343 			pf_counter_u64_init(&newrule->evaluations, M_WAITOK);
3344 			for (int i = 0; i < 2; i++) {
3345 				pf_counter_u64_init(&newrule->packets[i], M_WAITOK);
3346 				pf_counter_u64_init(&newrule->bytes[i], M_WAITOK);
3347 			}
3348 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
3349 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
3350 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
3351 			newrule->cuid = td->td_ucred->cr_ruid;
3352 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
3353 			TAILQ_INIT(&newrule->rpool.list);
3354 		}
3355 #define	ERROUT(x)	ERROUT_IOCTL(DIOCCHANGERULE_error, x)
3356 
3357 		PF_CONFIG_LOCK();
3358 		PF_RULES_WLOCK();
3359 #ifdef PF_WANT_32_TO_64_COUNTER
3360 		if (newrule != NULL) {
3361 			LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist);
3362 			newrule->allrulelinked = true;
3363 			V_pf_allrulecount++;
3364 		}
3365 #endif
3366 
3367 		if (!(pcr->action == PF_CHANGE_REMOVE ||
3368 		    pcr->action == PF_CHANGE_GET_TICKET) &&
3369 		    pcr->pool_ticket != V_ticket_pabuf)
3370 			ERROUT(EBUSY);
3371 
3372 		ruleset = pf_find_kruleset(pcr->anchor);
3373 		if (ruleset == NULL)
3374 			ERROUT(EINVAL);
3375 
3376 		rs_num = pf_get_ruleset_number(pcr->rule.action);
3377 		if (rs_num >= PF_RULESET_MAX)
3378 			ERROUT(EINVAL);
3379 
3380 		/*
3381 		 * XXXMJG: there is no guarantee that the ruleset was
3382 		 * created by the usual route of calling DIOCXBEGIN.
3383 		 * As a result it is possible the rule tree will not
3384 		 * be allocated yet. Hack around it by doing it here.
3385 		 * Note it is fine to let the tree persist in case of
3386 		 * error as it will be freed down the road on future
3387 		 * updates (if need be).
3388 		 */
3389 		if (ruleset->rules[rs_num].active.tree == NULL) {
3390 			ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT);
3391 			if (ruleset->rules[rs_num].active.tree == NULL) {
3392 				ERROUT(ENOMEM);
3393 			}
3394 		}
3395 
3396 		if (pcr->action == PF_CHANGE_GET_TICKET) {
3397 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
3398 			ERROUT(0);
3399 		} else if (pcr->ticket !=
3400 			    ruleset->rules[rs_num].active.ticket)
3401 				ERROUT(EINVAL);
3402 
3403 		if (pcr->action != PF_CHANGE_REMOVE) {
3404 			if (newrule->ifname[0]) {
3405 				newrule->kif = pfi_kkif_attach(kif,
3406 				    newrule->ifname);
3407 				kif = NULL;
3408 				pfi_kkif_ref(newrule->kif);
3409 			} else
3410 				newrule->kif = NULL;
3411 
3412 			if (newrule->rtableid > 0 &&
3413 			    newrule->rtableid >= rt_numfibs)
3414 				error = EBUSY;
3415 
3416 #ifdef ALTQ
3417 			/* set queue IDs */
3418 			if (newrule->qname[0] != 0) {
3419 				if ((newrule->qid =
3420 				    pf_qname2qid(newrule->qname)) == 0)
3421 					error = EBUSY;
3422 				else if (newrule->pqname[0] != 0) {
3423 					if ((newrule->pqid =
3424 					    pf_qname2qid(newrule->pqname)) == 0)
3425 						error = EBUSY;
3426 				} else
3427 					newrule->pqid = newrule->qid;
3428 			}
3429 #endif /* ALTQ */
3430 			if (newrule->tagname[0])
3431 				if ((newrule->tag =
3432 				    pf_tagname2tag(newrule->tagname)) == 0)
3433 					error = EBUSY;
3434 			if (newrule->match_tagname[0])
3435 				if ((newrule->match_tag = pf_tagname2tag(
3436 				    newrule->match_tagname)) == 0)
3437 					error = EBUSY;
3438 			if (newrule->rt && !newrule->direction)
3439 				error = EINVAL;
3440 			if (!newrule->log)
3441 				newrule->logif = 0;
3442 			if (newrule->logif >= PFLOGIFS_MAX)
3443 				error = EINVAL;
3444 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
3445 				error = ENOMEM;
3446 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
3447 				error = ENOMEM;
3448 			if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
3449 				error = EINVAL;
3450 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
3451 				if (pa->addr.type == PF_ADDR_TABLE) {
3452 					pa->addr.p.tbl =
3453 					    pfr_attach_table(ruleset,
3454 					    pa->addr.v.tblname);
3455 					if (pa->addr.p.tbl == NULL)
3456 						error = ENOMEM;
3457 				}
3458 
3459 			newrule->overload_tbl = NULL;
3460 			if (newrule->overload_tblname[0]) {
3461 				if ((newrule->overload_tbl = pfr_attach_table(
3462 				    ruleset, newrule->overload_tblname)) ==
3463 				    NULL)
3464 					error = EINVAL;
3465 				else
3466 					newrule->overload_tbl->pfrkt_flags |=
3467 					    PFR_TFLAG_ACTIVE;
3468 			}
3469 
3470 			pf_mv_kpool(&V_pf_pabuf, &newrule->rpool.list);
3471 			if (((((newrule->action == PF_NAT) ||
3472 			    (newrule->action == PF_RDR) ||
3473 			    (newrule->action == PF_BINAT) ||
3474 			    (newrule->rt > PF_NOPFROUTE)) &&
3475 			    !newrule->anchor)) &&
3476 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
3477 				error = EINVAL;
3478 
3479 			if (error) {
3480 				pf_free_rule(newrule);
3481 				PF_RULES_WUNLOCK();
3482 				PF_CONFIG_UNLOCK();
3483 				break;
3484 			}
3485 
3486 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
3487 		}
3488 		pf_empty_kpool(&V_pf_pabuf);
3489 
3490 		if (pcr->action == PF_CHANGE_ADD_HEAD)
3491 			oldrule = TAILQ_FIRST(
3492 			    ruleset->rules[rs_num].active.ptr);
3493 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
3494 			oldrule = TAILQ_LAST(
3495 			    ruleset->rules[rs_num].active.ptr, pf_krulequeue);
3496 		else {
3497 			oldrule = TAILQ_FIRST(
3498 			    ruleset->rules[rs_num].active.ptr);
3499 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
3500 				oldrule = TAILQ_NEXT(oldrule, entries);
3501 			if (oldrule == NULL) {
3502 				if (newrule != NULL)
3503 					pf_free_rule(newrule);
3504 				PF_RULES_WUNLOCK();
3505 				PF_CONFIG_UNLOCK();
3506 				error = EINVAL;
3507 				break;
3508 			}
3509 		}
3510 
3511 		if (pcr->action == PF_CHANGE_REMOVE) {
3512 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
3513 			    oldrule);
3514 			RB_REMOVE(pf_krule_global,
3515 			    ruleset->rules[rs_num].active.tree, oldrule);
3516 			ruleset->rules[rs_num].active.rcount--;
3517 		} else {
3518 			pf_hash_rule(newrule);
3519 			if (RB_INSERT(pf_krule_global,
3520 			    ruleset->rules[rs_num].active.tree, newrule) != NULL) {
3521 				pf_free_rule(newrule);
3522 				PF_RULES_WUNLOCK();
3523 				PF_CONFIG_UNLOCK();
3524 				error = EEXIST;
3525 				break;
3526 			}
3527 
3528 			if (oldrule == NULL)
3529 				TAILQ_INSERT_TAIL(
3530 				    ruleset->rules[rs_num].active.ptr,
3531 				    newrule, entries);
3532 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
3533 			    pcr->action == PF_CHANGE_ADD_BEFORE)
3534 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
3535 			else
3536 				TAILQ_INSERT_AFTER(
3537 				    ruleset->rules[rs_num].active.ptr,
3538 				    oldrule, newrule, entries);
3539 			ruleset->rules[rs_num].active.rcount++;
3540 		}
3541 
3542 		nr = 0;
3543 		TAILQ_FOREACH(oldrule,
3544 		    ruleset->rules[rs_num].active.ptr, entries)
3545 			oldrule->nr = nr++;
3546 
3547 		ruleset->rules[rs_num].active.ticket++;
3548 
3549 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
3550 		pf_remove_if_empty_kruleset(ruleset);
3551 
3552 		PF_RULES_WUNLOCK();
3553 		PF_CONFIG_UNLOCK();
3554 		break;
3555 
3556 #undef ERROUT
3557 DIOCCHANGERULE_error:
3558 		PF_RULES_WUNLOCK();
3559 		PF_CONFIG_UNLOCK();
3560 		pf_krule_free(newrule);
3561 		pf_kkif_free(kif);
3562 		break;
3563 	}
3564 
3565 	case DIOCCLRSTATESNV: {
3566 		error = pf_clearstates_nv((struct pfioc_nv *)addr);
3567 		break;
3568 	}
3569 
3570 	case DIOCKILLSTATESNV: {
3571 		error = pf_killstates_nv((struct pfioc_nv *)addr);
3572 		break;
3573 	}
3574 
3575 	case DIOCADDSTATE: {
3576 		struct pfioc_state		*ps = (struct pfioc_state *)addr;
3577 		struct pfsync_state_1301	*sp = &ps->state;
3578 
3579 		if (sp->timeout >= PFTM_MAX) {
3580 			error = EINVAL;
3581 			break;
3582 		}
3583 		if (V_pfsync_state_import_ptr != NULL) {
3584 			PF_RULES_RLOCK();
3585 			error = V_pfsync_state_import_ptr(
3586 			    (union pfsync_state_union *)sp, PFSYNC_SI_IOCTL,
3587 			    PFSYNC_MSG_VERSION_1301);
3588 			PF_RULES_RUNLOCK();
3589 		} else
3590 			error = EOPNOTSUPP;
3591 		break;
3592 	}
3593 
3594 	case DIOCGETSTATE: {
3595 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
3596 		struct pf_kstate	*s;
3597 
3598 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
3599 		if (s == NULL) {
3600 			error = ENOENT;
3601 			break;
3602 		}
3603 
3604 		pfsync_state_export((union pfsync_state_union*)&ps->state,
3605 		    s, PFSYNC_MSG_VERSION_1301);
3606 		PF_STATE_UNLOCK(s);
3607 		break;
3608 	}
3609 
3610 	case DIOCGETSTATENV: {
3611 		error = pf_getstate((struct pfioc_nv *)addr);
3612 		break;
3613 	}
3614 
3615 #ifdef COMPAT_FREEBSD14
3616 	case DIOCGETSTATES: {
3617 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
3618 		struct pf_kstate	*s;
3619 		struct pfsync_state_1301	*pstore, *p;
3620 		int			 i, nr;
3621 		size_t			 slice_count = 16, count;
3622 		void			*out;
3623 
3624 		if (ps->ps_len <= 0) {
3625 			nr = uma_zone_get_cur(V_pf_state_z);
3626 			ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3627 			break;
3628 		}
3629 
3630 		out = ps->ps_states;
3631 		pstore = mallocarray(slice_count,
3632 		    sizeof(struct pfsync_state_1301), M_TEMP, M_WAITOK | M_ZERO);
3633 		nr = 0;
3634 
3635 		for (i = 0; i <= pf_hashmask; i++) {
3636 			struct pf_idhash *ih = &V_pf_idhash[i];
3637 
3638 DIOCGETSTATES_retry:
3639 			p = pstore;
3640 
3641 			if (LIST_EMPTY(&ih->states))
3642 				continue;
3643 
3644 			PF_HASHROW_LOCK(ih);
3645 			count = 0;
3646 			LIST_FOREACH(s, &ih->states, entry) {
3647 				if (s->timeout == PFTM_UNLINKED)
3648 					continue;
3649 				count++;
3650 			}
3651 
3652 			if (count > slice_count) {
3653 				PF_HASHROW_UNLOCK(ih);
3654 				free(pstore, M_TEMP);
3655 				slice_count = count * 2;
3656 				pstore = mallocarray(slice_count,
3657 				    sizeof(struct pfsync_state_1301), M_TEMP,
3658 				    M_WAITOK | M_ZERO);
3659 				goto DIOCGETSTATES_retry;
3660 			}
3661 
3662 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
3663 				PF_HASHROW_UNLOCK(ih);
3664 				goto DIOCGETSTATES_full;
3665 			}
3666 
3667 			LIST_FOREACH(s, &ih->states, entry) {
3668 				if (s->timeout == PFTM_UNLINKED)
3669 					continue;
3670 
3671 				pfsync_state_export((union pfsync_state_union*)p,
3672 				    s, PFSYNC_MSG_VERSION_1301);
3673 				p++;
3674 				nr++;
3675 			}
3676 			PF_HASHROW_UNLOCK(ih);
3677 			error = copyout(pstore, out,
3678 			    sizeof(struct pfsync_state_1301) * count);
3679 			if (error)
3680 				break;
3681 			out = ps->ps_states + nr;
3682 		}
3683 DIOCGETSTATES_full:
3684 		ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3685 		free(pstore, M_TEMP);
3686 
3687 		break;
3688 	}
3689 
3690 	case DIOCGETSTATESV2: {
3691 		struct pfioc_states_v2	*ps = (struct pfioc_states_v2 *)addr;
3692 		struct pf_kstate	*s;
3693 		struct pf_state_export	*pstore, *p;
3694 		int i, nr;
3695 		size_t slice_count = 16, count;
3696 		void *out;
3697 
3698 		if (ps->ps_req_version > PF_STATE_VERSION) {
3699 			error = ENOTSUP;
3700 			break;
3701 		}
3702 
3703 		if (ps->ps_len <= 0) {
3704 			nr = uma_zone_get_cur(V_pf_state_z);
3705 			ps->ps_len = sizeof(struct pf_state_export) * nr;
3706 			break;
3707 		}
3708 
3709 		out = ps->ps_states;
3710 		pstore = mallocarray(slice_count,
3711 		    sizeof(struct pf_state_export), M_TEMP, M_WAITOK | M_ZERO);
3712 		nr = 0;
3713 
3714 		for (i = 0; i <= pf_hashmask; i++) {
3715 			struct pf_idhash *ih = &V_pf_idhash[i];
3716 
3717 DIOCGETSTATESV2_retry:
3718 			p = pstore;
3719 
3720 			if (LIST_EMPTY(&ih->states))
3721 				continue;
3722 
3723 			PF_HASHROW_LOCK(ih);
3724 			count = 0;
3725 			LIST_FOREACH(s, &ih->states, entry) {
3726 				if (s->timeout == PFTM_UNLINKED)
3727 					continue;
3728 				count++;
3729 			}
3730 
3731 			if (count > slice_count) {
3732 				PF_HASHROW_UNLOCK(ih);
3733 				free(pstore, M_TEMP);
3734 				slice_count = count * 2;
3735 				pstore = mallocarray(slice_count,
3736 				    sizeof(struct pf_state_export), M_TEMP,
3737 				    M_WAITOK | M_ZERO);
3738 				goto DIOCGETSTATESV2_retry;
3739 			}
3740 
3741 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
3742 				PF_HASHROW_UNLOCK(ih);
3743 				goto DIOCGETSTATESV2_full;
3744 			}
3745 
3746 			LIST_FOREACH(s, &ih->states, entry) {
3747 				if (s->timeout == PFTM_UNLINKED)
3748 					continue;
3749 
3750 				pf_state_export(p, s);
3751 				p++;
3752 				nr++;
3753 			}
3754 			PF_HASHROW_UNLOCK(ih);
3755 			error = copyout(pstore, out,
3756 			    sizeof(struct pf_state_export) * count);
3757 			if (error)
3758 				break;
3759 			out = ps->ps_states + nr;
3760 		}
3761 DIOCGETSTATESV2_full:
3762 		ps->ps_len = nr * sizeof(struct pf_state_export);
3763 		free(pstore, M_TEMP);
3764 
3765 		break;
3766 	}
3767 #endif
3768 	case DIOCGETSTATUSNV: {
3769 		error = pf_getstatus((struct pfioc_nv *)addr);
3770 		break;
3771 	}
3772 
3773 	case DIOCSETSTATUSIF: {
3774 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
3775 
3776 		if (pi->ifname[0] == 0) {
3777 			bzero(V_pf_status.ifname, IFNAMSIZ);
3778 			break;
3779 		}
3780 		PF_RULES_WLOCK();
3781 		error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
3782 		PF_RULES_WUNLOCK();
3783 		break;
3784 	}
3785 
3786 	case DIOCCLRSTATUS: {
3787 		pf_ioctl_clear_status();
3788 		break;
3789 	}
3790 
3791 	case DIOCNATLOOK: {
3792 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
3793 		struct pf_state_key	*sk;
3794 		struct pf_kstate	*state;
3795 		struct pf_state_key_cmp	 key;
3796 		int			 m = 0, direction = pnl->direction;
3797 		int			 sidx, didx;
3798 
3799 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
3800 		sidx = (direction == PF_IN) ? 1 : 0;
3801 		didx = (direction == PF_IN) ? 0 : 1;
3802 
3803 		if (!pnl->proto ||
3804 		    PF_AZERO(&pnl->saddr, pnl->af) ||
3805 		    PF_AZERO(&pnl->daddr, pnl->af) ||
3806 		    ((pnl->proto == IPPROTO_TCP ||
3807 		    pnl->proto == IPPROTO_UDP) &&
3808 		    (!pnl->dport || !pnl->sport)))
3809 			error = EINVAL;
3810 		else {
3811 			bzero(&key, sizeof(key));
3812 			key.af = pnl->af;
3813 			key.proto = pnl->proto;
3814 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
3815 			key.port[sidx] = pnl->sport;
3816 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
3817 			key.port[didx] = pnl->dport;
3818 
3819 			state = pf_find_state_all(&key, direction, &m);
3820 			if (state == NULL) {
3821 				error = ENOENT;
3822 			} else {
3823 				if (m > 1) {
3824 					PF_STATE_UNLOCK(state);
3825 					error = E2BIG;	/* more than one state */
3826 				} else {
3827 					sk = state->key[sidx];
3828 					PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
3829 					pnl->rsport = sk->port[sidx];
3830 					PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
3831 					pnl->rdport = sk->port[didx];
3832 					PF_STATE_UNLOCK(state);
3833 				}
3834 			}
3835 		}
3836 		break;
3837 	}
3838 
3839 	case DIOCSETTIMEOUT: {
3840 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
3841 		int		 old;
3842 
3843 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
3844 		    pt->seconds < 0) {
3845 			error = EINVAL;
3846 			break;
3847 		}
3848 		PF_RULES_WLOCK();
3849 		old = V_pf_default_rule.timeout[pt->timeout];
3850 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
3851 			pt->seconds = 1;
3852 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
3853 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
3854 			wakeup(pf_purge_thread);
3855 		pt->seconds = old;
3856 		PF_RULES_WUNLOCK();
3857 		break;
3858 	}
3859 
3860 	case DIOCGETTIMEOUT: {
3861 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
3862 
3863 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
3864 			error = EINVAL;
3865 			break;
3866 		}
3867 		PF_RULES_RLOCK();
3868 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
3869 		PF_RULES_RUNLOCK();
3870 		break;
3871 	}
3872 
3873 	case DIOCGETLIMIT: {
3874 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
3875 
3876 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
3877 			error = EINVAL;
3878 			break;
3879 		}
3880 		PF_RULES_RLOCK();
3881 		pl->limit = V_pf_limits[pl->index].limit;
3882 		PF_RULES_RUNLOCK();
3883 		break;
3884 	}
3885 
3886 	case DIOCSETLIMIT: {
3887 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
3888 		int			 old_limit;
3889 
3890 		PF_RULES_WLOCK();
3891 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
3892 		    V_pf_limits[pl->index].zone == NULL) {
3893 			PF_RULES_WUNLOCK();
3894 			error = EINVAL;
3895 			break;
3896 		}
3897 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
3898 		old_limit = V_pf_limits[pl->index].limit;
3899 		V_pf_limits[pl->index].limit = pl->limit;
3900 		pl->limit = old_limit;
3901 		PF_RULES_WUNLOCK();
3902 		break;
3903 	}
3904 
3905 	case DIOCSETDEBUG: {
3906 		u_int32_t	*level = (u_int32_t *)addr;
3907 
3908 		PF_RULES_WLOCK();
3909 		V_pf_status.debug = *level;
3910 		PF_RULES_WUNLOCK();
3911 		break;
3912 	}
3913 
3914 	case DIOCCLRRULECTRS: {
3915 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
3916 		struct pf_kruleset	*ruleset = &pf_main_ruleset;
3917 		struct pf_krule		*rule;
3918 
3919 		PF_RULES_WLOCK();
3920 		TAILQ_FOREACH(rule,
3921 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
3922 			pf_counter_u64_zero(&rule->evaluations);
3923 			for (int i = 0; i < 2; i++) {
3924 				pf_counter_u64_zero(&rule->packets[i]);
3925 				pf_counter_u64_zero(&rule->bytes[i]);
3926 			}
3927 		}
3928 		PF_RULES_WUNLOCK();
3929 		break;
3930 	}
3931 
3932 	case DIOCGIFSPEEDV0:
3933 	case DIOCGIFSPEEDV1: {
3934 		struct pf_ifspeed_v1	*psp = (struct pf_ifspeed_v1 *)addr;
3935 		struct pf_ifspeed_v1	ps;
3936 		struct ifnet		*ifp;
3937 
3938 		if (psp->ifname[0] == '\0') {
3939 			error = EINVAL;
3940 			break;
3941 		}
3942 
3943 		error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ);
3944 		if (error != 0)
3945 			break;
3946 		ifp = ifunit(ps.ifname);
3947 		if (ifp != NULL) {
3948 			psp->baudrate32 =
3949 			    (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
3950 			if (cmd == DIOCGIFSPEEDV1)
3951 				psp->baudrate = ifp->if_baudrate;
3952 		} else {
3953 			error = EINVAL;
3954 		}
3955 		break;
3956 	}
3957 
3958 #ifdef ALTQ
3959 	case DIOCSTARTALTQ: {
3960 		struct pf_altq		*altq;
3961 
3962 		PF_RULES_WLOCK();
3963 		/* enable all altq interfaces on active list */
3964 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
3965 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
3966 				error = pf_enable_altq(altq);
3967 				if (error != 0)
3968 					break;
3969 			}
3970 		}
3971 		if (error == 0)
3972 			V_pf_altq_running = 1;
3973 		PF_RULES_WUNLOCK();
3974 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
3975 		break;
3976 	}
3977 
3978 	case DIOCSTOPALTQ: {
3979 		struct pf_altq		*altq;
3980 
3981 		PF_RULES_WLOCK();
3982 		/* disable all altq interfaces on active list */
3983 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
3984 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
3985 				error = pf_disable_altq(altq);
3986 				if (error != 0)
3987 					break;
3988 			}
3989 		}
3990 		if (error == 0)
3991 			V_pf_altq_running = 0;
3992 		PF_RULES_WUNLOCK();
3993 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
3994 		break;
3995 	}
3996 
3997 	case DIOCADDALTQV0:
3998 	case DIOCADDALTQV1: {
3999 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4000 		struct pf_altq		*altq, *a;
4001 		struct ifnet		*ifp;
4002 
4003 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
4004 		error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
4005 		if (error)
4006 			break;
4007 		altq->local_flags = 0;
4008 
4009 		PF_RULES_WLOCK();
4010 		if (pa->ticket != V_ticket_altqs_inactive) {
4011 			PF_RULES_WUNLOCK();
4012 			free(altq, M_PFALTQ);
4013 			error = EBUSY;
4014 			break;
4015 		}
4016 
4017 		/*
4018 		 * if this is for a queue, find the discipline and
4019 		 * copy the necessary fields
4020 		 */
4021 		if (altq->qname[0] != 0) {
4022 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
4023 				PF_RULES_WUNLOCK();
4024 				error = EBUSY;
4025 				free(altq, M_PFALTQ);
4026 				break;
4027 			}
4028 			altq->altq_disc = NULL;
4029 			TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
4030 				if (strncmp(a->ifname, altq->ifname,
4031 				    IFNAMSIZ) == 0) {
4032 					altq->altq_disc = a->altq_disc;
4033 					break;
4034 				}
4035 			}
4036 		}
4037 
4038 		if ((ifp = ifunit(altq->ifname)) == NULL)
4039 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
4040 		else
4041 			error = altq_add(ifp, altq);
4042 
4043 		if (error) {
4044 			PF_RULES_WUNLOCK();
4045 			free(altq, M_PFALTQ);
4046 			break;
4047 		}
4048 
4049 		if (altq->qname[0] != 0)
4050 			TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
4051 		else
4052 			TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
4053 		/* version error check done on import above */
4054 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4055 		PF_RULES_WUNLOCK();
4056 		break;
4057 	}
4058 
4059 	case DIOCGETALTQSV0:
4060 	case DIOCGETALTQSV1: {
4061 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4062 		struct pf_altq		*altq;
4063 
4064 		PF_RULES_RLOCK();
4065 		pa->nr = 0;
4066 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
4067 			pa->nr++;
4068 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
4069 			pa->nr++;
4070 		pa->ticket = V_ticket_altqs_active;
4071 		PF_RULES_RUNLOCK();
4072 		break;
4073 	}
4074 
4075 	case DIOCGETALTQV0:
4076 	case DIOCGETALTQV1: {
4077 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4078 		struct pf_altq		*altq;
4079 
4080 		PF_RULES_RLOCK();
4081 		if (pa->ticket != V_ticket_altqs_active) {
4082 			PF_RULES_RUNLOCK();
4083 			error = EBUSY;
4084 			break;
4085 		}
4086 		altq = pf_altq_get_nth_active(pa->nr);
4087 		if (altq == NULL) {
4088 			PF_RULES_RUNLOCK();
4089 			error = EBUSY;
4090 			break;
4091 		}
4092 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4093 		PF_RULES_RUNLOCK();
4094 		break;
4095 	}
4096 
4097 	case DIOCCHANGEALTQV0:
4098 	case DIOCCHANGEALTQV1:
4099 		/* CHANGEALTQ not supported yet! */
4100 		error = ENODEV;
4101 		break;
4102 
4103 	case DIOCGETQSTATSV0:
4104 	case DIOCGETQSTATSV1: {
4105 		struct pfioc_qstats_v1	*pq = (struct pfioc_qstats_v1 *)addr;
4106 		struct pf_altq		*altq;
4107 		int			 nbytes;
4108 		u_int32_t		 version;
4109 
4110 		PF_RULES_RLOCK();
4111 		if (pq->ticket != V_ticket_altqs_active) {
4112 			PF_RULES_RUNLOCK();
4113 			error = EBUSY;
4114 			break;
4115 		}
4116 		nbytes = pq->nbytes;
4117 		altq = pf_altq_get_nth_active(pq->nr);
4118 		if (altq == NULL) {
4119 			PF_RULES_RUNLOCK();
4120 			error = EBUSY;
4121 			break;
4122 		}
4123 
4124 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
4125 			PF_RULES_RUNLOCK();
4126 			error = ENXIO;
4127 			break;
4128 		}
4129 		PF_RULES_RUNLOCK();
4130 		if (cmd == DIOCGETQSTATSV0)
4131 			version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
4132 		else
4133 			version = pq->version;
4134 		error = altq_getqstats(altq, pq->buf, &nbytes, version);
4135 		if (error == 0) {
4136 			pq->scheduler = altq->scheduler;
4137 			pq->nbytes = nbytes;
4138 		}
4139 		break;
4140 	}
4141 #endif /* ALTQ */
4142 
4143 	case DIOCBEGINADDRS: {
4144 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4145 
4146 		PF_RULES_WLOCK();
4147 		pf_empty_kpool(&V_pf_pabuf);
4148 		pp->ticket = ++V_ticket_pabuf;
4149 		PF_RULES_WUNLOCK();
4150 		break;
4151 	}
4152 
4153 	case DIOCADDADDR: {
4154 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4155 		struct pf_kpooladdr	*pa;
4156 		struct pfi_kkif		*kif = NULL;
4157 
4158 #ifndef INET
4159 		if (pp->af == AF_INET) {
4160 			error = EAFNOSUPPORT;
4161 			break;
4162 		}
4163 #endif /* INET */
4164 #ifndef INET6
4165 		if (pp->af == AF_INET6) {
4166 			error = EAFNOSUPPORT;
4167 			break;
4168 		}
4169 #endif /* INET6 */
4170 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
4171 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
4172 		    pp->addr.addr.type != PF_ADDR_TABLE) {
4173 			error = EINVAL;
4174 			break;
4175 		}
4176 		if (pp->addr.addr.p.dyn != NULL) {
4177 			error = EINVAL;
4178 			break;
4179 		}
4180 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
4181 		error = pf_pooladdr_to_kpooladdr(&pp->addr, pa);
4182 		if (error != 0)
4183 			break;
4184 		if (pa->ifname[0])
4185 			kif = pf_kkif_create(M_WAITOK);
4186 		PF_RULES_WLOCK();
4187 		if (pp->ticket != V_ticket_pabuf) {
4188 			PF_RULES_WUNLOCK();
4189 			if (pa->ifname[0])
4190 				pf_kkif_free(kif);
4191 			free(pa, M_PFRULE);
4192 			error = EBUSY;
4193 			break;
4194 		}
4195 		if (pa->ifname[0]) {
4196 			pa->kif = pfi_kkif_attach(kif, pa->ifname);
4197 			kif = NULL;
4198 			pfi_kkif_ref(pa->kif);
4199 		} else
4200 			pa->kif = NULL;
4201 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
4202 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
4203 			if (pa->ifname[0])
4204 				pfi_kkif_unref(pa->kif);
4205 			PF_RULES_WUNLOCK();
4206 			free(pa, M_PFRULE);
4207 			break;
4208 		}
4209 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
4210 		PF_RULES_WUNLOCK();
4211 		break;
4212 	}
4213 
4214 	case DIOCGETADDRS: {
4215 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4216 		struct pf_kpool		*pool;
4217 		struct pf_kpooladdr	*pa;
4218 
4219 		pp->anchor[sizeof(pp->anchor) - 1] = 0;
4220 		pp->nr = 0;
4221 
4222 		PF_RULES_RLOCK();
4223 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
4224 		    pp->r_num, 0, 1, 0);
4225 		if (pool == NULL) {
4226 			PF_RULES_RUNLOCK();
4227 			error = EBUSY;
4228 			break;
4229 		}
4230 		TAILQ_FOREACH(pa, &pool->list, entries)
4231 			pp->nr++;
4232 		PF_RULES_RUNLOCK();
4233 		break;
4234 	}
4235 
4236 	case DIOCGETADDR: {
4237 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4238 		struct pf_kpool		*pool;
4239 		struct pf_kpooladdr	*pa;
4240 		u_int32_t		 nr = 0;
4241 
4242 		pp->anchor[sizeof(pp->anchor) - 1] = 0;
4243 
4244 		PF_RULES_RLOCK();
4245 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
4246 		    pp->r_num, 0, 1, 1);
4247 		if (pool == NULL) {
4248 			PF_RULES_RUNLOCK();
4249 			error = EBUSY;
4250 			break;
4251 		}
4252 		pa = TAILQ_FIRST(&pool->list);
4253 		while ((pa != NULL) && (nr < pp->nr)) {
4254 			pa = TAILQ_NEXT(pa, entries);
4255 			nr++;
4256 		}
4257 		if (pa == NULL) {
4258 			PF_RULES_RUNLOCK();
4259 			error = EBUSY;
4260 			break;
4261 		}
4262 		pf_kpooladdr_to_pooladdr(pa, &pp->addr);
4263 		pf_addr_copyout(&pp->addr.addr);
4264 		PF_RULES_RUNLOCK();
4265 		break;
4266 	}
4267 
4268 	case DIOCCHANGEADDR: {
4269 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
4270 		struct pf_kpool		*pool;
4271 		struct pf_kpooladdr	*oldpa = NULL, *newpa = NULL;
4272 		struct pf_kruleset	*ruleset;
4273 		struct pfi_kkif		*kif = NULL;
4274 
4275 		pca->anchor[sizeof(pca->anchor) - 1] = 0;
4276 
4277 		if (pca->action < PF_CHANGE_ADD_HEAD ||
4278 		    pca->action > PF_CHANGE_REMOVE) {
4279 			error = EINVAL;
4280 			break;
4281 		}
4282 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
4283 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
4284 		    pca->addr.addr.type != PF_ADDR_TABLE) {
4285 			error = EINVAL;
4286 			break;
4287 		}
4288 		if (pca->addr.addr.p.dyn != NULL) {
4289 			error = EINVAL;
4290 			break;
4291 		}
4292 
4293 		if (pca->action != PF_CHANGE_REMOVE) {
4294 #ifndef INET
4295 			if (pca->af == AF_INET) {
4296 				error = EAFNOSUPPORT;
4297 				break;
4298 			}
4299 #endif /* INET */
4300 #ifndef INET6
4301 			if (pca->af == AF_INET6) {
4302 				error = EAFNOSUPPORT;
4303 				break;
4304 			}
4305 #endif /* INET6 */
4306 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
4307 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
4308 			if (newpa->ifname[0])
4309 				kif = pf_kkif_create(M_WAITOK);
4310 			newpa->kif = NULL;
4311 		}
4312 #define	ERROUT(x)	ERROUT_IOCTL(DIOCCHANGEADDR_error, x)
4313 		PF_RULES_WLOCK();
4314 		ruleset = pf_find_kruleset(pca->anchor);
4315 		if (ruleset == NULL)
4316 			ERROUT(EBUSY);
4317 
4318 		pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
4319 		    pca->r_num, pca->r_last, 1, 1);
4320 		if (pool == NULL)
4321 			ERROUT(EBUSY);
4322 
4323 		if (pca->action != PF_CHANGE_REMOVE) {
4324 			if (newpa->ifname[0]) {
4325 				newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
4326 				pfi_kkif_ref(newpa->kif);
4327 				kif = NULL;
4328 			}
4329 
4330 			switch (newpa->addr.type) {
4331 			case PF_ADDR_DYNIFTL:
4332 				error = pfi_dynaddr_setup(&newpa->addr,
4333 				    pca->af);
4334 				break;
4335 			case PF_ADDR_TABLE:
4336 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
4337 				    newpa->addr.v.tblname);
4338 				if (newpa->addr.p.tbl == NULL)
4339 					error = ENOMEM;
4340 				break;
4341 			}
4342 			if (error)
4343 				goto DIOCCHANGEADDR_error;
4344 		}
4345 
4346 		switch (pca->action) {
4347 		case PF_CHANGE_ADD_HEAD:
4348 			oldpa = TAILQ_FIRST(&pool->list);
4349 			break;
4350 		case PF_CHANGE_ADD_TAIL:
4351 			oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
4352 			break;
4353 		default:
4354 			oldpa = TAILQ_FIRST(&pool->list);
4355 			for (int i = 0; oldpa && i < pca->nr; i++)
4356 				oldpa = TAILQ_NEXT(oldpa, entries);
4357 
4358 			if (oldpa == NULL)
4359 				ERROUT(EINVAL);
4360 		}
4361 
4362 		if (pca->action == PF_CHANGE_REMOVE) {
4363 			TAILQ_REMOVE(&pool->list, oldpa, entries);
4364 			switch (oldpa->addr.type) {
4365 			case PF_ADDR_DYNIFTL:
4366 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
4367 				break;
4368 			case PF_ADDR_TABLE:
4369 				pfr_detach_table(oldpa->addr.p.tbl);
4370 				break;
4371 			}
4372 			if (oldpa->kif)
4373 				pfi_kkif_unref(oldpa->kif);
4374 			free(oldpa, M_PFRULE);
4375 		} else {
4376 			if (oldpa == NULL)
4377 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
4378 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
4379 			    pca->action == PF_CHANGE_ADD_BEFORE)
4380 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
4381 			else
4382 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
4383 				    newpa, entries);
4384 		}
4385 
4386 		pool->cur = TAILQ_FIRST(&pool->list);
4387 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
4388 		PF_RULES_WUNLOCK();
4389 		break;
4390 
4391 #undef ERROUT
4392 DIOCCHANGEADDR_error:
4393 		if (newpa != NULL) {
4394 			if (newpa->kif)
4395 				pfi_kkif_unref(newpa->kif);
4396 			free(newpa, M_PFRULE);
4397 		}
4398 		PF_RULES_WUNLOCK();
4399 		pf_kkif_free(kif);
4400 		break;
4401 	}
4402 
4403 	case DIOCGETRULESETS: {
4404 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
4405 		struct pf_kruleset	*ruleset;
4406 		struct pf_kanchor	*anchor;
4407 
4408 		pr->path[sizeof(pr->path) - 1] = 0;
4409 
4410 		PF_RULES_RLOCK();
4411 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4412 			PF_RULES_RUNLOCK();
4413 			error = ENOENT;
4414 			break;
4415 		}
4416 		pr->nr = 0;
4417 		if (ruleset->anchor == NULL) {
4418 			/* XXX kludge for pf_main_ruleset */
4419 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4420 				if (anchor->parent == NULL)
4421 					pr->nr++;
4422 		} else {
4423 			RB_FOREACH(anchor, pf_kanchor_node,
4424 			    &ruleset->anchor->children)
4425 				pr->nr++;
4426 		}
4427 		PF_RULES_RUNLOCK();
4428 		break;
4429 	}
4430 
4431 	case DIOCGETRULESET: {
4432 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
4433 		struct pf_kruleset	*ruleset;
4434 		struct pf_kanchor	*anchor;
4435 		u_int32_t		 nr = 0;
4436 
4437 		pr->path[sizeof(pr->path) - 1] = 0;
4438 
4439 		PF_RULES_RLOCK();
4440 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4441 			PF_RULES_RUNLOCK();
4442 			error = ENOENT;
4443 			break;
4444 		}
4445 		pr->name[0] = 0;
4446 		if (ruleset->anchor == NULL) {
4447 			/* XXX kludge for pf_main_ruleset */
4448 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4449 				if (anchor->parent == NULL && nr++ == pr->nr) {
4450 					strlcpy(pr->name, anchor->name,
4451 					    sizeof(pr->name));
4452 					break;
4453 				}
4454 		} else {
4455 			RB_FOREACH(anchor, pf_kanchor_node,
4456 			    &ruleset->anchor->children)
4457 				if (nr++ == pr->nr) {
4458 					strlcpy(pr->name, anchor->name,
4459 					    sizeof(pr->name));
4460 					break;
4461 				}
4462 		}
4463 		if (!pr->name[0])
4464 			error = EBUSY;
4465 		PF_RULES_RUNLOCK();
4466 		break;
4467 	}
4468 
4469 	case DIOCRCLRTABLES: {
4470 		struct pfioc_table *io = (struct pfioc_table *)addr;
4471 
4472 		if (io->pfrio_esize != 0) {
4473 			error = ENODEV;
4474 			break;
4475 		}
4476 		PF_RULES_WLOCK();
4477 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
4478 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4479 		PF_RULES_WUNLOCK();
4480 		break;
4481 	}
4482 
4483 	case DIOCRADDTABLES: {
4484 		struct pfioc_table *io = (struct pfioc_table *)addr;
4485 		struct pfr_table *pfrts;
4486 		size_t totlen;
4487 
4488 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4489 			error = ENODEV;
4490 			break;
4491 		}
4492 
4493 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4494 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4495 			error = ENOMEM;
4496 			break;
4497 		}
4498 
4499 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4500 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4501 		    M_TEMP, M_WAITOK);
4502 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4503 		if (error) {
4504 			free(pfrts, M_TEMP);
4505 			break;
4506 		}
4507 		PF_RULES_WLOCK();
4508 		error = pfr_add_tables(pfrts, io->pfrio_size,
4509 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4510 		PF_RULES_WUNLOCK();
4511 		free(pfrts, M_TEMP);
4512 		break;
4513 	}
4514 
4515 	case DIOCRDELTABLES: {
4516 		struct pfioc_table *io = (struct pfioc_table *)addr;
4517 		struct pfr_table *pfrts;
4518 		size_t totlen;
4519 
4520 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4521 			error = ENODEV;
4522 			break;
4523 		}
4524 
4525 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4526 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4527 			error = ENOMEM;
4528 			break;
4529 		}
4530 
4531 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4532 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4533 		    M_TEMP, M_WAITOK);
4534 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4535 		if (error) {
4536 			free(pfrts, M_TEMP);
4537 			break;
4538 		}
4539 		PF_RULES_WLOCK();
4540 		error = pfr_del_tables(pfrts, io->pfrio_size,
4541 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4542 		PF_RULES_WUNLOCK();
4543 		free(pfrts, M_TEMP);
4544 		break;
4545 	}
4546 
4547 	case DIOCRGETTABLES: {
4548 		struct pfioc_table *io = (struct pfioc_table *)addr;
4549 		struct pfr_table *pfrts;
4550 		size_t totlen;
4551 		int n;
4552 
4553 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4554 			error = ENODEV;
4555 			break;
4556 		}
4557 		PF_RULES_RLOCK();
4558 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4559 		if (n < 0) {
4560 			PF_RULES_RUNLOCK();
4561 			error = EINVAL;
4562 			break;
4563 		}
4564 		io->pfrio_size = min(io->pfrio_size, n);
4565 
4566 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4567 
4568 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4569 		    M_TEMP, M_NOWAIT | M_ZERO);
4570 		if (pfrts == NULL) {
4571 			error = ENOMEM;
4572 			PF_RULES_RUNLOCK();
4573 			break;
4574 		}
4575 		error = pfr_get_tables(&io->pfrio_table, pfrts,
4576 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4577 		PF_RULES_RUNLOCK();
4578 		if (error == 0)
4579 			error = copyout(pfrts, io->pfrio_buffer, totlen);
4580 		free(pfrts, M_TEMP);
4581 		break;
4582 	}
4583 
4584 	case DIOCRGETTSTATS: {
4585 		struct pfioc_table *io = (struct pfioc_table *)addr;
4586 		struct pfr_tstats *pfrtstats;
4587 		size_t totlen;
4588 		int n;
4589 
4590 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
4591 			error = ENODEV;
4592 			break;
4593 		}
4594 		PF_TABLE_STATS_LOCK();
4595 		PF_RULES_RLOCK();
4596 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4597 		if (n < 0) {
4598 			PF_RULES_RUNLOCK();
4599 			PF_TABLE_STATS_UNLOCK();
4600 			error = EINVAL;
4601 			break;
4602 		}
4603 		io->pfrio_size = min(io->pfrio_size, n);
4604 
4605 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
4606 		pfrtstats = mallocarray(io->pfrio_size,
4607 		    sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT | M_ZERO);
4608 		if (pfrtstats == NULL) {
4609 			error = ENOMEM;
4610 			PF_RULES_RUNLOCK();
4611 			PF_TABLE_STATS_UNLOCK();
4612 			break;
4613 		}
4614 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
4615 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4616 		PF_RULES_RUNLOCK();
4617 		PF_TABLE_STATS_UNLOCK();
4618 		if (error == 0)
4619 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
4620 		free(pfrtstats, M_TEMP);
4621 		break;
4622 	}
4623 
4624 	case DIOCRCLRTSTATS: {
4625 		struct pfioc_table *io = (struct pfioc_table *)addr;
4626 		struct pfr_table *pfrts;
4627 		size_t totlen;
4628 
4629 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4630 			error = ENODEV;
4631 			break;
4632 		}
4633 
4634 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4635 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4636 			/* We used to count tables and use the minimum required
4637 			 * size, so we didn't fail on overly large requests.
4638 			 * Keep doing so. */
4639 			io->pfrio_size = pf_ioctl_maxcount;
4640 			break;
4641 		}
4642 
4643 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4644 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4645 		    M_TEMP, M_WAITOK);
4646 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4647 		if (error) {
4648 			free(pfrts, M_TEMP);
4649 			break;
4650 		}
4651 
4652 		PF_TABLE_STATS_LOCK();
4653 		PF_RULES_RLOCK();
4654 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
4655 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4656 		PF_RULES_RUNLOCK();
4657 		PF_TABLE_STATS_UNLOCK();
4658 		free(pfrts, M_TEMP);
4659 		break;
4660 	}
4661 
4662 	case DIOCRSETTFLAGS: {
4663 		struct pfioc_table *io = (struct pfioc_table *)addr;
4664 		struct pfr_table *pfrts;
4665 		size_t totlen;
4666 		int n;
4667 
4668 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4669 			error = ENODEV;
4670 			break;
4671 		}
4672 
4673 		PF_RULES_RLOCK();
4674 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4675 		if (n < 0) {
4676 			PF_RULES_RUNLOCK();
4677 			error = EINVAL;
4678 			break;
4679 		}
4680 
4681 		io->pfrio_size = min(io->pfrio_size, n);
4682 		PF_RULES_RUNLOCK();
4683 
4684 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4685 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4686 		    M_TEMP, M_WAITOK);
4687 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4688 		if (error) {
4689 			free(pfrts, M_TEMP);
4690 			break;
4691 		}
4692 		PF_RULES_WLOCK();
4693 		error = pfr_set_tflags(pfrts, io->pfrio_size,
4694 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
4695 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4696 		PF_RULES_WUNLOCK();
4697 		free(pfrts, M_TEMP);
4698 		break;
4699 	}
4700 
4701 	case DIOCRCLRADDRS: {
4702 		struct pfioc_table *io = (struct pfioc_table *)addr;
4703 
4704 		if (io->pfrio_esize != 0) {
4705 			error = ENODEV;
4706 			break;
4707 		}
4708 		PF_RULES_WLOCK();
4709 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
4710 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4711 		PF_RULES_WUNLOCK();
4712 		break;
4713 	}
4714 
4715 	case DIOCRADDADDRS: {
4716 		struct pfioc_table *io = (struct pfioc_table *)addr;
4717 		struct pfr_addr *pfras;
4718 		size_t totlen;
4719 
4720 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4721 			error = ENODEV;
4722 			break;
4723 		}
4724 		if (io->pfrio_size < 0 ||
4725 		    io->pfrio_size > pf_ioctl_maxcount ||
4726 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4727 			error = EINVAL;
4728 			break;
4729 		}
4730 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4731 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4732 		    M_TEMP, M_WAITOK);
4733 		error = copyin(io->pfrio_buffer, pfras, totlen);
4734 		if (error) {
4735 			free(pfras, M_TEMP);
4736 			break;
4737 		}
4738 		PF_RULES_WLOCK();
4739 		error = pfr_add_addrs(&io->pfrio_table, pfras,
4740 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
4741 		    PFR_FLAG_USERIOCTL);
4742 		PF_RULES_WUNLOCK();
4743 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4744 			error = copyout(pfras, io->pfrio_buffer, totlen);
4745 		free(pfras, M_TEMP);
4746 		break;
4747 	}
4748 
4749 	case DIOCRDELADDRS: {
4750 		struct pfioc_table *io = (struct pfioc_table *)addr;
4751 		struct pfr_addr *pfras;
4752 		size_t totlen;
4753 
4754 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4755 			error = ENODEV;
4756 			break;
4757 		}
4758 		if (io->pfrio_size < 0 ||
4759 		    io->pfrio_size > pf_ioctl_maxcount ||
4760 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4761 			error = EINVAL;
4762 			break;
4763 		}
4764 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4765 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4766 		    M_TEMP, M_WAITOK);
4767 		error = copyin(io->pfrio_buffer, pfras, totlen);
4768 		if (error) {
4769 			free(pfras, M_TEMP);
4770 			break;
4771 		}
4772 		PF_RULES_WLOCK();
4773 		error = pfr_del_addrs(&io->pfrio_table, pfras,
4774 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
4775 		    PFR_FLAG_USERIOCTL);
4776 		PF_RULES_WUNLOCK();
4777 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4778 			error = copyout(pfras, io->pfrio_buffer, totlen);
4779 		free(pfras, M_TEMP);
4780 		break;
4781 	}
4782 
4783 	case DIOCRSETADDRS: {
4784 		struct pfioc_table *io = (struct pfioc_table *)addr;
4785 		struct pfr_addr *pfras;
4786 		size_t totlen, count;
4787 
4788 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4789 			error = ENODEV;
4790 			break;
4791 		}
4792 		if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
4793 			error = EINVAL;
4794 			break;
4795 		}
4796 		count = max(io->pfrio_size, io->pfrio_size2);
4797 		if (count > pf_ioctl_maxcount ||
4798 		    WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
4799 			error = EINVAL;
4800 			break;
4801 		}
4802 		totlen = count * sizeof(struct pfr_addr);
4803 		pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
4804 		    M_WAITOK);
4805 		error = copyin(io->pfrio_buffer, pfras, totlen);
4806 		if (error) {
4807 			free(pfras, M_TEMP);
4808 			break;
4809 		}
4810 		PF_RULES_WLOCK();
4811 		error = pfr_set_addrs(&io->pfrio_table, pfras,
4812 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
4813 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
4814 		    PFR_FLAG_USERIOCTL, 0);
4815 		PF_RULES_WUNLOCK();
4816 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4817 			error = copyout(pfras, io->pfrio_buffer, totlen);
4818 		free(pfras, M_TEMP);
4819 		break;
4820 	}
4821 
4822 	case DIOCRGETADDRS: {
4823 		struct pfioc_table *io = (struct pfioc_table *)addr;
4824 		struct pfr_addr *pfras;
4825 		size_t totlen;
4826 
4827 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4828 			error = ENODEV;
4829 			break;
4830 		}
4831 		if (io->pfrio_size < 0 ||
4832 		    io->pfrio_size > pf_ioctl_maxcount ||
4833 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4834 			error = EINVAL;
4835 			break;
4836 		}
4837 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4838 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4839 		    M_TEMP, M_WAITOK | M_ZERO);
4840 		PF_RULES_RLOCK();
4841 		error = pfr_get_addrs(&io->pfrio_table, pfras,
4842 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4843 		PF_RULES_RUNLOCK();
4844 		if (error == 0)
4845 			error = copyout(pfras, io->pfrio_buffer, totlen);
4846 		free(pfras, M_TEMP);
4847 		break;
4848 	}
4849 
4850 	case DIOCRGETASTATS: {
4851 		struct pfioc_table *io = (struct pfioc_table *)addr;
4852 		struct pfr_astats *pfrastats;
4853 		size_t totlen;
4854 
4855 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
4856 			error = ENODEV;
4857 			break;
4858 		}
4859 		if (io->pfrio_size < 0 ||
4860 		    io->pfrio_size > pf_ioctl_maxcount ||
4861 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
4862 			error = EINVAL;
4863 			break;
4864 		}
4865 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
4866 		pfrastats = mallocarray(io->pfrio_size,
4867 		    sizeof(struct pfr_astats), M_TEMP, M_WAITOK | M_ZERO);
4868 		PF_RULES_RLOCK();
4869 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
4870 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4871 		PF_RULES_RUNLOCK();
4872 		if (error == 0)
4873 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
4874 		free(pfrastats, M_TEMP);
4875 		break;
4876 	}
4877 
4878 	case DIOCRCLRASTATS: {
4879 		struct pfioc_table *io = (struct pfioc_table *)addr;
4880 		struct pfr_addr *pfras;
4881 		size_t totlen;
4882 
4883 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4884 			error = ENODEV;
4885 			break;
4886 		}
4887 		if (io->pfrio_size < 0 ||
4888 		    io->pfrio_size > pf_ioctl_maxcount ||
4889 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4890 			error = EINVAL;
4891 			break;
4892 		}
4893 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4894 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4895 		    M_TEMP, M_WAITOK);
4896 		error = copyin(io->pfrio_buffer, pfras, totlen);
4897 		if (error) {
4898 			free(pfras, M_TEMP);
4899 			break;
4900 		}
4901 		PF_RULES_WLOCK();
4902 		error = pfr_clr_astats(&io->pfrio_table, pfras,
4903 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
4904 		    PFR_FLAG_USERIOCTL);
4905 		PF_RULES_WUNLOCK();
4906 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4907 			error = copyout(pfras, io->pfrio_buffer, totlen);
4908 		free(pfras, M_TEMP);
4909 		break;
4910 	}
4911 
4912 	case DIOCRTSTADDRS: {
4913 		struct pfioc_table *io = (struct pfioc_table *)addr;
4914 		struct pfr_addr *pfras;
4915 		size_t totlen;
4916 
4917 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4918 			error = ENODEV;
4919 			break;
4920 		}
4921 		if (io->pfrio_size < 0 ||
4922 		    io->pfrio_size > pf_ioctl_maxcount ||
4923 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4924 			error = EINVAL;
4925 			break;
4926 		}
4927 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4928 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4929 		    M_TEMP, M_WAITOK);
4930 		error = copyin(io->pfrio_buffer, pfras, totlen);
4931 		if (error) {
4932 			free(pfras, M_TEMP);
4933 			break;
4934 		}
4935 		PF_RULES_RLOCK();
4936 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
4937 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
4938 		    PFR_FLAG_USERIOCTL);
4939 		PF_RULES_RUNLOCK();
4940 		if (error == 0)
4941 			error = copyout(pfras, io->pfrio_buffer, totlen);
4942 		free(pfras, M_TEMP);
4943 		break;
4944 	}
4945 
4946 	case DIOCRINADEFINE: {
4947 		struct pfioc_table *io = (struct pfioc_table *)addr;
4948 		struct pfr_addr *pfras;
4949 		size_t totlen;
4950 
4951 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4952 			error = ENODEV;
4953 			break;
4954 		}
4955 		if (io->pfrio_size < 0 ||
4956 		    io->pfrio_size > pf_ioctl_maxcount ||
4957 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4958 			error = EINVAL;
4959 			break;
4960 		}
4961 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4962 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4963 		    M_TEMP, M_WAITOK);
4964 		error = copyin(io->pfrio_buffer, pfras, totlen);
4965 		if (error) {
4966 			free(pfras, M_TEMP);
4967 			break;
4968 		}
4969 		PF_RULES_WLOCK();
4970 		error = pfr_ina_define(&io->pfrio_table, pfras,
4971 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
4972 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4973 		PF_RULES_WUNLOCK();
4974 		free(pfras, M_TEMP);
4975 		break;
4976 	}
4977 
4978 	case DIOCOSFPADD: {
4979 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
4980 		PF_RULES_WLOCK();
4981 		error = pf_osfp_add(io);
4982 		PF_RULES_WUNLOCK();
4983 		break;
4984 	}
4985 
4986 	case DIOCOSFPGET: {
4987 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
4988 		PF_RULES_RLOCK();
4989 		error = pf_osfp_get(io);
4990 		PF_RULES_RUNLOCK();
4991 		break;
4992 	}
4993 
4994 	case DIOCXBEGIN: {
4995 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
4996 		struct pfioc_trans_e	*ioes, *ioe;
4997 		size_t			 totlen;
4998 		int			 i;
4999 
5000 		if (io->esize != sizeof(*ioe)) {
5001 			error = ENODEV;
5002 			break;
5003 		}
5004 		if (io->size < 0 ||
5005 		    io->size > pf_ioctl_maxcount ||
5006 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5007 			error = EINVAL;
5008 			break;
5009 		}
5010 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5011 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5012 		    M_TEMP, M_WAITOK);
5013 		error = copyin(io->array, ioes, totlen);
5014 		if (error) {
5015 			free(ioes, M_TEMP);
5016 			break;
5017 		}
5018 		/* Ensure there's no more ethernet rules to clean up. */
5019 		NET_EPOCH_DRAIN_CALLBACKS();
5020 		PF_RULES_WLOCK();
5021 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5022 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5023 			switch (ioe->rs_num) {
5024 			case PF_RULESET_ETH:
5025 				if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) {
5026 					PF_RULES_WUNLOCK();
5027 					free(ioes, M_TEMP);
5028 					goto fail;
5029 				}
5030 				break;
5031 #ifdef ALTQ
5032 			case PF_RULESET_ALTQ:
5033 				if (ioe->anchor[0]) {
5034 					PF_RULES_WUNLOCK();
5035 					free(ioes, M_TEMP);
5036 					error = EINVAL;
5037 					goto fail;
5038 				}
5039 				if ((error = pf_begin_altq(&ioe->ticket))) {
5040 					PF_RULES_WUNLOCK();
5041 					free(ioes, M_TEMP);
5042 					goto fail;
5043 				}
5044 				break;
5045 #endif /* ALTQ */
5046 			case PF_RULESET_TABLE:
5047 			    {
5048 				struct pfr_table table;
5049 
5050 				bzero(&table, sizeof(table));
5051 				strlcpy(table.pfrt_anchor, ioe->anchor,
5052 				    sizeof(table.pfrt_anchor));
5053 				if ((error = pfr_ina_begin(&table,
5054 				    &ioe->ticket, NULL, 0))) {
5055 					PF_RULES_WUNLOCK();
5056 					free(ioes, M_TEMP);
5057 					goto fail;
5058 				}
5059 				break;
5060 			    }
5061 			default:
5062 				if ((error = pf_begin_rules(&ioe->ticket,
5063 				    ioe->rs_num, ioe->anchor))) {
5064 					PF_RULES_WUNLOCK();
5065 					free(ioes, M_TEMP);
5066 					goto fail;
5067 				}
5068 				break;
5069 			}
5070 		}
5071 		PF_RULES_WUNLOCK();
5072 		error = copyout(ioes, io->array, totlen);
5073 		free(ioes, M_TEMP);
5074 		break;
5075 	}
5076 
5077 	case DIOCXROLLBACK: {
5078 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5079 		struct pfioc_trans_e	*ioe, *ioes;
5080 		size_t			 totlen;
5081 		int			 i;
5082 
5083 		if (io->esize != sizeof(*ioe)) {
5084 			error = ENODEV;
5085 			break;
5086 		}
5087 		if (io->size < 0 ||
5088 		    io->size > pf_ioctl_maxcount ||
5089 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5090 			error = EINVAL;
5091 			break;
5092 		}
5093 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5094 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5095 		    M_TEMP, M_WAITOK);
5096 		error = copyin(io->array, ioes, totlen);
5097 		if (error) {
5098 			free(ioes, M_TEMP);
5099 			break;
5100 		}
5101 		PF_RULES_WLOCK();
5102 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5103 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5104 			switch (ioe->rs_num) {
5105 			case PF_RULESET_ETH:
5106 				if ((error = pf_rollback_eth(ioe->ticket,
5107 				    ioe->anchor))) {
5108 					PF_RULES_WUNLOCK();
5109 					free(ioes, M_TEMP);
5110 					goto fail; /* really bad */
5111 				}
5112 				break;
5113 #ifdef ALTQ
5114 			case PF_RULESET_ALTQ:
5115 				if (ioe->anchor[0]) {
5116 					PF_RULES_WUNLOCK();
5117 					free(ioes, M_TEMP);
5118 					error = EINVAL;
5119 					goto fail;
5120 				}
5121 				if ((error = pf_rollback_altq(ioe->ticket))) {
5122 					PF_RULES_WUNLOCK();
5123 					free(ioes, M_TEMP);
5124 					goto fail; /* really bad */
5125 				}
5126 				break;
5127 #endif /* ALTQ */
5128 			case PF_RULESET_TABLE:
5129 			    {
5130 				struct pfr_table table;
5131 
5132 				bzero(&table, sizeof(table));
5133 				strlcpy(table.pfrt_anchor, ioe->anchor,
5134 				    sizeof(table.pfrt_anchor));
5135 				if ((error = pfr_ina_rollback(&table,
5136 				    ioe->ticket, NULL, 0))) {
5137 					PF_RULES_WUNLOCK();
5138 					free(ioes, M_TEMP);
5139 					goto fail; /* really bad */
5140 				}
5141 				break;
5142 			    }
5143 			default:
5144 				if ((error = pf_rollback_rules(ioe->ticket,
5145 				    ioe->rs_num, ioe->anchor))) {
5146 					PF_RULES_WUNLOCK();
5147 					free(ioes, M_TEMP);
5148 					goto fail; /* really bad */
5149 				}
5150 				break;
5151 			}
5152 		}
5153 		PF_RULES_WUNLOCK();
5154 		free(ioes, M_TEMP);
5155 		break;
5156 	}
5157 
5158 	case DIOCXCOMMIT: {
5159 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5160 		struct pfioc_trans_e	*ioe, *ioes;
5161 		struct pf_kruleset	*rs;
5162 		struct pf_keth_ruleset	*ers;
5163 		size_t			 totlen;
5164 		int			 i;
5165 
5166 		if (io->esize != sizeof(*ioe)) {
5167 			error = ENODEV;
5168 			break;
5169 		}
5170 
5171 		if (io->size < 0 ||
5172 		    io->size > pf_ioctl_maxcount ||
5173 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5174 			error = EINVAL;
5175 			break;
5176 		}
5177 
5178 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5179 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5180 		    M_TEMP, M_WAITOK);
5181 		error = copyin(io->array, ioes, totlen);
5182 		if (error) {
5183 			free(ioes, M_TEMP);
5184 			break;
5185 		}
5186 		PF_RULES_WLOCK();
5187 		/* First makes sure everything will succeed. */
5188 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5189 			ioe->anchor[sizeof(ioe->anchor) - 1] = 0;
5190 			switch (ioe->rs_num) {
5191 			case PF_RULESET_ETH:
5192 				ers = pf_find_keth_ruleset(ioe->anchor);
5193 				if (ers == NULL || ioe->ticket == 0 ||
5194 				    ioe->ticket != ers->inactive.ticket) {
5195 					PF_RULES_WUNLOCK();
5196 					free(ioes, M_TEMP);
5197 					error = EINVAL;
5198 					goto fail;
5199 				}
5200 				break;
5201 #ifdef ALTQ
5202 			case PF_RULESET_ALTQ:
5203 				if (ioe->anchor[0]) {
5204 					PF_RULES_WUNLOCK();
5205 					free(ioes, M_TEMP);
5206 					error = EINVAL;
5207 					goto fail;
5208 				}
5209 				if (!V_altqs_inactive_open || ioe->ticket !=
5210 				    V_ticket_altqs_inactive) {
5211 					PF_RULES_WUNLOCK();
5212 					free(ioes, M_TEMP);
5213 					error = EBUSY;
5214 					goto fail;
5215 				}
5216 				break;
5217 #endif /* ALTQ */
5218 			case PF_RULESET_TABLE:
5219 				rs = pf_find_kruleset(ioe->anchor);
5220 				if (rs == NULL || !rs->topen || ioe->ticket !=
5221 				    rs->tticket) {
5222 					PF_RULES_WUNLOCK();
5223 					free(ioes, M_TEMP);
5224 					error = EBUSY;
5225 					goto fail;
5226 				}
5227 				break;
5228 			default:
5229 				if (ioe->rs_num < 0 || ioe->rs_num >=
5230 				    PF_RULESET_MAX) {
5231 					PF_RULES_WUNLOCK();
5232 					free(ioes, M_TEMP);
5233 					error = EINVAL;
5234 					goto fail;
5235 				}
5236 				rs = pf_find_kruleset(ioe->anchor);
5237 				if (rs == NULL ||
5238 				    !rs->rules[ioe->rs_num].inactive.open ||
5239 				    rs->rules[ioe->rs_num].inactive.ticket !=
5240 				    ioe->ticket) {
5241 					PF_RULES_WUNLOCK();
5242 					free(ioes, M_TEMP);
5243 					error = EBUSY;
5244 					goto fail;
5245 				}
5246 				break;
5247 			}
5248 		}
5249 		/* Now do the commit - no errors should happen here. */
5250 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5251 			switch (ioe->rs_num) {
5252 			case PF_RULESET_ETH:
5253 				if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) {
5254 					PF_RULES_WUNLOCK();
5255 					free(ioes, M_TEMP);
5256 					goto fail; /* really bad */
5257 				}
5258 				break;
5259 #ifdef ALTQ
5260 			case PF_RULESET_ALTQ:
5261 				if ((error = pf_commit_altq(ioe->ticket))) {
5262 					PF_RULES_WUNLOCK();
5263 					free(ioes, M_TEMP);
5264 					goto fail; /* really bad */
5265 				}
5266 				break;
5267 #endif /* ALTQ */
5268 			case PF_RULESET_TABLE:
5269 			    {
5270 				struct pfr_table table;
5271 
5272 				bzero(&table, sizeof(table));
5273 				(void)strlcpy(table.pfrt_anchor, ioe->anchor,
5274 				    sizeof(table.pfrt_anchor));
5275 				if ((error = pfr_ina_commit(&table,
5276 				    ioe->ticket, NULL, NULL, 0))) {
5277 					PF_RULES_WUNLOCK();
5278 					free(ioes, M_TEMP);
5279 					goto fail; /* really bad */
5280 				}
5281 				break;
5282 			    }
5283 			default:
5284 				if ((error = pf_commit_rules(ioe->ticket,
5285 				    ioe->rs_num, ioe->anchor))) {
5286 					PF_RULES_WUNLOCK();
5287 					free(ioes, M_TEMP);
5288 					goto fail; /* really bad */
5289 				}
5290 				break;
5291 			}
5292 		}
5293 		PF_RULES_WUNLOCK();
5294 
5295 		/* Only hook into EtherNet taffic if we've got rules for it. */
5296 		if (! TAILQ_EMPTY(V_pf_keth->active.rules))
5297 			hook_pf_eth();
5298 		else
5299 			dehook_pf_eth();
5300 
5301 		free(ioes, M_TEMP);
5302 		break;
5303 	}
5304 
5305 	case DIOCGETSRCNODES: {
5306 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
5307 		struct pf_srchash	*sh;
5308 		struct pf_ksrc_node	*n;
5309 		struct pf_src_node	*p, *pstore;
5310 		uint32_t		 i, nr = 0;
5311 
5312 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5313 				i++, sh++) {
5314 			PF_HASHROW_LOCK(sh);
5315 			LIST_FOREACH(n, &sh->nodes, entry)
5316 				nr++;
5317 			PF_HASHROW_UNLOCK(sh);
5318 		}
5319 
5320 		psn->psn_len = min(psn->psn_len,
5321 		    sizeof(struct pf_src_node) * nr);
5322 
5323 		if (psn->psn_len == 0) {
5324 			psn->psn_len = sizeof(struct pf_src_node) * nr;
5325 			break;
5326 		}
5327 
5328 		nr = 0;
5329 
5330 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
5331 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5332 		    i++, sh++) {
5333 		    PF_HASHROW_LOCK(sh);
5334 		    LIST_FOREACH(n, &sh->nodes, entry) {
5335 
5336 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
5337 				break;
5338 
5339 			pf_src_node_copy(n, p);
5340 
5341 			p++;
5342 			nr++;
5343 		    }
5344 		    PF_HASHROW_UNLOCK(sh);
5345 		}
5346 		error = copyout(pstore, psn->psn_src_nodes,
5347 		    sizeof(struct pf_src_node) * nr);
5348 		if (error) {
5349 			free(pstore, M_TEMP);
5350 			break;
5351 		}
5352 		psn->psn_len = sizeof(struct pf_src_node) * nr;
5353 		free(pstore, M_TEMP);
5354 		break;
5355 	}
5356 
5357 	case DIOCCLRSRCNODES: {
5358 		pf_clear_srcnodes(NULL);
5359 		pf_purge_expired_src_nodes();
5360 		break;
5361 	}
5362 
5363 	case DIOCKILLSRCNODES:
5364 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
5365 		break;
5366 
5367 #ifdef COMPAT_FREEBSD13
5368 	case DIOCKEEPCOUNTERS_FREEBSD13:
5369 #endif
5370 	case DIOCKEEPCOUNTERS:
5371 		error = pf_keepcounters((struct pfioc_nv *)addr);
5372 		break;
5373 
5374 	case DIOCGETSYNCOOKIES:
5375 		error = pf_get_syncookies((struct pfioc_nv *)addr);
5376 		break;
5377 
5378 	case DIOCSETSYNCOOKIES:
5379 		error = pf_set_syncookies((struct pfioc_nv *)addr);
5380 		break;
5381 
5382 	case DIOCSETHOSTID: {
5383 		u_int32_t	*hostid = (u_int32_t *)addr;
5384 
5385 		PF_RULES_WLOCK();
5386 		if (*hostid == 0)
5387 			V_pf_status.hostid = arc4random();
5388 		else
5389 			V_pf_status.hostid = *hostid;
5390 		PF_RULES_WUNLOCK();
5391 		break;
5392 	}
5393 
5394 	case DIOCOSFPFLUSH:
5395 		PF_RULES_WLOCK();
5396 		pf_osfp_flush();
5397 		PF_RULES_WUNLOCK();
5398 		break;
5399 
5400 	case DIOCIGETIFACES: {
5401 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5402 		struct pfi_kif *ifstore;
5403 		size_t bufsiz;
5404 
5405 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
5406 			error = ENODEV;
5407 			break;
5408 		}
5409 
5410 		if (io->pfiio_size < 0 ||
5411 		    io->pfiio_size > pf_ioctl_maxcount ||
5412 		    WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
5413 			error = EINVAL;
5414 			break;
5415 		}
5416 
5417 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5418 
5419 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
5420 		ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
5421 		    M_TEMP, M_WAITOK | M_ZERO);
5422 
5423 		PF_RULES_RLOCK();
5424 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
5425 		PF_RULES_RUNLOCK();
5426 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
5427 		free(ifstore, M_TEMP);
5428 		break;
5429 	}
5430 
5431 	case DIOCSETIFFLAG: {
5432 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5433 
5434 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5435 
5436 		PF_RULES_WLOCK();
5437 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
5438 		PF_RULES_WUNLOCK();
5439 		break;
5440 	}
5441 
5442 	case DIOCCLRIFFLAG: {
5443 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5444 
5445 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5446 
5447 		PF_RULES_WLOCK();
5448 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
5449 		PF_RULES_WUNLOCK();
5450 		break;
5451 	}
5452 
5453 	case DIOCSETREASS: {
5454 		u_int32_t	*reass = (u_int32_t *)addr;
5455 
5456 		V_pf_status.reass = *reass & (PF_REASS_ENABLED|PF_REASS_NODF);
5457 		/* Removal of DF flag without reassembly enabled is not a
5458 		 * valid combination. Disable reassembly in such case. */
5459 		if (!(V_pf_status.reass & PF_REASS_ENABLED))
5460 			V_pf_status.reass = 0;
5461 		break;
5462 	}
5463 
5464 	default:
5465 		error = ENODEV;
5466 		break;
5467 	}
5468 fail:
5469 	CURVNET_RESTORE();
5470 
5471 #undef ERROUT_IOCTL
5472 
5473 	return (error);
5474 }
5475 
5476 void
5477 pfsync_state_export(union pfsync_state_union *sp, struct pf_kstate *st, int msg_version)
5478 {
5479 	bzero(sp, sizeof(union pfsync_state_union));
5480 
5481 	/* copy from state key */
5482 	sp->pfs_1301.key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5483 	sp->pfs_1301.key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5484 	sp->pfs_1301.key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5485 	sp->pfs_1301.key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5486 	sp->pfs_1301.key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5487 	sp->pfs_1301.key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5488 	sp->pfs_1301.key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5489 	sp->pfs_1301.key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5490 	sp->pfs_1301.proto = st->key[PF_SK_WIRE]->proto;
5491 	sp->pfs_1301.af = st->key[PF_SK_WIRE]->af;
5492 
5493 	/* copy from state */
5494 	strlcpy(sp->pfs_1301.ifname, st->kif->pfik_name, sizeof(sp->pfs_1301.ifname));
5495 	bcopy(&st->rt_addr, &sp->pfs_1301.rt_addr, sizeof(sp->pfs_1301.rt_addr));
5496 	sp->pfs_1301.creation = htonl(time_uptime - (st->creation / 1000));
5497 	sp->pfs_1301.expire = pf_state_expires(st);
5498 	if (sp->pfs_1301.expire <= time_uptime)
5499 		sp->pfs_1301.expire = htonl(0);
5500 	else
5501 		sp->pfs_1301.expire = htonl(sp->pfs_1301.expire - time_uptime);
5502 
5503 	sp->pfs_1301.direction = st->direction;
5504 	sp->pfs_1301.log = st->act.log;
5505 	sp->pfs_1301.timeout = st->timeout;
5506 
5507 	switch (msg_version) {
5508 		case PFSYNC_MSG_VERSION_1301:
5509 			sp->pfs_1301.state_flags = st->state_flags;
5510 			break;
5511 		case PFSYNC_MSG_VERSION_1400:
5512 			sp->pfs_1400.state_flags = htons(st->state_flags);
5513 			sp->pfs_1400.qid = htons(st->act.qid);
5514 			sp->pfs_1400.pqid = htons(st->act.pqid);
5515 			sp->pfs_1400.dnpipe = htons(st->act.dnpipe);
5516 			sp->pfs_1400.dnrpipe = htons(st->act.dnrpipe);
5517 			sp->pfs_1400.rtableid = htonl(st->act.rtableid);
5518 			sp->pfs_1400.min_ttl = st->act.min_ttl;
5519 			sp->pfs_1400.set_tos = st->act.set_tos;
5520 			sp->pfs_1400.max_mss = htons(st->act.max_mss);
5521 			sp->pfs_1400.set_prio[0] = st->act.set_prio[0];
5522 			sp->pfs_1400.set_prio[1] = st->act.set_prio[1];
5523 			sp->pfs_1400.rt = st->rt;
5524 			if (st->rt_kif)
5525 				strlcpy(sp->pfs_1400.rt_ifname,
5526 				    st->rt_kif->pfik_name,
5527 				    sizeof(sp->pfs_1400.rt_ifname));
5528 			break;
5529 		default:
5530 			panic("%s: Unsupported pfsync_msg_version %d",
5531 			    __func__, msg_version);
5532 	}
5533 
5534 	if (st->src_node)
5535 		sp->pfs_1301.sync_flags |= PFSYNC_FLAG_SRCNODE;
5536 	if (st->nat_src_node)
5537 		sp->pfs_1301.sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5538 
5539 	sp->pfs_1301.id = st->id;
5540 	sp->pfs_1301.creatorid = st->creatorid;
5541 	pf_state_peer_hton(&st->src, &sp->pfs_1301.src);
5542 	pf_state_peer_hton(&st->dst, &sp->pfs_1301.dst);
5543 
5544 	if (st->rule.ptr == NULL)
5545 		sp->pfs_1301.rule = htonl(-1);
5546 	else
5547 		sp->pfs_1301.rule = htonl(st->rule.ptr->nr);
5548 	if (st->anchor.ptr == NULL)
5549 		sp->pfs_1301.anchor = htonl(-1);
5550 	else
5551 		sp->pfs_1301.anchor = htonl(st->anchor.ptr->nr);
5552 	if (st->nat_rule.ptr == NULL)
5553 		sp->pfs_1301.nat_rule = htonl(-1);
5554 	else
5555 		sp->pfs_1301.nat_rule = htonl(st->nat_rule.ptr->nr);
5556 
5557 	pf_state_counter_hton(st->packets[0], sp->pfs_1301.packets[0]);
5558 	pf_state_counter_hton(st->packets[1], sp->pfs_1301.packets[1]);
5559 	pf_state_counter_hton(st->bytes[0], sp->pfs_1301.bytes[0]);
5560 	pf_state_counter_hton(st->bytes[1], sp->pfs_1301.bytes[1]);
5561 }
5562 
5563 void
5564 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st)
5565 {
5566 	bzero(sp, sizeof(*sp));
5567 
5568 	sp->version = PF_STATE_VERSION;
5569 
5570 	/* copy from state key */
5571 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5572 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5573 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5574 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5575 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5576 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5577 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5578 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5579 	sp->proto = st->key[PF_SK_WIRE]->proto;
5580 	sp->af = st->key[PF_SK_WIRE]->af;
5581 
5582 	/* copy from state */
5583 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
5584 	strlcpy(sp->orig_ifname, st->orig_kif->pfik_name,
5585 	    sizeof(sp->orig_ifname));
5586 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
5587 	sp->creation = htonl(time_uptime - (st->creation / 1000));
5588 	sp->expire = pf_state_expires(st);
5589 	if (sp->expire <= time_uptime)
5590 		sp->expire = htonl(0);
5591 	else
5592 		sp->expire = htonl(sp->expire - time_uptime);
5593 
5594 	sp->direction = st->direction;
5595 	sp->log = st->act.log;
5596 	sp->timeout = st->timeout;
5597 	/* 8 bits for the old libpfctl, 16 bits for the new libpfctl */
5598 	sp->state_flags_compat = st->state_flags;
5599 	sp->state_flags = htons(st->state_flags);
5600 	if (st->src_node)
5601 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
5602 	if (st->nat_src_node)
5603 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5604 
5605 	sp->id = st->id;
5606 	sp->creatorid = st->creatorid;
5607 	pf_state_peer_hton(&st->src, &sp->src);
5608 	pf_state_peer_hton(&st->dst, &sp->dst);
5609 
5610 	if (st->rule.ptr == NULL)
5611 		sp->rule = htonl(-1);
5612 	else
5613 		sp->rule = htonl(st->rule.ptr->nr);
5614 	if (st->anchor.ptr == NULL)
5615 		sp->anchor = htonl(-1);
5616 	else
5617 		sp->anchor = htonl(st->anchor.ptr->nr);
5618 	if (st->nat_rule.ptr == NULL)
5619 		sp->nat_rule = htonl(-1);
5620 	else
5621 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
5622 
5623 	sp->packets[0] = st->packets[0];
5624 	sp->packets[1] = st->packets[1];
5625 	sp->bytes[0] = st->bytes[0];
5626 	sp->bytes[1] = st->bytes[1];
5627 
5628 	sp->qid = htons(st->act.qid);
5629 	sp->pqid = htons(st->act.pqid);
5630 	sp->dnpipe = htons(st->act.dnpipe);
5631 	sp->dnrpipe = htons(st->act.dnrpipe);
5632 	sp->rtableid = htonl(st->act.rtableid);
5633 	sp->min_ttl = st->act.min_ttl;
5634 	sp->set_tos = st->act.set_tos;
5635 	sp->max_mss = htons(st->act.max_mss);
5636 	sp->rt = st->rt;
5637 	if (st->rt_kif)
5638 		strlcpy(sp->rt_ifname, st->rt_kif->pfik_name,
5639 		    sizeof(sp->rt_ifname));
5640 	sp->set_prio[0] = st->act.set_prio[0];
5641 	sp->set_prio[1] = st->act.set_prio[1];
5642 
5643 }
5644 
5645 static void
5646 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
5647 {
5648 	struct pfr_ktable *kt;
5649 
5650 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
5651 
5652 	kt = aw->p.tbl;
5653 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
5654 		kt = kt->pfrkt_root;
5655 	aw->p.tbl = NULL;
5656 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
5657 		kt->pfrkt_cnt : -1;
5658 }
5659 
5660 static int
5661 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters,
5662     size_t number, char **names)
5663 {
5664 	nvlist_t        *nvc;
5665 
5666 	nvc = nvlist_create(0);
5667 	if (nvc == NULL)
5668 		return (ENOMEM);
5669 
5670 	for (int i = 0; i < number; i++) {
5671 		nvlist_append_number_array(nvc, "counters",
5672 		    counter_u64_fetch(counters[i]));
5673 		nvlist_append_string_array(nvc, "names",
5674 		    names[i]);
5675 		nvlist_append_number_array(nvc, "ids",
5676 		    i);
5677 	}
5678 	nvlist_add_nvlist(nvl, name, nvc);
5679 	nvlist_destroy(nvc);
5680 
5681 	return (0);
5682 }
5683 
5684 static int
5685 pf_getstatus(struct pfioc_nv *nv)
5686 {
5687 	nvlist_t        *nvl = NULL, *nvc = NULL;
5688 	void            *nvlpacked = NULL;
5689 	int              error;
5690 	struct pf_status s;
5691 	char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES;
5692 	char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES;
5693 	char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES;
5694 	PF_RULES_RLOCK_TRACKER;
5695 
5696 #define ERROUT(x)      ERROUT_FUNCTION(errout, x)
5697 
5698 	PF_RULES_RLOCK();
5699 
5700 	nvl = nvlist_create(0);
5701 	if (nvl == NULL)
5702 		ERROUT(ENOMEM);
5703 
5704 	nvlist_add_bool(nvl, "running", V_pf_status.running);
5705 	nvlist_add_number(nvl, "since", V_pf_status.since);
5706 	nvlist_add_number(nvl, "debug", V_pf_status.debug);
5707 	nvlist_add_number(nvl, "hostid", V_pf_status.hostid);
5708 	nvlist_add_number(nvl, "states", V_pf_status.states);
5709 	nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes);
5710 	nvlist_add_number(nvl, "reass", V_pf_status.reass);
5711 	nvlist_add_bool(nvl, "syncookies_active",
5712 	    V_pf_status.syncookies_active);
5713 	nvlist_add_number(nvl, "halfopen_states", V_pf_status.states_halfopen);
5714 
5715 	/* counters */
5716 	error = pf_add_status_counters(nvl, "counters", V_pf_status.counters,
5717 	    PFRES_MAX, pf_reasons);
5718 	if (error != 0)
5719 		ERROUT(error);
5720 
5721 	/* lcounters */
5722 	error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters,
5723 	    KLCNT_MAX, pf_lcounter);
5724 	if (error != 0)
5725 		ERROUT(error);
5726 
5727 	/* fcounters */
5728 	nvc = nvlist_create(0);
5729 	if (nvc == NULL)
5730 		ERROUT(ENOMEM);
5731 
5732 	for (int i = 0; i < FCNT_MAX; i++) {
5733 		nvlist_append_number_array(nvc, "counters",
5734 		    pf_counter_u64_fetch(&V_pf_status.fcounters[i]));
5735 		nvlist_append_string_array(nvc, "names",
5736 		    pf_fcounter[i]);
5737 		nvlist_append_number_array(nvc, "ids",
5738 		    i);
5739 	}
5740 	nvlist_add_nvlist(nvl, "fcounters", nvc);
5741 	nvlist_destroy(nvc);
5742 	nvc = NULL;
5743 
5744 	/* scounters */
5745 	error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters,
5746 	    SCNT_MAX, pf_fcounter);
5747 	if (error != 0)
5748 		ERROUT(error);
5749 
5750 	nvlist_add_string(nvl, "ifname", V_pf_status.ifname);
5751 	nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum,
5752 	    PF_MD5_DIGEST_LENGTH);
5753 
5754 	pfi_update_status(V_pf_status.ifname, &s);
5755 
5756 	/* pcounters / bcounters */
5757 	for (int i = 0; i < 2; i++) {
5758 		for (int j = 0; j < 2; j++) {
5759 			for (int k = 0; k < 2; k++) {
5760 				nvlist_append_number_array(nvl, "pcounters",
5761 				    s.pcounters[i][j][k]);
5762 			}
5763 			nvlist_append_number_array(nvl, "bcounters",
5764 			    s.bcounters[i][j]);
5765 		}
5766 	}
5767 
5768 	nvlpacked = nvlist_pack(nvl, &nv->len);
5769 	if (nvlpacked == NULL)
5770 		ERROUT(ENOMEM);
5771 
5772 	if (nv->size == 0)
5773 		ERROUT(0);
5774 	else if (nv->size < nv->len)
5775 		ERROUT(ENOSPC);
5776 
5777 	PF_RULES_RUNLOCK();
5778 	error = copyout(nvlpacked, nv->data, nv->len);
5779 	goto done;
5780 
5781 #undef ERROUT
5782 errout:
5783 	PF_RULES_RUNLOCK();
5784 done:
5785 	free(nvlpacked, M_NVLIST);
5786 	nvlist_destroy(nvc);
5787 	nvlist_destroy(nvl);
5788 
5789 	return (error);
5790 }
5791 
5792 /*
5793  * XXX - Check for version mismatch!!!
5794  */
5795 static void
5796 pf_clear_all_states(void)
5797 {
5798 	struct epoch_tracker	 et;
5799 	struct pf_kstate	*s;
5800 	u_int i;
5801 
5802 	NET_EPOCH_ENTER(et);
5803 	for (i = 0; i <= pf_hashmask; i++) {
5804 		struct pf_idhash *ih = &V_pf_idhash[i];
5805 relock:
5806 		PF_HASHROW_LOCK(ih);
5807 		LIST_FOREACH(s, &ih->states, entry) {
5808 			s->timeout = PFTM_PURGE;
5809 			/* Don't send out individual delete messages. */
5810 			s->state_flags |= PFSTATE_NOSYNC;
5811 			pf_unlink_state(s);
5812 			goto relock;
5813 		}
5814 		PF_HASHROW_UNLOCK(ih);
5815 	}
5816 	NET_EPOCH_EXIT(et);
5817 }
5818 
5819 static int
5820 pf_clear_tables(void)
5821 {
5822 	struct pfioc_table io;
5823 	int error;
5824 
5825 	bzero(&io, sizeof(io));
5826 	io.pfrio_flags |= PFR_FLAG_ALLRSETS;
5827 
5828 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
5829 	    io.pfrio_flags);
5830 
5831 	return (error);
5832 }
5833 
5834 static void
5835 pf_clear_srcnodes(struct pf_ksrc_node *n)
5836 {
5837 	struct pf_kstate *s;
5838 	int i;
5839 
5840 	for (i = 0; i <= pf_hashmask; i++) {
5841 		struct pf_idhash *ih = &V_pf_idhash[i];
5842 
5843 		PF_HASHROW_LOCK(ih);
5844 		LIST_FOREACH(s, &ih->states, entry) {
5845 			if (n == NULL || n == s->src_node)
5846 				s->src_node = NULL;
5847 			if (n == NULL || n == s->nat_src_node)
5848 				s->nat_src_node = NULL;
5849 		}
5850 		PF_HASHROW_UNLOCK(ih);
5851 	}
5852 
5853 	if (n == NULL) {
5854 		struct pf_srchash *sh;
5855 
5856 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5857 		    i++, sh++) {
5858 			PF_HASHROW_LOCK(sh);
5859 			LIST_FOREACH(n, &sh->nodes, entry) {
5860 				n->expire = 1;
5861 				n->states = 0;
5862 			}
5863 			PF_HASHROW_UNLOCK(sh);
5864 		}
5865 	} else {
5866 		/* XXX: hash slot should already be locked here. */
5867 		n->expire = 1;
5868 		n->states = 0;
5869 	}
5870 }
5871 
5872 static void
5873 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
5874 {
5875 	struct pf_ksrc_node_list	 kill;
5876 
5877 	LIST_INIT(&kill);
5878 	for (int i = 0; i <= pf_srchashmask; i++) {
5879 		struct pf_srchash *sh = &V_pf_srchash[i];
5880 		struct pf_ksrc_node *sn, *tmp;
5881 
5882 		PF_HASHROW_LOCK(sh);
5883 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
5884 			if (PF_MATCHA(psnk->psnk_src.neg,
5885 			      &psnk->psnk_src.addr.v.a.addr,
5886 			      &psnk->psnk_src.addr.v.a.mask,
5887 			      &sn->addr, sn->af) &&
5888 			    PF_MATCHA(psnk->psnk_dst.neg,
5889 			      &psnk->psnk_dst.addr.v.a.addr,
5890 			      &psnk->psnk_dst.addr.v.a.mask,
5891 			      &sn->raddr, sn->af)) {
5892 				pf_unlink_src_node(sn);
5893 				LIST_INSERT_HEAD(&kill, sn, entry);
5894 				sn->expire = 1;
5895 			}
5896 		PF_HASHROW_UNLOCK(sh);
5897 	}
5898 
5899 	for (int i = 0; i <= pf_hashmask; i++) {
5900 		struct pf_idhash *ih = &V_pf_idhash[i];
5901 		struct pf_kstate *s;
5902 
5903 		PF_HASHROW_LOCK(ih);
5904 		LIST_FOREACH(s, &ih->states, entry) {
5905 			if (s->src_node && s->src_node->expire == 1)
5906 				s->src_node = NULL;
5907 			if (s->nat_src_node && s->nat_src_node->expire == 1)
5908 				s->nat_src_node = NULL;
5909 		}
5910 		PF_HASHROW_UNLOCK(ih);
5911 	}
5912 
5913 	psnk->psnk_killed = pf_free_src_nodes(&kill);
5914 }
5915 
5916 static int
5917 pf_keepcounters(struct pfioc_nv *nv)
5918 {
5919 	nvlist_t	*nvl = NULL;
5920 	void		*nvlpacked = NULL;
5921 	int		 error = 0;
5922 
5923 #define	ERROUT(x)	ERROUT_FUNCTION(on_error, x)
5924 
5925 	if (nv->len > pf_ioctl_maxcount)
5926 		ERROUT(ENOMEM);
5927 
5928 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
5929 	if (nvlpacked == NULL)
5930 		ERROUT(ENOMEM);
5931 
5932 	error = copyin(nv->data, nvlpacked, nv->len);
5933 	if (error)
5934 		ERROUT(error);
5935 
5936 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
5937 	if (nvl == NULL)
5938 		ERROUT(EBADMSG);
5939 
5940 	if (! nvlist_exists_bool(nvl, "keep_counters"))
5941 		ERROUT(EBADMSG);
5942 
5943 	V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters");
5944 
5945 on_error:
5946 	nvlist_destroy(nvl);
5947 	free(nvlpacked, M_NVLIST);
5948 	return (error);
5949 }
5950 
5951 unsigned int
5952 pf_clear_states(const struct pf_kstate_kill *kill)
5953 {
5954 	struct pf_state_key_cmp	 match_key;
5955 	struct pf_kstate	*s;
5956 	struct pfi_kkif	*kif;
5957 	int		 idx;
5958 	unsigned int	 killed = 0, dir;
5959 
5960 	NET_EPOCH_ASSERT();
5961 
5962 	for (unsigned int i = 0; i <= pf_hashmask; i++) {
5963 		struct pf_idhash *ih = &V_pf_idhash[i];
5964 
5965 relock_DIOCCLRSTATES:
5966 		PF_HASHROW_LOCK(ih);
5967 		LIST_FOREACH(s, &ih->states, entry) {
5968 			/* For floating states look at the original kif. */
5969 			kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
5970 
5971 			if (kill->psk_ifname[0] &&
5972 			    strcmp(kill->psk_ifname,
5973 			    kif->pfik_name))
5974 				continue;
5975 
5976 			if (kill->psk_kill_match) {
5977 				bzero(&match_key, sizeof(match_key));
5978 
5979 				if (s->direction == PF_OUT) {
5980 					dir = PF_IN;
5981 					idx = PF_SK_STACK;
5982 				} else {
5983 					dir = PF_OUT;
5984 					idx = PF_SK_WIRE;
5985 				}
5986 
5987 				match_key.af = s->key[idx]->af;
5988 				match_key.proto = s->key[idx]->proto;
5989 				PF_ACPY(&match_key.addr[0],
5990 				    &s->key[idx]->addr[1], match_key.af);
5991 				match_key.port[0] = s->key[idx]->port[1];
5992 				PF_ACPY(&match_key.addr[1],
5993 				    &s->key[idx]->addr[0], match_key.af);
5994 				match_key.port[1] = s->key[idx]->port[0];
5995 			}
5996 
5997 			/*
5998 			 * Don't send out individual
5999 			 * delete messages.
6000 			 */
6001 			s->state_flags |= PFSTATE_NOSYNC;
6002 			pf_unlink_state(s);
6003 			killed++;
6004 
6005 			if (kill->psk_kill_match)
6006 				killed += pf_kill_matching_state(&match_key,
6007 				    dir);
6008 
6009 			goto relock_DIOCCLRSTATES;
6010 		}
6011 		PF_HASHROW_UNLOCK(ih);
6012 	}
6013 
6014 	if (V_pfsync_clear_states_ptr != NULL)
6015 		V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname);
6016 
6017 	return (killed);
6018 }
6019 
6020 void
6021 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed)
6022 {
6023 	struct pf_kstate	*s;
6024 
6025 	NET_EPOCH_ASSERT();
6026 	if (kill->psk_pfcmp.id) {
6027 		if (kill->psk_pfcmp.creatorid == 0)
6028 			kill->psk_pfcmp.creatorid = V_pf_status.hostid;
6029 		if ((s = pf_find_state_byid(kill->psk_pfcmp.id,
6030 		    kill->psk_pfcmp.creatorid))) {
6031 			pf_unlink_state(s);
6032 			*killed = 1;
6033 		}
6034 		return;
6035 	}
6036 
6037 	for (unsigned int i = 0; i <= pf_hashmask; i++)
6038 		*killed += pf_killstates_row(kill, &V_pf_idhash[i]);
6039 }
6040 
6041 static int
6042 pf_killstates_nv(struct pfioc_nv *nv)
6043 {
6044 	struct pf_kstate_kill	 kill;
6045 	struct epoch_tracker	 et;
6046 	nvlist_t		*nvl = NULL;
6047 	void			*nvlpacked = NULL;
6048 	int			 error = 0;
6049 	unsigned int		 killed = 0;
6050 
6051 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
6052 
6053 	if (nv->len > pf_ioctl_maxcount)
6054 		ERROUT(ENOMEM);
6055 
6056 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6057 	if (nvlpacked == NULL)
6058 		ERROUT(ENOMEM);
6059 
6060 	error = copyin(nv->data, nvlpacked, nv->len);
6061 	if (error)
6062 		ERROUT(error);
6063 
6064 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6065 	if (nvl == NULL)
6066 		ERROUT(EBADMSG);
6067 
6068 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6069 	if (error)
6070 		ERROUT(error);
6071 
6072 	NET_EPOCH_ENTER(et);
6073 	pf_killstates(&kill, &killed);
6074 	NET_EPOCH_EXIT(et);
6075 
6076 	free(nvlpacked, M_NVLIST);
6077 	nvlpacked = NULL;
6078 	nvlist_destroy(nvl);
6079 	nvl = nvlist_create(0);
6080 	if (nvl == NULL)
6081 		ERROUT(ENOMEM);
6082 
6083 	nvlist_add_number(nvl, "killed", killed);
6084 
6085 	nvlpacked = nvlist_pack(nvl, &nv->len);
6086 	if (nvlpacked == NULL)
6087 		ERROUT(ENOMEM);
6088 
6089 	if (nv->size == 0)
6090 		ERROUT(0);
6091 	else if (nv->size < nv->len)
6092 		ERROUT(ENOSPC);
6093 
6094 	error = copyout(nvlpacked, nv->data, nv->len);
6095 
6096 on_error:
6097 	nvlist_destroy(nvl);
6098 	free(nvlpacked, M_NVLIST);
6099 	return (error);
6100 }
6101 
6102 static int
6103 pf_clearstates_nv(struct pfioc_nv *nv)
6104 {
6105 	struct pf_kstate_kill	 kill;
6106 	struct epoch_tracker	 et;
6107 	nvlist_t		*nvl = NULL;
6108 	void			*nvlpacked = NULL;
6109 	int			 error = 0;
6110 	unsigned int		 killed;
6111 
6112 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
6113 
6114 	if (nv->len > pf_ioctl_maxcount)
6115 		ERROUT(ENOMEM);
6116 
6117 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6118 	if (nvlpacked == NULL)
6119 		ERROUT(ENOMEM);
6120 
6121 	error = copyin(nv->data, nvlpacked, nv->len);
6122 	if (error)
6123 		ERROUT(error);
6124 
6125 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6126 	if (nvl == NULL)
6127 		ERROUT(EBADMSG);
6128 
6129 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6130 	if (error)
6131 		ERROUT(error);
6132 
6133 	NET_EPOCH_ENTER(et);
6134 	killed = pf_clear_states(&kill);
6135 	NET_EPOCH_EXIT(et);
6136 
6137 	free(nvlpacked, M_NVLIST);
6138 	nvlpacked = NULL;
6139 	nvlist_destroy(nvl);
6140 	nvl = nvlist_create(0);
6141 	if (nvl == NULL)
6142 		ERROUT(ENOMEM);
6143 
6144 	nvlist_add_number(nvl, "killed", killed);
6145 
6146 	nvlpacked = nvlist_pack(nvl, &nv->len);
6147 	if (nvlpacked == NULL)
6148 		ERROUT(ENOMEM);
6149 
6150 	if (nv->size == 0)
6151 		ERROUT(0);
6152 	else if (nv->size < nv->len)
6153 		ERROUT(ENOSPC);
6154 
6155 	error = copyout(nvlpacked, nv->data, nv->len);
6156 
6157 #undef ERROUT
6158 on_error:
6159 	nvlist_destroy(nvl);
6160 	free(nvlpacked, M_NVLIST);
6161 	return (error);
6162 }
6163 
6164 static int
6165 pf_getstate(struct pfioc_nv *nv)
6166 {
6167 	nvlist_t		*nvl = NULL, *nvls;
6168 	void			*nvlpacked = NULL;
6169 	struct pf_kstate	*s = NULL;
6170 	int			 error = 0;
6171 	uint64_t		 id, creatorid;
6172 
6173 #define ERROUT(x)	ERROUT_FUNCTION(errout, x)
6174 
6175 	if (nv->len > pf_ioctl_maxcount)
6176 		ERROUT(ENOMEM);
6177 
6178 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6179 	if (nvlpacked == NULL)
6180 		ERROUT(ENOMEM);
6181 
6182 	error = copyin(nv->data, nvlpacked, nv->len);
6183 	if (error)
6184 		ERROUT(error);
6185 
6186 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6187 	if (nvl == NULL)
6188 		ERROUT(EBADMSG);
6189 
6190 	PFNV_CHK(pf_nvuint64(nvl, "id", &id));
6191 	PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid));
6192 
6193 	s = pf_find_state_byid(id, creatorid);
6194 	if (s == NULL)
6195 		ERROUT(ENOENT);
6196 
6197 	free(nvlpacked, M_NVLIST);
6198 	nvlpacked = NULL;
6199 	nvlist_destroy(nvl);
6200 	nvl = nvlist_create(0);
6201 	if (nvl == NULL)
6202 		ERROUT(ENOMEM);
6203 
6204 	nvls = pf_state_to_nvstate(s);
6205 	if (nvls == NULL)
6206 		ERROUT(ENOMEM);
6207 
6208 	nvlist_add_nvlist(nvl, "state", nvls);
6209 	nvlist_destroy(nvls);
6210 
6211 	nvlpacked = nvlist_pack(nvl, &nv->len);
6212 	if (nvlpacked == NULL)
6213 		ERROUT(ENOMEM);
6214 
6215 	if (nv->size == 0)
6216 		ERROUT(0);
6217 	else if (nv->size < nv->len)
6218 		ERROUT(ENOSPC);
6219 
6220 	error = copyout(nvlpacked, nv->data, nv->len);
6221 
6222 #undef ERROUT
6223 errout:
6224 	if (s != NULL)
6225 		PF_STATE_UNLOCK(s);
6226 	free(nvlpacked, M_NVLIST);
6227 	nvlist_destroy(nvl);
6228 	return (error);
6229 }
6230 
6231 /*
6232  * XXX - Check for version mismatch!!!
6233  */
6234 
6235 /*
6236  * Duplicate pfctl -Fa operation to get rid of as much as we can.
6237  */
6238 static int
6239 shutdown_pf(void)
6240 {
6241 	int error = 0;
6242 	u_int32_t t[5];
6243 	char nn = '\0';
6244 	struct pf_kanchor *anchor;
6245 	struct pf_keth_anchor *eth_anchor;
6246 	int rs_num;
6247 
6248 	do {
6249 		/* Unlink rules of all user defined anchors */
6250 		RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) {
6251 			/* Wildcard based anchors may not have a respective
6252 			 * explicit anchor rule or they may be left empty
6253 			 * without rules. It leads to anchor.refcnt=0, and the
6254 			 * rest of the logic does not expect it. */
6255 			if (anchor->refcnt == 0)
6256 				anchor->refcnt = 1;
6257 			for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6258 				if ((error = pf_begin_rules(&t[rs_num], rs_num,
6259 				    anchor->path)) != 0) {
6260 					DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: "
6261 					    "anchor.path=%s rs_num=%d\n",
6262 					    anchor->path, rs_num));
6263 					goto error;	/* XXX: rollback? */
6264 				}
6265 			}
6266 			for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6267 				error = pf_commit_rules(t[rs_num], rs_num,
6268 				    anchor->path);
6269 				MPASS(error == 0);
6270 			}
6271 		}
6272 
6273 		/* Unlink rules of all user defined ether anchors */
6274 		RB_FOREACH(eth_anchor, pf_keth_anchor_global,
6275 		    &V_pf_keth_anchors) {
6276 			/* Wildcard based anchors may not have a respective
6277 			 * explicit anchor rule or they may be left empty
6278 			 * without rules. It leads to anchor.refcnt=0, and the
6279 			 * rest of the logic does not expect it. */
6280 			if (eth_anchor->refcnt == 0)
6281 				eth_anchor->refcnt = 1;
6282 			if ((error = pf_begin_eth(&t[0], eth_anchor->path))
6283 			    != 0) {
6284 				DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth "
6285 				    "anchor.path=%s\n", eth_anchor->path));
6286 				goto error;
6287 			}
6288 			error = pf_commit_eth(t[0], eth_anchor->path);
6289 			MPASS(error == 0);
6290 		}
6291 
6292 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
6293 		    != 0) {
6294 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
6295 			break;
6296 		}
6297 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
6298 		    != 0) {
6299 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
6300 			break;		/* XXX: rollback? */
6301 		}
6302 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
6303 		    != 0) {
6304 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
6305 			break;		/* XXX: rollback? */
6306 		}
6307 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
6308 		    != 0) {
6309 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
6310 			break;		/* XXX: rollback? */
6311 		}
6312 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
6313 		    != 0) {
6314 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
6315 			break;		/* XXX: rollback? */
6316 		}
6317 
6318 		error = pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
6319 		MPASS(error == 0);
6320 		error = pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
6321 		MPASS(error == 0);
6322 		error = pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
6323 		MPASS(error == 0);
6324 		error = pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
6325 		MPASS(error == 0);
6326 		error = pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
6327 		MPASS(error == 0);
6328 
6329 		if ((error = pf_clear_tables()) != 0)
6330 			break;
6331 
6332 		if ((error = pf_begin_eth(&t[0], &nn)) != 0) {
6333 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth\n"));
6334 			break;
6335 		}
6336 		error = pf_commit_eth(t[0], &nn);
6337 		MPASS(error == 0);
6338 
6339 #ifdef ALTQ
6340 		if ((error = pf_begin_altq(&t[0])) != 0) {
6341 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
6342 			break;
6343 		}
6344 		pf_commit_altq(t[0]);
6345 #endif
6346 
6347 		pf_clear_all_states();
6348 
6349 		pf_clear_srcnodes(NULL);
6350 
6351 		/* status does not use malloced mem so no need to cleanup */
6352 		/* fingerprints and interfaces have their own cleanup code */
6353 	} while(0);
6354 
6355 error:
6356 	return (error);
6357 }
6358 
6359 static pfil_return_t
6360 pf_check_return(int chk, struct mbuf **m)
6361 {
6362 
6363 	switch (chk) {
6364 	case PF_PASS:
6365 		if (*m == NULL)
6366 			return (PFIL_CONSUMED);
6367 		else
6368 			return (PFIL_PASS);
6369 		break;
6370 	default:
6371 		if (*m != NULL) {
6372 			m_freem(*m);
6373 			*m = NULL;
6374 		}
6375 		return (PFIL_DROPPED);
6376 	}
6377 }
6378 
6379 static pfil_return_t
6380 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6381     void *ruleset __unused, struct inpcb *inp)
6382 {
6383 	int chk;
6384 
6385 	chk = pf_test_eth(PF_IN, flags, ifp, m, inp);
6386 
6387 	return (pf_check_return(chk, m));
6388 }
6389 
6390 static pfil_return_t
6391 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6392     void *ruleset __unused, struct inpcb *inp)
6393 {
6394 	int chk;
6395 
6396 	chk = pf_test_eth(PF_OUT, flags, ifp, m, inp);
6397 
6398 	return (pf_check_return(chk, m));
6399 }
6400 
6401 #ifdef INET
6402 static pfil_return_t
6403 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6404     void *ruleset __unused, struct inpcb *inp)
6405 {
6406 	int chk;
6407 
6408 	chk = pf_test(PF_IN, flags, ifp, m, inp, NULL);
6409 
6410 	return (pf_check_return(chk, m));
6411 }
6412 
6413 static pfil_return_t
6414 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6415     void *ruleset __unused,  struct inpcb *inp)
6416 {
6417 	int chk;
6418 
6419 	chk = pf_test(PF_OUT, flags, ifp, m, inp, NULL);
6420 
6421 	return (pf_check_return(chk, m));
6422 }
6423 #endif
6424 
6425 #ifdef INET6
6426 static pfil_return_t
6427 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
6428     void *ruleset __unused,  struct inpcb *inp)
6429 {
6430 	int chk;
6431 
6432 	/*
6433 	 * In case of loopback traffic IPv6 uses the real interface in
6434 	 * order to support scoped addresses. In order to support stateful
6435 	 * filtering we have change this to lo0 as it is the case in IPv4.
6436 	 */
6437 	CURVNET_SET(ifp->if_vnet);
6438 	chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp,
6439 	    m, inp, NULL);
6440 	CURVNET_RESTORE();
6441 
6442 	return (pf_check_return(chk, m));
6443 }
6444 
6445 static pfil_return_t
6446 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
6447     void *ruleset __unused,  struct inpcb *inp)
6448 {
6449 	int chk;
6450 
6451 	CURVNET_SET(ifp->if_vnet);
6452 	chk = pf_test6(PF_OUT, flags, ifp, m, inp, NULL);
6453 	CURVNET_RESTORE();
6454 
6455 	return (pf_check_return(chk, m));
6456 }
6457 #endif /* INET6 */
6458 
6459 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook);
6460 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook);
6461 #define	V_pf_eth_in_hook	VNET(pf_eth_in_hook)
6462 #define	V_pf_eth_out_hook	VNET(pf_eth_out_hook)
6463 
6464 #ifdef INET
6465 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
6466 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
6467 #define	V_pf_ip4_in_hook	VNET(pf_ip4_in_hook)
6468 #define	V_pf_ip4_out_hook	VNET(pf_ip4_out_hook)
6469 #endif
6470 #ifdef INET6
6471 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
6472 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
6473 #define	V_pf_ip6_in_hook	VNET(pf_ip6_in_hook)
6474 #define	V_pf_ip6_out_hook	VNET(pf_ip6_out_hook)
6475 #endif
6476 
6477 static void
6478 hook_pf_eth(void)
6479 {
6480 	struct pfil_hook_args pha = {
6481 		.pa_version = PFIL_VERSION,
6482 		.pa_modname = "pf",
6483 		.pa_type = PFIL_TYPE_ETHERNET,
6484 	};
6485 	struct pfil_link_args pla = {
6486 		.pa_version = PFIL_VERSION,
6487 	};
6488 	int ret __diagused;
6489 
6490 	if (atomic_load_bool(&V_pf_pfil_eth_hooked))
6491 		return;
6492 
6493 	pha.pa_mbuf_chk = pf_eth_check_in;
6494 	pha.pa_flags = PFIL_IN;
6495 	pha.pa_rulname = "eth-in";
6496 	V_pf_eth_in_hook = pfil_add_hook(&pha);
6497 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6498 	pla.pa_head = V_link_pfil_head;
6499 	pla.pa_hook = V_pf_eth_in_hook;
6500 	ret = pfil_link(&pla);
6501 	MPASS(ret == 0);
6502 	pha.pa_mbuf_chk = pf_eth_check_out;
6503 	pha.pa_flags = PFIL_OUT;
6504 	pha.pa_rulname = "eth-out";
6505 	V_pf_eth_out_hook = pfil_add_hook(&pha);
6506 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6507 	pla.pa_head = V_link_pfil_head;
6508 	pla.pa_hook = V_pf_eth_out_hook;
6509 	ret = pfil_link(&pla);
6510 	MPASS(ret == 0);
6511 
6512 	atomic_store_bool(&V_pf_pfil_eth_hooked, true);
6513 }
6514 
6515 static void
6516 hook_pf(void)
6517 {
6518 	struct pfil_hook_args pha = {
6519 		.pa_version = PFIL_VERSION,
6520 		.pa_modname = "pf",
6521 	};
6522 	struct pfil_link_args pla = {
6523 		.pa_version = PFIL_VERSION,
6524 	};
6525 	int ret __diagused;
6526 
6527 	if (atomic_load_bool(&V_pf_pfil_hooked))
6528 		return;
6529 
6530 #ifdef INET
6531 	pha.pa_type = PFIL_TYPE_IP4;
6532 	pha.pa_mbuf_chk = pf_check_in;
6533 	pha.pa_flags = PFIL_IN;
6534 	pha.pa_rulname = "default-in";
6535 	V_pf_ip4_in_hook = pfil_add_hook(&pha);
6536 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6537 	pla.pa_head = V_inet_pfil_head;
6538 	pla.pa_hook = V_pf_ip4_in_hook;
6539 	ret = pfil_link(&pla);
6540 	MPASS(ret == 0);
6541 	pha.pa_mbuf_chk = pf_check_out;
6542 	pha.pa_flags = PFIL_OUT;
6543 	pha.pa_rulname = "default-out";
6544 	V_pf_ip4_out_hook = pfil_add_hook(&pha);
6545 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6546 	pla.pa_head = V_inet_pfil_head;
6547 	pla.pa_hook = V_pf_ip4_out_hook;
6548 	ret = pfil_link(&pla);
6549 	MPASS(ret == 0);
6550 	if (V_pf_filter_local) {
6551 		pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6552 		pla.pa_head = V_inet_local_pfil_head;
6553 		pla.pa_hook = V_pf_ip4_out_hook;
6554 		ret = pfil_link(&pla);
6555 		MPASS(ret == 0);
6556 	}
6557 #endif
6558 #ifdef INET6
6559 	pha.pa_type = PFIL_TYPE_IP6;
6560 	pha.pa_mbuf_chk = pf_check6_in;
6561 	pha.pa_flags = PFIL_IN;
6562 	pha.pa_rulname = "default-in6";
6563 	V_pf_ip6_in_hook = pfil_add_hook(&pha);
6564 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6565 	pla.pa_head = V_inet6_pfil_head;
6566 	pla.pa_hook = V_pf_ip6_in_hook;
6567 	ret = pfil_link(&pla);
6568 	MPASS(ret == 0);
6569 	pha.pa_mbuf_chk = pf_check6_out;
6570 	pha.pa_rulname = "default-out6";
6571 	pha.pa_flags = PFIL_OUT;
6572 	V_pf_ip6_out_hook = pfil_add_hook(&pha);
6573 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6574 	pla.pa_head = V_inet6_pfil_head;
6575 	pla.pa_hook = V_pf_ip6_out_hook;
6576 	ret = pfil_link(&pla);
6577 	MPASS(ret == 0);
6578 	if (V_pf_filter_local) {
6579 		pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6580 		pla.pa_head = V_inet6_local_pfil_head;
6581 		pla.pa_hook = V_pf_ip6_out_hook;
6582 		ret = pfil_link(&pla);
6583 		MPASS(ret == 0);
6584 	}
6585 #endif
6586 
6587 	atomic_store_bool(&V_pf_pfil_hooked, true);
6588 }
6589 
6590 static void
6591 dehook_pf_eth(void)
6592 {
6593 
6594 	if (!atomic_load_bool(&V_pf_pfil_eth_hooked))
6595 		return;
6596 
6597 	pfil_remove_hook(V_pf_eth_in_hook);
6598 	pfil_remove_hook(V_pf_eth_out_hook);
6599 
6600 	atomic_store_bool(&V_pf_pfil_eth_hooked, false);
6601 }
6602 
6603 static void
6604 dehook_pf(void)
6605 {
6606 
6607 	if (!atomic_load_bool(&V_pf_pfil_hooked))
6608 		return;
6609 
6610 #ifdef INET
6611 	pfil_remove_hook(V_pf_ip4_in_hook);
6612 	pfil_remove_hook(V_pf_ip4_out_hook);
6613 #endif
6614 #ifdef INET6
6615 	pfil_remove_hook(V_pf_ip6_in_hook);
6616 	pfil_remove_hook(V_pf_ip6_out_hook);
6617 #endif
6618 
6619 	atomic_store_bool(&V_pf_pfil_hooked, false);
6620 }
6621 
6622 static void
6623 pf_load_vnet(void)
6624 {
6625 	V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
6626 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
6627 
6628 	rm_init_flags(&V_pf_rules_lock, "pf rulesets", RM_RECURSE);
6629 	sx_init(&V_pf_ioctl_lock, "pf ioctl");
6630 
6631 	pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
6632 	    PF_RULE_TAG_HASH_SIZE_DEFAULT);
6633 #ifdef ALTQ
6634 	pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
6635 	    PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
6636 #endif
6637 
6638 	V_pf_keth = &V_pf_main_keth_anchor.ruleset;
6639 
6640 	pfattach_vnet();
6641 	V_pf_vnet_active = 1;
6642 }
6643 
6644 static int
6645 pf_load(void)
6646 {
6647 	int error;
6648 
6649 	sx_init(&pf_end_lock, "pf end thread");
6650 
6651 	pf_mtag_initialize();
6652 
6653 	pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
6654 	if (pf_dev == NULL)
6655 		return (ENOMEM);
6656 
6657 	pf_end_threads = 0;
6658 	error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
6659 	if (error != 0)
6660 		return (error);
6661 
6662 	pfi_initialize();
6663 
6664 	return (0);
6665 }
6666 
6667 static void
6668 pf_unload_vnet(void)
6669 {
6670 	int ret __diagused;
6671 
6672 	V_pf_vnet_active = 0;
6673 	V_pf_status.running = 0;
6674 	dehook_pf();
6675 	dehook_pf_eth();
6676 
6677 	PF_RULES_WLOCK();
6678 	pf_syncookies_cleanup();
6679 	shutdown_pf();
6680 	PF_RULES_WUNLOCK();
6681 
6682 	/* Make sure we've cleaned up ethernet rules before we continue. */
6683 	NET_EPOCH_DRAIN_CALLBACKS();
6684 
6685 	ret = swi_remove(V_pf_swi_cookie);
6686 	MPASS(ret == 0);
6687 	ret = intr_event_destroy(V_pf_swi_ie);
6688 	MPASS(ret == 0);
6689 
6690 	pf_unload_vnet_purge();
6691 
6692 	pf_normalize_cleanup();
6693 	PF_RULES_WLOCK();
6694 	pfi_cleanup_vnet();
6695 	PF_RULES_WUNLOCK();
6696 	pfr_cleanup();
6697 	pf_osfp_flush();
6698 	pf_cleanup();
6699 	if (IS_DEFAULT_VNET(curvnet))
6700 		pf_mtag_cleanup();
6701 
6702 	pf_cleanup_tagset(&V_pf_tags);
6703 #ifdef ALTQ
6704 	pf_cleanup_tagset(&V_pf_qids);
6705 #endif
6706 	uma_zdestroy(V_pf_tag_z);
6707 
6708 #ifdef PF_WANT_32_TO_64_COUNTER
6709 	PF_RULES_WLOCK();
6710 	LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist);
6711 
6712 	MPASS(LIST_EMPTY(&V_pf_allkiflist));
6713 	MPASS(V_pf_allkifcount == 0);
6714 
6715 	LIST_REMOVE(&V_pf_default_rule, allrulelist);
6716 	V_pf_allrulecount--;
6717 	LIST_REMOVE(V_pf_rulemarker, allrulelist);
6718 
6719 	MPASS(LIST_EMPTY(&V_pf_allrulelist));
6720 	MPASS(V_pf_allrulecount == 0);
6721 
6722 	PF_RULES_WUNLOCK();
6723 
6724 	free(V_pf_kifmarker, PFI_MTYPE);
6725 	free(V_pf_rulemarker, M_PFRULE);
6726 #endif
6727 
6728 	/* Free counters last as we updated them during shutdown. */
6729 	pf_counter_u64_deinit(&V_pf_default_rule.evaluations);
6730 	for (int i = 0; i < 2; i++) {
6731 		pf_counter_u64_deinit(&V_pf_default_rule.packets[i]);
6732 		pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]);
6733 	}
6734 	counter_u64_free(V_pf_default_rule.states_cur);
6735 	counter_u64_free(V_pf_default_rule.states_tot);
6736 	counter_u64_free(V_pf_default_rule.src_nodes);
6737 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, V_pf_default_rule.timestamp);
6738 
6739 	for (int i = 0; i < PFRES_MAX; i++)
6740 		counter_u64_free(V_pf_status.counters[i]);
6741 	for (int i = 0; i < KLCNT_MAX; i++)
6742 		counter_u64_free(V_pf_status.lcounters[i]);
6743 	for (int i = 0; i < FCNT_MAX; i++)
6744 		pf_counter_u64_deinit(&V_pf_status.fcounters[i]);
6745 	for (int i = 0; i < SCNT_MAX; i++)
6746 		counter_u64_free(V_pf_status.scounters[i]);
6747 
6748 	rm_destroy(&V_pf_rules_lock);
6749 	sx_destroy(&V_pf_ioctl_lock);
6750 }
6751 
6752 static void
6753 pf_unload(void)
6754 {
6755 
6756 	sx_xlock(&pf_end_lock);
6757 	pf_end_threads = 1;
6758 	while (pf_end_threads < 2) {
6759 		wakeup_one(pf_purge_thread);
6760 		sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
6761 	}
6762 	sx_xunlock(&pf_end_lock);
6763 
6764 	pf_nl_unregister();
6765 
6766 	if (pf_dev != NULL)
6767 		destroy_dev(pf_dev);
6768 
6769 	pfi_cleanup();
6770 
6771 	sx_destroy(&pf_end_lock);
6772 }
6773 
6774 static void
6775 vnet_pf_init(void *unused __unused)
6776 {
6777 
6778 	pf_load_vnet();
6779 }
6780 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6781     vnet_pf_init, NULL);
6782 
6783 static void
6784 vnet_pf_uninit(const void *unused __unused)
6785 {
6786 
6787 	pf_unload_vnet();
6788 }
6789 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
6790 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6791     vnet_pf_uninit, NULL);
6792 
6793 static int
6794 pf_modevent(module_t mod, int type, void *data)
6795 {
6796 	int error = 0;
6797 
6798 	switch(type) {
6799 	case MOD_LOAD:
6800 		error = pf_load();
6801 		pf_nl_register();
6802 		break;
6803 	case MOD_UNLOAD:
6804 		/* Handled in SYSUNINIT(pf_unload) to ensure it's done after
6805 		 * the vnet_pf_uninit()s */
6806 		break;
6807 	default:
6808 		error = EINVAL;
6809 		break;
6810 	}
6811 
6812 	return (error);
6813 }
6814 
6815 static moduledata_t pf_mod = {
6816 	"pf",
6817 	pf_modevent,
6818 	0
6819 };
6820 
6821 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
6822 MODULE_DEPEND(pf, netlink, 1, 1, 1);
6823 MODULE_VERSION(pf, PF_MODVER);
6824